Journal of Structural Geology - Special Issue - Fault Zones

Photograph of the Month
JSG special issue Fault Zones Photo of the month

The St. Jean Mas dAlary fault system in the Permian Lodve
Basin, S. France may be viewed as a zone of faults at the outcrop/
eld scale, or a fault zone at the map/satellite photo scale.
Complexity in the system is established from the role of reactivat-
ing basement (Hercynian) heterogeneities during Permian exten-
sion and the tilting of the Lower Permian strata as faulting
proceeded, causing some fault segments to die and new ones to
form (Wibberley et al., Journal of the Geological Society, 2007).
Activity of different segments at different times during the evolu-
tion of the system, along with the strongly heterogeneous mechan-
ical stratigraphy, encouraged the development of a variety of fault
zone structures such as lenses, fault-propagation folds and relay
zones (Van der Zee et al., Geol. Soc. Lond. Spec. Pub. 299, 2008).
C.A.J. Wibberley*
TOTAL, CSTJF, Av. Larribau, Pau, France
* Tel.: 33 559 83 57 93; fax: 33 559 83 56 15.
E-mail address: christopher.wibberley@total.com.
27 October 2010
Available online 18 November 2010
Contents lists available at ScienceDirect
Journal of Structural Geology
j ournal homepage: www. el sevi er. com/ l ocat e/ j sg
0191-8141/$ see front matter
doi:10.1016/j.jsg.2010.10.005
Journal of Structural Geology 32 (2010) 1553
Editorial
Fault zones: A complex issue
As recognized by John Wesley Powell as early as 1870 (Fig. 1),
faults are complex 4-D structures: volumes of complexly deformed
rock that evolve in their structural and uid ow properties
through time. Fault zones are composed of rocks in variable degrees
of deformation states, often with a lower strain fracture-dominated
damage zone surrounding a more highly strained heterogeneous
core zone containing one or several slip zones, gouge and brec-
cias, and oblique Riedel shears. A zone of faults at the larger scale
on the other hand, is often considered as a discrete slip plane in
a volume of otherwise intact rock. However, this intact rock is rarely
really intact, and studies over the last twenty or so years have
shown that this depends critically upon the scale of observation:
the closer you look, the smaller the faults that you can see, hence
the more faults you can see. As deformation evolves, many of the
smaller faults in a zone switch off leaving the larger ones to carry
on getting larger (widening the core zone), often cannibalizing
some of the smaller faults in the process.
Recent advances in modelling, sub-surface and eld data acqui-
sition have shed light on the spatial variability of fault structures
and the time scales and length scales of changes in fault properties.
Perhaps the main advances since the 1870s have been that we
appreciate much more the signicance and impact of complex
structure inzones of faults and fault zones. We thus arrive at a cross-
roads today where we understand that the complexity we see when
we look closely at zones of faults at the large scale exists also at
the scale of an individual fault (Fig. 2). Further, we understand
that the division is really rather arbitrary and that the processes
contributing to the growth and internal structural development of
a single fault zone are similar to that of a zone of faults: propagation
and linkage can increase the length of the fault and change the
fault pattern, but can also result in brecciation and the production
of a fault rock and the generation of a recognizable zone of britt-
ley-deformed rock associated with shear. Yet we do not know
how to capture this complexity in a way in which it may be used
for predictive tools in the many elds in which fault zones impact
us, such as seismogenesis, sealing and leaking in hydrocarbon and
CO
2
systems, hydrological owand radioactive waste management.
Although there is increasing interest in fault and fracture-
related uid ow in the Earths crust driven by these applications,
there is a general lack of awareness of the exciting research being
conducted across these traditionally diverse elds. The conference
Fault Zones: Structure, Geomechanics and Fluid Flow was there-
fore convened at the Geological Society of London in September
2008 with the objective of bringing together scientists from these
communities who were interested in fault growth, fault zone prop-
erties and their effects on other processes such as uid ow
and earthquake processes. The conference was attended by
a wide range of scientists and was run in parallel with the meeting
of the Clay Club.
This Journal of Structural Geology special issue presents 23 arti-
cles that stemfromthe conference and cover the state-of-the-art in
methods and applications involved in studying fault zone structure,
and its impact on geomechanical and uid ow properties. The
papers fall naturally into sections that concentrate on structure,
on geomechanics, and on uid ow properties, although there
is of course a great deal of overlap and, as indeed is the point of
this special issue, nearly all the papers cover more than one of these
topics. An introductory review paper (Faulkner et al.) discusses
Fig. 1. An example of early recognition of the 4-D complexity of fault zone structure. Excerpt from an article by William Powell The Canyons of the Colorado in Scientic American,
v. 5, no. 124, 1878. Reprinted with kind permission from Scientic American.
0191-8141/$ see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jsg.2010.10.006
recent work showing howfault zones and fault systems control the
structure, mechanics and uid ow properties of the Earths upper
crust. These three aspects of faults are intimately related and
cannot be considered in isolation.
Some of the papers in this special issue examine the initiation
and distribution of faulting and related fracturing. These studies
give the emerging picture of an intimate relationship, on both the
geological and co-seismic timescale, between fault propagation
and fault zone fracturing and enhanced uid ow. Fault initiation
and distributions in a system, whilst strongly dependent on the
intermediate principal stress (Haimson & Rudnicki) also depend
on the tectonic context of shortening versus extension (Saillet &
Wibberley) and presence of pre-existing oblique faults (Henza
et al.). This obliquity of pre-existing structure to the new remote
applied stress also controls the style of segment linkage and related
fracturing (Moir et al.), as does fault friction (Soliva et al.) and the
elastic modulii of wall rock and fault zone (Aydin & Berryman).
Such segment linkage-related fracturing greatly increases fault
permeability, and in mudstones this preferentially occurs in over-
consolidated contexts, such as during exhumation (Ishii et al.). Other
processes such as co-seismic rupture tip propagation may cause
fracture damage, which along with radiated seismic energy, is larger
when slip weakening distance is smaller due to the larger wall rock
strains involved in slip pulse propagation (Savage & Cooke). Coseis-
mic fracturing may also lead to entire brecciation of rocks by decom-
pressionboiling and precipitation as sudden hydraulic connection to
geothermal reservoirs occurs (Caine et al.), and such rapid fracturing
and later sealing processes can also be induced by human activity as
in the case of tunneling excavations in the Boom Clay (Van Marke &
Bastiaens).
The presence of damage zones of fractures around a highly-
sheared core of fault rock is known to affect the uid ow and
mechanical behavior of the fault zone, but little quantication is
available to help us predict these affects. Dockrill & Shipton show
combined eld and geochemical evidence for up-fault CO
2
-charged
groundwater owin the fracture damage zone in multiple sporadic
events, and emphasize that algorithms often used in industry for
predicting across-fault oware not appropriate to up-fault damage
zone ow prediction. Medeiros et al. show how clusters of defor-
mation band faults around larger faults, although causing local
deviations in groundwater ow, do not seriously compartmentalize
the aquifers insingle-phase ow. The mechanical impact of different
stiffness between host rock, fracture damage zone and a softer core
zone suggests that further fractures are more likely to propagate
Fig. 2. Block diagram showing the evolution, from left to right, of a network of individually-nucleating faults to a single, large, complex fault zone. The fault zone grows both by the
linkage of the individual faults, incorporating the slices of host rock in between as lenses into the fault which then get broken up, and by generating new damage around the fault
which then gets eaten up as the fault continues to grow. A few simple uid trapping and migration pathways are also shown for illustration.
Editorial / Journal of Structural Geology 32 (2010) 15541556 1555
inwards before being arrested (Gudmundsson), leading to sugges-
tion that fault zone widening may be limited in discrete steps.
If damage zone widening may be limited by such mechanisms,
then increased deformation must be accommodated by higher
strains in the central part of the fault zone, the fault core, during
these periods, thus localizing further deformation in this zone. This
localization is corroborated by data on carbonate fault cores by Bas-
tesen & Braathen for normal faults with throws above around
100 m. This does not mean that these high-strain core zones
become less complex as displacement increases, although struc-
tures such as lenses of host rock are perhaps in some way related
to bed thickness throw relationships. Other factors affecting
complexity and the presence of structures such as lenses in the
core zone are high competence contrast between different beds
in a multilayered sequence (Schmatz et al.) and effective normal
stress (Cuisiat & Skurtveit), both of which inhibit the formation of
relatively simple clay smears.
Seen at the scale of a large fault zone, individual slip zones
within it are generally assumed to be homogeneous. However,
even the most localized slip zones (of sub-millimetric to centimet-
ric widths) in granular fault zones may be heterogeneous and
complex, with deformation mechanisms evolving as the proper-
ties change with strain, such as porosity reduction by progressive
grain size reduction during compactional cataclastic ow (Bal-
samo et al.). This is argued by Hadizadeh et al. to be, in itself,
a precursor to further brittle shear localization. Indeed this gets
even more complex when hydrothermal uids and temperatures
are considered, with experimental work in synthetic muscovite
gouges (Van Diggelen et al.) generating continuous networks of
ne-grained hardening cataclastic bands anastomosting around
lenses of lower strain (in some cases plastically-deformed) mate-
rial. These bands gradually eat up the lenses as deformation
evolves, in the same way as lenses of wall rock are incorporated
into fault zones in layered sedimentary sequences and eventually
get broken down with increasing strain.
Whilst many of these papers contribute to industrial applica-
tions concerning sub-surface uid ow in the Earths crust, correct
structural interpretation an essential foundation for reliable
studies of fault sealing and fracture-related owprediction in reser-
voirs is still a primary concern. This is discussed by Freeman et al.
who propose guidelines for fault interpretation from seismic using
wall rock strains based on published data on fault dimensions and
displacements to provide useful guidelines for structural interpre-
tation of sub-surface data. Geomechanical applications are nicely
illustrated by Cuisiat et al. (b) who describe a geomechanical study
of fault stability during depletion of the Statfjord eld in the North
Sea. Their nding that the peak shear strength of the fault (core)
material is the biggest source of uncertainty should continue to
spur work on the mechanical properties of fault materials and
upscaling to fault strength. Fault sealing is also addressed by Woods
& Norris in terms of the migration of an exponentially decaying
source migrating into a leaky seal, their numerical modelling illus-
trating the parameters controlling ow.
Thus beyond the direct interpretations of deformation processes
contributing to fault zone growth described in this special issue,
many of the descriptive observations may also serve to characterize
fault zone complexity (e.g. statistical distributions of faults, fault
dimensions, fault rock distributions such as the size and shape of
high and/or low-permeability zones like host rock lenses, core-to-
damage zone relative thicknesses) and constrain dynamic uncer-
tainty models for fault property and uid ow prediction. The trick
will then be how to put such observational data, in quantitative
form, into models for uncertainty in fault zone property prediction,
and accounting for fault zone complexity.
Christopher A.J. Wibberley*
TOTAL, CSTJF, Av. Larribau, Pau, France
* Tel.: 33 559 83 57 93.
E-mail address: christopher.wibberley@total.com.
Zoe K. Shipton
Dept. Civil Engineering, Univ. Stathclyde, Glasgow, U.K.
27 October 2010
Available online 30 October 2010
Editorial / Journal of Structural Geology 32 (2010) 15541556 1556
Review Article
A review of recent developments concerning the structure, mechanics and uid
ow properties of fault zones
D.R. Faulkner
a,
*
, C.A.L. Jackson
b
, R.J. Lunn
c
, R.W. Schlische
d
, Z.K. Shipton
e
, C.A.J. Wibberley
f
,
M.O. Withjack
d
a
Department of Earth and Ocean Sciences, University of Liverpool, Liverpool, UK
b
Department of Earth Science and Engineering, Imperial College London, UK
c
Department of Civil Engineering, University of Strathclyde, Glasgow, UK
d
Department of Earth and Planetary Sciences, Rutgers University, Piscataway, New Jersey, USA
e
Department of Geographical and Earth Sciences, University of Glasgow, Glasgow, UK
f
Total France EP, CSTJF, Av. Larribau, Pau, France
a r t i c l e i n f o
Article history:
Received 11 December 2009
Received in revised form
1 June 2010
Accepted 16 June 2010
Available online 11 August 2010
Keywords:
Faults
Structure
Fluid ow
Mechanics
Earthquakes
a b s t r a c t
Fault zones and fault systems have a key role in the development of the Earths crust. They control the
mechanics and uid ow properties of the crust, and the architecture of sedimentary deposits in basins.
We review key advances in the study of the structure, mechanics and uid ow properties of fault zones
and fault systems. We emphasize that these three aspects of faults are intimately related and cannot be
considered in isolation. For brevity, the review is concentrates on advances made primarily in the past 10
years, and also to fault zones in the brittle continental crust. Finally the paper outlines some key areas for
future research in this eld.
2010 Published by Elsevier Ltd.
1. Introduction
Fault zones control a wide range of crustal processes. Although
fault zones occupy only a small volume of the crust, they have
a controlling inuence on the crusts mechanical and uid ow
properties. Much recent work has concentrated on describing and
understanding the importance of the structure, mechanics and
uid ow properties of fault zones. This has involved eld obser-
vations, laboratory experiments, seismology, hydrogeology, and
analytical and numerical modelling.
Brittle fault zones are lithologically heterogeneous, anisotropic
and discontinuous. Faults are complex zones composed of linked
fault segments, one or more high strain slip surfaces nested within
regions of high and low strain (often called fault core and damage
zone), Riedel shears, splay faults, dilational and contractional jogs,
and relay ramps (Rawling et al., 2001; Shipton and Cowie, 2001;
Faulkner et al., 2003; Childs et al., 2009). Individual fault zones
commonly show signicant variation in complexity along strike or
down dip, even over relatively short distances (Schulz and Evans,
2000; Shipton and Cowie, 2001; Kirkpatrick et al., 2008; Lunn
et al., 2008). Fault zone structure, mechanics and permeability can
vary strongly both over geological time (e.g. Eichhubl et al., 2009)
and at timescales relevant to a variety of industrial applications.
The strength of the lithosphere varies with depth, temperature
and mineralogy (Kohlstedt et al., 1995) but a major load-bearing
region is likely present at the base of the seismogenic zone at w15
to 20 km depth for normal continental crust. The mechanical
properties of faults at this depth are thus inferred to control to
some extent the strength of the entire crust. This inference is
supported by observations of crustal stress magnitudes at shal-
lower crustal levels that appear to be limited by the typical fric-
tional strength of faults (Townend and Zoback, 2000). A related goal
to characterizing the mechanical properties of faults at depth is to
understand the earthquake process from nucleation and propaga-
tion to arrest.
Faults play an important role in controlling the migration of
crustal uids. One example of this is hydrocarbon migration,
accumulation and leakage in sedimentary basins. At the basin scale,
* Corresponding author.
E-mail address: faulkner@liv.ac.uk (D.R. Faulkner).
0191-8141/$ e see front matter 2010 Published by Elsevier Ltd.
doi:10.1016/j.jsg.2010.06.009
Journal of Structural Geology 32 (2010) 1557e1575
faults and fault-related folds control subsidence patterns and hence
the distribution of thermally mature zones (e.g. Brister et al., 2002).
Faults are also a key component of many hydrocarbon plays; they
also may control discrete subsurface pressure cells (e.g. Grauls et al.,
2002). At a smaller scale, a better understanding of the role of faults
in compartmentalizing elds will yield better estimates of hydro-
carbon production (e.g. Manzocchi et al., 1999). Increasingly, the
recognition of high transient permeability along faults induced by
hydrocarbon production (Losh and Haney, 2006) has focussed
interest on the temporal variation of fault-zone permeability. Ore
deposits are also commonly related to fault zones due to episodic,
localized hydrothermal ows that occur during and immediately
after periods of fault movement (e.g. Cox et al., 2001; Sibson, 2001).
Characterizing the uid ow properties of the crust is necessary
to facilitate the development of deep-waste storage repositories (e.g.
Ferrill et al., 1999; Douglas et al., 2000), to allow sequestration of
industrially-produced greenhouse gases (Streit and Hillis, 2004;
Dockrill and Shipton, 2010) and to realize the potential of geothermal
energy in appropriate locations (Rowland and Sibson, 2004; Fairley,
2009). The physical characteristics and properties of faults will play
an important role in regional crustal uid owthat might affect such
applications.
This paper concentrates on three primary aspects of fault zones
and fault systems; their structure, mechanics and uid ow prop-
erties. We emphasize that these three aspects are inextricably
coupled and this is highlighted in Fig. 1. Much recent research
reects efforts to understand the nature and processes behind this
coupling. For instance, fault rocks are commonly altered (Evans and
Chester, 1995) and are not simply a granulated product of their
protolith. In fact, in the upper crust, many fault rocks may be viewed
as low- to medium-grade metamorphic rocks, with authigenic
growth of clays and other minerals. A close-knit coupling exists
between deformation, mechanics and uid ow in fault zones by
deformation- and reaction-driven changes in porosity and perme-
ability, and uids causing changes in deformation mechanisms
throughuid-rockinteractions infault zones (e.g. Rutter andBrodie,
1995; Wibberley and McCaig, 2000; Eichhubl et al., 2005; Holyoke
and Tullis, 2006; OHara, 2007). The impact on fault rheology, such
as byuid-enhancedreactionsoftening, maybe considerable(Imber
et al., 2001; Gueydan et al., 2003; Jefferies et al., 2006a).
Given the importance of fault zones and fault systems, a vast
body of work covers their development and properties. In this
paper we review important classical concepts but concentrate on
key advances in our understanding of fault zones over the past w10
years. Further, we limit our discussion to upper crustal faults
developed in the continental crust (i.e. those in the top 20 kmof the
crust). The review is perhaps biased towards the interests of the
authors and concentrates on what we see as important issues
reading fault zones. A number of excellent review articles have
been published that refer to earlier work and the reader will be
directed to these for not only a more historical view on the devel-
opment of studies in this eld, but more detail regarding the
different disciplines involved.
This reviewconcentrates rst on the structure of fault zones and
fault systems, including scaling relationships that can help in
deciphering how fault zones develop and grow. Second, we discuss
the mechanics of fault zones; this involves topics ranging from
laboratory experimentation to modelling of earthquake rupture.
Last, we address the topic of uid-ow properties of faults,
involving studies from in situ observations of uid ow, laboratory
experiments and modelling. We conclude by highlighting some key
outstanding questions.
2. Structure and development of fault zones and fault
systems
2.1. Typical fault zone structure
A simple conceptual model for fault zone structure, developed
over the past 20 years, involves strain that is localized in a fault core
surrounded by a distributed zone of fractures and faulting in the
damage zone (Fig. 2, see references in Wibberley et al., 2008). The
fault core generally consists of gouge, cataclasite or ultracataclasite
(or a combination of these), and the damage zone generally consists
of fractures over a wide range of length scales and subsidiary faults.
Strain may be homogeneously distributed across the fault core
(Rutter et al., 1986; Faulkner et al., 2003) or may be highly localized
onto discrete slip surfaces (Chester and Logan, 1986; De Paola et al.,
2008). Brittle fault rocks have previously been considered chaotic,
but detailed observations show that they are highly structured,
commonly containing P foliations, Riedel shears and Y shears
(Logan et al., 1979; Chester et al., 1985; Rutter et al., 1986; Jefferies
et al., 2006b). Additionally, fault breccias, such as those described
by Caine et al. (2010) can occur as part of the fault core in areas
where movement involving fault jogs results in dilatation.
Recent research has questioned the general applicability of this
simple model. Fault zones may contain a single fault core (some-
times with branching subsidiary faults), or the fault core may
branch, anastomose and link, entraining blocks or lenses of frac-
tured protolith between the layers (Fig. 2b; Faulkner et al., 2003). A
comparison between the Punchbowl fault (Chester et al., 1993) and
the Carboneras fault (Faulkner et al., 2003), two 40 km-displace-
ment strike-slip faults that operated at similar depths, illustrates
the effect of the contrast in mechanical properties between fault
zone and protolith. The widths of the core zones of these faults
(50 cm for the Punchbowl fault in granite and low-porosity sand-
stones versus multiple strands over a 1 km wide zone for the Car-
boneras fault in predominantly mica schists) may reect
differences in the mechanical strength contrast between protolith
and the deformation zone. The second style of fault zone structure
has been observed in the SAFOD borehole, where multiple fault
strands, individually up to several metres thick, exist and at least
two of these strands of fault gouge are moving simultaneously
(Zoback et al., 2010). The schematic diagrams in Fig. 2 purposely
have no scale to illustrate that these fault zone structures occur at
a variety of scales. The internal structure of fault zones might
provide a useful indication of their mechanical properties (Faulkner
and Rutter, 2003; Biegel and Sammis, 2004; Faulkner et al., 2008).
Fault Zone
Composition
Structure Fluid Flow
Mechanics
Permeability
Mode of Failure Mode of Failure
Segmentation
Roughness
Dilatancy
Compaction
Reaction
Alteration
Fig. 1. Flow diagram showing inter-relationships among the three main topics of
structure, mechanics and uid ow. Mode of failure refers to whether or not seismic
slip occurs.
D.R. Faulkner et al. / Journal of Structural Geology 32 (2010) 1557e1575 1558
Localization may be further enhanced where a fault juxtaposes two
protoliths of highly contrasting competence, such as the Median
Tectonic Line, Japan (e.g. Wibberley and Shimamoto, 2003).
Fault zone structure depends on the depth of formation (e.g.
Ishii et al., in press), protolith, tectonic environment (e.g. strike-
slip, extension or compressional), magnitude of displacement and
uid ow. For instance, faults in low porosity rocks (Balsamo et al.,
2010) generally have a ne-grained fault core surrounded by
a fracture-dominated damage zone. In contrast, coarser grained,
high porosity rocks commonly develop by the formation and
amalgamation of low porosity deformation bands followed by the
nucleation and propagation of high permeability slip surfaces
(Fossen et al., 2007).
Several authors provide qualitative descriptions of fracture
damage zones surrounding a fault core (e.g. McGrath and Davison,
1995; Shipton and Cowie, 2003; Kim et al., 2004; Berg and Skar,
2005; Cembrano et al., 2005; Johansen et al., 2005; Cook et al.,
2006; de Joussineau and Aydin, 2007; Fossen et al., 2007).
Damage zones contain fractures at a range of different scales from
grain-scale microfractures to macrofractures that may accommo-
date small shear offsets and a small quantity of cataclasite. It can be
difcult to distinguish between subsidiary fault structures (which
may be viewed as faults in their own right) and fault damage zone
fractures. In the tip zones of large displacement faults in particular
the complexity of deformation is marked (Kirkpatrick et al., 2008).
The orientation of the macroscopic deformation surrounding fault
tips may include horsetail geometries, wing cracks and synthetic or
antithetic subsidiary faults (e.g. de Joussineau and Aydin, 2007;
Moir et al., 2010). In lowporosity rocks or under loweffective stress
conditions, damage zones consist of dilatant fractures (Blenkinsop,
2008), whereas higher porosity rocks often develop structures in
their damage zones such as compaction bands in sandstone
(analogous to anticracks) or cataclastic deformation bands (e.g.
Johansen et al., 2005; Fossen et al., 2007).
Quantitative studies of damage zones, commonly involve
determining the density of fractures (usually fromline counting) as
a function of distance from the fault core. For low porosity rocks,
macrofractures (mesoscale features that may be readily identied
in the eld) and microfractures (measured from orientated thin-
sections) commonly show an exponential decrease with distance
from the fault core (Vermilye and Scholz, 1998; Wilson et al., 2003;
Mitchell and Faulkner, 2009) (Fig. 3). This relationship has been
linked to the decay of stress away from a fault tip predicted from
fracture mechanics models. Microfractures in deformation band-
dominated damage zones in high porosity sandstones show no
observable change of microfracture density surrounding faults
(Anders and Wiltschko, 1994; Shipton and Cowie, 2001) due to the
different micromechanics of deformation-band faulting and
a b
core multiple cores damage zone damage zone
Fig. 2. Typical fault zone structures. (a) Shows a single high-strain core surrounded by a fractured damage zone (after Chester and Logan, 1986) and (b) shows multiple cores model,
where many strands of high-strain material enclose lenses of fractured protolith (after Faulkner et al., 2003).
Fig. 3. An example of (a) microfracturing and (b) macrofracturing surrounding three
strike-slip fault zones in northern Chile. The fault zones are in low porosity crystalline
rocks and the Caleta Coloso fault has w5 km of displacement, the Cristales fault has
220 m of displacement and the Blanca fault 35 m of displacement. From Mitchell and
Faulkner (2009).
fracture-dominated faulting (see Fig. 4). Nevertheless, where
damage zones are dominated by deformation bands, these may
show an exponentially decreasing density from faults in high-
porosity sandstone (Berg and Skar, 2005; de Joussineau and Aydin,
2007), although in at least some cases, fault localization occurs
once the damage zone has already developed (Saillet and
Wibberley, 2010).
A maximum microfracture density is often attained immedi-
ately adjacent to the fault core that is dependent on rock type but
independent of the fault displacement (Fig. 3). Although only three
faults are shown in Fig. 3, three additional faults with smaller
displacements (of 13 cm to 2 m) showed a similar maximum frac-
ture density (of w20 per mm) (Mitchell and Faulkner, 2009). This is
suggestive of a critical amount of fracture damage that the rock can
accumulate at the boundary between the fault core and the damage
zone. The same type of fracture damage distribution occurs at the
small scale (i.e. mm) in experiments (Janssen et al., 2001). A
recently recognized type of fracture damage surrounding the
surface trace of large active fault zones involves the in situ frag-
mentation of the protolith around the fault core, such as inferred
fromobservations at several localities of the same fault (Reches and
Dewers, 2005; Dor et al., 2006). The original structure and textures
of the protolith are preserved, but pervasive microfracturing of the
protolith has lead to complete incohesion. They have been termed
pulverized rocks.
2.2. Fault zone scaling and development
As summarized in the previous sub-section, fault zone structure
depends on a variety of factors. Hence caution must be used in the
scaling relations among fault zone characteristics such as fault
thickness or damage zone width. This problem can be particularly
acute when datasets from different areas are combined. The way in
which different authors have dened damage zone width (mean
width at one site, single measurement along a scanline or maximum
damage zone width envelope) makes comparing studies very dif-
cult (Shipton et al., 2006). Another issue is that the data may be
biased by differences in the methods used to measure the fault
width (e.g. should the width of the San Andreas fault zone as
measured from a geological map or satellite photo include all the
low-strain rock in between the various strands of the fault system?)
(Schulz and Evans, 2000). Where faults with a range of displace-
ments or lengths formed in the same protolith, at approximately the
same depth and environmental conditions and under a similar stress
eld, then useful scaling relationships may be determined (e.g.
Childs et al., 2009; Mitchell and Faulkner, 2009).
Faults initiate from the coalescence of a number of growing
mode I cracks (Reches and Lockner, 1994; Healy et al., 2006).
Commonly, the initiation of faults occurs on pre-existing planes of
weaknesses such as cooling joints (Martel et al., 1988; Di Toro and
Pennacchioni, 2005), dykes (dAlessio and Martel, 2005) or tectonic
joints (Wilkins et al., 2001; de Joussineau and Aydin, 2007; Moir
et al., 2010).
As fault cores develop, wear models have suggested that
progressive damage leads to the development of a wider fault core.
Broad trends may be seen in fault core thickness e displacement
relationships (Scholz, 1987), but the wide scatter in these data may
reect the fact that these compilations are fromfaults developed in
wide range of tectonic regimes and which formed in variable
environmental conditions. Although wear models are physically
appealing from a mechanical perspective and make intuitive sense,
the complexities arising in natural fault zones produce a wide range
of resultant fault core internal structures (Evans, 1990; Shipton
et al., 2006; Faulkner et al., 2008). In summary, quantitative
models for fault core development may not be universally appli-
cable and may, more accurately, provide an upper bound on fault
core width.
The growth of fault damage zones and their scaling with fault
displacement is an important topic, as the fracture damage zones
surrounding faults may provide uid ow pathways of economic
signicance and, in addition, act as an energy sink during quasi-
static fault growth or dynamic rupture. Vermilye and Scholz (1998)
made measurements of small displacement faults and concluded
that fault damage zone width (dened by the points where fracture
damage falls below background levels on either side of the fault)
scales with fault displacement. Mitchell and Faulkner (2009)
studied damage zone width dened by macro- and micro-frac-
ture densities for faults developed within the same granodioritic
batholith over nearly ve orders of magnitude of displacement
(Fig. 5). The damage zone width scales with fault displacement,
however at higher displacements the rate of growth of the damage
zone width with displacement decreases after a few hundred
Fig. 4. Comparison of microfracture density with distance from the main fault from high porosity rocks with deformation bands (Anders and Wiltschko, 1994; Shipton and Cowie,
2001) with microfracture density from low porosity rocks (Anders and Wiltschko, 1994; Vermilye and Scholz, 1998). The average host rock porosity for each site is given in brackets.
Linear regressions show that in low porosity rocks a logarithmic decay of microfracture density occurs with distance from the fault zone. Conversely, in high porosity rocks,
microfracture densities drop to background levels at all points outside the fault zone.
metres of displacement (Fig. 5). Micarelli et al. (2006) found
a similar decrease in the growth of fracture damage zone width at
displacements above 1e5 m in high-porosity carbonate sequences
in Sicily. However, in this case, displacements of 1e5 m coincide
with the onset of cataclastic fault rock generation, which may be
controlled by the scale of the bedding in these rocks which is of the
same order of magnitude. Savage and Brodsky (submitted for
publication) compiled fault width versus displacement data from
a range of sources (with inherent limits of comparison) and also
suggest that the growth of the macrofracture damage zone width
decreases markedly for larger (>100 m) displacement faults.
Many authors have attempted to relate the fracture damage
around faults and the width of the damage zone to the mechanics
responsible for their formation. Such studies are particularly
attractive for, if the underlying physics are understood, predictions
of damage zone properties can be made for any faults where the
mechanical properties of the protolith are known. Fig. 6 summa-
rizes the primary processes that have been suggested for the
production of fracture damage surrounding faults (Wilson et al.,
2003; Blenkinsop, 2008; Mitchell and Faulkner, 2009). Many
damage zone models have concentrated on the link to process zone
mechanics (Anders and Wiltschko, 1994; Vermilye and Scholz,
1998) or the repeated slip on small patches along a single fault
(Shipton and Cowie, 2003). The intensity of tensile damage can be
related to slip distribution surrounding the fault tip (Savage and
Cooke, 2010) or stress-release at extensional relays (Soliva et al.,
2010). The various types of models have, in general, predictable
fracture populations and orientations. However, many of the pre-
dicted fracture orientations are very similar and distinguishing
between them is difcult. Given that many of these processes
operate at various times on any given fault, the pattern of fracturing
may be very difcult to interpret (Fig. 6f). The conclusions from
most recent studies have supported the idea that the fracture
damage surrounding faults accumulates from a combination of
processes (Shipton and Cowie, 2003; Wilson et al., 2003;
Blenkinsop, 2008; Mitchell and Faulkner, 2009).
2.3. Fault-system development
This review has largely concentrated thus far on the develop-
ment of individual faults. This review has largely concentrated thus
far on the development of individual faults. To understand the
growth of faults and interpret relationships between displacement
and length, or displacement and damage-zone development, we
must also consider the interaction of faults or fault segments.
For a fault zone with signicant displacement, it is difcult or
impossible to condently identify the incremental steps which
occurred during its development. Where syn-tectonic (growth)
strata are preserved, however, the stratigraphic architecture of
these units may provide constraints on the development of fault
systems (e.g. Gawthorpe et al., 1997). For example, high-resolution
3D seismic surveys (e.g. Baudon and Cartwright, 2008; Lohr et al.,
2008) image the geometry of growth strata. Additionally, numer-
ical and analogue modelling allows systematic study of fault-zone
and fault-system development with known boundary conditions
(see papers by Henza et al., 2010; Moir et al., 2010; Schmatz et al.,
2010).
As proposed by many authors (Walsh and Watterson, 1988;
Peacock and Sanderson, 1991; Walsh et al., 2002; Childs et al.,
2009), larger faults are the result of the growth and linkage of
smaller faults. This growth and linkage process commonly produces
a variety of fault-related folds, which are the continuous (ductile)
deformation that accompanies and is genetically related to faulting
(e.g. Anastasio et al., 1997; Cosgrove and Ameen, 2000; Wilkerson
et al., 2002). Fault-related folds include folds that accommodate
displacement variations on faults (e.g. relay ramps), folds resulting
from fault-tip propagation, and folds due to movement on non-
planar faults (which may form due to the linkage of non-coplanar
fault surfaces). Fault-related folds are associated with reverse and
thrust faults (Suppe, 1983; McClay, 2004; Shawet al., 2005), normal
faults (Schlische, 1995; Janecke et al., 1998; Withjack et al., 2002),
strike-slipfaults (Christie-BlickandBiddle, 1985; Harding, 1990) and
oblique-slip faults (Tindall and Davis, 1999; Cristallini and
Allmendinger, 2001; Schlische et al., 2002). The rock properties in
these folds are commonly modied; thus, some fault-related folds
may be considered as part of the faults damage zone.
The formation of larger faults through the linkage of smaller
faults inuences scaling relationships (Dawers and Anders, 1995;
Kim and Sanderson, 2005). For example, when two smaller faults
link, the resultant length might be longer than if the fault grew as
a single structure. Walsh et al. (2002) note that once some of the
faults in a system become larger than others by segment linkage,
they tend to dominate further deformation by increasing
displacement rates; this typically results in the cessation of the
activity on the smaller faults.
Early attempts to infer the scaling relationships of faults relied
largely on logelog plots of maximum displacement versus length.
These datasets indicate that similar displacement e length ratios
occur for a wide range of scales, suggesting some scale-invariant
behaviour to fault propagation and growth (Walsh and Watterson,
1988, also see Bonnet et al., 2001 for a review). The recognition of
the importance of segment linkage in fault growth at least partially
explained the wide scatter in the data, even on logelog plots (e.g.
Cartwright et al., 1995; Dawers andAnders, 1995; KimandSanderson,
2005). Interestingly, analyses of the thickness of fault cores showed
similar scale-invariant behaviour on logelog plots of displacement
versus thickness (Section 2.2). In both cases, the validity of this
approach for inferring the mechanics of fault growth or fault core-
zone development (e.g. by wear or abrasion) is difcult because of the
hazards of comparingdatasets fromdifferent lithological andtectonic
settings. Nevertheless, in the case of fault-length scaling, general-
izations of data sets can provide useful upper limits to feasible
displacement gradients which are important as rules-of-thumb for
quality control of structural interpretations from seismic data,
particularly in the case of spare 2D lines (Freeman et al., 2010).
In sedimentary basins, fault linkage typically involves the
formation and eventual breaching of relay zones (see Walsh et al.,
Displacement (m)
10
-2
10
-1
10
0
10
1
10
2
10
3
10
4
D
a
m
a
g
e
z
o
n
e
w
i
d
t
h
(
m
)
10
-2
10
-1
10
0
10
1
10
2
10
3
Range of data from Savage
and Brodsky (2009)
Fig. 5. The scaling of damage zone width (for both macrofractures and microfractures)
as a function of displacement. The lled circles are microfracture damage zone data
(collected from a single low porosity, crystalline lithology) from Mitchell and Faulkner
(2009); the shaded area shows the extent of data compiled by Savage and Brodsky
(submitted for publication) from a range of different faults, using macrofracture
damage zones. The data from Mitchell and Faulkner are tted with a hyperbolic
function shown by the solid line.
2002 and references therein). Recent studies of overlap and relay
zones have shown how, at a variety of scales, they may evolve into
fault-bounded lenses as the deformation evolves and the two
segments fully connect to form a single fault (e.g. Kristensen et al.,
2008; Childs et al., 2009). Similarly, asperity bifurcation (splay-
ing) such as at zones of enhanced wall rock fracturing, and along-
strike reconnection of these splays results in isolation of lenses of
low-strainwall rock between the splays in a growing fault zone (e.g.
Van der Zee et al., 2008). Both of these processes are controlled by
heterogeneities of characteristic length scales ranging from grain
size through bed (e.g. Wilkins and Gross, 2002; Soliva and
Benedicto, 2005) and joint scale (Van der Zee et al., 2008) to
structures such as sidewall rip-outs at the scale of strike-slip
faults in the upper crust (e.g. Swanson, 2005). The scale of these
might be controlled by variations in fault orientation, possibly due
to fault linkage. Thus, in reality, fault thickness growth may occur
by a number of discrete scale-dependent steps (Wibberley et al.,
2008). Recent studies using geomorphology (Whittaker et al.,
2008; Roberts et al., 2009), shallow seismic surveys (Bull et al.,
2006) and numerical modelling (Cowie et al., 2006) have investi-
gated the accumulation of slip at breached relays and the rela-
tionship between throw rate and the longevity of sip minima at
linkage sites. These results have important implications for basin
development and earthquake rupture propagation through fault
linkage zones (Roberts et al., 2009; Walker et al., 2009).
3. Mechanics of fault zones and fault systems
Quantication of the mechanics of faulting and earthquakes
comes fromin situ crustal measurements, laboratory friction testing,
eld measurements and inversion of seismic data (Scholz, 2002).
Relative mechanical properties (without quantifying the absolute
stresses) may be derived from eld observations and seismic data.
Classic work conducted by Byerlee (1978) provides a range of friction
coefcients under which fault slip should occur at crustal depths
where brittle deformation dominates. A compilation of crustal stress
measurements by Townend and Zoback (2000) has highlighted the
general applicability of Byerlees law. Similarly, an analysis of the dip
of seismogenic normal faults indicate that they generally fall at the
lower end of the range of friction values suggested by Byerlee
(Collettini and Sibson, 2001), and a similar observation holds for
seismogenic reverse faults (Sibson and Xie, 1998). In this section we
highlight instances where crustal fault strength appears to vary
considerably from the picture outlined above. The mechanics of the
earthquake process is then addressed.
3.1. Weak faulting
Compelling observations indicate that some faults slip under
anomalously low friction coefcients, much less than those pre-
dicted by Byerlee. The notion of weak faults derives mainly from
the orientation of the fault plane with respect to the maximum
principal stress. Weak faults are classied as those that appear to
slip even when frictional lock-up is predicted by Byerlees law. One
issue relates to explaining continued slip on a fault rotated out of its
ideal orientation. However, another problem is to explain how
Fig. 6. Processes that are responsible for creating off-fault damage resulting in damage
zones (after Mitchell and Faulkner, 2009). The damage in the gure is shown as mode I
fractures surrounding a fault. Left-hand side boxes show the initial state and right-
hand side boxes show a more evolved state. (a) Shows damage from the coalescence of
microfractures; (b) shows damage from linking of structures (c) shows damage from
fault growth involving a process zone (d) shows damage from continued displace-
ment on wavy faults; (e) shows co-seismic fracture damage, where V
r
is the rupture
velocity and (f) illustrates how a combination of all these can produce a complicated
pattern of fracture damage surrounding a fault core (based on work from Wilson et al.,
2003; Blenkinsop, 2008; Mitchell and Faulkner, 2009). Note that damage generating
processes highlighted in this gure can be active at different stages during the
evolution of a fault zone.
some faults appear to have formed at such a non-optimal orien-
tation (e.g. Collettini and Holdsworth, 2004).
The debate surrounding the problem of weakness on large
thrusts is probably the oldest of those regarding weak faults. The
paradox is that the maximum fault area that a relatively thin thrust
sheet can load without breaking up internally is much less than
observed in natural examples. Hubbert and Rubey (1959) suggested
a solution where basal friction can be overcome by high uid pres-
sure in the thrust fault zone. Price (1988) on the other hand, showed
that the paradox is based on the assumption that slip occurs on all of
the thrust surface simultaneously, whereas evidence from earth-
quakes suggests that thrusts, like other active faults, operate by the
overall accumulation of displacement on slip patches on different
parts of the fault at different times. Furthermore, thrust zones in
crystalline basement show extensive evidence for uid-enhanced
reaction weakening, encouraging a change in deformation mecha-
nism regime from frictional sliding to shear by diffusive mass
transfer and/or crystalline plasticity (Wibberley 2005).On the topic
of reverse faults, Sibson (2009) recently showed that normal faults
on the eastern side of the Sea of Japan reactivated in a reverse sense,
as the back-arc basin has started to contract, are unfavourably
oriented with respect to the regional stress eld, suggesting they are
weak.
Another class of faults that may be viewed as weak are low-
angle normal faults (Axen, 2007). The existence of these structures
is still debated, but compelling recent geological and geophysical
data support their occurrence. One example can be found in central
Italy, where extension and uplift is migrating fromwest to east, and
previously active extensional structures have been exhumed and
are exposed on the Island of Elba. Geological evidence suggests that
the Zuccale fault on Elba, currently in a low angle orientation
(w10
), was active in this same position during movement. The

evidence for this includes the formation of contemporaneous
footwall conjugate extensional faults (Smith et al., 2007) and the
orientation of original compressional structures and vertical
opening mode extensional veins (Collettini and Holdsworth, 2004).
Farther to the west, from the surface to depths around 8 km,
Chiaraluce et al. (2007) imaged a microseismically active zone, the
Alto Tiberina fault, oriented at a lowangle. They interpret this to be
a currently active low-angle extensional detachment, of which the
Zuccale fault to the east is an ancient, uplifted example.
For strike-slip faults, the San Andreas fault in California is
inferred to be weak based on heat-ow data, seismological
constraints and stress orientation. Early work showed that there
was a conspicuous absence of any elevated heat ow fromctional
heating that essentially limits the average shear stress on the fault
to typical stress drops in earthquakes, 10e20 MPa (Brune et al.,
1969; Lachenbruch and Sass, 1980). These data, combined with
the observation that the angle between the maximum principal
stress and the San Andreas fault is high (most latterly by Hickman
and Zoback, 2004; Boness and Zoback, 2006), suggest that the fault
moves under friction coefcients of w0.2 or less. The interpretation
and the quality of both the heat owand the stress orientation data
are controversial (see Scholz, 2006 and references therein). Results
fromthe San Andreas Fault Observatory at Depth (SAFOD) provided
more data close to the San Andreas fault at 3 km depth at the
southern limit of the creeping section near Parkeld. These data
suggest the fault is indeed weak, at least in the creeping section
(Hickman and Zoback, 2004; Scholz, 2006).
Regardless of whether or not weak faults exist, the debate has
driven a large body of work regarding the strength of faults in
general. Laboratory studies have concentrated on the search for
a weak mineral phase that might realistically be present over
seismogenic depths in sufcient quantities to promote weakening.
The frictional strength of most single phyllosilicate phases is less
than predicted by Byerlees law (Rutter et al., 1986; Logan and
Rauenzahn, 1987; Morrow et al., 1992; Scruggs and Tullis, 1998;
Bos and Spiers, 2001; Saffer et al., 2001; Saffer and Marone,
2003; Moore and Lockner, 2004; Tembe et al., 2006; Takahashi
et al., 2007; Crawford et al., 2008; Ikari et al., 2009), although this
is not always the case (van Diggelen et al., 2010). What is not so
clear at present is how the mixing of weak clay material with other
phases, such as quartz affects gouge strength, and how the uid
phase may affect whether sliding is dominated by friction or
another, viscous, rheology. There are few studies that have
addressed the rst question, but those that have suggest a gradual
decrease in frictional strength with the addition of clay (Takahashi
et al., 2007; Crawford et al., 2008). It also appears that natural fault
rock microstructures, with inter-connected networks of weak
phyllosilicate phases, are important to produce some form of
weakening (Holdsworth, 2004). Collettini et al. (2009a) compared
the frictional strength of intact wafers of natural fault gouge from
the Zuccale fault (Elba, Italy) to mineralogically identical powders
and found the material retaining the in situ microstructure is
signicantly weaker than its powdered counterpart. This suggests
that previous experiments where powdered natural gouge was
used provide an upper bound on the strength at best. However, the
experiments on single mineralogical components should still
provide a useful guide to the lower strength bound for natural
gouge containing these phases.
Phyllosilicate phases the required strength to explain weak
faulting, such as montmorillonite or chrysotile, tend to strengthen at
laboratory pressures and temperatures greater than that equivalent
to a few kilometres depth (e.g. Morrow et al., 1992; Moore et al.,
2004). One exception is talc, which shows remarkable weakness
over the entire temperature and pressure conditions of the seis-
mogenic crust (see Fig. 7a; Escartin et al., 2008; Moore and Lockner,
2008). For this reason the discovery of serpentinites and associated
talc within cuttings fromthe SAFODborehole was signicant (Moore
and Rymer, 2007). However, it is not clear if the talc is concentrated
in sufcient quantities or along narrow slip surfaces in the two
currently active strands of the San Andreas fault at the SAFOD site.
Testing of powdered cuttings from the SAFOD borehole (with
inherent limitations, as noted above) showed that friction coef-
cients are lower than those predicted by Byerlee, but not sufcient to
account for the observations of fault weakness (Tembe et al., 2006).
Similarly, the presence of talc in the Zuccale low-angle normal
fault has been noted. It is explained by reactions involving dolomite
and silica-rich uids (Collettini et al., 2009b). However, frictional
studies of powdered natural gouge from the Zuccale fault suggest
that the friction coefcients are too high to explain movement,
although these might represent an upper bound of strength as
powdered samples were used (Smith and Faulkner, 2010). Similar
results apply to other lowangle normal fault systems (Numelin et al.,
2007). More importantly, the discontinuous nature of the talc-rich
fault gouge layers in the Zuccale fault show that, even with the
presence of talc, the geometry cannot explain slip (Smith et al., 2007;
Smith and Faulkner, 2010). In summary, although intrinsically weak
minerals do exist within fault zones, it is currently unclear whether
these can be the sole source of weakening.
Laboratory testing generally occurs at room temperature
conditions and at strain rates that far exceed their natural coun-
terparts for fault creep. One possible way to weaken faults is that
additional mechanisms other than frictional ones are responsible
for fault slip. These would only operate at much slower strain rates
than those achievable in the laboratory (Gratier et al., 1999). The
overall dominance of pressure solution creep over frictional
processes in fault zones remains a possibility (Rutter and
Mainprice, 1979). Geological observations of many faults suggest
that the mineralogical changes during deformation in large faults
might promote this type of behaviour (Wintsch et al., 1995; Imber
et al., 2001; Holdsworth, 2004). If pressure solution is operative in
very mature fault zones, the soluble phases may be removed
entirely, leading renewed frictional behaviour. Experiments on rock
analogues (halite and clay) showed that both frictional processes
and pressure solution are can operate simultaneously (e.g. Bos
et al., 2000). One problem is that characteristic microstructures
indicating that pressure solution creep was operative in natural
fault rocks are difcult and sometimes impossible to nd. Pressure
solution experiments on natural rocks, or those that are likely to be
present in natural fault zones, are few, and the diffusion rates for
quartz, for example, are slow (Hickman and Evans, 1995; Gratier
et al., 2009). Consequently it is difcult to assess the possible
contribution of pressure solution creep on fault behaviour.
However, if used to explain possible weakness, it will be more
applicable to faults with low displacement rates (e.g. w1 mm/year)
such as those predicted for low-angle detachment faults, or in the
interseismic period between earthquakes (e.g. Renard et al., 2000).
Another possibility for intrinsic fault weakening arises from the
concentration of phyllosilicate phases in mature fault cores.
Wintsch et al. (1995) suggested that a layer of well-orientated
phyllosilicate basal planes might provide a weak horizon within
faults. Experiments on phyllosilicates shows they have signicant
mechanical anisotropy (e.g. Mares and Kronenberg, 1993; Mariani
et al., 2006 and references therein). Mariani et al. (2006) showed
that with muscovite polycrystals at elevated temperature and very
low laboratory strain rates deformation approached linear viscous
and showed no grain size dependence, suggesting some form of
Harper-Dorn creep. Clearly additional experiments at lower strain
rates are necessary to investigate these processes further.
High pore uid pressures are another mechanism that may
weaken faults (Fig. 7b; Hickman et al., 1995). One model involves
pore pressure that varies over the seismic cycle. Immediately
following an earthquake the permeability of the fault is high
(Sibson, 1990). Over time, permeability falls and compaction of the
fault leads to overpressure and promotes slip (Blanpied et al., 1992;
Byerlee, 1993). A second model involves a fault zone that contin-
uously acts as an impermeable barrier and high pore uid pressure
is maintained by a uid ux from mid-crustal levels (Byerlee, 1990;
Rice, 1992). The rst model is more applicable to seismogenic faults,
where as the second is better suited to faults undergoing creep.
Geological evidence supports both models. For the second model,
lowpermeability fault gouges and a favourable fault zone structure
suggests that long term sealing is possible if uid sources at depth
are available (Faulkner and Rutter, 2001). One drawback with these
models is that the level of pore uid pressure required to promote
slip on very unfavourably oriented faults generally exceeds the
minimum principal stress (Rice, 1992). Thus, the pore pressure
would always be limited by hydrofracturing and uid pressure loss
before slip occurs on the fault. This problem is circumvented if the
stress state within the fault zone is different to the remotely applied
stress eld. This condition is possible, with mechanical continuity,
if the mechanical properties of the fault zone are different from the
country rock (Rice, 1992; Chery et al., 2004; Faulkner et al., 2006).
One nal fault weakening mechanism operates during seismic
slip (dynamic weakening; Fig. 7c). If operative, dynamic weakening
mechanisms can explain the lack of any heat ow anomaly but, in
the absence of any other weakening mechanism, cannot explain the
nucleation of dynamic events on a severely misoriented fault. The
issue of dynamic rupture will be discussed in Section 3.3.
3.2. Earthquake nucleation
The issue of how earthquakes nucleate is important because it
might produce measurable precursory phenomena that may be
used in short-term earthquake prediction (Ellsworth and Beroza,
Fig. 7. Weak fault models. (a) Talc strength over the brittle crustal conditions. (b) An example of transient uid pressure weakening based on a crack seal cycle (e.g. Byerlee, 1993);
and (c) dynamic weakening, where the dynamic stress drop (not measurable in earthquake events) is much larger than the measured static stress drop, requiring that much less
energy is dissipated in the earthquake than might be expected.
1995; Beroza and Ellsworth, 1996). In recent years, modelling of the
build up to instability has generally utilized the rate- and state-
dependent friction laws that predict a period of stable, aseismic slip
preceding instability. The rate and state formulation also allows
a wide variety of observed fault behaviour to be modelled including
aseismic fault creep, slow earthquakes and dynamic rupture
(Scholz, 1998).
The principles of rate and state friction are well-known and are
only briey described here (for reviews on rate and state friction
behaviour see Dieterich and Kilgore, 1996; Marone, 1998). Although
extrapolated to accelerating slip and instability in the case of
earthquakes, the formulation is based on laboratory experiments
where the response of the dynamic friction coefcient to a step
increase sliding velocity is measured (see Fig. 8). There is an
instantaneous response of the friction coefcient (or equally the
shear traction) to the change in velocity (the rate effect), followed
by time-dependent evolution over a particular slip displacement
(the state effect) (Fig. 8). The rate and state friction law (Eq. (1)) is
purely phenomenological and the physical processes responsible
for the observed behaviour, particularly for the state evolution, are
poorly known.
m m
0
aln
V
V
0
b ln
V
0
q
D
c
(1)
where m and m
0
are the friction coefcient and the initial friction
coefcient respectively, a and b are experimentally derived
constants, V and V
0
are the new sliding velocity and the initial
sliding velocity respectively, q is the state variable and D
c
is the slip
weakening distance.
The evolution of the state variable is a function of time, normal
stress and displacement and has units of time. It has been explained
in terms of the age of the load-supporting contacts and the time
required for a new set of contacts to develop following a pertur-
bation of the system (e.g. displacement rate, change in normal
stress, etc.). Dieterich and Kilgore (1994) showed direct evidence
that the state variable is related to the evolution of the area of the
load supporting contacts over time and various normal stresses,
termed asperity creep. Two formulations are widely used to model
the state evolution, the aging (or slowness) law and the slip law
(Eqs. (2) and (3)).
_
q 1
Vq
D
c
Aging; Dieterich or Slowness law (2)
_
q
Vq
D
c
ln
Vq
D
c
Slip or Ruina law (3)
These two formulations produce quite different styles of
nucleation and rupture (e.g. Ampuero and Rubin, 2008).
The range of geological materials that have been characterized
in a rate-and-state framework are few. Most comprehensive studies
involve quartz and crushed granite powders. These granular
materials generally show velocity weakening behaviour (where
aeb is negative, see Fig. 8) at low slip rates, opening the possibility
for unstable slip (Green and Marone, 2002). Most natural fault
zones contain at least a proportion of phyllosilicate minerals but
experimental studies on these materials are even sparser (Morrow
et al., 1992; Scruggs and Tullis, 1998; Reinen, 2000; Saffer et al.,
2001; Moore and Lockner, 2008; Ikari et al., 2009; Smith and
Faulkner, 2010). These studies generally suggest that phyllosili-
cate-rich gouges exhibit velocity strengthening behaviour at low
slip rates (where aeb is positive) and thus may be associated with
fault creep (Faulkner et al., 2003). Talc, in particular, exhibits
inherently stable, velocity-strengthening behaviour under all
conditions tested (Moore and Lockner, 2008), although these
conditions do not include seismic slip speeds (w0.1 to 1 m/s). Note,
however, that montmorillonite and serpentinite gouge can both
exhibit velocity weakening behaviour (Reinen, 2000; Saffer et al.,
2001). Indeed, a recent compilation of experimental work
suggests that even materials which exhibit velocity-strengthening
behaviour at lower slip rates become velocity-weakening above
w0.1 m/s (Wibberley et al. 2008). Some experimental studies have
shown that phyllosilicates can exhibit negative b values (Saffer and
Marone, 2003; Ikari et al., 2009; Smith and Faulkner, 2010), which
are difcult to interpret physically as a negative b value is generally
assumed to indicate an increase in contact surface area with faster
slip. Karner et al. (1997) and Blanpied et al. (1998) also report
negative b values for granular quartz and granite.
The microphysical processes responsible for the observed rate-
and state-dependent behaviour are thermally activated and follow
Arrhenius-type behaviour (Chester, 1994; Blanpied et al., 1998;
Nakatani, 2001; Rice et al., 2001). They presumably include sub-
critical crack growth, crystal plasticity, diffusion and possibly
reaction at grain contacts. However, the rate and state formulation
does not include temperature. Chester (1994) showed that the
activation energy required for wet quartz gouge was consistent
with sub-critical crack growth at asperity contacts. This is sup-
ported by the conclusions of Frye and Marone (2002) that relative
humidity plays a role in the granular friction of quartz and alumina
as a result of chemically assisted mechanisms. This is clearly an
important area for future research, for if the physical mechanisms
responsible for frictional behaviour are known and characterized,
then the behaviour at a wider set of environmental conditions can
be better predicted. However, we note that future progress in this
area of research probably needs to combine experiments with
detailed microscopy (SEM and TEM) and theoretical and compu-
tational studies due to the complexity of the interactions that occur
at the nanoscale (Szlufarska et al., 2008; Mo et al., 2009).
3.3. Mechanics of dynamic rupture
The properties of faults during rupture are typically studied
using inversion of seismic data recorded during earthquakes to
compute the slip distribution of rupture events (kinematic model).
Various models allow computation of the stress eld, from which
the physical properties of the rupture are inferred (see review by
Kanamori and Brodsky, 2004). Recently, complementary laboratory
studies and eld measurements have led to a better understanding
of dynamic rupture, although this is currently a rapidly developing
area of research.
The slip history during large earthquakes appears to take the
form of a slip pulse rather than a self-similar crack-like rupture
(Heaton, 1990). In the slip-pulse model, only part of the fault plane
ruptures at any one time. Most seismological models now assume
source-time functions of small nite duration and thus implicitly
apply the slip-pulse model. However, this mode of slip still raises
Slow, V
1
Fast, V
2
Slip
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
a*ln(V /V )
2 1
b*ln(V /V )
2 1
D
c
Fig. 8. Idealized rate- and state-dependent frictional behaviour, in which the
mechanical response is dened by a stepwise change in velocity.
a number of unanswered questions. First, the spectral signature of
large earthquakes suggests that the corner frequency (lowest
frequency seismic waves) scales with rupture area which runs
contrary to the idea of a similar-size slip pulse regardless of the size
of the rupture. Additionally, in the early stages of failure the slip-
pulse model must accumulate approximately the correct amount of
slip appropriate to the overall rupture size. This leads to the ques-
tion of whether an earthquake event knows a priori how large it
will be (Marone and Richardson, 2006). Finally, modelling using
a rate-and-state framework has shown that slip pulses can only
develop under a restricted set of conditions where the initial shear
stress on the fault is low(Zheng and Rice, 1998). Lowstress (weak)
faults might well exist (Section 3.1), but a large proportion of faults
are inferred to have high stress levels (Townend and Zoback,
2000). Does this mean that slip pulse-type behaviour is not
possible for these faults?
Seismic records allow modelling of the mechanical properties of
rupture, but, in recent years, laboratory and eld measurements
have provided independent constraints. All these data yield an
understanding of the rupture process in terms of the energy
balance for an earthquake (Eq. (4); see Kanamori and Rivera, 2006).
The change in potential energy DU
e
(the sum of the elastic strain
energy and gravitational energy during earthquake slip) is the sum
of surface energy U
s
(to produce new crack surface area), kinetic
energy U
k
(radiated as seismic waves) and frictional energy U
f
(dissipated as heat):
DU
e
U
s
U
k
U
f
: (4)
Constraints on the kinetic energy are well known from model-
ling of seismic waves. Chester et al. (2005) estimated the fracture
energy for the Punchbowl fault in California by accounting for all
the fracture surface energy in both the damage zone and, more
importantly, the core zone, where nano-sized particles are present
and account for a signicant proportion of the fracture area. Hert-
zian fracture models (where fracturing occurs at grain contacts of
spherical grains) suggest that it is mechanically very difcult to
produce such small particle sizes. Sammis and Ben-Zion (2008)
suggested that shock loading and sub-critical crack growth under
compressive stress, or high strain rate tensile stress may be
responsible. The fracture energy expended during the formation of
pulverized rocks (Section 2.1) that are thought to develop during
seismic slip near the surface is not currently known. However,
Biegel et al. (2008) showed that off-fault damage will affect the
velocity of an earthquake slip pulse.
Di Toro et al. (2005) showed that a simple analysis of pseudo-
tachylyte veins provides broad constraints on the dynamic stress
during rupture. This analysis indicated very low(in terms of Byerlee
friction) shear stresses driving rupture. Di Toro et al. (2006)
corroborated these results with measurements, using the same
protolith, of dynamic friction at seismogenic slip velocity. While
frictional melting might result in low shear stresses driving slip
(after the effects of viscous braking have been overcome; Tsutsumi
and Shimamoto, 1997), the friction of other granular materials at
high velocity were unknown until recently.
Technical developments over the past 15 years (notably in the
laboratories of Shimamoto and co-workers) have allowed
measurement of the stresses during high-velocity frictional testing
in rotary shear apparatus. Fig. 9 shows typical results froma friction
experiment conducted at seismogenic slip velocity. These data
complement the data modelled from natural earthquake ruptures.
A key feature of the data in Fig. 9 is the dramatic weakening from
Byerlee levels of friction down to levels between 0.1 and 0.2. The
reasons for this weakening are many, and dependent on the
material tested. They include ash heating at asperity contacts
(Bowden and Tabor, 1950), silica gel formation (Goldsby and Tullis,
2002), thermal pressurization (Hirose and Bystricky, 2007), fric-
tional melt lubrication (Di Toro et al., 2006) and thermal decom-
position (Han et al., 2007). Recent modelling of the earthquake
process has started to combine and incorporate some of these
additional thermal factors into rate- and state-frictional frame-
works (Rempel and Rice, 2006; Rudnicki and Rice, 2006; Segall and
Rice, 2006; Noda, 2008; Noda et al., 2009).
The results in Fig. 9 have important implications. First, they
explain the long-standing debate on the development of frictional
melting in fault zones. Simple analyses show that extreme
temperatures are quickly reached due to frictional heating (see
Rice, 2006 for a summary). All these models assume friction coef-
cients commensurate with Byerlees law. If the shear traction
required for seismic slip reduces dramatically then the frictional
energy converted to heat is also dramatically reduced. It can also
explain the lack of any heat ow anomaly (Section 3.1) over the
seismogenic parts of the San Andreas fault.
Another feature of high velocity laboratory tests is the magnitude
of D
c
which is on the order of decimetres to metres. In slowfrictional
testing (Section 3.3) to determine rate- and state-friction parame-
ters, the slip-weakening distance is on the order of microns. A long-
standing debate focuses on the apparent discrepancy of D
c
values
derived from the laboratory versus values inferred from seismolog-
ical data, the latter being onthe order of a metre. This parameter may
scale with fault roughness or fault thickness (Scholz, 1988; Marone
and Kilgore, 1993). The emergence of models that account for the
different physical processes that occur during seismic slip as
opposed to slow laboratory frictional sliding seem to provide an
answer to the discrepancy. It appears the seismically derived D
c
is
a different parameter to that measured at slow slip rates in the
laboratory, and involves fundamentally different physics. This is
supported by the range of behaviour observed in high velocity
experiments, modelling and from other geological observations that
include thermal pressurization (Wibberley and Shimamoto, 2005;
Bizzarri and Cocco, 2006), ash heating, thermal decomposition
and frictional melting. The value of D
c
might be envisaged to increase
as thermally activated processes continue to produce weakening
with continued slip (in the same way that the boundary between
surface energy and heat varies during slip as suggested by Tinti et al.,
2005). It is clear that smaller critical slip distances must exist,
otherwise small earthquakes could never nucleate and that a sensi-
tive balance exists between the energy required for the work of
fracture and that converted to heat. Unfortunately, testing these
hypotheses with seismological data is hampered by the limited
frequency bandwidth from which kinematic models of rupture are
derived and also the constraints on the resolution and uniqueness of
Fig. 9. Results from a typical high-velocity friction experiment on kaolinite (unpub-
lished data from Faulkner, Mitchell, Hirose and Shimamoto). It shows a number of
characteristic features including the peak friction, slip weakening distance and steady-
state sliding. The strength recovery upon deceleration of slip is also shown.
parameters such as D
c
(Spudich and Guatteri, 2004; Tinti et al.,
2009). Field observations by Kirkpatrick and Shipton (in press)
conrm that slip weakening mechanisms are likely to be spatially
and temporally variable across an earthquake fault surface.
4. Fluid ow in fault zones
In recent years our understanding of uid ow through faults
has advanced greatly. The typical structure of fault zones (Section
2.1), with a core and damage zone, has provided the framework
within which to place laboratory measurements of the uid ow
properties of natural and synthetic fault products into context
(Fig. 10). Fault-related uid ow has also been investigated via
a number of indirect data sources such as migrating seismicity at
depth, shallow reservoir-induced seismicity, springs, geysers and
geothermal systems. These sources have provided some rst-order
constraints on the rates of uid ow in natural fault zones at depth,
and at length scales unavailable to lab experiments.
In the original Caine et al. (1996) fault core and damage zone
model of fault architecture the fault core was visualised as
providing an across-fault barrier to ow and the fractured damage
zone as an along/up-fault conduit. However, the varying fault
architectures outlined in Section 2.1 gives rise to a much more
complex set of fault zone hydraulic behaviours. The intricate
structure of lowand high permeability features within a fault zone
can lead to extreme permeability heterogeneity and anisotropy.
The permeability of a fault zone, both in-plane and perpendicular to
the plane (across-fault) is governed both by the permeability of the
individual fault rocks/fractures and, critically, by their geometric
architecture in three dimensions (e.g. Lunn et al., 2008). For
example, rocks from the fault core are commonly rich in phyllosi-
licates, which typically have low permeability, but only form
barriers to ow if they are continuous throughout the fault plane
(Faulkner and Rutter, 2001). Open fractures and slip surfaces (both
within the fault core and the surrounding damage zone) have
a permeability governed by their aperture distribution, which is in
turn inuenced by their orientation to the present day local stress
eld. Such fractures and slip surfaces may have a substantially
greater permeability than the host rock; however, their net effect
on bulk along-fault and across-fault ow, depends entirely on their
connectivity and ability to cross-cut other lower permeability units.
4.1. The hydraulic properties of the fault core and its inuence on
uid ow
In natural faults two distinct types of gouge are present. The rst
are granular materials composed of broken, irregular but roughly
equant clasts (in the sense that their long and short axes are
approximately equal), and the second are gouges that contain some
proportion of phyllosilicate material. Relatively few data on the
permeability of granular gouges are available but they tend to
develop a characteristic grain size distribution (Sammis et al., 1987;
Marone and Scholz, 1989) that may suggest a similar permeability
development for all these materials. Zhang and Tullis (1998)
measured the permeability development in synthetic quartz
gouge at a normal stress of 25 MPa. They found that at shear strains
up to w10 the permeability is reduced by two to three orders of
magnitude. This is in agreement with more recent ndings of
Crawford et al. (2008) and Main et al. (2000). Beyond this shear
strain (to a shear strain w200), Zhang and Tullis (1998) found the
permeability dropped by a further two to three orders of magnitude
and that a permeability anisotropy of one order of magnitude
developed. This was due to the formation of localized, ne-grained
Y shears. These laboratory data are in agreement with eld obser-
vations and permeability measurements from boreholes that
suggest a signicant drop in cross-fault permeability in deforma-
tion band-dominated faults as the fault core develops through-
going slip surfaces (Shipton et al., 2002, 2005).
Fault zones rich in phyllosilicate material tend to have lower
permeabilities than quartz and/or framework silicate-rich gouges.
Information on the uid ow properties of phyllosilicate-rich fault
zones is necessary to understand uid ow associated with fault
creep (e.g. Faulkner and Rutter, 2001) and earthquake slip (e.g.
Wibberley and Shimamoto, 2005; Yamashita and Suzuki, 2009), as
many large faults contain signicant proportions of clays. Where
the uid-ow properties of fault zones are needed to evaluate the
robustness of a fault-bounded hydrocarbon prospect or the eld
compartmentalizing effects of intra-reservoir faults, estimating the
possible phyllosilicate content of the fault zone is critical, along
with reservoir juxtaposition geometry. Based on eld-observations
of fault zones, the two main mechanisms that entrain phyllosili-
cates (typically shale and/or clays) into fault zones in layered
sandstone e shale sequences are shale or clay smearing (e.g. van
der Zee and Urai, 2005) and abrasional mixing (e.g. Yielding et al.,
fracture
density
permeability
a b
Fig. 10. Some physical properties of fault zones related to their structure (damage zone and fault core). (a) Single fault core and (b) multiple fault core, which illustrates the resulting
complexity in characterizing the resultant properties.
1997). The necessity of estimating fault properties from limited
datasets led to algorithms for estimating fault zone composition
which assume one or the other mechanism is operative. For shale/
clay smearing, the Clay Smear Potential (Weber et al., 1978) or the
Shale Smear Factor (Lindsay et al., 1993) rely on parameters such as
the thickness of the shale/clay source bed, distance of a point on the
fault from that source bed, and/or throw. These algorithms only
predict whether or not the smear along the fault is discontinuous
(likely leading to leakage) for a given fault throw and, if so, where.
On the other hand, the abrasional mixing mechanism led Yielding
et al. (1997) to propose the Shale Gouge Ratio (SGR) algorithm,
a ratio or percentage of shale in a silicilastic fault zone, which
simply assumes that any one point on the fault has a composition
identical to the average composition of the sequence past which
that point has slipped. In terms of sandstone eshale sequences, this
is extremely practical to implement, because the net volume of clay
(Vcl) logs from nearby wells can be extrapolated onto the fault (in
cases of simple stratigraphy), from which SGR is calculated for all
points on the fault for which wall-rock Vcl data exist. In reality, fault
zones are much more complex and local small-scale variations can
exist even in abrasive fault zones where the rule generally holds.
However, quantitatively constraining this variation may in future
help predict uncertainties in SGR-based evaluations.
In evaluating the sealing potential of a fault, analysis of juxta-
positions of reservoir/carrier beds against other reservoir beds is
critical. Basic geometric evaluation of such likely leak points is
easily done by fault-plane mapping of hanging-wall / footwall
juxtapositions, commonly called Allan diagrams (Allan, 1989).
However, eld-based studies show that faults are commonly both
segmented (in both dip and strike, e.g. Nemser and Cowan, 2009),
have multiple slip surfaces and involve varying degrees of ductile
deformation (fault-related folding; Section 2.3). Thus the net throw
observed on one fault at the scale of seismic resolution is in reality
often divided over two or more slip surfaces which share the
displacement at the sub-seismic scale. Such fault zone structure
therefore implies there might be reservoir-reservoir juxtapositions
across individual fault strands even where the entire fault
completely offsets the reservoir. Thus, better prediction of the likely
segmentation and slip zone bifurcation is needed, particularly the
incorporation of lenses of host-rock into the fault zone (e.g. Van der
Zee et al., 2008, Schmatz et al., 2010).
The recognition that fault zones in siliclastic sedimentary basins
are typically sand-shale gouges with uid barrier/transmission
behaviour governed by similar principles as shale-rich top-seals, led
to the application of capillary sealing theory. By using calibration
data for given fault-rock clay compositions and using laboratory
measurements (e.g. Sperrevik et al., 2002) or eld post-mortem
results (e.g. Bretan et al., 2003), the SGR method can be used to map
estimated capillary threshold pressures on the fault, yielding the
likely maximum hydrocarbon column height trapped against the
fault. The morerecent recognitionthat fault zone heterogeneitymay
lead to connected weak points which may provide a pathway for
hydrocarbon leakage has emphasized the need to predict better the
variability in fault zone structure and composition.
Similarly, fault zone permeability, particularly for incorporating
fault impact into reservoir simulators via the transmissibility
multiplier (Manzocchi et al., 1999), is commonly estimated from
SGR e permeability algorithms. Generally, there is a non-linear
dependence of the permeability on the fault zone clay content
under hydrostatic conditions due to the grain size difference and
compaction characteristics. For example, as the clay content
increases to between 25 and 40 volume % (the theoretical porosity
minimum, Revil et al., 2002) the clay particles sit in the pore space
between quartz, and the compaction characteristics are largely
controlled by the quartz framework. The permeability is strongly
controlled by the fraction of clay (Takahashi et al., 2007; Crawford
et al., 2008). At larger percentages of clay, the permeability is less
sensitive to the magnitude of the clay fraction. As the clay compacts
more readily than the quartz, the porosity minimum varies with
effective pressure (Crawford et al., 2008). Shear-enhanced
compaction is much less pronounced for the clay component than
for the quartz end-member. As a result, the clay-rich mixtures do
not reduce their permeability by much in comparison to the quartz-
rich gouges, which undergo a signicant permeability reduction
(Takahashi et al., 2007; Crawford et al., 2008).
Many measurements of natural clay-rich fault rocks are avail-
able (Faulkner and Rutter, 2000; Faulkner et al., 2003; Wibberley
and Shimamoto, 2003; Tsutsumi et al., 2004; Mizoguchi et al.,
2008). They all demonstrate the very low permeability of this
material. Natural gouge can have permeability anisotropy of up to
three orders of magnitude (Faulkner and Rutter, 2000). For
synthetic phyllosilicate gouges this value appears to be much lower
(Zhang et al., 1999), presumably due to the nature and distribution
of the authigenic clay phases that develop in natural gouges. Recent
work has measured the intensity of clay fabrics by, for example, X-
ray texture goniometry (Solum and van der Pluijm, 2009; Haines
et al., 2009). The gouges have a generally weak preferred orienta-
tion of the clays. Solum and van der Pluijm suggest that this indi-
cates that the clay fabrics are localized phenomena. Haines et al.
(2009) imply that the lack of clay fabric indicates the perme-
ability anisotropy must also be low. However, previous work on clay
fabrics shows that the permeability anisotropy is not due to clay
alignment; indeed this can only account for a permeability
anisotropy less than an order of magnitude (Faulkner and Rutter,
1998). Alternating microlayers of porous granular material and
ne grained clay-rich material are observed in the microstructure
and explain the anisotropy (Faulkner and Rutter, 1998; Faulkner,
2004). Furthermore, authigenic clay growth observed in TEM
images shows growth randomly in pore space in the stress shadows
of larger relict grains (Rutter et al., 1986).
The pressure dependence of the permeability is remarkably
similar for most clay-rich fault gouges, whether synthetic or natural
(see Faulkner, 2004). Faulkner and Rutter (2000, 2003) showed that
temperature and pore uid chemistry can strongly affect the
permeability of natural clay-rich rocks by altering physico-chemical
interactions between the rocks and aqueous pore uid.
The stress history of fault gouge, particularly clay-rich gouge,
has been shown to be an important control on the permeability.
Bolton et al. (1998) and Zhang et al. (1999) have shown that
permeability is reduced in sheared gouges that have undergone
normal consolidation, but may increase their permeability if they
have undergone overconsolidation and are subsequently sheared at
lower effective pressure conditions. Indeed the log-linear rela-
tionship of both porosity and permeability with mean effective
stress in clay-rich gouges fromthe Median Tectonic Line, Japan, and
dependence on the anisotropy of the stress regime suggest that
uid owmodelling in such fault zones may be based around a Cam
Clay -type soil mechanics framework (Wibberley et al., 2008).
The temporal evolution of the permeability of fault gouges,
particularly at hydrothermal conditions, has been a recent focus of
research. Pure quartz gouges (Nakatani and Scholz, 2004a;
Yasuhara et al., 2005; Giger et al., 2007) and quartzofeldspathic
mixtures (Olsen et al., 1998; Tenthorey et al., 1998; Morrow et al.,
2001) have been tested experimentally. Permeability reduction is
found to be initially rapid and progressively slows with time. The
rate of permeability reduction increases with increasing tempera-
ture (Olsen et al., 1998; Tenthorey et al., 2003; Nakatani and Scholz,
2004a; Yasuhara et al., 2005; Giger et al., 2007). The processes
responsible for the permeability reduction are interpreted to be
solution-precipitation processes (Nakatani and Scholz, 2004a;
Giger et al., 2007) and also precipitation of authigenic minerals
following the breakdown of quartz and feldspar (Tenthorey et al.,
1998). These processes have been successfully modelled
(Aharonov et al., 1998; Nakatani and Scholz, 2004b). These studies
have demonstrated that sealing will occur on time scales
commensurate with those of earthquake recurrence times (Morrow
et al., 2001).
4.2. Damage zone permeability
The permeability of fault damage zones is governed by the host-
rock permeability and the presence and geometric composition of
both macro-scale fracture networks (which will increase perme-
ability), and of low permeability deformation and compaction
bands (which will decrease permeability). Fault damage zone
permeability in low porosity rocks (Balsamo et al., 2010) is gener-
ally fracture-dominated and governed by the connectivity of the
macro-scale fracture network (Fig. 10). This contrasts with high
porosity rocks where the damage zone may be more complex and
permeability is governed by the frequency and connectivity of both
low permeability deformation bands and of high permeability slip
surfaces (Lunn et al., 2008). Both macro-scale fracture networks
and deformation bands have been shown to decrease in frequency
with increasing distance from the fault core (Rawling et al., 2001;
Shipton et al., 2002; Wilson et al., 2003; Mitchell and Faulkner,
2009). Wibberley and Shimamoto (2003) measured the perme-
ability of rocks collected from the surface trace of the Median
Tectonic line in Japan. Their measurements showed permeability
variations over several orders of magnitude, which can be
explained by the lithological variation of the fault zone and also the
structural complexity. Fig. 10 shows how this structural complexity
might result in permeability heterogeneity from the variation in
microfracture densities observed around multiple fault cores.
The permeability of the host rock within the damage zone is
controlled by the frequency and orientation of microfractures. One
potential problem with direct measurement of damage zone
permeability from rocks collected from the surface outcrop of fault
zones is that they may be subject to weathering or modication
since the fault related microfracture network was produced
(MorrowandLockner, 1994). Measurements of rocks recoveredfrom
depth fromfault-zone drilling projects may overcome this problem.
Experiments indicate permeability of initially lowporosity rocks
taken to failure increase by two to three orders of magnitude
(Simpson et al., 2001; Oda et al., 2002; Uehara and Shimamoto,
2004; Mitchell and Faulkner, 2008). For initially high porosity
rocks, the permeability may signicantly decrease with deforma-
tion in the damage zone (Main et al., 2000). Measurements of
permeability fromthe intact state of rocks to their failure stress can
be scaled to the levels and distribution of damage seen surrounding
fault zones (see Section 2.1), if a common factor between the two
can be found. For example one such common factor is the micro-
fracture density which can be measured in the eld and then
compared to that in experiments at various stages of damage,
where the permeability may be readily measured.
4.3. Estimating bulk fault zone permeability
The physical characteristics of fault damage zones are described
in Section 2.1. Estimates of the bulk fault zone permeability, for fault-
perpendicular and fault-parallel ow, are derived from numerical
models that simulate owthrough the fault zone (Brown and Bruhn,
1998; Jourde et al., 2002; Matthai and Belayneh, 2004; Odling et al.,
2004; Lunn et al., 2008) Such models showsignicant channelling of
ow within fault zones into a small number of focussed ow paths.
Similar channelling effects are observed in ow experiments within
individual fractures (Brown et al., 1998; Beeler and Hickman, 2004).
Observations from boreholes that penetrate faults, as well as along-
fault and across-fault pumping tests, are necessary to determine bulk
fault-zone permeability and to validate numerical models (Evans
et al., 2005; Medeiros et al., 2010).
Very few studies have measured bulk fault zone permeability
directly using boreholes. However, a number of secondary data
sources allow estimation of along-fault permeability. Talwani et al.
(1999) estimated the permeability of a shallow fault zone using
sinusoidal pressure oscillations in boreholes from lake level uctua-
tions. Their analysis shows fault zone permeability, in faults that are
subject to reservoir-induced seismicity, are between 1.1 10
15
and
1.78 10
15
m
2
. Tadokoro et al. (2000) estimated along-strike fault
(damage) zone permeabilities around 1e10 10
15
m
2
from the
migration rates of induced seismicity during borehole injection
experiments in the Nojima fault zone following the Kobe 1995
earthquake. At deeper levels, Shapiro et al. (1997) found crustal
permeability to be w10
16
m
2
in the KTB borehole in the depth
interval 7.5e9 km. Acompilation of fault zone permeabilities derived
from reservoir-induced seismicity data can be found in Talwani et al.
(2007). Usingthesametechniques of migratingpatterns of seismicity,
but at deeper levels in the brittle crust, Miller et al. (2004) estimated
fault zone permeability to be 4 10
11
m
2
immediately following the
M6 Colorito earthquake sequence of 1997 in central Italy. Noir et al.
(1997) inferred a higher fault zone permeability of 10
8
m
2
for the
Dobi earthquake sequence in Afar in 1989.
4.4. Spatial and temporal variability of fault zone permeability
A number of recent studies have shown that fault zone
permeability is highly heterogeneous both spatially and temporally.
Studies examining the distribution of geothermal spring tempera-
tures along the Borax Lake fault, Oregon, USA, show that neigh-
bouring springs a few metres apart can have widely different
temperatures (Fairley and Hinds, 2004). These springs suggest that
high permeability pathways exist as discrete structures in the Borax
fault damage zone, and that individual ow paths have the capa-
bility to transport uids rapidly from depth parallel to the fault
plane. Dockrill and Shipton (2010) use observations of natural
leakage of CO
2
along faults in Utah, USA, to show that along-fault
ow is occurring at a few discrete locations along strike, and that
these discrete along-fault ow channels have migrated over time.
The presence of a modern oil seep at one location on the fault also
indicates the existence of discrete unconnected along-fault pipes
that provide pathways for uid owfromlithologies at depth to the
surface and are unconnected to shallower horizons. Evans et al.
(2005), during an injection test in the Soultz borehole, observed
that some 95% of the ow entered the rock mass at just 10 major
owing fractures. Do Nascimento et al. (2005) show that pressure
changes as low as 0.5 kPa (equivalent to 5 cm of water head) are
enough to trigger transient changes in permeability that are
spatially correlated, and related to seismicity below a water
reservoir.
In the hydrocarbon exploration and production industry,
evidence for the spatial and temporal variation of fault perme-
ability exists but is usually limited to indirect inferences from
reservoir uid and pressure data either side of the faults, because
faults are generally avoided when drilling as they give rise to
a number of drilling problems (e.g. Grauls et al., 2002). Neverthe-
less, studies of charge timing in fault-bounded blocks (e.g. Residual
Salt Analyses of uid inclusions in reservoir pore cements) can
show an increase in the hydrocarbon buoyancy pressure differen-
tial across the fault through time, attributed to increasing sealing
properties as compaction affects the fault during progressive burial.
Transient increases in up-fault permeability during periodic
reactivation may lead to leakage of fault-bounded hydrocarbon
traps, as has been found for several cases in the northern North Sea
for example (Wiprut and Zoback, 2002). On production time scales,
faults which initially sealed signicant hydrocarbon-related pres-
sure differences may become leaking as one compartment is
depleted (e.g. Dincau, 1998) and stress changes related to depletion
may render the fault unstable in certain cases, leading to up-fault
leakage during slip (Cuisiat et al, 2010).
One exception to the general paucity of direct industry
measurements of fault permeability is the Pathnder well on
Eugene Island, Gulf of Mexico, which shows an active normal
growth fault in an overpressured siliclastic setting to have a rela-
tively high up-fault permeability of w1 mD which sharply
increases to around 1D as uid pressure is further increased
towards the minimum effective principal stress (Losh and Haney,
2006). A similar observation was made from data from an ODP
borehole through the basal thrust of the Barbados accretionary
prism (Screaton et al., 1990). Seismic data shot on Eugene Island at
two different times over a seven year interval image the same
overpressure pulse in two different places e around 1 km apart,
leading to large-scale permeability estimates of around 0.1 Darcy
and the suggestion that faults can burp such overpressure pulses
upwards (Haney et al., 2005).
Healing of macro and microfracture networks in damaged rocks,
in the same way as for healing of fault gouge, is an important
temporal process in fault zones. The lifetime of fracture networks is
needed to predict cyclic fault zone permeability at depth. Seismo-
logical evidence has shown that the recovery of P and S wave
velocities (presumably related to healing of fracture damage)
following the 1992 Landers earthquake is quite rapid (<10 years)
(Vidale and Li, 2003). This recovery may only partially heal fault
fracture damage as the low-velocity zone surrounding faults
appears to be long-lived (Cochran et al., 2009).
5. Concluding remarks
Recent advances in the study of the structure, mechanics and
uid ow properties of faults and fault systems have been
reviewed. The importance of the interplay of these three properties
was emphasized. We conclude this work by highlighting some of
the key areas of ongoing research.
In terms of fault zone structure, the heterogeneity and along
fault variability are still poorly known. For example, seismological
methods are necessary to determine the structure of fault zones at
depth. Hence, if we are to reconcile the structure of fault zones at
depth with that observed at the surface, we must better understand
the dependence of seismological parameters on the physical
properties of the fault rocks. Basin-scale fault systems can benet
from revised models of fault growth related to geometric and
kinematic coherence. The impact of new 3D seismic datasets has
the potential to greatly enhance our understanding of fault growth
by providing a detailed view of growth strata that are intimately
associated with the growth of faults and development of fault
systems.
Our understanding of the mechanics of faulting and earthquakes
is still limited. The currently unresolved question of weak faulting
has helped to focus efforts, but direct observation via fault-drilling
projects is necessary to resolve 2010. Even then, the drill hole has to
penetrate a representative portion of the fault zone, with all the
inherent problems of fault-zone heterogeneity previously
mentioned. We can improve our knowledge of the mechanics of
earthquakes by integrating data from seismology, experiments and
eld geology. A key aspect is to improve our understanding of the
manifestations of seismic slip and its thermal effects in natural fault
rocks, if this is possible. This may be aided in the future with
comparison with microstructures developed in high velocity
experiments. Experiments (at both high and lowvelocity) must aim
to understand the underlying physics of slip.
Fluid ow around faults is dictated by 3D fault zone structure
and, as previously mentioned, this is likely to be heterogeneous.
Investigations of along-fault ow, in particular within crystalline
rocks at depth, showthat owcan be dominated by a small number
of fractures within the surrounding damage zone. Recent efforts
have helped to characterize the damage surrounding faults and we
can potentially reconcile this with laboratory measurements of
various fault-zone components (e.g. fault core or damage zone).
However, there remains a pressing need for a greater number of in
situ measurements over large areas, such as those exposed within
tunnels that characterize both the permeability of individual
features and whole fault zones at depth. Our understanding of the
temporal evolution of fault-zone permeability is still limited, but
we can address this by a combination of eld and laboratory
measurements.
Overall, as many aspects of faulting and fault systems are highly
interrelated, an integrated approach is necessary to make progress
in understanding their structure, mechanics and uid ow prop-
erties. This integrated approach will necessarily involve many
different disciplines, from eld geology, laboratory measurements,
geophysical measurements, modelling (numerical and experi-
mental), and direct observation of faults through drilling.
Acknowledgements
DRF thanks the Departamento de Geologia, Universidad de Chile
for their hospitality during the preparation of this manuscript. We
thank Bob Holdsworth for providing a thorough review that
improved the manuscript.
References
Aharonov, E., Tenthorey, E., Scholz, C.H., 1998. Precipitation sealing and diagenesis e
2. Theoretical analysis. Journal of Geophysical Research e Solid Earth 103 (B10),
23969e23981.
Allan, U.S., 1989. Model for hydrocarbon migration and entrapment within faulted
structures. American Association of Petroleum Geologists Bulletin 73, 803e811.
Ampuero, J.P., Rubin, A.M., 2008. Earthquake nucleation on rate and state faults -
aging and slip laws. Journal of Geophysical Research e Solid Earth 113 (B1).
Anastasio, D.J., Erslev, E.A., Fisher, D.M., 1997. Preface: fault-related folding. Journal
of Structural Geology 19, vevi.
Anders, M.H., Wiltschko, D.V., 1994. Microfracturing, paleostress and the growth of
faults. Journal of Structural Geology 16 (6), 795e815.
Axen, G.J., 2007. Research focus: signicance of large-displacement, low-angle
normal faults. Geology 35 (3), 287e288.
Balsamo, F., Storti, F., Salvini, F., Silva, A., Lima, C., 2010. Structural and petrophysical
evolution of extensional fault zones in low-porosity, poorly lithied sandstones
of the Barreiras Formation, NE Brazil. Journal of Structural Geology 32 (11),
1806e1826.
Baudon, C., Cartwright, J., 2008. Early stage evolution of growth faults: 3D seismic
insights from the Levant Basin, Eastern Mediterranean. Journal of Structural
Geology 30 (7), 888e898.
Beeler, N.M., Hickman, S.H., 2004. Stress-induced, time-dependent fracture closure
at hydrothermal conditions. Journal of Geophysical Research 109.
Berg, S.S., Skar, T., 2005. Controls on damage zone asymmetry of a normal fault
zone: outcrop analyses of a segment of the Moab fault, SE Utah. Journal Of
Structural Geology 27 (10), 1803e1822.
Beroza, G.C., Ellsworth, W.L., 1996. Properties of the seismic nucleation phase.
Tectonophysics 261 (1e3), 209e227.
Biegel, R.L., Sammis, C.G., 2004. Relating fault mechanics to fault zone structure.
Advances in Geophysics 47, 65e111.
Biegel, R.L., Sammis, C.G., Rosakis, A.J., 2008. An experimental study of the effect of
off-fault damage on the velocity of a slip pulse. Journal of Geophysical Research
e Solid Earth 113 (B4).
Bizzarri, A., Cocco, M., 2006. A thermal pressurization model for the spontaneous
dynamic rupture propagation on a three-dimensional fault: 2. Traction evolu-
tion and dynamic parameters. Journal of Geophysical Research e Solid Earth 111
(B5).
Blanpied, M.L., Lockner, D.A., Byerlee, J.D., 1992. An earthquake mechanism based on
rapid sealing of faults. Nature 358 (6387), 574e576.
Blanpied, M.L., Marone, C.J., Lockner, D.A., Byerlee, J.D., King, D.P., 1998. Quantitative
measure of the variation in fault rheology due to uiderock interactions.
Journal of Geophysical Research e Solid Earth 103 (B5), 9691e9712.
Blenkinsop, T.G., 2008. Relationships between faults, extension fractures and veins,
and stress. Journal of Structural Geology 30 (5), 622e632.
Bolton, A.J., Maltman, A.J., Clennell, M.B., 1998. The importance of overpressure
timing and permeability evolution in ne-grained sediments undergoing shear.
Journal of Structural Geology 20 (8), 1013e1022.
Boness, N.L., Zoback, M.D., 2006. A multiscale study of the mechanisms controlling
shear velocity anisotropy in the San Andreas Fault Observatory at depth.
Geophysics 71 (5), F131eF146.
Bonnet, E., Bour, O., Odling, N.E., Davy, P., Main, I., Cowie, P., Berkowitz, B., 2001.
Scaling of fracture systems in geological media. Reviews of Geophysics 39 (3),
347e383.
Bos, B., Peach, C.J., Spiers, C.J., 2000. Frictional-viscous ow of simulated fault gouge
caused by the combined effects of phyllosilicates and pressure solution. Tec-
tonophysics 327 (3-4), 173e194.
Bos, B., Spiers, C.J., 2001. Experimental investigation into the microstructural and
mechanical evolution of phyllosilicate-bearing fault rock under conditions
favouring pressure solution. Journal of Structural Geology 23 (8), 1187e1202.
Bowden, F.P., Tabor, D., 1950. The Friction and Lubrication of Solids. Oxford
University Press.
Bretan, P., Yielding, G., Jones, H., 2003. Using calibrated shale gouge ratio to estimate
hydrocarbon column heights. AAPG Bulletin 87 (3), 397e413.
Brister, B.S., Stephens, W.C., Norman, G.A., 2002. Structure, stratigraphy, and
hydrocarbon system of a Pennsylvanian pull-apart basin in north-central Texas.
AAPG Bulletin 86 (1), 1e20.
Brown, S., Caprihan, A., Hardy, R., 1998. Experimental observation of uid ow
channels in a single fracture. Journal of Geophysical Research e Solid Earth 103
(B3), 5125e5132.
Brown, S.R., Bruhn, R.L., 1998. Fluid permeability of deformable fracture networks.
Journal of Geophysical Research e Solid Earth 103 (B2), 2489e2500.
Brune, J.N., Henyey, T.L., Roy, R.F., 1969. Heat ow, stress, and rate of slip along the
San Andreas Fault, California. Journal of Geophysical Research 74, 3821e3827.
Bull, J.M., Barnes, P.M., Lamarche, G., Sanderson, D.J., Cowie, P.A., Taylor, S.K.,
Dix, J.K., 2006. High-resolution record of displacement accumulation on an
active normal fault: implications for models of slip accumulation during
repeated earthquakes. Journal of Structural Geology 28 (7), 1146e1166.
Byerlee, J., 1978. Friction of rocks. Pageoph 116, 615e626.
Byerlee, J., 1990. Friction, overpressure and fault normal compression. Geophysical
Research Letters 17 (12), 2109e2112.
Byerlee, J., 1993. Model for episodic ow of high-pressure water in fault zones
before earthquakes. Geology 21 (4), 303e306.
Caine, J.S., Evans, J.P., Forster, C.B., 1996. Fault zone architecture and permeability
structure. Geology 24 (11), 1025e1028.
Caine, J.S., Bruhn, R.L., Forster, C.B., 2010. Internal structure, fault rocks, and infer-
ences regarding deformation, uid ow, and mineralization in the seismogenic
stillwater normal fault, Dixie Valley, Nevada. Journal of Structural Geology 32
(11), 1576e1589.
Cartwright, J.A., Trudgill, B.D., Manseld, C.S., 1995. Fault growth by segment
linkage - an explanation for scatter in maximum displacement and trace length
data from the canyonlands grabens of SE Utah. Journal of Structural Geology 17
(9), 1319e1326.
Cembrano, J., Gonzalez, G., Arancibia, G., Ahumada, I., Olivares, V., Herrera, V., 2005.
Fault zone development and strain partitioning in an extensional strike-slip
duplex: A case study from the Mesozoic Atacama fault system, Northern Chile.
Tectonophysics 400 (1e4), 105e125.
Chery, J., Zoback, M.D., Hickman, S., 2004. A mechanical model of the San Andreas
fault and SAFOD pilot hole stress measurements. Geophysical Research Letters
31 (15).
Chester, F.M., 1994. Effects of temperature on friction: constitutive equations and
experiments with quartz gouge. Journal of Geophysical Research 99 (B4),
7247e7261.
Chester, F.M., Logan, J.M., 1986. Implications for mechanical-properties of brittle
faults from observations of the Punchbowl fault zone, California. Pure and
Applied Geophysics 124 (1-2), 79e106.
Chester, F.M., Friedman, M., Logan, J.M., 1985. Foliated cataclasites. Tectonophysics
111 (1e2), 139e146.
Chester, F.M., Evans, J.P., Biegel, R.L., 1993. Internal structure and weakening
mechanisms of the San-Andreas fault. Journal of Geophysical Research e Solid
Earth 98 (B1), 771e786.
Chester, J.S., Chester, F.M., Kronenberg, A.K., 2005. Fracture surface energy of the
Punchbowl fault, San Andreas system. Nature 437 (7055), 133e136.
Chiaraluce, L., Chiarabba, C., Collettini, C., Piccinini, D., Cocco, M., 2007. Architecture
and mechanics of an active low-angle normal fault: Alto Tiberina Fault,
northern Apennines, Italy. Journal of Geophysical Research e Solid Earth 112
(B10), B10310. doi:10.1029/2007JB005015.
Childs, C., Manzocchi, T., Walsh, J.J., Bonson, C.G., Nicol, A., Schopfer, M.P.J., 2009. A
geometric model of fault zone and fault rock thickness variations. Journal of
Christie-Blick, N., Biddle, K.T., 1985. Deformation and basin formation along strike-
slip faults. Strike-slip deformation, basin formation, and sedimentation. pp. 1e34.
Cochran, E.S., Li, Y.G., Shearer, P.M., Barbot, S., Fialko, Y., Vidale, J.E., 2009. Seismic
and geodetic evidence for extensive, long-lived fault damage zones. Geology 37
(4), 315e318.
Collettini, C., Holdsworth, R.E., 2004. Fault zone weakening and character of slip
along low-angle normal faults: insights from the Zuccale fault, Elba, Italy.
Journal of the Geological Society 161, 1039e1051.
Collettini, C., Sibson, R.H., 2001. Normal faults, normal friction? Geology 29 (10),
927e930.
Collettini, C., Niemeijer, A., Viti, C., Marone, C., 2009a. Fault zone fabric and fault
weakness. Nature.
Collettini, C., Viti, C., Smith, S.A.F., Holdsworth, R.E., 2009b. Development of inter-
connected talc networks and weakening of continental low-angle normal
faults. Geology 37 (6), 567e570.
Cook, J.E., Dunne, W.M., Onasch, C.A., 2006. Development of a dilatant damage zone
along a thrust relay in a low-porosity quartz arenite. Journal of Structural
Geology 28 (5), 776e792.
Cosgrove, J.W., Ameen, M.S., 2000. Forced folds and fractures. In: Geological Society
of London, Special Publications, vol. 169.
Cowie, P.A., Attal, M., Tucker, G.E., Whittaker, A.C., Naylor, M., Ganas, A.,
Roberts, G.P., 2006. Investigating the surface process response to fault inter-
action and linkage using a numerical modelling approach. Basin Research 18
(3), 231e266.
Cox, S.F., Knackstedt, M.A., Braun, J., 2001. Principles of structural control on
permeability and uid ow in hydrothermal systems. Reviews in Economic
Geology 14, 1e24.
Crawford, B.R., Faulkner, D.R., Rutter, E.H., 2008. Strength, porosity, and perme-
ability development during hydrostatic and shear loading of synthetic quartz-
clay fault gouge. Journal of Geophysical Research e Solid Earth 113 (B3).
Cuisiat, F., Jostad, H.P., Andresen, L., Skurtveit, E., Skomedal, E., Hettema, M.,
Lyslo, K., 2010. Geomechanical integrity of sealing faults during depressuriza-
tion of the Statfjord eld. Journal of Structural Geology 32 (11), 1754e1767.
Cristallini, E.O., Allmendinger, R.W., 2001. Pseudo 3-D modeling of trishear fault-
propagation folding. Journal of Structural Geology 23 (12), 1883e1899.
dAlessio, M., Martel, S.J., 2005. Development of strike-slip faults from dikes,
Sequoia National Park, California. Journal of Structural Geology 27 (1), 35e49.
Dawers, N.H., Anders, M.H., 1995. Displacement-length scaling and fault linkage.
Journal of Structural Geology 17 (5), 607.
de Joussineau, G., Aydin, A., 2007. The evolution of the damage zone with fault
growth in sandstone and its multiscale characteristics. Journal of Geophysical
Research 112.
De Paola, N., Collettini, C., Faulkner, D.R., Trippetta, F., 2008. Fault zone architecture
and deformation processes within evaporitic rocks in the upper crust. Tectonics
27 (4).
Di Toro, G., Pennacchioni, G., 2005. Fault plane processes and mesoscopic structure
of a strong-type seismogenic fault in tonalites (Adamello batholith, Southern
Alps). Tectonophysics 402 (1e4), 55e80.
Di Toro, G., Nielsen, S., Pennacchioni, G., 2005. Earthquake rupture dynamics frozen
in exhumed ancient faults. Nature 436, 1009e1012.
Di Toro, G., Hirose, T., Nielsen, S., Pennacchioni, G., Shimamoto, T., 2006. Natural and
experimental evidence of melt lubrication of faults during earthquakes. Science
311 (5761), 647e649.
Dieterich, J.H., Kilgore, B., 1996. Implications of fault constitutive properties for
earthquake prediction. Proceedings of the National Academy of Sciences of the
United States of America 93 (9), 3787e3794.
Dieterich, J.H., Kilgore, B.D., 1994. Direct observation of frictional contacts - new
insights for state-dependent properties. Pure and Applied Geophysics 143
(1e3), 283e302.
Dincau, A.R., 1998. Prediction and timing of production induced fault seal break-
down in the South Marsh Island 66 gas eld. Gulf Coast Association of
Geological Societies Transactions XLVIII, 21e32.
Do Nascimento, A.F., Lunn, R.J., Cowie, P.A., 2005. Modeling the heterogeneous
hydraulic properties of faults using constraints from reservoir-induced seis-
micity. Journal of Geophysical Research e Solid Earth 110 (B9).
Dockrill, B., Shipton, Z.K., 2010. Structural controls on leakage from a natural CO
2
geologic storage site: central Utah. U.S.A. Journal of Structural Geology 32 (11),
1768e1782.
Dor, O., Ben-Zion, Y., Rockwell, T.K., Brune, J., 2006. Pulverized rocks in the Mojave
section of the San Andreas Fault Zone. Earth and Planetary Science Letters 245
(3e4), 642e654.
Douglas, M., Clark, I.D., Raven, K., Bottomley, D., 2000. Groundwater mixing
dynamics at a Canadian Shield mine. Journal of Hydrology 235 (1e2), 88e103.
Eichhubl, P., DOnfro, P.S., Aydin, A., Waters, A., McCarty, D.K., 2005. Structure, petro-
physics, and diagenesis of shale entrained along a normal fault at Black Diamond
Mines, California e implications for fault seal. AAPG Bulletin 89 (9), 1113e1137.
Eichhubl, P., Davatzes, N.C., Becker, S.P., 2009. Structural and diagenetic control of
uid migration and cementation along the Moab fault, Utah. AAPG Bulletin 93
(5), 653e681.
Ellsworth, W.L., Beroza, G.C., 1995. Seismic evidence for a earthquake nucleation
phase. Science 268 (5212), 851e855.
Escartin, J., Andreani, M., Hirth, G., Evans, B., 2008. Relationships between the
microstructural evolution and the rheology of talc at elevated pressures and
temperatures. Earth and Planetary Science Letters 268 (3e4), 463e475.
Evans, J.P., 1990. Thickness displacement relationships for fault zones. Journal of
Evans, J.P., Chester, F.M., 1995. Fluiderock interaction in faults of the San-Andreas
system e inferences from San-Gabriel fault rock geochemistry and micro-
structures. Journal of Geophysical Research e Solid Earth 100 (B7),
13007e13020.
Evans, K.F., Genter, A., Sausse, J., 2005. Permeability creation and damage due to
massive uid injections into granite at 3.5 km at Soultz: 1. Borehole observa-
tions. Journal of Geophysical Research e Solid Earth 110 (B4).
Fairley, J.P., 2009. Modeling uid ow in a heterogeneous, fault-controlled hydro-
thermal system. Geouids 9 (2), 153e166.
Fairley, J.P., Hinds, J.J., 2004. Field observation of uid circulation patterns in
a normal fault system. Geophysical Research Letters 31, 1e4.
Faulkner, D.R., 2004. A model for the variation in permeability of clay-bearing fault
gouge with depth in the brittle crust. Geophysical Research Letters 31 (19), 5.
Faulkner, D.R., Lewis, A.C., Rutter, E.H., 2003. On the internal structure and
mechanics of large strike-slip fault zones: eld observations of the Carboneras
fault in southeastem Spain. Tectonophysics 367 (3e4), 235e251.
Faulkner, D.R., Mitchell, T.M., Healy, D., Heap, M.J., 2006. Slip on weak faults by the
rotation of regional stress in the fracture damage zone. Nature 444 (7121),
922e925.
Faulkner, D.R., Mitchell, T.M., Rutter, E.H., Cembrano, J., 2008. On the structure and
mechanical properties of large strike-slip faults. In: Wibberley, C.A.J., Kurz, W.,
Imber, J., Holdsworth., R.E., Collettini, C. (Eds.), Structure of Fault Zones:
Implications for Mechanical and Fluid-ow Properties. Geological Society of
London Special Publication, vol. 299, pp. 139e150.
Faulkner, D.R., Rutter, E.H., 1998. The gas permeability of clay-bearing fault gouge at
20
C. In: Jones, G., Fisher, Q., Knipe, R.J. (Eds.), Faults, Fault Sealing and Fluid
Flow in Hydrocarbon Reservoirs. Geological Society of London, Special Publi-
cation, vol. 147, pp. 147e156.
Faulkner, D.R., Rutter, E.H., 2000. Comparisons of water and argon permeability in
natural clay-bearing fault gouge under high pressure at 20 degrees C. Journal of
Geophysical Research e Solid Earth 105 (B7), 16415e16426.
Faulkner, D.R., Rutter, E.H., 2001. Can the maintenance of overpressured uids in
large strike-slip fault zones explain their apparent weakness? Geology 29 (6),
503e506.
Faulkner, D.R., Rutter, E.H., 2003. The effect of temperature, the nature of the pore
uid, and subyield differential stress on the permeability of phyllosilicate-rich
fault gouge. Journal of Geophysical Research e Solid Earth 108 (B5).
Ferrill, D.A., Winterle, J., Wittmeyer, G., Sims, D., Colton, S., Armstrong, A.,
Morris, A.P., 1999. Stressed rock strains groundwater at Yucca Mountain,
Nevada. GSA Today 9, 1e8.
Fossen, H., Schultz, R.A., Shipton, Z.K., Mair, K., 2007. Deformation bands in sand-
stone: a review. Journal of the Geological Society 164, 755e769.
Freeman, B., Boult, P., Yielding, G., Menpez, S., 2010. Using empirical geological rules
to reduce structural uncertainty in seismic interpretation of faults. Journal of
Frye, K.M., Marone, C., 2002. Effect of humidity on granular friction at room
temperature. Journal of Geophysical Research e Solid Earth 107 (B11).
Gawthorpe, R.L., Sharp, I., Underhill, J.R., Gupta, S., 1997. Linked sequence strati-
graphic and structural evolution of propagating normal faults. Geology 25 (9),
795e798.
Giger, S.B., Tenthorey, E., Cox, S.F., Gerald, J.D.F., 2007. Permeability evolution in
quartz fault gouges under hydrothermal conditions. Journal of Geophysical
Research e Solid Earth 112 (B7).
Goldsby, D.L., Tullis, T.E., 2002. Low frictional strength of quartz rocks at subseismic
slip rates. Geophysical Research Letters 29.
Gratier, J.P., Guiguet, R., Renard, F., Jenatton, L., Bernard, D., 2009. A pressure solu-
tion creep law for quartz from indentation experiments. Journal of Geophysical
Research e Solid Earth 114.
Gratier, J.P., Renard, F., Labaume, P., 1999. How pressure solution creep and frac-
turing processes interact in the upper crust to make it behave in both a brittle
and viscous manner. Journal of Structural Geology 21 (8e9), 1189e1197.
Grauls, D., Pascaud, F., Rives, T., 2002. Quantitative fault seal assessment in
hydrocarbon-compartmentalised structures using uid pressure data. In:
Koestler, A.G., Hunsdale, R. (Eds.), Hydrocarbon Seal Quantication. NPF Special
Publication, vol. 11, pp. 141e156.
Green, H.W., Marone, C., 2002. Instability of deformation. Reviews in Mineralogy
and Geochemistry 51, 181e199.
Gueydan, F., Leroy, Y.M., Jolivet, L., Agard, P., 2003. Analysis of continental mid-
crustal strain localization induced by microfracturing and reaction-softening.
Journal of Geophysical Research e Solid Earth 108 (B2).
Haines, S.H., van der Pluijm, B.A., Ikari, M.J., Saffer, D.M., Marone, C., 2009. Clay
fabric intensity in natural and articial fault gouges: implications for brittle
fault zone processes and sedimentary basin clay fabric evolution. Journal of
Geophysical Research e Solid Earth 114.
Han, R., Shimamoto, T., Hirose, T., Ree, J.H., Ando, J., 2007. Ultralow friction of
carbonate faults caused by thermal decomposition. Science 316 (5826),
878e881.
Harding, T.P., 1990. Identication of wrench faults using subsurface structural data:
criteria and pitfalls. American Association of Petroleum Geologists Bulletin 74
(10), 1590e1609.
Haney, M.M., Snieder, R., Sheiman, J., Losh, S., 2005. A moving uid pulse in a fault
zone. Nature 437, 46.
Henza, A.A., Withjack, M.O., Schlische, R.W., 2010. Normal-fault development
during two phases of non-coaxial extension: An experimental study. Journal of
Heaton, T.H., 1990. Evidence for and implications of self-healing pulses of slip in
earthquake rupture. Physics of the Earth and Planetary Interiors 64 (1), 1e20.
Hickman, S., Zoback, M., 2004. Stress orientations and magnitudes in the SAFOD
pilot hole. Geophysical Research Letters 31 (15).
Hickman, S.H., Evans, B., 1995. Kinetics of pressure solution at halite-silica interfaces
and intergranular clay lms. Journal of Geophysical Research e Solid Earth 100
(B7), 13113e13132.
Hickman, S., Sibson, R., Bruhn, R., 1995. Introduction to special section e mechanical
involvement of uids in faulting. Journal of Geophysical Research e Solid Earth
100 (B7), 12831e12840.
Hirose, T., Bystricky, M., 2007. Extreme dynamic weakening of faults during dehy-
dration by coseismic shear heating. Geophysical Research Letters 34 (14).
Holdsworth, R.E., 2004. Weak faults e rotten cores. Science 303 (5655), 181e182.
Holyoke, C.W., Tullis, J., 2006. The interaction between reaction and deformation:
an experimental study using a biotite plus plagioclase plus quartz gneiss.
Journal of Metamorphic Geology 24 (8), 743e762.
Hubbert, M.K., Rubey, W.W., 1959. Role of uid pressure in mechanics of overthrust
faulting 1. Mechanics of uid lled porous solids and its application of over-
thrust faulting. Geological Society of America Bulletin 70 (2), 115e166.
Ikari, M.J., Saffer, D.M., Marone, C., 2009. Frictional and hydrologic properties of
clay-rich fault gouge. Journal of Geophysical Research 114.
Imber, J., Holdsworth, R.E., Butler, C.A., Strachan, R.A., 2001. A reappraisal of the
Sibson-Scholz fault zone model: the nature of the frictional to viscous (brit-
tleeductile) transition along a long-lived, crustal-scale fault, Outer Hebrides,
Scotland. Tectonics 20 (5), 601e624.
Ishii, E., Funaki, H., Tokiwa, T., Ota, K. Relationship between growth mechanism of
faults and permeability variations with depth in siliceous mudstone. Journal of
Structural Geology, in press.
Janecke, S.U., Vandenburg, C.J., Blankenau, J.J., 1998. Geometry, mechanisms and
signicance of extensional folds from examples in the Rocky Mountain Basin
and Range province, USA. Journal of Structural Geology 20 (7), 841e856.
Janssen, C., Wagner, F.C., Zang, A., Dresen, G., 2001. Fracture process zone in granite:
a microstructural analysis. International Journal of Earth Sciences 90 (1), 46e59.
Jefferies, S.P., Holdsworth, R.E., Shimamoto, T., Takagi, H., Lloyd, G.E., Spiers, C.J.,
2006a. Origin and mechanical signicance of foliated cataclastic rocks in the
cores of crustal-scale faults: examples from the Median Tectonic Line, Japan.
Jefferies, S.P., Holdsworth, R.E., Wibberley, C.A.J., Shimamoto, T., Spiers, C.J.,
Niemeijer, A.R., Lloyd, G.E., 2006b. The nature and importance of phyllonite
development in crustal-scale fault cores: an example from the Median Tectonic
Line, Japan. Journal of Structural Geology 28 (2), 220e235.
Johansen, T.E.S., Fossen, H., Kluge, R., 2005. The impact of syn-faulting porosity
reduction on damage zone architecture in porous sandstone: an outcrop
example from the Moab Fault, Utah. Journal of Structural Geology 27 (8),
1469e1485.
Jourde, H., Flodin, E.A., Aydin, A., Durlofsky, L.J., Wen, X.H., 2002. Computing
permeability of fault zones in eolian sandstone from outcrop measurements.
AAPG Bulletin 86 (7), 1187e1200.
Kanamori, H., Brodsky, E.E., 2004. The physics of earthquakes. Reports on Progress in
Physics 67 (8). doi:10.1088/0034-4885/67/8/R03 pii: S0034-4885(04)25227-7.
Kanamori, H., Rivera, L., 2006. Energy partitioning during an earthquake. In:
Abercrombie, R., McGarr, A., Kanamori, H., Di Toro, G. (Eds.), Earthquakes:
Radiated Energy and the Physics of Faulting. Geophysical Monograph Series,
vol. 170, pp. 3e15.
Karner, S.L., Marone, C., Evans, B., 1997. Laboratory study of fault healing and lith-
ication in simulated fault gouge under hydrothermal conditions. Tectono-
physics 277 (1e3), 41e55.
Kim, Y.S., Peacock, D.C.P., Sanderson, D.J., 2004. Fault damage zones. Journal of
Kim, Y.S., Sanderson, D.J., 2005. The relationship between displacement and length
of faults: a review. Earth-Science Reviews 68 (3e4), 317e334.
Kirkpatrick, J.D., Shipton, Z.K., Evans, J.P., Micklethwaite, S., Lim, S.J., McKillop, P.,
2008. Strike-slip fault terminations at seismogenic depths: the structure and
kinematics of the Glacier Lakes fault, Sierra Nevada United States. Journal of
Geophysical Research e Solid Earth 113 (B4).
Kohlstedt, D.L., Evans, B., Mackwell, S.J., 1995. Strength of the lithosphere - con-
staints imposed by laboratory experiments. Journal of Geophysical Research e
Solid Earth 100 (B9), 17587e17602.
Kristensen, M.B., Childs, C.J., Korstgard, J.A., 2008. The 3D geometry of small-scale
relay zones between normal faults in soft sediments. Journal of Structural
Geology 30 (2), 257e272.
Lachenbruch, A.H., Sass, J.H., 1980. Heat-ow and energetics of the San-Andreas
fault zone. Journal of Geophysical Research 85 (Nb11), 6185e6222.
Lindsay, N.G., Murphy, F.C., Walsh, J.J., Watterson, J., 1993. Outcrop studies of shale
smear on fault surfaces. In: International Association of Sedimentology, Special
Publications, vol. 15 113e123.
Logan, J.M., Friedman, M., Higgs, N.G., Dengo, C.A. and Shimamoto, T. 1979. Exper-
imental studies of simulated gouge and their application to studies of natural
fault zones. Analysis of actual fault zones in bedrock. U.S. Geological Survey
Open File Report 79-1239, pp. 305e343.
Logan, J.M., Rauenzahn, K.A., 1987. Frictional dependence of gouge mixtures of
quartz and montmorillonite on velocity, composition and fabric. Tectonophysics
144 (1e3), 87e108.
Lohr, T., Krawczyk, C.M., Oncken, O., Tanner, D.C., 2008. Evolution of a fault surface
from 3D attribute analysis and displacement measurements. Journal of Struc-
tural Geology 30 (6), 690e700.
Losh, S., Haney, M., 2006. Episodic uid ow in an aseismic overpressured growth
fault, northern Gulf of Mexico. In: Abercrombie, R., McGarr, A., Di Toro, G.,
Kanamori, H. (Eds.), Earthquakes: Radiated Energy and the Physics of Faulting.
American Geophysical Union Geophysical Monograph Series, vol. 170, pp.
199e206.
Lunn, R.J., Willson, J.P., Shipton, Z.K., Moir, H., 2008. Simulating brittle fault growth
from linkage of preexisting structures. Journal of Geophysical Research e Solid
Earth 113 (B7).
Main, I.G., Kwon, O., Ngwenya, B.T., Elphick, S.C., 2000. Fault sealing during defor-
mation-band growth in porous sandstone. Geology 28 (12), 1131e1134.
Manzocchi, T., Walsh, J.J., Nell, P., Yielding, G., 1999. Fault transmissibility multipliers
for ow simulation models. Petroleum Geoscience 5 (1), 53e63.
Mares, V.M., Kronenberg, A.K., 1993. Experimental deformation of muscovite.
Journal of Structural Geology 15 (9e10), 1061e1075.
Mariani, E., Brodie, K.H., Rutter, E.H., 2006. Experimental deformation of muscovite
shear zones at high temperatures under hydrothermal conditions and the
strength of phyllosilicate-bearing faults in nature. Journal of Structural Geology
28 (9), 1569e1587.
Marone, C., 1998. Laboratory-derived friction laws and their application to seismic
faulting. Annual Review of Earth and Planetary Sciences 26, 643e696.
Marone, C., Kilgore, B., 1993. Scaling of the critical slip distance for seismic faulting
with shear strain in fault zones. Nature 362 (6421), 618e621.
Marone, C., Richardson, E., 2006. Do earthquakes rupture piece by piece or all
together? Science 313 (5794), 1748e1749.
Marone, C., Scholz, C.H., 1989. Particle-size distribution and microstructures within
simulated fault Gouge. Journal of Structural Geology 11 (7), 799e814.
Martel, S.J., Pollard, D.D., Segall, P., 1988. Development of simple strike-slip fault
zones, Mount Abbot Quadrangle, Sierra Nevada, California. Geological Society of
America Bulletin 100 (9), 1451e1465.
Matthai, S.K., Belayneh, M., 2004. Fluid ow partitioning between fractures and
a permeable rock matrix. Geophysical Research Letters 31 (7).
McClay, K.R., 2004. Thrust tectonics and hydrocarbonsystems. AAPGMemoir 82, 667.
McGrath, A.G., Davison, I., 1995. Damage zone geometry around fault tips. Journal of
Medeiros W.E., do Nascimento, A.F., Alves da Silva, F.C., Destro, N., Demtrio, J.G.A.
Evidence of hydraulic connectivity across deformation bands from eld
pumping tests: Two examples from Tucano Basin, NE Brazil. Journal of Struc-
tural Geology, this issue.
Micarelli, L., Benedicto, A., Wibberley, C.A.J., 2006. Structural evolution and
permeability of normal fault zones in highly porous carbonate rocks. Journal of
Miller, S.A., Collettini, C., Chiaraluce, L., Cocco, M., Barchi, M., Kaus, B.J.P., 2004.
Aftershocks driven by a high-pressure CO
2
source at depth. Nature 427 (6976),
724e727.
Mitchell, T.M., Faulkner, D.R., 2008. Experimental measurements of permeability
evolution during triaxial compression of initially intact crystalline rocks and
implications for uid ow in fault zones. Journal of Geophysical Research e
Solid Earth 113 (B11).
Mitchell, T.M., Faulkner, D.R., 2009. The nature and origin of off-fault damage
surrounding strike-slip fault zones with a wide range of displacements: A eld
study from the Atacama fault zone, northern Chile. Journal of Structural
Geology 31, 802e816.
Mizoguchi, K., Hirose, T., Shimamoto, T., Fukuyama, E., 2008. Internal structure and
permeability of the Nojima fault, southwest Japan. Journal of Structural Geology
30 (4), 513e524.
Mo, Y.F., Turner, K.T., Szlufarska, I., 2009. Friction laws at the nanoscale. Nature 457
(7233), 1116e1119.
Moir, H., Lunn, R., Shipton, Z., Kirkpatrick, J. Simulating brittle fault evolution from
networks of pre-existing structures. Journal of Structural Geology, this issue.
Moore, D.E., Lockner, D.A., 2004. Crystallographic controls on the frictional behavior
of dry and water-saturated sheet structure minerals. Journal of Geophysical
Moore, D.E., Lockner, D.A., 2008. Talc friction in the temperature range 25 degrees-
400 degrees C: relevance for fault-zone weakening. Tectonophysics 449 (1e4),
120e132.
Moore, D.E., Lockner, D.A., Tanaka, H., Iwata, K., 2004. The coefcient of friction of
Chrysotile gouge at seismogenic depths. International Geology Review 46 (5),
385e398.
Moore, D.E., Rymer, M.J., 2007. Talc-bearing serpentinite and the creeping section of
the San Andreas fault. Nature 448 (7155), 795e797.
Morrow, C.A., Lockner, D.A., 1994. Permeability differences between surface-derived
and deep drillhole core samples. Geophysical Research Letters 21 (19),
2151e2154.
Morrow, C.A., Moore, D.E., Lockner, D.A., 2001. Permeability reduction in granite
under hydrothermal conditions. Journal of Geophysical Research e Solid Earth
106 (B12), 30551e30560.
Morrow, C.A., Radney, B., Byerlee, J.D., 1992. Frictional strength and the effective
pressure law of montmorillonite and illite clays. In: Evans, B., Wong, T.-F. (Eds.),
Fault Mechanics and Transport Properties of Rocks. Academic Press, pp. 69e88.
Nakatani, M., 2001. Conceptual and physical clarication of rate and state friction:
frictional sliding as a thermally activated rheology. Journal of Geophysical
Research e Solid Earth 106 (B7), 13347e13380.
Nakatani, M., Scholz, C.H., 2004a. Frictional healing of quartz gouge under hydro-
thermal conditions: 1. Experimental evidence for solution transfer healing
mechanism. Journal of Geophysical Research e Solid Earth 109 (B7).
Nakatani, M., Scholz, C.H., 2004b. Frictional healing of quartz gouge under hydro-
thermal conditions: 2. Quantitative interpretation with a physical model.
Nemser, E.S., Cowan, D.S., 2009. Downdip segmentation of strike-slip fault zones in
the brittle crust. Geology 37 (5), 419e422.
Noda, H., 2008. Frictional constitutive law at intermediate slip rates accounting for
ash heating and thermally activated slip process. Journal of Geophysical
Noda, H., Dunham, E.M., Rice, J.R., 2009. Earthquake ruptures with thermal weak-
ening and the operation of major faults at low overall stress levels. Journal of
Geophysical Research 114.
Noir, J., Jacques, E., Bekri, S., Adler, P.M., Tapponnier, P., King, G.C.P., 1997. Fluid ow
triggered migration of events in the 1989 Dobi earthquake sequence of Central
Afar. Geophysical Research Letters 24 (18), 2335e2338.
Numelin, T., Marone, C., Kirby, E., 2007. Frictional properties of natural fault gouge
from a low-angle normal fault, Panamint Valley, California. Tectonics 26 (2).
OHara, K., 2007. Reaction weakening and emplacement of crystalline thrusts:
diffusion control on reaction rate and strain rate. Journal of Structural Geology
29 (8), 1301e1314.
Oda, M., Takemura, T., Aoki, T., 2002. Damage growth and permeability change in
triaxial compressiontests of Inada granite. Mechanics of Materials 34 (6), 313e331.
Odling, N.E., Harris, S.D., Knipe, R., 2004. Permeability scaling properties of fault
damage zones in siliclastic rocks. Journal of Structural Geology 26 (9),
1727e1747.
Olsen, M.P., Scholz, C.H., Leger, A., 1998. Healing and sealing of a simulated fault
gouge under hydrothermal conditions: implications for fault healing. Journal of
Peacock, D.C.P., Sanderson, D.J., 1991. Displacements, segment linkage and relay
ramps in normal fault zones. Journal of Structural Geology 13 (6), 721.
Price, R.A., 1988. The mechanical paradox of large overthrusts. Geological Society of
America Bulletin 100 (12), 1898e1908.
Rawling, G.C., Goodwin, L.B., Wilson, J.L., 2001. Internal architecture, permeability
structure, and hydrologic signicance of contrasting fault-zone types. Geology
29 (1), 43e46.
Reches, Z., Dewers, T.A., 2005. Gouge formation by dynamic pulverization during
earthquake rupture. Earth and Planetary Science Letters 235 (1e2), 361e374.
Reches, Z., Lockner, D.A., 1994. Nucleation and growth of faults in brittle rocks.
Journal of Geophysical Research 99 (B9), 18159e18173.
Reinen, L.A., 2000. Slip styles in a spring-slider model with a laboratory-derived
constitutive law for serpentinite. Geophysical Research Letters 27 (14),
2037e2040.
Rempel, A.W., Rice, J.R., 2006. Thermal pressurization and onset of melting in fault
zones. Journal of Geophysical Research e Solid Earth 111 (B9).
Renard, F., Gratier, J.P., Jamtveit, B., 2000. Kinetics of crack-sealing, intergranular
pressure solution, and compaction around active faults. Journal of Structural
Geology 22 (10), 1395e1407.
Revil, A., Grauls, D., Brevart, O., 2002. Mechanical compaction of sand/clay mixtures.
Rice, J.R., 1992. Fault stress states, pore pressure distributions, and the weakness of
the San Andreas fault. In: Evans, B., Wong, T.-F. (Eds.), Fault Mechanics and
Transport Properties in Rocks. Academic Press, p. 28.
Rice, J.R., 2006. Heating and weakening of faults during earthquake slip. Journal of
Rice, J.R., Lapusta, N., Ranjith, K., 2001. Rate and state dependent friction and the
stability of sliding between elastically deformable solids. Journal of the
Mechanics and Physics of Solids 49 (9), 1865e1898.
Roberts, G.P., Houghton, S.L., Underwood, C., Papanikolaou, I., Cowie, P.A., van
Calsteren, P., Wigley, T., Cooper, F.J., McArthur, J.M., 2009. Localization of
Quaternary slip rates in an active rift in 10(5) years: an example from central
Greece constrained by U-234eTh-230 coral dates from uplifted paleoshore-
lines. Journal of Geophysical Research e Solid Earth 114.
Rowland, J.V., Sibson, R.H., 2004. Structural controls on hydrothermal ow in
a segmented rift system, Taupo Volcanic Zone, New Zealand. Geouids 4 (4),
259e283.
Rudnicki, J.W., Rice, J.R., 2006. Effective normal stress alteration due to pore pres-
sure changes induced by dynamic slip propagation on a plane between
dissimilar materials. Journal of Geophysical Research e Solid Earth 111 (B10).
Rutter, E.H., Brodie, K.H., 1995. Mechanistic interactions between deformation and
metamorphism. Geological Journal 30 (3e4), 227e240.
Rutter, E.H., Maddock, R.H., Hall, S.H., White, S.H., 1986. Comparative microstruc-
tures of natural and experimentally produced clay-bearing fault gouges. Pure
and Applied Geophysics 124 (1e2), 3e30.
Rutter, E.H., Mainprice, D.H., 1979. On the possibility of slow fault slip controlled by
a diffusive mass transfer process. Gerlands Beitrage zur Geophysik 88, 154e162.
Saffer, D.M., Frye, K.M., Marone, C., Mair, K., 2001. Laboratory results indicating
complex and potentially unstable frictional behavior of smectite clay.
Geophysical Research Letters 28 (12), 2297e2300.
Saffer, D.M., Marone, C., 2003. Comparison of smectite- and illite-rich gouge frictional
properties: application to the updip limit of the seismogenic zone along
subduction megathrusts. Earth and Planetary Science Letters 215 (1-2), 219e235.
Saillet, E., Wibberley, C. Evolution of cataclastic faulting in high porosity sandstone,
Bassin du Sud-Est, Provence, France. Journal of Structural Geology, this issue.
Sammis, C., King, G., Biegel, R., 1987. The kinematics of gouge deformation. Pure and
Applied Geophysics 125 (5), 777e812.
Sammis, C.G., Ben-Zion, Y., 2008. Mechanics of grain-size reduction in fault zones.
Savage, H.M., Brodsky, E.E. Collateral damage: capturing slip delocalization in
fracture proles. Journal of Geophysical Research, submitted for publication.
Savage, H.M., Cooke, M.L., 2010. Unlocking the effects of friction on fault damage
zone models. Journal of Structural Geology 32 (11), 1732e1741.
Schlische, R.W., 1995. Geometry and origin of fault-related folds in extensional
settings. AAPG Bulletin e American Association of Petroleum Geologists 79 (11),
1661e1678.
Schlische, R.W., Withjack, M.O., Eisenstadt, G., 2002. An experimental study of the
secondary deformation produced by oblique-slip normal faulting. American
Association of Petroleum Geologists Bulletin 86 (5), 885e906.
Schmatz, J., Vrolijk, P., Urai, J., 2010. Clay smear in normal faults - the effect of
multilayers and clay cementation in water-saturated model experiments.
Scholz, C.H., 1987. Wear and Gouge formation in brittle faulting. Geology 15 (6),
493e495.
Scholz, C.H., 1988. The critical slip distance for seismic faulting. Nature 336 (6201),
761e763.
Scholz, C.H., 1998. Earthquakes and friction laws. Nature 391 (6662), 37e42.
Scholz, C.H., 2002. The Mechanics of Earthquakes and Faulting. Cambridge
University Press, Cambridge.
Scholz, C.H., 2006. The strength of the San Andreas fault: a critical analysis.
Earthquakes: Radiated Energy and the Physics of Faulting 170, 301e311.
Schulz, S.E., Evans, J.P., 2000. Mesoscopic structure of the Punchbowl Fault,
Southern California and the geologic and geophysical structure of active strike-
slip faults. Journal of Structural Geology 22 (7), 913e930.
Screaton, E.J., Wuthrich, D.R., Dreiss, S.J., 1990. Permeabilities, uid pressures, and
ow rates in the Barbados ridge complex. Journal of Geophysical Research e
Solid Earth and Planets 95 (B6), 8997e9007.
Scruggs, V.J., Tullis, T.E., 1998. Correlation between velocity dependence of friction
and strain localization in large displacement experiments on feldspar, musco-
vite and biotite gouge. Tectonophysics 295 (1e2), 15e40.
Segall, P., Rice, J.R., 2006. Does shear heating of pore uid contribute to earthquake
nucleation? Journal of Geophysical Research e Solid Earth 111 (B9).
Shapiro, S.A., Huenges, E., Borm, G., 1997. Estimating the crust permeability from
uid-injection-induced seismic emission at the KTB site. Geophysical Journal
International 131 (2), F15eF18.
Shaw, J.H., Connors, C., Suppe, J., 2005. Seismic interpretation of contractional fault-
related folds: an AAPG seismic atlas. American Association of Petroleum
Geologists Studies in Geology 53, 157.
Shipton, Z.K., Cowie, P.A., 2001. Damage zone and slip-surface evolution over mm to
km scales in high-porosity Navajo sandstone, Utah. Journal of Structural
Geology 23 (12), 1825e1844.
Shipton, Z.K., Cowie, P.A., 2003. A conceptual model for the origin of fault damage
zone structures in high-porosity sandstone. Journal of Structural Geology 25 (8),
1343e1345.
Shipton, Z.K., Evans, J.P., Robeson, K.R., Forster, C.B., Snelgrove, S., 2002. Structural
heterogeneity and permeability in faulted eolian sandstone: implications for
subsurface modeling of faults. AAPG Bulletin 86 (5), 863e883.
Shipton, Z.K., Evans, J.P., Thompson, L.B., 2005. The geometry and thickness of
deformation band fault core, and its inuence on sealing characteristics of
deformation band fault zones. American Association of Petroleum Geologists
Memoir 85, 181e195.
Shipton, Z.K., Soden, A.M., Kirkpatrick, J.D., Bright, A.M., Lunn, R.J., 2006. How thick
is a fault? Fault displacement-thickness scaling revisited. Earthquakes: Radiated
Energy and the Physics of Faulting 170, 193e198.
Sibson, R.H., 1990. Conditions for fault-valve behaviour. In: Geological Society of
London, Special Publication, vol. 54 15e28.
Sibson, R.H., 2001. Seismogenic framework for hydrothermal transport and ore
deposition. Reviews in Economic Geology 14, 25e50.
Sibson, R.H., 2009. Rupturing in overpressured crust during compressional inver-
sion-the case from NE Honshu, Japan. Tectonophysics 473 (3e4), 404e416.
Sibson, R.H., Xie, G.Y., 1998. Dip range for intracontinental reverse fault ruptures:
truth not stranger than friction? Bulletin of the Seismological Society of
America 88 (4), 1014e1022.
Simpson, G., Gueguen, Y., Schneider, F., 2001. Permeability enhancement due to
microcrack dilatancy in the damage regime. Journal of Geophysical Research e
Solid Earth 106 (B3), 3999e4016.
Smith, S.A.F., Faulkner, D.R., 2010. Laboratory measurements of the frictional
properties of a natural low-angle normal fault: The Zuccale fault, Elba Island,
Italy. Journal of Geophysical Research 115, B02407.
Smith, S.A.F., Holdsworth, R.E., Collettini, C., Imber, J., 2007. Using footwall struc-
tures to constrain the evolution of low-angle normal faults. Journal of the
Geological Society 164, 1187e1191.
Soliva, R., Benedicto, A., 2005. Geometry, scaling relations and spacing of vertically
restricted normal faults. Journal of Structural Geology 27 (2), 317e325.
Soliva, R., Maertan, F., Petit, J.-P., Auzias, V. Fault static friction and fracture orien-
tation in extensional relays; insight from eld data, photoelasticity and 3D
numerical modeling. Journal of Structural Geology, this issue.
Solum, J.G., van der Pluijm, B.A. Quantication of fabrics in clay gouge from the
Carboneras fault, Spain and implications for fault behavior. Tectonophysics 475
(3e4), 554e562.
Sperrevik, S., Gillespie, P.A., Fisher, Q., Halvorsen, T., Knipe, R.J., 2002. Empirical
estimation of fault rock properties. In: Koestler, A.G., Hunsdale, R. (Eds.),
Hydrocarbon Seal Quantication. NPF Special Publications, vol. 11, pp. 109e125.
Spudich, P., Guatteri, M., 2004. The effect of bandwidth limitations on the inference
of earthquake slip-weakening distance from seismograms. Bulletin of the
Seismological Society of America 94 (6), 2028e2036.
Streit, J.E., Hillis, R.R., 2004. Estimating fault stability and sustainable uid pressures
for underground storage of CO2
in porous rock. Energy 29 (9e10), 1445e1456.
Suppe, J., 1983. Geometry and kinematics of fault-bend folding. American Journal of
Science 283 (7), 684e721.
Swanson, M.T., 2005. Geometry and kinematics of adhesive wear in brittle strike-
slip fault zones. Journal of Structural Geology 27 (5), 871e887.
Szlufarska, I., Chandross, M., Carpick, R.W., 2008. Recent advances in single-asperity
nanotribology. Journal of Physics D e Applied Physics 41 (12).
Tadokoro, K., Ando, M., Nishigami, K., 2000. Induced earthquakes accompanying the
water injection experiment at the Nojima fault zone, Japan: seismicity and its
migration. Journal of Geophysical Research e Solid Earth 105 (B3), 6089e6104.
Takahashi, M., Mizoguchi, K., Kitamura, K., Masuda, K., 2007. Effects of clay content
on the frictional strength and uid transport property of faults. Journal of
Talwani, P., Cobb, J.S., Schaeffer, M.F., 1999. In situ measurements of hydraulic
properties of a shear zone in northwestern South Carolina. Journal of
Talwani, P., Chen, L., Gahalaut, K., 2007. Seismogenic permeability, k(S). Journal of
Tembe, S., Lockner, D.A., Solum, J.G., Morrow, C.A., Wong, T.F., Moore, D.E., 2006.
Frictional strength of cuttings and core from SAFOD drillhole phases 1 and 2.
Geophysical Research Letters 33 (23).
Tenthorey, E., Scholz, C.H., Aharonov, E., Leger, A., 1998. Precipitation sealing and
diagenesis e 1. Experimental results. Journal of Geophysical Research e Solid
Earth 103 (B10), 23951e23967.
Tenthorey, E., Cox, S.F., Todd, H.F., 2003. Evolution of strength recovery and
permeability during uid-rock reaction in experimental fault zones. Earth and
Planetary Science Letters 206 (1e2), 161e172.
Tindall, S.E., Davis, G.H., 1999. Monocline development by oblique-slip fault-prop-
agation folding: the East Kaibab monocline, Colorado Plateau, Utah. Journal of
Tinti, E., Cocco, M., Fukuyama, E., Piatanesi, A., 2009. Dependence of slip weakening
distance (D-c) on nal slip during dynamic rupture of earthquakes. Geophysical
Journal International 177 (3), 1205e1220.
Tinti, E., Spudich, P., Cocco, M., 2005. Earthquake fracture energy inferred from
kinematic rupture models on extended faults. Journal of Geophysical Research
e Solid Earth 110 (B12).
Townend, J., Zoback, M.D., 2000. How faulting keeps the crust strong. Geology 28
(5), 399e402.
Tsutsumi, A., Shimamoto, T., 1997. High-velocity frictional properties of gabbro.
Geophysical Research Letters 24 (6), 699e702.
Tsutsumi, A., Nishino, S., Mizoguchi, K., Hirose, T., Uehara, S., Sato, K., Tanikawa, W.,
Shimamoto, T., 2004. Principal fault zone width and permeability of the active
Neodani fault, Nobi fault system, Southwest Japan. Tectonophysics 379 (1e4),
93e108.
Uehara, S., Shimamoto, T., 2004. Gas permeability evolution of cataclasite and fault
gouge in triaxial compression and implications for changes in fault-zone
permeability structure through the earthquake cycle. Tectonophysics 378 (3e4),
183e195.
van der Zee, W., Urai, J.L., 2005. Processes of normal fault evolution in a siliciclastic
sequence: a case study from Miri, Sarawak, Malaysia. Journal of Structural
Geology 27 (12), 2281e2300.
Van der Zee, W., Wibberley, C.A.J., Urai, J.L., 2008. The inuence of layering and pre-
existing joints on the development of internal structure in normal fault zones:
the Lodve basin, France. In: Wibberley, C.A.J., Kurz, W., Imber, J.,
Holdsworth, R.E., Collettini, C. (Eds.), The Internal Structure of Fault Zones:
Implications for Mechanical and Fluid Flow Properties. Geological Society of
London, Special Publication, vol. 299, pp. 57e74.
van Diggelen, E. W., De Bresser, J.H., Peach, C.J., Spiers, C.J. High shear strain
behaviour of synthetic muscovite fault gouges under hydrothermal conditions.
Journal of Structural Geology, this issue.
Vermilye, J.M., Scholz, C.H., 1998. The process zone: a microstructural view of
fault growth. Journal of Geophysical Research e Solid Earth 103 (B6),
12223e12237.
Vidale, J.E., Li, Y.G., 2003. Damage to the shallow Landers fault from the nearby
Hector Mine earthquake. Nature 421 (6922), 524e526.
Walker, J.P.F., Roberts, G.P., Cowie, P.A., Papanikolaou, I.D., Sammonds, P.R.,
Michetti, A.M., Phillips, R.J., 2009. Horizontal strain-rates and throw-rates
across breached relay zones, central Italy: implications for the preservation of
throw decits at points of normal fault linkage. Journal of Structural Geology 31
(10), 1145e1160.
Walsh, J.J., Watterson, J., 1988. Analysis of the relationship between displacements
and dimensions of faults. Journal of Structural Geology 10 (3), 239e247.
Walsh, J.J., Nicol, A., Childs, C., 2002. An alternative model for the growth of faults.
Weber, K.J., Mandl, G., Pilaar, W.F., Lehner, F., Precious, R.G. 1978. The role of faults in
hydrocarbon migration and trapping in Nigerian growth fault structures. In:
10th Annual Offshore Technology Conference Proceedings 4, pp. 2643e2653.
Whittaker, A.C., Attal, M., Cowie, P.A., Tucker, G.E., Roberts, G., 2008. Decoding
temporal and spatial patterns of fault uplift using transient river long proles.
Geomorphology 100 (3e4), 506e526.
Wibberley, C.A.J., 2005. Initiation of basement thrust detachments by fault-zone
reaction weakening. In: Bruhn, D., Burlini, L. (Eds.), High Strain Zones: Structure
and Physical Properties, vol. 245. Geological Society, London, Special Publica-
tions, pp. 347e372.
Wibberley, C.A.J., McCaig, A.M., 2000. Quantifying orthoclase and albite muscovi-
tisation sequences in fault zones. Chemical Geology 165 (3e4), 181e196.
Wibberley, C.A.J., Shimamoto, T., 2003. Internal structure and permeability of major
strike-slip fault zones: the Median Tectonic Line in Mie Prefecture, Southwest
Japan. Journal of Structural Geology 25 (1), 59e78.
Wibberley, C.A.J., Shimamoto, T., 2005. Earthquake slip weakening and asperities
explained by thermal pressurization. Nature 436 (7051), 689e692.
Wibberley, C.A.J., Yielding, G., Di Toro, G., 2008. Recent advances in the under-
standing of fault zone internal structure; a review. In: Wibberley, C.A.J.,
Kurz, W., Imber, J., Holdsworth, R.E., Collettini, C. (Eds.), Structure of Fault
Zones: Implications for Mechanical and Fluid-ow Properties. Geological
Society of London Special Publication, vol. 299, pp. 5e33.
Wilkerson, M.S., Fischer, M.P., Apotria, T., 2002. Fault-related folds: the transition
from 2-D to 3-D e preface. Journal of Structural Geology 24 (4), 591e592.
Wilkins, S.J., Gross, M.R., 2002. Normal fault growth in layered rocks at Split
Mountain, Utah: inuence of mechanical stratigraphy on dip linkage, fault
restriction and fault scaling. Journal of Structural Geology 24 (9), 1413e1429.
Wilkins, S.J., Gross, M.R., Wacker, M., Eyal, Y., Engelder, T., 2001. Faulted joints:
kinematics, displacement-length scaling relations and criteria for their identi-
cation. Journal of Structural Geology 23 (2e3), 315e327.
Wilson, J.E., Chester, J.S., Chester, F.M., 2003. Microfracture analysis of fault growth
and wear processes, Punchbowl Fault, San Andreas System, California. Journal
of Structural Geology 25 (11), 1855e1873.
Wintsch, R.P., Christoffersen, R., Kronenberg, A.K., 1995. Fluid-rock reaction weak-
ening of fault zones. Journal of Geophysical Research e Solid Earth 100 (B7),
13021e13032.
Wiprut, D., Zoback, M., 2002. Fault reactivation, leakage potential, and hydro-
carbon column heights in the northern North Sea. In: Koestler, A.G.,
Hunsdale, R. (Eds.), Hydrocarbon Seal Quantication. NPF Special Publication,
vol. 11, pp. 203e219.
Withjack, M.O., Schlische, R.W., Olsen, P.E., 2002. Rift-basin structure and its
inuence on sedimentary systems. In: Society for Sedimentary Geology Special
Publication, vol. 73 57e81.
Yamashita, T., Suzuki, T., 2009. Quasi-static fault slip on an interface between
poroelastic media with different hydraulic diffusivity: a generation mechanism
of afterslip. Journal of Geophysical Research e Solid Earth 114.
Yasuhara, H., Marone, C., Elsworth, D., 2005. Fault zone restrengthening and fric-
tional healing: the role of pressure solution. Journal of Geophysical Research e
Solid Earth 110 (B6).
Yielding, G., Freeman, B., Needham, D.T., 1997. Quantitative fault seal prediction.
AAPG Bulletin e American Association of Petroleum Geologists 81 (6),
897e917.
Zhang, S.Q., Tullis, T.E., 1998. The effect of fault slip on permeability and perme-
ability anisotropy in quartz gouge. Tectonophysics 295 (1e2), 41e52.
Zhang, S.Q., Tullis, T.E., Scruggs, V.J., 1999. Permeability anisotropy and pressure
dependency of permeability in experimentally sheared gouge materials. Journal
of Structural Geology 21 (7), 795e806.
Zheng, G., Rice, J.R., 1998. Conditions under which velocity-weakening friction
allows a self-healing versus a cracklike mode of rupture. Bulletin of the Seis-
mological Society of America 88 (6), 1466e1483.
Zoback, M., Hickman, S., Ellsworth, W., 2010. Scientic drilling into the San Andreas
fault zone. Eos. Transactions American Geophysical Union 91 (22), 197e199.
Internal structure, fault rocks, and inferences regarding deformation, uid ow,
and mineralization in the seismogenic Stillwater normal fault, Dixie
Valley, Nevada
Jonathan Saul Caine
a,
*
, Ronald L. Bruhn
b
, Craig B. Forster
b
a
U.S. Geological Survey, P.O. Box 25046, MS 964, Denver, CO 80225, USA
b
Department of Geology and Geophysics, University of Utah, 115 South 1460 East, Salt Lake City, UT 84112, USA
Article history:
Received 15 January 2009
28 January 2010
Accepted 9 March 2010
Available online 17 March 2010
This work is dedicated to the memory of
Craig B. Forster who died in a tragic accident
on December 28, 2008. It is a reection of
his exceptional enthusiasm and dedication
to bringing students and colleagues
together in the pursuit of collaborative
scientic research.
Keywords:
Fault zone
Seismicity
Fluid ow
Hydrothermal
Breccia textures
Silicication
a b s t r a c t
Outcrop mapping and fault-rock characterization of the Stillwater normal fault zone in Dixie Valley,
Nevada are used to document and interpret ancient hydrothermal uid ow and its possible relationship
to seismic deformation. The fault zone is composed of distinct structural and hydrogeological compo-
nents. Previous work on the fault rocks is extended to the map scale where a distinctive fault core shows
a spectrum of different fault-related breccias. These include predominantly clast-supported breccias with
angular clasts that are cut by zones containing breccias with rounded clasts that are also clast supported.
These are further cut by breccias that are predominantly matrix supported with angular and rounded
clasts. The fault-core breccias are surrounded by a heterogeneously fractured damage zone. Breccias are
bounded between major, silicied slip surfaces, forming large pod-like structures, systematically
oriented with long axes parallel to slip. Matrix-supported breccias have multiply brecciated, angular and
rounded clasts revealing episodic deformation and uid ow. These breccias have a quartz-rich matrix
with microcrystalline anhedral, equant, and pervasively conformable mosaic texture. The breccia pods
are interpreted to have formed by decompression boiling and rapid precipitation of hydrothermal uids
whose ow was induced by coseismic, hybrid dilatant-shear deformation and hydraulic connection to
a geothermal reservoir. The addition of hydrothermal silica cement localized in the core at the map scale
causes fault-zone widening, local sealing, and mechanical heterogeneities that impact the evolution of
the fault zone throughout the seismic cycle.
Published by Elsevier Ltd.
1. Introduction
The presence and ow of uids in the upper crust has a major
impact on the mechanics of faulting (Hubbert and Rubey, 1959; Nur
and Booker, 1972; Sibson, 1977, 1981, 1990, 1996; Power and Tullis,
1989; Bruhn et al., 1990, 1994; Parry and Bruhn, 1990; Scholz, 2002;
Chester et al., 1993; Rice, 1992; Byerlee, 1993; Keller and Loaiciga,
1993; Evans and Chester, 1995; Caine et al., 1996; Miller et al.,
1996; Seront et al., 1998; Tanaka et al., 2001; Wibberley,
2002;Faulkner et al., 2006; Lockner et al., 2009). Fluid ow and
its interactions with heterogeneous permeability structures in
a fault zone can control the magnitude of local principal stresses
(Nemcok et al., 2002). This, in turn, affects local uid-pressure
gradients, mechanical failure, propagation of pressure transients,
uid inltration into and out of a fault zone via fault-valve mech-
anisms (e.g., Sibson, 1992), and fault-zone sealing and healing (e.g.,
Faulkner et al., 2008). Fluid ow in fault zones can control the
location, emplacement, and evolution of economic mineral
deposits and geothermal systems (e.g., Newhouse, 1942; Cox et al.,
2001; Sibson, 2001; Micklethwaite, 2009), and may also impact the
locations and magnitudes of foreshock, earthquake and aftershock
distributions (Miller et al., 2004). Yet fault zones are heterogeneous
geological and hydrological structures that commonly are not well
exposed. Even in well-exposed fault zones direct links between
internal structure, fault rocks, and mineral assemblages that
uniquely indicate a seismogenic origin are uncommon (cf. Sibson,
1986b; Cowan, 1999; Ujiie et al., 2007; Woodcock et al., 2007;
Smith et al., 2008). Thus, the study of exposed, seismogenic fault
zones that may record uid ow-related processes associated with
earthquakes remains important for understanding the mechanics
of faulting.
E-mail address: jscaine@usgs.gov (J.S. Caine).
0191-8141/$ e see front matter Published by Elsevier Ltd.
doi:10.1016/j.jsg.2010.03.004
Fault zones are commonly composed of distinct, three-dimen-
sional, mappable components that include a fault core and damage
zone within relatively undeformed protolith (Chester and Logan,
1986; Smith et al., 1990; Forster et al., 1991; Caine et al., 1996).
Most of the strain is accommodated in a fault core indicated by
rocks such as fault-related breccias and clay-rich gouge. Fault zones
can also have multiple core zones interspersed with pods of
heterogeneously deformed host rock (cf. Faulkner et al., 2006). A
damage zone is the mappable network of subsidiary structures that
surrounds a fault core or fault-core zone and is related to the
nucleation, evolution, and growth of the fault zone (Chester and
Logan, 1986; Scholz, 2002; Caine et al., 1996; Knipe et al., 1998).
Damage-zone fracture networks commonly have orientations
mechanically related to the master fault and are of higher intensity
than found in the protolith (e.g., Caine and Forster, 1999). The fault
core and damage zone are surrounded by the protolith where fault-
related structures are generally absent.
The bulk permeability structure and strength of a fault zone are
controlled by preexisting and newly developed structures, the
regional and local stress state, fault-zone geometry, and changes
in lithology resulting from the coupling of mechanical, thermal,
uid ow, and reactive geochemical processes. For example, the
creation of new hydraulically contrasting lithologies and struc-
tures, such as clay-rich cataclasites and complex fracture
networks, has been documented to result from as well as impact
uid ow in diverse brittle fault-zone settings (Sibson, 1986a;
Chester and Logan, 1986; Scholz, 2002; Bruhn et al., 1994;
Antonellini and Aydin, 1994; Goddard and Evans, 1995; Caine
et al., 1996; Faulkner et al., 2006). These fault-related physical
attributes in the upper crust create hydraulic and mechanical
heterogeneity and anisotropy that have a signicant impact on
rupture and the arrest of failure (Parry et al., 1991; Byerlee, 1993;
Miller et al., 1996; Seront et al., 1998) as well as growth and
widening of a fault zone.
Previous theoretical research in earthquake mechanics has
focused on the role of uid circulation and hydrothermal alteration
associated with faulting processes (Sibson, 1981; Bruhn et al., 1994;
Parry et al., 1991; Rice, 1992; Byerlee, 1993; Scholz, 2002; Unsworth
et al., 1997). Although there have been studies of exhumed and
well-exposed seismogenic fault zones (e.g., Hancock and Barka,
1987; Ghisetti et al., 2001), details regarding the physical
pathways along which uid ow occurs, and the characteristics of
structures and rock types that result fromcoupled deformation and
uid ow remain sparsely documented.
This paper describes eld observations fromthe Mirrors locality
of the Stillwater Fault Zone (SFZ) in Dixie Valley, Nevada (Fig. 1).
This is an area of geological interest due to exposures of exhumed
portions of the footwall of this normal fault with a record of historic
earthquakes and surface ruptures associated with the fault. There
are also epithermal gold deposits, and a productive geothermal
reservoir hydraulically connected to the fault zone. Outcrop
mapping, hand-sample and thin-section fault-rock studies are used
to extend previous work and (1) document the internal structure
and geometry of part of the fault zone, (2) infer the paleo-perme-
ability structure, (3) document the textural attributes, composition,
and spatial and temporal distribution of fault rocks, and (4) infer
deformation-related uid owprocesses associated with seismicity
and growth of the fault zone.
2. Geologic setting and previous work
The SFZ is historically active and capable of generating magni-
tude (M) > 6 earthquakes. Ground surface rupture associated with
the 1954 M 6.8 earthquake was 30e40 km long (Caskey et al.,
1996). The SFZ, also called the Dixie Valley Fault, is the eastern
range-bounding fault between the Stillwater Mountains and Dixie
Valley graben (Fig. 1). Fault segments that range from several
kilometers to a few tens of kilometers in length form the SFZ. The
SFZ is one segment in a 300 km long belt of normal and normal-
oblique slip faults (Wallace and Whitney, 1984; Caskey et al., 1996).
The Stillwater Range is composed of Mesozoic metasedimentary,
plutonic, and volcanic rocks. Igneous rocks include the Jurassic
gabbroic Humboldt igneous complex, Cretaceous granites, a multi-
phase Oligocene graniteegranodioriteequartz monzonite complex,
and various extrusive rocks of Oligocene and Miocene age (Page,
1965; Speed and Armstrong, 1971; Wilden and Speed, 1974;
Speed, 1976).
The SFZ has been seismically active since Oligocene to early
Miocene times (Parry et al., 1991; Bruhn et al., 1994; Caskey et al.,
1996; Seront et al., 1998). The Holocene fault scarps that cut the
basin ll along the eastern base of the Stillwater Range are ground
surface ruptures probably formed during major earthquakes within
Surface trace of Stillwater
fault zone, bar and ball on
hanging wall.
Tertiary granitic rocks
Tertiary volcanic rocks
Jurassic granodiortic to
gabbroic rocks
Jr/Tr sedimentary rocks
Box
Canyons
Oxbow
Geothermal
Plant
The Mirrors
0 10 20 30 40
Km
N
Nevada
Stillwater
Range
40 N Latitude
D
i
x
i
e
V
a
l
l
e
y
118 W Longitude
Explanation
S
t
i
l
l
w
a
t
e
r
R
a
n
g
e
Fig. 1. Location map for the Stillwater Fault Zone in Dixie Valley, Nevada showing the primary study area of the Mirrors Locality. Modied from Parry and Bruhn (1990).
J.S. Caine et al. / Journal of Structural Geology 32 (2010) 1576e1589 1577
the last 12,000 years (Wallace and Whitney, 1984; Caskey et al.,
1996). The eastern slope of the Stillwater Range is the exhumed
footwall of the fault zone. An along-strike differential surface-uplift
history of the Stillwater Range is suggested by rocks exhumed from
a minimumdepth of 6 km(Parry et al., 1991) in the southern part of
the range and as little as 2 km at temperatures less than 270

C in
the central part of the range (Power and Tullis, 1989).
The rocks exposed along the SFZ and at the Mirrors locality in
the central part of the Stillwater Range have undergone varying
degrees of hydrothermal alteration that includes the following
mineral assemblages and paragenetic sequence: biotite K-feld-
spar, chlorite sphene epidote magnetite; mixed-layer
chloriteesmectite smectite goethite; quartz illite
calcite ferroan dolomite calcite barite; quartz kaolinite;
and quartz calcite (Parry et al., 1991; Lutz et al., 1997). The
assemblages in this paragenetic sequence were interpreted to
preserve a record of exhumation, with the earlier assemblages also
representing deeper crustal conditions. Power and Tullis (1989)
present a detailed analysis of slip surfaces at the Mirrors and
make estimates of deformation temperatures based on the ther-
modynamic stability of the quartz kaolinite mineral assemblage
in distinctive fault-related breccias and the tectonostratigraphic
position of this locality with respect to overlying volcanic rocks.
They observed crystallographic preferred orientation of quartz c-
axes in samples of the slickensided surfaces and their preferred
interpretation of this alignment involves pressure solution and
growth/dissolution rate anisotropy during continuous, low
temperature, low strain-rate deformation. Additionally, Power and
Tullis (1989) interpreted multiple generations of fault-related
brecciation and thus interpreted alternating periods of continuous
deformation during the interseismic phases of the earthquake cycle
followed by discontinuous deformation during coseismic phases of
the earthquake cycle.
Hydrothermal alteration in the Oligocene granitic complex
along the base of the Stillwater Range reects the interplay of
faulting, uid circulation, and exhumation associated with the SFZ
(Parry et al., 1991; Lutz et al., 1997). At some localities this alteration
may reect fault-related uid interactions with an ancient, and
presently producing, geothermal reservoir as well as a number of
surcial expressions of the geothermal system such as hot springs,
sinter mounds, and fumaroles (e.g., Lutz et al., 2002). Barton et al.
(1997) and Hickman et al. (1997, 1998, 2000) studied borehole log
and in situ stress data near the Oxbow geothermal plant, several
kilometers northeast of the Mirrors locality (Fig. 1). They found that
the hydraulically conductive fractures in the geothermal reservoir
were also critically stressed for shear failure in the regional stress
eld. Finally, structurally favorable sites and fault-related fracture
networks also create loci for hydrothermal uid ow that produced
economic epithermal gold deposits (e.g., Vikre, 1994).
3. The Mirrors map area exposure, overview, and methods
Each fault-zone component is well exposed in the bedrock at the
Mirrors locality (Figs. 2e4). This footwall remnant of the SFZ
extends approximately 250 m vertically upward from the base of
the Stillwater Range. Although the hanging wall is composed of
Quaternary basin ll in depositional contact with the crystalline
footwall and no recent ground surface rupturing fault scarps were
observed directly against it, we have mapped this as a fault contact
to portray the crystalline fault rocks in juxtaposition with the ll
(Figs. 2 and 3). It is likely that recent fault scarps exist in the ll but
mapping these was beyond the scope of our work. There are no
exposures of the crystalline hanging wall at the Mirrors. The
exposed fault core is complex but ranges in thickness from w1 to
5 m. However, partial erosion of the core makes it difcult to
estimate the in situ thickness or geometry of the entire fault core.
Although previously mapped as hornblende gabbros and anor-
thosites of the Jurassic Humboldt igneous complex (Speed, 1976),
the protolith at the Mirrors locality is a ne to medium-grained
granodiorite with chloritization of biotite and hornblende in the
protolith and variable chloritic to argillic to silicic alteration in the
damage zone.
Semiquantitative X-ray diffraction (XRD) patterns were
obtained from representative, whole-rock samples of each fault
zone component (Table 1). The samples were X-rayed with a quartz
standard and Cu-Ka radiation from 5
to 90
two-theta. Spectra
1
3
5
3
DZ
68
FC
DZ 68
DZ
PL
76
64
PL Protolith Geologic Contact
DZ Damage Zone
FC Fault Core Topographic Contour
Qal Quaternary Alluvium
Strike and Dip of Contact
64
70
N
~ Contour Interval 37 meters
0 ~ Scale (meters) 305
Qal
Qal
PL
PL
FC
72
69
60
FC
1
2
0
7
The Stillwater
Fault Zone
Dashed where gradational.
Bar and ball on hanging wall.
FI IG G. .
6AA
FI IG G. .
4
Fig. 2. Simplied geologic map of the Stillwater Fault Zone at the Mirrors Locality. Note that the protolith/damage-zone contact is gradational (dashed line) and the damage-zone/
fault-core contact is abrupt (solid line). The approximate locations of Figs. 4 and 6A are shown for reference.
were manually matched to library peak intensity data for individual
mineral determination and the estimated error for each reported
mineral is 10 weight percent.
The contacts between, and physical properties of, the fault core,
damage zone, and protolith were mapped using standard tape and
compass techniques using a 1:24,000 topographic base map blown
up to a scale of approximately 1:6,000 (Figs. 2 and 3). Detailed
fracture and vein data (e.g., orientation, intensity, and failure mode
interpretation, i.e., shear or extension fracture) were also collected
and used to help dene each architectural component. Fracture
data were collected using standard scanline and grid counting
methods (e.g., Priest, 1993) in three perpendicular directions
related to strike, dip, and slip direction of the fault zone (e.g., Seront
et al., 1998; Caine and Forster, 1999) at a variety of representative
outcrops. Data were integrated across a number of scales of
observation with centimeter-scale petrographic analyses, meter-
scale outcrop mapping, and at tens to hundreds of meters-scales
using low-elevation aerial photographs.
4. Fault zone structure and fault rocks
4.1. Fault zone orientation, component contact relationships,
and mineralogy
The SFZ strikes northeast to east-northeast and dips from 32
to
70
southeast (Figs. 1e5). The broad range in orientation of major

slip surfaces shown in Fig. 5 reects signicant map to outcrop-
scale variation in the topography of multiple slip surfaces (cf. Power
and Tullis, 1989; Sagy et al., 2007). Slickenlines indicate that the
fault is dominantly normal dipeslip with minor components of left
and right lateral strikeeslip (Fig. 5).
Figs. 2 and 3 show that the fault core and damage zone are
tabular to curvitabular bodies. The contact between the fault core
and damage zone is marked by an abrupt transition between the
cataclastic rocks and distinct breccia bodies of the core and highly
fractured and veined granodiorite of the damage zone (Figs. 2e4).
Portions of this contact are marked by a single curviplanar, pol-
ished and striated slip surface, whereas at other locations the
fault rocks at the contact are breccias. Some polished slip surfaces
are on the order of 100 m
2
, exposed over a strike distance of
about 100 m (Fig. 6A). Additionally, distinct and overlapping slip
surface sheets of polished and striated quartz within the core
show different lineation directions. In some locations several,
anastomosing primary slip surfaces exist at the contact and it is
unclear if only one or all of these were active during deformation.
The fault-core and damage-zone contact is both a structural and
a lithologic contact, as the fault-core rocks are highly silicied
and the damage-zone rocks generally are not. In contrast to the
distinct fault-core/damage-zone contact, the contact between the
damage zone and protolith is gradational and marked by
decreasing fracture intensity, silicication, and argillic alteration
(Figs. 2e4).
Whole-rock XRD analysis of representative samples from each
fault-zone component shows spatially systematic changes in
composition from the fault core to the protolith at the map to
outcrop scale. Quartz is the dominant mineral in the fault core
whereas plagioclase feldspar is dominant in the damage zone and
protolith (Table 1). Kaolinite, carbonates, barite, hematite, chlorite,
biotite, and titanium oxides are also present in the fault core and
damage zone. The damage zone shows mineralogical evidence of
argillic alteration compared with the protolith.
4.2. Protolith and damage-zone fracture networks
Several macroscopic fracture types were interpreted from the
orientations and morphology of distinctive fracture sets relative to
the average orientation of the master fault zone, assuming an
Andersonian stress regime typical for normal dipeslip faults (Fig. 5;
cf. Caine and Forster, 1999). Fractures that strike subparallel to the
fault zone but are subvertical and planar are interpreted to be
extension fractures. Curviplanar fractures oriented parallel to the
fault zone that have normal dipeslip slickenlines, and (or) minor
normal offsets and those that strike parallel to the fault zone but
have antithetic dips, are interpreted to be shear fractures. A set of
subhorizontal fractures are termed step fractures as they result in
stepped shaped outcrops. They are generally rough but planar to
curviplanar, and are oriented at high angles to the average orien-
tation of the master fault zone. Step fractures have an uncertain
origin but may be similar to linking fractures and small faults found
in compound fault zones of the Sierra Nevada of central California
Protolith
Qal
0 ~ Scale (meters) 305
Qal
Damage Zone (DZ)
Protolith
Clast
Supported
Breccias
Pods of Matrix
Supported
Breccias
Core
DZ
~N
Fig. 3. Schematic block diagram showing breccia pods, clast and matrix-supported
breccias in the fault core. Note the abrupt contact between the fault core and damage
zone and the gradational contact between the damage zone and the protolith. The
approximate location of Fig. 6A is shown for reference. Qal Quaternary alluvium.
FFa auul lt t
CCoor ree
PPrroot tool li it thh
DDaammaagge e
ZZo on ne e
~ N NEE
~~ 1 1 mmeetteerr
QQa all
Fig. 4. Photograph of the footwall, fault-zone components and contacts. Diagonally
from left to right in the photograph are Quaternary alluvium, the brecciated and
silicied fault core, the highly fractured damage zone, and the gradational contact
between the damage zone and the relatively less fractured protolith.
(Martel, 1990). Mixed-mode fracturing is suggested locally by the
presence of horsetail-like or feather-like splay fractures (Fig. 6, cf.
Micklethwaite, 2009). There are three sets of fractures that appear
to have no systematic orientation with the master fault zone but
cross all but the shear fracture sets and thus are termed cross
fractures (Fig. 5). They are typically steeply to moderately dipping,
are often smooth walled, and planar. Interestingly, the critically
stressed and hydraulically conductive fractures associated with the
geothermal reservoir observed by Barton et al. (1997) and Hickman
et al. (1997) have reasonably similar orientations to two of the cross
fracture sets and to the silicied slip surfaces found in the exhumed
core of the SFZ (Fig. 5).
Macroscopic cross fractures are primarily found in the protolith
but are also found in the damage zone (Fig. 5). The median fracture
intensity inclusive of all fracture types observed from 11 different
locations in the protolith is 16 fractures per meter with a minimum
of 10 and maximum of 25 per meter. Microfracturing is not abun-
dant in the protolith (Fig. 6). Because cross fractures are generally
Table 1
Mirrors locality X-ray diffraction data averaged for each fault zone component. Major rock forming mineral concentrations reported in weight percent, 10 percent error.
Feldspars are predominantly plagioclase and accessory minerals include chlorite, biotite, and hornblende.
Sample type (n number of samples) Quartz Plagioclase K-feldspar Kaolinite Carbonate FE and TI oxides Amorphous or accessory
Fault-core rocks
Pink slip surface/rounded clast breccia (n 1) 88 0 3 1 0 8
Clast-supported breccias (n 3) 63 0 25 4 5 3
Matrix-supported breccia whole rock (n 3) 79 0 10 3 3 5
Matrix supported breccia matrix only (n 1) 74 0 22 2 0 2
Damage zone (n 4) 24 45 9 15 5 2
Protolith (n 6) 10 71 6 2 3 6
D
Shear
Fractures
Extension
Fractures
Step
Fractures
Cross
Fractures
Master
Fault
Zone
3

