You are on page 1of 10

Available online at www.sciencedirect.

com

Acta Materialia 61 (2013) 13941403


www.elsevier.com/locate/actamat

Mechanisms of dislocation multiplication at crack tips


Erik Bitzek a,b,, Peter Gumbsch b,c
a

Department of Material Science and Engineering, Institute I: General Materials Properties, Friedrich-Alexander-Universitat Erlangen-Nurnberg,
Martensstr. 5, 91058 Erlangen, Germany
b
Institut fur Angewandte Materialien IAM, Karlsruher Institut fur Technologie KIT, 76131 Karlsruhe, Germany
c
Fraunhofer-Institut fur Werkstomechanik IWM, 79108 Freiburg, Germany
Received 21 September 2012; received in revised form 9 November 2012; accepted 12 November 2012
Available online 14 December 2012

Abstract
Whether a stressed material fractures by brittle cleavage or ductile rupture is determined by its ability to convert elastic strain energy
to plastic deformation through the generation and motion of dislocations. Although it is known that pre-existing dislocations play a
crucial role in crack tip plasticity, the involved mechanisms are unclear. Here it is demonstrated by atomistic simulations that individual
pre-existing dislocations may lead to the generation of large numbers of dislocations at the crack tip. The newly generated dislocations
are usually of dierent types.
The processes involved are fundamentally dierent for stationary cracks and propagating cracks. Whereas local crack front reorientation plays an important role in propagating cracks, the multiplication mechanism at stationary cracks is connected with cross-slip in
the highly inhomogeneous stress eld of the crack. Analysis of the forces acting on the dislocations allows to determine which dislocations multiply and the slip systems they activate. These results provide the necessary physical link between pre-existing dislocations and
the generation of dislocations at crack tips.
2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Fracture; Dislocations; Crack-dislocation interaction; Brittle-to-ductile transition; Atomistic simulations

1. Introduction
The resistance of a material against crack propagation is
undoubtedly one of the most important properties of structural materials. It is quantied by the fracture toughness
KIc. The temperature dependence of the fracture toughness
and the strain-rate dependence of the brittle-to-ductile
transition (BDT) have both been shown to correlate well
with dislocation mobility [13]. Although this behavior is
indicative of dislocation motion playing a mayor role in
both toughness and the BDT, our mechanistic understand-

Corresponding author at: Department of Material Science and


Engineering, Institute I: General Materials Properties, Friedrich-Alexander-Universitat Erlangen-Nurnberg, Martensstr. 5, 91058 Erlangen, Germany. Tel.: +49 (0)9131 8527507; fax: +49 (0)9131 8527504.
E-mail address: erik.bitzek@ww.uni-erlangen.de (E. Bitzek).

ing is not suciently advanced to predict the occurrence of


either phenomenon.
Fracture toughness is determined by the competition of
the dynamics of the atomic bond-breaking processes [46]
and plastic deformation in the vicinity of the crack [7].
Fracture toughness is known to critically depend on materials microstructure [811], and particularly on how dislocations are generated or multiplied at the crack tip [2,12].
Dislocation nucleation from the crack tip [1321] and
ledges at the crack front [12,2224] have been studied intensively, but there are still many open questions, including
how the experimental setup inuences the selection of the
activated crack-tip slip systems [12,25,26]. The general outcome of these studies, however, is that the barrier for dislocation nucleation is usually too large for homogeneous
nucleation of dislocations on inclined and oblique planes
[8,16,2224]. The observation that no dislocation emission
takes place in molecular dynamics (MD) simulations of

1359-6454/$36.00 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.actamat.2012.11.016

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

fracture in typical crack systems of Sieven at 1500 K


[21]conrms these theoretical considerations. Heterogeneous dislocation nucleation at crack tip ledges is energetically the most favorable alternative, in particular if the
emitted dislocations are of screw character [8,23,24]. However, the ledges must be of signicant height (about 50 nm
in the case of iron [23]), and it is not clear how ledges could
account for the complicated slip geometries involving dislocations of dierent Burgers vectors on multiple slip planes,
all emanating from one point at the crack tip [27,28]. Furthermore, in detailed investigations of the dislocations
emitted from crack tips [12] it was shown that not all dislocations could be traced back to crack front ledges or other
defects at the crack tip.
Although it has long been known that predeformation
and the availability of dislocations in a material strongly
inuence fracture toughness, only recently has attention
been directed towards the mechanisms of interaction of
existing dislocations with the crack tip. In situ experiments
by X-ray topography [2932] or transmission electron
microscopy and tomography [3335] elucidate how dislocations interact with a crack-tip in otherwise dislocation-free
silicon. While many dislocations either move away from
the crack tip or do not seem to interact strongly, some individual dislocations impinging on the crack front stimulated
the generation of large numbers of new dislocations on
multiple glide systems [2932]. One of the most imminent
questions with regard to the interaction of preexisting dislocations and cracks therefore is which dislocations lead to
the avalanche-like multiplication and why. The goal of this
paper therefore is to elucidate the mechanisms responsible
for stimulated emission of dislocations from a crack front
caused by individual pre-existing dislocations.
The atomistic simulations reported here reveal several
mechanisms by which pre-existing dislocations interact
with an atomically sharp crack and multiply. It is shown
here that such multiplication necessarily requires cross-slip
at the crack tip, which is substantially aided by a signchange in the driving force of some dislocations in the eld
of the crack. For propagating cracks, the interaction with
dislocations can lead to a local reorientation of the crack
front, which enables the activation of additional slip
systems.
2. Simulation methods
Because of the wealth of experimental details available
on dislocation sources in silicon, this seems like the material
of choice for any such simulation. However, dislocation
motion in silicon is thermally activated and very sluggish
at reasonable temperatures due to the relatively high activation energy for dislocation motion [36]. Clearly for atomistic
simulations, which are limited to short time periods of the
order of nanoseconds, silicon cannot be used unless unrealistically high loads and temperatures are used. Therefore,
the EAM potential of Mishin [37] for face-centered cubic
(fcc) nickel was used as the model material, since it has

