CHAPTER 6: Vlasov Equation Part 1

6 1 Introduction
6.1
Distribution Function, Kinetic Equation, and Kinetic Theory :
The distribution function f (x,v,t ) gives the particle density of a
certain
t i species
i in
i the
th 6-dimensional
6 di i l phase
h space off x andd v att time
ti t.
Thus, f (x,v,t )d 3 xd 3v is the total number of particles in the differen-
tial volume d 3 xd 3v at point (x,v ) and time t.
A kinetic equation describes the time evolution of f (x,v,t ). The
kinetic theories in Chs. 3-5 derive various forms of kinetic equations.
In most cases,, however,, the plasma
p behavior can be described byy an
approximate kinetic equation, called the Vlasov equation, which
simply neglects the complications caused by collisions.
By ignoring collisions,
collisions we may start out without the knowledge of
Chs. 3-5 and proceed directly to the derivation of the Vlasov equation
(also called the collisionless Boltzmann equation). 1

6.1 Introduction (continued)

The Vlasov Equation : As shown in Sec. 2.9, a collision can

result in an abrupt change of two colliding particles
particles' velocities and
their instant escape from a small element in the x-v space, which
contains the colliding particles before the collision.
However if collisions are neglected (valid on a time scale  the
However,
collision time, see Sec. 1.6), particles in an element at position A in
the x-v space will wander in continuous curves to v
position
iti B (see
( figure).
fi ) Thus,
Th theth total
t t l number
b iin the
th B
element is conserved, and f (x,v,t ) obeys an equation A
of continuity, which takes the form (see next page): x
 f (x,v,t )    [ f (x,v,t )(x , v )]  0 (1)
t x, v

In (1),  x, v [  ( x , y , z , v , v , v )] is a 6 - dimensional

x y z
divergence operator and (x , v ) [ ( x , y , z, vx , v y , vz )] can be regarded
as a 6-dimensional "velocity" vector in the x-v space.
2
 6.1 Introduction (continued)

To show that t f (x,v,t )   x, v  [ f (x,v,t )(x , v )]  0 [(1)] implies

v
conservation of particles, we integrate it over
ds6
an arbitrary volume V6 enclosed by surface S6 V6 S
6
in the x-v space: x
d
dt V6
f (x, v, t )d 3 xd 3v  V  x, v  [ f (x,v,t )(x , v )]d 3 xd 3v  0
   6

N (t )
 f (x,v,t )(x , v )ds6
s6
(by 6-dimensional divergence theorem)
 d N (t ) 
dt  s6 f (x,v,t )(x , v )  ds6  0, (2)
where N (t ) is the total number of particles in V6 , f (x,v,t )(x , v ) is the
6-dimensional "particle flux" in the x-v space, ds 6 is a 6-dimensional
differential surface area of S6 , with a direction normal to S6 pointing
outward. Thus, (2) states that the rate of increase (decrease) of the
total number of particles in V6 equals the partcle flux into (out of) V6 .
3

6.1 Introduction (continued)

Rewrite:  f (x,v,t )    [ f (x,v,t )(x , v )]  0 (1)

t x, v
(1),  x, v  [ f (x,v,t )(x , v )]
I (1)
In
 x ( fx )  y ( fy )  z ( fz )  v ( fvx )  v ( fv y )  v ( fvz )
x y z
  x  ( fx )   v  ( fv ) (6 1)
(6.1)
x  v
 f  x  v + v  x f  f  v  v + v  v f
where, because v is an independent variable, we have  x  v = 0 (3)
q
In general,  v  v does not vanish. But for v  m (E  1c v  B),
 v  E  0 [E does not depend on v]
we have 
 v  ( v  B)  B  ( v  v )  v ( v  B)  0  0  0
q
which gives  v  v = m  v  (E  1c v  B)  0 (4)
q
Thus,,  x, v  [ f (x, v, t )(x , v )]  v  x f + m (E  1c v  B)  v f
and (1) becomes  f  v  f + q (E  1 v  B)  f  0, (6.5)
t x m c v
which is the Vlasov equation. 4
 6.1 Introduction (continued)
Physical Interpretation of the Vlasov Equation :
The Valsov equation t f  v  x f + m (E  1c v  B)  v f  0 (6
q
(6.5)
5)
q
can be written
v m (E  1c v  B)
d f (x,v,t )   f  dx 
 x f  ddtv 
 v f  0,
0 (5)
dt t dt
df
where dt is the total time derivative of f . The total time derivative
(also called a convective derivative) follows the orbit
orbit* of a particle in
the x-v space. It evaluates the variation of f due to the change of the
particle position in the x-v space as well as the explicit time variation
of f . Thus,
Thus (6.5)
(6 5) or (5) can be interpreted as: Along a particle's
particle s orbit
in the x-v space, the particle density f (x,v,t ) remains unchanged.
*In the 6-dimensional x-v space, particles at a point (x,v ) have the
same velociy v. Hence
Hence, the orbit of a particle is also the orbit of an
infinistesimal element containing the particle. In the 3-dimensional
x-space, by contrast, particles at a point x have a range of velocities. 5

6.1 Introduction (continued)

This interpretation is consistent with the fact that the sum of

(3) and (4) gives  x, v  (x , v )   x  x   v  v = 00, (6)
i.e. the 6-dimensional divergence of the 6-dimensional "velocity"
(x , v ) vanishes. This implies that a collisionless v
plasma
l is
i incompressible.
i ibl Hence,
H the
th volume
l off
B
an elements (thus the particle density f ) will be A
unchanged as it moves from A to B. x
A specific example :
Consider the simple case of a group of particles initially located in
a square
q in the x-vx space
p (area
( A, lower figure).
g ) If the pparticles are
force free, those on the upper edge will move at the fastest (equal and
constant) speed, while those on the lower edge vx
move at the slowest (equal and constant) speed. speed B
Then, some time later, the square will become a A
parallelogram of the same area (area B, lower figure). x 6
 6.1 Introduction (continued)

Effect of collisions :
If there are collisions, they will cause a variation of f at the
symbolic rate of ( t f )coll , which should be added to (6.5) to
give
 f  v  f  q (E  1 v  B)   f  (  f ) ,
t x m c v t coll
while the specfic
p form of ( t f )coll depends
p on interparticle
p forces.
Throughout this course, the ( t f )coll term will be neglected.

6.1 Introduction (continued)

Complete Set of Equations : We now have the following set

of self
self-consistent,
consistent coupled particle and field equations:
 q
 t f  v f  m (E  c v  B)  v f  0
1 (6.5)
  B  0 (7)
 For simplicity, we shall

  E  4 henthforth denote  x by . (8)
 1 
  E   c t B (9)
 1  4
  B  c t E  c J (10)
  (x, t )   q  f (x, v, t )d 3v (11)
 
where 
J (x, t )  

q  f (x, v, t ) vd 3v (12)

Each particle species, denoted by the subscript " ", is governed

by a separate Valsov equation, and q carries the sign of the charge.
8
 6.2 Equilibrium Solutions
General Form of Equilibrium Solutions : As shown in (5),
the Vlasov equation can be written as a total time derivative:
d f (x,v,t )   f  dx f  dv  f  0, (5)
dt t dt dt v

where dtd follows the orbit of a particle whose position and velocity
at time t is x and v, respectively. Thus, any fuction of constants of the
motion along the orbit of the particle, Ci  Ci ( x, v,t ), is a solution of
d f (C , C ...)   f dCi  0,
the Vlasov equation, i.e. dt 1 2 C dt i i
because,, by
y the definition of constant of the motion,,
dCi
dt
 t Ci  ddtx Ci  ddtv  v Ci  0.
The equilibrium
q solution ((denoted by y subscript p "0")) of interest to
us is a steady-state solution formed of constants of the motion that do
not depend explicitly on t , i.e. f 0  f 0 (C1 , C2 ...) with Ci  Ci (x, v ). 9

6.2 Equilibrium Solutions (continued)

Examples of Constants of the Motion :
1 If B 0  E0  00, vx , v y , and vz are constants of the motion.
1. motion
2. If E0  0, B 0  B0e z  const., v , vz are constants of the motion.
3. The motion of a charged particle (mass m and charge q ) in EM
fi ld (represented
fields ( t d by t ti l A andd  ) is
b potentials i governedd by b Lagrange's
L '
equation: [Goldstein, Poole, & Safko, "Classical Mechanics," 3rd ed., p. 21]
d L  L , i  1, 2, 3 q
(13)  m dv  qE  c v  B (13)
dt qi qi dt
See Goldstein, Poole, &
qi is a position coordinate. Safko, Sec. 1.5.
where  q
 L  2 mv  c v  A  q [L: Lagrangian; v: particle velocity]
1 2

In cylindrical coordinates, we have qi  (r ,  , z ), qi  (r, , z ),

v 2  r 2  r 2 2  z 2 , and v  A  rA
 r  r A  zA
 z . If the fields A and 
are independent
i d d off  , (13) gives i d L  L  0. 0 Hence,
H
dt  
L  mrv  q rA  const [canonical angular momentum] (14)
 c 
 10
 6.2 Equilibrium Solutions (continued)
Examples of Equilibrium Solutions:
In contrast to the Boltzmann equation,
equation which has only one
equilibrium solution (the Maxwellian distribution), the Vlasov
equation has an infinite number of possible equilibrium solutions.
But they exist on a time scale short compared with the collision time.
time
The choice of the equilibrium solution depends on how the
plasma is formed. For example, if we inject two counter streaming
electrons
l l i v0 ez andd v0 ez into
off velocity i a neutralizing
li i background
b k d
of cold ions, we have the following equilibrium solutions for the
electrons and ions (of equal density n0):
 f e 0  1 n0 (vx ) (v y )[ (vz  v0 )   (vz  v0 )]
2
 (15)
 i 0
f  n0 ( v x ) ( v y ) ( v z )
which correctly represent the electron/ion distributions on a time
scale short compared with the collision time.
11

6.2 Equilibrium Solutions (continued)

Given sufficient time, collisions will first randomize electron
velocities and eventually equalize electron and ion temperatures.
temperatures
The fianl state will be an equilibrium solution in the form of the
Maxwellian distribution for both the electrons and ions:
2 n0 2
f 0 (v)  n0 ( m )3/2 exp( mv )  exp( v 2 ) (16)
2 kT 2kT (2 ) vT
3/2 3 2vT
where vT  kT / m is the thermal speed, T is the same for both
i , m  me for
species f the
th electrons
l t andd m  mi for
f the
th ions.
i I (16)
In (16),
f 0 has been normalized to give a uniform particle density of n0
in x-space [  f 0 (v)d 3v  n0 ].
 5(2n1) 
  x 2ne ax2 dx  13
 0 2n1 an a
Useful formulae:  ( a  0) (17)
 2 n 1  ax2
 0 x e dx  n1 n!
 2a
 0

e ax dx  12
2
; 
x 2e ax dx  41a a ;
2 
x 4e ax dx  32 a (18)12
2
a 0 0 8a
 6.2 Equilibrium Solutions (continued)
Discussion: In constructing the equilibrium solution f 0 , we must
also consider the self self-consistency
consistency of the solution. For example, using
(11) and (12), we find that both (15) and (16) give 0  e 0  i 0  0
and J 0 = J e 0  J i 0  0. Hence, the plasma produces no self fields and
the assumption of E0  B 0  0 (which makes vx , v y , and vz constants
of the motion) is valid. However, if f e 0  12 n0 (vx ) (v y )[ (vz  v0 ) 
 (vz  v0 )] in (15) is replace by f e0  n0 (vx ) (v y ) (vz  v0 ), then
J ez  0 and there will be a self magnetic field, in which vx , v y , vz are
no longer constants of the motion. As a result, f e 0 (vx , v y ,vz ) does not
satisfy the Vlasov equation
equation. In other words
words, the complete equilibrium
solution includes not just f 0 , but also the self-consistent fields.
When the air in equilibrium [(16)] is disturbed, sound waves will be
generated When a plasma in equilibrium is disturbed
generated. disturbed, a great variety
of waves may be generated. Some may even grow exponentially. These
are subjects of primary interest in plasma studies. 13

6.3 Electrostatic Waves

Rewrite the Vlasov-Maxwell equations:
 q
 t f  v f  m (E  c v  B)  v f  0
1 (6.5)
  B  0 (7)


  E  4 (8)
 1 
  E   c t B (9)
 1  4
  B  c t E  c J (10)
  (x, t )   q  f (x, v, t )d 3v (11)
 
where 
J (x, t )  
q  f (x, v, t ) vd 3v (12)

Below we present a kinetic treatment

Below, treatment of the problem in Sec.
Sec 1.4.
14
The electrons now have a velocity spread and, as a result, we will find
that the electrostatic plasma oscillation becomes an electrostatic wave.14
 6.3 Electrostatic Waves (continued)
At high frequencies (e.g.  ~  pe ), the ions cannot respond fast
enough to play a significant role. So we assume that the ions form a
stationary background of uniform density n0 . Since we consider only
electron dynamics, the species subscript " " in f will be dropped.
Equilibrium (Zero-Order) Solution :
Assume there is no field (including external field) at equilibrium.
then, v x , v y , and v z are constants of the motion and any fuction of v,
f 0  f 0 ( v )  f 0 ( v x , v y , v z )), is an equilibrium
eq ilibri m solution
sol tion for the electrons,
electrons
i.e.  f ( v )  v  f ( v )  e ( E  1 v  B )   f ( v )  0

t 

0
 
0 me 0 c 0 v 0
0 0 0
0
provided f 0 ( v ), which represents a uniform distribution in real space,
is normalized to the ion density n0 and it gives rise to zero current,
  f 0 ( v )d v  n0
3 For clarity, all equilibrium quantities
i
i.e.  are denoted by subscript "0". They (19)
  0
f ( v ) v d 3
v  0 are treated as zero-order quantities.
so that there are no fields at equilibrium (E0  B0  0). 15

6.3 Electrostatic Waves (continued)

First-Order Solution (Linear Theory) of Electrostatic Waves
by the Normal-Mode Method :
Consider small deviations from the equilibrium solution in (19)
[f  f 0 ( v ), E0  B0  0] and specialize to waves without a magnetic
(thus,   E =  1c t B  0  E =   )). W
fi ld (th
field We may th
then write
it
 f ( x, v, t )  f 0 ( v )  f1 ( x, v, t ) For clarity, we denote all
 small quantities by subscript
 ( x, t )  1 ( x, t ) "1"
1 . They are treated as
(20)
E( x, t )  E1 ( x, t )  1 ( x, t ) first-order quantities.
Sub. (20) into the Vlasov equation:
0
 f  v  f  e ( E  1 v  B)   f  0,
0 (6.5)
(6 5)
t me c v
we find the zero-order terms vanish. Equating the first-order terms,
we obtain  f  v  f   e   f (21)
t 1 1 me 1 v 0

Sub. E1  1 and 1   e  f1d 3v into the first-order field

equation,   E1  41 , we obtain 21  4 e  f1d 3v (22)16
 6.3 Electrostatic Waves (continued)

  f1  v f1   me  1 v f 0 (21)

Rewrite:  t2 e

 1  4 e  f1d v
3
(22)
Consider a normal mode (denoted by subscript "k" ) by letting
 f1 (x, v, t )  f1k ( v )eikz z it [f1k ( v ) : a small function of v ] (23)
 ik z it
1 (x, t )  1k e z [1k : a small constant] (24)
where it is understood that the LHS is given by the real part of the RHS.
The normal-mode analysis is general because a complete solution
can be expressed as a superposition of any number of normal modes.
f 0 ( v )
((21),
), ((23),
3), aandd (24) g ve (i  ik z vz ) f1k ( v )   mee ik z1k
( ) give
vz
f ( v )
 f1k ( v )  mee  kkz v 1k 0v (25)
z z z
f 0 ( v )
4 n e2 vz
(22)-(25) give  k z21k 
me
0
n0 k z1k
1
  kz vz
d 3v (26)
 2pe 17

