PHYS 643 Week 4: Compressible fluids — Sound waves and

shocks
Sound waves
Compressions in a gas propagate as sound waves. The simplest case to consider is a
gas at uniform density and at rest. Small perturbations in the density, velocity, and
pressure
ρ → ρ + δρ, v → v + δv, P → P + δP
then obey the equations
∂δρ
= −ρ∇ · δv (1)
∂t
and
∂δv
= −∇δP,
ρ (2)
∂t
where we have kept only terms first order in the perturbations. These equations show
the physics of the wave: compression leads to a local increase in density and therefore
pressure; the pressure gradient acts as a restoring force trying to remove the compres-
sion.
To see that there is a wave, we assume that the perturbations are rapid enough that
there is no time for heat to flow into or out of a fluid element, so that the perturbations
are adiabatic, with
δP δρ
=γ .
P ρ
In that case, equations (1) and (2) can be combined into a wave equation

∂2 v γP 2
2
= ∇ v = c2s ∇2 v,
∂t ρ

where the adiabatic sound speed1 cs is given by c2s = γP/ρ. This is the sound speed that
we usually think of – looking up values for atmospheric pressure ≈ 105 Pa, density of
air at STP ≈ 1.2 kg m−3 , and γ = 7/5 for a diatomic gas, I get 340 m/s.
Looking for plane wave solutions, ie. perturbations ∝ e−iωt+k·r , we find a dispersion
relation
ω 2 = c2s k2 .
The linear dispersion relation ω ∝ k means that these waves are non-dispersive. They
have frequency-independent and equal phase and group velocities: the phase velocity
is ω/k = cs and group velocity is dω/dk = cs .
1 Note that in general, the sound speed is
∂P
c2s =
∂ρ
with the partial derivative taken under whatever conditions are appropriate for the perturbations. We
considered adiabatic perturbations so the derivative is taken at constant entropy. When heat transfer
is very rapid for example, we would keep temperature constant when taking the derivative, giving the
isothermal sound speed c2T = P/ρ.

1
 Things get more complicated when the fluid is magnetized. As we discussed in
the first week, a magnetized plasma has a magnetic pressure that acts perpendicular
to the field lines. Acoustic waves that are travelling across the magnetic field lines
experience an extra restoring force and travel more quickly. For the perpendicular
case k ⊥ B, the dispersion relation is

ω 2 = k2 (c2s + v2A )

where
1
vA = 󰁳 B
4πρ
is the Alfven velocity. This mode is known as the fast magnetosonic mode. An acoustic
wave travelling along the field direction k 󰀂 B does not feel the magnetic pressure
and has the usual dispersion relation ω 2 = c2s k2 . These are known as slow magnetosonic
modes.
Just to give a bit of the flavour of the calculation, the magnetic field enters through
the J × B force. If the background field is uniform with J = 0, the perturbations give a
J × B force
δJ × B (∇ × δB) × B i
= = (k × δB) × B.
c 4π 4π
We also need the induction equation

iωδB = ∇ × (δv × B).

With these two extra ingredients, you can show that (try it!)
󰀅 󰀆
δv(ω 2 − (k · v A )2 ) − (k · δv) k(c2s + v2A ) − v A (k · v A ) + k(k · v A )(δv · v A ) = 0.

This is quite a complicated dispersion relation, which why it helps to think about
particular limits. Setting k · v A = 0 makes several terms vanish, and you can straight-
forwardly show that ω 2 = k2 (c2s + v2A ), the fast magnetosonic mode mentioned above.
If instead we assume k 󰀂 v A , then things simplify to

δv(ω 2 − k2 v2A ) = (k · δv)k(c2s − v2A ). (3)

Dotting this equation with k gives ω 2 = k2 c2s the slow magnetosonic wave mentioned
earlier.
We should mention that there is also a non-compressive wave in the magnetized
case, the Alfven wave. The tension of magnetic field lines supports a transverse wave
similar to a wave on a string. To see this, set k · δv = 0 (an incompressible perturba-
tion) in equation (3). There is a solution if

ω 2 = v2A k2

which is the dispersion relation for Alfven waves. You can use the induction equation
to show that for these waves k · δB = 0, ie. they are transverse to the magnetic field.
They propagate at the Alfven speed v A .

