Waves in Plasma

The notes given here, closely follow pgs. 79-154 of Chen, "Introduction to
Plasma Physics and Controlled Fusion", (Plenum, NY, 1985).

1 Introduction
Plasma is a complex fluid that supports many plasma wave modes. Restoring
forces include kinetic pressure and electric and magnetic forces. Wave phenom-
ena are important for heating plasmas, instabilities and diagnostics etc.
In vacuum, there is only one wave mode - the electromagnetic wave with
ω/k = c and having oscillating E and B components perpendicular to k. In
air, both sound waves and electromagnetic waves propagate. In plasma, both
electrostatic waves and electromagnetic waves will propagate. In the former
case, the electric field perturbation associated with the wave is parallel to the
wave propoagation direction E//k so that there are no magnetic perturbations
associated with the wave:

∇ × E = ik × E = iωB = 0
While in air, sound waves propagate through collisions, in a highy ionized
plasma, these collisions occur through the wave electric fields.
There are a great variety of possible plasma waves modes, since the wave
phase velocity depends on both the wave frequency and its angle of propaga-
tion with respect to the background magnetic field. Important characteristic
frequencies are ω pe , ω ce and ω ci .

2 Waves
Any sinusoidally oscillating quantity — say, the density n — can be represented
as follows:
n = n0 exp[i(k · r − ωt)]
where, in Cartesian coordinates,

k · r =kx x + ky y + kz z

Here n0 is a constant defining the amplitude of the wave, k is called the

propagation constant. If the wave propages in the x direction, k has only an x
component:
n = n̄ei(kx−ωt )
By convention, the exponential notation means that the real part of the
expression is to be taken as the measurable quantity. Let us choose n̄ to be real;
we shall see that this corresponds to a choice of the origins of x and t. The real
part of n is then

1
 Re(n) = n̄ cos(kx − ωt)
The exponential notation is useful for analysis of linear systems where Fourier
synthesis and superposition are valid.

2.1 Phase velocity

The phase velocity is the velocity on the wave of a point of constant phase.
Thus (d/dt)(kx − ωt) = 0 or

dx ω
= = υφ
dt t
is the phase velocity.
Consider now another oscillating quantity in the wave, say the electric field
E. Since we have already chosen the phase of n to be zero, we msut alow E to
have a different phase δ :

E = Ē cos(kx − ωt + δ) or E = Ēei(kx−ωt+δ)
where Ē is a real, constant vector.
It is customary to incorporate the phase information into Ē by allowing Ē
to be complex. We can write

E = Ēeiδ ei(kx−ωt) ≡ Ēc ei(kx−ωt)

Where Ēc is a complex amplitude.

Im(Ēc )
tan δ =
Re(Ēc )

From now on, we shall assume that all amplitudes are complex and drop the
subscript c. Any oscillating quantity g will be wriiten

g = g exp[i(k · r − ωt)]
so that g can stand for either the complex amplitude or the entire expression.
There can be no confusion, because in linear wave theory the same exponential
factor will occur on both sides of any equation and can be cancelled out.

2.2 Group velocity

Information is usually encoded on a carrier wave as either a modulation of its
phase or amplitude (or polarization). A simple amplitude modulated wave can
be constructed by combining two carriers of slightly different frequency ω-dω
and ω+dω. The resulting beat pattern (the information) travels at the group
velocity:

2
 Figure 1:

dω
υg = <c
dk
Let these waves be

E1 = E0 cos[(k + ∆k)x − (ω + ∆ω)t]

E2 = E0 cos[(k − ∆k)x − (ω − ∆ω)t]

E1 + E2 = 2E0 cos[(∆k)x − (∆ω)t] cos(kx − ωt)

This is a sinusoidally modulated wave. The envelope of the wave, given by
(∆k)x−(∆ω)t, is what carries information; it travels at velocity ∆ω/∆k. Taken
the limit ∆ω → 0, we thus define the group velocity.

The group velocity is closely related to the concept of Poynting flux. The
distinction between phase and group velocities is shown
schematically as follows:

3 Plasma Oscillations
If the electrons in plasma are displaced froma uniform background of ions, elec-
tric fields will be built in such a direction as to restore the neutrality of the
plasma by pulling the electrons back to their original positions. Becasue of
their inertia, the electrons will overshoot and oscillate around their equilibrium
positions with a characteristic frequency known as the plasma frequency.
Assumptions:

3
 1. There is no magnetic field
2. There are no thermal motions ( kT = 0)
3. the ions are fixed in space in a uniform distribution;
4. the plasma is infinite in extent
∂
5. the electron motions occur only in the x direction. (∇ = x̂ ∂x ; E = E x̂)

