You are on page 1of 49

AAS

1. INTRODUCTION
1.1 Sources of sound
These notes are concerned with the mechanisms of sound production and propagation.
Music, calm speech, gently owing streams or leaves uttering in a breeze on a summer day
are pleasant and desirable sounds. Noise, howling winds, explosions and angry voices are
possibly less so. In all such cases we seek to understand the basic sources of the sound by
careful analysis of the equations of motions. Most mechanical sources are very complex, very
often involving ill-dened turbulent and perhaps combusting ows and their interactions
with vibrating structures, and the energy released as sound is usually a tiny fraction of
the structural and hydrodynamic energy of the source region. It is therefore important to
develop robust mathematical models that reliably incorporate this general ineciency of
the sound generation mechanisms; small errors in source modelling can evidently lead to
very large errors in acoustic prediction.
To do this we consider a uid that can be regarded as continuous and locally
homogeneous at all levels of subdivision. At any time t and position x = (x
1
, x
2
, x
3
) the
state of the uid is dened when the velocity v and any two thermodynamic variables are
specied for each of the uid particles of which the uid may be supposed to consist. The
distinctive uid property possessed by both liquids and gases is that these uid particles
can move freely relative to one another under the inuence of applied forces or other
externally imposed changes at the boundaries of the uid. Five scalar partial dierential
equations are required to determine these motions. They are statements of conservation of
mass, momentum and energy, and they are to be solved subject to appropriate boundary
and initial conditions, dependent on the problem at hand. These equations will be used to
formulate and analyse a wide range of acoustic problems; our main task will be to simplify
these problems to obtain a thorough understanding of source mechanisms together with a
quantitative estimate of the radiated sound.
M. S. Howe 1 1.1 Sources of sound
AAS
1.2 Equations of motion of a uid
The state of a uid at time t and position x = (x
1
, x
2
, x
3
) is dened when the velocity
v and any two thermodynamic variables are specied. Five scalar equations are therefore
required to determine the motion. These equations are statements of the conservation of
mass, momentum and energy. They must normally be supplemented by a thermodynamic
equation of state.
1.2.1 The material derivative
Let v
i
denote the component of uid velocity v in the x
i
-direction, and consider the rate
at which any function F(x, t) varies following the motion of a uid particle. Let the particle
be at x at time t, and at x+x a short time later at time t +t, where x = v(x, t)t + .
At the new position of the uid particle
F(x + x, t + t) = F(x, t) + v
j
t
F
x
j
(x, t) + t
F
t
(x, t) + ,
where the repeated sux j implies summation over j = 1, 2, 3. The limiting value of
(F(x + x, t + t) F(x, t))/t as t 0 denes the material (or Lagrangian) derivative
DF/Dt of F:
DF
Dt
=
F
t
+ v
j
F
x
j

F
t
+v F. (1.2.1)
DF/Dt measures the time rate of change of F as seen by an observer moving with the uid
particle that occupies position x at time t.
1.2.2 Equation of continuity
A uid particle of volume V and mass density has a total mass of V (V )(x, t),
where x denotes the position of the centroid of V at time t (Figure 1.2.1). Conservation of
mass requires that D(V )/Dt = 0, i.e. that
1

D
Dt
+
1
V
DV
Dt
= 0. (1.2.2)
Now
1
V
DV
Dt
=
1
V
_
S
v dS, where the integration is over the closed material surface S forming
the boundary of V, on which the vector surface element dS is directed out of V. It is the
M. S. Howe 2 1.2 Equations of motion
AAS
fractional rate of increase of the volume of the uid particle, and becomes equal to div v as
V 0. In this limit (1.2.2) can therefore be cast in any of the following equivalent forms of
the equation of continuity
1

D
Dt
+ div v = 0,

t
+ div(v) = 0,

t
+

x
j
(v
j
) = 0.
_

_
(1.2.3)
Figure 1.2.1
1.2.3 Momentum equation
The momentum equation is derived in a similar manner by consideration of the rate of
change of momentum D(V v)/Dt (V )Dv/Dt of a uid particle subject to pressure p
and viscous forces acting on the surface S of V, and body forces F per unit volume within
V. It is sucient for our purposes to quote the form of the resulting Navier-Stokes equation

Dv
Dt
= p curl +
_

+
4
3

_
div v +F, (1.2.4)
where = curl v is the vorticity, and and

are respectively the shear and bulk


coecients of viscosity, which generally vary with pressure, temperature and position in
the uid. Viscous forces are important predominantly close to solid boundaries, where the
frictional drag is governed by the shear viscosity . It is then a good approximation to
adopt the Stokesian model in which the contribution of the bulk viscosity is ignored.
Values of , and = / (the kinematic viscosity) for air and water at 10

C and one
atmosphere pressure are given in the Table 1.2.1:
M. S. Howe 3 1.2 Equations of motion
AAS
kg/m
3
kg/ms m
2
/s
Air 1.23 1.764 10
5
1.433 10
5
Water 1000 1.284 10
3
1.284 10
6
Table 1.2.1
1.2.4 Energy equation
Consideration of the transfer of heat by molecular diusion across the moving material
boundary S of the uid particle of Figure 1.2.1, and of the production of heat by frictional
dissipation within its interior volume V leads to the energy equation
T
Ds
Dt
= 2
_
e
ij

1
3
e
kk

ij
_
2
+

e
2
kk
+ div
_
T
_
, (1.2.5)
where T, s are respectively the temperature and specic entropy (i.e. entropy per unit
mass), is the thermal conductivity of the uid, and
e
ij
=
1
2
_
v
i
x
j
+
v
j
x
i
_
(1.2.6)
is the rate of strain tensor, which accounts for changes in shape of the uid particle. Because
e
kk
div v, the term involving the bulk coecient of viscosity

in (1.2.5) determines the


dissipative production of heat during compressions and rarefactions of the uid.
To understand the signicance of the rate of strain tensor, observe that the velocity v

relative to the centroid of the moving uid particle at vector distance x

from its centroid,


is given to rst order by (see A.7)
v

i
= x

j
v
i
x
j
x

j
1
2
_
v
i
x
j
+
v
j
x
i
_
+ x

j
1
2
_
v
i
x
j

v
j
x
i
_
=
1
2

i
_
e
jk
x

j
x

k
_
+
1
2
( x

)
i
where e
ij
and the vorticity are evaluated at the centroid. The term in therefore repre-
sents relative rigid body rotation of the particle about the centroid, at angular velocity
1
2
,
M. S. Howe 4 1.2 Equations of motion
AAS
with no change of shape. The gradient term, however, represents an irrotational distortion
of V, and is responsible for frictional forces and the conversion of mechanical energy into heat.
1.2.5 Equation of state
In the presence of velocity and pressure gradients a uid cannot be in strict
thermodynamic equilibrium, and thermodynamic variables require special interpretation.
The density and the total internal energy e per unit mass can be dened in the usual
way for a very small uid particle without the need for thermodynamic equilibrium, such
that and e are the mass and internal energy per unit volume. The pressure and all
other thermodynamic quantities are then dened by means of the same functions of
and e that would be used for a system in thermal equilibrium, and the relations between
the thermodynamic variables are then the same as for a uid in local thermodynamic
equilibrium, dened by equations of the form
p = p(, s), p = p(, T), s = s(, T), etc. (1.2.7)
The equations of state (1.2.7) permit any thermodynamic variable to be expressed in terms
of any two variables, such as the density and temperature, although in applications it may
be more convenient to use other such equations. For sound propagation in an ideal uid it
is usual to neglect dissipation and to assume homentropic ow: s = s
o
= constant. This
permits the uid motion to be determined from the equations of continuity and momentum
and the equation of state p = p(, s
o
), the energy equation being ignored. In more general
situations it is necessary to retain the energy equation to account for coupling between
macroscopic motions and the internal energy of the uid.
Note, however, that the thermodynamic pressure p = p(, e) dened in this way is
generally no longer the sole source of normal stress on any surface drawn in the uid.
This is the case in a uid of non-zero bulk viscosity (

,= 0), whose molecules possess


rotational (or other internal) degrees of freedom whose relaxation time required for the
M. S. Howe 5 1.2 Equations of motion
AAS
re-establishment of thermal equilibrium after, say, a compression, is large relative to the
equilibration time of the translational degrees of freedom. When compressed (during an
interval in which div v < 0) the temperature must rise, but the corresponding increase
in the rotational energy lags slightly behind that of the molecular translational energy
responsible for normal stress; the thermodynamic pressure p accordingly is smaller than the
true normal stress (which equals p

div v).
1.2.6 Croccos equation
The momentum equation (1.2.4) can be recast by introducing the specic enthalpy
w = e + p/ of the uid, in terms of which the rst law of thermodynamics supplies the
relation
dw =
dp

+ Tds. (1.2.8)
The vector identity (v )v = v +
_
1
2
v
2
_
and (1.2.8) permit the momentum equation
to be put in Croccos form
v
t
+B = v + Ts curl +
_

