Anda di halaman 1dari 46

Advanced Plasma Physics Notes 2008

Spike
April 6, 2008
Contents
1 Lecture 1: Kinetic Theory 3
1.1 Liouvilles Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 -space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 -space vs. -space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 The BBGKY Hierarchy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Lecture 2: Derivation of BBGKY Hierarchy 4
2.1 BBGKY Hierarchy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Lecture 3 6
3.1 Dilute Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Boltzmann Collision Integral from the BBGKY Viewpoint . . . . . . . . . . . . . . . . . . 7
3.3 Times Arrow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4 Lecture 4 8
4.1 Boltzmann Collision Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.1.1 Boltzmann Collision Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5 Lecture 5 10
5.1 Electron Scattering by Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.2 Fokker-Planck Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6 Lecture 6 12
6.1 More on the collision operator: BGK (Bahatnagar-Gross-Krook) Equation . . . . . . . . . 12
6.2 Vlasov Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6.3 Debye Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
7 Lecture 7 15
7.1 Weakly and Strongly Coupled Plasmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
7.2 Dielectric Properties of a Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
7.3 Charge and Current Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
8 Lecture 8 17
8.1 Landau Prescription . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
8.2 Charge Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
8.3 Plasma Dispersion function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
9 Lecture 9 20
9.1 Plasma Dielectric Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
9.2 Plasma Dispersion Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
9.3 Dispersion Relation for Longitudinal Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 21
9.4 Dispersion Relation for Transverse waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
9.5 Transverse Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1
9.6 Longitudinal Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
9.7 Physical Origin of Landau Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
10 Lecture 10: MHD 23
10.1 Hydrodynamic limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
10.2 Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
10.3 Momentum Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
10.4 Multi-Element Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
10.5 Bulk Fluid Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
11 Lecture 11 27
11.1 Ohms Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
11.2 Energy Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
11.3 Transport Coecients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
11.4 Hydro-Magnetic Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
12 Lecture 12: Ideal MHD Approximation 30
12.1 Pressure Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
12.2 Incompressible Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
12.3 Bernoullis Equation - Steady Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
12.4 Kelvins Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
12.5 Alfven Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
13 Lecture 13: Magnetosonic Waves 33
14 Lecture 14: Magneto-Hydrodynamic Waves 35
14.1 M.H.D. Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
15 Lecture 15: M.H.D. Shocks 38
15.0.1 Alfven Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
15.1 Boltzmanns Equation as a Fokker-Planck (Landaus Equation) . . . . . . . . . . . . . . . 39
16 Lecture 16 41
16.1 Rosenbluth, Macdonald and Judd Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
16.1.1 Relationship with Landaus Formula . . . . . . . . . . . . . . . . . . . . . . . . . . 45
17 Lecture 17: Calculation of the Transport Coecients 46
2
1 Lecture 1: Kinetic Theory
The motion of individual particles can be considered as an ensemble. We take as our starting point
Hamiltons equation for the dynamics of particles. Consider a system of N particles, where N is very
large. Each particle has q
i
, p
i
( r, v). We can now consider a space, , which consists of the co-ordinate
axis of q
i
, p
i
of all of the particles. (A 6N dimensional space). The particle state (as determined by q
i
, p
i
)
appears as a point, and the particle trajectory appears as a line in the space. The probability function
gives the probability that a particle will be found in the box formed by dq
1
dq
2
. . . dq
N
at q
1
p
1
. . . q
N
p
N
is given by Prob = F(q
1
p
1
q
2
p
2
. . . q
N
p
N
)dq
1
dp
1
. . . dq
N
dp
N
.
If the particle position is initially at some known point, we can say that
F(q
1
p
1
q
2
p
2
. . . q
N
p
N
, t) =
N
t=1
(q
i1
Q
i1
(t))(p
i1
P
i1
(t)) . . . (1)
where Q
i1
(t), P
i1
(t) are the integrals of the Hamiltonian equation in time, Q
i1
(t), P
i1
(t) is the classical
trajectory of particle i.
In fact, although we cannot describe the initial conditions exactly, we do know a probability distri-
bution for the particles. We can use this to form an ensemble of paths in phase space (6N, q
i
, p
i
). We
follow the motion of the ensemble in phase space.
1.1 Liouvilles Theorem
As we follow the motion of the probability cloud, Liouvilles theorem states that:
DF
DT
= 0 =
F
t
+
N
i=1
q
i
_
F
q
i
+ p
i
F
p
i
_
(2)
This follows directly from Hamiltons equation for a conservative system:
p
i
=
H
q
i
; q
i
=
H
p
i
(3)
This only works if the system is conservative. If we add a magnetic eld, so that the system is
non-conservative, Hamiltons equations are still applicable, and Liouvilles theorem is valid.
N.B. So for, the motion of the system is completely thermodynamically reversible. However, a real
plasma is not reversible. How is this?
1.2 -space
This is a six-dimensional phase space: (r, v, t).
The probability distribution function, f (r, v, t) gives the probability of nding any particle in the
volume element (dr, dv) at (r, v) as f (r, v, t) dr dv.
1.3 -space vs. -space
Consider the probability in -space of nding particle 1 in r
1
, v
1
, dr
1
, dv
1
to be Adr
1
, dv
1
and the other
particles anywhere in phase space.
A =

N
F(v
1
, r
1
, v
2
, r
2
, . . . , v
N
, r
N
, t)dr
2
, dv
2
, . . . , dr
N
, dv
N
(4)
N.B. There is no dr
1
, dv
1
in this equation.
But: the particles are indistinguishable; all particles have the same probability of being in the same
space element.
f
(1)
(r
1
, v
1
, t) = f (r
1
, v
1
, t) = NA = N

N
F(v
1
, r
1
, v
2
, r
2
, . . . , v
N
, r
N
, t)dr
2
, dv
2
, . . . , dr
N
, dv
N
(5)
But: the integration from r
2
r
N
removes all of the correlation information, for example f (r
1
, v
1
, t)
contains no enhanced probability for nding particle 1 in the neighbourhood of 2. In this case, the
probability distributions of the particles are totally independent of each other.
3
Consider the probability of nding a particle in r
1
, dr
1
, v
1
, dv
1
and a second in r
2
, dr
2
, v
2
, dv
2
si-
multaneously. This is the two-particle distribution function: f
(2)
(r
1
, v
1
, r
2
, v
2
, t)dr
1
, dv
1
, dr
2
, dv
2
. This
distribution includes correlation, i.e the probability of nding one particle is inuenced by that of a
second, the distributions of the particles are not independent, one depends on the other.
Correlation is very important. There are two types of correlation in plasmas. Collisional correlation
is short-range, and there is also a long-range correlation. In dense plasmas there may also be a third,
short-range correlation.
In terms of -space distribution:
f
(2)
(r
1
, r
2
, v
1
, v
2
) =
N!
(N 2)!2!

N
F(v
1
, r
1
, v
2
, r
2
, . . . , v
N
, r
N
, t)dr
3
, dv
3
, . . . , dr
N
, dv
N
(6)
where
N!
(N 2)!2!
=
N(N 1)
2
(7)
This can be generalised to any number of particles.
The general particle distribution function: r
1
, v
1
, dr
1
, dv
1
, . . . , r
x
, v
x
, dr
x
, dv
x
gives:
f
(q)
(r
1
, v
1
, . . . , r
q
, v
q
) =
N!
(N q)!q!

q+1

N
F(r
1
, v
1
, . . . , r
q
, v
q
, r
q+1
, v
q+1
, . . . , r
N
, v
N
)dr
q+1
, dv
q+1
, . . . , dr
N
, dv
N
(8)
gives a set of N-fold distribution functions in -space. The complete set contains all the information
in the -space distribution. Calculating this is very dicult.
If the particles are independent (i.e. there is no interaction between the particles) then
f
(2)
(r
1
, v
1
, r
2
, v
2
) = f
(1)
(r
1
, v
1
)f
(1)
(r
2
, v
2
) (9)
If the particle-particle interactions are weak, the motions are nearly independent, and the one-particle
distribution particle is nearly adequate. We can therefor work with a truncated set of distributions.
1.4 The BBGKY Hierarchy
A set of equations which enable the calculation of f
(q)
from f
(q1)
, f
(q+1)
. Eventually this set must be
truncated with some f
(q+1)
assumed to be small. For dilute gas plasmas we can truncate at f
(2)
, which
removes correlations. However, the full set of -space distributions are still reversible. The truncation
must also introduce irreversibility.
2 Lecture 2: Derivation of BBGKY Hierarchy
Forces on a particle:
Externally applied force: F
i
. e.g. gravity. (N.B. F = distribution function. Note lack of indices)
Internal forces: forces from mutual interactions: F
ij
, in general a function of the separation of the
particles: F
ij
([r
i
r
j
[).
F
t
+ v
i

F
r
i
+ v
i

F
v
i
=
N
i=1

N
j=1,j=i
F
ij
m

F
v
i
(10)
Remember: F is a symmetric function w.r.t. the interchange of particles.
Multiply (10) by
N!
(Nn)!
Integrate over dr
n+1
dv
n+1
. . . dr
N
dv
N
to yield f
(n)
(r
1
, v
1
, . . . , r
n
, v
n
)
N!
(N n)!

F
t
dr
n+1
dv
n+1
. . . dr
N
dv
N
=
f
(n)
t
(r
1
, v
1
, . . . , r
N
, v
N
) (11)
4
Integrals for i > n which contain derivatives with respect to one of the variables of integration, may
be integrated by parts to give a surface integral at the boundary and a dierent volume integral, whose
values are zero, for example

v
i

F
r
i
dr
i
=
_

dS
i
v
i
F

dr
i
F
v
i
r
r
_
= 0
since lim
r
F = 0, and v
i
and r
i
are independent variables.
N!
(N n)!

v
i

F
r
i
dr
n+1
dv
n+1
. . . dr
N
dv
N
=
_
_
_
v
i

f
(n)
r
i
(r
1
, v
1
, . . . , r
N
, v
N
) if i n
0 if i > n
_
_
_
Noting that the only external force which depends on the velocity is the Lorentz force due to an
external magnetic eld v B, there is no component parallel to the velocity and therefore F
i
/v
i
= 0
N!
(N n)!
1
m

F
i

F
v
i
dr
n+1
dv
n+1
. . . dr
N
dv
N
=
_
_
_
1
m
F
i

f
(n)
v
i
(r
1
, v
1
, . . . , r
N
, v
N
) if i n
0 if i > n
_
_
_
The mutual interaction between particles is treated by breaking the sum into three parts

N
i=1

N
j=1,j=i
=
n
i=1

n
j=1,j=i
+
n
i=1

N
j=n+1,j=i
+
N
i=n+1

N
j=1,j=i
(12)
N!
(N n)!

F
ij
(r
i
r
j
)
m

F
v
i
dr
n+1
, dv
n+1
, . . . , dr
N
, dv
N
=
F
ij
(r
i
r
j
)
m

f
(n)
v
i
(13)
where i n and j n.
N!
(N n)!

F
ij
m

F
v
i
dr
n+1
dv
n+1
, . . . , dr
N
, dv
N
=
1
(N n)

F
ij
(r
i
r
j
)
m

f
(n+1)
v
i
dr
j
dv
j
(14)
where i n and j > n.
Total contribution to the sum: multiply by (N-n) to take into account identical particles.
N!
(N n)!