1395

the same slip systems as silicon but less sluggish dislocation


motion. It thus allows the study of dislocation processes in
the short time-scale available in MD simulations.
The simulation setup is shown in Fig. 1a. The simulations are performed in the so-called c-orientation [12] on
(1 1 0) cracks with a crack front aligned along the [0 0 1]
direction. This crack system has been extensively studied
in silicon [12,21,25,26,28,30,3335,3840], and pre-existing
dislocations were observed to cause stimulated dislocation
emission in this crack orientation [31,32]. Calculation of
the driving force caused by the stress eld of the crack on
all possible slip systems shows three distinct classes of dislocations in the c-orientation [12,41] (see Fig. 1b). The dislocations in each class are listed in the table in Fig. 1c.
Screw dislocations can change between classes I and II by
cross-slip, whereas dislocations of class III cannot change
to another class. Dislocations of class III have never been
observed to be emitted from c-oriented cracks in silicon
under mode I loading [12,3033,35].1 The present study
therefore focuses on dislocations of class I and II (for the
dislocations of class III, see Ref. [41]). Both were observed
as the result of stimulated emission in experiments [31,32].
The simulated system (Fig. 1a) corresponds to a cubic
crystal with side lengths of 75 nm (about 38 million atoms).
Fixed boundary conditions are used for the atoms in the
outermost boundary layers, except along the crack front
direction, where the motion of the atoms in the boundary
layers is only restricted in the z-direction. In this conguration a crack is introduced by application of the displacement eld of an atomically sharp crack in a rigidly
clamped thin strip [43] to all atoms. Using the FIRE algorithm [44], the crack is relaxed at the Grith strain
(G = 2.7%, determined from molecular static simulations
as the strain at which the strain energy which would be
released upon crack propagation just equals the surface
energy) where it remains stationary. The crack can in principle be further loaded or unloaded by appropriately scaling the displacement eld. However, we have performed
all simulations reported here with cracks loaded at or just
above the Grith load. A straight dislocation line is then
introduced in front of the crack tip by linear superposition
of the displacement eld of a relaxed dislocation to the displacement eld of the crack [41]. This setup mimics a scenario often found in brittle materials below the BDT: a
brittle (moving) crack encounters a lattice dislocation. In
preliminary simulations [41] we used a blunted crack to
start with, which is computationally simpler but did not
allow for a proper denition of the critical loading at the
crack tip, and might furthermore preclude certain interaction mechanisms between the dislocation and the crack tip.
The interaction between the crack and the dislocation is
studied in microcanonical MD simulations starting from
an initial temperature close to 0 K. During the simulations,
1

Dislocations of class III are sometimes observed when the crack is


created by indentation, which, however, leads to additional mode II and
III loading [42].

1396

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

atoms fixed in x,y,z

class I

atoms fixed in z

class II
(d)

(a)

[110]

(c)

[1-10]x

[0

01

class III

c
I

AC(d)=BD(c)

BC(d)=AD(c)

II

AC(b)=BD(a)

BC(a)=AD(b)

III

AB(d)=DC(a)

AB(c)=DC(b)

Fig. 1. (a) Simulation setup and crystallographic orientation of the crack in a model fcc material. As shown in the Thompson tetrahedron, the {1 1 1} glide
planes (a) and (b) contain a Burgers vector DC normal to the cleavage plane; planes (c) and (d), oblique to the cleavage plane and the crack front, contain
the crack propagation direction (parallel to direction BA). (b) The relative magnitude and sign of the glide component of the PeachKoehler force on a
dislocation segment of unit length placed in the vicinity of the crack front for dierent classes of dislocations. It is independent of the line direction and
calculated from the anisotropic solution of the stress eld of a sharp crack. Red means a positive force in glide direction, blue corresponds to a negative
force. The symmetry relations between slip systems are summarized in table (c). Dislocations in the same class experience the same forces in the stress eld
of the crack, as shown in (b) Within the same class, the behavior of dislocations on the right and on the left side of the table is mirror symmetric with
respect to the crack plane.

