Vous êtes sur la page 1sur 9

Weakly relativistic quantum kinetic theory for electrostatic wave modes in magnetized

plasmas
Azhar Hussain, Martin Stefan, and Gert Brodin

Citation: Physics of Plasmas (1994-present) 21, 032104 (2014); doi: 10.1063/1.4867490
View online: http://dx.doi.org/10.1063/1.4867490
View Table of Contents: http://scitation.aip.org/content/aip/journal/pop/21/3?ver=pdfcov
Published by the AIP Publishing




















This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
Weakly relativistic quantum kinetic theory for electrostatic wave modes in
magnetized plasmas
Azhar Hussain,
1
Martin Stefan,
2
and Gert Brodin
2
1
Department of Physics, GC University Lahore, 54000 Lahore, Pakistan
2
Department of Physics, Umea University, SE-901 87 Umea, Sweden
(Received 25 October 2013; accepted 20 February 2014; published online 5 March 2014)
We have derived the electrostatic dispersion relation in a magnetized plasma using a recently
developed quantum kinetic model based on the Dirac equation. The model contains weakly
relativistic spin effects such as Thomas precession, the polarization currents associated with the
spin and the spin-orbit coupling. It turns out that for strictly electrostatic perturbations the non-
relativistic spin effects vanish, and the modication of the classical dispersion relation is solely
associated with the relativistic terms. Several new wave modes appear due the electron spin effects,
and an example for astrophysical plasmas are given. VC
2014 AIP Publishing LLC.
[http://dx.doi.org/10.1063/1.4867490]
I. INTRODUCTION
During the last 510 years there has been an increased
interest in the eld of quantum plasmas, see e.g., the review
articles.
14
The interest is motivated by emerging applica-
tions in, e.g., quantum wells,
5
plasmonics,
6,7
spintronics,
8
astrophysics,
9
and ultra-cold plasmas.
10
The requirements
for quantum effects to be of importance are typically that the
plasma is of high density and not of too high temperature,
see e.g., Refs. 14 for detailed discussions. The theoretical
descriptions of quantum plasmas vary in scope and complex-
ity. Typically hydrodynamic approaches cover effects due to
particle dispersion and the Fermi pressure,
11,12
whereas
much of the kinetic treatments are based on the Wigner-
Moyal equation.
1
More accurate treatments based on the
Kadanoff-Baym kinetic equations that also includes relaxa-
tion processes
13,14
are also common. Moreover, the magnet-
ization currents and magnetic dipole force due to the
electron spin have been included in hydrodynamic
15
as well
as kinetic theories.
16,17
In the present paper, we will start from a recently devel-
oped quantum kinetic model,
17
extending previous quantum
mechanical models,
16
to include weakly relativistic effects
such as spin-orbit interaction and Thomas precession.
18
The
model is used to analyze linear electrostatic waves propagat-
ing at an arbitrary angle with respect to an external magnetic
eld. A general dispersion relation is derived which reduces
to the well-known classical result when the limit h 0 is
taken. One of the main effects of the full model is to allow
for new electrostatic wave modes that is absent unless the
spin-orbit interaction is present. The dispersion relation is
solved numerically for an example of this kind. It turns out
that the new wave modes are localized to rather narrow fre-
quency bands, unless the plasma density is very high and/or
the external magnetic eld is very strong.
The organization of the paper is as follows: In Sec. II we
present the model kinetic equation together with the current
densities in a semi-relativistic regime. Next the general
theory for electrostatic waves involving magnetization and
polarization currents due to the electron spin is studied in
Sec. III. In Sec. IV the kinetic equation of Sec. II is solved
and the general electrostatic dispersion relation is found.
Section V presents a numerical solution of the dispersion for
parameters of astrophysical interest, and nally our result is
discussed and summarized in Sec. VI.
II. THE WEAKLY RELATIVISTIC MODEL
In a recent work, Asenjo et al.
17
derived a weakly rela-
tivistic model based on the Dirac equation. By making suc-
cessive FoldyWouthuysen transformations
19
of the Dirac
Hamiltonian, a weakly relativistic Hamiltonian can be
calculated that only contains the positive energy states
(i.e., the electrons). The density matrix can then be con-
structed, whose evolution are described by the weakly rela-
tivistic Hamiltonian. By making a Wigner-Stratonovich
transformation
20
of the density matrix, together with a
Q-transformation (see Eq. (12) in Ref. 17) a scalar quasi
distribution function can be dened. As a result, the kinetic
evolution equation for a weakly relativistic spin 1/2 colli-
sionless plasma in the long scale length limit, to order c
2
and h
2
, where c is the speed of light in vacuum and h is
Plancks constant, is found to be
0 =
@f
@t

p
m
e

l
2m
e
c
E (s \
s
)
_ _
\
x
f
q E
1
c
p
m
e

l
2m
e
c
E (s \
s
)
_ _
B
_ _
\
p
f

2l
h
s B
p E
2m
e
c
_ _
\
s
f l s \
s
( )
@
i
x
B
p E
2m
e
c
_ _ _ _
@
i
p
f ; (1)
where f = f (x; p; s; t) is the quasi-distribution function
dened on a phase space extended by two spin dimensions
(denoted s) on the unit sphere, in addition to the traditional
space and momentum coordinates x and p. We use the nota-
tion m for the mass and l is the magnetic moment of the par-
ticle (= geh=4m
e
for electrons, where g = 2:002319 is the
1070-664X/2014/21(3)/032104/8/$30.00 VC
2014 AIP Publishing LLC 21, 032104-1
PHYSICS OF PLASMAS 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
spin g-factor), q is its charge. The index x, p, or s on the
nabla operator indicates that it acts on the respective coordi-
nates. When writing down this equation the last term of the
equation derived in Ref. 17 was omitted. This term is associ-
ated with the Darwin term in the Hamiltonian, and was
dropped since in a long scale length expansion it is smaller
than the other terms.
21
This Vlasov-like equation is coupled to Maxwells
equations
\ E = 4pq
T
; (2)
\B =
1
c
@E
@t

