Chapter 1

Introduction
1.1
1.1.1

What is a Plasma?
An ionized gas

A plasma is a gas in which an important fraction of the atoms is ionized, so that the electrons

and ions are separately free.

When does this ionization occur? When the temperature is hot enough.

Balance between collisional ionization and recombination:

Figure 1.1: Ionization and Recombination

Ionization has a threshold energy. Recombination has not but is much less probable.

Threshold is ionization energy (13.6eV, H). i

Integral over Maxwellian distribution gives rate coecients (reaction rates). Because of

the tail of the Maxwellian distribution, the ionization rate extends below T = i . And in

equilibrium, when

nions
< i v >
=
,
(1.1)
< r v >
nneutrals

Figure 1.2: Ionization and radiative recombination rate coecients for atomic hydrogen
the percentage of ions is large ( 100%) if electron temperature: Te >
i /10. e.g. Hydrogen

is ionized for Te >

1eV
(11,600
k).
At
room
temp
r
ionization
is
negligible.

For dissociation and ionization balance gure see e.g. Delcroix Plasma Physics Wiley (1965)
gure 1A.5, page 25.

1.1.2

Plasmas are QuasiNeutral

If a gas of electrons and ions (singly charged) has unequal numbers, there will be a net charge
density, .
= ne (e) + ni (+e) = e(ni ne )
(1.2)
This will give rise to an electric eld via
.E =

e
= (ni ne )
0
0

(1.3)

Example: Slab.
dE

=
dx
0
x
E =
0
7

(1.4)
(1.5)

Figure 1.3: Charged slab

This results in a force on the charges tending to expel whichever species is in excess. That
is, if ni > ne , the E eld causes ni to decrease, ne to increase tending to reduce the charge.
This restoring force is enormous!
Example
Consider Te = 1eV , ne = 1019 m3 (a modest plasma; c.f. density of atmosphere nmolecules
3 1025 m3 ). Suppose there is a small dierence in ion and electron densities n = (ni ne )
so = n e

(1.6)

Then the force per unit volume at distance x is

Fe = E = 2

x
x
= (n e)2
0
0

(1.7)

Take n/ne = 1% , x = 0.10m.

Fe = (1017 1.6 1019 )2 0.1/8.8 1012 = 3 106 N.m3

(1.8)

Compare with this the pressure force per unit volume p/x : p ne Te (+ni Ti )
Fp 1019 1.6 1019 /0.1 = 16N m3

(1.9)

Electrostatic force >> Kinetic Pressure Force.

This is one aspect of the fact that, because of being ionized, plasmas exhibit all sorts of col
lective behavior, dierent from neutral gases, mediated by the long distance electromagnetic
forces E, B.
Another example (related) is that of longitudinal waves. In a normal gas, sound waves are
propagated via the intermolecular action of collisions. In a plasma, waves can propagate
when collisions are negligible because of the coulomb interaction of the particles.
8

1.2
1.2.1

Plasma Shielding
Elementary Derivation of the Boltzmann Distribution

Basic principle of Statistical Mechanics:

Thermal Equilibrium Most Probable State i.e. State with large number of possible ar

rangements of microstates.

Figure 1.4: Statistical Systems in Thermal Contact

Consider two weakly coupled systems S1 , S2 with energies E1 , E2 . Let g1 , g2 be the number of
microscopic states which give rise to these energies, for each system. Then the total number
of microstates of the combined system is (assuming states are independent)
g = g1 g2

(1.10)

If the total energy of combined system is xed E1 + E2 = Et then this can be written as a
function of E1 :

and

g = g1 (E1 )g2 (Et E1 )

dg
dg1
dg2
g2 g1
.
=
dE
dE
dE1

The most probable state is that for which

dg
dE1

(1.11)
(1.12)

= 0 i.e.

1 dg1
1 dg2
d
d
=
or
ln g1 =
ln g2
g1 dE
g2 dE
dE
dE
Thus, in equilibrium, states in thermal contact have equal values of

(1.13)
d
dE

ln g.

One denes ln g as the Entropy.

d
And [ dE
ln g]1 = T the Temperature.

Now suppose that we want to know the relative probability of 2 microstates of system 1 in
equilibrium. There are, in all, g1 of these states, for each specic E1 but we want to know
how many states of the combined system correspond to a single microstate of S1 .
Obviously that is just equal to the number of states of system 2. So, denoting the two values
of the energies of S1 for the two microstates we are comparing by EA , EB the ratio of the
number of combined system states for S1A and S1B is
g2 (Et EA )
= exp[(Et EA ) (Et EB )]
g2 (Et EB )
9

(1.14)

Now we suppose that system S2 is large compared with S1 so that EA and EB represent very
small changes in S2 s energy, and we can Taylor expand

g2 (Et EA )
d
d
exp EA
+ EB
g2 (Et EA )
dE
dE

(1.15)

Thus we have shown that the ratio of the probability of a system (S1 ) being in any two
microstates A, B is simply

(EA EB )
,
(1.16)
exp
T
when in equilibrium with a (large) thermal reservoir. This is the wellknown Boltzmann
factor.
You may notice that Boltzmanns constant is absent from this formula. That is because of
using natural thermodynamic units for entropy (dimensionless) and temperature (energy).
Boltzmanns constant is simply a conversion factor between the natural units of temperature
(energy, e.g. Joules) and (e.g.) degrees Kelvin. Kelvins are based on C which arbitrarily
choose melting and boiling points of water and divide into 100.
Plasma physics is done almost always using energy units for temperature. Because Joules
are very large, usually electronvolts (eV) are used.
1eV = 11600K = 1.6 1019 Joules.

(1.17)

One consequence of our Botzmann factor is that a gas of moving particles whose energy is
2
1
mv 2 adopts the MaxwellBoltzmann (Maxwellian) distribution of velocities exp[ mv
].
2
2T

1.2.2

Plasma Density in Electrostatic Potential

When there is a varying potential, , the densities of electrons (and ions) is aected by it.
If electrons are in thermal equilibrium, they will adopt a Boltzmann distribution of density
ne exp(

e
) .
Te

(1.18)

This is because each electron, regardless of velocity possesses a potential energy e.

Consequence is that (g 1.5) a selfconsistent loop of dependencies occurs.
This is one elementary example of the general principle of plasmas requiring a selfconsistent
solution of Maxwells equations of electrodynamics plus the particle dynamics of the plasma.

1.2.3

Debye Shielding

A slightly dierent approach to discussing quasineutrality leads to the important quantity

called the Debye Length.
Suppose we put a plane grid into a plasma, held at a certain potential, g .
10

Figure 1.5: Selfconsistent loop of dependencies

Figure 1.6: Shielding of elds from a 1D grid.

Then, unlike the vacuum case, the perturbation to the potential falls o rather rapidly into
the plasma. We can show this as follows. The important equations are:
Poisson s Equation
Electron Density

d2
e
=
(ni ne )

dx2
0
ne = n exp(e/Te ).
2 =

(1.19)
(1.20)

[This is a Boltzmann factor; it assumes that electrons are in thermal equilibrium. n is

density far from the grid (where we take = 0).]
Ion Density

ni = n .

(1.21)

[Applies far from grid by quasineutrality; we just assume, for the sake of this illustrative
calculation that ion density is not perturbed by perturbation.]
Substitute:

d2
en
e
=
exp
1 .
dx2
0
Te

(1.22)

This is a nasty nonlinear equation, but far from the grid |e/Te | << 1 so we can use a Taylor
expression: exp e
1 + e
. So
Te
Te
d2
e2 n
en e

=
dx2
0 Te
0 Te

11

(1.23)

Solutions: = 0 exp(|x|/D ) where

0 Te
D 2
e n

12

(1.24)

This is called the Debye Length

Perturbations to the charge density and potential in a plasma tend to fall o with charac

teristic length D .

In Fusion plasmas D is typically small. [e.g. ne = 1020 m3 Te = 1keV

20m]

D = 2 105 m =

Usually we include as part of the denition of a plasma that D << the size of plasma. This

ensures that collective eects, quasineutrality etc. are important. Otherwise they probably

arent.

1.2.4

PlasmaSolid Boundaries (Elementary)

When a plasma is in contact with a solid, the solid acts as a sink draining away the plasma.
Recombination of electrons and ions occur at surface. Then:
1. Plasma is normally charged positively with respect to the solid.

Figure 1.7: PlasmaSolid interface: Sheath

2. There is a relatively thin region called the sheath, at the boundary of the plasma,
where the main potential variation occurs.
Reason for potential drop:
Dierent velocities of electrons and ions.
If there were no potential variation (E= 0) the electrons and ions would hit the surface at
the random rate
1
nv
per unit area
(1.25)
4

12

[This equation comes from elementary gaskinetic theory. See problems if not familiar.]
The mean speed v =

8T
m

T
.
m

mi
faster and hence would drain out of plasma
Because of mass dierence electrons move m
e
faster. Hence, plasma charges up enough that an electric eld opposes electron escape and

reduces total electric current to zero.

Estimate of potential:

Ion escape ux
Electron escape ux

1
n v
4 i i
1
n v
4 e i

Prime denotes values at solid surface.

Boltzmann factor applied to electrons:
ne = n exp[es /Te ]

(1.26)

where s is solid potential relative to distant () plasma.

Since ions are being dragged out by potential assume ni n (Zi = 1). [This is only

approximately correct.]

Hence total current density out of plasma is

1
1
j = qi ni vi + qe ne ve
4
4

en
es
{
vi exp
=
ve }
4
Te

(1.27)
(1.28)

This must be zero so

Te
Ti me
vi
Te 1
ln
=
ln | | =
e
ve
e 2
Te mi

me
Te 1
[if Te = Ti .]
=
ln
e 2
mi

For hydrogen

mi
me

= 1800 so

1
2

Te
e

1.2.5

(1.30)

e
ln m
= 3.75.
mi

The potential of the surface relative to plasma is approximately 4

[Note

(1.29)

Te
.
e

is just the electron temp r in electronvolts expressed as a voltage.]

Thickness of the sheath

Crude estimates of sheath thickness can be obtained by assuming that ion density is uniform.
Then equation of potential is, as before,
d2
en
=
2
dx
0

e
exp
1
Te

13

(1.31)

We know the rough scalelength of solutions of this equation is

0 Te
D = 2
e n

21

the Debye Length.

(1.32)

Actually our previous solution was valid only for |e/Te | << 1 which is no longer valid.
When e/Te > 1 (as will be the case in the sheath). We can practically ignore the electron
density, in which case the solution will continue only quadratically. One might expect,
therefore, that the sheath thickness is roughly given by an electric potential gradient

T 1
e D

(1.33)

extending sucient distance to reach S = 4 Tee i.e.

distance x 4D
This is correct for the typical sheath thickness but not at all rigorous.

1.3

The Plasma Parameter

Notice that in our development of Debye shielding we used ne e as the charge density and
supposed that it could be taken as smooth and continuous. However if the density were so
low that there were less than approximately one electron in the Debye shielding region this
approach would not be valid. Actually we have to address this problem in 3d by dening
the Plasma Parameter, ND , as
ND = Number of particles in the Debye Sphere.

4
= n. 3D
3

T2
1

n2

(1.34)

If ND <
1 then the individual particles cannot be treated as a smooth continuum. It will be
seen later that this means that collisions dominate the behaviour: i.e. short range correlation
is just as important as the long range collective eects.
Often, therefore we add a further qualication of plasma:
ND >> 1 (Collective eects dominate over collisions)

1.4

(1.35)

Summary

Plasma is an ionized gas in which collective eects dominate over collisions.

[D << size ,
14

ND >> 1 .]

(1.36)

1.5

Occurrence of Plasmas

Gas Discharges:

Fluorescent Lights, Spark gaps, arcs, welding, lighting

Controlled Fusion

Ionosphere:

Ionized belt surrounding earth

Interplanetary Medium:
Stellar Astrophysics:
Ion Propulsion:

Stars. Pulsars. Radiationprocesses.

Advanced space drives, etc.

& Space Technology

Gas Lasers:

Magnetospheres of planets and starts. Solar Wind.

Interaction of Spacecraft with environment

Plasma discharge pumped lasers: CO2 , He, Ne, HCN.

Materials Processing:

Surface treatment for hardening. Crystal Growing.

Semiconductor Processing: Ion beam doping, plasma etching & sputtering.

Solid State Plasmas:

Behavior of semiconductors.

For a gure locating dierent types of plasma in the plane of density versus temperature see

for example Goldston and Rutherford Introduction to Plasma Physics IOP Publishing, 1995,

gure 1.3 page 9. Another is at http://www.plasmas.org/basics.htm

1.6

Dierent Descriptions of Plasma

1. Single Particle Approach. (Incomplete in itself). Eq. of Motion.

2. Kinetic Theory. Boltzmann Equation.

+ v.
+ a.
f =
t
t
x
v

(1.37)
col.

3. Fluid Description. Moments, Velocity, Pressure, Currents, etc.

Uses of these.
Single Particle Solutions Orbits
Kinetic Theory Solutions Transport Coefs.
Fluid Theory Macroscopic Description
All descriptions should be consistent. Sometimes they are dierent ways of looking at the
same thing.

15

1.6.1

Equations of Plasma Physics

.E =

E =

.B = 0

B
t

B = 0 j +

1 E
c2 t

(1.38)

F = q(E + v B)

1.6.2

Self Consistency

In solving plasma problems one usually has a circular system:

The problem is solved only when we have a model in which all parts are self consistent. We

need a bootstrap procedure.

Generally we have to do it in stages:

Calculate Plasma Response (to given E,B)

Get currents & charge densities
Calculate E & B for j, p.
Then put it all together.

16

Chapter 2

Motion of Charged Particles in Fields

Plasmas are complicated because motions of electrons and ions are determined by the electric

and magnetic elds but also change the elds by the currents they carry.

For now we shall ignore the second part of the problem and assume that Fields are Prescribed.

Even so, calculating the motion of a charged particle can be quite hard.

Equation of motion:

dv
v
B )
= q ( E +
dt
charge Eeld velocity Beld

Rate of change of momentum

Lorentz Force
m

(2.1)

Have to solve this dierential equation, to get position r and velocity (v= r ) given E(r, t), B(r, t).
Approach: Start simple, gradually generalize.

2.1

Uniform B eld, E = 0.
mv = qv B

2.1.1

(2.2)

Qualitatively

in the plane perpendicular to B: Accel. is perp to v so particle moves in a circle whose

radius rL is such as to satisfy
mrL 2 = m

2
v
= |q |v B
rL

is the angular (velocity) frequency

2
1st equality shows 2 = v
/rL2 (rL = v /)

17

(2.3)

Figure 2.1: Circular orbit in uniform magnetic eld.

Hence second gives m v 2 = |q |v B
|q|B
.
m

i.e. =

(2.4)

Particle moves in a circular orbit with

|q |B
the Cyclotron Frequency
m
v
and radius rl =
the Larmor Radius.

angular velocity =

2.1.2

(2.5)
(2.6)

By Vector Algebra

Particle Energy is constant. proof : take v. Eq. of motion then

d 1 2
mv.v =
mv
dt 2

= qv.(v B) = 0.

(2.7)

Parallel and Perpendicular motions separate. v = constant because accel ( v B)

is perpendicular to B.
Perpendicular Dynamics:
Take B in z direction and write components
mv x = qvy B , mv y = qvx B
Hence
qB
qB
v y =
m
m

vx =
Solution: vx = v cos t

vx = 2 vx

(2.8)

(2.9)

(choose zero of time)

Substitute back:
vy =

m
|q |
v x = v sin t
qB
q
18

(2.10)

Integrate:
x = x0 +

v
sin t ,

y = y0 +

q v
cos t
|q|

(2.11)

Figure 2.2: Gyro center (x0 , y0 ) and orbit

This is the equation of a circle with center r0 = (x0 , y0 ) and radius rL = v /: Gyro Radius.

[Angle is = t]

Direction of rotation is as indicated opposite for opposite sign of charge:

Ions rotate anticlockwise. Electrons clockwise about the magnetic eld.

The current carried by the plasma always is in such a direction as to reduce the magnetic

eld.

This is the property of a magnetic material which is Diagmagnetic.

When v is nonzero the total motion is along a helix.

2.2

Uniform B and nonzero E

mv = q(E + v B)

(2.12)

Parallel motion: Before, when E = 0 this was v = const. Now it is clearly

v =

qE
m

(2.13)

Constant acceleration along the eld.

Perpendicular Motion

Qualitatively:

Speed of positive particle is greater at top than bottom so radius of curvature is greater.

Result is that guiding center moves perpendicular to both E and B. It drifts across the

eld.

Algebraically: It is clear that if we can nd a constant velocity vd that satises

E + vd B = 0
19

(2.14)

Figure 2.3: E B drift orbit

then the sum of this drift velocity plus the velocity
vL =

d
[rL ei(tt0 ) ]
dt

(2.15)

which we calculated for the E = 0 gyration will satisfy the equation of motion.
Take B the above equation:
0 = E B + (vd B) B = E B + (vd .B)B B 2 vd
so that
vd =

EB
B2

(2.16)

(2.17)

does satisfy it.

Hence the full solution is

v=

v +
vd
+ vL
parallel crosseld drift Gyration

where
v =

qE
m

(2.18)

(2.19)

and
vd (eq 2.17) is the E B drift of the gyrocenter.
Comments on E B drift:
1. It is independent of the properties of the drifting particle (q, m, v, whatever).
2. Hence it is in the same direction for electrons and ions.
3. Underlying physics for this is that in the frame moving at the E B drift E = 0. We
have transformed away the electric eld.
4. Formula given above is exact except for the fact that relativistic eects have been
ignored. They would be important if vd c.
20

2.2.1

Drift due to Gravity or other Forces

Suppose particle is subject to some other force, such as gravity. Write it F so that
1
mv = F + q v B = q( F + v B)
q

(2.20)

This is just like the Electric eld case except with F/q replacing E.
The drift is therefore
vd =

1FB
q B2

(2.21)

In this case, if force on electrons and ions is same, they drift in opposite directions.

This general formula can be used to get the drift velocity in some other cases of interest (see

later).

2.3

NonUniform B Field

If Blines are straight but the magnitude of B varies in space we get orbits that look quali
tatively similar to the E B case:

Figure 2.4: B drift orbit

Curvature of orbit is greater where B is greater causing loop to be small on that side. Result
is a drift perpendicular to both B and B. Notice, though, that electrons and ions go in
opposite directions (unlike E B).
Algebra
We try to nd a decomposition of the velocity as before into v = vd + vL where vd is
constant.

We shall nd that this can be done only approximately. Also we must have a simple expres

sion for B. This we get by assuming that the Larmor radius is much smaller than the scale

length of B variation i.e.,

rL << B/|B |
(2.22)
21

in which case we can express the eld approximately as the rst two terms in a Taylor
expression:
B B0 + (r.)B
(2.23)
Then substituting the decomposed velocity we get:
dv
= mv L = q(v B) = q[vL B0 + vd B0 + (vL + vd ) (r.)B]
dt
or 0 = vd B0 + vL (r.)B + vd (r.)B
m

(2.24)
(2.25)

Now we shall nd that vd /vL is also small, like r|B |/B. Therefore the last term here is
second order but the rst two are rst order. So we drop the last term.
Now the awkward part is that vL and rL are periodic. Substitute for r = r0 + rL so
0 = vd B0 + vL (rL .)B + vL (r0 .)B

(2.26)

We now average over a cyclotron period. The last term is eit so it averages to zero:
0 = vd B + vL (rL .)B .

(2.27)

To perform the average use

So

v
q
rL = (xL , yL ) =
sin t,
cos t

|q |

q
vL = (x L , y L ) = v cos t,
sin t
|q |
d
[vL (r.)B]x = vy y B
dy
d
[vL (r.)B]y = vx y B
dy

(2.28)
(2.29)
(2.30)
(2.31)

(Taking B to be in the ydirection).

Then
v2
vy y = cos t sin t
= 0

2
q
v2
1 v
q
vx y =
cos t cos t =
|q |

2 |q |
So
vL (r.)B =
Substitute in:
0 = vd B
22

2
q 1 v
B
|q | 2

2
q v

B
|q
| 2

(2.32)

(2.33)

(2.34)

(2.35)

and solve as before to get

vd =

2
1 v
|q| 2

B B
=

B2

2
q v

B B
|q
| 2 B 2

(2.36)

or equivalently
vd =

2
B B
1 mv
B2
q 2B

(2.37)

This is called the Grad B drift.

2.4

Curvature Drift

When the Beld lines are curved and the particle has a velocity v along the eld, another
drift occurs.

Figure 2.5: Curvature and Centrifugal Force

Take |B | constant; radius of curvature Re .
To 1st order the particle just spirals along the eld.
In the frame of the guiding center a force appears because the plasma is rotating about the

center of curvature.

This centrifugal force is Fcf

Fcf = m

v2
pointing outward
Rc

as a vector
Fcf = mv
2

Rc
Rc2

(2.38)

(2.39)

[There is also a coriolis force 2m( v) but this averages to zero over a gyroperiod.]
Use the previous formula for a force
vd =

mv2 Rc B
1 Fcf B
=

q B2
qB 2 Rc2
23

(2.40)

This is the Curvature Drift.

It is often convenient to have this expressed in terms of the eld gradients. So we relate Rc

to B etc. as follows:

Figure 2.6: Dierential expression of curvature

(Carets denote unit vectors)
From the diagram
2 b
1 = R
c
db = b

(2.41)

d = Rc

(2.42)

c
Rc
db
R
=
= 2
Rc
Rc
dl

(2.43)

db

= (B.)
b
dl

(2.44)

and
So

But (by denition)

So the curvature drift can be written

vd =

2.4.1

mv2 Rc
mv2 B (b.)
B
b

=
2
2
2
q Rc B
q
B

(2.45)

Vacuum Fields

Relation between B & Rc drifts

The curvature and B are related because of Maxwells equations, their relation depends
on the current density j. A particular case of interest is j = 0: vacuum elds.

