THE
DYNAMICAL THEOKY
OF
SOUND
BY
HORACE LAMB, SC.D., LL.D, F.R.S.
PROFESSOR OF MATHEMATICS IN THE VICTORIA UNIVERSITY OF MANCHESTER ;
FORMERLY FELLOW OF TRINITY COLLEGE, CAMBRIDGE
LONDON
EDWARD ARNOLD
1910
[All Eights reserved]
PREFACE
A COMPLETE survey of the theory of sound would lead into
^^ many fields, physical, physiological, psychological, aesthetic.
The present treatise has a more modest aim, in that it is
devoted mainly to the dynamical aspect of the subject. It is
accordingly to a great extent mathematical, but I have tried to
restrict myself to methods and processes which shall be as
simple and direct as is possible, regard being had to the nature
of the questions treated. I hope therefore that the book may
fairly be described as elementary, and that it may serve as a
stepping stone to the study of the writings of Helmholtz and
Lord Rayleigh, to which I am myself indebted for almost all
that I know of the subject.
The limitation of methods has involved some sacrifices.
Various topics of interest have had to be omitted, whilst
others are treated only in outline, but I trust that enough
remains to afford a connected view of the subject in at all events
its more important branches. In the latter part of the book
a number of questions arise which it is hardly possible
to deal with according to the stricter canons even of mathe
matical physics. Some recourse to intuitional assumptions is
inevitable, and if in order to bring such questions within the
scope of this treatise I have occasionally carried this license
a little further than is customary, I would plead that this is
not altogether a defect, since attention is thereby concentrated
on those features which are most important from the physical
point of view.
257816
iv PBEFACE
Although a few historical notes are inserted here and there,
there is no attempt at systematic citation of authorities. The
reader who wishes to carry the matter further will naturally
turn in the first instance to Lord Rayleigh's treatise, where
full references, together with valuable critical discussions, will
be found. I may perhaps be allowed to refer also to the
article entitled " Schwingungen elastischer Systeme, insbeson
dere Akustik," in the fourth volume of the Encyclopddie der
mathematischen Wissenschaften (Leipzig, 1906).
I have regarded the detailed description of experimental
methods as lying outside my province. I trust, however, that
no one will approach the study of the subject as here treated
without some firsthand acquaintance with the leading pheno
mena. Fortunately, a good deal can be accomplished in this
way with very simple and easily accessible appliances; and
there is, moreover, no want of excellent practical manuals.
I have to thank Mr H. J. Priestley for kind assistance in
reading the proofsheets.
H. L.
January, 1910.
CONTENTS
INTRODUCTION
AKT. PAGE
1. Simple Vibrations and Pure Tones 1
2. Musical Notes 3
3. Musical Intervals. Diatonic Scale . .5
CHAPTER I
THEORY OF VIBRATIONS
4. The Pendulum 8
5. Simple Harmonic Motion 9
6. Further Examples 11
7. Dynamics of a System with One Degree of Freedom. Free
Oscillations 12
8. Forced Oscillations of a Pendulum 16
9. Forced Oscillations in any System with One Degree of
Freedom. Selective Resonance 20
10. Superposition of Simple Vibrations 22
11. Free Oscillations with Friction ...... 24
12. General Dissipative System with One Degree of Freedom.
Effect of Periodic Disturbing Forces .... 27
13. Effect of Damping on Resonance 32
14. Systems of Multiple Freedom. Examples. The Double
Pendulum 34
15. General Equations of a Multiple System . . . . 41
16. Free Periods of a Multiple System. Stationary Property . 44
17. Forced Oscillations of a Multiple System. Principle of
Reciprocity 47
18. Composition of SimpleHarmonic Vibrations in Different
Directions 48
19. Transition to Continuous Systems 52
20. On the Use of Imaginary Quantities 53
21. Historical Note 58
CONTENTS
CHAPTER II
STKINGS
ART. PAGE
22. Equation of Motion. Energy ...... 59
23. Waves on an Unlimited String 61
24. Eeflection. Periodic Motion of a Finite String ... 64
25. Normal Modes of Finite String. Harmonics ... 68
26. String excited by Plucking, or by Impact .... 72
27. Vibrations of a Violin String 75
28. Forced Vibrations of a String 80
29. Qualifications to the Theory of Strings .... 81
30. Vibrations of a Loaded String ...... 82
31. Hanging Chain . . . .... . . 84
CHAPTER III
FOURIER'S THEOREM
32. The SineSeries . ... . . . . . 87
33. The CosineSeries 92
34. Complete Form of Fourier's Theorem. Discontinuities . 92
35. Law of Convergence of Coefficients ... . . . 94
36. Physical Approximation. Case of Plucked String . 96
37. Application to Violin String . . . . . . 98
38. String Excited by Impact 99
39. General Theory of Normal Functions. Harmonic Analysis 101
CHAPTER IV
BARS
40. Elementary Theory of Elasticity. Strains . . . . 106
41. Stresses 108
42. Elastic Constants. Potential Energy of Deformation . . 110
43. Longitudinal Vibrations of Bars . . ... . . 114
44. Plane Waves in an Elastic Medium . . . . . 118
45. Flexural Vibrations of a Bar . . . . . .120
46. Freefree Bar . . . . . . . .... 124
47. Clampedfree Bar 127
48. Summary of Results. Forced Vibrations . . ... 130
49. Applications . 131
50. Effect of Permanent Tension . ... . . . . 132
51. Vibrations of a Ring. Flexural and Extensional Modes . 133
CONTENTS
CHAPTER V
MEMBRANES AND PLATES
ABT. PAGE
52. Equation of Motion of a Membrane. Energy . . . 139
53. Square Membrane. Normal Modes . . . . . 142
54. Circular Membrane. Normal Modes .'"''/ . . . 144
55. Uniform Flexure of a Plate ...'.... 150
56. Vibrations of a Plate. General Results . . k /. 152
57. Vibrations of Curved Shells . 155
CHAPTER VI
PLANE WAVES OF SOUND
58. Elasticity of Gases . . . ... . , . . 157
59. Plane Waves. Velocity of Sound . . . . . . 160
60. Energy of Sound Waves . . . ... ... 163
61. Reflection . . . . 168
62. Vibrations of a Column of Air . . . , . . 170
63. Waves of Finite Amplitude . / ... . . 174
64. Viscosity . . 183
65. Effect of Heat Conduction . . . ; . . . . 187
66. Damping of Waves in Narrow Tubes and Crevices . . 190
CHAPTER VII
GENERAL THEORY OF SOUND WAVES
67. Definitions. Flux. Divergence 197
68. Equations of Motion . 200
69. VelocityPotential 201
70. General Equation of Sound Waves 204
71. Spherical Waves 205
72. Waves resulting from a given Initial Disturbance . . 212
73. Sources of Sound. Reflection 214
74. Refraction due to Variation of Temperature . . . 216
75. Refraction by Wind 219
CHAPTER VIII
SIMPLEHARMONIC WAVES. DIFFRACTION
76. Spherical Waves. PointSources of Sound .... 223
77. Vibrating Sphere 228
78. Effect of a Local Periodic Force . 233
viii CONTENTS
ART. PAGE
79. Waves generated by Vibrating Solid 236
80. Communication of Vibrations to a Gas .... 237
81. Scattering of Sound Waves by an Obstacle . . . 240
82. Transmission of Sound by an Aperture .... 244
83. Contrast between Diffraction Effects in Sound and Light.
Influence of WaveLength . . . . . "''.'!. 248
CHAPTER IX
PIPES AND RESONATORS
84. Normal Modes of Rectangular and Spherical Vessels . 254
85. Vibrations in a Cylindrical Vessel 259
86. Free Vibrations of a Resonator. Dissipation . . . 260
87. Corrected Theory of the Organ Pipe 266
88. Resonator under Influence of External Source. Reaction
on the Source ... . . . t v, . . 270
89. Mode of Action of an Organ Pipe. Vibrations caused by
Heat . . . . 276
90. Theory of ReedPipes . . . . . . . . 278
CHAPTER X
PHYSIOLOGICAL ACOUSTICS
91. Analysis of Sound Sensations. Musical Notes . . . 284
92. Influence of Overtones on Quality 286
93. Interference of Pure Tones. Influence on the Definition
of Intervals . . . 287
94. Helmholtz Theory of Audition . . . . . . 289
95. CombinationTones 292
96. Influence of CombinationTones on Musical Intervals ' . 297
97. Perception of Direction of Sound . . . ' . . 298
INDEX . 301
THE DYNAMICAL THEOEY
OF SOUND
INTRODUCTION
1. Simple Vibrations and Pure Tones.
In any ordinary phenomenon of sound we are concerned,
first with the vibrating body, e.g. a string or a tuning fork or
a column of air, in which the disturbance originates, secondly
with the transmission of the vibrations through the aerial
medium, next with the sensations which the impact of the
waves on the drum of the ear somehow and indirectly produces,
and finally with the interpretation which, guided mainly and
perhaps altogether by experience, we put upon these sensations.
It is in something like this natural order that the subject
will be discussed in the following pages, but the later stages
involving physiological and psychological questions can only be
touched upon very lightly.
As few readers are likely to take up this book without
some previous knowledge of the subject we may briefly re
capitulate a few points which will be more or less familiar, with
the view of fixing the meaning of some technical terms which
will be of constant occurrence. Many of the matters here
referred to will of course be dealt with more fully later.
The frontier between physics and physiology is reached at
the tympanic membrane, and from the physical standpoint it is
to the variations of pressure in the external earcavity that we
must in the last resort look, under normal (as distinguished
from pathological) conditions, for the cause of whatever sensations
of sound we experience. These variations may conveniently
be imagined to be exhibited graphically, like the ordinary
variations of barometric pressure, by a curve in which the
abscissae represent times and the ordinates deviations of the
L. 1
2 , DYNAMICAL .THEOKY OF SOUND
pressure on one side or other of the mean, the only difference
being that the horizontal and vertical scales are now enormously
magnified.
The variety of such curves is of course endless, and it is
impossible to suppose that a distinct provision is made in the
ear for the recognition of each, or even of each of the numerous
classes into which they might conceivably be grouped. It is
therefore necessary to analyse, as far as possible, both the
vibration forms and the resulting sensations into simpler
elements which shall correspond each to each.
As regards the vibrationforms, there is one mode of
resolution which at once claims consideration on dynamical
grounds. The fundamental type of vibration in Mechanics is
that known as "simpleharmonic," which is represented graphic
ally by a curve of sines (Fig. 3, p. 10). This is met with in
the pendulum, and in all other cases of a freely vibrating body
or mechanical system having only one degree of freedom. It
can moreover be shewn that the most complicated oscillation of
any system whatever may, so far as friction can be neglected, be
regarded as made up of a series of vibrations of this kind, each
of which might be excited separately by suitable precautions.
The reason for the preeminent position which the simple
harmonic type occupies in Mechanics is that it is the only type
which retains its character absolutely unchanged whenever it
is transmitted from one system to another. This will be ex
plained more fully in the following chapter.
The analysis of sensations is a much more delicate matter,
and it was a great step in Acoustics when Ohm* in 1843
definitely propounded the doctrine that the simplest and
fundamental type of soundsensation is that which corresponds
to a simpleharmonic vibration. This implies that all other
soundsensations are in reality complex, being made up of
elementary sensations corresponding to the various simple
harmonic constituents into which the vibrationform can be
resolved. The statement is subject to some qualifications, in
particular as to the degree of independence of elementary
* G. S. Ohm (17871854), professor of physics at Munich 184954, known
also as the author of " Ohm's Law " of electric conduction.
INTRODUCTION 3
sensations very near to one another in the scale, but these need
not detain us at present. It may be regarded as in the main
fully established, chiefly in consequence of the labours of
Helmholtz*. The sensation corresponding to a simpleharmonic
vibration is called a " simple tone " or a " pure tone," or merely
a "tone." The sound emitted by a tuning fork fitted with
a suitable resonator, or by a wide stopped organ pipe, gives the
best approach to it.
Since the form of the vibration curve is fixed, the distinction
between one simple tone and another can only be due to
difference of frequency or of amplitude. The " frequency," i.e.
the number of complete vibrations per second, determines the
" pitch," greater frequency corresponding to higher pitch. The
lower and upper limits of frequency for tones audible to the
human ear are put at about 24 and 24,000 respectively; the
range employed in music is much narrower, and extends only
from about 40 to 4000. As between tones of the same pitch,
the amplitude, or rather its square, determines the rate of
supply of energy to the ear and so the relative "intensity,"
but it will be understood that it is physical rather than
subjective intensity that is here involved. Between tones of
different pitch only a vague comparison of loudness is possible,
and this may have little relation to the supply of energy. Near
the limits of audibility the sensation may be feeble, even though
the energysupply be relatively considerable.
2. Musical Notes.
From the chaos of more complex sounds there stands out a
special class, viz. that of musical "notes." The characteristic of
such sounds is that the sensation is smooth, continuous, and
capable (at least in imagination) of indefinite prolongation
without perceptible change. The nature of the corresponding
vibrations is well ascertained. If we investigate any contrivance
* Hermann Helmholtz (1821 94), successively professor of physiology
(Konigsberg 1849), anatomy (Bonn 1855), physiology (Heidelberg 1858) and
physics (Berlin 1871). Reference will often be made to his classical work : Die
Lehre von den Tonempfindungen als physiologische Grundlage filr die Theorie der
Musik, Brunswick, 1862. An English translation from the third edition (1870)
was published by A. J. Ellis under the title Sensations of Tone, London, 1875.
12
4 DYNAMICAL THEOEY OF SOUND
by which a note of good musical quality is actually produced,
we find that the vibration can be resolved into a series of simple
harmonic components whose frequencies stand to one another
in a certain special relation, viz. they are proportional to the
numbers 1, 2, 3, .... Individual members of the series may be
absent, and there is practically a limit on the ascending side,
but no other ratios are admissible. It is evident from the
above relation that the resultant vibrationform is necessarily
periodic in character, recurring exactly at intervals equal to the
period in which the first member of the series goes through its
phases. It must be remembered, however, that the ear has no
knowledge of the periodic character as such, and it must not be
supposed that every periodic vibration will necessarily produce a
sensation which is musically tolerable. The superposition of
simpleharmonic vibrations to produce periodic vibrationforms
is illustrated by some of the diagrams given below in
Chapter III.
One musical note may differ from another in respect of
pitch, quality, and loudriess. The pitch is usually estimated
as that of the first simpleharmonic vibration in the series, viz.
that of lowest frequency, but if the amplitude of this first
component be relatively small, and especially if it fall near the
lower limit of the audible scale, the estimated pitch may be
that of the second component.
By " quality " is meant that unmistakable character which
distinguishes a note on one instrument from the note of the
same pitch as given by another. Every musical instrument
has as a rule its own specific quality*, which is seldom likely to
be confused with that of another. Everyone recognizes for
instance the difference in character between the sound of a
flute, a violin, a trumpet, and the human voice, respectively.
It is obvious that difference of quality, so far as it is not due to
adventitious circumstances f, can only be ascribed to difference
of vibrationform, and so to differences in the relative amplitudes
and phases of the simpleharmonic constituents. According to
* French timbre ; German Klangfarbe.
f Such as the manner in which the sound sets in and ceases ; this is different
for instance in the violin and the piano.
INTRODUCTION 5
Helmholtz the influence of phase is inappreciable. This has
been contested by some writers, but there can be no doubt that
in most cases the difference of quality is a question of relative
amplitudes alone.
Comparisons of loudness can only be made strictly between
sounds of the same quality and about the same pitch.
It follows from the preceding that, so far as Ohm's law is
valid, the sensation of a musical note must be complex, and made
up of the simpler sensations, or tones, which correspond to the
various simpleharmonic elements in the vibrationform. This
doctrine has to contend with strong and to some extent
instinctive prepossessions to the contrary, and some preliminary
training is usually necessary before it is accepted as a fact of
personal experience. We shall return to this question later; at
present we merely record that that element in the sensation
which corresponds to the gravest simpleharmonic constituent
is called the " fundamental tone," and that the others are termed
its " overtones " or " harmonics."
3. Musical Intervals. Diatonic Scale.
There are certain special relations, familiar to trained ears,
in which two notes or two simple tones may stand to one
another. These are the various consonant and other "intervals."
Physically they are marked by the property that the frequencies
corresponding to the respective pitches are in a definite
numerical ratio, which can be expressed by means of two small
integers. The names of the more important consonant intervals,
with the respective ratios, are as follows :
Unison 1 : 1 Octave 1 : 2
Fifth 2 : 3 Fourth 3 : 4
Major Third 4 : 5 Minor Sixth 5 : 8
Minor Third 5 : 6 Major Sixth 3 : 5.
The ear has of course no appreciation of the numerical
relations themselves ; but each interval is more or less sharply
" defined," in the sense that a slight mistuning of either note is
at once detected by the beats, and consequent sensation of
roughness, which are produced. The explanation of these latter
peculiarities must be deferred for the present.
6 DYNAMICAL THEORY OF SOUND
The names given to the various intervals are in a sense
accidental, and refer to the relative positions of the notes on
the ordinary " diatonic scale." This is based on the " major
chord," which is a combination of three notes forming a Major
and a Minor Third ; i.e. their frequencies are as 4 : 5 : 6. If we
start from any arbitrary note, which we will call C, as keynote,
the two notes which lie a Fifth above and below it are called
the " dominant" (G) and the " subdominant " (F,) respectively.
If we form the major chord from C we get the notes E = f C,
and G =  C. Again if we form the major chord from G we get
the notes B = f G =  1 / C, and d = f G = f C. The latter falls
outside the octave beginning with C ; the corresponding note
within the octave is D = f C. Lastly, forming the major chord
from F, we get A, = f F = f x f C = f C, the octave of which is
A =  C, and C itself. We thus obtain the scale of seven notes
whose frequencies are proportional to the numbers here given :
C D E F G A B
i I I f f S
24 27 30 32 36 40 45
This is continued upwards and downwards in octaves ; the same
letters are repeated as the names of the notes, but the various
octaves may be distinguished by difference of type, and by
accents or suffixes. The precise pitch of the keynote is so far
arbitrary; it determines, and is determined by, that of any
other note in the scale. Among musicians the standard has
varied in different places and at different times, the general
tendency being in the direction of a rise. German physical
writers, including Helmholtz, have followed a standard which
assigns to a certain A a frequency of 440*. On this basis we
have the following frequencies for a certain range of the
scale :

<j
f^>
~~Z3 ^~
1
"Z?"
&
cs
c
D
E
F
G
A
B c
d
e
/ g
a b
66
74
82$
88
99
110
123 132
148$
165
176 198
220 247$
* This makes c' = 264. Physical instrument makers now often take c' = 256,
which is convenient on account of its continued divisibility by 2.
INTRODUCTION
V d' J f tf a' V d' d" e" f" g" a" b"
264 297 330 352 396 440 495 528 594 660 704 792 880 990
Underneath the ordinary musical symbols we have placed
the convenient literal notation employed by German writers.
This may be continued upwards by means of additional accents
(c'", c*, ...), and downwards by suffixes (C,, C,,, ...).
If in the construction of the scale we had used, instead of
the major, the minor chord, which consists of a Minor and a
Major Third in ascending order, the frequencies being as
10 : 12 : 15, we should have required three notes not included
in the above scheme. And if, starting from any note already
obtained (other than C) as a new keynote, we proceed to
construct a major or a minor scale, further additional notes are
required. In the case of the violin, or of the human voice, or of
some other windinstruments which allow of continuous varia
tion of pitch, this presents no difficulty. But in instruments
like the piano or organ the multiplication of fixed notes beyond
a moderate limit is impracticable. It is found, however, that
by a slight tampering with the correct numerical relations the
requirements of most keys can be fairly well met by a system
of twelve notes in each octave, which are known as
C C* D D* E F F* G GS A A3 B.
This process of adjustment, or compromise, is called "tempera
ment"; on the usual system of " equal " temperament the
intervals between the successive notes are made equal, the
octave being accordingly divided into twelve steps for each of
which the vibrationratio is 2 T X Thus the ratio of G to C is
made to be 2 T *= T4983 instead of 15.
CHAPTEE I
THEORY OF VIBRATIONS
4. The Pendulum.
A vibrating body, such as a string or a bar or a plate,
cannot give rise to a sound except in so far as it acts on the
surrounding medium, which in turn exerts a certain reaction
on the body. The reaction is however in many cases so slight
that its effects only become sensible after a large number of
oscillations. Hence, to simplify matters, we begin by ignoring
it, and investigate the nature of the vibrations of a mechanical
system considered as completely isolated.
The theory of vibrations begins, historically and naturally,
with the pendulum. With this simple apparatus
we are able to illustrate, in all essentials, many
important principles of acoustics, the mere differ
ences of scale as regards amplitude and period,
enormous as they are, being unimportant from the
dynamical point of view.
A particle of mass M, suspended from a fixed
point by a light string of length I, is supposed
to make small oscillations, in a vertical plane,
about its position of equilibrium. If the inclina
tion of the string to the vertical never exceeds
a few degrees, the vertical displacement of the
particle may (to a first approximation) be neg
lected, and the tension (P) of the string may be
equated to the gravity Mg of the particle. Since the horizontal
displacement (x) is affected only by the horizontal component
of the tension, we have
M =  P  =  Ma  m
Fig. 1.
THEORY OF VIBRATIONS
If we put n* = gll, (2)
d?x
this becomes rf^ +w2a? = ^' ^
and the solution is
x A cos nt + B sin nt, (4)
where the constants A, B may have any values. That this
formula really satisfies (3) is verified at once by differentiation ;
and since it contains two arbitrary constants A, B, we are able
to adapt it to any prescribed initial conditions of displacement
and velocity. Thus if, when < = 0, we are to have #=o? ,
dxjdt = u Q , we find
Un .
cos nt + sin nt.
n
.(5)
It is of course necessary, in the application to the pendulum,
that the initial conditions should be such as are consistent with
the assumed " smallness " of the oscillations. Thus in (5) we
must suppose that the ratios x /l and u /nl are both small. In
virtue of (2) the latter ratio is equal to */(u */gl), so that u
must be small compared with the velocity " due to " half the
length of the pendulum.
5. SimpleHarmonic Motion.
If in 4 (4) we put
A ~D /"I \
as is always possible by a suitable choice of a and e, we get
The particular type of vibration represented by this formula
is of fundamental importance.
It is called a "simpleharmonic,"
or (sometimes) a "simple"
vibration. Its character is best
exhibited if we imagine a
geometrical point Q to describe
a circle of radius a with the
constant angular velocity n.
The orthogonal projection P of
Q on a fixed diameter AOA'
will move exactly according to
10 DYNAMICAL THEORY OF SOUND
the formula (2), provided it be started at the proper instant.
The angle nt+e(=AOQ) is called the "phase"; and the
elements a, e are called the " amplitude " and the " initial
phase," respectively. The interval Zir/n between two suc
cessive transits through the origin in the same direction is
called the " period." In acoustics, where we have to deal with
very rapid vibrations, it is usual to specify, instead of the
period, its reciprocal the " frequency " (N), i.e. the number of
complete vibrations per second ; thus
In the case of the pendulum, where n = *J(g/l), the period
is 2ir^/(l/g). As in the case of all other dynamical systems
which we shall have occasion to consider, this is independent
of the amplitude so long as the latter is small (.
The velocity of P in any position is
............ (3)
as appears also by resolving the velocity (na) of Q parallel
to OA.
Fig. 3.
In all cases of rectilinear motion of a point the method of
graphical representation by means of a curve constructed with
* The want of a separate name for the angular velocity n in the auxiliary
circle is sometimes felt. In the theory of the tides the term "speed" was
introduced by Lord Kelvin. As an alternative term in acoustics the word
"rapidity" may perhaps be suggested.
f This observation was made by Galileo in 1583, the pendulum being a
lamp which hangs in the cathedral of Pisa.
THE OK Y OF VIBRATIONS 11
the time t as abscissa and the displacement x as ordinate is
of great value. This is called the " curve of positions," or the
" spacetime curve." In experimental acoustics numerous
mechanical and optical devices have been contrived by means
of which such curves can be obtained. In the present case
of a simpleharmonic vibration, the formula (2) shews that the
curve in question is the wellknown " curve of sines."
6. Further Examples.
The governing feature in the theory of the pendulum is
that the force acting on the particle is always towards the
position of equilibrium and (to a sufficient approximation)
proportional to the displacement therefrom. All cases of
this kind are covered by the differential equation
and the oscillation is therefore of the type (2) of 5, with
n z = K/M. The motion is therefore simpleharmonic, with
the frequency
determined solely by the nature of the system, and independent
of the amplitude. The structure of this formula should be
noticed, on account of its wide analogies. The frequency
varies as the square root of the ratio of two quantities, one
of which (K) measures the elasticity, or the degree of stability,
of the system, whilst the other is a coefficient of inertia.
Consider, for example, the vertical oscillations of a n
mass M hanging from a fixed support by a helical
spring. In conformity with Hooke's law of elasticity,
we assume that the force exerted by the spring is
equal to the increase of length multiplied by a certain
constant K, which may be called the "stiffness" of
the particular spring. In the position of equilibrium
the tension of the spring exactly balances the gravity
Mg\ and if M be displaced downwards through a
space x, an additional force Kx towards this position
is called into play, so that the equation of motion is of
12 DYNAMICAL THEORY OF SOUND
the type (1). The inertia of the spring itself is here
neglected*.
Again, suppose we have a mass M attached to a wire which
is tightly stretched between
two fixed points with a ten
sion P. We neglect gravity
and the inertia of the wire Flg> 5 '
itself; and we further assume the lateral displacement (x) to
be so small that the change in tension is a negligible fraction
of P. If a, b denote the distances of the attached particle
from the two ends, we have
which is of the same form as 4 (3), with n? = P (a + b)/Mab.
The frequency is therefore
ab
This case is of interest because acoustical frequencies can
easily be realized. Thus if the tension be 10 kilogrammes,
and a mass of 5 grammes be attached at the middle, the
wire being 50 cm. long, we find N = 63.
7. Dynamics of a System with One Degree of Freedom.
Free Oscillations.
The above examples are all concerned with the rectilinear
motion of a particle, but exactly the same type of vibration
is met with in every case of a dynamical system of one degree
of freedom oscillating freely, through a small range, about
a configuration of stable equilibrium.
A system is said to have "one degree of freedom" when
the various configurations which it can assume can all be
specified by assigning the proper values .to a single variable
element or "coordinate." Thus, the position of a cylinder
(of any form of section) rolling on a horizontal plane is defined
by the angle through which it has turned from some standard
position. A system of two particles attached at different points
of a string whose ends A, B are fixed has one degree of freedom
* A correction on this account is investigated in 7.
THEORY OF VIBRATIONS 13
if it be restricted to displacements in the vertical plane through
A, B, for the configuration may be specified by the inclination
of any one of the strings to the horizontal. Again, the con
figuration of a steamengine and of the whole train of machinery
which it actuates is defined by
the angular coordinate of the
flywheel. The variety of such
systems is endless, but if we
exclude frictional or other dis
sipative forces the whole motion
of the system when started ^7
anyhow and left to itself is
governed by the equation of energy. And in the case of
small oscillations about stable equilibrium, the differential
equation of motion, as we shall see, reduces always to the
type 6 (1).
We denote by q the variable coordinate which specifies
the configuration. As in the case of Fig. 6, this may be
chosen in various ways, but the particular choice made is
immaterial. From the definition of the system it is plain
that each particle is restricted to a certain path. If in
consequence of an infinitesimal variation Bq of the coordinate
a particle ra describes an element Ss of its path, we have
8s = a$q, where a is a coefficient which is in general different
for different particles, and also depends on the particular
configuration q from which the variation is made. Hence,
dividing by the timeelement St, the velocity of this particle
is v = adq/dt, or in the fluxional notation *, v = aq.
Hence the total kinetic energy, usually denoted by T, is
T=&(m*) = la#, (1)
where a = 2 (ma 2 ), (2)
the summation X embracing all the particles of the system.
The coefficient a is in general a function of q ; it may be
called the " coefficient of inertia " for the particular configura
tion q. For example, in the case of the rolling cylinder referred
* The use of dots to denote differentiations with respect to t was revived by
Lagrange in the Mecanique Analytique (1788), and again in later times by
Thomson and Tait. We write q for dqjdt and q for
14 DYNAMICAL THEORY OF SOUND
to above, it is the (usually variable) moment of inertia about
the line of contact with the horizontal plane, provided q
denote the angular coordinate.
The potential energy of the system, since it depends on the
configuration, will be a function of q only. If we denote it
by V, the conservation of energy gives
%aq*+ F=const., ..................... (3)
provided the system be free from extraneous forces. The
value of the constant is of course determined by the initial
circumstances. If we differentiate (3) with respect to t, the
resulting equation is divisible by q, and we obtain
which may be regarded as the equation of free motion of the
system, with the unknown reactions between its parts elim
inated. In the application to small oscillations it greatly
simplifies.
In order that there may be equilibrium the equation (4)
must be satisfied by q = const. This requires that d V/dq = ;
i.e. an equilibrium configuration is characterised by the fact
that the potential energy is "stationary" in value for small
deviations from it. By adding or subtracting a constant, we
can choose q so as to vanish in the equilibrium configuration
which is under consideration, whence, expanding in powers of
the small quantity q, we have
F= const. + c 2 +..., .................. (5)
the first power of q being absent on account of the stationary
property. The constant c is positive if the equilibrium con
figuration be stable, and V accordingly then a minimum*. It
may be called the " coefficient of stability."
If we substitute from (5) in (4), and omit terms of the
second order in q, q, we obtain
aq + cq = Q, ........................ (6)
where a may now be supposed to be constant, and to have the
value corresponding to the equilibrium configuration.
* In the opposite case the solution of (6) below would involve real exponen
tials instead of circular functions, indicating instability.
THEORY OF VIBRATIONS 15
Since (6) is of the same type as 6 (1), with
n* = c/a, (7)
the variation of q is simpleharmonic, say
q = C cos (nt + e), (8)
with the frequency
AT
Moreover, since the displacement of any particle of the
system along its path, from its equilibrium position, is pro
portional to q (being equal to aq in the above notation), we see
that each particle will execute a simpleharmonic vibration of
the above frequency, and that the different particles will keep
step with one another, passing through their mean positions
simultaneously. The amplitudes of the respective particles are
moreover in fixed ratios to one another, the absolute amplitude,
and the phase, being alone arbitrary, i.e. dependent on the
particular initial conditions.
The kinetic and potential energies are respectively
T = i of = \ n*aC* sin 2 (nt + e), )
V= I c(? = IcC 2 cos 2 (nt + e), j
the sum being
T+ V=\n\iV* = \cG\ ............... (11)
in virtue of (7). Since the mean values of sin 2 (w + e) and
cos 2 (nt he) are obviously equal, and therefore each =J, the
energy is on the average half kinetic and half potential.
The application of the theory to particular cases requires
only the calculation of the coefficients a and c, the latter being
(in mechanical problems) usually the more troublesome. In
the case of a body attached to a vertical wire, and making
torsional oscillations about the axis of the wire, a is the moment
of inertia about this axis, and c is the modulus of torsion,
i.e. cq is the torsional couple when the body is turned through
an angle q.
Again in the case of a mass suspended by a coiled spring
(Fig. 4), if we assume that the vertical displacement of any
point of the spring is proportional to its depth z below the
16 DYNAMICAL THEOBY OF SOUND
point of suspension in the unstrained state, the kinetic energy
is given by
..................... (12)
if p be the linedensity, I the unstretched length, and q the
displacement of the weight. The inertia of the spring can
therefore be allowed for by imagining the suspended mass to be
increased by onethird that of the spring.
8. Forced Oscillations of a Pendulum.
The vibrations so far considered are " free," i.e. the system
is supposed subject to no forces except those incidental to its
constitution and its relation to the environment. We have
now to examine the effect of disturbing forces, and in particular
that of a force which is a simpleharmonic function of the time.
This kind of case arises when one vibrating body acts on
another under such conditions that the reaction on the first
body may be neglected.
For definiteness we take the case of a mass movable in a
straight line, the subsequent generalization ( 9) being a very
simple matter. The equation (1) of 6 is now replaced by
(1)
the last term representing the disturbing force, whose amplitude
F, and frequency p/Zir, are regarded as given*. If we write
f, .................. (2)
we have j+n?x=fcospt ................... (3)
The complete solution of this equation is
x = A cos nt + B sin nt 4 ~  cos pt t ...... (4)
7i 2 p z
as is easily verified by differentiation.
The first part of this, with its arbitrary constants A } B,
represents a free vibration of the character explained in 5,
* The slightly more general case where the force is represented by F cos (pt + a)
can be allowed for by changing the origin from which t is reckoned.
THEORY OF VIBRATIONS
17
with the frequency H/^TT proper to the system. On this is
superposed a " forced vibration " represented by the last term.
This is of simpleharmonic type, with the frequency p/2?r of the
disturbing force, and the phase is the same as that of the force,
or the opposite, according as p $ n, i.e. according as the imposed
frequency is less or greater than the natural frequency.
The above theory is easily illustrated by means of the
pendulum. If the upper end of the string, instead of being
fixed, is made to execute a horizontal motion in which the
displacement at time t is (Fig. 7), the equation of motion (1)
of 4 is replaced by
.(5)
or
.(6)
This is the same as if the upper end were fixed, and the bob
were subject to a horizontal force whose accelerative effect is
?i 2 f . If as a particular case we take
acospt, ........................ (7)
The annexed Fig. 8 repre
sents the forced oscillation in the two cases of p < n and p > n y
respectively. The pendulum oscillates as if C were a fixed
L. 2
we get the form (3), with /= ?i 2 ct.
18 DYNAMICAL THEOKY OF SOUND
point, the distance CP being equal to the length of the simple
pendulum whose free period is equal to the imposed period
This example is due to Young*, who applied it to illustrate
the dynamical theory of the tides, where the same question of
phase arises. It appears from this theory that the tides in an
imagined equatorial belt of ocean, of a breadth not exceeding
a few degrees of latitude, and of any depth comparable with
the actual depth of the sea, would be "inverted," i.e. there
would be low water beneath the moon, and high water in
longitudes 90 E. and W. from it, the reason being that the
period of the disturbing force (viz. 12 lunar hours) is less than
the corresponding free period, so that there is opposition of
phase.
The arbitrary constants in the complete solution (4) are
determined by the initial conditions. Suppose, for example,
that the body starts from rest in the zero position at the instant
t = 0. We find
x j4 a (cos nt cos pt), ......... . ..... (8)
as may be immediately verified.
When the imposed frequency p/2?r is nearly equal to the
natural period, the last term in (4) becomes very large, and it
may be that the assumption as to the smallness of x on which
the equation (1) is usually based (as in the case of the pendulum)
is thereby violated. The result expressed by (4) is then not to
be accepted without reserve, but we have at all events an indica
tion of the reason why an amplitude of abnormal amount ensues
whenever there is approximate agreement between the free and
the forced period.
In the case (p = ri) of exact coincidence between the two
periods, the solution (4) becomes altogether unmeaning, but an
intelligible result may be obtained if we examine any particular
* Dr Thomas Young (1773 1829), famous for his researches on light, and
other branches of physics. The elementary theory of free and forced oscilla
tions was given by him in an article on " A Theory of the Tides, including the
consideration of Resistance," Nicholsons Journal, 1813 ; Miscellaneous Works,
London, 1855, vol. n., p. 262.
THEORY OF VIBRATIONS 19
case in which the initial conditions are definite. Thus, in the
case of (8), the formula may be written
/ sin \ ( p n) t .
and as p approaches equality with n this tends to the limiting
form
(10)
This may be described (roughly) as a simple vibration
whose amplitude increases proportionally to t. For a reason
just indicated this is only valid as a representation of the earlier
stages of the motion.
The case of a disturbing force of more general character
may be briefly noticed. The differential equation is then of
the form
+ *=/() ................... (11)
The method of solution, by variation of parameters, or
otherwise, is explained in books on differential equations. The
result, which may easily be verified, is
x =  sin nt I f(t) cos nt dt cos nt I f(t) sin nt dt. (12)
It is unnecessary to add explicitly terms of the type
A cos nt + B sin nt, which express the free vibrations, since
these are already present in virtue of the arbitrary constants
implied in the indefinite integrals.
If the force f(t) is only sensible for a certain finite range of
t, and if the particle be originally at rest in the position of
equilibrium, we may write
x =  sin nt I f(t) cos nt dt  cos nt I f(t) sin nt dt, (13)
n J ao n J _QO
since this makes x 0, dx/dt = for t =  oo . The vibra
tion which remains after the force has ceased to be sensible is
accordingly
x = A cos nt + B sin nt, ............... (14)
where
= 1 f(t)smntdt, B = T f(i)cosntdt. (15)
nj oo nj _aj 4
22
20 DYNAMICAL THE OB Y OF SOUND
For example, let
* ..................... < 16 >
this represents a force which is sensible for a greater or less
interval on both sides of the instant t = 0, according to the
value of r, the integral amount or impulse being //,*. By
making r sufficiently small we can approximate as closely as
we please to the case of an instantaneous impulse. Since
cosntdt_7r r smntdt_ m
~ ~
LL6 ** T
we have x=  sin nt ................... (18)
The exponential factor shews the effect of spreading out
the impulse. This effect is greater, the greater the frequency
of the natural vibration.
9. Forced Oscillations in any System with One Degree
of Freedom. Selective Resonance.
The generalization of these results offers no difficulty. When
given extraneous forces act on a system with one degree of
freedom, whose coordinate is q, the work which they perform in
an infinitely small change of configuration, being proportional to
8q, may be denoted by QSq. The quantity Q is called the
"force" acting on the system, "referred to the coordinate q."
For instance, if q be the angular coordinate of a body which can
rotate about a fixed axis, Q is the moment of the extraneous
forces about this axis.
It follows that in any actual motion of the system the rate
at which extraneous forces are doing work is Qq. The equation
of energy now takes the form
j t (T+V)=Qq, ..................... (1)
whence, inserting the value of T from 7 (1), we have
'
* The graph of this function is given, for another purpose, in Fig. 14, p. 33.
t The former of these integrals is evaluated in most books on the Integral
Calculus.
THEORY OF VIBRATIONS 21
When dealing with small motions in the neighbourhood of
a configuration of equilibrium we may neglect terms of the
second order as before. Hence, substituting the value of V
from 7 (5), we find
aq+cq = Q ......................... (3)
When Q is of simpleharmonic type, varying (say) as cos pt,
the forced oscillation is given by
which is of course merely a generalized form of the last term in
8 W
Two special cases may be noticed. When p is very small,
(4) reduces to q = Q/c. This may be described as the "equili
brium" value* of the displacement, viz. it is the statical
displacement which would be maintained by a constant force
equal to the instantaneous value of Q. In other words, it is
the displacement which would be produced if the system were
devoid of inertia (a = 0). Denoting this equilibrium value by
q, we may write (4) in the form
where, as in 7, n denotes the speed of a free vibration.
When, on the other hand, p is very great compared with n,
(4) reduces to
q = Q/p*a, ..................... (6)
approximately. This is almost the same as if the system were
devoid of potential energy, the inertia alone having any sensible
influence.
When two or more disturbing forces of simpleharmonic
type act on a system, the forced vibrations due to them may be
superposed by mere addition. Thus a disturbing force
Q =/ cos ( Pl t + d) +/ 2 cos (pj, + oj + ...... (7)
will produce the forced oscillation
a 2 )H.... (8)
* The name is taken from the theory of the tides, where the equilibrium
tideheight is denned as that which would be maintained by the disturbing
forces if these were to remain permanently at their instantaneous values.
22 DYNAMICAL THEOKY OF SOUND
It will be observed that the amplitudes of the various terms
are not proportional to those of the corresponding terms in the
value of Q, owing to the difference in the denominators.
This is an illustration of a remark made in 1 that the
simpleharmonic type is the only one which is unaltered in
character when it is transmitted, the character of the composite
vibration represented by (8) being different from that of the
generating force. In particular if one of the imposed speeds
p lt p z , ... be nearly coincident with the natural speed n, the
corresponding element in the forced vibration may greatly
predominate over the rest. This is the theory of selective
"resonance," so far as it is possible to develop it without
reference to dissipative forces.
10. Superposition of Simple Vibrations.
The superposition of simpleharmonic motions in the same
straight line has many important applications. For instance,
the height of the tide at any station is the algebraic sum of a
number of simpleharmonic com
ponents, the most considerable
(at many stations) being those
whose periods are half a lunar
and half a solar day, respectively.
The composition of two
simple vibrations may be illus
trated by the geometrical
method of Fig. 2. If two
points Q lt Q 2 describe concentric
circles with the angular velo
cities TH, n^ their projections
on a fixed diameter will execute simpleharmonic vibrations
of the forms
#! = aj cos (nj + e^, # 2 = a 2 cos (n + e 2 ), (1)
where Oj , a 2 are the radii of the two circles, and ej , e 2 are the
initial inclinations of the radii OQi, OQ 2 to the axis of x. The
result of the superposition is
(2)
THEORY OF VIBRATIONS 23
and it appears that the value of x is the projection of OR, the
diagonal of the parallelogram determined by OQ l} OQ 2 .
If Tij = 7i 2 , the two component vibrations have the same period,
the angle QiOQ 2 is constant, and the resultant vibration is
simpleharmonic of the same period.
But if Wj, 7*2 are unequal, the angle QiOQ 2 will vary between
and 180, and OR will oscillate between the values c^ a 2 .
In Lord Kelvin's "tidal clock," the "hands" OQ l} OQ 2 revolve
in half a lunar and half a solar day, respectively, and the sides
QtR, Q 2 R of the parallelogram are formed of rods jointed to
these and to one another. The projection of R then indicates
the tideheight due to the superposition of the lunar and solar
semidiurnal tides.
If the periods Sir/r^, 27r/n z are very nearly though not
exactly equal, the angle QiOQ 2 will vary very little in the course
of a single revolution of OQ l or OQ 2 , and the resultant vibration
may be described, in general terms, as a simple vibration whose
amplitude fluctuates between the limits c^ a z . The period
of a fluctuation is the interval in which one arm OQ l gains four
right angles on the other, or STT/^ n 2 ). Inverting, we see
that the frequency of the fluctuations is the difference of the
frequencies of the two constituent vibrations. We have here
the reason for the alternation of " spring " and " neap " tides,
according as the phases of the lunar and solar semidiurnal
tides agree or are opposed. In acoustics we have the important
phenomenon of " beats " between two tones of slightly different
pitch. The contrast between the maximum and minimum
amplitudes is of course greatest when the amplitudes a lf a^ of
Fig. 10.
the primary vibrations are equal. We then have
x = o^cos (n^t + ej + a? cos (nj f e 2 )
= 2acos {(7*!  n^) t + (e l  e 2 )} cos {J (n^ + w 2 ) t + J (e l I e 2 )}. (3)
This may be described, in the same general manner as before,
24 DYNAMICAL THEORY OF SOUND
as a simple vibration whose period is 27T/J (% + w 2 ), and whose
amplitude oscillates between the limits and 2a, in the time
7T/ (n^ n z ). This is illustrated graphically, with x as ordinate
and t as abscissa, in Fig. 10, for the case of n^ : n z = 41 : 39.
11. Free Oscillations with Friction.
The conception of a dynamical system as perfectly isolated
and free from dissipative forces, which was adapted provisionally
in 4 10, is of course an ideal one. In practice the energy of
free vibrations is gradually used up, or rather converted into
other forms, although in most cases of acoustical interest the
process is a comparatively slow one, in the sense that the
fraction of the energy which is dissipated in the course of a
single period is very minute.
To represent the effects of dissipation, whether this be due
to causes internal to the system, or to the communication of
energy to a surrounding medium, we introduce forces of resist
ance which are proportional to velocity. The forces in question
are by hypothesis functions of the velocity*, and when the
motion is small, the first power only need be regarded.
The equation of free motion of a particle about a position of
equilibrium thus becomes
,, d z x rr dx
M M =  Kx  R di< .................. < J >
where R is the coefficient of resistance. If we write
k, .................. (2)
weget
The solution of this equation may be made to depend on
that of 4 (3) by the following artifice f. We put
//IX
(4)
* We shall see at a later stage (Chap. VIII) that the resistance of a medium
may introduce additional forces depending on the acceleration. These have
the effect of a slight apparent increase of inertia, and contribute nothing to
the dissipation. It is unnecessary to take explicit account of them at present.
t Another method of solution is given in 20.
THEORY OF VIBRATIONS 25
and obtain, on substitution,
We have now three cases to distinguish. If the friction be
relatively small, more precisely if k < 2n, we may put
r^ = 7i 2 J#, ..................... (6)
and the solution of (3) is
y = A cos n't + B sin n't, ............... (7)
whence x = e " **' (A cos n't + B sin n't) ............. (8)
Changing the arbitrary constants, and putting
r = 2/&, ........................... (9)
we have x = ae~^ T cos(n' + e) ................ (10)
This may be described as a modified simpleharmonic vibration
in which the amplitude (ae ~ ^ r ) sinks asymptotically to as t
increases. The time T in which the amplitude is diminished
in the ratio l/e is called the " modulus of decay." The
relation between x and t is exhibited graphically in Fig. 11,
Fig. 11.
where the dotted lines represent portions of the exponential
curves x = ae ~ . For the sake of clearness the rapidity
of decay is here taken to be much greater than it would be in
any ordinary acoustical example.
DYNAMICAL THEOEY OF SOUND
We have seen that a true simpleharmonic vibration may
be regarded as the orthogonal projection of uniform motion in
a circle. An analogous representation of the modified type (10)
is obtained if we replace the circle by an equiangular spiral
described with constant angular velocity ri about the pole 0, in
the direction in which the radius vector r decreases*. The
formula (10) is in fact equivalent to # = rcos#, provided
r = ae~ il \ 6 = n't + (11)
Eliminating t we have
r=ae~ f "> (12)
where = (n'r)~ l , a. ae^ . This is the polar equation of the
spiral in question. The curve in Fig. 12 corresponds in scale
with Fig. 11.
In most acoustical applications the fraction k/2n, or 1/nr, is
a very small quantity.
In this case, the dif
ference between n and
ri is a small quantity
of the second order,
and may usually be ig
nored ; in other words,
the effect of friction on
the period is insensible.
It may be noted that
the quantityl/nr, whose
square is neglected, is
the ratio of the period
27T/71 to the time 27TT
in which the amplitude is diminished in the ratio e~ * or ^.
If k be greater than 2n the form of the solution of (3) is
altered, viz. we have
*, (13)
Fig. 12.
y
whence
if
Ae
at
(14)
* This theorem was given in 1867 by P. G. Tait (18311901), Professor of
Natural Philosophy at Edinburgh (18601901).
THEORY OF VIBRATIONS 27
The particle comes asymptotically to rest but does not oscillate;
in fact we may easily see that it passes once at most through
its zero position. This type of motion is realized in the case of
a pendulum swinging in a very viscous liquid, and in "deadbeat"
galvanometers and other electrical instruments, but it is ' of
little interest in acoustics.
If k = 2n, exactly, the solution of (3) is of the form
x = (A +Bt)e~ nt , .................. (16)
as to which similar remarks may be made.
12. General Dissipative System with One Degree of
Freedom. Effect of Periodic Disturbing Forces.
The effect of dissipation on the free motion of any system
having one degree of freedom is allowed for by the assumption
that there is a loss of mechanical energy at a rate proportional
to the square of the generalized velocity, so that in the notation
of 7
J <#) = &?, ............... (1)
whence aq + bq + cq = Q ................... (2)
This is of course the same as if we had introduced a frictional
force Q =  bq in 9 (3).
The equation (1) has the same form as 11 (3), and the
results will correspond if we put
n 2 = c/a, r = 2a/6 ................... (3)
When the dissipation is small, the rate of decay of the
amplitude can be estimated by an independent method, due to
Stokes*, which we shall often find useful. The period being
practically unaffected by vicosity, a considerable number of
oscillations can be fairly represented by
q = G cos (nt + e), ..................... (4)
provided C and e be gradually changed so as to fit the altering
circumstances. The average energy over such an interval will
be Jrc 2 a(7 2 , approximately, by 7 (11); and the rate of dissipa
tion will be
bcf = iw 2 60 2 (1  cos 2 (nt + e)},
* Sir George Gabriel Stokes (18191903), Lucasian Professor of Mathematics
at Cambridge (18491903).
28 DYNAMICAL THEOEY OF SOUND
the mean value of which is ^ri*bC\ Equating the mean rate of
decay of the energy to the mean dissipation, we get
ifrtC', .................. (5)
< 6 >
or tf=<7 <T t/T , ........................ (7)
if T = 2a/6, as in (3).
When there are given extraneous forces in addition to the
dissipative influences, the equation of energy takes the form
& ............ (8)
whence aq + bq + cq = Q ...................... (9)
As in 9 we consider specially the case of a disturbing
force of simpleharmonic type, say
Q = Ccospt ...................... (10)
A particular solution of (9) is then obtained in the form
q = Fcospt + Gsinpt, ............... (11)
provided the constants F, G are properly chosen. The necessary
conditions are found on substitution to be
 .
If we put
Rcosa, pb=Rsina, ......... (13)
we find F=cosa, = sina, ............... (14)
c
whence* q = ^ cos (pt a) (15)
The values of R and a are determined by
/nTi
(16)
R is to be taken positively, and a may be assumed to lie
between and TT.
* A more rapid way of obtaining this solution is explained in 20.
THEORY OF VIBRATIONS 29
The equation (9) is still satisfied if we add to (15) terms
representing a free oscillation; and these added terms are
necessary in order to constitute a complete solution capable of
adjustment to arbitrary initial conditions. The free vibration
dies out, however, asymptotically, so that after the lapse of a
sufficient time the forced vibration (15) is alone sensible.
The circumstances which affect the amplitude and phase of
this forced vibration require careful attention. The amplitude
is a maximum when R? is least, i.e. when
(!. a)
and the maximum amplitude is accordingly
In most cases of interest 6 2 /oc is a small quantity of the
second order; the maximum is then C/nb, and occurs when
p = n, very approximately.
Again, it appears from (15) and (16) that the phase of q lags
behind that of the disturbing force by an angle a, which lies
between and JTT, or between TT and TT, according as p 2 is less
or greater than c/a, i.e. according as the imposed frequency
is less or greater than the natural frequency. If, keeping p
constant, we diminish the dissipationcoefficient b, a. tends to
the limit or TT, respectively, in accordance with 8, where we
found exact agreement or opposition of phase in the absence
of resistance. But if, keeping b constant, we make p approach
the value n ( = \/(c/a)) which determines the frequency in the
absence of dissipation, a tends to the limit ^TT, and the phases
of q and Q differ by an amount corresponding to a quarterperiod.
This means that the maxima of the disturbing force are now
synchronous with the maxima of the velocity q.
Some light is thrown on these relations if we examine
the case of a pendulum whose bob receives equal positive
and negative instantaneous impulses alternately at regular
intervals. It is seen at once from Fig. 13 that an impulse in
the direction of motion accelerates or retards the phase of an
otherwise free vibration, according as it precedes or follows
30
DYNAMICAL THEORY OF SOUND
(within the limits of a quarterperiod) the instant of maximum
velocity. Thus if when the particle is at P, on its way to 0,
the velocity be increased in the ratio of PQ to PQi, the phase
is accelerated by the angle QOQ l} whilst a similar impulse at P'
would retard the phase by the angle Q'OQ\.
In order that no effect may be produced on the phase it
is necessary that the impulse be delivered at the instant of
passing through 0. If we imagine that a small assisting
impulse is given at every such passage, as in the case of the
ordinary clock escapement, we have an illustration of the
circumstances of maxi
The
mum resonance,
period of the disturbing
force is exactly equal to
the natural period, and
the force synchronizes
with the velocity. The
amplitude is deter
mined by the considera
tion that the work done
by the impulses must
balance that lost by
friction. The result is
not essentially different
if the impulse be dif
Fig. 13.
fused symmetrically about 0, as in the case of a simple
harmonic force, since the acceleration of phase on one side of
is cancelled by the retardation on the other.
Next suppose that the assisting impulses are given
each time the bob passes the symmetrically situated points
P, P' inwards. There is an acceleration of phase at each
impulse, and the period is shortened. This illustrates the case
of a disturbing force whose period is less than the natural
period, and whose maxima and minima precede the maxima and
minima of the velocity. If on the other hand the impulses are
given as the bob passes the points P and P' outwards, there is
a repeated retardation of phase, and the period is lengthened.
This corresponds to the case of a disturbing force whose period
THEORY OF V1BKATIONS 31
is greater than the natural period ; the maxima and minima
of the force now follow those of the velocity. The reader is
recommended to follow out in detail the argument here sketched,
and to examine the effect of substituting a continuous simple
harmonic force for the series of disconnected impulses. An
explanation may also be found, on the same principles, of the
fact that a small frictional force varying as the velocity has no
sensible effect on the free period.
We return to the analytical discussion. A difference of
phase between the force and the displacement is essential in
order that the disturbing force may supply energy to compensate
that which is continually being lost by dissipation. When, as
in 9, there is complete agreement (or opposition) of phase
between q and Q, the force is, in astronomical phrase, "in
quadrature " with the velocity q, that is, the phases differ by \ir,
and the total work done in a complete period is zero. Under
the present circumstances the disturbing force is at any instant
doing work at the rate
Qq = ^D~ s i n (pt ~ cos^tf
= g {sin a sin (2^ a) j, ............ (19)
the mean value of which is
(20)
The same expression is of course obtained as the mean value
of bq 2 , since the energy supplied by the disturbing force must
exactly compensate, on the average, that which is continually
being lost by dissipation, the mean energy stored in the system
being constant.
It follows from (16) and (20) that the dissipation is greatest
when OL^TT, or p = n, i.e. when the imposed frequency coincides
with that of the free vibration in the absence of resistance.
The maximum value is %C*/b, being greater, of course, the
smaller the value of b.
32 DYNAMICAL THEORY OF SOUND
13. Effect of Damping on Resonance.
The abnormal amplitude and dissipation which ensue
whenever the imposed period is equal, or nearly equal, to the
natural period constitute the phenomenon of "resonance,"
already referred to in 8, of which we shall have many
acoustical examples in the sequel. It may be illustrated
mechanically by giving a slight toandfro motion of suitable
period to the point of suspension of a simple pendulum, or
better by means of a double pendulum ( 14), i.e. an arrange
ment in which two weights are attached at different points to
a string hanging vertically from a fixed point. If the upper
weight (M ) be considerable, whilst the lower one (ra) is relatively
small, M will swing almost exactly like the bob of a simple
pendulum, the reaction of ra being slight. Under these
conditions the motion of ra is practically that of a pendulum
whose point of suspension has an imposed simpleharmonic
vibration ( 8), and if the length of the lower portion of the
string be properly adjusted, a violent motion of ra may ensue.
