89

17. WAVES AND SHOCKS IN MHD where c2s = γkB To /m is the square of the (unper-
turbed) sound speed.
Nonmagnetized flows have one important characteris-
These equations (17.3) can be combined, algebra-
tic signal speed – that’s the sound speed, as we saw in
ed, and eventually made into one equation for the dis-
Chapter 7. When we add a B field, things get more
turbance velocity, v1 :
interesting ... there is more than one possible “signal-
carrying” wave. ∂ 2 v1
= c2s ∇(∇ · v1 )
A. MHD Waves ∂t2 (17.4)
Bo
In Chapter 7 we found that a density perturbation prop- = [∇ × (∇ × (v1 × Bo ))] ×
4πρo
agates at a speed cs = (∂p/∂ρ)1/2 . This is a criti-
cal quantity in fluid dynamics – the speed at which in- If we now assume a plane-wave solution,
formation can propagate in a gas. For a magnetized
plasma, the situation is (not surprisingly) more com- v1 (r, t) = v1 ei(k·r−ωt) (17.5)
plex. There are two characteristic waves, Alfven and
our linearized equation reduces to
magnetosonic. To start, we carry out a linear analysis
to discover what types of perturbations propagate, and
at what speeds. ω 2 v1 =c2s k(k · v1 )
Bo (17.6)
1. BASIC STRUCTURE : LINEAR ANALYSIS + [k × (k × (v1 × Bo ))] ×
4πρo
Start with a set of basic, unperturbed equations:
Now, finally, we’re in position to analyze basic wave
Dρ modes for this system.
+ ρ∇ · v = 0
Dt As a prelude, note that if Bo = 0, (17.6 reduces to
Dv 1
ρ = −∇p + (∇ × B) × B
Dt 4π  ω 2 v1 = c2s k(k · v1 ) (17.7)
D p
=0 Dotting this with k, and assuming k · v1 6= 0 (i.e.the
Dt ργ
∂B perturbed velocity has some component parallel to the
= ∇ × (v × B) wavevector), this gives
∂t
∇·B = 0 (17.1) ω = cs k (17.8)

Linearize these in the usual way. That is, take for the dispersion relation. This should make you feel
good – we have recovered sound waves.
ρ = ρo + ρ1 ; v = v1
2. ALFVEN WAVES
p = po + p1 ; B = Bo + B1 (17.2)
These are waves in which the magnetic field dominates.
where the unperturbed (“zero” subscript) state is taken It exerts the restoring force; fluctuations in the plasma
as uniform and homogeneous. (The more general case density and pressure are either exactly zero, or unim-
of background structure, won’t be addressed here). portant. We can start by guesstimating the likely wave
With this, we get speed. Recall that waves in an elastic wire propagate
∂ρ1 due to the tension; as the field lines in a plasma exert a
+ (v1 · ∇)ρo + ρo (∇ · v1 ) = 0 tension Bo2 /4π, one might expect a wave speed
∂t
∂v1 1 Bo
ρo = −∇p1 + (∇ × B1 ) × Bo vA = (17.9)
∂t 4π   (4πρo )1/2
∂p1 2 ∂ρ1
+ (v1 · ∇)po − cs + (v1 · ∇)ρo = 0
∂t ∂t This is the Alfven speed, and it is, indeed, a useful scal-
∂B1 ing speed for waves in a magnetized plasma. We can
= ∇ × (v1 × Bo ) also note directly that ∇ · B = 0 ⇒ k · B1 = 0; so
∂t
(17.3) that the magnetic field perturbation must be normal to
∇ · B1 = 0 the wavevector.
 90

Following Priest, we now specify to the magnetized- angle terms explicit letting θ be the angle between k
only limit, that is, dropping the po terms from (17.3). and Bo ), and letting b̂o be the unit vector along Bo , we
This is called the cold-plasma limit, and means we’re have the Alfvenic wave dispersion relation:
ignoring any internal-pressure effects (wo we no longer
have sound waves). Doing this, rewriting to make the