6.3 Electrostatic Waves (continued)

f 0 ( v )
vz
R it (26) :  k z21k   2pe n10 k z1k 
Rewrite  k z v z d v
3
(26)
f 0 ( v )
2pe  vz
For 1k  0,, we must have 1 n0  vz   d v
1 3
((27))
kz2 kz
The v x and v y integrations in (27) may be immediately carried out
to result in a one-dimensional distribution function g 0 ( v z ) [which,
[ , byy
(19), is normalized to 1]: g ( v )  1 f ( v )dv dv
n0  0
0 z x y (28)
dg0 ( v )
 2pe  dvz
Then, (27) becomes 1  2  v   dv z  0 
dispersion 
(29)
kz z k  relation 
z
(29) has a singularity at v z   / k z , which will be addressed later.
For now, we circumvent this difficulty by assuming v z   / k z for the
majority of electrons so that g0 (v z ) is negligibly small at v z   / k z .
18
 6.3 Electrostatic Waves (continued)

Since g0 ( v z  )  0, integrating (29) by parts gives

2  g0 ( vz ) 2pe  g0 ( vz )
kz2  ( vz   )2 z 2  (1 k
1  pe dv  1  dv z  0 (6.25)
zvz )2
k z
Note: The z -direction here is the x -direction in Nicholson. g0 , k z ,
and v z are, respectively, g , k , and u in Nicholson.
Expanding (1  kzvz ) 2 and keeping terms up to second order in

kzvz , we obtain 1  2pe  g ( v )[1  2 kzvz  3( kz vz )2 ]dv  0 (6.26)
  2  0 z   z

A specific example : If the equilibrium soultion is Maxwellian:

n0 vx2 v2y vz2
f0 ( v)  exp(  ) (6.23)
(2 )3/2 vTe
3 2
2vTe
  ax
dx  12 a , we obtain
2
then, using the formula in (18): 0 e
2
g0 ( v z )  n1  f 0 ( v )dv x dv y  1 exp(  vz2 ) (6.24), (30)
0 2 vTe 2vTe 19

6.3 Electrostatic Waves (continued)

With g0 ( v z )  1 exp(  vz2 ) [(30)] and the formulae:

2 vTe 2
2vTe
 
 ax
dx  12 a and x 2 e  ax dx  41a a [(18)], we obtain
2 2
0 e 0
  
 g0 (vz )dv z  1;;  v 0 ( v z ) dv z  0;  v z g 0 ( v z ) dv z  vTe
2 2
z g 
(31)
( )
an odd function of v z
2pe  k z vz k z vz 2
Sub. ((31)) into 1 
 
2  g 0 ( v z )[1  2   3(
(  ) ]dv z  0 [(6.26)],
[( )],
2 k 2v 2  2
we obtain 1  pe2  3 z Te4 pe  0, (6.27)
 
B assumption
By (30)   k z vTe . Thus,
ti (30), Th to t lowest
l t order,
d (6.27)
(6 27) gives
i
   pe and, to next order, we obtain the dispersion relation for the
Langmuir
g wave : (See
( Sec. 7.3 for a fluid treatment))
 2   2pe  3k z2 vTe
2
[     pe (1  23 k z2 vTe
2
)] (6.28)
20
 6.4 Landau Contour
We now address the singularity encountered in (29). First, a review
off relevant
l t definitions
d fi iti andd theorems
th involving
i l i complex l variables.
i bl
(Reference: Mathews and Walker, "Math. Methods of Phys.," 2nd ed.)
Laplace Transform : (M&W, Sec. 4.3)
 L[ (t )]     (t )e pt dt   ( p ) A tilde "  " on top of a symbol
 0
indicates a p-space quantity.
 
initial 
value

 L[ (t )]  p ( p )   (t  0)) ((32))

  
initial  
value 
initial 
value
 L[ (t )]  p 2 ( p )  p  (t  0)   (t  0)
 Im( p)
Inverse Laplace transform:
p i
 (t )  21 i  p 0i  ( p )e pt dp poles
(33)
0 of  ( p)
Note: p0 ( 0) is sufficiently large so that all Re( p)
p path of
the poles of  ( p ) lie to the left of the path of 0
p-integration
p -integration. Hence,  (t )  0 if t  0. (why?) 21

6.4 Landau Contour (continued)

Analytic Function : (M&W, Appendix A)

A function f ( z ) in the complex z -plane (z  x  iy  rei ) is said
to be analytic at a point z if it has a derivative there and the derivative
f ( z  h) f ( z ) y
f ( z )  lim [h: a complex number]
h 0 h z
r
is independent of the path by which h approaches 0.  x
0
The necessary and sufficient conditions for a function
W ( z )  U ( x, y )  iV ( x, y )
to be analytic are: U  V and V   U (34)
x y x y
Examples: W  z 2 , W  z , and W  e z are all analytic functions.
W  z * (z*: complex conjugate of z ) is not an analytic function.

22
 6.4 Landau Contour (continued)

Single-Valued Function : (M&W, Appendix A)

A function W ( z ) in the complex z -plane
plane is single
single-valued
valued if
W ( z  rei )  W ( z  rei (  2 n ) ) [n  1, 2,] y
W  z 2 iss a ssingle-valued
g e v ued function.
u c o . r
z
E l : 
Examples 
W  z is not a single-valued function. 0
1/ 2 x
For W  z1/ 2 , we may draw a branch cut from y
z  0 to
t  in
i the
th z -plane
l andd forbid
f bid z to
t cross it.
it z
r
Then, W  z is single-valued in the z -plane
1/ 2  x
0
where  is restricted to the range 0    2 , i.e. branch cut
the  value would have a 2 jump if z were to cross the branch cut.
Regular Function : (M&W, Appendix A)
A function is said to be regularg in a region
g R if it is both analytic
y
and single-valued in R. Thus, W  z 2 is regular in the z -plane with an
arbitray  and W  z1/ 2 is regular in the z -plane with a branch cut. 23

6.4 Landau Contour (continued)

Cauchy' Theorem : (M&W, Appendix A)

If a function f ( z ) is regular in a region R, then  c f ( z )dz  0,0
where C is any closed path lying within R. Hence, the line integral
z2 y
z1 f ( z ) dz is independent of the path of integration z2

from z1 to z2 if the path lies within region R. z1 

x
Theorem of Residues : (M&W, Sec. 3.3 & Appendix A)
If f ( z ) is regular in a region R, except for a finite number of poles,
es
then,  c f ( z )dz  2 i  residues inside C , (35)
where C is any closed path (in the counterclockwise direction) within R
and of order n at z  z   1 {( d )n1[( z  z0 ) n f ( z )]}z  z0 (36)
the residue of a pole
 0  (n1)! dz
y
Example:  z  z dz  2 ig ( z0 ) 
g ( z) g ( z ) is regular 
z
c 0  with
i h no poles. l  0
x
b g ( x)  x0  g ( x )
Principal Value : P a x x dx  lim a x x dx  x  x x dx (37)
b g ( x) 
0    0 0 0  24
 6.4 Landau Contour (continued)

Identity Theorem : (M&W, Appendix A)

If two functions are each regular in a region R, and having the
same values for all points within some subregion or for all points
along an arc of some curve within R, then the two functions are
identical everywhere in R. For example plane e z is
example, in the the z -plane,
the unique function in the z -plane which equals e x on the x-axis.
Analytic Continuation : (M&W, Appendix A) R1
If f1 ( z ) andd f 2 ( z ) are analytic
l ti in
i regions
i R1 andd R2 , f1(z)
respectively, and f1  f 2 in a common region (or line), f1  f2
then f 2 ( z ) is the analytic continuation of f1 ( z ) into R2 . f2(z)
By identity theorem, it is the unique analytic continuation. R2
Example 1: f1  1  z  z 2  z 3   is analytic in the region z  1.
f 2  1/(1  z ) is analytic everywhere except at the pole z  1. Since
f1  f 2 in the common region z  1, f 2 is the unique analytic
continuation of f1 into the z  1 region (except for the pole at z  1). 25

6.4 Landau Contour (continued)

Example 2: Consider the following 2 analytic functions of p:

 g ( vz ) analytic
l ti in th upper  vz -plane
i the
f1 ( p )   dv z ip / k z (38)
 vz ip / k z  half plane, Re( p )  0
(k z is real and positive.) ppath of
integration
f 2 ( p )   dv z
g ( vz ) analytic in the  (39)
L vz ip / k z entire p -plane  vz -plane
ip / k z

 g ( vz )
L d
Landau
contour   
dv z
vz ip / k z
, Re( p )  0
 
 g ( vz ) ip
  P  dv z   ig
g ( ), Re(( p )  0 ((40))
 vz ip / k z kz
definition  
of Landau   dv z g ( vz )  2 ig ( ip ), Re( p )  0
contour   vz ip / k z kz Landau contour
Since f1 ( p )  f 2 ( p ) in the upper half plane, f 2 ( p ) is the (unique)
analytical continuation of f1 ( p ) into the lower half plane. 26
 6.4 Landau Contour (continued)

Electrostatic Waves by the Method of Laplace Transform :

(Ref Krall & Trivelpiece,
(Ref. Trivelpiece Secs.
Secs 8.3
8 3 and 8.4)
8 4)
  f1  v  f1   mee  1  v f 0 (21)
Return to (21) and (22):  t2
 1  4 e  f1d v
3
((22))
Direct substitution of the nomal mode [(23), (24)] into (21) and
(22) results in a singularity in (29). Landau resolved this problem by
treating (21) and (22) as an initial value problem in t , while analyzing
a spatial Fourier component in z (denoted by subscript "k").
 f1 ( x, v, t )  f1k ( v, t )eikz z  By assumption, the wave  (41)
Let   has no x, y -variation.
 variation 

1 ( x, t )  1k (t )e
ikk z z
(42)
Sub. (41), (42) into (21), (22), we obtain
  f ( v, t )  ik v f ( v, t )   e ik  (t ) f 0 ( v ) (43)
 t 1k z z 1k me z 1k vz
 2
  k z 1k (t )  4 e  f1k ( v, t )d v
3
(44)27

6.4 Landau Contour (continued)

  f ( v, t )  ik v f ( v, t )   e ik  (t ) f 0 ( v )
 1k (43)
Rewrite  t vz
z z 1k me z 1k
k z 1k (t )  4 e  f1k ( v, t )d v
2 3
(44)
Perform a Laplace transform on (43) and (44) [see (32)], we obtain
 pff ( v, p )  f ( v, t  0)  ik v f ( v, p )   e ik  ( p ) f 0 ( v ) (45)
 1k 1k z z 1k me z 1k vz
 2
k z 1k ( p )  4 e  f1k ( v, p )d v
  3
(46)
 f ( v, p )   f ( v, t )e pt dt
where 
 1k 0 1k (47)
  pt
1k ( p )  0 1k (t )e dt (48)
f ( v )
f1k ( v,t 0) mee ikz 0 1k ( p )
(45)  f1k ( v, p )  vz (49)
pikz vz
A note on notations: Subscripts "0" 0 and "1" 1 indicate,
indicate respectively,
respectively
zero-order and first-order quantities. Subscript "k" indicates a Fourier
component in z. Symbols with a "  " sign on top are p-space quantities.28
 6.4 Landau Contour (continued)
Sub. (49) into (46), we obtain
 g1k (vz ,t 0)
 f ( v,t 0) 3 i 4 n3 0e  dvz
4 e  1k d v kz  vz  ip
i
 pikz vz
1k ( p )   kz
(50)
ikz f 0 ( v ) dg0 (vz )
 2pe  n0 vz  2pe 
 kz2  vz  ip
dvz
k z2[1  d 3v] 1 i dvz
kz2 p ikz vz
kz
 g1k (vz , t  0)  n1  f1k ( v, t  0)dvx dv y Im( p )
where  0
(51)
 1

 0 z
g ( v ) n0 0 f ( v )dv x dv y poles
l
of  ( p)
Inverse Laplace transform: 1k
p0 i Re( p)
By (33), 1k (t )  1  1k ( p )e pt dp, of (52)
2 i p0 i p
0 path
p
p-integration
where p0 ( 0) is a real number. In (50), k z is
real by assumption. If k z  0, we see from (50) Im(vz )
that the pole (ip / k z ) of the vz -integrals
integrals lies ip / k z
above the Re(vz ) axis. Hence, 1k ( p ) & 1k (t ) Re(vz )
path of vz - integration 29
are valid solutions without any singularity.

6.4 Landau Contour (continued)

p0 i
Rewrite (52): 1k (t )  1   ( p)e dp, pt
(52)
2 i p0 i 1k
 g (vz ,t 0)
i 4 n3 0e  1k dvz Im( p)
kz  vz  ip
where 1k ( p )  kz poles (50)
dg0 (vz ) of 1k ( p)
 2pe 

dvz
1 dvz Re( p)
kz2 vz  ip p path of
kz 0
p-integration
The mathematical solution (52) is in a form
in which the physics (e.g. normal modes and Im(vz )
dispersion relation, etc.) is not transparent. We ip / k z
Re(vz )
need more work to obtain a physics solution.
solution
path of vz - integration
This will require a detour of the path of the
p -integration into the Re( p)  0 region, which implies that the pole at
i / k z will
ip ill cross the Re(vz ) axis, hich 1k ( p ) is sing
a is on which singular.
lar Th
Thus, s the

analytic region of 1k ( p ) is bounded by the Re(p )  0 line, and our first
step is to analytically continue 1k ( p ) from Re(p)  0 into Re( p)  0. 30
 6.4 Landau Contour (continued)

Step 1 : Analytic continuation : By the method in (38)-(40), we

may analytically continue 1k ( p ) from Re(p )  0 into Re(p )  0 by
by changing the path of vz -integration in (50) from the straight line

 dvz to the Landau contour: L dvz . Im(vz )
ip / k z
g1k (vz ,t 0)
i 4 n3 0e L vz  ip dvz
Re(vz )
kz original path of
Thus, 1k ( p )  kz
vz - integration (53)
dg
g0 (vz )
 2pe
L
dvz
1 dvz
kz2 vz  ip vz -plane
kz ip / k z
Note that this path change will not affect the same as
p0 i original path
value of 1k (t )  1   ( p )e pt dp in (52)
2 i p0 i 1k
because 1k ( p ) in (52) is evaluated along the
Re(p )  p0 ( 0) line, for which the Landau path
is the same as the original vz -path (see figures). Landau contour 31

6.4 Landau Contour (continued)

S ( p ,k z )
Write (53) as 1k ( p ) 
vz -plane
, (54)
D ( p ,k z ) ip
p / kz
 4 n0e g1k (vz ,t 0)
kz3 L vz  ip
 S ( p , k z )  i dvz
 implying kz
where  "source"
source dg0 (vz )
 D( p, k )  1   2pe
kz2 L vz  ip
dvz
dvz (55)
 z
 kz Landau contour
p0  i S ( p,k ) pt
2 i  p0 i D ( p,kz )
then, (52)  1k (t )  1 z
e dp (56)
Im( p)
Since 1k ( p ) is now regular (analytic and
poles
single valued) in the entire p-plane
single-valued) plane (except of 1k ( p)
at poles), we are free to deform the path of Re( p)
p-integration in (56) (by Cauchy's theorem), p
0 path of
provided
id d the
th new path th does
d nott cross any pole. l p integration
-integration
Note: We now have 2 complex planes: p -plane and vz -plane, and
there are integrals along complex paths in both planes. 32
 6.4 Landau Contour (continued)

Step 2 : Deformation of the p -contour in (56): Im( p)

p0 i S ( p,k ) ptt
1k (t )  21 i  z
e dp poles (56)
p0 i D ( p ,k z )
of 1k ( p)
To bring out the physics in (56), we deform Re( p)
the p -contour
the contour as shown to the right. right Cauchy
Cauchy'ss p
0 original
p-contour
theorem requires the new path to encircle
(rather than cross) the poles it encounters. deformed p-contour, Im( p) p  i
A
Assume th
thatt all
ll the
th poles l off 1k ( p ) radius   0

