38

7. SIGNAL PROPAGATION Thus, δv > 0 if δρ > 0; the passage of a compression

In this chapter we introduce a very important basic tool:
the speed at which a disturbance propagates through
a fluid. Because the disturbance we’re considering is
compressive – think of a locally overpressure region –
one example is a sound wave. Thus the signal speed is
called the sound speed: we’ll see that it’s given by
∂p
c2s = (7.1)
∂ρ
First we’ll derive this important speed, then look at how
causality can dramatically change the nature of super-
sonic flows.
A. Sound Waves and the Signal Speed
This is worth two derivations; one physical, and another
one more mathematical.
SOUND WAVES : A PHYSICAL APPROACH
wave leaves behind a fluid moving in the direction of
the wave. Now, apply momentum balance: the net force
on some control volume, from the pressure difference,
equals the rate of change of momentum in that volume.
That is,
p − (p + δp) = ρcs [(cs − δv) − cs ]
so that, again to first order small, we have
δp ≃ ρcs δv (7.3)
Combining (7.2) and (7.3), we get a condition on the
wave speed (to allow mass and momentum balancs):
δp
c2s = (7.4)
δρ
Thus, we have one justification of our expression (7.1)
for cs .

We can also demonstrate that cs is the speed of a travel- SOUND WAVES : A FORMAL APPROACH
ling wave. Let some perturbation (δρ, δp, δT ) be mov-
We can also use a more mathematical approach, deriv-
ing at some cs . Ahead of the wave the fluid has v = 0;
ing a formal wave equation by linearizing our two basic
behind the wave the fluid has δv, in the same direction
equations, continuity & momentum:
as the wave motion.
∂ρ
+ ∇ · (ρv) = 0 (7.5)
∂t
(assuming no body forces)
∂v
ρ + ρ(v · ∇)v = −∇p (7.6)
∂t
Our question is, at what speed does a small, compres-
sional disturbance in the gas travel? We start with a
uniform, static unperturbed state, described by ρo , po
and vo = 0. We add small perturbations, ρ1 , p1 and v1 .
If we put these into the mass and momentum equations,
(7.5) and (7.6), noting that the “o” terms are constant in
space and in time, we get
∂ρ1
Figure 7.1. Propagation of a sound wave: (a) into a still + v1 · ∇ρo + (ρo + ρ1 )∇ · v = 0
∂t
fluid; (b) a stationary wave. From Kundu figure 15.1. Note,
u here is the same as v in the text. and
∂v1
Mass conservation at the wave front, in a frame moving (ρo + ρ1 ) + (ρo + ρ1 )v1 · ∇v1 = −∇p1
with the wave front, gives ∂t

ρcs = (ρ + δρ)(cs − δv) In these equations, we now drop terms which are sec-
ond order in the perturbed quantities, and we write
and to lowest order small, this gives ∇p = (∂p/∂ρ)∇ρ. This gives us
δρ ∂ρ1
δv ≃ cs (7.2) + ρo ∇ · v1 = 0 (7.7)
ρ ∂t
 39

and B. Why is the sound speed important?

 
∂v1 ∂p I’m storing two important ideas in this section.
ρ0 + ∇ρ1 = 0 (7.8)
∂t ∂ρ
1. WHEN CAN WE ASSUME INCOMPRESSIBLE
Two approaches are possible here. FLOW ?
• Method # 1. This is more physically intuitive.
From this pair of equations we can, for instance, elimi- In the first five chapters, we assumed incompressible
nate v1 , to get a second order DE in ρ1 : flow, and noted that this is a good approximation when
the flow speed is much less than the sound speed. To
∂ 2 ρ1
 
