You are on page 1of 12

Chapter 4

Waves in ideal MHD and


two-fluid theory

A plasma immersed in a magnetic field can support a great variety of waves.


A first insight can be gained in the framework of ideal MHD, i. e. starting
from the set of Eqs. (3.16,3.19,3.20,3.23) discussed in the previous Chapter. In
particular, the problem can be simplified assuming that the wave represents
just a small perturbation of an otherwise homogeneous and static equilibrium.
Schematically, this amounts to a procedure involving the following basic steps:

1. Define a uniform equilibrium, i. e. all quantities are independent of space


and time.

2. Introduce a small perturbation which can vary in space and time, and
linearize the equations.

3. Substitute a plane wave for all perturbed quantities and reduce the equa-
tions to algebraic equations.

4. Now eliminate all the quantities but one, to have a linear equation in only
one of the perturbed quantities.

5. The linear equation has a trivial solution (the equilibrium), but to have
a non-trivial solution in general a constraint has to be satisfied. This will
give the dispersion relation, i. e. the relation between the wave frequency
and the wave vector of the wave admitted in the system (in general more
than one wave can exist).

At the end of this chapter, we will also consider an example of perturbation of a


non-uniform background (differently from item 1 above), which is particularly
important in the frame of plasma turbulence, as we will see later.

45
46 CHAPTER 4. WAVES IN IDEAL MHD AND TWO-FLUID THEORY

4.1 MHD Equilibrium and linearized equations


As mentioned above, we suppose the wave to be a small perturbation of a
homogeneous and static equilibrium. This perturbation varies in space and time
according to a plane-wave ansatz. We will denote the perturbed quantities by
a tilde, so that in general for a given quantity g we will now have to substitute

g(r, t) = g0 + g̃(r, t)

in the original equations. Since we are interested in linear waves we will assume
all perturbations to be small and consistently neglect all products involving
two or more fluctuating quantities. This procedure can be demonstrated for
the continuity equation

∂(ρ0 + ρ̃)
+ ∇ · [(ρ0 + ρ̃)(u0 + ũ)] = 0.
∂t
This equation now considerably simplifies because the partial derivatives of ρ0
are zero for a static homogeneous medium; moreover, to the lowest order the
velocity is assumed to vanish, u0 ≡ 0. The term ρ̃ũ is of second order and will
be neglected. Therefore, the continuity equation becomes

∂ ρ̃
+ ρ0 ∇ · ũ = 0. (4.1)
∂t
The remaining equations of ideal MHD can be treated in a similar way. The
result for the equation of motion is (remember that µ0 j0 = ∇ × B0 = 0 because
B0 is uniform)
∂ ũ
ρ0 = −∇p̃ + j̃ × B0 . (4.2)
∂t
The pressure and magnetic-field perturbation read respectively

∂ p̃
= −γp0 ∇ · ũ (4.3)
∂t
and
∂ B̃
= ∇ × (ũ × B0 ) . (4.4)
∂t
Deriving Eq. (4.2) with respect to time (where obviously j̃ = ∇ × B̃/µ0 ) and
employing Eqs. (4.3,4.4) yields a linear differential equation for ũ:

∂ 2 ũ 1
ρ0 = γp0 ∇(∇ · ũ) + ∇ × [∇ × (ũ × B0 )] × B0 . (4.5)
∂t2 µ0

This is our linearized wave equation, where we make the Fourier-ansatz ũ =


û exp[i(k·r−ωt)]. Without loss of generality we can choose our reference frame
such that B0 = B0 ẑ and k lies in the yz plane, i. e. kx = 0, ky = k⊥ , kz = kk .
4.1. MHD EQUILIBRIUM AND LINEARIZED EQUATIONS 47

After some simple vector algebra one can show that {k × [k × (û × ẑ)]} × ẑ =
kk2 ûx x̂ + k 2 uy ŷ. Introducing the Alfvén speed

2 B02
vA ≡ (4.6)
µ0 ρ 0
and the sound speed
γp0
c2s ≡ (4.7)
ρ0
Eq. (4.5) can be written in matrix form as
 2
ω − kk2 vA
2
0 0
 
ûx
 0 ω 2 − k⊥
2 2
cs − k 2 vA
2
−kk k⊥ c2s   ûy  = 0. (4.8)
2
0 −kk k⊥ cs ω 2 − kk2 c2s ûz

This equation always has a trivial solution û = 0. To have a nontrivial solution,


the determinant of the matrix acting on û must vanish. This leads to the
dispersion relation
 h i
2
ω 2 − kk2 vA ω 4 − k 2 c2s + vA
2
 2
ω + kk2 k 2 c2s vA
2
= 0.

