Nonlinear Alfvén waves in extended magnetohydrodynamics

, ,
Hamdi M. Abdelhamid and Zensho Yoshida

Citation: Phys. Plasmas 23, 022105 (2016); doi: 10.1063/1.4941596

View online: http://dx.doi.org/10.1063/1.4941596
View Table of Contents: http://aip.scitation.org/toc/php/23/2
Published by the American Institute of Physics
 PHYSICS OF PLASMAS 23, 022105 (2016)

n waves in extended magnetohydrodynamics

Nonlinear Alfve
Hamdi M. Abdelhamid1,2,a) and Zensho Yoshida1,b)
1
Graduate School of Frontier Sciences, University of Tokyo, Kashiwanoha, Kashiwa, Chiba 277-8561, Japan
2
Physics Department, Faculty of Science, Mansoura University, Mansoura 35516, Egypt
(Received 5 November 2015; accepted 27 January 2016; published online 10 February 2016)
Large-amplitude Alfven waves are observed in various systems in space and laboratories,
demonstrating an interesting property that the wave shapes are stable even in the nonlinear regime.
The ideal magnetohydrodynamics (MHD) model predicts that an Alfven wave keeps an arbitrary
shape constant when it propagates on a homogeneous ambient magnetic field. However, such
arbitrariness is an artifact of the idealized model that omits the dispersive effects. Only special
wave forms, consisting of two component sinusoidal functions, can maintain the shape; we derive
fully nonlinear Alfven waves by an extended MHD model that includes both the Hall and electron
inertia effects. Interestingly, these “small-scale effects” change the picture completely; the large-
scale component of the wave cannot be independent of the small scale component, and the coexis-
tence of them forbids the large scale component to have a free wave form. This is a manifestation
of the nonlinearity-dispersion interplay, which is somewhat different from that of solitons. VC 2016

AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4941596]

I. INTRODUCTION equilibriums) on Casimir leaves, we derive nonlinear wave

solutions. The dispersion relation is exactly that of the linear
Alfven waves are the most typical electromagnetic
theory, while the wave amplitude may be arbitrarily large.
phenomena in magnetized plasmas. In particular, nonlinear
The wave function is composed of two components bearing
Alfven waves deeply influence various plasma regimes in
distinct length scales.
laboratory as well as in space, which have a crucial role in
plasma heating,1,2 turbulence,3–5 reconnection,6 etc. As an  WAVES IN IDEAL MHD
II. ALFVEN
interesting property of the Alfven waves, the amplitudes as
well as wave forms are totally arbitrary when they propagate Here, we consider an ideal MHD plasma obeying
on a homogenous ambient magnetic field.7,8 In fact, we often
@q
observe large-amplitude Alfven waves in orderly propaga- ¼ r  ðVqÞ; (1)
tion (for example, Ref. 9). To put it in theoretical language, @t
the set A of Alfven waves after an appropriate transformation @V
¼  ðr  V Þ  V þ q1 ðr  BÞ  B
(see Ref. 10) is a closed linear subspace, i.e., every linear @t  
combination of the members of A gives solution to the fully r V 2 =2 þ h ; (2)
nonlinear wave equation. Needless to say, the set of general
solutions to a linear equation is, by definition, a linear sub- @B
¼ r  ðV  B Þ; (3)
space. However, it is remarkable that the nonlinear magneto- @t
hydrodynamics (MHD) equation has such a linear subspace where q is the density, V is the velocity, B is the magnetic
A of solutions. field, and h(q) is the total enthalpy. All variables are written
Here, we investigate the underlying mechanism produc- in the standard Alfven units. The ideal MHD equation has a
ing such solutions in the context of a more accurate noncanonical Hamiltonian structure.14
framework, generalized MHD. When we take into account Hamiltonian and Poisson operator of MHD
dispersion effects (we consider both ion and electron inertial
effects11,12), the wave forms are no longer arbitrary (remem- ð   2  
jVj jBj2 3
ber that the ideal MHD model is dispersion free). Yet, we H¼ q þ U ðqÞ þ d x; (4)
X 2 2
find that the generalized MHD system has a linear subspace 0 1
of nonlinear solutions. The Casimir invariants of the system 0 r 0
are the root cause of this interesting property.13 B C
B C
We start by reviewing how the nonlinear Alfven waves J MHD ¼ B B r q 1
ðr  VÞ  䊊 q ðr  䊊Þ  B C:
1
C
are created in ideal MHD; we put the problem in the perspec- @ A
1
tive of Hamiltonian mechanics. We then formulate the 0 r  ð䊊  q BÞ 0
generalized MHD system in a Hamiltonian form. Via con- (5)
structing equilibrium solutions (so-called Beltrami
The existence of Casimir invariants is the signature of the
a)
Electronic mail: hamdi@ppl.k.u-tokyo.ac.jp noncanonicality, by which the orbits in the phase space are
b)
Electronic mail: yoshida@k.u-tokyo.ac.jp restricted to stay on the Casimir leaves (the level-sets of the