∂p start, we justify this assumption. We begin with the
2
= ∇2 ρ1 (7.9)
∂t ∂ρ continuity equation:
This, of course, has travelling wave solutions: the dis- ∇ · (ρv) = 0 (7.13)
turbances travel at a speed given by
 1/2 This reduces to the incompressible condition, ∇·v = 0,
∂p if (in 1D Cartesian for simplicity),
cs = (7.10)
∂ρ
∂ρ ∂v
This recovers our guess in (7.1). v ≪ρ ; vδρ ≪ ρδv (7.14)
∂x ∂x
• Method # 2. This is more formal, and a sim-
Now, the momentum/Euler equation becomes,
ple example of a general method which is the standard
approach when the situation is more complex. (This 1
will be useful in the homework for instance). Refer vδv ≃ δp (7.15)
ρ
back to equations (7.7). They are linear in the per-
turbations, which allows us to use Fourier techniques. If we now use the sound speed, from (7.1), we have
That is, we can consider one simple perturbation, say δp ≃ c2s δρ, and thus
ρ1 , v1 ∝ ei(k·x−ωt) ; and we know that any arbitrary
perturbation can be expressed as a sum of these (ω, k) δρ v 2 δv
≃ 2 (7.16)
waves. ρ cs v
With this, taking (7.1) as a definition (only that, for We thus find that the density changes are negligible –
the moment), and dropping down to a 1D perturbation and thus we can work in the incompressible limit –
to simplify, equations (7.7) and (7.8) become when
v2
−iωρ1 + ρo ikv1 = 0 = M2 ≪ 1 (7.17)
(7.11) c2s
−iωρo v1 + ikc2s ρ1 = 0
Here, we have introduced the Mach number, M =
From these we immediately find the dispersion rela- v/cs .
tion: 2. THE IMPORTANCE OF CAUSALITY
2
ω = c2s k2 (7.12) We have just demonstrated that the sound speed is the
Thus, we have verified that we have non-dispersive speed at which information propagates; it is also critical
waves propagating at speed cs . This again verifies our to the dynamics of the flow.
guess in (7.1). There are important difference between subsonic
The derivative in (7.1) is usually taken assuming the and supersonic flows. Subsonic flows can be thought of
disturbance is adiabatic, so that p/ργ is constant, and as quasi-hydrostatic. That is, the flow field is strongly
c2s = γp/ρ. This corresponds, physically, to the heat- influenced by pressure gradients which are determined
ing and cooling times for the perturbation to be long by conditions a long distance away (such as at bound-
compared to the wave travel time. An alternative choice aries).
which is sometimes used, is to consider isothermal per- Supersonic flows, however, are quasi-ballistic. This
turbations: T = constant – which is the limit in which distinction is due to the fact that information travels at a
the local cooling and heating times are short compared finite speed, the speed of a simple sound wave. Both 1D
to the wave travel time. In this case, c2s = p/ρ. and 3D cartoons, in Figures 7.1 and 7.2, can illustrate
 40

this point. Pressure gradients have only a limited range • hypersonic: M ∼ < 3, say. Qualitatively the same as

of influence, and conditions far away have little or no
effect on a solution locally. We’ll see that supersonic
flows can (and usually do) contain discontinuous jumps
in the flow properties (shocks). They can violate our
subsonic intuition, for instance a supersonic flow in a
diverging channel will accelerate (as we’ll see below).
supersonic flow, but some interesting new effects – such
as strong heating and ionization of boundary layers –
come into play. Shocks can be analyzed in the strong-
shock limit (chapter 9).
C. Weak Waves and Causality

perturbation
How can we use the signal speed – the sound speed –
Undisturbed flow to understand a flow? One way to approach this, fol-
cs cs lowing Currie, is to consider “weak waves”. Specify to
reverse waves forward waves
a 1D system, and let co be the undisturbed value of cs .
Figure 7.2. Physical illustration of simple waves. The Equations (7.7) and (7.8) become
information that the flow has been “whacked” at point a, ∂ρ1 ∂v1 ∂v1 ∂ρ1
propagates by simple sound waves, moving at speed cs + ρo = 0 ; ρo + c2s =0
relative to the fluid in the pipe. Following Thompson figure ∂t ∂x ∂t ∂x
8.6 In the previous derivation, cs can be a function of den-
sity, and thus can vary within the wave. Here, we sim-
plify by taking cs to be a constant, co . But now: be-
cause ρo is assumed constant, and because vo = 0, we
ct
s can write
vt vt

∂ρ ∂v ∂v ∂ρ
cst
+ ρo = 0 ; ρo + c2o =0 (7.18)
∂t ∂x ∂t ∂x
M<1 M>1
M=0

Figure 7.3. Mach’s construction for the propagation of a

Compare these to the originals, (7.7) and (7.8): we have
disturbance. Consider a point source of sound (Thompson quietly removed the bothersome, nonlinear terms. This
suggests a bumblebee) in a moving medium. If the source is valid only as long as we continue to assume a small
and flow are stationary, the sound propagates spherically perturbation – a “weak wave”. We can now divide the
from the source. If the source/flow are moving subsonically. first of (7.18) by ρo , the second by ρo co , and add and
the motion only distorts the spherical wavefronts. If, subtract, to get
however, the motion is supersonic. all disturbances are    
confined to a Mach cone; an observer located outside of this ∂ v ρ ∂ v ρ
cone does not receive any information about the bee. The + + co + =0 ;
opening angle of the cone is called the Mach angle:
∂t co ρo ∂x co ρo
   
sin µ = 1/M. ∂ v ρ ∂ v ρ
− − co − =0
∂t co ρo ∂x co ρo (7.19)
Some authors use the following classifications,
These now have the form of a total (Lagrangian) deriva-
based on the Mach number M = v/cs (I follow Kundu,
tive (compare equation 1.5). Thus, we have an impor-
for instance):
tant result:
• incompressible: M ∼ < 0.3 everywhere; can ignore  
density variations due to pressure changes. v ρ
+ = constant on x − co t = constant
• subsonic: 0.3 ∼< M < 1 everywhere. Need to be co ρo
∼ (7.20)
careful with density fluctuation (now above the 10%
and
level), but no shock waves in the flow.  
• transonic: 0.8 ∼< M < 1.2, say; M is “around v ρ
∼ − = constant on x + co t = constant
unity”. Shock waves appear, increasing drag. These are co ρo
the hardest flows to analyze analytically, as the nonlin- (7.21)
ear terms in the governing equations are important, but An alternate form of (7.20 and 7.21) can be found with
the simplifying effects of M > ∞ flow can’t be used a bit of algebra. Using (7.7) and ρ1 ≪ ρo , it’s easy to
yet. express ρ/ρo in terms of p/po and γ, to write
< M < 3, typically. Shock waves
 