+
4
3

_
div v +
F

, (1.2.9)
where

/ and
B = w +
1
2
v
2
(1.2.10)
is the total enthalpy. In a perfect gas w = c
p
T = p/( 1), where c
p
is the specic
heat at constant pressure and = c
p
/c
v
, c
v
being the specic heat at constant volume.
Croccos equation nds application in the acoustics of turbulent, heat conducting ows.
M. S. Howe 6 1.2 Equations of motion
AAS
1.3 Sound waves in an ideal uid
The intensity of an acoustic pressure p in air (relative to the mean atmospheric pressure)
is usually measured in decibels by the quantity
20 log
10
_
[p[
p
ref
_
,
where the reference pressure p
ref
= 2 10
5
Pa. Thus, p = p
o
1 atmosphere (= 10
5
Pa)
is equivalent to 194 dB. A very loud sound 120 dB corresponds to
p
p
o

2 10
5
10
5
10
(
120
20
)
= 2 10
4
1.
Similarly, for a deafening sound of 160 dB, p/p
o
0.02. This corresponds to a pressure of
about 0.3 lbs/in
2
, and is loud enough for nonlinear eects to begin to be important.
The passage of a sound wave in the form of a pressure uctuation is accompanied by a
back-and-forth motion of the uid in the direction of propagation at the acoustic particle
velocity v, say. It will be seen (1.3.2) that
acoustic particle velocity
acoustic pressure

o
speed of sound
,
where
o
is the mean air density. The speed of sound in air is about 340 m/sec. Thus,
v 5 cm/sec at 120 dB ; at 160 dB v 5 m/sec.
1.3.1 The wave equation for an ideal uid
In most applications the acoustic pressure is very small relative to p
o
, and sound
propagation is studied by linearizing the equations of motion. We consider rst the
simplest case of sound propagation in an ideal uid i.e. a homogeneous, inviscid,
non-heat-conducting uid of mean pressure p
o
and density
o
, which is at rest in the
absence of the sound. The energy equation (1.2.5) implies that Ds/Dt = 0 so that sound
propagation is homentropic (adiabatic) with s = s
o
s(p
o
,
o
) = constant throughout uid.
The implication is that, in an ideal uid there is negligible dissipation of the organized
M. S. Howe 7 1.3 Ideal uid
AAS
mechanical energy of the sound by heat and momentum transfer by molecular diusion
between neighboring uid particles.
The departures of the pressure and density from their undisturbed values are denoted
by p

where p

/p
o
1,

/
o
< 1. The linearized form of the momentum equation
(1.2.4) for an ideal uid ( =

= 0) then becomes

o
v
t
+p

= F. (1.3.1)
Before linearizing the continuity equation (1.2.3) it is useful to make an articial
generalization by inserting a volume source distribution q(x, t) on the right hand side:
1

D
Dt
+ div v = q; (1.3.2)
q is the rate of increase of uid volume per unit volume of the uid, and might represent, for
example, the eect of volume pulsations of a small body in the uid (1.4). The linearized
equation is then
1

t
+ div v = q. (1.3.3)
Eliminate v between (1.3.1) and (1.3.3):

t
2

2
p

=
o
q
t
div F. (1.3.4)
An equation determining the pressure p

alone in terms of q and F is obtained by invoking


the homentropic approximation p = p(, s
o
), where p
o
= p(
o
, s
o
) in the undisturbed state.
Therefore
p
o
+ p

= p(
o
+

, s
o
) p(
o
, s
o
) +

(, s
o
), (1.3.5)
where the derivative is evaluated at the undisturbed value
o
of the density. It has the
dimensions of (velocity)
2
and its square root denes the speed of sound
c
o
=

_
_
p

_
s
, (1.3.6)
M. S. Howe 8 1.3 Ideal uid
AAS
where dierentiation is performed at s = s
o
, and evaluated at =
o
. In air c
o
340 m/s;
in water c
o
1500 m/s.
From (1.3.5):

= p

/c
2
o
, and substitution for

in (1.3.4) yields the inhomogeneous


wave equation
_
1
c
2
o

2
t
2

2
_
p =
o
q
t
div F, (1.3.7)
where the prime

on the acoustic pressure has been discarded. This equation governs the
production of sound waves by the volume source q and the force F. When these terms are
absent the equation describes sound propagation from sources on the boundaries of the
uid, such as the vibrating cone of a loudspeaker.
The volume source q and (with the exception of gravity) the body force F would never
appear in a complete description of sound generation in a real uid. They are introduced
only when we think we understand how to model mathematically the real sources of
sound in terms of idealized volume sources and forces. In general this can be a dangerous
procedure because, as we shall see, small errors in specifying the sources of sound in a uid
often result in very large errors in the predicted sound. This is because only a tiny fraction
of the available energy of a vibrating uid or structure actually radiates away as sound.
When F = 0 equation (1.3.1) implies the existence of a velocity potential such that
v = , in terms of which the perturbation pressure is given by
p =
o

t
. (1.3.8)
It follows from this and (1.3.7) (with F = 0) that the velocity potential is the solution of
_
1
c
2
o

2
t
2

2
_
= q(x, t). (1.3.9)
Causality can be invoked to justify the neglect of any a time-independent constants of
integration. Equation (1.3.9) is the wave equation of classical acoustics.
In the propagation zone, where the source terms q = 0, F = 0, the velocity v and the
perturbations in p, (and in other thermodynamic quantities such as the temperature T,
M. S. Howe 9 1.3 Ideal uid
AAS
internal energy e and enthalpy w, but not the specic entropy s, which remains constant and
equal to s
o
) propagate as sound governed by the homogeneous form of (1.3.9). The velocity
uctuation v produced by the passage of the wave is the acoustic particle velocity.
1.3.2 Plane waves
A plane acoustic wave propagating in the x-direction satises
_
1
c
2
o

2
t
2


2
x
2
_
= 0, (1.3.10)
which has the general solution (DAlembert 1747)
=
_
t
x
c
o
_
+
_
t +
x
c
o
_
, (1.3.11)
where , are arbitrary functions that respectively represent waves travelling at speed
c
o
without change of form in the positive and negative x-directions. The acoustic particle
velocity v = is parallel to the propagation direction (the waves are longitudinal).
The solutions (1.3.11) and the linearized (source-free) equations of motion can be used
to show that uctuations in v, p,

, T

and w

in a plane wave propagating parallel to the


x-axis are related by
v =
p

o
c
o
,

=
p
c
2
o
, T

=
p

o
c
p
, w

=
p

o
, (1.3.12)
where p is the acoustic pressure, c
p
is the specic heat at constant pressure, and the sign
is taken according as the wave propagates in the positive or negative x-direction.
1.3.3 Speed of sound
In a perfect gas p = RT and s = c
v
ln(p/

), where R = c
p
c
v
is the gas constant,
and in the linearized approximation
c
o
=
_
p
o
/
o
=
_
RT
o
. (1.3.13)
M. S. Howe 10 1.3 Ideal uid
AAS
Typical approximate speeds of sound in air and in water, and the corresponding acoustic
wavelength = c
o
/f at a frequency of f = 1 kHz are given in Table 1.3.1
c
o
at 1 kHz
m/s f/s km/h mph metres feet
Air 340 1100 1225 750 0.3 1
Water 1500 5000 5400 3400 1.5 5
Table 1.3.1 Speed of sound and acoustic wavelength
Example 1. Waves in a uniform tube generated by an oscillating piston The end
x = 0 of an innitely long, uniform tube is closed by a smoothly sliding piston executing
small amplitude normal oscillations at velocity u
o
(t) (Figure 1.3.1a). If x increases along
the tube, linear acoustic theory and the radiation condition require that p = (t x/c
o
).
At x = 0 the velocities of the uid and piston are the same, so that
u
o
(t)
(t)

o
c
o
, . . p =
o
c
o
u
o
(t x/c
o
) for x > 0.
In practice a solution of this kind, where energy is conned by the tube to propagate in
waves of constant cross-section, becomes progressively invalid as x increases, because of the
accumulation of small eects of ow nonlinearity. Nonlinear analysis reveals that at a point
in the wave where the particle velocity is v the wave actually propagates at speed c
o
+ v,
so that wave elements where v is large and positive produce wave steepening, resulting
ultimately in the formation of shock waves. This type of behaviour is important, for
example, for waves generated in a long railway tunnel by the piston eect of an entering
high-speed train.
Example 2. Reection at a closed end (Figure 1.3.1b) Let the plane wave
p = p
I
(t x/c
o
) approach from x < 0 the closed, rigid end at x = 0 of a uniform,
M. S. Howe 11 1.3 Ideal uid
AAS
semi-innite tube. The reected pressure p
R
(t + x/c
o
) is determined by the condition that
the (normal) uid velocity must vanish at x = 0. Therefore p
I
(t)/
o
c
o
p
R
(t)/
o
c
o
= 0,
and the overall pressure within the tube is given by
p = p
I
(t x/c
o
) + p
I
(t + x/c
o
), x < 0.
Reection at the rigid end causes pressure doubling at the wall where p = 2p
I
(t).
Figure 1.3.1
Example 3. Reection at an open end (Figure 1.3.1c) When the wavelength of the
sound is large compared to the radius R of an open ended circular cylindrical tube, the rst
approximation to the condition satised by the acoustic pressure at the open end (x = 0) is
that the overall pressure p = 0. Indeed, because of the free expansion of the uid outside
the tube, the pressure outside may be assumed to vanish compared to the incident pressure
M. S. Howe 12 1.3 Ideal uid
AAS
p
I
(t x/c
o
) p
I
(t). Then, integration of the linearized momentum equation over the
uid contained in a spherical region V of radius R
s
, where R R
s
= the acoustic
wavelength, reveals that
R
2
p