Figure 2.7: Local polar coordinates in a vacuum eld

B=0

(static case)
24

(2.46)

Consider the zcomponent

1
(rB ) (Br = 0 by choice).
r r
B
+
r

0 = ( B)z =
=

B
r

(2.47)
(2.48)

or, in other words,

(B)r =

B
Rc

(2.49)

[Note also 0 = ( B) = B /z : (B)z = 0]

and hence (B)perp = B Rc /Rc2 .
Thus the grad B drift can be written:
vB =

2
mv
B B
mV2 Rc B
=
2q
B3
2q Rc2 B 2

(2.50)

and the total drift across a vacuum eld becomes

1 2 Rc B
1
mv2 + mv
.
q
2
Rc2 B 2

vR + vB =

(2.51)

Notice the following:

1. Rc & B drifts are in the same direction.
2. They are in opposite directions for opposite charges.
3. They are proportional to particle energies
4. Curvature Parallel Energy ( 2)
B Perpendicular Energy
5. As a result one can very quickly calculate the average drift over a thermal distribution
of particles because
T
1
mv2 =
2
2
1 2
mv
= T
2
Therefore

(2.52)
2 degrees of freedom

(2.53)

b
2T Rc B 2T B b.

vR + vB =
=
q Rc2 B 2
q
B2

25

(2.54)

2.5 Interlude: Toroidal Connement of Single Parti

cles
Since particles can move freely along a magnetic eld even if not across it, we cannot ob
viously conne the particles in a straight magnetic eld. Obvious idea: bend the eld lines
into circles so that they have no ends.

Figure 2.8: Toroidal eld geometry

Problem
Curvature & B drifts
1
vd =
mv2 +
q

1
|vd | =
mv2 +
q

1
2 R B
mv

2
R2 B 2

1
2
1
mv

2
BR

(2.55)

(2.56)

Ions drift up. Electrons down. There is no connement. When there is nite density things

Figure 2.9: Charge separation due to vertical drift

are even worse because charge separation occurs E E B Outward Motion.

2.5.1 How to solve this problem?

Consider a beam of electrons v =
0 v = 0. Drift is
mv2 1
vd =
q BT R
26

(2.57)

What Bz is required to cancel this?

Adding Bz gives a compensating vertical velocity

v = v
We want total

Bz
BT

for Bz << BT

mv2 q
Bz
vz = 0 = v
+
BT
q BT R

(2.58)

(2.59)

So Bz = mv /Rq is the right amount of eld.

Note that this is such as to make
rL (Bz ) =

|mv |
=R .
|qBz |

(2.60)

But Bz required depends on v and q so we cant compensate for all particles simultaneously.
Vertical eld along cannot do it.

2.5.2

The Solution: Rotational Transform

Figure 2.10: Tokamak eld lines with rotational transform

Toroidal Coordinate system (r, , ) (minor radius, poloidal angle, toroidal angle), see gure
2.8.

Suppose we have a poloidal eld B

Field Lines become helical and wind around the torus: gure 2.10.

27

In the poloidal crosssection the eld describes a circle as it goes round in .

Equation of motion of a particle exactly following the eld is:
r

d
B
B B
B
=
v =
v =
v
dt
B
B B
B

(2.61)

r = constant.

(2.62)

and
z direction.
Now add on to this motion the cross eld drift in the

Figure 2.11: Components of velocity

d
B
=
v + vd cos
dt
B
dr
= vd sin
dt

(2.63)
(2.64)

Take ratio, to eliminate time:

1 dr
=
r d

ud sin
+ vd cos

B
v
B

(2.65)

Take B , B, v , vd to be constants, then we can integrate this orbit equation:

[ln r] = [ ln |

B v
+ vd cos |] .
B

(2.66)

Take r = r0 when cos = 0 ( = 2 ) then

Bvd
r = r0 / 1 +
cos
b v
If

Bvd
B v

(2.67)

<< 1 this is approximately

r = r0 cos

where =

Bvd
r
B v 0

This is approximately a circular orbit shifted by a distance :

28

(2.68)

Figure 2.12: Shifted, approximately circular orbit

Substitute for vd
1
2
) 1
B 1 (mv2 +
2 mv

r0
B q
v
B R
1
2
2
1 mv
+ 2 mv rp

qB
v
R
mv r0
r0
If v = 0
=
= rL ,
qB R
R

(2.69)
(2.70)
(2.71)

where rL is the Larmor Radius in a eld B r/R.

Provided is small, particles will be conned. Obviously the important thing is the poloidal

rotation of the eld lines: Rotational Transform.

Rotational Transform
poloidal angle
1 toroidal rotation
poloidal angle
(transform/2 =)
.
toroidal angle

rotational transform

(2.72)
(2.73)

(Originally, was used to denote the transform. Since about 1990 it has been used to denote
the transform divided by 2 which is the inverse of the safety factor.)
Safety Factor
qs =

1
toroidal angle
.
=

poloidal angle

(2.74)

Actually the value of these ratios may vary as one moves around the magnetic eld. Denition

strictly requires one should take the limit of a large no. of rotations.

qs is a topological number: number of rotations the long way per rotation the short way.

Cylindrical approx.:

qs =

rB
RB

29

(2.75)

In terms of safety factor the orbit shift can be written

|| = rL

r
B r
= rL
= r L qs
R
B R

(2.76)

(assuming B >> B ).

2.6 The Mirror Eect of Parallel Field Gradients: E =

0, B B

Figure 2.13: Basis of parallel mirror force

In the above situation there is a net force along B.
Force is
< F > = |qv B| sin = |q|v B sin
Br
sin =
B

(2.77)
(2.78)

Calculate Br as function of Bz from .B = 0.

.B =

(rBr ) + Bz = 0 .
r r
z

Hence
rBr =
Suppose rL is small enough that

Bz
z

[rBr ]r0L
So

Bz
dr
z

(2.79)

(2.80)

const.

rL

rdr

1 Bz
Bz
= rL2
z
z
2

1 Bz
Br (rL ) = rL
2 z
rL 1 Bz
Br
=+
sin =
B
2 2 z

30

(2.81)

(2.82)
(2.83)

Hence

1
mv 2 Bz
v rL Bz
.
(2.84)
= 2
B z
2 z
As particle enters increasing eld region it experiences a net parallel retarding force.

< F >= |q|

Dene Magnetic Moment

1 2
/B .
mv
2
Note this is consistent with loop current denition
= AI = r
L2 .

|q |rL v
|q |v
=
2rL
2

(2.85)

(2.86)

Force is F = . B
This is force on a magnetic dipole of moment .
F = . B

(2.87)

Our always points along B but in opposite direction.

2.6.1

Force on an Elementary Magnetic Moment Circuit

Consider a plane rectangular circuit carrying current I having elementary area dxdy = dA.
Regard this as a vector pointing in the z direction dA. The force on this circuit in a eld
B(r) is F such that
Bz
x
Bz
= Idx[Bz (y + dy) Bz (y)] = Idydx
y
= Idx[By (y + dy) By (y)] Idy[Bx (x + dx) Bx (x)]

Bz
Bx By
= Idxdy
+
= Idydx
x
y
z

Fx = Idy[Bz (x + dx) Bz (x)] = Idydx

(2.88)

Fy

(2.89)

Fz

(2.90)
(2.91)

(Using .B = 0).
Hence, summarizing: F = IdydxBz . Now dene = IdA = Idydx
z and take it constant.
Then clearly the force can be written
F = (B.)

[Strictly = (B).]

(2.92)

is the (vector) magnetic moment of the circuit.

The shape of the circuit does not matter since any circuit can be considered to be composed

of the sum of many rectangular circuits. So in general

= IdA
31

(2.93)

and force is
F = (B.)

( constant),

(2.94)

We shall show in a moment that || is constant for a circulating particle, regard as an

elementary circuit. Also, for a particle always points in the B direction. [Note that this
means that the eect of particles on the eld is to decrease it.] Hence the force may be
written
F = B
(2.95)
This gives us both:
Magnetic Mirror Force:
F = B

(2.96)

and
Grad B Drift:
vB =

2.6.2

1 FB
B B
=
.
2
q B
q B2

(2.97)

is a constant of the motion

Adiabatic Invariant
Proof from F
Parallel equation of motion
m
So
mv
or

dv
dB
= F =
dt
dz

dv
d 1
dB
dB
= ( mv2 ) = vz
=
dt
dt 2
dz
dt
d 1 2
dB
( mv ) +
= 0 .
dt 2
dt

(2.98)

(2.99)
(2.100)

Conservation of Total KE
d 1 2 1 2
( mv + mv ) = 0
dt 2 2
d 1 2
=
( mv + B) = 0
dt 2
Combine

d
dB
(B)
= 0
dt
dt
d
=
=0
As required
dt
32

(2.101)
(2.102)

(2.103)
(2.104)

Angular Momentum
of particle about the guiding center is
2
mv
2m 12 mv
mv =
|q|B
|q | B
2m
=
.
|q |

rL mv =

(2.105)
(2.106)

Conservation of magnetic moment is basically conservation of angular momentum about the

guiding center.
Proof direct from Angular Momentum
Consider angular momentum about G.C. Because is ignorable (locally) Canonical angular
momentum is conserved.
p = [r (mv + qA)]z

conserved.

(2.107)

Here A is the vector potential such that B = A

the denition of the vector potential means that
Bz =
rL A (rL ) =

1 (rA ))
r
r

rL

r.Bz dr =

(2.108)
rL2
m
Bz =
2
|q |

(2.109)

Hence
q
m
rL v m + q
|q |
|q |
q
=
m.
|q|

p =

(2.110)
(2.111)

So p = const = constant.

Conservation of is basically conservation of angular momentum of particle about G.C.

2.6.3

Mirror Trapping

F may be enough to reect particles back. But may not!

Lets calculate whether it will:

Suppose reection occurs.

At reection point vr = 0.

Energy conservation

1 2
1
2
2
m(v0
+ v0
) = mvr
2
2
33

(2.112)

Figure 2.14: Magnetic Mirror

conservation
1
2
mv0
2

B0

1
2
mvr
2

(2.113)

Br

Hence
Br 2
v
B0 0
v2
= 2 0 2
v0 + vo

2
2
v0
+ v0
=

B0
Br

2.6.4 Pitch Angle

(2.114)
(2.115)

v
v

tan =

(2.116)

2
v0
B0
= 2
= sin2 0
2
Br
v0 + v0

(2.117)

So, given a pitch angle 0 , reection takes place where B0 /Br = sin2 0 .
If 0 is too small no reection can occur.
Critical angle c is obviously

c = sin1 (B0 /B1 ) 2

(2.118)

Loss Cone is all < c .

Importance of Mirror Ratio: Rm = B1 /B0 .

2.6.5

Other Features of Mirror Motions

Flux enclosed by gyro orbit is constant.

=

rL2 B

2
m2 v
= 2 2 B
q B

34

(2.119)

Figure 2.15: Critical angle c divides velocity space into a losscone and a region of mirror
trapping
2
2m 12 mv
=
q2
B
2m
=
= constant.
q2

(2.120)
(2.121)

Note that if B changes suddenly might not be conserved.

Figure 2.16: Flux tube described by orbit

Basic requirement
rL << B/|B |

(2.122)

Slow variation of B (relative to rL ).

2.7

Time Varying B Field

(E inductive)

Particle can gain energy from the inductive E eld

B
t

or E.dl = B.ds
=
dt
s
E =

(2.123)
(2.124)

Hence work done on particle in 1 revolution is

w =

|q |E.d = +|q|

2
B.ds
= +|q |
= |q|Br
L
dt
s

35

(2.125)

Figure 2.17: Particle orbits round B so as to perform a line integral of the Electric eld
(d and v q are in opposition directions).

1 2
mv
2

2 =
= |q|Br
L
=

Hence

2 B
.
||

(2.127)

(2.128)

d
d 1 2
mv =
(B) .
dt 2
dt

(2.129)

d
= 0.
dt

(2.130)

but also

Hence

2m
,
q2

(2.126)

|| 1 2
db
d 1 2

mv =
mv =
2
dt
dt 2
2

Notice that since =

orbit is conserved.

2
1 mv
2 Bm
2
|q|B
B

this is just another way of saying that the ux through the gyro

Notice also energy increase. Method of heating. Adiabatic Compression.

2.8

Time Varying Eeld

(E, B uniform)

Recall the E B drift:

EB
B2
when E varies so does vEB . Thus the guiding centre experiences an acceleration
vEB =

v EB

d
=
dt

EB
B2

(2.131)

(2.132)

In the frame of the guiding centre which is accelerating, a force is felt.

d
Fa = m
dt

EB
B2

(Pushed back into seat! ve.)

36

(2.133)

This force produces another drift

1 Fa B
m d EB
=
=
B
q B2
qB 2 dt
B2

m d
=
(E.B) B B 2 E
qB dt
m
=
E
qB 2

vD

(2.134)
(2.135)
(2.136)

This is called the polarization drift.

EB
m
E
+
B2
qB 2
EB
1
=
E
+
2
B
B

vD = vEB + vp =

(2.137)
(2.138)

Figure 2.18: Suddenly turning on an electric eld causes a shift of the gyrocenter in the
direction of force. This is the polarization drift.
Startup eect: When we switch on an electric eld the average position (gyro center) of
an initially stationary particle shifts over by 12 the orbit size. The polarization drift is this
polarization eect on the medium.
Total shift due to vp is
r

2.8.1

m
m
E dt =
[E ]
2
qB
qB 2

vp dt =

Direct Derivation of

dE
dt

(2.139)

eect: Polarization Drift

Consider an oscillatory eld E = Eeit ( r0 B)

m

dv
= q (E + v B)
dt

= q Eeit + v B

(2.140)
(2.141)

Try for a solution in the form

v = vD eit + vL
37

(2.142)

where, as usual, vL satises mv L = qvL B

Then

Solve for vD :

xit

m(ivD = q (E + vD B)

(1)

(2.143)

Take B this equation:

(2)

mi (vD B) = q E B + B2 .v|D B B 2 vD

(2.144)

add mi (1) to q (2) to eliminate vD B.

m2 2 vD + q 2 (E B B 2 vD ) = miqE

or :

vD

mi
EB
m2 2
1 2 2 = 2E+
q B
qB
B2

(2.146)

iq
EB
2
E+
1 2 =
B|q|
B2

(2.147)

i.e.

vD

(2.145)

this is the same formula as we had before: the sum of polarization and
Since i t
E B drifts except for the [1 2 2 ] term.

This term comes from the change in vD with time (accel).

Thus our earlier expression was only approximate. A good approx if << .

2.9

Non Uniform E
m

(Finite Larmor Radius)

dv
= q (E(r) + v B)
dt

(2.148)

Seek the usual soltuion v = vD + vg .

Then average out over a gyro orbit
m

dvD
= 0 = q (E(r) + v B)
dt
= q [E(r) + vD B]

(2.149)
(2.150)

Hence drift is obviously

vD =

E(r) B
B2

(2.151)

So we just need to nd the average E eld experienced.

Expand E as a Taylor series about the G.C.
x2 2
y2 2
E + cross terms + .
E(r) = E0 + (r.) E +
+
2!x2
2! y 2

38

(2.152)

(E.g. cross terms are xy xy

E).

Average over a gyro orbit: r = rL (cos , sin , 0).

Average of cross terms = 0.

Then

E(r) = E + (rL .)E +

rL2 2
E.
2!

(2.153)

linear term rL = 0. So
rL2 2
E
4
Hence E B with 1st niteLarmorradius correction is
E(r) E +

r2
EB
= 1 + L 2
.
r
B2

vEB

(2.154)

(2.155)

[Note: Grad B drift is a nite Larmor eect already.]

Second and Third Adiabatic Invariants
There are additional approximately conserved quantities like in some geometries.

2.10

Summary of Drifts
vE =
vF =
vE =
vB =
vR =
vR + vB =
vp =

EB
Electric Field
B2
1FB
General Force
q B2

rL2 2 E B
1+
Nonuniform E
B2
4
2
mv
B B

GradB
2q
B3
mv2 Rc B
Curvature
q Rc2 B 2

1
1 2 Rc B
2
mv
+ mv
Vacuum Fields.
q
2
Rc2 B 2
q E

Polarization
|q | ||B

(2.156)
(2.157)
(2.158)
(2.159)
(2.160)
(2.161)
(2.162)

Mirror Motion

2
mv
2B

is constant

Force is F = B.
39

(2.163)

Chapter 3
Collisions in Plasmas
3.1

Binary collisions between charged particles

Reduced-mass for binary collisions:

Two particles interacting with each other have forces
F12 force on 1 from 2.
F21 force on 2 from 1.
By Newtons 3rd law, F12 = F21 .
Equations of motion:
m1r1 = F12

m2r2 = F21

(3.1)

Combine to get
r1 r2 = F12

1
1
+
m1 m2

(3.2)

which may be written

m1 m2 d2
(r1 r2 ) = F12
(3.3)
m1 + m2 dt2
If F12 depends only on the difference vector r1 r2 , then this equation is identical to the
m2
moving at position r r1 r2 with
equation of a particle of Reduced Mass mr mm11+m
2
respect to a fixed center of force:
mr r = F12 (r) .
(3.4)
This is the equation we analyse, but actually particle 2 does move. And we need to recognize
that when interpreting mathematics.
If F21 and r1 r2 are always parallel, then a general form of the trajectory can be written
as an integral. To save time we specialize immediately to the Coulomb force
F12 =

q1 q2 r
40 r3

Solution of this standard (Newtons) problem:

40

(3.5)

1: projectile
v1

r
b Impact Parameter
1
2: target
Figure 3.1: Geometry of the collision orbit
Angular momentum is conserved:
mr r2 = const. = mr bv1
Substitute u

1
r

then =

bv1
r2

( clockwise from symmetry)

(3.6)

= u2 bv1

Also
r

d 1
1 du
du
= 2
= bv1
dt u
u d
d
2
d2 u
d
u
bv1 2 = (bv1 )2 u2 2
d
d

(3.7)
(3.8)

Then radial acceleration is

2

r r = (bv1 ) u

d2 u
+u
d2

|F12 |
mr

(3.9)

i.e.

d2 u
q1 q2
1
+u=
2
d
40 mr (bv1 )2
This orbit equation has the elementary solution
u

(3.10)

1
q1 q2
1
= C cos
r
40 mr (bv1 )2

(3.11)

The sin term is absent by symmetry. The other constant of integration, C, must be deter1
= 0. So
mined by initial condition. At initial (far distant) angle, 1 , u1 =
0 = C cos 1
There:
r1 = v1 = bv1
Hence
tan 1 =

sin 1
=
cos 1

q1 q 2
1
40 mr (bv1 )2

(3.12)

du
|1 = +bv1 C sin 1
d
1/Cb

q1 q2
1
/C
40 mr (bv1 )2

41

b
b90

(3.13)

(3.14)

where
b90

q1 q2
1
40 mr v12

(3.15)

Notice that tan 1 = 1 when b = b90 . This is when 1 = 45 and = 90 . So particle

emerges at 90 to initial direction when
90 impact parameter00

b = b90

(3.16)

Finally:
1
1
b2
C = cosec1 =
1 + 90
b
b
b2

3.1.1

!1
2

(3.17)

Frames of Reference

Key quantity we want is the scattering angle but we need to be careful about reference
frames.
Most natural frame of ref is Center-of-Mass frame, in which C of M is stationary. C of
M has position:
m1 r1 + m2 r2
(3.18)
R=
m1 + m2
and velocity (in lab frame)
m1 v1 + m2 v2
(3.19)
V=
m1 + m2
Now
m2
r1 = R +
r
(3.20)
m1 + m2
m1
r2 = R
r
(3.21)
m1 + m2
So motion of either particle in C of M frame is a factor times difference vector, r.
Velocity in lab frame is obtained by adding V to the C of M velocity, e.g.

m2 r
m1 +m2

+ V.

Angles of position vectors and velocity differences are same in all frames.
Angles (i.e. directions) of velocities are not same.

3.1.2

Scattering Angle

In C of M frame is just the final angle of r.

21 + =

(3.22)

(1 is negative)
= + 21

;
42

1 =

.
2

(3.23)

1
1
1

Figure 3.2: Relation between 1 and .

tan 1 = tan
=
2
2


= cot

(3.24)

So

b
=
2
b90

b90
tan
=
2
b
cot

(3.25)
(3.26)

But scattering angle (defined as exit velocity angle relative to initial velocity) in lab frame
is different.
Final velocity in CM frame
0
vCM
= v1CM (cos c , sin C ) =

m2
v1 (cos c , sin c )
m1 + m2

(3.27)

[ c and v1 is initial relative velocity]. Final velocity in Lab frame

vL0

0
vCM

m2 v1
m2 v1
+V = V +
cos c ,
sin c
m1 + m2
m1 + m2


(3.28)

So angle is given by
cot L =

m2 v1
cos c
m1 +m2
m2 v1
sin x
m1 +m2

V +

V m1 + m2
cosecc + cot c
v1
m2

(3.29)

For the specific case when m2 is initially a stationary target in lab frame, then
V
cot L

m1 v1
and hence
m1 + m2
m1
=
cosecc + cot c
m2
=

(3.30)
(3.31)

This is exact.
Small angle approximation (cot 1 , cosec

gives

1
m1 1
1
m2
=
+
L =
c
L
m2 c c
m1 + m2
So small angles are proportional, with ratio set by the mass-ratio of particles.
43

(3.32)

Center-of-Mass Frame

y
x
c

Particle 1
m
_r v
m1 0

Particle 2

m
_r v
m2 0

C-of-M

Laboratory Frame,
Stationary Target

Particle 1
v0

C-of-M

Particle 2

Figure 3.3: Collisions viewed in Center of Mass and Laboratory frame.