3
Damage Zone
n=594 1
2
4c
3
4a
4b
1
2
B
4c
N
Protolith
n=147
4b
4a
4c
4c
4b
A
N
N
Geothermal
Reservoir
Fractures
C
Fault Core
n
surfaces
=34
n
lineations
=21
SFZ
Fig. 5. Lower hemisphere, equal area projections showing orientations of structures in each fault-zone component at the Mirrors locality (A and B, Kamb contoured poles to
fractures, contour interval 2.0 sigma with the mean SFZ great circle shown in black, modied from Caine and Forster, 1999; C, great circles and slip lineations). A) Data from
protolith only. B) Data from damage zone only. In A and B, fractures broken into sets from raw data are assigned to a mode of formation based on interpretation of mechanical
compatibility with an Andersonian model of a normal fault. Relative to the master fault zone the following sets are dened: 1 extension; 2 shear; 3 step; 4a, 4b, and 4c cross
fractures. C) Gray great circles are slip surfaces and small gray dots are the corresponding slickenlines. The black great circle and large black dot show the mean slip surface and
slickenline orientations and the large black square is the pole to the mean SFZ slip surface. The gray star shows the approximate mean pole to hydraulically conductive and critically
stressed fractures derived from borehole logging in the Dixie Valley geothermal reservoir within several kilometers of the Mirrors locality (from Barton et al., 1997). D) Schematic,
eld-based cross-section of the traces of idealized fracture sets associated with an Andersonian normal fault (modied from Caine and Forster, 1999).
cut by fault-related shear fractures, the cross fractures may have
formed prior to faulting. Additionally, some cross fractures are lled
with a distinctive beige quartz kaolinite mineral assemblage that
emanates from the fault core, through the damage zone and into
the protolith suggesting they were possibly hydraulically
conductive during faulting. Extension, shear, and step fractures do
not generally occur in the protolith, suggesting that they are
mechanically related to faulting as their presence is largely
restricted to the damage zone (Fig. 5).
The damage zone consists of intensely fractured and veined
granodiorite (Figs. 2e7). The median fracture intensity inclusive of all
fracture types observed from 16 localities in the damage zone is 50
fractures per meter with a minimum of 32 and maximum of 99 per
meter. Althoughnosuitable outcrops were found where macroscopic
fracture intensity changes could be continuously measured from the
damage zone into the protolith, the change did not appear to be
systematic from one location to another where partial observation
could be made. Alteration and microfracture intensity also increase
towardthe fault core andsilicicationinthe damage zoneis generally
restricted to veins (Fig. 7; Table 1). Damage-zone veins at all sites are
lled with the same quartz kaolinite mineral assemblage found in
the fault core and many commonly have angular fragments of wall
rock suspended in the matrix (Fig. 7). The veins range in width,
perpendicular to their walls, from a few millimeters up to about
10 cmand have trace lengths of just under a meter up to nearly 10 m.
They occur at intensities of approximately 1 vein per meter to
stockworks with many tens of veins per meter. Veins are planar to
curviplanar but when they occur as stockworks they can have quite
irregular and jagged shapes.
4.3. Fault rocks
A representative sample from the fault core is described from
hand-sample to thin-section scales (Figs. 8 and 9). This sample
contains a major silicied slip surface similar to those found
bounding some of the breccia pods as well as the range of different
types of fault-related breccias with important cross-cutting rela-
tionships representative of the Mirrors locality. Sections cut parallel
and perpendicular to slip were cut to document variations in
textures, microstructures, and fault-rock compositions.
The fault core at the Mirrors locality comprises several struc-
turally and texturally distinct fault-rock types described using
elements of classications by Sibson (1977, 1986a), Sillitoe (1985),
Jbrak (1997), and Mort and Woodcock (2008) including: breccias
with rounded clasts; clast-supported breccias with highly angular
and interlocking clasts; matrix-supported breccias with angular
clasts that could be pieced back together if the matrix was
removed; breccias with open spaces between the clasts; and fault
gouge described in detail below.
4.3.1. Breccias with rounded clasts
Breccias with rounded clasts are predominantly matrix sup-
ported but are also locally clast supported. They are poorly sorted
with respect to size, composed of well-rounded clasts and sub-
angular clasts, and are typically associated with cross-cutting slip
surfaces. These textures are observed in the fault rocks at both the
hand-sample and thin-section scales (Figs. 8 and 9). The uppermost
millimeter of the hand sample shown in Fig. 8 is a polished and
striated, silicied slip surface representative of those described
above. In the rst centimeter of the sample, from the slip surface
toward the base of the polished face (Figs. 8 and 9), the sample
exhibits an ultra-ne-grained, highly indurated, whiteepink
colored breccia matrix. Most of the clasts in this matrix show
distinct clasts within clasts textures. This part of the sample is
nearly identical to samples studied by Power and Tullis (1989).
In thin section, quartz is the primary mineral with minor
amounts of kaolinite within about a centimeter of the slip surface
(Fig. 9A and D). In Fig. 9D the section is oriented perpendicular to
the strike of the fault and parallel to the slip direction. The silicied
hanging wall material shows crystallographic preferred orientation
Fig. 6. A) Exposure of the main Mirrors slip surfaces (SS) and large breccias pods (BP)
outlined in dashed yellow. Approximate contacts between Quaternary alluvium (Qal),
fault core (FC), damage zone (DZ), and protolith (PL) are shown with black lines. The
large corrugated pod is composed of all breccia types, bounded by polished and stri-
ated curviplanar slip surfaces, and is elongated parallel to the average slip vector
indicated by the red arrows. B) Outcrop photograph of a small (compass for scale)
breccia pod with matrix-supported breccias. The long axis of the pod is subparallel to
the average orientation of the master fault zone shown by the red arrow. A possible
feeder vein (FV) connected to a network of very long and wide veins can be seen at the
bottom central portion of the pod. C) Breccia pod veins with jigsaw puzzle breccia.
Note the en chelon clasts and closing taper of the veins away from a larger breccia
pod to the right of the photograph. Each photograph is a northward to northwest-
erward-looking view along the strike of the Stillwater Fault Zone.
(CPO) of quartz, indicated by linear domains of common extinction
in synthetic and antithetic microfaults as described in detail by
Power and Tullis (1989) and is typical of breccia pod bounding slip
surfaces. Scanning electron microscopy shows quartz occurs as
microcrystalline prisms on the order of 10 mm long with approxi-
mate alignment with the synthetic microfaults (Fig. 9E). Quartz
with CPO also occurs as micron-scale coatings surrounding many of
the clasts, with common extinction at average angles of about
60e120
from the slip surface (Fig. 9A). There are also numerous,
discrete rounded clasts and areas of matrix with internal domains
of CPO of quartz that do not make regular angles with the slip
surface consistent with observations by Power and Tullis (1989).
Many of the silicied clasts have varying degrees of optical contrast
with the matrix suggestive of varying degrees of silicication and
Fig. 8. A) Photograph of a polished hand-sample from the fault core. The top of the
sample is a slip surface and the polished face is parallel to the slip direction. The three
major breccia types are exhibited. Note that the breccia with rounded clasts grades into
the clast-supported breccia. In contrast, the contact between the clast-supported
breccia and the matrix-supported breccia is relatively abrupt indicating the cross-
cutting of the former by the latter. B) Example of breccia with vuggy open space. Each
iron oxide stained clast is coated with a rind of carbonate with euhedral crystals. Note
the 10 hand lens for scale.
Fig. 7. Photomicrographs of representative rock samples from the protolith and
damage zone. A) Polarized light image of protolith granodiorite. Equigranular
quartz plagioclase potassium feldspar biotite, which is altered to chlorite. Note
the absence of microfracturing. B) Polarized light image of damage-zone micro-
fractures in granodiorite w10 m from the fault core. C) Polarized light image of
damage-zone granodiorite w5 m from the fault core shows extensive microfracturing,
horsetail fractures, and silicied veins with cataclastic particles.
clay mineral content that result in translucent gray to brown,
layered to clotted, felt-like masses. Little to no carbonate is
observed in these breccias which grades into clast-supported
breccias as the polished face of the representative hand sample is
traversed further down, where the carbonate content increases
(Fig. 8). In other samples and exposures the breccias with rounded
clasts abruptly cut clast-supported breccias or are themselves cut
by discrete slip surfaces (e.g., Fig. 9D).
4.3.2. Clast-supported breccias
In outcrop and hand samples the most common fault-related
breccias are the predominantly clast-supported breccias with local
matrix-supported textures. These breccias have an average matrix-
to-clast ratio of w30:70. The clasts are poorly sorted, angular to
subangular, show interclast veining, and are predominantly
composed of silicied granodiorite protolith (Figs. 8 and 9B). These
breccias show little evidence for cataclastic transport of rock, such
as relative movement or rounding of clasts. This rock type is the
background fault rock that is cut by slip surfaces, breccias with
rounded clasts, matrix-supported breccias, and breccias with open
space. In some samples, clasts of brecciated protolith in the clast-
supported breccias are recognized in thin section and appear
brecciated multiple times. These types of textures are the exception
and are found where the breccias are locally more matrix sup-
ported. Some clasts are locally incorporated into a ne-grained,
quartz-rich breccia but none were observed with domains of quartz
CPO. The clasts and silicic matrix are both cut by carbonate veins,
which are also cut by ne-grained quartz kaolinite veins. The
latter assemblage is cut again by a set of young calcite veins. The
veins commonly include clasts of the protolith, preexisting vein
materials, breccias, and other fault-related rocks that show
multiple brecciation events.
4.3.3. Matrix-supported breccias
Matrix-supported breccias are found in distinctive pod-like
bodies and in veins that emanate from and (or) connect to other
breccia bodies. These breccias are also characterized by angular
breccia clasts of local wall rock that could be pieced back together
if the matrix that separates them was removed. Yet, locally they
have subrounded clasts of the other breccia types discussed
Fig. 9. Polarized light and scanning electron microscope (SEM) micrographs of the three breccias found in Fig. 8. A) Silicied breccia with rounded clasts that formed directly under
a polished and striated slip surface shown by a white arrow. Note the rounded and angular clasts, completely silicied relict clasts (RRC), partially clast-supported texture, and
crystallographic preferred orientation (CPO) of quartz. B) Moderately silicied predominantly clast-supported breccia showing subangular clasts and poor sorting. C) Highly silicied
matrix-supported breccia characterized by angular and rounded clasts sitting in an undeformed microcrystalline, quartz-rich matrix. D) Oriented section of the sharp juxtaposition
of two different breccias with rounded clasts (BRC) separated by a polished and striated silicied slip surface (the section separated along the thick dark irregular line with the white
arrow in it). E) SEM, secondary electron image of the predominantly quartz (Q) matrix with minor kaolinite (K) from the hanging wall of the sample shown in D (mineralogy
determined by energy dispersive spectroscopy). The image is roughly oriented as image D. F) Cathodoluminescence image of a clast within the microcrystalline quartz matrix (black
background). The triangular clast is interpreted as deformed quartz and the background quartz is interpreted as undeformed quartz. Note that the clast is laced with several
generations of healed microfractures (black and gray stringers of quartz).
above, some with exotic compositions from the wall rock. In
either case the clasts are generally widely separated from one
another (Figs. 6, 8 and 9C). Breccias with angular clasts that can be
pieced back together have been previously referred as jigsaw
puzzle breccias and implosion or explosion breccias (cf. Sibson,
1986a; Jbrak, 1997; Power and Tullis, 1989; Mort and
Woodcock, 2008). At the Mirrors, these breccias are matrix sup-
ported and the matrix has a distinctive beige color. They are
poorly sorted, highly silicied, lack veins or fractures, and exhibit
no open, macroscopic pore space. The breccia pods in which the
matrix-supported breccias occur abruptly and irregularly cut the
other breccia types (Fig. 8).
The clasts in the matrix-supported breccias range from 1 mm to
3 cm in size, contain multiple clasts of varying composition with
cross cutting internal brecciation but are compositionally distinct
from the beige matrix (Fig. 8). Clast size distribution is variable but
most clasts are orders of magnitude larger than the matrix (Figs. 8
and 9C). Clasts generally contain deformed angular quartz grains
visible in thin section. Cathodoluminescence analysis of the matrix-
supported breccia samples shows extensive fracturing and multiple
vein-lling events isolated within the clasts with no such defor-
mation or veining in the matrix (Fig. 9F). Clast lithologies include
adjacent wall rock breccias in the jigsawpuzzle breccias. Other clast
lithologies include protolith granodiorite, porphyritic volcanic
rocks, iron-stained and zoned dolomite vein crystals, relicts of
highly silicied breccia clasts in contact with their matrix material,
polished and striated fault surface materials, and possible meta-
sedimentary rock fragments. Calcite occurs as intraclast vein llings
in association with ferroan dolomite and as a late-stage precipitate
that lls some isolated void spaces, small stringer-like veins, and as
dismembered vein fragments.
The clasts are encased in a distinctive ultra-ne-grained
(w0.1e10 mm), beige colored, microcrystalline quartz-rich matrix
with minor amounts of intergrown kaolinite, trace amounts of
carbonate and amorphous material (Table 1). The matrix shows
rather uniformly sized anhedral to subhedral interlocking quartz
crystals with salt and pepper or mosaic-like microtexture (Fig. 9C).
Sub-equant, poorly developed quartz prisms with anhedral quartz
intergrowths are occasionally observed with somewhat reticulate
texture. All quartz grains are completely intergrown and conform
to every surface they are in contact with, showing no evidence for
preferred crystallographic orientation, spherulitic, or feathering
textures. Quartz crystals have uniform to domainal extinction
within individual grains and within masses of grains. The matrix
shows no evidence of brecciation, layering, overgrowths, cross-
cutting veins, or crack-seal textures. Carbonate appears to ll minor
void space as late-stage precipitate.
Maximum syndeformational porosity in the breccia pods was
estimated from representative hand samples of the matrix-sup-
ported breccias. Syndeformation porosity was estimated by digital
imaging of serial sections from several slabbed and polished
samples ranging in size from about 20 by 10 cm to about 3 by 5 cm
(e.g., the lower part of Fig. 8). The samples were scanned and a gray-
scale histogram was captured for each image where the dark
histogram signal was assumed to represent the clasts and the light
histogram signal assumed to represent the matrix. The measure-
ments also assume that the matrix of these breccias represent the
pore space formed during a single deformation event, consistent
with it cross-cutting other main breccias types. This is also
reasonable because the matrix itself is compositionally homoge-
nous, unlike the clasts, and shows no evidence for subsequent
deformation. The average, maximum syndeformation porosity
from this analysis is 54 percent and the analysis conrms the
largely matrix-supported texture using this pseudo three-dimen-
sional approach.
4.3.4. Breccias with open space
At several locations in exposures of the fault core there are
breccias that exhibit signicant open space and vugs (Fig. 8B).
These breccias include generally angular, iron oxide stained clasts
of other breccia types and are clast-supported fault rocks. The
volume of the interclast open space is on the order of microns to
several tens of cubic centimeters. Many of the spaces are coated
with banded and multilayered carbonate rinds that consist of
euhedral calcite crystals that conformably overlay previous layers
and the clasts. There also are euhedral carbonate veins that cut
through the clasts and in some cases connect to the euhedral rinds.
4.3.5. Fault gouge
The nal fault-rock category found at the Mirrors is unconsoli-
dated fault gouge (cf. Sibson, 1977). Only one thin seam, less than
10 cm in thickness, of this material was found. The gouge is ne to
medium grained and contains silicied clasts of the breccia types
described above. The gouge is not clay-rich as is commonly found in
some shallowly formed, brittle faults (cf. Morrow et al., 1984;
Foxford et al., 1998; Heynekamp et al., 1999; Vrolijk and van der
Pluijm, 1999; Aydin and Eyal, 2002). Although the exposure of
gouge cuts the clast-supported breccias its temporal relationships
to the other breccias types were not observed.
4.4. Internal structure of the fault core
A characteristic feature of the fault core is the occurrence of
matrix-supported breccias and other fault rocks that form pod-like
bodies (Figs. 3 and 6). The pods are generally lenticular with their
long axes subparallel to the average orientation of the slip vector for
the master fault zone. The short axes of the breccia pods are typi-
cally orthogonal to slip in the plane of the fault. Large slip surfaces
have elongated scoop-like depressions and associated ridges
parallel to the slip direction. The walls of the breccia pods that are
not bounded by slip surfaces are irregular and where exposed, they
are enveloped in fractured protolith. One pod-shaped structure
shows an along-strike dimension of several meters, a down-dip
dimension of w35 m, and a thickness of less than w5 m(Fig. 6). The
bulk of the pod is predominantly composed of silicied clast-sup-
ported breccias with interior meter-scale pods of the beige matrix-
supported breccias. This pod was one of the largest in outcrop; pods
on the order of w2 m or less in length are more typical.
The dimensions of exposed curviplanar or corrugated slip
surfaces that bound breccia bodies were characterized using
wavelength and amplitude measurements made with a tape
measure in the plane of the fault orthogonal to the slip direction.
The dimensions range from a maximum of approximately 35 m in
length to 3 m in amplitude at the map scale and 4 m in length to
0.5 m in amplitude at the outcrop scale. There is insufcient
exposure to make full measurements of anisotropy parallel to slip
at the map scale; however, at the outcrop scale on average the pods
vary in length from roughly 2 m to 15 m, with length to amplitude
ratios between 7:1 and 20:1. There is a difference in the wavelength
to amplitude ratio of breccia body corrugations in a direction
perpendicular to slip versus parallel to slip. At the outcrop scale,
corrugations parallel to slip may be up to three times as long as
those perpendicular to slip. In spite of the crude nature of these
estimates, they are consistent with other more sophisticated
measurements of fault surface roughness for the Mirrors and other
localities over a number of scales (cf. Lee and Bruhn, 1996; Sagy
et al., 2007).
En chelon, highly angular beige matrix-supported breccia
clasts in veinlets that extend from the breccia pods, indicate wall
rock fracture, shear, and subsequent uid ow within the breccia
pod interiors prior to mineral precipitation and sealing (Fig. 6).
Breccia-lled veins at the margins of the pods taper away from the
pods and terminate. Some of these veins, similar to that seen in
Fig. 6, are connected to an extensive network of veins that are
irregularly shaped, show variable thickness, and extend into the
damage zone.
5. Discussion
The SFZ at the Mirrors locality comprises a complex set of
structures and rock types that include multiple slip surfaces,
distinct fault rocks and breccia bodies, fault and non-fault-related
fracture networks. We infer that several processes acted together
during the formation of the fault zone and that some of the rocks
may record coseismic deformation and associated uid ow. In the
following discussion we present key observations and a conceptual
model that links deformation and uid ow to architecture,
permeability structure, and fault-zone growth.
5.1. Origin of the breccias and breccia pods
The key observations and interpretations that bear on the origin
of the different breccia types and breccia pods as recording
coseismic processes, and the growth of the SFZ include the
following. The breccia pods are discontinuous, elongate, and
tapered bodies aligned subparallel to the slip direction of the
master fault zone. These pods are primarily composed of clast-
supported breccias with local matrix-supported breccias. Silicied
slip surfaces and associated breccias with rounded clasts cut the
clast-supported breccias but are also locally brecciated. The matrix-
supported breccias are emplaced into the surrounding clast-sup-
ported breccias, slip surfaces, and breccias with rounded clasts
within the pods. Pod morphology, spatial restriction to the fault
core, kinematic compatibility with normal faulting, and large esti-
mated syndeformational porosity of the matrix-supported breccias
suggest that there was a component of dilatational strain during
deformation and pod formation. Slip surfaces and some matrix-
supported breccias with en chelon clasts also record a component
of shear deformation.
The topography of many of the large slip surfaces that bound the
pods includes elongated scoop-like depressions that may have
formed along local extensional fault jogs. These relatively small-
scale depressions may also record local dilatational strain and
permeability enhancement during coseismic deformation (e.g.,
Power and Tullis, 1989). Sibson (1986a) inferred that brecciation
could be caused by pore uid-pressure differentials between the
wall rock and an incipient, dilatant opening within a fault zone. If
the effective stress acting on a fault zone exceeds the tensile
strength of the wall rock, catastrophic failure can cause implosion
of the wall rock into openings. Many of the breccia pods, however,
do not occur near slip surfaces and are larger than openings due to
different topography along adjacent slip surfaces. This, and the
fault-slip parallel morphology of the larger breccia pods, suggests
that some pods may have originated by processes other than
dilatational strain in jogs caused by variations in slip-surface
topography. These larger breccia pods may have formed along
mechanically weak sites within the fault core such as preexisting
polished and striated slip surfaces, macroscopic fractures that were
partially mineral-lled, microfractured regions, or zones of
compositional variation.
The clast-within-clast textures found in the breccias provide
a record of multiple faulting or slip events localized in the core. The
matrix-supported jigsaw puzzle breccias have clasts that are
composed of the adjacent wall rock and these are generally angular
suggesting little transport. However, locally in the breccia pods,
where there are widely separated clasts in the same matrix as the
jigsaw puzzle breccias, the clasts are composed of exotic parent
lithologies and are subangular to subrounded indicating signicant
transport. The rounded clasts may also indicate mechanical attri-
tion fromgrain-size reduction given their proximity to slip surfaces.
Clast rounding may also have occurred during hydrothermal uid
owand associated abrasional wear of clasts against one another in
saturated, silica-rich rock our (cf. Sillitoe, 1985; Jbrak, 1997).
The matrix of breccia pods with matrix-supported clasts is
composed of distinctive beige colored, homogenous, ne-grained,
quartz with minor kaolinite and carbonate. The quartz has a mosaic
texture with uniformly sized (w0.1e10 mm) anhedral to subhedral
interlocking crystals that form domains of variable optical extinc-
tion and that completely inll and conform to all available space.
Lovering (1972), Fournier (1985), Saunders (1994), and Dong et al.
(1995) provide evidence that this mosaic texture forms by recrys-
tallization of amorphous silica that precipitated from the boiling of
hydrothermal uids and associated processes at temperatures
<180

C. At the temperatures of the modern geothermal reservoir,
a rapid pressure drop could result in such decompressional (iso-
enthalpic) boiling, phase separation, cooling, and rapid precipita-
tion of amorphous silica (Hedenquist and Henley, 1985; Saunders,
1994; Rimstidt, 1997). Such pressure drops could have occurred
in seismically induced dilatant openings localized in the fault zone
(e.g., Sibson, 1986a) and that could have also hydraulically con-
nected the reservoir to the surface. Decompressional boiling is
consistent with the minor kaolinite and carbonate minerals found
in small interstices of the mosaic quartz matrix. These phases are
also found in fault rocks elsewhere along the SFZ (Parry et al., 1991).
Loss of CO
2
to the vapor phase during boiling, with attendant
increase in pH, will cause clay and carbonate to precipitate (e.g.,
Fournier, 1985). Such decompression coupled with rapid amor-
phous silica deposition required rapid ow of uid and vapor up
through, and laterally within, the fault zone. Surcial expression of
this type of uid movement is evident as sinter mounds and
terraces, hot springs, and fumaroles such as those found at the
ground surface in association with the fault zone (cf. Caine and
Forster, 1999; Lutz et al., 2002).
Breccia-lled veins, with the same quartzekaolinite matrix as
the breccia pods emanate from and taper outward from the pod
margins and terminate. This may indicate that uid ow and the
pressure gradient were directed outward fromthe pod interior. The
pods then likely became permeability heterogeneities within the
fault core leading to a heterogeneous spatial distribution of
cemented or sealed portions of the fault zone. This, in turn, may
have lead to heterogeneities in fault-zone strength as the fault zone
evolved.
Observations of the juxtaposition of polished slip surfaces
against breccias with rounded clasts, cutting across clast-supported
breccias with angular clasts, then cut by matrix-supported breccias
with jigsaw puzzle and rounded clasts, open space breccias, the
irregular character of some of the larger breccia-lled veins, and
a seam of clay-rich gouge indicate that different mechanisms of
failure may have been operative in the fault zone over time. Some of
these mechanisms may be attributed to a non-static stress eld
changing dynamically during deformation or attributed to varia-
tions in uid pressures within the fault zone at different times (cf.
Caskey and Wesnousky, 1997; Micklethwaite, 2009). Multiple,
anastomosing, layered slip surfaces with curviplanar geometry and
different lineation directions indicate some portion of slip during
any deformation event was distributed along any number of pre-
existing surfaces from previous events. However, no evidence was
found to evaluate the nature of the actual slip distribution on an
event-by-event basis.
If breccia pod formation was seismogenic, it is unclear if they are
related to hypocentral earthquake nucleation sites or possibly
record rupturing in more distal and shallower parts of the fault
zone. The breccia pods and associated rocks may be unique
remnants of the deformation products linked to the distribution of
earthquake foreshocks, aftershocks, and microseismicity as
rupturing initiated and ceased, and pore pressure adjusted to
structural changes throughout the fault zone (cf. Nur and Booker,
1972; Byerlee, 1993; Miller et al., 1996, 2004). Episodic deforma-
tion and resulting fault rocks evolved into a hydraulically hetero-
geneous fault zone with local permeability contrasts great enough
to possibly have trapped uids under varying pressures. Byerlee
(1993) proposed a uid-pressure compartment model for fault
zones associated with earthquakes where uid sealed under high
pressure in hydraulically isolated compartments partially drains
into adjacent compartments of initially lower uid pressure during
a deformation event. Although not a direct comparison, Parry and
Bruhn (1990) and Parry et al. (1991) showed evidence for tran-
sient variations in uid pressure from uid inclusion data from the
SFZ about 40 km south of the Mirrors locality. The breccia pods are
not likely remnants of Byerlee-like pressure compartments per se,
however, the locations, sizes, shapes, and breaching of the breccia
pods were likely inuenced by variations in permeability, pore uid
pressure, and mechanical heterogeneities; features that may also
help to explain the variable uid pressures observed by Parry and
Bruhn (1990).
5.2. Heterogeneity and anisotropy within the fault zone
Several types of structural heterogeneities and anisotropies may
act together to impact uid ow, mechanics, and growth of the SFZ.
Fault-related permeability anisotropy in crystalline rock is
controlled by fracture networks and heterogeneously distributed
fault-rock types, such as the breccia bodies, with varying perme-
abilities (e.g., Seront et al., 1998; Caine and Forster, 1999). The
corrugated topography of the major slip surfaces that bound the
fault core and minor slip surfaces within the fault core may have
had a role in controlling the shapes and limited the growth of
breccia bodies. The elongation of these bodies parallel to the slip
direction should also contribute to permeability heterogeneity.
Anisotropic slip-surface topography may also be an important
cause of changes in local stress elds (Ferrill et al., 1999). The
distribution of slip surface and slip vector orientation data suggests
that most of the oblique slip is related to the accommodation of
deformation along the corrugated slip surfaces consistent with
attendant localized changes in the local stress eld (Fig. 5).
Outcrop-scale fracture network heterogeneities, such variations
in total fracture intensity between the damage zone and protolith,
the apparent unsystematic manner in which fracture intensity
changes with distance fromthe fault core, or internal heterogeneity
of the number of shear fractures and minor slip surfaces in any
given location, suggest that deformation and slip for any given
deformation event were heterogeneously distributed. Thus, fault
properties such as strain, strength, and permeability depend on the
range of scales over which fault-rock and damage-related hetero-
geneities exist. For example, the elongate nature of the bodies of
fault-core breccias, the contrast of fracture network intensity
between the fault core and damage zone, and the variable degree of
silica cementation in the fracture networks are evidence of signif-
icant hydrogeologic and mechanical heterogeneities and anisot-
ropies, implying there is signicant heterogeneity of fault-zone
strength and permeability from at least the outcrop scale to the
map scale.
Seront et al. (1998) measured permeability values on repre-
sentative core-plug samples that ranged from as high as 10
8
m
2
in
the damage zone to as low as 10
20
m
2
in the fault-core matrix.
Given the map-scale to micro-scale change in fracture intensity is
higher in the damage zone than in the fault core and that the core is
the locus of silicication, a heterogeneous combined conduit-
barrier permeability structure is suggested for the fault zone during
interseismic periods (cf. Caine and Forster, 1999). Although specu-
lative, this permeability heterogeneity could also lead to pore-
pressure heterogeneity and consequently strength heterogeneity
throughout the seismic cycle.
Seront et al. (1998) also found no appreciable difference in
mechanical strength of small laboratory samples of each fault-zone
component at the Mirrors. Their experimental results highlight the
importance of scale when considering the relative mechanical
strength within a fault zone composed of discrete components. For
example, uncemented, smooth-walled and discontinuous shear
fractures observed along the fault-core/damage-zone contact, and
within the damage zone, would provide favorably oriented failure
surfaces for later deformation and uid ow events (e.g., Barton
et al., 1997). The mechanical strength of these macro-scale frac-
tures would depend on frictional properties of the fractures rather
than grain-scale cohesion or microfracture properties that control
strength at the core-plug scale.
5.3. Model for deformation induced uid ow
We propose the following conceptual model that relates defor-
mation, growth, fault-related uid ow, and stress cycling to
architecture and permeability structure (Fig. 10). The proposed
model is built on the work of Power and Tullis (1989), Parry and
Bruhn (1990), Parry et al. (1991), Muir-Wood (1994), Seront et al.
(1998), and Sibson (2001) and assumes a constant maximum and
decreasing minimum principal stress in an Andersonian normal
fault regime.
During interseismic periods, uids were stored in an ancient
equivalent of the modern geothermal reservoir, in primary pore
space, and secondary macroscopic fracture networks in and
surrounding the fault zone (Fig. 10). The hydrothermal silica
cements in the fault core were likely precipitated from uids that
initially equilibrated with rocks in this reservoir. Aseismic defor-
mation along slip surfaces may have occurred as hypothesized by
Power and Tullis (1989). As differential stress increased, preexist-
ing fractures primarily in the damage zone, could have opened due
to decreasing minimum compressive stress. New fractures may
also have formed which mobilized uids and possibly dissolved
silica from comminuted by-products formed in preexisting frac-
tures in the surrounding host rock.
During coseismic brittle failure events new hydraulic connec-
tions to the geothermal reservoir were formed as ruptures propa-
gated likely releasing seismic energy. Quartz-supersaturated uid
and vapor inltration, ow, boiling and (or) phase separation
occurred in these new, highly permeable pathways (cf. Miller et al.,
2004). This resulted in rapid quartz matrix precipitation and
suspension of clasts forming the matrix-supported breccias in
hybrid dilatant and shear related openings localized in the fault
core. The effective minimum compressive stress also increased,
closing optimally oriented damage-zone fractures further mobi-
lizing uids into the fault core with accompanying shear along both
preexisting and newly formed discrete fractures and fracture
networks (cf. Muir-Wood and King, 1993; Seront et al., 1998;
Fig. 10). Comminution and the development of breccias with
rounded clasts were localized along constrictions between discrete
slip surfaces. Adjacent to slip surfaces and in the wall rock, in situ
distributed failure, possibly crushing, resulted in predominantly
clast-supported breccias. Some of the clast-supported breccias with
localized areas of matrix-supported textures may be older equiva-
lents of breccias similar to the beige matrix-supported breccias.
This process appears to have been repeated numerous times and
may have resulted in an increase in width of the fault zone as well
as heterogeneous sealing of the fault core. Dilatancy hardening may
also have resulted in associationwith uid-pressure drops (cf. Parry
et al., 1991).
Alternatively, the formation of the matrix-supported breccias
in discrete pods and some of the clast-supported breccias may
have formed by the fault-valve mechanism (Sibson, 1990) where
failure was driven by pore uid pressure rather than tectonic
stress. The present geothermal reservoir near the Mirrors locality
appears to be in a heterogeneous pressure regime largely at or
near hydrostatic pressure but with areas of enhanced pressure
(Hickman et al., 2000). Whereas, deep wells drilled outside of the
productive geothermal reservoir have artesian pressures (Dick
Benoit, personal communication, 2009). It may have been possible
that there were volumes of the ancient reservoir with localized
areas of overpressure connected to the fault zone or zones within
the fault with anomalous uid pressures possibly due to perme-
ability heterogeneity as suggested above. In either case, the
various breccia textures do not provide unique evidence for
a tectonic stress versus pore uid-pressure mechanism for their
origin.
Exhumation of the Mirrors portion of the fault zone brought the
fault zone into progressively different temperature, pressure, and
uid chemical conditions. The breccias with open space may have
originated from low pH ascending uids and yet clasts coated with
euhedral carbonate rinds are consistent with downwelling, silica
undersaturated uids both possible processes during interseismic
periods. Cross-cutting seams of fault gouge like represent yet
another phase of deformation largely lacking signicant localized
uid ow and associated hydraulic connection.
Fault
Core
Damage
Zone
Protolith
Damage Zone
Protolith
Interseismic Period
3
Decreases
Fluids Accumulate
Coseismic Period
3
Increases
Hybrid Failure
Fluids Mobilize
Fault Core
Damage Zone
Protolith
Damage Zone
Protolith
1
Constant
3

3
1
Induced
Hydraulic
Connections
to Geothermal
Reservoir
Matrix
Supported
& Jigsaw Clast
Supported
Rounded
Clasts
Major Breccia Types
Fig. 10. Schematic diagram and conceptual model of deformation, breccia formation, hybrid shear and dilatant fault growth, and uid ow inferred from the Mirrors locality
(hydrologic effects adapted from Muir-Wood, 1994 and specic structures from this work). The regional maximum principal stress (s
1
) is assumed to be constant and vertical
throughout the evolution of the fault zone. The minimum principal stress (s
3
) is assumed to be horizontal. At an early time (top diagram) in its evolution, the fault core is
predominantly composed of clast-supported breccias cut by breccias with rounded clasts (middle gray ll) and associated anastomosing slip surfaces (black lines that bound and
weave through the fault core). In the interseismic period when s
3
decreases as differential stress increases, optimally oriented fracture networks dilate (stubby arrows indicate local
stress), take on and store uids (curved arrows indicate uid transport direction). During a coseismic period, mixed-mode deformation (shear and dilatancy) is localized in the fault
core creating new hydraulic connections to an ancient geothermal reservoir accompanied by uid ow into breccia pods (light gray ll), development of breccias with rounded
clasts, and rapid precipitation of quartz due to decompressional boiling, locally sealing the fault zone. Because effective s
3
is increased, uids may also ow into the high
permeability fault core from optimally oriented damage-zone fracture networks.
6. Conclusions
Field mapping along the Stillwater Fault Zone provides
a detailed view of the internal structure of the footwall of a seis-
mogenic normal fault zone. Distinct mechanical and hydraulic
components include a fault core, damage zone, and protolith (cf.
Seront et al., 1998). The different components are distinguished by
distinctive variations in lithology, amount and type of hydro-
thermal alteration, fracture intensity, matrix-scale hydraulic prop-
erties, and structure. The fault core shows evidence for localized
deformation along silicied slip surfaces near large dilatant breccia
pods indicative of distributed deformation (cf. Power and Tullis,
1989). The core is surrounded by damage-zone fracture networks
also indicative of distributed deformation. The juxtaposition of
these distinct deformational styles leads us to characterize the
architecture of the fault zone at the Mirrors locality as a composite
deformation zone (Caine and Forster, 1999) where different fault-
related breccias and other fault rocks are indicative of multiple
deformation events possibly under different stress and (or) uid-
pressure conditions as the fault zone evolved.
Laboratory-derived permeability values indicate up to approxi-
mately four orders of magnitude difference in permeability
between the fault rock, damage zone, and protolith samples at the
core-plug scale. This and fracture network modeling (Caine and
Forster, 1999) in combination with the mapped fault-zone
components lead us to infer that the bulk permeability structure of
the present day fault zone is a heterogeneous and anisotropic
combined conduit-barrier ow system (e.g., Caine et al., 1996).
During seismic events, much of the fault zone behaves as a complex
and heterogeneous conduit where ow paths are likely controlled
to a large degree by mechanical heterogeneities, and the local
stresses they may induce (e.g., Faulkner et al., 2006). These
heterogeneous features are formed during progressive deformation
and associated sealing events. The source of uids, uid tempera-
tures, and component chemical constituents also play a large role in
the distribution of permeability and mechanical heterogeneity
from one seismic cycle to another.
The breccia pods contain remnants of likely coseismic, outcrop
to map-scale openings with components of hybrid dilatant and
shear deformation. Distinctive matrix-supported and jigsaw puzzle
breccias within these pods have microcrystalline, quartz-rich
mosaic textures that may record rapid, hydrothermal uid owand
mineral precipitation induced by the coseismic dilatancy and
hydraulic connection to an ancient geothermal reservoir. Alterna-
tively, the matrix-supported breccias and associated pods may have
formed from uid overpressures either in connection with the
geothermal reservoir or locally within the core of the fault. In either
case, the various breccia textures do not provide unique evidence
for a tectonic stress versus pore uid-pressure mechanism for their
origin. Yet, if either hypothesis is correct, these distinctive fault
rocks are formed during high enough strain rates to qualify as being
coseismic, and thus the breccias represent seismogenic rock
textures (cf. Cowan, 1999). Although, the fault rocks at the Mirrors
locality may be coincidental and unique to the Stillwater Fault Zone,
the textures and associated inferences regarding their formation
may be underappreciated where similar combinations of rocks,
structures, and geothermal uid reservoirs occur.
Acknowledgements
Funding for this work was provided by a U.S. Geological Survey,
National Earthquake Hazards Reduction Program Grant # 1434-93-
G-2280 to Forster and Bruhn who supported Caine with a graduate
research assistantship at the University of Utah, Department of
Geology and Geophysics from 1992 through 1995. We thank Bill
Parry, Laurel Goodwin, Darrel Cowan, George Breit, Don Sweetkind,
and Albert Hofstra, for helpful comments on an early version of this
manuscript. Lyndsay Ball, Dan Faulkner, David Lockner, Steven
Micklethwaite, and Rick Sibson provided critical reviews that
greatly improved the manuscript. We also thank Bennet Leeper,
Lori Chadwell, Kathleen Royster, and Bernard Seront who provided
eld assistance.
References
Antonellini, M., Aydin, A., 1994. Effect of faulting on uid ow in porous sandstones;
petrophysical properties. AAPG Bulletin 78, 355e377.
Aydin, A., Eyal, Y., 2002. Anatomy of a normal fault with shale smear; implications
for fault seal. AAPG Bulletin 86 (8), 1367e1381.
Barton, C.A., Hickman, S., Morin, R., Zoback, M.D., Finkbeiner, T., Sass, J., Benoit, W.R.,
1997. Fracture permeability and its relationship to in-situ stress in the Dixie
Valley, Nevada, geothermal reservoir. Report SGP-TR-155. In: Proceedings of the
22nd Workshop on Geothermal Reservoir Engineering. Stanford University,
Stanford, California, pp. 147e152.
Bruhn, R.L., Yonkee, W.A., Parry, W.T., 1990. Structural and uid-chemical properties
of seismogenic normal faults; earthquake source processes. In: 19th IUGG
General Assembly, Symposium on Earthquake Source Processes, Vancouver, BC,
Canada, Aug. 19e21, 1987, vol. 175, pp. 139e157.
Bruhn, R.L., Perry, W.T., Yonkee, W.A., Thompson, T., 1994. Fracturing and hydro-
thermal alteration in normal fault zones; faulting, friction, and earthquake
mechanics; part 1. Pure and Applied Geophysics 142, 609e644.
Byerlee, J.D., 1993. Model for episodic ow of high-pressure water in fault zones
before earthquakes. Geology 21, 303e306.
structure. Geology 24, 1025e1028.
Caine, J.S., Forster, C.B., 1999. Fault zone architecture and uid ow; insights from
eld data and numerical modeling. In: Haneberg, W.C., Mozley, P.S., Moore, J.C.,
Goodwin, L.B. (Eds.), Faults and Sub-surface Fluid Flow in the Shallow Crust.
AGU Geophysical Monograph, vol. 113, pp. 101e127.
Caskey, S.J., Wesnousky, S.G., Zhang, P., Slemmons, D.B., 1996. Surface faulting of the
1954 Fairview Peak (M
s
7.2) and Dixie Valley (M
s
6.8) earthquakes, central
Nevada. Bulletin of the Seismological Society of America 86, 761e787.
Caskey, S.J., Wesnousky, S.G., 1997. Static stress changes and earthquake triggering
during the 1954 Fairview Peak and Dixie Valley earthquakes, central Nevada.
Bulletin of the Seismological Society of America 87 (3), 521e527.
Chester, F.M., Evans, J.P., Biegel, R.L., 1993. Internal structure and weakening
mechanisms of the San Andreas Fault. Journal of Geophysical Research 98 (B1),
771e786.
Chester, F.M., Logan, J.M., 1986. Implications for mechanical properties of brittle
faults from observations of the Punchbowl fault zone, California. Pure and
Applied Geophysics 124, 79e106.
Cowan, D.S., 1999. Do faults preserve a record of seismic slip? A eld geologists
opinion. Journal of Structural Geology 21, 995e1001.
Cox, S.F., Knackstedt, M.A., Braun, J., 2001. Principles of structural control on
permeability and uid ow in hydrothermal systems. Reviews in Economic
Geology 14, 1e24.
Dong, G., Morrison, G., Jaireth, S., 1995. Quartz textures in epithermal veins,
Queensland; classication, origin and implication. Economic Geology and the
Bulletin of the Society of Economic Geologists 90, 1841e1856.
Evans, J.P., Chester, F.M., 1995. Fluiderock interaction in faults of the San Andreas
system: inferences from San Gabriel fault rock geochemistry and microstruc-
tures. Journal of Geophysical Research 100 (B7), 13007e13020.
Faulkner, D.R., Mitchell, T.M., Healy, D., Heap, M.J., 2006. Slip on weak faults by
the rotation of regional stress in the fracture damage zone. Nature 444,
922e925.
Faulkner, D.R., Mitchell, T.M., Rutter, E.H., Cembrano, J., 2008. On the structure and
mechanical properties of large strikeeslip faults. In: Wibberley, C.A.J., Kurz, W.,
Imber, J., Holdsworth, R.E., Collettini, C. (Eds.), The Internal Structure of Fault
Zones: Implications for Mechanical and Fluid Flow Properties. Geological
Society, London, Special Publications, vol. 299, pp. 139e150.
Ferrill, D.A., Stamatakos, J.A., Sims, D., 1999. Normal fault corrugation; implications
for growth and seismicity of active normal faults. Journal of Structural Geology
21, 1027e1038.
Forster, C.B., Evans, J.P., Torgersen, T., 1991. Hydrogeology of thrust faults and
crystalline thrust sheets; results of combined eld and modeling studies.
Geophysical Research Letters 18, 979e982.
Fournier, R.O., 1985. The behavior of silica in hydrothermal solutions. In:
Berger, B.R., Bethke, P.M. (Eds.). Geology and Geochemistry of Epithermal
Systems. Reviews in Economic Geology 2, 45e61. Society of Economic
Geologists.
Foxford, K.A., Walsh, J.J., Watterson, J., Garden, I.R., Guscott, S.C., Burley, S.D., 1998.
Structure and content of the Moab Fault Zone, Utah, USA, and its implications
for fault seal prediction. In: Jones, G., Fisher, Q.J., Knipe, R.J. (Eds.), Faulting, Fault
Sealing, and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London,
Special Publications, vol. 147, pp. 87e103.
Ghisetti, F., Kirschner, D., Vezzani, L., Agosta, F., 2001. Stable isotope evidence for
contrasting paleouid circulation in thrust faults and normal faults of the
central Apennines, Italy. Journal of Geophysical Research 106, 8811e8825.
Goddard, J.V., Evans, J.P., 1995. Chemical changes and uiderock interaction in
faults of crystalline thrust sheets, northwestern Wyoming, U.S.A. Journal of
Structural Geology 17, 533e547.
Hancock, P.L., Barka, A.A., 1987. Kinematic indicators on active normal faults in
western Turkey. Journal of Structural Geology 9, 573e584.
Hedenquist, J.W., Henley, R.W., 1985. Hydrothermal eruptions in the Waiotapu
geothermal system, New Zealand; their origin, associated breccias, and relation
to precious metal mineralization. Economic Geology and the Bulletin of the
Society of Economic Geologists 80, 1640e1668.
Heynekamp, M.R., Goodwin, L.B., Mozley, P.S., Haneberg, W.C., 1999. Controls on
fault-zone architecture in poorly lithied sediments, Rio Grande Rift, New
Mexico; implications for fault-zone permeability and uid ow; faults and
subsurface uid ow in the shallow crust. Geophysical Monograph 113,
27e49.
Hickman, S.H., Barton, C.A., Zoback, M.D., Morin, R., Sass, J.H., Benoit, R., 1997. In situ
stress and fracture permeability along the Stillwater fault zone, Dixie Valley,
Nevada. International Journal of Rock Mechanics and Mining Sciences & Geo-
mechanics Abstracts 34, 414.
Hickman, S.H., Barton, C.A., Zoback, M.D., Morin, R.H., Benoit, R., et al., 1998. In-situ
stress and fracture permeability along the Stillwater fault zone, Dixie Valley,
Nevada. Eos, Transactions, American Geophysical Union 79, 814.
Hickman, S.H., Zoback, M.D., Barton, C.A., Benoit, R., Svitek, J., Summers, R., 2000.
Stress and permeability heterogeneity within the Dixie Valley geothermal
reservoir: recent results from well. Report SGP-TR-165. In: Proceedings of the
25th Workshop on Geothermal Reservoir Engineering. Stanford University,
Stanford, California, pp. 82e85.
Hubbert, M.K., Rubey, W.W., 1959. Mechanics of uid-lled porous solids and its
application to overthrust faulting, [part] 1 of role of uid pressure in mechanics
of overthrust faulting. Geological Society of America Bulletin 70, 115e166.
Jbrak, M., 1997. Hydrothermal breccias in vein-type ore deposits; a review of
mechanisms, morphology and size distribution. Ore Geology Reviews 12 (3),
111e134.
Keller, E.A., Loaiciga, H.A., 1993. Fluid-pressure induced seismicity at regional scales.
Geophysical Research Letters 20, 1683e1686.
Knipe, R.J., Jones, G., Fisher, Q.J., 1998. Faulting, fault sealing and uid ow in
hydrocarbon reservoirs: an introduction. In: Jones, G., Fisher, Q.J., Knipe, R.J.
(Eds.), Faulting, Fault Sealing, and Fluid Flow in Hydrocarbon Reservoirs.
Geological Society, London, Special Publications, vol. 147, pp. viiexxi.
Lee, J.-J., Bruhn, R.L., 1996. Structural anisotropy of normal fault surfaces. Journal of
Lockner, D.A., Tanaka, H., Ito, H., Ikeda, R., Omura, K., Naka, H., 2009. Geometry of
the Nojima fault at Nojima-Hirabayashi, Japan e I. A simple damage structure
inferred from borehole core permeability. Pure and Applied Geophysics 30,
1649e1667.
Lovering, T.G., 1972. Jasperoid in the United States; Its Characteristics, Origin, and
Economic Signicance. U.S. Geological Survey Professional Paper, Report: P
0710, 164 pp.
Lutz, S.J., Moore, J.N., Benoit, D., 1997. Geologic framework of Jurassic reservoir rocks
in the Dixie Valley geothermal eld, Nevada: implications from hydrothermal
alteration and stratigraphy. Proceedings e Workshop on Geothermal Reservoir
Engineering 155, 131e139.
Lutz, S.J., Caskey, S.J., Mildenhall, D.C., Browne, P.R.L., Johnson, S.D., 2002. Dating
sinter deposits in northern Dixie Valley, Nevada; the paleoseismic record and
implications for the Dixie Valley geothermal system. Proceedings e Workshop
on Geothermal Reservoir Engineering 171, 284e290.
Martel, S.J., 1990. Formation of compound strikeeslip fault zones, Mount Abbot
Quadrangle, California. Journal of Structural Geology 12, 869e882.
Micklethwaite, S., 2009. Mechanisms of faulting and permeability enhancement
during epithermal mineralisation; Cracow goldeld, Australia. Journal of
Miller, S.A., Nur, A., Olgaard, D.L., 1996. Earthquakes as a coupled shear stress e high
pore pressure dynamical system. Geophysical Research Letters 23, 197e200.
Miller, S.A., Collettini, C., Chiaraluce, L., Cocco, M., Barchi, M., Kaus, B.J.P., 2004.
Aftershocks driven by a high-pressure CO
2
source at depth. Nature 427,
724e727.
Morrow, C.A., Shi, L.Q., Byerlee, J.D., 1984. Permeability of fault gouge under
conning pressure and shear stress. Journal of Geophysical Research 89 (B5),
3193e3200.
Mort, K., Woodcock, N.H., 2008. Quantifying fault breccia geometry: Dent Fault, NW
England. Journal of Structural Geology 30, 701e709.
Muir-Wood, R., King, G.C.P., 1993. Hydrological signatures of earthquake strain.
Journal of Geophysical Research 98 (B12), 22035e22068.
Muir-Wood, R., 1994. Earthquakes, strain-cycling and the mobilization of uids. In:
Geological Society, London, Special Publications, vol. 78 85e98.
Nemcok, M., Henk, A., Gayer, R.A., Vandycke, S., Hathaway, T.M., 2002. Strikeeslip
fault bridge uid pumping mechanism: insights from eld-based
palaeostress analysis and numerical modeling. Journal of Structural Geology 24,
1885e1901.
Newhouse, W.H., 1942. Ore Deposits as Related to Structural Features. Hafner
Publishing Co., New York, London, 280 pp.
Nur, A., Booker, J.R., 1972. Aftershocks caused by pore uid ow? Science 175,
885e887.
Page, B.M., 1965. Preliminary geologic map of a part of the Stillwater Range,
Churchill County, Nevada. In: Map 28-Nevada Bureau of Mines and Geology.
Parry, W.T., Bruhn, R.L., 1990. Fluid pressure transients on seismogenic normal
faults. Tectonophysics 179, 335e344.
Parry, W.T., Hedderly-Smith, D., Bruhn, R.L., 1991. Fluid inclusions and hydrothermal
alteration on the Dixie Valley fault, Nevada. Journal of Geophysical Research 96
(B12), 19733e19748.
Power, W.L., Tullis, T.E., 1989. The relationship between slickenside surfaces in ne-
grained quartz and the seismic cycle. Journal of Structural Geology 11, 879e893.
Priest, S., 1993. Discontinuity Analysis for Rock Engineering. Chapman and Hall.
Rice, J.R., 1992. Fault stress states, pore pressure distributions, and the weakness of
the San Andreas Fault. In: Evans, B., Wong, T.-F. (Eds.), Fault Mechanics and
Transport Properties of Rocks; a Festschrift in Honor of W.F. Brace. Academy
Press, pp. 475e503.
Rimstidt, J.D., 1997. Gangue mineral transport and deposition. In: Barnes, H.L. (Ed.),
Geochemistry of Hydrothermal Ore Deposits. John Wiley & Sons, New York, NY,
pp. 487e516.
Sagy, A., Brodsky, E.E., Axen, G.J., 2007. Evolution of fault-surface roughness with
slip. Geology 35, 283e286.
Saunders, J.A., 1994. Silica and gold textures in bonanza ores of the Sleeper Deposit,
Humboldt County, Nevada; evidence for colloids and implications for epi-
thermal ore-forming processes. Economic Geology and the Bulletin of the
Scholz, C.H., 2002. The Mechanics of Earthquakes and Faulting, second ed. Cam-
bridge University Press, Cambridge.
Seront, B., Wong, T.-F., Caine, J.S., Forster, C.B., Bruhn, R.L., Fredrich, J.T., 1998.
Laboratory characterization of hydromechanical properties of a seismogenic
normal fault system. Journal of Structural Geology 20, 865e881.
Sibson, R.H., 1977. Fault rocks and fault mechanisms. Journal of the Geological
Society of London 133, 191e213.
Sibson, R.H., 1981. Fluid ow accompanying faulting; eld evidence and models. In:
Simpson, D.W., Richards, P.G. (Eds.), Maurice Ewing Series, no.4, pp. 593e603.
Sibson, R.H., 1986a. Brecciation processes in fault zones; inferences from earth-
quake rupturing. Pure and Applied Geophysics 124, 159e175.
Sibson, R.H., 1986b. Earthquakes and rock deformation in crustal fault zones.
Annual Review of Earth and Planetary Sciences 14, 149e175.
Sibson, R.H., 1990. Conditions for fault-valve behaviour. In: Knipe, R.J., Rutter, E.H.
(Eds.), Geological Society, London, Special Publications, vol. 54, pp. 15e28.
Sibson, R.H., 1992. Implications of fault-valve behaviour for rupture nucleation and
recurrence. In: Mikumo, T., Aki, K., Ohnaka, M., Ruff, L.J., Spudich, P.K.P. (Eds.),
Earthquake Source Physics and Earthquake Precursors. Tectonophysics 211,
283e293.
Sibson, R.H., 1996. Structural permeability of uid-driven fault-fracture meshes.
Journal of Structural Geology 18, 1031e1042.
Sibson, R.H., 2001. Seismogenic framework for hydrothermal transport and ore
deposition. Reviews in Economic Geology 14, 25e50.
Sillitoe, R.H., 1985. Ore-related breccias in volcanoplutonic arcs. In: Sawkins, F.J.,
Sillitoe, R.H. (Eds.), Economic Geology and the Bulletin of the Society of Economic
Geologists, 80. Economic Geology Publishing Company, Lancaster, pp. 1467e1514.
Smith, L., Forster, C., Evans, J., 1990. Interaction of fault zones, uid ow, and heat
transfer at the basin scale. In: Neuman, S.P., Neretnieks, I. (Eds.), Hydrogeology,
vol. 2, pp. 41e67.
Smith, S.A.F., Collettini, C., Holdsworth, R.E., 2008. Recognizing the seismic cycle
along ancient faults: CO
2
-induced uidization of breccias in the footwall of
a sealing low-angle normal fault. Journal of Structural Geology 30, 1034e1046.
Speed, R.C., Armstrong, R.L., 1971. Potassiumeargon ages of some minerals from
igneous rocks of western Nevada. Isochron/West 1, 1e8.
Speed, R.C., 1976. Geologic Map of the Humboldt Lopolith and Surrounding Terrane,
Nevada.
Tanaka, H., Hinoki, S.-I., Kosaka, K., Lin, A., Takemura, K., Murata, A., Miyata, T.,
Shimamoto, T., Fujimoto, K., Wibberley, C.A.J., 2001. Deformation mechanics and
uid behavior in a shallow, brittle fault zone during coseismic and interseismic
periods; results from drill core penetrating the Nojima Fault, Japan. Island Arc
10, 381e391.
Ujiie, K., Yamaguchi, A., Kimura, G., Toh, S., 2007. Fluidization of granular material in
a subduction thrust at seismogenic depths. Earth and Planetary Science Letters
259 (3e4), 307e318.
Unsworth, M.J., Malin, P.E., Egbert, G.D., Booker, J.R., 1997. Internal structure of the
San Andreas Fault at Parkeld, California. Geology 25, 359e362.
Vikre, P.G., 1994. Gold mineralization and fault evolution at the Dixie Comstock
Mine, Churchill County, Nevada. Economic Geology and the Bulletin of the
Vrolijk, P., van der Pluijm, B.A., 1999. Clay gouge. Journal of Structural Geology 21
(8e9), 1039e1048.
Wallace, R.E., Whitney, R.A., 1984. Late Quaternary History of the Stillwater Seismic
Gap, Nevada.
Wibberley, C.A.J., 2002. Hydraulic diffusivity of fault gouge zones and implications
for thermal pressurization during seismic slip. Earth. Planets and Space 54 (11),
1153e1171.
Wilden, R., Speed, R.C., 1974. Geology and mineral resources of Churchill County,
Nevada. Nevada Bureau of Mines and Geology Bulletin 83, 95.
Woodcock, N.H., Dickson, J.A.D., Tarasewicz, J.P.T., 2007. Transient permeability and
reseal hardening in fault zones; evidence from dilation breccia textures. In:
Lonergan, L., Jolly, R.J.H., Rawnsley, K., Sanderson, D.J. (Eds.), Fractured reser-
voirs, 270. Geological Society Special Publications, London, pp. 43e53.
Evolution of cataclastic faulting in high-porosity sandstone, Bassin du Sud-Est,
Provence, France
Elodie Saillet
a,
*
, Christopher A.J. Wibberley
b
a
Geosciences Azur, UMR 6526, 250 Av. Albert Einstein, 06560 Valbonne, France
b
TOTAL EP, CSTJF, Av. Larribau, 64018 Pau, France
Article history:
Received 14 April 2009
9 February 2010
Accepted 17 February 2010
Available online 6 March 2010
Keywords:
Reservoirs
Porous sandstones
Cataclastic Deformation Bands (CDBs)
Bassin du Sud-Est
a b s t r a c t
Cataclastic deformation structures in Cretaceous high-porosity sands in the Bassin du Sud-Est, SE France
were surveyed by scan-lines to examine: (i) the role of tectonic loading path on cataclastic deformation
band (CDB) network development, and (ii) the development of larger ultracataclastic faults as strain
increases. Deformation during PyreneaneProvenal shortening resulted in a persistent high density
(w10/m
2
) of conjugate reverse-sense CDB zones (displacements up to w30 cm), with no generation of
larger faults. Highelow-density undulations occur for each pair of the conjugate set in an alternating
manner, suggestive of network hardening, with a wavelength of several tens of metres being in the order
of mechanical bed thickness. For two study areas which experienced signicant OligoceneeMiocene
extension, a moderate, undulating background density (w4/m
2
) of normal-offset CDBs was recorded,
which became focussed in places into clusters (w50/m
2
) a few metres wide. Thus tectonic loading path
may strongly inuence strain distribution. CDB zones develop by the addition of successive bands at the
edges until, at a thickness of around 5 cm, new bands tend to stray further away from the zone edges.
Coarser sands have thicker CDB zones, suggesting that host grain size, along with mechanical bed
thickness, could be an important contributor to the scale limit in CDB zone growth. Larger ultracataclastic
faults and discrete slip zones localised within or at the edges of some clusters of CDB zones, post-date
cluster development rather than inducing it. This stage of deformation evolution is only reached in
extension, not in shortening, suggesting the infeasibility of achieving the critical state line during
horizontal compression.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
Fluid circulation in the crust, and in particular hydrocarbon
migration in reservoirs, is highly dependant on fault geometrical
and hydromechanical properties (e.g. Manzocchi et al., 1998;
Matthi et al., 1998; Wibberley et al., 2008). Faulting in porous
sandstone often produces zones of deformation bands rather than
planar fracture surfaces (Aydin, 1978; Aydin and Johnson, 1978,
1983; Underhill and Woodcock, 1987; Antonellini et al., 1994;
Davis, 1999; Fossen et al., 2007). Cataclastic deformation bands
(CDBs) are brittle shear zones that form through a combination of
compaction and cataclasis. Porosity and grain size reduction asso-
ciated with CDB formation are thought to cause strain hardening,
further deformation then being accommodated by deformation of
the wall rock, adjacent to the initial band (Aydin, 1978; Aydin and
Johnson, 1978, 1983; Underhill and Woodcock, 1987). Continued
deformation may possibly result in the development of localised
slip surfaces at the edge of deformation band zones (Aydin and
Johnson, 1983; Antonelini and Aydin, 1995; Shipton and Cowie,
2001). Some, but not all of these different eld observations have
been understood through laboratory experiments (Wong et al.,
1997; Mair et al., 2000; Torabi et al., 2007).
Descriptions of such deformation distributions in high-porosity
sandstones are quite varied, ranging from examples of deformation
band localisation as damage zones around larger faults, in relay
zones (often expressed as ladder zones cf. Schultz and Balasko,
2003) or in fault-tip folds, to zones of deformation distributed
over distances much greater than typical damage zone widths (e.g.
w100 m) with no obvious relationship to larger structures (Aydin,
1978; Underhill and Woodcock, 1987; Jamison and Stearns, 1982;
Shipton and Cowie, 2001; Du Bernard et al., 2002a,b; Wibberley
et al., 2007). Yet although recent advances have been made in
understanding the mechanics of yielding to generate a single
deformation band (e.g. Schultz and Siddharthan, 2005; Aydin et al.,
2006; Wibberley et al., 2007), no unifying mechanical model exists
for explaining and predicting distributions of deformation bands in
* Corresponding author. Tel.: 33 492942682.
E-mail address: saillet@geoazur.unice.fr (E. Saillet).
0191-8141/$ e see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jsg.2010.02.007
terms of regional controls such as tectonic loading paths. Further-
more, the evolution of a network, and mechanical controls on this
evolution, are still unclear, particularly with respect to localisation
processes (a) at the scale of a single deformation band e what
controls deformation localisation into a band, and deformation
band growth? (b) at the scale of a cluster of deformation bands edo
they become clustered around previously formed larger faults as
fault damage zones, or do the clusters of deformation bands form
rst, by early deformation localisation, with continued localisation
of deformation generating the larger faults within or at the edges of
these clusters? Some example of these behaviours exist (Fossen
et al., 2000, 2007; Shipton and Cowie, 2001), yet understanding
the mechanical controls on fault distribution is fundamental in
order to provide better fault distribution prediction in reservoir
ow simulations from limited structural input data (Saillet, 2009).
This paper presents a statistical and quantitative eld study
aimed at providing a model of deformation patterns and fault
growth mechanisms in high-porosity sandstones. This study pres-
ents three different cases of Cretaceous high-porosity sands and
sandstones in the Bassin du Sud-Est, Provence, France (Fig. 1). The
eld data were recorded along scan-lines from excellent quarry
exposures from the Orange, Massif d'Uchaux and Bdoin areas
(Fig. 1). These eld data allow us to provide detailed descriptions of
the distribution of deformation and its evolution, leading to inter-
pretations in terms of fault growth mechanisms and fault network
development in high-porosity sandstones. This part of the study
concerns only the geometrical evolution of the deformation. The
impact of deformed structures in sandstones on uid ow, evalu-
ated from permeability and geometric properties of the structures,
will be addressed in a separate publication.
2. Geological setting and data acquisition
2.1. Regional context of the Bassin du Sud-Est
The Bassin du Sud-Est is a triangular region between the Massif
Central to the North West, the Alps to the East, and the Mediter-
ranean Sea to the South. It is a Mesozoic cratonic basin on the edge
of the Alpine orogen, w200 km long and 100e150 km wide. The
total thickness of the sedimentary units is up to 10,000 m in the
central area, but this thickness decreases to 2000e3000 m towards
the edge of the basin (Delfaud and Dubois, 1984). From the Triassic
to the Cretaceous, sedimentary deposits are essentially marine,
corresponding to basin rifting related to Tethys opening. In the
middle Cretaceous, the sedimentary units correspond to detrital
sands deposited during the beginning of basin inversion. From late
to end Cretaceous, the sands were deposited only in a continental
environment, with locally high sedimentary rates (Debrand-
Passard et al., 1984). In the West of the Bassin du Sud-Est, much of
the resulting CenomanianeTuronian deposits are high-porosity
sands and poorly to moderately consolidated sandstones.
This paper presents a combined study of three areas in the
Bassin du Sud-Est, the Bdoin, Massif d'Uchaux and Orange areas
(Fig. 1). These studies were carried out on high-porosity sand and
sandstone outcrops in active and abandoned quarries (Figs. 2e4),
which provide excellent 2-D and 3-D exposures of deformation
band networks and larger faults. These sands/sandstones are
composed of a large range of quartz grain sizes which vary between
the study areas. They also contain a fewclay lamellae and, in places,
limestone beds containing a few shallow marine fossils. These
sands generally have a marine origin from deltaic to beach sands,
Fig. 1. Simplied geological and structural map of the study areas in the Bassin du Sud-Est, Provence, Southeast France. Locations of the three study areas are denoted by boxes.
Modied from Wibberley et al. (2007).
E. Saillet, C.A.J. Wibberley / Journal of Structural Geology 32 (2010) 1590e1608 1591
with some beds of aeolian origin. The sand outcrops show a low to
moderate cohesion between the quartz grains due to a lack of
cement, but can formvertical quarry faces for some months to years
before erosion and outcrop collapse sets in. These sand units have
undergone a low depth of burial, possibly less than 800 m (Delfaud
and Dubois, 1984). Deformation is expressed in these sands and
sandstone units as cataclastic deformation bands (CDBs), larger,
ultracataclastic (mature) faults and discrete slip surfaces
(Wibberley et al., 2007). The orientation and kinematics of these
structures are highly variable, but, fromregional knowledge, can be
attributed to the three different regional tectonic events in Pro-
vence: (i) NortheSouth PyreneaneProvenal shortening accom-
modated by regional foreland thrusting, giving the Cretaceous
outcrop an EasteWest structural trend; (ii) NWeSE Oligocene-
eearly Miocene rifting which caused normal faulting in parts of the
Upper Cretaceous strata; (iii) Miocene left-lateral strike-slip reac-
tivation of some of the major pre-existing NEeSW faults in the
region.
2.2. Deformation features
The deformation features in the Cretaceous high-porosity sands
and sandstones of the Bassin du Sud-Est consist of cataclastic
deformation bands (CDBs) and larger ultracataclastic faults or
discrete slip surfaces. At the outcrop scale the CDBs appear as single
bands with millimetric offsets, usually grouped into zones of
narrowly spaced bands with offsets from a few millimetres to tens
of centimetres (Fig. 5). In most cases, the ne-scale sedimentary
lamellae make it easy to determine apparent offsets. At the
microscopic scale these CDBs are characterized by grain crushing
and compaction (Wibberley et al., 2007). Such grain crushing and
compaction resulted in a signicant decrease in the porosity and
permeability between the host sand and the deformed sand as
previously described elsewhere (e.g. Underhill and Woodcock,
1987; Antonelini and Aydin, 1994; Fowles and Burley, 1994). The
observations of these CDBs in the Bassin du Sud-Est suggest an
evolution with displacement increase and thickness growth of the
structures, similar to previous observations (Aydin and Johnson,
1978; Underhill and Woodcock, 1987; Antonelini and Aydin,
1995). The thinner features are deformation bands with very low
displacements (e.g. <10 mm). These bands correspond to a single
cataclastic fault strand with a thickness of one to a few millimetres.
The SEM observations of these CDBs (Wibberley et al., 2007) show
a preferential occurrence of fractures at grain contact points, sug-
gesting a process of Hertzian fracturing (Gallagher et al., 1974). This
observation also shows that, for a single CDB, this Hertzian frac-
turing process involves a zone typically not greater than ve quartz
grains wide. Scanning electron microscope (SEM) observations
show that the processes of fracture and cataclasis produce a new
generation of very ne grained fragments inlling the porosity
space.
Thicker features are multiple strand zones of cataclastic
deformation bands (e.g. 10e300 mm in displacement, 10e100 mm
wide) formed by generation of successive adjacent single bands,
referred to hereafter as CDB zones. The adjacent bands within
these CDB zones can be generated with a variable spacing. At the
outcrop scale it is not easy to distinguish the different bands within
a CDB zone, except where there are lenses of host sand between the
different bands. Field observations show that the number of indi-
vidual bands in a CDB zone correlates with offset, up to the scale
limit of the deformation bands, around 30 cm offset (Wibberley
et al., 2000). This nding contrasts with studies of the number of
discrete slip surfaces in fault core zones in high-porosity sand-
stones which show no correlation with offset (Shipton et al., 2005).
Larger faults, with offsets greater than around 1 m, are also
present in the high-porosity sands and sandstones of the Bassin du
Sud-Est. These faults appear to have been generated separately from
the smaller-offset CDBs, in some cases during a later tectonic event
Fig. 2. (a) Location of the study area at Orange. Modied from the 1/50,000 geological map of Orange, BRGM. (b) Location of the scan-line within the sandstones from the Quartier
de l'tang quarry.
than those forming the cataclastic deformation bands. These larger
faults present drastic differences in their properties and charac-
teristics compared to the CDBs. On the eld scale, these faults show
a fairly homogeneous deformation zone of ultracataclastic fault
rock, with sometimes one or more discrete slip surfaces in the fault
core, and a cluster of anastomosing cataclastic strands around the
edges of the fault zone (Fig. 6). At the microscopic scale there are
also large differences between these faults and the CBD micro-
textures. SEM photomicrographs (Wibberley et al., 2007) show
some rounded quartz grains of original sedimentary size and
a higher proportion of ultrane material than in CDBs. These
differences seem to reect a transition in fault growth mechanisms
fromCDBs to larger faults, yet there are a relatively small number of
these ultracataclastic fault zones in the study areas. In the Orange
and Bdoin areas, these mature faults are normal faults, post-dating
the earlier reverse-sense CDBs. In the Massif d'Uchaux area they are
large strike-slip faults, cross-cutting the earlier normal-sense CDBs.
Thus, in some cases at least, deformation localisation into
ultracataclastic fault zones is encouraged by the presence of an
earlier set of deformation bands of high density (Wibberley et al.,
2007).
2.3. Field data acquisition
In order to characterise the deformation distribution in each
area, the CDBs were sampled by scan-lines made on each outcrop.
The position of all visible CDBs or larger faults was recorded along
one metre-wide scan-lines (thus structural density, rather than
frequency, is recorded), perpendicular to the strikes of the main
structures. This methodology is similar to at least some of the
previous studies of this type elsewhere (e.g. Du Bernard et al.,
2002b). Because of our rectilinear quarry topography, it is not
necessary to make any statistical correction in the CDB positions on
the outcrop, although density corrections were performed for dip
obliquity between (horizontal) scan-line and non-vertical struc-
tures (Priest and Hudson, 1981). For each CDB, the orientation, the
Fig. 3. (a) Location of the study area in the Massif d'Uchaux. Modied from the 1/50,000 geological map of Orange, BRGM. (b) Location of the different scan-lines within the
Boncava quarry.
thickness, and the number of CBDs in a CDB zone and their offset
were recorded in order to statistically analyse each outcrop. CDB
zones are grouped into a single structure provided that the outer-
most strands are no further from the remaining structure than the
thickness of the remaining structure, and that they converge to
touch the structure. A total scan-line length of 717 m was recorded
for the 3 study areas combined.
3. The distribution of deformation
3.1. Outcrop description
3.1.1. Orange sandstones
The Orange study area is located in the abandoned Quartier de
l'Etang quarry, on the south side of the town (Fig. 2). The outcrop is
composed of different sedimentary units of Cenomanian age. The
majority of the quarry is composed of a cohesive (not disintegrating
to the touch) quartz sandstone unit, mineralogically mature, but
texturally immature, with a high porosity partly due to the small
proportion of quartz cement. This sandstone unit has a marine to
beach origin, and is composed of medium-size quartz grains of
average size of 550 mm. At outcrop, it is possible to observe some
sedimentary channel structures formed in a marine environment,
and in other parts of the outcrop some benthic shallow marine
fossils are present. The sandstone unit is covered by a calcarenite
unit, the contact between these two units dipping around 8
to the
South-West. The deformation of the sandstone unit of the Quartier
de l'Etang quarry consists of conjugate reverse-sense CBDs with
a low angle of dip (34
to 37
) and an ESEeWNW trend thought to

be related to NeS PyreneaneProvenal shortening (Figs. 5, 7a). At
the Orange outcrop, there are also a small number of high-dip
(around 80
) normal CDBs which were formed during a later

tectonic event (Fig. 7a). A mature, larger normal fault (10 m
displacement) is also present. One single scan-line 258 m long has
been recorded from the Orange study area.
3.1.2. Massif d'Uchaux sands
The Massif d'Uchaux study focuses on the excellent exposure
provided in the Boncavai quarry, to the northeast of the town of
Mornas. The study area is made up of different Turonian (late
Cretaceous) sands and sandstones. All the outcrops studied are
situated within the same sand units. These sand units have a very
low cohesion (crumble to the touch) and are essentially composed
of moderate to large quartz grains, with a few clay lamellae. These
incohesive sands are organized in centimetre-scale ning-upwards
units. These observations suggest sand deposition in an inter-tidal
deltaic environment. The Boncavai quarry area is situated on the
south side of an anticline structure of WNWeESE trend (Fig. 3). At
the scale of the quarry, the deposits have a constant dip of a few
degrees to the Southeast. The deformation of the Turonian sands is
essentially expressed as normal-sense conjugate CDBs trending
NNEeSSW corresponding to ESEeWNW Oligocene extension
(Fig. 7b). There is widespread evidence that these CDBs act as
barriers to recent and present-day groundwater ow, because of
the haematite-coloured grain-coating deposits in the sand adjacent
to the faults. This coloration marking the local limits of uid
percolation down through the sand in the vadose zone and/or
Liesegang ring type diffusion, although the mechanism of uid-
rock interaction responsible for this is currently not clear and is the
subject of parallel research. A few low-angle reverse CDBs were cut
by the normal structures, probably corresponding to the Pyr-
eneaneProvenal shortening (Fig. 7b). A mature strike-slip fault is
also present cross-cutting the normal CDBs, probably related to the
regional Miocene strike-slip faulting (Fig. 7b). Data from this study
area were collected from ve different rectilinear scan-lines along
the outcrops, with a total length of 209 m.
3.1.3. Bdoin sands
The Bdoin study area centres on the Le Cros quarry which is still
periodically active, to the North West of the village of Bdoin, South
of Mt. Ventoux (Fig. 4). The Le Cros quarry is composed of a single
sand unit several tens of metres thick, with a low cohesion
(crumble to the touch), of Cenomanian age. The Le Cros quarry
sands have very ne to intermediate quartz grain size and a few
clay lamellae. The deposition of this unit probably occurred in
a beach to marine environment, with occasional evidence for
transition to an aeolian-type deposition. This unit has a constant
low dip (a few degrees) to the South-West. As in the Massif
d'Uchaux area, the CDBs of Bdoin exert an inuence on present-
day groundwater ow, as evidenced by the haematite-red deposits
of grain-coatings adjacent to some of the CDB clusters. The sands of
the Bdoin area showdeformation structures generated during three
different tectonic events. A set of reverse, conjugate low-angle CDBs
Fig. 4. (a) Location of the study area in the region of Bdoin. Modied from the 1/
50,000 geological map of Vaison la Romaine, BRGM. (b) Location of the different scan-
lines within the sands from the Le Cros quarry.
is present, with a broadly EeW to NEeSW trend, formed during
NeS PyreneaneProvenal shortening. This rst fault generation is
cross-cut by a large population of steeply dipping normal, conju-
gate CDBs trending WSWeENE probably corresponding to Oligo-
ceneeMiocene extension (Fig. 7c). This second tectonic event is also
responsible for the presence of a larger, mature normal fault zone
with a localised slip plane and a throw on the order of 10 m. Sub-
vertical CDBs with a NWeSE trend (Fig. 7c) post-date the other
structures and are interpreted as having formed during later strike-
slip faulting during the Miocene. Data were collected from four
different rectilinear scan-lines along the outcrops, with a total
length of 250 m.
3.2. Description of CDB densities
In this sub-section, data on the distribution of CDBs and larger
faults from the outcrops for the three study areas are presented, in
order to investigate the controls on the density of developing arrays
of deformation bands and faults. For each outcrop the CDB density
is dened as the number of CDBs per square metre plotted as
a function of distance along the scan-line.
3.2.1. Orange sandstones
The deformation recorded in the Orange study area corresponds
mostly to one single tectonic event, expressed as low-angle
conjugate reverse faulting. A later extensional event generated only
a small number of normal CDBs and larger mature ultracataclastic
faults. Data corresponding to these reverse CDBs were obtained
along one single scan-line 258 m long. Because of the angle of 40
between orientation-dip of the structures and the scan-line, it is

important to make a correction to the eld-measured density
values (D
eld
). The angle correction (D
corr
) is obtained (Priest and
Hudson, 1981) by using the equation (1):
D
corr
D
field
1=Sin 35
(1)
On the density proles, the D
corr
and the D
eld
are given. The
density proles obtained illustrate three important features
(Fig. 8):
1. The CDB density scan-line corresponding to the total data set of
the conjugate reverse fault population shows a persistently
high density (generally in the range 7e15/m
2
), and a more-or-
less continuous distribution of the deformation along the
outcrop, with a clustering in the South (Fig. 8a). The clusters of
CDBs on the South side of the outcrop correspond to the
presence of ladder zones of very thin CDBs and do not impact
signicantly on the overall distribution of strain.
2. The CDB density variations corresponding to the two different
(opposing dips) sets of the conjugate fault population show
a more heterogeneous distribution of deformation (Fig. 8b, c).
The density prole corresponding to the SW-dipping conjugate
population shows a high CDB concentration in the south,
middle and north of the outcrop but low-density voids
around 100 and 160 m (Fig. 8b). The density prole corre-
sponding to the opposite set (North-dipping) of the conjugate
population shows a high CDB concentration around 100 and
160 m but important voids in the South, in the centre at
around 120 m, and in the North (Fig. 8c). As illustrated by the
vertical dashed arrows in Fig. 8b and c, these voids or troughs
in the NE-dipping set density prole coincide closely with the
Fig. 5. Field appearance of conjugate arrays of cataclastic deformation bands in the Upper Cretaceous sandstones at the Quartier de l'Etang quarry, Orange. a) Detailed eld view of
the alternating chronology between structures of two conjugate sets. (i) Field photo; (ii) interpretative eld sketch. b) Broader eld view of the generally synchronous, but locally
alternating, relationships between two conjugate sets of reverse CDBs and CDB zones. (i) Field photo; (ii) interpretative eld sketch.
positions of high-density zones of the SW-dipping set, and vice
versa, i.e. that there is a spatial correlation between the high
density of one of the conjugate sets and a low density of the
other. A statistical analysis of deformation distribution, based
on the clustering tendency as used in Wibberley et al. (2007),
improved to analyse a range of data bin sizes, shows that each
of the two conjugate sets is much more clustered than the
overall population treated as a single ensemble (Fig. 9).
Furthermore, the cluster analysis is strongest for bin sizes in the
order of w60 m along the scan-line (Fig. 9), converting to
a wavelength of around 30 m when the dip correction is
applied to these low-angle structures, i.e. on the order of bed
thickness.
3. The larger normal fault, on the North side of the outcrop
(Fig. 2), corresponding to a later tectonic event, has no spatial
correlation with any of the reverse structures. It has a throw of
w10 m and is an ultracataclastic fault zone around 75 cmwide
where measured. The microstructures show an intense
decrease of grains and pores size, which can signicantly retard
uid ow (Saillet, 2009). Additionally, Wibberley et al. (2007)
show that larger faults form preferentially in regions which
have already recorded an earlier deformation event.
3.2.2. Massif d'Uchaux sands
The deformation of the Massif d'Uchaux is essentially expressed
by a population of normal-sense conjugate CDBs and later strike-
slip faults. A lownumber of small low-angle reverse-sense CDBs are
present, probably generated by an earlier tectonic event. The
density proles (Fig. 10) corresponding to the different outcrops
show essentially the conjugate and normal CDBs but some strike-
slip faults are also present including a large mature ultra-
cataclastic strike-slip fault which crosses the entire quarry outcrop
(Fig. 3). On each prole the position of this large strike-slip fault is
indicated. The density proles (Fig. 10) illustrate three important
features:
1. The different density proles show similar patterns of spatial
deformation distribution. The most striking feature of this
pattern is the relatively persistent, in relation to precedent
publications (Du Bernard et al., 2002b), yet undulating back-
ground distribution of the deformation over the length of the
45e60 m scan-lines.
2. For all the outcrops there are a few zones which present an
anomalous CDB density increase in addition to the moderate
overall density (Fig. 10). These zones of high-density concen-
tration are ladder zone features such as in the middle of the
outcrop South-1. The anomalous weighting effect of small
ladder zones on density proles has been checked by also
plotting proles of cumulative thickness per square metre (not
presented in this paper for simplicity). In such a case the
outcrop shows a more even deformation distribution because
each ladder zone is a high density of very thin features,
impacting much less on cumulative thickness proles. In other
Fig. 6. Field appearance of relationships between thin CDBs, large clusters and a slip zone in the Upper Cretaceous sands of the Le Cros quarry at Bdoin. a) Field view of the
relationships between earlier thin structures and two larger CDB clusters with strike-slip kinematics. (i) Field photo; (ii) interpretative eld sketch. b) Field view of the relationships
between a cemented slip zone within a large CDB cluster of normal fault kinematics. The eld observations show that earlier CDBs cluster is cut by the slip zone. (i) Field photo;
(ii) interpretative eld sketch.
parts of the outcrop there are increases of CDB density unre-
lated to specic structures such as a larger fault.
3. All the outcrops studied are cut by the same large ultra-
cataclastic strike-slip fault zone (with bedding-parallel stria-
tions) which is systematically correlated with a large CDB
density increase. Nevertheless, it is difcult to determine the
kinematics of these CDBs which are present around the larger
fault and hence infer the tectonic event which generated them.
The CDBs clustered around the main strike-slip fault zone show
normal apparent offset in the vertical outcrop views, but due to
the geometry of the dip of the CDBs and the dip of the stra-
tigraphy they may in reality be normal or strike-slip structures.
Hence the relative timing of background CDBs with respect to
clusters is not inferred by these observations.
All the Massif d'Uchaux scan-lines show relatively similar
deformation distribution patterns and CDB densities. These density
distribution diagrams illustrate three important features: (i) The
overall CDB density is moderate (mean values are given in Fig. 10),
with the predominance of single CDBs and thin CDB zones; (ii) A
few patches present anomalous CDB density increases, where CDB
zones are organized in clusters; (iii) These clusters are not
systematically associated with main slip surfaces or larger ultra-
cataclastic fault zones, but those slip surfaces or fault zones that are
present do exist within or at the edges of clusters of CDB zones.
3.2.3. Bdoin sands
Three different events have been recorded by the outcrops of
Bdoin and are evidenced along each of the four scan-lines recor-
ded in the study area (Fig. 4): the PyreneaneProvenal shortening,
the OligoceneeMiocene extension and the strike-slip faulting
during the Miocene. However, offset along the CDBs and CDB zones
is of a similar magnitude to the CDB zone thickness, and displace-
ment is only visible to the naked eye for the larger features. Hence it
Fig. 7. Lower-hemisphere equal-area stereographic projections (stereograms) of poles to planes for cataclastic deformation bands and larger faults from the studied areas. (a) Data
from the Orange area; (b) data from the Massif d'Uchaux area; (c) data from the Bdoin area.
is very often not possible to determine the kinematics of any one
CDB structure. The density proles obtained on the scan-lines
(totalling 250 m in length) illustrate three important features
similar to those of the Massif d'Uchaux area (Fig. 11):
1. There is generally an overall deformation pattern of undulating
moderate CDB density which constitutes a background
deformation along the outcrops.
2. All the different outcrops from the Bdoin study area show
some patches with an increase in the density of structures
corresponding to a localisation of the deformation within a few
clusters of CDB zones. These clusters are 5e10 m wide with
a CDB density comprised between 30 and 50 structures per
square metre, for a general density average comprised between
6 and 10 structures. These clusters do not generally coincide
with the position of any larger fault, except in one case (scan-
line 3, Fig. 11).
3. Although there are not many larger faults visible at the
outcrops in the Bdoin study area, it is possible to identify two
larger structures. The rst one is situated on scan-line number
3 (Figs. 4 and 11), where it is possible to observe a progressive
increase of the CDB density to a maximum in the centre of the
cluster, beyond which the density drops sharply to dene the
other edge of the cluster. Localisation of a discrete slip surface
occurred at this sharp transition region at the edge of the
cluster. On this rst structure it is difcult to determine the
relative timing of cluster and slip surface because of the poor
condition of the outcrop. The second structure is situated on
the south side of the quarry and was not sampled by any scan-
line because of the orientation of the fault with respect to the
outcrop. Nevertheless, it is also possible in this case to observe
a cluster of CDB zones around a central slip surface with the
same orientation and dip. Given that many clusters exist with no
localised slip planes or larger fault zones, it seems logical to
suggest that the clusters evolved rst, only some of which then
localised deformation into more evolved fault and/or slip zones.
The Bdoin scan-lines show relatively similar deformation
distribution patterns to the Massif d'Uchaux outcrops: (i) The
overall CDB density is undulating but generally moderate, with
Fig. 8. Graphs of CDB densities for the conjugate, reverse structures recorded along the 258 m long outcrop of the Quartier de l'Etang quarry in the Orange area. Density is dened
as the number of CDBs per square metre, directly measured on the eld or with angle correction (see text for more details). (a) The overall sum of the CDBs present in the scan-line
of the outcrop. (b) Density distribution of the SW-dipping CDB set. (c) Density distribution of the N-dipping CDB set. For each graph n is the total number of deformed structures and
D
av
is the mean average density directly measured on the eld (without angle correction). Cf is the cluster factor in each case (see text for details).
mean scan-line averages between 6 and 10 structures per m; (ii) A
few zones of anomalously high CDB density exist (with density of
30e50 structures per m), where CDB zones are organized into
clusters; (iii) These clusters are not systematically associated with
discrete slip surfaces or ultracataclastic fault zones, but ultra-
cataclastic fault zones are sometimes present within the maximum
density of CDB zones. In the Bdoin area the deformation features
correspond to three different tectonic events and in some cases it is
not possible to determine the kinematics or displacement with the
naked eye.
4. Characterisation of individual CDB properties based
on eld data
4.1. Fault thickness data
Previous studies of the relationship between fault zone thick-
ness and fault displacement (Robertson, 1983; Scholz, 1987; Hull,
1988; Evans, 1990) have generally shown that although there is
an overall linear relationship between fault zone thickness and
displacement, it is necessary to take account of various problems
like the denition of the edges of the fault zone or the thickness
variation along a single fault when interpreting data. Fault zone
thicknessedisplacement data are presented in Fig. 12, for single
CDBs, CDB zones and larger ultracataclastic fault zones in the study
area of the sandstones from Orange and the incohesive sands from
Massif d'Uchaux and Bdoin. The data are presented with two
different thickness data measurements. Thicknesses using method
1 are given as the total distance between the outermost edges of the
outermost strands constituting the multiple-band CDB zones. With
this rst measurement method the host sand contained in between
the different strands of a multiple-band CDB zone is included in the
measurement. Thicknesses using method 2 are given as the sum of
the thicknesses of each individual band constituting a multiple-
band CDB zone. This second measurement method denes an
effective thickness of deformed material, taking into account that
material between the bands is undeformed host sand.
There are sufcient similarities between the data fromthe three
study areas to be able to generalize the observations in the
following way:
1. Fault (or CBD) thickness is often equal to (with a thick-
nessedisplacement ratio approximately equal to 1) or smaller
than displacement (with a thicknessedisplacement ratio
between 1 and 0.1), with a general linear relationship despite
this variation.
2. For the larger structures (d >w100e500 mm), a proportional
increase between thickness and displacement does not clearly
exist.
3. Where thickness data were also collected using method 2 or
effective thickness data, these thicknesses recalculated as the
sumof the individual strands showa steady general decrease in
ratio of effective thickness of deformed material to
displacement as displacement increases. In other words, as
displacement increases, additional increments of displacement
result in decreasing amounts of added volume of deformed
material.
4. Larger normal faults (throw>1 m) and CDBs/CDB zones are
plotted on the same graph for comparison. These faults fall in
a different class of structures and they show different micro-
structure and deformation mechanisms. These ultracataclastic
fault zones have much lower thicknessedisplacement ratios
than CDBs and CDB zones.
5. Where clay layers are intersected, the structures have thick-
nessedisplacement ratios at the lower end of the range of data.
6. Where structures were formed in different tectonic events at
the same outcrop, no signicant differences are detectable
between data corresponding to the different events.
4.2. Fault thickness distribution data
Fig. 13 shows the fault thickness distribution data for all of the
outcrops studied in the Bassin du Sud-Est. For each different study
area, all of the structures recorded along the scan-lines are pre-
sented. Data are not divided into kinematic groups, because kine-
matics was not discernable in many cases, and so leaving such data
out would generate articial trends, particularly in the exclusion of
the single CDBs and thinner CDB zones for which offsets were most
difcult to discern with the naked eye. The three different study
areas exhibit the same general pattern, that there is a predomi-
nance of thin CDB zones of small displacement. For example, only
20% of the structures have thickness greater than around 1 cm
(0.6e2 cm depending on the study area). There is a much lower
proportion of larger (wider) CDB zones and ultracataclastic fault
zones than small CDB zones. Differences in the fault thickness
distribution data between the three study areas relates closely to
the initial host sand grain size: the larger host sand grain size has
thicker structures, and the lower host sand grain size has thinner
ones (Fig. 13).
4.3. CDB, cluster and fault zone internal structure
Fig. 14 presents information on the internal structure of CDB
zones in terms of the number of individual cataclastic bands
identiable to the naked eye within a CDB zone in relation to its
thickness. These data are presented separately for the three
different study areas. The data corresponding to the Orange area
have been recorded only using thickness measurement method 1
(Fig. 14a), i.e. the overall width of the CDB zone fromone edge to the
other. Method 2 was not used because it was not possible to
precisely distinguish the limits of the zones of cemented host grains
from the white strands of the deformation band structure with the
naked eye. It was, however, possible to detect strands as whiter
zones and tentatively count them. Data from the Massif d'Uchaux
and Bdoin areas have been presented using both method 1
Fig. 9. Graphs of cluster factor versus bin size using the denition of Wibberley et al.
(2007) adapted for analysing the size of the most likely periodic clustering by
repeating the cluster factor analysis for a range of different bin sizes, displacing the bin
along the scan-line at 1 m intervals. Results are theoretically independent of sample
size, hence the result that each conjugate set has a cluster factor twice as high as the
ensemble of the conjugate fault population (peaking at around 50e60 m bin size)
suggests that the more even distribution of CDBs in the overall population is the result
of summing the more clustered distribution of peaks and troughs of the two conjugate
sets together.
(Fig. 14b(i) and c(i)) and method 2 (Fig. 14b(ii) and c(ii)) thickness
measurements. Thicknesses using method 2 are given as the sumof
the thicknesses of each of the individual bands constituting
a multiple-band CDB zone or individual high-strain zones in an
ultracataclastic fault zone.
The data on the cataclastic bands in the CDB zones corre-
sponding to the three different study areas show the same overall
patterns of increase in the number of strands with increase in
thickness. In general, the CDB zones show a linear dependence of
the number of bands with increase in thickness, suggesting that
CDB zones develop by formation of additional bands successively at
the edges of the structure as deformation continues. Multiple-band
CDB zones with identiable undeformed sand in between the
bands are typically recorded as thicker structures for a given
displacement than those CDB zones with no host sand within the
structure (Fig. 14b(i), c(i)) but when the effective thickness of fault
rock (thickness method 2) is plotted, it seems clearer that the
thickness of fault rock increases as a function of displacement in the
same way for the two types of structures: the best-t relation for
the effective thickness for both types of structure together (Fig. 14b
(ii), c(ii)), being very close to the best-t relation of overall
thickness (method 1) for the CDB zones with no host sand within
them. For the data from Bdoin (Fig. 14c), the CDB zones thicker
than around four centimetres a different trend is apparent:
a disproportionate thickness increase with addition of new bands
(Fig. 14b(i), c(i)) collapses onto a single linear trend when the data
are re-plotted in terms of effective thickness of deformed fault rock
(Fig. 14b(ii), c(ii)). This indicates that CDB zones growing wider than
around 50 mm do so by addition of successive bands further and
further away into the host sand until they are considered as sepa-
rate structures entirely.
The different eld observations show an evidence for a chro-
nology in the development of isolated CDBs, CDB clusters and slip
surfaces or narrow slip zones. Generally, the observations show
that clusters are cut by the slip surfaces and narrow slip zones
(Fig. 6b). The different scan-lines (Figs. 10 and 11) also show that
slip surfaces and narrow ultracataclastic fault zones can exist in the
middle or at the edge of clusters of CDB zones, but they never exist
without clusters. On the other hand, many clusters exist with no
associated slip surface or narrow ultracataclastic zone. These
observations suggest that the clusters of CDB zones formed rst,
some of which then went on to localise the deformation by
Fig. 10. Graphs of CDB densities for the normal and strike-slip structures recorded along scan-lines with a total length of 209 m in the Boncavai quarry in the Massif d'Uchaux area.
Density is dened as the number of CDBs per square metre. The CDB density overall is moderate, with localisation of deformation within clusters of deformation bands. For each
outcrop, n is the total number of deformed structures and D
av
is the mean average density.
generation of discrete slip planes or narrow slip zones of highly
deformed ultracataclasite.
5. Discussion
Spatial distributions and geometric properties of CDBs and
larger faults were measured in three different areas of the Bassin du
Sud-Est along scan-lines of totalling 717 m in length. Systematic
recording of the deformation structures has allowed us to better
understand the mechanisms which control the distribution of
deformation and fault growth in high-porosity sandstones.
5.1. Development of individual CDBs
The systematic recording of the deformation features along the
different outcrops allow us to characterise the deformation in the
high-porosity sands and sandstones and to propose a conceptual
model for CDB growth (Fig. 15). First, our results showa predominance
Fig. 11. Graph of CDB densities for the reverse, normal and strike-slip structures recorded along scan-lines with a total length of 250 m in the Le Cros quarry in the Bdoin area.
Density is dened as the number of CDBs per square metre. As for the Massif d'Uchaux case, the CDB density overall is moderate but variable, with localisation of deformation within
clusters of deformation bands. For each outcrop, n is the total number of deformed structures and D
av
is the mean average density.
of thinner CDB zones, with 80% of all CDB zones having a thickness
less than around a centimetre, as discussed in the preceding section
(Fig. 13). The frequency distribution of thickness/displacement,
skewed towards the narrower structures, is signicant in consid-
ering fault growth. The thinner features grow during continued
displacement with a proportional increase in thickness up to
around 5 cm, by an increase of the number of individual bands
which constitute a CDB zone (Stages 1 and 2 in Fig. 15a and b).
Additional bands become spaced further and further away fromthe
CDB zone, so that beyond around 100 mm thickness, later strands
are far enough away to be counted as separate structures (see
denition in the section on data acquisition). Nevertheless, the eld
data presented (Figs. 12 and 14) suggest an upper limit of CDB zone
thickness of around 0.1e0.2 m, beyond which deformation jumps
to a new band entirely.
There is, on average, a difference in the thicknessedisplacement
ratio between the smaller CDB zones and the larger faults, for all
those structures with displacements above around 0.5 m. This
difference is clearest for the data from Orange (Fig. 12a). For the
smaller features (d <200 mm), there is a proportional increase in
thickness with displacement, albeit with a scatter in thick-
nessedisplacement ratios between 0.1 and 2. The larger structures
(d >0.5 m) showthickness-to-displacement ratios less than 0.2, i.e.
at or below the thicknessedisplacement ratio range of the smaller-
displacement structures (Fig. 12), suggesting that the growth
process does not operate as efciently at larger displacements.
Indeed, the development of ultracataclastic fault rock textures in
the larger-displacement structures gives the impression of even
more localised deformation. Studies showthat ultracataclastic fault
core zones in high-porosity sandstones elsewhere also have much
narrower thickness-to-displacement ratios than associated CDBs
(e.g. Shipton et al., 2005). Where clay is involved, thick-
nessedisplacement faulted clay values have, on average, lower
ratios than the structures without clay, demonstrating that the
Fig. 12. Thicknessedisplacement relationships for CDBs, CDB zones and larger faults in the high-porosity Cretaceous sands and sandstones of Provence. (a) Data from the Orange
area; (b) data from the Massif d'Uchaux area; (c) data from the Bdoin area. Data from the Massif d'Uchaux and Bdoin areas have been recorded based on two different
measurement methods: (i) thickness dened as the distance between the outside edges of the outermost strands (method 1); (ii) thickness dened as the sum of individual strand
thickness (method 2). The cohesion of the host rock seems to have a very important role in the thicknessedisplacement relationships (difference of one order of magnitude between
sands and sandstones), and hence on the micro-mechanisms of deformation.
incorporation of even small amounts of clay into the deformation
bands may preferentially localise deformation or at least partially
inhibit wall rock wear and deformation band widening.
The observation that CDB zones grow by the addition of new
peripheral bands instead of continued deformation on the existing
ones may, at rst sight, be taken as evidence of work hardening (e.g.
Underhill and Woodcock, 1987). Nevertheless, stress perturbations
around the CDB zone which encourage localisation of the new
bands adjacent to the CDB zone, coupled with intragranular
microcracking around the CDB zone as it develops (Wibberley et al.,
2007), suggest that other factors may also play a role, particularly as
laboratory experiments do not showa strength increase of samples
as individual strands form in a developing CDB zone (Mair et al.,
2000). The data corresponding to the larger CDB zones (with
a thickness more than w10 cm) show a diminishing thickness
growth as displacement progressively increases, by a decreasing
role played by the addition of new bands to the CDB zone structure.
Thus the data fromthe different sand and sandstone outcrops show
two different growth mechanisms, which are scale dependent. The
thinner CDB zones grow by systematic addition of new bands,
according to a work-hardening process and/or by stress perturba-
tions encouraging adjacent intragranular fracturing. For CDB zones
wider than around 10 cm however, the data suggest that these
work hardening and/or other processes cease to dominate. The
reason for such a transition may be one or more of the following
possibilities:
1. The larger CDB zones are typically longer e their development
may be limited by bed thickness and mechanical properties of
adjacent beds (e.g. Schultz and Fossen, 2002) as predicted by
numerical modelling of instability development during multi-
layer deformation (e.g. Schueller et al., 2005). However, beds
are often thicker than the height of the outcrop (e.g. several
metres to several tens of metres), making this difcult to
assess.
2. Stress perturbations may vary with CDB zone thickness, length
and displacement gradient. It is possible that wider, longer CDB
zones will have wider zones of stress perturbation, so addi-
tional bands could be generated further and further away from
the original CDB zone to the point where they are no longer
recorded as part of the same single (multistrand) CDB zone.
3. Grain size and grain point contact geometry could be control-
ling factors (e.g. Cundall, 1989), particularly given that smaller
average host sand grain size seems to cause narrower CDBs
(Fig. 13).
5.2. Development of CDB networks
The three study areas show deformation patterns which can be
related to one or more different tectonic events. For the sandstones
from the Orange area, the structures are mostly conjugate reverse
CDB faults corresponding to PyreneaneProvenal shortening. Here,
only a few later, normal faults are present, related to Oligoce-
neeMiocene extension. Evidence of deformation structures corre-
sponding essentially to one single tectonic event at this outcrop is
highly signicant in terms of allowing us to make interpretations of
how this network of structures developed during the single
tectonic event. In terms of the spatial distribution of the CDBs
documented from the scan-line data (Fig. 8), two main interpre-
tations can be made:
1. Firstly, the distribution of the deformation along the outcrop is
relatively homogeneous with a moderately high density of
CDBs and an addition of high density of deformation which
corresponds to a localisation of the deformation in a few
clusters of CDBs such as ladder zones. Such a persistent
density distribution over a transect of c. 250 m is important
because it suggests an overall relatively homogeneous bulk
strain (shortening) accommodated by these CDBs over a rela-
tively large distance, in comparison to previous studies which
mostly suggest CDBs are formed as clusters of CDBs around
larger faults, such as damage zones developed by progressive
displacement along the fault (e.g. Shipton and Cowie, 2001; Du
Bernard et al., 2002b; Johansen and Fossen, 2008).
2. Secondly, there is a general inverse relationship between the
undulating density of one of the conjugate sets and the density
of the other (Fig. 8), with alternating density peaks and troughs
of one set mirroring the troughs and peaks of the conjugate set.
The sum of the two conjugate distributions is a much more
continuous deformation distribution than the clustered
tendency of each of the two conjugate sets individually (Fig. 9).
These eld observations allow us to suggest that deformation
proceeded by development of two conjugate sets of CDBs, of
opposing dips, during the same tectonic event. At the begin-
ning of the tectonic event the deformation is accommodated by
a random distribution of each conjugate set. As deformation
proceeds, further generation of the CDBs of one set is inhibited
in regions where earlier CDBs of the other set are already
present (Fig. 15a). The areas not deformed early on in the
process may experience a higher number of CDBs generated
later. The result of this growth process is that regions with
a high density of faults fromone set have a lower density in the
other, and vice-versa. This CDB growth process by competition
between two opposing conjugate sets during a single tectonic
event results in a broadly homogeneous total (albeit undu-
lating) deformation distribution along the outcrop.
Previous studies have suggested that deformation bands form
by work-hardening processes (e.g. Aydin and Johnson, 1983;
Underhill and Woodcock, 1987). Furthermore, work-hardening
processes at fault tips and relay zones have been evidenced by the
generation of ladder zones between segment tips (e.g. Schultz and
Balasko, 2003; Okubo and Schultz, 2005, 2006). However, our eld
observations show that the work-hardening nature inferred for
CDB generation may inuence the population at the network scale
as well as at the scale of the fault tip, thus causing an increase in the
mechanical strength of the rock mass as well as the individual
deformation band structure. Hence, it is easier for a conjugate set to
develop in a region with low density of earlier deformation struc-
tures. Although overall the two sets are generally synchronous, as
shown by the mutual cross-cutting relationships in different places
Fig. 13. Frequency distribution of CDB zone and fault zone thicknesses for cataclastic
deformation bands and larger faults in the three different study areas of Provence. The
thickness is dened as the distance between the outside edges of the outermost bands
(method 1). The average grain size of the host rock is given for each outcrop by the
symbol f.
(Fig. 5), the preferential development of one set in a certain region
probably depends on it arriving in that region by chance before the
other set. Despite this seeming randomness, it is nevertheless
possible that the spacing of peaks and troughs in the densities of
each conjugate set, in this case around 60 m along the scan-line, so
around 30 m perpendicular to any one set, is related to mechanical
unit thickness, although the base of the unit is below ground level
so we can not assess this.
It was not possible to test this model for conjugate CDB devel-
opment on the other study areas because they have recorded
deformation from two or three superimposed tectonic events, and
in many cases it is not possible to attribute an individual CDB to
a particular tectonic event. Because of this, the statistical clustering
analysis performed on the Orange data (Section 3.2.1) was not
performed on the other data sets as it would not have been
meaningful. However, all the Massif d'Uchaux outcrops expressed
similar patterns. For each outcrop there is a generally moderate
density of deformation structures with the addition of clusters of
CDB zones which localised the deformation (Fig. 10). The same
deformation patterns exist also in the outcrops fromthe incohesive
sands of the Bdoin area (Fig. 11). In these cases, the majority of the
CDBs are normal-sense structures generated during extension. It
therefore seems likely that the difference in clustering tendency
between these outcrops and the distributed nature of the defor-
mation at Orange is simply due to kinematic type of deformation,
with shortening encouraging further generation of CDBs
throughout the volume by system hardening during compression
(Fig. 15a). However, other factors such as differences in burial depth
Fig. 14. Relationship between number of bands and thickness for CDB zones and larger faults in the high-porosity Cretaceous sands and sandstones of Provence. (a) Data from the
Orange area; (b) data from the Massif d'Uchaux area; (c) data from the Bdoin area. Data from the Massif d'Uchaux and Bdoin areas have been recorded based on two different
measurement methods: (i) thickness dened as the distance between the outside edges of the outermost bands (method 1); (ii) thickness dened as the sum of individual band
thickness (method 2). The linear regressions presented on these graphs are the lines of best-t, most of which are constrained to pass through the origin; the equations are given for
each one. For the graphs (c), the second linear regression corresponds to the lines of best-t of the largest structures, which are not constrained to pass through the origin. The
regression in (c ii) is given as a dotted line because of the small number of these structures.
at the time of deformation, which is difcult to constrain here, may
also play a role. It seems likely that differences in tectonic loading
path have had an effect on the spatial distribution of the resulting
deformation.
5.3. Development of higher-displacement mature faults
and slip planes
Previous qualitative and semi-quantitative work on some of the
outcrops presented here found that larger ultracataclastic fault
zones and localised slip surfaces often formed where there was
a high density of CDBs from a previous tectonic event (Wibberley
et al., 2007). The new work expands on the previous study by the
thorough collection of statistical data from scan-lines, such as
spatial variations in density, as well as studying additional
outcrops. It has been found that larger-displacement fault zones
and localised slip surfaces can also form in the same tectonic event
as the CDBs, but in this case they exist within, or at the edge of
clusters of CDB zones. The precise relative chronology of CDBs and
larger faults is difcult to discern at outcrop in most cases, and it is
possible to suggest two different hypotheses for their genetic
relationship: either (a) all the background CDBs formed rst, with
a relatively constant density across the outcrop, followed by
localisation into the clusters; or (b) the CDBs initiated in a few
places, and continued deformation allowed both further develop-
ment into clusters and spreading of the deformation elsewhere to
form the other individual CDBs further away from the clusters.
The fact that all larger-displacement faults occur within regions
of high CDB density, but that many high-density clusters exist
without larger faults, suggests that the clusters formed rst, only
some of which suffered continued deformation localisation and
generation of ultracataclastic faults and discrete slip surfaces
(Fig. 15b, stage 4). In other words, a larger fault is not a prerequisite
for generating peripheral clusters of CDBs (as in a damage zone)
in the study areas examined, but these larger faults localise on pre-
existing cluster zones to accommodate further strain. In fact, in
extensional contexts such as those which generated the structures
studies in the Urchaux and Bdoin areas, it is much easier to form
a slip zone. In the case of reverse deformation, it seems that it is
much more difcult to obtain a localised slip plane or narrow slip
zone e the example of deformation bands accommodating short-
ening at Orange never achieved a state of localisation to evolve into
a large structure; deformation is likely to be transferred into
a different lithological layer before such a high deviatoric stress is
achieved.
The progressive localisation of deformation through time during
extension, from an undulating background CDB density pattern
and the clustering of CDBs through to formation of larger-
displacement ultracataclastic fault zones and initiation of discrete
slip surfaces may therefore be at least partly a function of the
extensional tectonic regime and associated loading path. It is often
tempting to associate such deformation localisation with material
softening of the slip zones, yet in reality localisation is also
controlled by other factors such as system geometric constraints
Fig. 15. Series of block diagrams showing the sequential evolution of deformation from a single band to different sets in a network, with possible localised faulting, during a single
tectonic event. (a) A compressive tectonic event. (Stage 1) Random distribution of the deformation at the beginning of the tectonic event. (Stage 2) Evolution of single CDBs to larger
CDB zones. (Stage 3) Further deformation results in cross-cutting of CDB zones by conjugate structures, but the previous structures inhibit further development of the conjugate set
in places. (Stage 4) As deformation continues, the structural network evolves towards a broadly constant density of conjugate structures. (b) An extension tectonic event. (Stage 1)
Initiation of single CDBs at the beginning of the tectonic event. (Stage 2) Evolution of single CDBs to larger CDB zones. (Stage 3) Localisation of the deformation within some clusters
of CDB zones. (Stage 4) As deformation continues, a slip surface or narrow slip zone grows within a zone of deformation band clustering.
and is not simply a function of material softening (e.g. Hobbs et al.,
1990). In the case of extension, it has been suggested that elastic
unloading of the wall during normal faulting may inhibit the
continued generation of new deformation bands around the fault,
whilst the main fault zone continues to deform. As the faults grow
up-dip and down-dip they will be controlled by other system
parameters such as bed thickness and the mechanical contrast with
adjacent beds, so that eventually the mechanical properties of the
fault will be controlled by system constraints at a larger scale,
possibly with damage zone accumulation contributing to further
fault zone development.
The importance of loading path on deformation structural
evolution can be illustrated by QeP
0
maps (Fig. 16). These plot the
stress paths in terms of differential stress (Q) against effective mean
stress (P
0
) as commonly done in soil mechanics. On such plots, the
range of stress states at the onset of plastic deformation can be
plotted (a yield envelope), as can the elds in which deformation
occurs by compaction, dilation or the critical state line of constant-
volume deformation between the two. Plastic yield envelopes
continuously evolve as the sand is subjected to compaction and/or
tectonic deformation until they reach a clasticeplastic yield
envelope at which point cataclastic deformation bands are gener-
ated (Bolton and McDowell, 1997; Wibberley et al., 2007). For this
reason, several authors have recently used a framework of soil
mechanics to better understand the initiation of cataclastic defor-
mation bands, seen as a phenomenon of plastic deformation which
may be either compactant, dilatant or constant-volume (e.g.
Schultz and Siddharthan, 2005; Aydin et al., 2006; Wibberley et al.,
2007). Deformation accommodating much larger slip than typical
deformation bands, usually by discrete slip surfaces and/or ultra-
cataclastic fault zones, is considered to occur on the critical state
line of isovolumetric deformation. Although there is no a priori
connection between the critical state line (as a range of stress
states) and localisation (as a description of deformation distribu-
tion), microstructural studies of the ultracataclastic fault zones
suggest constant-volume granular ow operated during shearing,
supported by classical soil mechanics experiments after a given
shear strain (e.g. Mandl et al., 1977).
Fig. 16 illustrates four hypothetical loading paths on a QeP
0
map of differential stress (Q) against effective mean stress (P
0
), the
rst path (A) being tectonic compression, the other ones being
different cases of extension (BeD). In all cases, cataclastic defor-
mation bands initiate at the clasticeplastic yield envelope (stage
i), in most cases on the right hand side (cap side) of the Q peak
in the yield envelope, suggesting compaction. The strength of the
CDB itself denes a new yield envelope, being stronger than the
host sand. In the case of shortening, CDBs continue to be gener-
ated until they start to cross each other, at which point network
hardening occurs and the stress path moves up towards the CDB
yield envelope (stage ii on stress path A). A critical density is
achieved at which the bulk stress state lies in fact on the CDB yield
envelope (stage iii). Further increase in stress is required to
continue deformation, but it is uncertain whether this can
continue to the critical state line (Fig. 16) e it is more likely that
failure of an adjacent bed will lead to a thrust fault propagating
into the porous sand bed.
In extension, on the other hand, the exact sequence and type of
structures generated is suggested to depend on the starting
position of the effective tension loading with respect to the clas-
ticeplastic and CDB yield envelopes. In loading path B (Fig. 16), the
path continues to increase differential stress and hence generate
CDBs until, like path A, a critical density is reached (network
hardening). Beyond this, differential stress continues to increase
until the critical state line of the (bulk) deformation band material
is reached at which point faulting occurs by ultracataclasic in
a zone with granular ow, and/or by forming discrete slip surfaces.
Field evidence indicates that slip surfaces in this context localise
within clusters of CDB zones or at the competence contrast
between CDB clusters and host sand. In loading path C (Fig. 16),
the critical state line for the sand is reached as the stress path
Fig. 16. A QeP
0
map of generation of deformation structures in high-porosity sand. CSL indicates Critical State Line. Four different hypothetical stress paths are indicated (labelled
AeD), each producing a different range of structures. See text for details.
continues to move through the compactant regime during
network hardening before the CDB yield envelope is reached,
suggesting that faulting will occur in the sand by localised ultra-
cataclasis. Finally, in loading path D (Fig. 16), the CDB initiation
occurred at greater differential stress than the critical state line
(and on the dilatant side to the left of the yield envelope peak),
suggesting that further deformation will quickly result in defor-
mation localisation in the sand on discrete slip surfaces along with
unloading. This softening effect may be enhanced by non-associ-
ated Coulomb plasticity softening (Mandl, 1988) as described by
Wibberley et al. (2007).
6. Conclusions
The statistical analyses of the spatial distributions and
geometrical properties of cataclastic deformation bands and larger
ultracataclastic fault zones in Cretaceous high-porosity sands and
sandstones from Provence, SouthEast France, allow us to evaluate
the controls on deformation distribution and fault growth mecha-
nisms. The main results can be summarized as follows:
1. CDB density distributions along the Orange outcrop suggest
that deformation was accommodated by two conjugate sets,
each of heterogeneous distribution, during the same tectonic
event. In any one place, where there is a higher density of faults
from the rst set, there is a lower density in the second, and
where there is a lower density of faults from the rst set, there
is a higher density in the second. The result of this fault growth
process is a broadly homogeneous total deformation along the
outcrops which developed by a system-hardening mechanism
during shortening. The frequency of the density undulations of
each of the two conjugate sets may be related to mechanical
bed thickness.
2. The Massif d'Uchaux and Bdoin outcrops have moderate,
undulating densities of mainly normal-sense CDBs associated
with extension. Some zones are present with anomalously
high CDB densities, where CDB zones are organized into
clusters, some of which contain larger-displacement ultra-
cataclastic fault zones or discrete slip surfaces. This distribu-
tion differs from the reverse-sense CDBs at Orange. Thus for
the rst deformation event to strongly affect the outcrop, the
clustering tendency is dependant on kinematics and therefore
tectonic loading path. Thickness growth of CDB zones,
however, is independent of these, but dependent on host grain
size with thicker CDB zones forming in coarser sand/
sandstone.
3. CDB zones grow by a proportional increase of thickness with
displacement due to the addition of new individual bands
within the zone, corresponding to a work-hardening process
at the scale of the zone. As thickness achieves a certain
value on the order of w10 cm, the generation of additional
bands does not continue in proportion to thickness growth.
In these larger CDB zones, the new outer bands are gener-
ated further and further into the host sand away from the
zone. Hardening processes cease to dominate for these
larger CDB zones, and further deformation occurs with
clustering of CDB zones followed in some cases by genera-
tion of the larger-displacement ultracataclastic faults and
discrete slip surfaces which grow with increasing local-
isation of deformation.
Acknowledgements
The rst author was supported by a doctorate student MRT
stipend from the French Ministry for Higher Education and
Research. Access to the active quarries was granted by the oper-
ating company SIFRACO who is gratefully acknowledged. Thorough
reviews by Geoffrey Rawling and Richard Schultz helped clarify
much of the text, and we are also grateful to Zoe Shipton for
editorial guidance.
References
Antonelini, M., Aydin, A., 1994. Effect of faulting on uid ow in porous sandstones:
petrophysical properties. American Association of Petroleum Geologists
Bulletin 78, 335e377.
Antonelini, M., Aydin, A., 1995. The effect of faulting on uid ow in porous
sandstones: geometry and spatial distribution. American Association of Petro-
leum Geologists Bulletin 79, 642e671.
Antonellini, M.A.A., Aydin, A., Pollard, D.D., 1994. Microstructure of deformation
bands in porous sandstones at Arches National Park, Utah. Journal of Structural
Aydin, A., 1978. Small faults formed as deformation bands in sandstones. Pure and
Applied Geophysics 116, 913e930.
Aydin, A., Johnson, D., 1978. Development of faults as zones of deformation bands
and as slip surfaces in sandstone. Pure and Applied Geophysics 116, 931e942.
Aydin, A., Johnson, D., 1983. Analysis of faulting in porous sandstones. Journal of
Aydin, A., Borja, R.I., Eichhubl, P., 2006. Geological and mathematical framework for
failure modes in granular rock. Journal of Structural Geology 28, 83e98.
Bolton, M.D., McDowell, G.R., 1997. Clastic mechanics. In: Fleck, N.A., Cocks, A.C.F.
(Eds.), IUTAM Symposium on Mechanics of Granular and Porous Materials.
Kluwer, Dordrecht, pp. 35e46.
Cundall, P.A., 1989. Numerical experiments on localization in frictional materials.
Ingenieur-Archiv 59, 148e159.
Davis, G.H., 1999. Structural geology of the Colorado Plateau regional of southern
Utah, with special emphasis on deformation bands. Geological Society of
America Special Paper 342.
Debrand-Passard, S., Courbouleix, S., Lienhardt, M.J., 1984. Synthse gologique du
Sud-Est de la France. Mmoire du Bureau de Recherches Geologiques et Mini-
res, (BRGM), 125, Orleans, France.
Delfaud, J., Dubois, P., 1984. Le basin du Sud-Est. In: Gubler, Y. (Ed.), Dynamique des
bassins sdimentaires. Livre jubilaire BRGM, Orleans, France.
Du Bernard, X., Eichhbul, P., Aydin, A., 2002a. Dilation bands: a new form of
localized failure in granular media. Geophysical Research Letters 29, 2176.
doi:10.1029/2002GL015966.
Du Bernard, X., Labaume, P., Darcel, C., Davy, P., Bour, O., 2002b. Cataclastic slip band
distribution in normal fault damage zones, Nubian sandstones, Suez rift. Journal
of Geophysical Research 107, 103e114.
Evans, J.P., 1990. Thicknessedisplacement relationships for fault zones. Journal of
Fossen, H., Odinsen, T., Frseth, R.B., Gabrielsen, R.H., 2000. Detachments and low-
angle faults in the northern North Sea rift system. In: Nttvedt, A. (Ed.),
Dynamics of the Norwegian margins. Special Publications, Geological Society,
London, vol. 167, pp. 105e131.
stone: a review. Journal of the Geological Society of London 164, 755e769.
Fowles, J., Burley, S., 1994. Textural and permeability characteristics of faulted, high
porosity sandstones. Marine and Petroleum Geology 11, 608e623.
Gallagher Jr., J.J., Friedman, M., Handin, J., Sowers, G.M., 1974. Experimental studies
relating to microfracture in sandstone. Tectonophysics 21, 203e247.
Hobbs, B.E., Mhlhaus, H.-B., Ord, A., 1990. Instability, softening and localization of
deformation. In: Knipe, R.J., Rutter, E.H. (Eds.), DeformationMechanisms, Rheology
and Tectonics. Geological Society Special Publications, vol. 54, pp. 143e165.
Hull, J., 1988. Thicknessedisplacement relationships for fault zones. Journal of
Jamison, W.R., Stearns, D.W., 1982. Tectonic deformation of Wingate Sandstone,
Colorado National Monument. American Association of Petroleum Geologists
Bulletin 66, 2584e2608.
Johansen, T.E.S., Fossen, H., 2008. Internal geometry of fault damage zones in
interbedded siliclastic sediments. In: Wibberley, C.A.J., Kurz, W., Imber, J.,
Implications for Mechanical and Fluid-Flow Properties. Geological Society
Special Publication, vol. 299, pp. 34e55.
Mair, K., Main, I., Elphick, S., 2000. Sequential growth of deformation bands in the
laboratory. Journal of Structural Geology 22, 25e42.
Mandl, G., 1988. Mechanics and Tectonics of Faulting: Models and Basic Concepts.
Elsevier, Amsterdam.
Mandl, G., de Jong, L.N.J., Maltha, A., 1977. Shear zones in granular material. Rock
Mechanics 9, 95e144.
Manzocchi, T., Ringrose, P.S., Underhill, J.R., 1998. Flow through fault systems in
high-porosity sandstones. In: Coward, M.P., Daltaban, T.S., Johnson, H. (Eds.),
Structural Geology in Reservoir Characterisation. Geological Society, London,
Matthi, S.K., Aydin, A., Pollard, D.D., Roberts, S.G., 1998. Numerical simulation of
departures from radial drawn in faulted sandstone reservoirs with joints and
deformation bands. In: Jones, G., Fisher, Q., Knipe, R.J. (Eds.), Faulting, Fault
Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London,
Okubo, C.H., Schultz, R.A., 2005. Evolution of damage zone geometry and intensity
in porous sandstone: insight gained from strain energy density. Journal of the
Geological Society of London 162, 939e950.
Okubo, C.H., Schultz, R.A., 2006. Near-tip stress rotation and the development of
deformation band stepover geometries in mode II. Geological Society of
America Bulletin 118, 343e348.
Priest, S.D., Hudson, J.A., 1981. Estimation of discontinuity spacing and tracing
length using scanline survey. International Journal of Rock Mechanics and
Mining Sciences and Geomechanics Abstracts 18, 183e197.
Robertson, E.C., 1983. Relationship of fault displacement to gouge and breccia
thickness. Mining Engineering 35, 1426e1432.
Saillet, E., 2009. La localisation de la dformation dans les grs poreux: car-
actrisation d'un analogue de rservoir faill dans le Bassin du Sud-Est,
Provence, France. Doctorate thesis, University of Nice-Sophia Antipolis,
France, 273 p.
Geology 23, 1825e1844.
Shipton, Z.K., Evans, J.P., Thompson, L.B., 2005. The geometry and thickness of
deformation-band fault core and its inuence on sealing characteristics of
deformation-band fault zones. In: Sorkhabi, R., Tsuji, Y. (Eds.), Faults, Fluid Flow
and Petroleum Traps. American Association of Petroleum Geologists Memoir,
vol. 85, pp. 181e195.
Scholz, C.H., 1987. Wear and gouge formation in brittle faulting. Geology 15,
493e495.
Schultz, R.A., Fossen, H., 2002. Displacementelength scaling in three dimensions:
the importance of aspect ratio and application to deformation bands. Journal of
Schultz, R.A., Balasko, C.M., 2003. Growth of deformation bands into echelon and
ladder geometries. Geophysical Research Letters 30, 2033. doi:10.1029/
2003GL018449.
Schultz, R.A., Siddharthan, R., 2005. A general framework for the occurrence and
faulting of deformationbands inporous granular rocks. Tectonophysics 411, 1e18.
Schueller, S., Gueydan, F., Davy, P., 2005. Brittleeductile coupling: role of ductile
viscosity on brittle fracturing. Geophysical Research Letters 32, L10308.
doi:10.1029/2004GL022272.
Torabi, A., Braathen, A., Cuisiat, F., Fossen, H., 2007. Shear zones in porous sand:
insights from ring-shear experiments and naturally deformed sandstones.
Tectonophysics 437, 37e50.
Underhill, J.R., Woodcock, N.H., 1987. Faulting mechanisms in high porosity sand-
stones; New Red Sandstone, Arran, Scotland. In: Jones, M.E., Preston, R.M.F.
(Eds.), Deformation of Sediments and Sedimentary Rocks. Geological Society,
London, Special Publications, vol. 29, pp. 91e105.
Wibberley, C.A.J., Petit, J.P., Rives, T., 2000. Mechanics of cataclastic deformation
band faulting in high-porosity sandstone, Provence. Comptes Rendus de
l'Academie des Sciences, Paris 331, 419e425.
Wibberley, C.A.J., Petit, J.P., Rives, T., 2007. The mechanics of fault distribution and
localization in high-porosity sands, Provence, France. In: Lewis, H., Couples, G.D.
(Eds.), The Relationship between Damage and Localisation. Geological Society,
London, Special Publication, vol. 289, pp. 19e46.
Wibberley, C.A.J., Yielding, G., DiToro, G., 2008. Recent advances in the under-
standing of fault zone internal structure: a review. In: Wibberley, C.A.J.,
Kurz, W., Imber, J., Holdsworth, R.E., Collettini, C. (Eds.), The Internal Structure
of Fault Zones: Implications for Mechanical and Fluid-Flow Properties.
Geological Society Special Publication, vol. 299, pp. 5e33.
Wong, T.-F., David, C., Zhu, W., 1997. The transition frombrittle faulting to cataclastic
ow in porous sandstones: mechanical deformation. Journal of Geophysical
Research 102, 3009e3025.
Extensional faults in ne grained carbonates e analysis of fault core lithology
and thicknessedisplacement relationships
Eivind Bastesen
a, b,
*
, Alvar Braathen
b, c
a
Centre for Integrated Petroleum Research, University of Bergen, 5020 Bergen, Norway
b
Department of Earth Science, University of Bergen, 5020 Bergen, Norway
c
University Centre in Svalbard, 9171 Longyearbyen, Norway
Article history:
Received 4 July 2009
26 August 2010
Accepted 18 September 2010
Available online 25 September 2010
Keywords:
Thicknessedisplacement
Extensional faults
Carbonates
Fault core
Fault facies
a b s t r a c t
A study of 103 extensional faults hosted by ne grained carbonates in western Sinai, Svalbard and Oman
reveals that faults vary geometrically between simple cores and cores comprising fault splays, lenses,
segment linkages and overlap structures. Fault core rocks are typically carbonate breccias, carbonate and
shale gouge, shale smear, secondary calcite cement and veins, and host rock lenses.
There is a signicant scatter in the core thickness for any given displacement, but the overall pattern is
that the thickness increases with displacement. This increase best ts a power law function (0.29D
0.56
)
that describes a gradual decrease in the thickness/displacement relationship for increasing slip along
faults. In more detail, the general function can be seen as the sum of two (power law) trend lines; the
rst representing thin localized fault cores with generally simple and planar geometry, the second
representing thicker fault cores with complex geometry of lenses and overlap structures and with fault
rock membranes.
Thestudiedfaults showasignicant changeincompositionandgeometryfromsmall (0e1m), tomoderate
(1e10m) andtolarge offset faults (10e400m). Theoverall patternis that fault initiates as fractures lledwith
calcite veins and thin shear fractures that hosts gouge membranes. With increased fault offset, complexity
increases with breakdownof veins, more extensive fault rock membranes, anda trend towards development
of lenses. Whenoffset exceeds 100 m, cores become complex, withmultiple slipzones, cementedbreccia and
shale smear membranes, and various types of lenses. We envision that the fault development as reected by
offset is dominated by forces (extension, compression) acting in the fault, mechanical heterogeneity of wall
rocks, the core lithologies and their developing rheology, and especially geometric effects arising from fault
irregularities.
1. Introduction
Characterization and quantication of fault zones in outcrops is
a fundamental requirement for modelling and forecasting structural
reservoir heterogeneity in carbonate reservoirs. Key parameters for
fault characterization include fault thickness, composition, geom-
etry and displacement (e.g. Yielding et al., 1997; Manzocchi et al.,
1999; Braathen et al., 2009). A fault can be dened as a zone of
focused deformation that can be subdivided into domains/sub-
zones termed core and damage zone(s) (e.g. Chester and Logan,
1986). Alternatively, a fault can be considered an array of hard-
linked and soft-linked fault segments of various scales that affect
a restricted rock volume or fault envelope (e.g. Peacock, 2002;
Childs et al., 2009; Braathen et al., 2009). Descriptions of fault cores
(e.g. Caine et al., 1996; Childs et al., 1996; Lindanger et al., 2007;
Bonson et al., 2007; Wibberley et al., 2008; Bastesen et al., 2009;
Braathen et al., 2009) show a number of recurring elements such
as slip surfaces, fracture/deformation band sets, fault rocks (gouge,
breccias and cataclasites), shale smears, and lenses of protolith or
fault rock. Bulk strain of the core is semi-penetrative to penetrative,
and core elements in most cases exhibit signicantly altered uid
conductivity compared to the host rock fromwhich they are derived.
In contrast, bulk strain in the damage zones anking the core is non-
penetrative and hosts discrete structures including minor faults,
fractures and/or deformation band sets. Studies addressing the
width of the fault zone envelope vs. fault displacement have
revealed a substantial degree of variation and uncertainty (e.g. Hull,
1988; Knott, 1994; Shipton et al., 2006; Childs et al., 2009). In most
cases the thickness/displacement ratio (T/D) shows that thickness
* Corresponding author. Centre for Integrated Petroleum Research, Uni Research,
5020 Bergen, Norway. Tel.: 47 99230248.
E-mail address: Eivind.bastesen@Uni.no (E. Bastesen).
doi:10.1016/j.jsg.2010.09.008
varies by up to three orders of magnitude for a given displacement,
reecting the geometric complexity arising from the presence of
linked and unlinked segments (Childs et al., 2009).
In this paper we present characteristics and scaling laws for
faults in ne grained carbonates. The dataset includes 103 faults
described using the fault facies characterization concept (Braathen
et al., 2009). Fault facies refers to any feature or rock body
deriving its properties from tectonic deformation (Tveranger et al.,
2005), and includes the main elements such as lenses, membranes
and fractures. Fault facies can be characterized in terms of dimen-
sions, geometry, internal structure and petrophysical properties,
thus facilitating quantication, pattern recognition and statistical
handling of structural elements in fault envelopes (Tveranger et al.,
2005; Braathen et al., 2009). In the present study we have dened
fault facies according to composition and geometry. Fault core
geometry and distribution of fault facies are analysed in relation to
the T/D ratio as established for each studied fault.
Fault architecture and related uid ow properties in carbonate
rocks have in the last years received increased attention (e.g.
Agosta and Kirschner, 2003; Cello et al., 2003; Micarelli et al., 2003;
Storti et al., 2003; Labaume et al., 2004; Agosta and Aydin, 2006;
Graham Wall et al., 2006; Bonson et al., 2007; Benedicto et al.,
2008; Bastesen et al., 2009; Putz-Perrier and Sanderson, 2010).
However, studies addressing scaling relationships of faults in such
rocks are scarce, and restricted to faults with small displacements
(Billi et al., 2003; Micarelli et al., 2006; Soliva and Benedicto, 2005),
or case studies (Micarelli et al., 2003; Agosta and Aydin, 2006;
Bonson et al., 2007; Bastesen et al., 2009). In this paper we
present a database of extensional faults from shallow buried
(<2 km) carbonates from three different regions: western Sinai
(Egypt), Central Spitsbergen, Svalbard (Arctic Norway), and Central
Oman (Adams Foothills). The bulk of the data were collected in
western Sinai, whereas data from Spitsbergen and Oman were
collected for comparison purposes (i.e. different tectonic regimes
and/or protoliths). All three areas exhibit thick successions of
sedimentary carbonates which are truncated by well exposed
extensional faults at different scales, with fault displacements
ranging from a fewcentimetres to several hundred metres. In Sinai
(e.g. Moustafa, 2004) and Spitsbergen (e.g. Steel and Worsley,
1984; Maher and Braathen, in press), faulting is related to
regional rifting events, whereas in Oman, extensional faults are
found along the crest of regional anticlines and domes, of which
the folding is controlled by deep-seated thrusting and salt move-
ments (e.g. Hanna, 1990). Protoliths range from massive homoge-
nous limestone to layered heterogeneous shale-rich carbonates
and marls.
2. Terminology
The studied fault cores exhibit several fault core facies associa-
tions (Braathen et al., 2009), which allow an identication and
a description of fault facies using lithology and fault core geometry
as descriptive parameters. Lithologically, fault core facies associa-
tions can be subdivided into shale smear (SS) (Lindsay et al., 1993;
Yielding et al., 1997), carbonate breccia (CB) (Billi, 2005; Micarelli
et al., 2003, 2006), secondary calcite (SCa) (Benedicto et al.,
2008), or composite cores; the latter displaying two or all three
fault core facies associations (Fig. 1).
The fault rocks observed in the present study formed at shallow
depths and at low temperature, and in most cases bear resem-
blance to the primary non-cohesive breccia series described by
Sibson (1977) and Braathen et al. (2004). Clast materials are frag-
ments of carbonate formed by brittle failure. The breccia matrix
consists of either nely crushed limestone fragments in a gouge
(mud fraction fault rock), ne breccia (Billi, 2005) exhibiting
varying degrees of cementation (CB), or shale; the latter giving rise
to shale supported breccias (SCB) (Fig. 1b,c). In this study gouge
was observed as very ne, crushed carbonate material and thin
(<1 mm) membranes of sheared calcareous clay along faults. Shale
smears are shale layers dragged into the fault, aligned parallel to
fault dip and connected to a source shale layer (Lindsay et al.,
1993). Secondary mineral precipitation was observed in the
shape of calcite veins and void llings and as pore space ll in
breccias (described above). In many places precipitated calcite
forms fault-parallel membranes and lenses, displaying the typical
crack-seal vein appearance advocated by Petit et al. (1999). In the
following descriptions the compositional elements are arranged
into fault facies such as lenses (Childs et al., 1997; Gabrielsen and
Clausen, 2001) and membranes (Braathen et al., 2009). Fault len-
ses are elongate pods of host rocks, fault rocks and/or calcite veins
which are completely separated from the surrounding fault
elements by slip surfaces with wall rocks (slip zones) (Fig. 1a).
Membranes are continuous, semi-continuous or patchy fault-
parallel sheets consisting of carbonate breccias, gouge, shale smear
or veins.
Overall fault geometries are classied in Fig. 2. The Type 1
geometry corresponds to faults with a simple geometry, which
further divides into planar geometry (1a) or fault cores that exhibit
jogs or bends of either releasing (1b) or restraining (1c) character.
Secondary shear fractures are classied according to their slip
direction relative to the main fault orientation (Petit, 1987; Bastesen
et al., 2009). These are termed R, P or R
0
shears and correspond to
Type 2 a, b and c, respectively. Type 3 geometries are soft-linked
releasing (3a) or restraining (3b) overlap structures (Rykkelid and
Fossen, 2002; Ferrill and Morris, 2003), while Type 4 geometries
represent breached relays. Fault lenses are assigned to the Type 5.
The Type 6 geometry includes faults with multiple slip surface
zones, several lenses and complex intrinsic composition.
3. Methods
3.1. Database
The database analysed in this study includes a total of 103
extensional faults: 68 from western Sinai, 22 from Oman and 13
from Central Spitsbergen. Fault displacements range from2.6 cmto
400 m, and fault core thicknesses from 1 mm to 11 m. For the
purpose of the present study, faults are termed small when the
displacement is smaller than 1 m, moderate when the displace-
ment is in the range of 1e10 m, and large for the range of
10e400 m. The faults are mainly exposed in 2D cliff sections. For
each fault core thickness, composition and the overall geometry
were recorded (Figs. 1 and 2). The thickness was measured normal
to the dip of the fault. Due to along-fault variations, the thickness
was measured in several places along individual fault outcrops,
covering the maximum and minimum thicknesses of the fault. This
yielded a database of 423 thickness points for a given displacement,
with assigned composition and geometry.
3.2. Fault core thickness
The precise boundaries of the fault core are in some cases
difcult to establish accurately (Childs et al., 2009). In such cases we
dene the core as the part of the fault envelope accommodating the
bulk of the displacement and delimits it by identifying intervals
where sedimentary structures are signicantly displaced (lenses)
or pervasively deformed (brecciated). The core can also consist
entirely of precipitated calcite along one or more fractures (Fig. 1a).
In some cases, the fault core and host rock are separated by a slip
surface, in other cases this transition may be gradual, forming
E. Bastesen, A. Braathen / Journal of Structural Geology 32 (2010) 1609e1628 1610
a fault core-damage zone transition (Billi et al., 2003). Relay
structures and fault bends require additional criteria, in that they
offer increased complexity. Breached relays were measured as
single fault strand; in un-breached relays the two fault strands
were measured individually.
3.3. Fault displacement
Measuring displacement is fairly straightforward, in cases
where the fault displacement does not exceed the height of the
outcrop. For larger faults, with displacements exceeding the height
Fig. 1. a) Schematic fault core with lithologies and geometries typically encountered in small to moderate offset extensional faults in ne grained carbonates. The fault core area is
dened by the bold lines, whereas the surrounding rock is deformed in the damage zone. The upper part of the fault core consists mainly of calcite veins and shale gouge (see inset),
where the calcite veins are truncated by slip surfaces coated with gouge to form sheared calcite veins and lenses. The middle part of the core is dominated by membranes of
carbonate breccias, while the lower part hosts shale smear and associated breccias that are arranged in composite lenses. b) Ternary compositional facies diagram based on three
end members; shale smear (SS), secondary calcite (SCa) and carbonate breccia (CB). Fault cores with combinations of these elements are represented within the diagram. In cases
where all three elements are observed, the fault core is described as a composite facies core that plots in the middle of the diagram. c) Table explaining the abbreviations of the fault
core composition.
of the exposure, local and/or regional stratigraphy and formation
thickness data must be used. Thus stratigraphic offset can be esti-
mated by applying the overall geometry of the fault, such as dip,
position of relay structures, and fault drag. The accuracy of this
method is dependent on the level of stratigraphic details available
and, consequently, the precision of the estimated displacement
commonly decreases with increasing displacement.
4. Geological setting
4.1. Western Sinai
The studied faults are located in carbonates exposed in the
eastern, exhumed ank of the Suez Rift (Figs. 3a and 4). The region
is characterized by large fault blocks (e.g. Hammam Faraoun and El
Qaa) bounded by basement-involved, west-facing extensional
master faults (Coastal fault belt and Eastern boundary fault belt)
with kilometre-scale displacement (Moustafa and Abdeen, 1992;
Sharp et al., 2000; Jackson et al., 2006). These large fault blocks
are broken up by subsidiary faults with maximum displacement of
a few hundred metres. The main period of fault movement, as
recorded by syn-rift deposits, occurred during the Oligocene to
mid-Miocene (e.g. Robson, 1971; Patton et al., 1994; Bosworth et al.,
2005). Faults included in the present database were found in the
Hammam Faraoun and El Qaa fault blocks (Fig. 4a).
Carbonates form a substantial part of the w500 m thick late
Cretaceous to early Tertiary El Egma Group (Moustafa and Abdeen,
1992; Bosworth et al., 2005), which can be further subdivided into
Fig. 2. Geometric classication scheme, showing possible geometries encountered in extensional faults. The diagram spans from planar faults (1a) to fault with complexities; such
as fault bends (1b,c), complex faults with splay structures (2), overlap structures (3), breached overlap (4), lenses (5), and multiple fault strands (6).
the Sudr, Esna, Thebes, Darat and Tanka formations (Fig. 4b). Most
data points for the present study were collected from faults in the
Sudr, Thebes, Darat and Tanka formations. The Thebes Formation
(Said, 1960; Kuss et al., 2000) is a massive, deep water, fossiliferous
limestone, characterized by abundant bands and concretions of
chert and layers of marl. The formation exhibits signicant lateral
variation, both in composition and thickness (Moustafa and
Abdeen, 1992), changing from deep marine, massive micrite
horizons in the north to cherty micritic wackestone in the south
(Moustafa, 2004). The unit exhibits successions of massive lime-
stone and heterogeneous intervals of bedded chert, marl, lime-
stones and 10e50 cm calcareous clay beds. The Thebes Formation
forms a minor reservoir, cap rock and potential source to the
hydrocarbon elds of the central Suez Rift (Alsharhan, 2003). The
Darat Formation consists of a succession of massive chalky lime-
stones and 0.5e2 mthick limestone beds intercalated with 1e5 cm
muddy to clayey limestone and up to 0.5 m thick calcareous clay
units. The Tanka Formation is a brilliantly white, bedded, chalky
limestone with thin layers of aky marl, deposited in a shallow
intertidal environment. This unit is only exposed in the northern
and central part of the Hammam Faraoun block, and is missing in
southern parts of the region due to uplift and rift-shoulder erosion
(Moustafa, 2004).
4.2. Central Oman
The study site is located in the carbonate platform of the
Adams Foothills. The investigated faults are situated in the two
mountain areas separated by a distance of 100 km; the Jebel
Quasaybah and the Jebel Madar (Fig. 3b) (Grlaud et al., 2006).
Jebel Quasaybah is located at the western end of a 70 km long,
EeW oriented anticline (Fig. 3b), whereas Jebel Madar is formed
by a local salt dome (Immenhauser et al., 2007). Outcrops are
found in the hillsides and along valleys cutting into folded and
faulted carbonates of Jurassic to Cretaceous age (Scott, 1990;
Wagner, 1990). The structural conguration reects the early
Tertiary plate-scale closure of the southeastern Persian Gulf
(Hanna, 1990), with major detachment folds located above blind
thrusts located in Paleozoic salt layers (Al-Kindi et al., 2006).
Within the major folds, there are smaller faults, distinguishable
as steeply dipping normal and strike-slip faults and thrust faults.
Faults are also located along the hinges of regional folds associ-
ated with salt diapir-driven exuring (Montenat et al., 2000;
Immenhauser et al., 2007).
Faults at the studied sites are hosted by early to middle Creta-
ceous carbonates (Grlaud et al., 2006). These can be divided into
three units, from base to top: the Shuaiba, Nahr Umr, and Natih
formations (Alsharhan and Nairn, 1997; Grlaud et al., 2006). The
Shuaiba Formation (40e125 m thick) is characterized by massive
limestones consisting of packstones and wackestone beds, whereas
the Nahr Umr Formation (150 mthick) is a thick green shale section
with some thin beds of micritic and marly limestones. The upper-
most Natih Formation (350 m) consists of layered mudstone and
wackstones/packstones (Alsharhan and Nairn, 1997).
4.3. Central Spitsbergen (Svalbard)
The dataset from Spitsbergen was collected from a rift-basin
found in the inner parts of the Billefjorden area (Fig. 3c). Several
phases of tectonic activity have been reported from the basin
bounding master fault system, the NeS trending Billefjorden fault
zone (e.g. McCann and Dallmann, 1996). Of special interest to this
study is mid- to late-Carboniferous to Permian (?) extension
causing the formation of the Billefjorden Trough; a more than
2000 m deep and 30e40 km wide, asymmetric rift-basin lled by
mixed clastic, carbonates and evaporites (Johannessen and Steel,
1992; Maher and Braathen, in press). Syn-rift deposits are found
in the Ebbadalen and Minkinfjellet formations, whereas the Wor-
diekammen Formation constitutes the late-rift succession. The
latter unit reects a transition to a regional, stable platform setting
that prevails in Spitsbergen and the Barents Shelf (Pickard et al.,
1996; Samuelsberg et al., 2003). The faults studied are formed in
the late Carboniferous to Permian Wordiekammen Formation. This
unit consists of 1e10 m thick micrite layers that may be subdivided
into the so-called black crags (Pickard et al., 1996). The crags are
separated by m-thick calcareous shales and wacke/packstones. The
base of the Wordiekammen Formation is in many places charac-
terized by breccia pipes known as the Fortet Breccia Formation.
These pipes represent paleo-karst breccias formed due to collapse
into cavities which were formed due to extensive karstication of
Fig. 3. a). Location of the Sinai study area (star), showing the current plate tectonic setting of the region. Arrows indicate relative plate motion. b) Tectonic map of the eastern part of
the Arabian Peninsula, showing the Hajjar mountain chain and the anticline structures of the study areas of the Adams Foothills in central Oman. The Jebel Madar (east) and Jebel
Quasaybah (west) study areas are denoted with stars. c) Tectonic map of the Spitsbergen island, showing large faults, including the Billefjorden fault zone (BFZ) and the map
distribution of pre-Carboniferous, late Palaeozoic and Mesozoic to Tertiary bedrock units. The star locates the study area.
underlying gypsum in the late Carboniferous Minkinfjellet Forma-
tion (Eliassen and Talbot, 2003).
5. Field data
5.1. Faults in Western Sinai
The majority of the studied extensional faults in western Sinai
are oriented NWeSE, i.e. parallel to the regional structural grain
(Fig. 4c). Some faults oriented NeS and NEeSWwere also observed.
Most faults juxtapose pre-rift carbonates, with a few exceptional
faults that juxtapose the Tanka Formation with syn-rift sedimen-
tary rocks. The majority of fault outcrops were observed in the
Thebes and Darat formations, but some faults juxtapose the Sudr
and/or Esna formations (shale) with the Thebes Formation. The
limestone beds fall into two categories; 1) massive- to bedded-
limestone (Thebes, Sudr and Tanka formations) and 2) interbedded
limestone shale (Darat Formation). The displacement on the
studied faults is listed in Table 1.
5.1.1. Small offset faults
Small faults in the massive limestone beds of Darat, Thebes, Sudr
and Tanka formations have a geometry characterized by straight
and slightly curved fault segments (Fig. 5a,b). Lenses and fault
splays appear in breached fault overlaps and near fault bends. In
most cases the fault core in these carbonates is composed of
membranes of calcite veins and thin (mm) clay gouge along slip
surfaces. Fault cores dominated by veins of crystalline calcite were
identied as a fault core facies in approximately 70% of all studied
small faults in Sinai. Most of these veins are cut by multiple fault-
parallel shear fractures, and are therefore classied as sheared
calcite veins. In many cases, these sheared calcite veins are asso-
ciated with thin clay gouge membranes and calcite gouge
membranes associated with striated slip surfaces. The latter are
commonly positioned along the fault core to host rock boundary.
Fig. 4. a) Geological map of the Sinai eld area, adopted from Moustafa (2004). Red squares indicate areas where faults have been studied. CBF e coastal fault belt, EBFB e eastern
boundary fault belt. b) Simplied stratigraphical column of western Sinai, highlighting the carbonate succession of the El Egma Group. c) Stereographic plot (lower hemisphere, equal
area stereo net) showing the orientation of the studied faults. (For interpretation of the references to colour in this gure legend, the reader is referred to the webversion of this article).
Table 1
Summary of fault core data from the Sinai study area. For each studied fault the displacement, maximum and minimum thicknesses, fault core composition and geometry are
listed. Abbreviations are given in Fig. 1.
Fault
loc.
Displacement
(cm)
Thickness
(cm)
Protolith Fm. Fault
geometry
Fault core
composition
Small offset faults
(>1 m)
Si1 4 0.3e0.7 Calcareous clay-limestone Darat 1a SCa G
Si2 11 0.7e3.3 Calcareous clay-limestone Darat 1a SCa SS
Si3 11 0.7e1.9 Calcareous clay-limestone Darat 5 SCa SS
Si4 2.6e12 0.2e2.5 Bedded limestone Tanka 1c&5 SCa SS
Si5 13 0.5e1.3 Calcareous clay-limestone Darat 1b SCa SS
Si6 14 1.5e2.8 Bedded limestone Tanka 3a, 1c SCa SS
Si7 17 2e2.4 Calcareous clay-limestone Darat 1a SCa SS
Si8 21.5 0.4e5 Calcareous clay-limestone Darat 1a&5 SCa SS
Si9 4e23 1e22 Limestone-clay-chert Darat 1b&3a CB G
Si10 25 0.5e2 Bedded limestone Tanka 1a SCa G
Si11 2.6e27.3 1.3e3.25 Bedded limestone Tanka 1c SCa
Si12 10e29 0.7e1.2 Calcareous clay-limestone Darat 1a SCa SS
Si13 32 1.8e3 Bedded limestone Tanka 1a SCa
Si14 25e33 0.2e2.4 Calcareous clay-limestone Tanka 4 SCa SS
Si15 35 0.3e1.5 Bedded limestone Tanka 2c SCa G
Si16 26e37 2.6e9.5 Calcareous clay-limestone Darat 1b CB SS
Si17 37 9e23 Bedded limestone Tanka 3a CB G SCa
Si18 40 2.8e7.3 Bedded limestone Tanka 1a SCa
Si19 3e41 0.1e1.5 Calcareous clay-limestone Darat 2a & 5 G SS
Si20 38e45 0.1e0.4 Bedded limestone Tanka 1a G
Si21 49 0.2e2.7 Bedded limestone Tanka 1a SCa G
Si22 40e59 1.5e13 Bedded limestone Thebes 4 Composite
Si23 60 1e10 Calcareous clay-limestone Darat 1b Composite
Si24 53e68 0.2e10 Calcareous clay-limestone Darat 2c & 5 SCa SS
Si25 75 2e4.7 Calcareous clay-limestone Tanka 5 SCa SS
Si26 75 0.8e9.5 Bedded limestone Tanka 1a SCa
Si27 80 1.5e9 Calcareous clay-limestone Darat 1a Composite
Si28 80 2e7 Massive limestone Thebes 3a SCa G
Si29 80 0.5 Massive limestone Thebes 1a G
Si30 80 1.2e6.5 Bedded limestone Tanka 1a&5 SCa
Si31 35e95 2.2e4.5 Calcareous clay-limestone Tanka 1a&2a SCa SS
Si32 95 0.4e7 Calcareous clay-limestone Thebes 1b Composite
Moderate offset
faults (1 me10 m)
Si33 100 1.3e12 Bedded limestone Tanka 1b CB SCa
Si34 105 2.9e13 Bedded limestone Tanka 1a&5 SCa SS
Si35 68e109 2.5e12 Calcareous clay-limestone Thebes 1a&5 SS Composite
Si36 3e114 0.2e13.1 Calcareous clay-limestone Tanka 4 SCa SS
Si37 115 2e13 Massive limestone Thebes 5 CB SCa
Si38 70e135 0.1e14 Calcareous clay-limestone Darat 2a & 5 CB SCa SS
Si39 120e140 1e15 Shale-limestone Darat 1a Composite
Si40 140 3e5 Bedded limestone Tanka 1b CB SS Composite
Si41 100e150 2e6 Shale-limestone Darat 2a SS
Si42 150 0.8e10 Massive limestone Thebes 1c, 5 G
Si43 150 0.1e6 Massive limestone Thebes 4 Sca G
Si44 150 1.8e5.1 Bedded limestone Tanka 2a&2b Sca G
Si45 190 2.3e5.8 Calcareous clay-limestone Darat 1a SCa SS
Si46 200 1.1e1.4 Massive limestone Thebes 2c G
Si47 220 1.3e13 Massive limestone Thebes 1a Sea
Si48 250 0.2e6 Massive limestone Thebes 1a CB G
Si49 250 2.6e11.5 Bedded limestone Tanka 1a SCa SS
Si50 100e260 0.6e55 Calcareous clay-limestone Darat 2c & 5 SCa SS G
Si51 150e280 3e50 Massive limestone Thebes 3b CFR
Si52 350 2e7 Limestone-clay-chert Thebes 2b & 5 CB SCa G
Si53 400 2e9.9 Shale-limestone Darat 1a SS Composite
Si54 450 5e50 Calcareous clay-limestone Darat 5 CB Composite
Si55 450e500 4.5e7.8 Bedded limestone Darat 1a Composite
Si56 560 1.3e27 Massive limestone Thebes 2b & 5 CB SCa G
Si57 600 10e25 Limestone-clay-chert Thebes 5 Composite
Si58 600 10.4 Shale-limestone Darat 1a SS
Large offset
faults (<10 m)
Si59 1000 6e28.6 Bedded limestone Tanka 1a&5 CB SCa
Si60 1600 1.4e15 Massive limestone Thebes 5,2a CB G
Si61 2800 15e20 Limestone-clay-chert Thebes 1a Composite
Si62 3000 1e38 Limestone-clay-chert Thebes 1a, 5 CB SCa G
Si63 3500 4e30 Massive limestone Thebes 5 CB SCa
Si64 5000 21e23 Limestone-clay-chert Thebes 5 CB
Si65 23000 70e195 Limestone-chalk-shale Sudr-Thebes 7 CB Composite
Si66 25000 15e110 Bedded limestone Tanka 7 CB
Si67 35000 1100 Shale (Esna) Sudr -Thebes 7 SS
Si68 40000 100e490 Limestone shale Tanka-Thebes 7 CB SS
Calcite vein membranes are generally fairly continuous along
the fault core. Their thickness varies from 0.5 to 13 cm, with the
largest thickness related to fault bends where lens shaped vein
bodies are formed. In most of these lenses the vein lamination
curves similar to the lens shape, and the central part of the calcite
lled lens exhibits open voids with sparry calcite.
In thin section, sheared calcite veins are seemed to consist of
a matrix of crystalline calcite cross-cut by fractures that, in some
Fig. 5. Examples of small and moderate offset faults observed in Sinai, Egypt. a) Fault with 20 cm displacement hosted in marly limestone, chert and chalk layers (Thebes
Formation). The fault has jogs, which are characterized by thick breccia pods next to chert layers. Thin slip zones characterize more planar parts of the fault in the limestone layers.
b) Fault with 1.5 m displacement showing a fault core of calcite veins (SCa) and shale supported carbonate breccia (SCB). c) Mosaic of photomicrographs that displays the
microstructures of a sheared calcite vein collected from a fault with w2 m displacement. The upper part is a cemented gouge (SCa G), neighbouring a zone of calcite crystals (SCa)
cut by numerous curved fractures (slip surfaces). The fractures are associated with thin slivers of gouge and protolith rock, probably derived from shearing along the vein-protolith
contact. The lowermost part is a cemented carbonate breccia (CCB). d) Fault with 4 m displacement cutting through marl and shale prone parts of the Darat Formation. The fault has
a prominent slip surface with well developed shale smears (SS), gouge (G) and breccia (CB) membranes. Note lens formation near the fault bend and the drag of layering in the
footwall and hanging wall.
cases, are coated with a thin (w1 mm) membrane of gouge (Fig. 5c).
Inside the crystalline calcite vein, <1 mm thin lenses/slivers of host
rock limestone can be observed.
Carbonate breccias are commonly found in lenses and typically
consist of coarse, clast-supported breccia that shows extensive
calcite cementation. These lenses are especially common near fault
bends related to chert bands of the Thebes Formation (Fig. 5a).
In shale prone carbonates (Darat and parts of Thebes formations),
small faults appear as bifurcated, splayed and bended (Fig. 5a). In
these faults, lenses occur as partly shattered limestones that
commonly are found along complex fault overlap zones. Shale
smears, together with fault rocks lenses and calcite veins, locally
form composite fault cores. The shale smears show variable
smearing potential, depending on the thickness of the shale proto-
liths. Shale smear factors (Lindsay et al., 1993) are typically around 4.
5.1.2. Moderate offset faults
In faults with moderate offset, calcite veins occur as broken up
lenses and/or juxtaposed with other fault core facies, such as
breccias and shale smears (Fig. 5b,d). Carbonate breccias are more
common compared to small faults, constituting 35% of the core
facies. The breccia matrix mostly consists of calcite gouge that is
cemented, but in the shale prone units breccias are also shale
supported. The carbonate breccias are arranged in semi-continuous
membranes or as highly elongated lenses, bound by slip zones,
often in association with shale smears and calcite veins.
5.1.3. Large offset faults
Faults featuring 10e50 m displacements were only observed in
the massive parts of the Thebes Formation. These are characterized
by relatively thin fault cores (2e50 cm) consisting of membranes of
ne grained (sub-mm clast size) carbonate gouge/breccias, thin
clay gouge, calcite veins and fault rock lenses (Fig. 6a). The lenses
are 2e5 m long and may be 20e50 cm thick, consisting of ne
grained to coarse grained breccias.
In large faults with more than 100 m displacement, shale smears
of the approximately 50 mthick Esna and Thal formations (between
Darat and Tanka formations) are common (Fig. 6b,c). These faults
may be totally dominated by shale smear, and the thickness of
observed shale layers in cores are up to 11 m in large faults juxta-
posing the upper Sudr to the Darat formations. However, common
thicknesses are around 0.5e3 m. In some cases these shale smears
are associated with thick layers of coarse grained fault breccias,
partly shattered limestone and calcite veins along slip surfaces. They
are classied as composite fault cores (Fig. 6b). In larger faults or in
positions far away from the source shale layer of the Esna shale, the
displacement exceeds the smearing potential of the shale. Fault
cores in these cases are dominated by breccia membranes and thin
and patchy lenses/pockets of shale (Fig. 6c). All large faults are
accompanied by a well-dened damage zone consisting of fractures,
small and partly moderate faults, and, locally stylolites.
5.2. Faults in central Oman
Faults studied at the Jebel Madar locality are mostly NeS and
NEeSW oriented, steep and dening horst-and-graben structures
parallel to the salt diapir fold crests (Fig. 7a,b). In this locality the
displacement of the studied faults varies between 10 cmand 300 m,
but may be much larger in other parts of the Jebel Madar (Fig. 7a and
Table 2a). In the Jebel Quasaybah area, data were collected from an
array of extensional faults, dipping 60
e80
, bounding NeS trend-

ing narrowgraben structures. These faults are found in well bedded
Natih Formation units (Fig. 7b). The range of displacements range
fromthis area is between 10 cmand 50 m. Similar to the Sinai faults,
geometry and compositional characteristics vary as a function of the
Fig. 6. a) Photographof fault withca. 25 mdisplacement hosted inthe Thebes Formation.
The fault is oriented NWeSE and dips towards the SW. Details of the fault core are: (i)
Localized fault core with secondary calcite veins and gouge, (ii) distributed slip zones
bounding a protolith lens, and (iii) carbonate fault rock lens. b) Large (w200 m displace-
ment) fault juxtaposing the Thebes and Sudr formations. The fault core has two main
elements; shale smear layers derived fromthe Esna shale and layers of cemented breccias
derived from limestones and cherts of the Thebes Formation. c) Large fault (w250 m
displacement) juxtaposingthe Thebes Formationwiththe Sudr Formation. The upper part
of the exposure is characterized by fault slip zones bounding lenses of carbonate breccia
and host rock. The lower part consists of shale smear derived from the Esna Formation.
protolith carbonate rock. For example, the mud-rich carbonates of
Natih Formation (in Jebel Madar), exhibit faults with thin and
localized fault cores. The faults typically feature a curved and well-
dened slip surface, similar to some of the thin, large faults in the
massive parts of Thebes Formation of Sinai. In contrast, faults in the
massive and stiff limestone layers of the Shuaibah Formation
exclusively host fault cores of massive thick membranes of cemen-
ted breccias and/or sheared calcite veins. Shale smear is not
observed to the same degree as in the faults of Sinai.
5.2.1. Small and moderate offset faults
Calcite precipitations dominate small faults observed in the
Shuaibah and Natih formations of Jebel Qusaybah. They are char-
acterized by sheared veins similar to those studied in Sinai.
Carbonate breccias are formed as discontinuous lenses in small
faults (Fig. 7c) and formsemi-continuous membranes in faults with
more than 2 m displacement. At the Jebel Qusaybah site, the
moderate offset faults are composed of coarse breccias. Typically,
limestone breccia clasts are separated by a pattern of conjugated
Riedel and anti Riedel veins of red/brown coloured calcite (Fig. 7d).
Fault cores with this pattern are characterized by a gradual tran-
sition fromfault core to damage zone, without any clear slip surface
boundary. Conversely, in the Natih Formation of the Jebel Madar
site, moderate to large offset faults appear as narrow zones con-
sisting of discrete slip surfaces associated with a thin fault gouge,
and are rarely associated with calcite precipitations. Carbonate
breccias are commonly discontinuously distributed as lenses or
pockets along the fault core margins.
With increasing slip, the fault cores of the large faults of Jebel
Quasaybahare associatedwithseveral slipsurfaces boundingunits of
ne grained breccias embedded in coarse grained breccias of
red/brown calcite (Fig. 7d). Typically, ne grained breccias appear
along slip surfaces towards the footwall. A second type of large offset
fault consists entirely of slip surfaces hosting crystalline calcite veins.
These fault cores are observed in the Jebel Madar site and have
thicknesses ranging from 50 to 300 cm. And some veins extend for
approximately 100 minstrike direction. The calcite veins are sheared
by numerous parallel and corrugated slip surfaces. This shearing
causes a compartmentalization of sheets with different calcite
textures, from ne grained crushed and brecciated calcite to large
undeformed blocky and acicular crystals that can be 30 cmin length.
Theblockycrystals areorientedperpendicular totheslipsurfaces, and
growinto fault-parallel voids, whereas acicular crystals exhibit radial
growth directions, initiating from protolith clasts. The slip zones are
characterized by sub-mm gouge membranes and thicker zones con-
sisting entirely of crystalline calcite breccias.
5.3. Faults in Central Spitsbergen
The majority of faults are observed in intact, but highly fractured
and semi-brecciated limestone of the Wordiekammen Formation
and Fortet Breccia member. Pressure solution contacts are seen in
both lithologies, indicating a deeper maximum burial of the rocks,
in contrasts to the Sinai and Oman areas. This is expected since the
rocks are older and have undergone several burial and un-roong
events (Steel and Worsley, 1984; Maher and Braathen, in press).
Fig. 7. a) Map of the Jebel Madar area modied from Immenhauser et al. (2007). The
Jebel Madar forms a dome shaped anticline, with associated extensional faults. The
studied localities are indicated by boxes. b) Stereo plots of faults studied in the Jebel
Quasaybah (i) and Jebel Madar (ii) localities, showing a steep NeS orientation of the
faults. c) Fault from the Jebel Madar with 1.20 m displacement hosted in the Shuaiba
Formation. The fault core is dominated by irregular secondary calcite veins and patches
of cemented carbonate breccia and gouge. d) Fault with approximately 25 m
displacement hosted in well bedded Natih Formation at the Jebel Quasaybah site. The
fault core consists of a thick layer of well cemented carbonate breccias, there is a ne
grained breccia close to the footwall, grading into a coarse grained proto-breccia
towards the hanging wall.
The dataset covers faults trending NWeSE, NeS and NEeSW,
which is parallel to the local and/or regional structural grain
(Fig. 8a,b). The faults are located near Pyramiden and at Fortet,
which both host karst breccias. At the Pyramiden site, one master
fault (major segment within the Billefjorden fault zone) cuts
through the entire Wordiekammen Formation, and stratigraphic
correlation suggests a throw exceeding 250 m. Of the 13 faults
studied, most reveal geometries of overlapping and breached
segments, fault splays and lenses (Table 2b).
5.3.1. Small and moderate offset faults
In contrast to the faults in Sinai and Oman, the small faults in
Spitsbergen exhibit breccias and generally lack the extensive
calcite veins. The overall observation is steep, localized fault cores
consisting of carbonate breccias, with a jagged boundary towards
a densely fractured (jointed) carbonate host rock. The fault orien-
tation and core geometry seem well connected with the orienta-
tion of joints. Fault breccias are mostly shale supported, but
bordered by a thin membrane of clay gouge (Fig. 8c). In some cases
there are cemented coarse grained breccias bound by striated slip
surfaces. Stylolites are localized to the contact between breccia
clasts, consistent with local pressure solution and calcite precipi-
tation in breccias. The bulk of the dataset shows fault core thick-
ness variations ranging from 0.1 to 15 cm.
The outcrop of the master fault at the Pyramiden site
(250e300 m displacement) reveals a fault core and a hanging wall
damage zone comprising a 60 m wide zone of brecciated and
intensively fractured limestone. Bedding in both the footwall and
hanging wall dips 20
e40
towards the southeast, increasing