There is, therefore, no fluctating magnetic field; this is an electrostatic os-

cillation.
The electron equations of motion and continuity are

∂υ e
mne [ + (υ e · ∇)υ e ] = −ene E (1)
∂t
∂ne
+ ∇ · (ne υ e ) = 0
∂t
We use Poisson’s equation to find E:

0∇ · E = e(ni − ne )

Linearization: We separate the dependent variables into two parts: an

"equilibrium" part indicated by a subscript 0, and a small "perturbation" part
indicated by a subscript 1:

ne = n0 + n1 υ e = υ 0 + υ 1 E = E0 + E1

∂υ1
m[ ] = −eE1
∂t
∂n1
+ n0 ∇ · υ1 = 0
∂t
0 ∇ · E1 = eni

where we have already dropped high order perturbation terms and assumed
a uniform neutral plasma at equilibrium.
The oscillating quantities are assumed to behave sinusoidally:

υ1 = υ 1 ei(kx−ωt) x̂
n1 = n1 ei(kx−ωt)
E = Eei(kx−ωt) x̂

Use
∂
←→ −iω ∇ ←→ ik
∂t

4
 ⇒

−imωυ 1 = −eE1
−iωn1 = −n0 ikυ 1
ik 0 E1 = −en1

Eliminating n1 and E1 :

n0 e2
−imωυ 1 = −i υ1
0ω

If υ1 does not vanish, we must have

n0 e2 1/2
ω 2p = ( )
0m

— Plasma frequency
This frequency, denpending only on the plasma density, is one of the fun-
damental parameters of a plasma. Because of the smallness of m, the plasma
frequency is usually very high. For instance, in plasma of density n = 1018 m−3 ,
we have fp = ω p /2π ' 9 GHz.
In particlular, ω does not depend on k,
dω
υg = =0
dk
The disturbance does not propagate! They can be pictured as independent
oscillators.

4 Electron Plasma Waves

There is another effect that can cause plasma oscillations to propagate, and
that is thermal motion. Electrons streaming into adjacent layers of plasma with
their thermal velocities will carry information about what is happening in the
oscillating region. The plasma oscillation can the properly be called a plasma
wave.
We can easily treat this effect by adding a term -∇pe to the equation of
motion (equation 1).
For one-dimension, N = 1 (freedom), γ = (N + 2)/N = 3

∂n1
∇pe = 3kTe ∇ne = 3kTe ∇(n0 + n1 ) = 3kTe x̂
∂x
and the linearized equation of motion is
∂υ 1 ∂n1
mn0 = −en0 E1 − 3kTe
∂t ∂x
⇒ imωn0 υ1 = −en0 E1 − 3kTe ikn1

5
 Equations about E1 and n1 are the same as above, and we have
−e n0 ik
imωn0 υ1 = [en0 ( ) + 3kTe ik] υ1
ik 0 iω

n0 e2 3kTe 2
ω2 υ1 = ( + k )υ1
0m m
3
ω 2 = ω 2p + k 2 υ 2th
2
where υ 2th = 2kTe /m. The frequency now depend on k, and the group
velocity is finite:
dω 3k 2 3 υ2th
υg = = υ th =
dk 2ω 2 υφ
Note:

1. υg < c (why?)
2. At large k (small λ), information travels essentially at the thermal velocity.
At small k, infomation traves more slowly than υth even though υφ is
greather than υth . This is because the density gradient is small at large λ,
and thermal motions carry very little net momentum into adjacent layers.

Experiments to test the theory:

1. Looney and Brown, 1954

2. A more recent experiment: Barrett, Jones, and Frankling,

5 Sound Waves
As an introduction ion waves, let us briefly review the theory of sound waves in
ordinary air.
∂υ γp
ρ[ + (υ · ∇)υ] = −∇p = − ∇ρ
∂t ρ
∂ρ
+ ∇ · (ρυ) = 0
∂t
Linearizing about a stationary equilibrium with uniform p0 and ρ0 , we have

γp0
−iωρ0 υ 1 = − ikρ1
ρ0
−iωρ1 + ρ0 ik · υ 1 = 0

6
 For a plane wave with k = kx̂, υ = υx̂, we find, upon elimianting ρ1
γp0 ρ0 ikυ 1
−iωρ0 υ1 = − ik
ρ0 iω
γp0
ω 2 υ 1 = k2 υ1
ρ0
or
ω γp0 1/2 γkT 1/2
=( ) =( ) = cs
k ρ0 M
This is the expression for the velocity cs of sound waves in a neutral gas. The
waves are pressure waves propagationg from one layer to the next by collisions
among the air molecules. In a plasma with no neutrals and few collisions, an
analogous phenomenon occurs. This is called an ion acoustic wave, or, simply
an ion wave.