3.2

Differential Cross-Section for Scattering by Angle

Rutherford Cross-Section

By definition the cross-section, , for any specified collision process when a particle is passing
through a density n2 of targets is such that the number of such collisions per unit path length
is n2 .
Sometimes a continuum of types of collision is considered, e.g. we consider collisions at
different angles () to be distinct. In that case we usually discuss differential cross-sections
d
(e.g d
) defined such that number of collisions in an (angle) element d per unit path length
d
d
is n2 d dx. [Note that d
is just notation for a number. Some authors just write (), but I
find that less clear.]
Normally, for scattering-angle discrimination we discuss the differential cross-section per unit
solid angle:
d
.
(3.33)
ds
This is related to scattering angle integrated over all azimuthal directions of scattering by:
44

r0

r0 sin

ds

Figure 3.4: Scattering angle and solid angle relationship.

So that since

we have

ds = 2 sin d

(3.34)

d
d
ds =
d
ds
d

(3.35)

d
1
d
=
ds
2 sin d

(3.36)

Now, since is a function (only) of the impact parameter, b, we just have to determine the
number of collisions per unit length at impact parameter b.










































































b













v 

























































































db





















 


 


 


 



 

 

 

 



 


 


 


 


 

 

 

 



 


 


 


 


 

 

 

 






 


 


 






 


 


 






 


 


 



 


 


 


 
dl

Figure 3.5: Annular volume corresponding to db.

Think of the projectile as dragging along an annulus of radius b and thickness db for an
elementary distance along its path, d`. It thereby drags through a volume:
d`2bdb .

(3.37)

Therefore in this distance it has encountered a total number of targets

d`2bdb . n2
45

(3.38)

at impact parameter b(db). By definition this is equal to d` d

dbn2 . Hence the differential
db
cross-section for scattering (encounter) at impact parameter b is
d
= 2b .
db

(3.39)

Again by definition, since is a function of b



d
d
d
d  db 
dx =
db
=
=  .
 d 
d
db
d
db

(3.40)

[db/d is negative but differential cross-sections are positive.]

Substitute and we get


d
1
d  db 
b  db 
=

 =

 .
ds
2 sin db  d 
sin  d 

(3.41)

[This is a general result for classical collisions.]

For Coulomb collisions, in C of M frame,

b
=
2
b90

(3.42)

db
d

b90

= b90
cot
= cosec2
.
d
d
2
2
2

(3.43)

 

cot

Hence

b90 cot 2 b90

d

=
cosec2
ds
sin
2
2

2
1
b cos 2 / sin 2
= 90

2 2 sin 2 cos 2 sin2 2

b290
=
4 sin4 2

(3.44)
(3.45)
(3.46)

This is the Rutherford Cross-Section.

d
b290
=
ds
4 sin4

(3.47)

for scattering by Coulomb forces through an angle measured in C of M frame.

Notice that

d
ds

as 0.

This is because of the long-range nature of the Coulomb force. Distant collisions tend to
dominate. ( 0 b ).

46

3.3

Relaxation Processes

There are 2 (main) different types of collisional relaxation process we need to discuss for a
test particle moving through a background of scatterers:
1. Energy Loss (or equilibrium)
2. Momentum Loss (or angular scattering)
The distinction may be illustrated by a large angle (90 ) scatter from a heavy (stationary)
target.
If the target is fixed, no energy is transferred to it. So the energy loss is zero (or small if
scatterer is just heavy). However, the momentum in the x direction is completely lost in
this 90 scatter.
This shows that the timescales for Energy loss and momentum loss may be very different.

3.3.1

Energy Loss

For an initially stationary target, the final velocity in lab frame of the projectile is
vL0

m 2 v1
m2 v1
ml v1
+
cos c ,
sin c
=
m1 + m2 m1 + m2
m1 + m2


(3.48)

So the final kinetic energy is

K

1
1
=
m1 vL02 = m1 v12
2
2

(

m1
m1 + m2

2

2m1 m2
cos c
(m1 + m2 )2



m22
2
2
+
cos

+
sin

c
c
(m1 + m2 )2
(
)
1
2m1 m2
2
=
m 1 v1 1 +
(cos c 1)
2
(m1 + m2 )2
(
)
1
2m1 m2
2
2 c
=
m 1 v1 1 +
2 sin
2
2
(m1 + m2 )2

(3.49)

(3.50)
(3.51)
(3.52)

Hence the kinetic energy lost is K = K K 0

1
4m1 m2
2 c
m1 v2
2 sin
2
2
(m1 + m2 )
1
4m1 m2
1
m1 v12
=
2
2 
b
2
(m1 + m2 )
+1
b90

(3.53)

[using cot

c
b
=
]
2
b90

(3.54)

(exact). For small angles  1 i.e. b/b90  1 this energy lost in a single collision is
approximately
!2


1
4m1 m2
b90
2
m 1 v1
(3.55)
2
b
(m1 + m2 )2
47

If what we are asking is: how fast does the projectile lose energy? Then we need add up the
effects of all collisions in an elemental length d` at all relevant impact parameters.
The contribution from impact parameter range db at b will equal the number of targets
encountered times K:
n
d`2bdb}
| 2 {z

encounters

1
4m1 m2
m1 v12
2
(m1 + m2 )2
|

b90
b

!2

{z

(3.56)
}

Loss per encounter (K)

This must be integrated over all b to get total energy loss.

Z
4m1 m2
dK = n2 d`K
(m1 + m2 )2

so

b90
b

!2

2bdb

(3.57)

dK
m1 m2
= K n2
8b290 [ln |b|]max
min
d`
(m1 + m2 )2

(3.58)

We see there is a problem both limits of the integral (b 0, b ) diverge logarithmically.

That is because the formulas we are integrating are approximate.
1. We are using small-angle approx for K.
2. We are assuming the Coulomb force applies but this is a plasma so there is screening.

3.3.2

Cut-offs Estimates

1. Small-angle approx breaks down around b = b90 . Just truncate the integral there;
ignore contributions from b < b90 .
2. Debye Shielding says really the potential varies as

exp

r
D

instead of

1
r

(3.59)

so approximate this by cutting off integral at b = D equivalent to

bmin = b90 .
dK
d`

= Kn2

bmax = D .

(3.60)

m1 m2
2
2 8b90 ln ||
(m1 + m2 )

D
=
=
b90

0 Te
ne2

48

 12 ,

q1 q 2
40 mr v12

(3.61)
!

(3.62)

So Coulomb Logarithm is ln
D
=
=
b90

0 Te
ne2

 12 ,

q1 q2
40 mr v12

(3.63)

Because these cut-offs are in ln term result is not sensitive to their exact values.
One commonly uses Collision Frequency. Energy Loss Collision Frequency is
K v1

m1 m2
1 dK
8b290 ln ||
= n2 v1
K dL
(m1 + m2 )2

(3.64)

Substitute for b90 and mr (in b90 )

"

Collision time

q1 q2
m1 m2
= n2 v1
m2 2
2 8
40 mm11+m
v
(m1 + m2 )
2 1
q12 q22
8
ln
= n2
2
(40 ) m1 m2 v13

ln

(3.65)
(3.66)

K 1/K

Effective (Energy Loss) Cross-section

1 dK
K d`

K = K /n2 v1 =

3.3.3

#2

= K n2

8
q12 q22
ln
2
(40 ) m1 m2 v14

(3.67)

Momentum Loss

Loss of x-momentum in 1 collision is

0
)
px = m1 (v1 vLx

(3.68)

m1
m2
= m 1 v1 1
+
cos c
m1 + m2 m1 + m2
m2
= px
(1 cos c )
m1 + m2
m2
2c
m2
2b290
' px
= px
m1 + m2 2
m1 + m2 b2




(3.69)
(3.70)
(3.71)

(small angle approx). Hence rate of momentum loss can be obtained using an integral
identical to the energy loss but with the above parameters:
Z bmax
dp
m2
2b290
= n2 p
2bdb
d`
m1 + m2 b2
bmin
m2
= n2 p
4 b290 ln
m1 + m2

(3.72)
(3.73)

Note for the future reference:

dp
dp
m1 m2
=v
= n2 v 2
4b290 ln .
dt
d`
m1 + m2
49

(3.74)

Therefore Momentum Loss.

Collision Frequency
p = v1

1 dp
m2
= n 2 v1
4b290 ln
p d`
m1 + m2
"

m2
q 1 q2
= n 2 v1
4
m2 2
m1 + m2
40 mm11+m
v
2 1
= n2

Collision Time

(3.75)

#2

ln

q12 q22 4 (m1 + m2 )

ln
m2 m21 v13
(40 )2

(3.76)
(3.77)

p = 1/p

Cross-Section (effective)

= p /n2 v1

Notice ratio
Energy Loss K
2
=
Momentum loss p
m1 m2

m1 + m2
2m1
=
2
m2 m1
m1 + m2

(3.78)

This is
' 2
= 1
 1

if
if
if

m1 >> m2
m1 = m2
m1  m2 .

(3.79)
(3.80)
(3.81)

Third case, e.g. electrons shows that mostly the angle of velocity scatters. Therefore
Momentum Scattering time is often called 90 scattering time to diffuse through 90 in
angle.

3.3.4

Random Walk in angle

When m1 << m2 energy loss << momentum loss. Hence |vL0 ' v1 . All that matters is the
scattering angle: L ' c ' 2b90 /b.
Mean angle of deviation in length L is zero because all directions are equally likely.
But:
Mean square angle is
2

= n2 L

bmax

2 2bdb

(3.82)

bmin

= Ln2 8 b290 ln

(3.83)

Spread is all round when 2 ' 1. This is roughly when a particle has scattered 90 on
average. It requires
Ln2 8b290 ln = 1 .
(3.84)
50

So can think of a kind of cross-section for 90 90 scattering as such that

n2 L90 = 1 when Ln2 8 b290 ln = 1
i.e. 90 = 8 b290 ln (= 2p )

(3.85)
(3.86)

This is 8 ln larger than cross-section for 90 scattering in single collision.

Be Careful! 90 is not a usual type of cross-section because the whole process is really
diffusive in angle.
Actually all collision processes due to coulomb force are best treated (in a Mathematical
way) as a diffusion in velocity space
Fokker-Planck equation.

3.3.5

Summary of different types of collision

The Energy Loss collision frequency is to do with slowing down to rest and exchanging
energy. It is required for calculating
Equilibration Times (of Temperatures)
Energy Transfer between species.
The Momentum Loss frequency is to do with loss of directed velocity. It is required for
calculating
Mobility: Conductivity/Resistivity
Viscosity
Particle Diffusion
Energy (Thermal) Diffusion
Usually we distinguish between electrons and ions because of their very different mass:
Energy Loss
[Stationary Targets]
Momentum Loss
K

ee =

ei =

ii =

ie =

e4
ne
(40 )2
Z 2 e4
ni
(40 )2
Z 2 e4
ni
(4eo )2
Z 2 e4
ne e 2
(40 )

8
ln
m2e ve3
8
ln
me mi ve3
8
ln
m2i vi3
8
ln
mi me vi3

me + me
ee = ee
=1
2me


me + mi
mi
p
ei =K ei
'
2me
2me


mi + mi
p
K
ii = ii
=1
2mi


me + mi
1
p
K
ie = ie =
'
2mii
2
p

(3.87)

Sometimes one distinguishes between transverse diffusion of velocity and momentum loss.
The ratio of these two is
p21
p2 L


,
 p 
k 



 pL 

d2L
=
dL



,
 1 dp 




 p dL 
2

m2

m1 +m2 c
2c
m2
m1 +m2 2

51

(3.88)
2m2
.
m1 + m2

(3.89)

So
2m2
90
=
p
m1 + m2

=1

like particles

'2
m1 << m2
2m2
'
m2 << m1 .
m1

(3.90)
(3.91)
(3.92)

Hence

ee =

ei

ii

ie

ee =

ee (= ee !!)
ni
= 2p ei = K ee Z 2 (= Zee ) (= ei )
ne
p
K
= ii = ii (= ii !!) (Like Ions)
2me p
me K
=
ie =
ie = K ii = ii
mi
mi

(3.93)
(3.94)
(3.95)
(3.96)

[But note: ions are slowed down by electrons long before being angle scattered.]

3.4

Thermal Distribution Collisions

So far we have calculated collision frequencies with stationary targets and single-velocity
projectiles but generally we shall care about thermal (Maxwellian) distributions (or nearly
thermal) of both species. This is harder to calculate and we shall resort to some heuristic
calculations.

3.4.1

ei

Very rare for thermal ion velocity to be electron. So ignore ion motion.
Average over electron distribution.
Momentum loss to ions from (assumed) drifting Maxwellian electron distribution:


fe (v) = ne

me
2Te

3
2

m (v vd )2
exp
2T
"

(3.97)

Each electron in this distribution is losing momentum to the ions at a rate given by the
collision frequency
q 2 q 2 4 (me + mi )
ln
(3.98)
p = ni e i 2
mi m2e v 3
(40 )
so total rate of loss of momentum is given by (per unit volume)
dp Z
= fe (v) p (v) me v d3 v
dt
To evaluate this integral approximately we adopt the following simplifications.

52

(3.99)

1. Ignore variations of ln with v and just replace a typical thermal value in =

D /b90 (v1 ).
2. Suppose
q that drift velocity vd is small relative to the typical thermal velocity, written
ve te /me and express fe in terms of u vve to first order in ud vved :
1
exp
(u ud )2
2

fe = ne

(3.100)

(2) 2 ve3
"
#
ne
u2
'
(1 + u.ud ) exp
= (1 + ux ud ) fo
3
2
(2) 2 ve3

(3.101)

taking x-axis along ud and denoting by fo the unshifted Maxwellian.

Then momentum loss rate per unit volume
Z
dpx

=
fe p me vx d3 v
dt
= p (vt )me

(1 + ux ud )fo

= p (vt )me vd

ve3
vx d3 v
v3

(3.102)

u2x
fo d3 v
u3

To evaluate this integral, use the spherical symmetry of fo to see that:

Z

u2x
1 Z u2x + u2y + u3z
1 Z u2
3
3
f
d
v
=
f
d
v
=
fo d3 v
o
o
u3
3Z
u3
3 u3
1 ve
=
fo 4v 2 dv
3 0 v
2 Z
=
ve
fo 2vdv
3
0
!
Z
v 2
2
ne
exp
dv 2
=
ve
2ve2
3 (2) 32 v 3 0
e
2 ne
2
=
3 2 =
1 ne .
3 (2) 2
3 (2) 2

Thus the Maxwell-averaged momentum-loss frequency is

1 dp
2

ei =
1 p (vt )
p dt
3 (2) 2

(3.103)

(3.104)

(where p = me vd ne is the momentum per unit volume attributable to drift).

ei =
=

2
1

3 (2) 2
2
3 (2)

1
2

ni
ni

qe2 qi2 4 (me + mi )

ln e
mi m2e ve3
(40 )2
ze2
40

!2

4
1
2

3
2

ln e

me Te
(substituting for thermal electron velocity, ve , and dropping
This is the standard form of electron collision frequency.
53

me
mi

(3.105)
(3.106)

order term), where Ze = qi .

3.4.2

ie

Ion momentum loss to electrons can be treated by a simple Galilean transformation of the
e i case because it is still the electron thermal motions that matter.

Ions

Electrons

Ions

Electrons

vi

vi = 0

vd = vi

Figure 3.6: Ion-electron collisions are equivalent to electron-ion collisions in a moving reference frame.
Rate of momentum transfer,

dp
,
dt

is same in both cases:

Hence pe ei = pi ie or
ie =

dp
= p
dt

(3.107)

pe
ne me
ei =
ei
pi
ni mi

(3.108)

(since drift velocities are the same).

Ion momentum loss to electrons is much lower collision frequency than e i because ions
possess so much more momentum for the same velocity.

3.4.3

ii

Ion-ion collisions can be treated somewhat like e i collisions except that we have to
account for moving targets i.e. their thermal motion.
Consider two different ion species moving relative to each other with drift velocity vd ; the
targets thermal motion affects the momentum transfer cross-section.
Using our previous expression for momentum transfer, we can write the average rate of
transfer per unit volume as: [see 3.74 note for future reference]

m1 m2
dp Z Z
=
vr
vr 4 b290 ln f1 f2 d2 v1 d3 v2
dt
m1 + m2

(3.109)

where vr is the relative velocity (v1 v2 ) and b90 is expressed

b90 =

q1 q2
1
40 mr vr2
54

(3.110)

and mr is the reduced mass

m1 m2
.
m1 +m2

Since everything in the integral apart from f1 f2 depends only on the relative velocity, we
proceed by transforming the velocity coordinates from v1 , v2 to being expressed in terms of
relative (vr ) and average (V say)
vr v1 v2

m1 v1 + m2 v2
.
m1 + m2

(3.111)

Take f1 and f2 to be shifted Maxwellians in the overall C of M frame:



fj = nj

mj
2T

3
2

mj (vj vdj )2
exp
2T
"

(j = 1, 2)

(3.112)

where m1 vd1 + m2 vd2 = 0. Then

f1 f2

3 

3

m2 2
m1 v12 m2 v22
exp

= n1 n2
2T
2T
2T


v1 .m1 vd1 v2 .m2 vd2
1+
+
T
T


m1
2T

"

(3.113)

to first order in vd . Convert CM coordinates and find (after algebra)



f1 f2 = n1 n2

M
2T


 32 

1+

mr
2T

3

mr
vd .vr
T

M V 2 mr vr2
exp

2T
2T
"

(3.114)

where M = m1 + m2 . Note also that (it can be shown) d3 v1 d3 v2 = d3 vr d3 V . Hence

dp Z Z
=

dt

3 

3

M 2 mr 2
vr mr vr 4
ln n1 n2
2T!
2T
!

2
2 
MV
mr vr
mr
exp
1+
vd .vr d3 vr d3 V
exp
2T
2T
T


b290

(3.115)

and since nothing except the exponential depends on V , that integral can be done:
dp Z
mr

= vr mr vr 4 ln n1 n2
dt
2T


3
2

mr vr2
exp
2

!

mr
1+
vd .vr d3 vr
T


(3.116)

This integral is of just the same type as for e i collisions, i.e.

Z 2
dp
ux
= vd vrt mr 4 b290 (vrt ) ln t n1 n2
fo (vr )d3 vr
dt
u3
2
= vd vrt mr 4 b290 (vrt ) ln t n1 n2
2
3 (2) 2

55

(3.117)

where vrt mTr , b290 (vrt ) is the ninety degree impact parameter evaluated at velocity vtr ,
and fo is the normalized Maxwellian.
dp
2

=
1
dt
3 (2) 2

q 1 q2
40

2

4
ln t n1 n2 mr vd
3
m2r vrt

(3.118)

This is the general result for momentum exchange rate between two Maxwellians drifting at
small relative velocity vd .
To get a collision frequency is a matter of deciding which species is stationary and so what the
momentum density of the moving species is. Suppose we regard 2 as targets then momentum
density is n1 m1 vd so
2

1
q 1 q2
dp
2
=
=
1 n2
n1 m1 vd dt
40
3 (2) 2


12

4 ln t
.
3
mr vrt
m1

(3.119)

This expression works immediately for electron-ion collisions substituting mr ' me , recovering previous.
For equal-mass ions mr =

m2i
mi +mi

= 12 mi and vrt =

T
mr

2T
.
mi

Substituting, we get
ii =

1
3


1
2

ni

q1 q2
40

2

4
1
2

3
2

ln

(3.120)

mi Ti

1
2

that is,
times the e i expression but with ion parameters substituted. [Note, however,
that we have considered the ion species to be different.]

3.4.4

ee

Electron-electron collisions are covered by the same formalism, so

ee =

1
3

1
2

ne

e2
40

!2

4
1

ln .

(3.121)

me2 Te2

However, the physical case under discussion is not so obvious; since electrons are indistiguishable how do we define two different drifting maxwellian electron populations? A more
specific discussion would be needed to make this rigorous.

Generally ee ei / 2 : electron-electron collision frequency electron-ion (for momentum

loss).

