1917 - Treatise Analytical Dynamics

A TREATISE
ON THE
ANALYTICAL DYNAMICS
OF PARTICLES AND RIGID BODIES
CAMBRIDGE UNIVERSITY PRESS

C.
FETTER LANE, E.G.

100 PRINCES STREET
Soulum:
lirfu gorft:
Bombag,
Calcutta
CLAY, MANAGER
F.
G.
PUTNAM S SONS
MACMILLAN AND CO.,
DENT AND SONS, LTD.
P.
anto Jftafcras:
Toronto:
Til
J.
1C
M.
MARUZEN-KABUSHIKI-KAISHA
All rights reserved
LTD.
A TREATISE
ON THE
ANALYTICAL DYNAMICS
OF PARTICLES AND RIGID BODIES;
WITH AN INTRODUCTION TO THE
PROBLEM OF THREE BODIES
,
BY
E.
T.
Hon.
WHITTAKER
Sc.E).
(Dubl.)
Professor of Mathematics in
F.R.S.
the University of Edinburgh
SECOND EDITION
CAMBRIDGE
AT THK UNIVERSITY
1917
PRESS
Uf
irst
Edition 1904
Second Edition 1917
PREFACE TO THE SECOND EDITION
TN
revising this book for a second edition, I have endeavoured to givereferences to,
and in some cases accounts
of,
the
numerous
original
Dynamics which have been published by various investigators

I have moreover added some historical
since the first edition appeared.
researches in
matter, and rewritten

in
detail,
many sections. It is not necessary to specify these

but perhaps I may mention that the new explanation of the
125 sprang from a desire to do

justice to the earliest great work of Hamilton s genius that the changes in
69 (the motion of a body about a rixed point under no forces) arose from
transformation-theory of Dynamics in
my
opinion that the Jacobian functions are preferable to the Weierstrassian
in
the
numerical computations
a fuller proof of
and that
Sundman s theorem
should have liked
181), but thought
to
give
it
better to give
only such an account as might impel the reader to consult
Mr Sundman s
own
accessible
and readable memoir.
I wish again to
record
my
obligations to the staff of the
Cambridge
University Press.
E. T.
WHITTAKER.
EDINBURGH,
August, 1916.
a3
CONTENTS
CHAPTER
I.
KINEMATICAL PRELIMINARIES.
PAGE
SECTION
The displacements
3.
of rigid bodies
Euler s theorem on rotations about a point
The theorem of Rodrigues and Hamilton
4.
The composition
1.
2.
5.
6.
7.
8.
......
.......
and opposite rotations about parallel axes

Chasles theorem on the most general displacement of a rigid body
Halphen s theorem on the composition of two general displacements
of equal
Analytic representation of a displacement

The composition of small rotations
2
3
...
13.
Euler s parametric specification of rotations round a point

The Eulerian angles
Connexion of the Eulerian angles with the parameters
?;,
^
The connexion of rotations with homographies the Cayley-Klein parameters
Vectors
14.
Velocity and acceleration
15.
Angular velocity its vectorial character

Determination of the components of angular velocity of a system in terms
of the Eulerian angles, and of the symmetrical parameters
Time-flux of a vector whose components relative to moving axes are given
Special resolutions of the velocity and acceleration
15
MISCELLANEOUS EXAMPLES
22
9.
10.
11.
12.
16.
their vectorial character
18.
CHAPTER
II.
THE EQUATIONS OF MOTION.
10
11
13
14
17.
16
17
18
Contents
vii
PAGE
SECTION
26.
Lagrange s form of the equations of motion of a holonomic system
27.
Conservative forces
28.
The
29.
30.
31.
32.
...
34.
35.
Impulsive motion
36.
The Lagrangian equations

34
38
explicit form of Lagrange s equations

Motion of a system which is constrained to rotate uniformly round an axis
The Lagrangian equations for quasi-coordinates
Forces derivable from a potential-function which involves the velocities
Initial motions
.
the kinetic potential
Similarity in dynamical systems

Motion with reversed forces
33.
39
40
41
44
45
47
47
......
of impulsive motion
...
48
50
51
CHAPTER
III.
PRINCIPLES AVAILABLE FOR THE INTEGRATION.

38.
Problems which are soluble by quadratures

Systems with ignorable coordinates
39.
Special cases of ignoration
40.
The general theorem

The energy equation
37.
41.
42.
43.
52
.
......
momentum and
momentum
integrals of
of angular
angular
momentum
54
58
61
62
Reduction of a dynamical problem to a problem with fewer degrees of

freedom, by means of the energy equation
Separation of the variables dynamical systems of Liouville s type
64
69
CHAPTER
67
IV.
THE SOLUBLE PROBLEMS OF PARTICLE DYNAMICS.
.... ......
47.
The particle with one degree of freedom the pendulum

Motion in a moving tube
Motion of two interacting free particles
Central forces in general
Hamilton s theorem
48.
The
49.
Motion under the Newtonian law

The mutual transformation of fields of central force and
44.
45.
46.
50.
76
77
..........
............
integrable cases of central forces

and elliptic functions
71
74
problems soluble in terms of circular

80
86
fields of parallel
"
force
93
51.
Bonnet s theorem
52.
Determination of the most general field of force under which a given curve
or family of curves can be described
95
53.
The problem
97
54.
Motion on a surface
Motion on a surface of revolution
55.
of
two centres of gravitation
elliptic functions
56.
99
...........
...........
Joukovsky s theorem
94
cases soluble in terms of circular and
103
109
111
Contents
vii.i
CHAPTER
V.
THE DYNAMICAL SPECIFICATION OF BODIES.

SECTION
PAGK
57.
Definitions
58.
The moments
59.
Derivation of
60.
about a parallel axis through the centre of gravity is known

Connexion between moments of inertia with respect to different sets of axes
61.
The
some simple bodies

the moment of inertia about any
of inertia of
axis
when the moment
117
-.
118
of
121
inertia
.......
through the same origin
122
63.
principal axes of inertia

Cauchy s momenta! ellipsoid
Calculation of the angular momentum of a moving rigid body
Calculation of the kinetic energy of a moving rigid body
64.
Independence of the motion of the centre of gravity and the motion relative
62.
to
124
124
126
127
it
129
CHAPTER
VI.
THE SOLUBLE PROBLEMS OF RIGID DYNAMICS.

65.
The motion
of systems with one degree of freedom
motion round a fixed

131
axis, etc.
66.
The motion
67.
Initial
68.
The motion
69.
Motion
70.
Poinsot s
71.
Motion of a top on a perfectly rough plane
72.
73.
Determination of the remaining Eulerian angles, and of the Cayley-Klein

parameters the spherical top
Motion of a top on a perfectly smooth plane
74.
Kowalevski
top
164
75.
Impulsive motion
167
two degrees of freedom
of systems with
137
....
motions
of systems with three degrees of freedom
of a body about a fixed point under no forces
............
.............
.........
............
kinematical
representation of
the motion
the
CHAPTER
152
155
159
163
169
144
determination of the Eulerian
143
and
polhode
herpolhode
angle 6
141
VII.
THEORY OF VIBRATIONS.
76.
77.
...........
Vibrations about equilibrium

Normal coordinates
...
177
178
183
79.
Sylvester s theorem on the reality of the roots of the determinantal equation

Solution of the differential equations -the periods
stability
80.
Examples of vibrations about equilibrium
187
81.
Effect of a
78.
new
83.
constraint on the periods of a vibrating system

The stationary character of normal vibrations
Vibrations about steady motion
84.
The
82.
......
185
191
192
193
85.
integration of the equations

Examples of vibrations about steady motion
86.
Vibrations of systems involving moving constraints
-07
208
19;">
203
ix
Content*
CHAPTER
VIII.
NON-HOLONOMIC SYSTEMS. DISSIPATIVE SYSTEMS.

PAGE
SECTION
87.
88.
89.
90.
91.
92.
93.
94.
s equations with undetermined multipliers

manner
Equations of motion referred to axes moving in any
Application to special non-holonomic problems
Vibrations of non-holonomic systems
Lagrange
217
frictional forces
Dissipative systems
Resisting forces which depend on the velocity
;
Rayleigh s dissipation-function
Vibrations of dissipative systems
95.
Impact
96.
Loss of kinetic energy in impact
97.
Examples of impact
.......
......
CHAPTER
IX.
THE PRINCIPLES OF LEAST ACTION AND LEAST CURVATURE.

98.
The
99.
Hamilton
....
trajectories of a
102.
dynamical system
conservative holonomic systems
The principle of Least Action for conservative holonomic systems
Extension of Hamilton s principle to non-conservative dynamical systems
Extension of Hamilton s principle and the principle of Least Action to
103.
Are the stationary
100.
101.
s principle for
non-holonomic systems
104.
105.
..........
Kinetic foci
integrals actual minima?
Representation of the motion of dynamical systems by means of geodesies
The least-curvature principle of Gauss and Hertz
.
.....
106.
Expression of the curvature of a path in terms of generalised coordinates
107.
Appell s equations
Bertrand s theorem
108.
245
245
247
248
249
250
253
254
256
258
260
261
CHAPTER
X.
HAMILTONIAN SYSTEMS AND THEIR INTEGRAL-INVARIANTS.

109.
110.
111.
112.
s form of the equations of motion

Equations arising from the Calculus of Variations
Hamilton
114.
115.
116.
On systems which
117.
118.
119.
relative integral-invariant
263
265
267
Integral-invariants
The variational equations
Integral-invariants of order one

Relative integral-invariants
113.
.....
.........
which
268
269
271
is
possessed by
all
Hamiltonian systems
possess the relative integral-invariant
The expression of integral-invariants

The theorem of Lie and Koenigs
The last multiplier
in
|2pS</
terms of integrals
272
272
274
275
276
Content*
120.
Derivation of an integral from two multipliers

Hainiltonian systems
279
use of a
121.
Application of the last multiplier to
122.
single known integral

Integral-invariants whose order is equal to the order of the system
Reduction of differential equations to the Lagrangian form
123.
124.
a.se
in
which the kinetic energy
is
280
quadratic in the velocities
283
284
285
...
286
CHAPTER XL
THE TRANSFORMATION-THEORY OF DYNAMICS.
125.
Hamilton
126.
127.
Contact-transformations in space of any number of dimensions

The bilinear covariant of a general differential form
128.
The conditions
129.
The conditions
130.
131.
expressions
Poisson s bracket-expressions
The conditions for a contact-transformation expressed by
132.
The sub-groups
133.
Infinitesimal contact-transformations
134.
The
s Characteristic
Function and contact-transformations
288
292
....
for a contact-transformation expressed
by means
of
296
the
297
bilinear covariant
............
.........
........
.............
........
.........
.........
for a contact-transformation in
terms of Lagrange
bracket-
means of Poisson
tions
135.
136.
138.
139.
140.
300
Mathieu transformations and extended point-transforma
new view
of
resulting
dynamics
Helmholtz s reciprocal theorem
Jacobi s theorem on the transformation of a given dynamical system into
another dynamical system

137.
299
bracket-expressions
of
298
CHAPTER
304
304
305
307
......
...
310
Transformations in which the independent variable

New formulation of the integration-problem
.
302
....
Representation of a dynamical problem by a differential form

The Harniltonian function of the transformed equations
301
is
changed
309
310
311
XII.
PROPERTIES OF THE INTEGRALS OF DYNAMICAL SYSTEMS.
.............
.......
............
............
141.
Reduction of the order of a Harniltonian system by use of the energy
142.
143.
Hamilton s partial differential equation

Hamilton s integral as a solution of Hamilton
144.
The connexion
145.
Poisson s theorem
146.
The constancy
147.
Involution-systems
integral
s partial differential equation

of integrals with infinitesimal transformations admitted by
the system
of Lagrange s bracket-expressions
318
320
321
322
Contents
xi
PAGE
SECTION
a dynamical problem when
half the integrals are
known
148.
Solution of
149.
Levi-Civita s theorem
150.
Systems which possess integrals linear in the momenta

Determination of the forces acting on a system for which an integral
151.
323
325
....
.
328
is
331
known
152.
153.
.......
Application to the case of a particle whose equations of motion possess an

integral quadratic in the velocities
General dynamical systems possessing integrals quadratic in the velocities
332
336
335
CHAPTER
XIII.
THE REDUCTION OF THE PROBLEM OF THREE BODIES.

339
154.
Introduction
155.
The
156.
Jacobi s equation
Reduction to the 12th order, by use of the integrals of motion of the centre
157.
............
........
differential equations of the
of gravity
Reduction to the 8th order, by use of the integrals of angular

and elimination of the nodes
159.
Reduction to the 6th order
160.
161.
Alternative reduction of the problem from the 18th to the 6th order
The problem of three bodies in a plane
162.
The
163.
Extension to the problem of n bodies
344
347
348
351
.......
problem of three bodies
343
momentum
158.
restricted
340
342
problem
353
356
356
CHAPTER
XIV.
THE THEOREMS OF BRUNS AND POINCARE.

164.
Bruns theorem
Statement of the theorem
(i)
(ii)
Expression of an integral
358
in
terms of the essential coordinates of
......
....
358
Expression of the integral as a quotient of two real polynomials

Derivation of integrals from the numerator and denominator of
the quotient
361
(vii)
Proof that
366
(viii)
Proof that
does not involve the irrationality

is a function only of the momenta and the
integrals
momentum
371
the problem
(iii)
An
(iv)
Only one irrationality can occur
(v)
(vi)
must involve the momenta
integral
<
of angular
(ix)
(x)
(xi)
in the integral
Proof that
<
is
....
a function of
7",
Z, J/,
Deduction of Brims theorem, for integrals which do not involve

Extension of Brims result to integrals which involve the time
359
360
362
374
t
.
376
378
Contents
xii
165.
Poincare
(i)
theorem
The equations
(iii)
Statement of
Proof that
(iv)
Proof that
(ii)
(v)
4>
4>
.......
of motion of the restricted problem of three bodies

:
Poincare"
is
theorem
not a function of
q%
......
Proof that the existence of a one-valued integral

the result of (iii) in the general case
Removal
(vii)
Deduction of Poincare s theorem
of
is
the restrictions on the coefficients
m>
CHAPTER
381
381
Iff,
cannot involve the variables q,
(vi)
380
382
inconsistent with
,,, 2
383
384
385
XV.
THE GENERAL THEORY OF
ORBITS.
166.
Introduction
386
167.
Periodic solutions
386
168.
Poincare
169.
170.
171.
172.
normal variables
for a periodic orbit
criterion for the discovery of periodic orbits
three particles
of
Lagrange s particles
Stability
The differential equation of the
Lagrange
387
.
390
periodic orbits in the vicinity

normal displacement from an orbit
...........
..........
.......
.
173.
Korteweg s theorem
174.
The index
175.
177.
Characteristic exponents
Properties of the characteristic exponents
Attractive and repellent regions of a field of force
178.
Application of the energy integral to the problem of stability
179.
Application of integral-invariants to investigations of stability
176.
of stability
388
394
395
397
398
400
401
403
406
407
107
CHAPTER
XVI.
INTEGRATION EY TRIGONOMETRIC SERIES.

180.
The need
series
.............
for series
which converge
181.
The
182.
Trigonometric series
Removal of terms of the
183.
184.
185.
for all values of the
regularisation of the problem of three bodies
time
Poincare
410
411
412
first degree from the energy function

Determination of the normal coordinates by a contact-transformation
Transformation to the trigonometric form of If
.
........
413
414
417
419
420
188.
Other types of motion which lead to equations of the same form

Removal of a periodic term from
Removal of further periodic terms from //
189.
Reversion to the original coordinates
423
424
186.
187.
INDEX OF AUTHORS QUOTED

INDEX OF TERMS EMPLOYED
4-2-2
4^5
.
428
CHAPTER
KINEMATICAL PRELIMINARIES
The displacements of rigid
1.
bodies.
The name
Analytical Dynamics is given to that branch of knowledge

in which the motions of material bodies, considered as due to the mutual
interactions of the bodies, are discussed
by the aid of mathematical
analysis.
It is natural to begin this discussion by considering the various possible

types of motion in themselves, leaving out of account for a time the causes
to which the initiation of motion may be ascribed this preliminary enquiry
;
The
object of the present chapter is

of kinematical theorems which will be required in the
constitutes the science of Kinematics.

to establish a
number
rest of the work.
Kinematics
is in itself
an extensive subject,
for a complete account of
which the student
In
dealing exclusively with it, e.g. that of Koenigs (Paris, 1897).
shall confine our attention to theorems which are of utility in the
appli
is referred to treatises
what
follows
we
cations of Kinematics to Dynamics.
We
shall say that a material body is rigid when the mutual distance of
every pair of specified points in it is invariable, so that the body does not
expand or contract or change its shape in any way, although it may change
its
position with reference to surrounding objects.
If a rigid body is moved from one position to another, the

change of
Certain special kinds of
position is called a displacement of the body.
displacement have received specific names; thus,
if
of every point of the
body which
the displacement
called a rotation about the line
is
lies
the position in space
on some straight
line
L;
if
is
unchanged,
the position in
of the body is unchanged, the displacement is called

space of some point
a rotation about the point P; and if the lines
joining the initial and final
of
each
of
the
of
the
are
a
set of parallel
positions
lines
points
body
straight
of length
so that the orientation of the
body in space
unaltered, the
displacement is called a translation parallel to the direction of the lines,
through a distance I.
w. D.
I,
is
Kinematical Preliminaries
[CH.
Elder s theorem on rotations about a point*.
2.
Consider a rigid body, one of whose points is made immoveable by some

attachment suppose that the body is free to turn about this point in any
manner, and let any two possible configurations of the body be taken for
;
We
now
shall
call
to the configuration
through the fixed point,
to
and the configuration Q.

these the configuration
shew that it is possible to bring the body from the configuration
convenience we shall
simply rotating it about some definite line

that a rotation about a point is always equivalent
Q by
i.e.
rotation about a line through the point.
To
was first given by Euler), denote the fixed

be
the
positions, in the configuration P, of two lines
point by 0;
the
fixed
which
are
fixed in the body and move with it let
through
point
OA OB be the positions of the same lines in the configuration Q. Draw
establish this result (which
OA OB
let
AOA
and bisects the angle

perpendicular to the plane
.ACM/: and draw also the plane which is perpendicular to the plane BOB
the plane which
is
and bisects the angle BOB. Let 00 be the line of intersection of these two
if they are coincident, we
planes, supposing them to be not coincident
denote by 00 the line of intersection of the planes OAB and OA B
;
Then
clearly in either case the line
in exactly the same way as

and
say, the angles
AGO
B OG.
it
OG
OA
and
is
related to the lines
OA
OB
and
OA OB
,
that
is to
A OG and
respectively equal to the angles

in such a
is rotated about
system
OB come
00
related to the lines
BOG are
It follows that if the
that the lines
is
OABC
OA
into the positions
and
OB
OC
way
respectively,
is therefore
The line
position unchanged.
unaffected by the displacement in question, and so the displacement can
be represented by a rotation through some angle round 06 which proves
then
will
retain
its
the theorem.
When
a body
is
continuously moving round one of
its
fixed in space, the displacement from its position at time
time
some
At, can,
through the fixed point.
interval Atf
The
line,
indefinitely diminished,
axis of rotation of the body at the time t.
When
a body
is
points, which is
to its position at
by Euler s theorem, be obtained by rotating the body about
definite line
when the
is
limiting position of this

called the instantaneous
is
continuously moving round one of its points, which is fixed, the locus
body is a cone, whose vertex is at the fixed point: the
of the instantaneous axis in the
locus of the instantaneous axis in space is also a cone whose vertex is at the fixed point.
Shew that the actual motion of the body can be obtained by making the former of these
cones (supposed to be rigidly connected with the body)

be fixed in space).
roll
on the latter cone (supposed to

(Poiusot.)
similar proof shews that if any two positions of a plane figure in the
are given, the displacement from one position to the other can be
same plane
Novi Comment. Petrop. xx.
(1776), p. 189,
25.
2-4]
regarded as a rotation about some point in the plane.

the centre of rotation.
When
3
This point
is
called
the body
is regarded as
continuously moving, the small displace
position to the position which succeeds it after an infinitesimal
interval of time can therefore be accomplished by a rotation round a
point
ment from one
this point is called the instantaneous centre of rotation.
lamina moves in any manner in its plane. Prove that the locus at any
Example 1.
instant of points which are at inflexions of their paths is a circle, which touches the loci
in the lamina and in space of the centre of instantaneous rotation.
(Coll. Exam.)
rigid body in two dimensions is subjected successively to two finite

Example 2.
If D 2 be the line joining the centres of displacement, and if
displacements in its plane.
DI be the line which is brought into the position
2 by half the first displacement (i.e.
by rotation through half the angle), and if 3 be the position to which D 2 is brought by
half the second displacement, shew that the centre of the total displacement of the rigid
is the intersection of DI and
3
(Coll. Exam.)
body
3.
Z>
The theorem of Rodrigues and Hamilton*.
Any two successive rotations about a fixed point can be compounded into
a single rotation by means of a theorem, which may be stated as follows
:
Successive rotations about three concurrent lines fixed in space, through twice
of the planes formed by them, restore a body to its original position.
the angles
Draw. Op, Oq, Or per

QOR, ROP, POQ respectively. Then if a body is
rotated through two right angles about Oq, and afterwards through two
right
angles about Or, the position of OP is on the whole unaffected, while Oq is
moved to the position occupied by its image in the line Or; the effect is
therefore the same as that of a rotation round OP through twice the
angle
between the planes PR and PQ, which we may call the angle RPQ. It
For
let
the lines be denoted by OP, OQ, OR.
pendicular to the planes
follows that successive rotations
round OP, OQ,
OR
through twice the angles
RPQ, PQR, QRP,
respectively, are equivalent to successive rotations through

two right angles about the lines Oq, Or, Or, Op, Op, Oq; but the latter
rotations will clearly on the whole produce no displacement; which establishes
the theorem.
4.
The composition of equal and opposite rotations about parallel axes.
A case of special interest is that in which a body is subjected in turn to

two rotations of equal amount in opposite senses about two parallel axes.
In neither displacement is any point of the body displaced in a direction
parallel to the axes, and this is therefore true of the total displacement.
Moreover, if any line be taken in the body in a plane perpendicular to the
*
0. Rodrigues, Journ. de Math. v. (1840), p. 380; Hamilton, Lectures on Quaternions,

the proof here given is due to Burnside, Acta Math. xxv.
(1902).
12
344
[CH.
axes, this line in the first displacement will be turned through an angle
equal to the angle of rotation, and in the second displacement will be turned
back through the same angle so its final position

original position; which evidently can be the case
;
exception,
only
translation.
when the
It follows that
total
displacement
is
be parallel to its
every line without
will
for
equivalent to a simple
two successive equal and opposite rotations about
parallel axes are equivalent to a translation in a direction perpendicular to

the axes; or, in other words, a rotation about
any axis is equivalent to
a rotation through the same angle about any axis parallel to

a simple translation in a direction perpendicular to the axis.
it,
together with
The converse
of this, namely the theorem that a rotation

of a rigid
about
body
any axis, preceded or followed by a translation in a direction
perpendicular to the axis, are together equivalent to a rotation of the body
about a parallel axis, is also true, being essentially the same as the result
stated in
2, that any displacement in a plane can be regarded as a rotation
round some point in the plane. By considering the angle between the
initial and final positions of
any line which is perpendicular to the axis and
moves with the body, we see that the angles of rotation round the two axes
are equal.
5.
Chasles theorem on the most general displacement of a rigid body*.
We
shall
now, consider displacements of a more general character. It is

evident that a free rigid body can be moved from
any one selected con
figuration P in space to any other Q by first moving some selected point of
its position in the
configuration P to its position in the
each of the other points of the body
being moved by a simple
translation parallel to this (so that the body is oriented in the same
way after
the operation as before), and secondly rotating the body about this point into
the configuration Q.
By Euler s theorem, this latter operation can be
performed by simply rotating the body about a line through the point so
the body from
configuration Q,
we
see that the most general displacement

of a rigid body can be obtained by
first translating the body, and then rotating it about a line.
We
shall
now shew
that the line about which the rotation takes

place can
motion of translation is parallel to this line. For let A
be the initial position of any point of the body, and B the position to which
this point is brought by the motion of translation.
Let
be the line
A
to
the
line
round
which
the
rotation
takes
parallel
through
place, and let
be the foot of the perpendicular from B on AK. Then the motion of
be so chosen, that the
AK
translation can evidently be accomplished in two

stages, the first of which
is a translation parallel to the line about which the rotation takes
place,
*
Mozzi, Discorso matematico sopra ilrotamento momentaneo del corpi, Naples, 1763; Cauchy,
Exercices de Hath. n. (Paris, 1827), p. 87; Oeuvres, (2) vn. p. 94; Chasles, Bulletin Unit: des
Sciences (Ferussac), xiv. (1830), p. 321 ; Comptes Rendus de VAcad. xvi.
(1843), p. 1420.
4-6]
bringing the point A to the position K, and the second of which is

a translation perpendicular to the line about which the rotation takes place,
to the position B.
But by 4, the second translation,
bringing the point
with
the
rotation
which
follows
it, are together
together
equivalent to
a simple rotation about a
new
axis parallel to the first one.
If therefore any
point on this axis be taken as base-point, the whole displacement can be

accomplished by a translation of the body parallel to a certain line through
together with a rotation about this line; this establishes the
this point,
theorem.
This combination of a translation and a rotation round a line parallel to

is called a screw
the ratio of the distance of
the direction of translation
It is
translation to the angle of rotation is called the pitch of the screw.
clear that in a screw displacement, the order in which the translation and
rotation take place
is
indifferent.
Halphen s theorem on the composition of two general displacements.