F
ij
(r
i
r
j
)
m
=
F
v
i
dr
n+1
dv
n+1
. . . , dr
N
dv
N
(15)
Now, substitute these into the Liouville equation:
f
(n)
t
+
n
i=1
v
i

f
(n)
r
i
+
n
i=1
F
i
m

f
(n)
v
i
+
n
i=1

j=i,j=i
F
ij
m

f
(n)
v
i
=
n
i=1

N
j=n+1

r
j
v
j
F
ij
(r
i
r
j
)
m

f
(n+1)
v
i
(16)
2.1 BBGKY Hierarchy
To obtain f
(n)
you need f
(n+1)
. The complete set f
1)
. . . f
(N)
are equivalent to the Liouville equation.
We need to introduce closure: Limit the number of n, usually to n = 1 (or n = 2 for dilute systems).
Using n = 1 we get:
f
(1)
t
+v
f
(1)
r
1
+
F
m

f
(1)
v
=
1
m

F
12
(r
1
r
2
)

r
1
_
f
(2)
(r
1
, v
1
, r
2
, v
2
)
_
dr
2
dv
2
. .
(17)
where the third term is the collision term, giving the interaction force. If it is assumed that there is no
external force the Boltzmann equation is recovered.
5
f
(2)
t
+ v
1

f
(2)
r
1
+ v
2

f
(2)
r
2
+
1
m
_
F
12
(r
1
r
2
)
f
(1)
v
1
+F
21
(r
2
r
1
)
f
(2)
v
2
_
=

_
F
13
(r
1
r
3
)
f
(n+1)
r
1
+F
23
(r
2
r
3
)
f
(n+1)
r
2
_
dv
3
dr
3
(18)
Find an approximation based on physical considerations for the collision term. What are the con-
straints on the physics to get a sensible collision term?
3 Lecture 3
Consider the streaming derivative for the one particle distribution
Df
(1)
Dt

f
(1)
t
+v
1

f
(1)
r
1
+
F
1
m

f
(1)
v
1
. .
(19)
In the uid dynamic analogy, this is the Lagrangian time derivative for the uid owing in phase
space.Driving term for the velocity, namely the acceleration, is determined by the external forces alone.
In addition the time variation of the one particle distribution depends on an additional term determined
by the interaction forces, which itself depends of the two particle derivative.
n = 1 can also be written as:
Df
(1)
Dt
=
f
(1)
t

coll
(20)
3.1 Dilute Systems
If the system is dilute there are three well separated time scales:
1. Collision time:
0

a
v
where a = range of intermolecular force.
2. Collision frequency (time between collisions): t
0


v
where = the mean free path.
3. Macro. time:
0

L
vs
where L = Lab scale, and V
s
= speed of sound.
Since
The range of the interparticle force a is very small
The mean free path is relatively large a is large
The scale of the laboratory apparatus is extremely large L
Consequently
0
t
0

0
We prepare the system in some state at t = 0, not in equilibrium. Rapid changes occur over times
0
due to the interparticle force. The multiple particle distribution functions f
(s)
s 2 change rapidly over
these time scales, and establish an equilibrium amongst themselves. On the other hand the one particle
distribution f
(1)
on changes more slowly due to the streaming derivative and changes only over as time
t
0
as interactions take place.
We are usually not interested in changes over
0
, as it is too small and gives a level of detail, which
is not required. However changes over t
0
determining the one particle distribution directly relate to
experimentally measurable quantities such as temperature/density etc. We can therefore average the
detailed uctuations over times
0
without losing any important information.
When this averaging is carried out, the behaviour of the multiple particle distribution functions
rapidly achieve the equilibrium forms associated with the instantaneous one particle distribution. Since
6
this equilibration time is very short, we imagine that multiple particle distributions, on averaging, have
a time depending, whose time dependence is determined by that of the one particle distribution
f
(s)
(r
1
, v
1
, . . . , r
s
, v
s
, t) f
(s)
_
r
1
, v
1
, . . . , r
s
, v
s
)

f
(1)
(t)
_
(21)
where s 2.
As a result the equation for n = 1 becomes
Df
(1)
Dt
= A
_
r, v,

f
(1)
_
=
f
(1)
t

coll
(22)
We still have reversibility problems. There is no arrowof time. A loss of reversibility can be introduced
by introducing a direction of time by which before and after are clearly distinguished. This assumes
molecular chaos:
1. Before a collision (2-body) the 2 bodies are statistically independent.
2. After the collision they are correlated.
3. In time the correlation is lost between collisions. .
As a result, the particles involved in the collisions are always uncorrelated before the collision.
3.2 Boltzmann Collision Integral from the BBGKY Viewpoint
For dilute gases collisions are binary and completed over very short timescales.
Assumptions:
i 3-body (and higher) terms are negligible.
ii Time variation during collision is negligible.
iii Spatial variation over the collision range (a) is negligible.
iv Change in velocity due to external forces during collisions is negligible.
Take n = 2:

(ii)
f
(2)
t
+v
f
(2)
r
+v

f
(2)
r

+
F
m

f
(2)
v
+
F

m

f
(2)
v

+
$
$
$
$
$
$
$
$X
(iv)
F(r r

m

f
(2)
v
+
$
$
$
$
$
$
$
$$X
(iv)
F(r r

)
m

f
(2)
v
1
=
1
m

_
$
$
$
$
$
$
$
$$X
(i)
F(r r

)
f
(3)
r
+
$
$
$
$
$
$
$
$
$X
(i)
F(r

)
f
(3)
r

_dr

dr

(23)
Since there is no spatial inhomogeneity (iii), a small displacement leaves the two particle distribution
unchanged.
f
(2)
(r, v, r

, v

, t) = f
(2)
((r +), v, (r

+), v

, t) (24)
We can now dierentiate with respect to :
f
(2)
r
+
f
(2)
r

= 0 (25)
So (23) becomes:
(v v

)
f
(2)
r

+
F(r r

)
m
f
(2)
v
+
F(r

r)
m
f
(2)
v

= 0 (26)
For n = 1:
7
f
(1)

t
+v
f
(1)

r
+
F
m
f
(1)

v
=

..
_
F(r

r)
m
f
(2)
v
_
dr

dv

. .
(27)
where the left hand side is J, and the integrand of the right hand side can be substituted from (26) to
give:
J =

_
(v v

)
f
(2)
r


F(r

r)
m

f
(2)
v

_
dr

dv

(28)
integrate the second term over v

, using Gausss theorem.

F(r

r)
m

f
(2)
v

dv

= 0 (29)
Integrate the rst term over r

, can use Gauss to give:


J =

dv

dS

(v v

)f
(2)

dr

$
$
$
$
$
$
$$X
0
f
(2)

r

(v v

) (30)
It is possible to cancel the second term on the left hand side as v & v

are not functions of r

. The
surface integral is over the collision volume.
Consider the collision volume as a sphere of radius a, centred on the centre of mass. V is the relative
velocity, where V = vv

,which is unchanged in magnitude in an elastic collision. We consider (dierent)


pairs of particles with the same velocities after the collision as those before.

r

is the radius vector of
the outgoing particle at the surface on top (exit),

r

is the radius vector of the incoming particle at


the surface on the bottom (entry). The element dS

of the surface corresponding to the entry and exit


particles is projected into d on the equatorial plane. Thus d is the element of the collision cross
section.
3.3 Times Arrow
Assumption (v) Molecular chaos. Incoming particles are statistically independent.
f
(2)
(r, v, r

, v

) = f
(1)
(r, v) f
(1)
(r

, v

) (31)
This applies to the entry, but not the exit, particles.
For an elastic collision: Conservation of total momentum and energy, therefore centre-of-mass velocity
is constant and relative velocity is rotated through a scattering angle . can be specied from the
cross-section as = (v, )d. Hence the velocities after the collision v and v

can be calculated.
From Liouvilles theorem the two particle distribution function after the collision must be equal to
its value before the collision for the same two particles. Thus if r, v and r

, v

are the initial particle


positions and velocities which give rise to velocities r, v and r

, v

respectively after collision:


f
(2)
(r, v, r

, v

) = f
(1)
(r, v) f
(1)
(r

, v

) (32)
4 Lecture 4
4.1 Boltzmann Collision Operator
From the closing of the BBGKY Hierarchy:
J =

dv

V = f
(2)
(r, v,

, v

, t) f
(2)
(r, v,

, v

, t)dw (33)
where v = relative velocity, dw = dierential cross-section for collision through an angle .
We make the additional assumption of molecular chaos, where two particles have such a long history
before the collision that they are essentially uncorrelated before the collision.
8
Before Collision:
f
(2)
(r, v,

, v

) = f
(1)
(r, v) f
(1)
(r

, v

) (34)
Liouvilles Theorem: Distribution function f
(2)
is unchanged through the collision.
After collision assuming Molecular Chaos:
f
(2)
(r, v,

, v

, t) = f
(1)
(r, v, t)f
(1)
(r

, v

, t) (35)
r, v and r

, v

after collision result from r, v, r

, v

before collision.
Since the collision volume is very small, r = r

= r = r

=

r

=

r

J =

dv

dw
_
f
(1)
(r, v, t) f
(1)
(r, v

, t) f
(1)
(r, v, t) f
(1)
(r, v

, t)
_
(36)
where v and v

lead to v and v

after the collision.


J =

dv

d(V, )
_
f
(1)
(r, v) f
(1)
(r, v

) f
(1)
(r, v) f
(1)
(r, v

)
_
(37)
where dw = (v, )d is the dierential cross section for scattering through solid angle d for a relative
velocity V and angle of scatter .
4.1.1 Boltzmann Collision Operator
Boltzmann Collision Operator (for a 1 particle distribution function):
f
t
+v
f
r
+
F
m
f
v
=
f
t

coll
(38)
where f f
(1)
and
f
t

coll
= J. [Note that henceforward we shall omit distribution function superscripts
unless necessary].
The rate of change of f due to collisions = the rate of collisions bringing particles into the phase
space volume drdv - rate of collisions exiting drdv.
The number of collisions exiting drdv per unit time = f (r, v)drdv

f (r

, v

)V dv

dr

Number of collisions entering the volume element r v r

r v r

f (r, v)f (r

, v

) V dr dv dr

dv

The volume element is given by the Jacobian, for which drdvdr

dv

= drdvdr

dv

follows from the


dynamics.
Hence we obtain the same result as before
f
t

coll
=

_
f (r, v)f (r, v

) f (r, v)f (r, v

V dv

d (39)
All occur at the same place in space (due to the collision volume being small) so r is the same for all.
4.2 Plasma
Long range forces between particles (Coulomb Force):
Coulomb Force: (r, r

)
1
|rr

|
violates the short-range assumption.
Impose closure on the BBGKY: Only at short range is the Coulomb force strong. Assume a dilute
plasma - can separate a strong eld region and consider two body collisions only in this region. Outside
this region we consider the eld as weak, we can therefore treat the eld as a perturbation, following
many body collisions.
Strong collision region: collisions are essentially binary. Impact parameter less than of order of 90
o
scattering by a Coulomb eld, b b
min
, Landau length.
Weak collision region: Multi-particle, cooperative region Impact parameter b
min
b
D
(Debye
Length).(e.g. plasma waves).
9
Collective region: Impact parameter b
D
. Interaction characterised by collective behaviour re-
sulting in plasma waves due to non-local self-generated elds.
For small collisions: The particle wanders through the plasma with numerous small deections - many
body collisions: (interaction through the coulomb eld.
Electron Ion scattering: Momentum change of the particle is the sum of many terms. p = p
1
+
p
2
+ + p
N
. Interacting particles are randomly distributed (molecular chaos) giving p
x
0 but
p
2
x
,= 0. The statistical distribution of the particle deections. These are randomly distributed in angle,
Gaussian with variance proportional to length of path and proportional to time. The R.M.S. Deection
is proportional to the square root of time.
In the remainder of this course we shall consider dilute plasma in which the particle density is
suciently small that correlations are almost unimportant. However we cannot neglect them altogether
as they form the basis for both collisions and collective behaviour. For dilate plasma we may consider
only the one particle distribution function (f)
(1)
greatly simplifying calculations.
5 Lecture 5
5.1 Electron Scattering by Ions
We consider the scattering of electrons by randomly distributed stationary ions.
The impulse of each ion on the electron: p
i
Total impulse due to many ions i =
i
p
i
If the electron is reasonably well separated from the ions (i.e. the ions are far apart) the impulse from
each ion is small, and can be treated as a perturbation on the motion of the electron. The motion is
therefore not much disturbed by each impulse. Since each perturbation is small it acts independently of
the others despite being simultaneous. (This condition is only true for distant collisions). Since the ions
are randomly distributed, the net impulse on the electron will be zero, as long as the path is suciently
long. However
i
[p
i
[
2
,= 0. (In a particular direction j,
i
[p
ij
[
2
yields the mean square impulse in
the direction j.)
p
2
j
=
(
i
p
ij
)
2
N
(40)
where N is the number of ions encountered
Hence the mean squared impulse in time t:
p
2
x
=

i
(p
ix
)
2
+
$
$
$
$
$
$
$
$
$X
0
2
i=j
(p
ix
p
jx
)
N
(41)
=

[p
x
(b)]
2
Nvbdbddt (42)
where theta is the azimuthal angle and b the impact parameter.
The mean squared deection angle is given by:

