Isotropic Scale-Invariant Dissipation of Solar Wind Turbulence

K. H. Kiyani,1, ∗ S. C. Chapman,1 Yu. V. Khotyaintsev,2 A. Turner,1 B. Hnat,1 and F. Sahraoui3

1
Centre for Fusion, Space and Astrophysics; University of Warwick, Coventry, CV4 7AL, United Kingdom
2
Swedish Institute of Space Physics, Uppsala, Sweden
3
Laboratoire de Physique des Plasmas, Observatoire de Saint-Maur, Saint-Maur-Des-Fosss, 94107 France
The anisotropic nature of solar wind magnetic fluctuations is investigated scale-by-scale using
high cadence in-situ magnetic field measurements spanning five decades in scales from the inertial
to dissipation ranges of plasma turbulence. We find an abrupt transition at ion kinetic scales to a
single isotropic stochastic process that characterizes the dissipation range on all observable scales.
In contrast to the inertial range, this is accompanied by a successive scale-invariant reduction in the
ratio between parallel and transverse power. We suggest a possible phase space mechanism for this,
based on nonlinear wave-particle interactions, operating in this scale-invariant isotropic manner.

PACS numbers: 94.05.Lk, 52.35.Ra, 95.30.Qd, 96.60.Vg

In-situ measurements of fields and particles in the in-
terplanetary solar wind provide unique observations for
the study of a magnetohydrodynamic (MHD) turbulent
plasma [1–3]. An inertial range of turbulence is suggested
by the observations of a power-law power spectral den-
sity (PSD) over a finite range of scales [1, 4] set within
a highly structured solar wind flow [5, 6]. Unlike neu-
tral fluids where dissipation is carried out by viscosity,
interplanetary space plasmas are virtually collisionless in
the classical sense. On scales characteristic of the ions
there is a transition in the PSD from the inertial range
at lower frequencies to a steeper power-law spanning up
to two decades in scale [7–11]. This second interval of
scaling in the PSD was dubbed the dissipation range in
analogy with hydrodynamic turbulence [12]. The nature
of the fluctuations on these kinetic scales, the mecha-
nisms by which the turbulent energy is dissipated [8, 13]
and the role of dispersion of linear wave modes [14–16]
are all hotly debated [17, 18] as vital ingredients of any
future model of this dissipation range.
Since the background magnetic field orders both wave
propogation in configuration space and kinetic physics in
phase space [19], anisotropy [16, 20–23] is a central fea-
ture of plasma turbulence in the solar wind. Magnetic
field fluctuations are known [24, 25] to contain compo-
nents transverse and parallel to the background magnetic
field. The seminal study of [24] used Mariner 5 observa-
tions to investigate the field variance tensor projected
onto an orthonormal ‘field-velocity’ coordinate system.
They found the majority of the power is in the trans-
verse fluctuations with a ratio between transverse and
parallel components ∼ 9 : 1; this decreases to ∼ 5 : 1 in
the dissipation range as first seen in high cadence WIND
spacecraft data [12].
In this Letter we conduct a scale-by-scale study of the
anisotropy in transverse and parallel fluctuations over
five decades in temporal scales from a few hours to tens of
milliseconds. We show that the findings of [12, 24] are in
fact the result of a successive scale-invariant reduction in
the power ratio between the two components as we move
from one scale to the next in the dissipation range. By
comparing the probability density functions (PDF) for
the parallel and transverse fluctuations scale-by-scale we
will show that the dissipation range is characterised by a
single isotropic stochastic process; in contrast to the iner-
tial range where the corresonding signature is anisotropic.
Moreover, the transition in these processes between the
inertial and dissipation range is abrupt, occuring at ion
kinetic scales.
We stress that the anisotropy discussed in this Letter
is in the full vector direction of field fluctuations δB pro-
jected parallel and transverse to the local magnetic field
direction (also known as variance anisotropy [23, 24, 26]).
This is in contrast to the anisotropy measured in the vec-
tor r which characterises the scale of the fluctuations (or
wavevector k in Fourier space) [10, 20, 21, 25, 27]. In
the context of the latter anisotropy, the intervals stud-
ied here are dominanted by wavevectors strongly oblique
(60◦ - 90◦ ) to the mean background magnetic field. In
such intervals the inertial range is ubiquitously seen to
have a spectral index of ∼ −5/3, whilst in the dissipation
range this steepens to ∼ −2.9 [7].
We discuss an interval of quiet ambient solar wind for
which there are observations from both the Cluster and
ACE (Advanced Compositon Explorer) space missions
[28, 29]. The Cluster interval at 450 Hz cadence [7] is
of an hour duration 2007/01/30 00:10-01:10 UT when
the instruments were operating at burst mode, and will
primarily be used to study the dissipation range at space-
craft frequencies above 1 Hz. We construct a combined
data set from the DC magnetic field (sampled at 67 Hz)
of the Flux Gate Magnetometer (FGM) for frequencies
below 1 Hz, with the high frequency (oversampled at 450
Hz) search-coil data from the STAFF-SC experiment for
frequencies above 1 Hz, using the wavelet reconstruc-
tion method [10]. The ACE interval at 1 Hz cadence
is over two days 2007/01/30 00:00 UT to 2007/01/31
23:59 UT and samples the inertial range at spacecraft
frequencies below 1 Hz. Both these intervals are in sta-
tionary fast wind (≃ 667 km s−1 ) with similar plasma