One very important point remains to be mentioned. As the
interval p/n between the forced and the natural frequencies
diverges from unity (on either side), the dissipation falls off
from its maximum the more rapidly, the smaller the value of
the frictional coefficient b. In other words, the greater the
intensity of the resonance in the case of exact coincidence of
frequencies, the narrower the range over which it is approxi
mately equal to the maximum. For example, a tuning fork, even
when mounted on a "resonance box," requires very perfect tuning
in order that it may be excited perceptibly by the vibrations of
another fork in the neighbourhood, whereas the column of air
in a nearly closed vessel (e.g. a bottle or an organ pipe) will
respond vigorously to a much wider range of frequencies. To
elucidate the point, we notice that the expression (20) of 12
for the dissipation may be written
,,
26 S1
where = %nb/c = I/WT, ........ . ............ (2)
THEORY OF VIBRATIONS 33
in the notation of 12 (3). The second factor has its maximum
value 1/fi when p = n, and evidently diminishes more rapidly,
as p/n deviates from unity, the smaller the value of 0. The
question may be conveniently illustrated graphically by con
structing a curve which shall shew the dissipation corresponding
to different frequencies. As regards the abscissa, it would in
strictness be most proper to take, not the ratio p/n, but its
logarithm, since equal intervals (in the musical sense) then
correspond to equal lengths of the axis of x. We might
therefore write
,(3)
but when, as usually happens, the sensible resonance is confined
to a small range of p/n, we may use the simpler formulae
ft
Hr*
1
.(4)
The curve represented by the latter equation is symmetrical
about the axis of y, and approaches the axis of x asymptotically
as x increases. It is evident that if $ be increased in any
ratio, the new curve is obtained by increasing all the abscissae in
that ratio, and diminishing the ordinates in the inverse ratio,
the area (TT) included between the curve and the axis of x being
L. 3
34 DYNAMICAL THEORY OF SOUND
unaltered. The intensity sinks to onehalf its maximum when
a? = fp, or
ll (5)
n nr
Thus if the damping be such that a free vibration would have
its amplitude diminished in the ratio l/e in 10, 100, 1000
periods*, respectively, the corresponding values of the interval
p/n at which the dissipation would be reduced to onehalf the
maximum would be 1 '016, 1 '0016, 1 '00016. The curve
in (4) is shewn in Fig. 14.
The above argument deals with the dissipation, which is the
most important feature. The consideration of the square of the
amplitude, or of the energy stored in the system, leads to very
similar results, especially when the damping is slight.
14. Systems of Multiple Freedom. Examples. The
Double Pendulum.
We approach the consideration of systems having any finite
number of degrees of freedom. A system is said to have ra
such degrees when m independent variables, or " coordinates,"
are required and are sufficient to specify the various configura
tions which it can assume. The notion, first brought into
formal prominence by Lord Kelvin f, has a wide application
in mechanism and in theoretical mechanics. In the case of
the telescope of an altazimuth instrument or of an equatorial
we have m = 2; in the gyroscope, or (more generally) in any
case of a rigid body free to turn about a fixed point, m = 3 ;
for a rigid structure or frame movable in two dimensions
m = 3; for a rigid structure freely movable in space m = 6.
The choice of the coordinates in any particular case can be
made in an endless variety of ways, but the number is always
determinate. Thus in technical mechanics we have the pro
position that a rigid frame movable in one plane can be fixed by
* In an experiment by Lord Eayleigh, the number of periods for a particular
tuning fork of 256 v.s. was about 5900. When a resonator was used the number
fell to 3300. Theory of Sound, vol. n., p. 436.
t William Thomson, afterwards Lord Kelvin (18241907), Professor of
Natural Philosophy at Glasgow 184699. The matter is explained in Thomson
and Tait's Natural Philosophy, 2nd ed., 195201 (1879).
THEORY OF VIBRATIONS 35
means of three links connecting any three points of it to any
three fixed points in the plane*. Similarly any rigid three
dimensional structure can be anchored firmly by six links
connecting six points of it with six points fixed relatively
to the earth.
Proceeding to the vibrations of a multiple system about
a configuration of equilibrium, we begin as before with the
examination of a few particular cases.
Take first the oscillations of a particle in a smooth bowl of
any continuous shape. By means of suitable constraints, the
particle may be restricted to oscillate in any given vertical
plane through the lowest point 0, e.g. by confining it between
two frictionless guides infinitely close to one another. In
general there will be a lateral pressure on one or other of these
guides, which will however vanish if the plane in question
passes through either of the principal directions of curvature
at 0. Hence two modes of free simpleharmonic vibration, in
perpendicular directions, are possible, with speeds
nxVto/A), n, = ^(g/R,\ (1)
where B^, R 2 , are the radii of curvature of the principal sections
at 0. On account of the assumed smallness of the motion,
these vibrations may be superposed. The result is, if x, y be
horizontal rectangular coordinates through 0,
x = A l cos n t t + A 2 sin rz^,]
y = Bj cos n^t + B 2 sin nj.)
Since this contains four arbitrary constants, we can adjust
the solution to given initial values of x, y, x, y.
This case is very neatly illustrated by Blackburn's pen
dulum ( (Fig. 15). A weight hangs by a string CP from a point
C of a string A CB whose ends A , B are fixed. The strings being
supposed destitute of inertia, the point P will always be in the
same plane with A, B, C. Under this condition the locus of
P is the ringshaped surface generated by revolving a circle
* Provided the directions of the three links be not concurrent (or parallel).
There is a proviso of a more complex character in the case which follows ; but
such details need not occupy us here.
t H. Blackburn, Professor of Mathematics at Glasgow 184979.
32
36
DYNAMICAL THEORY OF SOUND
with centre C and radius CP, in the plane ACB, about AB as
an axis; and the principal radii of curvature at the lowest point
are E l  CO, R z = EO, where E is the point of A B vertically
above 0. The corresponding directions of vibration are re
spectively in and perpendicular to the plane ABO.
Fig. 16.
PC
Fig. 15.
Fig. 17.
Another very simple case is that of two equal particles M
attached symmetrically at distances a from the ends of a tense
string, whose total length is, say, 2 (a + 6), so that 26 denotes
the length of the central portion. One obvious mode of
simpleharmonic vibration is that in which the deflections of the
two particles are always equal and of the same sign (Fig. 16).
If P be the tension of the string, the equation of motion of
either particle is then
*5s >?. (8)
and the speed is therefore
(4)
In another mode the two deflections are equal in magnitude
and opposite in sign, so that the middle point of the string is
stationary (Fig. 17). The circumstances are therefore exactly
the same as in 6, and the speed is
'P a + b\
^r)' (5)
greater, as we should expect, than 7^. If we denote the
THEORY OF VIBRATIONS
37
deflections of the two particles by x, y, the superposition of
the two modes gives
x = A cos (nj + eO + B cos (n + e 2 ),)
\ (b)
y = A cos (%< 4 61) B cos (n z t 4 e 2 ),
where the four constants JL, 5, e 1} e 2 are arbitrary.
In the case of three attached particles the nature of the
various modes is not so immediately obvious, even in the case
of symmetry. We will suppose that the masses are equal,
and that they divide the line into four equal segments a.
Denoting the deflections by x, y, z, we have
P
dt~ a a ' (
dt 2 a, a )
If we put, for shortness, fi = P/Ma, these may be written
.(7)
.(8)
To ascertain the existence of modes of vibration in which
the motion of each particle is simpleharmonic, with the same
period and phase, we assume, tentatively,
x = A cos (nt f e), y = B cos (nt + e), z C cos (nt f e). (9)
It appears, on substitution in (8), that the equations will be
satisfied provided
(10)
These three equations determine the two ratios A : B : C
and the value of ri 2 . Eliminating the former ratios we have
............. (11)
38
DYNAMICAL THEORY OF SOUND
This is a cubic in n\ One root is nf = 2/z, and we find on
reference to (10) that this makes B l = 0, A l =  C 1} and there
fore
x = A l cos (njt 4 6j), i/=0, z A 1 cos(n l t + e l ). ...(12)
This mode might have been foreseen, and its frequency
determined at once, as in the preceding example. The
remaining roots of (11) are
and it appears from (10) that these make
A 2 =C 2) B 2 = </2A 2 , and A 3 =C 3 , B 3
respectively. The corresponding modes are therefore
x = A 2 cos (n + 6 2 ), y =  V2 A 2 cos (n 2 t + e 2 ),
z A 2 cos (n 2 t + 6 2 ),
and
x A 3 cos (n s t + e 3 ), y = V2 A 3 cos (n 3 t + e 3 ),
^ = J. 3 cos (n 3 ^ + e 3 ).
These are shewn, along with the former mode, in Fig. 18.
The complete solution of the equations is obtained by super
position of (12), (13) and (14), and contains the six arbitrary
constants A lf A 2 , A S) e l} e 2 , e 3 .
.(13)
.(14)
Fig. 18.
We conclude these illustrations with the case of the double
pendulum, where we are entirely dependent on general method.
A mass M hangs from a fixed point by a string of length a,
and a second mass ra hangs from M by a string of length b.
For simplicity we suppose the motion confined to one vertical
THEORY OF VIBRATIONS
plane. The horizontal excursions x, y of M, m respectively
being supposed small, the tensions of the upper and lower
strings will be (M + m)g and mg, approximately. The equa
tions of motion are therefore
d*x
m ^ = ~ m 9
x
(15)
To find the possible modes of simpleharmonic vibration we
assume
x = A cos (nt + e), y B cos (nt + e) ....... (16)
The equations are satisfied provided
...(17)
.(18)
where /ji =
Eliminating the ratio A : B, we find
.(19)
which is a quadratic in n z . The condition
for real roots, viz.
r
ab
.(20)
is obviously always fulfilled. It is further
easily seen that both roots are positive, so
that n also is real.
The problem includes a number of inter
esting special cases, but we will only notice
one or two. If the ratio /*, = m/(M 4 m),
be small, the two roots of (19) are nf^g/a,
nf=g/b, approximately. In the former
case M oscillates like the bob of a simple
pendulum of length a, whilst m executes
what may be regarded as a forced oscillation
40 DYNAMICAL THEOKY OF SOUND
of the corresponding frequency ; this case has already been
referred to in 13. In the second mode the ratio A : B is small,
as appears from the second of equations (17); M is then nearly
at rest, whilst m oscillates like the bob of a pendulum of
length b.
Since the expression on the lefthand side of (20) cannot
vanish, the two frequencies can never exactly coincide, but they
become approximately equal if a = 6, nearly, and //, is small.
A curious phenomenon may then present itself. The motion
of each mass, being made up of two superposed simpleharmonic
vibrations of nearly equal period, may fluctuate greatly in
extent, and if the amplitudes of the two vibrations are equal
we have periods of approximate rest, as explained in 10. The
motion then appears to be transferred alternately from m to M,
and from M to m, at regular intervals*.
If, on the other hand, M is small compared with m, p is nearly
equal to unity, and the two roots of (19) are ?i 2 = g/(a + b) and
n* = mg/M . (a f b)/ab, approximately. The former root makes
B/A = (a + b)/a, nearly, so that the two masses are always
nearly in a line with the point of suspension, m now oscillating
like the bob of a pendulum of length a + b. In the second
mode the ratio B/A is small, so that m is approximately at
rest ; the motion of M is then like that of a particle attached
to a string which is stretched between fixed points with a
tension mg (cf. 6).
Another case of interest is obtained if we make a infinite.
One root of (19) then vanishes, and the other is
which makes A/B =  m/M. This indicates that if the support
of a simple pendulum yield horizontally, but without elasticity,
the frequency is increased in a certain ratio which is of course
* The influence of dissipation is of course here neglected. If m be subject
to a frictional resistance, and especially if the modulus of decay be less than
the period of the fluctuation given by the above theory, the phenomena are
modified, and the illustration of the theory of resonance ( 12) is improved.
There is now a continual, though possibly a slow, drain on the original energy
otM.
THEOKY OF VIBKATIONS 41
smaller the greater the inertia of the support. This is however
more easily seen directly.
15. General Equations of a Multiple System.
The general theory of the small oscillations of a multiple
system can only be given here in outline. In the case of one degree
of freedom ( 7) it was possible to base the theory on the equation
of energy alone, but when we have more than one dependent
variable this is no longer sufficient, and some further appeal
must be made to Dynamics. For brevity of statement we will
suppose that there are two degrees of freedom, but there is
nothing in the argument which cannot at once be extended to
the general case.
We imagine, then, a system such that every configuration
which we need consider can be specified by means of two
independent geometric variables or "coordinates" q l9 q 2 . If in
any configuration (q l} q^) the coordinate q 1 (alone) receive an
infinitesimal variation 8q lt any particle ra of the system will
undergo a displacement 88 1 = a 1 8q 1 in a certain direction.
Similarly if q z alone be varied m will be displaced through a
space Bs 2 = *q* in a certain direction, different in general from
the former. The resultant displacement Bs when both variations
are made is given by
8s* = Bs, 2 + 2&! &? 2 cos 6 + &? 2 2
= di 2 Bqi* + 2a 1 2 cos 6 8q 1 Bq^ + c^ 2 fy 2 2 , (1)
where 6 denotes the angle between the directions of 8s lt Bs 2 .
If we divide by Bt*, we obtain the square of the velocity v
of the particle m, in any motion of the system through the con
figuration (q lt q 2 ), in terms of the generalized "components of
velocity" q lt q 2 , thus
v z = ct^ 2 + 2^0, cos Oqfa + a 2 2 ? 2 2 (2)
The total kinetic energy of the system is therefore given by
2T = 2 (mi; 2 ) = a u tf + 2a 12 ^ 2 4 o^ 2 , (3)
where
a n = 2, (rav), a 12
the summation 2 extending over all the particles m of the
42 DYNAMICAL THEOBY OF SOUND
system. The coefficients a n , a 12 , a^ are in general functions of
q lt q 2 ; they are called the "coefficients of inertia" for the par
ticular configuration considered.
Next, let FI denote the total force acting on m, resolved
in the direction of s ly and let F 2 have the corresponding meaning
for the direction of Bs 2 . The work done on the system in any
infinitesimal displacement will therefore be
2 (FM + F 2 Ss 2 ) = 2 (F&) % + 2 (F^) 8q 2 . ..... .(5)
If there are no extraneous forces, this work is accounted for
by a diminution in the potential energy V of the system. When
extraneous forces act we have in addition the work due to these,
which we may suppose expressed in the form
The coefficients Q lt Q 2 are called, by an obvious analogy, the
generalized "components of (extraneous) force." Hence
. . .(6)
whence
In the application to small oscillations we assume that q lt q 2
are small quantities vanishing in the configuration of equi
librium, and for consistency we must also suppose that the
disturbing forces Q lt Q 2 are small. The quantities a l9 2 and
therefore also a n , a l2 , a& may now be treated as constants.
The velocity of the particle m is made up of components a^,
a 2 <?2 in the directions & x and 8s 2 , respectively; and if we neglect
the squares of small quantities its acceleration is made up in
like manner of components a^, ,#.,*. Hence resolving in the
direction of Bsi the forces acting on m we have
m (,& + ct 2 q 2 cos 6) = F l ,
and similarly m (a^ cos 6 + o^q 2 ) = F 2 .)
* The former of these two quantities is (to the first order) the acceleration
calculated on the supposition that q\ alone varies, and the latter is the accelera
tion when #2 alone varies. It is only on the hypothesis of infinitely small
motions that the resultant acceleration is obtained by superposition of these.
THEORY OF VIBRATIONS 43
If we multiply the former of these equations by ^ and the
second by a^, and sum for all the particles of the system, we
find, with the notation of (4),
dV .
.(9)
and similarly a^ + a&q 2 = 5 h Q
oq%
where 021 is of course identical with a 12 .
When there are no extraneous forces these equations are by
hypothesis satisfied by q^ = 0, q a = 0. The configuration of
equilibrium is therefore characterized by the property that
i =0 ' f (10)
in other words, the potential energy is stationary for all infini
tesimal displacements therefrom. Hence if V be expanded in
powers of q lt q 2 , the terms of the first order will be absent, and
we may write with sufficient approximation
2 V = Caq? + 2c 12 <? 1 <7 2 + Ca^jj 2 , (11)
a constant term being omitted. The quantities c u , c 12 , c^ are
called the " coefficients of stability."
Hence (9) may be written
a^i + ^22^2
where c 21 = c 12 .
If we look back to any of the special problems of 14 we
shall recognize that the equations of motion are in fact of this
type. For example, in the case of the doublependulum we
have
...(13)
The formulae therefore correspond if we put
ft = ar, q* = y, \
a n = M, 12 = 0, 022 = m, V ...(14)
c n = (M + m) g/a + mg/b, c u =  mg/b, C& = mg/b.J
44 DYNAMICAL THEOKY OF SOUND
The general case of m degrees of freedom hardly differs
except in the length of the formulae. We have then m equations
of the type
a*i?i + a# 2 + .  + a sm q m + c gl ^ + c^ + . . . 4 C 8m q m = Q s , (15)
where s is any one of the integers 1, 2, 3,...m.
16. Free Periods of a Multiple System. Stationary
Property.
In the case of free vibrations we have Qi = 0, Q 2 = Q, and
the solution of 15 (12) then follows exactly the same course
as in the particular examples already given. We assume
q 1 = A l cos(nt + e), q 2 = A 2 cos(nt + e), (1)
and obtain (c u  n z a u ) A, f (c 12  n 2 a l2 ) A 2 = 0,1
(c 21  ?i 2 (A 21 ) A l + (c sst  n^a^) A^ = 0.)
Eliminating the ratio A l :A 2) we obtain
 c u w 8 a n , C 12 /i 2 a 12
C 21 rc 2 a 21 , CJB 71*022 ~
where (it is to be noticed) the determinant is of the " sym
metrical " type. This equation gives the two admissible values
of n*. Adopting either of these we obtain a solution in which
the ratio of A l to A 2 is determined by either of the equations
(2). The mode of vibration thus ascertained involves therefore
two arbitrary constants, viz. the absolute value of (say) A 1} and
the initial phase e. The second root of (3) leads to another
solution of like character.
The extension of the method to the general case is obvious,
but it may be well to state the results formally. In any
conservative system of m degrees of freedom there are in
general m distinct " normal modes " of free vibration about
a configuration of stable equilibrium, the frequencies of which
are given by a symmetrical determinantal equation of the mih
order in n 2 , analogous to (3), and so depend solely on the con
stitution of the system. In each of these modes ihe t system
oscillates exactly as if it had only one degree of freedom, the
coordinates q^,q^, ... q m being in constant ratios to one another,
and the description of 7 therefore applies. The directions of
motion of the various particles and the relative amplitudes are
THEOKY OF VIBRATIONS 45
in any one mode determinate, though usually different for
different modes, the only arbitrary elements being the absolute
amplitude and the phaseconstant.
The equations of motion being necessarily linear, since
products and squares of the coordinates and their differential
coefficients with respect to the time are expressly excluded, it
follows that the different solutions may be superposed by
addition of the corresponding expressions. This has been
sufficiently illustrated in the preceding examples. By super
posing in this way the m normal modes, each with its arbitrary
amplitude and phase, we obtain a solution involving 2m
arbitrary constants, which is exactly the right number to
enable us to represent the effect of arbitrary initial values of
the coordinates q l} q^, ... q m and velocities q lt q z , ... q m . In
other words, the most general free motion of the system about
a configuration of stable equilibrium may be regarded as made
up of the m normal modes with suitable amplitudes and initial
phases. This principle dates from D. Bernoulli* (1741).
In particular cases it may happen that two (or more) of the
natural periods of the system coincide. There is then a corre
sponding degree of indeterminateness in the character of the
normal modes. The simplest example is furnished by the
spherical pendulum, or by a particle oscillating in a smooth
spherical bowl. The normal modes may then be taken to
correspond to any two horizontal directions through the position
of equilibrium. From the theoretical standpoint such coinci
dences may be regarded as accidental, since they are destroyed
by the slightest alteration in the constitution of the system
(e.g. if the bowl in the above illustration be in the slightest
degree ellipsoidal), but in practice they often lead to interesting
results. Cf. 53 below.
An important characteristic of the normal modes, first
pointed out by Lord Rayleigh in 1883, has still to be referred
* Daniel Bernoulli (17001782), one of the younger members of the
distinguished family of Swiss mathematicians. Professor of mathematics at
St Petersburg (172533), and of physics at Bale (175082). His chief work
was on hydrodynamics, on the theory of vibrating strings, and on the flexure
of elastic beams.
46 DYNAMICAL THEORY OF SOUND
to. If, by the introduction of frictionless constraints which do
no work, the system be restricted to vibrate in a mode only
slightly different from one of these, the period will be altered
only by a small quantity of the second order. In other words
the periods of the several normal modes are " stationary " when
compared with those of slightly different constrained modes.
Suppose, for instance, that the normal mode in question is such
that in it the coordinate q l alone varies. We have, then, in (2),
a 12 = 0, c 12 = 0, and the natural frequency is determined by
n* = c u /a 11 . If the constraint be expressed by q 2 = \qi, the
condition that the constraining forces shall do no work, viz.
ftfc + Q2fc = 0, or ft + XQ 2 = 0, leads to
(a 11 + X s toa)& + (c 11 + X s c B )g 1 = 0, ............ (4)
and the speed (p) is accordingly given by
When X is small, this differs from n 2 by a small quantity of
the second order. The proof, although limited to two degrees,
is easily generalized. Owing to our liberty of choice of the
coordinates, we can always arrange that q l shall be the only
coordinate which varies in the mode in question, and that
the constraint shall be expressed by a system of relations of
the type q 2 = \q l} q 3 = fj,q lf q 4 = vq lt ....
For an obvious illustration we may have recourse again
to the particle on a smooth surface. If the constrained path
be a vertical section through the lowest point, the period is
%wJ(Rlg\ where R is the radius of curvature of the section, and
it is known that R is a maximum or minimum for the principal
sections.
The equation (5) shews further that the constrained period is
(as in the particular case) intermediate between the two natural
periods ; this property can also be generalized.
It follows that even when it is not easy to ascertain the
precise character of a particular normal mode, a close approxi
mation to the frequency can often be obtained on the assumption
of an assumed type which we can judge on independent grounds
to be a fairly good representation of the true one. And in the
THEORY OF VIBRATIONS 47
case of the gravest natural mode the frequency thus obtained
will be an upper limit. Take, for instance, the case of three
equal particles attached at equal intervals to a tense string
( 14), and consider an assumed type of symmetrical vibration
in which x = z \y. The kinetic energy is then given by
2T = M(d? + y*+ &) = M(I + 2\*)f, ......... (6)
so that the inertiacoefficient is M(l + 2X 2 ). For the potential
energy we have
2 , (7)
a a
as is found by calculation of the work required to stretch the
string (as in 22), or otherwise. The coefficient of stability is
therefore P/a . (4\ 2  4\ 4 2). For the speed ( p) we then have
P 4X*4X + 2
This is stationary for X = + ^ \/2, and the corresponding speeds
are as in 14. In this case it was evident beforehand that the
assumed type would include the true natural modes of sym
metrical character.
It is unnecessary for the purposes of this book to discuss in
detail the theory of dissipation in a multiple system. The
general effect is the same as in 12 ; the free vibrations
gradually die out, but if the dissipative forces be relatively
small the periods are not sensibly affected.
17. Forced Oscillations of a Multiple System. Prin
ciple of Reciprocity.
The theory of forced oscillations is sufficiently illustrated if
in 15 (12) we assume that Qi varies as cospt, whilst (? 2 = 0.
The equations will be satisfied if we assume that q l and q t both
vary as cospt, provided
qi + (cvpdu) ? 2 = ft,
ft, I
0. J
These determine the (constant) ratios of q 1 and q 2 to Q l ; thus
48 DYNAMICAL THEOEY OF SOUND
where A (p 2 ) is the determinant on the lefthand side of 16
(3), with p 2 written for n 2 . The general conclusion is that when
a periodic force of simpleharmonic type acts on any part of
the system, every part will execute a simpleharmonic vibration
of the same period, with synchronism of phase, but the
amplitude will of course be different in different parts. When
the period of the forced vibration nearly coincides with that
of one of the free modes, an abnormal amplitude of forced
vibration will in general result, owing to the smallness of the
denominator in the formulae (2). For a complete account of
this matter we should have to take dissipative forces into
consideration, as in 12.
A remarkable theorem of reciprocity, first proved by Helmholtz
for aerial vibrations, and afterwards greatly extended by Lord
Rayleigh, follows from (2). If we imagine a second case of
forced vibration (distinguished by accents) in which Q/ =
whilst Q 2 ' varies as cos pt, we shall have
Comparing with (2), we see that
,:ft = 9,':ft'. ..................... (4)
The interpretation is most easily expressed when the "forces"
Q 1 and Q 2 ' are of the same character, e.g. both ordinary statical
forces, or both couples, in which case we may put Qi = Q 2 ', and
obtain q 2 = qi'. In words: The vibration of type 2 due to a
given periodic force of type 1 agrees in amplitude and phase
with the vibration of type 1 due to an equal force of type 2.
An example from the theory of strings will be found in 28.
The above proof is easily extended to the general case of
ra degrees of freedom.
18. Composition of SimpleHarmonic Vibrations in
Different Directions.
We recur to the subject of composition of simpleharmonic
vibrations which, though not so important as in Optics, claims a
little further attention. If in a freely vibrating system we fix
our attention on a particular particle, the directions in which it
THEORY OF VIBRATIONS 49
oscillates in the several normal modes will in general be different.
The superposition then takes place of course according to the
law of geometrical or vector addition.
It will suffice to consider the case of two degrees of freedom,
where we have independent simpleharmonic vibrations in the
directions corresponding to the Bs l} &s 2 of 15. The result is a
plane orbit, usually of a complicated character. For instance,
in the case of Blackburn's pendulum ( 14), we have
x = A cos (nj + ej), y = E cos (n z t + e 2 ), (1)
where x, y are rectangular coordinates. The orbit is here
contained within the rectangle bounded by the lines x A,
y = E. If n lt HZ are commensurable, the values of x, y and
x, y will recur after the lapse of an interval equal to the least
common multiple of the two periods, and the path will be
reentrant. The resulting figures, obtained in this and in other
ways, are associated with the name of Lissajous*, who has had
many followers in a region which is very attractive from the
experimental point of view.
The simplest case is that of rij = n^ If we eliminate t in
(1) we then obtain
s(e 1 e 2 ) + ^ = sm 2 (e 1  2 ) (2)
This represents an ellipse which, if the initial phases e lt e 2 coincide,
or differ by TT, degenerates into a straight line (Fig. 20). The
simplest mechanical illustration is furnished by the spherical
pendulum. When the relation is that of the octave (r^ 2n?)
we have a curve with two loops, which may degenerate into one
or other of two parabolic arcs (Fig. 21). The curves in these and
in other cases of commensurability are easily traced from the
formulae (1) with the help of tables. A simple geometrical
construction is indicated in Fig. 22, where the circumferences of
the auxiliary circles are divided into segments corresponding to
equal intervals of time in the two simpleharmonic motions
which are to be compounded. If we start at a corner of any
* J. A. Lissajous (182280). Professor of physics at the Lyce"e St Louis
1850 74; rector of the Academy of Chambe'ry 1874 5, and of Besancon
1875 9. His chief memoir, Etude aptique des mouvements vibratoires, was
published in 1873.
L. 4
50
DYNAMICAL THEORY OF SOUND
THEORY OF VIBRATIONS
51
one of the rectangles in the figure, and proceed diagonally, we
pass through a succession of points, equidistant in time, on a
curve of the system.
Fig. 22.
Another conception of these figures, also due to Lissajous,
may be mentioned. If we write 6 for nj, and adjust the origin
of t, the formulae (1) are equivalent, on the hypothesis of
commensurability, to
y
(# a),
.(3)
where p/q is a fraction in its lowest terms. These equations,
when combined with
z asinO, ........................ (4)
represent a curve of sines traced on the surface of the circular
cylinder
x * + 2 * = a * ........ ................ (5)
and going through its period p times in q successive circuits
of the cylinder. The Lissajous curve (3) is the orthogonal
projection of this curve on a plane (z = 0) through the axis of
42
52 DYNAMICAL THEORY OF SOUND
the cylinder. This is illustrated by Fig. 22, where the dotted
branch may be regarded as the projection of that part of the
sinecurve which lies on the rear half of the curved surface. A
change in the relative phase in (1) is equivalent to a change in
the angle a, and may be represented by a rotation of the cylinder
about its axis, of corresponding amount. This, again, may be
illustrated from Fig. 22 by starting the curve one step further to
the right or left. When the ratio of the periods is nearly, but
not exactly, that of two integers, the orbit gradually passes
through the various phases of the commensurable case, in a
recurring cycle*. Thus in the case of approximate unison, or
approximate octave, the cycle includes the phases shewn in
Fig. 21 or 22, followed by the same in reverse order. The same
result is obtained by a continuous rotation of Lissajous'
cylinder.
19. Transition to Continuous Systems.
The space which we have devoted to the study of dynamical
systems of finite freedom is justified by the consideration that
we here meet with principles, in their primitive and most easily
apprehended forms, which run through the whole of theoretical
acoustics. In the subsequent chapters we shall be concerned
with systems such as strings, bars, membranes, columns of air,
where the number of degrees of freedom is infinite. Mathematic
ally, it is sometimes possible to pass from one of these classes to
the other by a sort of limiting processes when D. Bernoulli (1732)
discussed the vibrations of a hanging chain as a limiting form
of the problem where a large number of equal and equidistant
particles are attached to a tense string whose own mass is
neglected. In any case, there can be no question that the
general principles referred to retain their validity. The main
qualification to be noticed is that the normal modes are now
infinite in number. It is usual to consider them as arranged
* In Lissajous' method the vibrations which are optically compounded are
those of two tuning forks. The figures obtained when the tones sounded by the
forks form any one of the simpler musical intervals give a beautiful verification
of the numerical relations referred to in 3. In the case of unison, when the
tuning is not quite exact, the cycle of changes synchronises with the beats
which are heard ; see 10.
THEORY OF VIBRATIONS 53
in ascending order of frequency ; the mode of slowest vibration
may still be called the " fundamental," and is generally the
most important.
Before leaving the general theory it may be desirable to
emphasize once more the importance of the simpleharmonic
type of vibration from the dynamical point of view. We have
seen that it is the characteristic type for a frictionless system of one
degree of freedom, or (more generally) for a system oscillating
as if it possessed only one degree, as in the case of the normal
modes. It is also the only type of imposed vibration which is
accurately reproduced, on a larger or smaller scale, in every
part of the system. If a force of perfectly arbitrary type act at
any point, the vibrations produced in other parts of the system
have as a rule no special resemblance to this or to one another;
it is only in the case of a periodic force following the simple
harmonic law of variation with the time that the induced
vibrations are exactly similar, and keep step with the force.
Moreover it is only in so far as the disturbing force is simple
harmonic, or contains simpleharmonic constituents, that it is
capable of generating a forced vibration of abnormal amplitude
when a critical frequency is approached. It is in these circum
stances that Helmholtz found the clue to his theory of audition,
to which we shall have to refer at a later stage.
20. On the Use of Imaginary Quantities.
The treatment of dynamical equations can often be greatly
simplified by the use of socalled "imaginaries." As we shall
occasionally have recourse to this procedure, it may be convenient
to explain briefly the principles on which it rests.
The reader will be familiar with the geometrical representa
tion of a "complex" quantity a+ ib, where a, b are real and i
stands for \/( I), by a vector drawn from the origin to the
point whose rectangular coordinates are (a, b), and with the
fact that addition of imaginaries corresponds to geometrical
addition (or composition) of the respective vectors. The
symbol a + ib when applied as a multiplying operator to any
vector denotes the same process by which the vector a + ib may
be supposed to have been derived from the vector 1, viz. it
54
DYNAMICAL THEORY OF SOUND
alters the length in a certain ratio r, and turns it through a
certain angle a. These quantities are defined by
rcosa = a, r sin a. = b, (1)
the quadrant in which a lies being determined by the sign
attributed to cos a or sin a by (1). We have then
a + ib = r (cos a I i sin a) (3)
Hence a symbol of the form cos a + i sin a denotes the opera
tion of turning a vector
through an angle a without
alteration of length; in par
ticular the symbol i denotes
the operation of turning
through a right angle in the
positive (counterclockwise)
direction.
The symbol
w = cos + i sin (4)
may be represented by a unit
vector OP drawn from in
the direction 6. If we regard Fi g . 23.
this as a function of 0, and if
w + Bw be represented by OP', the angle POP' will be equal
to BO. The vector PP' which represents Bw will therefore have
a length BO, and since it is turned through a right angle
relatively to OP, its symbol will be iBO.w. Hence
.(5)
It is easily shewn that the only solution of this equation
which fulfils the necessary condition that z = 1 for 6= 0, is
w = e, (6)
where e ie is to be taken as denned by the ordinary exponential
series. Thus
e* = cos0 + *sin0 (7)
We may add that the "additiontheorem" of the exponential
function can now be derived immediately from the geometrical
representation.
THEORY OF VIBRATIONS 55
It has been thought worth while to recapitulate these ele
mentary matters because they have interesting illustrations in
the present subject. Thus if x, y be rectangular coordinates,
and we write
z = x + iy, ........................... (8)
the equation z = Ce int , ........................... (9)
where C may of course be complex, expresses that the vector C
is turned in a time t through an angle nt in the positive direc
tion. It therefore represents uniform motion in a circle, with
angular velocity n, in the positive direction. The radius of this
circle is given by the "absolute value" of G, which is often
denoted by  G ; thus if C = A + iB, where A and B are real, we
have C \ = \I(A Z + B 2 ). In the same way the equation
z =G' e int ........................ (!0)
represents uniform motion in a circle, with angular velocity n,
in the negative (or clockwise) direction.
We come now to the application to linear differential
equations with constant coefficients. From our point of view
the simplest case is the equation
of 4. In order that every step of the work may admit of
interpretation, we associate with this the independent equation
0, ................. ....(12)
as in the theory of the spherical pendulum. The two may be
combined in the one equation
' +*. ..................... < 13 >
which may indeed be regarded as representing directly, without
the intermediary of (11) and (12), the law of acceleration in the
spherical pendulum and similar problems. To solve (13) we
assume z = Ce**, and we find that the equation is satisfied
provided X 2 f n 2 = 0, or X = in. Since different solutions can
be added, we obtain the form
int , .................. (14)
56 DYNAMICAL THEORY OF SOUND
with two complex arbitrary constants (7, G f . These can be
determined so as to identify z and z, at the instant t = 0, with
the vectors which represent the initial position and velocity of
the point (x, y). It appears from (14) that the most general
motion of a point subject to (13) may be obtained by the
superposition of two uniform circular motions in opposite direc
tions. The same problem (virtually) has been treated in 18,
where the path was found to be an ellipse. This resolution
of an "elliptic harmonic" vibration into two circular vibrations
in opposite directions has important applications in Optics.
The solution of the equation (11) may be derived from (14)
by taking the " real " part of both sides, i.e. by projecting the
motion on to the axis of x. Since <7, C' are of the forms
C=A+iB, C'=A' + iB', (15)
it might appear at first that the result would involve four
arbitrary constants. These occur, however, in such a way
that they are really equivalent only to two. Thus we find
x = (A+A')cosvt(BB')smnt (16)
The kinematical reason for this is that, as regards their
projections on a straight line, righthanded and lefthanded
circular motions are indistinguishable. An important practical
corollary follows. We should have obtained equal generality,
so far as the solution of (11) is concerned, if we had contented
ourselves with either solution of (13), for example
z=Ce int , (17)
and taken the real part
x A cosnt Bsmnt (18)
This conclusion is obviously not restricted to the particular
differential equation (11) with which we started. The use of
an adjunct equation such as (12) has only been resorted to
in order to remove the suspicion of anything that can truly
be called " imaginary " in the work. Such assistance can
always be invoked mentally, but it is as unnecessary as it
would be tedious always formally to introduce it. If in any
case of a linear differential equation between x and t, with
constant real coefficients, we seek for a solution of the type
x = Cte xt , the imaginary values (if any) of X will occur in
THEORY OF VIBRATIONS 57
conjugate pairs of the form m in, and we may assert that
the part of the solution corresponding to this pair of roots
will be given with sufficient generality if we make use of one
only of these, writing, for instance,
x=Ce (m + in}t , ..................... (19)
and taking the real part.
We may apply these considerations, for example, to the
equation
of resisted motion about an equilibrium position ( 11). If we
put x = Ce , we have
A. 2 + &X + /4 = ...................... (21)
Hence \ = \kiri, ..................... (22)
where TO' = VO"i& 2 ), ..................... (23)
provided k 2 < 4//,. On the above principle a sufficient solution
or, in real form,
x=e~* kt (Acosn'tBsmn't\ ......... (24)
which is equivalent to 11 (8).
The same method can be followed with regard to the
equation of forced oscillations, say
?SBS /<8jtf ............. (25)
Instead of this we take the equation
g + ** + / u.^ ................ (26)
the implied adjunct equation being of the type (25) with
fsinpt instead of /cos pt on the right hand. A particular
solution is
z=Ce ipt , ........................ (27)
provided (/z  p 2 + ikp) G =/. .................. (28)
fj pt
Hence * =  ^ sr ................... (29)
*
58 DYNAMICAL THEORY OF SOUND
If we put /jup* = Rcos a, kp = Rsina, (30)
this becomes z = ^e i(pt ~ a \ (31)
the real part of which is
x = ^cos(pta) (32)
This may be compared, for brevity, with the process of 12.
21. Historical note.
The theory of vibrations has a long and rather intricate
history, in which Pure Mathematics and Mechanics have
reacted on one another with great advantage to the progress
of both sciences. Various special problems of great interest
had been solved by the Bernoullis, Euler*, and other mathe
maticians, but it is to Lagrange  that we owe the general
theory of the small oscillations of a system of finite freedom
treated by means of generalized coordinates. The work of
Lagrange was purposely somewhat abstract in formj; the
full dynamical interpretation was reserved for Thomson and
Tait (Natural Philosophy, 1867), to whom we also owe the
now current terminology of the subject. The theory has
been very greatly extended by Lord Rayleigh, and systematic
ally applied to acoustics as well as other branches of physics,
in various writings, most of which (down to the year 1896)
are incorporated in his Theory of Sound .
* Leonhard Eoler, born at Bale 1707, died at St Petersburg 1783. He wrote
extensively on most branches of mathematics and mechanics, and fixed to
a great extent the notations now in use.
t Joseph Louis Lagrange, born at Turin 1736, died at Paris 1813, "the
greatest mathematician since the time of Newton."
J "On ne trouvera point de Figures dans cet Ouvrage. Les m^thodes que
j'y expose ne demandent ni constructions, ni raisonnemens ge'ome'triques ou
me'chaniques, mais seulement des operations alge"briques, assujeties a une
marche reguliere et uniforme." (Preface to the Mecanique Analytique, 1788.)
1st ed. London 1877, 2nd ed. London 1894 6. See also his Scientific
Papers, Cambridge 18991902.
CHAPTER II
STRINGS
22. Equation of Motion. Energy.
We proceed to the more or less detailed study of the
vibrations of various types of continuous systems. Amongst
these the first place must for many reasons be assigned to
the transverse vibrations of a uniform tense string. Historically,
this was the first problem of the kind to be treated theoretically.
The mathematical analysis is simple, and various points of the
general theory sketched in the preceding chapter receive
interesting illustrations, which are moreover easily verified
experimentally. Again, the sequence of the natural periods
of free vibration has the special "harmonic" relation which
has long been recognized as in some way essential to good
musical quality, although the true reason, which is ultimately a
matter of physiology, has only in recent times been investigated.
The mathematical theory has further suggested some remarkable
theorems, as to the resolution of a vibration of arbitrary type
into simpleharmonic constituents, which are of farreaching
significance. Finally it is to be noted that in the propagation
of a disturbance along a uniform string we have the first and
simplest type of wavemotion.
The string is supposed to be of uniform linedensity p,
and to be stretched with a tension P. The axis of x is taken
along the equilibrium position, and we denote by y the trans
verse deflection at the point x, at time t. It is assumed that
the gradient dy/dx of the curve formed by the string at any
instant is so small that the change of tension may be
60 DYNAMICAL THEOEY OF SOUND
neglected. Under these conditions the equation of motion
of an element Bx is
(1)
where ir denotes the inclination of the tangent line to the
axis of x. The righthand side is, in fact, the difference of
the tensions on the two ends of the element, when resolved in
the direction of y. In virtue of the assumption just made we
may write sin \r = tan i/r = dy/dx, so that (1) becomes
where c 2 =P/p ............................ (3)
It is easily seen that the constant c has the dimensions of
a velocity.
The kinetic energy of any portion of the string is given by
T=lpjfdx ..................... (4)
taken between the proper limits. The potential energy may
be calculated in two ways. In the first place we may imagine
the string to be brought from rest in its equilibrium position
to rest in any assigned form by means of lateral pressures
applied to it. For simplicity suppose that at any stage of the
process the ordinates all bear the same ratio (k) to their final
values y, so that the successive forms assumed by the string
differ only in amplitude. The force which must be applied to
an element Sx to balance the tensions on its ends is
(P sin ir) &c,
ox
where sin ^ is now to be equated to kdy/dx', and the displace
ment when k increases by &k is y 8k. The total work done on
this element is therefore
where the accents indicate differentiations with respect to x.
The potential energy is accordingly
v (5)
STKINGS 61
In the alternative method we calculate the work done in
stretching the string against the tension P. The increase in
length of an element &e is
approximately, so that
F=4P/y'=<fo. ..................... (6)
The formulae (5), (6) lead to identical results when applied to
the whole disturbed extent of the string. For by a partial
integration we have
Syy"dx = [yy r l+Sy'*dx, ............... (7)
where the first term refers to the limits. It vanishes at the
extremities of the disturbed portion, since y is there = 0.
23. Waves on an Unlimited String.
The solution of 22 (2) is
y=f(ctx) + F(ct + x) ................ (1)
where the functions /, F are arbitrary. It is easily verified
by differentiation that this formula does in fact satisfy the
differential equation, and we shall see presently that by means
of the two arbitrary functions which it contains we are able to
represent the effect of any given initial distribution of displace
ment (y) and velocity (y\ It was published by d'Alembert*
in 1747.
The two terms in (1) admit of simple interpretations.
Taking the first term alone, we see that so far as this is
concerned the value of y is unaltered when x and ct are
increased by equal amounts ; the displacement therefore which
exists at the instant t at the point x is found at a later instant
t 4 r in the position x 4 CT. Hence the equation
y=f(ctx) ..................... (2)
represents a waveform travelling unchanged with the velocity
c in the direction of ^positive. The equation
y = F(ct + x) ..................... (3)
represents in like manner a wave travelling with the same
velocity in the direction of ^negative. And it appears that
* J. le Kond d'Alembert (171783), encyclopaedist and mathematician ; he
made important contributions to dynamics and hydrodynamics.
62 DYNAMICAL THEORY OF SOUND
the most general free motion of the string may be regarded as
made up of two such wavesystems superposed.
The form of the expression \/(P/p) for the wavevelocity is
to be noticed. As in all analogous cases the wave velocity
appears as the square root of the ratio of two quantities, one
of which represents (in a general sense) the elasticity, and the
other the inertia, of the medium concerned.
A simple proof of the formula for the wavevelocity has
been given by Prof. Tait*. Imagine a string to be drawn with
constant velocity v through a smooth curved tube, the portions
outside the tube being straight and in the same line. Since
there is no tangential acceleration the tension P is uniform.
Also the resultant of the tensions on the ends of an element
Ss, at any point of the tube, will be a force PSs/R in the
direction of the normal, where R is the radius of curvature.
This will balance the "centrifugal force" p$s.v 2 /R (fv z =P/p.
Under this condition the tube may be abolished, since it exerts
no pressure, and we have a standing wave on a moving string.
If we now impress on everything a velocity v in the opposite
direction to the former, we have a wave progressing without
change of form, on a string which is otherwise at rest, with the
velocity \/(P/p). It will be noticed that this investigation does
not require the displacements to be small.
The motion of an unlimited string consequent on arbitrary
initial conditions
y=4>(x), y = + (x\ [* = 0], ............ (4)
may be deduced from (1), but it will be sufficient to write down
the result, viz.
rx+ct
+(z)dz. (5)
J xct
This may be immediately verified.
If the initial disturbance be restricted to a finite extent
of the string, the motion finally resolves itself into two
distinct waves travelling without change in opposite directions.
In these separate waves we have
* Encyc. Brit. 9th ed. Art. " Mechanics."
STRINGS 63
as is seen at once by considering two consecutive positions of the
waveform. Thus if in Fig. 24 the curves A, B represent the
positions at the instants t, t + St, we have PQ = c&t, RP = ySt,
RP/PQ = y', whence the former
of the relations (6). The same
thing follows of course from
differentiation of (2). Con
versely, it is easily seen from
(5), or otherwise, that if the
initial conditions be adjusted so that either of the relations (6)
is everywhere satisfied, a single progressive wave will result.
When the string is started with initial displacement, but
no initial velocity, the formula (5) reduces to
y = i((ac*)+(0 + cO} 00
The two component waveforms resemble the initial profile, but
are of half the height at corresponding points. It is easily seen
without analysis that this hypothesis satisfies the condition of
zero initial velocity.
It appears from (6) that in any case of a single progressive
wave the expressions (4) and (6) of 22 for the kinetic and
potential energies are equal. Lord Rayleigh has pointed out
that this very general characteristic of wave motion may be
inferred otherwise as follows. Imagine the wave as resulting
from an initial condition in which the string was at rest, and
the energy E therefore all potential, in the manner just
explained. The two derived waves have half the amplitude (at
corresponding points) of the original form, and the potential
energy of each is therefore J E. Since the total energy of each
wave must be ^ E, it follows that the kinetic energy of each
must be \E.
In mathematical investigations it is not unusual to find the
effect of dissipation represented by^jt he hypothesis that each
element of the string is resisted by a force proportional to its
velocity, so that the differential equation takes the form
dt 
As regards the theory of stringed instruments this particular
64 DYNAMICAL THEOKY OF SOUND
correction has no importance, the direct influence of the air
being quite insignificant; but the solution of (8) when k is
small is of some interest from the standpoint of wavetheory,
and may therefore find a place here. If the square of k be
neglected, the equation may be written
This is of the same form as 22 (2), and therefore
y = e~* kt f(ctx) + e* kt F(ct+a;) ....... (10)
This represents two wavesystems travelling in opposite
directions with velocity c; but there is now a gradual diminu
tion of amplitude in each case as time goes on, as is indicated
by the exponential factor. Again, since the functions are
arbitrary, we may replace f(ct x) and F(ct + x) by
e W*to f(ct x ) and e* k(t+xlc) F(ct+x),
respectively, so that the solution may also be written
y = e  lkxlc f(ctx) + e* lixlc F(ct + x) ....... (11)
This form is appropriate when a prescribed motion is maintained
at a given point of the string. Thus if the imposed condition
be that y = (f)(i) for x = 0, the waves propagated to the right
of the origin are given by
(12)
The exponential shews the decrease of amplitude as the waves
reach portions of the string further and further away from the
origin.
24. Reflection. Periodic Motion of a Finite String.
If a point of the string, say the origin 0, be fixed, we must
have y at this point for all values of t. Hence, in 23 (1),
f(ct) + F(ct) = 0, or F(z) = f(z).
The solution therefore takes the form
y=f(ctx)f(ct + x) ................ (1)
As applied for example to the portion of the string which
lies to the left of 0, this indicates the superposition of a direct
STRINGS 65
or "incident" wave represented by the first term, and a "re
flected" wave represented by the second. The amplitude of
the reflected wave is equal, at corresponding points, to that
of the incident wave, so that there is no alteration in the
energy, but the sign of y is reversed. It is otherwise obvious
that if on an unlimited string we start two waves which are
antisymmetrical with respect to 0, in opposite directions, the
 Fig. 25.
point of the string which is at will remain at rest, even if
it be free. Hence by the crossing of the waves the circum
stances of reflection at a fixed point are exactly represented.
It will be noticed that a lateral force is exerted on the fixed
point during the process of reflection.
In the case of a finite string whose ends are (say) at the
points x 0, x = I, we have the further condition that
f(ctl)f(ct + l) = (2)
for all values of t. If we write z for ct I, this becomes
/(*)/(* + SJ) (3)
so that f(z) is a periodic function, its values recurring when
ever z increases by 2/. It follows that the motion of the string
is periodic with respect to t, the period 2l/c being the time
which a wave would take to travel twice the length. It is
otherwise evident that a disturbance starting from any point
P of the string, in either direction, will after two successive
reflections at the ends pass P again, in the same direction as
at first, with its original amplitude and sign.
L. 5
66 DYNAMICAL THEORY OF SOUND
When the initial data are of displacement only, i.e. with
zero initial velocity, the successive forms assumed by the string
in the course of a period can be obtained by a graphical con
struction. We suppose the initial form y </> (x), where <f> (as) is
originally defined only for values of x ranging from to I, to be
continued indefinitely both ways, subject to the conditions
4>(_#) = </>(tf), (Z + aj) = (Zaj) (4)
If we imagine curves of the type thus obtained to travel
both ways with velocity c, and if we take at each instant the
arithmetic mean of the ordinates, in accordance with 23 (7), it
is evident that the varying form thus obtained will represent
Fig. 26.
a possible motion on an unlimited string, in which the points
x = 0, x = I, x = + 21, . . . remain at rest. The portion between
x = and x = I will therefore satisfy all the conditions of the
question. The process is illustrated in the annexed Fig. 26 ;
the initial form here consists of two straight pieces meeting at
an angle, and the result after an interval l/8c is ascertained.
In this way we might trace (after Young) the successive
forms assumed by a string excited by " plucking," one point of
the string being pressed aside out of its equilibrium position,
and then released from rest, but the actual construction can in
such a case be greatly simplified. It is easily seen that the
form of the string at any instant consists in general of three
portions; the outer portions have the same gradients as the
two pieces into which the string was initially divided, whilst
the gradient of the middle portion is the arithmetic mean of
STRINGS
67
these, account being taken of sign. The line of this middle
portion moves parallel to itself, with constant velocity, back
wards and forwards between the two corners of the parallelogram
of which the initial form constitutes two adjacent sides.
Fig. 27.
In the annexed Fig. 27, which corresponds with Fig. 26, the
plucking is supposed to take place at a distance of onefourth
the length from one end, and the phases shewn follow one
another at intervals of onesixteenth of a complete period, the
successive forms being APB, AQ&B, AQ^B, AQ 3 R 3 B, AQ 4 R 4 B,
and so on. It is evident on inspection of the figure that any
point of a plucked string moves backwards and forwards with
constant velocity between two extreme positions, in which it
rests alternately during (in general) unequal intervals. The
spacetime diagrams of the middle point, and of the point
plucked, under the conditions of Fig. 27, are given in Fig. 28.
Fig. 28.
52
68 DYNAMICAL THEOKY OF SOUND
In the latter case one of the intervals of rest vanishes*.
It is of course with the vibrations of a finite string that
we are chiefly concerned in acoustics. The string is usually
stretched with considerable tension between the two points which
limit the vibrating portion. At one at least of these points the
string passes over a bridge resting on a soundingboard, whose
function it is to communicate the vibrations to the surrounding
air. The direct action of the string in generating airwaves
is quite insignificant, but by the alternating pressure on the
bridge the whole area of the soundingboard is set into forced
vibration. This implies of course a certain reaction on the
string itself, which is however, in the first approximation,
usually negligible, for the reason given in 4.
For experimental purposes an arrangement called a " mono
chord " is used. The soundingboard here forms the upper face
of a rectangular " resonance chamber." The distance between
the bridges can be varied and measured, and the tension, being
produced by a weight attached to one end of the wire, which
passes over a smooth pulley, can be regarded, at all events
approximately, as known. For purposes of comparison one or
more additional wires may be stretched alongside the former,
their tension being adjusted, as in the pianoforte, by means of
pegs at the extremities.
25. Normal Modes of Finite String. Harmonics.
The preceding investigations have been given on account of
their historical importance, and for the sake of the analogies with
other types of wavemotion which we shall meet with later.
From the purely acoustical point of view they are however of
secondary interest. The ear knows nothing of the particular
geometrical forms assumed by the string, and is concerned
solely with the frequencies and intensities of the simple
harmonic constituents into which the vibration can be resolved.
* The theoretical vibrationforms have been verified experimentally by
KrigarMenzel and Kaps, Wied. Ann., vol. L., 1893, so far as the initial stages
of the motion are concerned. After a few vibrations the form is seen to be
undergoing a gradual change. This is attributed to a slight yielding of the
supports of the string, in consequence of which the normal frequencies are not
exactly commensurable, and the resulting motion therefore not accurately
periodic. The construction in Fig. 27 is also due to these writers.
STRINGS
69
To ascertain the normal modes of vibration of a finite string
we may have recourse to the general procedure explained in
Chapter I. In any such mode y will vary as a simpleharmonic
function of the time, say cos (nt + e). This makes y = ri*y,
and the equation (2) of 22 therefore assumes the form
.(i)
(2)
The solution of this, exhibiting the timefactor, is
/ . nx T. . nx\ f . ^
y = I A cos  1 B sin 1 cos (nt f e). ...
\ c c /
The fixed ends of the string being at x = 0, # = I, we must
have A=Q, sin (nl/c) = 0, whence
nl/7rc = l, 2,3, ...................... (3)
This gives the admissible values of n. In any one normal mode
we have, therefore,
n . STTX /STTCt \ ,..
y = C 8 sm cos( + 6j, ............... (4)
where 5 is an integer, and the amplitude C g and initial phase
s are arbitrary. The gravest, or fundamental mode, which
determines the pitch of note sounded, corresponds to s = l.
The string then oscillates in the form of the curve of sines
between the two extreme positions shewn in the upper part of
Fig. 29. The frequency is
70 DYNAMICAL THEOKY OF SOUND
and so varies inversely as the length and as the square root of
the linedensity, and directly as the square root of the tension.
These statements, which were formulated as the result of
experiment long before the mathematical theory had been
developed, are known as Mersenne's laws*. The determination
of absolute pitch by the formula (5) does not admit of very
great accuracy owing to the difficulty in measuring the tension,
which is apt (owing to friction) to be slightly different on the
two sides of a bridge.
The principles that the frequency diminishes with increase
of length and with increase of linedensity have a familiar
illustration in the pianoforte, where longer and intrinsically
heavier strings are used for the graver notes. If the relation
of pitch were adjusted by length alone the strings corresponding
to the lower notes would have to be at least 100 times as long
as those belonging to the highest. In order to secure a suffi
ciently low pitch within practical limits of length, and with
a sufficient degree of tension, the string is loaded with a coil of
wire wrapped closely round it. This has the effect of increasing
the inertia without seriously impairing the flexibility, which is
an essential point. The influence of tension, again, is illustrated
in the process of tuning, which consists in tightening up the
wires when these have stretched, or the pegs have yielded, so
that the instrument has fallen in pitch, or become " flat."
In the next normal mode after the fundamental the middle
point x \l is at rest (Fig. 29). And in the 5th mode, whose
frequency is by (3) s times that of the fundamental, there are
s 1 internal points of rest, or " nodes," in addition to the
ends. Midway between these we have the points of maximum
amplitude, or " loops." Each segment into which the string is
divided by the nodes vibrates as in the fundamental mode of a
string of 1/sth the length.
As already stated ( 2) the sequence of simple vibrations
with frequencies proportional to the natural numbers 1,2,3,...,
which we here meet with, has important properties, musically
* M. Mersenne (15881648), a Franciscan friar, was a schoolfellow and
lifelong friend of Descartes, and maintained an extensive correspondence with
him and other men of science of the day.
STRINGS 71
and physiologically. Its occurrence in vibrating systems is of
course quite exceptional. Even in the present case, if the
string deviate appreciably from uniformity or from perfect
flexibility, the above scale of frequencies is at once departed
from *.
We were led in 16 to the conclusion, on physical grounds,
that in any system of finite extent the effect of the most
general initial conditions consistent with its constitution may
be obtained by superposition of the several normal modes, with
suitable amplitudes and phaseconstants. We infer that the
most general motion of a finite string can be represented by
the formula
/?x
............ (6)
provided the constants C 8 , e g be properly determined, the
summation S extending over all integral values of s. An
equivalent form is
STTCt n . S7TCt\ . STTX /h _ x
,cosj +4amjJ smp, ...... (7)
where A 8 = C 8 cose 8) B 8 =C 8 s'm s ............. (8)
If the string start from rest in a given position at the
instant = the coefficients B 8 will vanish; if it be started
with given velocities from the equilibrium position (y 0)
the coefficients A 8 will vanish.