[ S ( p, k z ) / D ( p, k z )] are at the (1st-order)

growing
roots p j (j  1,2, ) of D ( p, k z )  0. Then, mode
Re( p)
1k (t )   ( p  p j ) D ( p,kz )  e j 
S ( p,k ) p t residues 
damped integrand
j  z   p  p  at poles  mode p t ip t
j e r i
 transient effects [integrations away from poles] (57)
p0  i
Path integrations away from the poles result in
1k (t )  0 as t   because e pt oscillates rapidly with pi [  Im( p )]. 33

6.4 Landau Contour (continued)

dg0 (vz )
 2pe
L
dvz
Normal modes: Rewrite: 1 dvz  0 (55)
kz2 vz  ip
i
kz
Define   ip so that 1k (t )  e it
and 1 ( x, t )  eit ikz z (58)
dg0 (vz )
 2pe dispersion 
L
dvz
then, (55) can be written 1  vz   dv z  0  relation  (59)
kz2 k z

(corresponding Landau contour: vz vz vz )

i  0 i  0 i  0
and (57) can be written as a sum of normal modes
S ( ,k )  i t
1k (t )   1i (   j ) D ( ,kz )  e j   transient  (60)
j  z    effects 
j
S ( ,kz )  i t ik z z  transient 
 1 (x, t )   1i (   j ) e j  (61)
j  D ( ,kz )   effects 
j
where frequencies  j (  ip j , j  1, 2, ) of the normal modes (in
gereral complex numbers) for a given k z can be found from (59). 34
 6.4 Landau Contour (continued)

Summary of techniques and theorems used :

A Laplace transform brings us into the complex variable territory.
Cauchy's theorem is then used to deform the p-contour (left figure).
This requires the analytic continuation of 1k ( p) to the entire p-plane,
which in turn leads to the Landau contour for the vz -integration
integration
(right figure). With the deformed p -contour, we are able to apply the
residue theorem to extract the essential physics by isolating the normal
modes
d from
f th
the (non-essential)
( ti l) transient
t i t effects.
ff t
deformed Im( p ) vz -plane
p -contour, p0  i ip / k z
radius  

growing
mode
Re( p )
damped integrand
mode p t ip t
e r i
Landau contour
p0  i 35

6.4 Landau Contour (continued)

A Recipe for Handling Singularities in Normal -Mode Method :

Rewrite the solution [(29)] obtained by the normal-mode method
in Sec. 6.3 and the new solution [(59) and (61)] obtained by the
Laplace transform method in this section.
dg0 (vz )
 2pe  By comparison, we find that

dvz
1 dv  0 (29)
kz2  vz   z the Laplace-transform method
kz
gives the additional information
 dg0 (vz ) off mode
d amplitude
li d in
i terms off
1   pe
2

kz2 L vz  k
dvz
  dvz  0 the initial perturbation [see (61)]. (59)
 z
 1 (   ) S ( ,kz )  i j t ik z z  transient
t i t
 ( x , t )   D ( ,kz )  
e 
effects 
(61)
j 
1 i j
 j

We find that ((29)) and ((59)) have the same form except p for the path
p
of the vz -integration. This provides a simple recipe for removing the

singularity in (29): replacing  dvz with the Landau contour: L dvz . 36
 6.4 Landau Contour (continued)
The above recipe is of general applicability; namely, we may
solve
so ve a variety
va ety of
o problems
p ob e s by the
t e (simpler)
(s p e ) normal-mode
o a ode method,et od,
then remove similar singularities in the solutions by replacing the

 dvz contour with the Laudau contour L dvz .
A question may arise as to whether the pole should remain above
or below the Landau contour. As just shown, this depends on whether
ip
the original position of the pole k is above or below the Re(v z )-axis
z
 g ( vz ,t 0)
i 4kn30e  1k dv z
z  vz  kip
in (50): 1k ( p )  z
dg0 ( vz )
(50)
2pe 

dvz
1 dv z
kz2 vz  kip
z
p0 i
b f
before we deform
d f the
h p -contour in (52) 1k (t )  21 i 
i (52):  ( p )e pt dp
d
p0 i 1k
ip
i.e. we determine the original position of k by setting Re( p )  p0  0.37
z

6.4 Landau Contour (continued)

 (t )  1 p0 i  ( p )e pt dp,
2 i  p0 i 1k
(52)
 1k
  g (vz ,t 0)
 i 4 n3 0e  1k dvz
Rewrite  k z  vz  ip

1k ( p ) 
kz
((50))
dg0 (vz )
  2pe  dvz
kz2  vz  ip
 1  dvz
 kz
where
h R ( p )  p0  00. W
Re( We see, if k z  0, 0 the
th original
i i l position
iti off the
th
ip
pole k lies above the path of vz -integration, and should remain so
z
when the p-contour in (52) is deformed and the vz -contour in (50) is
changed to the Landau contour. If k z  0, then the original position
ip
of the pole k lies below, and should remain below, the vz -contour.
z
If we convert the variable p to a new variable   ip so that
the solution has the form of a normal mode: exp(  it  ik z z ), the
original position of the pole can be similarly determined as follows: 38
 6.4 Landau Contour (continued)

Recipe for Landau path in normal -mode analysis:

k z  0 (poles remain above vz -contour) :
 vz vz (62)
 vz
it ikz z i  0   0 i  0
e dependence:  i

k z  0 (poles remain below vz -contour) :


 vz vz vz (63)
i  0 i  0 i  0

k z  0 (poles remain below vz -contour) :


vz v z (64)
 vz
it ikz z i  0   0   0
e dependence:  i i

k z  0 (poles remain above vz -contour) :


 vz vz vz (65)
i  0 i  0 i  0
39

6.5 Landau Damping

We have shown that a Laplace transform elegantly resolves the
singularty problem in normal-mode analysis. The recipe is in the
form of the Landau contour for the v z -integration [(62)-(65)].
Waves considered so far are electrostatic in nature (i.e. without
a B-field component). The Langmuir wave derived in Sec. 6.3 is
only one example of such waves. In this section, we reconsider the
Langmuir
g wave byy properly
p p y accountingg for the singularity
g y through
g
the use of the Landau contour. This leads to a very important new
phenomenon known as Landau damping.
We will then go beyond the scope implied by the section title
with an examination of two types of electostatic instabilities, the
Landau growth and the two-stream instability, which occur in a
plasma with a non-Maxwellian
non Maxwellian electron distribution.
distribution We will also
consider a different type of (low-frequency) electrostatic wave, the
ion sound wave, which involves both the electrons and ions. 40
 6.5 Landau Damping (continued)
Landau Damping in a Plasma with a Maxwellian g0 ( v z ) :
2pe dg0 ( vz )
Rewrite (59): 1
k2z L vz 1 kz dvz
dv z  0, (59)

which was derived under the e it ikz z dependence. Thus, for k z  0,
the v z -contour is [see (62)] vz vz vz
i  0 i  0 i  0
If i  0, we may take the path vz and write (see
Nicholson, pp. 279-284)
dg0 ( vz )  dg0 ( vz ) dg0 ( vz )
 1
L vz  
kz
dvz
dv z  P  vz 1 kz dvz
dv z   i dv
z v  / k
z z
(66)

If  / k z  vTe , we may expand the principal g 0 (v z )


dg0 ( vz ) kz
value term to obtain L vz 1 kz dvz
dv z vT e
vz
 dg
  kz  [1    (  )  (  )  ]dv z   i dvz
k z vz z z 2 kvz z 3 kv 0
(67)
vz  / k z 41

6.5 Landau Damping (continued)

dg0 ( vz )
Rewrite (67): L vz 1 kz dvz
dv z
 dg dg
  kz  dvz0 [1    (  )  (  )  ]dv z   i dvz
k z vz z z 2 kvz z 3 kv 0
(68)
vz  k
z
 g ( v )  1 exp(  vz2 ) [see (30)]
 0 z 2 vTe 2
2vTe
For a Maxwellian g 0 : 
dg0 ( vz )

vz exp(  vz2 )
 dv (69)
 z 2 vTe
3 2
2vTe
 
we use 0 x 2 e  ax dx  41a a and 0 x 4 e  ax dx  8a32 a [(18)] to
2 2

dg0 ( vz ) dg
L vz 1 kz d z  ( kz )2  3( kz )4 vTe   i dv 0
2
obtain
b i dvz
dv (70)
z v 
z k
z
Sub. (70) into (59) and using (69), we obtain
 2
2 k 2v 2   2 kz2vTe2 2
1 (1  3 z 2Te )  i 2
pe
e 0 pe
(71)
2  z Te k v3
3
42
 6.5 Landau Damping (continued)
 2
2 kz2vT2e  2  2 k 2v 2
Rewrite (71): 1  pe (1  3 )  i 2 3pe3 e z Te  0 (71)
 2
 2
kz vTe
1



1
  
Let   r  ii and i  1  12  12 (1  i i ) 2  12 (1  2i i )
r  r r  r r
Keeping terms up to first order, (71) gives
r2
 2 2 2  
2  r 2 kz2vTe
2
1  pe2 (1  3 z 2Te )  2i pe2 i  i 2 pe
k v 2
e 0 (72)
r r r r kz3vTe
3

We may solve (72) by the method of iteration. To the lowest order,

(72) gives r   pe . To first order, the real part of (72) gives
k 2v 2 k 2v 2
r2   2pe (1  3 z 2Te )  r   pe (1  23 z 2Te ) (73)
pe pe
and the imaginary part of (72) gives
 2pe
 3
 4
i   8
2 k z2vTe
2 2
pe
e [k z  0] (74)
k 3v3
z Te 43

6.5 Landau Damping (continued)

The electron Debye length (De ) can be written

kTe kTe me vT2e
De
2

4 n0e2
 
me 4 n e2  2pe (75)
0
Thus, (73) and (74) can be written
 2 2
3 k z vTe )   (1  3 k 2  2 )

 r
  pe (1  2  2pe pe 2 z De

  2pe 3 (76)
   21 2  3
   pe e 2 k z vTe     pe e 2 k z De
4 2 2 2 2
i   8 kz3vTe 3 8 k z3D3 e
where r and i agree with (6.52) and (6.53) in Nicholson.
D
Discussion :
(i) (76) is derived under the e it ik z z dependence with k z  0.
 t
Hence,, i is a negative
g number and   e i , which implies p
that the wave is damped even though the plasma is assumed to be
collisionless. This is known as Landau damping.
44
 6.5 Landau Damping (continued)
Rewrite (76):
 2 2
3 k z vTe )   (1  3 k 2  2 )

 r   pe (1  2  2pe pe 2 z De

  2 (76)
 2pe2  3  21 2  3
   pe e 2 k z vTe     pe e 2 k z De
4 2 2
i   8 kz3vTe 3 8 k z3D3 e
(ii) In the limit of Te  0, we have r   pe and i  0. Thus,
((76)) reduce to the (undamped)
( p ) pplasma oscillation of a cold pplasma.
discussed in Sec. 1.4. This shows that the plasma temperature is
responsible for both the Landau damping and the change from an
oscillation phenomenon to a wave phenomenon.phenomenon
(iii) Mathematically, contribution to i comes g 0 (v z )
from the residue of the pole at v z   / k z in the 
v z -integration.
integration Physically,
Physically this implies that k z
vTe
Landau damping is due to resonant electrons vz
moving at the phase velocity ( / k z ) of the wave. 45

6.5 Landau Damping (continued)

Question : How will the result be changed if k z  0, while still

assuming the e it ikz z dependence?
Answer :
In this case, from (63), we use the contour v z instead of
vz . Thus, the only change is to replace "+ i" in (66):
dg0 ( vz )  dg0 ( vz ) dg0 ( vz )
L vz 1 kz dvz
dv z  P  vz 1 kz dvz
dv z   i dvz v  / k
z z
with "   i". This will result in the same expression for r , but with
a sign change in i i.e. g 0 (v z )
 2pe
 3  1 3 
4  2 k z2De
i  8 3 pe3 e  8 3 pe3 e
2 2
2 k z2vTe 2 2
kz v
kz vTe kz De Te
vz
Since k z  0, i is still a negative number and the wave will be
damped at the same rate as is expected from symmetry considerations.
46
 6.5 Landau Damping (continued)

Landau Growth in a Plasma with a Bump-in-Tail Distribution :

C id a plasma
Consider l whose
h electrons
l t consist
i t off 2 spatially
ti ll uniform
if
components with densities n0a and n0b , and equilibrium distributions
g 0a (v z ) and g 0b (v z ) (see figure below). Assume that (1) n0a  n0b ;
(2) g 0a (v z ) has a v z -spread of vTa centered at v z  0; and (3) g 0b (v z )
has a v z -spread of vTb centered at v z  vb  vTa . Thus, g 0b (v z ) looks
like a small "bump" p in the "tail" pportion of g 0a (v z )).
To be self-consistent, we assume further that the ion density is
equal to the total electron density, and the ions drift to the right with
a current equal and opposite to the electron current.current Thus Thus, there is
no electric or magnetic field at equilibrium.

g 0a ( v z ) vTb
g 0b ( v z )
vTa
vb vz
0 47

6.5 Landau Damping (continued)

it ik z z
Assuming e dependence with k z  0 and treating each
component separately as before,
before we obtain the dispersion relation:
pa
2 dg0 a ( vz ) 2pb dg0b ( vz )
1
kz2  1
L vz  
k
dvz
dv z  2
kz L vz 1 kz dvz
dv z  0 (77)

z  
( A) ( B)
For term (A), we assume  /k z  vTa (no resonant electrons). Thus,
g 0a (v z )   (v z ). An integration by parts gives: Term (A)  k z2 /  2 (78)
Since g 0b (v z )  g 0a (v z ), the real part of term (B ) is negligible
compared with term (A). However, 
g 0a ( v z ) kz
we must keep the imaginary part of
g 0b ( v z )
term (B ) because it determines i . vTa
vb
vz
dg0b ( vz ) 0
Thus, Term (B )   i dv . (79)
z vz  / k z
pa
2 pb
2
dg b ( vz )
(77)-(79) give 1  2   i 2 0dv 0 (80)
 kz
z  / k z
z v 48
 6.5 Landau Damping (continued)

 pa
2 2pb dg0b ( vz )
Rewrite (80): 1 
2   i
kz2 dvz v  / k
0 (80)

z z

Writing   r  ii and assuming i  1, we obtain by

r
  
expansion: 12  12 (1  i i )  12 (1  2i i ). Then (80) gives
2
 r r  r r

r   pa r : due to g 0 a ; r : due to g 0b


   pa ppb dg0b ( vz )   ppa n0 b dg0b ( vz )
2 3 (81)
i  2 kz 2 dvz v  / k
 n
2 kz 0 a dvz v  / k
2
 z r z z r z

(81) shows that the sign of i depends on the sign of dg0b / dv z

at v z  r / k z . Thus,
Thus i  0 (Landau growth) if r / k z falls on the
positive slope of g0b , and i  0 (Landau damping) if r / k z falls
on the negative slope of g0b . The   Landau 
g ( v ) k z  growth 
Landau growth is our first example 0a z
g 0b ( v z )
of unstable equilibrium solutions. vT a
vb v z 49
0

