University of Warwick 2020-21

PX392 Plasma Electrodynamics

Professor Valery M Nakariakov

Contents
1 Foundations of Plasma Physics 1
1.1 What is a Plasma? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Electron Plasma Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Debye Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Three Plasma Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Gyration and Gyrofrequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Drift Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
 1.6.1 Guiding centre drift due to non-magnetic forces . . . . . . . . . . . . . . 10
1.6.2 Drift in a perpendicular electric field . . . . . . . . . . . . . . . . . . . . 11
1.6.3 Drift in a perpendicular gravitational field . . . . . . . . . . . . . . . . . 11
1.6.4 ∇B0 drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6.5 Magnetic curvature drift . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Thermal effects in electron plasma oscillations . . . . . . . . . . . . . . . . . . . 13

2 Dielectric Description of Magnetised Plasmas 14

2.1 EM Wave Equation in a Plasma with B0 6= 0 . . . . . . . . . . . . . . . . . . . 14
2.2 Microscopic description of EM wave interaction with a magnetised plasma . . . 16
2.3 Matrix representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Elements of the dielectric tensor ˆ . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Dispersion Relations for High frequency EM Waves in a Cold Plasma 18

3.1 The dispersion relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 EM Waves in a Plasma without B0 . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Parallel, High Frequency Waves in Cold Magnetised Plasmas . . . . . . . . . . 20
3.4 Whistler Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Faraday Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.6 Plasma Diagnostics Using Faraday Rotation . . . . . . . . . . . . . . . . . . . . 23
3.7 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.8 Perpendicular, High Frequency Waves in Cold Magnetised Plasmas . . . . . . . 24
3.9 Effects of finite temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4 Elements of Plasma Kinetics 28

4.1 The Vlasov Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Landau Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Bump-on-tail instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5 Magnetohydrodynamics (MHD) 34
5.1 MHD Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.2 MHD Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3 MHD Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.3.1 Alfvén waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.3.2 Magnetoacoustic waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.4 MHD Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

1 Foundations of Plasma Physics

1.1 What is a Plasma?

Plasma is the 4th state of matter.

Irving Langmuir, 1923: “Plasma is an ionised gas, or a gas of electrically charged particles”.
Usually, with the increase in the internal energy, the matter goes through the following states:
solid, liquid, gaseous and plasma. The plasma is created by ionisation of a gas (while there
more exotic kinds of a plasma, e.g. the electron “gas” in conductors and semi-conductors can
also be described as a plasma).
The most common state of the Universe: 99.999% of the visible matter in the Universe is a
plasma. Examples of plasmas:

• Earth’s ionosphere: the part of the Earth atmosphere, from about 50 km to about
1000 km (yes, the International Space Station is in the ionosphere), consisting mainly of
e− , O+ , NO+ , O+
2 , and neutral atoms of nitrogen and oxygen.

• Earth’s magnetosphere: the outer part of the Earth atmosphere, from the ionosphere to
about 15 Earth’s radii on the dayside, and an extended magnetotail (up to 200 Earth
radii) on the nightside. The plasma is mainly hydrogen, consisting of e− and p+ .
• Solar and stellar coronae: the outer part of the atmosphere of the Sun, mainly hydrogen,
consisting of e− and p+ .
• Solar and stellar winds, also mainly hydrogen; interstellar and intergalactic plasmas.
• Various laboratory devices, e.g. used for controlled fusion. For example, the magnetically
confined plasmas in tokamaks, stellarators, z-pinch machines etc., and the laser confined
plasmas.
• Plasmas in pulsar atmospheres: e− and p+ (“pair plasmas”).
• Dusty plasmas: e− and p+ , and positively and negatively charged dust grains.

If we apply an electric field to a plasma, electrons and protons (and other ions) would move in
the opposite directions → the electric charge separation → internal electric field that affects
the charged particle motion (other charged particles of the plasma “feel” it too) that affects the
charged particle’s motion → moving charged particles constitute electric current that creates
a magnetic field affecting the charged particles by the Lorentz force → complicated collective
dynamics as the main feature of a plasma.
Main physical characteristics of plasma:

• Concentration: the number of particles of a certain type in unit volume. For example,
ne is the concentration of free electrons (which are not parts of neutral atoms and ions),
ni is the concentration of ions of a certain type.

1
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

• Mass density: the mass per unit volume, e.g. in a hydrogen plasma, ρ = mp np .

• Degree of ionisation: α ≡ nions /(nions + nneutrals ). In particular, a gas with α = 1% (e.g.

the lower ionosphere) behaves, sometimes, as a plasma.

• Temperature: the measure of the internal energy. Different species (free electrons, various
ions, neutrals) may have different temperatures. But, the plasma should be in thermal
equilibrium.

An important property of a plasma is quasi-neutrality: the overall electric charge is 0, while

locally, at small scales, there may be a charge separation and hence a local electric field.

Plasma concentration range is more than 30 orders of magnitude; plasma temperature range
is more than 10 orders of magnitude. (1 eV ≈ 11,600 K).

1.2 Electron Plasma Oscillations

Assume that the plasma is neutral (all electrons are situated near the positive ions) and neglect
the thermal motion. If we somehow displace some electrons from the equilibrium, it will cause
the charge separation and generate the electric field. Let us restrict our attention to the hydro-
gen plasma for simplicity. As electrons are much lighter than protons, the electrons respond
to the electric field generated by the charge separation much more rapidly than the inertial
protons. Hence, the motion of protons can be neglected. This field will provide the electrons
with the force that will return them to the equilibrium. But, when the electrons reach the equi-
librium, they will have some kinetic energy because of their motion in the electric field. Thus,
they will overshoot the equilibrium, and will move further, generating the electric field, until
they are stopped by the field, and dragged by it back to the equilibrium. Hence the electrons
will oscillate around the equilibrium: the phenomenon of electron plasma oscillations (also
known as the Langmuir oscillations, electrostatic oscillations, etc.). This behaviour resembles
the oscillations of a pendulum with the periodic conversion of the gravitational potential energy
to the kinetic energy and back.

2
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Electron plasma oscillations is one of the intrinsic plasma phenomena. What is the frequency
of the oscillations?
Consider a 1D, along the x axis motions:

Let us displace a vertical slab of width x of electrons in the x-direction. The electric charge in
this slab is Q = −e × (the number of electrons in the slab) = −ene xL, where ne is the electron
concentration. Here we assumed that the length of the slab in the direction perpendicular to
the sketch is 1.
Determine the electric field E generated by the charge separation, using Gauss’ theorem. The
electric field through a closed surface is determined by the electric charge confined within this
surface S (shown by the green dashed line in the sketch),
Z
E.ds = Q/0 . (1)
S

According to the sketch above, the electric field goes only through the left side of the surface,
of the length L, and is everywhere normal to it. Also, by the symmetry of the problem, the
field is constant. Thus, we obtain Z
E.ds = −EL, (2)
S
where E is the x-component of the electric field vector. Therefore, we have
ene xL ne e
− EL = − →E= x. (3)
0 0
The equation of motions for the electrons in the slab is
ne e
me ẍ = −eE = −e x, (4)
0
or
2
ẍ + ωpe x = 0, (5)

3
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

where
s
ne e 2
ωpe ≡ . (6)
me 0

It is the harmonic oscillator equation, with the general solution

x(t) = A0 sin(ωpe t + φ0 ), (7)

where A0 and φ0 are constants. Thus, the displaced and released electron slab will oscillate har-
monically with the frequency ωpe that is called the cyclic electron plasma frequency (sometimes
simply “plasma frequency”), corresponding to the electron plasma frequency fpe = ωpe /2π.
The electron plasma frequency is determined by the concentration of free electrons in the
1/2 p
plasma (as ne ). It can be estimated as fpe (Hz) ≈ 9 ne (m−3 ).
Typical values of the electron plasma frequency:

Plasma Electron Plasma

system concentration Frequency
Ionosphere 1011 m−3 3 MHz
Magnetosphere 109 m−3 300 kHz
Solar corona 1015 m−3 300 MHz
Solar wind at 1 AU 107 m−3 30 kHz
Tokamak 1020 m−3 90 THz
Interstellar 106 m−3 10 kHz
Inertial confinement 1028 m−3 1015 Hz
Metals 1029 m−3 3 × 1015 Hz

Plasma oscillations manifest the collective nature of a plasma. The plasma particles are “con-
nected” one to another via self-consistent electromagnetic forces. This collective behaviour
makes the plasma different from gases.

1.3 Debye Shielding

Insert a point positive charge into a plasma. It would attract electrons and repel positive ions,
which will form a cloud around the test charge.

4
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

These charged particles will create an electric field that will counteract the field of the inserted
point charge. Thus, at some distance, the test point charge becomes shielded from the rest of
the plasma. It is the effect of Debye shielding (or screening). It is another manifestation of the
collective behaviour of a plasma. What is the distance of the Debye shielding from the test
charge?
Recall that the electrons and ions have some temperatures Te and Ti , respectively (are in
thermal motion). By the way, in a plasma the electron and ion temperatures are often different
from each other. Some of the electrons can escape from the electron cloud around the test
point charge. Thus, the edge of the cloud is determined by the balance of the potential energy
of the electrons in the electric field and the thermal energy of electrons.
Poisson’s equation:
∇ · E = e(ni − ne )/0 . (8)
Here we assumed that the ions are one-time ionised, i.e. their charge is e. In the static case,
the electric field can be expressed via the electrostatic potential φ, as E = −∇φ. Consider the
field in the x direction,
d2 φ e(ni − ne )
2
=− . (9)
dx 0
In the presence of a potential field φ, the potential energy of a particle with the charge q is
W = qφ, and the particles are distributed according to the Maxwell–Boltzmann law, as
 qφ 
n(W ) = n0 exp − , (10)
kB T
where kB is the Boltzmann constant and T is the temperature of the particles.
Thus, for the electrons and ions in the cloud we have
 eφ 
ne (W ) = n0 exp + , (11)
kB Te
 eφ 
ni (W ) = n0 exp − .
kB Ti
Here we used the quasi-neutrality condition.