Since the value of every term in (6) or (7) recurs whenever
t is increased by 2/c, the vibration is essentially periodic, as
already proved in  24. In all other respects the motion of the
string when started in an arbitrary manner is, from the present
point of view, of a complex character, being made up of an
endless series of simpleharmonic vibrations. The resulting
note is accordingly made up of a series of pure tones, consisting
(in general) of a fundamental, its octave, twelfth, double octave,
and so on.
It is not altogether easy to excite a string in such a way
* The fact that a particular sequence of notes, musically related to one
another, is associated with lengths of string proportional to the quantities
1 i> i> i> was known to the Greeks, and was the origin of the name
"harmonic" as applied to the numerical series.
72 DYNAMICAL THEOEY OF SOUND
that the resulting motion shall be strictly simpleharmonic,
and the sensation accordingly that of a pure tone. But, as
will be shewn more fully in 39, it is possible to suppress
all the tones below any assigned rank (s) by checking the
vibration at a node of the 5th mode, as, for instance, by
contact with a camelhair pencil. The remaining nodal points
of this constituent are then points of rest, whilst halfway
between them there is vigorous vibration. The experiment,
which is very striking, is easily made with the monochord.
The energy in any normal mode is easily calculated. We
find
...... (9)
...(10)
The coefficients are equal, in virtue of 22 (3), and the total
energy in this mode is
T+v
It is further easily proved that the whole energy of the
string is the sum of the energies corresponding to the various
normal modes, viz.
T+V^S.*Of^*f(4*+Bf) ....... (12)
This is a general property of the normal modes of a vibrating
system. The proof, in the present case, depends on the fact
that
[ l . STTX . STT3C ..''* /, ox
I sin j sm = dx = 0, ............ (13)
JO * '
if s, s be any two unequal integers. See 32 (4).
26. String excited by Plucking, or by Impact.
The relative amplitudes of the various modes is of course
a matter of importance, as on it the quality of the note
depends ( 2). Usually a string is excited in one of three
ways, viz. by plucking (as in the harp, zither, &c.) f by striking
with a hammer (pianoforte), or by bowing (violin, violon
cello, &c.).
STRINGS 73
If the string be pulled aside through a small space , at
a distance a from the end a? = 0, and then be released, the
values of the coefficients in 25 (7) are found to be
/ ,~ . srra . STHZ? STTC
whence y , y,  x 2  sin = sin = cos  * . . . .(2)
7r 2 a(/ a) ,s* I I I
The mode of calculation will be explained in the next chapter
(see 36). We notice that the harmonic of order 8 will be
altogether absent if sin (sirajl) 0, i.e. if the point of plucking
be at one of its nodes; this was remarked by Young (1841).
Thus if the string be plucked at the centre, all the harmonics
of even order will be absent. The formula (1) combined with
25 (12) shews that, apart from a trigonometrical factor which
lies between and 1, the intensities of the successive harmonics
will vary as 1/s 2 . The higher harmonics are therefore relatively
feebly represented in the actual vibration of the string.
The effect of the impact of a hammer depends on the
manner and duration of the contact, and is more difficult to
estimate. The question is indeed, strictly, one of forced
vibrations ( 28); but in the somewhat fictitious case where
the duration is so small that the impact has ceased before
the disturbance (travelling with the velocity c) has had time
to spread over any appreciable fraction of the length, we
may treat the problem as one of free motion with given initial
velocity concentrated on a short length. The result is
where a is the distance from the origin to the point struck,
and //. represents the total momentum communicated by the
impact. Hence
2u _., 1 . Sira . STTX . Sirct
y =  2  sin y sin = sin = .......... (4)
TTpC Sill
As in the previous problem, the sih mode is absent if the
origin be at one of its nodes. Apart from the trigonometrical
factor on which this circumstance depends, the intensities of
74 DYNAMICAL THEOBY OF SOUND
the successive modes are, according to 25 (12), now of the
same order of magnitude. The unreal character of the pre
ceding hypothesis betrays itself in this result ; but we may at
all events infer that in the case of a very brief impact the
higher harmonics are relatively much more in evidence than
in the former problem.
In reality the impact, even in the case of a metallic
hammer, is far from instantaneous, the time of contact, though
very short as measured by ordinary standards, being at all
events comparable with the period of vibration of the string*.
The effect of an impulse of finite duration has been calculated
by Helmholtz, to whom most of the present theory is due, on
the supposition that the pressure begins at the instant t = 0,
and lasts for a time T, during which it rises from zero to a
maximum and falls to zero again, according to the law sin (TT/T).
A somewhat simpler result is obtained if we imagine the law
of pressure to be
where /* represents the timeintegral of the force from t = oo
to t + oo . This law, whose graphical representation has the
form of the curve in Fig. 14, p. 33, has the defect that there is
no definite instant of beginning or ending, but as the true law
is in any case unknown, it may serve for purposes of illustration.
The interval of time during which the force is sensible is
comparable with r, and can be made as narrow as we please
by diminishing T. The details of the calculation will more
conveniently find a place in the next chapter ( 38). The
result is
(6)
When T is infinitesimal this agrees with (3). In other cases
the intensities of the higher harmonics vary as e ~ 8irCT ' , if we
omit the trigonometrical factor.
Although the pressure is thus rendered less abrupt as
regards its variation with the time, it is still assumed to be
* Kaufmann, Wied. Ann., vol. LIV. (1896).
STKIJJGS 75
concentrated at a point. If we were to imagine it distributed
continuously over a short length of the string this would
further increase the relative weight of the lower harmonics
(see 38).
According to a general principle, which is here exemplified,
and which will be further referred to in the next chapter,
the higher harmonics are excited in greater relative intensity,
the more abrupt the character of the originating disturbance.
From a musical point of view the harmonics after about the
sixth are to be discouraged, since they come sufficiently
near to one another in the scale to be mutually discordant.
In the pianoforte the hammers are covered with layers of softer
material, so that the variation of pressure during the impact
is rendered more gradual.
The point at which the blow is delivered is also a matter
of importance. To obtain a note of rich musical quality the
lower harmonics should be present in considerable force, and
the middle regions of the string are on this account to be
avoided. On the other hand, the harmonics of higher order
than the sixth are prejudicial, as already stated. Both re
quirements are met by fixing the striking point at a distance
of about oneseventh of the length from one end. The partial
tones which have nodes at or near this point will then not
be excited at all, or only with comparatively feeble intensity.
27. Vibrations of a Violin String.
The theory of the vibrations of a string when excited by
bowing is somewhat difficult, but the main features have been
elucidated by Helmholtz. Since the pitch is found to be that
natural to the string, the vibrations are to be regarded as in
a sense " free," the function of the bow being to maintain the
motion by supplying energy to make up for the losses by
dissipation. In the case of the violin &c., where the strings
are of light material and pass over a bridge resting on a very
sensitive surface (of the resonance cavity), these losses may
be relatively considerable. The mode of action of the bow
appears to be that it drags the string with it for a time by
friction, until at length the latter springs back ; after a further
76 DYNAMICAL THEOEY OF SOUND
interval the string is carried forward again, and so on*, the
complete cycle taking place in the period of vibration.
In order to obtain data for mathematical analysis Helmholtz
began by an experimental study of the character of the vibration
at various points. The device was an optical one, of the kind
employed by Lissajous ( 18), by which the rectilinear vibration
of the point examined is compounded with an independent
vibration at right angles, whose period is commensurable, or
nearly so, with that of the string. A microscope whose axis is
horizontal is directed to the point to be studied, the string
itself being vertical. The eyepiece of the microscope is fixed,
but the objective is carried by one of the prongs of a tuning
fork and vibrates in a vertical direction. When the fork alone
vibrates the image of a bright point on the string is drawn
out into a vertical line; when the string alone vibrates the
appearance is that of a horizontal line. When both vibrations
coexist the result would be a closed curve if the periods were
exactly commensurable. For example, if the period of the fork
were exactly commensurable with that of the string, and if the
vibration of the point examined were simpleharmonic, the
result would be one of the corresponding series of Lissajous
figures ( 18); whilst if the relation between the periods were
inexact, the curve would pass in succession through the various
phases of the series. In the actual circumstances the forms of
the curves are modified, and it is possible from the result to
make inferences as to the true nature of the vibration studied.
Fig. 30.
The interpretation is facilitated by the ideal representation
of the successive phases as orthogonal projections of a curve
traced on a revolving cylinder. It was found that the space
* In order that work may be done it is necessary to suppose that the
frictional force is greater in the first stage than in the second. This is
consistent with the known law that friction of (relative) rest is greater than
friction of motion. The remark is due to Lord Rayleigh.
STRINGS 77
time diagram of the point of the string
under examination has the simple form
shewn in Fig. 30. The nature of the
modification in the Lissajous figures may
be illustrated in the case of unison
between fork and string. If the portion
of the broken line in Fig. 30 which lies
between A and B be wrapped round a
cylinder whose circumference is equal to
A B y its projections on planes through
the axis will include such forms as are
here shewn (Fig. 31)*.
The period of vibration of the point
examined is made up of two intervals,
usually of unequal duration, during
which the point moves backwards and
forwards, respectively, with constant but
(in general) unequal velocities. The
ratio of the two intervals is further
ascertained to be equal to that of the
two segments into which the string is
divided by the point. These results have
been confirmed by subsequent observers
who have obtained the spacetime dia
gram in a more direct mannerf. In
order that they may come out clearly
some precautions are necessary. Some
thing depends on the skill with which
the bow is used, and apparently on the
quality of the instrument. In order, also,
that the diagram should be free from
minor irregularities the bow should be
* In the actual experiments of Helmholtz the
frequency of the string was four times that of the
fork. The circumference of the cylinder in the
above mode of representation then includes four
periods of the zigzag line in Fig. 30.
t ErigarMenzel and Raps, Wied. Ann., vol.
XLIV. (1891).
78 DYNAMICAL THEOKY OF SOUND
applied at a node of one of the harmonics, and the point
observed should be at another node of the same.
Except at the two instants in each period when the velocity
suddenly changes, the acceleration of the point (P) examined is
zero. It follows from 22 (2) that the curvature of the string
in the neighbourhood of P vanishes, and that the form of the
string at any instant is accordingly made up of straight pieces.
Fig. 32.
It appears that all the conditions of the problem can be satisfied
if we assume that the form is always that of two such pieces
meeting at a variable point Q. In Fig. 32 let AB (= I) be the
undisturbed position of the string, and let a (= AN) and /3
(= NQ) be the coordinates of Q referred to A as origin and AB
as axis of abscissae. The equations of the two portions of the
string are
yi = M> y* = P(ix)l(i*), ......... (i)
and the difference of the velocities near Q on the two sides
is accordingly
In the time St a length d8t of the string is traversed by the
point Q, so that a mass pd&t has its velocity increased by the
above amount. This is the effect of the transverse force
where P is the tension, acting for the time St. Equating the
change of momentum to the impulse of the force we find
* ...................... (4)
STRINGS 79
The point of discontinuity Q (of the gradient) must therefore
travel right or left with the velocity c.
Let us suppose that Q starts from A at the instant t = 0,
and that & is at first positive. The observations of Helmholtz
shew that the velocity at a point x, viz.
is during an interval x/c constant, whence
= (7a(Za), ..................... (6)
no additive constant being admissible, since ft must vanish with
a. This is the equation of a parabolic arc passing through A, B.
The conditions of the problem are therefore all fulfilled if we
imagine Q to travel backwards and forwards along two such
arcs, with velocity c, in the manner indicated in Fig. 32. In
terms of the maximum displacement j3 we have G = 4/8 /Z 2 ,
and the equations of the two portions of the string at any
instant are therefore
y,^(i), ^<l*> ...... (7)
It only remains to resolve this motion into its simple
harmonic constituents. The details of the calculation are
given in 37. The result is
STTCt
where the summation embraces all integral values of s. Com
paring with 25 (7) we have
A = 0, B.3 ................... (9)
These results, and indeed the whole investigation, take no
account of the position of the point to which the bow is applied.
It is plain, however, that the position of the bow must have
some influence on the character of the vibration; and it is
found in fact that those normal modes are absent which have a
node at the point in question. It is for this reason that the
somewhat idealized vibrationform which is adopted as a basis
of calculation is only obtained in its purity at corresponding
nodes.
80 DYNAMICAL THEORY OF SOUND
28. Forced Vibrations of a String.
The simplest case of forced vibration is where a given
simpleharmonic motion
y = 0coa(pt + a) .................. (1)
is imposed at a point (x = a). The portions of the string on the
two sides of this point are to be treated separately. The
results are
. px
sin
cos (pt fa) [0 < x < a] ,
sm
sn
for these satisfy the general differential equation 22 (2), they
make y 1 = for x = 0, and y z = for x = I, and they agree with
(1) when x = a. The amplitude of y l or y 2 becomes very great,
owing to the smallness of the denominator, whenever pa/c or
p(l a)/c is nearly equal to a multiple of TT, i.e. when the
imposed period 2ir/p approximates to a natural period of a
string of length a or I a, respectively. To obtain a practical
result in such cases we should have to take account of dissipative
forces.
The case is illustrated by pressing the stem of a vibrating
tuning fork on a piano string. The sound swells out powerfully
whenever the portion of the string between the point of contact
and either end has a natural mode in unison with the fork.
This plan is recommended by Helmholtz as a means of producing
pure tones, since the higher modes of the fork, not being
harmonic with the fundamental, are not reinforced.
When a transverse force of amount Y per unit length acts
on the string, the equation (2) of 22 is replaced by
In general Y will be a function both of x and t.
The case of a periodic force F cos (pt + a) concentrated on an
infinitely short length of the wire at x a may be deduced from
STRINGS 81
the formulae (2). The value of ft in terms of F is found from
the consideration that the force must just balance the pull of
the string on this point, i.e.
F C os(pt + ct)=P yi 'Py 2 ' ............... (4)
for x = a. This leads to
. px . p(l a)
sm sin 
sn
c c
The formula for y z differs only in that the letters x and a are
interchanged ; we have here an instance of the reciprocal theorem
of 17, according to which the vibration at a point x due to a
periodic force at a must be the same as the vibration at the
point a due to an equal force (of the same period) at x.
The amplitude becomes as a rule great when sm(pljc) is
small, i.e. when the imposed period approaches a natural period
of the whole string. An indeterminate case occurs when
sin (pale) and sin (pile) = simultaneously, the point x = a
being then a node.
29. Qualifications to the Theory of Strings.
We have in 26, 27 considered the relative amplitudes of
the different harmonics when a string is excited in various ways,
but we must not assume that the corresponding relative inten
sities are accurately reproduced in the resulting soundwaves,
which are started indirectly through the sounding board.
If we neglect the reaction on the string, which may for a
considerable number of vibrations be insensible, we may regard
the string as exerting on each bridge a force proportional to
the value of dy/dx there*, as given by the respective formula.
The differentiation introduces a factor s in the coefficient of
the 5th harmonic, and so increases the importance of the
higher modes. On the other hand, the amplitude of vibra
tion of the sounding board due to a simpleharmonic force
of given amplitude, will vary somewhat with the frequency,
* Thus in the case of the plucked string it appears from Fig. 27 that
the pressure on each end alternates between two constant values of opposite
sign.
L. 6
82 DYNAMICAL THEOKY OF SOUND
on the general principle illustrated in 9*. This is probably
to the relative advantage of the lower modes. The effect of
yielding of the bridges in modifying the natural frequencies
of the string has been discussed by Rayleighf; it is probably
in practice very slight.
Another cause which must be mentioned as affecting our
results to some extent is the imperfect flexibility of the string,
or wire. In the case of the higher normal modes the segments
into which the string is divided may be so short that flexural
couples come into play, and tend to raise the frequency by
increasing the potential energy of a given deformation. This
will be referred to later ( 50). A further point is that the
abrupt forms postulated in the theory of plucked or bowed
strings are not exactly realized, and that such investigations as
those of 26, 27 are to be viewed as approximations, which are
however quite adequate so far as the determination of the ampli
tudes of the graver and more important harmonics is concerned.
30. Vibrations of a Loaded String.
We conclude this chapter with the discussion of one or two
problems which, besides being of some interest in themselves,
may serve to remind us again that the harmonic scale of fre
quencies is after all an exceptional phenomenon, even in the
case of strings.
Take first the case of a string, otherwise uniform, loaded
with a mass M at its centre. It is obvious that those normal
modes of the unloaded string which have a node at this point
are unaffected. Leaving these on one side, we consider only
those vibrations in which there is at every instant complete
symmetry with regard to the centre. If the lateral displacement
of M be {3 cos (nt + e), we have, for the first half of the string,
(1)
* Some interesting experiments bearing on these questions have been made
by Barton and Garrett, Phil. Mag. (6), vol. x., 1905. See also Barton, Text
Book of Sound, London, 1908, 361.
t Theory of Sound, 135.
STRINGS
The equation of motion of M is
(2)
where after the differentiations we must suppose x = I. This
gives
nl nl I
2c 2c = b' (3)
where b is written for M/p, i.e. b is the length of string whose
mass would be equal to that of the attached particle. The
frequencies are therefore determined by
where x 1) # 2 , x z , ... are the roots of the transcendental equation
x tan x = l/b (5)
Equations more or less of this character occur in many branches
of mathematical physics, and can often be solved approximately
by graphical construction. Thus in the present instance if we
trace the curves
b
2/ = cot#, y= I ., (6)
the abscissae give the roots. If 6 be relatively small these
fall a little short of JTT, TT, TT, ..., respectively, and the
bl
Fig. 33.
65
84 DYNAMICAL THEOKY OF SOUND
frequencies are therefore slightly lower than in the symmetric
modes of an unloaded string. As b increases the frequencies
all diminish, the physical reason being of course the increased
inertia. Finally, when M is very large compared with the mass
pi of the string, l/b is small, and the lowest root of (5) is given
approximately by a? = 1/b, whence
n =
in agreement with 6 (4).
31. Hanging Chain.
The contrast with previous continuous systems is still more
marked in the case of the small oscillations of a uniform chain
hanging vertically from its upper end, which is fixed. This
has no immediate acoustic importance, but it is interesting
historically*, and is, from the standpoint of general theory,
instructive in various ways.
We take the origin at the equilibrium position of the free
end. The tension at a height x above this point will be
P = gpx, the vertical motion being neglected as of the second
order. Hence if y denote the horizontal deflection, we have
or
a*
3 / dy
I*K
(*y\
rat/ 1
To ascertain the normal modes we assume that y varies as
cos (nt + e), and obtain
i( x d /] + n y = Q ................... (3)
dx \ dxj g *
This can be integrated by a series, but the solution assumes
a somewhat neater form if we introduce a new independent
variable in place of as. The wavevelocity on a string having
a uniform tension equal to that which obtains at the point x
would be *J(Plp) or *J(gx). Hence if r denotes the time which
* It appears to have been the first instance in which the various normal
modes of a continuous system "were determined, viz. by D. Bernoulli (1732).
The Bessel's Function also makes its first appearance in this connection.
STKINGS 85
a point moving always with this local wavevelocity would take
to travel from the lower end to the point x, we have
In terms of r as independent variable the equation (3) becomes
0 ... ............... (5)
For the present purpose we do not require the complete
solution, but only that solution which remains finite when
T = 0. This is
where G is arbitrary, as may be verified by actual differentia
tion, and substitution in (5). The function defined by the
series in brackets presents itself in many physical problems;
it is called the "Bessel's Function of Zero Order," and is
denoted by J Q (nr)*. Hence, inserting the timefactor,
y = CJ (riT)cos(nt + 6) (7)
The value of T corresponding to the upper end (x I) is
T 1 = 2V(%), (8)
and the condition that this end should be fixed gives
Jo (nil) = (9)
This determines the admissible values of n. The first few
roots are given by
WTl /7r = 7655, 17571, 27546,..., (10)
where the numbers tend to the form s J, s being integral. In
the modes after the first, the values of T corresponding to the
lower roots give the nodes. Thus in the second mode there is
a node at the point T/T, = 7655/17571, or a;/l = T*lT l * = '190.
The gravest period is 2?r/w = 5'225 \J(llg\ whereas the period
of oscillation of a rigid bar of the same length is 5130 ^(Ijg).
The comparison verifies a general principle referred to in 16,
* Elaborate numerical tables of the Bessel's Functions, calculated by
Meissel and others, are given by Gray and Mathews, Treatise onBessel Functions,
London, 1895. A convenient abridgment is included in Dale's FiveFigure Tables
of Mathematical Functions, London, 1903.
86
DYNAMICAL THEORY OF SOUND
according to which any constraint has the effect of quickening the
gravest oscillation. The first two modes are shewn (on different
scales) in Fig. 31, the two nodal points representing the point of
suspension in the two cases.
Fig. 34.
CHAPTER III
FOURIER'S THEOREM
32. The SineSeries.
The study of the transverse vibrations of strings has
already suggested a remarkable theorem of pure mathematics,
to which some further attention must now be given. The
theory of the normal modes has led us ( 25) to the conclusion
that the free motion of a string of length I, started in any
arbitrary manner, can be expressed by a series of the form
STTCt n . S7TCt\ . STTX
8 cosj + 5 f sm Jsinj, ...... (1)
where s = 1, 2, 3, ..., provided the constants A 8) B 8 be properly
determined. In particular if the string be supposed to start
from rest at the instant t = in the arbitrary form y = f(x), it
should be possible to determine the coefficients A s so that
(2)
for values of x ranging from x = to x = I. This is a particular
case of "Fourier's Theorem*." Since I is at our disposal we
may conveniently replace it (for general purposes) by TT, and
the statement then is that an arbitrary function f(x) can be
expressed, for values of x ranging from to TT, in the form
f(x) = A l sin x + A 2 sin 2# f . . . + A 8 sin sx + ....... (3)
* J. B. J. Fourier (17681830). The history of the theorem is closely
interwoven with that of the theory of strings, and of the theory of heat
conduction. Fourier's own researches are expounded in his Theorie de la
Chateur, Paris, 1822. An outline of the history is given in Prof. Carslaw's
book cited on p. 96. The subject is treated most fully by H. Burkbardt in his
report entitled Entwickelungen nach oscillirenden Funktionen..., Leipzig, 1908.
88 DYNAMICAL THEOKY OF SOUND
The reasoning by which we have been led to this result is
of a physical rather than a mathematical nature, and we have
moreover not referred to the restrictions which physical con
siderations alone would impose on the character of the arbitrary
function f(x). Leaving such points for the moment, and as
suming the theorem provisionally, we proceed to the deter
mination of the coefficients. If we multiply both sides of (3)
by sin sx, and integrate from x to x = ir y we get on the right
hand a series whose general term is
A r I sin rx sin sx dx
Jo
= %A r l {cos (r s) x cos (r 4 s) x} dx. ...(4)
J o
When the integers r, s are unequal this vanishes, since each
cosine goes through its cycle of values, positive and negative,
once or oftener within the range of integration. But when
r = s, the first cosine is replaced by unity, and the result is
Hence
2 f v
A g = \ f(x)$msxdx ................ (5)
The process may be illustrated by a few examples. Take,
first, the case of
/(*) = (*), ..................... (6)
which is represented by an arc of a parabola. We find, after
a series of partial integrations,
2 C w 4
A 8 =\ x (TT x) sin sx dx = (1 COSSTT). ...(7)
7T J o TrS"*
This is equal to or 8/?rs 3 , according as s is even or odd. The
theorem therefore becomes
x (TT x)  ( sin x + sin 3# +  sin 5x + . . . ) . (8)
7T \ O O /
If we put x J TT in this we obtain the formula
32"" 3 3 5 3 '"' '
which is known on other grounds to be correct. The equality
in (8) may also be tested graphically. It is found that the
discrepancy between the graph of x (TT x) and that of the
FOURIER'S THEOREM
function represented by the first three terms on the right hand
is so slight that it would be barely perceptible on a scale
suited to the pages of this book.
In the next example the graph of f(x) consists of two straight
lines through the points x = 0, x = TT, respectively, meeting at
an angle at the point x = a. If we assume the ordinate at the
latter point to be unity, we have
/(*) = #/ [0 <*<], j
/(#) = (TT  at) / (IT a) [*<*<*].)
We find, after some reductions,
2 f a 2 f 71 "
A 8 xsinsxdx\  .  ^1 (TT x) sin sx dx
Tra.'o 7r(7ra)J a
1
sinsa. ...(11)
a(7ra) V
Fig. 35.
90 DYNAMICAL THEOKY OF SOUND
Thus
2 / 1
f(x) = j I sin a sin x 4 ~ 2 sin 2a sin 2x
+ i 2 sin3asin3#+...y ...(12)
As a check on this result we may put a = JTT, x \ir\ this
gives
which is known to be right. This example is of interest in
connection with the theory of the plucked string ( 26, 36).
Fig. 35 shews the graph of f(x) together with that of the
function represented by the first eight terms of the series on
the right hand of equation (12), in the case of a = f?r. The
fourth and eighth terms contribute nothing to the result in
this case, since they correspond to modes having a node at the
point plucked.
Again, let /(#) = TT a? ...................... (14)
2 f 77 2
We find A s = \ (TT #)sins#c& =  ............. (15)
TTJO s
The theorem therefore asserts that
TT  x = 2 (sin x + \ sin 2# f sin 3# +...). . . .(16)
If we put x \ TT, we obtain
which is Euler's formula for the quadrature of the circle.
The formula (16) also verifies obviously for #=TT; but if we
put x we see that there is some limitation to its validity.
The necessary modification is stated in 34. The series is
moreover much more slowly convergent than in the preceding
case; this is illustrated by Fig. 36, which shews the graph
of TT x together with that of the function represented by
the first eight terms of the series. For any value of x other
than we can obtain an approximation as close as we please,
provided we take a sufficient number of terms, but the smaller
the value of x the greater will be the number of terms required
to attain a prescribed standard.
FOUKIER'S THEOREM
91
Fig. 36.
The preceding illustrations, with the diagrams, afford at
all events a presumption in favour of the theorem in question,
but shew at the same time that it is subject to some restric
tions. The theorem admits of independent mathematical proof
under certain conditions as to the nature of the "arbitrary"
function f(x). We shall, however, not enter upon this, but
shall content ourselves with the following formal statement :
If we form the sum of the first m terms of the series (2),
and write
fm (#) = AI sin x + A z sin 2x + ... + A m sin mx, (18)
where
A 8 = I f(x)sinsxdx, ............ (19)
it may be shewn that, for any assigned value of x in the range
from to TT, the sum f m (x) will tend with increasing m to the
limit f(x), provided the function f(x) is continuous throughout
the above range, has only a finite number of maxima and
minima, and vanishes for x = and x = IT.
It will be noticed that the conditions here postulated are
92 DYNAMICAL THEOKY OF SOUND
fulfilled as a matter of course by any function which it is
natural to assume as representing the initial form, or the
initial velocity, of a tense string. We also see that the
difficulty met with in the case of (16) can be accounted for
by the fact that the function does not vanish for x = 0.
An extension of the statement to meet such cases will be
given presently ( 34).
33. The Cosine Series.
The theory of the longitudinal vibrations of rods, or of
columns of air, leads, in addition, to a similar theorem relating
to the expansion of an arbitrary function in a series of cosines.
The formal statement is now as follows :
If we write
fm (#) = A + A l cos x + AZ cos 2% + . . . + A m cos mx, (1)
where
whilst for s >
2
= l f(x)co$sxdx, (3)
it may be shewn that as m increases the sum f m (x) will tend to
the limit /(#), provided f(x) is continuous throughout the
range from to TT, and has at most a finite number of
maxima and minima. There is now no restriction as to the
values of /(O) and /(TT).
If the determination of the effect of special initial conditions
in a longitudinally vibrating bar which is free at both ends
were as interesting a problem as it is in the case of strings
we should have recourse to the cosineseries.
34. Complete Form of Fourier's Theorem. Discon
tinuities.
The question arises as to what is represented by the
sineseries or the cosineseries, supposed continued to infinity,
when x lies outside the limits and TT. The answer is supplied
by the consideration that both series are periodic functions
of x } the period being 2?r, whilst the former is an odd, the
FOURIER'S THEOREM 93
latter an even function of x*. This is illustrated by the
annexed graphical representations, in which f(x) is given
primarily only for the range TT, but is continued in one
case as an odd and in the other as an even periodic function
of x. It will be noticed that in the former case the stipulation
that f(x) is to vanish for x = and x = TT is necessary if
discontinuities are to be avoided.
Since any function f(x) given arbitrarily for values of x
ranging (say) from TT to TT can be resolved into the sum of an
even and an odd function, viz.
/(*) = i !/(*)+/(*)} +i {/(*)/< *)). <i)
we derive the more general theorem that the sum
f m (x) A + A! cos x + A z cos 2# + . . . + A m cos mx
4 B! sin x + B 2 sin 2x + ... + B m sin mx, ...(2)
where
* An "odd" function is one which is simply reversed in sign with x,
like x 3 or sinx. An "even" function is one which is unaltered in value
when the sign of x is changed, like x or cos x.
94 DYNAMICAL THEORY OF SOUND
whilst for s >
If* 1 1 f" \
A s =  I {/(#) +/( a))} cos sxdx = I f(x) cos s#cfo,
TTJo " TTJ  n I ^
i t w i r* 1
B 8 =\ \f(x) f( x)} sin s#cfo? =  I /(#) sin s#d#,
TTJo * TTJ _ ff
tends with increasing m to the limit /(a?), provided f(x) is
continuous from X TT to # = TT and has at most a finite
number of maxima and minima, and provided also that
/( TT) =/(TT). For values of x outside this range the limit
represents, under these conditions, a periodic function of
period 2?r. This is the complete form of Fourier's Theorem,
and includes the others as special cases.
We should be led directly, on physical grounds, to this form
of the theorem if we were to investigate the "longitudinal"
vibrations of the column of air in a reentrant circular tube.
We have so far supposed the function /(#) to be continuous,
as well as finite, even when continued beyond the original
range as a periodic function. But the theorems hold, with
a modification to be stated immediately, even if f(x) have
a finite number of isolated discontinuities. In such a case
the series f m (x) still converges, with increasing m, to the value
of /(#), except at the points of discontinuity. But if a be
a point where f(x) abruptly changes its value, the sum f m (a)
tends to the limit
where f(a 0) and/(a + 0) represent the values of f(x) at
infinitesimal distances to the left and right, respectively, of the
point a. For example, in the case of the sineseries 32 (3),
if f(x) does not vanish when x = or when X = TT, there is
discontinuity at these points in the periodic function, and
the series / m (0), for example, has the limit 0, which is
the arithmetic mean of the values of the continued function
on the two sides of the point #=0. This is illustrated in
Fig. 36.
35. Law of Convergence of Coefficients.
It remains to say something as to the law of decrease of
the successive terms. It is evident at once that under the
FOURIER'S THEOREM 95
litions laid down the values of the coefficients A 8 and B 8
must ultimately diminish indefinitely as s increases, owing to
the more and more rapid fluctuation in sign of cos so; and
sin sx, and the consequent more complete cancelling of the
various elements in the definite integrals of 32 (5) and
33 (3).
More definite results have been formulated by Stokes.
The following statement must be understood to refer to the
function as continued in the manner above explained; and
care is necessary, in particular cases, to see whether discon
tinuities of f(x) or its derivatives are introduced at the
terminal points of the various segments:
If f(x) have (in a period) a finite number of isolated
discontinuities, the coefficients converge ultimately towards
zero like the members of the sequence
1, 2> 3> 4> !
This is exemplified by 32 (16) and Fig. 36.
If f(x) is everywhere continuous, whilst its first derivative
f'(x) has a finite number of isolated discontinuities, the con
vergence is ultimately that of the sequence
ill 1
' 2 2 ' 3 2 ' 4 2 ' *
This is illustrated by 32 (12) and Fig. 35.
If f(x\ f(x) are continuous, whilst /"(#) is discontinuous
at isolated points, the sequence of comparison is
l > %3> ^3> #>
as in the case of 32 (8). And, generally, if f(x) and its
derivatives up to the order n 1 inclusive are continuous, whilst
the nth derivative has (in a period) a finite number of isolated
discontinuities, the convergency is ultimately as
1 1 1
The nature of the proof, which is simple, may be briefly
96 DYNAMICAL THEORY OF SOUND
indicated for the case of the sineseries. We have, by a partial
integration,
2 f 77
A s I f(x) sin sx dx
1 F2 "1 2 C n
=  /O)coss# + I f(x)w$sxdx, ...(4)
where the integrated term is to be calculated separately for
each of the segments lying between the points of discontinuity
of /(#), if any, which occur in the range extending from x = to
# = TT inclusively. For example, if as in 32 (14) the only
discontinuity is at x 0, its value is 2/(0)/S7r. In any case
there is, for all values of s, an upper limit to the coefficient of
l/s in the first part of (4); we denote this limit by M. The
definite integral in the second term tends ultimately to
zero, as s increases, owing to the fluctuations in sign of cos sx.
Hence A 8 is ultimately comparable with M/s. If there is no
discontinuity of /(#), even at the points x = 0, x TT, the first
term in the above value of A 8 vanishes, and continuing the
integration we find
A s =  \ f O)sin sx\  f fjT<*) sin ** dx. . . .(5)
r \_Tr J s "^J Q
In the first part, regard must be had to the discontinuities of
/'(#), if any. Denoting by M the upper limit of the coefficient
of l/s 2 , we see that A s is ultimately comparable with M/s*, the
second term in (5) vanishing in comparison, by the principle
of fluctuation. The further course of the argument is now
sufficiently apparent.
36. Physical Approximation. Case of Plucked String.
It has been thought worth while to state Fourier's theorem
with some care, although we do not enter into the details of the
mathematical proof, which is necessarily somewhat intricate,
owing to the various restrictions which are involved*.
From a physical point of view the matter may be dealt with,
and perhaps adequately, in a much simpler manner. To explain
this, it is best to take a definite problem, for instance that of
* The most recent English treatise on the subject is that of Prof.
H. S. Carslaw, Fourier's Series and Integrals, London, 1906.
FOURIER'S THEOREM 97
the plucked string ( 26). The differential equation, and the
terminal conditions, are satisfied by the finite series
. TTX Tret . fax
y=A l sm j cos j 4 A z sin j cos . \
. mirct
+ A m sm j cos j , ...(1)
each term of which represents a normal mode of vibration.
This makes the initial velocity zero, whilst the initial form is
. TTX . fax . rrnrx
y = A 1 sm j 4 A 9 mn ; + ... + A m smj . ...(2)
II V
The question we now have to consider is, how to determine the
coefficients A lt A z , ... A m so that (2) may represent, as closely
as may be, a prescribed initial form
y/() ......................... (3)
There are many reasons why, from the physical point of
view, we may be content with an approximate solution of the
problem. Leaving aside such questions as the resistance of the
air and the yielding of the supports at the ends of the string, we
have still to remember that in substituting a mathematical line
of matter, capable only of exerting tension, we have considerably
overidealized the circumstances. In the higher normal modes,
at all events, the imperfect flexibility, and the uncertainty as to
the true nature of the terminal conditions, render this representa
tion somewhat inadequate, so that a solution which professes to
determine these modes accurately is open to the criticism that
it attempts too much. Again, the assumed initial form in
which two straight pieces meet at a point, is one which can
only be approximately realized ; if we go too far in this direction
we should produce a permanent bend, or kink, in an actual
wire.
The determination of the coefficients in the finite series
(2) will depend on the kind of approximation aimed at. For
example, we might divide the length of the string into ra + 1
equal parts, and choose the coefficients 'so that the functions (2)
and (3) should be equal at the m dividing points. The curves
represented by these equations will then intersect in m points in
addition to the ends. Another method is to make the sum of
L. 7
98 DYNAMICAL THEORY OF SOUND
the squares of the errors involved in the substitution of (2) for
(3) as small as possible. Thus if, for shortness, we replace I by
TT, we have to choose the coefficients so as to make the integral
/(#) ( A l sin x + AZ sin 2# 4 . . . + A m sin mx)} z dx (4)
a minimum. If we differentiate with respect to A 8 we get .
)(A 1 smx+A z sin2x + ...+A m smmx)}smsxdx=(), (5)
or, by 32 (4),
2 /**'
A 8 = \ f(x)smsxdx ................ (6)
TTjQ
Hence this method of least squares, applied to the expression
(2) consisting of a finite number of terms, gives precisely the
values of the coefficients which were obtained by Fourier's
process*. Each coefficient is determined by itself, and the
effect of adding more terms to (2) is to improve the approxima
tion, without affecting the values of the coefficients already
found. If we revert to general units, the formula (6) is
replaced by
2 l / s S7TX
S= ~
In the case of the plucked string, the form to which we
endeavour to approximate is
y = Pxla [0<x<a], y = # (I x)l(l a) [a<x<l\ (8)
The result is obtained at once from 32 (11) if we write jrx/l
for x, and therefore irajl for a, I for TT, and introduce the factor fi.
Thus
, 2/9Z 2 . STTO,
as stated in 26 (1). The nature of the approximation is
illustrated in Fig. 35.
37. Application to Violin String.
To apply the method to the problem of the violin string
( 27), we take as origin of t the instant when the point Q in
* This theorem is due to A. Toepler (1876).
FOURIER'S THEOREM
Fig. 32 starts from A to describe the upper parabolic arc. At
this instant we have y = 0, everywhere, whilst
(i)
We therefore begin with the finite series
. 7TX . TTCt .
y A l smj sin j + A^ sm j sm j I ...
. mirx . mirct /sn
4 A in sm sm . ...(2)
This satisfies the differential equation, and makes y = for t = 0.
It only remains so to determine the constants that the series
. D . D . mirx /ON
1 sm j + J5 2 sm j~ + . . . + .O TO sin y , ...(3)
where B t = A s , ..................... (4)
may represent the initial distribution (1) of velocity, as nearly
as possible. The determination of B 8 has virtually been made
in 32 (15). With the necessary modifications of notation we
find
as stated in 27. The graph of the initial velocity, and the
approximation attained by taking the first eight terms of the
series (3), are shewn in Fig. 36.
It will be noticed that our approximation has even an
advantage over the result obtained by carrying the series to
infinity. In the latter case, the initial velocity, as represented
by (1), is discontinuous when x = 0, being zero for # = 0, but
equal to 4y9 c/Z when x differs ever so little from 0. The
idealized representation of the motion in 27 is in this
respect imperfect ; the parabola in Fig. 32 should be slightly
modified so as to touch the line AB at its extremities.
38. String Excited by Impact.
As a final example we take the case of a string started by
an impact, as in 26. We begin with the case of a force
72
100 DYNAMICAL THEORY OF SOUND
distributed continuously in space and in time, the differential
equation being
as in 28. Suppose, in the first place, that
y * /*\ rjrx /\  ^Tne .c *+\ m7nr . .
 =/j (t) sm j +/, (0 sin j + . . . +f m (t) sin y , (2)
the coefficients being known functions of t. The equation (1)
is then satisfied by
7TO) . ?? mirx
 + 77 2 sm + ...+7; m sin  , ...(3)
provided ***/. ................... (4)
The solution of this equation has been given in 8. If we
assume that r) s = 0, 17, = for t = oo , and that/^ (t) is sensible
only for a finite range of t, the resulting value of rj 8 is
57TC^ f 00 ,. ,.. . STTCt
If as a particular case we put
we have ^ = e^^sin, ........... (7)
STTC L
by 8 (18). As a function of t, Y now follows the special law
indicated by the last factor in (6), at every point of the string,
but we have not yet made any special assumption as to the
distribution of the force over the length. Its timeintegral is
given by
1 T 00
P J 
TT7 , ~ . ~ . ^ . mirx
Ydt= Ojsmy + O 2 sm 7  + ... + C m sm y . (8)
We may now seek to determine the values of the coefficients
so that this expression may be sensible only in the neighbour
hood of the point x a. We assume, then, that
PJ oo
FOURIER'S THEOREM " *101
where <f> (x) vanishes except between the limits a e and a + e,
say. The formula (7) of 36 then gives
2 [ l . , N . STTX , 2 f a+t , , . . S7TX , , .
C g = j <f> O) sm r <te = 7 # W sm ~T~ * ( 10 )
& ./ ^ *J at
If be small, the series thus obtained converges at first very
slowly, and a great many terms might have to be taken to
secure a reasonable approximation. In the terms of lower
order we have
2 . STra f a+e . . , 2u . sira
0,jgUiyJ <0r)<fo=^smy, ...(11)
/a+c
<f>(x)dx, (12)
a
i.e. /Lt represents the total impulse. The corresponding term in
the value (3) of y is
2/4 1 .. . STra . STTX . sirct /1Q ,
. e*"* 11 . sm j . sm = . sm = . . . .(13)
PTTC s III
But however small e may be, so long as it is not evanescent,
the value of G 8 given by (10) will ultimately tend to zero with
increasing 5, owing to the more and more complete mutual
destruction of positive and negative elements under the integral
sign. This shews the effect of diffusing the impulse over a
small but finite portion of the string.
The case of an instantaneous local impulse is obtained by
putting r=0(cf. 26).
39. General Theory of Normal Functions. Har
monic Analysis.
The space which has been devoted to Fourier's theorem is
no more than is warranted by its importance, especially in rela
tion to the theory of strings, but it is well to remember that
from the standpoint of the theory of vibrations the theorem is only
one out of an infinite number which can be based on the same
kind of physical considerations. Every vibratory system has its
own series of "normal functions," as they are called, which
express the configuration of the system in the various normal
modes. In the case of a uniform string, or of the doublyopen
organ pipe, these functions happen to have he simple form
102 DYNAMICAL THEORY OF SOUND
sin (STTX/I), or cos (sirx/l), respectively. More complicated forms
will be met with when we come to the theory of transverse
vibrations of bars, and to that of membranes ; and even in the
cases just mentioned the simplicity of type would at once
disappear if the uniformity of linedensity, or of crosssection,
respectively, were departed from. In some problems, indeed,
of considerable interest, e.g. that of the vibrations of a rectangular
plate, the precise form of the functions has still to be discovered.
But in any case the functions theoretically exist ; and on the
principle that any free motion whatever of the system consists
of some combination or other of the various normal modes, it
must be possible to express any arbitrary initial state, and
therefore any arbitrary function of position in the system, by a
series of normal functions. Such preeminence as attaches to
Fourier's theorem is, from the present point of view, due merely
to the fact that in it we have the simplest exemplification of
this principle in the case of a continuous system, and the one
where the physical induction has been most fully corroborated
by independent mathematical proof. It may also be added that
it is only in the case of strings that the calculation of the
effect of particular initial conditions has any great interest.
There is however another point of view from which the
resolution of a function into a series of sines or cosines of the
variable is of peculiar importance, viz. when we are dealing with
functions of the time. The dynamical reason for this has
already been dwelt upon ( 19).
When a function f(t) is known to be periodic, of period T,
its resolution by Fourier's theorem is
/v,x r
j (t) = A + A l cos  f A z cos  h A a cos  [ . . .
f B l sin   + J5 2 sin h B 3 sin ^^ + . . . , (1)
T T T
where A =*( T f(t)dt, (2)
whilst for s > 0,
/A p nH fl4 T) / f(t\ oin /O\
^t^ oos ac, xj,  i y ^c^ sin . v* 5 /
FOURIER'S THEOREM 103
This is of course merely a restatement of the theorem of
34, with the necessary changes of notation. It will be noticed
that A represents the mean value of the function.
We have already been led to formulae of the type (1) as
expressing the motion at any assigned point of a freely vibrat
ing string, the period r being equal to 2l/c. Another important
acoustical application is to the analysis of a periodic current of
air, as in the siren, or the reedstops of an organ ( 90). Again,
in the case of electromagnetically driven tuning forks, a periodic
current can powerfully excite, not only a fork in unison with
itself, but also others whose natural frequencies are respectively,
twice, three times, ... as great. This is due to the fact that the
disturbing force is of the type (1), the selective resonance taking
place according to the principles of 9.
Various mechanical contrivances for resolving a given
periodic curve into its simpleharmonic constituents, and
conversely for compounding a number of independent sine
and cosinecurves whose periods are as 1, , J, ..., have been
devised by Lord Kelvin and others. From the standpoint of
the present subject the most remarkable of these is perhaps
the machine constructed by Prof. A. A. Michelson, in which
provision is made for as many as 80 constituents*.
It is hardly necessary to say explicitly that the resolution
of a periodic function of t in the form (1) can only be effected
in one way, the values of the coefficients as given by (2) and (3)
being determinate. In particular, a series of the above type
cannot vanish for all values of t unless its coefficients severally
vanish. Thus in a freely vibrating string, if the motion at any
given point x be prevented, as by touching with a camelhair
pencil, the coefficients of cos(s7rct/l) and sin(s7rct/l) in the
general formula (7) of 25 must be zero, i.e. we must have
. ' . STTX . . STTX _
A 8 smj = Q, B 8 smj = (4)
for all values of s. Unless x be commensurable with I this
requires that A a = Q, B a = Q, and the whole string will be
* Phil. Mag. (5), vol. XLV. (1898) ; this paper contains a number of most
interesting examples of results obtained. The construction is also explained in
his book On Light Waves and their Uses, Chicago, 1903.
104 DYNAMICAL THEORY OF SOUND
reduced to rest. In the excepted case the conditions (4) are
satisfied independently of the values of A s and B s whenever
sin (STTX/I) = 0, i.e. those normal modes remain unaffected
which have a node at the point touched.
A question arises as to the effect of nonperiodic forces on
a dynamical system. For the reason already so often insisted
upon, it is convenient, whenever possible, to resolve the force
into a series of terms of the type
A cos pt + B sin pt ................... (5)
Each element A cospt or Bsmpt then produces throughout
the system its own effect, viz. an oscillation of the same type
and period, the configuration and its amplitude depending on
the speed p. In some cases the resolution presents itself quite
naturally, as for example in the theory of the tides. The
disturbing effect of the sun and moon, when account is taken
of their varying declinations, and of the inequalities in their
orbital motions, can be sufficiently represented by a series of
terms of the type (5). It follows that the tideheight at any
particular place must be expressed by a series of like character,
in which the values of p are known. The theoretical determina
tion of the coefficients is out of the question for the actual
ocean, with its variable depth and irregular boundaries, but
their values can be inferred a posteriori with more or less
accuracy from a comparison of the formula with observation,
and when once ascertained can be used for prediction*.
When the disturbing force is perfectly arbitrary in character,
without any obvious periodic elements, the question is more
complicated. There is a form of Fourier's theorem specially
appropriate to this case, but its application is usually difficult,
and it is simpler to have recourse, as in 38, to the formula (12)
of 8. The objection that this implies a knowledge of the
whole previous history of the system is met if we introduce
the consideration of damping, which is in reality always present.
The equation
.................. (6)
* For an elementary account of the matter see Sir G. H. Darwin, The Tides,
London, 1898.
FOURIER'S THEOREM 105
may be written
provided n' 2 = n 2 JA; 2 ...................... (8)
Hence, by the formula referred to,
cos n'tdt
~~ **'
e~ **' cos n't je* kt f() sin n'tdt. . . .(9)
If we have x 0, x for t oo , the limits of integration
are oo and t. For instance, the value of x when t =
becomes
x = l,r e* kt f(i)sinn'tdt. (10)
Owing to the presence of the exponential factor it is only for
a certain range of negative values of t that the function under
the integral sign has as a rule an appreciable value. In other
words, the effects of the action of the force prior to a certain
antecedent epoch have practically disappeared.
CHAPTER IV
BARS
40. Elementary Theory of Elasticity. Strains.
We require a few elementary notions from the theory of
elasticity. As regards the purely geometrical study of de
formations, or "strains," it is usual to begin with the con
sideration of a body in a state of uniform or "homogeneous"
strain. This is sufficiently defined by the property that any
two lines in the substance which were originally straight and
parallel remain straight and parallel, although their direction
relative to other lines in the substance is usually altered.
A parallelogram therefore remains a parallelogram, and it
easily follows that the lengths of all finite parallel straight
lines are altered in the same ratio ; this ratio will however
usually be different for different directions in the substance.
It can be shewn that there are three mutually perpendicular
directions in the substance which remain mutually perpen
dicular after the deformation; these are called the "principal
axes" of the strain. It is unnecessary, for our purposes, to
give the formal proof, as the existence of such axes will be
in evidence in such simple cases as we shall meet with.
It follows from this theorem that any originally spherical
portion of the substance is deformed into an ellipsoid whose
axes are in the directions of the principal strainaxes.
If PQ, P'Q' denote any straight line in the substance,
before and after the strain, the ratio of the increase of length
to the original length, viz.
PQ
BARS
107
is called the "extension"; it will in general be different for
different directions of PQ. In the theory of elastic solids, e is
always a very minute fraction. We denote by e 1} e 2 , 6 8 , the
extensions in the directions of the principal axes.
The ratio of the increase of volume to the original volume
is called the " dilatation." Denoting it by A, and considering
the change of volume of a cubical block whose edges are along
the principal axes, we have
or A = 6! +6 2 + 6 3 , (2)
the products of small quantities being neglected.
There are two special types of homogeneous strain which
require notice. First, suppose 6! = 6 2 = 6 3 , = 6, say. Any origin
ally'^pherical portion of the sub
stance then remains spherical, and
the extension is therefore the same
in all directions. The strain may
accordingly be described as one of
uniform extension; and we note
that = JA.
Again, take the case of e l = 6 2 ,
= 6, say, whilst 6 8 = 0, and therefore
A = 0. A square whose diagonals
AOC, BOD are parallel to the axes
1, 2 is converted into a rhombus A'B'C'D ', and since
to the first order, the lengths of the sides are unaltered. Also
...(3)
tan A' OB' = it? = tan (J TT + e),
1e
so that the angles of the rhombus are \ IT 2e. Another view
of this state of strain is obtained if we imagine the rhombus
A'B'C'D' to be moved in its own plane so that A'B' coincides
with AB. This is legitimate, since no displacement of the
body as a whole affects the question. We then see that any
two planes of the substance parallel to AB and the axis 3 are
displaced relatively to one another, without change of mutual
108
DYNAMICAL THEORY OF SOUND
distance, by an amount proportional to this distance. This
kind of strain is called a " shear," from the fact that it is of the
type which tends to be set up by the action of the two edges
of a pair of shears. The " amount " (77) of the shear is specified
by the relative displacement per unit of mutual distance, i.e. by
the ratio DD'/AD, or 2e, in the first part of Fig. 39. Again,
by moving B'C' into coincidence with BC, we might prove that
the strain is also equivalent to a shearing of planes parallel to
BC and the axis 3, in the direction of BC. This is shewn in
the second half of the figure.
D D'
C C A A
D D
A B B C
Fig. 39.
41. Stresses.
The name " stress " is applied to the mutual action which
is exerted across any ideal surface S drawn in a body, between
the portions of matter immediately adjacent to $ on either side.
We are here concerned with molecular actions sensible only over
an exceedingly short range, so that the portions of matter in
question are confined to two exceedingly thin strata, whose
common boundary is S. The resultant force on a small portion
of either stratum may then be taken to be ultimately pro
portional to its area, and the intensity of the stress is
accordingly specified by the force per unit area. This force
may be of the nature either of a push or a pull, and may be
normal or oblique, or even tangential to the area.
For simplicity, it is usual to begin with the notion of a
state of uniform or " homogeneous " stress, i.e. the stress over
any plane is assumed to be uniform, and the same in direction
and intensity for any two parallel planes. It will of course in
general be different for planes drawn in different directions.
It may be shewn that there are then three mutually perpendic
BARS
109
ular sets of planes such that across each of these the stress is
in the direction of the normal ; but for a reason already indicated
we need not stop to prove this theorem. The planes in question
are called the " principal planes " of the stress, and the corre
sponding stressintensities are called the "principal stresses."
They are usually reckoned as positive when of the nature of
tensions; we denote them by Pi,p 2 ,p 3 .
There are certain special types of stress to be noticed.
First let j9 1 =j9 2 = PS The stress across every plane is then in
the direction of the normal, and of uniform intensity, as in
hydrostatics.
Next, let^! = pt,=p, say, whilst p 3 = 0. Consider a unit
cube whose faces are parallel to the principal planes. The
portion included between the faces represented by AB, DA, in
the figure, and the diagonal plane represented by BD, is in
equilibrium under three forces. Two of these forces are
parallel and proportional to DA and A B, viz. the forces on AB
and DA, respectively. The third force is therefore along and
Fig. 40.
Fig. 41.
proportional to BD\ and its amount (CT) per unit area is p.
A similar result holds with respect to the diagonal plane A C.
A cube four of whose faces are parallel to these diagonal planes
is in equilibrium under tangential stresses, in the manner
shewn. This type is accordingly called a "shearing stress."
Its amount (CT) is specified by the tangential force per unit
area on the planes in question.
110 DYNAMICAL THEOEY OF SOUND
In general the states of strain and stress in a body are not
uniform, or " homogeneous," but vary continuously from point
to point; but the above notions are still applicable to the
infinitely small elements into which the body may be conceived
to be divided.
42. Elastic Constants. Potential Energy of Defor
mation.
The theory of strains is a matter of pure geometry ; that of
stresses one of pure statics. When we come to connect the two
we require some physical assumption. The usual hypothesis,
known as " Hooke's law," * is that the stresses are linear
functions of the strains. This law ceases to hold, even approxi
mately, when the strains exceed certain values called the
"elastic limits"; but for the purposes of acoustics it may be
adopted without hesitation, on account of the excessive minute
ness of the varying strains with which we are concerne^f.
In an "isotropic" substance, i.e. one in which there is no
distinction of properties between one direction and another, the
principal axes of strain must evidently coincide with those of
stress. Moreover the principal stress p^ must involve the
principal strains e 2 , e 3 symmetrically, and so on. The most
general assumption consistent with this requirement, and with
Hooke's law, is of the form
(1)
where X, /j, are constants depending on the nature of the
material]:. It will be noticed that e l5 e a , e 3 are pure ratios and
that the dimensions of \, p are therefore those of stress, or force
* Robert Hooke (1635 1703), professor of geometry at Gresham College
16651703.
t If Hooke's law were sensibly departed from, the frequencies of the normal
modes of a vibrating bar would no longer be independent of the amplitude.
Since the ear is very sensitive to variation of pitch, this would easily be detected.
This remark is due to Stokes.
J There is a great diversity of notation as regards these constants. The
above symbols are those introduced by G. Lame" (1795 1870), professor of
physics at the Ecole poly technique 1832 44.
BAKS, 111
divided by area, viz. [ML^T^] if [M], [L\ [T] denote the units
of mass, length, and time.
There are various combinations of the constants X, p which
are important in physics, as well as in technical mechanics. In
a uniform dilatation we have p 1 = p 2 = p 3 ( = p, say), l = e 2 = e s
(= JA), whence
p = (X + ^)A ...................... (2)
Hence if we write * = \ + ft, ........................ (3)
K will denote the " volumeelasticity " or " cubical elasticity " of
the substance, i.e. the ratio of the uniform stress to the dilatation
which it involves.
Next suppose that l =  e 2 = e, e 3 = 0, and therefore
Pi = p^=p, p 3 = 0, which is the case of a pure shear,
involving a shearing stress. According to the investigations
of 40, 41 the shearing stress is vr = p, and the shear is
rj = 2e. Hence, from (1),
*r = /"7, ........................... (4)
i.e. fj. denotes the ratio of the stress to the strain (appropriately
measured) in a pure shear. It is called the " rigidity " of the
substance.
Again, suppose we have a bar stretched lengthways, but free
from lateral stress. We put, then, in (1), p% = 0, p 3 = 0. This
leads to
pi = Ee l9 .............. . ............ (5)
where ^ ................