4p
c
J
T
; (3)
where the total charge and current density are given by
q
T
= q
F
\ P; (4)
J
T
= J
F
\M
@
@t
P: (5)
Here q
F
= q
_
dXf is the free charge density and the free cur-
rent density, the polarization and magnetization are given by
J
f
= q
_
dX
p
m
e

3l
2m
e
c
E s
_ _
f ; (6)
P = 3l
_
dX
s p
2m
e
c
f ; (7)
M= 3l
_
dXsf ; (8)
respectively, where dX = d
3
p sin h
s
dh
s
d/
s
using spherical
coordinates h
s
and /
s
in spin space.
While the above model may look complicated at rst, it
is possible to relate it to the classical Vlasov equation. The
rst difculty is to relate the momentum variable to the ki-
netic velocity. From Eq. (6), we observe that the kinetic ve-
locity v and the momentum are related through
v = v(x; p; s; t) =
p
m
e

3l
2m
e
c
E s; (9)
which also is the expression used by Ref. 17. This makes the
momentum-spin combination that appears in second and third
term of Eq. (1) easy to interpret if we make the substitution
s \
s
3s. This may seem somewhat unjustied, and it
must be stressed that this replacement is not generally possi-
ble. However, such a substitution is nevertheless not without
merit, as demonstrated by the calculation in Ref. 16 where it
is shown that s \
s
3s applies in case the particles are in
a coherent spin-up or spin-down state. Once the spin contribu-
tion to the velocity momentum relation is understood, Eq. (1)
is a rather straightforward generalization of Eq. (64) in Ref.
16. We note that here that the fourth term of (1) describing
spin precession and the fth term describing the magnetic
dipole force is modied according to B (B p E=2mc),
where the well-known Thomas factor of 1/2 appears.
18
It
should be stressed here the spin contribution to the momentum
can be neglected in the Thomas factor, since it is of higher
order in an expansion in 1=mc
2
:
For a more complete discussion of the model presented
here containing, e.g., conservation laws see Ref. 17.
III. GENERALTHEORY FOR ELECTROSTATIC WAVES
Before we start to consider the kinetic theory of Sec. II
in any detail, it is convenient to derive some fairly general
results for electrostatic wave propagation in homogeneous
media. In addition to well-known standard results we will
here account for magnetization currents and polarization cur-
rents that occur as a result of the electron spin properties. In
general for linear waves in homogeneous medium we can
write the current density as
j
i
= r
ij
E
j
; (10)
where the conductivity tensor is a function of frequency and
wavevector, i.e., r
ij
= r
ij
(x; k) and we have used the sum-
mation convention. The conductivity tensor as dened here
includes all currents, i.e., free currents, magnetization cur-
rents, and polarization currents. Eliminating the magnetic
eld in Amperes law using Faradays law we obtain
1
k
2
c
2
x
2
_ _
d
ij

k
i
k
j
c
2
x
2

ir
ij
e
0
x
_ _
E
j
= 0: (11)
The non-trivial solutions is obtained by putting the determi-
nant to zero, which give the dispersion relation as
det 1
k
2
c
2
x
2
_ _
d
ij

k
i
k
j
c
2
x
2

ir
ij
e
0
x
_ _
= 0: (12)
While for an oblique direction of propagation, the waves are
typically not exactly electrostatic, we can let k
2
c
2
=x
2