2
Compressible vs. incompressible flow
An important point to make is that compressibility is a flow property as well as a
material property. Flows that are subsonic, with fluid velocities much smaller than the
sound speed, are incompressible with ∇ · v ≈ 0, even though the material itself may
be compressible. A way to think of this is that there is plenty of time for compressions
to be smoothed out by propagation of sound waves if the flow is subsonic.
A simple illustration is given by a steady 1D isentropic flow. Isentropic means
that we can write the pressure gradient term as ∂P/∂x = c2s ∂ρ/∂x, where cs is the
isentropic sound speed. The momentum equation is then
dv c2 dρ
v =− s
dx ρ dx
v dρ v2
⇒ =− 2
ρ dv cs
1 d v2
⇒ (ρv) = 1 − 2 . (4)
ρ dv cs
Equation (4) shows that for subsonic flow, the mass flux ρv increases with velocity.
This is what we would expect for an incompressible flow: at constant density, if you
move faster the mass flux is larger. But note what happens at speeds faster than the
sound speed. Then, the mass flux decreases as the flow speed increases. Despite
moving faster, the density drops giving a smaller mass flux.
“Real-life” examples of these two limits are a river, which flows faster when the
river narrows or slower when the river widens, and traffic on the freeway, which flows
faster when the road widens and slows when the road narrows.

Steepening
When deriving the sound speed, we considered linear waves, ie. small perturbations
to a background state. However, we know that the fluid equations have a non-linear
term (v.∇)v, so that for large amplitudes it is not very useful to write the flow as a sum
of plane waves. Whereas in a linear problem the plane waves evolve independently,
and so it makes sense to use a Fourier decomposition, the non-linear terms couple the
amplitudes of the different modes.
An important effect of the non-linear terms is that they lead to steepening of the
velocity profile. We can see this by looking at the equation
∂v ∂v
+v = 0. (5)
∂t ∂x
The general solution to this equation is
v = f ( x − vt) = f (ξ ),
where f (ξ ) is some arbitrary function of ξ = x − vt. To see this, change variables
󰀕 󰀖
∂v d f ∂ξ ′ ∂v
= = f −v − t
∂t dξ ∂t ∂t

3
 ∂v −v f ′
⇒ = .
∂t 1 + f ′t
Similarly,
∂v f′
= . (6)
∂x 1 + f ′t
Combining these derivatives, we see that equation (5) is indeed satisfied.
More importantly, we see an interesting behaviour in the spatial derivative given
by equation (6). An initial profile with ∂v/∂x |t=0 = f ′ < 0 will reach ∂v/∂x → ∞
after a time 󰀏 󰀏 󰀏 󰀏
󰀏 1󰀏 󰀏 1 󰀏
t = 󰀏󰀏− ′ 󰀏󰀏 = 󰀏󰀏− 󰀏
f ∂v/∂x |t=0 󰀏
which we can think of as a local “turnover time” for the fluid.
The profile steepens in as illustrated in the sketch below.

A shock forms in which the velocity v changes its value on a very short lengthscale.
The thickness of the shock is set by the viscous term in the momentum equation which
becomes important as dv/dx becomes large. Viscous stresses act to smooth out the
velocity gradient and eventually will balance the steepening from the non-linear term.
The lengthscale on which this happens is very short, of order the microscopic mean
free path. In fact, we don’t need to understand the details of what happens inside the
shock, we can instead treat the shock as a discontinuity and relate the fluid velocity,
density and temperature on each side using conservation of mass, momentum and
energy. We do that in the next section.

4
 A classical example of a situation in which a shock forms is the “shock tube” in
which a piston moves into a cylinder. A shock propagates ahead of the cylinder, ac-
celerating the fluid from rest to the speed of the cylinder, and at the same time com-
pressing the gas. Another situation is supersonic flow around an object. A shock
forms which acts to slow the fluid from supersonic to subsonic. The fact that the flow
is subsonic near the object means that the sound crossing time can be shorter than the
flow time, in this way the fluid flow around the obstacle.