6.5 Landau Damping (continued)

A Qualitative Interpretation of Landau Damping and

Landau Growth :
vz
Assume that an electrostatic wave v ph  /kz
with phase velocity  /k z is present in
the plasma. An electron moving with
velocity v z sees the wave at the Doppler-shifted frequency   :
    k z vz ((82))
If      k z v z  0 (i.e. electron velocity  phase velocity), the
electron experiences almost a DC electric field. In this field, it will
gain or lose energy for an extended period of time (  2 /  ). This
phenomenon is known as resonant interaction.
Divide the electrons into  a slow electron
 a fast electron
slow electrons:      k z v z  0 v ph  /kz

fast electrons:      k z v z  0
50
 6.5 Landau Damping (continued)
For both slow and fast electrons, some
 a slow electron
will lose energy to the wave and some will  a fast
f t electron
l t
gain energy from the wave, depending on v ph  /kz
the position of the electron relative to the
phase of the wave
wave.
If a slow electron loses energy in the resonant interaction, its v z
decreases. Hence, its   (    k z v z  0), which is a positive number
b
becomes greater.
t As A a result,
lt the
th time
ti forf sustained
t i d interaction
i t ti ( 2 )
becomes shorter. This will result in weaker resonance.
On the other hand, if a slow electron gains energy in the resonant
i t ti v z increases
interaction, i andd   becomes
b smaller.
ll Hence,
H the
th time
ti
for sustained interaction becomes longer (stronger resonance). This
give the electrons in the energy-gaining phase the advantage and, on
average, slow electrons gain energy from the wave.
Similarly, fast electrons will, on average, lose energy to wave.
51

6.5 Landau Damping (continued)

We have just concluded that, on average, slow electrons (relative

to the phase velocity of the wave) gain energy from the wave and
fast electrons lose energy to the wave. Thus, if the plasma contains
more slow electrons than fast electrons (i.e. a negative slope of g 0
at  /k z , see left
l f figure),
fi ) the
h net effect
ff is i an energy transfer
f ffrom the h
wave to the electrons (Landau damping).
By similar argument, if the plasma contains more fast electrons
than slow electrons (see right figure), there will be a net energy
transfer from the electrons to the wave (Landau growth).
g 0 (v z ) 
 g 0a ( v z ) kz
kz g 0b ( v z )
vTe v T a
vz vz
0
more slow electrons more fast electrons
than fast electrons than slow electrons 52
 6.5 Landau Damping (continued)

Kinetic Treatment vs Fluid Treatment :

We have just considered a case in which details of the particle
distribution function determine whether a wave grows or damps.
On the other hand, fluid equations (derived in Sec. 1.4 of lecture
notes
t byb a simple
i l method)
th d) are formally
f ll derived
d i d from
f the
th Vlasov
Vl
equation (Sec. 7.2) by an integration procedure over the velocity
space, in which details of the distribution function are lost. Hence,
a fluid
fl id treatment will
ill miss
i the
h Landau
L d damping/growth
d i / h (Sec.
(S 7.3)
7 3)
and other effects sensitive to the distribution function, but results
of fluid equations are implicit in kinetic equations.
Chapter 7 contains a fluid treatment of important plasma modes
and instabilities. Here, as a supplement to Ch. 6, we will cover some
of these topisc in the framework of the Vlasov equation.
Our first study of fluid modes is on the two-stream instability.
More will be considered in subsequent sections of this chapter.
53

6.5 Landau Damping (continued)

Two-Stream Instability I :
g the bump-in-tail
Consider again p model for Landau ggrowth ((upper
pp
figure). The dispersion relation obtained by the kinetic approach is
 2pa dg0 a (vz )  2pb dg0b (vz )
1
k2
z  1
L vz  
k z
dvz
dv z 
kz2 L vz 1 kz dvz
dvz  0 (77)

Suppose the velocity spreads (vTa , vTb ) of the 2 components vanish

(lower figure). We then have a situation where one component streams
through
h h anothher component. Integrating
I i (77) by b parts andd letting
l i
g 0a   (vz ), g 0b   (vz  vb ), we
g 0a ( v z ) vTb
 2  2pb g 0b (v z )
obtain 1  
pa
0 vT a (83)
 ( kzvb )2
2 v z
0 v b
This example shows that the
the kinetic result [(77)] can be  (v z )
 (v z  vb )
reduced to the fluid result [(83)]
vb vz
in the proper limit. 0 54
 6.5 Landau Damping (continued)
 2pa  2pb
The dispersion relation: 1    0 [(83)] can be
 2 ( kzvb )2
 2pa kzvb 2  2pb
written (1  2 ) (1   )  2  (84)
    
  Da (, k z )
Da ( , k z ) Db ( , kz ) 1
 pa
  2pa
 Da ( , k z )  1   2 Db ( , k z )
where  kz
 D ( , k )  (1  kzvb ) 2 
 b z  pa / vb

(84) can be regarded as the coupling between the plasma mode

(Da  0) andd th
the beam
b moded (Db  0).
0) The
Th coupling
li is i strongest
t t
near the intersection of the two modes (see Figure). The intersecting

point is at    pa and k z  vpa , which are solutions of
b
 Da ( , k z )  0

 Db ( , k z )  0 55

6.5 Landau Damping (continued)

2pa kzvb 2 2pb
Rewrite (84): (1  2 )(1   )  2 (84)
 
To show that there is an instability, we will only look for the 
value at k z   pa /vb , i.e. at the point of strongest interaction. Letting
1  1 (1   );
   pa  , we gett  1 1 
 pa  pa ) 2  2pa (1  2  pa ) (85)
Sub. (85) and k z   pa /vb into (84) and keeping terms up to first
order in small quantities  and  pb , we obtain
 2  2pb n0b 2
2 2
pa  pa 2pa  ( n0a )
n i 
2 2in
1  ( n0b ) 3 e 3 , n  1,2,3
  3  12  3pa ( nb )2    21/3
a pa n
0a

1
2
1  ( n0b ) 3  1
    2  i 2
3 (86)
21/3 pa n0 a
 1  i 3 Unstable mode
 2 2 56
 6.5 Landau Damping (continued)
From (86), we find the frequency of the unstable mode:
n 2
   pa     pa [1  1/1 3 ( n0b ) 3 ( 12  i 23 ))]
2 0a
 n0b 23
r   pa [1  24/ 3 ( n0 a ) ]
1
 2
(87)
n
  3  ( 0b ) 3 [growth rate]
 i 24/ 3 pa n0 a
Likee thee case
c se of
o Landau
d u growth
g ow (upper
(uppe figure),
gu e), thee free
ee energy
e e gy
available in a non-Maxwellian plasma drives a two-stream instability
(lower figure) . However, in the latter vTb
g 0a ( v z )
case there are no electrons at exactly
case, g 0b ( v z )
the phase velocity of the wave. Thus, vTa
vz
there is no singularity problem and a 0
r /k z vb
fluid treatment will also be adequate.
adequate  (v z )
(See Sec. 7.13 for a fluid treatment of
r /k z  (v  v )
z b
a slightly different two-stream model.) 0 v v z57
b

6.5 Landau Damping (continued)

Comparing the Landau growth rate in (81) and the two-stream
   pa n0b dg0b (vz )
3
 
 i 2 kz2 n0 a dvz v  / k [Landau growth]
growth rate in (87):  z r z
2
  3  ( n0b ) 3 [Two
[Two-streamstream instability]
 i 24 / 3 pa n0 a
we may show that i [Landau growth]  i [Two-stream instability].
This is because, in the Landau growth (upper figure), electrons
drifting slower than the wave absorb energy from the wave, whereas
in the two-stream instability (lower
figure), all electrons deliver energy g 0a ( v z ) vTb
to the wave because they all drift g 0b ( v z )
vTa
faster than the wave. vz
0
r /k z vb
As the g 0b spread
p increases from 0
 (v z )
to a large value, the fluid instability r /k z  (v  v )
z b
will transition to a kinetic instability. v
0 vb z58
 6.5 Landau Damping (continued)

Two-Stream Instability II :
 ( v z  vb )  (v z  vb )
Consider two cold electron beams
of equal density nb streaming in an ion
neutralizing background in opposite  vb vb vz
0
directions (vb and  vb ). In (83), we
 2pb
have already obtained the term for the forward stream: .
( kz vb )2
By symmetry, the backward beam will have a similar form, with
vb replaced by  vb . Thus, following the same treatment leading to
(83), we obtain the dispersion relation:
 2pb  2pb
1   0, (88)
( kz vb )2 ( kz vb )2
which gives a quadratic equation in  2 :
 4  2( 2pb  k z2vb2 ) 2  2 2pb k z2vb2  k z4vb4  0 (89)
59

6.5 Landau Damping (continued)

Rewrite  4  2( 2pb  k z2vb2 ) 2  2 2pb k z2vb2  k z4vb4  0 (89)

The solution for  2 as obtained from (89) is
1
 2   2pb  k z2vb2   pb ( pb
2
 4k z2vb2 ) 2 (90)
It can be shown from (90) that for, k z2  2 2pb / vb2 , all value of
 are real (no instability). However, for k z2  2 pb
2
/ vb2 , two values
of  will be a pair of complex conjugates and one of them gives rise
to an instability.
We may find the wave number (k zmax ) for which the growth rate
maximizes by finding the value of k z2 for which d  2 / dk z2  0. The
3 pb
result is k zmax   2v , which corresponds to a maximum growth
b
rate of i   pb / 2. We will return to this problem again in Special
Topic II in connection with a new subject: the absolute instability. 60
 6.5 Landau Damping (continued)

Ion-Acoustic Waves : (See Sec. 7.3 for a fluid treatment)

At low frequencies,
frequencies the ion contribution to the dispersion relation
cannot be ignored. For low frequency electrostatic waves, we simply
add to (59) the ion term (the species subscript "e" or "i" is also added),
 2pe dge 0 (vz )  2pi dgi 0 (vz )
1
k2 z 
L vz  
k
1
dvz
z
dv z 
kz2  1
L vz  
kz
dvz
dvz  0 (91)

Assume a Maxwellian distribution for both the electrons and ions:

 g (v )  1 exp(  vz2 ) (92)
 e0 z gi 0 ( v z )
2 vTe 2vTe 2
 vz2  / kz ge0 (vz )
 g i 0 (v z )  1 exp(( 2 ) (93)
 2 vTi 2vTi vz
with vTe  vTi . Assume further vTe   / k z  vTi (see figure) so that
there is negligible electron Landau damping (because dg e 0 / dvz  0)
and negligible ion Landau damping (because  / k z  vTi ). We will
therefore neglect the imaginary part of both integrals in (91). 61

6.5 Landau Damping (continued)

dge 0 (vz ) vz dge 0 (vz )
For the electrons, sub. dvz
  2 g 0 into
vTe L vz 1 kz dvz
dvz

and neglect k (  vTe ) in the denominator, we obtain

z
 (75)
dge 0 (vz )
 v
L z  
1
d z
dv
kz
dvz   12
vTe  g0 (vz )dvz   v1Te2    2pe1De2 (94)

dvz , we assume k  vz and

dgi 0 (vz )
For the ion integral: L vz 1 kz dvz z

follow the same steps leading to (70). This gives the cold ion limit:
dgi 0 (vz )
L vz 1 kz
kz 2
dvz
dv z  ( ) (95)
 2pi 1
Sub. (94) and (95) into (91), we obtain 1   0, (96) 
kz2De
2 2
which is the most basic form of the dispersion relation because ion
thermal effects and electron Landau damp have all been neglected.
From (96), we find    pi , i.e. this is indeed a low-frequency wave. 62
 6.5 Landau Damping (continued)

1  2pi
The dispersion relation 1  0 (94)
kz2De
2 2
kz2De
2 2
kz2 kTe 4 n0e2 kz2 kTe
gives 2  pi
 
1kz2De2 1kz2De
2 4 n e2
0
mi 1kz2De
2 mi

  kzCs
kT
 with Cs  me (95)
1kz De
2 2 i

Physically, when the ions are perturbed, electrons tend to follow

the ions to shield their electric field, thereby reducing the restoring
forces on the ions. So the wave has a frequency lower than  pi ( pi
would be the ion oscillation frequency if the electrons were immobile).
However, the electrons cannot effectively shield the ion electric field
if  (wavelength)  De (see Sec. 1.2). Thus, when   De , or k z De
( ) shows that  will approach
 1,, (94) pp  pi .
Since the plasma motion is longitudinal and Cs is similar to the
sound speed of a neutral gas, the wave is called an ion acoustic wave. 63
 6.10 General Theory of Linear Vlasov Waves Part 2
We have so far treated only electrostatic waves in the absence of
an external
t l field.
fi ld In
I this
thi section,
ti we lay l the
th groundwork
d k for
f a generall
theory of linear waves, both electrostatic and electromagnetic, in an
infinite and uniform plasma. We assume that the plasma is immersed
i a uniform
in if externall magnetic
i field
fi ld along
l the i B 0  B0e z , but
h z -axis: b
there is no external electric field (E0  0).
Equilibrium (Zero-Order) Solution :
An equilibrium solution f 0 ( v ) must satisfy the zero-order Vlasov
equation: t f 0 ( v )  v f 0 ( v )  m (E0  1c v  B0e z )  v f 0 ( v )  0
q
  
  
0 0
0
 ( v  B0e z )  v f 0 ( v )  0 (101)
Thus, any function of the form f 0 (v ,vz ) satisfies (101), provided
that the total charge and current densities of all species vanish so that
there is no net self field at equilibrium. This in turn makes v and vz
constants of the motion in the only field present: B0e z . 1

6.10 General Theory of Linear Vlasov Waves (continued)

Examples of equilibrium solutions (normalized to n 0 ) are:

f  n0 v2
 3/2 v3 exp(  2v2 ) [Maxwellian] (102)
 0
(2 ) T T
 v2 2
n0
 f 0  (2 )3/2 v2 v exp( p( 2  vz2 ) [[bi-Maxwellian] ((103))
2vT  2vTz
 T  Tz
 f  n0  (v  v ) (v ) (104)
 0 2 v  0 z

In (103),
(103) the
h particles
i l have
h two temperatures, vT  andd vTz . In
(104), all particles have the same v ( v0 ) and vz ( 0). (104) is
approximately self-consistent if the self magnetic field due to the
gyrating particles is negligible.
First -Order Equations :
The linear p properties
p of a pplasma is contained in the dispersion
p
relation. To obtain the dispersion relation, we first linearize the set
of Vlasov/Maxwell equations by writing 2
 6.10 General Theory of Linear Vlasov Waves (continued)

 f (x, v, t )  f 0 ( v )  f 1 (x, v, t ) (105)

E(x, t )  E (x, t ) (106)
 1
As before, first-order
B(x, t )  B0e z + B1 (x, t ) quantities are denoted
(107)
  (x, t )   (x, t ) byy subscript
p "1". ((108))
 1
J (x, t )  J1 (x, t ) (109)
Sub. (105)-(109) into the Vlasov/Maxwell equations. Zero-order
terms give the equilibrium solution.
solution Equating the firstfirst-order
order terms
terms,
  f 1  v f 1  mqc ( v  B0e z )  v f 1
 t 
 q
  m (E1  1c v  B1 )  v f 0 (110)
  B  0 (111)
 1
we obtain 
  E1  41 (112)

  E1   1c  B1 (113)
 t
 1  4
  B1  c t E1  c J1 (114) 3

6.10 General Theory of Linear Vlasov Waves (continued)

 1 (x, t )   q  f 1 (x, v, t ) d 3v (115)

 
where
h 
J1 (x, t )  
q  f 1 (x, v, t ) vd 3v (116)

Note: In Nicholson (6.145)-(6.153), zero-order fields depends on

x and t. But from (6.154) on, E0  0 and B 0  B0e z , as in our model.
Particle dynamics: These first-order equations (110)-(116) are
coupled.
p To examine the particle
p dynamics,
y , we start from ((110):
)
 f  v f  q ( v  B e )  f
t  1  1 m c 0 z v 1
q
  m (E1  1c v  B1 )  v f 0 ((110))
The LHS is a total time derivative [ dtd f ] along the zero-order
1
q
orbit because the acceleration force is c v  B0e z . Thus, (110) can
be written d f   q (E  1 v  B )  f (117)
dt  1 m 1 c 1 v 0
4
 6.10 General Theory of Linear Vlasov Waves (continued)