ω2 2 2
h i
2 v1 = k cos θv1 − (k · v1 )k cos θ b̂o + (k · v1 ) − k cos θ(b̂o · v1 ) k (17.10)
vA

Dotting this with bˆo gives bˆo · v1 = 0; thus, waves in an Alfvenic disturbance described by
this limit just have the perturbed velocity normal to the
ambient magnetic field. Dotting this with k gives B1
v1 = − ; |Bo + B1 | = constant
(4πρo )1/2
ω 2 − k2 vA
2


k · v1 = 0 (17.11) (17.14)
satisfies the full equations, 17.1), without the need to
This has two separate solutions, as follows.
linearize. One important consequence of these waves,
B0 Priest notes, is that finite-amplitude waves do not tend
mag tension
v B1 to steepen, and so dissipate must less readily than other
1
wave modes.
B
mag tension
• compressional alfven waves The second solution
mag tension
to (17.11) is
Figure 17.1. Schematic of (shear) Alfven waves; the
perturbed B1 and v1 terms are perpendicular to the ω = kvA (17.15)
background field Bo . Following Cravens Figure 4.16.
which describes compressional Alfven waves. For these
• (shear) alfven waves. This is the case of an in- waves, the linearized equations (17.3) show that v1 is
compressible perturbation: normal to Bo , and lies in the (k, Bo ) plane. It therefore
∇ · v1 = 0 ; k · v1 = 0 (17.12) has components both along and transverse to k, and so
gives rise to density and pressure fluctuations.
which is, of course, one solution to (17.11). These are,
thus, transverse waves (particle motion transverse to k, 3. MAGNETOSONIC WAVES

B1 ⊥ Bo ). Using this, (17.10) gives When the gas pressure is dynamically comparable
ω = kvA cos θ (17.13) to the magnetic field, the wave nature is different.
We might expect the wave speed to be a mixture of
for Alfven waves (sometimes called shear Alfven compressive effects (through cs ) and magnetic effects
waves). These waves have a phase speed vA cos θ, (through vA ).
which agrees with our simple argument for waves prop-
B
agating exactly along Bo . Putting this solution back 1
k
into the linearized equations (17.3), the first two show v1

that there are no density or pressure perturbations: ρ1 = Bo

p1 = 0. So, the plasma just moves back and forth with

B1 , without any compressive effects. (The same equa-
tion sshow that v1 = B1 / (4πρo )1/2 , so that v1 and B1
Figure 17.2. Schematic of magnetosonic waves,
are in the same direction). illustrating a compressive wave propagating at right angles
An interesting aside is that these solutions also de- to the background Bo . Following Cravens figure 4.17.
scribe finite-amplitude waves. Most wave solutions re-
quire small-amplitude perturbations (as we began with Returning to (17.6) and keeping the pressure and
here). However, with some algebra one can verify that compresssive terms, the ruling equation here is
 91

ω2 2 2
h i
ˆo + (1 + c2 /v 2 )(k · v1 ) − k cos θ(bˆo · v1 ) k
2 v1 = k cos θv1 − (k · v1 )k cos θ b s A (17.16)
vA

Dotting this with k and bo , in turn, gives two useful B. MHD Shocks; Jump Conditions
relations:
In Chapter 9, we found that density perturbations –
−ω 2 + k2 c2s + k2 vA
2
(k · v1 ) = k3 vA
2
cos θ(bˆo · v1 )