2
x
=
p
2
x
(mv)
2
(43)
where (mv)
2
is the momentum of the electron.
Since the electron-ion interaction is weak, the electron path is approximately a straight line z = vt.
The transverse impulse at impact parameter b is:
p
xi
=
qQ
4
0

bdz
(b
2
+z
2
)
3
2
=
2qQ
4
0
bv
(44)
p
i
=
2qQ
4
0
bmv
2
(45)
where Q is the ion charge and q the electron charge.
10

2
x
=

bdb

d
_
2qQcos
4
0
bmv
2
_
2
Nvdt (46)
=
4q
2
Q
2
N
(4
0
)
2
m
2
v
3

db
b
dt (47)
where

db
b
is a logarithmic singularity, giving problems for b both large and small.
We now look at the case where b is large. Here we start to get some screening of ion potential by
other electrons, up to b >
D
(see later). (If b >
D
there is no impulse, this gives an upper limit on the
integral of
D
.)
Turning our attention to the case where b is small, the deection will be very large, and the assumption
of small angle deection is violated. We can treat the problem as a series of two particle collisions using
the Rutherford scattering formula (equivalent to a Boltzmann collision term). It is found that this result
is equivalent to a cut-o at b
min
(the Landau parameter/distance). This is the distance that corresponds
to 90
o
scattering.
b
min
=
qQ
4
0
mv
2
(48)

2
x
=
4q
2
Q
2
N
(4
0
)
2
m
2
v
3

D
bmin
db
b
dt (49)
=
4q
2
Q
2
N
(4
0
)
2
m
2
v
3
ln
_

D
b
min
_
(50)
Let ln = ln
_
D
bmin
_
known as the Spitzer parameter/Coulomb logarithm. ( 5 10). The physical
signicance of is that it is the ratio of the number of small-angle large-impact collisions to the number
of large-angle small-impact collisions. Post-hoc justication of the use of perturbation is provided by
being large.
p
2
x
,
2
x
lead directly to the Fokker-Planck equation, a method for calculating the distribution
function.
5.2 Fokker-Planck Equation
Multiple small interactions that change the distribution. We are interested in the probability function
that an electron with velocity v undergoes a collision that changes its velocity by (v, dv) namely
(v)dv in a time dt.
f (r, v, t) =

_
f (r, (v v), (t dt)

(v, v)dv (51)


If v is small we can expect, via the Taylor series:
f (r, v, t) =

d(v)f (r, v, (t dt))(v, v)


v
_
f
v
+ f

v
_
+
1
2
v
i
v
j
_


2
f
vivj
+ 2
f
vi

vj
+ f

2

vivj
_

(v, v)d(v) = 1 (52)


f
t

coll
=

v
i
_
f v
i

+
1
2

2
v
i
v
j
_
f v
i
v
j

(53)
where the rst term in square brackets is the coecient of dynamical friction and the second is the
coecient of diusion.
_
v
i
v
i
v
j
_
=
1
t

(v, v)
_
v
i
v
i
v
j
_
d(v)
11
Referring to the mean change in the velocity on the top line, and the mean change in the product of two
velocities on the bottom.
For electron scattering from stationary ions, we may use the expressions obtained earlier:
v
x
=
y
0 (54)
v
2
x
= v
2
y
, 0 (55)
v
x
v
y
0 (56)
where x and y are perpendicular to the electron motion. Writing for the total perpendicular velocity
component and | for the parallel we note that because the collisions are elastic:
1
2
m(v
2

+ 2v
2

) = is
unchanged. Hence
v

2
=
8q
2
Q
2
N ln
(4
0
)
2
m
2
v
v

=
1
2
1
v
v
2

=
(4q
2
Q
2
N ln
4
0
)
2
m
2
v
2
v
2

0
(57)
If we add these back into (53) we get the Fokker-Planck equation.
We will discuss electron-electron collisions later.
If the plasma is non-isotropic v
x
may not be zero, and v
x
v
y
may also not be zero.
In equilibrium (i.e. thermodynamic balance) (53):
f
t

coll
= 0 (58)
which xes the numerical constants.
6 Lecture 6
6.1 More on the collision operator: BGK (Bahatnagar-Gross-Krook) Equa-
tion
Dening the relaxation time that the distribution takes to go into a Maxwellian:
f
t

coll
=
1

(f f
0
) (59)
Where f
0
is Maxwellian, and is the relaxation time.
6.2 Vlasov Equation
Thus far we have considered collisional eects, where the electron is scattered by short range eects. (i.e.
by randomly distributed scattering elds, l <
D
).
Now, we look at co-operative eects: scattering by self-generated elds. If the interaction is over
longer distances (l >
D
) the system acts more like a uid, and we lose graininess. This gives rise to
the Vlasov equation describing the distribution function (f)
(1)
.
Imagine that the plasma is subdivided.
e
m
is constant, where e is the electronic charge and m the
electron mass. There are no changes on the macroscopic scale, and we hold the total charge and mass
constant. Now let the number of particles, N, tend to innity, but with total charge and mass constant.
e, m 0. Consider the static elds, which will produce an acceleration:
a
(0)
i
=
e
m
_
E
(0)
+v
i
B
(0)
_
(60)
where
e
m
is a constant.
Now consider the interactions between two particles, i and j:
12
a
(j)
i
=
e
2
4
0
m
(r
i
r
j
)
[r
i
r
j
[
3

e
2
m
0 (61)
In this limit two body eects are small, so we assume that the particles are weakly correlated, i.e.
independent of one another.
f
(2)
(r
1
, v
1
, r
2
, v
2
) = f
(1)
(r
1
, v
1
)f
(1)
(r
2
, v
2
) (62)
Put this into the 1st term of the BBGKY:
f
(1)
t
+v
1

f
(1)
r
1
+
_
a
(0)
+

a
(2)
1
f
(1)
(r
2
, v
2
, t)dr
2
dv
2
_

f
(1)
v
1
= 0 (63)
where a
(0)
is the static eld term, and the integral is over the self-consistent eld term.
Substituting for a
(2)
1
we introduce the self-consistent electric eld:
E
self(r,t)
=
e
4
0

r r

[r r

[
3
. .
f
(1)
(r

, v

, t)dr

dv

(64)
where the terms over the brace are equal to a
(2)
.
Similarly we may include a self-generated magnetic eld arising from the current in the plasma B
self
,
which will add an additional acceleration v B
self
to a
(2)
.
f
t
+v
f
r
+a
f
v
= 0 (65)
where we have omitted the (unnecessary) superscript on the distribution function, and
a =
e
m
_
E
0
+ E
self
+v (B
0
+B
self
)
_
(66)
and E
self
and B
self
are found by integrating over the local charge density and current density throughout
the plasma.
This is the Vlasov Equation.
The Vlasov equation is collision free, uid-like long range interactions. The interaction term is
collective, but non-collisional.
Empirically the collisions can be added by adding an extra term to the Vlasov equation:
f
t
+v
f
r
+a
f
v
=
f
t

coll
(67)
The collision free Vlasov equation gives wave solutions. The collision term adds damping.
6.3 Debye Length
Length over which individual particle graininess exists. Consider a plasma of temperature T
i
for the
ions and T
e
for the electrons. The ion charge is Ze and the electron density is n
0
e
. The plasma will be
quasi-neutral: n
i

ne
Z
. The electrons moving around the ions experience an attractive electro-static
eld. We would therefore expect the density of electrons close to the ions to be greater than the ambient
density:
n
e
= n
0
e
exp
_
e
kT
e
_
; n
i
= n
0
i
exp
_
Ze
kT
i
_
(68)
Consider the value around a single ion: charge Ze at r = 0. Assume that the interparticle ion distance
is small compared to the penetration distance of the eld, i.e. that electrons are not xed to particular
ions. Over long times and large distances we will lose the graininess of the individual particles. The
statistical distribution of the electrons will give a slowly varying potential, , described by Poissons
equation:

2
=
1

0
[Zn
i
e n
e
e] (69)
13
Spherically symmetric near an ion:
1
r
2
d
dr
_
r
2
d
dr
_
=
1

0
n
0
e
_
exp
_
Ze
kT
i
_
exp
_
e
kT
e
__
(70)
Boundary condition 1: Near the ion: =
Ze
40r
as r 0
Boundary condition 2: Far from the ion eld must vanish: 0 as r .
1
r
2
d
dr
_
r
2
d
dr
_
=
n
0
e
e
2

0
_
1
kT
e
+
Z
kT
i
_
. .
=

2
D
(71)
Substitute:
=
1
x
f (72)
x =
r

D
(73)
to obtain
d
2
f
dx
2
= f (74)
subject to
f (0) = 1 (75)
f (x) 0 as x (76)
(77)
whose solution is
f (x) exp x (78)
and thus
=
Ze
4
0
r
2
exp
_

D
_
(79)
where
2
D
=
n
0
e
e
2
0
_
1
kTe
+
Z
kTi
_
is the Debye length.
We may write this expression in alternative convenient forms:

2
D
=

0
kT
e
n
e
e
2
_
1 +
ZkT
e
kT
i
_
(80)
Divide this into Electron and Ion parts:

2
e
=
0kTe
nee
2

2
i
=
0kTi
nie
2
_
1

2
D
=
1

2
e
+
1

2
i
The eld falls o rapidly as r
D
, screening the Coulomb eld, and reducing the range of the eld
from to
D
.
The electron screening of elds is more general over
D
. Imagine a wall at the edge of the plasma.
The electron rate to the wall is
1
4
nv, where the electrons are lost. This produces an electron-decient
area adjacent to the wall, which will create a eld. This eld will retard electrons and accelerate ions,
and the area over which this violation of charge neutrality occurs is known as the sheath. It extends a
distance
D
from the wall.
The Debye length measures the penetration of electric elds into the plasma. However, the Debye
length only well formed if the number of particles in the Debye Sphere is large, or alternatively that the
Debye length is much larger than the interparticle separation.
14
7 Lecture 7
7.1 Weakly and Strongly Coupled Plasmas
The analysis leading to the Debye shielding concept is only applicable to a weakly coupled plasma where
the thermal energies are large compared to electro-static energies
The conditions for Debye Length analysis:
1. Dilute Plasma
2. Over a Debye Length the particles should form a statistically averaged continuum. (i.e. n

1
3
i
<<

D
).
If the plasma is dense:
D
n

1
3
i
we get a strongly correlated plasma. These plasmas are highly
structured. (We can also get dusty plasmas). The strength of correlation is dened via the correlation
coecient:
n

1
3
i

D
=
_

_
n

1
3
i

_

0
kT
e
Z
2
n
i
e
2
_
_

_
2
(81)
=
Z
2
n
1
3
i
e
2

0
kT
e
(82)
If > 1 then the plasma is strongly correlated.