Xf p
This ratio of the longitudinal stress to the corresponding
extension is called "Young's modulus" of elasticity; its
technical importance is obvious. We also find
2 = e 8 = o 1 , ........................ (7)
Where
This fraction accordingly measures the ratio of lateral con
traction to longitudinal extension under the circumstances
supposed ; it is known as " Poisson's ratio." *
* 8. D. Poisson (17811840). His chief contributions to acoustics relate to
the vibrations of membranes and plates, and to the general theory of sound
waves in air.
112 DYNAMICAL THEORY OF SOUND
By solving the equations (1) we can express e l} e 2 , e 3 as linear
functions of p lt p z , p a . It is obvious, however, that the formula
for e l must involve p z and p s symmetrically; and from this
consideration, and from the physical meanings of the constants
E and cr, it follows immediately that the result must be
equivalent to
E t *>p l #(p t +p t ), ;
............... (9)
Of the various elastic constants and their combinations,
one or other may appear specially important, according to the
nature of the question in view, and this may account for the
great diversity of notations which has arisen. In any case two
independent quantities are necessary and sufficient to define the
elastic behaviour of an isotropic substance. From a physical
standpoint K and p, might appear to be the most fundamental ;
whilst as regards facility of direct measurement preference may
be given to E and //,, whence K and cr can be derived by the
formulae
fiE E
" = 9 P 3E' ^V" 1 ' ............ (1
which follow easily from (3), (6) and (7). On a particular
hypothesis as to the ultimate structure of an elastic solid
Poisson was led to the conclusion that the two elastic constants
are not independent, but are connected by an invariable relation,
which in our notation is expressed by X = /JL. This makes
=/*, E=ln, cr = J ................ (11)
On experimental grounds Wertheim (1848) proposed the
relation X = 2/A, which makes
* = f ft E = K, <r = J ................ (12)
More accurate methods of measurement, introduced by
Kirchhoff* and others, support the view, which has been con
sistently held by English physicists f that there is no necessary
* G. E. Kirchhoff (182487), professor of physics at Heidelberg 185475,
at Berlin 1875 87 ; famous for his share in the discovery of spectrum analysis,
but the author also of important memoirs on the theory of elasticity and its
applications to the vibrations of bars and plates.
t Notably by Green (17931841), Stokes, and Lord Kelvin.
BARS
113
definite relation between X and fj,, and consequently no universal
value of cr. We may note that in an absolutely incompressible
medium we should have
* = oo, E=%n, <r = i (13)
The following table gives the results of a few determinations
by Everett (1867). The second column gives the volume
density in grammes per cubic centimetre. The next three
columns give the respective elastic constants, in dynes per
square centimetre. These are followed in the last column by
the corresponding values of a. The last two rows illustrate the
fact that the elastic constants may vary appreciably in different
specimens of nominally the same substance.
P
E
/*
K
ff
Steel
7849
2139 x 10 12
819 xlO 11
1841 XlO 12
310
Iron (wrought)
7677
1963x1012
769 xlO 11
1456 XlO 12
275
Iron (cast)
7235
l349x!0 12
532 x 10 11
964 xlO 12
267
Copper
8843
1234 xlO 1 *
447 x 10 11
1684 xlO 12
378
Glass (1)
2942
603 x 10 12
240 x 10 11
415 x 10 12
258
Glass (2)
2935
574 x 10 12
235 x 10"
347 xlO 12
229
For technical purposes the elastic constants E, K, fj, are
often expressed in gravitation measure, e.g. in grammes per
square centimetre. The corresponding numbers in the above
table are then divided by g. Another mode of specification,
employed by Young, is in terms of the length of a bar of the
particular substance, whose weight per unit area of crosssection
would be equal to the modulus in question when expressed in
gravitation measure; this is called the "lengthmodulus." Thus
if L be the lengthmodulus of extension of a bar free to contract
laterally we have
E= ffP L (14)
Taking g 981, the above table gives, in the case of steel,
L = 278 x 10 6 centimetres.
114 DYNAMICAL THEORY OF SOUND
The potential energy (W) per unit volume of a strained
isotropic substance may be found by calculating the work done
by the stresses on the faces of a unit cube, on the hypothesis
that the strains increase from zero to their final values keeping
their mutual ratios unchanged. The average stresses are then
onehalf the final stresses.
Thus in the case of a uniform dilatation A we have
In the case of a pure shear 77,
Tr=4^ = J^ ................... (16)
In the extension of a bar, with freedom of lateral con
traction,
W=J ft e, = itf,' ......... . ......... (17) i
In the general case we have
, + 2 + 3 ) 2 + p (ef + 6 2 2 + 6 3 2 )
+ J /* {(e 2  6 3 ) 2 + (e 3  6l ) + (e t  6 2 ) 2 ). . . .(18)
This shews that in order that the potential energy may be
a minimum in the unstrained state K and p must be positive.
It is otherwise obvious from the meaning of the symbols that
if either of these were negative the unstrained state would be
unstable.
43. Longitudinal Vibrations of Bars.
We take the axis of x along the bar, and denote by x + f
the position at time t of that crosssection whose undisturbed
position is a?, so that f denotes the displacement. An element of
length is then altered from Sac to S(#+ f), or (1 + ')Sa;, where
the accent denotes differentiation with respect to x. Equating
this to (1 + e) 8x, we have
.!* ............................ CD
dx
The tension across the sectional area (&>) is therefore Eeco.
The acceleration of momentum of the mass included between
the two crosssections corresponding to x and x + Sx is p&Sx . %.
BARS 115
Equating this to the difference of the forces on the two ends,
we have
If the section be uniform, this reduces to
^^ (3)
~**
where c* = E/p ........................... (4)
It will be noticed that in this investigation it is not
necessary to assume the substance of the bar to be isotropic,
provided the proper value of the Young's modulus be taken*.
The result is also unaffected if the bar, or wire, be subject
to a permanent longitudinal tension, since by Hooke's law
the stress due to the extension (1) may be superposed on the
permanent tension, so long as the limits of perfect elasticity
are not transgressed.
As in 23 the general solution of (3) is
g = f(ctx) + F(ct+x), ............... (5)
representing two wavesystems travelling unchanged in opposite
directions with the velocity c given by (4). In terms of the
lengthmodulus, we have by 42 (14)
c = V(<?); ........................ (6)
this is the velocity due to a fall from rest through a height J L.
Some numerical values of c are given in the last column of the
table on p. 119.
The application to particular problems may be treated very
briefly. The various cases that arise present themselves in
a more interesting form when we come to the vibrations of
columns of air.
In the case of a rod or wire fixed at both ends, we have
f = for x = and x = I (say) ; and the mathematical theory
* In an "aelotropic" or crystalline solid the values of E will be different for
bars cut in different directions from the substance.
82
116 DYNAMICAL THEORY OF SOUND
is exactly the same as in the case of the transverse vibrations of
a string. The frequencies of the various modes are given by
N = sc/2l, 00
where s = l, 2, 3, .... The result is unaffected by permanent
tension in the wire.
When the rod is free, the condition of zero stress at the
ends gives f ' = for x = and x = I. Introducing this condition
in (5) we find
F' ( ct ) = f (ct), F'(ct + l)=f(ctl), (8)
for all values of t. The former of these gives on integration
F(ct)=f(ct), (9)
no explicit additive constant being necessary since it may be
supposed included in the value of f(ct). The second relation
then gives
f(ct + l)=f(ctl)+C. (10)
The constant G is connected with the total momentum of the
bar in the direction of its length. We have, from (9) and (10),
f f dx=c \\f\ct  x) +f(ct + x)} dx = cC. . . .(11)
Jo Jo
Since nothing essential is altered if we superpose any uniform
velocity in the direction of the length, we may assume the
masscentre to be at rest, in which case (7=0. The formula
(10) then shews that the residual motion is periodic, since
everything recurs when t increases by 2l/c.
In the analytical process for ascertaining the normal modes
we assume that f varies as cos (nt f e), whence
and
(WIT W 'T'N
4 cos +sin ) cos (nt + e) (13)
c c /
The conditions that 9j/3a? = for x and x = I require B 0,
sin (nl/c) = 0, whence
nJ/c = 67r, (14)
where 5 = 0, 1, 2, 3, ..., the scale of periods being harmonic.
The nodes (f = 0) are given by cos (sirx/l) = 0, and the loops,
BARS 117
or places of zero stress, by sin (STTX/I) = 0. In the gravest mode
(s 1) we have a node at the centre*.
On the principles explained in 16, 32 the most general
free motion of the bar, under the present conditions, may be
expressed by a series
STTCt D . S7TCt\ S7TX
cos j + B 8 sin = \ cos y, ...... (15)
where 5 = 0, 1 , 2, 3, .... Thus if the bar be started from rest
in the state of strain defined by
= /<*) [ = 0], .................. (16)
we have B 8 = Q; and we infer that it must be possible to
determine the coefficient A a so that
^ / A
= 2,(A a
(17)
for values of x ranging from to /. This is the result referred
to by anticipation in 33.
The longitudinal vibrations of bars or wires have hardly
any practical application of importance, except in some primitive
forms of telephone. As regards bars, the pitch is very high
compared with that of the transverse vibrations, which it is
difficult to avoid exciting simultaneously. Again, if we compare
the frequencies of longitudinal vibration of a tense wire with
those of the corresponding transverse modes, the ratio will be
that of the wavevelocities, i.e. of *J(E/p) to ^(P/pco), where P
denotes the permanent tension. If e be the extension due
to P, we have P = Ee .(o, and the ratio is l/\/e , which is
usually very great f. Longitudinal vibrations may be elicited
on the monochord by rubbing the wire lengthwise with a piece
of leather sprinkled with resin ; the resulting note is very
shrill.
It is assumed in the preceding theory that the extension
and the accompanying stress are at any instant uniform over
the crosssection ; in other words, we have assumed that the
* The case * = needs, in strictness, separate examination. It leads to
= AQ (l + at), which may be interpreted as an oscillation of infinitely long
period. If the masscentre be at rest we have A =0.
t This comparison is due to Poisson (1828).
118 DYNAMICAL THEOEY OF SOUND
lateral contraction adjusts itself instantaneously through the
thickness. This is not quite exact, as there is a certain degree
of lateral inertia, but the error is insignificant so long as the
wavelength is large compared with the diameter. In the
modes of very high order it might become sensible, but these
are in any case of no importance from the point of view
of acoustics. ~~ A correction has been investigated by Lord
Rayleigh.
44. Plane Waves in an Elastic Medium.
The theory of plane waves in an unlimited isotropic elastic
medium is so closely analogous to that of longitudinal waves
in a rod that it may be briefly noticed here. It is assumed
that the state of things is at any instant uniform over any
plane perpendicular to the direction of propagation (#).
Such waves may be of two types, which are distinguished
as " dilatational " or " longitudinal," and " distortional " or
" transversal," respectively. In the former class the displace
ment is wholly in the direction of propagation. Denoting it
by f, we have, in the notation of 42,
6! = dg/diK, 6 2 = 0, 3 = 0,
and therefore p l = (\ + 2/i) l = (K + f /*) d/dx .......... (1)
Considering the portion of matter corresponding to unit area
of a stratum of thickness 8x, we have
whence =
< 2 >
if tt = (* + iri)lp ...................... (3)
Some numerical values of the wavevelocity a are given on
the next page, and it will be observed that they are in all
cases greater than the corresponding values of c, as was to be
expected, since the potential energy due to a given extension
8f/3# is now greater owing to the absence of lateral yielding.
In the second type of plane waves the displacement is
everywhere at right angles to the direction of propagation.
It may be resolved into two components parallel to y and z,
respectively, which may be treated separately. Considering
BARS
119
the former component (17) alone, we see that the strain at
any point consists in a shear of amount drj/dx. The consequent
stress across any plane perpendicular to Ox is parallel to Oy,
and its intensity is fidrj/dx. Hence forming the equation of
motion of a portion of matter defined as before we have
or
if V=dp (5)
Some values of the wavevelocity b are tabulated below.
Wavevelocities (metres per second).
a
b
c
Steel
611 xlO 3
323 x lO 3
522 x 10 3
Iron (wrought)
568 x 10 3
316x103
506 x 10 3
Iron (cast)
481 x 10 3
271 x 10 3
432 x 10 3
Copper
508 xlO 3
225 xlO 3
374X10 3
Glass (1)
500 xlO 3
286 x 10 3
453X10 3
Glass (2)
474 x 10 3
283 x 10 3
442 x 10 3
It may be shewn that any local disturbance in an unlimited
elastic medium breaks up into two waves, diverging with the
velocities a and 6, which tend ultimately to assume the
" longitudinal " and " transverse " characters, respectively.
The theory is historically important in relation to Optics,
but in our present subject great caution is necessary in
drawing inferences as to the propagation of waves in limited
solids. We have already seen that in a cylindrical or prismatic
rod the velocity of longitudinal waves is quite distinct from a,
and the theory becomes altogether different in the case of
flexural vibrations, to be referred to presently. In these cases
a modification was of course to be expected, since the wave
length is understood to be large compared with the dimensions
120 DYNAMICAL THEOBY OF SOUND
of the crosssection. But even in the other extreme, when all
the dimensions of the body are large compared with the wave
length, the circumstances may be profoundly modified by the
existence of a free boundary. A new type of waves, called
after the discoverer the " Rayleigh waves " (1885), make their
appearance, and under some conditions may become, from the
observational point of view, predominant. These are surface
waves in which the agitation penetrates only to a relatively
small depth. Their velocity is somewhat less than that of
the distortional waves ; thus for an incompressible solid it
is *9554 b, whilst on Poisson's hypothesis (cr = J) it is '91946.
In modern observations of the tremors due to distant earth
quakes three phases of the disturbance are often recognized.
The first is interpreted as due to the arrival of the dilatational
waves, propagated directly through the substance of the earth,
the second as due to that of the distortional waves, also
propagated directly, and the third to that of the Rayleigh
waves, which have travelled over the surface and are therefore
delayed more than in proportion to the difference of wave
velocity*. The latter waves as they spread over the surface
are less attenuated than the former, which diverge in three
dimensions. It has even been attempted to deduce estimates
of the volumeelasticity and rigidity of the materials of the
earth from the various wavevelocities, as inferred from the
seismic records f.
45. Plexural Vibrations of a Bar.
We proceed to the transverse jv&F
vibrations of a bar naturally straight.
To avoid unnecessary complications
we will suppose that the bar has a
longitudinal plane of symmetry, and
that the flexure takes place parallel
to this plane. We will also assume
for the present that the total longi
tudinal stress on any section is zero.
The resultant stress at a section there Fig. 42.
* E. D. Oldham, Phil. Tram. A, 1900.
t Prof. A. E. H. Love, Phil. Trans. A, vol. ccvu., p. 215 (1908).
^%
( M+bN
^ *r
121
fore reduces to a, transverse " shearing force " F, and a couple
or " bending moment " M. These will be functions of x, the
longitudinal coordinate. If rj denote the lateral displacement,
parallel to the plane of symmetry, then, resolving transversally
the forces acting on an element Sx of the length, we have
Y)
or P &) o^ = o~ ...................... (1)
dt z dx
Again, if K denote the radius of gyration of the area of the
crosssection G> about an axis through its centre of gravity,
normal to the plane of flexure, the element of mass is
ultimately a disk of area o>, thickness 8x, and moment of
inertia pwSx . K 2 *. Since the axis of this disk has been turned
through a small angle drj/dx from the position of equilibrium,
the equation of angular motion is
whence pmK = + F . .................. (2)
Tf we eliminate F between (1) and (2) we have
provided the sectional area 6> be uniform.
We have next to express M in terms of the deformation
of the bar. Consider in the first instance the case of a bar
uniformly bent, so that its axis becomes an arc of a circle.
It is evident from symmetry that the shearing force F now
vanishes, and it hardly needs calculation to shew that the
strain in any part of the crosssection will be proportional to
the curvature. Hence by Hooke's law the resultant couple M
will also vary as the curvature, or
M=WIR, ........................ (4)
where R is the radius of curvature, and 23 is a constant
depending on the shape and size of the crosssection, and on
the elastic properties of the material.
* The symbol K is not required at present in its former sense as an elastic
constant.
122
DYNAMICAL THEORY OF SOUND
J
J
The value of 23 is found as follows. We take rectangular
axes Gy, Gz in the plane of a
section, the origin being at the
centre (i.e. the centre of gravity
of the area), and the axis of z
normal to the plane of flexure.
Assuming the axis of the bar,
i.e. the line through the centres
of the sections, to be unex
tended, we see that if R denote
the radius of the circle into
which it is bent, the length
of a longitudinal linear element
whose distance from the plane
xz is y is altered in the ratio
of R + y to R, and that the
extension is accordingly y/R.
The corresponding stress per
unit area of the section is JEy/R,
where E is the appropriate Young's modulus. The total
longitudinal tension is therefore
E
Fig. 43.
This justifies the provisional assumption that the axis (as
above defined) is on the present hypothesis unextended. For
the bending moment we have, taking moments about Gz,
E
Except in the special case just considered, viz. that of a bar
bent statically into an arc of uniform curvature, there will be
a shearing of crosssections relative to one another, and also
a warping of the sections so that these do not remain accurately
plane. An exact investigation is out of the question, but
enough is understood of the matter to warrant the statement
that the additional sfam'rcs thus introduced are as a rule small
compared with those taken account of in the preceding calcula
tion. We therefore adopt the formula (5) as sufficiently
BARS 123
accurate in all cases, provided R denote the radius of curvature
at the point considered.
In the present application drj/dx is a small quantity, so that
we may put R~* = &rj/dx?, and therefore
(6)
Substituting in (3) we obtain
(7)
For most purposes this equation may be simplified by the
omission of the second term, as we shall see immediately.
The kinetic energy of the bar is
The second term, which represents the energy of rotation of the
elements, is usually negligible.
The potential energy is found, in accordance with 42 (16),
by integrating the expression %Ee z , =\Ey' i lJ&, first over the
area of the crosssection, and then over the length; thus
. .............. .(9)
Consider for a moment the propagation of a system of waves
of simpleharmonic profile along an unlimited rod, assuming
v)=Ccosk(ctx) ................... (10)
Since everything here recurs whenever # is increased by 2Tr/fc,
the constant k is connected with the wavelength A. by the
relation
fc=27T/X ......................... (11)
On substitution the equation (7) is found to be satisfied
provided
This gives the wavevelocity c, which is seen not to be a definite
quantity fixed by the constitution of the rod, but to depend also
on the wavelength. To trace the progress of a wave of any
type other than (10), it would be necessary to resolve the wave
124 DYNAMICAL THEORY OF SOUND
form into simpleharmonic functions of x. Each of these would
travel with its own velocity, so that the resultant waveprofile
would continually alter. For this reason it would be hopeless
to look for a general solution of (7), or even of the modified
form (13) below, of the same simple character that we met
with in the theory of strings ( 23), and again in that of the
longitudinal vibrations of rods.
A further remark is that when we substitute from (10)
in (7), the second term is of the order &V as compared with
the first. When the wavelength is large compared with the
dimensions of the crosssection this is a very small quantity,
and the term in question, which arose through taking account
of the rotatory inertia of the elements of the bar, viz. in
equation (2), may be neglected. It is easy to see, and it may
be verified a posteriori, that the same simplification is legitimate
in discussing the vibrations of a finite bar, at all events so long
as the distance between successive nodes is large compared
with K. We accordingly take the equation
P
as the basis of our subsequent work, together with the formulae
M = E**%{, FWE^. ...(14)
3# a doc W
46. FreeFree Bar.
To ascertain the normal modes of a finite bar we assume as
usual that 77 varies as cos (nt + e). The equation (13) of the
preceding section then reduces to
where m* = ri i p/K 2 E. ........................ (2)
It is to be noted that m is of the nature of the reciprocal of
a line. The solution of (1) is
77 = A cosh mx + B sinh mx f C cos mx + D sin mx, (3)
the timefactor being for the present omitted. The three ratios
A:B:C:D, and the admissible values of m, and thence of n\
are fixed by the four terminal conditions, viz. two for each end.
BARS
125
Take first the case of a perfectly free bar, of length I, say.
If we take the origin at the middle*, these conditions are, by
45(14),
," = 0, ,'"=0 [ci<] (4)
The normal modes fall naturally into two classes ; in one of
these 77 is an even, in the other an odd function of x. For the
symmetrical vibrations we have
77 = A cosh mx + C cos mx, (5)
with the terminal conditions
A cosh J ml G cos ^ml = 0, j
A sinh ^ml + Csin w = 0,J
whence tanh ^ml= tan ^ml (7)
+ 1
1
Fig. 44.
The roots of this equation are easily found approximately by
graphical construction, viz. as the abscissae of the intersections
of the curves y = tan #, y = tanh x, the latter of which is
* This improvement on the ordinary procedure is due to Sir A. G. Greenhill,
Mess, of Math. vol. xvi., p. 115 (1886).
126 DYNAMICAL THEOKY OF SOUND
asymptotic to the line y = 1. The figure shews that we have
approximately
) .................. (8)
where 6? = 1, 2, 3, ..., and a s is small. It follows from (2) that
the frequencies of the successive normal modes of symmetrical
type are approximately proportional to 3 2 , 7 2 , II 2 , .... For a
more exact computation of the roots we have
_ __ m _
ia *
where s = e~ sir ...................... (10)
Hence a 8 = ten> (&' 2a >) = ^ s e~^^e~^+ .... (11)
Since ? g is small, even for s = 1 (viz. & = "00898), this is easily
solved by successive approximation.
In the asymmetric modes we have
rj = Bsinhnuc + Dsmmx, ............... (12)
with the terminal conditions
B sinh J ml D sin \ml 0,j
cosh imJD cos JmZ = 0,J
whence tanh \ml = tan \ml ................... (14)
The roots of this are given by the intersections of the curves
y = tan x, y = tanh X, the latter of which is asymptotic to the
line y = 1 ; see Fig. 44. It appears that
JmJ = ( + J)ir&, .................. (15)
where 5 = 1, 2, 3, ..., and ft is small. The corresponding
frequencies are approximately proportional to 5 2 , 9 2 , 13 2 , ....
For the more exact calculation we have
1+ tan mn. tanh
where ^ = e~ 2s7r ~ i7r ........ , .......... ...(17)
Hence ft = tan 1 (^ 2ft )= ?^ 8 K^~ 6 ^ + ....... (18)
Since i = "00039, the approximation is very rapid, even for
BAES 127
Combining the results for the two classes it is found that
wZ/7T = 150562, 2*49975, 3'50001, ..., ...(19)
where the values for the symmetric and asymmetric types
alternate. The subsequent numbers are adequately represented
by s + . The fact that the frequencies are approximately
proportional to 3 2 , 5 a , 7 2 , ... was ascertained, from observation
alone, by Chladni*.
To examine the form assumed by the bar in any normal
mode, we require the ratio of the arbitrary constants, as deter
mined by (6) or (13). Thus in the case of symmetry we have
rj = C (cos  ml cosh mx + cosh ml cos mx) cos (nt + e), (20)
where m is a root of (7). The curve may be traced with the
help of a table of hyperbolic functions, and the positions of the
nodes found by interpolation. The form assumed in the
gravest mode is shewn in Fig. 45. The nodes here are at a
distance of '224 of the length from the ends.
Fig. 45.
The corresponding formula for the asymmetric modes is
77 = G (sin J ml sinh mx + sinh ml sin mx) cos (nt 4 e), (21)
where m is determined by (14).
47. Clampedfree Bar.
The next most interesting case is that of a bar clamped at
one end and free at the other. Here also there is an advantage
in taking the origin at the middle point of the length ). The
terminal conditions then are
^=0, v = o [* = jq (i)
* E. F. F. Chladni, born at Wittenberg 1756, died at Breslau 1827.
Distinguished by his experimental researches in acoustics. These are recorded
in his book Die Akustik, Leipzig, 1802.
t Greenhill, 1. c.
128 DYNAMICAL THEORY OF SOUND
at the clamped end, and
," = 0, ,'" = O = JZ] (2)
for the free end. In one class of vibrations we have
77 = A cosh mx f D sin mx, (3)
with the conditions
A cosh \ml D sin \ml 0,^1 ...
A sinh ^ml + Dcos ^ ml = 0, 1
whence coth \ml = tan \ml (5)
This is solved graphically by the intersections of the curves
y tan cc, y = coth x, the latter of which has y 1 as an
asymptote ; see Fig. 44. We have, approximately,
iwiZ = (* + i)7r + a; (6)
where s = 0, 1, 2, 3, ..., and a/ is small. This leads to
tana/^ 2 "'', (7)
where f f = e~ 2 " iir , (8)
whence a/= ^~ 2a<! 'i ?.'*~ 6 ; + , (9)
which can easily be solved by successive approximation, except
in the case of the first root (s = 0). For this special methods
are necessary*. In the remaining type of vibrations we have
77 = B sinh mx + Ccosmx, (10)
with Bsinh ml +
.(11)
B cosh \ ml + C sin  ml = 0, 1
whence coth \rnl tan \ml ................ (12)
The intersections of the curves y = tan ,r., y = coth x are also
shewn in Fig. 44. The roots of (12) are given by
imJ = (*i)9T&', ............... (13)
where s = 1, 2, 3, .... Hence
? s 6 2 ^, .................. (14)
where f. = e"" ir , .................. (15)
and therefore
A' = &*'  J ?. w + ............. (16)
* One such method will be indicated later in connection with the radial
vibrations of air in a spherical vessel ( 84). Another very powerful method is
explained in Rayleigh's treatise.
BAES 129
The frequencies of the whole series of normal modes, after the
first, are approximately proportional to 3 2 , 5 2 , 7 2 , . . . , as found
experimentally by Chladni. The accurate solution gives, to
five places,
ml/7r = 59686, V49418, 250025, (17)
In the modes which follow the first we have respectively one,
two, three, ... internal nodes. The annexed figure shews the
gravest mode.
Fig. 46.
Other problems, which are however of less interest, may be
obtained by varying the terminal conditions. We will only
notice the case where both ends are " supported," i.e. fixed in
position but free from terminal couples. The conditions then
are, by 45 (14),
77 = 0, 7/"=o >=iq (is)
In the symmetrical class we have
77 = G cos mx . cos (nt f e), (19)
with cos ^ml = 0, whence
mll7r = l, 3, 5, (20)
In the asymmetric class
77 = G sin mx . cos (nt + e), (21)
with raZ/7r = 2, 4, 6, (22)
The frequencies are, by 46 (2), proportional to the values of
m 2 , and so to the squares of the natural numbers.
The foundations of the theory of the transverse vibrations
were laid by D. Bernoulli (1735) and Euler (1740). The
latter also gave the numerical solution of the period equation
in a few of the more important cases. In more recent times
L. 9
130 DYNAMICAL THEORY OF SOUND
the calculations, including the determination of the nodes &c.,
have been greatly extended by Lissajous (1850), Seebeck* (1848)
and Lord Rayleigh.
48. Summary of Results. Forced Vibrations.
In any one of the preceding cases, and in any particular
mode, ra varies inversely as I, and therefore, by 46 (2), the
period 27r/n will for bars of the same material vary as Z 2 //c.
Hence for bars which are in all respects similar to one another
(geometrically) the period will vary as the linear scale. For
bars of the same section the period is as the square of the
length. As regards the shape and size of the cross section,
everything depends on the radius of gyration K\ thus for bars of
rectangular section the frequency varies as the thickness in the
plane of vibration, and is independent of the lateral dimension.
This latter statement needs, however, some qualification ; it is
implied that the breadth is small compared with the length of
the bar, or (more precisely) with the distance between con
secutive nodes. When this condition is violated the problem
comes under the more complex theory of plates ( 55).
It is of interest to compare the frequencies of transverse and
longitudinal vibration of a bar in corresponding cases. For a
bar free at both ends we have, in the gravest transverse mode,
?i 2 = / ^(m0 4 = ^ / ^x(l50562) 4 , (1)
whilst in the gravest longitudinal mode
(2)
Hence , = 7'122y (3)
71 I
This explains the relative slowness of the transversal modes.
The comparison is due to Poisson.
We pass over the question of determining the motion
consequent on arbitrary initial conditions, by means of the
normal functions. In the case of the freefree bar, for example,
these are given by the expressions in brackets in equations (20)
and (21) of 46.
* L, F. W, A, Seebeck (180549), professor of physics at Leipzig.
BARS 131
The theory of forced vibrations again, is of little acoustical
interest, although it has some technical importance. A simple
example is furnished by the coupling rod which connects the
wheels of a locomotive. Attending only to the vertical com
ponent of the motion, and treating the bar as uniform, we have
to solve the equation (13) of 45 subject to the conditions
where n is the angular velocity of the wheels, and & is the
vertical amplitude. The forced oscillation is evidently of
symmetrical type, and we therefore assume
rj = ( A cosh mx + C cos mx) cos (pt + a) (5)
This satisfies the differential equation, provided
m*=p*p/K*E', (6)
whilst the terminal conditions give
A cosh^ml + Ocosra = /8, ) ^
A cosh \mlG cos \ml = 0, J
the latter equation expressing the absence of terminal couples
(" = 0). Hence
2 cosh ml'
The oscillations would become dangerously large if c
were small, i.e. if the imposed frequency (>/2?r) were to ap
proximate to that of one of the symmetrical free modes of the
bar when " supported " at the ends ( 47 (20).
49. Applications.
The use of transverse vibrations of bars in music is re
stricted by the fact that the overtones are not harmonic to the
fundamental. If a flat bar, otherwise free, be supported at the
nodes of the fundamental (Fig. 45), and struck with a soft
hammer, the production of overtones is, however, in some
measure discouraged, and musical instruments of a kind (such
as the " glass harmonica ") have been constructed on this plan.
The most important application is in the tuning fork.
92
132 DYNAMICAL THEOKY OF SOUND
Theory and observation alike shew that the effect of curving
a bar is to lower the pitch of the gravest mode and to make the
nodes approach the centre. It was found by Chladni that
when the bar takes the form of an elongated U, the nodes are
very close to the bend. The amplitude of vibration at the
centre of the bend will therefore be small compared with that
at the end of the prongs. The circumstances are somewhat
modified by the attachment of the stem, but the transmission
of energy is comparatively slow, and the vibrations have con
siderable persistence. A fork may also be compared to a couple
of bars each clamped at one end, and the formula (2) of 46,
with ml/7r = '59686, may be used to estimate the frequency
theoretically. If this analogy were exact there would of course
be no loss of energy of the kind just referred to.
Massive forks are usually set into vibration by means of
a violoncello bow applied to one prong near the free end. The
production of overtones having nodes in this neighbourhood is
thus discouraged. The fundamental is further reinforced re
latively to the other modes if the stem be screwed into the
upper face of a resonance box of suitable dimensions.
When a fork is excited in this or in other ways, it often
happens that the motion is not in the first instance symmetrical
with respect to the medial plane. In that event the vibration
may be regarded as made up of a symmetrical and an unsym
metrical component. These will in general have slightly
different frequencies, and beats may be produced. But unless
the stem be very firmly fixed the vibrations of the latter class
are rapidly dissipated by being communicated to the support,
since they involve an oscillation of the centre of mass of the
fork.
The first overtone of a fork may be elicited in considerable
intensity by bowing one of the prongs near the bend ; the note
produced is very shrill.
50. Effect of Permanent Tension.
In the theory developed in 45 it was assumed that the
longitudinal tension, when integrated over the area of the
crosssection, vanishes. It is easy to see that the effect of
BARS 133
a permanent tension P is merely to add a term P?/' to the
equation (13) of 45, so that
where c z P/pa> (2)
This equation has been employed to estimate the effect of
stiffness of a pianowire on the sequence of proper tones, but
the matter is complicated by the uncertainty as to the nature
of the terminal conditions. A wire, where it passes over a
bridge, cannot be quite accurately regarded either as merely
"supported" or as "clamped." The question will perhaps be
sufficiently illustrated if we consider a wavesystem
7] = Ccosk(ctx) (3)
on an unlimited wire. We find, on substitution in (1)
c 2 = c 2 +cr (4)
E
where c^ = .k z /c', (5)
i.e. Cj is the velocity of transverse waves of length 27T/& on a bar
free from tension. We have seen that in the case of a piano
string E/p is large compared with c 2 ; on the other hand K is
usually an exceedingly minute fraction of the wavelength. In
the graver modes of a pianostring this second influence pre
dominates, and (Cj/Co) 2 is small ; the wavevelocity is practically
unaffected by stiffness, and the harmonic sequence is not
disturbed. It is only in the case of the modes of very high
order, where the length is divided into a large number of
vibrating segments, that a sensible effect could be looked for.
It has already been stated that in the pianoforte such modes are,
so far as may be, discouraged on independent grounds. In any
case it appears from (4) that the effect of stiffness is relatively
less important, the greater the value of c , i.e. the tighter the
wires are strung.
51. Vibrations of a Ring. Flexural and Extensional
Modes.
The theory of the vibrations of a circular ring is important
as throwing light on some later questions which can only be
134 DYNAMICAL THEORY OF SOUND
dealt with imperfectly in this book, owing to the difficulties of
an exact investigation. As various points of interest arise, we
treat the matter somewhat fully.
The ring is supposed to be uniform, and the section is
assumed to be symmetrical with respect to a plane perpen
dicular to the axis. We further consider only vibrations
parallel to this plane. Let u, v be the displacements of an
element of the ring along and at right angles to the original
radius vector, so that the polar coordinates of the element are
changed from (a, 0) to (a + u, 6 + v/a). We require expressions
for the extension, and for the change of curvature. In con
sequence of the assumed smallness of the displacements, we
may calculate the instalments of these quantities which are due
to u and v separately, and add the results. The radial displace
ment by itself changes the length of an element from a&O to
(a + u) 80, and so causes an extension u/a. The transverse
displacement obviously contributes dv/adO. The total extension
is therefore
.(1)
Again, in consequence of the radial displacement alone the
normal to the curve is rotated backwards so as to make an
angle du/adO with the radius, and the mutual inclination of the
normals at the ends of an element a BO is accordingly diminished
by fru/adfr . S0. Dividing the angle between the normals by
the altered length (a + u) BO we get the altered curvature, thus
(BO Tr^e>0)Ma
a a 2 \80 2
Since the transverse displacement v by itself contributes
nothing, the increase of curvature is
a?
The resultant stress across any section may be resolved into
a radial shearing force P, a tangential tension Q, and a bending
moment M. On the principles of 43, 45 we have
BARS
135
the bending moment being now proportional to the increase of
curvature.
Resolving along and perpendicular to the radius vector the
forces on a masselement pwaSO, we have (see Fig. 47)
P a>aS0. u = SP QSe, pvaW.v = BQ + PS0 ;
and, taking moments about a normal to the plane of the ring,
the rotational inertia being neglected as in the case of a straight
bar ( 45). Thus
8 2 w 8P n
a* = ;SC. P^ a ^ =
dt* 80
.(4)
and
80
= Pa.
(5)
P+bP
These, together with (3), are
the equations of our problem.
It is easily seen that they cannot
be satisfied on the assumption
that the tension Q vanishes, and
that accordingly some degree of
extension is involved in any
mode of vibration. This is
readily accounted for, a stress
of this kind being necessarily
called into play by the inertia of
the different portions swinging
in opposite directions. It may be shewn however that in the
" flexural " modes to be referred to presently the corresponding
strains are small compared with those involved in the change of
curvature.
Eliminating P, Q, M between (3), (4), and (5), we find
E ( dv
(6)
Fig. 47.
,
** +
_ L _ u
pa 2 + ~ + =
To ascertain the normal modes we assume that u and v vary
as cos (nt + e). Again, the ring being complete, u and v are
136 DYNAMICAL THEOKY OF SOUND
necessarily periodic functions of 6, the period being 2?r, and
can accordingly be expanded by Fourier's theorem in series of
sines and cosines of multiples of 6 ; moreover it is easily
proved that the terms of any given rank in the expansion must
satisfy the equations separately. We find, in fact, that a
sufficient assumption for our purpose is
u = A cos s6 . cos (nt + e), v = B sin sB . cos (nt + e), (7)
where 5 is integral or zero. This leads to
(8)
 2 )5 = 0,
/
where {3 = ri*a?p/E. (9)
Hence
Since tc/a is small, the sum of the roots of this quadratic in ft is
s 2 + 1, approximately, whilst the product s 2 (s 2 I) 2 /c 2 /a 2 is small.
The two roots are therefore
<?2 / Q 2 _ 1 \2 ,.2
0*+i, *'^J, ............ (ID
approximately.
On reference to (8) we see that the former root makes
B = sA nearly. The corresponding modes are closely analogous
to the longitudinal modes of a straight bar, the potential
energy being mainly due to the extension ; and the frequencies,
which are given by
2 = (* 2 + l) / . .................. (12)
are, for similar dimensions, of the like order of magnitude. The
case s = is that of purely radial vibrations.
The vibrations corresponding to the second root are more
important. We then have, from (8), A + sB = 0, nearly ; thus
A
u = A cos s& . cos (nt + e), v =  sin s& . cos (nt + e), (13)
s
with n'. ...(14)
2 4
BARS
137
It follows from (1) that the extension is negligible, and the
energy mainly flexural. The frequencies are in fact comparable
with those of transverse vibration of a bar. In the mode of
order s there are 2s nodes, or places of vanishing radial motion,
but these are not points of rest, the tangential motion being
there a maximum*. In the case 5=1 the circle is merely
displaced as a whole, without deformation, and the period is
Fig. 48.
accordingly infinite. The most important case is that of s = 2,
where the ring oscillates between two slightly elliptical extreme
forms. The arrows in the annexed figure shew the directions of
motion at various parts of the circumference at two epochs,
separated by half a period, when the ring passes through its
equilibrium position. The dotted lines pass through the nodes
of the radial vibration.
One farther point is to be noticed. Owing to the assumed
uniformity of the ring the origin of 6 is arbitrary, and other
modes, with the same frequencies, are obtained by adding a
constant to 0. In particular we have the flexural mode
u = A sin sO . cos (nt + e), v = cos sd . cos (nt + e), (15)
s
with the same value of n 2 as in (14). We have here an instance
of the kind referred to in 16, where two distinct normal modes
* This point is illustrated by the vibrations of a fingerbowl when excited by
drawing a wetted finger along the edge. The point of rubbing is a node as
regards the radial vibration, and the crispations on the contained water are
accordingly most conspicuous at distances of 45 on either side, where the radial
motion is a maximum.
138 DYNAMICAL THEOKY OF SOUND
have the same frequency, and the modes themselves accordingly
become to some extent indeterminate. The case would be
altered at once if the ring were not quite uniform, e.g. if it were
slightly thicker at one point. The normal modes in which
there is a node or a loop respectively, of radial vibration, at
this point would differ somewhat in character, and have slightly
different frequencies. Accordingly when both modes are excited
we should have beats between the corresponding tones. This
is a phenomenon often noticeable in the case of bells (and
fingerbowls), the inequality being due to a slight defect of
symmetry.
The vibrations of a ring in its own plane were first investi
gated by R. Hoppe (1871) ; a simplified treatment of the flexural
modes was subsequently given by Lord Rayleigh. The theory
of vibrations normal to the plane is more intricate, since torsion
is involved as well as flexure. The problem has been solved by
J. H. Michell (1889), who finds, in the case of circular cross
section,
s 2 + 1 + a ' pa 4 '
where a is Poisson's ratio.
CHAPTER V
MEMBRANES AND PLATES
52. Equation of Motion of a Membrane. Energy.
The vibrations of membranes are not very important in
themselves, and the conditions assumed for the sake of mathe
matical simplicity are, moreover, not easily realized experi
mentally. The theory is however, for a twodimensional system,
comparatively simple, and the results help us to understand in
a general way the character of the normal modes in other cases
where the difficulties of calculation are much greater, and
indeed often insuperable.
The ideal membrane of theory is a material surface such that
the stress across any lineelement drawn on it is always in the
tangent plane. We shall consider only cases where the surface
in its undisturbed state is plane, and is in a state of uniform, or
"homogeneous," stress; i.e. it is assumed that the stresses across
any two parallel and equal lines are the same in direction and
magnitude. We further suppose, for simplicity, that the stress
across any lineelement is perpendicular to that element. It
follows, exactly as in hydrostatics, from a consideration of the
forces acting on the contour of a triangular area, that the stress
(per unit length) is the same for all directions of a lineelement.
This uniform stress is called the " tension " of the membrane ;
we denote it by P. Its dimensions are those of a force divided
by a line, or [MT~*].
We take rectangular axes of x, y in the plane of the
undisturbed membrane, and denote by f the displacement
normal to this plane. The surfacedensity (i.e. the mass per
140 DYNAMICAL THEOEY OF SOUND
unit area) is assumed to be uniform, and is denoted by p. To
form the equations of motion we calculate the forces on the
sides of a rectangular element BxSy having its centre at (x, y).
In the displaced position, the gradient of a line parallel to x is
3?/da?, and that of a line parallel to y is d/dy. Hence the stress
across a line through the centre of the element parallel to By,
when resolved in the direction of the normal to the plane xy, is
Pd^/dx . Sy. The corresponding components of force on the two
edges % of the rectangle are
where the upper signs relate to the edge whose abscissa is
x + J&z?, and the lower to the edge x ^x. The sum of these
gives P9 2 f/3# 2 . SxSy. A similar calculation for the two edges &z?
gives Pd*/dy 2 . &x$y. The resultant force on the rectangle is
therefore
<
The above may be compared with the investigation by
which, in the theory of Capillarity, the tensions across the
boundary of an element 8S of a soapfilm are shewn to be
equivalent to a normal force
where B lf R 2 are the principal radii of curvature of the surface.
It is shewn in books on solid geometry that, if f denote distance
from the plane xy, we have
R, R, 8# 2 a^ 2
at points where the inclination of the tangent plane to xy is
small.
Equating the expression (1) to the acceleration of momentum
of the element, viz. p&xSy . , we obtain the equation of motion
This is due to Euler (1766).
MEMBRANES AND PLATES 141
The kinetic energy is given by
taken over the area of the membrane.
The potential energy is found most easily as the work
required to stretch the membrane. As in the theory of
capillarity this is equal to the tension P multiplied by the
increase of area. Now if a prism be constructed on a rectangular
element Sx&y of the plane xy as base, this will cut out from the
displaced membrane a nearly rectangular portion whose sides
are
~ VKS'h
and whose area is therefore, to the second order,
The same expression is obtained by calculating, from the
expression (1), the work done by normal pressures applied
(as in 22) to deform the membrane into its actual shape, the
ratio of f to its final value being, at any stage of the process,
the same all over the membrane. The result is
The reader who is familiar with the theory of attractions will
recognize that this is equal to
where in the first term the integration extends over all the
elements Ss of the contour, and 8n is an element of the normal
to &s drawn inwards, in the plane of the membrane. Since
at a fixed edge f = 0, the formula agrees with (5).
142 DYNAMICAL THEOKY OF SOUND
53. Square Membrane. Normal Modes.
To ascertain the normal modes of a limited membrane we
assume as usual that f varies as cos (nt 4 e), so that
where k* = n*p/P ......................... (2)
At a fixed boundary we must have f = 0. It is found that
the solution of (1) subject to this condition is possible only
for a series of definite values of k, which determine, by (2), the
corresponding frequencies.
In the case of a rectangular membrane, we take the origin
at a corner, and the axes of x, y along the edges which meet
there. The equations of the remaining edges being, say,
x = a, y = b, the equation (1) and the boundary condition
are satisfied by
f=0sin sin^cos(ri$ + e), ......... (3)
CL (J
where s, s' are integers, provided
It may be shewn, by an easy extension of Fourier's theorem,
that (3) is the only admissible type of solution in the present
case ; it was given by Poisson in 1829.
In any normal mode for which s or s' > 1, we have nodal
lines parallel to the edges. It appears from (4) that if the
ratio a? : b 2 is not equal to that of two integers, the frequencies
are all distinct, and the nodal lines are restricted to these
forms. But if a 2 : 6 2 is commensurable, some of the periods
coincide, and the corresponding modes may be superposed
in arbitrary proportions ( 16). The nodal lines may then
assume a great variety of forms. The simplest instance is
that of the square membrane (a = 6), when
) ...................... (5)
MEMBKANES AND PLATES
143
Thus by superposition of the modes for which s = 2, s' = 1 and
s = 1, s' = 2, respectively, we get
oe sin
.Try . irx .
sm   + X sin sm
a a a a
. TTX . iry ( TTX Try\ , .
sm sm   cos h Xcos 1 , (6)
a a \ a a J
where X may have any value. For example, in the cases X= 1
the diagonals a; + y = a, x y = 0, respectively, are nodal lines.
The figure shews the cases X = 0, X = , X = 1, which
give a sufficient indication of the various forms that may
arise.
A"i
Fig. 49.
Again, by superposition of the cases s = 3, s f = 1 and s
s' = 3, we get
. OTTX . fry . TTX . tTry
sm sm  f X sm sm
a a a a
a a ( a V a )) '
The cases X = 0, X = J, X = + 1 are shewn in Fig. 50 ;
intermediate forms are readily supplied in imagination.
A still greater variety is introduced by the fact that a
number which is the sum of two squares can sometimes be
so resolved in more than one way. For example, the modes
for which
s = 4, 7, 1, 8,
s = 7, 4, 8, 1,}
respectively, have all the same frequency.
144
DYNAMICAL THEOEY OF SOUND
Fig. 50.
54. Circular Membrane. Normal Modes.
In the case of the circular membrane we naturally have
recourse to polar coordinates, with the origin at the centre.
The differential equation may be obtained by transformation
of 52 (3), but a more direct process is preferable.
Take first the case of the symmetrical vibrations where
is a function of r, the distance from 0, only. The stress across
a circle of radius r has a resultant P . ZTTT . 9f/9r normal to the
plane of the undisturbed membrane, and the difference of the
stresses on the edges of the annulus whose inner and outer
radii are r and r + Sr gives a force
,
Equating this to p . 2?rr5r . f, which is the acceleration of
momentum of the annulus, we get
If f varies as cos (nt f e), this reduces to
where & 2 = n 2 p/P, as before.
(2)
MEMBRANES AND PLATES
145
If we assume, as is necessarily the case when the origin is
included within the region to which (2) applies, that f can be
expanded in a series of ascending powers of r, the coefficients
(after the first) may be found by substitution in (2), and we
obtain
(3)
provided
2 2 .4 2
.(4)
This is the Bessel's Function* of zero order, "of the first kind,"
which we have already met with in 31 ; it is represented
graphically in Fig. 51. If a be the radius of the boundary,
Fig. 51.
supposed fixed, the admissible values of k and thence of n are
determined by the equation
J (ka) = 0,_ ,...(5)
viz. we have
kaln = 7655, 17571, 2'7546, 3'7534, (6)
the numbers tending to the form ra J, where m is integral.
The first of these roots corresponds to the gravest of all the
normal modes of the membrane. In the rath mode there are
in 1 nodal circles, in addition to the edge, whose radii are
given by the roots of lower order. Thus in the case of the
second root we have for the nodal circle kr/7r = '7655, whence
r/a = *4356. The characters of the various normal modes will
be understood from Fig. 51, which may be taken to represent
a section through the centre, normal to the plane of the
membrane.
* F. w.
181046.
I*.
(17841846), director of the observatory at Konigsberg
10
146 DYNAMICAL THEOEY OF SOUND
The complete solution of the differential equation (2),
which is of the second order, would consist of the sum of
two definite functions of kr, each multiplied by an arbitrary
constant; but the second solution, which is called a Bessel's
Function " of the second kind," becomes infinite for r = 0, and
is therefore inapplicable to a complete circular area. In the
case of an annular membrane, however, bounded by concentric
circles, both solutions would be admissible, and both would be
required in order to satisfy the conditions at the two edges*.
The theory of the symmetrical vibrations of a circular
membrane was given by Poisson (1829), who also calculated
approximately a few of the roots of the periodequation (5).
When the vibrations are not symmetrical we may begin by
calculating the forces on a quasirectangular element of area
bounded by two radii vectores and two concentric circles, the
sides being accordingly $r and rB0. The stresses on the
curved sides give a resultant
normal to the plane, whilst the stresses on the straight sides
produce
Equating the sum of these expressions to pr$0Sr . , we obtain
p^pJil^U 8 ^ (7)
p tf (r9rV 8rJ r 2 90 2 j'
or, in the case of simpleharmonic vibrations,
with the same meaning of & 2 as before.
* On account of the frequent occurrence of the Bessel's Functions in
mathematical physics, especially in twodimensional problems, great attention
has been devoted to them by mathematicians. The difficulty in investigating
their properties is much as if we had to ascertain all the properties of the
cosinefunction from the series
and were ignorant of its connection with the circle.
MEMBRANEiS AND PLATES 147
Since f is a periodic function of 0, of period 2Tr, it can be
expanded (for any particular value of r) in a series of sines and
cosines of multiples of 0, thus
f = R Q + R l cos 6 + $ sin + ...
+ .RsCoss0f &sms0+ ..., ...(9)
by Fourier's theorem; and this formula will apply to the
whole membrane if the coefficients be regarded as functions
of r. Moreover on substitution in (8) it appears that each
term must satisfy the equation separately. Thus we have
a typical .solution
.cos(nt+e), ............ (10)
provided + i ! + *_ fi. _0. ...(11)
2 z
r
The solution of this, which is finite for r = 0, can be found in
the form of an ascending series. In the accepted notation we
have R 8 = A 8 J 8 (kr), where the function /, is defined by
This is known as the Bessel's Function of the sth order, of the
first kind. As in the case of (2) there is a second solution
which becomes infinite for r = 0, but in the case of the complete
circular membrane this of course is inadmissible. We have then
the normal modes
=AJ s (kr)coss0.cos(nt + e), ......... (13)
where k is determined by
J 9 (ka)=0 ...................... (14)
Similarly, taking a term 8 g sin sO from (9) we should have been
led to the modes
=BJ 8 (kr)sms0.cos(nt + e), ......... (15)
with the same determination of k. Owing to the equality of
periods the normal modes are to some extent indeterminate.
Thus, for any admissible value of k, we may combine (13)
and (15) in arbitrary proportions, and obtain
?= CJ, (kr) cos (80 + a) . cos (nt + e) ....... (16)
102
148
DYNAMICAL THEOKY OF SOUND
We have here s nodal diameters, given by
s0 + a= \ir, f?r, ...,
and accordingly arranged at intervals of TT/S. Again for every
value of k after the lowest we have one or more nodal circles
whose radii are given by the roots of lower order. In the case
s = l, where there is one nodal diameter, we have
ka/7r= 12197, 2'2330, 3'2383, 4'2411, (17)
the numbers tending to the form ra + J . The characters of
the corresponding modes may be gathered from the annexed
10
10
Fig. 52.
graph of the function J l (z) ; this may be supposed to represent
a section through the centre, normal to the nodal diameter. In
the second of the above modes, the radius of the nodal circle is
given by
r/a = 12197/22330 = '546.
Fig. 53 shews in plan the configuration of the nodal lines
in the first three modes of the types s = 0, 5 = 1, s 2, re
spectively. The + and signs distinguish the segments
\vhich are at any instant in opposite phases of vibration.
Whatever the form of the boundary, the value of f in the
neighbourhood of any point of a membrane must admit of
expression in the form (9), with
R 8 = A 8 J.(kr), S 8 = B 8 J 8 (kr), (18)
the factor cos (nt + e) being of course understood. If be on
a nodal line we must have f = for r = 0, and therefore A = 0.
The form of the membrane near is then given by
^ = (A 1 cosO + B 1 sme)J l (kr), (19)
ultimately, and the direction of the nodal line at is accord
MEMBRANES AND PLATES
149
ingly given by tan 6 = A l /B l . If all the coefficients of order
less than s vanish, we have, for small values of r,
(20)
The node has then s branches passing through 0, making equal
angles TT/S with one another, their directions being given by
tan s0 = AgjBg. This is illustrated in the preceding diagrams ;
for instance the cases s = 2, s = 3, s = 4 all occur in Fig. 50.
Fig. 53.
According to a general theorem stated in 16 it must be
possible by combination of the various normal modes of a
membrane in suitable proportions, and with proper relations
of phase, to represent the effect of arbitrary initial conditions.
We do not enter into this; and the theory of the forced vibrations
must also be passed over except fora simple example.
150 DYNAMICAL THEORY OF SOUND
When a force Z per unit area acts on a circular membrane,
the equation (1) is replaced by
it being supposed, for simplicity, that there is symmetry as
regards the distribution of Z and the consequent displacements
If, further, Z vary as cos (pt + a), we have
Z
(23)
If Z be independent of r, so that the impressed force is
uniform over the membrane, the solution of (22) is obviously
t=lp + CJ t (kr), ............... (24)
and determining the constant C so that f =0 for r = a, we find
(25)
(ka)
The amplitude becomes very great whenever fca approximates
to a root of (5), i.e. whenever the imposed frequency approaches
that of one of the symmetrical free modes. When, on the
other hand, the imposed vibration is relatively slow, ka is
small, and we have by (4)
(26)
approximately. This is the statical deflection corresponding to
the instantaneous value of the disturbing force.
55. Uniform Flexure of a Plate.
The theory of the transverse vibrations of plates stands in
the same relation to that of bars as the theory of membranes
to that of strings. The reader will understand from this com
parison that the mathematical difficulties are considerable, arid
will not be surprised to learn that some of the most interesting
and, at first sight, simple problems remain unsolved. On the
other hand the subject readily admits of experimental illustra
tion. If the plate be horizontal, and fixed at one point, the
MEMBRANES AND PLATES 151
configuration of the nodal lines can be exhibited by means
of a little sand previously strewn on the surface. When any
particular normal mode is excited, the sand is shaken away
from the places of vigorous motion, and accumulates in the
neighbourhood of the nodal lines. Usually the plate is set
into vibration by bowing at right angles to the edge, and the
desired mode is favoured by touching the edge with the fingers
at one or more nodal points. If, as in the case of a rectangular
plate fixed at the centre, the point of support is a nodal point
>f several normal modes, a great variety of beautiful figures
nay be obtained. An extensive series of diagrams of results
obtained in this way were given by Chladni; many of these are
eproduced in the current manuals of experimental acoustics.
In the theoretical treatment it is assumed that one of the
principal axes of strain and stress is normal to the faces of the
plate, and that the corresponding stress vanishes. Putting,
then, p s = in the formulae (9) of 42, we find, for the
remaining principal stresses,
)> p> = E'( > + a ei ) t ......... (1)
where E f = E/(l  <r 2 ) ...................... (2)
If R l , R 2 be the principal radii of curvature at any point of the
plate, when bent, we have, by an investigation similar to that
of 45,
C^Z/R!, e 2 = z/R 2 , .................. (3)
where z denotes distance from the medial plane. If we consider
a rectangular element of the plate bounded by lines of curvature,
and denote by h the halfthickness, this leads to bending
moments
per unit length of the respective edges, in the planes of the two
principal curvatures.
The potential energy per unit volume is
ef) .......... (5)
152 DYNAMICAL THEOBY OF SOUND
If we substitute from (3), and integrate over the thickness, we
find for the potential energy per unit area of the plate
The formulae (4) may be applied to the case of a flat bar of
rectangular section, uniformly bent by two opposing couples
MJ), where b denotes the breadth. Along the free edges we
have M 2 = 0, and therefore
JZr' = rJZr' (7)
The bending moment is accordingly
M 1 b=%Ebh 3 /R l) (8)
by (4). This agrees, as it must, with 45 (5), since o> = 2bh,
K * == *h?. The formula (7) shews that when a bar of rectangular
section is bent in a plane parallel to one pair of faces, an opposite
or " anticlastic " cur
vature is produced in
the plane of the cross
section, the ratio of
the curvatures being
identical with Pois
son's ratio &. This
circumstance has
been made the basis
of practical methods Fig. 54.
of determining <r, by
Cornu* (1869) and Mallock (1879), the curvatures being
measured by optical or other means.