Halphen has shewn* how to determine geometrically the resultant of any two screw6.
displacements as a screw-displacement.
Let
and A 2 denote the axes of the two screws, and
Let BI be the line which
brought to the position
is
-4 ]2
A 12
common
their
by half the
perpendicular.
first
displacement
and rotation through half the angle), and let B2 be the line to
whose position A^ is brought by half the second displacement; let C denote the common
Halphen s result is that the axis of the resultant
perpendicular to the lines BI and B2
(i.e.
half the translation,
screw-displacement
position
For
B2
let
position DI
to DI and
is C,
and
which brings the line
the displacement is twice that
to the
DI and D2 be lines such that half the given displacements will bring A 12 to the
and D2 to the position A 12 respectively, and let
be the common perpendicular
C"
The figure thus obtained, and that which is obtained from it by rotating
right angles about A i2 , evidently coincide whence we have the relations
Intercept
Intercept
Intercept
made on B by A { and C= Intercept made on

made on B2 by A 2 and C= Intercept made on
made on C by BI and B2 = Intercept made on
1
Angle between the planes

Angle between the planes
Angle between B 1 and B 2
AiB B C= Angle
A 2 B2 B2 C= Angle
1
it
through two
= Angle
Z>
C"
by AI and
by A 2 and C
by DI and D 2
C",
between the planes AI D lt

between the planes A 2 D2
,
between
and
DiC
DC
screw about A 1 brings C to the position of

produced, the inter
being brought to the position of the intersection of D^ and C ; and
then the screw about A 2 brings
to the position of C produced, the intersection of
so C is the axis of the resultant
2 and C being brought to the intersection of B 2 and C
It follows that the
section of
and
C"
C"
and 2
screw, and the amount of the translation is twice the intercept made on C by
Also the line BI, which by the first screw is brought to the position
is by the second
brought to a position making the same angle with B2 that B 2 makes with B l ; and therefore
.
Z>j,
Nouvelles Annales de Math.
Mesa, of Math. xix. p. 104 (1889).
(3) i. p.
298 (1882).
The proof given here
is
due to Burnside,
the rotation of the resultant screw
Halphen
is
twice the angle between
[CH.
Bz and B.
This establishes
theorem.
Shew
Example.
by the composition
that any infinitesimal displacement of a rigid body can be obtained

two infinitesimal rotations round lines, and that one of these lines
of
can be arbitrarily chosen.
7.
Analytic representation of a displacement.
We
shall
now
see
how any displacement
of a rigid
body can be represented
analytically.
Let rectangular axes Oxyz be taken, fixed in space these will be supposed
form a right-handed system, i.e. if the axes are so placed that Oz is directed
vertically upwards and Oy is directed to the northern horizon, then Ox will
be directed to the east. Let the displacement considered be equivalent to a
:
to
rotation through an angle
&>
about a line whose direction-angles are
(a, /3, 7),
and which passes through a point
whose coordinates are (a, b, c), together

with a translation through a distance d parallel to this line. The
angle
must be taken with its appropriate sign, the sign being positive when the line
&>
(a, /3,
7) being directed vertically upwards, the rotation from the southern
is round
by the east. Let the point P whose
be brought by the displacement to the position of the
point Q (X, Y, Z); and let the point P be brought by the translation alone
to the position of the point R (, 77, ) then we have
evidently
horizon
to
the
coordinates are
northern
(x, y, z)
= x + d cos a,
77
= y + d cos ft,
%=z
+ d cos 7.
Let
rotation,
be the foot of the perpendicular from R (or Q) on the axis of

and let L be the foot of the perpendicular from Q on KR. Then
we have
X g= projection of
the broken line
RLQ
on the axis Ox,
being understood that projections have their appropriate signs, so that the
XA ), not (XA XB ).
projection of a line AB on the axis of a; is (XB
it
Now
the projection of
(projection of
Ox
is
A K on
the axis Ox)
- a - cos a {( - a) cos a + (T? - b) cos

|
or
and as
axis
KR on the axis
RL =
Ox
(1
cos
co)
KR,
it
/3
+ (f - c) cos 7],
follows that the projection of
RL
on the
is
- (1 - cos tu) [f - a - cos a {( - a) cos a + (17 - b) cos @ +

Moreover, the line
LQ
is
normal to the plane
RKA,
and
its
- c) cos 7}].
direction-cosines
are therefore proportional to the quantities

(
c) cos /3
6) cos 7,
(77
(??
b) cos
a) cos 7
a) cos
/3,
c) cos a,
6-8]
and since the sum of the squares of these three quantities, divided by the
the quantity sirfRAK,
expression {(-a) + 0?- &) + (-c) }, represents
1
and the three
,
follows that the sum of the three squares is equal to
are the projections on the axes of a length +
quantities themselves
sin w, the projection of LQ
measured along the line LQ. Since
2
it
KR
KR
LQ= KR
on the axis Ox
is
therefore
sin
&)
c) cos ft
{(
b) cos 7}.
(r)
On considering a special case, e.g. supposing that the axis of rotation

axis Oz, we see that the upper sign is correct and thus we have
is
the
Z - = - (1 - cos
&>)
{(- ) - cos
cos a cos /3(r)
+
Substituting
for
sin
to
{cos /3
cos
c)
rj,
(1
- cos w)
+ sin to
7 (77.
z,
c)}
6)j.
we have
cos a cos y(z
b)
(cos /3(z
cos
c)
y(y
c)}
&)}.
we have
F = y + d cos /3 - (1 - cos &))
{(y
- cos /3 cos y(z

4-
sin
Z= z + d cos 7 - (1
ft)
cos w) {(z
sin
&)
2
6) sin /3
a.
a)
- c)
(x
(cos a.(y
cos
c)
[cos 7 (#
cos 7 cos
sin 2
b)
cos a (x
cos a.(z
coordinates x, y, z of the original

which define the displacement.
The composition of small
a)}
c)}
7
cos
a)
/3
7 cos /3(y
cos ft(x
These equations give the new coordinates X, Y,
8.
7(
2
a) sin a
{(x
cos a cos (3(y
and
cos a cos
b)
their values in terms of x, y,
X = x + d cos a
Similarly
(- a)
b}}
a)}.
terms of the
in
position of the point and the quantities
rotations.
We shall now apply the last result to the case in which the rotation is
infinitesimal, the axis of rotation passing through the origin and there being
for co, where S-v/r
no motion of translation. We shall write
The equations of the
quantity whose square can be neglected.
now become
= x + (z cos /3 y cos 7) 8-v/r,
is
a small
8>|r
last article
F = y + (x cos
Z z + (y cos
z cos a)
x cos
/3)
S-^r,
8^.
are the equations which we should obtain if we successively (in

a
ty about Ox,
any order) subjected the body to infinitesimal rotations cos
It follows that any small rota
cos /3 8-^r about Oy, and cos 7 8^ about Oz.
But these
tion Sty about
line
OK is
equivalent to successive small rotations Sty cos

.
KOx
[CH.
about Ox, Si/r cos KOy about Oy, and &$ cos KOz about Oz, where Ox,
Oy, Oz
are any three mutually perpendicular lines which intersect
in one
its
.
OK
of
points, 0.
Enters parametric
9.
round a point*.
specification of rotations
The analytic expressions for the translational part of a

displacement are,
as we have seen,
extremely simple; but the expressions for the rotational
part are not so simple, and these will now be further considered.
Suppose
then that a rigid body is rotated
about a line through
through an angle
the origin, whose
are
a, ft, 7.
7, the coordinates
direction-angles
By
(X, Y, Z) of the new position of a point whose original coordinates were
&>
(x, y, z)
are given
(X = x
by the equations
2 sin a
\w
2
(x sin a
cos a cos
+
i
F= y
2 sin |a) (y sin
z cos
ft
=z
2 sin 2
sin 2
G) (
Now
introduce parameters
=cosasina>,
77
ft
cos
77,
= cos /3 sin |
\w {z cos
x cos ft cos
ft
cos
2 sin ^co cos
cos
^&>
y cos
7),
z cos
a),
a)
2 sin i&) cos ô) (x cos 7
x cos 7 cos a
z cos a cos 7)
ft
2 sin -iw cos
/3)
(y cos a
cos
/3).
^, defined by the equations
= cos 7 sin
&>,
i&>,
x=
cos&>;
these parameters evidently satisfy the identical relation
r+7?
+r+ %
and the above equations can be written
=i,
in the form
If therefore the coordinate axes are denoted
by OXYZ, and
if
moveable
axes which originally coincide with these are brought into the
position Oxyz
by the given rotation, the direction-cosines of the two sets of axes with
reference to each other are given by the following scheme
:
Novi Comment. Petrop. xx. (1776),
p. 208,
6 sqq.
8-rlO]
It is readily seen that the
two successive displacements
parameters
(
r,
f *
,
(",
xx
x"=
f",
17",
and (,
77,
corresponding to the resultant of

x ), are given by the equations
x"),
f,
-^f-w -f-
These formulae (which were discovered independently at different times by Gauss,

Rodrigues, Hamilton, and Cayley) really constitute the theorem for the multiplication of
For ^,
?;,
may be regarded as the components of a quaternion*
quaternions.
k
where
+
i,
satisfy the equations
j,
X+i +y
,
i*
=j* = k*=-\, ij=-ji=k, jk=-kj =
and the above formulae are then

X"
+ Z i + jj +
all
t"k
comprehended
X+&+W
i,
ki=-ik=j;
in the single equation
(X
+ & + JJ +
*)
The reader who is acquainted with quaternions will observe that the effect of the
1
where q denotes the
rotation on any vector p is to convert it into the vector qpq
,
quaternion x + &+n} +
&">
The Eulerian
10.
the quaternion
itself is not
the rotational operator.
angles.
various methods of parametrically

practically useful of the
round a fixed
a
of
the
rigid body due to a rotation
displacement
representing
is likewise due to Eulerf: it has the disadvantage of being unsympoint
metrical, but is otherwise very simple and convenient.
The most
be the fixed point round which the rotation takes place, and let
be a right-handed system of rectangular axes _fixed in space. Let
Oxyz be rectangular axes fixed relatively to the body and moving with it,
and Oxyz
and such that before the displacement the two sets of axes
Let
OXYZ
OXYZ
OK
be perpendicular to the plane zOZ,

Let
are coincident in position.
drawn so that if OZ is directed to the vertical and the projection of Oz
is directed to the east.
to OZ is directed to the south, then
OK
perpendicular
by 6,
-^r,
respectively: these are
as the three Eulerian angles defining the position of the axes Oxyz
Denote the angles zOZ, YOK,
known
with reference to the axes
yOK
<,
OXYZ.
In order to find the direction-cosines of Ox, Oy, Oz, with respect to OX,
that these are equal to the projections on Ox, Oy, Oz, respectively,
Now this unit length has projections
of a unit length measured along OX.
where
OL is the intersection of the
and
sin
cos $ along OL
along OK,
we observe
<
planes
XOY and Z0z\
but a length
cos<
along
OL
has projections cos sin 6

is the intersection of the
<f>
$ cos 6 along OM, where OM

cos
along OM has projections
planes xOy and ZOz; and a length cos
cos
cos 6 sin -fy along Oy also, a length
cos $ cos 6 cos A|T along Ox and
sin
cos i/r
Ox and
sin
sin<sim/r along
along OK has projections
along Oz and cos
<
<
(f>
* This
quaternion will have its tensor equal to unity.
t Novi Comment. Petrop. xx. (1776), p. 189.
Kinemat ical Preliminaries
10
along Oy.
Hence
OX
cos
d>

sin
-vjr
alonsf
Ox
r
cos ty along Oy,
sin 6
cos
are
cos
on Ox, Oy, Oz respectively of the
finally the projections
unit length measured on
[CH.
along Oz.
(/>
Proceeding in this way, we obtain for the direction-cosines of the two sets of
axes OX YZ and Oxyz with respect to each other the
following scheme
:
Connexion of the Eulerian angles with the parameters
11.
The
^.
77,
between the Eulerian angles 6,

and the parameters
\|r
f
9
be
obtained
the
schemes
of directionf/j
%
may
by comparing
cosines which have been
in
and
10
9
given
they may however be obtained
relations
<,
>
directly as follows
Let OX YZ and Oxyz be the fixed axes and the axes derived from these
by the rotation w round a line OR, whose direction-angles are (a, ft, 7).
Draw
a sphere of unit radius with the point

as centre, so that planes
intersect
the
in
passing through
sphere
great circles, and lines intersect the
in
Then
in
the
sphere
points.
spherical triangle RZz, the sides are 7, 7, 0,
and the angle at
is
whence we have the relation
o>
sn
RZ
is
let v
Then the
<
6 about the pole of Zz, and

into an arc
7 sin
denote the angle RZY, so RZz = ITT

v.
brought to the position Rz by successive rotations
Moreover,
arc
sin
-v/r
about z
making an angle
but the
v
first
<
about Z,
of these transforms
v with Zz, at
or ITT
RZ
the
\-rr
second rotation transforms this into an arc making the same angle ^TT
v
with Zz, but passing through z and the third rotation transforms it into an
(f>
</>
arc
to
making an angle
TT
RzZ, or
TT
^TT
RZz, or
+ ty
TT
with Zz, at
(^TT
<j>
or
v),
TT
z.
But
or \tr
v
this angle
-f
</>
so
must be equal
we have
10-12]
RZX
Hence, since in the spherical triangle

angle at
Z is
-|TT
the sides are
+ $), we have
(TT
cos a = sin 7 sin \
v or
for sin
and the
$).
7 from the equation already found,
cos a sin
\w =
sin
sin
or
^6
sin
\Q
sin |
Similarly from the spherical triangle

cos
and again eliminating
a, 7, ^TT,
-v^
(-v/r
Substituting
11
ft
RZY
= sin 7 cos
ft
$),
(i/r
<).
we have
0),
(i/r
we have
sin 7,
cos
|(^
this gives
sin ^tu
or
?;
= sin cos (i|r

= sini#cosi(^
0),
<).
Moreover, since we have shewn that in the spherical triangle RZz the
and the angles are (rr ^r
we have
(?r
i/r
sides are 7, 7, 0,
<),
<),
&>,
the relations
= COS-i0COS-J(-V/r + 0),
=
+
7 cos ^0 sin %
COS i ft)
and
sin
cos
|&>
(-v/r
x = cos 1
= cos ^
or
The four parameters

angles
6,
<f>,
g,
77,
sin l
(-v/r
f
77
0).
are Aus expressed in terms of the Eulerian
sin 1
sin i(-^
-</>),
= sin ^0 cos |(i|r
1
f = cos
=
^ cos ^
(i^ +
cos Y (^ +
^>),
sin
(/>),
0).
7%e connexion of rotations with homographies ;j
Consider
0),
ty by the relations
(
12.
<),
+
+
cos |(\/r
now a
sphere, on the surface of which
the Cayley- Klein parameters.
we shall call 8} are

Let these figures be stereographically projected on a plane (e.g. by taking the
highest point of the sphere as vertex of projection and the tangent-plane at the lowest
Now let the
point of the sphere as the plane) we shall call the projected figures P.
sphere be rotated through a definite angle about some axis through its centre, so that the
let the figures in their new
be
figures on its surface are shifted to new positions
any
figures (which
drawn.
positions
the stereographic projections of the figures 8 (with the same vertex and
Then corresponding to the rotation of the
plane of projection as before) be called P
sphere, which changes 8 to S we have a transformation in the plane, which changes the
called
and
let
P into the figures P We shall now examine this transformation more closely.
If one of the figures P is a circle in the plane, we know that the
corresponding figure S
figures
must be a
circle traced
on the sphere, since by stereographic projection a circle is changed

S must also be a circle; and hence P must -also be a circle.
into a circle: therefore
Thus we
sphere,
see that the transformations of the plane, which correspond to rotations

of the
be such as to change any circle in the plane into another circle in the
plane.
must
12
may be shewn*
It
in the following
z=x+y J
Let
plane
that any transformation of this kind
where x and y are the rectangular coordinates of any point in the

complex variable z.
= x + i/ \/ - 1, where x and y refer to the point into which the point (x, y)
1,
circles into circles^,
Then any
a, 6,
c,
one-to-one transformation of the plane, which
be defined by
may
an equation of the type
_
=
are (real or complex} constants;
kind combined with a reflexion in one of
represented analytically
changed by the transformation.
changes all
where
may be
so that to this point there corresponds a definite value of the
Similarly let z
is
way
[CH.
or else by
a transformation of
this latter
the axes of coordinates.
transformation represented by an equation of the type

_ az + b
,
cz
+d
a homographic transformation, or homography. It appears therefore that homo

graphies in a plane correspond to rotations of a solid body about a fixed point, in such a
way that if two homographies correspond respectively to two rotations, the homography
compounded of these corresponds to the rotation compounded of the two rotations J.
is called
We
shall
now
see
how the connexion between
represented analytically.
Let us replace the parameters
77,
f,
rotations
and homographies may be
^ by new parameters
a,
/3,
y,
8,
defined by the
equations
*"~
/3--y
94
~ q9--
_/3 + y
l~
9,
Z*
2*
so that they are connected with the Eulerian angles

.
d,
_a + 8
X~~a
Z
>
(j>,
^ by the
equations
y=ism-.e
= cos-.e 2
These
"
"
Cayley-Klein
parameters clearly satisfy the relation

aS
and replacing the quantities

their values in terms of
of
a, /3, y,
a, /3,
8 the following
-y,
77,
8,
scheme
^
we have
Cf. L. K.
scheme of direction-cosines given
in
for the values of the direction-cosines in
9 by
terms
/3y=
in the
Ford, An introduction to the theory of automorphic functions (London, 1915).

straight line is to be regarded as a particular kind of circle.
J Klein, Math. Ann.
ix. (1875), p.
183; Cayley, Math. Ann. xv. (1879),
p.
238.
12, 13]
It may readily be shewn that the parameters

resultant of two successive displacements (a /3 y
equations
a"
y"
=a a+y
= ya +8y
(a",
t3,
ft",
and
= aft
+ft8
8"=yft
+ 88
ft"
y",
(a,
13
corresponding to the
8) are given by the
8")
y,
ft,
These equations shew that the transformation

a"z+ff
y"z
is
8"
the result of performing in succession the two substitutions
J=
,
+ ft
z=
,
an d
yz + 8
*,,
az
+ ft
^j
yz+8
and the connexion between rotations and homographic transformations
is
thus evident
analytically.
One advantage
(|,
r),
f,
x),
of the Cayley-Klein parameters, as compared with the parameters

that they retain some of the simplicity of the quaternion calculus, while
is
using the *J
of ordinary algebra instead of the
is
fixed axes
by f1}
and
<,
OXYZ)
**
before the motion, and (di, fa
tan \6{ by
/,
"**
"fj
Example
2.
If
after the motion.
Denoting
^ tan d
| t
1
008*0- sin $ 6
e-^s
from the equations
2
the quantities X-p,
Xi 2 are formed, and
2
and
the
symbols
2
quantities X^, X^,
1
,
these quantities are regarded as umbral

2
x 2 2 x l x 2 are replaced by -Y+iX,
shew
that
the equations obtained are
respectively,
XX
Z,
shew that
^=f
Y+iX,
k of Hamilton s quaternions.
j,
Let (6,
1.
\^) denote the Eulerian angles.
Suppose that a point in space
carried about with the axes Oxyz has the vectorial angles (0 lt
to the
t ) (referred
Example
which
i,
if
%i
-y +ix, y + ix, z,
- Y+ iX= a 2 - y + ix] + 2apz + /3 2 (y + ix\
Y+ iX = y 2 (-y + ix] + 2y8z + 8 2 (y + ix),
Z = ay (-y + ix + (a8 + py)z + p8(y + w),
f
and that these are the three equations connecting the coordinates (X,
referred to the axes
.Example
3.
OXYZ with
its
coordinates
(x, y, z)
Y, Z) of a point
referred to the axes Oxyz.
If
y + ix: y + ix
and
z=\\
Y+iX Y+iX Z=\

:
\1
:1
:
shew that
and \^=
y
13.
We
Vectors.
now proceed
displacement by
to
consider
the
essential
features
involved in
the
simple translation of a rigid body.
The operation
of translation in
itself,
considered apart from the body
translated, evidently possesses the following properties
14
1.
by any one of the equal and parallel

which have a given length (viz. the distance of the translation)
It can be specified completely
lines of space
and given direction

furnishes
2.
[OH. i
all
If
its
joining
to the sum
(viz.
since such a line
ACDE...KB
be a broken line
the direction of the translation)
the data which describe the operation.
AB
be one of these
lines,
and
AB is
equivalent
KB.
number of
operations
extremities, then the operation represented by

of the operations represented by
C, CD, DE,
These properties
and 2 are possessed by a large
and quantities other than the operation of translation an operation or

quantity which possesses them is called a vector quantity.
;
By 2,
a vector
AB is equivalent
to the
sum
of three vectors
A K, KL, LB,
respectively parallel to three given rectangular axes, and forming a broken

and B. These three vectors are called the com
line joining the points
ponents of the vector AB along the given axes.

cos a, Zcos/3,
be the length and
(a, ft,
7)
the lengths of the component vectors are clearly

cosy), being in fact the projections of A B on the axes.
the direction-angles of
(I
If
AB,
single vector which

called their resultant.
is
equivalent to any
number
of given vectors
is
is conceived as varying in dependence on a

parameter (e.g. the
between
the
vectors
difference
the
to
time),
corresponding
any two values of
the parameter is also a vector, and hence the rate of change of the vector
If a vector
with respect to the parameter
also a vector,
is
whose components are the

This
rates of change of the corresponding components.

of the vector with respect to the parameter.
14.
Velocity
and acceleration ;
Consider now a body which
is
is
called the flux
their vectorial character.
being continuously translated (though not
necessarily always in the same direction) without any change of orientation.

Its total translation to any time t is a vector quantity, and hence the rate at
which this changes with the time, i.e. its time-flux, is also a vector quantity,
if x, y, z are the coordinates referred
is called the velocity of the body
to fixed axes of any point fixed in the body and moving with it, then the com
which
ponents of the velocity referred to these axes are the rates of change of x, y, z,
i.e. are x, y, z (where dots denote differentiations with respect to the time
t}.
Similarly the rate of change of the velocity is again a vector, whose
components are x, y, z (two dots indicating second derivatives with respect
this vector is called the acceleration of the body.
to the time)
;
It is clear that if
and Q are two moving
points, the vector
which
represents the translation (or velocity, or acceleration) of Q is the sum of the

vector which represents the translation (or velocity, or acceleration, as the
and the vector which represents the translation (or velocity,
case may be) of
or acceleration) of
relative to P,
moves with P, and whose
i.e.
of
referred to axes whose origin
directions are invariable.
13-15]
15.
Angular
velocity
its vectorial
15
character.
Consider next a body which is rotating continuously about a line. Let 6

denote the angle turned through at any time t then Q represents the
speed
of turning at the time t.
If from any point on the line round which the
rotation takes place a segment whose length represents 6 is measured
along
the line, this segment will evidently furnish a complete
specification of the
nature of the rotation at the instant t, or (as it is generally
expressed) of
:
the angular velocity of the body.

The direction in which the segment is
measured from the base-point is to be connected with the sense of rotation
by the usual convention, namely that when the segment

upwards the rotation from the southern horizon
is
directed vertically
to the northern is
round by
the east.
An
and
angular velocity
direction.
Now
is
therefore represented
a body one
if
by a
line of definite length
of whose
is fixed
8,
by
points
experiences a small rotation Sty round any line OK, this displacement is
equivalent to successive small rotations Sty cos a round Ox, Sty cos ft round
Oy, and Sty cos 7 round Oz, where Ox, Oy, Oz are any three mutually
and (a, fi, 7) are the direction-angles
perpendicular lines passing through
of
with reference to Oxyz. From this it is clear that we can
regard an
OK
angular velocity represented by a length ty measured on

to
velocities
OK
as equivalent
ty cos j3, ty cos 7,
angular
represented by lengths ty cos a,
measured along Ox, Oy, Oz, respectively.
But this is essentially the fundamental property of vectors, and can be
expressed by the statement that angular velocities can be resolved and
compounded according to the vectorial law.

It must be observed however that an
angular velocity does not fulfil all
the conditions which enter into the definition of a vector, for an
angular
velocity about one line is not equivalent to an angular velocity of the same
magnitude about a parallel line. Angular velocity must therefore be regarded
as a vector which is localised
along a definite line.
Example. A right circular cone of semi-vertical angle $ rolls without sliding on a plane.
To find its instantaneous axis of rotation, and to determine its
about this
angular
velocity
axis in terms of the angular

velocity of the line of contact in the plane.
Since
all
points of the generator which
at rest (for there
the cone.
is
no
sliding), this
is in
contact with the plane are

instantaneously
is the instantaneous axis of rotation of
generator
Let
denote the angular velocity of the cone about this
generator, and let 6
denote the angular velocity of the line of contact in the
Then the motion of the
plane.
axis of the cone can be represented
6 round the normal to the
by an angular
o>
velocity
plane,
and the whole motion of the cone
compounded of this together with a rotation

that the component of angular
velocity of the cone
is
round the axis of the cone. It follows

about a line through the vertex of the cone
perpendicular to the axis is 6 cos /3
but
this must equal the resolved
part of co in this direction, which is wsin/3.
We have
;
therefore
w=6
which
is
the required relation between
o>
cot
and
/3,
6.
16
[CH.
Determination of the components of angular velocity of a system in

16.
terms of the Eulerian angles, and of the symmetrical parameters.
The position at any time of a rigid body which is continuously moving
about a fixed point
is
most conveniently described by taking two
sets
OXYZ
are fixed in space, while the

of rectangular axes, of which one set
the
other set Oxyz are fixed relatively to the body, and move with it
then
the
three
Eulerian
of
the
^,
specified by
body being
angles 0,
position
;
</>,
which define the position of the axes Oxyz relatively to the axes OXYZ.
We shall now determine the components, along the moving axes, of the
angular velocity of the body at any instant.
Let OK denote the line of intersection of the planes XOY and xOy the
;
angular velocity of the system is evidently compounded of angular velocities

about OZ, and -^r about Oz. Of these, the first can be
6 about OK,
law by angular velocities 6 sin i/r about Ox
replaced according to the vectorial
<j>
and 6 cos
about
about Oy
ty
Ox,
sin
sin
<f>
and the second can be resolved into

cos 6 about Oz.
about Oy, and
-fy
<
So
<j>
sin 6 cos
finally
\|r
if
denote the components of angular velocity of the body about the

axes Ox, Oy, Oz, respectively, we have
G>
I} eo 2 ,
o>
= 6 sin -Jr
sin 6 cos
= va cos y
sm a sin y,
i
a
1a) = y (p COS
ilr,
<j>
&>
ft) 3
From
t>
<6
c7.
-|-
we can
these expressions
at once deduce the values of
in terms of the symmetrical parameters
??,
%, of
for
twj,
o>
o>
we have
cf>=-
Similarly
we have
==
"4?
and we have
cos d
~H
^,
v2
$^.
>
+%
Substituting these values in the equation
The values
principle of
&>
a
.
= ty +
cos
#,
we have
and
can be at once obtained from this by the
2
and
thus
we have the components of angular velocity
symmetry
of o^
<w
given by the equations
16, 17]
17.
Time-flux of a vector whose components relative
17
to
moving axes are
given.
Suppose now that a vector quantity is specified by its components f, 77,

any instant t with reference to the instantaneous position of a right-handed
system of axes Oxyz which are themselves in motion and let it be required
to find the vector which represents the rate of
change of the given vector.
at
Let
&) 3 denote the

2
components of the angular velocity of the
resolved
the
instantaneous position of the axes Ox, Oy, Oz
system Oxyz,
along
themselves.
&>
&>!,
The time-flux of the given vector is the (vector) sum of the time-fluxes
of the components
taken separately. But if we consider the vector
?/,
it is increased in
length to g + %dt in the infinitesimal interval of time dt,
and at the same time is turned by the motion of the axes, so that (owing to
,
the angular velocity round Oy) it is displaced through an angle co 2 dt from its
position in the original plane zOx, in the direction away from Oz, and also
(owing to the angular velocity round Oz) it is displaced through an angle
(o s dt from its
position in the original plane xOy, towards Oy. The coordinates
of its extremity at the end of the interval of time dt, referred to the
positions
of the axes at the commencement of the interval dt, are therefore
(neglecting
infinitesimals of order higher than the first)
and
so the
components of the vector which represents the time-flux of

