Oxford MMathPhys Lectures: Plasma Kinetics and MHD 21

(a) (b)
Figure 3. Lev Landau (1908-1968), great Soviet physicist, quintessential theoretician, author
of the Book, cult figure. It is a minor feature of his scientific biography that he wrote the two
most important plasma-physics papers of all time (Landau 1936, 1946). He also got a Nobel
Prize (1962), but not for plasma physics. (a) Cartoon by A. A. Yuzefovich (from Landau &
Lifshitz 1976); the caption says “[And] Dau spake. . . ” (. . . unto the students, also depicted).
(b) Landau’s mugshot from NKVD prison (1938), where he ended up for seditious talk and from
whence he was released in 1939 after Peter Kapitsa’s personal appeal to Stalin.

Note that because the nonlinear term couples perturbations at different k’s (scales),
this theory will lead to multi-scale (usually, power-law) fluctuation spectra.

2.4.4. Strong-Turbulence Theory

If perturbations manage to grow to a level at which
tnl ∼ ω −1 , (2.17)
we are a facing strong turbulence. This is actually what mostly happens (including in
WT systems, where turbulence often transitions into the strong regime at small enough,
or large enough, scales). Theory of such regimes tends to be of phenomenological/scaling
kind, often in the spirit of the classic Kolmogorov (1941) theory of hydrodynamic
turbulence. No one really knows how to move very far beyond this sort of approach—and
not for lack of trying. I will return to this topic in §§8.6 and 12.2.

3. Linear Theory: Waves, Landau Damping and Kinetic Instablities

Enough idle chatter, let us calculate! In this section, we are concerned with the
linearised Vlasov–Poisson system, (2.13) and (2.9):
∂δfkα qα ∂f0α
+ ik · v δfkα = ϕk ik · , (3.1)
∂t mα ∂v
Z
4π X
ϕk = 2 qα dv δfkα . (3.2)
k α

For compactness of notation, I will drop both the species index α and the wave number
k in the subscripts, unless they are necessary for understanding.
We will discover that electrostatic perturbations in a plasma described by (3.1) and
(3.2) oscillate, can pass their energy to particles (damp) or even grow, sucking energy
from the particles. We will also discover that it is useful to know some complex analysis.

3.1. Initial-Value Problem

We shall follow Landau’s original paper (Landau 1946) in considering an initial-value
problem—because, as we will see, perturbations can be damped or grow, so it is not
22 A. A. Schekochihin

Figure 4. Layout of the complex-p plane: δfˆ(p) is analytic for Re p > σ. At Re p < σ, δfˆ(p)
may have singularities (poles).

appropriate to think of them over t ∈ [−∞, +∞] (and—NB!!!—the damped perturbations

are not pure eignenmodes; see §5.3). So we look for δf (v, t) satisfying (3.1) with the initial
condition
δf (v, t = 0) = g(v). (3.3)
It is, therefore, appropriate to use the Laplace transform to solve (3.1):
Z ∞
δfˆ(p) = dt e−pt δf (t) . (3.4)
0

It is a mathematical certainty that if there exists a real number σ > 0 such that
|δf (t)| < eσt as t → ∞, (3.5)
then the integral (3.4) exists (i.e., is finite) for all values of p such that Re p > σ. The
inverse Laplace transform, giving us back our distribution function as a function of time,
is then
Z i∞+σ
1
δf (t) = dp ept δfˆ(p), (3.6)
2πi −i∞+σ
where the integral in the complex plane is along a straight line parallel to the imaginary
axis and intersecting the real axis at Re p = σ (Fig. 4).
Since we expect to be able to recover our desired time-dependent function δf (v, t)
from its Laplace transform, it is worth knowing the latter. To find it, we Laplace-
transform (3.1):
Z ∞ Z ∞
−pt ∂δf
 −pt ∞
l.h.s. = dt e = e δf 0 + p dt e−pt δf = −g + p δfˆ,
0 ∂t 0
q ∂f0
r.h.s. = −ik · v δfˆ + ϕ̂ ik · . (3.7)
m ∂v
Equating these two expressions, we find the solution:
 
1 q ∂f0
δfˆ(p) = i ϕ̂(p) k · +g . (3.8)
p + ik · v m ∂v
 Oxford MMathPhys Lectures: Plasma Kinetics and MHD 23

Figure 5. New contour for the inverse Laplace transform.