to get the electrostatic limit, assuming the conductivity ten-


sor to be of order unity in this expansion parameter. To order
k
6
c
6
=x
6
we get an identity that is always fullled and to
order k
4
c
4
=x
4
we get
k
2
c
4
x
4
[k
2
l
r
xx
ie
0
x ( ) k
l
k
z
(r
xz
r
zx
) k
2
z
r
zz
ie
0
x ( )[ =0;
(13)
where the wave-vector is assumed to lie in the xz-plane for
convenience. Eq. (13) shows that only four conductivity ten-
sor components are needed, and agrees with the dispersion
relation one gets by assuming E = \U from the start, and
using Poissons equation rather than the full set of
Maxwells equations. While the above derivation of the elec-
trostatic limit did not rely on any specic model for the par-
ticles such as, e.g., the Vlasov equation, it still does not
apply directly to the spin-kinetic model we are interested in.
The reason is that the assumption that r
ij
is of order unity in
our expansion parameter k
2
c
2
=x
2
does not apply. In particu-
lar, certain terms in r
ij
as k
2
c
2
=x
2
. To address
our situation, it is therefore convenient to divide the current
density into different contributions as
032104-2 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
r
ij
= r
free
ij
r
magn
ij
r
pol
ij
; (14)
where the three conductivity tensor components are the ones
resulting from the free current density, the magnetization
current density and the polarization current density, respec-
tively. The terms in the conductivity tensor diverging when
k
2
c
2
=x
2
is the ones proportional to the magnetization
current density, i.e., certain terms in r
magn
ij
. Specically the
diverging terms r
d
ij
can be written
22
r
d
ij
= S
c
2
x
2
k
2
z
ik
2
z
k
l
k
z
ik
2
z
k
2
z
ik
l
k
z
k
l
k
z
ik
l
k
z
k
2
l
_
_
_
_
_
_; (15)
where S is a non-diverging scalar and we have assumed that
the wave vector lies in the xz-plane, i.e., k = k
l
^ x k
z
^z.
However, although the individual terms become large, the
structure of the matrix r
d
ij
implies that its contribution to the
total dispersion relation nevertheless vanish to the order
k
4
c
4
=x
4
.
Next, we note that the magnetic eld gives a contribu-
tion to the free current density and also to the polarization
current density. These contributions come partly from the
magnetic dipole force term and partly from the spin-
precession term in Eq. (1). Thus, we can divide
r
free
ij
= r
freeE
ij
r
freeB
ij
; (16)
where the second term comes from the source term that ex-
plicitly depend on the perturbed magnetic eld in (1). A sim-
ilar division can be made of the polarization current density.
However, since B
i
= e
ijk
k
j
E
k
=x, the contribution r
freeB
ij
will
be of the form
j
freeB
i
= v
freeB
ij
B
j
= v
freeB
ij
e
jkl
k
j
E
k
=x; (17)
such that r
freeB
ij
= v
freeB
ij
e
jkl
k
j
where v
freeB
ij
is a magnetic
susceptibility tensor whose detailed properties are irrelevant
for our discussion. The important point is that the antisym-
metry property of the Levi-Chivita tensor e
ijk
means that the
contribution from r
freeB
ij
cancel when the electrostatic limit
is taken (i.e., the contributions to the general dispersion rela-
tion in Eq. (12) cancel to order k
4
c
4
=x
4
). An identical argu-
ment holds for r
polB
ij
. Thus, in order to calculate the general
dispersion relation, we need to nd four tensor components
(r
xx
; r
xz
; r
zx
and r
zz
as in Eq. (13)), and only the direct con-
tribution from the electric eld, r
freeE
ij
rather than r
free
ij
, and
r
polE
ij
rather than r
pol
ij
. Taking this into account keeping
terms up to order k
4
c
4
=x
4
we nd that Eq. (12) reduces to
sin
2
h r
freeE
xx
r
polE
xx
_ _
sin h cos h r
freeE
xz
r
polE
xz
r
freeE
zx
r
polE
zx
_ _
cos
2
h r
freeE
zz
r
polE
zz
_ _
= ie
0
x; (18)
where the angle h is dened by k = k
l
^ x k
z
^z:
= k(sin h^ x cos h^z). Eq. (18) can be derived more directly
(but less rigorously) from Poissons equation. It should be
noted that while the full tensor r
ij
is Hermitian when the
pole contributions are neglected, the same is not true for the
sub-parts, r
freeE
ij
, etc. As a result we cannot invoke symme-
try arguments for the tensor and must compute both r
freeE
xz
and r
freeE
zx
, and similarly for the polarization current
contribution.
IV. THE ELECTROSTATIC DISPERSION RELATION
A. Solutions of the linearized kinetic equation
Here, we assume that the magnetic eld can be written
as B = B
0
B
1
, where subscripts 0 and 1 are used for equi-
librium and perturbed quantities. (The electric eld is simply
written as E
1
, since the equilibrium part is taken to be zero,
i.e., E
0
= 0:) We also write the total distribution function as
f = f
0
f
1
. First, we discuss the thermodynamic expressions
for the background spin and momentum distribution given
by f
0
. In Eq. (60) by Ref. 16 a general expression for f
0
is
given for a magnetized plasma, where the z-axis is taken
along the unperturbed magnetic eld, i.e., B
0
= B
0
^z. The
general formula is very complicated due to the effect of
Landau quantization, which is signicant whenever the
Zeeman energy lB
0
is comparable to the kinetic energy E
k
(which can be either k
B
T or k
B
T
F
, depending on whether the
temperature T or the Fermi temperature T
F
is larger). In
order to simplify the expressions, we will consider the limit
when lB
0
< E
k
, such that the effect of Landau quantization
is reasonably small. For this case, the distribution function in
thermodynamic equilibrium can be written as
f
0
= f
0
(p
2
; h
s
) =
^
f
0
(p
2
)[c

(1 cos h
s
) c

(1 cos h
s
)[
(19)
independent of the ordering between T and T
F
. Here,
^
f
0
(p
2
)
is normalized such that
_
^
f
0
(p
2
)d
3
p = 1, and the constants
c