<"
are
Similarly the components of the vectors which represent the time-fluxes

of the vectors ?; and
are respectively
&>
and
3 ?7,
o>
Adding
This
we have
these,
vector in the
finally the components of the time-flux of the given
form
result
can
be
immediately applied
acceleration of a point whose coordinates

reference to axes moving with an
to
find
(x, y, z) at time
the velocity and

t are
given with
angular velocity whose components along
the axes themselves at time
are
(a>
&>
&>
3 ).
For substituting in the above formulae, we see that the components of

the velocity are
x
w. D.
yw-s
+ zw
zo>!
+ xa)
18
[CH.
Now
applying the same formulae to the case in which the vector whose
time-flux is sought is the velocity, we have the components of the accelera
tion of the point in the form
--
-yCtt
(x
yco 3
(^
#o>
+ za}
-I-
2)
(t>
(y
ZW-L
(#
2/
ft>
ya>i)
+ xo) ) + o)
s
(z
xw z +
2/&>i),
In the case in which the motion takes place in a plane, which we may
take as the plane Oxy, there will be only two coordinates (x, y}, and only one
component of angular velocity, namely 6, where 6 is the angle made by the
moving axes with their positions at some fixed epoch the components of
;
velocity are therefore (putting
z,
each equal to zero in the above
expressions)
+ xO,
y6 and y
and the components of acceleration are

x - 2y6
-yO- x6
and y + 2x6
+ x 0-
yfc.
Example. Prove that in the general case of motion of a rigid body there is at each
instant one definite point at a finite distance which regarded as invariably connected with
the body has no acceleration at the instant, provided the axis of the body s screwing
motion be not instantaneously stationary
in direction.
18.
Special resolutions of the velocity
The
results obtained
and
in the last article
(Coll.
Exam.)
acceleration.
enable us to obtain formulae,
use, relating to the components of the velocity

moving point in various special directions.
which are frequently of

acceleration of a
(i)
Velocity
and
and
acceleration in polar coordinates.
Let the position of a point be defined by
its
polar coordinates
r,
6,
0,
connected with the coordinates (X, Y, Z) of the point referred to fixed

rectangular axes OXYZ by the equations
X = r sin 6 cos
Y = r sin 6 sin
Z = r cos 6
<,
(j),
let it be required to determine the components of velocity and

acceleration of the point in the direction of the radius vector r, in the
direction which is perpendicular to r and lies in the plane containing r and-
and
OZ (this
plane
is
generally called the meridian plane), and in the direction
perpendicular to the meridian plane these three directions are frequently

described as the directions of r increasing, 6 increasing, and
increasing,
;
(f>
19
17, 18]
Take a line through the origin 0, parallel to the direction of

respectively.
a line through 0, parallel to the
increasing, as a moving axis Ox and take
of r
line
and
a
axis
direction of
parallel to the direction
Oy,
increasing, as
the
determine
which
The three Eulerian angles
increasing as axis Oz.
;
<p
OXYZ
are
reference to the fixed axes
position of the moving axes Oxyz with
of angular velocity of the system Oxyz,
(6, d), 0); so ( 16) the components
the axes Ox, Oy, Oz themselves, are
resolved
along
eoj
The coordinates
sin
of the
6,
&>
0,
<w
(p
cos
6.
point, referred to the
moving
moving
axes, are
17 the components of velocity of the point resolved

by
(0, JO, r);
to
the
moving axes are
parallel
and
so
sin
rd,
r,
6,
r<j>
and the components of acceleration in the directions of 6 increasing,

of
17) are
increasing, and r increasing (again using the formulae
2
-T-
]
!
(r0)
r<
sin 6 cos
r0, or r0
.
-r-
dt
(rd) sin 0)
and
fcf>
2r0
or
cos
rd] sin
+ r66 cos 0,
sin
r sin
d
-j-
at
2
2
(r sin
0,
6d>),
rd) 2 sin 6.
r&*
motion of the point is in a plane, we can take the initial line in this
in these formulae
and the quantities denoted by r and
If the
plane as axis Oz,
become ordinary polar coordinates in the plane since (p is now zero, the
components of velocity and acceleration in the directions of r increasing and
;
increasing are
(r,
and
(r
Velocity
(ii)
and
r0
2
,
rd),
r0
2f 0).
acceleration in cylindrical coordinates.
Consider now
a point whose position is defined by its cylindrical

coordinates z, p, $, connected with the coordinates (X, Y, Z) of the point
referred to fixed rectangular axes
by the equations
OXYZ
X = p cos
^p
d),
si
n
$>
Z = z\
be required to find the components of the velocity and acceleration

of the point in the direction parallel to the axis of z, in the direction of the
line drawn from the axis of z to the point, perpendicular to the axis of z, and
and
let it
These three directions are

in the direction perpendicular to these two lines.
z
the
direction of p increasing,
the
called
direction
increasing,
of
generally
and the direction of
<p
increasing
and the coordinate
is
called the azimuth

22
20
[CH.
The components
increasing.
along the axes Oxyz
of angular velocity of the system Oxyz, resolved

themselves, are clearly
COj
==
==
0,
d>2
0,
61)3
(D,
and the coordinates of the moving point, referred to the moving axes, are
It follows by
17 that the components of velocity of the point in
(p, 0, 2).
these directions are
and the components of acceleration are
Velocity
(iii)
and acceleration
in arc-coordinates.
Another application of the formulae of 17 is to the determination of the

components of velocity and acceleration of a point which is moving in any
way in space, resolved along the tangent, principal normal, and binormal to
its
path.
first the case of a particle moving in a plane

and take lines
to
a
fixed
the
0,
point
parallel respectively
through
tangent and inward
normal to the path, as moving axes Ox and Oy. These axes are rotating
is the angle made by the
round
with angular velocity (ft, where
tangent
Consider
<
some
If v denotes the velocity of

fixed line in the plane.
at
and p the radius of
the
of
described
time
arc
the path
the point, s
t,
curvature of the path at the point, we have
to the path with
ds
v
dt
ds
dtp
and the angular velocity of the axes can therefore be written in the form
(v,
the same axes are
(v, v
v\
.-}
pJ
Since
it
v/p.
Since the components of the velocity parallel to the moving axes are
17 that the components of the acceleration parallel to
0), it follows from
_dv
ds dv
dv
dt
dt ds
ds
follows that the acceleration of the
dv
tangent to
normal
its
is
path
ia
-j-
moving point
in the direction of the
and the acceleration in the direction of the inward
Now the velocity of a moving point is determined by the knowledge of

two consecutive positions of the moving point, and the acceleration is therefore
determined by the knowledge of three consecutive positions so even if the
path of the point is not plane, it can for the purpose of determining its
;
acceleration at
any instant be regarded as moving in the osculating plane of
21
18]
contains three consecutive positions of the point.

path, since this plane
Hence the components of acceleration of the point, in the directions of the
its
and binormal
tangent, principal normal,
dv
v2
ds
to its
path, are
Acceleration along the radius and tangent.
(iv)
acceleration of a point which is in motion in a plane may be expressed

in the following form*; let r be the radius vector to the point from a fixed
from the origin on the tangent to the
origin in the plane, p the perpendicular
at
time t, p the radius of curvature of
described
of
the
arc
s
the
path
path,
at the point, and v or s the velocity of the point at time t and let
the
The
path
h denote the product pv.
to the
the acceleration of the point
along the radius vector
into components
tangent
Then
to the
origin
can be resolved
and
~r
along the
path.
For the acceleration can be resolved into components vdv/ds along the
now a vector F directed outwards along
tangent and v jp along the normal
into
vectors
be
resolved
can
vector
radius
the
Fp/r along the inward normal
2
a
v
so
vector
the
and Fdr/ds along
/p along the inward normal can be
tangent,
2
rv 2 dr
rv
T- along the tangent.
inwards along the radius vector and
resolved into
pp ds
pp
The acceleration is therefore equivalent to components
rv 2 dr
dv
v -p H
-r- along the tangent,
on ds
ds
pp
z
rv 2
- inwards along the radius vector.
and
pp
The
latter
component
1 dv 2
2 ds
v2
which establishes Siacci

Example
tion of
1.
and the former can be written
pp
Id (v p*)
z
dp
ds
2p
ds
h dh
ds
s result.
Determine the meridian, normal, and transverse components of the accelera
a point moving on
is
the surface of the anchor-ring
be the centre of the anchor-ring and C the centre

and let
on which P lies. The polar coordinates of C relative to
and the polar coordinates of P relative to C are (a, 0, $) ; so the components
be the point (6, $),

Let
of the meridian cross-section
are
(c,
<p),
of acceleration of
and
relative to
are
transverse
ctj)
2
c<p
outwards from the
axis,
i.e.
and
*
Dae
2
c<
sin 6 along the normal,
c0 cos 6 along the meridian.
to Siacci, Atti della R. Ace. di Torino, XIY. p. 750.
22
The components
P relative
of acceleration of
fad -
2
a<j>
sin Q cos 6
C are
along the meridian,
(sin
-j-
sin 6 dt
to
[CH.
transverse.
.
<)
sin 2
normal.
t<j>
Thus
finally the
components
a6
2
a<j)
and
of acceleration of
(c + a sin
ad
cd)
6}
sin 6
-.
+ sm
6
-.
dt
sin 6
space are
normal,
c<f>
2
(sin 6
-y-
P in
along the meridian,
cos 6
transverse,
.
d>)
Example 2. If the tangential and normal components of the acceleration of a point

moving in a plane are constant, shew that the point describes a logarithmic spiral.
In this case
v -r-=a, where a
ds
so
i)
a constant,
=as.
Also
is
c,
where
a constant,
c is
= Cp,
where
C is
so
or
s=C-j-r, where
A and B
Example
are constants
To find
3.
is
(f>
the angle
made by
the tangent with a fixed
line.
we have
Integrating this equation,
where
a constant,
and
this is the intrinsic equation of the logarithmic spiral.
the acceleration
of a point which describes a logarithmic spiral with
constant angular velocity about the pole.
By
and
Siacci s theorem, the
-=5
2
so the
ds
components of acceleration are
along the tangent: but
if to is
53
PP
along the radius vector
the constant angular velocity, we have h =
u>r
components of acceleration are

co
2
?"
p
Since -
7*
7"
and
CiT
-5-
and
2o)
2 3
r dr
-j-
are constant in the spin

spiral,
;
as
jr
we
see that each of these
components of
acceleration varies directly as the radius vector.
MISCELLANEOUS EXAMPLES.
1.
If the instantaneous axis of rotation of a
in the body,
shew that
it is
also fixed in space,
body moveable about a fixed point is

the motion is a rotation round a
i.e.
fixed
fixed
axis.
A point is referred to rectangular axes Ox, Oy revolving about the origin with
2.
2 2
x (distance),
angular velocity to; if there be an acceleration to x=a, y = 0, of amount ?i
shew that the path relative to the axes can be constructed by taking (i) a point
<o
23
i]
x = ri2a/(n 2 -l), (ii) a uniform circular motion with angular velocity (n-l)co about this,
but in the opposite
and (iii) a uniform circular motion with angular velocity (n +
this
last.
about
(Coll. Exam.)
sense,
l)a>,
3.
The velocity of a point moving in a plane is the resultant of a velocity v along the
Prove that the
radius vector to a fixed point and a velocity v parallel to a fixed line.
corresponding accelerations are
dv
vv
6 being the angle that the radius vector
4.
point moves
a fixed line in the plane, where
ponent
in a plane,
and
dv
^ + -cos0,
s+
and
makes with the
vv
fixed direction.
is referred to Cartesian axes
Exam.)
making angles
Shew
are given functions of the time.
a, /3
(Coll.
a, /3
that the
with
com
velocities of the point are
x - xd cot
(/3
- a) - y$ cosec
and obtain expressions
for the
- a),
(/3
component
y +y$
- a) + xd
cot
cosec
accelerations.
(/3
- a),
(Coll.
Exam.)
5.
A point is moving in a plane Q is the logarithm of the ratio of its distances from
also 2k is the distance
is the angle between them
two fixed points in the plane, and
between the fixed points. Shew that the velocity of the point is
:
<
cosh 6
- cos
(Coll.
Exam.)
(f)
two different descriptions of a curve by a moving point, the product of the

that the
places in the two descriptions is constant, shew
accelerations at corresponding places in the two descriptions are as the squares of the
the normal to the curve, in
velocities, and that their directions make equal angles with
(J- von Vieth.)
opposite senses.
6.
If in
velocities at corresponding
7.
A point is moving in a parabola of latus rectum 4,
and when
its
distance from the
the velocity is v shew that its acceleration is compounded of accelerations

and N, along the radius vector and normal respectively, where
focus
is r,
R=v ^
dr
N=^,% dr
%- (v*r).
(Coll.
Shew
o>
is
\|/-
and
yyo>2
9.
0,
The
+ (.coi +
velocity of a point
with a fixed
line.
is
2.o>
Exam.)
2r
that if the axes of x and y rotate with angular velocities coi,

the
angle between them, the component accelerations of the point
^
to the axes are
cosec \^,
x xo)i 2 (x&i + 2# 1 ) cot
+
2)
8.
and
1)
made up
cosec
(?/o>2
\|f
respectively,
(x,
y) parallel
2/o>
+ (yw2 +
of
2?/o>2)
cot
\/r.
(Coll.
Exam.)
u, v in directions
components
making angles
Prove that the components /, / in these directions of the accelera
tion of the point will be given
by
f= u u6 cot x~ v $ cosec x,
f = v + ud cosec x + cot x,
v<}>
X being the inclination of the two
directions.
Being given that the lines joining a moving point to two fixed points are r, s in length
in inclination to the line joining the two fixed points, determine the acceleration
6,
of the point in terms of
the rates of increase of 0, (f).
(Coll. Exam.)
and

24
|_
CIL
If A, B, C be three fixed points, and the component velocities of a moving point

the
directions PA, PB, PC be u, v, and w shew that the accelerations in the same
along
10.
directions are
/
cosAPB
cosAPC\
and two similar expressions.
(Coll.
Exam.)
11.
The movement of a plane lamina is given by the angular velocity and the com
resolved along axes Ox, Oy traced on the lamina.
ponent velocities u, v of the origin
Find the component velocities of any point (x, y] of the lamina. Shew that the equations
a>
tan
-T-
1
(
at
"-
)=
o>
one being the locus of those points which are passing

cusps on their curve loci in space and the other being the locus of the centres of curvature
of the envelopes in space of all straight lines of the lamina.
(Coll. Exam.)
represent circular loci on the lamina
Shew
12.
that
when a point
describes a space-curve,
acceleration can be resolved
its
two components, of which one acts along the radius vector from the projection of a
fixed point on the osculating plane, and the other along the tangent and that these are
into
respectively
L
P3
^
P
TdT + T*qdq
fi~ds
p*
~d7
where p is the radius of curvature, q the distance of the fixed point from its projection on
the osculating plane, r and p are the distances of this projection from the moving point
and the tangent, T is an arbitrary function (equal to the product of p and the velocity) and
s is the arc.
(Siacci.)
A circle, a straight line, and a point lie in one plane, and the position of the point
13.
determined by the lengths t of its tangent to the circle and p of its perpendicular to
the line.
Prove that, if the velocity of the point is made up of components u, v in the
is
directions of these lengths
and
if
their
mutual inclination be
6,
the component accelerations
be
will
uvcosd/t,
v+uv/t.
(Coll.
Exam.)
A particle moves in a circular arc. If r, r are the distances of the particle at P

14.
from the extremities A, B of & fixed chord, shew that the accelerations along AP, BP are
respectively
dv
-j- H
dt
where
v,
vv
rr
dv
cos a), and

ivr-f
rH
x
dt
v are the velocities in the directions of
r,
vv
rr
(r
^
and a
r cos
is
a),
the angle
APB.
point describes a semicircle under accelerations directed to the extremities of a

diameter, which are at any point inversely as the radii vectores r, r to the extremities of
the diameter. Shew that the accelerations are
4a4 F 2
where a
is
the radius of the circle and
V the
4a4
72
velocity of the point parallel to the diameter.

(Coll.
Exam.)
25
i]
The motion
15.
of a rigid
C and
in
body
two dimensions
is
defined by the velocity
(u, v} of
Determine the coordinates relative to

angular velocity
points
the point 7 of zero velocity, and shew that the direction of motion of any other point
one of
its
its
<o.
of
is
perpendicular to PI.
Find the coordinates of the point
P in terms of its
J of zero
and express the acceleration of

(Coll. Exam.)
acceleration,
coordinates relative to J.
A point on a plane is moving with constant velocity V relative to it, the plane at
16.
Shew
the same time turning round a fixed axis perpendicular to it with angular velocity
that the path of the point is given by the equation
o>.
and 6 being referred
to fixed axes,
and a being the shortest distance
of the point
axis of rotation.
from the
Exam.)
(Coll.
where
The acceleration of a moving point Q is represented at any instant by
17.
Prove that the
a fixed point and a describes uniformly a circle whose centre is
is a fixed point and p describes
velocity of Q at any instant is represented by Op, where
<oa,
is
a circle uniformly
18.
and determine the path described by Q.

(Camb. Math. Tripos, Part
222
point moves along the curve of intersection of the ellipsoid ~g
the hyperboloid of one sheet -^

d
~"
+O
?/
A
>.
+C
- *~~
"2^T
=1 and
the curve meets the hyperboloid

of two sheets ,2
J1
a -/i
where h
is
constant.
+ b T--
l~
21
~~?
2
-=
1902.)
+W+- =
an d
the P oint where
its velocit y at
>
I,
*s
/i
Prove that the resolved part of the acceleration of the point along
the normal to the ellipsoid
is
h 2 abc (n
X)
-7=..
/ri
(Coll.
Exam.)
19.
A rigid body is rolling without sliding on a plane, and at any instant its angular
2 along the tangent to the lines of curvature at the point
velocity has components coj,
of contact, and 3 along the normal shew that the point of the body which is at the point
of contact has component accelerations
o>
o>
where RI, R% are the principal

of contact.
radii of curvature of the surface of the
body at the point

(Coll. Exam.)
CHAPTER
II
THE EQUATIONS OF MOTION

The ideas of rest and motion.
In the previous chapter we have
frequently used the terms
19.
"
"
fixed
and
"
"
as applied to
systems.
moving
kinematical considerations, it is
significance of these words; all
So long
as
we
are occupied with

purely
to
enter into the ultimate
unnecessary
that is meant is, that we consider the
"
displacement of the moving systems, so far as

with respect to the systems which are called
"
it
affects their
configuration
"fixed,"
question of what
leaving on one side the
meant by absolute fixity."

When however we come to consider the motion
"
is
of bodies as due
no longer be disregarded.
In popular language the word
is
generally used of
to specific
causes, this question can
"fixed"
terrestrial
to
denote
invariable
objects
position relative to the surface of the earth at
the place considered.
But the earth is rotating on its axis, and at the same

time revolving round the Sun, while the Sun in
turn, accompanied by all
the planets, is
moving with a large velocity along some not very accurately
known direction in space. It seems hopeless therefore to
attempt to find
rest."
anything which can be really considered to be
"at
In the nineteenth century
was supposed -that the aether of

space (the
and magnetic actions) was (apart from small
vibratory motions) stagnant, and so was capable of providing a basis for
absolute fixity. But this doctrine has been subverted
by the modern Principle
which
asserts
that
even
in
the
of Relativity*,
domain of electromagnetic
it
vehicle of light
and of
phenomena
impossible to distinguish absolute rest from a state of uniform

common to all the members of a system.
it is
electric
translatory motion
Accordingly in dynamics, although
we always imply
be
called,
that there
is
some
when we speak
of the motion of bodies
set of axes, or frame
with reference to which the motion
of reference as
it
may
and
which we apply the conventional word fixed,"
it
must
not
be
yet
supposed
that absolute fixity has
thereby been discovered. When we are
to
is
regarded as taking place,
"
considering
Cf.
or
Whittaker
Conway
History of the Theories of Aether and Electricity, ch.

Relativity (London, 1915).
s
xii.
(London, 1910)
27
The Equations of Motion
19, 20]
the motion of terrestrial bodies at some place on the earth s surface, we shall
take the frame of reference to be fixed with reference to the earth, and it is
then found that the laws which will presently be given are sufficient to
explain the phenomena with a sufficient degree of accuracy in other words,
the earth s motion does not exercise a sufficient disturbing influence to make
it necessary to allow for its effects in the majority of cases of the motion of
;
terrestrial bodies.
It is also necessary to consider the
meaning
to
be attached to the word
which in the previous chapter stood merely for any parameter

time,"
varying continuously with the configuration of the systems considered. The
Principle of Relativity reveals the great difficulties that attend any attempt
to elucidate the idea of time
in particular, it is by no means easy to define
simultaneity, i.e. to explain what is meant by saying that two events at
"
at the same time."

However, a system of
time-measurement which is intelligible from the point of view of ordinary
instrumental work, and which is sufficient for our present purpose, is the
following we suppose that the angle through which the earth has rotated on
its axis (measured with reference to the fixed stars, whose small motions we
can for this purpose neglect), in the interval between two events, measures
the time elapsed between the events in question. This angular measure can
different points of space
"
happen
be converted into the ordinary measure in terms of mean solar hours, minutes,
and seconds at the rate of 360 degrees to 24 x 365/366 hours.
20.
The laws which determine motion*.
Considering
now
the motion of terrestrial objects, and taking the earth as
the frame of reference, it is natural to begin by investigating the motion of a

very small material body, or particle as we shall call it, when moving in vacuo
and entirely unconnected with surrounding objects. The paths described by
such a particle under various circumstances of projection may be observed,

and the methods of the preceding chapter enable us, from the knowledge
thus acquired, to calculate the acceleration of the particle at any point of any
It is found that for all the paths the acceleration
particular observed path.
is
always directed vertically downwards. This

and is generally denoted by the letter g its
in Great Britain, about 981 centimetres per second per second.
of constant amount,
acceleration
amount
is,
is
known
and
is
as gravity,
The knowledge of this experimental fact is theoretically sufficient to

enable us to calculate the path of any free terrestrial particle in vacuo, when
the circumstances of its projection are known the actual calculation will not
:
belongs more properly to a later chapter.

motion which is next in simplicity is that of two
be given here, as
The
case of
it
particles
which are connected together by an extremely light inextensible thread, and

*
The laws
of motion are due to
Newton
Principia, p. 12 (ed. 1687).
^8
move
are free to
[OH.
So long as the thread is

moves with the acceleration gravity, just as if the other
But when the thread is taut, the two particles influence
in vacuo at the earth s surface.
slack, each particle
were not present.
We can now as before observe the path of one of the

and hence calculate the acceleration by which at any instant its
each other s motion.

particles,
motion
is
being modified. We thereby arrive at the experimental fact, that

can be represented at any instant by the resultant of two
this acceleration
of which one represents the acceleration g and the other
vectors,
along the instantaneous position
is directed
the thread.
of
The
influence of one particle on the motion of the other consists there

fore in superposing on the acceleration due to gravity another acceleration,
which acts along the line joining the particles and which is compounded
with gravity according to the vectorial law of composition of accelerations.

Denoting the particles by A and B, we can at any instant calculate, from the
observed paths, the magnitudes of the accelerations/! and/2 thus exerted by
B on A and by A on B respectively and this calculation immediately yields
;
the result that the ratio of f\ to /2 does not vary throughout the motion.
On
the
motions
which
result
from
various
modes
of
investigating
projection, at
various temperatures etc., we are led to the conclusion that this ratio is an
invariable physical constant of the pair of bodies
On
and B*.
consideration of the motion of more complex systems
it
is
found
that the experimental laws just stated can be generalised so as to form

a complete basis for all dynamics, whether terrestrial or cosmic.
This
generalised statement is as follows: If any set of mutually connected particles
are in motion, the acceleration with which any one particle moves is the
resultant of the acceleration with which it would move if perfectly free, and
accelerations directed along the lines joining it to the other particles which
constrain its motion.
Moreover, to the several particles A, B, C, ..., numbers
can
be
due to the
fÂ)
assigned, such that the acceleration along
m m
n>
AB
c>
on
influence of
in the ratio
on
is
to
the acceleration along
m^-.m^.
The ratios of
BA
these
due to
numbers
the influence
mA mB
,
...
of
are
invariable physical constants of the particles.
The evidence
for the truth of this
statement
agreement of the calculations based on
it,
is
to be found in the universal
such as those given later in this
book, with the results of observation.

It will
be noticed that only the ratios of the numbers m^, m$, me, ... are
it is convenient to take some definite
particle A as
determined by the law
a standard, calling it the unit of mass, and then to call the numbers
B me,
?7ic/m A ... the masses of the other particles
,
m^jm^,
The ratio is in i act, equal to the ratio of the weight of B to the weight of A the ratio of
the weights of two terrestrial bodies, as observed at the same place on the earth s surface, is a
perfectly definite quantity, and does not vary with the place of observation.
;
29
20, 21]
The mass
of the
compound particle formed by uniting two
or
more
particles
sum
of the masses of the separate particles.

is found to be equal to the
mass of a finite
Owing to this additive property of mass, we can speak of the
to take as our unit of
body of any size or shape and it will be convenient
known as the
mass the mass of the
part of a certain piece of platinum
this unit will be called a gramme, and the number
standard
;
y^th
kilogramme;
mass of any other body to this unit mass

representing the ratio of the
called the mass of the body in grammes.
21.
is
Force.
have seen that in every case of the interaction of two particles A and
acceleration
B, the mutual influence consists of an acceleration fA on A and an
BA respec
and
AB
directed
vectors
on B, these accelerations being
along
We
fii
and being inversely proportional to the masses m A and m B It follows

that the vector quantity m A fA is equal to the vector quantity m s fB but has
The vector m A fA is called the force exerted by the
the reverse direction.
A and similarly the vector m B fB is called the force
particle B on the particle
exerted by the particle A on the particle B.
.
tively,
With
this terminology, the
mutual action of a connected
law of the
on each other
system of particles can be stated in the form the forces exerted
This
is often
and
are
connected
opposite.
equal
particles
by every pair of
:
called the
Law
of Action and Reaction.
If the various forces which act on a particle A as a result of its connexion

with other particles are compounded according to the vectorial law, the
resultant force gives the total influence exerted by them on the particle A
this force divided by
A is the acceleration induced in A by the other
and the resultant of this acceleration and the acceleration which the
free (due to such causes as gravitation) is
particle A would have if entirely
the actual acceleration with which the particle A moves.
particles
is induced in
In general, if an acceleration represented by a vector
due to
called
vector
the
a particle of mass
the/orce*
mfis
by any agency,
this cause acting on the particle and the resultant of all the forces due to
It follows
various agencies is called the total force acting on the particle.
axes
of the
fixed
to
the
are
if
that
rectangular
components parallel
(X, Y, Z)
are
the
com
and
at
the
on
total force acting
(x, y, z)
any instant,
particle
at
that
described
is
which
its
being
path
ponents of the acceleration with
instant, then
we have the equations
mx = X,
Two
my =
Y,
mz =
other terms which are frequently used
Z.
may
conveniently be defined
at this point.
*
Force
is
the vis matrix of
Newton
Principia,
i.
def. 8.
30
[OH.
The product
of the number which represents the magnitude of the com

of
a
ponent
given force perpendicular to a given line L and the number
which represents the perpendicular distance of the line of action of the force
from the line
is
called the
moment
of the force about the line L.
If the three components (X, Y, Z) of the force

acting on a single free
particle are given functions of the coordinates (x, y, z) of the particle, they
are said to define & field offorce.
22.
Work.
Consider
now any system
of particles,
whose motion
is
either quite free or
by given connexions between the particles, or constraints due to

other particles which are not regarded as forming part of the system.
Let m
be the mass of any one of the particles, whose coordinates referred to fixed
restricted
rectangular axes in any selected configuration of the system are (x, y, z) and
let (X, Y, Z} be the components,
parallel to the axes, of the total force
;
acting on the particle in this configuration.
Let (x + Sx, y + By, z + 8z) be the coordinates of any point very near to
the point (x, y, z), such that the displacement of the particle
from one
to
the
other
does
not
violate
of
the
constraints
is
point
(for instance, if
any
move on a given surface, the two

on the surface). Then the quantity
constrained to
X&K +
points must both be situated
1% + ZSz
m
called the work* done on the particle

by the forces acting on
infinitesimal displacement from the position (x, y, z) to the position
is
(x
Bx,
y+
Sy, z
it
in the
8z).
This expression can evidently be interpreted physically as being the

product of the distance through which the particle is displaced and the com
ponent of the force (X, Y, Z) along the direction of this displacement.
Since forces obey the vectorial law of composition, the sum of the com
ponents in a given direction of any number of forces acting together on a
particle is equal to the component in this direction of their resultant and
hence the work done by a force which acts on a particle in a given displace
ment is equal to the sum of the quantities of work done in the same displacement
:
by any
set
offorces into which
Suppose now
this
force can be resolved.
that in the course of a motion of the system, the particle
gradually displaced from any position (which we can

to some other position at a finite distance from the
is
its final position).