Rewrite (117):  f  0 ( v  ,v z )

d f ( x, v, t )   q [E ( x, t )  1 v  B (x, t )]  f ( v ) , (117)
dt  1 m 1 c 1 v 0
d follows the zero-order orbit of a particle, which we denote
where dt
b x(t ) and
by d v(t )). U
Under
d ththe condit
ditions:
i x(t   t )  x andd v(t   t )  v,
vx (t )  v cos[   (t   t )] y
v (t )  v sin[   (t   t ))] v
 y  
v (t )  vz 
 z x
we have  x(t )  x  v sin[   (t   t )]  v sin  (118)
   
 y(t )  y  v  v
cos[   (t   t )]   cos 
   
 z (t )  vz (t   t )  z
q B
where   m c0 . (118) reduces to (1.25) if wes set  (polar angle of
v  )   / 2 and t  0. At t   t , (118) gives x = x and v = v as required.5

6.10 General Theory of Linear Vlasov Waves (continued)

Change the variable t in (117) to t  and note that x(t   t )  x and

v(t   t )  v. A t -integration
integration of (117) from   to t gives
t

d f [x(t ), v(t ), t ]dt   f (x, v, t )  f [x(t ), v(t ), t ]
dt  1 1 1 t 
q t
  m
 
dt {E1[x(t ),
) t ]  1c v(t )  B1[x(t ),
) t ]} 
 v f 0 ( v) (119)
We now consider a normal mode by assuming
E1 (x, t ) E1k 
B (x, t ) B  it ik x
1
 J ( x, t )  1k
e (120)
 1  J 1k  As before, subscript "k"
  1
f ( x , v , t )   1k 
f ( v ) denotes a normal mode.
mode
where E1k , B1k , J1k are complex constants, and f 1k ( v ) is a complex
function of v. J1k can be expressed in terms f 1k ( v ) as
J1k   q  f 1k ( v ) vd 3v (121)

Question: Why is 1 ( x, t ) not included in (120)? (see note below) 6
 6.10 General Theory of Linear Vlasov Waves (continued)

Sub. (120) into

f 1 (x, v, t )  f 1[ x(t )), v(t )), t ]t 
q t
  m 
dt {E1[x(t ), t ]  1c v(t )  B1[x(t ), t ]}  v f 0 ( v) (119)
we obtain
f 1k ( v )e it ik x  f 1k [ v(t )]e it ik x
t  
 it  i (t   t )  ik  x  ik  ( x  - x)
t



dt [E1k  1c v(t )  B1k ]e it ik x  v f 0 ( v)
q
  m 
t
  m eit ik x  dt [E1k  1c v(t )  B1k ]  v f 0 ( v)
q


 e i (t t )ik ( x-x) (122)

(122) the opertaor operates on f 0 ( v) only.
Note : In (122), only

6.10 General Theory of Linear Vlasov Waves (continued)

We assume i ( Im  )  0, then f 1k [ v(t )]e it ik x  0 and

t 
(122) becomes
f 1k ( v )e it ik x
t
  m e it ik x  dt [E1k  1c v(t )  B1k ]  v f 0 ( v)
q


 e i (t t )ik ( x-x)

Factoring out the eit ik x dependence, we obtain
q t
f 1k ( v )   m 

dt [E1k  1c v(t )  B1k ]  v f 0 ( v)
(123) ei (t t )iikk ( x-x))
Note:
(i) 1 (x, t ) is not in (120) because, with the assumed e it ik x
d
dependence,
d 1k iis implicit
i li it in
i J1k though
th h the
th contitinuity
it equation:
ti

  J  t  0  ik  J1k  i1k  0  1k  1 k  J1k 8
 6.10 General Theory of Linear Vlasov Waves (continued)

(ii) Although (123) is derived under the assumption of i  0,

the dispersion relation to be obtained from (123) can be analytically
continued to an arbitrary  by the use of Landau contour for the vz -
g [[see ((135)].
integral )] The argument
g follows the treatment of Landau
damping.
(iii) v as a vector is a function of t . However, it is understood
that f 0 ( v) is a function of scalars v and vz , both being constants
of the motion. Hence, in writing f 0 ( v) in (119), (122), and (123),
v is not displayed as a function of t .
(iv) The method we employed to obtain (123) is called "method method
of characteristics" or "integrating over unperturbed orbit".

6.10 General Theory of Linear Vlasov Waves (continued)

Field equation: From the linearized Maxwell equations:
  E1   1c  B1 (113)
t
  4
  B1  c t E1  c J1
1 (114)
we obtain   (  E1 )   1c t   B1   c12 t 2 E1  4c2 t J1
2
(124)
E1 (x, t ) E1k 
B (x, t ) B  it ik x
For a normal mode:  1   1k e (120)
J1 (x, t ) J1k
  
 f 1 (x, v, t )  f 1k ( v ) 
k  (k  E1k )  c2 E1k   4c2i J1k
2
((113)) and ((114)) ggive ((125))
A note on notations: Subscripts "0" and "1" indicate, respectively,
zero and first order quantities. Subscript "k" indicates a normal mode.
S b i t " " iindicates
Subscript di t particle
ti l species.
i
(123) and (125) together with the orbit equations (118) form the
basis for our treatment of linear plasma waves in Secs. 6.11 and 6.12. 10
 6-11 Linear Vlasov Waves in Unmagnetized Plasma
q t
) f 1k ( v )   m
Rewrite ((123): 
dt [E1k  1c v(t )  B1k ]  v f 0 ( v)
 e i (t t )ik ( x-x) (123)
Assume the absence of an external magnetic field (  0). Then,
 v(t )  v  const
(118) reduces to  , and (123) can be written
x(t )  x  v (t   t )
t
 v f 0 ( v )  dt e i (t t )ik v (t t ) (126)
q
f 1k ( v )   m (E1k  1c v  B1k ) 


Again, assuming i  0, we obtain from (126)

q (E1k  1c vB1k ) v f 0 ( v )
f 1k ( v )  m i ( kv ) (127)
Note: (127) can be readily derived by setting B0  0 and sub. the
normal mode (120) into the linearized Vlasov equation (110):
 f  v f  q ( v  B e )  f   q (E  1 v  B )  f
t  1  1 m c 0 z v 1 m 1 c 1 v 0
but in the presence of B0e z , we must use (123) (see next section). 11

6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)

q (E1k  1c vB1k ) v f 0 ( v )

Rewrite f 1k ( v)  m i ( kv ) (127)
Assume f 0 ( v ) is an isotropic function, i.e. f 0 ( v)  f 0 (v).
Then, (v  B1k )  v f 0 (v)  0. Since there is no external magnetic
field and f 0 (v ) is isotropic, the plasma properties are also isotropic.
Without loss of generality, we assume k  k z e z . Thus, (127) becomes
q E1k 
 v f 0 ( v )
f 1k ( v )  m i ( kz vz ) (128)

Sub. (128) into J1k   q  f 1k ( v) vd 3v [(121)], we obtain


q2 E  v f (v )
J1k   m  v 1ik( kv0 ) d 3v (129)
 z z

The field equation: k  (k  E1k )  c2 E1k   4c2i J1k [(125)]

2

k z2 E1kz e z  ( k z2  c2 )E1k   4c2i J1k

2
may be written (130)
12
 6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)
Electrostatic waves :
 q2 E1k  v f 0 (v ) 3
J 
 1k  m  v i ( kz vz ) d v ( )
(129)
Rewrite 
k z2 E1kz e z  (k z2  22 )E1k  4 2i J1k  0 (130)
 c c
For electrostatic waves, E1k  k ( k z e z ). So we set
E1k  E1kz e z (131)
(129) then gives f (v ) 0
  i m E1kz  (vx e x  v y e y  vz e z )  vkz v d 3v
q2
J1k (132)
 z z
1
(132) f 0 (v)  f 0 [(vx2  v 2y  vz2 ) 2 ] iis an even ffunction
IIn (132), i off vx
and v y . Hence, the x and y components vanish upon, respectively,
vx and v y integrations,
g and we have
f 0 (v )
vz
  i m E1kz e z   kvvz d 3v
q2
J1k (133)
 z z 13

6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)

f 0 (v )
vz
  i m E1kz e z   kvvz d 3v
q2
Rewrite
i J1k (133)
 z z

Defining g 0 (vz )  n1 0  f 0 (v)dvx dv y [as in (28)], we obtain

dg (v )
v 0 z
 z
n 0q2
J1k   i m E1kz e z   dv z
kz vz
dv z

Writing  vkz v  k1 ( 1   k v ), we have
z z z z z

J1k   i m0  E1kz e z k1  (1   k v ) dv0 z dvz ,
n q 2 dg (v )
 z z z z
The first term vanishes upon vz -integration because g 0 (vz )  0
at vz  . Thus,
dg 0 (vz )
n 0q2  
J1k   i m k 2 E1kz e z  v dv

z
 dv z , (134)
 z z k z 14
 6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)
dg 0 (vz )

Sub E1k  E1kz e z and J1k   i m0  k2 E1kz e z 
n q2 dvz
Sub. vz   dv z
 z kz

k z2 E1kz e z  ( k z2  c2 )E1k   4c2i J1k

2
into the field equation: (130)
we obtain the dispersion relation for electrostatic waves:
dg 0 (vz )
 2p
L
1  dvz
 kz2 vz   dvz  0,, ((135))
kz

where, by the recipe in (62), we have replaced  dvz with L dvz .
(127) we have assumed i  0.
In deriving (123) and (127), 0 With the
Landau contour,  in (135) can have any value provided the pole
 /k z does not cross the Landau contour.
(135) agrees with the electrostatic disperson relations in Sec.
Sec 6.5
65
for the Langmuir wave, Landau damping/growth, two-stream insta-
bilities, and ion acoustic waves. 15

6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)

Discussion: Rewrite   B1  1c t E1  4c J1 (114)

For the normal mode in (120), the RHS of (114) gives
1  E  4 J  ( i E  4 J )e it ikx (136)
c t 1 c 1 c c 1k
1k
dg 0 ( vz )
n 0q2  
Inserting J1k   i m k 2 E1kz  v dv

z
 dv z e z [(134)] and
 z z k z
E1k  E1kz
k e z [(131)] into (136),
(136) we find
dg 0 ( vz )

i E  4 J  ci E1kz e z  4c  i m0  k2 E1kz e z  vdv
n q 2
z
z  k
c 1k c 1k dv z
 z
z
dg 0 ( vz )
   2p

 c E1kz e z {1   2
i dvz dv }  0
k L vz  
 z
kz 
z

0 by (135)
This shows that, the displacement current and particle current
exactly cancel out. Hence, we have an electrostatic wave. 16
 6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)

Electromagnetic waves :
 q2 E1k 
 v f 0 ( v ) 3
J 
 1k  m  v i ( kz vz ) d v
 (129)
Rewrite 
 k z2 E1kz e z  (k z2  2 )E1k  4 2i J1k  0
2

 c c
(130)
For electromagnetic waves, E1k  k (  k z e z ). So, without loss of
generality (because the plasma is isotropic), we set E1k  E1ky e y (137)
Then (129) and (130) give
Then,
 f 0 (v )
J   i  E q 2 vy
 ( v e  v e  v e )  
d 3v (138)
 1k

m  1ky x x y y z z k z vz
(k 2   2 ) E e  4 i J  0 (139)
 z c2 1ky y c2 1k
f 0 (v) is an even function of vx . Hence, the x-component of (138)
vanishes upon vx -integration. v f 0 (v) is an odd function of v y .
y
Hence, the z -component of (138) vanishes upon v y -integration. 17

6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)

We are then left with only the y component of (138):
v y f 0 (v )
q2 v
J1k   i m E1ky e y   k vy d 3v (140)
 z z

Integrating (140) by parts of over v y yields

q2 f (v )
J1k   i m E1ky e y  0k v d 3v
 z z

Using the one-dimensional equilibrium distribution function:

g (v )  1 f (v)dv dv ,
0 z n 0  0 x y
n q2 g (v )
we mayy write ((140)) as J1k   i m0  E1ky e y  0k vz dvz ((141))
 z z

Sub. (141) and E1k  E1ky e y into

 k z2 E1kz 2 4 i
k e z  ( k z  c2 )E1k  c2 J1k  0
2
(130)

k z2  2     2p  0k vz dvz  0

2 g (v )
we obtain (142)
c  z z 18
 6-11 Linear Vlasov Waves in Unmagnetized Plasma (continued)

Rewrite k z2  2     2p  0k vz dvz  0

2 g (v )
(142)
c  z z

EM waves in a plasma have a phase velocity  c [see (143) below].

Hence, we may assume  / k z  vz and neglect the the k z vz term in
the demonamitor of the integral in (142)
(142). 
g (v )
  0k vz dvz  1  g 0 (vz )dvz  1
z z

This results in the dispersion relation:  pe

kz
 2  k z2c 2   2pe (143)
where we have neglected the small ion contribution.
The  vs k z plot (see figure) is similar to that of the waveguide.
There is a cutoff frequency  pe , below which EM waves can not
ppropagate.
p g Short radio waves (  10 MHz)) are hence reflected from
the ionosphere. This has been exploited for long-range broadcasting.
By comparison, the free space is non-dispersive with   k z c. 19

6-12 Linear Vlasov Waves in Magnetized Plasma

(Ref.: Krall and Trivelpiece, Sec. 8.10)
Dispersion Relation : We begin this section with a derivation of
the general dispersion relation for waves in an infinite, uniform, and
magnetized plasma on the basis of the following linearized equations
derived in Sec. 6.10 for a normal mode with eit ik x dependence:
 f ( v )   q t dt [E  1 v(t )  B ]  f ( v)
  1k m  1k c 1k v  0

  e i (t t )ik ( x-x) (123)
  4 i
 k  (k  E1k )  c2 E1k   c2 J1k
2
(125)

where we have assumed a uniform external magnetic field B 0  B0e z ,
and shown that the equilibrium distribution function in such a field is
f 0 ( v )  f 0 (v ,vz )). So the plasma is isotropic in the x, y -dimensions,
-dimensions
but it is 3-dimensional anisotropic. Thus, we expect the conductivity
 
to be in the form of a tensor  : J1k    E1k . 20
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

General form of the dispersion ralation:

 xx  xy  xz   E 
    1kx 
Write J1k    E1k   yx  yy  yz   E1ky  (144)
  
 zx  zy  zz   E1kz
k 
Without loss of generality (for a plasma isotropic in x, y ), we let
k  k e x  k z e z (145)
Sub. (144) and (145) into the field equation:
k  (k  E )   E   4 i J
2
1k c2 1k c2 1k (125)
the x, y, z components are
 (1  kz2c2  4 i  ) E  4 i  E  ( k kz c2  4 i  ) E  0
 2  xx 1kx  xy 1ky 2  xz 1kz
 4 i kc2 2
4 i 4 i
   yx E1kx  (1   2    yy ) E1ky    yz E1kz  0 (146)
 k k c2
 ( z2  4 i  zx ) E1kx  4 i  zy E1ky  (1  k c2  4 i  zz ) E1kz  0 21
2 2

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

(146) can be written
D Dxy Dxz   E1kx 
  xx  
D  E1k   Dyx Dyy Dyz   E1ky   0 or (147)
D  
 zx Dzy Dzz   E1kz 
1  kz2c2  4 i  4 i  k kz c2 4 i  E 
  2  xx  xy  2    xz   1kx 
 4 i  1  kc2  4 i  yy
2 2
4 i    E   0 (148)
  yx  yz   1ky 
 k kz2c  4 i  4 i  1   2  4 i  zz   E 
2 kc2 2

   zx  zy
  
1kz 
For (147) or (148) to be solvable,
solvable the determinent
determinent of D must vanish:
Dxx Dxy Dxz