towards the fault core, indicating a normal drag towards the fault.
Extensional faults observed in the damage zone of the hanging wall
are therefore assumed to be rotated, in that extensional faulting
took place prior to or during folding; many of these faults are
described in the small to moderate offset fault section above. The
core thickness of this large fault is approximately 3.5 m. It consists
of a matrix-supported breccia with limestone clasts of 0.1e1 cm, cut
by numerous slip surfaces with mm-thick gouge membranes.
Table 2
Summary of fault core data from the Oman (A) and Spitsbergen (B) study areas. For
each studied fault the displacement, maximum and minimum thicknesses, and fault
core characteristics are listed. Abbreviations are as follows; G-gouge membrane;
SCa-secondary calcite; CFR-carbonate fault rock; SS-shale smear.
Displacement
(cm)
Thickness
(cm)
Protolith Fault
geometry
Fault core
composition
Oman
Om1 18 13e22 Bedded limestone 3b SCa
Om2 40 18 Bedded limestone 1c SCa
Om3 70 5e13.5 Bedded limestone 1a SCa
Om4 70 2e10 Bedded limestone 3a CB
Om5 100 20 Bedded limestone 1b SCa
Om6 100 10 Bedded limestone 1c CB
Om7 100e130 5e38 Bedded limestone 1b&5 CB SCa G
Om9 300 20e25 Bedded limestone 1a SCa SS
Om11 100e350 12e23 Bedded limestone 2a SCa
Om12 400 2e40 Bedded limestone 2c CB
Om13 630 21e29 Bedded limestone 2b SCa
Om14 1000 15e25 Bedded limestone 1a SCa CB
Om16 1500 2e16 Bedded limestone 1a CB G
Om18 2000 170e200 Bedded limestone 7 CB
Om19 2000 75 Bedded limestone 7 CB
Om20 4200 150e170 Bedded limestone 7 CB
Om21 11 500 70e200 Bedded limestone 7 CB SCa
Om22 28 800 400 Bedded limestone 7 CB
Spitsbergen
Sv1 15 6e9 Limestone shale 4&5 CB
Sv2 17 0.1e1.1 Limestone shale 1a CB G
Sv3 22 5e8 Limestone shale 4 CB
Sv4 34 0.5e15 Limestone shale 1a CB SCa
Sv5 45 0.6e13 Limestone shale 4&2a,1b CB
Sv6 50 0.5e8 Limestone shale 2c & 4 CB G
Sv7 80 2e4 Limestone shale 4 CB
Sv8 100 17 Breccia pipe 2a CB
Sv9 130 80 Breccia pipe 2a CB
Sv10 135 0.3e17 Limestone shale 4&5 CB G
Sv11 400 0.3e40 Breccia pipe 5&1a CB G
Sv12 540 7.5e59 Limestone shale 5&1a CB G
Sv13 20 000 350 Collapse breccia
and limestone
7 CB G
Fig. 8. Field localities from the Central Spitsbergen area. a) Geological map of the inner
Billefjorden area, with a simplied stratigraphic legend. b) Stereographic plot of
orientation of faults studied in Billefjorden, indicating a general NeS trend parallel to
the regional grain. c) Fault with 1.35 m displacement that juxtaposes a fossiliferous
grainstone with a micrite. The measuring stick is 1 m. The left inset photograph shows
a complex fault core composed of thick shale supported breccia and gouge
membranes. The right inset photograph shows a localized fault core within
a restraining bend. The fault core is here a thin membrane consisting of a ne grained
shale supported carbonate breccia (SCB).
6. Facies and thicknessedisplacement analysis
In this section we summarize the distribution of fault facies for
the faults found in carbonate rocks of the three study areas. We
then evaluate the thickness/displacement relationships. Finally, we
compare the combined relationships between thickness/displace-
ment vs. geometry and composition.
6.1. Facies analysis
6.1.1. Composition
Of all the fault cores, regardless of eld area and displacement,
34% consist of carbonate breccias, 22% are composed of shale
smears and 20% consist of sheared cement veins. Furthermore, 20%
of the cores have a combination of all the above mentioned
elements (composite core) (Fig. 9a). Only 5% of the faults consist of
gouge; however this element is commonly a minor volumetric
constituent in fault cores dominated by calcite veins or by slip
surfaces in larger faults. Separating the data into different fault
scales reveals that there is a gradual change of fault core facies
from small to large faults.
In the small faults the most common facies combination is that
of calcite veins and gouge (30%). Combinations of calcite veins and
shale smear are also common occupying 28% of the dataset,
whereas fault cores dominated by breccias constitute 21%.
In moderate offset faults a mixture of compositional elements,
such as smeared shale layers, cemented breccias and veins of
calcite, become more common (28%). Also, shale smear is more
frequent, probably related to more prominent thicker shale layers
that are dragged along faults. Furthermore, the calcite veins
become less apparent (13%) and carbonates breccias (35%) become
more common.
Fig. 9. Facies distribution of fault core lithologies and geometries divided into displacement intervals. a) Lithological facies diagram similar to the classes of Fig. 1. with colour coding
b) Pie chart displaying the composition of all faults studied. c) and d) Pie charts displaying the percentage of compositional and geometrical facies for fault offset ranges of 0e1 m,
1e10 m, and 10e400 m.
For large faults, breccias are the most common constituent
occupying 58%. There are fewer fault cores with only shale smear
(11%); however, many of the faults with composite facies elements
(16%) have a signicant part consisting of shale. Cores with calcite
veins in combination with carbonate breccia occupy 16% of the
composition of this displacement range.
6.1.2. Geometry
The analysed dataset shows 30% planar fault segments, i.e.
segments without complexities (Fig. 9b). On the other hand, 70% of
the faults show complexities, ranging from fault splays, to lenses
and to multiple arrays of slip surfaces. Most of the planar fault
segments are found for small faults (37%), whereas this geometry
decreases with increasing displacement, constituting 25% for
moderate offset faults, and 16% of the large faults. The largest fault
with a planar geometry has 50 m displacement. Fault lenses are
observed in 20% of the small faults, whereas for moderate offset
faults lenses comprise 35% of the dataset. In large faults, lenses
again become less common. For large faults, nearly 50% of the cores
show multiple slip surfaces, which corresponds to the entire
dataset of faults with more than 100 m displacement.
6.2. Thicknessedisplacement relationships
From the 103 faults, 423 thickness measurements were made.
Out of these, 334 are fromSinai, 58 fromOman and 31 fromSvalbard
(Fig. 10). The displacement along the observed faults ranges from
4 cm to 400 m, of which the majority (80%) of the faults exhibit
displacements below 10 m. Of these, a signicant number (40%)
have displacements of less than 1 m (Fig. 10a). The thickness of the
fault cores increases with increasing displacement; for displacement
below 1 m the fault cores display thicknesses between 0.1 cm and
23 cm, for moderate offset faults the thickness varies between
0.1 cm and 55 cm, and for large faults the thickness varies between
1 cmand 11 m. For faults with more than 200 mof displacement, the
minimum thickness is 50 cm, and thicknesses above 3 m are always
related to smearing of major shale layers.
The thickness/displacement relationship of all fault core
measurements shows that there is a positive correlation between
thickness and displacement with a linear correlation of
0.01 D 0.067 (Fig. 10b). This trend is validated by a regression
value of R
2
0.45. For the same dataset, a power law line can be
calculated (aX
b
0.29D
0.56
), where a 95% prediction belt indicates
a scatter covering more than three orders of thickness values.
Regression value (R
2
) for the power law line is 0.44. The exponent of
0.56 for the power law equation is consistent with smaller faults
having a relatively highthicknessedisplacement ratio, witha gradual
decrease inthis ratiofor larger faults. Inmore detail, the average plots
of the thicknessedisplacement (T/D) relationship for each of the
studied fault scales showa decrease fromD/32 of small faults withan
averageof 40cmdisplacement, toD/81for moderateoffset faults with
anaverageof 2.3mdisplacement, andtoD/267for largefaults withan
average of 79 m displacement (Fig. 10c).
6.3. Facies combinations and thickness/displacement relationship
By combining the thickness/displacement data with the compo-
sitional and geometrical information we can go deeper into the fault
core characteristics (Fig. 11 and Table 3). In this analysis we use
arithmetic graphs to further investigate the relationship between
thickness/displacement and facies. These graphs are divided into
small (Fig. 11a,b), moderate (Fig. 11c,d) and large offset faults
(Fig. 11e,f). The combination of geometrical and compositional facies
is then compared (Table 3), for example by the average thickness
value for each displacement range that represents a reference value.
Fig. 10. Displacement vs. thickness data from the three study areas. A) Histogram showing the distribution of the displacement in intervals covering all the analysed faults. B)
Thicknessedisplacement diagram with logarithmic axes. Dashed lines illustrate orders of thickness/displacement relationships. The regression lines are displaying the 95%
prediction belt with basis in the best t power law trend line. Note that the condence belt shows a scatter of data over three orders of magnitude for any given displacement. C)
Average displacement against average thickness to displacement ratio of three displacement ranges (0e100, 100e1000, 1000e40000 cm). This plot suggests that there is a gradual
decrease in thickness to displacement ratio for increasing displacement.
Faults withthickness values abovetheaveragearetermedthickfaults,
whereas faults with values beloware termed localized faults.
For the small faults (Fig. 11a,b) in the range of 4 cme100 cm
displacement, there is a poor correlation between the thickness and
the displacement, as indicated by the high scatter in data points
(Fig. 11a). The thickness range is from 1 mm to 23 cm. There is also
large variability of geometrical and compositional facies. However,
combinations such as planar faults with calcite veins and gouge are
far more frequent (79%) in the localized faults (thickness below
3.8 cm) than any other combination. For thick fault cores (above the
average thickness of 3.8 cm) the dominant facies combinations are
carbonate breccias and calcite veins located to releasing bends, and
fault splays often having a lens shape. Thick planar cores are
dominated by smears of relatively thick shale layers and signicant
accumulations of calcite veins.
Moderate offset faults (Fig. 11c,d) ranges in thickness from 1 mm
to 80 cm with an average of 11 cm. The thickness distribution is
skewed towards lower displacements, and there is an overall poor
correlation in the relationship between thickness and displace-
ment. Localized faults consist most commonly of planar segments
Fig. 11. A) Arithmetic displacement vs. thickness plotted using the same data points as shown in Fig. 10. The points are denoted with the lithological and geometrical characteristics,
as observed in outcrops. For analysis purposes the plots are divided into small (a,b) (0e1 m), moderate (c,d) (1e10 m) and large offset (e,f) (10e400 m) faults. Dashed lines in the
four lower plots indicate the level of average thickness and the transition between planar localized fault cores and thick complex fault cores. For further description see the text.
with calcite veins, gouge and shale smear (59%). Another trend is
that planar faults hosting breccia membranes and composite facies
become more frequent (16%) than for small faults. For the thick
faults, lenses consisting of calcite veins, breccias and host rocks
combined with composite facies are the most common constitu-
ents (38%), whereas the combination of planar fault segments with
breccias, calcite veins and shale smear occupies 28% of the dataset.
The large faults are analysed in two groups faults with 10e100 m
displacement and major faults with more than 100 mdisplacement
(Fig. 11e,f). The reason for this separation is that there is a clear
difference in thickness and composition between these two groups,
with an average thickness of 43 cm for the faults between 10 and
100 m, and 223 cm for faults with displacement between 100 and
400 m. The separation between thick and localized faults of the
Table 3
Table showing the combination of geometrical and compositional data of all the measured thickness points. The data is shown in the plots of Fig. 11, where the thickness/
displacement data is divided into small, moderate and large offset faults. The small faults have an average thickness of 3.8 cm; faults thicker than this are termed thick faults,
while faults with lower thicknesses are termed localized faults. The same principle is used for the moderate offset faults, where the average is 11 cm. Few measurements from
large faults make discrimination between localized and thick cores uncertain. Instead this group is divided into faults of 10e100 m displacement and faults of 100e400 m
displacement.
N % N %
Small offset faults (N [201)
Localized fault cores (T < 3.8 cm) Thick fault cores (T > 3.8 cm)
Planar Breccia 3 2% Planar Breccia 2 4%
Planar Gouge 26 19% Planar Cement veins 9 18%
Planar Cement veins 61 45% Planar Shale smear 6 12%
Planar Shale smear 20 15% Releasing bend Breccia 11 22%
Releasing bend Gouge 2 1% Releasing bend Gouge 1 2%
Releasing bend Shale smear 1 1% Releasing bend Cement veins 6 12%
Releasing bend Cement veins 1 1% Releasing bend Shale smear 2 4%
Restraining bend Gouge 2 1% Releasing bend Composite 2 4%
Restraining bend Cement veins 3 2% Restraining bend Cement veins 1 2%
Fault splay Cement veins 4 3% Fault splay Breccia 1 2%
Fault splay Shale smear 2 1% Fault splay Cement veins 4 8%
Overlap Breccia 1 1% Fault splay Shale smear 1 2%
Overlap Cement veins 1 1% Overlap Breccia 2 4%
Overlap Shale smear 3 2% Overlap Lenses 1 2%
Segment link Breccia 2 1%
Segment link Gouge 1 1%
Segment link Cement veins 1 1%
Lenses Shale smear 2 1%
Lenses Composite 1 1%
Moderate offset faults (N [ 160)
Localized fault cores (T < 11 cm) Thick fault cores (T > 11 cm)
Planar Breccia 9 9% Planar Breccia 4 8%
Planar Cement veins 37 37% Planar Shale smear 1 2%
Planar Shale smear 12 12% Releasing bend Cement veins 2 4%
Planar Composite 6 6% Fault splay Breccia 9 18%
Fault bends Breccia 2 2% Fault splay Cement veins 2 4%
Fault bends Gouge 2 2% Overlap Shale smear 1 2%
Fault bends Composite 1 1% Overlap Composite 2 4%
Fault splay Breccia 1 1% Overlap Host rocks 1 2%
Fault splay Gouge 2 2% Lenses Breccia 4 8%
Fault splay Cement veins 3 3% Lenses Composite 6 12%
Fault splay Shale smear 1 1% Lenses Host rocks 8 16%
Fault splay Composite 2 2% Lenses Cement veins 1 2%
Segment link Cement veins 1 1%
Segment link Shale smear 2 2%
Lenses Breccia 1 1%
Lenses Gouge 2 2%
Lenses Calcite 2 2%
Lenses Composite 2 2%
Lenses Host rocks 2 2%
Large offset faults (10e50 m disp.) All large faults (N [60)
Planar Cement veins 7 16% Planar Breccia 5 8%
Planar Breccia 5 11% Planar Gouge 6 10%
Planar Composite 1 2% Planar Composite 1 2%
R splay Breccia 2 5% R splay Breccia 2 3%
R splay Breccia 4 9% R
0
splay Breccia 4 7%
Lenses Breccia 9 20% Lenses Breccia 9 15%
Lenses Host rock lens 5 11% Lenses Host rocks 5 8%
Multiple slip Breccia 5 11%
Major faults 100e440 disp
Multiple slip Breccia 9 56% Multiple slip Breccia 14 23%
Multiple slip Cement veins 3 19% Multiple slip Cement veins 3 5%
Multiple slip Shale smear 2 13% Multiple slip Composite 2 3%
Multiple slip Composite 2 13% Multiple slip Shale smear 2 3%
entire dataset is inuenced by this difference; smaller displace-
ment large faults occupy the domains of localized faults, whereas
the larger displacement faults occupy the thick fault domain. Faults
between 10 and 100 moffset have a distributed facies combination;
with 41% of the dataset showing planar faults with breccias, gouge
and calcite veins. Faults with lenses of breccias and host rocks
occupy 31% of the dataset. The remaining data points consist of
breccia membranes cut by multiple slip surfaces, or faults
geometrically inuenced by R splays.
The outcrops of the largest offset faults all display multiple
slip surfaces. Faults with a composition of breccias are by the far
most common (56%). The remaining faults are composed of
composite facies (13%), shale smear (13%), and calcite veins (19%).
Fault cores dominated by shale smear occupy the largest thick-
nesses (3.5 and 11 m).
7. Discussion
7.1. Displacementethickness relationships
The presented thickness/displacement dataset has a signicant
scatter of data points (Figs. 10 and 11), revealing up to three orders
of magnitude of thicknesses for any given displacement. A similar
variation in the thickness/displacement relationships is commonly
observed and has been related to varying protoliths, mechanical
layering, depth of deformation, fault zone architecture (lenses and
anastomosing fracture network), strain widening and localization
processes (hardening vs. softening), and observation criteria
(Blenkinsop, 1989; Evans, 1990; Shipton et al., 2006; Wibberley
et al., 2008; Childs et al., 2009). This variation in thickness offers
signicant uncertainty for example in the selection of input
parameters for reservoir fault seal assessments (Yielding et al.,
1997; Manzocchi et al., 1999, 2010; Sperrevik et al., 2002).
We address this variation by further classifying faults into
different compositional and geometrical facies, by that exploring
some factors controlling the thickness/displacement relationship.
Firstly, when comparing small, moderate and large offset faults, the
increase in thickness guided by displacement is lower for large
faults than for small faults (Fig. 10c). There is also a clear difference
in the average thickness/displacement ratio for these three
displacement ranges. Contrary to this, most publications address-
ing thickness/displacement relationships of faults indicate a linear
increase in thickness guided by displacement, i.e. the thickness/
displacement relationship is similar for all fault scales (e.g. Otsuki,
1978; Robertson, 1983; Scholz, 1987; Hull, 1988; Walsh et al., 1998),
which has been ascribed to a constant growth of the fault core
controlled by the fault rock rheology and strain hardening (Scholz,
1987; Hull, 1988). Our conclusion is that a power law correlation
with an exponent of approximately 0.6 can best represent the
general thickness to displacement relationship of fault cores in ne
grained carbonates. This trend line agrees with similar datasets,
such as that of Shipton et al. (2006), Braathen et al. (2009), Balsamo
and Storti (in press), and an unpublished study from fault core
thickness in sandstone of Sinai (Braathen and Skar, pers. comm.).
The variation in thickness for any given displacement (Fig. 10) can
be somewhat reduced by applying two different trend lines; one
line representing fault cores with a simple planar geometry and
a related simple composition, and a second trend line representing
complex, thicker fault cores commonly associated with fault jogs or
relay structures, or larger displacement (Fig. 12).
7.2. Fault styles and displacement
A number of studies address fault style in various sedimentary
rocks of varying degree of lithication (e.g. Childs et al., 1997;
Heynekamp et al., 1999; Agosta and Kirschner, 2003; Storti et al.,
Fig. 12. Schematic representation of stages and mechanism for development of faults in carbonates, based on literature and observations from the three study areas. (i) Increase in
thickness due to mechanical hardening of fault rocks related to crushing and frictional wear. (ii) Increase in carbonate fault rock thickness due to linkage, breaching and breakdown
of asperities and relay structures. (iii) Increase in calcite vein thickness due to repeated dilational fracturing and precipitation of calcite cement. (iv) Thickening of the core by
incorporating several shale layers.
2003; Myers andAydin, 2004; Labaume et al., 2004; Vander Zee and
Urai, 2005; Agosta and Aydin, 2006; Bonson et al., 2007; Benedicto
et al., 2008; Braathenet al., 2009; Bastesenet al., 2009). According to
Braathen et al. (2009), many of the fault characteristics are recurring
and therefore are globally representative, with the distinctive fault
zone elements represented by discrete structures, membranes and
lenses. Herein, we address the spatial patterns and distribution of
these elements in fault cores of ne grained carbonates.
Small offset faults (0,1e1 m) commonly have thin and localized
fault cores along straight segments, with membranes consisting of
calcite veins, gouge along shear fractures, and shale smear (Figs. 9
and 11). They often reveal linkage of segments where the fault
core becomes thicker, but these irregularities show a similar
composition as found along the localized segments.
For moderate offset faults (1e10 m), composite cores become
more common. Such cores are dominated by both host rock and
fault rock lenses surrounded by slip surfaces or slip zones offering
both gouge and breccias (Figs. 9 and 11).
For larger faults, a transition exists between 50 m and 100 m of
displacement. Faults with 10e50 m offset consist of relatively thin
localized membranes of sheared veins and ne grained breccias,
and with host rocks lenses similar to moderate offset faults. Larger
faults (>100 m offset) show thicker cores with multiple slip zones,
hosting signicant breccias and shale membranes (Figs. 9 and 11).
Further, these cores show extensive calcite precipitation as cement
in breccias and in voids, and a prominent development of lenses.
In summary, for fault cores in ne grained carbonates the overall
pattern is that of fault initiation dominated by fractures lled with
calcite veins and thin shear fractures hosting gouge membranes
(Fig. 13). With increased fault offset, complexity increases with
breakdown of veins, more extensive fault rock membranes, and
a trend towards development of lenses. When offset exceeds 100 m,
cores become complex with multiple slip zones, breccia and shale
smear membranes, and with lenses. There is also extensive calcite
cementation of rocks in the core. This general pattern reects the
development of the fault with repeated activation during accu-
mulation of offset, as discussed below.
7.3. Core development related to fault displacement
Both the temporal and spatial development of fault cores, and
the related link to facies composition, is intricate and therefore only
possible to discuss in general terms. The observed development in
style of extensional fault cores of ne grained carbonate rocks
seems inuenced by the interplay of e (i) calcite dissolution and
precipitation, (ii) disaggregation of rocks (rheology), (iii) shale
smearing, and (iv) fault geometry (Figs. 1, 2 and 13).
The dominance of calcite veins in combination with gouge and
shale smear observed in the small and moderate offset faults is
likely related to the crack seal-slip-mechanism, which may be
ascribed to over pressurized uids in fractures and small faults
(Petit et al., 1999). This mechanism is believed to occur in
mechanically strong host rocks, where evidence of multiple phases
of calcite vein precipitation suggests recurring fracturing and
subsequent re-sealing. Repeated slip-surface activation appears to
have taken place along the boundary between the host rock and the
calcite vein(s), as evidenced by the millimetre thick lenses of host
rock limestone and fault gouge surrounded by a crystalline calcite
matrix, which are observed along the slip surface. Our study
supports Petit et al.s (1999) conclusion that these veins become
less common with increased fault slip, which could mimic higher
uid permeability in larger offset faults. This is based on the
observation that the typical crack-seal related veins are common
for small faults. For larger faults, the veins become shattered, and
lenses of breccia and overlap structures lled with secondary
calcite become more common. However, there are contradicting
cases; in the Jebel Madar (Oman) area extensive calcite minerali-
zation was found along one of the large faults and similar miner-
alization is also described for faults and karst caves in the northern
part of the Jebel Madar (Montenat et al., 2000; Immenhauser et al.,
2007). This extensive calcite precipitation is assumed to be recent;
in that dating of calcite from the latter study suggests a Pleistocene
age. The secondary calcite is believed to have formed due to mixing
of deep-seated uids with meteoric uids (Immenhauser et al.,
2007). This suggests recent fault (re-)activation and that these
faults formed major uid pathways during their active phases (see
also, Roberts and Stewart, 1994 and Bastesen et al., 2009).
Another factor affecting the precipitation of cement is the depth
of faulting, which controls the degree of pressure solution (Carrio-
Schaffhauser and Gaviglio, 1990; Peacock and Sanderson, 1995;
Micarelli et al., 2005; Benedicto et al., 2008). Pressure solution
may localize the fault zone, by dissolving carbonate material and
leaving insoluble shale gouge along the fault surface (Micarelli
et al., 2005; Benedicto et al., 2008). In this study, stylolites are
observed between clasts of the core but not along slip surfaces.
Further, for large faults, stylolitic surfaces are observed in the
damage zones. The fact that many faults experience a high degree
of calcite precipitation associated with shale gouge could also be
related to pressure solution around insoluble clay. This may explain
why carbonates with low shale content have fault cores with thin
membranes of shale gouge along slip surfaces, together with
sheared calcite veins.
Childs et al. (2009) advocate the link between fault develop-
ment and fault zone thickness. They show that fault zone width
varies from relay structures to fault cores, with the range of thick-
nesses encountered for relays observed to be as much as seven
orders of magnitude higher than for fault cores. Similarly, the fault
core thickness is distributed over three orders of magnitude (e.g.
Knott, 1994; Sperrevik et al., 2002). To explain this scatter Childs
et al. (2009) suggested that the fault is in general weakened by
the process of segmental growth of fault arrays due to linkage of
fault segments and progressive breakdown to fault rocks. In
parallel, the relative thickness is decreasing fromun-breached fault
segment geometries to fault rock layers. Similar conclusions can be
drawn from this study; (i) in general, cores thicken with displace-
ment, and (ii) geometrically-complex faults display thicker cores
than individual planar fault segments.
Fig. 13. Average thickness/displacement plot of thick, geometrically-complex fault
cores (triangles) and thin localized fault cores (circles). Trend lines are indicated for the
two groups.
There are, however, changes in the pattern linked to the
displacement. Lenses are for instance more common for moderate
and large offset faults, contrary to small faults. The observed
threshold for lenses between small and moderate offset faults
could have several reasons; (i) the lenses of small faults are
mostly abraded from the walls of slip surfaces and are therefore
very small. Thereby, they are observed as rock fragments in gouge
rather than as separate elements. (ii) Small faults more commonly
consist of isolated segments within a fault array, with fault prop-
agation that is layer/rheology controlled. As these segments link
up with increased fault offset, lenses develop in segment linkage
areas (Lindanger et al., 2007). At the same time, the fault develops
towards a through going structure less affected by wall rock
rheology (e.g. Peacock and Sanderson, 1991; Ferrill and Morris,
2003; Schpfer et al., 2006). As pointed out by Childs et al.
(2009), there is no objective means of determining whether
a fault-bounded host rock lens was formed by linkage of fault
segments and splays or breakdown of asperities and relay struc-
tures, in other words related to fault irregularity. But the observed
increased likelihood of encountering lenses in larger faults gives
general guidelines.
For moderate offset faults (1e10 m), composite fault cores
become more common. This could relate to shale layers that are
dragged along the fault and become juxtaposed with pre-existing
calcite veins and breccias. Alternatively, the smears are developed
in segment overlaps by shale layer rotation perhaps in parallel with
disintegration of stronger rocks into breccias along the slip zones.
Another trend is that the moderate offset faults have breccias that
form lenses/pockets, not through going membranes as commonly
found in large offset faults. This could be explained by the break-
down and disintegration of lenses with given positions within the
core, which results in localized breccias. Along a similar line,
breccias could relate to segment and splay linkage causing direct
disintegration of (strong) rocks in linkage areas. With increasing
offset, the development of the fault rocks in the core connects to the
number of disintegrated lenses and shattered linkage/irregularity
areas. If lenses and irregularities appear throughout the fault
evolution, the fault rocks would gradually become more wide-
spread and thereby start to form a multi-layer, progressively more
continuous membrane along the fault core. Further, the same fault
rocks would experience rejuvenation during fault events, devel-
oping internal thin shear zones or slip zones, hosting ner grained
fault rocks that surround fault rock lenses. These shear zones
commonly thin the fault rocks membranes. Both the shear zones in
fault rocks and the thinning of such rocks are frequently observed
in this study for faults with displacement between 10 and 100 m.
For faults with an offset exceeding 100 m, the common char-
acteristics is that of thicker cores with multiple slip zones and/or
slip surfaces, signicant breccia and shale membranes, and
a prominent occurrences of lenses. In that lenses are common for
the large offset faults, the process of lens formation and lens and
fault irregularity breakdown seems continuous through the fault
evolution, as envisioned by Childs et al. (2009).
One exception to this is some extensional faults formed in
anticlines of Oman. There, faults show thick breccias for relatively
small displacements, which could relate to dilation across the crest
of anticlines at shallow level in the crust (Ferrill and Morris, 2003).
This illustrates that the forces (extension, compression) acting in
the fault core could inuence both the core thickness and its
intrinsic composition. However, the fault core characteristics are
probably experiencing a stronger guidance by the mechanical
heterogeneity of wall rocks, especially for smaller offsets. With
increasing offset and better developed fault cores, the core lithol-
ogies and their developing rheology would also contribute (e.g.
Peacock and Sanderson, 1991; Schpfer et al., 2006; Childs et al.,
2009). Finally, geometric effects arising from fault irregularities,
splaying and segmentation would be an important factor.
8. Conclusions
This work addresses the relationship between thickness and
displacement of fault cores in ne grained carbonates. We relate
this relationship to the fault geometry, and the fault core and host
rock lithology:
1) Fault in ne grained carbonates has geometries varying
between simple fault cores and fault cores comprising fault
splays, lenses, segment linkages and overlap structures.
2) Typical rocks representing the thickness of the fault core are
carbonate breccias, carbonate and shale gouge, shale smear,
secondary calcite cement and veins, and host rock lenses.
3) There is a large scatter in fault core thickness vs. displacement.
Generally, the fault core thickness increases with displacement
that may be expressed by a power lawfunction (0.29D
0.56
). This
trend describes a gradual decrease in the thickness/displace-
ment relationship for increasing slip along faults. In more
detail, the general function is the sumof two (power law) trend
lines; the rst representing thin localized fault cores with
generally simple and planar geometry, the second representing
thicker fault cores with complex geometry of lenses and
overlap structures and with rock membranes.
4) There is a signicant change in composition and geometry from
small (0e1 m), to moderate (1e10 m) and to large offset faults
(10e400 m).
Small faults have thin, localized fault cores along straight
segments, with membranes consisting of calcite veins, gouge
along shear fractures, and shale smear. They often reveal
thicker segment linkage areas with a similar composition.
Moderate faults commonly have composite cores dominated
by both host rock and fault rock lenses surrounded by slip
surfaces or slip zones with gouge and breccias.
Large faults with 50 m and 100 m offset consist of relatively
thin localized membranes of sheared veins and ne grained
breccias, and with host rocks lenses. In contrast, larger faults
(>100 m offset) show thicker cores with multiple slip zones,
long breccia and shale membranes that often are cemented,
and lenses.
5) We envision that fault development is guided by forces
(extension, compression) acting in the fault, mechanical
heterogeneity of wall rocks, the core lithologies and their
developing rheology, and geometric effects arising from fault
irregularities.
Acknowledgement
Ian Sharp and David Hunt are gratefully thanked for providing
excellent maps and for introducing us to the Sinai and Oman eld
areas. The manuscript benetted signicantly through critical
reviewof Fabrizio Balsamo and Roger Soliva. Roy H. Gabrielsen, Atle
Rotevatn, Jan Tveranger and Simon Buckley are thanked for great
efforts in reviewing and discussing the paper. For eld compan-
ionship and exceptionally insightful discussions the authors wish to
thank Haakon and Sigurd Fossen, Harmon Maher, Sylvie Schueller
and Walter Wheeler.
The study was nanced by the Fault Facies project at the Centre
for Integrated PetroleumResearch, University of Bergen, by support
from ConocoPhillips, StatoilHydro, the Research Council of Norway,
and the University Centre on Svalbard.
References
Agosta, F., Kirschner, D.L., 2003. Fluid conduits in carbonate hosted seismogenic
normal faults of central Italy. Journal of Geophysical Research 108 (No B4), 13.
Agosta, F., Aydin, A., 2006. Architecture and deformation mechanism of a basin-
bounding normal fault in Mesozoic platform carbonates, central Italy. Journal of
Al-Kindi, M., Casey, M., Butler, R.W.H., 2006. Structural evolution and fracture
patterns in front range of northern Oman Mountains. In: Middle East Confer-
ence and Exhibition Manama, Bahrain.
Alsharhan, A.S., Nairn, A.E., 1997. M. Sedimentary Basins and Petroleum Geology of
the Middle East. Elsevier, Amsterdam, pp. 843.
Alsharhan, A.S., 2003. Petroleum geology and potential hydrocarbon plays in the
Gulf of Suez rift basin, Egypt. American Association of Petroleum Geologists 87,
143e180.
Balsamo, F., Storti, F. Grain size and permeability evolution of soft-sediment
extensional sub-seismic and seismic fault zones in high-porosity sediments
from the Crotone basin, southern Apennines, Italy. Marine and Petroleum
Geology, in press.
Bastesen, E., Braathen, A., Nttveit, H., Gabrielsen, R.H., Skar, T., 2009. Extensional
fault cores in micritic carbonates, a case study from Gulf of Corinth, Greece.
Benedicto, A., Plagnes, V., Vergly, P., Flott, N., Schultz, R.A., 2008. Fault and uid
interaction in a rifted margin: integrated study of calcite-sealed fault-related
structures (Southern Corinth margin). In: Wibberley, C.A.J., Kurz, W., Imber, J.,
Implications for Mechanical and Fluid-Flow Properties. Geological Society,
London, Special Publications, vol. 299, pp. 257e275.
Billi, A., Salvini, F., Storti, F., 2003. The damage zone fault core transition in
carbonate rocks: implications for fault growth, structure and permeability.
Billi, A., 2005. Grain size distribution and thickness of breccia and gouge zones from
thin (<1 m) strike-slip fault cores in limestone. Journal of Structural Geology 27,
1823e1837.
Blenkinsop, T.G., 1989. Thicknessedisplacement relationships for deformation
zones: discussion. Journal of Structural Geology 11, 1051e1054.
Bonson, C.G., Childs, C., Walsh, J.J., Schopfer, M.P.J., Carboni, V., 2007. Geometric and
kinematic controls on the internal structure of a large normal fault in massive
limestones: the Maghlaq Fault, Malta. Journal of Structural Geology 29 (2),
336e354.
Bosworth, W., Huchon, P., McClay, K., 2005. The Red Sea and Gulf of Aden Basins.
Journal of African Earth Sciences 43 (1e3), 334e378.
Braathen, A., Osmundsen, P.T., Gabrielsen, R.H., 2004. Dynamic development of fault
rocks in a crustal-scale detachment: an example from western Norway.
Tectonics 23 (No. 4 TC4010), 1e21.
Braathen, A., Tveranger, J., Fossen, H., Skar, T., Cardozo, N., Semshaug, S.L.,
Bastesen, E., Sverdrup, E., 2009. Fault facies and its applications to sand-
stone reservoirs. American Association of Petroleum Geologists Bulletin 93,
891e917.
structure. Geology 24, 1025e1028.
Carrio-Schaffhauser, E., Gaviglio, P., 1990. Pressure solution and cementation stim-
ulated by faulting in limestones. Journal of Structural Geology 12 (8), 987e994.
Cello, G., Tondi, E., van Dijk, J.P., Mattioni, L., Micarelli, L., Pinti, S., 2003. Geometry,
kinematics and scaling properties of faults and fractures as tools for modelling
geouid reservoirs: Examples from the Appennines, Italy. In: Niewland, D.A.
(Ed.), New Insights into Structural Interpretation and Modelling. Geological
Society, London, Special Publications, vol. 212, pp. 7e22.
Chester, F.M., Logan, J.M., 1986. Composite planar fabric of gouge from the Punch-
bowl fault zone, California. Journal of Structural Geology 9, 621e634.
Childs, C., Watterson, J., Walsh, J.J., 1996. A model for the structure and development
of fault zones. Journal of the Geological Society of London 153, 337e340.
Childs, C., Walsh, J.J., Watterson, J., 1997. Complexity in fault zone structure and
implications for fault seal prediction. In: Mller-Pedersen, P., Koestler, A.G.
(Eds.), Hydrocarbon Seals: Importance for Exploration and Production.
Norwegian Petroleum Society Special Publication, vol. 7, pp. 61e72.
Childs, C., Manzocchi, T., Walsh, J.J., Bonson, C.G., Nicol, A., Schpfer, M.P.J., 2009. A
Eliassen, A., Talbot, M.R., 2003. Sedimentary facies and depositional history of the
mid Carboniferous Minkinfjellet Formation, central Spitsbergen. Norwegian
Journal of Geology 83, 299e318.
Evans, J.P., 1990. Thicknessedisplacement relationships for fault zones. Journal of
Ferrill, D.A., Morris, A.P., 2003. Dilational normal faults. Journal of Structural
Geology 25 (2), 183e196.
Gabrielsen, R.H., Clausen, J.A., 2001. Horses and duplexes in extensional regimes:
a scale modelling contribution. In: Koyi, H.A., Mancktelow, N.S. (Eds.), Tectonic
Modeling: A Volume in Honor of Hans Ramberg. Geological Society of America
Memoir 193, pp. 219e233.
Graham Wall, B.R., Girbacea, R., Mesonjesi, A., Aydin, A., 2006. Evolution of fracture
and fault controlled uid pathways in carbonates of the Albanides fold and
thrust belt. American Association of Petroleum Geologists Bulletin 90,
1227e1249.
Grlaud, C., Razin, P., Homewood, P.W., Schwab, A.M., 2006. Development of inci-
sions on a periodically emergent carbonate platform (Natih Formation late
Cretaceous Oman). Journal of Sedimentary Research 76, 647e669.
Hanna, S.S., 1990. The Alpine deformation of the CentralOman Mountains. In:
Geological Society, London, Special Publications, vol. 49, pp. 341e359.
Mexico: implications for fault-zone permeability and uid ow. In:
Goodwin, L.B., Mozley, P.S., Moore, J.M., Haneberg, W.C. (Eds.), Faults and
Subsurface Fluid Flow in the Shallow Crust. Geophysical Monograph 113.
American Geophysical Union, Washington, pp. 27e49.
Hull, J., 1988. Thicknessedisplacement relationships for deformation zones. Journal
of Structural Geology 10, 431e435.
Immenhauser, A., Dublyansky, Y.V., Verwer, K., Fleitman, D., Pashenko, S.E., 2007.
Textural, elemental and isotopic characteristics of Pleistocene phreatic cave
deposits (Jabal Madar, Oman). Journal of Sedimentary Research 77, 68e88.
Jackson, C.A.L., Gawthorpe, R.L., Leppard, C.W., Sharp, I.R., 2006. Rift-initiation
development of normal fault blocks: insights from the Hammam Faraun fault
block, Suez Rift, Egypt. Journal of the Geological Society 163 (1), 165e183.
Johannessen, E.P., Steel, R., 1992. Mid-carbonferous extension and rift-inll
sequences in the Billefjorden through, Svalbard. Norsk Geologisk Tidskrift 72,
35e48.
Knott, S.D., 1994. Fault zone thickness versus displacement in the Permo-Triassic
sandstones of NW England. Journal of the Geological Society 151, 17e25.
Kuss, J., Scheiber, C., Gietl, R., 2000. Carbonate platform to basin transition along an
Upper Cretaceous to Lower Tertiary Syrian Arc uplift, Galala Plateaus, Eastern
Desert of Egypt. GeoArabia 5, 405e424.
Labaume, P., Carrio-Schaffhauser, E., Gamond, J.-F., Renard, F., 2004. Deformation
mechanisms and uid-driven mass transfers in the recent fault zones of the
Corinth Rift (Greece). Comptes Rendus Geosciences 336 (4e5), 375e383.
Lindanger, M., Gabrielsen, R.H., Braathen, A., 2007. Analysis of rock lenses in
extensional faults. Norwegian Journal of Geology 87, 361e372.
smears on fault surfaces. In: Flint, S.S., Bryant, I.D. (Eds.), The Geological
Modelling of Hydrocarbon Reservoirs and Outcrop Analogues. Special Publica-
tions, International Association of Sedimentologists, vol. 15, pp. 113e123.
Maher, H.D. Jr., Braathen, A. Lvehovden fault and Billefjorden rift basin segmen-
tation and development, Spitsbergen. Geological Magazine, in press.
for ow simulation models. Petroleum Geoscience 5, 53e63.
Manzocchi, T., Childs, C., Walsh, J.J., 2010. Faults and fault properties in hydrocarbon
ow models. Geouids 10, 94e113.
McCann, A.J., Dallmann, W.K., 1996. Reactivation history of the long-lived Bill-
efjorden Zault Zone in the north Central Spitsbergen, Svalbard. Geological
Magazine 133, 63e84.
Micarelli, L., Moretti, I., Daniel, J.M., 2003. Structural properties of rift-related
normal faults: the case study of the Gulf of Corinth, Greece. Journal of Geo-
dynamics 36 (1e2), 275e303.
Micarelli, L., Benedicto, A., Invernizzi, C., Saint-Bezar, B., Michelot, J.L., Vergely, P.,
2005. Inuence of P/T conditions on the style of normal fault initiation and
growth in limestones from the SE-Basin, France. 2006. Journal of Structural
Geology 27, 1577e1598.
Micarelli, L., Benedicto, A., Wibberley, C.A.J., 2006. Structural evolution and
permeability of normal fault zones in highly porous carbonate rocks. Journal of
Montenat, C., Soudet, H.J., Barrier, P., Cherau, A., 2000. Karstication and tectonic
evolution of the Jabal Madar (Adam Foothills, Arabian platform) during the
Upper Cretaceous Bulletin Centre de Recherches. ELF Exploration Production
22, 161e183.
Moustafa, A.R., Abdeen, M.M., 1992. Structural setting of the Hamman Faraoun
block, eastern side of the Suez rift. Journal of the University of Kuwait (Science)
19, 291e309.
Moustafa, A.R., 2004. Geologic maps of the eastern side of the Suez Rift (Western
Sinai Peninsula), Egypt. American Association of Petroleum Geologists Map
Series.
Myers, R., Aydin, A., 2004. The evolution of faults formed by shearing across joint
zones in sandstone. Journal of Structural Geology v. 26, 947e966.
Otsuki, K., 1978. On the relationship between the width of shear zone and the
displacement along fault. Journal of the Geological Society of Japan 84, 661e669.
Patton, T.L., Moustafa, A.R., Nelson, R.A., Abdine, S.A., 1994. Tectonic evolution and
structural setting of the Suez Rift. In: Landon, S.M. (Ed.), Interior Rift Basin
American Association of Petroleum Geologists Memoir 59, pp. 7e55.
Peacock, D.C.P., Sanderson, D.J., 1991. Displacements, segment linkage and relay
ramps in normal fault zones. Journal of Structural Geology 13, 721e733.
Peacock, D.C.P., Sanderson, D.J., 1995. Pull aparts, shear fractures and pressure
solution. Tectonophysics 241, 1e13.
Peacock, D.C.P., 2002. Propagation, interaction and linkage in normal fault systems.
Earth Science Reviews 58 (1e2), 121e142.
Petit, J.P., 1987. Criteria for the sense of movement on fault surfaces in brittle rocks.
Journal of Structural Geology 9 (5e6), 597e608.
Petit, J.-P., Wibberley, C.A.J., Ruiz, G., 1999. Crack-seal, slip: a new fault valve
mechanism? Journal of Structural Geology 21 (8e9), 1199e1207.
Pickard, N.A.H., Eilertsen, F., Hanken, N.M., Johansen, T.A., Lny, A., Nakrem, H.A.,
Nilsson, I., Samuelsberg, T.J., Somerville, I., 1996. Stratigraphic frameworkof upper
Carboniferous (Moscovian-Kasimovian) strata in BnsowLand. Central Spitsber-
gen: palaeogeographic implications. Norsk Geologisk Tidskrift 76, 169e185.
Putz-Perrier, M.W., Sanderson, D.J., 2010. Distribution of faults and extensional
strain in fractured carbonates of the North Malta Graben. American Association
of Petroleum Geologists 94 (No 4), 435e456.
Roberts, G., Stewart, I., 1994. Uplift, deformation and uid involvement within an
active normal fault zone in the Gulf of Corinth, Greece. Journal of the Geological
Society of London 151, 531e541.
thickness. Mining Engineering 35, 1426e1432.
Robson, D.A., 1971. The structure of the Gulf of Suez (Clysmic) rift, with special
reference to the eastern side. Journal of the Geological Society 127 (3), 247e271.
Rykkelid, E., Fossen, H., 2002. Layer rotation around vertical fault overlap zones:
observations from seismic data, eld examples, and physical experiments.
Marine and Petroleum Geology 19 (2), 181e192.
Said, R., 1960. Planktonic foraminafera from the Thebes formation, Luxor, Egypt.
Micropaleontology 6 (3), 277e286.
Samuelsberg, T., Elvebakk, G., Stemmerik, L., 2003. Late Palaeozoic evolution of the
Finnmark platform, southern Norwegian Barents Sea. Norwegian Journal of
Scholz, C.H., 1987. Wear andgouge formationinbrittle faulting. Geology 15, 493e495.
Scott, R.W., 1990. Chronostratigraphy of the Cretaceous carbonate shelf, southeastern
Arabia. In: Geological Society, London, Special Publications, vol. 49, pp. 89e108.
Schpfer, M.P.J., Childs, C., Walsh, J.J., 2006. Localisation of normal faults in multi-
layer sequences. Journal of Structural Geology 28, 816e833.
Sharp, I.R., Gawthorpe, R.L., Armstrong, B., Underhill, J.R., 2000. Propagation history
and passive margin rotation of mesoscale normal faults: implications for synrift
stratigraphic development. Basin Research 12, 285e305.
Shipton, Z.K., Soden, A.M., Kirkpatrick, J.D., Bright, A.M., Lunn, R.J., 2006. How
thick is a fault? Fault displacementethickness scaling revisited. In: Aber-
crombie, R., McGarr, A., Di Toro, G., Kanamori, H. (Eds.), Radiated Energy and
the Physics of Faulting. American Geophysical Union Monograph Series 170,
pp. 193e198.
Sibson, R.H., 1977. Fault rocks and fault mechanisms. Jornal of the Geological Society
of London 133, 191e213.
Soliva, R., Benedicto, A., 2005. Geometry scaling relations and spacing of vertically
restricted normal faults. Journal of Structural Geology 27, 317e325.
Steel, R.J., Worsley, D., 1984. Svalbard`s post Caledonian strata e an atlas on sed-
imentational patterns and palaeogeographic evolution. In: Graham and Trot-
man, Petroleum Geology of the North European Margin. Norwegian Petroleum
Society, pp. 109e135.
Storti, F., Billi, A., Salvini, F., 2003. Particle size distributions in natural carbonate
fault rocks: insights for non-self-similar cataclasis. Earth and Planetary Science
Letters 206 (1e2), 173e186.
Sperrevik, S., Gillespie, P.A., Fisher, Q.J., Halvorsen, T., Knipe, R.J., 2002. Empirical
Hydrocarbon Seal Quantication. Norwegian Petroleum Society, Special Publi-
cation, vol. 11, pp. 109e125.
Tveranger, J., Braathen, A., Skar, T., Skauge, A., 2005. Centre for Integrated Petroleum
Research e research activities with emphasis on uid ow in fault zones.
Norwegian Journal of Geology 85, 63e71.
Van der Zee, W., Urai, J.L., 2005. Processes of normal fault evolution in a siliciclastic
Geology 27, 2281e2300.
Wagner, P.D., 1990. Geochemical stratigraphy and porosity controls in cretaceous
carbonates near the Oman Mountains. In: Geological Society, London, Special
Publications, vol. 49, pp. 127e137.
Walsh, J., Watterson, J., Heath, A.E., Childs, C., 1998. Representation and scaling of
faults in uid ow models. Petroleum Geoscience 4, 241e251.
Geological Society of London, Special Publication, vol. 299, pp. 5e33.
American Association of Petroleum Geologists Bulletin 81, 897e917.
Analysis of the growth of strike-slip faults using effective medium theory
Atilla Aydin
a,
*
, James G. Berryman
b
a
Rock Fracture Project, Department of Geological and Environmental Sciences, Stanford University, Stanford, CA, USA
b
Lawrence Berkeley National Laboratory, Earth Science Division, One Cyclotron Road, MS/90R1116, Berkeley, CA, USA
Article history:
Received 14 March 2009
9 November 2009
Accepted 16 November 2009
Keywords:
Fault growth
Fault scaling
Fault linkage and coalescence
Fault damage zone
Cataclastic deformation
Effective moduli
Effective medium model
a b s t r a c t
Increases in the dimensions of strike-slip faults including fault length, thickness of fault rock and the
surrounding damage zone collectively provide quantitative denition of fault growth and are commonly
measured in terms of the maximum fault slip. The eld observations indicate that a common mechanism
for fault growth in the brittle upper crust is fault lengthening by linkage and coalescence of neighboring
fault segments or strands, and fault rock-zone widening into highly fractured inner damage zone via
cataclastic deformation. The most important underlying mechanical reason in both cases is prior
weakening of the rocks surrounding a faults core and between neighboring fault segments by faulting-
related fractures. In this paper, using eld observations together with effective medium models, we
analyze the reduction in the effective elastic properties of rock in terms of density of the fault-related
brittle fractures and fracture intersection angles controlled primarily by the splay angles. Fracture
densities or equivalent fracture spacing values corresponding to the vanishing Youngs, shear, and quasi-
pure shear moduli were obtained by extrapolation from the calculated range of these parameters. The
fracture densities or the equivalent spacing values obtained using this method compare well with the
eld data measured along scan lines across the faults in the study area. These ndings should be helpful
for a better understanding of the fracture density/spacing distribution around faults and the transition
from discrete fracturing to cataclastic deformation associated with fault growth and the related
instabilities.
1. Introduction
Strike-slip faults, similar to other types of faults, typically have
complex architectures with numerous segments or strands of
various trace lengths separated by steps or relays of various sizes
(Fig. 1a). This discontinuous characteristic of strike-slip faults has
been reported for simple incipient faults (Segall and Pollard, 1980,
1983; Gamond, 1983; Sibson, 1986; Willemse et al., 1997; Peacock
and Sanderson, 1995) as well as for mature crustal-scale faults
(Aydin and Nur, 1982; Barka and Kadinsky-Cade, 1988; Wesnousky,
1988; Stirling et al., 1996; Kim et al., 2004) and is thought to be
pertinent to a number of properties of strike-slip fault systems
including their permeability structure (Sibson, 1985; Martel and
Peterson, 1991; Aydin, 2000; Odling et al., 2004), the dynamics and
size of earthquake ruptures (Aki, 1989; Harris and Day, 1999; Harris
et al., 1999; Wesnousky, 2006; Shaw and Dieterich, 2007), the
spatial and temporal evolution of earthquakes (Dewey, 1976;
Tokso z et al., 1979; Stein et al., 1997), and growth and scaling of
faults (de Joussineau and Aydin, 2009; Scholz, 2002).
One of the fault scaling relationships concerns fault length (L) to
maximum fault slip or displacement (D). Various studies of mostly
normal faults (Watterson, 1986; Walsh and Watterson, 1987; Cowie
and Scholz, 1992; Schlische et al., 1996; Scholz, 2002) concluded
that the lengthslip relationship has the form, L D
n
, where n was
proposed to be between 1 and 2.
Neighboring segments of strike-slip faults are separated by
steps (Fig. 1a). These steps have self-similar geometry regardless of
the sense of stepping and sense of shearing (Aydin and Schultz,
1990; Aydin and Nur, 1982). However, the failure modes and the
distribution of the shearing-related structures may be different
from one sense of step to another depending on loading, stress
perturbations, rheology, and the geometry of initial pre-faulting
discontinuities (Kim et al., 2004; Myers and Aydin, 2004; Peacock
and Sanderson, 1991, 1995; Burgmann and Pollard, 1994; Sibson,
1986; Gamond, 1983; Rispoli, 1981).
A data set collected by Wesnousky (1988) from crustal-scale
strike-slip faults suggests that the number of steps per kilometer
along strike-slip faults decreases as fault slip increases. Although it
E-mail address: aydin@stanford.edu (A. Aydin).
doi:10.1016/j.jsg.2009.11.007
is difcult to dene uncertainty regarding the fault dimensions
measured from published geologic and seismologic maps, recent
experimental (Otsuki and Dilov, 2005) and site-specic eld data
from the same tectonic region and lithology and similar resolution
(de Joussineau and Aydin, 2009) appear to conrm this trend. It is
also interesting to note that larger size of steps is associated with
faults having larger maximum slip magnitudes (Aydin and Nur,
1982; de Joussineau and Aydin; 2009), possibly related to the ability
of fault slip to jump from one fault segment to next over the fault
steps between them (Shaw and Dieterich, 2007; Harris and Day,
1999; Harris et al., 1999).
The fault length-slip, and step count per kilometer length
relationships, regardless of their exact form, imply that faults, like
other types of geological structures having different senses of
displacement discontinuity, start small in length and growlarger in
time and space. As faults grow longer, they are able to interact with
the neighboring faults at greater distances. Consequently, faults
extendtheir lengths bylinkage andcoalescence of smaller segments
through fault steps in order to accommodate larger amount of slip
(Segall and Pollard, 1983; Martel et al., 1988; Martel, 1990; Peacock
and Sanderson, 1995; Cartwright et al., 1995; Dawers and Anders,
1995; Pachell and Evans, 2002; Scholz, 2002; Myers and Aydin,
2004; de Joussineau and Aydin, 2009). It follows that the fault
length-maximum fault slip plots for faults which grew by linkage
and coalescence for any single fault zone is not actually continuous
but rather have sharp jumps coinciding with large increases in
lengths at the merger of neighboring segments and ats
corresponding to the time span between the consecutive merger
instances, in which fault lengths stay nearly constant while fault
slips increase tothelimitinglength/slipratio(Cartwright et al., 1995)
Another type of fault scaling relationship illuminates how fault
zones become wider as they grow (Fig. 1b). Field data (Hull, 1988;
Robertson, 1983; Knott et al., 1996) and theoretical considerations
(Scholz, 2002) suggest that the width or thickness of faults
increases linearly with fault slip. Agosta and Aydin (2006) and de
Joussineau and Aydin (2007) proposed that fault rock zones growor
widen perpendicular to their trend at the expense of highly
fractured inner damage zone via cataclastic deformation. This
widening is also inuenced by the width of the steps along faults
(Kim et al., 2004; Childs et al., 2009), which are precursors of fault
cores.
An important consequence of lengthening of faults by linkage
and coalescence is that larger magnitude of slip takes place in
merged or composite segments which tend to straighten the overall
through-going fault trace with respect to the earlier segmented or
discontinuous trace. This process, which appears to be a second
order shear localization phenomenon immediately after the fault
zone attains the next composite conguration, is referred to as
through-going faulting, fault straightening, and fault zone simpli-
cation (Cox and Scholz, 1988; Reches and Lockner, 1994; Le Pichon
et al., 2001; Scholz, 2002; Ben-Zion and Sammis, 2003).
As this short introductory account indicates, the discontinuous
geometry of strike-slip faults, their segmentation, the geometry
and scaling of the segments and steps, and their impact on earth-
quake rupture, uid ow and mineralization have attracted
considerable interest in the literature. However, aside from
a number of papers addressing the stress state between neigh-
boring faults and the type and orientation of the linkage structure
(Segall and Pollard, 1980; Pollard and Segall, 1987; Du and Aydin,
1993; Crider and Pollard, 1998; De Bremaecker and Ferris, 2004),
very little attention has been paid to quantication of the elastic
parameters leading to the growth of the fault dimensions. To this
end, only a handful of studies address these criteria. The rst group
of these papers includes those dealing with calculation of the
critical damage parameters at fault steps in terms of strain invari-
ants (Lyakhovsky and Ben-Zion, in review; Lyakhovsky et al., 1997).
The second category is rather empirical and is based on a eld
survey of normal faults and subsequent analysis of displacement-
segment separation ratio to dene those elds with unlinked and
linked congurations (Soliva and Benedicto, 2004). A large number
of publications deal with calculating effective moduli of fractured
materials (Lockner and Madden, 1991; Sayers and Kachanov, 1991;
Berryman and Grechka, 2006; Grechka and Kachanov, 2006;
Berryman, 2008; Grifth et al., 2009). However, these do not
address directly the subject of the actual fault growth and the
related problems.
In this paper, we present natural and idealized fracture patterns
around faults and within fault steps from the Valley of Fire State
Park, Nevada, and use effective medium theory to investigate how
rocks around and between fault segments weaken by faulting-
related fractures and, consequently, how these intensely fractured
rock masses may facilitate the growth of fault dimensions. Effective
medium theory (sometimes called mixture theory, averaging,
up-scaling or homogenization) is a collection of analytical methods
designed to capture average properties and behaviors of very
complex and heterogeneous media (Berryman, 2008). These
methods have been applied to almost any type of complex system
and physical property, but they tend to work best for quasi-static
behaviors such as those considered in the present paper. A related
paper (Berryman and Aydin, in press) dealing with the method-
ology in more detail for calculating effective moduli of fractured
media will be presented somewhere else.
Fig. 1. Idealized diagrams showing: (a) segmentation along strike-slip faults and their
segment lengths, heights and step numbers. Overlaps (or step lengths) and step widths
(or fault separations) are also shown in inset. (b) Internal architecture of a strike-slip
fault showing fault core, which includes fault rock and slip surfaces, and the
surrounding damage zone.
A. Aydin, J.G. Berryman / Journal of Structural Geology 32 (2010) 16291642 1630
2. Geological background
For more than a decade, one of us (A.A.) and his students have
been studying the initiation, interaction and growth of brittle,
primarily strike-slip faults in the Jurassic eolian Aztec Sandstone
withexcellent exposures inthe Valleyof Fire State Park, about 60 km
northeast of Las Vegas, Nevada (Fig. 2) (akir andAydin, 1994; Flodin
et al., 2005; Myers and Aydin, 2004; Flodin and Aydin, 2004; de
Joussineau and Aydin, 2007, 2009; de Joussineau et al., 2007). We
have documented that these Cenozoic faults (Bohannon, 1983) with
both left- and right-lateral offsets ranging froma fewmillimeters to
a fewkilometers initiatedfroma systemof joint zones byshearingof
the joints, formation of new splay fractures and their subsequent
shearing. Bymappingof strike-slipfaults withincreasingmagnitude
of slip, the mechanism of fault growth was established to be the
linkage and coalescence of initially sheared joint zones and of fault
segments at progressively greater scales (Myers and Aydin, 2004;
Flodin and Aydin, 2004; de Joussineau and Aydin, 2007). It was also
concluded that the pattern and orientations of these faults within
the Valley of Fire and the surrounding area are reminiscent of the
large size strike-slip faults in the southeastern Basin and Range
province (akir et al., 1998). The conceptual models and the eld
data in the next two sections ultimately rest upon these early
publications as well as some new data and synthesis.
3. Conceptual models
Fig. 1a is an idealized diagram illustrating fundamental
geometric attributes of strike-slip faults. Along-strike view shows
segments with various lengths (l
1
,l
2
,.,l
n
), which include the slip
vector direction by denition. Hence, identifying segments assures
a basis for determining mean and maximum segment lengths. This
view also shows discontinuities along the trace length in the form
of steps (s
1
,s
2
,.,s
n 1
) with overlaps (o
1
,o
2
,.,o
n 1)
and widths
(w
1
,w
2
,.,w
n 1)
or separations (Fig. 1a inset). These parameters
provide the bases for calculating the number of steps per length,
and the size of the steps along a given strike-slip fault.
Down-dip view also includes segments with steps. We will not
consider down-dip segmentation and steps because these steps do
not provide much resistance to strike-slip motion due to the slip
vector being horizontal and therefore the linkage is relatively
simple. In vertically anisotropic lithologies, the problem becomes
more complex by the presence of inelastic rocks such as shale,
which is out of the scope of this paper.
Fig. 1b shows a detailed view of fault zone architecture with
a fault core, made up of fault rock and slip surfaces. Fault rock is the
product of fragmentation and cataclasis when a highly fractured
rock is disaggregated, fracture-bounded blocks rotate and translate
and grains crush. Slip surfaces are discrete structures that accom-
modate shear displacement along polished and striated surfaces
which usually run through, or occur adjacent to, the fault rock. Fault
cores are anked on both sides by a damage zone which includes
a complex fracture systemof splay joints and sheared splay joints of
various generations. We note that, as depicted in the diagram, the
distributions of fault rock and damage zone are highly irregular. The
diagraminFig. 1balsoillustrates the notionthat one of the major slip
surfaces withinthe fault core maybe continuous fromone endtothe
other, whichis meant torepresent a through-goingfault surface, and
accommodates a large portion of the total slip across the fault zone.
4. Field data
Here, we focus on dimensional attributes and growth processes
of a network of strike-slip faults exposed in the Aztec Sandstone
cropping out in the Valley of Fire Sate Park and its surroundings
(Fig. 2). Although the geometric and mechanical properties of the
strike-slip faults at this location appear to be similar to those
strike-slip faults from different regions (de Joussineau and Aydin,
2009) as summarized in Section 1, we restrict our statistical and
conceptual models to the cases that we studied in some detail at
the Valley of Fire State Park for two reasons. One is that the
mechanisms of fault initiation and growth are well known and the
precision of measurements is very good and fairly uniform in
a wide range of scales.
4.1. Fault segment length, step number and dimensions, and fault
width
Data from about 20 well-exposed faults in the study area show
that the mean segment length increases with the maximum fault
Fig. 2. Location and simplied map of the faults of Cenozoic age in the Valley of Fire region of southern Nevada. Heavy lines are faults. Arrows indicate predominant sense of slip.
LMFS, Lake Mead Fault System; LVVSZ, Las Vegas Valley Shear Zone. Rectangle marks the location of the Valley of Fire State Park.
slip (de Joussineau and Aydin, 2009) as idealized in Fig. 3a. Here, we
use a simple idealized trend because the exact nature of the
relationship is not the focus. Data collected from the same faults
also show that the number of steps per kilometer decreases as
a function of the maximumfault slip. Stating this result in a simpler
way; when n number of segments is reduced to n 1 segments by
linkage of the two neighboring segments, then, n 1 number of
steps will be reduced down to n 2 steps by destruction of one of
the steps.
The damage zone and fault rock widths increase with increasing
fault slip (de Joussineau and Aydin, 2007, 2009; Flodin et al., 2005;
Myers and Aydin, 2004). These relationships are not locally smooth,
but regardless of the actual forms, can be idealized as shown in
Fig. 3b. These trends are more meaningful for considering either
mean or maximum widths versus maximum slip values. For
example, the damage zone widths are controlled primarily by the
location, angle and length of the splay fractures (de Joussineau
et al., 2007) whose distribution about the fault may be highly
nonuniform.
The nature of fracturing at small fault steps can be characterized
as initial splay fracturing in response to shearing of the echelon
joints (Fig. 4a) and then shearing of the 1st generation of splay
fractures and formation of a 2nd generation splay fractures that
connect the sheared 1st generation splay fractures (Fig. 4b). The
intersection angle between the sheared joints and the 1st genera-
tion splay fractures varies from about 15
to 85
with an average
value of about 19
for isolated sheared joints, and about 50
for
subparallel interacting sheared joints (Fig. 5a,b and Table 1a).
Terminal areas of a fault zone generally reect the incipient
stages of fault development. In this regard, Fig. 6 shows a portion of
a fault zone that has w65 cm maximum observable right-lateral
slip elucidating the transition from an echelon sheared-joints array
to a through-going fault formation. Similar to the cases shown in
Fig. 4, shearing of the initial joint systemresulted in splay fracturing
and continued shearing facilitated the formation of multiple sets of
sequential splays localizing into discontinuous pockets of high
density fractures, and, in places, fragmentation zones. The incipient
short slip surfaces eventually go through these pockets of weak-
ened rock at fault steps.
The photographs and maps in Fig. 7a,b showa well-exposed fault
of about 14 m left-lateral slip, which displays several characteristic
architectural elements common to all strike-slip faults in the study
area: Fault rock, slip surfaces, and damage zone. Fig. 7a shows
domains of different deformation zones and of fracture densities,
which allow one to see a simpler picture of elongated, noncolinear
bodies of the fault rock and the adjacent areas of high fracture
density. Fig. 7a,b also shows slip surfaces invarious orientations and
sizes, one of whichis continuous fromone endof the mappedarea to
the other going through the elongated bodies of fault rock. There are
also short slip surfaces terminating at an acute angle to the
rectilinear strings of fault rock (Fig. 7b). We interpret these diagonal
short slip surfaces as relics of the initial sheared joints and the high
Fig. 3. Idealized diagrams summarizing general trends of (a) mean segment length
and number of steps per kilometer, and (b) mean fault rock and mean damage zones
widths as the maximum fault slip increases.
Fig. 4. (a) Incipient right-lateral shearing (w2 cm) of a series of echelon joints with right steps. Splay fractures at high-angle to the sheared joints are localized near the tips of the
segments at the steps. From Myers and Aydin (2004). (b) Two sets of splay fractures at and around a right step along a strike-slip fault with about 80 cm right-lateral slip. The two
sets have a range of intersection angles from 30
to 60
. From de Joussineau and Aydin (2009).

intensity fracture zones, to a large extent, fracture localization
betweenthese shearedjoints. Then, the fault rockandthe associated
slip surfaces subparallel to the fault core represent the architecture
of a newprogressionin fault growth with a through-going fault core
and slip surface representing the latest linkage and coalescence
structure, for this portion of the fault. This elucidation makes an
important point that anygivenstage of progressive faulting includes
overprinted patterns of fault zone elements, which poses a major
challenge in predicting a faults architecture without sufcient
information for its deformation history.
The 14 m fault described above also displays characteristics of
its damage zone. Fig. 8a,b shows photos and details of fractures on
one side of this fault. The fault core and the inner and outer damage
zones are shown in Fig. 8a. The details of the inner damage zone are
shown in Fig. 8b. Here, several generations of splay fractures are
identied based on their abutting relationships and marked by
different color codes. The younger ones ll in between the echelon
sheared joints of the initial stage (de Joussineau and Aydin, 2007).
Fig. 8c shows a photo and a detailed map of ne- and coarse-
grained fault rocks and major through-going slip surfaces along
a strike-slip fault with about w25 m left-lateral slip. Some of the
earlier fractures within the coarse-grained fault rock can still be
identied (dotted lines). Also important to point out are the
triangular pockets of ne-grained fault rock protruding into the
damage zone on the left hand side of the fault core, where
the fracture intensity appears to be high. The triangular zones
between one of the main slip surfaces and sheared splay fractures
(one marked in the map as x) are known to be the location of
higher fracture concentrations based on observations at other
locations with similar geometric settings (Flodin and Aydin, 2004).
A conceptual model depicting linkage and coalescence of fault
segments or strands which result in longer segment lengths,
reduced number of fault steps per kilometer, and wider damage
zones and fault core with increasing fault slip is shown in Fig. 9. In
a simple way, the model illustrates how strike-slip faults grow via
an interrelated series of processes including splay fracturing,
shearing of splay fractures, segment linkage, and formation of
through-going slip surfaces in a hierarchical manner. Progressive
lengthening of the linked segments, in turn, drives splay fractures
to farther distances from the main body of the fault and thus
increases the size of potential steps and eventually the widths of
fault cores and fault damage zones.
The types of eld data crucial for the fault growth are the
density or spacing of the fault-related fractures as well as their
patterns. In the study area, a vast amount of data is available for the
distribution of the fault-related fractures (de Joussineau and Aydin,
2007, 2009; de Joussineau et al., 2007; Flodin and Aydin, 2004;
Myers, 1999). Fig. 10a displays one of these for the left-lateral fault
Fig. 5. Histograms showing the distribution of the intersection angles between faults
and their splay fractures (splay angles) for more than 750 measurements. (a) For
isolated faults, and (b) for closely spaced, interacting subparallel faults. From de
Joussineau et al. (2007).
Table 1
(a) The average fracture intersection angles for isolated faults and their splays (19
),
and for closely spaced interacting subparallel faults and their splays (50
). From de
Joussineau et al. (2007). (b) The spacing range dened by the best t line to the
smallest end of the spacing distribution of the fault-related fractures in the study
area. From de Joussineau and Aydin (2007). The spacing values obtained by this
method are between w1 and 5 cm with the largest concentration between 1 and
2 cm (see inset 1). The angular differences between the scanline and fault-related
fractures for 14 m fault (inset 2 in which the bins represent intervals of 09, 1019,
etc.). More than 75% of the fractures make angles larger than 50
to the scanline (see

the diagram within inset 2 for designing the angle between scanline and a fracture
set (a) and the true and apparent spacing (S
t
and S
a
, respectively)). This requires
a maximum correction factor of about 0.77.
a
Pattern
type
Average
splay
angle
Inset 1
Isolated/single
fault
19
Interacting
faults
50
b
Fault Spacing
range
(cm)
Inset 2
80 cm 1.518
8 m 2.039
14 m 2.051
80 m 1.4110
Mixed
Scanline #1 2.052
Scanline #2 0.914
Scanline #3 5.038
with 14 m slip referred to earlier in this manuscript (see Figs. 7 and
8). The spacing distribution for the fault-related fractures for this
fault has been determined along 8 scanlines perpendicular to the
fault trace in intervals about 26 m recording the distance,
orientation and length of the fractures with a resolution of 0.5 cm.
The spacing has been calculated as the distance between consec-
utive fractures. Of interest here is the spacing distribution on the
smaller end of the spectrum identied by the linear trend in the
spacing distribution plot which denes a range of spacing values
from 2 to 51 cm (see Fig. 10a). Fig. 10b shows the fracture spacing
distribution obtained froma single scanline across many faults with
aggregate slip on the order of a few hundreds of meters. Here the
range of the smallest linear spacing trend is 538 cm. In both cases,
we focus on the smallest ends of the ranges, 2 cm in Fig. 10a and
5 cm in Fig. 10b) which represent the smallest fracture spacing for
a statistically signicant number of data points measured near the
fault cores. Similarly, in Table 1b, the spacing ranges dened by the
linear ts to the smallest slopes in the distribution of data, and the
corresponding minimum spacing values, for three other faults and
two additional scanlines across a number of faults are given. As
shown in the histogram summary (inset 1) a majority of the
minimumspacing values falls between 1 and 2 cm. Considering the
minimum measurable spacing was 0.5 cm, these numbers are well
above the minimum resolvable spacing.
5. Analysis using effective medium models
The premise of this study is that a certain degree of high
intensity fracturing at fault steps and fault damage zones weakens
the rock masses thereby facilitating fault lengthening through
linkage and coalescence of neighboring segments and fault zone
widening by incorporation of the fractured and fragmented
material into the fault rock via a cataclastic process.
Next, we will use effective medium models to investigate the
degradation of the strength and reduction of resistance to
cataclastic deformation, which presumably pave the way for the
setting of through-going faults. Given the complexity of the
Fig. 6. Detailed map of the end of a small shear zone of about 65 cm observable
maximum right-lateral slip showing a set of slightly sheared and highly overlapped
echelon joints and sheared joints with many splay joints at high-angle to the sheared
joints. Gray shading marks narrow pockets of fragmentation and thick lines show
incipient through-going slip surfaces orientated at a small-angle to the sheared initial
echelon joints. Slightly revised from Davatzes and Aydin (2004).
Fig. 7. (a and b). Detailed maps of a strike-slip fault with about 14 m left-lateral slip. (b) shows the orientations, lengths, and intersections of damage zone fractures (splay joints and
sheared splay joints) around the fault core (Myers and Aydin, 2004) whereas (a) shows a new reinterpreted version of the same fault architecture in which noncolinear pockets of
fault rocks and highly fractured domains of occasionally fragmented damage zone are delineated in the eld. One through-going slip surface (thick solid line) and several short slip
surfaces (dotted lines) diagonal to the through-going slip surface are highlighted. The geometry and distribution of many of short diagonal slip surfaces resemble the initial echelon
sheared joints observed along faults with smaller slip in the area. The original map by R. Myers (1999); the present version was revised from Davatzes and Aydin (2004).
fractures around faults, the problem is obviously rather difcult
and, at this stage, further simplication is desirable for applications
to an effective medium theory. We rst idealize the common
fracture patterns in terms of their orientation-intersection angle,
length, and density (Fig. 11a). We then study parametrically the
effective elastic moduli of such a conguration as a function of
fracture density for each idealized fracture pattern dened by the
angle between fracture sets in order to assess the degree of
degradations in the moduli values as the fracture density increases.
The details of the effective medium theory that we shall employ
have been recently described by Berryman and Aydin (in press) and
is based on the earlier work by Backus (1962), Schoenberg and Muir
(1989), and Berryman and Grechka (2006).
Fig. 11a shows an idealized fracture network which is consistent
in principle with the sequential formation of two fracture sets and
their ultimate pattern, the examples of which can be seen in Figs. 4,
6, 7, and 8. Here the lengths (l) of the fractures, the density (r) or
spacing (s) of the fractures, and angle (F
F
) between the two sets of
fractures characterize the pattern in a layer with a thickness, h.
One of the most commonly used fracture density concepts goes
back to Bristow (1960) and Budiansky and OConnell (1976). For
a set of rectangular at (or a ribbon-shaped) fractures, which is the
most pertinent to physical properties of fractured media such as
resistivity, uid ow, and elasticity, is
r nh
2
l (1)
where n N/V, with N and V being the number of fractures and the
rock volume, respectively. The volume, V is equal to lhs where l, h,
and s are average fracture length, height, and spacing, respectively.
Taking t as the average fracture thickness or fracture aperture, the
porosity of a system of rectangular at fractures with an average
Fig. 8. Damage zone characteristics around the strike-slip fault with w14 m left-lateral slip. (a) Two fractured domains were distinguished: The inner damage zone of high fracture
density right next to the fault core; and the outer damage zone of signicantly lower fracture density. (b) A detailed map of the inner damage zone shows that several generations of
splay fractures (marked by different color codes) ll in between the echelon sheared joints of the initial stage. (a,b) From de Joussineau and Aydin (2007). (c) Detail map showing
ne- and coarse-grained fault rock and major through-going slip surfaces along a left-lateral fault with about 25 m left-lateral slip. Some of the earlier fractures within the coarse-
grained fault rock can still be identied (dotted lines). Also important to point out triangular pockets of ne-grained fault rock protruding into the damage zone in some locations on
the left hand side of the fault core, where the fracture intensity is the highest.
spacing value, s, is dened as a fraction of the spacing distance
occupied by the pores:
4 lht=V t=s (2)
Then, the fracture density is given as
r h=twh=s (3)
Based on Eqs. (2) and (3), we nd that the fracture density as
dened here is proportional to height or bed thickness over spacing
and is dimensionless. The fracture density of, for example, 1.0,
corresponds to a commonly observed average spacing for one set of
opening mode fractures in a bed, for which the average spacing, s,
scales with the bed thickness, h, for well-developed fracture
systems (Wu and Pollard, 1995; Bai and Pollard, 2000). Thus, for
a bed thickness of 5 cm, the average spacing is 5 cm. For two sets of
overlapping fracture systems of equal density in a bed, the value for
the density approaches 2.0, and the average spacing is h/2. For a bed
of 5 cm thick, the equivalent spacing for the two overlapping sets is
2.5 cm.
5.1. Compliance matrix and the corresponding Youngs and shear
moduli components
The quasi-static equation of elasticity using Voigt notation is
(Nye, 1985; Pollard and Fletcher, 2005):
Fig. 9. Conceptual model showing linkage and coalescence of fault segments or strands which result in longer segment lengths, reduced number of fault steps, larger step sizes, and
wider damage zones and fault rock zones with increasing fault slip. From de Joussineau and Aydin (2007).
Fig. 10. Fracture spacing distributions: (a) for the 14 m fault and (b) across an area which included many faults with an aggregate slip of a few hundred meters. The line ts to the
data at the smallest ends of the spacing range are also shown. The smallest spacings dened by these linear trends are taken as the critical values below which the systems are
thought to be unstable. Slightly changed from de Joussineau and Aydin (2007).
0
B
B
B
B
B
B
@
3
11
3
22
3
33
3
23
3
31
3
12
1
C
C
C
C
C
C
A
0
B
B
B
B
B
B
@
S
11
S
12
S
13
S
14
S
15
S
16
S
21
S
22
S
23
S
24
S
25
S
26
S
31
S
32
S
33
S
34
S
35
S
36
S
41
S
42
S
43
S
54
S
45
S
46
S
51
S
52
S
53
S
54
S
55
S
56
S
61
S
62
S
63
S
64
S
65
S
66
1
C
C
C
C
C
C
A
0
B
B
B
B
B
B
@
s
11
s
22
s
33
s
23
s
31
s
12
1
C
C
C
C
C
C
A
(4)
where 3 and s are the six independent components of strain and
stress, respectively, and S is the symmetric 6-by-6 compliance
matrix. The numbers 1, 2, 3 always indicate Cartesian axes (say, x, y,
z respectively). Elastic extension in the x- or 1-direction is denoted
by 3
11
, etc., while a shearing (torsion or twisting) strain around the
x- or 1-axis is represented by 3
23
, etc. Similarly, the normal stress or
tension in the x-direction is s
11
, and the shear stress around the
x-axis is symbolized by s
23
, etc.
For any system, the full compliance matrix (Eq. (4)), or its
inverse, the stiffness matrix, has six eigenvectors, each of which is
a 1 6 matrix and is associated with a scalar eigenvalue. If S is the
matrix, v is the eigenvector, and c is the eigenvalue, then by de-
nition Sv cv. This means that when matrix S is multiplied by
vector v, the result is a vector proportional to v, and the constant of
proportionality is the eigenvalue c. There are always 6 distinct
eigenvectors. However, eigenvalues may or may not all be distinct.
For an isotropic system, ve of these eigenvalues are for shearing
modes and one is for pure compression/tension mode. Of the ve
shearing modes, three are the independent torsional or twisting
motions and/or the corresponding stresses; for example, in an
isotropic system, 3
23
, couples simply to s
23
, while all the off--
diagonal compliances and/or stiffnesses involving subscripts 4, 5, 6
vanish identically. Two other types of shear modes are eigenmodes
for an isotropic system; for example, when s
22
s
11
, we have
a comparable push-pull or pure shear mode resulting in the
eigen-response 3
22
3
11
for the strain. For the isotropic case,
there are three distinct versions of these pure shear behaviors that
give analogous results, but for nonisotropic systems usually only
one of these will actually be an eigenmode the most common
example of this behavior being for transversely isotropic systems.
In the presence of a set of perfectly parallel fractures in an
otherwise isotropic elastic medium, the elastic matrix becomes
transversely isotropic. The plane of the parallel fractures is the
plane of symmetry, and the direction perpendicular to this plane is
the axis of symmetry. Elastic behavior strictly within the plane of
symmetry (i.e., two-dimensional behavior in this plane) remains
isotropic, which is the origin of the term transverse isotropy
this type of isotropic behavior thus occurring transversely to the
axis of symmetry.
When analyzing such systems in three-dimensional space, it is
common to choose the axis of symmetry to coincide with one of the
spatial axes, x, y, and z, or 1, 2, and 3, respectively. This choice makes
no difference to the nal results but makes some difference to the
level of difculty in obtaining those results. In particular, making
a good choice of axes can simplify the matrix of elastic coefcients
somewhat, so that, for a set of parallel fractures, we have
a compliance matrix in the Voigt (Nye, 1985) 6 6 matrix form of
the elastic tensor notation:
S
0
B
B
B
B
B
B
B
B
B
@
1
E11
n
12
E11
n
13
E33
n
12
E11
1
E22
n
23
E33
n
13
E33
n
23
E33
1
E33
1
G44
1
G55
1
G66
1
C
C
C
C
C
C
C
C
C
A
(5)
Thus, for orthorhombic symmetry, the diagonal components of
the matrix; the Youngs moduli E
11
, E
22
, and E
33
and the shear
moduli G
44
, G
55
, and G
66
are inversely related to these diagonal
components. Note that the zero matrix elements were left blank
in Eq. (5) for simplicity as is standard practice. A particular
modulus that we call qGp, for quasi-pure shear mode is also
calculated because it is likely to play a role in the failure of the
systems we examine here, and it is well known in the geological
sciences. The mode qGp is actually an eigenvector of the system
considered, but its physical interpretation is not simple, because it
is not exactly any one of the six standard modes of a simple elastic
system pointed out earlier. However, its behavior is very close to
that of a pure shear mode and that is why the term quasi is used
here.
In calculating these components of the effective elastic moduli
for a medium with Poissons ratio of 0.4375 appropriate for sand-
stone, which has two fracture sets (Fig. 11a), we followan approach
based primarily on layer averaging methods of Backus (1962) and
Schoenberg and Muir (1989). The details of the mathematical
analysis of the effective properties of such a composite system are
given by Berryman and Aydin (in press). Basically, two different
layers each containing one set of fractures with the same density (r)
but possibly differing distributions, are considered for the effective
moduli calculations (Fig. 11b). After constructing one layer with one
of the fracture sets, this layer is rotated in such a way that the
combined fracture system will have the desired angle between the
Fig. 11. (a) A simplied and idealized fracture pattern at fault steps and within inner
damage zones. The pattern is dened by the angle (F
F
) between the two fracture sets
and lengths (l) and density (r) of the fracture systems. Fracture density is proportional
to the ratio of layer thickness or fracture height (h) to average spacing (s). (b,c) Block
diagrams showing the procedure to represent layers with each fracture set and the
congurations of the layers for averaging the effective properties in z- and x-directions
or 3- and 1-directions, respectively. The block diagrams represent stacking up layers
vertically in z-direction ((b) sandwich conguration) and arranging the layers side by
side in x-direction ((c) contiguous conguration).
two fracture sets. This is done by rotating each layer plus/minus one
half of the angle between the two fracture sets. In this paper, we
investigate cases where the angles between the two fracture sets
(F
F
) are 15
, 30
, 45
and 60
. The two layers are either stacked up,

which we call sandwich conguration (Fig. 11b) or are placed side
by side, which we call contiguous conguration (Fig. 11c). The
former is used for averaging in the z- or 3-axis whereas the latter is
used for averaging along the x- or 1-axis. Another side by side
contiguous conguration similar to that in Fig. 11c is used for
averaging along the y- or 2-direction. Both the contiguous and
sandwich congurations represent interacting but not intersecting
fracture arrays.
Fig. 12 shows the plots of the calculated effective moduli; (a) for
the Youngs moduli, (b) for the shear moduli, and (c) for the quasi-
shear moduli for pure shear for each of the four different fracture
congurations dened by the angle F
F
15
, 30
, 45
, and 60
and
for a range of fracture densities from 0 to 0.2. This range of density
is constrained by the availability (from previous work of Berryman
and Grechka, 2006) of the fracture inuence coefcients which are
required for the effective property calculations. The results show
that the E
11
components of the Youngs moduli for all four fracture
congurations do not change at all (all four lines overlap along the
top blue line in Fig. 12a) with increasing fracture density up to 0.2.
This is because E
11
corresponds to uniaxial loading in the x- or
1-direction and the changes of the angles and densities of fractures,
as seen from this direction edge-on, make no difference on the
effective moduli. The E
22
and E
33
components show systematic
decrease for all congurations as the fracture densities increase. We
note that the E
22
for the conguration F
F
60
and 45
experiences
greater decreases for the range of densities than fracture sets with
other intersection angles, whereas E
33
shows greater decreases for
congurations with smaller intersection angles, F
F
15
and 30
.
For example, the effective Youngs modulus, E
33
, corresponding to
F
F
15
at a fracture density of 0.2 shows about 34% reduction

with respect to the value for a medium without any fractures
(r 0). We note that the plots for E
22
and E
33
have segments with
different slopes indicating the nonlinear nature of the moduli
variations as the density increases. Since the change of slope occurs
at the fracture densities for which the calculations were performed,
the change would have been smoother if more runs with inter-
mediate fracture density values were performed.
The plots for the shear moduli components for the four fracture
congurations are given in Fig. 12b. They all are nearly linear except
one (G
44
-shearing about the 1- or x-axis). G
66
and G
55
(shearing
about the vertical 3- or z-axis and the 2- or y-axis, respectively) get
monotonically weaker for the density range used. However, these
moduli show the largest decreases for the fracture congurations
60
and 15
, respectively. The greatest decrease of the shear moduli

occur in the G
44
and G
55
components (shearing about the 1- or
x-axis and 2- or y-axis, respectively) corresponding to the fracture
congurations with the lowest angle, F
F
15
. However, this
decrease amounts to about 20% of the modulus for a medium
without any fractures. The curves for G
44
components have
a crossover at a fracture density, r, between 0.1 and 0.15. This
crossover is curious and remains to be investigated further.
Fig. 12c shows the variation in the effective quasi-shear modulus
for pure shear (qGp) as being one of the special cases. This
parameter shows a smaller variation of about 5% (with respect to
the modulus for the no-fracture state) for F
F
60
at the highest
fracture density (0.2) used in the calculations.
5.2. Extrapolations
Because the concept of elasticity is based on energy storage in
the elastic material/medium, there is an elastic conservative energy
associated with the elastic system. Each eigenvalue is a measure of
the elastic energy that can be stored in the system associated with
its elastic matrix. Since these stored energies must be positive
quantities, it follows that the eigenvalues themselves must all be
positive. If any elastic eigenvalue for a system vanishes, then this
means that it is impossible to store energy in this particular mode
and that the strain of the systemincreases without additional stress
if it is attempted to excite this mode. This condition denes
a mechanical instability in the system. So it is reasonable to use this
condition as one denition of mechanical system failure, and this is
why we look for the appearance of such failed modes in our analysis
of elastic system response. However, the vanishing values that we
want lie outside the range of values in our plots. This is because, as
pointed out earlier, the required fracture inuence coefcients are
not presently available for values outside of the range of densities
considered here.
The ways in which the shear and quasi-shear moduli vary with
increasing fracture density for each conguration in our models are
nearly linear. This may warrant extrapolations using the last
segment of the curves (for r between 0.1 and 0.2) to estimate the
critical fracture densities corresponding to the vanishing values of
shear moduli components. Hence, G
55
and G
44
plots for the inter-
section angle of 15
provide 0.9 for the critical density which is the

upper bound for cataclastic failure. The qGp for pure shear gives the
highest critical fracture density on the order of about 4, which
implies that the failure will occur earlier due to a more signicant
weakening of the other elastic parameters. Although the Youngs
moduli curves showing greater decreases (E
33
) appear to be highly
nonlinear, again the last linear segments are used to approximate
the densities corresponding to the vanishing value of this compo-
nent. For these cases the critical values of densities are close to each
other for each angular conguration being between 0.88 and 1.01.
Just to provide a spacing value to which the JSG reader can relate to:
These roughly correspond to a single set fracture spacing of about
5 cm for a 5 cm thick bed and 2.5 cm for two sets fracture spacing.
However it is possible that cataclastic failure may occur at spacings
lower than that for thinner layers and having the bedding interfaces
fail as part of the fragmentation process which has not been
considered in our analyses.
6. Discussion
The premise behind the present study is that the patterns of
fractures at strike-slip fault steps and inner damage zones adjacent
to fault rocks is complicated but can be simplied to represent the
local damage fairly well. Due to high density of fractures, the
mechanical properties of the rock masses at these locations are so
altered that a new paradigm is required to analyze the conditions
leading to the growth of faults by linkage of neighboring segments
and by enlargement of fault rock into the adjacent inner damage
zone through cataclastic deformation. The rationale for this
premise is three fold. First, laboratory studies on porous granular
rocks showa nearly linear trend between uniaxial strength and the
Youngs modulus (Palchik, 1999) and shear strength and the shear
modulus (Holt et al., 1987). Second, there is a theoretical basis in
linear elastic fracture mechanics (LEFM) for the fracture toughness
being equal to the Youngs modulus times the strain energy release
rate. Thus, the toughness is proportional to the Youngs modulus for
a given fracture extension though in homogeneous elastic medium
(Lawn and Wilshaw, 1975). Third, as has been referred to in this
paper, the compliance matrix for anisotropic elastic materials links
stresses to strains and should be crucial for determining the
deformation of rocks including their yielding or failure.
Thus, this premise separates the present study from those using
single linkage structure whether in opening, closing, or shearing
modes in pristine rock within the context of the LEFM (see for
example, Segall and Pollard, 1980; Du and Aydin, 1993, 1995; Crider
and Pollard, 1998; De Bremaecker and Ferris, 2004). In this regard,
the underlying reasoning in our approach is similar to that of the
damage concept of Lyakhovsky et al. (1997); and Lyakhovsky and
Ben-Zion (in review) if the fracture density is a proxy for the
damage parameter in their model. On the other hand, the linkage
criteria for the normal fault relays in map view investigated by
Soliva and Benedicto (2004) is not quite analogous for the strike-
slip fault congurations considered in our study for the simple
reason that the map viewof normal faults does not contain the slip
vector. However, the displacement/segment separation ratio,
which is a measure of the shear strain across fault steps, used by
these authors to characterize various stages of fault linkage may be
related to the fracture density used in our study for strike-slip fault
steps. It is likely that the fracture density is related to the shear
strain across the zone, however, the nature of such a possible
relationship between these parameters is not yet known.
It appears that the underlying mechanical principle for fault
growth processes and many of the related scaling relationships is
controlled by the stress concentration at fault tips and its length
dependence. First, for a simple mode II fracture, the stress
components at a point in the regions away from the fracture tips
decay as (a/r)
2
for 2D (Pollard and Segall, 1987) and (a/r)
3
for 3D
(Ben-Zion and Sammis, 2003), where a is half fault length and r is
the radial distance to the center of the fault. Second, fault interac-
tion is an important factor in the nal fault geometry of discon-
tinuous strike-slip faults (Aydin and Schultz, 1990). It turns out that
the relative locations of the neighboring fracture tips do have
a strong impact on echelon mode II fracture geometry but this
inuence is more or less independent of the sense of echelon steps.
The data on the step size and distribution are also consistent with
the earlier results in that larger steps are associated with longer
fault segments and, presumably, higher slip magnitudes.
Steps or relay ramps between echelon strike-slip faults include
various structures (joints, pressure solutions, other faults, and
folds) and eventually are cut by a through-going fault connecting
the echelon fault segments. In our study area, the failure structures
at steps and around the fault core are generally mode I fractures
formed either under tensile local stresses (Segall and Pollard, 1980)
or compressive local stresses (Horii and Nemat-Nasser, 1985).
However, shear bands (Aydin et al., 2006; Shipton and Cowie, 2003)
are occasionally observed at narrow contractional steps (Davatzes
et al., 2003), which are neglected in this study.
Earlier experimental and theoretical studies have proposed
buckling (Peng and Johnson, 1972) and bending (Renshaw and
Fig. 12. Plots showing the variation in elastic moduli of rocks with prescribed fracture patterns (dened by the angle, F
F
, between two fracture sets as 15
, 30
, 45
and 60
) and
a range of fracture densities (r from 0.0 to 0.2). (a) The three components of the Youngs moduli (E
11
, E
22
, and E
33
) with increasing fracture density. The nearly horizontal solid blue
line near the top represents all four E
11
plots for the density range and for all four angles between the fracture sets (the lines just overlap). (b) The three components of the shear
moduli (G
44
, G
55
, and G
66
) for the four angles with increasing fracture density. (c) The quasi-shear moduli for pure shear for the same four angular congurations and density range.
This moduli show the least decrease in magnitude with respect to that for no-fracture state.
Schulson, 2001) of thin and slender rock slabs between a set of
fractures as a mechanism for through-going shear fracture forma-
tion. It is difcult to identify these mechanisms in the eld. As the
eld data show, most fracture-bounded blocks have diamond
shapes due to dihedral intersection angle between the fracture sets
and may not be favorable either for buckling or bending. Rather, the
triangular areas at the intersection of the fractures appear to be
most prone to further fracturing and fragmentation. The rotation
and translation of some fracture-bounded rock blocks with respect
to the neighboring blocks can be identied in advanced stages of
deformation, particularly within slivers preserved within fault
cores. However, the relative timing of these rotations and relative
motions with respect to the shear zone evolution cannot be
determined.
The presence of multiple sets of fractures formed by splaying
within fault steps and inner damage zones with intersection angles
different than 90
, requires that all but the youngest set of fractures

are sheared. However, it is reasonable to assume that even
a sheared joint would be associated with a series of narrow
pull-aparts due to common joint surface roughness introduced by
hackle and rib marks. Thus, these slightly sheared joints would
have collinear open fractures along their lengths and, therefore,
they can be represented as a series of rectangular at fractures for
the purpose of the effective properties calculations.
The fracture intersection angles in the study area are controlled
by the splay angles (Fig. 5) which show a broad range of variation
(de Joussineau et al., 2007). This variation is in part due to potential
interaction between closely spaced subparallel faults. For example,
the average value of the splay angles is 19
for isolated single faults

whereas the average splay angle reaches 50
for closely spaced

interacting subparallel faults (Table 1a). It sufces to say that these
average values are well within the range (1560
) used in the
effective medium models in our study. However, the effective
properties based on the high end of the intersection angles
(4560
) are more appropriate for fractured slabs between closely

spaced subparallel faults.
The critical fracture density of about 1.0 corresponds to the
lowest vanishing values of shear moduli components (G
55
and G
44
)
for the smallest intersection angle (15
) and one of the Youngs

moduli components, E
33
, for all four intersection angles used in the
modeling study. Thus, for this density, the average spacing value for
one set of fractures is 5 cm for a 5 cm thick bed (see Eq. (3)) and
2.5 cm for two overlapping sets as deduced earlier. However, the
problem with the critical density or spacing values is that they can
never be captured in the eld because once the instability is
reached, the fracture systems are obliterated. We just dened an
envelope by tabulating the lowest fracture spacing without
disturbance or obliteration around a series of individual faults.
Comparing the theoretical results with the minimumspacing range
(w15 cm) dened by the rst linear segment of the spacing
distribution data from the study area (Table 1b), indicates that the
calculated and measured spacing values match quite well.
However, the way that fracture spacing was measured in the eld is
sensitive to the direction of scanline with respect to the fracture
orientation. Because the scanlines are approximately perpendicular
to the individual faults for single faults, and to the dominant faults
for a system of faults, one set of fractures is perpendicular to this
orientation and they would provide true spacing. Whereas, the
other fracture set would differ from the right angle orientation to
the scanline as much as the range of splay angles given in Table 1a.
Considering the average values for the most common splay angles
of 1950
, the true spacing values would defer from the measured

values by a factor of 0.95 and 0.64, respectively. However, the actual
difference between the calculated and measured fracture spacing
depends on the fracture angles from the scanline as well as their
frequency as illustrated for one of the well-studied faults, the 14 m
fault (Table 1, inset 2). Here more than 75% of the measured fracture
angles are at high-angle (5090
) to the scanline orientation

requiring a maximumcorrection factor of about 0.77. Evenwith this
correction, the theoretical and eld values for the critical fracture
spacing are in the same order of magnitude. These results may also
imply that the lack of a large number of fracture spacing under
a critical value of 12 cm in fault-related fracture spacing distri-
bution plots may be due to the destruction of those fractures with
spacing value equal to or smaller than that of the critical value. This
type of tapering in frequency plots is commonly interpreted in
terms of the minimum resolution and its impact on sampling error.
Albeit, considering the minimum resolvable spacing measurement
of 0.5 cm in our survey, it is likely that the initial tapering is not
because of limited sampling of this range of spacing but the
destruction of those fractures with spacing under a critical value.
We should also point out that cataclastic failure may occur at
fracture densities corresponding to nonzero values of the moduli
and having bedding interfaces fail as part of the fragmentation
process which has not been considered here. The quasi-shear
modulus for pure shear gives the highest critical fracture density of
slightly larger than 4 for one fracture set. In light of other results,
this implies that the failure will occur earlier due to weakening of
the other elastic parameters for a given fracture density.
7. Conclusions
In this paper, we conceptualized fault growth in terms of
increasing dimensions of fault zones with fault slip and provided
examples from the same structural and lithological setting. We
idealized complex fracture geometries at strike-slip fault steps and
inner damage zones in order to use the effective medium models
for gaining an insight in the inuence of the faulting-related
fractures on the mechanical properties of rocks along and around
the faults. Our results indicate that a signicant reduction
(w2034%) in most components of the Youngs, shear, and quasi-
shear (for pure shear) moduli of the fractured rock masses should
occur as the fracture density increases modestly. The extrapolated
values of the Youngs and shear moduli for the fracture congura-
tion which resulted in the greatest moduli reduction would suggest
an upper bound value for a critical density of about 1. This corre-
sponds to a critical fracture spacing value on the order of 5 cm for
a bed thickness of 5 cm for one set, and 2.5 cm for two overlapping
sets. The spacing values dened by the minimumvalues associated
with the line t to the lower ends of the measured spacing distri-
butions from the study area suggest that the undisturbed smallest
fracture spacing is about 12 cmwith a possible correction factor of
0.77, which still amounts to the same order of magnitude as the
modeling results. This also implies that the lack of a large number of
fracture spacing below 12 cm in fault-related fracture spacing
distribution plots may be due to the destruction of most fractures
with spacing under a critical fracture spacing value.
Acknowledgements
The works at the Valley of Fire State Park by many former
graduate students and postdocs who studied with A. Aydin at
Stanford University formed the foundation for establishing the
conceptual models in this paper. Among these, Ghislain de Joussi-
neaus work has been heavily relied upon. A partial list of other
students and postdocs includes R. Myers, E. Flodin, N. Davatzes, and
P. Eichhubl. A. Aydin is supported by the US DOE Basic Energy
Science, Division of Chemical Sciences, Geosciences and Bio-
Sciences, Grant no. DE-FG03-94ER14462. Work of J.G. Berryman
performed under auspices of the US DOE by the University of
California Lawrence Berkeley National Laboratory under contract
no. DE-ACO2-05CH11231. A. Aydin is grateful to the Valley of Fire
State Park personnel for their support of the eld campaigns
through many years. Comments by Christopher Wibberley, Roy
Schliche, Zoe Shipton, and Roger Soliva improved the manuscript.
References
Agosta, F., Aydin, A., 2006. Architecture and deformation mechanism of a basin
bounding normal fault in Mesozoic platform carbonates, central Italy. Journal of
Structural Geology 28, 24452467.
Aki, K., 1989. Geometric features of a fault zone related to the nucleation and
termination of an earthquake rupture, in: Proceedings of Conference XLV Fault
Segmentation and Controls of Rupture Initiation and Termination. US Geolog-
ical Survey Open File Report 89-315, pp. 19.
Aydin, A., 2000. Fractures, faults, and hydrocarbon entrapment, migration and ow.
Marine and Petroleum Geology 17, 797814.
Aydin, A., Nur, A., 1982. Evolution of pull-apart basins and their scale independence.
Tectonics 1, 91105.
Aydin, A., Schultz, R.A., 1990. Effect of mechanical interaction on the development of
strike-slipfaults withechelonpatterns. Journal of Structural Geology12, 123129.
Aydin, A., Borja, R.I., Eichhubl, P., 2006. Geological and Mathematical Framework for
failure modes in granular rock. Journal of Structural Geology 28, 8398.
Backus, G.E., 1962. Long-wave elastic anisotropy produced by horizontal layering.
Journal of Geophysical Research 67, 44274440.
Bai, T., Pollard, D.D., 2000. Fracture spacing in layered rocks: a new explanation
based on the stress transition. Journal of Structural Geology 22, 4357.
Barka, A., Kadinsky-Cade, K., 1988. Strike-slip fault geometry in Turkey and its
inuence on earthquake activity. Tectonics 7, 663684.
Ben-Zion, Y., Sammis, C.G., 2003. Characterization of fault zones. Pure and Applied
Geophysics 160, 677715. 00334553/03/040677-39.
Berryman, J.G., 2008. Elastic and transport properties in polycrystals of cracked
grains: cross-property relations and microstructure. International Journal of
Engineering Science 46, 500512.
Berryman, J.G., Aydin, A., Quasi-static analysis of elastic behaviour for some higher
density crack systems. International Journal of Numerical and Analytical
Methods in Geomechanics, in press, doi:10.1002/nag.874.
Berryman, J.G., Grechka, V., 2006. Random polycrystals of grains containing cracks:
Model of quasistatic elastic behavior for fractured systems. Journal of Applied
Physics 100, 113527.
Bohannon, R.G., 1983. Mesozoic and Cenozoic tectonic development of the
Muddy, North Muddy, and northern Black Mountains, Clark County, Nevada.
In: Miller, D.M., Todd, V.R., Howard, K.A. (Eds.), Tectonic and Stratigraphic
Studies in the Eastern Great Basin. Geological Society of America Memoir, 157,
pp. 125148.
Bristow, J., 1960. Microcracks and the static and dynamic elastic constants of
annealed and heavily cold-worked metals. British Journal of Applied Physics 11,
8185.
Budiansky, B., OConnell, R.J., 1976. Elastic moduli of a cracked solid. International
Journal of Solids and Structures 12, 8197.
Burgmann, R., Pollard, D.D., 1994. Strain accommodation about strike-slip fault
discontinuities in granitic rock under brittle-to-ductile conditions. Journal of
akir, M., Aydin, A., 1994. Tectonics and Fracture Characteristics of the Northern
Lake Mead, SE Nevada, Field Guide Book. Proceedings of the Stanford Rock
Fracture Project Workshop.
akir, M., Aydin, A., Campagna, D.J., 1998. Deformation pattern around conjoining
strike-slip faults systems in the Basin and Range, southeast Nevada: the role of
strike-slip faulting in basin formation and inversion. Tectonics 17, 344359.
Cartwright, J.A., Trudgill, B.D., Manseld, C.S., 1995. Fault growth by segment
linkage: an explanation for scatter in maximum displacement and trace length
data from the Canyonlands Grabens of SE Utah. Journal of Structural Geology 17,
13191326.
Childs, C., Manzocchi, T., Walsh, J.J., Bonson, C.G., Nicol, A., Schopfer, M.P.J., 2009. A
Structural Geology 31 (2), 117127.
Cowie, P.A., Scholz, C.H., 1992. Displacement-length scaling relationship for faults:
data synthesis and discussion. Journal of Structural Geology 14, 11491156.
Cox, S.J.D., Scholz, C.H., 1988. Rupture initiation in shear fracture of rocks: an
experimental study. Journal of Geophysical Research 93, 33073320.
Crider, J.G., Pollard, D.D., 1998. Fault linkage: three-dimensional mechanical inter-
action between echelon normal faults. Journal of Geophysical Research 103,
2437324391.
Davatzes, N.C., Aydin, A., 2004. Fault linkage and evolution of the fault core in
sandstone, Valley of Fire State Park, Nevada: a preliminary report. Rock Fracture
Project Workshop Vol. 15, E1E10.
Davatzes, N.C., Aydin, A., Eichhubl, P., 2003. Overprinting faulting mechanisms
during the development of multiple fault sets in sandstone, Chimney Rock,
Utah. Tectonophysics 363, 118.
Journal of Structural Geology 17, 607614.
De Bremaecker, J.-C., Ferris, M.C., 2004. Numerical models of shear fracture prop-
agation. Engineering Fracture Mechanics 71, 21612178.
de Joussineau, G., Aydin, A., 2007. The evolution of the damage zone with fault
growth and its multiscale characterization. Journal of Geophysical Research 112,
B12401,. doi:10.1029/2006JB004711.
de Joussineau, G., Aydin, A., 2009. Segmentation along strike-slip faults and their
self-similar architecture. Pure and Applied Geophysics 166, 15751594.
doi:10.1007/s00024-009-0511-4.
de Joussineau, G., Mutlu, O., Aydin, A., Pollard, D.D., 2007. Characterization of strike-
slip fault-splay relationships in sandstone. Journal of Structural Geology 29,
18311842.
Dewey, J.W., 1976. Seismicity of Northern Anatolia. Bulletin Seismological Society of
America 66, 843868.
Du, Y., Aydin, A., 1993. The maximum distortional strain energy density criterion for
shear fracture propagation with applications to the growth paths of en echelon
faults. Geophysical Research Letters 20, 10911094.
Flodin, E.A., Aydin, A., 2004. Evolution of a strike-slip fault network, Valley of Fire,
southern Nevada. Geological Society of America Bulletin 116, 4259.
doi:10.1130/B25282.1.
Flodin, E.A., Gerdes, M., Aydin, A., Wiggins, W.D., 2005. Petrophysical properties of
cataclastic fault rock in sandstone. In: Sorkhabi, R., Tsuji, Y. (Eds.), Faults, Fluid
Flow, and Petroleum Traps. American Association of Petroleum Geologists
Memoir, 85, pp. 197227.
Gamond, J.F., 1983. Displacement features associated with fault zones: a comparison
between observed examples and experimental models. Journal of Structural
Geology 5, 3345.
Grechka, V., Kachanov, M., 2006. Seismic characterization of multiple fracture sets:
Does orthotropy sufce? Geophysics 71, D93D105.
Grifth, W.A., Sanz, P.F., Pollard, D.D., 2009. Inuence of fault damage zone fractures
on the effective stiffness of fault damage zone rocks. Pure and Applied
Geophysics 166, 15951627, doi:10.1007/s00024-009-0519-9.
Harris, R.A., Day, S.M., 1999. Dynamic 3D simulation of earthquakes on en echelon
Harris, R.A., Archuleta, R.J., Day, S.M., 1999. Fault steps and the dynamic rupture
process: 2-d simulation of a spontaneously propagating shear fractures.
Geophysical Research Letters 18, 893896.
Holt, R.M., Ingsy, P., Mikkelsen, M., 1987. Rock mechanical analysis of North Sea
reservoir formations. SPE #16796. 62nd Annual Technical Conference, Dallas, TX.
Horii, H., Nemat-Nasser, S., 1985. Compression-induced microcrack growth in brittle
solids: axial splitting and shear fracture. Journal of Geophysical Research 90,
31053125.
Hull, J., 1988. Thickness-displacement relationships for deformation zones. Journal
of Structural Geology 10, 431435.
Kim, Y., Peacock, D.C.P., Sanderson, D.J., 2004. Fault damage zones. Journal of
Knott, S.D., Beach, A., Brockbank, P.J., McCallum, J.E., Welbon, A.I., 1996. Spatial and
mechanical controls on normal fault population. Journal of Structural Geology
18, 359372.
Lawn, B.R., Wilshaw, T.R., 1975. Fracture of Brittle Solids. Cambridge University
Press, 204 pp.
Le Pichon, X., S engo r, A.M.C., Demirbag, E., Rangin, C.,
_
Imren, C., Armijo, R.,
Go ru r, N., agatay, N., Mercier de Lepinay, B., Meyer, B., Saatilar, R., Tok, B.,
2001. The active Main Marmara Fault. Earth and Planetary Sciences Letters
192, 595616.
Lockner, D.A., Madden, T.R., 1991. A multiple-crack model of brittle fracture 1. Non-
time-dependent simulations. Journal of Geophysical Research 96, 19623
19642.
Lyakhovsky, V., Ben-Zion, Y., in review. Evolving fault zone structures in a damage
rheology model. Geochemistry, Geophysics, Geosystems, in review.
Lyakhovsky, V., Ben-Zion, Y., Agnon, A., 1997. Distributed damage, faulting, and
friction. Journal of Geophysical Research 102, 2763527649.
Martel, S.J., 1990. Formation of compound strike-slip fault zones, Mount Abbot
quadrangle, California. Journal of Structural Geology 12, 869882.
Martel, S.J., Peterson Jr., J.E., 1991. Interdisciplinary characterization of fracture
systems at the US/BK site, Grimsel Laboratory, Switzerland. International
Journal of Rock Mechanics and Mining Science and Geomechanical Abstracts
28, 259323.
Martel, S.J., Pollard, D.D., Segall, P., 1988. Development of simple strike-slip fault
zones in granitic rock, Mount Abbot quadrangle, Sierra Nevada, California.
Geological Society of America Bulletin 99, 14511465.
Myers, R., 1999. Structure and hydraulics of brittle faults in sandstone, PhD thesis,
Stanford University.
zones in sandstone. Journal of Structural Geology 26, 947966.
Nye, J.F., 1985. Physical Properties of Crystals: Their Representation by Tensors and
Matrices. Oxford Science Publications.
Odling, N.E., Harris, S.D., Knipe, R.J., 2004. Permeability scaling properties of
fault damage zones in siliclastic rocks. Journal of Structural Geology 26,
17271747.
Otsuki, K., Dilov, T., 2005. Evolution of self-similar geometry of experimental fault
zones; implications for seismic nucleation and earthquake size. Journal of
Geophysical Research 110, B03303. doi:10.1029/2004JB003359.
Pachell, M.A., Evans, J.P., 2002. Growth, linkage, and termination processes of
a 10-km-long strike-slip fault in jointed granite: the Gemini fault zone, Sierra
Nevada, California. Journal of Structural Geology 24, 19031924.
Palchik, V., 1999. Inuence of porosityandelastic modulus onuniaxial strengthinsoft
brittle porous sandstone. Rock Mechanics and Rock Engineering 32, 303309.
Peacock, D.C.P., Sanderson, D.J., 1991. Displacement, segment linkage and relay
ramps in normal fault zones. Journal of Structural Geology 13, 721733.
Peacock, D.C.P., Sanderson, D.J., 1995. Strike-slip relay ramps. Journal of Structural
Geology 17, 13511360.
Peng, S., Johnson, A.M., 1972. Crack growth and faulting in cylindrical specimens of
Chelmsford Granite. International Journal of Rock Mechanics. Mineral Sciences
and Geomechanics Abstracts 9, 3786.
Pollard, D.D., Fletcher, R.C., 2005. Fundamentals of Structural Geology. Cambridge
University Press, Cambridge, UK, 500 pp.
Pollard, D.D., Segall, P., 1987. Theoretical displacements and stresses near fractures
in rock: with applications to faults, joints, veins, dikes, and solution surfaces.
In: Atkinson, B.K. (Ed.), Fracture Mechanics of Rock. Academic Press, London,
pp. 277349.
Reches, Z., Lockner, D.A., 1994. Nucleation and growth of faults in brittle rocks.
Renshaw, C.E., Schulson, E.M., 2001. Universal behaviour in compressive failure of
brittle materials. Nature 412, 897899.
Rispoli, R., 1981. Stress elds about strike-slip faults inferred from stylolites and
tension gashes. Tectonophysics 75, 729736.
thickness. Mineral Engineering Transactions. American Institute of Mining
Engineering 35, 14261432.
Sayers, C.M., Kachanov, M., 1991. A simple technique for nding effective elastic
constants of cracked solids for arbitrary crack orientation statistics. Interna-
tional Journal of Solids and Structures 27, 671680.
Schlische, R.W., Young, S.S., Ackermann, R.V., Gupta, A., 1996. Geometry and scaling
relations of a population of very small rift-related normal faults. Geology 24,
683686.
Schoenberg, M., Muir, F.A., 1989. A calculus for nely layered anisotropic media.
Geophysics 54 581489.
bridge University Press, Cambridge, UK, 471 pp.
Segall, P., Pollard, D.D., 1980. Mechanics of discontinuous faults. Journal of
Geophysical Research 85, 43374350.
Segall, P., Pollard, D.D., 1983. Nucleation and growth of strike slip faults in granite.
Shaw, B.E., Dieterich, J.H., 2007. Probabilities for jumping fault segment stepovers.
Geophysical Research Letters 34, L01307. doi:10.1029/2006GL027980.
Shipton, Z.K., Cowie, P.A., 2003. A conceptual model for the origin of fault damage
zone structures in high-porosity sandstone. Journal of Structural Geology 25,
333344.
Sibson, R.H., 1985. Stopping of earthquake ruptures at dilational fault jogs. Nature
316, 248251.
Soliva, R., Benedicto, A., 2004. A linkage criterion for segmented normal faults.
Stein, R.S., Barka, A.A., Dieterich, J.H., 1997. Progressive failure on the North
Anatolian fault since 1939 by earthquake stress triggering. Geophysical Journal
International 128, 594604.
Stirling, M.W., Wesnousky, S.G., Shimazaki, K., 1996. Fault trace complexity,
cumulative slip, and the shape of the magnitude-frequency distribution for
strike-slip faults: a global survey. Geophysical Journal International 124,
833868.
Tokso z, N., Shakal, A.F., Michael, A.J., 1979. Space-time migration of earthquakes
along the North Anatolian fault zone and seismic gap. Pure and Applied
Geophysics 117, 12581270.
Walsh, J.J., Watterson, J., 1987. Distribution of cumulative displacement and seismic
slip on a single normal fault surface. Journal of Structural Geology 9, 10391046.
Watterson, J., 1986. Fault dimensions, displacement and growth. Pure and Applied
Wesnousky, S.G., 1988. Seismological and structural evolution of strike-slip faults.
Nature 335, 340342.
Wesnousky, S.G., 2006. Predicting the endpoints of earthquake ruptures. Nature
444, 358360.
Willemse, E.J.M., Peacock, D.C.P., Aydin, A., 1997. Nucleation and growth of strike-
slip faults in limestones from Somerset, U.K. Journal of Structural Geology 19,
14611477.
Wu, H., Pollard, D.D., 1995. An experimental study of the relationship between joint
spacing and layer thickness. Journal of Structural Geology 17, 887905.
Effects of internal structure and local stresses on fracture propagation, deection,
and arrest in fault zones
Agust Gudmundsson
a,
*
, Trine H. Simmenes
b,1
, Belinda Larsen
b
, Sonja L. Philipp
c
a
Department of Earth Sciences, Royal Holloway, University of London, Egham, Surrey TW20 0EX, UK
b
Department of Earth Science, University of Bergen, Norway
c
Geoscience Centre, University of Gottingen, Germany
Article history:
1 August 2009
Accepted 20 August 2009
Keywords:
Damage zone
Fault core
Crustal stresses
Toughness
Fractures
Crustal uids
a b s t r a c t
The way that faults transport crustal uids is important in many elds of earth sciences such as
petroleum geology, geothermal research, volcanology, seismology, and hydrogeology. For understanding
the permeability evolution and maintenance in a fault zone, its internal structure and associated local
stresses and mechanical properties must be known. This follows because the permeability is primarily
related to fracture propagation and their linking up into interconnected clusters in the fault zone. Here
we show that a fault zone can be regarded as an elastic inclusion with mechanical properties that differ
from those of the host rock. As a consequence, the fault zone modies the associated regional stress eld
and develops its own local stress eld which normally differs signicantly, both as regard magnitude and
orientation of the principal stresses, from the regional eld. The local stress eld, together with fault-rock
heterogeneities and interfaces (discontinuities; fractures, contacts), determine fracture propagation,
deection (along discontinuities/interfaces), and arrest in the fault zone and, thereby, its permeability
development. We provide new data on the internal structure of fault zones, in particular the fracture
frequency in the damage zone as a function of distance from the fault core. New numerical models show
that the local stress eld inside a fault zone, modelled as an inclusion, differ signicantly from those of
the host rock, both as regards the magnitude and the directions of the principal stresses. Also, when the
mechanical layering of the damage zone, due to variation in its fracture frequency, is considered, the
numerical models show abrupt changes in local stresses not only between the core and the damage zone
but also within the damage zone itself. Abrupt changes in local stresses within the fault zone generate
barriers to fracture propagation and contribute to fracture deection and/or arrest. Also, analytical
solutions of the effects of material toughness (the critical energy release rate) of layers and their
interfaces show that propagating fractures commonly become deected into, and often arrested at, the
interfaces. Generally, fractures propagating from a compliant (soft) layer towards a stiffer one tend to
become deected and arrested at the contact between the layers, whereas fractures propagating from
a stiff layer towards a softer one tend to penetrate the contact. Thus, it is normally easier for fractures to
propagate from the host rock into the damage zone than vice versa. Similarly, it is easier for fractures to
propagate from the outer, stiffer parts of the damage zone to the inner, softer parts, and from the stiff
host rock to the outer damage zone, than in the opposite directions. These conclusions contribute to
increased understanding as to how fractures propagate and become arrested within fault zones, and how
the fault zone thickness is conned at any particular time during its evolution.
1. Introduction
In recent years, there has been considerable geological work on
the internal structure of major fault zones (e.g., Byerlee, 1993;
Bruhn et al., 1994; Caine et al., 1996; Sibson, 1996; Evans et al., 1997;
Gutmanis et al., 1998; Sibson, 2003; Gudmundsson, 2004; Shima-
moto et al., 2004; Berg and Skar, 2005; Agosta and Aydin, 2006;
Faulkner et al., 2006; Bradbury et al., 2007; Li and Malin, 2008). This
work has partly focused on analysing the fault rocks themselves,
E-mail addresses: a.gudmundsson@es.rhul.ac.uk, rock.fractures@googlemail.
com (A. Gudmundsson).
1
Present address: StatoilHydro Research Center, Sandsliveien 90, 5020 Bergen,
Norway.
doi:10.1016/j.jsg.2009.08.013
their structure and mechanical properties, and partly on the
permeability structure and its maintenance in fault zones. This is
because of the importance that uid transport by fault zones has in
many elds of earth sciences. Inparticular, the in situ bulk hydraulic
characteristics of fault zones have been measured in boreholes (e.g.,
Ahlbom and Smellie, 1991; Barton et al., 1995; Fisher et al., 1996;
Braathen et al., 1999; Nativ et al., 1999; Lin et al., 2007; Tanaka et al.,
2007) and modelled (e.g., Barton et al., 1995; Lopez and Smith,
1995; Bredehoeft, 1997; Faulkner et al., 2006; Healy, 2008; Li and
Malin, 2008), the results suggesting that during non-slip periods
the damage zone is the main conductor of uids (cf. Gudmundsson,
2000; Gudmundsson et al., 2002).
Despite this work, the mechanical and permeability properties
of major fault zones, including associated fracture propagation in
the damage zone, are still not well understood, making it difcult to
construct realistic numerical models. This is partly due to major
fault zones being mechanically heterogeneous and, commonly,
layered parallel with the fault plane. Thus, Youngs modulus of
a fault zone is likely to vary signicantly with distance from the
fault plane itself, that is, from the core and through the various
subzones of the damage zone to the host rock (Gudmundsson,
2004; Gudmundsson and Brenner, 2003; Faulkner et al., 2006). As
a consequence, fault zones tend to develop local stresses, many of
which may be widely different from the associated regional stress
elds (Gudmundsson and Brenner, 2003). Variations in local
stresses are, in fact, universal features of mechanically layered
rocks, whether the layering is parallel with the fault plane, and thus
often steeply dipping or vertical, or gently dipping or horizontal as
is many sedimentary basins and composite volcanoes (Gud-
mundsson, 2006; Gudmundsson and Philipp, 2006). In a fault zone,
the local stress elds largely determine the fracture propagation
and arrest, and associated seismic events, and thereby much of the
fault-zone permeability.
This paper is on the internal mechanical structure of fault zones
and how it affects local stresses, fracture development and arrest.
The implications for fault-zone permeability are briey discussed,
but the focus is on the solid-mechanical aspects. In particular, the
paper has three main aims. The rst is to present results on the
internal structure of fault zones and how they function as general
elastic inclusions. The results derive fromeld studies of fault zones
of various types. A second aim is to present new numerical models
on the local stresses in fault zones. These models use eld
observations of internal structures of fault zones as a basis, focusing
on the effects that different fracture frequencies have in generating
subzones with different mechanical properties and local stresses
within the main fault zones. The third aim is to explore the reasons
why most fractures in fault zones remain short in comparison with
the strike dimension of the fault zone itself. The explanation offered
here is that the heterogeneous and anisotropic mechanical prop-
erties and local stresses within such fault zones, together with
numerous interfaces/discontinuities (contacts, existing fractures),
tend to deect and, commonly, arrest most of the fractures after
comparatively short propagation.
2. Internal structure of a fault zone
From a distance, fault zones appear as lineaments (Fig. 1).
Indeed, fault zones are commonly viewed as lineaments with little
or no internal structure and heterogeneity. As a consequence, fault
zones have for a long time been modelled as single, elastic cracks or
dislocations (Steketee, 1958; Press, 1965). While simple crack
models can be very useful for understanding faultfault interaction
and fault effects on regional stresses, they are less useful for
understanding the local stresses around and within the fault zone
itself. Since these local stresses largely control the slip and fracture
development and thus the permeability of the fault zone, the
internal mechanical structure of the fault zone must be considered
with a view of understanding its uid-transport properties.
Detailed eld observations of well-exposed fault zones show
that they normally consist of two main structural units, namely
a fault core and a fault damage zone (Fig. 2). The core takes up most
of the fault displacement and it is also referred to as the fault slip
zone (Bruhn et al., 1994; Sibson, 2003). Although the core contains
many small faults and fractures, its characteristic features are
breccias and cataclastic rocks. Commonly, the core rock is crushed
and altered into a porous material (Fig. 3) that behaves as ductile or
semi-brittle except at very high strain rates such as during seis-
mogenic faulting. In the core, there are commonly numerous veins
lled with secondary minerals spaced at centimetres or milli-
metres, that form dense networks. These networks, when trans-
porting uids, give the core a granular-media structure at the
millimetre or centimetre scale, thereby supporting its being
modelled as a porous medium.
While the eld description in this paper of the fault core and
damage zone focus on large fault zones, it should be emphasised
Fig. 1. View west, two parallel fault zones seen as lineaments (marked by arrows)
dissecting layers of limestone and shale in the Bristol Channel at Kilve, the Somerset
Coast, England. The distance between the faults at the location of the arrows is about
25 m.
Fig. 2. Schematic illustration of a fault core and fault damage zone of a (strike-slip)
fault. The core consists primarily of breccia and cataclastic rock (Fig. 3). The damage
zone, located on each side of the core, commonly contains some cataclastic rocks and
breccias but is characterised by numerous faults and fractures (Figs. 5 and 8), many of
which are eventually lled with secondary minerals (Fig. 4).
A. Gudmundsson et al. / Journal of Structural Geology 32 (2010) 16431655 1644
that the same units are seen in much smaller fault zones. In fact,
laboratory experiments on small rock samples may produce very
similar units, that is, a thin core and a thicker damage zone where
the frequency of fractures changes irregularly, but generally
decreases, with increasing distance from the core (Shimada, 2000).
In major fault zones, the thickness of the core is commonly from
several metres to a fewtens of metres (Fig. 3; Gudmundsson, 2004;
Berg and Skar, 2005; Agosta and Aydin, 2006; Tanaka et al., 2007; Li
and Malin, 2008). Very large faults zones, such as many transform
faults, may, however, develop several fault cores and damage zones
(Faulkner et al., 2006; Gudmundsson, 2007). The overall perme-
ability of the core is very low during most of the interseismic (non-
slip) period, so that the core commonly acts as a barrier to uid
ow, except during periods of high strain rates such as are associ-
ated with fault slip (Gudmundsson, 2000).
The damage zone, also referred to as the transition zone (Bruhn
et al., 1994), consists partly of lenses of breccias and other hetero-
geneities, and partly of sets of extension fractures, and to a lesser
degree, shear fractures (Gudmundsson et al., 2002). Many, and
presumably most, of the extension fractures are hydrofractures that
eventually become mineral-lled veins (Fig. 4). The fractures and
faults, and other discontinuities, generally make the damage zone
much more permeable than the fault core. For example, laboratory
measurements indicate that hydraulic conductivities in the damage
zone are as much as several orders of magnitude greater than those
of either the fault core or the host rock (Evans et al., 1997; Seront
et al., 1998).
There is normally not a sharp boundary between the damage
zone and the host rock. In the host rock, also referred to as the
protolith (Bruhn et al., 1994; Seront et al., 1998), the number of
fractures is generally less than that in the damage zone (Fig. 5;
Shimada, 2000). Although the boundaries between the fault core
and the damage zone are sharper than those between the damage
zone and the host rock, all these boundaries vary along the length
of the fault and change, in time and space, with the evolution of the
fault zone (Figs. 2 and 6).
Some general results of recent studies on the hydromechanical
properties of the fault core, the damage zone, and the host rock may
be briey summarised as follows. The damage zone is the main
conduit for ow of water along a major fault zone. Laboratory
measurements of small samples indicate a permeability of the
damage zone as much as 10,000-times higher than that of the core
or the host rock. Evans et al. (1997) suggest that while the labora-
tory samples from the core yield hydraulic conductivity values that
may not be much lower than the bulk in situ values, the laboratory
values of hydraulic conductivity for the damage zone are likely to
be considerably lower than the corresponding bulk in situ values.
This follows because the larger, highly conductive fractures that are
common in the damage zone (Figs. 4 and 5) are not represented in
the small, relatively non-fractured laboratory samples. This
conclusion is supported by reported in situ measurements giving
fault zone permeabilities as much as 1000-times greater than the
maximum laboratory values of Evans et al. (1997). Thus, the in situ
permeability difference between the core and the damage zone
may be even greater than the cited laboratory values would
indicate.
3. Local stresses in fault zones
From a mechanical point of view, a fault zone may be regarded
as an elastic inclusion (Fig. 6). As dened here, an elastic inclusion is
Fig. 3. Fault core and damage zone of a part of the Husavik-Flatey Fault, a transform
fault partly exposed on land in North Iceland (Gudmundsson, 2007). View west, the
10-m-thick core (see the person for scale) strikes N62
W and is mainly of breccias

(crushed basaltic lava ows), whereas the damage zone is characterised by tilted lava
ows (dipping 40
NW) and fractures of various sizes and types, many of which are
lled with secondary minerals (Fig. 4). Only a small part of the damage zone (which is
many hundred metres thick) is seen in the photograph.
Fig. 4. Mineral-vein network in a part of the damage zone of the Husavik-Flatey Fault
(Fig. 3). Some 80% of the veins are pure extension fractures, driven open by uid
overpressure (Gudmundsson et al., 2002).
Fig. 5. View southeast, an example of the variation in fracture frequency with distance
from the core of a fault in Vaksdal, West Norway. The highest number of fractures is at
the contact between the core and the innermost part of the damage zone. From there,
the fracture number decreases, in an irregular fashion, towards the host rock (gneiss),
at about 10 m from the core. Many fracture frequency proles of this type are provided
by Simmenes (2002) and Larsen (2002).
a three-dimensional body with elastic properties that differ from
those of the host material. More specically, an elastic inclusion is
a body with material properties that contrast with those of the
surrounding material, commonly referred to as the matrix, to
which the inclusion is welded.
The concept of an elastic inclusion as described here is well
established in the classical elasticity and rock mechanics literature
(Eshelby, 1957; Savin, 1961; Jaeger et al., 2007). The more recent
literature on micromechanics, however, uses inhomogeneities
rather than elastic inclusions for the concept dened above
(Nemat-Nasser and Hori, 1999; Qu and Cherkaoui, 2006). Here an
elastic inclusion denotes a material body hosted by a larger body
with different elastic properties (Gudmundsson, 2006), so that any
rock body, such as a fault zone, hosted by a larger body with
different properties is regarded as an inclusion.
The presence of an elastic inclusion modies the regional stress
eld so as to generate a local stress eld that operates both within
the inclusion and in its vicinity (Fig. 7). This follows because the
elastic properties of the inclusion, particularly its Youngs modulus
or stiffness, differ from those of the host rock. Thus, during any
loading (stress, displacement, or pressure), the responses of
the rocks constituting the inclusion differ from those of the
surrounding rocks. For example, if the rocks that constitute the
inclusion (the fault zone) are stiffer (higher Youngs modulus) than
the host rock, then the inclusion takes on most of the loading and
becomes subject to either relative tensile stresses (if the loading is
in extension) or compressive stresses (if the loading is in
compression). By contrast, if the inclusion rocks are more
compliant or softer than the host rock, then most of the loading is
taken up by the host rock which, thereby, develops locally high
relative tensile or compressive stresses depending whether the
loading is in extension or compression.
As is indicated above, many, and perhaps most, fault zones are
composed of a core and a damage zone that are widely different in
mechanical properties (Figs. 25). In addition, the damage zone
itself is commonly composed of subzones with different mechan-
ical properties, partly attributable to variations in fracture
frequencies (Figs. 5 and 8). Thus, the local stresses are likely to vary
not only between the host rock and the fault zone, or between the
core and the damage zone, but also within the damage zone itself.
To take the difference in stiffness between the host rock and the
fault zone into account, and how these change the local stresses of
fault zones, consider rst the model in Fig. 9. This model is based on
a normal fault zone in Vaksdal, close to Bergen in West Norway
(Fig. 8). The model divides the fault zone into four main subzones.
In the centre there is the fault plane itself, modelled as an internal,
compliant elastic spring. Based on estimates from open fractures
Fig. 7. Fault zone modelled as a simple elastic inclusion (Fig. 6). The nite-element
(www.Ansys.com; Zienkiewicz, 1977) model can be viewed either as a sinistral strike-
slip fault (lateral section) or as a normal fault (vertical section). Youngs modulus of the
fault zone is 1 GPa, a typical generalised value (Figs. 10 and 11), and that of the host
rock 40 GPa. The horizontal tension is 5 MPa, a value close to the maximum tensile
strength of solid rocks (Haimson and Rummel, 1982; Schultz, 1995; Amadei and
Stephansson, 1997), and may thus be regarded as a typical loading before fault slip in
active rift zones. The trends of the stress trajectories of s
3
(the minimum principal
compressive, maximum tensile, stress) change at the contact between the fault zone
and the host rock, indicating that the fault zone has a local stress eld different from
the regional eld of the host rock.
Fig. 8. Fracture frequency as a function of distance from the core of the fault modelled
in Figs. 9 and 10. The measurements are from a major normal fault zone in Vaksdal,
close to Bergen in West Norway (Simmenes, 2002). The inner part of the damage zone
has 24 fractures per unit area but at a distance of 20 m from the core, the outer part of
the damage zone has 14 fractures per unit area. Then at 80 m from the core, the
fracture frequency has fallen to 4 per unit area and is the same at a distance of 110 m;
these two latter areas are thus regarded as part of the host-rock fracture frequency.
Fig. 6. Schematic illustration of a fault zone as an elastic inclusion (inhomogeneity).
Normally, the elastic properties of the fault rock (damage zone and core) differ from
those of the host rock, so that there will be stress concentration around the fault zone,
as well as a local stress eld inside it. It is this local stress eld that controls fracture
development and fault slip in the fault zone and, therefore, largely its permeability.
(Gudmundsson and Brenner, 2003), the stiffness of the spring is
taken as 6 MPa m
1
. The stiffness of an elastic spring is determined
from a stressdisplacement curve, whereas Youngs modulus is
determined from a stressstrain curve. Thus, while Youngs
modulus has the units of (M)Pa, the spring has the units of
(M)Pa m
1
.
The fault plane is surrounded by the fault core, whose Youngs
modulus is taken as 0.1 GPa. This value is based on typical Youngs
moduli of unconsolidated rocks as well as in situ measurements
from various fault cores worldwide with common values between
0.1 and 1 GPa (Hoek, 2000; Schon, 2004). The Youngs modulus of
the inner damage zone is 1 GPa, that of the outer damage zone
10 GPa, and that of the host rock 50 GPa. These values reect the
decreasing number of fractures (Fig. 8) with increasing distance
from the inner damage zone to the host rock. The rock itself is
gneiss, but in accordance with well-known effects of fractures and
other cavities on Youngs modulus (Farmer, 1983; Priest, 1993;
Nemat-Nasser and Hori, 1999; Sadd, 2005), Youngs modulus is low
for the highly-fractured inner damage zone (Fig. 9), somewhat
higher for the less-fractured outer damage zone, and highest for the
normally fractured host rock.
The stressconcentration results (Fig. 10) indicate that, because
of the lower Youngs modulus inside the fault zone than outside it,
for the given loading conditions there will be lower von Mises shear
stresses in the fault zone than in the host rock. This may seem
surprising given that, when generalised, the fault slip is mostly
conned to the fault zone rather than the host rock. However, the
von Mises shear stresses reach the typical stress drops/driving
stresses for seismogenic fault slip, which are mostly 112 MPa
(Scholz, 1990), and the slip would occur in the fault zone simply
because it already has a weak fault plane and, most likely, a much
higher pore-uid pressure than the host rock. It is well-known that
tectonic earthquakes are usually related to zones of high-uid
pressure, so that, using the modied Coulomb criterion, the driving
shear stress for seismogenic fault slip, s, becomes:
s = 2T
0
f (s
n
P) (1)
where T
0
is the tensile strength of the rock, f is the coefcient of
internal friction, s
n
is the normal stress on the fault plane, and P is
the total uid pressure on the fault plane at the time of slip. When
the uid pressure approaches or equals the normal stress, the term
f(s
n
P) approaches or equals zero (for a higher uid pressure the
term may, in fact, become negative), so that the driving shear stress
for slip becomes 2T
0
. Since the in situ tensile strength of rocks is
commonly in the range of 0.56 MPa (Haimson and Rummel, 1982;
Schultz, 1995; Amadei and Stephansson, 1997), it follows that, for
high-uid-pressure fault zones, the driving shear stress for slip
should be 112 MPa, which is in agreement with common stress
drops (Kasahara, 1981; Scholz, 1990). Thus, even if the low-Youngs
modulus in the damage zone and core results in comparatively low
shear stresses in many active fault zones, they tend to slip because
of the existing weak fault plane (or planes), the high-uid-pressure
(and thus low friction), and the low effective normal stress on the
fault plane.
The local stresses in a fault zone do not depend only on the
stress magnitudes, but also on the directions of the stress vectors,
as represented by the trajectories of the principal stresses. The
model in Fig. 7 considers the fault as a single zone, an inclusion, but
as we have seen (Figs. 25, 8) there is commonly a signicant
difference in mechanical properties between the core and the
damage zone, as well as between the various subzones of the
damage zone itself. This is taken into account in the model in
Fig. 10, and also in the model below (Fig. 11).
In the model in Fig. 11 the fault zone is divided into ve
subzones. One, in the centre, represents the core of the fault zone
and has a Youngs modulus of 1 GPa, similar to many compliant or
soft breccias and unconsolidated rocks (Hoek, 2000; Schon, 2004).
Then comes the inner part of the damage zone, on either side of the
core, with a Youngs modulus of 5 GPa. This is, again, similar to the
Youngs modulus of many fractured rocks, as is indicated above. The
other part of the damage zone has a stiffness of 10 GPa, which is
similar to many fractured rocks where the fractures are not very
dense (Gudmundsson and Brenner, 2003). Finally, the host rock has
a Youngs modulus of 40 GPa, which is typical for many solid rocks
(Bell, 2000; Nilsen and Palmstro m, 2000).
The loading is extension oblique to the fault (Fig. 11). Viewed in
a vertical section, the loading would be appropriate for a reverse
fault, whereas viewed in a lateral section, the loading would be
appropriate for a dextral strike-slip fault. In either case, the oblique
loading combined with the variation in stiffness (Youngs modulus)
towards the centre of the fault (through the damage zone and to the
Fig. 9. Set-up of the model in Fig. 10 is largely based on the internal structure of the
fault zone in Fig. 8. The fault trends N40
E and is hosted by gneiss. The loading is EW

compressive stress of magnitude 15 MPa, as inferred from stress data from Norway
(Hicks, 1996). The fault plane (the fracture represented by a red, thick line) has
a stiffness of 6 MPa m
1
. The core has a Youngs modulus of 0.1 GPa, the inner damage
zone 1 GPa, the outer damage zone 10 GPa, and the host rock 50 GPa. The model is
fastened in the corners (indicated by crosses) so as to avoid rigid-body rotation and
translation (Simmenes, 2002).
Fig. 10. Boundary-element (Beasy, 1991) model showing the von Mises shear-stress
concentration (in MPa) around the fault zone in Fig. 9. The white line represents the
fault plane. Clearly, the fault core and inner damage zone have comparatively low
shear stress, 15 MPa, although high enough for slip, whereas the outer damage zone
has comparatively high shear stress. The stress transfer from the stiffer parts of the
damage zone, and into the host rock, is one way by which fault zones may grow
(through fracture formation) in thickness over time.
core) results in rotation of the principal stresses. Rotations of the
principal stresses within a fault zone are observed in the eld, and
are partly reected in the different trend of fractures close to the
fault plane or the core and away fromthe fault plane (in the damage
zone). One such example is provided by a fault in the Bristol
Channel of the UK (Fig. 12).
Models indicate that rotation of principal stresses is common in
layered rocks, whether the layers are subhorizontal, inclined, or
subvertical (Gudmundsson, 2006; Gudmundsson and Brenner,
2004; Faulkner et al., 2006; Gudmundsson and Philipp, 2006).
Rotation of the principal stresses between layers of different
mechanical properties is also common in the eld outside fault
zones, and is well demonstrated by changes in the orientation of
extension fracture (joints, mineral veins) between the layers
(Fig. 13).
The principal conclusion is that the fault zone develops a local
stress eld that controls its mechanical behaviour, slip, and
permeability. Because of the mechanical layering inside the fault
zone, the local stresses will vary between its subzones and, as
a consequence, there will rarely be uniformstresses over large parts
of the fault zone. It follows that stress eld homogenisation over
extensive parts of the fault zone, a necessary condition for large-
scale fault slip (Gudmundsson and Homberg, 1999), is only rarely
reached. Thus, most fault slips, both along the main fault itself as
well as along smaller faults in the damage zone, remain small.
Because of the variations in local stresses and rock properties, stress
elds favouring a particular type of fracture propagation (such as
a normal fault) are usually only reached within a comparatively
small region within a single subzone of the fault zone. It follows
that as soon as the particular fracture tries to propagate beyond
that stress-homogenised region, the fracture enters regions that
commonly have unfavourable local stresses, so that the fracture
propagation becomes deected and, often, arrested.
4. Fracture deection and arrest
Most fractures propagate for only very short distances before
they become arrested. This applies to crustal fractures in general,
and fractures within the damage zones of fault zones in particular.
The following terms are frequently used in the discussion below:
a discontinuity, a contact, and an interface. The last one, interface, is
the standard term in materials science for a contact between
dissimilar (or similar) material layers or phases (solid, gas, liquid)
(Sutton and Balluf, 1995). In earth sciences, a contact between
similar or dissimilar rocks has essentially the same meaning,
whereas a discontinuity denotes a contact or a fracture with
a negligible tensile strength (Priest, 1993). These terms will be used
as appropriate, with interface being the most general one.
When a propagating fracture meets an interface or a disconti-
nuity, such as a weak or open contact between dissimilar rocks or
an earlier fracture, the propagating fracture may do one of the
following (Fig. 14):
v become arrested so as to stop its propagation;
v penetrate the discontinuity;
v become deected along the discontinuity, in one or two
directions.
These three scenarios are well-known fromeld observations of
rock fractures. Here we propose that fracture deection at
a discontinuity, as well as fracture arrest, can be understood in
terms of three related parameters, namely:
v the induced tensile stress ahead of the propagating fracture tip;
v rotation of the principal stresses at the discontinuity;
v the material toughness or critical energy release rate of the
discontinuity in relation to that of the adjacent rock layers.
4.1. Local stresses at a discontinuity
For a homogeneous, isotropic material the fracture-induced
tensile stress ahead of and parallel to a propagating mode I crack, an
extension fracture such as are common in fault zones (Fig. 4;
Gudmundsson et al., 2002), is about 20% of the tensile stress ahead
of and perpendicular to the crack (Cook and Gordon, 1964; Thouless
and Parmigiani, 2007). Thus, the tensile stress induced by a fracture
may open up a discontinuity ahead of the fracture tip if the tensile
strength of the discontinuity is less than about 20% of the fracture-
perpendicular tensile strength of the adjacent rock layers. For an
average in situ rock tensile strength, 23 MPa (Haimson and
Rummel, 1982; Schultz, 1995; Amadei and Stephansson, 1997), the
discontinuity thus opens up if its tensile strength is less than 0.4
0.6 MPa. Since the minimum in situ tensile strength is about
0.5 MPa (Schultz, 1995), this is a possible mechanism for the
formation of T-shaped fractures at contacts (Fig. 15; Zhang et al.,
2007) and fracture arrest in heterogeneous fault zones and layered
rocks in general.
The deection of a fracture into a T-shape on meeting
a discontinuity such as a fracture or a contact (Figs. 14 and 15) is not
limited to discontinuities at right angles to the propagating frac-
ture. In fault zones, the fractures generally show a range in strikes
(Fig. 12), suggesting that earlier fractures may be oblique to, and
Fig. 11. Finite-element (www.Ansys.com; Zienkiewicz, 1977) model of the stress
trajectories of s
3
(the minimum principal compressive, maximum tensile, stress) in
a fault zone composed of a core with a Youngs modulus of 1 GPa, an inner damage
zone with a Youngs modulus of 5 GPa, and an outer damage zone with a Youngs
modulus of 10 GPa and located in a host rock with a Youngs modulus of 40 GPa. The
fault zone is subject to oblique loading, a tensile stress of 5 MPa. The trends of s
3
differ
between the core and the damage zone and the damage zone and the host rock, as well
as between the subzones of the damage zone. This model demonstrates the variation
in the local stresses that may occur within a typical layered fault zone.
contribute to the arrest of, subsequently formed fractures. Also, the
various contacts in heterogeneous fault zones will commonly make
an angle to a propagating fracture (Fig. 11). This aspect of the Cook
Gordon mechanism is best illustrated through simple numerical
models, using the Beasy program (Beasy, 1991; Brebbia and Dom-
inguez, 1992; www.beasy.com) where the weak discontinuity is
modelled as an internal spring, that is, a rock layer that behaves as
compliant, but still elastic, material (Fig. 9). Some modelling results
of an oblique discontinuity deecting, and presumably arresting,
the propagation of fractures are given in Fig. 16.
Experiments on dynamic crack propagation indicate that Cook
Gordon debonding is a common mechanism of fracture deection
and arrest, referred to as delamination in composite materials (Xu
et al., 2003; Xu and Rosakis, 2003; Wang and Xu, 2006). The results
suggest that it is primarily the tensile strength of the discontinuity
itself which determines if the debonding takes place (Wang and Xu,
2006). When the fracture-induced tensile stress has opened up the
discontinuity, the propagating fracture, on meeting the disconti-
nuity, may become deected along the discontinuity (Figs. 1416),
provided the stress eld is favourable to such a path change.
Experimental results (Xu et al., 2003) support the theoretical
results of He and Hutchinson (1989) in that a mode I fracture, such
as many mineral veins in fault zones (Fig. 4; Gudmundsson et al.,
2002), that becomes deected into the discontinuity changes into
a mixed-mode fracture.
The local stress elds in the layers on either side of a discon-
tinuity or layer contact may also decide whether a propagating
fracture becomes deected on meeting the discontinuity. The local
stress change, rotation and change in magnitude of the principal
stresses, may happen even if the layers are welded together and
results in the formation of a barrier to the propagation of fractures
of a certain type. Many studies have been made of barriers due to
principal-stress rotation in recent years (e.g., Gudmundsson, 2006;
Gudmundsson and Brenner, 2004; Faulkner et al., 2006; Gud-
mundsson and Philipp, 2006). The numerical examples for faults
zones provided here (Figs. 7, 10, 11) show that the stress trajec-
tories (directions) of s
3
(the minimum principal compressive,
maximum tensile, stress) and the stress magnitudes inside the
fault zone are different from those outside the fault zone and also
different between subzones of the damage zone (Figs. 10 and 11).
In addition, the contacts between the zones, particularly between
the core and the inner damage zone, may be such as to encourage
fracture deection and arrest (Figs. 1416). Thus, many fractures in
the damage zone may become deected and/or arrested on
meeting the contacts between the damage zone and the host rock
(Figs. 7 and 10), or between subzones of the damage zone (Figs. 10
and 11), because of properties of the contacts themselves and
because of changes in the magnitudes and rotation of the principal
stresses.
4.2. Toughness of a discontinuity
The CookGordon and stress rotation (stress barrier) mecha-
nisms cause many fractures in heterogeneous rocks in general, and
fault zones in particular, to become deected and/or arrested at
discontinuities and contacts between dissimilar rocks. But the
difference in material toughness between the interface/disconti-
nuity/contact and the adjacent rock layers is also of fundamental
importance. Generally, material toughness (critical energy release
rate) is a measure of the energy needed to propagate a fracture
through a material (Hull and Clyne, 1996; Chawla, 1998). Thus,
a tough material has a larger area under the stressstrain curve
before failure than a brittle material. Material toughness is dened
as the energy (in joule) absorbed per unit area of crack. The term
fracture toughness is sometimes regarded as synonymous with
material toughness (Hutchinson, 1996), but fracture toughness is
mostly used for the critical stress-intensity factor, K
c
(Broek, 1978;
Karihaloo, 1995). Thus, even if critical stress-intensity factor K
c
and
the critical strain energy release rate G
c
are related and are both
a measure of fracture resistance, they have different units and are
Fig. 12. Many hydrofractures (mineral veins) injected from the main fault plane (broken line) differ in trend from the fault plane. This is shown on the inset histogram, giving the
angle between the veins and the fault plane. This indicates that the local stresses in the damage zone, during vein injection, where different from those associated with the fault
plane and also that the stiffness variation may have been more irregular with distance than that in the modelled and observed faults zones in Figs. 5 and 811. Location: the Bristol
Channel at Kilve, the Somerset Coast, England.
best regarded as distinct. Here, fracture toughness denotes the
critical stress-intensity factor for a crack topropagate, withthe units
of stress (pascal) times square root of crack length, whereas material
toughness has the units of energy per unit area (Kobayashi, 2004;
Broberg, 1999; Anderson, 2005; Rice, 2006; Rice and Cocco, 2007).
The total strain energy release rate G
total
in a mixed-mode
fracture propagation is given by:
G
total
= G
I
G
II
G
III
=
1 n
2
K
2
I
E

1 n
2
K
2
II
E

(1 n)K
2
III
E
(2)
where G
IIII
are the material toughnesses for the ideal crack-
displacement modes IIII (Broberg, 1999; Anderson, 2005), E is
Youngs modulus (compliance or stiffness), n is Poissons ratio,
and K
IIII
are the associated stress-intensity factors. The critical
value of the stress-intensity K
c
denotes the fracture toughness.
Eq. (2) assumes plane-strain conditions; in the case of plane-
stress, the term (1 n
2
) =1. By their nature and loading, fractures
that become deected into discontinuities or interfaces
(contacts) are generally of a mixed-mode (Hutchinson, 1996;
Xu et al., 2003).
As regards pure crack-displacements, the opening (extension)
mode is denoted by I, the in-plane shear mode by II, and the out-of-
plane (anti-plane) shear mode by III (Broberg, 1999; Anderson,
2005). In geology, a mode I crack model is suitable for extension
fractures whereas mode II is suitable for many dip-slip faults
(normal and reverse) and mode III for strike-slip faults. All of these
fracture types and modes, IIII, are common in the damage zones of
fault zones (Figs. 35).
If the subzones or layers on either side of a discontinuity have
the same mechanical properties, such as is sometimes approxi-
mately the case in parts of a faults zone, the condition for an
extension fracture to penetrate the discontinuity (Fig. 14B) is that
the strain energy release rate G
p
, (with subscript p for penetration)
reaches the critical value for fracture extension, namely the mate-
rial toughness of the layer, G
L
(with subscript L for rock layer). Thus,
from Eq. (2) the conditions become:
G
p
=
1 n
2
K
2
I
E
= G
L
(3)
By contrast, the fracture will kink at or deect into the discon-
tinuity if the strain energy release rate reaches the material
toughness of the discontinuity itself, G
D
(with superscript D for
discontinuity). Since the fracture propagates in a mixed-mode
(mode I and II) along the discontinuity (Hutchinson, 1996; Xu et al.,
2003; Wang and Xu, 2006), it follows from Eq. (2) that deection
into the discontinuity occurs if:
G
d
=
1 n
2
K
2
I
K
2
II
= G
D
(4)
where the stress-intensity factors K
I
K
II
now refer to the discon-
tinuity. From Eqs. (3) and (4), the extension fracture penetrates the
discontinuity if:
G
d
G
p
<
G
D
G
L
(5)
but becomes deected into the discontinuity if:
G
d
G
p
_
G
D
G
L
(6)
Equations (3)(6) are likely to control, partly at least, whether
a fracture penetrates or becomes deected along a discontinuity,
such as a fracture or a contact, in some fault zones.
When there is an abrupt change in the mechanical properties at
interfaces such as contacts or discontinuities (Figs. 25, 8), an
elastic mismatch, the assumption of the rock layers on either side of
the interface being with the same properties is not warranted. The
magnitude of the mechanical change across a discontinuity or an
interface is commonly indicated by the Dundurs (1969) elastic
mismatch parameters. The two Dundurs parameters, a and b, may
be given as (cf. He and Hutchinson, 1989; Hutchinson, 1996; Freund
and Suresh, 2003):
a =
E
*
1
E
*
2
E
*
1
E
*
2
(7)
b =
1
2
m
1
(1 2n
2
) m
2
(1 2n
1
)
m
1
(1 n
2
) m
2
(1 n
1
)
(8)
where m is shear modulus, n is Poissons ratio, and the plain strain
Youngs modulus is E
*
=E/(1 n
2
). The subscript 2 is used for the
modulus of the rock hosting the fracture and subscript 1 for the
material on the other side (the far side with respect to the fracture
tip) of the discontinuity. Generally, a is a measure of mismatch in
the extensional or uniaxial stiffness and b in the volumetric or areal
stiffness (Freund and Suresh, 2003).
The strain energy release rate associated with fracture pene-
tration into the layer above the discontinuity, G
p
, is given by (He
and Hutchinson, 1989; He et al., 1994):
Fig. 13. View north, a section of limestone and shale layers in the Bristol Channel in
Wales. The joints (extension fractures, many with mineral llings) differ in trends
between the limestone layers, as indicated. For example, where the person is sitting, the
joints in the layer at her feet differ by about 20
from those in the next layer above, as

indicated by the joint-parallel arrows. Since extension fractures are perpendicular to s
3
(the minimum principal compressive, maximum tensile, stress), its trend in different
limestone layers is likely to have differed by this amount at the time of joint formation.
G
p
=
1 n
1
2m
1
K
2
I
=
1 n
1
2m
1
c
2
k
2
1
a
12l
(9)
where a is the length of fracture penetration (Fig. 14), k
1
is an
amplitude factor, proportional to the loading (here the driving
stress or uid overpressure for fracture propagation), l is real and c
a non-dimensional complex-valued functions, both of which
depend on the Dundurs parameters (Eqs. (7) and (8)). The strain
energy release rate associated with fracture deection into the
discontinuity or interface, G
d
, is given by (He and Hutchinson, 1989;
He et al., 1994):
G
d
= [(1 n
1
)=m
1
(1 n
2
)=m
2
[
K
2
I
K
2
II
.
4 cos h
2
p3
(10)
with
K
2
I
K
2
II
= k
2
1
a
12l
h
[d[
2
[e[
2
2R
e
(de)
i
(11)
where d and e are non-dimensional complex-valued functions that
depend on the Dundurs parameters. The ratio G
d
/G
p
is independent
of k
1
as well as the fracture-segment length a (Fig. 14) and is given
by (He and Hutchinson, 1989):
G
d
G
p
=
1 b
2
1 a

[d[
2
[e[
2
2R
e
(de)
c
2
(12)
By analogy with Eqs. (5) and (6), the fracture is likely to pene-
trate the discontinuity or interface between the dissimilar layers if:
G
d
G
p
<
G
D
(j)
G
1
L
(13)
but more likely to become deected into the discontinuity (and
often arrested) if:
G
d
G
p
_
G
D
(j)
G
1
L
(14)
the subscript for the material toughness being for layer 1 (Figs. 14
and 17) and j is a measure of the relative proportion of mode II to
mode I, namely, j =tan
1
(K
II
/K
I
) so that j =0
is for pure mode I

and j =90
for pure mode II.

For a given fracture-segment length a (Fig. 14), the energy
release rate depends on a (assume b =0), and the ratio G
d
/G
p
(Eqs.
(12)(14)) can be plotted as a function of a (Fig. 17). In the area
belowthe curves the ratio G
d
/G
p
favours deection of a fracture into
the discontinuity, whereas in the area above the curves the ratio
Fig. 14. On meeting an interface or discontinuity such as a contact, a fracture may (A) become arrested, (B) penetrate layer 1 above (or on the other or far side of) the contact, or
become doubly (C) or singly (D) deected into the discontinuity. In case C, the result is a T-shaped fracture (Fig. 15). Modied from Hutchinson (1996).
Fig. 15. T-shaped fracture at a contact between soft shale and stiff limestone in the
Bristol Channel in Wales. The fracture penetrates the cm-thick soft shale layer and then
becomes doubly deected (Fig. 14) to generate a T-shape.
favours vertical penetration of the fracture through the disconti-
nuity and into layer 1. Also, when stiffness of layer 1 is equal to that
of layer 2, the Dundurs parameter a =0 and G
d
/G
p
=0.26. Then
a fracture deects along the discontinuity or interface only if the
material toughness of the discontinuity (G
D
) is less than 26% of the
material toughness of layer 1 (G
L
1
). This latter condition is probably
uncommon, which may partly explain why fractures tend to
penetrate layered rocks, rather than become deected or arrested
at the layer contacts, where all the layers have similar mechanical
properties.
Furthermore, the curves in Fig. 17 show that the conditions for
a single-directed and a double-directed fracture propagation
(a T-shaped fracture, Fig. 15) along the discontinuity are very
similar for most values of a. Thus, for practical purposes, the
tendency for a fracture to be deected in one or two directions
along the discontinuity may be regarded as the same. When a is
negative, that is, the stiffness of layer 1 is less than that of layer 2,
there is generally much less tendency for deection of a fracture
along the discontinuity than when a is positive. When the positive
value of a increases and layer 1 becomes stiffer in relation to layer 2
(Eq. (8)), there is a greatly increased tendency for a fracture to
deect into the discontinuity.
5. Discussion
One of the principal results of this study is that fractures in fault
zones propagate only when and where the local stresses are
favourable to that type of fracture propagation (Fig. 18). Because
a fault zone is normally very heterogeneous as regards its
mechanical properties, the local stresses within the fault zone are
also likely to be heterogeneous and change abruptly from one part
of the zone to another. In particular, the commonly observed
mechanical layering inside a fault zone (Figs. 25, 8, 9) is likely to
result in local stresses that vary between its subzones (Figs. 10 and
11). It follows that uniform stresses over large parts of the fault
zone, a stress eld homogenisation (Gudmundsson and Homberg,
1999), which is a necessary condition for the propagation of large
fractures or fault slip along large parts of, or the entire, fault zone
are rarely reached. This is one reason why most fracture-propaga-
tion paths remain short and why most fault slips, both along the
Fig. 17. When a fracture meets an interface, the ratio of strain energy release rate for
fracture deection (G
d
) to that of fracture penetration (G
p
) controls the fracture
propagation. The ratio is here shows as a function of the Dundurs elastic mismatch
parameter a (Eqs. (7) and (8)). There is little difference in the elastic strain energy
release rate for a single or double deection (cf. Figs. 1416). For negative values of a,
layer 2 (the fracture-hosting layer) is stiffer than layer 1 and there is little tendency to
fracture deection along the interface. However, as the stiffness of layer 2 decreases in
relation to that of layer 1, the tendency to fracture deection along the interface greatly
increases. Modied from He et al. (1994).
Fig. 18. Fault zone is normally very heterogeneous as regards its mechanical proper-
ties, so that its local stresses are likely to be heterogeneous and change abruptly from
one part of the zone to another, as indicated schematically in this illustration.
Consequently, many fractures in the damage zone become arrested after a short
propagation when they enter layers, or meet with interfaces/contacts, that are
unfavourable to their propagation.
Fig. 16. Boundary-element (Beasy) model of a propagating fracture (here a hydro-
fracture) meeting with an oblique, weak interface or discontinuity (here a fault zone).
These results, showing the (asymmetric) opening and the tensile stresses in mega-
pascals, are completely general. If an extension fracture of any kind meets with a weak
interface (for example, a low-tensile strength discontinuity) at an angle, it tends to
open up the interface. If the interface trends perpendicular to the propagating
extension fracture, the opening is symmetric and may result in a T-shaped fracture
(Fig. 15, Gudmundsson, 2003); if the interface is oblique to the advancing extension
fracture, the opening is asymmetric, commonly resulting in a singly deected fracture
(Fig. 14D). Fracture propagation inside fault zones, particularly between layers within
the damage zone, between the damage zone and the core, and between the damage
zone and the host rock may become similarly deected and, commonly, arrested.
main fault itself as well as along smaller faults in the damage zone,
remain small.
Thus, because of the variations in local stresses and rock prop-
erties within fault zones, stress elds favouring a particular type of
fracture propagation (for example, an extension fracture) are
usually only satised within a comparatively small region or rock
volume at any one time (Fig. 18). A fracture formed, or slipping,
within that region will, as soon as it tries to propagate beyond that
region, enter fault zone parts with different properties and, nor-
mally, different local stresses. Commonly, these different stresses
are unfavourable to that type of fracture propagation (Fig. 18), so
that the fracture tends to become deected and, often, arrested.
One of the principal mechanismby which a propagating fracture
becomes deected and/or arrested at a discontinuity or an interface
is through meeting with stress barriers. Such barriers are simply
layers or rock units with local stresses that are unfavourable to the
propagation of the particular type of fracture. A barrier of this type
is primarily generated through rotation and changes in magnitude
of the principal stresses in the layer on the far side of an interface or
a discontinuity. Many studies have been made of the relevance of
this mechanism for fracture deection and arrest, and all indicate
that it is a viable and common mechanism in layered rocks (Gud-
mundsson, 2006; Gudmundsson and Brenner, 2004; Faulkner et al.,
2006; Gudmundsson and Philipp, 2006). The results of the
numerical models are supported by numerous eld observations
(Figs. 13 and 15). This mechanism is doubtless important for frac-
ture deection and arrest in heterogeneous and layered fault zones,
as is indicated by model results (Figs. 7, 10, 11). The mechanism is
likely to contribute to fractures in the damage zone becoming
deected and/or arrested on meeting the contacts between the
damage zone and the host rock (Figs. 7 and 10), or at contacts
between subzones of the damage zone (Figs. 10 and 11).
The conditions for fracture deection and arrest at interfaces
between dissimilar layers, however, depend on two additional
factors: rst, the mechanical properties of the interface itself in
relation to those of the adjacent rocks and, second, on the direction
of the fracture propagation in relation to the stiffnesses of the rocks
through which it propagates. Analytical solutions (Eqs. (5), (6) and
(13), (14)) indicate that the probability of fracture becoming
deected and/or arrested at an interface, rather than penetrating
the interface, depend on the ratios between the material tough-
nesses of the rock layer on the opposite side of the interface and
that of the interface itself in relation to the energy release rates
associated with fracture deection and penetration. In particular,
when the energy release rate ratio is below certain critical values,
deection into the interface is favoured whereas above those values
fracture penetration of the interface and into the layer on the other
side of it is favoured (Fig. 17).
Thus, a fracture propagating through a soft pyroclastic or sedi-
mentary layer towards a stiff basaltic lava ow would be more
likely to deect into the discontinuity than a fracture propagating
form a stiff lava ow towards a soft pyroclastic layer. This conclu-
sion is supported by many experiments on fracture propagation
and arrest at discontinuities between dissimilar layers (Kim et al.,
2006), and in geological analogue experiments (Kavanagh et al.,
2006). When the deection is not possible because of the orien-
tation of the principal stresses (Fig. 10), then the fracture propa-
gating from a soft towards a stiff layer would tend to become
arrested at the discontinuity. This is exactly what is commonly seen
in the eld (Figs. 13 and 15), that is, the discontinuity or contact acts
as a trap and arrests the fracture propagation.
The analytical results also indicate that a fracture propagating
from a layer with a lower Youngs modulus towards a layer with
a higher Youngs modulus has a strong tendency to become
deected along the contact or interface between the layers and,
commonly, arrested. By contrast, when the fracture propagates
froma high-Youngs modulus layer towards a low-Youngs modulus
layer, the fracture tends to penetrate the contact or interface. These
results are supported by materials-science experiments (Kim et al.,
2006) and numerous eld observations of rock fractures. For
example, it is common to see dykes and other rock fractures
propagating through relatively compliant (soft) rocks, such as
basaltic breccias and other pyroclastic rocks, to be deected and
arrested at contacts with stiffer rocks such as basaltic lava ows
(Gudmundsson, 2003).
These results apply equally well to fault zones and suggest that it
is more difcult for fractures to propagate from the compliant core
into the adjacent subzones of the damage zone (Fig. 12) than from
the damage zone into the core. By analogy, it is also easier for
fractures coming from the outer, stiffer parts of the damage zone to
penetrate the softer inner parts of the damage zone (Figs. 9 and 10)
than for fractures propagating in the opposite direction. Also, while
fractures may comparatively easily propagate from the host rock
into the damage zone, they would tend to become deected and/or
arrested when propagating fromthe damage zone rock towards the
host rock, thereby conning the fault zone thickness at any
particular time (Figs. 511).
6. Conclusions
The main conclusions of this study may be summarised as
follows:
v A fault zone may be regarded as an elastic inclusion with
mechanical properties that differ from those of the host rock.
The fault zone normally develops its own local stresses which
differ from the associated regional stresses.
v The local stresses of the fault zone and its heterogeneities and
interfaces and discontinuities (fractures, contacts) to a large
extent determine propagation, deection, and arrest of the
fractures in the fault zone.
v Numerical models, provided here, show that the magnitudes
and directions of the principal stresses inside a fault zone differ
signicantly from those in the host rock. For a mechanically
layered damage zone, there are abrupt changes in local stresses
not only between the core and the damage zone but also
between the layers or subzones of the damage zone itself.
v Abrupt changes in local stresses within the fault zone may
generate barriers to fracture propagation and contribute to
fracture deection and arrest at interfaces and discontinuities.
Arrested (layer-bound) fractures contribute little to perme-
ability development of a fault zone and the associated uid
transport.
v Analytical solutions on the material toughnesses of interfaces
such as discontinuities and contacts between mechanically
dissimilar layers within a fault zone show that fractures
commonly become deected into, and often arrested at,
interfaces.
v Fractures propagating froma softer layer towards a stiffer layer
tend to become deected and/or arrested at the contact
between the layers, whereas fractures propagating from a stiff
layer towards a soft one tend to penetrate the contact. It is thus
normally easier for a fracture to propagate from the damage
zone into the soft core than in the opposite direction.
v It is also normally more difcult for a fracture to propagate
from the inner, softer parts of the damage zone, to the outer
stiffer parts, and fromthe outer stiffer parts of the damage zone
into the host rock, than in the opposite directions.
v Fracture propagation, deection, and arrest at interfaces and,
alternatively, penetration of the interfaces, have large effects
on how fault damage zones and cores grow and change their
permeability and mechanical structure with time.
Acknowledgements
We thank Isabel Bivour, Gabriele Ertl, Oktawian Ewiak, Sonja
Geilert, and Kristine Nilsen for running some of the numerical
models and for help with the gures and the reviewers, Valerio
Acocella and Derek Keir, for very helpful comments. This work was
supported by the Norwegian Research Council Petromaks project
no. 163316/S30 Carbonate Reservoir Geomodels.
References
Agosta, F., Aydin, A., 2006. Architecture and deformation mechanism of a basin-
bounding normal fault in Mesozoic platform carbonates, central Italy. J. Struct.
Geol. 28, 14451467.
Ahlbom, K., Smellie, J.A.T., 1991. Overview of the fracture zone project at Finnsjo n.
Sweden J. Hydrol. 126, 115.
Amadei, B., Stephansson, O., 1997. Rock Stress and its Measurement. Chapman &
Hall, London.
Anderson, T.L., 2005. Fracture Mechanics: Fundamentals and Applications, third ed.
Taylor & Francis, London.
Barton, C.A., Zoback, M.D., Moos, D., 1995. Fluid ow along potentially active faults
in crystalline rock. Geology 23, 683686.
Beasy, 1991. The Boundary Element Analysis System User Guide. Computational
Mechanics, Boston.
Bell, F.G., 2000. Engineering Properties of Soils and Rocks, fourth ed. Blackwell,
Oxford.
Berg, S.S., Skar, T., 2005. Controls on damage zone asymmetry of a normal fault
zone: outcrop analyses of a segment of the Moab fault, SE Utah. J. Struct. Geol.
27, 18031822.
Braathen, A., Berg, S.S., Storro, G., Jaeger, O., Henriksen, H. and Gabrielsen, R., 1999.
Fracture-zone geometry and groundwater ow; results from fracture studies
and drill tests in Sunnfjord. Geological Survey of Norway, Report 99.017, 68 pp.
(in Norwegian).
Bradbury, K.K., Barton, D.C., Solum, J.G., Draper, S.D., Evans, J.P., 2007. Mineralogical
and textural analyses of drill cuttings from the San Andreas Fault Observatory
(SAFOD) boreholes: initial interpretations of fault zone composition and
constraints on geologic models. Geosphere 3, 299318.
Brebbia, C.A., Dominguez, J., 1992. Boundary Elements: An Introductory Course.
Computational Mechanics, Boston.
Bredehoeft, J.D., 1997. Fault permeability near Yucca Mountain. Water Resour. Res.
33, 24592463.
Broberg, K.B., 1999. Cracks and Fracture. Academic Press, New York.
Broek, D., 1978. Elementary Fracture Mechanics, second ed. Nordhoff, Leiden.
Bruhn, R.L., Parry, W.T., Yonkee, W.A., Thompson, T., 1994. Fracturing and hydro-
thermal alteration in normal fault zones. Pure Appl. Geophys. 142, 609644.
Byerlee, J., 1993. Model for episodic ow of high-pressure water in fault zones
before earthquakes. Geology 21, 303306.
structure. Geology 24, 10251028.
Chawla, K.K., 1998. Composite Materials: Science and Engineering, second ed.
Springer, Berlin.
Cook, J., Gordon, J.E., 1964. A mechanism for the control of crack propagation in all-
brittle systems. Proc. R. Soc. Lond., Ser. A282, 508520.
Dundurs, J., 1969. Edge-bonded dissimilar orthogonal wedges. J. Appl. Mech. 36,
650652.
Eshelby, J.D., 1957. The determination of the elastic eld of an ellipsoidal inclusion,
and related problems. Proc. R. Soc. Lond. A241, 376396.
Evans, J.P., Forster, C.B., Goddard, J.V., 1997. Permeability of fault-related rocks, and
implications for hydraulic structure of fault zones. J. Struct. Geol. 19, 13931404.
Farmer, I., 1983. Engineering Behaviour of Rocks, second ed. Chapman and Hall,
London.
rotation of regional stress in the fracture damage zone. Nature 444, 922925.
Fisher, A.T., Zwart, G., Ocean Drilling Program Leg 156 Scientic Party, 1996. Relation
between permeability and effective stress along a plate-boundary fault,
Barbados accretionary complex. Geology 24, 307310.
Freund, L.B., Suresh, S., 2003. Thin Film Materials: Stress, Defect Formation and
Surface Evolution. CUP, Cambridge.
Gudmundsson, A., 2000. Active fault zones and groundwater ow. Geophys. Res.
Lett. 27, 29932996.
Gudmundsson, A., 2003. Surface stresses associated with arrested dykes in rift
zones. Bull. Volcanol. 65, 606619.
Gudmundsson, A., 2004. Effects of Youngs modulus on fault displacement. C.R.
Geosci. 336, 8592.
Gudmundsson, A., 2006. Howlocal stresses control magma-chamber ruptures, dyke
injections, and eruptions in composite volcanoes. Earth Sci. Rev. 79, 131.
Gudmundsson, A., 2007. Infrastructure and evolution of ocean-ridge discontinuities
in Iceland. J. Geodyn. 43, 629.
Gudmundsson, A., Brenner, S.L., 2003. Loading of a seismic zone to failure deforms
nearby volcanoes: a new earthquake precursor. Terra Nova 15, 187193.
Gudmundsson, A., Brenner, S.L., 2004. How mechanical layering affects local
stresses, unrests, and eruptions of volcanoes. Geophys. Res. Lett. 31.
doi:10.1029/2004GL020083.
Gudmundsson, A., Homberg, C., 1999. Evolution of stress elds and faulting in
seismic zones. Pure Appl. Geophys. 154, 257280.
Gudmundsson, A., Philipp, S.L., 2006. How local stress elds prevent volcanic
eruptions. J. Volcanol. Geotherm. Res. 158, 257268.
Gudmundsson, A., Fjeldskaar, I., Brenner, S.L., 2002. Propagation pathways and uid
transport of hydrofractures in jointed and layered rocks in geothermal elds. J.
Volcanol. Geotherm. Res. 116, 257278.
Gutmanis, J.C., Lanyon, G.W., Wynn, T.J., Watson, C.R., 1998. Fluidowinfaults: a study
of fault hydrogeology inTriassic sandstone and Ordovician volcaniclastic rocks at
Sellaeld, north-west England. Proc. Yorkshire Geol. Soc. 52, 159175.
Haimson, B.C., Rummel, F., 1982. Hydrofracturing stress measurements in the
Iceland research drilling project drill hole at Reydarfjordur. Iceland J. Geophys.
Res. 87, 66316649.
He, M.Y., Hutchinson, J.W., 1989. Crack deection at an interface between dissimilar
elastic materials. Int. J. Solids Struct. 25, 10531067.
He, M.Y., Evans, A.G., Hutchinson, J.W., 1994. Crack deection at an interface
between dissimilar elastic materials: role of residual stresses. Int. J. Solids
Struct. 31, 34433455.
Healy, D., 2008. Damage patterns, stress rotations and pore uid pressure in strike-
slip fault zones. J. Geophys. Res. 113, B12407.
Hicks, E.C., 1996. Crustal Stresses in Norway and Surrounding Areas as Derived from
Focal Mechanisms Solutions and in situ Stress Measurements. Cand. Scient.
Thesis, University of Oslo, Oslo.
Hoek, E., 2000. Practical Rock Engineering. http://www.rockscience.com.
Hull, D., Clyne, T.W., 1996. An Introduction to Composite Materials, second ed. CUP,
Cambridge.
Hutchinson, J.W., 1996. Stresses and Failure Modes in Thin Films and Multilayers.
Notes for a Dcamm Course. Technical University of Denmark, Lyngby, pp. 145.
Jaeger, J.C., Cook, N.G.W., Zimmerman, R.W., 2007. Fundamentals of Rock
Mechanics, fourth ed. Blackwell, Oxford.
Karihaloo, B.L., 1995. Fracture Mechanics and Structural Concrete. Longman, Brunt
Mill, Harlow, UK.
Kasahara, K., 1981. Earthquake Mechanics. CUP, Cambridge.
Kavanagh, J.L., Menand, T., Sparks, R.S.J., 2006. An experimental investigation of sill
formation and propagation in layered elastic media. Earth Planet. Sci. Lett. 245,
799813.
Kim, J.W., Bhowmick, S., Hermann, I., Lawn, B.R., 2006. Transverse fracture of brittle
bilayers: relevance to failure of all-ceramic dental crowns. J. Biomed. Materials
Res. 79B, 5865.
Kobayashi, T., 2004. Strength and Toughness of Materials. Springer, Berlin.
Larsen, B., 2002. Fracture systems, active faults and groundwater potential on the
island of Bmlo, West Norway. MSc thesis, University of Bergen, Bergen, 150 pp.
(in Norwegian).
Li, Y.G., Malin, P.E., 2008. San Andreas Fault damage at SAFOD viewed with fault-
guided waves. Geophys. Res. Lett. 35, L08304.
Lin, A., Maruyama, T., Kobayashi, K., 2007. Tectonic implications of damage zone-
related fault-fracture networks revealed in drill core through the Nojima fault,
Japan. Tectonophysics 443, 161173.
Lopez, D.L., Smith, L., 1995. Fluid ow in fault zones: analysis of the interplay
between convective circulation and topographically driven groundwater ow.
Water Resour. Res. 31, 14891503.
Nativ, R., Adar, E.M., Becker, A., 1999. Designing a monitoring network for
contaminated ground water in fractured chalk. Ground Water 37, 3847.
Nemat-Nasser, S., Hori, M., 1999. Micromechanics: Overall Properties of Heteroge-
neous Materials, second ed. Elsevier, Amsterdam.
Nilsen, B., Palmstro m, A., 2000. Engineering Geology and Rock Engineering.
Norwegian Soil and Rock Engineering Association (NJFF), Oslo.
Press, F., 1965. Displacements, strains, and tilts at teleseismic distances. J. Geophys.
Res. 70, 23952412.
Priest, S.D., 1993. Discontinuity Analysis for Rock Engineering. Chapman and Hall,
London.
Qu, J., Cherkaoui, M., 2006. Fundamentals of Micromechanics of Solids. Wiley, NJ.
Rice, J.R., 2006. Heating and weakening of faults during earthquake slip. J. Geophys.
Res. 111, B05311.
Rice, J.R., Cocco, M., 2007. Seismic fault rheology and earthquake dynamics. In:
Handy, M.R., Hirth, G., Horius, N. (Eds.), Tectonic Faults: Agents of Chance on
a Dynamic Earth. The MIT Press, Cambridge, MA, pp. 99137.
Sadd, M.H., 2005. Elasticity: Theory, Applications, and Numerics. Elsevier,
Amsterdam.
Savin, G.N., 1961. Stress Concentration Around Holes. Pergamon, New York.
Scholz, C.H., 1990. The Mechanics of Earthquakes and Faulting. CUP, New York.
Schon, J.H., 2004. Physical Properties of Rocks: Fundamentals and Principles of
Petrophysics. Elsevier, Oxford.
Schultz, R.A., 1995. Limits on strength and deformation properties of jointed basaltic
rock masses. Rock Mech. Rock Eng. 28, 115.
Seront, B., Wong, T.F., Caine, J.S., Forster, C.B., Bruhn, R.L., 1998. Laboratory charac-
terisation of hydromechanical properties of a seismogenic normal fault system.
J. Struct. Geol. 20, 865881.
Shimamoto, T., Noda, H., Tanikawa, W., Wibberley, C.A.J., Uehara, S., 2004. Fault-
zone permeability structures and their implications for earthquake
mechanisms and geo-engineering problems. In: Ohnishi, Y., Aoki, K. (Eds.),
Contributions of Rock Mechanics to the New Century. Millpress Science, Rot-
terdam, pp. 10211026.
Shimada, M., 2000. Mechanical Behavior of Rocks Under High Pressure Conditions.
Balkema, Rotterdam.
J. Struct. Geol. 18, 10311042.
Sibson, R.H., 2003. Thickness of the seismic slip zone. Geol. Soc. Am. Bull. 93,
11691178.
Simmenes, T.H., 2002. Fracture systems: Fault development and uid transport in
Vaksdal, West Norway. MSc thesis, University of Bergen, Bergen, 143 pp.
Steketee, J.A., 1958. Some geophysical applications of the elasticity theory of
dislocations. Can. J. Phys. 36, 11681198.
Sutton, A.P., Balluf, R.W., 1995. Interfaces in Crystalline Materials. OUP, Oxford.
Tanaka, H., Omura, K., Matsuda, T., Ikeda, R., Kobayashi, K., Murakamai, M.,
Shimada, K., 2007. Architectural evolution of the Nojima fault and identication
of the activated slip layer by Kobe earthquake. J. Geophys. Res. 112, B07304.
Thouless, M.D., Parmigiani, J.P., 2007. Mixed-mode cohesive-zone models for
delamination and deection in composites. In: Srensen, B.F., Mikkelson, L.P.,
Lilhot, H., Goutianos, S., Abdul-Mahdi, F.S. (Eds.), Proceedings of the 28th Ris
International Symposium on Material Science: Interface Design of Polymer
matrix Composites, pp. 93111. Roskilde, Denmark.
Wang, P., Xu, L.R., 2006. Dynamic interfacial debonding initiation induced by an
incident crack. Int. J. Solids Struct. 43, 65356550.
Xu, L.R., Rosakis, A.J., 2003. An experimental study of impact-induced failure events
in homogeneous layered materials using dynamic photoelasticity and
high-speed photography. Optic Laser Eng. 40, 263288.
Xu, L.R., Huang, Y.Y., Rosakis, A.J., 2003. Dynamics crack deection and penetration
at interfaces in homogeneous materials: experimental studies and model
predictions. J. Mech. Phys. Solids 51, 461486.
Zhang, X., Jeffrey, R.G., Thiercelin, M., 2007. Deection and propagation of uid-
driven fractures at frictional bedding interfaces: a numerical investigation.
J. Struct. Geol. 29, 396410.
Zienkiewicz, O.C., 1977. The Finite Element Method. McGraw-Hill, New York.
Normal-fault development during two phases of non-coaxial extension:
An experimental study
Alissa A. Henza
*
, Martha O. Withjack, Roy W. Schlische
Department of Earth & Planetary Sciences, Rutgers University, 610 Taylor Road, Piscataway, NJ 08854-8066, USA
Article history:
24 June 2009
Accepted 26 July 2009
Available online 4 August 2009
Keywords:
Normal faults
Experimental modeling
Multiple phases of extension
Fault reactivation
a b s t r a c t
We use scaled experimental (analog) models to study the effect of a pre-existing fault fabric on fault
development during extension. In our models, a homogeneous layer of wet clay undergoes two non-
coaxial phases of extension whose directions differ by up to 45
. The normal faults that develop during

the rst phase create a pronounced fault fabric that inuences normal-fault development during the
second phase. In all models, the pre-existing faults are reactivated during the second phase of extension.
Their sense of slip depends on the angle between the rst- and second-phase extension directions.
Specically, the component of dip slip relative to strike slip decreases as the angle between the rst- and
second-phase extension directions increases. New normal faults also form during the second phase of
extension in all models. The number of new fault segments increases as the angle between the rst- and
second-phase extension directions increases. The orientations of the new normal-fault segments are
both orthogonal and oblique to the second-phase extension direction, indicating that both the second-
phase extension direction and the pre-existing fault fabric control the orientation of new fault segments.
Some of the new normal faults cut and offset the pre-existing faults, whereas others terminate against
them, producing complex fault patterns and interactions. The modeling results explain fault interactions
observed in the Jeanne dArc rift basin of offshore Newfoundland, Canada, and the reactivation of
abyssal-hill normal faults at outer highs near subduction zones.
1. Introduction
Many extensional provinces have undergone more than one
episode of deformation. For example, researchers have recognized
multiple phases of non-coaxial extension in the Jeanne dArc rift
basin (e.g., Enachescu, 1987; Tankard and Welsink, 1987; Sinclair,
1995a, b; Sinclair and Withjack, 2008), the North Sea (e.g., Badley
et al., 1988; Bartholomewet al., 1993; Frseth, 1996), Thailand (e.g.,
Morley et al., 2004, 2007), and the East African rift system (e.g.,
Strecker et al., 1990). Although these studies provide valuable
information about fault orientations and interactions in areas with
complex extensional histories, several critical questions remain.
How does the fault pattern that forms during an early episode of
extension affect the fault patterns that form during subsequent
episodes of extension? What factors determine whether faults are
reactivated or new faults form during subsequent deformation?
How do pre-existing normal faults affect the initiation, propaga-
tion, and geometry of newly formed normal faults? The goal of our
research is to address these critical questions. Specically, we use
scaled experimental (analog) models to simulate two phases of
non-coaxial extension and study how the angle between the
extension directions affects the resultant deformation patterns. We
also compare the modeling results with natural fault patterns in
regions that have undergone multiple phases of extension.
2. Experimental approach
2.1. Modeling materials
Most scaled experimental models of extension use either dry
sand or wet clay as the primary modeling material (e.g., Eisenstadt
and Sims, 2005; Withjack et al., 2007, and references therein).
Deformation patterns in sand and clay models of extension have
similarities and differences (Withjack and Callaway, 2000; Eisen-
stadt and Sims, 2005; Withjack and Schlische, 2006; Withjack et al.,
2007). In both sand and clay models, normal faults form that strike
90
to the extension direction. Fault-zone widths, however, are

much greater in sand models (>1.0 mm) than in clay models
(<0.1 mm). Also, most deformation is localized on a few major
normal faults in sand models, whereas deformation is distributed
* Corresponding author. Tel.:1 732 445 2125.
E-mail addresses: ahenza@rci.rutgers.edu (A.A. Henza), drmeow3@rci.
rutgers.edu (M.O. Withjack), schlisch@rci.rutgers.edu (R.W. Schlische).
doi:10.1016/j.jsg.2009.07.007
among major normal faults, minor normal faults, and folds in clay
models. In this study, we use wet clay as the modeling material to
provide a more detailed view of fault interactions and evolution.
The wet clay is similar to that used in other experimental modeling
studies (e.g., Withjack and Callaway, 2000; Eisenstadt and Sims,
2005; Withjack and Schlische, 2006; Withjack et al., 2007). It is
composed mainly of kaolinite particles (<0.005 mm in diameter)
and water (w40% by weight) and has a density of 1.551.60 g cm
3
.
Its coefcient of internal friction is w0.6, and its cohesive strength
is w 50 Pa.
To ensure dynamic similarity between the experimental models
and natural prototypes, two conditions must be satised (e.g.,
Hubbert, 1937; Nalpas and Brun, 1993; Weijermars et al., 1993).
First, the modeling material must have a similar coefcient of
friction to that of rocks in nature. For most sedimentary rocks, the
coefcient of friction ranges between 0.55 and 0.85 (e.g., Handin,
1966; Byerlee, 1978). Thus, the wet clay in our models (with
a coefcient of friction of w0.6) is a suitable modeling material.
Second, the models must obey the scaling relationship:
C
*
r
*
L
*
g
*
(1)
where C
*
, r
*
, L
*
, and g
*
are the model to prototype ratios for
cohesion, density, length, and gravity, respectively. In our models,
the value of r
*
is 0.7 and g
*
is 1. Thus, C
*
and L
*
must have similar
magnitudes to ensure dynamic similarity. In nature, C ranges from
less than 1 MPa (for loosely compacted sedimentary rocks) to more
than 10 MPa (for intact igneous or metamorphic rocks) (Handin,
1966; Schellart, 2000; and references therein). Additionally, C can
be signicantly less than 1 MPa for fractured rocks (e.g., Byerlee,
1978; Brace and Kohlstedt, 1980). As mentioned previously, the wet
clay in our models has a cohesive strength of w50 Pa, resulting in
a value of C
*
between 10
4
and 10
6
. Therefore, L
*
ranges between
10
4
and 10
6
in our models, depending on the cohesion of the
natural prototype. If the clay simulates a layer of loosely compacted
sedimentary rock, then 1 cm in the model represents w100 m in
nature. Alternatively, if the clay simulates intact crystalline rock,
then 1 cm in the model represents about w10 km in nature.
2.2. Experimental set-up
Our experimental set-up resembles that in previous experi-
mental models of oblique extension (e.g., Withjack and Jamison,
1986; Tron and Brun, 1991; McClay and White, 1995; Bonini et al.,
1997; Keep and McClay, 1997; Clifton et al., 2000). The base of the
apparatus consists of an 8-cm wide rubber sheet attached to two
rigid sheets (one xed and one mobile) (Fig. 1a). A 0.5-cm thick
layer of silicone polymer, with a viscosity of w10
4
Pa s (Weijermars,
1986), overlies the rubber sheet. A layer of wet clay covers the layer
of silicone polymer, the xed rigid sheet, and the mobile rigid sheet
(Fig. 1b). It is 3.5 cm thick above the layer of silicone polymer and
4.0 cm thick above the rigid sheets. During the experiments, the
mobile rigid sheet moves outward, stretching the attached rubber
sheet and the overlying silicone polymer (Fig. 1a). In response,
a deformation zone develops within the clay layer above the rubber
sheet and silicone polymer. The silicone polymer serves two func-
tions in our models: it localizes deformation within the clay layer
above the rubber sheet (e.g., Bellahsen et al., 2003), and it decou-
ples the clay layer from the rubber sheet (allowing the base of the
clay layer to move vertically).
Based on Withjack and Jamison (1986), a is the clockwise angle
measured from the trend of the deformation zone within the clay
layer to the displacement direction of the mobile rigid sheet
(Fig. 1a). Oblique deformation with both extensional and shear
components results when a s0
, a s90
, or a s180
(Withjack
and Jamison, 1986). Furthermore, when a s90
, the maximum
extension direction and displacement direction differ: the
maximum extension direction lies midway between the displace-
ment direction and the normal to the trend of the deformation zone
(see Withjack and Jamison (1986) for details). Previous models of
oblique extension (e.g., Withjack and Jamison, 1986; Tron and Brun,
1991; Clifton et al., 2000) have shown that only normal faults form
when 45
a 135
(Fig. 2a). For a <45
and a >135
, normal,
oblique-slip and/or strike-slip faults develop.
All models inthis study have two phases of deformation(Fig. 2b).
During the rst phase, the mobile sheet moves outward at a rate of
4 cmh
1
(1.1 10
3
cms
1
) in a prescribed direction (a
1
45
) for
a prescribeddisplacement (3.5 cm). Inresponse, a pervasive(but not
continuous) fabric consisting of normal faults develops throughout
the deformation zone in the clay layer. During the second phase, the
mobile plate moves outward at a rate of 4 cmh
1
in a different
prescribed direction (a
2
135
, 120
, 105
, or 90
; see Fig. 2b) for

aprescribeddistance(3.5 cm). Thus, theanglebetweentherst- and
second-phase displacement directions varies from 90
to 45
in our
models (Figs. 1, 2), whereas the angle between the rst- and second-
phase extension directions varies from 45
to 22.5
(Fig. 2b).
2.3. Model analysis
Photographs of the top surface of the clay layer, taken at regular
time intervals, record the surface deformation through time during
both phases of extension. To exclude edge effects, we analyze only
the central part of the top surface of the deformation zone. Offsets
of supercial linear markers on the top surface of the clay layer
indicate the sense of slip on faults during both phases of extension
(Fig. 3a). We conrm the sense of slip by observing corrugations on
the fault surfaces at the end of each experiment (Fig. 3b). Corru-
gations are grooves on fault surfaces that parallel the slip direction
(Granger, 2006; Granger et al., 2006, Granger et al., 2008). To
determine the numbers and orientations of fault segments, we t
straight lines to individual fault segments (Fig. 3c, d). Fitting lines to
fault segments, rather than drawing lines from fault tip to fault tip,
captures all segment orientations in areas where second-phase
faults have linked with rst-phase faults. To determine whether
deformation patterns changed signicantly with depth, we also
examined the bottom surface of a dried model with boundary
conditions identical to Model 1 (a
1
45
and a
2
135
).
Wet clay
Silicone polymer
3.5 cm 4 cm
8 cm Fixed rigid
sheet
Mobile rigid
sheet
Rubber sheet
a
60 cm
6
8
c
m
8 cm rubber sheet
Fixed
rigid
sheet
Mobile
rigid
sheet
b
Displacement
direction
Trend of
deformation zone
Fig. 1. Experimental set-up in (a) plan view and (b) cross section. Plan view set-up
shows denition of a (based on Withjack and Jamison, 1986).
A.A. Henza et al. / Journal of Structural Geology 32 (2010) 16561667 1657
3. Experimental observations
3.1. Large angle between extension directions
Model 1 has the maximum possible angular difference
between displacement directions (90
) for which only normal

faults form, and, therefore, the rst- and second-phase extension
directions are 45
apart (Fig. 2b). During the rst phase of

extension (a
1
45
), numerous faults form in the deformation

zone of the clay layer (Fig. 4a
1
). Offset markers on the top surface
of the clay layer and corrugations on the fault surfaces show that
these faults are normal faults. Their strike is orthogonal to the
Fig. 3. Methods used in fault analysis. (a) Photographs of the top surface of the clay layer of Model 1 showing supercial linear markers at the beginning of the experiment (left),
after the rst phase of extension (middle), and after the second phase of extension (right). Offsets of markers indicate sense of slip on faults. Arrows indicate extension direction for
each phase. (b) Photograph of slip-parallel corrugations on a fault surface. (c) Photograph of a segmented fault scarp (white on photograph). (d) Line drawing of segmented fault
showing straight-line t for each fault segment. Fitting lines to fault segments results in three distinct segments with three distinct orientations, whereas a tip-to-tip line yields one
fault with one orientation.
Model 1
= 45 = 135
= 45 = 90
= 45
= 45
= 120
= 105
Model 2
Displacement direction
Extension direction
E
1
E
2
E
1
E
2
E
1 E
2
E
1
E
2
= 45
1
st
phase displacement direction
2
nd
phase displacement direction
= 135
= 120
= 105 = 90
Range of values for which
normal faults form
b a
Model 3 Model 4
Fig. 2. (a) Diagram showing range of a values for which normal faults form (shaded area) and a values used in our models (a
1
denes the rst-phase displacement direction, and a
2
denes the second-phase displacement direction). (b) Displacement and extension directions for models (E
1
is the rst-phase extension direction, and E
2
is the second-phase
extension direction).
extension direction, and they dip w60
. The fault pattern at the

end of the rst phase of Model 1 resembles those in other models
of oblique extension with a 45
(e.g., Withjack and Jamison,