_
Z
2
e
2
_

0
n
1/3
i
_
kT
e


Electrostatic potential energy between particles
Thermal Energy per particle
(83)
Therefore the plasma is strongly correlated when the electrostatic energy is greater than the thermal
energy. Strongly coupled plasmas are unusual, and very hard to make plasmas where 1 100 ish.
In this case the distributing can no longer be written in terms of a single particle distribution function as
the higher order terms are not small. Calculations of these systems are very dicult, and Monte-Carlo
methods must often be used. (Note that an ionic crystal is essentially a strongly coupled plasma.)
7.2 Dielectric Properties of a Plasma
Unmagnetised, Warm, Dilute
Long Wavelengths (>>
D
)
Coherent/Cooperative motion Vlasov Equation
Consider an imposed E.M. wave eld with wave vector k and frequency .
E
0
exp [i(t k r)] (84)
The response of the plasma to this eld is given by:
f
t
+v
f
r
+
q
m
E
f
v
= 0
_
f
t

coll
_
(85)
Assume that the applied eld is weak: f = f
0
+f
1
(perturbation) where f
0
is the ambient distribution
(Maxwellian) and f
1
is the perturbation. This gives equations for the electrons and ions. As the ion mass
is large, the eect on ions is much weaker than that on electrons. We therefore neglect ion motion:
f
1
t
+v
f
1
r

e
m
E
0
exp i(t k r)
f
0
v
= 0 (86)
15
Writing f
1
(v, r, t) as a Fourier transform

f
1
(v, k, )
f
1
(v, r, t) =

f
1
(v, k, ) expi(t k r) (87)
i( r v)

f
1
=
e
m
E
0

f
0
v
(88)

f
1
=
e
m
E
0

f0
v
i( k v)
(89)
Denominator singular when v

(v component | to k) satises:
kv

= (90)
We can treat this in one of three ways to maintain physical reality:
1. An initial value problem: Switch eld on at t = 0 Laplace Transforms.
2. Switch on: slowly increase eld from t = . Introduce a term exp (t) into the amplitude.
3. Add in collisions via BGK term with a collision time (which can be very large).
All three cases change the denominator of (89) to [i( k v) + ] where = for case 2, or
1

for
case 3, but is a very small quantity (0). This explicitly introduces causality, and removes reversibility.
7.3 Charge and Current Density
We calculate the Fourier components of the charge and current density. These are wave terms with the
same frequency and wave number as the applied eld.
(r, t) = (k, )
j(r, t) =

j(k, )
_
exp [i(t k r)] (91)

= e


f
1
dv over all velocity space (92)
=
e
2
m
lim
0

E
0

f0
v
dv
[i( k v) +]
(93)

=
e
2
m
lim
0

E
0

f0
v
(i( k v) +)
( k v)
2
+
2
dv (94)
=
e
2
m
lim
0

E
0

f0
v
[i( k v)]
( k v)
2
+
2
dv

e
2
m
lim
0

E
0

f0
v
( k v)
2
+
2
dv (95)
=
e
2
m
P

E
0

f0
v
[i( k v)]
dv
e
m
2

E
0
f
0
v
dv

k
(96)
For the rst integral we introduce the Principal Value of the integral as 0
P

f (x)
x
dx = lim
0

xf (x)
(x
2
+
2
)
dx = lim
0
_

f (x)
x
dx +

f (x)
x
dx
_
(97)
provided the function f (x) is continuous at x = 0. For the second integral we note that for ( kv

)
not small, the integrand is small, therefore the dominant contribution occurs when kv

= . Hence
lim
0

E
0

f0
v
( k v

)
2
+
2
dv

= lim
0
E
0

f
0
v

dv

( k v

)
2
+
2
= E
0

f
0
V

#
1

k
(98)
16
using the standard integral

1
a
2
+x
2
dx =

a
This yields the Landau Prescription (standard):

1
x
= P

1
x
+i(x) (99)
where (x) is the Dirac delta function.
8 Lecture 8
8.1 Landau Prescription

L
1
x
= P

1
x
+i(x) (100)
1. Oscillating part of the waves: exp [i(t k r)]. If this is exp [i(k r t)] the plus in the Landau
Prescription becomes a minus.
2. (x) term with +i is a direct consequence of causality: expression of times arrow.
3. Singularity is when v

=

k
, when particle velocity matches the phase velocity - resonance.
8.2 Charge Density

= e

f dv =
e
2
m
lim
0

E
0

f0
v
dv
i( k v)
=
e
2
m

L
E
0

f0
V
dV
i( k v)
(101)
f
0
is symmetric in v . This is an even function of v.
f0
v

,
f0
v1
,
f0
v2
are all odd in v

, v
1
and v
2
respectively. (v
1
, v
2
) = v

.
v

f0
v

is even.
Integrating:
E
0

f0
V

1
dv
1
(i( k v))
= 0 =
E
0

f0
V

1
dv
1
(i( k v

))
(102)

=
e
2
m
E
0

L
f0
v

dv

(i( k v

))
(103)
j is the charge velocity.

j =
e
2
m

v E
0

f0
v
dv
i( k v

)
(104)

=
e
2
m
E
0
v

f0
v

dv
i( k v

)
(105)

j
1,2
=
e
2
m
E
01,2

v
1,2
f0
v1,2
dv
i( kv

)
(106)
The charge density and current densities are not independent, but must satisfy an equation of continuity

t
+ j = 0 (Current continuity equation) (107)
i( k

j) =
e
2
m
E
0

B
1
i( k v)
i( k v)

U
0
f
0
v

dv (108)
17
We have two decoupled terms:
| longitudinal parallel to the eld propagation direction, k
transverse, perpendicular parallel to the eld propagation direction, k, two components (1,2)
Introduce the dielectric constant: (k, ).
From electrostatics:
D = E =
0
E +P (109)
where P = dipole moment per unit volume.
j =
P
t
(Rate of change of polarisation),P is an oscillating term. (110)
= iP
0
(111)
D
t
=
0
E
t
+j (112)
=
E
t
(113)
i = i
0
+

j
E
0
(114)
where
2
p
=
nee
2
0m
, and

f
0
=
f0
ne
. f
0
is the number of electrons per unit volume of phase space, n
e
is the
electron density, and

f
0
is the probability of nding the electron per unit volume.

0
= 1 +

2
p

2
v

f0
v

dv
(1
kv

)
(115)
= 1 +

2
p

2
_
_
_

2
k
2
_
v

f0
v

dv
(

k
v

)
_
_
(116)

0
= 1 +

2
p

f0
v

dv
(1
kv

)
(117)
= 1

2
p


f
0
dv
(

k
v

)
(118)

0
= 1 +

2
p


1
_

k
_
(119)

0
= 1

2
p


2
_

k
_
(120)

1
(x) = x
2

f0
v

dv
(x v

)
(121)

2
(x) = x

f
0
(x v

)
dv (122)

1
,
2
are components of the plasma dispersion function:
8.3 Plasma Dispersion function
18
2 4 6 8
z
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
G
G
G
G
G
0
Figure 1: Plot of the real and imaginary components of the plasma dispersion function G(z)
and its derivative G

(z) for real values of the argument z.


1
9
G(z) =

1
2

L
e

2
(z )
d (123)
I(z) = P
_

1
2

2
(z )
d
_
(124)
= I(z) (125)
= 2e
z
2

z
0
e
t
2
dt (Dawsons Integral) (126)
G(z) = 2e
z
2

z
0
e
t
2
dt +i
t
1
2
e
z
2
(127)
G

(z) = 2

1
2

2
i(z )
= 2(1 zG(z)) (128)

0
= 1 +

2
p

2
z
2
G

(z)

0
= 1

2
p

2
zG(z)
_
_
_
z =

k
_
m
2kT
_1
2
T is the electron temperature
For small z:
G(z) i
1
2
e
z
2
+ 2z(1
2z
2
3
+
4z
4
15
. . . ) (129)
For large z:
G(z) =
$
$
$
$
$X
very small
i
1
2
e
z
2
+z
1
(1 +
1
2z
2
+
3
4z
4
+. . . ) (130)
where = 0, 1, or 2 if z < 0, = 0, or > 0
9 Lecture 9
9.1 Plasma Dielectric Constant
There are two. One parallel to the wave:

0
= 1 +

2
p

2
z
2

G(z) (131)
and one perpendicular to the wave:

0
= 1

2
p

2
zG(z) (132)
where:
z =

k
_
m
2kT
_1
2
(133)
G

(z) = 2(1 zG(z)) (134)


9.2 Plasma Dispersion Relation
This is the relationship between frequency and wavelength. Written as D(, k).
Start from Maxwells equations:
20
1

0
B = j +
0
E
t
(135)
E =
B
t
(136)
B = 0 (137)
E =

0
(138)
(E) =
0
j
t
+
0

0
E
t
(139)

2
E
0

2
E
t
2
=
0
j
t
+
1

0
p (140)
Add the plane waves:

j
tilde
_
_
_
exp [i(t k r]
_
k
2


2
c
2
_


E =
1

0
_
1
c
2

j k
_
(141)
9.3 Dispersion Relation for Longitudinal Waves
Take the scalar product k

E:
_
k
2


2
c
2
_
k

E =
i

0
_
1
c
2
k

j k
2

_
(142)
Apply equation of continuity of current:

t
+ j = 0 (143)
i( k

j) = 0 (144)

c
k

j k
2
=
1

2
c
2
k
2
_
k

j (145)
_
k
2


2
c
2
_
k

E =
1
i
0
_
k
2


2
c
2
_
k

j (146)
Now introduce the dielectric function:

=
0

E

+
1
i

(147)
_
k
2


2
c
2
_

k

E = 0 (148)
If k
2
=

2
c
2
this is an electro-magnetic wave in free space. Not the right solution.
If k

E = 0 there is no applied eld: Also a wrong solution.
Finally, if

(k, ) = 0 this is the dispersion relationship for longitudinal waves.


9.4 Dispersion Relation for Transverse waves
Take the vector product with k
21
_
k
2


2
c
2
_
k

E =
i

0
c
2
k

j =

2
c
2
1
i
0
k

j (149)

=
0

E

+
1
i

Use this to eliminate



j term above (150)
_
k
2


2
c
2

0
_
k

E = 0 (151)
If k

E = 0 there is no applied wave.
The dispersion relation for transverse waves is therefore given by:
k
2
=

2
c
2

0
(152)
9.5 Transverse Waves

k
=
c
_

0
(153)
If the phase velocity

k
>> Thermal Velocity
_
_
2kT
m
_
, then:
z =

k
_
2kT
m
1 (154)
and we can use the large z approximation for the plasma dispersion relation.
The large z approximation:
G(z) i
1
2
e
z
2
+z
1
_
1 +
1
2z
2
+
3
4z
4
+. . .
_
(155)
The imaginary part of this approximation is negligible, and there is no damping.

0
= 1

2
p

2
zG(z) (156)
= 1

2
p

2
_
1 +
1
2z
2
+
3
4z
4
+. . .
_
(157)
1

2
p

2
(Due to z being very large) (158)

k
=
c
_
_
1

2
p

2
_
(159)
> c (160)
The phase velocity greater than the speed of light ( >
p
). The phase velocity is real, otherwise
the light is strongly attenuated. Therefore light will only propagate if the density is less than the critical
density. (
p
= ).
The group velocity is given by:
d
dk
= c

_
1

2
p

2
_
(< c) (161)
The information is therefore transferred slower than the speed of light.
Transverse waves electro-magnetic waves in plasmas.
Collisional damping does occur, resulting in inverse bremsstrahling or collisional absorption. (In this
analysis the damping term is zero).
22
9.6 Longitudinal Waves

= 0 (162)
1
2
G

(z) +k
2

2
D
= 0 (163)
zG(z) = 1 +k
2

2
D
(164)
Here z is complex, and is complex due to the damping:
=
r
+ i
..
(165)
where the indicated term is the damping term.
Assume a small k
D
(i.e. the wavelength is long compared to the Debye length,
D
). This gives
that z is large.
G(z) = i

1
2
e
z
2
+z
1
_
1 +
1
2z
2
+
3
4z
4
+. . .
_
(166)
The imaginary term is small. The asymptotic form of G(z) gives

2
=
2
p
+
3
2
v
2

k
2
=
2
p
(1 + 3k
2

2
D
) (167)
where v

=
_
2kTe
m
. This result is the Bohm-Gross frequency. In a cold plasma
2
=
2
p
cold plasma
wave. 3k
2

2
D
is the warm plasma correction. Warm plasma changes the resonant frequency, introducing
a wavelength dependence.
Landau damping is given by:
= () =
_