Shock jump conditions

To derive the shock jump conditions, also known as the Rankine-Hugoniot relations,
we first move into the frame of the shock as illustrated below. On the left, you see the
shock moving to the right at speed vs ; on the right in the shock frame the unshocked
fluid is moving to the left at speed vs . Across the shock, the fluid changes velocity
from v1 = −vs to v2 , and density and pressure change from values ρ1 and P1 to ρ2 and
P2 .

5
 We then integrate the fluid equations across the shock. For a steady 1D flow, con-
tinuity is
∂
(ρv) = 0.
∂x
Integrating,
󰁝 󰂃
∂
dx (ρv) = [ρv]󰂃−󰂃 = 0
−󰂃 ∂x
or
ρ1 v1 = ρ2 v2 . (7)
Momentum is
dv d dP
ρv = (ρv2 ) = − ,
dx dx dx
which when integrated gives

P1 + ρv21 = P2 + ρv22 . (8)

The total energy equation is

󰀗 󰀕 󰀖󰀘
d 1 2
v ρv + ρe + P =0
dx 2
1 P 1 P2
⇒ v21 + e1 + 1 = v22 + e2 + .
2 ρ1 2 ρ2
For an ideal gas, P = (γ − 1)ρe, so we can rewrite this
1 2 γ P1 1 γ P2
v1 + = v22 + . (9)
2 γ − 1 ρ1 2 γ − 1 ρ2
Equations (7)-(9) are the shock jump conditions, relating the “upstream” conditions
(v1 , ρ1 , P1 ) to the “downstream” ones (v2 , ρ2 , P2 ) .
The jump conditions can be combined to derive a number of useful results. One of
them is
ρ2 v (γ + 1) M12
= 1 =
ρ1 v2 2 + (γ − 1) M12
where M1 = u1 /c1 is the upstream Mach number, the shock velocity divided by the
upstream sound speed. This shows that there is a maximum compression which oc-
curs for a strong shock (M1 ≫ 1), ρ2 /ρ1 = (γ + 1)/(γ − 1). This compression factor
is 4 for a monatomic gas (γ = 5/3).

6
 While the compression is limited, note that the pressure and therefore temperature
jump can be large. The pressure jump is
P2 2γM12 − (γ − 1)
=
P1 γ+1
which is ∝ M12 for a strong shock.
The P2 –ρ2 relation is known as the shock adiabat or the Hugoniot curve. But note
that the flow across the shock is definitely not adiabatic! There is a large jump in
entropy as the ordered kinetic energy of the rapid upstream flow is converted into
heat in the compressed slow-moving gas downstream. For example, for a strong shock
with γ = 5/3 you should be able to show that the downstream temperature is
k B T2 3
= v2s .
µ1 m p 16
More complex cases that you could look at are:
• an oblique shock, in which the flow direction is not perpendicular to the shock.
These occur in flow around an object, where the shocks help to redirect the fluid.
• a magnetized shock. As you might expect from our discussion of fast and slow
magnetosonic waves, the direction of the magnetic field relative to the shock
front makes a difference. A magnetic field perpendicular to the flow and paral-
lel to the shock is compressed and gives an extra pressure that must be included
in the jump conditions. There is also a jump condition on B coming from inte-
grating the induction equation across the shock. For example, you can show that
the ratio B/ρ is the same on both sides when the magnetic field is parallel to the
shock. Compression of the fluid also implies a larger field strength because of
magnetic flux conservation.
• a radiative shock. Shocks in astrophysics are often very radiative: the tempera-
ture immediately after the shock is so great that it leads to rapid cooling of the
shocked gas. The net result can be much larger compression factors than in the
strong shock case. A limit to consider is the isothermal shock in which the cooling
is strong enough to equalize the temperature of the pre-shock and post-shock
gas. Then the compression ratio is ρ2 /ρ1 = u21 /c2T where c T is the isothermal
sound speed.

Papers
• Goldreich, Murray, & Kumar (1994) “Excitation of acoustic modes in the Sun”
http://adsabs.harvard.edu/abs/1994ApJ...424..466G
• Bazer and Ericson (1957) “Hydromagnetic shocks”
http://adsabs.harvard.edu/abs/1959ApJ...129..758B
• Taylor (1950) “The formation of a blast wave by a very intense explosion”
http://adsabs.harvard.edu/abs/1950RSPSA.201..159T
http://adsabs.harvard.edu/abs/1950RSPSA.201..175T