1986; Tron and Brun, 1991; McClay and White, 1995; Clifton et al.,
2000).
During the second phase of extension (a
2
135
), many of the
rst-phase normal faults are reactivated with oblique slip (right-
lateral and normal components) (Fig. 4a
2
). The strike-slip compo-
nent is signicantly larger than the dip-slip component (Table 1).
New normal faults also develop during the second phase of
Fig. 4. Photographs of the top surface of the clay layer showing central region of the models after rst and second phases of extension. Arrows indicate extension direction, and
dashed box shows location of line drawings in Fig. 5. Fault scarps dipping toward the top of the page appear bright; fault scarps dipping toward the bottom of the page appear dark.
extension (Figs. 4a
2
, 5a
2
). Most new normal faults initiate at rst-
phase faults and propagate away from them (Fig. 6). The strike of
the new normal faults ranges from orthogonal to oblique to the
second-phase extension direction, and many new faults change
orientation during fault propagation. These faults are initially
orthogonal to the pre-existing fault, and become approximately
orthogonal to the second-phase extension direction as they prop-
agate away fromthe pre-existing fault (Fig. 6). Where a newnormal
fault encounters a pre-existing fault, the new fault either cuts
across the pre-existing fault or terminates against it (Fig. 7). In
Model 1, more newnormal faults cut across pre-existing faults (61%
Table 1
Fault-population statistics for second-phase deformation.
Model
number
Angle between extension
directions during 1st and
2nd phases
% of Fault population
that form during
2nd phase
% of Dip-slip (normal)
motion on reactivated
faults during 2nd phase
1 45
42 43
2 37.5
33 50
3 30
27 65
4 22.5
12 70
Fig. 5. Summary of modeling results. Line drawings of fault heaves are from the central region of the models (locations shown in Fig. 4). Rose diagrams show orientations of fault
segments obtained using the method shown in Fig. 3. Arrows on rose diagrams show extension directions for each phase. Bin size for rose diagrams is 5
; the outside circle of each

rose diagram equals 50% of the length-weighted fault segment population; n is the number of fault segments in a 40 cm by 12 cm area on the top surface of the clay layer from the
central region of each model.
of total fault interactions) than terminate against them (Fig. 7d).
Generally, the new normal faults are signicantly shorter than
faults that form under identical conditions in the absence of a pre-
existing fault fabric (Fig. 8b). In addition, the displacement on most
of the new normal faults is greatest adjacent to the pre-existing
fault and decreases outward (Fig. 8a). The interactions between the
new faults and the rst-phase faults create a complex fault pattern
observable on both the top and bottom of the clay layer (Fig. 8c).
3.2. Moderate angles between extension directions
In Models 2 and 3 (which have 37.5
and 30
between the rst-

and second-phase extension directions, respectively), second-
phase deformation is accommodated by reactivating the pre-
existing normal faults and forming new normal faults (Fig. 4b
2
, c
2
).
In both models, many of the rst-phase faults are reactivated as
oblique-slip faults with both normal and right-lateral strike-slip
components. The magnitude of the normal component relative to
the strike-slip component is larger in Models 2 and 3 than in Model
1 (Table 1), increasing as the angle between the directions of the
rst- and second-phase extension decreases. New normal faults
formduring the second phase of extension of both models (Fig. 5b
2
,
c
2
), although fewer new fault segments develop in Models 2 and 3
than in Model 1 (Table 1). The strike of the new normal-fault
segments ranges from orthogonal to oblique to the second-phase
extension direction (Figs. 4, 5), and many new faults change
orientation as they propagate. As in Model 1, new normal faults
either cut across or terminate against the pre-existing faults. More
new normal faults cut across pre-existing faults in Model 2 than in
Model 3 (Fig. 7d).
3.3. Small angle between extension directions
In Model 4 (which has 22.5
between the rst- and second-

phase extension directions), fewnewnormal faults formduring the
second phase of extension (Figs. 4d
2
, 5d
2
; Table 1). Instead, reac-
tivation of the rst-phase normal faults accommodates most
2.4 cm
Orthogonal to
applied extension
direction
2.2 cm
2.0 cm
1.8 cm
1.6 cm
1c m
1.4 cm
Orthogonal to
first-phase fault
Fig. 6. Example of propagation of new normal faults during the second phase of
extension for Model 1. Arrows show second-phase extension direction. Line drawings
are from photographs of the top surface of the clay layer for progressively greater
values of displacement of the movable sheet. As the new fault propagates outward
from the pre-existing fault, its strike changes from orthogonal to the pre-existing fault
to orthogonal to the extension direction.
b
c
Pre-existing fault
New fault
a
b
22 ccmm
c
%
o
f
n
e
w
f
a
u
l
t
s
t
h
a
t
c
u
t
a
n
d
o
f
f
s
e
t
p
r
e
-
e
x
i
s
t
i
n
g
f
a
u
l
t
s
d
10
20
30
40
50
60
70
80
90
100
25 30 35 40 45 20
Angle between extensional phases
1 2
3
4
1
st
2
nd
Fig. 7. Examples of fault interactions in the models. (a) Photograph of part of the top surface of the clay layer of Model 1. Fault scarps dipping toward the top of the page appear
bright; fault scarps dipping toward the bottom of the page appear dark. (b) Line drawing showing a new normal fault terminating against a pre-existing fault. (c) Line drawing
showing a new normal fault cutting and offsetting a pre-existing fault. (d) Graph showing the percentage of second-phase faults that cut and offset pre-existing faults (rather than
terminate against pre-existing faults) for models. Numbers adjacent to the data points indicate the model designation.
deformation (Fig. 4d
2
). Motion on the reactivatied faults is mostly
dip slip during the second phase, with a minor right-lateral strike-
slip component (Table 1). Although corrugations are present on
fault surfaces in all models, the two sets of corrugations associated
with the two phases of extension are most prominent for this
model (Fig. 9). New normal-fault segments constitute only a small
part (12%) of the fault population in Model 4. The strike of these
new normal-fault segments ranges from orthogonal to oblique to
the second-phase extension direction. Although few new normal
faults are present in this model, an almost equal percentage of the
new normal faults cuts across or terminates against the rst-phase
faults (Fig. 7d).
4. Discussion
4.1. Summary of modeling results
The experimental models show that the normal faults that
form during one episode of extension affect the faults that
develop during subsequent episodes of extension. Specically, our
models show that the angle between the rst and second phases
of extension inuences the sense of slip on reactivated rst-phase
faults and the abundance and orientation of second-phase normal
faults. Both fault reactivation and new fault formation occur
during the second phase of extension in our models (Figs. 4, 5).
The reactivated faults have oblique slip, with the strike-slip
component decreasing as the angle between the directions of the
rst- and second-phase extension decreases (Table 1). The
number of new normal-fault segments also decreases as the angle
between the rst- and second-phase extension directions
decreases (Fig. 5, Table 1). In Model 4, with an angle of 22.5
between the extension directions, most second-phase deforma-

tion is accommodated by fault reactivation with little formation of
new normal faults (Figs. 4, 5).
Multiple fault populations are present in all of the models. After
the rst phase of extension, the strike of most fault segments is
approximately orthogonal to the rst-phase extension direction.
Fault segments whose strike is oblique to the rst-phase extension
direction likely reect fault linkage, and constitute only a small part
Fig. 8. (a) Photograph of the top surface of the clay layer of Model 1. (b) Photograph of
the top surface of the clay layer of a different model subjected to the same second-
phase extension as Model 1 but no rst-phase extension. (c) Photograph of the bottom
surface of the clay layer of a different model with the same boundary conditions as
Model 1. Fault scarps dipping toward the top of the page appear bright; fault scarps
dipping toward the bottom of the page appear dark. Fault dip directions appear
reversed in (c) because it shows the bottom surface of the clay layer.
Fig. 9. Oblique-view photograph of fault surface in Model 4 showing two sets of
corrugations. The corrugations parallel the slip direction during each phase of exten-
sion. This fault formed during the rst phase of extension as a normal fault and was
reactivated during the second phase as an oblique-slip fault (with normal and right-
lateral strike-slip components).
(less than 5%) of the fault segment population (Fig. 5). After the
second phase of extension, each model has normal-fault segments
with strikes approximately orthogonal to the rst-phase extension
direction, approximately orthogonal to the second-phase extension
direction, and oblique to both extension directions (Figs. 4, 5). The
latter fault segments, with strikes oblique to both extension
directions, likely reect local perturbations in the stress/strain state
near pre-existing faults (e.g., Homberg et al., 1997, 2004; Katten-
horn et al., 2000). As new normal faults propagate away from pre-
existing faults, their strikes commonly change from orthogonal to
the pre-existing faults to orthogonal to the applied extension
direction (Fig. 6). However, many second-phase normal faults are
conned by closely spaced, pre-existing faults, and never leave the
perturbed stress/strain domain. This compartmentalization also
limits the length of many second-phase normal faults.
4.2. Fault reactivation
The strength of most upper crustal rocks increases with depth,
obeying a MohrCoulomb criterion of failure (e.g., Byerlee, 1978).
According to this criterion,
s C ms
n
(2)
where s and s
n
are, respectively, the shear and normal stresses on
a potential fault surface, C is the cohesive strength, and m is the
coefcient of internal friction. Once a fault surface forms, frictional-
sliding controls failure according to the equation:
s m
s
s
n
(3)
where s and s
n
are the shear and normal stresses on a potential
fault surface and m
s
is the coefcient of sliding friction along the
fault plane. Pre-existing zones of weakness are likely to reactivate if
the ratio of shear stress to normal stress on the surface of the pre-
existing zone of weakness exceeds m
s
.
In all of our models, the orientation of s
1
(the maximum
principal stress) is vertical during both phases of extension. The
orientations of s
2
and s
3
(the intermediate and minimum principal
stresses), however, change between the rst and second phases of
extension. The minimum principal stress, s
3
, parallels the exten-
sion direction at the start of each extensional phase (ignoring any
stress variations associated with the pre-existing faults). Thus, the
rst-phase normal faults, with dips of w60
in the rst-phase s
1
/
s
3
reference frame, have apparent dips between 51
(Model 1) and
58
(Model 4) in the second-phase s

1
/s
3
reference frame (Fig. 10a,
b). Although the exact value for the coefcient of sliding friction
for clay is unknown, we assume that the frictional-sliding enve-
lope parallels the MohrCoulomb failure envelope for intact clay
(with m w0.6 and C w50 Pa) (Fig. 10c). Thus, the loss of cohesion
on the fault surface is the only difference between the intact clay
and the faulted clay after the rst phase of extension. With this
assumption, we expect faults with an apparent dip of about 45
1
2
3
4
51 54 56 58
102 108 112 116
51 54
56
58
45 37.5 30 22.5
100 200 300 400 500 600
100
200
n
New faults during 1
st
phase
Reactivated faults during 2
nd
phase
(number indicates model designation)
c
(Pa)
(Pa)
a
b
Pre-existing fault surface
Cross sections in (a)
= C +
n
=
s n
1
2
3
4
(see b)
* is the angle between and the normal to the fault plane
1
3
51
3
54
1
3
56
1
3
58
1
Model 1 Model 2 Model 3 Model 4
2
Cross section parallel to
3
during 2
nd
phase
Angle between
extension directions
Apparent dip of reactivated
frame during 2
nd
phase
faults in
1
/
3
reference
2 = 102
(faulting)
(frictional sliding)
Fig. 10. (a) Stress regime for reactivated faults at the beginning of the second phase of extension for all models. (b) Block diagram showing orientation of cross sections shown in (a).
(c) Mohr-circle diagram for stress state in the experimental models at the base of the 3.5-cm thick clay layer. Numbers 14 refer to the four different models. Black line shows the
MohrCoulomb failure envelope for wet clay, and the grey line shows the assumed frictional-sliding failure envelope for wet clay. The two failure envelopes differ only by the value
of cohesion.
and higher to be reactivated in our models (Fig. 10c). This agrees
with the observations that all rst-phase faults are reactivated as
oblique-slip faults during the second phase of extension in our
models.
4.3. Comparison to previous modeling studies
Bonini et al. (1997) and Keep and McClay (1997) simulated two
non-coaxial phases of extension using a layer of dry sand overlying
a basal layer of silicone polymer. Their models differ from our
models in terms of the modeling material (dry sand vs. wet clay)
and the width of the basal layer of silicone polymer (the entire base
of the model vs. a narrow zone). Despite these differences, the
modeling results are qualitatively similar. Their models show that
faults that form during the rst phase of extension inuence the
fault patterns that develop during the second phase of extension.
Specically, the reactivation of the rst-phase faults during the
second phase, the development of new second-phase faults, and
the slip and attitude of the second-phase faults depend on the
angle between the rst- and second-phase extension directions.
Reactivation is more likely when the angle is small, whereas new
fault development is more likely when the angle is large. Our
models support these conclusions. In addition, our models show
that the angle between the rst- and second-phase extension
directions controls the relative components of dip slip and strike
slip on the reactivated faults and the number of new faults that
form during the second phase of extension. Furthermore, our
models allow for the observation of small-scale features (such as
the interaction between pre-existing faults and newfaults) because
faults are more numerous, have smaller displacements, and are
more closely spaced in clay models than in sand models (e.g.,
Withjack et al., 2007). Our work, in combination with that of Bonini
et al. (1997) and Keep and McClay (1997), illustrates that a pre-
existing fault fabric substantially inuences the fault patterns that
develop during subsequent episodes of extension.
4.4. Natural examples of multi-phase extension
The fault interactions in our models are similar to those
observed in nature. In the Terra Nova region of the Jeanne dArc rift
basin of offshore Newfoundland, two main fault orientations (NS
and EW) are present (e.g., Enachescu, 1987; Sinclair, 1995a, b)
(Fig. 11a). Movement on N-striking normal faults occurred during
the Late Jurassic through the early Early Cretaceous, whereas
movement on E-striking normal faults occurred during the late
Early Cretaceous (Sinclair, 1995a). Well and 3D seismic-reection
data show that some of the younger, E-striking normal faults cut
and offset the older, N-striking normal faults, whereas others
terminate against the older, N-striking normal faults (McIntyre
et al., 2004) (Fig. 11b, c). The presence of both types of fault inter-
actions in the Jeanne dArc rift basin matches our experimental
observations that both types of fault interactions are likely to occur
with multiple phases of non-coaxial extension. In addition, in many
parts of the Jeanne dArc basin, displacement on the younger,
E-striking normal faults is greatest adjacent to the older, N-striking
normal faults (Fig. 11b). This displacement variation is common in
our models where new faults initiate at pre-existing faults and
propagate outward (Fig. 6). The modeling results suggest that the
5 km
a
b
Pre-existing faults
New faults
Atlantic
Ocean
c
c
b
Newfoundland
N
Fig. 11. (a) Map of the Terra Nova region of the Jeanne dArc rift basin, offshore Newfoundland, showing faults offsetting the B marker (Early Cretaceous). Map modied from
McIntyre et al. (2004). Insert shows location of the Jeanne dArc basin (box) relative to Newfoundland, Canada. (b) Younger, E-striking normal faults terminating against older, N-
striking normal faults. (c) Younger, ESE-striking normal faults cutting and offsetting older, N-striking normal faults. Locations are shown in (a).
younger, E-striking normal faults initiated at and propagated
outward from the older, N-striking normal faults.
Our modeling results also agree with natural observations of
normal-fault reactivation. Studies of abyssal-hill normal faults in
subduction zones (e.g., Masson, 1991; Moritera-Guitie rrez et al.,
2003; Billen et al., 2007) show that the formation of new outer-
slope normal faults (which form to accommodate bending-induced
extension of subducting plates) depends on the orientation of the
pre-existing abyssal-hill normal faults relative to the orientation of
the trench axis. Abyssal-hill normal faults format spreading centers
to accommodate extension of newly forming crust, becoming
permanent features of the oceanic crust (e.g., Rea, 1975; Kriner
et al., 2006; and references therein). Moritera-Guitie rrez et al.
(2003) and Billen et al. (2007) show that, if the angle between the
trench axis and the strike of the pre-existing abyssal-hill faults is
less than 25
(i.e., the angle between the rst- and second-phase

extension directions is less than 25
), then abyssal-hill normal

faults are reactivated and no new normal faults form (Fig. 12b). If
the angle between the trench axes and the strike of the abyssal-hill
faults is greater than 25
, however, new outer-slope normal faults

form that strike perpendicular to the bending-induced extension
direction (i.e., parallel to the trench) (Fig. 12a). Observations from
the Izu-Bonin Trench (Renard et al., 1987), the Middle American
Trench (Masson, 1991), and the Aleutian Trench (Moritera-Gui-
tie rrez et al., 2003) show that the pre-existing abyssal-hill normal
faults may also be reactivated (in addition to new normal fault
formation) if the angle between the trench and the abyssal-hill
faults is greater than 25
. Our modeling results corroborate these

observations: specically, as the angle between the directions of
the two phases of extension decreases from 30
to 22.5
, the strain
accommodation changes from fault reactivation and new normal-
fault formation (Model 3) to mainly fault reactivation (Model 4).
This transition, however, is gradual, occurring between 30
and
22.5
. In addition, our models suggest that some of the reactivated

abyssal-hill normal faults are likely to be oblique-slip faults.
4.5. Predicting strain state from fault orientations
Fault orientations are commonly used to estimate the paleo-
strain states in extensional provinces by assuming that the strike of
the normal faults is perpendicular to the direction of the maximum
horizontal extension (and presumably the direction of s
3
) as pre-
dicted by Anderson (1951). Our modeling results illustrate that, in
areas with multiple extensional episodes, fault orientations are
inuenced by local perturbations in the stress/strain state associ-
ated with pre-existing faults and, thus, are not reliable paleo-strain
indicators (Fig. 13).
In addition, our modeling results suggest that fault interactions
do not necessarily provide information about the order of exten-
sional phases. For example, the second-phase normal faults in our
models either cut the rst-phase normal faults or terminate against
them. Second-phase faults that terminate against rst-phase faults
may be misinterpreted as coeval releasing faults (as dened by
Destro, 1995), or they may appear to be cut by the rst-phase faults
and erroneously considered to be older. Fault orientations, fault
slip, and timing of fault activity (based on the presence or absence
of growth beds) are all necessary to determine the paleo-strain
state. Without this information, the directions and relative timing
of extension determined from fault interactions represent only one
of many possible deformational histories.
5. Conclusions
We have used scaled experimental models to study how a pre-
existing fault fabric affects subsequent deformation during
multiple phases of extension. The models indicate that pre-existing
faults are reactivated for a range of second-phase extension direc-
tions. In addition, pre-existing faults inuence the fault patterns
that form during subsequent extensional episodes by controlling
new fault orientations, lengths, and locations.
t
r
e
n
c
h
b
Pre-existing abyssal-hill faults
Newly formed outer-slope faults
Reactivated abyssal-hill faults
t
r
e
n
c
h
a
Fig. 12. Schematic map patterns showing development of outer-slope normal faults.
(a) New outer-slope normal faults form if the angle between the trench axis and the
strike of pre-existing abyssal-hill normal faults is greater than 25
. (b) Abyssal-hill
normal faults reactivate and no new outer-slope normal faults form if the angle
between the trench axis and the strike of the abyssal-hill normal faults is less than 25
.
Modied from Billen et al. (2007).
2 cm
Inferred extension directions
Actual extension directions
45
75
Fig. 13. Line drawing of the top surface of the clay layer of Model 1 after both phases of
extension (location shown in Fig. 4). Arrows indicate inferred extension directions
determined by assuming that extension is orthogonal to the strike of the normal faults.
The angle between the inferred extension directions is 75
, although the actual angle

between the extension directions is 45
.
The angle between the strike of the pre-existing normal faults
and the second-phase extension direction controls the sense of
slip on the reactivated faults, with the component of dip slip
relative to strike slip decreasing as the angle between the rst-
and second-phase extension directions increases.
New normal faults form during the second phase of extension
in all models. These faults are generally shorter than faults that
formin models with no pre-existing fault fabric. The number of
new fault segments varies between models, with more new
fault segments forming as the angle between the rst- and
second-phase extension directions increases.
The orientations of new normal-fault segments are both
orthogonal and oblique to the second-phase extension direc-
tion. New normal faults with strikes oblique to the second-
phase extension direction likely reect local perturbations in
the stress/strain state near pre-existing faults.
New normal faults commonly initiate at pre-existing normal
faults and propagate outward. Displacement on these new
normal faults is greatest adjacent to the pre-existing faults.
Where new normal faults encounter pre-existing faults, the
newfaults either cut across the pre-existing faults or terminate
against them.
Fault interactions and fault reactivations in the models are
similar to those observed in nature, including the Jeanne dArc
rift basin of offshore Newfoundland and at outer highs near
subduction zones.
Interpretations of extensional histories based solely on fault
orientations and fault interactions represent only one of many
possible deformation histories. Additional information, such as
fault slip and the timing of fault activity (i.e., the presence or
absence of growth beds), is necessary to constrain the paleo-
strain state.
Acknowledgments
We thank our colleagues Michael Durcanin, Iain Sinclair, and
Judith McIntyre for valuable discussions and insights about
modeling and fault interactions and Hemal Vora for his assistance
in the laboratory. We also thank Marco Bonini, Chris Jackson, and
Bruno Vendeville for their detailed and helpful reviews of the
manuscript. We gratefully acknowledge the support of the Struc-
tural Modeling Laboratory at Rutgers University by the National
Science Foundation (EAR-0838462 and EAR-0408878), Husky
Energy Inc., and Petrobras S.A.
References
Anderson, E.M., 1951. The Dynamics of Faulting and Dyke Formation with Appli-
cations to Britain. Oliver & Boyd, Edinburgh.
Badley, M.E., Price, J.D., Dahl, C.R., Agdestein, T., 1988. The structural evolution of the
northern Viking Graben and its bearing upon extensional models of basin
formation. Journal of the Geological Society, London 145, 455472.
Bartholomew, I.D., Peters, J.M., Powell, C.M., 1993. Regional structural evolution of
the North Sea: oblique slip and the reactivation of basement lineaments. In:
Parker, J.R. (Ed.), Petroleum Geology of Northwest Europe: Proceedings of the
Fourth Conference. The Geological Society, London, pp. 11091122.
Bellahsen, N., Daniel, J.M., Bollinger, L., Burov, E., 2003. Inuence of viscous layers on
the growth of normal faults: insights from experimental and numerical models.
Billen, M., Cowgill, E., Buer, E., 2007. Determination of fault friction from reac-
tivation of abyssal-hill faults in subduction zones. Geology 35, 819822.
Bonini, M., Souriot, T., Boccaletti, M., Brun, J.P., 1997. Successive orthogonal and
oblique extension episodes in a rift zone: laboratory experiments with appli-
cation to the Ethiopian Rift. Tectonics 16, 347362.
Brace, W.F., Kohlstedt, D.L., 1980. Limits on lithospheric stress imposed by labora-
tory experiments. Journal of Geophysical Research 85, 62486252.
Byerlee, J., 1978. Friction of rocks. Pure and Applied Geophysics 116, 615626.
Clifton, A.E., Schlische, R.W., Withjack, M.O., Ackermann, R.V., 2000. Inuence of rift
obliquity on fault-population systematics: results of clay modeling experi-
ments. Journal of Structural Geology 22, 14911509.
Destro, N., 1995. Release fault: a variety of cross fault in linked extensional fault
systems, in the Sergipe-Alagoas Basin, NE Brazil. Journal of Structural Geology
17, 615629.
Eisenstadt, G., Sims, D., 2005. Evaluating sand and clay models: do rheological
differences matter? Journal of Structural Geology 27, 13991412.
Enachescu, M.E., 1987. Tectonic and structural framework of the northeast
Newfoundland continental margin. In: Beaumont, C., Tankard, A.J. (Eds.),
Sedimentary Basins and Basin-Forming Mechanisms. Canadian Society of
Petroleum Geologists Memoir 12, 117146.
Frseth, R.B., 1996. Interaction of Permo-Triassic and Jurassic extensional fault-
blocks during the development of the northern North Sea. Journal of the
Geological Society, London 153, 931944.
Granger, A.B., 2006. Inuence of basal boundary conditions on normal-fault systems
in scaled physical models. Masters thesis, Rutgers University.
Granger, A.B., Withjack, M.O., Schlische, R.W., 2006. Undulations on normal-fault
surfaces: insights into fault growth using scaled physical models of extension.
Geological Society of America Abstracts with Program 38, p. 480.
Granger, A.B., Withjack, M.O., Schlische, R.W., September 2008. Fault surface
corrugations: insights from scaled experimental models of extension. In: Fault
Zones: Structure, Geomechanics, and Fluid Flow Conference, Abstracts
Volume. Geological Society of London. 38.
Handin, 1966. Strength and ductility. In: Clark, S.P. (Ed.), Handbook of Physical
Constants, 97. Geological Society of America Memoir, pp. 233289.
Homberg, C., Hu, J.C., Angelier, J., Bergerat, F., Lacombe, O., 1997. Characterization of
stress perturbations near major fault zones: insights from 2D distinct-element
numerical modelling and eld studies (Jura mountains). Journal of Structural
Geology 19, 703718.
Homberg, C., Angelier, J., Bergerat, F., Lacombe, O., 2004. Using stress deections to
identify slip events in fault systems. Earth and Planetary Science Letters 217,
409424.
Hubbert, M.K., 1937. Theory of scale models as applied to the study of geological
structures. Geological Society of America Bulletin 48, 14591519.
Kattenhorn, S.A., Aydin, A., Pollard, D.D., 2000. Joints at high angles to normal fault
strike: an explanation using 3-D numerical models of fault-perturbed stress
elds. Journal of Structural Geology 22, 123.
Keep, M., McClay, K.R., 1997. Analogue modeling of multiphase rift systems. Tec-
tonophysics 273, 239270.
Kriner, K.K., Pockalny, R.A., Larson, R.L., 2006. Bathymetric gradients of lineated
abyssal hills: inferring seaoor spreading vectors and a new model for hills
formed at ultra-fast rates. Earth and Planetary Science Letters 242, 98110.
Masson, D.G., 1991. Fault patterns at outer trench walls. Marine Geophysical
Researches 13, 209225.
McIntyre, J., DeSilva, N., Thompson, T., 2004. Mapping of key geological markers in
the Jeanne dArc basin based on 3-D seismic. Canadian Society of Petroleum
Geologists Annual Meeting.
McClay, K.R., White, M.J., 1995. Analogue modeling of orthogonal and oblique rift-
ing. Marine and Petroleum Geology 12, 137151.
Moritera-Guitie rrez, C.A., Scholl, D.W., Carlso, R.L., 2003. Fault trends on the
seaward slope of the Aleutian Trench: implications for a laterally changing
stress eld tied to a westward increase in oblique convergence. Journal of
Geophysical Research 108. doi:10.1029/2001JB001433.
Morley, C.K., Gabdi, S., Seusutthiya, K., 2007. Fault superimposition and linkage
resulting from stress changes during rifting: examples from 3D seismic data,
Phitsanulok Basin, Thailand. Journal of Structural Geology 29, 646663.
Morley, C.K., Harayana, C., Phoosongsee, W., Pongwapee, S., Kornsawan, A.,
Wonganan, N., 2004. Activation of rift oblique and rift parallel pre-existing
fabrics during extension and their effect on deformation style: examples from
the rifts of Thailand. Journal of Structural Geology 26, 18031829.
Nalpas, T., Brun, J.P., 1993. Salt ow and diapirism related to extension at crustal
scale. Tectonophysics 228, 349362.
Rea, D.K., 1975. Model for the formation of topographic features of the East Pacic
Rise Crest. Geology 3, 7780.
Renard, V., Nakamure, K., Angelier, J., Azema, J., Bourgois, J., Deplus, C., Fujioka, K.,
Hamano, Y., Huchon, P., Kinoshita, H., Labaume, P., Ogawa, Y., Seno, T.,
Takeuchi, A., Tanahashi, M., Uchiyama, A., Vigneresse, J.L., 1987. Trench triple
junction off Central Japan preliminary results of the FrenchJapanese 1984
Kaiko cruise, Leg 2. Earth and Planetary Science Letters 83, 243256.
Schellart, W.P., 2000. Shear test results for cohesion and friction coefcients for
different granular materials: scaling implications for their usage in analogue
modeling. Tectonophysics 324, 116.
Sinclair, I.K., 1995a. Transpressional inversion due to episodic rotation of exten-
sional stresses in Jeanne dArc Basin, offshore Newfoundland. In: Buchanan, J.G.,
Buchanan, P.G. (Eds.), Basin Inversion. Geological Society Special Publication 88,
pp. 249271.
Sinclair, I.K., 1995b. Sequence stratigraphic response to AptianAlbian rifting in
conjugate margin basins: a comparison of the Jeanne dArc Basin, offshore
Newfoundland, and the Porcupine Basin, offshore Ireland. In: Scrutton, R.A.,
Stoker, M.S., Shimmield, G.B., Tudhope, A.W. (Eds.), The Tectonics, Sedimenta-
tion, and Palaeoceanography of the North Atlantic Region. Geological Society
Special Publication 90, pp. 2949.
Sinclair, I.K., Withjack, M.O., 2008. Mid to Late Cretaceous structural and sedi-
mentary architecture at the Terra Nova oileld, offshore Newfoundland
implications for tectonic history of the North Atlantic. In: Brown, D.E. (Ed.),
Central Atlantic Conjugate Margins. Dalhousie University, Halifax, Nova Scotia,
pp. 125141.
Strecker, M.R., Blisniuk, P.M., Eisbacher, G.H., 1990. Rotation of extension direction
in the central Kenyan Rift. Geology 18, 299302.
Tankard, A.J., Welsink, H.J., 1987. Extensional tectonics and stratigraphy of Hibernia
oil eld, Grand Banks, Newfoundland. AAPG Bulletin 71, 12101232.
Tron, V., Brun, J.P., 1991. Experiments on oblique rifting in brittle-ductile systems.
Tectonophysics 188, 7184.
Weijermars, R., 1986. Flow behaviour and physical chemistry of bouncing putties
and related polymers in view of tectonic laboratory applications. Tectonophy-
sics 124, 325358.
Weijermars, R., Jackson, M.P.A., Vendeville, B.C., 1993. Rheological and tectonic
modeling of salt provinces. Tectonophysics 217, 143174.
Withjack, M.O., Callaway, J.S., 2000. Active normal faulting beneath a salt layer: An
experimental study of deformation in the cover sequence. AAPG Bulletin 84,
627652.
Withjack, M.O., Jamison, W.R., 1986. Deformation produced by oblique rifting.
Withjack, M.O., Schlische, R.W., 2006. Geometric and experimental models of
extensional fault-bend folds. In: Buiter, S.J.H., Schreurs, G. (Eds.), Analogue and
Numerical Modelling of Crustal-Scale Processes. Geological Society (London)
Special Publication 253, pp. 285305.
Withjack, M.O., Schlische, R.W., Henza, A.A., 2007. Scaled experimental models of
extension: dry sand vs. wet clay. Houston Geological Survey Bulletin 49 (8), 3149.
Using empirical geological rules to reduce structural uncertainty in seismic
interpretation of faults
Brett Freeman
a,
*
, Peter J. Boult
b
, Graham Yielding
a
, Sandy Menpes
c
a
Badley Geoscience Ltd., North Beck House, North Beck Lane, Hundleby, Lincolnshire PE23 5NB, UK
b
GINKGO ENPGNG, 57, Seventh Avenue, St. Morris SA 5068, Australia
c
Palaeosearch, 41 Walker Avenue, Heatheld SA 5153, Australia
Article history:
Received 3 February 2009
21 October 2009
Accepted 2 November 2009
Keywords:
Normal fault
Displacement
Displacement gradient
Wall-rock strain
Seismic interpretation
a b s t r a c t
Good seismic interpretation of faults should include a workow that checks the interpretation against
known structural properties of fault systems. Estimates of wall-rock strains provide one objective means
for discriminating between correct and incorrect structural interpretations of 2D and 3D seismic data
implied wall-rock strain should be below a geologically plausible maximum. We call this the strain
minimisation approach. Drawing on the large body of published data for strike dimension and maximum
displacement for faults we suggest a realistic upper limit of wall-rock shear strain of 0.05, and 0.1 for
maximum longitudinal strain when measured in the displacement direction. Small-scale variation of
fault wall-rock strain also adheres to this rule, except in specic areas of strain localisation such as relay
zones. As a case study we review an existing structural interpretation of 2D seismic surveys. Mapping of
shear and longitudinal strain on the fault planes show values commonly greater than 0.05 and 0.1
respectively. Thus the model is deemed inadmissible. We then reinterpreted the area in an iterative
manner using the strain minimisation approach. By using regions of implied high wall-rock strain as an
indicator of high uncertainty in the interpretation, we were able to break out two self-consistent fault
sets, each of which had geologically plausible wall-rock strains, where previously there had only been
one fault set with highly implausible wall-rock strains.
1. Introduction
It has been established for more than twenty years that the
displacement on geological fault surfaces varies in a smooth,
continuous and consistent manner. Rippon (1985) and Barnett et al.
(1987) rst demonstrated this for isolated normal faults from the
English coalelds. They used precise survey data from coal mine
plans to measure the throw (vertical component of dip separation)
for a number of coal seams at varying spatial locations. When the
throwvalues were plotted on a strike projection of the fault surface,
the contours of throw were concentric and sub-parallel, with
a maximumthrowclose to the centre of the fault surface. Moreover
the boundary, or tip, of the fault surface was approximately ellip-
tical. These important observations have stimulated an enormous
amount of research into the form and scaling relationships of
displacement distributions, the quantitative systematics of fault
geometry and speculation on fault growth mechanisms. The
simplicity of the observations implies both a consistency and a limit
to the strain in the wall rocks. Barnett et al. (1987), Bouvier et al.
(1989) (normal faults) and then Chapman and Meneilly (1991)
(reactivated normal fault with net reverse displacement) demon-
strated similar patterns from seismic interpretation. Although the
precision of the structural information from seismic data is
considerably poorer than the surveyed data of Rippon (1985), these
early examples of displacement distributions are also characteris-
tically continuous and smooth.
A priori knowledge of the shape/form of the displacement
distribution and its gradients can be useful as an aid to seismic
interpretation. Barnett et al. (1987) suggested that it could serve
both as a quality control metric and a means to predict quantita-
tively the location of lithological layers (horizons) and faults where
data is limited. In other words it can be used to manage interpre-
tation uncertainty. In faulted reservoirs, structural uncertainty
arises from two major sources of error: the systematic error of the
seismic method, and the human error of the interpreter. For good
quality 3D seismic data the order of error in lateral positioning of
structures is approximately the same as the error in the vertical
dimension and both are dominantly systematic. However, when
the spacing between fault traces from line samples (e.g. seismic
lines) is closer than the spacing between the line samples
* Corresponding author. Tel.: 44 (0) 1754 890390; fax: 44 (0) 1790 753527.
E-mail address: brett@badleys.co.uk (B. Freeman).
doi:10.1016/j.jsg.2009.11.001
themselves, the lateral correlation of faults is equivocal (e.g.
Freeman et al., 1990). So for 2D seismic data, reconnaissance
mapping, poor quality 3D seismic data and for small faults in 3D
seismic data, the pattern of faulting becomes a serious interpretive
issue. The balance of the error, or uncertainty, is then strongly one-
sided and the effects of systematic errors become secondary to
those inherent in the interpreters model. Freeman et al. (1990)
introduced a methodology that used displacement patterns to
distinguish likely fault plane correlations from possible and
impossible correlations. In a similar vein, Needham et al. (1996)
showed howthis kind of analysis was valuable for validating three-
dimensional structural models. Traditionally the analytical part of
the process has taken the form of visual inspection of the throw
contours. If the resulting pattern is smooth and continuous, the
fault may be judged as a valid interpretation, otherwise the
correlation is deemed to be suspect. Although ostensibly objective,
the effectiveness of the above approach is dependent on the skill or
experience of the interpreter in being able to identify bad contour
patterns. We know of no published work that actually quanties
what constitutes either a good or a bad contour pattern. In this
paper we suggest that the above basic validation procedure can be
improved by (1) quantifying the strains that are implied by the
contour patterns and (2) setting out reasonable limits for the
magnitudes of these strains.
We show that there is a simple relationship between strain and
the displacement gradient. Drawing from a large database of pub-
lished information on the shapes of displacement proles and the
scaling relationship between displacement and fault dimension, we
can suggest reasonable limits on the amount of strain that is
admissible in the walls of a fault. Implied strains that exceed these
empirical limits indicate aws in the structural model. The result-
ing strategy for interpretation leads to a structural model that
minimizes the strain attributable to faulting.
As an example we show how an analysis of a 2D seismic inter-
pretation from South Australia consistently implies erratic and
unrealistically large strains. An iterative structural reinterpretation
using our minimum strain approach provides a solution that is
geologically more feasible.
2. Displacement and wall-rock strain
There is a simple relationship between the displacement gradient
and the strain of the wall rocks in the plane parallel to the fault.
Fig. 1a shows the deformation associated with the faulting of a pre-
existing uniform horizontal layer (i.e. the fault is not a growth fault).
The element, E (Fig. 1a and b), is dened by the position of the layer
in the undeformed state with the top of the layer at p and the base of
the layer at q (Fig. 1b). For the sake of argument we assume that
displacement is in the dip direction of the fault, and that strain is
partitioned equally in the two walls of the fault. The layer is then
faulted such that, inthe dip direction (parallel to y), p moves to p
0
and
q moves to q
0
. The stretch in the dip direction is then
1 e
u
q
q
u
p
p
q p
(1)
where e is the unit extension, u
p
and u
q
are the absolute displace-
ments for one side of the fault (half the total, relative displace-
ment). This can be re-written as
1 e
1
1
2
Du
Dy
(2)
where the factor of 1/2 means that u refers to the total relative
displacement across the fault. At the limit as Dy approaches zero
1 e
1
1
2
vu
vy
(3)
In other words the unit extension is equal to half the displacement
gradient. Using an alternative formulation it is easy to show
that the stretch in the upthrown layer is the reciprocal of the stretch
in the downthrown layer and that the undeformed layer thickness
is the average of the upthrown and downthrown thicknesses
(cf. gure 1 from Barnett et al., 1987).
Referring back to Fig. 1b we can also see that, for each wall of the
fault, the strain g for shear in the dip direction is given by
g
u
s
u
p
=s p
1
2
Du
Dx
(4)
then as Dx approaches zero
g
1
2
vu
vx
(5)
Eqs. (3) and (5) are useful results because (1) they are independent
of the form of the displacement distribution, and (2) they give us
a direct way to measure and represent strain from information that
is almost universally available from seismic interpretations.
If we can place realistic limits on the strain values, we then have
a method for distinguishing between good and bad fault interpre-
tation that is entirely quantitative and objective.
Fig. 1. (a) Schematic of an idealized fault plane (strike dimension L, dip dimension L/2)
showing the absolute displacements from a horizontal, unfaulted layer, to the faulted,
upthrown and downthrown positions. The element E in the unfaulted state is trans-
lated and strained to E
0
. (b) Analysis of the change in shape of the rectangular element
E. x is the strike direction of the fault and y is the dip direction, u
p
, u
q
and u
s
are
absolute, dipslip displacements.
B. Freeman et al. / Journal of Structural Geology 32 (2010) 16681676 1669
2.1. Ideal displacement patterns for unrestricted faults
Although direct measurements of wall-rock strains is probably
beyond normal eld techniques, the simple fact that unrestricted
faults have tip lines means that the wall rocks must be differentially
strained. Eshelby (1957) and Pollard and Segall (1987) suggest the
slip on a dislocation in a linear elastic solid is characterised by
a semi-elliptical slip prole. In other words, a straight marker line
in a wall of a fault and initially perpendicular to the slip direction
will have a deformed shape of a semi-ellipse and reects directly,
the differential wall-rock strains. This type of slip prole equates to
a single earthquake event. However, Nicol et al. (1996) and many
others show that natural examples of unrestricted faults have
approximately linear normalized proles i.e. triangular. Further-
more Manzocchi et al. (2006) argue that this feature of geological
faults seems to hold irrespective of the growth mechanism or the
form of the slip prole for an individual event. This is a convenient
conclusion because it means the gradient of displacement on an
unrestricted fault surface is approximately constant.
2.2. Limits on displacement
Various compilations of data for D
max
/L (D
max
is the maximum
displacement, L is the strike dimension of a fault) have been pub-
lished (e.g. Bailey et al., 2005; Kim and Sanderson, 2005; Schultz
et al., 2008). Although there remains debate about the exact nature
of the power-law distribution of D
max
vs L, it seems that 0.1 repre-
sents the naturally observed upper bound for all types of faults over
all measured scales (Fig. 2). However, if we focus our attention on
the scale range imaged on seismic data, we can rene the limit to
0.05 (Fig. 2). Then if we assume for an unrestricted fault that the
displacement prole is triangular, (1/2) (D
max
/2)/(L/2) 0.05 places
a natural limit of 0.05 on the shear strain in each wall.
Unfortunately there is no similar database for the relationship
between displacement and the dip dimension of a fault. In this
respect we make a further assumption that the aspect ratio of our
unrestricted fault is 2 (e.g. Nicol et al., 1996), then (1/2) (D
max
/2)/
(L/4) 0.1 represents the limit of the longitudinal strain in each
wall. These suggested limits for shear and longitudinal strain are
consistent with detailed measurements of coaleld fault displace-
ment gradients by Walsh and Watterson (1989). For tip restricted or
half restricted faults the displacement prole is steepened towards
the tips (Nicol et al., 1996). Potentially this could increase the shear
strain by a factor of two or more.
3. Method
All the analysis has been performed using the TrapTester
software (www.badleys.co.uk/products/traptester.htm). Fault
planes have been generated as triangulated meshes from the
vertices of fault sticks picked on seismic sections. The horizon
cutoffs at the faults are calculated from seismic interpretation of
the horizons. It is rare for seismic horizon picks to tie exactly
with the fault picks and therefore some extrapolation is required
to make the cutoff lines. In fact we make three-dimensional
surface models of both the upthrown and downthrown sides of
the fault. Each of these models is extrapolated so that it extends
beyond the fault in both directions and the cutoff is calculated as
the intersection between the fault surface and the horizon
surface model. If faults are joined at branch lines, the horizon
surface model is based on the structurally coherent horizon data
that is conned to the appropriate parts of the interpretation as
bounded by all relevant fault planes. The quality of t of the
resulting polygons has been assessed visually and is, in all cases,
a fair representation of where an expert geoscientist might draw
the cutoff by hand.
Immediately prior to analysis all information for a fault plane
is referred to a coordinate frame that is specic to each particular
fault. We call this a natural coordinate system (NCS). The xy
plane (x along strike, y down-dip) of the NCS represents a best t
plane through all the data points that dene the fault plane
topography. The upthrown and downthrown cutoffs form
a polygon and this is the basis for the structural measurements.
We measure the dip separation on a set of sweep lines (constant x)
at an interval of 50 m in the x direction of the NCS. The
raw measurements are interpolated on to a 50 m50 m grid
using the multi-level B-Spline method of Lee et al. (1997). This
interpolation scheme honours the raw data and produces
a smoothly interpolated surface everywhere else. We calculate the
strike and dip gradients of dip separation using a central difference
formula.
It should be noted that the displacement gradient is measured
and recorded midway between the corresponding upthrown and
downthrown layers i.e. at the location in the undeformed state.
However, the implied strains refer to the layers themselves i.e. the
deformed state.
4. Interpretation example from South Australia
4.1. Gambier Embayment, Otway basin
The Otway Basin is a passive margin forming a large part of the
Eastern Australian Rift System that resulted from the separation of
Australia fromAntarctica during the Late Jurassic to Late Cretaceous
(Lovibond et al., 1995). Along with the rest of the Australian margin,
the Otway Basin has a complex Mesozoic to present day structural
history, with multiple rifting events causing principal stress/strain
directions and magnitudes to change repeatedly during its devel-
opment. The Gambier Embayment is a Tertiary sub-basin of the
Otway Basin. It is bounded to the north and east by the Tartwaup
Hinge Zone, to the west by the continental shelf and to the east it
Fig. 2. Compilation of maximum displacement (D
max
) and maximum fault length (L)
adapted from Schultz et al. (2008). The data are contoured in lines of constant D
max
/L.
The 0.05 contour is our putative upper bound for D
max
/L ratio at the range of scales of
normal faults imaged on seismic data.
merges (in Victoria) into the Tertiary NWSE trending Portland
Trough.
Within the Gambier Embayment the Late Cretaceous section
comprises deep- to marginal-marine and deltaic sediments up to
2 km thick onshore and over 3 km offshore. The Tertiary section is
up to 1000 mthick in the Gambier Embayment and is up to 2700 m
in the Portland Trough. It comprises uvial to deltaic sediments
overlain by marl and limestone up to 400 m thick in the Gambier
Embayment and 900 m thick in the Portland Trough (Boult, 1999;
Boult and Hibburt, 2002).
4.2. Seismic data
The raw data for this study is based entirely on 2D time-
migrated seismic data with approximate line spacings of 1.5 km for
dip lines and 5 km for strike lines (see Fig. 3a). For the purpose of
piecewise depth conversion we have used an average velocity of
3000 ms
1
over the entire two way time (TWT) interval. There are
a number of different vintages of seismic and the quality of the
reection data is variable. For the most part fault offsets are clearly
imaged in the upper part of the sections but they become less easy
to interpret as the data reaches about 2000 ms. Similarly, seismic
reectors are relatively well imaged in the upper sections becoming
less well imaged at depth. An example of the variability of quality is
illustrated in Fig. 4.
4.3. Initial structural model
Interpretation (Essential Petroleum Resources Ltd., 2006) had
been conducted using a typical industry strategy on 2D seismic
panels and horizon maps and as far as we are aware, made no
deliberate attempt to adhere to the basic rules of displacement
continuity as outlined in the introduction to this paper. The
resulting structural model comprises a set of fault planes and ve
horizons over a TWT range of about 2500 ms. A summary of the
initial model at top reservoir level (Fig. 3b) shows an eastward
shallowing structure dissected by a set of large NWSE trending
normal faults (Essential Petroleum Resources Ltd., 2006). These
faults are spaced at about 2 km, they have maximum throws of
about 200 ms (approximately 300 m) and maximum dip separa-
tions of the order 400 m. Throughout the discussion of these data
we use dip separation as the measure of displacement in the fault
plane. The top reservoir has a TWT range of the order 300 ms about
an absolute TWT of 1900 ms. At this level and deeper, it becomes
more challenging to tie horizons from line to line and across faults.
It is therefore very important to make efcient use of the more
Fig. 3. (a) Base map of the 2D seismic shot lines from the Gambier Embayment, Otway Basin, South Australia. The line with the asterisks is shown in Fig. 4. (b) Prospect map at the
level of top reservoir. Colour code dark to light indicates shallow to deep.
reliable structural information above the top reservoir in order to
constrain the fault model at depth. Fig. 5be illustrates that the
fault planes themselves are picked over the full depth range.
However, displacement information (from the horizon cutoffs) is
limited to a relatively small area of the fault planes (approximately
50%) since not all of the ve horizons are picked persistently on all
lines and there were no picks at the top of the fault plane. Again,
Fig. 5be shows the regions where the displacement information is
reliable i.e. in the close vicinity of horizon picks.
Bulk measurements for the faults fall in a geologically sensible
cluster with 0.1 <D
max
/L< 0.001. Because the maximum displace-
ments are small relative to the interpreted fault spacing we might
expect there to be minimal interference between the faults hence
we might also expect simple displacement contour patterns.
Notwithstanding these two observations, displacement mapping
clearly shows that all of the fault planes have erratic contour
patterns (Fig. 5a). They show multiple bulls-eyes (highs and lows)
and exhibit both sub-vertical and sub-horizontal valleys in the
displacement magnitudes. In terms of displacement gradients,
hence strain, all the faults exhibit multiple lateral swings in the sign
of shear strain and longitudinal strain (Fig. 5be). We expect shear
strain to be highest at the tips, positive at the left of the strike
projection and negative at the right. In fact we see the polarities
inverted, locally and globally, and the high shear strains concen-
trated towards the centre of the faults. Equally importantly, nearly
all the faults have areas where the magnitudes of the strains lie
outside our bounding, acceptable threshold values (red and
magenta in Fig. 5be). In general we take the (implied) high shear
strain anomalies to indicate a locationwhere the current fault plane
should either be split in to two separate faults or where two faults
join at a branch line. The anomalous (implied) longitudinal strains
and associated ips in polarity usually indicate that a horizon has
been picked persistently in the wrong part of the waveform on one
or other side of the fault.
4.4. Revised structural model
One of us (PJB) undertook a reinterpretation of the seismic
data with a view to producing (1) a model that was geologically
more acceptable and (2) a model that minimized the implied
strains. We picked the same ve horizons with the exception of
the top reservoir where, instead, we picked a reection 200 ms
above the original. (With sufcient picks the shape of the
displacement distribution should be independent of the actual
horizons that are chosen.) The majority of faults intersect at least
three of the ve horizons. In most cases the upper ends of the
traces (sticks) are observed to be at zero displacement so, in
addition to horizon-based displacement information, the upper
tips have been explicitly assigned displacements of zero.
Although all the original interpreted fault traces were reused
(they received minor lateral shifts on some of the lines) our
reinterpretation identied many more fault traces on each of the
sections. In all we more than doubled the amount of fault
information. A possible reason for the inadequacy of the original
model is that the interpretation strategy was driven by picking
the faults with the largest offsets and then forcing them into
common-trend, lateral, correlations. In fact it is equally impor-
tant to identify the fault traces with small offsets, since a fault
plane can be interpreted as nearing its tip, only if it is correlated
to traces with minimal displacement. In this work we have used
an interpretation strategy that is both iterative and incremental
inasmuch as the structure is validated as it evolves. The proce-
dure is summarized in Boult et al. (2008).
Fig. 6 summarizes the results of the reinterpretation and we
outline the major contrasts below:
(1) The new interpretation has a denser fault pattern. Unlike the
original, we identify two major fault sets: one set in a NWSE
orientation (c.f. Fig. 5a), the other trending, broadly, WNW
ESE.
(2) There are no faults that transect the entire survey area. In
general the new faults have smaller map dimensions than the
originals. We identify isolated, en echelon and linked structures.
(3) The displacement contours are, largely, smooth and continuous
and they are void of bulls-eyes.
(4) In general, the displacement gradients and strain patterns are
also smooth and lack the erratic polarity ips that we see in
Fig. 5. In particular, the implied strains are relatively low. The
majority of surface area on each fault has strain values well
below our upper bounds.
Fig. 4. Example of the 2D seismic data from line bu85-38_r9 (see Fig. 3 for location). The sub-vertical lines are fault segments, the sub-horizontal lines are horizon picks. All
segments and picks are from the revised interpretation. Dashed vertical lines mark the intersection with other seismic lines.
(5) Values of longitudinal strains mostly lie in the range
0.1 <e <0.1 and they are tightly clustered around smaller
values. The signs are uniformly positive in the upper parts of
the fault and negative at depth. However, on several of the
faults we see persistent bands of high implied strain that lie
outside our upper limit (red colours in Fig. 6b and c). Initially
this appears to put the interpretation in question but in fact,
the top horizon is picked above an unconformity. This gives the
effect of anomalously high displacement gradients and we take
these high values to indicate that the faults were at least
partially active prior to the unconformity.
(6) The two fault trends (above) can also be distinguished on the
basis of the displacement and longitudinal strain patterns. The
NWSE set displays a symmetry that we associate with
displacement dying out towards both an upper and lower tip
while the WNWENE set has displacement increasing
downwards.
(7) The shear strain patterns show an obvious symmetry with low
values in the central part of the fault, increasing to maximum
values at each lateral tip.
(8) To a large degree the shear strain maps conform to our notion
of maximum strain. The polarities are mostly consistent and
the extents of regions where the strains are beyond our limits
of g 0.05 are conned to the fault tip regions. This increase in
gradient and strain towards the tips is likely to be a conse-
quence of interference between two or more faults.
5. Discussion and conclusions
We have shown that measurement of displacement and
mapping of displacement gradients leads to the notion of limits
to the wall-rock strain either side of a fault. Applying these limits
allows us to make objective judgements about the validity of
fault interpretation from seismic data and thus reduces the
uncertainty inherent in the interpretation process. In practice it is
the bounding values that are important and it is of little conse-
quence whether the actual metric that is used is the displace-
ment gradient or the strain. The displacement gradient is, of
course, the measured quantity but its meaning is slightly less
tangible than strain. Strain is a quantity that is more commonly
used in the literature and more likely to be linked, at least
intuitively, to other phenomenon. For example, one might predict
there to be a correlation between wall-rock strain and the degree
of fracture damage in the close vicinity of a fault.
The brief and simple analysis (Eqs. (3) and (5)) is based on dip
slip relative motion. Therefore its application is most suitable for
unrestricted, normal or reverse faults. However, dip separation is
always a minimumestimate of the true slip magnitude; for dipslip
faults it is exact but for oblique slip faults it is less than the true slip.
Similarly our estimate of both the longitudinal strain and the shear
strain will underestimate the values in the true slip direction. We
would suggest that for minor obliquity the dip slip thresholds may
still be usefully applied but it would be more accurate if either (1)
the dip slip values were to be corrected for the rake of the slip
vector or (2) the offsets were to be measured along the slip direc-
tion. But precise slip direction is almost always impossible to
determine from seismic reection data.
Given the generality of approximately triangular displacement
proles, we believe the upper limit we place on shear strain of
g 0.05 for isolated faults at the seismic scale is reasonably robust
since it is based on a large collection of D
max
/L data. As we have
already discussed, the limit of D
max
/L 0.05 seems to hold over the
scale range observed on seismic reection data. However, we
should note that other workers show that small faults (up to
L 5 m) in incompletely lithied sand have D
max
/L bounded at 0.1.
The contrast in mechanical properties with completely lithied
sandstone in elasto-plastic models of fault propagation can explain
such increases (Wibberley et al., 1999). Similarly, at the upper end
of the scale thrust faults also seem to be bounded by D
max
/L 0.1
(see Fig. 2), but this may be due to other processes such as tip
propagation blocking by basalt layers (Puentes Hills), or ductile
deformation mechanisms accommodating along-strike decolle-
ment strains (Rocky Mountains thrusts). Although there are issues
involving sampling of the principal axis of a fault, the strike
dimension and maximum offset are relatively straightforward to
measure. The main source of uncertainty is whether or not the
faults in the compilation (Fig. 2) are truly isolated. If the highest
D
max
/L values are due to faults that are horizontally restricted, then
our estimate of maximum shear strain will be too high. We should
also note that in all our natural examples described here the shear
strain gradient is lower at the centre of the faults than at the tips
and that the gradient is highest between the fault centre and the
tips i.e. the pattern is actually more of a bell shape than linear.
Moreover, in the vicinity of overlapping tips, we record higher
shear strains than the upper limit we expect for isolated faults. This
is consistent with other natural examples (Nicol et al., 1996) and
with the effect of mechanical interaction between two faults.
Relative to an unrestricted fault, the shear strain, where faults
overstep, increases with size of overlap, dip dimension and
decreasing distance between the faults (Willemse, 1997). Again,
these high strain zones may give an indication of sub-seismic scale
fracturing and thus have some impact on the local uid ow
behaviour.
Maximum allowable longitudinal strain in the dip slip direc-
tion is less robust. Schultz and Fossen (2002) argue that the
displacement scales not only with length but with aspect ratio
and that the highest D
max
/L corresponds to the smallest aspect
ratios. Soliva et al. (2005) report a similar phenomenon that for
a given constant fault height, D
max
/L decreases with length. Both
of these studies refer to outcrop scale observations where the
structures are conned to single layers. Beyond outcrop scale the
dip dimension of faults is, by comparison with the strike
dimension, more difcult to constrain. For example on seismic
reection data it is common for the upper tips to be truncated at
unconformities and/or the lower tips to be not imaged clearly.
Consequently there is little in the way of published data that
incorporates D
max
, L and aspect ratio at the scale of seismically
imaged faults. Nicol et al. (1996) show that the aspect ratios of
unrestricted faults, over the scale range of 10s of metres to 10s of
kilometers, lie between 1 and 3 but unfortunately they provide
no information on the maximum displacement for the same
faults. In setting our upper limit we have tried to embrace both
of these sets of observations, so a maximum longitudinal strain
of e 0.1 is based on D
max
/L 0.05 and the ellipticity of a fault
being a maximum of 2:1.
Our upper limit of dip-direction longitudinal strain leads to
a maximum layer thickness ratio of approximately 1.2 for corre-
sponding layers either side of the fault. Note that we are describing
fault-related strain of pre-faulting layers, not thickness changes in
syn-faulting layers (i.e. growth sequences). Increasing the aspect
ratio of the ellipse and maintaining the same displacement and
strike dimension (in contrast to Soliva et al. (2005), above)
dramatically increases the implied longitudinal strain. For example
an ellipticity of 3:1 implies a maximum longitudinal strain of
e 0.15 and differential thicknesses of 1.35. In our experience
apparent strains of this magnitude are always associated with
sedimentary growth or they are found near the tops of faults that
have been truncated by unconformities. In either case they are not
real strains. On the other hand choosing an aspect ratio of unity
reduces the upper limit by a factor of two to e 0.05. We believe
that the routine examination of longitudinal strain should provide
a valuable quality control metric but we also suggest that the
published database for D
max
, L and aspect ratio needs to be
enhanced.
All our analysis and discussion is based on upper limits to strain.
It is also possible, from the point of view of 2D seismic interpre-
tation, for faults to be over-correlated laterally with large fault
strike lengths but only very small displacements. It seems that this
type of error is more difcult to quantify in terms of strain limits
since the range of known D
max
/L implies that strains could easily be
at least 100 times less than our upper bounds.
There remains the problem of anomalous implied strains in the
walls of faults which, in all other respects, have been interpreted
using this minimum strain approach. Depending on scale and
quality of the seismic data it is always possible that relatively small
structures are missed from the interpretation. In which case, these
anomalies may be the best indication of additional, real, structure.
Although we strongly caution against the invention of structure it
may be that such indirectly observed features could be incorpo-
rated into end member structural models. These would have
particular signicance if the three-dimensional structural model is
to be used for uid ow simulation.
Fig. 5. (a) Perspective viewof the faults from the original interpretation, colour coded and contoured in dip separation. The fault surfaces are displayed only where the displacement
information is present. Details of faults I, II, III, and IV are shown in strike projection in b through e. In each panel (be) the top image shows the entire fault surface (grey) and the
region where displacement information is present is colour coded in dip separation. Beneath is the map of shear strain, beneath that is the map of longitudinal strain. In the legend
the strains can be converted to displacement gradient by multiplying by two.
5.1. Conclusions
(1) The problem of validating fault interpretation from seismic
data can be addressed using displacement and strain analysis.
(2) There is a simple relationship between instantaneous displace-
ment gradients and wall-rock shear and longitudinal strains.
(3) For unrestricted faults, reasonable natural upper limits to the
magnitudes of shear strain and longitudinal strain in the dip
slip direction are 0.05 and 0.1 respectively.
(4) Faults with strains that lie above these bounds are unlikely to
be correlated correctly.
(5) Fault interpretations that minimize the wall-rock strains
provide the most feasible geological solution.
Acknowledgements
We would like to thank the many colleagues at Badley Geo-
science Ltd and the Fault Analysis Group whose ideas have helped
to shape this work. The manuscript has beneted from editing by
Chris Wibberley and the careful and constructive reviews of Matt
Pachell and one anonymous reviewer.
References
Bailey, W.R., Walsh, J.J., Manzocchi, T., 2005. Fault populations, strain distribution
and basement fault reactivation in the East Pennines Coaleld, UK. Journal of
Barnett, J.A., Mortimer, J., Rippon, J.H., Walsh, J.J., Watterson, J., 1987. Displacement
geometry in the volume containing a single normal fault. American Association
of Petroleum Geologists Bulletin 71, 925937.
Boult, P.J., 1999. Maturity Modelling of the Casterton Formation and Killara Coals in
PEP 111 and PEP 101. Boral Energy Resources Ltd., Otway Basin, Victoria.
Unpublished report.
Boult, P.J., Hibburt, J.E. (Eds.), 2002. The Petroleum Geology of South Australia. Vol.
1: Otway BasinPetroleum Geology of South Australia Series, second ed., vol. 1.
Department of Primary Industries and Resources, South Australia.
Boult, P, Freeman, B., Yielding, G., Menpes, S., Diekman, L.J., 2008. A minimum-strain
approach to reducing the structural uncertainty in poor 2D seismic data,
Gambier Embayment, Otway Basin, Australia. In: Proceedings of the Third
Eastern Australasian Basins Symposium, Sydney, September 14th17th.
Fig. 6. (a) Perspective view of faults from the reinterpretation, colour coded and contoured in dip separation. (b) Shear strain (upper) and longitudinal strain (lower) maps of the
WNWESE fault set. (c) Shear strain (upper) and longitudinal strain (lower) maps of the NWSE fault set.
Bouvier, J.D., Kaars-Sijpesteijn, C.H., Kluesner, D.F., Onyejekwe, C.C., Van der Pal, R.C.,
1989. Three-dimensional seismic interpretation and fault sealing investigations,
Nun River Field, Nigeria. American Association of Petroleum Geologists Bulletin
73, 13971414.
Chapman, T.J., Meneilly, A.W., 1991. The displacement patterns associated with
a reverse-reactivated, normal growth fault. In: Roberts, A.M., Yielding, G.,
Freeman, B. (Eds.), The Geometry of Normal Faults. Geological Society of Lon-
don, Special Publication, vol. 56, pp. 183191.
Eshelby, J.D., 1957. The determination of the elastic eld of an ellipsoidal inclusion and
relatedproblems. Proceedings of theRoyal Societyof London, SeriesA241, 376396.
Essential Petroleum Resources Ltd., 2006. Internal Presentation to the PEL 72 Joint
Venture Technical Committee.
Freeman, B., Badley, M.E., Yielding, G., 1990. Fault correlation during seismic
interpretation. First Break 8, 8795.
Kim, Y.-S., Sanderson, D.J., 2005. The relationship between displacement and length
of faults: a review. Earth-Science Reviews 68, 317334.
Lee, S., Wolberg, G., Shin, S.Y., 1997. Scattered data interpolation with multilevel B-
splines. IEEE Transactions on Visualization and Computer Graphics 3, 228244.
Lovibond, R., Suttill, R.J., Skinner, J.E., Aburas, A.N., 1995. The hydrocarbon potential
of the Penola Trough, Otway Basin. APEA Journal 35, 358371.
Manzocchi, T., Walsh, J.J., Nicol, A., 2006. Displacement accumulation from
earthquakes on isolated normal faults. Journal of Structural Geology 28,
16851693.
Needham, D.T., Yielding, G., Freeman, B., 1996. Analysis of fault geometry and
displacement patterns. In: Buchanan, P.G., Nieuwland, D.A. (Eds.), Modern
Developments in Structural Interpretation, Validation and Modelling. Geolog-
ical Society of London, Special Publication, vol. 99, pp. 189199.
Nicol, A., Watterson, J., Walsh, J.J., Childs, C., 1996. The shapes, major axis orienta-
tions and displacement patterns of fault surfaces. Journal of Structural Geology
18, 235248.
Pollard, D.D., Segall, P., 1987. Theoretical displacements and stresses near fractures
in rock: with applications to faults, joints, veins, dikes, and solution surfaces. In:
Atkinson, B.K. (Ed.), Fracture Mechanics of Rock. Academic Press, London, pp.
277349.
Rippon, J.H., 1985. Contoured patterns of the throw and hade of normal faults in the
Coal Measures (Westphalian) of north-east Derbyshire. Proceedings of the
Yorkshire Geological Society 45, 147161.
Schultz, R.A., Soliva, R., Fossen, H., Okubo, C.H., Reeves, D.M., 2008. Dependence of
displacementlength scaling relations for fractures and deformation bands on
the volumetric changes across them. Journal of Structural Geology 30, 1405
1411.
Schultz, R.A., Fossen, H., 2002. Displacementlength scaling in three dimensions:
the importance of aspect ratio and application to deformation bands. Journal of
Soliva, R., Schultz, R.A., Benedicto, A., 2005. Three-dimensional displacement
length scaling and maximum dimension of normal faults in layered rocks.
Walsh, J.J., Watterson, J., 1989. Displacement gradients on fault surfaces. Journal of
Wibberley, C.A.J., Petit, J.-P., Rives, T., 1999. Mechanics of high displacement gradient
faulting prior to lithication. Journal of Structural Geology 21, 251257.
Willemse, E.M.J., 1997. Segmented normal faults: correspondence between three-
dimensional mechanical models and eld data. Journal of Geophysical Research
102 (B1), 675692.
Excavation induced fractures in a plastic clay formation: Observations at the
HADES URF
Philippe Van Marcke
*
, Wim Bastiaens
1
SCKCEN, Boeretang 200, 2400 Mol, Belgium
Article history:
1 December 2009
Accepted 20 January 2010
Available online 13 February 2010
Keywords:
Geological disposal
Boom Clay
Tunnel excavation
Excavation induced fractures
Hydro-mechanical behaviour
Self-sealing
a b s t r a c t
The Belgian research programme for geological disposal of radioactive waste focuses on the Boom Clay as
the potential host rock formation. To examine the feasibility of constructing an underground repository
in this clay layer, an underground research facility HADES has been constructed in several stages since
1980. The two galleries most recently excavated, the Connecting Gallery in 2002 and the Praclay Gallery
in 2007, were constructed by means of an industrial method using a tunnelling machine. During these
excavations the hydro-mechanical response of the clay was characterised.
A fracture pattern was observed consistently during the excavation of both galleries. The extent of this
fractured zone was determined for the Connecting Gallery, but requires some further study. A strong
hydro-mechanical coupling and a clear time dependency were noticed, even at an unexpectedly large
distance from the excavation. Furthermore the Boom Clay responds in an anisotropic manner to the
excavation due to anisotropy in the in situ stress state and the Boom Clay characteristics. Self-sealing
processes were observed and appear to occur relatively fast.
1. Introduction
The geological disposal of radioactive waste has been studied in
Belgium since the early seventies by the Belgian Nuclear Research
Centre (SCKCEN). The research is focused on the Boom Clay:
a plastic clay layer that is found from a depth of 190 m under the
site of SCKCEN in Mol (in the northeast of Belgium) where it has
a thickness of about 100 m. It has a low hydraulic conductivity (in
the order of 10
12
m/s) and displays a plastic behaviour which
results in self-sealing properties and a relatively high convergence
when excavating galleries in it.
In 1980 SCKCEN started the construction of the underground
facility HADES at a depth of 225 m (Fig. 1). Its main purpose was to
examine the feasibility of constructing such a repository and to
provide SCKCEN with an underground infrastructure for experi-
mental research on the geological disposal of radioactive waste. Not
much knowledge and experience on excavating in a deep plastic
clay formation were available at that time. The work during this
phase is therefore considered to be pioneering.
In 2002 the second shaft was connected with the existing
underground infrastructure by the Connecting Gallery (80 m
long and 4.8 m in external diameter). This was done in an
industrial manner by the use of a tunnelling machine. Several
measurement and research programmes were carried out before,
during and after the construction works to characterise the
hydro-mechanical response of the clay around the repository
(Bastiaens et al., 2003; Bernier et al., 2003). In particular the
fracture pattern resulting from the excavation was characterised.
In 2007 the Praclay Gallery (45 m long and 2.5 m in external
diameter) was constructed perpendicular to the Connecting
Gallery. Again, the hydro-mechanical response was measured
and characterised.
This paper discusses these measurements and observations.
First the Boom Clay characteristics are given, then the used exca-
vation technique is described after which the hydro-mechanical
observations are presented.
2. Boom Clay characteristics
The Boom Clay is a silty clay characterised by a structure of
bands that are several tens of centimetres thick, reecting mainly
cyclical variations in grain size (silt and clay content) due to uc-
tuations in the wave action on the sedimentation medium and to
variations in the carbonate and organic matter contents. Typical
concretions, known as septaria, are found in the marly bands
occurring throughout the thickness of the formation.
* Corresponding author. Tel.: 32 14 33 29 88; fax: 32 14 32 37 09.
E-mail addresses: pvmarcke@sckcen.be (P. Van Marcke), wbastiae@sckcen.be
(W. Bastiaens).
1
Tel.: 32 14 33 27 90; fax: 32 14 32 37 09.
doi:10.1016/j.jsg.2010.01.010
The formation belongs to the Rupelian which is the geological
part of the Tertiary Period with an age between 36 and 30 million
years. It is found at a depth of about 190 m under the SCKCEN site
of Mol where it has a thickness of about 100 m (see Fig. 2). The
Boom Clay layer is almost horizontal (it dips 1e2% towards the NE)
and water bearing sand layers are situated above and below it.
Due to its vertical lithological heterogeneity the mineralogy of
the Boom Clay is characterised by a wide variation in the content of
clay minerals (from 30 to 70% volume, dry matter). In descending
Fig. 1. Construction history of the underground research facility HADES.
Fig. 2. Geological section under the Mol site.
Table 1
Undrained geomechanical characteristics of the Boom Clay at the depth of HADES.
Property Unit Value
Young's modulus tangential at the origin E 200e400 MPa
Poisson's ratio n 0.40e0.45
Unconned compressive strength UCS 2 MPa
Angle of friction 4 4
Cohesion c 0.5e1.0 MPa

Plastic limit w
p
23e29%
Liquid limit w
l
55e80%
Plastic index IP 32e51%
Hydraulic conductivity k w10
12
m/s
Porosity n 39%
Water content w 30e40 vol%
Fig. 3. Tunnelling machine used for the excavation of (a) the Connecting Gallery and
(b) the Praclay Gallery.
P. Van Marcke, W. Bastiaens / Journal of Structural Geology 32 (2010) 1677e1684 1678
order of importance, the non-argillaceous fraction of the sediment
consists of quartz, feldspars, carbonates and pyrite. The organic
matter content ranges from 1 to 3% weight, dry matter. The water
content ranges from 30 to 40% volume.
The Boom Clay displays furthermore a visco-elasto-plastic
behaviour (Bastiaens et al., 2006; Coll et al., 2007; Cui et al., 2009;
Le et al., 2007), such that its convergence in the longer termis high.
At the level of HADES, total stress and pore water pressure are
respectively 4.5 and 2.2 MPa. The vertical stress is estimated to be
slightly higher than the horizontal ones (K
0
w 0.9). The undrained
geomechanical characteristics of the Boom Clay at the depth of the
existing facilities are given in Table 1 (Horseman et al., 1987).
3. Excavation by means of a tunnelling machine
Both the excavation of the Connecting Gallery and the Praclay
Gallery were performed by the use of a tunnelling machine (Fig. 3).
This machine consists of a roadheader with drillhead which is
placed under the protection of a shield.
Fig. 4 shows the different steps in the excavation procedure. The
roadheader moves along the clay front and the majority of the face
is removed (Fig. 4a). Once all the clay within the reach of the
roadheader is removed (Fig. 4b), the complete tunnelling machine
is pushed forward against the already placed lining (Fig. 4c). By
doing so the outer rim of the excavation face is cut by the edges of
the shield. This ensures a smooth and circular excavation prole.
When the tunnelling machine has progressed over a sufciently
large distance, a new lining ring can be placed by the use of an
erector (Fig. 5). The placement of the lining as soon as possible after
the excavation minimises the convergence. Also the use of a rigid
and expanding type of lining (see further) avoids further conver-
gence and the need to perform post-grouting.
The lining of the galleries consists of concrete wedge blocks
(Fig. 6a). The segments are unreinforced and unbolted and each
ring is independent. By introducing shorter, wedge shaped
segments the ring can be expanded against the circular excavated
prole and in that way a post-stressing effect in the lining can be
induced (Fig. 6b). The dimensions of the key segment can be
Fig. 4. Excavation phases: (a) the roadheader is moved along the clay front to remove the clay; (b) all the clay is removed within the reach of the roadheader; (c) the tunnelling
machine is pushed forward against the already built lining. Once the tunnelling machine has sufciently progressed, a new lining ring can be installed.
adjusted by sawing the segment at its front or at its back end. This
allows adjusting the diameter of the expanding lining to the
diameter of the excavated prole.
The Connecting Gallery was excavated at a rate of 3 m/day while
the excavation rate of the Praclay Gallery was 2 m/day. Achieving
a mean excavation rate of 10 m/day is assumed to be realistic for
a future repository when also a larger access shaft will be available
for transporting cuttings and material.
Fig. 5. Erector of the tunnelling machine used to place the gallery lining segments: (a) bird-wing erector used when constructing the Connecting Gallery; (b) rotary erector used for
the construction of the Praclay Gallery.
Fig. 6. (a) Picture of the Connecting Gallery showing the concrete wedge blocks
making up the gallery lining. (b) By using key segments the lining can be expanded and
the diameter of the lining can be adjusted according to the diameter of the excavated
prole.
Fig. 7. Modelled stresses along the gallery axis ahead of the excavation front: (a) axial,
radial and differential stresses; (b) differential stresses.
4. Hydro-mechanical response of the Boom Clay to the
excavation
The excavation of a gallery results in differential stresses that
are relatively high compared to the strength of the Boom Clay
(Table 1). Fig. 7 shows the stresses around an excavation as
obtained by a perfect elasto-plastic model using the parameters
from Table 1. The nite difference code FLAC2D was used to
compute the stresses for this model.
These high differential stresses result in shear fractures which
were, during the excavation of both the Connecting as the Praclay
Gallery, systematically observed on the sidewalls of the excavation
as well as on the excavation front (Fig. 8).
These observations revealed a fracture pattern similar in both
excavations (Fig. 9). The pattern consists of two conjugated
fracture planes: one in the upper part dipping towards the
excavation direction and the other in the lower part dipping
towards the opposite direction. These fracture planes intersect at
mid-height of the gallery and they are slightly curved. Their
intersection both with a vertical and a horizontal plane passing
through the gallery axis is a curve. Due to the anisotropy of the
stresses in the clay e the vertical stresses are higher than the
horizontal stresses e this curve is much more pronounced
vertically than horizontally. The distance between successive
fractures is a few decimetres. These observations are consistent
with observations from earlier excavations such as the Test Drift
(Mertens et al., 2003, 2004).
The radial extent of the fractured zone around the Connecting
Gallery was determined from cores taken around the gallery.
Fractures were found up to a distance of about 1 m(Fig. 9). The axial
extent was found by modelling and appeared to be about 6 m. This
gure was conrmed when the Connecting Gallery reached the
Test Drift (Fig. 1): about 6 m from the end of the Test Drift fractures
induced by the excavation of the latter were observed.
The extent of the fractured zone around the Praclay Gallery still
needs to be evaluated. This will be done when instrumentation
around the gallery is placed. However, piezometers placed parallel
to the Praclay Gallery at a horizontal distance of 0.75 mdid not drop
to atmospheric pressure and thus suggest that the fractures do not
extend up to 0.75 m in a horizontal plane.
Since the Praclay Gallery is constructed perpendicular to the
Connecting Gallery, the excavation of the rst passes through the
fractured zone induced by the construction of the latter. These
fractures were observed in the rst fewmetres of the Praclay Gallery
excavation (Fig. 10). At rst sight, with the naked eye, the fractures
appeared to curve in a direction opposite to that expected. This is
because the fracture is the intersection between the fracture plane
and the excavation front which is spherical. These fractures were
observed up to an excavation distance of 6 m. This is well beyond the
formerly determined radial extent of 1 m. Possibly microfractures up
to that extent were induced during the excavation of the Connecting
Gallery, but were not observed in the cores taken radially around the
Connecting Gallery. As a result of the stress redistribution created by
the excavation of the Praclay Gallery these microfractures might
have formed macrofractures. Another hypothesis is that the fractures
induced by the excavation of the Connecting Gallery in 2002 were
reactivated and their extent increased due to additional stress
redistribution by the Praclay Gallery excavation. At the crossing, and
especially at the level of the fractures, other stress orientations exist
ahead of the face than those that are present beyond the start-up
zone during the excavation.
Fig. 11 shows the pore water pressure as function of the radial
distance to the Connecting Gallery. The pore water pressures are
measured by two piezometers around the gallery and are plotted as
a percentage of the undisturbed in situ pressure at their location.
The gallery lining is shown as well. The pore water pressure
distribution around the gallery is remarkably large and anisotropic.
The hydraulically disturbed zone (HDZ) in a horizontal plane seems
to range up to about 20 m into the host rock. In a vertical plane
however, an equilibrium pressure is not reached even at 40 m. The
values are all signicantly lower than 100%. This is probably due to
the disturbance caused by the installation of the piezometer itself.
Because of the decreased stresses near the excavation (Fig. 7),
the clay will converge towards the excavation. An instantaneous
convergence results from the elasticity of the clay while the creep
behaviour of the clay gives rise to a time-dependent convergence.
Fig. 8. Fractures observed (a) on the sidewalls (Connecting Gallery) and (b) the
excavation front (Praclay Gallery).
Fig. 9. Fracture pattern observed in the Connecting Gallery. A similar pattern was
observed for the Praclay Gallery. The radial extent of the fracture pattern around the
Connecting Gallery was estimated at 1 m, the axial extent at 6 m. The extent of the
fractures around the Praclay Gallery still needs to be evaluated.
This dilatation, combined with the undrained conditions on the
short term, can result in negative pore water pressures.
In spite of the higher vertical stresses, the convergence appeared
to be higher in the horizontal direction than in the vertical direc-
tion. This can be explained by the shape of the fracturation pattern.
Since the dip directions of the fractures are roughly parallel to the
gallery axis, vertical de-stressing can assumed to be partly affected
by fracturing ahead of the excavation. This in turn reduces the
vertical convergence afterwards. However, the clay at the sides of
the gallery converges as if no fracturing had occurred and thus the
horizontal convergence is larger.
Once the lining is placed, the convergence of the clay will result
in an increase of the pressure exerted by the clay on the lining
(Fig. 12). This pressure build-up is monitored through embedded
vibrating wire strain gauges. Four gallery lining rings are assembled
with these strain-gauged segments. The strain measured by these
gauges is converted into mechanical stress in the segments, which
on its turn can be related to the total pressure in the host rock
acting on the lining.
During the consolidation of the clay around the excavation, self-
sealing occurs as the open fractures close progressively due to the
creep behaviour of the clay and to its swelling when it rehydrates.
Evidence of the self-sealing of BoomClay was observed fromcoring
and instrumentation campaigns as the clay appeared to close
spontaneously against the borehole casings. Open boreholes close
up completely and if an open casing (not closed at the end) is
installed in a borehole, the BoomClay ows inside the tube (Bernier
et al., 2006).
Within the SELFRAC EC project (Bernier et al., 2006), the fast
self-sealing capacity of Boom Clay has been demonstrated and
visualised by means of X-ray CT. Fig. 13a shows the CT image of
a Boom Clay sample shortly after an articial fracture was created
and before saturation of the sample. Fig. 13b gives an image of the
sample 4.5 h after saturation. Closure of the fracture is clearly
observed. Sealing was conrmed by measuring the hydraulic
conductivity measurement of the sample which returned to the
same order of magnitude as the undisturbed value (w10
12
m/s)
after a few months.
In situ experiments have also revealed sealing processes in the
Boom Clay. Two years after the excavation of the Connecting
Gallery, sealing had reduced the radial extent of the fractured zone
from 1 m to less than 0.6 m.
An impact of the ventilation of the gallery on the fracturing or
a desaturation of the Boom Clay around the gallery has not been
observed in the HADES URL. Moreover, a laboratory mock-up test,
PHEBUS (Robinet et al., 1998), was setup to study the desaturation
of the Boom Clay due to ventilation. In the test a controlled
humidity air (33% humidity) was cycled in an orice drilled in
a cylinder sample and the hydric exchange between the air and
Boom Clay was measured. After dismantling also the moisture
prole of the mock-up was measured. The desaturation zone
appeared to be very limited.
Fig. 10. (a) Excavation front of the Praclay Gallery 0.5 m from the Connecting Gallery.
(b) Excavation front of the Praclay Gallery 4.0 m from the Connecting Gallery.
Fig. 11. Pore water pressure in function of the distance to the excavation. The pore water pressure is expressed as a percentage of the undisturbed in situ value.
5. Conclusions
In the framework of the Belgian research programme for the
geological disposal of radioactive waste, an underground
research facility has been built in the Boom Clay layer at a depth
of 225 m. The response of the Boom Clay to the excavation of
a gallery was observed and characterised during the excavation
of the Connecting Gallery in 2002 and the Praclay Gallery in
2007.
The excavation induces fractures in the surrounding clay layer.
These fractures are created according to a pattern consisting of
two conjugated fracture planes: one in the upper part dipping
towards the excavation direction, the other in the lower part
dipping towards the opposite direction. The fractures curve
towards the excavation. The extent of this fractured zone depends
on the diameter of the excavated gallery. For a gallery of 4.8 m, the
radial extent appeared to be 1 m. The extent of the fractures around
the Praclay Gallery (2.5 m in external diameter) still needs to be
determined.
Furthermore in situ measurements and numerical modelling
were performed to characterise the geomechanical behaviour of the
clay. A strong hydro-mechanical coupling and a clear time depen-
dency were noticed, even at an unexpectedly large distance fromthe
excavation.
Once the lining is placed, the consolidation of the clay results
in an increase of the pressure exerted by the clay on the lining
and open fractures close progressively. In other words, self-seal-
ing processes occur in the Boom Clay and they occur relatively
fast. These processes reduce the extent of the fractured zone
around the Connecting Gallery from 1 m to 0.6 m within one
year.
Finally, the hydro-mechanical response of the clay appeared to
be anisotropic. This was attributed to anisotropy in in situ stress
conditions and in the hydro-mechanical properties of the Boom
Clay.
Acknowledgements
This work is performed in close cooperation with, and with the
nancial support of ONDRAF/NIRAS, the Belgian Agency for
Radioactive Waste and Fissile materials, as part of the program on
geological disposal of high level/long lived radioactive waste that is
carried out by ONDRAF/NIRAS.
Fig. 12. Evolution of the average pressure exerted on a lining ring of the Connecting Gallery.
Fig. 13. (a) CT image of an articially fractured Boom Clay sample. (b) CT image of the
same sample after a few hours of saturation.
References
Bastiaens, W., Bernier, F., Buyens, M., Demarche, M., Li, X.L., Linotte, J.M., Verstricht, J.,
2003. The Connecting Gallery e the Extension of the Hades Underground
Research Facility at Mol, Belgium. EURIDICE report 03-294. Mol: ESV EURIDICE.
Bastiaens, W., Bernier, F., Li, X.-L., 2006. An overview of long-term HM measure-
ments around HADES URF. In: EUROCK 2006, Multiphysics Coupling and Long
Term Behaviour in Rock Mechanics, Lige, Belgium, 9e12 May 2006/Belgian
Rock Mechanics Society. British Indian Ocean Territory, Taylor & Francis/Bal-
kema, London, pp. 15e26.
Bernier, F., Li, X.L., Verstricht, J., Barnichon, J.D., Labiouse, V., Bastiaens, W., Palut, J.M.,
Ben Slimane, K., Ghoreychi, M., Gaombalet, J., Huertas, F., Galera, J.M., Merrien, K.,
Elorza, F.J., Davies, C., 2003. CLIPEX: Clay Instrumentation Programme for the
Extension of an Underground Research Laboratory, Final Report, EUR 20619 EN.
Bernier, F., Li, X.L., Bastiaens, W., Ortiz, L., Van Geet, M., Wouters, L., Frieg, B.,
Blmling, P., Desrues, J., Viaggiani, G., Coll, C., Chanchole, S., De Greef, V., Hamza,
R., Malinsky, L., Vervoort, A., Vanbrabant, Y., Debecker, B., Verstraelen, J.,
Govaerts, A., Wevers, M., Labiouse, V., Escofer, S., Mathier, J.F., Gastaldo, L.,
Bhler, Ch., 2006. Fractures and Self-Healing within the Excavation Disturbed
Zone in Clays. EC report on the SELFRAC project.
Coll, C., Collin, F., Radu, J.P., Illing, P., Schroeder, Ch., Charlier, R., 2007. Long
Term Behaviour of Boom Clay: Inuence of Clay Viscosity on the Far
Field Pore Pressure Distribution. Final report of cooperation project with
Ulg.
Cui, Y.-J., Le, T.T., Tang, M., Delage, P., Li, X.-L., 2009. Investigating the time-depen-
dent behaviour of Boom Clay under thermomechanical loading. Gotechnique
59 (4), 319e329.
Horseman, S.T., Winter, M.G., Entwistle, D.C., 1987. Geotechnical Characterisation of
Boom Clay in Relation to the Disposal of Radioactive Waste. Final report. EUR
10987. Commission of the European Communities, , Luxembourg.
Le, T.T., Cui, Y.-J., Delage, P., Li, X.-L., 2007. Creep Behaviour in Thermal and
Mechanical Consolidation Tests on Boom clay. Paper submitted to the 11th
Congress of the International Society for Rock Mechanics, Lisbon, Portugal,
Organised by ISRM; 9e13 July 2007.
Mertens, J., Vandenberghe, N., Wouters, K., Sintubin, M., 2003. The origin and
development of joints in the BoomClay formation (Rupelian) in Belgium. In: Van
Rensbergen, P., Hillis, R.R., Maltman, A.J., Morley, C.K. (Eds.), Subsurface Sediment
Mobilization. Geological Society Special Publications, 216, pp. 309e321.
Mertens, J., Bastiaens, W., Dehandschutter, B., 2004. Characterisation of induced
discontinuities in the Boom Clay around the underground excavations (URF,
Mol, Belgium). Applied Clay Science 26, 413e428.
Robinet, J.C., Pasquiou, A., Palut, J.M., Noynaert, L., 1998. Etudes des transferts hydri-
ques entre le massif argileux et les excavations dans un laboratoire souterrain
pour le stockagede dchets radioactifs. Essai Phebus, Rapport nal, EUR17792FR.
High shear strain behaviour of synthetic muscovite fault gouges under
hydrothermal conditions
Esther W.E. Van Diggelen
*
, Johannes H.P. De Bresser, Colin J. Peach, Christopher J. Spiers
Faculty of Geosciences, Utrecht University, P.O. Box 80.021, 3508 TA Utrecht, Netherlands
Article history:
24 July 2009
Keywords:
Frictional behaviour
Muscovite
Simulated fault gouge
Rotary shear experiments
Hydrothermal conditions
a b s t r a c t
Major continental fault zones typically contain phyllosilicates and have long been recognised as zones of
persistent weakness. To establish whether the presence of micas can explain this weakness, we studied
the frictional behaviour of simulated muscovite fault gouge by performing rotary shear experiments in
the temperature range 20700

C, under constant effective normal stresses of 20100 MPa, a xed uid
pressure of 100 MPa and at sliding velocities of 0.033.7 mm/s, reaching shear strains up to 100. Cataclasis
causes substantial grain size reduction up to 600

C. With increasing strain, both pervasive and localized
cataclasis and related compaction result in strain hardening, until steady state is reached. This reects
the progressive development of a continuous network of ne grained, hardening bands. Coarse grained
relict lenses between these bands show folded and kinked muscovite grains indicative of ductile
mechanisms. Samples deformed at 700

C show evidence for chemical alteration and partial melting.
Since our data suggest that muscovite gouge strengthens with depth and strain, it is questionable
whether its presence can contribute to the long-term weakness of major crustal fault zones, unless
a substantial decrease in strength occurs at shear strain rates lower than attained in our study.
1. Introduction
Major continental fault zones have long been recognised by eld
geologists as zones of highly localized deformation and hence of
persistent weakness (Balfour et al., 2005; Holdsworth et al., 2001;
Imber et al., 1997; Jefferies et al., 2006b; Rutter et al., 2001; White
et al., 1986; Zoback et al., 2007; Zoback et al., 1987). The implication
is that the quartz-, feldspar-, and phyllosilicate-rich fault rocks
often found within such faults must be weaker than the
surrounding country rock. Laboratory rock friction experiments
have conrmed that typical continental fault rocks are indeed weak
compared to intact quartzo-feldspathic host rocks (Bos et al., 2000;
Dieterich, 1978; Hickman, 1991; Holdsworth, 2004; Lachenbruch
and Sass, 1980; Logan and Rauenzahn, 1987; Morrow et al., 2000;
Morrow et al., 1992; Niemeijer and Spiers, 2005, 2006; Shimamoto
and Logan, 1981; Takahashi et al., 2007).
Lab data for long-term fault strength are usually expressed in
terms of Byerlees Rule for fault friction (friction coefcient m 0.6
0.9) giving way to plastic ow at depths of 1015 km, thus
producing the classical Christmas Tree strength prole (Bos and
Spiers, 2002; Byerlee, 1978; Goetze and Evans, 1979; Holdsworth
et al., 2001; Niemeijer and Spiers, 2005). Such proles t well with
the depth distribution of seismicity on major faults (Scholz, 2002;
Sibson, 1983). However, laboratory-based strength proles are not
fully consistent with other geophysical observations on major
mature fault zones, like the San Andreas fault zone. In the case of
the San Andreas fault, the lack of a positive heat owanomaly, and
the high angles (w70
) measured between the in-situ principal

stress s
1
and the fault surface (Townend and Zoback, 2004; Zoback
et al., 2007), imply a low mean resolved shear stress on the fault of
around 1020 MPa (Lachenbruch and Sass, 1980; Zoback et al.,
1987) and a coefcient of friction of only w0.2. This is far less than
measured for (simulated) fault rocks in laboratory experiments
(Blanpied et al., 1995; Bos and Spiers, 2000; Byerlee, 1978;
Carpenter et al., 2009; Moore and Lockner, 2004; Morrow et al.,
2000; Morrow et al., 1992; Nakatani and Scholz, 2004; Niemeijer
and Spiers, 2005; Tembe et al., 2006a; Tembe et al., 2006b).
A widely proposed explanation for the inferred long-term
weakness of major crustal faults is that high pore uid pressure
reduces the effective shear stress required for slip. Fluids released
during compaction and/or dehydration reactions at depth might
locally increase uid pressures to approach lithostatic values
(Byerlee, 1990; Collettini and Barchi, 2002; Faulkner and Rutter,
2001; Hickman et al., 1995; Miller et al., 1996; Miller and Olgaard,
1997; Sibson, 2004; Sleep, 1995). However, uid pressures inside
a fault can only increase if the zone has a low permeability, caused
* Corresponding author. Tel.: 31 30 2535079; fax: 31 30 2537725.
E-mail address: diggelen@geo.uu.nl (E.W.E. Van Diggelen).
doi:10.1016/j.jsg.2009.08.020
by ubiquitous phyllosilicates or continuous wall rock cementation
(Schleicher et al., 2008; Schleicher et al., 2009b; Zhang et al., 2001).
Faulkner and Rutter (2001) showed that dehydration of phyllosili-
cates (assuming 50% phyllosilicate content at 1520 km depth) is
sufcient to maintain an elevated uid pressure for w12 ky, but
insufcient to weaken a large fault on geologic timescales. More-
over, in-situ measurements in fault zones show little evidence for
continuously high uid pressures (Wintsch et al., 1995). In partic-
ular, uid pressure measurements in the SAFOD (San Andreas Fault
Observatory at Depth) drill hole indicate near-hydrostatic rather
than lithostatic uid pressures at depths up to 3.5 km (Tembe et al.,
2006b; Zoback et al., 2007). It thus seems unlikely that high uid
pressures alone can account for the apparent long-term weakness
of large scale fault zones.
Alternatively, the presence of relatively weak reaction product
minerals, such as clays, micas, chlorite or talc may explain the
inferred weakness of major faults (Arancibia and Morata, 2005;
Jefferies et al., 2006a; Logan and Rauenzahn, 1987; OHara, 2007;
Shea and Kronenberg, 1992; Wibberley, 1999; Wintsch et al.,
1995). Experiments have shown that many such phyllosilicates are
signicantly weaker than quartz and feldspars (Ikari et al., 2007;
Logan and Rauenzahn, 1987; Mariani et al., 2006; Moore and
Lockner, 2004; Moore and Lockner, 2007; Moore et al., 1997;
Morrow et al., 2000; Morrow et al., 1992; Shimamoto and Logan,
1981) and are stable up to 15 km depth. It follows that a contin-
uous, through-going foliation of phyllosilicate minerals may
strongly inuence fault zone rheology on a crustal scale (Holds-
worth et al., 2001; Imber et al., 1997; Ranalli, 1995; Rutter et al.,
2001; Rutter et al., 1986). Reaction-softening of this type has
frequently been proposed as an explanation for the weakness of
the San Andreas fault zone (Evans and Chester, 1995; Moore and
Rymer, 2007; Wintsch et al., 1995). Recently, core material
retrieved during SAFOD phase 3 drilling has shown that the
actively deforming strands of the fault contain a grayish-black,
foliated, phyllosilicate-rich fault gouge. Visible clasts make up
about 5% of the rock volume and consist of fragments of serpen-
tinite, sandstone and siltstone protolith (SAFOD core atlas, 2007).
These observations conrm that phyllosilicate foliation develop-
ment and reaction weakening are key processes in determining
fault zone rheology.
Additionally, uid-assisted deformation processes, such as
pressure solution, have been put forward as a mechanism for
weakening of fault rocks (Blanpied et al., 1995; Chester, 1995; Col-
lettini and Holdsworth, 2004; Hickman et al., 1995; Jefferies et al.,
2006b; Kanagawa, 2002; Kanagawa et al., 2000; Lehner and
Bataille, 1984; Schleicher et al., 2009a; Wu et al., 1975). Experi-
ments on wet quartz gouge under hydrothermal conditions
(Kanagawa et al., 2000; Niemeijer et al., 2002), and on wet granular
halite used as an analogue (Bos et al., 2000), show competition
between pressure solution, compaction and cataclasis, leading to
high frictional strength (m 0.80.9) at low slip rates. However,
experiments on simulated gouges consisting of halite plus
kaolinite, or of halite plus muscovite, show a reduction in m-values
to 0.30.4 (Bos and Spiers, 2000, 2001; Niemeijer and Spiers, 2005).
This is due to the development of a through-going phyllosilicate
foliation on which frictional slip occurs with accommodation by
pressure solution of the intervening halite clasts. Bos and Spiers
(2002) and Niemeijer and Spiers (2005, 2007) developed a micro-
physical model explaining their experimental results and predict-
ing upper crustal strengthdepth proles for foliated quartz-mica
fault rock some 25 times weaker than obtained using Byerlees
Rule. The low strength predicted is due to the serial effect of easy
pressure solution of quartz clasts at low sliding rates and the low
frictional strength (m 0.3) assumed for the enveloping phyllosili-
cate foliation. In the experimental study of Mariani et al. (2006),
pure muscovite fault gouge shows strain rate insensitive strain
hardening behaviour at shear strain rates faster than 1.4 10
5
s
1
at 700

C and at all strain rates tested (10
7
to 10
3
s
1
) at
temperatures of 400600

C. This behaviour is primarily attributed
to mutual misalignment of mica akes with contributions from
progressive porosity reduction and formation of oblique shear
features (Mariani et al., 2006). At 700

C and strain rates lower than
1.4 10
5
s
1
, the shear strength drops dramatically. Deformation
of the muscovite gouge under these conditions has linear viscous
characteristics, possibly due to viscous glide of basal dislocations in
the mica becoming rate controlling (Mariani et al., 2006). The above
studies imply that both pressure solution and phyllosilicate folia-
tion development may play an important role in weakening major
fault zones, and demonstrate a crucial need for laboratory data on
the frictional and plastic ow strength of phyllosilicates under
conditions close to those in observed in nature.
As discussed, laboratory studies have shown that most phyllo-
silicate gouges are characterized by signicantly lower frictional
strength (m 0.20.5) than Byerlees Rule predicts (Ikari et al.,
2007; Mariani et al., 2006; Moore and Lockner, 2004; Morrowet al.,
1992; Niemeijer and Spiers, 2005; Scruggs and Tullis, 1998), espe-
cially in the presence of water (Moore and Lockner, 2004; Moore
et al., 1997; Morrow et al., 2000). Most laboratory experiments on
phyllosilicates have been performed at roomtemperature and/or at
low shear strains (g <15), without a pore uid. By contrast, upper
and mid-crustal fault rocks typically experienced large shear strains
at elevated temperatures (50350

C) in the presence of a chemi-
cally active pore uid. In order to improve our understanding of
fault zone deformation processes and the role of phyllosilicates
under upper and mid-crustal conditions, it is, therefore, necessary
to perform rock friction experiments on realistic gouge composi-
tions under hydrothermal conditions, and towards shear strains
that better correspond with natural settings. High temperatures
(300700

C) and elevated uid pressures (up to 103 MPa) were
attained by Mariani et al. (2006) in their experimental study of
muscovite fault gouge, but the saw-cut deformation geometry they
applied did not allow high shear strains to be reached (g for most
experiments in the order of 2.3), and constant normal stress was
difcult to maintain (He et al., 2007, 2006). In contrast, large shear
strains under controlled normal stress were reached in various
types of gouges by Scruggs and Tullis (1998) using an experimental
rotary shear set-up, but no experiments other than at room
temperature were performed.
The present study focuses on determining the frictional
behaviour of simulated muscovite fault gouge at shear strains
above 20 and occasionally even up to 100, under hydrothermal
conditions in the temperature range 20700

C. In our tests we
applied constant, controlled effective normal stresses of 20
100 MPa, at a xed uid pressure of 100 MPa with deionised water
as the pore uid. Samples were typically deformed at sliding
velocities in the range 0.033.7 mm/s, corresponding to shear strain
rates of w10
5
to 10
3
s
1
. We chose to investigate muscovite
because it is a principal phyllosilicate constituent of many exhumed
upper and mid-crustal faults, and because relatively little data
exists to support the m-value of 0.3 for muscovite used in recent
models for fault strength (Bos and Spiers, 2002; Niemeijer and
Spiers, 2005; Niemeijer and Spiers, 2007). The effects of normal
stress, sliding velocity, displacement (i.e. shear strain) and
temperature on the frictional behaviour of the muscovite gouges
were determined and combined with the results of microstructural
analysis of the deformed samples to elucidate the operative
microphysical processes. The mechanical data obtained are used to
construct a strength prole for a muscovite-dominated fault,
assuming a thermal gradient and other conditions similar to those
pertaining to the San Andreas fault zone.
E.W.E. Van Diggelen et al. / Journal of Structural Geology 32 (2010) 16851700 1686
2. Experimental approach
Our experiments consisted of high strain rotary shear experi-
ments performed on ne grained granular muscovite samples
(simulated muscovite fault gouge) under hydrothermal conditions
(P
f
100 MPa, T 20700

C). We primarily aimed to determine
friction coefcient and microstructural development as a function
of shear strain (g up to 100), effective normal stress (s
eff
20
100 MPa), temperature (T 20700

C), and sliding velocity
(V 0.033.7 mm/s).
2.1. Starting material
The synthetic fault gouge material used in this study is a natural
muscovite powder (Mica S), commercially obtained fromInternatio
B.V. and originally mined in Aspang, Austria. The chemical
composition of the muscovite is K
1.97
Na
0.07
(Al
3.02
Fe
0.94
Mg
0.50
)
(Si
7.27
Al
1.51
) O
20
(OH)
4
, determined using a Jeol JXA-8600 superp-
robe microprobe and assuming four formula units for the hydroxyl
group and normalized to oxygen. This material was previously used
in the experiments reported by Niemeijer and Spiers (2005, 2006).
Particle size analysis, performed using a Malvern particle sizer,
showed that the median grain size of the muscovite was w13 mm
(equivalent spherical diameter), with 90% of the grains lying
between 3 and 50 mm in size. Microprobe and Energy Dispersive X-
ray (EDX) analysis showed that the bulk material also contained
quartz (<10 vol%), apatite (<0.1 vol%), zircon (<0.1 vol%), and
calcium-carbonate (<0.1 vol%).
2.2. Experimental apparatus
The hydrothermal rotary shear apparatus, described in detail by
Niemeijer et al. (2008), was used to perform our high shear strain
experiments on the synthetic muscovite fault gouges (Fig. 1). In this
machine, the synthetic fault gouge sample is located between a pair
of opposing internal pistons made of Rene 41 with a mullite core to
reduce axial heat loss. The initially w1 mm thick gouge sample is
kept in place by a set of two mildly hardened stainless steel
conning rings (Thyssen type 1.4122), with an inner radius of 14
and 11 mm, respectively (Fig. 1c and d). To prevent localization of
slip along the pistonsample interface, a cross-hatch pattern has
been machined on the piston surfaces in contact with the sample.
This pattern consists of two sets of grooves 0.5 mm apart and
roughly 0.2 mm depth, resulting in asperities with a surface area of
w0.25 mm
2
, and has been roughened further by dry-sandblasting
with glass beads (diameter 120 mm) for extra grip on the gouge.
During a test, the piston-sample assembly is positioned inside
a 300 MPa pressure vessel, which in turn is located in an Instron
1362 loading frame (Fig. 1a). The top of the piston-sample assembly
is coupled to a pressure-compensated piston in the upper sealing
head. This piston in turn locks into the upper forcing block and the
attached torque and axial force gauge couple. Axial force is applied
to the forcing block via a low friction bearing and measured using
an in-line Instron load cell. The lower piston ts into a rotation-
proof slot in the bottomof the pressure vessel, which is itself rigidly
mounted in the lower forcing block (Fig. 1a). This is coupled to the
underlying servo-controlled, rotational drive system, which in turn
is rigidly coupled to the vertical Instron loading ram.
Normal stress is applied using the Instron loading ram, and is
measured externally by means of the (100 kN) Instron load cell that
can be held constant to within 0.2 MPa. Pore uid pressure is
applied directly to the sample by pressurizing the sealed vessel
with deionised water using a manually driven high pressure pump.
The uid pressure is measured externally using a pressure trans-
ducer with a resolution of 0.005 MPa. The sample is heated using an
internally mounted 3 kW Inconel-sheathed furnace element
(Fig. 1b). Temperatures up to 700

C are continuously measured to
within 1

C using two K-type thermocouples, one located in the
furnace element and the other mounted w5 mm below the sample
at a similar distance from the furnace element as the gouge. The
main vessel seals (upper lower O-rings and outer skin) are cooled
by means of a water cooling system at a ow rate of typically 3 l/
min. Vertical displacement of the upper (pressure-compensated)
loading piston, and hence compaction/dilatation normal to the
synthetic fault gouge, is externally measured using a linear variable
differential transformer (LVDT) located in the Instron loading ram
assembly, with a resolution of 0.005 mm.
The servo-controlled motor and gearbox system, located on the
Instron ram, rotates the vessel and the internally xed lower piston
at a xed rate, producing sliding velocities within the sample of
0.033.7 mm/s (equivalent to shear strain rates of the order of 10
5
to 10
3
s
1
), while the upper piston and coupled pressure-
compensated piston remain stationary with respect to the Instron
loading frame. Shear displacements up to w80 mm (equivalent to
a shear strain g of up to 130) are measured using an external
potentiometer with an accuracy of 0.001 mm, connected to the
lower (rotating) forcing block. Shear stress is measured externally
with an equivalent accuracy of 0.2 MPa (zero machine-friction
case), using the torque gauge couple mounted on the upper forcing
block. The measured shear stress reects the total mechanical
resistance to rotation and thus needs to be corrected for apparatus
friction, i.e., the frictional force exerted by the seals and by the
conning rings surrounding the gouge.
2.3. Experimental procedure
In each experiment, 0.30 g of sample material was loaded onto
one of the piston surfaces with the conning rings in place. The
material was then uniformly distributed over the piston surface
and lightly pre-compacted creating a layer w1 mm in thickness,
after which the opposing piston was lowered into the conning
rings and xed with the retaining screw (Fig. 1d). The piston-
sample assembly was subsequently lowered into the waterprimed
vessel, which in turn was sealed and placed in the Instron 1362
loading frame. The vessel was then moved upwards using the
Instron ram, and the piston assembly was engaged with the upper
forcing block plus the torque and axial force gauges.
After heating to the desired test temperature (<30 min) and
applying a uid pressure of 100 MPa, a normal stress up to 200 MPa
was applied using the Instron. The system was then left to equili-
brate thermally for 45 min. During this time, the sample compacted
from about 40% to 2530% porosity (see Section 2.4), i.e., to
a thickness of w0.6 mm. After equilibration, the rotary drive was
typically switched on to give a sliding velocity of 1.0 mm/s until
a displacement of w10 mm was reached, beyond which the effec-
tive normal stress or sliding velocity were systematically varied.
The following types of experiments were performed at a xed
uid pressure of 100 MPa:
(1) sliding experiments at constant effective normal stress
(s
eff
100 MPa) and temperature (T 500

C), and at xed
sliding velocities of V 0.1, 1.0 and 3.7 mm/s;
(2) experiments at constant effective normal stress (s
eff
100 MPa) and sliding velocity (V 1.0 mm/s), and at
a constant temperature of 100, 300, 500 and 700

C;
(3) normal stress-stepping experiments (s
eff
20, 40, 60, 80 and
100 MPa) at constant sliding velocity (V 1.0 mm/s) and at
xed temperature (T 20, 150, 225, 300, 400, 500, 600 and
700

C); and
(4) velocity-stepping experiments (V 0.03, 0.1, 0.5, 1, 3.7 mm/s) at
constant effective normal stress (s
eff
100 MPa) and at xed
temperature (T 20, 150, 225, 300, 400, 500, 600, and 700

C).
For control purposes, one additional compaction experiment
(no shear displacement) was performed at T 700

C and
s
eff
100 MPa. The main purpose of this experiment was to seek
evidence for any chemical breakdown or partial melting of
muscovite, or for possible healing mechanisms that might other-
wise be destroyed during shear. To maximize the chance of
detecting such effects, the duration of the experiment was 166 h
(about 1 week), which is 520 times longer than the other exper-
iments performed.
Note that the normal stress-stepping experiments performed at
V 1.0 mm/s (type 3 above) were carried out to determine the
coefcient of friction of the samples at different temperatures.
Prior to the rst step in stress, the samples were sheared at an
effective normal stress of 20 MPa until a shear strain of 1316
(displacement of 810 mm) was reached with respect to the initial
gouge thickness of 0.62 mm. The effective normal stress was then
stepped up, with a step size of 20 MPa, to 100 MPa and back down
to 20 MPa (Niemeijer and Spiers, 2007; Niemeijer et al., 2008). A
shear strain of w1.6 (displacement of 1 mm) was typically suf-
cient to reach steady state shear stress levels in an individual stress
step. In the velocity-stepping experiments (type 4 above), the
gouges were sheared prior to the stepping sequence at an effective
normal stress of 100 MPa and at a constant sliding velocity of
1.0 mm/s. After a shear strain of w16 (displacement of 10 mm) was
reached, the velocity was typically stepped to 3.7, 0.1, 1.0, 0.03, 1.0
and 0.5 mm/s, respectively. In most steps, a shear strain of w1.6
(displacement of 1 mm) was again sufcient to reach steady state
shear stress levels.
After each shear experiment, the rotary drive and furnace were
turned off, resulting in an immediate decrease in temperature and
pore uid pressure. The system cooled to room temperature
within 20 min, after which the axial load was removed and the
water cooling was switched off. The pistons and sample material
were then taken out and left to dry in an oven at 60

C for at least
45 min. All experiments and corresponding conditions are listed
in Table 1.
Fig. 1. Hydrothermal rotary shear apparatus used in the present study. (a) The rotary shear drive system and pressure vessel mounted inside the Instron loading frame. (b) The two
internal pistons assembled with the pressure-compensated upper sealing head. The insulating mullite tubes sheathed with stainless steel and the furnace element are normally
located inside the vessel prior to assembly. (c) The two internal pistons with sample conning rings. The internal pistons are made of Rene 41 and have a core of mullite to reduce
heat loss along the pistons. The conning rings were made of mild hardened stainless steel type 1.4122, suitable for use even at temperatures up to 700

C. (d) Schematic cross
section of the piston assembly showing the position of the fault gouge sample.
2.4. Data acquisition and processing
All temperature, uid pressure, force, torque and rotational
displacement signals were logged as a function of time, using a 16
bit A/D converter connected to a computer. Shear stresses and
friction coefcients were calculated from the externally measured
torque and axial force gauge signals using the method described
below to correct for seal friction (Niemeijer et al., 2008). Slip
displacements and rates were calculated directly from the
measured rotary displacements making no correction for apparatus
twist, as this effect was small compared to the displacements
imposed.
Shear strains were calculated by dividing the measured rotary
displacement by the initial gouge thickness, taken to be 0.62 mm
in all tests. This was based on compaction experiments performed
at room temperature and 100 MPa effective normal stress,
showing that muscovite powder compacts to a porosity of w27%
under these conditions. This value agrees well with the 2630%
initial porosity measured by Mariani et al. (2006) for muscovite
gouges of the same grain size, cold-pressed in a die at room
temperature. It is noted that subsequent hot-pressing of these
samples by Mariani et al. (2006), for 30 h at 500700

C, reduced
the porosity to 18%. In our shear experiments, however, the
samples were pre-compacted under the deformation conditions
for only w45 min, resulting in a porosity at the beginning of each
experiment that is most likely still rather close to the initial value,
even at elevated temperatures. We, therefore, take 27% as the
initial porosity of all our samples, implying a value for the sample
thickness of 0.62 mm. During shear, porosity is further reduced,
resulting in a decrease in sample thickness which might differ
depending on temperature. Because of the uncertainties involved
in precise determination of this thickness, we use the initial
thickness for shear strain estimates. All reported shear strains,
hence, are minimum values.
Typical results of a normal stress-stepping experiment showed
a linear relation between the raw (uncorrected) shear stress and
the effective normal stress data (Fig. 2), described as:
s
m
m
m
$s
m
C
m
s
s
s
f
(1)
where s
m
is the measured shear stress, m
m
the slope of the shear
stress-normal stress graph, s
m
the measured normal stress and C
m
represents a constant. Since the shear stress is measured externally
in our experimental set-up, the measured values s
m
are not only
dependent on the shear and cohesive stresses supported by the
simulated fault gouge s
s
, but also upon the frictional stress
Table 1
Sliding experiments at constant conditions (non-stepping), s
eff
100 MPa
Experiment Temperature
T (
C)
Sliding
velocity
V (mm/s)
Final
shear
strain g
Apparent
compaction
(mm)
Final steady
state friction
coefcient m
MUS17 500 0.1 79.2 0.20 0.61
MUS18 500 3.7 63.7 0.18 0.68
MUS21 500 1 103.3 0.19 0.63
MUS04 500 1 ND ND ND
MUS05 300 1 36.9 0.12 0.70
MUS07 100 1 142.0 0.22 0.630.70
MUS11 500 1 133.4 0.19 0.89
MUS36 700 1 7.3 0.16 0.37
Normal stress-stepping experiments, V 1.0 mm/s, s
eff
20, 40, 60, 80, 100 MPa
T (
C)
Final
shear
strain g
Final apparent
compaction
(mm)
Steady state
friction
coefcient m
Notes
MUS01 20 29.1 0.08 0.43
MUS02 150 25.0 0.08 0.43
MUS03 500 28.6 0.15 0.51 Stick-slip
MUS06 400 28.7 0.08 0.60
MUS08 300 ND ND 0.54
MUS09 600 27.4 0.06 0.74
MUS12 500 27.8 0.06 0.76 Stick-slip
MUS16 100 30.3 0.22 0.38
MUS22 700 31.7 0.14 0.55 Initial dilatation
MUS28 225 22.6 0.16 0.37
MUS31 300 29.0 0.20 0.38
Velocity-stepping experiments, s
eff
100 MPa, V 0.03, 0.1, 0.5, 1, 3.7 mm/s
T (
C)
Final
shear
strain g
Final apparent
compaction
(mm)
Final steady
state friction
coefcient m
Notes
MUS13 500 32.6 0.12 0.90 Stick-slip, initial
dilatation
MUS14 300 37.0 0.14 ND
MUS19 700 31.0 0.26 0.43
MUS20 20 37.7 0.12 0.38
MUS26 150 42.2 0.11 0.46
MUS30 500 44.8 0.14 0.62 Stick-slip, initial
dilatation
MUS35 400 40.2 0.16 0.54
Compaction experiment, s
eff
100 MPa, P
f
100 MPa
MUS23 700 0 0.11
0
20
40
60
80
100
120
0 2 4 6 8 10 12 14 16 18
Shear displacement (mm)
S
h
e
a
r
-

o
r

n
o
r
m
a
l

s
t
r
e
s
s

(
M
P
a
)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 5 10 15 20 25
Shear strain
C
o
m
p
a
c
t
i
o
n

d
u
r
i
n
g

s
h
e
a
r

(
m
m
)
Shear stress
Normal stress
Compaction
= 0.38 + 15.713
R
2
= 0.9993
= 0.38 +16.636
R
2
= 0.9985
0
10
20
30
40
50
60
0 20 40 60 80 100 120
Normal stress (MPa)
S
t
e
a
d
y

s
t
a
t
e

s
h
e
a
r

s
t
r
e
s
s

(
M
P
a
)
Step up Step down
Average = 0.38
a
b
Fig. 2. Typical example of a normal stress-stepping experiment (MUS31), T 300

C,
P
f
100 MPa, s
eff
20100 MPa, V 1.0 mm/s and g 29. (a) Shear stress, normal
stress and apparent compaction as function of shear displacement and shear strain.
Note that the apparent compaction is not corrected for apparatus distortion. (b)
Correlated shear stress vs. normal stress diagram. The slope of the linear best t
through the data points represents a measure for the mean steady state friction
coefcient of the sample. The intercept of the linear best t at zero effective normal
stress is the effect of the friction of the O-ring seals and the conning rings (Niemeijer
et al. (2008), see Section 2.4). These intercept values were used to correct the
measured shear stresses obtained from velocity-stepping experiments.
contribution s
f
exerted by the O-ring seals on the pressure-
compensated piston and by the conning rings surrounding the
sample (Fig. 1c and d). Friction calibrations performed using an
internal normal force gauge have shown that at measured normal
stresses of 100 MPa the contribution of the O-ring friction to the
normal stress is constant and negligible compared to the applied
normal stress s
m
, so that the normal stress on the sample s
s
s
m
.
Therefore, the measured friction coefcient (m
m
in Eq. (1) repre-
sents true sample behaviour m
s
s
s
=s
s
provided the O-ring
contribution to the measured torque is independent of normal
stress. We performed the following calibration experiments to
allow meaningful processing of the data obtained.
In this procedure, two annular PEEK (PolyEtherEtherKetone)
rings of w0.3 mm in thickness were inserted into the apparatus in
the place of the usual fault gouge samples. Friction experiments
were performed using these PEEK samples (i) with a stainless
steel outer/inner conning ring, (ii) with a PEEK outer/inner
conning ring, and (iii) without an outer conning ring. The
frictional behaviour of the contact between the two solid PEEK
samples was measured in all of these cases for effective normal
stresses in the range 20100 MPa. The data obtained show well-
dened linear relations between shear stress and normal stress,
indicating PEEK-on-PEEK friction coefcients m
s
in the range 0.03
0.06. No systematic change in m
s
was seen in relation to the type of
inner/outer conning ring used (steel, PEEK or no ring), implying
that the frictional forces generated by the conning rings are
minor. The m
s
values 0.030.06 are low compared to the value of
0.1 (peek on peek) quoted by the manufacturer (see Niemeijer
et al., 2008), but t well with the low end of the range of 0.050.5
given in the literature (Burris and Sawyer, 2006a, 2006b; Theiler
and Gradt, 2008). This agreement implies that the frictional forces
caused by the seals and conning rings are unlikely to be signif-
icantly inuenced by the normal stress. We, therefore, conclude
that linear ts to the shear stress vs. normal stress data measured
in our experiments on muscovite (Fig. 2) yield true friction coef-
cients m
s
for the fault gouge. This is consistent with previous
conclusions based on experiments in the same apparatus by
(Niemeijer et al., 2008).
From the data of Niemeijer et al. (2008), the cohesion C
s
of our
muscovite samples can be taken as near zero (Niemeijer et al., 2008,
who measured 0.16 MPa for muscovite gouge), meaning that C
m
in
Eq. (1) is made up almost entirely of O-ring friction C
f
. Values of C
m
have been determined for every normal stress-stepping test per-
formed. Although they show some variability (1323 MPa), no
systematic relationship with temperature was apparent. We,
therefore, used the average value obtained for C
m
(17 MPa) to
correct the measured shear stress s
m
for O-ring and (minor)
conning ring friction. Dividing the result by the applied effective
normal stress s
m
(cf. Eq. (1)) then yields the internal friction coef-
cient of the sample:
m
s

s
m
C
m
s
m
(2)
Curves of corrected shear stress (s
m
C
m
) vs. displacement can
thus easily be recast into m
s
vs. shear strain g curves for our samples.
Compaction and dilatation measurements were determined
from the position of the Instron loading ram. Since the heated
length of the loading pistons is very limited, we can assume that
temperature does not inuence elastic apparatus distortion. Axial
machine distortion, therefore, depends only on normal stress. We
did not correct for this. However, we compare measured compac-
tion/dilatation data for different experiments only fromthe onset of
shear deformation and at xed effective normal stress. Any differ-
ences between experiments, therefore, reect sample behaviour
and are unrelated to apparatus behaviour.
2.5. Sample handling and microstructural methods
After the experiments, most samples broke into arc-shaped
fragments during removal of the sheared gouge from the set-up.
The fragments varied in size between 3 and 14 mmin arc length, by
3 mm wide and w0.5 mm thick. All were dried for w45 min at
60

C after extraction. The dried samples were then vacuum
impregnated with epoxy resin for about 15 min and left to dry for
a period of 24 h. The impregnated fragments were nally sectioned
normal to the shear plane and parallel to the shear direction, and
polished. Mineral content and microstructure were investigated
using a Scanning Electron Microscope (SEM) equipped with Energy
Dispersive X-ray analysis system (EDX). Grain sizes and mineral
fractions in the deformed samples were obtained from point
counting analyses.
3. Mechanical and microstructural results
We performed 8 sliding experiments at xed conditions, 14
normal stress-stepping experiments, 9 velocity-stepping experi-
ments and 1 compaction experiment. The experiments and corre-
sponding conditions are represented in Table 1.
3.1. Effect of shear strain
Fig. 3a and b shows the evolution of friction coefcient with
shear strain for a number of samples deformed at 500

C and at
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 5 10 15 20 25 30
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0 5 10 15 20 25 30 35 40 45
Shear strain
(

y
t
i
c
o
l
e
v

g
n
i
d
i
l
S
)
s
/
m
Shear stress
Sliding velocity
V = 1.0 m/s 1.0 m/s 1.0 m/s
0.5 m/s
3.7 m/s
0.1 m/s0.03 m/s
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 10 20 30 40 50 60 70
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
0 20 40 60 80 100
Shear strain
MUS17 V = 0.1 um/s
MUS18 V = 3.7 um/s
MUS21 V = 1.0 um/s
MUS30 V-stepping experiment
V = 0.1 m/s
V = 3.7 m/s V = 1.0 m/s
Unstable
stick-slip behaviour
Stable sliding
behaviour
a
b
Fig. 3. (a) Friction coefcient vs. shear displacement (i.e. shear strain) obtained from
three sliding experiments under constant conditions (non-stepping) at T 500

C,
P
f
100 MPa, s
eff
100 MPa and one velocity-stepping experiment under the same
pressure and temperature conditions for comparison. The samples deformed at 0.1 and
1.0 mm/s showunstable stick-slip behaviour until sliding stabilizes at g 16 and g 24,
respectively. The velocity-stepping experiment also shows unstable sliding in most
steps, while the sample deformed at 3.7 mm/s exhibited stable sliding throughout the
experiment. (b) Friction coefcient data from velocity-stepping experiment MUS30
compared with the steps in sliding velocity.
100 MPa effective normal stress. In the early stages of the experi-
ments (g up to 3), the friction coefcient strongly increased with
shear strain, then gradually approached a steady state value of
w0.6 at g of 3540. The samples compacted more or less contin-
uously during shear (Fig. 4a) with little effect of velocity (Fig. 4b).
Stick-slip behaviour was observed in some experiments at 400 and
500

C, but no systematic relationship with strain or velocity was
found. Stick-slip was not observed in tests at temperatures outside
this temperature range. Fig. 5 shows values of friction coefcient vs.
shear strain at g >7, as obtained from a representative set of our
velocity-stepping experiments. Note that data obtained at constant
temperature (i.e. in a given stepping test) show a minor increase in
friction coefcient with strain within the range g 1035, and little
effect of velocity.
Fig. 6ad showthe microstructures developed at shear strains of
w29, 40, 64 and 103, in samples deformed at 400 and 500

C,
100 MPa uid pressure and a sliding velocity of 1.0 mm/s. Of these
samples, three were deformed under 100 MPa normal stress and
one at 20100 MPa in a normal stress-stepping experiment
(MUS32). The microstructures developed at g 29 and g 40 at
400

C (Fig. 6a, b and e) are rather similar. They show a muscovite
foliation oblique to the shear direction and the presence of thin
(w2 mm) anastomosing shear bands, dominantly in Y-shear orien-
tation. Most muscovite grains have sizes of 520 mm, similar to the
starting material and some grains show occasional folding or kink-
ing. The thin bands contain grains of 12 mm, and are sometimes
bounded by thicker (310 mm) zones of 14 mm size grains. Signi-
cant porosity is visible at the tips of the muscovite grains, which are
often jagged and broken in appearance. About 5% quartz porphyr-
oclasts (520 mm) are present, often showing extensional fractures.
The effect of shear strain can best be assessed using Fig. 6c and
d for g 64 and g 103, both at 500

C and similar normal stress
and uid pressure. Shear bands are wider now than seen at lower
shear strain at 400

C, and their width increases with strain, up to
20 mm in the sample deformed to g 103. The bands anastomose
around dense lenses of coarser muscovite (Fig. 6f), showgrain sizes
<1 mm (Fig. 6g), and occupy roughly 2035% of the gouge volume.
The intervening lenses do not show a clear foliation, but do show
local folding and kinking in the muscovite grains and have a grain
size of 220 mm. The lenses also contain occasional quartz por-
phyroclasts with a grain size of 210 mm.
3.2. Effect of normal stress
Data on the measured shear stress, the effective normal stress
and apparent compaction (uncorrected vertical displacement) vs.
rotary displacement, obtained from a typical normal stress-step-
ping experiment (MUS31), are illustrated in Fig. 2a (see also Table
1). The corresponding shear stress vs. normal stress diagram, from
which the mean steady state coefcient of friction is calculated, is
shown in Fig. 2b. In individual stress steps, the shear stress initially
increased rapidly, but reached steady state within w1 mm of
displacement (Fig. 2a). Also, the apparent compaction increased
instantly when normal stress was stepped up, while an instant
expansion occurred when normal stress was stepped down. Rough
stiffness calibrations have shown that these vertical displacements
were largely due to elastic distortion of the apparatus, though net
compaction of the gouge was recorded at the end of the experi-
ments (Table 1). We do not take the apparent compaction obtained
from the normal stress-stepping experiments into account in the
interpretation of our results, since we cannot correct accurately for
apparatus stiffness. Nonetheless, friction coefcients obtained in
the normal stress-stepping tests ranged from 0.37 to 0.60
depending on temperature (see Section 3.4).
Fig. 6a and b show microstructures of samples that reached
different shear strains, at T 400

C, but that also differed in
effective normal stress. The micrograph shown in Fig. 6a is obtained
from an experiment in which the effective normal stress was
stepped from 20 MPa to 100 MPa and back (in 20 MPa-steps).
Fig. 6b represents a sample deformed at constant normal stress of
100 MPa. The microstructures of the two samples are rather similar
(see also Section 3.1), hence, there seems to be no signicant effect
a
b
-0.05
0
0.05
0.1
0.15
0.2
0.25
0 10 20 30 40 50 60 70
)
m
m
(

r
a
e
h
s

g
n
i
r
u
d

n
o
i
t
c
a
p
m
o
C
MUS17 V = 0.1 um/s
MUS18 V = 3.7 um/s
MUS21 V = 1.0 um/s
MUS30 V-stepping experiment
V = 0.1 m/s
V = 3.7 m/s
V = 1.0 m/s
V- stepping
-0.05
0
0.05
0.1
0.15
0.2
0.25
0 5 10 15 20 25 30
)
m
m
(

r
a
e
h
s

g
n
i
r
u
d

n
o
i
t
c
a
p
m
o
C
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
(

y
t
i
c
o
l
e
v

g
n
i
d
i
l
S
)
s
/
m
Compaction
Sliding velocity
V = 1.0 m/s 1.0 m/s 1.0 m/s
0.5 m/s
3.7 m/s
0.1 m/s 0.03 m/s
Fig. 4. (a) Compaction vs. shear displacement diagram obtained from the three
constant velocity experiments and a single velocity-stepping experiment as in Fig. 3.
Only the compaction recorded from the onset of shearing is indicated here, at constant
effective normal stress of 100 MPa. The irregularities in the data signal are due to a low
signal to noise ratio and are not related to stick-slip behaviour of the sample. (b)
Compaction data from velocity-stepping experiment MUS30 compared with the steps
in sliding velocity.
0
1 . 0
2 . 0
3 . 0
4 . 0
5 . 0
6 . 0
7 . 0
8 . 0
0 3 5 2 0 2 5 1 0 1 5 0
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
5 4 0 4 5 3 0 3 5 2 0 2 5 1 0 1 5 0
Shear strain
MUS35, T = 400 C MUS30, T = 500 C
MUS19, T = 700 C MUS20, T = 20 C
20 C
700 C
400 C
500 C
Fig. 5. Diagram illustrating the effect of shear strain on the coefcient of friction
obtained from velocity-stepping experiments at different temperatures. At 700

C,
steady state shear stress levels were not reached within a single velocity step. The
reported values for this temperature are only represented as an indication and should
be considered with care.
of normal stress on the development of microstructures, at least not
at 400

C.
3.3. Effect of sliding velocity
From Figs. 35, it has already been noted that sliding velocity
does not have an important effect on the mechanical behaviour of
the gouges. This is largely conrmed in Fig. 7, where the steady
state friction coefcients obtained in our velocity-stepping tests are
explicitly plotted against sliding velocity. Clearly, there is no
signicant effect of velocity on friction coefcient at xed
temperatures up to 500600

C, though m does increase with
temperature (see Section 3.4). At 700

C, the generally lower fric-
tion coefcient seems to peak at 1.0 mm/s, even though steady state
shear stress levels were not reached under these conditions. We did
not observe a systematic effect of sliding velocity on the compac-
tion behaviour of the gouges (Fig. 4a and b).
The microstructures typifying samples deformed at constant
sliding velocities ranging from0.1 to3.7 mm/s are illustratedinFig. 8,
for T 500

C and for shear strains of about 6379 (samples MUS17,
18 and 21, Table 1). The microstructures developed under these
conditions are again characterized by ne grained, elongate lenses
and oval clasts measuring typically 2050 by 10 mm (Fig. 8a and c)
consisting of often folded and kinked muscovites some 220 mm in
length(Fig. 8bandd). Inthe SEM, the lenses showa lowporosityand
are separated by a through-going anastomosing network of ultra-
ne grained (1 mm) bands, oriented parallel to and, though less
well developed, at roughly 2040
to the shear direction (Fig. 8a and

c). These ultra-ne grained bands not only contain muscovite and
(rare) quartz clasts ranging from1 to 10 mmin size, but also some
fragments which are agglomerated grains and contain both
muscovite and quartz (see detail image Fig. 6d).
Onthis basis, the mainmicrostructural features donot seemtobe
strongly affected by sliding velocity. However, the dense muscovite
lenses developed in the sample sheared at 0.1 mm/s showabundant
rounded folds and only a few chevron type folds (Fig. 8b), whereas
the sample sheared at high velocity (3.7 mm/s; Fig. 8d) shows more
prominent chevron type folds and more signs of brittle deformation
Fig. 6. SEM backscatter images showing the development of microstructure with increasing shear strain in samples deformed at 400 and 500

C, 100 MPa uid pressure and shear
strain as indicated. The shear direction is sinistral and roughly horizontal. Note the progressive development of ne grained bands. At low strain, the muscovite grains (m) form
a foliation oblique to the shear direction, while at high shear strain no foliation is observed and the bands anastomose around lenses of relatively less deformed material. Some
quartz porphyroclasts (q) are present in all samples. (a) Obtained from normal stress-stepping experiment MUS32, T 400

C, s
eff
20100 MPa and V 1.0 mm/s. (b) Obtained
from velocity-stepping experiment MUS35, T 400

C, s
eff
100 MPa and V 0.033.7 mm/s. (c) Obtained from non-stepping experiment MUS18, T 500

C, s
eff
100 MPa and
V 1.0 mm/s. (d) Obtained from non-stepping experiment MUS21, T 500

C, s
eff
100 MPa and V 1.0 mm/s. (e) Detail of Fig. (a) showing thin, ne grained bands. (f) Detail of
a lense from sample MUS21. (g) Detail of Fig. (d) showing a thick, ne grained band.
at the hinges. Material sheared at 1.0 mm/s under similar conditions
(MUS21) showed an intermediated microstructure.
3.4. Effect of temperature
Fig. 9a shows the effect of temperature on friction coefcient as
obtained at different shear strains from our velocity-stepping
experiments at constant temperature. These data illustrate the
signicant increase in friction coefcient that occur up to g 10
and the steady state attained at shear strains of 3040 (see Section
3.3). The steady state coefcient of friction increases from a mean
value of 0.37 at 20

C to 0.56 at 300

C, remaining around this value
up to 600

C. At 700

C, the coefcient of friction decreases again, to
a mean value of 0.38 in the velocity-stepping tests. Fig. 9b shows
the friction coefcient data obtained from the uncorrected normal
stress-stepping experiments. These m-values are not taken at
a specic shear strain, but over the entire stepping sequence and
thus incorporate an effect of shear strain. The absolute values
shown in Fig. 9b increase from 0.38 at low temperature up to 0.60
at higher temperature. Although, these average values differ
slightly from the friction coefcients obtained from the velocity-
stepping tests, the trend in both data sets is very similar except at
700

C. For comparison, the low shear strain data obtained at 400
700

C by Mariani et al. (2006) are plotted in Fig. 9a. Note the good
agreement with our low strain data, again except at 700

C.
As illustrated in Fig. 10, the amount of apparent vertical
compaction measured in the rst portion of our velocity-stepping
experiments, that is in the rst 10 mm displacement at 1.0 mm/s,
decreased slightly with increasing temperature up to 500

C. The
rate of compaction was highest in the early stages of shearing at
150300

C, but quickly decreased to a much lower rate at higher
shear strain. At 500 and 600

C (Figs. 4 and 10), limited initial
dilatation is followed by gradual compaction at a rate that
decreases with shear strain. The compaction behaviour at 700

C
differs from that at lower temperature; it reaches much higher
values and appears to continue at a higher rate.
0
0.2
0.4
0.6
0.8
0.01 0.1 1 10
Sliding velocity ( m/s)
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
MUS20, T = 20 C MUS35, T = 400 C
MUS30, T = 500 C MUS19, T = 700 C
Fig. 7. Diagram illustrating the changes in friction coefcient with sliding velocity at
different temperatures. At all temperatures but 700

C, steady state shear stress levels,
hence steady state friction coefcients, were reached within the amount of shear strain
(1.54) obtained in individual steps.
Fig. 8. SEM backscatter images showing the main structure developing in samples sheared at 500

C, s
eff
100 MPa, P
f
100 MPa and sliding velocities as indicated. The shear
direction is sinistral and roughly horizontal. (a) Microstructure obtained from non-stepping experiment MUS17 (g 79), characterized by dense, elongate lenses, which are
separated by a network of ultra-ne grained zones. (b) Detail of microstructure from sample MUS17, showing folding of the mica sheets inside one of the massive elongate lenses. (c)
Obtained from non-stepping experiment MUS18 (g 64), which is very similar to Fig. (a). (d) Detail of microstructure from sample MUS18, showing an example of more chevron
type folds that appear to be more prominent in samples sheared at 3.7 mm/s than at 0.1 mm/s.
The effect of temperature on the development of the micro-
structure is illustrated in Fig. 11. The samples shown in the gure
have shear strains from a relatively narrow range (g 2942),
except for the sample deformed at 500

C (g 64). Fig. 11a and
b show the microstructure of samples deformed at low tempera-
ture (MUS20, T 20

C; MUS26, T 150

C). These samples show
thin shear bands (thickness <5 mm) roughly parallel and at a low
angle to the shear direction, consisting of very ne material
(<5 mm) separating lenses of larger muscovite grains (520 mm).
Some of the lenses as well as some of the muscovite grains show
elongated sigmoidal shapes. Quartz porphyroclasts up to 50 mm
occur and often show extensional fractures.
Fig. 11c shows the microstructure of a sample sheared at 300

C
(MUS14). This micrograph displays dense looking lenses separated
by shear bands that crosscut the sample. These ne grained zones
vary in thickness from 5 to 50 mmand contain some large clasts (up
to 40 mm) embedded in very ne (<2 mm) material. Some of these
clasts appear to be clusters of agglomerated grains. At 500

C
(Fig. 11d), the microstructure is increasingly characterized by
elongate lenses (2050 by 10 mm) of ne folded and kinked grains
(510 mm). The lenses are separated by a through-going anasto-
mosing network of ne grained zones (grain size <1 mm), oriented
at zero to roughly 30
to the shear direction.

In contrast to the samples deformed at lower temperature, the
microstructure developed at 600

C (Fig. 11e) displays no major
zones with a reduced grain size. The structure is dominated by an
oblique foliation dened by the muscovite grains oriented at an
angle of roughly 20
to the sample boundary and shear direction.

Quartz porphyroclasts are somewhat reduced in size (up to 5 mm)
and some show of truncations suggestive of diffusive mass transfer
processes such as pressure solution (Fig. 12).
The different shades in the SEM image of Fig. 13a and b at g 7
correspond to a different atomic number, indicative of different
mineral phases or alteration. Quartz appears dark and muscovite
light grey. The overall grain size of the material is small (<5 mm),
but the larger mica grains dene a wavy foliation parallel to the
shear plane. In contrast to the foliation and banding developed at
lower temperature, the foliation seen at 700

C often shows sharp,
wavy surfaces suggestive of slip surfaces. The quartz grains are
fractured but form elongated, sigmoidal lenses with a stair-step-
ping character and long, drawn-out tails not seen at 600

C or
below. A third phase is visible lling fractures plus voids in many
quartz clasts (Fig. 13b and c). This other phase is present in amounts
too small to be determined given the resolution of our EDX system.
At 700

C, clusters of agglomerated grains occur (Fig. 11f). The
clusters are about 2050 mm in size, while the grains in these
agglomerates range from 1 to 10 mm in diameter.
Fig. 13d shows a micrograph obtained from compaction exper-
iment MUS23 (T 700

C) The mica grains are aligned normal to
the compaction direction, some displaying intra-crystalline
porosity. The contacts between the quartz and muscovite grains
shows truncations and indentations and a number of muscovite
grains appear corroded.
4. Discussion
4.1. Deformation mechanisms and gouge evolution
The trends in frictional and volumetric behaviour seen with
changing conditions, notably strain and temperature, can be
explained in terms of the microstructures developed in the sheared
gouges. A characteristic element of the microstructure is the
generally reduced size of the muscovite grains compared to the
starting grain size, the angular shapes of many of the ne grains,
and the progressive development with strain of anastomosing
(ultra) ne grained bands parallel to the shear plane at least up to
500

C. From these features, from the jagged ends of the coarser
muscovites, and from the clearly fractured quartz clasts, we
conclude that cataclasis is the grain size reducing mechanism at
20500

C. The ne grained bands suggest a tendency for locali-
zation. Cataclasis generally produces a broadening grain size
a
b
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 100 200 300 400 500 600 700
Temperature (C)
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
Shear strain ~ 0.8 Shear strain ~ 17
Shear strain ~ 30 Shear strain ~ 40
Mariani et al. (2006)
Obtained from normal stress stepping
experiments in steps at 1.0 m/s
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 100 200 300 400 500 600 700
Temperature (C)
F
r
i
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
Fig. 9. Diagrams showing the effect of temperature on friction coefcient. (a) Friction
coefcients obtained at different shear displacements from velocity-stepping experi-
ments at an applied normal stress of 100 MPa and a uid pressure of 100 MPa. Data are
corrected for ring- and seal friction as described in Section 2.5. (b) The friction coef-
cients obtained from normal stress-stepping experiments under s
eff
20100 MPa,
P
f
100 MPa and at V 1.0 mm/s. Note that these friction coefcients are mean values,
obtained at different shear strains and from uncorrected shear- and normal stress data.
They are used here to illustrate the trends rather than the absolute values.
5 0 . 0 -
0
5 0 . 0
1 . 0
5 1 . 0
2 . 0
5 2 . 0
2 1 0 1 8 6 4 2 0
C
o
m
p
a
c
t
i
o
n

d
u
r
i
n
g

s
h
e
a
r

(
m
m
)
C 0 0 3 = T 4 1 S U M
C 0 5 1 = T 6 2 S U M
C 0 0 6 = T 5 2 S U M
C 0 0 7 = T 9 1 S U M
T = 150 C
T = 700 C
T = 300 C
T = 600 C
Fig. 10. Diagram showing the apparent compaction vs. shear displacement in the rst
portion of our velocity-stepping experiments, i.e., at V 1.0 mm/s. Note the effect of
temperature. At 600

C the sample shows dilatation in the rst 2 mm displacement,
while the other samples continuously compact. The sample deformed at 700

C shows
most compaction.
distribution (Rawling and Goodwin, 2003; Sammis et al., 1987),
which results in a decrease in porosity, an increase in density and
hence on-going compaction (Saileswaran and Panchanathan, 1973),
in agreement with our ndings (Fig. 4). Such compaction will in
general increase friction coefcient due to an increasing dilatation
angle (Niemeijer and Spiers, 2006).
In an attempt to explain the strain hardening behaviour seen
in our experiments, we infer that with increasing strain, both
pervasive and localized cataclasis, and related compaction, rst
result in hardening, which leads to widening of the ner anas-
tomosing bands (cf. Scruggs and Tullis, 1998), at the cost of the
low strain lenses. Steady state is then approached as the micro-
structure becomes dominated by the development of a contin-
uous network of ne grained bands and oblique foliation. The
increase in width of these ne grained cataclastic bands with
temperature appears to be associated with an increase in friction
coefcient (at high strain), perhaps implying that compaction
and hence hardening are promoted by thermally activated
Fig. 11. SEM backscatter images showing the effect of temperature on the microstructure at 100 MPa uid pressure. Most fault gouges have fragmented after retrieval from the
shear apparatus. The shear direction is sinistral and roughly horizontal and temperature is as indicated. (a) Microstructure obtained from velocity-stepping experiment MUS20,
s
eff
100 MPa, V 0.033.7 mm/s and g 38 showing large quartz porphyroclasts (q) and fractured grains. The mica grains (m) dene an oblique foliation and the overall structure
is cross cut by ne grained shear bands. (b) Structure obtained from velocity-stepping experiment MUS26, at conditions similar to (a), only at a different temperature. (c) Structure
obtained from velocity-stepping experiment MUS14, s
eff
100 MPa, V 0.033.7 mm/s and g 37, showing denser lenses, surrounded by more ne grained zones, widening with
increasing shear strain. (d) Micrograph from the sample deformed in non-stepping experiment MUS18, at s
eff
100 MPa, V 3.7 mm/s and g 64. This structure again shows dense
lenses, surrounded by a network of ne grained bands. (e) Structure obtained from normal stress-stepping experiment MUS24, sheared at s
eff
20100 MPa, V 1.0 mm/s and
g 29, showing no major grain size reduction compared to the starting material. The muscovite grains dene a strong, foliation oblique to the shear direction. (f) Fragment obtained
from sample MUS19, s
eff
100 MPa, V 0.033.7 mm/s and g 31 showing agglomerated muscovite grains.
diffusive or plastic ow processes in the bands. The coarse lenses
show folded and kinked muscovite grains indicative of active
crystal plastic mechanisms. Observations on these mechanisms,
however, are not diagnostic enough to draw conclusions
regarding the contribution of plastic deformation to the bulk
mechanical behaviour.
Upon reaching 600

C, the ne grained bands become a less
prominent microstructural feature than at 20500

C (Fig. 11e),
though the overall microstructure remains that of a gouge with
a grain size reduced compared to that of the starting material. The
strength of the gouge is still directly dependent on normal stress (as
in Fig. 2), demonstrating that there was no major change in
deformation mechanismgoing fromlowto high temperature at the
strain rates applied in this study. Moreover, the friction coefcient
at 600

C is not dissimilar from that at 400 and 500

C, suggesting
that the difference in microstructure (homogeneous matrix vs.
broad bands, respectively see Figs. 11e and 6) does not result in
a difference in bulk mechanical behaviour.
At 700

C, the chemical variations visible on the SEM images
(Fig. 13ac) together with the corroded appearance of muscovite
grains (Fig. 13d) suggest that under these conditions muscovite
breakdown reactions may have played a role. These then would
have changed the gouge such that a decrease in friction coefcient
resulted, from0.56 at 600

C to 0.38 at 700

C. The sigmoidal quartz
clasts are mostly fractured and a number of these cracks appear to
have been lled by a different phase, which also forms long (50
100 mm) tails (Fig. 13ac) and might be indicative of partial melting
of the gouges.
With respect to the <10% quartz present in our samples, we note
the following. At all observed scales, the quartz grains occurred as
isolated clasts and did not form a connected network. Niemeijer
and Spiers (2005) showed that the presence of up to 20% of halite in
Fig. 12. SEM backscatter image obtained from normal stress-stepping experiment
MUS24 (T 600

C, g 29). The quartz porphyroclasts (q) in the micrograph is fractured
and shows truncations (t), suggestive of the operationof mass transfer processes. The rest
of the structure consists of ne muscovite (m) grains. No marked reactionproducts due to
the breakdown of muscovite quartz are visible at the muscovitequartz interfaces.
Fig. 13. SEM backscatter images showing the microstructure of muscovite gouge at 700

C. The shear direction is roughly horizontal and the shear sense is sinistral. Figures ac are
obtained from non-stepping experiment MUS36 sheared at T 700

C, V 1.0 mm/s and g 7. (a) Micrograph showing a dense, foliated microstructure with muscovite (light grey, m),
quartz (dark, q) and a fewimpurities (white). Some grains are fractured and the overall grain size appears smaller than the starting grain size, though some large muscovites remain. (b)
Detail from (a) showing the altered appearance of fractured quartz grains. A possible third phase lling the cracks is indicated with an arrow. (c) Detail from (b). Note the third phase
present between the fractured quartz grains, indicated with arrows, possibly due to partial melting of the sample. (d) Micrograph from compaction experiment MUS23. The micro-
structure has not beensheared, andnofoliationor shear bands have developed. Some grains appear corroded, possibly due tochemical alterationandthe graincontacts are clearly visible.
a muscovite fault gouge results in a slightly higher shear stress,
without changing the mechanical behaviour and velocity-depen-
dence of the mixture. Assuming that our isolated quartz grains do
not behave much different in the muscovite gouge than the low% of
halite of Niemeijer and Spiers, we infer that the presence of quartz
in our samples did not strongly inuence the measured friction
coefcients.
4.2. Stability of muscovite: decomposition vs. melting
Dehydroxylation reactions of dry muscovite can be expected to
start around 400

C. These reactions, however, are substantially
inhibited by applying a pore uid pressure in the order of 100 MPa
(Mariani et al., 2006). Since we have applied 100 MPa uid pressure
in all experiments, it is unlikely that muscovite dehydroxylation
played an important role in our experiments.
At around 600

C and uid pressures of 100 MPa, decomposition
of muscovite plus quartz may take place following the reaction
(Chatterjee and Johannes, 1974): Muscovite Quartz /K-feld-
spar Andalusite Water. Since our samples contain <10% of
quartz, we expected to see reaction products around the quartz
grains in samples deformed at 600

C. However, no evidence for the
decomposition or reaction of muscovite and quartz was found (see
Fig. 12). Around 700

C and water pressures of 100 MPa, decom-
position of muscovite occurs following the reaction (Chatterjee and
Johannes, 1974): Muscovite /K-feldspar CorundumWater.
Our compaction experiment at 700

C (MUS23) showed altered/
corroded muscovite grains, indicating that chemical alteration of
the muscovite indeed occurred (Fig. 13d). Although the different
shades on the SEM images (Fig. 13ac) from the sample sheared at
700

C suggest the presence of new mineral phases, the limited
amount of reaction product did not allow meaningful EDX analysis,
given the resolution of our set-up. Mariani et al. (2006) performed
detailed TEM analysis on muscovite samples deformed at this high
temperature and reported various reaction products following the
reaction: Muscovite /K-feldspar Biotite Mullite Water. The
exact decomposition reaction is strongly inuenced by the chem-
ical composition of the starting material. For example, Na-for-K
substitution reduces the temperature stability of white micas (Deer
et al., 1962). However, K-feldspar is produced in all decomposition
reactions of muscovite at this high temperature. It is, therefore, not
unlikely that K-feldspar was formed.
The presence of quartz in combination with muscovite and
water lowers the melting temperature of the aggregate to w700

C
at 5 kbar pressure (Huang and Wyllie, 1974), suggesting an even
lower melting temperature under the conditions reached in this
study. It is thus possible that the phase in the cracks (Fig. 13ac) is
partial melt lubricating the grain contacts. Previous high-velocity
friction experiments on fault gouge have shown that the presence
of a small amount of melt reduces the friction of a fault surface
signicantly (Di Toro et al., 2006).
4.3. Comparison with previous work
The coefcient of friction for muscovite obtained in this study
was found to be 0.37 at room temperature, which is in reasonably
good agreement with values reported from previous room
temperature experiments on muscovite fault gouge: 0.38 (Scruggs
and Tullis, 1998), 0.42 (Moore and Lockner, 2004; Morrow et al.,
2000) and 0.4 (Mares and Kronenberg, 1993). Scruggs and Tullis
(1998) sheared wetted muscovite gouge without uid pressure
under 25 MPa effective normal stress up to shear displacements up
to 150200 mm. Moore and Lockner (2004) and Morrow et al.
(2000) performed experiments on wet muscovite fault gouge at
low shear strains and Mares and Kronenberg (1993) deformed dry,
single crystals of muscovite under different pressures. The differ-
ence between the values of the friction coefcients resulting from
the various studies is probably due to the different experimental
set-ups, the presence or absence of pore uid, use of different
starting materials, and variations in shear strains reached. Despite
these differences, all reported friction coefcients for wet musco-
vite at room temperature fall in the range m 0.39 0.03.
Strain hardening behaviour is often observed in experiments on
muscovite and other phyllosilicates (Logan and Rauenzahn, 1987;
Mariani et al., 2006; Moore and Lockner, 2004; Moore and Rymer,
2007; Morrow et al., 2000; Rutter and Maddock, 1992; Rutter et al.,
1986). We attribute the strain hardening behaviour observed at
g <10 in our experiments to pervasive cataclasis coupled with the
development of cataclastic shear bands (made of ne grained
material) and an oblique foliation, cf. Scruggs and Tullis (1998). We
suggest that the hardening is caused by grain size reduction and
associated compaction, leading to a change in dilatation angle and
an increase in friction coefcient. These bands are at an angle of
roughly 2040
to the shear direction, regardless of shear strain

and normal stress, suggesting that they are dynamically stable. The
bands widen at the cost of the lenses, until a steady state shear
stress is reached.
We turn now to the effect of temperature on m. Moore and
Lockner (2004) observed a positive correlation between the value
for the coefcient of friction of wet sheet silicate gouges at room
temperature and the electrostatic separation energy of the
minerals. Sheets silicates with a relatively strong bonding between
the individual sheets are likely to have a higher coefcient of fric-
tion than those with a weak bonding between the sheets. These
authors infer that water forms thin, structured lms between the
sheet surfaces. The polar water molecules are bonded to the plate
surfaces in proportion to the minerals surface energy. The friction
coefcient of the sample reects either the strength of the bonding
between the water and the different mineral surfaces or the
stresses required to shear through these water lms themselves.
Consequently, the coefcient of friction of a given sheet structure
mineral depends on the properties of the mineral surfaces and/or
the properties of the uid phase and changes therein. An increase
in temperature could drive the adsorbed water off the mineral
surfaces, resulting in an increase in friction coefcient, as previ-
ously observed in serpentine gouges (Moore et al., 1997). Although
the friction coefcient of our samples increases with temperature,
the constant uid pressure of 100 MPa is likely to maintain the
presence of thin water lms between the sheet surfaces. We,
therefore, attribute the observed change in friction coefcient to
the development of the observed networks of shear bands rather
than to a change in the bonding between the pore water and the
mica surfaces.
Mariani et al. (2006), who performed direct shear (saw-cut)
experiments on simulated muscovite gouge at temperatures of
300700

C, reported friction coefcients that are in good agree-
ment with our absolute values for m at g <10 (see Fig. 9a). Since the
samples deformed by Mariani et al. (2006) have a very-low (<9%)
starting porosity it is unlikely that compaction during shear
attributed to the observed strain hardening in their experiments.
Mariani et al. inferred that misaligned mica grains can serve as
obstacles to sliding, causing the hardening. They suggest that as the
microstructure evolves with shear strain and more and more grains
align themselves, the hardening rate is reduced and a steady state
shear stress may be reached. Because of the low strains attained in
the experiments of Mariani et al., such steady state was never
reached in their tests, but is clearly approached in our experiments.
However, the microstructures of our samples do not show an
evolving alignment of muscovite grains parallel to the shear
direction. Rather, a dynamically stable microstructure consisting of
Y-shear bands of cataclastic material and an oblique foliation
developed. This suggests that at lowshear strain, in samples where
little syn-deformational compaction occurs (cf. Mariani et al.,
2006), misaligned grains may result in hardening of the gouge,
whereas at higher shear strains in gouges that densify during
deformation (this study), cataclasis and the formation of ne
grained bands can explain the observed strain hardening
behaviour.
Sliding velocity and thus shear strain rate did not signicantly
effect the friction coefcients under the explored conditions in
good agreements with Mariani et al. (2006) and Niemeijer and
Spiers (2005). However, a change from rate-independent to rate-
dependent behaviour was reported at shear strain rates of 10
6
s
1
by Mariani et al. (2006) obtained during both constant displace-
ment rate and relaxation tests, which is at slower rates than the
10
5
to 10
3
s
1
applied in the present study. This rate-dependent
behaviour was attributed to viscous glide of basal dislocations
becoming competitive at low strain rates, but was not seen in our
faster experiments. Stick-slip behaviour at temperatures of 400
500

C was observed by Mariani et al. (2006), and conrmed under
similar pressure and temperature conditions in our experiments.
Mares and Kronenberg (1993) performed shortening experi-
ments on single crystals of muscovite using a Heard-type triaxial
gas apparatus at temperatures of 20400

C. These authors suggest
that deformation of muscovite crystals may occur by dislocation
glide at 45
to (001), kink generation at 0
to (001) or fracture at 90
to (001). They observed evidence of dislocation glide in samples

deformed at 45
to (001) at temperatures of 20, 200 and 400

C.
Evidence for plastic deformation in muscovite single crystals at
room temperature and pressure has been previously reported
during basal plane slip (Meike, 1989). Our results, including those at
higher temperature also show some evidence of ductile deforma-
tion, notably the folding of muscovite grains, but always subordi-
nate compared to the role of cataclasis. This is most likely due to the
relatively high strain rates used in our experiments. Moreover, the
reported observations of plastic deformation in muscovite at room
temperature were obtained from experiments on single crystals
(Mares and Kronenberg, 1993; Meike, 1989), whereas our samples
consist of ne grained fault gouge. Mariani et al. (2006) observed
plastic deformation in muscovite fault gouge at elevated tempera-
tures and low strain rates. We, therefore, suggest that the defor-
mation of ne grained fault gouge of grain sizes used here was by
cataclasis and grain boundary sliding at high strain rates, while
plastic mechanisms may control deformation at low strain rates or
in coarser material, less likely to slide and fracture.
At 700

C, muscovite breakdown reactions result in a different
mechanical behaviour of the gouges and associated development of
modied microstructures. Mariani et al. (2006) observed at 700

C
a sudden weakening, similar to our results but also a change from
rate-independent to rate-dependent behaviour. However, they
observed the growth of K-feldspar grains as a result of the chemical
breakdown of the muscovites and did not observe evidence for
partial melting. We attribute this difference in chemical behaviour
at 700

C to the relatively high percentage of quartz impurities in
our samples (<10% compared to <1%). These quartz impurities
reduce the melting temperature of our samples to w700

C (Huang
and Wyllie, 1974), while the melting temperature in the samples of
Mariani et al. is well outside the range of their experimental
conditions.
4.4. Implications for natural fault zones
Our experiments indicate that muscovite fault gouges
strengthen with increasing temperature up to 600

C, showing an
increasing in friction coefcient from 0.37 to 0.56. In other words,
muscovite gouge becomes increasingly more resistant to sliding
with increasing temperature at the sliding velocities studied (strain
rates 10
5
to 10
3
s
1
). In order to obtain a rst-order assessment of
the importance of muscovite strength in crustal faults, we used the
data presented in Fig. 9a to construct a frictional strength prole for
the crust that allows comparison with Byerlees Rule (Fig. 14). For
the strain rates applied in our study, the shear strength of musco-
vite appeared independent of sliding velocity, hence we did not
include a rate effect on m beyond the range of sliding velocities
investigated. It is noted, however, that rate-dependent behaviour,
was observed by Mariani et al. (2006) for strain rates lower than
ours, but only at 700

C, i.e., at depths 2045 km depending on
geothermal gradient (lowermost part of our prole). It is unclear if
and to what extent the rate-dependent behaviour of Mariani et al.
occurs at slow strain rates at lower temperature than 700

C.
We calculated our crustal strength prole using the approach
previously used by Bos and Spiers (2002) and Niemeijer and Spiers
(2005). We assumed a geothermal gradient of 30
km
1
and
a surface temperature of 25

C, an average crustal density of
2.7 gcm
3
, and a Byerlee friction coefcient of 0.75. A frictional
strength prole corresponding to m 0.2 is added to represent the
strength expected for the San Andreas fault zone on the basis of
heat ow measurements (Lachenbruch and Sass, 1980). The crystal
plastic ow strength of muscovite is added (Fig. 14) following
(Kronenberg et al., 1990) for shear strain rates of 10
14
, 10
12
, and
10
10
s
1
and assuming an activation enthalpy of 270 kJmol
1
(Mariani et al., 2006). The owstrength of quartz is added since this
is lower than the ow strength of muscovite under low-crustal
conditions (Fig. 14) following an empirical equation for dislocation
creep (Gleason and Tullis, 1995).
The shear strength of muscovite gouge inferred from our
experimental data, at depths of 510 km, is 3070 MPa, while
a strength of 1020 MPa is implied for the San Andreas fault zone
(Lachenbruch and Sass, 1980). Based on these results, it seems
Transcurrent fault
0
5
0 1
5 1
0 2
5 2
0 0 2 0 5 1 0 0 1 0 5 0
D
e
p
t
h

z

(
k
m
)
Byerlee's law
Muscovite
0 1
4 1 -
s u m
0 1
2 1 -
s u m
0 1
0 1 -
s u m
0 1
4 1 -
z t q
0 1
2 1 -
z t q
0 1
0 1 -
z t q
Muscovite
breakdown
Shear strength (MPa)
= 0.75
= 0.2
Fig. 14. Crustal strength prole comparing Byerlees Rule with the frictional strength
prole for muscovite, drawn using data presented in Fig. 9a. We assumed a geothermal
gradient of 30
Ckm
1
and a surface temperature of 25

C, an average crustal density of
2.7 gcm
3
, and a Byerlee friction coefcient of 0.75. The frictional strength prole
corresponding to m 0.2 is added to represent the strength expected for the San
Andreas fault zone on the basis of heat ow measurements (Lachenbruch and Sass,
1980). The plastic ow strength of muscovite is added following Kronenberg et al.
(1990) for various strain rates assuming an activation energy of 270 kJmol
1
after
Mariani et al. (2006). The plastic ow strength of wet quartz is added following
Gleason and Tullis (1995), since it is lower than the ow strength of muscovite under
low-crustal conditions.
unlikely that the presence of muscovite alone, with the rate-inde-
pendent frictional strength as taken in our study, can account for
the inferred weakness of large scale crustal fault zones. Muscovite,
however, shows stick-slip behaviour, which is the laboratory
equivalent of earthquakes, at temperatures between 400 and
500

C, corresponding to a depth of 1317 km in a transcurrent
fault zone. This unstable sliding behaviour might be of some
signicance in controlling fault behaviour towards the base of the
seismogenic zone.
5. Conclusions
The present study aimed to determine the frictional behaviour
of simulated muscovite fault gouge at high shear displacements
and under hydrothermal conditions in the temperature range 20
700

C, at shear strain rates w10
5
to 10
3
s
1
. The following
conclusions were reached:
(1) all our samples showed strong strain hardening at low shear
strains, followed by a more gradual increase in strength, until
steady state was reached. The steady state coefcient of friction
increased from 0.37 at room temperature to 0.56 at 300

C,
remaining around this value up to 600

C. At 700

C, the coef-
cient of friction decreased again, to a value of 0.38;
(2) all samples showed substantial grain size reduction at T up to
600

C due to pervasive and localized cataclasis, which resulted
in continuous compaction and hardening of the gouge. This
hardening is due to the progressive development of an anas-
tomosing network of ne grained, cataclastic shear bands,
gradually widening and hardening. Coarse grained relict lenses
between the cataclastic bands show folded and kinked
muscovite grains indicative of active ductile mechanisms;
(3) due to the presence of quartz impurities (<10%) in the gouge it
is possible that partial melting occurred at 700