D  Dyx Dyy Dyz  0 (149)
Dzx Dzy Dzz
(149) is the most comprehensive form of the dispersion relation. 22
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Particle dynamics: In (149), the conductivity tensor  is still
unknown To obtain the specific expression of the dispersion relation
unknown. relation,
we need to work on the equations for particle dynamics.
Define   t  t  and rewrite (123) and (118) in terms of 
q 0
f 1k ( v )   m  d [E1k  1c v ( )  B1k ]   v f 0 ( v )


 ei ik [ x( )-x] (150)

y
vx ( )  v cos(    )
v  ( )  v sin(    ) v
 y  
v  ( )  v 
x
 z z
 x ( )  x  v sin(    )  v sin  (151)
  
 v v
 y ( )  y   cos(    )   cos 

 z ( )  vz  z 23

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Using k  k e x  k z e z [(145)] and the orbit equations in (151),
we may write k  [x ( )  x]  k [ x ( )  x]  k z [ z ( )  z ]
kv
 [ sin(    )  sin  ]  k z vz

i ik [ x( )-x ] i ( kz vz ) i kv [sin(  )sin  ]

 e e

Using the Bessel function identity: eix sin    J s ( x )e is ,
s  
 i kv sin(  )
  J s (   )eis (  )
 k v
e 
 
we obtain  k v s
 i   sin
i  kv is
 e   J s  (  )e
s
i ik [ x( )-x i ( kz vz ) i kv [sin(  )sin  ]
x] 
 e e
i ( ss )
   J s (   ) J s (   )ei ( kz vz  s )
k v k v
(152)
s s   24
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Since f 0 ( v )  f 0 (v , vz ) and v ( v ), vz ( vz ) are constants

off th
the motion,
ti we have
h  v f 0 ( v )   v f 0 (v , vz )
y v  / v
e f  f f
  0 e   v1  0 e   0 e z
v v   vz
 f f
x  2 20 v   2vz 20 e z (153)
v vz
Thus,
h E1k   v f 0 ( v )  2(( E1kx e x  E1ky e y  E1kz e z )
f f
 ( 20 v   vz 20 e z )
v vz
f f
 2( E1kx vx  E1ky v y ) 20  2 E1kz vz 20 (154)
v vz
(120) we obtain B1k  
From (113) and (120), c k  E . Then,
1k Then
v  B  c v  (k  E )  c [( v  E )k  (k  v )E ]
1k  1k  1k 1k (155)
25

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

(153) and (155) give
1 ( v  B )   f  2 [( v  E )k  (k  v )E ]  ( f 0 v  v f 0 e )
c 1k v  0  1k 1k v2  z vz2 z
 2 [( v  E )(k  v )  (k  v )( v  E )] f 0
 1k   1k v2

ff 
 [( v  E1k )k z vz  (k  v )vz E1k ] 20 
vz 
 2 [(v E  v E  v E )k v
 x 1kx y 1ky z 1kz  x
 f
 (k vx  k z vz )(vx E1kx  v y E1ky )] 20
v
f 
 [(vx E1kx  v y E1ky  vz E1kz )k z vz (k vx  k z vz )vz E1kz ] 20 
vz 
2 [(k v E  k v E  k v E )v f 0
 z x 1kx z y 1ky  x 1kz y v2

f
 (k z vx E1kx  k z v y E1ky  k vx E1kz )vz 20 ] (156)
vz 26
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Combining (154) and (156), we obtain

vz

(E1k  1c v   B1k )   v f 0  2vx X  2v y Y  2 vz Z
 v [ei (  )  ei (  ) ] X
 iv [ei (  )  ei (  ) ]Y  2vz Z (157)
 f 0 vz f 0 f 0
 X  E1kx v2   (k z E1kx  k E1kz )( vz2  v2 )

 f v f f
where Y  E1ky 20  z k z E1ky ( 20  20 ) (158)
v vz v

 f 0
 Z  E1kz vz2

Note: (i) v , vz , and v are constants of the motion, but v  ,vx
and v y are time dependent.
(ii) X , Y , and Z are functions of constants of the motion. 27

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Combining (152) and (157) gives

(E1k  1c v   B1k )   v f 0 ei ik[ x( ))-x]
i ( ss )
 v X [ei (  )  ei (  ) ]  J s J s ei ( kz vz s )
s s
i ( s s )
 ivY [e i ( 
 )
e i ( 
 )
]  J s J s ei ( kz vz s )
s s
i ( kz vz  s ) i ( s s )
 2vz Z   J s J s e (159)
s s
ei (  )  i ( k v s ) i ( s s )
Write   J s J s  i (   )  e z z  
s s 
e 
ei[ kz vz ( s 1) )] i ( ss 1) 
   J s J s 
s s i[ kz vz ( s 1) )] i ( s s 1) 
 n  s  1 e 
 n  s  1
   J J s  i ( kz vz n ) i ( sn)
    n1   e (160)
n s  J n1 J s 
  28
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Sub. (160) into (159), we obtain

(E1k  1c v   B1k )   v f 0 e i ik[ x( )-x]
   [v X ( J n1  J n1 )  ivY ( J n1  J n1 )  2vz ZJ n ]J s
n s
i ( kz vz n ) i ( sn)
 e  e (161)
the only factor that depends on 
q 0
S b (161) iinto f 1k ( v )   m  d [E1k  1c v ( )  B1k ]
Sub.

  v f 0 ( v )e i ik[ x( )-x] (150)

and carrying out the  -integration,
integration we obtain
v X ( J n 1 J n 1)ivY ( J n 1 J n 1) 2vz ZJ n
J s ei ( sn)
q
f 1k ( v )  m    i (  k v  n )
n s  z z  (162)
This is Eq. (8.10.8) in Krall & Trivelpiece. Note that all Bessel
kv
functions have the same argument:   .
 29

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

The conductivity tensor: The perturbed current [(121)] can be written:
 2 
J1k   q  f 1k ( v ) vd 3 v   q 0 v dv
d  0 d   dv
d z f 1k ( v ) v
 
First consider the x -component of J1k : v cos  e x  v sin  e y  vz e z

J1kx   q 0 v dv 0 d   dvz f 1k ( v ) 12 v (ei  e i )

 2 
(163)

By (144), J1kx can be expressed in terms of the conductivity tensor
as J1kx   xx E1kx   xy E1ky   xz E1kz (164)
Then,  xx is the coefficient of the sum of all E1kx terms in (163)
[which can be found from (162) and (158)]:
q2  2 
 xx   m 0 v dv 0 d  dvz 12 v (ei  ei )

f 0
v[ (1 kzvz ) kzvz f20 ]
v vz
( J n1  J n1 ) J s ei ( sn) (165)
2

n s i ( kz vz n ) 30
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Using the Bessel function identities:

 J  x   J  x   2n J n  x 
n1 n1 x
 (166)
 J n1  x   J n1  x   2 J n  x 
kv kv 2n kv
we may write J n1 (   )  J n1 (   )  J (  ) (167)
  kv n 
Sub. (167) into (165), we obtain
q2  2  nv 
 xx   m  v dv  d  dvz  
0 0  k

f 0
(1 kzvz ) kzvz f20
v vz
J n J s ei ( sn1)
 ei ( sn1)  ,
2

n s i ( kz vz n ) 
(168)
where we see that only the s   n  1 terms in the s  sum will survive
the  -integration.
31

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Carrying out the  -integration in (168), we obtain

2 q2   nv 
 xx   m    v dv  dvz  
 k
 n 0 2n
f 0 kz vz kz vz f 0  J
((1  )  kv n
v 2 vz2 
  J n ( J n 1  J n1 )
i ( kz vz n )
f 0 kz vz kz vz f 0
(1  ) 
2 q2   n22 2 v2 vz2
  m  n 0 2v dv  dvz Jn 
  k2 i ( kz vz n )
f 0 kz vz kz vz f 0
(1  ) 
 2p   n22 2 v2 vz2
       z k2 n i(kzvz n  ) ,
2 v dv dv J
 2 n 0
(169)
where f 0  n1 f 0 [hence  f 0 d 3 v  1].
0 32
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

The general dispersion relation: Rewrite the dispersion relation:

Dxx Dxy Dxz

D  D yx D yy D yz
Dzx Dzy Dzz
k 2c 2  i 
1  z 2  4 4 i  kkzc2  4 i 
 xx  xy 2  xz
 4 i  k 2c 2
i 
 4 4 i   0 (149)
 yx 1  2 yy  yz
kkzc2  4 i  4 i  1
k2c2 i 
 4
2  zx  zy 2 zz
k 2c2  i  , wee obtain
S b  xx from (169) into Dxx  1  z 2  4
Sub.
 xx

Dxx  1  z 2  2
k 2c2  
 

 p2  0
n
2v dv 

dvz

f 0 (v ,vz ) kz vz kz vz f 0 (v ,vz )

(1  ) 
n22 2 kv v2 vz2
 2 Jn ( ) (170)33
k  kz vz  n 

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

By similar method, we obtain the other elements of the dispersion

t
tensor. The
Th complete
l t results
lt are (see
( Krall
K ll & Trivelpiece,
Ti l i pp. 405-406)
405 406)
n22 2 kv
Dxx  1  z 2  2
k 2c2
   2p
J (
k2 n  
) (171)
 n

Dxy   2 i   2 n v J ( kv ) dJn (kv / )  (172)

 n
p k n 
 d (kv /  ) 
n vz 2 kv
Dxz  z 2  2   2p
k k c2 Jn ( ) (173)
  n k  
D yx   Dxy (174)
(k2 kz2 )c2 dJ n (kv /  ) 2
D yy  1   2   2p v2 [ ]  (175)
2  n d (kv /  )
 i   2 v v J ( kv ) dJn (kv /  ) 
D yz  2 (176)
 n
p z  n 
 d (kv /  ) 

34
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

n 2 kv
Dzxx  z 2  2
k k c2
   2p vz J (
k n  
) (177)
 n
 i   2 v v J ( kv ) dJn ( kv / ) 
Dzy   2 (178)
p z  n  d (k v /  ) 
 n    
k2c2  2 kv
Dzz  1 
2    2p vz2 J n2 (
 
) (179)
 n
 F (v , v )   2v dv  dv F (v ,v )
  z 0    z kzvz nz  (180)

 f 0 (v , vz ) kzvz )  kzvz f 0 (v , vz )
where    v2
(1   
vz2
(181)

 f 0 (v , vz ) n f 0 (v , vz ) f 0 (v , vz )
    [  ] (182)
 vz2 vz2 v2
Question: The plasma is isotropic in x and y. Why are Dxx and
D yy unequal? 35

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Waves Propagating Along B0ez (k = kzez): ex

The dispersion relation for waves propagating B0
along B0 = B0ez may be obtained by letting ez
k
k  0 in (171)-(179). ey
F a small
For ll argument,t the
th Bessel
B l functions
f ti Jn(x)
( ) and
d Jn(x)
( ) can be
b
approximately written
x0
 J n ( x)  1 ( x )n
 n! 2
 all x x0 (183)
 J ( x)  ( 1) n J ( x)   (1)n ( x )n
  n n n! 2
For n  0 and n  1, we have in the limit x  0,
 J 0 ( x)  1,, J1 ( x)  x , J 1 ( x)   x
 2 2
 (184)
 J 0 ( x)   2 , J1( x)  2 , J 1 ( x)   2
x 1 1
36
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
kv
In (171)-(180), the argument of all Bessel functions is .

U i (183) and
Using d (184)
(184), we fi
find
d th
thatt iin th li it k  0,
the limit 0
Dxz  Dzx  Dzy  Dzy  0 (185)
and the other elements become
 D  1  kz2c2  2  2 1  v 2 
  p 4    n1  v  n1 
2  (186)
 xx 2 

D
 yy  D xx ((187))
 2 i
 Dxy   D yx      p 4  v  n1  v  n1 
2 1 2 2  (188)
 
2 
 Dzz  1     p vz2 
2
(189)
  n 0
Thus, the dispersion relation (149) reduces to
Dxx Dxy 0
2 2
D yx D yy 0  Dzz ( Dxx D yy  Dxy D yx )  Dzz ( Dxx  Dxy )0 (190)
0 0 Dzz 37

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

2
Rewrite the dispersion relation: 2
Dzz ( Dxx  Dxy )0 (190)
Several modes are contained in (190). To find these modes, we
assume, for simplicity, that the plasma is isotropic in all 3 dimensions,
i.e. f 0 (v , vz )  f 0 (v).
Electrostatic waves:
One of the solutions of (190) is Dzz  0, which, by (189), (180),
andd (182)
(182), can bbe written
i
vz2  2 f 0
  vz
1  2   2p 0 2v dv  dvz 0
 k z vz 
vz d g 0 (vz )
 dv
or 1   p  dvz kzzvz   0,
1 2 (191)


where g 0 (vz )  0 2 v dv f 0 (v , vz ).
38
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
The integral in (191) can be written
vz d g 0 (vz ) [ ( kz vz )] d g 0 (vz )
 dvz  dvz
 dvz kzvz   k1z  dvz kz vz 


 d g 0 (vz ) 
d z
dv
 1  dvz  1  dvz d g 0 (vz )
kz kz vz  kz  dv

z 
d g (v ) 0
 dv  0 z
 2  dvz z  (192)
kz vz 
kz d g (v )
 2p  dvz  0 z
(191), we get 1   2  dvz  00,
vz  
Sub (192) into (191)
Sub.
 kz kz
which agrees with the electrostatic dispersion relation [(135)] for an
unmagnetized plasma.
plasma This is because electrostatic waves involve
particle motion along B 0 , and hence are unaffected by the magnetic
field. The mode considered below will provide an opposite example. 39

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

2 2
Electromagnetic waves: Rewrite Dzz ( Dxx  Dxy )0 (190)
 Dxx  iDxy  0 (193)
(190) gives two other solutions:  D  iD  0
 xx xy (194)


We recall that the dispersion relation D  0 [(149)] is based on

the condition for the solvability of the field equations in D  E1k  0
[(147)] For k  0,
[(147)]. 0 we have Dxz  Dzx  Dzy  Dzy  0, 0 Dxx  D yy ,
and D yx   Dxy [see (187)]. Then, for solutions (193) and (194),
 Dxx Dxyy   E1kx   Dxx Dxyy   E1kx 
(147) gives
i D D   E    D D  E   0 (195)
 yx yy   1ky   xy xx   1ky 
 E   Dxy E
 Dxx E1kx  Dxy E1ky  0  1kx Dxx 1ky
    (196)
  Dxy E1kx  Dxx E1ky  0
D
 E1kx  Dxx E1ky
 xy
40
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

 E   Dxy E
 1kx Dxx 1kyy
R
Rewrite:
it  Dxx
(196)
 E1kx  D E1ky
 xy
The following information about the modes in (193) and (194) can
be immediately learned from (196):
2 2
(1) The 2 equations in (196) are consistent only when Dxx   Dxy ,
or when (190) is satisfied.
satisfied This is a specific example which shows the
dispersion relation as the condition for solvability of field equations.
We also find that (196) gives the relative amplitude (not the absolute
values) of the field components,
components as is typical of linear solutions.
solutions
(2) The fields in (196) are in the x - y plane. With k  k z e z and
E (x, t )  E e it ikz z [see (120)], we find
1 1k
  E1 ( x, t )  ik z e z  E1k e it ikz z  0.
Thus, the two solutions represent electromagnetic waves. 41

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Dxy
(3) Either of the equations in (196) gives E1kx   D E1ky . Hence,
xx

iE1ky for Dxx  iDxy  0  E [e x  ie y ] (198)

E1kx    E1k   1k
iE for Dxx  iDxy  0  E1k [e x  ie y ] (199)
 1ky
where E1k  ( E12kx  E12ky )1/ 2 .
y
Thus, both waves are circularly polarized*. Without
E1
loss of g
generality
y ((explained
p later),
), we assume ppositive x
 and B0 . At any fixed position z, the field in (198)
E1k (e x  ie y )eit ikz z
rotates opposite to the gyration of electrons in B0 e z . left circularly
In another view,
view it rotates in the direction of left
left-hand
hand polarized
y
fingers if the thumb points to the direction of B0 e z ,
E1
hence the name "left circularly polarized wave". In x
contrast (199) gives a "right
contrast, right circularly polarized
E1k (e x  ie y )eit ikz z
wave" rotating in the same sense as the electrons. right circularly
*Note: The circular polarization is due to Dxx  D yy . polarized 42
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