It follows from the above that a perfectly free rectangular
plate cannot vibrate after the manner of a bar, with nodal lines
parallel to one pair of opposite edges, since couples would be
required, about the remaining edges, to counteract the tendency
to anticlastic curvature.
56. Vibrations of a Plate. General Results.
In a vibrating plate the directions and amounts of the
principal curvatures will in general vary from point to point.
* A. M. Cornu (1841 1902), professor of physics at the Ecole Polytechnique
1871 1902. Famous for his experimental determination of the velocity of light,
and for other important contributions to optics.
MEMBEANES AND PLATES 153
Shearing forces will also be called into play normal to the plane
of the plate. The circumstances are somewhat complicated,
but the deduction of the equation of motion for the body of the
plate is a straightforward matter, and presents no real difficulty.
A more serious question arises when we come to the conditions
to be satisfied at a free edge. It appears that the simple
condition of strain which has been postulated as the basis of
the formulae (4) of 55 cannot be assumed to hold, even
approximately, right up to the edge. In the immediate neigh
bourhood of the edge, i.e. to a distance inwards comparable
with the thickness, a peculiar state of strain in general exists,
one remarkable result of which is a shearing force on sections
perpendicular to the edge, of quite abnormal amount.
For the further development of the subject reference must
be made to other works*. We merely quote a few of the more
important results which have been obtained, relating chiefly to
plates whose edges are free.
It is found that for a plate of given lateral dimensions the
frequency (n/2?r) of any particular normal mode is given by
*'.#.*, (1)
where, as in 46, m is a constant, of the nature of the reciprocal
of a line, given by a certain transcendental equation, and p
denotes the volumedensity. For plates with geometrically
similar boundaries the frequency accordingly varies as the
thickness, and inversely as the square of the lateral dimensions.
In the case of a perfectly free circular disk the nodal lines
are circles and equidistant
diameters. In the symmetrical
modes, which were investigated
to some extent by Poisson
(1829), we have nodal circles
alone. Thus in the gravest
mode of this type we have a
nodal circle of radius '678a, where a is the radius of the disk ;
in the next mode there are two nodal circles of radii '39 2a
* See Lord Kayleigh. Theory of Sound, chap. 10 ; Love, Theory of Elasticity,
Cambridge, 1906, chap. 22.
154
DYNAMICAL THEORY OF SOUND
and '842a, and so on, the numbers varying slightly however
with the value adopted for or. According to Poisson, the
values of m for the above modes are given by
ra 2 a 2 = 88897, 3836, (2)
on the hypothesis that a = J.
The complete theory of the free circular plate was worked
out by Kirchhoff in a celebrated memoir (1850). It appears
that the gravest of all the normal modes has two nodal
diameters, and no nodal circle. Its frequency is
5234. /(
according as we adopt the value cr \ or a = % of Poisson's
ratio. The figure shews the
configuration of the nodal lines
in the simplest cases of one and
two nodal diameters.
The theory of a circular
plate clamped at the edge has
been treated by Poisson and
others. In the first two symmetrical modes it is found that
ra 2 a 2 = 102156, 39'59, (3)
respectively. In the second of these modes there is a nodal
circle of radius '38 la. The theory has been applied by Lord
Rayleigh to calculate the natural frequencies of a telephone
plate.
Fig. 56.
Fig. 57.
In the case of a square plate we have to depend almost
wholly on observation, there being at present no exact theory.
As in the case of the square membrane ( 53), the nodal lines
MEMBRANES AND PLATES 155
may assume a great variety of forms, owing to the superposition
of different modes having the same frequency. The gravest
mode of a free plate is that in which the nodal lines form a
cross through the centre, with arms parallel to the sides.
The figure shews other cases in which possible forms can be
assigned to the nodal lines from considerations of symmetry.
57. Vibrations of Curved Shells.
When we proceed to the vibrations of curved plates, or
shells, we meet with further complications due to the fact that
no absolutely sharp line can be drawn between flexural and
extensional modes. This has been already exemplified in the
case of the ring ( 51). It appears, however, that as the thick
ness is (in imagination) reduced the normal modes tend to fall
into two distinct categories. In one of these the frequencies
tend to definite limits, the deformations being mainly
extensional, and so analogous to the longitudinal vibrations
of a bar, where the dimensions of the crosssection were found
to be immaterial. In the second category the frequencies
diminish without limit, being ultimately proportional to the
thickness, as in the flexural vibrations of a bar or plate.
It will be understood that, acoustically, the flexural vibra
tions are alone of real interest. When the shape is one of
revolution about an axis, the nodal lines will evidently be
parallels of latitude and equidistant meridians. As in the case
of 51 these are not lines of absolute rest, the tangential motion
being there relatively at its greatest. This has an application
to bells. A theoretical calculation of the frequencies of an
actual bell is of course out of the question ; but it is somewhat
remarkable that no systematic experimental study appears to
have been made until the subject was taken up by Lord
Rayleigh in 1890. Some unexpected results were obtained.
To quote a typical case, the normal modes of a particular bell,
when arranged in ascending order of frequency, were found to
have the following numbers of nodal meridians and parallels,
and the pitches indicated :
(4,0) (4,1) (6,?) (6,?) (8,?)
e c" f" + &"(, d'".
156 DYNAMICAL THEORY OF SOUND
Of these the only one which has any relation to the nominal
pitch (d"} of the bell is the fifth in order, and this is out by an
octave. A mistake of an octave in judging pitch is not
uncommon, for physiological reasons, but it is surprising that
the presence of the lower dissonant tones should be so easily
disregarded. It is conceivable that the mode of striking may
be in some degree unfavourable to the production of the more
discordant elements.
The vibrations of an elastic solid whose dimensions are all
of the same order of magnitude are from our present point of
view of subordinate interest. The only case which has been
worked out is that of the sphere. In the most important
mode one diameter extends and contracts whilst the perpen
dicular diameters simultaneously contract and expand, respec
tively. The frequency of this mode is, for such values of a as
are commonly met with,
(1)
about, where a is the radius. This is the lowest of all the
natural frequencies. For a steel ball one centimetre in radius,
this makes N= 136000.
CHAPTER VI
PLANE WAVES OF SOUND
58. Elasticity of Gases.
In any fluid there is a definite relation between the pressure
p, the density p, and the temperature 0, and any two of these
quantities accordingly serve to specify the physical state of
the substance. It is often convenient to use in place of p its
reciprocal v, the volume of unit mass.
In thermodynamical investigations the two quantities
usually chosen as independent variables are p and v. In
Watt's " indicator diagram " these are taken as rectangular
coordinates, p being the ordinate and v the abscissa. Any
particular state is then represented by a point on the diagram,
and any succession of states by a continuous line. We may
imagine the unit mass of the fluid to be enclosed in a deform
able envelope, and that an infinitesimal change of volume is
produced by a displacement of the boundaiy in the direction
of the normal, whose amount is (say) v for any given surface
element 8S. The work done by the contained gas in this
process is 2 (p&S . v), or pSv, since 2 (v8S) = Sv. Hence the
work done in any succession of changes, represented by a curve
on the diagram, will be given by fpdv, i.e. by the area included
between the curve, the axis of abscissae, and the first and last
ordinates. This area is of course to be taken with its proper
sign, according as the work is positive or negative.
There are two kinds of successions of states which are
specially important. In the first of these the temperature does
not vary, and the representative lines are therefore called
" isothermals." By means of a system of isothermal lines
drawn at sufficiently small intervals the properties of the
158 DYNAMICAL THEOKY OF SOUND
substance can be completely mapped out. The other suc
cessions referred to are those in which there is no gain or loss
of heat to the substance, as if it were enclosed in a vessel (o:
variable volume) whose walls are absolute nonconductors. The
corresponding lines are therefore called " adiabatics."
In a perfect gas we have
p = R P 0, or pv = R0, ............... (1)
where 6 is the absolute temperature on the gas thermometer
and R is a constant depending on the nature of the gas. The
isothermal lines pv = const, are therefore rectangular hyperbolas
asymptotic to the coordinate axes. As regards the adiabatics
the heat required to increase the pressure by &p when the
volume is constant will be given by an expression of the form
PSp. If c denote the specific heat (per unit mass) at constant
volume, this must be equal to c&0, where $0 is the corre
sponding change of temperature. Now when $v = we have
Sp/p = $0/0, whence, comparing, P = c0/p. Again, the heat
required to' augment the volume by Iv when the pressure is
constant may be denoted by QSv, which must be equal to c'&0
where c' is the specific heat at constant pressure. Since, when
fy = we have &v/v = $0/0, we find Q = c0/v. The heat ab
sorbed when both pressure and volume are varied infinitesimally
is therefore
............... (2)
and the differential equation of the adiabatics is therefore
$M* = 0. ...(3)
p c v
The ratio c'/c. of the two specific heats is practically constant.
Denoting it by 7, .we have
\ogp 4 7 log v = const.,
or pv? = const., ........................ (4)
as the equation of the adiabatic lines. The value of 7 as found
by direct experiment is about 1*41 for air, oxygen, nitrogen,
and hydrogen. The figure shews the isothermal and adiabatic
lines through a point P of the diagram, the latter curve being
the steeper.
PLANE WAVES OF SOUND
159
When the pressure and volume vary in any connected
manner, the ratio of
the increment &p of
the pressure to the
" compression," i.e. the
negative dilatation
Sv/v, may be called
the " elasticity of
volume." Its value
will depend not only
on the particular state,
but on the manner in
which the variations
from that state are
supposed to take place,
i.e. on the direction
of the corresponding
p.
curve on the diagram.
If the tangent at the
point P meet the axis of p in U, and NU be the projection of
P U on this axis, we have
dv
.(5)
this projection therefore represents the elasticity under the
particular condition. On the isothermal hypothesis, to which
these letters refer in the figure, the elasticity is equal to the
pressure p, as follows at once from (1), or from the fact that
the tangent to a rectangular hyperbola is bisected at the
point of contact. If the variations are subject to the adiabatic
law, the elasticity, as deduced from (4), is yp, and so greater
than in the former case. This is represented by NU' in the
figure. Even in the case of solid and liquid bodies we ought,
in strictness, to discriminate between isothermal and adiabatic
coefficients of elasticity, but the differences happen not to be
very important.
The work done by unit mass of a gas in expanding between
any two adjacent states is easily read off from a diagram as
ov), or Po(v v) + %(pp )(v v), (6)
160
DYNAMICAL THEORY OF SOUND
correct to the second order of small quantities.
states are a finite distance apart we
require, to know the manner of transi
tion. For changes along an isothermal
line pv = p v we have
When the two
dv=p Q v \og ...(7)
For variations along an adiabatic
V %
Fig. 59.
59. Plane Waves. Velocity of Sound.
The theory of plane waves of sound is very similar to that
of the longitudinal vibrations of rods ( 43). We assume that
the motion is everywhere parallel to the axis of x, and is the
same at any given instant over any plane perpendicular to this
axis. We denote displacement from the equilibrium position
by f . The symbols p, p, % are supposed to refer at the time
t to that plane of particles whose undisturbed position is
x] they are therefore functions of the independent variables
x and t. The constant equilibrium values of p, p are dis
tinguished as PQ, p .
The dilatation A was defined in 40 as the ratio of the
increment of volume to the original volume, viz.
(1)
In the present branch of the subject it is usual to introduce
a symbol s to denote the " condensation," i.e. the ratio of the
increment of density to the original density ; thus
(2)
Since v=l/p, we have
.(3)
The stratum of air which was originally bounded by the
planes x and#+& is at time t bounded by the planes x+% and
PLANE WAVES OF SOUND 161
x + f t &B + Sf , and its thickness is therefore changed from 8x
to &c f 8f , or (1 + 9f/9#) 8#, and the dilatation is accordingly
 ' A  *= ...................... <*>
Hence, in the case of infinitely small disturbances, we have,
by(3)>
In forming the equation of motion we assume that the
pressure varies with the density according to some definite law.
We have then, for small values of s,
p=p Q + KS, ........................ (6)
where K is a coefficient of cubic elasticity. t Considering the
acceleration of momentum of unit area of a stratum originally
bounded by the planes x and x f &c, we have
where 8p represents the excess of pressure on the anterior face.
Hence, by (5) and (6),
***** (7)
dP'^da?'
where c = V(*/?o) ......................... (8)
The solution of (7) is as in 23, 43
%=f(ctx) + F(ct + x), ............... (9)
and represents two systems of waves travelling in opposite
directions with the velocity c*.
If we assume, as Newton f did, that, the expansions and
contractions of a gas, as a soundwave passes, take place
isothermally, i.e. without variation of temperature, the relation
between p and p is given by Boyle's law, viz. p/p Q = p/p = 1 4 s,
whence K=p , as already proved. This makes
Now for air at C. we may put, as corresponding values,
p = 76 x 1360 x 981, p 9 = '00129,
* The analytical theory of plane waves of sound is due to Euler (1747) and
Lagrange (1759).
t The investigation is given in Prop. 48 of the second book of the Principia
(1726).
L. 11
162 DYNAMICAL THEORY OF SOUND
in absolute centimetregrammesecond units, whence c = 280
metres per second. This is considerably less than the observed
velocity.
The discrepancy was first fully accounted for by Laplace
and Poisson*. When a gas is rarefied or condensed the
temperature tends to fall or rise, except in so far as the
process is mitigated by the supply or abstraction of heat. In
ordinary soundwaves the condensation s changes sign so fre
quently, and the temperature consequently rises and falls so
rapidly, that there is no time for sensible transfer of heat
between adjacent portions of the gas. The flow of heat has
hardly set in from one element to another before its direction is
reversed, and the conditions are therefore practically adiabatic.
The formula
becomes, for small values of $,
p=p Q (I+ys), ..................... (12)
whence K = yp 0) as in 58, and
c = V(7po/Po) ...................... (13)
Putting 7=1*41 we find that the Newtonian velocity of sound
in air must be increased in the ratio 1'187, whence c = 332
metres per second at 0C. This is in good agreement with
direct observation.
As there is now no question as to the soundness of this
explanation, and as the direct determination of 7 is a matter
of considerable difficulty, the formula (13) is often used in the
inverse manner, as a means of deducing the value of 7 for
various gases from the observed velocities of soundwaves in
them. For example, it was in this way that in 1895 the value
of 7 for the newly discovered gas argon was found by Lord
Rayleigh to lie between 1'6 and 17. The experimental method
(due to Kundt) is referred to in 62 below.
Since p /p = R6 , the velocity of sound as given by (13) is
independent of the actual density, but will vary as the square
root of the absolute temperature. Also, so far as 7 has the
same value, the velocity of sound in different gases will vary
* About, or before, the year 1807.
PLANE WAVES OF SOUND 163
inversely as the square root of the density, provided the com
parison be made at the same pressure. These conclusions are
in agreement with observation.
The formula (8) will of course apply to any fluid medium,
provided the proper value of K be taken. In liquids the
difference between the isothermal and adiabatic elasticities
3 may be neglected. For water at 15 C. we may put
K = 222 x 10 10 , p = l, in C.G.S. units, whence c = 1490 metres
; per second. The number found by Colladon and Sturm (1826)
I by direct observation, in the water of the lake of Geneva, was
! 1435, at a temperature of about 8 C.
Another formula for the velocity of sound may be noticed.
If H denote the height of a " homogeneous atmosphere," i.e. of
bi column of uniform density p Q whose weight would produce the
actual pressure p per unit area, we have p = gpoH, and the
Newtonian formula (10) becomes
c = V(<7#); ..................... (14)
cf. 43 (6). The velocity is accordingly that due to a fall from
rest through a height %H. It appears from 58 (1) that for
a given gas, and at a given place, H depends only on the
temperature. The corresponding adiabatic formula is
(15)
60. Energy of Sound Waves.
The kinetic energy of a system of plane waves is, per unit
area of the wavefronts,
where the integration extends over the space which was occupied
by the disturbed air in the equilibrium state.
The work done by unit mass in expanding through a small
range was found in 58 to be given accurately, to the second
order, by the expression
where the suffix refers to the final state. If we form the sum
of the corresponding quantities for all the masselements of the
system, the first term disappears whenever the conditions are
such that the total change of volume is zero. Again, in the
112
164 DYNAMICAL THEORY OF SOUND
second term we may put, with sufficient accuracy, pp = KS,
V Q V = V O S, and obtain ^KS Z .V O . The expression %/cs* is there
fore to be integrated over the volume occupied in the undis
turbed state. So far nothing is stipulated as to the hypothesis
to which K relates; but it is only in the case of adiabatic
expansion that the result can be identified with the potential
energy in the strict sense of this term. We then have
V = ^ K fs 2 dx } ..................... (3)
where K = <yp 0) per unit area of wavefront. If K refer to the
isothermal condition, the expression on the right hand is what
is known in thermodynamics as the " free energy."
It is unnecessary to repeat what has been said in 23 as to
the resolution of an arbitrary initial disturbance into two
wavesystems travelling in opposite directions. In a single
progressive wavesystem, say
?=/((*), ..................... (4)
we have by 59 (5) = cs, ........................... (5)
where denotes the particle velocity in the direction of propa
gation. Since f has the same sign as s, an airparticle moves
forwards (i.e. with the waves) as a phase of condensation passes
it, and backwards during a rarefaction. It appears moreover,
from (1), (3), and (5), that the total energy is half kinetic
and half potential. This also follows independently from the
general argument given in 23.
The case of a simpleharmonic train of progressive waves is
specially important. The formula
.................. (6)
represents a train of amplitude a, frequency n/2?r, and wave
length X = 27rc/ft. We find
n (t  ]
sin 2 n t   dx
(7)
The mean value of the second term under the integral sign is
zero, and the average kinetic energy per unit volume is therefore
and the average value of the total energy
PLANE WAVES OF SOUND 165
Since iia is the maximum particlevelocity, we see that the
energy in any region including an exact number of wavelengths
is the same as the kinetic energy of the whole mass when
animated with the maximum velocity of the airparticles. If s l
be used to denote the maximum condensation, we have 81 = na/c,
and the average energy per unit volume may therefore also be
expressed by ^potfsf.
We can also estimate, incidentally, the nature of the approxi
mation involved in the derivation of the equation of motion 59
(7). The approximation consisted in neglecting the square of s,
or 9f/9a?. Since 81 = 27ra/\, this means that the amplitude a is
assumed to be small compared with X/27T, a condition which is
abundantly fulfilled in all ordinary soundwaves.
So far we have traced the course of waves regarded as
already existent, without any reference to their origin. As an
example, though a somewhat artificial one, of the manner in
which waves may be supposed to be generated, imagine a long
straight tube, of sectional area , in which a piston is made
to move to and fro through a small range, in any arbitrary
manner. The origin of x being taken at the mean position
of the piston, the forced waves in the tube, to the right,
due to a prescribed motion
?=/(*) (8)
of the piston, will evidently be given by
In particular, if % = acosnt, (10)
we have = acosnU ) (11)
The rate at which work is being done by the piston on the au
to the right is
4f/i 2/Y 2
= p 9 <ona sm nt \  o>sm*nt. ......(12)
c
The mean value of the first term is zero, whilst that of the
second is
(13)
166 DYNAMICAL THEORY OF SOUND
This is exactly the mean energy contained in a volume we of
the space occupied by the wavetrain (11). The result may
perhaps at first sight appear to be a mere truism. It may be
argued that in each unit of time fresh waves are generated
which occupy a length c of the tube, and that the piston must
as a matter of course supply the corresponding amount of
energy. It must be remembered, however, that an infinitely
long train of waves of the type (11) would take an infinite time
to establish, and that in the case of a finite train the suggested
line of argument would require us to examine into what is
taking place at its front. In the present instance the result
would, it is true, be unaffected, but the case would be altered
if the wavevelocity were different for different wavelengths,
as it is for example in dispersive media in optics, in deepwater
waves in hydrodynamics, and in the case of flexural waves on a
long straight bar ( 45). There is then a distinction between
the wave velocity (for a particular wavelength) and the "group
velocity" which determines the rate of propagation of energy.
In the above problem force must be applied to the piston in
order to maintain the vibration (8) against the reaction of the
air. If the piston be free, the store of energy which it origin
ally possessed will be gradually used up in the generation of
airwaves. Suppose, for example, that the piston is attached
to a spring, and that in the absence of the air the period of its
free vibrations would be ^TT/U. Under the actual conditions,
its equation of motion will be of the form
M(Z + n^) = (p Po )co, (14)
where the variable part of the pressure alone appears, since the
constant part merely affects the equilibrium position. From
the general theory of progressive waves we have
pp^KS^tcg/c, (15)
and the equation (14) becomes
5 + ?+^ = o (16)
This is of the form discussed in 11, and the solution is
f .=(7<?'/ T cos(rc'+ e), (17)
PLANE WAVES OF SOUND 167
provided T = ZMc/tca = ^Mjp^c, n' z = n*l/r*. i ..... (18)
When nr is large the effect on the period may be neglected.
The condition for this is that 2M* must be large compared
with p o>X/27r, where \ is the wavelength. The inertia of the
piston must therefore be great compared with that of the air
contained in a length X/2n of the tube. The same law of
decay would be given also by the indirect method explained
in 12.
We have seen in (13) that the rate of propagation of energy
across unit area of wavefront in a progressive system of waves
of simpleharmonic type is Jp ri 2 a a c, or J/^oC 3 ^ 2 , if s l denote the
maximum condensation. The result was obtained for plane
waves, but will hold for all kinds of wave at a sufficient distance
from the source. Consequently if W denote the total emission
of sonorous energy per second from a source near the ground,
the value of s a , at a distance r, will be given by the relation
W =i/3 c 3 5 1 2 x 27rr 2 = 7rp c 3 r 2 5 1 2 .......... (19)
This formula was applied by Lord Rayleigh to estimate the
limit of audibility of a sound of given pitch. The value of W,
as inferred from the power spent in actuating the source
(a whistle), is the product of the current into the pressure, and if
r be the distance at which the sound is just audible, the formula
will give a value of s lf which is necessarily, however, greater
than the true limit, since the value of W is too high, not all
the energy being spent in sound. In this way it was ascer
tained that sounds could be heard in which Sj_ was certainly less
than 4 x 10~ 8 . The corresponding amplitude as deduced from
the formula ncu cs^ was 8 x 10~ 8 cm. By an independent
method, in which the above source of uncertainty was avoided,
the limit of audibility was fixed at about s l = 6 x 10" 9 . Subse
quent experiments by Wien (1903) and Rayleigh f indicate an
increase of sensitiveness with rise of pitch, for tones near the
middle of the ordinary musical scale.
* The factor 2 would disappear if the piston were supposed to generate waves
on both sides.
t Phil. Mag. (6), vol. xiv. (1907).
168 DYNAMICAL THEORY OF SOUND
61. Reflection.
When there is a fixed barrier at the origin the general
solution is replaced, as in 24, by
g = f(cta;)f(ct + x) (1)
Considering, for example, the region to the left of the origin,
the first term may be interpreted as representing a primary
wavesystem approaching the barrier; the second term then
represents the reflected system. The latter has the same
amplitude at corresponding points ; the velocity j is reversed,
but the condensation s (= di~/dx) has its sign unchanged. We
have here, in its simplest form, the explanation of echoes.
There is another case of reflection which it is important to
consider. Suppose that at one point (say x = 0) the condition
of unvarying pressure (s = 0) is imposed. We must have then,
in 59 (9),
F'(et)=f(ct), (2)
which shews that the functions f, F must differ only by
a constant. Since this constant would merely represent a
displacement common to the whole mass, which is without
influence on the question, it may be ignored. We have then
f = f(ct"x)+f(ct + x) t (3)
where as before the first term may be taken to represent an
incident, and the second a reflected wavesystem, in the region
lying to the left of 0. The velocity f is here reflected un
changed, but the sign of s is reversed. The conditions would
be realized if the air were in contact at the plane as = with
a medium capable of exerting pressure, but destitute of inertia.
This is of course an ideal case, but the condition of invariable
pressure is approximated to in some degree at the open end of
a pipe. The present investigation has also an application to
the reflection of longitudinal waves at the free end of a rod
( 43).
The general problem of (direct) reflection at the common
boundary of two distinct fluid media is hardly more complicated.
The origin being taken in the boundary, a wavesystem ap
proaching from the left will give rise to a reflected wave on the
left and a transmitted wave on the right. We distinguish
PLANE WAVES OF SOUND 169
quantities relating to the incident and reflected wave by the
suffixes 1 and 2, respectively, whilst those relating to the
transmitted wave are indicated by (grave) accents. Since the
velocity and the pressure must be the same for the two media
at the origin, we have
?i + ? 2 = i\ *s 1 + *s 2 = *Y [>=0], ......... (4)
the equilibrium pressure p being necessarily the same. Now
^ = 0$!, J 2 = cs 2 , c> whence
c(s l s 2 ) = cs\ ic (! + * 2 ) = *V [# = 0] ....... (5)
^
Hence S "
These formulae relate in the first instance to the state of
things at the origin, on the two sides; but it is easily seen that
they will also represent the ratios of amplitudes at correspond
ing points in the respective waves. If the inertia of the second
medium were infinite, we should have c = 0, and therefore
$2 = $!, as in the case of reflection at a rigid barrier. On the
other hand, if the inertia of the second medium were evanescent,
we should have c = oo and s 2 = i, as above.
The energies of corresponding portions of the various waves
are proportional to KS^C, KSC, #W, since the lengths occupied
by these portions will vary as the respective wavevelocities.
The conservation of energy therefore requires
KSi*C = KS 2 *C + KS"*C', .................. (7)
this is easily verified from (6).
If we put K = p c z , K = poC*, we have, from (6),
*i A> c + p c
As an example, take the case of air waves incident normally
on the surface of water. We have p /p x = '00129, c/c' = '222,
about; whence s z /s l = '99943. There is therefore almost com
plete reflection, with hardly any transmission.
In the case of two gaseous media having the same ratio of
specific heats, and therefore the same elasticity (K = yp ), the
formulae simplify ; thus
* = ^, S =^ (9)
S, C + C 5 X C + C
170 DYNAMICAL THEOKY OF SOUND
These are identical with Fresnel's formulae for the amplitudes
of reflected and transmitted light in the case of normal inci
dence on the common boundary of two transparent media.
62. Vibrations of a Column of Air.
When we come to the free oscillations of the air contained
in a pipe of finite length, the question definitely arises as to the
condition to be satisfied at an open end. There is here a
transition, more or less rapid, from plane waves in the tube
to diverging spherical waves in the external space, which it is
difficult to allow for exactly. In the usual rudimentary theory,
which dates from D. Bernoulli, Euler, and Lagrange, it is
assumed that the variation of pressure in the tube, at the open
end, may be neglected. As already stated, this would be
accurately the case if the external air were replaced by a
substance capable of exerting pressure (p ) but devoid of
inertia. There would then be no loss of energy on reflection
at the open end ( 61), and the vibrations in the tube, once
excited, would be persistent. The hypothesis is obviously
an imperfect representation of the facts ; the condition s =
can only be approximately fulfilled, and energy must con
tinually be spent in the generation of waves diverging outwards
from the mouth, so that the vibrations if left to themselves will
be sensible only for a very limited time ; this may however
cover hundreds of periods. We shall return to these questions
later (Chapter IX) ; at present we content ourselves with
tracing out the consequences of the approximate theory.
The periodic character of the motion in a finite pipe can be
inferred from the theory of waves, exactly as in the case of
strings ( 24). Suppose for example that a wave ,of limited
extent is started in either direction from a point P of a tube
AB. After two reflections, at A and B, the wave will pass P
again in the same direction as at first. If both ends be closed,
the sign of s is unaltered at either reflection, whilst that of j is
twice reversed. Hence after the interval 2l/c, where l AB,
the initial circumstances are exactly reproduced. The same
result holds if both ends be open, since there have now been
PLANE WAVES OF SOUND 171
two reversals of s and none of f in the interval in question.
But if one end be closed and the other open, the signs of s and
f at P have each undergone one reversal only in the interval
2l/c, and a further interval of like duration must elapse before
the original state of things at P is restored.
The foregoing theory explains one or two important points
in the theory of organpipes. Thus the frequency, in the
gravest mode, is inversely proportional to the length, and is
lower by an octave for a "stopped" pipe, i.e. a pipe closed
at one end, than for an " open " pipe, i.e. one open at both ends,
of the same length. It is, again, directly proportional to the
velocity of sound, and so increases with rise of temperature.
In the analytical method for determining the normal modes
we assume as usual that f varies as cos (nt + e). The equation
59 (7) then becomes
the solution of which is
/ . nx D . nx\ /ON
t=\A cos h B sin cos (nt + e), (2)
\ c c J
as in 25. The corresponding wavelength of progressive waves
in free air is X = 2?rc/n. Hence in any system of standing waves
there is a series of nodes ( = 0) at intervals of X, and a series
of loops, or places of zero condensation, (df/dx = 0), halfway
between these.
For a tube closed at both ends (x = 0, x = I) we have
4=0, sin(y/c) = 0, (3)
and therefore
~ . rmrx frmrct \
I f = w sm j cos!^ + m ), (4)
\ 6 /
where m 1, 2, 3, ..., the normal modes forming a harmonic
series.
For a pipe open at both ends, the condition that s=d^/dxQ
for x = and x = I gives
= 0, sin(^/c) = 0, (5)
and the typical solution is
n mirx frmrct \
= G m cos j cos j + e m , (6)
t \ 6 /
172 DYNAMICAL THEORY OF SOUND
where ra = l, 2, 3, .... Here, again, the sequence of normal
modes is harmonic. The figure illustrates the cases m=l,
m = 2. The arrows shew the direction of motion at the loops,
whose position is indicated by the dotted lines, in two opposite
phases ; the nodes are indicated by the full transverse lines.
Fig. 60.
In the case of a pipe closed at # = and open at x=l,
we have
.4=0, cos(ttZ/c) = 0, .................. (7)
whence nl/c ^mjr, the integers m being odd. We thus obtain
~ mirx (rmrct
=
where m= 1, 3, 5, .... The absence of the harmonics of even
Fig. 61.
order determines the characteristic " quality " of stopped pipes
( 91). The figure shews the cases m = 1, m = 3.
The formula (2) can be applied also to the case of forced
PLANE WAVES OF SOUND 173
vibrations of given frequency (n/27r). Thus if a prescribed
vibration
f = J. cos(nt + e) .................. (9)
be maintained at x = 0, and if the tube be closed at x = I, the
motion of the gas is given by
sin  cos (nt + e). . . .(10)
c
The amplitude becomes abnormally great, even when we take
account of dissipative forces, if sin (nl/c) = 0, or I = %m\, where
m is integral. This is the principle of a method due to Kundt
(1868) by which the velocity of sound in various gases can be
compared by smallscale experiments. The wavelengths are
found by measuring the distances between the nodes, whose
position is indicated by the heaping up of lycopodium powder
previously scattered in the tube. The vibrations are excited
in the two tubes (containing the two gases to be compared) by
disks fitted to the two ends of a longitudinally vibrating rod.
If the end x= I is open, the formula (10) is replaced by
g= ^008^=^ 008 (lit + 6), ,..(11)
cos (nl/c) c
and the condition of strongest resonance is cos (nl/c) = 0, or
/ = Jm\, where m is an odd integer.
The preceding investigations would apply also to the
vibrations of a column of water, or other liquid, contained
in a tube, provided the material of the tube were absolutely
rigid. In practice, however, the yielding of the walls has
an appreciable effect; the potential energy corresponding to
a given strain (dg/dx) of the fluid is diminished, and the wave
velocity is lowered. The fact was observed by Wertheim (1847),
but the true explanation is due to Helmholtz (1848). The
question has been further investigated by Korteweg (1878)
and the present writer. Owing to the much greater velocities
( 44) of elastic waves in solids such as glass or steel, as
compared with the sound velocity in water, the stresses in
the walls adjust themselves so rapidly that it is legitimate
to assume that the deformation of the tube has the statical
value corresponding to the instantaneous distribution of
174 DYNAMICAL THEOEY OF SOUND
pressure in the liquid. If c be the theoretical velocity of
sound in the liquid, as given by 59 (8), c the actual velocity,
it is found that in the case of a tube of small thickness h
...(12)
where a is the internal radius, K is the volumeelasticity of
the liquid, and E is the value of Young's modulus for the
material of the tube. Thus in the case of water (K = 2'22 x 10 10 )
contained in a glass tube (^=6*03 x 10 11 ) whose thickness is
onetenth of the radius, we find c '759c . Even in the other
extreme, when the walls are very thick, it is found that
where /A is the rigidity. The value of /z, for glass is, roughly,
about 10 times the value of K for water; this would give
a diminution of about 5 per cent, in the wavevelocity.
63. Waves of Finite Amplitude.
The laws of sound propagation, as they are investigated in
this and succeeding chapters, are subject to some qualifications
which may best be considered in relation to plane waves, where
the theory is simplest.
In the first place, it has been assumed that the conden
sation s may be treated as infinitely small. This hypothesis
is adequate for most purposes, but there are certain "second
order " effects which are of some theoretical importance.
It is easy to shew that a progressive wave of finite (as
distinguished from infinitely small) amplitude cannot be pro
pagated without change of type, except on the hypothesis
of a certain special relation between pressure and density.
Assuming, for a moment, that a wave of permanent type is
in progress, we may in imagination impress on the whole
mass of air a velocity equal and opposite to that of the
wave. In this way we obtain a condition of " steady motion "
as it is called, in which the velocity, pressure, and density at
any point of space are constant with respect to the time.
For definiteness we may fix our attention on the air contained
in a long straight tube of unit sectional area. The velocity u
PLANE WAVES OF SOUND 175
being now a function of the spacecoordinate x only, the
acceleration of the airparticles will be given by udu/dx as
in ordinary dynamics. Hence, considering the acceleration of
momentum of the mass which at the instant considered lies
between the planes x and x + 8x, we have
du dp
pu r = / ...................... (1)
dx dx
Also, since the same amount of matter crosses each section in
unit time, we have
pu = const. =??i, ..................... (2)
say. Hence mdu/dx dp/dx, and
p= C mu, ........................ (3)
or pp = m(u u) = m* , ......... (4)
\Po pJ
where the symbols p Q , p , u refer to the parts of the medium
which in the original form of the question were undisturbed.
This gives the special relation referred to. In terms of the
volume per unit mass we have
pp =m*(v Q v), .................. (5)
which is the equation of a straight line on the indicator
diagram. A relation of this type does not hold for any
known substance, whether under the adiabatic or the iso
thermal condition, and could in any case only apply to a
limited range, since the volume would otherwise shrink to
nothing under a certain finite pressure.
If, however, the range of density be small, the equation (5)
can be identified with 59 (6) provided m z =Kp . Since m=p ^o>
where u is the wave velocity in the original form of the problem,
this gives u Q 2 = ic/p , in agreement with 59 (8). The process
is equivalent to choosing m so that the straight line (5) shall
be a tangent at the point (v , p ) to the curve which on the
indicator diagram gives the effective relation between p and v.
The condition (5) was obtained in different ways by
Earnshaw (1860) and Rankine* (1870).
To ascertain the character of the continual change of type
* W. J. M. Rankine (1820 72), professor of engineering at Glasgow,
185572.
176 DYNAMICAL THEORY OF SOUND
which must take place in soundwaves propagated in actual
fluids, we must have recourse to accurate equations of motion.
On the plan of 59 we have
and P = po/(l + A) = p /(l f ^ (7)
Hence, on the adiabatic hypothesis that
P/Po ~ (p/po) 7 > (8)
we find by elimination of p and p
2\Y+1
, (9)
where c 2 = ypo/po as before.
For illustrative purposes it is sufficient to consider the
isothermal case, which is derived from the above by putting
7=1, so that
We have seen in 60 that on the hypothesis of infinitely
small vibrations there is a definite relation between particle
velocity and condensation in a progressive wave. Following
Earnshaw, we assume (tentatively) that the same thing holds
in the general case, and write accordingly
(11)
where the form of the function is to be determined. From this
we deduce
da?'
and therefore = f ................ (IS)
Hence (10) is satisfied provided
PLANE WAVES OF SOUND 177
no additive constant being necessary if we assume that f =
in the parts of the medium not affected by the wave. This
may also be written
?= + clog(l+s), .................. (16)
by 59 (3). Another form is
P/P* = e^' c ...................... (17)
When s is infinitesimal the formula (16) reduces to f = + cs, in
agreement with 60.
To find the rate at which any particular value of s is
propagated, in either of these cases, we note that the value of
9f/9a? which is associated with the particle x at the instant t
will have been transmitted to the particle x + $>x at the instant
t + $t, provided
f 8M.ga.o,
oxdt da?
i.e. by (12) and (14),
&c c(l+s)& = ................ (18)
The phase s is therefore propagated with the velocity
J = + c(l + S ) .................. (19)
relative to the undisturbed medium. To find the rate of
propagation in space we have to take account of the total
variation of x + , which is
The required velocity is therefore
('+DM S ........ ;
The lower sign relates to a wave travelling in the direction of
# positive. It appears from (16) that positive values of f are
then associated with positive values of s, as in the approximate
theory of 60 ; but the formula (20) shews that the velocity
of propagation is greater, the greater the value of s. The parts
of the wave where the density is greater therefore gain con
tinually on those where it is less. Thus if the relation between
s and x be exhibited graphically, the curve A in the annexed
L. 12
178
DYNAMICAL THEORY OF SOUND
figure takes after a time some such form as B*. The wave
becomes, so to speak, continually steeper in front, and slopes
more gradually in the rear, until a time arrives at which the
gradient at some point becomes infinite. After this stage the
analysis ceases to have any real meaning.
Fig. 62.
The adiabatic hypothesis leads to results of the same
general character. The reader will find no difficulty in verifying
the following statement. The formula (16) is replaced by
and the velocity of propagation of a particular value of s is
Tc(l+*)* (7 + 1) .................. (22)
relative to the undisturbed medium, or
in space. In the latter formula the particlevelocity is added
to the velocity of sound proper to the actual density, which is
on the adiabatic hypothesis dependent on the degree of con
densation and consequent change of temperature. The general
conclusions are as before.
* It is not very important here whether the coordinate x be supposed (as in
the previous part of this investigation) to refer to the undisturbed medium, or
to be an ordinary spacecoordinate. In either case the tendency is the same.
PLANE WAVES OF SOUND 179
It must be remembered that since the equation of motion
(9) is not linear, distinct solutions, such as those representing
waves travelling right and left, respectively, which we have
just been considering, cannot be superposed by mere addition.
It may however be remarked that, as a result of a more
complete investigation, Riemann* found (1860) that a localized
arbitrary initial disturbance does eventually resolve itself into
two waves of the above kinds, travelling in opposite directions.
To follow exactly the career of waves of finite amplitude
generated in any given manner is a problem of considerable
difficulty; but some indications may be obtained by methods
of approximation. This procedure was adopted by Airy f (1845)
in his work on the dynamical theory of the tides, where similar
questions arise with respect to tides in shallow seas and
estuaries.
Suppose, for instance, we have a long straight tube in which
a piston (at x = 0) is made to move in an arbitrary manner
?=/(<) ....................... ..(24)
The equation (9) becomes, if we neglect terms of the third
order in the derivatives of f,
If we omit the last term, we have as in 60 the first
approximation
(26)
Substituting this value of f in the small term of (25) we
obtain
The solution of this which is consistent with (26) is
....... <2S>
as is easily verified. The correction to the first approximation
* Bernhard Riemann (1826 66), professor of mathematics at Gottingen
185766.
t Sir George Biddell Airy (1801 92), Plumian professor of astronomy at
Cambridge 182835, astronomer royal 183581.
122
180 DYNAMICAL THEORY OF SOUND
(26) is proportional to x, and to the square of the ratio of the
velocity of the piston to the velocity of sound. This latter ratio
may in practice be exceedingly small, but as we travel to the
right the correction continually increases in importance, until at
length the neglect of terms of the third and higher orders
would no longer be justified. This is what we should expect
from the results of Earnsnaw's investigation.
When the motion of the piston is simpleharmonic, say
/(0 = acosnt, (29)
the formula (28) gives
g = a cos n (t } + (v + V n9a * % \ _ C os2rc (t }l. (30)
\ C/ oC \ C/j
The displacement of any particle is no longer simpleharmonic,
but consists of a part independent of t together with two
simpleharmonic terms, one having the frequency of the
imposed vibration (29), and the other a frequency twice as
great. This illustrates the implied limitation to infinitely
small motions in the usual theory of forced oscillations ( 17).
Again, if the given vibration of the piston be made up of
two simpleharmonic components, say
f(t) = ! cos nj + a 2 cos n 2 t, (31)
we find
f = ttj cos nj (t   j + a 2 cos n z (t J
f 1 ( / #\
 x \ n?a? + n 2 2 a 2 2 nfaf cos 2w x It
kr ( V cj
n 2 W cos 2n 2 (t
V c
f 2/1^2 Oi a z cos (/*! n 2 ) { t 
2n 1 n 2 a 1 o 2 cos (^ + ^2) It JY .
.(32)
We thus learn that in addition to the vibrations of double
frequency, other simpleharmonic vibrations whose frequencies
are respectively the difference and the sum of the primary
frequencies now make their appearance. In acoustical language,
two simple vibrations of sufficient amplitude may give rise not
PLANE WAVES OF SOUND 181
only to the corresponding pure tones, but to their octaves, as
well as to certain "combination tones," whose occurrence
reminds us again, that the principle of superposition is no
longer valid. We shall have occasion to refer to this investi
gation at a later period (Chap. X).
The analogous phenomenon in tidal theory is the production
of "over tides," which are in fact appreciable, and have to be
provided for in the Harmonic Analysis referred to in 39.
We have seen that the main effect of finite amplitude is
that in a progressive wave the gradients, both of pressure and
of density, tend to become infinite. This has suggested the
question whether a wave of discontinuity might not finally be
established, analogous to a "bore" in waterwaves. To examine
into the possibility of such a wave we take the question in its
simplest form, and assume that the circumstances are everywhere
uniform, except for the sudden transition at the plane of dis
continuity. Further, by the superposition of a certain uniform
velocity, we reduce the problem to one of steady motion in
which the plane in question is fixed.
The symbols p , p , u will then be supposed to refer to the
region to the left of this plane, whilst the values of the corre
sponding quantities on the
right are denoted by p, p, u.
Since in every unit of time
the same mass (m) of fluid
crosses any unit area normal
to the direction of flow, we have
pu = p u =m, or u = mv, u Q = mv (33)
Again, since in unit time a mass m has its velocity changed from
UQ to u, the momentum of the portion of air included between
two planes in the positions indicated by the dotted lines in
Fig. 63 is increasing at the rate m (u  u ), whence
p p = m(uu Q \ (34)
or, in virtue of (33),
p p = m?(v v ), (35)
in agreement with (5). If we now superpose a uniform velocity
182 DYNAMICAL THEOEY OF SOUND
MO, we get the case of a wave advancing into a region
previously at rest. The wavevelocity is given by
V Q V pp po
as first found by Stokes (1848), and afterwards independently
by Earnshaw, Riemann, and Rankine. A difficulty, first pointed
out by Lord Rayleigh, arises, however, as to the conservation of
energy. The rate at which work is being done on the portion
of air above considered is p u Q pu, whilst that at which the
kinetic energy is increasing is \ m (u? u<?). The difference is
p u pu\m (u 2  uf) = \m ( p + p)  v). . . .(37)
If the two points (v, p), (v , p ) on the indicator diagram be
denoted by P, P , respectively, the expression (37) is m times
the area of the trapezium bounded by the straight line P P,
the axis of v, and the ordinates p 0) p. If the transition be
effected without gain or loss of heat, the points P , P will lie
on the same adiabatic, and the gain of intrinsic energy will be
represented by the area included between this curve, the axis
of v, and the same two ordinates. Since the adiabatics are con
cave upwards, the latter area is (in absolute value) less than the
former. It appears on examination of the signs to be attributed
to the areas that if v > v the work done is more than is accounted
for by the increase of the kinetic and intrinsic energies, whilst if
VQ < v the work given out would be more than is equivalent to
the apparent loss of energy.
It is evident that no complete theory of waves of discon
tinuity can be attempted without some reference to viscosity
and to thermal conduction, since at the point of transition
the gradients of velocity and temperature are infinite.
It does not appear probable that under ordinary conditions
the modifications due to finite amplitude are of serious im
portance. In equation (30), for instance, the ratio of the
amplitude of the vibration of the second order to that of the
primary vibration is comparable with tfax/c 2 , or with n*a/g . x/H,
where H is the height of the homogeneous atmosphere. With
ordinary amplitudes a, and ordinary distances x, this ratio will
be very small. In three dimensions the effect must be very
PLANE WAVES OF SOUND 183
much less, owing to the diminution of amplitude by spherical
divergence.
64. Viscosity.
The essence of viscosity is that in a moving fluid the stresses
differ from a state of pressure uniform in all directions about a
point, by quantities depending on the rates of deformation. It
is usually assumed that these quantities are linear functions of
the rates of strain ; from our present standpoint this is
sufficiently justified by the fact that the strain velocities are
regarded as infinitely small. As in 40 there will at any
instant, and at any given point, be three principal axes of the
deformation which is taking place, and these will naturally be
the principal axes of the corresponding stress. We therefore
write, by analogy with 42 (1),
............... (1)
where lt e 2 > *s are the principal strain velocities, and
A = 1 + 2 + 3 ...................... (2)
By the same kind of proof as in 41, // is recognized as the
coefficient of viscous resistance to a shearing motion in parallel
planes; viz. if TJ denote the rate of shear, and TS the corre
sponding stress, we have
r = //i) ......................... (3)
The value of // has been determined with considerable accuracy
for a number of fluids, gaseous as well as liquid.
It will be noticed that the meaning of the symbol p, and
consequently the value of V, is so far indeterminate, since
nothing is altered in the shape of the formulae (1) if we
incorporate in p any constant multiple of A. In the case of
liquids it is in fact usual so to incorporate the second terms in
(1). In the application to gases it is convenient to regard p as
defined by the gaseous laws (p = DpO). There is at present no
experimental evidence as to how far the mean stress about a
point, viz.
i (Pi +p* +p,) = P + (*' + I A*') A,
184 DYNAMICAL THEORY OF SOUND
differs, in a moving gas, from p, as thus fixed ; but from
considerations based on the kinetic theory of gases Maxwell*
inferred (1866) that the two things are identical, and that
accordingly
V }/*' W
As we are interested chiefly in the order of magnitude of
the effects, the precise determination of X' is not of much
consequence to us; accordingly Maxwell's view is adopted for
simplicity in what follows.
The dimensions of // are those of a stress multiplied by
a time, or [ML~ 1 T~ 1 ]. It is found that /*' is independent of
the density, but (in gases) increases with rise of temperature.
Its value for air at C. is about '000170 in absolute c.G.s.
units. It will appear however immediately that the effect of
viscosity in modifying motion depends not so much on the
value of fA as on its ratio to the inertia of the fluid. This
ratio
v = p/po (5)
is therefore called by Maxwell the " kinematic " coefficient of
viscosity; its dimensions are [L 2 T~~ 1 ]. For air at C. its value
is about '132 c.G.s.
The rate at which the stresses on the faces of a unit cube
are doing work in changing its size and shape is given by
X'A 2 + 2fl' (tf + 6 2 2 + 6 3 2 )
' {( 2  6 3 ) 2 + (63  <0 2 + (^  6 2 ) 2 }. . . .(6)
The term p& represents the rate at which the intrinsic
energy is increasing. The remaining terms, which are essenti
ally positive, indicate a dissipation of energy at the rate
f/{fe*s) 2 + ( *0 2 + (, <U 2 } (7)
per unit volume. The mechanical energy thus lost is converted
into heat. It will be noticed that (7) vanishes in the case of
uniform expansion (e t = e 2 = 6 3 ) ; this is a necessary consequence
of our previous assumption as to the value of the constant X'.
* James Clerk Maxwell (1831 79), professor of experimental physics at
Cambridge (187179) ; author of the electromagnetic theory of light.
PLANE WAVES OF SOUND 185
In the case of a pure shearing motion (17), the formula (6)
takes the shape
w^ = /*V ......................... (8)
In plane waves of sound we have e 2 = 0, e 3 = 0, and therefore
from (1) and (4)
p l ^p\^fJi l = p KS + ^fJL / l .......... (9)
Moreover, in the notation of 59,
2 * ............... <'>
The equation of motion, viz.
o^ 9 ^ (11)
*!* ..................... l " J
therefore becomes
*
To obtain a solution appropriate to the case of free waves
we put
f=Pcosfcp, ..................... (13)
where P is a function of t, to be determined. We find that
(12) will be satisfied, provided
<">
This has the form of 11 (3), and the solution is therefore
P = ta<<' T cos(7tfHe), ............... (15)
provided r = 3/2i/& 2 , n? = & 2 c 2  1/r 2 ............. (16)
In all cases of interest cr is a considerable multiple of the
wavelength (X = 2?r/A?), so that n = kc, practically, the friction
having as usual no appreciable effect on the period. Thus
f = Ce~ tir cos (kct + e) . cos kx .......... (17)
This represents a system of standing waves with fixed nodes
and loops. There is a similar solution in which cos kx is
replaced by sin&#, and by superposition of the two we can
construct a progressive wavesystem
% = Ce*l r wak(ctx) ................ (18)
Putting i/ = 132 for the case of air, we find T = '288X 2 , the
units being the second and the centimetre.
186 DYNAMICAL THEOEY OF SOUND
The solution of (14) may also be effected concisely by means
of imaginary quantities. Thus in investigating forced simple
harmonic vibrations of prescribed frequency we assume that
fc _ ftgi nt+ mx (19)
whence, on substitution,
n 2
(20)
c 2 I 1 ivn
The ratio vn/c* is usually very small; thus for n = 1500 its value
is, with previous data, about 1*8 x 10~ 7 . Hence
Taking the lower sign, which corresponds to waves travelling in
the direction of ^positive, and rejecting the imaginary part of
(19), we find
............... (22)
provided l=3(?/2vn* ...................... (23)
This represents a system of waves generated to the right of
the origin by a prescribed motion f = a cos nt at this point (as
by a piston in a tube if we neglect the friction at the sides).
The waves advance, with (sensibly) the usual velocity c, but
diminish exponentially in amplitude as they proceed*. The
linear magnitude I measures the distance over which the waves
travel before the amplitude is diminished in the ratio \\e. In
terms of the wave length we have
I = (Sc/STr 2 !/) . X 2 , .................. (24)
or, with previous data, I = 9'56X 2 x 10 3 . The effect of viscosity
in stifling the vibrations is therefore very slight except in the
case of sounds of very high frequency and consequently short
wavelength. Even for \ = 10 cm. the value of / is nearly 10
kilometres. When we come to the discussion of three
dimensional waves it will be clear that the effect of viscosity
may for most purposes be ignored in comparison with the
diminution of intensity due to spherical divergence. It is,
however, of some interest to observe that there is a physical
* This calculation was first made by Stokes (1845).
PLANE WAVES OF SOUND 187
limit to the frequency of vibrations which are capable of
propagation for more than a very moderate distance.
The viscosity being small, the rate at which work is done
per unit area by the piston in maintaining the wavesystem
(22) must have sensibly the value p n 2 a 2 c found in 60. Since
the energy in the medium to the right is now finite and on the
average constant, this must be equal to the rate of dissipation
of energy by viscosity. The equality is easily verified. The
dissipation is, by (7),
= J // ^rf V** cos 2 "(**) dx, . . .(25)
approximately, if we keep only the most important term.
Writing
and taking the mean value with respect to the time, we obtain
by (23).
65. Effect of Heat Conduction.
A further cause of dissipation of energy is to be found in
the thermal processes consequent on the alternate expansions
and rarefactions of the air. If indeed these succeed each other
with sufficient rapidity, the variations are almost accurately
adiabatic, as explained in 59 ; but, as was first pointed out by
Kirchhoff (1868), the residual conduction of heat is in any case
of equal importance with viscosity. On the kinetic theory of
gases the coefficients of " thermometric " conductivity (v) and
of kinematic viscosity are in fact of the same order of magnitude ;
according to Maxwell the relation is v=^v. For this reason
the preceding calculations of the effect of viscosity on air waves
must not be looked upon as more than illustrative. A complete
investigation, in which both influences are taken into account,
shews that the effect is equivalent to an increase in the
kinematic viscosity, but the order of magnitude is unaffected.
188 DYNAMICAL THEORY OF SOUND
If on the other hand the alternations of density were to
take place with extreme slowness, as in the case of very long
waves of simpleharmonic type, there would be time for
practical equalization of temperature, and the dissipative
influence of conduction as well as viscosity would again be
insignificant. Since the expansions are here nearly isothermal,
the wavevelocity will approximate to the Newtonian value
(I 59 (10))
In intermediate cases the theory shews that the wave
velocity would no longer be constant, but perceptibly dependent
on the frequency. Since no such effect is observed, we infer
that in all ordinary cases the conditions are practically adiabatic.
It appears also that in such intermediate cases the dissipation
would be very greatly increased. The investigation of Stokes
(1851), which is here referred to, relates to the effect of
radiation ; the extension to conduction was made independently
by Kirchhoff and Lord Rayleigh. It is probable that the
effects of radiation alone are of subordinate importance.
The detailed calculation must be passed over, but the
general explanation of the manner in which thermal processes
may operate to produce dissipation of energy has been stated
with such admirable clearness by Stokes that it is worth while
to reproduce the passage in question. The explicit reference is
to radiation, but the same principles are involved in the case of
conduction also.
" Conceive a mass of air contained in a cylinder in which an
airtight piston fits, which is capable of moving without friction,
and which has its outer face exposed to a constant atmospheric
pressure ; and suppose the air alternately compressed and
rarefied by the motion of the piston. If the motion take place
with extreme slowness, there will be no sensible change of
temperature, and therefore the work done on the air during
compression will be given out again by the air during expansion,
inasmuch as the pressure on the piston will be the same when
the piston is at the same point of the cylinder, whether it be
moving forwards or backwards. Similarly, the work done in
rarefying the air will be given out again by the atmosphere as
the piston returns towards its position of equilibrium, so that
PLANE WAVES OF SOUND 189
the motion would go on without any permanent consumption
of labouring force. Next, suppose the motion of the piston
somewhat quicker, so that there is a sensible change of tempera
ture produced by condensation and rarefaction. As the piston
moves forward in condensing the air, the temperature rises, and
therefore the piston has to work against a pressure greater than
if there had been no variation of temperature. By the time
the piston returns, a good portion of the heat developed by
compression has passed off, and therefore the piston is not
helped as much in its backward motion by the pressure of the
air in the cylinder as it had been opposed in its forward motion.
Similarly, as the piston continues its backward motion, rarefying
the air, the temperature falls, the pressure of the air in the
cylinder is diminished more than corresponds merely to the
change of density, and therefore the piston is less helped in
opposing the atmospheric pressure than it would have been had
the temperature remained constant. But by the time the
piston is returning towards its position of equilibrium, the cold
has diminished in consequence of the supply of heat from the
sides of the cylinder, and therefore the force urging the piston
forward, arising, as it does, from the excess of the external over
the internal pressure, is less than that which opposed the piston
in moving from its position of equilibrium. Hence in this case
the motion of the piston could not be kept up without a
continual supply of labouring force. Lastly, suppose the piston
to oscillate with great rapidity, so that there is not time for any
sensible quantity of heat to pass and repass between the air and
the sides of the cylinder. In this case the pressures would be
equal when the piston was at a given point of the cylinder,
whether it were going or returning, and consequently there
would be no permanent consumption of labouring force. I do
not speak of the disturbance of the external air, because I am
not now taking into account the inertia of the air either within
or without the cylinder. The third case, then, is similar to the
first, so far as regards the permanence of the motion; but there
is this difference ; that, in consequence of the heat produced by
compression and the cold produced by rarefaction, the force
urging the piston towards its position of equilibrium, on
190 DYNAMICAL THEOKY OF SOUND
whichever side of that position the piston may happen to be, is
greater than it would have been had the temperature remained
unaltered.