6 Ion Waves
Ions can still transmit vibrations to each other in the absence of collisions be-
cause of their charge. Acoustic waves can occur through the intermediary of
electric field. Since the motion of massive ions will be involved, these will
be low-frequency oscillations, and we can use the plasma approximation, i.e.
ni = ne = n insteady of Possion’s equation. The ion fluid equation in the
absence of a magnetic field is
∂υ i
M n[ + (υi · ∇)υi ] = enE − ∇p = −en∇φ − γ i kTi ∇n (2)
∂t
Linearizing and assuming plane waves, we have
−iωM nn0 υ i1 = −en0 ikφ1 − γ i kTi ikn1
As for the electrons, we may assume m = 0. The balance of forces on
electrons, therefore, requires
ne = n = n0 exp(eφ1 /kTe ) = n0 (1 + eφ1 /kTe + ...)
The perturbation in density of electrons, and therefore, of ions, is then
eφ1
n1 = n0
kTe
The linearized ion equation of continuity
iωn1 = n0 ikυ i1
Substituting for φ1 and n1 in terms of υ i1
kTe n0 ikυ i1
iωM n0 υi1 = (en0 ik + γ i kTi ik)
en0 iω

7
 kTe γ kTi
ω2 = k2 ( + i )
M M
ω kTe + γ i kTi 1/2
=( ) ≡ υs
k M
This is the dispersion relation for ion acoustiv waves; υs is the sound speed
in a plasma.
Note:

1. Since the ions suffer one-dimensional compressons in the plane waves,

γ i = 3.The elctrons move so fast relative to these waves that they have
time to equalize their temperature everywhere; therefore, the electrons are
isothermal, and γ e = 1.
2. υg = υ φ
3. When kTi → 0, ion waves still exist.

7 Validity of the Plasma Approximation

In deriving the velocity of ion waves, we used the neutrality conditions ni = ne
while allowing E to be finite. To see what error was engendered in the process,
we now allow ni to differ from ne and use the linearized Possion equation:
2
0∇ · E1 = 0k φ1 = e(ni1 − ne1 )

The electron density is given by the linearized Boltzmann relation:

eφ1
n1 = n0
kTe
Thus

2 n0 e2
0 φ1 (k + ) = eni1
0 kTe
2 2 2
0 φ1 (k λD + 1) = eni1 λD

The ion density is given by the linearized ion continuity equation:

k
ni1 =
n0 υ i1
ω
Inserting into the ion equation of motion (Equation 2)

en0 ik eλ2D k
iωM nn0 υi1 = ( + γ i kTi ik) n0 υi1
0 1 + k 2 λ2D ω

8
 2
k2 n0 e2 −1
0 λD
ω2 = ( 2 + γ i kTi ) (3)
M 1 + k 2 λD
ω kTe 1 γ kTi 1/2
=( 2 + i )
k M 1 + k λD
2 M
This is the same as we obtained previously except for the factor 1 + k 2 λ2D .
Our assumption ni = ne has given rise to an error of order k2 λ2D = (2πλD /λ)2 .
Since λD is very samll in most experiments, the plasma approximation is valid
for all except the shortest wavelength waves.

8 Comparision of Ion and Electron Waves

If we consider these short-wavelength waves by taking k 2 λ2D >> 1, Eq 3 becomes

n0 e2 n0 e2
ω2 = k2 2
= ≡ Ω2p
0M k 0M
We have, for simplicity, also taken the limit Ti → 0. Here Ωp is the ion
plasma frequency. For high frequencies the ion acoustic wave turns into a
constant-frequency wave. There is thus a complementary behavior between
electron plasma waves and ion acoustic waves:
electron plasma waves: basically constant frequency, but become constant
velocity at large k
ion acoustic waves: basically constant velocity, but become constant fre-
quency at large k.

9 Electrostatic Electron Oscillations Perpendic-

ular to B
Up to now, we have assumed B = 0,. When a magnetic field exists, may more
types of waves are possible. We shall examine only the simplest cases, start-
ing with high-frequency, electrostatic, electron oscillations propagating at right
angles to the magnetic field.
Some definitions:

Parallel / perpendicular : denote the direction of k relative to the undisturbed

magnetic field B0 .
Longitudianl / transverse: refer to the direction of of k relative to the oscil-
lating electric field E1.
electrostatic / electromagnetic: the oscillating magnetic field B1 = 0 / B1 6= 0

9
 The last two sets of terms are related by Maxwell’s equation

∇ × E1 = −Ḃ1

or
k × E1 = ωB1
If a wave is longitudianl, k × E1 vanishes, and the wave is also electrostatic.
If the wave is transevese, B1 is finite, and the wave is electromagnetic. It is of
course possibel for k to be at an arbitrary angle to B0 and E1 , then one would
have a mixture of the principal modes presented here.
Electron oscillations perpendicular to B0 .
Assumptions:

1. Ions fixed
2. kTe = 0

The motion of electrons in then governed by the following linearized equa-

tions:
∂υ e1
m = −e(E1 + υ e1 × B0 )
∂t
∂ne1
+ n0 ∇ · υ e1 = 0
∂t
0 ∇ · E1 = −ene1

We shall consider only longitudianl waves with k//E1 . Without loss of

generality, we can choose the x axis to lie along k and E1 , and the z axis lie
along B0 .
Dropping the subscripts 1 and e and separating the momentum equations
into components

−iωmυ x = −eE − eυ y B0
−iωυy = eυx B0
−iωmυ z = 0

⇒
ieB0
iωmυ x = eE + eB0 υx
mω
eE/imω
υx =
1 − ω 2c /ω 2

The linearized continuity equations:

k
n1 = n0 υx
ω

10
 Linearizing the Possion’s equation and using the last two results

k eE ω2
ik 0 E = −e n0 (1 − c2 )−1
ω imω ω
ω 2c ω 2p
(1 − )E = E
ω2 ω2
The dispersion relation is therefore

ω 2 = ω 2p + ω 2c ≡ ω 2h

The frequency ω h is called the upper hybrid frequency. Electrostatic electron

waves across B have this frequency, while whose along B are the usual plasma
oscillation with ω = ω p . The group velocity is again zero as long as thermal
motions are neglected.
Physical picture: There are two restoring forces acting on the electrons: the
electrostatic field and the Lorentz force. The increased restoring force makes
the frequency larger than that of a plasma oscillation.

10 Electrostatic Ion Waves Perpendicular to B

We next consider what happens to the ion acoustic wave when k is perpendicular
to B0 .
Assumptions:

1. n0 and B0 constant and uniform and v0 = E0 = 0.

2. Ti = 0
3. Electrostaic waves with k × E = 0, so that E = −∇φ
4. E = E1 x̂ and ∇ = ikx̂
5. let k be almost perpendicualr to B0 , allowing the eclectrons to preserve
charge neutrality by flowing along the B lines.

For the ion equation of motion, we have

∂υ i1
M = −e∇φ1 + eυ i1 × B0
∂t
Assuming plane waves propagating in the x direction and separating into
components, we have

−iωM υ ix = −eikφ1 + eυ iy B0
−iωM υ iy = −eυ ix B0

11
 Solving as before, we find

ek Ω2
υ ix = φ1 (1 − 2c )−1
Mω ω
where Ωc = eB0 /M is the ion cyclotron frequency. The ion equation of
continuity yields, as usual.
k
n1 = n0 υ ix
ω
Assuming the electrons can move along B0 ,we can use the Boltzmann rela-
tion for electrons.
ne1 eφ
= 1
n0 kTe
The plasma approximation ni = ne now closes the system of equations.

Ω2c ek kB Te n0 k
(1 − 2
)υ ix = υix
ω M ω en0 ω

kB Te
ω 2 = Ω2c + k 2
M

ω 2 = Ω2c + k 2 υ 2s
This is the dispersion relation for electrostatic ion cyclotron waves.

Physical picture: the ions undergo an acoustic-type oscillation, but the

Lorentz force constitutes a new restoring force giving rise to the Ω2c .

11 The Lower Hybrid Frequency

We now consider what happens when θ is exactly π/2, and the electrons are
not allowed to preserve charge neutrality by flowing along the lines of force.
Insteady of obeying Boltzmann’s relation, they will obey the full equation of
motion.
The ion equation of motion us unchanged:

ek Ω2
υ ix = φ1 (1 − 2c )−1
Mω ω
By changing e to -e, M to m, and Ωc to -ω c , we can write down the result
for electrons, with Te = 0

−ek ω2
υex = φ1 (1 − c2 )−1
mω ω
The equations of continuity give
k k
ni1 = n0 υ i1 ne1 = n0 υ e1
ω ω

12
 The plasma approximation ni = ne , then requires υi1 = υ e1

Ω2c ω2
M (1 − 2
) = −m(1 − c2 )
ω ω
1 1
ω 2 (M + m) = mω 2c + M Ω2c = e2 B 2 ( + )
m M

e2 B 2
ω2 = = ω c Ωc
Mm

ω = (ω c Ωc )1/2 ≡ ω l
This is called the lower hybrid frequency. If we had used Poisson’s equation
instead of the plasma approximation, we would have obtained:
1 1 1
2 = + 2
ωl ω c Ωc Ωp

In low-density plasmas the latter term actually dominates. The plasma

approximation is not valid at such frequencies. Lower hybrid oscillations can be
observed only is θ is very close to π/2.

13