t
_
V

o
v
1
d
3
x
fV p
I
(t)
c
o
,
where v
1
O(p
I
(t)/
o
c
o
) is the particle velocity parallel to the tube, p is the net pressure
within the tube close to the open end, and f is the frequency of the sound.
Therefore, near the open end
p
fV
R
2
c
o
p
I
(t)
R
3
s
R
2

p
I
(t) p
I
(t) as =
c
o
f
.
Thus, relative to the incident pressure, the pressure at the open end when R may be
assumed to vanish. The pressure wave reected back into the tube at the end is therefore
approximately p
I
(t + x/c
o
),
p p
I
(t x/c
o
) p
I
(t + x/c
o
), x < 0,
and the acoustic particle velocity in the mouth of the tube 2p
I
(t)/
o
c
o
, twice that
attributable to the incident wave alone.
Example 4. Low frequency resonant oscillations in a pipe with open ends (Figure
1.3.1d) A pressure wave of complex amplitude p

and radian frequency propagating in


the x-direction has the representation p = Rep

e
i(tx/co)
. Therefore, the combination
p = Re
_
p

e
i(tx/co)
p

e
i(t+x/co)
_
vanishes at x = 0, and vanishes also at x = for those frequencies satisfying
e
i/co
e
i/co
= 0, i.e. sin(k
o
) = 0
where k
o
= /c
o
is called the acoustic wavenumber. This equation accordingly determines
the resonance frequencies of an open-ended tube of length in the low frequency
approximation in which the pressure is assumed to vanish at the ends, viz
f =

2

nc
o
2
, n = 1, 2, 3, . . . .
M. S. Howe 13 1.3 Ideal uid
AAS
Thus, the minimum resonance frequency of a pipe of length = 1 m is about 170 Hz
(c
o
340 m/s in air), whose wavelength c
o
/f = 2 = 2 m does not depend on the speed of
sound.
The simple, one dimensional theory neglects energy losses from the ends of the tube
by radiation into the ambient atmosphere, and neglects also viscous and thermal losses
in acoustic boundary layers at the walls of the tube. These cause the wave amplitude
to decay after several wave periods, so that resonant oscillations within the tube actually
persist only for a nite time after the source of excitation is removed.
M. S. Howe 14 1.3 Ideal uid
AAS
1.4 Low frequency pulsations of a sphere
Small amplitude, irrotational motion produced by radial pulsations of a sphere of mean
radius a with centre at the origin satisfy the homogeneous equation
1
c
2
o

t
2

2
= 0, r = [x[ > a. (1.4.1)
If the oscillations occur at frequency (proportional to e
it
, for example), then /t ,
and very close to the sphere the two terms on the left of this equation are respectively of
orders k
2
o
and /a
2
, where k
o
= /c
o
is called the acoustic wavenumber. If k
o
a 1 the
sphere is said to be acoustically compact. The characteristic wavelength 2/k
o
of
the sound produced by the pulsations is then much larger than the radius a. More generally,
a body is said to be acoustically compact when its characteristic dimension is small
compared to the wavelengths of the sound waves it is producing or with which it interacts.
In this limit the unsteady motion in the immediate neighbourhood of the sphere is governed
by the Laplace equation obtained by discarding the rst term on the left of (1.4.1):

2
= 0, r > a. (1.4.2)
This is just the continuity equation div v = 0 for incompressible ow. The corresponding
pressure and density perturbations p and satisfy

p
c
2
o
.
In an incompressible uid the pressure changes by the action of external forces (moving
boundaries, etc), but the density must remain xed. Thus, the formal limit of incompressible
ow corresponds to setting c
o
= .
M. S. Howe 15 1.4 Pulsating sphere
AAS
1.4.1 Pulsating sphere in incompressible uid
Let the normal velocity on the mean position r = a of the surface of the sphere be v
n
(t).
When the motion is incompressible we have to solve

2
= 0, r > a,
/r = v
n
(t), r = a
_

_
where r = [x[.
Figure 1.4.1
The solution must be radially symmetric, so that

2
=
1
r
2

r
_
r
2

r
_
= 0, r > a.
Hence
=
A
r
+ B, where A A(t), B B(t) are functions of t.
B(t) can be discarded, because the pressure uctuations (
o
/t) must vanish as
r , and a constant value of B has no physical signicance. Applying the condition
/r = v
n
, r = a we then nd
=
a
2
v
n
(t)
r
, r > a. (1.4.3)
Thus, the pressure
p =
o

t
=
o
a
2
r
dv
n
dt
(t)
M. S. Howe 16 1.4 Pulsating sphere
AAS
decays like 1/r with distance from the sphere, and exhibits the unphysical characteristic
of changing instantaneously everywhere when dv
n
/dt changes its value. For any time t
the volume ux q(t) of uid is the same across any closed surface enclosing the sphere.
Evaluating it for any sphere S of radius r > a we nd
q(t) =
_
S
dS = 4a
2
v
n
(t),
and we may also write
=
q(t)
4r
, r > a. (1.4.4)
1.4.2 Point source in incompressible uid
The incompressible motion generated by a volume point source of strength q(t)
concentrated at the origin is the solution of equation (1.3.9) with c
o
= and
q(x, t) = q(t)(x):

2
= q(t)(x), where (x) = (x
1
)(x
2
)(x
3
). (1.4.5)
The solution must be radially symmetric and given by
=
A
r
, for r > 0. (1.4.6)
To nd A equation (1.4.5) is integrated over the interior of a sphere of radius r = R > 0,
and the divergence theorem is applied on the left:
_
r<R

2
d
3
x =
_
S
dS, where S is
the surface of the sphere. Then
_
S
dS
_
A
R
2
_
(4R
2
) = q(t).
Hence A = q(t)/4 and = q(t)/4r, which agrees with the solution (1.4.4) for the
sphere with the same volume outow in the region r > a = radius of the sphere. This
indicates that when we are interested in modelling the eect of a pulsating sphere at large
M. S. Howe 17 1.4 Pulsating sphere
AAS
distances r a, it is permissible to replace the sphere by a point source (a monopole)
of the same strength q(t) = rate of change of the volume of the sphere. This conclusion is
valid for any pulsating body, not just a sphere. However, it is not necessarily a good model
(especially when we come to examine the production of sound by a pulsating body) in the
presence of a mean uid ow past the sphere.
The solution (1.4.6) for the point source is strictly valid only for r > 0, where it satises

2
= 0. What happens as r 0, where its value is actually undened? To answer this
question we write the solution in the form
= lim
0
q(t)
4(r
2
+
2
)
1
2
, > 0, in which case
2
= lim
0
3
2
q(t)
4(r
2
+
2
)
5
2
.
The last limit is just equal to q(t)(x). Indeed when is small 3
2
/4(r
2
+
2
)
5
2
is also
small except close to r = 0, where it attains a large maximum 3/4
3
. Therefore, for any
smoothly varying test function f(x) and any volume V enclosing the origin
lim
0
_
V
3
2
f(x)d
3
x
4(r
2
+
2
)
5
2
= f(0) lim
0
_

3
2
d
3
x
4(r
2
+
2
)
5
2
= f(0)
_

0
3
2
r
2
dr
(r
2
+
2
)
5
2
= f(0),
where the value of the last integral is independent of . This is the dening property of the
three-dimensional -function.
Thus the correct interpretation of the solution
=
1
4r
of
2
= (x) (1.4.7)
for a unit point source ( q = 1) is
1
4r
= lim
0
1
4(r
2
+
2
)
1
2
, r 0, (1.4.8)
where

2
_
1
4r
_
= lim
0

2
_
1
4(r
2
+
2
)
1
2
_
= lim
0
3
2
4(r
2
+
2
)
5
2
= (x). (1.4.9)
M. S. Howe 18 1.4 Pulsating sphere
AAS
1.4.3 Low frequency pulsations of a sphere in compressible uid
Let us next calculate the radially symmetric sound produced by the acoustically compact
pulsating sphere of 1.4.1. Setting r = [x[ and observing that radial symmetry implies that