3.4.5

Summary of Thermal Collision Frequencies

For momentum loss:

ei

!2
2
Ze2
4
= ni
ln e .
3
40 me12 Te32
56

(3.122)

1
ee ' ei . (electron parameters)
2
ne me
ie =
ei .
ni mi




1
2
2
qi qi0 2 4
mi0
ii0 = ni0
ln i
1
3
3
40 m 2 T 2 mi + mi0
i

Energy loss

(3.123)
(3.124)
(3.125)

related to the above (p ) by

K

2mi
p
.
m1 + m2

(3.126)

Transverse diffusion of momentum , related to the above by:

3.5
3.5.1

2m2 p
.
m1 + m2

(3.127)

Applications of Collision Analysis

Energetic (Runaway) Electrons

Consider an energetic ( 21 me v12 >> T ) electron travelling through a plasma. It is slowed down
(loses momentum) by collisions with electrons and ions (Z), with collision frequency:
p

ee = ee = ne

ei =

e4
8
ln
2
(40 ) m2e v13

1
Z ee
2

(3.128)
(3.129)

Hence (in the absence of other forces)

d
(me v) = (p ee + p ei ) mp v
dt


Z
= 1+
ee me v
2

(3.130)
(3.131)

This is equivalent to saying that the electron experiences an effective Frictional force
d
Z
(me v) = 1 +
ee me v
=
dt
2


Z
e4
8 ln
= 1+
ne
2
2
(40 ) me v 2


Ff
Ff

57

(3.132)
(3.133)

Notice
1. for Z = 1 slowing down is

2
3

on electrons

1
3

ions

2. Ff decreases with v increasing.

Suppose now there is an electric field, E. The electron experiences an accelerating Force.
Total force
d
Z
e4
8 ln
(mv) = eE + Ff = eE 1 +
ne
2
dt
2
(40 ) me v 2


F =

(3.134)

Two Cases (When E is accelerating)

1. |eE| < |Ff |: Electron Slows Down
2. |eE| > |Ff |: Electron Speeds Up!
Once the electron energy exceeds a certain value its velocity increases continuously and the
friction force becomes less and less effective. The electron is then said to ahve become a
runaway.
Condition:

1
Z
e4
8 ln
me v 2 > 1 +
ne
2
2
2
2eE
(40 )


3.5.2

(3.135)

Plasma Resistivity (DC)

Consider a bulk distribution of electrons in an electric field. They tend to be accelerated by

E and decelerated by collisions.
In this case, considering the electrons as a whole, no loss of total electron momentum by
e e collisions. Hence the friction force we need is just that due to ei .
If the electrons have a mean drift velocity vd (<< vthe ) then
d
(me vd ) = eE ei me vd
dt

(3.136)

Hence in steady state

vd =

eE
.
me ei

(3.137)

The current is then

ne e2 E
(3.138)
me ei
Now generally, for a conducting medium we define the conductivity, , or resistivity, , by
j = ne evd =

j = E

j = E
58

1
=

(3.139)

Therefore, for a plasma,

=

ne e2
1
=

me ei

(3.140)

Substitute the value of ei and we get

1

ni Z 2
e2 me2 8 ln

'
ne (4 )2 32 Te32
0

(3.141)

Ze2 me2 8 ln
=
3

(40 )2 3 2 Te2

(for a single ion species).

(3.142)

Notice
1. Density cancels out because more electrons means (a) more carriers but (b) more
collisions.
2. Main dependence is Te3/2 . High electron temperature implies low resistivity (high
conductivity).
3. This expression is only approximate because the current tends to be carried by the
more energetic electrons, which have smaller ei ; thus if we had done a proper average
over f (ve ) we expect a lower numerical value. Detailed calculations give
ln

= 5.2 105

(3.143)

(Te /eV ) 2
for Z = 1 (vs. ' 104 in our expression). This is Spitzer resistivity. The detailed
calculation value is roughly a factor of two smaller than our calculation, which is not
a negligible correction!

3.5.3

Diffusion

For motion parallel to a magnetic field if we take a typical electron, with velocity vk ' vte it
will travel a distance approximately
`e = vte / ei

(3.144)

before being pitch-angle scattered enough to have its velocity randomised. [This is an orderof-magnitude calculation so we ignore ee .] ` is the mean free path.
Roughly speaking, any electron does a random walk along the field with step size ` and step
frequency ei . Thus the diffusion coefficient of this process is
Dek '

`2e

2
vte
ei '
.
ei

(3.145)

Similarly for ions

Dik ' `2i ii '
59

vti2
ii

(3.146)

Notice
ii / ei '
Hence

me
mi

1
2

'

vti
vte

(if Te ' Ti )

(3.147)

`e ' `i

Mean free paths for electrons and ions are same.

The diffusion coefficients are in the ratio
Di
me
'
De
mi


1
2

: Ions diffuse slower in parallel direction.

(3.148)

Diffusion Perpendicular to Mag. Field is different

GC
rL

GC
rL

Figure 3.7: Cross-field diffusion by collisions causing a jump in the gyrocenter (GC) position.
Roughly speaking, if electron direction is changed by 90 the Guiding Centre moves by
a distance rL . Hence we may think of this as a random walk with step size rL and
frequency ei . Hence
v2
2
De ' rLe
ei ' te2 ei
(3.149)
e
Ion transport is similar but requires a discussion of the effects of like and unlike collisions.
Particle transport occurs only via unlike collisions. To show this we consider in more detail
the change in guiding center position at a collision. Recall mv = qv B which leads to
v =

q
rL B
m

(perp. velocity only).

(3.150)

This gives
rL =

B mv
qB 2

(3.151)

At a collision the particle position does not change (instantaneously) but the guiding center
position (r0 ) does.
r00 + r0L = r0 + rL r0 r00 r0 = (r0L rL )

60

(3.152)

Change in rL is due to the momentum change caused by the collision:

r0L rL =

B
B
0

m(v

v)

(mv)
qB 2
qB 2

So
r0 =

B
(mv).
qB 2

(3.153)

(3.154)

The total momentum conservation means that (mv) for the two particles colliding is equal
and opposite. Hence, from our equation, for like particles, r0 is equal and opposite. The
mean position of guiding centers of two colliding like particles (r01 + r02 )/2 does not change.
No net cross field particle (guiding center) shift.
Unlike collisions (between particles of different charge q) do produce net transport of particles
of either type. And indeed may move r01 and r02 in same direction if they have opposite
charge.
v2
2 p
ie ' ti2 p ie
(3.155)
Di ' rLi
i
2
2
Notice that rLi
/rLe
' mi /me

ie / ei

'

me
mi

So Di /De ' 1 (for equal temperatures). Collisional diffusion rates of particles are same
for ions and electrons.
However energy transport is different because it can occur by like-like collisins.
Thermal Diffusivity:
2
2
e rLe
( ei + ee ) rLe
ei
2 p
2
i rLi ( ie + ii ) ' rLi ii
1

mi me2
mi
r2 ii
2Li
'
1 =
rLe ei
me m 2
me
i


i /e

( ei ee )
( ii >> ie )

(3.156)
(3.157)

1
2

(equal T)

(3.158)
1

Collisional Thermal transport by Ions is greater than by electrons [factor (mi /me ) 2 ].

3.5.4

Energy Equilibration

If Te 6= Ti then there is an exchange of enegy between electrons and ions tending to make
Te = Ti . As we saw earlier
2me p
me
K
ei =
ei =
ei
(3.159)
mi
mi
So applying this to averages.
K

ei

'

2me
ei
mi

61

(' ie )

(3.160)

Thermal energy exchange occurs me /mi slower than momentum exchange. (Allows Te 6=
Ti ). So
dTe
dTi
(3.161)
=
= K ei (Te Ti )
dt
dt
From this one can obtain the heat exchange rate (per unit volume), Hei , say:
d 3
d
ne Te =
dt 2
dt
3 d
= n (Te Ti ) =
4 dt

Hei =

3
ni Ti
2

(3.162)

3 K
n ei (Te Ti )
2

(3.163)

Important point:
K

ei

1
me
'
Zee ' 2
mi
Z

Me
mi

 12

ii .

(3.164)

Electrons and Ions equilibrate among themselves much faster than with each other.

3.6

Some Orders of Magnitude

1. ln is very slowly varying. Typically has value 12 to 16 for laboratory plasmas.

3

2. ei 6 1011 (ni /m3 ) / (Te /eV ) 2

(ln = 15, Z = 1).
5 1
20 3
e.g. = 2 10 s (when n = 10 m and Te = 1keV.) For phenomena which happen
much faster than this, i.e.  1/ei 5s, collisions can be ignored.
Examples: Electromagnetic Waves with high frequency.
3. Resistivity. Because most of the energy of a current carrying plasma is in the B field
not the K.E. of electrons. Resistive decay of current can be much slower than ei . E.g.
Coaxial Plasma: (Unit length)
Inductance L = o ln ab
Resistance R = 1/a2
L/R decay time
o a2
b
ne e2
b
ln '
o a2 ln

a
me ei
a
2 2
2
2
a
ne e a 1
1
1

= p2
>>
.
2
me 0 c ei
c
ei
ei

(3.165)

Comparison 1 keV temperature plasma has same (conductivity/) resistivity as a slab

of copper ( 2 108 m).
Ohmic Heating Because Te3/2 , if we try to heat a plasma Ohmically, i.e. by simply
passing a current through it, this works well at low temperatures but its effectiveness falls
off rapidly at high temperature.
62

Result for most Fusion schemes it looks as if Ohmic heating does not quite yet get us to the
required ignition temperature. We need auxilliary heating, e.g. Neutral Beams. (These slow
down by collisions.)

63

Chapter 4

Fluid Description of Plasma

The single particle approach gets to be horribly complicated, as we have seen.

Basically we need a more statistical approach because we cant follow each particle separately.

If the details of the distribution function in velocity space are important we have to stay

with the Boltzmann equation. It is a kind of particle conservation equation.

4.1

Particle Conservation (In 3d Space)

Figure 4.1: Elementary volume for particle conservation

Number of particles in box xyz is the volume, V = xyz, times the density n.
Rate of change of number is is equal to the number owing across the boundary per unit
time, the ux. (In absence of sources.)

[xyz n] = Flow Out across boundary.

t

(4.1)

Take particle velocity to be v(r) [no random velocity, only ow] and origin at the center of
the box refer to ux density as nv = J.
Flow Out = [Jz (0, 0, z/2) Jz (0, 0, z/2)] xy + x + y .
64

(4.2)

Expand as Taylor series

Jz (0, 0, ) = Jz (0) +

Jz .
z

(4.3)

So,

(nvz )zxy + x + y
z
= V . (nv).

ow out

(4.4)

Hence Particle Conservation

n = .(nv)
t
Notice we have essential proved an elementary form of Gausss theorem

.Ad r =

A.dS.

(4.5)

(4.6)

The expression: Fluid Description refers to any simplied plasma treatment which does
not keep track of vdependence of f detail.
1. Fluid Descriptions are essentially 3d (r).
2. Deal with quantities averaged over velocity space (e.g. density, mean velocity, ...).
3. Omit some important physical processes (but describe others).
4. Provide tractable approaches to many problems.
5. Will occupy most of the rest of my lectures.
Fluid Equations can be derived mathematically by taking moments1 of the Boltzmann Equa
tion.
0th moment
d3 v

1st moment
vd3 v

2nd moment
vvd3 v

(4.7)
(4.8)
(4.9)

These lead, respectively, to (0) Particle (1) Momentum (2) Energy conservation equations.
We shall adopt a more direct physical approach.
1

They are therefore sometimes called Moment Equations.

65

4.2

Fluid Motion

The motion of a uid is described by a vector velocity eld v(r) (which is the mean velocity
of all the individual particles which make up the uid at r). Also the particle density
n(r) is required. We are here discussing the motion of uid of a single type of particle of
mass/charge, m/q so the charge and mass density are qn and mn respectively.
The particle conservation equation we already know. It is also sometimes called the Conti
nuity Equation

n + .(nv) = 0
(4.10)
t
It is also possible to expand the . to get:

n + (v.)n + n.v = 0
t

(4.11)

The signicance, here, is that the rst two terms are the convective derivative of n
D
d

+ v.
Dt
dt
t

(4.12)

so the continuity equation can be written

D
n = n.v
Dt

4.2.1

(4.13)

Lagrangian & Eulerian Viewpoints

There are essentially 2 views.

1. Lagrangian. Sit on a uid element and move with it as uid moves.

Figure 4.2: Lagrangean Viewpoint

2. Eulerian. Sit at a xed point in space and watch uid move through your volume
element: identity of uid in volume continually changing

t
D
Dt

means rate of change at xed point (Euler).

d
dt

v. =

+ v. means
t

dx

+ dy
+ dz
t x
t y
t z

rate of change at moving point (Lagrange).

: change due to motion.

66

Figure 4.3: Eulerian Viewpoint

Our derivation of continuity was Eulerian. From the Lagrangian view
d N
N d
1 dV
D
n=
=
V = n
2
Dt
dt V
V dt
V dt

(4.14)

since total number of particles in volume element (N ) is constant (we are moving with
them). (V = xyz.)
Now

d
V
dt

dx
dy
dz
yz +
zx +
yx
dt
dt
dt

1 dx
1 dy
1 dz
= V
+
+
x dt
y dt
x dt
=

d (x)
= vx (x/2) vx (x/2)
dt
vx
x
etc. . . . y
...z
x

But

Hence

(4.15)
(4.16)

(4.17)
(4.18)

d
vx vy vz
V = V
+
+
= V .v
dt
x
y
z

(4.19)

and so

D
n = n.v
(4.20)
Dt
Lagrangian Continuity. Naturally, this is the same equation as Eulerian when one puts
D

= t
+ v..
Dt

The quantity .v is the rate of (Volume) compression of element.

4.2.2

Momentum (Conservation) Equation

Each of the particles is acted on by the Lorentz force q[E + ui B] (ui is individual particles

velocity).

Hence total force on the uid element due to EM elds is

(q [E + ui B]) = N q (E + v B)

67

(4.21)

(Using mean: v =

u/N.)

EM Force density (per unit volume) is:

FEM = nq(E + v B).

(4.22)

The total momentum of the element is

mui = m N v = V mnv

(4.23)

so Momentum Density is mnv.

If no other forces are acting then clearly the equation of motion requires us to set the time

derivative of mnv equal to FEM . Because we want to retain the identity of the particles

under consideration we want D/Dt i.e. the convective derivative (Lagrangian picture).

In general there are additional forces acting.

(1) Pressure
(2) Collisional Friction.

4.2.3

Pressure Force

In a gas p(= nT ) is the force per unit area arising from thermal motions. The surrounding
uid exerts this force on the element:

Figure 4.4: Pressure forces on opposite faces of element.

Net force in x direction is
p (x/2) yz + p (x/2) yz
p
p
xyz
= V
= V (p)x
x
x

(4.24)
(4.25)

So (isotropic) pressure force density (/unit vol)

Fp = p

(4.26)

How does this arise in our picture above?

Answer: Exchange of momentum by particle thermal motion across the element boundary.
Although in Lagrangian picture we move with the element (as dened by mean velocity v)
individual particles also have thermal velocity so that the additional velocity they have is
wi = ui v

peculiar velocity
68

(4.27)

Because of this, some cross the element boundary and exchange momentum with outside.
(Even though there is no net change of number of particles in element.) Rate of exchange
of momentum due to particles with peculiar velocity w, d3 w across a surface element ds is
f (w)mw d3 w

momm density at w

w
. ds

ow rate across ds

(4.28)

Integrate over distrib function to obtain the total momentum exchange rate:
ds.

mwwf (w)d3 w

(4.29)

The thing in the integral is a tensor. Write

p =

mwwf (w)d3 w

(Pressure Tensor)

(4.30)

Then momentum exchange rate is

p . ds

(4.31)

Actually, if f (w) is isotropic (e.g. Maxwellian) then

pxy =
and pxx =

m wx wy f (w)d3 w = 0

mwx2 f (w)d3 w nT (= pyy = pzz = p)

etc.

(4.32)
(4.33)

So then the exchange rate is pds. (Scalar Pressure).

Integrate ds over the whole V then x component of momm exchange rate is
x
x
p
yz p
yz = V (p)x
2
2

(4.34)

and so
Total momentum loss rate due to exchange across the boundary per unit volume is
p

(= Fp )

(4.35)

In terms of the momentum equation, either we put p on the momentum derivative side or

Fp on force side. The result is the same.

Ignoring Collisions, Momentum Equation is

D
(mnV v) = [FEM + Fp ] V
Dt
Recall that nV = N

D
(N )
Dt

(4.36)

= 0; so

L.H.S. = mnV

Dv
dt

(4.37)

Thus, substituting for F s:

Momentum Equation.

Dv
v
mn
= mn
+ v.v = qn (E + v B) p
Dt
t
69

(4.38)

4.2.4

Momentum Equation: Eulerian Viewpoint

Fixed element in space. Plasma ows through it.

1. E.M. force on element (per unit vol.)
FEM = nq(E + v B)

as before.

(4.39)

2. Momentum ux across boundary (per unit vol)

= .

m(v + w)(v + w) f (w) d3 w

vw +
wv +ww) f (w) d3 w}

integrates to 0
= .{mnvv + p}
= mn(v.)v + mv [.(nv)] + p

= .{

m(vv +

(4.40)
(4.41)

(4.42)
(4.43)

(Take isotropic p.)

3. Rate of change of momentum within element (per unit vol)
=

(mnv)
t

(4.44)

Hence, total momentum balance:

(mnv) + mn(v.)v + mv [.(nv)] + p = FEM

t

(4.45)

Use the continuity equation:

n
+ .(nv) = 0 ,
t
to cancel the third term and part of the 1st:

n
v
v
(mnv) + mv (. (nv)) = mv{
+ . (nv)} + mn
= mn
t
t
t
t

(4.46)

(4.47)

Then take p to RHS to get nal form:

Momentum Equation:

v
mn
+ (v.) v = nq (E + v B) p .
t
As before, via Lagrangian formulation. (Collisions have been ignored.)

70

(4.48)

4.2.5

Eect of Collisions

First notice that like particle collisions do not change the total momentum (which is averaged

over all particles of that species).

Collisions between unlike particles do exchange momentum between the species. Therefore

once we realize that any quasineutral plasma consists of at least two dierent species (elec

trons and ions) and hence two dierent interpenetrating uids we may need to account for

another momentum loss (gain) term.

The rate of momentum density loss by species 1 colliding with species 2 is:

12 n1 m1 (v1 v2 )

(4.49)

Hence we can immediately generalize the momentum equation to

m1 n1

v1
+ (v1 .) v1 = n1 q1 (E + v1 B) p1 12 n1 m1 (v1 v2 )
t

(4.50)

With similar equation for species 2.

4.3

The Key Question for Momentum Equation:

What do we take for p?

Basically p = nT is determined by energy balance, which will tell how T varies. We could

write an energy equation in the same way as momentum. However, this would then contain

a term for heat ux, which would be unknown. In general, the k th moment equation contains

a term which is a (k + 1)th moment.

Continuity, 0th equation contains v determined by

Momentum, 1st equation contains p determined by

Energy,
2nd equation contains Q determined by . . .
In order to get a sensible result we have to truncate this hierarchy. Do this by some sort of
assumption about the heat ux. This will lead to an
Equation of State:
pn = const.
(4.51)
The value of to be taken depends on the heat ux assumption and on the isotropy (or
otherwise) of the energy distribution.
Examples
1. Isothermal: T = const.:

= 1.

2. Adiabatic/Isotropic: 3 degrees of freedom = 35 .

3. Adiabatic/1 degree of freedom

= 3.
71

4. Adiabatic/2 degrees of freedom

In general, n(/2)T = p(V /V )

= 2.
(Adiabatic degrees)

T
V
=
2 T
V

= +

n
n

(4.52)

So
p
n T
2
=
+
= 1+
p
n
T

i.e.

n
,
n

pn(1+ ) = const.

(4.53)

(4.54)

In a normal gas, which holds together by collisions, energy is rapidly shared between 3
spacedegrees of freedom. Plasmas are often rather collisionless so compression in 1 dimension
often stays conned to 1degree of freedom. Sometimes heat transport is so rapid that the
isothermal approach is valid. It depends on the exact situation; so lets leave undened
for now.

4.4

Summary of TwoFluid Equations

Species j
Plasma Response
1. Continuity:
nj
+ .(nj vj ) = 0
t

(4.55)

vj
+ (vj .) vj = nj qj (E + vj B) pj jk nj mj (vj vk )
t

(4.56)

2. Momentum:

mj nj

3. Energy/Equation of State:
pj n
j = const..

(4.57)

(j = electrons, ions).
Maxwells Equations
.B = 0
1 E
B = o j + 2
c t

72

.E = /o
B
E=
t

(4.58)
(4.59)

With

= qe ne + qi ni = e (ne + Zni )
j = qe ne ve + qi ni vi = e (ne ve + Zni vi )
(Quasineutral)
= ene (ve vi r)

(4.60)
(4.61)
(4.62)

Accounting
Unknowns
n e , ni
2
ve , vi
6
p e , pi
2
E, B
6
16

Equations
Continuity e, i
Momentum e, i
State e, i
Maxwell

2
6
2
8
18

but 2 of Maxwell (. equs) are redundant because can be deduced from others: e.g.

(.B)
t

1
1

(.E) = 2
+ .E
and . ( B) = 0 = o .j + 2
c t
o
c t
. ( E) = 0 =

(4.63)
(4.64)

So 16 equs for 16 unknowns.

Equations still very dicult and complicated mostly because it is Nonlinear
In some cases can get a tractable problem by linearizing. That means, take some known
equilibrium solution and suppose the deviation (perturbation) from it is small so we can
retain only the 1st linear terms and not the others.

4.5

TwoFluid Equilibrium: Diamagnetic Current

Slab:
Straight Beld:

,
y z

= 0

= 0.

B = Bz.

Equilibrium:

Collisionless:

0.

=0

(E = )

Momentum Equation(s):
mj nj (vj .)vj = nj qj (E + vj B) pj

(4.65)

Drop j sux for now. Then take x, y components:

d
dp
vx = nq(Ex + vy B)
dx
dx
d
mn vx vy = nq(0 vx B)
dx

mn vx

73

(4.66)
(4.67)

Eq 4.67 is satised by taking vx = 0. Then 4.66

nq(Ex + vy B)
i.e.
vy =

dp
= 0.
dx

Ex
1 dp
+
B
nqB dx

(4.68)

(4.69)

or, in vector form:

v=

EB

2
B

EB drift

p B

nq B 2

(4.70)

Diamagnetic Drift

Notice:
In magnetic eld () uid velocity is determined by component of momentum equation
orthogonal to it (and to B).
Additional drift (diamagnetic) arises in standard F B form from pressure force.
Diagmagnetic drift is opposite for opposite signs of charge (electrons vs. ions).
Now restore species distinctions and consider electrons plus single ion species i. Quasineu
trality says ni qi = ne qe . Hence adding solutions
ne qe ve + ni qi vi =

EB
B
(ni qi + ne qe ) (pe + pi ) 2
2

B
B

(4.71)

B
B2

(4.72)

=0

Hence current density:

j = (pe + pi )

This is the diamagnetic current. The electric eld, E, disappears because of quasineutrality.