This equation has three real positive solutions for ω 2 corresponding to

ω 2 = kk2 vA
2
(4.9)

and
1h 2 2  q i
ω2 = k cs + vA 2 2 )2 − 4k 2 k 2 c2 v 2 =
∓ k 4 (c2s + vA k s A
2 √
 1∓ 1−δ
= k 2 c2s + vA
2
, (4.10)
2
where it is immediate to show that δ ≡ (kk2 /k 2 )4c2s vA
2
/(c2s + vA
2 2
) is always ≤ 1,
since both factors are never larger than unity.

~ ~ Perturbed magnetic field lines


vB

Unperturbed mag-
netic field
ez
k, B
48 CHAPTER 4. WAVES IN IDEAL MHD AND TWO-FLUID THEORY

Fig. 1: The shear-Alfvén wave.

Let us discuss the the solution given by Eq.(4.9) first. In this case the wave
equation (4.8) implies ũx 6= 0, ũy = ũz = 0. From Eq.(4.4) follows after Fourier
transform −iω B̃ = ik×(ũ×B0 ẑ), from which we deduce that the magnetic-field
perturbation B̃ is in the x-direction, i. e. this solution describes a transverse
wave called the shear-Alfvén wave, as shown in Fig. 1. It is an incompressible
wave, since ∇ · ũ → k · ũ = 0 (remember that we fixed the reference frame such
that kx = 0). The restoring force leading to the oscillation of the plasma (and
of the magnetic field) is given by the tension of the field lines discussed in the
previous chapter.
The solutions corresponding to the two dispersion relations given by Eq.(4.10)
involve both the sound and the Alfvén speeds and are therefore called the slow
and the fast “magnetoacoustic” waves (where slow and fast refer to the minus
and plus sign in Eq. (4.10), respectively). In this case ũx and B̃x vanish. To
obtain an intuitive physical picture of these waves, let us suppose that either
the parallel or the perpendicular components of the wave vector are zero. If
kk → 0, i. e. k⊥ → k, we have δ → 0. In this case the fast magnetoacoustic
wave has a frequency ω 2 = k 2 (c2s + vA2
) and the only non-vanishing component
of the velocity ũ is in the y direction. Employing again Eq. (4.4), we discover
that in this case the magnetic-field perturbation is aligned with the background
field and points in the z direction.
Perturbed magnetic field lines
Unperturbed magnetic field
~
v

~
ez B
Fig. 2: The compressional Alfvén wave.
This situation is depicted in Fig. 2. This wave is called the compressional
Alfvén wave and resembles the well-known sound wave in a compressible gas,
the main difference being the fact that in this case also the magnetic pressure
contributes or even dominates if the plasma β introduced in the previous chap-
ter, cf. Eq. (3.32), is small (it is noted that the case β ≪ 1 corresponds to
c2s ≪ vA2
, see Eqs. (4.6, 4.7)). For kk → 0, the slow magnetoacoustic wave has
ω → 0 and does not propagate.
If we take now the limit k⊥ → 0, i. e. kk → k, we have 1 − δ → (c2s −
vA ) /(c2s + vA
2 2 2 2
) . For the fast magnetoacoustic wave, we recover the features
4.2. TWO-FLUID DESCRIPTION 49

already discussed for the case of the shear-Alfvén wave, applying now the same
considerations to the y direction.

~
v
~
p

Fig. 3: The sound wave.