The Laplace transform of the potential, ϕ̂(p), itself depends on δfˆ via (3.2):
Z ∞ Z
4π X
ϕ̂(p) = dt e−pt
ϕ(t) = 2 qα dv δfˆα (p)
0 k α
Z  
4π X 1 qα ∂f0α
= 2 qα dv i ϕ̂(p) k · + gα . (3.9)
k α p + ik · v mα ∂v
This is an algebraic equation for ϕ̂(p). Collecting terms, we get
" #
X 4πq 2 Z 1 ∂f0α 4π X
Z
gα
α
1− 2m
i dv k· ϕ̂(p) = 2 qα dv . (3.10)
α
k α p + ik · v ∂v k α
p + ik · v
| {z }
≡ (p, k)
The prefactor in the left-hand side, which I denote (p, k), is called the dielectric function,
because it encodes all the self-consistent charge-density perturbations that plasma sets
up in response to an electric field. This is going to be an important function, so let us
write it out beautifully:
X ωpα2 Z
i 1 ∂f0α
(p, k) = 1 − 2
dv k· , (3.11)
α
k nα p + ik · v ∂v

where the plasma frequency of species α is defined by [cf. (2.2)]

2 4πqα2 nα
ωpα = . (3.12)
mα
The solution of (3.10) is
Z
4π X gα
ϕ̂(p) = 2
qα dv . (3.13)
k (p, k) α p + ik · v
To calculate ϕ(t), we need to inverse-Laplace-transform ϕ̂: similarly to (3.6),
Z i∞+σ
1
ϕ(t) = dp ept ϕ̂(p). (3.14)
2πi −i∞+σ
How do we do this integral? Recall that δfˆ and, therefore, ϕ̂ only exists (i.e., is finite) for
24 A. A. Schekochihin
Re p > σ, whereas at Re p < σ, it can have singularities, i.e., poles—let us call them pi ,
indexed by i. If we analytically continue ϕ̂(p) everywhere to Re p < σ except those poles,
the result must have the form
X ci
ϕ̂(p) = + A(p), (3.15)
i
p − pi

where ci are some coefficients (residues) and A(p) is the analytic part of the solution. The
integration contour in (3.14) can be shifted to Re p → −∞ but with the proviso that it
cannot cross the poles, as shown in Fig. 5 (this is proven by making a closed loop out of
the old and the new contours, joining them at ±i∞, and noting that this loop encloses no
poles). Then the contributions to the integral from the vertical segments of the contour
are exponentially small,11 the contributions from the segments leading towards and away
from the poles cancel, and the contributions from the circles around the poles can, by
Cauchy’s formula, be expressed in terms of the poles and residues:
X
ϕ(t) = c i e pi t . (3.16)
i

Thus, in the long-time limit, perturbations of the potential will evolve ∝ epi t , where pi
are poles of ϕ̂(p). In general, pi = −iωi + γi , where ωi is a real frequency (giving wave-
like behaviour of perturbations), γi < 0 represents damping and γi > 0 growth of the
perturbations (instability).
Note that we need not be particularly interested in what ci ’s are because, if we set
up an initial perturbation with a given k and then wait long enough, only the fastest-
growing or, failing growth, the slowest-damped mode will survive, with all others having
exponentially small amplitudes. Thus, a typical outcome of the linear theory is ϕ(t)
oscillating at some frequency and growing or decaying at some unique rate. Since this
is a solution of a linear equation, the prefactor in front of the exponential can be scaled
arbitrarily and so does not matter.
Going back to (3.13), we realise that the poles of ϕ̂(p) are zeros of the dielectric
function:
(pi , k) = 0 ⇒ pi = pi (k) = −iωi (k) + γi (k). (3.17)
To find the wave frequencies ωi and the damping/growth rates γi , we must solve this
equation, which is called the plasma dispersion relation.