1070-664X/2016/23(2)/022105/5/$30.00 23, 022105-1 C 2016 AIP Publishing LLC

V
022105-2 H. M. Abdelhamid and Z. Yoshida Phys. Plasmas 23, 022105 (2016)

Casimir invariants).15 The equilibrium points are, then, the normalized to the ambient filed B0, the velocity to the Alfven
pffiffiffiffiffiffiffi
stationary points of the Hamiltonian
Ð (energy) on the Casimir speed (VA ¼ B0 = l q (l8 the vacuum permeability)), time
leaves. The cross helicity Ccross ¼ X V  B d 3 x (X is the total to the ion gyroperiod x1 ci , and the space variables to the ion
volume of the plasma) is one of the Casimir invariants of skin depth di.
MHD, which is relevant to the present purpose of construct- Equations (10)–(12) with the total energy
ing nonlinear Alfven waves. ð   2  
Minimizing the Hamiltonian H with the constraint on jVj B  B 3
H¼ q þ U ðqÞ þ d x; (14)
Ccross, we obtain X 2 2

V ¼ 6B: (6) and Poisson operator

0 1
Evidently, every B, being combined with V ¼ 6B, is an 0 r 0
equilibrium (@ t ¼ 0) solution of the ideal incompressible B C
B C
MHD equations. B ðrVÞ  䊊 ðr 䊊Þ B C
B C
We can convert these equilibrium stats to Alfven waves Br  C
B q q C
propagating on a homogeneous ambient magnetic field B0 B C
B " ! C
J ¼B   C;
(which can be arbitrarily chosen).13 Let us rewrite B and B ð䊊 B Þ ðr 䊊Þ B C
V ¼ 6B as B 0 r r C
B q q C
B !#C
B C
B ¼ B0 þ b; V ¼ 6B0 þ v: (7) B ðr 䊊Þ  ðrVÞ C
@ þde2 r A
q
Boosting the coordinate x ! x7B0 t, we find that the decom-
posed component (which is the wave component) satisfies (15)

@v   constitute a noncanonical Hamiltonian system in which the

¼ ðr  vÞ  v þ ðr  bÞ  ðb þ B0 Þ  r V 2 =2 þ P ;
@t phase space is spanned by the dynamical variables q, V, and
(8) B*. The Jacobi identity of the Poisson bracket defined by this
@b J was proved in Ref. 12. The reader is referred to Ref. 16,
¼ r  ½v  ðb þ B0 Þ; (9) where the relations among the generalized MHD, Hall
@t
MHD, and ideal MHD are delineated by suitable variable
which are exactly the Alfven wave equations with an ambi- transformations.
ent field B0. Notice that the wave component b and v propa- The extended MHD has three independent Casimirs
gate with the Alfven velocity 6B0. ð  
1
C1 ¼ B  V  2 A d3 x; (16)
III. EXTENDED MAGNETOHYDRODYNAMICS X 2de
ð
Now we will proceed to extend our analysis to a gener- 1    
C2 ¼ B  A þ de2 V  ðr  V Þ d 3 x; (17)
alized MHD. We start with the dimensionless extended 2 X
MHD equations,11,12 which are composed of ð
C3 ¼ q d 3 x: (18)
@q X
¼ r  ðqV Þ; (10)
@t
@V IV. BELTRAMI EQUILIBRIA
¼  ðr  V Þ  V þ q1 ðr  BÞ  B
@t  To construct the Beltrami equilibria, we start from the
2
r h þ V 2 =2 þ de2 ðr  BÞ =2q2 ; (11) energy-Casimir functional of the extended MHD system,
which reads as
@B  
¼ r  ðV  B Þ  r  q1 ðr  BÞ  B
@t X
3
  Hl ðuÞ ¼ HðuÞ  ln Cn ðuÞ: (19)
þ de2 r  q1 ðr  BÞ  ðr  V Þ ; (12)
n¼1