2

1
r
2

r
_
r
2

r
_

1
r

2
r
2
(r) ,
it follows that (1.4.1) reduces to the one dimensional wave equation for r
1
c
2
o

2
t
2
(r)

2
r
2
(r) = 0, when r > a. (1.4.10)
The general solution r = (t r/c
o
) +(t +r/c
o
), for arbitrary functions and , yields
=

_
t
r
co
_
r
+

_
t +
r
co
_
r
, r > a. (1.4.11)
The terms on the right respectively represent spherically symmetric disturbances
propagating in the directions of increasing and decreasing values of r at the speed of
sound c
o
. Causality requires the incoming wave to be omitted, i.e. that = 0, because
(t + r/c
o
)/r necessarily represents sound arriving from r = and must be absent for
pulsations that started at some nite time in the past. This statement of the causality
principle is equivalent to imposing a radiation condition that sound waves must radiate
away from their source.
The outgoing wave function is determined from the boundary condition /r = v
n
(t)
at r = a, which gives
v
n
(t) =

r
_
_

_
t
r
co
_
r
_
_
r=a
=
1
a
2

_
t
a
c
o
_

1
ac
o

t
_
t
a
c
o
_
. (1.4.12)
But for a typical component of (t) e
it
of radian frequency ,
e
i(ta/co)
e
it
and
a
c
o

t
k
o
a 1,
when the sphere is compact (k
o
a 1). Therefore, the time delay a/c
o
on the right of
(1.4.12) can be neglected and the time-derivative term discarded, giving (t) a
2
v
n
(t),
M. S. Howe 19 1.4 Pulsating sphere
AAS
and

q(t r/c
o
)
4r
, r > a (1.4.13)
where q(t) = 4a
2
v
n
(t) is the rate of volume outow from the sphere. The acoustic potential
at time t at a distant point r is seen to be the same as for incompressible ow (equation
(1.4.4)) except that it is delayed by the time of travel r/c
o
of sound from the sphere.
This result is typical of compact, pulsating bodies; the sound can always be expressed in
form (1.4.13) terms of the pulsational volumetric ow rate q(t).
M. S. Howe 20 1.4 Pulsating sphere
AAS
1.5 Sound produced by an impulsive point source
The impulsive point source of strength q = (x)(t) is non-zero for an innitesimal
time at t = 0. The usual convention in acoustics, however, is to reverse the sign of the
source (so that q(x, t) = (x)(t)), and to consider the corresponding inhomogeneous
wave equation (1.3.9) in the form
_
1
c
2
o

2
t
2

2
_
= (x)(t). (1.5.1)
Because the source vanishes for t < 0 we are interested only in the causal solution, which is
non-zero only for t > 0.
The solution is radially symmetric and of outgoing wave form, so that we can put
=
(t r/c
o
)
r
for r = [x[ > 0. (1.5.2)
The functional form of can be determined by the method of 1.4.2, or more simply by
noting that

(t r/c
o
)
r
= lim
0
(t)
(r
2
+
2
)
1
2
when r 0, and therefore that temporal derivatives /t become negligible compared
to spatial derivatives /r. In other words the solution must resemble that for an
incompressible uid very close to the source. Hence we must have (t) = (t)/4 and the
causal solution of (1.5.1) becomes
(x, t) =
1
4r

_
t
r
c
o
_

1
4[x[

_
t
[x[
c
o
_
. (1.5.3)
The sound wave consists of a singular spherical pulse that is non-zero only on the surface of
the sphere r = c
o
t > 0 expanding at the speed of sound; it vanishes everywhere for t < 0.
M. S. Howe 21 1.5 Impulsive point source
AAS
1.6 Free space Greens function
The free space Greens function G(x, y, t, ) is the causal solution of the wave equation
generated by the impulsive point source (x y)(t ), located at the point x = y at
time t = . The formula for G is obtained from the solution (1.5.3) for a source at x = 0 at
t = 0 simply by replacing x by x y and t by t . In other words, if
_
1
c
2
o

2
t
2

2
_
G = (x y)(t ), where G = 0 for t < , (1.6.1)
then
G(x, y, t, ) =
1
4[x y[

_
t
[x y[
c
o
_
. (1.6.2)
This represents an impulsive, spherically symmetric wave expanding from the source at y
at the speed of sound. The wave amplitude decreases inversely with distance [x y[ from
the source point y.
1.6.1 The retarded potential
Greens function is the fundamental building block for solutions of the inhomogeneous
wave equation (1.3.7) of linear acoustics in an unbounded medium. Let us write this
equation in the form
_
1
c
2
o

2
t
2

2
_
p = T(x, t), (1.6.3)
where the generalized source T(x, t) is assumed to be generating waves that propagate
away from the source region, in accordance with the radiation condition.
This source distribution can be regarded as a distribution of impulsive point sources of
the type on the right of equation (1.6.1), because
T(x, t) =
__

T(y, )(x y)(t )d


3
yd.
The outgoing wave solution for each constituent source of strength
T(y, )(x y)(t )d
3
yd is T(y, )G(x, y, t, )d
3
yd,
M. S. Howe 22 1.6 Greens function
AAS
so that by adding these individual contributions we obtain
p(x, t) =
__

T(y, )G(x, y, t, )d
3
yd (1.6.4)
=
1
4
__

T(y, )
[x y[

_
t
[x y[
c
o
_
d
3
yd (1.6.5)
i.e. p(x, t) =
1
4
_

T
_
y, t
|xy|
co
_
[x y[
d
3
y. (1.6.6)
The integral formula (1.6.6) is called a retarded potential; it represents the pressure at
position x and time t as a linear superposition of contributions from sources at positions
y which radiated at the earlier times t [x y[/c
o
, [x y[/c
o
being the time of travel of
sound waves from y to x.
1.6.2 Greens function in one or two space dimensions: method of descent
The Greens function for plane waves that propagate in one dimension (in a uniform
duct, for example) say parallel to the x
1
-axis, is the causal solution G(x
1
, y
1
, t, ) of
_
1
c
2
o

2
t
2


2
x
2
1
_
G = (x
1
y
1
)(t ), where G = 0 for t < . (1.6.7)
This equation can be obtained formally by integrating the corresponding three dimensional
equation (1.6.1) over the plane of uniformity < x
2
, x
3
< . This is Hadamards
(1952) method of descent to a lower space dimension. Using the formula (1.6.2) for G in
three dimensions, we nd that (1.6.7) is satised by
G(x
1
, y
1
, t, ) =
__

1
4[x y[

_
t
[x y[
c
o
_
dy
2
dy
3
=
_

0

_
_
t
_
[x
1
y
1
[
2
+
2
c
o
_
_
d
2
_
[x
1
y
1
[
2
+
2
=
c
o
2
H
_
t
[x
1
y
1
[
c
o
_
, (1.6.8)
M. S. Howe 23 1.6 Greens function
AAS
where H(x) is the Heaviside unit step function (= 1, 0 according as x
>
<
0; see Appendix B).
In two dimensions sound waves propagate cylindrically as functions of (x
1
, x
2
) and t,
and Greens function can be found by descent, by integrating the three dimensional formula
over < x
3
< to obtain
G(x, y, t, ) =
H(t [x y[/c
o
)
2
_
(t )
2
[x y[
2
/c
2
o
, x = (x
1
, x
2
), y = (y
1
, y
2
), (1.6.9)
the causal solution of
_
1
c
2
o

2
t
2

2
_
G = (x
1
y
1
)(x
2
y
2
)(t ), G = 0 for t < .
In three dimensions G consists of a spherically spreading singular pulse that vanishes
everywhere except at the wavefront. The corresponding one dimensional Greens function
is nite and consists of two simple discontinuities propagating in both directions from the
source at the speed of sound, to the rear of which the amplitude is constant and equal to
c
o
/2. The behaviour of G in two dimensions exhibits certain intermediate characteristics:
the wavefront consists of a circular cylindrical singular pulse radiating outwards from the
source at the speed of sound, but followed by a slowly decaying tail extending back to the
source point where its amplitude decreases like 1/(t ). From the view point of an observer
in three dimensions, the two dimensional source (x
1
y
1
)(x
2
y
2
)(t ) is equivalent to
a uniform, innitely long line source (parallel to the x
3
axis). The tail can be attributed
to the arrival of sound from distant points on this line source, which persists for all time
after the passage of the wave front, which can be regarded as produced by components of
the line source in the immediate neighbourhood of its intersection with the x
1
x
2
plane.
M. S. Howe 24 1.6 Greens function
AAS
1.7 Initial value problem for the wave equation
Our rst application of the retarded potential integral is to the solution of Cauchys
problem, where it is required to calculate the sound at times t > 0 in terms of the state
of the acoustic medium at t = 0. This includes, for example, the problem of determining
the sound produced by the sudden rupture of a closed material envelope containing air at
high pressure (a bursting balloon). Alternatively, the state of a system of sound waves
generated by sources in the distant past might be specied and it is desired to determine
their subsequent propagation.
The nature of required initial data at t = 0, say, may be deduced directly from the
homogeneous form of the wave equation (1.6.3), whose solution is required for t > 0. To do
this the equation is multiplied by the Heaviside function H(t), making use of the identity
H(t)