3.2. Calculating the Dielectric Function: the “Landau Prescription”

In order to be able to solve (p, k) = 0, we must learn how to calculate (p, k) for any
given p and k. Before I wrote (3.15), I said that ϕ̂, given by (3.13), had to be analytically
continued to the entire complex plane from the area where its analyticity was guaranteed
(Re p > σ), but I did not explain how this was to be done. In order to do it, we must

11
They are exponentially small in time as t → ∞ because the integrand of the inverse Laplace
transform (3.14) contains a factor of eRe pt , which decays faster than any of the “modes” in (3.16).
If ϕ̂(p) does not grow too fast at large p, the integral along the vertical part of the contour may
also vanish at any finite t, but that is not guaranteed in general: indeed, looking ahead to the
explicit expression (3.27) for ϕ̂(p), with the Landau prescription for analytic continuation to
Re p < 0 analogous to (3.20), we see that ϕ̂(p) will contain a term ∝ Gα (ip/k), which can be
large at large Re p, e.g., if Gα (vz ) is a Maxwellian. Note also that we need the (wildly oscillating
in time) integral of ept ϕ̂(p) over the horizontal segments with Im p → ±∞ to vanish. This is fine
provided ϕ̂(±i∞) = 0, which is usually OK.
 Oxford MMathPhys Lectures: Plasma Kinetics and MHD 25

(a) Re p > 0 (b) Re p = 0 (c) Re p < 0

Figure 6. The Landau prescription for the contour of integration in (3.18).

Figure 7. Proof of Landau’s prescription [see (3.19)].

learn how to calculate the velocity integral in (3.11)—if we want (p, k) and, therefore,
its zeros pi —and also how to calculate the similar integral in (3.13) containing gα .
First of all, let us turn these integrals into a 1D form. Given k, we can always choose
the z axis to be along k.12 Then
Z Z Z Z
1 ∂f0 1 ∂
dv k· = dvz k dvx dvy f0 (vx , vy , vz )
p + ik · v ∂v p + ikvz ∂vz
| {z }
≡ F (vz )
Z +∞ 0
F (vz )
= −i dvz . (3.18)
−∞ vz − ip/k
Assuming, reasonably, that F 0 (vz ) is a nice (analytic) function everywhere, we conclude
that the integrand in (3.18) has one pole, vz = ip/k. When Re p > σ > 0, this pole is
harmless because, in the complex plane associated with the vz variable, it lies above the
integration contour, which is the real axis, vz ∈ (−∞, +∞). We can think of analytically
continuing the above integral to Re p < σ as moving the pole vz = ip/k down, towards
and below the real axis. As long as Re p > 0, this can be done with impunity, in the
sense that the pole stays above the integration contour, and so the analytic continuation
is simply the same integral (3.18), still along the real axis. However, if the pole moves so
far down that Re p = 0 or Re p < 0, we must deform the contour of integration in such
a way as to keep the pole always above it, as shown in Fig. 6. This is called the Landau
prescription and the contour thus deformed is called the Landau contour, CL .
Let me prove that this is indeed an analytic continuation, i.e., that the integral (3.18),
12
NB: This means that in what follows, k > 0 by definition.
26 A. A. Schekochihin
adjusted to be along CL , is an analytic function for all values of p. Let us cut the
Landau contour at vz = ±R and close it in the upper half-plane with a semicircle CR
of radius R > σ/k (Fig. 7). Then, with integration running along the truncated CL and
counterclockwise along CR , we get, by Cauchy’s formula,
F 0 (vz ) F 0 (vz )
Z Z  
0 ip
dvz + dvz = 2πi F . (3.19)
CL vz − ip/k CR vz − ip/k k
Since analyticity is guaranteed for Re p > σ, the integral along CR is analytic. The right-
hand side is also analytic, by assumption. Therefore, the integral along CL is analytic—
this is the integral along the Landau contour if we take R → ∞. Q.e.d.
With the Landau prescription, our integral is calculated as follows:
 Z +∞
F 0 (vz )

 dvz if Re p > 0,
vz − ip/k

−∞






Z 0
F (vz )  Z +∞

F 0 (vz )
 
0 ip
dvz = P dvz + iπF if Re p = 0, (3.20)
CL vz − ip/k  −∞ vz − ip/k k




 Z +∞
F 0 (vz )
 