and c

should produce the number density of the spin-up


and spin-down states, respectively, i.e.,
_
^
f
0
(p
2
)c
6
(16cos h
s
)
dX = n
06
, where the total number density is n
0
=n
0
n
0
.
First picking the case T
F
T (corresponding to a large and
negative chemical potential in Eq. (60) of Ref. 16) we obtain
c
6
=n
06
=4p =n
0
exp(6lB
0
=k
B
T)=[cosh(lB
0
=k
B
T)4p[. As
a result, the degree of spin polarization in the unperturbed
state is found to be n
0
n
0
=n
0
tanh(lB
0
=k
B
T) for
T
F
T. Note that since we have assumed the Zeeman
energy to be small, we should not apply this expression for a
too large argument of the tanh-function, such that a rst or
second order Taylor-expansion will sufce in most cases.
Next, we consider the opposite regime T
F
T, in which
case the chemical potential should be equated with the Fermi
energy k
B
T
F
, and as a consequence
^
f
0
(p
2
) changes from a
Maxwellian to a Fermi-Dirac distribution (i.e.,
^
f
0
(p
2
)
exp(p
2
=m
e
v
2
t
) changes to
^
f
0
(p
2
) being constant for
p
2
<p
2
F
=m
e
k
B
T
F
=2). Also using lB
0
k
B
T
F
(to make
Landau quantization negligible) we then get
c
6
=n
06
=4p (n
0
=2)[163lB
0
=2k
B
T
F
[=4p. Now the spin
polarization in the unperturbed state becomes
n
0
n
0
=3n
0
lB
0
=2k
B
T
F
. As pointed out above, the
expressions presented here are the thermodynamic equilib-
rium expressions. If the unperturbed magnetic eld changes
faster than the spin relaxation time, it may be adequate to
032104-3 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
apply another degree of spin polarization in the unperturbed
state (see, e.g., Ref. 22). In the present work, we will limit
ourselves to a background state in thermodynamic equilib-
rium, however.
The linearization of eq. (1) in the electrostatic limit
gives us
0 =
@f
1
@t

p
m
e
\
x
f
1

2l
h
s B
0
\
s
f
1
l@
i
x
[B
0
(s \
s
)[@
i
p
f
1
qE
1
\
p
f
0

l
2m
e
c
@
i
x
(p E
1
) (s \
s
) [ [@
i
p
f
0

ql
2m
e
c
2
[E
1
(s \
s
)[ B
0
\
p
f
0

l
hm
e
c
s (p E
1
) \
s
f
0
: (20)
In the electrostatic limit terms proportional to B
1
are dropped
by denition. However, generally for waves propagating at
an angle to B
0
, the perturbed magnetic eld B
1
is nonzero.
Nevertheless for sufciently short wavelengths it is possible
to neglect these term as demonstrated in Sec. III.
To solve Eq. (20) for the perturbed distribution function
f
1
we make the ansatz
f
1
=

a;b=
g
ab
(p
l
; p
z
; h
s
)w
a
(/
p
; p
l
)
1

2p
_ exp(ib/
s
);
(21)
where
w
a
(/
p
; p
l
) =
1

2p
_ exp[i(a/
p
k
l
p
l
sin /
p
=x
ce
)[
=
1

2p
_

l=
J
l
k
l
p
l
m
e
x
ce
_ _
exp[i(l a)/
p
[; (22)
x
ce
is the electron cyclotron frequency and J
l
is the Bessel
function of order l. Using cylindrical coordinates for the mo-
mentum variable (p
l
; /
p
; p
z
) and spherical coordinates for
the spin variable (h
s
; /
s
), and following the standard techni-
ques
22
based on the orthogonality of the w
a
eigenfunctions
(see, e.g., Ref. 22 for technical details), we have
f
1
=