the increase in temperature due to conversion of the elastically stored energy of the starting conguration into the
motion of atoms was less than 10 K. For visualization only
atoms with a potential energy deviating by more than 2%
from the bulk energy are displayed. The Burgers vectors
of the dislocations were determined using a slip vector
analysis [45]. In addition to the simulations on atomically
sharp cracks shown here, we also studied the interaction
of dislocations with blunted cracks at dierent loads and
using various box sizes (see also [41,46]). The general mechanisms reported here are representative of over 20 simulations of dislocations interacting with the crack tip.
Together with our preliminary results reported in Ref.
[41], our simulations show that the characteristic mechanisms of the dislocationcrack tip interactions do not
depend strongly on the box size or the loading conditions
others than the magnitude of the load.
3. Results
The dislocations of class I were shown previously to just
cut the crack front without causing any dislocation emission, cross-slip or dislocation reactions [41]. Dislocations

of class I were often reported as the product but never


reported to initiate stimulated dislocation emission in
experiments [30,31]. We therefore focus here on the interaction of dislocations of class II with the crack.
If an initially straight screw dislocation with Burgers
vector DB is placed on the (a)-plane Dx = 6 nm in front
of the crack tip, a segment of the screw dislocation immediately cross-slips onto the (c)-plane (see Fig. 2a). This segment is classied as a class I dislocation. It is attracted
towards the crack tip and cuts it like other class I dislocations without any further dislocation reactions. The dislocation line in front of the crack also cross-slips onto the
(c)-plane. The remaining initial dislocation above the crack
surface does not cross-slip but bows down and intersects
the crack surface where it stays pinned. This segment then
acts as a pole around which the cross-slipped dislocation
revolves, forming a spiral source. The boundary conditions
do not allow the newly generated dislocations to escape,
which causes a back-stress on the source and eventually
stops its operation. The entire process is shown in Video
1 in the Supplementary Material.
The situation is dierent when a dislocation with
the same Burgers vector and identical glide plane but

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

t=4 ps

t=19 ps

vacancy row

t=28 ps

B
D

1397

Fig. 2. Evolution of an initially straight screw dislocation DB(a) in front of the crack tip: (a) cross-slip on the (c)-plane. A jog on the dislocation line
produced a vacancy row; (b) formation of a spiral source around the pole formed by the remaining original dislocation; (c) dislocation structure formed by
the emission of additional Dc-dislocations from the crack tip and dislocation reactions.

non-screw line orientation is placed close to the crack tip.


Fig. 3 and Supplementary Video 2 show an example in
which an initially straight 60 dislocation with Burgers vector DB(a) placed in front of the atomically sharp crack tip
(Dx = 6 nm) approaches a crack, loaded at 0.995 of the
Grith strain. As can be seen in Fig. 3a, parts of the dislocation close to the crack tip are strongly curved. Once the
dislocation intersects the crack front, the upper part of the
leading partial dislocation changes its glide plane to the (b)plane by creating a stair-rod partial dislocation between the
(a)- and the (b)-plane, according to the reaction
Da ! Db + ba. This essentially pins the upper part of the
dislocation at the crack front, while the lower part of the
dislocation proceeds to intersect the lower crack surface.
On the upper side, however, a new leading partial dislocation Dc is generated on the (c)-plane from the intersection
of the stair-rod dislocation ba and the crack front. The
emission of further leading partial dislocations on the (c)and (d)-planes lead to the complex dislocation structure
in Fig. 3c. The cross-slip process of an individual partial
dislocation is usually referred to as the Fleischer mechanism [47].
Other simulations under dierent loading conditions as
well as in smaller simulation boxes have all conrmed these
processes. In particular, for the kind of dislocations studied
here, stimulated emission of dislocations was always
observed for strains larger than 0.9G and was always con-

t=5 ps

nected with partial dislocation cross-slip and the formation


of a stair-rod dislocation at the crack tip.
It is, however, interesting to note that the partial dislocation cross-slip at the crack tip seems to be a necessary
but not sucient condition for the stimulated emission of
dislocations from the crack tip. We investigated this with
the 60 dislocation described above. When the dislocation
is pinned further away from the crack so that it can barely
reach the crack tip or is subjected to low loads, e.g. 0.82G
(which can only be reached with blunted crack tips), partial
dislocation cross-slip takes place without the occurrence of
stimulated dislocation emission. To elaborate on this
aspect further we also performed static simulations of dislocations impinging on cracks. However, in no case did the
static simulations lead to stimulated emission of dislocations. Even a dislocation placed directly at the crack front,
relaxed and subsequently equilibrated at 100 K, showed
cross-slip but no new dislocation was nucleated. The stress
eld and interaction of the dislocation and the crack therefore does not seem to be sucient to stimulate the dislocation emission process. The dynamics of the process which is
captured in the MD simulations where the dislocation
reaches the crack tip with velocities of the order of 1
2 nm ps1 seems to be important.
The interaction of a propagating atomically sharp crack
with the 60 DB(a) dislocation placed Dx = 8 nm in front
of the crack tip at an overload of 1.04G is shown in

t=11 ps

t=30 ps

crack front

[001]

Fig. 3. Molecular dynamics simulation of an initially straight 60 dislocation DB(a) in front of the crack tip. (a) Close-up of the curved dislocation during
its approach to the crack front. (b) Close-up of the stimulated emission of Dc and the partial dislocation cross-slip Da ! Db + ba at the crack front by
which a stair-rod dislocation ba is created when the leading partial dislocation changes its glide plane. (c) Complex dislocation structure formed by the
emission of additional Dc-dislocations as well as the trailing partial dislocation cB from the crack tip and subsequent dislocation reactions.