5
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

As the cloud of the charged particles “shields” the electric field of the test charge from the rest
of the plasma, at the edge of the cloud eφ  kB T . Hence, in Eqs. (11) we can use the Taylor
expansion, exp y ≈ 1 + y for |y|  1,

n0 e 2  1 1
e(ni − ne ) ≈ − + φ. (12)
kB Ti Te

Substituting it in Poisson’s equation, we obtain

d2 φ n0 e2  1 1
= + φ = λ−2
D φ, (13)
dx2 kB 0 Ti Te

with the solution  x 

φ ∝ exp − . (14)
λD

The quantity λD has the dimension of length, and is the size of the cloud around the test
charge. It is the measure of the effect of the test charge on the surrounding plasma, called the
Debye length (or “Debye” radius),

n0 e 2  1 1
λ−2
D ≡ + , (15)
kB 0 Ti Te

and it is determined by the plasma concentration and temperature.

If we neglect the motion of heavy, inertial ions,
r
1 kB Te
λD ≈ . (16)
ωpe me

It can be shown that the electric potential around the sample charge is almost not influenced
by the plasma particles at the distances x  λD , and is the potential of the test charge in
vacuum, while at x > λD the potential decreases exponentially, that is, faster than any power.
Thus, while the electric force is long range in free space, in a plasma it only extends to Debye
length.
In other words, the Debye shielding effect can be defined as the exponential decrease of the
electric potential around a sample charge in a plasma, that is, faster than in the vacuum. The
Debye length determines the distance at which the electric field of the charge in the plasma
decreases by the factor of e ≈ 2.718.
It can be estimated as p
λD [m] ≈ 70 Te [K]/ne [m−3 ]. (17)

In a 3D problem, the shielding cloud has a spherically symmetric shape, and is called the Debye
sphere.
Plasmas do not contain strong electric fields as they reorganise to shield from them.

6
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

1.4 Three Plasma Criteria

In nature, gases always have some populations of ions and free electrons, appearing, e.g.,
because of photoionisation. But, not every gas is a plasma. To be a plasma, it should satisfy
three criteria:

• The characteristic size of the volume occupied by the gas should be  λD .

• The number of free electrons in the Debye sphere, the quantity known as the plasma
parameter, ne 4πλ3D /3, should be  1.

• For the free electrons to be unaffected by the collisions with neutrals, the mean time be-
tween electron-neutral collisions should be  2π/ωpe = the period of plasma oscillations.

1.5 Gyration and Gyrofrequency

Plasmas are often penetrated by external magnetic fields. Consider motion of an electron in a
given magnetic field B0 . The equation of motion is
−e
V̇ = V × B0 . (18)
me

In the following we restrict ourself to the non-relativistic case.

1) If there is no the external field, V̇ = 0, hence V is constant.
2) Case B0 6= 0.
Consider the case of a constant magnetic field B0 .
Calculate the scalar product of the equation of motion and me V:
−e
me V.V̇ = me V.V × B0 . (19)
me
d 
me V 2 = 0, mV 2 = const,
dt
thus the kinetic energy is constant. In the external magnetic field the electron does not gain
energy.
Calculate the scalar product of the equation of motion and B0 :
−e
B0 .V̇ = B0 .V × B0 . (20)
me
B0 .V̇ = 0, → V̇k = 0 → Vk = const,

where the symbol k shows the direction along B0 . Thus the electron’s speed along the field is
constant.

7
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Separating the motion to the motion in the plane perpendicular to B0 ||z, that is described by
the perpendicular components of the velocity, Vx and Vy , and the motion along the field, Vz .
The RHS of the equation of motion is

x y z
V × B0 = Vx Vy Vz (21)
0 0 B0

Calculating the determinant, we obtain for the components of the velocity,


e
 V̇x = − me B0 Vy ,

V̇y = + mee B0 Vx , (22)

 V̇ =
z 0.

Thus, we confirm the previously obtained result that along the field the electron moves at the
constant speed Vz .
Differentiating the equations for Vx and Vy with respect to time, and using the original equa-
tions for Vx and Vy we obtain two harmonic oscillation equations,

eB0 2
 
V̈x + Vx = 0, (23)
me
eB0 2
 
V̈y + Vy = 0. (24)
me

The coefficients in the second terms have the dimension of frequency squared,

eB0
ωce ≡ , (25)
me

which is known as the cyclic electron gyrofrequency (also called the electron cyclotron frequency
or Larmor frequency). We can also define an electron gyrofrequency as fce = ωce /2π. Similar
frequencies can be introduced for other charged species of the plasma, e.g. ions,

|qi |B0
ωci ≡ , (26)
mi
where qi and mi are the ion’s charge and mass, respectively.
We can see that the gyrofrequency is independent of the charged particle energy (or the plasma
temperature): all particles of the same kind gyrate with the same frequency.
Thus, the trajectory of the electron is

x(t) = x0 + ..., (27)

y(t) = y0 + ...., (28)
z(t) = z0 + ... (29)

8
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

We choose the cos and sin solutions of the harmonic ODE to ensure that |V| = const (the
kinetic energy of the electron remains constant).
As the Lorentz force V × B0 is perpendicular to the velocity vector, in the plane perpendicular
to the magnetic field the particle path is a circle. The radius of the circle that the particle
motion is in the perpendicular plane is called gyroradius (also called Larmor radius), rLi ≡
V⊥ /ωci .

Such a motion is called gyration. Positively and negatively charged particles rotate in the
plane perpendicular to the field in the opposite directions. Specifically, the positively charged
particles orbit in a left-hand orbit, while the negatively charged particles orbit in a right-hand
orbit.
Thus, the motion of the electron (as well as any other charged particle) in the plasma with
a constant magnetic field is a combination of two motions: the gyro rotation in the plane
perpendicular to the field, and a steady motion along the field. In other words the particle has
a helical trajectory:

The pitch angle of the helical curve is defined as tan α = V⊥ /Vk . The centre of the circular orbit

9
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

in the perpendicular plane is called the guiding centre. The particle motion can be considered
then as a combination of the gyration and the motion of the guiding centre.
The electron gyrofrequency and gyroradius can be estimated as fce [Hz] ≈ 2.8 × 106 B0 [G] and
rLe [m] ≈ 2.4 × 10−2 (T [K]/11605)1/2 B0−1 [G], respectively, where the magnetic field is measured
in G (1 T = 104 G).

1.6 Drift Motion

1.6.1 Guiding centre drift due to non-magnetic forces

Consider the case when, in addition to the external magnetic field, the charged particle is
affected by another, non-magnetic constant force F acting in the plane perpendicular to the
magnetic field B0 . The dynamics of a charged particle is described by the equation of motion,
q F
V̇ = V × B0 + . (30)
m m
As in Section 1.5, the particle motion can be considered as a sum of two motions, V = V⊥ +Vk ,
perpendicular and parallel to the magnetic field, respectively.
The parallel component of the equation of motion is
V̇k = 0, (31)
and, consequently, Vk is constant.
In the plane perpendicular to the magnetic field the equation of motion is
q F
V̇⊥ = V⊥ × B0 + . (32)
m m
Define a new variable,
F × B0
u = V⊥ − = V⊥ − VD , (33)
qB02
where
F × B0
VD ≡ . (34)
qB02
Using u we can rewrite Eq. (32) as
q
u̇ = u × B0 . (35)
m
We can prove it by substituting Eq. (33) in (35), and using that VD is constant, F ⊥ B0 , and
the vector identity −(F × B0 ) × B0 = B0 × (F × B0 ) = B02 F − (F.B0 )B0 .
Eq. (35) coincides with Eqs. (18) that are the equations of motion for a charged particle in a
constant magnetic field with no non-magnetic forces. The solution to Eqs. (18) is the gyration
with the gyrofrequency. Thus, the solution for u gyrates with the gyrofrequency. Returning
back to the laboratory frame of reference, with the use of Eq. (33), we see that the particle
experiences

10
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

• gyration with the gyrofrequency, and

• drift with the drift velocity VD . As VD k (B0 × F), the drift motion is perpendicular to
both the magnetic field B0 and the force F.

1.6.2 Drift in a perpendicular electric field

Consider the case when the non-magnetic force F is the force of a static electric field
E ⊥ B0 , thus F = qE. Substituting this expression in Eq. (34) we obtain the drift
velocity,
E × B0
VD = , (36)
B02
perpendicular to both E and B0 .
Note that in this case the drift velocity is independent of the electric charge, and hence
both electrons and positive ions drift in the same direction, perpendicular to both E and
B0 .

1.6.3 Drift in a perpendicular gravitational field

Consider the case when the non-magnetic force F is the gravitational force, F = mg.
The drift velocity is
m g × B0
VD = , (37)
q B02
perpendicular to both g and B0 . This time the drift velocity does depend on the sign
of charge q. Hence the electrons and positive ions drift in the opposite directions. It
causes an electric charge separation and an associated electric current known as the drift
current. Its density can be estimated as j = n+ q+ VD+ − n− q− VD− , with the indices
indicating the positive and negative charges.

1.6.4 ∇B0 drift

In realistic situations the external magnetic field B0 is not uniform.

If it varies slowly in space, we can Taylor-expand it as
B0 (r) = B00 + (r.∇)B0 (r) + ..., (38)
with all the terms evaluated at the guiding centre. Thus, the equation of motion becomes
q q  
V̇ ≈ V × B00 + V × (r.∇)B0 (r) . (39)
m m
The second term on the RHS can be considered as an extra term responsible for the drift.
Consider a magnetic field that is straight, but has varies in the perpendicular direction.
Thus, its gradient ∇B0 ⊥ B0 . Thus, the gyroradii are different in different locations
across the field. For example:

11
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

We see that the guiding centre of the motion goes upwards, in the direction perpendicular
to both the field and its gradient. It is the ∇B0 -drift (also called the gradient drift). In
this case
1
VD ∝ B0 × ∇B0 . (40)
q

As the drift speed depends upon the sign of the electric charge, the ∇B0 -drift produces
an electric current.

1.6.5 Magnetic curvature drift

Consider a curved magnetic field. Locally the plasma particles experience gyration and
a steady motion of the guiding centre along the field.