*
The work done on
Newton defined the Actio Agentis
along the direction of motion

p. 25 (ed. 1687).
it is
call its initial position)

first
(which we can call

which act on
the particle by the forces
as the product of the velocity into the component of force

evidently the time-flux of the work done. Cf. Principia, i.
21-23]
31
during this finite displacement is defined to be the sum of the quantities

work done in the successive infinitesimal displacements by which we can
regard the finite displacement as achieved. The work done in a finite dis
it
of
is
placement
therefore represented
where the integration

the arc
is
by the integral
dx
v dy
dz\
as
(LS
(is /
taken between the
initial
and
final positions
along
described in space by the particle during displacement.
These definitions can now be extended to the whole set of particles which
form the system considered the system being initially in any given con
;
we
consider any mode of displacing the various particles of the

system which is not inconsistent with the connexions and constraints; the
sum of the quantities of work performed on all the particles of the system in
figuration,
the displacement is called the total work done on the system in the displace
ment by the forces which act on it.
23.
Forces which do no work.
There are certain classes of forces which frequently occur in dynamical

systems, and which are characterised by the feature that during the motion
they do no work on the system.
Among
1.
these
The
may
be mentioned
reactions of fixed smooth surfaces:
that the reaction
is
normal to the
the term smooth implies

in each infinitesimal
and therefore
surface,
displacement the point of application of the reaction is displaced in a direction

perpendicular to the reaction, so that no work is done.
The
reactions of fixed perfectly rough surfaces: the term perfectly

rough implies that the motion of any body in contact with the surface is one
of pure rolling without sliding, and therefore the point of application of the
2.
reaction
is
(to the first order of small
infinitesimal displacement, so that no
quantities) not displaced in each

is done.
work
The mutual
reaction of two particles which are rigidly connected

y z^) and (#2 y2 z2 ) are the coordinates of the particles,
and (X, Y, Z) are the components of the force exerted by the first particle on
the second, so that
Y,
Z) are the components of the force exerted
3.
together: for if (x1}
lt
(X,
by the second
particle on the
an arbitrary displacement
X (8x
But
since the distance

Cj
j/
first,
the total work done by these forces in
is
Sx,)
+ Y (Sy
\ v
Si/i)
3 */
between the particles

\2
/.
\2
_]_
+ Z (8z
is
&2-A
/
*
invariable,
\2\
we have
32
and since the
have
force acts in the direction of the line

joining the particles,
X:Y:Z=(x 2
Combining the
[CH.
last
a?,)
(y.2
- y,)
(z2
n
we
- z,).
two equations, we have
X (8x
and therefore no work
&O +
is
F(S//2
- SyJ + Z (8z - 82,) =

2
0,
done in the aggregate by the mutual forces between
the particles.
rigid body is regarded from the dynamical point of view as an

of
particles, so connected together that their mutual distances are
aggregate
invariable.
It follows from 3 that the reactions between the particles which
4.
are called into play in order that this condition may be satisfied (or molecular
forces as they are called, to distinguish them from external forces such as
gravity) do, in the aggregate, no
The
5.
work
in
any displacement of the body.
reactions at a fixed pivot about which a body of the system can
turn, or at a fixed hinge, or at a joint between two bodies of the system, are
similarly seen to belong to the category of forces which do no work.
In estimating the total work done by the forces acting on a dynamical

system in any displacement of the system, we can therefore neglect all forces
of the above-mentioned types.
24.
The coordinates of a dynamical system.
Any
material system
constituted of a
is
regarded from the dynamical point of view as
number
of particles, subject to interconnexions and con

a rigid body being regarded as a collection of
kinds
of
various
straints
;
particles,
which are kept at invariable distances from each other by means
of suitable internal reactions.
When
the constitution of such a system (i.e. the shape, size, and mass of
the various parts of which it is composed, and the constraints which act on
given, its configuration at any time can be specified in terms of a
of quantities which vary when the configuration is altered,
which will be called the coordinates of the system; thus, the position of a
them)
certain
and
is
number
single free particle in space is completely defined by its three rectangular

the position of
(x, y, z) with reference to some fixed set of axes
coordinates
a single particle which is constrained to move in a fixed narrow tube, which has
the form of a twisted curve in space, is completely specified by one coordinate,
namely the distance s measured along the arc of the tube to the particle from
the position of a rigid
fixed point in the tube which is taken as origin
is fixed, is completely determined by three co
of
whose
one
points
body,
of
10 the position of
ordinates, namely the three Eulerian angles 6,
-v/r
some
<,
two particles which are connected by a taut inextensible string can be defined
by five coordinates, namely the three rectangular coordinates of one of the
23-25]
33
and two of the direction-cosines of the string (since when these five
quantities are known, the position of the second particle is uniquely deter
mined) and so on.
particles
State the
Example.
number
any instant of a
smooth surface.
configuration at
a given fixed
of independent coordinates required to specify the

body which is constrained to move in contact with
rigid
We
shall generally denote by n the number of coordinates

required to
the
configuration of a system, and shall suppose the systems con
specify
sidered to be such that n is finite. The coordinates will
generally be denoted
If the system contains moving constraints

... q n
(e.g. if it consists
of a particle which is constrained to be in contact with a surface which in
turn is made to rotate with constant angular velocity round a fixed axis),
by q l} q 2
it
<?i>
may be
qn
<?2,
>
necessary to specify the time t in addition to the coordinates

in order to define completely a
configuration of the system.
The quantities q 1} g.2

to the coordinates q l} q2
...
q n are frequently called the velocities corresponding
qn
flexible string, free to move in space, is an

example of a dynamical system
excluded by the limitation that n is to be finite; for the configuration of the
string cannot be expressed in terms of a finite number of parameters.
heavy
which
is
25.
Holonomic and non-holonomic systems.
It is
now necessary
to call attention to a distinction
between two kinds
of dynamical systems, which is of great importance in the

analytical discussion
of their motion this distinction may be illustrated
a
by
simple example.
we consider the motion of a sphere of given radius, which is constrained

move in contact with a given fixed plane, which we can take as the plane
If
to
of xy, the configuration of the sphere at
any instant is completely specified

namely the two rectangular coordinates (x, y) of the
centre of the sphere and the three Eulerian
angles 0,
10, which
fy of
by
five coordinates,
<f>,
The sphere can take

specify the orientation of the sphere about its centre.
up any position whatever, so long as it is in contact with the
the five
coordinates
If
(x, y, 6,
<f>,
now the plane
by the coordinates
-v/r)
is
plane
can therefore have any
arbitrary values.
smooth, the displacement from any position, defined

to any adjacent
i/r),
position, defined by the
9
+
4+
B0,
By,
80,
B-^r), where Bac,
(x, y, 6, $,
coordinates (x + 8x, y+
By,
8^
are arbitrary independent infinitesimal quantities, is a
possible displacement,
i.e. the
sphere can perform it without violating the constraints of the system.
But if the plane is perfectly rough, this is no longer the case when 8x,
By, 86,
8$, 8-^r are arbitrary; for now the condition that the displacement of the
point of contact is zero (to the first order of small quantities) must be
satisfied, and this implies that the quantities 8x, By, 80,
8^ are no
must be such
longer independent, but are mutually connected (in fact,
B<f>,
"f
B<f>,
B<f>,
they
w. D.
34
[CH.
as to satisfy two non-integrable linear equations) so that in the case of the

sphere on the perfectly rough plane, a displacement represented by arbitrary
;
infinitesimal changes in
coordinates
the
is
not necessarily a possible
dis
placement.
dynamical system for which a displacement represented by arbitrary

infinitesimal changes in the coordinates is in general a possible displacement
(as in the case of the sphere on the smooth plane) is said to be holonomic
when
this condition is not satisfied (as in the case of the sphere

plane) the system is said to be non-holonomic.
on the rough
fyn) are arbitrary infinitesimal increments of the coordinates

in a dynamical system, these will define a possible displacement if the system
is holonomic, while for non-holonomic systems a certain number, say m, of
If
(&7i>
&/ 2
equations must be satisfied between them in order that they may correspond
The number (n m) is called the number of
to a possible displacement.
Holonomic systems are therefore charac
of
the
system.
degrees of freedom
by the
terised
number
fact that the
number
of degrees of freedom
is
equal to the
of independent coordinates required to specify the configuration
of the system.
Lagrange s form of the equations of motion of a holonomic system*.

We shall now consider the motion of a holonomic system with n degrees
of freedom. Let (q lt q 2 ... q n ) be the coordinates which specify the con
at the time t.
figuration of the system
26.
Let
(xi,
2/t,
nii
Zi)
of the system, and let

typify the mass of one of the particles
its coordinates, referred to some fixed set of rectangular axes.
be
These coordinates of individual particles are (from our knowledge of the

constitution of the system)
the system, and possibly of
known
t
also
functions of the coordinates q 1} q2
...
q n of
be
dependence
expressed by the
let this
equations
Let (Xi, Yi, Zi) be the components of the total force (external and
then the equations of motion of this
t
molecular) acting on the particle
are
particle
mi
ar.i
it
niiyi
=Y
m^ = Z
i}
Multiply these equations by
*
first
/i
?fc
*b
dq r
dg r
dq r
Lagrange, Mecanique Analytique (1788), Seconde Partie, Section

suggested in one of his earlier papers, Miscell. Taurin. n. (1760).
iv.
The equations were
25, 26]
respectively,
add them, and sum
for all
35
the particles of the system.
We
thus have
5-
2,Wi
&
4. ;;
Vi 5
g +
a
dfi
xi
d
y
h Zi
^
^\
?(y
9/i
]**2,[JLf
\
i
dqr J
dqr
I-
T^
^ 4. A
7
8<
ifl 5
dqr
t-
dq r J
where the symbol S denotes summation over all the particles of the system
be either an integration (if the particles are united into rigid bodies)
this can
or a
summation over a
discrete aggregate of particles.
But we have
a
7i
^J^fZL^^^/
<-<
_ Xi
..
SO
..
Xi ^
^T"
dq r
dq r
_d_(.Xi 3jci\_
~dt( dqJ
dq
~_ d
dt
i. a*A
Xii
Xi
^/a
dt\d
/
Xii
\d qi dqr
dqj
dq 2 dqr
and therefore we have
Now
the quantity
i2*n(#+#+#)
sum
masses of the particles of the system, each

this is called the Kinetic
half
the
multiplied by
square of its velocity
of
the constitution of the
of
the
From
our
Energy
knowledge
system*.
represents the
of the
system, the kinetic energy can be calculated f as a function of

<?i,
we
shall
denote
it
T(q lt
and
shall
qn, q\,
<?a,
<J
qn ,t;
by
suppose that I
is
q2
...
qn q lt
known
q.2 ,
...
function of
qn
its
t),
arguments.
Since
The mass of a particle multiplied by the square of its velocity was called the vis viva by
Leibnitz (Acta erud., 1695).
t The methods of performing this calculation for rigid bodies are given in Chapter V.
32
*f
(,~r
^r
36
and
and
iji
Z{
[CH.
are likewise linear functions of q l} q 2 ... q n we see that T is a

do not involve the
...
q n if the functions /,
,
quadratic function of q ly q
2>
</>,
time explicitly (as is generally the case if there are no moving constraints
in the system), the quantities x, y, z are homogeneous linear functions of
q l} q2
...
q n and then
>
From
T is
a homogeneous quadratic function of q lt q 2
...
qn
it follows that the kinetic energy of a system is essentially

therefore a positive definite quadratic form in glt q 2 ... n and so satisfies
the conditions that its discriminant and the principal minors of every order of its
the definition
positive;
is
<j
discriminant are positive.
We
have thus derived from the equations of motion the equation
ftT\
dt(dqj
dr_y( Y dA+V
+i d-^+7
dq ^
dqr~i(
dqr
"
>
-+i\
9g
and the expression on the left-hand side of this equation does not involve the
individual particles of the system, except in so far as they contribute to the
We
have now to see
the right-hand side of the equation

can also be brought to a form in which the individuality of the separate
kinetic energy T.
particles
is
if
lost.
For this purpose, consider that displacement of the system in which the
coordinate q r is changed to q r + Sq r while the coordinates
,
q 1 ,q.2,...q,-i,qr+i,---qn
and the time
(so far as this is required for the specification of the system) are
Since the system is holonomic, this can be effected without
In this displacement, the coordinates of the
violating the constraints.
unaltered.
particle
m; are changed to
and therefore the
total
work done
act on the particles of the system
in the displacement
by
the forces which
all
is
*~\
Now
-*
f~\
ogv
09V
"!.
*-\
d<
of the forces which act on the system, there are several kinds which
Among these are, as was seen in 23,
do no work.
1.
The molecular
forces
which act between the particles of the rigid
bodies contained in the system
2. The pressures of connecting-rods of invariable length, the reactions

at fixed pivots, and the tensions of taut inextensible strings
:
The
reactions of any fixed smooth surfaces or curves with which

bodies of the system are constrained to remain in contact or of perfectly
rough surfaces, so far as these can enter into holonomic systems
3.
37
26]
The
4.
reactions of
any smooth surfaces or curves with which bodies
of the system are constrained to remain in contact, when these surfaces or

curves are forced to move in some prescribed manner for the displacement
;
made on the
supposition that t, so far as it is required for

the specification of the system, is not varied, i.e. that such surfaces or curves
so that this case reduces to the
are not moved during the displacement
considered above
is
preceding.
The
forces acting
on the system, other than these which do no work, are

It follows that the quantity
called the external forces.
k
--4.r V 8(
=
9
^
/<
dq
is
4-p
JL i
ZAI
d ^*
-=
oq r
dqr
the work done by the external forces in the displacement which corresponds
change of q r to qr + $qr the other coordinates being unaltered. This is
to a
a quantity which (from our knowledge of the constitution of the system, and
of the forces at work) is a known function of q lt q. ... qn t; we shall denote
,
2>
it
by
Qr (qi,
We
q2
qn t)8q r
have therefore
fdT\ _3T__
~n
dt(dqj
dq r
^"
This equation is true for all values of r from 1 to n inclusive we thus

have n ordinary differential equations of the second order, in which qi,q2 ,...q n
;
are the dependent variables
and
t is
the independent variable
as the
number
of differential equations is equal to the number of dependent variables, the

equations are theoretically sufficient to determine the motion when the
circumstances are given.

be thus stated
initial
We
have now arrived at a result which may
Let
T denote
the kinetic energy of a
+ Qz$qa +
Qi&?i
denote the
(&?!,
8q
work done by
...
8q n ), so that
the
T,
dynamical system, and

...
external forces
lt
...
let
+ Qn $q n
in
o,re,
an arbitrary displacement
from our knowledge of
constitution of the system, known functions of q lt q2 ... q n q lt q2 ... q n

then the equations which determine the motion of the system may be written
,
the
,
t;
dT
dT\
These are known as Lagranges equations of motion. It will be observed

that the unknown reactions (e.g. of the constraints) do not enter into these
equations.
The determination of these reactions forms a separate branch of

* so we can
is known as Kineto-statics
say that in Lagranges
mechanics, which
equations the kineto-statical relations

*
Cf.
of
the
Heun, Deutsche Math. Ver.
problem are altogether eliminated.
ix.
(Heft 2) (1900), p.
1.
38
27.
Conservative forces
[CH.
the Kinetic Potential.
have the property that the work done by the forces

of the field in a displacement of a dynamical system from one
configuration
to another depends only on the initial and final
of
the system,
configurations
Certain
fields of force
being the same whatever be the sequence of infinitesimal displacements by

which the finite displacement is effected.
a conspicuous example of a field of force of this character the work done
motion of a particle of mass m from one position at a height h to
another position at a height k above the earth s surface is mg(k-k), and this does not
depend in any way on the path by which the particle is moved from one position to the
Gravity
by gravity
is
in the
other.
Fields of force of this type are said to be conservative.
Let the configuration of any dynamical system be specified

by n
coordinates q l} q 2 ... q n
Choose some configuration of the system, say
that for which
.
q,-=a r
(r=
1, 2, ... n),
as a standard configuration then if the external forces

acting on the system
are conservative, the work done by these forces in a
displacement of the
;
system from the configuration
(q 1} q2 ... q n ) to the standard configuration is a

definite function of q lt q^, ... q n , not depending on the mode of
displacement.
Let this function be denoted by V(q } 2 ... q n ); it is called the Potential
,
<?
Energy* of the system in the configuration (q lt q2 ... q n ). In this case the

work done by the external forces in an arbitrary displacement
,
(%,
is
Sq 2
...
Bqn )
evidently equal to the infinitesimal decrease in the function V, corresponding

i.e. is
equal to the quantity
to the displacement,
~~
9F
5~
~~
<7i
dql
Lagrange
dV.
____
we introduce a new
then Lagrange
function
3q r
dqr
of the variables q lt q 2
d /dL\
UT-
dL =
5
0,
dqr
The Potential-function was introduced by Lagrange

is
...
q n q lt
,
...
qn
t,
equations can be written
dt \dq r j
Potential
L = T-V,
-7i
name
oq n
equations of motion therefore take the form
defined by the equation
dq n
dt \dqr )
If
8F,
#2
s~~
dq 2
due
to
Green
(1828).
(r=
1. 2, ... n).
in 1773 (Oeuvres, vi. p. 335).
The
27, 28]
The function L
is
called the Kinetic Potential, or
39
Lagrangian function
completely specifies, so far as dynamical investigations

are concerned, a holonomic system for which the forces are conservative.
this single function
28.
The
We
shall
form of Lagranges
now shew how the second
explicit
equations.
derivatives of the coordinates with
respect to the time can be found explicitly from
Lagrange
equations.
Let the configuration of the dynamical system considered be specified by

coordinates q1} q 2 ... q n
we shall suppose that the configuration can be
of these coordinates alone, without t, so that
in
terms
completely specified
;
the kinetic energy of the system is a homogeneous quadratic function of

As was seen in
26, this is always the case when the
qlt q2 ,...qn
constraints are independent of the time, but not in general when the
.
constraints have forced motions (as for instance in the case of a particle
constrained to move on a wire which is made to rotate in a given way).
Suppose then that the kinetic energy
T= |
where a H
is
2,a k iqk qi,
k=n=i
a {k and where the coefficients

,
<%
are
known
functions of
of motion for the system are
= l,
2, ...n),
(r=l,
2, ...w),
(r
dt
dq
dq
or
*vJi^V.""*i?iika
w
n
^vj~l
v
c*
c
m
ZOnqt+2, 2,
r \qiq
I"/
"
or
;=im=i
s=i
where the symbol
1,
which
rt
= Qr,
= 1,
i
(r
2, ... n),
called a Christoffel s symbol*, denotes the
is
expression
1 /8$;r
9$mr
2 V3gm
\dqm
9^
dq
_ 5$;m \
9g r /
dq
These equations, being linear in the accelerations, can be solved

In fact, let D denote the determinant
quantities qs
$12
$13
$21
$22
$23
$31
$32
U-ll
It
for the
was introduced by
Christoffel,
theory of quadratic differential forms.
Journal filr Math. LXX. (1869), and
is
of importance in the
40
and
let
rs
be the minor of a rs in this determinant.
of the above system by
lv
A 2v
A nv
...
[CH.
Multiply the n equations
respectively,
and add them:
re-
membering that the quantity 2

has the value
when
A rv ar8
equal to
s is
v,
is
s is different
from
v,
and
we have
when
zero
^~|
|~
This equation is true for all values of v from 1 to n inclusive; and

these n equations, in which q\, q 2 ... n are
q
given explicitly as functions of
,
q l} q2 ... q
of motion.
,
n>
q lt q 2
...
qn
can be regarded as replacing Lagrange
Motion of a system which
29.
is
equations
constrained to rotate uniformly round an
axis.
In many dynamical systems, some

part of the system is compelled by an
external agency to revolve with constant
angular velocity w round a given
fixed axis the motion of a bead on a wire which is made to rotate in this
;
way
is
a simple example.
direct application of
is
holonomic; but
it is
is,
as
we have
seen,
no objection to the
equations to such cases, provided the system

often more convenient to use a theorem which we
s
and which reduces the consideration of systems of this kind

that of systems in which no forced rotation about the
given axis takes
shall
to
now
There
Lagrange
obtain,
place.
Suppose that, independently of the prescribed motion round the axis, the
system has n degrees of freedom, so that if the given axis is taken as axis of
z, and any plane through this axis and turning with the
prescribed angular
is measured, the
velocity is taken as the plane from which the azimuth
</>
cylindrical coordinates of
terms of n coordinates q lt
any
q2
particle
. . .
q n these expressions not involving the time t.

in the actual motion be T, and if the
,
Then if the kinetic energy of the system

work done by the external forces in an
be Qi&qi
+ QSq,+
...
+ Qn &qn
of the system can be expressed in
arbitrary infinitesimal displacement
where Q lt Q 2
...,
Qn
will
be supposed to
depend only on the coordinates q lt q2 ..., q n and if the kinetic energy of

the system when the forced angular
velocity is replaced by zero be denoted
T
we
have
lt
by
,
Now
2?wr2 will be a function of q lt q 2 ..., q n which is

determined by our knowledge of the constitution of the
system denote it by
W. The quantity 2rar2 will also be a known function of 1} 2 ..., ni
q
q
q
the quantity
28-30]
41
linear in q 1} q2 ..., qn it will be zero if, when co is zero, the

qi, .., qn being
motion of every particle has no component in the direction of
increasing
while if n is equal to unity, so that there is only one coordinate q, it will be
.
</>
the perfect differential with respect to t of a function of q

cases of most frequent occurrence, and we shall include
2
suming that ^m?^
dY
of the form -T-
is
<
where
is
these are the two
them both by
as-
a given function of the
CLT>
coordinates q lt q2
We
...,
qn
have therefore
and the Lagrangian equations
d /dT\
dT
dY\
can be written in the form
dY\
dT,\
dT,
dW
(
dt
These equations shew

motion
the
is
the
dqr
were
,,..,.
dqr
that, subject to the
assumption already mentioned,

angular velocity were zero, and
2
2
contain an additional term
a)
In this
same as if
the potential energy
rj
to
the prescribed
^mr
way, by modifying the potential energy, we are enabled to pass from a

system which is constrained to rotate about the given axis to a system for
which
this rotation does not take

The term centrifugal forces is
place.
sometimes used of the imaginary forces introduced in this way to represent
the effect of the enforced rotation.
The Lagrangian equations for quasi-coordinates.

In the form of Lagrange s equations given in
26, the variables are n
coordinates q 1} q2 ..., q n and the time t; the knowledge of these
quantities,
30.
together with a knowledge of the constitution of the system, is sufficient to

determine the position of any particle in any configuration of the
system,
which may be expressed by saying that q 1} q.2 ..., q n are true coordinates of
the system.
We shall now find the form which is taken by the equations
when the variables used are no longer restricted to be true coordinates of
the system*.
,
Consider a system denned by n true coordinates q lt q2 ...,

q n the
T and the work done by the external forces in a
,
kinetic energy being

*
known to Lagrange and Euler: the

due to Boltzmann (Wien. Sitzungsbcrichte, 1902) and Hamel
Particular cases of the theorem of this article were
general form of the equations

(Zcitschrift fiir
Math.
u.
is
Phys. 1904).
42
&%
%)
+ Q2 8q,
...,
being
Lagrangian equations of motion of the system are
displacement (8ql} 8q2
Let
<?i>
o) 2 ,
&>i,
fy,
qn
..., to n
be
?i
+ Q n fy
...
so that the
n>
independent linear combinations of the velocities
defined by the relations
a>
where
[OH.
=a
lr
+ a 2r q2 +
...
+ anr q n
(r=l,
2, ..., w)...(2),
ann are given functions of q lt qz ..., qn and let dir^, dir^

..., d7r n be n linear combinations of the differentials
dq lt dq2 ..., dq n defined
the
relations
by
,
or
21
...,
d7rr
where the
These
C$
a lr dq l
same as
coefficients a are the

last
2rd<? 2
. . .
ct
nr dq n
(r
...,),
1, 2,
in the previous set of equations.
equations would be immediately integrable
the relations
if
C^
= V^ were satisfied
^~
oqm
oq
for all values of K,
r,
and m, and in that case variables
ir r would exist which would be true coordinates

we shall not however
suppose the equations to be necessarily integrable, so that dir^, cfor 2 ..., d7r n
will not
necessarily be the differentials of coordinates 7r a 7r2 ..., ir n we shall
call the quantities d^, dir^, ..., d-rr n
differentials of quasi- coordinates.
;
Suppose that the relations

equations
q<
= /3Kl w +
l
when
(2),
/3 K2 &) 2
...
solved for q l}
/3 Kn
Multiplying the Lagrangian equations (1) by

and adding, we obtain the equation
3T\
(d
Now ^Q K SqK
is
wn
&
(r
r
2i .,
q.2
...,
I, 2,
qn
give the
...,n)...(3).
..., /8 Br ,
respectively,
the work done by the external forces on the system in an
arbitrary displacement, so
2 /3 Kr Q K S7rr
is
the work done in a displacement
in which
all the
If therefore the work
quantities STT are zero, except S?rr
done by the external forces on the system in an
arbitrary infinitesimal dis
placement (Swx, S?r2 ..., S7r n ) is n^TTj + II 2 S7r2 + ... + U n 87Tn W6 have
.
d
2/3
K
By means
"
of equations (3)
function T, so that
=n
~
f^_^l
\dqj dq
K)
(dt
we can
suppose for simplicity that

of T be denoted by T.
...,
w ni
q n from the
2
...,
q ly q2 ..., qn (we
not contained explicitly in T) let this form
is
&>
O/TT
Then we have
eliminate q1} q2
becomes a function of w
rl^P
=2r
CL*.
30]
43
and therefore
dT
^-
d /df
sj
But
*5<ft
Kr a KS
dT
-jda KS
at
dco s
dq K
is
zero or unity according as r
/ÛXS*
+
different from, or equal
is
to, s
so
we have
We
also
7.
dt (fo r )
sdT
"25"
aT
Sfl
5i
-TT
have
dT__dT
v 8f
~r ** o
o
-a
oqK
oq
a&>
_ af
dw s dq K
"i
dq
^ v af
8ams
d(o s
Qmy
"r^^
dq K
Now
Sy8 Kr ^
coordinate
we
or
shall
2=
s* -, would represent
5
denote
by the symbol
it
if
?r r
whether
were a true
TT,.
a true
is
OTT,-
coordinate or not.
Also the expression
depends only on the connexion between the true coordinates and the dif
ferentials of the quasi-coordinates, and is
independent of the nature or
motion of the dynamical system considered we shall denote this expression
:
by
We
7rz-
have therefore
at
rg
5
\0r/
ii
-- =
Od) s
= Ur
(r
C7Tr
= 1,
2,
ri).
These n equations are the equations of motion expressed in terms of the

when the quasi-coordinates are true coordinates, the
quasi-coordinates
7^
quantities
are all zero, since the conditions
5-^
oqm
-,.-
are satisfied, and
oq K
the equations reduce to the ordinary Lagrangian equations
dT\
dT
Example. A rigid body is free to turn about one of its points 0, which is fixed, so
that the coordinates of the body can be taken to be the three Eulerian
angles 6,
^,
which ( 10) specify the position of axes Oxyz, fixed in the body and moving with it, with
reference to axes
fixed in space.
Let an arbitrary displacement (80, 8$, 8-^) of the
body be equivalent to the resultant of small rotations ST^, STT,, 8*3 round Ox, Oy, Oz,
respectively, so that diri, dir.2 o?7r 3 can be taken as the differentials of quasi-coordinates
<,
OXYZ
let
0)1,
o>
o>
be the components about the axes Oxyz of the angular

velocity of the body
44
at
instant, so that diri, dir 2
any
are the differentials of quasi-coordinates corre

Shew that the equations of motion of
2
3
d-n- 3
sponding respectively to the velocities
[CH.
i,
&>
the body are
dT
(-(-}dt
dT
1-
dT =
<BO
^r
IIi,
\o<ai/
d /dT\
ffl
where
7* is
n n n
:
2,
ar
--
-- oay = H
a?7
3
OQ)
3,
OTs
the kinetic energy of the body, expressed in terms of 1?