8
_1
2
p
(k
3

3
D
)
exp
_

1
2k
2

2
D

3
2
_
(168)
This is weak for long wavelengths. We get strong damping when k
D
1. No wave is formed for very
short wavelengths (if k
D
1) as it is damped before oscillation.
9.7 Physical Origin of Landau Damping
The imaginary part is due to the resonance where v

=

k
(phase velocity). This depends on the gradient
of the density distribution at v

=

k
_
f0
v

k
_
.
The wave tends to accelerate (pull along...) particles with v |<

k
, whilst it slows particles with
v

>

k
. The slow particles are accelerated, the fast decelerate. However, there are more slow particles
than fast, which leeches energy from the wave, giving wave damping.
In the frame of the wave, the eld has a well, in which particles can become trapped. The wave
changes phase slightly due to the trapped electrons, which introduces phase incoherence. This also
damps the wave.
10 Lecture 10: MHD
10.1 Hydrodynamic limit
Consider the system in an averaged sense - averaged over the distribution:
n =

f (v)dv (169)
where n is the number density. This gives the particle numbers per unit density.
nmu =

mvf (v)dv (170)


23
where nmu is the momentum per unit volume.
nE =

1
2
mv
2
f (v)dv (171)
where nE is the energy of a particle. The velocity of a particle has 2 parts. The averaged ow velocity
(u), and the random (thermal) velocity (w).
v = u +w (172)
The thermal energy is given by:
n =

1
2
mw
2
f (v)dv (173)
Since the mean of w is zero:
nmw =

wf (v)dv = 0 (174)
Therefore energy is given by:
E = +
1
2
mu
2
(175)
where
=
1
2
mw
2
=
3
2
nkT
if the distribution is isotropic.
Use the Boltzmann equation as a starting point to form averages. First dene a quantity Q(v):
Q(r, t) =

Q(v)f (r, v, t)dv


n(r, t)
(176)
The force on the particles:
F = ma = q(E +v B) m (177)
Where is the gravitational potential. Consider the moments of v B about terms in the Boltzmann
equation:

Q
f
t
dv =

t

Qf dv =
(Qn)
t
(178)

Qv
f
r
v =

r

Qvf (v)dv =

r
(nQv) (179)

Qa
f
v
dv =

Qa
i
f
vi
dv =

v
i
(Qa
i
)dv (180)
after interaction by parts, noting that boundary terms at v : f 0. Since the only forces on
the particle depending on velocity are those due to the magnetic eld, which have no component in the
direction x
i
depending on v
i
. Therefore
a
i
x
i
= 0 (181)

Qa
f
v
dv = na
Q
v
(182)

Q
f
t

coll
dv =

t
(nQ)

coll
(183)
since the particle number, momentum and energy are conserved in the collision. Inserting the momentum
into the Boltzmann equation gives:

t
(nQ) +

r
(nQv) na
Q
v
=

t
(nQ)

coll
(184)
24
10.2 Continuity Equation
Q(v) = 1
n
t
+

r
(nv) = 0 (185)
This excludes ionisation events, and is the equation for both electrons and ions (dierent number density).
10.3 Momentum Equation
Q(v) = mv (186)
For a single element uid (e.g. a gas):

t
(nmv) +

r
j
(nmv
i
v
j
)
nF
i
m

v
j
(mv
j
) = 0 (187)
in Cartesian tensor notation. F is the total force on each particle, m its mass.
v
i
v
j
= u
i
u
j
+w
i
w
j
+

B
0
u
i
w
j
+

B
0
w
i
u
j
(188)
where nmu
i
u
j
is the momentum ux, nmw
i
w
j
is the momentum transfer (stress tensor), and the other
terms cancel to zero because the average of w is zero.
nmw
i
w
j
=
ij
(189)
where
i,j
is the total stress tensor.
If the uid is assumed to be isotropic: w
i
w
j
= 0 where i ,= j. and the only stress is the hydrodynamic
pressure, acting inwards:

ij
= p
ij
=
1
3
nmw
2

ij
(190)
where

i,j
=
_
1 if (i = j)
0 otherwise
is the Kroneker delta
There is another stress term, associated with viscosity in a velocity gradient, when the uid is non-
isotropic:

ij
= 2(
ij

1
3

ii

ij
) +
ij

ij
(191)

ij
=
1
2
_
u
i
r
j
+
u
j
r
i
_
(192)
and hence the total stress

i,j
=
i,j
p
i,j
(193)
Making use of the equation of continuity we obtain the familiar equation from uid mechanics
nm
u
t
+nm(u )u

x
i

i,j
nF = 0 (194)
25
10.4 Multi-Element Fluid
Momentum is exchanged between particles in a collision:

t
_
nQ

coll


t
[nmv]
coll
,= 0 (195)
Introduce an exchange term: P
ij
= average momentum exchange from particle i to particle j.
(sec/unit volume). From particle i:
n
i
m
i
_
u
i
t
+
_
u
i


r
_
u
i
_

&
&
&
&b
+

r
p
i

r
l
[
il
] n
i
F
i
= P
ij
(196)
In plasma viscosity is generally negligible:

r
l


r
(p
i
) = p
i
(197)
Replacing the external force by that due to the electric and magnetic elds
n
i
m
i
_
u
i
t
+ (u
i
)u
i
_
+p
i
= zen
i
[E +u
i
B] +P
il
(198)
n
e
m
e
_
u
e
t
+ (u
e
)u
d
_
+p
d
= en
e
[E +v
i
B] +P
il
(199)
where the rst equation is for ions, and the second is for electrons. The viscous terms and gravity are
neglected.
10.5 Bulk Fluid Equation
n
e
zn
i
quasi-neutrality Debye length is short.
Equation of continuity

t
(n
i
m
i
) +

t
(n
e
m
e
)

t
+ (u) = 0 (200)
u =
(miui+meu
e
)
mi+me
Mean ow velocity (201)
= n
i
m
i
+n
e
m
e
Density (202)
Momentum Equation

_
u
t
+ (u )u
_
= p +
..
qE +j B (203)
is Eulers Equation, with added E.M. eld force.
q = zen
i
en
e
0 Charge Density (204)
j = zen
i
u
i
en
e
u
i
Current Density (205)
m
e
m
i
(206)
u u
i
(207)
If the plasma is quasi-neutral:
u
i
u
e
(208)
and u
i
and u
e
dier only by a rotational vector, i.e. the dierence in velocity vector forms closed loops in
the plasma. Since the bulk velocity is approximately equal to the ion velocity is small, these are current
loops of electrons independent of bulk motion, governed by Ohms Law:
26
m
e
m
i
(209)
m
e
n
e
e
2
_
zen
i
u
i
t
en
e
u
e
t
_
=
m
e
n
e
e
2
j
t
(210)
= E +u B
j B
n
e
e
+
p
e
n
e
e
+
P
ie
n
e
0 (211)
u
e
= u
j
n
e
e
(212)
u
e
u
i
(213)
The
j
t
term is usually negligible. The generalised eld
E

= E +u B
j B
n
e
e
+
p
e
n
e
e
(214)
In an isotropic (non-magnetised plasma), the momentum exchange
P
ie
n
e
e
(u
i
u
e
) (215)
j (216)
=
j

(where is the conductivity) (217)


= E

(eqn.213) (218)
11 Lecture 11
11.1 Ohms Law
This derives from the current equation. The current is carried by electrons, as they are lighter than the
ions, and is caused by the dierence between the electron velocity and mean speed:
j = en
e
(u u
e
) (219)
m
e
n
e
e
2
_
z
e
en
i
n
i
t
eu
e
u
e
t
_
=
m
e
n
e
e
2

!
0(almost always)
j
t
(220)
=
eective electrical eld (with electron pressure)
..
E
..
E-Field
+ u B
. .
Lorentz force (induced)

j B
n
e
e
. .
Hall term
+
P
ie
n
e
e
+
p
e
n
e
e
..
electron pressure
(221)
Ohms law:
E =
j

(222)
where is the conductivity (scalar in an unmagnetised plasma, second order tensor when magnetised).
We expect P
ie
j (current density), leading to, assuming that the medium is isotropic, Ohms law:
1

=
P
ie
n
e
e
(223)
This is negative because we are dealing with negative charges.
27
Thermo-electricity, the Thompson eect, is current driven by a temperature gradient.
P
ie
n
e
e
=
j

T (224)
where is the thermo-electric coecient. In a magnetic eld , are both tensors with three independent
components. (Again, assume a non-isotropic medium in the presence of a magnetic eld):
=
_
_

0 0
0

_
_
where

= electrical eld along magnetic eld

= electrical eld perpendicular to the magnetic eld

= cross-product component arising when E and B are normal to each other


11.2 Energy Equations
Set Q =
1
2
mv
2
, and ions, electrons and E.M. eld, we get the force balance equation:
U
t
+ W = 0 (225)
Where U is the energy density, and w is the energy ux:
U =
3
2
n
i
kT
i
+
3
2
n
e
kT
e
+
1
2
u
2
+

0
E
2
2
+
B
2
2
0
(226)
W =
_
3
2
n
i
kT
i
+p
i
_
. .
Enthalpy of the ions
u
i
+
_
3
2
n
e
kT
e
+ p
e
_
. .
Enthalpy of the electrons
u
e
+
_
1
2
m
i
n
i
u
2
i
_
. .
ion K.E. density
u
i
+
_
1
2
m
e
n
e
u
2
e
_
. .
electron K.E. density
u
e
(227)
+
1

0
E B
. .
Poynting vector

ij
v
j
. .
viscous work done
+ q
..
thermal ux
(228)
Where q is the energy ux due to thermal transport, and is the heat transfer term in the above equation.
q = T E

(229)
Where E

is the net eective electric eld. In a magnetised, non-isotropic plasma and are tensors,
where
= T
e
+
5
2
kT
e
e
(230)
where is the same as that in equation(224) as required by the Onsager relations (reciprocal kinetic
process).
11.3 Transport Coecients
For a small perturbation, the ux is proportional to the appropriate driving force: Conductivity, Thermal
Conductivity, Thermo-electric Coecient, Viscosity
1
. These are all collisional terms, and, if there are
magnetic elds as well, these are also all tensors.
If you apply a force to a particle distribution, the distribution will be distorted away from the
Maxwellian. Provided the force is weak, the distortion will be small (perturbation), and the ux will be
linearly proportional to the force. If the force is strong, this linearity will break down.
Distribution function can be written with a perturbation f
1
on a steady state Maxwellian f
0
1
Only associated with ions, usually neglected
28
f = f
0
+ f
1
(231)
Applied to electrons for , , , and to ions for . If we assume that f
1
is small, the conditions on f
are given by:
0 =
f
0
t

c
(232)
This gives a Maxwellian for f
0
.
f
0
t
+v
f
0
r
+a
f
0
v
. .
Force
= C(f
0
[f
1
))
. .
Collision term
(233)
where C(f
0
[f
1
) is the rst order perturbation term introduced by f
1
. Hence knowing that (f)
0
is
Maxwellian yields the l.h.s. from which the perturbed distribution function f
1
can be calculated. The
solution of f
1
gives the perturbed distribution, and f
1
gives you the ux.
11.4 Hydro-Magnetic Approximation
We assume that the plasma has innite conductivity, (or, more correctly, that the magnetic
Reynolds number R
m
(= Lu) , where L is the characteristic length, and u is the characteristic
velocity). Since j is nite, this gives E

= 0. We assume that there is no Hall term, and no electron


pressure term E +u B = 0. This leads to Faradays Law (by taking the curl) for the ux :
B
t
= (u B) (234)
We now consider through an arbitrary loop S which is xed in the uid, i.e. made up of the same
uid particles for all time:
=

A
B dA (235)
where A is the area vector, which changes in time as the uid moves. The area changes due to the
motion:
dA
dt
=