parameters: average magnetic field B ≃ 4.5 nT , pro-
ton plasma β ≃ 1.2, plasma density ne ≃ 4 cm−3 ,
Alfvén speed VA ≃ 50 km s−1 , perpendicular ion tem-
perature Ti⊥ ≃ 24 eV , proton and electron temperatures
Tp ≃ 15 eV and Te ≃ 22 eV respectively, and ion and
electron gyroradii ρi ≃ 111.2 km and ρe ≃ 1.7 km re-
spectively.
We use the Undecimated Discrete Wavelet Trans-
form (UDWT) method (also known as the translation-
invariant wavelet transform [30]). Unlike the standard
discrete wavelet transform, the UDWT does not down-
sample the data at each stage of the transform. Thus
the UDWT provides information about the signal at each
observation time and so retains event information. This
timing information is necessary to project the fluctua-
tions onto a mean background field at each observation
time. We use the Daubechies 2 wavelet (db2), which is
a 4-tap high/low pass filter pair, chosen due to the com-
promise between time-frequency compactness, smoother
spectral index estimate, and as its first two moments are
zero it captures spectral indices as steep as -5 [31].
At each scale τ we high-pass filter the magnetic field
B(tj ) (sampled at discrete times tj ) to give a fluctuation
δB(tj , τ ) in terms of the wavelet coefficient (at time tj
based on a temporal scale τ ):
N
δB(tj , τ ) =
X
B(tk )
√ 
τ ψ(τ tk − tj ) ,
k=1
where ψ is the wavelet function, τ = 2i ∆ : i =
{0, 1, 2, 3, . . .} is the dyadic scale parameter and ∆ is the
sampling interval √ of the observations. Note that with-
out the factor τ and choosing ψ to be the db1 (Harr)
wavelet, the definition in eq. (1) is identical to calcu-
lating increments which are used in generalised stucture
function (GSF) analysis of intermittency studies in tur-
bulence [31]. The scale τ can be related to a central
2
the velocity direction unit vector time-averaged over
the entire interval. Since the solar wind velocity is
in a very fast and steady stream, and is within ∼ 3◦
of the GSE x direction in both the ACE and Clus-
ter intervals, it is reasonable to take a time averaged
global , rather than a local, velocity field. Together
[e⊥1 (tj , f ), e⊥2 (tj , f ), ek (tj , f )] form a time and scale de-
pendent orthornormal basis. The fluctuations parallel
δBk (tj , f ) and transverse [δB⊥1 (tj , f ) , δB⊥2 (tj , f )] to
the mean field are then given by (scalar product) pro-
jections onto this new basis. In keeping with Parseval’s
theorem for the conservation of the L2 -norm (energy con-
servation) the PSD is:
N
2∆ X
P SDk(⊥) (f ) = δBk(⊥) (tj , f )2 , (2)
N j=1
p
where δB⊥ (tj , f ) = 2 (t , f ) + δB 2 (t , f ) is the
δB⊥1 j ⊥2 j
magnitude of the total transverse fluctuation and N is
the sample size at each frequency. For the Cluster in-
terval the P SDk and P SD⊥ are shown in fig.1. The
lower panel of fig.1 shows that not only do these results
recover the 9 : 1 anisotropy ratio of [24] in a systematic
scale-by-scale way, they also show that the decrease in the
anisotropy observed by [12] in the dissipation range is ac-
tually a scale-free progression to isotropy. This progres-
(1) sion of the anisotropy in the power ratios P SDk/⊥ begins
at the ion gyroradius, ρi , and follows the power-law rela-
tionship P SDk/⊥ ∼ f 0.27 to the electron gyroradius, ρe .
The spectral indices obtained are ≃ −1.55±0.01 for both
components in the inertial range, and ≃ −2.67±0.01 and
≃ −2.94±0.01 for parallel and transverse components re-
spectively, in the dissipation range.
We next calculate higher-order statistics given by the
absolute moments of the fluctuations (the GSF) [7]. The
mth order wavelet structure function [31] is given by
N  m
frequency f in Hz [32] of a dyadic frequency band and so m 1 X  δBk(⊥) (tj , τ ) 
Sk(⊥) (τ ) = √  . (3)
we can write δB(tj , f ). The result of the low-pass filter
at each stage can similarly be written as B̃(tj , f ) using
the wavelet conjugate scaling function [30]. It is impor-
tant to note that both δB(tj , f ) and B̃(tj , f ) live in a
function space spanned by the bases constructed from
shifts and dilations of the db2 wavelet and scaling func-
tions (i.e. shift tj and dilation τ as in eq. (1)). We use
the inverse UDWT [30] operation on B̃(tj , f ) in order to
obtain the mean background field B(tj , f ), on which we
N j=1  τ
Scale-invariance is indicated by Sim (τ ) ∝ τ ζ(m) ; where
ζ(m) are the scaling exponents. The structure functions
and corresponding exponents are shown in fig. 2 for both
the inertial and dissipation ranges using the ACE and
Cluster intervals respectively. The scaling in the inertial
and dissipation ranges are distinct. The inertial range
shows multi-exponent scaling as evidenced by a nonlin-
then project the fluctuations δB(tj , f ) . The mean  field
 ear ζ(m) characteristic of solar wind turbulence at MHD
direction unit vector is ek (tj , f ) = B(tj , f )/ B(tj , f ).
This is distinct from the mean field coordinate presented
in [12, 24] as the mean field here is a locally varying
scale-dependent field consistent with the scale depen-
dent fluctuations.  The otherDtwo Eperpendicular
scales [1]. In contrast, the dissipation range is monoscal-
ing i.e. characterised by a linear ζ(m) = Hm and single
exponent H, consistent with [7]. Notably both paral-
lel and transverse fluctuations in the dissipation range
 DaxesE show monoscaling with similar exponents, suggestive of
are e⊥1 (tj , f ) = ek (tj , f ) × V̂ / ek (tj , f ) × V̂ 
  an isotropic underlying mechanism.
D E We finally look scale-by-scale at the individual PDFs
and e⊥2 (tj , f ) = ek (tj , f ) × e⊥1 (tj , f ), where V̂ is for the transverse (e⊥1 direction) and parallel fluctua-
 3