2
3
6, 0, *//;
moments about the axes Ox, Oy, Oz, respectively, of the external forces
a>
a>
o>
are the
acting on the body
and
ar
of d&
stands for ^- ^
--
ot
OTT,.
It will
appear later that
Forces
31.
00)!
a>3/
T depends
derivable
from
j.
af J21
OTT,.
0
a 1}
VTr r
o) 2
&>
dy
87^
so the terms =
are zero.
uir r
a potential-function
which
involves
the
velocities.
In certain cases the conception of a potential-energy function can be

extended to dynamical systems in which the acting forces depend not
only
on the position but on the velocities and accelerations of the bodies.
For consider a dynamical system whose configuration is specified

by
..., q n and suppose that the work done by the external
forces in an arbitrary displacement (Bq lt 8q 2 ..., 8qn ) is
coordinates q 1} q2
Then
if
Qr
can be expressed in the form
dv
d /dv
jr-
-r^-.dt
where
V is a
tions of
given function of q lt q2
_
if
...,
qn q lt
...,
qn
= 1,
2,
n\
the Lagrangian equa
motion are
dt(dqj
and
(r
\dq r j
dq r
a kinetic potential
~
dqr
dqr
dt(dqr )
be defined by the equation
L = T-V,
the equations take the customary form
d fdL\
dL
U-.dt \dq r j
-T,
The
function.
function
)-a-=0
dq
can be regarded as a generalised potential

energy
of such a system is furnished
by the motion of a
An example
30-32]
45
law of attraction*
particle subject to Weber s electrodynamic
mass
on
the
unit
the
force
acting
particle being
per
point,
/
-2
r2 \
__
2r
f>
the distance of the particle from the centre of force

is defined by the equation
the function
where r
to a fixed
is
in this case
Example.
If the forces
Q lt ^2
...,
Qn
of a dynamical system which
is specified
by
coordinates q ly q 2 ...,q n are derivable from a generalised potential-function F, so that

,
_dt
oq r
shew that Q l} Q2
...,
QH must
be linear functions of q\,
q-2 , ...,
q n satisfying the n(2n

,
relations
On the general
may be made to
conditions for the existence of a kinetic potential of forces, reference

.
Helmholtz, Journal
fiir
Math., Vol.
c.
(1886).
Mayer, Leipzig. Berickte, Vol. XLVIII. (1896).

Hirsch, Math. Annalen, Vol. L. (1898).
Initial motions.
32.
differential equations of motion of a dynamical system cannot in

It is how
be
solved in a finite form in terms of known functions.
general
ever always possible (except in the vicinity of certain singularities which
need not be considered here) to solve a set of differential equations by power-
The
series,
i.e.
to obtain for the
dependent variables q lf q 2
...,
q n expressions of
the type
*/2
the coefficients a, b, ... can in fact be obtained by substituting these series in

the differential equations, and equating to zero the coefficients of the various
*
W. Weber, Annalen
of -Aether
d.
Phys. LXXIII. (1848), p. 193.
and Electricity, pp. 226
231.
Cf.
Whittaker
History of the Theories
46
the expansions will converge in general for values of

some definite circle of convergence in the tf-plane.
powers of
It
[OH.
within
plain that these series will give any information which may be
initial character of the motion (t being measured from the
is
required about the
commencement
of the motion), since

is the initial value of
q lt 6 X is the
This method of discussing the initial motion
value of q l9 and so on.
of a system is illustrated by the following example.
i
initial
Consider the motion of a particle of unit mass, which is free to move in a

rest, and which is acted on by a field of force whose components
Example.
plane and initially at
parallel to fixed rectangular axes at
determine the
Let (# +
so that
T)
initial
any point (x, y} are (X, Y)

radius of curvature of the path.
y + *i) be the coordinates
may
of
any point adjacent
be regarded as small quantities
.#)
and
let it
be required to
to the initial point
(.z;
y),
then the equations of motion are
3Y(x,y}
8y
If therefore
we assume
for
and
77
the expansions
to include terms of lower order than t 2 since the quantities

T)
(it is not necessary
TJ,
are initially zero), and substitute in these differential equations, we find, on comparing
the coefficients of various powers of t, the relations
,
y),
The path
where
Now
= 0,
of the particle near the point
(X
y) is therefore given
by the
series
denotes the quantity \&.

if
the coordinates | and
77
of
the radius of curvature at the point u
any curve are expressed

is
known
in
terms of a parameter
above expressions,
u,
to be
so the radius of curvature corresponding to the zero value of u, for the curve given
by the
is
(x ~ +
V
and
ox
r) x- (x *-+ Y ^\ Y
dy )
ctx
dy )
this is the required radius of curvature of the path of the particle at the initial point.
32-34]
47
Similarity in dynamical systems*.
33.
any system of connected particles and rigid bodies is given, it is

possible to construct another system exactly similar to it, but on a different
If now the masses and forces in the two systems, which we can call
scale.
the pattern and model respectively, bear certain ratios to each other, the
workings of the two systems will be similar, though possibly at speeds which
are not the same but bear a constant ratio to each other.
If
To find the relation between the various ratios involved, let the linear
dimensions of the model and pattern be in the ratio x 1, let the masses of
corresponding particles be in the ratio y 1, let the rates of working be in the
ratio z 1, so that the times elapsed between corresponding phases are in the
:
z, and let the forces be in the ratio w.\.

have an equation of motion of the form
ratio 1
mx = X
so if
is
I,
for each particle
we
x is altered in the
we must have
altered in the ratio y
altered in the ratio
Then
ratio xz*
1,
1,
and
is
w = xyz*,
and
this is the required relation
between the numbers
x, y, z,
w.
If the forces acting are those
due to gravity, we have w=y, and conse

Example.
2
quently xy = I, so that the rates of working are inversely as the square roots of the linear
dimensions.
If the forces acting are the mutual gravitations of the
particles, every particle
attracting every other particle with a force proportional to the product of the masses and
2
2
the inverse square of the distance, we have
are in
/x so that the rates of
w=y
the ratio y\
working
x%.
Motion with reversed forces.
34.
A special case of similarity is that in which the ratio w has the value 1.
We have seen that the motion of any dynamical system which is subjected
to constraints independent of the time, and to forces
the positions of the particles, is expressed
the
which depend only on

Lagrangian equations
by
dT
d /8T\
7
where the kinetic energy I

velocities q\, j a
and Q
is
a homogeneous quadratic function of the

involving the coordinates q lt q 2 ..., qn in any way,
..., q n
a function of q l
Introduce a
is
q2
...,
new independent
q n only.
variable defined
and
let
accents denote
*
it,
differentiations
with
Newton, Principia, Book n. Sect.
by the equation
where i = V
regard
7,
to
Prop. 32.
r.
Then
1,
since
48
d /dT\
dt
beC
hom g eneous
d fc^\
me
&>
-2
in dt, the above equations
drdq;)-^ =-$
is
..-,
But
of degree
a<T
where
3i,
dT
are
\M )
the same function of
qn, qi, q*,
-..,
[CH.
q,
qz
qn
...,
1,2, ...,),
<r=
>
q lt
...,
q n that
is
of
T (instead of t) be now
interpreted as denoting the time, these last
are
the
of
motion
of the same system when
equations
equations
subjected to
the same forces reversed in direction.
Moreover, if a 1} a2 ..., an
if
n are the initial values of q lt q,,
ft, /32
...,
q
respectively, in any
particular case of the motion of the original system, then a l} a 2 ..., a n -ifr,
...,
iftn will be the corresponding quantities in the transformed
i/3
thus have the theorem that in
problem.
any dynamical system subjected
to constraints independent
of the time and to forces which depend only on the
position of the particles, the integrals of the equations of motion are still real
...,
n>
q lt q
2>
...,
n>
2>
We
if
be replaced by
- V-
*J
-It and
the initial velocities
&,
..., /3 n
- V-
by
- V^l&,
l/32 ...,
l{j n respectively; and the expressions thus obtained repre
sent the motion which the same
with the same initial
system would have
,
if,
conditions,
it
were acted on by the same forces reversed in direction.
35.
Impulsive motion.
In certain cases (e.g. in the collision of
rigid bodies) the velocities of the
particles in a dynamical system are changed so rapidly that the time occupied
in the process may, for analytical
purposes, be altogether neglected.
The laws which govern the impulsive motion
of a system bear a close

analogy to those which apply in the case of motion under finite forces they
can be formulated in the following way*.
:
The number which represents the mass of a particle,

multiplied by the
vector which represents its
velocity at any instant, is a vector quantity
(localised in a line through the particle) which is called the momentum of
the particle at that instantf; the three
components parallel to rectangularmomentum of a particle of mass
at the point (x,
are
axes Oxyz of the
y, z)
therefore (mx, my, mz). If

any number of particles form a dynamical system,
the sum of the components in any
given direction of the momenta of the
particles is called the
component in that direction of the momentum of the
The impulsive changes
of velocity in the various

particles of a
connected system can be regarded as the result of sudden communications
of momentum to the particles.
system.
The
*
effect of
an agency which causes

impulsive motion in the system
They were involved
in the discovery of the laws of

impact in 1668 by Wallis
and Wren,
Phil. Trans. No. 43, pp. 864, 867.
t Momentum is the quantitas motus of

traced back to Descartes.
Newton
s Principia,
Book
i.
Def.
2.
The
idea can be
34, 35]
49
momentum which
it would communicate to a
single
are the components of velocity of a
particle of mass m, referred to fixed axes in space, before the impulsive
communication of momentum to the particle, and if (u, v, w) are the com
will
be measured by the
If therefore
free particle.
(?/ 0)
w)
ponents of velocity of the particle after the impulse, then the vector quantity
(localised in a line through the particle) whose components are
m (u
m (v
-w ),
m (w w
),
represents the impulse acting on the particle.
For the discussion of the impulsive motion of a connected system of

particles, it is clearly necessary to have some experimental law analogous to
the law of Action and Reaction of finite forces such a law is contained in
the statement that the total impulse acting on a particle of a connected
;
of the external impulse on the particle (i.e. the

impulse communicated by agencies external to the system, measured by the
momentum which the particle would acquire if free) together with impulses
directed along the lines which join this particle to the other particles which
is
system
constrain
to the resultant
equal
its
motion; and the mutually induced impulses between two connected
particles are equal in magnitude
and
opposite in sign.
If we regard the components of an

impulse as the time-integrals of the
components of an ordinary finite force which is very large but acts only for
a very short time, the law just stated
agrees with the law of Action and
Reaction
for finite forces.
Change of kinetic energy due
The change
given set of impulses
Let an impulse
reference are
from v
in
to impulses.
in kinetic energy of a
(A,
/,
^,
may
dynamical system whose particles are acted on by a
be determined in the following way.
directed along a line whose direction-cosines referred to fixed axes of

be communicated to a particle of mass m, changing its velocity
j/),
a direction whose direction -cosines are (Z
The equations
direction-cosines are (L, J/, JV).
NO), to v, in a direction
of impulsive motion are
,
whose
Multiplying these equations respectively by
$(vL + vQ L
$(vM+v
),
),
and
and adding, we have
The change in kinetic energy of the particle is therefore equal to the

product of the
impulse and the mean of the components, before and after the impulse, of the velocity of
the particle in the direction of the impulse.
Now consider any dynamical system of connected particles and rigid bodies, to which
given impulses are communicated applying this result to each particle of the system, and
summing, we see that the change in the kinetic energy of the system is equal to the sum of the
impulses applied to it, each multiplied by the mean of the components, before and
after the
;
communication of the impulse, of the velocity of its

point of application in the direction of the
In this result we can clearly neglect the
impulse.
impulsive forces between the molecules
of any rigid body of the system.
W. D.
50
[CH.
The Lagrangian equations of impulsive motion.
36.
The equations
impulsive motion of
of
system can be
Lagrangian equations of motion for
expressed in a form* analogous to the

finite forces, in the following way.
a dynamical
Let (X i} Yi, Zi) be the components of the total impulse (external and
molecular) applied to a particle m; of the system, situated at the point
The equations of impulsive motion of the particle are
(#,-, yi, z^.
(an
-x )=X
io
iy
(y
-y
io )
=T
i}
(z {
- z^) = Z
where (x^, y io z^) and (x i} y i} z^ denote the components of velocity of the

particle before and after the application of the impulse.
,
If
^ 2 ..., qn denote the n independent coordinates in terms of which
the configuration of the system can be expressed, we have therefore
,
<?!,
2 m,-
- asfe) =
\
(x t
"""I
h (yi
\
y*>)
vst
^- +
,
(&i
it}
Yv
)
5
9gv
??/
(JiXj
-*
cq r
where the summation
Now
it
is
in
extended over
is
OZ
T ^f ^
cqr/
oq r
all
forming the summation on the right-hand side of this equation,

26 that the molecular impulses between particles of the
seen as in
system can be omitted
the quantity
dq r
dqr
dqr
can therefore readily be found when the external impulses are known: we
shall denote it by the symbol Qr
We have consequently
.
2m,
t
But
as in
(& t
- 4-o)
+ (y t
2/ io )
oq r
(*t
oq r
- ^)
= Qr
dq r )
26 we have
da^_dxi
dq,.~dq
.
"bxi_
Vi
dqr
_ d (
-dq r
and similarly
.
dxi
_9^
*dqr~dqU
where q ro and q r denote the velocities of the coordinate q r before and after
the impulse respectively. Thus if
Due
to
Lagrange, Mec. Anal.
(2
ed.), n. p. 183.
36]
51
denotes the kinetic energy of the system after the impulse, the above
equation can be written in the form
dqr
/rl
where
\dq r /Q
(~
)
\oqr/ o
/}
T1
denotes the quantity corresponding to ^-, but relating to the

oc[r
instant before the impulse.
Similar equations can be found for the rest of the coordinates q 1} g 2

set of n equations
>
>
qnl
and thus we obtain the
fdT\
which are known as the Lagrangian equations of impulsive motion.
These are algebraical equations

terms of q w qw
determination of q lt q2
for the
...,
qn in
qm they are not differential equations like the Lagrangian

of motion for finite forces, since the second derivates of the
,
.,
equations
coordinates with respect to the time do not enter.
1.
Two rigid bodies moving in space are constrained only by a taut inextensible string
joining a given point of one body to a given point of the other, and one of the bodies is
constrained to roll without sliding on a given fixed surface. How many degrees of freedom
has the system, and how many independent coordinates are required to specify its con
figuration
2.
point
is
referred to curvilinear coordinates
velocity is
Shew that
a,
b,
c,
and the square of
its
2 T= A a? + Bb 2 + Cc2 + 2Fbc + 2Gcd + 2Hdb.

p, q,
r,
the component accelerations in the directions of the tangents to the
coordinate lines, are given by three equations of the type

d_
f
F\
dT
fiT\
- --- =
3-. )
dt (d
\dd
oaj
j
3.
field
da
..
HTT
(~*
p JA + -, _ q2 + -T--, r.
-
JB
(Coll.
Exam. )
v/6
particle which is free to move in space is initially at rest at the origin, and is in
of force whose components (X, F, Z) at any point (x, y, z) are given by the expansions
JT= a +bx+ quadratic and higher terms

Y= cx+ quadratic and higher terms
Z=
dx2 + cubic and higher terms
in x, y, z
in x, y,
Find the radii of curvature and torsion of the orbit at the
in x, y, z;
z.
origin.
42
CHAPTER
III
Problems which are soluble by quadratures.
37.
The determination of the motion of a holonomic dynamical system with

number of degrees of freedom has in the preceding chapter been
shewn to depend on the solution of a set of ordinary differential equations.
If n denotes the number of degrees of freedom, and (q 1} q2 ..., q n ) are the
a finite
coordinates specifying the configuration of the system at the time t, then

the set of equations consists of n differential equations, each of the second
order, with q lt q 2 ...,q n as dependent variables and t as independent variable.
This set of equations is said to be of order 2n, the order being defined to
,
be the sum of the orders of the highest derivates of the dependent vari
ables occurring in the equations.
It is a well-known result of the theory
of ordinary differential equations that the number of arbitrary constants of
integration in the solution of a set of differential equations is equal to
the order of the system whence it follows that there are 2n constants of
;
integration in the general solution of a holonomic dynamical problem with
n degrees of freedom.
Now
set of differential equations of order
any given
k can be reduced
to the form
-^
X X
where
l}
...,
new
are
= X r (ae
known
1}
x2
..,
xk
(r
t},
= 1,
2,
. .
.,
k\
functions of their arguments, by taking as
variables (x1 x2 ..., xk ) the original dependent variables together with

their derivates up to (but not including) the highest derivates occurring in
Thus e.g. the set of equations
the original set of equations.
,
Q and Q2
(where
order
dx
~T7
4,
are any functions of the arguments indicated) which
is
can be reduced to the set
dx2 _
_
^3>
~~rr
dx3
^4>
~jT
f)
^Jl
\&lt MS*
x
^3>
&4/I
dx4 _ ^
^2
~~JT
\&lt
-%2>
fiat
^4)5
of
53
Principles available for the integration
37]
by taking
= 01,
#1
=&,
a*
#3
= 0i,
^4
02-
The form
= Xr (x x2
1
-^
therefore be regarded
equations of order k.
as the
may
...,
(r
fc>
form
typical
for a
= l,
&)
2, ...,
set of differential
rlf
If a function
f{x
x2
...,xk
~-
such that
t) is
zero
is
when
(#,,
#2
>
>
#*)
tt-t
are any functions of
whatever which satisfy these differential equations, the
equation
/(#i, x2
is
#,
...,
t)
= Constant
an integral of the system. The condition that a given function

furnish an integral of the system is easily found
for the equation
called
may
dfldt
gives
ôx-i
*- 4 +
4 + cx
o
d^j
and
this relation
<*
4-
identically satisfied in order that the equation
f(x
may
L.
oa;2
must be
= 0,
8
8^Y ^
= ^;
+ 57
+ 5 / Â
dt
dx
57
dt
dxk
V Y + ^ -^2+
T
or
+ ~- xk +
oc 2
..., aek ,
t)
Constant
be an integral of the system of differential

equations.
Sometimes the function
called
an
solution of the set of differential equations of order k
is
/ itself (as
distinct
from the equation /= constant)
is
integral of the system.
The complete
furnished by k integrals
fr
On
2,
afc, t)
= ar
(r
1, 2,
k),
where a lt a z ..., ak are arbitrary constants, provided these integrals are

distinct, i.e. no one of them is algebraically deducible from the others.
For
let the values of as1} xz ..., xk obtained from these
equations as functions of
a 2 .. a k be
t, a,!
x r = <,.(!, a,, ..., ak t),
(r--l, 2, ..., k};
,
then
if
(a-j,
x2
...,
xk ) are any particular
set of functions of
which
satisfy
the differential equations, it follows from what has been said above that
by giving to the arbitrary constants a r suitable constant values we can make
the equations
/r
(ar,,
x2
...,
a! k
true for this particular set of functions (xl}

of functions (x ly xZ) ..., xk ) will be included
the equations x r
of freedom
may
t)
= ar
x2
...,
(r=l,
2, ...,
k)
xk ); and therefore this set
among
the functions defined by
The
solution of a dynamical problem with n

degrees
therefore be regarded as
equivalent to the determination of
r
<j>
2w integrals of a set of differential

equations of order
2n.
54
[OH. in

Thus the
differential equation
?=-?>
which
is
two integrals
of the second order, possesses the
tan- 1
and
a.2
are arbitrary constants.
?_;= a
?
where
+ ? 2 =,,
On
solving these equations for q
and these equations constitute the solution of the
and
q,
we have
differential equation.
The more elementary
division of dynamics, with which this and the

immediately succeeding chapters are concerned, is occupied with the dis
cussion of those dynamical problems which can be solved completely in
terms of the known elementary functions or the indefinite integrals of such

functions. These are generally referred to as problems soluble by quadratures.
The problems of dynamics are not in general soluble by quadratures and in
;
those cases in which a solution by quadratures can be effected, there must

always be some special reason for it, in fact the kinetic potential of the
problem must have some special character.
The
object of the
present
to discuss those peculiarities of the kinetic potential which are

chapter
most frequently found in problems soluble by quadratures, and which in fact
is
are the ultimate explanation of the solubility.
Systems with ignorable coordinates.
38.
We
have seen ( 27) that the motion of a conservative holonomic

dynamical system with n degrees of freedom, for which the coordinates are
.., q n and the kinetic potential is L, is determined by the differential
22
<?!,
equations
d fdL\
U-T-
-T.
at \dqr J)
- 5dL = 0,
= 1,
(r
2,
. . .
n).
dqr
7)T
The quantity
coordinate qr
^-roq r
is
generally called the
momentum corresponding
to the
that some of the coordinates, say q ly q z ..., q k are not

in L, although the corresponding velocities
contained
explicitly
q q%, ..., qk
are so contained.
Coordinates of this kind are said to be ignorable or cyclic
It
may happen
appear in the following chapters that the presence of ignorable

coordinates is the most frequently-occurring reason for the solubility of
it
will
particular problems
by quadratures.

coordinates are
of motion which correspond to the k ignorable
55
37, 38]
and on integration, these can be written
where
/3j, /3 2 , ..., /3&
These
are constants of integration.
equations are
last
evidently k integrals of the system.
We shall now shew how these k integrals can be utilised to reduce the
order of the set of Lagrangian differential equations of motion*.
k
Let
denote the function
dL
qr
r=\
~-r-
By means
(r-1,
f|-A,
we can express the k quantities q ly
of the k equations
v$r
2,
<j
..., qk,
which are the
2, ...,*),.
velocities cor
responding to the ignorable coordinates, in terms of

qk+i>
we
shall
qk+2>
>
<ln,
<ik+i,
qk+2,
Qn, @i, $*,
R is
suppose that in this way the function
->
ftk\
expressed in terms of the
latter set of quantities.
Now
tities
qk+1
Sf denote the increment produced in any function / of the quan
qk+l qk+2
,
...,
let
Bqn Bq 1}
,
qn
...,
&,
...,
2
qn
...,
fr,
$0
qn (or of the quantities q k+1) qk+2

by arbitrary infinitesimal changes Bqt+1 Sq
q2
...,
Sqn in its arguments.
ar
= b(L
by the definition of R.
...,
k+2>
qn
...,
Then we have
v
2,
r =\
3L\
qr
oqr/
But
and
since
We
have therefore
r=k + ivqr
r=k+lOqr
r=\
and since the infinitesimal quantities occurring on the right-hand side of this
equation are arbitrary and independent, the equation is equivalent to the
/
is
The transformation which follows is really a case of the Hamiltonian transformation, which
discussed in Chapter X; it was however first separately given by Eouth in 1876, and somewhat
later
by Helmholtz.
56
in
[CH.
system of equations
(r
dq r
dq r
dL
dR
= k+l,
+ 2,
..., n),
Substituting these results in the Lagrangian equations of motion,
we
have
m
Now E
is
^-
f^\ _ O
^L
7
ac \oqr /
I
"~~
<"\
n
^
ft
*V
l"
J-
....
A/ ~|
7fc ).
oqr
a function only of the variables qk+l q k+2

...,q n qk+l
,
...,q n
and the constants &, y32

so this is a new Lagrangian
ftk
system of
equations, which we can regard as defining a new dynamical problem with
,
only (n
new
k) degrees of freedom, the
coordinates being q k+1 qk+2

,
...,
qn
and the new kinetic potential being R. When the variables k+l k+2 ..., n
q
q
q
have been obtained in terms of t by solving this new
dynamical problem, the
remainder of the original coordinates, namely q l} 2 ...,
q
q can be obtained
from the equations
,
k>
Hence a dynamical problem with n degrees of freedom, which has k

coordinates, can be reduced to a
degrees of freedom.
ignoralle
dynamical problem which has only (n
This process
is
k)
called the ignoration of coordinates.
The essential basis of the ignoration of coordinates is in the theorem that when the
kinetic potential does not contain one of the coordinates
q r explicitly, although it involves
the corresponding velocity
namely
= constant.
be given
q,.,
This
is
an integral of the motion can be at once written

down,
a particular case of a
much more
general theorem which
when a dynamical system admits a known infinitesimal

contact-transformation, an integral of the system can .be immediately obtained.
will
later, to the effect that
If the original problem relates to the motion of a conservative

dynamical
system in which the constraints are independent of the time, we have seen
that its kinetic potential L consists of a
part (the kinetic energy) which is
a homogeneous
function
and which involves

q., ..., q n
with
a
qk+i, qk+z,
any way, together
part (the potential energy with
sign reversed) which involves qk+1 qk+2 ..., q n only. But in the new
dynamical system which is obtained after the ignoration of coordinates, the
,
quadratic
of q lt
q n in
kinetic potential
cannot be divided into two parts in this
will in general contain terms linear in the velocities.
And
way
in fact,
more generally
when (as happens very frequently in the more advanced parts of
Dynamics)
the solution of one set of Lagrangian differential
equations is made to depend
57
38]
on that of another set of Lagrangian differential equations with a smaller

number of coordinates, the kinetic potential of this new system is not
necessarily divisible into two groups of terms corresponding to a kinetic and
a potential energy.
shall sometimes use the word natural to denote those
of
systems
Lagrangian equations for which the kinetic potential contains
We
in the velocities, and non-natural to denote

only terms of degrees 2 and
those systems for which this condition is not satisfied.
As an example of the ignoration of coordinates, consider a dynamical system with two

degrees of freedom, for which the kinetic energy is
and the potential energy

where
a, 6,
c,
is
are given constants.
It is evident that q^ is
or
an ignorable coordinate, since
it
does not appear explicitly in
V.
The
kinetic potential of the system is
and the integral corresponding to the ignorable coordinate
where
ft is
The
a constant, whose value
kinetic potential of the
and the problem
is
is
determined by the
is
initial
circumstances of the motion.
new dynamical system obtained by ignoring the coordinate
now reduced
to the solution of the single equation
(3R
dt\dq
or
As
this is a linear differential equation with constant coefficients, its solution
immediately written down
where
A and
of the motion.
are constants of integration, to be determined by the initial circumstances

This equation gives the required expression of the coordinate 2
terms
<?
of the time
can be
it is
the value of qi in terms of
can then be deduced from the equation
which gives
8i n 2
and so completes the solution of the system.
58

39.
Special cases of ignoration; integrals of
[CH.
in
momentum and angular
momentum.
We
now
shall
consider specially the two
commonest types
of ignorable
coordinates in dynamical problems.
Systems possessing an integral of momentum,.