2
p
t
2
=

2
t
2
_
pH(t)
_


t
_
(t)p
_
(t)
p
t
.
Then equation (1.6.3), with T(x, t) 0, becomes
_
1
c
2
o

2
t
2

2
_
(pH) =
1
c
2
o

t
_
(t)p
_
+
1
c
2
o
(t)
p
t
. (1.7.1)
This equation is formally valid for all time < t < , with pH(t) 0 for t < 0.
The outgoing wave solution calculated using the retarded potential (1.6.6) determines
p(x, t)H(t) p(x, t) for t > 0 in terms of the initial pressure and velocity distributions:
p = f(x),
p
t
=
o
c
2
o
div v = g(x) at t = 0, (1.7.2)
where f(x), g(x) represent the prescribed initial values (p/t =
o
c
2
o
div v is just the
linearised continuity equation).
To evaluate the retarded potential integral for t > 0 we start from (1.6.5):
p(x, t) =
1
4c
2
o
__
_

_
f(y)()
_
+ g(y)()
_

_
t
[x y[
c
o
_
d
3
yd
[x y[
. (1.7.3)
M. S. Howe 25 1.7 Initial value problem
AAS
Introduce polar coordinates (r = [x y[, , ), with origin at the observer position x, so
that d
3
y = r
2
dr d, where d = sin dd is the solid angle element. Then the integral
involving f(y) becomes
1
4c
2
o
__

_
f(r, )()
_

_
t
r
c
o
_
rdrd =
1
4c
2
o

t
_
f(r, )
_
t
r
c
o
_
rdrd
=

t
_
tf(c
o
t, )
d
4
=

t
_
t

f(c
o
t)
_
,
where

f(c
o
t) =
_
f(c
o
t, )
d
4
is the mean value of f(y) f(r, ) on the surface of a sphere of radius r = c
o
t centred on
the observer position x (P in Figure 1.7.1).
Combining this with a similar calculation for the term in g(y), leads to Poissons (1819)
solution
p(x, t) =

t
_
t

f(c
o
t)
_
+ t g(c
o
t). (1.7.4)
Figure 1.7.1
If the initial values f(x), g(x) of the pressure and its time derivative are non-zero only
within a nite region bounded by a closed surface S (Figure 1.7.1), the mean values on the
right of (1.7.4) vanish except when the expanding spherical surface r = c
o
t cuts across S.
The pressure perturbation at P outside S is therefore non-zero only for r
1
< c
o
t < r
2
, where
M. S. Howe 26 1.7 Initial value problem
AAS
r
1
, r
2
are the respective radii of the smallest and largest spheres centred on P that just
touch S. The acoustic pressure radiating from S is therefore non-zero only within a shell-like
region of space; at time t the outer surface of this shell is the envelope of the family of
spheres of radius c
o
t whose centres lie on S. Equation (1.7.4) also gives the solution for an
observer within S. For suciently small time the inner envelope of the expanding family
of spheres forms a wavefront collapsing into the interior of S this eventually crosses itself,
emerges from the other side of S and expands to form the inner boundary of the radiating
shell.
Note that the right hand side of (1.7.1) can be cast in the form of the general source of
the linear acoustic equation (1.3.7)

o
q
t
div F
where q(x, t), F(x, t) are impulsive volume source and body force distributions
q(x, t) =
1

o
c
2
o
p(x, 0)(t), F(x, t) =
o
v(x, 0)(t), (1.7.5)
and where the formula for F is obtained from the linearised continuity equation
1
c
2
o
p
t
+ div(
o
v) = 0.
Also, when the initial disturbance is non-zero only within a nite region bounded by the
surface S, the linearised momentum equation (1.3.1) implies that at any point x outside
S
_

0
p(x, t)dt =
o
[v(x, t)]

t=0
0, because v = 0 before and after the passage of the
sound wave. Therefore
_

0
p(x, t)dt = 0 for all x, (1.7.6)
because the integral must vanish at innitely large distances from S. Thus, the acoustic
pressure variation at x during the time interval r
1
/c
o
< t < r
2
/c
o
occupied by the wave
must always involve equal and opposite net compressions and rarefactions. This conclusion
is true for waves propagating in three and two dimensions, but not for one-dimensional
propagation. It is illustrated by the following example.
M. S. Howe 27 1.7 Initial value problem
AAS
1.7.1 Sound radiated from a spherical region of initial high pressure
Let the initial uniform high pressure f(x) = p
0
> 0 be conned to stationary uid in the
interior of a sphere S of radius a, so that g(x) = 0. Evidently, for the exterior point P in
Figure 1.7.2 at distance r from the centre O of the sphere, r
1
= r a, r
2
= r + a, and the
pressure at P is non-zero only for r a < c
o
t < r + a. During this time

f(c
o
t) =
p
0
4
,
where =
_
2
0
d
_
(t)
0
sin d = 21 cos (t) is the solid angle subtended at P by the
spherical cap AB formed by the intersection of S and the sphere of radius c
o
t centred on P.
Hence, using the cosine formula (a
2
= r
2
+ c
2
o
t
2
2rc
o
t cos ) for the triangle OAP, we nd

f(c
o
t) =
p
0
4rc
o
t
_
a
2
(c
o
t r)
2
__
H(c
o
t r + a) H(c
o
t r a)
_
,
and therefore
p(r, t) =

t
_
t

f(c
o
t)
_
=
p
0
2r
(c
o
t r)
_
H(c
o
t r + a) H(c
o
t r a)
_
, r > a, (1.7.7)
which represents an outgoing spherical wave conned to the shell c
o
t a < r < c
o
t +a that
decreases in amplitude like 1/r.
Figure 1.7.2
A similar calculation performed when P lies within the initial high pressure region
(r < a) gives
p(r, t) = p
0
_
1
(c
o
t + r)
2r
H(c
o
t + r a) +
(c
o
t r)
2r
H(c
o
t r a)
_
, r < a. (1.7.8)
M. S. Howe 28 1.7 Initial value problem
AAS
The pressure at r inside S remains equal to p
0
until the arrival of the rarefaction wave (the
second term in the large brackets) at time t = (a r)/c
o
, which is subsequently reected
from the centre r = 0 as a compression wave, after the passage of which the pressure is
reduced to zero.
The sequence of events is illustrated in Figure 1.7.3, where the nondimensional pressure
p/p
0
is plotted against r/a for a set of increasing values of c
o
t/a. The perturbation
pressure is initially uniform within S at c
o
t/a = 0 and vanishes elsewhere. Compression
and rarefaction waves radiate respectively into the exterior and interior of S as c
o
t/a
increases towards 1 . When c
o
t/a 1 the negative rarefaction peak becomes very large, and
ultimately p/p
0
(c
o
t/a1) at r = 0. Long before this happens in a real uid, however,
the large rarefaction is suppressed by nonlinear actions that increase the propagation speed
of the higher pressure sections of the inward propagating wave, thereby inhibiting the
formation of a deep negative pressure. After reection at r = 0 at c
o
t/a = 1 the whole
disturbance becomes outgoing of N-wave prole occupying a shell of thickness 2a, the
compression wavefront being at r = c
o
t + a and the pressure vanishing within the shell
at r = c
o
t. Spherical spreading causes the wave amplitude to decrease like 1/r. It is also
follows from formulae (1.7.7), (1.7.8) for the pressure that
_

0
p(r, t)dt = 0, even though the
initial pressure distribution p
0
> 0 within r < a.
M. S. Howe 29 1.7 Initial value problem
AAS
Figure 1.7.3
1.7.2 Initial value problem in one dimension
The initial value problem for one dimensional propagation of sound parallel to the x
axis in unbounded uid is governed by
_
1
c
2
o

2
t
2


2
x
2
_
(pH) =
1
c
2
o

t
_
(t)f(x)
_
+
1
c
2
o
(t)g(x), < x < , (1.7.9)
where f(x), g(x) are the respective initial values of p and p/t. This is solved by means
of the Greens function (1.6.8), using the results (see equation (B.1.6) of the Appendix):
__
1
c
2
o

