WDS'14 Proceedings of Contributed Papers — Physics, 298–303, 2014.

ISBN 978-80-7378-276-4 © MATFYZPRESS

Plasma Dispersion Relation and Instabilities in Electron

Velocity Distribution Function
A. Tomori
Charles University, Faculty of Mathematics and Physics, Prague, Czech Republic.

Abstract. Properties of possible electrostatic and electromagnetic wave propa-

gation in hot space plasmas can be described by plasma dispersion relation. This
can be obtained using plasma kinetic theory and Maxwell’s equations. Analyti-
cal solutions exist for some special cases, however general solution needs use of
numerical methods, with the input including velocity distribution function. Non-
Maxwellian electron velocity distribution functions can lead to wave generation and
amplification. Examples of such instabilities leading to various types of waves are
described.

Introduction
In this overview paper we briefly describe methods of derivation and calculation of the
dispersion relation for electromagnetic waves in a hot collisionless magnetized plasma.
We start with the dispersion relation of electrostatic waves in a hot unmagnetized plasma
(or in a magnetized plasma with parallel propagation). Then we continue with a general case
of the dispersion relation of electromagnetic waves in a hot plasma with an ambient magnetic
field present. Also instabilities in electron velocity distribution function are discussed.
In the last part of the paper we discuss some applications, which include analysis of wave-
particle interactions, mechanisms of wave generation and transfer of energy between different
electron populations.
The intent of the paper is to be a summary of information the author came across while
writing a code of a numeric dispersion relation solver, hence the new findings can be expected
in subsequent papers.

Dispersion relation
Dispersion relation provides a relationship between the wave vector and the frequency of a
wave and describes under which conditions the wave can propagate and under which conditions it
cannot propagate. The dispersion relation can usually be obtained as a condition for non-trivial
solutions of a homogeneous set of equations which describe given waves, and it is usually written
in the form D(k, ω) = 0. Equations describing a wave are usually in a form of a homogeneous
set of partial differential equations, but after applying an integral transform (Fourier or Laplace)
a set of algebraic equations has to be solved. Roots of the dispersion relation correspond to the
actual modes of wave propagation.
The dispersion relation contains information not only on the wavelength and the frequency,
but also on the direction of wave propagation. As the frequency is considered complex, the imag-
inary part of frequency =ω = γ describes the growth or damping rate of the wave, depending
on the sign. For a given wave vector, generally more solutions of a complex frequency can exist.
Phase and group velocities can be obtained from the dispersion relation.
The dispersion relation depends on the properties of a plasma, namely on phase space
distribution functions of plasma particles, properties of plasma particles (mass and charge) and
electric and magnetic field.
In order to be able to derive the dispersion relation for waves in a plasma, some assumptions
are made. Hot homogeneous collisionless plasma in a magnetic field is assumed. The dispersion
relation is derived using kinetic theory and linearization. It means that the starting point is a
homogeneous (collisionless) version of the Boltzmann equation (also called the Vlasov equation)

298
 TOMORI: PLASMA DISPERSION RELATION

10

Langmuir

9 Bernstein
R
X
O
f / fce
8
L
X

6
10-4 10-3 10-2 10-1 1
k (m-1)

Figure 1. Example of the dispersion relation for two directions of wave propagation – parallel
propagation (solid lines) and perpendicular propagation (dotted lines). Parallel to a magnetic
field propagate electromagnetic R and L waves (the second R mode is not visible on the picture)
and an electrostatic Langmuir mode wave. Perpendicular to a magnetic field propagate two ex-
traordinary modes X, one ordinary mode O and electrostatic Bernstein modes for high values
of k. Only real part of frequency is shown. The distribution function consists of three electron
populations – a core population (n = 47.24 cm−3 , vth = 10 eV, T⊥ /Tk = 0.08), a warm popula-
tion (n = 0.82 cm−3 , vth = 1180 eV, T⊥ /Tk = 0.98) and a warm electron population with a loss
cone (n = 0.41 cm−3 , vth = 1890 eV, T⊥ /Tk = 1.16 and the loss-cone depth parameter β defined
in Rönnmark [1982] β = 0.2). Electron cyclotron and plasma frequencies are fce = 7 875 Hz and
fpe = 62 505 Hz. Parameters taken from Grimald and Santolı́k [2010].

with the set of Maxwell equations. Linearization means that varying quantities are expressed
as a sum of a mean value and its first order perturbation. Solutions of waves are expected in a
form of exp [i (k · r − ωt)] (planar wave). As a consequence of a linearization, solutions are valid
only during a linear phase until non-linear processes take over (a wave cannot grow indefinitely).
Example of the dispersion relation for a plasma consisting of several electron populations
with the distribution function exhibiting the loss cone anisotropy for ωpe > ωce can be seen in
Fig. 1.