 0 ip

 dvz + i2πF if Re p < 0,
−∞ vz − ip/k k

where the integrals are again over the real axis and the imaginary bits come from the
contour making a half (when Re p = 0) or a full (when Re p < 0) circle around the pole.
In the case of Re p = 0, or ip = ω, the integral along the real axis is formally divergent
and so we take its principal value, defined as
Z +∞ "Z Z +∞ #
ω/k−ε
F 0 (vz ) F 0 (vz )
P dvz = lim + dvz . (3.21)
−∞ vz − ω/k ε→0 −∞ ω/k+ε vz − ω/k

The difference between (3.21) and the usual Lebesgue definition of an integral is that the latter
would be
" Z +∞ #
F 0 (vz ) F 0 (vz )
Z +∞ Z ω/k−ε1
dvz = lim + lim dvz , (3.22)
−∞ vz − ω/k ε1 →0 −∞ ε2 →0 ω/k+ε
2
vz − ω/k

and this, with, in general, ε1 6= ε2 , diverges logarithmically, whereas in (3.21), the divergences
neatly cancel.
The Re p = 0 case in (3.20),
F 0 (vz ) F 0 (vz )
Z Z +∞ ω
dvz =P dvz + iπF 0 , (3.23)
CL vz − ω/k −∞ vz − ω/k k
which tends to be of most use in analytical theory, is a particular instance of Plemelj’s formula:
for a real ζ and a well-behaved function f (no poles on or near the real axis),
Z +∞ Z +∞
f (x) f (x)
lim dx =P dx ± iπf (ζ), (3.24)
ε→+0 −∞ x − ζ ∓ iε −∞ x −ζ
also sometimes written as
1 1
lim =P ± iπδ(x − ζ), (3.25)
ε→+0 x − ζ ∓ iε x−ζ

Finally, armed with Landau’s prescription, we are ready to calculate. The dielectric
 Oxford MMathPhys Lectures: Plasma Kinetics and MHD 27
function (3.11) becomes
2
X ωpα Fα0 (vz )
Z
1
(p, k) = 1 − 2
dvz , (3.26)
α
k nα CL vz − ip/k

and, analogously, our Laplace-transformed solution (3.13) becomes

Z
4πi X Gα (vz )
ϕ̂(p) = − 3 qα dvz , (3.27)
k (p, k) α CL vz − ip/k
R R
where Gα (vz ) = dvx dvy gα (vx , vy , vz ).

3.3. Solving the Dispersion Relation: the Limit of Slow Damping/Growth

A particularly analytically solvable and physically interesting case is one in which, for
p = −iω + γ, γ  ω and γ  kvthα , i.e., the case of the damping or growth time of the
waves being longer than either their period or the time particles take to cross them. In
this limit, the dispersion relation (3.17) is
∂
(p, k) ≈ (−iω, k) + iγ (−iω, k) = 0. (3.28)
∂ω
Setting the real part of (3.28) to zero gives the equation for the real frequency:

Re (−iω, k) = 0 . (3.29)

Setting the imaginary part of (3.28) to zero gives us the damping/growth rate in terms
of the real frequency:
 −1
∂
γ = −Im (−iω, k) Re (−iω, k) . (3.30)
∂ω

Thus, we now only need (p, k) with p = −iω. Using (3.23), we get
2
X ωpα Fα0 (vz )
Z
1
Re  = 1 − P dv z , (3.31)
α
k 2 nα vz − ω/k
X ωpα 2
π 0 ω
Im  = − F . (3.32)
α
k 2 nα α k