_
()f(y)
_
G(x, y, t, )dyd =
1
2c
o
_
f(y)
_
t
[x y[
c
o
_
dy
=
1
2
_
f(x c
o
t) + f(x + c
o
t)
_
__
1
c
2
o
()g(y)G(x, y, t, )dyd =
1
2c
o
_
g(y)H
_
t
[x y[
c
o
_
dy
=
1
2c
o
_
x+cot
xcot
g(y) dy.
The combination of these results yields DAlemberts solution
p(x, t) =
1
2
_
f(x c
o
t) + f(x c
o
t)
_
+
1
2c
o
_
x+cot
xcot
g(y) dy. (1.7.10)
M. S. Howe 30 1.7 Initial value problem
AAS
A simple application of this formula is illustrated in Figure 1.7.4, for the case where the
uid is initially at rest (g(x) 0) with p = f(x) ,= 0 only within the interval a < x < a,
where it has the triangular waveform
f(x) = p
max
_
1
[x[
a
_
H
_
1
[x[
a
_
with peak pressure p
max
at x = 0. This pressure distribution splits symmetrically (see
gure), forming equal waves propagating without change of form to x = , and given for
t > 0 by
p(x, t)
p
max
=
1
2
__
1
[x c
o
t[
a
_
H
_
1
[x c
o
t[
a
_
+
_
1
[x + c
o
t[
a
_
H
_
1
[x + c
o
t[
a
__
.
Figure 1.7.4
M. S. Howe 31 1.7 Initial value problem
AAS
1.8 Monopoles, dipoles and quadrupoles
A volume point source q(t)(x) of the type considered in 1.4 as a model for a pulsating
sphere is also called a point monopole. For a compressible medium the corresponding
velocity potential it produces is the outgoing solution of
_
1
c
2
o

2
t
2

2
_
= q(t)(x). (1.8.1)
The solution can be written down by analogy with the solution (1.6.6) of equation
(1.6.3) for the acoustic pressure. Replace p by in (1.6.6) and set T(y, ) = q()(y).
Then
(x, t) =
q
_
t
|x|
co
_
4[x[

q
_
t
r
co
_
4r
. (1.8.2)
This coincides with the corresponding solution (1.4.3) for a compact pulsating sphere.
Changes in the motion of the sphere (i.e. in the value of the volume outow rate q(t)) are
communicated to a uid element at distance r after a time delay r/c
o
required for sound to
travel outward from the source.
1.8.1 The point dipole
Let f = f(t) be a time dependent vector, then a source on the right of the acoustic
pressure equation (1.6.3) of the form
T(x, t) = div
_
f(t)(x)
_


x
j
_
f
j
(t)(x)
_
(1.8.3)
is called a point dipole (located at the origin). The repeated subscript j in this equation,
implies a summation over j = 1, 2, 3 (see Appendix A). Equation (1.3.7) shows that
the point dipole is equivalent to a force distribution F(x, t) = f(t)(x) per unit volume
applied to the uid at the origin.
The sound produced by the dipole can be calculated from (1.6.6), but it is easier to use
(1.6.5):
p(x, t) =
1
4
__

y
j
_
f
j
()(y)
_

_
t
|xy|
co
_
[x y[
d
3
yd.
M. S. Howe 32 1.8 Monopoles, dipoles . . .
AAS
Integrate by parts with respect to each y
j
(recalling that (y) = 0 at y
j
= ), and note
that

y
j

_
t
|xy|
co
_
[x y[
=

x
j

_
t
|xy|
co
_
[x y[
.
Then
p(x, t) =
1
4
__

f
j
()(y)

x
j
_
_

_
t
|xy|
co
_
[x y[
_
_
d
3
yd
=
1
4

x
j
__

f
j
()(y)
_
_

_
t
|xy|
co
_
[x y[
_
_
d
3
yd
i.e.
p(x, t) =

x
j
_
_
f
j
_
t
|x|
co
_
4[x[
_
_
. (1.8.4)
The same procedure shows that for a distributed dipole source of the type T(x, t) =
div f(x, t) on the right of equation (1.6.3), the acoustic pressure is
p(x, t) =
1
4

x
j
_

f
j
_
y, t
|xy|
co
_
[x y[
d
3
y. (1.8.5)
A point dipole at the origin orientated in the direction of a unit vector n is entirely
equivalent to two point monopoles of equal but opposite strengths placed a short distance
apart (much smaller than the acoustic wavelength) on opposite sides of the origin on a line
through the origin parallel to n. For example, if n is parallel to the x-axis, and the sources
are distance apart, the two monopoles would be
q(t)
_
x

2
_
(y)(z) q(t)
_
x +

2
_
(y)(z) q(t)

(x)(y)(z)

x
_
q(t)(x)
_
.
This is a uid volume dipole. The relation p =
o
/t implies that the equivalent dipole
source in the pressure equation (1.3.7) or (1.6.3) is

x
_

q
t
(t)(x)
_
.
M. S. Howe 33 1.8 Monopoles, dipoles . . .
AAS
1.8.2 Quadrupoles
A source distribution involving two space derivatives is equivalent to a combination of
four monopole sources whose net volume source strength is zero, and is called a quadrupole.
A general quadrupole is a source of the form
T(x, t) =

2
T
ij
x
i
x
j
(x, t) (1.8.6)
in equation (1.6.3). The argument above leading to expression (1.8.5) can be applied twice
to show that the corresponding acoustic pressure is given by
p(x, t) =
1
4

2
x
i
x
j
_

T
ij
(y, t [x y[/c
o
)
[x y[
d
3
y. (1.8.7)
1.8.3 Solution of the general linear acoustic equation
The sources on the right of the general linear acoustic equation (1.3.7)
_
1
c
2
o

2
t
2

2
_
p =
o
q
t
div F
are respectively of monopole and dipole type. The solution with outgoing wave behaviour
is therefore
p(x, t) =

o
4

t
_

q(y, t [x y[/c
o
)
[x y[
d
3
y
1
4

x
j
_

F
j
(y, t [x y[/c
o
)
[x y[
d
3
y.
(1.8.8)
1.8.4 Vibrating sphere
Let a rigid sphere of radius a execute small amplitude oscillations at speed U(t) in
the x
1
-direction (Figure 1.8.1a) at suciently low frequencies that it may be assumed to
be acoustically compact. Take the coordinate origin at the mean position of the centre.
The sphere pushes uid away from its advancing front hemisphere, and the retreating
rear hemisphere draws in uid from its wake. The sphere therefore resembles a dipole
source, and it is shown in 2.7 that the motion induced in an ideal uid is equivalent to
M. S. Howe 34 1.8 Monopoles, dipoles . . .
AAS
that produced by a point volume dipole of strength 2a
3
U(t) at the position of its centre
directed along the x
1
-axis. The velocity potential is the solution of
_
1
c
2
o

2
t
2

2
_
=

x
1
_
2a
3
U(t)(x)
_
. (1.8.9)
By analogy with (1.8.3) and (1.8.4), we have
(x, t) =

x
1
_
_
2a
3
U
_
t
|x|
co
_
4[x[
_
_
. (1.8.10)
Now

x
j
[x[ =
x
j
[x[
. (1.8.11)
Applying this formula for j = 1, we nd (putting r = [x[ and x
1
= r cos )
=
a
3
cos
2r
2
U
_
t
r
c
o
_

a
3
cos
2c
o
r
U
t
_
t
r
c
o
_
near eld far eld
The near eld term is dominant at suciently small distances r from the origin that
1
r

1
c
o
1
U
U
t

f
c
o
where f is the characteristic frequency of the oscillations of the sphere. But, sound of
frequency f travels a distance
c
o
/f = one acoustic wavelength
in one period of oscillation 1/f. Hence the near eld term is dominant when
r .
The motion becomes incompressible when c
o
. In this limit the solution reduces
entirely to the near eld term, which is also called the hydrodynamic near eld; it decreases
in amplitude like 1/r
2
as r .
M. S. Howe 35 1.8 Monopoles, dipoles . . .
AAS
The far eld is the acoustic region that only exists when the uid is compressible.
It consists of propagating sound waves, carrying energy away from the sphere, and takes
over from the near eld when r . There is an intermediate zone where r in which
the solution is in a state of transition from the near to the far eld. This accords with the
assumption that the sphere is compact and can be replaced by the point dipole: the motion
close to the sphere is essentially the same as if the uid is incompressible i.e. the diameter
of the sphere is much smaller than the acoustic wavelength (a ).
Figure 1.8.1
The intensity of the sound generated by the sphere in the far eld is proportional to
2
:

a
6
4c
2
o
r
2
_
U
t
_
2
t
r
co
cos
2
, r .
The dependence on determines the directivity of the sound. For the dipole it has the
gure of eight pattern illustrated in Figure 1.8.1b, with peaks in directions parallel to the
dipole axis ( = 0, ); there are radiation nulls at =