• supersonic: 1 ∼ ∼ + v 1 p
J = + = constant
are usually present. Analysis is easier because informa- co γ po (7.22)
tion propagates along well-defined directions in (x, t)
space, called characteristics. on x − co t = constant
 41

diaphragm
and
 
v 1 p p1 p0
J− = − = constant
co γ po (7.23)
on x + co t = constant t

The interpretation is simple for (7.20) or (7.22). The x = -ct x = ct

lines x = ±co t are the loci of forward and backward (3) P (3)

propagating sound waves. Our results say that the quan- (2)
tities on the left of equations (7.21) and (7.23) are con- (1)

stant along these trajectories.

D. Two examples of simple waves x
Figure 7.4. Illustrating the geometry of the
Two standard examples show how this analysis can be ruptured-diaphragm problem, in the x,t plane. The two solid
used. lines are the forward and reverse waves which start at the
origin; the two dotted lines are the waves which intersect
1. SHOCK TUBE the observation point P. Following Currie Figure 11.2.

Consider a gas confied to a 1D tube. A diaphragm at 2. PISTON PROBLEM

x = 0 separates high-pressure (p1 , x < 0) gas from
low-pressure ( p0 , x > 0) gas. At t = 0 the diaphragm Consider another 1D tube filled with gas; at t = 0 a
breaks, and the high-pressure gas begins to expand into piston is moved into the gas, at velocity U . How does
the low-pressure gas. The information that this has hap- the gas respond?
pened can only travel at the sound speed, co . Thus, a In Figure 7.5 we show this problem in the (x, t)
compression wave will move to the right, and an ex- plane. Once again, we expect the gas to be undisturbed
pansion wave to the left, both at co . The problem: what (pressure po , zero velocity) ahead of the forward-wave
is the velocity and pressure of the gas everywhere, as a which starts from the piston at t = 0. Also again, con-
function of time? sider an observation point P (x, t). It connects the the
Figure 7.4 shows an (x, t) diagram for this system, forward x-axis by a reverse wave; we can evaluate the
with the two wave paths as solid lines. Gas ahead of constant J − , because we know v = 0, p = po at the
the forward wave must be undisturbed: it has p = x-axis. The point P also connects to the piston by a
po , v = 0. Similarly, gas behind the reverse wave has forward wave. We cannot evaluate J + + yet, however;
p = p1 , v = 0. What of the middle region? Con- we know the velocity vp = U , but not the pressure, pp ,
sider an observation point P (xp , tp ). Two wave lines at the piston. We thus need a third wave, another re-
intersect P : one forward wave which originates from verse one which connects the piston to the x-axis. We
(x < xp , t = 0), and a reverse wave which originates thus have three algebraic equations:
from (x > xp , t = 0). But now: we know v = 0 ev- U 1 pp 1
erywhere at t = 0, so we can evaluate the constants in − =−
co γ po γ
(7.22, 7.23) for each of these wave lines. For the re-
verse one, we have J − = −1/γ; and for the forward v 1 p U 1 pp
+ = + (7.25)
one, we have J + = p1 /γpo (verify this for yourself!). co γ po co γ po
Thus, at point P we know that v 1 p 1
− =−
v 1 p 1 p1 v 1 p 1 co γ po γ
+ = and − =−
co γ po γ po co γ po γ This now solves the system: we can find pp , and then
solve for conditions in the gas between the piston and
These contain enough information to solve the system.
the forward wave:
We find that
co p1 − po p1 + po p U
v= ; p= (7.24) v=U ; =γ +1 (7.26)
2γ po 2 po co
Thus: the region inbetween the two waves has a uni- 3. WAVES AT BOUNDARIES
form pressure, equal to the mean of p1 and po ; and it
moves to the right at a uniform speed, proportional to Finally, a comment about wave reflection at boundaries.
the initial pressure difference. Waves reflect in like manner off solid boundaries. This
 42

U
(piston)

t
x=Ut

P
x=ct
(v=U)

x
(v=0, p=p0)
Figure 7.5. Illustrating the geometry of the piston
problem, in the x,t plane. The two solid lines are the locus
of the piston (assumed to be close to vertical; U ≪ co ), and
the forward wave starting from the origin. The dotted lines
are the two waves which intersect the observation point P ,
and the third (reverse) wave used to connect the piston to
the x > 0 axis. Following Currie Figure 11.5.

means that at a solid wall, a compression wave reflects

as a compression wave, and an expansion wave reflects
as an expansion wave. Free boundaries are different;
waves reflect in an unlike manner. That means that ex-
pansion waves reflect as compression waves, and vice
versa. The reason is that along a solid boundary, the
flow direction is the boundary condition, while off a
free boundary the pressure is the boundary condition.

References
Good references here include Thompson, Currie,
Kundu and Faber.