The dispersion relation for the slow wave reads ω 2 = k(k)


2
c2s . Again, since
ũz 6= 0 and B̃ = 0, the oscillations do not involve the magnetic field, as shown
in Fig. 3. This longitudinal wave is the standard sound wave. In the general
case, both kk and k⊥ are finite and the resulting waves are a superposition of
the fundamental modes presented above.

4.2 Two-fluid description

An investigation of plasma waves in the framework of a two-fluid description


(typically electrons and one ion species) is very instructive, since it allows to
appreciate the differences with respect to a single-fluid (MHD) approach, like
the possibility of a parallel electric field or the inclusion of the electron inertia.
In the following, we discuss how the sound wave, i. e. a wave which does
not involve any oscillation in the magnetic field, changes when the motion of
ions and electrons is treated separately. First, the oscillation is assumed to
be mainly carried by the ions (in which case the electrons can be regarded as
adiabatic, as we will see), then high-frequency plasma waves sustained by the
motion of the electrons are considered, while the ions can be assumed to form a
static neutralizing background. Since we are treating fluid oscillations parallel
to B0 , the Lorentz force q ũ × B0 does not contribute to the dynamics. It is
noted that charge separations arising from the motion along the field lines give
rise to an oscillating electric field parallel to the wave vector, k × Ẽ = 0. Such a
longitudinal wave is also called electrostatic, since a Fourier transformation of
Faraday’s law for the wave fields implies ω B̃ = k × Ẽ = 0, i. e. the oscillating
magnetic field of the wave vanishes. Since the magnetic field is not involved
in the physical processes considered below, we will set it to zero from the very
beginning.
50 CHAPTER 4. WAVES IN IDEAL MHD AND TWO-FLUID THEORY

4.2.1 Ion oscillations


For the case of ion oscillations, the dispersion relation can be obtained by
linearizing the equations for the ion fluid (3.1,3.10,3.11), which for B = 0 read

∂t ni = −∇ · (ni ui )
mi ni (∂t + ui · ∇) ui = −∇pi + Zeni E
pi pi0
= = const.
nγi nγi0

which become (again we set u0 = 0 and suppose the electric field to be due to
the perturbation, E0 = 0)

∂t ñi = −ni0 ∇ · ũi (4.11)


mi ni0 ∂t ũi = −∇p̃i + Zeni0 Ẽ (4.12)
 
pi0 γ ñi
pi0 + p̃i = n 1+γ (4.13)
nγi0 i0 ni0

From the last equation we learn that p̃i = γkB Ti0 ñi . Since the first equation
can be used to express ñi as a function of ũi , the only quantity which still
is to be determined in the equation of motion is the electric field. It must be
computed from Maxwell’s equation ∇ · Ẽ = (Zeñi − eñe )/ǫ0 which we can write
as the Poisson equation for the potential substituting Ẽ = −∇φ̃. The perturbed
electron density needed for the determination of the perturbed charged density
can be obtained starting from the assumption that the electron dynamics is
much faster than the ion dynamics we want to investigate. In other words, we
assume that the electron response is so fast that we can neglect the electron
inertia. The electron momentum balance is then

0 = −∇pe − ene E

or in linearized form (assuming isothermal electrons, Te = Te0 )

−kB Te0 ∇ñe + ene0 ∇φ̃ = 0.

From this equation we obtain an adiabatic electron density perturbation ñe /ne0 =
eφ̃/kB Te0 . The Poisson equation reads then
!
2 e eφ̃
∇ φ̃ = − Z ñi − ne0 . (4.14)
ǫ0 kB Te0

Now we assume that all the perturbed quantities in Eqs.(4.11,4.12, 4.14) satisfy
a plane-wave ansatz so that their space and time dependencies are of the form
exp[i(k · r − ωt)]. As a result, the ion density perturbation is

k · ũi
ñi = ni0 (4.15)
ω
4.2. TWO-FLUID DESCRIPTION 51

and the perturbed electrostatic potential is

1 kB Te0 k · ũi
φ̃ = 2 2 ,
1 + k λDe e ω

where the definition of the electron Debye length, Eq. (1.5), and the quasi-
neutrality condition ne0 = Zni0 have been exploited together with Eq. (4.15).
Eq. (4.12) can be finally cast in the matrix form
   