Let us consider a two-species plasma, consisting of electrons and a single species of

ions. There will be two interesting limits:
p
• “High-frequency” electron waves: ω  kvthe , where vthe = 2Te /me is the “thermal
speed” of the electrons;13 this limit will give us Langmuir waves (§3.4), slowly damped
or growing (§3.5).
• “Low-frequency” ion waves: a particularly tractable limitpwill be that of “hot”
electrons and “cold” ions, viz., kvthe  ω  kvthi , where vthi = 2Ti /mi is the “thermal
speed” of the ions; this limit will give us the sound (“ion-acoustic waves”; §3.8), which
also can be damped or growing (§3.9).
13
This is a standard well-defined quantity for a Maxwellian equilibrium distribution
√
Fe (vz ) = (ne / π vthe ) exp(−vz2 /vthe
2
), but if we wish to consider a non-Maxwellian Fe , let
vthe be some typical speed characterising the width of the equilibrium distribution, defined
by, e.g., (3.36).
28 A. A. Schekochihin
3.4. Langmuir Waves
Consider the limit
ω
 vthe , (3.33)
k
i.e., the phase velocity of the waves is much greater than the typical velocity of a particle
from the “thermal bulk” of the distribution. This means that in (3.31), we can expand in
vz ∼ vthe being small compared to ω/k (higher values of vz are cut off by the “thermal”
fall-off of the equilibrium distribution function). Note that ω  kvthe also implies ω 
kvthi because
r
vthi Ti me
= 1 (3.34)
vthe Te mi
as long as Ti /Te is not huge.14 Thus, (3.31) becomes
" 2  3 #
X ωpα 2 Z 
1 k 0 kvz kvz kvz
Re  = 1 + P dvz Fα (vz ) 1 + + + + ...
α
k 2 nα ω ω ω ω
X ωpα 2  Z Z
1 0 k 1
=1+ dvz Fα (vz ) − dvz Fα (vz )
α
kω nα ω nα
| {z } | {z }
=0 =1
2
k3 1
Z Z 
k 1
−2 dvz vz Fα (vz ) −3 3 dvz vz2 Fα (vz ) + . . .
ω 2 nα ω nα
| {z } | {z }
=0 2
= vthα /2
2
ωpα 3 k 2 vthα
2
X  
=1− 1+ + ... , (3.35)
α
ω2 2 ω 2

where we have integrated by parts everywhere, assumed that there are no mean flows,
hvz i = 0, and, in the last term, used
2
vthα
hvz2 i = , (3.36)
2
which is indeed the case for a Maxwellian Fα or, if Fα is not a Maxwellian, can be viewed
as the definition of vthα .
The ion contribution to (3.35) is small because
2
ωpi Zme
2
=  1, (3.37)
ωpe mi
so ions do not participate in this dynamics at all. Therefore, to lowest order, the dispersion
relation (3.29) becomes
2
ωpe 4πe2 ne
Re  ≈ 1 − =0 ⇒ ω 2 = ωpe
2
= . (3.38)
ω2 me

This is the Tonks & Langmuir (1929) dispersion relation for what is known as Langmuir,
or plasma, oscillations. This is the formal derivation of the result that we already had,
on less mathematically rigorous, physical grounds, in §2.1.

14
p
For hydrogen plasma, mi /me ≈ 42, the answer to the Ultimate Question of Life, Universe
and Everything (Adams 1979).
 Oxford MMathPhys Lectures: Plasma Kinetics and MHD 29
We can do a little better if we retain the (small) k-dependent term in (3.35):
2 
ωpe 3 k 2 vthe
2

Re  ≈ 1 − 2 1 + 2
=0 ⇒ ω 2 ≈ ωpe
2
(1 + 3k 2 λ2De ) , (3.39)
ω 2 ω
| {z }
use
ω 2 ≈ ωpe
2

√ p
where λDe = vthe / 2 ωpe = Te /4πe2 ne is the “electron Debye length” [cf. (1.6)].
Equation (3.39) is the Bohm & Gross (1949a) dispersion relation, describing an upgrade
of the Langmuir oscillations to dispersive Langmuir waves, which have a non-zero group
velocity (this effect is due to the electron pressure perturbation joining the electric field
in providing the restoring force for the waves: see Exercise 3.1).
Note that all this is only valid for ω  kvthe , which we now see is equivalent to
kλDe  1 (3.40)
(the wave length of the perturbation is long compared to the Debye length).