A
u dl (236)
where the integral is around the loop
Hence
d
dt
=

A
B
t
..
_1
dA +

B u dl
. .
_2
(237)
The rst term _1 is due to time variation in the magnetic eld, the second _2 the result of ux swept
out by the motion.
Thus changing the order of terms in the scalar triple product and using Stokes Theorem:
d
dt
= 0 (238)
and the ux remains constant through the loop.
Consider a tube of force with walls parallel to B. Put a loop into a wall of the tube. is therefore 0
since the normal component of B is zero . This remains constant (= 0) as the plasma moves, as there is
no B through the loop. The loop is arbitrary, moves with the uid along a stream line, and

B dA = 0
for all such loops, so the uid, which ties the loop, also follows a ux line in addition to the stream lines.
It may also be said that the ux lines are tied to the motion, the frozen in condition.
29
(u B) = (B )u (u )B +
$
$
$
$X
0
u B B u (239)

t
+ (u ) = u (240)
d
dt
_
B

_
=
d
dt
_
B

_
+u
_
B

_
(241)
=
_
B


_
u (242)
Consider the movement of an element l xed in the eld:
d
dt
[l] =

t
[l] +u [l] = (l )u (243)
If l initially lies on a eld line, then it will continue to do so.
B l (244)
12 Lecture 12: Ideal MHD Approximation
12.1 Pressure Balance
If the plasma is at rest, the acceleration term can be rewritten as:

du
dt
= 0 (245)
= p +j B (246)
p = j B (247)
If we use Maxwells laws:
j =
1

0
(B) (248)
p = j B (249)
=
1

0
(B) B (250)
=
1

0
_

_
1
2
(B
2
)
. .
Mag. Pressure Term
(B )B
. .
Longitudinal Stress
_

_
(251)
where the longitudinal stress acts along the eld lines.
For many applications (B )B = 0, as we are only interested in the ability of the magnetic eld to
constrain the plasma. The pressure balance equation is therefore:
p +
1
2
0
B
2
= Constant (252)
The plasma , which, for magnetic connement devices, relates to the stability etc, is dened as:
=
P
max
_
B
2
max
20
_ (253)
For all conned plasmas, < 1, and should be signicantly less than one.
30
In the the adiabatic limit when: , and q 0, i.e no dissipation or entropy increase the energy
equation reduces to the adiabatic gas law:
3
2
dp
dt
+
5
2
p

d
dt
= 0 (254)
where
d
dt
=

t
+ucdot (255)
is the time derivative moving with the plasma.
Hence
_
p

_5
3
= constant
12.2 Incompressible Approximation
The equation of motion:

_
u
t
+ ( u) u +
1
2
(u
2
)
. .
(u)u
_

_
=
_
p +
1
2
0
B
2
_
+
1

0
(B )B (256)
and the vorticity = v. If the last term is zero, i.e. j B can be written as a potential, the ow
may be described in terms of the usual uid mechanics relations.
12.3 Bernoullis Equation - Steady Flow
For a uid:
1
2
u
2
+
p

= Constant along a streamline (257)


If we take the scalar product of u with the equation of motion:
u
_
p

+
B
2
2
0

+
1
2
u
2
_
=
1

0
u (B )B
. .
Mag. Stress Term
(258)
where the magnetic stress term must be much less than 1 for Bernoulli to apply. If we integrate along a
streamline, we nd that:
p

+
B
2
2
0
+
1
2
u
2
= Constant (259)
We now re-write the stress term in terms of a component along the eld lines, and another normal
to the eld lines. Let b be the unit vector in the eld direction, and u be the unit vector normal to the
eld line:
1

0
u (B )B =
1

0
_

_
u u
B
2
R
. .
Normal to the Field line
+ (u

b)(

b (
1
2
B
2
)
. .
Along the eld line
_

_
(260)
Where R = radius of curvature of the eld line.
31
12.4 Kelvins Theorem
The circulation around a loop xed in the eld is constant, and given by:
=

u dl =

u
. .
Vorticity
ds (261)
We now take the curl of out equation of motion ( = 0):

_
u
t
+ u
_
=
1

0
[(B )B] (262)
If we integrate the left hand side over the loop:
d
dt

ds = 0 recover Kelvins theorem (263)


But the right hand side gives a non-zero contribution (Kelvins theorem fails in the plasma). The
plasma generates vorticity. In contrast to a uid, a magnetised plasma can support shear, i.e. the eld
imposes a degree of stiness to the plasma. In consequence although uids only support longitudinal
waves (only sound waves, no transverse waves), a plasma allows transverse waves (via vorticity) which
have vorticity associated with them, in addition to the longitudinal waves.
Transverse:
B
x
z
,= 0 Non-zero curl = Vorticity (264)
Longitudinal:
u
x
z
= 0 : u
x
= 0 (265)
Solids also support both longitudinal and transverse waves (e.g. earthquakes).
12.5 Alfven Waves
The simplest possible case of transverse waves: uniform incompressible innite conductivity plasma,
moving at a constant velocity, u, in a uniform eld,B. We move to the rest frame of the plasma. The
plasma is perturbed by small changes in velocity u

and eld B

. The linearised equations are:

t
= p

+
1

0
(B

) B (266)
Adding Maxwell gives:
B

t
= E = (B

)u

(267)
where E = u B

.
We now dierentiate (266) with respect to time, and substitute for (267):

2
u

t
2
=
_
p

t
_
+
1

0
[(B )u

] B (268)
Now take the curl:

t
2
=
1

0
(B )[(B )

] (269)
N.B. B is a constant.
Introduce a plane wave:

=
0
exp [i(t k r)] (270)
where is the vorticity, and is the angular frequency.
The phase velocity is given by:
32
V =

k
= C
A
cos (271)
Where C
A
is the Alfven velocity, and is the angle between k and B, the angle of propagation with
respect to the eld lines.
C
A
=
B

(272)
String tension:
B
2
0
, String mass: .

t
2
=
1

(B )[(B )

] = Vorticity (273)
The only direction in which

varies is along B. Since = u, this implies that particle velocity u

is perpendicular to the magnetic eld superimposed on the steady background motion.


For transverse waves: u

= 0, as these are incompressible waves anyway, and there is no need to


make the incompressible approximation.
13 Lecture 13: Magnetosonic Waves
An incompressible plasma can only support transverse waves (Alfven waves). Plasmas are, however,
compressible. This adds longitudinal (sonic) waves. We expect that the new waves will be mixed
longitudinal and transverse. Consider a perturbation on an ambient condition assuming that the plasma
is initially at rest:
Density:
0
+ ( is the perturbation,
0
is the ambient condition
Pressure: p
0
+p
Magnetic eld: B
0
+B
Velocity: v (v
0
= 0 after changing the frame to the plasma rest frame)
Displacement: , where v =
d
dt
.
Linearising the M.H.D. equations, and assuming adiabatic ow:

t
+
0
v = 0 (274)

v
t
= p +
1

0
(B) B
0
(275)
p
t
=
p
0

0
t
(276)
E +v B = 0 (277)
B
t
= E (278)
v =
d
dt
=

t
The other term is neglected here) (279)
Here, (274) becomes:
+
0
= 0 (280)
If we combine (276) and (280) we get:
p =
p
0

0
= p
0
(281)
33
Combining (277), (278) and (279) gives:
B
t
=
0
_

t
B
0
_
(282)
B = ( B
0
) (283)

t
2

p
0

0
( )
1

0
(B) B
0
= 0 (284)
Let:
(B) B
0
=
_

_
B
2
0
(B
0
)(B
0
)

+ (B
0

2
) B
0
_
(B
0
)( )
_
(285)
Therefore:

t
2
+p
0
( )
1

_
B
2
0
(B
0
)(B
0

0
(B
0
)
2
+
1

0
B
0
(B
0
)( ) = 0(286)
We now introduce the Alfven speed:
C
2
A
=
B
2
0

0
(287)
The sound speed:
C
2
s
=
p
0

0
(288)
And the ratio:
=
C
2
s
C
2
A
=
_

2
_
(289)
where =
p0

B
2
2
0

This gives:
1
C
2
A

t
2

_
(1 +) (b )(b )

(b )
2
+ (b )( )b = 0 (290)
where b =
B
0
|B
0
|
, and is the unit vector in the B-direction.
N.B.

2

t
2
is not parallel to , so the wave is not longitudinal. Indeed, it must include an Alfven wave,
which is transverse.
We now introduce a plane wave: =
0
exp[i(t k r)]. The equation will contain 3 vector com-
ponents. We could go to the three Cartesian coordinates are obtain equations of the form:
V
2
C
2
A

i
+

j
A
ij

j
= 0 (291)
where V is the phase velocity, and A
ij
(b, k, ) is what we need to solve for the wave.
Forming the components of the wave in three directions: k, ( to k in plane (k, b)), and ( to k and
b).
Form k : gives components parallel to k.
Form b : gives components parallel to b.
Form (b k) : gives components perpendicular to both b and k.
34
1st:
k
_
V
2
C
2
A
(1 +)
_
k + (b k)(b ) = 0 (292)
2nd:
b
V
2
C
2
A
b (b k)(k ) = 0 (293)
3rd:
(b k)
_
V
2
C
2
A
(b k)
2
_
b k = 0 (294)
These three equations must be satised by any general wave. If we have a wave with parallel to
b k, i.e. perpendicular to both b and k, we get:
_
V
C
A
_
= b k (295)
The phase velocity V = C
A
cos where () is the angle between b and k. These are the Alfven waves.
They are characterised by having displacement, perpendicular to b and k. Two coupled equations
where is in the plane of k and b:
_
V
C
A
_
4
(1 +)
_
V
C
A
_
2
+ cos
2
= 0 (296)
Obtaining consistence for the equations for (k ) and (b ). This is a quadratic equation for V
2
. There
are two distinct waves, each of which can move forwards or backwards (fast or slow waves).
V
2

=
1
2
_

_
(C
2
A
+C
2
s
)
_
(C
2
A
+C
2
s
) 4C
2
A
C
2
s
cos
2

. .
Two waves, neither all longitudinal nor all transverse
_

_
(297)
If = + this is a fast wave, if = this is a slow wave, with the condition that k, b, all couple.
14 Lecture 14: Magneto-Hydrodynamic Waves
1
C
2
A

t
2
[(1 +) (b )(b )] (b )
2
+ (b )( )b = 0 (298)
Introduce a plasma wave: =
0
exp
_
i
2

(V t k r)

where v is the phase velocity, and k is the unit


vector in the wave direction.
Form the terms k , b , (b k) ).
_
V
2
C
2
A
(1 +)
_
k + (b k)(b ) = 0 (299)
V
2
C
2
A
b (b k)(k ) = 0 (300)
_
V 62
C
2
A
(b k)
2
_
b k = 0 (301)
Case 1: perpendicular to b and k:
k = b = 0 (302)
b k ,= 0 (303)
V
2
= (b k)
2
C
2
A
(304)
35
Let be the angle between b and k.
V
2
= C
2
A
cos
2
Alfven waves (305)
Case 2: lies in the plane of b and k.
b k = 0 (306)
The remaining two equations are coupled.
There are two waves, since there are two solutions. We require consistency between the two equations.
_
V
2
C
2
A
(1 +)
_
V
2
C
2
A
+(b k)
2
= 0 (307)
V
4
C
4
A
(1 +)
V
2
C
2
A
+cos
2
= 0 (308)
(309)
Solutions:
V
2

=
1
2
_
(C
2
A
+C
2
s
)
_
[(C
2
A
+C
2
s
) 4C
2
A
C
2
s
cos
2
]
_
(310)
Two waves, fast (+) and slow (). These have both longitudinal and transverse components.
If = 0 k is parallel to b.
V
+
=
1
2
_
C
2
A
+C
2
s
+[C
2
A
C
2
s
[ = Larger of C
A
or C
s
(311)
V