γkB Ti0 1 ZkB Te0
ω2I − + kk · ũi = 0.
mi 1 + k 2 λ2De mi

The solvability condition leads to the following dispersion relation


 
γkB Ti0 1 ZkB Te0
ω2 = + k2 .
mi 1 + k 2 λ2De mi

Assuming Te0 = Ti0 = T0 and Z = 1, we can easily calculate the wave frequency
in the limiting case k 2 λ2De ≪ 1 (long wavelength) for which we obtain ω 2 = c2s k 2
(electrons are here isothermal, so that in their contribution to the sound speed
the adiabaticity index is equal to unity). As we see, in the long-wavelength
limit we recover the MHD sound wave. In the opposite limit, k 2 λ2De ≫ 1, it is
easy to show that if Ti0 ≪ Te0 it is ω 2 = Z 2 e2 ni0 /ǫ0 mi ≡ ωpi , i. e. we have
ion plasma oscillations. Including finite Ti0 leads to the ion counterpart of the
Bohm-Gross dispersion relation derived below for the electrons.

4.2.2 Electron oscillations


For oscillations occurring on the time scale typical of the electron motion, we
can neglect the ion response, so that no perturbed ion quantities are considered.
In this case we linearize the electron fluid equations

∂t ne = −∇ · (ne ue )
me ne (∂t + ue · ∇) ue = −∇pe − ene E
pe pe0
= const. = γ
nγe ne0

and obtain

∂t ñe = −ne0 ∇ · ũe (4.16)


me ne0 ∂t ũe = −∇p̃e − ene0 Ẽ (4.17)
 
pe0 γ ñe
pe0 + p̃e = n 1+γ . (4.18)
nγe0 e0 ne0

In the case under consideration, the Poisson equation is simply ∇·Ẽ = −eñe /ǫ0 .
For a plane wave it is immediate to determine k · Ẽ in terms of ñe and ñe in
52 CHAPTER 4. WAVES IN IDEAL MHD AND TWO-FLUID THEORY

terms of k · ũe . Taking the scalar product of Eq.(4.17) with k one obtains then
an equation (dispersion relation) for ω as a function of k, namely

kB Te0
ω 2 = ωpe
2
+ γk 2
me

or equivalently
ω2
2
= 1 + γk 2 λ2De
ωpe
which is called the Bohm-Gross dispersion relation. In this case, the long-
wavelength limit k 2 λ2De ≪ 1 leads to the plasma oscillations we have met in
the first chapter. Thermal effects modify the wave frequency. In the short-
wavelength limit, the wave propagates with constant phase and group velocity
(of the order of the electron thermal speed) in a similar way as a sound wave,
where the dynamics, however, is mediated by both the pressure gradient and
the electric field perturbation.
A particularly interesting role in the oscillations discussed above is played
by the particles moving at a speed comparable to the wave velocity. We will
investigate this point in the frame of the kinetic theory and discover that the
waves can undergo a collisionless damping called the Landau damping.

4.3 Inhomogeneous background: The drift wave


Up to now, we limited our discussion to perturbations of uniform equilibria.
We now relax this assumption and consider the case of a plasma with an equi-
librium density gradient to show that a new basic plasma wave, called the drift
wave, can develop in this case. Its importance is related to the fact that this
oscillation, and the possible related instability, can be considered as a sort of
paradigm for those plasma perturbations (often referred to as “drift instabili-
ties”) which are mainly responsible for turbulent transport of particle and heat
in magnetically confined plasmas. For these instabilities, the drift ordering dis-
cussed in Sec. 3.1.1 is assumed to apply. The fundamental underlying picture
is that the parallel dynamics is dictated by the electrons, which can respond
much more quickly than the ions to perturbation along the field lines. The
perpendicular fluid motion of the ions, on the other hand, is described to the
lowest order by the E × B and diamagnetic velocities as in Eq. (3.13) and to
the next order by the polarization drift, cf. Sec. 3.1.1. We assume moreover
that the fluid motion is slow enough that the inductive part of the electric field
can be neglected and we write E = −∇φ.
With reference to Fig. 4, we choose the background density gradient to
point in the negative x-direction, while the magnetic field is in z-direction,
B = Bẑ. Before linearizing the relevant fluid equations, let us consider the
continuity equation for the ions under the assumption that the perpendicular
flow is given by Eq.(3.13). Splitting the parallel and perpendicular components
4.3. INHOMOGENEOUS BACKGROUND: THE DRIFT WAVE 53