C;
(4) on the basis of our results, it seems unlikely that the presence
of muscovite can signicantly contribute to the long-term
weakness of large scale crustal fault zones, unless its strength
dramatically decreases with decreasing sliding velocity or
shear strain rate.
Acknowledgements
This research was funded by the Netherlands Research Centre
for Integrated Solid Earth Science, project AM 2.1. We thank Thony
van der Gon Netscher and Gert Kastelein for constructing and
adjusting the hydrothermal rotary shear apparatus. We would also
like to thank Eimert de Graaff and Peter van Krieken for their
technical support and also Andre Niemeijer and Gill Pennock for
the useful discussions. Finally, we gratefully acknowledge Eli-
sabetta Mariani and Diane Moore for the constructive reviews.
References
Arancibia, G., Morata, D., 2005. Compositional variations of syntectonic white-mica
in low-grade ignimbritic mylonite. Journal of Structural Geology 27, 745767.
Balfour, N.J., Savage, M.K., Townend, J., 2005. Stress and crustal anisotropy in
Marlborough, New Zealand: evidence for low fault strength and structure-
controlled anisotropy. Geophysical Journal International 163, 10731086.
Blanpied, M.L., Lockner, D.A., Byerlee, J.D., 1995. Frictional slip of granite at hydro-
thermal conditions. Journal of Geophysical Research 100, 1304513064.
Bos, B., Peach, C.J., Spiers, C.J., 2000. Slip behavior of simulated gouge-bearing faults
under conditions favoring pressure solution. Journal of Geophysical Research
105, 1669916717.
Bos, B., Spiers, C.J., 2000. Effect of phyllosilicates on uid-assisted healing of gouge-
bearing faults. Earth and Planetary Science Letters 184, 199210.
Bos, B., Spiers, C.J., 2001. Experimental investigation into the microstructural and
mechanical evolution of phyllosilicate-bearing fault rock under conditions
favouring pressure solution. Journal of Structural Geology 23, 11871202.
Bos, B., Spiers, C.J., 2002. Frictional-viscous ow of phyllosilicate-bearing fault rock:
microphysical model and implications for crustal strength proles. Journal of
Geophysical Research 107, 2028. doi:10.1029/2001JB000301.
Burris, D.L., Sawyer, W.G., 2006a. Improved wear resistance in alumina-PTFE
nanocomposites with irregular shaped nanoparticles. Wear 260, 915918.
Burris, D.L., Sawyer, W.G., 2006b. A low friction and ultra low wear rate PEEK/PTFE
composite. Wear 261, 410418.
Byerlee, J., 1990. Friction, overpressure and fault normal compression. Geophysical
Research Letters 17, 21092112.
Byerlee, J.D., 1978. Friction of rocks. Pure and Applied Geophysics 116, 615626.
Carpenter, B.M., Marone, C., Saffer, D.M., 2009. Frictional behavior of materials in
the 3D SAFOD volume. Geophyscial Research Letters 36. doi:10.1029/
2008GL036660.
Chatterjee, N.D., Johannes, W., 1974. Thermal stability adn standard thermodynamic
properties of synthetic 2M1-muscovite, KAl2[AlSi3O10(OH)2]. Contributions to
Mineralogy and Petrology 48, 89114.
Chester, F.M., 1995. A rheologic model for wet crust applied to strike slip faults.
Collettini, C., Barchi, M.R., 2002. A low-angle normal fault in the Umbria region
(Central Italy): a mechanical model for the related microseismicity. Tectono-
physics 359, 97115.
Collettini, C., Holdsworth, R.E., 2004. Fault zone weakening and character of slip
along low-angle normal faults: insights from the Zuccale fault, Isle of Elba, Italy.
Journal of the Geological Society of London 161, 10391051.
Deer, W.A., Howie, R.A., Zussman, J., 1962. An Introduction to the Rock Forming
Minerals: Micas. Longman Group Limited, London.
Di Toro, G., Hirose, T., Nielsen, S., Pennacchioni, G., Shimamoto, T., 2006. Natural and
experimental evidence of melt lubrication of faults during earthquakes. Science
311, 647649.
Dieterich, J.H., 1978. Time-dependent friction and the mechanics of stick-slip. Pure
and Applied Geophysics (Historical Archive) 116, 790806.
Evans, J.P., Chester, F.M., 1995. Fluid-rock interaction in faults of the San Andreas
tures. Journal of Geophysical Research 100, 1300713020.
Faulkner, D.R., Rutter, E.H., 2001. Can the maintenance of overpressured uids in
large strike-slip fault zones explain their apparent weakness? Geology 29,
503506.
Gleason, G.C., Tullis, J., 1995. A ow law for dislocation creep of quartz aggregates
determined with the molten salt cell. Tectonophysics 247, 123.
Goetze, C., Evans, B., 1979. Stress and temperature in the bending lithosphere as
constrained by experimental rock mechanics. Geophysical Journal International
59, 463478.
He, C., Wang, Z., Yao, W., 2007. Frictional sliding of gabbro gouge under hydro-
thermal conditions. Tectonophysics 445, 353362.
He, C., Yao, W., Wang, Z., Zhou, Y., 2006. Strength and stability of frictional sliding of
gabbro gouge at elevated temperatures. Tectonophysics 427, 217229.
Hickman, S.H., 1991. Stress in the lithosphere and the strength of active faults.
Review of Geophysics 29, 759775.
Hickman, S.H., Sibson, R.H., Bruhn, R., 1995. Introduction to special section:
mechanical involvement of uids in faulting. Journal of Geophysical Research
100, 1283112840.
Holdsworth, R.E., 2004. Weak faults rotten cores. Science 303, 181182.
Holdsworth, R.E., Stewart, M., Imber, J., Strachan, R.A., 2001. The structure and
rheological evolution of reactivated continental fault zones: a review and case
study. In: Miller, J.A., Holdsworth, R.E., Buick, I.S., Handy, M.R. (Eds.), Continental
Reactivation and Reworking, vol. 184. Geological Society of London, pp. 115137.
Huang, W.L., Wyllie, P.J., 1974. Melting relations of muscovite with quartz and
sanidine in the K2O-Al2O3-SiO2-H2O system to 30 kilobars and an outline of
paragonite melting relations. American Journal of Science 274, 378395.
Ikari, M.J., Saffer, D.M., Marone, C., 2007. Effect of hydration state on the frictional
properties of montmorillonite-based fault gouge. Journal of Geophysical
Research 112. doi:10.1029/2006JB004748.
Imber, J., Holdsworth, R.E., Butler, C.A., Lloyd, G.E., 1997. Fault-zone weakening
processes along the reactivated Outer Hebrides Fault Zone, Scotland. Journal of
the Geological Society of London 154, 105109.
Jefferies, S.P., Holdsworth, R.E., Shimamoto, T., Takagi, H., Lloyd, G.E., Spiers, C.J.,
2006a. Origin and mechanical signicance of foliated cataclastic rocks in the
cores of crustal-scale faults: examples from the Median Tectonic Line, Japan.
Journal of Geophysical Research 111. doi:10.1029/2005JB004205.
Jefferies, S.P., Holdsworth, R.E., Wibberley, C.A.J., Shimamoto, T., Spiers, C.J.,
Niemeijer, A.R., Lloyd, G.E., 2006b. The nature and importance of phyllonite
development in crustal-scale fault cores: an example from the Median Tectonic
Line, Japan. Journal of Structural Geology 28, 220235.
Kanagawa, K., 2002. Frictional behavior of synthetic gouge-bearing faults under the
operation of pressure solution. Earth Planets and Space 54, 11471152.
Kanagawa, K., Cox, S.F., Zhang, S., 2000. Effects of dissolutionprecipitation
processes on the strength and mechanical behavior of quartz gouge at high-
temperature hydrothermal conditions. Journal of Geophysical Research 105,
11,11511,126.
Kronenberg, A.K., Kirby, S.H., Pinkston, J., 1990. Basal slip and mechanical anisotropy
of biotite. Journal of Geophysical Research 95, 1925719278.
Lachenbruch, A.H., Sass, J.H., 1980. Heat ow and energetic of the San Andreas fault
zone. Journal of Geophysical Research 85, 61856223.
Lehner, F.K., Bataille, J., 1984. Nonequilibrium thermodynamics of pressure solution.
Pure and Applied Geophysics 122, 5385.
Logan, J.M., Rauenzahn, K.A., 1987. Frictional dependence of gouge mixtures of
quartz and montmorillonite on velocity, composition and fabric. Tectonophysics
144, 87108.
Mares, V.M., Kronenberg, A.K., 1993. Experimental deformation of muscovite.
Mariani, E., Brodie, K.H., Rutter, E.H., 2006. Experimental deformation of muscovite
shear zones at high temperatures under hydrothermal conditions and the
strength of phyllosilicate-bearing faults in nature. Journal of Structural Geology
28, 15691587.
Meike, A., 1989. In situ-deformation of micas: a high-voltage electron-microscope
study. American Mineralogist 74, 780796.
Miller, S.A., Nur, A., Olgaard, D.L., 1996. Earthquakes as a coupled shear stress-high
pore pressure dynamical system. Geophysical Research Letters 23, 197200.
Miller, S.A., Olgaard, D.L., 1997. Modeling seismicity clustering and fault weakness
due to high pore pressures. Physics and Chemistry of The Earth 22, 4348.
Moore, D.E., Lockner, D.A., 2004. Crystallographic controls on the frictional behavior
of dry and water-saturated sheet structure minerals. Journal of Geophysical
Research 109. doi:10.1029/2003JB002582.
Moore, D.E., Lockner, D.A., 2007. Comparative deformation behavior of minerals in
serpentinized ultramac rock: application to the slabmantle interface in
subduction zones. International Geology Review 49, 401415.
Moore, D.E., Lockner, D.A., Shengli, M., Summers, R., Byerlee, J.D., 1997. Strengths of
serpentinite gouges at elevated temperatures. Journal of Geophysical Research
102, 1478714801.
Moore, D.E., Rymer, M.J., 2007. Talc-bearing serpentinite and the creeping section of
the San Andreas fault. Nature 448, 795797.
Morrow, C.A., Moore, D.E., Lockner, D.A., 2000. The effect of mineral bond strength
and adsorbed water on fault gouge frictional strength. Geophysical Research
Letters 27, 815818.
Morrow, C.A., Radney, B., Byerlee, J.D., 1992. Frictional strength and the effective
pressure law of montmorillonite and illite clays. In: Evans, B., Wong, T.-F. (Eds.),
Fault Mechanics and Transport Properties of Rocks. Academic, San Diego, CA,
pp. 6988.
Nakatani, M., Scholz, C.H., 2004. Frictional healing of quartz gouge under hydro-
thermal conditions: 1. Experimental evidence for solution transfer healing
mechanism. Journal of Geophysical Research 109. doi:10.1029/2001JB001522.
Niemeijer, A.R., Spiers, C.J., 2005. Inuence of phyllosilicates on fault strength in the
brittle-ductile transition: insights from rock analogue experiments. In:
Bruhn, D., Burlini, L. (Eds.), High Strain Zones: Structure and Physical Properties,
vol. 245. Geological Society of London, pp. 303327.
Niemeijer, A.R., Spiers, C.J., 2006. Velocity dependence of strength and healing behav-
iour in simulated phyllosilicate-bearing fault gouge. Tectonophysics 427, 231253.
Niemeijer, A.R., Spiers, C.J., 2007. A microphysical model for strong velocity weak-
ening in phyllosilicate-bearing fault gouges. Journal of Geophysical Research
112. doi:10.1029/2007JB005008.
Niemeijer, A.R., Spiers, C.J., Bos, B., 2002. Compaction creep of quartz sand at 400
600 C: experimental evidence for dissolution-controlled pressure solution.
Earth and Planetary Science Letters 195, 261275.
Niemeijer, A.R., Spiers, C.J., Peach, C.J., 2008. Frictional behaviour of simulated
quartz fault gouges under hydrothermal conditions: results from ultra-high
strain rotary shear experiments. Tectonophysics 460, 288303.
OHara, K., 2007. Reaction weakening and emplacement of crystalline thrusts:
diffusion control on reaction rate and strain rate. Journal of Structural Geology
29, 13011314.
Ranalli, G., 1995. Rheology of the Earth. Chapman & Hall. 413p.
Rawling, G.C., Goodwin, L.B., 2003. Cataclasis and particulate ow in faulted, poorly
lithied sediments. Journal of Structural Geology 25, 317331.
Rutter, E.H., Holdsworth, R.E., Knipe, R.J., 2001. The nature and tectonic signicance
of fault zone weakening: an introduction. In: Holdsworth, R.E., Strachan, R.A.,
Magloughlin, J.F., Knipe, R.J. (Eds.), The Nature and Tectonic Signicance of Fault
Zone Weakening, vol. 186. Geological Society of London, pp. 111.
Rutter, E.H., Maddock, R.H., 1992. On the mechanical properties of synthetic kao-
lintie/quartz fault gouge. Terra Nova 4, 489500.
Rutter, E.H., Maddock, R.H., Hall, S.H., White, S.H., 1986. Comparative microstruc-
tures of natural and experimentally produced clay-bearing fault gouges. Pure
and Applied Geophysics (Historical Archive) 124, 330.
SAFOD core atlas, v.., 2007. Available from: www.earthscope.org.
Saileswaran, N., Panchanathan, V., 1973. Compaction of grains. General parameter
evaluation. Powder Technology 8, 1926.
Sammis, C.G., King, G., Biegel, R.L., 1987. The kinematics of gouge deformation. Pure
and Applied Geophysics 125, 777812.
Schleicher, A.M., Tourscher, S.N., van der Pluijm, B.A., Warr, L.N., 2009a. Constraints
on mineralization, uid-rock interaction, and mass transfer during faulting at
23 km depth from the SAFOD drill hole. Journal of Geophysical Research 114.
Schleicher, A.M., van der Pluijm, B.A., Warr, L.N., 2008. What Controls Creep on the
San Andreas Fault at the SAFOD Drillhole? Eos Transactions AGU 89 (53) Fall
Meet. Suppl. Abstract T13A1910.
Schleicher, A.M., Warr, L.N., van der Pluijm, B.A., 2009b. On the origin of mixed-
layered clay minerals from the San Andreas Fault at 2.53 km vertical depth
(SAFOD drillhole at Parkeld, California). Contributions to Mineralogy and
Petrology 157, 173187.
bridge University Press Cambridge, UK. 471pp.
Scruggs, V.J., Tullis, T.E., 1998. Correlation between velocity dependence of friction
and strain localization in large displacement experiments on feldspar, musco-
vite and biotite gouge. Tectonophysics 295, 1540.
Shea, W.T., Kronenberg, A.K., 1992. Rheology and deformation mechanisms of an
isotropic mica schist. Journal of Geophysical Research 97, 1520115237.
Shimamoto, T., Logan, J.M., 1981. Effects of simulated clay gouges on the sliding
behavior of Tennessee sandstone. Tectonophysics 75, 243255.
Sibson, R.H., 1983. Continental fault structure and the shallow earthquake source.
Journal of the Geological Society of London 140, 741767.
Sibson, R.H., 2004. Controls on maximum uid overpressure dening conditions for
mesozonal mineralisation. Journal of Structural Geology 26, 11271136.
Sleep, N.H., 1995. Ductile creep, compaction, and rate and state dependent friction
within major fault zones. Journal of Geophysical Research 100, 1306513080.
Takahashi, M., Mizoguchi, K., Kitamura, K., Masuda, K., 2007. Effects of clay content
on the frictional strength and uid transport property of faults. Journal of
Tembe, S., Lockner, D.A., Solum, J.G., Morrow, C.A., Wong, T.-F., Moore, D.E., 2006a.
Frictional strength of cuttings and core from SAFOD drillhole phases 1 and 2.
Geophyscial Research Letters 33. doi:10.1029/2006GL027626.
Tembe, S., Lockner, D.A., Wong, T.-F., 2006b. Strength of SAFOD fault gouge under
hydrothermal conditions. Eos Transactions AGU 87 Abstract S23C-0181.
Theiler, G., Gradt, T., 2008. Inuence of the temperature on the tribological
behaviour of PEEK composites in vacuum environment. Journal of Physics:
Conference Series 100. doi:10.1088/1742-6596/100/7/072040.
Townend, J., Zoback, M.D., 2004. Regional tectonic stress near the San Andreas fault
in central and southern California. Geophysical Research Letters 31. doi:10.1029/
2003GL018918.
White, S.H., Bretan, P.G., Rutter, E.H., 1986. Fault-zone reactivation: kinematics and
mechanisms. Philosophical Transactions of the Royal Society of London A317,
8197.
Wibberley, C.A.J., 1999. Are feldspar-to-mica reactions necessarily reaction-soft-
ening processes in fault zones? Journal of Structural Geology 21, 12191227.
Wintsch, R.P., Christoffersen, R., Kronenberg, A.K., 1995. Fluid-rock reaction weak-
ening of fault zones. Journal of Geophysical Research 100, 1302113032.
Wu, F.T., Blatter, L., Roberson, H., 1975. Clay gouges in the San Andreas Fault System
and their possible implications. Pure and Applied Geophysics (Historical
Archive) 113, 8795.
Zhang, S., Tullis, T.E., Scruggs, V.J., 2001. Implications of permeability and its
anisotropy in a mica gouge for pore pressures in fault zones. Tectonophysics
335, 3750.
Zoback, M.D., Hickman, S.H., Ellsworth, W., Kirschner, D., Pennell, N.B., Chery, J.,
Sobolev, S., 2007. Preliminary Results from SAFOD Phase 3: implications for the
state of stress and shear localization in and near the San Andreas Fault at depth
in central California. Eos Transactions AGU 88 (52) Fall Meet. Suppl. Abstract
T13 G-03.
Zoback, M.D., Zoback, M.L., Mount, V.S., Suppe, J., Eaton, J.P., Healy, J.H.,
Oppenheimer, D., Reasenberg, P., Jones, L., Rayleigh, C.B., Wong, I.G., Scotti, O.,
Wentworth, C., 1987. New evidence on the state of stress of the San Andreas
fault system. Science 238, 11051111.
The effect of the intermediate principal stress on fault formation
and fault angle in siltstone
Bezalel Haimson
a,
*
, John W. Rudnicki
b
a
University of Wisconsin, Madison, WI 53706, USA
b
Northwestern University, Evanston, IL 60208, USA
Article history:
5 July 2009
Keywords:
Bifurcation
Fault angle
Faulting
Siltstone
MohrCoulomb
Shear localization
Strength criterion
True triaxial test
a b s t r a c t
We conducted true triaxial compression tests on specimens prepared from two siltstone core sections,
one above and one below the Chelungpu Fault, Taiwan. For different constant s
2
and s
3
magnitudes, the
maximum principal stress (s
1
) was raised until a post failure stage was reached, and a through-going
fault had developed. Despite differences between the properties of the two cores, in all tests peak s
1
increased as s
2
was set at higher levels than s
3
, in contrast to MohrCoulomb condition predictions. The
faultnormal vector was aligned with the s
3
direction and made an angle (q) with s
1
direction. The angle
q, which corresponds to fault dip in case of normal faulting, increased monotonically with s
2
for xed s
3
,
a variation that is also inconsistent with MohrCoulomb theory.
The results of shear band localization theory are used with fault angles observed for axisymmetric
compression and deviatoric pure shear to infer properties of the inelastic constitutive behavior. These
properties are signicantly different for the two cores. Using them to predict q for other deviatoric stress
states yields good agreement with the observations for core II and acceptable agreement for core I. The
results are used to predict the angle variation for constant mean normal stress (q decreases as the
deviatoric stress state varies from axisymmetric extension to axisymmetric compression) and at xed
deviatoric stress state (q decreases monotonically with increasing mean normal stress).
1. Introduction
Laboratory experiments simulating compressive failure and
faulting in rocks are typically conducted on cylindrical specimens
subjected to constant lateral conning pressure and a rising axial
load until brittle fracture occurs. These conventional triaxial tests
replicate only a special case of crustal condition, that inwhich two of
the principal stresses are equal. Conventional triaxial tests on rocks
were conducted as early as the turn of the last century (Von Ka rma n,
1911). They gained acceptance because of the relatively simple
equipment, specimen preparation, and testing procedure. The
ubiquity of conventional triaxial testing can also be traced to the
assumption that the intermediate principal stress (s
2
) has a negli-
gible effect on rock failure characteristics as expressed, for example,
in the Mohr or MohrCoulomb failure criteria (Jaeger et al., 2007).
However, indications from the three major types of faulting
encountered in the eld, and results of numerous in situ stress
measurements at depths reaching several kilometers (McGarr and
Gay, 1978; Brace and Kohlstedt, 1980), point to a state of stress in
the earths crust that is fully three-dimensional (s
1
ss
2
ss
3
).
Murrell (1963), Handin et al. (1967), and Mogi (1967) compared
results of conventional triaxial compression tests with those of
conventional triaxial extension and deduced that the differences in
rock resistance to faulting between the two modes of loading were
due to the different magnitudes of s
2
applied. Inspired by this
evidence, Mogi (1971) introduced a true triaxial testing machine in
which rectangular prismatic specimens were subjected to three
different principal stresses. He found that indeed s
2
affects the
stress level at which faulting occurs, and hence the rock strength
criterion, as well as the angle at which the fault develops. Little
follow-up on Mogis seminal work took place until Haimson and co-
workers carried out similar true triaxial tests in igneous rocks
(Westerly granite, Haimson and Chang, 2000; Pohang rhyolite,
Chang and Haimson, 2007); a metamorphic rock (KTB amphibolite,
Chang and Haimson, 2000; Haimson and Chang, 2002) and a sedi-
mentary rock (siltstone, Oku et al., 2007). They found that strength,
deformability, and fault angle, were affected by s
2
in all the tested
rocks. An exception was found in tests of Long Valley (California)
* Corresponding author. Tel.: 1 608 262 2563.
E-mail address: bhaimson@wisc.edu (B. Haimson).
doi:10.1016/j.jsg.2009.08.017
ultra ne-grained hornfels and metapelite (Chang and Haimson,
2005), unusual rocks that appear to have no dilatancy, and develop
no visible microcracks.
Independently, building upon the antecedents of Hadamard
(1903), Mandel (1966), Thomas (1961) and Hill (1962), Rudnicki
and Rice (1975) (also, Rice, 1976; Besue lle and Rudnicki, 2004)
suggested a description of failure as a bifurcation from homoge-
neous (spatially uniform) deformation that predicted a strong
dependence on s
2
(via the deviatoric stress state). More specically,
Rudnicki and Rice (1975) established conditions for which a solu-
tion corresponding to concentrated deformation in a planar band
was an alternative to continued homogeneous deformation. The
appearance of this localized mode of deformation is often essen-
tially coincident with failure by development of a through-going
fault or fracture, but in other cases it may be the precursor to a more
extended evolution of localized deformation that ultimately
requires signicant additional strain for failure. Analysis based on
this approach yields a relation among constitutive parameters
required for the onset of bifurcation and, hence, depends strongly
on how the homogeneous deformation prior to bifurcation, espe-
cially the inelastic portion, is modeled.
The predictions of the failure stress by the bifurcation approach
depend strongly on certain details of the constitutive behavior that
are difcult to determine experimentally and, for various other
reasons, are difcult to compare with experimental observations of
failure (Besue lle and Rudnicki, 2004). However, the prediction for
the fault angle is much less sensitive to these details and is more
easily compared with observations in terms of the constitutive
parameters for homogeneous deformation just prior to bifurcation.
Rudnicki (2008a,b) has used and extended results from the bifur-
cation theory to interpret observations of failure plane inclinations
in true triaxial tests of Westerly Granite (Haimson and Chang,
2000) and to infer aspects of the constitutive behavior.
In this paper we describe two series of true triaxial tests con-
ducted on samples of siltstone taken fromthe hanging wall and the
footwall of the Chelungpu fault, Taiwan. The tests reveal a clear
dependency of strength and fault angle on the magnitude of s
2
. The
results are compared with predictions based on shear localization
theory incorporating a yield surface and plastic potential that
depend on three stress invariants (rather than two, as in Rudnicki
and Rice (1975)). Dependences of the yield surface and plastic
potential on mean stress are inferred fromthe fault angles observed
in axisymmetric compression and deviatoric pure shear. These
dependences are used to compare the predicted fault angles with
observations for other deviatoric stress states and to predict the
variation that would be observed with mean stress for xed
deviatoric stress state and with deviatoric stress state for xed
mean stress.
2. Rocks tested
Rock specimens used in the true triaxial tests described here
came from core recovered from the scientic hole A, near the
northern end of the Chelungpu fault. The hole was drilled as part of
the Taiwan Chelungpu Fault Drilling Project (TCDP). The project
was undertaken to study the faulting mechanism behind the
destructive Chi-Chi earthquake (1999; M
w
7.6), characterized as
a thrust motion across the North-South striking Chelungpu fault
(Shin and Teng, 2001; Lin et al., 2003). Core made available to us
came from short sections centered at the depth of 891 m (core I)
and 1252 m (core II), straddling the active fault, which was inter-
cepted at 1111 m. Core I is a siltstone belonging to the early Pleis-
tocene Cholan Formation, which persists to a depth of 1013 m; core
II is also a siltstone, belonging to the Pliocene Chinshui Formation,
which prevails at depths of 10131313 m. Thus, core II is
representative of the rock traversed by the active Chelungpu fault.
Core I comes from a somewhat younger formation. As shown
below, there are distinct differences in mineral content, as well as in
mechanical behavior between the two siltstones. We cannot tell
whether the differences are related to their juxtaposition with
respect to the fault and its activity or, what appears more likely, the
consequence of their different deposition ages.
The siltstone in core I contains 68% quartz, 19.5% clay, 9.5%
feldspar, and 3% biotite; the siltstone in core II consists of 65%
quartz, 25.5% clay, 7.5% feldspar, and 2% biotite. The only major
difference is the amount of clay, and this may have a role in the
disparity between the two core sections in some of their physical
and mechanical properties (Table 1).
The 891 m siltstone is both stronger and stiffer in compression
than the 1252 m core. Also notable is that core I rock has signi-
cantly larger grain size than core II. Although the siltstone is
a sedimentary rock, no bedding planes were visible, and neither
core showed signs of inhomogeneity. We examined the degree of
anisotropy by running unconned (uniaxial) compression tests in
which cylindrical specimens (25.4 mm in diameter) were drilled
out of the vertical core I at inclinations of 0
, 30
, 60
, and 90
. The
compressive strengths in specimens from the four inclinations
were all within 3% of the mean 79.4 MPa. Hence, we considered the
siltstone to be basically isotropic with respect to strength. With
respect to elastic modulus, the largest difference from the mean
13.7 GPa was less than 7%, a sign of a rather mild degree of
anisotropy. These results enabled us to consider the siltstone
practically isotropic.
3. Experiment setup and procedure
The apparatus used in our tests was a recently fabricated true
triaxial testing machine (Haimson and Chang, 2000). As stated
above, it had been successfully employed to characterize mechan-
ical properties under the most general 3Dstate of stress of Westerly
granite, KTB amphibolite, Long Valley hornfels and metapelite, and
Pohang rhyolite. The true triaxial cell facilitates the application of
three principal stresses to a rectangular prismatic specimen (size
19 19 38 mm
3
) by use of three independent servo-controlled
units. The application of the maximum principal stress (s
1
) in the
axial direction of the specimen and of the intermediate principal
stress (s
2
) in one of the two lateral directions is carried out by use of
two pairs of hydraulically driven pistons. The minimum principal
stress (s
3
) is directly applied to the other pair of lateral faces by
conning uid pressure inside the cell. Details of the testing
system, and its calibration, can be found in Haimson and Chang
(2000).
In selected specimens strain gages for measuring strains in the
direction of s
1
and s
2
were afxed to the faces subjected to s
3
loading. Strain in the s
3
direction was measured using a beryllium
copper strain-gaged beam mounted on one of the specimen s
3
faces.
Table 1
Some physical and mechanical properties of core I and core II.
Property Core I (891 m) Core II (1252 m)
Mean grain size, mm 56 (200) 44 (200)
Dry density, kg/m
3
2594 (28) 2587 (28)
Effective porosity, % 6.9 (15) 6.1 (15)
Unconned compressive strength
(UCS), MPa
79.5 (2) 63.4 (2)
Youngs modulus, GPa 13.7 (2) 9.2 (2)
Poissons ratio 0.13 (2) 0.2 (2)
Brazilian tensile strength, MPa 5.4 (11) 4.2 (3)
Note: Numbers in parenthesis refer to the amount of measurements conducted.
B. Haimson, J.W. Rudnicki / Journal of Structural Geology 32 (2010) 17011711 1702
The tests reported here were conducted under dry conditions.
For that purpose the s
3
faces, as well as the spaces between the
edges of the anvils applying the other two loads were coated with
a thin layer of polyurethane so that the conning uid could not
penetrate into the specimen. The scarcity of available core pre-
vented us from testing saturated siltstone under pore pressure.
However, that condition would have only affected the strength
magnitudes, not the general mechanical behavior.
Testing procedure consisted of raising s
1
at a constant strain rate
of 8 10
6
s
1
while holding s
2
and s
3
at their preset magnitudes,
until a post failure stage was reached, at which s
1
had declined
about 10% from its peak level. Upon unloading, tested specimens
were sectioned along the s
1
s
3
plane in order to record fault atti-
tude. In selected samples, sections were also prepared for SEM
inspection.
4. Triaxial strength
4.1. Conventional triaxial strength
Conventional triaxial tests (s
2
s
3
) were rst carried out in
order to establish the Mohr strength criterion in terms of principal
stresses, for later comparison with the true triaxial compressive
strength (s
2
>s
3
). The Mohr criterion was obtained by tting
a monotonically increasing power function to the experimental
data (Fig. 1). In core I the best-tting criterion is expressed by
a power function:
s
1
76:89 9:34s
0:80
3
R 0:997 (1a)
In core II the Mohr criterion takes the form:
s
1
73:23 6:36s
0:84
3
R 0:998 (1b)
The tting is excellent in both cases, with core II showing lower
conventional triaxial strength. The experimental results in both
cores can also be tted by linear regression (MohrCoulomb
criterion):
s
1
94:9 3:6s
3
R 0:992 for core I (2a)
s
1
84:2 3:0s
3
R 0:992 for core II (2b)
Again, core II proves to be weaker throughout the range of s
3
(Fig. 2).
4.2. True triaxial strength
True triaxial tests were conducted for several constant magni-
tudes of s
3
. For a given s
3
, a series of tests were run, each at
different s
2
between s
2
s
3
and s
2
approaching peak s
1
. The
results in terms of peak s
1
versus s
2
for the different preset s
3
are
plotted in Fig. 3a and b. The solid line in the plots represents the
Mohr criterion determined from tests in which s
2
s
3
. Dashed
curves showthe trend of the true triaxial strength for each constant
s
3
. The plots reveal a common characteristic of gradually increasing
strength with the rise in s
2
until a top level is reached, followed by
a gradual decline, as predicted theoretically by Wiebols and Cook
(1968) and conrmed by tests in other rocks by Mogi (1971),
Haimson and Chang (2000), and Chang and Haimson (2000, 2007).
The higher compressive strength when s
2
>s
3
as compared with
that at s
2
s
3
reveals the inadequacy of the Mohr (or Mohr
Coulomb) criterion to predict rock failure under the most general
state of stress.
We attempted to determine the true triaxial strength criteria for
siltstones by rst employing Nadais (1950) proposed relationship.
He suggested a true triaxial strength criterion for brittle materials,
such as rocks, in terms of the two stress invariants, octahedral shear
stress (s
oct
) and octahedral normal stress, or mean stress (s
oct
). He
related these invariants by a function (f) dependent on the rock
material properties in the form of s
oct
f s
oct
. Plotting all test
results shown in Fig. 3 in the Nadai domain (Fig. 4) yields the
following best-t power function curves:
s
oct
5:27s
0:60
oct
0
100
200
300
400
500
0 20 40 60 80 100 120
1

(
M
P
a
)
2
=
3
(MPa)
1
= 76.89 + 9.338
3
0.80
R = 0.997
Core I
1
= 73.23 + 6.36
3
0.84
R = 0.998
Core II
Fig. 1. Maximum compressive principal stress (s
1
) at failure as a function of the least
principal stress (s
3
) under conventional triaxial stress condition (s
2
s
3
), and the best-
tting power function strength criterion (Mohr) for each of the two siltstone cores.
0
100
200
300
400
500
0 0 2 40 60 80 100 120
1

(
M
P
a
)
2
=
3
(MPa)
1
= 84.2 + 3.0
3
R = 0.992
Core II
1
= 94.9+ 3.6
3
R = 0.992
Core I
Fig. 2. Maximum compressive principal stress (s
1
) at failure as a function of the least
principal stress (s
3
) under conventional triaxial stress condition (s
2
s
3
), and the best-
tting linear strength criterion (MohrCoulomb) for each of the two siltstone cores.
s
oct
5:10s
0:58
oct
where
s
oct

_
_
s
1
s
2
2
s
2
s
3
2
s
3
s
1
2
_
1=2
__
3 (4)
and
s
oct
s
1
s
2
s
3
=3 (5)
The Nadai criterion appears to t the data reasonably well, albeit
with considerable scatter.
Mogi (1971) adjusted the Nadai criterion to correspond with the
mode of compressive failure in brittle rock, which does not occur
over the entire volume (as expressed by Eqs. (3)), but is restricted to
faulting along a plane aligned with the s
2
direction. For this reason
he replaced s
oct
with the mean stress acting on the plane of failure
(s
m,2
). All data points in Fig. 3 are well-tted by a power function in
Mogis domain (Fig. 5):
0
100
200
300
400
500
600

(
M
P
a
)
2
(MPa)
Mohr criterion (
2
=
3
)
2
=
1
Uniaxial compressive strength C
0
25 MPa
40 MPa
60 MPa
100 MPa
3
= 10 MPa
Core I
0
100
200
300
400
500
600
0 100 200 300 400 500 0 100 200 300 400 500

(
M
P
a
)
2
(MPa)
Mohr criterion (
2
=
3
) 2
=
1
3
= 10 MPa
40 MPa
60 MPa
100 MPa
Uniaxial compressive strength C
0
Core II
a b
Fig. 3. Variation of peak compressive stress s
1
as a function of s
2
for different constant values of s
3
(a) in the core I and (b) in core II.
0
50
100
150
200
0 50 100 150 200 250 300 350
o
c
t
(
M
P
a
)
oct
(MPa)
oct
(MPa) =5.27
oct
0.60
R = 0.927
Core I
oct
(MPa) = 5.10
oct
0.58
R = 0.915
Core II
Fig. 4. A true triaxial strength criterion for the siltstone (cores I and II), based on all the
experimental results shown in Fig. 3, in the Nadai (1950) domain of s
oct
versus s
oct
.
0
50
100
150
200
0 50 100 150 200 250 300 350
o
c
t
(
M
P
a
)
m,2
(MPa)
oct
(MPa) = 2.32
m,2
0.75
R = 0.995
Core I
oct
(MPa) = 2.86
m,2
0.69
R = 0.991
Core II
Fig. 5. A true triaxial strength criterion for the siltstone (cores I and II), based on all the
experimental results shown in Fig. 3, in the Mogi (1971) domain of s
oct
versus s
m,2
.
s
oct
2:32s
0:75
m;2
s
oct
2:86s
0:69
m;2
where
s
m;2
s
1
s
3
=2 (7)
Comparison between Figs. 4 and 5 makes it clear that Mogis
relationship (Eqs. (6)) better represents the true triaxial strength
criteria for both cores. A physical interpretation of Eqs. (6) is that
brittle failure, or faulting, occurs when the distortional strain
energy reaches a critical value that increases monotonically with
the mean normal stress on the failure plane.
5. Fault angle
Failure in true triaxial tests took the form of single or conju-
gate through-going faulting. The faultnormal direction and the
angle (q) between it and s
1
were carefully measured at the
conclusion of each test. Under conventional triaxial stresses fault
normal direction was random since s
2
s
3
. Conventional triaxial
strength data appear to be tted well by the MohrCoulomb
criterion (Fig. 2). The linear relationship represented by this
criterion enables the computation of the unique angle of internal
friction f:
f 34:6
for core I (8a)

f 30:3
for core II (8b)

Subsequently, the theoretical fault angle q is also unique,
regardless of s
2
or s
3
magnitudes:
q 45
34:6=2 62:3
for core I (9a)

q 45
30:3=2 60:2
for core II (9b)

However, the fault angles measured in tested specimens that have
been subjected to conventional triaxial stress conditions show
a gradual decrease from 73
to 59
in core I, and from 76
to 52
in
core II, as s
2
s
3
increases from 10 MPa to 100 MPa (Fig. 6). A
similar trend of fault angle decrease with conning pressure
increase has been documented in other rocks, such as Fontaine-
bleau sandstone (Haied et al., 2000), Westerlygranite (Haimson and
Chang, 2000), and KTB amphibolite (Chang and Haimson, 2000).
This variation of the fault angle with lateral stress contradicts the
prediction based on the MohrCoulomb criterion (Eqs. (9)).
In true triaxial tests the trend in both cores is for the fault angle
to increase with s
2
for constant s
3
. The total increase is limited to
about 10
in the two siltstones, less than in the Westerly granite

(Haimson and Chang, 2000) or the KTB amphibolite (Chang and
Haimson, 2000), but still signicant. Fig. 7a and b shows the results
of all measured fault angles. The increase in fault angle with s
2
for
constant s
3
cannot be explained by Mohr theory (Eqs. (1) or (2)),
which neglects the effect of s
2
on rock failure. Using the power
function strength criterion (Mohr) in Fig. 3 would introduce
a dependence of the fault angle on s
2
but not on s
3
.
6. Fault angle prediction from bifurcation theory
Rudnicki and Rice (1975) gave a prediction for the fault angle for
the most general form of a rate-independent, elastically isotropic
plastic constitutive relationwith a smoothyield surface (the surface
that forms the boundary of the stress states that cause only elastic
deformation) and plastic potential (which gives the direction of the
inelastic strain increments in stress space) that depend only on two
of the three stress invariants. This form included the possibility of
dependence of the yield stress in shear on mean stress (either
positive, as on a frictional yield surface or negative, as on a yield
surface cap) and inelastic volume change (either compaction or
dilatancy). Rudnicki and Olsson (1998) gave a more convenient
rearrangement of the Rudnicki and Rice result. This expression also
incorporates a correction (Perrin and Leblond, 1993; also, noted by
Ottosen and Runesson, 1991) for the limiting cases in which the
band is predicted to be perpendicular to the most compressive
stress (compaction band) or the least compressive stress (dila-
tion band) rather to the intermediate principal stress, which is the
most common case.
Ottosen and Runesson (1991) derived an expression for the
predicted fault angle for a yield surface and plastic potential that
depended on all three stress invariants. Rudnicki (2008a) noted
that the prediction could be obtained by replacement of certain
quantities in the original Rudnicki and Rice (1975) expression. This
result is explained in more detail in Rudnicki (2008b) and
summarized here.
From the experimental point of view, it is natural to express
results in terms of the principal stresses, since these are the
quantities applied or measured. Conceptually, however, it is more
natural to use the stress invariants. For an isotropic material these
two representations are completely equivalent. For example, the
Nadai criterion (Eq. (3)) can be expressed in terms of two invari-
ants whereas the MohrCoulomb or Mogi criterion (Eq. (6))
requires three. Using the stress invariants has the advantage that
they are independent of the particular choice of axes and, in
applications; this avoids the necessity of determining principal
stress directions at each point. More importantly, from a physical
point of view, we expect that the shear (deviatoric) stress and
mean (hydrostatic stress) have different effects on rock behavior.
These contributions are clearly identied by the invariants,
whereas changing a particular principal stress changes both the
deviatoric and mean stresses.
The theory is based on forms for the yield surface and the plastic
potential surface that depend on all three stress invariants. The
invariants used here are chosen as follows: The vonMises equivalent
50
55
60
65
70
75
80
0 20 40 60 80 100 120 140
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
2
=
3
(MPa)
(deg) = 74 - 0.16
3
R = 0.908
Core I
(deg) = 79 - 0.26
3
R = 0.993
Core II
Fig. 6. Fault angle (q) variation with s
2
s
3
under conventional triaxial stress condi-
tions (cores I and II).
stress s which is equal to

3=2
_
s
oct
(Eq. (4)), s
oct
(Eq. (5)), and the
Lode angle q
L
. The Lode angle is dened as 3q
L
arcsin
27
p
J
3
=2s
3
where J
3
dets
ij
is the thirdinvariant of the deviatoric stress. (The
difference in sign from Rudnicki (2008a,b) occurs because stresses
are taken as positive in compression here). In the space of principal
stresses, q
L
is the angle in planes that are normal to the hydrostat
s
1
s
2
s
3
and for which s
oct
is constant. The angle q
L
is zero for
deviatoric pure shear s
2
s
1
s
3
=2, and varies between
p=630
for axisymmetric extension s

1
s
2
and p=630
for axisymmetric compression s

3
s
2
. For an isotropic material
the remaining ve sectors (of 60
) are given by symmetry.

In terms of the three invariants just dened, the yield condition
and plastic potential are given by equations of the form
Fs; s
oct
; q
L
0 and Gs; s
oct
; q
L
0, respectively. Both will
generally be functions of additional parameters that keep track of
the history of inelastic deformation and are not displayed here.
Values of constitutive parameters will evolve with inelastic defor-
mation but those that enter the fault angle predictions are those
just prior to bifurcation.
The direction of the inelastic strain increments (in stress space)
coincides with the normal to the plastic potential surface
P
ij
vG=vs
ij
. The normal to the yield surface is Q
ij
vF=vs
ij
and is
not in the same direction as the normal to the plastic potential. For
geomaterials it is typically the case that P
kk
sQ
kk
or that F
s
sG
s
(where the subscript denotes partial differentiation with respect to
s
oct
). Geometrically, in a space where the principal stresses are
taken as axes, the normal to the yield surface differs from the
direction of inelastic strain increments in planes that contain the
hydrostatic axis s
1
s
2
s
3
. Physically, this means that the ratio of
increments of s to s
oct
does not equal the ratio of the increments of
inelastic volume strain (positive in dilation) to equivalent inelastic
shear strain, dened by dg
p
2de
p
ij
de
p
ij
1=2
where de
p
ij
is the
deviatoric portion of the inelastic strain increment. The direction of
inelastic strain increments is taken to be normal to the projection of
the yield surface in deviatoric planes (planes normal to the
hydrostatic axis) because there are neither observational evidence
nor physical grounds indicating otherwise. As a result, the devia-
toric parts of the normals to the yield surface and plastic potential
are identical,P
0
ij
Q
0
ij
, where the prime denotes the deviatoric part,
and G
s
F
s
and G
q
L
F
q
L
.
With these denitions and assumptions the fault angle is given
by
q p=4 1=2arcsin a (10)
where
a
1=31 nP
kk
Q
kk
Q
0
2
1 2n
4
_
Q
0
ij
Q
0
ij
_
3
_
Q
0
2
_
2
_ (11)
v is Poissons ratio and Q
0
2
is the intermediate principal deviatoric
value of Q
ij
Eq. (11) can be expressed in terms of derivatives of the
yield function and plastic potential as
a
1 ncos jF
s
G
s
=F
s
1 2nsinj q
L
=
3
p
cosj q
L
(12)
by using the following relations:
50
55
60
65
70
75
80
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
2
(MPa)
3
= 10 MPa
25 MPa
40 MPa
60 MPa
100 MPa
core I
50
55
60
65
70
75
80
0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
2
(MPa)
3
= 10 MPa
40 MPa
60 MPa
100 MPa
Core II
a b
Fig. 7. Measured fault angle (q) variation with s
2
for different magnitudes of constant s
3
(a) for core I, and (b) for core II.
50 100 150 200 250 300 350
52
56
60
64
68
72
76
core I
10 MPa
25 MPa
60 MPa
40 MPa
100 MPa
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
Mean Compressive Stress,
0ct
(MPa)
core II
10 MPa
40 MPa
60 MPa
100 MPa
Fig. 8. Measured fault angle (q) variation with s
oct
for different magnitudes of constant
s
3
(different symbols) for core I (open) and core II (solid).
P
kk
3G
s
; Q
kk
3F
s
; tan j
F
q
L
=s
F
s
; 2Q
0
ij
Q
0
ij
_
F
q
L
=s
_
2
F
s
2
(13)
and
Q
0
2
Q
0
ij
_
Q
0
ij
=2

2
3
p sinj q
L
(14)
Eq. (12) reduces to the Rudnicki and Olsson (1998) form of the
Rudnicki and Rice (1975) expression by setting j 0, noting that
F
s
=F
s
3m, G
s
=F
s
3b and sin q
L

3
p
=2N, where
N s
2
=s is the deviatoric stress state parameter used by Rudnicki
and Rice (1975) (again, the minus sign occurs here because stresses
are positive in compression).
To use Eq. (12) to make predictions for comparisonwith data it is
necessary to adopt a specic form for the yield condition. Rudnicki
(2008a) showed that although limited agreement with data on
Westerly granite could be achieved with a two invariant constitu-
tive model (no dependence on q
L
), better agreement was possible
with a yield condition of the following form:
4
27
_
A sin3q
L
_
s
s
0
_
3
_
s
s
0
_
2
1 0 (15)
where 0 A 1. If A 0, then the shape of the yield surface in the
deviatoric plane reduces to a circle with the radius determined by
s
0
s
oct
. When A 1, the shape reduces to a triangular Rankine
type model, in which yield is determined by a critical value of the
least compressive stress. Rudnicki (2008b) has discussed how
assuming that s
0
is proportional to s
oct
and particular choices for
the constant of proportionality and A duplicate the forms suggested
by Lade and Duncan (1975) and Matsuoka and Nakai (1974) (also
described in Borja et al., 2003).
In general, A, in addition to s
0
, could depend on the octahedral
normal stress, s
oct
, but we use a constant value of 0.7 here. Limited
experimentation indicated no strong dependence on A (as long as it
was not near its limits) but we made no attempt to optimize the
choice to agree with the data.
Because the rst termin Eq. (15) vanishes for q
L
0, s
0
(s
oct
) gives
the mean stress dependence of the yield stress in deviatoric pure
shear s
2
s
1
s
3
=2. Evaluating the third of Eqs. (13) for Eq.
(15) yields an expression for tan j and the ratio F
s
=sF
s
is given by
52
56
60
64
68
72
76
80
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
2 sin(
L
) 2 sin(
L
)
10 MPa
25 MPa
40 MPa
60 MPa
100 MPa
core I
-0.8 -0.4 0.0 0.4 0.8 1.2 -1.2 -0.8 -0.4 0.0 0.4 0.8 1.2
52
56
60
64
68
72
76
10 MPa
40 MPa
60 MPa
100 MPa
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
Core II
a
b
Fig. 9. Measured fault angle (q) variation with deviatoric stress state, from axisymmetric extension (1, left) to axisymmetric compression (1, right) for different magnitudes of
constant s
3
(different symbols) for core I (a) and core II (b).
50 100 150 200 250 300
60
64
68
72
76
Axisym Comp
(
L
= /6))
76.7 - 0.089 * x
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
F
a
u
l
t

a
n
g
l
e

(
d
e
g
)
oct
(MPa)
Core I
Dev Pure Shear
(
L
= 0)
79.7 - 0.051 * x
50 100 150 200 250 300 350
52
56
60
64
68
72
76
80
Dev Pure Shear
L
= 0
84.2 - 0.087*x
Axisym Com
L
= /6
84.6 - 0.160*x
Core II
oct
(MPa)
a
b
Fig. 10. Fault angle (q) variation with s
oct
for axisymmetric compression (q
L
p/6) and deviatoric pure shear (q
L
0) for core I (a) and core II (b).
F
s
sF
s
s
s
0
ds
0
s
oct
ds
oct
(16)
because s s
0
for pure shear q
L
0, Eq. (16) gives a friction coef-
cient (which will be different for other deviatoric stress states, e.g.,
axisymmetric compression or extension). A similar ratio could be
inferred from the data shown in Fig. 4, but would pertain to the
slope of the failure surface rather than to the yield function. The
failure surface can be quite different from yield surface (even
evaluated at failure) (see, e.g., Holcomb and Rudnicki (2001) or
Besue lle and Rudnicki (2004)).
A consequence of the requirement that normality be satised in
the deviatoric plane (see end of paragraph preceding Eq. (10)) is
that the yield function and plastic potential can differ only by
a function of s s
oct
that we denote by H(s). A dilatancy factor b,
the ratio of the inelastic increment of volume strain to inelastic
increment of shear strain (dened earlier) is given by
b
3
2
1
1 A
2
=3
_
_
2
ds
0
ds
oct
s
0
H
0
s
oct
_
(17)
As for Eq. (16), this expression applies for pure shear and would be
slightly different for other deviatoric stress states.
7. Application to TCDP data
Because the yield surface and predictions for the band angle are
expressed in terms of invariants of the stress rather than the
principal stresses themselves, we plot the observed fault angle
versus the mean compressive stress s
oct
(in Fig. 8) and the Lode
angle q
L
(in Fig. 9). If both the mean stress and Lode angle are
known then the value of s is determined by the yield condition. In
Fig. 8, both core I and core II showan approximately linear increase
in fault angle with s
oct
for xed values of least compressive stress,
but an overall decrease in band angle with increasing mean stress.
Figs. 9a (core I) and Fig. 9b (core II) showthe fault dip angle plotted
against 2 sin q
L
which varies from 1 on the left for axisymmetric
extension, through 0 for deviatoric pure shear, to 1 for axisym-
metric compression on the right. Data from both cores appear to
show a slightly decreasing dip angle for increasing Lode angle, but
this is more evident for core II.
As already noted, the tests were conducted for xed values of
the least compressive stress. Consequently, neither s
oct
nor q
L
is
constant in the tests and it is difcult to infer the dependence on
either from Figs. 8 and 9. Nevertheless, within both data sets are
several tests for axisymmetric compression s
2
s
3
; q
L
p=6
and near pure shear s
2
s
1
s
3
=2; q
L
0. Here near
means j2 sin q
L
j 0:05 for core I and 0.08 for core II. Fig. 10a (core I)
and Fig. 10b (core II) show the fault angles for axisymmetric
compression and pure shear against s
oct
and linear ts through the
data for each deviatoric stress state. All the slopes are negative, but
the magnitudes are greater for core I than for core II. For each core
the magnitude of the slope is larger for axisymmetric compression
than for deviatoric pure shear, by nearly a factor of 2 for core II.
Substituting the linear relations shown in Fig. 10 into Eq. (12),
using Eq. (15), and evaluating for pure shear (q
L
0) and axisym-
metric compression (q
L
p/6) yields two linear equations for the
unknown functions of s: ds
0
=ds (Eq. (16)) and H
0
s (Eq. (17)).
Fig. 11 shows the solutions for the two data sets. For both the
variation is greater for core II than for core I.
Values of ds
0
=ds, which has the interpretation of a friction
coefcient, are reasonable, though, perhaps, on the low side, for
both core I and core II. The slight increase with mean stress is,
however, surprising as friction coefcient nearly always decreases
with increasing mean stress. For comparison, a friction coefcient
-50 50 1001 50 200 250 300 350 400
-1.2
-0.8
-0.4
0.0
0.4
0.8
1.2
oct
(MPa)
d
0
/d
Dilatancy Factor
0
Fig. 11. Solutions for ds
0
=ds, Eq. (16), and the dilatancy factor b, Eq. (17), as functions
of s
oct
based on linear ts in Fig. 10. Solid lines show results for core I and dashed for
core II.
60
64
68
72
76
10 MPa
25 MPa
40 MPa
60 MPa
100 MPa
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
Mean Compressive Stress (MPa)
Core I
50 100 150 200 250 300 350 50 100 150 200 250 300
50
55
60
65
70
75
80
10 Mpa
40 MPa
60 MPa
100 MPa
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
Core II
a
b
Fig. 12. Comparison of predicted (open symbols) with observed (lled symbols) fault angles plotted against mean compressive stress (s
oct
) for core I (a) and core II (b).
calculated as the slope of the curve for core I in Fig. 4 decreases from
0.66 at 50 MPa to 0.30 at 350 MPa. That for core II decreases from
0.57 to 0.25 over the same range of compressive stress. The increase
shown in Fig. 11 is, however, not large and given the idealizations of
the model and variation in the data, it has questionable
signicance.
The values for the dilatancy factor decrease substantially with
mean stress. This is as expected but the magnitude of the values at
lowand high conning stresses are surprisingly large. The values at
low mean stress are around 1 and indicate very strong dilatancy.
Values for core I decrease to a small negative value, about 0.3 at
the highest mean stress. Although a decrease of dilatancy with
mean stress is reasonable, it is less likely that compaction would
occur, even at the highest mean stress, given the relatively low
porosity and the history of signicant shear. The very strong
compaction predicted for core II at high mean stress is not realistic.
Some aspects of the functions plotted in Fig. 11 are in accord
with expected behavior, but it seems difcult to assign them more
than qualitative signicance. It is, however, interesting that the
behavior inferred for the two cores is signicantly different. Despite
the questionable aspects of the plots in Fig. 11, they are based on the
observed decrease in band angle with mean stress for pure shear
and axisymmetric compression shown in Fig. 10 and will, of course,
reproduce this behavior. This is a signicant improvement over
constitutive models based on only two of the stress invariants. In
addition, the predicted band angle (see Eq. (12)) depends primarily
on the sum of the two functions. For this reason, it seems that the
slight increase with mean stress predicted for the friction coef-
cient is compensated by an excessive decrease in the dilatancy
factor. At this point, it is unclear what improvement might lead to
a more reasonable division.
The functions of mean stress shown in Fig. 11 are then re-
substituted into Eq. (12) using Eq. (15) and used to evaluate the
predicted band angles for all the Lode angles in the measured data.
Fig. 12a (core I) and Fig. 12b (core II) show the fault angle data and
predictions plotted against the mean compressive stress (with the
open symbols showing the predictions). Clearly, the agreement is
much better for core II (correlation 95%) than for core I (correlation
80%). Fig. 13 plots the same data against the Lode angle parameter
2 sin q
L
. Not surprisingly, for both the agreement is better for
positive values since the mean stress variation used the data for
pure shear q
L
0 and axisymmetric compression (q
L
p/6).
Figs. 14 and 15 show the predicted fault angles if tests were
conducted at xed values of mean stress or of Lode angle q
L
. Fig. 14
plots the fault angle against mean compressive stress for ve values
of q
L
corresponding to xed deviatoric stress states. As expected,
-0.8 -0.4 0.0 0.4 0.8 1.2
52
56
60
64
68
72
76
80
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
2 sin(
L
) 2 sin(
L
)
10 MPa
25 MPa
40 MPa
60 MPa
100 MPa
core I
-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2
50
55
60
65
70
75
80
10 Mpa
40 MPa
60 MPa
100 MPa
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
Core II
a
b
Fig. 13. Comparison of predicted (open symbols) with observed (lled symbols) fault angles plotted against deviatoric stress state 2 sin q
L
for core I (a) and core II (b).
-50 0 50 100 150 200 250 300 350 400
40
50
60
70
80
90
Axisym Comp (
L
= )
L
=
Pure Shear (
L
= )
L
=
Axisym Ext (
L
= )
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
Core I
-50 0 50 100 150 200 250 300 350 400
20
30
40
50
60
70
80
90
AxiSym Comp (
L
/6)
L
= /12
Pure Shear (
L
= 0)
L
= /12
AxiSym Ext (
L
/6)
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
Mean Compressive Stress(MPa)
Core II
a b
Fig. 14. Predicted variation of the fault angle against mean compressive stress (s
oct
) for constant deviatoric stress states 2 sin q
L
the fault angles decrease with increasing mean stress. The straight
lines for axisymmetric compression and pure shear reect the use
of these data in the tting. For both cores dilation bands (perpen-
dicular to least compressive stress; extension fractures) are pre-
dicted for axisymmetric extension at the lowest mean normal
stresses, though stress states corresponding to q <q
L
are not well
populated by the data. For axisymmetric compression at the
highest mean stresses the angles extend down to slightly less than
50
for core I. For core II they extend down to 30
, suggesting the
possibility of compaction band formation at mean stresses some-
what higher than achieved in the tests. This prediction is also
consistent with the strong compaction inferred for core II at higher
mean stresses (Fig. 11b).
Fig. 15 plots the fault angle against the deviatoric stress state
parameter for ve constant values of the mean compressive stress.
The band angle decreases as the deviatoric stress state varies from
axisymmetric extension (left side) to axisymmetric compression
(right side). This trend is consistent with the predictions of the
simpler constitutive relation used by Rudnicki and Rice (1975)
although the decrease here is not so large. Because the fault angle
also decreases with mean compressive stress, the variations with
mean stress and deviatoric stress states can offset or augment each
other. The results for core II predict a dilation band for the lowest
mean stress. Although neither result shows a compaction band, the
fault angle does decrease rapidly approaching axisymmetric
compression for the highest mean stress values shown. Rudnicki
(2004) has suggested this rapid decrease as a possible reason for
the infrequent observation of dip angles in this range.
8. Conclusions
True triaxial tests on two siltstone cores from the TCDP hole A
reveal that the intermediate principal stress is a signicant
contributor to their compressive strength, and bring into question
the suitability of the Mohr and MohrCoulomb strength criteria,
which neglect the effect of s
2
. Rather, strength criteria in terms of
the invariants octahedral shear stress and the 2-D mean stress in
the s
1
s
3
plane t well the experimental data.
The angle of the fault created upon brittle fracture, is also
strongly affected by s
2
. It is found that the angle rises with s
2
for
constant s
3
, further questioning the adequacy of the Mohr-type
criteria, which predict a fault angle independent of s
2
.
The variation of the fault angle with mean compressive stress
and deviatoric stress state is modeled using localization theory
with a three invariant form for the yield function and plastic
potential. Calibration of the results using subsets of the data for
deviatoric pure shear and axisymmetric compression yields two
inelastic properties that are functions of the mean stress. Incor-
porating these inferred functions, the predictions agree well with
the entire data set for core II and acceptably with that for core I. The
results are then used to predict the variation for the band angle for
true triaxial tests conducted at constant mean stress and xed
deviatoric stress state. The fault angle at constant mean normal
stress is predicted to decrease as the deviatoric stress state varies
from axisymmetric extension to axisymmetric compression. The
fault angle at xed deviatoric stress state is predicted to decrease
monotonically with increasing mean normal stress.
Although the inferred inelastic constitutive functions are
reasonable, there is no independent verication of their form. Nor
are there true triaxial tests available at constant mean stresses or
deviatoric stress state. Nevertheless the variation of observed fault
angle is clearly not well-described by the simple MohrCoulomb
type theory. Comparison with the more elaborate theory, despite
the uncertainties involved, yields predictions that could be evalu-
ated by further testing and offers insight into the inelastic consti-
tutive behavior of the rock and its relation to failure.
Acknowledgements
Partial nancial support for JWR was provided by the US Dept. of
Energy, Ofce of Science, Basic Energy Sciences, Geosciences
Program through grant DE-FG02-93ER14344/A016 to North-
western University. Partial nancial support for BHwas provided by
NSF grant EAR-0346141. Thanks are extended to H. Oku for carrying
out the laboratory experiments and to Florent Gimbert for many
helpful discussions about three invariant constitutive relations.
Core fromthe TCDP was obtained courtesy of Professor Sheng-Rong
Song, National Taiwan University. The authors also thank Philip
Benson and an anonymous reviewer for comments that improved
the paper.
References
Besue lle, P., Rudnicki, J.W., 2004. Localization: shear bands and compaction bands.
In: Gue guen, Y., Boute ca, M. (Eds.), Mechanics of Fluid Saturated Rocks. Inter-
national Geophysics Series, 89. Academic Press, London, pp. 219321.
Borja, R.I., Sama, K.M., Sanz, P.F., 2003. On the numerical integration of three-
invariant elastoplastic constitutive models. Computer Methods in Applied
Mechanics and Engineering 192, 12271258.
-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2
30
40
50
60
70
80
90
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
2 sin(
L
)
Core I
50 MPa
150 MPa
250 MPa
350 MPa
450 MPa
-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2
0
10
20
30
40
50
60
70
80
90
F
a
u
l
t

A
n
g
l
e

(
d
e
g
)
2 sin(
L
)
Core II
50 MPa
150 MPa
250 MPa
350 MPa
450 MPa
a b
Fig. 15. Predicted variation of the fault angle against deviatoric stress state 2 sin q
L
at ve constant values of s
oct
tory experiments. Journal of Geophysical Research 85 (B11), 62486252.
Chang, C., Haimson, B.C., 2000. True triaxial strength and deformability of the KTB
deep hole amphibolite. Journal of Geophysical Research 105, 1899919014.
Chang, C., Haimson, B.C., 2005. Nondilatant deformation and failure mechanism in
two Long Valley Caldera rocks under true triaxial compression. International
Journal Rock Mechanics and Mining Sciences 42, 402414.
Chang, C., Haimson, B., 2007. Effect of uid pressure on rock compressive failure in
a nearly impermeable crystalline rock: implication on mechanism of borehole
breakouts. Engineering Geology 89, 230242.
Hadamard, J., 1903. Leons sur la Propagation de Ondes et Les Eqs. de LHy-
drodynamique, Paris.
Haied, A., Kondo, D., Henry, J.P., 2000. Strain localization in Fontainebleau sand-
stone. Mechanics of CohesiveFrictional Materials 5, 239253.
Haimson, B., Chang, C., 2000. A new true triaxial cell for testing mechanical prop-
erties of rock, and its use to determine rock strength and deformability of
Westerly granite. International Journal of Rock Mechanics and Mining Sciences
17, 285296.
Haimson, B., Chang, C., 2002. True triaxial strength of the KTB amphibolite under
borehole wall conditions and its use to estimate the maximum horizontal in
situ stress. Journal of Geophysical Research 107 (B10) ETG 15-1 to 14.
Handin, J., Heard, H.C., Magouirk, J.N., 1967. Effect of the intermediate principal
stress on the failure of limestone, dolomite, and glass at different temperature
and strain rate. Journal of Geophysical Research 72, 611640.
Hill, R., 1962. Acceleration waves in solids. Journal of the Mechanics and Physics of
Solids 19, 116.
Holcomb, D.J., Rudnicki, J.W., 2001. Inelastic constitutive properties and shear
localization in Tennessee marble. International Journal for Numerical and
Analytical Methods in Geomechanics 25, 109129.
Jaeger, J.C., Cook, N.G.W., Zimmerman, R., 2007. Fundamentals of Rock Mechanics,
fourth ed. Blackwell Publishers, 475 pp.
Lade, P.V., Duncan, J.M., 1975. Elasto-plastic stressstrain theory for cohesionless
soil. Journal of the Geotechnical Engineering Division American Society of Civil
Engineers 101 (GT10), 10371053.
Lin, C.-W., Lee, Y.-L., Huang, M.-L., Lai, W.-C., Yuan, B.-D., Huang, C.-Y., 2003. Char-
acteristics of surface ruptures associated with the Chi-Chi earthquake of
September 21, 1999. Engineering Geology 71, 1330.
Mandel, J., 1966. Conditions de stabilte et postulat de Drucker. In:
Kravtchenko, J., Sirieys, P.M. (Eds.), Rheology and Soil Mechanics. Springer
Verlag, pp. 5868.
Matsuoka, H., Nakai, T., 1974. Stressdeformation and strength characteristics of soil
under three different principal stresses. Proceedings of Japan Society of Civil
Engineers 232, 5970.
McGarr, A., Gay, N.C., 1978. State of stress in the earths crust. Annual Review of
Earth and Planetary Sciences 6, 405436.
Mogi, K., 1967. Effect of the intermediate principal stress on rock failure. Journal of
Mogi, K., 1971. Fracture and owof rocks under high triaxial compression. Journal of
Murrell, S.A.F., 1963. A criterion for brittle fracture of rocks and concrete under
triaxial stress, and the effect of pore pressure on the criterion. In: Fairhurst, C.
(Ed.), Proceedings of the Fifth Symposium on Rock Mechanics. Pergamon Press,
pp. 563577.
Nadai, A., 1950. Theory of Flowand Fracture of Solids, vol. 1. McGraw-Hill, New York.
Oku, H., Haimson, B., Song, S.R., 2007. True triaxial strength and deformability of the
siltstone overlying the Chelungpu fault (Chi-Chi earthquake), Taiwan.
Ottosen, N.S., Runesson, K., 1991. Properties of discontinuous bifurcation solutions
in elasto-plasticity. International Journal of Solids and Structures 27, 401421.
Perrin, G., Leblond, J.B., 1993. Rudnicki and Rices analysis of strain localization
revisited. Journal of Applied Mechanics 60, 842846.
Rice, J.R., 1976. The localization of plastic deformation. In: Koiter, W.T. (Ed.), Theo-
retical and Applied Mechanics, Proceedings of the 14th International Congress
on Theoretical and Applied Mechanics. North-Holland Publishing Company,
Delft, The Netherlands, pp. 207220.
Rudnicki, J.W., 2004. Shear and compaction band formation on an elliptic yield cap.
Rudnicki, J.W., Olsson W.A., 1998. Reexamination of fault angles predicted by shear
localization theory. In: Proceedings of Third North American Rock Mechanics
Symposium (NARMS98), Rock Mechanics in Mining, Petroleum and Civil
Works, 35 June, 1998, Cancun, Mexico. Extended Abstract in International
Journal of Rock Mechanics and Mining Sciences 35(415), 512513.
Rudnicki, J.W., Rice, J.R., 1975. Conditions for the localization of deformation in
pressure-sensitive dilatant materials. Journal of the Mechanics and Physics of
Solids 23, 371394.
Rudnicki, J.W., 2008a. Localized failure in brittle rock. In: Shao, J.F., Burlion, N. (Eds.),
Thermo-Hydromechanical and Chemical Coupling in Geomaterials and Appli-
cations, Proceedings of Third International Symposium GeoProc2008. Wiley,
pp. 2540.
Rudnicki, J.W., 2008b. Failure of Brittle Rock in the Laboratory and in the Earth, To
Appear in Proceedings of XXII International Congress on Theoretical and
Applied Mechanics, Adelaide, Australia, 2430 August.
Shin, T.-C., Teng, T.-L., 2001. An overview of the 1999 Chi-Chi, Taiwan, earthquake.
Bulletin of the Seismological Society of America 91, 895913.
Thomas, T.Y., 1961. Plastic Flow and Fracture in Solids. Academic Press.
Von Ka rma n, T., 1911. Festigkeitsversuche Unter all Seitigem Druck. Z. Verin Deut.,
Ingr. 55, 17491759.
Wiebols, G.A., Cook, N.G.W., 1968. An energy criterion for the strength of rock in
polyaxial compression. International Journal of Rock Mechanics and Mining
Sciences 5, 529549.
Porosity and particle shape changes leading to shear localization
in small-displacement faults
Jafar Hadizadeh
a,
*
, Reza Sehhati
b, c
, Terry Tullis
d
a
Department of Geography and Geosciences, University of Louisville, Louisville, KY, USA
b
Department of Civil and Environmental Engineering, Washington State University, Pullman, WA, USA
c
Berger/ABAM Engineering Inc., Federal Way, WA 98003, USA
d
Department of Geological Sciences, Brown University, Providence, RI, USA
Article history:
17 August 2010
Accepted 23 September 2010
Available online 1 October 2010
Keywords:
Shear localization
Gouge porosity
Particle shape and size
Particle size distribution
a b s t r a c t
A microstructural study of shear localization in fault gouge was carried out in small-displacement faults
so there would be minimum masking effects from a complex deformation history. We studied particle
size, shape, and porosity changes in gouge adjacent to zones of shear localization in natural and synthetic
gouges subjected to shear displacements d, of up to 1.2 m. Scanning electron microscope images were
used for estimating image porosity F
I
, and measuring particle size of the deformed and undeformed
gouges. The particle size data were used for calculating simulated porosity F
S
from computer-generated
simple fractal gouge model of each sample. Modeled microstructures contained round grains and
a fractal distribution matched to that of the measured natural samples. Changes in F
I
, F
S
, and F
I
/F
S
with
increasing d were used for tracking changes in particle shape and porosity of the gouges precursory to
shear localization. The F
I
and F
S
values for the natural and synthetic gouges converge at d w 0.1 m,
suggesting that gouge particles adjacent to shear localization sites tend to become rounded. Porosity for
such densied regions of the gouge adjacent to Y-shear zones was determined to be <1% at large
displacements. In the same regions, the porosity reductions were also associated with decreased sorting
coefcient and fractal dimensions D > 2.6. The study suggests that brittle shear localization may involve
favorably-oriented micro porous pockets of gouge that result from competing changes in particle shape
and particle size, which tend to affect gouge porosity in different ways.
1. Introduction
Brittle shear localization microstructures such as slip surfaces,
shear bands, and cataclastic foliation are commonly found within
the core of many natural fault zones. The localization process
results in mechanical weakening of the fault and is often associated
with the most comminuted and densied regions in fault gouge
(e.g. Evans and Chester, 1995; Chester and Chester, 1998; Boullier
et al., 2004; Hayman, 2006; Rawling and Goodwin, 2006;
Rockwell and Ben-Zion, 2007; Tanaka et al., 2007; Brogi, 2008).
Shear localization has been observed in relatively unaltered small-
displacement faults with displacements typically <1 m as in most
experimental faults as well as in mature natural fault zones
involving a variety of alteration products. Experimental fault gouge
studies have shown that 0.1 m of shear displacement results in
shear localization, although microstructural changes leading to
shear localization are not well understood. A number of studies of
natural, experimental and computer simulated gouge deformation
conclude that shear localization is primarily a particle size and
particle-size distribution driven process (Dieterich, 1981; Marone
and Scholz, 1989; Biegel et al., 1989; Logan et al., 1992; Gu and
Wong, 1994; Billi, 2007; Keulen et al., 2007; Sammis and Ben-
Zion, 2008). Experimental studies by Mandl et al. (1977),
Vardoulakis (1980), Marone and Scholz (1989), Mair and Marone
(1999), and Mair et al. (2002) indicate that gouge attains a critical
strain or particle size distribution prior to shear localization. Mandl
et al. (1977) based on experimental data, suggested that shear
localizes in favorably-oriented bands of gouge that have achieved
a critical PSD through conned comminution. This is possible
because an increase in the proportion of ne particles in gouge,
while not reducing cohesive forces, might reduce the friction.
Subsequently, at a lowthreshold value of internal friction there will
be a drastic reduction in boundary shear by development of a slip
plane and growth of a shear zone. In the powder industry this
E-mail address: hadizadeh@louisville.edu (J. Hadizadeh).
doi:10.1016/j.jsg.2010.09.010
phenomenon is attributed to rounding and size equalizing of
particles which tend to reduce interlocking resistance of a granular
material (Lowrison, 1974). The microstructural aspects of the model
described above draws support from a number of observations.
Gouge deformation simulations by Morgan and Boettcher (1999)
showed that a sharp drop in sliding contacts accompanied local-
ized failure as fewer particles were involved in the zone of defor-
mation. Mair and Marone (1999) studied the controlling effect of
particle size on shear localization and noted that PSD evolution in
ne and coarse gouge differed only by a shear strain g of 3.4. They
suggested that higher fracture toughness might have inhibited
further comminution of ne gouge as its PSD became more
uniform. Scarpelli and Wood (1982), Moore et al. (1989), Logan
et al. (1992) reported the same sequence of microstructural
development preceding shear localization in calcite and halite
gouges. In a model presented by Shipton and Cowie (2001) a critical
amount of comminution or strain was necessary for slip surface
nucleation, but the slip surfaces accommodated further strain
without appreciable amount of comminution. The experimental
study of Marone and Scholz (1989) concluded that transition from
pervasive to localized shear occurs at a critical strain or PSD in
addition to the inuence of gouge density.
This study attempts to provide a microstructural model for shear
localization by investigating the combined effect of particle shape
and particle size on porosity changes that precede shear localiza-
tion in small-displacement natural and experimental faults. The
effect of porosity changes on shear localization in granular material
with varied particle size and PSD has been studied previously. The
work of Mead (1925) and Frank (1965) shows that since in conned
comminution dilatancy is suppressed, shear localization must
occur in regions of gouge with least dilatancy rate. Marone and
Scholz (1989) reported dilatant behavior at the onset of shear
localization in their experimental quartz gouge. The actual shear
localization occurred on R
1
Riedel shear bands, and the dilatancy
was believed to be the result of unpacking of over consolidated
gouge. Marone and Scholz (1989) noted that particles within their
Fig. 1. Natural and synthetic gouges used in the study. (a) Typical small-displacement splay fault in the Aztec sandstone. Outcrop picture is labeled with the apparent displacement
vector (d
a
), and trend of the slickenside lineation on the fault plane (dotted line). Compaction bands serve as displacement markers. Typical undeformed texture of the sandstone is
shown on the right. (b) Rotary shear sample ring consisting of a 2 mm gouge layer (not to scale). The close-up view of the gouge layer on the left is an actual section across
undeformed simulated Westerly granite gouge compacted to 25 MPa pressure. The gouge is held between granite forcing blocks g, forming the shear zone boundaries; Q quartz;
K potassium feldspar; P plagioclase feldspar; lightest shade particles are phyllosilicates. Textures are back-scattered SEM images.
J. Hadizadeh et al. / Journal of Structural Geology 32 (2010) 1712e1720 1713
experimental shear bands did not follow a fractal size distribution,
but the relationship between the shear band microstructures and
the unpacking process was not discussed.
2. The studied fault gouges and methods of data collection
2.1. The natural gouge
Gouge samples were collected from splay faults of the Lonewolf
fault zone in the Valley of Fire State Park (VOF), Nevada, U.S.A. The
splay faults are believed to be sheared joints, the coalescence of
which lead to the development of the Lonewolf fault zone (Flodin
and Aydin, 2004; Myers and Aydin, 2004; Davatzes et al., 2003).
The splay faults ranging in measurable displacement fromw0.01 m
to 1.5 m were found in the upper domain of the Aztec sandstone
Formation described in detail by Flodin et al. (2003). Due to soft
iron oxide/clay mineral cementation of the sandstone and weath-
ering effects, in-situ impregnation by clear resin was necessary
before the samples could be lifted from the desired locations along
the splay faults. The sampling method thus preserved the gouge
microstructures. The Aztec sandstone in the sampling area is an
aeolian feldspathic quartz arenite made of rounded to well-
rounded quartz grains forming 95% of the rock with average
undeformed grain size of 230 mm. The measured porosity (Flodin
et al., 2003 via Helium porosimetry) ranged from 21.2% to 23.8%
with the presence of a small amount of pore-lling quartz or iron
oxide cement. Our sampling area corresponds to porosity samples
54e56 in Table 2, Flodin et al. (2003).
The strike-slip and normal faulting in the VOF area had taken
place during the Miocene Basin and Range tectonic activity. It has
been suggested that in the initial phase of the activity the Aztec
Formation had been buried by at least 1.6 km of sediments
(Bohannon, 1983), andpossibly byanadditional 1e4kmof overlying
Sevier-related thrust sheet (Brock and Engelder, 1977). Based on the
overburden thickness estimates and studies by Flodin and Aydin
(2004) the possible range of overburden pressure at the time of
faulting could be 10e40 MPa assuming an overburden density of
2700 kg/m
3
. The effective pressure for the VOF gouges might have
been closer to the middle of the range because for approximately
same shear strains the overall degree of cataclasis in the VOF shear
bands was similar to those observed in our experimental shear
bands deformed at 25 MPa. We found no optical evidence of crystal
plastic deformation in quartz grains of the sampled gouge, which
suggested temperatures <200

C for the observed deformation.
Compaction bands that formed prior to faulting in the Aztec
sandstone (Flodin et al., 2003) were offset by the studied splay faults
and provided excellent displacement markers (Fig. 1a). The measured
apparent displacement d
a
(the band offset), together with the pitch R
of slickenside lineation on the fault plane was used to determine true
displacement d, as d
a
/cos R. This relationship holds for the studied
faults with a strike-slip displacement component since the displace-
ment markers (compaction bands) were oriented vertical, or near
vertical (see Fig. 1a). The apparent displacement varied along the
strike, but the faults were sampled exactly where the structural
measurements were made. The true displacements thus calculated
were17mm, 116mm, 348mmand1219mmfor thefour samples used
inthis study. The gouge zone thicknesses measured inthinsections at
right angles to the shear zone borders ranged from1.2 0.08 mm to
15.5 5 mm, positively correlating with the displacement values. A
summary of the sample data is presented in Table 1.
2.2. The synthetic gouge
Synthetic gouge samples (WGK) with particle size 88 mm
consisting of 28% quartz, 35% microcline, 32% plagioclase, 5% mica,
and <1% opaque were prepared by grinding Westerly granite. The
study uses deformed synthetic gouge froma previous experimental
study. The gouge had been deformed at room temperature and
25 MPa normal stress in a rotary shear apparatus. The sliding
velocity was stepped between 1 mm/s (for 1 mm distance), and
10 mm/s (for 10 mmdistance) in all experiments. Technical specics
of the apparatus are described elsewhere (Tullis and Weeks, 1986;
Beeler et al., 1996). For each experiment approximately 1 g of the
material was packed in a ring-shaped sample holder forming
a w2 mm layer of gouge with w35% initial porosity (Fig. 1b). A
compaction test showed that raising normal stress to 25 MPa at the
beginning of each experiment reduces the gouge layer thickness by
about 5.5% (w110 mm). The compaction run also provided refer-
ences for the initial texture, PSD, and porosity. The gouge layers
were deformed to 44 mm, 79 mm, and 387 mm of shear
displacement. While the simulated gouge was almost entirely
velocity weakening up to the largest displacements, the effect of
sliding rate on shear localization was not tested.
2.3. Particle size measurements
The thin section areas selected for imaging, and the subsequent
particle size and porosity measurements, were thoroughly inspec-
ted for surface damage and plucked grains. The particle size
measurements other than for characterizing the undeformed
materials were made on areas of gouge adjacent to zones of shear
localization. It was assumed that the intensely deformed gouge
within 500 mm distance either side of Y-shear zones reect the
particle shape, particle size distribution, and porosity that existed at
the onset of shear localization regardless of shear strain in the
samples. It has been shown that shear localization as a result of
comminutionoccurs during the early increments (g
0
s 1e10) of shear
strain (Marone and Scholz, 1989; Gu and Wong, 1994; Mair and
Marone, 1999; Wolf et al., 2003). The minimum shear strain in
VOF and WGK gouges were 14 and 22 respectively. Furthermore, we
assume that as the gouge approaches the shear localization stage
the regions around potential shear zone deform more intensely
thaninthe bulk gouge. This maybe so because shear localizes where
gouge is weakening. Thus unlike the bulk gouge, microstructures
near the localized shear zone are expected to record the process of
weakening by developing a somewhat different set of textural
attributes. We avoided selecting areas for microstructural
measurements where relatively undeformed gouge (e.g. coarse
gouge near the fault zone margins in Fig. 2a and b) was in contact
with shear bands. Such contacts might indicate the possibility of
shear localization due to some preexisting microstructural
inhomogeneity. The PSD data sets were used in computer simula-
tions, and for determining fractal dimension of the gouges. The
measurements were conducted on polished petrographic thin
sections cut perpendicular to the shear planes and viewed in back-
scattered mode in a Zeiss Supra-35VP scanning electron micro-
scope (SEM). Optical microscopy included transmitted and reected
Table 1
A summary of structural data for the studied splay faults of the Lonewolf fault zone
in the Aztec sandstone.
Sample Fault plane attitude and
sense of shear
d
a
, mm R
d, mm T, mm
VOF4A 284, 39NE; RL 16.5 17 17 1.2
VOF2A 340, 74SW; Rl 70 53 116 1.4
VOF01 342, 72NE; LL 90 75 348 3.6
VoF5A 015, 76NW; LL 1275 17 1219 15.5
d
a
apparent displacement; R
pitch of the slickenside lineation on the fault

plane; d true shear displacement; T average gouge layer thickness; RL and LL
respectively refer to right-lateral and left-lateral sense of shear ascertained from
compaction band offset and Reidel shear sets in the fault gouge.
light imaging. Outline of gouge particles were manually traced on
digital images from selected regions of deformed and undeformed
samples. The tracing was carried out with a 2 pixel digital tip at 3
screen magnication and transformed into calibrated line drawing
overlays using Adobe Photoshop. Gouge particle size S, repre-
sented by the equivalent circle diameter of particle image area A,
was measured on the overlays using relationship S 2(A/p)
.
Measurements were carried out using SigmaScan Pro and MAT-
LAB Image Processing Toolbox applications. To gain a wider
particle size range we pooled size data for each deformed gouge
sample from 2 (VOF samples) or 3 (WGK samples) sets of images
taken telescopically at increasing magnication. Asingle SEMimage
each of the undeformed VOF sandstone and WGK simulated gouge
was used for reference particle size measurements. In telescopic
imaging, the microscope magnicationwas varied by a factor of two
from 250 to 32000 depending on the particle size. Duplicate
particle sizes (same values to 3 decimal places) in two consecutive
images in each set were discarded. As mentioned earlier, particle
size measurement of the deformed gouge was based on selected
areas adjacent to Y-shear zones (Fig. 2). Particle size data for this
study consisted of 3469 measurements (particle outline traces)
acquired from 9 SEM images for the VOF samples and 5868
measurements acquired from10 SEMimages for the WGK samples.
The difference in number of measurements (including the unde-
formed materials) reects the smaller average particle size of the
WGK gouge that resulted in a larger number of particles per unit
area compared to that for the VOF gouge.
The size-number data was used for determining the fractal
dimension D, as the slope of the logelog size (S) e number (N)
distribution given by NS cS
D
, where c is a constant (Turcotte,
1986). We present D values here as the 3D, or the volume fractal
dimension by adding 1.0 to the calculated D values (Falconer, 1985).
The changes in PSD with increased shear displacement were
analyzed using the sorting coefcient Q [S
q1
/S
q3
]
, where S
q1
and
S
q3
are the rst quartile (25% of distribution >S
q1
) and the third
quartile (75% of the distribution >S
q3
) in a given cumulative size
Fig. 2. Typical areas of gouge used for particle size and porosity measurements. The dotted line labeled Y is trace of the Y-shear determined from a larger area of the sample than
shown here. Whole shear zones are shown on top with insets showing location of enlarged areas at the bottom. (a) Shear zone in the Aztec sandstone sample with 348 mm of
displacement (VOF01 in Table 1). (b) Synthetic Westerly granite gouge layer after 44 mm of displacement (sample WGK258).
distribution respectively (Krumbein and Sloss, 1951). Decreasing Q
values indicate reduced particle size range, improved sorting, and
increased porosity while increasing Q values indicate widening of
the size range, poor sorting, and decreased porosity (Rogers and
Head, 1961; Marone and Scholz, 1989). The gouge porosity here
referred to as 2D image porosity F
I
, is the total porosity estimated
via digital analysis of the SEM images. Pore spaces as conrmed by
secondary electron images appeared as dark featureless areas
between particles on backscattered SEM images (see Fig. 1). Thus
F
I
[SA
(pore space)
100]/A
(image)
, where A is the measured areas on
the image. The image processing method and its validation with
respect to volume porosity in actual rock material has been dis-
cussed elsewhere (e.g. Antonellini et al., 1994; Anselmatti et al.,
1998; Solymar and Fabricius, 1999; Talukdar et al., 2002;
Johansen et al., 2005). The sources of error in area measurements
included image quality and magnication, and the thresholding
process. Thresholding refers to a digital manipulation of the image
that involves differentiating a certain range of pixel intensities from
the full image pixel intensity distribution. For our purpose, the pore
space pixel range was isolated by thresholding binarized images
using SigmaScan Pro application. The thresholding average
values were 0.1% and 2% of the total measured pore areas at the
highest (64000) and the lowest (250) image magnications
respectively. After thresholding, but prior to the measurements,
pore-like pixel areas within particles were removed from the
threshold copy of the image. Since the particle-pore borders con-
sisted of only a zone of 2e5 pixels, an average of the highest and
lowest pore space area was used in individual images. The image
porosity was then estimated by averaging F
I
values fromindividual
images in a telescopic series that represented each sample. A 3-
image telescopic set was used for the highest displacement samples
(WGK262 and VOF5A) while a 2-image telescopic set, taken at
250and 1K, was used in all other samples; 9 (VOF) and 7 (WGK)
SEM images were used.
2.4. Porosity from computer-generated models
A computer program for generating simple fractal gouge (SFG)
models was written in VC. The basic objective of the simulations
was to obtain reasonable estimates for the limiting values of
porosity as gouge particles become more rounded with increased
comminution. The program was not written to simulate realistic
gouge microstructures; it was intended for providing estimates of
porosity in a hypothetical gouge with spherical particles and fractal
particle size distribution. The F
S
values, generated based on our real
gouge PSD data, were an approximation which we considered more
accurate than a hypothetical PSD. The nature of the PSD approxi-
mation was that we expected a deviation to occur from the so-
called steady-state fractal size distribution (D 2.6) within the
shear bands. The approximation is reasonable since the PSD data
used for the simulations were collected from selected areas adja-
cent to the shear bands.
The algorithmtakes as input a PSD data set acquired fromgouge
images as described earlier. The size and total number of particles
remained unchanged from the original distribution, and the
particles were simulated as perfect circles. As a fractal distribution,
the particles for each simulated texture were assembled such that
least number of same size particles touched. This was achieved by
packing particles according to tangent circle solutions for neigh-
boring particles. The program produced maximum packing density
for the given data set under the described conditions. The packing
density of the SFG models could exceed the maximum packing
density achievable with uniform size circles in 2D space given by
p/O12 (Hecht, 2004). The algorithm was capable of processing
an unlimited number of particles with unlimited size range in
a descending order. The 2D porosity of the SFG was calculated by
summing up the void space areas within the perimeter of the entire
simulated mass. Although in this study only 2D results are reported
we note that the 3D porosity tends to be higher than the 2D
Fig. 3. Examples of the simple fractal gouge (SFG) model based on particle size data from gouge areas adjacent to Y-shear zones. Models are for (a) Aztec sandstone gouge area
shown in Fig. 2a, N 817 particles, D 3.306, and (b) Synthetic gouge area shown in Fig. 2b, N 1386 particles, D 3.284. Scale bars on images are approximately true for the
models.
porosity for the same packing arrangement. The main source of
error in the simulation originated in the manual particle outline
tracing operations that provided the particle size data set. This was
in turn dependent upon magnication and image quality. The
programnumeric output included SFGporosity, F
S
, and a simulated
texture generated via a visualizer program (examples shown in
Fig. 3). We note that both F
S
and F
I
should be considered reason-
able proxies for the gouge porosity rather than actual porosity.
3. Results
The results are shown in plots of the Figs. 4e6. The measured
porosity of 21e23% in the undeformed Aztec sandstone is compa-
rable to its SFG model porosity of 28.55%. The difference is in part
due to non-fractal PSD of the undeformed sandstone and the
presence of pore-lling cement. In the deformed sandstone gouge,
F
I
and F
S
values rst diverge at d > 20 mm and then converge at
d > 300 mm (Fig. 4a). This transient deviation from continuous
porosity reduction is due to comminution of the round undeformed
particles into angular particles during initial increments of shear
displacement on the fault. The porosity of the sandstone gouge
drops with further comminution and rounding of the cataclastic
particles. For the synthetic Westerly granite gouge with initially
angular particles, F
S
and F
I
values have their largest difference in
the undeformed gouge, but the difference is reduced monotonously
with increasing displacement (Fig. 4b). In terms of particle shape,
Fig. 4c shows that conned comminution in the synthetic gouge
results in rounding of the particles as well as size reduction,
although an aggregate with initially round particles will do so by
rst transforming to an angular aggregate. Although in both gouge
types the porosity is signicantly reduced by comminution, and F
I
and F
S
values tend to converge at d > 0.1 m, the porosity reduction
in the synthetic gouge occurs at a higher rate with respect to
displacement presumably due to higher normal stresses. For
example, at dw360 mm F
I
is w0.45% and w1.7% for the synthetic
gouge and the sandstone gouge respectively. We note that as
expected, dF
S
/dd is nearly identical for both gouge types.
The evolution of porosity with shear displacement in the two
gouge types is compared by presenting the changes as F
I
/F
S
ratio in
C
b
a
Fig. 4. Changes in porosity with shear displacement of areas adjacent to Y-shear zones
in (a) The Aztec sandstone gouge and (b) Westerly granite synthetic gouge. The
porosity is represented by the 2D image porosity F
I
, and its corresponding SFG model
porosity F
S
. Undeformed gouge in both cases is assigned a nominal displacement value
of 1 mm (c). SEM backscattered image of simulated Westerly granite gouge texture
showing abundance of particles with round to sub rounded shapes bordering the main
Y-shear zone in the experiment with 387 mm shear displacement.
Fig. 5. Plot comparing change in F
I
/F
S
ratio with shear displacement in Aztec sand-
stone and Westerly granite synthetic gouges. The undeformed gouge is assigned
a nominal displacement value of 1 mm.
Fig. 5. The F
I
/F
S
for the two deformed gouge types converge with
increasing shear displacement while remaining >1, which indicates
a higher porosity for the gouges compared to their SFG models. The
plots in Fig. 5 conrm that lower porosities adjacent to zones of
shear localization result from rounding as well as ning of the
particles. The particle shape changes are associated with increasing
D values and decreasing sorting coefcient Q (decreasing size
range). The interrelationship between sorting, porosity, and D,
based on our data, is shown in Fig. 6. The plots indicate that
densication of gouge at potential shear localization sites is
concurrent with particle rounding and improved sorting.
4. Discussion
The studied gouges had not undergone appreciable hydro-
thermal alterations or pressure solution and lack signicant phyl-
losilicates fractions, all of which are shown to be important factors
in deformation of gouge in mature fault zones. As suggested by
studies of clayequartz gouge mixtures (e.g. Crawford et al., 2008),
the microstructural and mechanical effects of phyllosilicates in
quartz-feldspatic gouge are minimal if phyllosilicates make 5% of
the gouge contents. Experimental studies also show that the
process of shear localization is affected by the magnitude of the
mean stress mostly in the dilatational phase (Marone and Scholz,
1989; Gu and Wong, 1994; Besuelle, 2001). This is mainly because
the strain per fracture rule applicable to a fractal PSD breaks down
within the shear bands (Sammis et al., 1987) and there is less
number of contact points in a non-fractal aggregate. The shear
localization model discussed here is relevant to microstructures
within the shear bands in the post dilatational stage of the shear
localization.
The results provide both visual and numeric conrmation of the
correlation between shape of particles in the studied gouges and
the porosity values calculated from their corresponding SFG
models. The undeformed Aztec sandstone with round particles has
F
I
/F
S
of w1, while the undeformed synthetic Westerly granite
gouge with highly angular particles has F
I
/F
S
of w9. The changes in
F
I
and F
S
values and F
I
/F
S
ratio with increasing shear displacement
showthat using the model results to infer changes in particle shape
of the deformed gouges is also reasonable. In the regions adjacent
to Y-shear zones, porosities calculated from SFG models change
from 2% to 0.2% with increasing displacement as the corresponding
gouge porosity is reduced from 5% to 0.3%. Furthermore, we
showed that despite mineralogical differences of the two gouge
types the microstructural attributes of the gouge adjacent to shear
bands converge with increased comminution.
The particle size distribution adjacent to Y-shear zones in both
gouge types consistently yields D values greater than what is
considered to be an ideal gouge PSD of 2.6 produced through
conned comminution (Sammis et al., 1987; Sammis and Biegel,
1989; Blenkinsop, 1991). The gouge regions with D > 2.6, having
already achieved a high packing density, might be viewed as having
a critical PSDin connectionwith the shear localization processes. To
further dene the condition we consider porosity (packing density)
and sorting characteristics of the gouge from the studied regions.
The sorting ratio of the gouge was shown to drop continuously with
increasing displacement as the average ratio of the largest to
smallest size particles within the zones was reduced from 3 to 1.3
prior to shear localization. However, to interpret the decreasing Q
as an increasing porosity appears to contradict the decreasing F
I
and F
s
values we report in the same gouge over the same range of
shear displacements. A possible explanation is that porosity
reductions through particle rounding are offset by the porosity
gains through decreasing Q. The SFG models clearly showthat prior
to shear localization the rounding effect continues with increasing
shear displacement. The local dilation rate, therefore, may depend
on net porosity gain or loss due to the competing effects of the
changes in size and shape of particles within a densifying gouge.
A shear localization model based on the presented data and
arguments above appears to be in general agreement with the
localizationmodel discussedbyMarone andScholz (1989). For shear
localization in gouge with the critical PSD, a sufcient density of
regions with lower dilation rate dF/dg, is required. The shear
localization model illustrated in Fig. 7 thus involves unpacking (loss
of cohesion at micro-scales) of densied gouge along a zone of
favorably oriented pockets of relatively porous (microporous)
gouge. Based on our analyses of the critical PSD above, the net
porosity gain within a shear band region would occur only if the
comminution process (including particle boundary attrition and
transgranular fracture) becomes more effective in eliminating
particle size differences than rounding the particles. The micro-
mechanics of the competing effect could not be ascertained from
this study, but the data indicates that such microstructural state is
reached at fractal dimensions D > 2.6 as depicted in the schematic
Fig. 6. Change in image porosity F
I
and sorting ratio Q of gouge with increasing fractal
dimension. Vertical error bars represent range of porosity measurements about the
average. (a) Aztec sandstone gouge (b) Synthetic gouge. Higher Q values for sandstone
reect its higher average grain size compared to synthetic gouge. Data points for
undeformed material are shown for reference on the left side in each case.
plot of Fig. 7a. It is likely, however, that shear localization in the
densied gouge involves a range of D values at D > 2.6 rather than
requiring a unique D value for the entire length of a potential shear
band. This is simply because of the non-uniform nature of commi-
nutionintensity withina deforming gouge (Hadizadehand Johnson,
2003). The model gouge microstructure at the onset of shear local-
ization includes pockets or streaks of porous gouge, some of which
are favorablyorientedparallel to Y-shear orientationas illustratedin
Fig. 7b. Based on a sorting criterion Marone and Scholz (1989)
showed that in the bulk gouge with fractal size distribution sorting
remained poor and porosity remained low since small particles
tended to ll in between the larger particles. Within the bulk gouge,
regions with non-fractal PSD (e.g. within shear bands) tend to have
reduced particle size range, improved sorting and a higher porosity.
It is assumed that regions with lower rates of porosity reduction
(represented by dashed segment of porosity curve in Fig. 7a) are
initially highly localized, and that these regions need not to spread
throughout a signicant thickness of the gougebeforetheunpacking
occurs. This assumption is consistent with the common observation
that thickness of slipsurfaces andshear bands oftenconstituteavery
small fractionof the total gouge zone thickness. At particle-scale, the
model is supported by previous work that shows comminution in
shear localization sites tends to eliminate larger particles (Marone
and Scholz, 1989; Blenkinsop, 1991; Mair and Marone, 1999) prob-
ably because the growing number of smaller particles of all shapes
require larger stresses to fracture (Kendall, 1978). A uniformly
distributed porosity in the densied gouge would assist the shear
localization process by providing weak links between the porous
pockets.
The shear strength of an incipient unpacking surface is depen-
dent upon surface roughness (Biegel et al., 1992), which in this case
is mainly determined by the maximum size and spacing of the
pores in the dense gouge. Shearing of the sub-micron asperities
along unpacking surfaces may explain presence of thin bands of
extremely ne particles observed within shear bands and along
slickenside surfaces (e.g. Yund et al., 1990; Power and Tullis, 1989).
5. Conclusions
1. Conned comminution generates similar particle shape and
size distributions with increasing shear displacement in the
two studied gouge types with different mineral composition
and initial textures.
2. Adjacent to sites of shear localization the gouge particles are
rounder and particle size range is signicantly narrowed with
fractal dimensions D >2.6. This observation suggests that shear
localization must involve unpacking of a densied gouge.
3. The microstructural data and simple fractal gouge models
indicate that unpacking of the gouges might be the result of
highly localized porosity variations within the densied gouge,
caused by the competing effects of changes in particle shape
and particle size.
Acknowledgements
We wish to thank the editors of the JSG for their valuable
comments. Lori Kennedy and the anonymous reviewers of this
manuscript are thanked for their constructive comments. Discus-
sions with Judy Chester and Joseph C. White resulted in signicant
improvements in the manuscript. We wish to thank David Goldsby
and Anoaur Koncachbaev for their assistance with the rotary shear
experiments at Brown Geosciences Department, and Joseph Wil-
liams for his help with electron microscopy at the University of
Louisville. This research was partially supported by the US National
Science Foundation grant NSF-EAR-0229654 to Jafar Hadizadeh.
References
Antonellini, M.A., Aydin, A., et al., 1994. Microstructure of deformation bands in
porous sandstones at Arches National Park, Utah. J. Struct. Geol. 16, 941e959.
Anselmatti, F.S., Luthi, S., Eberli, G.P., 1998. Quantitative characterization of
carbonate pore systems by digital image analysis. AAPG Bull. 82, 1815e1836.
Beeler, N.M., Tullis, T.E., Weeks, J.D., 1996. Frictional behavior of large displacement
experimental faults. J. Geophys. Res. 101, 8697e8715.
Besuelle, P., 2001. Evolution of strain localization with stress in sandstone: brittle
and semi-brittle regimes. Phys. Chem. Earth (A) 26, 101e106.
Fig. 7. A model for shear localization based on competing changes in particle shape
and particle size of the gouge. (a) Behavior of microstructural variables porosity F, and
sorting coefcient Q with increased fractal dimension of gouge shown in an ideal
extrapolation of data. Shear localizes in regions of gouge where porosity increase by
narrowing particle size range (decreasing Q) is greater than porosity reduction by
particle rounding (decreasing F). Porosity is expected to drop sharply within shear
localization microstructures such as shear bands. It is assumed that changes in Q and F
take place at D > 2.6. (b) Highly schematic representation of gouge microstructure at
the onset of shear localization in gouge (dashed line segments in a) in an approxi-
mately 20 15 mm area. Pockets of well-sorted porous gouge (angular particles) are
bordered by dense gouge (round particles) with well distributed micro porosity. The
PSD in this region of gouge has D > 2.6. The particle shapes are simplied and exag-
gerated, and sense of shear is arbitrary.
Biegel, R.L., Wang, W., Scholz, C.H., Boitnott, G.N., Yoshioka, N., 1992. Micro-
mechanics of rock friction, 1. Effects of surface roughness on initial friction and
slip hardening in Westerly granite. J. Geophys. Res. 97, 8951e8964.
Biegel, R.L., Sammis, C.G., et al., 1989. The frictional properties of a simulated gouge
having fractal particle distribution. J. Struct. Geol. 11, 827e846.
Billi, A., 2007. On the extent of size range and power law scaling for particles of
natural carbonate fault cores. J. Struct. Geol. 29, 1512e1521.
Blenkinsop, T.G., 1991. Cataclasis and the processes of particle size reduction.
PAGEOPH 136, 59e86.
Bohannon, R.G., 1983. Mesozoic and Cenozoic tectonic development of the muddy,
north muddy, and northern black mountains, Clark County, Nevada. Geol. Soc.
Am. Mem. 157, 125e148.
Boullier, A.-M., Fujimotob, K., Ohtanib, T., Roman-Rossa, G., Lewinc, E., Itob, H.,
Pezardd, P., Ildefonsed, B., 2004. Textural evidence for recent co-seismic
circulation of uids in the Nojima fault zone, Awaji Island, Japan. Tectonophys.
378, 165e181.
Brock, W.G., Engelder, T., 1977. Deformation associated with the movement of the
muddy Mountain overthrust in the Bufngton Windows, S. Nevada. Geol. Soc.
Am. Bull. 88, 1667e1677.
Brogi, A., 2008. Fault zone architecture and permeability features in siliceous
sedimentary rocks: insights from the Rapolano geothermal area (Northern
Apennines, Italy). J. Struct. Geol. 30, 237e256.
Chester, F.M., Chester, J.S., 1998. Ultracataclasite structure and friction processes of
the Punchbowl fault, San Andreas system, California. Tectonophys. 295,
199e221.
Crawford, B.R., Faulkner, D.R., Rutter, E.H., 2008. Strength, porosity, and perme-
ability development during hydrostatic and shear loading of synthetic
quartzeclay fault gouge. J. Geophys. Res. 113 (B03207).
during the development of multiple fault sets in sandstone, Chimmney rock
fault array, Utah, USA. Tectonophys. 363 (1e2), 1e18.
Dieterich, J.H., 1981. Constitutive properties of faults with simulated gouge.
Mechanical behavior of crustal rocks, the Handin volume. AGU. Monog. 24,
103e120.
Evans, J.P., Chester, F.M., 1995. Fluid-rock interaction in faults of the San Andreas
tures. J. Geophys. Res. 100, 13007e13020.
Falconer, K.J., 1985. Fractal Geometry: Mathematical Formulations and Applications.
Cambridge University Press.
Flodin, E., Aydin, A., 2004. Evolution of a strike-slip fault network, Valley of Fire
state park, southern Nevada. Geol. Soc. Am. Bull. 116, 42e59.
Flodin, E., Prasad, M., Aydin, A., 2003. Petrophysical constraints on deformation styles
in Aztec sandstone, southern Nevada, USA. Pure Appl Geophys 160, 1589e1610.
Frank, F.C., 1965. On dilatancy in relation to seismic sources. Rev. Geophys. 3,
485e553.
Gu, Y., Wong, T.-F., 1994. Development of shear localization in simulated quartz
gouge: effect of cumulative slip and gouge particle size. Pure Appl Geophys 143,
387e423.
Hadizadeh, J., Johnson, W.K., 2003. Estimating local strain due to comminution in
experimental cataclastic textures. J. Struct. Geol. 25, 1973e1979.
Hayman, N.W., 2006. Shallow crustal fault rocks from the Black Mountain detach-
ments, Death Valley, CA. J. Struct. Geol. 28, 1767e1784.
Hecht, C.A., 2004. Geomechanical models for clastic grain packing. Pure Appl
Geophys 163, 331e349.
example from the Moab Fault, Utah. J. Struct. Geol. 27, 1469e1485.
Kendall, K., 1978. The impossibility of comminuting small particles by compression.
Nature 272, 710e711.
Keulen, N., Heilbronner, R., Stnitz, H., Boullier, A.-M., Ito, H., 2007. Grain size
distributions of fault rocks: a comparison between experimentally and natu-
rally deformed granitoids. J. Struct. Geol. 29, 1282e1300.
Krumbein, W.C., Sloss, L.L., 1951. Stratigraphy and Sedimentation. Freeman.
Logan, J.M., Dengo, C.A., Higgs, N.G., Wang, Z.Z., 1992. Fabrics of experimental fault
zones: their development and relationship to mechanical behavior. In: Fault
Mechanics and Transport Properties of Rocks. Academic Press Ltd., pp. 33e66.
Lowrison, G.C., 1974. Crushing and Grinding: The Size Reduction of Solid Materials.
Butterworth Publishers, London, 286 pp.
Mair, K., Marone, C., 1999. Friction of simulated fault gouge for a wide range of
velocities and normal stresses. J. Geophys. Res. 104, 28899e28914.
Mair, K., Frye, K.M., et al., 2002. Inuence of grain characteristics on the friction of
granular shear zones. J. Geophys. Res. 107 (ECV4), 1e9.
Mandl, G., De Jong, L.N.J., Maltha, A., 1977. Shear zones in granular materials. Rock
Mech. 9, 95e144.
simulated fault gouge. J. Struct. Geol. 11, 799e814.
Mead, W.J., 1925. The geologic role of dilatancy. J. Geol. 33, 685e698.
Moore, D.E., Summers, R., et al., 1989. Sliding behavior and deformation textures of
heated illite gouge. J. Struct. Geol. 11, 329e342.
Morgan, J.K., Boettcher, M.S., 1999. Numerical simulations of granular shear zones
using the distinct element method 1. Shear zone kinematics and micro-
mechanics of localization. J. Geophys. Res. 104, 2703e2719.
zones in sandstone. J. Struct. Geol. 26, 947e966.
Power, W.L., Tullis, T.E., 1989. The relationship between slickenside surfaces in ne-
grained quartz and the seismic cycle. J. Struct. Geol. 11, 879e894.
Rawling, G.C., Goodwin, L.B., 2006. Structural record of the mechanical evolution of
mixed zones in faulted poorly lithied sediments, Rio Grande rift, New Mexico,
USA. J. Struct. Geol. 28, 1623e1639.
Rockwell, T.K., Ben-Zion, Y., 2007. High localization of primary slip zones in large
earthquakes from paleoseismic trenches: observations and implications for
earthquake physics. J. Geophys. Res. 112 (B10304), 1e12.
Rogers, J.J.W., Head, W.B., 1961. Relationship between porosity, median size, and
sorting coefcients of synthetic sands. J. Sediment. Petrol. 31, 467e470.
Sammis, C.G., Ben-Zion, Y., 2008. Mechanics of grain-size reduction in fault zones.
J. Geophys. Res. 113.
Sammis, C.G., Biegel, R.L., 1989. Fractals, fault-gouge, and friction. Pure Appl Geo-
phys 131, 255e271.
Sammis, C.G., King, G., Biegel, R., 1987. The kinematics of gouge deformation. Pure
Appl Geophys 125, 777e812.
Scarpelli, G., Wood, D.M., 1982. Experimental observations of shear band patterns in
direct shear tests. In: Vermeer, P.A., Luger, H.J. (Eds.), Deformation and Failure of
Granular Materials. IUTAMDelft, Balkema, Rotterdam, Netherlands, pp. 473e484.
Shipton, Z.K., Cowie, P.A., 2001. Damage zone and slip-surface evolution over
micron to km scales in high-porosity Navajo sandstone, Utah. J. Struct. Geol. 23,
1825e1844.
Solymar, M., Fabricius, I.L., 1999. Image analysis and estimation of porosity and
permeability of Arnager Greensand, upper Cretaceous, Denmark. Phys. Chem.
Earth (A) 24, 587e591.
Talukdar, M.S., Torsaeter, O., Ioannnidis, M.A., Howard, J.I., 2002. Stochastic recon-
struction of chalk from 2D images. Transport Porous Media 48, 101e123.
Tanaka, H., Omura, K., Matsuda, T., Ikeda, R., Kobayashi, K., Murakami, M.,
Shimada, K., 2007. Architectural evolution of the Nojima fault and identication
of the activated slip layer by Kobe earthquake. J. Geophys. Res. 112 (B07304),
1e20.
Tullis, T.E., Weeks, J.D., 1986. Constitutive behavior and stability of frictional sliding
of granite. Pure Appl Geophys 124, 384e414.
Turcotte, D.L., 1986. Fractals and fragmentation. J. Geophys. Res. 91, 1921e1926.
Vardoulakis, I., 1980. Shear band inclination and shear modulus of sand in biaxial
tests. Int. J. Numer. Anal. Meth. Geomech. 4, 103e119.
Wolf, H., Konig, D., et al., 2003. Experimental investigation of shear band patterns in
granular material. J. Struct. Geol. 25, 1229e1240.
Yund, R.A., Blanpied, M.L., Tullis, T.E., Weeks, J.P., 1990. Amorphous material in high
strain experimental fault gouges. J. Geophys. Res. 95, 15589e15602.
Field evidences for the role of static friction on fracture orientation in extensional
relays along strike-slip faults: Comparison with photoelasticity and 3-D
numerical modeling
Roger Soliva
a,
*
, Frantz Maerten
a, b
, Jean-Pierre Petit
a
, Vincent Auzias
c
a
Universite Montpellier II, Lab. Geosciences Montpellier, UMR 5243, Place E. Bataillon, 34095 Montpellier cedex, France
b
IGEOSS, Parc Euromedecine, 340 rue Louis Pasteur, 34790 Grabels, France
c
BERKINE SONATRACH ANADARKO, Rte de Cina, 16001 Hassi Messaoud, Algeria
Article history:
Received 25 May 2009
14 January 2010
Available online 28 January 2010
To the memory of Maurice Mattauer,
professor at the University of Montpellier II,
who left us in April 2009
Keywords:
Fault
Friction
Relay
Wing cracks
Damage zone
a b s t r a c t
Fault friction is a parameter that is difcult to assess along fault zones since its determination depends on
the knowledge of any factor controlling the state of stress around faults. In brittle homogeneous rocks,
a limited number of these factors, such as the shape of the fault surface, the vicinity of fault tips or the
remote stress ratio, are crucial to constrain for this determination. In this paper, we propose to analyse
a eld example in which all these properties are met and where the nature of the slipped structure
suggest differences in static friction. We compare the orientations of branching fractures at strike-slip
relay zones between en echelon stylolites and en echelon joints both reactivated in shear. The eld data
are compared with both photoelastic and 3-D numerical models that consider the remote stress
conditions and the role of the geometry of the strike-slip segments. Based on eld observations, these
analyses quantitatively demonstrate the signicant role of fault friction on the local stress eld orien-
tation and subsequent fracture formation. This work points out that estimations of fault friction based on
analyses of fracture patterns or in situ stresses must be accompanied with a thorough investigation of the
3-D fault shape, its segmentation and the remote stress state.
1. Introduction
Static friction along faults is an extremely important parameter
for the understanding of the seismic cycle, the distribution of
stresses, fracture patterns and damage zones around faults. In the
past decades many efforts have been made to estimate fault friction
along natural faults (e.g. Hanks, 1977; Zoback and Zoback, 1980;
Brace and Kohlstedt, 1980; Lachenbruch and Sass, 1980; Zoback
and Healy, 1984; Mount and Suppe, 1987; Brudy et al., 1997;
Zoback et al., 1987; Scholz, 2000). The measure of static friction
estimated using laboratory tests on fault gouges is scale-limited, i.e.
on gouge samples from a bore hole cutting crossing the fault, and
therefore may not represent the frictional state of the whole
surface. Other approaches, based on the analyses of the heat ow
(Brune et al., 1969; Lachenbruch and Sass, 1980; dAlessio et al.,
2003) or numerical modeling (e.g. Parsons, 2002; Lovely et al.,
2009), allow discussion on the state of friction along the fault but
are quite indirect. The analysis of in situ stresses from bore hole
measurements or fracture patterns are considered as the best
indicator of the frictional state along a fault, (Zoback and Healy,
1984; Zoback et al., 1987; Scholz, 2000).
Assuming that fault cohesion can be close to zero on an active
fault (Byerlee, 1978), the static friction has been approximated by
Amontons rst law, in which the frictional cfcient (m) is
expressed as a function of the shear (F) and normal (N) components
of the forces applied to a frictional surface.
F m*N (1)
This law states that the friction coefcient of an innitely long
fault surface is directly related to the orientation and the
magnitude of the stresses close to this surface (Fig. 1a). This
reveals that the analysis of the stress eld around a fault can be
used to determine the static friction along a fault, in cases where
the remote ratio of stresses applied to the sliding surface is
known. Therefore, any indicators of the stress eld around faults
E-mail addresses: roger.soliva@gm.univ-montp2.fr (R. Soliva), fmaerten@igeoss.
com (F. Maerten), vincent_auzias@berkine.com (V. Auzias).
doi:10.1016/j.jsg.2010.01.008
(e.g. bore hole analysis, faults or fracture patterns) provide the
opportunity to quantify the static friction. However, this analytical
approach, based on Amontons rst law, assumes that the fault
plane is rectilinear and that the fault tips are innitely far from
the study area. Such a rst order approximation is quite unreal-
istic for natural faults having tips, being irregular, segmented or
more complex in shape (Fig. 1bd). The local orientation and
magnitude of the stress eld around a fault does not rely only on
fault friction, which makes its determination non-unique unless
we have knowledge of the other factors perturbing the local stress
eld.
In homogeneous rocks, the rst parameter that has been
considered as acting on the local stress eld, and more precisely on
the crack angle to a fault, is the static friction coefcient (e.g. Petit
and Barquins, 1988; Barquins et al., 1997; Martel, 1997; Ohlmacher
and Aydin, 1997; Willemse and Pollard, 1998; Zhou, 2006; Mutlu
and Pollard, 2008). However, the remote stress angle has been
considered as very important (Barquins et al., 1992; Ohlmacher and
Aydin, 1997) as well as the remote stress ratio (Auzias et al., 1997;
Katternhorn et al., 2000; Zhou, 2006). Others factors more related
to the geometry and behaviour of the fault surface also seem to be
very inuential, as the 3-D geometry of the faults (e.g. Segall and
Pollard, 1980; King et al., 1994; Willemse, 1997; Maerten et al.,
2002; Bourne and Willemse, 2001), its spatial/temporal evolution
(Willson et al., 2007; Lunn et al., 2008; Moir et al., 2009), and fault
opening (Kattenhorn and Marshall, 2006). Therefore, any analysis
of fault zones that aims to estimate the role of fault friction on the
stress eld, or in contrast to determine the state of friction from
stresses analysis, must know any of these factors that can perturb
the local stress eld.
In this paper, we analyse a eld example in which these factors
can be estimated. Drastic differences in fracture orientation
between reactivated frictional stylolites (i.e. structures of high
friction coefcient) and frictionless joints (i.e. structures of low
friction coefcient) suggest that friction is a prominent property
inuencing the stress perturbation at the close vicinity of a fault.
We chose to study fracture orientation at extensional relay zones
because the stress orientation has been described as quite stable in
space along a relay zone (compared to outside) due to the juxta-
position of the two extensive fault quadrants (see Fig. 1c) (e.g.
Auzias et al., 1997; Ohlmacher and Aydin, 1997). We compare the
eld data to photoelastic and 3-Dnumerical models to demonstrate
and quantify the signicant role of static friction on the stress and
fracture orientation at extensional relay zones.

= /
n
a b
Rectilinear fault
without tips
c
Rectilinear not frictional
segmented fault with tips
Rectilinear not frictional
fault with tips
1
1st Amonton's law
Stress trajectory = 0
= 0
= 0
= 0
n
= 0
=
0
= 0
n = 0
n = 0
n = 0
n = 0
n = 0
d
Complex not frictional
fault shapes with tips
1
Fig. 1. Comparison between the stress perturbation due to fault friction (a) and the stress perturbation due to different examples of fault geometry (b, c and d). (b), (c) and (d) are s
1
stress patterns inferred from photoelastic modeling (Joussineau et al., 2003). The remote stress applied is uniaxial. Dots represent fault tips. This gure shows that even for m 0, the
orientation of s
1
can be oblique and even parallel to the fault surface, rather than perpendicular as suggested by Amontons rst Law.
R. Soliva et al. / Journal of Structural Geology 32 (2010) 17211731 1722
2. Field data
2.1. Geological setting
The studied exposure, located close to Les Matelles (15 kmNorth
of Montpellier, France, Fig. 2), is a suitable site for the study of
brittle tectonics in limestones and stress perturbations around
meso-scale faults (Rispoli, 1981; Fletcher and Pollard, 1981; Petit
et al., 1999). The brittle tectonic structures observed (Fig. 3a)
were formed during multiphase compressive tectonics allowing the
formation of joints and stylolites. These structures of similar
dimension and orientation have been reactivated as slip surfaces
during a late tectonic event (Petit and Mattauer, 1995). Because of
their different roughness, joints and stylolites are expected to be of
different frictional properties during slip. Most of them show
secondary fracturing and linkage at relay zones (Fig. 3b), that can be
used as indicators of the palaeostress orientation (e.g. Rispoli et al.,
1981; Petit and Mattauer, 1995). It is therefore worthwhile to
address with particular care on the geological setting and history of
the brittle structures that will be used to constrain the role of fault
friction on fracture orientation.
The studied exposure has been fully described in a number of
previous studies (e.g. Rispoli et al., 1981; Taha, 1986; Petit and
Mattauer, 1995; Petit et al., 1999). This area is located in the
vicinity of a fault branch called the Lirou fault (Fig. 2b). The Matelles
fault zone, like many faults in the area, had both left-lateral strike
slip related to the Pyrenean shortening and normal slip related to
the Oligocene rift extension in the Languedoc. Middle cretaceous
normal slip along the MatellesCorconne fault zone is also expec-
ted during the Durancian tectonic events.
The brittle deformation sequence described by Petit and
Mattauer (1995) begins by a vertical jointing stage of the lime-
stone layers with two principal trends, N020 and N140. The second
stage is a rst generation of stylolite formation oriented N040. The
third stage, the most important for our study, is the reactivation of
the previous structures as sinistral and dextral strike slips due to
a last shortening creating wing cracks, en echelon veins and
a second generation of stylolites around the reactivated defects. As
shown by this last generation of joints and stylolites formed, the
last shortening stage occurs with the maximumprincipal stress (s
1
)
oriented NorthSouth. As suggested by rock experiments, photoe-
lastic models, numerical and analytical solutions (see Wawersik
and Brace, 1971; Petit and Barquins, 1988; Barquins and Petit,
1992; Chaker and Barquins, 1996; Lunn et al., 2008) the presence
of wing cracking around reactivated defects (see Fig. 3b) implies
remote stress conditions close to uniaxial loading (s
1
/s
3
10).
Conditions close to horizontal uniaxial stresses are possible at
shallowdepths, i.e. for little conning pressure. The expected depth
of faulting in the upper Jurassic limestone was probably less than
the thickness of the lower cretaceous series (w200 m), which was
potentially yet well eroded during the Pyrenean shortening. This
local stress state condition (high ratio of maximum to minimum
principal stresses, s
H
/s
h
) and reorientation of s
1
axis has been
related to a restraining bend along the Les Matelles fault during
Pyrenean strike-slip movements (Rawnsley et al., 1992; Petit et al.,
1999).
2.2. Extensional relay geometries
The last stage event provides the opportunity to analyse the
geometry of branching at relay zones between slipped overlapping
stylolites vs. slipped overlapping joints (Figs. 4 and 5a). The angle b,
dened as the angle between the orientation of remote s
1
relative
to the joints or stylolites reactivated in shear (Fig. 5c), is quite
variable (variation of w40
). A wide overlap of b angles is therefore

found for reactivated joints and stylolites containing relay zones.
For similar b angles, the branching angle a (dened as the angle
between the slipped structure and the branching jog) is quite
different with respect to the nature of the reactivated structure.
More precisely, for similar b angles, a is larger for joints than sty-
lolites (Fig. 5a and b). These observations are veried on a a vs.
b graph, in which additional measures from the literature were
reported for comparison. Field data from stylolites and joints are
consistent with the general scatter of all the data, and t in two
specic elds of branching congurations as described above.
Because of their different roughness (see Delair and Leroux, 1978;
Raynaud and Carrio-Schaffhauser, 1992, for the analysis of non-
reactivated stylolites), joints and stylolites are expected to be of
different frictional properties during slip. Therefore, the question
arising from these observations and treated in the next sections is
the following: Is this difference of branching geometry really due to
the frictional properties of the slipped structures? The local
orientation of relay branching fractures (a) gives a good approxi-
mation of the local orientation of s
1
during fault slip of all the faults
measured and at the same tectonic event (e.g. Auzias, 1995;
Ohlmacher and Aydin, 1997; Kattenhorn et al., 2000). The wide
range of a angle observed in the eld therefore reveals wide vari-
ation in the ratio of shear stress/normal stress, probably due to
Les Matelles
Pic St Loup
Lirou fault
branch
2 km
N
0 50
Gulf of Lion
Mesozoc cover
Cenozoc cover
km
Basement
Study
area
a
b
Camargue
basin
Jurassic
series
Cretaceous
series
Tertiary
series
Figure part b
Thrust
Faults undifferentiated
Strike-slip Normal fault
Fig. 2. Location and geological context of the study area. (a) Structural scheme of the
study area. (b) Geological map of the study area showing the Matelles fault and the
Lirou fault branch, modied from the geological map of St Martin de Londres, 1/50,000.
variations in static friction of the slipped structures. This hypothesis
will be tested below using analogue and numerical modeling.
3. Photoelastic modeling
3.1. Photoelastic method
A way to test the effect of friction is to analyse the stress eld
orientation using photoelastic experiments around opened (not
frictional, i.e. m 0) and closed defects (frictional). Photoelasticity is
an optical method of stress analysis within elastic translucent
materials like polymethylmethacrylate (PMMA), which present
accidental birefringence when loaded. These materials have the
property of resolving the light which falls on them at normal
incidence into two components, each one coinciding with a prin-
cipal plane of stress. This property, due to a higher density of the
material in these stress orientations, implies a light transmitted at
right angles (Hete nyi, 1966). If a photoelastic sample is placed
between crossed polarizers, black fringes named isoclinics (light
extincted, see Fig. 6b for an example) are observed on the second
polarizer, i.e. the analyzer. Isoclinics correspond to locations where
the plane of the incident polarized light coincides with one direc-
tion of principal stress within the sample. These isoclinics move
when the polarizers rotate together. By rotating the polarizers from
0
to 90
, and drawing the corresponding isoclinics, it is possible to

map the orientation of the two principal stresses. A thorough
description of the same experimental device (Fig. 6a) and addi-
tional details about the photoelasticity method can be found in
Joussineau et al. (2003).
To simulate the effect of fault friction on the stress eld of
extensional relay zone, two types of PMMA models were
compared: a rst one composed of closed defects and assumed to
have friction, and the other one with opened defects and assumed
to have negligible friction. To produce open defects, the thin plates
of PMMA (0.5 cm thick, 6.5 cm width and 10 cm high) were sawn-
off with a 300 mm thick micro-saw from central drill holes on each
interacting segments of 500 mm diameter. Closed defects were
produced as planar fractures propagated along linear traces drawn
with a cutter on the PMMA surface. These two types of models are
subjected to loading with different b angle such as 20
, 45
and 70
.
Here b is the angle between the axial loading and the planar defects
forming the relay zone.
The models are subjected to uniaxial conditions in order to be
consistent with the eld conditions expected (see Section 2). The
axial compressive load is imposed by an electromechanic testing
machine (Davenport 30 kN) and no lateral pressure is added. To
prevent bending of the PMMA plate under vertical loading, the
samples are maintained between vertical tighteners.
3.2. Experimental results of extensional relay stress pattern
The two types of models (frictional vs. not frictional) show
signicant differences in their local stress eld distribution. Fig. 7
presents stream lines of s
1
for not frictional (a) and frictional
conguration (b) subjected to a vertical loading with b 20
. The
stress eld is less perturbed in orientation for the frictional case.
Without friction, the orientation of s
1
is normal to the faults
(deviation of 70
from the remote s

1
) and quite stable along the
relay zone. Also note that the stress eld is perturbed outside of the
relay zone. In the frictional case, the stress eld changes from
the tip, where it is close to vertical, to the center of the relay zone,
where it reaches its maximum deviation of 45
. Additional tests,
not presented here and done for variable overlap and constant
spacing between the defects, showthe same maximumvalues of s
1
deviation and a better stability of s
1
orientation in the relay zone as
the overlap increases.
All the studied tests show results generally consistent with eld
observations. Fig. 8 exhibits the compilation of the a and b angles
data for all the tests done with constant relay geometry. Note that
a here corresponds to the angle between the slipped defect and s
1
at the center of the relay zone. Tests with no friction lie in the graph
area of high a and relatively lowb angles, which corresponds to the
zone of slipped joints (of low friction compared to slipped stylo-
lites). In contrast, frictional tests data lie in the area of lower a and
relatively high b angles, which corresponds to the zones of slipped
stylolites (high friction).
4. Numerical modeling
The numerical code used to investigate fault friction is a 3-D
Boundary Element Method (BEM) called Poly3D (Thomas, 1993). It
relies on the analytical solution of an angular dislocation in
a homogeneous elastic whole- or half-space (Comninou and
Dundurs, 1975). As opposed to the Okadas code (Okada, 1985),
Fig. 3. Field photographs of the study area. (a) Outcrop overview showing the layered upper Jurassic mudstones damaged by a dense pattern of calcite sealed fractures and
stylolites. (b) First generation stylolites reactivated as sinistral strike slips showing wing cracks and branched stylolites (see Rispoli, 1981). The length of the scale bar is 20 cm.
which uses rectangular elements, Poly3D discretizes faults and
fractures using triangular elements, and therefore avoids the
creation of overlaps and gaps between adjacent elements which
perturb the solution (Maerten et al., 2005). Mixed traction
displacement boundary conditions can be used for each constitu-
tive element of the model (tractions are shear and normal stresses
resolved on the fault surface). When traction boundary conditions
are specied, we have to solve for the corresponding unknown
displacement discontinuity according to the initially prescribed
traction values. As soon as all displacement discontinuities are
known (i.e. the slip patches), strain, stress and displacement can be
computed at any observation point within the elastic eld. Note
that transient variations in friction coefcient or the dynamic stress
eld are not considered (e.g., Poliakov et al., 2002).
Inorder tohave a frictional behaviour, the codehas beenextended
to support inequality constraints on traction and displacement.
Specically, the static Coulomb friction has been implemented as
a traction inequality constraint and validated by comparison with
analytical and numerical solutions (Maerten et al., 2009). For a given
fault surface, thecoefcient of frictionandcohesioncanbeprescribed
globally onto a fault surface or locally, each constitutive element
having their own coefcients. Traction boundary conditions are
imposed along the three axis of each triangular element local coor-
dinate system (dip, strike and normal directions).
Fig. 4. (a) Examples of joints (left side) and stylolites (right side) reactivated as left-lateral strike slips. (b) Interpretation in terms of stress orientation using the remote orientation
of syn-kinematic joints and stylolites. The remote stress orientation is slightly different in the two cases because they were not measured at the same location and that the larger
Lirou fault probably modify the stress eld orientation at this NorthSouth last compressive stage.
For a model subjected to a compressive far eld stress, inter-
penetration of the elements has to be avoided. This is achieved by
using the displacement inequality constraint u
z
0, where u
z
represents the computed normal displacement of a triangular
element. Again, traction boundary conditions are imposed along
the three axes of each triangular element local coordinate system
(Maerten et al., 2009).
4.1. Model set up
Fig. 9a and b depicts the model congurations used for the BEM
modeling for joints and stylolites, respectively, and are built upon
eld observations (Fig. 4). All the veins and stylolites traces in the
vicinity of the zone of interest have been carefully mapped and
then vertically extruded in depth all along the limestone layer
thickness, giving rise to the 3-D triangulated surfaces. As the
Coulomb friction relates the shear to the normal components of the
forces applied to a frictional surface, traction boundary condition
for the three local axes (dip, strike and normal) of each constitutive
triangular element is used.
4.2. Modeling of joints reactivated in shear
The joints model, depicted in Fig. 9a, is based on the eld
observations shown in Fig. 4, left side. These rst generation joints
are subjected to a far eld remote stress with uniaxial compressive
condition and s
1
(in this area) oriented N170 as suggested by the
presence of surrounding joints and stylolites (third brittle defor-
mation stage, see Section 2). In order to display the stress orien-
tation within the extensional relay resulting from the computed
displacement discontinuities, an observation grid is placed close to
the top of the model (Fig. 9). Then, two simulations are performed:
a rst one, with a constant coefcient of friction m 0.6 for all
discontinuities and no cohesion, and a second without any friction
but with non-interpenetration as a unique constraint. The elastic
material properties used for the surrounding limestone are n 0.25
and E 1 GPa (see Hatheway and Kiersch, 1989).
Fig. 10a and b displays the frictional and not frictional models,
respectively. The orientation of s
1
axis t better with the strike of
the branching fracture in the case where the frictional coefcient
equals zero. Since s
1
should be parallel to the strike of the
branching fracture, these models suggest that, at the initiation of
the linkage, the slipping joints were preferably not frictional. This is
consistent with the absence of macroscopic irregularities along
these rectilinear structures.
4.3. Modeling of stylolite reactivated in shear
For the stylolites model depicted in Fig. 9b, the uniaxial
compressive far eld stress is oriented N015, as proved by the
Fig. 5. Variation in the geometry of extensional jogs between stylolites and joints reactivated in shear. (a) and (b) are eld examples of reactivated stylolites and joints, respectively.
Coloured dashed lines represent the orientation of fractures in the relay zones. (c) Graph of a and b angles for all the reactivated stylolites and joints measured in the eld. a and
b angles are represented in a small scheme at the left. Field observations show higher b angles for reactivated stylolites.
presence of surrounding joints and stylolites (see Fig. 4, right hand
side, and Fig. 9b in Petit and Mattauer, 1995). Since this model is
composed of two relays, two observation grids are placed in the
vicinity of them close to the top of the model. A rst simulation is
done using only the non-interpenetration constraint (i.e. with
m 0), whereas a second one employs a constant coefcient of
friction m 0.6 without cohesion.
Fig. 11a and b displays the results on the two observation grids
for the not frictional and frictional models, respectively. As opposed
to the previous joint modeling, the linking structures are more
consistent with high friction stress orientations.
4.4. Parametric analysis
A series of models have been done for variable friction and
constant fault geometry consistent with the overlapping segments
of the experimental PMMA model. The 3-D shape of the model is
shown in Fig. 12a. The results are analyzed on the observation grid
which allows to compare a conguration close to the eld and the
photoelastic modeling (Fig. 12b). The models were performed with
variation of static friction coefcient and b angles as shown in
Fig. 12c. The elastic material properties used are the same than
above since a large part of the eld data used for comparison were
measured in limestone (Fig. 5).
The results are in good agreement both with eld and experi-
mental data. Fig. 13 exhibits the compilation of the a and b angles
data for all the tests done. As for the experimental analysis,
a corresponds to the angle between the slipped surface and s
1
at
the center of the relay zone. The numerical models with no or little
friction lay in the graph area of high a and relatively low b angles,
which corresponds to the zone of slipped joints (frictionless
structures). In contrast, frictional models t in the area of lowa and
relatively high b angles, which corresponds to the zone of slipped
stylolites (frictional structures).
Fig. 6. (a) Experimental device of the photoelastic modeling. (b) Example of isoclinic and isochromic fringes obtained in a vertical uniaxial loading experiment of slipping over-
lapping open defects.
Fig. 7. Drawing of s
1
obtained from the analysis of isoclinic fringes for (a) uniaxial
vertical loading of open defects, i.e. non-frictional and (b) closed defects, i.e. frictional.
Open defects (no friction)
Closed defects (frictional)
Limestone (Liu, 1983)
Limestone (Taha, 1986)
Granite (Raynaud, 1978)
80
70
60
50
40
30
20
20 40 60 80
10
Slipped stylolites
Slipped joints
()
()
Experimental data Field data
Frictional
Frictionless
Present
study
Present study
Fig. 8. Comparison between a and b angles obtained by photoelastic modeling with
the dataset measured in the eld.
5. Discussion
5.1. Stress perturbation and friction of the slipping defects
The models are in good agreement with the eld observations,
however they do not cover the entire range of data, especially for
the low a angles (Fig. 13). This point can be discussed with respect
to a limited number of unconstrained factors that may inuence the
stress eld around the slipping defects.
Inelastic deformations can modify the magnitude of residual
stresses in the host rock around faults, but this is probably not the
explanation of the scatter observed in Fig. 13 for two reasons. First,
with respect to the brittle subsurface conditions of deformation,
the studied limestone probably has negligible inelastic behaviour
preceding its shear yielding strength (Rispoli, 1981; Petit and
Mattauer, 1995). Second, inelastic deformation around fault, if
any, has probably a larger inuence on the stress magnitude than
on the orientation (Bu rgmann and Pollard, 1994), which one close
to the fault must be directly related to fault friction. This suggests
that elastic models are relevant to simulate residual stresses related
to fault slip in this geological context, and that the spreading of eld
data compared to the model is mainly due to others factors.
The effect of 3-D fault geometry, especially the fault aspect ratio,
on the stress distribution around fault has probably little inuence.
Soliva et al. (2006) show that the dimension of the area of stress
perturbation around a fault scales linearly with the fault length
since the fault growth is radial, and tends to be limited to a certain
distance when the fault height reaches a constant value. This means
that for vertically restricted fault by strata bounds, the 3-D shape
can inuence the orientation of the stress eld at about a distance
around the fault equivalent to the layer thickness. In the present
study, for all the measures of angles made in the eld (in the relay
zones), the distance of the fractures around the faults is always
lower than the layer thickness potentially restricting the faults
(tens of cm). We therefore work in a window around the faults
where the stress perturbation should not be inuenced by the fault
aspect ratio (3-D shape), and that all the eld measurements could
be compared to 2-D photoelastic models or 3-D full space models
proposed.
The reason of the difference of scatter between the eld data
and the models is potentially purely geometric. For all the model
results presented in Fig. 13 (both parametric and photoelastic), the
fault congurations are idealized as two planar surfaces with
constant overlap and spacing, whereas the eld data are fromfaults
more complex in shapes, with variable overlap and spacing, curved,
with multiple segments and potentially more complex in 3-D.
We have shown that friction is the main factor controlling the
stress perturbation and the orientation of linking fractures.
However the physical cause for this variation in friction needs to be
discussed. Obviously, this cause can be reasonably ascribed to the
difference of surface roughness between the stylolites and the
joints. However, the analysis of the roughness of the slipped defects
is not very relevant on faulted stylolites since after faulting they
show a smoothed irregularity that is certainly different than the
initial one. The measure of roughness has been done on non-
reactivated stylolites (see Delair and Leroux, 1978; Raynaud and
Carrio-Schaffhauser, 1992, for the quantitative analysis of stylo-
lites roughness in the same study area). However, these stylolites
were not reactivated potentially because of a threshold friction,
then different than the initial state on the faulted stylolites. On the
other hand, efforts in measuring the surface roughness of the
slipped defects cannot be very conclusive since it represents
the nite strain.
5.2. Estimation of fault friction and upscaling
From three different approaches: (1) eld study, (2) experi-
mental modeling and (3) numerical modeling, we have shown that
the angle of fracture branching in strike-slip relay zones is highly
dependent on the frictional state of the overlapping faults.
It is worthwhile to note that the static friction estimated at the
relay zone corresponds to the friction of the faults in the vicinity of
the relay zone and at the time of the fault interaction through the
relay zone. This frictional property may therefore have evolved
through time and space with the progression of fault coalescence.
The quasi-static friction estimated must be therefore considered
as the time integrated friction of the period of fault interaction
through their stress eld. We also must keep in mind that this
Fig. 9. 3-D view of the geometry of the defects reactivated in shear used in Poly3D for the numerical simulation. (a) 3-D geometry of the reactivated joints shown in Fig. 4 with the
observation grid (dark square) on which the stresses are represented in Fig. 10. The position of the observation grid corresponds to the position of the top of the limestone layer
observed in the eld. (b) 3-D geometry of the reactivated stylolites shown in Fig. 4. The observation grids (not shown here), on which the stress eld is presented in Fig. 11, are
placed at the same level than in (a).
approach is only suitable along faults if the remote stress conditions
are well known, since the ratio of s
H
/s
h
is very important for the
stress orientation in the relay zone (Auzias, 1995). Any indicator of
the remote and relay zone stress eld are worth considering.
With regard to estimation of fault friction on large scale active
faults, particular care must be taken with the stress eld deter-
mined from data close to the Earths surface. For depth shallower
than 300 m, the orientation of the principal stress may be different
than the tectonic stresses predominant at depth (e.g. Engelder,
1993). On large faults (kilometric scale length), it seems therefore
more appropriate to provide an estimation of fault friction based on
the tectonic stress orientation or the fracture patterns measured in
deep bore holes, which are more representative of the brittle crust
stress state.
This approach, based on eld observation and numerical
modeling at the relay zone, seems therefore relevant for the esti-
mation of fault friction along active fault segments interacting
through their stress eld. Its main advantage compared to rock test
measures, is the in situ estimation of the friction in its own
geological context. We integrate a large part of the fault surface
around the relay zone (and it can be done outside as well), as
opposed to tests done on fault rocks, which correspond to a specic
location of the fault surface crossed by the bore hole. Moreover, this
approach provides an overall value of friction of the entire active
fault zone, which may be composed of compartment with various
fault rocks, as for example coarse cataclasites or gouges which can
be difcult to analyse in laboratory tests. On the other hand,
a limitation of this method is that permanent deformation (e.g.
measured by GPS) can not be used to estimate friction with a quasi-
static elastic model. Visco-elastic simulation of the lithosphere
could be more appropriated if it allows to simulate the precise
geometry of the fault segments.
6. Conclusion
In situ static friction can be estimated along a fault plane if its
shape, the far eld stress conditions and the stresses at its vicinity
are well known. Joints and stylolites reactivated in shear show
roughly different angles of linking fractures at their extensional
relay zones. The irregular shape of the stylolites and the rectilinear
trace of the joints suggest that different frictional behaviour may
explain these differences in branching angles. Photoelastic and
numerical modeling conrm this phenomenon. For the same
remote stress conditions, variation of the static friction along
simulated faults explains a wide part of the range of branching
angle measured at relay zones. In particular, our paper reveals four
main points:
Fig. 10. Model results for the joints reactivated in shear shown in Fig. 9. (a) Modeling
result for m 0. (b) Modeling result for m 0.6. The small arrows on the observation
grid show the local orientation of s
1
.
Fig. 11. Model results for the stylolites reactivated in shear shown in Fig. 9. (a) Model
result for a m 0. (b) Model result for a m 0.6. The small arrows on the observation
grid show the local orientation of s
1
.
(1) A simple Amontons rst law cannot be used systematically to
infer the static friction along natural faults,
(2) To discuss the amount of friction along a fault, the analysis of
the local stress eld must be compared to elasto-static
approaches that integrate the effect of mechanical interac-
tions along ended faults, irregular in shape, segmented or more
complex,
(3) Both eld data, photoelasticity and numerical modeling show
that wide variations of friction can explain a large part of the
variation in the angle of secondary fracturing in the relay zones,
(4) Shear-reactivated joints have lower estimated static friction
than shear-reactivated stylolites.
Acknowledgments
We which to particularly thank and dedicate this paper to
Maurice Mattauer who left us in April 2009. He discovered the
studied outcrop and recently participated to discussions about this
work at the laboratory and also in the eld. The eld work from
Roger Soliva was supported by an Action Structurante 2006 grant
from the laboratory Geosciences Montpellier UMR5243. W. Ashley
Grifth and Roy Schlische are thanked for their helpful comments.
References
dAlessio, M.A., Blythe, A.E., Bu rgmann, R., 2003. No frictional heat along the San
Gabriel fault, California: evidence from ssion-track thermochronology.
Geology 31, 541544.
Auzias, V., 1995. Photoelastic modeling of stress perturbations near faults and of the
associated fracturing: petroleum industry application, II: Mechanism of 3D joint
development in a natural reservoir analogue: the at-lying Devonian Old Red
Sandstone of Caithness (Scotland). Ph.D. thesis, Universite Montpellier II, p. 311.
Auzias, V., Rives, T., Rawnsley, K.D., Petit, J.-P., 1997. Fracture orientation modeling in
the vicinity of a horizontal well. Bulletin Elf Aquitaine Production F64018,
381397.
Barquins, M., Chaker, C., Petit, J.-P., 1997. Inuence du frottement sur le branche-
ment de ssures a` partir de de fauts obliques soumis a` une compression uni-
axiale. Compte Rendu de LAcademie des Sciences T324, 2936.
Barquins, M., Petit, J.-P., 1992. Kinetic instabilities during the propagation of
a branch crack: effects of loading conditions and internal pressure. Journal of
Barquins, M., Petit, J.-P., Maugis, D., Ghalayini, K., 1992. Path and kinetics of
branching from defects under uniaxial and biaxial compressive loading. Inter-
national Journal of Fracture 54, 139163.
Bourne, S.J., Willemse, E.J.M., 2001. Elastic stress control on the pattern of tensile
fracturing around a small fault network at Nash Point, UK. Journal of Structural
Geology 23, 17531770.
tory experiments. Journal of Geophysical Research 85, 62486252.
Brudy, M., Zoback, M.D., Fuchs, K., Rummel, F., Baumgartner, J., 1997. Estimate of the
complete stress tensor to 8 km depth in the KTB scientic drill holes: impli-
cations for crustal strength. Journal of Geophysical Research 102, 18,45318,475.
Brune, J.N., Henyey, T.L., Roy, R.F., 1969. Heat ow, stress, and rate of slip along the
San Andreas fault, California. Journal of Geophysical Research 74, 38213827.
Bu rgmann, R., Pollard, D.D., 1994. Strain accommodation about strike-slip fault
discontinuities in granitic rock under brittle-to ductile conditions. Journal of
Byerlee, J.D., 1978. Friction of rocks. Pure and Applied Geophysics 116, 615626.
Chaker, C., Barquins, M., 1996. Sliding effect on branch crack. Physics and Chemistry
of the Earth 21, 319323.
Comninou, M., Dundurs, J., 1975. The angular dislocation in a half space. Journal of
Elasticity 5, 203216.
Delair, J., Leroux, C., 1978. Me thodes de quantication de la disparition de matie` re
au niveau de stylolites tectoniques et me canismes de la deformation cassante
des calcaires. Bulletin de la Socie te Ge ologique de France 7, 137144.
Engelder, T., 1993. Stress Regimes in the Lithosphere. Princeton University Press,
Princeton, New Jersey, U.S.A., 475 pp.
Fletcher, R.C., Pollard, D.D., 1981. Anticrack model for pressure solution surfaces.
Geology 9, 419424.
Hanks, T.C., 1977. Earthquake stress drops, ambient tectonic stress, and the stresses
that drive plate motion. Pure and Applied Geophysics 115, 441458.
Hatheway, A.W., Kiersch, G.A., 1989. Engineering properties of rock. In:
Carmichael, R.S. (Ed.), Practical Handbook of Physical Properties of Rocks and
Minerals. CRC Press, Boca Raton, FL, pp. 672715.
Hete nyi, M., 1966. Handbook of Experimental Stress Analysis. Wiley, New York.
Joussineau, G., Petit, J.-P., Gauthier, B.D.M., 2003. Photoelastic and numerical
investigation of stress distributions around fault models under biaxial
compressive loading conditions. Tectonophysics 363, 1943.
Kattenhorn, S.A., Aydin, A., Pollard, D.D., 2000. Joints at high angles to normal fault
strike: an explanation using 3D numerical model of fault perturbated stress
eld. Journal of Structural Geology 22, 123.
Kattenhorn, S.A., Marshall, S.T., 2006. Fault-induced perturbed stress elds and
associated tensile and compressive deformation at fault tips in the ice shell of
Europa: implications for fault mechanics. Journal of Structural Geology 28 (12),
22042221.
King, G.C.P., Stein, R.S., Lin, J., 1994. Static stress changes and the triggering of
earthquakes. Bulletin of the Seismological Society of America 84 (3), 935953.
Lachenbruch, A., Sass, J., 1980. Heat ow and energetics of the San Andreas fault
zone. Journal of Geophysical Research 85, 61856222.
Open defects with contact points
(few friction)
Closed defects (frictional)
Limestone (Liu, 1983)
Limestone (Taha, 1986)
Granite (Raynaud, 1978)
80
70
60
50
40
30
20
20 40 60 80
10
Slipped stylolites
Slipped joints
= 1
()
()
= 0.8
= 0.6
= 0.4
= 0.2
= 0
Numerical data
Experimental data Field data
Frictional
Frictionless
Present
study
Present study
Present study
Fig. 13. Comparison of a and b angles between the data obtained by numerical
modeling, photoelastic modeling and the dataset measured in the eld. A wide part of
a and b spreading can be explained by variations in frictional coefcient of the slipped
defects.
Fig. 12. Input conditions for the parametric modeling. (a) Conguration of the 3-D
model geometry with an example of computed displacement contours in color. (b)
Horizontal view of the model conguration showing the angles a and b. (c) Variables
used in the parametric study.
Lovely, P.J., Pollard, D.D., Mutlu, O., 2009. Regions of reduced static stress drop near
fault tips for large strike-slip earthquakes. Bulletin of the Seismological Society
of America 99, 16911704.
from linkage of preexisting structures. Journal of Geophysical Research 113,
B07403. doi:10.1029/2007JB005388.
Maerten, L., Gillepsie, P., Pollard, D.D., 2002. Effect of local stress perturbation on
secondary fault development. Journal of Structural Geology 24, 145153.
Maerten, F., Resor, P.G., Pollard, D.D., Maerten, L., 2005. Inverting for slip on three-
dimensional fault surfaces using angular dislocations. Bulletin of the Seismo-
logical Society of America 95, 16541665.
Maerten, F., Maerten, L., Cooke, M., 2009. Solving 3D boundary element problems
using constrained iterative approach. Computational Geosciences. doi:10.1007/
s10596-009-9170-x.
Martel, S.J., 1997. Effects of cohesive zones onsmall faults andimplications for secondary
fracturing and fault trace geometry. Journal of Structural Geology 19, 835847.
Moir, H., Lunn, R.J., Shipton, Z.K., Kirkpatrick, J.D., 2009. Simulating brittle fault
evolution from networks of pre-existing joints within crystalline rock. Journal
of Structural Geology. doi:10.1016/j.jsg.2009.08.016.
Mount, V., Suppe, J., 1987. State of stress near the San Andreas fault: implications for
wrench tectonics. Geology 115, 11431146.
Mutlu, O., Pollard, D.D., 2008. On the patterns of wing cracks along an outcrop scale
aw: a numerical modeling approach using complementarity. Journal of
Ohlmacher, G.C., Aydin, A., 1997. Mechanics of veins, fault and solution surface
formation in the Appalachian valley, U.S.A.: implications for fault friction, state
of stress and uid pressure. Journal of Structural Geology 19, 927944.
Okada, Y., 1985. Surface deformation due to shear and tensile faults in a half-space.
Bulletin of the Seismological Society of America 75, 11351154.
Parsons, T., 2002. Nearly frictionless faulting from unclamping in long-term inter-
action models. Geology 30, 10631066.
Petit, J.-P., Barquins, M., 1988. Can natural faults propagate under mode II condi-
tions? Tectonics 7, 12431256.
Petit, J.-P., Mattauer, M., 1995. Palaeostress superimposition deduced from meso-
scale structures in limestone: the Matelles exposure, Languedoc, France. Journal
of Structural Geology 17, 245256.
Petit, J.P., Wibberley, C.A.J., Ruiz, G., 1999. Crack-seal, slip: a new fault valve
mechanism? Journal of Structural Geology 21, 11991207.
Poliakov, A.N.B., Dmowska, R., Rice, J.R., 2002. Dynamic shear rupture interactions
with fault bends and off-axis secondary faulting. Journal of Geophysical
Research 107 no. B11, 2295. doi:10.1029/2001JB000572, ESE 6-1 6-18.
Rawnsley, K.D., Rives, T., Petit, J.P., Hencher, S.R., Lumsden, A.C., 1992. Joint development
in perturbed stress elds near faults. Journal of Structural Geology 14, 939951.
Raynaud, S., Carrio-Schaffhauser, E., 1992. Rock matrix structures in a zone inu-
enced by a stylolite. Journal of Structural Geology 14, 973980.
Rispoli, R., 1981. Stress elds about strike-slip faults inferred from stylolites and
tension gashes. Tectonophysics 75, T29T36.
Scholz, 2000. Evidence for a strong San Andreas fault.
Segall, P., Pollard, D.D., 1980. Mechanics of discontinuous faulting. Journal of
Soliva, R., Benedicto, A., Maerten, L., 2006. Spacing and linkage of conned faults:
the importance of mechanical thickness. Journal of Geophysical Research 111,
B01402. doi:10.1029/2004JB003507.
Taha, M., 1986. Apport de la microtectonique cassante au proble` me des trajectoires
de contraintes et de leurs perturbations. Exemples du Nord de Montpellier,
the` se dE
tat, universite de Montpellier, p. 155.

Thomas, A.L., 1993. Poly3d: a three-dimensional, polygonal element, displacement
discontinuity boundary element computer program with applications to frac-
tures, faults, and cavities in the earths crust. Master thesis, Stanford University.
Wawersik, W.R., Brace, W.F., 1971. Post-failure behavior of a granite and diabase.
Rock Mechanics 3, 6185.
Willemse, E.J.M., 1997. Segmented normal faults: correspondence between three
dimensional mechanical models and eld data. Journal of Geophysical Research
102, 675692.
Willemse, E.J.M., Pollard, D.D., 1998. On the orientation and patterns of wing cracks
and solution surfaces at the tips of a sliding aw or fault. Journal of Geophysical
Research 103, 24272438.
Willson, J.P., Lunn, R.J., Shipton, Z.K., 2007. Simulating spatial and temporal evolu-
tion of multiple wing cracks around faults in crystalline basement rocks. Journal
of Geophysical Research-Solid Earth 113 (2007). doi:10.1029/2007JB005388
B07403.
Zoback, M.D., Healy, J., 1984. Friction, faulting, and in situ stress. Annales Geo-
physicae 2, 689698.
Zoback, M.L., Zoback, M.D., 1980. State of stress in the conterminous United States.
Zoback, M.D., Zoback, M.L., Mount, V., Eaton, J., Healy, J., Oppenheimer, D.,
Reasonberg, P., Jones, L., Raleigh, B., Wong, I., Scotti, O., Wentworth, C., 1987.
New evidence on the state of stress of the San Andreas fault system. Science
238, 11051111.
Zhou, X.P., 2006. Triaxial compressive behavior of rock with mesoscopic heterog-
enous behavior: strain energy density factor approach. Theoretical and Applied
Fracture Mechanics 45, 4663.
Unlocking the effects of friction on fault damage zones
Heather M. Savage
a,
*
, Michele L. Cooke
b
a
Earth and Planetary Science Department, University of California, Santa Cruz, CA 95064, USA
b
Geosciences Department, University of Massachusetts, Amherst, MA, USA
Article history:
10 August 2009
Keywords:
Fault zone evolution
Slip-weakening friction
Off-fault damage
Boundary element modeling
a b s t r a c t
Two-dimensional, numerical models of a linear fault embedded within a linear elastic medium show the
generation of off-fault tensile failure that results from inelastic slip along the fault. We explore quasi-
static models with slip-weakening friction to assess the effects of spatially and temporally variable
friction on the damage patterns. Tensile fractures form where tangential normal stresses along the fault
exceed the tensile strength of the rock. These stresses result from locally high slip gradients at the
rupture tip. Because faults of different displacement history and rock type could have varying slip-
weakening distances, we examine the effect of changing the slip-weakening distance on the damage
pattern and nd that this parameter is important in determining off-fault fracture intensity and conti-
nuity along strike. Faults with short slip-weakening distance produce greater off-fault damage and
signicantly greater seismic radiated energy than faults with longer slip-weakening distances. We also
investigate the effect of pre-existing damage on the subsequent development of fractures in second
generation slip episodes and nd that damage localizes onto pre-existing patches. These results could
guide eld studies of small faults as to whether the fault failed in one slip event or multiple small events.
1. Introduction
Halos of pervasive cracking around faults, called damage zones,
are ubiquitous features (e.g. Brock and Engelder, 1977; Chester and
Logan, 1986; Faulkner et al., 2006). Although damage zones have
important implications for rupture dynamics (Rice et al., 2005) and
uid ow in fault zones (Caine et al., 1996), our understanding of
their initiation and development is limited. The processes at work
during fault slip that could contribute to damage of the host rock
include: change in the quasi-static stress eld, ground shaking due
to seismic waves, fault geometry (such as fault bends), and fric-
tional variability along the fault due to the frictional evolution of
fault surfaces and/or gouge layers. However, damage zones
nucleate and evolve from repeated deformation and displacement
along faults (Sibson, 1977; Chester and Logan, 1986; Cowie and
Shipton, 1998; Shipton and Cowie, 2003; Kim et al., 2004; Okubo
and Schultz, 2005) and the overprinting of multiple events
obscures howdamage is generated in a single event, when the host
rock is relatively intact. In order to understand how the damage
zone develops in early stages of fault displacement when damage
patterns should be more straightforward, we perform numerical
experiments of the incipient stages of off-fault fracture
development, on several-meter long faults that slip fractions of
centimeters.
This study highlights some potential differences in macroscopic
damage patterns and intensity between faults with varying initial
roughness. Surface roughness will determine the critical slip
distance, which is the slip distance required for a fault to renew
asperity contacts so that friction can evolve between steady-state
values. In this paper we are specically modeling this process solely
as a function of slip and will refer to this distance as the slip-
weakening distance (L). This distance is most likely a function of
fault maturity on natural faults. Incipient faults have rougher
surfaces that smooth with shear displacement (Sagy et al., 2007).
For faults with gouge layers, localization of slip into shear bands
will have smaller critical slip distances compared to gouge zones in
which the entire layer participates in shear (Marone and Kilgore,
1993). The length of the critical slip distance affects the variability
of friction along the fault during failure. As the fault begins to slip,
patches of the fault that have slipped more than the critical slip
distance will be weaker than patches that have not and therefore
faults with small critical slip distances will have greater difference
in friction between adjacent segments. This frictional variability
will in turn inuence the slip gradient along the fault (e.g. Burg-
mann et al., 1994). When slip gradients are high enough, local
tensile stresses can arise and off-fault fracturing commences
(Cooke, 1997). Although critical slip distances may scale non-line-
arly with displacement along faults at large slips (Ohnaka, 2000;
* Corresponding author. Tel.: 1 831 459 5263.
E-mail address: hsavage@pmc.ucsc.edu (H.M. Savage).
doi:10.1016/j.jsg.2009.08.014
Abercrombie and Rice, 2005), they should be constant in terms of
displacement for the small slip events we discuss in this paper.
Previous models of damage zone generation show that tensile
tail cracks form in the tensile quadrants of the fault during
displacements (e.g. Rispoli, 1981; Martel, 1997). Quasi-static models
predict that fractures generally format crack tips, which we refer to
as fault tips in this paper in order to distinguish from the crack tips
associated with off-fault damage. If present, an inelastic process
zone can allowcracks to forminboard of the fault tip due to the slip
gradient created at this change in frictional strength (e.g. Burgmann
et al., 1994; Cooke, 1997). Dynamic slip studies have shown that the
slip gradient preceding a rupture front produces fractures far
inboard of the fault tip (Yamashita, 2000; Dalguer et al., 2003;
Andrews, 2005). In these models, the zone of damage generated is
thinnest where the rupture begins and widest at the end of the
rupture in the tensile quadrants. The angle of the tail crack with
respect to the main shear fracture can be a function of the strength
of the process zone (Cooke, 1997) or in the case of dynamic rupture,
slip velocity (Broberg, 1999). Yamashita (2000) predicted that joints
formed in the tensile quadrant make a more oblique angle to the
fault than joints formed in the compressive quadrant, however
their fractures must fall along a prescribed regular mesh. These
models predict that the generation of off-fault damage slows
rupture propagation during an earthquake, as energy is absorbed
through the creation of new fractures (e.g. Andrews, 2005).
In this paper, we directly simulate off-fault damage in the form
of tensile fractures generated along small faults. Our model is quasi-
static and has fault elements with a slip-weakening failure crite-
rion. By studying a quasi-static slip-weakening frictional fault, we
can isolate the effects of friction from the effects of dynamic
stresses included in elastodynamic rupture models (e.g. Dalguer
et al., 2003; Rice et al., 2005). We investigate fracture patterns for
different slip-weakening lengths in an effort to distinguish howthe
early frictional properties of the fault inuence the development of
a damage zone. Although slip-weakening distance is a proxy for
fault surface roughness, it does not reect changes in fault
planarity. Because our model is not constrained by a pre-existing
mesh along which fractures have to form, these numerically
generated fracture patterns can be compared with eld-based
measurements of macroscopic fractures.
Pre-existing damage may affect the slip distribution as well. The
formation of fractures alter the properties of the fault zone that effect
rupture propagation and subsequent failures, such as: reducing
stiffness of the host rock, absorbing energy through the creation of
newsurfaces, and, once the rock becomes highly fractured, providing
additional material tocrushintogouge. Rupturebranchingmayresult
fromactivation of shear fractures off the mainfault (Rice et al., 2005).
Manighetti et al. (2004) suggested that slip proles along a fault are
modulated by pre-existing damage. We investigate the second
episode of slip by inserting the damage patterns from the initial
models in our study and re-slipping the fault.
In addition to the fracture analysis, we investigate the effect of
damage zone generation, as well as frictional variability on the fault
system and the propagating rupture, by assessing the mechanical
work budget. Work budget analyses examine the balance between
the external work done on the system (stemming from stress and
displacement at the boundaries) and the work consumed within
the system by different deformational processes (e.g. Mitra and
Boyer, 1986; Kanamori and Heaton, 2000; Cooke and Murphy,
2004; Abercrombie and Rice, 2005; Chester et al., 2005; Del Cas-
tello and Cooke, 2007; Ismat, 2008). Examining how elements of
the work budget change as off-fault fractures grow provides
insights into the tradeoffs among the deformational processes.
Additionally, we can assess how development of damage zones
changes the mechanical efciency of faults on a systemic level.
2. Methods
2.1. Numerical method
Numerical models based upon continuum mechanics have been
used to model various geologic processes such as earthquake trig-
gering (e.g. Stein, 1999) and fault interaction (e.g. Willemse et al.,
1996; Maerten et al., 1999; Savage and Cooke, 2004; Marshall et al.,
2008). Mechanical models are based upon the three governing
equations of continuum mechanics, i.e. the equilibrium, compati-
bility and constitutive equations (Crouch and Stareld, 1990).
The Boundary Element Method (BEM) is a numerical formula-
tion of the governing equations. BEM models consider disconti-
nuities (e.g. faults) in an otherwise elastic homogenous space. In
two dimensions, fault surfaces and boundaries are discretized into
linear elements. Tractions or displacements are prescribed for each
element. Analytical functions calculate the effect of an elements
traction or displacement on the rest of the elements within the
model. These analytical functions form a systemof linear equations
that determines the resultant displacement or traction when the
prescribed displacement or traction is summed with the inuence
of the other elements. Once the internal and external boundary
tractions or displacements are known, the tractions or displace-
ment for any point within the body can be calculated. This method
is less computationally expensive than other methods such as
Finite Element Method codes, which discretize the entire elastic
body.
Fric2D is a two-dimensional, open-source BEM code that
simulates deformation around fractures using the displacement
discontinuity method (Crouch and Stareld, 1990) and incorporates
a frictional failure criterion for fault slip, as well as fracture prop-
agation (Cooke, 1997). In the Fric2D code, inelastic slip begins when
the shear stress along the fault equals or exceeds the frictional
strength of the element as dened by the Coulomb criterion:
s
c
c s
n
m
s
(1)
where c is cohesion, m
s
is static friction, s
N
is normal traction
(compression is positive) and s
c
is shear strength. Here, we have
modied the failure criterion so that once the static frictional
strength has been exceeded and the element commences inelastic
slip, the frictional strength of the element decreases linearly as
a function of slip, s:
m m
s
sm
s
m
d
L
(2)
where m
d
is a prescribed dynamic friction value and L is the slip-
weakening distance. Once the element has slipped the length of L,
the frictional strength of the element will remain at the dynamic
friction value. Slip-weakening friction is a simplistic friction law in
the sense that it does not capture velocity dependence or memory
effects on friction (Dieterich, 1979; Ruina, 1983). However, slip-
weakening friction is adequate for dynamic rupture simulations
where earthquake cycles are not relevant and has been used
previously to study off-fault fracturing (Dalguer et al., 2003).
2.2. Model setup
We chose our boundary conditions to reect the conditions on
a fault at seismogenic depths subjected to shear stresses that are
close to the frictional strength of the center patch of the fault
(reecting that this part of the fault is critically stressed). Our
boundary conditions simulate simple shear conditions with
constant shear and normal displacements along the top of the body
and linearly decreasing displacements along the sides of the body,
H.M. Savage, M.L. Cooke / Journal of Structural Geology 32 (2010) 17321741 1733
with zero displacement along the bottom edge (Fig. 1). Normal
displacements are calculated to reect the lithostatic stress asso-
ciated with a fault buried approximately 5 km (125 MPa).
Displacement is applied in one step and changes in stress and
displacement in the ensuing iterations represent the system
reaching convergence. At each iteration, the friction coefcient
evolves along the fault (Eq. (2)), elements along the fault slip (Eq.
(1)) and new elements are added to simulate damage production.
The fault is 15.1 m long and horizontal. The fault and the bound-
aries are discretized into equal-length 5 cm long linear elements.
This element length provides satisfactory resolution of rupture
advancement while limiting computational load. Along the fault,
we prescribe a 3.1 m long center patch that is slightly frictionally
weaker, so that the center patch fails rst. In this way, we create
a nucleation patch along which the rupture begins and propagates
toward either end of the fault. The center patch is long enough to
induce unstable sliding along the entire fault. The shear displace-
ments along the boundaries of the model are chosen so that the
center weak patch (m 0.28) is at failure. The subsequent reduction
of friction coefcient from0.28 to 0.2 along the center patch during
slip provides a shear stress drop of 10 MPa. The sides of the fault
have higher prescribed friction coefcient, 0.32, and slip in
response to the stress drop on the center patch.
Because rocks are weakest in tension, we choose to investigate
areas likely to produce opening-mode cracks. New tensile fractures
grow where the tangential normal stresses along a fault element
exceed the tensile strength of the host rock, prescribed here as
15 MPa. Tensile stresses along the fault occur due to slip gradients
between elements. Because our model does not have a pre-existing
mesh, fractures are free to form at any orientation and nucleate
perpendicular to the local maximumtensile direction. Fractures can
form at the nodes between every other element and can grow by
one element length during each iteration of frictional slip. The
minimum spacing of off-fault fractures is 10 cm in these models.
Propagation continues until the stress intensity factor (K
I
) at the tip
of the fractures is less than the fracture toughness (K
Ic
), prescribed
here as 2.5 MPa m
1/2
. Newfractures are permitted to open and slide
with frictional resistance equal to the static friction value of the
fault element that spawned the new fracture. Fractures are not
allowed to interpenetrate.
2.3. Analysis of mechanical work
Total work of the fault system describes all of the energy
expended during tectonic deformation of the fault and the
surrounding host rock. Energy is consumed during deformation
from work against gravity (W
grav
), propagation of new surfaces
(W
prop
), work to overcome frictional resistance to sliding along the
fault (W
fric
), work that promotes ground motion in the form of
seismic radiation (W
seis
), and nally work that goes into off-fault
deformation which we refer to as an internal strain energy (W
int
).
The total work reects the summation of each of these
components:
W
tot
W
grav
W
prop
W
fric
W
seis
W
int
(3)
Each component of the total work done on the fault system
can be evaluated from our model. In our analyses, we do not
consider the effects of gravity because our fault is horizontal and
our surface has no topography. The deformational work budget
can be delineated in a variety of ways. Here we follow that used
by Mitra and Boyer (1986), Cooke and Murphy (2004), Del Cas-
tello and Cooke (2007) and Ismat (2008). The result is very
similar to the energy budget delineated by Kanamori and Heaton
(2000) and Abercrombie and Rice (2005). Both approaches
consider the same energy budget but divide the energy terms up
in slightly different ways based on a difference in observables. As
we describe each term we point out the differences in the
notations.
The external work represents the amount of work applied at the
external boundaries of our system. The complete external work
term is integrated along both the boundary and the applied
displacements (u
j
);
W
ext

ZZ
s
ij
u
j
; x
u
j
du
j
dx (4)
where s
ij
is the stress along the boundary due to u
j
and x is position
along the external boundary. In a closed system, the total work of
the system must equal the external work. In our models, the
boundaries are not permitted to move so the external work does
not change during rupture propagation; the only changes are the
partitioning of work amongst the different work components
within the system.
In order for a tectonic fault to slip, the shear stress along the
fault must overcome the frictional strength of the fault. The work
done against frictional resistance at a single fault segment is
calculated as:
W
fric
s
N
msA (5)
where s
N
is normal stress, m is the coefcient of friction, s is slip and
A is the ruptured areas of the fault. When stresses along the fault
are tensile so that normal stresses are zero or positive, the work
done against friction is zero. The complete frictional work in two
dimensions is integrated over both the loading path and the length
of the fault, l:
W
fric

ZZ
s
N
u
i
; lmsu
i
; ldu
i
dl: (6)
Frictional work depends on the coefcient of friction, which in
our slip-weakening model changes with increasing displacement.
Until the slip-weakening distance is reached, the m in the frictional
work termis a function of displacement. After a displacement equal
to L has been achieved, m is equal to the dynamic friction value. W
fric
is similar to the E
F
notation used by Kanamori and Heaton (2000) to
describe the frictional energy loss except that our frictional work
integrates over the decrease in shear stress as slip increases from
zero to L. E
F
only considers the frictional work done under the
dynamic shear stress. Consequently, W
fric
is equivalent to E
F
E
G
of
Kanamori and Heaton (2000) notation, where E
G
represents the
energy consumed along the fault as slip increases to the critical slip
distance. With our delineation of work terms, the frictional work
produced by the rupture is expected to depend on the slip-weak-
ening length of the fault, while the seismological frictional work
delineation does not.
The work done in the creation of new surfaces through the
nucleation and propagation of off-fault tensile cracks is a function
of the surface energy of a crack, G
c
, and the total area of new
fracture surface created, S.
E = 20 GPa;
=
0.2
u
n
= 2.81 cm; u
s
= 2.1 cm
2 m
Fig. 1. Schematic diagram of model setup. Black horizontal line is the 3.1 m long
critically stressed portion of the fault, which is otherwise shown as grey line.
W
prop
G
c
S (7)
The surface energy for rocks has been empirically estimated in
a variety of ways through laboratory (e.g. Wong, 1982; Cox and
Scholz, 1988) and eld analyses (e.g. Olgaard and Brace, 1983;
Chester et al., 2005). These estimations provide a wide range of
values. Analytically, the propagation energy can be directly calcu-
lated from the prescribed fracture toughness because the fractures
only grow when the stress intensity factor exceeds the fracture
toughness of the rock. The plane strain relationship between
energy release rate G
Ic
and K
Ic
provides a means to calculate W
prop
.
W
prop
SG
Ic
S
1 n
2
K
2
Ic
E
(8)
where E and n are the Youngs Modulus and Poissons ratio
respectively of the material. For the material property values
chosen for this modeling study (E 20 GPa; n 0.2;
K
Ic
2.5 MPa m
1/2
), the surface energy is 300 J/m
2
. This formulation
of W
prop
differs fromthat of the fracture energy parameter E
G
in the
Kanamori and Heaton (2000) formulation because we explicitly
solve for the surface energy involved in creating the off-fault frac-
tures in the damage zone.
The energy lost to ground shaking during an earthquake is
proportional to the shear stress drop during slip. Although our
quasi-static model cannot explicitly account for the energy that
would go into the seismic waves, we can approximate this term
based on the stress drop that occurs during slip-weakening. This
stress drop represents the release of some portion of the stored
elastic strain that accumulates as a fault is stressed, however stress
drop may represent only a small fraction of the total shear stress on
the fault. We approximate the seismic energy released during a slip
event as:
W
seis

ZZ
Ds
u
j
; l
u
j
; l
du
j
dl (9)
where Ds is shear stress drop during slip-weakening. W
seis
is
similar to the E
R
in the notation used by Kanamori and Heaton
(2000). We will investigate if faults with different roughnesses
release different amounts of seismic energy during rupture prop-
agation. Stress drop in a fully dynamic model maybe higher for the
given conditions than our model results, but the trends we see in
the seismic work for changing slip-weakening distance should be
applicable for a fully dynamic model.
The internal work of the fault system is measured strain energy
density. Timoshenko and Goodier (1951) derived the total strain
energy for a two-dimensional system to be the sum of stress
multiplied by strain over an innitely small increment of strain. The
integral of the strain energy over the entire two-dimensional body
yields:
W
int

ZZ
1
2
s
xx
3
xx
s
zz
3
zz
2s
xz
3
xz
dxdz (10)
Although the internal strain energy represents elastic (and
therefore recoverable) strain, the internal work term also repre-
sents the energy available for consumption by inelastic processes
such as the production of off-fault damage. Prior to any slip along
the fault, W
fric
W
prop
W
seis
0 so that W
int
equals external
work. We expect that W
int
will decrease with slip and damage
production along the modeled faults. In our study, the internal
work is sampled at observation points distributed throughout the
model. These observation points often fall within areas of
concentrated stresses near the tips of the off-fault damage. Near the
displacement discontinuity elements, the local stress singularity is
overestimated (i.e. r
1
instead of r
1/2
) so that sampling in these
regions produces articially high internal work. A more reliable
method of calculating W
int
is to subtract the other work terms from
W
ext
.
3. Results
3.1. Slip-weakening distance and off-fault fracture patterns
We compare the off-fault fracture patterns and slip proles
generated along faults with varying slip-weakening distances due
to the application of displacements at the boundaries of our model
(Fig. 2). The fracture patterns are shown for each fault at the time
when the rupture reaches the tip of the modeled fault. New frac-
tures develop perpendicular to the direction of greatest tensile
stress and at positions along the fault where local tensile stress
exceeds tensile strength. This occurs at the rupture front where an
element that slipped juxtaposes an element that has not; the high
slip gradient produces locally high tension on one side of the fault.
Fractures form mostly in the tensile quadrants of the rupture tip
and sub-perpendicular (approximately 7085
) to the fault. In
some cases, a few cracks develop in the overall compressive
quadrants of the fault when local tensions arise during rupture. As
new fracture tips continue to propagate, they grow in various
directions, highlighting the locally changing stress elds due to the
presence of other nearby fractures. The resulting sawtooth fracture
trace is element size dependent, however the average fracture
angle for a given crack is not. The lack of perfect symmetry in the
fracture pattern arises from slight asymmetry in boundary condi-
tions to prevent rigid body motion. Once a small degree of fracture
asymmetry is introduced, the asymmetry of the model is further
enhanced.
Faults with the longest L (Fig. 3A) show fracture patterns
resembling static friction fault models (Martel, 1997) where
fractures are located in the tensile quadrants at the fault tips.
Decreasing the slip-weakening length creates more fracturing
inboard of the fault tip, resembling fully dynamic models of
tensile fracture zones, with the fractured area forming a wedge
shape that tapers towards the center of the fault when new
fractures are allowed to continue to grow after the rupture has
reached the fault tip (Fig. 3B; Dalguer et al., 2003). This same
wedge-shaped pattern is predicted by analytical models that
predict zones of activated off-fault damage but do not explicitly
generate off-fault fractures (Andrews, 2005; Rice et al., 2005;
Templeton and Rice, 2008). However for our comparison, we
restrict our analysis of the fracture patterns to the iteration at
which the rupture reached the fault tip. Because the models
presented here, as well as other models of off-fault damage, do
not allow for the fault tip to propagate when the rupture reaches
the fault tip, the resulting off-fault damage pattern may not be
meaningful past this iteration.
The initial fault roughness (i.e. the slip-weakening distance)
has a large effect on resultant fracture density and clustering of
the damage zone (Fig. 2). Fracture density along the length of
the fault decreases as a function of increasing slip-weakening
distance (Fig. 4). The fractures form in clusters, with the number
of clusters decreasing with larger slip-weakening distances. The
clusters represent deviations in the slip prole from ellipticity.
For longer slip-weakening distances, the slip prole along the
fault maintains a mostly elliptical shape (Fig. 2). However, as L
decreases, small toes of slip extend from the rupture front that
represent the number of elements whose frictional strength is
falling from the static to the dynamic value in that iteration
(Fig. 4B; online supplementary material). The clusters form
between the element closest to the tip that has weakened to its
dynamic friction value and the elements along which friction is
falling. When the slip-weakening distance is small, fewer
elements are in transition between static and dynamic friction,
resulting in damage clusters that are closer together. The growth
length of fractures in a single iteration and spacing of the
clusters depend on element size, but not the pattern of
clustering.
Smaller slip-weakening distance means that elements reach
dynamic friction levels more quickly and the rupture reaches the
fault tip in fewer iterations for the smoother faults (Fig. 5). The
increased time spent slipping at higher coefcients of friction slows
the rupture speed on the rougher faults. The fault rupture propa-
gates on the order of centimeters per iteration, which can be
thought of as a unit of time. The growth of mode I fractures can
grow one element per iteration, which in these models is 5 cm.
Therefore the Mode II rupture speed is similar to the Mode I
propagation speed in these models.
An interesting point to note is that although aspects of the slip
patterns vary while the rupture is propagating, the nal slip
proles, as well as average and maximum slip values, are very
similar (Fig. 2). According to our models, faults with similar slip
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0
A
B
2 4 6 8 10 12 14 16 18 20
Position (m)
L = 0.1 mm, Iteration 050
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
L = 1 mm, Iteration 070
Fig. 3. Faults allowed to continue fracturing after the rupture reached the fault tip show that faults with small slip-weakening distances resemble models of static friction (A) and
faults with longer slip-weakening distances resemble models of dynamic slip (B).
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
L = 1 mm, Iteration 029
0.0
0.5
S
l
i
p

(
c
m
)
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
0.0
0.5
S
l
i
p

(
c
m
)
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
0.0
0.5
S
l
i
p

(
c
m
)
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
0.0
0.5
S
l
i
p

(
c
m
)
Fig. 2. Fracture patterns and slip proles generated along faults of varying slip-weakening distances: A) 0.001 mm, B) 0.01 mm, C) 0.1 mm, and D) 1 mm. Faults are compared at the
iteration at which the rupture front reaches the end of the fault. More off-fault damage occurs on smoother faults.
proles but different initial surface roughness would have very
different fracturing intensity within the damage zone, at least for
the initial stages of damage zone development. The maximum
width of the damage zone is similar between all models, and
damage zones are wider per unit of slip than predicted by fault
scaling models (Scholz, 2002).
3.2. Mechanical work
In an effort to assess the mechanical efciency of the fractured
fault zones, we analyze the mechanical work associated with
a variety of slip-weakening distances along the modeled faults. We
examine both the components of work when the rupture reaches
the tip of each fault, as well as how different components of the
total work change as the rupture propagates.
3.2.1. Change in work with rupture propagation
We investigate how the work against frictional resistance and
seismic energy release components of the work budget evolve
throughout the rupture process. Fig. 6 shows how frictional and
seismic work increase over the course of a slip event in two models
with differing slip-weakening lengths. At the onset, no slip has
A
B
2
3
4
5
6
5.0
10
15
20
25
0 0.5 1 1.5
Density
Clusters
L(mm)
N
u
m
b
e
r

o
f

C
l
u
s
t
e
r
s
D
e
n
s
i
t
y

(
f
r
a
c
t
u
r
e
s
/
m
)
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
Iteration 015
0
1
F
r
i
c
t
i
o
n
0.0
0.5
S
l
i
p

(
c
m
)
0.4
0.0
0.4
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
Iteration 014
0
1
F
r
i
c
t
i
o
n
0.0
0.5
S
l
i
p