=
1
2
_
C
2
A
+C
2
s
[C
2
A
C
2
s
[ = Smaller of C
A
or C
s
(312)
If =

2
b is perpendicular to k.
V
+
=
_
(C
2
A
+C
2
s
) (313)
V

= 0 (314)
This is a purely longitudinal wave, with no second wave. The magnetic pressure adds to the kinetic
pressure, like a sound wave.
The longitudinal and transverse component (to the magnetic eld) are given by: k

for longitudinal,
and k

for transverse.
k

=
_
V
2
C
2
A

_
_
V
2
C
2
A
(1 +)
_k

(315)
k

=
(1 +)
_
[(1 +)
2
4cos
2
]
(1 +)
_
[(1 +)
2
4cos
2
]
(316)
This will be greater than zero for a fast wave, and less than zero for a slow wave.
The major dierence between the waves is that in fast waves the longitudinal and transverse parts
are in phase, but in slow waves they are out of phase. Given an arbitrary initial condition, displacement
and velocity, the relative phases x the initial fast and slow components.
36
14.1 M.H.D. Shocks
In uid dynamics a large amplitude sound wave goes into a shock:
c
2
=
p

(317)
p
(
1

)
(318)
c
2

(1)

(319)
Thus as the density increases, the sound speed increases and waves from the back build up into a
sharply rising front - shock wave.
Large amplitude magneto-sonic waves can develop into shock waves. We consider shocks with normal
in the x direction. The complexity of the shock wave is much increased due to the magnetic eld. i.e.
the wave is not only longitudinal, which means that the tangential symmetry of the the gas dynamics
is lost. We get ows along the surface of the shock (y,z directions). Let us look at the jump conditions
in the rest frame of the shock, remembering that on either side of the shock front we have p
1
,
1
, B
1
, u
1
and p
2
,
2
, B
2
, u
2
, where u is the ow velocity, u = (u, v, w).
Conservation of mass

1
u
1
=
2
u
2
(A)
Conservation of momentum normal to the shock front
p
1

1
u
2
1
+
B
2
y1
+B
2
z1
2
0
= p
2

2
u
2
2
+
B
2
y2
2
0
+
B
2
z2
2
0
(B)
Conservation of momentum transverse to the shock

1
u
1
v
1

B
x1
B
y1

0
=
2
u
2
v
2

B
x2
B
y2

0
y-direction

1
u
1
w
1

B
x1
B
z1

0
=
2
u
2
w
2

B
x2
B
z2

0
z-direction
_

_
(C)
Energy:

1
u
1
_
h
1
+
1
2
(u
2
1
+v
2
1
+w
2
1
)
_
+
E
1
B
1

x
=
2
u
2
_
h
2
+
1
2
(u
2
2
+v
2
2
+w
2
2
)
_
+
E
2
B
2

x
(D)
where h = +
p

= enthalpy per unit mass, or specic enthalpy.


The transverse electric and normal magnetic elds are both continuous across the shock:
E
t1
= E
t2
and B
x1
= B
x2
(E)
We can simplify these equations. The velocity components may be set by an appropriate transforma-
tion to lie in the (x, y) plane only. A further transformation in y allows either the transverse momentum
ux or the transverse electric eld to be zeroed (normally the transverse electric eld). We can assume
innite conductivity - Ideal M.H.D.- so that E+v B 0. The transverse eld equations now become:
u
1
B
y1
v
1
B
x1
= u
2
B
y2
v
2
B
x2
u
1
B
z1
w
1
B
x1
= u
2
B
z2
w
2
B
x2
_
(F)
As a result the energy equation (E) becomes :

1
u
1
_
h
1
+
1
2
(u
2
1
+v
2
1
+

w
2
1
)
_
=
2
u
2
_
h
2
+
1
2
(u
2
2
+v
2
2
+

w
2
2
)
_
+
u
1

0
_
B
2
y1
+
&
&
B
2
z1
_
+
u
2

0
_
B
2
y2
+
&
&
B
2
z2
_
v
1
B
x1
B
y1

w
1

0
B
x1
B
z1
v
2
B
x2
B
y2

w
2

0
B
x2
B
z2
(D)
37
following the frame transformations.
Example 1: Flow normal to shock on Entry:
v
1
= 0 (320)
B
y1
= 0 (321)
B
x1
= B
x2
= B
x
(322)
B
y2
=
0

2
u
2
v
2
B
x
via momentum (323)
=
v
2
B
x
u
2
via transverse electric eld (324)
Either v
2
= B
y2
= 0, a normal gas dynamic shock, or u
2
2
=
B
2
x2
02
B
y2
,= v
2
,= 0, switch on shock.
15 Lecture 15: M.H.D. Shocks

1
u
1
=
2
u
2
= j (325)
p
1

1
u
2
1
+
B
2
y1
2
0
+
B
2
z1
2
0
= p
2
+
2
u
2
2
+
B
2
y2
2
0
+
B
2
z2
2
0
(326)

1
u
1
v
1

B
x1
B
y1

0
=
2
u
2
v
2

B
x2
B
y2

0
(327)

1
u
1
w
1

B
x1
B
z1

0
=
2
u
2
w
2

B
x2
B
z2

0
(328)

1
u
1
_
h
1
+
1
2
(u
2
1
+v
2
1
+w
2
1
)
_
+
E
1
B
1

x
=
2
u
2
_
h
2
+
1
2
(u
2
2
+v
2
2
+w
2
2
)
_
+
E
2
B
2

x
(329)
Due to the assumption that the plasma has innite conductivity:
E
t1
= E
t2
(330)
B
x1
= B
x2
(331)
E +v B = 0 (332)
u
1
B
y1
v
1
B
x1
= u
2
B
y2
v
2
B
x2
(333)
u
1
B
z1
w
1
B
x1
= u
2
B
z2
w
2
B
x2
(334)
If v
1
= 0, B
y1
= 0 There is no upstream variation in either ow or eld.
B
y2
=

0

2
u
2
v
2
B
x
(335)
=
v
2
B
x
u
2
(336)
u
2
2
=
B
2
x

2
(337)
or v
2
= B
y2
= 0 (338)
Where (337) and (338) represent two possible solutions:
v
2
= B
y2
= 0 No transverse eld, same as an ordinary gas dynamic shock.
u
2
=
Bx

02
Transverse elds, switched-on shock.
In practice, the solution is determined by the stability and evolution of the shock. 1st get the less
likely shock, and then get the second shock. The shock involving the largest entropy change dominates
the interaction (by experiment). For weak shocks, this is the ordinary gas dynamics shock. Past a critical
point, increased strength = switched-on shock.
The three types of sonic waves, Alfven, Fast and Slow, can all produce shocks, and all have their own
type of shock. This produces very complex situations.
38
15.0.1 Alfven Wave
Consider v, w B
y
, B
z
: therefore v/B
y
/v = w/B
z
= . Assume that
B
2
y1
+B
2
z1
= B
2
y2
+B
2
z2
and therefore
v
2
y1
+v
2
z1
= v
2
y2
+v
2
z2
Dening the change across the shock by , thus
B
y
= B
y2
B
y1
Hence we obtain from the jump condition (C)
B
y
B
y
= B
z
B
z
v =
1

o
j
B
x
B
y
w =
1

0
j
B
x
B
z
Making use of the initial assumptions and conditions (F) we obtain
u
1
B
y
1
v
1
B
x
=
1
B
y
1
= u
2
B
y
2
v
2
B
x
=
2
B
y
2
u
1
B
z1
w
1
B
x
=
1
B
z1
= u
2
B
z2
w
2
B
x
=
2
B
z2
where
j =
1
u
1
=
2
u
2
;
1,2
= u
1,2
and
v
w
=
B
y
B
z
Thus the eld terms in (D) may be equated to

_
u(B
2
x
+B
2
y
) v(B
x
B
y
) w(B
x
B
z
)

= 0
It is easy to show that there exist solutions where the ow variables normal to the shock remain
unchanged, i.e. the disturbance is transverse along the shock,

1
=
2
u
1
= u
2

1
=
2
p
1
= p
2
Hence the wave speed is given by
u B
y
= B
x
v =
B
2
x

o
u
B
y
(339)
The wave speed,
_
B
2
x
0
, is the Alfven speed. This results in a discontinuous jump wave moving at
the Alfven speed, a nite amplitude Alfven wave. Thus far, we have not included collisions, and are
considering only collision free shocks. Where does the entropy change originate? Microscopic turbulence.
15.1 Boltzmanns Equation as a Fokker-Planck (Landaus Equation)
Boltzmanns equation is a binary collision equation. The Fokker-Planck equation is a many body equa-
tion. Our route here will be to start with weak collisions and small impulses, and add velocity slowly and
linearly. A large number of small binaries is a many body collision. We use Boltzmann, but with weak
collisions only, as there are many more weak collisions than strong collisions. We assume two identical
39
particles, here electron-electron scattering, with velocity v
1
, v
2
, relative velocity u = v
2
v
1
, and centre
of mass velocity V =
v
1
+v
2
2
. The cross-section for coulomb scattering through is:
=
_
e
2
2m
r
_
2
_
u sin
_

2
__
4
(340)
where m
r
=
1
2
m - electron reduced mass.
The Boltzmann collision integral:
I =
_
e
2
m
_
2

dv
2

sin dd
u
_
u sin
_

2
_
4
_
f
_
v
1

1
2
u
_
f
_
v
2
+
1
2
u
_
f (v
1
)f (v
2
)
_
(341)
u
f
= Ou O is a rotation (342)
u = u
f
u (343)
[u
f
[ = [u[ (344)
v
1

_
v
1

1
2
u
_
(345)
v
2

_
v
2

1
2
u
_
(346)
Unit vector l is parallel to u. Normal to l are orthogonal unit vectors l are m.n.
u = 2u sin
_

2
__
l sin
_

2
_
+mcos
_

2
_
cos () +nsin
_

2
_
sin ()
_
(347)
since magnitude of u is unchanged magnitude on scatter (elastic collision).
If the scatter is weak, u is small, we can expand terms via the Taylor series:
f
_
v
1

1
2
u
_
f
_
v
2
+
1
2
u
_
f (v
1
)f (v
2
)
=
$
$
$
$
$
f (v
1
)f (v
2
)
$
$
$
$
$
f (v
1
)f (v
2
) +
_
f (v
1
)
f
v

v
2
+ f (v
2
)
f
v

v
1
_
1
2
u
+
1
2
_
f (v
1
)

2
f
v
i
v
j

v
2
2
f
v
i

v
1
f
v
j

v
2
+ f (v
2
)

2
f
v
i
v
j

v
2
_
1
2
u
i
1
2
u
j
(348)
Integration over the azimuthal angle eliminates terms in cos , sin , cos sin since

2
0
dcos =

2
0
dsin =

2
0
dsin cos = 0
Hence substituting
I = 8
_
e
2
m
_
2

dv
2

0
d
_

2
_
cos
_

2
_
_
u
i
u
3
sin
_

2
_
_
f (v
1
)
f
v
i

v
2
f (v
2
)
f
v
i

v
1
+
+
1
2
_
f (v
1
)

2
f
v
i
v
j

v
2
2
f
v
i

v
1
f
v
j

v
2
+ f (v
2
)

2
f
v
i
v
j

v
2
_

1
u
_
u
2
u
1
u
2
sin
_

2
_
+
1
2
(m
i
m
j
+n
i
n
j
)
_
1
sin
_

2
_ sin
_

2
_
____
(349)
40
But

0
cos (

2
)d(

2
)
sin (

2
)
diverges at small and large . Physically this is due to the neglect of correlations
at small 0 (Debye shielding). As before we cut-o at the integral for small impact parameters at the
Debye length corresponding to

min


D
b
min
(350)
Hence
ln
_

_
sin
_

max
2
_
sin
_

min
2
_
_

_ = ln (351)
ln is large.
16 Lecture 16
I = 8
_
e
2
m
_
2

dv
2

0
d
_

2
_
cos
_

2
_
_
u
i
u
3
sin
_

2
_
_
f (v
1
)
f
v
i

v
2
f (v
2
)
f
v
i

v
1
+
+
1
2
_
f (v
1
)

2
f
v
i
v
j

v
2
2
f
v
i

v
1
f
v
j

v
2
+ f (v
2
)

2
f
v
i
v
j

v
2
_

1
u
_
u
2
u
1
u
2
sin
_

2
_
+
1
2
(m
i
m
j
+n
i
n
j
)
_
1
sin
_

2
_ sin
_

2
_
____
(352)