(General case j qj nj vj = ( pj )B/B 2 ).

4.6 Reduction of Fluid Approach to the Single Fluid

Equations
So far we have been using uid equations which apply to electrons and ions separately.
These are called Two Fluid equations because we always have to keep track of both uids
separately.
A further simplication is possible and useful sometimes by combining the electron and ion
equations together to obtain equations governing the plasma viewed as a Single Fluid.

74

Recall 2uid equations:

nj
+ .(nj vj ) = 0.
(4.73)
t

mj nj
+ vj . vj = nj qj (E + vj B) pj + Fjk (4.74)
t

Continuity (Cj )
Momentum (Mj )

(where we just write Fjk = vjk nj mj (vj vk ) for short.)

Now we rearrange these 4 equations (2 2 species) by adding and subtracting appropriately
to get new equations governing the new variables:

Mass Density
m
C of M Velocity
V
Charge density
q
Electric Current Density j

=
=
=
=
=
p =

Total Pressure

ne me + ni mi
(ne me ve + ni mi vi ) /m
qe n e + qi n i
qe ne ve + qi ni vi
qe ne (ve vi ) by quasi neutrality
pe + pi

(4.75)
(4.76)
(4.77)
(4.78)
(4.79)
(4.80)

1st equation: take me Ce + mi Ci

m
+ . (m V) = 0
t

(1)

Mass Conservation

(4.81)

2nd take qe Ce + qI Ci
(2)

q
+ .j = 0
t

Charge Conservation

(4.82)

3rd take Me + Mi . This is a bit more dicult. RHS becomes:

nj qj (E + vj B) pj + Fjk = q E + j B (pe + pi )

(4.83)

(we use the fact that Fei Fie so no net friction). LHS is

mj nj

+ vj . vj
t

(4.84)

The diculty here is that the convective term is nonlinear and so does not easily lend itself
to reexpression in terms of the new variables. But note that since me << mi the contribution
from electron momentum is usually much less than that from ions. So we ignore it in this
equation. To the same degree of approximation V vi : the CM velocity is the ion velocity.
Thus for the LHS of this momentum equation we take

mi ni

+ vj . vj m
+ V. V
t
t
75

(4.85)

so:

(3)
qe
Me
me

Finally we take

n j qj

qi
Mi
mi

+ V. V = q E + j B p
t

(4.86)

to get:

nj qj2
qj
qj

+ (vj .) vj =
(E + vj B)
pj +
Fjk }
{
t
mj
mj
mj
j

(4.87)

Again several diculties arise which it is not very protable to deal with rigorously.
Observe

that the LHS can be written (using quasineutrality ni qi + ne qe = 0) as m t jm provided

we discard the term in (v.)v. (Think of this as a linearization of this question.) [The
(v.)v convective term is a term which is not satisfactorily dealt with in this approach to
the single uid equations.]
In the R.H.S. we use quasineutrality again to write
nj qj2

mj

nj qq2

mj

E =

n2e qe2

1
1
mi ni + me ne
q e qi
+
E = n2e qe2
E=
m E, (4.88)
ne me ni mi
ne me ni mi
me mi

ne qe2
ni qi2
ve +
vi
me
mi

qe qi n e qe m i
ni qi me

=
{
ve +
vi }
me mi
qe
qi

qe qi
mi me
=
{ne me ve + ni mi vi
+
(qe ne ve + qi ni vi )}
me mi
qe
qi

qe qi
mi me
=
{m V
+
j}
qe
me mi
qi

vj =

(4.89)

Also, remembering Fei = ei ne mi (ve vi ) = Fie ,

qj
me
Fjk = ei ne qe ne qi
(ve vi )
mi
j mj

= ei

qe m e
1
j
qi m i

(4.90)

So we get

m
t

j
m

qe qi
mi me
=
m E + m V
+
j B
me mi
qi
qe

qi
qe m e
qe
pe
pi 1

ei j
me
mi
qi m i

Regroup after multiplying by

(4.91)

me mi
:
q e q i m

me mi
j
1 mi me
E+VB =
+
+
jB
m q i
qe
qe qi t m

qi
me mi
qe m e m e m i
qe
pe +
pi
1
ei j

mi
m qe qi
q i m i q e q i m
me
76

(4.92)

Notice that this is an equation relating the Electric eld in the frame moving with the uid

(L.H.S.) to things depending on current j i.e. this is a generalized form of Ohms Law.

One essentially never deals with this full generalized Ohms law. Make some approximations

recognizing the physical signicance of the various R.H.S. terms.

me mi
qe qi t

j
m

arises from electron inertia.

it will be negligible for low enough frequency.

1
m

mi me
+
j B is called the Hall Term.
qi
qe

and arises because current ow in a Beld tends to be diverted across the magnetic eld.
It is also often dropped but the justication for doing so is less obvious physically.
qi
qe
pi term
pe for comparable pressures,
mi
me
and the latter is the Hall term; so ignore qi pi /mi .
Last term in j has a coecient, ignoring me /mi c.f. 1 which is
me ei
me mi ei
= 2
=
qe qi (ni mi )
qe n e

the resistivity.

(4.93)

Hence dropping electron inertia, Hall term and pressure, the simplied Ohms law becomes:
E + V B = j

(4.94)

Final equation needed: state:

i
e
pe n
+ pi n
= constant.
i
e

Take quasineutrality ne ni m . Take e = i , then

p
m = const.

4.6.1

(4.95)

Summary of Single Fluid Equations: M.H.D.

Mass Conservation :
Charge Conservation :

Momentum :
Ohm s Law :
Eq. of State :

m
+ (m V) = 0
t
q
+ .j = 0
t

(4.96)
(4.97)

+ V. V = q E + j B p
t

(4.98)

E + V B = j

(4.99)

p
m = const.

(4.100)

77

4.6.2

Heuristic Derivation/Explanation

Mass Charge: Obvious.

M om

+ V. V =
t

jB

Electric
body force

rate of change of
total momentum density

q E +

Magnetic Force
on current

(4.101)

Pressure

Ohms Law
The electric eld seen by a moving (conducting) uid is E + V B = EV electric eld in
frame in which uid is at rest. This is equal to resistive electric eld j:
EV = E + V B = j

(4.102)

The q E term is generally dropped because it is much smaller than the j B term. To see
this, take orders of magnitude:
.E = q /0

1 E
B = 0 j + 2
c t

so q E0 /L

(4.103)

so E = j B/0 L

(4.104)

Therefore
q E
0

jB
L

B
0 L

L0
L2 /c2

=
B2
(0 L2 )2

light transit time

resistive skin time

(4.105)

This is generally a very small number. For example, even for a small cold plasma, say Te = 1
eV ( 2 103 mho/m), L = 1 cm, this ratio is about 108 .
Conclusion: the q E force is much smaller than the j B force for essentially all practical
cases. Ignore it.
Normally, also, one uses MHD only for low frequency phenomena, so the Maxwell displace
ment current, E/c2 t can be ignored.
Also we shall not need Poissons equation because that is taken care of by quasineutrality.

4.6.3

Maxwells Equations for MHD Use

.B = 0 ; E =

B
; B = o j .
t

(4.106)

The MHD equations nd their major use in studying macroscopic magnetic connement
problems. In Fusion we want somehow to conne the plasma pressure away from the walls
of the chamber, using the magnetic eld. In studying such problems MHD is the major tool.
On the other hand if we focus on a small section of the plasma as we do when studying
shortwavelength waves, other techniques: 2uid or kinetic are needed. Also, plasma is
approx. uniform.
78

Macroscopic Phenomena MHD

Microscopic Phenomena 2Fluid/Kinetic

4.7

MHD Equilibria

Study of how plasma can be held by magnetic eld. Equilibrium V =

equations reduce. Mass and Faradays law are automatic. We are left with
(M omm ) Force Balance 0 = j B p
Ampere

B = o j

= 0. So

(4.107)
(4.108)

Plus .B = 0, .j = 0.

Notice that provided we dont ask questions about Ohms law. E doesnt come into MHD

equilibrium.

These deceptively simple looking equations are the subject of much of Fusion research. The

hard part is taking into account complicated geometries.

We can do some useful calculations on simple geometries.

4.7.1

pinch

Figure 4.5: pinch conguration.

So called because plasma currents ow in direction.

Use MHD Equations

Take to be length, uniform in zdir.

By symmetry B has only z component.

By symmetry j has only comp.

By symmetry p has only r comp.

79

So we only need

(j B)r (p)r = 0
( B) = (o j)

p=0
i.e.
j Bz
r

Bz = o j
r
Bz Bz p

= 0
o r
r

Force
Ampere

Eliminate j :

(4.109)
(4.110)
(4.111)
(4.112)
(4.113)

i.e.
Bz2
+p = 0
2o
Bz2
Solution
+ p = const.
2o

(4.114)
(4.115)

Figure 4.6: Balance of kinetic and magnetic pressure

Bz2
B2
+ p = z ext
2o
2o

(4.116)

[Recall Single Particle Problem]

Think of these as a pressure equation. Equilibrium says total pressure = const.
Bz2
2o

kinetic pressure

magnetic pressure

80

= const.

(4.117)

Ratio of kinetic to magnetic pressure is plasma .

=

2o p
Bz2

(4.118)

measures eciency of plasma connement by B. Want large for fusion but limited by
instabilities, etc.

4.7.2

Zpinch

Figure 4.7: Zpinch conguration.

so called because j ows in zdirection. Again take to be length and uniform.
z
j = jz e

Force
Ampere

B = B e

(4.119)

(j B)r (p)r = jz B
( B)z (o j)z =

p
=0
r

1
(rB ) o jz = 0
r r

(4.120)
(4.121)

Eliminate j:
B
p
(rB )
=0
o r r
r
or

B2
o r

Extra Term

+
r

B2
2o

(4.122)

p =0

(4.123)

Magnetic+Kinetic pressure

Extra term acts like a magnetic tension force. Arises because Beld lines are curved.

81

Can integrate equation

B2 dr
B2
+
+ p(r)
o r
2o

=0

(4.124)

If we choose b to be edge (p(b) = 0) and set a = r we get

Figure 4.8: Radii of integration limits.

p(r) =

B2 (b) B2 (r) b B2 dr
+

2o
2o
r o r

(4.125)

Force balance in zpinch is somewhat more complicated because of the tension force. We
cant choose p(r) and j(r) independently; they have to be self consistent.
Example

j = const.
1
o jz
(rB ) = o jz B =
r
r r
2

(4.126)

Hence
b
1 o jz 2 2
{b r2 +
2r dr }
2o
2
r
o jz2 2
{b r2 }
=
4

p(r) =

(4.127)
(4.128)

Figure 4.9: Parabolic Pressure Prole.

Also note B (b) =

o jz b
2

so
p=

2
Bb
2 2
{b r2 }
2o b2

82

(4.129)

4.7.3

Stabilized Zpinch

Also called screw pinch, z pinch or sometimes loosely just zpinch.

Zpinch with some additional Bz as well as B
j Bz jz B

(Force)r

=0
r

Bz = o j
r

Ampere :

(4.130)
(4.131)

1
(rb ) = o jz
r r

(4.132)

Eliminate j:

or

Bz Bz
B
p

(rB )
= 0
o r r
r
o r

B2
o r

+
r

B2
2o

Mag Tension
only

4.8

(4.133)

= 0

(4.134)

Mag (+z)+Kinetic pressure

Some General Properties of MHD Equilibria

4.8.1 Pressure & Tension

j B p = 0

B = o j

(4.135)

We can eliminate j in the general case to get

1
( B) B = p.
o

(4.136)

Expand the vector triple product:

p =
put b =

B
|B|

1
1
(B.) B
B 2
o
2o

(4.137)

so that B = Bb = Bb + bB. Then

1 2
1
{B (b.) b + Bb (b.) B}
B 2
o
2o
B2
1
=
( b (b.)) B 2
(b.) b
o
2o

B2
B2
=
(b.) b
o
2p

p =

83

(4.138)
(4.139)
(4.140)

B
Now 2
is the perpendicular (to B) derivative of magnetic pressure and (b.)b is the
o
curvature of the magnetic eld line giving tension.

| (b.)b | has value

4.8.2

1
.
R

R: radius of curvature.

Magnetic Surfaces
0 = B. [j B p] = B.p

(4.141)

*Pressure is constant on a eld line (in MHD situation).

(Similarly, 0 = j.[j B p] = j.p.)

Figure 4.10: Contours of pressure.

Consider some arbitrary volume in which p = 0. That is, some plasma of whatever shape.
Draw contours (surfaces in 3d) on which p = const. At any point on such an isoberic surface
p is perp to the surface. But B.p = 0 implies that B is also perp to p.

Figure 4.11: B is perpendicular to p and so lies in the isobaric surface.

Hence

B lies in the surface p = const.

In equilibrium isobaric surfaces are magnetic surfaces.

[This argument does not work if p = const. i.e. p = 0. Then there need be no magnetic

surfaces.]

84

4.8.3

Current Surfaces

Since j.p = 0 in equilibrium the same argument applies to current density. That is

j lies in the surface p = const.

Isobaric Surfaces are Current Surfaces.

Moreover it is clear that

Magnetic Surfaces are Current Surfaces.

(since both coincide with isobaric surfaces.)

[It is important to note that the existence of magnetic surfaces is guaranteed only in the

MHD approximation when p =

0 > Taking account of corrections to MHD we may not

have magnetic surfaces even if p = 0.]

4.8.4

Low equilibria: ForceFree Plasmas

In many cases the ratio of kinetic to magnetic pressure is small, << 1 and we can approx
imately ignore p. Such an equilibrium is called force free.
jB=0

(4.142)

j = (r)B

(4.143)

implies j and B are parallel.

i.e.
Current ows along eld lines not across. Take divergence:
0 = .j = . ( (r) B) = (r) .B + (B.)
= (B.) .

(4.144)
(4.145)

The ratio j/B = is constant along eld lines.

is constant on a magnetic surface. If there are no surfaces, is constant everywhere.

Example: ForceFree Cylindrical Equil.

j B =
j = (r)B
B = o j = o (r)B

(4.146)
(4.147)

This is a somewhat more convenient form because it is linear in B (for specied (r)).
B = o B

Constant :
leads to a Bessel function solution

85

(4.148)

Bz = Bo Jo (o r)
B = Bo J1 (o r)

(4.149)
(4.150)

for o r > 1st zero of Jo the toroidal eld reverses. There are plasma connement schemes
with const. Reversed Field Pinch.

4.9

Toroidal Equilibrium

Bend a zpinch into a torus

Figure 4.12: Toroidal zpinch

B elds due to current are stronger at small R side Pressure (Magnetic) Force outwards.
Have to balance this by applying a vertical eld Bv to push plasma back by j Bv .

Figure 4.13: The eld of a toroidal loop is not an MHD equilibrium. Need to add a vertical
eld.
Bend a pinch into a torus: B is stronger at small R side outward force.
Cannot be balanced by Bv because no j . No equilibrium for a toroidally symmetric pinch.

Underlying Single Particle reason:

Toroidal pinch has B only. As we have seen before, curvature drifts are uncompensated

in such a conguration and lead to rapid outward motion.

86

Figure 4.14: Chargeseparation giving outward drift is equivalent to the lack of MHD toroidal
force balance.
We know how to solve this: Rotational Transform: get some B . Easiest way: add j . From
MHD viewpoint this allows you to push the plasma back by j Bv force. Essentially, this
is Tokamak.

4.10

Plasma Dynamics (MHD)

When we want to analyze nonequilibrium situations we must retain the momentum terms.
This will give a dynamic problem. Before doing this, though, let us analyse some purely
Kinematic Eects.
Ideal MHD
Set eta = 0 in Ohms Law.
A good approximation for high frequencies, i.e. times shorter than resistive decay time.
Ideal Ohm s Law.

E + V B = 0.

(4.151)

Also

B
Faraday s Law.
(4.152)
t
Together these two equations imply constraints on how the magnetic eld can change with
time: Eliminate E:
B
+ (V B) = +
(4.153)
t
This shows that the changes in B are completely determined by the ow, V.
E=

4.11

Flux Conservation

Consider an arbitrary closed contour C and spawning surface S in the uid.

Flux linked by C is
=

B.ds

Let C and S move with uid:

Total rate of change of is given by two terms:

87

(4.154)

Figure 4.15: Motion of contour with uid gives convective ux derivative term.

B
.ds
+
B.(V dl)
C
t

Due to changes in B Due to motion of C

=
=

E.ds

(V B).dl

(4.155)

(4.156)

(E + V B).dl = 0 by Ideal Ohm s Law.

(4.157)

Flux through any surface moving with uid is conserved.

4.12

Field Line Motion

Think of a eld line as the intersection of two surfaces both tangential to the eld everywhere:

Figure 4.16: Field line dened by intersection of two ux surfaces tangential to eld.
Let surfaces move with uid.
Since all parts of surfaces had zero ux crossing at start, they also have zero after, (by ux

conserv.).

Surfaces are tangent after motion

88

 Their intersection denes a eld line after.

We think of the new eld line as the same line as the old one (only moved).
Thus:
1. Number of eld lines ( ux) through any surface is constant. (Flux Cons.)
2. A line of uid that starts as a eld line remains one.

4.13

MHD Stability

The fact that one can nd an MHD equilibrium (e.g. zpinch) does not guarantee a useful
connement scheme because the equil. might be unstable. Ball on hill analogies:

Figure 4.17: Potential energy curves

An equilibrium is unstable if the curvature of the Potential energy surface is downward
d2
away from equil. That is if dx
2 {Wpot } < 0.

In MHD the potential energy is Magnetic + Kinetic Pressure (usually mostly magnetic).

If we can nd any type of perturbation which lowers the potential energy then the equil is

unstable. It will not remain but will rapidly be lost.

Example Zpinch

We know that there is an equilibrium: Is it stable?

Consider a perturbation thus:

Figure 4.18: Sausage instability

Simplify the picture by taking the current all to ow in a skin. We know that the pressure
2
is supported by the combination of B 2 /2o pressure and Bo r tension forces.
89

Figure 4.19: Skincurrent, sharp boundary pinch.

At the place where it pinches in (A)

B and 1r increase Mag. pressure & tension increase inward force no longer balance

by p perturbation grows.

At place where it bulges out (B)

B & 1r decrease Pressure & tension perturbation grows.

Conclusion a small perturbation induces a force tending to increase itself. Unstable ( W <
0).

4.14

General Perturbations of Cylindrical Equil.

Look for things which go like exp[i(kz + m)]. [Fourier (Normal Mode) Analysis].

Figure 4.20: Types of kink perturbation.

0
Generally Helical in form (like a screw thread). Example: m = 1 k =

4.15

zpinch

General Principles Governing Instabilities

(1) They try not to bend eld lines. (Because bending takes energy).
90

Perturbation

Figure 4.21: Driving force of a kink. Net force tends to increase perturbation. Unstable.

Figure 4.22: Alignment of perturbation and eld line minimizes bending energy.
(Constant surfaces) lies along magnetic eld.
Example: pinch type plasma column:

Figure 4.23: Flute or Interchange modes.

Preferred Perturbations are Flutes as per Greek columns Flute Instability. [Better
name: Interchange Instability, arises from idea that plasma and vacuum change places.]
(2) Occur when a heavier uid is supported by a lighter (Gravitational analogy).

Why does water fall out of an inverted glass? Air pressure could sustain it but does not

because of RayleighTaylor instability.

Similar for supporting a plasma by mag eld.

(3) Occur when | B | decreases away from the plasma region.

91

Figure 4.24: Inverted water glass analogy. Rayleigh Taylor instability.

B2
BA2
< B
2o
2o

(4.158)

Perturbation Grows.
(4) Occur when eld line curvature is towards the plasma (Equivalent to (3) because of
B = 0 in a vacuum).

Figure 4.25: Vertical upward eld gradient is unstable.

92

Figure 4.26: Examples of magnetic congurations with good and bad curvature.

93

4.16

Quick and Simple Analysis of Pinches

pinch
| B | = const. outside pinch
No eld line curvature. Neutral stability
| B | away from plasma outside
zpinch
Bad Curvature (Towards plasma) Instability.

Generally it is dicult to get the curvature to be good everywhere. Often it is sucient

to make it good on average on a eld line. This is referred to as Average Minimum B.

Tokamak has this.

General idea is that if eld line is only in bad curvature over part of its length then to perturb

in that region and not in the good region requires eld line bending:

Figure 4.27: Parallel localization of perturbation requires bending.

But bending is not preferred. So this may stabilize.

Possible way to stabilize conguration with bad curvature: Shear

Shear of Field Lines

Figure 4.28: Depiction of eld shear.

Direction of B changes. A perturbation along B at z3 is not along B at z2 or z1 so it would

have to bend eld there Stabilizing eect.

General Principle: Field line bending is stabilizing.

Example: Stabilized zpinch

Perturbations (e.g. sausage or kink) bend Bz so the tension in Bz acts as a restoring force

to prevent instability. If wave length very long bending is less. Least stable tends to be

longest wave length.

Example: Cylindrical Tokamak

94

Tokamak is in some ways like a periodic cylindrical stabilized pinch. Longest allowable wave
length = 1 turn round torus the long way, i.e.
kR = 1 :

= 2R.

(4.159)

Express this in terms of a toroidal mode number, n (s.t. perturbation exp i(n + m):
= Rz n = kR.
Most unstable mode tends to be n = 1.

[Careful! Tokamak has important toroidal eects and some modes can be localized in the

1).

bad curvature region (n =

Figure 4.29: Ballooning modes are localized in the outboard, bad curvature region.