Electrostatic waves
The starting point is the Vlasov equation
dfs (r, v, t) ∂fs qs
= + v · ∇fs + (E + v × B) · ∇v fs = 0, (1)
dt ∂t ms
where fs is the phase space distribution function of species s. By linearizing it (fs (v) =
f0 (v) + f1 (v), f1 is the first-order perturbation), B = 0, with electric field being completely
described by potential E = −∇Φ and by performing a Fourier transform in both time and
space, which brings us from spatial and temporal variables r and t to k and ω, yields the
dispersion relation for electrostatic waves in a hot plasma without a magnetic field [Gurnett
and Bhattacharjee, 2005]
X ωps2 Z ∞
∂Fs0 /∂vk
D(k, ω) = 1 − 2
dvk = 0. (2)
s
k −∞ vk − ω/k

ωps denotes the plasma frequency of species s with an electric charge qs , density ns and a
mass RR 2 = n q 2 /(ε m ). We introduced the reduced distribution function F (v ) =
ms , ωps s s 0 s s0 k
∞ R
1/n0 −∞ f0 (v) dvx dvy , such that Fs0 dvk = 1.

299
 TOMORI: PLASMA DISPERSION RELATION

This problem was solved by Vlasov [1938]. However, it has a problem when a resonance
between the wave and particles moving at the phase speed of the wave occurs. It does not give
us any advice on how to deal with the integral containing a pole. Therefore this equation can
be solved only for the distribution function with no particles moving at the phase velocity of
the wave. This is the reason this dispersion relation cannot lead to resonant instabilities.
Landau [1946] managed to solve the problem shortly after Vlasov. He uses Fourier transform
in space, but Laplace transform in time. The dispersion relation looks the same as in the Vlasov
case (2), but complex frequency ω is replaced by ip. It is not the final form, though. That form
is valid only for the positive half of a complex plane p, <p > 0, and it needs to be analytically
continued to the left half plane. The discontinuity when real part of p equals zero (<p = γ = 0)
is related to the problem in the Vlasov approach.
The analytical continuation is performed by distorting the integration contour (originally
running along the vk axis from −∞ to ∞) such that it always passes below the pole (see Fig. 2a).
We integrate along the parallel velocity axis vk , but we have to extend it to complex velocity
so we can make a turn around the pole. Therefore in case of γ < 0 we have to calculate
a residue. This involves evaluating the distribution function at a complex parallel speed. It
is straightforward for analytically defined distribution function like Maxwellian, but not so
obvious for the distribution function defined by data measured by a particle analyser on-board
a spacecraft.
Having done the analytical continuation, the dispersion relation for electrostatic waves in
a hot plasma without a magnetic field in case of <p = γ > 0 has the form from the Vlasov
approach (2), but in case of <p = γ < 0 it is
X ωps 2 Z ∞ 2
∂Fs0 /∂vk k ωps ∂Fs0
D(k, p) = 1 − 2
dv k − 2πi 2
= 0, (3)
s
k −∞ vk − ip/k |k| k ∂vk v =v = ip
Res
k k

and it can be written more generally as

2 Z
X ωps ∂Fs0 /∂vk
D(k, p) = 1 − 2
dvk = 0, (4)
s
k C vk − ip/k

given that the contour C runs along the vk axis and always stays below the pole.
The complex residue contributes to the imaginary part of the dispersion relation =D and
was lacking in the Vlasov approach (the real part of the dispersion relation <D in the first order
approximation equals the dispersion relation obtained in the Vlasov approach (2)). Shortly we
will see that the imaginary part of the dispersion relation =D is responsible for resonant growth
(damping).
When assuming the small growth rate |γ|  |ω| it is possible to derive a formula for the
growth rate for a given wave. Using the Taylor expansion of D(k, p) around p0 = <p = −iω
and stating D = 0 leads to the expression for the growth rate
−=D
γ= , (5)
∂<D/∂ω
which, after expressing =D, leads to
2
ωps
k ∂Fs0
γ=π . (6)
|k| k 2 ∂<D/∂ω ∂vk vk =ω/k

The growth rate is proportional to the slope of the reduced distribution function evaluated
at the phase velocity of the wave. Specifically for Langmuir waves it can be shown that the
∂<D/∂ω term is always positive and for velocity distributions with a single maximum centered
at zero the growth rate is always negative (because of the negative slope, see Fig. 2b). This is
the Landau damping. The energy of the wave is transfered into the energy of particles. This

300
 TOMORI: PLASMA DISPERSION RELATION

(a) (b)