sound waves – will steepen and can develop into
(17.17) shocks. Just as there are more than one signal-carrying
and type of MHD wave, there are more than one type of
MHD shock.
k cos θc2s (k · v1 ) = ω 2 (bˆo · v1 ) (17.18) Also in chapter 9, we applied conservation laws
to determine the jump conditions at hydrodynamic
If k · v1 = 0, we recover the Alfven wave solution shocks. We will follow the same path here, and we will
(17.13). If this isn’t zero, we can combine these two re- find that there is more than one soltuion to the jump
lations to find the dispersion relation for magnetosonic conditions – more than one type of shock. In particular,
waves: magnetosonic waves are compressive, and can steepen
into shocks; the two MS modes, fast and slow, connect
ω 4 − ω 2 k2 c2s + vA2 2 4
+ c2s vA k cos2 θ = 0 (17.19)


to two types of MHD shocks. Alfven waves, not being
For forward-pointing (ω/k > 0) waves, there are two compressive, do not steepen into shocks (although there
distinct solutions: are formal discontinuities that one can find, associated
with Alfven waves.)
ω2 1 2 2

1 4 4

1/2
2
= cs + vA ± cs + vA − 2c2s vA
2
cos 2θ We work in a frame in which the shock is at rest; this
k 2 2 makes everything steady state. We also ignore dissipa-
(17.20)
tion; it matters within the shock, but here we idealize to
The minus sign in (17.20) gives what is known as an infintely thin jump. We use to the notation of Chap-
the slow mode; the plus sign gives the fast mode. The ter 9, that is [[A]] is the jump in A across a boundary.
Alfven speed lies inbetween these two wave speeds; The unit vectors are n̂, normal to the boundary, and t̂,
thus the Alfven wave is occasionally called the inter- tangential to it. We start with the basic equations, and
mediate mode. for each write down the jump across the shock.
These two magnetosonic modes may be thought of • Maxwell. We know ∇ · B = 0; this
(as Priest notes) as a sound wave, modified by the mag-
netic field, and a compressional Alfven wave, modified [[B · n̂]] = 0 ; Bn = constant (17.21)
by the plasma pressure. If the B field becomes small
(vA → 0), the slow mode disappears, and the fast wave
is the continuity of the normal component, Bn , of the
becomes a sound wave. If the gas pressure becomes
B field. (You recall that the jump in Bt can be finite if
small (cs → 0), the slow mode disappears again, and
there is a surface current).
the fast wave becomes a compressional Alfven wave.
• Mass flux. We have, in a steady state, ∇ · (ρv) =
4. VALIDITY OF MHD WAVE THEORY 0, so that
An important limitation on this MHD wave theory is [[ρv · n̂]] = 0 ; ρvn = constant (17.22)
that it assumes simple, collective particle motion. At
higher frequencies, the single-particle motions in mixed
is the continuity of mass flux.
magnetic and electric fields cannot keep up with the
driving wave. The wave modes become more com- • Induction. In steady state, ∇ × (v × B) = 0.
plicated as a result, and require a plasma-physical ap- Across the boundary, this gives
proach (following individual particles, or at least indi-
vidual species). The common usage is that we can treat [[(v × B)t ]] = 0 ; [[vn Bt ]] = Bn [[vt ]] (17.23)
waves by MHD methods for wave frequencies below
Ωi = eBo /mi c, the ion gyrofrequency. where we’ve used Bn =constant in the last.
 92

• Momentum flux. • Parallel Shocks have the shock normal parallel

to B. These are no different from unmagnetized
shocks. The fluid can flow along the B field with-
B · n̂ B 2
  
out hindrance. There is no v × B term to maintain
ρv(v · n̂) + pn̂ − B − n̂ =0
4π 4π an EMF, nor is there a current (as B is constant by
(17.24) assumption).
is the basic momentum flux. A useful alternative to this • Perpendicular (Normal) Shocks have the shock
(resembling Bernoulli’s equation) comes from dotting normal perpendicular to B. These are the simplest
it with n̂: cases of magnetized shocks.
• Oblique Shocks are, as before, the most com-
  plicated. My presentation follows Priest, and
2 1 2

2
p + ρ(v · n̂) + (B · t̂) − (B · n̂) =0 Woods; the latter is the most thorough. There does
8π not seem to be an extensive literature on oblique
(17.25) MHD shocks – the complexity of the results may
We can also scalar multiply with t̂ to get be why.