12
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Because of the curvature that is locally characterised by the curvature radius R directed
in the direction opposite to the normal unit vector n, the guiding centre experiences a
centripetal force,
mVk2
Fc = − n. (41)
R
It is another example of an external force. Substituting it in Eq. (34) we obtain for the
drift velocity
Fc × B0 mVk2 n × B0
VD = = − . (42)
qB02 qR B02
This drift is called a curvature drift.
We see that the curvature drift speed depends upon the sign of the electric charge, hence
it leads to the charge separation and induces an electric current.
For example, in the Earth’s magnetosphere, the dipolar magnetic field varies in the radial
direction and the field lines are curved. Thus, it induces both the gradient and curvature
drifts and hence an electric current. This current is called a ring current. It is directed
along the equator in the westward direction at a height of 3–5 Earth’ radii:

The magnetic field created by the ring current is opposite to the Earth’s magnetic field,
and so causes a decrease in the total magnetic field in this region with important geo-
physical implications. During a geomagnetic storm, the number of particles in the ring
current increases, and magnifying its geophysical effect.

1.7 Thermal effects in electron plasma oscillations

Let us reconsider the electron plasma oscillations discussed in Sec. 1.2, accounting for
the thermal effects, i.e. the effects associated with the finite temperature.
When we displaced the slab of electrons from the quasi-static equilibrium, we considered
the slab affected by the electric field induced by the electric charge separation. However,
an additional physical quantity modified by this displacement is the density of the elec-
trons. If we account for the thermal effects, the local increase in the electron density is
accompanied by the local increase in the electron pressure. More specifically, it can be
obtained from the state equation (e.g. the ideal gas law, pe ∝ ρe = me ne , provided the

13
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

temperature does not change strongly). This effect vanishes when the temperature goes
to zero (i.e. when the thermal effects can be neglected in comparison with other, e.g.
electromagnetic effects). This local increase in the electron pressure is not compensated
by a decrease in the proton pressure, as the protons remain stationary.
This local increase in the electron pressure constitutes an additional restoring force −∇pe ,
that pushes the electrons back to the equilibrium. This force is the well-known force
that is responsible for the existence of, e.g., sound waves, drives the wind between the
geographical locations with the increased and decreased air pressure, etc.
As this −∇pe force depends on the steepness of the gradient of the pressure, it affects
perturbations of the same amplitude (i.e., the change of the pressure) but different wave-
lengths differently. It should affect the wave propagation.
Without the derivation, in the presence of thermal effects, the dispersion relation for
electron plasma oscillations becomes
 3 
ω 2 = ωpe
2
1+ λ D k 2
, (43)
(2π)2

where k is the wave number and λD is the Debye length. From this equation one can
calculate the phase and group speeds of the perturbations, that are finite this time. Thus,
thermal effects lead to the propagation of electron plasma oscillations: the phenomenon
of electron plasma waves (also called “electrostatic” or “Langmuir” waves). These waves
are essentially longitudinal, as the plasma moves along the direction of the wave vector.
The propagation speed is usually low, about the sound speed, as it is controlled by the
electron temperature via the temperature dependence of the Debye length.

2 Dielectric Description of Magnetised Plasmas

2.1 EM Wave Equation in a Plasma with B0 6= 0

In the majority of natural and laboratory plasma systems there is an external magnetic
field. Consider dynamics of EM waves in such a plasma.
Since B0 6= 0 there is anisotropy; the properties of the medium depend on the angle made
with the magnetic field, by (for example) the wave-vector of an EM wave. Depending on
the nature of the background field, particles will in general experience gyration and drifts.
How do EM waves interact with plasmas in the presence of the background magnetic field
B0 ?
We assume that

– B0 is ‘straight’ and constant.

– |B0 |  |B|, the magnetic field of the EM wave.

If we have an electric field external to the plasma, for example due to an EM wave passing
through it, then the electrons and protons move with respect to each other → have a

14
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

current. This is described by the current density, j,

j ∝ E → j = σE.

This is Ohm’s law with σ the conductivity. Since we have B0 6= 0, the motion of particles
consists of field-aligned motion, gyration and drift, then σ should be a tensor (3 × 3).
 
σxx σxy σxz
σ̂ =  σyx σyy σyz  → j = σ̂E.
σzx σzy σzz

How do EM waves propagate in a plasma with a tensor conductivity?

We begin with Ampère’s law

∂E
∇ × H = ε0 + j.
∂t
We should be using D, not E, in this form of Ampère’s law, but since we know nothing
of the polarisation P we shall proceed using E with the polarisation effects included in j.
We now use
B = µ0 H and µ0 ε0 = 1/c2
and let
E and B ∝ e−iωt
∂
such that → −iω and
∂t
 
iω i
⇒ ∇×B=− 2 Î + σ̂ ·E, (44)
c ε0 ω
| {z }
ˆ

where here Î is the identity matrix and ˆ is the dielectric tensor.

Thus
iω
∇×B=− ˆ · E. (45)
c2
Now use
∂B
∇×E=− = iωB
∂t
and combine it with equation (45), getting

ω2
∇2 E − ∇ (∇ · E) + ˆ · E = 0. (46)
c2
This is an EM wave equation, with all of the information about the medium encoded in
ˆ. There is no plasma yet in our equations. To introduce it, we should determine the
dielectric tensor ˆ from micro-physics.

15
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

2.2 Microscopic description of EM wave interaction with a magnetised

plasma

Consider dynamics of an electron in a plasma with an external magnetic field B0 and in

the field of an EM waves. In general, dynamics of a non-relativistic electron is described
by the equation of motion
mV̇ = F.

Let the background magnetic field be parallel to the z axis, i.e. B0 = B0 ẑ. Then

mV̇ = −eE(t) − eV × (B0 + B(t))

If |V|  c then |B(t)|  |B0 | (see the problem sheet)

⇒ mV̇ = −eE(t) − eV × B0

Split into components parallel and perpendicular to B0 :

mV̇z = −eEz (t), (47)

mv̇⊥ = −eE⊥ (t) + mω ce × v⊥ (t), (48)

eB0
where ω ce = is a vector with the magnitude equal to the electron gyrofrequency
m    
Vx Ex
and directed along the magnetic field, and V⊥ = and E⊥ = .
Vy Ey
In the following we assume that ω 6= ωce .

2.3 Matrix representation

∂
Let ∂t = −iω then from equations (47) and (48), differentiating Eq. (48) with respect
to time, and substituting the time derivative of V⊥ on the right hand side of the new
equation by the expression that we obtain from Eq. (48), we get

2 e
− ω 2 V⊥ = − (−iωE⊥ + ωce × E⊥ ) ,


ωce
m
e
iωVz = Ez .
m
These set of three algebraic equations can be written as a matrix equation

V = α̂ · E (49)

16
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

with
iω ωce
 
0
 ωce − ω 2
2 ωce − ω 2
2 
e  ωce iω

α̂ = − 2 0 
 
m ωce − ω 2 ωce − ω 2
2


 i 
0 0 −
ω
N.B.
1. α̂ is close to symmetric but there is anisotropy (note sign of off diagonal elements)
due to the background field B0 in the medium.

2. In this discussion and the following we do not consider the case where ω = ωce . This
corresponds to a resonance known as the gyro resonance or the electron cyclotron
resonance (ECR).

2.4 Elements of the dielectric tensor ˆ

In the derivation it assumed that protons and other ions are heavy and hence stationary
(do not repsond to the applied EM fields). Thus in our consideration the frequency of
EM wave is sufficiently high, much higher than e.g. the proton gyrofrequency. Under
these assumptions, the current density is
j = qnv = −ene v.
Using equation (49)
j = −ene α̂ E
| {z }
σ̂
which is Ohm’s law.

N.B. All of the above is based on the assumption that ω  Ωi = ωpi or the plasma
frequency of ions is much less than the frequency of the EM waves. ⇒ can neglect ion
contribution to the current.

Substituting into the expression for ˆ we obtain its matrix elements

 2
ωpe 2 
ωce ωpe
 1 + ω 2 − ω 2 −i ω ω 2 − ω 2 0   
 ce
2
ce
2
 1 −i2 0
 ω
ce ωpe ωpe 
 =  i2
ˆ
=  i 1+ 2 0 1 0 
2
 ω ωce − ω 2 ωce − ω 2


2 
0 0 3
 ωpe
0 0 1− 2
ω
r
ne
where we identify ωpe = e , the electron plasma frequency. The matrix ˆ contains
ε0 me
all of the information about the medium.

17
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

3 Dispersion Relations for High frequency EM Waves in a

Cold Plasma

3.1 The dispersion relation

Here a cold plama is one where thermal motion is negligible and high frequency again
implies ω  ωpi . Equation (46) can be simplified by the assumption E ∝ exp (ik · r)
then ∇ → ik and equation (46) becomes algebraic (that is much simpler than a second
order PDE). We have a set of three homogenous equations

ω2
−k 2 E + k(k · E) + ˆ · E = 0
c2

⇐⇒ M̂ · E = 0 (50)

where
1 − Ny2 − Nz2 −i2 + Nx Ny
 
Nx Nz
ω2
M̂ = 2  i2 + Nx Ny 1 − Nx2 − Nz2 Ny Nz 
c 2 2
Nx Nz Ny Nz 3 − Nx − Ny

with N = ckω and Nx,y,z the corresponding components of this vector. The absolute value
of N being the refractive index of the medium.
Non-trivial solutions to equation (50) exist if det(M̂ ) = 0, which hence is the dispersion
relation for high frequency EM waves in a cold magnetised plasma. It shows that the
EM wave propagation is anisotropic, it depends on the angle between the wave vector k
and the ambient magnetic field vector B0 .

For waves parallel to the field (k||B0 ||ẑ):

Nx = Ny = 0, and the dispersion relation reduces to

det(M̂ ) = 3 [(1 − Nz2 )2 − 22 ] = 0. (51)

For waves perpendicular to the field (k ⊥ B0 |ẑ):

Let k||x̂, then Ny = Nz = 0 and the dispersion relation is

det(M̂ ) = (3 − Nx2 )[1 (1 − Nx2 ) − 22 ] = 0.