 Dxx  iDxy  0 (193)

g to the dispersion
Turning p relations: 
 Dxx  iDxy  0 (194)
Dxx and Dxy are given by (186) and (188), respectively.
 kz2c2  2  2 1  v 2  
D
 xx  1 
 2   p  4    
 v2 
 
(186)
 n 1 n 1

 Dxyy   2 i   2p 1  v 2   v2  
  4    
 
(188)
 n 1 n 1

 Dxx  iDxy  1  z 2  2   2p [ v2 

k 2c2  v2 
  n1 n1
 v2   v2  ]
n1 n1

 1 z 2  
k 2c2
    2p v2 
n1
0 (200)

Similarly,
Dxx  iDxy  1  z 2  
k 2c2
    2p v2 
n1
0 (201)
43


6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Using (181) and (182), (200) and (201) can be written

 f 0 kz vz kz vz f 0
 2 (1  ) 
v2 vz2 3
  k z c  2    p  v dv dvz  0 (202)
2 2 2
  kz vz 

 [l ft circularly
[left i l l polarized
l i d wave]]

 f 0 kz vz kz vz f 0
(1  ) 
 2 v2 vz2 3
  k z c  2    p  v dv dvz  0 (203)
2 2 2
  kzvz 

 [right circularly polarized wave]
These two dispersion relations in their present forms allow an
anisotropic f 0 [e.g. (103) and (104)], which may lead to an
instability (an example will be provided at the end of this section).
plasma, f 0  f 0 (v2  vz2 ) [e
For an isotropic plasma [e.g.
g (102)]
(102)], wee have
ha e
f 0 f 0
 2 . Then, (202) and (203) reduce to
v2 vz 44
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

 f 0
 2 v
  k z c      p   k v  v dv dvz  0
2 2 2 2
(204)
  z z 
 [left circularly polarized wave]

 f 0
 2  k 2 c 2     2 v
 z  p   kzvz v2 dv dvz  0 (205)


 [right circularly polarized wave]
Integrating by parts with respect to v , we obtain
 2 f 0
  k z c  2    p   kz vz  v dv dvz  0
2 2 2
(206)
 
 [left circularly polarized wave]

 2  k 2 c 2  2   2 f 0
 z  p    kzvz  v dv dvz  0 (207)


 [right circularly polarized wave] 45

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

The basic properties of the waves can be most clearly seen in a

cold
ld plasma.
l So,
S we let
l t f 0  21v  (v ) (vz ) (208)

 2 
Note:  f 0 d 3 v   v dv 
0 0
d 

dvz f 0  1
Then, (206) and (207) give
 2  2p
  k z c      0
2 2
[left circularly polarized] (209)
  

 2  k 2 c 2   2p
 z 

   0 [right circularly polarized] (210)
As an exercise in kinetic treatment of plasma waves, we have
gone through great length to arrive at the above dispersion relations
for
o a cocoldd pplasma.
as a. In fact,
act, ((209)
09) aandd ((210)
0) ca
can be readily
ead y dederived
ved
from the fluid equations [see Nicholson, Sec. 7.10; Krall &
Trivelpiece, Sec. 4.10].
46
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
Assume the plasma contains only one ion species of charge e and
q B
mass mi . In
I all
ll equations,
ti  ( m c0 ) carries
i the
th sign
i off q andd B0 .

eB eB
To be more explicit, we define the notations: e  m c0 ; i  m c0
e i
Then (209) and (210) can be written
Then,
 2  2pe  2pi
  k z c  [      ]  0 [left circularly polarized] (211)
2 2
 e i
 
 2  k 2 c 2  [  pe  pi ]  0 [right circularly polarized] (212)
2 2

 z  e  i
Each equation can be put in the form of a 4th order polynomial in .
So, for a given k z , there are 4 solutions for . However, with  changed
to   , one equation become the other equation. Thus, a negaitive-
solution
l ti off one equation ti is i identical
id ti l to iti  solution
t a positive- l ti off the
th other
th
equation. So there must be 2 positive- and 2 negative- solutions for
each equation. This results in a total of 4 independent solutions. 47

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Furthermore, with a change of the sign of B0 , the two equations

also reverse
reverse. So,
So without loss of generality [confirming the statement
following (199)], we may restrict our consideration to positive-
solutions for a positive B0 (i.e. B 0 is in the positive z direction).
The 4 independent, positive  solutions of (211) and (212) in a
independent positive-
positive B0 are shown in Fig. 1 or 2 in four branches, ranging from
very low to very high frequencies. Various waves in these branches
will
ill be
b classified
l ifi d below
b l accordingdi to their
h i frequency
f range.
(See Nicholson, Sec. 7.10 & 7.11; Krall & Trivelpiece, Sec. 4.10).
Fig. 1 Fig.
g 2
2
1
2
e e
(right circularly polarized)  (right circularly polarized)
1
whistler
i i
(left circularly polarized) (left circularly polarized)
kz k z48
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

A. High frequency electromagnetic waves - Faraday rotation

For high frequency waves (  i ), ) the ion terms in (211) and
(212) can be neglected. Thus,
 2  2pe
  k z c     0 [[left circularlyy ppolarized]]
2 2
((213))
e
  2pe
 2  k 2 c 2 
 z  e  0 [right circularly polarized] (214)

The high frequency branches are plotted in the top two curves
in Figs. 1 and 2.
Fig.
g 1 Fig.
g 2
2
1
2
e e
(right circularly polarized)  ( i ht circularly
(right i l l polarized)
l i d)
1
whistler
i i
(left circularly polarized) (left circularly polarized)
kz k z 49

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Setting k z  0, we find the cut-off frequencies of the two branches:


1   2  e2  e
2 pe 4 2
(215)

The 2 figures differ in plasma densities. When  pe  2e , we

have 1  e (Fig. 1). When  pe  2e , we have 1  e (Fig. 2).
As B0  0 (e  0), the 2 branches coalesce with the same cutoff
frequency  pe and the same dispersion relation  2  k z2 c 2   2pe  0,
consistent with (143).
Fig.
g 1 Fig.
g 2
2
1 2
e e
((right y ppolarized)) 
g circularly (right circularly polarized)
1
whistler
i i
(left circularly polarized) (left circularly polarized)
kz k z50
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

As shown in Figs. 1 and 2, at a given frequency, the right circularly

polarized wave has a greater phase velocity than the left ciucularly
polarized wave. Hence, if a linearly polarized wave is injected into
the plasma, it may be regarded as the superposition of a right circularly
polarized wave and a left circularly polarized wave of equal amplitude,
amplitude
each traveling at a different phase velocity. The combined wave is still
linearly polarized but its E field (i.e. its polarization) will rotate as the
wave propagates.
t This
Thi isi called
ll d the
th Faraday
F d rotation t ti andd is
i exploited
l it d
for plasma density measurement because the degree of polarization
rotaion depends on the plasma density.
In an unmagnetized plasma, there is no electromagnetic wave
below the cutoff frequency  pe . A magnetized plasma, however,
can support other branches of electromagnetic waves at frequencies
below the cutoff frequencyies of the top two branches, as discussed
below.
51

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

B. Intermediate frequency electromagnetic waves - whistler wave

and electron cyclotron wave
In the intermediate frequency range, we still have   i , hence
(213) and (214) still apply. (213) has no other solution in this range.
(214) has a solution marked as "whistler"
whistler & "electron
electron cyclotron wave
wave"
in Figs. 1 and 2. The electron cyclotron wave can be exploited for
electron cyclotron resonance heating since it has the same frequency
as the
th electron
l t cyclotron
l t frequency
f andd it rotates
t t ini the
th same sense as
the electrons.
Fig.
g 1 Fig.
g 2
2
1 2
e e
((right y ppolarized)) 
g circularly (right circularly polarized)
1
whistler
i i
(left circularly polarized) (left circularly polarized)
kz k z52
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

For the whistler wave, the group velocity vg ( d  / dk z ) increases

as  increases.
increases When a lightning stroke on earth generates a pluse of
EM waves containing many frequencies, the pulse may reach the
ionosphere and propagate along the earth magnetic field as a whistler
wave Some the wave will eventually leave the ionosphere to impinge
wave.
on the earth, where it can be received by a radio and generate a sound
like that of a whistle, hence the name whistler wave. The radio signal
h a dduration
has ti longer
l than
th the
th original
i i l pulse
l because
b components t att
different frequencies travel at differnet vg in the ionosphere.
Fig.
g 1 Fig.
g 2
2
1 2
e e
((right y ppolarized)) 
g circularly (right circularly polarized)
1
whistler
i i
(left circularly polarized) (left circularly polarized)
kz k z53

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

C. Low frequency electromagnetic waves - Alfven wave and ion

cyclotron
l wave
As the frequency gets lower, the ions participate more and more.
At frequencies near or below i , the ions play a major role and we
must use (211) and (212). In the vicinity of i , (211) gives the low-
frequency end of the whistler wave (slightly modified by the ions),
and (212) gives a new wave called the ion cyclotron wave.

Fig. 1 Fig. 2
2
1 2
e e
(right circularly polarized)  (right circularly polarized)
1
whistler
i i
(left circularly polarized) (left circularly polarized)
kz kz
54
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
When   i , we have
 1 
    (1   )   (1   )
1 1 1 (216)
e,i e,i e,i e,i e,i
 1 
 e,i  e,i (1  e,i )  e,i (1  e,i )
1 1 1 (217)

Sub. (216) and (217) into either (211) or (212), we get the same
results (i.e the two low-frequency branches merge into one):
 2pe  2pi
  kz c   (
2 2 2 2
 )  0, (218)
e2 i2
 2ppe  2ppi  2ppi
Since   , we neglect the electron term to get
e2 ei i2
 2pi 4 ni 0e2 mi2c2
  kz c  
2 2 2 2
  2  k z2 c 2   2 mi e2 B2  0 ((219))
i2
  0
4 ni 0mc
i
2
B02 55

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Defining a speed VA (called the Alfven speed) in terms of B0

and the ion mass density i  ni 0 mi :
B0
VA  , (220)
4i
we obtain from (219) the diepersion relation of the Alfven wave:

  k z c   c2  0 or  2 
2 2 2
2 2 k z2VA2
(221)
VA 1  VA2 / c 2

Fig. 1 Fig. 2
2
1 2
e e
(right circularly polarized)  ((right
g circularlyy polarized)
p )
1
whistler
hi l
i i
(left circularly polarized) (left circularly polarized)
kz k z56
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Physics of the Alfven wave : We first develop the useful concept of

magnetic pressure and magnetic tension. Using
the static law:   B  4c J (approximately
applicable at very low frequencies), we may
express the magnetic force density f (force per
unit volume) entirely in terms of the B-field: B-field
a solenoid
f  1c J  B  41 (  B)  B lines
a uniform
if
( a  b )  ( a   ) b + ( b   ) a + a  (   b ) + b  (   a ) electron beam
2
  8B  41 (B )B (222)
  
magnetic pressure magnetic tension force density,
force density as if a curved B-field line B-field
tended to become a straight line.
line B lines
li
In regions where J  0, we have f  0, i.e. the
uniform current 57
pressure and tension force densities cancel out.

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Return to the Alfven wave. Since the two lower branches merge
at   i , the left and right circularly polarized waves have the
same phase velocity. So a linearly polarized wave will remain linearly
polarized (no Faraday rotation). The figure below shows a linearly
polarized wave with B1 in the y -direction
direction and E1 in the x-direction.
direction
Since   i , e , the electron and ion behavior can be described
by their E1  B 0 drift motion (same speed and same direction). The
wave electric fi ld E1xe x cause both
l t i field b th the
th electrons
l t andd ions
i to
t drift
d ift in
i
the  y -direction, while the wave magnetic z total B-field
field B1 y e y bends the external B 0 in the
direction of the plasma drift (see figure).
A quantitative analysis (Nicholson, y
p. 163)) shows that the field lines and
p Linearly polarized
E1 = E1xe x ; B1  B1ye y
the plasma move together as if the field
lines were "frozen" to the plasma (or the plasma frozen to field lines). 58
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

On the other hand, when the magnetic field lines are bent, there
is a "tension
tension force density"
density on the plasma,
plasma which acts as a restoring
force to drive the plasma back so that the field lines (which are frozen
to the plasma) become straight. As the field lines are straightened, the
momentum of the plasma carries the field lines further back, back thus
bending the field lines again, in the opposite direction. The tension
force then acts again to start another oscillation cycle.
N t th
Note thatt we have
h assumedd k  k z e z ; hence,
h all
ll quantities
titi vary
only with the z -variable. This implies z total B-field
that, at a given time, the E1x e x  B 0
drift in the y -direction has the same
speed at all points along y. Thus, y
the drift motion will not compress/ Linearly polarized
E1 = E1xe x ; B1  B1ye y
decompress the plasma to produce
a density variation. The plasma remains uniform in the processes.
59

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Alternative derivation of the Alfven wave dispersion relation:

Th E1  B drifts
The d ift cause the
th plasma
l electron
l t andd ions
i to
t move in
i
the same direction with the same speed, hence generating no current.
However, there is another drift motion due to the time variation of E1 ,
which results in a polarization drift current given by (2.43) of Sec. 2.5:
 c2 E1 imc2
J p  m2  E1k eit ikz z
B0 t B02
imc2
or J pk  E1k , (223)
B02
where  m  ni 0 mi  ne 0 me is the plasma mass density.
density Note that the
polarization drift speed is much greater for the ions than electrons.
J pk plays a critical role in the Alfven wave. It generates the wave
magnetic
ti field
fi ld B1k andd hence
h the
th magnetic
ti tension
t i force
f density.
d it In I
fact, we may derive the dispersion relation based on (223).
60
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

imc2
Sub. J1k  J pk  E1k [(223)], E1k  E1kx e x , and k  k z e z
B02
into the field equation derived earlier:
k  (k  E )   E   4 i J
2
1k c2 1k c2 1k (125)
we obtain
4m 2
(k z2  2 
2
) E1kx e x  0,
c B02
which gives the same dispersion relation as (221):
 2  k z2c 2   c2  0
2 2
VA

61

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Waves propagating perpendicular to B0 e z (k z  0) :

Assume isotropic distribution, f  0  f  0 (v2  vz2 ), we have
f 0 f 0
 2 . Then, with k z  0, we obtain from (173)-(178)
vz2 v
ff
vz J n2 20
n   vz
Dxz   2   2p  2 v dv  dv (224)
 n k 0    z n 
dJn f 0
vz v J n
 i   2  2v dv  dv d (kv /  ) vz2
D yz  2 p 0    z n 
(225)
 n
f
vz J n2 20
n  v
Dzx   2   2p


 n

k 0
2v  dv 
dv z n 
(226)
dJn ff 0
vz v J n
 i   2  2v dv  dv d (kv /  ) v2
Dzy   2 p 0    z n 
(227)
 n 62
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Factoring out the vz -integrals from (224) and (225), we have

 f
D yz , Dxz   dvz vz 20
vz
 f 
 12  dvz v 0  12 f 0 (vz ) 0
z 
Factoring out the vz -integrals from (226) and (227), we have
 f 0
Dzx , Dzy   dvz vz 0
v2
because f 0 is an even function of vz .
Thus,, D yz  Dxz  Dzx  Dzy and ((147)) reduces to