(a i.) &'()*)+,-+./0+/.),1/20+342- (b i.) &'()*)+,-+./0+/.),1/20+342-

C! ,
* "$

+.'2-().-)
+.'2-().-)
"$
$ "!
"!
%$

*45%!6:;9
*45 6: 9
;
%$
%

%!
%!
%!
&'()%*+,-*./0$*12!)3

<'.'**)*
<'.'**)*
$ $

!$ ! !

!$ !$ ,
! " # !" ! "
*45 6!,7-)0-89 *45%!6!,7-)0-89
%!
!#
(a ii.) @2).+3'*,.'25),-0'*325,)A<42)2+- (b ii.) F3--3<'+342,.'25),-0'*325,)A<42)2+-
%=# , $ ,
67:7&&9& <'.'**)*
!" %=" <'.'**)*
<:7/5C9:59 #
GH!=?I
+.'2-().-)
597:8D*8'E&*/'E59 % +.'2-().-)
GH!=J>
C
!! !=?

"6;9
"6;9
*
!=> "
!%4#
&'()%*+,-AAB!

DE:D,-+/11
!=#
!%4" %
!="
!%4!
!, !,
! " # ! " #
!) B4;)2+,; B4;)2+,;

!$ !) % )
&'( *567898:7;<*;:7=9*;:9>?9/8@*.123 Figure 2: (a i.) Transverse and parallel wavelet structure
)% functions of order 1−5 (from the bottom) and (a ii.) resultant
scaling exponents for the inertial range using the ACE data
interval. Structure functions have been vertically shifted for
Figure 1: (Upper panel) PSD (from Cluster) of the transverse clarity. (b i.) and (b ii.) similar to (a i.) and (a ii.) but in
and parallel components spanning both the inertial and dis- the dissipation range using the Cluster data interval.
sipation range. Sample size variance errors are smaller than
the markers. The search-coil noise floor PSD is obtained from
the z-component (spacecraft SR2 coordinates) of a very quiet
period in the magnetotail lobes (2007/06/30 15:00-15:05 UT)
as a proxy for the instrumentation noise. (Lower panel) Ratio # "

of parallel over transverse PSD. Vertical dashed and dashed- " Inertial !# Dissipation
67870090 67870090
dotted lines indicate ρi and ρe respectively; Doppler-shifted :87,*;98*9
Range :87,*;98*9
Range
!# !!
012#"(3*4!)*5+"(,-!#/

012#"(3*4!) 5+"(,- /

to spacecraft frequency using the Taylor hypothesis.