where The critical points on the Casimir leaves are found by setting
@u Hl ¼ 0 which yields
B ¼ B þ de2 r  q1 ðr  BÞ; (13)
 
l
de ¼ c/(xpedi) is the normalized electron skin depth normal- r  B ¼ l1 r  V þ l2  21 B ; (20)
de
ized to ion skin depth di ¼ c/(xpi), xpe and xpi are the elec-
tron and ion plasma frequencies, and c is the speed of light. qV ¼ l1 B þ l2 de2 r  V; (21)
Notice that for simplicity, barotropic pressure assumption is 2
used here. The above system of equations is normalized V2 ðr  BÞ
þ hðqÞ þ de2  l3 ¼ 0; (22)
in Alfvenic units defined as follows: the magnetic field is 2 2q2
022105-3 H. M. Abdelhamid and Z. Yoshida Phys. Plasmas 23, 022105 (2016)

where l1, l2, and l3 are Lagrange multipliers. Notice that Beltrami solutions, see Ref. 13. Let (k0 ¼ 0), two consequen-
(22) is Bernoulli’s equation. Now, consider the incompressi- ces immediately follow from (25)
ble flow (rV ¼ 0) with a constant mass density q ¼ 1. Then,
combining (20) and (21) with the aid of (13), we get the tri- l1 ¼ de2 l2 ¼ l; (28)
ple curl Beltrami equation k1 þ k2 ¼ g1 ;
r  r  r  B  g1 r  r  B þ g2 r  B  g3 B ¼ 0; k1 k2 ¼ g2 : (29)
(23)
Now, solving (29) yields
where  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
  k6 ¼ 16 1  4de2 ðl2  1Þ ; (30)
l 2lde2
g1 ¼ 2  21 D;
de l2
  ! where we chose kþ ¼ k1 and k ¼ k2. Under this conditions,
1 l1 l1 the general flow solution becomes
g2 ¼ l2 þ 2  2 1 þ D;
de l2 de l2
   
1
V ¼ la0 G0 þ ða1 G1 þ a2 G2 Þ: (31)
l1
g3 ¼ 1  2 de2 D ; l
de l2
   Based on the geometry (xyz—plane/space), the eigenfunc-
l l
D ¼ de2 l2  21 1 þ 1 : tions G1 and G2 are naturally a sinusoidal functions. To sat-
de l2
isfy the single Beltrami condition, the eigenfunctions are
The general solution of (23) can be expressed in terms of a given in the form of a circularly polarized wave
single Beltrami fields Gl(l ¼ 0, 1, 2), such that 0 1 0 1
sin ðk1 zÞ sin ðk2 zÞ
ðcurl  kl ÞGl ¼ 0 ðin XÞ; B C B C
G1 ¼ @ cos ðk1 zÞ A; G2 ¼ @ cos ðk2 zÞ A:
n  Gl ¼ 0 ðon XÞ; 0 0

for more details, see Refs. 10 and 17. Then, (23) can be writ- An immediate generalization to a more complex
ten as Beltrami field, so called ABC flow is possible [cf. Ref. 19].
These V and B have oscillatory amplitudes, thus the
ðcurl  k0 Þðcurl  k1 Þðcurl  k2 ÞB ¼ 0; (24)
Bernoulli condition (22) demands a non-constant h(q). We
where the eigenvalues k0, k1, and k2 are given by assume that the sound velocity is sufficiently large so that q
may be assumed to be constant and consistent to (23).
k0 þ k1 þ k2 ¼ g1 ; Considering density perturbation and discussing the coupling
k0 k1 þ k1 k2 þ k2 k0 ¼ g2 ; (25) between the resultant nonlinear Alfven wave and ion-sound
wave will be found in a forthcoming article; the methodol-
k0 k1 k2 ¼ g3 :
ogy can be found in Refs. 13 and 18, respectively. On the
Now constructing the general solution which is the linear other hand, Beltrami solutions (26) and (31) imply that the
combination of three eigenfunctions given as magnetic field and the flow velocity are not necessarily
aligned, unless l ¼ 61. Additionally, we observe that the
B ¼ a0 G0 þ a1 G1 þ a2 G2 ; (26) solutions are expressed as a combination of three Beltrami
eigenfunctions Gl, in which two of them have a large scale
where al’s are arbitrary constants. Substituting in (20) and (compared with the electron skin depth de), whereas the
(21), the corresponding flow is given by third is in scale hierarchy of de. Since in the Hall MHD limit
" # de ! 0, one of the eigenvalues (kþ ! (1  l2)/l) is finite,
 d2 l
2 2
V ¼ r 1 þ de k0 þ e 2
k0 a0 G0 whilst the other (k ! 1) is singular and therefore the
l1 corresponding eigenstate G2 is divergent; see Fig. 1. This
" # singularity can be removed by setting the arbitrary constant
 d2 l
2 2 e 2 (a2) associated with the divergent eigenstate (G2) to zero.
þ r 1 þ de k1 þ k1 a1 G1
l1 The physical insight and the derivation of the condition that
" # remove the singular part of the solutions will be the subject
 d2 l
2 2 e 2 of further publication.
þ r 1 þ de k2 þ k2 a2 G2 ; (27)
l1