  C. Perpendicular (Normal) Shocks

Bn Bt
ρvn vt − =0 (17.26) That is, let Bn = 0, but allow non-zero vn . A useful
4π
fact: from (17.28), we know [[vt ]] = 0 in this case. We
can then, always, transform to a frame with vt = 0;
These last two can be turned into more useful forms if this simplifies things (and I drop the subscript on vn ).
we isolate the (conserved) mass flux, ρvn : The jump conditions now become, for perpendicular
shocks,
 
1 1  2 
[[p]] + ρ2 vn2 + Bt = 0 (17.27) ρ1 v1 = constant
ρ 8π
B1 v1 = constant
and B12 (17.31)
ρ1 v12 + p1 + = constant
8π
Bn 1 2 B2
ρvn [[vt ]] − [[Bt ]] = 0 (17.28) v1 + h1 + 1 = constant
4π 2 4πρ

where “constant” means the quantity is the same up-

• Energy flux. There are several useful forms here. stream (subscript 1) and downstream (subscript 2) of
A basic form is the shock.
The geometry is illustrated in the Figure.
 
1 2 γ 1 111
000 B
ρv v + pv − (v × B) × B = 0 B
1 000
111
000
111
2
2 γ−1 4π 000
111
000
111
000
111
000
111
(17.29) 000
111
000
111
000
111
As before (chapter 6), we can define h = e + p/ρ = 000
111
000
111
000
111
000
111
γp/(γ − 1)ρ, with e = p/(γ − 1)ρ; and again use con- 000
111
000
111
000
111
000
111
tinuity of Bn and mass flux, to write this as 000
111
000
111
v 000
111 v
1 000
111
000
111 2
000
111
000
111
000
111
000
111
000
111
vn2 v2 B2 000
111
 
Bn 000
111
ρvn +h+ t + t [[vt · Bt ]]
= 000
111
S
2 2 4πρ 4π Figure 17.3. A plane perpendicular shock; the magnetic
(17.30) field is parallel to the shock front, and perpendicular to the
As before, the nature of the shock depends on its shock normal.
orientation relative to the flow; and also, in this case,
on its orientation to the B field. We can think of three Again, let’s call the compression ratio X = ρ2 /ρ1
cases. (where “2” is downstream and “1” is upstream”), and
 93

let the upstream B field be scaled by b = B 2 /4πρc2s = Solving (17.33) shows us that a perpendicular, mag-
vA2 /c2 . The various equations can be written as, netized shock is compressive only if the upstream flow
s
(relative to the shock) is above the fast magnetosonic
ρ2 speed. That is, X ≥ 1, if M2 ≥ 1 + vA 2 /c2 ; the up-
=X s
ρ1 stream flow must satisfy v12 ≥ c2s1 + vA1
2 . Compared to

v1 the HD solutions (Chapter 7), the effect of the B field is

=X
v2 to reduce the density jump X below its hydrodynamic
B2 (17.32) value (since the flow energy can go to magnetic as well
=X as thermal energy). At the same time, the pressure
B1
  and temperature jumps are larger than for the B = 0
p2 2 X −1 γb
1 − X2


= γM +1+ case. These shocks have the same limiting behavior as
p1 X 2 M → ∞: the density and field jumps asymptote to
(γ + 1)/(γ − 1); while the pressure and temperature
where the compression ratio is the solution of
jumps rise without bound.
b (γ − 2)X 2 − γX
 