"Now the first case is analogous to that of the sonorous
vibrations of air when the heat and cold produced by sudden
condensation and rarefaction are supposed to pass away with
great rapidity. For we are evidently concerned only with the
relative rates at which the phase of vibration changes, and the
heat causing the excess of temperature passes away, so that
it is perfectly immaterial whether we suppose the change of
motion to be very slow, or the cooling of heated air to be very
rapid. The second case is analogous to that of sound, when we
suppose the constants q* and n comparable with each other; and
we thus see how it is, that, on such a supposition, labouring force
would be so rapidly consumed, and the sound so rapidly stifled.
The third case is analogous to that of sound when we make the
usual supposition, that the alternations of condensation and
rarefaction take place with too great rapidity to allow a given
portion of air to acquire or lose any sensible portion of heat by
radiation. The increase in the force of restitution of the piston,
arising from the alternate elevation and depression of tempera
ture, is analogous to the increase in the forces of restitution
of the particles of air arising from the same cause, to which
corresponds an increase in the velocity of propagation of
sound."
66. Damping of Waves in Narrow Tubes and Crevices.
A somewhat greater effect of viscosity may be looked for
when the air is in contact with a solid body, as at the walls of
a pipe or resonator, owing to the practically infinite resistance
which the surface opposes to the sliding of the fluid immedi
ately in contact with it. It seems in fact to be wellestablished
that the relative velocity vanishes at the surface, whereas in
our theoretical investigations we assume for the most part that
sliding takes place quite freely. A closer examination shews
however that in the case of rapid vibrations, such as we are
concerned with in acoustics, the effect is mainly local, being
* [q is a constant of radiation.]
PLANE WAVES OF SOUND 191
confined, practically, to a very thin layer of air near the
surface, and is except in very narrow spaces unimportant.
The matter may be sufficiently illustrated by a very simple
case. Suppose that the fluid above the plane y = is acted on
by a periodic force
X = fcosnt, (1)
per unit mass, parallel to Ox, the plane forming a rigid
boundary. The consequent motion being everywhere parallel
to Ox and independent of the coordinate x, there is no variation
of density, and the deformations which are taking place are of
the nature of shearing motions parallel to y = 0. Denoting the
velocity f by u, the rate of shear will be
and the shearing stress on a plane parallel to y is accordingly
pdu/dy. The stratum bounded by the planes y and y + By
therefore experiences a resultant force
a
per unit area, parallel to x, and the equation of motion is of the
form
du d*u
We have to solve this under the condition that u = Q for
y = 0. For conciseness we put X = fe int , and reject (in the
end) the imaginary part of our expressions. The equation is
then satisfied by
u = (l n + Ae'y)e i ', .................. (4)
provided m 2 = in/v, or
m=(l+0& ..................... (5)
where = vW 2 ") ......................... ( 6 )
Since we are looking for a solution which shall be finite for
y = oo we take the lower sign. Also, the condition that a =
for y = requires that A = f/in. Hence
192 DYNAMICAL THEORY OF SOUND
or, keeping only the real part,
(8)
Tl fit
a result which is easily verified. When fty is moderately large
the value of u reduces practically to the first term, which is
the same as if there had been no friction. The rigid boundary
accordingly acts as a drag only on a thin stratum ; for example
when y Sir/ft the velocity falls short of its value at a great
distance from the surface by about one part in 535.
In actual problems of acoustics (relating for example to
vibrations in pipes) the force pX per unit volume is replaced
by the negative pressuregradient dp/da, and we have of course
changes of density to take into account, but the results have
a similar interpretation. The linear magnitude
h = 2ir/l3 = )/(4eirp.2irln) ............... (9)
may be taken to measure the extent to which the dragging
effect penetrates into the fluid. With the previous data its value
in centimetres is about 1'29/N*, where N is the frequency; thus
for N = 256 we find h = "80 mm.
We may apply the above investigation to obtain an estimate
of the effect of viscosity on the wavevelocity in a tube, on the
supposition that the diameter is small compared with the
wavelength but large compared with the quantity h. The
tangential stress on the fluid at the boundary y = is, in the
case of (7),
by (9), the timefactor e int being understood. The total tan
gential force exerted by the walls of a cylindrical tube of radius
a on the contained air may therefore be equated to
per unit length, where p denotes the mean pressure over the
section (?ra 2 ). Hence if u be the mean velocity, we have,
calculating the forces on the air contained in an element Bx
of the length,
9 du dp . /n ., , dp
7rp a? =  ira? + $(Ii)ha^,
r dt dx * ^ dx
PLANE WAVES OF SOUND 193
1 dp L , ~ h
, ~
or ^ =  f 11 (!{)= } ............. (11)
dt '
To this we must add the relations
..................... (12)
The elimination of p and s between these equations leads to
............. < 14 >
It is already assumed that the time enters through a factor
e int ; and the solution of (14) is therefore of the type
u=Ce int+mx , ...(15)
with ra 2 =
or m
approximately, on account of the assumed smallness of h/a.
For waves propagated in the direction of ^positive we take
the lower sign, and write
m = in/c' a, (18)
where c'
and a = nh/4s7rac (20)
We have, then u=Ce* .&< xlc '\ (21)
or, in real form, u Ce""* cos n It >j (22)
The wavevelocity is therefore diminished in the ratio given
by (19). The exponential factor in (22) expresses the law of
decay of the waves as they advance. If I be defined as in
64 (23) it will be found that al is of the order \*/ah. The
rate of decay is therefore much greater under the present
conditions than in the case of sound waves in the open.
A formula equivalent to (19) was published without demon
stration by Helmholtz in 1863. The above proof is a variation
of that given by Lord Rayleigh in his Theory of Sound.
L. 13
194 DYNAMICAL THEOKY OF SOUND
A more complete investigation was instituted by Kirchhoff
(1868) in which thermal processes are considered, as well as
viscosity. The effects are thereby increased, as already explained,
but remain of the same order of magnitude.
As already stated, it is implied in the above calculation that
the diameter of the tube greatly exceeds the quantity h. When
on the other hand the diameter is comparable with, or less
than h, the walls have relatively a much greater hold on the
vibrating mass, and the character of the motion is entirely
altered by the friction. In particular, when h is large com
pared with the width the mere inertia of the fluid ceases to
have any appreciable influence, the mean velocity over a
crosssection being determined by an approximately statical
equilibrium between the pressuregradient (in the direction of
the length) and the friction of the walls. We have, then,
(23)
where R is a coefficient of resistance, depending on the nature
of the fluid, and on the shape and size of the crosssection.
Again, by Boyle's law,
=.pb(l + 5), ..................... (24)
the isothermal hypothesis being adopted as now the most
appropriate, since, owing to the assumed narrowness of the
tube, transfer of heat can take place freely. Eliminating p and
s between (13), (23), and (24), we find
du_p d*u
dt~ RW
This has the same form as the equation of linear conduction of
heat. Assuming that
u=Ce int+rn ^, .................. (26)
we have m* = inR/p , and therefore
m=(l+*). .................. (27)
if ^ = ^nR/ Po ................... (28)
Taking the lower sign we obtain
MCkws+^newa^ ............... (29)
or, in real form, u = Ce~ wx cos (nt ^x) ............. (30)
PLANE WAVES OF SOUND 195
The value of R will be sensibly the same as if the fluid were
incompressible. Its determination is therefore the same as in
the case of the steady flow of a liquid under pressure through
a capillary tube. In this case, if the section be circular, the
shearing stress per unit length on a coaxial cylindrical surface
of radius r is 27n* . ^"dujdr, and the resultant of the longitudinal
forces on the two curved faces of a cylindrical shell of thick
ness 8r is therefore
per unit length. The sectional area of the shell being Zirrdr,
the requisite pressuregradient is
, (31)
drj
which is independent of x. There being no radial motion, we
have dp/dr = 0, so that p, and therefore dp/dx, is also independent
of r. The equation (31) is then satisfied by u = A+Br !t t
provided B be properly determined. The constant A is fixed
by the consideration that there is no slipping at the wall
(r = a) of the tube. In this way we find
The mean velocity over the area of the section is therefore
?ra 2 .' o 9# &P>' '
Hence, for a circular section,
12 = V/a (34)
The formula (33) contains Poiseuille's* law of efflux of liquid
through a capillary tube, viz. that the discharge per second
varies as the pressuregradient and as the fourth power of the
diameter. It may be made the basis of an experimental method
of determining p.
* J. L. M. Poiseuille (1799 1869), a practising physician in Paris, who was
interested in the capillary circulation of the blood. The date of the memoir
referred to is 1844.
132
196 DYNAMICAL THEORY OF SOUND
The case of an elliptic section can be solved in a similar
manner. The result, first given by Boussinesq (1868), is
E = V(a 2 + 6 2 )/a 2 & 2 , (35)
where a, b are the semiaxes. If we put a = oo we get the
case of a narrow crevice, bounded by parallel planes, the
breadth being 26, viz.
E = V/6 2 (36)
This can of course be obtained more easily by an independent
process.
The formula (30), when combined with (34) or (36), agrees
with the result of the more complete investigation given by
Lord Rayleigh (1883). It appears that u goes through its
cycle of phases in a distance 27r/r, but that within this space
the amplitude is diminished in the ratio e~ 2n = 1/535. In the
case of circular section we have
OT 2 = ifi'n/pda*, (37)
by (28) and (34). Hence when the circumstances are such that
the ratio v/na? is large, the distance in question is small com
pared with the wavelength (X = ^TTC/U) in the open ; for we
have
(Xs7/27r) 2  *7 2 c 2 /tt 2 = 4>v/na 2 (38)
Hence in a sufficiently narrow tube the waves are rapidly
stifled, the mechanical energy lost being of course converted
into heat.
The investigation has been employed by Lord Rayleigh to
illustrate the absorption of sound by porous bodies. When
a soundwave impinges on a slab which is permeated by a large
number of very minute channels, part of the energy is lost, so
far as sound is concerned, by dissipation within these channels,
in the way just explained. The interstices in hangings and
carpets act in a similar manner, and it is to this cause that the
effect of such appliances in deadening echoes in a room is to be
ascribed, a certain proportion of the energy being lost at each
reflection. It is to be observed that it is only through the
action of true dissipative forces, such as viscosity and thermal
conduction, that sound can die out in an enclosed space, no mere
modification of the waves by irregularities being of any avail.
CHAPTER VII
GENEKAL THEORY OF SOUND WAVES
67. Definitions. Flux. Divergence.
In respect of notation it is convenient now to take a point
of view somewhat different from that adopted in the preceding
chapter. We denote by u, v y w the component velocities,
parallel to rectangular axes, considered as functions of position
(x, y, z) and of time t. With each point of space there is
accordingly associated, at any given instant, a vector (u, v, w),
and the whole assemblage of such vectors gives an instantaneous
picture of the distribution of velocity*. On the other hand
the variations of u, v, w with the time, for given values of
x, y, z, give the history of what goes on at a particular place f,
but supply in the first instance no information as to the
careers of the various particles which (so to speak) successively
cross the scene.
When we proceed to calculate the component accelerations
of the particle which at the instant t is in the position (x, y, z)
we have to take account of the fact that after the lapse of a
time however short its velocities u, v, w will be given by the
respective functions of the altered position as well as the altered
epoch. Suppose that at two successive instants ^, ^ a particle
occupies the positions P and P', respectively, and that the
corresponding values of the ^component of the velocity are
* M. Marey and others have taken photographs, of short exposure, of a two
dimensional current of water carrying suspended motes. The image of each
mote is drawn out into a short line, which indicates the direction and magnitude
of the corresponding velocity.
t As if we were to view the surface of a stream through a narrow tube,
198 DYNAMICAL THEOEY OF SOUND
u l} u 2 at P and w/, u% at P'. The ascomponent of the
acceleration of this particle will be the limit of
The limit of the first term on the right is du/dt, the rate of
change of u at P. Again u% u z is the difference of simul
taneous velocities at the points P, P', so that, ultimately,
*,), ............ (2)
where du/ds is a spacedifferentiation in the direction PP', and
q is the resultant velocity \/(u 2 + v* + w z ). The final expression
for the acceleration parallel to x is therefore
du du ,
Similar values are obtained in like manner for the other
components. If (I, m, n) be the directioncosines of PP',
we have
du du dx du dy du dz
ds dx ds dy ds dz ds
j du du du
= t^+ra + 7i , (4)
dx dy dz
Philst u=lq, v = mq, w = nq (5)
!ence we may write (3) in the form
du du du du ._.
dt* U dx +V ty+ W 3z>
which is familiar to students of Hydrodynamics.
It has been thought worth while, as a matter of principle,
to accentuate the changed point of view, but in the application
to motions which are treated as infinitely slow the distinction
loses its importance. The second term in (3) is then of the
second order in the velocities, and the component particle
accelerations may be identified with du/dt, dv/dt } dw/dt. The
extent of the error here involved, in acoustical questions, may
be estimated as in 60 by a reference to plane waves of sound.
If
= a cos k (ct x), (7)
GENERAL THEORY OF SOUND WAVES 199
the ratio of the maximum value of udu/dx to du/dt is ka. The
restriction to " infinitely slow " motions therefore means that
the amplitude must be small compared with X/27T.
If we fix our attention on any geometrical surface, open or
closed, drawn in the region occupied by the fluid, the expression
(lu + mv + nw) BS . Bt,
where (/, m, n) is the direction of the normal drawn from an
elementary area BS of the surface, towards one side, measures
the volume which in the infinitely short time Bt crosses BS.
The coefficient of Bt in this expression is called the "flux"
across BS, and its integral
(lu + mv + nw) dS, . (8)
taken over the surface, is called the total flux across the latter
towards the side on which the normals are supposed drawn. It
measures the rate at which fluid is being carried across the
surface, expressed in terms of volume per unit time.
To calculate the flux outwards across the boundary of an
elementary rectangular region BxByBz having its centre P at
the point (x, y, z), we note that the average velocities parallel
to x, over the faces ByBz, being equal to the values of u at the
centres of these faces, will be
respectively. The difference of the fluxes, from left to right,
across these faces is accordingly du/dx .BxByBz. Adding the
corresponding terms for the other pairs of faces, we obtain the
result
C r
(c
ou dv dw\ ~ .
~ + 5 + )Ba:ByBz (9)
dy oz
The expression in brackets gives a sort of measure of the rate
at which the substance in the neighbourhood of P is on the
whole flowing away from P. It is therefore called the
" divergence " of the vector (u, v, w), and is denoted by
div (u, v, w) ; thus
. du dv dw
, W ) = + + (10)
200 DYNAMICAL THEOEY OF SOUND
By dividing any finite region into rectangular elements we
see that the total flux outwards across the boundary must be
equal to the volumeintegral of the divergence, or
This can of course be proved mathematically without attributing
any kinematical meaning to the symbols.
68. Equations of Motion.
To form the dynamical equations, we fix our attention on
that portion of matter which at the instant t occupies the
rectangular space &xyz. On the hypothesis of infinitely
slow motion its acceleration of momentum parallel to x is
p Sx &y$z .du/dt, where p is the density. The mean pressures
on the respective faces may be taken to be the pressures at the
centres of those faces, and the total pressures on the two faces
perpendicular to x are therefore
The difference gives a force dp/dx. Sx&y&z in the direction of
^positive. Equating this to the acceleration of momentum, we
obtain the first of the following system of equations :
du _ dp dv _ dp dw _ _dp ,.
p dt~~d~x' p dt~~dy' p ~dt~~dz'
Since the variations of p when multiplied by du/dt, ..., ... may
be neglected, we may replace p by its equilibrium value p , but
it will not always be necessary to preserve the suffix.
As in 59 we write
} ........................ (2)
where s denotes the condensation (p po)/po, and K is the cubic
elasticity of the fluid. If we further write
c 2 = */po, ........................ (3)
as before, we obtain
du_ 8s dv_ _ 8s <^__ c2 ds ( ft
dt~ dx> dt~ d' dt~ dz ....... W
GENEKAL THEOEY OF SOUND WAVES 201
If A denote the dilatation of volume of the fluid which at
the instant t fills the space SasSySz, as compared with its
equilibrium condition, we evidently have
^ = div (u, v, w\ (5)
or since, in the case of small motions, s = A,
The equations (4), (6) are fundamental in the present branch of
our subject. The purely kinematical relation (6) is sometimes
called the " equation of continuity."
69. VelocityPotential.
If we integrate the equations (4) of 68 with respect to t
we obtain
,1
OX j  v_i/ j u i / 1 \
where U Q , v 0t W Q are the values of u, v, w at the point (a?, y, z) at
the instant t = 0. In a large class of cases, these initial values
of u, v, w can be expressed as the partial differential coefficients
of a singlevalued function of (x, y, z), thus
Throughout any region to which this statement applies, the
values of u, v, w at any subsequent instant t can be similarly
expressed; thus, from (1),
 ' .........
rt
where <^> = c 2 I sdt+fa ................... (4)
J o
This function <f> is called a " velocitypotential," owing to its
analogy with the potentialfunction which occurs in the theories
of Attractions, Electrostatics, &c. It was introduced into
hydrodynamics by Lagrange.
202 DYNAMICAL THEORY OF SOUND
The instantaneous configuration of the " equipotential sur
faces" </> = const, indicates at once the distribution of velocity,
as regards both magnitude and direction.
Suppose two consecutive surfaces to be
drawn, for which the values of < differ by
S<. Let PP' be drawn normal to these, and
PP l parallel to x\ and let PP'=v. Ac
cording to (3) the velocity at P, resolved Flg< 64 '
in the direction PP ly is
3 PP' 8
ultimately, if I denote the cosine of the angle which the normal
PP' makes with Ox. From this, and from the analogous forms
of v, w t it is seen that the velocity at P is normal to the equi
potential surface passing through that point, and is equal in
magnitude to the limiting value of </>/*/. Hence if a system
of surfaces be drawn corresponding to values of <f> which differ
by equal infinitesimal amounts, the velocity is everywhere
orthogonal to these, and inversely proportional to Sv, the distance
between consecutive surfaces. More precisely, the velocity is
everywhere in the direction in which < decreases* most rapidly,
and is equal in absolute value to the gradient of <.
If we draw a linear element PQ (= Bs) in any other direction,
the velocity resolved in the direction of PQ is equal to the limit of
. ............... (6)
or  d(f>/ds.
The cases in which a velocitypotential exists include all
those where, in the region considered, the fluid was initially at
rest, for we may then put </> = 0, simply, and the subsequent
value is
4>=c 2 f sdt ........................ .(?)
J o
This will hold whenever the motion has been originated by the
vibration of solid or other bodies.
* It should be mentioned that in many books is taken with the opposite
sign; thus u = d<J>jdx, &c.
GENERAL THEORY OF SOUND WAVES 203
The real meaning of the property which differentiates the
present type of motion from all others is most clearly expressed
in terms of the "circulation" round a closed curve. If we divide
the curve into infinitesimal linear elements, and multiply the
length of each element by the tangential component of the
velocity, estimated always in the same direction round the
curve, the result is the "circulation" referred to. It may be
denoted by
\( u^ +v ^ \wjJds, or \(udx + vdy + wdz). ...(8)
On the present hypothesis the tangential velocity is d<f>/ds, and
the integral of this, taken round the circuit, is zero, the first and
last values of < being the same. The circulation is therefore
zero in every circuit which can be drawn in the region in
question. For a reason which may be understood by reference
to the case of an infinitesimal circuit, the type of motion now
under consideration is called "irrotational." The name has the
advantage of calling attention to a geometrical property rather
than to an analytical form of expression.
A dynamical interpretation can also be given to the
velocitypotential. The equations (3), when written in the
forms
p u = pfi^/dx, p v =  pdd<f>/dy, p w = pfifydz, (9)
shew that < is the potential per unit mass of a system of
extraneous impulsive forces which would generate the actual
motion of the fluid instantaneously from rest.
The theorem as to the persistence of the irrotational character
is most important; but it is necessary to observe the restrictions
under which it has been proved. It was implied, in the first
place, that the fluid was frictionless, and this is essential.
Again the medium has been supposed free from extraneous
forces, but the restriction is easily removed in the case of forces
which, like gravity, have a potential (per unit mass). Finally,
the assumption has been made that the motion is infinitely
small. This simplifies the proof, and covers most cases which are
of interest in acoustics. A more rigorous investigation would
shew that the circulation is (under the above condition) still
204 DYNAMICAL THEOKY OF SOUND
constant round any circuit, provided we imagine the circuit to
move with the fluid. If initially zero for every circuit which
can be drawn in a finite portion of the fluid, it will remain zero
for every such circuit.
70. General Equation of Sound Waves.
We postulate henceforth the existence of a velocity potential,
at all events in the case of a uniform medium, to which we
confine ourselves for the present. We have then, from
68 (6)
This symbol V 2 is called the "Laplacian operator," from its
constant occurrence in the analytical theory of attractions as
first developed by Laplace. Again, by differentiation of 69 (4)
with respect to t we get
Finally, by elimination of s,
w
This may be regarded as the general differential equation of
sound waves in a uniform medium. If a solution can be
obtained which gives prescribed initial values to </> and s
(or dQ/dt), and satisfies the other conditions of the problem, the
subsequent value of s is given by (3), and the values of u, v, w
by 69 (3).
We may stop for a moment to notice the form assumed by
the equations when the fluid is incompressible. This may be
regarded as an extreme case, in which c is made infinite, whilst
s is correspondingly diminished, in such a way that c 2 s, which
= (p p )/p , remains finite. The equation of continuity, 68
(6), takes the form
(5)
GENEKAL THEOEY OF SOUND WAVES 205
which is otherwise obvious from the meaning of "divergence."
In the case of irrotational motion, this becomes
V 2 < = 0, ........................... (6)
which is identical with "Laplace's equation" in the theory of
attractions. The same equation occurs in the theory of steady
electric (or thermal) conduction in metals. If, for example, $
denote the electric potential, the formulae (3) of 69 give the
components of current, provided the specific resistance of the
substance be taken to be unity. This analogy will be found
useful in the sequel.
The theory of the motion of incompressible fluids is capable
of throwing more light, occasionally, on acoustical phenomena
than might at first sight be anticipated. We are apt to forget
that the velocity with which changes of pressure are propagated
in water is after all only four or five times as great as in air,
and that the visible (or at all events easily imaginable) motions
of water, under circumstances where the compressibility has
obviously little influence, may supply a valuable hint as to the
behaviour of a gaseous substance under similar conditions. This
remark will have frequent illustration in the following chapters.
The kinetic energy of a system of sound waves is
.
The potential energy, as given by the argument of 60, is
...(8)
The integrations extend over the region affected.
71. Spherical Waves.
In the case of plane waves with fronts perpendicular to Ox
the equation (4) of 70 reduces to
**<** m
dt*~*dtf'
whence <f> f(ct x) + F(ct + x) ................ (2)
This need not be further discussed.
206 DYNAMICAL THEORY OF SOUND
The case which comes next in importance is that of
symmetrical spherical waves. If <j> be a function of the
distance r from the origin and of t, only, the velocity is d<t>/dr
outwards, in the direction of the radius, and is uniform over
any spherical surface having the origin as centre.
Instead of applying the general equation to the present
circumstances it is simpler to form the kinematical relation
corresponding to TO (1) de novo. The flux outwards across
a sphere of radius r is d<f>/dr . 4Trr 2 , and the difference of flux
across the outer and inner surfaces of a spherical shell of thick
ness Br is accordingly
The volume of the shell being 47rr 2 8r, this must be equal to
A . 4?rr 2 Sr or s . 47rr 2 Sr, whence
dt c
bmce c2s== ^i (4)
dt
i i. <P& c 2 d / 96\
as usual, we have : =  5 r 2 ^ (5)
dc r dr \ d/*/
This may also be written
The solution of this equation, viz.
represents the superposition of two wavesystems travelling
outwards and inwards, respectively, with the velocity c. In
the case of a diverging wave system
r<t>=f(ctr) (8)
we have, by (4), crs=f(ct r) (9)
Any value of rs is propagated unchanged ; the condensation A
therefore diminishes in the ratio l/r as it proceeds, and the
potential energy per unit volume diminishes as l/r 2 . For the
particlevelocity we have
3JL 1 1
r) (10)
GENERAL THEORY OF SOUND WAVES 207
The law of dependence on distance is here more complicated,
but as the wave spreads outwards the first term ultimately
predominates ; the velocity at corresponding points of the wave
then varies as 1/r, and the kinetic energy per unit volume
as 1/r 2 .
In a diverging wavesystem we have, from (9),
an = l(r+), .................. (11)
and similarly, in a converging wavesystem.
These relations correspond to (5) of 60, which is indeed a
particular case, since as r increases our spherical waves tend to
become ultimately plane.
The general argument of 23 can be adduced to prove that
in a diverging (or a converging) wavesystem by itself the
energy is half kinetic and half potential.
The solution (7) can be applied to a region included
between concentric spheres, or to a region having only one
finite spherical boundary, internal or external. In any case,
the conditions to be satisfied at the boundaries, whether finite
or infinite, must be given in order that the problem may be
determinate. In particular, even when the region is otherwise
unlimited, the point r = is to be reckoned as an internal
boundary ; . this point might for instance be occupied by a
" source " of sound ( 73). When there is no source there, the
flux across a small spherical surface surrounding must vanish,
i.e. we must have
(13)
r =0 \ v
When applied to (7) this condition gives
f(ct) + F(ct) = 0, .................. (14)
for all values of t, and the general solution therefore takes the
shape
r$ = F(ct + r)F(ctr) ............. (15)
This formula may be used to determine the motion con
sequent on arbitrary initial conditions which are symmetrical
208 DYNAMICAL THEOKY OF SOUND
about 0, in an unlimited medium. Suppose that when t =
we have
(16)
The former of these functions determines the initial distribution
of velocity, and the latter that of condensation. The function
F must now satisfy the conditions
F(r)F(r)=r^(r) ................ (IV)
(18)
It is to be noted that the variable r is essentially positive ; this
explains why two equations are necessary to determine F for
positive and negative values of the argument.
Suppose, for example, that there is no initial velocity
anywhere, but only an initial condensation, so that </> (r) = 0.
From (17) and (18) we deduce
F' (r)=J"( r) = \ r  x .(r) .......... (19)
The condensation at time t is given by
Ity F'(ct+r)F'(ctr)
= c *tt ~^r
This takes different forms according as ct is less or greater than
r. In the former case
<*)), ...(21)
and in the latter
As a particular case, suppose we have an initial condensation
which is uniform ( = s ) throughout the interior of a sphere of
radius a, and vanishes for r > a ; and let us examine the
subsequent variations of s at points outside the originally
disturbed region. Since % (r) vanishes by hypothesis for r > a,
the first part of the solution (21) or (22) disappears in the
GENERAL THEORY OF SOUND WAVES 209
present case. So long as ct < r  a, the second part of (21) will
also vanish, but when ct lies between r a and r we shall have
(23)
When ct > r, the "second formula (22) applies, and we find
that, so long as ct < r + a, the result (23) will still hold.
Finally, when ct > r + a we have again 5 = 0. The results
are shewn graphically in the following figure which exhibits
the variation of s with t at a particular point, and the space
cba,
ct+a
Fig. 65.
distribution of s at a particular instant, respectively. It
appears that after the lapse of a certain time (2a/c) we have a
diverging wave in the form of a spherical shell of thickness 2a,
and that s is positive through the outer half, and negative
through the inner half of the thickness. The changes in the
velocity may be inferred by means of the formula q = d<j>/dr.
For values of t between (r a)/c and (r + a)/c, i.e. during the
time of transit of the wave across the point considered, we find
......... (24)
whilst for other values of t we have <f> = 0. Hence within the
aforesaid limits of time we have
(25)
When r is large compared with a this changes sign for t = r/c,
approximately, the velocity being directed outwards in the
outer half, and inwards in the inner half of the shell. At the
boundaries of the disturbed region, where r = ct a, we have
q = + cas /2r. As the diverging wave reaches any point the
velocity suddenly rises from zero to the former of these values,
and as it leaves it the velocity falls suddenly from the latter
(negative) value to 0, The origin of the discontinuities in this
L. 14
210 DYNAMICAL THEORY OF SOUND
solution is to be sought of course in the discontinuity of the
initial distribution of density. Any difficulty which may be
felt on such grounds may in general be removed by substituting
in imagination an initial distribution in which the discontinuity
is replaced by a very rapid but continuous transition.
The solution of (6) in terms of the general initial con
ditions (16) may be investigated in a similar manner, but it
must suffice to quote the results. It may easily be verified
that they satisfy all the conditions of the question. They are
n/> = i (r + ct) fa(r + ct) + $(r ct) fa (r  ct)
1 fr+ct
+ il %<*)<* (26)
&&J rct
for ct < r, and
r$ =  (ct + r) fa(ct + r)(ct r) fa(ctr)
1 rct+r
+ 2J ()*... (27)
*C J ctr
for ct > r.
Since the origin evidently occupies an exceptional position
in the theory of spherical waves it is desirable to calculate the
value of (f> there, more especially as the result will be of service
presently when we come to the solution of the general equation
70 (4) of sound waves. The result may be deduced from
(27), or more directly from (15). We find
...(28)
and therefore from (17) and (18)
(29)
For example, in the special problem above considered, where
fa (r) = 0, whilst % (r) c\ or according as r a, we find
$ = c\t or according as t a/c. The consequent value of
s at is s for t < a/c and zero for t > a/c, whilst at the instant
t = a/c it is negative infinite. To escape this result we must
slightly modify the data, replacing the original distribution
of density by a continuous one. The figure is an attempt to
GENERAL THEOEY OF SOUND WAVES
211
shew an initial distribution of s which varies rapidly but
continuously from s to in the neighbourhood of r = a,
together with the consequent timevariation of s at 0.
cb
Fig. 66.
The problem which we have discussed exhibits a marked
contrast with the theory of plane waves, in that the wave
resulting from an arbitrary disturbance contains both con
densed and rarefied portions, even when there is no initial
velocity and the initial disturbance of density has everywhere
the same sign. The statement is easily generalized by means
of equations (1) of 69. If we take the integral of the value
of s at any point P over a time which covers the whole transit
of the wave, so that the values of u, v, w vanish at both limits,
we find that its spacederivatives are all zero. The integral
has therefore the same value for all positions of P. And by
taking P at an infinite distance, so that s becomes infinitely
small by spherical divergence, we see that the value is in fact
zero, i.e.
sdt = 0.
.(30)
The mean value of s at any point is therefore zero. This result
is of course not limited to the case of spherical waves.
142
212 DYNAMICAL THEOKY OF SOUND
72. Waves resulting from a given Initial Disturbance.
We have next to trace the effect of initial conditions in an
unlimited region, in the general case. We suppose that at the
instant t = we have
where the functions are arbitrary. To deduce the effect at any
subsequent instant, at any assigned point P, we consider in the
first instance the average value of < over a sphere of radius
r described with P as centre. This will be denoted by
(2)
if So> represent the elementary solid angle (SS/r 2 ) subtended at
P by any elementary area $S of the sphere. In the same way
we write
* 47T
//* (3)
This, like (2), will be a function of the variables r and t only.
If in 70 (3) we multiply both sides by 8&)/4?r, and integrate
over the aforesaid sphere of radius r, we find
It is also evident that the average normal velocity over the
sphere will be d(j>/dr. The argument by which the rate
of change of s was in 71 inferred from the consideration
of the total flux out of the region bounded by the spheres
r and r + &r can then be applied to prove that in the present
case
_
B; t*Sr
Eliminating s, we have
Sf^
dt'
which is identical in form with (5) of 71. We recognize then
that < is the velocitypotential of the system of spherical waves
GENERAL THEORY OF SOUND WAVES 213
which would result from initial distributions of velocity and
condensation expressed by
. ............... (7)
these functions of r being the average values of < (a, y, z) and
^o (x, y, z) taken over the aforesaid sphere. It follows from
71 (29) that the value of <f> at P is given by
(8)
This gives a rule for calculating the value of < for a point P at
any given instant t. It may be stated in words as follows :
To find the part of < due to the given initial distribution
of condensation, we describe about P a sphere of radius ct, and
calculate the average of the given initial values of d<f>/dt, i.e. of
the function ^o (#> y> z\ at the points of space through which
this surface passes, and multiply by t. To find the part due to
the initial velocities we replace the average of the given values
of d<f>/dt by the average of the given initial values of <, i.e. of
the function < (x, y, z), and differentiate the result, as thus
modified, with respect to t.
The theorem contained in (8) was given by Poisson (1819);
the actual form (8) and the interpretation are due to Stokes
(1850). It will be seen that the result, as thus stated, is in
reality very simple, if regard be had to the great generality of
the circumstances which are taken into account.
To trace the sequence of events at P we employ a series of
spheres whose radii (ct) increase continually from zero. If P
be external to the region which is the locus of the initial
disturbance, no effect is produced so long as the spheres do
not encroach on this region. If r lt r 2 be the least and greatest
distances of P from the boundary, the disturbance at P will
begin after a time r^c, will last for a time (r 2 rj/c, and will
then cease.
If with the various points of the boundary of the originally
disturbed region as centres we describe a series of spheres of
radius ct, the outer sheet of the envelope of these spheres will
mark out the boundary of the space which has been invaded by
\
214 DYNAMICAL THEORY OF SOUND
the disturbance up to the instant t. The envelopes corre
sponding to successive values of t will form a series of what are
known in geometry as "parallel surfaces"; in other words, the
boundary of the disturbed region spreads everywhere normal
to itself with the constant velocity c.
As a simple application of the formula (8) we may take the
problem already discussed in 71, where an initial uniform
condensation s was supposed to extend throughout the interior
of a sphere of radius a
having the origin as
centre. When a spherical
surface of radius ct, de
scribed with P as centre,
intersects the boundary of
the originally disturbed
region, as in the figure,
,1 P ,1 ,. Fig. 67.
the area of the portion
included within the latter is 2?r . PQ Z (1  cos OPQ), and the
average of the given initial values of s over the whole surface
(47r . PQ 2 ) is therefore
where r = OP. Hence, by the rule,
in agreement with 71 (24).
73. Sources of Sound. Reflection.
The very useful conception of a "pointsource" was
introduced into the subject by Helmholtz. We may imagine
(with Maxwell and Lord Rayleigh) that at such a point fluid
is introduced or abstracted at a certain rate, and that the
"strength" of the source is measured by the volume thus
introduced per unit time. The wavetrain due to a source of
strength f(t) at the origin is accordingly represented by
(1)
GENERAL THEORY OF SOUND WAVES 215
since this makes
If we differentiate the general equation of sound waves ((4)
of 70) with respect to x or y or z, we recognize that if <j> is a
solution so also is d<t>/dx, or d(j>/dy, or d<f>/dz. Thus from (1) we
derive the solution
which satisfies the general differential equation except at the
singular point r = 0. The value of <f> thus obtained may be
interpreted as the velocitypotential of a " double source " due
to the juxtaposition of two simple sources which are always in
opposite phases. This will be explained more fully in 76, in
the particular case where the variation with time is simple
harmonic.
The problem of reflection of sound by a rigid infinite plane
is readily solved by the method of "images." If with every
source P of sound on the near side of the boundary we associate
a similar source at the geometrical image P' of P with respect
to the plane, it is obvious that the condition of zero normal
velocity over the plane would still be fulfilled if the boundary
were abolished. Hence, in the actual case, the motion on the
near side will be made up of that due to the given sources P
and of that due to the images P'. It may be mentioned that
the present case of a rigid plane boundary is the only one where
the physical " image " of a pointsource is itself accurately a
pointsource.
The problem of reflection at the plane boundary of two
distinct fluid media has been discussed in 61, in the case of
direct incidence. The case of oblique reflection was solved by
Green (1847). The results are chiefly of interest for the sake
of the optical analogies, but one curious point, noticed by
Helmholtz, may be mentioned. Owing to the greater velocity
of sound in water, the conditions for total reflection may occur
when the waves are incident from air on water (in fact when
ever the angle of incidence exceeds about 13), but not in the
converse case. This is of course the reverse of what holds with
regard to light.
216 DYNAMICAL THEORY OF SOUND
74. Refraction due to Variation of Temperature.
Questions relating to wavepropagation in heterogeneous
media can only be discussed in a general way, and with the
help of conceptions borrowed from geometrical optics. If at
any surface there is an abrupt change of properties the law of
propagation is of course altered. If the dimensions of the
surface, and its radii of curvature, are large compared with
the wavelength, we have phenomena of regular reflection and
refraction, as in optics. Cases of absolute discontinuity are of
course not met with in the atmosphere, but the theory would
be practically unaffected if the change of properties were
effected within a space which is small compared with the
wavelength.
When on the other hand we have a continuous variation
such that the change of properties within a wavelength is
negligible, the case is analogous to that of atmospheric
refraction of light, which is discussed in books on optics and
astronomy. In an atmosphere of the same gas, at rest, a
variation in the velocity of sound can only arise through a
variation of temperature ( 59). The refraction due to varia
tion of temperature with altitude was first discussed by Osborne
Reynolds (1876). Suppose that, as usually happens, the
temperature diminishes upwards. Since the velocity of sound
varies as the square root of the absolute temperature, the lower
portions of a wavefront will be propagated faster than the upper
ones, so that a front which was originally vertical gets tilted
upwards more and more as it proceeds. The sound will there
fore, for the most part, pass over the head of an observer at
a sufficient distance, such residual effects as he perceives being
referable to diffraction. On the other hand, whenever the
temperature increases upwards the waves will be tilted down
wards, and the effect at a distance will be greater than if the
temperature had been uniform. This latter condition of the
atmosphere sometimes prevails on a clear night following a
warm day, when, owing to the cooling of the ground by
radiation, the lower strata of the atmosphere are reduced in
temperature relatively to the upper ones.
The theory has been further developed by Lord Rayleigh,
by means of the conception of rays of sound. The surfaces of
GENEKAL THEORY OF SOUND WAVES
217
equal wave velocity being supposed to be horizontal, each ray
will travel in a vertical plane. The cur
vature of a ray may be calculated directly
by a method due to Prof. James Thomson*.
If R be the radius of curvature, the two
wavefronts passing through the extremities
of an element 8s of the path will be
inclined at an angle 8s/R, and if 8s be the
length intercepted on an adjacent ray in
the same vertical plane, we have
are
(2)
.(3)
where 8n denotes the distance between
the two rays, the standard case being
that shewn in the figure. Since the elements
described in the same time we have
8s _ 8s
iT+8c"~"c'
whence, by comparison with (1),
1 1 dc
IR ~ ~cdn'
When the temperature diminishes upwards, 9c/9n is negative
and the curvature l/R is positive, as in the figure, and the rays
are curved upwards. But if the temperature increase upwards,
the curvature is downwards, so that an observer at the level of
the source may hear sounds which would otherwise have been
intercepted by obstacles.
The formula (3) leads to the ordinary law of refraction. If
>Jr be the inclination of the ray to the horizontal we may write
dc dc dc dc .
5 = j cos y, = jsm\lr, (4)
on dy ds dy
if y be the vertical coordinate. Hence, along the course of
a ray,
i?*lr 1 1 tJ.r.
.(5)
ds
1
^D
R
Idc
y
cds
* James Thomson (1822 92), professor of engineering at Belfast 1857 72,
and at Glasgow 1872 89.
218 DYNAMICAL THEOEY OF SOUND
or c sec Jr = const., ..................... (6)
which is the law in question. Conversely, from (6) we can
derive the formula (3). When c is known as a function of y
the equation (6) determines the paths.
The simplest hypothesis is that the temperature decreases
(or increases) upwards with a uniform gradient. This includes
the particular case of an atmosphere in " convective equili
brium " under gravity, where the gradient is
 7 " 1  (7)
'IP
H being the height of the homogeneous atmosphere ( 59)
corresponding to the temperature 6*. This is at the rate of
about 1C. per 100 metres. If a law of uniform decrease were
to hold without limitation, we should at a certain altitude
meet with a zero temperature (absolute). If for a moment
we take the origin at this level, and draw the axis of y
downwards, the temperature will be proportional to y t and the
wave velocity c to y?. Hence by (6) we have, along any ray,
...................... (8)
The paths are therefore cycloids, the generating circles of which
roll on the under side of the line y = 0. If on the other hand
the temperature increases upwards with a uniform gradient, the
paths are the cycloids whose generating circles roll on the upper
side of the line which corresponds to the zero of temperature.
In any practical case we are concerned only with the portions
of the curves near the vertices. The arcs may therefore be
taken to be circular, with a radius double the distance below
(or above) the level of zero temperature. In the extreme case
of upward diminution to which the formula (7) refers, this
radius will therefore be (roughly) 2 x 273 x 100 = 54600 metres,
for a temperature of C.
* It was pointed out by Lord Kelvin (1862) that this is the condition into
which the atmosphere would be brought by the free play of convection currents
alone, without conduction or radiation. It is therefore one of neutral equilibrium.
If the temperature diminish upwards at a greater rate the equilibrium becomes
unstable.
GENERAL THEORY OF SOUND WAVES 219
75. Refraction by Wind.
Another interesting question is that of refraction by wind.
A uniform motion of the medium introduces of course no
complication, the relative motion of the sound waves being
exactly the same as if the medium were at rest. Usually,
however, the wind velocity near the ground is less than above,
the motion of the lower layers of air being obstructed. Hence
when a wavefront travels with the wind, the upper portions
are propagated (in space) somewhat faster than the lower, the
velocity of the wind being superposed on that of sound. The
front is therefore continually being tilted downwards. For a
similar reason a wavefront travelling against the wind gets
tilted upwards, so that the sound tends to pass over the
head of an observer at a distance. This explanation of the
familiar fact that sound can be heard better, and further from
the source, when this lies to windward than when it is to
leeward of the observer, was first given by Stokes (1857).
The only previous suggestion had been that a sound which
has travelled a certain distance with the wind has really
traversed a shorter length of air. and has consequently become
less attenuated by spherical divergence, than if the wind had
been absent. Owing to the smallness of wind velocities in
comparison with that of sound, this cause is quite inadequate
to explain the very marked effects which are observed. The
true theory was discovered independently by Reynolds (1874),
and confirmed by a number of interesting experiments.
If we proceed to apply optical methods to the question,
it is necessary to dis
tinguish, as in the theory
of aberration, between the
direction of a ray and that
of a wavenormal. Let S l
represent the position of a
wavefront at time t, S'
the position at time t + 8t
of those particles which
were on S l} and 8 2 the
new position of the wave
220 DYNAMICAL THEOKY OF SOUND
front. Let P l be any point on Si, and P' the corresponding
point on S f , so that Pf is the path of a particle of the
medium in the time Bt. On the principles of optics, the new
position $ 2 of the wavefront is obtained as the envelope of
a system of spheres of radius c$t, described with the various
points P' of S' as centres. If P 2 be that point on the
envelope which corresponds to P', P^P Z will be an element of
a ray, and P'P Z an element of the wavenormal. Also since
Pf U8t, where U is the velocity of the medium, the " ray
velocity" (PiPJfo) is the resultant of the wave velocity and
the velocity of the medium.
In the present question the velocity U is horizontal, and
a function of the altitude (y) only. If i/r, < denote the
inclinations to the horizontal of the ray and the wavenormal,
respectively, we have
n V9 T)~ c s Q r> ()
or <f> = TT H simjr, (2)
c
if U/c be small, as will usually be the case.
To ascertain the law governing
the change of direction of the ray,
consider first the case of refraction at
the common horizontal boundary of
two uniform currents U, U'. If <, <'
be the inclinations of the wavenormal
on the two sides of the plane of
discontinuity, we have
c sec < 4 U = c sec </>' + U', (3)
each side expressing the horizontal velocity of the trace of
the wavefront on the plane in question. Since a continuous
variation of U can be approximated to by a series of small
discontinuities, we infer that (3) will still hold if <, U and
<f>', U' refer to any two positions on the same ray. This gives
the altered law of refraction. Lord Rayleigh points out that
since sec $ <%. 1, <f>' will become imaginary if
(U' Z7)/c> sec </>  1 (4)
GENERAL THEORY OF SOUND WAVES 221
There is therefore total reflection, at the stratum to which the
accents refer, of all wavefronts whose initial inclination (0) to
the vertical falls short of a certain limit.
Along any one ray we have
sec < H = const., (5)
c
or, by (2), sec i/r + sec 2 ^ = const., ....(6)
provided ^ be not too great. If we differentiate this with
respect to the arc s, and put d^/ds = 1/R, dy/ds = sin ir, we
find
1 /, 20" .\ IdU /lyx
r . 1 +  sec \Ir =__ (7)
E\ c r J c dy
The ray is therefore curved downwards or upwards, according
as dU/dy is positive or negative, i.e. according as the ray is
travelling with or against the wind. If the gradient dU/dy
be uniform, the rays have
all the same uniform curva ri
ture, approximately, owing
to the smallness of U/c,
unless indeed the inclination Fig 71
vjr becomes considerable. It
will be noticed that in this problem the path of a ray is
not reversible.
This is a convenient place for a reference to what is
known as " Doppler's principle " *. Suppose, for instance,
that a periodic source of sound is approaching a stationary
observer. The number of maxima of (say) the condensation s
which strike the ear of the latter in a second is increased, and
the pitch is therefore raised. The diminution in the period is
to the period when the source is at rest in the ratio of the
velocity of approach to the velocity of sound. When the
source recedes from the observer, this ratio is negative, and
the pitch is lowered. When the motion of the source is
oblique to the rays by which the sound is heard, the com
* Christian Doppler (1803 54), an Austrian mathematician, professor of
physics at Vienna 1851.
222 DYNAMICAL THE OK Y OF SOUND
ponent of its velocity in the direction of the ray is alone
effective. Analogous effects are produced when the source is at
rest and the observer in motion. The principle is exemplified
in the apparent change of pitch of the whistle of a locomotive
as a train dashes through a station ; but its most striking and
fruitful applications are met with in the theory of radiation.
CHAPTER VIII
SIMPLEHARMONIC WAVES. DIFFRACTION
76. Spherical Waves. PointSources of Sound.
From this point it is convenient to consider specially the
case of simpleharmonic vibrations. In problems relating to
the impact of sound waves on obstacles, or their transmission
by apertures in a screen, and so on, the results will vary in
character with the pitch, the determining element being the
relation between the wavelength and the linear dimensions of
the obstacles, &c.
It will be desirable, for the sake of conciseness, to use
imaginary quantities somewhat more freely than in the pre
ceding chapters. Thus we assume that the velocitypotential
< varies as e int , or e ikct , where
..................... (1)
if X be the wavelength of plane waves of the same period 2?r/n.
The general equation of sound waves ( 70 (4)) therefore be
comes
Vty + Afy = ...................... (2)
In the case of plane waves whose fronts are perpendicular
to the axis of x, we have
+"*<> ..................... < 3 >
the solution of which may be written
<f> = Ae ikx + Be ikx , ..................... (4)
or <f> = C cos kx + D sin kx, ............... (5)
224 DYNAMICAL THEOBY OF SOUND
the timefactor e int being understood. Thus a train of simple
harmonic waves travelling in the direction of ^positive is
represented by
When we proceed to calculations of energy it is of course
necessary to revert to real forms. Thus, taking the real part
of (6), we have
<I> = A cos k(ct x). '. ................. (7)
The mean energy per unit volume, as given by 70 (7), (8),
is ^pk*A 2 , and the mean energy transmitted per unit time,
per unit area of the wavefront, is
%pk 2 cA 2 , or $pn*/c.A* ................ (8)
We may call this the " energyflux " in the wavesystem (7).
The equation of symmetrical spherical waves, 71 (6), now
takes the form
^ + *W)0, .............. ....(9)
and the solution is
r<t> = Ae~ ikr + Be ikr , .... .............. (10)
or r<f> G cos kr + D sin AT, ............ (11)
the timefactor being understood as before. The two terms
in (10) correspond to waves diverging from, or converging to,
the origin, respectively. In particular, the diverging waves
due to a source Ae ikct at the origin are represented by
(12)
or, in real form, <t> = ~ A oosnff ] ............ (13)
\ c)
This is of course a particular case of 73 (1).
The maintenance of such a source in an unlimited medium
requires a certain expenditure of energy. The work done per
unit time at the surface of a sphere of radius r, on the fluid
outside, is the product of the pressure, the area, and the
outward velocity, or
. (14)
SIMPLEHAKMONIC WAVES. DIFFRACTION 225
It is evident that p contributes nothing to the average effect,
since the mean value of d<f>/dr at any point is zero. If we
substitute from (13) we find that the average of the remaining
part is
This quantity W is independent of r, as was to be anticipated,
since the mean energy in the space included between two
concentric spheres is constant. It measures the emission of
energy (per unit time) by the source. The formula may also
be inferred from the consideration that at a great distance the
waves may be regarded as plane. If in (8) we replace A by
the A/4>7rr of (13), and multiply by 4nrr*, we obtain the result
(15).
It must be remembered that this calculation of the energy
emitted applies only to an isolated source in free space. A
source placed in an enclosure with rigid walls does no work on
the whole, since the energy of the gas is constant. Even in an
open space the emission of energy may be greatly modified by
the neighbourhood of an obstacle. Thus in the case of a source
P close to a rigid plane boundary the amplitude of vibration at
any point is doubled by the reflection as from the image P'
( 73); the intensity is quadrupled, and the emission (on one side)
is therefore twice that of an equal source in free space.
The equation j^ ****>, .................. (16)
7? / v\
or, in real form, </>= ^  cos n U + ),
^rirT \ c/
may likewise be interpreted as representing a " sink " of sound,
i.e. a point where energy is absorbed, under similar conditions,
at the rate ri 2 pB*l&Trc. This conception is however of no great
assistance in acoustics.
The notion of a simple source, valuable as it is for theoretical
purposes, is seldom realized even approximately in practice.
A vibrating body such as a membrane, or either prong of a
tuning fork, is tending at any instant to produce a condensation
of the air in contact with it on the one side and a rarefaction
L. 15
226
DYNAMICAL THEORY OP SOUND
on the other, and is therefore more adequately represented, in
the simplest cases, by a combination of two simple sources near
together but in opposite phases. Idealizing this a little further
we are led to the mathematical conception of a "double source."
We begin with a simple source of strength m at a point 0,
and a simple source of strength f m at an adjacent point 0',
the signs indicating the oppo
sition of phase. If we next
imagine m to become infinitely
great, whilst the distance 00'
becomes infinitely small, in such
a way that the product m.OO'
remains finite, we have the ideal
" double source " of theory. The
direction 00' is called the
" axis," and the limit of m.OO'
is called the " strength." The resulting motion is evidently
symmetrical about the axis.
If the direction 00' be that of the axis of x t and be
taken as origin, the velocitypotential at P due to simple
sources + m at ' and 0, respectively, will be given by
p ikr\
 V)'
where r= OP, r' = O'P. If we draw PP' equal and parallel to
O'O, we have /= OP', and the expression in brackets is equal
to the change of value of the function e~ ikr /r caused by a
displacement of P to P'. Hence, ultimately, if P'P = 8x,
Putting m&x=I, we deduce the formula for a unit double
source at 0, having its axis along Ox, viz.
(20)
this is a particular case of 73 (3). When x alone is varied,
whilst y and z are constant, it appears from the figure that
SIMPLE HARMONIC WAVES. DIFFRACTION 227
Sr = cos &c, where 6 denotes the inclination of OP to Ox.
Hence d/dx = cos 6 9/9r, and
(21)
Performing the differentiation, we find
(22)
For small values of kr, i.e. within distances from which are
small compared with X/2w, this becomes
...................... (23)
On the other hand, for large values of kr,
,,ikr
tfc  cos0, (24)
r
so that along any one radius vector the condensation (s =
varies ultimately as 1/r. The radial and transverse components
of the velocity are to be found by the formula (6) of 69 ;
viz. they are 9$/9r and 9</>/r90, respectively. It appears
that near the origin these are of the same order of magnitude,
whilst at a great distance the lateral velocity is less than the
radial in the ratio 1/Ar.
Introducing the factor Ce int in (24), and taking the real
part, we find that the velocitypotential due to a double source,
of strength G cos nt y at a great distance, is
bC! / r\
 sin n (* cos ............. (25)
4?rr \ cj
The waves sent out in any direction are therefore ultimately
plane, of the type (7), provided A=kCcos6/4<7rr, the mere
difference of phase being disregarded ; and the flux of energy
(across unit area) will therefore be pk*cC 2 cos* ^/32?r 2 r 2 . Multi
plying by 27rrsin 6. rSO, which is the area of a zone of a
spherical surface of radius r bounded by the circles whose
angular radii are and 6 f 0, and integrating from 6 to
6 = TT, we find that the total emission of energy by the double
source C cos nt is
<>
152
228 DYNAMICAL THEOEY OF SOUND
It will be noticed that as the wavelength X is increased,
and k accordingly diminished, the fundamental equation (2)
tends to assume the form
W> = ........................ (27)
which is met with in the dynamics of incompressible fluids, and
in the theories of attractions and of electric and thermal con
duction. This assimilation may come about in two ways, either
through a diminution in the frequency (n/Zir), or by an increase
in the elasticity of the medium and consequently in the wave
velocity. Under the same condition the formula (12)
approximates to the form
which is the expression for the potential of a magnetic pole, or
for a source of electricity, and so on ; whilst in the case of the
double source (21) the limiting form is (23), which is recognized
as the potential of an infinitely small magnet.
A further remark of great importance is that within any
region, free from sources, whose dimensions are small compared
with X, the configuration of the equipotential surfaces <= const.
is at any instant sensibly the same as if the fluid were incom
pressible. For the value of </> due to an external source differs
from its value in the case of incompressibility chiefly by a
factor e~ ikr , where r denotes distance from the source. If
6 denote the greatest breadth of the region, this factor can at
most vary in the ratio e~ ikb , which differs very little from unity
when kb is small.
77. Vibrating Sphere.
By means of the fiction of a double source, of suitable
strength, at the centre, it is possible to calculate the waves
generated in the surrounding air by a vibrating solid sphere of
any radius. As this is almost the only problem of the kind
which can be completely solved we devote some space to it. The
work is simple, and the results throw a good deal of light on
other cases.
For reasons just referred to, it is instructive to look first at
the case where the fluid is incompressible. We take the origin
SIMPLEHAKMONIC WAVES. DIFFRACTION 229
at the mean position of the centre of the sphere, and the axis of
x along the line of its vibration ; and we denote its velocity
by U. The velocity of the fluid in contact with the sphere at
any point P, resolved in the direction of the normal, must be
equal to the normal component of the velocity of the point P of
the sphere itself, i.e. to UcosO, where is the angle POx.
This gives
 d ^=Ucos0 [r=a], ................ (1)
if a be the radius. The velocity due to a double source at in
an unlimited mass of incompressible fluid is of the form
sfl; ..................... (2)
and in order that this may be consistent with (1) we must have
C=27ra*U. ........................ (3)
With this determination of C the effect of the sphere on the
fluid is exactly that of the double source, and the solution of
our problem is
*=g 3 cos0. ..................... (4)
This depends only on the instantaneous value of U, as we should
expect, since under the present hypothesis disturbances are
propagated with infinite velocity. It should also be noted that
there is so far no assumption that U is small.