3.2 EM Waves in a Plasma without B0

If there is no an external magnetic field, B0 = 0, the electron gyrofrequency ωce = 0, and

thus 1 = 3 = 1 − ωpe2 /ω 2 , and  = 0. Therefore, the dielectric tensor becomes diagonal,
2
and can be written as
 2 
ωpe
 = 1 − 2 I.
ˆ (52)
ω

18
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

In this case Eq. (51) becomes

!2
2  2
 ωpe ωpe k 2 c2
1− 2 1 − 2 − z2 = 0. (53)
ω ω ω

It has two solutions:

The solution ω 2 = ωpe

2 that corresponds to the electron plasma oscillations. The second

solution,
ω 2 = c2 k 2 + ωpe
2
, (54)
where we omitted the index z in the expression for the wave number, corresponds to the
electromagnetic waves. Without the external magnetic field the plasma is isotropic, and
the wave propagation is independent of the direction.
Dispersion relation (54) describes EM waves in the plasma. When the medium is a
vacuum, ωpe = 0, the dispersion relation reduces to ω 2 = c2 k 2 , with the phase and group
speeds, ω/k and dω/dk, respectively, both equal to the speed of light and to each other.
Dispersion relation (54) shows that in a plasma, the lowest frequency of the propagating
EM waves is ωpe . This frequency is thus a cut-off frequency. Generally, at a cut-off
frequency, waves experience reflection.
For ω > ωpe , the phase and group speeds of electromagnetic waves are not equal to
each other, and depend on the wave q number and hence, on the frequency. Re-writing
the dispersion relation as ω = ± c2 k 2 + ωpe 2 , calculate the phase speed V
ph = ω/k =
q
± c2 + ωpe 2 /k 2 . The group speed V
gr = dω/dk is calculated using the rule for the
differentiation of an implicit function,
dω
2ω = 2c2 k,
dk
dω c2 k c2 k
= = ±q
dk ω c2 k 2 + ω 2pe
.

19
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

For ω < ωpe the wave number k becomes imaginary, and the wave cannot propagate in
the plasma. However, it still penetrates the plasma, while decreasing exponentially with
the distance,
1q 2 2 =±
iq 2
k=± ω − ωpe ωpe − ω 2 . (55)
c c
Thus,
 the wave amplitude depends on the spatial coordinate z exponentially, ∝ exp (ikz) =
√ 2 2 
ω −ωpe 2 )−1/2 . The parameter δ deter-
exp ∓ c z = exp (−z/δS ), where δS = c(ω 2 − ωpe S

mines the depth of the penetration of the EM wave in the plasma, in the case ω < ωpe . It
is called the skin depth (also the exponentiation distance, e-folding length). For ω  ωpe ,
the scale δS becomes c/ωpe , which is referred to as the “electron inertial length”.

3.3 Parallel, High Frequency Waves in Cold Magnetised Plasmas

Consider the case when the waves propagate parallel to the magnetic field. The dispersion
relation for the waves parallel to the magnetic field describes three wave modes:

• 3 = 0
2
ωpe
1− = 0 ⇐⇒ ω = ωpe
ω2
the electrostatic oscillations, like it has been discussed in Electrodynamics.
• (1 − Nz2 )2 − 22 = 0, which has two roots
!1/2
2
ωpe
ω
k= 1−
c ω(ωce + ω)

which describes left hand circularly polarised EM waves and

!1/2
2
ωpe
ω
k= 1+
c ω(ωce − ω)

which describes right hand circularly polarised EM waves.

20
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

√
Figure 1: The dispersion curves are shown on the ωk-plane (in the case ωpe /ωce < 2). Note,
that this description is incomplete, because there are other waves modes well below the electron
plasma frequency ωpe and the electron Larmor frequency ωce . To describe these low frequency
wave modes, we should take into account the motion of ions.

21
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Applying the condition k = 0, we can find the cut-off frequencies as the solutions to these
equations. The cut-off frequencies of left hand and right hand circularly polarised EM
waves are ω1 and ω2 respectively, where
s 
4ω 2
ωce  pe
ω1 = 1 + 2 − 1 , (56)
2 ωce
s 
4ωpe2
ωce 
ω2 = 1 + 2 + 1 . (57)
2 ωce

A wave mode can propagate in the plasma with given ωce and ωpe if the wave frequency
is higher than the cut-off frequency for this wave mode. Also, there is another right
hand circularly polarised mode, with the frequency ω < ωce . This wave mode is called
a whistler mode. When the frequency ω of the right hand circularly polarised mode
approaches ωce from below, the wave number k tends to infinity. This phenomenon is
called resonance. In this case, the resonance occurs because the frequency of the right
hand circularly polarised waves approaches the Larmor (gyro) frequency of electron gy-
ration, which is also a right-handed motion. The phenomenon is called electron cyclotron
resonance (ECR).
Which wave, left hand or right hand polarised is faster? It depends upon what we mean
by “faster ”- the phase speed or the group speed.

3.4 Whistler Mode

We examine the whistler wave mode (ω < ωce ), for right hand circularly polarised EM
2
ωpe
waves. Try ω  ωce and assume ωce ω  1 such that
!1/2
2
ωpe
ω
k= 1+
c ω(ωce − ω)
reduces to
ωpe √
k ≈
√ ω
c ωce
ck 2
 
⇒ω ≈ ωce
ωpe
⇒ ω ∝ k2 .

The group velocity is not constant.

dω 2ωce c2
vg = ≈ 2
k.
dk ωpe
We therefore observe dispersion, with higher frequency components travelling faster than
lower frequency ones. A signal composed of multiple frequency components will be
‘stretched out’ as it travels through a plasma in the whistler mode (with ω  ωce ).

22
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

3.5 Faraday Rotation

The dispersion relations give us kR = fR (ω) that is the wave number of the RH circularly
polarised EM wave, and kL = fL (ω) that is the wave number of the LH circularly
polarised EM wave.
Note that both have the same argument ω. In an electromagnetic wave E is a combination
of both LH and RH circularly polarised waves
h    i
E = ER eikR z + EL eikL z x̂ + i ER eikR z − EL eikL z ŷ e−iωt ,

where EL and ER are the amplitudes of the LH and RH circularly polarised waves
respectively. We can also write
!
Ex 1 + EL /ER ei(kL −kR )z
= −i , .
Ey 1 − EL /ER ei(kL −kR )z

We can introduce the polarisation angle

 
−1 Ey
ϕ ≡ tan .
Ex

On the plane perpendicular to the wave vector, this is the angle between e.g. the ver-
tical direction and the line along which the electric field in the EM wave changes. The
polarisation angle is determined by the phase shift between the perturbations of Ey and
Ex in the wave. Because of the dispersion, the polarisation angle evolves with the spatial
coordinate, in this case z, since ω/kL 6= ω/kR (different phase speeds).
Let us assume for simplicity that ER = EL , considering linearly polarised waves only.
Let us also assume that ω  ωpe and ω  ωce . Thus, we obtain that
2
ωce ωpe
∆k = kL − kR ∝ ∝ B0 ne .
cω 2

So, in the case of our linearly polarised wave, propagating through the plasma parallel
to a background magnetic field, the plane of the perturbation of the electric field vector
E of the EM wave rotates, its orientation dependant on the spatial coordinate z. This is
the effect known as Faraday Rotation.

3.6 Plasma Diagnostics Using Faraday Rotation

Thus, during the propagation of an EM wave through the plasma, its polarisation angle
changes proportional to the difference between the left hand and right hand polarised
waves, and the distance z traveled,

∆ϕ = (kL − kR )z.

23
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Hence for the plasma with the electron concentration ne and the magnetic field B0 , the
change of the polarisation angle with distance is

2
dϕ ωce ωpe e3 ne B0
= (kL − kR ) = ∆k = =
dz 2cω 2 2ε0 m2e c ω 2
This is the change of the polarisation angle with the distance. (Mind that if there is no
magnetic field (B0 = 0) or there is no plasma (ne = 0), the angle remains constant: there
is no the effect of Faraday rotation.
By observing dϕdz and knowing ω we can determine ne , B0 .
If ne (z) and B0 (z) are slowly varying functions i.e. the characteristic distance of the
2π 2π
variation in ne (z) and B0 (z) is  , then
kL kR
Z
∆ϕ = ϕ − ϕ0 = ∆kdz
LOS

where LOS is the Line-Of-Sight or the path of the wave. We have

e3
Z
1
∆ϕ = ne (z)B0 (z)dz
2ε0 m2e c ω 2 LOS
This is the formula used for plasma diagnostics by Faraday Rotation of EM waves.

3.7 Applications
1. Interplanetary Scintillation (IPS)
The interplanetary plasma can be analysed by observations of the EM radiation
emitted by extra-galactic sources. In this way, Parker’s model for the heliospheric
magnetic field produced by the sun was verified, see figure 2.
2. Microwave probing of laboratory plasmas
3. Active probing of space plasmas
For example, the ionosphere can be studied by sending signals between satellites,
or from a ground base transmitter.

3.8 Perpendicular, High Frequency Waves in Cold Magnetised Plasmas

In section 3.3 and subsequent sections up to now, we have discussed the special case
where k|B0 . Now we briefly discuss the case where k ⊥ B0 |ẑ. The appropriate dispersion
relation (for k|x̂) is given in section 3.1 as

(3 − Nx2 )[1 (1 − Nx2 ) − 22 ] = 0.

One solution is
3 − Nx2 = 0 ⇒ ω 2 = c2 k 2 + ωpe
2
,

24
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 2: Parker’s spiral, the observed form of the heliospheric magnetic field and the plasma
density. Left: distribution of the electron density in the ecliptic plane measured by IPS. Right:
a sketch of the magnetic field structure.

which coincides with the dispersion relation for EM waves in a plasma in the absence
of a background magnetic field. Hence this is known as the “Ordinary Mode” (or an
“o-mode”) and has a cut-off frequency at ω ≈ ωpe .
The ordinary mode is “ordinary”, in other words it does not “feel” the magnetic field,
because it perturbs the electrons in the direction strictly parallel to the external magnetic
field (the electric field in this mode is parallel to the external magnetic field, see Fig. 4).
Hence, the Lorentz force is zero. Thus, this mode is strictly linearly polarised.
Another solution is
1 (1 − Nx2 ) − 22 = 0

s
1 (ω 2 − ω12 )(ω 2 − ω22 )
⇒k=
c ω 2 − ωU2 H
q
where ωU H = ωpe 2 + ω 2 and is known as the “Upper Hybrid Frequency”. By inspection
ce
we can see that there are cut-offs at ω = ω1 and ω = ω2 where ω1 , ω2 are given in
equations 56 and 57. There is also a resonance at ω → ωU H , known as the upper hybrid

25
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

resonance. The dispersion curves are shown in figure 3. This mode is known as an “extra-
ordinary” mode (or an “x-mode”). It is elliptically polarised. Also, it has a component of
the electric field perturbation parallel to the wave vector (see Fig. 4). Hence, this mode
is partly longitudinal (electrostatic) and partly transverse.
In general, EM waves can be propagating at any angle to B0 and their properties can be
obtained from the general dispersion relation det(M̂ ) = 0

Figure 3: The dispersion curves for EM waves propagating perpendicular to the background
magnetic field are shown on the ωk-plane.