Let the coordinates of a conservative holonomic dynamical system with
n degrees of freedom be q l} q
and let T be the kinetic energy of the
.., q n
and
V
the
system,
potential energy, so that the equations of motion of the
(i)
2>
system are
dT
dV
~-~~
d (dT\
Suppose that one of the coordinates, say q l} is ignorable, and moreover is

such that an alteration of the value of q 1 by a quantity I, the
remaining
coordinates qz qs ..., q n being unaltered, corresponds to a
simple translation
of the whole system through a distance I parallel to a certain fixed direction
,
in space we shall take this to be the direction of the #-axis in a

system of
fixed rectangular axes of coordinates.
;
Since
q-i
is
an ignorable coordinate, we have the integral

O/TT
-z-r
Constant,
dq l
and we
shall
We
have
now
discuss the physical
ab-*Sb*
dT
is
miW + *f + if
>
extended over
all
the particles of the system,
$-M^T*
dAi
<
difi
+ v-^ + z-
^
= Zmtiki,
^}
*^/
y<
as,
3ft
,
bv
y
9#o
since in this case ^-
dq
Now
the
this equation.
O/TT
where the summation
meaning of
8
^
26
^=
9vi
1,
^=
9^j
0,
0.
dq 1
oqi
-axis of
Smgfej represents ( 35) the component parallel to the
of the system of particles m^, and consequently this is the
momentum
r
ri
physical
meaning of the quantity
r
in the present case.
0Ji
~\m
The
integral
= Constant
9g,
When a dynamical system can be

interpreted thus
translated as if rigid in a given direction without violating the constraints,
and the potential energy is thereby unaltered (the way in which the kinetic
can
therefore
be
energy depends on the velocities
is
obviously unaltered by this translation, so
ignorable), then the component parallel to this

of the system is constant,
the corresponding coordinate

direction of the
which
momentum
is
law of conservation of momentum*, and systems

applies are said to possess an integral of momentum.
This result
to
59
39]
it
is
called the
Systems possessing an integral of angular momentum.
(ii)
Again taking a system with coordinates q 1} q2 ..., q n and kinetic and

potential energies T and V respectively, let us now suppose that the
coordinate q is ignorable, and moreover is such that an alteration of q by
,
a quantity a, the other coordinates remaining unchanged, corresponds to

a simple rotation of the whole system through an angle a round a given
fixed line in space
we shall take this line as the axis of z in a system
of fixed rectangular axes of coordinates.
:
Since q
is
an ignorable coordinate, we have the integral

r
?i r
............................. (1),
;r-r- = Constant
9?i
and we have to determine the physical interpretation of
We
have as before
dT
/
j
where the summation

if
we
this equation.
is
=-- yt dyi
^.
r-
dqi
all
dqj
9?i
extended over
dzA
+ Zi =
But
write
Xi^ncosfa,
yi
dfa = dq
we have
d%i
~
= dxi
=
rism<f>i,
riSm>i=:
..
and therefore
3-r-
oq
Now
= 2m
(-
dayt
+ yiXi)
........................... (2).
r denote the distance of any particle of mass

from a given
and
if
denote the angular velocity of the
any instant,
particle about the line, the product mr*to is called the angular momentum
of the particle about the line.
if
straight line at
Let
moving
be any point, and

particle, the interval
<a
let
P,
of time
be two consecutive positions of the

between them being dt. Then the
This has been evolved gradually from the observation of Newton, Principia, Book i. iutrod.
if any number of bodies are acted on
only by their mutual attractions, their
common centre of gravity will either be at rest, or move uniformly in a straight line.
to Sect. XL, that
60
angular momentum about any line

value of the ratio
OK
??z
-7-
x Twice the area of the projection of the triangle

a plane perpendicular to
so if
OPP
on
OK,
OK and
m, n) are the direction-cosines of

cosines of the normal to the triangle
(I,
clearly the limiting
is
through
[CH.
OPP
if (X, p, v)
we
momentum about OK is
angular momentum about
are the direction-
that the angular

mp + nv) into the
see
equal to the product of (l\ +

the normal to the plane
It is evident from
this that if the angular momenta of a
about
particle
any three rectangular
axes Oxyz at any time are A,, A2 A 3 respectively, then the
angular momentum
OPP
about any line through

whose direction-cosines referred to these axes are
is
mh
nh
we may express this by saying that angular
Ih-i +
(I, m, n}
2 +
s
momenta about axes through a point are compounded according to the vectorial
;
law.
The angular momentum
of a dynamical system about a given axis
is
sum
of the angular momenta of the separate particles of

the system about the given axis in particular, the
angular momentum of
a system of particles
typified by a particle of mass m, whose coordinates are
defined to be the
(xi,
yiy
Zi},
about the axis of z
and the summation

expression for
is
is
Sm^r^, where
extended over
the angular
momentum
all the particles of the system

this
of a system can be written in the form
;
2m* (ijiXi -
xî),
and on comparing
momentum of
this with equation (2)
we have the
result that the angular

rlT
the system considered, about the axis
of
z, is
~-r
dî
The equation
(1) implies therefore that the angular momentum of the

system about the axis of z is constant and we have the following result
When a dynamical system can be rotated as if rigid round a given axis with
:
out violating the constraints,
and
angular momentum of the system

This result is known as
momentum*.
the potential energy is thereby unaltered, the
about this axis

the
theorem
is constant.
of conservation
of angular
Example. A system of n free particles is in motion urider the influence of their

mutual forces of attraction, these forces being derived from a kinetic potential V, which
contains the coordinates and components of velocity of the particles, so that the equations
of motion of the particles are
_ar
mrXr ~
..
Kepler
d /3F\
e
law, that the radius from the sun to a planet sweeps out equal areas in equal times,
from this the general
to all cases of motion under a central force
was extended by Newton
theorem of conservation of angular
momentum has
gradually developed.
61
39, 40]
shew that these equations possess the integrals

(
m r xr +^- J = Constant,
9F\
.,
yr + K-i
m r z r + ^r- = Constant,
cz r /)
r V
(
2 \m r (yr zr - z r yj
2
+y
m r (zr xr - x,. zr + zr
)
(^
may
be regarded as
.-
zr
9F)
= Constant,
^.-j
F-
JTTox r
an =
Constant,
xr 3
cz r
9 F!
F
=
C~yrgj}- Constant,
momentum and
generalisations of the integrals of
2 \m r (x r y r -y r x r ) + x r
which
3F -
Constant,
yJ
r \
F\
momentum.
The general theorem of angular momentum.
40.
The
result,
angular
(Levy.)
integral of angular
which
may be
momentum
is
a special case of a more general
obtained in the following way.
Consider a dynamical system formed of any number of free or connected

and interacting particles if they are subjected to any constraints other than
the mutual reactions of the particles, we shall suppose the forces due to these
constraints to be counted among the external forces.
:
Take any
line fixed in space,
and choose one of the coordinates which
to be such that a change in

specify the configuration of the system (say q^
in
the
other
coordinates,
implies a simple
q 1} unaccompanied by any change
rotation of the system as if rigid round the given line, through an angle equal
to the
change in q lf
We
suppose the constraints to be such that this
is
possible displacement of the system.
The Lagrangian equation
for
the coordinate q
is
dtdq
and
this reduces to
since the value of q : (as distinguished from
q^ cannot have any
a/77
the kinetic energy, and therefore

j
must be
zero.
Now
effect
on
T^T
the angular
is
-^-r
momentum of the system about the given line; and QiS^j is the work done
on the system by the external forces in a small displacement 8q 1} i.e. a small
rotation of the system about the given line through an angle 8q from which
1
it is easily
line.
We
moment
of the external forces about the given

Q
have therefore the result that the rate of change of the angular
seen that
is
the
62
momentum
[CH. in
of a dynamical system about any fixed line is equal to the moment

The law of conservation of angular
when the moment of the external
of the external forces about this line.

momentum obviously follows from this
forces is zero.
Similarly
to
dynamical system parallel

parallel to this
of the total external forces acting on the system.
line,
For impulsive motion

results
the rate of change of the momentum of a

any fixed direction is equal to the component,
we can shew that
it
easy to establish the following analogous
is
The impulsive increment of the component of momentum of a system in any

fixed direction is equal to the component in this direction of the total external
impulses applied
to the system.
The impulsive increment of the angular momentum of a system round any

axis is equal to the moment round that axis of the external impulses applied to
the system.
The Energy equation.

now introduce an integral which plays a great part in dynamical
investigations, and indeed in all physical questions.
41.
We
shall
In a conservative dynamical system let q i} q2 ..., qn be the coordinates

we shall suppose that the constraints
let L be the kinetic potential
,
and
L is a given function of the variables

not
We shall not, at
only,
qn, q\, q-2,
involving t explicitly.
qi, #2,
L
further
so
that
the
discussion
will apply
restrict
conditions,
first,
by any
to the non-natural systems obtained after ignoration of coordinates, as well as
are independent of the time, so that
>
>
<jn
to natural systems.
We
have
dL =
~T~
at
..
r =\
..
2,
qr
r =i
Integrating,
"r
dL
;r-r-
Or o
dqr
-
qr r Urr,
r =i
by the Lagrangian equations
n
v<Jr
a constant.
This equation
energy or
we have
l
is
at \oqr J
dL^
*q,
r=
where h
d idL\
+ 2
vqr
/
dL
i
r=i
uq r
"
9Z
q? D^~
is
an integral of the system, and
is
called the integral of
law of conservation of energy*.
Galileo was acquainted with the fact that the velocity of a particle sliding down an inclined
plane from rest depends only on the vertical height through which it has descended. From this
elementary particular case the principle was gradually evolved by Huygens, Newton, John
and Daniel Bernoulli, and Lagrange.
63
40, 41]
We
have seen that in natural systems, in which the constraints do not

involve the time, the kinetic potential L can be written in the form T
V,
where T (the kinetic energy of the system) is homogeneous and of degree 2
V is
in the velocities, while
In this
a function of the coordinates only.
case,
therefore, the integral of energy becomes

n
^
h=Zq
r=i
.
= 2T
2 r
Olj
^
LT
oqr
T +V,
since
T is
homogeneous of degree 2
It follows that in conservative natural systems, the
potential energies is constant.
This constant value h
in qlt
sum of
the kinetic
and
called the total energy
is
of the system.
This latter result can also be obtained directly from the elementary
equations of motion. For from the equations of motion of a single particle,
namely
m x = Xi
i
m yi = Y
m ^ = Z{
{
we have
Em*
(xi xi
where the summation
+ yiiji + Zi Zi) = 2 (XiXi + Yiiji + Z^t),

extended over
is
d S I m t (x?
.
y?
all
the particles of the system, or
+ zf) = 2 (Xdx + Ydy + Zdz},
increment of the kinetic energy of the system, in any infinitesimal

path, is equal to the work done by the forces acting on the system
so that the
part of
its
in this part of the path, and therefore is equal to the decrease in the
potential
of the
energy of the system. The sum of the kinetic and potential
energies
system
is
therefore constant.
The equation
of energy
m & + f + p) = Xdx + Ydy + Zdz

(
(where for simplicity we suppose the system to consist of a single particle) is true not
only when (#, y, z) denote coordinates referred to any fixed axes, but also when they
denote coordinates referred to axes which are moving with any motion of translation
in a fixed direction with constant velocity.
For let (, ;, f ) denote the coordinates of the particle referred to axes

and parallel to the moving axes Oxyz, so that
x=-at, y = T)-bt,
where
-ct,
the constant components of velocity of the origin

Then the result already proved is that
a, b, c are
d .4 m (e + ^ + C 2 ) = Xd +
fixed in space
Yd?,
of the
moving
+ (aX+bY+cZ)dt.
axes.
64
Now we
[OH.
have
d.
= m (at; + bij + c dt
= (aX+bY+cZ)dt,
)
and therefore
d.im(x*+y2 + z = Xdx+ Ydy + Zdz,

i
which establishes the theorem.

It may be noted that from this result the three equations of motion of the particle
at etc., and subtracting the equation of energy in the
can be derived, by taking x =
coordinates (#, y, 2) from the equation of energy in the coordinates (, 77, ).
Reduction of a dynamical problem
42.
a problem with fewer degrees of
to
freedom, by means of the energy-equation.
When
a conservative dynamical system has only one degree of freedom,

the integral of energy is alone sufficient to give the solution by quadratures.
For
if
q be the coordinate, the integral of energy

.
dL
-TT-.
L=h
dq
a relation between q and q
is
from this equation, so that
.if therefore q be found

explicitly in terms of q
takes the form
it
9 =/(?),
we can
integrate again and obtain the equation

t
-.^r-r
constant.
which constitutes the solution of the problem.

When the system has more than one degree of freedom, the integral of
energy is not in itself sufficient for the solution but we shall now shew that
;
can be used for the same purpose as the integrals corresponding to ignorable coordinates were used, namely to reduce the system to another dynamical
it
system with a smaller number of degrees of freedom*.

In the function L, replace the quantities q.2 q 3 ..., q n by ftq2 ftq^,
,
ft
q^, respectively,
by
n(<j,,
<//,
qs
q.2,
where qr denotes
...,
...,
qn
qn,
,
<?i,
q-2
ft, q*,
dL
wehave
...,
qn )-
qn )
an
q2
r
differentiating the equation

,
q3
aii
(-1^,3,...,,)
*
...,
and denote the resulting function
Then
= &(ft,
Whittaker, Mess, of Math. xxx. (1900).
......... (3).
41, 42]
Equations (1) and (2) give
**.*
ir
,-=2 î
oq r
80 _ *A
+ I
~
o
.
^-1
0$
oqi
Now
65
in the integral of energy

n
dL
r=l
qr
~]r-L
Off
........................... \^)
~i~ ^TT~
= h,
replace gv by^gv for all values of r from 2 to w inclusive, and then from this
equation obtain ^ as a function of the quantities (q2 q3 ...,q n qi,
qn)
and by using this expression for ql} express the function
,
>
>
<?2>
dL
g,
r=l 9^r ?i
in
terms of (^2
be denoted by
^3
...,
qn
q 1} q 2
...,
Let the function thus obtained
q n ).
then from (4) we see that
r)O
the same as ^-r
is
but
5î
differently expressed.
Differentiating the equation of energy, which by (4)

the form
an -
written in
li=h,
(ft
1
may be
fyi
and regarding
the variables
But
it
(<?/,
as a relation which implicitly determines

q.,
...,
qn
q lt q 2
...,
as a function of
differentiating the equation

r,
an
8^
regarded as an identity in the variables (g/, q3
9X
Comparing equations
(5)
_j^
and
(7),
az/
and comparing equations (6) and
we have
i
(8),
ao
Wr
we have
1
30
^Wr
...,q n
a^na^
^8"
dL
W. D.
qn ), we have
"
q-i,
qz
-,
qn ),
we have
66
Combining these with equations
(2)
dL^_dL
*\
-i
dqr
we
and
anQ
>
dq r
(3),
in
[CH.
we have
dL _
_
dq r
?i
dL
.
dqr
Substituting from these equations in the Lagrangian equations of motion,

obtain the system
d (dL \
-T. (5,
dt\dqr j
q
*l
dL
~
0,
(r
= 2,
(r
3,
. .
2, 3,
TO),
n).
dq r
or finally
Now
#zese
may
L
system in which
and q plays
be
dL
regarded as the equations of motion of a new dynamical
is the kinetic potential,

(qz ,
q3
...,
qn ) are the coordinates,
The new system

part of
obtained
of
coordinates, be in general
systems
by ignoration
i.e.
L
will
in the
not consist solely of terms of degrees 2 and
non-natural,
velocities (q 2 q3 ..., qn ) but on account of its possession of the Lagrangian
form, most of the theorems relating to dynamical systems will be applicable
the time as the independent variable.
the
will, like the
it.
The integral of energy thus enables us to reduce a given dynamical
system with n degrees of freedom to another dynamical system with only (n
1)
to
degrees offreedom.
The new dynamical system
will not in general possess an integral of

since
the
variable
energy,
independent
q l occurs explicitly in the new kinetic
L
if
But
an
is
potential
ql
ignorable coordinate in the original system,
.
then q l
will not occur explicitly in
any stage of the above process, and there

fore will not occur explicitly in
From this it follows that the new system
will also possess an
integral of energy,
namely
2
r=2
qr 3
Oqr
L =
constant,
and this can in its turn be used to reduce further the number of degrees of
freedom of the system.
The preceding theorems shew

n degrees
of freedom
and (n
that any conservative dynamical system with
be completely
integrated by quadratures we can proceed either (a) by first performing the
process of ignoration of the coordinates, so arriving at a system with only one
degree of freedom, which possesses an integral of energy and can therefore be
1) ignorable coordinates can
solved in the
manner indicated
at the beginning of the present article or

use the integral of energy to lower the number of degrees of
freedom by unity, then use the integral of energy of the new system to lower
the number of degrees of freedom again by unity, and so on, obtaining
(@) we can
first
finally
a system with one degree of freedom which again can be solved in the manner
indicated.
42, 43]
The
Example.
Shew
kinetic potential of a dynamical system
Jj
and where
=^,
is
is
given by the differential
__~
__
5ft \3$iY
where
is
that the relation between the variables l and z

q
q
equation
67
9? 2
denned by the equation
Shew that the non-natural dynamical system represented by the last differential
equation possesses an integral of energy, and hence solve the system by quadratures.
43.
fs
is
Separation of the variables ; dynamical systems of Liouville s type.
dynamical equations which are obviously soluble by quadratures

constituted by the equations of those systems for which the kinetic
energy
of the form
class of

where
vlt v2
...,
vn
V=
w w
lt
is
of the form
u>i
+ W (q ) + + wn (q n
w n are arbitrary functions
(?i)
...,
),
arguments so that the kinetic potential breaks

of which involves only one of the variables.
;
For in
this case the
Lagrangian equations of motion are
^ K (qr)
or
up
of their- respective
into a sum of parts, each
qr ]
v r (q r ) q r
- \ vr
+%
(q r) q,?
vr (qr ) qr2
= - ff [ (qr
= - w/ (qr \
i
(r
),
1 , 2,
2,
1, 2,
= 1,
(r
n),
n).
n),
These equations can be immediately integrated, and

give
%vr (q r ) qr
.
where d,
+ wr (q r ) = c r
^/
(r
are constants of integration; these

equations can be
further integrated, since the variables r and t are
q
separable, and we thus
obtain
c2
..., c n
^,
where ylt y a ...,yn are new constants of
integration.
,
(r-!,
These
2,
last
...,),
equations
constitute the solution of the problem.
An
important extension of this class of dynamical systems was made by

who shewed that all dynamical problems for which the kinetic
and potential energies can
respectively be put in the forms
Liouville*,
T=
V=
K
Wl
+ Ui (q ) + ...+* (qn )}
w
+
+ --- + wn (qn )
(?,)
^ ^
Vl (ft)
^+v
(q2 ) qs*+...
+Vn (qn ) j n*}
can be solved by quadratures.

*
Journal de Math. xiv. (1849),
p. 257.
52
[CH. in
68
For by taking
>Jv
where
vi
(ft)>
ft
ft
...,
V2
kinetic
are
qn
new
(q r )
dq r
= qr
we can
variables,
v n (qn)
we
(r
replace
all
1, 2,
.
,
w),
the functions
suppose this done, so that the
shall
by unity
(ft),
and potential energies take the form
>
T=
^(ft +
2
ft
+...+ft> ),
V = - K (ft) + w
where u stands
for
+ wn (q n }},
(ft)
the expression
i
(ft)

d
u 2 (ft)
+un (qn
...
).
for the coordinate q 1 is
ar = _
fdr\ _
dt \9ft /
3^
dv
dql
Multiplying this equation throughout by 2uqlt we have
~
(%>
But from the
*g w+ ^+
.) =
integral of energy of the system,
%u(q +
2
q->+
...
where h is a constant. The equation

be written in the form
Integrating,
..:
2 ft
qn*)
{h^
(?j)
we have
= h-
for the
V,
coordinate
- w,
q^
can therefore
(q,)}
we have
2
û q^ =
AM, (ft)
- Wj
(ft)
7j,
a constant of integration.
We obtain similar equations for each
the
...,
n
(ft, ft,
q )
corresponding constants (7!, j.2 ..., yn )
must satisfy the relation
where
71 is
of the coordinates
7i
72
+ 7 = 0,
in virtue of the integral of energy of the system.
69
43]
These equations give

(qj
{hu-t
- Wt
(<?i)
dq
71}
[hu 2
(g- 2 )
(q 2 )
*)%}""*
dq2 =
...
- wn (qn ) + yn
{hu n (q n )
~%
]
dq n
this set of equations, which can be immediately integrated since the

variables are separated, furnishes the solution of the system.
and
For further investigations on this subject cf. Hadamard, Bull, des

and Burgatti, Rom. Ace. L. Rend. (5) xx. (1911), p. 108.
Sc.
Math. xxxv. (1911),
p. 106,
components (X, Y} of the force acting on a particle of unit mass at the
a
in
plane do not involve the time t, shew that by elimination of t from the
y)
point
differential equations the solution of the problem is made to depend on the differential
1.
If the
(#,
equation of the third order
dx
dx*
2.
if
rigid
at once be written
3.
free particles is in motion,
down
is
unaltered
h, k,
The
and
is
is
where a, b, c, d are constants.

an equation of the form
4.
when the system
In a dynamical system with two degrees of freedom the kinetic energy
where
and their potential energy, which depends

in any configuration is translated
through any distance in any direction. What integrals of the motion can
system of
only on their coordinates,

as
--
&y
Shew
that the value of q 2 in terms of the time
is
given by
are constants.
kinetic potential of a dynamical system is
where
a, b, c
where
e is
are given constants
shew that
an arbitrary constant and
$>
q^ is given in
terms of
by the equation
denotes a Weierstrassian elliptic function.
Prove that in a system with ignorable coordinates the kinetic energy is the sum
of a quadratic function T of the velocities of the non-ignored coordinates and a quadratic
5.
function
K of the cyclic momenta.
In the case where there are three coordinates x, y,
<
and one coordinate
investigate the equations of motion of the type

8
/30\)
oy \dxjl
is
ignored
70
[CH. in
where Fis the potential energy, Jc is the cyclic momentum, and the differential coefficients
with respect to x, and $ are calculated from the Jinear equation by which k is
expressed in terms of x, i/, 0.
(Camb. Math. Tripos, 1904.)
of
</>
The
6.
kinetic potential of a dynamical system with
two degrees of freedom
is
By using the integral of energy, shew that the solution depends on the solution of the
problem for which the kinetic potential is
and by using the integral of energy of

qi and q 2 is of the form
where
and
this latter system,
are constants of integration, and
|P
shew that the
relation between
denotes the Weierstrassian
elliptic-
function.
The
7.
kinetic energy of a dynamical system
T=
2
ri
(qi
is
2
)
<72
(<?1
+ ?22
)>
is
F=-J
Shew (by use
and q2
is
where
a, b,
8.
and
of Liouville s
theorem, or otherwise) that the relation between q l
a 2 qi 2 + b 2 q%2 + 2abq l q 2 cos y = sin 2 y,
The
y are constants of
integration.
kinetic energy of a particle
its potential
energy
whose rectangular coordinates are
(x,
2
2
y) is ^ (x +^ ),
is
where (A, A 5, B C~) are constants and where (r,

are the distances of the particle
from the points whose coordinates are (c, 0) and (c, 0), where c is a constant. Shew
that when the quantities J(V + /) and \(r-rf } are taken as new variables, the
system is
of Liouville s type, and hence obtain its solution.
,
9.
The observation that
A system,
r")
"
a cat always
instant resume
Shew that
its
on
its feet
"
suggested the problem

completely specified by the position and velocity
Can it at a subsequent
initially without velocity in free space in vacua.
initial configuration but with a
different orientation in space ?
whose state at any instant
of each element, is
falls
is
the system is not conservative, or if the forces are derived from a

potential
not one-valued, the reply is affirmative but if the system is conservative with a
one-valued potential, the reply is negative.
which
if
is
(Of. Painleve,
Comptes Rendus, cxxxix. (1904),
p. 1170.)
CHAPTER
IV
THE SOLUBLE PROBLEMS OF PARTICLE DYNAMICS

The particle with one degree offreedom
44.
As examples of the methods described

now discuss those cases of the motion of a
the
pendulum.
in the foregoing chapters,
we
shall
single particle which can be solved
by quadratures.
We
motion of a particle of mass m, which is free

to move in the interior of a given fixed smooth tube of small bore, under
the action of forces which depend only on the position of the particle in the
shall consider first the
The tube can
tube.
in the
most general case be supposed
to
have the form
of a twisted curve in space.
Let s be the distance of the particle at time t from some fixed point of
the tube, measured along the arc of the curve formed by the tube: and let
f(s) be the component of the external forces acting on the particle, in the
direction of the tangent to the tube.
The
and
kinetic energy of the particle
its
potential energy
is
is
evidently
-f f(s)d,
where
is
a constant.
The equation
of energy
is
therefore
rs
2
|-ms
f(s)ds +
c,
J So
where
c is
a constant.
where
we have
another constant of integration. This equation represents the

solution of the problem, since it is an integral relation between s and t,
I
is
involving two constants of integration.
The Soluble Problems of Particle Dynamics
72
The two constants

initial
time
and
can be physically interpreted in terms of the

thus if the particle starts at
circumstances of the particle s motion

= t from the point s = s with velocity
,
values in the equation of energy,
u,
then on substituting these
we have
and on substituting the same values
we have
[OH. iv
in the final equation

connecting s
and
t,
t^.
The most famous problem

in this case the tube
of this type is that of the simple

pendulum
supposed to be in the form of a circle of radius a
is
whose plane
is vertical, and the

only external force acting on the particle is
gravity*.
Using 9 to denote the angle made with the downward vertical by
the radius vector from the centre of the circle to the particle, we have
= ad
so the equation of energy
a$ 2
= 2$r cos
and f(s) =
+ constant =
Suppose that when the particle

a 2 /9 2
quantity -= has the value
sin
Taking
Now
^9 = y
this
mg sin
is
is
Then
h.
4#
sin 2
\Q +
constant.
at the lowest point of the circle, the
this last equation can be written
becomes
pendulum-problem there are two distinct types of motion,

namely the oscillatory," in which the particle swings to and fro about the
lowest point of the circle, arid the circulating," in which the velocity of the
in the
"
"
enough to carry it over the highest point of the circle, so

moves round and round the circle, always in the same sense. We
particle is large
that
it
shall consider these cases separately.
In the oscillatory type of motion, since the particle comes to rest

(i)
before attaining the highest point of the circle, y must be zero for some value
of y less than unity, and therefore h/2a must be less than
unity.
Writing
where k
is
new
2ak2
positive constant less than unity, the equation
In actual pendulums, the tube
is
centre of the circle, which serves the

the circle.
becomes
replaced by a rigid bar connecting the particle to the

particle to describe
same purpose of constraining the
The isochronism of small oscillations of the pendulum was discovered by Galileo in 1632,
and the formula for the period was given by Huygens in 1673. Oscillations of finite
amplitude
were
first
studied by Euler in 1736.
73
44]
the solution of this is*
where
is
an arbitrary constant.
This equation represents the solution of the pendulum-problem in the

the two arbitrary constants of the solution are t and k, and
oscillatory case
:
these
must be determined from the
initial
conditions.
From
the
known
is periodic, its
properties of the elliptic function sn, we see that the motion
on which
occasions
two
consecutive
of
time
between
the
interval
period (i.e.
the
is
pendulum
in the
same configuration with the same
velocity) being
4.(O^jr
K, where
,
K=\
(ii)
case h
(1
- kH y% dt.
z
)~? (1
in this
Next, suppose that the motion is of the circulating type
2
=
k
will be less
A&
2a
the
if
we
write
than
so
2a,
quantity
greater
;
is
than unity.
The
the solution of which
and in
this
now becomes
is
and k are the two constants which must be determined in
accordance with the initial conditions.