!#

!! !'
*

!' !&

!& !%
tions. These are shown in fig. 3 where we have used the !% !$
scaling operation Ps (δBi σ −1 ) = σP (δBi , τ ) to rescale
!$ !<
the fluctuations by their standard deviation so to of- !!" !#"
*
"
!() (+,-."/
#" !" !'" !!" !#"
*
"
!() (+,-."/
#" !" '"

fer a comparison of the functional form of the PDFs

for different τ . We show overlaid rescaled PDFs cor- # "

responding to four values of τ in the dissipation range, " Parallel !# Transverse

>=**=67:=1,(87,29 fluctuations >=**=67:=1,(87,29 fluctuations
τ = {0.036, 0.071, 0.142, 0.284} seconds, and in the in- !# =,98:=70(87,29
!! =,98:=70(87,29
012#"(3*4!)*5+"(,-!#/

012#"(3*4!)*5+"(,-!#/

ertial range, τ = {16, 32, 64, 128} seconds. From the !!

!'

point of view of a stochastic process, a given scaling be- !'

!&
haviour (exponents) and PDF functional form will corre- !&
!%
spond to a particular process – and thus by extension the !%

stochastic signature of a particular physical mechanism. !$ !$

In fig. 3 we see that in the dissipation range the same !<

!!" !#" " #" !"
!<
!'" !!" !#" " #" !" '"
!()*(+,-."/ !()*(+,-."/
PDF is obtained over the range of τ for both parallel
and transverse fluctuations – suggesting that the dissi-
pation range is in this sense ‘process isotropic’. This is Figure 3: Recaled transverse and parallel PDFs of fluctua-
in contrast to the inertial range where we see that the tions in the inertial and dissipation ranges (with Poisonnian
transverse and parallel PDFs are clearly different as we error bars). Four values of τ were used in both ranges.
would anticipate from their distinct ζ(m) shown in fig.