(
c
m
)
Toe
Fig. 4. A) Along-strike fracture density decreases as a function of slip-weakening distance. This trend results in less continuous damage along the fault, so that there are fewer
clusters of fractures. B) Clustering is a function of slip-weakening distance because it represents the length of the transition zone from locked to slipping, and therefore the region
that is subjected to local tensile stress. This region can be seen as a toe on the slip prole in Iteration 15.
occurred and therefore no work has gone into friction or seismic
energy; all of the external work is expressed as internal work
within the material surrounding the fault. As slip progresses, W
fric
and W
seis
increase at the expense of internal work. The smoother
fault produces greater seismic work and greater frictional work at
each iteration of fault slip (Fig. 6). The seismic work and frictional
work are generated not just on the main fault but also along the off-
fault fractures as they slide. This accounts for irregularities along
the curves in Fig. 6. Because the rougher fault (L 1 mm) takes 4
more iterations than the smoother fault (L 0.01 mm) to reach the
fault tip, the frictional work on the rough fault slightly exceeds that
of the smooth fault once the rupture is at the tip of both modeled
faults.
3.2.2. Sensitivity of work components to slip-weakening distance
The total work of fracture propagation scales with the length/
area of new fracture surface produced (Eq. (9)). Faults with longer
slip-weakening distance produce less off-fault damage and
consume less work of fracture propagation (Fig. 7). The anoma-
lously large damage produced by the model with L 1 mmis due to
tensile fractures that develop within the compression quadrant of
this fault.
The work against frictional resistance to sliding and the seismic
radiated energy are plotted for faults when the rupture has just
reached the ends of the modeled faults. At this point, each of the
models has similar slip prole. The frictional work increases
modestly with increasing slip-weakening distance. Although faults
with L 1 mm have greater off-fault damage than faults with
L 2 mm, the L 1 mm faults require less frictional work. Fric-
tional work depends on both slip and friction coefcient. Because
the fault with longer slip-weakening distance slips while the fric-
tion coefcient is higher than the fault with shorter L, the fault with
longer L requires greater frictional work. Rough faults may require
slightly greater frictional work to slip than smoother faults;
however these differences may be small and impossible to discern
in the eld. In addition to slip and friction coefcient, frictional
work also depends on normal stress (Eq. (6)). While the normal
stress is the same for the faults modeled, it may differ signicantly
for faults in the eld.
The seismic radiated energy calculated from shear stress drops,
along both the primary fault and along the off-fault damage,
decreases sharply with increasing slip-weakening distance
(Fig. 7b). With smooth faults, the coefcient of friction and subse-
quently the shear stress can have greater drop between iterations
than along faults with long slip-weakening distance. The larger
shear stress drops between rupture propagation iterations
produces larger seismic radiated energy. Augmenting this trend is
the tendency for faults with smaller L to produce greater damage
(Fig. 7c). The development of off-fault fractures provides a means to
transfer stored internal work to seismic radiated energy. Together
these processes imply that rupture along smoother faults should
produce more shaking than ruptures along rougher faults.
3.3. Second generation of damage
Our model results suggest that fault surface evolution should
be accompanied with greater production of damage, slightly
lesser frictional work and signicantly greater seismic energy
release. These trends neglect the inuence of pre-existing off-
fault damage. To begin to address this issue we investigate the
propagation of rupture and development of damage along a fault
that already has some off-fault damage. We use the fracture
pattern from the L 1.5 mm fault for the initial damage pattern
and reapply the boundary conditions. Before we allow the
rupture to start along the central weak patch, we rst apply the
boundary displacements and let the faults slip and the cracks
grow to their full extent under the applied loading. Once this is
complete, the friction on all faults is brought to the static friction
levels and the friction along the central patch lowered to induce
rupture.
Table 1 presents the number of iterations to reach the modeled
fault tip and work values for the rst and second rupture episodes
along the L 1.5 mm fault. The number of slip iterations for the
rupture to reach the fault tip increases when the fault is anked by
existing fractures. The pre-existing damage deforms as the rupture
propagates along the fault, slowing down the rupture. Fracture
clusters fromthe rst episode are made longer but fewnewclusters
form (Fig. 8; online supplement). The growth of fractures per unit
slip also increases; whereas the total slip on the fault doubles when
adding the two episodes, the length of the longest fractures
quadruples. The total length of damage increases, which is reected
in the near doubling of the work of fracture propagation. The
presence of off-fault damage also increases the seismic radiated
Fig. 5. Iterations to the modeled fault tip for faults with differing slip-weakening
distance, L. Rupture propagates more slowly on rougher faults and takes more itera-
tions to reach the fault tip.
0.8
0.6
0.4
0.2
0.0
w
o
r
k

(
M
J
)
iterations of slip
10 20 0 30
W
seis
W
fric
W
seis
W
fric
L=1 mm L=0.1 mm
Fig. 6. Increase of work against frictional resistance to sliding (W
fric
) and work of
seismic energy release (W
seis
) during iterations of slip along two faults with different
slip-weakening distance, L. At each iteration of rupture propagation, W
fric
and W
seis
for
the smoother fault exceed the work of the rough fault. Once rupture reaches both
modeled fault tips, the ctional work is slightly greater for the rougher fault.
energy by a small amount compared to the rst episode of slip. The
internal work calculated by subtracting all the other work terms
from the external work decreases slightly with the second episode
of slip.
4. Discussion
4.1. Fracture patterns
The qualitative analysis of the off-fault damage pattern resulting
from fault slip has some interesting applications for eld
observations. The models show that the initial surface roughness
on the fault will have a considerable effect on the continuity of the
damage zone along strike. Faults with smaller slip-weakening
distances would have a more continuous damage zone, whereas
faults with large slip-weakening distances would have a less dense
network of fractures. However, we should note that a fault that has
greater geometric roughness (non-planarity) would concentrate
damage at asperities. The formation of off-fault damage diminishes
slip on the adjacent fault patch from an expected elliptical slip
distribution. The absolute magnitude of the critical slip distance is
difcult to assess for tectonic faults. In the laboratory, where
surfaces in general are smooth and gouge zone thicknesses in the
millimeter range, critical slip distance values are generally
measured to be 10 s of microns with the expectation of scaling up to
earthquake faults. Seismic slip inversions have estimated the crit-
ical slip distance on the order of 10100 cm. Although critical slip
distance should generally decrease as fault motion wears down
asperities (Sagy et al., 2007), the difculty of measuring the
evolution of this parameter in nature hampers our understanding
of these processes.
a b c
Fig. 7. Sensitivity of (a) frictional work, (b) and seismic energy and (c) propagation energy with slip-weakening distance along the modeled faults. We calculate the error of the
work values by examining the difference in work between these models and models with twice the element size. Open symbols denote models that produced tensional cracks
within the contractional quadrants of the fault. Frictional work increases slightly with L whereas seismic work decreases dramatically with increasing slip-weakening distance. The
amount of damage decreases with slip-weakening distance.
Table 1
Pre-existing fractures slow the speed of rupture and alter the work budget. More
work goes into seismic radiation and fracture propagation, whereas damage
decreases frictional work.
#Iterations
to fault tip
W
seis
(MJ) W
fric
(MJ) W
prop
(MJ) W
int
(MJ)
L 1.5 mm rst 34 0.10 0.86 0.0049 41.34
L 1.5 mm second 48 0.14 0.84 0.0075 41.31
1
0
1
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
0.0
0.5
S
l
i
p

(
c
m
)
B
1
0
1
P
o
s
i
t
i
o
n

(
m
)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
0.0
0.5
S
l
i
p

(
c
m
)
A
Fig. 8. Fault with pre-existing damage subjected to second episode of slip shows enhanced damage in areas where damage had already localized, but little new fracturing.
The models demonstrating the second generation of slip imply
that faults in the eld with more than one slip episode will have
high density, clustered fracture patterns, due to the ease of prop-
agating a fracture relative to forming a new one. This results in
a larger ratio between damage zone width and slip than a fault that
has one slip episode where total slip equaled the sum of the two
smaller episodes. However, using damage zone width per unit of
slip to estimate number of slip events would need to be limited to
comparing faults within the same eld site. Additionally, signicant
interseismic healing of fractures through processes such as mineral
precipitation would mitigate the effect of pre-existing damage on
subsequent ruptures.
4.2. Mechanical work analysis
Within the models of this study, we hold the external work
constant; no additional work is added to the system during the
propagation of the rupture. Because the left hand side of Eq. (3) is
constant during slip, the changes in work that we observe in the
models reects the transfer of work from internal stored work to
the non-conservative frictional heating, seismic radiated energy
and propagation energy. Of these non-conservative terms, the
greatest work is consumed in frictional heating along the primary
fault and its damage structures. In contrast, the work of fracture
propagation is several orders of magnitude smaller than the other
work terms, however processes such as rock pulverization could
require much more propagation energy than the cracks formed in
these models (Wilson et al., 2005). The total work of the system is
about 42.3 MJ so by far the greatest component of work is the
internal work stored within the system. Within these models only
2% of the internal work is converted to non-conservative work
terms. Our models showthat internal work decreases slightly upon
the second episode of slip along the fault. This suggests that further
ruptures along the fault could continue to transfer stored energy
within the host rock into frictional heating, seismic radiated energy
and the creation of new fault surfaces. With successive rupture
events we expect the slip-weakening distance along the fault to
generally decrease. Smoother faults are more effective than rough
faults at transferring work from internal work to the non-conser-
vative work terms.
5. Conclusions
Two-dimensional linear elastic models of frictional faults
suggest that the frictional slip-weakening distance (L) has signi-
cant effects on the tensile, off-fault damage pattern in a fault zone.
Faults with smaller slip-weakening distance have more continuous
along-strike damage whereas faults with large slip-weakening
distances concentrate fracturing at fault tips. Fractures form along
the fault in small clusters, with the number of clusters increasing as
function of L, thereby making along-strike fracturing more
continuous. Slip-weakening distance also affects how work is
consumed within the fault zone, with longer slip-weakening
distance resulting in more work done against friction and less
radiated seismic energy. Pre-existing damage further localizes
fracturing and consumes more internal work. Because initial
damage may be related to fault roughness, this implies that incip-
ient fault roughness controls along-strike fracture density even
after many episodes of slip.
Acknowledgements
We would like to thank two anonymous reviewers whose
insights greatly improved the paper. This work was supported by
NSF Grant EAR-0349070 to Michele Cooke.
Appendix. Supplemental data
Supplemental data associated with this article can be found in
online version at doi:10.1016/j.jsg.2009.08.014.
References
Abercrombie, R.E., Rice, J.R., 2005. Can observations of earthquake scaling constrain
slip-weakening? Geophysical Journal International 162. doi:10.1111/j.1365-
246X.2005.02579.x.
Andrews, D.J., 2005. Rupture dynamics with energy loss outside the slip zone.
Broberg, K.B., 1999. Cracks and Fractures. Academic Press, San Diego.
Brock, W.G., Engelder, J.T., 1977. Deformation associated with the movement of the
Muddy mountain overthrust in the Bufngton window, southeastern Nevada.
Bulletin of the Geological Society of America 88, 16671677.
Burgmann, R., Pollard, D., Martel, S., 1994. Slip distributions on faults: effects of
stress gradients, inelastic deformation, heterogeneous host-rock stiffness, and
fault interaction. Journal of Structural Geology 12, 16751690.
Chester, F.M., Logan, J.M., 1986. Implications for mechanical properties of brittle
faults from observations of the Punchbowl Fault, California. Pure and Applied
Chester, J.S., Chester, F.M., Kronenberg, A.K., 2005. Fracture surface energy of the
Punchbowl fault, San Andreas system. Nature 437. doi:10.1038/nature03942.
Cooke, M.L., 1997. Fracture localization along faults with spatially varying friction.
Journal of Geophysical Research 102, 22,425422,434.
Cooke, M.L., Murphy, S., 2004. Assessing the work budget and efciency of fault
systems using mechanical models. Journal of Geophysical Research 109.
doi:10.1029/2004JB002968.
Cowie, P.A., Shipton, Z.K., 1998. Fault tip displacement gradients and process zone
dimensions. Journal of Structural Geology 20, 983997.
Cox, S.J.D., Scholz, C.H., 1988. Rupture initiation in shear fracture of rocks: an
experimental study. Journal of Geophysical Research 93 (B4), 33073320.
Crouch, S.L., Stareld, A.M., 1990. Boundary Element Methods in Solid Mechanics.
Unwin Hyman, Boston, Mass.
Dalguer, L.A., Irikura, K., Riera, J.D., 2003. Simulation of tensile crack generation by
three-dimensional dynamic shear rupture propagation during an earthquake.
Del Castello, M., Cooke, M.L., 2007. The underthrusting-accretion cycle: work
budget as revealed by the boundary element method. Journal of Geophysical
Research 112, B12404. doi:10.1029/2007JB004997.
Dieterich, J.H., 1979. Modeling of rock friction: 1. Experimental results and consti-
tutive equations. Journal of Geophysical Research 84, 21612168.
rotation of regional stress in the fracture damage zone. Nature 444. doi:10.1038/
nature05353.
Ismat, Z., 2008 Energy budget during fold tightening of a multilayer fold. Journal of
Structural Geology 31. doi:10.1016/j.jsg.2008.10.006.
Kanamori, H., Heaton, T.H., 2000. Microscopic and macroscopic physics of earth-
quakes. In: Rundle, J., Turcotte, D., Klein, W. (Eds.), Geocomplexity and the
Physics of Earthquakes. Geophysical Monograph, vol. 20. American Geophysical
Union, Washington D.C., pp. 127141.
Kim, Y.-S., Peacock, D.C.P., Sanderson, D.J., 2004. Fault damage zones. Journal of
Maerten, L., Willemse, E.J.M., Pollard, D.D., Rawnsley, K., 1999. Slip distributions on
intersection normal faults. Journal of Structural Geology 21, 259271.
Manighetti, I., King, G., Sammis, C.G., 2004. The role of off-fault damage in the
evolution of normal faults. Earth and Planetary Science Letters 217. doi:10.1016/
S0012-821X(03)00601-0.
Marone, C., Kilgore, B., 1993. Scaling of the critical slip distance for seismic faulting
within shear strain in fault zones. Nature 362. doi:10.1038/362618a0.
Marshall, S.T., Cooke, M.L., Owen, S.E., 2008. Effects of non-planar fault topology and
mechanical interaction on fault slip distributions in the Ventura basin, CA.
Bulletin of the Seismological Society of America 98, 11131127. doi:10.1785/
0120070159.
Martel, S.J., 1997. Effects of cohesivezones onsmall faultsandimplications for secondary
Mitra, G., Boyer, S.E., 1986. Energy balance and deformation mechanisms of
duplexes. Journal of Structural Geology 8, 291304.
Okubo, C.H., Schultz, R.A., 2005. Evolution of damage zone geometry and intensity
in porous sandstone: insight gained from strain energy density. Journal of the
Geological Society of London 162, 939949.
Olgaard, D.L., Brace, W.F., 1983. The microstructure of gouge from a mining-induced
seismic shear zone. International Journal of Rock Mechanics and Mining
Sciences and Geomechanics Abstracts 20, 1119.
Ohnaka, M., 2000. A physical scaling relation between the size of an earthquake and
its nucleation zone size. Pure and Applied Geophysics 157, 22592282.
Rice, J.R., Sammis, C.G., Parsons, R., 2005. Off-fault secondary failure induced by
a dynamic slippulse. Bulletinof theSeismological Societyof America 95, 109134.
Rispoli, R., 1981. Stress elds around strike-slip faults inferred from stylolites and
tension gashes. Tectonophysics 75, T29T36.
Ruina, A., 1983. Slip instability and state variable friction laws. Journal of
Sagy, A., Brodsky, E.E., Axen, G.J., 2007. Evolution of fault surface roughness with
slip. Geology 35, 283286.
Savage, H.M., Cooke, M.L., 2004. An investigation into the role of fault interaction on
fold pattern. Journal of Structural Geology 26, 905917.
Scholz, C.H., 2002. The Mechanics of Earthquakes and Faulting. Cambridge
University Press, New York.
Shipton, Z.K., Cowie, P.A., 2003. Aconceptual model for theoriginof fault damagezone
structures inhigh-porositysandstone. Journal of Structural Geology 25, 333344.
Society 133, 191213.
Stein, R.S., 1999. The role of stress transfer in earthquake occurrence. Nature 402,
605609.
Templeton, E.L., Rice, J.R., 2008. Off-fault plasticity and earthquake rupture
dynamics: 1. Dry materials or neglect of uid pressure changes. Journal of
Timoshenko, S.P., Goodier, J.N., 1951. Theory of Elasticity. McGraw-Hill, New York.
Willemse, E.J.M., Pollard, D.D., Aydin, A., 1996. Three-dimensional analyses of slip
distributions on normal fault arrays with consequences for fault scaling. Journal
of Structural Geology 18 295209.
Wilson, B., Dewars, T., Reches, Z., Brune, J., 2005. Particle size and energetics of
gouge in from earthquake rupture zone. Nature 434, 749752.
Wong, T.-f., 1982. Shear fracture energy of westerly granite from postfailure
behavior. Journal of Geophysical Research 87, 9901000.
Yamashita, T., 2000. Generation of microcracks by dynamic shear rupture and its
effects on rupture growth and elastic wave radiation. Geophysical Journal
International 143, 395406.
Simulating brittle fault evolution from networks of pre-existing joints
within crystalline rock
Heather Moir
a,
*
, Rebecca J. Lunn
a
, Zoe K. Shipton
b
, James D. Kirkpatrick
b
a
Department of Civil Engineering, University of Strathclyde, Glasgow, Scotland, UK
b
Department of Geographical and Earth Sciences, University of Glasgow, Glasgow, Scotland, UK
Article history:
17 August 2009
Keywords:
Numerical modelling
Fault-zone evolution
a b s t r a c t
Many faults grow by linkage of smaller structures, and damage zones around faults may arise as a result
of this linkage process. In this paper we present the rst numerical simulations of the temporal and
spatial evolution of fault linkage structures from more than 20 pre-existing joints, the initial positions of
which are based on eld observation. We show how the constantly evolving geometry and local stress
eld within this network of joints contribute to the fracture pattern. Markedly different fault-zone trace
geometries are predicted when the joints are at different angles to the maximum compressive far-eld
stress ranging from evolving smooth linear structures to complex stepped fault-zone trace geometries.
We show that evolution of the complex fault-zone geometry is governed by: (1) the strong local vari-
ations in the stress eld due to complex interactions between neighbouring joints; and (2) the orien-
tation of the initial joint pattern with respect to the far-eld stress.
1. Introduction
Several authors have proposed that faults evolve under imposed
stress by the linkage of pre-existing structures (Segall and Pollard,
1983; Martel, 1990; Bergbauer and Martel, 1999; Pachell et al.,
2003). The pre-existing structures from which faults nucleate are
commonly open or mineral-lled joints that are weaker than the
surrounding rock (Segall and Pollard, 1983; Bergbauer and Martel,
1999; Pachell and Evans, 2002). When pre-existing features expe-
rience compressive loading, stress concentrations (both tensile and
shear) develop around the tip of the feature. Shearing of these pre-
existing features often results in the formation of secondary frac-
tures at (or near) the tip of the feature. These secondary fractures
have different names including: tail cracks/fractures (Cruikshank
and Aydin, 1994; Willemse et al., 1997), splay fractures (Pachell and
Evans, 2002; Myers and Aydin, 2004), horsetail fractures (Granier,
1985; Kim et al., 2004) and wing cracks (Crider and Peacock, 2004).
In this paper all fractures (tension or shear) associated with faulting
at (or near) the tip of a pre-existing feature are termed wing cracks.
Conceptual models of fault evolution through the development of
wing cracks (Martel, 1990; Martel and Boger, 1998) are supported
by eld observations of wing crack evolution from single joints or
faults (Kattenhorn and Marshall, 2006; Joussineau et al., 2007) and
by observations of linking fractures that have developed between
pairs of isolated faults (Peacock and Sanderson, 1995; Kim et al.,
2004). Wing cracks developing with shear displacement are also
commonly observed in eld data, for instance the sheared dyke in
Fig. 1a and the large splay faults in Kirkpatrick et al. (2008, their
Fig. 8).
In this paper, we focus on fault-zone development in crystalline
rocks. Natural exposures of fault-zone traces within crystalline
rocks can have many geometries, from smooth, approximately
planar features (Fig. 1a) where faults appear to develop along strike,
to complex stepped structures (Fig. 1b) where adjacent faults are
linked at stepovers, or a combination of both (Fig. 1c). Key questions
are: what governs the geometry of the evolving fault-zones? How
are fractures within the fault-zone linked?
A series of numerical models simulating fault growth, support
these conceptual models for fault-zone evolution. These models
have simulated the evolution of wing cracks from the tips of pre-
existing structures (Shen and Stephansson, 1993; Bu rgmann et al.,
1994; Kattenhorn et al., 2000; Willson et al., 2007) or the linkage of
pairs of faults with extensional and contractional geometries (Du
and Aydin, 1995; Bremaecker and Ferris, 2004; Lunn et al., 2008).
These simple, two-dimensional (2D) models have enabled predic-
tion of the orientation of linkage fractures and their mode of failure,
for a single fracture or pair of fractures in an ideal homogeneous
medium. However, these simulations, derived from one or two
fractures, are not sufcient to understand the range of complex
geometries observed in the eld (Fig. 1). Within this paper we
* Corresponding author at: Department of Civil Engineering, University of
Strathclyde, 16 Richmond Street, Glasgow G1 1XQ, Scotland, UK.
E-mail address: heather.moir@strath.ac.uk (H. Moir).
doi:10.1016/j.jsg.2009.08.016
extend current knowledge by simulating fault-zone evolution in
granite from a network of more than 20 joints. We show that
evolution of the resulting fault-zone geometry is governed by:
(1) the strong local variations in the stress eld due to complex
interactions between neighbouring joints; and (2) the orientation
of the initial joint pattern with respect to the far-eld stress.
2. Methodology
We use the computer code Modelling Of Permeability Evolution
in the Damage Zone surrounding faults (MOPEDZ) (Willson et al.,
2007) to simulate spatial and temporal evolution of complex
patterns of linking fractures. MOPEDZ was developed using the
commercially available nite-element software COMSOL which is
called from within the MATLAB code. The COMSOL nite-element
routines assume plane strain during the simulations. MOPEDZ is
a 2D nite-element model which solves Naviers equation in
a series of quasi steady-states and uses a combined Mohr Coulomb
and tensile failure criteria. Elements within the nite-element
mesh are either intact host rock or fractured host rock. Elements
which contain fractures (including the initial joints) are repre-
sented by lower effective material values (10% of the host) for
Youngs modulus, Poissons ratio and material strength, in a similar
approach toTang (1997). Representing the accumulation of damage
within each element by altering that elements material properties
is consistent with other damage mechanics models (Jing, 2003).
The initial conguration for all MOPEDZ simulations is similar to
that illustrated in Fig. 2 with the host rock (granodiorite) having the
Fig. 1. Field examples of mapped sections from fault-zones. (a) A segment of the outcrop map from NE of Neves lake in the Italian Alps showing a section of fault-zone with smooth
planar features (Pennacchioni and Mancktelow, 2007). (b) A segment of the outcrop map from the Waterfall region in the Sierra Nevada, California (Martel, 1990). (c) Map of
fractures in an exposure of the Lake Edison granodiorite in the Bear Creek region in the Sierra Nevada, California, UTM coordinates are: 0333075 4136569.
H. Moir et al. / Journal of Structural Geology 32 (2010) 17421753 1743
properties listedinTable 1andanyelements containingpre-existing
joints having reduced material properties (10% of the host) (Willson
et al., 2007). The simulated maximum compressive far-eld stress
direction, s
1
, is parallel totheyaxis (i.e. top-to-bottominall MOPEDZ
gures) and the minimum, s
3
, is parallel to the x axis (i.e. left-to-
right inall MOPEDZgures) (Fig. 2). Note that inthe eldboths
1
and
s
3
are horizontal. Initially all boundaries are displaced inward
holding s
1
2s
3
, (in compression) however following the rst
failure (either Mohr Coulomb or tensile) the s
3
boundaries are held
constant andfromthis point ononly the s
1
boundaries are displaced
towards each other, i.e. s
1
progressively increasing with s
3
held
constant. All simulations presented here are in compression.
Throughout this paper s
1
and s
3
refer to the far-eld stress imposed
by the boundaries of the nite-element model and s
1
Local
and s
3
Local
refer to the local stress eld around damaged cells. All simulations
use square nite-elements; the number of mesh elements in the
simulations presentedinthis paper varies from6400to136,500; the
size of each cell is approximately 13 mm
2
.
As an element fails (in either shear or tension) its material
properties are altered. Although the rst failures are triggered by
displacement of the boundaries, the alteration of the material
properties of those failed cells causes a change in both the direction
and magnitude of s
1
Local
and s
3
Local
(Lunn et al., 2008). This alteration
of the local stress may be sufcient to trigger additional failures
without any further displacement of the model boundaries. These
subsequent failures can be adjacent to previous failures, i.e. rep-
resenting the lengthening of a macroscopic fracture, or they can
occur in locations that are disconnected from any previous failure,
or they may be further fracturing of the same element or any
combination of these. MOPEDZ iteratively reduces the values of the
material properties as elements are predicted to fail; this reects
increasing damage to the host rock (host rock elements containing
pre-existing joints start with the lowest values, 10% of host rock).
Each element can fail up to a maximum of six times (resulting in
a reduction of strength, Youngs modulus and Poissons ratio) in
a geometric sequence (Willson et al., 2007) until they reach the
lowest value permitted (equivalent to those elements containing
the initial joints). We emphasise that each element in the mesh
may represent, at a sub-element scale, any number of micro or
macroscopic failures in the eld (in these simulations a sub-
element scale is smaller than 13 mm
2
).
Fig. 2. Typical initial setup showing the orientation of s
1
and s
3
(simulated far-eld
stress). Gray area is host rock, black is host rock containing joints (n.b. the pixellated
nature of the pre-existing joints is a product of the model). The model boundaries (red)
are under displacement control, following the initial failure only the top and bottom
boundaries are displaced. To avoid consideration of structures generated at the
boundary in the large simulations, only the central window (within the blue box) was
presented in the results. For all small simulations no window was taken and all results
within the red model boundaries are presented. The number of mesh elements varies
from 6400 to 136,500 depending on the size of the simulation. (For interpretation of
the references to colour in this gure legend, the reader is referred to the web version
of this article.)
Table 1
MOPEDZ simulation parameters for brittle rock.
Rock property Value Reference
Host Rock Youngs
modulus
60 GPa Martin (1997)
Turcotte and Schubert (2002)
Host rock Poissons
ratio
0.2 Turcotte and Schubert (2002)
Youngs modulus of
fractured element
1.2 GPa Segall and Pollard (1983)
Poissons ratio of
fractured element
0.02
Co (shear strength) 130 MPa Martin (1997)
m (coefcient of friction) 0.6 Byerlee (1967)
To (tensile strength) 10 MPa Martin (1997)
Number of cells permitted
to fail in any one step of
the MOPEDZ code
6
Where relevant the right hand column contains the reference from which the value
of the mechanical property was derived.
Fig. 3. Cartoon showing the evolution of restraining and releasing bends for a pair of
overlapping and under-lapping pre-existing joints with either contractional or
extensional relationship, (i) is the initial orientation of the joints, and sequences (iiiv)
show evolution of the predicted structure. All slipped joints are left lateral.
Fig. 4. (a) Small section from map shown in Fig. 1c. (b) Mapped joints oriented at 60
to the maximum principal stress (s

1
) (model requires s
1
to be parallel to the y axis). (c) Finite-
element mesh containing initial pre-existing joints (n.b. the pixelated nature of the pre-existing joints is a product of the model).
Fig. 5. Damage plot showing six frames from a simulation consisting of 350 steps which illustrate the temporal evolution of the linking fractures predicted by MOPEDZ from (i) the
initial joints through to (vi) the nal structure (nite-element mesh 390 350). The joints are oriented at approximately 60
to s
1
. Linkages at Location A are in a different
orientation to the rest of the simulation (see Fig. 7). At Location B overlapping joints in an extensional geometry link in a similar way as those in Fig. 3c. At Location C under-lapping
joints in a contractional geometry link in a similar way as those shown in Fig. 3b. At Location D two closely spaced joints in a contractional geometry link with a third more distant
joint which is in an extensional geometry. At Location E joints under-lapping and in an extensional geometry link in a similar way as those shown in Fig. 3a. All slipped joints are left
lateral.
Once a steady-state solution has been achieved for a given
boundary displacement, the top and bottom boundaries undergo
further displacement towards each other and the whole solution
process is repeated. Duringanyoneiterationof the code, onlya small
number of elements (<6inthesesimulations) are permittedtofail to
ensure stability of the model solution and provide an estimation of
the temporal propagation of the fractures.
Earlier research using MOPEDZ to examine failure from a single
joint (Willson et al., 2007) shows that fracture-trace geometries are
not sensitivetotheinitial mechanical properties of thehost rock(Table
1), e.g. if Youngsmodulus of thecrystallinehost rockis lower, thesame
tracegeometries areformedbut at lower values of the displacement of
the boundaries. Fracture-trace geometries are principally determined
bylocal variations intheYoungs modulus(i.e. damagedelements), the
orientation of the pre-existing joint to the far-eld maximum
compressive stress and the ratio of s
1
s
3
. In simulations where
s
1
[s
3
, failure was predominately in tension, for those where s
1
is
close to 2s
3
, simulated failure was predominately in shear. The mode
of failure results in different orientations of the evolving linkage
fractures (Lunn et al., 2008). MOPEDZ simulations of fault linkage
involving just two initial joints (Lunn et al., 2008) showed that frac-
ture-geometries develop in a predictable way summarised in Fig. 3.
Four initial stepover geometries were modelled: (a) under-lapping
extensional; (b) under-lapping contractional; (c) overlapping exten-
sional; and (d) overlapping contractional. The geometries of linkage
structures are governed by three key factors: (1) the ratio of s
1
s
3
; (2)
the initial relative positions of the joints, specically, contractional vs.
extensional geometries and overlapping vs. under-lapping joints; and
(3) the orientation of the most compressive principal stress direction
(s
1
) relative to the initial pair of joints.
In the following simulations we explore fault-zone evolution
through a large population of over 20 initial joints with gradually
increasing displacement of the s
1
boundaries of the model while s
3
boundaries are held constant. We start from an initial condition for
Fig. 6. Small simulation (80 80 nite-element mesh) with the joints in the same
orientation as Location E. (a) Surface plot of the principal stress prior to failure showing
interaction of the compressional quadrants of both joints. (b) Initial joint pattern
entered into MOPEDZ (overlap between j
1
and j
3
of 38 mm). (c) Damage plot of the
nal structure obtained.
Fig. 7. Small simulation with joints in the same orientation as Location A. (a) The
spatial and temporal evolution of the linking fractures predicted by MOPEDZ; black
represents elements of the of the nite-element mesh which contains fractures. (b)
Plots of the norm of the strain tensor which give a scalar representation of the
magnitude of the strain tensors; the darker the colour the higher the strain.
far-eld stresses of s
1
2s
3
. Within the initial joint population, all
four congurations of joint stepovers (as seen in Fig. 3) are locally
present. The simulations are conducted withthe joints at two angles
to s
1
(approximately 60
and 30
to the initial joints). To explore the

effect of local stress perturbations within a large joint network (>20
joints), simulation results are compared to the linkage structures
predicted for pairs of joints illustrated in Fig. 3.
2.1. Specifying the location and orientation of the pre-existing joints
The distribution of the pre-existing joints for the following
MOPEDZ simulations are based on part of an exposure of the Lake
Edison granodiorite in the Mount Abbot Quadrangle in the central
Sierra Nevada, California (Figs. 1c and 4a). Martel (1990), Evans et al.
(2000) and Kirkpatrick et al. (2008) show that faults in granites in
theSierraNevada have fault cores denedbyzones of cataclasite and
ultracataclasite and damage zones consisting of joints and minor
faults. These faults are thought tohave developed throughslip along
a population of joints (Segall and Pollard, 1983). These joints were
most likely formed during the cooling of the plutons (Bergbauer and
Martel, 1999). Observations suggest that slip along the joints was
accompanied by the development of wing cracks and linkage
structures forming small fault-zones (Martel et al., 1988). Field
observations were used to approximate the sections of the small
faults that may have comprised the original joints (before some
joints were reactivated). Un-reactivated joints were identied as
fractures that exhibit zero shear offset (through observations of
aplite dykes or mac enclaves) and lack any association with wing
cracks. The locations of reactivatedjoints were thendenedas those
portions of the small faults that have similar trace orientations
(within 10
) to the un-reactivated joints. The initial joint population

input into the following MOPEDZ simulations includes both the un-
reactivatedandreactivatedjoints. We emphasise that the purpose of
the following simulations is not to reproduce the detailed fault trace
geometries observed in the eld (since the actual locations of orig-
inal joints are not known) but to investigate conditions that may
promote differing styles of fault-zone development.
The initial joint pattern that was input into MOPEDZ and its rela-
tionship to the fracture-traces mapped in the eld is shown in Figs.
4ac. With the exception of the nal simulation the joints are at
Fig. 8. Small simulation with joints in the same orientation as at Location D. (a) Initial orientation of joints for MOPEDZ simulation (80 80 nite-element mesh, the pixelated
nature of the pre-existing joints is a product of the model). (b) Surface plot of the principal stress immediately prior to rst failure. (c) Plots of the norm of the strain tensor (scalar
representation of the magnitude of the strain tensors) illustrating the predicted evolution of the fractures.
approximately 60
to s
1
. Stress around pre-existing joints which
intersect the boundaries may be inuenced by the proximity of the
boundary. For the simulations with>20 joints, to avoidconsideration
of anystructures whichmight result fromboundaryeffects, results are
displayed and discussed only for an internal area in the centre of the
nite-element mesh, the edges of which are dened in Fig. 4b. Small-
scale simulations are also presented, to investigate behaviour at
specic locations within the larger mesh. These smaller scale simu-
lationsdisplayresults over thewholemodel domain(i.e. nowindowis
taken). In each case, initial damage predictions for the small mesh
were compared with those within the larger mesh to conrm that
predicted structures were similar, and hence that boundary condi-
tions were not having a substantial effect on model results.
2.2. Presentation of simulation results
Simulation results are illustrated using three types of maps. (1)
Damage plots show the elements that have failed grey indicates
intact host rock and black indicates fractured host rock (since
individual elements may fail multiple times in shear and/or
tension, modes of failure are not shown). (2) Strain plots show the
Euclidean norm of the strain tensor, which is one of the methods of
representing the scalar magnitude of a strain tensor (Mathews and
Fink, 2004). Plots of the normof the strain tensors for each element
elucidate a more detailed structure than the damage plots, since
they also highlight elements which are under a high strain but that
have not yet failed. The norm of the strain tensor presented here
may not be appropriate for direct comparison with eld data since
we start all simulations from an initial condition of zero strain. (3)
Surface plots of the local principal stress show the spatial distri-
bution of s
1
Local
; these plots have the same colour scale to alloweasy
comparison between simulations. Note that surface plots of the
local principal stress were produced within COMSOL in which
compression is negative and tension positive (the opposite
convention is usually adopted within the geological literature).
3. Results
3.1. Development of linkage structures
The spatial and temporal evolution of the fracture develop-
ment and linkage predicted by MOPEDZ, for the joint pattern in
Fig. 4c, is shown in Fig. 5 as a damage plot. The initial joints are at
approximately 60
to s
1
(Fig. 5i). At rst wing cracks begin to
develop on some but not all joints (Fig. 5ii). The orientation of the
propagating wing cracks are similar to those predicted for iso-
lated joint pairs in Fig. 3. As the simulation continues (Fig. 5iiivi)
several types of linkage structures are observed which are similar
to those in Fig. 3; note that only six frames are shown from
a simulation consisting of 350 steps. At Location A the linking
structure is similar to that for an overlapping pair of joints in an
extensional orientation. At Location B the structure is similar to
that for under-lapping joints in a contractional geometry (the
joints under-lap by a single mesh element). At Location C the
Fig. 9. Surface plot of the principal stress prior to failure for individual pre-existing structures in the same orientation as those at Location D, Fig. 5.
structure is the same as that for under-lapping joints in an
extensional geometry. At Location D stepover geometries for both
overlapping extensional joints and under-lapping contractional
joints are represented, the predicted linkage structure is similar
to that for overlapping joints in an extensional orientation.
At Location E the linkage structure that develops is different to
that predicted for an isolated pair of under-lapping contractional
joints (Fig. 3b); at E initial failure occurs in the host rock between
the two joints as opposed to propagating from the joint tips.
Because processing time for the large simulation (<20 joints) was
56 days, a small (6400 element mesh) simulation investigated
local behaviour at Location E using three joints in the same relative
positions; both the physical size represented by each nite-element
and the boundary conditions (progressive displacement of the s
1
boundaries starting from an initial value of s
1
2s
3
) remain the
same as that for the large simulation in Fig. 5. The stress eld
(Fig. 6a) shows that the relative positions of the pre-existing joints
facilitates interaction of the compressional quadrants of the two
joints (j
1
and j
3
), which results in linkage due to shear failure. In the
simulation joints j
1
and j
3
shown in Fig. 6b overlap by 38 mm. If the
tip of joint j
3
is adjusted (by at least 38 mm either way) either to
clearly over- or under-lap j
1
, linkage geometries are similar to those
expected for over- or under-lapping contractional joints (Fig. 3b;
Fig. 12 in Segall and Pollard, 1980).
Three joints circled at Location A form two stepovers. The left
stepover is extensional and the right stepover is contractional. The
damage evolution (Fig. 7a) and strain evolution (Fig. 7b) in a small-
scale simulation (6400 elements) with the joints in the same
relative locations shows two types of linkage structure. Initially, the
pair on the left behaves as the extensional geometry in Fig. 3c.
However, as the fracture propagating from the middle of the upper
joint lengthens, it begins to interact with the joint on the right of
the gure, changing its orientation and eventually resulting in
linkage of the pair of joints on the right that are in a contractional
geometry (Fig. 3d).
Location B evolves a linkage structure similar to that predicted
in Fig. 3b for under-lapping contractional geometries (the joints
are displaced from being collinear by approximately 1 cm). At
Location D, despite the upper two joints being closer together
than those at Location B and in a more pronounced under-lapping
contractional geometry, a similar linkage structure does not
develop. Instead, a wing crack propagates from the much more
distant, extensionally-related joint below. This geometry illus-
trates the effect of neighbouring joints, which is investigated by
a small-scale simulation of the three joints at Location D. The
initial joint conguration is shown in Fig. 8a and the magnitude of
s
1
prior to failure is shown in Fig. 8b. Comparison of Fig. 8b with
the stress elds which are predicted for each of the three joints if
simulated separately (Fig. 9) shows that having all three joints
present reduces the magnitude and extent of the region of
compressional stress surrounding the interacting tips, and
increases the magnitude and extent of the tensional stress, most
notably between joints p
2
and p
3
. This explains the linkage
structure that develops between the more distant extensional pair
of joints, evident from the strain evolution (Fig. 8c). The resulting
fracture pattern is similar to that seen in a geometrically similar
conguration of starter joints in Segall and Pollard (1980, their
Fig. 12).
3.1.1. The inuence of joint length on linkage
The extent of the local stress perturbation associated with
a fault has previously been related to its trace length (e.g. Segall
and Pollard, 1980). To explore the effect of joint length on evolving
fault linkage, small-scale simulations were performed adjusting
the joint conguration at Location D. The lengths of the upper
Fig. 10. Spatial and temporal evolution of strain predicted by MOPEDZ with pre-
existing structures in the same relative positions as Location D but with (a) the initial
length of p
1
doubled (p
1.1
) and (b) initial length of p
1
and p
2
were both doubled (p
1.1
and p
1.2
respectively). Note that wing cracks only develop on p
3
when the upper wing
cracks reach the boundary; had it not done so the growth of wing cracks fromp
3
would
have been suppressed. (All faults are left lateral.)
joints (p
1
and p
2
) were increased, keeping the relative position of
the adjacent joint tips constant (Fig. 10). When the length of
either one (Fig. 10a) or both (Fig. 10b) of the upper joints are
doubled, joints p
1
and p
2
link in a manner similar to that for
under-lapping contractional joint pairs. Previously, in Fig. 8,
linkage was between the extensional pair of joints p
2
and p
3
. From
these simulations it is apparent that the length of the joints
affects the magnitude and extent of the stress eld at the joint
tips, thus enhancing or decreasing the likelihood of linkage
between a pair of joints.
3.1.2. Separation between joints
Small-scale simulations consisting of up to three joints were
used to explore the effect of joint separation; this was investigated
by increasing the distance between joints in the y direction. The
position of the lower joint (based on the conguration at Location
D) was adjusted until it was separated enough to allow the upper
two (p
1
and p
2
) to link (Fig. 11a). The stress perturbations due to
each joint were then explored by systematically removing each
fromthe simulation (leaving only two joints in the simulation). This
produced varying joint linkage geometries from joints linking
rapidly (Fig. 11b) through joints that fail to link, but show an
evolving structure within the linkage zone (Fig. 11c) to joints which
do not link (Fig. 11d). Note that in this nal simulation (Fig. 11d) the
wing crack which develops fromp
3
does so at a different angle (40
from s
1
) than for the original simulation (24
from s
1
). Simulations
illustrated in Fig. 11 show that the proximity of neighbouring joints
affects both the location and orientation of the linkage structures
that develop.
3.2. Exploring the effect of orientation of the regional stress eld
Simulations of linkage from isolated pairs of joints in Lunn
et al. (2008) showed that one of the key factors controlling the
fault-zone geometry was the orientation of the joints to s
1
. In the
case of joints at a low angle to s
1
wing cracks were found to
propagate back into the compressional quadrant, similar to
structures observed in the eld (Vermilye and Scholz, 1998). To
explore what effect the orientation of s
1
can have on fault-zone
evolution from a complex joint pattern, the joints in Fig. 4 were
oriented at an angle of approximately 30
to s
1
(Fig. 12i). The
predicted evolution of linkage structures through time is shown in
Figs. 12iivi; a visual comparison of Fig. 12vi and Fig. 5vi shows
the nal geometries to be very different. Critically, joint traces
that were approximately co-linear now progressively link up
along strike to form long smooth linear fault traces (e.g. Locations
Fig. 11. (a) Spatial and temporal evolution of strain predicted by MOPEDZ with joints in the same relative positions as Location D but with p
3
displaced away from the upper two
joints. Temporal evolution of strain predicted by MOPEDZ (b) if p
3
is removed, (c) if p
1
is removed and (d) if p
2
is removed. Note that the angle of the wing crack propagating from
the p
3
is 24
for simulation (a) but is 40
for simulation (d).

B and D). Further, the few wing cracks that do evolve in Fig. 12
only develop once neighbouring joints have linked to form
through-going faults and propagate back into the compressional
quadrant (e.g. Locations A and F). The only exception to this is at
Location C, (Fig. 12iv) where an extensional stepover of the under-
lapping joints develops.
The alteration of s
1
LOCAL
and s
3
LOCAL
in both magnitude and
direction at the tips of a propagating fracture (or the original joint)
results from a combination of factors (Fig. 13).
The shape of the stress eld around a fracture or joint is
affected by its orientation with respect to the simulated far-
eld stresses s
1
and s
3
(Fig. 13a).
The stress elds of neighbouring faults and/or fractures interact
such that they can act to enhance or diminish local perturba-
tions in the stress eld (Fig. 13b).
Wing cracks which have a different orientation to that of the
initial joint will generate new perturbations in the local stress
eld (Fig. 13b(ii) for 60
).
These factors that control the perturbations in the local stress
eld have an inuence on the different styles of linkage structures
which develop between initially co-linear joints.
Three key observations are apparent from a comparison of the
simulation results for pre-existing joints at 60
and 30
to s
1
:
For the simulation at 60
to s
1
(Fig. 5), the linking wing cracks
are at a much wider variety of angles (e.g. see Locations B and D
in Fig. 5) than those at 30
to s
1
(Fig. 12).
For the simulation at 60
to s
1
(Fig. 5) many joints develop
multiple wing cracks at individual joint tips and linkage
structures tend to exhibit more damage than those at 30
to s
1
(e.g. compare Location D in Figs. 5 and 12).
With structures at 30
to s
1
, approximately 60% of the joints
link along strike forming smooth linear features which span
the model domain (Fig. 12); these features do not form for the
simulations at 60
to s
1
,
4. Discussion
The simulations show three key ndings: (1) local spatial and
temporal variations in the stress eld have a signicant effect on
the location, orientation and timing of wing crack development,
resulting in signicantly different patterns than those predicted
from consideration of single fractures or pairs of fractures. (2) A
signicant difference in resulting fault-zone geometry is predicted
when s
1
is oriented at 30
with respect to the initial joint pattern

(Figs. 5 and 12). (3) The spatial distribution (lengths, separation,
overlap, under-lap, spacing) of the original joints is a key control on
the predicted locations, orientations and timing of wing crack
development.
4.1. Local variations in the stress eld
The simulations focus on fault formation via linkage of pre-
existing original joints and show that the proximity of neighbour-
ing joints, and their effect on the local stress eld, affects both the
location and orientation of the linkage structures that develop. The
large-scale simulations represent an area approximately 4.5 m
wide interestingly, similar results are predicted if the model
domain represents a larger physical size (e.g. several kilometres
wide), where the initial features are 20 en-echelon pre-existing
faults (the only difference being the physical size represented by
each element in the nite-element mesh and the load required to
create the same number of damaged elements).
Currently researchers in many elds use predictions of static
stress distribution around an existing fault network to predict the
locations and orientations of likely fracture zones (Maerten et al.,
2002; Micklethwaite and Cox, 2004). Our simulations demonstrate
that the locations of active fracture zones associated with faulting
are likely to be critically affected by the constantly evolving local
stress eld as the fracture network develops. Simulations suggest
that, in some cases, fracture zones will not begin to develop until
adjacent through-going faults have fully formed. Further, the
orientations of these fractures will be inuenced by the evolving
Fig. 12. Damage plot showing six frames from a simulation consisting of 350 steps which showthe spatial and temporal evolution of the fractures predicted by MOPEDZ from (i) the
initial joint pattern through to (vi) the nal structure. The joints are oriented at 30
to s1. Locations AE indicated on (i) correspond to those in Fig. 5i. Simulation was carried out
with the same initial conditions as that shown in Fig. 5. All faults are left lateral.
geometry and proximity of neighbouring features within the
network. Hence, predictions of the location and orientation of
fracture zones, such as those by Maerten et al. (2002) and Mick-
lethwaite and Cox (2004) may be improved by incorporating
simulation of the constantly evolving local stress.
4.2. Orientation of the maximum compressive far-eld stress
For a simulated s
1
at a high angle to the original joints (Fig. 5),
faults are principally formed by slip on pre-existing joints which
then grow in length as wing cracks evolve and link originally
discontinuous adjacent fault traces, resulting in a stepped fault-
zone geometry. A comparison of these results with Fig. 1 shows
them to be similar to the complex fault-zone geometry in Fig. 1b
(Martel, 1990); a large number of wing cracks and linkage struc-
tures are present, at a variety of angles, with few through-going
features. Our simulations suggest that the faults from the Waterfall
region of the Sierra Nevada (Fig. 1b) were formed under s
1
at
approximately 60
to the original joints. This angle to s

1
differs
from the 25
30
derived for the same eld site using linear elastic

fracture mechanics (LEFM) by Segall and Pollard (1983) and Martel
(1997); this may be due to the assumptions inherent in LEFMwhere
failure is inferred from the steady-state stress distribution local to
a single innitely thin fracture within an innite elastic domain.
For s
1
at a low angle (30
) to the original joints, smooth

linear faults are predicted by the simulations with a small
number of linking fractures propagating back into the compres-
sional quadrant. A comparison of Fig. 12 with the eld obser-
vations shows the predicted fault geometry to be similar to that
in Fig. 1a (Pennacchioni and Mancktelow, 2007) for the small
NE-striking faults (these are labelled as fractures but offset the
aplite). We suggest that these NE-striking faults may have
formed by linkage of small joints and fractures when s
1
was
oriented approximately NNE.
Smooth co-linear fault traces such as those in Fig. 1a are
commonly observed in eld exposures. However, they are not
generally interpreted from eld data as having evolved from
linkage of co-linear joint traces, but instead are mapped as long,
small offset faults. In many cases this is likely to be because it is
difcult from eld data alone to distinguish whether individual
Fig. 13. Smoothed stress contours around the tip of a joint (or fault). (a) Orientation of the joint to the far-eld stress affects the orientation of the local stress eld around the tip
(red contours are extensional stress contours, blue are compressional stress contours). (b) Different styles of linkage structures which develop between initially co-linear joints due
to different interaction of the local stress eld. Here joints at 30
to s
1
link up approximately along strike, those at 60
develop wing cracks which later link through additional

fracturing when the local stress eld associated with the wing cracks interact. (For interpretation of the references to colour in this gure legend, the reader is referred to the web
version of this article.)
along strike sections of a fault originated as co-linear joints or
formed through along strike linkage. Multiple offset markers could
be useful because sites of linkage commonly remain as displace-
ment minima along a fault that develops by segment linkage (e.g.
Bu rgmann et al., 1994). However, the aplite dykes used as offset
markers are rare in both of the eld areas in Fig. 1. It is possible that
the fault gouge within a slipped joint will have a different miner-
alogy, grain size distribution or fabric than that along a linkage
fracture.
The faults from the Bear Creek region of the Sierra Nevada
(Fig. 1c) are a combination of frequent through-going smooth
faults and multiple wing cracks (typical of more stepped fault-
zones). We suggest that at this location one or more rotations of s
1
may have occurred during fault-zone evolution as discussed by
Segall and Pollard (1983). Rotation of s
1
may be caused by an actual
change in the regional stress eld or by more local evolution of the
stress eld during development of larger scale structures (e.g. due
to nearby linkage of faults at a scale one order of magnitude greater
than those simulated here).
5. Summary and conclusions
This paper applies a nite-element model, MOPEDZ, to inves-
tigate fault-zone evolution frommore than 20 pre-existing joints in
granite. We simulate fault-zone evolution for maximum compres-
sive far-eld stress (s
1
) at 30
and 60
to the orientation of the pre-

existing joints. Key ndings from the simulations are:
(1) Local spatial and temporal variations in the stress eld arise
from interactions between neighbouring joints that have
a signicant effect on the predicted locations, orientations and
timings of wing crack development.
(2) A clear difference in fault-zone geometry is predicted
depending on the orientation of s
1
with respect to the initial
joint pattern: for s
1
at an angle of 30
to the initial joints, co-

planar joints progressively link up along strike to form long
smooth linear faults with few additional wing cracks; for an
orientation of s
1
of 60
to the initial joints, a more fractured

complex stepped fault-zone geometry evolves with multiple
wing cracks forming linkage structures at a variety of angles.
Existing eld data show that both smooth and stepped fault-
zone trace geometries are commonly observed in crystalline
rocks.
(3) Local spatial and temporal stress perturbations affect predic-
tions of zones of enhanced fracturing within fault networks at
larger scales.
Acknowledgements
Heather Moir is supported by a University of Strathclyde Faculty
of Engineering scholarship. James Kirkpatrick is supported by the
Natural Environment Research Council (NE/E005365/1). Insightful
and thorough reviews by Steve Martel and Michael Gross and
personal conversations with Steve Martel have signicantly
improved the manuscript.
References
Bergbauer, S., Martel, S.J., 1999. Formation of joints in cooling plutons. Journal of
Bremaecker, D.J.C., Ferris, M.C., 2004. Numerical models of shear fracture propa-
gation. Engineering Fracture Mechanics 71, 21612178.
Bu rgmann, R., Pollard, D.D., Martel, S.J., 1994. Slip distributions on faults: effects of
stress gradients, inelastic deformation, heterogeneous host-rock stiffness, and
fault interaction. Journal of Structural Geology 16, 16751690.
Byerlee, J.D., 1967. Frictional characteristics of granite under high conning pres-
sure. Journal of Geophysical Research 72, 36393648.
Crider, J.G., Peacock, D.C.P., 2004. Initiation of brittle faults in the upper crust:
a review of eld observations. Journal of Structural Geology 26, 691707.
Cruikshank, K.M., Aydin, A., 1994. Role of fracture localization in arch formation at
Arches National Park, Utah. Geological Society of America Bulletin 106, 879891.
Du, Y.J., Aydin, A., 1995. Shear fracture patterns and connectivity at geometric
complexities along strike-slip faults. Journal of Geophysical Research-Solid
Earth 100 (B9), 1809318102.
Evans, J.P., Shipton, Z.K., Pachell, M.A., Lim, S.J., Robeson, K.R., 2000. The structure
and composition of exhumed faults, and their implications for seismic
processes. In: Bokelmann, G., Kovach, R.L. (Eds.), Proceedings of the 3rd
Conference on Tectonic Problems of the San Andreas Fault System, vol. 21.
Stanford University Publications, Geological Sciences, pp. 6781.
Granier, T., 1985. Origin, damping and pattern of development of faults in granite.
Tectonics 4, 721737.
Jing, L., 2003. A review of techniques, advances and outstanding issues in numerical
modelling for rock mechanics and rock engineering. International Journal of
Rock Mechanics & Mining Sciences 40, 283353.
Joussineau, G., Mutlu, O., Aydin, A., Pollard, D.D., 2007. Characterization of strike-
slip fault-splay relationships in sandstone. Journal of Structural Geology 29,
18311842.
Kattenhorn, S.A., Marshall, S.T., 2006. Fault-induced perturbed stress elds and asso-
ciated tensile and compressive deformation at fault tips in the ice shell of Europa:
implications for fault mechanics. Journal of Structural Geology 28, 22042221.
Kattenhorn, S.A., Aydin, A.A., Pollard, D.D., 2000. Joints at high angles to normal
fault strike: an explanation using 3-D numerical models of fault-perturbed
stress elds. Journal of structural Geology 22, 123.
Kim, Y.J., Peacock, D.C.P., Sanderson, D.J., 2004. Fault damage zones. Journal of
Kirkpatrick, J.D., Shipton, Z.K., Evans, J.P., Micklethwaite, S., Lim, S.J., McKillop, P.,
2008. Strike-slip fault terminations at seismogenic depths: the structure and
kinematics of the Glacier Lakes fault, Sierra Nevada United States. Journal of
Geophysical Research, B04304. doi:10.1029/2007jb005311.
B07403. doi:10.1029/2007JB005388.
Maerten, L., Gillespie, P., Pollard, D.D., 2002. Effects of local stress perturbation on
secondary fault development. Journal of Structural Geology 24, 145153.
Martel, S.J., 1990. Formation of compound strike-slip fault zones, Mount Abbot
Quadrangle, California. Journal of Structural Geology 12, 869882.
Martel, S.J., 1997. Effects of cohesive zones onsmall faults andimplications for secondary
Martel, S.J., Boger, W.A., 1998. Geometry and mechanics of secondary fracturing
around small three-dimensional faults in granitic rock. Journal of Geophysical
Research 103 (B9), 2129921314.
Martel, S.J., Pollard, D.D., Segall, P., 1988. Development of simple fault zones in
granitic rock, Mount Abbot quadrangle, Sierra Nevada, California. Geological
Society of America Bulletin 100, 14511465.
Martin, C.D., 1997. Seventeenth Canadian geotechnical colloquium: the effect of
cohesion loss and stress path on brittle rock strength. Canadian Geotechnical
Journal 34, 698725.
Mathews, J.H., Fink, K.D., 2004. Numerical Methods using Matlab, fourth ed. Pear-
son Prentice Hall, New Jersey.
Micklethwaite, S., Cox, S.F., 2004. Fault-segment rupture, aftershock-zone uid ow
and mineralization. Geology 32, 813816.
Pachell, M.A., Evans, J.P., 2002. Growth, linkage, and termination processes of
a 10-km-long strike-slip fault in jointed granite: the Gemini fault zone, Sierra
Nevada, California. Journal of Structural Geology 24, 19031924.
Pachell, M.A., Evans, J.P., Taylor, W.L., 2003. Kilometer-scale kinking of crystalline
rocks in a transpressive convergent setting, Central Sierra Nevada, California.
GSA Bulletin 115, 817831.
Peacock, D.C.P., Sanderson, D.J., 1995. Strike-slip relay ramps. Journal of Structural
Geology 17, 13511360.
Pennacchioni, G., Mancktelow, N.S., 2007. Nucleation and initial growth of a shear
zone network within compositionally and structurally heterogeneous granitoids
under amphibolitefacies conditions. Journal of Structural Geology 29, 17571780.
Segall, P., Pollard, D.D., 1980. Mechanics of discontinuous faults. Journal of
Shen, B., Stephansson, O., 1993. Numerical-analysis of mixed mode-I and mMode-II
fracture propagation. International Journal of Rock Mechanics and Mining
Sciences 30, 861867.
Tang, C.A., 1997. Numerical simulation of progressive rock failure and associated seis-
micity. International Journal of Rock Mechanics and Mining Sciences 34, 249261.
Turcotte, D., Schubert, J., 2002. Geodynamics, seconnd ed. Cambridge University
Press, Cambridge.
Vermilye, J.M., Scholz, C.H., 1998. The process zone: a microstructural view of fault
growth. Journal of Geophysical Research 103, 1222312237.
Willemse, E.J.M., Peacock, D.C.P., Aydin, A., 1997. Nucleation and growth of strike-slip
faults inlimestones fromSomerset, U.K. Journal of Structural Geology19, 14611477.
of Geophysical Research-Solid Earth 113, B07403. doi:10.1029/2007JB005388.
Geomechanical integrity of sealing faults during depressurisation of the
Statfjord Field
F. Cuisiat
a,
*
, H.P. Jostad
a
, L. Andresen
a
, E. Skurtveit
a
, E. Skomedal
b
, M. Hettema
b
, K. Lyslo
b
a
Norwegian Geotechnical Institute, P.O. Box 3930 Ullevl Stadion, N-0806 Oslo, Norway
b
Statoil, N-4035 Stavanger, Norway
Article history:
7 January 2010
Keywords:
Geomechanics
Production
Reservoir
Fault
Sealing integrity
a b s t r a c t
In this paper the results of geomechanical analyses of fault behaviour at the Statfjord Field carried out as
part of Statfjord Late Life Project are presented. The objective was to assess the sealing integrity of the
horst structure between the Statfjord and Snorre elds during nal depressurisation of the Statfjord
Field. According to eld pressure observations the Brent Fault is to date still acting as a pressure seal
between the Brent Field and the Statfjord Field, despite the large present-day depressurisation of the
Brent Field. These observations were used as a calibration and verication of the stress conditions that
can be sustained without modifying the seal integrity of the fault. Based on the calculated stress changes
in the horst structure which are equal to or less critical than the calculated present stress changes on the
Brent Fault, it is concluded that the mechanical effects associated with the planned depressurisation of
the Statfjord Field during its late life phase will not affect signicantly the hydraulic resistance of the
horst structure. A parametric study was conducted to investigate the sensitivity of the calculated stress
changes to various input parameters for fault geometry and properties. The largest uncertainty relates to
the peak shear strength of the fault (core) zone.
1. Introduction
In this paper the stress response of a major sealing fault to
a planned reduction in reservoir pressure during production is
analysed. The fault and the horst structure it bounds appear as
sealing pressure boundaries between Statfjord and the neigh-
bouring eld, Snorre, in the Norwegian North Sea. In response to
the operators concern, the main focus of the study is to investigate
the possibility to develop zones with shear or tensile failure during
production, as such zones may be subjected to signicant changes
in hydraulic resistance.
As documented in the literature, uid injection or uid with-
drawal can both induce active faulting in oil and gas reservoirs
(Grasso, 1992). Re-activation of reservoir bounding faults can cause
a loss of seal integrity (Wiprut and Zoback, 2000), and slip on active
faults can also cause shearing of production wells (Maury et al.,
1992) or wellbore instability during drilling (Willson et al., 1999).
The analyses presented in this paper are carried out within the
context of Statfjord Late Life Project, which was started by the
operator to increase gas recovery by reducing the reservoir
pressure (Boge et al., 2005). In order to assess the stress response
caused by this pressure reduction and the sealing integrity of the
horst structure, geomechanical models have been constructed to
predict production-induced stress changes during the nal phase of
eld production.
A calibration of the numerical approach is performed by calcu-
lating the stress changes in the Brent Fault due to present depres-
surisation of the Brent Field. From eld observations, the Brent
Fault, which acts as a hydraulic barrier between the Brent Field and
the Statfjord Field has not experienced any signicant change in the
hydraulic communication across it, despite the present pore pres-
sure depletion of nearly 30 MPa at the Brent Field. In our approach,
the maximum shear stress and minimum principal effective stress
due to the present pore pressure reduction at the Brent Field, are
used as an indicator of the stress condition that does not give
signicant change in the mechanical sealing integrity of the Brent
Fault. It is then argued that the properties of the Brent Fault and
horst are similar, such that the above stress condition for the Brent
Fault can be extrapolated to the horst structure during the
depressurisation phase planned for the Statfjord Field.
Two dimensional (2D) plane strain, geomechanical analyses are
carried out by looking at characteristic 2D cross sections through
a geological model of the area. It is considered that 2D analyses are
sufcient as a rst-order approximation, given the general geometry
E-mail address: fabrice.cuisiat@ngi.no (F. Cuisiat).
doi:10.1016/j.jsg.2010.01.006
of the problem and uncertainties in some of the input data (see
discussioninSection2.1). The geomechanical analyses are performed
using the numerical code Plaxis v 8.2 (2004). Plaxis is a nite element
program specially developed for geomechanical applications. In
order to apply Plaxis to reservoir-type problems, a user-denedporo-
elastic constitutive model has been implemented into Plaxis in order
to take into account the compressibility of the grains due to changes
in pore pressure (i.e. the Biot effect), as well as the compressibility of
the pore uid in shales during undrained deformation.
A critical input in the analyses is the mechanical properties of
faults. Based on results from laboratory experiments on intact and
faulted material, as well as an estimation of clay content within
fault planes, fault properties are suggested for the analyses. A key
feature that controls the stress changes in a fault zone is the degree
of drainage of the fault zone for the timescale considered. It is
shown that for the actual production time histories, the fault core
can be considered to be drained.
In light of the uncertainties related to geometries and
geomechanical properties of fault zones, a parametric study is
performed in order to identify the main parameters controlling
shear and normal stress changes and therefore, the possibility of
developing shear and tensile failure of the faults. Based on the
results of the parametric study, a base case model of the Brent Fault
is dened for geomechanical assessment of the structure.
1.1. Description of the Statfjord Field
The Statfjord Field is located in the Tampen area in the northern
part of the North Sea, between the Brent and Snorre elds (Fig. 1 top
right). The water depth in the area is about 150 m. The Statfjord
Field is located in a large, faulted block, which is tilted towards the
west and comprises a number of smaller, faulted compartments
along its east ank. Towards the south west the eld is bounded by
the Brent Fault, and in the north east a horst structure marks the
boundary with the Snorre eld (Fig. 1 bottom). Within the Statfjord
Field there are two reservoir units; the Brent Group at around
2500 m below mean sea level and the Statfjord Formation at
a depth of around 2700 mbelowmean sea level. The two reservoirs
are separated by the Dunlin Group, which consists mainly of shale.
The eld can be separated into a relatively undeformed main
eld area and an eastern ank which is heavily deformed by rota-
tional slide blocks. The main tectonic event in the area is related to
Fig. 1. Location of the Statfjord Field (top right). Regional stratigraphy of the Statfjord and Brent elds (top left). Global section (55 km wide) through the Statfjord Field including
part of the Brent Field close to the Brent Fault, the horst structure and part of the Snorre eld close to the horst structure (bottom). The section shows the seabed, Viking group (red),
Brent group (yellow), Dunlin group (light blue) and Statfjord formation (dark blue). Vertical scale exaggerated by a factor of 3 compared to the horizontal scale. The Brent Fault
between the Statfjord Field and the Brent Field and the Horst Structure between the Statfjord Field and the Snorre Field are highlighted. (For interpretation of the references to color
in this gure legend, the reader is referred to the web version of this article.)
F. Cuisiat et al. / Journal of Structural Geology 32 (2010) 17541767 1755
the opening of the Viking Graben which started in the Middle
Jurassic. Subsequent deposition of the Draupne Formation caused
gravitational instabilities along the crest of the eld and the
formation of rotationaly slide blocks in the Upper Triassic and
Jurassic sections (Hesthammer et al., 1999).
1.2. Statfjord late life project
The Statfjord Late Life project started in 2005 in order to improve
recovery from the Statfjord Field by converting the eld from an oil
eld with associated gas, into a gas eld with associated oil (Boge
et al., 2005). This is achieved with a very extensive well programme,
and modications of the platforms and associated Tampen link
pipeline to export gas. Gas export began in 2007. Production from
the Statfjord Field is expected to continue until 2020.
The production strategy for the eld was previously based on
injecting gas and water for enhanced oil recovery and reservoir
pressure maintenance. This resulted in over 60% crude oil recovery
from the stock tank oil originally in place (STOOIP). In the late life
phase of the eld, the remaining non-recoverable oil is produced
together with large volumes of previously-injected gas by reducing
pressure in the reservoirs. A pressure reduction of about 30 MPa
(300 bar) is planned, thereby reducing the pressure differences
with the Brent Field, and inducing a large planned pressure drop
through the horst connecting it with the Snorre eld.
2. Description of numerical models
In this section, the basic elements of the numerical models are
reviewed. The main input data to the geomechanical analyses
consist of:
- Geometry (reservoir layers, fault structures, intra-reservoir
shale layers, overburden and underburden)
- Initial vertical and horizontal stress, and pore pressure
distribution
- Geomechanical properties (strength and stiffness) of intact,
clay-rich materials (i.e. overburden, intra-reservoir shale
layers, fault core)
- Geomechanical properties (stiffness and strength) of intact
sand-rich materials (i.e. reservoir layers)
- Properties of faults (stiffness, strength and permeability)
- Pore pressure depletion histories
- Properties of uid in place (uid bulk modulus)
- Boundary conditions
2.1. Model geometry
The geometry of the geomechanical models is based on two-
dimensional cross-sections through a geological model based on
seismic data interpretation. Four different models were established:
one global model and three local models (i.e. Brent Fault, horst struc-
ture and a simplied model of the Brent Fault for parametric studies).
In the following, the geometry of the different models is reviewed.
The global model which is 60 km wide, extends from the Brent
Field to the Snorre eld. It is used to check the impact of boundary
conditions in smaller local models around the faults considered
(Fig. 2). The global model includes the Brent Fault separating the
Statfjord Field and the Brent Field to the south-west, and the horst
structure between the Statfjord Field and the Snorre eld towards
the northeast. The water depth varies from about 130 m in the
south-west to 330 m in the northeast.
Fig. 2. Global 2D nite element model (top) with detail of the mesh around the Brent Fault (bottom left) and horst structure (bottom right).The Brent and Statfjord reservoirs are
indicated on the gure.
Because of the variation in Brent Group thickness due to
erosion and variation in fault dip along the strike of the fault, two
vertical sections, Section 1 and Section 2, orientated roughly
northeastsouth-west through the Brent Fault, are considered.
The main geometrical characteristics of the two sections are
summarized in Table 1. The major difference between the two
sections is the thickness of the Brent Group on the footwall side of
the fault. In order to account for stiffness variation over the
heights of the reservoir layers in the geomechanical model, the
Brent Group is divided into two equally thick layers, Brent 1 (top)
and Brent 2 (bottom). Similarly the Statfjord formation is divided
into the Statfjord 1 (top) and Statfjord 2 (bottom), with thick-
nesses equal to 35% and 65% of the total unit thickness respec-
tively. The nite element model used for Section 2 is illustrated in
Fig. 3. A similar model to the Brent Fault model is used for the
sensitivity study.
Three vertical sections orientated roughly north-east south-
west through the horst structure are used in the numerical
analyses. The main differences between the three sections are the
juxtaposition window of the Statfjord Formation across the fault
zones and the thickness of the top Brent Group within the horst
structure. For brevity, only data and results from Section 1
through the horst structure and illustrated in Fig. 4 are discussed
in the paper. As for the Brent Fault model, the reservoir is divided
into two layers due to variation in material properties with
depth.
As shown in Figs. 3 and 4, simple planar fault geometries are
assumed in this study. Furthermore, only two-dimensional plane
strain models are used. This is considered reasonable given the
general elongated geometry of the elds (NESW), which means
that two-dimensional plane strain conditions should occur in the
direction orthogonal to the elongated trend (i.e. roughly NWSE
and normal to the sections chosen for the analyses). The major
assumptions underlying the models are justied by the following
observations:
a) the fault is interpreted as being fairly planar along-strike at the
scale considered here; in the absence of cross-cutting faults (i.e.
parallel to chosen sections), stress variations in the out-of-
plane are not expected and two-dimensional models can be
used.
b) Three dimensional stress redistribution close to fault ends
(along-strike) are not considered.
c) The behaviour of the faults and the general deformation mode
are largely controlled by the global deformation of the com-
pacting reservoirs (i.e. the driving forces), and the stiffness
contrast between reservoirs and surrounding formations.
In fact, since the focus of the analyses is to study local stress
conditions at the intersection of the faults and the elongated elds,
the models should be regarded as two-dimensional cross-sections
oriented normal to the fault planes. Hence, one can consider that
locally the true fault orientation is transformed to be normal to the
considered section, even though there might be some variation in
fault orientation at a larger scale.
2.2. Initial stresses and pore pressures
The in situ stress conditions are characterized by a stress regime
consistent withnormal faulting, i.e. the vertical stress is the maximum
principal stress. The stress regime is conrmed by the operators
drilling experience in the eld. The minimum horizontal stress is
obtained from extended leak-off tests (Raaen et al., 2006) and mini-
fracture tests. There is no eld evidence indicating large horizontal
stress anisotropy; hence, the two horizontal stresses are assumed
equal. Fig. 5 shows the total vertical stress, the total horizontal stress
and initial pore pressure proles as used in the geomechanical anal-
yses. The hydrostatic pressure line is included for reference.
2.3. Material properties for overburden and surrounding shale
The overburden and surrounding shale formations are modelled
as poro-elastic undrained material characterized by the following
parameters:
- Shear modulus G
- Drained Poissons ratio n
- Bulk modulus of solid grains K
s
- Porosity n
- Fluid bulk modulus K
f
The drained Poissons ratio (n) is taken equal to 0.2, i.e. repre-
sentative of low porosity shales. In fact, the value has a limited
effect in the analyses since, for shales, the excess pore pressure
induced by volumetric deformation in undrained conditions is
controlled by Skemptons coefcient B. The variation of B is small
when n varies between 0.1 and 0.3.
The bulk modulus of pore uid (K
f
) and solid (K
s
) is taken to be
2.2 GPa (i.e. brine) and 45 GPa respectively. For simplicity, the shear
modulus G is assumed to vary linearly from zero at seabed to
a given value at the top of the reservoir; this is determined fromthe
results of two undrained triaxial tests performed in NGIs soil and
rock mechanics laboratory on Viking Group and Burton Formation
shale core samples:
G 0:8D 0:145 (1)
where G is the shear modulus in GPa, and D is the true vertical
depth below mean sea level (TVD msl) in kilometres. Note that Eq.
(1) is eld specic and not general for shales.
Eq. (1) gives a higher stiffness than the empirical model
proposed by Horsrud (2001), which relates the Youngs modulus E
to the measured P-wave interval transit time Dt
p
according to:
E 0:076
304:8=Dt
p
3:23
(2)
The discrepancy may be explained by the dataset used to
establish Eq. (2), which is mostly based on younger, mechanically
weaker Tertiary age shales from the North Sea.
2.4. Material properties for reservoir layers
The reservoir layers are modelled as drained, poro-elastic
formations characterized by the shear modulus G, Poissons ratio n,
and bulk modulus of the grains K
s
. The drained bulk modulus of the
rock (K) is expressed as:
K 2G
1 n
31 2n
(3)
and Biots coefcient a as:
a 1 K=K
s
(4)
Table 1
Geometrical characteristic of two vertical sections through Brent fault used in two-
dimensional numerical analyses.
Section 1 Section 2
Average fault inclination (deg.) 54 47
Thickness of Brent Group
Footwall Hanging wall (m)
37183 204212
Thickness of Statfjord Formation
Footwall Hanging wall (m)
283319 268314
The values of K, a, G and n are inferred from triaxial tests per-
formed in the laboratory on reservoir core samples. A nearly
continuous core of some 720 m through the Brent and Dunlin
Groups, and the Statfjord Formation was taken for rock mechanic
studies. Special care was taken to minimise the damage caused by
the coring process (Hettema et al., 2002). The interpreted values of
the elastic properties are summarized in Table 2.
2.5. Properties of faults
In this section, the main geological characteristics of the faults
are briey presented, together with the approach used for model-
ling faults in the numerical analyses.
2.5.1. Fault geometry
The fault is dened by its throw, the fault core thickness, and the
thickness of the damage zones on footwall and hanging wall sides
(Sperrevik et al., 2002). Fault throw, inferred from seismic data,
together with empirical correlations between damage zone thick-
nesses, fault core thickness and fault throw developed from outcrop
studies (Beach et al., 1997, 1999), were used to estimate geometrical
properties for the Brent Fault and the faults bounding the horst
structure. These data are tabulated inTable 3 for two vertical sections
throughthe Brent Fault and inTable 4 for one vertical sectionthrough
thehorst structure. Inaddition, the ShaleGouge Ratio(SGR) proposed
by Yieldinget al. (1997), whichis a measure of the percentage of shale
or clay in the slipped interval, is given in the tables.
Fig. 3. Finite Element model of Brent Fault Section 2, with detailed view of reservoir section and fault.
Fig. 4. Details of Finite Element model for horst structure Section 1. Note the faults Z0 (Snorre side) and Z5 (Statfjord side) dening the horst. Dp is the pressure reduction
applied in the model.
Variation in the fault thickness is expected throughout the
sandstone shale sequence, with thicker damage zones observed
in competent shale units than in sandstone units (Sperrevik et al.,
2002). Signicant drag of the adjacent layers to the fault due to
shear deformation may also occur. However, in the absence of data
related to the drag, in particular of the relationship between drag
and fault throw, the drag component of deformation is not included
in the base case numerical model. As shown later in a simplied
parametric study, fault shear mobilisation is less when drag effects
are included, such that the base case is more conservative.
2.5.2. Fault mechanical properties
Shale Gouge Ratio (SGR) is used as an indication of the volume of
clay within the fault plane. This is a simplication of fault zone
complexity as SGR does not represent the detailed internal struc-
ture of a fault where clay smear, cataclasites, etc, might be present
together (Wibberley et al., 2008). Table 3 indicates that for the
Brent Group, the sandsand juxtaposition window is sealed by
a high clay concentration. Petrophysical and thin section analyses
on fault rock material from the Brent Group reservoir show most
fault rock may be classied as disaggregation zones/proto-cata-
clasites, indicating deformation at a relatively early stage in the
reservoirs burial history (Sverdrup and Bjrlykke, 1997; Fisher and
Knipe, 1998). In the Statfjord Formation, in view of the low SGR
values in the sandsand juxtaposition window (Table 3), it is
assumed that the Brent Fault zone, which is acting as a seal, consists
of a cataclastic fault rock, possibly with quartz cementation and
higher shear strength. In the horst structure, due to high SGR values
in the Statfjord Formation, low shear strength cannot be excluded
(Table 4).
The shear strength of the clay-rich fault core zone is dened
from the results of undrained triaxial tests carried out on core
samples from Viking Group and Burton Formation shale. Due to
limited data, a simple MohrCoulomb failure criterion is used. The
drained peak shear strength is found to vary between lower and
upper bounds as dened by the values of effective cohesion c
0
and
friction angle 4
0
given by (c
0
3 MPa, 4
0
24
) and (c
0
8 MPa,
4
0
23
), respectively. The lower bound is obtained from speci-

mens oriented at or close to the most critical orientation with
respect to loading. From these tests, an average shear modulus G
50
of 2.0 GPa is also found, which is used for the fault core zone (base
case).
2.5.3. Modelling of faults in nite element models
Except for the global model, the fault is modelled with
continuum elements having a nite thickness. In the global model
the faults are represented with interface elements. The results from
the parametric study show that stress changes on the fault are not
very sensitive to fault core thickness. Hence a constant thickness of
the fault core is used in the analyses despite the variation indicated
in Tables 3 and 4. Similarly, the damage zones and geometrical drag
effects are not taken into account as the parametric analyses show
a relatively small and positive effect with respect to shear mobi-
lisation of the fault.
0
500
1000
1500
2000
2500
3000
3500
0 10 20 30 40 50 60 70
T
r
u
e

V
e
r
t
i
c
a
l

D
e
p
t
h

b
e
l
o
w

M
e
a
n

S
e
a

L
e
v
e
l

(
m
)
Initial Stresses and Pore Pressure (MPa)
Brent Group
Dunlin Group
Statfjord formation
Vertical Stress
Horizontal Stress
Pore Pressure
Hydrostatic Pressure
Fig. 5. Initial vertical and horizontal stresses and pore pressure for the Statfjord Field based on pore pressure prognosis, density logs, drilling experience, reservoir pressure
measurements and mini-fracture tests (in the reservoir). The hydrostatic pressure line is included for reference. The depth in metres (vertical axis) is given as true vertical depth
(TVD) below mean sea level.
Table 2
Elastic properties of the reservoir layers inferred from over 100 triaxial tests on core
samples from well 33/9-A-37b.
Formation/group Porosity (%) G (GPa) n () K (GPa) a
Upper Brent 27 3.7 0.21 5.1 0.86
Lower Brent 26 4.5 0.19 5.8 0.84
Upper Statfjord 25 4.4 0.21 6.1 0.83
Lower Statfjord 22 5.8 0.18 7.1 0.81
Table 3
Characteristics of two sections of Brent fault through Brent Group, Dunlin and
Statfjord Formations. Data for Section 1 is written in plain, data for Section 2 in bold
letters (data from Statoil).
Formation/Group Throw (m) Fault zone
thickness (m)
Damage zone
thickness (m)
SGR
Brent 1 Eroded
140160
Eroded
0.200.25
Eroded
36
Eroded
3040
Brent 2 155165
160190
0.200.30
0.250.35
40
50
4050
3050
Dunlin 155165
150230
0.200.30
0.200.40
40
3260
>50
>50
Statfjord 1 155165
190245
0.200.30
0.350.40
40
4055
<15
<15
Statfjord 2 135165
245260
0.150.3
0.400.45
3840
3740
1530
<15
2.5.4. Are faults drained or undrained during reservoir production?
A key feature that controls stress changes within the fault in
response to pressure depletion in the reservoir sandstones is the
consolidation behaviour of the fault, i.e. the ability for uids within
the fault to ow during the timescale investigated. This can lead to
changes in pore uid pressure within the fault. If the timescale is
too short to allow uid to ow, the stress changes in the fault zone
are those associated with the undrained response of the medium.
For a uid-saturated, poro-elastic material, the relationship
between stress and strain is equivalent to an ordinary elastic
material with undrained mechanical parameters (Rice and Cleary,
1976). Such conditions can be, for instance, encountered in the
evaluation of coseismic stress and strain changes within faults
(Cocco and Rice, 2002). However, for reservoir depletion-induced
stress changes, the timescale is usually in the order of decades. For
instance, the pore pressure histories for the Brent Group and
Statfjord Formation in this study showa pressure change over circa
25 years. For these long timescales, it is shown that the fault core is
in a drained condition, given its permeability (albeit small) and
thickness.
Let us consider the case where the fault is bounded by reservoir
formations on hanging wall and footwall sides (i.e. sandsand
juxtaposition). Furthermore it is considered that from an initial
condition, the reservoir pressure is reduced on one side only
(Fig. 6). The permeability of the reservoir formations is much
greater than that of the fault core material so that the reservoir
formations can be considered fully-drained. In response to the
pressure reduction, uid ow takes place across the fault, and the
uid pressure within the fault evolves towards steady state ow
conditions at a rate depending on fault thickness and uid diffu-
sivity (Fig. 6). Under steady state owconditions, the pore pressure
varies linearly across the fault (assuming uniform permeability
within the fault).
The time before reaching almost 100% consolidation (i.e. steady
state ow conditions) is given by:
t
100
zL
2
=c s (5)
where L is the length of the drainage path, estimated to be about
the thickness of the fault core zone, and c is the diffusivity (or
consolidation coefcient; Rice and Cleary, 1976) given by:
c
kK
u
KK 4=3G
m
f
$a
2
K
u
4=3G
m
2
=s
(6)
where: k Intrinsic fault core permeability (m
2
) K
u
Undrained bulk
modulus (MPa) K Drained bulk modulus (MPa) G Shear modulus
(MPa) m
f
Dynamic viscosity (MPa s) a Biot coefcient.
The Biot coefcient is given by Eq. (4). The undrained bulk
modulus can be inferred from the BiotGassmann relationship
(Mavko et al., 1998; Hettema and de Pater, 1998) written as:
K
u
K
a
2
K
s
K
f
nK
s
a nK
f
(7)
where K
s
is the bulk modulus of the solid grains constituting the
fault core, K
f
is the uid bulk modulus and n the fault core porosity.
Fault core specimens are rarely available for specic offshore
elds. In the absence of fault core material from the Statfjord Field,
the results from laboratory measurements of permeability and
diffusion coefcient on intact samples of Burton Formation shale
(Dunlin Group) were used to estimate the consolidation time in the
fault core. For a rst-order estimate, an average of the four tests
made parallel and normal to the lamination is used, disregarding
the effect of anisotropy. This gives a permeability (k) and diffusivity
coefcient (c) equal to 3.8 10
21
m
2
and 4.75 10
8
m
2
/s (1.5 m
2
/
yr), respectively.
Assuming a fault core zone thickness between 0.2 mand 0.45 m,
Eq. (5) gives t
100
between 10 and 50 days. Hence, in view of the
reservoir production timescale, segments of the fault core zone
which juxtapose sand-rich reservoirs can be considered fully
drained, with a pore pressure distribution through the fault (core)
zone varying approximately linearly between the uid pressures in
the adjoining formations.
Table 4
Characteristics of Section 1 of horst structure through Brent Group, Dunlin and
Statfjord Formation. Data for Fault Z0 is written in plain, data for Fault Z5 in bold
letters. Data from Statoil.
Formation/group Throw (m) Fault zone
thickness (m)
Damage zone
thickness (m)
SGR
Brent 1 6578
5665
0.050.15
0.050.10
28
24
<35
1025
Brent 2 6572
5662
0.050.15
0.050.10
28
24
>35
2550
Dunlin 6594
6070
0.050.15
0.100.15
32
30
>50
>50
Statfjord 1 9194
6078
0.050.15
0.100.15
32
30
>50
>40
Statfjord 2 2694
78102
0.050.15
0.100.20
1832
32
>50
>40
Fig. 6. Consolidation in a fault. The pressure in the reservoir on the left-hand side of the fault is depleted by Dp, while the pressure in the right hand side reservoir is unchanged. c is
the diffusivity coefcient of the fault, t is time since depletion started, p
o
the initial pressure and p
ss
the steady state pressure distribution in the fault (i.e. at end of consolidation).
Note that t
100
is inversely proportional to the permeability (k) of
the fault core. The permeability of the fault core zone is highly
uncertain, but a value much smaller than that of the intact shale is
thought to be unrealistic. Even if the fault core permeability is ten
times smaller than the value assumed above, the time to reach full
consolidation (100500 days) is still smaller than production time
of several decades.
For sandclay juxtaposition, i.e. for segments of the fault where
the reservoir is juxtaposed to (undrained) shale-rich formations, it
is assumed that the fault is fully drained, with pressure in the fault
equal to the reservoir pressure.
2.6. Pore pressure history used for modelling
The pore pressure histories used in the study are obtained from
history matched reservoir simulations using the black-oil reservoir
simulation tool Eclipse.
For modelling of the Brent Fault, pore pressure histories from
2005 (present day situation) are used. The resulting pore pres-
sure reductions on both sides of the Brent Fault in the Brent Group
and Statfjord Formation are given in Table 5. For modelling of the
horst structure, pore pressure prognoses for the year 2020 (Late
life) are considered. The resulting pore pressure reductions at both
sides of the horst structure in the Brent Group and Statfjord
Formation are given in Table 6.
The shale formation is assumed undrained during the depletion,
i.e. no pore uid ow takes place. The pore pressure distribution
through the fault (core) zone is assumed to vary linearly between
the pressures in the reservoir layer at both sides of the fault. The
pore pressure in the damage zone is assumed to be equal to the
pore pressure in the reservoir layer on the same side of the fault
zone. The pore pressure within the horst structure is assumed to be
constant during reservoir depletion.
2.7. Boundary conditions
Boundary conditions for all models are such that displacements
along the bottom boundary are fully-xed, the top boundary is free
to move, while displacements along the side boundaries are xed in
the horizontal direction and free in the vertical direction.
3. Results from global analysis and parametric study
3.1. Global model
The global 2D model is used to assess the effects of boundary
conditions in local models on the stress changes in the fault zone
during depletion.
Acomparison between a local and a global model shows that the
displacement eld is inuenced by the close boundary conditions
in a local model (Fig. 7). However, the shear and normal stresses
along the fault are unaffected by the boundary conditions in the
local model (Fig. 8). Hence, our local models may be used to study
the stress changes in the fault during reservoir depletion, without
signicant effect of the close boundary conditions.
3.2. Parametric study of stress changes in fault core zone
The objective of the parametric study is to investigate the
sensitivity of the stress response in a fault core zone to variations in
fault geometry and material stiffness parameters. The two-
dimensional nite element model used for the parametric analysis
resembles the Brent Fault model (Section 2) at the Statfjord
Formation depth.
From a base case scenario, variation of several parameters has
been performed which can be grouped into:
- reservoir stiffness properties;
- overburden and intra-reservoir shale stiffness properties;
- fault geometry (inclination, thickness, drag and juxtaposition);
- pressure distribution and drainage of fault core zone.
The results presented in Table 7 are analysed in terms of
maximumshear stress s
max
s
1
s
3
=2 and effective octahedral
stress s
0
oct
s
0
1
s
0
2
s
0
3
=3 in the critical location of the fault
core zone. The results are averaged from the ten most critical
integration points representing approximately 10 m of the fault
length.
The closeness of the stress state to a MohrCoulomb failure line
may be dened as the degree of shear mobilisation or CF (e.g.
Templeton and Rice, 2008) given as:
Table 5
Pore pressure reduction applied during modelling of the Brent fault (pressure
histories for 2005).
Pressure reduction Dp (MPa) Brent Field Statfjord Field
Brent Group 30 8
Statfjord Formation 26 7
Table 6
Pore pressure reduction applied during modelling of the horst structure (pressure
histories for year 2020).
Pressure reduction Dp (MPa) Statfjord Field Snorre Field
Brent Group 20 0
Statfjord Formation. 20 10
Fig. 7. Calculated contours of horizontal displacements (in centimetres) around the Brent Fault at Year 2005 (Present day conditions). Global 2D model (Far BC) and local model
(Close BC). Note that color contours on the left and right gures have a slightly different scale.
CF
s
max
sin f
0
$a s
0
m
(8)
where s
0
m
s
0
1
s
0
3
=2 is the in-plane mean effective stress, a the
attraction and f
0
the effective friction angle of the fault core
material. The attraction is related to the friction angle and the
cohesion c
0
by a c
0
=tan f
0
. By using the values of cohesion and
friction angle for clay/shale material presented in Section 2.5
(c
0
3 MPa, 4
0
24
), CF can be used to assess the positive or

negative effect of varying one parameter with respect to shear
mobilisation of the fault core. Note that the MohrCoulomb shear
failure line corresponds to CF 1. Since elastic analyses are per-
formed (due to uncertainties in the strength parameters), values for
CF in excess of 1 are possible.
Before depletion, s
max
and s
0
oct
are related to the in situ effective
vertical stress s
0
v
and horizontal stress s
0
h
by:
(
s
max

1
2
s
0
V
s
0
h
s
0
oct

1
3
s
0
V
2s
0
h
(9)
After reservoir depletion, the maximum shear stress in the fault
core zone is concentrated in the area with sandsand juxtaposition
(Fig. 9); the most critical point being at the bottom of the depleted
reservoir layer, except for a reverse fault (Case 12 in Table 7) where
the most critical point is located at the top of the depleted reservoir.
The degree of mobilisation (CF) increases from circa 0.6 to 1.3 (Base
case) after depletion, indicating that shear failure may occur in the
fault core zone. Note that the value of CF and the occurrence of
shear failure depend on the actual shear strength. In fact, for the
higher strength estimate, CF increases from circa 0.5 initially to 0.9
after depletion; this indicates no shear failure at all.
28
29
30
31
32
33
34
35
12 14 16 18 20 22
Effective normal stress fault [ MPa]
]
m
[

h
t
p
e
D
CLOSE BC
FAR BC
28
29
30
31
32
33
34
35
4 6 8 10 12
Shear stress fault [ MPa ]
]
m
[

h
t
p
e
D
CLOSE BC
FAR BC
Brent gr. Brent gr.
Dunlin fm. Dunlin fm.
Statfjord fm. Statfjord fm.

0
0
1

x

0
0
1

x
Fig. 8. Calculated effective normal and tangential (shear) stress along the Brent Fault (at the Brent Field side of the fault) at Year 2005. Global 2D model (Far BC) and local model
(Close BC).
Table 7
Stress conditions at critical point in fault during reservoir depletion results from
parametric study. Figures in bold indicate an increase in shear stress mobilisation
after variation of one input parameter from base case. E
res
: Youngs modulus of
reservoir layers, n
res
: Poissons ratio of reservoir layers, E
shale
: Youngs modulus of
shale material, E
fault
: Youngs modulus of fault, FZT: fault zone thickness, L
juxt
: Sand
sand juxtaposition length, b: fault inclination.
Case ID Description s
max
(MPa) s
0
oct
(MPa) CF
Initial conditions before depletion 6.7 16.6 0.64
1 Base case
E
res
10 GPa, n
res
0.2,
E
shale
4 GPa,
E
fault
4 GPa, FZT 10 m,
L
juxt
50 m, b 45
13.1 13.6 1.30

Rock properties
3 E
res
5 GPa 13.7 15.6 1.25
5 n
res
0.4 10.6 23.6 0.77
5b n
res
0.1 13.1 12.8 1.35
4 E
shale
8 GPa 12.7 16.3 1.14
Fault geometry and properties
6 FZT 20 m 12.6 14.0 1.24
7 FZT 5 m 12.7 13.2 1.29
9 E
fault
8 GPa 14.1 11.9 1.49
10 b 90
(vertical) 12.3 13.5 1.24

11 b 60
12.5 13.2 1.27

12 b 45
(reverse fault) 11.1 19.7 0.91

18 Presence of damage
zones (thick 20 m,
E
damage
20 GPa)
12.5 13.5 1.26
13 Reservoir juxtaposition
L
juxt
100 m
13.2 13.9 1.30
14 Drag of reservoir formations 11.9 16.7 1.07
Pressure distribution
2 Undrained fault 13.6 15.9 1.23
16 Fully depleted fault 15.5 23.9 1.06
17 Effect of differential pressure
across the fault
18.0 15.9 1.55
Fig. 9. Parametric study of maximum shear stress distribution in fault core zone. Base
case result. Contours of maximum shear stress inside the fault core zone due to
a 20 MPa pore pressure reduction in the footwall side of the fault. Maximum shear
stress concentration occurs at the bottom of the 50 m sandsand juxtaposition
window.
The maximum shear stress and effective octahedral stress
changes in the fault core zone as calculated from the 2D analyses
are relatively insensitive to variations in geometry and stiffness
parameters. From the initial stress values of s
max,0
6.7 MPa and
s
0
oct,0
16.6 MPa, the maximum shear stress and effective octa-
hedral stress fall within a range s
max
1214 MPa and s
0
oct
12
17 MPa after 20 MPa depletion for the majority of cases (Table 7).
Furthermore, the parametric study shows that:
- The undrained behaviour of the fault core zone is less critical
than a drained one (Case 2 versus Base case). Furthermore, as
noted earlier, fully-drained behaviour is more relevant in view
of production timescales and fault core permeability.
- The degree of shear mobilisation in the fault core zone is
relatively insensitive to lower values of Poissons ratio in the
reservoir rocks (Case 5b), whereas higher values of Poissons
ratio lead to a reduction in shear mobilisation.
- The variation of fault core zone thickness (Cases 6 and 7)
results in only minor stress changes, mostly due to slight
difference in nite element discretization between the
numerical models. This can be expected as long as the fault
zone is very thin compared to the thickness of the reservoir
layers and the fault stiffness is not signicantly lower than the
stiffness of the reservoir layers. The displacement eld is then
governed by the stiffness of the surrounding material rather
than the fault properties. Hence, there is no signicant conse-
quence of using an unrealistic (e.g. too large) fault core zone
thickness. Furthermore, the exact thickness of the fault core
zone does not need to be known, and variations of the fault
core zone width along the fault plane do not need to be
modelled in detail.
- The potential for shear failure increases for a stiffer fault (Case 9).
- Fault dip (Cases 1012) has a small effect on the fault stress
response.
- Sand juxtaposition has little effect on the fault stress response
and degree of shear mobilisation (Case 13).
- The effect of drag of reservoir formations along the fault plane
(Case 14) is signicant, but contributes to a decrease in the
potential for shear stress mobilisation on the fault plane; hence
it can be neglected for a rst-order approach.
- When pore pressure changes in the fault core zone are consid-
ered (Case 16, with equal pore pressure in the depleted reservoir
and the fault core zone), the degree of shear mobilisation mostly
decreases due to an increase in the effective octahedral stress.
- The degree of shear mobilisation depends on the actual pore
pressure history on both sides of the fault and not only on the
pressure difference across the fault. This is illustrated by Case 17
where a differential pore pressure of 20 MPa is imposed, with
a depletion of 30 MPa on the footwall side and a depletion of
10 MPa on the hanging wall side and in the fault core zone. This
case, which is more critical than a depletion on only one side of
20 MPa, is also more realistic for the present day situation for the
Brent Fault when compared to the Base case model.
- The presence of a stiffer damage zone (Case 18) represented
with a constant thickness of 20 mand a Youngs modulus twice
as that of the reservoir sandstone, has a small positive impact
on the degree of shear mobilisation. However, the consequence
of damage zone and fault zone complexity for uid behaviour,
and the coupling between stress changes and hydraulic
changes, is well-documented in the literature (e.g. Odling et al.,
2004; Fisher and Knipe, 2001; Zhang and Sanderson, 1998).
Assessing this is, however, beyond the scope of this study.
Fig. 10. Effective principal stresses (crosses) calculated at the year 2005 (todays situation) around the Brent Fault for Section 1 (left) and Section 2 (right). Pore pressure reduction
of 26 MPa (Statfjord formation) and 30 MPa (Brent group) in the Brent Field, of 7 MPa (Statfjord formation) and 8 MPa (Brent group) in the Statfjord Field. Note that the purpose of
the gure is to illustrate principal stress re-orientation rather than stress magnitudes for which contour plots should be used. Hence no scale is linked to the crosses.
Fig. 11. Contours of maximum shear stress (416 MPa with increments of 1 MPa). Section 1 (left) and Section 2 (right) through the Brent Fault. Zones with shear stresses less than
4 MPa have no contours.
3.3. Shear- or tension-induced softening behaviour
The results of the parametric study show, during depletion of the
reservoir, that the maximum shear stress in the fault core zone may
reach the peak shear strength of the material constituting the fault
core zone. Redistribution of maximumshear stress on the fault plane
due to brittle softening may lead to propagation of failure zones
along the fault. Similarly, propagation of tensile failure zones may
occur due to effective minor principal stress softening (tension cut-
off). These effects are analysed in two dedicated cases where; a)
residual shear strength parameters (4
res
15
, c
res
3 MPa) are
introduced in a predened area of the fault core zone and b) a Mohr
Coulomb model with tension cut-off T 0 MPa is utilised. The details
of the modelling are not presented here due to space limitation. The
main conclusions from these two cases are: 1) if shear failure is
initiated, the failure zone propagates along the fault dip direction
rather than in the cross-fault direction and 2) if tension failure occurs
it initiates on the non-depleted side of the fault core zone. Propa-
gation of tension failure through the fault is not feasible because the
fault is drained and the effective mean stress increases on the
depleted side of the fault core zone. Furthermore, it is found that
there is no signicant tendency for a tension zone to propagate along
the fault dip direction on the non-depleted side.
4. Stress conditions of the Brent Fault and horst structure
4.1. Present stress conditions of the Brent Fault
Fig. 10 shows the calculated effective principal stresses at the
year 2005 around the Brent Fault in Section 1 (left) and Section 2
(right). There is little rotation of the principal stresses along the
fault. The maximum shear stress after pressure reduction is equal
to circa 16 MPa at the most critical points located in Section 2 at
the bottom of the Statfjord Formation or Brent Group (Fig. 11
right, juxtaposition zones of Brent_2 to Brent_1 and Statfjord_2
to Statfjord_1). The corresponding maximum shear stress is
about 10 MPa (Brent Group) and 14 MPa (Statfjord Formation) in
Section 1.
The stress paths for two critical points in the fault core zone for
both Section 1 and Section 2 are shown in Fig. 12, together with the
upper and lower bounds for the peak shear strength of the shale.
During pressure reduction in the reservoir, the critical point within
the Brent Fault moves towards shear failure. Shear failure may be
expected in the Brent Group as the shear failure criterion is
exceeded, especially for the low shear strength estimate. Shear
failure is less probable in the Statfjord Formation, although the
modelling indicates stress levels in excess of the shear failure
criterion. This is because the actual strength for the Statfjord
Formation is probably signicantly higher than the estimate
plotted on the gure due to its low clay content and presumably
cataclastic fault rock material. Tensile failure may also occur as
indicated by the negative minimum principal stress (i.e. tensile
stress) at the critical points in Section 2.
4.2. Stress conditions within the horst structure during the late life
of the Statfjord Field
The results for the horst structure are shown for Section 1. This is
found to be the most critical section investigated. Only the stress
state at the end of production from the Statfjord Field (end of year
2020) is considered.
Critical stress paths in Brent Fault
0
5
10
15
20
25
30
35
-2 0 2 4 6 8 10 12
Effective min. principal stress (MPa)
p
i
c
n
i
r
p
.
x
a
m
e
v
i
t
c
e
f
f
E
a
)
a
P
M
(
s
s
e
r
t
s
Sec1,Statfjord fm.
Sec1,Brent fm.
Sec2, Statfjord fm.
Sec2, Brent fm.
Low strength
high strength
Critical stress paths in Brent Fault
0
2
4
6
8
10
12
14
16
18
0 2 4 6 8 10 12 14 16 18 20 22
Effective mean stress (MPa)
)
a
P
M
(
s
s
e
r
t
s
r
a
e
h
s
.
x
a
M
Sec1, Statfjord fm.
Sec1, Brent fm.
Sec2, Statfjord fm.
Sec2, Brent fm.
high strength
Low strength
Fig. 12. Stress paths for critical points in Sections 1 and 2 through the Brent Fault. Effective mean stress vs. maximum shear stress (left) and minimum principal effective stress vs.
maximum principal effective stress (right). Upper and lower bounds of shear strength are indicated for the Brent Group.
Fig. 13. Calculated effective principal stress crosses at the top of the horst structure (left) and at Fault Z5 in the juxtaposition of the Statfjord Formation (right) for Section 1. Note
that the purpose of the gure is to illustrate principal stress re-orientation rather than stress magnitudes for which contour plots should be used. Hence no scale is linked to the
crosses.
Fig. 13 (left) shows the calculated principal effective stress
redistribution after pressure reduction at the top of the section
through the horst structure. Fig. 13 (right) shows the same for Fault
Z5 (Statfjord side of the horst), around the juxtaposition windowof
the Statfjord Formation (i.e. area where Statfjord_1 juxtaposes to
Statfjord_2). Vertical tensile cracks may develop at the top sand-
stone within the horst structure, as the minimum effective prin-
cipal stress is close to zero and its orientation is nearly horizontal.
The extent of the zone with tensile stresses increases with the pore
pressure depletion in the Statfjord Field (Fig. 14). The development
of tensile stresses along the fault only occurs inside the horst
structure. Tensile stresses are prevented fromdeveloping inside the
fault towards the Statfjord Field due to the drained behaviour of the
fault and effective stress increase during depletion. However,
a zone with tensile stresses also develops from a singular point
above the horst structure. In this area, a continuous drainage path
would not be expected since tensile stresses are prevented from
developing to the west of the horst structure (i.e. the depleted
Statfjord Field side).
Stress paths for two critical points in Fault Z5 show that shear
failure may occur in the Brent Group and the Statfjord Formation, as
the low strength failure criterion is exceeded (Fig. 15). Tensile
failure may also occur in the Brent Group as indicated by the
negative minimum principal stress (i.e. tensile stress) at the critical
points. The maximum shear stress is less than 10 MPa and
concentrated in the juxtaposition windows (Fig. 16).
5. Discussion
5.1. Present stress conditions associated with the Brent Fault
Numerical analyses of the present (Year 2005) stress condition
in the Brent Fault with signicantly larger pore pressure depletion
in the Brent Field compared to the Statfjord Field shows that shear
failure may have occurred in the fault at the juxtaposition window
for the Brent Group (i.e. Brent_2 against Brent_1 in Fig. 11). The
peak shear strength in this zone is assumed to be similar to that of
the Burton Formation shale, which is between 10 MPa and 18 MPa
for an effective octahedral stress between 14 MPa and 20 MPa. This
strength depends rst of all on the orientation of the lamination
compared to the orientation of the critical shear stress. The lowest
strength is obtained when the critical stress is oriented parallel to
the lamination. The actual orientation of the lamination within the
fault core zones is, however, unknown.
As shown by the modelling, if shear failure occurs, it is initiated
on the Statfjord Field side of the Brent Fault, i.e. the side with the
lowest pore pressure depletion and thus also the lowest effective
octahedral stress and corresponding strength. When the strength is
reduced to the residual strength (strain softening), the failure zone
may propagate along the fault but not across the fault. Tensile
failure may occur on one side of the fault in areas of low effective
horizontal stress (i.e. with the highest pore pressure). However, it
cannot propagate across the fault zone due to the increased
Fig. 14. Calculated zones with tensile stresses in Section 1 through the horst structure. Pore pressure depletion of 20 MPa (left) and 30 MPa (right) in the Statfjord Field side of the
horst structure.
Stresses in Z5 fault
Section 1
0
2
4
6
8
10
12
14
0 2 4 6 8 10 12 14
Effective mean stress (MPa)
)
a
P
M
(

s
s
e
r
t
s

r
a
e
h
s

.
x
a
M
Statfjord fm.
Brent fm.
Low strength
High strength
Stresses in Z5 fault
Section 1
0
5
10
15
20
25
-2 0 2 4 6 8 10 12
Effective min. principal stress (MPa)

l
a
p
i
c
n
i
r
p

.
x
a
m

e
v
i
t
c
e
f
f
E
)
a
P
M
(

s
s
e
r
t
s
Statfjord fm.
Brent fm.
Low strength
Fig. 15. Stress paths for critical points in Fault Z5, Section 1 through horst structure. Effective mean stress vs. maximum shear stress (left) and minimum principal effective stress vs.
maximum principal effective stress (right). Upper and lower bounds of shear strength for the Brent Group are indicated.
effective horizontal stress on the other side (i.e. depleted side with
the lowest pore pressure). Hence, a continuous zone with shear or
tensile failure cannot develop across the fault core zone. As a result,
only minor changes in hydraulic properties of the fault zone are
expected as a result of present stress changes caused by pore
pressure depletion on the Brent Field side. This conclusion agrees
well with eld observations, indicating no signicant changes in
the seal integrity of the Brent Fault, which currently acts as
a hydraulic barrier between the Brent and Statfjord Fields.
5.2. Late life stress conditions on the horst structure
Analyses of the horst structure show similar stress changes in
the fault zone between the Statfjord Field and the horst structure
(Fault Z5) as identied for the Brent Fault. However, the corre-
sponding maximum shear stress in Fault Z5 is generally smaller
than in the Brent Fault, such that the predicted late life stress
situation is considered to be less critical than that already expe-
rienced by the Brent Fault. Hence, if mechanical and sealing
properties for the two faults are similar, it can be concluded that
the sealing integrity of the horst structure should not be altered
signicantly after full depletion of the Statfjord Field. Further-
more, the results of the modelling show that if failure takes place
in the horst structure, then the failure zones (either shear or
tensile failure) would propagate along and not across the fault (at
the side with lowest or no depletion). The development of
microcracks (and/or opening of pre-existing microcracks) is also
inhibited on the depleted side of the horst structure (towards
Statfjord) due to effective octahedral stress increase which occurs
during depletion.
It should be pointed out that only mechanical effects due to
effective stress changes were considered in this study. Other
mechanisms (e.g. capillary effects) might contribute to change the
hydraulic resistance of faults during pressure depletion.
6. Conclusions
We have presented the results of geomechanical analyses of
fault behaviour at the Statfjord Field as part of Statfjord Late Life
project. The objective was to assess the potential for developing
hydraulic communication between the Statfjord and Snorre elds
through a horst structure, during nal depressurisation of the
Statfjord Field. Two-dimensional plane strain geomechanical
analyses were carried out to calculate the deformation and
maximum shear stresses along the faults bounding the horst
structure, resulting fromcompaction and horizontal deformation of
the Statfjord Field due to pore pressure reduction.
The stress conditions at the Brent Fault separating the Brent and
Statfjord Fields were rst considered. According to pressure data,
the fault is acting as a pressure seal between the two elds. The
results of the modelling show that the calculated stress changes in
the horst structure are equal to or less critical than the calculated
present stress changes in the Brent Fault. It is therefore concluded
that the mechanical effects (i.e. stress changes) associated with the
planned depressurisation of the Statfjord Field during late life will
not affect signicantly the hydraulic resistance of the horst
structure.
Only two-dimensional plane strain models and simple fault
geometries were considered in this study, as the focus of the
analyses was to identify failure modes and mechanisms rather than
to predict absolute values of stress changes. The error by using
a two-dimensional approach instead of three-dimensional model-
ling is less critical as the results are used quantitatively to compare
two faults under similar conditions. This approach would not have
been valid if the geometry of one fault had been very different to
that of the other.
In light of the complexity and uncertainty associated with fault
parameters, a parametric study was conducted to investigate the
sensitivity of the modelled stress changes. This modelling can help
to test several geological scenarios (e.g. presence of drag, thickness
of fault) and assess the relative importance of particular mecha-
nisms (e.g. drainage conditions). It is found that the maximum
stress changes are not very sensitive to geometrical variations and
uncertainties in stiffness distributions. The largest uncertainty
relates to the peak shear strength of the fault (core) zone.
It should be pointed out that only the mechanical effects due to
effective stress changes were considered in this study. Other
mechanisms (e.g. capillary effects) might contribute to variations in
the hydraulic resistance of faults during pressure depletion.
Furthermore, the faults were modelled as single plane of weak-
nesses, which is clearly an over-simplication. Although the impact
of damage zone was not properly modelled in this study, it was
shown that its impact might not be so important for maximum
shear stress distribution in the fault.
The integrity of the horst structure, as presented in this paper,
was assessed relatively to the situation at the Brent Fault. Absolute
predictions in terms of changes in hydraulic resistance were
beyond the scope of this study. Further work would be required to
develop petrophysical models which could relate the hydraulic
properties of faults (e.g. permeability, capillary entry pressure) to
mechanical changes (e.g. fault dilation, grain and pore volume,
fracturing, tortuosity).
Instead, the approach chosen in this study assumed that the
sealing conditions of the Brent Fault could be used to calibrate the
methodology applied to the horst structure. This assumption is
supported by geological understanding of the structures. However,
the biggest uncertainty pertained to the strength of fault zones.
Without specic fault data, but from expected clay content and
deformation products in the fault zone, this study suggests that it is
acceptable to use the residual shear strength of shales as repre-
sentative of the strength of the fault (core) zone. Further work is
required to improve the determination of the strength of faults. For
instance, a systematic database could be developed by testing the
strength, stiffness and hydraulic properties of fault core samples in
the laboratory, for various fault deformation products sampled
from core from offshore elds. In that way, reliable data for future
analyses could be gathered.
Fig. 16. Contours of maximum shear stress in MPa (from 6 to 12 MPa with increments
of 1 MPa) for Section 1 through the horst structure at Year 2020 (Base Case). Areas with
maximum values are indicated with gures and arrows.
Acknowledgements
The authors acknowledge the license partners of the Statfjord
Field (Statoil, ExxonMobil, Shell, ConocoPhilips, Enterprise oil and
Centrica) for their permission to publish this paper. We thank Rune
Holt and Dave Dewhurst for reviewing the manuscript and
contributing with valuable comments, and Chris Jackson for nal
editorial comments.
References
Beach, A., Brown, J.L., Welbon, A.I., McCallum, J.E., Brockbank, P.J., Knott, S.D., 1997.
Characteristics of fault zones in sandstones from NW England: application to
fault transmissibility. In: Meadows, N.S., Trueblood, S.P., Hardman, R., Cowan, G.
(Eds.), Petroleum Geology of the Irish Sea and Adjacent Areas. Geological
Society, London, Special Publications, vol. 124, pp. 315324.
Beach, A., Welbon, A.I., Brockbank, P.J., McCallum, J.E., 1999. Reservoir damage
around faults: outcrop examples form the Suez rift. Petroleum Geoscience 5,
109116.
Boge, R., Lien, S.K., Gjesdal, A., Hansen, A.G., 2005. Turning a North Sea Oil Giant into
a Gas Field Depressurization of the Statfjord Field SPE 96403, Offshore Europe,
69 September, Aberdeen, United Kingdom.
Cocco, M., Rice, J.R., 2002. Pore pressure and poroelasticity effects in Coulomb stress
analysis of earthquake interactions. Journal of Geophysical Research in Solid
Earth 107, 2030.
Fisher, Q.J., Knipe, R.J., 1998. Fault sealing processes in siliciclastic sediments. In:
Jones, G., Fisher, Q.J., Knipe, R.J. (Eds.), Faulting, Fault Sealing and Fluid Flow in
Hydrocarbon Reservoirs. Geological Society, London, Special Publications, vol.
147, pp. 117134.
Fisher, Q.J., Knipe, R.J., 2001. The permeability of faults within siliciclastic petroleum
reservoirs of the North Sea and Norwegian Continental Shelf. Marine and
Petroleum Geology 18, 10631081.
Grasso, J.R., 1992. Mechanics of seismic instabilities induced by the recovery of
hydrocarbons. Pure and Applied Geophysics 139, 507534.
Hesthammer, J., Jourdan, C.A., Nielsen, P.E., Ekern, T.E., Gibbons, K.A., 1999. A tec-
tonostratigraphic framework for the Statfjord Field, northern North Sea.
Petroleum Geoscience 5, 241256.
Hettema, M.H.H., de Pater, C.J., 1998. The poromechanical behaviour of Felser
Sandstone stress and temperature-dependent SPE/ISRM 47270, SPE/ISRM
Rock Mechanics in Petroleum Engineering, 810 (July, Trondheim, Norway).
Hettema, M.H.H., Hanssen, T.H., Jones, B.L., 2002. Minimizing Coring-Induced
Damage in Consolidated Rock. SPE/ISRM 78156, SPE/ISRM Rock Mechanics
Conference, 2023 October, Irving, Texas.
Horsrud, P., 2001. Estimating mechanical properties of shale from empirical
correlations. SPE 56017. SPE Drilling and Completion 16, 6873.
Maury, V.M.R., Grasso, J.R., Wittlinger, G., 1992. Monitoring of subsidence and
induced seismicity in the lacq gas-eld (France) the consequences on gas-
production and eld operation. Engineering Geology 32, 123135.
Mavko, G., Mukerji, T., Dvorkin, J., 1998. The Rock Physics Handbook. Tools for
Seismic Analysis in Porous Media. Cambridge University Press.
Odling, N.E., Harris, S.D., Knipe, R.J., 2004. Permeability scaling properties of
fault damage zones in siliclastic rocks. Journal of Structural Geology 26,
17271747.
Plaxis Users Manual. Plaxis 2D Prof. v.8.2, Build 811, 2004. www.plaxis.nl.
Raaen, A.M., Horsrud, P., Kjorholt, H., Okland, D., 2006. Improved routine esti-
mation of the minimum horizontal stress component from extended leak-off
tests. International Journal of Rock Mechanics and Mining Sciences 43,
3748.
Rice, J.R., Cleary, M.P., 1976. Some basic stress-diffusion solutions for uid-saturated
elastic porous media with compressible constituents. Reviews of Geophysics
and Space Physics 14, 227241.
Sperrevik, S., Gillespie, P.A., Fisher, J.F., Halvorsen, T., Knipe, R., 2002. Empirical
Estimation of Fault Rock Properties. In: Norwegian Petroleum Society (NPF)
Special Publication, vol. 11 109125.
Sverdrup, E., Bjrlykke, K., 1997. Fault properties and the development of cemented
fault zones in sedimentary basins: eld examples and predictive models. In:
Mller-Pedersen, P., Koestler, A.G. (Eds.), Hydrocarbon Seals: Importance for
Exploration and Production. Norwegian Petroleum Society (NPF) Special
Publications, vol. 7, pp. 91106.
Templeton, E.L., Rice, J.R., 2008. Off-fault plasticity and earthquake rupture
dynamics: 1. Dry materials or neglect of uid pressure changes. Journal of
Geophysical Research 113, B09306.
Geological Society, London, Special Publications, vol. 299, pp. 533.
Willson, S.M., Last, N.C., Zoback, M.D., Moos, D., 1999. Drilling in South America:
a wellbore stability approach for complex geologic conditions SPE 53940, Latin
American and Caribbean Petroleum Engineering Conference, 2123 April,
Caracas, Venezuela.
Wiprut, D., Zoback, M.D., 2000. Fault reactivation and uid ow along a previ-
ously dormant normal fault in the northern North Sea. Geology 28,
595598.
American Association of Petroleum Geologists Bulletin 81, 897917.
Zhang, X., Sanderson, D.J., 1998. Numerical study of critical behaviour of deforma-
tion and permeability of fractured rock masses. Marine and Petroleum Geology
15, 535548.
Structural controls on leakage from a natural CO
2
geologic storage site:
Central Utah, U.S.A.
Ben Dockrill
a,
*
, Zoe K. Shipton
b
a
Department of Geology, Trinity College Dublin, Dublin 2, Ireland.
b
Department of Geographical and Earth Sciences, University of Glasgow, Glasgow G12 8QQ, Scotland
Article history:
Received 26 May 2009
26 November 2009
Keywords:
Fault zone
Damage zone
Fluid ow
Co
2
Hydrocarbons
a b s t r a c t
Faults and associated fracture networks can signicantly inuence regional ow of groundwater,
hydrocarbons and other uids. The distribution of CO
2
springs and seeps along the Little Grand Wash
fault and Salt Wash faults in central Utah is controlled by along-fault ow of CO
2
-charged groundwater
from shallow aquifers (<1 km deep). The same faults are the likely conduits that charge the shallow
aquifers with CO
2
from depth. We document fault zone trace geometry and architecture, and evidence
for palaeo-uid ow within the footwalls of both faults. Evidence for palaeo-uid ow consists of
extensive bleaching of sandstones and some siltstones, mineralisation of carbonates and celestine veins
and minor hydrocarbon staining. The eld evidence shows that the pathways for multiple phases of uid
ow were structurally controlled utilising the fracture network developed in the damage zone of the
faults. To investigate the likely effect of these faults on the regional uid-migration pathways at depth,
a 3D model of the faulted systemwas generated and a fault seal analysis applied to predict the cross-fault
sealing capabilities of the studied faults. Due to the scarcity of subsurface data, the results are not
conclusive but suggest probable multiple cross-fault leak points for uids to migrate across the fault, in
contrast to the eld observations that indicate fault-parallel ow. This comparison of eld observations
to the modelling approach demonstrates the inability of conventional seal analysis techniques to predict
fault-parallel uid leakage and highlight the effects fracture networks in the damage zone, especially at
structural complexities along the fault, have in producing pathways for vertical ow. Multiple uids have
utilised similar fault-parallel pathways over geological time demonstrating that such pathways have the
potential to cause long-term leakage from hydrocarbon reservoirs and CO
2
storage sites.
1. Introduction
Faults and associated fracture networks can play a signicant
role in the subsurface migration of various uids. Focussing of ow
related to fault geometry has been demonstrated in geothermal
elds (Curewitz and Karson, 1997; Rowland and Sibson, 2004),
hydrothermal/epithermal systems (Breit and Meunier, 1990;
Micklethwaite, 2009) and petroleum systems (Chan et al., 2000;
Garden et al., 2001; Gartrell et al., 2004). In most studies, ow is
concentrated in the fracture network (commonly referred to as the
damage zone) that is developed around a main zone of slip.
Complexities along a fault related to terminations and/or linkages
between fault segments are commonly domains of high fracture
density and connectivity and are therefore likely to focus ow
(Curewitz and Karson, 1997; Anderson and Fairley, 2008; Eichhubl
et al., 2009). Models that predict fault properties based on the
throwand host rocks cut by the fault (Yielding et al., 1997; Yielding,
2002; Bretan et al., 2003) generally rely on oversimplied fault
geometries and complexities, leading to the possibility of under-
estimating likely leakage points due to fault throwpartitioning and
simplied fault linkages (Childs et al., 1996, 1997), and do not
account for along-fault ow.
This study investigates a natural leaking CO
2
-rich system at the
northern end of the Paradox Basin in central Utah, United States. In
this locality, the Little Grand Wash fault and northern fault of the
Salt Wash graben provide lateral barriers to present-day cross-fault
ow, but provide pathways through the cap rock via damage zone
fractures to allow CO
2
and additional uid regimes to leak to the
surface (Shipton et al., 2004, 2005). Multiple mineralisation and
diagenetic products associated with past and present migration of
uids along both faults demonstrate structural controls inuencing
multiple uid regimes to vertically migrate through a thick inter-
bedded sandstoneshale stratigraphy. By combining outcrop
* Corresponding author. Present address: Chevron Australia, 250 St Georges
Terrace, Perth, WA, Australia. Tel.: 61 8 9216 4141; fax: 61 8 9216 4103.
E-mail address: ben.dockrill@chevron.com (B. Dockrill).
doi:10.1016/j.jsg.2010.01.007
analysis of the fault zone and associated uid-migration products
with analysis of fault rock integrity through modelling, we have
examined constraints on leakage from shallow reservoirs. Faults
can form barriers to ow either by juxtaposing reservoir rock
against low-permeability clay-rich non-reservoir rock to create
a juxtaposition seal, or when processes of fault rock generation
form a low-porosity and low-permeability fault rock. By analysing
the fault rocks and the processes and factors that contribute to the
failure of this system, we can identify and highlight the roles faults
can play in trapping and transmitting uids. These results
demonstrate the impact fault zones can have on fault-parallel
leakage from a robust structural trap and highlight the potential
risks when assessing seal integrity for structural traps in the
hydrocarbon and emerging CO
2
geologic storage industries.
The eld area is located at the northern end of the Paradox Basin
in the Colorado Plateau region of the United States (Fig. 1), a late
Palaeozoic to Mesozoic intracratonic basin inlled with a thick
sequence of evaporite, carbonate and clastic sediments (Hintze,
1993). The basin is dened by the areal extent of the Pennsylva-
nian Paradox Formation, which contains nearly 2 km of evaporates
(Doelling et al., 1988). The basin has been investigated for hydro-
carbons (Peterson, 1973, 1989; Hansley, 1995; Huntoon et al., 1999)
and mineral resources (Breit and Meunier, 1990; Morrison and
Parry, 1986). Multiple reservoirs have accumulated CO
2
for
extended periods of times (Allis et al., 2001, 2005; Moore et al.,
2005; Shipton et al., 2004, 2005; White et al., 2005). Some of
these are now currently being exploited, predominantly for
enhanced oil recovery (i.e. Bravo and McElmo domes Allis et al.,
2001). Other reservoirs leak CO
2
due to the inuence of faults
(Springerville-St Johns Moore et al., 2005; Hurricane fault
Nelson et al., 2009) and/or boreholes (Woodside Doelling, 1994).
The stratigraphy in the Paradox basin ranges from the Penn-
sylvanian Paradox Formation to the Mid Cretaceous Mancos Shale,
though only the Upper Jurassic to Mid Cretaceous succession crops
out in the eld area (Fig. 1). The Pennsylvanian and Permian
formations consist of marine carbonates and shales that are
potential sources of the CO
2
(Heath et al., in press; Wilkinson et al.,
2008) and hydrocarbons (Peterson, 1973, 1989; Huntoon et al.,
1999; Chan et al., 2000; Garden et al., 2001). The aeolian reser-
voir sandstones of the Permian White RimSandstone are capped by
uvial and lacustrine shales of the Triassic Moenkopi and Chinle
formations. The aeolian Lower Jurassic Wingate and Navajo sand-
stones are important regional aquifers separated by the uvial
Kayenta Formation aquitard. Forming a seal above the Navajo
Sandstone is the Mid-Jurassic Carmel Formation, a complex sabkha
sequence of sandstone, siltstone, mudstone, limestone, anhydrite
and gypsum. The youngest reservoir units in the basin are the Mid-
Jurassic Entrada and Curtis aeolian to marginal marine sandstones,
overlain by marine siltstones and shales of the Middle Jurassic
Summerville Formation. The remaining stratigraphic sequence is
dominated by shales with small, disconnected reservoir units. The
Upper Jurassic Morrison Formation consists of stacked uvial
sandstone channels, interspersed and overlain by lacustrine shales.
The lower Cretaceous Cedar Mountain Formation lacustrine shales
are overlain by conglomeritic uvial channels from the lower
Cretaceous Dakota Formation. The youngest rocks exposed in the
eld area are marine marls of the Middle Cretaceous Mancos Shale.
The shallowly north- to northwest-plunging, open Green River
anticline is one of a series of northwest-trending folds that have
growth histories related to salt movement in the Paradox
Formation since the Permian (Doelling et al., 1988). The Green River
anticline is cut by the Little Grand Wash fault and the Salt
Wash graben. Timing of movement along both faults is poorly
constrained, with the youngest faulted stratigraphy being the Mid
Cretaceous Mancos Shale, though evidence presented by Pevear
M
o
a
b
f
a
u
l
t
S
a
l
t

V
a
l
l
e
y
C
o
u
r
t
h
o
u
s
e
s
y
n
c
l
i
n
e
Green River
Anticline
g
r
a
b
e
n
Green
River
0 10 Kilometres
K
uJ
uJ
uJ
uJ
uJ
uJ
mJ
mJ mJ
mJ
lJ
lJ
lJ
Fault (tick on downthrown side)
Cretaceous
Mancos and Cedar Mountain
K
Upper Jurassic
Morrison, Summerville, Curtis
uJ
Mid Jurrasic
Entrada and Carmel
mJ
Lower Jurassic
Navajo, Kayenta and Wingate
lJ
Tr
P
Q
Q
K
K
Fig. 2a
Fig. 2b
2
3
3
4
5
2
3
5
6 3
4
7
25
b
c
a
Ten Mile
graben
Salt Wash
graben
Little Grand
Wash fault
30 20 10 0
100
200
300
0
distance along strike (km)
t
h
r
o
w

(
m
)
260m
30 20 10 0
100
200
300
0
distance along strike (km)
t
h
r
o
w

(
m
)
400 366m
165m
R
R
210m
154m
B
UTAH
Fold (arrow in plunge direction)
Triassic
Chinle and Moenkopi
Tr
Permian
Cutler
P
Quaternary Q
N
110 15
o
110 00
o
39 00
o
38 45
o
Fig. 1. Geological map of the study area in central Utah, USA. Structural data from eld work and McKnight (1940) and Doelling (2001, 2002). Dashed boxes show the positions of
Fig. 2. The letters a, b, c indicate position of the fault zone strip maps displayed in Fig. 3. Top right inset, throw distribution along the Little Grand Wash fault zone. Centre inset,
throw distribution along the northern (black circles) and southern (grey triangles) Salt Wash graben. Note that relay ramps along the Salt Wash graben are coincident with throw
minima.
B. Dockrill, Z.K. Shipton / Journal of Structural Geology 32 (2010) 17681782 1769
et al. (1997) and Shipton et al. (2004) suggest possible early Tertiary
to Quaternary movement. Where the fold axis of the Green River
anticline is cut by the faults, CO
2
-charged groundwater effuses from
a series of geysers and springs (Doelling, 1994; Shipton et al., 2004,
2005). A series of actively-forming spring deposits and remains of
ancient travertine deposits are found along or immediately to the
North (i.e. in the footwall) of both fault systems (Fig. 2). All of the
springs originate in the footwall though some of the travertine
deposits drape over the fault into the hanging-wall. Additional
CO
2
-charged geysers occur where hydrocarbon or water boreholes
have penetrated the footwall reservoirs, though some boreholes
that penetrate the footwall reservoirs at depth do not leak (Shipton
et al., 2005).
The faulting of the Green River anticline has created a series of
stacked three-way anticlinal closures in the footwall reservoirs of
the Little Grand Wash fault and northern fault of the Salt Wash
graben. Southeast-directed regional groundwater ow (Hood and
Patterson, 1984) suggests that CO
2
-charged meteoric uids have
been focussed up the north-plunging Green River anticline towards
the central sections of the faults. Carrier beds for the CO
2
-charged
water include the Lower Jurassic Wingate and Navajo sandstones
and the Mid-Jurassic Entrada and Curtis sandstones (see Heath
et al., in press). The cap rocks for these structural closures are
provided by clay-rich and evaporitic beds in the sequence, while
lateral barriers to ow are provided by the faults (Shipton et al.,
2004, 2005). Stable isotope data indicate that travertine deposits
along both faults have resulted from a common CO
2
-rich uid
(Shipton et al., 2005; Dockrill et al. in prep). Furthermore, leakage is
conned to the footwall of both faults with no isotopically similar
carbonates located elsewhere in the area. The distribution of
springs and travertines therefore suggests that the studied faults
are providing a long-lived pathway for a signicant proportion of
CO
2
-rich uids to migrate vertically through multiple cap rocks in
the footwalls of both faults.
3. Fault geometry
The eastwest trending south-dipping Little Grand Wash fault
juxtaposes late Jurassic and Cretaceous siliciclastics at the surface
(Fig. 1). The fault has a 30 km long, arcuate surface trace that splits
into two dominant fault strands in the central part of the structure,
extending from 3.2 km east to 0.1 km west of the Green River
(Fig. 2). The two sub-parallel fault strands introduce structural
complexities to the fault with a series of prominent fault bends,
branch points and relay ramps. Between and surrounding the two
main strands, a complicated array of minor faults and relays has
Fig. 2. Relationship between Little Grand Wash fault and Salt Wash graben structures and locations of CO
2
leakage (modern and ancient travertines) and evidence for past uid
ow. (a) Central part of the Little Grand Wash fault where it splits into two main strands and cuts the hinge of the Green River anticline. Dashed boxes showthe two locations for the
fracture analysis in Fig. 4. (b) Detailed structure of the Salt Wash graben. Equal angle stereonets show strike and dip of faults with lineations (triangles) indicating direction of fault
movement. N is the number of faults measured. Closed triangles correspond to northern fault lineations, while open triangles indicate southern fault lineations. (For interpretation
of the references to colour in this gure legend, the reader is referred to the web version of this article).
created multiple structural terraces with varying dips (Fig. 2; see
also Fig. 7 in Vrolijk et al., 2005). Away from the central zone,
a single fault strand is present. In the subsurface, the down-dip
extent and geometry of the fault is uncertain, though drilling
records for two exploration wells (Amerada Hess 1 and 2) that
intersect the fault at depth suggest that the fault may offset rocks as
old as Pennsylvanian in age.
The dip of the main fault strands along the Little Grand Wash
fault varies but is generally steep, averaging 70
to the south. Slip

indicators show that the fault is mostly dip-slip with some local
oblique right- and left-lateral movement in the structural terraces
(stereonets in Fig. 2). The maximum stratigraphic offset across the
fault is 260 m immediately east of the Green River with throw
systematically decreasing towards the exposed surface fault tips
(Fig. 1).
The Salt Wash graben forms part of a 31 km-long structure
consisting of two west-northwest-trending grabens (Salt Wash and
Ten Mile grabens) linked by a left-stepping relay ramp (Fig. 1). Both
grabens are bound by sub-parallel, inward-dipping normal faults
that overlap for a kilometre and result in a steep, east-dipping ramp
connecting the northern and southern footwalls between the two
down-dropped structures (Williams, 2005). At the surface the
graben juxtaposes Mid-Jurassic and Cretaceous siliciclastics. The
faults subsurface geometry and vertical penetration is unknown
due to the lack of subsurface data. However, reservoir-scale faults to
the SE (Moab fault and Salt Valley graben; Fig. 1) commonly
penetrate the Pennsylvanian Paradox Formation in the immediate
region (Doelling et al., 1988; Foxford et al., 1998). Further discussion
will be limited to the Salt Wash graben as the Ten Mile graben is off
structural trend in relation to the Green River anticline and asso-
ciated uid-owfeatures, and eld mapping shows no evidence for
present or past uid ow along the Ten Mile graben (Williams,
2005).
The Salt Wash graben is 20 km long, with an average width of
800 m. Bed dips within the graben range fromsub-horizontal in the
centre of the graben to up to 25
towards the bounding faults. The

northern fault is a single strand that dips steeply to the south,
averaging 73
(Fig. 2). Sporadic geometric complexities are

encountered along the fault trace in the form of relay ramps and
branch points. The southern fault dips steeply to the north,
averaging 75
and displays rare relay ramps and branch points