D. Oblique Shocks
(17.33)
= X (γ − 1)M2 + 2 − (γ + 1)M2 MHD shocks at general angles can be more complex, of
 
course. To specify the geometry, let x̂ be the direction
Solutions to this are shown in Figure 17.4. (Details: along the shock normal, and ŷ be in the plane of the
doing the algebra on equations (17.32), and eliminating shock, and in the plane of the B field.
p2 /p1 , gives a cubic equation in X. I’ve factored out
1. DO IT GENERALLY
(X − 1) (which is not zero in any interesting case), to
get this quadratic. Note also typos in two of Priest’s The general jump conditions (or conservation laws) are
equations have been fixed.) as follows. For mass

3
ρvx = constant (17.34)

(where “constant” means has the same value on both

sides of the shock). The ∇ · B Maxwell-eqn becomes
2
Bx = constant (17.35)

Two components of the force equation become

1
B 2 Bx2
p+ − + ρvx2 = constant (17.36)
8π 4π
0 and
Bx By
ρvx vy − = constant (17.37)
4π
-1
5 10 15 20 The induction equation becomes

Figure 17.4. Numerical solution of the jump conditions, vx By − vy Bx = constant (17.38)

17.11, for unmagnetized shocks and for two different values
2
of b = vA /c2s . See text for discussion. And, finally, the energy equation becomes

B2 1 2 B2
   
B·v
vx p+ − Bx + vx ρe + ρv + = constant (17.39)
8π 4π 2 8π
 94

These equations can, in principle, be solved for the [[vx ]] − m [[ν]] = 0 (17.42)
general shock jump conditions. But they are long and
complex, so most authors simplify by working in a use-
ful reference frame.
1
2. FIND A USEFUL REFERENCE FRAME n [ vx ]] + [[p]] + hBy i [[By ]] = 0 (17.43)
4π
At this point, it is very helpful to transform to a frame
moving along the shock face, at a speed
B1y 1
v1y = v1x (17.40) n [[vy ]] − Bx [[By ]] = 0 (17.44)
B1x 4π

Such a transform can always be found, if Bx 6= 0. In

this frame, the plasma velocity is parallel to the B field
1  2
on both sides of the shock; which in turn implies that vx + vy2 + [[h]] = 0

(17.45)
E = 0 on both sides. 2
This transform simplifies the situation, but the al-
gebra is still complex. It also turns out that three so- and
lutions are possible: fast shocks (which develop from
fast-mode MS waves); intermediate shocks (which de- Bx [[by ]] − m [[ν]] hBy i − mhνi [[By ]] = 0) (17.46)
velop from Alfven waves); and slow shocks (which de-
velop from slow-mode MS waves). I’ve seen two differ-
ent analytic approaches, & will summarize both here. To solve this system, Woods defines θ1 as the angle
between B1 and the shock normal; andintroduces the
3. NOW SOLVE THE SYSTEM : I “useful intermediate variables”
To follow convention, and shorten the notation a bit,
we define ν = 1/ρ as the inverse density (“specific vol- c2s [[p]]
S= 2 ; P =
ume”). Two important constants, across the shock, are vA p1
m = ρv = v/ν, the normal mass flux; and Bx , the (17.47)
[[By ]] 1S 1
normal field component. Recall [[A]] = A2 − A1 is the ǫ= ; χ= P+ ǫ
B1 γ ǫ 2
jump in A; and hAi = 12 (A1 + A2 ) is the mean value
of A. The jump conditions are, now,
In these variables, the jump conditions combine to a
[[Bx ]] = 0 (17.41) quadratic (taking γ = 5/3 explicitly):

   
2 2 5
2 sin θ1 − ǫ χ − 2 ǫ sin θ1 − 1 + S1 χ − (ǫ + 2S1 sin θ1 ) − 0 (17.48)
3 6

 
• method # 1. which can be solved directly for χ, given p2 5ǫ 1
=1+ χ− ǫ (17.50)
choices of S1 and θ1 (that is, choices for the upstream p1 3S1 2
magnetic field ratio and the flow angle). Slow shocks
correspond to χ < 0, and fast shocks to χ > 0. The and
more usual ratios can be found from the χ solution:
2
vx1 sin θ1 ǫ
2 = 1+ 2
(χ + sin θ1 ) + (χ + sin θ1 )
bx1 cos θ1 cos2 θ1
(17.51)
ρ2 χ + sin θ1 Woods presents a long discussion of the details of this
=1+ǫ (17.49)
ρ1 1 + χ sin θ1 system and some solutions.
 95