d2 l V. EXACT SOLUTIONS OF THE EXTENDED MHD
where r ¼ l1 þ l2 1  le 2 .
1

Now, setting one of the Beltrami eigenvalues equal to To examine the propagation of the wave component,
zero (k0 ¼ 0) (which implies that the corresponding eigen- we assume that G0 serve as an ambient field. Now, setting
function is a harmonic field) yields a special class of G0 ¼ ^e z and a0 ¼ 1, (G0 represents the normalized ambient
022105-4 H. M. Abdelhamid and Z. Yoshida Phys. Plasmas 23, 022105 (2016)

~  vÞ
@ ðB  þ r ~  ½ v  ðB  þ r
¼r ~  vÞ; (39)
@s
which are the Alfven wave equations with a homogeneous
ambient field B0 ¼ ^e z . Thence, on the boosted frame, the
fluctuated parts of the previous stationary solution appear as
propagating waves, which form exact solution of the incom-
pressible extended MHD equations. Here, we notice that the
wave components are a superposition of two Beltrami eigen-
functions, which imply that only definite wave functions
(sinusoidal functions) can propagate with a fixed shape.
Further, the phase velocity here is given by l, which from
(29) may be written as
" rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi#
1 k k2  
l6 ¼  2 2
  6 þ 1 þ de2 k2 ; (40)
1 þ de k 2 4

where the eigenvalue k: ¼ k1 or k2 serve as the wave number.

Then, the corresponding circularly polarized wave dispersion
relation ðx ¼ lð^e z  kÞÞ which in the case k ¼ k ^e z reads as
" rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi#
k k k2  
x6 ¼    6 þ 1 þ de2 k2 ; (41)
1 þ de2 k2 2 4

FIG. 1. The profiles of the eigenfunctions G1 ; G2 and their superposition G ¼ which represents the dispersion relation of the fully nonlin-
a1 G1 þa2 G2 for l¼2, a1 ¼a2 ¼1; (a) de ¼0.26 and (b) de ¼108. ear wave solutions. In the limit (de ! 0), (41) is reduced to
the dispersion of exact solution of Hall MHD.13,19 We also
magnetic field). From (31), the corresponding ambient flow observe that the inclusion of the electron inertia effect not
is V 0 ¼ l ^e z . The magnetic and flow fields become only modifies waves modes, remove the singularities associ-
ated with the exact solution of Hall MHD, but also captured
B ¼ b þ ^e z ; V ¼ v þ l^e z ; (32)
more of the physics of the full two-fluid model; see Fig. 2.
where b ¼ lv: (33)

Let us show explicitly that the Beltrami solution (32) and (33)
can be modified to wave solution by boosting the coordinate.
The Beltrami solution is the stationary solution satisfying

0 ¼ r  ½ðV  r  BÞ  B ; (34)

0 ¼ r  ½V  ðB þ r  VÞ; (35)
r  V ¼ 0; (36)

r  B ¼ 0; r  B ¼ 0: (37)

Transforming the system under Galilean-boost yields the

new coordinates

ðx; y; zÞ7!ðx; y; nÞ :¼ ðx; y; z  ltÞ;

where t 7! s: ¼ t and z 7! n: ¼ z  lt. The transformations of

derivatives with respect to the coordinates are

~ x;y;n ; @ @ @
rx;y;z 7!r 7!  l ;
@t @s @n
where for 3-vector X (with r X ¼ 0), l @X @n ¼ r 
ðl^e z  X Þ is true. Using (32), Equations (34) and (35) can be
boosted in the new coordinates into