95

Chapter 5

Electromagnetic Waves in Plasmas

5.1 General Treatment of Linear Waves in Anisotropic

Medium
Start with general approach to waves in a linear Medium: Maxwell:
B = o j +

1 E
B
; E =
2
c t
t

(5.1)

we keep all the mediums response explicit in j. Plasma is (innite and) uniform so we Fourier
analyze in space and time. That is we seek a solution in which all variables go like
exp i(k.x t) [real part of]

(5.2)

It is really the linearised equations which we treat this way; if there is some equilibrium eld

OK but the equations above mean implicitly the perturbations B, E, j, etc.

Fourier analyzed:

i
E ;
c2
Eliminate B by taking k second eq. and 1st
ik B = o j +

ik E = iB

ik (k E) = o j

i 2
E
c2

(5.3)

(5.4)

So

2
(5.5)
E + io j = 0
c2
Now, in order to get further we must have some relationship between j and E(k, ). This
will have to come from solving the plasma equations but for now we can just write the most
general linear relationship j and E as
k (k E) +

j = .E
96

(5.6)

 is the conductivity tensor. Think of this equation as a matrix e.g.:

jx
xx xy ...
Ex

j
...
...
...
=
y

Ey
jz
... ... zz
Ez

(5.7)

This is a general form of Ohms Law. Of course if the plasma (medium) is isotropic (same
in all directions) all odiagonal s are zero and one gets j = E.
Thus

2
E + io .E = 0
(5.8)
c2
Recall that in elementary E&M, dielectric media are discussed in terms of a dielectric con
stant and a polarization of the medium, P, caused by modication of atoms. Then
k(k.E) k 2 E +

o E =

Displacement

and .D =

Polarization

ext

(5.9)

externalcharge

and one writes

P=

o E

(5.10)

susceptibility
Our case is completely analogous, except we have chosen to express the response of the
medium in terms of current density, j, rather than polarization P For such a dielectric
medium, Amperes law would be written:
1
D

B = jext +
= o E,
o
t
t

if jext = 0 ,

(5.11)

where the dielectric constant would be = 1 + .

Thus, the explicit polarization current can be expressed in the form of an equivalent dielectric
expression if
E
E

j + o
= .E +
= o .E
(5.12)
t
t
t
or

= 1 +
(5.13)
io
Notice the dielectric constant is a tensor because of anisotropy. The last two terms come
from the RHS of Amperes law:

j + (o E) .
(5.14)
t
If we were thinking in terms of a dielectric medium with no explicit currents, only implicit (in

) we would write this t

(o E); the dielectric constant. Our medium is possibly anisotropic

so we need t (o .E) dielectric tensor. The obvious thing is therefore to dene

=1+

1
io c2
=1+

io

97

(5.15)

Then

2
.E = 0
c2
and we may regard (k, ) as the dielectric tensor.
k(k.E) k 2 E +

(5.16)

Write the equation as a tensor multiplying E:

D.E = 0

(5.17)

with

2
}
c2
Again this is a matrix equation i.e. 3 simultaneous homogeneous eqs. for E.
D = {kk k 2 1 +

(5.18)

Dxx Dxy ...

Ex

D
...
...
yx
Ey = 0
...
... Dzz
Ez

(5.19)

In order to have a nonzero E solution we must have

det | D |= 0.

(5.20)

This will give us an equation relating k and , which tells us about the possible wavelengths
and frequencies of waves in our plasma.

5.1.1

Simple Case. Isotropic Medium

= 1
= 1

(5.21)
(5.22)

Take k in z direction then write out the Dispersion tensor D.

0 0 0
k 2 0 0

c2

2
D = 0 0 0 0 k 0 + 0
0 0 kk
0 0 k2
0

kk

k2 1

k 2 +

c2

0
2
k +
0

Take determinant:
2
det |
D
|= k + 2
c

2
0

c2
2
0 c2
2
c2

0
2

c
2

c2
2

2
= 0.
c2

Two possible types of solution to this dispersion relation:

98

(5.23)

(5.24)

(A)
k2 +

0 0

0 0
0 0

2
= 0.
c2

(5.25)

0
Ex

0
E
y
=0
2
Ez

c2

Ez = 0.

(5.26)

Electric eld is transverse (E.k = 0)

Phase velocity of the wave is

c
=
k

(5.27)

This is just like a regular EM wave traveling in a medium with refractive index
N

kc
= .

(5.28)

(B)

Dxx

0
0

2
=0
i.e. = 0
c2

0 0
Ex

Dyy 0

E
y = 0
Ex = Ey = 0.
0 0
Ez

Electric Field is Longitudinal (E k = 0)

(5.29)
(5.30)

E k.

This has no obvious counterpart in optics etc. because is not usually zero. In plasmas

= 0 is a relevant solution. Plasmas can support longitudinal waves.

5.1.2

General Case

(k in zdirection)

N 2 + xx
xy
xz

yx
N 2 + yy yz
D =
2

c
zy
zx
zz
2

k 2 c2
N = 2

(5.31)

When we take determinant we shall get a quadratic in N 2 (for given ) provided is not
explicitly dependent on k. So for any there are two values of N 2 . Two modes. The
polarization E of these modes will be in general partly longitudinal and partly transverse.
The point: separation into distinct longitudinal and transverse modes is not possible in
anisotropic media (e.g. plasma with Bo ).
All we have said applies to general linear medium (crystal, glass, dielectric, plasma). Now
we have to get the correct expression for and hence by analysis of the plasma (uid)
equations.

99

5.2

High Frequency Plasma Conductivity

We want, now, to calculate the current for given (Fourier) electric eld E(k, ), to get the
conductivity, . It wont be the same as the DC conductivity which we calculated before
(for collisions) because the inertia of the species will be important. In fact, provided
ei

(5.32)

we can ignore collisions altogether. Do this for simplicity, although this approach can be
generalized.
Also, under many circumstances we can ignore the pressure force p. In general will
be true if k vte,i We take the plasma equilibrium to be at rest: vo = 0. This gives a
manageable problem with wide applicability.
Approximations:
Collisionless
ei = 0
Cold Plasma p = 0 (e.g. T 0)
Stationary Equil vo = 0

5.2.1

(5.33)

Zero Beld case

To start with take Bo = 0: Plasma isotropic Momentum equation (for electrons rst)

v
mn
+ (v.)v = nqE
t

(5.34)

Notice the characteristic of the cold plasma approx. that we can cancel n from this equation
and on linearizing get essentially the single particle equation.
m

v1
= qE
t

(Drop the 1 sux now).

(5.35)

This can be solved for given as

q
E
im
and the current (due to this species, electrons) is
v=

j = nqv =

(5.36)

nq 2
E
im

(5.37)

nq 2
m

(5.38)

So the conductivity is
=i
Hence dielectric constant is
i
nq 2
=1+
=1
o
mo

100

1
=1+
2

(5.39)

Longitudinal Waves (Bo = 0)

Dispersion relation we know is
nq 2
=0=1
mo

1
2

(5.40)

[Strictly, the we want here is the total including both electron and ion contributions to
the conductivity. But
mi
e

(for z = 1)
(5.41)
i
me
so to a rst approximation, ignore ion motions.]
Solution

ne qe2
.
me o

(5.42)

In this approx. longitudinal oscillations of the electron uid have a single unique frequency:

p =

ne e2
m e o

1
2

(5.43)

This is called the Plasma Frequency (more properly pe the electron plasma frequency).
If we allow for ion motions we get an ion conductivity
i =

ini qi2
mi

(5.44)

and hence
i
ne qe2
ni qi2
= 1+
(e + i ) = 1
+
o
o m e o m i

tot

1
2

(5.45)

2
2
= 1 pe
+ pi
/ 2

where

pi

ni qi2
o mi

1
2

(5.46)

is the Ion Plasma Frequency.

Simple Derivation of Plasma Oscillations
Take ions stationary; perturb a slab of plasma by shifting electrons a distance x. Charge
built up is ne qx per unit area. Hence electric eld generated
E=

n e qe x
o

101

(5.47)

Figure 5.1: Slab derivation of plasma oscillations

Equation of motion of electrons
me
i.e.

dv
ne q 2 x
= e ;
dt
o

d2 x
ne qe2
x=0
+
dt2
o me

(5.48)

(5.49)

Simple harmonic oscillator with frequency

pe =

ne qe2
o me

1
2

Plasma Frequency.

(5.50)

The Characteristic Frequency of Longitudinal Oscillations in a plasma. Notice

1. = p for all k in this approx.
2. Phase velocity

can have any value.

3. Group velocity of wave, which is the velocity at which information/energy travel is

vg =

d
= 0 !!
dk

(5.51)

In a way, these oscillations can hardly be thought of as a proper wave because they do
not transport energy or information. (In Cold Plasma Limit). [Nevertheless they do emerge
from the wave analysis and with less restrictive approxs do have nite vg .]
Transverse Waves (Bo = 0)
Dispersion relation:
k2 +

2
=0
c2

102

(5.52)

or

k 2 c2
2
2
2
=

=
1

pe
pi /
2
2
1 pe
/ 2

N2

(5.53)

Figure 5.2: Unmagnetized plasma transverse wave.

Figure 5.3: Alternative dispersion plot.

Alternative expression:
2 p2
k + 2 2 =0
c
c
2

103

(5.54)

which implies

2 = p2 + k 2 c2
=

p2 + k 2 c2

(5.55)
1

(5.56)

Meaning of Negative N 2 : Cut O

5.2.2

When N 2 < 0 (for < p ) this means N is pure imaginary and hence so is k for real .
Thus the wave we have found goes like
exp{ | k | x it}

(5.57)

i.e. its space dependence is exponential not oscillatory. Such a wave is said to be Evanescent
or Cut O. It does not truly propagate through the medium but just damps exponentially.
Example:

Figure 5.4: Wave behaviour at cuto.

A wave incident on a plasma with p2 > 2 is simply reected, no energy is transmitted
through the plasma.

5.3

Cold Plasma Waves

(Magnetized Plasma)

Objective: calculate ,D, k(), using known plasma equations.

Approximation: Ignore thermal motion of particles.

Applicability: Most situations where (1) plasma pressure and (2) absorption are negligible.

Generally requires wave phase velocity vthermal .

104

5.3.1

Derivation of Dispersion Relation

Can derive the cold plasma approx from uid plasma equations. Simpler just to say that
all particles (of a specic species) just move together obeying Newtons 2nd law:
m

v
= q(E + v B)
t

(5.58)

Take the background plasma to have E0 = 0, B = B0 and zero velocity. Then all motion
is due to the wave and also the waves magnetic eld can be ignored provided the particle
speed stays small. (This is a linearization).
m

v
= q(E + v B0 ),
t

(5.59)

where v, E exp i(k.x t) are wave quantities.

Substitute

i and write out equations. Choose axes such that B0 = B0 (0, 0, 1).
imvx = q(Ex + vy B0 )
imvy = q(Ey vx B0 )
imvz = qEz

(5.60)

Solve for v in terms of E.

q iEx Ey
m
2 2

q Ex + iEy
=
m
2 2
q i

=
Ez

vx =
vy
vz

(5.61)

0
where = qB
is the gyrofrequency but its sign is that of the charge on the particle species
m
under consideration.

Since the current is j = qvn = .E we can identify the conductivity tensor for the species

(j) as:

j =

qj2 nj
i
mj 2 2j
qj2 nj j
mj 2 2j

qj2 nj j
mj 2 2j
2
qj nj
i
mj 2 2j

0
0
iqj2

nj
m

(5.62)

The total conductivity, due to all species, is the sum of the conductivities for each
=

(5.63)

So
xx = yy =

q12 nj
j

mj

105

i
2j

(5.64)

 qj2 nj

xy = yx =

mj

zz =

j
2j

(5.65)

qj2 nj i
j

Susceptibility =

(5.66)

mj

1
.
io

xx xy 0
S iD 0

S
0
=
yx yy 0 = iD

0
0 zz
0
0
P

(5.67)

where
xx = yy = S = 1

ixy = iyx = D =
zz = P = 1

and
2
pj

2
pj
2 2j

(5.68)

2
pj
2 2j

(5.69)

j
2
pj
2

(5.70)

qj2 nj

o m j

(5.71)

is the plasma frequency for that species.

S & D stand for Sum and Dierence:
1
1
S = (R + L)
D = (R L)
2
2
where R & L stand for Righthand and Lefthand and are:
R=1

2
pj
( + j )

, L=1

(5.72)

2
pj
( j )

(5.73)

The R & L terms arise in a derivation based on expressing the eld in terms of rotating
polarizations (right & left) rather than the direct Cartesian approach.
We now have the dielectric tensor from which to obtain the dispersion relation and solve
it to get k() and the polarization. Notice, rst, that is indeed independent of k so the
dispersion relation (for given ) is a quadratic in N 2 (or k 2 ).
Choose convenient axes such that ky = Ny = 0. Let be angle between k and B0 so that
Nz = N cos ,
Then

Nx = N sin .

(5.74)

N 2 cos2 + S
iD
N 2 sin cos

2
+iD
N + S
0
D =

0
N 2 sin cos
N 2 sin2 + P

106

(5.75)

and

D = AN 4 BN 2 + C

(5.76)

A S sin2 + P cos2
B RL sin2 + P S(1 + cos2 )
C P RL

(5.77)
(5.78)
(5.79)

where

Solutions are
N2 =

BF
,
2A

(5.80)

where the discriminant, F , is given by

F 2 = (RL P S)2 sin4 + 4P 2 D2 cos2

(5.81)

after some algebra. This is often, for historical reasons, written in the equivalent form (called
the AppletonHartree dispersion relation)
N2 = 1

2(A B + C)
2A B F

(5.82)

The quantity F 2 is generally +ve, so N 2 is real propagating or evanescent no wave

absorption for cold plasma.

Solution can also be written

tan2 =

P (N 2 R) (N 2 L)
(SN 2 RL) (N 2 P )

(5.83)

This compact form makes it easy to identify the dispersion relation at = 0 &
and perpendicular propagation tan = 0, .
Parallel:

P = 0 , N2 = R

Perp:

N2 =

RL
S

i.e. parallel

N2 = L

N2 = P .

Example: Righthand wave

N 2 = R. (Single Ion Species).
N2 = 1

2
2
pi
pe

( |e |) ( + |i |)

(5.84)

This has a wave resonance N 2 at = |e |, only. Righthand wave also has a cuto at
R = 0, whose solution proves to be

|e | |i | |e | + |i |
= R =
+
2
2
107

1/2
2
2
+ pe
+ pi

(5.85)

Since mi me this can be approximated as:

2
|e |
R
1 + 1 + 4 pe2
2

|e |

1
2

(5.86)

This is always above |e |.

Figure 5.5: The form of the dispersion relation for RH wave.

One can similarly investigate LH wave and perp propagating waves. The resulting wave
resonances and cutos depend only upon 2 properties (for specied ion mass) (1) Density
mi
2
pe
factors.]
(2) Magnetic Field |e |. [Ion values pi , |i | are got by m
e
These resonances and cutos are often plotted on a 2D plane

| e |

p2
2

( B, n) called the

C M A Diagram.

We dont have time for it here.

5.3.2

Hybrid Resonances

Extraordinary wave

N2 =

N2 =

( + e ) ( + i )

2
pe

Perpendicular Propagation

RL
S

( + i )

2
pi

( + e )

( e ) ( i )

2
pe

( i ) ...

2 ( 2 2 ) 2 ( 2 2 )
( 2 2e ) ( 2 2i ) pe
i
pi
e

(5.87)
2

Resonance is where denominator = 0. Solve the quadratic in and one gets

2
2
2 + e2 pi
2i
2 + e2 + pi
+ 2i

pe
2 = pe
2
2

108

2
2 2
+ pe
pi

(5.88)

Neglecting terms of order

me
mi

(e.g.

2
pi
2 )
pe

one gets solutions

2
U2 H = pe
+ 2e
2
LH
=

2
2e pi
2
2e + pe

Upper Hybrid Resonance.

(5.89)

Lower Hybrid Resonance..

(5.90)

2
At very high density, pe
2e
2
LH
|e ||i |

(5.91)

geometric mean of cyclotron frequencies.

2
At very low density, pe
2e
2
2
LH
pi

(5.92)

ion plasma frequency

2
Usually in tokamaks pe
2e . Intermediate.

Summary Graph ( > p )

Figure 5.6: Summary of magnetized dispersion relation

Cutos are where N 2 = 0.
Resonances are where N 2 .
Intermediate angles of propagation have refractive indices between the = 0,
shaded areas.
109

lines, in the

5.3.3

Whistlers

(Ref. R.A. Helliwell, Whistlers & Related Ionospheric Phenomena, Stanford UP 1965.)
For N 2 1 the right hand wave can be written
N2

2
pe
( |e |)

(N = kc/)

(5.93)

Group velocity is
d
vg =
=
dk

dk
d

Then since
N=

d
=
d

(5.94)

1
2

N
c

(5.95)

(|e | ) 2
we have

1
1 2

d
d
p 2
2
2
(N ) =
=

+
p
1
1
3
1
2
d
d (|e | ) 2
(|e | ) 2
(|e | ) 2
p /2
=

{(|e | ) + }
3
1
(|e | ) 2
2
p |e |
/2
=

3
1
(|
e |
) 2
2

Thus

(5.96)

c 2 (|e | ) 2
2
vg =
p |e |

(5.97)

Group Delay is
L
1

1
3

vg

2 (|e | ) 2

| e |

1
2

|e |

Figure 5.7: Whistler delay plot

110

3
2

(5.98)

Plot with

L
vg

as xaxis.

Resulting form explains downward whistle.

Lightning strike function excites all frequencies.

Lower ones arrive later.

Examples of actual whistler sounds can be obtained from http://wwwistp.gsfc.nasa.

gov/istp/polar/polar_pwi_sounds.html.

5.4

Thermal Eects on Plasma Waves

The cold plasma approx is only good for high frequency, N 2 1 waves. If is low or N 2 1
one may have to consider thermal eects. From the uid viewpoint, this means pressure.
Write down the momentum equation. (We shall go back to B0 = 0) linearized
mn

v1
= nqE1 p1 ;
t

(5.99)

remember these are the perturbations:

p = p0 + p1

(5.100)

Fourier Analyse (drop 1s)

mn(i)v = nqE ikp

(5.101)

The key question: how to relate p to v

Answer: Equation of state + Continuity
State
pn = const. (p0 + p1 ) (n0 + n1 ) = p0 n
0

(5.102)

Use Taylor Expansion

(p0 + p1 ) (n0 + n1 )
Hence

p0 n
0

p1
n1
1+

p0
n0

(5.103)

p1
n1
=
p0
n0

(5.104)

n
+ . (nv) = 0
t

(5.105)

Continuity

Linearise:

n1
+ . (n0 v1 ) = 0
t

n
+ n0 .v = 0
t

(5.106)

Fourier Transform
in1 + n0 ik.v1 = 0
111

(5.107)

i.e.

n1 = n0

k.v

(5.108)

Combine State & Continuity

n0 k.v
n1
k.v

p1 = p0
= p0
= p0
n0
no

(5.109)

Hence Momentum becomes

mn (i) v = nqE

ikp0
k.v

(5.110)

Notice Transverse waves have k.v = 0; so they are unaected by pressure.

Therefore we need only consider the longitudinal wave. However, for consistency let us

proceed as before to get the dielectric tensor etc.

Choose axes such that k = k

ez then obviously:
vx =
vz =

iq
Ex
m

vy =

iq
Ey
m

(5.111)

q
Ez
m i + (ik 2 p0 /mn)

(5.112)

Hence

1 0
inq
0 1
=
m
0 0

0
0

=1+

=
0

1
k 2 p0
1
mn 2

(5.113)

1
p2
0
0
p2
0
0
1
2
w2
0
0
1
2 kp2 p0

(5.114)

mn

(Taking account only of 1 species, electrons, for now.)

We have conrmed the previous comment that the transverse waves (Ex , Ey ) are unaected.

The longitudinal wave is. Notice that now depends on k as well as . This is called spatial

dispersion.

For completeness, note that the dielectric tensor can be expressed in general tensor notation

as

2
1
= 1 p2 1 + kk
2 0 1

1 k 2 pmn

p2
1

= 1 2 1 + kk w2 mn

1
2
k p

(5.115)

This form shows isotropy with respect to the medium: there is no preferred direction in
space for the wave vector k.
112

But once k is chosen, is not isotropic. The direction of k becomes a special direction.
Longitudinal Waves: dispersion relation is
zz = 0
which is
1

(as before)
p2

(5.116)

=0 .

(5.117)

p0
mn

(5.118)

T
= p2 + k 2 vt2
m

(5.119)

k2 p0
mn

or
2 = p2 + k 2
Recall p0 = n0 T = nT ; so this is usually written:
2 = p2 + k 2

[The appropriate value of to take is 1 dimensional adiabatic i.e. = 3. This seems plausible
since the electron motion is 1d (along k) and may be demonstrated more rigorously by kinetic
theory.]
The above dispersion relation is called the BohmGross formula for electron plasma waves.
Notice the group velocity:
d
1 d 2
vg =
=
=
dk
2 dk

kvt2
p2 + k 2 vt2

1 = 0.

(5.120)

and for kvt > p this tends to 2 vt . In this limit energy travels at the electron thermal
speed.