a;l=
J
l
(g)e
i(la)/
p
2i
Ae
i/
s
xk
z
p
z
=m
e
ax
ce
x
cg
_

B
xk
z
p
z
=m
e
ax
ce

Ce
i/
s
xk
z
p
z
=m
e
ax
ce
x
cg
_
f
0
;
(23)
where g = k
l
p
l
=m
e
x
ce
and x
cg
= 2lB
0
=h = (g=2)x
ce
are
the spin precession factors. Note that the spin precession
frequency is very close to the electron cyclotron frequency.
The expressions for A, B, and C are exposed in the
Appendix A. Note that in the expressions for A, B, and C
and throughout the text below, it is understood that the
argument of the Bessel function is k
l
p
l
=m
e
x
ce
( ), unless
stated otherwise.
B. Conductivity components for the free and
polarization current densities
According to our notation, we let all currents within lin-
ear theory be computed from a conductivity tensor, j
i
= r
ij
E
j
,
where j
i
contain the free currents, the magnetization currents
as well as the polarization currents. Each of these three contri-
butions are well dened, and we may thus separate the differ-
ent contributions into r
ij
= r
free
ij
r
magn
ij
r
pol
ij
. Furthermore,
some of the contributions to r
ij
have the electric eld as a
direct source, whereas (if we avoid the electrostatic approxi-
mation) some have the perturbed magnetic eld as a source.
Since we can always rewrite the magnetic eld in terms of the
electric eld using Faradays law, the magnetic eld contribu-
tions can still be included in r
ij
, which is the convention we
adopt here. Nevertheless, it is of interest to separate the terms
depending on the underlaying source term, and thus all differ-
ent contributions (r
free
ij
; r
magn
ij
; r
pol
ij
) can be divided into those
originating from the electric eld and those originating from
the (perturbed) magnetic eld. Hence, we make the division
r
free
ij
= r
freeE
ij
r
freeB
ij
, where r
freeE
ij
originates from the
electric eld terms in Eq. (1) whereas r
freeB
ij
originates from
the magnetic eld terms. A similar division is made of r
pol
ij
.
As demonstrated in Sec. II we note that for short wavelengths
electrostatic approximations can be made, which means that
the terms r
magn
ij
r
freeB
ij
and r
polB
ij
do not contribute to the dis-
persion relation, and furthermore only four out of the nine
Cartesian tensor components are needed. Given this, we calcu-
late the four relevant components of r
freeE
ij
and r
polE
ij
.
A technical detail that must be kept in mind is that the
linearized free current density is not only computed from the
perturbed distribution function. From Eq. (6) we stress that it
is given by
J
f
= q
_
dX
p
m
e
f
1
J
f
=
3ql
2m
e
c
_
dXE sf
0
: (24)
Substituting Eq. (23) into Eq. (24), the required components
of the free part of the conductivity tensor can be read off,
and the result after some of the integrations have been car-
ried out is given in the Appendix B. Similarly, the linearized
expression for the polarization density is given by
P = 3l
_
dX
s p
2m
e
c
f
1
; (25)
and Eq. (23) is substituted into Eq. (25) to compute the line-
arized polarization current density J
p
= @p=@t: With help of
this, we can read off the polarization part of the conductivity
tensor, and the required four components are presented in
Appendix B.
In addition to polarization currents, we also have mag-
netization currents. However, as pointed out above and dem-
onstrated in Sec. II these terms do not contribute to the
dispersion relation in the electrostatic limit. Interestingly,
this is not because they are necessarily small compared to,
e.g., the polarization currents, but because the symmetry
properties of various parts of r
ij
means that cancellations
occur in the dispersion relation for short wavelengths.
032104-4 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
V. SOLUTIONS TOTHE DISPERSION RELATION
There are a number of parameters that determine the rel-
ative magnitude of the quantum contributions to the electro-
static dispersion relation Eq. (18) derived in Sec. II. These
include hx
p
=m
e
c
2
; hx
p
_
=k
B
T and l
B
B
0
=k
B
T, where x
p
is the
plasma frequency. The different quantum terms are generally
proportional to some combination of these factors. If the
Fermi temperature T
F
is higher than the electron tempera-
ture, T is typically replaced by T
F
when the order of magni-
tude of the quantum terms are estimated. It should be noted
that only the quantum relativistic effects contribute in the
electrostatic case. For the electromagnetic case, when the
non-relativistic spin contributions survive, the magnitude of
these terms can be signicantly larger (cf. Ref. 22).
Excluding the fully quantum relativistic regime (where
hx
p
=m
e
c
2
~ 1), the quantum terms of Eq. (18) may still be
signicant if we are close to any of the resonances of the
new terms. As an illustrative example, let us consider waves
with a frequency close to 2x
ce
propagating perpendicular to
the magnetic eld, i.e., we let k
z
= 0 in which case the gen-
eral dispersion relation (18) immediately reduces to
r
freeE
xx
r
polE
xx
= ie
0
x. Furthermore, since by assumption
x ~ 2x
ce
, in the sums over a we can focus on the terms with
small denominators. Due to the fact x
ce
x
cg
these are
x (3x
ce
x
cg
); x (x
ce
x
cg
), and x 2x
ce
, and as a
further consequence we are also dropping the ion contribu-
tion. Moreover, we pick the case with T T
F
, in which case
we have a Fermi-Dirac momentum distribution, and a ther-
modynamic equilibrium background spin polarization given
by n
0
n
0
= 3n
0
lB
0
=2k
B
T
F
. Under these conditions, the
dispersion relation can be written
1 =
12x
2
pe
(x 2x
ce
)
m
2
e
x
ce
k
2
l
p
2
F
_
1
1
du J
2
2
(f

1 u
2
_
)
(hx
pe
)
2
64m
2
e
c
4

3g
2
x
ce
x (x
cg
x
ce
)
_
1
1
du 4u
4
u
2
( )
J
2
1
(f

1 u
2
_
)

g
2
(hx
pe
)
2
64m
2
e
c
4
3g
2
x
ce
x (3x
ce
x
cg
)
_
1
1
du 4u
4
3u
2
( )
J
2
3
(f

1 u
2
_
); (26)
where v
F
= 2k
B
T
F
=m
e
( )
1=2
is the Fermi velocity, f =
kv
F
=x
ce
(with k = k
l
) and we have used the properties of
the Heaviside and delta functions to carry out one of the mo-
mentum integrations, and the remaining variable of integra-
tion is changed to u = p
z
=m
e
v
F
. In the above expression,
f