along strike. To the east, the southern fault bends eastward, trun-
cating the northern fault and terminating against the southern fault
of the Ten Mile graben. Rare slickensides indicate predominantly
dip-slip movement. The maximum stratigraphic offset across the
northern fault is 366 m and across the southern fault is 210 m
(Fig. 1). Variation in throw along strike of both faults denes
a throw prole consistent with a kinematically linked fault system
(e.g. Dawers and Anders, 1995).
4. Fault architecture
At the surface, both faults displace an interbedded sequence of
sandstones and shales. The surface exposures of both fault zones
contain a fault core, where most of the slip is accommodated,
surrounded by a damage zone of smaller-offset structures (c.f. Caine
et al., 1996). The fault cores are typically 13 m wide and contain
three principal deformation elements: (1) clay-rich gouge; (2)
entrained sections of host rock; and (3) slip zones (Fig. 3; Fig. 9 in
Vrolijk et al., 2005). The gouge forms centimetre- to metre-thick
zones with a highly foliated, fault-parallel fabric that locally
forms an SC fabric adjacent to slip zones. X-ray diffraction analyses
show that the gouge is composed of quartz and clay minerals
including illite, illitesmectite and kaolinite, with sporadic halite,
calcite, feldspar and haematite. Entrained sections of sand-rich host
rock include fractured, elongated lenses of sandstone and siltstone,
220 cm thick, and coherent but highly-fractured slabs up to 1 m
thick with occasional deformation bands and rare calcite veining.
These lenses and slabs are most evident where Mid to Late Jurassic
sandstones are located in the footwall. The isolated lenses are
enveloped by gouge and have long axes orientated sub-parallel to
the gouge foliation. The striated and intensely foliated clay-rich slip
zones are 5100 mm thick with highly polished striated surfaces,
and occasionally contain angular, sand-rich, pebble-sized clasts.
Slip zones dene the limit of the fault core, separating fault rock
from fractured host rock. Slip zones are also common inside the
fault core, where they separate fault gouge and sandstone pods
derived from different lithologies. These observations suggest that
the slip zones accommodate the majority of the slip within the
fault core.
Variations in fracture orientation and density in the damage
zone were characterised along transects in two relatively well-
Fig. 3. Structural transects across the fault core at (a) the Little Grand Wash fault western outcrop, (b) the northern fault of the Salt Wash graben and (c) the Little Grand Wash fault
eastern outcrop. The location of transects are shown in Fig. 1. Transects are at a 1:1 scale with respective scale bars in top right corners. Exposures were cleaned with trowels and
brushes to remove the top weathered rind of the clay-rich rocks. Mineralogy of fault core gouge was assessed using X-ray diffraction. I in I/S refers to illite in illite/smectite mixed
layers.
exposed areas along the Little Grand Wash fault (Fig. 4). The
Little Grand Wash fault damage zone is dominated by a broadly
eastwest trending set of fractures that locally reect the strike
of the fault (Fig. 4). Immediately adjacent to the fault core
a network of shear fractures and joints broadly diminishes in
frequency outward into the host rock. Where observed, there is
little to no displacement associated with the fractures and they
are occasionally inlled with carbonates or sulphates. Variations
from this trend are observed in a relay ramp between two fault
strands (scanline 4) and in the hanging-wall shales (scanline 7).
The relay ramp contains a diverse array of fracture orientations
that are likely due to complicated three-dimensional strain
during the development of the ramp. Distinct northwest-
trending fractures at the top of the ramp (scanline 3; Fig. 4a)
are likely the result of rotation and steepening of strata into the
fault zone. The varied fracture orientations in the hanging-wall
shales (scanline 7; Fig. 4b) may represent pre-existing fractures
overprinted by later fault-associated fractures, or differing strain
accommodation in the clay-rich strata compared to the sand-rich
footwall strata.
Fig. 4. Fracture analyses along the Little Grand Wash fault. (a) Five scanlines within and away from a partially breached footwall relay ramp along the northern of two main fault
strands with a combined throw of 245 m. (b) Two scanlines near two closely-spaced fault strands with a combined throw of 255 m. To limit bias due to transect orientation (Park
and West, 2002) where possible the scanlines were orientated perpendicular to the master fault. Fracture orientation (rose diagrams, R % of fractures in largest petal) and
frequency (histograms, 2 m bins, F mean number of fractures per 2 m bin) are shown for each scanline.
5. Evidence for uid ow and leakage
At present, CO
2
-charged groundwater is emanating fromsprings
and geysers along both faults and an active hydrocarbon seep is
located in the immediate footwall of the Little Grand Wash fault.
Additionally, the soil gas surveys of Allis et al. (2005) show that
focussed areas of high CO
2
ux occurs along the faults, not all of
which are associated with effusion of CO
2
-charged groundwater. At
the surface, the CO
2
-charged groundwater precipitates travertine
mounds. The positions of past CO
2
leak points are marked by partial
to complete remnants of ancient travertine mounds (Shipton et al.,
2005; Dockrill, 2006). Active and ancient travertine deposits are
concentrated in the footwall of both faults proximal to where the
Green River anticline axis is cut by the faults (Fig. 2). Stable isotope
studies indicate that the ancient travertines were precipitated by
the same CO
2
-rich uids that are currently leaking (Dockrill, 2006;
Dockrill et al. in prep.). The ancient travertine deposits form resis-
tant caps on top of numerous buttes (Fig. 5d) and are topographi-
cally higher than the actively-forming travertines (5f).
There are variations in the distribution and development of the
travertine deposits between the two faults. The Little Grand Wash
fault contains a series of discrete, well-developed, thick (110 m)
deposits that are located in various footwall lithologies from the
sand-rich Curtis Formation to clay-rich sections of the Morrison
Formation (Fig. 2a). The deposits form in the immediate footwall
but can drape over the fault into the hanging-wall. Additionally, the
deposits are conned to sections where the two main fault strands
are close together and/or there are pronounced structural
complexities such as fault bends and relay ramps. The northern
fault of the Salt Wash graben contains mainly thinner (0.54 m) and
less well-developed deposits compared to the Little Grand Wash
fault. They are located in the sand-rich Entrada and Curtis forma-
tions with deposits predominantly conned to the immediate
footwall, occasionally draping into the adjacent graben. However,
around the Green River fold axis deposits increase in number and
extend further into the footwall (Fig. 2b). The easternmost traver-
tine deposits along this fault are associated with a breached relay
ramp.
Bleaching of sandstones occurs to varying degrees in the foot-
walls of both faults, and is spatially associated with the travertine
deposits. The bleaching is characterised as the loss of haematite
rims mainly around sand grains (Fig. 5 a and b) resulting in a colour
change in the footwall sandstones from ubiquitous red of the
unbleached rock to yellow/white. The minor proportions of silt-
stones affected by bleaching also showa transition to paler colours.
The distribution of bleaching in the footwalls of both faults is
focussed in the faulted fold axis of the Green River anticline, though
its spatial extent varies depending on the footwall lithologies.
Along the more shale-rich Little Grand Wash fault footwall,
bleaching is mainly conned to the base of ancient travertine
deposits (Fig. 5c) or limited sections of the sand-rich Curtis
Formation (Fig. 5d). At the base of the ancient travertines, bleaching
is characterised by thin haloes (120 cm wide) around sub-vertical
fractures, commonly inlled with aragonite or calcite veins and
sub-horizontal tabular bleached bodies (5200 cm thick) that
extend laterally as dendritic stringers up to 200 m from the fault.
The thickness of the bleached haloes and tabular bodies is depen-
dant on the lithology; with clay-rich, lower porosity units (i.e.
Morrison Formation) having thinner bleached zones relative to the
sand-rich, higher porosity units (i.e. Curtis Formation). Bleaching
away from the travertines is constrained to lower sections of the
Curtis Formation, with the bottom 2 m of the outcropping sand-
stone partially to completely bleached, while the next 4m contains
sporadic, thin bleached horizons (110 cm thick) sub-parallel to
bedding and thin bleached haloes (15 cm thick) surrounding
steeply-dipping fractures that are sporadically inlled with calcite
or rare gypsumveins. Above this, bleaching is rare within the Curtis
Formation and the overlying more clay-rich lithologies away from
the travertines. Celestine and calcite veins are found associated
with the bleaching (Fig. 5j).
Along the more sand-rich rocks exposed in the footwall of the
northern fault of the Salt Wash graben, the red Entrada sandstones
are extensively bleached to pale yellow (Fig. 5e). In the lower
section of the exposed Entrada Formation (up to 8 m thick), sand-
stones are completely bleached, and the contact between bleached
and unbleached sandstone does not followbedding contacts (Fig. 5f
and g). Above this contact, bleaching is mainly located around
clusters of steeply-dipping fractures with bleached haloes up to 5 m
wide (Fig. 5h). Pervasive reduction of the lower parts of the Entrada
and Curtis sandstones extends up to 300 minto the footwall (Fig. 2).
Beyond this region, bleached sandstones are only observed at the
base of ancient travertine deposits and around steeply-dipping
fractures. Similar to the Little Grand Wash fault, bleaching at the
base of travertine deposits is characterised mainly by reduction
haloes (up to 50 cmwide) around fractures, sometimes inlled with
calcite, aragonite or gypsum (Fig. 5i) and limited sub-horizontal
tabular reduction bodies (up to 2 m thick). Additional bleaching is
also observed around steeply-dipping fracture clusters up to 400 m
into the footwall. Close to the fault, a signicant proportion of the
fractures within the Entrada and Curtis sandstones are accompa-
nied by reduction haloes, 5200 cmwide. Further into the footwall,
sporadic clusters of fractures, generally oblique to the fault, contain
thinner reduction haloes (1100 cm wide) that continue for as far
as the Curtis sandstone crops out at the surface. Away from the
anticline axis, the only signicant patch of bleaching occurs by the
breached relay where the easternmost ancient travertine is found
(Fig. 2).
Hydrocarbon staining is limited to sandstones of the Curtis
Formation (Fig. 6a) and the Salt Wash Member of the Morrison
Formation close to or between the main fault strands of the Little
Grand Wash fault. Hydrocarbon-stained sandstones are identied
by a greyish colour, a musty petroleum odour when freshly broken
and the occasional presence of bitumen veins. Hydrocarbon stain-
ing in the Curtis Formation is rare and limited to the same strati-
graphic section that contains the bleaching, with 520 cm thick
horizons of grey sandstone bordered by 15 cm thick, yellow to
white bleached haloes (Fig. 6b). Conversely, hydrocarbon staining
in the Salt Wash Member is more abundant with a 10 m thick
sequence of stained beds between the two main fault strands. The
stained sandstones are dark grey to white, are inundated with
Fig. 5. Iron-oxide reduction in the footwall of the Little Grand Wash fault and northern fault of the Salt Wash graben. Photomicrographs of Curtis Sandstone with (a) quartz (Qtz)
grains rimmed by haematite and (b) haematite rims removed from quartz grains. (c) Reduction haloes surrounding sub-vertical fractures in the Curtis Formation (pen is 15 cm long).
(d) Tabular reduction bodies (black arrows) extending into the Curtis Formation footwall. Yellow outline shows extent of overlying travertine deposit; dashed line is the fault trace.
(e) Sporadic reduction horizons in the Curtis formation in the footwall of the LGWF. (f) Viewof the reduced, yellow Entrada sandstones in the core of the Green River anticline facing
east. Away from the anticline, sandstones revert back to a red colour. Trace of the fault (nSWG) marked with a dashed line. (g) Entrada sandstone butte with lower half completely
reduced and upper half only reduced around a fracture cluster (red arrow). Note reduction is not bed conned with a varying upper contact (cf. black arrows). Height of butte
approximately 18 m. (h) Undulating reduction front in Entrada sandstone demonstrating that the colour change is not depositionally controlled. (i) Reduction haloes surrounding
fractures in Entrada sandstone at the base of a travertine deposit (scale card is 16 cm long). (j) Celestine vein with reduction halo (lens cap 8 cm in diameter). (For interpretation of
the references to colour in this gure legend, the reader is referred to the web version of this article.)
bitumen veins and contain many sub-vertical fractures with
a hydrocarbon coating (Fig. 6c). At the base of this sequence,
a hydrocarbon seep has been active since at least the beginning of
the 20th century (Lupton, 1914; Fig. 6d).
6. Fault modelling
To place the fault-parallel uid-migration observed in the eld
in context with potential for across-fault ow at depth, we need to
predict the likely fault rocks and their uid-ow properties at
depth. A simple three-dimensional model of the northern Paradox
basin area was generated using Midland Valley Explorations
2Dmove and 3Dmove software tools. 2D and 3DMove are structural
modelling software that uses geo-constrained data to build vali-
dated, balanced cross sections and surfaces, respectively. The model
was constructed from approximately 200 hydrocarbon and water
wells and 22 wire-line logs (Fig. 7a), and published (McKnight,
1940; Williams, 1964; Condon, 1997; Doelling, 2001, 2002) and
new eld data. These data were combined in 2DMove to generate
a series of isopach and structure maps (Fig. 7b) that were used as
templates to create 115 cross sections (e.g. Fig. 7c) covering the
study area. The sections were subsequently connected by linear
interpolation in 3DMove to create surfaces representing either
faults or stratigraphic tops (Fig. 7d). The resulting 3D model
consisted of 10 stratigraphic horizons representing the tops of units
from the White Rim Sandstone to the Cedar Mountain Formation,
based on major lithological changes and local formation divisions.
The horizons are juxtaposed by 5 fault surfaces representing the
Little Grand Wash fault and the bounding faults of the Salt Wash
and Ten Mile grabens. Due to the lack of subsurface data pene-
trating or adjacent to the faults, they were modelled as single
planar surfaces with constant dip. The three-dimensional model
was validated in 2Dmove through section balancing during the 2D
section and 3D surface construction (i.e. Midland Valley, 2004) and
constrained by surface geologic data. The detailed methodology of
the model construction can be found in Dockrill (2006).
To assess the likely sealing properties of the fault rocks at depth
we undertook a simplied fault seal analysis in Traptester software,
provided by Badley Geoscience. Regions of cross-fault reservoir
reservoir contact (potential cross-fault leak points) were identied
from juxtaposition of footwall and hanging-wall cutoffs of key
subsurface reservoirs against each fault using Allan diagrams
(Fig. 8). To indicate the potential for membrane seal development
by argillaceous smear or entrainment, Shale Gouge Ratio (SGR)
calculations were applied to reservoirreservoir contacts. SGR is
the predicted percentage % shale within the fault (gouge assuming
uniform mixing), calculated from the ratio of total shale bed
thickness and fault throw (Yielding et al., 1997). SGR was the
preferred smear algorithm used as it incorporates the effects of
argillaceous material distributed through all siliciclastic units as
opposed to individual argillaceous beds (cf. Clay Smear Potential
and Shale Smear Factor; Yielding et al., 1997; Yielding, 2002) and it
is the most commonly published algorithm with the best dened
threshold sealing values (Yielding et al., 1997; Foxford et al., 1998;
Sperrevik et al., 2000; Bretan et al., 2003; Gibson and Bentham,
2003). To establish the SGR along the fault, the V
shale
content
(volumetric shale content) of the modelled stratigraphy was esti-
mated from representative wire-line gamma logs spliced together
from wells proximal to the fault. The base shale and sand
measurements were normalised to the 100% and 0% shale value
respectively and a V
shale
cutoff of 0.5 was used to dene shale beds
(Fig. 7e). Without calibrating wire-line logs against core data there
is a risk of misinterpretation of argillaceous content; gamma logs, in
particular, are poor detectors of kaolin and mica (Bretan et al.,
2003). X-ray diffraction analyses of fault gouge from both faults
(Fig. 3) indicate that illite and illitesmectite assemblages are the
dominant clay minerals with minor amounts of kaolinite and no
mica. The error related to using only gamma logs to determine
V
shale
is therefore probably small for this study. However, without
Fig. 6. Hydrocarbon staining along the Little Grand Wash fault. (a) Spherical oil globules interspersed around grains and within the matrix of the Curtis sandstone. Scale bar is
1 mm. (b) Hydrocarbon-stained Curtis sandstone with a thin white reduction halo (scale card is 16 cm long). (c) Fractures with a distinct hydrocarbon stain (lens cap is 8 cm in
diameter). (d) Fresh oil seep at the base of the hydrocarbon-saturated Salt Wash sandstones. The width of the shallow pit is approximately 20 cm.
direct calibration with core data uncertainty remains, and
predicted fault rock values that are close to the threshold value for
a continuous clay-rich fault gouge should treated with caution
while higher values are considered more robust.
Potential carrier beds for the multiple uid regimes in the
subsurface include the Navajo, Wingate and White Rim sandstones.
An Allan diagram of the Little Grand Wash fault shows that apart
from self-juxtaposition towards fault tips, the Navajo Sandstone is
juxtaposed against the Entrada Sandstone and the Wingate Sand-
stone is juxtaposed against sections of the Entrada and Navajo
sandstones (Fig. 8a). Conversely, the White RimSandstone contains
a good juxtaposition seal against the shale-rich Chinle and Moen-
kopi formations. The northern fault of the Salt Wash graben has the
Navajo Sandstone juxtaposed against the Entrada Sandstone, the
Wingate Sandstone juxtaposed against the Entrada and Navajo
sandstones and the upper section of the White Rim Sandstone
Fig. 7. Stages in the construction of the three-dimensional model used for fault seal analysis. The red line indicates position of the representative cross section display in Fig. 7.
(a) Distribution of exploration well (circles) and wire-line log (crosses) data used to construct the model. (b) Structure contour map of the Navajo Sandstone created from
a combination of well, eld and published (McKnight, 1940; Doelling, 2001) data. (c) Representative cross section across the Little Grand Wash fault constructed in 2DMove from
structure contour and isopach maps. (d) Generation of a three-dimensional surface in 3DMove from a series of cross sections. (e) Representative V
shale
curve used in the model to
dene the shale content of the stratigraphic sequence. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
juxtaposed against the Wingate Sandstone (Fig. 8b). These juxta-
positions provide potential pathways for cross-fault ow if no
membrane seal is developed in the fault rocks.
Calculated SGR values at reservoirreservoir juxtapositions
identied along both faults indicate values as low as 25% for the
Navajo Sandstone, 19% for the Wingate Sandstone and 75% for
the White Rim Sandstone. Outcrop (Foxford et al., 1998) and
experimental (Sperrevik et al., 2000) studies suggest continuous
argillaceous smears can develop from SGR values greater than ca.
20% and published studies of proven hydrocarbon fault traps
indicate SGR between 15 and 25% generally represent the threshold
for a leakseal transition (Yielding et al., 1997; Yielding, 2002;
Gibson and Bentham, 2003). However, without calibration of the
SGR to analogues or nearby proven fault traps, the SGR threshold
value for the fault to seal remains unknown and the development of
a continuous argillaceous membrane seal that can maintain a uid
column is questionable. For this study, if a SGR value of 20% is used
as a threshold between a fault sealing and leaking, there is potential
for both faults to have cross-fault leak points in the footwall where
the Wingate Sandstone is juxtaposed against the Navajo Sandstone
and towards the fault tips where reservoirs are self-juxtaposed.
Thus, depending on where uids are sourced and their migration
pathways, both faults possibly provide opportunities for cross-fault
ow where development of a continuous argillaceous smear is
questionable due to low SGR values.
Sensitivity analyses of the fault seal model were also undertaken
using juxtaposition/triangle diagrams constructed in Traptester to
1) assess uncertainties related to the modelling of the faults from
a sparse data set, and 2) assess the impact of using alternative
smearing algorithms to calculate membrane seal potential. Fault
modelling uncertainties relate to constraints imposed by a lack of
subsurface data resulting in faults being modelled with constant
vertical throw and as single fault planes with no geometrical
complexity. The impact of changing the vertical throwprole along
both faults can be examined by determining the maximumvertical
change in throw of the modelled stratigraphic sequence calculated
by using an aspect ratio (fault strike length/fault dip length) of 2.15
(Nicol et al., 1996). Fig. 9a shows the maximum change in throw
will be 15 m for the Little Grand Wash fault and 20 m for the
northern fault of the Salt Wash graben and that these changes will
have minimal impact in juxtaposition relationships or SGR values.
Of greater concern is not being able to replicate the outcrop
expressions of the studied faults, with structural complexities
observed in the eld such as multiple fault strands and relay ramps
having the potential to dramatically change throw proles and
reservoir juxtaposition relationships. The effect of multiple fault
Fig. 8. Allan diagrams (Allan, 1989) of (a) the Little Grand Wash fault and (b) the northern fault of the Salt Wash graben. Coloured zones indicate SGR values at reservoirreservoir
juxtapositions. Solid and dashed lines indicate footwall and hanging-wall intersections along both faults, respectively. Key reservoirreservoir juxtapositions and their minimum
SGR values in the central sections of both faults are annotated for reference. Vertical exaggeration 6. (For interpretation of the references to colour in this gure legend, the reader
is referred to the web version of this article.)
strands is evident at the centre of the Little Grand Wash fault,
where the total throw (260 m) is partitioned between the northern
(150 m) and southern (110 m) fault strands (Fig. 9a). Compared to
the single fault modelled in the seal analysis, throw partitioning
between the two fault strands creates more areas of reservoir
reservoir juxtaposition and a general decrease in SGR values,
making the faulted system more exposed to cross-fault ow. Relay
ramps can also compromise the sealing capability of a fault through
also partitioning throw between hard-linked fault strands as well
as providing direct leakage pathways out of structural closures
between soft-linked faults (Hesthammer and Fossen, 2000;
Rotevatn et al., 2007). The along strike outcrop expression of the
Little Grand Wash fault shows multiple examples of hard- and soft-
fault linkage and throw partitioning between multiple fault
strands. The northern fault of the Salt Wash graben has a simpler
outcrop expression but both fault systems would be expected to
have more complex fault geometries in the subsurface than used in
the simplied model, enhancing the likelihood of further reservoir
reservoir juxtapositions and resultant cross-fault leak points in
addition to areas identied in the simplied fault seal model.
Alternative fault seal algorithms that attempt to model the
development of argillaceous smears within the faulted sequence
include Clay Smear Potential (CSP; Bouvier et al., 1989; Fulljames
et al., 1997) and Shale Smear Factor (SSF; Lindsay et al., 1993). The
CSP algorithm is derived from studies on non-lithied clays with
ductile smears while the SSF algorithm is based on lithied shales
with abrasive-type smears. Both algorithms are reliant on dening
smears resulting only from single argillaceous beds. For instances
where more than one argillaceous bed is contributing to the smear,
CSP and SSF represent only the bed with the greatest smear
contribution. Local burial history curves from Nuccio and Condon
(1996) indicate that faulting in the region would have occurred
only in a lithied environment (i.e. abrasive-type smears), limiting
the SSF algorithm as the only logical comparison to the SGR.
Previous studies using the SSF (Lindsay et al., 1993; Takahashi,
2003) indicate that smears become discontinuous and cross-fault
leakage can occur at values greater than 510, which would
represent the blue, yellow and green areas in Fig. 9b. Compared to
modelling results using the SGR (Fig. 9a), which predicts contin-
uous smears forming from 20%, the SSF indicates most of the
reservoirreservoir juxtapositions in the Navajo and Wingate
sandstones for both faults would most likely have cross-fault
leakage due to non-development of continuous shale smears.
Variations between the SGR and SSF in predicting continuous
argillaceous smear development on the studied faults can probably
be attributed to the SSF not incorporating argillaceous material
from non-dened shale beds (cf. SGR) and not modelling
compound smears (i.e. smears generated frommultiple shale beds).
Whether this suggests one smearing algorithm is better in
predicting fault seal potential than the other is speculative as there
is no denitive subsurface data available to conrm the cross-fault
sealing capabilities of both studied faults. However, it does
illustrate the variability in results depending on the smearing
algorithmused and the need for good calibrations to known sealing
and non-sealing faults in a region to dene the best algorithm to
use and its threshold values for sealing.
Based on these sensitivity analyses, the fault seal model gener-
ated for this study should be considered a best case scenario for seal
potential, with a high probability of additional cross-fault leak
points along both modelled faults. However, eld evidence of uid
migration in the footwalls of both faults suggests a signicant
proportion of uids are migrating parallel to the faults, a situation
that cannot be modelled in current fault seal analysis workows or
is generally considered due to the probable multiple cross-fault
leak points along both studied faults.
7. Discussion
7.1. Origin, timing and phase of uid regimes
The CO
2
-charged springs are a consequence of deep CO
2
mixing
with one or more regional aquifers and eventually leaking to the
surface in the footwalls of the studied faults. Previous studies on
the origin of the CO
2
have varied with Shipton et al. (2004, 2005)
Fig. 9. Juxtaposition diagrams (Knipe, 1997) exploring uncertainties related to the fault seal model. Black areas on the juxtaposition diagrams represent reservoir or seal juxtaposed
against seal and coloured areas indicate (a) shale gouge ratio and (b) shale smear factor values at reservoirreservoir juxtapositions. Seals are dened as stratigraphy with V
shale
(VSH) values greater than 0.5. Vertical blue lines represent the modelled maximum throw on the Little Grand Wash fault (LGWF) and northern fault of the Salt Wash graben (SWG).
Vertical purple lines represent maximum surface throw on the Little Grand Wash fault partitioned between two fault strands. White shaded areas indicate maximum changes in
down-dip throw along both faults. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
and Heath et al. (in press) suggesting a crustal source generated by
claycarbonate reactions in Palaeozoic source rocks such as the
Paradox Formation. Wilkinson et al. (2008) argue for an additional
mantle source from nearby igneous intrusions mixed with
a younger crustal source mainly derived from diagenetic reactions
in the shallower Navajo Sandstone. Regardless of its origin, all
authors indicate that a proportion of the CO
2
has mixed with brines
from as deep as the Paradox Formation, and that the resulting
CO
2
-rich brine has migrated vertically into the Navajo Sandstone
before leaking to the surface in the study area. Dating of the
travertines indicates this surface leakage and respective subsurface
migration of CO
2
has been occurring for at least the last 100,000
years (Dockrill, 2006; Burnside et al., 2007, 2009). If a mantle source
is invoked, CO
2
migration could date back to the Tertiary when the
intrusives where emplaced in the region (Ross, 1998). Either way,
CO
2
migration through this faulted stratigraphy has been a long-
lived and pervasive event.
How the CO
2
interacts with the multiple aquifers and in what
state it migrates through the faulted stratigraphy is also important
as this indicates the likelihood that CO
2
could form signicant
subsurface accumulations or migrate straight through the faulted
stratigraphy. If the CO
2
is undersaturated in the aquifers, it will
migrate in solution. In this case migration is governed by one-phase
ow and is controlled solely by rock permeability, limiting the
effectiveness of seals to form absolute barriers and maintain
signicant subsurface uid accumulations. Alternatively, if the CO
2
is oversaturated, it will migrate as a separate free phase; as super-
critical CO
2
at depths generally greater than 800 m and as a gas at
shallower depths (Herzog and Golomb, 2004). In two-phase ow,
top or fault seals will be absolute until the capillary entry pressure
of the seal rock is exceeded, and subsequent migration is controlled
by rock permeability (e.g. Yielding et al., 1997; Bretan et al., 2003).
Shipton et al. (2005) and Heath et al. (inpress) assumed the CO
2
was
sufciently oversaturated to migrate as a free phase through the
aquifers, forming multiple accumulations in the structural highs of
the regional faulted aquifers analogous to migrating hydrocarbons.
However, Wilkinson et al. (2008) used mass balance calculations to
indicate that CO
2
is generally undersaturated in respect to the
aquifers and that CO
2
migration is primarily in solution with a free
phase only forming in fractures above the Navajo Sandstone. This is
supported by Assayag et al. (2009) who used carbon isotopes to
indicate CO
2
is undersaturated in the Navajo Sandstone aquifer.
Based on the two later studies that considered aquifer geochem-
istry, CO
2
is likely to have migrated to the surface through multiple
aquifers primarily in solution, with limited free-phase CO
2
accumulations forming in the regional aquifers. Regardless of
whether CO
2
was a free phase or in solution, its migration pathway
from source to surface was directed by zones of enhanced rock
permeability associated with the studied faults.
The origin of the uid(s) responsible for the bleaching of
predominantly Mid-Jurassic sandstones in the footwalls of both
faults is not clear, as more than one extra-formational uid is
capable of generating the bleached zones, including hydrocarbons,
organic acids, methane and hydrogen sulphide (Surdamet al., 1993;
Chan et al., 2000; Garden et al., 2001; Eichhubl et al., 2009).
Previous studies of bleaching in the Paradox Basin have mainly
attributed the bleaching uid directly to hydrocarbons (Hansley,
1995; Chan et al., 2000; Garden et al., 2001; Beitler et al., 2003;
Parry et al., 2004) or uids directly associated with them
(Eichhubl et al., 2009). One exception is Haszeldine et al. (2005)
who suggest that water and CO
2
H
2
S is responsible for the
bleaching of the sandstone and that this uid migrated along
similar pathways to oil migration. The presence of a hydrocarbon
seep and oil-stained sandstones in the Curtis and Morrison
formations adjacent to the Little Grand Wash fault indicates that
hydrocarbons may have caused the bleaching. However, the
absence of hydrocarbons or associated features in the more
extensively reduced northern footwall of the Salt Wash graben
presents some doubt as to whether hydrocarbons were the
bleaching uid or if multiple uids were involved. If hydrocarbons
were responsible for the bleaching, the most likely hydrocarbon
source would be organic-rich black shales from the Pennsylvanian
Paradox Formation (Nuccio and Condon, 1996), though the Penn-
sylvanian Honaker Trail Formation, Permian Kaibab Limestone and
Triassic Sinbad Limestone are also potential sources (Peterson,
1989; Huffman et al., 1996). The timing of hydrocarbon migration
in the region has been conned to the early Tertiary along the Moab
fault to the East (Garden et al., 2001) and to the Mid Tertiary in the
Tar Sand Triangle to the South (Hansley, 1995; Huntoon et al., 1999).
The presence of an active seep on the Little Grand Wash fault
suggests that hydrocarbon migration continues to the present day,
or that one or more subsequent pulses of hydrocarbons have
charged the system from different sources. Migration of hydrocar-
bons from the Palaeozoic source through to the currently exposed
Jurassic reservoirs with bleaching and beyond will be governed by
two-phase ow.
Precipitation of the celestine veins (SrSO
4
, Fig. 5j) require a Sr
2
-
rich uid to mix with a SO
4
2
-rich uid or replace calcium sulphate
minerals (i.e. gypsum or anhydrite) in-situ (Scholle et al., 1990).
Precipitation of celestine probably results from Pennsylvanian-
derived, Sr
2
-rich uids rising through the stratigraphy and mix-
ing with shallow SO
4
2
-rich meteoric waters or directly replacing
gypsum precipitated in fractures. Analysis of meteoric waters
erupting from geysers and springs along both faults by Heath et al.
(in press) indicate the present-day waters are undersaturated in
respect to sulphate. Thus, sulphate precipitation must predate the
current uid regime. The presence of reduction haloes around
fractures inlled with sulphate suggests sulphate precipitation
occurred during or after iron-oxide reduction as fracture pathways
needed to be open to allow the migration of iron-reducing uids. If
hydrocarbons were responsible for iron-oxide reduction, sulphate
precipitation commenced during or after the early Tertiary and
ceased before start of the present-day meteoric conditions.
Considering the uid-ow products that have been identied in
the study area, it is apparent that varied uid regimes (i.e. CO
2
-rich
brines, hydrocarbons, Sr-rich uids) in differing uid states (i.e.
one- vs. two-phase ow) have utilised the same pathways to
vertically migrate through this faulted stratigraphy over an
extended period of time. This common migration pathway for
multiple uid regimes has been inuenced by zones of enhanced
permeability associated with the studied faults, which enable
vertical migration through multiple cap rocks (i.e. Chinle, Moen-
kopi, Carmel and Morrison formations) and eventual leakage of
uids out of the system.
7.2. Structural controls on uid migration
The trap geometry of the study area dictates that the studied
faults play a primary role in the migration of uids from source to
reservoir and eventual leakage from the system. Multiple uid
regimes have utilised a common pathway parallel to the faults to
migrate through the system. However, fault seal modelling of the
Little Grand Wash fault suggests there should be multiple leak
points along the juxtaposed reservoirs that permit across-fault
ow, primarily due to throw partitioning associated with fault
linkage and multiple fault strands. To comprehend why there is an
inconsistency between the eld and model-based approaches in
identifying migration pathways, we need to appreciate what is
encouraging fault-parallel owin this faulted system. The two most
likely options promoting vertical rather than across-fault ow are
1) incompetent cap rocks relative to the faults sealing capabilities,
or 2) enhanced fault-parallel permeability compared to the
surrounding country rock enabling migration through overlying
cap rocks.
Given that all the uid regimes were likely sourced fromat least
as deep as the Paradox Formation, there are multiple formations
that could act as cap rocks to inhibit vertical migration of uids. The
two most likely cap rock sequences are the Chinle and Moenkopi
formations above the White Rim Sandstone and the Carmel
Formation above the Navajo Sandstone. Combined, the Chinle and
Moenkopi formations form a thick (>200 m), shale-rich sequence
of uvial and lacustrine deposits with minor marine incursions
(Trimble and Doelling, 1978). Mercury injection measurements on
the Moenkopi Formation by White et al. (2005) indicate that the
unit could maintain a CO
2
column height of 100 m, suggesting the
combined formations could provide a competent overlying cap rock
sequence for any uids migrating through the White Rim Sand-
stone. The Carmel Formation is a heterogeneous sabhka sequence
of interbedded siliciclastics, carbonates and evaporates with the
upper half containing over 50% anhydrite and gypsum(Trimble and
Doelling, 1978). This unit is a proven seal for the nearby Farnham
Dome eld forming a competent cap rock above the Najavo Sand-
stone, which contains a 100 m CO
2
gas column. (Allis et al., 2001)
Thus, cap rocks in the study area appear to be competent seals, with
fault-parallel ow of multiple uids more likely reliant on
enhanced permeability associated with the faults rather than
inadequate cap rocks.
To understand how enhanced permeability associated with the
faults is promoting fault-parallel uid ow, the specic migration
pathways through the faulted stratigraphy needs to be identied.
The migration pathways of these uids can be determined through
identifying the spatial distribution of their ow products in varied
lithologies observed in the footwalls of both faults. In sand-rich
formations (Entrada and Curtis) prevalent in the northern foot-
wall of the Salt Wash graben, pervasive bleaching and numerous
travertine deposits extend into the footwall around the fold axis of
the Green River anticline. This distribution is consistent with uids
migrating up-dip and pooling at the crest of the faulted anticline.
Distal to this area or in clay-rich stratigraphy, bleaching and
travertine mineralisation is limited to the immediate footwall
suggesting fracture networks developed in the damage zone of the
fault play a role in the migration of uids. This is readily evident in
the clay-rich formations (Summerville and Morrison) of the Little
Grand Wash fault footwall, where travertine mineralisation and
bleaching is restricted to the immediate footwall, preferentially at
structural complexities (i.e. relay zones and fault bends). Therefore,
upon the migration of uids to the crest of the faulted anticline,
uids vertically migrate through low-porosity, clay-rich units by
utilising the fracture network developed in the damage zone of the
fault. Fluid migration within the damage zone is concentrated in
areas of structural complexity along the faults that include relay
ramps, fault intersections and/or fault bends.
The localisation of uid migration to structurally complex
sections of the fault is a well documented phenomenon (Kerrich,
1986; Caine et al., 1996; Sibson, 1996; Curewitz and Karson, 1997;
Gudmundsson et al., 2001; Gartrell et al., 2004; Eichhubl et al.,
2009). Simple geometric models and eld observations suggest
fault terminations and or linkages between fault segments are
domains of high fracture density and connectivity and are therefore
likely to focus uid ow (Curewitz and Karson, 1997). This was
broadly demonstrated in this study with scanlines along the Little
Grand Wash fault showing an increase in fracture density and range
of orientations (and therefore fracture connectivity) surrounding
a relay zone as opposed to a simpler section of the fault. Further-
more, distribution of uid-ow products in shale-rich footwall
sections of the studied faults were limited to areas of structural
complexity involving relay ramps or fault bends. It is inferred these
areas of structural complexity along the fault have enhanced
permeability generated by the development of a fracture network
due to complex stresses involved in fault linkage processes.
A signicant proportion of uids migrating through the study
area are migrating parallel to the fault, utilising the damage zone
fracture network of the fault with pathways focussed at areas of
structural complexity with enhanced fracture connectivity. This
study highlights the importance of a holistic approach when
considering the seal integrity of a structural closure that empha-
sises the need for detailed mapping of structures to fully assess the
risk of fault linkage creating unfavourable juxtapositions and ow
pathways for vertical migration of uids.
8. Conclusions
A protracted history of fault-related uid owis recorded within
the footwalls of the Little Grand Wash fault and northern fault of
the Salt Wash graben. Field evidence indicates hydrocarbons have
and are migrating along the Little Grand Wash fault with regional
studies indicating that migration commenced from the early
Tertiary. Iron-oxide reduction in the footwalls of the Little Grand
Wash fault and northern fault of the Salt Wash graben may be the
result of this hydrocarbon migration, though the absence of
hydrocarbons or hydrocarbon related features in the northern
footwall of the Salt Wash graben suggests alternate or multiple
uid regimes could be responsible. Sporadic sulphate veins derived
fromSO
4
2
rich meteoric waters (gypsum) and rising Pennsylvanian
brines (celestine) were precipitated during or after iron-oxide
reduction and before the start of present-day meteoric water
conditions. Precipitation of subsurface carbonate veins accompa-
nied precipitation at the surface of travertine from present-day
carbonate-rich meteoric waters. Our outcrop studies show that,
where the footwall Upper Jurassic units are juxtaposed against
shales in the hanging-wall, the fault core is dominated by low-
permeability clay-rich gouge. Conversely, the surrounding
damage zone is dominated by fractures both in the high-
permeability reservoir units and in the low-permeability seal
units. Furthermore, fracture orientation range and frequency are
enhanced in zones of structural complexity improving fracture
connectivity and providing possible pathways for uids to migrate
parallel to the fault. The presence of mineralisation and diagenesis
within and bordering fractures in the footwall of both faults
conrms that fault-parallel leakage occurred through the damage
zone via the cap rocks. Modelling of the fault rocks at depth fails to
predict such fault-parallel ow and underestimates the effect of
complex fault architectures. Modelling of the effects of faults on
uid ow must include an appreciation of likely fault architectures
and include an assessment of the relative permeabilities of the
predicted fault seal, the cap rock and the faulted cap rock.
Acknowledgements
Funding for this work came from The Carbon Capture Project
and the Trinity College Broad Curriculum Initiative. Midland Valley
and Badley Geoscience generously provided 2DMove and 3DMove
and Traptester software. John Walsh, Conrad Childs, Tom Man-
zocchi and Julian Strand provided constructive advice. Rick Allis
provided a helpful review of an early version of this manuscript,
and Peter Eichhubl, Chris Wibberley and an anonymous reviewer
provided constructive reviews for this version. The team from Utah
State University: Jim Evans, Jason Heath and Tony Williams
provided assistance and constructive discussion in the eld.
References
Allan, U.S., 1989. Model for hydrocarbon migration and entrapment within
faulted structures. American Association of Petroleum Geologists Bulletin 73,
803811.
Allis, R., White, S., Chidsey, T., Gwynn, W., Morgan, C. Adams, M., Moore, J., 2001.
Natural CO
2
reservoirs on the colorado plateau and southern rocky mountains:
candidates for CO
2
sequestration. In: Proceedings of the rst national confer-
ence on Carbon sequestration, Washington DC, May 2001.
Allis, R., Bergfeld, D., Moore, J., McClure, K., Morgan, C., Chidsey, T., Heath, J.,
McPherson B., 2005. Implications of results from CO
2
ux surveys over known
CO
2
systems for long-term monitoring. In: Proceedings of the Fourth Annual
Conference on Carbon Sequestration and Sequestration, Virginia, May 2005.
Anderson, T.R.,, Fairley, J.P., 2008. Relating permeability to the structural setting of
a fault-controlled hydrothermal system in southeast Oregon, USA. Journal of
Assayag, N., Bickle, M., Kampman, N., Becker, J., 2009. Carbon isotopic constraints on
CO
2
degassing in cold-water Geysers, Green River, Utah. Energy Procedia 1,
23612366.
Beitler, B., Chan, M.A., Parry, W.T., 2003. Bleaching of Jurassic Navajo sandstone on
colorado plateau laramide highs; evidence of exhumed hydrocarbon supergi-
ants? Geology 31, 10411044.
Bouvier, J.D., Kaars-Sijpesteijn, C.H., Kluesner, D.F., Onyejekwe, C.C., van der Pal, R.C.,
1989. Three-dimensional seismic interpretation and fault sealing investigations,
Nun River Field, Nigeria. American Association of Petroleum Geologists Bulletin
73, 13971414.
Breit, G.N., Meunier, J., 1990. Fluid inclusion, d18O, and 87Sr/86Sr evidence for the
origin of fault-controlled copper mineralization, Lisbon Valley, Utah, and Slick
Rock District, Colorado. Economic Geology 85, 884891.
Bretan, P., Yielding, G., Jones, H., 2003. Using calibrated shale gouge ratio to estimate
hydrocarbon column heights. American Association of Petroleum Geologists
Bulletin 87, 397413.
Burnside, N., Shipton, Z.K., Ellam, R.M., 2007. Using spring deposits to track the
evolution of uid pathways through faults: little Grand Wash fault and Salt
Wash graben, Utah, USA. Geological Society of America Abstracts with
Programs 39, 240.
Burnside, N., Dockrill, B., Shipton, Z.K., Ellam, R.M., 2009. Dating and constraining
leakage rates from a natural analogue for CO
2
storage: the Little Grand Wash
and Salt Wash faults, Utah, USA. in: Faults and Top Seals: From Pore to Basin
Scale, EAGE Conference volume extended abstract, Montpellier, France, 2124
September 2009.
Chan, M.A., Parry, W.T., Bowman, J.R., 2000. Diagenetic hematite and manganese
oxides and fault-related uid ow in Jurassic sandstones, Southeastern Utah.
Childs, C., Watterson, J., Walsh, J.J., 1996. A model for the structure and development
of fault zones. Journal of the Geological Society of London 153, 337340.
Childs, C., Walsh, J.J., Watterson, J., 1997. Complexity in fault zone structure and
implications for fault seal prediction. In: Mller-Pedersen, P., Koestler, A.G.
(Eds.), Hydrocarbon Seals - Importance for Exploration and Production.
Norwegian Petroleum Society (NPF), Special Publication 7, 6172.
Condon, S.M., 1997. Geology of the Pennsylvanian and Permian Cutler Group and
Permian Kaibab limestone in the Paradox basin, southeastern Utah and
southwestern Colorado. U.S. Geological Survey Bulletin 2000-P.
Curewitz, D., Karson, J.A., 1997. Structural settings of hydrothermal outow; fracture
permeability maintained by fault propagation and interaction. Journal of
Volcanology and Geothermal Research 79, 149168.
Dockrill, B., 2006. Understanding Leakage from a Fault-Sealed CO
2
Reservoir in East-
Central Utah: A Natural Analogue Applicable to CO
2
Storage. Ph.D thesis, Trinity
College Dublin.
Dockrill B., Kirschner, D.L., Shipton, Z.K., Internal structure and evolution of fault-
related active and ancient travertine deposits in East-Central Utah, U.S.A. (in
prep).
Doelling, H.H., 1994. Tufa deposits in western Grand County. Survey Notes Utah
Geological Survey 26, 810.
Doelling, H.H., Huntoon, P.W., Oviatt, C.G., 1988. Salt deformation in the Paradox
region. Utah Geological and Mineral Survey Bulletin 122.
Doelling, H.H., 2001. Geologic map of the Moab and eastern part of the San Rafael
Desert 30
0
60
0
quadrangles, Grand and Emery counties, Utah, and Mesa
County, colorado. Utah Geological Survey, Map 180.
Doelling, H.H., 2002. Interim Geological Map of the San Rafael Desert 30
0
60
0
quadrangle, Grand and Emery County. Utah: Utah Geological Survey Map
OFR-404.
Eichhubl, P., Davatzes, N.C., Becker, S.P., 2009. Structural and diagenetic Control of
uid migration and cementation along the Moab Fault, Utah. American
Association of Petroleum Geologists Bulletin 93, 653681.
Foxford, K.A., Walsh, J.J., Watterson, J., Garden, I.R., Guscott, S.C. and Burley, S.D..,
1998. Structure and content of the Moab fault zone, Utah, USA, and its impli-
cations for fault seal prediction. In Jones, G., Fisher, Q.J. and Knipe, R.J. (eds).
Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological
Society, London, Special Publications, 147, 87103.
Fulljames, R.J., Zijerverld, L.J., Franssen, R.C.M.W., 1997. Fault seal processes:
synthetic analysis of fault seals over geological and production time scales. In:
Moller, P., Koestler, A.G. (Eds.), Hydrocarbon Seals: Importance for Exploration
and Production. NPF Special Publication 7, 5159.
Gartrell, A., Zhang, Y., Lisk, M., Dewhurst, D., 2004. Fault intersections as critical
hydrocarbon leakage zones: integrated eld study and numerical modelling of
an example from the Timor Sea, Australia. Marine and Petroleum Geology 21,
11651179.
Garden, I.R., Guscott, S.C., Burley, S.D., Foxford, K.A., Walsh, J.J., Marshall, J., 2001. An
exhumed palaeo-hydrocarbon migration fairway in a faulted carrier system,
Entrada Sandstone of SE Utah, USA. Geouids 1, 195213.
Gibson, R.G., Bentham, P.A., 2003. Use of fault-seal analysis in understanding petro-
leum migration in a complexly faulted anticlinal trap, Columbus Basin, offshore
Trinidad. American Association of Petroleum Geologists Bulletin 87, 465478.
Gudmundsson, A., Berg, S.S., Lyslo, K.B., Skurtveit, E., 2001. Fracture networks
and uid transport in active fault zones. Journal of Structural Geology 23,
343353.
Hansley, P.L., 1995. Diagenetic and burial history of the lower Permian white rim
sandstone in the Tar sand triangle, Paradox basin, southeastern Utah. U.S.
Geological Survey Bulletin 2000-I.
Haszeldine, R.S., Quinn, O., England, G., Shipton, Z.K., Evans, J.P., Heath, J., Crossey, L.,
Ballentine, C.J., 2005. Natural geochemical analogues for carbon dioxide storage
and sequestration in deep geological porous reservoirsGasWater Interactions.
Special Issue of Oil & Gas Science and Technology (Revue de lInst Francais de
Petrole 60, 112.
Heath, J.E., Lachmar, T.E., Evans, J.P., Kolesar, P.T., William, A.P., 2009. Hydro-
geochemical characterization of leaking carbon dioxide-charged fault zones in
east-central Utah, with implications for geologic carbon storage. In:
McPherson, B., Sundquist, E. (Eds.), The Science and Technology of Carbon
Sequestration. American Geophysical Union (AGU) Monograph 18, 147158.
Herzog, H.J., Golomb, D., 2004. Carbon capture and storage from fossil fuel use. In:
Cleveland, C.J. (Ed.), Encyclopedia of Energy. Elsevier Science Inc, New York,
pp. 277287.
Hesthammer, J., Fossen, H., 2000. Uncertainties associated with fault sealing
analysis. Petroleum Geoscience 6, 3745.
Hintze, J.F., 1993. Geological history of Utah. Brigham Young University Geology
Studies, Special Publication 7.
Hood, J.W., Patterson, D.J., 1984. Bedrock aquifers in the northern San Rafael Swell
area, Utah, with special emphasis on the Navajo sandstone. Utah Department of
Natural Resources, Technical Publication 78.
Huffman, A.C., Lund, W.R., Godwin, L.H., 1996. Geology and resources of the Paradox
basin. Utah Geological Association Publication 25.
Huntoon, J.E., Hansley, P.L., Naeser, N.D., 1999. The search for a source rock for the
giant Tar Sand Triangle accumulation, southeastern Utah. American Association
of Petroleum Geologists Bulletin 83, 467495.
Kerrich, R., 1986. Fluid inltration into fault zones; chemical, isotopic, and
mechanical effects. Pure and Applied Geophysics 124, 225268.
Knipe, R.J., 1997. Juxtaposition and seal diagrams to help analyze fault seals in
hydrocarbon reservoirs. American Association of Petroleum Geologists Bulletin
81, 187195.
smears on fault surfaces. In: Flint, S.S., Bryant, I.D. (Eds.), The Geological
Modelling of Hydrocarbon Reservoirs and Outcrop Analogues. Special Publica-
tion of the International Association of Sedimentologists 15, 113123.
Lupton, C.T., 1914. Oil and gas near green River, Grand County, Utah. U.S. Geological
Survey Bulletin 541-D, 115133.
McKnight, E.T., 1940. Geology of the area between green and colorado rivers, Grand
and San Juan counties, Utah. U.S. Geological Survey Bulletin 908.
Micklethwaite, S., 2009. Mechanisms of faulting and permeability enhancement
during epithermal mineralisation: Cracow goldeld, Australia. Journal of
Midland Valley, 2004. 3DMove User Guide. Midland Valley, Glasgow, Scotland.
Moore, J., Adams, M., Allis, R., Lutz, S., Rauzi, S., 2005. Mineralogical and
geochemical consequences of the long-term presence of CO2
in natural reser-
voirs: an example from the SpringervilleSt. Johns Field, Arizona, and New
Mexico, U.S.A. Chemical Geology 217, 365385.
Morrison, S.J., Parry, W.T., 1986. Formation of carbonate-sulfate veins associated
with copper ore deposits from saline basin brines, Lisbon Valley, Utah; uid
inclusion and isotopic evidence. Economic Geology 81, 18531866.
Nelson, S.T., Mayo, A.L., Gilllan, S.M., Dutson, S.J., Harris, R.A., Shipton, Z.K.,
Tingey, D.G., 2009. Enhanced fracture permeability and accompanying uid
ow in the footwall of a normal fault: the Hurricane fault at Pah Tempe hot
springs, Washington County, Utah. Geological Society of America Bulletin 121,
236246.
Nicol, A., Watterson, J., Walsh, J.J., Childs, C., 1996. The shapes, major axis orienta-
tions and displacement patterns of fault surfaces. Journal of Structural Geology
18, 235248.
Nuccio, V.F., Condon, S.M., 1996. Burial and thermal history of the Paradox basin,
Utah and colorado, and the petroleum potential of the Middle Pennsylvanian
Paradox formation. U.S. Geological Survey Bulletin 2000-O, 41.
Park, H.J., West, T.R., 2002. Sampling bias of discontinuity orientation caused by
linear sampling technique. Engineering Geology 66, 99110.
Parry, W.T., Chan, M.A., Beitler, B., 2004. Chemical bleaching indicates episodes of
uid ow in deformation bands in sandstone. American Association of Petro-
leum Geologists Bulletin 88, 175191.
Peterson, P.R., 1973. Salt Wash eld. Utah Geological Survey Oil and Gas Field
Studies 4, 3.
Peterson, J.A., 1989. Geology and petroleum resources, Paradox basin Province. U.S.
Geological Survey Open File Report, OF 88-0450-U, p. 69.
Pevear, D.R., Vrolijk, P.J., Longstaffe, F.J., 1997. Timing of Moab fault displacement
and uid movement integrated with burial history using radiogenic and stable
isotopes. In: Hendry, J., Carey, P., Parnell, J., Ruffel, A., Worden, R. (Eds.),
Geouids II 1997 Extended Abstract Volume, pp. 4245.
Ross, M.L., 1998. Geology of the Tertiary intrusive centers of the La Sal Mountains,
Utah; inuence of preexisting structural features on emplacement and
morphology. U.S. Geological Survey Bulletin B-2158, 6183.
Rotevatn, A., Hesthammer J., Fossen, H., Aas, T.E., and Howell J.., 2007. Are relay ramps
conduits for uidow?Structural analysis of a relayrampinArches National Park,
Utah. In Lonergan, L., Rawnsley, R.J.H. & Sanderson, D.J. eds. Fractured Reservoirs.
Geological Society, London Special Publications, 270, 5571.
Rowland, J.V., Sibson, R.H., 2004. Structural controls on hydrothermal ow
in a segmented rift system, Taupo Volcanic Zone, New Zealand. Geouids 4,
259283.
Scholle, P.A., Stemmerik, L., Harpoth, O., 1990. Origin of major karst-associated
celestite mineralization in Karstryggen, central East Greenland. Journal of
Sedimentary Petrology 60, 397410.
Shipton, Z.K., Evans, J.P., Kirschner, D., Kolesar, P.T., Williams, A.P., and Heath, J..,
2004. Analysis of CO
2
leakage through low-permeability faults from natural
reservoirs in the Colorado Plateau, east-Central Utah. in, Baines, S.J. and R.H.
Worden, eds. Geological Storage of Carbon Dioxide. Geological Society, London,
Special Publications, 233, 359.
Shipton, Z.K, Evans, J.P., Dockrill, B., Heath, J., Williams, A.P., Kirschner, D., Kolesar,
P.T., 2005. Natural leaking CO
2
-charged systems as analogues for failed geologic
storage reservoirs. In, Benson, S.M., Oldenburg C., Hoversten M., S. Imbus, eds.,
Carbon Dioxide Capture for Storage in Deep Geological Formations- Results
from the CO
2
Capture Project, pp. 699712.
Sperrevik, S., Faerseth, R.B., Gabrielsen, R.H., 2000. Experiments on clay smear
formation along faults. Petroleum Geoscience 6, 113123.
Surdam, R.C., Jiao, Z.S., MacGowan, D.B., 1993. Redox reactions involving hydro-
carbons and mineral oxidants; a mechanism for signicant porosity enhance-
ment in sandstones. American Association of Petroleum Geologists Bulletin 77,
15091518.
Takahashi, M., 2003. Permeability change during experimental fault smearing.
Journal of Geophysical Research 108 (B5), 2235,. doi:10.1029/2002JB001984.
Trimble, L.M., Doelling, H.H., 1978. Geology and uranium-vanadium deposits of the
San Rafael River mining area, Emery County, Utah. Utah Geological and Mineral
Survey Bulletin 113, 112.
Vrolijk, P., Myers, R., Sweet, M.L., Shipton, Z.K., Dockrill, B., Evans, J.P., Heath, J.,
Williams, A.P., 2005. Anatomy of reservoir-scale normal faults in central Utah:
stratigraphic controls and implications for fault zone evolution and uid ow.
GSA Field Guide 6: Interior Western United States 6, 261282.
White, S.P., Allis, R.G., Moore, J., Chidsey, T., Morgan, C., Gwynn, W., Adams, M.,
2005. Simulation of reactive transport of injected CO
2
on the Colorado Plateau,
Utah, USA. Chemical Geology 217, 387405.
Wilkinson, M., Gilllan, S.M.V., Haszeldine, R.S., Ballentine, C.J., 2008. Plumbing the
depths testing natural tracers of subsurface CO
2
origin and migration, Utah,
USA. In: Grobe, M., Pashin, J.C., Dodge, R.L. (Eds.), Carbon Dioxide Sequestration
in Geological Media d State of the Science. AAPG Studies 59, 116.
Williams, P.L., 1964. Geology, structure, and uranium deposits of the Moab Quad-
rangle, Colorado and Utah. U.S. Geological Survey Miscellaneous Investigation
Series, Map I-360.
Williams, A.P., 2005. Structural Analysis of the Little Grand Wash and Salt Wash
faults to Investigate the Leakage of CO2 from a Natural Reservoir. M.S. thesis,
Utah State Univ.
Yielding, G., 2002. Shale Gouge ratio: calibration by geohistory. In: Koestler, A.G.,
Hunsdale, R. (Eds.), Hydrocarbon Seal Quantication. NPF Special Publication 11,
115.
Evidence of hydraulic connectivity across deformation bands from eld pumping
tests: Two examples from Tucano Basin, NE Brazil
W.E. Medeiros
a, c,
*
, A.F. do Nascimento
a, c
, F.C. Alves da Silva
b
, N. Destro
d
, J.G.A. Deme trio
e
a
Universidade Federal do Rio Grande do Norte, Centro de Ciencias Exatas e da Terra, Departamento de Geof sica, Programa de Pos-graduaa o em Geodinamica e Geof sica,
59072-970, Natal, RN, Brazil
b
Universidade Federal do Rio Grande do Norte, Centro de Ciencias Exatas e da Terra, Departamento de Geologia, Programa de Pos-graduaa o em Geodinamica e Geof sica,
59072-970, Natal, RN, Brazil
c
INCTGP, Instituto Nacional de Geof sica do Petroleo (CNPq), Brazil
d
Petrobras, Research and Development Center, Av. Horacio Macedo, 950, Cidade Universitaria, Ilha do Funda o, 21941-915, Rio de Janeiro, RJ, Brazil
e
Universidade Federal de Pernambuco, Centro de Tecnologia, Departamento de Geologia, Laboratorio de Hidrogeologia, LABHID, 50740-530, Recife, PE, Brazil
Article history:
12 August 2009
Keywords:
Pumping tests
deformation bands
Tucano Basin
hydraulic connectivity
permeability
a b s t r a c t
It is assumed that deformation bands may compartmentalize aquifers or hydrocarbon reservoirs because
these low-permeability structures may behave as barriers to uid ow. To address the question whether
there is, at a reservoir scale, hydraulic connectivity across a damage zone dominated by cataclastic
deformation bands, we present results of two pumping tests carried out in a uvial-deltaic phreatic
sandstone aquifer from the Ilhas Group at Tucano Basin (NE Brazil) where intense concentration of
deformation bands occurs. Both test sites are associated with macroscopic damage zones that are
approximately 1 km in length and 15 m thick. In situ permeability measurements show values of
2000 mD for host rock and 0.1 mD for deformation bands. GPR proles reveal good continuity of the
primary sedimentary structures with almost no deformation band vertical offsets (less than 10 cm). The
well locations for the pumping tests were chosen so that the damage zone is located between pumping
and monitoring wells. Pumping tests in both cases revealed hydraulic connectivity across the damage
zone since the observed stationary drawdown at monitoring wells was a considerable fraction of the
drawdown observed in the pumped well. In one experiment using eight monitoring wells the drawdown
cone is evolving through the damage zone instead of contouring it. Local deviations in the natural
groundwater ow allow to say that the damage zone dimensions are quite large both in horizontal and
vertical directions compared to the distances among the wells. The interpretation of all experimental
results is that deformation bands do not fully compartmentalize the aquifer. Generalization of this result
to hydrocarbon reservoirs has to take into account capillary effects which are not present in the studied
case.
1. Introduction
Deformation bands (Aydin, 1978; Antonellini and Aydin, 1994)
are one kind of frictional deformation structures found in the
uppermost Earths crust. They may be dened as tabular structures
of nite width resulting fromstrain localization commonly found in
very porous (1525% of porosity) granular material such as sand-
stones. Commonly, deformation bands exhibit localized porosity
reduction that lack shear offset (Du Bernard et al., 2002). Based on
microstructural analysis, Antonellini et al. (1994) classied the
deformations bands in three main groups: (i) deformation bands
without cataclasis: characterized by the almost complete lack of
crushed grains. They can exhibit positive or negative volume vari-
ation; (ii) deformation bands with cataclasis: characterized by
intense grain size and volume reduction, and (iii) deformation
bands with clay smearing or localized shear zones. For a compre-
hensive review on deformation bands see Fossen et al. (2007).
Deformation bands are thought to bafe uid ow during oil
production (e.g. Knipe et al., 1997; Gibson, 1994) and to act as seal
for hydrocarbon accumulations (e.g. Ogilvie and Glover, 2001)
because they tend to have lower permeability than their host rock
(e.g., Pittman, 1981; Jamison and Stearns, 1982; Antonellini and
Aydin, 1994; Knipe et al., 1997; Gibson, 1994; Fisher and Knipe,
2001; Lothe et al., 2002; Shipton et al., 2002). Tindall (2006) argued
that when connectivity across deformation bands (or damage zone)
is detected it is interpreted that this is due to open joints. However,
* Corresponding author. Fax: 55 84 3215 3683.
E-mail address: walter@geosica.ufrn.br (W.E. Medeiros).
doi:10.1016/j.jsg.2009.08.019
Fossen and Bale (2007) suggested that the impact of deformation
band on reservoir production is small or negligible in most cases
based on observation of paleouid fronts seen in the eld and on
mathematical considerations on the average permeability of rocks
affected by deformation bands.
In order to address the question whether there is, at a reser-
voir scale, connectivity across damage zones dominated by
deformation bands in porous sandstone, we show here results of
two pumping tests of a phreatic sandstone aquifer where intense
concentration of deformation bands occur. In both studied sites,
because deformation bands are more resistant to erosion than
the host rock they show a typical positive relief at the mesoscopic
scale (Fig. 1a). Microscopically they are characterized by cata-
clastic processes with grain size reduction (Fig. 1b). The degree of
cataclasis showed by the deformation bands of the Ilhas Group
can vary from moderate to high (Alves da Silva et al., 2005).
Sometimes they exhibit strong comminution with production of
gouge-like material in discontinuous stripes; in other occasions
survivor grains can be visualized within a ner tectonic matrix as
show in Fig. 1b. The deformations bands limits can be sharp or
gradational in respect to the normal sandstone and sometimes
one can nd both types of limits in the same structure, as also
shown in Fig. 1b.
Both test sites are associated to faults with a length of 1 km.
The associated damage zones have a thickness of approximately
15 m. As a general rule, the well locations were chosen so that
the deformation bands are located between pumping and
monitoring wells. We stress that the results of our pumping tests
must be interpreted in terms of piezometric head connectivity
between the wells and not in terms of mass transfer between
them.
We are going to present strong evidence that there is hydraulic
connectivity across the deformation bands/damage zones on
a reservoir scale. In the following sections we present the regional
geographical and geological contexts of the test sites, petrophys-
ical and geophysical characterization of the aquifer sandstone
where the tests were performed, description of the pumping tests
methodology and their results and nally discussions and
conclusions.
Fig. 1. Deformation bands in outcrops at Site 1: (a) Mesoscopic aspect of the defor-
mation band. Note the positive relieve of the structures due to their greater resistance
to erosion. (b) Microscopic view of a deformation band exhibiting evidences of cata-
clasis. The lower contact is sharper than the upper one and is shown by a dashed line.
Note the great reduction of grain size in the tectonic matrix. Some survivor grains (S)
are still seen along the deformation band.
Fig. 2. Geological sketch of Sites 1 and 2 (S1 and S2 respectively) in Tucano Basin, Bahia State, NE Brazil.
W.E. Medeiros et al. / Journal of Structural Geology 32 (2010) 17831791 1784
2. Tucano Basin, NE-Brazil: geographical and geological
contexts
In Tucano Basin (Bahia State, NE Brazil), porous sandstones
exhibit abundant deformation bands and related fault outcrops.
This basin is the central part of the intracontinental Reconcavo
TucanoJatoba rift (RTJ) which is composed by N-S and NE-SW half
grabens with upper Jurassic to lower Cretaceous sedimentary
inllings. Regionally speaking, the RTJ shows a NS trend related to
the stress eld that stretched the continental lithosphere in
association to the Gondwana break-up and formation of the South
Atlantic Ocean to the east of the rift system (Magnavita and
Cupertino, 1987; Magnavita, 1992).
The Tucano Basin is characterized by sets of NS trending normal
faults that tilted the sedimentary strata eastward displaying
a domino-like structural style (Magnavita and Cupertino, 1987;
Magnavita, 1992). The basin sedimentary inllings can be related to
distinct tectonic stages: deposition of the Brotas Group (Aliana and
Sergi Formations) and the basal portion of the Santo Amaro Group
(Itaparica Formation) represent the pre-rift stage whilst deposition
of Candeias Formation (Santo Amaro Group), Ilhas and Massacara
Groups, and Salvador Formation are associated to the rift phase
(Caixeta et al., 1994).
The pumping tests described hereafter were done in sites
dominated by the sandstones of the Ilhas Group (Fig. 2). A uvial-
deltaic sedimentation is proposed to this group (Cupertino, 1990).
The two chosen sites are located near the south margin of the Vaza-
Barris River (Fig. 2). Although the Vaza-Barris River is the main river
of the region it is intermittent because the climate in the area is dry.
The main rain season usually occurs from May to August.
In outcrops at the studied sites (Fig. 3ac) the normal faults
trend NNE (Fig. 3c) and dip steeply to the ESE (Figs. 3b and 4). In the
damage zone of such faults three main sets of deformations bands
are intensely developed (Fig. 4).
3. Petrophysical and geophysical characterization of the Ilhas
Group sandstone
Because of the excellent visual exposure, Site 1 was chosen for
conducting GPR (Ground Penetrating Radar) and in situ perme-
ability measurements. Fig. 5 shows the location of the GPR and
permeability measurement proles. This excellent visual exposure
makes Site 1 as a type-outcrop and a lot of care had to be taken to
preserve it, thus posing operational problems in particular for well
drilling.
Fig. 6 shows a GPR section transverse to the main deformation
bands. The survey was done using a 200 MHz antenna connected to
a GSSI SIR-System2. The data processing (Miranda, 2004) was done
in such a way to highlight the deformation bands. The GPR pro-
cessing ow begins with a trace editing to remove anomalous
amplitudes and signal sampling errors. Afterwards, a zero-offset
correction was done by removing aerial and direct wave arrivals.
Then, low frequency electromagnetic induction effects between
pairs of antennae were removed with a high-pass band lter. After
this preliminary processing spherical and exponential correction
together with spectral balancing were applied to the traces to
restore amplitude and frequency content due to attenuation (Xav-
ier Neto and Medeiros, 2006). Because noise content is also
enhanced in this operation, a low-pass lter was applied. Finally,
migration and topographic correction was done.
In Fig. 6, we interpret the abundant white subvertical stripes as
deformation bands. We indicate some of them by a black ellipsis.
We think this interpretation is correct based on two facts: (i)
correlation between scan line and GPR prole (Miranda, 2004)
and (ii) grain reduction associated to deformation band would
change porosity and water adsorption thus changing the GPR
response. Note that despite the pervasiveness of the interpreted
deformation bands they are poorly interconnected evidencing
that the undamaged portion of the sandstone remains inter-
connected regardless the presence of a complex set of deforma-
tion bands. Additionally, the GPR section reveals good continuity
of the primary sedimentary structures showing that there is
almost no vertical offset of the deformation bands. These two
observations are corroborated by eld data. We are condent that
if we show this section to an interpreter without telling him/her
Fig. 3. (a) Outcrop showing deformation bands in different orientations in Site 1.
(b) View of the damage zone in the Site 1 showing a swarm of deformation bands.
(c) Fault plane with part of the associated damage zone (upper part of the picture) in
the Site 2. The hammer handle points towards the North. U and D identify the foot
wall and hanging wall respectively.
about the presence of the deformation bands, chances are that the
interpreter would not infer these deformation bands. The dif-
culty to image the deformation bands was quite surprising for us
since we expected they would have a higher impact in the image.
Probably, the reason for that is twofold: (1) the deformation bands
are dominantly subvertical with no offset and (2) they occur in
rocks of the same lithology. Certainly a similar effect occurs with
seismic data in reservoir scale. If that is the case, conventional
processing and/or interpreting seismic data would not
Fig. 4. (a) Photo and (b) stereograms showing the three main sets of deformation bands in Site 1. Two sets are subvertical (1 and 3) and the other one (2) is subhorizontal. The
hammer points towards the North.
Fig. 5. Aerial photo of Site 1 showing the location of the GPR and permeability
measurement proles (continuous and dashed black lines, respectively) and pumping
well locations (B1, B2 and B3).
Fig. 6. Ground Penetrating Radar section (200 MHz) transverse to the deformation
bands in Site 1. Pervasive white stripes are interpreted as deformation bands, as shown
by the black ellipsis.
straightforwardly identify deformation bands. Therefore, even
when their presence is suspected, the impact of the deformation
bands in seismic data would potentially be better evaluated with
very specic data processing maybe similar to fracture and joint
detection (e.g. Liu et al., 2005).
In situ permeability measurements were done using
a portable PPP-250 minipermeameter from Core Laboratories
Company. The working principle of this device is based on the
diffusion of an air pulse injected into the rock (Goggin et al.,
1988). According to Antonellini and Aydin (1994) the sampled
rock volume with this device is about 5pr
3
where r is the radius
of the cylinder (gun) from where the air pulse is injected. In our
case, this volume is around 0.25 cm
3
. To assure reliability of the
measurements, we checked regularly the device using a ceramic
sample with known permeability provided by the manufacturer.
For example, in one of these tests, the known value of 7.5 mD of
the ceramic sample was reproduced with average 7.43 mD and
standard deviation 0.39 mD after 100 measurements. During the
permeability measurements on the rock, approximately one
check of the device with the ceramic sample was done at every
30 measurements. Fig. 7 shows a typical permeability scan line
measurement transverse to the deformation bands. The location
of the scan line is shown in Fig. 5. Before measurements the rock
surface was prepared to remove a thin weathering layer and to
allow better coupling between the minipermeameter and the
rock surface. Permeability measurements on the rock surface
were done at every 0.2 m and every value shown in Fig. 7 is the
mean of three measurements done in the same point. The high
values of permeability (up to 2000 mD) are associated to the host
rock whilst low values (down to 0.1 mD) are associated to
deformation bands. The high variability (up to four orders of
magnitude) revels that a single scan line crosses many defor-
mation bands at this scale.
4. Pumping tests
4.1. Site 1
Site 1 exhibits an excellent visual exposure of deformation
bands, especially on the horizontal cross section (Fig. 3b). Because
two dominant sets of deformation bands in the area are subvertical
(Fig. 4), the outcrop oor shows the horizontal cuts of the defor-
mation bands revealing a great amount of their intersections and
complexity (Fig. 3a, b). However, the wall of the main deformation
band (fault?) is not visible so that subvertical offsets could not be
measured and/or inferred. Judging also from the GPR data, there is
no signicant vertical offset.
Three wells were drilled around Site 1 (B1, B2 and B3 in Fig. 5).
All three wells were drilled and cased in the same manner so that
any of themcould either be used to pump water fromor to measure
the drawdown. The wells are 30 m long and are cased with 6
diameter steel tubes. The steel casing is hydraulically opened from
12 to 28 mdeep so that lters could be installed in this depth range.
The remaining parts of the wells are not connected to the aquifer.
Static water levels were measured in all three wells. They were
the same within 0.2 m. This difference is due to the natural
piezometric head. The local topography is at (maximum differ-
ence of 0.4 m) and the water table is in average 7 m deep. From the
hydrogeological viewpoint, the aquifer in the area may be assumed
as phreatic.
All pumping tests were done in December 2005 (the dry season
in the area) using the same equipment, technical team and
following similar procedures. While one of the wells was pumped,
Fig. 7. Scan line of in situ permeability measurements transverse to the deformation
bands in Site 1.
Fig. 8. Observed drawdown curves for pumping tests in Site 1. (a) At B2 and B3 when
pumping from B1. (b) At B1 and B3 when pumping from B2. (c) At B1 and B2 when
pumping from B3.
the observed drawdowns in all three wells were monitored. All
drawndown measurements were done with either automatic or
manual level loggers. Three constant ow rates of 1.60, 1.13 and
1.38 m
3
/h were used during the tests for B1, B2 and B3, respectively.
These values were determined with pre-tests in order to attain
stationary conditions. B1 and B2 were pumped during 24 and 12 h
followed by 12 and 6 h, respectively, of recovery in order to rees-
tablish the natural piezometric head. The pumping test from B3
was scheduled to occur in the same manner as pumping test from
B2 but due to operational problems, the test was interrupted after
6 h of pumping and the piezometric head reestablishment was not
achieved. Adequate disposal of the waste water was done in order
to avoid its inuence on the aquifer recharge.
Results of the pumping tests for each pair of monitoring wells
are shown in Fig. 8ac (Alves da Silva and Destro, 2008). Each of
these gures shows the drawdown measured in the two moni-
toring wells while pumping from a third one. As usual, the x-axis
variable in each of these plots is expressed as [time/distance
2
] in
order to take into account the combined effect of pumping time and
distance between pairs of pumping and observation wells (de
Marsily, 1986). In addition, the maximum drawdowns observed in
the pumping wells were 14.0, 6.5, and 16.5 m when pumping from
B1, B2, and B3, respectively.
According to local farmers, similar wells typically show yields
higher (w40 m
3
/h) than the ones obtained in our tests. This fact
reveals that the macroscopic values of the permeability and
transmissivity of the aquifer at the test site are relatively low
because of the presence of the deformation bands. Fig. 8ac reveals
hydraulic connectivity across the damage zone since the observed
stationary drawdowns at monitoring wells were considerable
fractions of the drawdowns observed in each pumped well.
In all pumping tests, samples of the waste water were collected
in order to measure their electric conductivity. Fig. 9 shows in
a single plot these values measured in the three different tests
(Alves da Silva and Destro, 2008). It is important to bear in mind
that these measurements were not simultaneous and the time in
the x-axis refers to the pumping interval time in each test.
Fig. 9 shows that the initial values for water conductivity are
quite different for the three wells. However, as pumping continues,
there is a tendency that water electric conductivity values reach
a common value around 360 mS/m. We give belowan interpretation
of this fact based on simple uid ow simulation in an equivalent
isotropic homogeneous aquifer with similar ow rates as those
obtained in the pumping tests. This simulation reveals that uid
particle motion near the pumping wells are limited to a radial
distance of about 3 m after 24 h of pumping. Thus, given that after
Fig. 10. Sketch map of Site 2 showing fault and damage zone, the Brejinho village and
the well locations (P1P9). U and D identify the foot wall and hanging wall,
respectively.
Fig. 11. Natural piezometric head (cm) in Site 2. Black arrow shows the inferred main
direction for the groundwater ow. U and D identify the foot wall and hanging
wall, respectively.
Fig. 9. Electrical conductivity measurements of the waste water from tests in Site 1.
Table 1
Information about the pumping tests in Site 2.
Pumping
well
Flow
rate
(m
3
/h)
Pumping
time
(h)
Time to
reestablish
the natural
piezometric
head
Monitoring wells
during
pumping time
Monitoring wells
measuring the
time to
re-establish the
natural
piezometric
head
P1 5.45 48 24 h All wells
including P1
P2, P3, P4, P6,
P7 and P8
P2 6.14 24 Ranging from
1 h 40 min
to 2 h
All wells
including P2
P1, P3, P4, P5,
P6 and P8
P5 4.64 24 10 h All wells
including P5
P1, P3, P4,
P7 and P8
24 h of pumping the water electric conductivity reaches the same
value (w360 mS/m), a volume of around 3
3
m
3
is a fair estimate of
the representative elementary volume (REV) for the entire aquifer.
Note that this REV is much larger than the usual values for homo-
geneous and isotropic aquifers (w10
3
m
3
). So, this is evidence that
there is difculty for water circulation and mixing inside the aquifer
due to the presence of the deformation bands and the REV is
comparatively larger to take into account the high heterogeneity
due to the deformation bands.
Pumping test results fromSite 1 suggest that deformation bands
do not fully compartmentalize the aquifer. Some issues may be
raised regarding this result:
1. At a meso-scale, the fault zone (and consequently its damage
zone) is not exposed 100 m away (northwards) from the wells
(Fig. 5). If the outcrop absence represents a termination of the
damage zone, where the intensity of the deformation band is
expected to diminish, the piezometric head connection
between pairs of wells in each side of the damage zone would
occur by contouring this termination. Because physical integ-
rity of Site 1 was a major constraint, we could not drill wells
between B1 and B2 in order to monitor the drawdown cone
evolution so that this possibility could not be ruled out in this
test.
2. The damage zone could have a small vertical extent when
compared with its horizontal dimension. Therefore, analo-
gously to the previous issue, piezometric head connectivity
would occur by contouring the hypothetical in-depth vertical
termination.
3. And isolated macroscopic subvertical fracture transverse to the
deformation bands was observed thus posing the possibility
that the observed piezometric head connectivity could be due
to (or enhanced by) this fracture.
Because these issues could not be addressed with the pumping
tests in Site 1, we decided to choose another site, for a second test,
carefully selected in order to honour two additional criteria in
relation to Site 1: (i) fracture absence at least from a visual
inspection view point in the site of the new test, and (ii) possibility
to drill a group of wells so that the cone of depression could be
better monitored and we could be more condent that the draw-
down is evolving through the damage zone instead of contouring it.
Site 2 (Fig. 2) ts both additional criteria above described.
Regarding the vertical dimension issue we raised above, we can
assert for now that the likelihood of this being the dominant factor
in two tests is reduced.
4.2. Site 2
Eight vertical wells were drilled on Site 2 (P1P8 in Fig. 10). In
addition, a pre-existing well (P9) that supplies water for the local
Brejinho community was used as a monitoring well. Wells P1P8
were drilled in the same manner with depths ranging from 30 to
35 m and only P1, P2 and P5 were fully cased with a 4 diameter
PVC tube because they were chosen to be pumping wells. The
remaining ones, used only for drawdown monitoring, were cased
only on their top 5 m to avoid collapse. PVC tubes in P1, P2 and P5
were open from approximately 15 to 27 m to allow hydraulic
connectivity to the aquifer. Regarding P9, unfortunately no infor-
mation on how this well was drilled and constructed was available.
All pumping tests were done using the same equipment, tech-
nical team and following similar procedures to those described in
Fig. 12. Piezometric head maps (m) after 24 (a) and 48 h (b) of pumping from P1 in
Site 2. U and D identify the foot wall and hanging wall respectively.
Fig. 13. Electrical conductivity measurements of the waste water from tests in Site 2.
Site 1. All tests were done in January 2007 still within the dry
season in order to maintain natural conditions similar to those at
the time of the tests in Site 1. The natural piezometric head is
shown in Fig. 11. One can observe that the piezometric head
gradients are small (around 1.6 cm/m) and that the main natural
water ux is towards the ESE direction. There are, however,
important deviations in direction resembling wave refraction,
indicating that the fault zone is large both in horizontal and vertical
directions. As consequence, the issue 3 raised above may be ruled
out. The water table static level is nearly the same as that found in
Site 1, which corroborates the hypothesis that a phreatic aquifer
model is valid in the region.
Table 1 shows pumping test information that has been gathered.
As in the previous tests in Site 1, adequate disposal of the waste
water was done. Despite the fact the ow rates are around four
times greater than those obtained from tests in Site 1, they are still
small when compared to typical yields in the region. For the sake of
clarity, we are presenting piezometric head maps after 24 and 48 h
of pumping from P1 only (Fig. 12a, b); results obtained when
pumping from P2 and P5 are similar.
Both piezometric head maps in Fig. 12a, b show that the draw-
down cone is indeed evolving through the damage zone instead of
contouring it. Moreover, the drawdown cone shapes reveal the
presence of a highly anisotropic structure. For instance, in Fig. 12b,
after 48 h of pumping fromP1, the isovalue line of 0.2 m reaches P5
and P9 despite the fact that the distance fromP1 to P9 is more than
twice the distance from P1 to P5. Since P1 and P9 are at the same
side of the fault/damage zone, this is consistent with our expecta-
tion that the drawdown cone will reach more easily P9 for a given
value of drawdown.
As in previous tests in Site 1, samples of the waste water were
collected in order to measure their electric conductivity (Fig. 13).
Regarding the measurements of electrical conductivity of the waste
water, the observation time was not enough to draw similar
conclusions to those described for the pumping tests in Site 1.
However, the fact that the initial values for the water electrical
conductivity observed in different wells may be different is also
valid here. If the hypothesis we raised about the local uid mobility
around the pumped wells is correct, we could say that for the case
of Site 2, the mobility is far less than the one inferred from the data
in Site 1. In fact, if convergence to a single background level of
electrical conductivity is to be attained in Site 2, the estimated
pumping time is 10
6
min (approximately 2 years).
5. Discussion and conclusion
Regarding the condence of the assertion that deformation
bands/damage zones do not fully compartmentalize the aquifer,
results from Site 2 fulll some aspects that remained open after
tests in Site 1. These open aspects are associated to the niteness
(compared to the distances among wells) either in horizontal or
vertical dimension of the damage zones and also to the presence of
open fractures. Results from Site 2 show that the drawdown is
indeed evolving through the damage zone and is not contouring it.
Additionally, local deviations in the natural groundwater ow are
consistent with the fact that the vertical dimensions of the damage
zone are large compared to the distances among wells. Finally, from
eld observations, there are no open fractures in Site 2.
We stress that the main strength of our results is that they are
based on real tests where the pumped aquifer is at a similar scale of
an oil reservoir. We are aware that in any real test it is impossible to
have control in all aspects that eventually may inuence the results.
Additionally, in aquifers, capillary effects are not important whilst
in hydrocarbon reservoir they play an important role on compart-
mentalization. For a comprehensive discussion on the effects of
capillary pressure in sandstones containing deformation bands,
please see Fossen and Bale (2007) and references therein.
The conclusion that deformation bands/damage zones do not
fully compartmentalize the studied aquifer is supported by the
inspection of the geometry of these structures. It is our experience
that analysis of deformation bands from aerial photos, outcrops,
GPR images and thin sections evidence that individually such
structures present terminations in all scales although the dimen-
sion of a group of these structures may continue and be quite larger
than individual structures. In other words, the undamaged portion
of the sandstone remains three-dimensionally interconnected
regardless the presence of an intricate set of deformation bands.
Thus, a percolation core of the undamaged reservoir rock is
preserved allowing at least partial hydraulic connection through it.
In fact, the hydraulic connection through the percolation core of
the undamaged rock is clear from the uid ow simulation results
presented by Lunn et al. (2008). As Lunn et al. (2008) state: .it is
the connected nature of highly tortuous high-permeability path-
ways that governs the bulk hydraulic properties of the fault zone.
Our main results are also in agreement with Fossen and Bale (2007)
suggestion that deformation bands may not inuence uid ow in
a production setting because, among other reasons, these struc-
tures may present low connectivity so that uids are free to ow
around and between bands. In this aspect, the deformation bands
here play a similar role to faults that despite their local sealing
nature may not fully compartmentalize a reservoir as demonstrated
by Medeiros et al. (2007) for the case of a carbonate reservoir.
Nonetheless the above arguments concerning the possible
general validity of our results, we are aware that at the moment
they are limited to the case of cataclastic deformation bands
occurring in sandstones of uvial-deltaic origin, which is the case of
the studied Ilhas Group sandstone. We propose that tests in aqui-
fers containing deformation bands of other types should be carried
out to verify if similar results are found.
Acknowledgements
The authors wish to thank PETROBRAS and FINEP for the
nancial support for this research. PETROBRAS is thanked again for
the permission to publish this paper. W.E. Medeiros (Proc. 301568/
2008-1), A.F. do Nascimento (Proc. 303706/2008-2), and F.C.A. Silva
(Proc. 309289/2006-8) thank CNPq for their research fellowships.
The authors also thank Emanuel Jardim de Sa , Patrcia Costa and
Ingred Guedes for their contribution with structural geology
insights and data. Jose Moreira and Jose Quirino are also thanked
for helping with the GPR survey. Hugo Miranda is thanked for the in
situ permeability tests and also for helping with the GPR survey. We
thank Olivar Lima for providing the minipermeameter. Pedro Souto
and his team are thanked for performing the wells drilling. We
express our gratitude also for the careful reviews fromVictor Bense,
Silje Berg, and Jerry Fairley because they raised many important
points and suggested literature, which improved a lot the original
manuscript. And nally, the last but not the least, we thank the
A
gua Branca Village community for their hospitality and support

along all the phases of the project.
References
Alves da Silva, F.C., Destro, N., 2008. Falhas e fraturas naturais: aplicaoes na car-
acterizaao de reservato rios. UFRN/FINEP/Petrobras. Petrobras Internal Report
No. 650-29397, 276 pp. (in Portuguese).
Alves da Silva, F.C., Ferreira, T.S., Costa, P.R.C., Guedes, I.M.G., Jardim de Sa , E.F., 2005.
Aspectos microtexturais, geome tricos e mecanismos de deformaao de bandas
de deformaao e fraturas em arenitos porosos da bacia de Tucano (BA): inu-
encia na circulaao de uidos. In: XXI Simpo sio de Geologia do Nordeste, Recife,
2005, Resumos Expandidos, pp. 194198 (in Portuguese).
Antonellini, M.A., Aydin, A., 1994. Effect of faulting on uid ow in porous sand-
stones: petrophysical properties. American Association of Petroleum Geologists
Antonellini, M.A., Aydin, A., Pollard, D.D., 1994. Petrophysical study of fault in
sandstone using petrographic image analysis and X-ray computerized tomog-
raphy. Pure and Applied Geophysics 143, 181201.
Aydin, A., 1978. Small faults formed as deformation bands in sandstone. Pure and
Applied Geophysics 116, 913929.
Caixeta, J.M., Bueno, G.V., Magnavita, L.P., Feijo , F.J., 1994. Bacias do Reconcavo, Tucano
e Jatoba . Boletim de Geociencias da Petrobras 8, 163172 (in Portuguese).
Cupertino, J.A., 1990. Esta gio Explorato rio das Bacias do Tucano Central, Norte e
Jatoba . Boletim de Geociencias da Petrobras 4, 4554 (in Portuguese).
de Marsily, G., 1986. Quantitative Hydrology. Academic Press, New York.
Du Bernard, X., Eichhubl, P., Aydin, A., 2002. Dilation bands: A new form of localized
failure in granular media. Geophysical Research Letters 29, 2176. doi:10.1029/
2002GL015966.
reservoirs of the North Sea and Norwegian continental shelf. Marine and
Fossen, H., Bale, A., 2007. Deformation bands and their inuence on uid ow.
stone a review. Journal of the Geological Society 164, 755769.
Gibson, R.G., 1994. Fault-zone seals in siliciclatic strata of the Columbus Basin,
offshore Trinidad. American Association of Petroleum Geologists Bulletin 78,
13721385.
Goggin, D.J., Thrasher, R.L., Lake, L.W., 1988. A theoretical and experimental
analysis of minipermeameter response including gas slippage and high
velocity ow effects. In situ 12, 79116.
Jamison, W.R., Stearns, D.W., 1982. Tectonic deformation of Wingate Sandstone,
Colorado National Monument. American Association of Petroleum Geologists
Bulletin 66, 25842608.
Knipe, R.J., Fisher, Q.J., Clennell, M.R., Farmer, A.B., Harrison, A.B., Kidd, B.,
McAllister, E., Porter, J.R., White, E.A., 1997. Seal analysis: successful method-
ologies, application and future directions. In: Mller-Pedersen, P., Koestler, A.G.
(Eds.), Hydrocarbon Seals: Importance for Exploration and Production.
Norwegian Petroleum Society (NPF) Special Publication 7, pp. 1540.
Liu, E., Chapman, M., Hudson, J.A., Tod, S.R., Maultzsch, S., Li, X.Y., 2005. Quantitative
determination of hydraulic properties of fractured rock using seismic tech-
niques. Geological Society Special Publications 249, 2942.
Lothe, A.E., Gabrielsen, R.H., Bjrnevoll-Hagen, N., Larsen, B.T., 2002. An experi-
mental study of the texture of deformation bands: effects on the porosity and
permeability of sandstones. Petroleum Geoscience 8, 195207.
Lunn, R.J., Shipton, Z.K., Bright, A.M., 2008. How can we improve estimates of bulk
fault zone hydraulic properties? In: Wibberley, C.A.J., Kurz, W., Imber, J.,
Implications for Mechanical and Fluid-Flow Properties Geological Society,
London, Special Publication, 299.
Magnavita, L.P., Cupertino, J.A., 1987. Concepao atual sobre as bacias do Tucano e
Jatoba , Jatoba , Nordeste do Brasil. Boletim de Geociencias da Petrobras 1, 119
134 (in Portuguese).
Magnavita, L.P., 1992. Geometry and kinematics of the ReconcavoTucanoJatoba
rift. NE Brazil. PhD thesis, University of Oxford.
Medeiros, W.E., Nascimento, A.F., do Antunes, A.F., Jardim de Sa , E.F., Lima Neto, F.F.,
2007. Spatial pressure compartmentalization in faulted reservoirs as a conse-
quence of fault connectivity: a uid ow modelling perspective, Xare u oil eld,
NE Brazil. Petroleum Geoscience 13, 341352.
Miranda, H.C.B., 2004. Interpretaao conjunta de dados de GPR e medidas de per-
meabilidade sobre um ana logo de reservato rio silicicla stico falhado na Bacia de
Tucano. NE do Brasil. Universidade Federal do Rio Grande do Norte, MSc Thesis,
in Portuguese.
Ogilvie, S.R., Glover, P.W.J., 2001. The petrophysical properties of deformationbands in
relationtotheir microstructure. EarthandPlanetaryScienceLetters 193, 129142.
Pittman, E.D., 1981. Effect of fault-related granulation on porosity and permeability
of quartz sandstones, Simpson Group (Ordovician), Oklahoma. American
Shipton, Z.K., Evans, J.P., Robeson, K., Forster, C.B., Snelgrove, S., 2002. Structural
heterogeneity and permeability in faulted aeolian sandstone: implications for
subsurface modelling of faults. American Association of Petroleum Geologists
Tindall, S.E., 2006. Jointed deformation bands may not compartmentalize reser-
voirs. American Association of Petroleum Geologists Bulletin 90, 177192.
Xavier Neto, P., Medeiros, W.E., 2006. A practical approach to correct attenuation
effects in GPR data. Journal of Applied Geophysics 59, 140151.
Relationship between fault growth mechanism and permeability variations with
depth of siliceous mudstones in northern Hokkaido, Japan
Eiichi Ishii
a,
*
, Hironori Funaki
a
, Tetsuya Tokiwa
a
, Kunio Ota
b
a
Horonobe Underground Research Unit, Japan Atomic Energy Agency, Hokushin 432-2, Horonobe-cho, Hokkaido 098-3224, Japan
b
Geological Isolation Research and Development Directorate, Japan Atomic Energy Agency, Muramatsu 4-33, Tokai-mura, Ibaraki 319-1144, Japan
Article history:
6 October 2009
Accepted 30 October 2009
Keywords:
Fault
Shear
Tensile
Stress
Permeability
Siliceous mudstone
a b s t r a c t
In order to assess the inuence of remote mean stress correlated with depth of burial on the principal
mode of failure at fault tips during fault slip in a lithologically homogeneous, fractured rock mass, the
growth mechanisms of strike-slip faults have been studied at outcrop-scale in the siliceous mudstones of
northern Hokkaido, Japan. We take a multifaceted approach combining i) geological characterization of
fractures by fracture mapping in outcrop and fracture logging of boreholes (drilling depth: 1020 m),
ii) rock mechanical characterization by laboratory tests on core samples, and iii) theoretical analyses
using the GrifthCoulomb criterion. These suggested that the principal mode of failure in the
mudstones is dependent, not only on rock strength, but also on remote mean stresses. During and/or
after uplift and erosion the faults grew mainly by linking with adjacent faults via numerous splay cracks,
formed by tensile failure above roughly 400 m depth. In contrast, below this depth, the faults grew
predominantly by shear failure. Such growth mechanisms are consistent with the fact that hydraulic tests
performed in boreholes show that highly permeable sections (hydraulic transmissivity: >10
5
m
2
/s) are
restricted to depths of less than 400 m.
1. Introduction
Decreasing permeability with depth in fractured rock masses
(e.g., at depths of less than 1 km), which is consistent with closure
of pore apertures with increasing pressure (e.g., Wei et al., 1995;
O
hman et al., 2005), has been widely recognized in deep drilling

programs worldwide (e.g., Rhe n et al., 2006; Andersson et al., 2007;
Nordqvist et al., 2008). However, clear exceptions to this trend exist
and numerous authors have suggested that permeability distribu-
tions in fractured rock masses are fundamentally dependent on
mechanisms of their brittle deformation (e.g., Caine et al., 1996;
Mazurek et al., 1996, 1998, 2003; Evans et al., 1997; Gutmanis et al.,
1998; Mazurek, 1998, 2000; Bossart et al., 2001).
Sheared-joint-based faulting associated with the numerous
tensile splay cracks that propagate fromthe tip of faults when faults
slip is a well known mechanism of brittle deformation (e.g., Segall
and Pollard, 1983; Martel and Pollard, 1989; Flodin and Aydin,
2004; Myers and Aydin, 2004). This style of faulting is known to
signicantly increase the permeability of a fractured rock mass
(e.g., Mazurek et al., 1998, 2003; Dholakia et al., 1998; dAlessio and
Martel, 2004; Lunn et al., 2008; Eichhubl et al., 2009). Mazurek
et al. (1998) showed that the transmissivity of groundwater inow
points in boreholes penetrating the highly consolidated and frac-
tured marls in the central Swiss Alps decreases with depth, and
indicated that the tensile splay cracks forming the fault linkages
may be the actual inow points. This is an example of an earlier
study that implied a relationship between splays and permeability
variations with depth. However, the relationship has not been
examined in detail.
In order for tensile splay cracks to form, tensile failure is
required near fault tips when fault slip occurs. However, such
tensile failure may not necessarily occur under all stress states; it
is possible that shear failure could occur. The principal mode of
failure is also known to depend on lithology, i.e., tensile failure
tends to occur in stronger rocks while shear failure tends to occur
in weaker rocks (e.g., Gross, 1995; Eichhubl and Boles, 2000;
Welch et al., 2009). Also concerning the principal mode of failure
near fault tips during sheared-joint-based faulting, the control of
failure mode by lithology is shown by eld observations on the
Monterey Formation of California (Dholakia et al., 1998). However,
it is indicated by numerical simulations using the Grifth
Coulomb criterion that the principal mode of failure near fault tips
due to fault slip depends not only on lithology but also on remote
mean stress at the time of fault slip (Bourne and Willemse, 2001),
* Corresponding author. Fax: 81 1632 5 2344.
E-mail address: ishii.eiichi@jaea.go.jp (E. Ishii).
doi:10.1016/j.jsg.2009.10.012
which suggests that tensile failure tends to occur under lower
remote mean stress and shear failure tends to occur under higher
remote mean stress. Therefore, remote mean stress is also an
important factor in failure mode, and, considering that remote
mean stress is generally correlated with burial depth (e.g., Twiss
and Moores, 2007), the difference in the principal failure mode
with increasing burial depth may also have a direct relationship to
variations in permeability with depth, as observed in fractured
rock masses.
In the Horonobe area, northern Hokkaido, Japan, folded
Neogene siliceous mudstone, the Wakkanai Formation (Figs. 1 and
2), is massive and lithologically homogeneous (Iijima and Tada,
1981) with very weakly developed bedding planes (Ishii et al.,
2006). Faults crosscutting bedding planes at a high angle (FCBs) and
bedding faults parallel to bedding are observed at outcrop-scale
(Ishii and Fukushima, 2006). The FCBs are mainly strike-slip faults
oblique to fold axes. Geological and hydrogeological data indicate
that some of the FCBs are the main owpaths (Ishii and Fukushima,
2006; Kurikami et al., 2008; Funaki et al., 2009). Furthermore,
hydraulic packer tests performed in twelve deep, vertical boreholes
(drilling depth: 1020 m) show that the sections with highest
permeability (hydraulic transmissivity: >10
5
m
2
/s) are restricted
to depths of less than about 400 m in the formation (Fig. 3).
In this study, a case study addressing the inuence of remote
mean stress (i.e. burial depth) on the principal mode of failure
Fig. 1. Geological map and geological cross-section in the Horonobe area (after Ishii et al., 2008), showing locations of the cleared outcrop and the boreholes. Plate boundaries and
directions of plate migration in location map of the left upper are from Wei and Seno (1998).
Fig. 2. Schematic columnar section in the study area (modied Ota et al., 2007).
Sb: Sarabetsu Formation; Yc: Yuchi Formation; Kt; Koetoi Formation; Wk: Wakkanai
Formation.
E. Ishii et al. / Journal of Structural Geology 32 (2010) 17921805 1793
near fault tips during fault slip in a lithologically homogeneous,
fractured rock mass is discussed. Taking it one step further, the
relationship between failure mode and permeability, that is,
the relationship between the growth mechanisms of the FCBs and
the permeability variations with depth of the Wakkanai Formation
is discussed. For this study, related to the parameter burial depth,
vertical borehole data are important. However, information on the
faults reliant solely on restricted information from borehole core
data is not enough. Hence, a multifaceted approach combining
outcrop surveying, borehole investigations, and theoretical anal-
yses is applied for the Wakkanai Formation. The study combines
i) geological characterization of fractures by fracture mapping in
outcrop and fracture logging of boreholes, ii) rock mechanical
characterization by laboratory tests on core samples for tensile
strength and angle of internal friction, and iii) theoretical analyses
using the GrifthCoulomb criterion.
The Horonobe area is located on the eastern margin of
a Neogene to Quaternary sedimentary basin located on the western
side of northern Hokkaido, in a Quaternary, active foreland fold-
and-thrust belt near the boundary between the Okhotsk and
Amurian plates (e.g., Yamamoto, 1979; Wei and Seno, 1998; Ikeda,
2002; Fig. 1). The basin grades upward stratigraphically from the
Wakkanai Formation (massive siliceous mudstones including opal-
CT), to the Koetoi Formation (massive diatomaceous mudstones
including opal-A, but not opal-CT), to the Yuchi Formation (ne to
mediumgrained sandstones), and lastly to the Sarabetsu Formation
(alternating beds of conglomerate, sandstone and mudstone,
intercalated with coal seams), overlain by late Pleistocene to
Holocene deposits (Figs. 1 and 2). The Wakkanai and Koetoi
Formations have the following characteristics: i) they have fairly
uniform muddy facies with rare intercalation of pyroclastic sedi-
ments (Mitsui and Taguchi, 1977; Tada and Iijima, 1982; Ishii et al.,
2008), ii) bedding planes are difcult to recognize due to their
homogeneity, though planes can be recognized, albeit weakly by
electrical micro-imaging (Ishii et al., 2006), iii) they are non-
calcareous, diatomaceous siliceous rocks but relatively argillaceous
(Iijima and Tada, 1981; Tada and Iijima, 1982) and have Al
2
O
3
content of 811% (Ishii et al., 2007; Hiraga and Ishii, 2008), iv)
mineral compositions consist of silica minerals (opal-A and/or opal-
CT) and small amounts of quartz, feldspar, clay minerals (kaolinite,
smectite, illite and/or chlorite), pyrite, and carbonates (siderite and/
or magnesite) (Ishii et al., 2007), and v) they resemble siliceous
shale units of the Monterey Formation of California (e.g., Murata
and Nakata, 1974; Isaacs, 1982) in rock type. This study focused on
the siliceous mudstone composing the Wakkanai Formation, which
was formed by induration of the diatomaceous mudstone
composing the Koetoi Formation during progressive burial
diagenesis of silica minerals, i.e., conversion of opal-A to opal-CT
(Fukusawa, 1985). Furthermore, stiff siliceous mudstones, termed
opaline chert, are observed in the upper part of the Wakkanai
Formation at the surface (Fukusawa, 1985). It was suggested by
Iijima and Tada (1981) that the opaline cherts formed in the upper
part of the opal-CT zone (near the opal-A zone) by additional silica
cementation during uplift of the Neogene noncalcareous siliceous
rocks in northern Japan. The opaline cherts are inter-layered with
siliceous mudstones in the upper part of the Wakkanai Formation
at the surface (e.g., Fig. 4), whereas, in core samples, such opaline
Fig. 3. Permeability of the Wakkanai Formation obtained by hydraulic packer tests in
twelve vertical boreholes (HDB-1 to HDB-11 and PB-V01). Data are from Kurikami
(2007) and Yabuuchi et al. (2008). Compared to the hydraulic transmissivity of similar
depths, higher permeability sections tend to correspond to fractured sections. The
difference in hydraulic transmissivity between a fractured section and an intact section
at similar depths may be ve orders of magnitude.
Fig. 4. Photograph of the horizontal bedrock exposure made by stripping with heavy equipment (modied Ishii and Fukushima, 2006). The opaline chert layer parallel to bedding
plane (strike/dip: N10
W/30
E) is displaced by the major FCBs (strike/dip: N40
56
W/54
90
N) in the center of the outcrop.

cherts are rarely observed and the harder siliceous mudstones are
common in the upper part of the Wakkanai Formation.
In the center of the Horonobe area, a map-scale fault, the
Omagari Fault, strikes NNWSSE, generally subparallel to major
folds developed in the area. The Omagari Fault is an oblique fault
with reverse dip-slip and sinistral strike-slip components, which
exhibits an eastwest step-over at surface that possibly converges
in the subsurface (Ishii et al., 2006). A previous study on burial
diagenesis in this area indicates that the Wakkanai Formation was
buried to at least a depth of 1 km. Flexural folding of the sediments
began between 2.2 and 1.0 Ma. The study area was exposed at
approximately 1.0 Ma, at the latest (Ishii et al., 2008).
3. Methodology
3.1. Geological characterization of fractures
Fracture mapping of an exposure of the Wakkanai Formation
bedrock (Loc. A in Fig. 1) and fracture logging of four boreholes
(HDB-6, HDB-9, HDB-11 and PB-V01; Fig. 1) were carried out for the
geological characterization of fractures. Fracture mapping of hori-
zontal exposures of bedrock is considered important for geomet-
rical analysis of strike-slip faults such as the FCBs. However,
because there are few bedrock exposures in this area due to the
abundance of vegetation and overburden, a horizontal exposure,
32.5 m25.5 m (Fig. 4), was cleared with heavy equipment for
fracture mapping (Ishii and Fukushima, 2006). Data on the
following fracture characteristics were collected: fracture orienta-
tions, lengths, widths and termination styles, lling materials,
failure modes, sense of displacement, and crosscutting relation-
ships. Stress directions for displacement along the oblique faults
were determined by the multiple inverse method of Yamaji (2000).
In the borehole investigations, the following data on natural frac-
tures in the core were collected: depth of intersections, frequencies,
fracture orientations and widths, lling materials, failure modes,
and sense of displacement. Fractures induced by drilling or
handling were excluded following criteria established by Kulander
et al. (1990). Fracture orientations were also determined by
detailed correlation between the core data and the acoustic
imagery data obtained by the borehole televiewer survey (BHTV).
Concerning BHTV, even if a borehole wall is composed of dark-
colored rock such as siliceous mudstone and/or is coated with
drilling mud, or even if a borehole is lled by cloudy borehole
water, BHTV can provide useful images of the borehole wall (Wil-
liams and Johnson, 2004). Furthermore, although BHTV detects
fewer fractures than those observed in cores owing to its spatial
resolution, BHTV is useful for determining the distribution of major
fracture orientations (Genter et al., 1997).
The following protocols and denitions were used in this study:
the determination of displacement using slickensteps (secondary
fractures) and the classication of fault rocks are after the criteria of
Petit (1987) and Sibson (1977), respectively. Also, faults with slip of
less than 45
rake measured in the plane of the fault are dened as

predominantly strike-slip faults, while, those faults at more than
45
rake, are predominantly dip-slip faults. No sense of slip

direction of the hanging wall, either up dip or down, is implied.
FCBs composed of fault planes with associated fault rocks such as
breccia are dened as major FCBs, whereas FCBs comprising fault
planes with associated slickenlines and/or slickensteps (secondary
fractures) but not fault rocks such as breccia are dened as minor
FCBs. FCBs include major FCBs and minor FCBs. Moreover, frac-
tures with natural plumose structure are dened as tensile frac-
tures. Sample core photographs are shown in Fig. 5.
3.2. Mechanical characterization of host rock
Intact siliceous mudstone cores from the Wakkanai Formation
were used for the rock mechanics studies. Tensile strength and
internal friction angles for theoretical analyses using the Grifth
Coulomb criterion were obtained by Brazilian tests, after the
Japanese industrial standards (JIS M 0303; Japanese Industrial
Standard Committee, 2000), and triaxial compression tests after
Kovari et al. (1983), respectively.
4. Results
Major FCBs, minor FCBs, tensile fractures, and bedding faults
such as occur in the cleared, horizontal outcrop were intersected by
the boreholes. The fault rocks associated with the major FCBs are
mostly fault breccias (thickness: <30 cm), and seldomoccur as fault
gouge (thickness: <3 mm). The tensile fractures are outcrop-scale
fractures which crosscut the bedding planes at high angles similar
to those of the FCBs. Although splay cracks propagating from major
FCBs were observed in the outcrop, they are classied as either
tensile fractures or minor FCBs.
4.1. Fracture mapping of the horizontal outcrop
The strikes and sense of displacement along FCBs in the bedrock
exposure are shown in Fig. 6a. The data on long, well linked faults
were collected at many points along strike. The strikes of the FCBs
are WNWESE, oblique to the direction of the fold axes, as is
evident in Fig. 1 (i.e. N10
W). The sense of displacement is sinistral,

predominantly strike-slip.
Crosscutting relationships between major FCBs and bedding
faults show that the former always displace the latter (Fig. 7),
generally with less than 0.6 m offset. Layered opaline chert is also
present in this outcrop, but whether or not major FCBs displace the
layered opaline chert is not clear because the layered opaline chert
boundaries are so indistinct (Fig. 4) that with fault displacements of
generally less than 0.6 m, any offset is difcult to recognize. Only
the major FCBs in the center of the outcrop exhibit large enough
Fig. 5. Examples of fractures in core. (a) major FCB (370.8 m depth in HDB-6), (b) minor FCB (277.1 m depth in HDB-6), and (c) tensile fracture (377.1 m depth in PB-V01).
offsets, (i.e. about 10 m) to determine that there has indeed been
displacement of the layered opaline chert (Fig. 4).
In the outcrop, many splay cracks propagate from major FCBs.
The splays are often in densely fractured zones at the major FCB tips
or within the major FCB linkages, where the fault appears to have
a left-step. The splay cracks strike 1050
counterclockwise from
the fault planes and have lengths of decimeters to meters, whereas
the lengths of the linked fault segments are from meters to deca-
meters (Figs. 8 and 9). Most splay cracks are tensile fractures and do
not show any evidence of shearing although some splay cracks,
particularly those within fault linkages, show evidence of shearing.
These shear splay cracks are thought to be sheared tensile fractures
which have originated as tensile fractures, because the orientation
of the shear splay cracks is the same as those of the adjacent tensile
splay cracks.
In addition, the major FCBs were shown to have slipped under
a horizontal stress state with s1 normal to the fold axis and s3
parallel to the fold axis, determined by the multiple inverse method
using the fault displacement data obtained at this outcrop (Ishii and
Fukushima, 2006). Under this stress regime the above orientation
of the evolving splay cracks is in agreement with those generally
predicted by the numerical simulations of Lunn et al. (2008).
4.2. Fracture logging of the boreholes
Siliceous mudstones of the Wakkanai Formation were inter-
sected by borehole HDB-6 from 262 m to 620 m depth; by HDB-9
from 0 m to 520 m depth; by HDB-11 from 460 m to 1020 m depth
and by PB-V01 from 237 m to 520 m depth. Fractures in the core
from these sections were examined. At shallower depths in bore-
holes HDB-6, HDB-11 and PB-V01, the diatomaceous mudstones of
the Koetoi Formation were intersected. FCBs were commonly
intersected by the boreholes; examination of recovered core indi-
cated that for every 10 m long section of core there were between
0 and 44 FCBs with an average of 9.4 FCBs over 10 m long sections
(frequency: 044 per 10 m, avg. 9.4 per 10 m). Additionally, 44% of
the FCB population were oriented by detailed correlation between
core and BHTV imagery. Moreover, 14% of the population of FCBs
observed in core are major FCBs (frequency: 09 per 10 m, avg. 1.3
per 10 m), and 23% of the major FCB population could be oriented
by detailed correlation with the BHTV imagery. Strikes of all the
oriented FCBs are predominantly EW, oblique to the fold axes (i.e.
N40
W: Fig. 1) though the FCBs with NNESWS-strikes, oblique to

the fold axes as well as EW-strikes, are also recognized in HDB-11
(Fig. 6d), and the sense of displacement is predominantly strike-slip
in each borehole (Fig. 6be). These observations are very similar to
those on the FCBs in the horizontal outcrop. For those oriented
major FCBs, their sense of displacement is predominantly strike-
Fig. 6. Histograms showing the distribution of fault strikes and sense of displacement
of FCBs. Upper and lower histograms in (a)(e) showthe data on major and minor FCBs
and only on major FCBs, respectively. However, in the boreholes, only the data on the
oriented faults determined by detailed correlation between core and BHTV imagery are
shown. Gray and white bars indicate predominantly strike-slip and predominantly
dip-slip, respectively.
Fig. 7. Photograph showing the crosscutting relationship between a younger, major
FCB and a bedding fault. Note the sinistral, predominantly strike-slip shear along the
major FCB displacing the bedding fault with an offset of about 0.6 m.
slip (Fig. 6be), similar to the major FCBs in the outcrop. There is no
systematic change in strike nor sense of displacement with depth.
In core from relatively great depths (e.g. below 400 m), en
echelon shear fractures (minor FCBs) were sometimes observed to
develop in extensions several centimeters long at and beyond the
fault tips in directions parallel to the existing minor FCBs, although
it should be remembered that the core diameter is only 86 mm.
Thus at depth, shear fractures appear to propagate fromfault tips in
directions subparallel to the existing fault planes and parallel shear
fractures formbeyond the fault tips. Sense of displacement of those
shear fractures is the same as for the existing minor FCBs. Examples
are shown in Fig. 10. En echelon shear fractures occur more often at
greater depths, e.g., in PB-V01, the average frequency is 1.0 per
100 m (1s 1.4) above 450 m, whereas, at greater depth, it is 6.0
per 100 m (one interval data). Similar trends are also observed in
the other boreholes; 0.5 per 100 m (1s 0.7) above 400 m vs 1.5
per 100 m (1s 0.7) below 400 m in HDB-6; 1.0 per 100 m
(1s 1.7) above 300 m vs 5.0 per 100 m (1s 1.4) below 300 m in
HDB-9; and 2.0 per 100 m (1s 1.9) below 500 m in HDB-11. In
addition, tensile fractures are also observed in core (frequency: 0
16 per 10 m, avg. 2.3 per 10 m, 1s 3.3). Frequencies of the tensile
fractures decrease with depth, e.g., in PB-V01, the average
frequency is 3.3 per 10 m (1s 3.6) above 450 m, while, at greater
depth, it is 1.6 per 10 m (1s 2.7). Similarities are found in the
other boreholes; 3.2 per 10 m (1s 4.0) above 400 m vs 1.9 per
10 m (1s 3.7) below 400 m in HDB-6; 3.2 per 10 m (1s 4.1)
above 300 mvs 2.0 per 10 m (1s 2.3) below300 m in HDB-9; and
0.8 per 10 m (1s 1.5) below 500 m in HDB-11. Thus, en echelon
shear fractures increase in number as depth increases, while
tension fractures decrease in number.
Fig. 8. (Left side) Sketch of fractures in the horizontal bedrock exposure except for bedding faults (modied from Ishii and Fukushima, 2006), (a) photograph of splay cracks
propagating from a major FCB and (b) photograph of splay cracks (linkage fractures) within a major FCB linkage. In the left gure, thick lines represent major FCBs, while thin lines
are minor FCBs and tensile fractures such as splay cracks. The sense of displacement of FCBs is sinistral, predominantly strike-slip. Note the left gure area corresponds to the
clockwise at a right angle of Fig. 4.
Fig. 9. Detailed sketches of the high density of fractures in the fault linkage zones shown in Fig. 8, where many splay cracks propagate from major FCBs (modied Ishii and
Fukushima, 2006). Thick lines represent major FCBs and thin lines represent tensile fractures and minor FCBs. The splay cracks appear to have grown as internal tensile fractures,
although some subsequently accommodated shear displacements. Eventually, their fractures form fault zones.
4.3. Rock mechanics tests: siliceous mudstone
Tensile strengths of 0.8 0.1 MPa to 3.2 0.3 MPa were
obtained by Brazilian tests on siliceous mudstone core in 30100 m
intervals in each borehole. The average of the tensile strengths is
1.5 MPa (1s 0.6), though some tensile strengths exceeding 2 MPa
were obtained (Fig. 11a). The core samples showing high tensile
strength are from 303 m and 348 m depth in HDB-6 and from
272 m, 303 m and 402 m depth in PB-V01. These stratigraphic
positions are in the upper part of the Wakkanai Formation. The fact
that such harder siliceous mudstones are observed in the upper
part of the Wakkanai Formation is consistent with the previous
indication that additional silica cementation, which hardened sili-
ceous mudstones, may often occur in the upper part of the opal-CT
zone (near the opal-A zone) in Neogene noncalcareous siliceous
rocks in northern Japan (Iijima and Tada, 1981).
Internal friction angles obtained by triaxial compression testing
of cores from 300 m and 550 m depth in HDB-6 have values of
27.3 1.3
and 27.5 1.0
respectively, (avg. 27.4
), though tensile
strengths on core from nearly the same depths are 3.1 0.2 MPa
and 0.9 0.1 MPa respectively. Internal friction angles were also
obtained from ten core samples of the Wakkanai Formation in the
other boreholes (HDB-1 and HDB-2) (Niunoya and Matsui, 2005),
and the average angle of internal friction for the twelve core
samples from these boreholes (HDB-1, HDB-2 and HDB-6) is
26.1 2.9
(Fig. 11b).
5. Discussion
5.1. Relative age of formation of the major FCBs
The relative timing of fault formation can be inferred from their
crosscutting relationships with other structures. The major FCBs
displace bedding faults without exception. The bedding faults are
considered to have formed synchronously with exural folding
(Ishii and Fukushima, 2006). Therefore, the later major FCBs are
thought to have developed during and/or after uplift and erosion
following folding. This deduction is also supported by observations
from the horizontal outcrop where a major FCB displaces layered
opaline chert, which is considered by Iijima and Tada (1981) to have
formed during uplift and erosion. This implies the major FCBs were
growing during and/or after uplift and erosion.
5.2. Stress state suitable for propagation of splay cracks in the
Wakkanai Formation
What stress state would lead to the development of the abun-
dant splay cracks that are seen to propagate from the major FCBs?
Fig. 10. En echelon shear fractures (minor FCBs) evolving at and beyond the minor FCB tips, FCB1 and FCB2 tips in this gure, at much lower angles to the existing fault planes
(a core sample from 469.6 m depth in PB-V01). (a) an overview shot of the core sample (b) occurrence at and beyond the FCB1 tip, (c) hanging wall surface of the FCB1, (d)
occurrence at and beyond the FCB2 tip, and (e) footwall surfaces of the shear fractures developing at and beyond the FCB1 tip. Splay fracturing was not observed near the FCB1 and
FCB2 tips. Sense of displacement of the FCB1, FCB2 and shear fractures are the same; sinistral, predominantly strike-slip, although slickenlines and slickensteps of the shear fractures
indicate only very weak shearing.
Most splay cracks propagating froma fault can be formed as tensile
fractures, particularly at the fault tips, by concentration of tensile
stress generated when a slip patch nucleates and propagates in
a fault (Martel and Pollard, 1989). But such tensile failure does not
necessarily occur at the fault tips under all initial stress states.
Previous numerical simulations suggest that shear failure can also
occur at the fault tips following fault slip, resulting in propagation
of shear fractures subparallel to the original fault plane or forma-
tion of shear fractures that propagated back into the compressive
quadrants (Shen and Stephansson, 1993; Bourne and Willemse,
2001; Willson et al., 2007; Lunn et al., 2008). Although an extension
of the initial fault in its own plane also has been previously simu-
lated by Du and Aydin (1993, 1995), tensile failure is suppressed in
their conceptual model of mechanical failure.
Following Bourne and Willemse (2001), the distance between
any prevailing stress state in a rock mass to either the tensile or the
shear failure conditions can be described by a stress quantity, X on
a Mohr diagram. In Fig. 12, X corresponds to the shortest distance
from the GrifthCoulomb failure envelope of rock strength to the
Mohr circle, the initial stress state. Brittle failure occurs when
the Mohr circle stresses reach the failure envelope, i.e. X 0, and
the mode of failure is determined by whichever failure condition is
met; either X
shear
(X
s
) 0 or X
tensile
(X
t
) 0. Furthermore, reduction
of effective normal stress and increase of shear stress near fault tips,
that are produced by fault slip (e.g., Martel and Pollard, 1989;
Bourne and Willemse, 2001), are important factors causing brittle
failure near fault tips.
When fault slip occurs under an initial stress state where X
s
is
smaller than X
t
(Fig. 13a, top), two cases can be assumed for failure
induced by reduction of effective normal stress and/or increase of
shear stress near the fault tip (Fig. 13b). In the rst case, shear
failure occurs before tensile failure can occur (Fig. 13b, top), and, at
least in the case of the Wakkanai Formation, en echelon shear
fractures can develop at and beyond the fault tips in the plane of
and thus parallel to the existing faults as shown in Fig. 10 and
Fig. 13c, top. Shear fractures that propagate back into the
compressive quadrants were not observed in the outcrop. Eventu-
ally, the en echelon shear fractures would coalesce into a fault
represented by fault rocks such as a major FCB. In the other case,
tensile failure occurs before shear failure can occur (Fig. 13b,
bottom) and splay cracks propagate from the fault tips by tensile
failure (Fig. 13c, bottom).
However, when fault slip occurs in the initial stress state where
X
s
is larger than X
t
(Fig. 13a, bottom), tensile failure will certainly
occur before shear failure due to reduction of effective normal
stress and increase of shear stress near the fault tip (Fig. 13b,
bottom), and splay cracks will propagate from the fault tips by
tensile failure. Therefore, the suitable stress state for propagation of
the splay cracks in the Wakkanai Formation is likely to be the case
where X
s
X
t
is positive, that is when X
s
>X
t
.
The above value of X
s
X
t
is written by the GrifthCoulomb
criterion as:
X
s
X
t

2S
T
cos f
i
0:5
1 sin f
i
s
r
m
where S
T
, f
i
and s
m
r
are the tensile strength, the angle of internal
friction, and the remote mean stress, respectively (Fig. 12). This
formula means that failure modes which occur at fault tips by the
fault slip depend on the tensile strength, the angle of internal
friction, and the remote mean stress at the time of fault slip, and, in
the case of rocks which do not signicantly vary in both the tensile
strength and the angle of internal friction, the lower remote mean
stress is suitable for propagation of tensile splay cracks. This
formula also means that higher values of the tensile strength
promote tensile failure near fault tips when fault slip occurs while
the lower values tend to induce shear failure, provided the remote
mean stresses are similar (and the angles of internal friction are
also). This can be consistent with the relationship between rock
strengths and failure modes suggested by eld observations of
siliceous shale units of the Monterey Formation of coastal California
(Gross, 1995; Dholakia et al., 1998; Eichhubl and Boles, 2000).
5.3. Estimation of X
s
X
t
distributions in the boreholes
In this study, the distributions of X
s
X
t
in the Wakkanai
Formation in each borehole were estimated using the above
formula, with tensile strengths and angles of internal friction
determined by the laboratory testing and the assumed remote
mean stresses. Tensile strengths were set as the strengths obtained
by Brazilian tests (0.83.2 MPa) at depths in 50100 m intervals in
each borehole (Fig. 14). Although the internal friction angles
Fig. 11. Histograms on rock mechanical properties of the Wakkanai Formation
measured by laboratory tests. (a) tensile strengths of core samples from HDB-6, HDB-9,
HDB-11 and PB-V01 examined in this study. (b) internal friction angles of core samples
from HDB-6 and boreholes HDB-1 and HDB-2 (Niunoya and Matsui, 2005).
Fig. 12. Proximity of stress state to brittle failure conditions is represented by the
smallest stress increment required to reach that stress state from either the shear
failure envelope, X
s
, or the tensile failure envelope, X
t
: see Bourne and Willemse
(2001); the cohesive strength, s
0
, is assumed to be equal to twice the tensile strength,
S
T
, following Brace (1960).
obtained by the triaxial compression tests in the boreholes are only
from two locations in HDB-6, the average angle of internal friction
for the twelve core samples including core samples from the other
boreholes (HDB-1 and HDB-2) is 26.1 2.9
(Fig. 11b). In this study,

considering variations in the internal friction angles, 20.3 and 31.9
,
which are lower and higher than the average degree by 2s (5.8),
were represented as the lowangle case (Case 1) and the high angle
case (Case 2) respectively (Fig. 14). For the remote mean stresses,
the following was assumed: i) the principal remote stresses are
horizontal and vertical based on the results of the multiple inverse
Fig. 13. Modes of brittle failure near the fault tip induced by reduction of effective normal stress and/or increase of shear stress that are produced by fault slip in the Wakkanai
Formation. The failure modes are determined by the initial stress state. See Fig. 12 for the envelopes and circles.
Fig. 14. The estimated depth proles of X
s
X
t
estimated in each borehole. Results in Case 1 and Case 2 are based on a low friction angle (20.3
) and a high internal friction angle

(31.9
) respectively, and the results are almost the same. The lithostatic load estimated by density logging and the tensile strengths determined by laboratory testing are also shown.
analyses using the fault slip data sampled in the horizontal outcrop,
ii) the vertical stress is lithostatic, iii) the pore pressure is hydro-
static since the present pore pressures of the sections examined in
the borehole are nearly under hydrostatic state (Kurikami, 2007;
Funaki et al., 2009) and further the elevated pore pressure, which
could have been raised during the early stage of folding following
burial diagenesis (e.g., Engelder, 1985), are inferred to be nearly
released by early fracturing predating the major FCBs development,
and iv) the remote mean stress is a vertical effective stress as the
sense of displacement of the major FCBs is predominantly strike-
slip (Fig. 6). The lithostatic load was calculated using data obtained
by density logging in each borehole (Fig. 14). Although the esti-
mation based on the above assumptions is accompanied by some
errors, the distribution of approximate X
s
X
t
could be determined.
The estimated depth proles of X
s
X
t
in Case 1 and Case 2 are
shown in Fig. 14. In this study, the domain where X
s
X
t
is positive
is dened as SDT (Suitable Domain for pervasive Tensile failure),
while, the domain where X
s
X
t
is negative is dened as UDT
(Unsuitable Domain for pervasive Tensile failure). The results using
upper and lower values for internal friction angle are almost the
same (Fig. 14). The SDT domains are estimated to be above 400 min
HDB-6, above 300 m in HDB-9, and above 450 m in PB-V01, while
the UDT domains were estimated to occur below these depths in
HDB-6, in HDB-9, and in PB-V01, respectively and below 500 m in
HDB-11 (the depth of the SDT domain in HDB-11 is unknown, but is
no deeper than 500 m). Thus, the SDT and UDT domains roughly
correspond to the rock mass above and below 400 m depth in each
borehole, respectively, if we take into consideration estimation
errors.
5.4. Comparison between the distribution of the SDT and UDT and
fractures in core
The presence of many splay cracks along a fault is an indication
of numerous slip events along a fault (Martel and Pollard, 1989).
Similarly, a fault associated with fault rocks such as breccia is likely
formed by several slip events. Hence, if tensile splay cracks are
easily developed by fault slip in an SDT, the number of tensile
fractures near major FCBs is expected to be higher in an SDT. Fig. 15
shows the frequencies (number per 10 m) of major FCBs and of
tensile fractures, the ratios of the frequencies of tensile fractures to
major FCBs, and the distributions, in terms of depth and bound-
aries, of the SDTand UDT in each borehole. Comparison of the ratios
of tensile fracture to major FCB frequencies and the SDT/UDT
Fig. 15. Frequency distributions of brittle fractures in the boreholes in the Wakkanai Formation. Frequency is expressed as a number per 10 m for FCBs, major FCBs, and tensile
fractures in each borehole. Also, tensile strengths, histograms showing the ratios of tensile fracture frequency to major FCB frequency, and the distribution of the SDT and UDT in
each borehole are provided.
distributions, indicate that the higher ratios (e.g. 5) are restricted,
with one exception in HDB-9, to the SDT domain. The frequencies of
tensile fractures also are higher in the SDT than the UDT. This
evidence supports the above hypothesis. Although a considerable
number of tensile fractures are observed also in the range of 470
480 m depth in HDB-6 in the UDT, it could be the result of the
following: i) mechanical effects of slip on a considerable number of
FCBs observed at 440460 m depth (since tensile failure also can
occur by fault slip even if within a UDT as shown in Fig. 13), or ii)
error in estimation of the location of the SDT/UDT boundary
(perhaps several tens meters) due to the assumptions relevant to
the remote mean stresses at the time of fault slip.
Furthermore, the ratios of tensile fracture frequency to major
FCB frequency are near zero throughout the dened UDT, though
the frequency of major FCBs in the UDT (09 per 10 m; avg. 1.3 per
10 m) is similar to that in the SDT (09 per 10 m; avg. 1.4 per 10 m).
This observation implies that the principal mode of failure in a UDT
is shear. This implication is also supported by the fact that the
occurrences of en echelon shear fractures developing at and
beyond minor FCB tips in cores, i.e., those not associated with
splays, were observed more often in the UDT than the SDT. This is
based on both the frequency distribution of en echelon shear
fractures mentioned in Section 4.2, and the SDT and UDT domains
dened in Section 5.3.
Depths above 400 m in HDB-6 and the depths above 450 m in
PB-V01 where tensile strengths are higher appear to correspond to
the zones of high ratios of tensile fracture to major FCB frequencies
and high frequencies of tensile fractures. But the zones with high
ratios of tensile fracture to major FCB frequencies and high
frequencies of tensile fractures are also recognized above 300 m
depth in HDB-9, where tensile strengths are not higher. This shows
that the failure mode near fault tips during faulting in the Wak-
kanai Formation, depends not only on rock strength, but also the
remote mean stresses.
5.5. Growth model of the major FCBs
Concerning fault growth models in brittle sedimentary rocks,
models in sandstones have been developed in previous studies.
Two main ways that faults grow in sandstones are recognized
(Davatzes et al., 2003; Flodin and Aydin, 2004): i) deformation
band-based faulting; and ii) sheared-joint-based faulting. In the
rst model, faults grow by localization and amalgamation of indi-
vidual deformation bands to form deformation band zones with
subsequent coalescence of the zones and discontinuous slip
surfaces nucleated along deformation bands to form through going
deformation band-style faults (e.g. Aydin and Johnson, 1978;
Antonellini and Aydin, 1995; Shipton and Cowie, 2001). In the
second model, faults grow by linking with neighboring faults via
linkage structures such as splay cracks, which formed near the fault
tips by stress concentrations following slip nucleation on preex-
isting structures (e.g., Flodin and Aydin, 2004; Myers and Aydin,
2004). Although fault growth models in mudstones are seldom
known, Dholakia et al. (1998) suggested that faults grew by linking
via splay cracks produced by shearing along initial joints, i.e.
the sheared-joint-based faulting model, in siliceous shale units of
the Monterey Formation. In the case of the siliceous mudstone, the
Wakkanai Formation, it is inferred by this study that, during and/or
after uplift and erosion the major FCBs grewmainly by linking with
adjacent faults via numerous splay cracks, which formed by tensile
failure as shown in Figs. 8, 9 and Fig. 13c-bottom, in an SDT (i.e.
above roughly 400 m depth). Such growth mechanism in an SDT
domain is similar to the sheared-joint-based faulting model. In
contrast, in a UDT (i.e. belowroughly 400 m depth), the major FCBs
grew predominantly by shear failure, and could develop in direc-
tions parallel to the fault planes through coalescence of the en
echelon shear fractures as shown by Fig. 10, Fig. 13c, top and Fig. 16
though the developing direction would not necessarily be limited
to the above directions. A fault growth model in mudstones like the
mechanism in a UDT domain has not been reported, however, it
may be similar to the idealized fault growth model suggested by
Cowie and Scholz (1992).
5.6. Comparison between SDT and UDT and permeability of the
Wakkanai Formation
Based on the above, the linking of adjacent, major FCBs via
numerous tensile splay cracks is assumed to develop in an SDT.
Previous studies pointed out that such structures associated with
tensile fractures have important hydraulic properties. Eichhubl
et al. (2009) showed by using the distribution of fault-related
calcite cement at the Moab fault system in southern Utah as an
indicator of paleouid migration, that uid ow was focused
especially at locations of structural complexity such as fault inter-
sections, extensional steps, and fault-segment terminations, where
many tensile fractures developed. Dholakia et al. (1998) indicated
that increased hydrocarbon concentrations in the Monterey
Formation in the southern San Joaquin Valley and coastal California
are almost exclusively associated with faults which grew through
linkage via many tensile splay cracks. Khang et al. (2004) indicated
by numerical simulations that geometry of the structures linking
faults via tensile splay cracks plays an important role in uid ow
through the fracture network. Mazurek et al. (1998, 2003) and Lunn
et al. (2008) also imply that linkage by structures such as tensile
Fig. 16. A growth model of the major FCBs in the Wakkanai Formation during and/or
after uplift and erosion. When fault slip occurs in an SDT domain, splay cracks form
near the fault tips by tensile failure, resulting in faults strongly interconnected via
numerous tensile splay cracks. In contrast, when fault slip is in a UDT domain, shear
failure is predominant near fault tips and en echelon shear fractures could form at and
beyond the fault tips.
splay cracks may represent preferential pathways for uid ow. In
consideration of the work of these authors, comparisons between
the distributions of SDT and UDT domains in each borehole and
variations in permeability in the Wakkanai Formation were done in
this study.
Fig. 17 shows the results of the comparisons in the each bore-
hole. The SDT domains generally have higher permeability (e.g.
hydraulic transmissivity: 10
5
m
2
/s) whereas the UDT domains
clearly show lower permeability (e.g. hydraulic transmissivity:
10
5
m
2
/s). However, some lower permeability sections are also
observed in the SDT domains in HDB-9 and PB-V01. These sections
have relatively lower frequencies of total number of FCBs and
tensile fractures and thus are not active in the main connected ow
paths in an SDT. This observation suggests that the hydrogeological
environment in an SDT differs from that in a UDT due to hydraulic
effects of the faults strongly interconnected via many tensile splay
cracks, and that these structures result in the restricted distribution
of the highly permeable sections to depths less than about 400 min
the Wakkanai Formation, as shown in Figs. 3 and 17.
Diagenetic alteration products can also provide some indication
of permeability of fractured rock masses (e.g., Milodowski et al.,
1998; Solumet al., 2005; Eichhubl et al., 2009). But in the Wakkanai
Formation, neither mineral precipitation on fractures nor alteration
along fractures has been observed in core samples though oxidized
zones along the major FCBs, which imply high permeability, are
observed at surface (Ishii and Fukushima, 2006). Moreover,
although the harder siliceous mudstones, which might be formed
due to additional silica cementation during uplift as suggested by
Iijima and Tada (1981), are observed in the subsurface, a relation-
ship between the harder siliceous mudstones and permeability of
the Wakkanai Formation has not been found. If the issue of whether
or not some diagenetic indicator for uid ow exists in the Wak-
kanai Formation is validated, further detailed X-ray analyses such
as SEM, XRD and XRF must be done on rock matrices and fracture
surfaces.
6. Conclusions
This paper discusses the growth mechanisms of the major FCBs
in the Wakkanai Formation by the following methods: i) geological
characterization by fracture mapping at an outcrop and by fracture
logging in several boreholes; ii) rock mechanical characterization
by laboratory tests on core samples for tensile strength and angle of
Fig. 17. Comparison of the SDT and UDT distributions and permeability in the Wakkanai Formation. Also, tensile strengths, the frequencies of total number of FCBs and tensile
fractures and the ratios of tensile fracture frequency to major FCB frequency in each borehole are shown.
internal friction, and iii) theoretical analysis using the Grifth
Coulomb criterion. This paper suggests the following ideas:
1. The principal mode of failure, which occurs near fault tips by
fault slip in the Wakkanai Formation, depends not only on rock
strength, but also on the remote mean stresses.
2. During and/or after uplift and erosion the major FCBs grew
mainly by linking with adjacent major FCBs via numerous splay
cracks formed by tensile failure in an SDT (i.e. above roughly
400 mdepth), while, in a UDT (i.e. belowroughly 400 mdepth),
the major FCBs predominantly grow by shear failure, and could
develop in directions parallel to the fault planes through coa-
lescence of en echelon shear fractures though the direction of
development is not necessarily limited to the above directions.
3. The hydrogeological environment in an SDT differs fromthat in
a UDT due to hydraulic effects of the faults strongly inter-
connected via many tensile splay cracks. The preferential
development of the linking structures results in the restricted
distribution of the high permeability sections to less than about
400 m depth in the Wakkanai Formation.
Such conclusions are useful also for three dimensional geolog-
ical modeling and for groundwater ow simulations (e.g., for the
geological disposal of high level radioactive waste). However, in
order to conduct the more detailed modeling and simulations for
the Wakkanai Formation, it may be necessary to solve unanswered
questions, such as the comprehensive deformation history, and the
detailed hydraulic effects of the faults strongly interconnected via
many tensile splay cracks.
Acknowledgements
We thank G. McCrank and W.R. Alexander for English editing
and their helpful suggestions on this manuscript. We also express
our gratitude to H. Moir, J.G. Solum and Z. Shipton for their critical
review of the manuscript.
Appendix
The acronyms and their denition used in this study are as
shown in Table A1.
References
Andersson, J., Ahokas, H., Hudson, J.A., Koskinen, L., Luukkonen, A., Lo fman, J., Keto, V.,
Pitka nen, P., Mattila, J., Ikonen, A.T.K., Yla -Mella, M., 2007. Olkiluoto Site
Description 2006. POSIVA Technical Report POSIVA 2007-03, Olkiluoto, Finland.
Antonellini, M., Aydin, A., 1995. Effect of faulting on uid ow in porous sandstone:
geometry and spatial distribution. American Association of Petroleum Geologist
Aydin, A., Johnson, A.M., 1978. Development of faults as zones of deformation
bands and as slip surfaces in sandstone. Pure and Applied Geophysics 116,
931942.
Bossart, P., Hermanson, J., Mazurek, M., 2001. Analysis of fracture network based on
the integration of structural and hydrogeological observations on different
scales. SKB Technical Report 01-21, SKB, Stockholm, Sweden.
Bourne, S.J., Willemse, E.J.M., 2001. Elastic stress control on the pattern of tensile
fracturing around a small fault network at Nash Point, UK. Journal of Structural
Geology 23, 17531770.
Brace, W.F., 1960. An extension of the Grifth theory of fracture to rocks. Journal of
Cowie, P.A., Scholz, C.H., 1992. Physical explanation for displacement-length rela-
tionship of faults using a post-yield fracture mechanics model. Journal of
dAlessio, M.A., Martel, S.J., 2004. Fault terminations and barriers to fault growth.
during the development of multiple fault sets, Chimney Rock fault array, Utah,
USA. Tectonophysics 363, 118.
Dholakia, S.K., Aydin, A., Pollard, D.D., Zoback, M.D., 1998. Fault-controlled hydro-
carbon pathways in the Monterey Formation, California. American Association
of Petroleum Geologist Bulletin 82, 15511574.
Du, Y., Aydin, A., 1993. The maximum distortional strain energy density criterion for
shear fracture propagation with applications to the growth paths of en e chelon
Du, Y., Aydin, A., 1995. Shear fracture patterns and connectivity at geometric
complexities along strike-slip faults. Journal of Geophysical Research 100,
1809318102.
Eichhubl, P., Boles, J.R., 2000. Focused uid ow along faults in the Monterey
Formation, coastal California. Geological Society of America Bulletin 112,
16671679.
Eichhubl, P., Davatzes, N.C., Becker, S.P., 2009. Structural and diagenetic control of
uid migration and cementation along the Moab fault, Utah. American Asso-
ciation of Petroleum Geologist Bulletin 93, 653681.
Engelder, T., 1985. Loading paths to joint propagation during a tectonic cycle: an
example from the Appalachian Plateau, U.S.A. Journal of Structural Geology 7,
459476.
Evans, J.P., Forster, C.B., Goddard, J.V., 1997. Permeability of fault-related rocks, and
implications for hydraulic structure of fault zones. Journal of Structural Geology
19, 13931404.
Flodin, E.A., Aydin, A., 2004. Evolution of a strike-slip fault network, Valley of Fire
State Park, southern Nevada. Geological Society of America Bulletin 116 (1/2),
4259.
Fukusawa, H., 1985. Late Neogene Formations in the Tempoku-Haboro region,
Hokkaido, Japan stratigraphic reinvestigation of the Wakkanai and Koetoi
Formations. The Journal of the Geological Society of Japan 91, 833849.
Funaki, H., Ishii, E., Tokiwa, T., 2009. Evaluation of the role of fracture as the major
water-conducting feature in Neogene sedimentary rocks. Journal of the Japan
Society of Engineering Geology 50, 239248.
Genter, A., Castaing, C., Dezayes, C., Tenzer, H., Traineau, H., Villemin, T., 1997.
Comparative analysis of direct (core) and indirect (borehole imaging tools)
collection of fracture data in the Hot Dry Rock Soultz reservoir (France). Journal
of Geophysical Research 102, 1541915431.
Gross, M.R., 1995. Fracture partitioning: failure mode as a function of lithology in
the Monterey Formation of coastal California. Geological Society of America
Bulletin 107, 779792.
Gutmanis, J.C., Lanyon, G.W., Wynn, T.J., Watson, C.R., 1998. Fluid ow in faults:
a study of fault hydrogeology in Triassic sandstone and Ordovician volcani-
clastic rocks at Sellaeld, north-west England. Proceedings of the Yorkshire
Geological Society 52, 159175.
Hiraga, N., Ishii, E., 2008. Mineral and Chemical Composition of Rock Core and
Surface Gas Composition in Horonobe Underground Research Laboratory
Project (Phase 1). JAEA Technical Report JAEA-Data/Code 2007-022. Japan
Atomic Energy Agency, Tokai-mura, Japan.
Iijima, A., Tada, R., 1981. Silica diagenesis of Neogene diatomaceous and volcani-
clastic sediments in northern Japan. Sedimentology 28, 185200.
Ikeda, Y., 2002. The origin and mechanism of active folding in Japan. Active Fault
Research 22, 6770.
Isaacs, C.M., 1982. Inuence of rock composition on kinematics of silica phase
changes in the Monterey Formation, Santa Barbara area, California. Geology 10,
304308.
Ishii, E., Fukushima, T., 2006. A case study of analysis of faults in Neogene siliceous
rocks. Journal of the Japan Society of Engineering Geology 47, 280291.
Ishii, E., Yasue, K., Tanaka, T., Tsukuwi, R., Matsuo, K., Sugiyama, K., Matsuo, S., 2006.
Three-dimensional distribution and hydrogeological properties of the Omagar
Fault in the Horonobe area, northern Hokkaido, Japan. The Journal of the
Geological Society of Japan 112, 301314.
Ishii, E., Hama, K., Kunimaru, T., Sato, H., 2007. Change in groundwater pH by
inltration of meteoric water into shallow part of marine deposits. The Journal
of the Geological Society of Japan 113, 4152.
Ishii, E., Yasue, K., Ohira, H., Furusawa, A., Hasegawa, T., Nakagawa, M., 2008.
Inception of anticline growth near the Omagari Fault, northern Hokkaido, Japan.
The Journal of the Geological Society of Japan 114, 286299.
Table A1
Acronyms and denitions.
Acronym Denition
FCB Fault crosscutting bedding planes at a high angle
Major FCB FCB composed of fault planes with associated
fault rocks such as breccia
Minor FCB FCB composed of fault planes with associated
slickenlines and/or slickensteps but not fault
rocks such as breccia
X Shortest distance from the GrifthCoulomb failure
envelope of rock strength to the Mohr circle
X
s
Shortest distance from the Coulomb failure envelope
of rock strength to the Mohr circle
X
t
Shortest distance from the Grifth failure envelope
of rock strength to the Mohr circle
SDT Suitable domain for pervasive tensile failure
UDT Unsuitable domain for pervasive tensile failure
Japanese Industrial Standard Committee, 2000. Method of Test for Tensile Strength
of Rock JIS M 0303. Japanese Standards Association, Tokyo, Japan.
Khang, N.D., Watanabe, K., Saegusa, H., 2004. Fracture step structure: geometrical
characterization and effects on uid ow and breakthrough curve. Engineering
Geology 75, 107127.
Kovari, K., Tisa, A., Einstein, H.H., Franklin, J.A., 1983. Suggested methods for
determining the strength of rock materials in triaxial compression: revised
version. International Journal of Rock Mechanics and Mining Sciences 20,
283290.
Kulander, B.R., Dean, S.L., Ward Jr., B.J., 1990. Fractured core analysis: interpretation,
logging, and use of natural and induced fractures in core. In: AAPG Methods in
Exploration Series 8. AAPG, Oklahoma, USA.
Kurikami, H., 2007. Groundwater Flow Analysis in the Horonobe Underground
Research Laboratory Project: Recalculation based on the Investigation until
Fiscal Year 2005. JAEA Technical Report JAEA-Research 2007-036. Japan Atomic
Energy Agency, Tokai-mura, Japan.
Kurikami, H., Takeuchi, R., Yabuuchi, S., 2008. Scale effect and heterogeneity of
hydraulic conductivity of sedimentary rocks at Horonobe URL site. Physics and
Chemistry of the Earth Parts A/B/C 33 (Suppl. 1), S37S44.
B07403. doi:10.1029/2007JB005388.
Martel, S.J., Pollard, D.D., 1989. Mechanics of slip and fracture along small faults and
simple strike-slip fault zones in granitic rock. Journal of Geophysical Research
94, 94179428.
Mazurek, M., 1998. Geology of the crystalline basement of northern Switzerland
and derivation of geological input data for safety assessment models. NAGRA
Technical Report NTB 93-12, NAGRA, Wettingen, Switzerland.
Mazurek, M., 2000. Geological and hydraulic properties of water-conducting features
in crystalline rocks. In: Stober, I., Bucher, K. (Eds.), Hydrogeology of Crystalline
Rocks. Kluwer Academic Publishers, Dordrecht, Netherlands, pp. 326.
Mazurek, M., Bossart, P., Eliasson, T., 1996. Classication and characterization of
water-conducting features at A
spo : results of investigations on the outcrop

scale. SKB International Cooperation Report ICR 97-01, SKB, Stockholm, Sweden.
Mazurek, M., Lanyon, G.W., Vomvoris, S., Gautschi, A., 1998. Derivation and applica-
tion of a geologic dataset for ow modeling by discrete fracture networks in low-
permeability argillaceous rocks. Journal of Contaminant Hydrology 35, 117.
Mazurek, M., Jakob, A., Bossart, P., 2003. Solute transport in crystalline rocks at A
spo
I: geological basis and model calibration. Journal of Contaminant Hydrology
61, 157174.
Milodowski, A.E., Gillespie, M.R., Naden, J., Fortey, N.J., Shepherd, T.J., Pearce, J.M.,
Metcalfe, R., 1998. The petrology and paragenesis of fracture mineralisation in
the Sellaeld area, west Cumbria. Proceedings of the Yorkshire Geological
Society 52, 215241.
Mitsui, K., Taguchi, K., 1977. Silica mineral diagenesis in Neogene tertiary shales in
the Tempoku district, Hokkaido, Japan. Journal of Sedimentary Petrology 47,
158167.
Murata, K.J., Nakata, J.K., 1974. Cristobalitic stage in the diagenesis of diatomaceous
shale. Science 184, 567568.
Niunoya, S., Matsui, H., 2005. The Investigation on Rock Mechanics in HDB-1
and HDB-2 Boreholes in order to Select the URL Area. JNC Technical Report
TN5400 2005-012. Japan Nuclear Cycle Development Institute, Tokai-mura,
Japan.
Nordqvist, R., Gustafsson, E., Andersson, P., Thur, P., 2008. Groundwater ow and
hydraulic gradients in fractures and fracture zones at Forsmark and Oskar-
shamn. SKB Technical Report R-08-103, SKB, Stockholm, Sweden.
O
hman, J., Niemi, A., Tsang, C.-F., 2005. Probabilistic estimation of fracture trans-
missivity from Wellbore hydraulic data accounting for depth-dependent
anisotropic rock stress. International Journal of Rock Mechanics and Mining
Ota, K., Abe, H., Yamaguchi, T., Kunimaru, T., Ishii, E., Kurikami, H., Tomura, G.,
Shibano, K., Hama, K., Matsui, H., Niizato, T., Takahashi, K., Niunoya, S., Ohara, H.,
Asamori, K., Morioka, H., Funaki, H., Shigeta, N., Fukushima, T., 2007. Horonoe
Underground Research Laboratory Project, synthesis of Phase I investigations
20012005. JAEA Technical Report JAEA-Research 2007-044. Japan Atomic
Energy Agency, Tokai-mura, Japan.
Petit, J.P., 1987. Criteria for the sense of movement on fault surfaces in brittle rocks.
Rhe n, I., Forsmark, T., Forssman, I., Zetterlund, M., 2006. Evaluation of hydro-
geological properties for Hydraulic Conductor Domain (HCD) and Hydraulic
Rock Domains (HRD). SKB Technical Report R-06-22, SKB, Stockholm, Sweden.
Shen, B., Stephansson, O., 1993. Numerical analysis of mixed Mode I and Mode II
fracture propagation. International Journal of Rock Mechanics and Mining
Sibson, R.H., 1977. Fault rocks and fault mechanisms. Journal of Geological Society
133, 191213.
Geology 23, 18251844.
Solum, J.G., van der Pluijim, B.A., Peacor, D.R., 2005. Neocrystallization, fabrics and
age of clay minerals from an exposure of the Moab Fault, Utah. Journal of
Tada, R., Iijima, A., 1982. Petrology and diagenetic changes of Neogene siliceous
rocks in northern Japan. Journal of Sedimentary Petrology 53, 911930.
Twiss, R.J., Moores, E.M., 2007. Structural Geology, second ed. W.H. Freeman and
Company, New York.
Wei, D., Seno, T., 1998. Determination of the Amurian plate motion. In: Flower, M.,
Chung, S.L., Lo, C.H., Lee, T.Y. (Eds.), Mantle Dynamics and Plate Interactions in
East Asia. Geodynamics Series 27. American Geophysical Union, Washington,
D.C., USA, pp. 337346.
Wei, Z.Q., Egger, P., Descoeudres, F., 1995. Permeability predictions for jointed rock
masses. International Journal of Rock Mechanics and Mining Sciences 32, 251261.
Welch, M.J., Davies, R.K., Knipe, R.J., 2009. A dynamic model for fault nucleation and
propagation in a mechanically layered section. Tectonophysics. doi:10.1016/
j.tecto.2009.04.025.
Williams, J.H., Johnson, C.D., 2004. Acoustic and optical borehole-wall imaging
for fractured-rock aquifer studies. Journal of Applied Geophysics 55,
151159.
of Geophysical Research 112, B08408. doi:10.1029/2006JB004815.
Yabuuchi, S., Kunimaru, T., Ishii, E., Hatsuyama, Y., Ijiri, Y., Matsuoka, K., Ibara, T.,
Matsunami, S., Makino, A., 2008. Horonobe Underground Research Laboratory
Project, Overview of the Pilot Borehole Investigation of the Ventilation Shaft
(PB-V01): Hydrogeological Investigation. JAEA Technical Report JAEA-Data/Code
2008-026. Japan Atomic Energy Agency, Tokai-mura, Japan.
Yamaji, A., 2000. The multiple inverse method: a new technique to separate
stresses from heterogeneous fault-slip data. Journal of Structural Geology
22, 441452.
Yamamoto, H., 1979. The geologic structure and the sedimentary basin off northern
part of the Hokkaido Island. Journal of the Japanese Association of Petroleum
Technologists 44, 260267.
Structural and petrophysical evolution of extensional fault zones in low-porosity,
poorly lithied sandstones of the Barreiras Formation, NE Brazil
F. Balsamo
a,
*
, F. Storti
a
, F. Salvini
a
, A.T. Silva
b
, C.C. Lima
b
a
Universita` degli Studi Roma Tre, Dipartimento di Scienze Geologiche, 00146 Roma, Italy
b
Cenpes, Petrobras, Rio de Janeiro, Brazil
Article history:
2 October 2009
Accepted 5 October 2009
Keywords:
Grain size reduction
Fractal distribution
Porosity
Cataclasis
Dilatancy
Fault-zone permeability
a b s t r a c t
We describe the structural and petrophysical evolution of extensional fault zones developed in low
porosity, poorly lithied, quartz-dominated sandstones from the Mio-Pliocene continental Barreiras
Formation, NE Brazil. We studied eight fault zones developed as sands were lithied. Fault displacement
ranges from a few centimetres to w50 m. A diagnostic feature of the studied fault zones is the lack of
deformation bands, which typically develop in high porosity sand(stone)s. Structural and microstructural
analyses, grain size and shape analyses, porosity and pore size analyses, and laboratory and in situ
permeability measurements show relationships between deformation processes and hydrologic prop-
erties. Undeformed rocks are very poorly sorted, medium- to ne-grained, clay-rich sandstones with an
average intergranular porosity of about 3%. Sandstones in damage zones record non-destructive dilatant
granular ow and formation of opening-mode intergranular extensional fractures, which increase
porosity, pore connectivity and permeability. Deformation in fault cores evolved from particulate ow to
compactional cataclastic ow, with progressive grain size reduction increasing the amount of silt- and
clay-size fractions. Porosity was dramatically reduced to an average value of 0.2% and permeability is
generally lower than the related protoliths. All this evidence highlights a conduit/barrier behaviour of the
studied fault zones, which signicantly differs from the sealing behaviour of deformation band fault
zones commonly observed in high-porosity sandstones.
1. Introduction
Porosity and its evolution through time are an important factor
controlling what kind of mesoscopic deformation structures
develop during rock failure (e.g. Dunn et al., 1973; Vernik et al.,
1993; Kwon et al., 2005; Fossen et al., 2007). In the last four
decades, considerable attention has been devoted to the under-
standing of fault zone evolution in high-porosity (>1015%), lithi-
ed to loose granular material both in the eld (Aydin, 1978; Aydin
and Johnson, 1978; Pittman, 1981; Lucas and Moore, 1986; Anto-
nellini and Aydin, 1994; Fowles and Burley, 1994; Fossen and
Hesthammer, 1997; Heynekamp et al., 1999; Cashman and Cash-
man, 2000; Shipton and Cowie, 2001; Rawling and Goodwin, 2003;
Flodin et al., 2003, 2005; Johansen et al., 2005; Minor and Hudson,
2006) and in experimental studies (Borg et al., 1960; Mandl et al.,
1977; Mene ndez et al., 1996; Wong et al., 1997; Zhu and Wong,
1997; Mair et al., 2000). This because of their important inuence
on uid ow in hydrocarbon reservoirs and groundwater aquifers
(e.g. Haneberg, 1995; Walsh et al., 1998; Heynekamp et al., 1999;
Aydin, 2000; Fisher and Knipe, 2001; Rawling et al., 2001;
Manzocchi et al., 2002; Nelson et al., 2009). Deformation in high-
porosity granular materials occurs by development of small
displacement deformation structures comprehensively referred to
as deformation bands, which evolve into zones of deformation
bands and slip surfaces with increasing offset (e.g. Aydin and
Johnson, 1978; Fowles and Burley, 1994; Shipton and Cowie, 2001;
Fossen et al., 2007). A typical result of deformation band faulting in
high-porosity sandstones is that their extensive development in
fault damage zones may reduce fault transmissibility, thus
providing an effective barrier to uid ow (e.g. Antonellini et al.,
1994, 1999; Sigda et al., 1999; Rotevatn et al., 2007). This hydraulic
behaviour differs from the typical conduit behaviour of fault
damage zones in low-porosity fully lithied rocks, where defor-
mation is dominated by opening-mode fracturing (e.g. Caine et al.,
1996; Billi et al., 2003; Kim et al., 2004).
The lower threshold porosity limit for deformation band
development is at about 1015% (e.g. Dunn et al., 1973; Flodin et al.,
2003; Wong et al., 1997). Below this threshold limit, shear strength
* Corresponding author. Tel.: 39 0657338049; fax: 39 0657338201.
E-mail address: balsamo@uniroma3.it (F. Balsamo).
doi:10.1016/j.jsg.2009.10.010
becomes a fundamental parameter controlling deformation
mechanisms. Joints and slip surfaces are expected to develop in
fully lithied sandstones (e.g. Johansen et al., 2005; Fossen et al.,
2007). On the other hand, deformation in low-porosity poorly
lithied sand(stone)s is still poorly understood. In this paper, we
attempt to bridge the gap by describing the structural and petro-
physical evolution of extensional fault zones developed in low-
porosity, poorly lithied quartz-dominated sandstones of the Bar-
reiras Formation, NE Brazil. The relative compositional maturity
and homogeneity of the Barreiras sandstones allow us to discount
the effects of clay smearing and tectonic mixing of strongly
different sedimentary units within fault zones (e.g. Antonellini and
Aydin, 1994; Gibson, 1998; Heynekamp et al., 1999; Caine and
Minor, 2009). Results of structural, microstructural, grain size, grain
shape, and porosity analyses and permeability measurements are
described with the aim of (1) inferring the deformation mecha-
nisms that governed the evolution of these extensional fault zones;
(2) proposing an evolutionary model of grain size, grain shape and
porosity changes during extensional faulting; and (3) assessing the
inuence of faulting on uid ow by establishing a relationship
between fault-related permeability variations and fault displace-
ment. The latter provides a useful tool for predicting the expected
permeability and transmissibility of sub-seismic and seismic fault
zones in sand-dominated clastic reservoirs.
2. Analytical methods
Structural analysis was used to constrain the mesoscale archi-
tecture and kinematics of the studied extensional fault zones. Where
offset markers were available, stratigraphic separations were
measured in the eld and then converted into true fault displace-
ment values by using fault kinematics (Butler and Bell, 1989). Fault
core thicknesses were measured to determine whether or not there
was a predictive statistical relationship that could be used for esti-
mating fault displacement from fault core width when direct
measurements were not possible (e.g. Walsh and Watterson, 1988).
Undeformed sandstones and rocks in damage zones and fault
cores were sampledat eachstudied eldsite. About 0.5 kgof material
was collected for each sample. After complete disaggregation and
chemical removal of Fe-oxides in the laboratory, grain size analysis of
about 60gof granular material was completedbycombiningstandard
sieve and laser diffraction analyses in order to account for coarser
(gravel- and sand-size) and ner (silt- and clay-size) fractions,
respectively (Selley, 2000). A total of 44 samples were analysed
including 17 undeformed sandstones, 16 fault core rocks, and 11
damage zone sandstones. Results of grain size data were plotted as
frequency distribution curves using the Phi scale arrangement
(Krumbein, 1934, 1938). Grain size distributions were quantitatively
described in terms of the following statistical parameters: (a) the
mean size Phi
m
(a measure of the average size of the curve); (b) the
standard deviation So (a measure of the size spread around the mean
value, or sorting); (c) the skewness Sk (a measure of the curve
symmetry around the mean value, or preferential spread to one side
of the mean); and (d) the kurtosis (the degree of concentration of the
grain sizes relative to the mean value) (Inman, 1952; Grifths, 1952).
These parameters were obtainedby mathematical methods (Folkand
Ward, 1957; Krumbein and Pettijohn, 1938) employing the entire
sample grain size populations (McManus, 1988). Grain size distribu-
tions were also transformed into equivalent particle numbers (e.g.
Storti et al., 2003) assuming spherical particles with a density of 2.65
g/cm
3
and plotted against the equivalent size classes in bilogarithmic
diagrams to obtain their fractal dimensions (D) as the slope of the
best-t lines (e.g. Blenkinsop, 1991).
Additional non-destructive sampling was carried out in the
three structural domains (undeformed sands, damage zones and
fault cores) for blue-dyed epoxying and thin section analysis.
Microstructural analyses were carried out with a standard petro-
graphic microscope connected to a digital photo camera.
Computer-based image analysis techniques (e.g. Francus, 1998;
Heilbronner and Keulen, 2006) were applied to thin section images
using the Optimas-6.5
commercial software (e.g. Shipton and

Cowie, 2001). Colour TIFF images (2568 1938 pixel resolution;
1pixel 2.6 mm) were acquired at constant 1.25 magnication
under plane polarised light, in order to ensure the appropriate
visualization of both microporosity and macroporosity (i.e. pores
with equivalent diameter lower and greater than 63 mm, respec-
tively; Choquette and Lloyd, 1970). A total of 45 images were ana-
lysed to characterise thin section sectors representative of
undeformed, damage zone and fault core rock fabrics, respectively.
Microfabric heterogeneities were addressed by acquiring multiple
images from the same thin section.
Images were smoothed by one cycle of 3 3 convolution mask
lter to remove the background noise and converted to binary
images. To avoid sampling errors, only the objects with a 3 3 pixel
minimum area (i.e. objects with an equivalent diameter greater
than 8.8 mm) were considered (Francus and Pirard, 2004). Quanti-
tative thin section analysis of unaltered granular material included:
(a) the total percentage of 2D intergranular porosity, calculated as
the ratio between void area and total area of the image 100; and
(b) the size of pores, calculated as equivalent diameters (i.e.
diameter of the circular area with the same area of the object) and
their size distributions. We assumed that 2D porosity is a reason-
able approximation of the 3D porosity (e.g. Johansen et al., 2005).
Grain shape data were acquired from the different architectural
elements of a characteristic fault zone among the studied ones (i.e.
undeformed, damaged and fault core domains). Image analysis was
applied to thin sections of 3 4 cm samples produced from loose
grains from the fault zone domains dispersed into blue epoxy. Five
standard sand-size classes were separately analysed for grain
shape: 1 mm; 0.5 mm; 0.250 mm; 0.125 mm; and 0.063 mm (i.e.
Phi 0, 1, 2, 3, 4, respectively). Angularity (A Perimeter
2
/Area) of
at least 100 grains was measured for any size class to ensure
a statistically signicant population (Anders and Wiltschko, 1994).
A total of 54 images from 15 standard thin sections were analysed.
Angularity data were represented as frequency histograms and
best-t polymodal Gaussian distributions were determined (Wise
et al., 1985) using the software Daisy 4.1 (Salvini, 2004). For
comparison, the obtained angularity values were calibrated with
the Krumbein (1941) roundness visual chart. As a reference an
angularity value of 17.7 corresponds to the threshold value between
sub-rounded and sub-angular particles of Krumbein (1941).
Air and gas permeameters were used to measure the perme-
abilities of protolith sandstones and fault zone structural domains
in the eld and laboratory, respectively. Laboratory measurements
on oriented samples were performed with a gas-permeameter
PDPK-400, which provides reliable results from 10
6
to greater
than 30 Darcy. In situ measurements along fault-perpendicular
transects starting and terminating in undeformed domains were
performed by a portable Tiny-perm II air minipermeameter, which
provides reliable permeability values in the range 10
4
up to 10
Darcy. Sampling sites were carefully scraped with a putty knife and
gently brushed to remove weathering effects prior to permeability
measurements. Permeability data were statistically analysed using
best-t polymodal Gaussian distributions.
The northeastern part of the Brazilian coastal plain where the
study area is located exposes a Precambrian crystalline basement
overlain by Cretaceous basinal rocks (Jandaira carbonates and Au
F. Balsamo et al. / Journal of Structural Geology 32 (2010) 18061826 1807
uvial sandstones) and a Cenozoic sedimentary cover (Fig. 1). The
latter formed late in the rifting history of the Atlantic ocean and
mostly consists of a continental clastic sequence called the Barrei-
ras Formation, which crops out along more than 4000 km of the
Brazilian littoral zone from the Amazon delta to the Rio de Janeiro
coast (e.g. Mabesoone et al., 1972; Suguio and Nogueira, 1999). The
Barreiras formation mainly consists of less-than 100-m thick,
poorly lithied, massive quartz-dominated sandstones. The degree
of lithication is highly variable and strongly depends on the
amount and type of weathering-induced Fe-oxide cementation,
which is responsible for the typical orange/yellow to red/violet
colours of outcropping sandstones. An early Miocene age of the
Barreiras Formation in the study area was obtained by dating
detrital and authigenic Fe-pisoliths (using the (UTh)/He method)
from the Barreiras sediments. The age interval is constrained by the
minimum age of the detrital pisoliths (22 Ma) and the maximum
age of the authigenic pisoliths (17 Ma) (Lima, 2008).
There is clear evidence of post-rift faulting in NE Brazil since the
late Tertiary (e.g. Bezerra and Vita-Finzi, 2000). Deformation
includes strike-slip and extensional faulting reactivating Precam-
brian shear zones and generating new structures which controlled
sediment deposition and coastal morphology (Bezerra et al., 2001).
Signicant historical and modern seismicity indicates that faulting
is ongoing in the region (Ferreira and Assumpao, 1983; Ferreira
et al., 1998).
The study area is located in the western part of the Potiguar
sedimentary basin (de Matos, 1992), northwest of Icapu village,
and extends for more than 10 km along the coast (Fig. 1). Cliff
exposures mainly consist of Barreiras Formation, although in few
Fig. 1. Simplied geological map of the Potiguar Basin, NE Brazil, showing the location
of the study area northwest of the Icapu village, in the continental Mio-Pliocene
Barreiras formation.
Fig. 2. The studied eld sites and lower hemisphere, equal area projections of the measured extensional faults and associated slickenlines. Geographic coordinates are listed. Poles
of all measured extensional fault surfaces are contoured at 3% intervals (equal area projection, lower hemisphere). Slickenlines (black dots) are projected in a lower hemisphere,
equal area projection. For regional location of the studied area see Fig. 1.
sites the Jandaira carbonates crops out in both stratigraphic and
tectonic contact with the overlying Barreiras sandstones.
4. Structural outline
The complex fault pattern exposed in the study area is charac-
terised by abundant fault segments showing different kinematics,
orientation, amount of displacement, and deformational styles
(Fig. 2). In particular, extensional fault segments have variable
orientations, though they exhibit dominant NE and secondary
NNW strikes. Cross-cutting relationships indicate that extensional
faults are generally overprinted by sub-vertical strike-slip faults.
The major extensional fault segments are generally associated with
faultward thickening of syn-tectonic sediments in the hanging-
walls, pointing to the occurrence of fault activity during the
deposition of the Barreiras sandstones. Field observations
(described in following sections) indicate that most of the major
extensional fault zones developed in soft sediment conditions,
while strike-slip faulting occurred in soft up to partially lithied
granular material, as indicated by the systematic presence of sub-
vertical extensional fractures in damage zones. For this study we
carefully selected eld sites where structures associated with
strike-slip fault zones were negligible or not visible at the outcrop
scale.
Fig. 3. (a) Structural architecture of an extensional fault zone with about 15 m of displacement consisting of a fault core and hanging-wall and footwall damage zones. The latter is
poorly developed and has variable thickness. Subvertical bleached fractures are associated with late stage strike-slip faulting. (b) Detailed view of a typical fault core structural
association, including sand pods incorporated during fault movement, foliated clay-rich material, and multiple anastomosing slip surfaces dening an SCC
0
array. Slip surfaces
bound yellow, irregularly stretched, sand blobs which lack original sedimentary fabric. (c) Mesoscopically ductile fabric and pervasive foliation within the fault core, surrounding
a weakly deformed, round-shaped sand pod. The fault core-damage zone boundary consists of a striated master slip surface. (d) Stretched and internally foliated yellow sand blob
within the fault core, showing cuspate boundaries with strongly foliated fault gouge.
4.1. Architectural elements
The internal structure of the studied fault zones (Fig. 3a) consists
of three main architectural elements: undeformed protolith,
damage zone, and fault core, which accommodated the majority of
the slip (e.g. Chester et al., 1993; Caine et al., 1996).
Fault cores exhibit macroscopically distributed ductile struc-
tures (cf. Rutter, 1986; Rawling and Goodwin, 2006) without
extensive localisation (Fig. 3ac). Their internal fabric is extremely
heterogeneous at the mesoscale due to the occurrence of closely
spaced and multiple, anastomosing slip surfaces oriented sub-
parallel to the master slip surfaces and forming SCC
0
structural
arrays (e.g. Passchier and Trouw, 1996) (Fig. 3b). Slip surfaces
separate strongly foliated red/orange ne-grained sand layers,
irregularly stretched yellow sand blobs, and sand pods (Fig. 3c
and d). Sand pods are 3080 cm wide and their shape varies from
elongate in the direction of fault movement to sub-rounded. Pods
have relatively sharp boundaries and locally preserve sedimentary
structures. Conversely, yellow sand blobs are 520 cm long and are
generally strongly elongate. Blobs show irregular globular to
cuspate boundaries and their internal sedimentary fabric is
partially or completely disrupted. Despite the initial relative
homogeneity of the Barreiras Formation, fault core width is locally
highly variable along both fault strike and dip within the same fault
zone.
Fault cores are typically bound by two master slip surfaces and
are generally surrounded by structurally heterogeneous hanging-
wall and footwall damage zones. In some cases, however, the
latter are partially developed or absent (Fig. 3a). Damage zones
typically include widespread mobilised sediments (e.g. Maltman
and Bolton, 2003) where sediments are transposed into foliation
(Fig. 3a) and partially cemented beds are attenuated, folded, and,
eventually, disrupted. These features resemble the mixed zones
described by Heynekamp et al. (1999) and Rawling and Goodwin
(2003) in high-porosity, poorly lithied sediments. The boundaries
between damage zones and the undeformed sandstones are
typically gradual and were placed at points where the sub-
horizontal undeformed bed attitude started to be reworked by
deformation. Zones of deformation bands sub-parallel to the main
fault surfaces (e.g. Aydin, 1978; Rawling and Goodwin, 2003) were
not recognised neither at the boundary between mobilised sedi-
ments and undeformed protoliths, nor adjacent to the master slip
surfaces. In some cases, the macroscopically ductile fabrics in the
damage zones are overprinted by minor faults and localised
dilatant fractures.
In the range of measured displacements (0.120 m), mean fault
core and damage zone thicknesses in 5 fault zones linearly increase
with increasing displacement (Fig. 4). For the construction of
diagrams in Fig. 4 we used mean thicknesses of cores and damage
zones, obtained by averaging several across-fault measurements.
The obtained scaling relationships were then used for calculating
fault displacement of fault zones where unambiguous reference
markers were not available or where fault displacement exceeded
the cliff height.
4.2. Microstructural features
From a textural and mineralogical point of view, undisturbed
materials are poorly sorted, grain-supported quartz-dominated
sandstones with a wide range in pore size and thin, pore-lling Fe-
oxide cement bonds. Coarser grains are sub-angular and elongate
showing a slight preferential orientation parallel bedding (Fig. 5a
and b).
The most common feature of the studied extensional fault
damage zones is the widespread presence of zones where larger
clasts are further apart than in the undeformed sand, and are
separated by ner grained material (Fig. 5c). Quartz grains show
a loose packing geometry and coarser grains do not show any
preferential orientation. Locally, poorly foliated sands occur along
subsidiary fault splays. Fe-oxide cement is patchily distributed
within damage zones. In some locations, localised opening-mode
deformation structures such as dilation bands (Du Bernard et al.,
2002) and extensional fractures were observed. Extensional frac-
tures developed as irregular intergranular opening-mode fractures
with sharp boundaries that mainly follow the coarser grain rims
(Fig. 5d). Extensional fractures mainly formed in tighter and/or
more cemented granular packages (Fig. 5df). Their apertures are
generally less than 100 mm. In some cases, intergranular exten-
sional fractures are lled by Fe-oxide cement (Fig. 5e), whereas in
other cases they cut cement bonds (Fig. 5f). Dilation bands typically
occur at intergranular extensional fracture tips and are charac-
terised by diffused, gradational boundaries with respect to the host
sandstone. They have an average thickness of about 0.51 mm
(Fig. 5g).
Thin section analysis on fault core rocks was performed on
foliated clay-rich sands, which generally exhibit variable grain size
and porosity. Key observations include: 1) a general decrease in the
mean grain size relative to damage zone samples, with coarse
elongate and angular survivor grains dispersed in a mediumto ne-
grained matrix (Fig. 6a); 2) fault-parallel alignment of elongate
coarse and ne grains which dene the foliation (Fig. 6b); 3) at the
grain scale, the coarser quartz grains exhibit distributed intra-
granular fracturing and aking (Fig. 6c). In low-displacement faults
(<1 m) dilational fabrics in damage zones are separated from
tighter packing in fault core rocks by diffuse boundaries (Fig. 6d).
Fe-oxide cement in fault core rocks was rarely observed.
Fig. 4. Fault core (a) and damage zone (b) thicknessdisplacement relationships
measured in the studied fault zones. Error bars represent minimum and maximum
values.
Fig. 5. Plane-polarised-light photomicrographs of undeformed (a, b) and damage zone (cg) sandstones impregnated with blue epoxy. (a) Undeformed sandstone of the Barreiras
formation showing the well cemented, low-porosity and quartz-dominated composition. Quartz sand grains (Qz) are sub-angular to very angular. (b) Example of 2D intergranular
primary porosity (in blue) consisting of microporosity (mP) (i.e. pores with equivalent diameters <63 mm), and macroporosity (MP) (i.e. pores with equivalent diameters >63 mm).
Quartz grains are bonded by Fe-oxide meniscus cement. (c) Loose packing fabric in the hanging-wall damage zone of a fault zone with displacement of w20 m. Note sub-angular to
very angular quartz grains. (d) Intergranular extensional fracture in the hanging-wall damage zone of a fault zone with estimated displacement of w35 m. Such structures typically
overprint a pre-existing fabric. (e) Fe-oxide-lled intergranular extensional fracture. (f) Intergranular extensional fracture developed in a well cemented sandstone. (g) Dilation band
developed in coarse-grained sandstone at the tip of an extensional fracture.
5. Undeformed sandstone characterisation
5.1. Grain size analysis
Results of grain size analyses on 17 samples of undeformed
sandstone collected at different eld sites are summarised as
frequency curves (weight %) in Fig. 7a, while sedimentological
statistical parameters are listed in Table A1 (see Appendix). The
percentage of gravel-size material (>2 mm) ranges from 0.1% to
7.7% with an average value of 1.76 2.53%. The sand-size fractions
(20.063 mm) are the most abundant ones and vary from 42.63%
up to 82.75%, with an average value of 66.2 6.39%. The silt-size
fractions (0.0630.004 mm) ranges from 10.22% to 21.63%, with an
average value of 11.8 2.95%. The cumulative residual clay-size
fractions (<4 mm) ranges from5.67% (clean sandstone) up to 35.68%
(impure or clay-rich sandstone) with an average value of
23.2 2.6%.
The mean size (Phi
m
) varies from 2.34 (ne sand) to 5.22
(medium silt) with an average value of 3.4 0.66 (very ne sand).
The most representative size classes (i.e. the modes) of the unde-
formed samples are Phi 2 (mediumsand, grain diameter between
0.5 and 0.25 mm) and Phi 3 (ne sand; 0.250.125 mm). Sample
sorting varies between 2.25 and 3.62 with a mean value of 3 0.32
(very poorly sorted). Skewness ranges from 0.19 (near symmet-
rical granulometric curve) up to 1.35 (very ne-skewed curve), with
a mean value of 0.6 0.3 (slightly coarse skewed). Kurtosis varies
between 1.43 (very platykurtic curve) and 4.21 (leptokurtic curve);
the average value is 1.6 0.16 (platykurtic curves). The fractal
dimension of the undeformed samples is quite homogeneous
varying from 2.42 up to 3.06 with an average value of 2.7 0.11.
5.2. Porosity and permeability
Five thin sections from ve undisturbed samples were selected
to quantify the intergranular porosity and the corresponding pore
size distributions. A total of 17 digital images were analysed. In all
the analysed thin sections, pore size mainly varies from 32 to 125
mm (Fig. 8a). Measured intergranular porosity ranges between 1.66
0.5% and 6.77 0.14% (Fig. 8b); thus, these are low porosity,
generally moderately cemented sandstones. The contribution of
micro and macroporosity is almost equivalent.
Permeability in the undeformed sandstones was measured at
six eld sites. Permeability statistics are summarised in Fig. 8c.
Measured permeability ranges from 0.63 to 233.69 md, with mean
values spanning two orders of magnitude from2.61 0.35 to 70.35
0.26 md.
6. Damage zone characterisation
Grain size analyses were performed on samples collected both
in the hanging-wall and footwall damage zones of ve extensional
faults with different displacement values. Grain size distributions
are more variable than those of undeformed samples since they
underwent different amounts and types of deformation during
Fig. 6. Plane-polarised-light photomicrograph from fault core samples impregnated with blue epoxy. (a) Tight packing fabric in the fault core of a fault zone with displacement of
w25 m. Coarse, angular, survivor quartz grains are surrounded by a ne-grained matrix. (b) Foliation imparted by the preferential alignment of coarse and ne grains surrounding
a narrow gouge seam. Note difference in grain-size distribution above and below the gouge. (c) Example of both rounded (R) and angular (A) coarse grains surrounded by very
angular and elongated ner fragments produced by intense crushing of the coarser grains (arrows). (d) Boundary (dotted line) between very low-porosity fault core (FC) and
relatively high-porosity damage zone (DZ) in a fault zone with displacement < 3 m. A angular grain, R rounded grain.
fault zone evolution (Fig. 7b). The most abundant fractions in the
analysed samples are the sand-size ones, which range from 30.09%
to 75.77%, with a mean value of 51.1 14.38% (Table A2 in the
Appendix). The gravel-size fractions are strongly subordinate,
ranging from 0% to 1.73%. The silt-size fractions range from 7.7% to
30.69%, with a mean value of 14.15 2.17%. The residual clay-size
fraction varies from 10.48% to 39.16% with a mean value of 32.8
4.3%, higher than in the undeformed samples.
The modes of the damage zone samples, minus the residual
clay-size fractions, range between Phi 2 and Phi 4. Mean grain
diameters range from 3.10 to 5.98 with a mean value of 4.9 0.66
(coarse silt). Sorting ranges between 2.4 and 3.33 with a mean value
of 2.80.33(verypoorlysorted). Skewness broadlyranges from0.64
(coarse skewed curve) to 1.01 (ne-skewed curve), with a mean
value of 0 0.27 (symmetrical curve). Kurtosis varies between
1.33 (very platykurtic curve) and 3.34 (mesokurtic curve); the
mean value is 1.7 0.2 (platykurtic curve). Fractal dimensions
are not homogeneous and vary from 2.62 to 3.65, with a mean
value of 2.9 0.17.
A common result of dilatant structures in damage zones is that
porosity is largely produced by pores with equivalent diameters >63
mm, reected by the asymmetry of curves in Fig. 8d. Intergranular
porosity is signicantly higher than in the pristine sandstones,
varying between 5.14 0.6% and 9.07 0.27%, with the exception of
samples affected by intergranular extensional fractures that have
porosities of about 2% (e.g. sample 12C-03) (Figs. 8b and 5d). Dilation
bands formed in sandstones with mean porosity of about 46%;
porosity within the dilation bands, however, is as high as 30%.
Permeability was measured in both footwall and hanging-wall
damage zones of ve fault zones (Fig. 8c). Permeability values span
over three orders of magnitude, ranging from 7.61 md to 1033.95
md. Higher values were generally measured in the hanging-wall
blocks (which, as mentioned above, typically exhibit better devel-
oped damage zones). The highest permeability values were recor-
ded across localised dilation bands and extensional fractures,
whereas the lowest ones were recorded across secondary fault
splays. The mean permeability (the main Gaussian peak value) in
the analysed damage zones varies from26.83 0.13 md to 172.81
0.59 md.
7. Fault core characterisation
Grain size distribution curves of fault core samples are heteroge-
neous (Fig. 7c), showing differences even within the same fault core
due to compositional and structural heterogeneities (e.g. Fig. 3b), as
well as different displacement values that presumably caused
different deformation intensities. Grain-size distribution curves are
unimodal with the exception of two cases. The sedimentological
characteristics of fault core samples are signicantly different than
the parent undeformed sandstones (Table A3 in the Appendix).
The most abundant grain-size fractions in 56% of the samples
analysed are the silt- and residual clay-size ones. Exceptions are
samples 5C-06 and 5D-06, which were taken from the stretched
yellow sediment incorporated in the fault core and foliated sand,
respectively; samples BR1A, B, and E, which are from a fault zone
with displacement of w7 m; and sample BR12A, where the sand
size-fractions are most abundant. The amount of gravel-size
material is quite small (0.78 1.51%); sand-size fractions span from
2.59% to 70.51%, with a mean value of 34.4 21.7%. The silt-size
fractions range between 14.13% and 41.18%, with a mean value of
17.9 1.4%. The clay-size fraction ranges from5.71% to 77.92%, with
a mean value of 41.3 20.56%. The calculated mean grain size
(Phi
m
) of fault core rocks have a wide range of variability, from 2.85
(medium sand) up to 7.56 (very ne silt). The mean value is 5.5
1.56. All samples are poorly (So 1.22) to very poorly (So 3.28)
sorted with a mean value of 2.7 0.31. Granulometric curves are
generally symmetrical to very coarsely skewed, with the exception
of sample 5C-06 and the samples taken from faults with
Fig. 7. (a) Grain size distributions of undeformed sandstones showing ne-skewed
unimodal curves of poorly sorted, medium to ne sandstones. (b) Grain size distri-
butions of footwall and hanging-wall damage zone sandstones showing the unimodal
symmetrical curves of the poorly sorted, ne to very ne granular material. (c) Grain
size distributions of fault core rocks. Curves are generally unimodal and symmetrical to
slightly coarse skewed, with the exception of less deformed samples (i.e. sand blob and
foliated sand).
displacement lower than 10 m. The shape of the granulometric
curves ranges from very platykurtic to very leptokurtic, with
a mean value of Ku 2 0.47. The fractal dimensions of the ana-
lysed samples range from 2.7 to 4.29; the mean value is 3.1 0.2.
In fault core rocks microporosity prevails with respect to mac-
roporosity: very few large pores were found, whereas the bulk of
pores show sizes generally lower than 63 mm (Figs. 6a,b and 8e).
Total intergranular porosity in the analysed sections varies from
2.25 0.7% to 0.13 0.08% (Fig. 8b). Permeability was measured in
fault core rocks of nine fault zones with different displacement
values. Measurements were carried out in several sites within each
fault core and their cumulative statistics are summarised in Fig. 8c.
Permeability values are heterogeneous and span over 7 orders of
magnitude from 0.0039 md to 5263 md. Statistical analysis indi-
cates that the mean permeability of each fault core varies from 0.14
0.19 md to 51.02 0.48 md.
8. Particle shape analysis
The particle angularity data of undeformed, damaged and fault
core sandstones from a fault zone with estimated displacement of
Fig. 8. Porosity and permeability data. (a) Pore area frequency distribution curves of the analysed images from undeformed samples. (b) Summary table showing the average
porosity percentages in all the analysed samples. The number (n) of analysed pores and analysed images is indicated. Fault displacement values (Displ.) are also indicated.
(c) Summary table showing results of Gaussian distribution statistics for permeability measurements in the undeformed, damage zone, and fault core rocks. All permeability values
in the studied fault zones were measured in directions parallel to the fault strike. Mean values correspond to the main Gaussian peaks. (d) Pore area frequency distribution curves of
the analysed images from damage zone samples, showing the relative abundance of macroporosity and microporosity. (e) Pore area frequency distribution curves of the analysed
images from fault core samples, showing the relative abundance of microporosity and macroporosity.
w25 m are shown as frequency histograms (Fig. 9ac) for the whole
range of sand-size fractions (i.e. 1.00.063 mm). The angularity
values obtained are extremely heterogeneous and broadly scattered
between 15 (rounded particle) and 30 (very angular particle) in all
the analysed size classes. Statistical data analysis shows the existence
of polymodal distributions represented by Gaussian curves.
In the undeformed domain (Fig. 9a) the angularity of very coarse
(1 mm) sand particles peaks at 18.5 0.94, with a subordinate peak
at 20.8 1.02. Coarse undeformed sand particles (0.5 mm) have
similar angularity with the main Gaussian peak at 18.7 0.83, and
two more angular peaks at 20.9 0.59 and 23.4 0.89. The main
angularity of medium undeformed sand particles (0.25 mm) peaks
at 19.5 0.84, with two subordinate Gaussian peaks at 17.9 0.67
and 21.9 0.1.41. Angularity values of ne sand particles (0.125
mm) peaks at 18 1.3, 20 0.37 and 21.8 0.51, respectively. Very
ne sand particles (0.063 mm) have the lower angularity values
which peaks at 17 0.94, with two subordinate Gaussian peaks at
19.2 0.66 and 23 1.18.
In the damage zone (Fig. 9b), the main angularity of very coarse
sand particles peaks at 18.2 0.95, with a subordinated population
at 21.6 0.81. Coarse sand particles have peak angularity values of
18.7 1.08, with two peaks at 16.8 0.46 and 21.1 1.59. Medium
sand particle angularity is well clustered at 18.2 0.36, with two
peaks at 16.3 0.51 and 21 2.37. The angularity values of ne
sand particles in the damage zone peaks at 18.1 0.49, 16.3 0.5
and 19.7 0.24. The angularity of very ne sand particles is 17.1
1.02, with a subordinate Gaussian peak at 19.2 1.5.
Particle angularity in the fault core (Fig. 9c) is signicantly
different with respect to the previously described domains. Angu-
larity of very coarse sand particles is 18.6 0.8; a subordinate more
rounded subpopulation shows at a mean value of 17 0.49, while
a third more angular subpopulation is 22.5 0.9. Most coarse sand
particles show angularity values averaged at 19.1 1.13, with
a subordinate rounded population at 17.2 0.69. Angularity of
medium sand particles peaks at 17.8 1.8 and 20.8 1.19,
respectively. Fine sand particles are generally more angular and
broadly scattered ranging from22.5 2.44 to 17.2 0.76. The main
angularity mode of very ne sand particles peaks at 21.5 1.2, with
two more rounded subpopulations showing means at 16.8 0.58
and 18.5 0.42, respectively.
9. Comparative analyses
9.1. Granulometric data
Cumulative grain size distributions representative of the unde-
formed, damage zone and fault core domains are compared in
Fig. 10a. Data points dening each curve were obtained by aver-
aging the corresponding frequency values for each Phi classes in the
plots in Fig. 7. As expected, fault deformation produces a slight
decrease in the mean grain size in the damage zones, especially in
the secondary faults. In addition, the mean grain size is signicantly
reduced in the fault core rocks to a very ne silt-size material (Phi
m
5.5 1.56) (Fig. 10b). In particular, the sand-size content in the
Fig. 9. Shape analyses of sand-size fractions showing the polymodal frequency distributions of grain angularity in the undeformed (a), damage zone (b), and fault core sandstones
(c). Grain angularity plotted against frequency shows a polymodal frequency distribution in all analysed size classes. Best-t Gaussian curves are superimposed on the corre-
sponding data histograms. Sample name, Gaussian peaks, and related standard deviations, are also indicated.
undeformed domain is signicantly higher than in the damage
zones and fault cores (Fig. 10c). These analyses therefore show
a relative decrease in the sand-size fractions from undeformed to
damage zone to fault core rocks and a corresponding increase in
silt- and residual clay-size fractions (Fig. 10ce). The latter reaches
a mean value of about 41% in fault cores. Sorting, however, does not
show signicant variation across the three structural domains; all
of the samples are very poorly sorted (So > 2) (Fig. 10f), but
undeformed sandstones are positively skewed toward the coarser
sizes (Sk > 0), whereas fault core samples display negative skew-
ness towards the ner fractions (Sk < 0) (Fig. 10g). Samples from
damage zones show symmetrical grain size distributions (Sk 0).
Mean kurtosis values show very platykurtic to platykurtic curves
for all the three domains (Ku < 2.55) (Fig. 10h). Samples from fault
cores show higher values of kurtosis, indicating grain size distri-
butions with a higher concentration around the mean size. The
mean fractal dimensions systematically increase from the unde-
formed rocks (D 2.7 0.11) to the damage zones (D 2.9 0.17)
to the more intensely deformed fault cores (D 3.1 0.2) (Fig. 10i).
9.2. Porosity data
Microstructural analyses indicate that, in the 0.00881 mmpore
size interval, porosity, pore size and pore size distribution are
different in the three structural domains. Sandstones in unde-
formed domains have an average primary intergranular 2D
porosity of about 3% (Fig. 11a), with large pores between sand- and
silt-sized grains and a lot of micropores between ner particles. The
ratio between macro and microporosity in undeformed sandstones
is between 1 and 2 (Fig. 11b). Average porosity in damage zones is
nearly three times of that of undeformed materials, even though
not all samples show such an increase in bulk porosity (i.e. more
compacted jointed samples) (Fig. 11a). The ratio between macro
and microporosity is generally higher than in the undeformed
sandstones, with values ranging between 1.4 and 4.4 (Fig. 11b).
Porosity and pore size in fault core samples are dramatically
reduced with respect to both undeformed and damage zone rocks.
In particular, average porosity in fault cores is about 6.6% of that in
the undeformed ones, and about 2.3% of average porosity in
damage zone rocks (Fig. 11a). Pores generally have dimensions less
than 63 mm and thus the ratio between macro and microporosity is
mostly lower than 1 (Fig. 11b).
9.3. Morphometric data
Undeformed sand grains are generally sub-angular to very
angular: the main Gaussian peaks for all the size classes show
angularity values (A) generally higher than 18, with the exception
of very ne (<0.1 mm) sand particles, which are sub-rounded
(Fig. 12a). The main Gaussian peaks (dark dots) show angularity
between A 17 and A 19.5 which tends to decrease slightly with
particle size. Particle angularity in the damage zone is not
Fig. 10. Comparative granulometric data analyses of undeformed (solid line), damaged (dashed line) and fault core (dotted line) sandstones obtained by averaging the statistical
parameters shown in Table A1, A2, and A3. (a) Grain size distributions curves, (b) mean grain size, (c) sand-size content, (d) silt-size content, (e) residual clay-size content, (f) sorting,
(g) skewness, (h) kurtosis, (i) fractal dimension.
signicantly different from respect to the undeformed sand frac-
tions (Fig. 12b): particles are generally angular to very angular with
main Gaussian peak values between A 17.1 and A 18.7. The only
difference is the loss of the very angular (A > 22) subpopulations in
the medium and ner sand size classes, which exhibit more
rounded particles. Particle angularity is signicantly different in the
fault core, where angularity increases with decreasing particle size
(Fig. 12c). Very coarse, coarse, and medium size sands in the core
are mainly sub-rounded to sub-angular, ranging between A 17.2
and A 18.6; very angular coarser particles are almost completely
absent and populations of newly formed rounded particles are
common. In contrast, ne and very ne sand size fractions mainly
consist of angular to very angular particles with a values higher
than 21, even where relict sub-rounded populations derived from
the undeformed sandstone are still present.
9.4. Permeability data
Measured permeability in undeformed sandstones ranges from
about 10
1
to 10
2
md, displaying two Gaussian peaks with mean
Gaussian peak values of 2.52 0.3 md and 43.68 0.6 md,
respectively. Best-t unimodal Gaussian curve provides a mean
value of 11.7 0.9 md (Fig. 13a). Damage zone permeability is
signicantly higher, ranging over 4 orders of magnitude from
about 10
0
to 10
3
md with a mean value of 86.9 0.4 md (i.e.,
nearly eight times that of the undeformed sandstones) (Fig. 13b).
Fault core permeability measurements taken from faults with
different displacement span over six orders of magnitude, ranging
from 10
2
to 10
3
md. Cumulative statistics for the core measure-
ments show a polymodal distribution with mean Gaussian peak
values at 0.18 0.3 md (w11% of data), 3.4 0.3 md (w44% of
data), 24.2 0.3 md (w30% of data), and 87.5 0.5 md (w15% of
data). Best-t unimodal Gaussian curve provides a mean value of
7.96 0.7 md, about 68% of the undeformed sandstone mean
permeability (Fig. 13c).
10. Across-fault data variability
Five fault zones were selected for horizontal transects across the
fault strike. Each transect coordinate is centred on the master slip
surface in the fault core. Samples and permeability measurements
were collected in the same mechanical unit. Sample locations for
grain size analyses are spaced every 4080 cm, whereas perme-
ability measurement points are spaced 2050 cm apart in damage
zones and 210 cm apart in fault cores.
10.1. Granulometric transects
Granulometric parameter variations along two representative
transects are displayed in Fig. 14. All the statistical parameters
change systematically across the fault zones, indicating changes in
Fig. 11. Comparison of porosity (a) and macroporosity/microporosity ratio (b) among
undeformed, damage zone and fault core sandstones, respectively. Dark grey lines
indicate the average 2D porosity values in the undeformed (squares), damage zone
(triangles), and fault core (circles) domains. A pore area.
Fig. 12. Comparative shape analyses of undeformed (a), damage zone (b) and fault core (c) sand-size fractions, respectively. Larger black dots indicate the values of the largest
Gaussian peaks, whereas larger clear dots indicate the values of the secondary Gaussian peaks. Smaller clear dots indicate the values of third- and fourth-order Gaussian peaks.
the shape of the corresponding granulometric curves, which are
particularly dramatic in the fault cores. Mean grain size decreases
(i.e., Phi
m
increases) with deformation most dramatically in the
fault core, where it sharply reaches the lowest values. Fine-skewed
undeformed sandstones contrast with symmetrically skewed
sandstones in the damage zones. Approaching the fault core, grain
size distributions abruptly become strongly coarse skewed, indi-
cating a signicant increase in the ner grain size fractions (Fig. 14a
and b). The very poor sorting of undeformed samples tends to
improve slightly with deformation in damage zones, and strongly
improves in the intensely deformed localised slip surfaces within
the fault cores. Platykurtic undeformed samples change abruptly to
meso to leptokurtic in the fault cores; no such signicant modi-
cations are evident in the damage zones (Fig. 14c and d). The
percentage of residual clay-size fractions tends to increase locally in
the damage zones (due to secondary fault spays) and abruptly
increases in the fault cores, as much as to 7080% for the most
intensely sheared samples. Finally, the fractal dimension of about
2.7 in the undeformed sandstones gradually increases between 3
and 3.5 in the damage zones, then abruptly increases to values
higher then 4 in the fault cores (Fig. 14e and f).
10.2. Permeability transects
Permeability measurements along ve fault-perpendicular
transects show values ranging between 10 and 100 md in the
undeformed sandstones, and rather scattered and heterogeneous
values in damage zones, that are mostly higher than in the adjacent
undeformed sandstones and range between 10 and 1000 md
(Fig. 15). Fault core permeability is typically reduced up to 1 order of
magnitude from the average permeability of the host rock,
although locally, the difference may be less or slightly more
(compare values in Fig. 8c).
11. Fault displacement versus petrophysical properties in
fault cores
11.1. Granulometric data
The ratio between mean grain sizes in fault cores and corre-
sponding undeformed sandstones (Phi
m (core)
/Phi
m (und)
) generally
increases with increasing displacement indicating a general
decrease in the mean grain size in the faulted sandstones
(Fig. 16a). The only exception is represented by the outlier at about
20 m of displacement, which is characterised by higher commi-
nution intensity. Sorting tends to slightly deteriorate in two fault
cores, while in the others it slightly improves with increasing
displacement (Fig. 16b). Skewness ratios are generally negative
and are not signicantly inuenced by fault displacement
(Fig. 16c). On the other hand, the residual clay-size fraction ratios
increase up to about 40 m of fault displacement, with the anom-
alously high value at about 20 m (Fig. 16d). Finally, the fractal
dimension ratios (D
(core)
/D
(und)
) increase with increasing fault
displacement, showing an outlier value at about 12 m of fault
displacement (Fig. 16e).
11.2. Permeability
The fault core-undeformed sandstone ratios (K
(core)
/k
(und)
) are
generally lower than 1, and irregularly decrease with increasing
fault displacement, with the exception of a fault with a few centi-
metres of displacement that showed a slight increase in fault core
permeability. Overall, they showan exponentially decreasing trend
(Fig. 16f). As a rst approximation, where fault zones accommodate
displacements less than about 20 m, the average permeability
reductions in the fault cores are of one order of magnitude. Where
faults have displacements higher than about 20 m, the average
permeability reductions locally reach two orders of magnitude, the
only exception being an outlier (unlled dot in the plot) that relates
to an anomalously low permeability in the corresponding unde-
formed sandstones.
12. Discussion
12.1. Deformation mechanisms
Several lines of evidence indicate that deformation mecha-
nisms in fault core and damage zone evolve from distributed to
localised deformation during progressive sediment lithication.
Fig. 13. Comparative permeability analyses of undeformed (a), damage zone (b) and
fault core (c) sandstones, respectively. Best-t unimodal Gaussian curves for unde-
formed and fault core permeability are indicated with a dotted line. Polymodal best-t
Gaussian curves are superimposed on the corresponding data histograms.
The loose packing fabric observed in thin section (Figs. 5c and
6d), the slight particle size (Fig. 10) and shape (Fig. 12a and b)
variability, the relative abundance of sub-angular undamaged
coarser grains (Figs. 5c and 6), and the higher porosity and
asymmetric pore size distribution toward macropores (Fig. 8d),
support deformation by non-destructive particulate ow of
unlithied sediments (cf. Mandl et al., 1977; Maltman, 1994). In
the fault core, this inference is supported by the macroscopically
ductile distributed fabric (e.g. Rutter, 1986; Rawling and Good-
win, 2003) (Fig. 3). Effective uid circulation is inferred to facil-
itate particle translation, rotation and grain boundary sliding
with subordinated intragranular grain breakage (e.g. Sibson,
1977; Borradaile, 1981). Fluid-assisted environment is supported
by the presence of elongate, foliated sand blobs with cuspate
boundaries within the fault core (Fig. 3b and d) which are
interpreted to represent an early incorporation of unlithied
sand and sediment intermingling during fault slip. Non-
destructive particulate ow is interpreted to occur in the early
stages of the fault core evolution, whereas it persists throughout
most of the damage zone history.
The progressive grain size reduction with increasing fault
displacement (Fig. 16a) is interpreted to represent the progression
from particulate ow with little cataclasis to localised cataclastic
deformation in the fault core, particularly within the master slip
surfaces (Fig. 14). Cataclastic deformation in fault core rocks is also
conrmed by 1) the tight, matrix-supported foliated fabric (Fig. 6a
and b); 2) intragranular fragmentation (Fig. 6c); 3) the reduction in
porosity and pore size (Fig. 8c); 4) the occurrence of very angular
ner grains surrounding coarser ones (Fig. 12c); and 5) the increase
in clay-size fractions and fractal dimension compared to unde-
formed protoliths (cf. Sammis et al., 1987) (Fig. 16d and e). Com-
pactional cataclastic ow is inferred to occur with progressive
fragmentation of coarse grain at the impingement points (e.g.
Gallagher et al., 1974) and chipping during rolling and translation,
both rounding their initial angular shape and producing very
angular small akes. Elongate particles are generally reoriented
into the sense of shear (cf. Goodwin and Tikoff, 2002). Similar
results were found by previous workers in both rocks (e.g. Sammis
et al., 1987; Blenkinsop, 1991; Storti et al., 2003) and poorly lithied
sediments (cf. Heynekamp et al., 1999; Sigda et al., 1999; Cashman
Fig. 14. Granulometric transects across two fault zones with displacements of about 12 m (Fault 4, left column) and 49 m (Fault 8, right column) illustrating the trends of mean size
(Phi
m
) and skewness (a, b), sorting and kurtosis (c, d), and clay-size content and fractal dimension (e, f), with increasing deformation. See text for details.
and Cashman, 2000; Rawling and Goodwin, 2003, 2006; Cashman
et al., 2007).
The non-linear trends of granulometric parameter ratios with
respect to fault displacement (Fig. 16ae) indicate that the fault
rocks reach a level of grain size maturity beyond which few new
grains are fractured, with the larger survivor grains presumably
cushioned by the ner matrix material (e.g. Blenkinsop, 1991).
Outliers may derive from non-uniformly distributed slip and
compositional heterogeneities in fault cores. The change from
distributed to localised cataclastic deformation in the fault core
Fig. 15. Permeability transects across ve fault zones with displacement of about 35 m (a), 20 m (b), 12 m (c) 15 m (d) and about 50 m (e). Permeability tends to be higher in the
damage zones relative to undeformed sandstones, with the exception of secondary faults. Higher values were recorded across intergranular extensional fractures. In contrast,
permeability is locally reduced more than one order of magnitude in fault core rocks.
occurred during sediment lithication. This is conrmed by the
presence of relatively rigid and internally undeformed sand pods
(Fig. 3c) which supports an interpretation of tectonic entrain-
ment and mixing of partially lithied sand units during late stage
faulting (cf. Heynekamp et al., 1999; Rawling and Goodwin,
2006). A similar inference can be made for damage zones by the
mutually overprinting relationships between intergranular
extensional fracturing and Fe-oxide cementation in low-porosity
sands (Fig. 5df), indicating progressive selective sediment
strengthening during late stage faulting (cf. Caine and Minor,
2009).
The intergranular extensional fractures described in this study,
developed by separation of the crack walls with jig-saw t geom-
etry along grain boundaries, can be considered a transitional failure
mechanism between dilation bands of high-porosity sand(stone)s
and transgranular extensional fractures of low-porosity, competent
rocks.
12.2. Fault zone evolution
Field and laboratory analyses support a four-stage evolu-
tionary model, where faulting of quartz-dominated granular
material occurs during progressive Fe-oxide cementation under
low conning pressure (Fig. 17). Nucleation of syn-sedimentary
extensional shear zones in unlithied material slightly modies
the sedimentary fabric without signicantly changing the pris-
tine grain size distribution. At this stage, fault core and damage
zone are not yet well dened and deformation results in dila-
tional zones (Fig. 17a). Pore-size increase favours uid drainage
from undeformed sediments toward the shear zones, which can
undergo transient overpressuring facilitating shear failure
(Maltman, 1994). Fault zone architecture starts developing and
shearing in the fault core is accommodated by particulate ow
(cf. Borradaile, 1981; Rawling and Goodwin, 2003), possibly in
a water-saturated environment (Fig. 17b). Grain breakage at this
stage is negligible and distributed deformation in the damage
zone is mostly achieved by dilatancy. This results in a general
increase in bulk porosity and pore sizes that favour Fe-oxide
cementation, likely occurring in a vadose environment as indi-
cated by the patchily distribution of cement within the fault
zone and meniscus cement in the undeformed sands (e.g. Lucia,
2007).
Incipient lithication promotes more localised, cataclastic
deformation in the fault core (Fig. 17c) (cf. Lucas and Moore,
1986; Bernabe et al., 1992; Johansen et al., 2005; Rawling and
Goodwin, 2006; Balsamo et al., 2008). Fault core thickness
increases and porosity is reduced to less than 1%, favouring
sealing behaviour. Widespread outcrop-scale folds develop and
become disrupted during sediment mobilization (Maltman and
Bolton, 2003) in the damage zone, facilitated to the loosening of
the original interlocking of the densely packed sand grains. Non-
destructive dilatant granular ow occurs in heterogeneous zones
where ne-grained granular material is inferred to have moved
into dilatant areas during deformation. As a result, porosity
tends to increase by as much as 10%. Macroporosity exceeds
microporosity due to the connection and growth of dilated
pores.
As a consequence of progressive consolidation, deformation,
cementation, and the increase of conning stress (cf. Engelder,
1974; Marone and Scholz, 1989), granular material strength
Fig. 16. Trends of the statistical granulometric parameters (ae) and permeability reduction (f) in fault cores as a function of fault displacement. Each point in the graphs is given by
the ratio between the mean value of a given parameter in the fault core and the mean value of the same parameter in the undeformed counterpart. Grey points relate to the fault
zones which displacement values were inferred from the fault core thickness-displacement data in Fig. 4. Unlled point in (f) represents a fault core where the mean permeability
was similar to the anomalously, low permeable undeformed sandstone. Error bars represent the minimum and maximum values obtained from the corresponding end-member
parameters. See text for details.
improves during deformation. Fault core and damage zone
thicknesses continue increasing and incorporation of damage-
zone material into the fault core produces overall grain-scale
mixing which feeds the cataclastic process with new
undamaged grains (Fig. 17d). Macroporosity is dramatically
reduced by mechanical compaction and the production of ner
particles during grain fracturing and abrasion. In the damage
zone, non-destructive granular ow is inhibited by the resulting
increase in sediment strength, and intergranular extensional
fractures and dilation bands start to develop (Fig. 17d). Devel-
opment of opening-mode structures results in an effective
network of secondary fracture porosity (e.g. Laubach and Ward,
2006) which strongly increases macroporosity and secondary
pore connectivity.
The time and space heterogeneous occurrence of dilatancy
episodes and intergranular extensional fracturing in the damage
zone are possibly favoured by transient uid overpressuring pul-
ses during fault history, which may be triggered by ineffective
drainage during progressive burial (e.g. Maltman and Bolton,
2003), and by the improving hydraulic differentiation between
low-permeable fault core and high permeable damage zone (e.g.
Davis, 1999).
Fig. 17. Not to scale cartoon showing the proposed structural and hydraulic evolutionary model for extensional fault zones developed in low-porosity, quartz-dominated
sand(stone)s during progressive lithication. Grain-scale deformation mechanisms inferred to have operated during fault zone evolution are indicated in circular magnied views of
fault rocks. Images are drawn at the same scale. FC fault core; FWDZ footwall damage zone; HWDZ hanging-wall damage zone; d fault displacement. See text for details.
12.3. Fault zone hydraulic behaviour
The proposed evolutionary model implies that, when
displacement is very low (<w1 m), fault zones in low porosity,
poorly lithied sediments are dominated by dilatant structures
and, consequently, are expected to behave as preferential
conduits for uid ow (Fig. 17). As low porosity, low permeability
fault core rocks develop, they provide increasingly efcient
barriers to cross-fault uid ow, and the hydraulic behaviour of
fault zones progressively changes to conduit-barrier systems. In
map view, uid ow is enhanced along the fault strike of such
systems, because the hanging-wall and footwall damage zones in
which permeability is enhanced are separated by low perme-
ability fault cores. Only fault tip regions may behave as outward
migrating transient dilational conduits, for as long as the sand-
stones remain poorly cemented, such that particulate ow is
possible.
Increasing fault displacement up to tens of meters during
progressive sand lithication enhances the permeability contrast
between fault cores and damage zones. In particular, fault cores
suffer a strong permeability decrease by mechanical reduction of
grain size, pore aperture size, and porosity as a result of cataclasis
(e.g.; Pittman, 1981; Antonellini and Aydin, 1994; Sigda et al., 1999).
Concomitant burial and cementation reduces primary porosity in
damage zones. However, the related strength increase promotes
the formation of localised dilatant structures, resulting in a further,
more localised increase in sandstone permeability in the damage
zones (Fig. 17).
The hydraulic behaviour described above is typical of low-
porosity, well lithied rocks deforming by brittle fracturing (e.g.
Caine et al., 1996; Storti et al., 2003), whereas poorly lithied (e.g.
Rawling and Goodwin, 2003) and lithied high-porosity (e.g.
Antonellini and Aydin, 1994; Shipton et al., 2002) sands and
sandstones are typically considered to deformby deformation band
faulting, and, in many cases, both damage zones and fault cores
behave as hydraulic barriers, despite exceptions have been
described (e.g. Fossen et al., 2007). The outcomes of this study show
that high permeability damage zones are expected to form in
poorly lithied sandstones that have low initial porosity, implying
that faults in such rocks can channelize uid ow. These observa-
tions could help to explain the existence of preferential fault-
parallel ux in poorly lithied petroleum reservoirs (e.g. Silva et al.,
2008). Moreover, the somewhat atypical hydraulic behaviour of the
studied low-porosity sandstones conrms that initial porosity plays
a fundamental role in determining the deformational structures
form during fault zone evolution as proposed by Fossen et al.
(2007) and, hence, fault zone hydrology. This is supported by the
evidence that, for comparable displacement values, deformation in
the studied fault cores caused an average permeability decrease of
about two orders of magnitude, whereas in high-porosity sand-
stones the permeability reduction in fault cores is reported to reach
at least four orders of magnitude (Antonellini et al., 1994; Sigda
et al., 1999; Ogilvie and Glover, 2001; Shipton et al., 2002; Flodin
et al., 2005).
13. Conclusions
The results of our structural, microstructural and petrophysical
study of extensional fault zones in the continental Barreiras
Formation are summarised as follows:
1) Deformation in faults formed in low-porosity, quartz-domi-
nated, poorly lithied sandstones occurs initially by dilatant,
non-destructive distributed particulate ow, which evolves
with increasing slip and tectonic compaction to compactional
cataclastic ow and deformation on localised slip surfaces.
Angular original grains are progressively fragmented by
transgranular fracturing and abraded as grain vertexes chip
during grain rolling and translation. Rounded, coarse
survivor grains are eventually cushioned by more angular,
ner grained akes in a tightly packed foliated rock.
Progressive grain size reduction (to very ne silt) and the
consequent increase in the silt- and clay-size fractions
produce coarse-skewed granular material in which porosity
(up to 0.2%) and average fault core rock permeability are
reduced up to two orders of magnitude. Deformation in fault
damage zones occurs initially by dilatant granular ow and
secondary faulting, which do not signicantly change grain
size and shape distributions, but signicantly alter grain
packing. With increase sediment lithication, intergranular
extensional fractures and dilation bands develop and result
in an overall increase in porosity up to about 9% and pore
connectivity, thereby increasing damage zone permeability.
No deformation bands develop throughout the evolution of
the studied fault zones.
2) Fault zones in low-porosity, poorly lithied sandstones produce
hydraulic systems that evolve from conduits to barrier-conduit
systems with progressive sandstone lithication and increasing
slip. The permeability of fault cores decreases dramatically
with slip up to w20 m of fault displacement, after which it
tends to decrease at a lower rate during further slip. This
evolution in hydraulic behaviour is more similar to that of well
lithied brittle rocks than that of deformation band-domi-
nated, high-porosity sandstones.
3) The difference between fault core rock permeability and low-
porosity, poorly lithied sandstone protoliths is generally less
than that described for fault zones in high-porosity sand-
stones. This evidence further points to the crucial role of initial
porosity and grain size properties (such as mean size and
sorting) of host sandstones in predicting the structural
evolution and hydraulic properties of fault zones in clastic
material.
4) The documented trends between fault displacement and
petrophysical properties (e.g., grain size distributions and
permeability in fault core rocks), provide a useful tool for
predicting grain size distribution, permeability, and trans-
missibility of subsurface fault zones in low-porosity, quartz-
dominated sandstone uid reservoirs. Our study shows that,
given the enhanced porosity and permeability observed in the
damage zones and the reduced porosity and permeability
observed in fault cores, fault zone deformation in poorly
lithied, low-porosity sandstones produce barrier-conduit
systems, which can channelize ow parallel to faults, consis-
tent with observations in some deep-water Brazilian clastic
reservoirs.
Acknowledgements
This work was carried out under the framework of the TRAFUR
and TRAFUR2 (Transmissibility of Faults in Unconsolidated Rocks)
research projects, funded by Petrobras, SA. We gratefully
acknowledge Petrobras for releasing this material for publication.
We are extremely grateful to Laurel Goodwin and Scott Minor for
their constructive and helpful reviews that allowed us to signi-
cantly improve the nal manuscript. Francisco Hilario Bezerra is
kindly thanked for helpful discussion in the eld. Josias Barreto is
kindly thanked for logistic support in Icapu`. Isabela de Oliveira
Carmo, Maria Lima and Paulo Vasconcelos are kindly thanked for
discussions on age of Barreiras Formation.
Appendix
Table A1
Size-fraction frequencies and related grain size statistical parameters describing the mean size, sorting, skewness, kurtosis and fractal dimension of the analysed undeformed
samples. R
2
is the coefcient of determination of the fractal dimension D, i.e. the square of the coefcient of correlation.
Sample Gravel % Sand % Silt % Clay % Mean size Phi
m
Sorting So Skewness Sk Kurtosis Ku Fractal D R
2
2A-06 0.43 63.43 14.88 21.26 3.89 3.01 0.50 1.74 2.82 0.9671
2D-06 0.72 81.88 6.21 11.19 2.57 2.54 1.33 3.85 2.42 0.9711
4A-06 0.06 42.63 21.63 35.68 5.22 3.03 0.19 1.43 3.06 0.9852
4B-06 0.10 62.73 13.36 23.81 3.93 3.10 0.48 1.62 2.79 0.9745
4C-06 0.47 60.47 15.83 23.23 3.95 3.16 0.38 1.58 2.80 0.9845
4D-06 0.57 76.26 10.24 12.94 2.98 2.65 1.09 3.02 2.63 0.9724
4L-06 0.50 62.25 11.31 25.94 4.31 2.90 0.36 1.82 2.77 0.9732
4M-06 2.37 66.07 8.14 23.43 3.46 3.22 0.61 1.87 2.57 0.9794
5A-06 0.25 68.27 11.11 20.37 3.54 2.98 0.70 2.01 2.68 0.9774
8A-06 2.29 65.01 9.17 23.53 3.51 3.23 0.55 1.84 2.59 0.9867
8B-06 0.85 67.98 9.94 21.23 3.41 3.10 0.72 1.96 2.65 0.9787
8J-06 6.61 55.97 10.58 26.84 3.49 3.62 0.33 1.54 2.61 0.9949
8L-06 0.61 60.81 14.30 24.28 3.87 3.22 0.43 1.56 2.77 0.9814
BR1C 6.10 73.90 12.98 7.01 2.34 2.63 0.93 3.22 2.61 0.9915
BR1D 7.75 69.81 16.76 5.67 2.50 2.60 0.69 2.94 2.71 0.9908
BR8B 0.08 73.70 14.73 11.49 3.31 2.52 0.94 2.83 2.79 0.9724
BR12D 0.10 82.75 10.22 6.94 2.64 2.25 1.35 4.21 2.63 0.9686
mean 1.76 66.2 11.8 23.2 3.47 3.00 0.60 1.60 2.70 0.98
st dev 2.53 6.39 2.95 2.6 0.66 0.32 0.30 0.16 0.11 0.01
Table A2
Size-fraction frequencies and related statistical parameters describing the mean size, sorting, skewness and kurtosis of the damage zone samples. The fractal dimension of the
analysed samples is also indicated, as well as their structural position. (HWhanging-wall damage zone; ms mobilised sediments; sf secondary fault; ext intergranular
extensional fractures). R
2
is the coefcient of determination of the fractal dimension D, i.e. the square of the coefcient of correlation.
m
2
Structural position
2E 06 0.30 60.64 12.85 26.21 4.80 2.60 0.37 1.76 3.13 0.9632 HW-ms
4H-06 0.03 30.09 30.69 39.19 5.98 2.45 0.29 1.54 3.65 0.9678 HW
4I-06 0.00 74.93 7.67 17.40 3.87 2.37 1.26 2.80 3.52 0.9158 HW
4J-06 0.84 48.03 18.74 32.40 5.13 2.75 0.02 1.71 3.10 0.9738 HW-sf
5F-06 1.73 51.84 13.28 33.15 4.53 3.33 0.02 1.46 2.78 0.9895 HW-ms
5G-06 0.00 51.03 15.48 33.49 4.70 3.20 0.06 1.33 2.93 0.9759 HW-ms
8C-06 0.43 56.32 14.89 28.36 4.50 2.99 0.21 1.61 2.86 0.9812 FW
8D-06 0.32 49.74 17.41 32.52 4.80 3.07 0.05 1.44 2.92 0.9855 FW
8E 06 0.00 34.12 13.31 52.57 5.92 3.13 0.64 1.73 2.93 0.9884 FW
8F-06 0.34 38.57 22.26 38.82 5.59 2.86 0.34 1.62 3.17 0.9844 FW
BR12C 0.25 75.55 13.72 10.48 3.10 2.40 1.01 3.34 2.62 0.9682 HW-ext
mean 0.39 51.10 14.15 32.8 4.90 2.80 0.00 1.70 2.90 0.97
st dev 0.51 14.38 2.17 4.30 0.66 0.38 0.27 0.20 0.17 0.02
Table A3
Size-fraction frequencies and related statistical parameters describing the mean size, sorting, skewness and kurtosis of the fault core samples. The fractal dimension of the
analysed samples is also indicated, as well as their structural position within the fault core. R
2
is the coefcient of determination of the fractal dimension D, i.e. the square of the
coefcient of correlation.
m
2
Structural position
2B-06 0.00 10.67 15.35 73.98 7.56 1.92 1.98 5.64 3.28 0.9898 Foliated sand
2C-06 0.02 28.35 17.43 54.20 6.32 2.81 0.83 2.10 3.13 0.9879 Fault gouge
4E 06 0.15 38.58 19.36 41.92 5.60 2.92 0.34 1.57 3.07 0.9844 Foliated sand
4F-06 0.00 8.15 33.14 58.71 7.28 1.77 1.18 2.91 4.29 0.9801 Fault gouge
4G-06 0.02 28.54 24.58 46.86 6.18 2.66 0.64 1.89 3.31 0.9839 Foliated sand
5C-06 5.61 61.15 14.15 19.09 3.35 3.23 0.45 1.93 2.70 0.9921 Sand blob
5D-06 0.70 49.16 18.24 31.89 4.65 3.28 0.05 1.45 2.85 0.9944 Foliated sand
5E 06 0.00 41.38 17.82 40.80 5.48 2.97 0.28 1.48 3.04 0.9848 Fault gouge
8G-06 0.00 31.23 17.51 51.26 6.19 2.75 0.62 1.72 3.40 0.9584 Fault gouge
8H-06 0.00 2.59 19.49 77.92 8.01 1.22 3.34 15.44 3.85 0.9946 Fault gouge
8I-06 0.12 29.45 17.92 52.51 6.29 2.80 0.80 2.08 3.16 0.9865 Foliated sand
BR1A 1.14 70.00 23.15 5.71 3.37 2.21 0.62 3.09 2.92 0.9791 Foliated sand
BR1B 2.99 70.51 18.15 8.34 2.85 2.70 0.73 2.59 2.73 0.9922 Foliated sand
BR1E 1.10 65.06 25.93 7.91 3.61 2.41 0.49 2.55 2.97 0.9832 Foliated sand
BR12A 0.00 60.70 14.13 25.17 4.31 2.74 0.59 1.79 3.06 0.9252 Foliated sand
BR12B 0.59 24.85 41.18 33.38 5.65 2.41 0.24 2.21 3.31 0.9739 Fault gouge
Average 0.78 34.40 17.90 41.30 5.50 2.70 0.20 2.00 3.10 0.98
St dev 1.51 21.07 1.40 20.56 1.56 0.31 0.70 0.47 0.20 0.02
References
Anders, M.H., Wiltschko, D.V., 1994. Microfracturing, paleostress and the growth of
faults. Journal of Structural Geology 16, 795815.
Antonellini, M., Aydin, A., 1994. Effect of faulting on uid ow in porous sandstones;
petrophysical properties. American Association of Petroleum Geologists
Antonellini, M.A., Aydin, A., Pollard, D.D., 1994. Microstructure of deformation
Geology 16, 941959.
Antonellini, M., Aydin, A., Orr, L., 1999. Outcrop-aided characterization of a faulted
hydrocarbon reservoir: arroyo Grande Oil Field, California, USA. In:
Haneberg, W.C., Mozley, P.S., Moore, J.C., Goodwin, L.B. (Eds.), Faults and
Subsurface Fluid Flow in Shallow Crust. Geophysical Monograph, vol. 113.
American Geophysical Union, pp. 726.
Aydin, A., 1978. Small faults formed as deformation bands in sandstone. Pure and
Aydin, A., 2000. Fractures, faults, and hydrocarbon entrapment, migration and ow.
Aydin, A., Johnson, A.M., 1978. Development of faults as zones of deformation bands
and as slip surfaces in sandstone. Pure and Applied Geophysics 116, 931942.
Balsamo, F., Storti, F., Piovano, B., Cifelli, F., Salvini, F., Lima, C.C., 2008. Time
dependent structural architecture of subsidiary fracturing and stress pattern in
the tip region of an extensional growth fault system, Tarquinia Basin, Italy.
Bernabe, Y., Fryer, D.T., Hayes, J.A., 1992. The effect of cement on the strength of
granular rocks. Geophysical Research Letters 19, 15111514.
Bezerra, H.F.R., Amaro, V.E., Vita-Finzi, C., Saadi, A., 2001. Pliocene-Quaternary fault
control of sedimentation and coastal plain morphology in NE Brazil. Journal of
South American Earth Science 14, 6175.
Bezerra, H.F.R., Vita-Finzi, C., 2000. How active is a passive margin? Paleoseismicity
in northeastern Brazil. Geology 28, 591594.
Billi, A., Salvini, F., Storti, F., 2003. The damage zone-fault core transition in
carbonate rocks: implications for fault growth, structure and permeability.
Blenkinsop, T.G., 1991. Cataclasis and processes of particle size reduction. Pure and
Borg, I., Friedman, M., Handin, J., Higgs, D.V., 1960. Experimental deformation of St.
Peter Sand: a study of cataclastic ow. Geological Society of America Memoir
79, 133191.
Borradaile, G.J., 1981. Particulate ow of rock and the formation of cleavage. Tec-
tonophysics 72, 305321.
Butler, B.C., Bell, J.D., 1989. Interpretation of Geological Maps, Longman Earth
Science Series.
Caine, J.S., Minor, S.A., 2009. Structural and geochemical characteristics of faulted
sediments and inferences on the role of water in deformation, Rio Grande Rift,
New Mexico. Geological Society of America Bulletin 121, 13251340.
Cashman, S., Cashman, K., 2000. Cataclasis and deformation-band formation in
unconsolidated marine terrace sand, Humboldt County, California. Geology 28,
111114.
Cashman, S.M., Baldwin, J.N., Cashman, K.V., Swanson, K., Crawford, R., 2007.
Microstructures developed by coseismic and aseismic faulting in near-surface
sediments, San Andreas fault, California. Geology 35, 611614.
Chester, F., Evans, J., Biegel, R., 1993. Internal structure and weakening mecha-
nisms of the San Andreas fault. Journal of Geophysical Research 98 (B1),
771786.
Choquette, P.W., Lloyd, C.P., 1970. Geologic nomenclature and classication of
porosity in sedimentary carbonates. American Association of Petroleum Geol-
ogists Bulletin 54, 207244.
Davis, G.H., 1999. Structural geology of the Colorado plateau region of southern
Utah, with special emphasis on deformation bands. Geological Society of
America Special Papers 342, 1157.
Du Bernard, X., Eichhubl, P., Aydin, A., 2002. Dilation bands: a new form of localized
failure in granular media. Geophysical Research Letters 29 (24), 2176.
doi:10.1029/2002GL015966.
Dunn, D.E., LaFountain, L., Jackson, R.E., 1973. Porosity dependence and mechanism
of brittle fracture in sandstones. Journal of Geophysical Research 78, 24032417.
Engelder, J.T., 1974. Cataclasis and the generation of fault gouge. Geological Society
of America Bulletin 85, 15151522.
Ferreira, J.M., Assumpao, M., 1983. Sismicidade no Nordeste do Brasil. Revista
Brasileira de Geosica 1, 6788.
Ferreira, J.M., Oliveira, R.T., Takeya, M.K., Asssumpao, M., 1998. Superposition of local
andregional stresses innortheast Brazil: evidencefromfocal mechanisms around
the Potiguar marginal basin. Geophysical Journal International 134, 341355.
reservoirs of the North Sea and Norwegian Continental Shelf. Marine and
Flodin, E., Prasad, M., Aydin, A., 2003. Petrophysical constraints on deformation
styles in Aztec sandstone, southern Nevada, USA. Pure and Applied Geophysics
160, 15891610.
Flodin, E.A., Gerdes, M., Aydin, A., Wiggins, W.D., 2005. Petrophysical properties of
cataclastic fault rock in sandstone. In: Sorkhabi, R., Tsuji, Y. (Eds.), Faults, Fluid
Flow, and Petroleum Traps. American Association of Petroleum Geologists
Memoir, vol. 85, pp. 197217.
Folk, R.L., Ward, W.C., 1957. Brazos river bar (Texas); a study in the signicance of
grain size parameters. Journal of Sedimentary Research 27, 326.
Fossen, H., Schultz, R., Shipton, K.Z., Mair, K., 2007. Deformation bands in sandstone
a review. Journal of the Geological Society 164, 755769.
Fossen, H., Hesthammer, J., 1997. Geometric analysis and scaling relations of
deformation bands in porous sandstone. Journal of Structural Geology 19,
14791493.
Fowles, J., Burley, S., 1994. Textural and permeability characteristics of faulted, high
porosity sandstones. Marine and Petroleum Geology 11, 608623.
Francus, P., 1998. An image-analysis technique to measure grain-size variation in
thin sections of soft clastic sediments. Sedimentary Geology 121, 289298.
Francus, P., Pirard, E., 2004. Testing for sources of errors in quantitative image
analysis. In: Francus, P. (Ed.), Image Analysis, Sediments and Paleoenviron-
ments, vol. 7. Kluwer Academic Publisher, pp. 87102.
Gallagher, J.J., Friedman, M., Handin, J., Sowers, G.M., 1974. Experimental studies
relating to microfracture in sandstone. Tectonophysics 21, 203247.
Gibson, R.G., 1998. Physical character and uid-ow properties of sandstone-
derived fault zones. In: Coward, M.P., Johnson, H., Daltaban, T.S. (Eds.), Struc-
tural Geology in Reservoir Characterization. Geological Society of London,
Special Publication, vol. 127, pp. 8397.
Goodwin, L.B., Tikoff, B., 2002. Competency contrast, kinematics, and the devel-
opment of foliations and lineations in the crust. Journal of Structural Geology
24, 10651085.
Grifths, J.C., 1952. Grain-size distribution and reservoirrock characteristics.
Haneberg, W.C., 1995. Steady-state groundwater ow across idealized faults. Water
Resources Research 31, 18151820.
Heilbronner, R., Keulen, N., 2006. Grain size and grain shape analysis of fault rocks.
Mexico: implications for fault-zone permeability and uid ow. In:
Haneberg, W.C., Mozley, P.S., Moore, J.C., Goodwin, L.B. (Eds.), Faults and
Subsurface Fluid Flow in Shallow Crust. Geophysical Monograph, vol. 113.
American Geophysical Union, pp. 2749.
Inman, D.L., 1952. Measures for describing the size distribution of sediments.
Journal of Sedimentary Research 22, 1257145.
example from the Moab Fault, Utah. Journal of Structural Geology 27,
14691485.
Kim, Y.S., Peacock, D.C.P., Sanderson, D.J., 2004. Fault damage zones. Journal of
Krumbein, W.C., 1934. Size frequency distributions of sediments. Journal of Sedi-
mentary Research 4, 6577.
Krumbein, W.C., 1938. Size frequency distributions of sediments and the normal Phi
curve. Journal of Sedimentary Research 3, 8490.
Krumbein, W.C., 1941. Measurement and geological signicance of shape and
roundness of sedimentary particles. Journal of Sedimentary Research 11, 6472.
Krumbein, W.C., Pettijohn, F.J., 1938. Manual of Sedimentary Petrography. Appleton-
Century Company, New York.
Kwon, Q., Ngwenya, B.T., Main, I.G., Elphick, S.C., 2005. Permeability evolution
during deformation of siliciclastic sandstones from Moab, Utah. In: Sorkhabi, R.,
Tsuji, Y. (Eds.), Faults, Fluid Flow, and Petroleum Traps. American Association of
Petroleum Geologists Memoir, vol. 85, pp. 219236.
Laubach, S.E., Ward, M.E., 2006. Diagenesis in porosity evolution of opening-mode
fractures, middle Triassic to lower Jurassic La Boca Formation, NE Mexico.
Lima, M., 2008. A histo ria do Intemperismo na provncia Borborema Oriental,
Nordeste do Brasil: implicaoes paleoclima ticas e tectonicas. 2008. PhD thesis,
Universidade Federal do Rio Grande do Norte, Brazil.
Lucas, S.E., Moore, J.C., 1986. Cataclastic deformation in accretionary wedges. Deep
Sea Drilling project Leg 166, southern Mexico, and on-land examples from
Barbados and Kodiak Islands. In: Moore, J.C. (Ed.), Structural Fabrics in Deep Sea
Drilling Project Cores from Forearcs. Geological Society of America Memoir, vol.
166, pp. 89104.
Lucia, F.J., 2007. Carbonate Reservoir Characterization. An Integrated Approach.
Springer, Berlin.
Mabesoone, J.M., Campos, E., Silva, A., Beurlen, K., 1972. Estratigraa e origem do
Grupo Barreiras em Pernambuco, Paraba e Rio Grande do Norte. Revista Bra-
sileira de Geociencia 2, 173190.
Mair, K., Main, I., Elphick, S., 2000. Sequential growth of deformation bands in the
laboratory. Journal of Structural Geology 22, 2542.
Maltman, A., 1994. Introduction and overview. In: Maltman, A.J. (Ed.), The
Geological Deformation of Sediment. Chapman & Hall, London, pp. 135.
Maltman, A.J., Bolton, A., 2003. How sediments become mobilized. In: Geological
Society of London Special Publications, vol. 216, pp. 920.
Mandl, G., de Jong, L.N.J., Maltha, A., 1977. Shear zones in granular material. Rock
Mechanics and Rock Engineering 9, 95144.
Manzocchi, T., Ringrose, P.S., Underhill, J.R., 2002. Flow through fault systems in
high-porosity sandstones. In: Holdsworth, R.E., Turner, J.P. (Eds.), Extensional
Tectonics: Faulting and Related Processes. The Geological Society of London,
Key Issues in Earth Sciences, vol. 2, pp. 281298. 2.
simulated fault gouge. Journal of Structural Geology 11, 799814.
de Matos, R.M.D., 1992. The northeast Brazilian rift system. Tectonics 11, 766791.
McManus, J., 1988. Grain Size Determination and Interpretation. Techniques in
Sedimentology. Backwell, Oxford.
Mene ndez, B., Zhu, W., Wong, T.F., 1996. Micromechanics of brittle faulting and
cataclastic ow in Berea sandstone. Journal of Structural Geology 18, 116.
Minor, S.A., Hudson, M.R., 2006. Regional survey of structural properties and cemen-
tation patterns of fault zones in the northern part of the Albuquerque basin, New
Mexico Implications for ground-water ow, U.S.G.S. Professional paper 1719.
Nelson, S.T., Mayo, A.L., Gilllan, S., Dutson, S.J., Harris, R.A., Shipton, Z.K.,
Tingey, D.G., 2009. Enhanced fracture permeability and accompanying uid
ow in the footwall of a normal fault: the Hurricane fault at Pah Tempe hot
springs, Washington County, Utah. Geological Society of America Bulletin 121,
236246.
Ogilvie, S.R., Glover, P.W.J., 2001. The petrophysical properties of deformation bands
in relation to their microstructures. Earth and Planetary Science Letters 193,
129142.
Passchier, C.W., Trouw, R.A.J., 1996. Microtectonics. Springer, Berlin.
Pittman, D.E., 1981. Effect of fault-related granulation on porosity and permeability
of quartz sandstones, Simpson Group (Ordovician), Oklahoma. American
Rawling, G.C., Goodwin, L.B., Wilson, J.L., 2001. Internal architecture, permeability
structure, and hydrologic signicance of contrasting fault-zone types. Geology
29, 4346.
Rawling, G.C., Goodwin, L.B., 2003. Cataclasis and particulate ow in faulted, poorly
lithied sediments. Journal of Structural Geology 25, 317331.
USA. Journal of Structural Geology 28, 16231639.
Rotevatn, A., Fossen, H., Hesthammer, J., Aas, T.E., Howell, J.A., 2007. Are relay ramps
conduits for uid ow? Structural analysis of a relay ramp in Arches National Park,
Utah. In: Geological Society, London Special Publications, vol. 270, pp. 5571.
Rutter, E.H., 1986. On the nomenclature of mode of failure transitions in rocks.
Storti, F., Billi, A., Salvini, F., 2003. Particle size distributions in natural carbonate
fault rocks: insights for non-self-similar cataclasis. Earth and Planetary Science
Letters 206, 173186.
Salvini, F., 2004. Daisy 4.1 Software The Structural Data Integrated System
Analyzer. Available at. University of Roma Tre. http://host.uniroma3.it/progetti/
fralab.
Sammis, C., King, G., Biegel, R., 1987. The kinematics of gouge deformation. Pure and
Selley, R.C., 2000. Applied Sedimentology. Academic Press, San Diego.
Geology 23, 18251844.
Shipton, Z.K., Evans, J.P., Robeson, K.R., Forster, C.B., Snelgrove, S., 2002. Structural
heterogeneity and permeability in faulted Eolian Sandstone: implications for
subsurface modeling of faults. American Association of Petroleum Geologists
Society 133, 191213.
Sigda, J.M., Goodwin, L.B., Mozley, P.S., Wilson, J.L., 1999. Permeability alteration in
small-displacement faults in poorly lithied sediments: Rio Grande Rift, Central
New Mexico. In: Haneberg, W.C., Mozley, P.S., Moore, J.C., Goodwin, L.B. (Eds.),
Faults and Subsurface Fluid Flow in Shallow Crust. Geophysical Monograph, vol.
113. American Geophysical Union, pp. 5168.
Silva, A.T., Lima, C.C., Faerstein, M., Salvini, F., Balsamo, F., 2008. Geomechanical
model helps to explain unexpected high well productivity offshore Brazil. In:
International Geological Congress, Oslo (abstract).
Suguio, K., Nogueira, A.C.R., 1999. Revisao crtica dos conhecimentos geolo gicos sobre
a Formaao (ouGrupo?) Barreiras do Neo geno e o seu possvel signicado como
testemunho de alguns eventos geolo gicos mundiais. Geociencias 18, 461479.
Vernik, L., Bruno, M., Bovberg, C., 1993. Empirical relations between compressive
strength and porosity of siliciclastic rocks. International Journal of Rock
Mechancs and Mining Science 30, 677680.
Walsh, J.J., Watterson, J., Heath, A., Gillespie, P.A., Childs, C., 1998. Assessment of
the effects of sub-seismic faults on bulk permeabilities of reservoir sequences.
In: Coward, M.P., Johnson, H., Daltaban, T.S. (Eds.), Structural Geology in
Reservoir Characterization. Geological Society of London, Special Publication,
vol. 127, pp. 99114.
Walsh, J.J., Watterson, J., 1988. Analysis of the relationship between displacements
and dimensions of faults. Journal of Structural Geology 10, 239247.
Wise, D.U., Funiciello, R., Parotto, M., Salvini, F., 1985. Topographic lineament
swarms: clues to their origin from domain analysis of Italy. Geological Society of
America Bulletin 96, 952967.
Wong, T., David, C., Zhu, W., 1997. The transition from brittle faulting to cataclastic
ow in porous sandstones: mechanical deformation. Journal of Geophysical
Research 102 (B2), 30093025.
Zhu, W., Wong, T., 1997. The transition from brittle faulting to cataclastic ow:
permeability evolution. Journal of Geophysical Research 102 (B2), 30273041.
Buoyancy driven ow from a waning source through a porous leaky aquifer
Andrew W. Woods
a,
*
, Simon Norris
b
a
BP Institute, University of Cambridge, Cambridge, CB3 OEZ, United Kingdom
b
Nuclear Decomissioning Authority, Harwell, Oxfordshire
Article history:
21 July 2009
a b s t r a c t
We develop a series of models to describe the migration of a buoyant uid through a layered permeable
rock following release from a localized waning source. In particular, if the uid is injected into a high
permeability layer, bounded above by a layer of lower permeability, a plume migrates along the interface,
with some draining into the low permeability layer if the current is sufciently deep to overcome the
capillary entry pressure. We show the motion of the uid is controlled by a number of key factors with
the dominant dimensionless numbers being the time-scale for the source to decay compared to the time-
scale for draining through the low permeability layer, the residual saturation of the gas and water in the
formation as that phase is displaced by the other phase, and the capillary entry pressure, as measured by
the critical depth of the current required for draining, as compared to the initial depth of the current.
Simplied analytical models are presented to illustrate some of the key controls on and transitions in the
ow, and the models are used to explore leakage and trapping prior to ow reaching a fault zone.
1. Introduction
There is growing interest in the migration of gas froma localized
source through a permeable rock owing to its relevance for the
dispersal of CO
2
sequestered in the subsurface, but more generally
for the dispersal of buoyant uids which may form in and migrate
from geological waste repositories (Bickle et al., 2007; Hesse et al.,
2007). In this latter situation, an interesting feature is that the rate
of generation of buoyant uid may progressively decay over
a period of several hundred years. The gradual waning of the source
leads to some interesting dynamical balances in the migrating
buoyant plume, and has a critical impact on the dispersal pattern of
the uid; this forms the topic of the present contribution.
Several models have been developed to describe the motion of
buoyant plumes of uid migrating through the subsurface (Bare-
nblatt, 1996; Bear, 1972; Huppert and Woods, 1995), and recently
these have been extended to include the equations for two-phase
ow in a permeable rock, including the effects of the relative
permeability between the two phases (Hesse et al., 2006; Nord-
botten and Celia, 2006). With a localized source of buoyant uid,
these models lead to the prediction that in a conned aquifer,
a buoyant plume develops adjacent to the upper boundary of the
aquifer and then spreads out along the aquifer (Hesse et al., 2006,
2008; Mitchell and Woods, 2006). If it comes into contact with
a fault/fracture system, then some of the ow may drain upwards
along the fault where it may then intersect another permeable
layer, enabling part of the leakage ux to continue spreading
laterally (Pritchard et al., 2001; Pritchard, 2007). If the source
wanes, the continuing nite plume will then develop a trailing
front. As this advances through the formation, there may be some
capillary trapping of the uid leading to a residual saturation
(Barenblatt, 1996; Hesse et al., 2006; Obi and Blunt, 2006; Kharaka
et al., 2006; Farcas and Woods, 2009a). As a result, the plume
becomes progressively depleted as it migrates through the forma-
tion, leaving the capillary trapped zone behind.
In the present contribution, we examine the motion of
a buoyant plume supplied by a waning source which spreads
through a permeable rock, bounded above by a less permeable thin
layer into which the uid may slowly drain off. We also account for
the capillary retention of a fraction of the uid at any point along
the plume where the owthickness decreases in time. This leads to
predictions of the fraction of the current which may remain trap-
ped in the original layer rather than leaking off higher into the
formation.
We note that our analysis is restricted to a two-dimensional
ow, in order to identify some of the key controls on the system,
although we note that three-dimensional cross-ow effects can
also arise, especially far upslope of the source (cf. Vella and Hup-
pert, 2007; Farcas and Woods, in press). However, with a long linear
source, the effects of three-dimensional spreading of the ow
E-mail address: andy@bpi.cam.ac.uk (A.W. Woods).
doi:10.1016/j.jsg.2009.08.018
upslope of the source to points beyond the extremities of the
source, may only become dominant once the ow has advanced
a substantial distance upslope, and so in that case, the present
modelling may provide a reasonable approximation to the ow in
the near eld. Also, in some situations, the ow may be structurally
conned such that the two-dimensional model may provide
a reasonable leading order model for the ow.
2. The model
We consider the migration of a buoyant uid of density r
through a permeable layer of rock, saturated with uid of density
r Dr, which is bounded above by a thin layer of lower perme-
ability. We assume the injected uid is only able to invade the low
permeability layer if it is sufciently deep, h(x,t), to overcome the
capillary entry pressure, h >h
c
. Otherwise it will continue to run
upslope through the high permeability layer, under the lower
boundary of this seal layer (Fig. 1). As the current spreads out along
this layer, of inclination to the horizontal q, the alongslope motion
is governed by the buoyancy forces acting on the ow, according to
the relation for the transport or Darcy ux m (cf Bear, 1972; Bare-
nblatt, 1996)
u
kDrg
m
vh
vx
cos q sin q
(1)
where h(x,t) is the thickness of the current, x is the alongslope
position, m is the viscosity and k is the effective permeability of the
uid as it migrates through the rock, where we account for
the effects of relative permeability in a very simple fashion with the
single permeability parameter. Most of the interest in this work is
in modelling gas or supercritical uid dispersion and we model the
motion through the rock in terms of an effective permeability. This
has been shown to give good leading order predictions compared to
the full two-phase ow relations for such buoyancy driven ows
(cf. Nordbotten and Celia, 2006; Hesse et al., 2006). The rate of loss
of uid from the current to the overlying low permeability layer,
through unit length of the boundary, depends on the permeability
k
b
, the thickness b of the seal layer, and the thickness of the current
(cf. Pritchard et al., 2001) according to the relation
Loss
k
b
gDrh bcos q
bm
(2)
These equations are then combined with the relation for the
conservation of mass, which in the invading ow has the form
f1 s
w
1 R
vh
vt