5.4.1

Refractive Index Plot

Bohm Gross electron plasma waves:

p2
c2
N =
1 2
2
e vte

(5.121)

Transverse electromagnetic waves:

2
N = 1 p2

(5.122)

These have just the same shape except the electron plasma waves have much larger vertical
scale:
On the EM wave scale, the plasma wave curve is nearly vertical. In the cold plasma it was
exactly vertical.
We have relaxed the Cold Plasma approximation.
113

Figure 5.8: Refractive Index Plot. Top plot on the scale of the BohmGross Plasma waves.
Bottom plot, on the scale of the EM transverse waves

5.4.2

Including the ion response

As an example of the dierent things which can occur when ions are allowed to move include
longitudinal ion response:
0 = zz = 1

2
pe

k2 pe e
me ne

2
pi

k2 pi i
mi ni

(5.123)

This is now a quadratic equation for 2 so there are two solutions possible for a given . One
2
will be in the vicinity of the electron plasma wave solution and the inclusion of pi
which is
2
pe will give a small correction.
Second solution will be where the third term is same magnitude as second (both will be
1). This will be at low frequency. So we may write the dispersion relation approximately
as:
2
2
pi
pi
k2 pe e
=0
(5.124)
2
2 kmpi inii
me ne
i.e.
2 =

2
k 2 pi i pi
k 2 pe e
+ 2
mi ni
pe me ne

114

i pi e pe
+
ni
ne

2 i Ti + e Te
= k
mi
= k2

1
mi

(5.125)

[In this case the electrons have time to stream through the wave in 1 oscillation so they
tend to be isothermal: i.e. e = 1. What to take for i is less clear, and less important
because kinetic theory shows that these waves we have just found are strongly damped
unless Ti Te .]
These are ionacoustic or ionsound waves
2
= c2s
k2
cs is the sound speed
c2s =

(5.126)

i Ti + Te
Te

mi
mi

(5.127)

Approximately nondispersive waves with phase velocity cs .

5.5

Electrostatic Approximation for (Plasma) Waves

The dispersion relation is written generally as

N (N E) + .E = N(N.E) N 2 E + .E = 0

(5.128)

Consider E to be expressible as longitudinal and transverse components E , Et such that

N E = 0, N.Et = 0. Then the dispersion relation can be written
N (N.E ) N 2 (E + Et ) + . (E + Et ) = N 2 Et + .Et + .E = 0

(5.129)

or

N 2 .Et = .E

(5.130)

Now the electric eld can always be written as the sum of a curlfree component plus a
divergenceless component, e.g. conventionally

E=

Curlf ree
Electrostatic

(5.131)

Divergencef ree
Electromagnetic

and these may be termed electrostatic and electromagnetic parts of the eld.
For a plane wave, these two parts are clearly the same as the longitudinal and transverse
parts because
= ik is longitudinal
(5.132)
= 0 (because .A = 0 (w.l.o.g.)) then k.A
= 0 so A
is transverse.
and if .A
115

Electrostatic waves are those that are describable by the electrostatic part of the electric
eld, which is the longitudinal part: |E | |Et |.
If we simply say Et = 0 then the dispersion relation becomes .E = 0. This is not the most
general dispersion relation for electrostatic waves. It is too restrictive. In general, there is
a more signicant way in which to get solutions where |E | |Et |. It is for N 2 to be very
large compared to all the components of : N 2 .
If this is the case, then the dispersion relation is approximately
N 2 Et = .E ;

(5.133)

Et is small but not zero.

We can then annihilate the Et term by taking the N component of this equation; leaving
N..E = (N..N) E = 0

k..k = 0 .

(5.134)

When the medium is isotropic there is no relevant dierence between the electrostatic dis
persion relation:
N..N = 0
(5.135)
and the purely longitudinal case .N = 0. If we choose axes such that N is along z, then the
mediums isotropy ensures the odiagonal components of are zero so N..N = 0 requires
zz = 0 .N = 0. However if the medium is not isotropic, then even if

N..N = N 2 zz = 0

(5.136)

there may be odiagonal terms of that make

.N = 0

(5.137)

In other words, in an anisotropic medium (for example a magnetized plasma) the electrostatic
approximation can give waves that have nonzero transverse electric eld (of order ||||/N 2
times E ) even though the waves are describable in terms of a scalar potential.
To approach this more directly, from Maxwells equations, applied to a dielectric medium
of dielectric tensor , the electrostatic part of the electric eld is derived from the electric
displacement
.D = . (0 .E) = = 0 (no free charges)
(5.138)
So for plane waves 0 = k.D = k..E = ik..k.
The electric displacement, D, is purely transverse (not zero) but the electric eld, E then
gives rise to an electromagnetic eld via H = D/t. If N 2 then this magnetic
(inductive) component can be considered as a benign passive coupling to the electrostatic
wave.
In summary, the electrostatic dispersion relation is k..k = 0, or in coordinates where k is
in the zdirection, zz = 0.

116

5.6

Simple Example of MHD Dynamics: Alfven Waves

Ignore Pressure & Resistance.

DV
=jB
Dt
E+VB=0

(5.139)
(5.140)

Linearize:
V = V1 ,

B = B0 + B1 (B0 uniform),

j = j1 .

V
= j B0
t
E + V B0 = 0

(5.141)
(5.142)

(5.143)

Fourier Transform:
(i)V = j B0

(5.144)

E + V B0 = 0

(5.145)

Eliminate V by taking 5.144 B0 and substituting from 5.145.

E+
or
E=

1
(j B0 ) B0 = 0
i

(5.146)

1
B02
{(j.B0 ) B0 B02 j} =
j
i
i

(5.147)

So conductivity tensor can be written (z in B direction).

1 0 0
i

=
2 0 1 0
B0
0 0

(5.148)

where implies that E = 0 (because of Ohms law). Hence Dielectric Tensor

= 1 +

i0
0 B 2

=1+

1 0 0

0 1 0 .
0 0

(5.149)

Dispersion tensor in general is:

D=

2
2
NN

N
+

c2

(5.150)

Dispersion Relation taking N = Nx , Ny = 0

N 2 + 1 +

0
|
D
| =

N N

0 B 2

0
2
2
N N + 1 +
0
117

0 B 2

N N

0
= 0

(5.151)

Figure 5.9: Compressional Alfven Wave. Works by magnetic pressure (primarily).

Meaning of is that the cofactor must be zero i.e.

N2

+1+
0 B 2

N + 1 +
0 B 2
2

=0

(5.152)

The 1s here come from Maxwell displacement current and are usually negligible (N2 1).
So nal waves are
1. N 2 =

0 B 2

Nondispersive wave with phase and group velocities

c
vp = vg =
=
N

where we call

B2
0

c2 0 B 2

1
2

B2
=
0

(5.153)

1
2

vA

the Alfven Speed

(5.154)

Vy = 0 Vx =
0 (Vz = 0)

(5.155)

Polarization:
E = Ez = 0,

Ex = 0. Ey =
0

Party longitudinal (velocity) wave Compression Compressional Alfven Wave.

2.

k 2 c2

N2 = 0B 2 = 2
Any has unique k . Wave has unique velocity in direction: vA .

Polarization

Ez = Ey = 0 Ex =
0

Vx = 0 Vy =
0 (Vz = 0)

Transverse velocity: Shear Alfven Wave.

Works by eld line bending (Tension Force) (no compression).
118

(5.156)

Figure 5.10: Shear Alfven Wave

5.7

NonUniform Plasmas and wave propagation

Practical plasmas are not innite & homogeneous. So how does all this plane wave analysis
apply practically?

If the spatial variation of the plasma is slow c.f. the wave length of the wave, then coupling

to other waves will be small (negligible).

Figure 5.11: Comparison of sudden and gradualy refractive index change.

For a given , slowly varying plasma means N/ dN
or kN/ dN
1. Locally, the plasma
dx
dx
appears uniform.
Even if the coupling is small, so that locally the wave propagates as if in an innite uniform
plasma, we still need a way of calculating how the solution propagates from one place to
the other. This is handled by the WKB(J) or eikonal or ray optic or geometric optics
approximation.
WKBJ solution
Consider the model 1d wave equation (for eld )
d2 E
+ k2E = 0
dx2
with k now a slowly varying function of x. Seek a solution in the form
E = exp (i (x))
119

(it implied)

(5.157)

(5.158)

 is the wave phase (= kx in uniform plasma).

Dierentiate twice
d2 E
d2
d
=
{i

dx2
dx2
dx

}ei

(5.159)

Substitute into dierential equation to obtain

Recognize that in uniform plasma

ignore this term.

d
dx

d2
dx2

= k2 + i

d2
dx2

(5.160)

= 0. So in slightly nonuniform, 1st approx is to

d
k(x)
dx
Then obtain a second approximation by substituting

(5.161)

d2
dk

2
dx
dx

(5.162)

so

d
dk
k2 i
dx
dx

d
i dk
k
dx
2k dx

Integrate:

Hence E is
E = ei =

(5.163)
using Taylor expansion.

kdx + i ln k 2
1

exp
i

(5.164)

(5.165)

kdx

(5.166)
1
k2
This is classic WKBJ solution. Originally studied by Green & Liouville (1837), the Green
of Greens functions, the Liouville of Sturm Liouville theory.

Basic idea of this approach: (1) solve the local dispersion relation as if in innite homogeneous

plasma, to get k(x), (2) form approximate solution for all space as above.

Phase of wave varies as integral of kdx.

In addition, amplitude varies as

5.8

1
1

k2

. This is required to make the total energy ow uniform.

Two Stream Instability

An example of waves becoming unstable in a nonequilibrium plasma. Analysis is possible

using Cold Plasma techniques.

Consider a plasma with two participating cold species but having dierent average velocities.

120

These are two streams.

Species1 Species2
.
.
M oving. Stationary.
Speed v

(5.167)

We can look at them in dierent inertial frames, e.g. species (stream) 2 stationary or 1
stationary (or neither).
We analyse by obtaining the susceptibility for each species and adding together to get total
dielectric constant (scalar 1d if unmagnetized).
In a frame of reference in which it is stationary, a stream j has the (Cold Plasma) suscepti
bility
2
pj
j =
.
(5.168)
2
If the stream is moving with velocity vj (zero order) then its susceptibility is
j =

2
pj

( kvj )2

. (k & vj in same direction)

(5.169)

Proof from equation of motion:

qj
v
= i ( kvj ) v
.
E=
+ v.
v = (i + ik.vj ) v
mj
t

(5.170)

j = j vj + j .
v + v
j.

(5.171)

Current density
Substitute in
.j +

= ik.
vj + ik.
v i = 0
t
k.
v
j = j
k.vj

(5.172)
(5.173)

in terms of E:
Hence substituting for v
j 0 .E = j =

j q j
k.E
,
mj i ( k.vj )2

(5.174)

which shows the longitudinal susceptibility is

j =

2
pj
j q j
1
=
mj 0 ( 2 kvj )2
( kvj )2

(5.175)

Proof by transforming frame of reference:

Consider Galileean transformation to a frame moving with the stream at velocity vj .
x = x + vj t ;
121

t = t

(5.176)

exp i (k.x t) = exp i (k.x ( k.vj ) t )

(5.177)

So in frame of the stream, = k.vj .

Substitute in stationary cold plasma expression:
2
2
pj
pj
j = 2 =
.

( kvj )2

(5.178)

Thus for n streams we have

=1+

j = 1

2
pj

( kvj )2

(5.179)

Longitudinal wave dispersion relation is

= 0.
Two streams
0==1

2
p1

( kv1 )2

(5.180)

P2 2
( kv2 )2

(5.181)

For given real k this is a quartic in . It has the form:

Figure 5.12: Twostream stability analysis.

If crosses zero between the wells, then 4 real solutions for . (Case B).

If not, then 2 of the solutions are complex: = r ii (Case A).

The time dependence of these complex roots is

exp (it) = exp (ir t i t) .

122

(5.182)

The +ve sign is growing in time: instability.

It is straightforward to show that Case A occurs if

2
3

2
3

|
k (v2 v1 )| < p1 + p2

3
2

(5.183)

Small enough k (long enough wavelength) is always unstable.

2
2
Simple interpretation (p2
p1
, v1 = 0) a tenuous beam in a plasma sees a negative if
<
|kv2 | p1 .
Negative implies charge perturbation causes E that enhances itself: charge (spontaneous)
bunching.

5.9

Kinetic Theory of Plasma Waves

Wave damping is due to waveparticle resonance. To treat this we need to keep track of the
particle distribution in velocity space kinetic theory.

5.9.1

Vlasov Equation

Treat particles as moving in 6D phase space x position, v velocity. At any instant a particle
occupies a unique position in phase space (x, v).
Consider an elemental volume d3 xd3 v of phase space [dxdydzdvx dvy dvz ], at (x, v). Write
down an equation that is conservation of particles for this volume

3 3
f d xd v =
t
+
+
+

[vx f (x + dx
x, v) vx f (x, v)] dydzd3 v
same for dy, dz
) ax f (x, v)] d3 xdvy dvz
[ax f (x, v + dvx x
same for dvy , dvz

Figure 5.13: Dierence in ow across xsurfaces (+y + z).

a is velocity space motion, i.e. acceleration.
123

(5.184)

Divide through by d3 xd3 v and take limit

f
t

(vx f ) +
(vy f ) +
(vz f ) +
(ax f ) +
(ay f ) +
(az f )
y
z
vx
vy
vz
x
= . (vf ) + v . (af )
(5.185)

[Notation: Use

; v
v ].

Take this simple continuity equation in phase space and expand:

f
+ (.v) f + (v.) f + (v .a) f + (a.v ) f = 0.
t
Recognize that means here

(5.186)

etc. keeping v constant so that .v = 0 by denition. So

f
f
f
+ v.
+ a.
= f (v .a)
t
x
v

(5.187)

Now we want to couple this equation with Maxwells equations for the elds, and the Lorentz
force
q
a=
(E + v B)
(5.188)
m
Actually we dont want to use the E retaining all the local eects of individual particles. We
want a smoothed out eld. Ensemble averaged E.
Evaluate
q
q
v .a = v . (E + v B) = v . (v B)
m
m
q
=
B. (v v) = 0.
m

(5.189)
(5.190)

So RHS is zero. However in the use of smoothed out E we have ignored local eect of one
particle on another due to the graininess. That is collisions.
Boltzmann Equation:
f
f
f
+ v.
+ a.
=
t
x
v

f
t

(5.191)
collisions

Vlasov Equation Boltzman Eq without collisions. For electromagnetic forces:

f
q
f
f
+ v.
+ (E + v B)
= 0.
t
x m
v

(5.192)

Interpretation:
Distribution function is constant along particle orbit in phase space:

d
f
dt

= 0.

d
f
dx f
v f
f=
+
.
+
.
dt
t
dt x
dt v
Coupled to Vlasov equation for each particle species we have Maxwells equations.
124

(5.193)

VlasovMaxwell Equations
fj
t

fj
qj
fj
+
(E + v B) .
=0
x
mj
vj
1 E
B
E =
, B = 0 j + 2
c t
t

, .B = 0
.E =
0
+ v.

(5.194)
(5.195)
(5.196)

Coupling is completed via charge & current densities.

=

qj n j =

j =

qj

fj d3 v

(5.197)

qj nj Vj =

qj

fj vd3 v.

(5.198)

Describe phenomena in which collisions are not important, keeping track of the (statistically

averaged) particle distribution function.

Plasma waves are the most important phenomena covered by the VlasovMaxwell equations.

6dimensional, nonlinear, timedependent, integraldierential equations!

5.9.2

Linearized Wave Solution of Vlasov Equation

Unmagnetized Plasma
Linearize the Vlasov Eq by supposing
f = f0 (v) + f1 (v) exp i (k.x t) , f1 small.
also E = E1 exp i (k.x t) B = B1 exp i (k.x t)
Zeroth order f0 equation satised by

,
t x

if1 + v.ikf1 +

(5.199)
(5.200)

= 0. First order:

q
f0
(E1 + v B1 ) .
= 0.
m
v

(5.201)

[Note v is not per se of any order, it is an independent variable.]

Solution:
f1 =

1
q
f0
(E1 + v B1 ) .
i ( k.v) m
v

For convenience, assume f0 is isotropic. Then

f1 =

f0
v

(5.202)

0
is in direction v so v B1 . f
=0
v

q
E . f0
m 1 v

i ( k.v)

(5.203)

We want to calculate the conductivity . Do this by simply integrating:

j=

0
q 2 v f
v
qf1 vd v =
d3 v .E1 .
im k.v

125

(5.204)

Here the electric eld has been taken outside the vintegral but its dot product is with
f0 /v. Hence we have the tensor conductivity,
0
q 2 v f
v
=
d3 v
im k.v

(5.205)

Focus on zz component:
1 + zz = zz = 1 +

f
zz
q 2 vz vz0 3
=1+
dv
i0
m0 k.v

(5.206)

Such an expression applies for the conductivity (susceptibility) of each species, if more than
one needs to be considered.
It looks as if we are there! Just do the integral!
Now the problem becomes evident. The integrand has a zero in the denominator. At least
we can do 2 of 3 integrals by dening the 1dimensional distribution function
fz (vz )
Then

f (v)dvx dvy

(k = kz)

f
q 2 vz vzz
=
dvz
m0 kvz

(5.207)

(5.208)

(drop the z sux from now on. 1d problem).

How do we integrate through the pole at v = k ? Contribution of resonant particles. Crucial
to get right.
Path of velocity integration
First, realize that the solution we have found is not complete. In fact a more general solution
can be constructed by adding any solution of
f1
f1
+v
=0
t
z
[We are dealing with 1d Vlasov equation:

f
t

+ v f
+
z

(5.209)
qE f
m v

= 0.] Solution of this is

f1 = g(vt z, v)

(5.210)

where g is an arbitrary function of its arguments. Hence general solution is

f1 =

q
0

E f
m
v

i ( kv)

exp i (kz t) + g (vt z, v)

(5.211)

and g must be determined by initial conditions. In general, if we start up the wave suddenly
there will be a transient that makes g nonzero.
126

So instead we consider a case of complex (real k for simplicity) where = r + ii and

i > 0.

This case corresponds to a growing wave:

exp(it) = exp(ir t + i t)

(5.212)

Then we can take our initial condition to be f1 = 0 at t . This is satised by taking

g = 0.

For i > 0 the complementary function, g, is zero.

Physically this can be thought of as treating a case where there is a very gradual, smooth
start up, so that no transients are generated.

Thus if i > 0, the solution is simply the velocity integral, taken along the real axis, with

no additional terms. For

f
q2
v v
=
dv
mo C kv

i > 0,

(5.213)

where there is now no diculty about the integration because is complex.

Figure 5.14: Contour of integration in complex vplane.

The pole of the integrand is at v =

which is above the real axis.

The question then arises as to how to do the calculation if i 0. The answer is by analytic

continuation, regarding all quantities as complex.

Analytic Continuation of is accomplished by allowing /k to move (e.g. changing the i )

but never allowing any poles to cross the integration contour, as things change continuously.

Remember (Fig 5.15)

F dz =

residues 2i

(5.214)

(Cauchys theorem)

Where residues = limzzk [F (z)/(z zk )] at the poles, zk , of F (z). We can deform the contour

how we like, provided no poles cross it. Hence contour (Fig 5.16)

127

Figure 5.15: Cauchys theorem.

Figure 5.16: Landau Contour

We conclude that the integration contour for i < 0 is not just along the real v axis. It
includes the pole also.

To express our answer in a universal way we use the notation of Principal Value of a

singular integral dened as the average of paths above and below

F
1
dv =
v v0
2

+
C1

C2

F
dv
v v0

(5.215)

Figure 5.17: Two halves of principal value contour.

Then

0
v f
1
f0
12
v
dv 2i 2
=
{
}

kv
2
k v
v=
mo

(5.216)

Second term is half the normal residue term; so it is half of the integral round the pole.
128

Figure 5.18: Contour equivalence.

Our expression is only shorthand for the (Landau) prescription:
Integrate below the pole.
(Nautilus).
Contribution from the pole can be considered to arise from the complementary function
g(vt z, v). If g is to be proportional to exp(ikz), then it must be of the form g = exp[ik(z
vt)]h(v) where h(v) is an arbitrary function. To get the result previously calculated, the value
of h(v) must be (for real )

q 1 f0

v
h(v) =
m k v
w
k

(5.217)

(so that

q
f0 q
2

vgdv = i 2
io
k v mo

.)

(5.218)

This Dirac delta function says that the complementary function is limited to particles with
exactly the wave phase speed k . It is the resonant behaviour of these particles and the
imaginary term they contribute to that is responsible for wave damping.

We shall see in a moment, that the standard case will be i < 0, so the opposite of the

prescription i > 0 that makes g = 0. Therefore there will generally be a complementary

function, nonzero, describing resonant eects. We dont have to calculate it explicitly

because the Landau prescription takes care of it.

5.9.3

Landaus original approach.

(1946)

Corrected Vlasovs assumption that the correct result was just the principal value of the inte

gral. Landau recognized the importance of initial conditions and so used Laplace Transform

approach to the problem

A(p) =
ept A(t)dt
(5.219)
0

The Laplace Transform inversion formula is

1 s+i pt
e A(p)dp
A(t) =
2i si

(5.220)

where the path of integration must be chosen to the right of any poles of A(p) (i.e. s large
enough). Such a prescription seems reasonable. If we make (p) large enough then the A(p)
integral will presumably exist. The inversion formula can also be proved rigorously so that
gives condence that this is the right approach.
129


If we identify p i, then the transform is A = e
it A(t)dt, which can be identied as
the Fourier transform that would give component A
eit , the wave we are discussing.
Making (p) positive enough to be to the right of all poles is then equivalent to making ()
positive enough so that the path in space is above all poles, in particular i > (kv). For
real velocity, v, this is precisely the condition i > 0, we adopted before to justify putting
the complementary function zero.

Either approach gives the same prescription. It is all bound up with satisfying causality.

5.9.4

Solution of Dispersion Relation

We have the dielectric tensor

0
f0
q 2 v f
v

dv i
2
,
=1+=1+
kv
k v

m0

(5.221)

for a general isotropic distribution. We also know that the dispersion relation is

N 2 + t
0
0

2
2
0
N
+

= N + t = 0
t
0
0

(5.222)

Giving transverse waves N 2 = t and

longitudinal waves = 0.