@B ~  BÞ  B ;
~  ½ ðv  r
¼r (38) FIG. 2. Normalized dispersion relation profiles for de ¼ 0 (dashed-red) and
@s de ¼ 0.0233 (blue); (a) (x) and (b) (xþ).
022105-5 H. M. Abdelhamid and Z. Yoshida
The previous sentence can be clarified by studying the
extreme limits of Equation (41):
1. for k  1
l6 ! 61; x6 ! 7k;
which represent the shear Alfven wave in the ideal MHD
limit.
2. for k > 1 but de2 k2  1
lþ ! 1=k; xþ ! 1;
l ! k; x ! k2 ;
xþ represents here ion gyrofrequency, whilst x repre-
sents whistler wave.
3. for k 1 and de2 k2 1
l6 ! h6 =k; x6 ! h6 ;
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where h6 ¼ ð16 1 þ 4de2 Þ=2de2 are constants (h
approximates the normalized electron gyrofrequency, hþ
approximates the normalized ion gyrofrequency). An im-
portant feature here is the dependence of h on the dimen-
sionless electron skin depth, which represent a direct
relation between the electron skin depth and ion(electron)
gyrofrequency.
VI. CONCLUSIONS
In summary, we have given exact wave solutions of the
fully nonlinear extended MHD system. The solutions consist
of two Beltrami eigenfunctions with different length scales
(macroscopic and microscopic). In contrast to the ideal
MHD predictions, inclusion of the small scale effects (Hall
and electron inertia effects) caused only two component
sinusoidal wave functions to preserve its shape. As shown in
(33), the partition of the wave energy between b and v is
determined by the phase velocity l that is a function of k
(wave number). We can observe that for k > 1 only whistler
wave has a magnetic energy more than the kinetic energy
Phys. Plasmas 23, 022105 (2016)
ðb ! kvÞ, otherwise the kinetic energy is the dominant
ðv ! kbÞ. This is also in a marked contrast to the ideal
Alfven waves in which the wave energy is equally parti-
tioned by b and v.
ACKNOWLEDGMENTS
The authors thank Professor P. J. Morrison and Dr. E.
Tassi for many useful discussions. H. M. Abdelhamid would
like to thank the Egyptian Ministry of Higher Education for
supporting his research activities. This work was partly
supported by JSPS KAKENHI Grant Nos. 23224014 and
15K13532.
1
A. Hasegawa and L. Chen, Phys. Rev. Lett. 32, 454 (1974).
2
S. W. McIntosh, B. De Pontieu, M. Carlsson, V. Hansteen, P. Boerner, and
M. Goossens, Nature 475, 477–480 (2011).
3
S. Sridhar and P. Goldreich, Astrophys. J. 432, 612 (1994).
4
P. Goldreich and S. Sridhar, Astrophys. J. 438, 763 (1995).
5
S. Boldyrev, Phys. Rev. Lett. 96, 115002 (2006).
6
B. N. Rogers, R. E. Denton, J. F. Drake, and M. A. Shay, Phys. Rev. Lett.
87, 195004 (2001).
7
A. Hasegawa and C. Uberoi, The Alfv en Wave (National Technical
8
Information Service, Springfield, VA, 1982).
N. F. Cramer, The Physics of Alfven Waves (Wiley, Berlin, 2001).
9
N. Nishizuka, M. Shimizu, T. Nakamura, K. Otsuji, T. J. Okamoto, Y.
Katsukawa, and K. Shibata, Astrophys. J. 683, L83 (2008).
10
S. M. Mahajan and Z. Yoshida, Phys. Rev. Lett. 81, 4863 (1998).
11
K. Kimura and P. J. Morrison, Phys. Plasmas 21, 082101 (2014).
12
H. M. Abdelhamid, Y. Kawazura, and Z. Yoshida, J. Phys. A: Math.
Theor. 48, 235502 (2015).
13
Z. Yoshida, Commun. Nonlinear Sci. Numer. Simul. 17, 2223 (2012).
14
P. J. Morrison and J. M. Greene, Phys. Rev. Lett. 45, 790 (1980).
15
P. J. Morrison, Rev. Mod. Phys. 70(2), 467 (1998).
16
M. Lingam, P. J. Morrison, and G. Miloshevich, Phys. Plasmas 22, 072111
(2015).
17
Z. Yoshida and S. M. Mahajan, Phys. Rev. Lett. 88, 095001 (2002).
18
S. Emoto and Z. Yoshida, Commun. Nonlinear Sci. Numer. Simul. 19,
53–59 (2014).
19
S. M. Mahajan and V. Krishan, Mon. Not. R. Astron. Soc. 359, L27–L29
(2005).