1398

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

Fig. 4 and Video 3 (Supplementary Material). The simulation is equivalent to the one shown in Fig. 3 just at a higher
load at which the crack propagates. The initial processes
(Fig. 4a) were similar to the stationary crack. However,
the evolution of the system was quite dierent since the
propagating crack was arrested at the point where it interacted with the dislocation. The rest of the crack front, however, kept propagating. Due to the emission of blunting
dislocations from the point of interaction, the crack front
was blunted along the Ba and bA directions and locally
aligned with these directions. This enabled the emission
of multiple dislocation half-loops on the (a)- and (b)planes, which were coupled by stair rod dislocations to partial dislocations on the (c)- and (d)-planes.

t=8 ps

a
D

D
[001]

D
B

t=12 ps

b
D

C
B

t=27 ps

c
D
D

C
D

Fig. 4. Snapshots of the interaction between a running crack and the 60


dislocation DB(a) at a strain of 1.04G (Dx = 8 nm). Only atoms with
increased potential energy are shown. The intersection of the crack with
the dislocation leads to crack front segments with orientations along Ba
and bA, from which partial dislocation half-loops such as Da are emitted.

4. Discussion
The simulation of the interaction of dierent types of
dislocations with a c-crack showed that only dislocations
of class II (see Fig. 1) interact with the crack tip in a
non-trivial way [41]. This is in agreement with X-ray
tomography results [32] where dislocations of class I were
often reported as the product but never reported to initiate
stimulated dislocation emission. Dislocations of class II
were reported as stimulating and as product dislocations.
All the characteristics of the dislocation sources
observed in this study can be related to the few experimental investigations of the detailed mechanisms of dislocation
nucleation and multiplication at crack tips.
First, the rare case of a single Burgers vector source
observed by George and Michot [12] was reported by them
to oftenbut not alwaysbe connected to crack tip
ledges. However, assuming that pre-existing ledges act as
the source appears unjustied since the ledges would
quickly be exhausted by the emission of the dislocations.
Furthermore, many visible crack tip ledges were not acting
as dislocation source [12]. Our simulations show a more
plausible mechanism, namely that cross-slip of a screw dislocation near the crack tip generates a spiral source. Such a
source can operate as long as the dislocations it has generated are able to move away from the crack tip. It can therefore generate large numbers of dislocations. Depending on
the precise orientation of the glide system such a source
canbut does not have togenerate ledges on the crack
faces behind the crack tip as the arm of the dislocation that
is directed backward moves away from the crack tip and
cuts the crack surface.
The more commonly observed dislocation source
according to Ref. [12] is a source which emits multiple Burgers vectors on dierent glide planes at once (see also Refs.
[2730]). Also X-ray tomography results clearly showed
multiple Burgers vectors being emitted from one source
or being stimulated from the interaction of one single dislocation with a crack tip [30,32]. This is also what we nd
here as a result of the interaction of the 60 dislocation with
the crack tip (see Fig. 3).
The crack arrest experiments by Gally and Argon
[8,26,25] were also conducted in the c-orientation and
revealed ^-shaped etch pit patterns at the arrest line which
were identied to emanate from potent dislocation sources
at the crack tip (see Fig. 11 in Ref. [26]). These sources produced suciently many dislocations to fully shield the
crack front and to arrest the propagating crack. The nucleated dislocations were rationalized to be screw dislocations
expanding backward from the crack tip toward the ank
regions of the crack [26]. Contrary to the sources at
stationary slightly blunted crack tips studied in Refs.
[12,32,38,39], the sources at propagating cracks in Ref.
[26] produced only dislocations on the (a)- and (b)-planes,
which have lower resolved shear stresses than the oblique
glide planes (c) and (d). Our simulation results on propagating cracks interacting with a pre-existing 60 DB(a)

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

dislocation (e.g. Fig. 4), as well as similar results of the


interaction of a propagating crack with tiny pores [46],
showed signicantly dierent results from the stationary
crack (Fig. 3). While the initial cross-slip event appeared
similar, the propagating crack then generated dislocations
with predominantly blunting character and clear V-shaped
traces along the crack front. If one imagines that the dislocations generated from several such V-shaped sources
meet, they would block each other and produce the same
conguration of etch pits as shown in Fig. 11 of Ref.
[26]. However, the location of the sources would then not
be as indicated there but at the opposite corners of the
crack front line marking the intersections of the (a)- and
(b)-planes with the cleavage plane. In our simulations the
emission of the blunting dislocations on the (a)- and (b)planes takes place in the forward direction of the propagating crack front, rather than backwards as suggested in Ref.
[25]. In hindsight, again our results appear more plausible.
The above dislocation source mechanism leading to the
V-shaped crack fronts is specic to propagating cleavage
cracks and cannot occur at a stationary crack. The observation of distinctly dierent nucleation processes at a stationary crack front as in Fig. 3 compared to a
propagating crack front as in Fig. 4 is clearly linked to
the fact that blunting dislocations can only be generated
when the crack front lies within an inclined glide plane.
Only then does the emission of dislocations with edge character lead to the blunting of the crack front. The blunting,
i.e. the increased curvature of the crack tip, is caused by the
formation of a ledge along the crack front direction by the
emitted dislocation. Since in the c-orientation the stationary crack front is not part of an inclined glide plane, the
nucleation of blunting dislocations is suppressed, and only
shielding dislocations are emitted, mostly on the highest
stressed set of glide planes. Propagating cracks, however,
can locally adopt dierent crack front orientations, e.g.
when a part of the crack front is held back by an obstacle
while the remaining crack front keeps propagating. When
the local crack front orientation is part of an inclined slip
system, blunting dislocations can be generated, which hinder the further propagation of the reoriented crack front,
while expanding in the forward direction of the propagating crack front which retained the initial orientation.
The possibility of activating new slip systems by local
reorientation of the crack front direction of propagating
cracks should be a general feature which applies to dierent
materials and crystal structures, as long as multiple blunting slip systems exist for the given crack orientation, which
intersect the crack front direction at an angle attainable by
local variations of the crack front orientation. These observations also solve one of the puzzling details in the literature, namely that dierent fracture experiments on the
same cleavage system, such as the loading of a stationary
crack [12,38,39] and the arrest of a running crack [25,26],
have shown very dierent dislocation emission processes.
One general consequence of this inherent dierence
between propagating and static crack is that the crack

