You are on page 1of 5

Advanced Topics in Electrodynamics

Seminar Topic: Vlasov paper


Matthias Baldursson Harksen

We will in this discussion be working in phase space. Which is a 6-dimensional space consisting of the
postition components of particles, ~x = (x, y, z) as well as their momentum components, ~v = (vx , vy , vz ). A
point in phase space is denoted by (~x, ~v ) = (x, y, z, vx , vy , vz ). We have a great interest in the number of
particles, dN , contained in a small infinitesimal phase space volume d3 ~xd3~v = dxdydzdvx dvy dvz and hence
we define the phase space particle density:
dN
f (~x, ~v , t) =
d3 ~xd3~v
This gives the number of particles in a given region of space which are contained in a region with particles
of a specific momentum. Then the corresponding particle density is found by integrating over all possible
momenta, i.e.
Z
dN
n(~x, t) = f (~x, ~v , t)d3~v = 3 .
d ~x
And the total number of particles in some region is found by integating over that volume region.

During our last lecture we had shown that as a consequence of Liouville’s theorem we have that phase space
flow is conserved and thus we found that the phase space particle density obeys:
df
= 0.
dt
In the absence of collisions. This is called the collisionless Boltzmann equation or the Vlasov equation. But
as f = f (~x, ~v , t) we find by making use of the chain rule that:

df ∂f dx ∂f dy ∂f dz ∂f dz ∂f dvx ∂f dvy ∂f dvz ∂f dvz ∂f


= + + + + + + + +
dt ∂t dt ∂x dt ∂y dt ∂z dt ∂z dt ∂vx dt ∂vy dt ∂vz dt ∂vz
∂f
= + ~v · ∇x f + ~a · ∇v f = 0.
∂t
For our plasma we have that the acceleration is due to the Lorentz-force and hence:
q ~ ~

~a = E + ~v × B
m
If we then introduce the following notation for the different species of particles. Let fs denote for species s the
distribution function for the number of particles within an infinitesimal phase space volume. We will denote
the disribution function for electrons by fe , the distribution function for ions by fp and the distribution
function for neutral particles by fn . The discussion in the Vlasov paper aims to show why we may skip
contributions from the collisions of particles. Incorprating collisions into our model the Vlasov equation
takes the form of:
 
df ∂f
=
dt ∂t collisions

1
In our case we must therefore consider all possible types of collissions between the particles. I.e. an electron
colliding with another elctron or an electron colliding with an ion or an electron colliding with a neutral
particle. Hence we find that the system of equations which we are interested in is given by:

      
∂fe ~ e ~ ~

~ ∂fe ∂fe ∂fe

 + ~v · ∇x fe + E + ~v × B · ∇v fe = + +
∂t me ∂t ee ∂t ep ∂t en









 ∂f      
~ x fp + Ze E ~ v fp = ∂fp ∂fp ∂fp
 
p ~ + ~v × B
~ ·∇
+ ~v · ∇ + +

 ∂t mp ∂t pe ∂t pp ∂t pn




      

 ∂fn ~ ∂fp ∂fp ∂fp

 + ~
v · ∇ x f n = + +
∂t ∂t ne ∂t np ∂t nn

Which we need to supplement with the Maxwell equations:



∇
 ~ ·E
~ = ρ,
 ε0
∇
 ~ ·B
~ = 0,
 ~ ×E
∇ ~ = − ∂ B~ ,

 ∂t

∇~ ×B~ = µ J~ + µ ε ~
∂E
0 0 0 ∂t .

where the density is given by:


Z Z
ρ = −e fe d3 v + Ze fp d3 v.

And the current is given by:


Z Z
~j = −e 3
~v fe d v + Ze ~v fp d3 v.

Our aim is to show under which assumptions we may neglect the contributions due to collisions. Therefore
we take a closer look at the collision term. Consider the collisions from a particle of species 1 colliding with
a particle of species 2. To simplify the analysis we should imagine that a particle of type 1 is traveling with
the relative velocity:
~vr = ~v1 − ~v2
In order to measure the total contribution due to the collisions we need to estimate the total number of
collisions during some time ∆t. The particle of species 1 will sweep out a cylinder of volume σvr ∆t during
this time where σ is the cross section of the interaction between the particles. Hence the total number of
collisions that one particle of speices 1 has during the time ∆t depends upon the number density of particles
of species 2 contained in this volume element:
Z
number of collisions = n2 σvr ∆t = f2 σvr ∆td3~v2

But this holds equally for all the particles of species 1 and hence we find that we should add a term of:
Z
f1 f2 σvr d3~v2

However, as is the case in classical mechancis when we have a ball bouncing of a rigid wall the force acting on
the ball during its contact with the wall is given by ∆p
∆t . Therefore we also need to consider the distributions
functions after the collision. We therefore conclude that in the equations of motion for a particle of species
1 where we are including terms with the collision with particles of species 2 we have:
  Z Z
∂f 0
= f10 f20 σvr0 d3~v2 − f1 f2 σvr d3~v2
∂t 12

2
It is more convenient to relate the cross section to the impact parameter. We have then that:
dσ = bdbdϕ

R σb denotes
where R σ the impact parameter and ϕ is the azimuthal angle. Therefore we find by noting that
σ = 0 dσ = 0 bdbdϕ the following expression:
  Z Z
∂f 0 0 0 3 0
= 2π f1 f2 vr b db d ~v2 − 2π f1 f2 vr b db d3~v2
∂t 12
And therefore we focus on understanding various impact parameters better to understand the frequency of
collisions. Consider to start with the hard sphere of radius R and a small projectile colliding with it with
some angle of deflection θ.