v
vx
hu Loss for
vh
vt
> 0 (3)
since as the invading gas advances into the formation, there is
a fraction s
w
of the pore space which remains saturated in the
original uid, and gas then dissolves into this uid, representing an
effective additional pore volume fs
w
R for the injected uid. Here
R denotes the mass fraction of gas dissolved in the original uid
(which occupies the fraction s
w
of the pore volume) multiplied by
the density of the original uid and divided by the density of the
free gas phase. Also, in this expression f denotes the porosity of the
rock.
In contrast, in any part of the ow where the depth of the
current decreases with time, then as the buoyant uid vacates the
pore space and is displaced with water, there will be some residual
gas trapped which occupies a fraction s
g
of the pore spaces. Here,
for simplicity, we model this as being a constant (cf. Barenblatt,
1996; Hesse et al., 2006, 2008) and so the conservation of mass
takes the form
f
1 s
w
s
g
vh
vt

v
vx
hu Loss for
vh
vt
< 0 (4)
In order to solve for the motion of the current we require some
boundary conditions. First, it follows that the nose of the current
propagates at the rate
dx
dt

u
f1 s
w
Rs
w
(5)
while we assume that the source ux, at x 0, gradually wanes, at
a rate
Qt Q
o
expt=s
kDrgsin q
m
1 cot q
vh0; t
vx
h0; t
(6)
Fig. 1. Cartoon of the ow geometry for the present problem.
A.W. Woods, S. Norris / Journal of Structural Geology 32 (2010) 18271833 1828
where t is the e-folding time over which the source ux decays. It is
this waning source ux, coupled with the dynamics of continuous
leakage of uid through the overlying seal rock, and the capillary
retention at the tail of the current, which provides the newanalytic
results of this paper.
From eq. (6), we deduce that there is no drainage if the source
ux Q
o
is smaller than a critical value
Q
o
<
h
c
gDrsin q
m
Qcrit (7)
3. Approximations and analytical solutions
With this system of equations, we can now develop a series of
solutions for the motion of the plume of gas along the inclined low
permeabilitylayer. Thesesolutions areuseful for exposingsomeof the
key controls on the distance travelled along the layer, and also how
the current partitions between that component which is retained in
the original layer and that component which migrates through the
low permeability partial seal layer and higher into the formation.
Before developing solutions for the motion, there are some
simplications which we can introduce which simplify the analysis,
and allow for an approximate analytical solution. First, as the
current spreads out and disperses into a relatively long and thin
ow, of lateral scale L and depth H say, such that L >Hcot q, then
the alongslope component of gravity, proportional to sin q, domi-
nates the force associated with the cross-slope component of
gravity which acts on variations in the alongslope depth of the
current and is proportional to cos q vh=vx . In this limit, h cot q <L,
the dynamical term proportional to cos qv=vxhvh=vx,which arises
in the rst term on the right had side of eqs. (3) and (4), as may be
inferred by combining these eqns with eq. (1), can be neglected.
Indeed, we demonstrate that our analytical solutions are consistent
with this approximation.
Also, in the limit that the critical current depth required for
draining, h
c
, satises h
c
>b, the depth of overlying lowpermeability
layer, then the numerator (h b) in the loss term can be approxi-
mated by h (>h
c
) since the loss only arises if the ow is sufciently
deep to overcome the capillary entry pressure (cf. Woods and Far-
cas, 2009a).
It is also convenient to introduce the scaling for the speed
U
kDrgsin q
mf
1 s
w
s
g
; (8)
the dimensionless ratio of the speed of the front and the tail of the
current, as given by
l
1 s
w
1 R
1 s
w
s
g
(9)
and the inverse of the time-scale for the draining ux
b U
k
b
cos q
kbsin q
(10)
With a waning source, these approximations lead to the governing
equations
l
vh
vt
U
vh
vx
bh for h > h
c
and
vh
vt
> 0 (11)
l
vh
vt
U
vh
vx
for h < h
c
and
vh
vt
> 0 (12)
and
vh
vt
U
vh
vx
bh for h > h
c
and
vh
vt
< 0 (13)
vh
vt
U
vh
vx
for h < h
c
and
vh
vt
< 0 (14)
with the boundary conditions that at the source,
1 s
w
s
g
Uh0; t Q
o
expt=s (15)
and that at the nose of the current, x x
n
(t),
hx
n
; t 0 and
dx
n
dt

U
l
(16)
It may be seen from the denition of l that l >1, and so the
advection speed of the current in the region in which the current is
invading new rock, U/l, is slower than the advection speed of the
current in the region in which it is receding from the rock, U. This
means that the nose of the current in which the depth decreases
from a maximum to zero, occurs across a localized region whose
detail depends on the cross-slope component of gravity. In this
simplied model, this is represented by a localised front at x x
n
(t).
The structure of the current behind this front depends on the
source ow rate compared to the draining rate, and we now
consider a range of cases in turn.
3.1. Small supply ux with no leakage current
If Q<Q(crit), then there is no drainage into the overlying layer,
and as the current moves forward along the boundary, the source
ux gradually wanes, leading to a waning plume. In this case, with
the draining term neglected, the solution of the equation for the
current depth, eq. (14), can be written in the form
hx; t h0; 0exp

t
s

x
Us
(17)
It follows that surfaces of constant depth advance forward at
a speed given by U (eg see Fig. 2 below). This is faster than the speed
of the leading front, U/l, and so the leading edge of the current
gradually becomes shallower with time. This is a result of the loss of
uid through capillary retention as the depth of the current at
a given point in space behind the leading front gradually decreases
in time.
Indeed, by direct substitution, it follows that the depth of the
current at the leading edge, x
n
Ut/l, has the form
distance
time
x=Ut /
dx/dt=U
Depth remains
constant along
Leading edge
Fig. 2. illustrates the evolution of the ow in terms of the motion of surfaces of
constant depth in the xt plane.
hUt=l; t h0; 0exp

l 1t
ls
(18)
We can also calculate the area of the zone of the rock which is
invaded by the current. This can be used to estimate the volume of
the residual uid which is trapped in the pore space once the
current recedes, although we note that, in time, this trapped uid
may dissolve into the water, and be carried off by any hydrological
ow. The solution above for the shape of the current as a function of
distance and time illustrates that at each point in space, the current
is deepest on rst arriving at that point. The current rst arrives at
each point x after a time lx/U (Fig. 2), and so the maximum depth of
the current at a distance x from the source, h
max
(x), is
h
max
x h0; 0exp

xl 1
Us
(19)
This curve describes the locus of the zone in which there may be
some residual plume uid once the plume has drained and moved
on, and hence in which there may be a possible source of
contaminant in a subsequent hydrological ow. In Fig. 3 below, we
illustrate the envelope of the zone contaminated with gas and
compare this with the instantaneous proles of the buoyant plume
at different times.
3.2. Larger source ux and drainage
With a larger source ux, Q>Q(crit), then initially there will be
some drainage into the overlying layer, with h >h
c
. In the region of
the current where h >h
c
the solution may be written in the form,
hx; t h0; 0exp

t
s

x
Us
1 bs
(20)
and the rate of propagation of surfaces of constant depth is now
faster than in the case with no draining, as the uid leaks off
through the overlying layer. We will now see that these new
solutions are very different fromthe case with no draining (sect 3.1)
3.2.1. Slow draining or rapid decay of the source ux
In the case 1 >bs, the depth of the current h increases with
distance from the source at a given time since the draining of the
uid is slow compared to the decay of the source and hence the
uid at the source has the smallest ux and so is shallowest; as
a consequence, the depth rst decreases to value h h
c
at x 0
when t t
c
, as given by (see Fig. 4 and 5)
t
c
s ln
h
c
h0; 0
(21)
Subsequently, as the ux continues to wane, the depth of the
current near the source decreases to values h <h
c
(Fig. 5) and the
location of the point at which the depth has value h
c
migrates away
from the source and has position x x
c
(t). In the region 0 <x <x
c
,
the current does not drain since h <h
c
. As the zone in which h <h
c
advances outwards from the source, the leading part of this region,
where h h
c
has position given by (Fig. 4)
x
c
t
Ut t
c
1 bs
(22)
Meanwhile the leading edge of the current has position x
n
Ut/l
(cf. eq. (16)), and so the zone in which draining occurs advances
progressively further from the source. Eventually, the front
x x
c
(t),which represents the closest point to the source at which
the depth has value h
c
and hence can drain, reaches the leading
edge of the ow. This occurs at time
t t
d
t
c
1
bs
l

1
l
1
(23)
Subsequently, there is no more draining anywhere in the ow
(Fig. 4). For times t >t
c
the closest point to the source at which the
depth has value h
c
is given by the front x x
b
(t) where
X
b
t Ut t
c
(24)
This lags behind the front x x
c
, and between these fronts, in
the region x
b
<x <x
c
, the depth has the constant value h
c
but there
is no draining. In the near source region, 0 <x <x
b
, in which the
current depth h <h
c
, the plume has shape (cf. eq. (17) and Fig. 4)
hx; t h0; 0exp

t
s

x
Us
(25)
In this near source region, the depth increases with distance
from the source, and reaches the critical depth h h
c
at the point
x
b
(t). Eventually, at time t
b
, the nearest point to the source at which
the current increases to depth h
c
, as given by x x
b
, reaches the
leading edge of the current, so that x
b
(t) x
n
(t). This occurs at time
t t
b
given by
t
b
lt
c
=l 1 (26)
0
0.2
0.4
0.6
0.8
1
1.2
0.00 0.50 1.00 1.50
d
i
m
e
n
s
i
o
n
l
e
s
s

h
e
i
g
h
t
dimensionless distance from source
time=0.5 1.0 1.5
Fig. 3. Comparison of the envelope of the zone invaded by the injected uid and the
instantaneous shape of the injected uid plume at times 0.5s, 1.0s and 1.5s.
Fig. 4. (x,t) plot to illustrate the evolution of the fronts x x
b
, x
c
and x
n
as they evolve
with time in the current. At times earlier than t
c
the current is deeper than h
c,
but for
times greater than t
b
the whole current is thinner than the critical depth h
c.
For
intermediate times, the near source region, x <x
b
, is shallower than h
c
while the more
distal parts of the current are either of depth h
c
, for x
b
<x <x
c
or of depth greater than
h
c
for x
n
>x >x
c
.
Subsequently the current is described by relation (17), and is
everywhere shallower than h
c
. This sequence of ow regimes is
illustrated in the Fig. 4 and 5 shown below.
3.2.2. Fast draining or slow decay of the source
In the case that 1 <bs, and with Q>Q
c
, the current is initially
deeper than the critical value for draining, h >h
c
, and the ow
initially advances with prole
hx; t h0; 0exp

t
s

x
us
1 bs
(27)
in the region 0 <x <Ut/l (Fig. 6). In this case, at a given time, the
current becomes shallower with distance from the source, as
a result of the draining occurring more rapidly than the rate of
decay of the source, so that the owfurther fromthe source has less
ux than that at the source (Fig. 7, dashed lines). As a result, the
leading front of the current eventually reaches the critical depth at
which draining ceases, h h
c
. This occurs when
t t
d
t
c
1
bs
l

1
l
1
(28)
Subsequently, the leading edge of the current continues forward
with depth h h
c
while the closest point to the source at which the
depth of the current h h
c
, as given by x x
c
, migrates backwards
towards the source according to the relationship (cf. eq. (22) and
Fig. 6)
x
c
U
t t
c
1 bs
(29)
This front eventually reaches the source when t t
c
. (Fig. 6).
Subsequently, for t >t
c
, the depth of the current at the source
decreases to values h <h
c
and is given by the original solution (17)
in the region 0 <x <x
b
(t). From this solution, it follows that the
point nearest to the source at which the current depth equals the
critical depth h
c
has position
x
b
Ut t
c
(30)
This front eventually catches up with the leading edge of the
current, which advances at the rate
X
n
Ut/l, at the time given by (cf. eq. (27) and Fig. 6)
t t
b
l
t
c
l 1
(31)
Subsequently, the whole ow evolves according to the simple
non-draining solution (17) (see Fig. 7, dotted line).
3.3. Fraction which drains
In general the fraction of the ow which drains depends on the
capillary pressure, which suppresses the draining, the source ow
rate, the ratio of the draining time to the decay time of the source,
bs, and also the residual saturation of the water and the gas at the
advancing and receding fronts, as expressed by l.
In general the expression is complex to calculate, but is found
by comparing the volume input at the source with the volume
which remains in the formation, with the difference represent-
ing the fraction which has drained. There is however a useful
limit when h
c
<h(0) in which case, to leading order, the fraction
retained in the original layer may be found by integration of eq.
(20) evaluated at t lx/u. This leads to the result that the frac-
tion of the source uid which remains trapped in the formation,
F, is given by
F
l 1
l 1 bs
(32)
where we note that l >1 (eq. (9)). This expression effectively
compares the process of capillary trapping at the nose and tail of
the ow with the drainage through the upper boundary. It illus-
trates that if the draining time 1/b is short compared to the decay
time of the source, t, then F will become relatively small, and much
of the injected uid can drain away, whereas if the draining time is
comparable to or longer than the decay time of the source, then
much of the injected uid remains in the original layer. As the
capillary pressure increases, this further restricts the fraction of the
ow which drains, and so the above expression provides a lower
bound on the fraction of the ow which remains trapped in the
original layer of the formation.
We illustrate the variation of F with bs for a series of represen-
tative values of l (0.05, 0.1 and 0.15) in Fig. 8 below.
Fig. 5. Illustration of the evolution of the current with time. Each prole corresponds
to a vertical line (ie constant time) in the (x,t) plane of Fig. 4; in this case, the proles at
t/4 and t/2 lie in the range (t
d
<t <t
b
), while the proles at 3t/4 and t correspond to
times greater than t
b
.
Fig. 6. Illustration of the structure of the current on an xt plot. The gure shows how
the various transition points in the current evolve with time. For t <t
d
the ow is
everywhere deeper than h
c
. For t
d
<t <t
c
the near source region is deeper than h
c
while the distal part of the ow, x >x
c
, has constant depth h
c
. For t >t
c
the near source
region has become shallower than h
c
, while for x >x
b
the ow has constant depth.
At late times, t >t
b
the ow is everywhere shallower than h
c.
4. Draining through faults
In comparison with the above results in which the draining
occurs through the upper boundary of the formation, we now
consider the case in which uid leaks off through a localised fault
which cuts across the layers. Typically faults are narrow compared
to the length scale of the ow, but provide a higher permeability
route to the surface. If the fault connects the ow in the lower
owing layer to a layer of high permeability above the seal layer,
then the ux through this fault will have the form (cf. Pritchard,
2007)
Q
fault

k
f
Drg cos q hw
mb
Uh (33)
where w is the width of the fault, and b the vertical thickness of the
fault, across which the gas pressure acts to drive the ow through
the fault. k
f
is the permeability of the fault, and h is the current
thickness just upstream of the fault. Here we assume that h >b so
that the hydrostatic pressure driving the ow through the fault is
associated with the buoyancy of the plume of injected uid in the
lower owing layer.
The fault ux Q
fault
represents a discontinuity in the ux of gas
along the layer. Since the alongslope ux scales as Uh, it follows that
the drainage through the fault dominates the ux along the
formation if U>U, and in this case, there will be no ux beyond the
fault, which for convenience we assume is located at x x
f
.
In this case, the fraction which remains in the formation is given
by the fraction which is trapped by capillary retention upstream of
the fault
F
1 exp

x
f
l 1
Us
(34)
Here, the critical balance is between the distance the fault lies
away from the source and the distance that would be travelled by
the plume over the time required for the source to decay, Us.
5. Application
It is useful to examine the implications of the model for a typical
example of the ow in a layered permeable rock. In the case of
a geological waste repository, there may be a ux of buoyant gas
with a decay time of order 300 years, and an initial ux of 10
5
m
2
/s
per unit length of the repository. If there is a layer of rock of
permeability 10
15
m
2
bounded above by a layer of permeability
10
17
m
2
, then with a porosity of 0.1 and a layer inclination of 10
o
,
the along layer velocity scale U has value of order 10
8
m/s with
uid of viscosity 10
4
Pa s. If the overlying seal layer has thickness
of order 1 m, then b has value of order 10
10
s
1
and so bs w1,
suggesting that the draining and the decay of the source occur over
approximately comparable times. If the capillary entry pressure to
the overlying layer is small, then the fraction of the owretained in
the layer is given by relation (32), within the simplied framework
of this model, and this has value of about F w0.10.2.
In the case of a rapidly decaying source or a current with slow
drainage rate, 1 <bs, then the drainage ux through the seal layer
F
D
(x,t), per unit length along the current, which is supplied to
points higher in the formation, is given by
F
D
x; t bh0; 0exp
t
s

x
Us
1 bs
for 0 < x < Ut=l

if t < t
c
and for
Ut t
c
1 bs
< x < Ut=l if t
c
< t < t
d
35
While in the case of a slowly decaying source or current with
high drainage rate, bs >1, the drainage ux, again given by the
same expression as in (35), is always located near to the source,
with drainage in the region 0 <x <Ut/l if t <t
d
and drainage in the
region 0 <x <U(t t
c
)/(1 bs) when t
d
<t <t
c
. Subsequently there
is no draining.
With a capillary entry pressure corresponding to a depth of
order 1 m, then with the above values for U, b and s it follows that
t
c
wt at which time the current has travelled a distance of order
100 m. The draining zone then evolves away from the source in the
slow draining case, until time t
d
w(1.11.2) t
c
. Similarly in the fast
draining case, the current will propagate about 100 m from the
source, while the draining persists, with the illustrative parameters
given in the example above. The plume will then cease to drain and
will migrate along the original layer as a thin, elongate ow. For
smaller capillary entry pressure, the ow may drain for times cor-
responding to several multiples of t, and hence the draining region
may extend several hundred metres alongslope.
In a different situation of CO
2
sequestration, the injection period
may only be of order 30 years. If the injectivity of the formation
D
i
m
e
n
s
i
o
n
l
e
s
s

D
e
p
t
h

e c n a t s i D s s e l n o i s n e m i D
Fig. 7. Illustration of the case in which for early times, t <t
d
, there is a draining zone
near the source (long and short dashed lines) and that as the current loses mass
through draining it reaches the critical depth at the nose of the ow, x x
n
, when t t
c
.
For t
c
>t >t
d
the location of the point closest to the source at which the depth equals
the critical value h
c
progressively migrates back to the origin which it reaches at t t
c
(dot-dashed line) . For t >t
c
, the depth is smaller than h
c
in the region x <x
b
, while the
more distal part of the ow, x >x
b
, has constant depth h
c
(dotted line). Eventually, for
t >t
b,
the whole ow is shallower than h
c.
0
0.2
0.4
0.6
0.8
1
0 0.2 0. 4 0.6 0. 8 1
f
r
a
c
t
i
o
n

r
e
t
a
i
n
e
d

i
n

t
h
e
o
r
i
g
i
n
a
l

l
a
y
e
r

Time of Decay of Source / Draining Time
0.15
0.05
Fig. 8. Illustration of the variation of the fraction of the ow which remains trapped in
the original layer of the formation as a function of the time of decay of the source
compared to the draining time across the thin partial seal layer.
becomes impeded with time, the continuing injection ux may
then decay with time, and the present model may give a guide to
the ow. Initially, the ux per unit length injected into a long
horizontal well may again be of order 10
5
m
2
/s, and if the
formation has similar properties to the example above, this would
correspond to the case bs w0.1 which represents a rapidly decay-
ing source compared to the drainage rate. Now the current would
continue draining for a period t
d
w5s which is about 150 years,
in which time it would propagate about 50 m from the source. We
note that in the case of a maintained steady ux, the drainage
dynamics are somewhat different, as described by Woods and
Farcas (2009).
From both these idealized examples, we see that with a decay-
ing source, the injected uid may rapidly spread alongslope and
hence thin out, thereby limiting the fraction of the ow which
drains into the overlying formation compared to the fraction which
becomes capillary trapped in the original owing layer.
6. Summary
Using a simplied approach, we have identied and modeled
some of the controls on the migration of buoyant uid through
a layered permeable rock issuing from a waning source of buoyant
uid. We have focussed on the dynamics of the current in a single
layer of the formation, examining the balance between leakage
from the layer, and lateral spreading of the current along that layer.
In modelling the leakage, we have accounted for the capillary
entry pressure into an overlying seal layer, and shown that this
leads to a localized region of leakage which evolves in time. As the
current wanes, the capillary entry pressure suppresses further
leakage and the remainder of the current migrates through the
original layer. Capillary trapping of the uid in the original layer
leads to a continual loss of uid from the ow, and as the plume
disperses, it is eventually trapped within this layer. The balance
between the fraction of the ow which is trapped in the original
layer, and the fraction which leaks into the overlying layer depends
on the ratio of draining time through the overlying layer compared
to the decay time of the source. With a rapidly decaying source,
most of the uid remains trapped in the original layer, whereas
with a slowly decaying source, much of the uid is able to leak into
higher parts of the geological formation.
We have also shown that if the current reaches a fracture, then
a signicant part of this current may be diverted through the
fracture and then migrate higher into the formation. The critical
controlling parameter in this case is the ratio of the distance of the
fracture from the source to the product of the Darcy ux and the
decay time of the source. The further the fracture from the source
the greater the fraction of the ow which is sequestered in the
original layer in which the buoyant uid is injected.
References
Barenblatt, G.I., 1996. Dimensional Analysis, Self-Similarity and Intermediate
Asymptotics. CUP.
Bear, J., 1972. Dynamics of Flow in Porous Media. Elsevier, pp. 1746.
Bickle, M., Chadwick, A., Huppert, H.E., Hallworth, M., Lyle, S., 2007. Modelling
carbon dioxide accumulation at Sleipner: implications for underground carbon
storage. Earth Planet. Sci. Lett. 255, 164176.
Farcas, A., Woods, A.W., 2009a. The effect of drainage on the capillary retention of
CO
2
in a layered permeable rock. J Fluid Mech. 618, 349359.
Farcas, A., Woodsm A.W. On the steady drainage of a gravity current, from a point
source, up a sloping layered permeable rock, sub-judice. J Fluid Mech. in press.
Hesse, M.A., Tchelepi, H.A., Orr, F.M. Jr., 2006. Scaling analysis of the migration of
CO
2
in aquifers. In: SPE 102796, Presented at the 2006 SPE Annual Technical
Conference and Exhibition, San Antonio, TX, 2427 Sept 2006.
Hesse, M.A., Orr Jr., F.M., Tchelepi, H.A., 2008. Gravity currents with residual trap-
ping. J Fluid Mech. 351, 3560.
Huppert, H.E., Woods, A.W., 1995. Gravity-driven ows in porous layers. J. Fluid
Mech. 292, 5569.
Kharaka, Y.K., Cole, D.R., Hovorka, S.D., Gunter, W.D., Knauss, K.G., Freifeld, B.M.,
2006. Gas water rock interactions in Frio formation following CO
2
injection:
implications for the storage of greenhouse gases in sedimentary basins.
Geology 34 (7), 577580.
Mitchell, V., Woods, A.W., 2006. Gravity driven ow in conned aquifers. J. Fluid
Mech. 566, 345355.
Nordbotten, J.M., Celia, M.A., 2006. Similarity solutions for uid injection into
conned aquifers. J. Fluid Mech. 561, 307327.
Obi, E.-O.I., Blunt, M.J., 2006. Streamline-based simulation of carbon dioxide storage
in a North Sea aquifer. Water Resour. Res. 42, W03414. doi:10.1029/
2004WR003347.
Pritchard, D., Woods, A.W., Hogg, A.J., 2001. On the slow draining of a gravity
current moving through a layered permeable medium. J. Fluid Mech. 444,
2347.
Pritchard, D., 2007. Gravity currents over fractured substrates in a porous medium.
J Fluid Mech. 584, 415431.
Vella, D., Huppert, H.E., 2007. Gravity currents in a porous medium at an inclined
plane. J Fluid Mech. 555, 353362.
Woods, A.W., Farcas, A., 2009. Capillary entry pressure and the leakage of gravity
currents through a sloping layered permeable rock. J Fluid Mech. 618,
361379.
Clay smear in normal fault zones The effect of multilayers and clay cementation
in water-saturated model experiments
J. Schmatz
a,
*
, P.J. Vrolijk
b
, J.L. Urai
a
a
Structural Geology, Tectonics and Geomechanics, Geological Institute, RWTH Aachen University, Lochnerstrasse 4-20, 52056 Aachen, Germany
b
ExxonMobil Upstream Research Co., P.O. Box 2189, Houston, TX, USA
Article history:
10 December 2009
Accepted 13 December 2009
Keywords:
Clay smear
Fault seal
Mechanical layering
Competence contrast
Experimental model
a b s t r a c t
We studied the evolution of fault zones in water-saturated model experiments consisting of sand and
clay layers above a normal fault dipping 70
in a stiff basal layer. The model is bounded below by a rigid

metal basement with a pre-cut 70
fault and above by a metal plate, also with a 70
cut, aligned in the

same plane as the basement fault. Quantitative analysis of particle displacements was undertaken with
PIV (Particle Image Velocimetry) software. In these models, the structure of initial localized deformation
evolves into a kinematically favorable fault zone. This evolution, which produces releasing or restraining
relays across the clay layer, has a major role in controlling fault-zone structure. We show that a high
competence contrast between sand and clay leads to a more complex fault zone due to the formation of
secondary shear zones and segmentation-induced fault lenses. A high competence contrast also
promotes a more complex temporal evolution of those shear zones. Weak clay layers are preferentially
enriched in fault zones, whereas strong, brittle clay initially fractures and forms clay boudins that rotate
in the deforming sand. With progressive deformation these boudins are abraded and transformed into
a soft-clay gouge. Thin, weak clays deform continuously over large displacements, and the volume of
clay-rich gouge increases as sand mixes into clay at the margins of the shear zone. Thus, we observe
a wide range of fault zone and fault gouge evolution by adjusting the mechanical properties of the clay.
Further physical insights into fault processes like those reached here may yield predictive models of
fault-zone evolution that will transcend empirical methods (e.g., shale-gouge ratio, SGR).
1. Introduction
Fault networks havea major effect onhydromechanical processes
in sedimentary basins. We have a basic understanding of the basic
physical processes of fault sealing, but the complex geometries and
many feedback processes make quantitative prediction difcult.
Fault zones have been widely investigated in eld studies, experi-
mental models and numerical simulations to better understand
their transport and sealing properties (e.g., Weber et al., 1978; Leh-
ner and Pilaar, 1991, 1997; Antonellini et al., 1994; Fulljames et al.,
1997; ClausenandGabrielsen, 2002; James et al., 2004; Egholmet al.,
2008; Mair and Abe, 2008; Urai et al., 2008).
Much of the research on faults in sandclay sequences is focused
on predicting the amount of clay incorporated into the fault gouge,
commonlycalledclaysmear. Claysmear, a looselydenedtermused
in hydrocarbon geology, describes the processes in which clay from
the wall rock is incorporated in a fault zone (Yielding et al., 1997). In
subsurface studies, numerous statistical algorithms are used for clay
smear analysis: clay-smear potential, shale-smear factor and shale-
gouge ratio (Lindsay et al., 1993; Fristad et al., 1997; Fulljames et al.,
1997; Yielding et al., 1997; Yielding, 2002). Most of these methods
are based on the assumption that fault gouge consists of a reworked
equivalent of the wall rocks offset by a fault (Holland et al., 2006)
without theadditionor removal of material (vanGent et al., inpress).
Because this method averages over the lithologic interval offset by
the fault, it fails to accurately predict the spatial distribution of clay
along the fault plane. In addition, the mechanical properties of the
rocks in the wall rock and fault zone are neglected. Clearly, a better
understanding of the processes involved during faulting would
improve the quality of fault-seal predictions.
When a segmented normal fault cuts heterogeneously layered
sequences with competence contrast, (e.g., clay and sand), the fault
develops a steeper dip in layers with a higher friction angle and
a shallower dip in layers with a lower friction angle (Peacock and
Sanderson, 1992). This results in vertically segmented faults with
steps at lithologic/competency boundaries (Childs et al., 1996).
* Corresponding author. Fax: 49 80 92358.
E-mail addresses: j.schmatz@ged.rwth-aachen.de (J. Schmatz), peter.vrolijk@
exxonmobil.com (P.J. Vrolijk), j.urai@ged.rwth-aachen.de (J.L. Urai).
doi:10.1016/j.jsg.2009.12.006
Other authors (e.g., Lehner and Pilaar, 1997; Van der Zee et al.,
2003) postulated that clay injection is able to enrich the fault
with clay over the amount expected by lithologic offset. The clay
injection mechanism proposed by Lehner and Pilaar (1997)
contains two essential elements: (1) a pull-apart structure forms
when the fault crosses a clay layer. (2) The clay bed is injected into
this pull-apart structure and subsequently sheared to form a thick
clay gouge. The injection process requires signicantly weaker clay
than the surrounding sand layers (Van der Zee et al., 2003). van
Gent et al. (in press) describe a fundamentally different process of
clay enrichment, which involves the vertical transport of clay along
dilatant fractures in brittle carbonates.
Experiments by Sperrevik et al. (2000) and Clausen and Gabri-
elsen (2002) showed a strong dependency of clay smear on the
mechanical properties of the materials. They used a ring-shear
apparatus to deform clay and sand layers to high strains at variable
normal stresses. They observed pronounced clay smear with
increasing stresses and decreasing clay strength. In a preliminary
set of experiments of layered sandclay sequences, Schmatz et al.
(2010) varied the clay-layer proportions, clay-layer thickness and
spacing, and clay composition; normal faulting created continuous
smear in weak, under-consolidated clay; in contrast, a strong, over-
consolidated clay rst deforms in a brittle mode and then some-
times becomes reworked to form a soft-clay gouge. Here we
present the results of water-saturated experiments with the added
boundary condition of a strong top layer containing a pre-cut fault
in a kinematically favorable orientation to the basement fault, and
investigate the effect of different multilayer congurations and clay
cementation on the structure of the evolving fault zone.
In experiments with a free top surface, deformation is less local-
ized at the topof the model (Fig. 1). Pilot experiments (Schmatz et al.,
2010) showthat a top plate and basement fault acting together form
two precursor faults: one initiating at the tip of the basement fault,
and one at the tip of the fault in the top plate (cf. quadshear, Welch
et al., 2009). With progressive deformation they link up across the
model to form a kinematically favored zone of deformation.
2. Methods
2.1. Setup
A sandbox was constructed to deform water-saturated, layered
(40 20 20 cm) (width/height/depth) sandclay models (Fig. 2).
In this series of experiments, the boundary conditions were chosen
to simulate a series of weaker layers sandwiched between two
strong and stiff layers cut by two co-planar faults dipping 70
(see
also Mandl, 2000; Van der Zee, 2002; Ferrill and Morris, 2003;
Adam et al., 2005). Water-saturated models allowed the deforma-
tion of wet clay and cohesionless sand together in one model
(Schmatz et al., 2010). The basement fault moved at 40 mm/h to
a maximum offset of 60 mm. The models were run between two
glass plates lubricated to minimize edge effects. At this deformation
rate, the thick clay layers were sheared under undrained condi-
tions, whereas pore pressures likely remained hydrostatic inside
the ne-grained sand. The resulting material properties were
characterized by a series of standard geotechnical measurements.
Full details of the methods for a free-surface boundary condition
are given in Schmatz et al. (2010). Here we summarize the most
important aspects and procedures that result from additional
boundary conditions and experiment designs.
2.2. Boundary conditions
We used 2 cm thick aluminum (r 2400 kg m
3
) top plates,
pre-cut at the same 70
angle as the basement fault, placed on top

of the models and aligned (co-planar) with the basement fault
(Fig. 1). The stiff base plate acts as a rigid guide compared to the
sediment, and the top plates rotate and move but without bending.
Approximately 40 experiments were run using variable clay
strengths and number and thicknesses of the clay layers to inves-
tigate fault-zone processes. The experiments were recorded with
time-lapse photographs (Schmatz et al., 2010) with constant time
interval and resolution, and are presented throughout the paper
cropped to the same section.
2.3. Sand
We used the same washed, well-sorted quartz sand with a grain
size of 0.10.4 mm as in Schmatz et al. (2010), with colored marker
horizons. The grain size range results in sand packing classied as
medium-dense; thus, the sand yields at a distinct peak shear
a
b
Fig. 1. (a) Sketch showing boundary conditions and fault evolution in an experiment
with a free surface. The steep precursor fault is shown with a dashed line. The fault
rotates into its kinematically favored position (bold line). Compare to Schmatz et al.
(2010). (b) Sketch showing boundary conditions and fault evolution in an experiment
with rigid top plates. Two precursor faults develop, initiating at the tip of the faults in
the basement and top plates. With progressive deformation, the zone of deformation
migrates into the center of the model (bold line).
J. Schmatz et al. / Journal of Structural Geology 32 (2010) 18341849 1835
strength and forms a localized shear band in the experimental
model.
2.4. Clay
The clay is a quartz (46.8%)kaolinite (42.0%) mixture with
water contents of w30, 40 and 50 wt.%, representing over-consol-
idated (rm), normal-consolidated (normal) and under-consoli-
dated (soft) clay, respectively. Direct shear tests indicate a ten times
higher undrained shear strength for rm clay than for soft clay
(Schmatz et al., 2010). In some experiments, small amounts (0.25
10 wt.%) of Portland cement were added to the soft clay. The
cement hardened (while water-saturated) the clay in w17 h.
Geotechnical measurements show undrained shear strength with
a linear increase from 2.6 kPa (1.0 wt.% cement) to 27 kPa (5.0 wt.%
cement). By comparison, rm but uncemented clay has a shear
strength of w2.4 kPa. To ensure that the cemented clay was
completely cured, most of the models were left overnight before
starting the deformation experiment.
2.5. Particle image velocimetry (PIV)
Time-lapse series of high-resolution digital images were pro-
cessed into movies and analyzed using PIV (e.g., Wolf et al., 2003;
Adamet al., 2005; Steingart andEvans, 2005; Hollandet al., 2006; van
Gent et al., in press). The PIV analysis documents displacements of
a granular structure on the particle scale and gives detailed velocity
elds. The resolution is given by an interrogationwindowsize of 128
128pixels decreasing toawindowof 3232pixels withanoverlap
of 75% producing a nal window size of 8 8 pixels, which corre-
sponds to approximately 16 sand grains in our experiments (Adam
et al., 2005; Schmatz et al., 2010). Movies of all experiments are
available at www.ged.rwth-aachen.de.
3. Experimental results
In this study we undertook nine systematic experimental series.
Parameters studied include: clay strength, layer thickness, and the
number of layers. To allow comparison with experiments without
top plates, we repeated a number of earlier experiments both with
and without the top plate. The full series of experiments is
summarized inTable 1. Here we describe a number of these in detail
(Figs. 316). To check the reproducibility of the experiments, one of
the experiments presented here (R-2-s-tp, Fig. 7, Table 1) was
repeated four times under identical conditions. We reproduce
many rst-order fault structures in these repeat experiments, but
sometimes small variations arise.
3.1. Effect of top plates
Counterclockwise rotation of the hanging wall top plate up to
1.8
was observed in most of the experiments. Measurements of

rotation show no correlation to the clay setup, clay type or layer
thickness.
3.1.1. Sand only
Our reference experiment is a water-saturated, sand-only model
(Fig. 3, Table 1, experiment P-1) with a base area of 40 20 cm and
15 cm high, with free surface, comparable to the boundary condi-
tion used in Horseld (1977) and also discussed in Adam et al.
(2005). Initial displacement along the basement fault creates
a precursor fault (Mandl, 2000) which migrates through the
material into a kinematically favorable position forming a trian-
gular shear zone (e.g., Erslev, 1991; Zehnder and Allmendinger,
2000; Cardozo et al., 2003; Fig. 3a and b). PIV visualization of the
incremental strain eld e
yx
(Fig. 3b) shows the migration of locali-
zation through the material and formation of a stable planar
deformation zone at a displacement of w14
mm. Adding the pre-

cut aluminum plates produces signicantly different results (Fig. 4,
Table 1, experiment P-2-tp). A precursor fault initiates from the
basement fault tip, dipping at approximately 45
in the direction of
the hanging wall. Almost immediately after 2 mm displacement on
the basement fault, deformation is also localized at the base of the
fault between the top plates, propagating downwards. With
progressive deformation these two zones interact, the precursor
fault becomes inactive, and deformation is localized in a stable,
planar deformation zone in the plane of the basement fault.
Localization is thus much more rapid than in the experiment with
a free top surface (displacement of w7
mm).
3.1.2. Sand and clay
Adding a 30 mm thick layer of rm clay to the model (Fig. 5,
Table 1, experiment N-9-f) has a major impact on the fault-zone
evolution. The experiment with a free surface is comparable to Figs.
9 and 11 of Schmatz et al. (2010). Here the fault-zone evolution is
complex. The steep precursor fault fails to propagate across the clay
layer, which instead deforms by bending. Fractures initiate at the
claysand interface at locations of high curvature and extension.
With increasing deformation the layer breaks into fragments of
various sizes which then rotate and become reworked into a ductile
clay gouge. The ow of sand grains along clay asperities is associ-
ated with asperity abrasion and mixing of sand and clay. Secondary
faults propagate fromfractures in the clay across the sand. A stable,
distinct planar fault plane fails to form for offset >30 mm (Fig. 5b).
a
b
Fig. 2. (a) Sketch of the underwater experimental apparatus with: 1) water-saturated
apparatus (inside), 2) waterproof container (outside), 3) rigid basement bottom with
drain perforation, 4) motor-driven fault offset, 5) basement fault dipping 70
, 6)
movable glass plate (inside), 7) movable glass plate (outside). (b) Sketch of setup
showing alternating layers of sand and clay overlying stiff basal block, covered with
pre-cut aluminum top plates. Red area indicates kinematically favored position of the
fault. (For interpretation of the references to colour in this gure legend, the reader is
referred to the web version of this article.)
In the same experiment with pre-cut top plates, both fault-zone
geometry and clay-gouge evolution are markedly different (same
vertical stress at the top of the clay layer; Fig. 6, Table 1, experiment
N-8-f-tp). Analogous to the experiment with sand only and top
plates, a precursor fault with a curvature towards the hanging wall
initiates at the basement fault tip, followed by a second fault that
initiates at the tip of the fault in the top plates and propagates
downwards. Bending and macroscopic brittle failure are absent.
The two fault segments link up across the clay layer and are
progressively straightened. At a displacement of 10 mm, a nearly
planar fault forms with a continuous, w2 mm thick clay gouge.
3.2. Effect of clay strength
3.2.1. Low competence contrast
Experiment R-2-s-tp contains a 40 mmthick sand layer between
two thin layers of soft clay (Fig. 7, Table 1); the clay beds compact
into normal clay before the experiment ends (Schmatz et al., 2010).
The competence contrast between sand and clay is thus low. The
rst fault initiates at the basement fault tip, followed by a second
fault initiating at the fault tip at the top plates. Belowthe lower clay
layer, the lower fault branched, with the branch on the footwall side
propagating through the clay. At the upper clay layer, there was
a mismatch between the two deformation bands arriving from
below and above, and the fault zone was initially segmented. The
nal, kinematically favored fault zone thus has a number of inactive
branches and is wider at the level of the clay layers with continuous
clay smear between the fault strands (Fig. 7b; cf. Fig. 3b of Van der
Zee et al., 2003).
Experiment C2-4-s-tp (Fig. 8, Table 1) has a similar initial stra-
tigraphy as the one described above, but the sand between the two
clay layers is 50 mm thick. To increase the competence contrast
slightly, 0.25 wt.% Portland cement was added to the soft-clay
mixture (this increased the strength only slightly from 0.2 kPa for
Table 1
Overview of experiment series.
Series New ID # Clay
layers
Thickness Layer setup (clay) Clay
type
Top
plates
#
Images
Velocity
[mm/min]
Water
content [%]
Hardening
[h]
Cement
[wt.%]
Figure SGR
diagram
Adam et al., 2005 P-1 0 150 133 0.7 4
1) Sand only P-2-tp 0 140 133 0.4 5, 19
P-3-tp 0 140 0.4
2) Changing two
parameters:
clay strength and layer
thickness
S-1-s 1 10 5/10/140 Soft 224 0.2 52 2
S-2-s 1 3 5/3/140 Soft 268 0.2 52 19
S-3-f 1 3 5/10/140 Firm 276 0.2 27
S-4-n 1 30 5/30/140 Normal 290 0.2 44 21
S-5-n 1 10 5/10/140 Normal 300 0.2 40 23
S-6-n 1 3 5/3/140 Normal 298 0.2 44 24
S-7-s 1 30 10/30/140 Soft 310 0.2 50 9
Schmatz et al., 2010 S-8-f 1 30 5/30/140 Firm 296 0.2 28 34
S-9-f 1 3 5/3/140 Firm 302 0.2 28 31
3) Changing four
parameters:
clay strength, layer
thickness,
number of layers, layer
orientation
N-1-s-tp 2 10 55/10/10/10/55 Soft 181 0.4 52 3
N-2-s-tp 1 30 55/30/55 Soft 183 0.4 52 12, 19 5
N-3-s-tp 1 30 30 Soft 185 0.4 52 13
N-4-n-tp 1 3 3 Normal 179 0.4 48 25
N-5-s-tp 1 3 3 Soft 146 0.4 52 17
N-6-s-tp 1 3 60/3/77 Soft 193 0.4 53 16
N-7-s-tp 2 3 50/3/6/3/78 Soft 182 0.4 52 10
N-8-f-tp 1 30 50/30/60 Firm 186 0.4 37 7 28
N-9-f- 1 30 50/30/60 Firm 189 0.4 34 6 29
N-10-n-tp 2 3 50/3/6/3/78 Normal 191 0.4 l45/u47 20
N-11-n-tp 1 10 50/10/80 Normal 191 0.4 44 22
4) Changing clay layer
distance
D-1-s-tp 2 3 40/3/30/3/64 Soft 187 0.4 53 8
D-2-s-tp 2 3 40/3/40/3/54 Soft 215 0.4 53 6
D-3-s-tp 2 3 40/3/50/3/44 Soft 203 0.4 53 18
5) Properties of cement C1-1-s-tp 1 2 60/2/78 Soft 195 0.4 50 72 10
C1-2-s-tp 1 3 60/3/77 Soft 100 0.4 53 72 1 35
C1-3-s-tp 1 3 60/3/77 Soft 158 0.4 52 72 5
C1-4-s-tp 1 5 60/5/75 Soft 181 0.4 52 48 10
C1-5-s-tp 1 3 60/3/77 Soft 126 0.4 51 22 10
C1-6-s-tp 1 3 60/3/77 Soft 157 0.4 53 2 10 10, 16 33
6) Reproduction R-1-s-tp 2 3 40/3/40/3/54 Soft 184 0.4 55 11
R-2-s-tp 2 3 40/3/40/3/54 Soft 181 0.4 52 8 7
R-3-s-tp 2 3 40/3/40/3/54 Soft 180 0.4 50 12
R-4-s-tp 2 3 40/3/40/3/54 Soft 191 0.4 51 14
R-5-s-tp 2 3 40/3/40/3/54 Soft 189 0.4 53 15
7) Cement content C2-1-s-tp 1 3 50/3/87 Soft 175 0.4 48 22 1 36
C2-2-s-tp 2 3 50/3/50/3/34 Soft 173 0.4 49 23 1 32
C2-3-s-tp 2 3 50/3/50/3/34 Soft 183 0.4 49 22 2 11 38
C2-4-s-tp 2 3 50/3/50/3/34 Soft 184 0.4 49 22 0 9 27
C2-5-s-tp 2 3 50/3/50/3/35 Soft 178 0.4 48 22 2 37
C2-6-s-tp 2 3 50/3/50/3/34 Soft 188 0.4 50 24 0 26
8) Multilayer M-1-s-tp 4 8 30/8/12/8/12/8/
12/8/30
Soft 186 0.4 44 13, 17 4
M-2-s-tp 4 Various 35/20/10/3/10/
20/10/2/40
Soft 185 0.4 50 14 1
9) Clay 2 50 20/50/5/50/40 Soft 163 0.4 48
10) Others K-1-f-tp 1 3 60/3/77 Firm 178 0.4 37 30
K-2-f-tp 1 3 50/3/87 Firm 133 0.4 34
K-3-f-tp 1 3 80/3/57 Firm 162 0.4 32
Fig. 3. (a) Image sequence showing experiment P-1 with 150 mm thick sand layer and
free top surface (see inset). The three stages record offset along the basement fault of
0.4, 15, and 22 mm. (b) Image sequence showing the corresponding PIV overlay to the
image sequence in (a) with the velocity vector eld displayed in the foreground and
the contour plot of incremental strain e
xy
in the background. Modied from Adam et al.
(2005). (c) Legend for insets in Figs. 318.
Fig. 4. (a) Image sequence showing experiment P-2-tp with a 140 mm thick sand
package and top plates (see inset). Fault offset is 2, 4, and 16 mm. Red lines indicate
location of rigid blocks at bottom and top. (b) Image sequence showing the corre-
sponding PIV overlay to image sequence in (a) with the velocity vector eld in the
foreground and a contour plot of the z-component of the incremental rotation eld in
the background. (For interpretation of the references to colour in this gure legend, the
reader is referred to the web version of this article.)
Fig. 5. (a) Image sequence showing experiment N-9-f with a 30 mm thick rm clay
layer in the center and a free top surface (see inset). Basement fault offset is 3, 6, and 35
mm. (b) Image sequence showing the corresponding PIV overlay to image sequence (a)
with the velocity vector eld in the foreground and a contour plot of the z-component
of the incremental rotation eld in the background.
Fig. 6. (a) Image sequence showing experiment N-8-f-tp with a 30 mm thick rm clay
layer in the center of the model and top plates (see inset). Basement fault offset is 4, 7
and 10 mm. (b) Image sequence showing the corresponding PIV overlay to image
sequence in (a) with the velocity vector eld in the foreground and a contour plot of
the z-component of the incremental rotation eld in the background.
Fig. 7. (a) Image sequence showing experiment R-2-s-tp with two 3 mm thick soft-
clay layers and top plates (see inset). Basement fault offset is 4, 6 and 20 mm. Red lines
indicate location of rigid blocks at bottom and top. (b) Image sequence showing the
corresponding PIV overlay to the image sequence in (a) with the velocity vector eld in
the foreground and a contour plot of the z-component of the incremental rotation eld
in the background. (For interpretation of the references to colour in this gure legend,
the reader is referred to the web version of this article.)
Fig. 8. (a) Image sequence showing experiment C2-4-s-tp with two 3 mm thick
cemented clay layers and top plates (see inset). Basement fault offset is 5, 9 and 36
mm. Red lines indicate location of rigid blocks at bottom and top. (b) Image sequence
showing the corresponding PIV overlay to the image sequence in (a) with the velocity
vector eld in the foreground and a contour plot of the z-component of the incre-
mental rotation eld in the background. (For interpretation of the references to colour
in this gure legend, the reader is referred to the web version of this article.)
soft clay to 0.25 kPa for soft clay with 0.25 wt.% cement). We
observe two precursor faults initiating fromthe basement and from
the top. The lower fault initiates with a steep dip, but, after inter-
secting the lower clay layer, it curves towards the footwall. The
upper fault evolves in its kinematically favored plane. Both faults
are active throughout the experiment, with a progressively thin-
ning but stable fault lens between the two active fault strands
(Fig. 8b). Continuous clay smear forms everywhere in the model.
The two experiments described above are therefore similar initially,
but the second experiment allows simultaneous displacement on
two fault surfaces whereas the rst experiment constrained
displacement to one fault strand.
3.2.2. High competence contrast
3.2.2.1. Strong clay. Experiment C1-6-s-tp (Fig. 9, Table 1) contains
a single, 3 mmthick clay layer 60 mmabove the basement. The clay
contains 10 wt.% cement and is cured only partly, so its strength is
approximately the same as the completely cured, 2 wt.% cemented
Fig. 9. Image sequence showing experiment C1-6-s-tp with one 3 mm thick cemented clay layer and top plates (see inset). Basement fault offset is 8, 16 and 28 mm. Red lines
indicate location of rigid blocks at bottom and top. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
Fig. 10. Image sequence showing experiment C2-3-s-tp with two cemented (1.5 wt.%) clay layers and top plates (see inset). Basement fault offset is 5, 25, and 45 mm. Red lines
indicate location of rigid blocks at bottom and top. (Transparent, reddish box in center of photos is a camera reection). (For interpretation of the references to colour in this gure
legend, the reader is referred to the web version of this article.)
clay. Deformation involves monoclinal bending of the clay followed
by fracturing and block rotation. The shear zones in the sand form
a wide lens with continuous activity on multiple fault strands. The
interior of the lens is highly deformed, in contrast to the previously
described experiments.
Experiment C2-3-s-tp (Fig. 10, Table 1) has two layers of 3 mm
thick clay. The clay was fully cured but with only 1.5 wt.% cement.
As in the previous experiment, initial deformation causes brittle
failure in both layers. Progressive deformation leads not only to
displacement and rotation of the fragments, but also to progressive
erosion of the fragment edges, forming a ductile clay gouge with
clay clasts and coremantle structure around the old fragments.
3.2.2.2. Weak clay. Experiment N-2-s-tp contains one 30 mmthick,
soft-clay layer (Fig. 11, Table 1). A near-vertical precursor fault
initiates from the basement fault tip. After 2 mm displacement on
the basement fault, deformation localizes at the tip of the fault
between the top plates and propagates downward. With progres-
sive deformation these two zones interact and overlap, forming
a restraining relay zone across the clay layer (Fig. 11a, center). Initial
Fig. 11. (a) Image sequence showing experiment N-2-s-tp with one 30 mm thick soft-
clay layer (see inset) and top plates. Fault offset is 4, 6 and 34 mm. Approximate fault
location is traced with red lines. (b) Image sequence showing the corresponding PIV
overlay to the image sequence in (a) with the contour plot of incremental rotation eld
and the velocity vector eld. (For interpretation of the references to colour in this
gure legend, the reader is referred to the web version of this article.)
Fig. 12. (a) Image sequence showing experiment experiment M-1-s-tp with four 8 mm
thick soft-clay layers and top plates (see inset). Fault offset is 2 and 11 mm. Red lines
indicate location of rigid blocks at bottom and top. Fig. 17 shows nal stage of this
experiment. (b) Image sequence showing the corresponding PIV overlay to the image
sequence with the contour plot of incremental rotation eld and the velocity vector
eld. (For interpretation of the references to colour in this gure legend, the reader is
deformation in the clay layer is in this restraining zone (cf. Egholm
et al., 2008). With increasing displacement, the top precursor fault
becomes inactive, and deformation switches to a new strand
forming a restraining relay zone relative to the fault below the clay
layer (Fig. 11b, bottom). At the same time, the fault below the clay
layer rotates clockwise, producing a gentle curving geometry at the
lower sandclay contact that is similar to normal-drag folds. This
fault evolution produces a stable, planar deformation zone with an
unusually thick clay gouge.
3.2.3. Effect of multiple layers and layer thickness
Experiment M-1-s-tp has four 12 mm thick soft-clay layers
alternating with 8 mm thick sand layers (Fig. 12, Table 1). The
fault-zone evolution in this experiment is similar to the reference
experiment with only sand (Fig. 4). The initially steep precursor
faults form a restraining zone across the sandclay sequence with
minor fault segmentation. Deformation in this sequence evolves
into the kinematically favored plane. Each clay layer forms
a continuous clay smear in the layered gouge, but the sand
between the clay layers thins and in some cases becomes bou-
dinaged (Fig. 12). A similar setup with four soft clay layers but
variable layer thickness (Fig. 13, Table 1, experiment M-2-s-tp)
illustrates the effect of layer thickness on fault-zone evolution.
Here, the total thickness of the clay-rich interval is more than in
the previous experiment (75 vs. 62 mm), and the claysand ratio is
also higher (3:2 vs. 2:3). The two initially steeply dipping
precursor faults connect in a zone of diffuse deformation with
a shallower dip than the initial segments. Localization of defor-
mation occurs on both sides of this zone with a less deformed lens
in between. With progressive deformation the sheared sand in the
fault zone becomes discontinuous as individual, sheared clay
layers coalesce.
4. Discussion
4.1. Effect of boundary conditions on fault-zone evolution
Faults in nature form under a variety of local boundary condi-
tions, but in many cases those boundary conditions might be less
than those imposed here (Schmatz et al., 2010). In this study we
explored normal fault development in weak materials layered
between two rigid layers faulted with the same 70
dip and lying in

the same plane. The resulting fault evolution differs dramatically
fromexperiments with a rigid lower boundary condition and a free
upper surface.
Fig. 13. (a) Image sequence showing experiment M-2-s-tp with four soft-clay layers
and top plates (see inset). Basement fault offset is 4, 15 and 55 mm. Red lines indicate
location of rigid blocks at bottom and top. (b) Image sequence showing the corre-
sponding PIV overlay to the image sequence in (a) with the contour plot of incremental
rotation eld and the velocity vector eld. (For interpretation of the references to
colour in this gure legend, the reader is referred to the web version of this article.)
a b
Fig. 14. Sketch of the rst-order differences in kinematics between experiments
containing soft (a) and rm (b) clay layers.
Imposing the rigid, faulted top boundary strongly controlled the
geometry of the resulting shear zone, producing kinematics similar
to those in a direct shear experiment (Zhang and Thornton, 2007).
Using geotechnical nomenclature, our setup is dened as a slanted
direct shear experiment. Post-mortem excavation of the fault
zones revealed few edge effects due to frictional drag along the
glass panels on the evolution of faults (Souloumiac et al., 2009), but
further study is ongoing.
As expected from previous work, the rst localization of defor-
mation along precursor faults started at the intersection of the
faults in the basement and top blocks. The position of these initial
faults is controlled by the stress elds around the fault tips, and
they rotate either continuously or in steps into the kinematically
favorable plane dened by the basement and top fault plane. In our
earlier experiments with a free top surface, this effect was only
present in the bottom part of the model. To some degree, the
experiments with the top blocks represent a symmetrical doubling
of the free-surface experiments. Localization from the top fault is
delayed from the initial basement offset as stresses propagate from
the basement upward. The kinematics of our experiments in some
cases resemble the Quadshear model (Welch et al., 2009) in which
a ductile interval sandwiched between two brittle layers becomes
sheared, although these models are purely kinematic and contain
no mechanical properties.
Reproduction of experiments with and without a top-plate
boundary condition (Figs. 3 and 5) reveals that the top plate
suppresses but fails to eliminate initial fault branching. The top
plate results in less segmentation and a less complex fault zone in
general (Childs et al., 1996; Van der Zee and Urai, 2005). A single,
planar fault zone forms earlier in the experiment when a top plate
is present, but the continuous history of deformation during early
displacement seems signicant to us, even though it is abbreviated
with the top plate. The folding that arises from the migration of
a shear surface appears similar to a normal-drag fold, but its
formation is different than the processes expected when a single,
planar fault propagates into a thick, layered sequence, such as the
slightly downwards-widening shear zone in Fig. 4.
4.2. Competence contrast
In our models, normal faulting starts from the tip of a preexist-
ing fault in a competent brittle bed (basement) where the overlying
layers accommodate a certain amount of strain (Mandl, 2000). In
experiments with soft and normally consolidated clay, the strain is
continuous without the formation of fractures, at least at the scale
of observation. In all cases, this produces continuous smear along
the fault surface in our 2D sections along the glass plate. These
observations agree with earlier studies by Sperrevik et al. (2000)
and Van der Zee and Urai (2005). Clay containing a small amount of
Portland cement (0.25 wt.% in soft clay) deforms in this same
manner because the increase in strength is insignicant compared
to the mean effective stress (van Gent et al., in press). Even normal
clay layers up to 10 mm thick have minimal effect on initial fault
localization, which we interpret to result from the lowcompetence
contrast between sand and clay. In experiments with a thick, soft-
clay layer (Figs. 11 and 13) the initial faults are longer-lived, leading
to the formation of a restraining step in the clay layer that
progressively thins out (Fig. 14). The clay in the restraining step is
strongly deformed until a new fault forms with a releasing geom-
etry, forming a nal, thick clay gouge. This structure is similar to
that in Fig. 10 of Van der Zee and Urai (2005) for which a similar
kinematic evolution was proposed. The early evolution is also
consistent with the DEM results of Egholm et al. (2008), although
no evidence of the overstepping process is presented. A full
mechanical explanation of the parameters that control this over-
stepping is the aim of an ongoing study using numerical and
experimental methods.
In experiments with very hard clay (rm, over-consolidated clay
or cemented clay), the fault-zone geometry is affected by fractures
in the clay that formin areas of high curvature as the layers are bent
(Figs. 9 and 10). In this case, a releasing relay zone forms between
the two faults propagating from the tips of the faults in the top and
bottomblocks (Fig. 14). This result compares with models by Mandl
(2000) and Lehner and Pilaar (1997) in which initial monoclinal
bending of the clay is interpreted to have changed the orientation
Fig. 15. PIV analysis of the experiment in Fig. 10, showing the formation of asperities and lens structures. Basement fault offset is 4, 6 and 13 mm. Red lines indicate location of rigid
blocks at bottom and top. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
and magnitude of the principal stresses. Changes in the stress
tensor allow the formation of secondary slip planes that cause
internal segmentation of the clay layer. Scho pfer et al. (2006)
investigated the deformation of multilayers under similar
boundary conditions using DEM techniques in materials where the
strong layers undergo tensile failure and the weaker layers deform
by faulting. They observed that, after the formation of monoclines,
the fault zone propagates in a staircase-like manner, with vertical
tensile fractures in the strong layers connected by shear fractures in
a complex fault zone. Although the details of the material proper-
ties are quite different, their results are best compared with our
results of cemented clay layers. The striking difference is that, in our
case, each strong layer develops two fractures, one in the plane of
the principal fault and the other along the plane of the initial
precursor fault. In the models of Scho pfer et al. (2006), most layers
initially develop only one fracture, but a second fracture develops in
the layer closest to the stiff basal layer, followed by block rotation of
a clay fragment, similar to the observations in our experiments.
In eld observations of normal faults by Van der Zee et al. (2008),
tensile fractures in the brittle layers are also attributed to the
evolution of the fault zone. In this case, however, the fractures are
interpreted to have formed before faulting and in different orien-
tations than the fault, adding additional geometric complexity.
One additional, important process is the reworking of brittle
clay fragments in the fault zone (Lindsay et al., 1993; Holland et al.,
2006, and others). In our experiments (Fig. 10), reworking requires
no more than a modest competence contrast between sand and
clay because of the relatively small displacements. Because abra-
sion is an irreversible process, we expect this process becomes
more important for faults with larger displacements (i.e. 10s of
meters).
4.3. Lens formation
Branches in fault zones dening a lens (or horse) are commonly
observed in both nature and experimental studies (Ramsay and
Huber, 1987; Childs et al., 1996, 1997; Walsh et al., 1999; Van der
Zee and Urai, 2005; Lindanger et al., 2007). In addition to fault
lenses caused by segment linkage, we observe lenses that formed
by asperity bifurcation (Lindanger et al., 2007). The formation of
asperities is common in most rm clay and cemented clay exper-
iments. As shown in Fig. 15 (experiment in Fig. 9, Table 1, experi-
ment C1-6-s-tp), these lenses evolve considerably through time.
Asperities are eroded progressively while continued localization
causes the formation of new asperities. Asperity-related lenses
may thin and tighten while new lenses are forming. The model of
Van der Zee (2002) suggests that material inside the lens becomes
more strongly deformed than the surrounding rock. The formation
of fault-bounded lens structures in our experiments is related to
the initial pattern of localization and the evolution of segmenta-
tion. In experiment C2-4-s-tp (Fig. 8, Table 1), a wide lens develops
after coalescence of the initial zones of localization. Although the
lens becomes eroded by deformation, its interior remains undis-
turbed as indicated by PIV (Fig. 8b). Clearly, the evolution of
deformation in fault-bounded lens structures is more complicated
than suggested by the models of Van der Zee and Urai (2005) and
Lindager et al. (2007).
4.4. Clay smear
According to Fulljames et al. (1997), the amount of clay smear at
a point on the fault reduces with distance to the source bed. In their
interpretation the smear forms a layered gouge containing clay
from each source bed: the greater the number and thickness of the
source beds, within the throw window, the greater the thickness of
Fig. 16. Photograph of experiment M-1-s-tp showing: 1) The SGR measuring method
after Fristad et al. (1997) applied to our experiments, with calculation point cp. SGR
P
shale bed thickness (h)/fault throw 100% (31% in this example); gouge clay
P
clay band thickness/shear zone width 100% (62% in this example); 2) a simple-
shear overlay on the lowermost clay layer indicating the accumulation of clay in the
shear zone; and 3) three selected areas of the fault zone enhanced to increase contrast.
The original clay layer shows much less variance in color than the material in the fault
zone, and where the color variance is proportional to the grain size of the sand
(wavelength). This is interpreted to be a result of a mixture of sand and clay in the fault
zone. (For interpretation of the references to colour in this gure legend, the reader is
the smear. Lindsay et al. (1993) distinguished among three types of
clay smear: (i) abrasion due to movement past sandstones,
(ii) shearing and ductile deformation between hanging wall and
footwall cut-offs of shale beds, and (iii) injection of clays during
uidization. Our experiments show that clay smear in the form of
a continuous clay band on the fault surface is caused by fault
segmentation followed by shearing (e.g. Figs. 7, 12 and 13). Asperity
abrasion and reworking of the clay also contribute to the formation
of clay smear (Fig. 10). Clausen and Gabrielsen (2002) investigated
clay smear processes with the help of a ring-shear apparatus and
concluded that the potential for developing a continuous clay
membrane increases as: (1) the normal stress increases, (2) the
water content of clay increases, and (3) the shear strength of clay
decreases. Our results are in agreement with these three inferences.
(1) Addition of the top plate causes continuous clay smear in
a model where clay fractures and becomes discontinuous without
the top plate. Although in these two experiments (Figs. 5 and 6) the
vertical stress at the top of the clay layer is the same, the presence
of the stiff aluminum plate, which resists bending, acts to increase
the mean stress in the region of the fault zone. (2) All soft-clay
experiments show continuous clay membranes. (3) Measurements
of shear strength (Schmatz et al., 2010) show the lowest shear
strength for soft clay.
4.5. Shale-gouge ratio (SGR)
The SGR method estimates the proportion of clay incorporated
into the fault gouge as a function of the sand/shale ratio which
moved past a point on a fault (Crawford et al., 2002). The shale
thicknesses are measured in a window with a height equal to the
throw (e.g., Yielding et al., 1997). We calculated SGR according to
this algorithm (Figs. 16 and 17) in each of the experiments in Table
1. The amount of gouge clay in the fault zone is dened as the clay
thickness as a percentage of the total shear zone thickness as
dened in PIV analyses (Fig. 16). SGR plotted against gouge clay in
the fault zone for most of our experiments (Table 1) illustrate that
there is no simple correlation between the two parameters (Fig. 17),
and in most cases there is either more or less clay in the fault than
expected from an SGR calculation. Normally consolidated clay is
both enriched and depleted in the fault. Soft clay leads to more clay
within the gouge zone than expected fromSGR (for 79% of the cases
there is more clay in the fault than expected from SGR). Brittle clay
tends to be depleted, albeit with a large scatter (in 85% of the
experiments, there is less and commonly far less clay in the fault
zone than expected from SGR). However, SGR yields only an
average of the amount of clay expected within the fault zone (Van
der Zee and Urai, 2005) and if all clay types are present, the average
SGR value will result. However, in settings where one clay type
dominates, there may be signicant bias in the calculated SGR
value. Moreover, the experiments illustrate large local variability
around the calculated SGR value within a consistent throwwindow.
4.6. Mixing
Fault zones typically exhibit heterogeneous, non-coaxial defor-
mation (Mair and Abe, 2008). In the case of simple shear, the
geometry of clay or sand within the fault zone is derived from the
fault-zone width, the original layer thickness, and the fault angle. A
simple-shear overlay on the nal gouge zone of experiment M-1-s-
tp (Fig. 16) shows that clay-gouge zone is thicker than expected
from simple shear, although there is no evidence of lateral clay
injection from the source layer (Van der Zee et al., 2003). One
process that contributes more clay to fault gouge than expected
from simple shear is mechanical mixing, in which clay moves into
the porosity of the sheared sand (Van der Zee, 2002). Clausen and
Gabrielsen (2002) reported three stages in the development of clay
membranes from ring-shear experiments. In their rst stage cor-
responding to low normal stress (6 kPa), clay membranes are
completely absent. Only occasional clay fragments occur in the fault
zone. In the second stage at higher normal stress (>25 kPa),
a mixture of sand and clay or patches of clay embedded in a sand
matrix occur. In the third stage, a semi-continuous clay membrane
develops (normal stress > 100 kPa). They inferred that parts of the
fault zone characterized by a claysand mixture are stable and
develop no other types of smear. They also speculated that mixing
of clay and sand occurs by local invasion of clay into sand pores
generated in dilation at an early stage of shear under low normal
stress. Clausen and Gabrielsen (2002) concluded that, after rupture
of the clay layers, the fault zone is dominated by plastic deforma-
tional processes, which include transport of single grains from sand
into the clay. Crawford et al. (2002) also reported on the generation
of anastomosing networks of microsmears around framework
grains or clasts. In our experiments mixtures of sand and clay
within the gouge zone are also present. A detailed investigation of
digitally enhanced high-resolution digital images (Fig. 16, experi-
ment M-1-s-tp) shows that clay in the shear zone differs in
composition from the undeformed source clay. A fault gouge con-
sisting of sand, clay and sandclay mixture replaces the initial
sheared layers of sand and clay.
4.7. Incipient faulting
Understanding localization in the layered sandclay experi-
ments requires a clear understanding of localization in a pure sand
experiment. Assuming that the initial deformation in our models is
non-localized, we sought evidence for this with PIV. We carefully
processed the results of the rst few photographs of a number of
experiments, plotting the velocity eld for basement fault
displacements up to 0.8 mm (Fig. 18). In the rst example we show
the incremental displacement vector eld in the lower 6 cm of
a sand-only experiment (Fig. 4). Here, there is initially a concentric
Fig. 17. Graph of SGR measured in all experiments (numbers refer to experiment
number in Table 1) plotted against the percentage of clay (gouge clay) in the fault zone.
vector eld centered on the tip of the basement fault, as predicted
in the analytical models of Patton and Fletcher (1995) for an elastic
material. Ongoing deformation makes the vector eld much more
discontinuous, reecting the initial localization of strain. Experi-
ment N-2-s-tp (Fig. 11) shows a similar early evolution. The semi-
concentric vector eld is observed in the initial phase, but with
some decoupling across the clay layer. With ongoing deformation,
localization is much more rapid in the sand below the soft-clay
layer.
4.8. Application to fault zones in nature
The results of our experiments simulate the formation of cm-
scale faults in saturated sandclay sequences with minimal over-
burden under very specic boundary conditions and therefore,
extrapolation to fault-zone evolution in layered sequences in
nature is complicated. True scaling would require knowledge of
model-to-nature ratios of length, density, gravity, and cohesion;
however, in nature these parameters exhibit a wide range of values.
Upscaling to faults cutting lithied sequences is even more prob-
lematic because, at this scale, the fault-zone widths (controlled by
the size of the sand grains) will likely be much larger than for
upscaled faults.
Nonetheless our experiments qualitatively compare with
nature when only minor length upscaling is required, such as for
small faults in thinly bedded sandclay sequences like those
described in Van der Zee and Urai (2005) (Fig. 19; see full
description in caption). Here, many of the structures in our
multilayers (Figs. 12 and 13, cf. Figs. 6, 9, 10, 14 of Van der Zee and
Urai, 2005) are similar, suggesting that the differences in
mechanical properties play a relatively minor role in controlling
the evolution. Considering constitutive properties, for both soft
and hard clay end-members embedded in sand, the patterns of
deformation are quite distinct, and seem to be reasonably consis-
tent, in experiment and in numerical simulation (Mair and Abe,
2008; Scho pfer et al., 2006; Egholm et al., 2008). This suggests that
these patterns are independent of small variations in material
properties and are therefore reasonably robust and might be
expected in nature, too.
The effect of different boundary conditions on how the
structures evolved is the most difcult to evaluate based on our
results. The initially segmented nature of normal faults has been
extensively documented, and Van der Zee et al. (2003) sug-
gested that it is in the segment links where the presence of soft
or hard clay layers causes the strongest deviations from a simple
deformation band. Results of this study show, in agreement with
earlier work (e.g. Childs et al., 1996; Scho pfer et al., 2006;
Egholm et al., 2008), that the presence of a mechanical stra-
tigraphy has a strong control on the location of segment linkage.
However, the change in importance of the precursor faults with
increasing burial and length scale is unclear and requires addi-
tional study.
Fig. 18. Images showing the incremental vector eld of the initial stages of experiments with top plates: (a) experiment P-2-tp and (b) experiment N-2-s-tp (a 30 mm thick soft-clay
layer), with the point of observation at the tip of the basement fault. The vector elds are for basement fault offsets of 0.4 and 2 mm. The vector eld for the clay layer is invalid and
covered with a transparent, blue mask. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
5. Conclusions
1. The mechanical properties of clay have a major effect on clay-
gouge evolution during fault-zone development in sandclay
sequences.
2. The structure of initially localized deformation evolves into
a kinematically favorable fault-zone structure. This evolution
has a major role in controlling fault-zone structures (e.g.,
releasing and restraining relay zones).
3. A high competence contrast between sand and clay leads to
a more complex fault zone due to the formation of secondary
localized zones and segmentation-induced fault lenses. A high
competence contrast also promotes coeval strain partitioning
between multiple shear zones.
4. Weak clay layers are preferentially enriched in normal fault
zones, whereas strong, brittle clay initially fractures and forms
clay boudins that rotate in the deforming sand. With progres-
sive deformation these boudins are abraded and transformed
into a soft-clay gouge.
5. Empirical methods such as SGR used to predict fault seal can be
improved by taking into account the mechanical properties of
the faulted rocks.
6. The low permeability clay gouge thickens with fault displace-
ment due to mixing of sand and clay.
Acknowledgements
This research would not be in its present state without the
contributions of a number of persons. In particular the authors
would like to thank M. Holland, S. Giese, M. Ziegler, W. van der Zee,
J. Wilden and K. Nienhaus. B. Maillot, A. Henza and especially R.
Schlische are acknowledged for their constructive reviews and
editorial help. This work was funded by ExxonMobil and, as part of
the ongoing research project UR 64/10-1, is supported by the
German Science Foundation DFG.
References
Adam, J., Urai, J.L., Wieneke, B., Oncken, O., Pfeiffer, K., Kukowski, N., Lohrmann, J.,
Hoth, S., van der Zee, W., Schmatz, J., 2005. Shear localisation and strain distri-
bution during tectonic faulting-new insights from granular-ow experiments
and high-resolution optical image correlation techniques. Journal of Structural
Geology 27, 283301.
Antonellini, M.A., Aydin, A., Pollard, D.D., 1994. Microstructure of deformation
Geology 16, 941959.
Cardozo, N., Bhalla, K., Zehnder, A.T., Allmendinger, R.W., 2003. Mechanical models
of fault propagation folds and comparison to the trishear kinematic model.
Journal of Structural 25, 118.
Childs, C., Nicol, A., Walsh, J.J., Watterson, J., 1996. Growth of vertically segmented
normal faults. Journal of Structural Geology 18, 13891397.
Childs, C., Walsh, J.J., Watterson, J., Mller-Pedersen, P., 1997. Complexity in fault
zone structure and implications for fault seal prediction. In: Mller
Pedersen, P., Koester, A.G. (Eds.), Hydrocarbon Seals. NPF special publication 7.
Elsevier, Amsterdam, pp. 6172.
Clausen, J.A., Gabrielsen, R.H., 2002. Parameters that control the development of
clay smear at low stress states: an experimental study using ring-shear appa-
ratus. Journal of Structural Geology 24, 15691586.
Crawford, B.R., Myers, R.D., Woronow, A., Faulkner, D.R., Rutter, E.H., 2002. Porosity
Permeability Relationships in Clay-bearing Fault Gouge. Society of Petroleum
Engineers, Irwing, Texas, 78214, pp. 13.
Egholm, D.G., Clausen, O.R., Sandiford, M., Kristensen, M.B., Korstgrd, J.A., 2008.
The mechanics of clay smearing along faults. Geology 36, 787790.
Erslev, E.A., 1991. Trishear fault-propagation folding. Geology 19, 617620.
Ferrill, D.A., Morris, A.P., 2003. Dilational normal faults. Journal of Structural
Geology 25, 183196.
Fristad, T., Groth, A., Yielding, G., Freeman, B., 1997. Quantitative fault seal prediction:
a case study from Oseberg Syd. In: MllerPedersen, P., Koester, A.G. (Eds.),
Hydrocarbon Seals. NPF special publication 7. Elsevier, Amsterdam, pp. 107124.
Fulljames, J.R., Zijerveld, L.J.J., Franssen, R.C.M.W., 1997. Fault seal processes:
systematic analysis of fault seals over geological and production time scales. In:
Moeller-Pedersen, Koester, A.G. (Eds.), Hydrocarbon Seals. NPF special publi-
cation 7. Elsevier, Amsterdam, pp. 5159.
van Gent, H.W., Holland, M., Urai, J.L., Loosveld, R., Evolution of fault zones in
carbonates with mechanical stratigraphy insights from scale models using
layered cohesive powder. Journal of Structural Geology: Special publication., in
press.
Holland, M., Urai, J.L., van der Zee, W., Stanjek, H., Konstanty, J., 2006. Fault gouge
evolution in highly overconsolidated claystones. Journal of Structural Geology
28, 323332.
Horseld, W.T., 1977. An experimental approach to basement-controlled faulting.
Geologie en Mijnbouw 56, 363370.
James, W.R., Fairchild, L.H., Nakayama, G.P., Hippler, S.J., Vrolijk, P.J., 2004. Fault-seal
analysis using a stochatic multifault approach. AAPG Bulletin 88, 885904.
Lehner, F.K., Pilaar, W.F., 1991. On a mechanism of clay smear-emplacement in-
synsedimentary normal faults (abstract). AAPG Bulletin 75, 619.
Lehner, F.K., Pilaar, W.F., 1997. The emplacement of clay smears in synsedimentary
normal faults: inference from eld observations near Frechen, Germany. In:
Moeller-Pedersen, Koester, A.G. (Eds.), Hydrocarbon Seals. NPF special publi-
cation 7. Elsevier, Amsterdam, pp. 3950.
Lindanger, M., Gabrielsen, R.H., Braathen, A., 2007. Analysis of rock lenses in
extensional faults. Norwegian Journal of Geology 87, 361372.
smears on fault surfaces. Special Publications. International Association of
Sedimentologists 15, 113123.
Mair, K., Abe, S., 2008. 3D numerical simulations of fault gouge evolution during
shear: grain size reduction and strain localization. Earth and Planetary Science
Letters 274, 7281.
Mandl, G., 2000. Faulting in Brittle Rocks. Springer, London, pp. 434.
Patton, T.L., Fletcher, R.C., 1995. Mathematical block-motion model for deformation
of a layer above a buried fault of arbitrary dip and sense of slip. Journal of
Peacock, D.C.P., Sanderson, D.J., 1992. Effects of layering and anisotropy on fault
geometry. Journal of the Geological Society of London 149, 793802.
Ramsay, J.G., Huber, M.I., 1987. The Techniques of Modern Structural Geology. In:
Folds and Fractures, vol. 2. Academic Press, London, pp. 391.
Schmatz, J., Holland, M., Giese, S., van der Zee, W., Urai, J.L., 2010. Clay smear
processes in mechanically layered sequences results of water-saturated model
experiments with free top surface. Journal of the Geological Society of India
(special issue 03)
Fig. 19. Field examples of normal faults formed at approximately 800 m in a deltaic sandclay sequence near Miri, Sarawak, Malaysia. (a) Complex clay-rich gouge in a fault with an
offset of 2.5 m. The sheared clay and sand develops into a banded fault gouge; coin for scale. (b) Small-offset fault in thinly layered sand and clay with thin clay gouge (lens cap for
scale). See Van der Zee and Urai (2005) for further information.
Scho pfer, M.P.J., Childs, C., Walsh, J.J., 2006. Localisation of normal faults in multi-
layer sequences. Journal of Structural Geology 28, 816833.
Souloumiac, P., Leroy, Y.M., Maillot, B., Krabbenhft, K., 2009. Predicting stress
distributions in fold-and-thrust belts and accretionary wedges by optimization.
Journal of Geophysical Research 114 B09404.
Sperrevik, S., Faerseth, R.B., Gabrielsen, R.H., 2000. Experiments on clay smear
formation along faults. Petroleum Geoscience 62, 113123.
Steingart, D.A., Evans, J.W., 2005. Measurements of granular ows in two-dimen-
sional hoppers by particle image velocimetry. Part I: experimental method and
results. Chemical Engineering Science 60, 589598.
Urai, J.L., Nover, G., Zwach, C., Ondrak, R., Scho ner, R., Kroos, B.M., 2008. Transport
processes. In: Littke, R., Bayer, U., Gajewski, D., Nelskamp, S. (Eds.), Dynamics of
complex intracontinental basins: The Central European Basin System. Springer,
Berlin Heidelberg, pp. 367388.
Van der Zee, W., 2002. Dynamics of fault gouge development in Layers sandclay
sequences. Shaker Verlag, Aachen, PhD thesis, pp. 155.
Van der Zee, W., Urai, J.L., Richard, P.D., 2003. Lateral clay injection into normal
faults. GeoArabia 8, 501522.
Van der Zee, W., Urai, J.L., 2005. Processes of normal fault evolution in a siliciclastic
Geology 27, 22812300.
Van der Zee, W., Wibberley, C.A.J., Urai, J.L., 2008. The inuence of layering and pre-
existing joints on the development of internal structure in normal fault zones:
the Lodeve basin, France. In: Wibberley, C.A.J., Kurz, W., Imber, J.,
Holdworth, R.E., Collerrini, C. (Eds.), The Internal Structure of Fault Zones:
Implications for Mechanical and Fluid-Flow Properties. Geological Society of
London Special Publications 299, pp. 5774.
Walsh, J.J., Watterson, J., Bailey, W., Childs, C., 1999. Fault relays, bends and branch-
lines. Journal of Structural Geology 21, 10191026.
Weber, K.J., Mandl, G., Pilaar, W.F., Lehner, F.K., Precious, R.G., 1978. The role of faults
in hydrocarbon migration and trapping in Nigerian growth fault structures. In:
Proc. 10th Annual Offshore Technology Conference, Houston, Texas 4, 2643
2653.
Welch, M.J., Knipe, R.J., Souque, C., Davies, R.K., 2009. A Quadshear kinematic model
for folding and clay smear development in fault zones. Tectonophysics,
186202.
Wolf, H., Konig, D., Triantafyllidis, T., 2003. Experimental investigation of shear
band patterns in granular material. Journal of Structural Geology 25, 1229
1240.
Yielding, G., 2002. Shale Gouge Ratio calibration by geohistory. In:
Koester, A.G.H.R. (Ed.), Hydrocarbon Seal Quantication. NPF Special Publication
11. Elsevier, Amsterdam, pp. 115.
AAPG Bulletin 81, 897917.
Zehnder, A.T., Allmendinger, R.W., 2000. Velocity eld for the trishear model.
Zhang, L., Thornton, C., 2007. A numerical examination of the direct shear test.
Geotechnique 57, 343354.
An experimental investigation of the development and permeability of clay
smears along faults in uncemented sediments
F. Cuisiat
*
, E. Skurtveit
Norwegian Geotechnical Institute, P.O. Box 3930, Ullevaal Stadion, N-0806 Oslo, Norway
Article history:
22 November 2009
Accepted 9 December 2009
Keywords:
Laboratory
Clay smear
Fault
Permeability
Ring shear
a b s t r a c t
The generation of clay smears along faults in uncemented sediments has been studied through labora-
tory experiments in a newly developed high stress ring shear apparatus. The main objective is to
investigate basic mechanisms involved in the deformation process of sediments during faulting and
formation of clay smears.
The experimental test program comprises ring shear tests on sand with embedded clay segments
(sandclay sequence) under constant effective normal stress. Visual inspection of the samples after
testing, analyses of thin sections and permeability measurements across the shear zone are used to
characterise geometrical continuity, thickness and sealing potential of the smear. Deformation processes
such as grain reorientation, clay smear and cataclasis are identied from the tests. The complexity of the
shear zone is observed to increase with the effective normal stress applied to the specimen and the
number of clay segments used in the ring (multilayered sandclay sequences). At low effective normal
stress, in clay-rich sediments, clay smear is the most efcient mechanism for permeability reduction. The
permeability across the smear decreases with ring rotation (or shear displacement) and effective normal
stress. A maximum decrease of two orders of magnitude compared to the permeability of the
surrounding sand is observed after 90
rotation under 10.5 MPa effective normal stress. Sandsand

juxtaposition shear is dominated by grain rolling causing only minor permeability reduction. At high
effective normal stress, permeability measurements across clay smear and sandsand juxtaposition yield
similar values indicating that the permeability reduction is dominated by grain size reduction in the
sand.
1. Introduction
Faults play a major role in the distribution and movement of
uids, hydrocarbons or groundwater in the Earths crust. Faults
may act as barriers, bafes, or mixed conduit-barrier systems to
uid ow depending on the composition of the faulted rock, the
stress conditions at the time of faulting, and post-faulting burial
and temperature history (Antonellini et al., 1994; Caine et al.,
1996; Sibson, 2000). An understanding of fault zone internal
structure and associated hydraulic properties is a requisite to
many elds such as petroleum exploration and production,
groundwater modelling, earthquake rupture, and sequestration of
CO
2
(e.g. Hickman et al., 1995; Sibson, 2000; Bense and Van Balen,
2004; Rutqvist et al., 2007; Annunziatellis et al., 2008; Wibberley
et al., 2008).
Faults in sediments can be classied upon the phyllosilicate
content of their host lithology and burial depth at time of faulting
(Fisher and Knipe, 1998, 2001). Faults in clean sandstones (<15%
clay) at shallow depth produce a disaggregation zone with only
local grain rearrangement, no grain-fracturing and similar or even
higher permeability than the host rock (Fisher and Knipe, 2001;
Bense and Van Balen, 2004). Faulting of the same sediments at
greater depth results in grain-fracturing (cataclasis) with clogging
of the pore space by smaller grain fragments. Experimental data
show that the onset of grain-fracturing may start at depths as low
as 500 m, or 5 MPa effective vertical stress (Chuhan et al., 2002).
Post-faulting burial may lead to quartz cementation for tempera-
tures greater than circa 90

C or circa 3 km in basins with normal
temperature gradient. Quartz cementation occurs most intensively
at burial depths between 3.5 and 5 km (120170
C) (Bjo rlykke and

Ho eg, 1997). In impure sands with clay content between 15 and
40%, or other contexts of similar mineralogy, faulting leads to a fault
gouge or phyllosilicate framework fault rock (Fisher and Knipe,
2001) with mixing of sand and clay often structured parallel to the
E-mail address: fabrice.cuisiat@ngi.no (F. Cuisiat).
doi:10.1016/j.jsg.2009.12.005
shear plane (Rutter et al., 1986; Wibberley and Shimamoto, 2003;
van der Zee and Urai, 2005).
In clay-rich sequences (i.e. clay content higher than 40%),
faulting is commonly associated with clay smears, that is shearing
of an offset clay layer into the fault zone (Fig. 1).
Smearing of low permeability clay has been presented as one of
the most efcient mechanism for fault sealing. From outcrop
observations, laboratory experiments, and numerical modelling
(e.g. Lindsay et al., 1993; Lehner and Pilaar, 1997; van der Zee and
Urai, 2005; Sperrevik et al., 2000; Clausen and Gabrielsen, 2002;
Egholm et al., 2008) several processes have been suggested to
explain the occurrence of clay smear such as: clay abrasion, lateral
clay injection from source layer and shearing within fault, and
material instabilities. Nevertheless, no real mechanics-based
predictive model for fault seals is yet available in the literature.
In this paper, we present the results of ring shear experiments to
study the development and permeability of clay smear along faults.
The use of a ring shear apparatus in geosciences is fairly wide-
spread, for instance to characterise the residual shear strength and
behaviour of soils and granular materials as well as the interaction
between soil and structure (e.g. Bishop et al., 1971; Tika et al., 1996;
Lupini et al., 1981; Hungr and Morgenstern, 1984; Sassa et al.,
2003). However, the use of a ring shear device as an analogue for
fault formation and clay smear is more limited, although some early
work was conducted by Mandl et al. (1977). The principle is illus-
trated in Fig. 2. The ring shear sample consists of an annular sand
specimen conned between upper and lower rings. One or several
segments of clay are embedded within the sand. After applying the
vertical (normal) stress s
n
onto the specimen, the lower ring is
rotated, thus dragging the clay material onto the shear plane
located at the separation of the upper and lower conning rings.
The linear displacement along the shear plane corresponds to
a throw on a fault plane, while the width of the embedded clay
segment corresponds to the thickness of a clay sequence through
which the fault develops. By performing several tests with different
conditions (effective normal stress, maximum rotation, sand
density, clay type, etc.) parameters may be varied to assess their
effect on the development of the shear band and clay smear
properties.
Early experimental studies dedicated to clay smear and shear
band formation carried out at the Norwegian Geotechnical Institute
(NGI) were performed in a classical geotechnical ring shear appa-
ratus which allowed for large deformations, but was limited to low
stress levels equivalent to ca. 50 m burial under hydrostatic
conditions (Sperrevik et al., 2000; Clausen and Gabrielsen, 2002;
Kvaale, 2002).
More recently, a new ring shear apparatus was designed and
constructed at NGI to investigate shear band formation and clay
smear in unconsolidated sediments at greater burial depths (Torabi
et al., 2007; Cuisiat et al., 2007). The new equipment allows for
normal effective stresses up to 20 MPa, which correspond to
a depth of ca. 2600 m under hydrostatic conditions or even higher
in over-pressured reservoirs. These depth ranges cover the depth at
which deformation (faulting) may have occurred for most North
Sea reservoirs. In the absence of carbonate cementation, North Sea
sediments remain loosely cemented until depths of 2.53 km until
diagenesis takes place (Bjo rlykke and Ho eg, 1997).
In the new ring shear apparatus, the ow resistance and
permeability can be measured across the specimen through 48
drainage points evenly distributed around the upper and lower
rings (Fig. 2 left) during testing. The possibility of running perme-
ability tests across the sheared specimen greatly improves the
quantication of the sealing potential of clay smear and other
deformation processes, and how this potential varies with normal
stresses and shear displacement (i.e. fault offset) for different
sediment types.
The main motivation of the work described in this paper is to
pursue the experimental work initiated earlier (Cuisiat et al., 2007;
Torabi et al., 2007) in order to increase our database of fault
properties. The work focuses on the mechanisms associated with
clay smear along faults in uncemented sediments as well as the
associated permeability changes. The nal aim is to develop an
experimental database for quantication of fault texture, petro-
physical properties, deformation and strength, and uid ow
properties of faults in unconsolidated sediments.
2. Clay smears: observation and prediction
Experimental studies (Weber et al., 1978; Sperrevik et al., 2000;
Clausen and Gabrielsen, 2002; Karakouzian and Hudyma, 2002) as
well as outcrop studies of clay smears in loose sandstone-shale
sequences (Weber et al., 1978; Lehner and Pilaar, 1997; Doughty,
2003), and lithied sandstoneshale sequences (Lindsay et al.,
1993), suggest that clay smearing is strongly dependent on the
original thickness of the shale beds, the competency contrast
between the shale and non-shale lithologies, the effective stress
(i.e. burial depth), and the magnitudes and rates of faulting. Thick
and weak source beds produce the thickest smears.
Conceptual models based on simple mechanical considerations
have been used to explain clay smearing (Lehner and Pilaar, 1997;
Mandl, 2000; Koledoye et al., 2003), and no mechanics-based
predictive tool is yet available to model the processes associated
with clay smear due to the difculties associated with the
complexity of the mechanisms (strain localisation, large displace-
ment, extrusion, loading rates). Recent developments include the
work by Gudehus and Karcher (2007) using hypoplasticity, and
Cardozo and Cuisiat (2008) on fault-propagation folding in strati-
ed media.
Practical predictive models are usually empirically based, by
relating the potential for clay smearing to properties easily
measured inwell logs and seismic sections, such as amount of shale
and fault throw across the sedimentary sequence (Yielding et al.,
1997). The most common ones are the clay smearing potential (CSP,
Bouvier et al., 1989), shale smear factor (SSF, Lindsay et al., 1993),
and shale gouge ratio (SGR, Yielding et al., 1997).
- The clay smear potential (CSP) (Bouvier et al., 1989) is given by
CSP cT
2
c
=d where T
c
is the thickness of the clay source bed,
Fig. 1. Deformation bands in sandstone formation on the east side of Courthouse
Canyon, Utah. Early soft-sediment deformation without cataclasis. Coin size is about
2 cm in diameter.
F. Cuisiat, E. Skurtveit / Journal of Structural Geology 32 (2010) 18501863 1851
d is the distance from the source bed, and c is a constant that
needs to be calibrated. The clay smear thickness can be
expected to correlate with CSP as shown by mechanical
considerations (Egholm et al., 2008).
- The shale smear factor (SSF) (Lindsay et al., 1993) is the fault
throw t divided by the vertical thickness of the offset shale
source layer d
cl
, i.e. SSF t=d
cl
.
- The shale gouge ratio (SGR) (Yielding et al., 1997) is the net
clay percentage in the slipped interval, i.e.
SGR
P
V
cl
Dz
t
100% where t is the local throw of the fault
and V
cl
is the clay volume factor in every interval Dz along the
fault.
The previous parameters should be calibrated against well and
outcrop data that document fault sealing. Based on an outcrop
study of small faults in sandstoneshale sequence (with typical
throws less than 1 m), Lindsay et al. (1993) concluded that with SSF
less or equal to 7 continuous shale smears are expected, and with
SSF less or equal to 11 continuous shale smears are probable.
Faerseth (2006) found that continuous smear along large faults can
develop when the shale smear factor (SSF), given by the fault throw
divided by the thickness of the shale source layer, is less or equal to
4. However, smaller faults (sub-seismic) may exhibit a continuous
smear for much higher smear factors, and display a greater varia-
tion in SSF (in the range of 150) compared to larger faults. Yielding
et al. (1997) dene an SGR limit of 1520% between sealing and
non-sealing faults, based on observations of pressure data across
faults from wells in various tectonic settings. Bense and Van Balen
(2004) obtained reasonable agreement of fault zone hydraulic
conductivity inferred the SGR approach, and independently from
calibrated numerical groundwater ow models.
Fault permeabilitySGR relationships have been proposed for
the North Sea based on well datasets (Manzocchi et al., 1999;
Sperrevik et al., 2002). Other approaches are based on extensive
experimental datasets fromtriaxial experiments, and usually relate
fault permeability to strain or porosity (e.g. Crawford et al., 2002;
Takahashi, 2003). However, the experiments are limited in terms of
maximum fault displacement that can be reproduced.
3. Laboratory experiments on clay smear formation
3.1. Sand and clay used for testing
The sand used in the experiments is Baskarp sand No 15,
a Holocene deposit fromVa tteren (Sweden) which consists of more
than 90% of angular to sub-angular well sorted quartz grains. The
sand is ne to medium grained (60% passing diameter
D
60
0.17 mm). The unit weight of grains g
s
is equal to 26.0 kN/m
3
.
The maximum and minimum dry unit weights g
dmax
and g
dmin
are
equal to 16.86 kN/m
3
and 14.28 kN/m
3
, respectively.
The clay used for the experiments is fromthe Troll eld offshore
Norway. The clay was sampled as part of a soil investigation
program for the Troll eld. Specimens from Unit I, 06 m depth
were selected for testing (Lunne et al., 2007). The geomaterial is
described as a very soft to rm, high plasticity and slightly over-
consolidated clay. Unit I is a glacio-marine sediment deposited
immediately after the de-glaciation of the area. Its deposition is
thought to have occurred in a low energy marine environment
while the sea level increased and water became deeper. The
average percentage of clay fraction (<2 mm) for Unit I is 44% and the
average percentage sand fraction (>60 mm) is 3%. The plasticity
index is on average circa 37%. The unit weight of solid particles g
s
is
equal to 26.8 kN/m
3
. The clay mineralogy is given in Table 1.
Before use in the ring shear apparatus, the Troll clay was
consolidated uniaxially up to 1.5 MPa effective vertical stress. After
unloading, the clay was then cut in segments shaped to t within
the conning rings of the ring shear apparatus. The water content
w
c
of the clay after compaction was measured to 26% giving an
Fig. 2. Use of ring shear device as an analogue to clay smear development during fault formation. Right: conceptual fault with clay smear. Middle: conguration of sand and clay
segments in ring shear experiments described in this paper. Permeability measurements across shear band (I) and clay smear (II) are illustrated. Left: ring shear cell lled with sand
(front part is omitted for clarity), after Torabi et al. (2007). Inner ring and outer ring diameters are equal to 25.4 and 30.48 cm, respectively. Moving parts are shown with red arrows.
initial porosity n
c
of 41%. The compaction behaviour of the Troll clay
as well as the time required for consolidation is investigated in an
oedometer cell in which the permeability can be continuously
interpreted from the excess pore pressure at the undrained bottom
of the specimen (Sandbkken et al., 1986). The permeability of the
clay at 3.5 MPa effective vertical stress (in oedometer conditions) is
roughly equal to 0.0025 mD.
3.2. Description of the high stress ring shear apparatus
A brief description of the ring shear cell is given here. The reader
is referred to Cuisiat et al. (2007) and Torabi et al. (2007) for more
information about the apparatus.
In the ring shear cell, the sample is contained within the annular
space created between the inner and outer conning rings, and the
lower and upper rings (Fig. 2 left). The width of the annular space
between the rings is 25.4 mm, and the maximum height of the
sample is 44 mm, which is limited by the overlap required to
hydraulically seal the gap between the upper ring and the conning
rings. The inner and outer diameters of the conning rings are
equal to 25.4 cm and 30.48 cm, respectively, giving a horizontal
cross-sectional area of the sample equal to 223 cm
2
. On the upper
and lower rings 48 knives (or grooves) are evenly distributed to
ensure transfer of the torque into the specimen. 48 orices are
evenly distributed between the knives on the upper and lower
rings for ow measurements. During shearing, the upper ring is
xed and only the lower ring rotates clockwise. Two reaction arms
on the upper frame provide the reaction forces against the rotation
forces of the lower assembly. Two load cells measure the reaction
forces, which in the absence of friction, sum to the shear forces
exerted on the specimen. The effective normal stress onto the
specimen is provided by a rigid loading frame. The normal
displacement is calculated from an average of the sample height
measured at 3 positions around the ring from three external linear
variable displacement transducers.
The porosity is calculated from the total volume and pore
volume based on the mass and dried densities of clay and sand. The
total volume is updated during an experiment from the measured
normal displacements. The porosity changes are calculated by
assuming no volumetric deformation in the clay.
3.3. Estimation of burial depth at time of failure
After initial normal loading and consolidation of the sample and
prior to shearing, a K
o
condition prevails in the sample. The major
and minor effective principal stresses correspond to the vertical
and horizontal effective stresses, respectively. During shearing, the
sample is loaded to failure by imposing horizontal displacements
and shear stresses, which results in a rotation and change in
magnitude of the principal stresses, while the normal effective
stress remains constant.
1
At failure assuming no volumetric defor-
mation, the rotation reaches 45
and vertical and horizontal

stresses are then equal. The horizontal direction is the direction of
maximum shear stresses in the sample (Nowacki, 1967), and the
maximum principal stress s
0
1
is inclined at 45
from the maximum

shear direction, as observed experimentally by Mandl et al. (1977).
In a sedimentary basin, under no lateral constraints, normal
faulting occurs as the Mohr-Coulomb slip (i.e. rst Riedel shears) at
an angle 45
f=2 froms
0
1
. The major principal effective stress s
0
1
is the vertical overburden stress s
0
v
in a basin under tectonic
extension, which can be converted to a depth of faulting assuming
a lithostatic gradient for the basin.
In the design of the experiments, we assume an effective
lithostatic gradient of 10 MPa/km and a friction angle of 30
to
calculate the normal stress to be applied so that failure occurs
approximately at a given burial depth. The actual depth of faulting
is back-calculated after the experiments from the measured
stresses at peak condition.
3.4. Test description
The experiments are described in Table 2. After mounting and
saturation inside the ring shear cell, the samples are rst consoli-
dated to nominal vertical stresses. A small cycling load (100 cycles,
5 kN amplitude) is used during normal loading to ensure good
contact between the sample and the knives in the ring shear device.
After consolidation, the samples are sheared under constant
effective normal stress with a prescribed rotation velocity of 1 deg/
5 min for a total length of 90 degrees (220 mmshear displacement).
The velocity has been chosen such that the estimated pore pressure
build-up at mid-height of the sample is negligible.
The permeability between two opposite openings across clay
smear and sandsand juxtaposition is measured during normal
loading and every 1530 degrees of shear rotation by imposing
a ow rate Q at the injection point and a constant pressure at the
outlet, and recording the pressure difference DP between inlet and
outlet after steady state conditions are reached. Note that shearing
is stopped during permeability measurements. Typically for
measuring permeability across the clay smear, two opposite
openings at the center of the fault throw are used, while the
Table 1
Mineralogy of the Troll clay fromX-ray diffraction analyses (fromLunne et al., 2007).
Mineral % Clay fraction (<2 mm) % Sand fraction (>60 mm)
Illite 54
Kaolinite 19
Smectite 13
Chlorite 6 10
Quartz 2 65
Feldspar 2 17
Calcite Minor
Pyrite 8
Table 2
Laboratory program for experiments dedicated to clay smear.
Test Effective normal
stress (MPa)
Over-consolidation Burial depth
a
(m)
RT15 3.5 Normally consolidated 500
RT17 7 Normally consolidated 1000
RT20 3.510.5 Overconsolidated max 1500
a
Estimated before testing.
Table 3
Strength data from ring shear experiments. The depth of burial at time of faulting is
estimated from peak stress and a pore pressure gradient of 10 MPa/km.
Test Effective normal
stress (MPa)
Peak shear
stress (MPa)
Depth of burial at
faulting (m)
RT15 3.5 2.20 570
RT16 0.7 0.79 149
RT17 7 3.90 1090
RT18 10.5 5.32 1582
RT19 3.5 2.19 569
RT20 10.53.5 2.46 596
1
This condition could differ in other experiments.
remaining openings are closed. The interpreted coefcient of
permeability is an average value of the sample containing unde-
formed sand, shear band, and clay smear. The permeability of the
sample can be back-calculated using Darcys law if one assumes
that the ow is one dimensional and that the involved ow cross
section is known, through the equation:
k
Qm
A
DP
H
(1)
where:
Q constant ow rate across sample (m
3
/s)
DP pressure difference from inlet to outlet (Pa)
k permeability (m
2
)
m uid viscosity (10
3
Pa s, water)
H sample height (m)
A assumed ow area (m
2
).
In practise the permeability is expressed in mD using
10
15
m
2
z 1mD. A constant ow rate Q equal to 2.3 10
7
m
3
/s is
used resulting in typical values of pressure difference equal to
58 10
3
Pa at steady state. Steady state owconditions are achieved
after two to three minutes.
The main uncertainty when using Darcys law relates to the size
of the owarea. In the following, we will assume that the owarea
is equal to 1/24th of the ring area for the interpretation of the
permeability experiments. Given that this assumption may only be
an approximation in reality, 3D nite difference modelling was
conducted of the ow conditions during a permeability test in the
ring shear apparatus. The modelling shows that the ow area may
be closer to the area of the clay smeared zone, depending on the
contrast between clay and sand permeability.
In tests RT15RT18, three segments of clay 120
apart fromeach
other are used. In tests RT19 and RT20 5 segments are used, 3 of
them being only separated 15
from each other to address the

formation of composite smear. One segment had only half
the width of the others to study the effect of clay source thickness.
The last segment serves as reference. In test RT20 the same
conguration as test RT19 is used except that the clay is over-
consolidated during shearing, i.e. shearing takes place at a normal
stress level lower than the maximum experienced by the clay
during normal loading. Samples for thin section analysis were
taken after each ring shear test.
4. Results from experiments
4.1. Effect of effective normal stress
The development of fault deformation products for different
effective normal stresses (i.e. simulating different burial depths) is
investigated through tests RT15RT18, which were performed
under normal effective stress varying between 0.7 MPa and
10.5 MPa (i.e. estimated burial depth varying between 100 m and
1500 m). A corresponding depth of burial at time of faulting is
inferred from peak stress conditions and a hydrostatic pore pres-
sure gradient. The inferred depth of burial is close to the estimated
one (Table 3).
The porosity changes during normal loading (i.e. burial of the
sediments) and shear loading (i.e. faulting) are shown in Fig. 3 and
Fig. 4, respectively. The normal loading curves show small porosity
variations from test to test, but the normal stiffness of the samples
is fairly similar for all tests especially after initial cyclic loading
(around 1000 kPa) indicating reasonable repeatability of the
preparation procedure of the samples.
Porosity loss increases with normal loading and shear
displacement due to a combination of grain rolling, packing and
crushing. The amount of porosity reduction during shearing clearly
increases with normal loading. Test RT18 (1500 m estimated burial
depth) shows important porosity reduction due to massive grain
crushing. Note that this mechanism is time dependent as shown by
continuous porosity reduction (Fig. 4) and stress relaxation (Fig. 5)
measured during phases where shearing was resumed (overnight
or weekend). Steady state conditions (i.e. shearing under no
volume changes) are achieved for most tests except at high effec-
tive normal stress for which shearing should have been pursued
beyond 90 degrees.
The effect of effective normal stress on the shear strength and
normalised shear strength (i.e. apparent friction) of the sample is
shown in Fig. 5. The peak shear stress correlates well with the
normal stress applied onto the sample. More ductile behaviour is
shown as normal stress increases. Note that test RT16 exhibits
35
35. 5
36
36. 5
37
37. 5
38
38. 5
39
39. 5
100 1000 10000 100000
Effective Normal Stress (kPa)
P
o
r
o
s
i
t
y

(
%
)

RT16 - 0.7 MPa
RT15 - 3.5 MPa
RT17 - 7 MPa
RT18 - 10.5 MPa
RT18 - 1500 m
RT17 - 1000 m
RT16 - 100 m
RT15 - 500 m
Fig. 3. Porosity changes during normal loading for clay smear tests. The corresponding
(estimated) burial depths are: 100 m (RT16), 500 m (RT15), 1000 m (RT17), 1500 m
(RT18).
29
30
31
32
33
34
35
36
37
38
0 5 0 100 150 200 250
Linear Displacement (mm)
P
o
r
o
s
i
t
y

(
%
)

RT16 - 0.7 MPa
RT15 - 3.5 MPa
RT17 - 7 MPa
RT18 - 10.5 MPa
RT17 - 1000 m
RT16 - 100 m
RT15 - 500 m
RT18 - 1500 m
Fig. 4. Porosity changes during shearing for clay smear tests. The corresponding
(estimated) burial depths at the time of faulting are: 100 m (RT16), 500 m (RT15),
1000 m (RT17), 1500 m (RT18). X-axis represents the linear displacement of the bottom
ring (i.e. fault throw).
abnormal high friction during shearing and two local peaks. This is
probably caused by local failure at the knives due to low internal
friction along the sidewalls within the ring cell. The low internal
friction is a result of the low normal stress applied which is not
suitable for the design of the high stress ring shear apparatus.
The residual friction angle, interpreted fromthe Coulomb failure
criterion s s
0
$ tan 4, gives a value around 25
for all tests, that is

more or less independent of the stress level and of the presence of
a clay smear. On the contrary, a similar clay smear experiment with
kaolin showed lower shear strength for faulting with clay smear
compared to a fault in pure sand only (Cuisiat et al., 2007), although
the residual friction of illite is lower than that of kaolinite (e.g.
Olson, 1974). The main reasons for the difference could be due to
the presence of sand and silt in the natural clay tested in our
experiments (circa 56%). The friction angle of a mixture of sand/silt
and clay is well known to increase with the sand/silt content in the
mixture (Lambe and Whitman, 1979).
The efciency of clay smear in terms of permeability across the
fault is illustrated in Fig. 6 where the ratio of clay smear
permeability to initial sand permeability (at the end of consolida-
tion) is shown versus linear displacement (220 mm 90 degrees
rotation) for different effective normal stresses. The initial sand
permeability varies from test to test depending on initial packing
and applied effective normal stress. For the reference case (RT15),
the initial sand permeability is equal to 3350 mD at 3.5 MPa
effective normal stress. For a given rotation in Fig. 6, the perme-
ability might be an average value of different measurements taken
at several locations along the smeared membrane. Note also that
the permeability across the clay smears is in fact the average
permeability across the whole specimen, in which several mecha-
nisms might contribute to permeability reduction (grain rear-
rangement and grain crushing in the sand, clay smear). The tests
show a dramatic permeability reduction with shearing, the reduc-
tion factor increasing with increasing effective normal stress. At
higher normal stresses (e.g. RT17 and RT18) corresponding to
greater depth of burial at the time of faulting, clay smear and sand
sand juxtaposition have similar permeability reductions for the
same amount of shear (Fig. 7). Visual post-mortem observations of
the sample (Fig. 8) show complex shear bands with occurrence of
grain crushing not only in the sandsand juxtaposition areas but
also along the clay smear zone. Grain crushing might therefore be
the dominant mechanism for permeability reduction at these
depths. At lower effective normal stress (e.g. RT16) corresponding
to shallow burial depth, the permeability across the clay smears is
0.001
0.01
0.1
1
0 5 0 100 150 200 250
P
e
r
m
e
a
b
i
l
i
t
y

R
a
t
i
o

k
s
m
e
a
r
/
k
s
a
n
d
,
i
n
i

RT16 - 0.7 MPa
RT15 - 3.5 MPa
RT17 - 7 MPa
RT18 - 10.5 MPa
RT18 - 1500m
RT17 - 1000m
RT15 - 500m
RT16 - 100m
Fig. 6. Ratio of clay smear permeability k
smear
over sand permeability before shearing
k
sand,ini
versus linear displacement for clay smear tests. The corresponding (estimated)
burial depths at the time of faulting are: 100 m (RT16), 500 m (RT15), 1000 m (RT17),
1500 m (RT18). Note that k
smear
is equal to k
sand,ini
at zero rotation by convention (no
smear).
1
10
100
1000
10000
0 2 4 6 8 10 12
Effective Normal Stress (MPa)
P
e
r
m
e
a
b
i
l
i
t
y

[
m
D
]

sand-sand juxtaposition
clay smear
Fig. 7. Permeability across sandsand juxtaposition k
sand
and clay smear k
smear
at the
end of shearing (90
) versus imposed effective normal stress during shearing for all

tests. The corresponding (estimated) burial depths at the time of faulting are: 100 m
(RT16), 500 m (RT15), 1000 m (RT17), 1500 m (RT18).
0
1
2
3
4
5
6
0 100 150 200 250
S
h
e
a
r

S
t
r
e
s
s

(
M
P
a
)

RT16 - 100m
RT15 - 500m
RT17 - 1000m
RT18 - 1500m
0
0. 2
0. 4
0. 6
0. 8
1
1. 2
1. 4
0 5
0 5 0 100 150 200 250
N
o
r
m
a
l
i
s
e
d

S
h
e
a
r

S
t
r
e
s
s

(

/

)

RT16 - 0.7 MPa
RT15 - 3.5 MPa
RT17 - 7 MPa
RT18 - 10.5 MPa
RT17 - 1000m
RT18 - 1500m
RT16 - 100m
RT15 - 500m
Fig. 5. Shear stress and normalised shear stress (i.e. ratio of shear stress over effective
normal stress) versus linear displacement for clay smear tests. The corresponding
(estimated) burial depths at the time of faulting are: 100 m (RT16), 500 m (RT15),
1000 m (RT17), 1500 m (RT18).
much lower than in the sandsand juxtaposition areas (Fig. 7).
Given the visible absence of sand shear (Fig. 8), clay smear seems to
be the main permeability-reducing mechanism for faults formed at
low burial depth. The contrast between permeability across clay
smears and sand sand juxtaposition clearly decreases with
increasing applied effective normal stress (Fig. 7).
The geometrical characteristics (i.e. thickness, continuity,
waviness) of the clay smears are qualitatively observed after testing
by removing progressively the sandclay material within the ring
cell. Clay smears show local buckling and variation in clay smear
thickness at low effective normal stress (0.7 MPa), see Fig. 8 (top).
At higher normal stresses (3.5, 7 and 10.5 MPa), clay smear occurs
within wider deformation bands where grain rolling and/or grain
crushing takes place. The thickness of the combined zone of clay
smear and sand shear seems to increase with effective normal
stress. Continuity of the clay smear is observed in all cases after
dismounting the sample, except for high normal stress (RT18)
where a local rupture of the clay smear is seen. The puncture
cannot be obviously correlated to changes in permeability
measurements. Thin sections taken from the samples were used to
qualitatively describe the amount of grain aking/splitting during
deformation (Fig. 8, bottom). The quality of the thin sections is
variable because of the difculty to saturate the clay-rich zone with
epoxy. Nevertheless, it can be seen that the degree of grain crushing
is increasing with increasing normal stress. At loweffective normal
stress, mixing of clay and sand grains is observed with diffuse
boundary between clay smear and sand. At higher normal stress,
sharper transitions are observed with pronounced grain commi-
nution within or along the clay smear. In the zone along the clay
smear, sand grains are oating in a matrix of crushed material.
A signicant amount of sand grains oating in the clay smear
can contribute to increase signicantly the permeability of the clay
membrane, compared to the permeability of a membrane made of
pure clay. Indeed the permeability of the clay smear, which is at the
most two orders of magnitude lower than the initial permeability of
the sand (Fig. 6), is much higher than the permeability of a pure
Troll clay sample tested in an oedometer cell (Section 3.1).
4.2. Inuence of clay source thickness and multiple clay layers
In tests RT19 and RT20, ve clay segments are embedded into
the sand to investigate the effect of overlapping clay smear and clay
Fig. 8. Vertical cross-sections of clay smear at the end of testing. The cross-sections are taken normal to shear direction in between upper and lower clay segments. Top:
Photographs during dismounting of sample. Bottom: Corresponding thin sections. Sample width in top gure is 2.54 cm, scale for thin sections is 500 mm. Note that the thin
fractures across and along the smear in thin sections of tests RT15, RT17 and RT18 are probably induced during the preparation of the thin sections. Apparent rupture on RT15 (top
gure) is due to loose grains during dismounting of sample. The corresponding (estimated) burial depths at the time of faulting are: 100 m (RT16), 500 m (RT15), 1000 m (RT17),
1500 m (RT18).
segment thickness within one test (Fig. 9, left). Three segments are
30
apart from each other to overlap during rotation of the ring.

One segment has half the thickness of the others, the last one is for
reference. In test RT20, the same conguration is used, but the
sample is rst normally loaded up to 10.5 MPa, and then unloaded
to 3.5 MPa to simulate burial and uplifting before faulting. As
a result, the clay and sand are in an over-consolidated state. Both
specimens RT19 and RT20 are sheared under a constant effective
normal stress of 3.5 MPa equivalent to a burial depth of 500 m.
Continuous clay smear is observed for both tests in all segments
except for the segment with half thickness in test RT19 where
a visible smear rupture circa 5 cm long is observed (Fig. 9 right). In
this case, only a very diffuse clay smear band, lling the pore space
between sand grains is observed (Fig. 10). The thickness of the clay
smear seems to increase and the transition between clay smear and
sand grains is more distinct when the thickness of the clay segment
increases, although zones with complete mixing of sand and clay
still remain. A 7 mm thick composite smear is observed for the
Fig. 9. Effect of clay thickness and composite smear. Shearing at 3.5 MPa effective normal stress (estimated burial depth at the time of faulting of 500 m). Left: Position of clay
segments in tests RT19 and RT20. Photograph width approximately 25.4 cm. Right: Cross-sections through clay smear at end of shearing for test RT19. Photograph width
approximately 2.54 cm.
Fig. 10. Thin sections from tests RT19 (normally consolidated) and RT20 (over-consolidated). Shearing at 3.5 MPa effective normal stress (estimated burial depth at the time of
faulting of 500 m). Note that the thin fractures observed in tests RT19 and RT20 are probably induced during the preparation of the thin sections.
overlapping clay segments, with clay issued from more than one
clay segment as indicated by embedded thin layers of sand within
the smear (Fig. 11). Sand shear is seen to a certain degree for all
segments (grain rolling or crushing), but seems to be more devel-
oped when the amount of clay available for smearing decreases.
Sand shear is also more developed in the over-consolidated sample
(RT20).
When clay segments overlap, a sand wedge can form between
the smear and the clay segment (Fig. 12). The resulting thick clay
smear is clearly visible with a marked layered structure of inter-
bedded clay/sand rich zones. The impact of overlapping segments is
clearly visible from the overall stress-displacement response
(Fig. 13). Tests RT15 and RT19 show very similar behaviour until
circa 110 mm linear displacement (i.e. 45
rotation), which corre-

sponds to onset of clay overlap in test RT19. Whereas the shear
strength of RT15 remains more or less constant with increasing
displacement, the shear strength of RT19 increases with linear
displacement, due to increased shear resistance with increasing
overlapping area. A similar behaviour is obtained for RT20, except
that the shear strength is generally higher because the sample is
over-consolidated. Note also a more brittle behaviour in test RT20
at the early stage of shearing, compared to RT19.
The results from permeability measurements show a higher
permeability of composite smear than for smear issued from
a single segment by the same ring rotation (see RT19 3 clay
segments compared to RT19 1 clay segment in Fig. 14). Further,
the permeability of composite smears does not decrease monoto-
nously with increased ring rotation. After a ring rotation of 60
(circa 145 mm linear displacement), the permeability increases

most probably due to the formation of interbedded sand wedges
between the smears with higher permeability than the smeared
clay. Note also that the permeability of the clay smear developed
from half a clay segment (RT19 clay segment in Fig. 14) is
slightly higher than the permeability of the smear for a full segment
(RT19 1 clay segment). Since a constant cross-ow area is
assumed in the interpretation of permeability Eq. (1), this could
indicate a larger amount of sand mix within the smear for a thinner
clay source. However, this result should be taken with caution as
the interpreted permeability depends highly on the value of the
cross-ow area used. If the true ow area is rather proportional to
the length of the smear membrane, for the same conditions of
rotation, ow rate and pressure gradient, the ratio between the
permeability of half a clay segment and that of a full segment is
then equal to (l w)/(l w/2), where l is the linear (shear)
displacement and w the width of the clay segment. This ratio
increases from 0.65 after 60 mm linear displacement to 0.93 after
225 mm.
The effect of over-consolidation is clearly shown in Test RT20:
the permeability across the smear for the over-consolidated spec-
imen (RT20) is lower than for the case of a normally consolidated
specimen (RT19) independently of segment thickness (Fig. 14).
5. Discussion
5.1. Conditions for clay smear formation
The kinematics and mechanics of clay smear remain poorly
understood. Outcrop eld studies (Lindsay et al., 1993; Lehner and
Pilaar, 1997; van der Zee and Urai, 2005; Bense et al., 2003) indicate
that clay smear develops from a combination of lateral clay injec-
tion and clay abrasion within the shear zone. Lehner and Pilaar
(1997) suggested that fault smear relates to pull-apart mechanics
creating normal fault relay zones across the shale beds, thereby
forming an extensional releasing bend. Egholm et al. (2008), from
discrete element modelling of fault zone evolution in a granular
material, show that contractional bends were more likely to be the
driving forces for clay smear. In our experiments, because of the
stress conguration inherent to the ring shear apparatus, it is likely
that abrasion smear is the dominant mechanism for clay smearing.
This is in line with the observed constant thickness of the smear
along the fault. In particular the thickness of the smear does not
seem to increase close to the clay segments.
Shear deformation in normally consolidated sediments results
in compaction and strain hardening due to interlocking of the
crushed grains, which may promote the spread of deformation into
adjacent sediments (Rawling and Goodwin, 2006). In the ring shear
apparatus this is hampered by the limited sample size. Neverthe-
less, the grain size distribution of deformation bands induced in the
Fig. 11. Thin sections from overlapping clay layer in RT19. Shearing at 3.5 MPa effective normal stress (estimated burial depth at the time of faulting of 500 m). (A) Layer enriched
with sand within the clay smear and (B) lens of crushed sand within the clay smear. Note that the thin fractures observed are probably induced during the preparation of the thin
sections.
ring shear apparatus is comparable to that of natural deformed
sands under similar burial conditions (Torabi et al., 2007).
5.2. Clay smear permeability from smear thickness
The clay smear permeability values interpreted from the ring
shear tests refer to average permeability values over the whole
height of the sample, i.e. of the underformed sand, the deformed
sand and the clay smear. In the following, we estimate the
permeability of the clay smear alone from a simple permeability
model. It is assumed that an average permeability can be dened
for the clay smear and the sand separately.
The average permeability k
avg
normal to a stratied medium
made of sand and clay smear is calculated from the harmonic
average:
k
avg
k
sand
1
1
k
sand
k
smear
t
smear
t
tot
t
smear
t
tot
(2)
where:
k
smear
is the permeability of the clay smear
k
sand
is the permeability of the sand
t
tot
is the total height of the sample
t
smear
is the thickness of the clay smear.
Fig. 12. Sketch showing formation of sand wedge during faulting in claysand layered sediments, and observed cross-sections at end of shearing in test RT19. Shearing at 3.5 MPa
effective normal stress (estimated burial depth at the time of faulting of 500 m). All sections are within 5 cm along the shear direction. The sand wedge thins in the direction of
rotation of the lower ring. Photograph width approximately equal to 2.54 cm.
The permeability ratio k
avg
/k
sand
is plotted in Fig. 15 for different
values of k
sand
/k
smear
, i.e. ratio between sand and clay smear
permeability. The results from the ring shear tests are plotted on
the gure, based on the measured permeability at end of shearing
(90
rotation) and clay smear thicknesses reported in Fig. 8. Rela-

tively small contrasts between sand and clay smear permeability
are back-calculated fromthe average permeability model, except at
shallow burial depth (test RT16) for which the clay smear perme-
ability is 100 times less than that of the sand. Thus, the model is
consistent with the idea that at shallow depth, clay smear is more
efcient than sandsand juxtaposition.
Note also that the inferred permeability of the smear is much
higher than that of pure clay measured in an oedometer cell
(Section 3.1). Permeability increase during shearing can result from
1) clay dilatancy causing changes in the microstructure (Bolton and
Maltman, 1998; Takizawa et al., 2005), 2) sand mixing in the clay
membrane.
5.3. Comparison with empirical models
The efciency of the developed clay smear to seal is assessed
from the empirical parameters presented earlier. The parameters
are calculated from the clay smear experiments in Table 4. In all
experiments, clay smear occurs along the shear plane with some
variation in clay smear thickness. Discontinuity in clay smear is
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 5 0 10 0 15 0 20 0 25 0
(

s

s

e

r

t

S

r

a

e

h

S

d

e

s

i

l

a

m

r

o

N

/

)

RT15 -3.5 MPa
RT19 -3.5 MPa
RT20 -3.5 MPa
RT15 -500 m
RT19 -500 m (CS)
RT20 -500m (CS/OC)
Fig. 13. Normalised shear stress versus linear displacement for tests RT15, RT19
(composite smear CS) and RT20 (composite smear and over-consolidated CS/OC). All
tests are sheared under 3.5 MPa effective normal stress.
0
500
1000
1500
2000
2500
0 50 100 150 200 250
P
e
r
m
e
a
b
i
l
i
t
y

a
c
r
o
s
s

s
m
e
a
r

(
m
D
)
RT19 - 1 clay segment
RT19 - 3 clay
segments
RT19 - 1/2 clay
segment
RT20 - 1 clay segment
RT20 - 1/2 clay segment
O
v
e
r
c
o
n
s
o
l
i
d
a
t
e
d
Fig. 14. Permeability across clay smear from single clay segment, half segment and
three segments versus linear displacement in tests RT19 and RT20 (over-consolidated).
Both tests are sheared under 3.5 MPa effective normal stress.
0.001
0.01
0.1
1
10
0 0.02 0.04 0.06 0.08 0.1
k
o
i
t
a
r
y
t
i
l
i
b
a
e
m
r
e
P
e
c
n
e
u
q
e
s
k
/
d
n
a
s
Normalised thickness of clay smear t
smear
/t
total
ksand/ksmear=10
ksand/ksmear=100
ksand/ksmear=1000
RT18
RT17
RT15
RT16
Fig. 15. Average permeability of sandclay smear layer sequence (harmonic average)
versus ratio of clay smear thickness to total thickness for different ratios of sand
permeability k
sand
/clay smear permeability k
smear
. The ring shear tests are plotted with
black lled symbols. The thickness of the clay smear is estimated after dismounting the
samples (Fig. 8).
Table 4
Shale Smear Factor SSF, Clay Smear Potential CSP and Shale Gouge Ratio SGR
calculated at mid-point along slipped interval from ring shear experiments on clay
smear. Tests RT19, RT15 and RT20 are sheared under 3.5 MPa effective normal stress,
but RT20 is over-consolidated. Tests RT16, RT17 and RT18 are sheared under 0.7, 7
and 10.5 MPa effective normal stress, respectively. The corresponding (estimated)
burial depths at the time of faulting are: 100 m (RT16), 500 m (RT15, RT19, RT20),
1000 m (RT17), 1500 m (RT18).
RT19, RT20 RT15RT20 RT19, RT20
Throw (mm) 219.5 219.5 219.5
Clay segment
thickness (mm)
15.5 31 31
Clay volume
factor (%)
100 100 100
Number of
clay segments
layer 1 layer 3 layers
Shale Smear
Factor (SSF)
14.2 7.1 2.4
Clay Smear
Potential (CSP)
1.1 4.4 39.4
Shale Gouge
Ratio (SGR)
7.1 14.1 42.4
Thickness of
smear (mm)
0.12 23 24 up to 7
Continuity of smear Discontinuous (RT19)
Continuous (RT20)
Continuous
except at high
burial depth (RT18)
Continuous
In bold gures: probable discontinuous smear predicted from empirical
relationships.
observed for test RT19 (half-segment width) in agreement with SSF
and SGR estimations. For one clay segment, the values of SSF and
SGR are close to the limit for breach of the smear (SSF 7 and
SGR 14). Continuous smear is observed in all experiments,
although a small puncture is observed at high burial depth
(1500 m, RT18).
The previous parameters do not provide quantitative relation-
ship between smear development, rock properties, stress condi-
tions and uid ow properties. Takahashi (2003) for instance
showed through laboratory experiments that a critical SSF for seal
breach depends on the effective normal stress applied to the fault
during smearing. Our experiments also indicate that the burial
depth (or normal stress) might control the continuity of the smear.
Nevertheless, our experiments are in agreement with eld
evidence showing a rst order correlation between SGR and seal
capacity (Naruk, 2008).
The results of the ring shear experiments are also compared to
the empirical relationship proposed by Sperrevik et al. (2002)
which relates fault rock permeability to maximum burial depth,
depth at the time of faulting and clay content in fault rock. In Fig. 16,
the empirical relationship is plotted for two values of maximum
burial depth (same as depth at the time of faulting in this case)
equal to 500 and 1500 m together with the permeability data from
ring shear tests with one clay segment. The ring shear test results
show a rst order agreement with the empirical relationship of
Sperrevik et al. (2002) based on data of nearly 100 normal faults
from clastic reservoirs in the North Sea (Fig. 16).
5.4. Fault permeability reduction with increasing burial depth
Despite uncertainties related to the interpretation of ow
measurements (in particular the owarea), the test results suggest
that clay smear and shear displacements are the main permeability
reduction mechanism for faults formed at low burial depths
(Fig. 17). Sandsand juxtaposition shear is dominated by grain
rolling causing only minor permeability reduction. At greater
depths grain crushing along the clay smear zone and in the sand
sand juxtaposition areas contributes as much to permeability
reduction as the formation of a clay smear (Fig. 17). The transition
between the two scenarios is likely to occur at around 500 m burial
depth. At this depth, some grain crushing is observed in the
experiments (e.g. test RT15), but not as developed as for experi-
ments conducted at 1000 mand 1500 mburial. In general the burial
depth at time of faulting has a larger inuence on the estimated
permeability change than the clay segment thickness and over-
consolidation ratio, within the range of parameter variation
investigated in the experiments.
6. Conclusions
In this paper, we have presented the results from experimental
work carried out with a high stress ring shear apparatus to inves-
tigate mechanisms of clay smear along faults in uncemented sedi-
ments at various burial depths, and its impact on uid ow
properties. The experiments consist of shearing a ring of sand with
embedded clay segments, thereby simulating faulting through
a layered sandclay sequence. Baskarp sand No 15 and a natural
glacio-marine clay from the Troll eld are used for testing. Visual
inspection of the samples after testing, analyses of thin sections and
permeability measurements across the shear zone are used to
describe clay smear continuity, thickness and permeability.
Deformation processes such as grain reorientation, clay smear
and cataclasis are identied from the tests. The complexity of the
shear zone is observed to increase with greater burial depth at time
of faulting. The experiments suggest that at shallow burial depth in
clay-rich sediments, clay smear is the most efcient mechanism for
permeability reduction. At this depth, sandsand juxtaposition
shear is dominated by grain rolling causing only minor perme-
ability reduction. At greater burial depths, permeability reduction is
dominated by grain crushing. Measurements of permeability both
across clay smear and sandsand juxtaposition yield similar values.
The observed thickness of clay smear is more sensitive to the
thickness of the sheared clay layers than other parameters tested
within the limited test program. Shearing of multiple clay layers (3
layers) produces a composite clay smear 23 times thicker than for
a single clay layer, whereas when reducing the clay layer thickness
to one half of the reference layer, a thin and discontinuous clay
smear is produced. The permeability across the clay smear is found
to increase as the thickness of the clay source decreases for single
clay layers, but the permeability for composite smear is more
complex and a clear trend is not found from the only two tests
performed. Over-consolidation of the sample prior to shearing has
no signicant inuence on the thickness and continuity of the clay
smear produced, but a reduction in permeability both for clay
smears and initial sand is found when compared to a normal
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E+02
1.E+04
1.E+06
1.E+08
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
)
d
m
(
y
t
i
l
i
b
a
e
m
r
e
P
Clay content in fault Vclay
1500 m
500 m
RT16 - 0.7 MPa
RT15 - 3.5 MPa
RT17 - 7 MPa
RT18 - 10.5 MPa
Fig. 16. Comparison of permeability data from ring shear tests (one clay segment) with
empirical relationship proposed by Sperrevik et al. (2002). The clay content in the fault
rock is taken equal to the SGR for the ring shear data. The corresponding (estimated)
burial depths at the time of faulting are: 100 m (RT16), 500 m (RT15), 1000 m (RT17),
1500 m (RT18).
Deep burial
depth
(~1500 m)
Shallow
burial depth
(~100m)
Bef ore
shearing
220 mm
Permeability
(mD)
~2500
~1200
~800
~1700
~10
~100
~500
110 mm
na
SGR 28 14
Shear displacement
Fig. 17. Synthesis of permeability trends from ring shear experiments. Values indicate
permeability in mD for clay smear (blue) and sandsand juxtaposition (yellow) or
a combination (yellow/blue). No ow measurement is available in the sandsand
juxtaposition for loweffective normal stress (i.e. low burial depth) at end of shear. Ring
shear displacements have been converted to shale gouge ratio (SGR). (For interpreta-
tion of the references to colour in this gure legend, the reader is referred to the web
version of this article.)
consolidated test. This implies that the initial density of the
material is important for the permeability measurements.
The results from the experiments show many similarities with
eld or outcrop natural observations. Next to clay smearing, drag or
injection of clay sand along the fault plane also occurs. Due to
mixing of clay and sand into the fault core, the ow properties of
the fault are expected to be anisotropic with higher permeability
along the shear direction (Bense and Van Balen, 2004). However,
more experimental data should be collected in order to develop
better trend lines and predictive models of fault properties. Field
and outcrop observations could be used actively to constrain the
experimental conditions to use in future tests. Future research will
complement the existing dataset as well as address formation of
phyllosilicate framework fault rocks in unclean sand with varying
clay content, clay mineralogy, burial depth, and fault throw. More
tests are also needed for improving the understanding of defor-
mation bands in multiple clay layers and the effect of composite
smear on fault permeability. Other types of experimental set-ups
such as a biaxial plane strain apparatus (Rykkelid and Skurtveit,
2008) may also be used to reproduce more accurately the stress and
strain conditions during basin extension and fault propagation
through layered sandclay sequences. Ultimately, through more
extensive databases, models can be implemented into reservoir
simulators to capture better the impact of faults on oil recovery
(Jolley et al., 2007).
Acknowledgements
The work presented in this paper was carried out as part of the
PROFUSE project (2007) with support from Total, BP and Sta-
toilHydro. Comments from Yves Leroy and an anonymous reviewer
are greatly acknowledged.
References
Annunziatellis, A., Beaubien, S.E., Bigi, S., Ciotoli, G., Coltella, M., Lombardi, S., 2008.
Gas migration along fault systems and through the vadose zone in the Latera
caldera (central Italy): implications for CO
2
geological storage. International
Journal of Greenhouse Gas Control 2, 353372.
Antonellini, M., Aydin, A., Pollard, D., 1994. Microstructure of deformation bands in
porous sandstones at Arches National Park, Utah. Journal of Structural Geology
16, 941959.
Bense, V., Van den Berg, E., Van Balen, R., 2003. Deformation mechanisms and
hydraulic properties of fault zones in unconsolidated sediments; the Roer
Valley rift system, The Netherlands. Hydrogeology Journal 11, 319332.
Bense, V.F., Van Balen, R., 2004. The effect of fault relay and clay smearing on
groundwater owpatterns in the lower rhine embayment. Basin Research
Volume 16, 397411.
Bishop, A.W., Green, G.E., Garga, V.K., Andersen, A., Browns, J.D., 1971. A new ring
shear apparatus and its application to the measurement of residual strength.
Geotechnique 21, 273328.
Bjo rlykke, K., Ho eg, K., 1997. Effects of burial diagenesis on stresses, compaction and
uid ow in sedimentary basins. Marine and Petroleum Geology 14, 267276.
Bolton, A., Maltman, A., 1998. Fluid-ow pathways in actively deforming sediments:
the role of pore uid pressures and volume change. Marine and Petroleum
Geology 15, 281297.
Bouvier, J.D., Kaars-Sijpesteijn, C.H., Kluesner, D.F., Onyejekwe, C.C., Van der Pal, R.C.,
1989. Three dimensional seismic interpretation and fault sealing investigations,
Nun River eld, Nigeria. AAPG Bulletin 73, 13971414.
Cardozo, N., Cuisiat, F., 2008. Fault propagation folding modeling with FLAC. In:
Proceedings of the 1st International FLAC/DEM Symposium on Numerical
Modeling, 2527 August 2008, Minneapolis, USA.
Chuhan, F.A., Kjeldstad, A., Bjorlykke, K., Hoeg, K., 2002. Porosity loss in sand by
grain crushing experimental evidence and relevance to reservoir quality.
Clausen, J.A., Gabrielsen, R.H., 2002. Parameters that control the development of
clay smear at low stress states: an experimental study using ring shear appa-
ratus. Journal of Structural Geology 24, 15691586.
Crawford, B.R., Myers, E.D., Woronov, A., Faulkner, D.R., Rutter, E.H., 2002. Porosity
Permeability Relationships in Clay-bearing Fault Gouge. Society of Petroleum
Engineers Inc. SPE/ISRM 78214, Rock Mechanics Conference, October 2002.
Cuisiat, F., Skurtveit, E., Cleave, R., 2007. Fault seal prediction in unconsolidated
sediments with a novel experimental apparatus. In: Proceedings of the 7th
ISRM Congress on Rock Mechanics, Lisboa, Portugal.
Doughty, P.T., 2003. Clay smear seals and fault sealing potential of an exhumed
growth fault, Rio grande rift, new Mexico. American Association of Petroleum
Geologists Bulletin 87, 427444.
Egholm, D.L., Clausen, O.R., Sandiford, M., Kristensen, M.B., Korstgard, J.A., 2008. The
mechanics of clay smearing along faults. Geology 36, 787790.
Faerseth, R.B., 2006. Shale smear along large faults: continuity of smear and the
fault seal capacity. Journal of the Geological Society 163, 741751.
Fisher, Q.J., Knipe, R.J., 1998. Fault sealing processes in siliciclastic sediments. In:
Jones, G., Fisher, Q., Knipe, R. (Eds.), Faults, Fault Sealing and Fluid Flow in
Hydrocarbon Reservoirs. Geological Society of London, special publication, vol.
147, pp. 117134.
reservoirs of the north sea and Norwegian continental shelf. Marine and
Gudehus, G., Karcher, C., 2007. Hypoplastic simulations of normal faults without
and with clay smears. Journal of Structural Geology 29, 530540.
Hickman, S., Sibson, R.H., Bruhn, R., 1995. Introduction to special section:
mechanical involvement of uids in faulting. Journal of Geophysical Research
100, 1283112840.
Hungr, O., Morgenstern, N.R., 1984. High velocity ring shear tests on sand. Geo-
technique 34, 415421.
Jolley, S.J., Dijk, H., Lamens, J.H., Fisher, Q.J., Manzocchi, T., Eikmans, H., Huang, Y.,
2007. Faulting and fault sealing in production simulation models: Brent Prov-
ince, northern North Sea. Petroleum Geoscience 13, 321340.
Karakouzian, M., Hudyma, N., 2002. A new apparatus for analog modeling of clay
smears. Journal of Structural Geology 24, 905912.
Koledoye, B.A., Aydin, A., May, E., 2003. A new process-based methodology for
analysis of shale smear along normal faults in the Niger Delta. American
Kvaale, T., 2002. Et stadium av reservoarsand i ringskjrapparat for simulere
oppfrselen i forkastningssoner. Master thesis, University of Oslo.
Lambe, T.W., Whitman, R.V., 1979. Soil Mechanics. John Wiley, New York.
Lehner, F.K., Pilaar, W.F., 1997. The emplacement of clay smears in synsedimentary
normal faults: inferences from eld observations near Frechen, Germany. In:
Mller-Pedersen, P., Koestler, A.G. (Eds.), Hydrocarbon Seals: Importance for
Exploration and Production. Norwegian Petroleum Society, Special Publication,
vol. 7, pp. 3950.
Lindsay, N.G., Murphy, F.C., Walsh, J.J., Waterson, J., 1993. Outcrop studies of shale
smears on fault surfaces. Special Publications International Association of
Sedimentologist 15, 113123.
Lunne, T., Long, M., Uzielli, M., 2007. Characterisation and engineering properties of
troll clay. In: Tan, Phoon, Hight, Leroueil (Eds.), Characterisation and Engi-
neering Properties of Natural Soils.
Lupini, J.F., Skinner, A.E., Vaughan, P.R., 1981. The drained residual strength of
cohesive soils. Geotechnique 31, 181213.
Mandl, G., 2000. Faulting in Brittle Rocks. Springer Verlag, Berlin, 434 p.
Mandl, G., de Jong, L.N.J., Maltha, A., 1977. Shear zones in granular material. An exper-
imental studyof their structureandmechanical genesis. RockMechanics9, 95144.
for ow simulation models. Petroleum Geoscience 5, 5363.
Naruk S.J., 2008. Empirical calibrations of proven sealing faults. Joint Meeting of the
Geological Society of America, Soil Science Society of America, American Society
of Agronomy, Crop Science Society of America, Gulf Coast Association of
Geological Societies with the Gulf Coast Section of SEPM.
Nowacki, E.H.F., 1967. Anvendelse av ringskjrapparatet for studier av omrrte
leirersegenskaper ved store deformasjoner. Det store eksamensarbeidet i
Geoteknikk og fundamenteringslre. NTH, Trondheim.
Olson, R.E., 1974. Shearing strength of kaolinite, illite and montmorillonite. Journal
of the Geotechnical Division, ASCE 100 (GT11), 12151299.
USA. Journal of Structural Geology 28, 16231639.
Rutqvist, J., Birkholzer, J., Cappa, F., Tsang, C.-F., 2007. Estimating maximum
sustainable injection pressure during geological sequestration of CO2
using
coupled uid ow and geomechanical fault-slip analysis. Energy Conversion
and Management 48, 17981807.
Rutter, E.H., Maddock, R.H., White, S.H., 1986. Comparative microstructures of
natural and experimentally produce clay-bearing fault gouges. Pure and
Rykkelid, E., Skurtveit, E., 2008. Experimental work on unconsolidated sand: the
effect of burial depth at time of deformation. In: Proceedings of the 42nd US
Rock Mechanics Symposium and 2nd US-Canada Rock Mechanics Symposium,
Paper No 08-236, San Francisco, June 29July 2.
Sandbkken, G., Berre, T., Lacasse, S., 1986. In: Yong, R.N., Townsend, F.C. (Eds.),
Oedometer Testing at the Norwegian Geotechnical Institute. Consolidation of
Soils: Testing and Evaluation, ASTM STP 892. American Society for Testing and
Materials, Philadelphia, pp. 329353.
Sassa, K., Gonghui, W., Fukuoka, H., 2003. Performing undrained shear test on
saturated sands in a new intelligent type ring shear apparatus. Geotechnical
Testing Journal 26, 19.
Sibson, R.H., 2000. Fluid involvement in normal faulting. Journal of Geodynamics
29, 469499.
Sperrevik, S., Frseth, R., Gabrielsen, R., 2000. Experiments on clay smear formation
along faults. Petroleum Geoscience 6, 113123.
Sperrevik, S., Gillespie, P.A., Fisher, Q.J., Halvorsen, T., Knipe, R.J., 2002. Empirical
Hydrocarbon Seal Quantication. Elsevier, Amsterdam, pp. 109125.
Takahashi, M., 2003. Permeability change during experimental fault smearing.
Takizawa, S., Kamai, T., Matsukura, Y., 2005. Fluid pathways in the shearing zones of
kaolin subjected to direct shear tests. Engineering Geology 78, 135142.
Tika, T.E., Vaughan, P.R., Lemos, L.J., 1996. Fast shearing of pre-existing shear zones
in soil. Geotechnique 46, 197233.
Torabi, A., Braathen, A., Cuisiat, F., Fossen, H., 2007. Shear zones in porous sand:
insights from ring-shear experiments and naturally deformed sandstones.
van der Zee, W., Urai, J.L., 2005. Processes of normal fault evolution in a siliciclastic
sequence: a case study from miri, sarawak, Malaysia. Journal of Structural
Geology 27, 22812300.
Weber, K.J., Mandl, G., Pilaar, W.F., Lehner, F., Precious, R.G., 1978. The role of faults in
hydrocarbon migration and trapping in Nigeriangrowth fault structures. In: 10th
Annual Offshore Technology Conference Proceedings, vol. 4, pp. 26432653.
Wibberley, C.A.J., Yielding, G., Di Toro, G., 2008. Recent Advances in the Under-
standing of Fault Zone Internal Structure: A Review. In: Geological Society,
London, Special Publications, vol. 299 533.
Wibberley, C.A.J., Shimamoto, T., 2003. Internal structure and permeability of major
strike-slip fault zones: the median tectonic line in mie prefecture, southwest
Japan. Journal of Structural Geology 25, 5978.
Yielding, G., Freeman, B., Needham, T., 1997. Quantitative fault seal prediction.

Journal of Structural Geology - Special Issue - Fault Zones

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Structural Geology - Special Issue - Fault Zones

Uploaded by

Copyright:

Available Formats

Photograph of the Month

JSG special issue Fault Zones Photo of the month

), was active in this same position during movement. The

southeast (Figs. 1e5). The broad range in orientation of major

) and an ESEeWNW trend thought to

) normal CDBs which were formed during a later

between orientation-dip of the structures and the scan-line, it is

, bounding NeS trend-

towards the southeast, increasing

for isolated sheared joints, and about 50

. From de Joussineau and Aydin (2009).

to the scanline (see

. The two layers are either stacked up,

at a fracture density of 0.2 shows about 34% reduction

, respectively. The greatest decrease of the shear moduli

provide 0.9 for the critical density which is the

, requires that all but the youngest set of fractures

for isolated single faults

for closely spaced

) are more appropriate for fractured slabs between closely

) and one of the Youngs

, the true spacing values would defer from the measured

) to the scanline orientation

W and is mainly of breccias

E and is hosted by gneiss. The loading is EW

from those in the next layer above, as

is for pure mode I

for pure mode II.

. The normal faults that develop during

to the extension direction. Fault-zone widths, however, are

(Fig. 2a). For a <45

; see Fig. 2b) for

) for which only normal

apart (Fig. 2b). During the rst phase of

), numerous faults form in the deformation

. The fault pattern at the

(e.g., Withjack and Jamison,

; the outside circle of each

between the rst-

between the rst- and second-

between the extension directions, most second-phase deforma-

(Model 4) in the second-phase s

(i.e., the angle between the rst- and second-phase

), then abyssal-hill normal

, however, new outer-slope normal faults

. Our modeling results corroborate these

. In addition, our models suggest that some of the reactivated

, although the actual angle

Cohesion c 0.5e1.0 MPa

) measured between the in-situ principal

to the shear direction (Fig. 8a and

to the shear direction.

to the sample boundary and shear direction.

to the shear direction, regardless of shear strain

to (001), kink generation at 0

to (001). They observed evidence of dislocation glide in samples

to (001) at temperatures of 20, 200 and 400

for core I (8a)

for core II (8b)

for core I (9a)

for core II (9b)

in core I, and from 76

in the two siltstones, less than in the Westerly granite

for axisymmetric extension s