1399

arrest toughness KIa could in principle show dierent values and a dierent orientation dependence as the crack initiation toughness KIc. The nucleation of blunting
dislocations from a locally reoriented crack front requires
a suciently long segment of the crack front to lie within
the inclined slip plane for a suciently long time. This
might not be the case for very rapidly propagating cracks,
or for cracks under cyclic loading, which advance homogeneously by small increments.
Investigating the mechanisms of dislocation multiplication in our simulations in more detail, it is found that dislocations of class II generally show strong local curvature
in front of the crack tip and react with the crack tip in a
non-trivial way. The reaction can be decomposed into three
dierent elementary processes which may occur individually or in combination with one another:
(a) the cross-slip of a dislocation segment that attains
screw orientation while approaching the crack tip,
(b) the partial cross-slip of dislocation segments directly
at the crack tip, usually followed by
(c) the stimulated emission of other dislocations.
The combination of these elementary processes leads to
dierent types of avalanche multiplication mechanisms of
dislocations at the crack tip. In all cases of multiplication
one of the two cross-slip processes (a) or (b) was involved.
Cross-slip of dislocations at or near the crack tip therefore
appears essential for the generation and multiplication of
dislocations at cracks.
The observed behavior of the dislocations in the stress
eld of the crackand particularly the crucial cross-slip
processescan be rationalized from the glide-component
of the PeachKoehler (PK) force FPK on a dislocation segment of unit length [48]. This driving force according to the
anisotropic elastic solution for the stress eld of a semi-innite c-oriented crack is shown in Fig. 1b. Dislocations of
class I generally experience higher driving forces than dislocations of class II. The force on class I dislocations is
mostly of the same sign and only shows one sign change
behind the crack tip. This can explain the observation that
class I dislocations usually just pass the crack tip without
signicant reactions [41]. In contrast, dislocations of class
II can encounter ve (class III four) sectors with alternating
signs of the driving forces on the dislocations. In particular
sign changes also occur directly at the crack tip in the singular part of the eld. Such a sign change in the singular
part of the crack tip eld will violently stop any incoming
dislocation at this location, where it may still see singular
stress components other than resolved shear stresses on
the original glide plane.
The cross-slip propensity of screw dislocation segments
can be rationalized as the dierence in the magnitude of
FPK on the screw dislocations in dierent slip planes
(Fig. 5a). For example, screw dislocations of class II on
the (a)-plane see large driving forces in front of the crack
to cross-slip to the (c)-plane. However, the sign of the

1400

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

Fig. 5. (a) The dierence in magnitude of the PeachKoehler force between dislocations of class I and class II. If a dislocation can cross-slip, it will
experience a larger driving force on the slip system belonging to class I in the blue areas, whereas the PeachKoehler force is larger on the same dislocation
on slip systems of class II in the red areas. (b) The similar dierence in magnitude of the PeachKoehler force on a partial dislocation for the case where
only the leading partial dislocation (here Da(a), respectively Db(b)) of a dislocation of class II can change the glide plane to the other glide plane of class II.

forces changes on the upper surface just behind the crack


tip. This is the reason why only a part of the screw segment
cross-slipped in Fig. 2. This change of sign is thus directly
related to the occurrence of the spiral source in Fig. 2. It
can also explain the observation of the cross-slip of a dislocation emitted on the (c)-plane onto the (a)-plane behind
the crack tip reported in Ref. [35] (e.g. Fig. 4 in Ref.
[35]). The change in sign of the PK force on the primary
glide plane itself further acts as a barrier to continued
motion of the dislocation on this plane, which further facilitates cross-slip [49].
Our simulations show that cross-slip always takes place
by the Fleischer mechanism [47,50,51] and often only partial cross-slip occurs, in which only the leading partial dislocation changes the glide plane, e.g. from the (a)-plane to
the (b)-plane in Figs. 3 and 4. It is worth recalling that dislocations of class II cannot cross-slip between the (a)- and
the (b)-planes. Dislocations of class II can only cross-slip
from the (a)- or (b)-plane to the (c)- or (d)-plane, thereby
changing their classication to class I dislocations (see
Fig. 1). The change of glide planes requires the creation
of a stair-rod dislocation, which is possible when the leading partial dislocation is oriented along the line of intersection between the two glide planes. If the leading partial is
part of a screw dislocation whose Burgers vector is parallel
to the line of intersection of the glide planes, e.g. the (a)and the (c)-plane, the cross-slip process can be completed
as in Fig. 2. However, in all other cases the Burgers vector
of the full dislocation is not parallel to the direction of the
stair-rod dislocation. Then the trailing partial dislocation
cannot react with the stair-rod dislocation and the crossslip process by the Fleischer mechanism cannot be completed. The consideration of the driving force for the
change of the glide plane is thus only relevant for the lead-