Figure 2. (a) Distorted integration contour in a complex velocity plane so it always passes
below the pole at vk = ip/k (diagram for k > 0; for k < 0 it would be made to pass above
the pole). (b) Assuming the weak growth rate approximation (|γ|  |ω|), the growth rate γ is
proportional to the slope of the reduced distribution function evaluated at the phase velocity
vk = ω/k of the wave. Images from Gurnett and Bhattacharjee [2005], vz corresponds to vk .

discovery was originally a bit surprising for a collisionless plasma. Contrary to equations for a
plasma with collisions, the Vlasov and Poisson’s equations are reversible in time, but lead to
irreversible processes [Mouhot and Villani, 2010].
For a plasma with Maxwellian distribution functions
"  2 #
1 vk
Fs0 (vk ) = √ exp − − vds (7)
πvths vths
q
with vths = 2kmBsTs being thermal velocity of species s and vds being drift velocity normalized
by vths , the dispersion relation (4) is often expressed using the so called plasma dispersion
function 2
e−z
Z
1
Z(ζ) = √ dz, (8)
π C z−ζ
where the integral is calculated along the contour C defined the same way as above. The
dispersion relation (4) can then be written as
X 1
D(k, p) = 1 + [1 + ζs Z(ζs )] = 0, (9)
s
(kλD s)2
q
1 0 kB Ts
where ζs = vths ip/k and λDs is the Debye length λDs = ns qs2
. The advantage of introducing
the plasma dispersion function Z(ζ) is also that it is already implemented in many numerical
√
packages (often by the so-called Faddeeva function w(ζ), Z(ζ) = i πw(ζ)) and one does not
need to deal with complex integration along a contour.

Electromagnetic waves
The wave equation of electromagnetic waves

ω2
k × (k × E) + K · E = 0, (10)
c2
σ
where K is the dielectric tensor K = 1 − iω 0
and σ the conductivity tensor, can be expressed as
a dot product of the dispersion tensor and the electric field, D · E = 0. The dispersion relation
is defined as a requirement for non-trivial solutions of E, which means that the determinant of
the dielectric tensor D equals zero, detD = 0. Calculating the actual modes of the dispersion
relation involves finding complex roots of a complex function detD.

301
 TOMORI: PLASMA DISPERSION RELATION

v|| (107m/s) 5

−5

−10
−10 −5 0 5
v⊥ (107m/s)

−25 −20 −15

log f (m−6s3)

(a) (b)

Figure 3. Example of the distribution function consisting of three electron populations (a core
population of cold electrons and two populations of hot electrons with bi-Maxwellian distribu-
tions) exhibiting the loss-cone anisotropy (the third electron population), color mapped graph
(a) and as a 3D graph (b). v⊥ and vk axes in panel (b) span across the same range as in (a). The
electron distribution function corresponds to electrons trapped in the Earth’s magnetosphere.

In order to compute the dielectric tensor the linearized Vlasov (zero- and first-order) and
Poisson’s equations are taken with new terms containing magnetic field [Gurnett and Bhat-
tacharjee, 2005]. The dielectric tensor can be expressed as [Rönnmark, 1982; Treumann and
Baumjohann, 1997b]
∞ 2 Z ∞Z ∞ kk ∂Fs0 + nΩs ∂Fs0
!
X ωps 2 ωps
X X ∂vk v⊥ ∂v⊥
K(k, ω) = 1 − 2
I− 2
Sns 2πv⊥ dv⊥ dvk , (11)
s
ω s n=−∞
ω 0 −∞ kk vk + nΩs − ω
 2 2 2

n v⊥ 2 nv⊥ 0 nv⊥ vk 2
J i J n J J
 βs2 n
2 βs n βs n

Sns (v⊥ , vk ) =  nv⊥
, (12)
0 2 0 2 −iv⊥ vk Jn Jn0 
−i Jn Jn v Jn β ⊥
nv⊥svk 2
βs Jn iv⊥ vk Jn Jn0 vk2 Jn2
where βs = k⊥ v⊥ /Ωs and Ωs is the gyrofrequency of species s defined by Ωs = qs B/ms . Jn and
Jn0 denotes Bessel function of the first kind and its derivative respectively and is a function of
βs . From the computational point of view calculating the dielectric tensor is the most difficult
part. It simplifies for some special conditions, specifically for parallel propagation only one
component of Sns tensor has to be computed for the electrostatic mode and two components for
the electromagnetic R and L modes with the other components being zero. For general oblique
wave propagation all components of Sns tensor have to be computed.