0
cos
_

2
_
d
_

2
_
sin
_

2
_ ln
_

_
sin
_

max
2
_
sin
_

min
2
_
_

_ = ln Spitzer/Coulomb Log. (353)

0
d
_

2
_
cos
_

2
_
sin
_

2
_
=
1
2
(354)
Assume ln is large.
m
i
m
j
+n
i
n
j
is the projection operator. It projects a vector, a, on to the plane perpendicular to g.
[m
i
m
j
+n
i
n
j
]a = m
i
(m a) +n
i
(n a) Component of a lying in m, n plane. (355)
= a l(l a) (356)
In tensor notation (357)
m
i
m
j
+n
i
n
j
=
ij
l
i
l
j
N.B. l is parallel to u. (358)
m
i
m
j
+n
i
n
j
=
ij

u
i
u
j
u
2
(359)
New Tensor: (360)

ij
=
1
u
_

ij

u
i
u
j
u
2
_
(361)

ij
u
i
=
2u
i
u
3
N.B. Repeated index = summation (362)
I = 2
_
e
2
m
_
2
ln

dv
2
_
2
_
f (v
1
)
f
v
i

v
2
f (v
2
)
f
v
i

v
1
_

ij
u
i
+

v
j

v
1
_
f (v
2
)
f
v
i

v
1
f (v
1
)
f
v
i

v
2
]
ij
+
_
f (v
1
)

2
f
v
i
v
j

v
2

f
v
i

v
1
f
v
j

v
2
_

ij
_
(363)
41
Noting

ij
u
i
=

ij
v
ij
=

ij
v
2j
and integrating by parts as necessary, we obtain Landaus equation:
I = 2
_
e
2
m
_
2
ln

v
ij

dv
2
_
f (v
2
)
f
v
i

v
1
f (v
1
)
f
v
i

v
2
_

ij
(364)
Landaus Equation

v
j

dv
2
_
f (v
2
)
f
v
i

v
1
f (v
1
)
f
v
i

v
2
_

ij
=

v
j
_
f (v
1
)
_

dv
2
f
v
i

v
2
_

ij
_
+

2
v
j
v
i
_
f (v
1
)
_
dv
2
f (v
2
)
ij
__


v
ij
_
f (v
1
)

dv
2
f (v
2
)
ij
v
ij
_
= 2

v
ij
_

_
f (v
i
)
_

dv
2
f
v
i

v
2

ij
. .
<vi,change in v
i
>
_

_
_

_
+

2
v
i
v
j
_

_
f (v
1
)
_
dv
2
f (v
2
)
ij
_
. .
<vivj>
_

_
(365)
(366)
Since

ij
v
1i
=

ij
v
2j
16.1 Rosenbluth, Macdonald and Judd Form
The detailed calculation of the collision terms foll wing Coulomb scattering is straightforward, but
complicated (Rosenbluth, Macdonald and Judd,Phys Rev 107, 1 1957). It leads to an equivalent result
as that of Landau for electron-electron collisions, but is more general as the particles may be dissimilar.
We consider the case of two particles of masses m and m

moving with velocities v and v

. As is well
known the collision occurs with scattering about the centre of mass with rotation of the relative velocity
vector V The velocity of each particle in terms of the centre of mass velocity U and the relative velocity
V are
v = U +
m

m+m

V
v

= U +
m
m+m

V
Hence the change in the velocity v as a result of the collision
v =
m

m+m

V
The scattering gives rise to velocity changes in the plane perpendicular,, and parallel, |, to the
initial relative velocity V as above. The cross section for scatter through an angle is modied by the
inclusion of the reduced mass m
m =
mm

m+m

42
but is otherwise unchanged. Therefore we may use the averages we have already calculated but with the
mass replaced by the reduced mass.
However the relative velocity V changes in direction and the centre of mass velocity U must be
removed before we can relate the particle velocities to their values in the laboratory frame. To do this
we identify a set of directions with respect to V , namely parallel to V and and perpendicular to
V so that
V

= V (1 cos ) = 2V sin
2
(/2)
V

= V sin cos = 2V sin(/2) cos(/2) cos


V

= V sin sin = 2V sin(/2) cos(/2) sin (367)


being the rotation angle in the plane , perpendicular to V .
Using either the small angle approach as in section 5.1 or directly from the full Rutherford cross
section as in section 15.1
() d =
Z
2
e
4
4 m
2
V
4
csc
4
(/2) d (368)
for particles of charge e and Ze. The resultant averages are carried out as described in section 5.1, but
in the centre of mass frame
V

) = V

) =
4Z
2
e
4
m
2
V
2
ln
V

) = V

) =
1
2
V

) = 0
V

2
) = V

2
) = 0
V

2
) = V

2
) =
1
2
V

2
) =
4Z
2
e
4
m
2
V
ln (369)
The unit vectors in the directions , and are

=
V
V
=

k V
V
x
2
+V
y
2
and =


where (,

j,

k) are the unit vectors in the (x, y, z) directions
Thus V in the laboratory set of Cartesian co-ordinates (x, y, z) may written
V
x
= (

)V

+ ( )V

+ ( )V

with similar expressions for V


y
and V
z
. Thus taking the average
V
x
) = (

)V

) + ( )
$
$
$$X
0
V

) + ( )
$
$
$$X
0
V

)
=
V
x
V
V

) (370)
with similar expressions for V
y
) and V
z
).
The product terms V
x
V
y
) etc. are obtained in a similar manner, noting that several terms cancel
to zero
V
x
V
y
) =
__
(

)V

+ ( )V

+ ( )V

_ _
(

)V

+ (

j )V

+ (

j )V

_ _
=
___

_ _

__

2
+
__

__

j
_
+
_

_
_

__
$
$
$
$$
V

+
_
_

_
_

j
__
V

2
+
_
( )
_

_
+
_

__

j
__
$
$
$
$$
V

+
_
( )
_

j
__
V

2
+
_
_

_
_

j
_
+ ( )
_

j
__
$
$
$
$$
V

_
=
_
_

_
_

j
_
+ ( )
_

j
__
1
2

2
_
43
Hence generalising

V
i
V
j
_
= [
i

j
+
i

j
]
1
2

2
_
= [
ij

i

j
]
1
2

2
_
(371)
since the components of the unit vectors

, , on the orthogonal set of unit vectors ,

j,

k namely
(
i
,
j
,
k
), (
i
,
j
,
k
) and (
i
,
j
,
k
) respectively satisfy the condition

j
+
i

j
+
i

j
=
ij
Returning to the laboratory frame, the change in the particle averages are given by

v
i
_
=
m

m+m

V
i
_
=
4 Z
2
e
4
ln
mm
V
i
V
3
=
4 Z
2
e
4
ln
mm

V
i
_
1
V
_
(372)
since V =
_
(v
i
v

i
) (v
i
v

i
) and

v
i
v
j
_
=
4 Z
2
e
4
ln
m
2
_

ij

V
i
V
j
V
2
_
=
4 Z
2
e
4
ln
m
2

2
V
V
i
V
j
(373)
These terms are written in terms of the functions
g(v) =

dv

f(v

)[v v

[ (374)
h(v) =

m+m

dv

f(v

)[v v

[
1
(375)
the sums being taken over all the perturbing species. The derivatives of g(v) and h(v) yield
g
v
i
=

dv

f(v

)
V
i
V

2
g
v
i
v
j
=

dv

f(v

)
_

ij

V
i
V
j
V
2
_
(376)
h
v
i
=

m+m

dv

f(v

)
V
i
V
3
(377)
Substituting we obtain the MacDonald, Rosenbluth and Judd form of the Fokker-Planck collision
operator.

v
i
_
=
4 Z
2
e
4
ln
m
2
h
v
i
(378)

v
i
v
j
_
=
4 Z
2
e
4
ln
m
2

2
g
v
i
v
j
(379)
In polar co-ordinates, without symmetry, a similar analysis gives basically a similar, but more com-
plicated, result.
44
16.1.1 Relationship with Landaus Formula
This result is easily related to the Landau formula for identical particles m = m

, section 15.1. Using


the same symbols as before u V , and v
1
v and v
2
v

we obtain I, the integral involved the nal


solution.

ij
=
1
u
_

ij

u
i
u
j
u
2
_
(380)

ij
u
ij
=

ij
v
ij
(381)
=

ij
v
2j
(382)
=
2u
i
u
3
(383)
Dene:
g(v
1
) =

dv
2
f (v
2
) u (384)
h(v
1
) = 2

dv
2
f
(
v
2
)u (385)
g
v
1i
=

dv
2
f (v
2
)
u
i
u
] (386)

2
g
v
1i
v
1j
=

dv
2
f (v
2
)
_

ij

u
i
u
j
u
2
_
(387)
=

dv
2
f (v
2
)
ij
(388)
h
v
1i
= 2

dv
2
f (v
2
)
u
i
u
(389)
=

dv
2
f (v
2
)

ij
v
1j
(390)
=

dv
2
f (v
2
)

ij
v
2j
(391)

3
g
v
1i
v
1j
v
1j
=

dv
2
f (v
2
)

ij
v
1j
(392)
=
h
v
1i
(393)

dv
2
_
f (v
0
)
f
v
1j
f (v
1
)
f
v
2j
_

ij
=

dv
2
_
f (v
2
)
f
v
1j
+ f (v
1
)
f
v
2j
_

ij
+2

dv
2
f (v
1
) f (v
2
)

ij
v
2j
(394)
=
f
v
1j

2
g
v
1i
v
1j
+ f (v
1
)

3
g
v
1j
v
1j
v
1j
2f (v
1
)
h
v
1j
(395)
Hence
I = 4
e
4
m
2
ln
_


v
1i
_
f (v
1
)
h
v
1i
_
. .
<vi>
+
1
2

2
v
1i
v
1j
_
f (v
1
)

2
g
v
1i
v
1j
_
. .
<vivj>
_

_
(396)
gives the dynamical friction and the diusion
Spherically symmetric distribution for f (v):
45
I = (4)
2
e
4
m
2
ln
1
v
2
d
dv
_
f +
df
dv
_
(397)
where:
=

v
0
f (v

)v
2
dv

(398)
=
1
3
_
1
v

v
0
f (v

)v
4
dv

+v
2


v
f (v

)v

dv

_
(399)
For dissimilar particles with mass m and m

, use the reduced mass:


mm

m+m

. The constant in h(v) is


changed to by
m+m

m
. i.e. replace the 2 by
m+m

m
in the front of the denition.)
17 Lecture 17: Calculation of the Transport Coecients
The transport coecients are: Electrical Conductivity, Thermal Conductivity, and Viscosity. Let us
imagine that the plasma is weakly perturbed by an external agency/force, such as a temperature gradient.
We now have the ambient thermal distribution, and the perturbation to the distribution induced by the
force. We assume that there is no net ux associated with the background (isotropic distribution). We
wish to nd the ux from the perturbation, this gives the properties of the plasma. We would like to
be able to calculate the perturbation due to a force on the distribution function, and calculate the ux
from this perturbation. We assume steady state before the perturbation:
f (v)
..
From Distribution
= f
0
(v)
. .
Initial Maxwellian
+ f
1
(v)
. .
Perturbation
(400)
Boltzmann equation:

!
0due to steady state
f
t
+v
f
r
a
0

f
v
=
f
t

c
I (401)
Apply (400), neglect terms of second order and higher To rst order:
f
0
_
mv
2
2kt

5
2
_
1
T
v
i
T
x
i
f
0
eZ
kT
E
i
v
i
= I(f
1
, f

0
) +I(f
0
, f

1
) (402)
Solve for f
1
. In general, the system has cylindrical symmetry if there is no magnetic eld present.
This means that f
1
cos , where is the poloidal angle with respect to the eld direction.
f
1
= D(v) cos (403)
where D is some function of v. For more details see Spitzer and H arm (Phys Rev 89, 977,1952), and
Cohen, Spitzer and Routly (Phys Rev 80, 230, 1950).
46

Anda mungkin juga menyukai