(iii)
Lastly, let h be equal to 2a, so that the particle just reaches the
vertex of the
The equation now becomes
circle.
is
was remarked by Appellf that an insight into the meaning of the imaginary period
which occur in the solution of the pendulum-problem is afforded
For we have seen that if the particle is set free with no initial
velocity at a point of the circle which is at a vertical height h above the lowest point,
the motion is given by
It
of the elliptic functions

by the theorem of 34.
where & 2 =
*
Cf.
Whittaker and Watson, Modern Analysis,
t Comptes Rendiis, LXXXVII. (1878).
22-11.
74
[OH. iv
and therefore by 34, if, with the same initial conditions, gravity were supposed to act
upwards, the motion would be given by
|(r-r,),
But the period

(2a
- A),
same as if the initial position were at a height

and the solution of this is
of this motion is the
gravity acting
downwards
y = V sn { A/1
The
motion has a
latter
A.
real period 4
(\t7
/j
(r
- TO ), V\
elliptic function sn
is
=1
- k2
and therefore the function
must have a period

so the function sn(w,
4^-j/r,
\t//
double periodicity of the
where
must have a period
4*/f.
The
thus inferred from dynamical considerations.
Example. A particle of unit mass moves on an epicycloid, traced by a point on the

circumference of a circle of radius b which rolls on a fixed circle of radius a. The
particle
is acted on
by a repulsive force pr directed from the centre of the fixed circle, where r is
the distance from this centre. Shew that the motion is
periodic, its period being
[This result
the form
is
most
easily obtained
when the equation
of the epicycloid
is
taken in
being the arc measured from the vertex of the epicycloid.]
Motion in a moving
45.
We
to
shall
move
now
tube.
some cases of the motion of a particle which is free

smooth tube, when the tube is itself constrained to move
discuss
in a given
in a given manner.
(i)
at
Tube rotating uniformly.
Suppose first that the tube

about a fixed axis in space.
is
constrained to rotate with uniform velocity

shall suppose that the particle is of unit
We
mass, as this involves no real loss of generality.
We
moreover suppose that the field of external force acting on the

derivable
from a potential- energy function which is symmetrical
particle
with respect to the fixed axis, and so can be expressed in terms of the
cylindrical coordinates z and r, where z is measured parallel to the fixed
shall
is
axis
and r
the
from the fixed axis for a particle

perpendicular
in the tube, this potential
can
therefore
be expressed in terms of
energy
the arc s we shall denote it by F(s), and the
equation of the tube will be
is
distance
written in the form

r
= g(s).
velocity
were
o>
term
motion of the particle is the same as

zero, and the potential energy were
29, the
By
^r
the form
&>
Hence we can
^
where
c is
2
-ia>
{#(s)}
the prescribed angular
to contain
down the equation
at once write
if
=
F(>)
an additional
of energy in
c,
a constant.
Integrating again,
and
75
44, 45]
we have
^ [2c +
between
this relation
[g (s)}
&>
and
- 2 V(s)Y^ds + constant,
s represents the solution of
the problem.
Example 1. If the rotating tube is plane, and the particle can describe it with
constant velocity when the fixed axis is vertical and in the plane of the tube, and the
field of force is that due to gravity, shew that the tube must be in the form of a parabola
with
its
and vertex downwards.
axis vertical
Example 2.
particle moves under gravity in a circular tube of radius a which
rotates uniformly about a fixed vertical axis inclined at an angle a to its plane ; if 6 be
the angular distance of the particle from the lowest point of the circle, shew that
fa cos
V
where the function
is
2a
formed with the roots
|jf>
aco 2
aw 2 cos a
cos a
ty
and
t(]
is
(ii)
cos a
>
Off
~3g~>
a constant.
Tube moving with constant acceleration parallel
Consider
now
to
a fixed
direction.
the motion of a particle in a straight tube, inclined at an
angle a to the horizontal, which is constrained to

plane with constant horizontal acceleration/.
move
in its
own
vertical
Taking the axis of x horizontal and that of y vertically upwards, with the
we have for the kinetic energy
origin at the initial position of the particle,
r-i(?+A
= y cot a + ^ft
where
so
T=i(ycota+/0 +
= -ly cosec a + y cot a .ft + %f*t
2
,
is
r=w.
The equation
of motion
d_ (dT_\ =
dt (dy )~
_ dV_
dy
2
,
76
[CH. iv
gives therefore
,.
"T.
or
(y cosec-a
+jt
gfcota) sin
we have, supposing the
Integrating,
and therefore
= \V
cot a)
=
2
g,
a.
particle to be initially at rest,
g cos a +/sin a)
sin
a.
These equations constitute the solution of the problem it will be observed

that in this system the kinetic
energy involves the time explicitly, so no
of
exists.
integral
energy
:
Motion of two interacting free
46.
particles.
We
shall next consider the motion of two

particles, of masses m^ and ra 2
respectively, which are free to move in space under the influence of mutual
forces of attraction or
repulsion, acting in the line joining the particles and
dependent on their distance from each other.
The system has six degrees of freedom, since the three rectangular coordi
nates of either particle can have any values whatever.
We shall take, as the
six coordinates defining the position of the
system, the coordinates (X, Y, Z}
of the centre of gravity of the particles, referred to
any fixed axes, and the
coordinates
(x, y, z)
at the particle
of the particle m.2 referred to
and which are
The coordinates
of
1}
parallel
moving axes whose
^+
TO2
TWj
(x
The
/T7
referred to the fixed axes, are
m +m
1
kinetic energy of the system

/
T=
The
is
Z+-2V
2
therefore
TT
m.2 /
m,
m + mj
raj
+ mj
m,
or
(wj
+ w ) (X- + Y +
2
Z"-)
m,J
*\
2
J^L L
771
m^ m
^+ ^+ ^
potential energy of the system depends only on the position of the

,
particles relative to each other, so can be expressed in terms of (x,

it
is
referred to the fixed axes, are
Z
and those of
origin
the fixed axes.
.to
be V(x,
y, z}.
//,
z)
let
45-47]
l =
m m.
m +m
l
The
first
..
of motion of the system are
_ _
Z = Q,
r=0,
0,
8F
m^m^
dx
77
m-L 4-
..
dV
J
z
m-jn^
m +m
l
"by
dV
..
dz
three of these equations shew that the centre of gravity moves in

velocity, and the other three equations shew that
a straight line with uniform

the motion of m relative to
same as if
m-^ is the
attracted to m^ with the force derived
from
and
luere fixed
--
were
V*.
lit/]
Example. If two free particles move in space under any law of mutual attraction,
shew that the tangents to their paths meet an arbitrary fixed plane in two points, the
line joining which passes through a fixed
point.
(Mehmke.)
47.
Central forces in general
The
last article
Hamilton s theorem.
shews that the problem of two interacting free

particles
is reducible to the
problem of the motion of a single free particle acted on by
a force directed towards or from a fixed centre. This is known as the
problem
of central forces. There is clearly no loss of generality if we suppose the
mass of the particle to be unity.
If the particle be projected in any way, it will
always remain in the plane
which passes through the centre of force and the initial direction of
projec
tion for at no time does any force act to remove it from this
plane. We can
:
therefore define the position of the particle by

polar coordinates (r, 6) in this
Let
denote the acceleration
plane, the centre of force being the origin.
directed to the centre of force.

is
We
shall not
suppose for the present that
necessarily a function of r alone.
The
is
T=^(r + r*6
2
and the work done by the

(o>,
BO)
force in
2
),
an arbitrary infinitesimal displacement
is
-PSr.
The Lagrangian equations of motion
(r -rfc
The
latter equation gives
of the particle are therefore
=- P,
on integration
r2 6
= h,
where h
is
a constant
this is the integral corresponding to the

ignorable coordinate 6,
physically interpreted as the integral of angular
about the centre of
momentum
force.
*
Newton, Principia, Book
i.
Sect. 11.
and can be
of the particle
78
To
find the differential equation of the
called the orbit or trajectory},

the relation
we
path described (which
eliminate dt from the
first
is
[OH. iv
generally
equation by using
h d
~dt~~r2
d0
we thus obtain the equation

h d
2
dr\
fh2
h2
r dd (r d~6)
writing u for
or,
_
~
"
r*
1/r,
dh*
h*u2
This
the differential equation of the orbit *, in polar coordinates its

integration will introduce two new arbitrary constants in addition to the
constant h, and a fourth arbitrary constant will occur in the determination of
t
is
by the equation
t
2
T Ir d0
constant.
fi J
The
equation of the orbit in (r, p) coordinates, (where p

denotes the perpendicular from the centre of force on the tangent to the
differential
orbit), is often of
(
use
which (since h
18),
it
is
may be obtained directly from

now constant) gives at once
Siacci s
theorem
h?r
PY
2
dp
PD = -.-f
dr
ft
or
which
is
the differential equation of the orbit.
Since h = vp, where v
is
the velocity in the orbit,

v*
which
may
=P
we have from
this equation
p
-P,
r

"-iPfr
where q
is
the chord of curvature of the orbit through the centre of force.
We
frequently require to know the law of force which must act towards a
given point in order that a given curve may be described this is given at once
by the equation
;
if
the equation of the curve
tion
is
given in
(r,
is given in polar coordinates

while if the equa
p) coordinates, the force is given by the equation
2
p h dp
;
p dr
*
This
is
Theorie de la
substantially given in Newton s Principia, Book

and in the above form in Whewell s
(1765)
Lune
i.
and
Dynamics
3,
and in Clairaut
(1823).
47]
If the equation of the curve

ceed as follows
is
79
given in rectangular coordinates, we pro
Take the centre of force as origin, and let f(x, y) = be the equation of
the given curve. The equation of angular momentum is
xyyx = h.
Differentiating the equation of the curve,
+fy y = 0,
fx
From
these two equations
we have
where
fx
stands for ~-
ooo
we obtain
_
~
Differentiating again,
.dx
<M>
j.
dx
dy
we have
hfy
xfx + yfy
d_
(_hfy\ ___ hfx
"dx\xfx
_8
f__hfy_
xfx + yfy dy \xfx + yfy
~+yfy)
Performing the differentiations, this gives
_
But the required
(fy fxx + tfxfyfxy ~ fxfyy)

2
"?
force is P,
where x
_ n?T {jiffxx
oc
P-
^J^Jy
this equation gives the required central force.
The most important

f(x,
2/)
is
case
In this case
result
is
that in which the
2/O, y) = ax- + Wixy + by- + 2gx +2fy

we find at once that the expression
Jxxjy
has, for points
of this
curve
a conic,
on the
+ c = 0.
~ tfxyjxjy +fyyjx
conic, the constant value
- (abc + 2fgh - of - bg* - ch

2
),
while the quantity
9/
+a
x^cx yf_,
dy
has the value
-(gv+fy + c},
and so is a constant multiple of the perpendicular from the point (x, y} on the
thus obtain, for the force
polar of the origin with respect to the conic.
We
under which a given conic can be described, an elegant expression due to

Hamilton*, namely that the force acting on the particle in the position (x, y)
*
Proc. Roy. Irish Acad. 1846.
80
[CH. iv
varies directly as the radius from the centre

of force to the point (x, y\ and
inversely as the cube of the perpendicular from (a;, y} on the polar of the centre
of force.
The two following theorems, the proof of which
regarded as the converse of Hamilton s theorem.
is left
to the student,
may
together be
If a particle moves under the action of a force directed to a fixed

(i)
point, varying
directly as the distance from the fixed point and inversely as the cube of the distance
from a given straight line, the orbit is always a conic.
If a
(ii)
particle
moves under the action of a
force
directed to the origin,
of
magnitude
where (x, y} are rectangular coordinates and p, a, /3, y are constants, the orbits are conies
which touch the lines
ax 2 + Zftxy + yy~ = 0.
Darboux (Comptes Rendus, LXXXIV. p. 936) has shewn that these two laws of force are
the only laws for which the orbits are always conies, if the force
depends only on the
Suchar (Nouv. Ann. 1 vi. p. 532) has found other laws of
position of the particle.
force,
which involve the components of velocity of the
Example
1.
If a conic be described
shew that the periodic time
is
-~ pj?,
particle.
^ given by Hamilton
under the force
where
is
theorem,
the perpendicular from the centre of
<V>
the conic on the polar of the centre of force.
Example
2.
Shew that
if
(Glaisher.)
the force be
a particle will describe a conic having
its
asymptotes parallel to the lines
Ax + 2Hxy + By* = 0,
2
if
properly projected.
48.
circular
The
and
(Glaisher.)
integrable cases of central forces
problems soluble in terms of
elliptic functions.
The most important
case of motion under central forces
the magnitude of the force depends

only on the distance
force by f(r), the differential
equation of the orbit is
<$>.(,
d& +
Integrating,
is
r.
_f(r)
~
hW
we have
du
where
c is
a constant
integrating this equation again,
~
J
h2
fy*
~r*\
r*
we have
that in which
Denoting the
47, 48]
81
When r has been

this is the equation of the orbit in polar coordinates.
found from this equation in terms of 8, the time is given by the integral
and
=T
r2 dO
hj
+ constant.
The problem of motion under central forces is therefore always soluble by

quadratures when the force is a function of the distance only.
Example. Shew that the differential equations of motion of a point P are always
integrable by a simple quadrature when the central force F is of the form
F=
where
is
a function of 6 only, while a and b are arbitrary constants.
(Armellini.)
We shall now discuss the cases in which the quadrature can be effected
terms of known functions, the central force being supposed to vary as some
positive or negative integral power,
say the nth, of the distance.
in
Let us
first find
those problems for which the integration can be effected

The above integral for the determination of
in terms of circular functions.
where
a, b, c
are constants
the term in u~ n
~1
second degree
this gives
(a
+ bu* + 0M-?-
except when n
1
)"*
1,
when a logarithm
replaces
If the profolem is to be soluble in terms of circular func

tions, the polynomial under the radical in the integrand must be at most of the
;
-w-l=0,
and consequently
n
The
case n
the case n
quadratic
Next,
let
or
or
2,
-3.
however excluded by what has already been said, and

be added, since in this case the irrationality becomes
taken as a new variable.
1 is
1 is to
when w
= -l, -2,
I,
is
us find the cases in which the integration can be effected by the
aid of elliptic functions*.

For this it is necessary that the irrationality to be
should
be
of
the
third or fourth degree -f- in the variable with
integrated
respect to which the integration
= 0,
n =
3,
4, or
5,
5,
when u
But
this condition is fulfilled if
taken as the independent variable

when w is taken as the independent variable.
is
or
7,
It follows that the
problem of motion under a central force which varies as
nth power of the distance
the
taken.
is
is soluble
by circular or
elliptic
functions in the
cases
n=
o, 3, 1, 0,
2,
- 3, -
4,
- 5, - 7.
These cases were first investigated by Legendre, Theorie des Fonctions Elliptiques
(1825)
and afterwards by J. F. Stader, Crelle s Journal, XLVI. (1853), p. 262.
t Whittaker and Watson, Modern Analysis,
22-7.
W. D.
82
Example.
Shew
that the problem
following fractional values
The general
soluble by elliptic functions
is
[CH. iv
when n has the
=_n2
_A2
_i3
case of fractional values of n
is
_T
_sS)
discussed by Nobile, Giornale di Mat. XLVI.
(1908), p. 313.
The
cases in which motion under a central force varying as a

power of
is soluble
by means of circular functions are of special interest.
the distance
They correspond,
n=
as
shewn above,
to the values
2 will be considered in the next article
1,
2,
the cases n
3 of n
the case
= 1 and
can be treated in the following way.

(i)
tt-1.
In this case the attractive force
is
becomes
so the equation of the orbit
f*Y
I
M - \~^
~
tf
dv,
cv
fl
where
u?
v,
so
_
or
2 (6
j)
arccos
where 7
\2
is
a constant of integration,
=
This
is
the equation of an ellipse (when fi

0) or hyperbola (when
The orbits are therefore conies whose centre
>
p<
referred to its centre.
is
0)
at
the centre of force*.

(ii)
=-
3.
In this case the attractive force
so the equation of the orbit
Newton found that
if
a body
centre of the ellipse, the force
Prop. x.
is
is
becomes
move
in
an ellipse under the action of a force directed to the
directly proportional to the distance
Principia,
Book
i.
2,
48]
Integrating,
we have
(u
=A
cos (kd
+ e), where
k2 =
= J.
cosh (kd
+ e), where
&2
\u
83
= A6 +
where in each case
when
//,
1,
when
p>h
when
=h
^
h
=^
e,
and
<
h2
/A
2
,
These curves are sometimes known as Cotes
spirals;
the last
is
the
reciprocal spiral*.
In connexion with forces varying as the inverse cube of the distance,
that
it
may
be observed
if
be an orbit described under a central force
where k
is
P (r)
to the origin, then the orbit
any constant, can be described under a central
force P(r)
+ T-=,
3
where
c is
the intervals of time between corresponding points, i.e. points for which the
radius vector has the same value, in the two orbits being the same.
a constant
For,
if
accented letters refer to the second orbit,
If therefore
we have
we choose the new constant of momentum h so that
K = hk
this equation implies that the intervals of
orbits are the same, since
which establishes the
it
can be written
time between corresponding points in the two
This
result.
is
we have
),
I* /
sometimes known as Newton s theorem of revolving
orbits.
types of central motion corresponding to
kThe
lead, as
n=5,
3,
0,
4,
has been shewn, to elliptic integrals
5,
on inverting the integrals, we

As an example we shall
obtain the solution in terms of

elliptic functions.
take the case of n =
5.
*
i.
2,
Prop.
ix.
R. Cotes,
Harmonia Mensurarum,
pp. 31, 98.
62
84
[OH. iv
be the force towards the centre of attraction we shall suppose

the particle initially projected with a velocity less than that which would be
acquired by a fall from rest at an infinite distance to the point of projection,
6
Let
fj,u
so that the total energy
is
negative
call this
Then the equation
^7.
quantity
of energy
together with the equation
sdr\
2
<y
it
Introducing in place of r a new variable p defined by the equation

r
= (^ v*
the differential equation becomes
-^
{^
The
roots of the quadratic
when 7 is positive their sum is

and the smaller [of them is less
Hence if the greater and less of the roots be denoted by e and e3
we have the relations
respectively, and if e2 denotes
are real
than
i.
<?i
61
so
+e +
es
62
63
>
>
0,
where
roots
e is
BI,
Now
a constant of integration, and the function
e z , e$.
is
real
and
finite
we
positive, and, as
cannot be greater than

has a
gj
formed with the
is
Thus we have
lower limit
At
/<-.
So
but when
>
e1
see from the equation of energy,
(6
>
e2
e)
>
es ,
is
real
and positive and
the function
p (#
e) is real
48J
and has a
finite
lower limit for
so e is purely real,
take
and
be
to
zero.
this is the
real values of
all
only
and by measuring d from a suitable

We have therefore
when
85
e
real
is
we can
initial line
equation of the orbit in polar coordinates*.
The time can now be determined from the equation

h.
dd
u,
or
Performing the integration, we have
where
(6) is the Weierstrassian zeta-functionf.
Example
1.
Shew
This equation determines
that the equation of the orbit of a particle which

5
^./r can be written in the form
t.
moves under the
influence of a central attractive force
r = asn
Kv
=z
or else in the form
a
-=
r
snf
4
is the
provided A
0, where h is the angular momentum round the origin and
4p,E
excess of the total energy over the potential
at
energy
infinity.
(Cambridge Math. Tripos, Part I, 1894.)
>
>
shew
particle is attracted to the origin with constant acceleration p.
Example 2.
that the radius vector, vectorial angle, and time, are
given in terms of a real auxiliary
angle u by equations of the type
;
o) 2
- a)
-ô(<D!
a>
+ a)
(Schoute .)
Among the points of special interest on an orbit are the points at which
the radius vector, after having increased for some time,
begins to decrease
or after having decreased for some time,
to
increase.
begins
point
belonging to the former of these classes is called an apocentre, while points
of the latter class are called pericentres

*
The
orbits are discussed
and
classified
both classes are included under the
by W. D. MacMillan, Amer. Journ. Math. xxx.
(1908), p. 282.
t Cf. Whittaker and Watson, Modern Analysis,
20-4.
86
At an
general term apse.

(e.g.
[CH. iv
apse, if the apse is not a singularity of the orbit
a cusp), we have
which implies that the tangent to the orbit
is
perpendicular to the radius
vector.
The words aphelion and perihelion are generally used instead of apocentre
and pericentre when the centre of force is supposed to be the Sun.
Example.
moves under an attraction
particle
r
r2
to a fixed centre
apses
+ r3
shew that the angle subtended
at the centre of force
by two consecutive
is
where h
is
49.
the constant of angular
momentum.
Motion under the Newtonian law*.
The remaining
case in which motion under a central force varying as an

of the distance can be solved in terms of circular functions is
integral power
This
that in which the force varies as the inverse square of the distance.
case is of great importance in Celestial Mechanics, since the mutual attractions
of the heavenly bodies vary as the inverse squares of their distances apart, in
accordance with the Newtonian law of universal gravitation.
(i)
The
orbits.
Consider then the motion of a particle which is acted on by a force

directed to a fixed point (which we can take as the origin of coordinates), of
from the fixed point.
pu?, where u is the reciprocal of the distance
magnitude
Let the particle be projected from the point whose polar coordinates are
with velocity v in a direction making an angle 7 with c so that the
(c, a)
;
angular
momentum
is
The
= CV
is
integral
sin 7.
differential equation of the orbit is
dhi
this
a linear differential
equation with
constant coefficients, and
is
,.
v 2 c 2 sin 2
*
i.
3,
Props. XL, xn.,
xm.
its
48, 49]
where

and
CT
87

This is the equation, in polar
whose focus is at the origin, whose eccentricity is e,
coordinates, of a conic
and whose semi-latus rectum
the constant
CT
given by the equation
is
v 2c2 sin 2
7
;
determines the position of the apse-line, and
is
called the
perihelion-constant.
The circumstance that the focus
of the conic
is
at the centre of force
is
in accord with
Hamilton s theorem for if the centre of force is at the focus of the conic the perpen
dicular on the polar of the centre of force is the
perpendicular on the directrix, which is
proportional to r, as by Hamilton s theorem the force must be proportional to 1/r2
;
To determine the constants

we observe that initially
#=a,
and
terms of the
or in
du
=
-775
dO
initial
data
c, a,
7, v
cot
c
substituting these values in the equation of the orbit and the equation
obtained by differentiating it with respect to 6, we. have
2
v c sin
2
\v
7=
/ti
/j.e
cos (a
c sin 7 cos 7 = pe sin (a -
Solving these equations for

6
and
or),
or).
we obtain
CT,
v 4 c 2 sin 2
2v
c sin
~~u?
cot (a
CT)
cv
sin
\-
-.
tan
7 cos 7
7.
The semi-major axis, when the conic is an ellipse, is generally

mean distance of the particle denoting it by a, we have
called the
1-e
and substituting the values of
and
already found,
determines a in terms of the
The time occupied

which
is
we have
a/
v.c
this equation
initial data.
in describing the whole circumference of the

ellipse,
generally called the periodic time,
T x area of
is
ellipse,
88
since h represents twice the rate at which the area
vector
the periodic time
is
therefore
is
where
[CH. iv
swept out by the radius

b is the
semi-minor
axis.
But we have
v c sin
I
so the periodic time is 2?r
by n
is
\/
V
Cl
fi S
1
.
It is usual to denote the quantity
n?a~
.3
the periodic time can then be written
called the
It
*/
V
7 = V pi
mean motion, being the mean value
of 8 for a complete period.
has been shewn by Bertrand and Koenigs that of all laws of force which give a zero
an infinite distance, the Newtonian law is the only one for which all the orbits are
force at
algebraic curves, and also the only one for which all the orbits are closed curves.
Example.
Shew
that
a centre of force repels a particle with a force varying as the

is a branch of a hyperbola, described about its
if
inverse square of the distance, the orbit

outer focus.
(ii)
The
Consider
velocity.
now
the case in which the orbit
is
an
ellipse
the equation
between the mean distance a and the velocity
establishes a connexion
and
radius vector c at the initial point of the path.

Since any point of the orbit
can be taken as initial point, we can write this equation
where
v is the velocity of
Similarly
and
if
if
the orbit
the orbit
It is clear
according as v
is
the particle at the point whose radius vector
is
a hyperbola, whose semi-major axis
we
find
a parabola, the relation becomes
from this that the orbit

2
is a,
is r.
= 2a
,
i.e.
is
an
ellipse,
parabola, or hyperbola,
according as the initial velocity of the particle
is
than, equal to, or greater than, the velocity which the particle would
acquire in falling from a position of rest at an infinite distance from the
less
centre of force to the initial position.
It can further
be shewn that the
a component j perpendicular
to the
dicular to the axis of the conic
For
if
S be
velocity at
any point can
be resolved into
radius vector and a component

^r perpen
each of these components being constant.
P the position of the moving particle,

P to the conic with the major axis, GL
on SP from G, and SY the perpendicular on the tangent at
the centre of force,
the intersection of the normal at
the perpendicular
89
49]
from
SPG
known that the
are respectively
sides of the triangle
of
the
velocity in the
perpendicular to the velocity and to the components
S, it is
two specified directions; and therefore we have
SP = h.SP = h
= v p~
~y p~ ^j
.
Component perpendicular
to the radius vector
_fji
=
Sfit
and Component perpendicular
to the axis
= -~p
x Component perpendicular
to the radius vector
_=
Bfl
which establishes the result stated.