Wave instabilities
Considering the physical mechanism involved, instabilities can be divided into reactive
(non-resonant or fluid type) and resonant or kinetic instabilities.
Reactive instability is a bunching instability [Cairns and Fung, 1988], where growing
waves create bunches of electrons in space, which enhance electric field. It leads to quasi-
monochromatic, very narrow band waves.
Kinetic instability is a resonant instability and involves interaction of the wave with par-
ticles, which move at speeds close to the wave phase velocity. It may lead to narrow or broad
band waves. Examples include inverse Landau or cyclotron damping.

302
 TOMORI: PLASMA DISPERSION RELATION

Unstable (resonant) electron distribution function must have more than one hump [Gur-
nett and Bhattacharjee, 2005]. This usually means at least two electron (or ion) components
(e.g. core and a beam population) [Treumann and Baumjohann, 1997a] or for example the
distribution function anisotropy (see Fig. 3 for the loss-cone anisotropy). It is valid for both
electrostatic and electromagnetic waves with the difference that in case of electrostatic waves
only the electric field is involved in resonant interaction, but in the case of electromagnetic
waves both the electric and magnetic fields are involved. Sometimes the magnetic field plays
the dominant role [Gurnett and Bhattacharjee, 2005].

Conclusion
Knowledge of the dispersion relation is crucial in analysis of wave-particle interactions,
mechanisms of wave generation and propagation, transfer of energy between electron populations
etc. Analysis of conditions of reactive and kinetic instabilities driven by streaming electrons,
which are responsible for narrow and wide band electrostatic emissions above and below plasma
frequency present in the electron foreshock region using the dispersion relation is discussed by
Fuselier et al. [1985], Gary [1985], and Cairns and Fung [1988].
Examples of a more recent research utilising the dispersion relation include study of acceler-
ation of high energy electrons in the outer Van Allen radiation belt by electromagnetic whistler
mode chorus. That, in turn, is generated by another electron population [Santolı́k et al., 2010].
Another example is an analysis of generation of whistler mode waves propagating along the
Saturn’s moon Rhea flux tube by the loss-cone anisotropy [Santolı́k et al., 2011]. Knowing the
approximate location of the source region of observed emissions and calculated value of max-
imum possible growth rate given by observed electron distribution function together with the
wave group velocity (also obtained from the dispersion relation) can give answers on generation
and propagation of waves, which are of interest.

Acknowledgments. The author acknowledge additional support from grant GAUK 678212.

References
Cairns, I. H. and Fung, S. F., Growth of electron plasma waves above and below f(p) in the electron
foreshock, J. Geophys. Res., 93 , 7307–7317, 1988.
Fuselier, S. A., Gurnett, D. A., and Fitzenreiter, R. J., The downshift of electron plasma oscillations in
the electron foreshock region, J. Geophys. Res., 90 , 3935–3946, 1985.
Gary, S. P., Electrostatic instabilities in plasmas with two electron components, J. Geophys. Res., 90 ,
8213–8221, 1985.
Grimald, S. and Santolı́k, O., Possible wave modes of wideband nonthermal continuum radiation in its
source region, J. Geophys. Res., 115 , 6209, 2010.
Gurnett, D. and Bhattacharjee, A., Introduction to plasma physics: with space and laboratory applica-
tions, Cambridge University Press, 2005.
Landau, L., On the vibration of the electronic plasma, J. Phys. USSR, 10 , 25, 1946.
Mouhot, C. and Villani, C., Landau damping, J. Math. Phys., 51 , 015 204–+, 2010.
Rönnmark, K., Waves in homogeneous, anisotropic multicomponent plasmas (WHAMP), Tech. rep.,
1982.
Santolı́k, O., Gurnett, D. A., Pickett, J. S., Grimald, S., Décreau, P. M. E., Parrot, M., Cornilleau-
Wehrlin, N., El-Lemdani Mazouz, F., Schriver, D., Meredith, N. P., and Fazakerley, A., Wave-particle
interactions in the equatorial source region of whistler-mode emissions, J. Geophys. Res., 115 , 0, 2010.
Santolı́k, O., Gurnett, D. A., Jones, G. H., Schippers, P., Crary, F. J., Leisner, J. S., Hospodarsky, G. B.,
Kurth, W. S., Russell, C. T., and Dougherty, M. K., Intense plasma wave emissions associated with
Saturn’s moon Rhea, Geophys. Res. Lett., 38 , 19 204, 2011.
Treumann, R. and Baumjohann, W., Advanced space plasma physics, Imperial College Press, 1997a.
Treumann, R. and Baumjohann, W., Basic Space Plasma Physics, Imperial College Press, 1997b.
Vlasov, A. A., On the oscillation properties of an electron gas, Zh. Èksper. Teoret. Fiz., 8 , 291–318,
1938.

303