The directions of motion at various points of the field may
be shewn by tracing the " lines of motion," which are lines
drawn from point to point, always in the direction of the
instantaneous velocity. In the case of small vibratory motion,
which we have especially in view, each particle oscillates
backwards and forwards through a short distance along the line
on which it is situate. If Sr, rS0 be the radial and transverse
projections of an element of such a line, these quantities must
be proportional to the radial and transverse components of
velocity, viz. d<f)/dr and d<l>/rd6, respectively. Hence
Sr rS0
cose Jsinfl' '
the integral of which is
........................ (6)
DYNAMICAL THEORY OP SOUND
where 6 is a parameter which varies from one line of motion
to another. The curves, which are identical in form with the
lines of force due to a small magnet, are shewn in Fig. 73.
Fig. 73.
To calculate the reaction on the sphere we divide the surface
into zones by planes perpendicular to Ox. The area of a zone
being 27ra 2 sin0S0, the resultant force on the sphere in the
direction of ^positive is
(7)
The constant part of the pressure contributes nothing to the
resultant. The variable part is, if terms of the second order in
the velocities be neglected,
.(8)
SIMPLEHAKMONIC WAVES. DIFFRACTION 231
since t enters only through U. Substituting we find
(9)
The remarkable point here is that the force is independent
of the velocity, and depends only on the acceleration of the
sphere. If the mass of the sphere be M, and if it be subject to
other extraneous force X, its equation of motion will be
(10)
or (*+f/^~Z ................... (11)
This is the same as if the fluid were abolished, and the inertia
of the sphere were increased by Trpa 3 , i.e. by half that of the
fluid which it displaces. It was shewn by Stokes (1843) that
this conclusion is accurate even when the restriction to small
motions is abandoned.
There is, as we shall see ( 79), nothing peculiar to the
sphere in the general character of the above result, but the
apparent addition to the inertia will vary of course with the
shape as well as the size of the solid, and will usually be
different for different directions of motion, as e.g. in the case of
an ellipsoid. The theory here touched upon has had a great
influence on recent physical speculations, and is responsible
ultimately for the suggestion that the apparent inertia of
ordinary matter may be partly or even wholly due to that of a
surrounding aetherial medium.
Turning now to the acoustical problem, let the velocity of
the sphere be expressed symbolically by
U = Ae int ...................... (12)
The surfacecondition will have the same form (1) as before.
The velocitypotential of a double source Ce in * at is
by 76 (21), the timefactor e int being omitted. The ratio of C
to A is then determined by (1).
232 DYNAMICAL THEOKY OF SOUND
The most interesting case is where the radius a of the
sphere is small compared with X/2?r, where X is the wavelength.
In the immediate neighbourhood of the sphere kr will then be
small, and the formula (13) is, for this region, practically
identical with (2). It follows that
<7=2iraM, ..................... (14)
nearly, and further that the lines of motion near the sphere
will have sensibly the configuration shewn in Fig. 73. The
apparent addition to the inertia of the sphere has very
approximately the same value f Trpa? as before. On the other
hand, at distances r which are comparable with, or greater
than, X, the motion of the fluid is altogether modified by the
compressibility. At sufficiently great distances we have, by
(13) and (14),
<t> = ^ika s A e ^cos0, ............... (15)
or, in real form,
S ff, ......... (16)
corresponding to a velocity
U = Acosnt ..................... (17)
of the sphere. The amplitude now varies ultimately as 1/r,
instead of 1/r 2 , as in the case of (4).
The investigation so far discloses nothing analogous to
a frictional resistance, whereas we know that owing to the
generation of waves travelling outwards a continual abstraction
of energy must take place. To calculate either the dissipative
resistance, or the work done, at the surface of the sphere, we
should have to use the complete formula (13); but the emission
of energy may be ascertained independently from the formula
(26) of 76. The strength of the equivalent double source
being given approximately by (14), we find
W = %7rpk*a 6 cA 2 ................... (18)
If p' denote the mean density of the sphere, its energy when
vibrating under the influence of (say) a spring will be
SIMPLE HARMONIC WAVES. DIFFRACTION 233
If, following a procedure explained in 11, we equate the rate
of decay of this energy to W, we find
and therefore A = A e T , ..................... (20)
8 o
provided T= /y ... ................... (21)
p
The ratio (nr/^Tr) of the modulus of decay to the period is
therefore usually very great.
78. Effect of a Local Periodic Force.
Corresponding results can, with the help of more or less
intuitive considerations, be obtained for other forms of vibrating
solid, but the work is much simplified by a preliminary theorem,
which has also an independent interest. This relates to the
effect of a periodic extraneous force concentrated about a point
in a gaseous medium.
An elementary proof can be derived at once from the pre
ceding investigation. The result will obviously be the same if
the force be imagined to act on an infinitely small sphere
having the same density as the surrounding fluid. The effect
is therefore that of a double source ; and if we now denote the
concentrated force, supposed acting parallel to x, by Pe int , we
find, putting M = f Trpa 3 in 77 (11),
P=2i7rpkca 3 A, ..................... (1;
and therefore, by 77 (15), for large values of kr,
p p ikr
<t> = T~ cos<9 ................... (2)
4f7TpC T
The following investigation is of a more formal character;
but it involves mathematical processes more intricate than
those which are employed in other parts of this book. The
work depends on the solution of the equation
V0+#* = <l>, ..................... (3)
where < is a given function of x, y, z which vanishes outside a
certain finite region R. In the theories of attraction, and of
thermal and electric conduction, we meet with the equation
*, ........................ (4)
234 DYNAMICAL THEORY OF SOUND
where <I> represents a distribution of density (p = <l>/47r), or
of sources of heat, &c. The solution of (4) appropriate to
infinite space (when there are no sources at infinity) is known,
viz. it is
where 4>' denotes the value of <3> at (of, y', z / ), r denotes distance
from this point to the point P, or (x, y, z), for which the value
of < is required, and the integration extends over all space for
which <E> differs from 0. For example, if we put <!>' = 4Trp',
we get the ordinary expression for the gravitation potential of
a continuous distribution of matter.
The analogous solution of (3) is
(6)
This represents a distribution of simple sources through R, the
strength per unit volume being 4>, and it is therefore obvious
at once that the equation V 2 < + A^ = is satisfied at all points
P external to R. The only question of any difficulty arises
when P is inside R. We then divide R into two regions R l
and R 2) of which R 2 encloses P and is ultimately taken to be
infinitely small in all its dimensions. The parts of at P due to
the sources in R l and R% , respectively, may be denoted by fa and
< 2  Since P is external to ^ we have V 2 ^ + A; 2 ^ = as before.
Within R 2 we may ultimately put e~ ikr l t and <f> 2 then
approximates to the gravitation potential of matter of density
4>/47r restricted to the space R 2 . We have then, ultimately,
on known principles, V 2 < 2 = <I> and < 2 = 0. Hence (1) is satisfied
by <f> = </>! + </> 2 . It is further evident that (6) is the solution
of (1) consistent with the condition that there are no sources of
sound except at the points to which <l> refers.
When forces X, Y, Z per unit mass act at the various
points of a gaseous medium, the equations (4) of 68 are
replaced by
9 <> +z , g = _ c ,^ + r , *?__<* + *. ...(7)
dt dx dt dt dt dz
SIMPLEHARMONIC WAVES. DIFFRACTION 235
If we eliminate u, v, w by the kinematical relation (1) of 70
we obtain
(8)
If X, Y, Z, s all vary as e ikct , this becomes
V*s + k*s = \div(X, Y,Z), ............ (9)
c
in the notation of 67. This is of the form (3), and the
solution is therefore
1 niftX' ar dZ'\e ik ' ,, ,,,
8 =  A ; *T + 57 + *i)  dxdy dz'
4>7rc z JJj\dx dy dz J r
The transformation is effected by partial integration of the
several terms, the integrated portions vanishing at infinity if
X', Y 7 , Z' do so. Also, since
r = J[(x  xj +(y yj + (z
we have
whence
 <>
From this the value of ^> follows by the relation
(12)
As a particular case, suppose that Y' = 0, Z' 0, and that
X' differs from only in a small region about the origin,
and put
pjjjX'dx'dy'dz' = P ................ (13)
We have < =  r ^ j (} , ............... (14)
dx\ r J
or, for large values of kr,
P pikr
cos(9, ............... (15)
r
as before. Comparing with 76 (24) we see that a concentrated
force Pe int has the effect of a double source of strength iP/pkc.
236 DYNAMICAL THEORY OF SOUND
79. Waves generated by Vibrating Solid.
We return to the problem of investigating the waves
generated by a vibrating body. In order not to complicate the
question too much we will assume that the body has some sort
of symmetry with respect ta an axis ; thus it may be a form
of revolution about this axis, or it may have two mutually
perpendicular planes of symmetry meeting in this axis, or
(again) a single plane of symmetry perpendicular to the axis.
In any case this axis is taken to be the direction (Ox) of
vibration.
The dimensions of the solid being supposed small in com
parison with \/27r, the motion of the fluid in the immediate
neighbourhood will be sensibly the same as in the case of
incompressibility, and the principal effect on the body will there
fore be equivalent to an increase of inertia. To establish this
latter point in a general manner, we note that the (irrotational)
motion of a frictionless liquid due to the motion of a solid in it
will have the velocity at every point in a determinate ratio
to the velocity U of the solid, and that the total kinetic energy
of the fluid may therefore be expressed by ^pQ'U 2 , where Q' is
a constant, of the nature of a volume, depending only on the
size and shape of the solid and the direction of its vibration.
Hence if M be the mass of the body, the equation of energy
takes the form
(l)
where the righthand member represents the rate at which
work is being done by the extraneous force X. Thus
X, .................. (2)
which shews that the inertia of the body is apparently increased
by the amount pQ'. An equivalent statement is that the
reaction of the liquid is equivalent to a force pQ'dU/dt.
In the actual case of the gaseous medium, it is plain that if
the solid were removed, and its place supplied by fluid, the
motion at a distance would be very approximately the same as
would be produced by a suitable periodic force from without,
SIMPLEHARMONIC WAVES. DIFFRACTION 237
acting on the substituted matter. Since this force has to
produce an acceleration of momentum pQdU/dt, where Q is
the volume displaced by the solid, as well as to balance the
reaction just referred to, its amount would be
, ...... (3)
if U=Ae int ......................... (4)
By 78 (2), the velocitypotential at a great distance r will
therefore be
cos 6 = Ae^ cos 6. ...(5)

4TT/oc r 4Trr
Comparing with 76 (24) we see that the effect of the vibrating
solid is equivalent to a double source of strength C = (Q+ Q')A,
and that the emission of energy is accordingly
(6)
by 76 (26). In the case of the sphere we have Q = %7ra?,
Q'= iQ, and the result accordingly agrees with 77 (18). It
can be shewn that for a circular disk of radius a, moving
broadside on, Q' = ^ira 3 , whilst Q of course =0.
80. Communication of Vibrations to a Gas.
The circumstances which govern the efficiency of a vibrating
body in generating sound waves, and the comparative effects in
different gases, were elucidated by Stokes in a classical memoir
" On the Communication of Vibrations from a Vibrating Body
to a surrounding Gas*." The starting point of the investigation
was an observation by Prof. J. Leslie (1837), who found that the
sound emitted by a bell vibrating in an atmosphere of hydrogen
was extremely feeble as compared with the effect in air. No
satisfactory explanation of this phenomenon was forthcoming
up to the time of Stokes' paper. The essence of the matter is
conveyed in the following quotation :
" When a body is slowly moved to and fro in any gas, the
gas behaves almost exactly like an incompressible fluid, and
* Phil. Trans. 1868. The passage which follows below is from the
'* abstract " in the Proc. Roy. Soc.
238 DYNAMICAL THEORY OF SOUND
there is merely a local reciprocating motion of the gas from the
anterior to the posterior region, and back again in the opposite
phase of the body's motion, in which the region that had been
anterior becomes posterior. If the rate of alternation of the
body's motion be taken greater and greater, or, in other words,
the periodic time less and less, the condensation and rarefaction
of the gas, which in the first instance was utterly insensible,
presently becomes sensible, and sound waves (or waves of the
same nature in case the periodic time be beyond the limits of
audibility) are produced, and exist along with the reciprocating
flow. As the periodic time is diminished, more and more of the
encroachment of the vibrating body on the gas goes to produce
a true sound wave, less and less a mere local reciprocating flow.
For a given periodic time, and given size, form, and mode of
vibration of the vibrating body, the gas behaves so much the
more nearly like an incompressible fluid as the velocity of
propagation of sound in it is greater ; and on this account the
intensity of the sonorous vibrations excited in air as compared
with hydrogen may be vastly greater than corresponds merely
with the difference of density of the two gases."
These remarks are exemplified in the results of 77 (13),
(14). If we fix our attention on a point at a distance from the
sphere, supposed vibrating with the velocity
U=Acoant, (1)
the motion there is given, when the period is sufficiently long,
by the formula
A a 3
<f> Y7 cos Q . cos nt, (2)
as if the fluid were incompressible. But when the frequency is
increased until the wavelength is small compared with the
distance r from the centre, the appropriate formula is
ka?A / r
?A A . f.r\ /ox
cos . sin n(t j , (3)
and the amplitude is accordingly greater than in the former
case in the ratio kr, or 2irr/\. For the same frequency, the
amplitude, which depends on k/c or n/c 2 , will in different gases
SIMPLEHARMONIC WAVES. DIFFRACTION 239
now vary inversely as the square of the wavevelocity. Again,
the emission of energy is, by 77 (17),
W = lirpkWcA* = lirpnW/c?.A*, (4)
and so varies (for the same gas) as the fourth power of the
frequency. The emission in different gases will (for the same
frequency) vary inversely as the fifth power of the wave
velocity, if we assume ( 59) that the latter varies inversely as
the square root of the density. For instance it will be about
1000 times less in hydrogen than in oxygen.
In order further to illustrate the effect of the lateral motion
of the gas, near the surface of the sphere, from the hemisphere
which is at the moment moving outwards to that which is
moving inwards, in weakening the intensity of the waves
propagated to a distance, we may calculate what the emission
would be if this lateral motion were prevented. For this
purpose we may (after Stokes) imagine a large number of fixed
partitions to extend radially outwards from near the surface.
In any one of the narrow conical tubes thus formed, the motion
will be of the same character as in the case of symmetrical
spherical waves. Now a uniform radial velocity C cos nt over
the surface of a sphere would be equivalent to a simple source
4Tra 2 C cos nt, and the corresponding emission per unit area
would be %tfa?pc C*, by 76 (15). If we now put G= A cos 6,
and integrate over the surface, we get the total emission in our
system of conical tubes. The result is
W ' = 1rrlcWpcA*, (5)
since the average of cos 2 6 for all directions in space is J . If
we compare this with (4), we see that the effect of the lateral
motion is to diminish the emission in the ratio J& 2 a 2 .
When, as for example in the case of a plate or a bell, the
surface is divided by nodal lines into a number of compart
ments vibrating in opposite phases, the opportunity of lateral
motion is increased, and the emission of energy correspondingly
weakened. For facility of calculation Stokes took the case of
a spherical surface, with various symmetrical arrangements of
nodal lines. In the problem of the oscillating sphere we have
one such line, viz. the great circle 6 = \ TT, and the emission, as
240 DYNAMICAL THEORY OF SOUND
we have just seen, is diminished by the lateral motion in the
ratio k' 2 a?. For a spherical surface with two nodal great circles
meeting at right angles the effect is much greater, the ratio
being fakta*. And as we increase the number of compartments
into which the sphere is divided, the ratio, already very small,
decreases with enormous rapidity.
For the sake of simplicity it has been assumed in the
preceding statements that the perimeter 2na of the sphere is
small compared with X. The influence of lateral motion is
however not confined to this case, but will make itself felt
whenever the dimensions of the compartments referred to are
small compared with X. In the case of the oscillating sphere
there is no difficulty in working out the result without any
restriction to the value of ka, starting from the formula (13) of
77.
Stokes has also investigated mathematically the case of a
cylinder vibrating at right angles to its length, where the same
cause is of course operative. In this way an estimate is
obtained of the direct effect of a vibrating string in generating
airwaves. This involves the ratio of the perimeter of the
crosssection of the string to the length of the airwaves, and
is in any practical case extraordinarily minute. As explained
in 24, almost the whole of the sound given out when a piano
string is struck comes from the sounding board.
81. Scattering of Sound Waves by an Obstacle.
We have next to consider the disturbance produced in a
train of sound waves by a rigid obstacle whose dimensions are
small compared with the wavelength. The scattered waves
which are sensible at a distance are due mainly to two causes.
If the obstacle were absent the space which it occupies would
be the seat of alternate condensations and rarefactions. The
effect of the obstacle in refusing to execute the corresponding
contractions and expansions of volume is, at a distance,
approximately the same as if in a medium otherwise at rest its
volume were to undergo a periodic change of just the opposite
character. The result is equivalent to a simple source. On
the disturbance thus produced there is superposed a second
SIMPLEHARMONIC WAVES. DIFFRACTION 241
wavesystem, which is due to the immobility of the obstacle.
If the latter were freely movable, and if it had moreover the
same density as the surrounding air, it would swing to and fro
with the airparticles, and the second wavesystem would be
absent. This system is accordingly the same as would be
produced if the obstacle were constrained to oscillate with
a motion exactly equal and opposite to that of the air in the
primary waves when undisturbed. The effect is, as we have
seen in 79, that of a double source. It might appear, at first
sight, that the former of these disturbing influences would be
much less important than the second, but in its effect at a
distance it becomes comparable, owing to the greater attenuation
by lateral motion of the waves proceeding from a double source.
If Q be the volume of the obstacle, the strength of the
simple source due to the first cause is
where s, <f> refer to the primary waves. In the case of a system
of plane waves
incident on a small obstacle at 0, this gives a velocitypotential
As regards the second cause, we will assume for simplicity
that the obstacle has the degree of symmetry postulated in
79 with respect to the direction (Ox) of the vibration in the
airwaves. If the wavesystem (2) were undisturbed, the
velocity of the airparticles at would be represented
symbolically by ikC, and the strength of the double source due
to the obstacle moving with this velocity reversed would be
ik(Q + Q') (7, in the notation of 79. The scattered waves at
a distance, due to the immobility, are therefore represented by
(4)
by 76 (24). The complete result is given by < = fa + < 2 .
It follows that the amplitude of the scattered waves at any
L. 16
242 DYNAMICAL THEORY OF SOUND
distant point is, for similar forms, directly proportional to the
volume of the obstacle and inversely proportional to the
square of the wavelength. This latter particular might have
been foreseen without calculation. The ratio to the original
amplitude must necessarily vary directly as the volume Q,
and inversely as the distance r, and in order that the result
may come out a pure number we must divide by X 2 , since X
is the only other linear magnitude involved. The emission
of energy, being proportional to the square of the amplitude,
will therefore vary as X~ 4 . This law of the inverse fourth
power holds also in optics, and for a similar reason, with respect
to the scattering of light by particles whose dimensions are
small compared with the dimensions of the lightwaves. The
blue of the sky, for instance, is attributed to the relative
preponderance of the shorter waves in the light scattered by
the molecules of air, and possibly by other particles ; in the
transmitted light, on the other hand, the longer waves pre
dominate. The theory is due to Lord Rayleigh, who has also
pointed to an acoustic illustration in what are called " harmonic
echoes." If a composite musical note, consisting of a funda
mental tone with its octave, &c., be sounded near a grove
of trees, for example, the ratio of the intensity of the octave to
that of the fundamental will in the scattered sound be 16 times
what it was in the original note. The scattered sound may
therefore appear to be raised in pitch by an octave.
The actual scattering of energy is found by adding the
results due to the simple and the double source. This may be
proved by calculating the work done at the surface of a sphere
of large radius r. The terms due to the combined action of the
two sources contain a factor cos 9, and so disappear when
integrated over the surface. Hence, by 76 (15), (26),
The energyflux in the primary waves being %pk 2 cC 2 , by
76 (8), the ratio which the energy scattered per second bears
to this is
(6)
SIMPLE. HARMONIC WAVES. DIFFRACTION 243
In the case of the sphere we found Q / = JQ =  7 ra 3 , and the
expression (6) therefore reduces to
J(*a).iro' ......................... (7)
In other words, the sphere scatters only the fraction  (&a) 4 of
the energy which falls upon it. For example, if the wave
length be a metre (which corresponds to a frequency of about
332), and the diameter of the sphere 1 mm., the fraction is
roughly 7 '6 x 10~ n . In the case of the circular disk, where
Q' = f Tra 8 , Q = 0, the ratio of the scattered to the incident
energy is ^(Ara) 4 .
The mathematical theory of the scattering by cylindrical
obstacles is more difficult. We will merely quote the result,
based on Lord Rayleigh's calculations, that when plane waves
are incident on a circular cylinder of radius a the fraction of the
incident energy which is scattered is f Tr^&a) 3 , approximately,
it being assumed as usual that ka is small. For a wire of
diameter 1 mm., and a wavelength of a metre, this
= 115 x 10 7 .
It is to be observed however that in the case of very minute
obstacles the order of magnitude of the results may be con
siderably modified by viscosity. The determining element here
is the ratio of the diameter of the obstacle to the quantity
h which was introduced in 66 as a measure of the thickness of
the airstratum, at the surface of the obstacle, whose motion
is appreciably affected by the friction. When the ratio in
question is moderately large the influence of viscosity on the
results will be very slight.
The distribution of velocity in the immediate neighbourhood
of the obstacle will be sensibly the same as in the case of a
uniform current of incompressible fluid flowing past the body.
In the case of the sphere it can be determined completely, but
the following approximation will be sufficient. We assume
(8)
where the first term represents the incident waves, and the
162
244 DYNAMICAL THEOEY OF SOUND
second is the form which the velocitypotential of a double
source assumes ( 76) when kr is small. This makes
 ik r + cos0, ......... (9)
and the condition of zero normal velocity for r = a is therefore
approximately satisfied provided B= l%ika?C. Hence in the
neighbourhood of the sphere we have
(10)
nearly. The velocities are therefore nearly the same as if the
fluid were incompressible. The pressure is given by
p=p + p<j>=p + inp(f> ............. (11)
This differs from the pressure (p + inpG) which would obtain
at the origin if the obstacle were absent by a term which
is small, of the order kr, in comparison. At points whose
distance r is a moderate multiple of a, whilst still small
compared with X, the pressure approximates even more closely
to that due to the incident waves alone.
82. Transmission of Sound by an Aperture.
In discussing the transmission of sound waves by an aperture
in a thin screen we will suppose, in the first instance, that the
dimensions of the aperture are small compared with the wave
length. This is of course the most interesting case from an
acoustical point of view.
The screen being supposed to occupy the plane x 0, and
the origin being taken in the aperture (S), let a wave train
represented by
be incident from the left. If we distinguish the functions
relating to the two sides of the screen by the suffixes 1 and 2,
we should have, if the screen were complete,
fr^Ce^ + Ce^*, < 2 = o, ............ (2)
the second term in <f> 1} which represents reflected waves, being
chosen so as to make dfa/da; = for x = 0.
In the actual problem the disturbance due to the aperture
will be confined mainly to the immediate neighbourhood of S,
and may be taken to be very small at distances from which,
SIMPLEHARMONIC WAVES. DIFFRACTION 245
though large as compared with the linear dimensions of S,
are small compared with X. Let two surfaces be drawn, on
the two sides, at some such distance from 0, each abutting
on the screen in the manner indicated by the dotted lines
in the figure. Within the region thus bounded, the fluid
oscillates backwards and forwards almost as if it were in
compressible, and the total flux ( 67) through the aperture
will therefore bear a constant ratio to the difference of the
velocitypotentials at the two surfaces. This will perhaps be
understood more clearly if we have
recourse to the analogy of electric
conduction. Suppose we have a /
large metallic mass, severed almost /
in two by a nonconducting parti /
tion occupying the place of the ;,'
screen. If this mass form part of \
an electric circuit, there will be \
little variation of potential in it \ x
except in the neighbourhood of the Xv
narrow neck which connects the two
portions. The electric potentials Fig" 74
at a distance on the two sides being
<>! and <f>. 2 , the current through the neck will be
*(*.&) ......................... (3)
where K may be called the " conductivity " of the neck, the
specific conductivity of the substance being taken to be unity.
In the hydrodynamical question, also, the quantity K may
appropriately be called the conductivity of the aperture. It is
easily seen that it is of the nature of a length.
At the two surfaces shewn in the figure we have ^ = 2(7,
<f> 2 0, approximately, and the total flux through the aperture
is therefore 2KC. If an equal flux were directed symmetrically
from the aperture on the lefthand side, the combination would
be equivalent, in an unlimited medium, to a simple source of
strength 4>KC. Hence, by 76 (12),
246 DYNAMICAL THEOEY OF SOUND
The corresponding velocitypotential on the near side is
evidently
e ikr . ..(5)
The energy ( W) transmitted by the aperture per second is
by the above reasoning onehalf that due to a simple source
at 0, whence, by 76 (15),
W = fccKWir ...................... (6)
The energyflux in the primary waves (1) being %pk*cC' 2 , the
ratio of W to this is 2^T 2 / 7r  ^ ^ s t be n ted that this is
independent of the wavelength X, so long, of course, as X is
large compared with the linear dimensions of S.
The exact calculation of K for various forms of aperture is
naturally a matter of some difficulty. For a circular aperture
of radius a it is found that K = 2a ; for other forms differing
little from a circle the value is sensibly the same as for a
circular aperture of the same area, the circle being evidently
a " stationary " form, in the sense in which this term is used
in the theory of maxima and minima. It appears then that
a circular (or nearly circular) aperture transmits the fraction
8/7T 2 , or '816, of the energy propagated across an equal area
(?ra 2 ) in the primary waves. This is, under the present
limitation as to size, very great compared with the energy
intercepted by a disk of the same dimensions ( 81). The
figure opposite gives the shapes of the surfaces of equal
pressure (< = const.), drawn for equidistant values of </>, in the
immediate neighbourhood of a circular aperture, and shews
how rapidly these tend to assume the spherical form. The
directions of vibration of the airparticles are of course normal
to these surfaces.
With regard to further problems of the kind we must
content ourselves with a few statements of results. In the
case of an aperture in the shape of a long narrow slit, whose
breadth is small compared with X, the energy transmitted is
again comparable with, and may even considerably exceed,
that corresponding to an equal area of wave front in the
primary waves. In the case of a grating composed of equal,
SIMPLEHARMONIC WAVES. DIFFRACTION 247
parallel, and equidistant slits in a thin screen, the fraction of
the total incident energy which is transmitted is found to be
1/(1 + A?/ 2 ), where k = STT/X, as usual, and
, a + b , Trb
U __, ogsec ___ (7)
where a denotes the breadth of an opening, and b that of each
intervening portion of the screen. As a numerical example,
Fig. 75.
suppose the wavelength to be ten times the interval a + 6
between the centres of successive apertures ; then even if the
apertures form only onetenth part of the whole area of the
screen, 88 per cent, of the sound will get through. In the
248 DYNAMICAL THEORY OF SOUND
case of a grating formed by equidistant bars of circular section,
the corresponding value of / is
l=7rb*/a, (8)
where b is the radius of the section, and a the distance between
the axes of consecutive bars. It is implied, however, that the
ratio b/a must not exceed (say) J.
83. Contrast between Diffraction Effects in Sound and
Light. Influence of WaveLength.
In the investigation of 82 an aperture was found to act as
a simple source from which sound diverges on the farther side
uniformly in all directions. This is in striking contrast with
what is usually observed in the case of light. We have so far
no indication of anything of the nature of beams or rays of
sound, just as when sound waves were incident on an obstacle
we found nothing of the nature of a soundshadow. The
difference in the results is due to the fact that the dimensions
of the aperture (or obstacle) have been supposed small in
comparison with the wavelength, whereas with light the
relation is usually the reverse.
We have avoided trespassing on the domain of Optics, but
as the dynamical conditions are in the present subject perfectly
definite, it may be permissible to examine this question of the
influence of wavelength a little more fully.
Consider the region of space lying to the right of the plane
x = 0. If this plane were a fixed boundary, and if there were
no sources of sound in the region, any disturbance would
ultimately pass away. Any steady periodic motion in the
region must therefore in the absence of internal sources be due
to motion of the boundary, and will be determinate when the
value of the normal component of the velocity at every point of
the latter is given. It can, moreover, be expressed in terms of
this distribution of normal velocity, as follows. The flux out
wards from an element 8$ of the plane is d<f>/dn . &S, if &n
denote an element of the normal drawn inwards from &S, and
if in imagination we associate with this an equal flux in the
opposite direction on the other side, the result is equivalent
SIMPLEHARMONIC WAVES. DIFFRACTION 249
to a source  29</3n . 8S in infinite space. The corresponding
velocitypotential at a point P is
where r denotes the distance of 8S from P. Integrating over
all the elements &S of the plane, we have
1 rr^^ a ikr
which is the required formula.
The motion to the right of the plane x = is also determinate
when the value of </> at every point of the plane, and thence
the pressure, is given, these two quantities being connected by
the relation p p + p<j> = p Q + ikcp<f>. Suppose for a moment
that in an otherwise unlimited medium we have a thin massless
membrane occupying the plane x = 0, and that on each element
of this a normal force X per unit area is exerted, which is
adjusted so as to produce the actual periodic pressure, and
therefore the actual value of $, on the positive face of the
membrane. By the theorem of 78 (15), the effect for an
element BS, will be equivalent to a double source, and the
corresponding velocitypotential at a point P will be
(2)
The variable parts of the pressures on the two faces of the
membrane, viz. + p(f> = ikcp<f>, must balance the force X, so
that X = %ikcp$. Substituting in (2), and integrating over the
plane x = 0, we obtain
* *
The structure of the integrals in (1) and (3) recalls the
process by which " Huygens' principle " is applied in optics to
find the disturbance at any point P in terms of "secondary waves"
supposed to issue from the various elements of a wavefront.
There was at one time much discussion as to the exact character
to be assigned to these secondary waves, more especially as to
the law of intensity in different directions. We now recognize
250 DYNAMICAL THEORY OF SOUND
that the problem has mathematically more than one solution ;
either of the above formulae will lead to an exact result, and
we might even use a combination of the two, in any arbitrary
proportions. This resolution of a historic controversy is due to
Lord Rayleigh.
As a verification of (3), suppose that the value of <j> at x =
is that due to a train of plane waves <f> = e~ ikx . Let OT denote
the distance of BS from the orthogonal projection of the point
P on the plane x 0, so that r 5 = # 2 4 w 2 . For the aggregate
of elements 88 forming a certain annulus of the plane we
may write 27ror&nr = 27rr&r. We have also dr/dx = a;/r. The
formula (3) therefore gives
r 9 fe~ ikr \
x ^ e )dr = e ik *. ...(4)
In the case of waves transmitted by an aperture in a plane
screen (x = 0), we have, in (1), 3</>/3n = except over the area of
the aperture. If, further, the dimensions of the aperture S are
small compared with X, then at a point P whose distance r is
large compared with X, the function e~ ikr /r will have sensibly
the same value for all the elements of S, and we may write
(5)
where the first factor represents the total flux through S.
Under these circumstances the aperture acts like a simple
source, as in 82.
It is understood of course that the expression d<f)/dn in
(1) or (5) represents the normal component of the velocity, as
modified by the action of the screen. When as in the case just
considered the aperture is relatively small, the distribution of
normal velocity over it will differ considerably from that due to
the primary waves alone. This distribution can be ascertained
approximately, in the case of plane waves incident directly on
a circular opening, from the electrical analogy of 82. The
lines of flow have the same configuration as the lines of force
SIMPLEHARMONIC WAVES. DIFFRACTION 251
due to an electrified disk*, and the normal velocity has the
distribution
_d$ = B
dn V( 2 ^ 2 )'
where tn denotes the distance of any point of the aperture from
its centre. The velocity becomes very great near the edge, and
is mathematically infinite ( at the edge itself (r = a), but it
appears on integration that the parts of the area near the edge
contribute little to the total flux, which is
(7)
If the incident waves be represented by
(8)
the same flux will as in 82 be expressed by 2KC, or 4a(7.
Hence, comparing,
In the other extreme, where the wavelength is only a
minute fraction of the dimensions of the aperture, the effect of
the screen in modifying the distribution of normal velocity over
the latter is practically confined to a distance of a few wave
lengths from the edge, and the corresponding part of the integral
in (1) is quite unimportant. In this case, the incident waves
being still expressed by (8), we can put d(f>/dn = ikC with
sufficient accuracy over the whole area of the aperture, whence
*(f^ .do)
For the methods of approximating to the value of this integral,
by the use of Huygens' or Fresnel's " zones," or otherwise, we
must refer to books on Optics. It is found that the amplitude
is nearly uniform within the space bounded by a cylindrical
surface whose generators are normal to the screen through the
edge of the aperture, and is nearly zero in the surrounding
region. Near the cylindrical boundary, on either side, we have
* See Fig. 75, p. 247, which represents the configuration of the equipotential
surfaces.
f The awkwardness of this conclusion may be avoided by giving the screen
a certain thickness, and rounding the edges.
252 DYNAMICAL THEOKY OF SOUND
the diffraction effects which are especially studied in the theory
of Light.
The question of the impact of waves on a plane lamina can
be treated in a similar manner. For this purpose the formula
(3) is most convenient. The lamina being in the plane x = 0,
and the primary waves being represented by (8), we may write
<t> = Ce~ ik * + x , ..................... (11)
where % is the velocitypotential due to a vibration of the
lamina normal to its plane with the velocity ikC, equal and
opposite to that in the primary waves. It is evident that the
values of this function at any two points which are symmetric
ally situated with respect to the plane x = will be equal in
magnitude but opposite in sign. We have then, to the right
of the lamina
This only requires a knowledge of the value of % at the positive
face of the lamina, the value at all other points of the plane
x = being obviously zero. The case where the dimensions of
the lamina are small compared with X has been noticed in
81 ; the scattered waves have then a much smaller intensity
than those transmitted by an aperture of the same size and
shape. In the opposite extreme, the value of % near the
positive face is, except near the edge, the same as in the case
of an infinite vibrating plate, viz. % = Ce ikx , so that we have
with sufficient accuracy
A detailed study of this integral would indicate, in the complete
solution expressed by (11), the existence of a soundshadow to
the right of the lamina. For large values of kr the formula
(13) may be replaced by
ikr
............ (14)
and for small obliquities 6 we may further put cos 6 = 1. The
formula then becomes, except as to sign, identical with (10),
shewing that the disturbance produced by the lamina is, under
SIMPLE HAKMONIC WAVES. DIFFRACTION 253
the conditions postulated, exactly opposite to that transmitted
by an aperture of the same dimensions. This is a familiar
fact in Optics ; but the preceding considerations shew that it
may be utterly wide of the mark when the wavelength is no
longer small compared with the linear dimensions concerned.
It need hardly be said that there are acoustical phenomena
where, as in the case of large reflecting or obstructing surfaces,
optical relations are approximated to. The results are then
analogous, the resemblance being more complete the higher
the pitch of the note sounded. By the use of a source of very
high pitch, and of a sensitive flame as a detector, Lord Rayleigh
has succeeded in imitating some of the most delicate phenomena
of physical optics.
In the above theoretical investigation we have been obliged
to rely to some extent on intuitive considerations, as e.g. in the
assumed distribution of velocity over the area of an aperture
when the wavelength is relatively small. It is therefore
desirable that such assumptions should be tested if possible by
exact calculation. The only instance, at present, where this has
been successfully carried out is that of waves incident on a
plane screen with a straight edge. The reflection by the screen,
the transmission past the edge, the formation of a shadow
behind the screen, and the diffraction phenomena near the
boundaries of the respective regions, all come out in practical
accordance with the usual theory. The investigation was
published by Sommerfeld in 1895*.
* A simplified version is given in the Proc. Lond. Math. Soc. (2), vol. iv.
(1906).
CHAPTER IX
PIPES AND RESONATORS
84. Normal Modes of Rectangular and Spherical
Vessels.
The main object in this chapter is to develop the laws of
vibration of air contained in cavities, such as those of resonators
and organ pipes, which are in communication with the external
atmosphere. A little space may however be devoted in the
first instance to some problems relating to the vibrations of air
in spaces which are completely enclosed by rigid walls. These
will at all events supply some interesting examples of the
general theory of normal modes ( 16).
The analytical process consists in finding solutions of the
equation
V 2 </> + #ty = ........................ (1)
consistent with the condition
which expresses that the component of the fluid velocity in
the direction of the normal (n) vanishes at the boundary. It
appears that, as in former analogous problems, this is only
possible for a certain sequence of values of k, which determine
the nature and the frequency of the respective normal modes.
In the case of a rectangular cavity we take the origin at a
corner, and the coordinate axes along the edges which meet
there. If the lengths of these edges be a, b, c, the condition
(2) is fulfilled by
TTTZ
</>= (7 cos cos j. cos 
d G
PIPES AND RESONATORS
255
where p, q, r are any integers; and the equation (1) is also
satisfied provided
(4)
If we put <? = 0, r = 0, the case degenerates into that of the
doubly closed pipe ( 62).
A more interesting case is that of a spherical cavity. The
symmetrical radial vibrations come under the methods of 71,
76. The formula (15) of 71, which implies that there is no
source at the origin, gives, in the case of simpleharmonic
vibrations,
^Ller ttr
(5)
...(6)
_
or, say,
sin kr
256
DYNAMICAL THEOEY OF SOUND
The condition (2) requires that d(f>/dr = for r a, the radius of
the cavity. Hence
tan kaka ......................... (7)
This is a transcendental equation to find k, and thence n (= kc).
The roots are obtained graphically (see Fig. 76) as the abscissae
of the intersections of the lines y = tan x, y=x y the zero root
being of course excepted as irrelevant. We have, approximately,
&a=(ra+i)7r, where m= 1, 2, 3, .... More accurate values of
the first three roots are
ka/ir = 14303, 24590, 3*4709 ............. (8)
The numbers give the ratio of the diameter 2a of the cavity to
the wavelength. In the modes after the first there are internal
spherical nodes (i.e. surfaces of zero velocity) whose relative
positions are indicated by the roots of inferior rank. In the
higher modes the nodal surfaces tend, as we should expect, to
become equidistant, since the conditions, except near the centre,
approximate to those of plane waves.
Equations of somewhat similar structure to (7) occur (as
we have seen) in various parts of our subject, as well as in
other branches of mathematical physics, and processes of
numerical solution have been de
vised by Euler, Lord Rayleigh
and others. There is one method,
of very general application, which
is so elegant, and at the same
time so little known, that it may
be worth while to explain it. It
is given by Fourier in his Theorie
de la Chaleur (1822). Starting
with a rough approximation, say
= 0?!, to a particular root of (7),
we calculate in succession the
quantities x z , x s , 0? 4 , ... determined
by the relations
# 2 = tan" 1 0?i , # 3 = tan" 1 # 2
The figure illustrates the manner in which these converge
towards the desired root as a limiting value, no matter from
Fig. 77.
tan" 1
(9)
PIPES AND RESONATORS 257
which side we start. Some fairly obvious precautions are
necessary in using the method, and it is easily seen that the
convergence will be slow if the two curves have nearly the
same inclination (in the same or in opposite senses) to the axis
of x. Expressed as multiples of TT, the successive approximations
obtained in this way to the first root of (7) are*
15, 1433435, T430444, 1430304, 1*430297, ....
The same analysis can obviously be applied to the theory of
vibrations in a conical pipe whose generating lines meet in 0.
If the tube extend from the origin to r = a, the usual approxi
mate condition (s = 0) to be satisfied at the open end gives
sin&a = 0, (10)
the same as for a doubly open pipe of length a ( 62). For the
case of a tube extending from r = a to r = b, and open at both
ends, we require the complete solution
r<f> = A cos kr + B sin kr. (11)
The conditions give
Acoska + Bsinka = Q, Acoskb + Bsinkb = Q, (12)
whence sin k (b a) = 0, (13)
as in the case of a doubly open pipe of length b a.
If x be any solution of the general equation (1), it appears
on differentiation throughout with respect to x that the equation
is also satisfied by </> = d%/3# . We have already had an example
of this in the general double source of 73. From (6) we
derive in this way the solution
d /sinferX ( .
* =G Wx()> (14)
or, if x = r cos 6,
ri d/smkr\  C ,, , 7 \ Q / ~\
6 = C=   cos 6  (kr cos kr sin kr) cos 6. (lo)
or \ T / r 1
This leads to another series of normal modes of the air con
* In calculations of this kind, and for the purposes of mathematical physics
generally, trigonometrical tables based on the centesimal division of the
quadrant are most convenient. A fourfigure table of this type is included in
J. Hoiiel's Eecueil des Formules et des Tables Numfriques, 3rd ed., Paris, 1885.
L. 17
258
DYNAMICAL THEOEY OF SOUND
tained in a spherical cavity. The condition 3</>/3r = is satisfied
for r = a, provided
tan ka =
2 
,(16)
The solution can be carried out as in the case of (7). The
annexed diagram of the curves y = cot x, y = (2 # 2 )/2#, shews
that the roots tend after a time to the form mnr. Approximate
values of the first few roots are
5, 1891, 2'930, 3'948, 4*959, ...(17)
Fig. 78.
the first of which alone gives any trouble. This root corresponds
to the gravest of all the normal modes of the cavity. The air
swings from side to side, much as in the case of a doubly closed
pipe, and the wavelength is X = 27T/A; = T509 x 2a. The
forms of the equipotential surfaces, to which the directions of
vibration of the airparticles are orthogonal, are shewn in Fig. 79.
In the next mode the radial velocity vanishes over the sphere
r/a= '6625/1891 ='350.
The study of the more complicated normal modes of vibra
tion in a spherical vessel would lead us too far. The problem
is fully discussed in Lord Rayleigh's treatise.
PIPES AND RESONATORS
259
Fig. 79.
85. Vibrations in a Cylindrical Vessel.
The theory of the purely transversal vibrations of the air
enclosed by a circular cylinder is very similar. As in 54, the
equation
v (p (Z> j i f\ /i \
^ j fcP(p = ( 1 )
03j Vll
where x, y are Cartesian coordinates in the plane of a cross
section, becomes in polar coordinates
and the typical solution, when there is no source at the
origin, is
x
GJ m (kr) cos W0.+<k ) ............... (3)
The admissible values of k are determined by the condition that
d<f>/dr = for r = a, or
J m '(ka)=0 ......................... (4)
For the radial vibrations (ra = 0) the earlier roots are given by
ka/ir = 1*2179, 22330, 32383,..., ......... (5)
172
j
260 DYNAMICAL THEOEY OF SOUND
the limiting form being integer + J. In the case m = 1, which
includes the gravest mode,
&a/7r=586, 1697, 2717,..., ............ (6)
the limiting form being integer J .
The purely longitudinal modes of a closed circular cylinder
come under 62. There remain the vibrations of mixed type.
The equation (2) has now to be modified by the inclusion of
a term 9 2 </>/d 2 , where z is the longitudinal coordinate. It is
found that the equation is satisfied by
(7)
provided k* = /3 2 + m'V 2 /^, .................. (8)
the origin being taken at the centre of one end. The condition
of zero normal velocity (d(f>/dz) at the other end (z = I) is
satisfied if m' be integral. The corresponding condition at the
cylindrical surface requires that /3 should be a root of
J m '(l3a) = ...................... (9)
86. Free Vibrations of a Resonator. Dissipation.
The foregoing examples are of theoretical rather than
practical interest, since the vibrations of a mass of air enclosed
by rigid walls would be completely isolated. For acoustical
purposes the vibrating mass must have some communication
with the external atmosphere ; on the other hand it is essential
that the communication should be so restricted that the frac
tion of the energy which is used up in a single period in
the generation of diverging waves shall still be very small.
Otherwise the free vibrations could hardly be regarded as
approximately simpleharmonic, and might even resemble
the "deadbeat" type (11).
The theory is simplest in the case of " resonators " such as
were employed by Helmholtz in his researches on the quality
of musical notes. These are nearly closed vessels, with an
aperture, and are used to intensify, by sympathetic vibration of
the enclosed air, the effect of a simple tone produced in the
neighbourhood. The precise form is not important; it may
be spherical or cylindrical, or almost any shape, so long as
PIPES AND RESONATORS 261
the least diameter considerably exceeds the dimensions of
the aperture. In his synthetic work on the vowel sounds
Helmholtz used cylindrical resonators having a circular opening
at the centre of one end. When the object was to detect and
to isolate a particular overtone in a complex sound, he used
the more convenient form shewn in Fig. 80. The small open
nipple opposite the mouth is inserted into the ear cavity, so
that the tympanic membrane becomes part of the internal wall
of the resonator.
Fig. 80. Fig. 81.
The theory of resonators was treated mathematically for the
first time by Helmholtz in 1860, and was afterwards greatly
simplified by Lord Rayleigh (1871). Suppose in the first place
that we have a vessel with a narrow cylindrical neck which is
occupied by a plug or piston freely movable to and fro (Fig.
81). Let Q denote the capacity of the vessel, I the length of
the neck, o> its sectional area, p the density of the piston. We
will assume that the period of vibration is so long that the
corresponding wavelength (X) in air is large compared with
the diameter of the vessel. Under this condition the con
densation s will at any instant be almost uniform throughout
the interior, and we may put s = cox/Q, where x denotes the
small displacement of the piston outwards from its mean
position. The resulting excess of pressure on the base of the
piston is /5C 2 so), or pc 2 (0 2 x/Q, and the equation of motion of
the system is, approximately,
(1)
DYNAMICAL THEORY OF SOUND
The motion is accordingly simpleharmonic, with a period 27r/n,
provided
The nature of the piston is of little importance, provided
its mass be sufficiently small. We may even replace it by air,
if the length I be small compared with X, for under this
condition the column of air in the neck will behave almost as if
it were incompressible. We have then p' = p, and
(3)
Even in the case of a resonator whose mouth consists of
a mere opening in the wall, without a neck, the theory is not
very different. It is only a question of obtaining a proper
measure of the inertia of the mass of air in the immediate
neighbourhood of the mouth, inside and outside, which takes
the place of the piston in the above problem. The flow
through the aperture at any instant is still regulated, ap
proximately, by the same laws as that of an incompressible
fluid, or of electricity in a uniform conductor. There being
little motion in the interior, the
value of $ there will be sensibly
uniform ; we denote it by fa. Out
side, at a short distance beyond the
mouth, we shall have <j> = 0, nearly.
If q denote the volume of air which
has passed through the aperture
outwards up to time t, the current,
or flux, outwards at this instant will
be q, and we have, by the electric analogy,
........................ (4)
where K is the " conductivity " ( 82), which depends, of course,
on the shape and size of the aperture and the configuration of
the wall in its neighbourhood. It is to be observed that this
relation (4) is purely kinematical ; from the point of view of the
generalized dynamics of a system of one degree of freedom
PIPES AND RESONATORS 263
( 7), it expresses the momentum (which may be symbolized by
pfa) in terms of the velocity q. The dynamical equation
c 2 s=<i ........................... (5)
of 70 (3) may in like manner be interpreted as expressing
the relation between change of momentum and force. If the
zero of q correspond to the equilibrium state, we have
s = q/Q ......................... (6)
Eliminating s and <, between (4), (5), and (6), we obtain
The motion is therefore of the type
q = Ccos(nt + e), ..................... (8)
provided n* = Kc?/Q ............ . ............ (9)
If we write n = kc, this gives
t? = K/Q, X = 2nV(Q/#) ................ (10)
The wavelength depends, as we should expect, solely on the linear
dimensions of the resonator and its aperture. For resonators
which are geometrically similar in all respects, it varies directly
as the linear dimension. This is in accordance with a general
principle which may be inferred from the differential equation
(2) of 76, or otherwise. The formula (9) indicates further that
the pitch of the resonator is lowered by contracting or partially
obstructing the aperture, whilst it is raised by diminishing the
internal capacity.
The kinetic energy, being mainly resident in the neighbour
hood of the mouth, may be calculated from the principles
applicable to an incompressible fluid. If the actual motion
were generated instantaneously from rest, the work required
would be the sum of half the products of the impulses into
the corresponding velocities. The equations (9) of 69 shew
that the requisite impulsive pressure is p^ hence
The potential energy is, by 70 (8),
F=4 / >cVQ = ( / >cVQ).3 2 ............. (12)
The coefficients in these expressions being known, the speed
n of the oscillations can be inferred at once by the general
264 DYNAMICAL THEORY OF SOUND
formula (7) of 7. It was under this form that the theory was
presented by Lord Rayleigh. It is to be noticed that the
inertiacoefficient is proportional to the " resistance " of the
aperture (in the electrical sense), whilst the coefficient of
stability, or elasticity, varies inversely as the capacity Q.
The preceding theory applies only to the gravest mode of the
resonator. In the higher modes the internal space is divided
into compartments by one or more " loop surfaces " (i.e. surfaces
of constant pressure, where <j> = 0), and the frequencies are
much greater. The wavelength is then at most comparable
with the linear dimensions, as in the problems of 84.
As already stated ( 82) the calculation of K is usually difficult.
For a circular aperture in a thin wall K is equal to the
diameter, and for any form differing not too much from a circle
we may put K=2 V(w/7r), approximately, where to is the area.
The frequency, as determined by (9), will then vary as a^/Q .
It is remarkable that this law was arrived at empirically by
Sondhauss at a date (1850) anterior to the theory. When the
aperture is fitted with a cylindrical neck, the conductivity is
limited mainly by the neck itself, and we may put K = w/l,
approximately, where I is the length. The formula (9) then
agrees with (3). It is implied that I is small compared with X,
and at the same time large compared with the diameter of the
channel.
We have in the above theory allowed for the inertia of the
external atmosphere, but not for its compressibility, and the
vibrations as given by (8) are accordingly persistent. In other
words, we have neglected the apparent* dissipation of the
energy of the resonator due to airwaves diverging outwards
from the neighbourhood of the mouth. This will have, in
general, no appreciable influence on the period, but will
manifest itself by a gradual decay of the amplitude.
The effect can be estimated with sufficient accuracy in
directly. The flux outwards at the mouth is, by (8),
q=nCsm(nt + e) (13)
* True dissipative influences such as viscosity and thermal conduction are
ignored in the present investigation. They probably play as a rule a wholly
subordinate part.
PIPES AND RESONATORS 265
If the resonator were practically isolated in space, then on
account of the assumed smallness of its dimensions as com
pared with \, the effect of the flux at a distance would be that
of a simple source of strength nC, and the rate of emission of
energy would accordingly be
W=n 4 pC*/Sirc, ..................... (14)
by the formula (15) of 76. The energy E of the motion,
being equal to the potential energy at its maximum, is,
approximately,
(15)
by (12). Equating, on the principles of 11, the rate of decay
of this energy to the emission W, we find
and therefore q = G e~ tlr cos (nt + e), ............... (17)
provided r = $7r(?/n'Q = SrrQ/K^ ............ (18)
in virtue of (9). The ratio of the modulus of decay to the
period (%7r/kc) is given by
(19)
Since K is at most comparable with the mean breadth of the
aperture, this ratio is usually very great, and the preliminary
assumptions implied in the above process are amply justified.
If the mouth of the resonator were furnished with an
infinite flange, i.e. one whose breadth is large compared with X,
the equivalent source would, as explained in 82, have double
the strength above assumed, and the emission of energy, now
operative in one half of the surrounding region, would be twice
as great. The modulus (18) would accordingly be halved.
As a numerical illustration of the theoretical results, take
the case of a spherical vessel 10 cm. in diameter, with a circular
aperture 1 cm. in radius, so that Q = 523*6, K 2. The wave
length, calculated from (10), is 101*6 ; and the frequency there
fore about 327. The modulus of decay, as given by (18), is
about onetenth of a second.
266 DYNAMICAL THEOEY OF SOUND
87. Corrected Theory of the Organ Pipe.
The same principles can be applied to obtain a correction
to the imperfect theory of the open pipe which was given in
62. We may begin by a brief examination of the slightly
simpler problem of reflection at an open end of an infinitely
long pipe ( 61).
Fig. 83.
Near the open end there is a certain region, whose dimensions
are small compared with the wavelength, within which the
transition takes place from plane waves within the tube to
diverging spherical waves outside*. We take the origin inside
the tube, near the mouth, but in the region of plane waves, and
the positive direction of the axis of x along the tube. For the
region of plane waves we may write
<f> = Ae ik * + Be ik *, ....(1)
where the first term may be taken to represent a train of waves
approaching the end, from the right, whilst the second term
* The figure, which is based on formulae given by Helmholtz in another
connection, relates to the twodimensional form of the problem. In three
dimensions the transition to a state of uniform radial flow outwards from the
mouth would be still more rapid.
PIPES AND RESONATORS 267
represents the reflected waves. The outward velocity at is
therefore represented by ik (A  B), and the flux is
q = ika>(AB\ ..................... (2)
where a> is the sectional area. The velocitypotential at is
A + B. The " resistance " between the section x = and the
external region to the left may be specified as equivalent to
that of a certain length a of the pipe, and is accordingly
denoted by a/to. Hence, by the electrical analogy,
to
(3)
u ^ lika ,
whence =   ...................... (4)
A I+ika.
If we put ka tan & fi, ..................... (5)
this may be written B/A =  e * k t ...................... (6)
Hence < = A {e ikx  e~* ***> ) ................ (7)
The reflected train is therefore equal in amplitude to the
incident one, as was to be expected, since the inertia only of the
external air is so far taken into account ; but there is a difference
of phase. In the theory of 61 the condition to be satisfied at
an open end was s = 0, or <f) = 0. Hence if we write (7) in the
form
= 4*^ {***> e*^} ............ (8)
we recognize that the circumstances are the same as if the pipe
were prolonged to the left for a length ft, and the reflection at the
mouth were to take place according to the rudimentary theory.
The wavelength being assumed to be large compared with the
diameter of the pipe, ka. will usually be small, so that ft = a,
nearly. But if the pipe be very much contracted or obstructed
at the mouth, ka may be considerable, and k/3 will in that case
approach \ir. We then have B = A, nearly, and the circum
stances approximate to those of reflection at a closed end.
The actual determination of a is a problem in electric
conduction which has at present only been solved, even
approximately, in a very few cases. Lord Rayleigh estimates
that for an accurately cylindrical tube fitted with an infinite
flange the value of a is about "82 of the sectional radius. For
268 DYNAMICAL THEOEY OF SOUND
an unflanged cylindrical tube experiment seems to indicate a
value of about '6 of the radius.
We will next suppose the pipe to be of finite length, and to
be closed at x = I, the origin being chosen as before, near the
mouth, in the region of plane waves. For this latter region we
may assume
<f> = A cos k(l x), .................. (9)
since d<f>/dx must vanish for x = l. The flux outwards at
the mouth is therefore
q = codcj>/dx = kco A sin kl, .................. (10)
and the potential at is A cos kl. Hence with the same
meaning of a as before we have
A cos klx kco A sin kl,
CO
or cotkl = ka ...................... (11)
This equation determines the wavelengths (^Tr/k) of the
various normal modes. Usually, ka. is small, and the solution
of (11) is then
kl = (m + J) TT ka,
or fc(/ + ) = (w + 4)9r, ............... (12)
where m is integral. The character of the normal modes is there
fore the same as on the rudimentary theory ( 62), provided we
imagine the length of the pipe to be increased by the quantity
a. In particular, the frequencies are as the odd integers
1, 3, 5, ... , so long as the wavelength remains large compared
with the diameter.
If the aperture be contracted the value of a is increased,
and the result tends to become less simple. In particular, the
harmonic relation of the successive frequencies is violated, as
may easily be seen from a graphical discussion of the equation
(11). When the pipe is almost closed, a is relatively great, and
the solution of (11) is kl = 1/ka, or k* = 1 /la. This agrees with
the formula (10) of 86, if we put col = Q, co/a = K.