1 u
2
_
is the argument of the Bessel functions. The rst
term of the right hand side gives the Bernstein resonances,
which is now complemented by the nearby spin resonances.
Solving the dispersion relation numerically for a magnetic
eld of 10
9
T and a number density 7 10
34
m
3
results in
three wave modes that are shown in Fig. 1. The frequencies
approach the precise resonances for very large wavenumber,
but deviates somewhat from the resonance frequencies for
smaller k. The middle curve is the classical Bernstein mode
modied by the electron degeneracy as well as by the cou-
pling to the nearby spin induced modes, which have slightly
higher and lower frequencies as seen in Fig 1. For these
modes to be coupled as in Fig. 1, very strong magnetic elds
are needed, as can be found, e.g., in an astrophysical con-
text.
9
It should be noted that the mode structure close to
other resonances are similar to that shown in Fig. 1. For each
cyclotron or Bernstein resonance there is one spin resonance
with slightly higher frequency, and one spin resonance with
slightly lower frequency.
VI. SUMMARYAND DISCUSSION
In the present paper, we have calculated the electrostatic
dispersion relation in a plasma accounting for weakly relativis-
tic effects associated with the electron spin using a kinetic
model. In addition to the non-relativistic spin effects (spin-pre-
cession, spin magnetization currents and the magnetic dipole
force) the model includes the Thomas precession, the polariza-
tion currents associated with the spin and the spin-orbit cou-
pling. The linearized theory has been solved in a magnetized
plasma for a homogeneous background. Due to the electrostatic
approximation, the general dispersion relation can be deter-
mined from 4 of the conductivity tensor components. In con-
trast to the non-relativistic case, a prominent contribution
comes from the polarization currents associated with the spin.
The weakly relativistic terms are typically smaller than
the non-relativistic terms. However, for electrostatic pertur-
bations the non-relativistic spin terms vanish and the
(weakly) relativistic remains. These terms are typically pro-
portional to the parameters hx
p
=m
e
c
2
and/or l
B
B
0
=m
e
c
2
(when compared with the classical terms) which are very
small unless the electron number density and/or magnetic
eld is large. To some extent the effect of the spin terms can
be enhanced due to the presence of resonances, which par-
tially reduce the need for extremely large densities and
FIG. 1. Normalized frequency plotted against normalized wavenumber for
the three roots to Eq. (26). The parameter values are B
0
= 10
9
T and
n
0
= 7 10
34
m
3
.
032104-5 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
magnetic eld strengths. As investigated in Sec. IV, these
spin resonances give raise to several new wave modes. Due
to the closeness of some of the spin resonances to the classi-
cal Bernstein modes, we also nd that the classical modes
can be inuenced by the spin resonances to a certain extent.
Our results suggest that the applications of the general
theory should primarily be sought in astrophysical environ-
ments such as pulsar and magnetar environments, where the
magnetic eld may have values in the range
B
0
~ 10
6
10
10
T
9,23
and the density can be of the order
n
0
~ 10
33
10
36
m
3
(i.e., a thick accretion disc or a neu-
tron star crust). To some extent inertial connement fusion
plasmas might be of interest, although the non-relativistic
spin terms is likely to be of more interest in that regime. It
should also be stressed that for the strongest magnetic elds
(i.e., magnetars with B
0
~ 10
10
T) effects such as Landau
quantization
16
should be included, and there is also a need
for a fully relativistic treatment of the spin, rather than the
weakly relativistic description used here.
ACKNOWLEDGMENTS
A. Hussain is very grateful to G. Brodin for his constant
support and hospitality at Umea University Sweden, and to
M. Stefan and J. Zamanian for fruitful discussions. This
work was fully supported by the Higher Education
Commission of Pakistan, under the International Research
Support Initiative Program (IRSIP) Award No:1-
8/HEC/HRD/2011/1722.
APPENDIX A: DISTRIBUTION FUNCTION
COEFFICIENTS
The results for the coefcients in Eq. (23) are
A =
l
m
e
c
p
z
E
x
sin h
s
cos h
s
@
@h
s
_ _
am
e
x
ce
k
z
p
z
( )J
a
p
l
E
z
sin h
s
cos h
s
@
@h
s
_ _ _
m
e
x
ce
2aJ
/
a
2a
m
e
x
ce
k
l
p
l
J
a
a
2
m
e
x
ce
k
l
p
l
J
a
J
/
a
_ _
k
z
p
z
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _ _ __
@
@p
2

qlB
0
p
l
m
e
c
2
E
z
sin h
s
cos h
s
@
@h
s
_ _
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _
_ _
@
@p
2

l
hm
e
c
E
z
p
l
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _
E
x
p
z
J
a
_ _
@
@h
s
; (A1)
B = 4qJ
a
amx
ce
k
l
E
x
E
z
p
z
_ _
@
@p
2

2l
m
e
c
p
l
E
x
cos h
s
sin h
s
@
@h
s
_ _
2am
e
x
ce
J
/
a

m
e
x
ce
k
l
p
l
J
a
_ _
k
z
p
z
J
/
a
_ _
_ _
@
@p
2

2qlB
0
m
e
c
2
am
e
x
ce
k
l
J
a
E
x
cosh
s
sinh
s
@
@h
s
_ _
@
@p
2
; (A2)
and
C =
l
m
e
c
p
z
E
x
sinh
s
cos h
s
@
@h
s
_ _
am
e
x
ce
k
z
p
z
( )J
a
p
l
E
z
sin h
s
cos h
s
@
@h
s
_ _ _
m
e
x
ce
2aJ
/
a
2a
m
e
x
ce
k
l
p
l
J
a
a
2
m
e
x
ce
k
l
p
l
J
a
J
/
a
_ _
k
z
p
z
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _ _ __
@
@p
2

qlB
0
p
l
m
e
c
2
E
z
sinh
s
cosh
s
@
@h
s
_ _
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _
@
@p
2

l
hm
e
c
E
z
p
l
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _
E
x
p
z
J
a
_ _
@
@h
s
; (A3)
respectively.
APPENDIX B: CONDUCTIVITY TENSOR COMPONENTS
The derivation of the conductivity tensor components is described in Sec. IVB. The result for the free conductivity tensor
components are
r
freeE
xx
=
16p
2
qx
ce
k
l

a
_
dp
l
dp
z
p
l
aJ
a
i(x k
z
p
z
=m
e
ax
ce
)
2qa
m
e
x
ce
k
l
p
l
p
l
J
a

l
m
e
c
p
l
2am
e
x
ce
J
/
a
a
m
e
x
ce ( )
2
k
l
p
l
J
a
k
z
p
z
J
/
a
_ _
(n
0
n
0
)
n
0
_ _
@
^
f
0
(p
2
)
@p
2
; (B1)
032104-6 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
r
freeE
xz
=
32p
2
k
l