Note that there are also low frequency EM waves propagating through magnetised plas-
mas, e.g. Alfvén waves, which are not described by this formalism.

3.9 Effects of finite temperature

Above, we neglected the effects of finite plasma temperature, considering the plasma to
be “cold”. Now, consider what is missing in that approximation.
Recall the derivation of the electron plasma oscillations. We displaced a slab of electrons
from the equilibrium. It led to the electric charge separation, and occurrence of the
restoring electrostatic force. Now, let us notice that at the new location of the displaced
electron slab, the electron concentration increases too (there are no the “old” electrons
plus the “new electrons”. In thermodynamics, the pressure is connected with the particle
concentration n via the state equation, i.e. p ∝ T n, where T is the temperature. In
our case, it is the electron pressure linked with the electron temperature and electron
concentration, pe ∝ Te ne . In the following consider the case Te = const.

26
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 4: The dispersion curves for EM waves propagating perpendicular to the background
magnetic field are shown on the ωk-plane.

The localised increase in ne results in the increase in pe (the ions are too heavy and are
not involved in this motion, hence we neglect the perturbation of the pressure associated
with them). It constitute the hydrodynamic force −∇pe directed from the region with
the enhanced electron concentration to the region with the decrease in ne . Thus, this
force acts in the same direction as the electrostatic force. This hydrodynamic force is
fully disconnected with the electrostatic force. In the “cold” approximation that force
did not appear, as Te = 0.
The hydrodynamic force depends on the spatial size of the displaced electron slab, λ,
which should not be confused with the displacement x that appeared in the derivation
of the electron plasma frequency,
ne
∇ne ≈ .
λ
The size of the displaced electron slab, λ is about the wavelength of the perturbation,
hence k = 2π/λ, where k is the wave number. Thus, the dispersion relation for electron
plasma oscillation becomes
ω 2 = ωpe
2
+ Cse2 2
k , (58)
where Cse is the electron sound speed determined by the electron temperature,
 γ k T 1/2
e B e
Cse = ,
me
and γe is the adiabatic index of electrons (i.e. the ratio of the heat capacity at constant
pressure to heat capacity at constant volume).
Dispersion relation (58) shows that the phase and group speeds of the perturbations
are finite, hence the oscillation propagates at a certain speed — it is an “electrostatic”
wave (also known as the Langmuir wave, electron plasma wave, etc.). This wave is

27
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

longitudinal, as the electrons moves in the direction of the wave propagation. The phase
and group speeds are not equal each other, hence the wave is dispersive. Equation (58)
is very similar to the dispersion relation for electromagnetic waves in a plasma without
the magnetic field. The only difference is the electron sound speed instead of the speed
of light.

4 Elements of Plasma Kinetics

We have already seen how a plasma could be described by single particle dynamics and
in terms of a dielectric tensor. Let us discuss another approach, based upon the methods
of statistical physics.

4.1 The Vlasov Equation

We use methods from statistical physics to study the kinetic properties of plasma, e.g.
the interaction of the plasma with particle beams, various resonances, wave–particle
interaction, etc. The distribution function f (r, v, t) gives the number density of particles
at position r with velocity v at a time t.
ZZZ ZZZ
f (r, v, t)dvdr = N, (59)
space velocity

where dr = dxdydz, dv = dvx dvy dvz and N is the total number of particles. Consider
a plasma system with a fixed total number of particles N so that dN/dt = 0, then
differentiating equation (59) with respect to time we have
ZZZ ZZZ
dN df
= dvdr = 0
dt dt
space velocity

df
⇒ = 0. (60)
dt
Equation (60) is the Boltzmann equation. Applying the chain rule we have
df ∂f ∂r ∂f ∂v ∂f
= + + =0
dt ∂t ∂t ∂r ∂t ∂v
∂f ∂f F ∂f
= +v + = 0.
∂t ∂r m ∂v

Now, if the we consider a collisionless plasma, then the only contribution to F is the
Lorentz force due to electromagnetism and we obtain the Vlasov equation,

df ∂f ∂f q ∂f
= +v + [E + v × B] = 0. (61)
dt ∂t ∂r m ∂v

28
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

The Vlasov equation should be supplemented by Maxwell equations for the EM field. The
Vlasov equation describes the interaction of EM waves and particles in a plasma, espe-
cially the effect connected with the departure from the thermal equilibrium. In general,
Vlasov equations should be written for each species (electrons, protons, ions) separately.
In collisionless plasmas these particles interact with each other via the perturbation of
the EM field.

4.2 Landau Damping

Consider a one dimensional (1D) longitudinal perturbation in a plasma with B = 0

(no background magnetic field). The unperturbed distribution is f = f0 (v) where f0 is
the equilibrium state given by some known function such as a Maxwellian distribution.
f = f0 (v) is equivalent to saying B, E = 0 since there is no dependance on time or
position for the distribution.
Now apply a perturbation f1 (r, v, t) such that

f (r, v, t) = f0 (v) + f1 (r, v, t).

We assume that f1  f0 . Substitution into the Vlasov equation (61) yeilds

∂f1 ∂f1 e ∂f0

+ vx − E1 = 0. (62)
∂t ∂x m ∂vx
In the following we assume that

– Everything depends only on the coordinate x (1D);

– f1  f0 such that we can perform linearisation1
– E1 is the electric field induced by the perturbation.

Let the perturbations be harmonic, then f1 ∝ exp(ikx − iωt) and equation (62) becomes

e ∂f0
−iωf1 + ikvx f1 = E1 ,
m ∂vx

ieE1 ∂f0 1
f1 = . (63)
m ∂vx ω − kvx

Now use the Poisson Equation ε0 ∇ · E1 = −en1 , where n1 is the electric charge density.
It can be obtained from the distribution function,
ZZZ
n1 = f1 dv.
velocity

1
Linearisation involves neglecting all terms containing f12 or higher powers. All terms linear in f1 or that
only contain f0n are retained;

29
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Here the integration is over all possible values and directions of the velocity.
ZZZ
ε0 ∇ · E1 = −e f1 dv (64)
velocity

From equations (63) and (64), eliminating f1 we find

e2
ZZZ
∂f0 /∂vx
1=− dv (65)
kmε0 ω − kvx
velocity

Equation (65) is a dispersion relation, it gives the relationship between k and ω. The
unperturbed (equilibrium) concentration n0 can be factorised out of the expression if we
replace f0 with a normalised distribution function F = f0 /n0 . The prefactor e2 n0 /mε0
can then be identified with the electron plasma frequency squared ωpe2 , and we have

2 Z+∞
ωpe
ZZ
∂F/∂vx
1=− dvx (66)
k ω − kvx
−∞

The integration with respect to vx and vy can be readily made e.g. the case of the
Maxwellian equilibrium, with the use of the identity
+∞
Z p
exp(−ax2 )dx = π/a. (67)
−∞

Now, consider the familiar limit of the cold plasma. The equilibrium distribution function
in this case is a δ-function of vx : it is zero everywhere except when vx = 0. Let us use
that  
∂ F ∂F/∂vx kF
= + . (68)
∂vx ω − kvx ω − kvx (ω − kvx )2
Integrating both left and right hand sides of this expression with respect to vx from −∞
to +∞ gives that the left hand side is zero: the derivative goes with the integral, and the
substitution of the limits gives zeros. Thus, we can express the first term on the right
hand side via the second term. Thus, equation (66) can be rewritten as
+∞
Z
2 F
1= ωpe dvx . (69)
(ω − kvx )2
−∞

Again, using that F ∝ δ(vx ) we see that the denominator only matters when vx = 0,
when it becomes ω 2 . Using the normalisation
+∞
Z +∞
Z
F dvx = 1 or f0 dvx = n0 , (70)
−∞ −∞

30
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

we arriving at
2
1 − ωpe /ω 2 = 0. (71)
Thus, we obtained the familiar expression for electrostatic plasma oscillations (see Sec-
tion 3.3).
N.B. Relativistic effects are ignored such that the integral over all velocities is from
−∞ to ∞ and that the one dimensional nature of the perturbations is exploited so that
dv → dvx .
If we account for the thermal effects, Equation (66) becomes the dispersion relation
for Langmuir waves (also called plasma or electrostatic waves). In these waves, the
wave vector is parallel to the perturbation velocity vx . The expression, (66), contains a
singularity when ω/k = vx . To fully solve the equation we need to borrow a result from
the ‘Theory of Complex Functions’
Z ∞ Z ∞
∂F/∂vx ∂F/∂vx ∂F
dvx = P dvx + iπ
−∞ vx − ω/k −∞ vx − ω/k ∂vx ω/k
R∞
where P −∞ denotes the “principle value of the integral”.
Then, without derivation,

2
ωpe 2 2
ωpe
Cse 2 ∂F
1− − k − iπ ' 0. (72)
ω2 ω2 k 2 ∂vx ω/k

2
ωpe
• The term 1 − describes oscillations of frequency ωpe ;
ω2
2
Cse
• The term k 2 describes propagation at around the speed of sound;
ω2
2
ωpe ∂F
• The term iπ 2 describes kinetic effects. Here the derivative (∂F /∂vx )|ω/k
k ∂vx ω/k
is evaluated in the vicinity of vx = ω/k.

We see that ω is complex, ω = Re(ω) + iIm(ω). (Previously we have had Im(ω) = 0 and
ω has always been real.) Again without derivation

2
πωpe
Im(ω) ∂F
' 2
(73)
Re(ω) 2k ∂vx ω/k

– If Im(ω) < 0, then

e−iωt = e−iRe(ω)t−iiIm(ω)t = e−iRe(ω)t+Im(ω)t

and the perturbations (Langmuir waves) decay with time.

– If Im(ω) > 0, then the solution grows exponentially with time.

31
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 5: The Maxwellian distribution

F . The particles in region 1 gain en-
ergy from the Langmuir wave since their
initial velocity vx is below that of the
wave. Those in region 2 transfer energy
to the wave since their initial velocity
vx is greater than that of the Langmuir
wave.