of the flow, Eq. (3.1) becomes


   
∂ni  E×B B × ∇pi
= −∇k ni uik − ∇ · ni − ∇ · . (4.19)
∂t B2 eB 2
In the diamagnetic flow (last term), the ion charge is supposed to be +e for
the sake of simplicity. As discussed above, we neglect the parallel dynamics of
the ions (first term on the right-hand side of the previous equation).

Fig. 4: The drift wave. The E × B-velocity uE is shown by red arrows.

Using Eqs. (B.6,B.11) it is straightforward to show that the diamagnetic


flow is divergence-free in the uniform-B case considered here. Using the same
vector identities it is shown that also the divergence of the E × B velocity is
zero, so that Eq. (4.19) reduces to
∂ni ∇φ × ẑ
= · ∇ni . (4.20)
∂t B
The procedure for the derivation of the electron response to the periodic density
perturbation depicted in Fig. 4 is the same as in Sec. 4.2.1. In the parallel
component of the equation of motion of the electrons, according to the drift
ordering considered here, the electron inertia is dropped and Eq.(3.10) becomes
simply
−∇k pe − ene Ek = 0. (4.21)
54 CHAPTER 4. WAVES IN IDEAL MHD AND TWO-FLUID THEORY

This equation shows that as a result of a pressure perturbation along the


magnetic-field lines, an electric field arises which is directed against the di-
rection of ∇k pe . In the following, a further simplification is introduced by
assuming the electrons to be isothermal, so that a pressure perturbation be-
comes effectively a density perturbation.
As in the previous sections of this chapter, Eqs. (4.20,4.21) are now lin-
earized and a plane-wave ansatz is substituted for the perturbed quantities. As
before, there is no equilibrium electric field (nor electrostatic potential). The
ion continuity equation reads then

iφ̃ dn0
−iωñi = (k × ẑ) · x̂
B dx
iky φ̃ dn0
= , (4.22)
B dx
where we took into account the fact that the equilibrium density profile is
supposed to depend on the x coordinate only. The linearized and Fourier-
transformed equation of motion for the electrons is simply
Te ñe
−ikk p̃e + ien0 kk φ̃ = 0 ⇒ φ̃ = (4.23)
e n0
(the last step being a consequence of the assumption of isothermal electrons,
T̃e = 0; moreover in this section we set for simplicity kB = 1 and treat the
temperatures as energies). Combining the last two equations and setting ñi =
ñe = ñ, we obtain the following dispersion relation
ky Te dn0
ω=− ≡ ω∗e . (4.24)
en0 B dx
This equation expresses the fact that the perturbation sketched in Fig. 4 moves
upwards (i.e. in the positive y-direction) at the diamagnetic speed of the elec-
trons. The displacement of the perturbation is due to the fact that the elec-
trostatic potential given by Eq. (4.23) leads to an E × B-velocity that advects
the plasma as shown in the figure. In the previous equation, ω∗e is called the
electron diamagnetic frequency.
Like for the oscillations described previously in this chapter, the frequency
of the drift wave derived in Eq. (4.24) is purely real, i.e. the perturbation is
neither growing nor damped. This is a consequence of the adiabatic electron
response as expressed by Eq.(4.23), which forces the electrostatic potential and
the density perturbation to oscillate with the same phase. Effects not retained
in Eq. (4.23), like finite electron inertia or electron-ion collisions, can lead to a
deviation from perfect adiabaticity, which can be modelled as
Te ñe iδ
φ̃ = e . (4.25)
e n0
For 0 < δ < π, the potential perturbation lags behind the density perturbation.
As a consequence, the E × B-velocity brings denser plasma into regions with
4.3. INHOMOGENEOUS BACKGROUND: THE DRIFT WAVE 55