a
_
dp
l
dp
z
p
l
p
z
ax
ce
J
2
a
i(x k
z
p
z
=m
e
ax
ce
)
@
^
f
0
(p
2
)
@p
2
; (B2)
r
freeE
zx
=
16p
2
q
m
e

a
_
dp
l
dp
z
p
l
p
z
J
a
i(x k
z
p
z
=m
e
ax
ce
)
2qa
m
e
x
ce
k
l
J
a

l
m
e
c
p
l
2am
e
x
ce
J
/
a
a
(m
e
x
ce
)
2
k
l
p
l
J
a
k
z
p
z
J
/
a
_ _
(n
0
n
0
)
n
0
_ _
@
^
f
0
(p
2
)
@p
2
; (B3)
and
r
freeE
zz
=
32p
2
q
2
m
e

a
_
dp
l
dp
z
p
l
p
2
z
J
2
a
i(x k
z
p
z
=m
e
ax
ce
)
@
^
f
0
(p
2
)
@p
2
: (B4)
For notational brevity the above expressions contain only the electron contribution. The ion part is obtained trivially by drop-
ping the spin terms proportional to l (since the ion spin corrections are much smaller than the electron ones) and by replacing
the electron charge and mass by the ion values. In obtaining Eqs. (B1)(B4), we have performed all angular integrations.
Similarly the results for the polarization components are
r
polE
xx
=
4l
m
e
c
xp
2

a
_
dp
l
dp
z
p
l
il
m
e
c(x k
z
p
z
=m
e
ax
ce
x
cg
)
J
2
a
p
2
z
am
e
x
ce
k
z
p
z
( )
@
@p
2

p
2
z
h
(n
0
n
0
)
n
0
_ _
_

ilp
2
z
m
e
c(x k
z
p
z
=m
e
ax
ce
x
cg
)
J
2
a
am
e
x
ce
k
z
p
z
( )
@
@p
2

(n
0
n
0
)
hn
0
_ _

2i
x k
z
p
z
=m
e
ax
ce
2qa
m
e
x
ce
k
l
p
l
p
2
l
J
/
a
J
a
(n
0
n
0
)
n
0

lp
2
l
m
e
c
2am
e
x
ce
J
/2
a
a
(m
e
x
ce
)
2
k
l
p
l
J
/
a
J
a
k
z
p
z
J
/2
a
_ _
_ _
@
@p
2
_
^
f
0
(p
2
); (B5)
r
polE
xz
=
4lxp
2
m
e
c

a
_
dp
l
dp
z
_
i
(x k
z
p
z
=m
e
ax
ce
x
cg
)
lp
l
p
z
m
e
c
2am
e
x
ce
J
a
J
/
a
a
(m
e
x
ce
)
2
k
l
p
l
(1 a)J
2
a
k
z
p
z
J
a
J
/
a
a
m
e
x
ce
k
z
p
z
k
l
p
l
J
2
a
_ _
@
@p
2
J
a
J
/
a
a
m
e
x
ce
k
l
p
l
J
2
a
_ _
(n
0
n
0
)
hn
0
_ _

i
(x k
z
p
z
=m
e
ax
ce
x
cg
)
lp
l
p
z
m
e
c
2am
e
x
ce
J
a
J
/
a
a
(m
e
x
ce
)
2
k
l
p
l
_ _
(1 a)J
2
a
k
z
p
z
J
a
J
/
a
a
m
e
x
ce
k
z
p
z
k
l
p
l
J
2
a
_ _
@
@p
2
J
/
a
a
m
e
x
ce
k
l
p
l
J
a
_ _
(n
0
n
0
)
hn
0
_

4i
(x k
z
p
z
=m
e
ax
ce
)
qp
l
p
z
J
a
J
/
a
(n
0
n
0
)
n
0
@
@p
2
_
^
f
0
(p
2
)p
l
; (B6)
r
polE
zx
=
4l
2
xp
2
3m
2
e
c
2

a
_
dp
l
dp
z
p
2
l
p
z
i
(x k
z
p
z
=m
e
ax
ce
x
cg
)
_
a
2
(m
e
x
ce
)
2
k
l
p
l
J
2
a
k
z
p
z
a
m
e
x
c
k
l
p
l
J
2
a
am
e
x
ce
J
a
J
/
a
k
z
p
z
J
a
J
/
a
_ _
@
@p
2
J
a
J
/
a
a
m
e
x
c
k
l
p
l
J
2
a
_ _
(n
0
n
0
)
hn
0
_ _

i
(x k
z
p
z
=m ax
ce
x
cg
)
am
e
x
ce
J
a
J
/
a
k
z
p
z
J
a
J
/
a
a
2
(m
e
x
ce
)
2
k
l
p
l
J
2
a
k
z
p
z
a
m
e
x
c
k
l
p
l
J
2
a
_ _
@
@p
2
J
a
J
/
a
a
m
e
x
c
k
l
p
l
J
2
a
_ _
(n
0
n
0
)
hn
0
_ _ _ _
^
f
0
(p
2
);
(B7)
and
032104-7 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56
r
polE
zz
=
4l
2
xp
2
3m
2
e
c
2

a
_
dp
l
dp
z
p
3
l
_
i
(x k
z
p
z
=m
e
ax
ce
x
cg
)
__
2am
e
x
ce
J
/2
a
a(1 3a)
(m
e
x
ce
)
2
k
l
p
l
J
a
J
/2
a
(1 a)