So for these plasma waves we see that there is either decay or an instability depending
on the sign of ∂F/∂vx .
The sign of the imaginary part of the frequency of the perturbation and hence the plasma
dynamics is defined by the sign of the derivative,

∂F
Im(ω) ∝ ,
∂vx ω/k

since all other parameters in equation (73) are either positive or squared.
When F is Maxwellian,
vx2
 
1
F =√ exp − 2
πvT e vT e
with vT e the thermal electron speed (the average speed of thermal motion), (see figure
5) therefore ∂F/∂vx is always < 0
The waves in this case experience decay, or damping. This is known as “Landau Damp-
ing” after Lev Landau. The plasma waves experience decay due to the presence of
“resonant particles” with velocities approximately equal to the phase speed of the wave.
Particles in region 1 of figure 5 gain energy from the Langmuir wave since they have vx
below the phase speed of the wave. The particles in region 2 of figure 5 transfer energy
to the wave since they have vx greater than the phase speed of the wave. As can be seen
from figure 5 the number of particles in region 1 > number of particles in region 2 (this
is prescribed by the sign of the derivate of the distribution function). Therefore the net
transfer of energy is from the wave to the particles, hence damping.

4.3 Bump-on-tail instability

What about the case where ∂F/∂vx > 0? Consider an example of a Maxwellian plasma
with a beam (figure 6).
The part of the distribution where ∂F/∂vx > 0 leads to an instability. The growth rate
of which can be obtained from Im(ω)

32
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 6: The by-Maxwellian distri-

bution created by the combination of
the original Maxwellian plasma and a
plasma beam with its own Maxwellian
velocity distribution. A part of the
curve has ∂F/∂vx > 0 and hence leads
to an instability within the plasma.

e−iωt = e−i(iIm(ω))t = eIm(ω)t = et/t0

So we would observe exponential growth of the instability with t0 = [Im(ω)]−1 . Such an

instability is known as “bump-on-tail” or “streaming” instability. The manifestation of
the instability is the increase in the level of electrostatic fluctuations in the plasma.
The bump-on-tail instability is an example of a plasma microscopic instability. Micro-
scopic instabilities (“microinstabilities”) operate in the velocity space, and occur in the
configuration space on a microscopic scale (e.g., gyroradius, inertial length). In con-
trast, macroscopic instabilities (“macroinstabilities”) operate in the configuration space
on much larger scales, e.g. the size of a plasma or magnetic field non-uniformity, which
is typically  than, e.g., gyroradius and inertial length.

An Example: Type III Radio Bursts

– Bursts of EM radiation at frequencies < 1 to 100 MHz are often seen coming from
the Sun (see Figure 8). These are generated by solar flares, when supra-thermal
electrons (v  vT e ) are ejected from active regions and follow open magnetic field
lines outwards the Sun.
– The distribution function for the velocities of the plasma particles becomes by-
Maxwellian, with a ‘bump’ on the tail of the distribution (the background plasma
plus the outwardly propagating beam of suprathermal electrons).
– There is “bump-on-tail”instability as electrons excite plasma oscillations along their
path. According to the dispersion relation, these oscillations have ω ≈ ωpe .
– The plasma oscillations excite EM waves with the same frequency, which can propa-
gate perpendicular to the magnetic field (e.g. Section 3.8). These waves are recorded
by our receivers.
– The plasma density in the corona is stratified (decreasing with increasing height),
therefore the electron-plasma frequency also decreases with height, ne (z) → ωpe (z)
– The net result is a time dependance of the arrival times of different frequencies, as
the flare propagates through the corona into the regions of continuously decreasing
ωpe . This is responsible for the frequency drift in Figure 8.

33
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 7: Type-III Radio Bursts Produced by the Oct. 28 and Nov. 4, 2003, X17 and X28
Solar Flares as detected by the Cassini Radio and Plasma Wave Science Instrument. (From
http://www-pw.physics.uiowa.edu

5 Magnetohydrodynamics (MHD)

Macroscopic (also called “fluid”) parameters of plasma can be obtained from the distri-
bution function fα (r, V, t). For example, the concentration of the particles of the kind
α (where the index α is, e.g., e for electrons, p for protons, etc.) is

Z+∞
ZZ
nα = fα (r, V, t)dV, (74)
−∞

where the integration is over all possible speeds.

Likewise, the bulk velocity of the plasma is
Z+∞
ZZ
1
Vα = Vfα (r, V, t)dV, (75)
nα
−∞

and the electric current density is

 +∞

X  ZZZ 
j= qα Vfα (r, V, t)dV , (76)
 
α −∞

34
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 8: An EUV image of the solar corona taken with the Extreme Ultraviolet Imager on the
ESA/NASA Solar and Heliospheric Observatory (the colour is artificial). MHD is a standard
tool for the study of physical processes in the solar coronal plasma.

where qα is the charge of the species α, and the summation is over all kinds of charged
particle species in the plasma.
Likewise, we can determine other microscopic parameters. But, the description of a
plasma in terms of the Vlasov equation is very difficult from the computational point of
view, and hence, there is a need for a simpler approach, in particular the analogue of the
fluid mechanics.

5.1 MHD Equations

MHD describes large scale, slow dynamics of plasmas. More specifically, we can apply
MHD when

1. Characteristic time  ion gyroperiod and mean free path time,

2. Characteristic scale  ion Larmor radius and mean free path length,
3. Plasma velocities are not relativistic.

In MHD, the plasma is considered as an electrically conducting fluid. Governing equa-

tions are equations of fluid dynamics and Maxwell’s equations. A self-consistent set of
MHD equations connects the plasma mass density ρ, the plasma velocity V, the thermo-
dynamic (also called kinetic) pressure p and the magnetic field B. In regorous derivation
of MHD, one should neglect the motion of electrons and consider only heavy ions. Thus,
in the following, we consider dynamics of ions (e.g., protons) only, considering the elec-
trons to be “massless”.

35
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

The 1st equation is the mass continuity

∂ρ
+ ∇(ρV) = 0, (77)
∂t
and it states that matter is neither created or destroyed. It can be obtained by taking
that the mass within an arbitrary, fixed volume v of the plasma can be changed only by
flow of the plasma into or out of that volume via its surface S:
∂
(mass) = mass flow flux, (78)
∂t
Z Z Z
∂
ρdv = − ρV · dS = − ∇ · (ρV)dv, (79)
∂t v S v
where we used the Gauss theorem, giving us
∂ρ
+ ∇ · (ρV) = 0. (80)
∂t

The 2nd is the equation of motion of an element of the fluid,

 
∂V
ρ + (V∇)V = −∇p + j × B, (81)
∂t
also called the Euler equation. The vector j is the electric current density which can be
expressed through the magnetic field B. This equation is a hydrodynamical analogue
of the 2nd law of Newton. This equation describes the conservation of the mechanical
momentum. On the right hand side we have the vector sum of the forces applied to
the plasma element. In the following we restrict ourself to the consideration of the gas
pressure gradient and the Lorentz forces only (the 1st and the 2nd terms on the right
hand side, respectively). But, in general, we may have other forces there too, for example
the gravity, the Coriolis force, the viscous force, etc. On the left hand side we have the
total derivate of the bulk flow velocity with respect to time. The LHS term can be
written as ρdV/dt.
The 3-rd equation is the energy equation, which in the simplest adiabatic case has the
form  
d p
= 0, (82)
dt ργ
where γ is the ratio of specific heats Cp /CV , and is normally taken as 5/3. This equation
describes the conservation of energy.
The temperature T of the plasma can be determined from the density ρ and the thermo-
dynamic pressure p, using the state equation. For example, in a pure hydrogen plasma,
this equation is
kB
p = 2 ρT, (83)
mi
where mi is the mass of a proton and kB is Boltzmann’s constant.

36
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Now, let us derive the equation for the magnetic field using Maxwell’s equations. Start
with Ohm’s law,
j = σE0 , (84)
where σ is electric conductivity (the physical quantity inverse to the resistivity) and E0
is the electric field experienced by the plasma (fluid) element in its rest frame. When
the plasma is moving (with respect to the external magnetic field) at the velocity V,
applying the Lorentz transformation we obtain

E0 = E + V × B. (85)

Now, Eq. (84) can be re-written as

1
j = E + V × B. (86)
σ
In the case of perfect conductivity, σ → ∞, we have

E = −V × B. (87)

Calculating the curl of the electric field E and using one of Maxwell’s equation,
∂B
∇×E=− , (88)
∂t
we can exclude the electric field and obtain the 4-th MHD equation,
∂B
= ∇ × (V × B), (89)
∂t

To close the set of MHD equations, we have to express the current density j through the
magnetic field B. Consider the other Maxwell’s equation,
1 ∂E
∇ × B = µ0 j + (90)
c2 ∂t
From Ohm’s law, we had E = −V × B. Consequently, we can estimate the electric field
as E ∼ V0 B, where V0 is a characteristic speed of the process. Consider the ratio of two
terms in Eq. (90):
1 ∂E
∇ × B and .
c2 ∂t
The first term is proportional to B/l0 , where l0 is a characteristic scale of the process, the
second to E/c2 t0 , where t0 is a characteristic time of the process, V0 = l0 /t0 . When the
process is not relativistic, V0  c, the first term is very much greater than the second,
and we have
1
j= ∇×B (91)
µ0
In addition, the magnetic field B must satisfy the condition ∇ · B = 0.
Thus, the closed set of MHD equations is

37
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

∂ρ
+ ∇(ρV) = 0, Mass Continuity Eq.,
∂t  
d p
= 0, Energy Eq.,
d t ργ
dV 1
ρ = −∇p − B × (∇ × B), the Euler Eq.,
dt µ0
∂B
= ∇ × (V × B), Induction Eq..
∂t

The equations are ideal, which means that all dissipative processes (finite viscosity, elec-
tric conductivity and thermal conductivity) were neglected.
One of the basic properties of ideal MHD equations is the effect of the “frozen-in” mag-
netic field. Plasma motions along the field lines do not change the field, but motions
transverse to the field carry the field with them:

– if a parcel of plasma moves, the magnetic field attached to the parcel moves along
with it;
– the plasma cannot move across magnetic field lines (though it remains free to move
along the field).