2
(the curve should be imagined to
be rotated about the x
1
-axis).
M. S. Howe 36 1.8 Monopoles, dipoles . . .
AAS
1.9 Acoustic energy equation
The acoustic energy equation in a stationary ideal uid is obtained by taking the scalar
product of the velocity v and linearised momentum equation (1.3.1). The result is written

t
_
1
2

o
v
2
_
+ div
_
p v
_
p div v = v F.
The term p div v on the left is rendered in a more useful form by substitution for div v from
the linearised equation of continuity (1.3.3)
div v = q
1

t
= q
1

o
c
2
o
p
t
.
The required energy equation is then obtained in the form

t
_
1
2

o
v
2
+
1
2
p
2

o
c
2
o
_
+ div
_
p v
_
= pq +v F. (1.9.1)
The term
c =
1
2

o
v
2
+
1
2
p
2

o
c
2
o
, (1.9.2)
is the acoustic energy per unit volume, the rst part of which is obviously the kinetic
energy density. The second term is the compressional or potential energy component,
calculated as follows. Let V be the volume occupied by unit mass of the uid, then V = 1,
and
_
V
Vo
p

dV is the compressional energy per unit mass, where V


o
= 1/
o
is the volume
occupied by unit mass in the absence of the sound, and p

is the perturbation pressure


produced by the volumetric change V V
o
. Then the compressional energy per unit volume
is (to second order, and using the adiabatic formula dp

= c
2
o
d)

o
_
V
Vo
p

dV =
o
_

o
p

d
_
1

_
=
1

o
c
2
o
_
p
0
p

dp

=
1
2
p
2

o
c
2
o
.
The divergence term in (1.9.1) governs the rate at which acoustic energy propagates
out of unit volume of uid. The terms on the right are respectively the rates of production
of acoustic energy by the volume and force distribution sources. The energy balance is
perhaps more clearly exhibited by integration of the energy equation over the uid region V
M. S. Howe 37 1.9 Acoustic energy
AAS
bounded by a large closed surface S (Figure 1.9.1) that contains all of the acoustic sources.
Then

t
_
V
c d
3
x +
_
S
pv dS =
_
V
(pq +v F) d
3
x, (1.9.3)
which equates the sum of the rate of accumulation of energy within V and the energy ux
out through S to the rate of working of the acoustic sources.
1.9.1 Calculation of the energy ux
At large distances r from a source region we generally have
p(x, t)

_
, , t
r
co
_
r
, r , (1.9.4)
where the function depends on the nature of the source distribution, and and are
polar angles determining the directivity of the sound. From the radial component of the
linearized momentum equation
v
r
t
=
1

o
p
r

1
r
2

_
, , t
r
c
o
_
+
1
c
o
r

t
_
, , t
r
c
o
_
. (1.9.5)
The rst term in the second line can be neglected when r , and therefore
v
r

1
c
o
r

_
, , t
r
c
o
_

o
c
o
. (1.9.6)
By considering the and components of the momentum equation we can show that the
corresponding velocity components v

, v

, say, decrease faster than 1/r as r . We


therefore conclude from this and (1.9.6) that the acoustic particle velocity is normal to
the acoustic wavefronts (the spherical surfaces r = c
o
t). In other words: sound consists of
longitudinal waves in which the uid particles oscillate backwards and forwards along the
local direction of propagation of the sound.
The acoustic power radiated by the source distribution is given by the surface
integral of equation (1.9.3), which we can take in the form
=
_
S
pv
r
dS =
_
S
p
2

o
c
o
dS, (1.9.7)
M. S. Howe 38 1.9 Acoustic energy
AAS
where the surface of integration S is that of a large sphere of radius r centered on the
source region. Because the surface area = 4r
2
we only need to know the pressure and
velocity correct to order 1/r on S in order to evaluate the integral. Smaller contributions
(such as that determined by the rst term in the second line of (1.9.5)) decrease too fast as
r increases to supply a nite contribution to the integral as r .
Figure 1.9.1
In acoustic problems we are therefore usually satised if we can calculate the pressure
and velocity in the acoustic far eld correct to order 1/r; this will always permit the
evaluation of the radiated sound power. The formula v
r
= p/
o
c
o
is applicable at large
distances from the sources, where the wavefronts can be regarded as locally plane, but it is
true identically for plane sound waves. In the latter case, and for spherical waves on the
surface of the large sphere of Figure 1.9.1, the quantity
I = pv
r
=
p
2

o
c
o
(1.9.8)
is called the acoustic intensity. It is the rate of transmission of acoustic energy per unit
area of wavefront. For a plane wave c = p
2
/
o
c
2
o
, so that I = c
o
c, i.e. the plane wave energy
ux is equal to the energy density multiplied by the speed of sound.
M. S. Howe 39 1.9 Acoustic energy
AAS
1.9.2 Example
For a transient acoustic source all of the wave energy radiates out through the distant
surface S in a nite time. Integration of the energy equation (1.9.3) over all times therefore
yields
_

_
S
pv dSdt =
_

_
V
(pq +v F) d
3
xdt. (1.9.9)
Let us verify this formula for the initial value problem of 1.7.1 of sound produced by the
sudden release of high pressure air from within the spherical region [x[ = r < a.
Outside the source region the pressure is given by equation (1.7.7), and the power
radiated through S is
=
_
S
p
2

o
c
o
dS = 4r
2
_
p
2
o
4r
2

o
c
o
(c
o
t r)
2
_
_
H(c
o
t r + a) H(c
o
t r a)
_
.
The total radiated acoustic energy E, say, determined by the left hand side of (1.9.9) is
therefore
E =
p
2
0

o
c
o
_
(r+a)/co
(ra)/co
(c
o
t r)
2
dt =
2a
3
3
p
2
0

o
c
2
o
. (1.9.10)
On the right of (1.9.9) we have, from equations (1.7.5), F = 0 and
q =
p
0

o
c
2
o
(t)H(a r).
This must be multiplied by the pressure given by equation (1.7.8) for r < a, following which
E is calculated by evaluation of E =
_

_
V
pq d
3
xdt. But in doing this it must be recalled
that the radiated pressure is the causal solution of equation (1.7.1), and what is actually
calculated is not p but pH(t). Thus,
E =
p
2
0

o
c
2
o
_

_
|x|<a
H(t) (t)
_
1
(c
o
t + r)
2r
H(c
o
t + r a) +
(c
o
t r)
2r
H(c
o
t r a)
_
d
3
xdt

p
2
0

o
c
2
o
_

_
|x|<a
H(t) (t)d
3
xdt =
4a
3
3
p
2
0

o
c
2
o
_

H(t) (t) dt
=
4a
3
3
p
2
0

o
c
2
o
_
1
2
H
2
(t)
_

=
2a
3
3
p
2
0

o
c
2
o
,
M. S. Howe 40 1.9 Acoustic energy
AAS
which agrees with (1.9.10).
The integration with respect to time has made use of the result dH(t)/dt = (t). Readers
uncomfortable with this formal step should observe that the solution of the initial value
problem (1.7.1) can be carried through with negligible change of details, by replacing H(t)
and (t) by their corresponding sequences H

(t),

(t) (see Appendix B)


H

(t) =
1
2
+
1

tan
1
_
t

_
,

(t) =

(
2
+ t
2
)
, > 0,
in which case
_

(t)

(t) dt = [
1
2
H
2

(t)]

=
1
2
.
M. S. Howe 41 1.9 Acoustic energy
AAS
1.10 Calculation of the acoustic far eld
We now discuss the approximations necessary to evaluate the sound in the far eld from
the retarded potential representation:
p(x, t) =
1
4
_

T
_
y, t
|xy|
co
_
[x y[
d
3
y. (1.10.1)
We assume that T(x, t) ,= 0 only within a nite source region (Figure 1.10.1), and take the
coordinate origin O within the region.
Figure 1.10.1
When [x[ and y lies within the source region (so that [x[ [y[)
[x y[
_
[x[
2
2x y +[y[
2
_1
2
= [x[
_
1
2x y
[x[
2
+
[y[
2
[x[
2
_1
2
[x[
_
1
x y
[x[
2
+ O
_
[y[
2
[x[
2
__
i.e. [x y[ [x[
x y
[x[
when
[y[
[x[
1. (1.10.2)
Similarly,
1
[x y[

1
_
[x[
xy
|x|
_

1
[x[
_
1 +
x y
[x[
2
_
. .
1
[x y[

1
[x[
+
x y
[x[
3
when
[y[
[x[
1. (1.10.3)
The approximation (1.10.3) shows that, in order to obtain the far eld approximation
of the solution (1.10.1) that behaves like 1/r = 1/[x[ as [x[ , it is sucient to replace
M. S. Howe 42 1.10 Acoustic far eld
AAS
[x y[ in the denominator of the integrand by [x[. However, in the argument of the source
strength T it is important to retain possible phase dierences between the sound waves
generated by components of the source distribution at dierent locations y; we therefore
replace [x y[ in the retarded time by the right hand side of (1.10.2). Hence,
p(x, t)
1
4[x[
_