(m
e
x
ce
)
3
(k
l
p
l
)
2
J
2
a
k
z
p
z
J
/2
a
2a
m
e
x
ce
k
z
p
z
k
l
p
l
J
a
J
/2
a
m
e
x
ce
k
z
p
z
k
l
p
l
_ _
2
J
2
a
_
@
@p
2

_
J
/2
a
2a
m
e
x
ce
k
l
p
l
J
a
J
/2
a
m
e
x
ce
k
l
p
l
_ _
2
J
2
a
_
(n
0
n
0
)
hn
0
_

i
(x k
z
p
z
=m
e
ax
ce
x
cg
)
__
2am
e
x
ce
J
/2
a
a(1 3a)
(m
e
x
ce
)
2
k
l
p
l
J
a
J
/2
a
(1 a)
(m
e
x
ce
)
3
(k
l
p
l
)
2
J
2
a
k
z
p
z
J
/2
a
2a
m
e
x
ce
k
z
p
z
k
l
p
l
J
a
J
/2
a
m
e
x
ce
k
z
p
z
k
l
p
l
_ _
2
J
2
a
_
@
@p
2

_
J
/2
a
2a
m
e
x
ce
k
l
p
l
J
a
J
/2
a
m
e
x
ce
k
l
p
l
_ _
2
J
2
a
_
(n
0
n
0
)
hn
0
_
_
^
f
0
(p
2
):
(B8)
1
G. Manfredi, How to model quantum plasmas, in Topics in Kinetic
Theory, Fields Institute Communication, edited by T. Passot, C. Sulem,
and P. L. Sulem (American Mathematical Society, Toronto, 2005), Vol.
46, pp. 263287.
2
S. H. Glenzer and R. Redmer, Rev. Mod. Phys. 81, 1625 (2009).
3
P. K. Shukla and B. Eliasson, Phys.-Usp. 53, 51 (2010).
4
P. K. Shukla and B. Eliasson, Rev. Mod. Phys. 83, 885 (2011).
5
G. Manfredi and P.-A. Hervieux, Appl. Phys. Lett. 91, 061108 (2007).
6
H. A. Atwater, Sci. Am. 296, 56 (2007).
7
M. Marklund, G. Brodin, L. Steno, and C. S. Liu, Europhys. Lett. 84,
17006 (2008).
8
S. A. Wolf, D. Awschalom, R. A. Buhrman, J. M. Daughton, S. von
Molnar, M. L. Roukes, A. Y. Chtchelkanova, and D. M. Treger, Science
294, 1488 (2001).
9
C. Kouveliotou, S. Dieters, T. Strohmayer, J. van Paradijs, G. J. Fishman,
C. A. Meegan, K. Hurley, J. Kommers, I. Smith, D. Frail, and T.
Murakami, Nature 393, 235 (1998); D. M. Palmer, S. Barthelmy, and N.
Gehrels, Nature 434, 1107 (2005); A. K. Harding and D. Lai, Rep. Prog.
Phys. 69, 2631 (2006).
10
M. P. Robinson, B. L. Tolra, M. W. Noel, T. F. Gallagher, and P. Pillet,
Phys. Rev. Lett. 85, 4466 (2000).
11
F. Haas, Phys. Plasmas 12, 062117 (2005).
12
J. Lundin, J. Zamanian, M. Marklund, and G. Brodin, Phys. Plasmas 14,
062112 (2007).
13
L. P. Kadanoff and G. Baym, Quantum Statistical Mechanics (Benjamin,
New York, 1962).
14
M. Bonitz, Quantum Kinetic Theory (B.G. Teubner, Stuttgart, Leipzig,
1998).
15
J. Zamanian, M. Stefan, M. Marklund, and G. Brodin, Phys. Plasmas 17,
102109 (2010).
16
J. Zamanian, M. Marklund, and G. Brodin, New J. Phys. 12, 043019
(2010).
17
F. A. Asenjo, J. Zamanian, M. Marklund, G. Brodin, and P. Johansson,
New J. Phys. 14, 073042 (2012).
18
J. D. Jackson, Classical Electrodynamics (Wiley, New York, 1975), chap.
11.8.
19
L. L. Foldy and S. A. Wouthuysen, Phys. Rev. 78, 29 (1950).
20
R. L. Stratonovich, Dokl. Akad. Nauk SSSR 1, 72 (1976); R. L.
Stratonovich, Sov. Phys.Dokl. 1, 414 (1976) (Engl. Transl.).
21
Equation (1) is valid for long wavelengths, which means scale lengths
much longer than the thermal de Broglie length and also much longer than
the Compton wavelength. The last condition is needed to drop the contri-
bution to Eq. (1) that would have originated from the Darwin term in the
Hamiltonian. At the same time we will consider the electrostatic approxi-
mation, which applies for sufciently short wavelengths. As seen from the
derivation in Sec. II, the conditions of sufciently short wavelengths and
sufciently long wavelengths can be fullled simultaneously.
22
J. Lundin and G. Brodin, Phys. Rev. E 82, 056407 (2010).
23
R. P. Drake, High-Energy-Density Physics (Springer-Verlag, 2006).
032104-8 Hussain, Stefan, and Brodin Phys. Plasmas 21, 032104 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
193.194.89.220 On: Sat, 08 Mar 2014 09:29:56

Vous aimerez peut-être aussi