In other words, any transverse displacement of the plasma leads to the change of the
magnetic field, and the other way around, any change of the magnetic field generates
plasma flows.

5.2 MHD Equilibrium

The static equilibrium conditions are:

∂
V = 0, = 0. (92)
∂t
These conditions identically satisfy the continuity, energy and induction equations.
From Euler’s equation we obtain the condition
1
− ∇p − B × (∇ × B) = 0, (93)
µ0
which is called the equation of magnetostatics. This equation should be supplemented
with the condition ∇.B = 0.
Eq. (93) can be re-written as

B2
 
1
−∇ p+ + (B.∇)B = 0. (94)
2µ0 µ0
The first term can be considered as the gradient of total pressure. The total pressure con-
sists of two terms, the kinetic (or thermodynamic) pressure P , and the magnetic pressure
B 2 /2µ0 . The second term is magnetic tension.

38
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 9: (Left:) Close-up a sunspot. (Right:) Sketch of the magnetic field in a sunspot.

Example: Sunspots

Sunspots appear as dark spots on the surface of the Sun. They typically last for several
days, although very large ones may live for several weeks. Sunspots are magnetic regions
on the Sun with magnetic field strengths thousands of times stronger than the Earth’s
magnetic field.
Consider a sunspot as a vertical magnetic flux tube (figure 10). The magnetic field B0 is
vertical. The kinetic pressure is p0 and pE inside and outside, respectively. The plasma
temperature is T0 inside the sunspot and TE outside.
Sunspots are long-durational objects with no fast flows of plasma. So, it is natural to
describe their structure in terms of magnetostatics. As the magnetic field is not bent,
the last term in Eq. (93), responsible for the magnetic tension, is zero. The equilibrium
condition becomes
B2
 
∇ p+ = 0, (95)
2µ0
This means that the total pressure must be equal inside and outside the sunspot,

B02
pE = p0 + . (96)
2µ0

Let us assume that the density of the plasmas inside and outside the sunspot are equal,
ρ0 = ρE . Now, we divide Eq. (96) by ρ0 ,

pE p0 B02
= + . (97)
ρE ρ0 2µ0 ρ0

39
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 10: Schematic diagram of a magnetic flux tube, such as those visible as sunspots on
the solar surface. (Mind the side-view!)

Using the state equations,

kB kB
pE = 2 ρE TE , p0 = 2 ρ0 T0 , (98)
mi mi
we obtain from Eq. (97)
2kB 2kB B02
TE = T0 + . (99)
mi mi 2µ0 ρ0
This gives us
T0 B2 mi B02
=1− 0 =1− (100)
TE 2µ0 2kB ρE TE 2µ0 pE
Thus, in a sunspot, TE > T0 . Indeed, temperatures in the dark centers of sunspots drop
to about 3700 K, compared to 5700 K for the surrounding photosphere. This is why
sunspots are seen to be darker than their surroundings.

5.3 MHD Waves

Ideal MHD connects the magnetic field B, plasma velocity V, pressure p and density ρ:
∂ρ
+ ∇(ρV) = 0, (101)
∂t
 
∂V 1
ρ + (V∇)V = −∇p − B × (∇ × B), (102)
∂t µ0

40
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

∂B
= ∇ × (V × B), (103)
∂t
 
d p
= 0. (104)
dt ργ

Consider an equilibrium, described by the conditions

∂
= 0, V = 0, (105)
∂t
which gives us the magnetostatic equation
1
∇p0 + B0 × (∇ × B0 ) = 0. (106)
µ0
The simplest possible solution of the magnetostatic equation is a uniform plasma:

p0 = const, B0 = const, (107)

and the equilibrium magnetic field B0 is straight.

Consider small perturbations of the equilibrium state:


B = B0 + B1 (r, t) 

V = 0 + V1 (r, t)

(108)
p = p0 + p1 (r, t) 

ρ = ρ0 + ρ1 (r, t)


Substitute these expressions into the MHD equations (101)–(104). Neglecting terms
which contain a product of two or more values with indices “1”, we obtain the set of
MHD equations, linearized near the equilibrium (107):

∂ρ1
+ ρ0 ∇V1 = 0, (109)
∂t
∂V1 1
ρ0 = −∇p1 − B0 × (∇ × B1 ), (110)
∂t µ0
∂p1 γp0 ∂ρ1
− = 0, (111)
∂t ρ0 ∂t
∂B1
= ∇ × (V1 × B0 ), (112)
∂t
Let the equilibrium magnetic field B0 be in xz-plane,

B0 = B0 sin α ex + B0 cos α ez , (113)

where α is the angle between the magnetic field and the unit vector ez .

41
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Consider plane waves, propagating along ez , so that all perturbed quantities are propor-
tional to exp(ikz − iωt). (This gives us ∂/∂t = −iω and ∇ = ik.) Projecting equations
(109)–(112) onto the axes, we have

− iωρ1 + ikρ0 Vz1 = 0, (114)

ikB0 cos α
− iωρ0 Vx1 − Bx1 = 0, (115)
µ0
ikB0 cos α
− iωρ0 Vy1 − By1 = 0, (116)
µ0
ikB0 sin α
− iωρ0 Vz1 + ikp1 + Bx1 = 0, (117)
µ0
− iωBx1 + ikB0 sin αVz1 − ikB0 cos α Vx1 = 0, (118)
− iωBy1 − ikB0 cos α Vy1 = 0, (119)
− iωBz1 = 0, (120)
iωγp0
− iωp1 + ρ1 = 0. (121)
ρ0

5.3.1 Alfvén waves

The set of equations (114)–(121) splits into two partial sub-sets. The first one is formed
by equations (116) and (119), describing By1 and Vy1 . The consistency condition gives
us
ω 2 − CA 2
cos2 α k 2 = 0, (122)
where CA = B0 /(µ0 ρ0 )1/2 is the Alfvén speed. This is dispersion relations for Alfvén
waves.
Physical processes in an Alfvén wave:

– An Alfvén wave can be considered as a transverse oscillation of a “heavy magnetic

string”;

42
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

– a localised transverse deformation of the field leads to the appearance of a magnetic

tension force, which tends to restore the equilibrium, i.e., to make the field straight;
– when the field is returning back towards the equilibrium configuration, it drags the
plasma with it, since the magnetic field is “frozen” in the plasma;
– during the return of the field towards equilibrium, the plasma gains kinetic energy
(as the field line is “heavy” because of the frozen-in condition);
– because of the inertia (the kinetic energy), the plasma overshoots the equilibrium,
and the field line becomes deformed in a direction opposite to the initial deformation,
until the magnetic tension force stops the deformation (when all the kinetic energy
has been converted into magnetic potential energy) and then the plasma motion is
revered, so that the field and plasma are returned back towards the equilibrium;
– the plasma again returns back to the equilibrium position with the kinetic energy,
and the process repeats;
– this perturbation propagates along the magnetic field at the Alfvén speed.
As this process is caused by the magnetic elasticity (which is proportional to the magnetic
field strength B0 ) and the plasma inertia which is proportional to its mass density ρ0 ,
the Alfvén speed is determined by B0 and ρ0 .
The main properties of Alfvén waves are as follows:
– Alfvén waves are transverse: the plasma is displaced in a direction perpendicular to
k. Thus, like any transverse wave, Alfvén waves can be polarised linearly, elliptically,
or circularly.
– In the linear regime, Alfvén waves are essentially incompressive: they do not modify
the density of the plasma, which results in ∇.v = 0.
– The group velocity (Vgr ) is always parallel to the magnetic field, but the phase
velocity (Vph ) can be oblique to the field. It implies that the wave vector and
the phase velocity may be non-parallel to the group velocity and the magnetic
field direction. To illustrate this, let us consider a guitar with identical strings
parallel to each other, separated by vacuum. One may produce a localised transverse
perturbation of those strings, which would develop in a form of transverse waves
propagating along the strings. The group speed is always along the direction of the
string, since it is the direction in which the energy and information are transferred
by the wave. But, the “wave front” produced by such a perturbation could be
oblique to the direction of the strings, for example, if the individual strings are
pitched at different distances from the guitar’s nut. In this case the phase speed of
the excited propagating perturbation is oblique to the strings, and hence the group
and phase velocities are not parallel to each other.
– The absolute value of the group velocity equals the Alfvén speed, CA .
In both linearly and elliptically polarised Alfvén waves, the absolute value of the magnetic
field varies in time. In circularly polarised Alfvén waves the absolute value remains
constant, while its direction varies in time and space.

43
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Example: Field line resonances in the Earth’s magnetosphere

The magnetic field of the Earth’s at small and medium distances from the ground can
be approximated as a dipole:

Above the ionosphere, the Earth’s magnetic field penetrates the magnetospheric plasma,
and large scale slow motions of the plasma are described by MHD. Different lines of
the field (or, more rigorously, magnetic surfaces shown in the figure above) can support
standing Alfvén waves. The perturbation of the electric field E in these waves has nodes
at the ionospheric footpoints of the field lines, and may have a different number of nodes
(or no nodes at all) along the magnetospheric part of the field line. Thus, the wavelengths
of these Alfvén waves are determined by the length along the field line between the
ionospheric footpoints. We can distinguish between the fundamental mode (no nodes
between the footpoints), the second harmonic mode (one node), the 3rd harmonics (two
nodes), etc.:

44
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

In this sketch the equatorial plane is in the horizontal direction, and b is the perturbation
of the magnetic field in the Alfvén wave. Mind that the perturbations of the electric field
and the magnetic field have different structures along the field line.

As the wavelength λ is fixed (e.g., for the fundamental mode it is double the length of
the magnetic field line between the ionospheric footpoints, c.f. the perturbations of a
string), the frequency of the Alfvénic oscillation is fixed too. If we assume the Alfvén
speed is constant along the field line (it can be achieved as both the magnetic field
and the plasma density decrease with height), the cyclic frequency of the oscillation is
ω = CA 2π/λ. Hence, at a certain geographical location one will detect an oscillation
with the frequency determined by this expression. This effect is the magnetospheric field
line resonance.

As at different geographical latitudes the field lines have different length, the resonant
Alfvén waves have different wave lengths and hence signals with different frequencies will
be detected. Typical periods of field line resonant Alfvénic oscillations are about one
minute.