ing partial dislocation. Fig. 5b shows that at the crack tip


the leading partial dislocation Da(a) of Fig. 3 has a large
driving force above the crack plane to change its glide
plane to the (b)-plane, whereas below the crack plane the
(a)-plane is favored. Because the cross-slip process cannot
be completed, the stair-rod dislocation and the trailing partial dislocation remain pinned at the crack front.
Stimulated emission of dislocations from the crack front
directly upon impinging of a lattice dislocation of class II
was always observed in conjunction with this sort of partial
dislocation cross-slip. In all cases stimulated dislocation
emission took place on the (c)- or (d)-planes, which again
is in excellent agreement with the observations of Scandian
and Michot that only dislocations of class II stimulated the
emission of dislocations which were mostly of class I
[31,32].
It was speculated in Refs. [29,30,39] that the stress and
strain eld in the core of a dislocation at the crack front
alone would be sucient to cause the emission of dislocations. It was implicitly assumed in Refs. [29,30,32,39] that
every dislocation which is attracted to the crack tip upon
intersection leads to the stimulated emission of dislocations. Our observations, however, show that only dislocations of class II lead to stimulated dislocation emission,
and that stimulated emission is linked to partial cross-slip
of the incoming dislocation [41]. It is important to note that
in order for partial dislocation cross-slip to take place, the
partial dislocation approaching the crack front has to experience a sudden change of driving force from attractive to
repulsive when intersecting the crack tip, i.e., not only
are the Burgers vector and the glide plane factors determining whether stimulated emission takes place, but the relative position of the approaching dislocation relative to
the crack tip is also important. As our previous work on

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

blunted cracks has shown, the same dislocation placed


above or below the crack can in one case lead to stimulated
dislocation emission but not in the other [41].
Our static simulations furthermore show that structural
defects and the stress eld caused by the presence of a dislocation at the crack front alone do not seem to be sucient to stimulate the dislocation emission observed in the
dynamic simulations. We conclude that the dynamics of
the dislocation, the inertia of the dislocation [52,53] and
the waves generated upon deceleration of the dislocation
by the abrupt reversal of the driving force from strongly
attractive to strongly repulsive at the crack tip seem to play
an important role in the stimulated dislocation emission, at
least in the case of our simulations.
Due to the short time-scales of MD simulations, thermally activated processes are usually prohibited. On experimental time-scales, however, the pinning of the impinging
dislocation at the crack front by the partial cross-slip
process increases the probability of thermally activated
dislocation nucleation in the stress eld of the dislocation
core. Dislocations which just cut the crack front, on
the contrary, might not remain at the crack tip for a suciently long time for thermally activated processes to take
place.
Although we have simulated the fracture behavior of
nickel we have compared most of our results to experimental observations in silicon. This seems to be far-fetched
since the fracture and deformation behavior of silicon
and nickel are very dierent. In particular nickel in experiments is found to be ductile even at low temperatures.
This, however, may be a consequence of the very processes
we study here, namely that pre-existing dislocations which
cannot be avoided in metals interact with the crack front
and generate sucient plasticity to prohibit crack propagation. Our atomistic model for nickel, however, can support
a brittle propagating crack if the crystal is defect free
[46,54,55]. However, silicon at temperatures close to the
BDT temperature has the same glide systems and similar
splitting of dislocations on the glide-planes into partial dislocations as fcc metals [48,56,57]. It can therefore be
assumed that the gist of the observed processes and their
discussion, which is based only on the available slip systems and the stress eld of the crack, can be transferred
to the situation in silicon. Namely, the development of a
plastic zone by stimulated dislocation emission and multiplication, involving cross-slip as the most important mechanism, should be a general phenomenon. The analysis of
the driving forces on dislocations in the stress eld of
cracks furthermore allows assessment of the propensity of
a given dislocation to stimulate the emission of new dislocations from the crack tip, which is again independent of
detailed mobility considerations. It should be noted that
recent experiments on silicon at lower temperatures and
high stresses showed undissociated dislocations on the
shue planes (see Ref. [58] for a review on the experimental literature and Ref. [59] on corresponding calculations).
As these experiments were performed using large hydro-