Exercise 3.1. Langmuir hydrodynamics.15 Starting from the linearised kinetic equation
for electrons and ignoring perturbations of the ion distribution function completely, work out
the fluid equations for electrons (i.e., the evolution equations for the electron density ne and
velocity ue ) and show that you can recover the Langmuir waves (3.39) if you assume that
electrons behave as a 1D adiabatic fluid (i.e., have the equation of state pe n−γ
e = const with
γ = 3). You can prove that they indeed do this by calculating their density and pressure directly
from the Landau solution for the perturbed distribution function (see §§5.3 and 5.6), ignoring
resonant particles. The “hydrodynamic” description of Langmuir waves will reappear in §8.

3.5. Landau Damping and Kinetic Instabilities

Now let us calculate the damping rate of Langmuir waves using (3.30):
2 2
∂Re  2ωpe ωpe π 0 ω  π ω3 1 0  ω 
≈ 3 , Im  ≈ − F ⇒ γ≈ F , (3.41)
∂ω ω k 2 ne e k 2 k 2 ne e k

where ω is given by (3.39) (if Fe is Maxwellian, the dispersive correction in ω 2 , when

substituted into the exponential, will change γ by a factor of order unity, hence it is
worth keeping). Provided ωF 0 (ω/k) < 0 (as would be the case, e.g., for any distribution
monotonically decreasing with |vz |; see Fig. 8a), γ < 0 and so this is indeed a damping
rate, the celebrated Landau damping (Landau 1946; it was confirmed experimentally two
decades later, by Malmberg & Wharton 1964).
The same theory also describes a class of kinetic instabilities: if ωF 0 (ω/k) > 0, then
γ > 0, so perturbations grow exponentially with time. An iconic example is the bump-
on-tail instability (Fig. 8b), which arises when a high-energy (vz  vthe ) electron beam
is injected into a plasma16 and whose quasilinear saturation we will study in great detail
in §6.
We see that the damping or growth of plasma waves occur via their interaction with
the particles whose velocities coincide with the phase velocity of the wave (“Landau
resonance”). Because such particles are moving in phase with the wave, its electric field
15
This is based on the 2017 exam question.
16
Here we are dealing with the case of a “warm beam” (meaning that it has a finite width). It
turns out that there exists also another instability, leading to growth of perturbations with ω/k
to the right of the bump’s peak, due to a different, “fluid” kind of resonance and possible even
for “cold beams” (i.e., beams of particles that all have the same velocity): see §3.7.
30 A. A. Schekochihin

(a) ωF 0 (ω/k) < 0: Landau damping (b) ωF 0 (ω/k) > 0: instability

Figure 8. The Landau resonance (particle velocities equalling phase speed of the wave vz = ω/k)
leads to damping of the wave if more particles lag just behind than overtake the wave and to
instability in the opposite case.

is stationary in their reference frame and so can do work on these particles, giving its
energy to them (damping) or receiving energy from them (instability). In contrast, other,
out-of-phase, particles experience no mean energy change over time because the field
that they “see” is oscillating. It turns out (§3.6) that the process works in the spirit of
socialist redistribution: the particles slightly lagging behind the wave will, on average,
receive energy from it, damping the wave, whereas those overtaking the wave will have
some of their energy taken away, amplifying the wave. The condition ωF 0 (ω/k) < 0
corresponds to the stragglers being more numerous than the strivers, leading to net
damping; ωF 0 (ω/k) > 0 implies the opposite, leading to an instability (which then leads
to flattening of the distribution; see §6).
Let us note again that these results are quantitatively valid only in the limit (3.33),
or, equivalently, (3.40). It makes sense that damping should be slow (γ  ω) in the
limit where the waves propagate much faster than the majority of the electrons (ω/k 
vthe ) and so can interact only with a small number of particularly fast particles (for a
2 2 2
Maxwellian equilibrium distribution, it is an exponentially small number ∼ e−ω /k vthe ).
If, on the other hand, ω/k ∼ vthe , the waves interact with the majority population and
the damping should be strong: a priori, we might expect γ ∼ kvthe .17

Exercise 3.2. Stability of isotropic distributions. Prove that if f0e (vx , vy , vz ) = f0e (v), i.e.,
if it is a 3D-isotropic distribution, monotonic or otherwise, the Langmuir waves at kλDe  1 are
always damped (this is solved in Lifshitz & Pitaevskii 1981; the statement of stability of isotropic
distributions is in fact valid much more generally than just for long-wavelength Langmuir waves:
see Exercise 4.2).