 Dxx Dxy 0   E1kx 

  
 D yx D yy k 0
0   E1ky (228)
( 8)
 0  
 0 Dzz   E 
 1kz 
63

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

where
 2p n22 2 f 0
Dxx  1  2
 
 n   2p 0 2v dv  dvz
k2
Jn
v2
(229)
 n
 2p n v
  2 i
 
Dxy   D yx   n  0 2v dv  dvz k
 n
dJ f
 J n d (k v n/  ) 20 (230)
   v
k2c2 2  2p  
D yy  1  2     2v dv  dv z v
2
  n n   0 

dJ f
 [ d ( k v n/  ) ]2 20 (231)
   v
k2c2  2  2p  f 0
Dzz  1 
2    n  0 2v dv  dvz vz2 J n2
vz2
(232)
 n
In (229) (232) the argument of all Bessel functions is k v /  .
(229)-(232),
In the limit of a cold plasma (v , vz  0), we only need to keep the
lowest-order, non-vanishing terms in the sum over n. 64
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

(1)n x n
Using (183): lim J n ( x)  1 ( x ) n ; lim J  n ( x)   ( ) ,
x 0 n! 2 x 0 n! 2
we find that the lowest-order, non-vanishing terms for Dxx are the
n  1 terms. Thus, the sum over n in Dxx is
 2p   n22 2 f 0
n n  0 2v dv 

dvz
k2
Jn
v2
 2p  2p   2 k2v2 f 0
[  ] 2v dv  dvz 
    0  k2 42 v2
 2p   f
 2 2  dv2  dvz v2 20
  0  v
 integration by parts over v2
 p  2 
2
  2 2  dv  dvz f 0
  0 
 forf a cold
ld plasma:
l f 0  21v  (v ) (vz )
 p2 
 (233)
2 ( 2 2 ) 65

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Sub. (233) into (229), we obtain

 2p
Dxx  1   (234)
  
2 2

y the lowest-order, non-vanishing

Similarly, g terms for Dxy , D yx , and
D yy are also the n  1 terms, and we obtain
 2p 
Dxy   D yx  i  ((235))
  ( 
2 2 )


k2c2  2p
D yy  1 2   2 2 (236)
   
The lowest-order, non-vanishing term for Dzz is the n  0 term,
k2c2   2p
which gives Dzz  1 
2 
 
2 (237)

66
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

 Dxx Dxy   E1kx 

((228) g ves 
8) gives
 D D   E   0 and a d Dzz E1kz k =00 (238)
( 38)
 xy yy   1ky 
Using (234)-(237), we find from (238) the dispersion relations:
 2p  2p 
1  2 2  i
Dxx Dxy      (  )
2 2
  0 (239)
 Dxy D yy  2p  k2c2  2p
i 2 2 ) 1   2    2 
  (    2
k2c2   2p
and Dzz  1 
2 
 
2 0 (240)

Note: We have assumed k  k e x  k z e z

ex
[(145)]. Thus, with k z  0, (239) and (240)
apply to waves with k  k e x . This explains k B0
ez
why Dxx and D yy are unequal, although the
ey
system is isotropic in x and y 67

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Ordinary mode: (see Nicholson, Sec. 7.9 for a fluid treatment)

k2c2  2
Rewrite Dzz  1 
2 
 
p
2 0

(240)

Since  ppe   pi
p , we may neglect
the ion contribution and get  pe
k
 2  k2 c 2   2pe  0 (241)
This is
Thi i the
th dispersion
di i relation
l ti for
f the
th ordinary
di mode.
d The
Th ordinary
di
mode is a pure electromagnetic mode, which propagates in a direction
perpendicular to B0 e z , with the electric field parallel to B0 e z . The
dispersion relation (241) has the same form as that of electromagnetic
waves in an unmagnetized plasma [see (143)] ex
because the electron motion is along B0 e z and k B0
hence is unaffected by the external magnetic ez
field. ey E 1
68
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
General properties of modes in Dxx D yy  D xy  0: 2

 Dxx Dxy   E1kx 

Rewrite the field equations: 
 D D  E   0 (238)
 xy yy   1ky 
and the dispersion relation: Dxx D yy  D xy
2
0 (242)
(238) gives the following information about the modes in (242):
(1) E1 [=E1kx e x +E1kyy e y ] of these modes lies on the x-y plane.
Dxy
(2) Under (242), either equation in (238) gives E1kx   D E1ky
xx
From (239), we see that Dxx is real and ex
Dxy is imaginary
imaginary. Thus
Thus, E1kx and E1ky differ k
by a factor "i", implying E1x and E1 y E1 B0
ez
are 900 out of pphase while havingg
unequal amplitudes ( E1x  E1 y ), e y An ellipse on the x-y plane
i.e. E1 is elliptically polarized. traced by the E1 vector 69

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

(3) As shown in the figure, we have
ex
E1 =E1kx e +E
k x 1ky k ye
 k
 k  k e
 x
Thus, in general, these modes are E1 B0
ez
neither
i h electrostatic
l i (k  E1  0)) nor
electromagnetic (k  E1  0), except e y An ellipse on the x-y plane
at p
particular frequencies
q (such
( as traced by the E1 vector
  ) or wave numbers (such as k  ). Consider, for example,
the relative amplitude of E1kx and E1ky in the relation:
Dxy
E1kx   D E1ky
xx
If, for some  or k , we have Dxx  0. Then, E1  E1kx e x and
the mode becomes electrostatic. If Dxy  0 at some  or k , then
electrostatic If,
E1  E1ky e y and the mode becomes electromagnetic. In either case,
E1 also becomes linearly polarized. 70
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
Extraordinary mode: (see Nicholson, Sec. 7.9 for a fluid treatment)
Att high
g frequencies,
eque c es, we may ay neglect
eg ect tthee ion o cocontribution.
t but o . Then,
e ,
 2pe  2pee
1 2 2 i
 e  ( 2 e2 )
(239) can be written  0, (243)
 2pee k2c2  2pe
-i 1 2  2 2
 ( 2 e2 )   e
eB
where, as before, e  m c0 . (243) gives
e
 2pe  2pe  4pee2
(1   2 2 )( 1  k c2   2 
2 2
)  0 (244)
e
2
e  2 ( 2 e2 )
After some alg gebra, ((244)) can be written
( 2  2pe )2  2e2 ( 2  2pe )2  2e2 identical to 
k2 c 2  = ,   (245)
 2  2pe e2  2 UH 2
(7.178) 
where UH , called the upper hybrid frequency,
2
frequency is defined as

UH
2
  2pe  e2 (246)
71

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

( 2  2pe )2  2e2
Rewrite k c  2 2
(245)

 2 UH 2

This is the dispersion relation for the extraordinary mode. It has

two branches with the following limiting frequencies:
e  e B0 /( me c)
  2  UH
k  0     1    2pe  e  e ; k  
2  4

2
k c
Thus, as shown in the figure,
the frequency of the lower branch  ordinary mode
UH
2
goes from 1 to UH and the
g  pe
frequency of the upper branch goes 1 extraordinary mode
from 2 to infinity. Note that at k
  UH , we have Dxx  0. Hence E1  E1x e x [see (238)] and the
0 Hence,
wave is electrostatic (called upper hybrid resonance). As   ,
we have Dxy  0  E1  E1 y e y and the wave is electromagnetic. 72
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Magnetosonic Wave: (see Nicholson, Sec. 7.12 for a fluid treatment)

 2p  2p 
1  2 2  i
Dxx Dxy      (  )
2 2
Rewrite   0 (239)
D yyx D yy  2p  k2c2  2p
i 1 2   2 2
  (  )    
2 2

At very low frequencies ( 2  i2 ), ions play a major role and we

mustt retain
t i the
th ion
i terms
t in
i (239).
(239) Under
U d the
th condition:
diti  2  i2
 2p  2p  2pe  2pi
Dxx  1    1    1  
     e2 i2
2 2 2

 2p  2p k2c2  pi

2
k2c2 k2c2
D yy  1 2   2 2  1 2   2  1 2  2
        i
 2p   2p i [  pe   pii ]  0
2 2
Dxy   D yx  i   i  =  e i
  (  )  
2 2

= 2pi / i 73

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Thus, (239) gives Dxx D yy  0. Because Dxx  0, we have

k2c2  pi
2
D yy  1 2  2  0 (247)
 i
(247) has a form identical to (219) if k z in (219) is replaced with k .
Thus, the solution is simply (221) with k z changed to k :
k2VA2
 2
(248)
1  VA2 / c 2
where the Alfven speed VA is defined in (220) as
B0
VA  [ i  ni 0 mi ] (220)
4i
(248) gives the dispersion relation for the ex
g
magnetosonic wave. Because D yy  0,, it has an k B0
ez
electric field E1  E1e y . Hence, k  E1  0 and B1
ey
the wave is electromagnetic with B1  B1e z . E1 74
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

A physical picture: Like the low-frequency Alfven wave, particle

dynamics can be described by 2 types of drift motion. The E1e y  B0e z
drifts move the plasma in the  x-direction. Since E1 (  eik x ) also
varies with x, the drift motion will compress/decompress the plasma,
resulting in a density variation along x. On the other hand
hand, the wave
magnetic field B1e z , when superposed with B0e z will cause a similar
density variation of the magnetic field lines (see figure). The field
lines are again frozen to the plasma, similar to the Alfven wave.
However, the restoring force (thus the oscillation mechanism) is now
provided by the magnetic pressure force density:  B 2 / 8 [(222)].
As in the Alfven wave
wave, there is ex
also a polarization drift current in the
k B1
y -direction due to the time variation
of E1e y . This current generates the B0 ez
wave magnetic field B1e z , hence E1 B1
the magnetic pressure. ey 75

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

Alternative derivation of the magnetosonic wave dispersion relation:

Th polarization
The l i ti drift
d ift currentt is
i given
i by
b
 c2 E1 imc2
J p  m2  E1k eit ik x
B0  t B02

imc2
or J pk  E1k , (249)
B02
imc2
Sub J1k  J pk 
Sub. E1k , E1k  E1ky e y , and k  ke x into
B02
the field equation: k  (k  E1k )  c2 E1k   4c2i J1k
2
(125)
4m 2
( k2  2 
2
we obtain ) E1ky e y  0
c B02
 2  k2 c 2   c2  0,
2 2
or
VA
which gives the same dispersion relation as (248):
76
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
Asymptotic behavior of the magnetosonic wave dispersion relation:
The dispersion relation for the magnetosonic wave:
 2  k2VA2 /(1  VA2 / c 2 ) (245)
is valid under the condition  2  i2 . It breaks down as k  . To
find the behavior at k  , we assume  2  i2 and  2  e2 .
 2p  2pe  2pi
Then, Dxx  1    1   and from
   e2  2
2 2

 2p  2p 
Dxx Dxy 1  
  2 2
 i 
  ( 2 2 )
 0 (239)
D yx D yy  2p  k2c2  2p
i 1 2   2 2
  (  )    
2 2

we find that, as k   but  remains finite, we must have

 2p  2pe  2pi
Dxx  1   2 2  1  2  2  0, (247)
   e 
which implies E1  E1x e x [see (238)]. Hence, the wave is electrostatic.77

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

 2pe  2pi
Rewrite 1  0 (247)
e2 2
(247) gives
if  2pe  e2
 2pie2  2pie2
  2 2  2  e i  LH
2 2
, [for k  ] (248)
e  pe  pe
where LH  e i ((249))
is called the lower hybrid frequency. extraordinary mode
This justifies the assumption:
ordinary
y mode
 2  i2 and  2  e2 2
UH
we made in obtaining (248).  pe
y all the pperpendicular
Finally, p 1
LH magnetosonic wave
modes (k  B0 ) discussed so
far are summarized in the figure. k
78
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)
Discussion :
((i)) We have covered a number of the most familiar modes in a
uniform plasma in the framework of the kinetic theory. These modes
are treated in Ch. 7 of Nicholson by the fluid theory. However, some
other familiar uniform
uniform-plasma
plasma modes have been left out,
out for example,
example
the Bernstein modes (Krall & Trivelpiece, Sec. 8.12.3).
(ii) We have only considered waves either along or perpendicular
to B0e z . In practice
practice, waves can exist at any angle to B0e z (Nicholson,
(Nicholson
p. 165), with complicated expressions and mixed properties. The
general dispersion relation (149) can be the basis for a detailed study
of such uncovered uniform
uniform-plasma
plasma modes
modes.
(iii) There are also modes which are not contained in (149). For
example, an inhomogeneous (equilibrium) distribution in density or
t
temperaturet introduces
i t d new modes,
d suchh as drift
d ift waves (Nicholson,
(Ni h l
Sec. 7.14; Krall & Trivelpiece, Secs. 8-15 and 8.16). Plasmas in
79

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

some devices (e.g. tokamaks) are futher complicated by a complex

magnetic field configuration.
configuration Such plasmas are usually the subjects
of research papers.
(iv) Modes considered in Sec. 6.12 are for a cold plasma. Basic
properties
ti andd the
th underlying
d l i physics
h i are more clearly
l l exhibited
hibit d in
i
this limit. However, cold modes are stable because there is no free
energy to drive an instability. In the next topic, we will demonstrate
show
h how h a mode d can become
b unstable
bl in
i an anisotropic
i i plasma.
l
(v) The relativistic Valsov equation can be derived by the same
steps as in the derivation of the Vlasov equation in Sec. 6.1. For the
case we considered (E0  0, B 0  B0e z ), the relativistic factor  is
a constant of the motion in zero-order orbit equations. Hence, the
derivation of the relativistic dispersion
dispersion relation takes exactly the same
steps which lead to (171)-(182). In Special Topis I, we will derive
the relativistic Vlasov equation and consider a relativistic instability. 80
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

A Slow -Wave Instability on the Electron Cyclotron Wave :

The dispersion relation for "right
right circularly polarized waves"waves is
 f 0 kzvz ) kzvz  f  0
v
(1   vz2 3
 2  k z2 c 2  2   2p

2
v dv dvz  0 (203)

 kzvz 
At high frequencies, we may neglect the ions. Then, for a cold
 2pe e B0
plasma, (203) reduces to  2  k z2 c 2   
e
 0, [   m ] (214)
e ec

which gives the 2 "right circularly polarized" branches in the figure.

Below, we will show that the
"electron cyclotron wave" portion
of the lower branch can be 2
e
destabilized by an anisotropy in 
1 (right circularly polarized)
velocity distribution, resulting i
in a velocity-space instability. (left circularly polarized)
k z 81

6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

For an anisotropic plasma, we cannot use (214), but must go back

to (203)
(203). Again
Again, assume high frequency and neglect the ions. ions (203)
 f e 0 (1 kzvz ) kzvz  f e 0
v2   vz2 3
gives  2  k z2 c 2  2 2pe
  kzvz e

v dv dvz  0 (250)

The integral I can be written I
 f e0 kz  f e0

(1 kzvz )  vz
 
v
I  12 
v2 dv dvz  12 v3 dv dv
 kzvz e  kzvz e   z
 integration by parts
f e0 (1 kzvz ) kz2
 
f
 e0
 v dv dv  1 v3 dv dv
 kz vz e   z 2 ( kz vz e )2   z
Now, assume f e0  1  (v  v0 ) (vz ), [see (104)] (251)
2 v
which represents a uniform distribution of electrons in random-phase
gyrational motion, with v  v0 and vz  0 for all electrons. 82
 6.12 Linear Vlasov Waves in Magnetized Plasma (continued)

kz2v20
Then, I  1 [ 1  1 ] (252)
2  
e 2 ( e )2
Sub. (252) into (250), we obtain the dispersion relation:
  1 kz v0 ]  0,,
2 2
 2  k z2 c 2   2pe [  
e 2 ( 
(253)
( )
e )2
which reduces to (214) as v0  0. frequency
The two right circularly polarized branches
are plotted for  pe  10e and v0  0.2c. The
top figure plots r (wave frequency) vs k z .
The upp pper branch is a fast wave (r / k z  c), )
while the lower branch is a slow wave (r / k z growth rate
 c). The bottom figure plots i (growth rate)
vs k z . We see that the slow wave is destabilized
by the gyrational particles, which feed energy
to the wave through cyclotron resonances. 83