# (
!#$%
!'
012 (3 4!() 5(+"(,- /
!'
!'$%
!" D?7EF@!F190(,19*?(
*
GE1<()H(F1<I1,?,C
!"$% "##J.#K.&#('%L##!'%L#%(M-
*
4<72,?C1C790(01N?*5
'#
!&
!&$%
678**97,(:7;9<8<(
=9>?09@11A(B*C9<7C?
!!
!!$% (
!! !" # "
*
!() (+,-."/
Figure 4: Rescaled PDF of Bz magnetic fluctuations from the
instrument noise proxy of fig. 1, at the scales τ as in fig. 3.
2 (a ii). Both inertial and dissipation range PDFs are
non-Gaussian. In contrast, the rescaled PDFs of the SC
instrument noise shown in fig. 4 is Gaussian; confirming
that our results are robust to noise contamination. We
have analysed other intervals (2007/20/01 12:00-13:15UT
and 2007/20/01 13:30-14:10UT) and find consistent re-
sults with those presented above.
In this Letter we have performed the first systematic
scale-by-scale decomposition w.r.t. anisotropy of mag-
netic fluctuations in the solar wind across both the iner-
tial and dissipation ranges, from fluid to electron scales.
Our analysis characterises the dissipation range as ‘pro-
cess isotropic’ in the sense that the PDF functional forms
for parallel and transverse fluctuations are the same, and
‘power anisotropic’ in that the PSD shows greater power
in transverse compared to parallel fluctuations. This is
distinct from the inertial range which is both process and
power anisotropic. On MHD scales, any scale invariant
process must occur in configuration space. Consequently
one envisages a turbulent cascade in the sense of coherent
field and bulk fluid motions – indeed, the observations of
an anisotropic multi-exponent inertial range process are
consistent with this. Below ion kinetic scales, however, it
is the coherent motions in the ion phase space that trans-
late to fluctuations in the current and the magnetic field
which we then observe. Thus the scale-invariance that we
observe in magnetic fluctuations in the dissipation range
could arise as a consequence of a scale-invariant process
in the ion phase space i.e. a phase space cascade syn-
onymous with a fully nonlinear wave-particle interaction
(WPI) [19]. Such a process would capture the proper-
4
ties that we observe: process isotropy, since both parallel
and transverse fluctuating currents are supported by a
single nonlinear set of phase space orbits, and monoscal-
ing or fractal scale-invariance which is a general property
of nonlinear resonant WPI. The scale-invariant progres-
sion to isotropy in power would naturally arise from a
gyro-resonant process: as on scales just below that of
the ion Larmor orbits the phase space orbit topology is
controlled by the background magnetic field; whereas, as
we move to smaller scales, the phase space amplitudes be-
come progressively dominated by the details of the non-
linear phase space orbits. An open question is whether
these results depend upon bulk plasma parameters and,
in particular, the angle of the plasma flow velocity to the
local mean magnetic field over which we compute the
PSD and longitudinal structure functions.
We have established that there is a scale-independent
! isotropic mechanism operating in the dissipation range.
If this is the process responsible for damping then this
implies that the dissipation mechanism is isotropic. This
provides an important insight into the nature of the ‘dis-
sipation’ occurring in the dissipation range; it is an ir-
reversible scale-invariant phase space mixing via a gyro-
resonant nonlinear WPI – a mechanism which can in-
crease entropy in the absence of collisions.
The authors acknowledge the ACE Science Centre and
the Cluster instrument teams for FGM, STAFF-SC, CIS
HIA, WHISPER and PEACE. We also acknowledge the
DSP group at Rice University for the Rice Wavelet MAT-
LAB Toolbox; and C. Foullon and T. Dudok De Wit for
discussions. This work was supported by the UK STFC.
∗
Electronic address: k.kiyani@warwick.ac.uk
[1] C.-Y. Tu and E. Marsch, Space Sci. Rev. 73 (1995).
[2] P. J. Coleman Jr., Astrophys. J. 153, 371 (1968).
[3] T. S. Horbury, M. A. Forman, and S. Oughton, Plasma
Phys. Control. Fusion 47, B703 (2005).
[4] S. C. Chapman and R. M. Nicol, Phys. Rev. Lett. 103,
241101 (2009).
[5] W. H. Matthaeus and M. L. Goldstein, Phys. Rev. Lett.
57, 495 (1986).
[6] K. Kiyani et al., Phys. Rev. Lett. 98, 211101 (2007).
[7] K. H. Kiyani et al., Phys. Rev. Lett. 103, 075006 (2009).
[8] F. Sahraoui et al., Phys. Rev. Lett. 102, 231102 (2009).
[9] O. Alexandrova et al., Phys. Rev. Lett. 103, 165003
(2009).
[10] C. H. K. Chen et al., Phys. Rev. Lett. 104, 255002 (2010).
[11] V. Carbone et al., Phys. Rev. Lett. 104, 181101 (2010).
[12] R. J. Leamon et al., J. Geophys. Res. 103, 4775 (1998).
[13] J. A. Araneda et al., Phys. Rev. Lett. 102, 175001 (2009).
[14] S. P. Gary, S. Saito, and H. Li, Geophys. Res. Lett. 35
(2008).
[15] S. D. Bale et al., Phys. Rev. Lett. 94, 215002 (2005).
[16] R. J. Leamon et al., J. Geophys. Res. 104, 22331 (1999).
[17] O. Alexandrova et al., ApJ 674, 1153 (2008).
[18] G. G. Howes et al., Phys. Rev. Lett. 100, 065004 (2008).

[19] T. Tatsuno et al., Phys. Rev. Lett. 103, 015003 (2009).
[20] T. S. Horbury, M. Forman, and S. Oughton, Phys. Rev.
Lett. 101, 175005 (2008).
[21] J. J. Podesta., ApJ 698, 986 (2009).
[22] S. D. Bale et al., Phys. Rev. Lett. 103, 211101 (2009).
[23] S. Chapman and B. Hnat, Geophys. Res. Lett. 34 (2007).
[24] J. W. Blecher and L. Davis Jr., J. Geophys. Res. 76, 3534
(1971).
[25] W. H. Matthaeus, M. L. Goldstein, and D. A. Roberts,
J. Geophys. Res. 95, 20673 (1990).
5
[26] K. Hamilton et al., J. Geophys. Res. 113, A01106 (2008).
[27] K. Osman and T. Horbury, ApJ Lett. 654, L103 (2007).
[28] T. L. Garrard et al., Space Sci. Rev. 86 (1998).
[29] C. P. Escoubet et al., Space Sci. Rev. 79, 11 (1997).
[30] R. T. Ogden, Essential Wavelets for Statistical Applica-
tions and Data Analysis (Birkhäuser, 1997).
[31] M. Farge and K. Schneider, in Encyclopedia of Mathe-
matical Physics (Elsevier, 2006), pp. 408–419.
[32] P. Abry, Ondelettes et Turbulences (Paris, 1997).