Landau’s method of working out waves and damping in collisionless plasmas, and in particular
his prescription for dealing with the singularities in the integrals, has always elicited a degree of
dissatisfaction in the minds of some mathematically inclined physicists and motivated them to
search for alternatives. Perhaps the earliest and best known such alternative is the formalism
due to van Kampen. His objective was more mathematical rigour—but even if this is of limited
appeal to you, the book by van Kampen & Felderhof (1967) is still a good read and a good
chance to question and re-examine your understanding of how it all works.
A key question that preoccupied van Kampen and many of those who re-examined Landau’s
solution later on was whether the initial-value problem for the linear evolution of perturbations
in a plasma could be solved in the usual way such things are done elsewhere in physics: by
decomposing the initial perturbation into some convenient set of normal modes, advancing each

17
This is indeed correct. You can confirm it numerically using (3.82) and (3.88).
 Oxford MMathPhys Lectures: Plasma Kinetics and MHD 31
of them in time, and then reassembling them back into the desired solution. The answer is
yes—van Kampen did find a complete set of modes, although they were not eigenfunctions of a
Hermitian operator and thus (arguably) were not very user-friendly. In a short paper by Ramos
& White (2018), you will find the most recent and the most transparent (in my view) scheme for
how to construct normal modes that are eigenfunctions of a Hermitian operator.18 One curious
corollary is that it is possible to cook up special initial perturbations that will not decay at the
Landau rate and, in fact, can have any time evolution that one cares to specify! If this intrigues
or disturbs you, follow the paper trail from Ramos & White (2018) backward in time.19

Landau damping became a cause célèbre in the hard-core mathematics community, as well as
in the wider science world, with the award of the Fields Medal in 2010 to Cédric Villani, who
proved (with C. Mouhot) that, basically, Landau’s solution of the linearised Vlasov equation
survived as a solution of the full nonlinear Vlasov equation for small enough and regular enough
initial perturbations: see a “popular” account of this by Villani (2014). The regularity restriction
is apparently important and the result can break down in interesting ways (Bedrossian 2016).
The culprit is plasma echo, of which more will be said in §12 (without any claim to mathematical
rigour).

3.6. Physical Picture of Landau Damping

The following simple argument (Lifshitz & Pitaevskii 1981) illustrates the physical mechanism
of Landau damping.
Consider an electron moving along the z axis, subject to a wave-like electric field:
dz
= vz , (3.42)
dt
dvz e e
=− E(z, t) = − E0 cos(ωt − kz)eεt . (3.43)
dt me me
I have given the electric field a slow time dependence, E ∝ eεt , but will later take ε → +0—this
describes the field switching on infinitely slowly from t = −∞. Let us assume that the amplitude
E0 of the electric field is so small that it changes the electron’s trajectory only a little over several
wave periods. Then the equations of motion can be solved perturbatively.
The lowest-order (E0 = 0) solution is
vz (t) = v0 = const, z(t) = v0 t. (3.44)
In the next order, let
vz (t) = v0 + δvz (t), z(t) = v0 t + δz(t). (3.45)
Equation (3.43) becomes
dδvz e e eE0
=− E(z(t), t) ≈ − E(v0 t, t) = − Re e[i(ω−kv0 )+ε]t . (3.46)
dt me me me
Integrating this gives
Z t
eE0 0
δvz (t) = − dt0 Re e[i(ω−kv0 )+ε]t
me 0

eE0 e[i(ω−kv0 )+ε]t − 1

=− Re
me i(ω − kv0 ) + ε
eE0 εeεt cos[(ω − kv0 )t] − ε + (ω − kv0 )eεt sin[(ω − kv0 )t]
=− . (3.47)
me (ω − kv0 )2 + ε2

18
This line of thinking has also made its way into galactic dynamics: see Lau & Binney (2021a,b).
19
Another amusing recent exercise is the paper by Heninger & Morrison (2018), where (following
up on Morrison 1994, 2000), van Kampen’s scheme is recast as a new transform (called
“G-transform”) to be used instead of the Laplace transform to solve Landau’s initial-value
problem.