In the case of a pipe open at both ends the period equation
is found to be
tan &Z = ( + '), ............... (13)
where a, a' are the corrections for the two ends, but the calcula
PIPES AND RESONATORS 269
tion implies that ka. and ka! are small. It is, however, only on
this condition that the conductivities at the two ends can, as a
rule, be estimated independently of one another. The equation
is then equivalent to
sin &( + + a 7 ) = 0, ............... (14)
and the frequencies are therefore those which are assigned to
a pipe of length I + a + a! by the rudimentary theory. The
harmonic relation between the various normal modes is pre
served, but it must be remembered that the approximation is
the more precarious, the higher the order of the harmonic.
The wavelengths of the proper tones are in all cases fixed
by the linear dimensions, but the frequencies, which vary as
the velocity of sound, will rise or fall with the temperature.
An "open" organ pipe is tuned by means of a contrivance
which increases or diminishes the effective aperture at the open
end, i.e. the end remote from the " mouth " proper. The pitch
of a " closed " pipe is regulated by adjusting the position of a
plug which forms the barrier.
To calculate the rate of decay of the free vibrations it will
be sufficient to take the case of the stopped pipe. The kinetic
energy corresponding to
(f) = A cos k (I x) cos nt ............... (15)
is given with sufficient accuracy by
T = Ipto dx = pk*c0l . A* cos 2 nt t . . .(16)
if ka. be small, since cos kl = 0, nearly. A more careful calcula
tion, taking account of the transition region between the plane
and the spherical waves, replaces I by I + a, approximately, in
this formula, but the correction is not important. The total
energy, being equal to the kinetic energy at its maximum, is
accordingly
E = \pteu>lA'> ...................... (17)
If the mouth be unflanged it acts, in relation to the external
space, as a simple source of strength kcoA sin kl, or kaA, nearly,
and the consequent emission of energy per second is accordingly
W = pfrco'cAt/STr, .................. (18)
270 DYNAMICAL THEOEY OF SOUND
by 76 (15). Equating the rate of decay of the energy to W,
we are led to the equation
dA k*a>c A _
"3* +
and the modulus of decay is therefore
r = 47rZ/A; 2 6>c ...................... (20)
The ratio of this to the period (2?r/A;c) is 21/kco, or (in the
gravest mode) ffi/ira), nearly. Since the moduli of the various
normal modes are proportional to the squares of the respective
wavelengths, the decay is the more rapid the higher the order.
For a flanged pipe the result (20) would be halved.
88. Resonator under Influence of External Source.
Reaction on the Source.
The theory of forced vibrations due to an external source
of sound, to which we now proceed, involves some rather
delicate considerations, and is often misunderstood. That
the mass of air contained in a resonator or an organ pipe
should be set into vigorous vibration by a source in approxi
mate unison with it is intelligible enough; but it is further
desirable to have some estimate of the amplitude of the forced
vibration, and in particular to understand why the sound which
is apparently emitted by the resonator should under certain
conditions enormously exceed that which would be produced
by the original source alone.
For simplicity we will suppose that this source is main
tained at constant amplitude by a suitable supply of energy,
so that the vibration of the air is everywhere steady. It is
evident at once that under this condition no work is done,
on the average of a whole period, at the mouth of a resonator
on the contained air, the energy of the latter being constant,
and consequently that no work can in turn be done by the
reaction of this mass on the external atmosphere. Any
increased propagation of sound to a distance must be due to
the changed conditions which the action of the resonator has
introduced in the neighbourhood of the original source. If
this source be not maintained constant, but merely started
with an initial fund of energy (as in the case of a tuning
PIPES AND RESONATORS 271
fork), this fund will under the influence of the resonator be
more rapidly consumed.
In order to treat the question in a form free from unessential
details, which may vary from one case to another, we take the
case of a resonator of the type considered in 86, whose
dimensions are small compared with the wavelength.
The theory is simplest when the frequency of the source
is very nearly equal to the natural frequency of the resonator,
as determined by 86 (9), so that the forced vibration in the
latter is at its strongest. It will perhaps make the matter
clearer if we imagine in the first instance that the resonator
has a short cylindrical neck in which a thin massless disk,
almost exactly fitting it, can be made to move to and fro by
a suitable application of force. Suppose then that the disk
is made to execute a vibration such that the volume swept
over by it outwards up to time t is
q = C cos nt ; (1)
and let the extraneous force which must be applied to the
disk to compensate the difference of the airpressures on the
two sides be denoted by
A cos nt + B sin nt, (2)
this expression being (say) positive when the force is outwards.
The component Acosnt which keeps step with the displace
ment is required to control the inertia of the air. From the
general theory of forced vibrations ( 9, 12) it appears that the
coefficient A can be made to have one sign or the other
by adjusting the value of n, the sign being the same as that
of C when the imposed vibration is relatively slow, and the
opposite when it is relatively rapid. We may therefore
suppose n to be so adjusted that A = 0. The circumstances
are then very nearly those of a free vibration, and the required
value of n is given by
n* = Kc*/Q, (3)
very approximately. The second component of the force (2),
which keeps step with the velocity (q), is required to maintain
the emission of energy outwards, which is, by 76 (15),
(4)
272 DYNAMICAL THEORY OF SOUND
This must be equal to the mean value of pq, where p is the
pressure at the outer face of the disk. Hence by comparison
we find that p must have the form
p = p + D cos nt r sin nt ............. (5)
The corresponding pressure on the inner face will be
p = p Q + D cos nt, ..................... (6)
simply, since no work is done, on the whole, on the air
contained in the resonator.
We may now invoke the action of the external source.
If this be such as would produce the pressure
P = P<> + ~A~ s ^ n nt .................. CO
at the mouth of the resonator if the disk were at rest, then
in the motion which is compounded of that due to the source
and that due to the disk no extraneous force will be required,
and the disk may therefore be annihilated without causing
any appreciable change in the conditions. If </> 2 be the
velocitypotential due to the source alone, at the mouth of
the resonator, we must have, in this case,
since (7) must be identical with p =p Q + p^.
The hypothesis of a rigid disk vibrating in a cylindrical
space was merely introduced for facility of conception, and
is in no way essential to the argument. The disk may, if
we please, be replaced by a flexible and extensible membrane
enclosing the aperture of the resonator, and abutting on the
external wall in the region of diverging waves.
Comparing (1) with (8) we see that to a disturbing potential
whose value at the mouth is
< 2 = Jcos(nte) .................. (9)
will correspond a vibration
PIPES AND RESONATORS 273
under the condition of maximum resonance, when n is given
by (3) approximately. The corresponding flux is
g = gsin(?rte) ................... (11)
The emission of energy is best calculated from a con
sideration of the circumstances at a great distance. The
velocitypotential will be compounded of that due to the
original source and that due to the flux q, and under certain
conditions the latter component may greatly preponderate.
The emission of energy is then
W=27rpcJ*, ..................... (12)
approximately, by 76 (15).
Thus if < 2 be due to a simple source A cos kct at a distance b
from the aperture, we have
&) ................ (13)
Hence J=A/4nrb,
and q = jj sin k (ct b) ................... (14)
This is equivalent to a source whose amplitude is to that of
the primary source in the ratio l/kb. If b be small compared
with X/27T this ratio is large ; and the emission of energy
exceeds that due to the original source in the ratio l/k?b z .
In the case of a double source B cos kct we may write,
if kb be small,
D
< 2 = 7 Tg c s a cos k(ctb), ............ (15)
by 76 (23), if a denote the angle which the axis of the source
makes with the line drawn from it to the aperture. Hence
/ = B cos a/4?r6 2 , and the emission, as given by (12), is
TF = pc J B 2 cos 2 a/87r& 4 ................ (16)
The emission due to the original (double) source alone would
be ptfcIP/ZiTr, by 76 (26). The ratio in which the emission
is increased is therefore 3 cos* a/fab*. Since the mean value
of cos 2 a is J, the mean value of this ratio, for all directions
of the axis of the double source, is 1/frb 4 . That the ratio
L. 18
274 DYNAMICAL THEORY OF SOUND
should, under the given conditions, be so much greater than
in the preceding case is due to the relatively smaller efficiency
of a double source, as compared with a simple one, in propagating
energy outwards ( 80).
It may be well to insist again that the increased output
of energy is an indirect consequence of the presence of the
resonator, which itself does no work. The whole energy is
supplied by the original source, where the motion takes place
against an augmented component of pressure in the same phase
with the velocity. The velocitypotential due to the flux q
outwards from the resonator, as given by (11), is
<>, = sin (nt kr e), (17)
kr
and the resultant pressure is
P=PQ\ cos (nt kr e). (1$)
r
In the case of a simple primary source we had J=A/4>7rb )
e = kb; hence, putting r = b, we find that the consequent
pressure in the neighbourhood of this source is
(19)
Since the imposed outward flux is A cos nt, the mean rate of
work against this part of the pressure is
The output is therefore greater than it would be in the
absence of the resonator, in the ratio cos 2kb/k 2 b 2 . This agrees
with the former result, obtained on the hypothesis that kb is
small.
The energy stored in the resonator under the conditions of
maximum vibration is, by 86 (15),
E = 8
This varies directly as the capacity Q, and is for apertures of
similar form inversely proportional to the area.
The effect of a resonator under the influence of a distant
PIPES AND RESONATORS 275
source in unison with it may be sufficiently illustrated on the
assumption that the incident waves are plane. If
fa = Jcosk(ctx), ............... (22)
the ratio of the energy scattered by the resonator, which is
given by (12), to the energyflux in the primary waves, viz.
JpAr'cJ" 2 , is 47T/& 2 , or X 2 /7r. The energy diverted per second,
at its maximum, is therefore equal to '318 of that which in
the primary waves is transmitted across a square area whose
side is the wavelength. It may be added that a similar
law is met with in the theory of selective absorption of
light.
When approximate agreement between the frequency
(n/2?r) of the source and the natural frequency (w /27r) of the
resonator is no longer assumed, the external pressure which is
required to maintain a steady vibration (1) through the aperture
will consist of two parts. In the first place we have a component
keeping step with the displacement, which is required in order
to control the inertia of the air. This is easily found by an
extension of the method of 86. If $1 denote the velocity
potential in the interior of the resonator, </> 2 that at a short
distance outside the aperture, in the region of approximately
spherical waves, we have
qKfaM .................. (23)
in accordance with the electrical analogy. In the interior we
have 5 = q/Q, c 2 s = ^, as before. Hence
q + n<?q = K<f> 2 , .................. (24)
where n 2 = ^Tc 2 /Q ...................... (25)
This gives, for the external pressure,
(26)
The second part, which is in the same phase as q, is needed in
order that there may on the average be no gain or loss of energy
to the air contained in the resonator, and is accordingly given
by (7). Hence we have, altogether,
smnt); ...(27)
182
276 DYNAMICAL THEORY OF SOUND
and the complete expression for the disturbing velocity potential
near the mouth must be
kK \ .
nt ^ cosnt \ ....... (28)
In the problem as it actually presents itself the value of </> 2
at the mouth is prescribed, say
</> 2 = Jcos(nJe); ............... (29)
and in order to identify this with (28) we must have
TcK nC r . / n 2 \ nC
_._, / sin = (l_j)^. (30)
Hence
This determines C in terms of J. If r denote the modulus of
decay of free vibrations, as given by 86 (18), the formula may
also be written
7"2 1 ( f n 2\2 A M 2^
^_ 2 = J_h _M +JL.1LI (32)
Except in the case of approximate synchronism the second
term within the brackets will be small compared with the
first. Hence for a given value of J, the value of nC (which is
the amplitude of the flux q) will be greatest when n = n ,
approximately. Moreover, for a given deviation of the ratio n/n
from unity the intensity of the resonance falls short of the
maximum in a greater proportion the greater the value of W O T,
i.e. the greater the ratio of the modulus of decay to the free
period. In other words, the smaller the damping of free
vibrations, the more sharply defined is the pitch of maximum
resonance. This is in accordance with the general theory
of 13.
The vibrations of a resonator under the influence of an
internal source of sound are discussed in 90 with special
reference to the theory of reedpipes.
89. Mode of Action of an Organ Pipe. Vibrations
caused by Heat.
Although the loss of energy in a single period may be small,
the free vibrations of the column of air contained in an organ
PIPES AND RESONATORS 277
pipe are practically dissipated in a fraction of a second ; this is
owing to the small inertia as compared with that of a piano
wire. For musical purposes some device for sustaining the
note is required. In the ordinary " flute pipe,"
the lower part of which is shewn in section in
Fig. 84, a thin stream of air is driven by pressure
from a windchest so as to strike against the
bevelled lip of the aperture. Under these cir
cumstances a very slight cause will make the jet
pass either wholly inside or wholly outside the
pipe. The precise mode of action is obscure, but
there can hardly be any doubt that in its main
features it is analogous to that of a clockescape
ment. Periodic impulses are given by the jet,
alternately inwards and outwards, to the air near
the mouth, always in the direction in which the
air is tending ; whilst the vibrating column itself
mainly determines the epochs at which these impulses shall
occur. The circumstances are accurately periodic, so that
the driving force can be resolved by Fourier's theorem into
a series of harmonic components whose frequencies are as
1, 2, 3, .... The relative amplitudes with which these are
reproduced in the vibrating column will depend on the
closeness of their frequencies to the natural frequencies. Thus
in a " closed " pipe, i.e. one closed at the upper end, the
harmonics of odd order are alone excited. Again the theory
of 87 indicates that in a sufficiently wide pipe the natural
frequencies may deviate sensibly from the harmonic relation,
in which case only the lower harmonics (after the funda
mental) will be sensible; in particular, a wide closed pipe
gives almost a pure tone. On the other hand a pipe which is
narrow in comparison with the length may give a note rich in
harmonics. Indeed, if such a pipe be blown with sufficient
force, the fundamental is not sounded at all, the period becoming
that of the first harmonic; if the strength of the blast be
further increased the note may jump to the next member of
the series, and so on. An explanation is probably to be found
in the sort of dynamical elasticity possessed by the jet.
278 DYNAMICAL THEORY OF SOUND
Metal pipes are richer in harmonics than wooden pipes of
the same dimensions. This may be partly due to the greater
fineness of the lip, which introduces a greater degree of abrupt
ness in the action of the jet, and so favours the amplitude of
the terms of higher order in the Fourier series which expresses
the driving force. Another source of the contrast in quality
may be found in the smaller rigidity and imperfect elasticity
of the walls of a wooden pipe, which may tend to absorb the
energy, especially in the case of the higher harmonics.
The " speaking " of a resonator of any kind, when a jet of
air is blown across its mouth, is to be explained on the same
principles. In a resonator of the usual type the normal modes
after the first are far removed in pitch from the fundamental,
and are not sensibly excited by the essentially periodic impulse.
The note obtained is therefore a pure tone.
The vibrations of a column of air may also be excited by the
periodic application of heat, as in the wellknown experiment
of the " singing flame," where a jet of hydrogen burns within an
open cylindrical pipe. For the maintenance of the vibration it
is necessary that heat should be supplied at a moment of con
densation, or abstracted at a moment of rarefaction. To explain
how the adjustment is effected, it would be necessary to take
account of the fact that the vibrating system includes the gas
contained in the supply tube of the jet, as well as the column
of air in the pipe. The matter is thus somewhat intricate,
but a satisfactory theory has been made out, which accounts
clearly for the several conditions under which the experiment
is found to succeed or to fail*.
90. Theory of ReedPipes.
The mechanism of the reed stops of the organ is quite
different. The current of air issuing from the windchest is
made intermittent by its passage through a rectangular
aperture in a metal plate, which is periodically opened and
closed (partially) by a vibrating metal tongue, or " reed." The
period is accordingly determined mainly by the elasticity and
inertia of the tongue itself. The vibrations of the latter were
* Lord Rayleigh, Theory of Sound, 322 h.
PIPES AND RESONATORS 279
found by Helmholtz, by direct observation, to be of the simple
harmonic type, but the fluctuations in the current of air are
necessarily of a more complex character. If the periodic
current be expressed by a Fourier series
C + Cicos(n {!)+ C 2 cos(2nt + 2 ) + ..., ...(1)
the coefficients C z , G 3 , ... are usually by no means insensible
as compared with C lt and accordingly if the sound is heard
directly it has a very harsh and nasal character. In practice,
the reed is fitted with a suitable resonator, or " soundpipe,"
which specially reinforces one or more of the lower elements in
the harmonic series (1).
For the purposes of mathematical treatment we may
idealize the question somewhat, and imagine that at a given
point in the interior of the resonator we have a simple source
of the type corresponding to one of the terms in (1). It
appears from the elementary theory of 62 that in the case of
a cylindrical pipe, with the source at one end, the frequencies
of maximum resonance are very approximately those of the
free vibrations when that end is closed. Hence a reed fitted
with a cylindrical soundpipe of suitable length will emit
a series of tones whose frequencies are proportional to the odd
integers 1, 3, 5, .... In a conical pipe, on the other hand, with
the source near the vertex, we have the complete series of
harmonics with frequencies proportional to 1, 2, 3, 4,... (see
84). But in either case the harmonics of high order are
discouraged by the increasing deviation of the frequencies of
maximum resonance from the harmonic relation which neces
sarily holds in the expression for the essentially periodic current
of air.
As the question is instructive in various ways it may be
worth while to examine more in detail the case of a cylindrical
sound pipe (of any form of section), applying the correction
for the open end, and allowing for the dissipation due to the
escape of sound outwards. The plan of the investigation is
similar to that of 87, the difference being that we now have
a source Ce int (say) at the end x = I. For simplicity we will
280 DYNAMICAL THEOKY OF SOUND
assume this source to be distributed uniformly over the cross
section, so that
Let us suppose for a moment that we have a flux
2 = A cos nt ........................ (3)
outwards from the mouth. The pressure at will consist of
two components. We have first the part necessary to control
the inertia of the 'air near the mouth ; the corresponding part
of the velocitypotential just inside is
6  q = cos nt, .................. (4)
CO * ft)
where a has the same meaning as in 87. Next we have the
part which is effective in generating diverging waves outside.
On the principles of 88 this is found to be
7 A cos nt, ..................... (5)
kA
corresponding to 4>^r~ sm nt > ..................... (6)
since k = n/c. The total velocitypotential at 0, corresponding
to (3), is therefore
............. (7)
Generalizing this, we may say that to a flux
q = Ae int ........................ (8)
corresponds <f> = o   e int , . ...(9)
\ 47T/ ft)
the expression (7) being in fact the real part of (9), when A
is real. The correspondence will hold even if A be complex,
since this is merely equivalent to a change in the origin of t.
We now assume, for the region of plane waves,
<t>=r~ {B cos k (I  x)  C sin k (I  x)}e int , . . .(10)
PIPES AND RESONATORS 281
where the constants have been adjusted so as to satisfy (2).
Comparing with (8) and (9) we find
j, vn *.
Hence
B = A {sin H ( ka   ) cos kl\ ,
(12)
!/ ik 2 (o\ )
cos & ( &a ) sin &Z [ .
V 47T/ J /
The latter equation gives A in terms of C. Considering only
absolute values we have
= (cos kl  ka sin kVf + (^)* sin2 kl  ( 13 )
Since fcco is usually a small fraction, the emission of energy,
which varies as \A*\, will be greatest for a given source (7cosn
when
cos kl = kassmkl, (14)
nearly, i.e. when the imposed frequency approximates to that
of one of the normal modes of the pipe when closed at x = I,
as determined by 87 (11). In the case of the reedpipe,
therefore, the tones which are specially reinforced consist of
the fundamental and the harmonics of odd order.
When (14) is satisfied, we have by (12)
C
This determines the relation between the flux outwards at the
mouth, and that constituting the source. The former now greatly
exceeds the latter in amplitude, and the factor i shews that it
differs in phase by a quarterperiod.
Again, from (10) and (11) we have
</> = T \ sin kx + ka cos kx cos kx ! e int . . . .(16)
282 DYNAMICAL THEOEY OF SOUND
When (14) holds, this reduces to
x) r sin kl cos kx e int
'kcoSmkir 47T
{ l"i
,*"? 011 \ cos A; (Z  a)  V sin M cos ^ r e * n ' (17)
) 2 sm 2 ^[ y 4?r j
The real part gives
<f> = = T =; ^ cos k(l x) sin w^ : sin &Z cos ^a? cos w^ [ ,
A^<w 2 sm 2 A;Z ( 4?r j
(18)
corresponding to a source Ccosnt. The variable part of the
pressure at the source (x = I) is
r sin H cos
The first part of this is by far the more considerable ; it is,
moreover, the only part which is effective in doing work. The
mean rate of work done at the source, i.e. the mean value of
pCcosnt, is
ZTTpcC*
"Jfc"Bin"'
It may easily be verified that this is equal to the work spent in
generating waves at the mouth, where, by (15),
tf ................ (21)
7
k*o> sin kl
It appears further from (19) that the maximum of pres
sure at xl synchronises almost with the maximum influx
of air, following it however by a short interval. There is
therefore a tendency slightly to lower the pitch of the reed,
which is, in the instruments here referred to, of the " in
beating type," i.e. the passage is opened when the reed swings
inwards, towards the wind chest. The fact that the resultant
force on the reed is approximately in the same phase with the
displacement indicates that the reed is vibrating with a
frequency somewhat less than that natural to it ( 12).
PIPES AND BESONATOBS 283
The reedstops of an organ fall in pitch as the temperature
rises, owing to the diminished elasticity of the metal tongues ;
this is the opposite of what happens with regard to the flute
pipes ( 62). A reedpipe is tuned by a contrivance which
alters the effective length of the vibrating tongue.
It should be mentioned that there is another class of
instruments in which the " reed " has a much smaller elasticity
and is mainly controlled by the reaction of the resonant
chamber, its own natural frequency being relatively low. The
reed is then of the "outbeating type," the aperture being
widest when the reed swings outwards, i.e. with the wind.
The human larynx comes essentially under this class.
CHAPTER X
PHYSIOLOGICAL ACOUSTICS
91. Analysis of Sound Sensations. Musical Notes.
The vibrations of elastic bodies and the propagation of
waves through the atmosphere are subject to wellascertained
mechanical laws, and the inferences drawn from these can be
controlled by more or less decisive experiments. But when
we approach the field where the human mechanism comes into
play, we are met by the peculiar difficulties which are inherent
in the observation and study of subjective phenomena. In
particular, when we endeavour to analyse a familiar complex
sensation into its elements, we are attempting a task for which
the experience of daily life has peculiarly unfitted us. Thus
we may have been accustomed to interpret the sensation in
question as indicating the presence of a particular object, or
the occurrence of a particular kind of event, in a particular
place. The elements of which it is made up give individually
little or no information; it is the combination which is significant,
and attention to the details would only distract from what is of
immediate practical interest. To use a rough and indeed an
utterly inadequate illustration, it is as if we were to insist upon
spelling every word we read.
The theory of senseperception, especially in relation to
optics and acoustics, is a fascinating subject, but it cannot be
dealt with here. The student who is unversed in it may be
referred to the writings of Helmholtz*.
* The theory is explained in its acoustical bearings in the Tonempfindungen,
already cited (p. 3). It is also discussed from the optical point of view in his
Handbuch der physiologischen Optik, 2nd ed., Hamburg and Leipzig, 1896.
Elementary expositions will be found in the two volumes of his Vortrage und
Reden, Brunswick, 1884, of which there is an English translation by E. Atkinson,
with the title: Popular Lectures on Scientific Subjects, 2nd ed., London, 1893.
PHYSIOLOGICAL ACOUSTICS 285
There are one or two questions, however, relating principally
to Ohm's Law ( 1), to which some reference is necessary. The
first point on which the student should satisfy himself is that
the various simpleharmonic vibrations which are as a rule
combined in the production of a musical note are really
represented by independent elements in the resulting sensation ;
that the latter can in fact be resolved into a fundamental tone
and a series of harmonics. For this a slight course of education
is necessary. A series of resonators of the type shewn in
Fig. 80, p. 261, tuned to the overtones which it is desired to
detect, are of great service for this purpose*. But such
assistance is not indispensable, and a good deal can be effected
with the piano or monochord. Take for instance the note c,
whose harmonics are c', g f , c", e", g ', If on the piano one
of these, say g' } be gently sounded, and the key then released,
so that the vibration is stopped, and if immediately afterwards
the note c be struck with full intensity, it is not difficult to
recognize in the compound sensation the presence of the
element previously heard. This is often more perceptible as
the sound dies away, the overtones being apparently extinguished
more slowly than the fundamental. A more immediately
convincing series of experiments can be made with the
monochord, or with a piano whose strings are horizontal and
therefore easily accessible. If a string be set into vibration
whilst damped at a nodal point of one of the harmonics by
contact with a hairpencil, the fundamental tone and all the
harmonics of lower rank may be reduced in intensity or
altogether extinguished, according to the degree and duration
of the pressure applied. In this way a whole series of types
of vibration can be produced in which the harmonic in question
is accompanied by a varying admixture of the fundamental, &c.
The occurrence throughout of the corresponding sensation as
an independent element in the resulting sound is in this way
easily appreciated. The piano also lends itself readily to the
* It may be noted that the external earcavity is itself a resonator,
responding most intensely to a certain tone, which varies for different
individuals but is usually in the neighbourhood of ff* g ir . The aperture
being relatively large, the damping and consequently the range of resonance
is considerable.
286 DYNAMICAL THE OK Y OF SOUND
analysis of compound notes by resonance. If the string c', for
example, be freed from its damper by holding down its key,
whilst c is sounded for a moment, the harmonic c is taken up
and continued by the firstmentioned string. If on the other
hand, the string c be free from its damper, whilst c is sounded
for a moment, the tone c' is taken up as a harmonic of the
lower string. These simple experiments, which (with others)
are recommended by Helmholtz, can of course be varied in
many ways. Again, when the ear has learnt to distinguish the
partial tones in a complex note, it is easy to note the absence
of a particular tone of the series when the corresponding simple
harmonic vibration is not excited. For instance, when a string
is struck at its middle point, the harmonics of even order are
wanting ( 26).
92. Influence of Overtones on Quality.
The quality of a musical note is determined ( 2) by the
number and relative intensities of the various tones which
compose it. The kind of influence which overtones of different
ranks exercise on the quality is summarised by Helmholtz,
somewhat as follows:
1. Pure tones like those of tuning forks with resonance
boxes, or of wide stopped organ pipes, are soft and pleasing,
smooth, but wanting in power.
2. Notes which contain a series of overtones up to the
fifth or sixth in rank are richer and more musical, and are
perfectly smooth so long as no higher overtones are sensible.
The notes of the piano and of open organ pipes are examples,
whilst those of the flute, and of the flutestops on the organ
when softly played, approximate more to the character of pure
tones. In the "mixture" stops of the organ the lower
harmonics are expressly provided in greater intensity by
auxiliary pipes which are played automatically along with that
which gives its name to the note.
3. When the harmonics of even order are absent, as in the
case of a stopped organ pipe, or a piano string struck at the
middle point, the note has a hollow, and even a nasal character,
if the odd harmonics are numerous.
PHYSIOLOGICAL ACOUSTICS 287
4. The sound may farther be described as "full," if the
fundamental tone be predominant, and as "empty" if it be
relatively feeble. This is exemplified in the difference of
quality between the sound of a pianowire when struck with
a soft or a hard hammer, respectively ( 26, 38).
5. When harmonics beyond the sixth or seventh are
present in considerable intensity, the sound is harsh and
rough, owing to the discords which these higher overtones
make with one another. If, however, the higher harmonics,
though present, are relatively weak, as in the case of the
stringed instruments of the orchestra, reedpipes, and the
human voice, they are useful as giving character and expression
to the sound. Brass instruments, on the other hand, with
their long series of powerful overtones, are as a rule only
tolerable in combination with others, or for the sake of
particular effects.
The analysis of the sounds of the human voice is naturally
a more difficult matter. In particular, the constitution of the
vowel sounds has been much debated, without any very definite
conclusion. The same vowel may be sung on a wide range of
notes, but preserves its peculiar character throughout ; and the
question arises, does this special quality depend solely on the
relative intensities of the various partial tones, or on the
predominance of one or more overtones of, or near to, a
particular pitch ? It will be remembered that the vibration of
the larynx is periodic, and that particular harmonics may be
reinforced by the resonance of the mouthcavity, as in the case
of a reedpipe ( 90). The balance of authority appears to
incline, though not very decisively, to the " fixedpitch " theory,
which is the second of the two alternatives above stated. A
review of the subject down to the year 1896 will be found in
the concluding chapter of Lord Rayleigh's treatise.
93. Interference of Pure Tones. Influence on the
Definition of Intervals.
It has so far been assumed that the sensations due to
two coexistent simpleharmonic vibrations are produced quite
independently of one another. This appears to be in fact the
288 DYNAMICAL THEOBY OF SOUND
case when the interval between the two tones is sufficiently
great ; but when the interval is small we have " interference,"
as we should expect from the analysis of 10, and the sensation
is in whole or in part intermittent. The phenomenon of" beats"
hardly needs description ; it is often met with in mistuned pairs
of pianowires, in the vibrations of fingerbowls, and so on. For
methodical study two pure tones are required of equal intensity,
as e.g. from two tuning forks (with resonators), or two stopped
organ pipes, which can be made to differ in pitch by a variable
amount. As unison is departed from, the beats (whose
frequency is always equal to the difference of the frequencies
of the primary tones) are at first slow and easily counted. As
the interval widens they become more rapid, and a sensation of
roughness or discord is experienced ; moreover, the primary
tones are now heard along with the beats. Finally, as the
interval is continually increased, the beats and the consequent
roughness gradually cease to be perceptible.
The intervals at which roughness begins and ceases, vary in
different parts of the scale. For the same interval the rough
ness is less, the higher the pitch ; on the other hand for a given
number of beats per second the roughness is greater in the
higher octaves.
In the case of two (or more) compound musical notes, we
may have beats and eventual roughness between any constituent
tones which are sufficiently near in the scale. We may even
have interference between the higher overtones of the same
note ; and it is for this reason that harmonics of higher order
than the sixth are prejudicial to good musical quality.
It is through the interference of pairs of overtones that
deviations from the consonant intervals ( 3) usually make
themselves felt. Thus in the case of the Octave cc' we have
tones with the frequencies
c = 132, 264, 396, 528, 660, 792, ...,
c' = 264, 528, 792, ...,
and if this be mistuned all the harmonics of c' are interfered
with by the even harmonics of c.
PHYSIOLOGICAL ACOUSTICS 289
In the case of the Fifth eg we have
c = 132, 264, 396, 528, 660, 792, ...,
g= 198, 396, 594, 792, ...,
and if this be mistuned the second tone of g beats with the
third tone of c, and so on. When the ratio of the vibration
numbers of the fundamentals is less simple, the harmonics
which can interfere are of higher order. Thus in the case of
the Major Third, where the ratio is 4:5, the first pair of
interfering overtones consists of the fifth tone of the lower note,
and the fourth of the higher. Since in many musical instruments
the fifth tone is very feeble, this consonance is less well defined
than the preceding ones. On the other hand the fundamentals
may fall, in the lower parts of the scale, within beating distance
(for example c= 132, e 165), so that this consonance is to be
reckoned also as less perfect than the former ones. Similar
remarks apply with greater force to such cases as the Minor
Third (5 : 6) and the Minor Sixth (5 : 8).
94. Helmholtz Theory of Audition.
The connection between primary sensations and simple
harmonic vibrations has still to be accounted for. The problem
is a physiological one; but the theory which Helmholtz has
framed to explain Ohm's law, so far as it holds, and the various
deviations from it, is in its essentials so simple, and is so
successful in binding together the facts of audition into a
coherent system, that a brief statement of it may be attempted.
In its simplest form the theory postulates the existence,
somewhere in the internal ear, of a series of structures each
of which has a natural period of vibration, and is connected
with a distinct nerveending. For brevity we will speak of
these structures as "resonators," since that is their proper
function. A particular resonator is excited whenever a
vibration of suitable frequency impinges on the ear; the
appropriate nerve is stimulated ; and the sensation is com
municated to the brain. In this way the resolution of a
musical note into its constituent tones is at once accounted
for.
It is necessary to suppose that the resonators are subject
L. 19
290
DYNAMICAL THEOEY OF SOUND
to a considerable amount of damping. If it were not so,
each resonator would go on vibrating, and the corresponding
sensation would persist, for an appreciable time after the
exciting cause had ceased. A similar interval of time would
elapse before the sensation reached its full intensity when the
cause first sets in. The effect would be that the sensations
due to a sufficiently rapid succession of distinct notes would
not be altogether detached from one another in point of time.
From considerations of this kind Helmholtz estimated that
the degree of damping must be such that the intensity (as
measured by the energy) of a free vibration would sink to
onetenth of its initial value in about ten complete vibrations.
It follows, as explained in 13, that each resonator will
respond to a certain range of frequencies on each side of
the one which has maximum effect. It is assumed, further,
that the difference of pitch of adjacent resonators is so small
that the same simpleharmonic vibration will excite a whole
group, the intensity falling off from the centre on either side.
This is illustrated by the annexed figure, repeated from 13,
which may now serve to exhibit the distribution of intensity
over a continuous series of resonators under the influence of
a given simpleharmonic vibration. The abscissa is p/n 1,
where p is now taken to represent the natural frequency of a
resonator, and n that of the imposed vibration. The horizontal
scale depends on the value of fi, or 1/rw, where r is the
PHYSIOLOGICAL ACOUSTICS 291
modulus of decay of a free vibration. On the above estimate
of Helmholtz we shall have
whence = '018. The intensity is therefore onehalf the
maximum for
2 = 1+ 018.
It will be observed that on the above view we ought in
strictness to speak of "simplest" rather than of "simple"
sensations of sound, absolutely simple sensations, in the strict
physiological meaning, being impossible to excite.
When two simpleharmonic vibrations, sufficiently far apart
in the scale, are in operation, the two groups of resonators
which are affected will be practically independent, and the
two sensations (of pure tones) will coexist. But when the
interval between the frequencies is sufficiently small, the two
groups will overlap, and the energy of vibration of those
resonators which are common to them will fluctuate in the
manner explained in 10. The excitation of the corresponding
nerveendings will therefore be intermittent, with a frequency
equal to the difference of those of the two originating vibrations.
This is, on the theory, the explanation of beats. As the interval
is increased, the beats become more rapid. The " roughness "
which is ultimately perceived, in spite of the diminishing
amplitude of the fluctuations, has a more remote physiological
explanation. According to Helmholtz, there is here an analogy
with the painful effect produced by a flickering light, and in
other cases where a nerve is stimulated repeatedly at intervals
of time which are neither too great nor too small. When the
intervals are sufficiently long, the nerve has time to recover
its initial sensibility, and so experiences the full effect of each
recurring stimulus. When on the other hand the intervals
are sufficiently short, the sensation tends to become continuous.
It is for this reason that beats exceeding, say, 132 per second
cease to produce the sensation of roughness, even although
the interval between the beating tones be such as would be
perceptibly discordant in a lower part of the scale.
192
292 DYNAMICAL THEOKY OF SOUND
The student of dynamics cannot fail to admire the beauty
of a theory which lends itself readily to the explanation of so
many complicated relations; but it is with the physiologist
and the anatomist that in the last resort it lies to decide
whether a mechanism of the kind postulated is really to be
found in the internal ear. In the original form of the theory
(1862) the resonators were identified with the structures known
as "Corti's rods," which are found arranged, some 3000 in
number, along the basilar membrane in the spiral cavity
of the cochlea. A disturbing discovery by Basse that these
structures do not occur in the ears of birds, to whom we can
hardly deny the perception of pitch, led to a modified form
of the theory. In the third edition of the Tonempfindungen
(1870) Helmholtz propounded the view that the resonators
consist of the various parts of the basilar membrane itself.
This membrane varies in breadth from one end to another,
like a very acuteangled triangle, and the tension appears to
be very much less in the direction of length than in that
of breadth. On this view the different parts could be set
into sympathetic vibration, much as in the case of a series
of strings of variable length placed side by side, except that
the independence of adjacent parts would be approximate
instead of absolute. For a full description of the complicated
structure of the internal ear, and for further speculations as
to the functions performed by its various parts, we must refer
to books on physiology.
95. CombinationTones.
In one important respect the theory as so far developed
is inadequate. The explanation of consonant intervals outlined
in 93 assumes that one at least, and generally both, of the
notes concerned is complex, and contains one or more overtones
in addition to the fundamental. It was in fact through the
interference of two tones, one at least of which is an overtone,
that departure from the exact relation of pitch was stated to
make itself manifest. When both tones are pure this means
of definition is wanting, and on the theory of audition sketched
in the preceding section there appears to be no reason why
PHYSIOLOGICAL ACOUSTICS 293
the octave (for example) should be distinguished by any
character of smoothness from adjacent intervals on either
side, the two groups of sensations being in any case quite
independent. Since the more consonant intervals at all events
are as a matter of fact easily recognized by the ear, even in
the case of apparently pure tones, and are thoroughly well
defined, the difficulty is a serious one. To meet it, Helmholtz
developed his theory of "combinationtones," which are assumed
to supply the function of the missing overtones.
In most of our investigations it has been assumed that the
amplitude of the vibrations may be treated as infinitely small,
so that disturbances due to different sources may be super
posed by mere addition. In the theory now in question this
assumption is abandoned ; the vibrations are regarded as
small, but not as infinitely small, and the interaction of the
disturbances due to different causes is, to a certain degree of
approximation, investigated.
We have already had an indication in 63 of the manner
in which two imposed simpleharmonic disturbing forces of
small but finite amplitude, with frequencies N lf N 2 respectively,
may generate in the air other simpleharmonic vibrations
whose frequencies are
2N lt 2# 2 , NiN,, Ni + N*,
and whose amplitudes involve the squares or product of the
amplitudes of the two primaries. If the approximation were
continued we should meet with further vibrations whose fre
quencies are of the type p l N l piN^ where p lt p 2 are integers.
In acoustical language, two simpleharmonic vibrations can, if
of sufficient intensity, give rise not only to the pure tones
usually associated with them, but also to a series of other pure
tones of higher order. The fact that a single harmonic vibration
can by itself give rise to a pure tone together with its octave, &c.
is itself of some importance, but the most interesting result is
due to the interaction, viz. the " differencetone " (N^ N 9 ).
The existence of differencetones was observed, apart from
all theory, by Sorge (1745) and Tartini (1754). The "sum
mationtone " (Ni + N 2 ) is more difficult to hear, and its
294 DYNAMICAL THEORY OF SOUND
existence has even been denied. It has however been objectively
demonstrated by Riicker and Edser*, by its effect on a tuning
fork of the same frequency.
Differencetones due to the causes just considered are most
easily perceptible where we have a mass of air which is subject
to the joint and vigorous action of the primary vibrations, as
in the harmonium and the siren; they can then, like other
tones, be reinforced by suitable resonators.
There is however a way in which combination tones may
conceivably be originated in the ear itself. To explain this
it is necessary briefly to consider the forced vibrations of an
unsymmetrical system. When a particle, or any system having
virtually one degree of freedom, receives a displacement x, the
force (intrinsic to the system) which tends to restore equilibrium
is a function of x, and may be supposed expressed, for small
values of x, by a series
An example is furnished by the common pendulum, where
the force of restitution is proportional to g sin 6, or
but here, on account of the symmetry with respect to the
vertical, the force changes sign with 6, so that
only odd powers of 6 occur. The correction for
small finite amplitudes depends therefore on the
term of the third order in 6. But if the system
be unsymmetrical, as in the case of a pendulum
hanging from the circumference of a horizontal
cylinderf, the term of the second order comes in,
and the correction is more important. Helmholtz Fi
lays stress on the fact that in the slightly
* Phil. Mag. (5), vol. xxxix. (1895).
f If a be the radius, and I the length of the free portion of the string when
vertical, the potential energy is
where .9 is the arc described by the bob from the lowest position. The restoring
force is therefore
dV_mg a lmga^ t
PHYSIOLOGICAL ACOUSTICS 295
funnelshaped tympanic membrane and its connections we
have precisely such an unsymmetrical system, the restoring
force being somewhat greater for inward than for outward
displacements of the same extent*. If we keep only the first
two terms in (1), the equation of motion is of the type
x+/juc = auc*+X, .................. (2)
where X represents the disturbing force f. The joint action
of two simpleharmonic forces will be represented by
X =f 1 cosn l t+f z cosnj ................ (3)
Neglecting, for a first approximation, the square of #, we have
x = l .. cos nj \   cos n 2 t, ......... (4)
/nn,* fjLnJ
the terms which represent the free vibrations being omitted,
since these are rapidly destroyed by dissipation. If we sub
stitute this value of x on the right hand of (2), and write for
shortness
AAfc %*)** / 2 /(/*^ 2 )=# 2 , ......... (5)
we obtain the differential equation
x + x = X  a + 2 2  af cos 2nJ  a. 2 cos
 <*9i92 cos (HJ  O t  ag,g z cos fa + w 2 ) t, . . .(6)
correct to the second order of /i,/ 2  The terms written in full
on the right hand may be regarded as a correction to the
disturbing force X. The solution of (6) gives, in addition to
(4), the terms
The first term merely indicates a shift of the mean position
* It may be noted that the same element of asymmetry is present in the
investigation of 63. When we proceed to the second order of small quantities,
the changes of pressure due to condensations s are no longer equal in
amount.
f It is unnecessary to take account of the variability of inertia, since this
can be got rid of by a proper choice of the coordinate x. In any case
it will not alter the general character of the results obtained in the second
approximation.
296 DYNAMICAL THEOEY OF SOUND
about which the oscillations take place. For the rest, we have
octaves of the primary tones, together with a difference and
a summationtone. If the approximation were continued we
should obtain combinationtones of higher order, as in the
former case.
When, as in the case of the tympanic membrane, the
free period 2?r/\//A is relatively long, the most important
combination tone is the differencetone (n^ n 2 ), on account of
the relative smallness of the corresponding denominator in (7).
The theory of combinationtones here reproduced has not
been accepted without question. The differencetones, as
already mentioned, were known as a fact since the time of
Tartini, and a plausible explanation had been given by
Thomas Young (1800). According to this view the beats
between the two tones, as the interval increases, ultimately
blend, as if they were so many separate impulses, into
a continuous tone having the frequency of the beats. The
difficulty of this explanation is that the actual impulses
during a beat are as much positive as negative, so that
it does not appear how any appreciable residual effect in
either direction could be produced, if the vibrating system
be symmetrical. It is true that if we turn to the figure on
p. 23, it is apparently periodic, with the period of the in
termittence ; but from the point of view of Fourier's theorem
the lower harmonics are all wanting, and the only two which are
present are precisely the two which are used in constructing the
figure. On the Helmholtz theory of audition the intermittent
excitation of a particular resonator m times a second is a wholly
different phenomenon from the excitation of an altogether
distinct resonator whose natural frequency is m. Young's
view appears indeed to be inadmissible on any dynamical
theory of audition, at least in the case of infinitely small
vibrations. On the other hand it is true, as we have seen,
that given a finite amplitude, and an unsymmetrical system,
a vibration of the type shewn in Fig. 10, p. 23, does actually
generate (among others) a vibration whose period corresponds
to the fluctuations there shewn. The distinction between the
two theories might therefore, from a merely practical point of
PHYSIOLOGICAL ACOUSTICS 297
view, be held to be almost verbal, were it not that Young's
theory fails to give an explanation of combinationtones other
than the first differencetone.
96. Influence of CombinationTones on Musical In
tervals.
A brief indication of the way in which combinationtones
may assist in defining the consonant intervals is all that can be
attempted here. Take first the case of (primarily) pure tones.
In the case of a slightly mistuned Octave, say ^ = 100,
N 9 = 201, we have N z N 1 = 101, which gives a differencetone
making 1 beat per second with N l9
For the Fifth, let N, = 200, N 2 =301. We have
giving combinationtones with 2 beats per second.
For the Fourth, let ^ = 300, N 2 = 401. Then
2^^=199, 2^2^
and the corresponding tones make 3 beats per second.
For the Major Third, let ^=400, JV 2 = 501. We have
2^2^ = 202, 3^2^=198, giving 4 beats per second.
We might proceed further in the list, but it will already
have been remarked that combinationtones of increasingly
high order are being invoked. This is quite in conformity
with the observed fact that the beats are, in all cases after the
octave, very faint unless the primaries be especially vigorous.
A more effective part is played by the combinationtones
when the notes concerned have one or two overtones, but not a
sufficient range of them to account for the definition on the
principles of 93. Take for instance the case of the Fifth,
when each note has a first harmonic in addition to the
fundamental. If the interval be slightly mistuned, we have say
the primary tones: 200, 400; 301, 602. These give the two
differencetones 301  200 = 101, 400  301 = 99, which inter
fere with one another.
The combination tones have an influence again, in the case
298 DYNAMICAL THEOKY OF SOUND
of consonant triads, especially of simple tones, but enough
has been said to shew their importance from the musical point
of view. For further developments reference must be made to
the work of Helmholtz*.
97. Perception of Direction of Sound.
One important question of physiological acoustics in which
dynamical principles are involved remains to be mentioned.
An observer, even when blindfolded, and with no adventitious
circumstances to guide him, is in general able to indicate with
great accuracy the direction from which a sound proceeds. In
the case of pure tones the discrimination between back and
front is indeed lost, as was to be expected, considering the
symmetry with respect to the medial plane of the head, but
right and left are clearly distinguished. For tones of small
wavelength this may be accounted for by the difference of
intensity of the sensation in the two ears, since the head acts
to some extent as a screen, as regards the further ear. But
when the wavelength of the sound much exceeds the peri
meter of the head the investigation given near the end of 81
shews that this difference must be very slight. According to
the most recent investigations of Lord Rayleigh^, the in
terpretation depends on the relative phase of the sounds as they
reach the two ears, a difference of even a fraction of a period
being effective. He found that if the same tone be led by
different channels to the two ears, and all extraneous dis
turbances be excluded, the sound can be made to appear to
come from the right or left at will, by adjusting the relative
phase. The origin of the sound was always attributed to that
side on which the phase is in advance (by less than half a
period). The result, which has been arrived at independently
by other observers, is at present unexplained. It has been
suggested that the phenomena may really be due to a differ
ence of intensity. A fraction of the sound may be transmitted
from each side to the opposite internal ear, through the bones of
* See also Sedley Taylor, Sound and Music, London, 1873.
t Phil. Mag. (6), vol. xm. (1907).
PHYSIOLOGICAL ACOUSTICS 299
the head, in which case the original difference of phase would
produce a slight difference of intensity on the two sides owing
to interference between the direct and transmitted vibrations*.
It is impossible to suppose, however, that this difference could
be other than exceedingly minute.
* Myers and Wilson, Proc. Roy. Soc. vol. LXXX. A, p. 260 (1908). This
hypothesis is discussed by Lord Kayleigh, Proc. Roy. Soc. vol. Lxxxm. A, p. 61
(1909).
INDEX
[The numerals refer to the pages]
Absorption of sound, 196
Adiabatic lines, 158
Airwaves, general theory of, 204
see also Sound waves
Amplitude, minimum audible, 167
Analysis of sound sensations, 2, 284
Anticlastic curvature of a flat bar, 152
Approximate solution of period
equations, 83, 125, 126, 128, 256,
258
Audibility, range of frequency for, 3
least amplitude for, 167
Audition, Helmholtz theory of, 289
Bars, longitudinal vibrations, 114
flexural vibrations, 120, 130
Beats, 23, 132, 138
relation of, to dissonance, 288
Bells, 155
Bessel's functions, 85, 145, 147, 259
Blackburn's pendulum, 35
Chain, vibrations of hanging, 84
Circular vibrations, 55
' Circulation ' denned, 203 *
Clampedfree bar, transverse vibrations
of a, 127
Combinationtones, 181, 292, 294
Communication of vibrations to a gas,
237
' Condensation ' defined, 160
Conduction of heat, effect of, on sound
waves, 187
' Conductivity ' of an aperture, 245
Conical pipe, normal modes of a, 257
Consonant intervals, 3, 283, 297
Cosineseries, 92
Curved shells, vibrations of, 155
Cylindrical vessel, normal modes of a,
259
Damping of vibrations, 25, 27, 57
effect of, on resonance, 32
of air waves by viscosity, 185, 186,
190
of a resonator, 265
of an organ pipe, 269
Degrees of freedom of a dynamical
system, 12, 14
Diatonic scale, 5
Diffraction of sound, 240, 244, 248
'Dilatation' defined, 107
Direction of sound, perception of,
298
Discontinuity, waves of, 181
Dissipation of energy by friction, 27,
184
Dissipation (apparent), by generation
of airwaves, 166, 225, 227, 232, 264,
269
'Divergence' defined, 199
Doppler's principle, 221
Double pendulum, 38
' Double source ' of sound, 215, 226
Elasticity, elementary theory of, 106
coefficients of, 110, 113
of gases, 159
Elliptic vibrations, 49
Emission of energy, by a simple source,
225
by a double source, 227
by a resonator, 265
by an open pipe, 269
Energy, of a simpleharmonic vibra
tion, 15
of a string, 60
of an elastic solid, 114
of a bar, 123
of a membrane, 141
of a bent plate, 151
of airwaves, 163, 205
'Extension' defined, 107
Extensional vibrations of a rod, 114
of a circular ring, 136
Finite amplitude, airwaves of, 174
Flexure, uniform, of a bar, 121
of a plate, 150
Flexural vibrations, of a bar, 12Q
of a ring, 136
of a plate, 153
'Flux,' defined, 199
< Flux of energy, '165, 224
302
INDEX
[The numerals refer to the pages']
Forced oscillations, 16, 20, 47
effect of friction on, 28, 57, 104
Fork, tuning, 132
Fourier's theorem, 87, 92
influence of discontinuities in, 92
law of convergence of coefficients in,
94
Freedom, degrees of, 12, 34
Free free bar, transverse vibrations of
a, 124
Free oscillations, 12
with friction, 24
general theory of, 44
Frequency, range of, for audibility, 3
Friction. See Dissipation
Gas, elasticity of a, 159
isothermal and adiabatic lines of a,
157, 158
Graphical solution of periodequations,
83, 125, 126, 128, 256, 258
Grating, transmission of sound by a,
247
Harmonic analysis, 101
Harmonics, 5
Heat, vibrations caused by, 278
Heatconduction, effect of, on sound
waves, 187
Hooke's law of elasticity, 11, 110
Huygens' principle, 249
Imaginaries, use of, 53
Impact, vibrations of a string due to,
73, 99
Indicator diagram, 157
Inertia, coefficients of, 42
' Irrotational ' motion defined, 203
Interference of simpleharmonic vibra
tions, 23
of pure tones, 287
Intervals, musical, 5, 288
degree of definition of, 297
Isothermal lines, 157
Laplace's equation, 205
Leslie's experiment, 237
Lines of motion, 229
Lissajous' figures, 49
Loaded string, normal modes of a, 36,
37,82
Local periodic force, effect of, in a
gaseous medium, 233
Longitudinal vibrations, of bars,
114
of columns of air, 170, 266
Loops, on a vibrating string, 70
in a pipe, 171
Membrane, transverse vibrations of a,
139
normal modes of a rectangular,
142
of a circular, 144
Mersenne's laws, 70
'Modulus of decay,' defined, 25
of airwaves, 185
of a vibrating sphere, 233
of a resonator, 265
of a pipe, 270
Modulus, Young's, 111
Multiple system, equations of motion
of a, 41, 44
normal modes of a, 44
forced vibrations of a, 47
Nodal lines, of a membrane, 143, 144,
148
of a plate, 153, 154
Nodes, in a vibrating string, 70
in a bar, 116, 127
in a pipe, 171
' Normal functions,' 101, 130
Normal modes of vibration, 44
Notes, musical, 3, 285
Ohm's law, 2, 285, 289
Organ pipe, normal modes of, 171
corrected theory of, 266
mode of action of, 276
Overtones, 5
influence of, on quality, 286
on the definition of consonant
intervals, 288
Pendulum, 8, 16
Blackburn's, 35
double, 38
Periodequations, graphical solution
of, 83, 125, 126, 128, 256, 258
Permanency of type, condition for, in
airwaves, 174
Pipe, normal modes of a, 171
modulus of decay of a, 270
velocity of sound in a narrow, 193
Plane waves in an elastic medium, 118
in air, 160, 174, 223
Plate, transverse vibrations of a cir
cular, 153 ; of a square, 154
Plucked string, theory of, 66, 98
Poiseuille's law, 195
Poisson's ratio, 111
'Quality' of musical notes, 4
influence of overtones on, 286
Beciprocity, principle of, 47, 81
INDEX
303
[The numerals refer to the pages]
Rectangular vessel, normal modes of
a, 254
Reedpipes, theory of, 278
Reflection of waves, 64, 168, 215,
267
Refraction of sound, due to variation
of temperature, 216
to wind, 219
Resonance, 18, 20, 22, 32, 270
Resonator, 261
free vibrations of a, 262
forced vibrations of a, 270
Ring, normal modes of a, 133
Scattering of sound waves by obstacles,
240
Sensations, analysis of, 2, 284
Shearing strain, 107
stress, 109
Shells, vibrations of curved, 155
' Simple source' of sound, 214, 224
Simpleharmonic vibrations, 2, 9
energy of, 15
superposition of, 22, 48
Sineseries, 87
Sound, velocity of, in air, 161, 162
in water, 163
Sound waves, plane, 160
spherical, 20o, 224
general, 204, 212, 214
of finite amplitude, 174
Sounding board, function of, 68, 81
Source of sound, simple, 214, 224
double, 215, 226
'Speed' of a simple vibration, 10
Sphere, waves produced by oscillating,
228
vibrations of an elastic, 156
Spherical vessel, normal modes of a,
255, 258
Stability, coefficients of, 43
Stationary property of normal modes,
45
Stiffness of pianowire, effect of, 82,
133
Strains, 106
Stresses, 108
String, transverse vibrations of a, 59
waves on a, 61, 64
normal modes of a finite, 68
forced vibrations of a, 80
String excited by plucking, 66, 72, 98
by impact 73, 99
by bowing, 75, 98
Superposition of vibrations, 22, 48
Temperament, equal, 7
Temperature, effect of unequal, on
propagation of sound, 216
Tension, effect of permanent, on the
vibrations of a bar, 132
Tones, pure, 1
interference of, 287
Transmission of sound by an aperture,
244
by a grating, 247
Transverse vibrations, of strings, 59
of bars, 120
of membranes, 139
of plates, 152
Tuning fork, 132
Velocity of sound, 161, 162
in a narrow pipe, 193
'Velocitypotential,' 201
Violinstring, 75, 98
Viscosity, 183
effect of, on air waves, 185, 186,
190
on waves in a narrow pipe, 193, 194,
196
Water, velocity of sound in, 163
vibrations of a column of, 173
Watt's indicator diagram, 157
Waves, on a string, 61, 64
in a bar, 115, 123
in an elastic medium, 118. See
also Sound waves
Wind, influence of, on sound propa
gation, 219
Young's modulus, 111
CAMBRIDGE : PRINTED BY JOHN CLAY, M.A. AT THE UNIVERSITY PRESS.
14 DAY USE
RETURN TO DESK FROM WHICH BORROWED
LOAN DEPT.
This book is due on the last date stamped below, or
on the date to which renewed.
Renewed books are subject to immediate recall.
REC'D
APR
LD 21A60m4,'64
(E4555slO)476B
General Library
University of California
_aer^ejev __
_^i3JJ
,
LD 2110m5,'4
YC 10940
THE UNIVERSITY OF CALIFORNIA LIBRARY