ñ > 0 and less dense plasma where ñ < 0. This situation corresponds thus
to an instability of the initial perturbation. In this case, the amplitude of
the perturbations will in general grow to a level for which the linearization
procedure employed throughout this chapter is no longer valid and the system
will evolve into a turbulent state, cf. Chapt. 6.
Before concluding this section, it is interesting to investigate the role of
the polarization drift introduced in Sec. 2.4 and Sec. 3.1.1 for the drift wave
dynamics. According to Eq. (3.14), the polarization-drift velocity is (for sim-
plicity we neglect here the contribution of the diamagnetic velocity, which is
sometimes referred to as the cold-ion limit, pi → 0)

mi dE⊥
u1⊥ = (4.26)
eB 2 dt
taking qi = e and considering only the ions because mi ≫ me . Although it is
formally of higher order with respect to the E × B-velocity uE , the polarization
velocity alters the previous description because it is not divergence-free, so that
in the linearized version of the continuity equation Eq.(3.1) we have now a term
proportional to −n0 ∇ · u1⊥ :

∂ ñi ∇φ̃ × ẑ mi n 0 ∂ 2
= · ∇n0 + ∇ φ̃, (4.27)
∂t B eB 2 ∂t ⊥

again with E = −∇φ̃. Taking the Fourier transform of the perturbed quantities
in the previous equation and proceeding as in the derivation of Eq. (4.24) we
obtain with φ̃ = (Te /e)(ñe /n0 ) and ñe = ñi :
2
ky Te dn0 2 mi T e
ω=− − ωk⊥ .
en0 B dx e 2 B 2 mi

The ratio Te /mi = c2s is the sound speed squared (in the limit Ti → 0 considered
here). Introducing the sound Larmor radius
cs
ρs = , (4.28)
Ωi

the previous equation leads to the dispersion relation


ω∗e
ω= 2 ρ2 , (4.29)
1 + k⊥ s

where the second term in the denominator contains the effect of the polarization
drift, which was absent in the derivation of Eq. (4.24). In the frame of the iδ
model mentioned above, i.e. assuming that the potential φ̃ deviates from perfect
adiabaticity according to Eq.(4.25), the result of the previous derivation would
be
ωe−iδ = ω∗e − ωk⊥ 2 2
ρs ,
56 CHAPTER 4. WAVES IN IDEAL MHD AND TWO-FLUID THEORY

2 2
which for δ ≪ 1 leads to ω = ω∗e /(1 + k⊥ ρs − iδ). Splitting real and imaginary
part we obtain again Eq. (4.29) for the wave frequency ω, while
ω∗e
γ= 2 δ (4.30)
2 ρ2 )
(1 + k⊥ s

expresses the growth rate of the drift wave (positive if δ > 0, as seen before).
2 2
For wavelengths which become smaller than ρs , the term k⊥ ρs dominates in
the denominators of both Eq. (4.29) and Eq. (4.30), so that both the frequency
2 2
and the growth rate tend to zero in the limit k⊥ ρs ≫ 1. This is linked to
the fact that while the advection term, i.e. the first term on the right-hand
side of Eq. (4.27), is proportional to ∇φ̃ ∼ ∇ñ, and is hence out of phase with
respect to the density perturbation ñ, leading to propagation of the density
disturbance, the polarization term, i.e. the second term on the right-hand side
of Eq. (4.27), is proportional to ∇2 φ̃ ∼ ∇2 ñ and does not lead to a drift. As
a consequence of Eqs. (4.29,4.30), we can expect that the maximum drift wave
frequency and growth rate correspond to wavelengths comparable to the sound
Larmor radius ρs , while ω, γ → 0 for both k⊥ ∼ ky → 0 and → ∞.

4.4 Literature
1. T. J. M. Boyd, J. J. Sanderson, The Physics of Plasmas, Cambridge
University Press, 2003.
2. F. F. Chen, Introduction to Plasma Physics, Plenum Press, 2nd edition,
1984.
3. U. Stroth, Plasmaphysik, 2. Auflage, Springer Spektrum, 2018.

You might also like