1401

static conning pressures, it is unclear whether dislocations


in the tensile stress eld of a crack at higher temperatures
might also be located in the shue-planes. If this would
be the case, cross-slip should be even more pronounced
for the undissociated dislocations, but also climb forces
would need to be considered in the analysis of the driving
forces on the dislocations [58].
Because of the relatively small simulation cell, the emission of a few dislocations already leads to a signicant
reduction of the load on the crack tip and the evolution
of a macroscopic plastic zone is prohibited. The generation
of dislocation avalanches from stimulated emission as
described in Refs. [2932,39] requires that the dislocations
emitted on the oblique planes (class I dislocations) crossslip onto the (a)- or (b)-planes where they are attracted to
the crack front and, as class II dislocations, can lead to further stimulated emission events, from which the process
can repeat itself. The cross-slip of the emitted dislocation
to the glide plane with lower resolved shear stress is
assumed to be triggered by the back-stress of the already
emitted dislocations on the oblique plane [32]. The computational capacities are currently not suciently developed
to study such processes on the atomic scale.
Model experiments and modeling of the BDT have largely focused on dislocation-free silicon or other single-crystalline materials. Taking these as a basis, macroscopic
simulations of the BDT have hitherto only considered dislocation motion within a homogeneous plastic zone. In
contrast, the fracture properties of technical materials are
of course dramatically inuenced by changes in the materials microstructure. Initial steps to add information about
the microstructure into the modeling of the BDT behavior
[60,61] have been scarce and not very successful. In addition, current multiscale approaches which use, for example,
dislocation dynamics models to describe the fracture processes on the microscopic scale [62] do not include mechanisms allowing for stimulated dislocation multiplication at
the crack tip. This means that although the shielding and
blunting eects of dislocations on the crack tip have been
included in microscopic fracture models, the feedback of
existing dislocations on the emission of new dislocations
has not yet become part of such modeling approaches.
Including stimulated dislocation emission is of course of
particular importance in the modeling of technical, dislocation-containing materials. The investigations reported here
provide examples showing how the missing information
about dislocation generation and multiplication at a crack
tip can be obtained. We have clearly demonstrated that dislocations from only a few glide systems lead to dislocation
multiplication. We provide furthermore a rst criterion to
discern which dislocations may lead to multiplication and
which may not. Certainly the evolution of the dislocation
microstructure beyond that point will be of crucial importance to develop a more complete physical picture of the
generation of a plastic zone. However, our investigations
open new ways of systematically going about the modeling
the generation of a plastic zone near a crack tip in techni-

1402

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403

cal, predeformed and therefore dislocation-containing


materials.
5. Summary
In summary, we have performed MD simulations of the
interaction of pre-existing dislocations with atomically
sharp cracks in the c-orientation. The simulations revealed
fundamentally dierent interaction mechanisms for stationary cracks compared to propagating cracks. The intersection of a propagating crack by a dislocation can lead to
the local reorientation of the crack front which enables the
emission of blunting dislocations on initially not available
slip planes. This observation can explain the dierences
observed in experiments on stationary vs. propagating
cracks.
Our simulations furthermore show that only dislocations of class II lead to the stimulated emission and multiplication of dislocations at the tip of a stationary crack.
Such stimulated dislocation emission and avalanche-like
dislocation multiplication have also been observed in
in situ experiments, and the dislocation products observed
in the simulations agree in detail with the experimental
results. The simulations show that cross-slip of dislocations
at or near the crack tip is essential for stimulated emission
and dislocation multiplication to take place. The analysis
of the cross-slip propensity of dislocations in the stress eld
of a crack shows that a sign change in the driving forces on
the dislocation at the crack tip is essential for the multiplication and allows to explain which incoming dislocations
lead to dislocation multiplication and which do not. This
opens new ways to include the interaction between dislocations and crack tips in physical models for the generation
of a plastic zone near a crack tip and subsequently the
BDT.
Acknowledgments
Financial support from the German Science Foundation
(DFG, Gu 367/30) is gratefully acknowledged. P.G. gratefully acknowledges KITP and Materials Department, UC
Santa Barbara for partial support of this work through
NSF Grants PH11-25915 and DMR-0843934. E.B. gratefully acknowledges the German Science Foundation
(DFG, HO 2187/6-1) for partial support of this work.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/
j.actamat.2012.11.016.
References
[1] Roberts S, Booth A, Hirsch P. Mater Sci Eng A 1994;176:918.
[2] Gumbsch P, Riedle J, Hartmaier A, Fischmeister HF. Science
1998;282:12935.

[3] Hartmaier A, Gumbsch P. Phys Rev B 2005;71:024108.