Figure 1: The impact parameter for a small projectile colliding with a large target of radius R.

By making use of the geometry we find that  the impact parameter is given by b = R sin α where π = θ + 2α
and hence we conclude that b = R cos θ2 is the impact parameter for the collision between two elastic
spheres. We could actually generalize this so that R would denote the sum of the radii of the two particles.

Figure 2: A couple of geometrical findings.

As we are only interested in the cross section we note that for the spheres:
R2
         
θ R θ θ θ
bdb = R cos − sin =− cos sin .
2 2 2 2 2 2

3
However, we would most likely be dealing with the Rutherford scattering of the Coulomb force. This cross
section is not easily derived. Here is my simplistic derivation (which looses a factor of 1/2 on the angle).
The impulse in the y-direction is given by:
" #+∞
Z +∞ Z +∞
kqQ kqQb t 2kqQ
∆py = 2 · sin ϕdt = 3/2
dt = kqQb p =
b2 b2 v02 t2 bv0

−∞ b2 + (v0 t)
2 −∞ (b2 + (v0 t)2 ) + −∞

Similarily we find for the impulse in the x-direction that:


Z +∞ Z +∞
kQq kQq
∆px = 2 · cos ϕdt = 3/2
· v0 tdt = 0
2 2
−∞ (b + (v0 t) ) −∞ (b2 + (v0 t)2 )

which vanishes since the integrand is an odd function in t. Therefore we conclude that:
 
2kqQ
vy ∆py bv0 2kQq
tan θ = = = =
vx p0 + ∆px mv0 bmv02

From which we find the (wrong) result:

2kQq
b= cot(θ)
mv02

A more formal derivation gives:


 
Qq θ
b= cot .
2πε0 mv02 2

Then we see that:


 2 θ

1 Qq cos 2
b db = − .
2 2πε0 mv02 sin3 ( θ2 )

At this point we should argue that for sufficiently large angles θ ∈ [ π2 , π] then sin3 ( θ2 ) ≈ 1. More specifically:

1 hπ i
∈ [1, 4], for θ ∈ ,π .
sin4 ( θ2 ) 2

For these angles it is a sufficiently good approximation to consider this as a collision with a solid sphere of
radius R where:
 
Qq
R=
2πε0 mv02

These are also those transits which I would be more worried about since these correspond to the particles
coming in close contact with each other. Next we consider the small angles for particles which feel the
Coulomb collision far away. By means of dimensional analysis Lev Landau argues that the mean free path
in this framework is given by:
2 2
(4πε0 ) (kB T ) bmax
`∼
= , where L= .
e4 Ln bmin
Where we have fixed the dimensionality of Landau’s result so that:
 2
s2 C 2
2
[ε0 ] [kB T ]
2
kg m3 J2
[`] = 4 = = m.
[e] [L][n] C 4 m13

4
2 2
We might even fix the result to ` =∼ (4πε02) 2(kB T ) but Landau clearly assumes (as does Vlasov) collisions
q Q Ln
from electrons with electrons. But we have then that:
nσ` = 1
i.e. that the number of particles inside the cylinder of cross section σ and length ` is 1. Since our cross-section
is supposed to correspond to a sphere of radius R we have σ = πR2 and thus:
r
1 2 1 −1/2 L Qq
nσ` = 1 =⇒ σ = =⇒ πR = = R = (πn`) = · .
n` n` π 4πε0 kB T
and using the definition of the velocity 32 kB T = 21 mv 2 we find that:
r r   r
9L Qq 1 9L Qq 9L
R= = = R0 = αR0
π 4πε0 mv02 2 π 2πε0 mv0 2 4π
Where R0 = 2πεQq
0 mv0
θ
2 is the radius of the cross section for large angles θ ∈ [ 2 , θ]. Next we need to show how

we may estimate α. The maximum value of the impact parameter, b, is obtained at the average distance
between the particles hence we have that:
bmax = n−1/3 .
While the smallest can be estimated as:
Qq
bmin =
2πε0 mv02
Thus we find that:
   
bmax 6πε0 kB T
L = ln = ln .
bmin n1/3 Qq
The angular eigenfrequencies of the plasma oscillations are given by:
 2 1/2
e n
ωpe =
ε0 m
While the angular frequency of collisions is given by:

ωcollisions =
τ
`
Where τ is the time between collisions. But we have that vτ = ` and hence τ = v and we find that:
2π 2πv
ωcollisions = = = 2πvnσ = 2π 2 vnα2 R02
τ `
In the article Vlasov considers the conditions of the ionosphere where the particle density is
nionosphere ∼ 106 cm−3 = 1012 m−3
and the temperature is T = 300 K. Then we find for electrons that:
L ∼ 7.898 =⇒ α = 2.378.
Which means that our assumptions are mostly okay. Finally we find that:
ωcollisions = 1.798 × 104 rad/s.
While
ωpe = 5.642 × 107 rad/s.
And therefore:
ωcollisions
= 0.032 %
ωpe
Which means that the plasma oscillates much faster than the frequency of collisions. Therefore we may
neglect this contribution in that case.

You might also like