Need to do the integral and hence get .

Presumably, if we have done this right, we ought to be able to get back the coldplasma

result as an approximation in the appropriate limits, plus some corrections. We previously

argued that coldplasma is valid if k vt . So regard kv

as a small quantity and expand:

2
1 f0
kv
kv

dv =
v
1+
+
+ ... dv

1 kv

0
v f
dv

1
2kv
kv
=
+3
fo 1 +

+ ...
dv

1
3nT k 2

n+
+ ...

m 2

(by parts)

(5.223)

Here we have assumed we are in the particles average rest frame (no bulk velocity) so that
f0 vdv = 0 and also we have used the temperature denition

nT =

mv 2 f0 dv

(5.224)

appropriate to one degree of freedom (1d problem). Ignoring the higher order terms we get:

2
T k2
2 1 f0

= 1 p2 1 + 3
+
i

m 2
k 2 n v
k

130

(5.225)

This is just what we expected. Cold plasma value was = 1

p2
.
2

We have two corrections

T k
1. To real part of , correction 3 m
= 3 vvpt due to nite temperature. We could have
2
got this from a uid treatment with pressure.

2. Imaginary part antihermitian part of dissipation.

Solve the dispersion relation for longitudinal waves = 0 (again assuming k real complex).
Assume i r then
T k2
2 1 f0
| }
+
i
m 2
k 2 n v k
T k2
r2 1 fo
+
i
|
r }
p2 {1 + 3
m r2
k 2 n v
k
1 2 r2 1 f0
r 1 f0

p i 2
| kr = p2
| r
2r i
k n v
2 k 2 n v k

(r + ii )2 r2 + 2r i i = p2 {1 + 3

Hence 1

(5.226)
(5.227)

For a Maxwellian distribution

m
f0 =
2T

f0
m
=
v
2T

1
2

r2
i p2
2 k3

mv 2
exp
n
2T

(5.228)

mv
mv 2

exp
n
T
2T

m
2T

1
2

mr2
m

exp
2T k 2
T

(5.229)

(5.230)

The dierence between r and p may not be important in the outside but ought to be
retained inside the exponential since
mp2
T k2
m p2
3
=
1+3
+
2
2
2
2
2T k
m p
2T k

So

i p

p3 1
mp2
3
exp

3
3
2
k vt
2T k
2

(5.231)

(5.232)

Imaginary part of is negative damping. This is Landau Damping.

Note that we have been treating a single species (electrons by implication) but if we need
more than one we simply add to . Solution is then more complex.

5.9.5

Direct Calculation of Collisionless Particle Heating

(Landau Damping without complex variables!)

We show by a direct calculation that net energy is transferred to electrons.

131

Suppose there exists a longitudinal wave

E = E cos(kz t)z

(5.233)

Equations of motion of a particle

dv
q
=
E cos(kz t)
dt
m
dz
= v
dt

(5.234)
(5.235)

Solve these assuming E is small by a perturbation expansion v = v0 + v1 + ..., z = z0 (t) +

z1 (t) + ... .

Zeroth order:

dvo
= 0 v0 = const ,
dt
where zi = const is the initial position.

z0 = zi + v0 t

(5.236)

First Order
dv1
q
q
=
E cos (kz0 t) = E cos (k (zi + v0 t) t)
dt
m
m

dz1

= v1
dt

(5.237)
(5.238)

Integrate:
qE sin (kzi + kv0 t)
+ const.
m
kv0
take initial conditions to be v1 , v2 = 0. Then
v1 =

v1 =

(5.239)

qE sin (kzi + t) sin (kzi )

m

(5.240)

where kv0 , is () the frequency at which the particle feels the wave eld.

sin kzi
qE cos kzi cos (kzi + t)
z1 =
t
2
m

(5.241)

(using z1 (0) = 0).

2nd Order

(Needed to get energy right)

dv2
qE
{cos (kzi + kv0 t t + kz1 ) cos (kzi + kv0 t t)}
=
dt

M
qE

=
kzi { sin (kzi + t)}
(kz1 1)
m

(5.242)

Now the gain in kinetic energy of the particle is

1 2 1
1 2
mv
mv = m{(v0 + v1 + v2 + ...)2 v02 }
2
2 0 2

1
= {2v0 v1 + v12
+ 2v0 v2 + higher order}
2
132

(5.243)

and the rate of increase of K.E. is

d 1 2
dv1
dv1
dv2
mv = m v0
+ v1
+ v0
dt 2
dt
dt
dt

(5.244)

We need to average this over space, i.e. over zi . This will cancel any component that simply
oscillates with zi .

d 1 2
dv1
dv2
dv1
mv
= v0
+ v1
+ v0
m
(5.245)
dt 2
dt
dt
dt

dv1
= 0
(5.246)
v0
dt

dv1
q 2 E 2 sin (kzi + t) sin kzi
v1
=
cos (kzi + t)
dt
m2

q 2 E 2 sin (kzi + t) sin (kzi + t) cos t + cos (kzi + t) sin t

=
m2

cos (kzi + t)
q 2 E 2 sin t
cos2 (kzi + t)
m2

q 2 E 2 1 sin t

(5.247)
m2 2

q 2 E 2
cos kzi cos (kzi + t)
sin kzi
kv0
sin (kzi + t)
t
2
2
m

q 2 E 2
sin t
cos t
2
sin (kzi + t)
kv0
t
m2
2

q 2 E 2 kv0
sin t
cos t

+t
(5.248)
m2 2
2

=
=

v0

dv2
dt

=
=
=

Hence

d1 2
mv
dt 2

q 2 E 2 sin t
sin t
cos t
=
kv0
+ kv0 t
2
2m

2 2
q E sin t
t
=
+
cos
t
+
t
cos
t
2m
2

(5.249)
(5.250)

This is the spaceaveraged power into particles of a specic velocity v0 . We need to integrate
over the distribution function. A trick identify helps:

sin t +
cos t + t cos t =
2

1 sin t
=
+ sin t
k v0

133

sin t
+ sin t

(5.251)
(5.252)

Hence power per unit volume is

d1 2
P =
mv f (v0 ) dv0
dt 2

q2E 2
sin t
=
f (v0 )
+ sin t dv0
2mk
v0

q 2 E 2 sin t
f
=
+ sin t
dv0
v0
2mk

(5.253)

As t becomes large, sin t = sin(kv0 )t becomes a rapidly oscillating function of v0 .

Hence second term of integrand contributes negligibly and the rst term,

sin t
sin t
=
t

(5.254)

becomes a highly localized, deltafunctionlike quantity. That enables the rest of the inte
grand to be evaluated just where = 0 (i.e. kv0 = 0).

Figure 5.19: Localized integrand function.

So:
P =

q 2 E 2 f sin x
|
dx
2mk k v k
x

(5.255)

x = t
= (kv0 )t.
sin x
and
dz = so
x

q 2 f0
|

(5.256)
2mk 2 v k
We have shown that there is a net transfer of energy to particles at the resonant velocity k
from the wave. (Positive if f
| is negative.)
v
P = E

5.9.6

Physical Picture

is the frequency in the particles (unperturbed) frame of reference, or equivalently it is

kv0 where v0 is particle speed in wave frame of reference. The latter is easier to deal with.
t = kv0 t is the phase the particle travels in time t. We found that the energy gain was of
the form

sin t
d (t) .
(5.257)
t
134

Figure 5.20: Phase distance traveled in time t.

This integrand becomes small (and oscillatory) for t 1. Physically, this means that
if particle moves through many wavelengths its energy gain is small. Dominant contribution
is from t < . These are particles that move through less than 12 wavelength during the
period under consideration. These are the resonant particles.

Figure 5.21: Dominant contribution

Particles moving slightly faster than wave are slowed down. This is a secondorder eect.

Figure 5.22: Particles moving slightly faster than the wave.

Some particles of this v0 group are being accelerated (A) some slowed (B). Because As are

then going faster, they spend less time in the down region. Bs are slowed; they spend more

time in up region. Net eect: tendency for particle to move its speed toward that of wave.

Particles moving slightly slower than wave are speeded up. (Same argument). But this is

only true for particles that have caught the wave.

Summary: Resonant particles velocity is drawn toward the wave phase velocity.

Is there net energy when we average both slower and faster particles? Depends which type

has most.

Our Complex variables wave treatment and our direct particle energy calculation give con

sistent answers. To show this we need to show energy conservation. Energy density of

135

Figure 5.23: Damping or growth depends on distribution slope

wave:
W =

1
1
[ 0 |E 2 | +
2 2

<sin2 > Electrostatic

1
n m|v2 | ]
2

(5.258)

P article Kinetic

Magnetic wave energy zero (negligible) for a longitudinal wave. We showed in Cold Plasma
qE
treatment that the velocity due to the wave is v = im
Hence
p2
1 0 E 2
W
1+ 2
2 2

(again electrons only)

(5.259)

When the wave is damped, it has imaginary part of , i and

dW
1 dE 2
=W 2
= 2i W
(5.260)
dt
E dt
Conservation of energy requires that this equal minus the particle energy gain rate, P . Hence
2

q f
P
+E 2 2mk2 v0 | k
2
2 1 f0

i =
=
| k

2 = p
2

2
2
0 E
2W
2
k n v

1 + p2
1 + p2
1

(5.261)

So for waves such that p , which is the dispersion relation to lowest order, we get
i =

p2

r 1 f0

2 k 2 n v r

(5.262)

This exactly agrees with the damping calculated from the complex dispersion relation using

the Vlasov equation.

This is the Landau damping calculation for longitudinal waves in a (magnetic) eldfree

plasma. Strictly, just for electron plasma waves.

How does this apply to the general magnetized plasma case with multiple species?

Doing a complete evaluation of the dielectric tensor using kinetic theory is feasible but very

heavy algebra. Our direct intuitive calculation gives the correct answer more directly.

136

5.9.7

Damping Mechanisms

Cold plasma dielectric tensor is Hermitian. [Complex conjugate*, transposeT = original

matrix.] This means no damping (dissipation).
The proof of this fact is simple but instructive. Rate of doing work on plasma per unit
volume is P = E.j. However we need to observe notation.
Notation is that E(k, ) is amplitude of wave which is really (E(k, ) exp i(k.x t)) and
similarly for j. Whenever products are taken: must take real part rst. So
P = (E exp i (k.x t)) . (j exp i (k.x t))
1

1 i
( = k.x t.)
=
Ee + E ei . jei + j ei
2
2

1
=
E.je2i + E.j + E .j + E .j e2i
4

(5.263)

The terms e2i & e2i are rapidly varying. We usually average over at least a period. These
average to zero. Hence
1
1
P = [E.j + E .j] = (E.j )
(5.264)
4
2
Now recognize that j = .E and substitute
P =

1
[E. .E + E ..E]
4

(5.265)

But for arbitrary matrices and vectors:

A.M.B = B.MT .A;

(5.266)

(in our dyadic notation we dont explicitly indicate transposes of vectors). So

hence

E. .E = E . T .E

(5.267)

1
P = E . T + .E
4

(5.268)

1
If = 1 + i
is hermitian T = , then the conductivity tensor is antihermitian
0
T = (if is real). In that case, equation 5.268 shows that < P >= 0. No dissipation.
Any dissipation of wave energy is associated with an antihermitian part of and hence .
Cold Plasma has none.

Collisions introduce damping. Can be included in equation of motion

m

dv
= q (E + v B) mv
dt

where is the collision frequency.

137

(5.269)

Whole calculation can be followed through replacing m(i) with m( i) everywhere.

This introduces complex quantity in S, D, P .

We shall not bother with this because in fusion plasmas collisional damping is usually neg

ligible. See this physically by saying that transit time of a wave is

Size
1 meter

3 109 seconds.
Speed
3 10+8 m/s

(5.270)

(Collision frequency)1 10s 1ms, depending on Te , ne .

When is the conductivity tensor Antihermitian?
Cold Plasma:

2
pj
j 2 2
j pj j
= j 2 2
j
2
pj
= 1 j 2

S =1

S iD 0

S
0
iD
=

0
0
P

where

D
P

(5.271)

This is manifestly Hermitian if is real, and then is antiHermitian.

This observation is sucient to show that if the plasma is driven with a steady wave, there

is no damping, and k does not acquire a complex part.

Two stream Instability

zz = 1

2
pj

( kvj )2

(5.272)

In this case, the relevant component is Hermitian (i.e. real) if both and k are real.

But that just begs the question: If and k are real, then theres no damping by denition.

So we cant necessarily detect damping or growth just by inspecting the dieletric tensor form

when it depends on both and k.

Electrostatic Waves in general have = 0 which is Hermitian. So really it is not enough to

deal with or . We need to deal with = io , which indeed has a Hermitian component

for the twostream instability (even though is Hermitian) because is complex.

5.9.8

Ion Acoustic Waves and Landau Damping

We previously derived ion acoustic waves based on uid treatment giving

zz = 1
Leading to 2 k 2

i Ti +e Te
mi

2
pe

k2 pe e
me ne

138

2
pi

k2 pi i
mi ni

(5.273)

Kinetic treatment adds the extra ingredient of Landau Damping. Vlasov plasma, unmagne

tized:
2
2
pe
1 foe dv pi
1 foi dv
zz = 1 2
2
(5.274)

k C v k v
n
k C v k v n
Both electron and ion damping need to be considered as possibly important.

Based on our uid treatment we know these waves will have small phase velocity relative to

electron thermal speed. Also cs is somewhat larger than the ion thermal speed.

Figure 5.24: Distribution functions of ions and electrons near the sound wave speed.
So we adopt approximations
vte

,
k

vti < (<)

(5.275)

and expand in opposite ways.

Ions are in the standard limit, so
2
3Ti k 2
2 1 foi
i pi2 1 +
+
i
|w/k
m 2
k 2 ni v

Electrons: we regard

(5.276)

as small and write

1
v

foe dv
1 foe dv

v n

v v
n

2 foe
=
dv
n v 2
2
me
=

foe dv
n
2Te
me
=
Te

for Maxwellian.
(5.277)

Write F0 = fo /n.
Contribution from the pole is as usual so

2
m
Foe
e + i
e =

pe

k2
Te
v /k

139

(5.278)

Collecting real and imaginary parts (at real )

2
2
pe
me pi
3Ti k 2
r (r ) = 1 + 2
2 1+
k Te
m r2
r

1
2 Foe
2 Foi
|/k
i (r ) = 2 pe
|/k + pi
k
v
v

(5.279)
(5.280)

The real part is essentially the same as before. The extra Bohm Gross term in ions appeared
previously in the denominator as
2
pi

2
3Ti k 2
pi2 1 +

mi 2

k2 pi i
mi

(5.281)

Since our kinetic form is based on a rather inaccurate Taylor expansion, it is not clear that
it is a better approx. We are probably better o using
2
pi
1
2 .
ik
2 1 3T
2
m

(5.282)

r2
Te + 3Ti
1
=
k2
mi
1 + k 2 2De

(5.283)

Then the solution of r (r ) = 0 is

as before, but weve proved that e = 1 is the correct choice, and kept the k 2 2De term (1st
term of r ).
The imaginary part of gives damping.
General way to solve for damping when small
We want to solve (k, ) = 0 with = r + ii , i small.
Taylor expand about real r :
d
|
d r

= (r ) + ii
(r )
r

() (r ) + ii

(5.284)
(5.285)

Let r be the solution of r (r ) = 0; then

() = ii (r ) + ii

(r ).
r

(5.286)

This is equal to zero when

i =

i (r )
(r )
r

140

(5.287)

If, by presumption, i r , or more precisely (in the vicinity of = 0), i /r r /r

then this can be written to lowest order:
i (r )
i = r (r )

(5.288)

Apply to ion acoustic waves:

2
r (r )
4Ti k 2
= pi3 2 + 4
r
r
mi r2

so

(5.289)

Foe
3
1
2 Foi
2
i = 2 2r
|/k + pi
|/k
2
pe
4T
k
i
k pi 2 + 4 m 2
v
v
i

(5.290)

For Maxwellian distributions, using our previous value for r ,

Foe
| r
v
k

me

2Te

1
=
2

me
Te

1
=
2

me
mi

me v m2Te v2
e
e
Te
Te + 3Ti
mi

v= kr
1
2

3T
me 1 + Tei

exp

2mi 1 + k 2 2D
1 + k 2 2D

3T 2
me 1 + Tei

,
Te 1 + k 2 2

(5.291)

De

where the exponent is of order me /mi here, and so the exponential is 1. And
1

3T 2
1
3T
Foi
1 mi 1 + Tei
Te 2
Te 1 + Tei

| kr =
exp
v
2Ti 1 + k 2 2D
2 Ti 1 + k 2 2 Ti

(5.292)

Hence
i

=
r
2

m
i
me
i
=
r

r2
k2

3Ti
Te

2
i k
2 + 4 3T
2
1 + k 2 2D
mi
r

me
mi

1
2

me mi
+

Te
Ti

1+

Te
Ti

3T

21

3T
Te 1 + Tei

exp
2Ti 1 + k 2 2D

(5.293)

1 + Tei

i
2 [1 + k 2 2 ] 32 2 + 4
Te3T
+3Ti
De

me
mi

electron

Te
+
Ti

23

3T
Te 1 + Tei
exp

.
2Ti 1 + k 2 2De

ion damping

141

(5.294)

[Note: the coecient on the rst line of equation 5.294 for i /r reduces to /8 for
Ti /Te 1 and kD e 1.]
Electron Landau damping of ion acoustic waves is rather small:

i
r

me
mi

1
.
70

Ion Landau damping is large, 1 unless the term in the exponent is large. That is
unless
Physics is that large
thermal velocity vti =

Te
T
i

Te
1 .
Ti

(5.295)

pulls the phase velocity of the wave:

Ti
.
mi

Te +3Ti
mi

= cs above the ion

If cs vti there are few resonant ions to damp the wave.

[Note. Many texts drop terms of order TTei early in the treatment, but that is not really
accurate. We have kept the rst order, giving extra coecient
3Ti
1+
Te

23

Te + 3Ti
3 Ti
1+
Te + 6Ti
2 Te

(5.296)

i
and an extra factor 1 + 3T
in the exponent. When Ti Te we ought really to use full
Te
solutions based on the Plasma Dispersion Function.]

5.9.9

Alternative expressions of Dielectric Tensor Elements

This subsection gives some useful algebraic relationships that enable one to transform to
dierent expressions sometimes encountered.
o
q2
fo

q 2 v f
v
dv = 2
dv
1
mo k C kv
v
mo C kv
q2 1
1 fo
dv

2
mo k C k v v
p2
1 1 fo
dv

k 2 C k v n v

p2
Fo
1 Fo

dv i
|
k2
v 2v
v k
k

zz =
=
=
=
where Fo =
of the form

fo
n

(5.297)
(5.298)
(5.299)
(5.300)

is the normalized distribution function. Other elements of involve integrals

o
vj f
mo
vl
d3 v .
=
q2
k.v
When k is in zdirection, k.v = kz vz . (Multi dimensional distribution f0 ).

jl

If (e.g., xy ) l = z and j = l then the integral over vl yields

vj

fo
dvj = fo dvj ,
vj

142

fo
dvl
vl

(5.301)

= 0. If j = l = z then
(5.302)

by parts. So, recalling the denition fz f dvx dvy ,

xx = yy

q2
foz
=
dvz
mo k.v
2
Foz
= p
dvz .
k.v

(5.303)

The fourth type of element is

xz

f
q 2 vx vzo 3
=
dv.
mo kz vz

(5.304)

This is not zero unless fo is isotropic (= fo (v)).

If f is isotropic
dfo v
vz dfo
fo
=
=
vz
dv vz
v dv

(5.305)

Then
fo

vx v
vx vz 1 dfo 3
3
z
dv =
dv
kz v z
kz vz v dv

vz
fo 3
=
d v=0
kz vz vx

(5.306)

(since the vx integral of fo /vx is zero). Hence for isotropic Fo = f0 /n, with k in the
zdirection,

Foz
C kvz dvz

0
2

Foz
C kvz dvz

0+

0
p2
k

Foz
1
C kvz vz dvz

(5.307)

(and the terms 0+ are the ones that need isotropy to make them zero).

t 0 0

=
0 t 0
0 0 l

(5.308)

where
p2
Foz
dvz
C kvz
2
1 Foz
= 1 2p
dvz
k C v k vz

t = 1

(5.309)

(5.310)

All integrals are along the Landau contour, passing below the pole.

143

5.9.10

Electromagnetic Waves in unmagnetized Vlasov Plasma

For transverse waves the dispersion relation is

p2 1
k 2 c2
foz dvz
2
=
N
=

=
1

t
2
n C ( kz vz )

(5.311)

This has, in principle, a contribution from the pole at kvz = 0. However, for a non
relativistic plasma, thermal velocity is c and the EM wave has phase velocity c. Con
sequently, for all velocities vz for which foz is nonzero kvz . We have seen with the cold
plasma treatment that the wave phase velocity is actually greater than c. Therefore a proper
relativistic distribution function will have no particles at all in resonance with the wave.
Therefore:
1. The imaginary part of t from the pole is negligible. And relativisitically zero.
2.
p2 1
kvz k 2 vz2
f
1
+
+ 2 + ... dvz
oz

2 n

2
2
p
k T
1 + 2 + ...
2

2
p
k 2 vt2
1+ 2
2

2
p
2

t 1
= 1
1
1

Thermal correction to the refractive index N is small because

k2 vt2
2

(5.312)

1.

Electromagnetic waves are hardly aected by Kinetic Theory treatment in unmagnetized

plasma. Cold Plasma treatment is generally good enough.

144