[4] Gumbsch P, Cannon RM. MRS Bull 2000;25:1520.
[5] Kermode JR, Albaret T, Sherman D, Bernstein N, Gumbsch P, Payne
MC, et al. Nature 2008;455:12247.
[6] Atrash F, Hashibon A, Gumbsch P, Sherman D. Phys Rev Lett
2011;106.
[7] Abraham FF, Walkup R, Gao H, Durchaineau M, De La Rubia TD,
Seager M. Proc Nat Acad Sci USA 2002;99:57837.
[8] Argon A. J Eng Mater Technol 2001;123:111.
[9] Warren P. Scripta Metall 1989;23:63742.
[10] Gumbsch P. J Nucl Mater 2003;323:30412.
[11] Giannattasio A, Roberts SG. Philos Mag 2007;87:258998.
[12] George A, Michot G. Mater Sci Eng A 1993;164:11834.
[13] Rice JR, Thomson R. Philos Mag 1974;29:7397.
[14] Rice JR. J Mech Phys Solids 1992;40:23971.
[15] Zhu T, Li J, Yip S. Phys Rev Lett 2004;93:025503.
[16] Michot G, Gonzalez BM. Int J Fracture 2006;139:35967 [7th
International conference on fundamentals of fracture, Nancy, France,
MAY 2326, 2005].
[17] Tanaka M, Tarleton E, Roberts SG. Acta Metall 2008;56:51239.
[18] Gordon P, Neeraj T, Luton MJ. Modell Simul Mater Sci Eng
2008;16:045006.
[19] Gordon PA, Neeraj T, Luton MJ. Modell Simul Mater Sci Eng
2009;17.
[20] Liu G, Xu G. J Mech Phys Solids 2009;57:107892.
[21] Thaulow C, Sen D, Buehler MJ. Mater Sci Eng A 2011;528:435764.
[22] Zhou SJ, Thomson R. J Mater Res 1991;6:63953.
[23] Xu G, Argon AS, Ortiz M. Philos Mag A 1997;75:34167.
[24] Xu G. Dislocation nucleation from crack tips and brittle to ductile
transition in cleavage fracture. Dislocations in solids, vol. 12. Elsevier;
2004. p. 83145.
[25] Argon AS, Gally BJ. Scripta Mater 2001;45:128794.
[26] Gally BJ, Argon AS. Philos Mag A 2001;81:699740.
[27] Michot G, de Oliveira MAL, George A. Mater Sci Eng
1994;A176:99109.
[28] Michot G, Loyola de Oliveira MA. Mater Trans 2001;42:149.
[29] Michot G, Loyola de Oliveira MA, Koizumi H. J Phys IV France
1989;8:Pr414553.
[30] Scandian C, Azzouzi H, Malou N, Michot G, George A. Phys Status
Solidi A 1999;171:6782.
[31] Scandian C. Conditions demission et de multiplication des dislocations a` lextremite dune ssure. Application au cas du silicium, Ph.D.
thesis, Institut National Polytechnique de Lorraine; 2000.
[32] Michot G. Acta Metall 2011;59:386471.
[33] Higashida K, Narita N, Tanaka M, Morikawa T, Miura Y, Onodera
R. Philos Mag A 2002;82:326373.
[34] Tanaka M, Sadamatsu S, Nakamura H, Higashida K. Mater Trans
2011;52:3527.
[35] Tanaka M, Sadamatsu S, Liu GS, Nakamura H, Higashida K,
Robertson IM. J Mater Res 2011;26:50813.
[36] George A, Champier G. Phys Status Solidi A 1979;53:52940.
[37] Mishin Y. Acta Metall 2004;52:145167.
[38] Loyola de Oliveira MA, George A, Michot G. J Phys D 1995;28:A3841.
[39] Michot G, Azzouzi H, Malou N, Loyola de Oliveira M, Scandian C,
George A. In: Lepinoux, J. et al., editors. Multiscale phenomena in
plasticity. Kluwer Academic Press; 2000. p. 11725.
[40] Sadamatsu S, Tanaka M, Honda M, Higashida K. J Phys Conf Ser
2010;240:012142.
[41] Bitzek E, Gumbsch P. J Solid Mech Mater Eng 2008;2:134859.
[42] Tanaka M. Private communication; 2012.
[43] Gumbsch P, Zhou S, Holian L. Phys Rev B 1997;55:344555.
[44] Bitzek E, Koskinen P, Gahler F, Moseler M, Gumbsch P. Phys Rev
Lett 2006;97:170201.
[45] Zimmerman JA, Kelchner CL, Klein PA, Hamilton JC, Foiles SM.
Phys Rev Lett 2001;87:165507.
[46] Bitzek E. Atomistic simulation of dislocation motion and interaction
with crack tips and voids. Schriftenreihe Werkstowissenschaft und
Werkstotechnik, vol. 39. Aachen: Shaker Verlag; 2007.

E. Bitzek, P. Gumbsch / Acta Materialia 61 (2013) 13941403


[47] Fleischer R. Acta Metall 1959;7:1345.
[48] Hirth JP, Lothe J. Theory of dislocations. 2nd ed. New York: John
Wiley & Sons; 1982.
[49] Kubin L, Hoc T, Devincre B. Acta Metall 2009;57:256775.
[50] Duesbery MS. Modell Simul Mater Sci Eng 1998;6:3549.
[51] Bitzek E, Brandl C, Derlet PM, Van Swygenhoven H. Phys Rev Lett
2008;100.
[52] Bitzek E, Gumbsch P. Mater Sci Eng A 2005;400401:404.
[53] Bitzek E, Gumbsch P. Mater Sci Eng A 2004;387389:115.
[54] Gumbsch P. J Mater Res 1995;10:2897907.

1403

[55] Gumbsch P, Beltz G. Modell Simul Mater Sci Eng 1995;3:597613.


[56] Gomez A, Cockayne D, Hirsch P, Vitek V. Philos Mag 1975;31:
10513.
[57] Rabier J, Demenet J. Scripta Mater 2001;45:125965.
[58] Rabier J. Phys Status Solidi A 2007;204:224855.
[59] Pizzagalli L, Beauchamp P. Philos Mag Lett 2008;88:4217.
[60] Jokl ML, Vitek V, McMahon CJ. Acta Metall 1980;28:147988.
[61] Mataga PA, Freund LB, Hutchinson JW. J Phys Chem Solids
1987;48:9851005.
[62] Wei YG, Xu GS. Int J Plasticity 2005;21:212349.

You might also like