T
_
y, t
[x[
c
o
+
x.y
c
o
[x[
_
d
3
y, [x[ . (1.10.4)
This is called the Fraunhofer approximation.
The source region may extend over many characteristic acoustic wavelengths of the
sound. By retaining the contribution x y/c
o
[x[ to the retarded time we ensure that any
interference between waves generated at dierent positions within the source region is
correctly described by the far eld approximation. In Figure 1.10.1 the acoustic travel
time from a source point y to the far eld point x is equal to that from the point labelled
A to x when [x[ . The travel time over the distance OA is just x y/c
o
[x[, so that
[x[/c
o
x y/c
o
[x[ gives the correct value of the retarded time when [x[ .
1.10.1 Dipole source distributions
By applying the far eld formula (1.10.4) to a dipole source T(x, t) = div f(x, t) we
obtain (from (1.8.5))
p(x, t)
1
4

x
j
_
1
[x[
_

f
j
_
y, t
[x[
c
o
+
x y
c
o
[x[
_
d
3
y
_

1
4[x[

x
j
_

f
j
_
y, t
[x[
c
o
+
x y
c
o
[x[
_
d
3
y, [x[ , (1.10.5)
because the dierential operator /x
j
need not be applied to 1/[x[ as this would give a
contribution decreasing like 1/r
2
at large distances from the dipole.
However, it is useful to make a further transformation that replaces /x
j
by the time
derivative /t, which is usually more easily estimated in applications. To do this we
M. S. Howe 43 1.10 Acoustic far eld
AAS
observe that
f
j
x
j
_
y, t
[x[
c
o
+
x y
c
o
[x[
_
=
f
j
t
_
y, t
[x[
c
o
+
x y
c
o
[x[
_


x
j
_
t
[x[
c
o
+
x y
c
o
[x[
_
=
f
j
t
_
y, t
[x[
c
o
+
x y
c
o
[x[
_

x
j
c
o
[x[
+
y
j
c
o
[x[

(x y)x
j
c
o
[x[
3
_

x
j
c
o
[x[
f
j
t
_
y, t
[x[
c
o
+
x y
c
o
[x[
_
as [x[ .
Hence, the far eld of a distribution of dipoles T(x, t) = div f(x, t) is given by
p(x, t) =
x
j
4c
o
[x[
2

t
_

f
j
_
y, t
[x[
c
o
+
x y
c
o
[x[
_
d
3
y. (1.10.6)
Note that
x
j
[x[
2
=
x
j
[x[
1
[x[
,
where x
j
/[x[ is the jth component of the unit vector x/[x[. Thus, the additional factor of
x
j
/[x[ in (1.10.6) does not change the rate of decay of the sound with distance from the
source (which is still like 1/r) but it does have an inuence on the acoustic directivity.
A comparison of (1.10.5) and (1.10.6) leads to the following rule for interchanging space
and time derivatives in the acoustic far eld

x
j

1
c
o
x
j
[x[

t
. (1.10.7)
1.10.2 Quadrupole source distributions
For the quadrupole (1.10.6)
T(x, t) =

2
T
ij
x
i
x
j
(x, t)
and
p(x, t) =
1
4

2
x
i
x
j
_

T
ij
(y, t [x y[/c
o
)
[x y[
d
3
y.
M. S. Howe 44 1.10 Acoustic far eld
AAS
By applying (1.10.4) and the rule (1.10.7), we nd that the acoustic far eld is given by
p(x, t)
x
i
x
j
4c
2
o
[x[
3

2
t
2
_

T
ij
_
y, t
[x[
c
o
+
x y
c
o
[x[
_
d
3
y, [x[ . (1.10.8)
1.10.3 Example
For the (1,2) point quadrupole
T(x, t) =

2
x
1
x
2
(T(t)(x))
equation (1.10.8) shows that in the acoustic far eld
p(x, t)
x
1
x
2
4c
2
o
[x[
3

2
T
t
2
_
t
[x[
c
o
_
, [x[ .
If we use spherical polar coordinates, such that
x
1
= r cos , x
2
= r sin cos , x
3
= r sin sin ,
we can write the pressure in the form
p(x, t)
sin 2 cos
8c
2
o
[x[

2
T
t
2
_
t
[x[
c
o
_
, [x[ .
The directivity of the sound ( p
2
) is therefore represented by sin
2
2 cos
2
. Its shape is
plotted in Figure 1.10.2 for radiation in the x
1
x
2
-plane ( = 0, ). The four lobe clover
leaf pattern is characteristic of a quadrupole T
ij
for which i ,= j.
M. S. Howe 45 1.10 Acoustic far eld
AAS
Figure 1.10.2
1.10.4 Far eld of a compact source distribution
Phase variations x y/c
o
[x[ of sound arriving from dierent parts of a generalised source
T(x, t) are small if the source region is acoustically compact. Therefore, the far eld
approximation ([x[ ) (1.10.4) becomes
p(x, t)
1
4[x[
_

T
_
y, t
[x[
c
o
+
x.y
c
o
[x[
_
d
3
y
1
4[x[
_

T
_
y, t
[x[
c
o
_
d
3
y.
This gives the principal component of the radiating sound unless it should happen that the
overall source strength is null, i.e., unless
_

T (y, t) d
3
y = 0. The amplitude of sound
waves in the far eld is then crucially dependent on the existence of small phase mismatches
between dierent parts of the source, and must be determined by expanding the Fraunhofer
approximation in powers of x y/c
o
[x[:
M. S. Howe 46 1.10 Acoustic far eld
AAS
p(x, t)
x
i
4c
o
[x[
2

t
_

y
i
T
_
y, t
[x[
c
o
_
d
3
y +
x
i
x
j
8c
o
[x[
3

2
t
2
_

y
i
y
j
T
_
y, t
[x[
c
o
_
d
3
y
+
= dipole + quadrupole + (1.10.9)
Each term in this multipole expansion is nominally of order /c
o
= k
o
1 relative to
its predecessor, where /t is the characteristic source frequency and the diameter
of the source region. Therefore, the expansion is halted at the rst non-zero term (see
Question 9 of Problems 1).
M. S. Howe 47 1.10 Acoustic far eld
AAS
Problems 1
1. Derive the relations (1.3.12) for plane sound waves.
2. Use the trial solution = (t |x|/c
o
) to solve the problem
_
1
c
2
o

2
t
2


2
x
2
_
= (x)(t), < x < ,
where = 0 for t < 0.
3. When the body force F = g in equation (1.3.1), where g is the acceleration due to gravity,
show that in the adiabatic approximation the linearised acoustic wave equation becomes
1

o
c
2
o

2
p

t
2


x
j
_
1

o
p

x
j
_
=
q
t
div
_
p

o
c
2
o
_
,
where p

= p p
o
, and the mean pressure, density and sound speed p
o
,
o
, c
o
vary with depth
in the atmosphere. Deduce that the gravitational term on the right hand side can be neglected
provided g/c
o
, where is a typical acoustic frequency.
4. Consider the initial value problem (1.7.1), (1.7.2) in which the pressure p = p
0
= constant and
p/t = 0 at time t = 0 in 0 < x
1
< h. Show that Poissons solution (1.7.4) predicts plane wave
propagation parallel to the x
1
direction, and that the pressure wave radiating through a typical
exterior point P distance r
1
from the source region has amplitude
1
2
p
0
, arrives at time t = r
1
/c
o
and occupies a plane shell of thickness h.
M. S. Howe 48 Problems 1
AAS
5. Derive DAlemberts solution (1.7.10) for the initial value problem in one space dimension from
Poissons formula (1.7.4).
6. At time t = 0 a plane acoustic wave p = p(x c
o
t) propagating in unbounded uid satises
p = p(x), p/t = c
o
p

(x) (the prime denoting dierentiation with respect to the argument).


Verify that Poissons solution (1.7.4) predicts that p = p(x c
o
t) when t > 0.
7. Calculate the acoustic power (1.9.7) radiated by an acoustically compact sphere of radius a exe-
cuting small amplitude translational oscillations of frequency and velocity U(t) = U
o
cos(t),
where U
o
= constant.
8. As for Problem 7 when the sphere executes small amplitude radial oscillations at normal velocity
v
n
= U
o
cos(t), U
o
= constant.
9. Consider the outgoing wave solution of
_
1
c
2
o

2
t
2

2
_
=
sgn(x)H(a |x|)U(t)
a
, a > 0,
where U(t) is known as a function of time. If the source region |x| < a is acoustically compact,
show that
(x, t)
a
3
cos
8c
o
|x|
U
t
_
t
|x|
c
o
_
, |x| ,
where cos = x/|x|.
10. A volume point source of strength q
o
(t) translates at constant, subsonic velocity U. The velocity
potential (x, t) of the radiated sound is determined by the solution of
_
1
c
2
o

2
t
2

2
_
= q
o
(t)(x Ut).
Show that
(x, t) =
q
o
(t R/c
o
)
4R(1 M cos )
, M =
U
c
o
,
where R is the distance of the reception point x fromthe source position at the time of emission of
the sound received at x at time t, and is the angle between U and the direction of propagation
of this sound.
M. S. Howe 49 Problems 1