4. NOW SOLVE THE SYSTEM : II 5. BACK TO THE PHYSICS

Alternatively, the conservation laws, (17.41)-(17.46), Whichever solution method one uses, three solutions of
can be written as these equations can be found.
One has X = 1, is called an intermediate or Alfven
ρvx = constant
wave, and isn’t of great interst as it isn’t compressive.1
Bx = constant In the intermediate wave, the tangential field compo-
! nent is simply reversed in sign, so the overall field ro-
By2 Bx (17.52) tates without changing magnitude; hence the name.
ρvv = p+ n̂ − B = constant
8π 4π The other two solutions are compressive, with X >
1, and are slow and fast shocks. The slow shock has
1 2
(v + vy2 ) + h = constant 2 , and also has B < B . That is, the magnetic
v12 ≤ vA1 2 1
2 x field bends towards the shock normal in a slow shock.
(recalling that h = γp/(γ − 1)ρ is the enthalpy). Similarly, in the slow shock the parallel flow is slowed
To solve these, we can (following Priest) again use down, v2y < v1y . Conversely, the fast shock has v12 ≥
XVA1 2 , and so B > B . Thus, the field bends away
the compression ratio, X = ρ2 /ρ1 . Keeping our defi- 2 1
nitions, c2s = γp/ρ, and vA2 = B 2 /4πρ (all evaluated from the shock normal in a fast shock. The parallel flow
upstream), the jump conditions become is speeded up in the fast shock, v2y > v1y .
v2x 1
=
v1x X
v2y 2
v 2 − vA1
= 21 2
v1y v1 − XvA1
ρ2
=X (17.53)
ρ1
(a) (b) (c) (d) (e)
B2y v 2 − vA1
2
= 21 2 X
Figure 17.5. Left, the changes in magnetic field
B1y v1 − XvA1 direction that are caused by the three types of oblique MHD
shock waves; (a), slow mode shock, in which B bends
(γ − 1)Xv12 X2 − 1
 
p2 toward the shock normal; (b) intermediate wave, in which B
=X+
p1 2c2s1 X2 “flips”; and (c) fast shock, in which B bends away from the
shock normal. From Priest figure 5.6. Right, the two special
where θ is the angle between the upstream magnetic cases of switch-off (d) and switch-on (e) shocks. From
field and the shock normal. The compression ratio X is Priest figure 5.7.)
now the solution of
There are two particular special cases of these. The
v12 − XVA12


F1 (X, θ)
(17.54) slow solution becomes a switch-off shock when v1 =
1 2 2 2 vA1 . In this limit, the tangential field vanishes behind
+ vA1 v1 sin θXF2 (X) = 0
2 the shock, B2y = 0. The fast solution becomes a
where switch-on shock when v1 = X 1/2 vA1 . In this limit, the
upstream field is parallel to the shock normal: B1y = 0.
F1 (X, θ) = Xc2s1
1 2 (17.55)
+ v1 cos θ [X(γ − 1) − (γ + 1)]
2
and References
The two best references I know for MHD waves and
F2 (X) = (γ + X(2 − γ)) v12
(17.56) shocks are Priest and Woods. I’ve followed both of
− 2
XvA1 ((γ + 1) − X(γ − 1)) them here.

Presumably this equation could be solved numerically

(picking values of cos θ of M = v1 /cs1 and also of
vA1 /cs1 ). However, I have not seen such solutions pre- 1
Apparently it was the first one found, and was at that point of
sented, nor have I felt the need to work them out myself. interest for laboratory plasmas; so I gather from Bazer & Ericson.