45
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

5.3.2 Magnetoacoustic waves

The second partial set of equations is formed by equations (114), (115), (117), (118) and
(121) and describes variables Vx1 , Vz1 , Bx1 , P1 and ρ1 . The consistency condition gives
us
(ω 2 − CA2
cos2 α k 2 )(ω 2 − Cs2 k 2 ) − CA
2
sin2 α ω 2 k 2 = 0, (123)

where Cs = (γp0 /ρ0 )1/2 . These waves are called magnetoacoustic, as they are essentially
compressive since they perturb the plasma density. It can be shown with the use of the
continuity equation, that in magnetoacoustic waves generally ∇.v 6= 0. Equation (123)
is bi-quadratic with respect to ω and k, and consequently has two pairs of roots for ω 2
or k 2 . Solving it with respect to ω 2 one obtains
  2
k
q
2 2
ω = (CA + Cs2 ) 2 2 2 2 2 2
± (CA + Cs ) − 4CA Cs cos α . (124)
2

As the argument of the square root is always positive, and as the right-hand side is
always positive, equation (124) has two pairs of roots that describe oscillatory motions.
The positive sign on the right hand side corresponds to the fast magnetoacoustic wave,
and the negative sign to the slow magnetoacoustic wave.
Physical processes that drive magnetoacoustic waves are associated with gradients of the
gas pressure and magnetic pressure, as well as the magnetic tension. For illustration
consider a plane fast magnetoacoustic wave, propagating across the field. In regions of
compression of the plasma, because of the frozen-in effect, there is also an increase in
the absolute value of the magnetic field. Hence, there are gradients of both gas and
magnetic pressures, directed outwards from the compressed regions, driving the plasma
(with the frozen-in magnetic field) outwards, towards regions of rarified plasma density
and decreased field. The force that attempts to restore the equilibrium is thus the
gradient of the total (gas plus magnetic) pressure. Because of the finite inertia of the
plasma, the plasma overshoots the equilibrium, and creates new regions of enhanced and
decreased gas and magnetic pressure. Thus, gradients of the total pressure occur, moving
the plasma, etc.
The main properties of magnetoacoustic waves are as follows:

– Magnetoacoustic waves are, in general, neither transverse nor longitudinal: the

induced plasma flow has a component along the wave vector.
– The waves are essentially compressive: they always perturb the density of the
plasma.
– Magnetoacoustic waves can, in general, propagate in all directions with respect to
the magnetic field, while their properties depend strongly upon the angle and the
plasma parameter β which is the ratio of the plasma pressure (p0 ) to the magnetic
pressure (B02 /(2µ)) and so is proportional to the ratio of the squares of the sound
and Alfvén speeds.

46
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

– The fast wave propagates in a direction perpendicular to the field at the phase
and group velocities Vph and Vgr with absolute values equal to the fast speed,
2 + C 2 1/2 . In the perpendicular direction, the phase speed of a slow wave


CF ≡ CA s
goes to zero, while its parallel group speed tends to the tube (or cusp) speed,
CA Cs
CT ≡ 2
. (125)
+ Cs2 )1/2
(CA

In all other directions the phase speeds of the fast and slow waves are greater or
lower than the Alfvén speed, respectively.
– In the β < 1 case, the fast wave cannot propagate along the field — if kkB0
the fast wave becomes incompressive and purely transverse, and hence degenerates
into the Alfvén wave. In the β > 1 case, the parallel fast wave is the acoustic
wave, propagating at the sound speed. In the β < 1 case, the parallel slow wave
propagates along the field at the speed Cs and does not perturb the magnetic field,
and hence degenerates to the usual sound wave propagating along the field. For all
values of β, the slow wave cannot propagate across the field.
– For the oblique slow wave the density and the absolute value of the magnetic field
are perturbed in anti-phase, while in the fast wave they are in phase.
– In the zero-β limit which describes well a number of important natural and labora-
tory plasmas, the slow wave ceases to exist. The fast wave propagates at the Alfvén
speed in all directions (while in the parallel propagation case it degenerates into an
Alfvén wave). Because of that, in the zero-β limit, the fast magnetoacoustic wave
is often referred to as a compressional Alfvén wave.
– In the β < 1 case, the largest perturbation of the plasma in all fast waves and the
Alfvén wave is transverse to the field.

The figure shows polar plots for phase speeds (Vph ) and group velocities (Vgr ), for β < 1:

Here the equilibrium magnetic field is directed horizontally. The red curves correspond
to the fast waves, the blue to the Alfvén and the green to slow. The distance from the

47
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

origin shows the speed. To determine the phase or group speed of MHD waves in a
certain direction, one should make a straight ray from the origin in this direction. The
distances at which this ray crosses the curves of different MHD waves gives the speeds.

In general, the fast magnetoacoustic wave propagates obliquely to the magnetic field.
The “fast” speed,
2

1/2
CF = CA + Cs2 , (126)

is always higher than both the sound and Alfvén speeds.

The slow magnetoacoustic wave propagates mainly along the magnetic field, or at some
narrow angle to the field.

In ideal MHD, in a uniform medium, MHD waves of all kinds are dispersionless, i.e.,
their phase and group speeds are independent of the frequency or wavelength.

5.4 MHD Instabilities

In the Chapter on the kinetic description of plasmas we have already discussed instabil-
ities of plasmas, that occur in the velocity space. Such instabilities are usually classified
as “micro-instabilities”. They result in a great enhancement of the level of fluctuations
in the plasma associated with the unstable mode. Much more violent instabilities occur
in the configuration space: “macro-instabilities”. They cause a major reconstruction of
the plasma configuration. A spectacular example of such an instability is a coronal mass
ejection (CME) observed in the solar corona.

Snapshots of the development of a mass ejection observed in the corona of the Sun from
the ESA Solar and Heliospheric Observatory (SoHO) spacecraft. The bright regions
correspond to the dense plasma highlighted by the solar light by the effect of Thom-
son scattering. The sequence of images shows the major reconstruction of the plasma
configuration, accompanied by the emergence of a plasma blob.

As an example, consider macroscopic instability of a very common configuration of a

plasma, a plasma pinch. In the simplest case, it is a plasma cylinder of a radius a,
embedded in vacuum, with an electric current along the axis of the cylinder:

48
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

It convenient to describe this configuration in cylindrical coordinates (r, θ, z) , with the

axis z coinciding with the axis of the cylinder. Consider the case of a uniform electric
current inside the cylinder. The current density is given by the expression

(0, 0, j), r < a,
j= (127)
(0, 0, 0), r > a.

The total current inside the plasma pinch is the integral of the current density over
the cross-sectional area, I = πa2 j. This current creates the magnetic field around the
cylinder, and also inside it. In MHD, the current density and magnetic field are connected
as j = ∇ × B/µ0 . The symmetry of the problem suggests that B = (0, Bθ (r), 0). Thus,
 
1 d
j = 0, 0, (rBθ ) , (128)
µ0 r dr

where we used the expression of the curl in cylindrical coordinates. Thus, we get
(
 µ0
1 d j, r < a, 2 jr, r < a,
(rBθ ) = and, hence, Bθ = µ0 a2 (129)
µ0 r dr 0, r > a, 2r j, r > a.

This is an equilibrium plasma configuration (see Problems): the plasma is confined

(“pinched”) to the cylinder by the force created by the magnetic field, generated by
the axial current. The magnetic forces counteract the gas pressure force that tries to
spread the plasma in the radial direction. But, is such a configuration stable?
Consider an axisymmetric (“sausage”) perturbation of the equilibrium, that makes a
localised compression of the cylinder (see Fig. 11, upper panel). Let us call the region
of the localised compression a “neck”. As the electric current at each position along
the cylinder should be constant, and as the current is given by the product of the local
current density and the cross-sectional area of the cylinder, in the neck the current density
must be increased. The increased current density produces the increased magnetic field,
Bθ = µ0 aj/2. The increased magnetic field causes increased magnetic tension (as the
radius of curvature of the field in the neck is smaller) squeezes the plasma in the neck (the
frozen-in effect). The compression of the plasma causes the increase in the gas pressure.
As the gas pressure is isotropic, its increase in the neck pushes the plasma along the
axis of the cylinder along the field. The field-aligned flows remove the plasma from the

49
University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 11: Sketches of a the physical processes in a plasma pinch, that cause the sausage (the
upper panel) and kink (the bottom panel) instabilities.

50
 University of Warwick — Nakariakov: PX392 Plasma Electrodynamics 2021–2022

Figure 12: Development of kink instability of a plasma loop in the corona of the Sun, recorded
in the EUV band with the Transition Region and Coronal Explorer (TRACE).

neck. Thus there is nothing left in the neck to counteract the magnetic forces squeezing
it. The perturbation increases and is not stopped. Thus, the equilibrium is unstable to
such a perturbation, and the plasma pinch is rapidly destroyed. This effect is known as
a “sausage” instability of the plasma pinch.
Moreover, the plasma pinch is also unstable to another perturbation, that is the dis-
placement of the axis of the cylinder (“kink” perturbation, see Fig. 11, bottom panel).
The perturbation makes the magnetic field at the point M stronger (by the absolute
value that is the “density” of the magnetic field lines in the sketch) than at the point N.
Thus, the magnetic pressure at M is higher than at N, that causes a force that amplifies
the perturbation. The perturbation is unbalanced by any forces and hence destroys the
plasma configuration. This is the “kink” instability, see Fig. 12.
Other important examples of macro-instabilities of plasmas are the Kelvin–Helmholtz
instability that appears in shear plasma flows, the Rayleigh–Taylor instability that ap-
pears when a denser plasma is situated above more rarified plasma (and also when one
plasma is accelerated in another one), the Jeans instability of self-gravitating plasmas,
radiative or thermal instability that can lead to localised condensations of cool plasmas,
and many others.
A standard approach to the analysis of the stability of a plasma equilibrium requires
dispersion relation for the perturbations. The sign of the imaginary part of the frequency
shows whether the perturbation grows (the configuration is unstable) or decreases (the
configuration is stable). The combination of plasma parameters that correspond to the
imaginary part of the frequency becoming zero shows the threshold of instability. The
arrangement of the plasma parameters that make the imaginary part of the frequency
zero is known as “stabilisation” of the configuration. In particular, the plasma pinch
considered above can be stabilised by adding a magnetic field along the axis of the
cylinder.

51