Example 1. Shew that in elliptic motion under Newton s law, the projections, on the
external bisector of two radii, of the velocities corresponding to these radii, are equal.
Shew also that the sum of the projections on the inner bisector is equal to the projection
of a line constant in magnitude and direction.
(Cailler.)
Example
where
the
mean
2.
Shew
that in elliptic motion under
Newton s
law, the quantity
Tdt,
denotes the kinetic energy, integrated over a complete period, depends only on
distance and not on the eccentricity.
(Grinwis.)
Example
constant p
3.
At a
certain point in an elliptic orbit described under a force
/i/r
the
suddenly changed by a small amount. If the eccentricities of the former and

new orbits are equal, shew that the point is an extremity of the minor axis.
(iii)
is
The anomalies in
elliptic motion.
If a particle is
describing an ellipse under a centre of force in the focus S,
the vectorial angle
of the point P at which the particle is situated on
the ellipse, measured from the
apse A which is nearer to the focus, is called
ASP
the true anomaly of the particle and will be denoted

by Q\ the eccentric
to
the
is
called
the eccentric anomaly of the
angle corresponding
point
and the quantity nt, where n is the

particle, and will be denoted by u
mean motion and t is the time of describing the arc AP, is called the mean
anomaly of the particle. We shall now find the connexion between the three
:
anomalies.
90
The
relation between
We
have
and u
is
found thus
an d
=a
where x
ex,
is
[CH. iv
e cos 6,
the rectangular coordinate of
referred
to the centre of the

ellipse as origin,
r= a(l-ecosu).
or
Hence
(1
-ecosu)
(1
e cos 6)
= 1 -e\
an equation which can also be written in the forms
/l-e\4
sin-*.
+
and
e cos
The
relation between
We
have
u and nt can be obtained
ft
rea
circle
= I\J\A/
^^ where Q
in the following
is
way
the P oint on tne auxiliary
corresponding to the point
[Area ACQ - Area SGQ}, where C
is
on the
ellipse
the centre of the
ellipse
a2
w
wa 2 f-jr
(2
nt= u e sin
80
This
is
known
a 2e
x-
sm wV
w.
as Kepler s equation.
for the solution of this

equation is described
The solution
(1907), p. 83.
iiomogram
Congres,
Keims
by H. Chretien, Assoc. Franf.

analytical expansion has been discussed by
being that by Levi-Civita, Atti delta R. Ace.
by
writers, an important recent memoir

del Lincei, Rendiconti,
(5) xin. (1904), p. 260.
many
Lastly, the relation between
We
have
nt
Replacing u by
nt
which
is
its
- e^ sin
(1
1 + e cos
the required relation
moving
as follows
= u-e sin u.
value in terms of
arcsm
vectorial angle of the
and nt can be found
0, this
0}
j
becomes
e(l- e^ sin
1
+ e cos
this equation gives the
time in terms of the
particle.
solution of the problem of

calculating the
based on a geometrical deduction, was found
True Anomaly from the Mean Anomaly,

the unpublished papers of Newton.
among
91
49]
Example
Shew
1.
that
*
7"
r=l T
where the symbols
denote Bessel coefficients*.
For we have
1
du
dt
\e cos u
d(nt}
<
2jr
- e cos u
2
,.
=1
cos rnt f 27r cos rnt
cosrnt
.j
du
+ ~2
TT
[2*
r=i
<S7r_/o
TT
= 1 + 22 Jr (re) cos r?z^ J
jo
d (nt)
ecosu
-,
cos r (w - e sin u)}
aw
jo
.
Y,
.,
by Fourier s theorem t
r=l
we have the required
Integrating,
Example
Shew
2.
result.
that
2
ê sin 2n
Example
and
In hyperbolic motion under the Newtonian law, shew that
3.
in parabolic motion,
where p
is
+ ....
shew that
the distance from the focus to the vertex.
Example
In elliptic motion under
4.
Newton
s law,
shew that the sum of the four
times (counted from perihelion) to the intersections of a circle with the ellipse is the same
for all concentric circles, and remains constant when the centre of the circle moves parallel
to the
major
(iv)
axis.
(Oekinghaus.)
Lambert s theorem.
Lambert
in
1731 shewed that in
elliptic
motion under the Newtonian
law, the time occupied in describing any arc depends only on the major axis,
the sum of the distances from the centre of force to the initial and final
points,
three
and the length of the chord joining these points so that if these
elements are given, the time is determinate, whatever be the form
:
of the ellipse
*
The name
of Bessel is
commonly connected with
this expansion
Lagrange, Oeuvres, in. p. 130.

t Of. Whittaker and Watson, Modern Analysis, Chapter
Ibid. Chapter xvn.
but
it is
really
due to
ix.
Lambert s original demonstration was geometrical and synthetic the theorem was proved
and generalised by Lagrange in 1778 (Oeuvres de Lagrange, iv. p. 559).
:
analytically
92
Let u and
[CH. iv
be the eccentric anomalies of the points then we have

n x the required time = u
e sin u
e sin u)
(u
iif
Now
(u
Ze
u)
sm
u
cos
u -+ u
.
"2
be the length of the chord, and r and r be the radii vectores,
if c
we have
+r =
and
e cos
a 2 (cos u
e cos
4a2 sin 2
^2V
u +- u
2e cos
% -
b 2 (sin
2
u
cos
-u
,
sin u} 2
-e
(l
e cos
(1
;=
i = 2 sin U
a
so
cos u) 2
cos 2
^-Y.
2
J
Hence we have
--
+r +
+cc = .
2
2 cos
(u
fw
and
+ r -c = 2
2 cos
-u
-w,
/
f
H arccos
e cos
u -u
---2
- + u \]
U-
h arccos
e cos
yj
and therefore*
r
2 arcsin 5
+ r + c\ = u - u
a
2\
o
2
v-,^
and
arcsm 51(r
Z
.
+r
Thus
.a =
sm
lfr
fl
quantities a and
if
-c\% =
I
/3
1-
arccos
e cos
+ arccos
^~
2
are denned
+ r +c\$
[e cos
u+u\
2
by the equations
1
/r
+r -
the last equations

give
a
Thus
finally
pO = u
0.
u,
and cos
=a=
is
Example
It will
(a
j3
-2
cos
- sin
=-- sin -~
2
a)
(/3
sin
/3).
Lambert s theorem.
1.
that the orbit

*
= e cos
we have
n x the required time
This
+-
Examine the
limiting case
when the minor axis
of the ellipse vanishes, so
is rectilinear.
be noticed that owing to the presence of the radicals, Lambert s theorem is not free
of sign. The reader will be able to determine without
difficulty the interpretation
from ambiguity
of sign corresponding to
any given position of the
initial
and
final points.
49, 50]
Example
2.
To obtain
the
form of Lambert s theorem applicable
to
parabolic motion.
we suppose the mean distance a to become large, the angles a and

small, so Lambert s theorem can be written in the approximate form
a3 - 83
If
Required
tirne
{(?
+ + c) *
?
6M *
is
become very
/3
=
and this
93
-(?
+ - c) *
?
},
the required form*.
Example 3. Establish Lambert s theorem for parabolic motion directly from the formulae
of parabolic motion.
The mutual transformation of
50.
fields of central force
and
fields of
parallel force.
If in the general problem of central forces we
suppose the centre of force
to be at a very great distance from the
part of the field considered, the lines
of action of the force in different
positions of the particle will be almost
parallel to each other; and on passing to the limiting case in which the
centre of force
is regarded as
being at an infinite distance, we arrive at the
of
the
motion
of
a
problem
particle under the influence of a force which is
to
a
fixed
direction.
always parallel
given
For the discussion of
this problem, take

rectangular axes Ox, Oy in the
plane of the motion, Ox being parallel to the direction of the force and let
(x) be the magnitude of the force, which will be
to be
;
supposed
of the coordinate y.
The equations
,
and the motion
is
of motion are
= X(x\
a,
b,
c,
0,
therefore represented by the

equations
x)
where
independent
dx
c }~4
are the constants of

integration;
determined by the circumstances of

projection,
i.e.
dx
I,
the values of these are
by the
initial
values of
x, y, x, y.
While the problem of motion in a

parallel field of force is a limiting case
problem of motion under central forces, it is not difficult to reduce the
latter more
general problem to the former more
one.
of the
special
For
if
a particle
is
in
motion under a force of

magnitude
directed to
This result was given by Euler in his Determinate Orbitae

Cometae Anni 1742
Lambert published the general theorem.
(1743), before
94
we may take
a fixed centre (which

of motion are
[CH. iv
as origin of coordinates), the equations
y-
The angular momentum
of the particle round the origin (which is con

let this be denoted by A.
Introduce new coordinates X, Y,
stant) is xy
yx
defined by the homographic transformation
:
X-?,
F-i,
and
let
be a new variable defined by the equation
*-/?
Then we have
W/Â.
-_
**s
VUV
\
1
dt\y)
dT
dF_^/l\
**
^^
jx
rfT
Tî
d2
I
/
U/
C}/-\
feU
||
Vi
1/Tfc
U*JU
dt^_ _y_
7^
^.2
y
f)
\y
tx
if
f>
tx
3/T
_ V
J/-7/
V
-^
/-"*
-*
These equations shew that a particle whose coordinates are (X, Y) would,
interpreted as the time, move as if acted on by a force parallel to
T were
the axis of
problem
P*
and of magnitude
^-
As the
will yield the solution of the original
general problem of motion under central forces

motion in a parallel field of force.
Example 1.
alone
is
Shew
solution of this transformed
problem,
is
it
follows that the
reducible to the problem of
that the path of a free particle moving under the influence of gravity
its axis vertical and vertex upwards.
a parabola with
Example
2.
Shew
the curve /(a?, y) =
that the magnitude of the force parallel to the axis of

is a constant multiple of
x under which
can be described
/a/\
-3
\dxj
_ ay /a/y
j / dxy v _ ay /8/yi
2
"
84%
dx* \dy)
ty
dy*
\dx}
If a parallel field of force is such that the path described by a free particle
is a conic whatever be the initial conditions, shew that the force varies as the inverse cube
Example 3.
of the distance from
51.
some
line perpendicular to the direction of the force.
Bonnet s theorem.
We
now proceed to discuss the motion of a particle which is simultaneously

by more than one centre of force. An indefinite number of particular
of motion of this kind can be obtained by means of a theorem due to
attracted
cases
Bonnet*, which may be stated thus

*
Liouville
Journal,
ix. (1844), p.
Mec. Anal. (Oeuvres de Lagrange, xn.
113 and Note

p. 353).
iv.
of
t.
n. of the last edition of Lagrange
50-52]
If a given
orbit
95
can be described in each of n given fields of force, taken
separately, the velocities at any point

of the orbit being v 1} v2 ..., vn
respectively, then the same orbit can be described in the field of force which
is obtained by superposing all these fields, the
velocity at the point
being
,
Or + v +
n )4
For suppose that in the field of force which is obtained by superposing

the original fields, an additional normal force R is required in order to make
the particle move on the curve in question and let it be projected from
;
a point
so that the square of its velocity at
is
equal to the
sum
of the
in the original fields of force.

Then on adding
squares of its velocities at
the equations of energy corresponding to the original motions, and comparing
with the equation of energy for the motion in question, we see that the
kinetic energy of the motion in question is the sum of the kinetic energies
of the original motions, i.e. that the velocity at any point
is
Hence, resolving along the normal to the
orbit,
we have
m is
where
and
F?,
the mass of the particle, p the radius of curvature of the orbit,

...,
n are the normal components of the original fields of force
at P.
and therefore
of force which
Example.
zero the given orbit is therefore a free path in the

obtained by superposing the original fields.
is
is
Shew
that
an
ellipse
field
can be described if forces
respectively act in the directions of the foci.
This result follows at once from Bonnet

forces are equivalent to forces
a force
-^x distance
and
~ acting
theorem when
it is
observed that the given
in the directions of the foci, together with
acting in the direction of the centre of the ellipse.
52.
Determination of the most general field of force under which a given
curve or family of curves can be described.
Let
(/>
be the equation of a curve
(x,
y)
on varying the constant
We
c,
this equation will
shall now find an expression for the

represent a family of curves.
most general field of force (the force being supposed to depend only on the
96
on
position of the particle
which
which
acts) for
it
[CH. iv
this family of curves is
a family of orbits of a particle.

v denote the velocity of the particle, and (X,
mass parallel to the coordinate axes.
Let
force per unit
1 dv^
normal components of acceleration being ^
Substituting for
its value,
we have
v^
and
-yCtS
O respectively,
we have
namely
4>y*<l>xx
v-
Writing
and replacing
j- by
2
((}>
X=U
Now
Y} the components of
The tangential and
(4>
+
X
= - u ($x +
*
(/y)~
x
<f>
~
^>yy
<j>y
<f)
X y)
</),
<j>
^}
+\$y jg
this equation
(<j>X
becomes
</>/)
depends on the velocity with which the given

and F are to be functions of the position, of
we can take u to be an arbitrary function of x and y we have
u is arbitrary, since it
orbits are described; and as
the particle,
therefore
X = U (QxQw ~
and similarly
Y=
4>y$xy)
U
(4>y<j>xx
<t>x<l>xy)
~
^(f>y
+ %$x
(<f>xUy
U>
(<&/
^y^x),
<t>
x U y ),
These expressions for the field

is an arbitrary function of x and y.
of force under which the curves of the given family are orbits were first given
by Dainelli*.
where u
Shew that a particle can describe a given curve uy\der any arbitrary forces
1.
directed to given fixed points, provided these forces satisfy the relations
Example
PI,
...
vohere r k is the radius
points,
and where
Pk
97\
as \
2
rk
and pk
is the
the perpendicular on the tangent, from the kth of the given fixed
radius of curvature of the given curve.
For the tangential and normal components of force on the particle are
T=-?Pk ~
fc
k
*
and
els
Giornale di Mat.
N= SP &,
r
k
k
xvm.
(1880), p. 271.
so
97
52, 53]
from the equation
we have
A particle
can describe a given curve under the single action of any one
Shew that the condition to
acting in given (variable) directions.
be satisfied in order that the same curve may be described under the joint action of forces
respectively, is
2
..., acting in the directions of $1, $ 2
F\,
Example
of the forces
2.
<i,
...,
>
^]-0u
j,
where
ck is
the chord of curvature of the curve in the direction of
Example
function
>
3.
point moves
that an equipotential curve
in a field of force in
V shew
is
equation
$t
a possible path, provided
is
(Curtis.)
two dimensions of which the work
V satisfy
the
wn
The problem of two centres of gravitation.

of motion of a particle moving in a plane under arbitrary
The most
forces cannot be integrated by quadratures in the general case.
famous of the known soluble problems of this class, other than problems of
53.
The equations
central motion, is the problem of two centres of gravitation, i.e. the problem
of determining the motion of a free particle in a plane, attracted by two fixed
Newtonian centres of force in the plane its integrability was discovered by
;
Euler*.
Let 2c denote the distance between the two centres of force
and take
the point midway between them as origin, and the line joining them as axis
of x, so that their coordinates can be taken to be (c, 0) and (
The
c, 0).
of
the
is
taken
to
be
therefore
mass
is
potential energy
particle (whose
unity)
F- - p (0 - cY +
where
and
yu,
p,
2/
}~*
ft
{(x
+ cr + f\
~*
are constants depending on the strength of the centres of
attraction.
Now
ellipse or
any
possible orbit
theorem
when
hyperbola with the two centres of force as
either centre of force acts alone,
a possible orbit
foci is
and therefore by Bonnet
a
s
when both
centres of force are acting.

It
is therefore natural, in
the
of
to
the particle,
position
replace the
defining
rectangular coordinates (x, y) by elliptic coordinates (, ?;), defined by the
equations
it
is
x=
cosh
cos
?;,
sinh
sin
77.
Euler, Mem. de Berlin, 1760, p. 228; Nov. Cornm. Pctrop. x. (1764), p. 207;
p. 152
Lagrange, M4m. de Turin, iv. (1706-9), pp. 118, 215, or Oeuvres, n. p. 67.
xi.
(1765),
w. D.
98
= Constant and
The equations
ellipses
and hyperbolas whose
[CH. iv
= Constant
then represent respectively

are at the centres of force; and these are
foci
77
a particular family of orbits.
The
when expressed
potential energy,
T is
and the kinetic energy
cos
(cosh
77)
and
in terms of
(cosh f
cos
becomes
77,
77)
This problem is evidently of Liouville s type ( 43), and can therefore

be integrated by the method applicable to this class of questions.
The
for the coordinate
Lagrangian equation
is
J
c
2
f cos 77)
{(cosh
-jCut
-c
cosh
sinh
i}
^ T T-
=)
(jc
or
c
or,
-j- {(cosh
- cos
r;)
2c 2 cosh
using the equation of energy

c
-cos 2 77)2
^- {(cosh
= - 2 (cosh
=2
- F)
(A
- cos
(cosh f
Og
2
r)
77)
(cosh
A (cosh 2 1
Integrating,
c
- +
2 (A
- cos
77)}
cos 2
77)
- F)
^ (cosh
cos 2
17)
+C
(cosh
cosh
00377)
+
C
cos
(cosh
77) [
we have
(cosh
is
(_
= 2-j-(h cosh
where 7
77)
>
"2
T + V = h,
Og
- cos
sinh
- cos
77)
A cosh 2
tt
"
+ u!
-
cosh f
7,
Subtracting this from the equation of energy, which can be written

|-
(cosh
- cos
=
(I
+ f)
h (cosh 2 f
cos 2
77)
77)
(cosh
+ cos 77) +
(cosh
cos
77),
99
53, 54]
we have
_
p2
Eliminating
cos 2
(cosh
cfa
2
77 )
?7
__
A cos 2
cos
+ 7.
77
between these equations, we have

it
Aicosh-ff-1-
/U/
,,.
cosh
h cos 2
f(
./
A,
/i
cos 2 ?;
U,
cos
77
+7
therefore
c,
~^
At
cos
77
elliptic integrals,
we have
w=N
u,
---- cosh
h cosh 2
-fJU
77
Introducing an auxiliary variable
These are
---
T;
dtj.
>
and we can therefore express
and
77
as elliptic
functions of the parameter u, say
= % (M
(>
),
17
=
<t>
O)-
These equations determine the orbit of the particle, the

7
?) being expressed in terms of the parameter u*.
Motion on a surface
54.
We
elliptic coordinates
f.
next proceed to consider the motion of a particle which

to move on a smooth surface, and is acted on by any forces.
shall
is free
Let (X, Y, Z) be the components, parallel to fixed rectangular axes, of

the external force on the particle, not including the pressure of the surface
let (x, y, z) be the coordinates and v the
velocity of the particle, s the arc and
:
p the radius of curvature of its path, ^ the angle between the principal
normal to the path and the normal to the surface, and (X, p, v) the directioncosines of the line
which
lies in
perpendicular to the path at time
the tangent-plane to the surface and is

the mass of the particle is taken as unity.
The
acceleration of the particle consists of components vdv/ds along the

2
tangent to the path and v /p along the principal normal the latter component
2
can be resolved into (v /p)sin^ along the line whose direction-cosines are
;
(X,
p.,
v)
and
(v /p) cos
^ along the normal
to the surface.
We
have therefore
the equations of motion
dv
r
ds
dz
dy
X-dx
r + Y ds
f- + z ds
ds
, ,
YfM
*
.................. (A),
+ Zv ........................... (B),
Some
generalisations of the problem of two fixed centres will be found in a paper by

Amer. Journ. Math, xxxni. (1911), p. 337.
t The earliest investigation of motion on a surface was Galileo s study of the motion of a
heavy particle on an inclined plane (Discourses, Third Dialogue, 1638). The motion of a heavy
particle moving in a horizontal circle on a sphere was examined by Huygens (Horolog. oscill.,
Hiltebeitel,
1673).
72
100
[CH. iv
and
these, together with the equation of the surface, are sufficient to determine
the motion for the equation of the surface may be regarded as giving z in
;
terms of x and
y,
and by using
this value for z
quantities occurring in equations (A)
we can
express
all
the
and (B)
equations (A) and (B) thus become a system

fourth order for the determination of x and y
in terms of x, y, x, y, x, y
of differential equations of the
in terms of
t.
If the forces are conservative, the expression
- Xdx - Ydy - Zdz

will be the differential of a
potential-energy function V(x, y, z); equation (A)
can therefore be integrated, and gives on
integration the equation of energy
$iP
where
in (B),
c is
a constant.
V(ae
y, z}
c,
2
Substituting the value of v given by this equation
we have
This
is
(on eliminating z by means of the equation to the surface) a

x and y, and is in fact the
differential equation of the second order between

differential equation of the orbits on the surface.
The differential equations of motion on a surface are not

integrable by
quadratures in the general case there are however two cases in which the
problem can be formulated in such a way as to utilise results obtained in
:
other connexions.
(i)
Motion under no forces.

i
When
the orbit
geodesic
no external forces act on the particle, equation (B) gives ^ = 0, so

* the
is a geodesic on the
surface
integral of energy shews that this
;
is
Example.
of striction
described with constant velocity.
is the
particle moves under no forces on the fixed smooth ruled surface whose line
axis of z, the direction-cosines of the generator at the
point z being
.
sin a cos
respectively.
To determine
sin a sin
cos
a,
the motion.
Let v denote the distance of the point on the surface whose coordinates are (x, y, z)
from the line of striction, measured along the generator, arid let (0, 0, f) be the coordinates
of the point in which this generator meets the line of striction.
Then we have
x= v sin a cos
This theorem
is
due
y = vsinasin
to Euler,
Mechanica
= +v cos a.
(1736), n. cap. 4.
54]
The
101
kinetic energy of the particle is
We can take v and f as the two coordinates which define the position of the particle
evident that the coordinate
is ignorable, and the
corresponding integral is
dT =
r
k,
it is
where k
is
a constant,
where h
is
a constant.
+ v cos a=k.
The
integral of energy
is
T=/i,
Eliminating f between these two integrals, we have
v 2 (v 2 +
m = 2hv + (2h - F) m
2
If v is initially sufficiently large
shall
compared with
suppose this to be the case, and shall write
2
(2h
f,
cosec 2 a.
the quantity (2A
cosec 2 a = 2AX 2
where X
F)
is
is
positive
new constant
the equation thus becomes
The
defined
integration of this equation can be effected

by the equation
by introducing a
real auxiliary variable M,
Writing v = \m.v~^, this becomes
and this
is
equivalent to the equation

*-f>()-ij
where the roots
ex
32
e 3 of
the function
e2 = X 2
The connexion between the variables

v
|p (M)
are real and are defined by the equations
= m2
ex
es
and u
is
e1
e.,
+ e3 = 0.
therefore expressed by the equation
= \m{^>(u)-e
}~^.
Substituting this value of v in the equation which connects v and
=
Cf.
e%u
f (M +
!)
t,
+ Constant.
20-33.
we have
102
This equation expresses the time

conjunction with the equation
gives the connexion between v
and
in
terms of the auxiliary variable
and thus
u,
in
t.
Motion on a developable surface.
(ii)
If the surface on
the
[CH. iv
known theorems
which the particle moves
is
we can
developable,
that the arc s and the quantity -
utilise
are unaltered
by
developing the surface on a plane: these results, applied to the equations of

motion given above, shew that if in the motion of a particle on a developable
surface under any forces the surface is developed on a plane, the particle will
describe the plane curve thus derived from its orbit with the same velocity as
before, provided the force acting in the plane-motion is the same in amount
and
direction relative to the curve as the component of force in the

tangentplane to the surface in the surface- motion.
Example
1.
whose axis
is
Prove that
the
is projected
along the surface of a right circular cone,
vertex upwards, icith the velocity due to the depth below the vertex.
traced out on the cone, when developed into a plane, icill be of the
form
smooth particle
vertical
path
and
r% sin | 6 = a*.
(Coll.
Exam. )
For on developing the cone, the problem becomes the same as that of motion in a plane
under a constant repulsive force from the origin, and with the velocity compatible with
We therefore have the integrals
rest at the origin.
r*
+ r2 2 = Cr,
r 2 = h,
where
C is
where h
is
a constant,
a constant.

1
Cr3
dr\*
= -g
so
if
say,
where a
is
a new constant,
u = -, we have
r
du\ 2
a3 w3
and therefore
*-(*
0=1
J
(l-a
(l
V2 )2
=
which
is
equivalent to the equation
sin
.
r * sin
where v = u-a?,
~1
v,
3
%6 = a?.
Example 2. If in the motion of a point P on a developable surface the tangent IP to

the edge of regression describes areas proportional to the times, shew that the component
of force perpendicular to
IP and
in the tangent-plane is proportional to
the radius of curvature of the edge of regression.
where p
(Hazzidakis.)
is
Motion on a surface of revolution
55.
and
103
54, 55]
elliptic
cases soluble in terms of circular
functions *.
The most important
case of surface-motion which
is
soluble
by quadra
the motion of a particle on a smooth surface of revolution, under

forces derivable from a potential-energy function which is symmetrical with
respect to the axis of revolution of the surface.
tures
is
Let the position of a point in space be denned by cylindrical coordinates

where z is a coordinate measured parallel to the axis of the surface,
(z, r,
is the
r is the perpendicular distance of the point from this axis, and
<),
<f>
The
azimuthal angle made by r with a fixed plane through the axis.

z and r, say
equation of the surface will be a relation between
r -/(*),
and the potential energy will be a function of z and r (it cannot involve
since it is symmetrical with respect to the axis), which for points on the
surface can, on replacing r by its value f(z\ be expressed as a function of z
can be taken as unity.
only, say V(z); the mass of the particle
<f>,
The
kinetic energy
The coordinate
is.
is
<f>
by
18,
evidently ignorable; the corresponding integral
k.
where k
is
a constant,
is
or
can be interpreted as the integral of angular

the axis of the surface.
this equation
The equation
and substituting
of energy is
for
(j>
T + V = h,
where A
in this equation from the preceding,
integrating this equation,
momentum
is
about
a constant,
we have
we have
i
Constant.
The
of r
relation
and
</>
between
and z
is
thus given by a quadrature
the values
are then obtained from the equation of the surface and the
equation
(/(*)}*-*,
respectively.
h
Jook
The motion
i.
of a particle on a surface of revolution
Section 10.
was investigated by Newton, Principia,
104
We
shall
now
[OH. iv
discuss the motion on those surfaces for which this
rature can be effected by means of known functions,

surface is vertical (z being measured
positively
when the
quad
axis of the
upwards) and gravity
is
the
only external force, so that
The circular cylinder.
(i)
When
the
surface
the circular cylinder r
is
= a,
the
above integral
becomes
t
and
the origin of coordinates
if
is
so chosen that 2ha-
= k we
2
have
* ==
or
to
^g(t
y>,
where
is
a constant.
is
a constant.
The equation
then gives
<o
</>
where
t ),
(t
The sphere.
The case in which the surface
<
(ii)
the sphere
is
called the
problem of the spherical pendulum*, and can be realised by

a
supposing
heavy particle attached to a fixed point by a light rigid wire
of
capable
moving freely about the point.
is
In this case the quadrature
for
becomes
~
t=l
or
The
integral
which we
integral,
by
*i,
on the right-hand side of this equation is an elliptic

shall now reduce to Weierstrass canonical form.
Denote
#2 Z3 the roots of the cubic

.
since the expression
(/i
- gz) (I* - z*)-k*
Lagrange, Mecavique Analytiqve. The complete solution in terms of (Jacobian) elliptic

functions was obtained by A. Tissot, Liouville s Journal, (1) xvn. (1852), p. 88 Jacobi s own
solution of the problem of a rotating rigid body in terms of elliptic functions had been
published
The analysis connected with the spherical pendulum is essentially that for
previously, in 1839.
:
Lame
equation of order
2.
55]
105
I of z, and
negative for the values I and
positive for very large positive
values of z and also for the values of z which occur in the
problem considered
is
I and + 1, since the

(which must necessarily lie between
particle is on the
we
see
that
one
of
the roots (say z^ is greater than I and the other
sphere)
two (say z2 and z3 where z2 zs ) are between I and
I.
The values of z in
>
the actual motion will
lie
between
and z3 since
,
for
them the cubic must
be positive.
Write
so that e 1} e2
e s are
new
6j
4-
relation
is
new
"
variable,
-1.2.3)
constants, which satisfy the relation

e.>
+e =
<
--
also satisfy the inequalities e l
The
where t
*-4 + f
->d
and
6g
between
>
e?
>
es
and z now becomes
or
where
roots
a constant of integration and the function
e is
e-i,
e2
e3
is
formed with the
g>
Now when e lt e2 e3 are real and in descending order of magnitude,

$(u) and @ (u) are both real when u is real, in which case $(u) is greater
than e lt and also when u is of the form
+ a real quantity, where 3 is
,
&>
&>..,
the half-period corresponding to the root es

in this latter case,
lies
between e2 and e 3
Since in the actual motion z lies between 2 and z3 it
follows that f lies between 2 and e3 and therefore the constant e must con
;
#>(t)
<?
of an imaginary part
3 and a real part depending on the instant from
which time is measured by a suitable choice of the origin of time, we can
take this real part of e to be zero, and we then have
sist
&>
*=^+7
?<
+<**
This equation gives the connexion between z and t.

determine the azimuth
For this we have the equation
<j>.
-,
"
7,
Icclt
dt
where
is
</>
We
have now to
106
To
t 4-
&>
are
we take X and
effect the integration,
I
corresponding to the values I and
constants defined by the equations
[CH. iv
to be the (imaginary) values of

of z respectively so that X and /*
/*
new
these equations give
>
integral
W=
21*
___
now becomes
(x)
The
~-
_
-l
dt
[
4J* J
{$>
~ kg
+w -
(t
3)
(X)}
{>
(t
+ 0,3) - 9 00}
dt
dt
a(j
$\t
>
(X) dt
00 dt
%>
-\
But* we have
(X)
and therefore
- t(\)\t
a
this equation expresses the angle
(t
o) 3
/JL)
a- (t
as a function of
t,
o>
and
</>
X)
so completes the
solution of the problem.
We
see that
when
increases
by
2a>
1(
fa increases
by
2^ (X -/*).
Example. When the bob of the spherical pendulum is executing periodic oscillations
between two parallels on the sphere, shew that one of the points reached on the higher
parallel, and the point on the lower parallel at which the bob arrives after a half-period,
have a difference of azimuth which always lies between one and two right angles.
(Puiseux and Halphen.)
The problem
of the spherical pendulum has been discussed from the standpoint of

by F. R. Moulton, Palermo Rend. xxxn. (1911), p. .338.
periodic solutions
(iii)
The paraboloid.
Consider next the problem of motion on the paraboloid, whose equation

r
In this case the quadrature

r
for
i
=
t
2a?
z%.
becomes
A*
~
\
t= (a+z)*
(2hz-2gz*-^-\
*
Cf.
dz.
% 20-53, Ex. 2.
is
55]
To obtain the
107
solution of the problem in terms of elliptic functions,

v, defined by the equation
we
introduce an auxiliary quantity

v
If a and
= ( \a + z) ~a
(where a
^ ft) denote
2hz
we can write
Define a
and
new
let e lf e 2 , e 3
respectively of z
and
it is
(~f)
variable
2gz*
- ^-}
dz.
the roots of the quadratic

2gz*
- k~ =
0,
4a
form
this integral in the
v=
(2hz
/*{4(*+)(*
by the equation
be the values of
corresponding to the values of -a,
then the integrals become
ft,
easily proved that the quantities ely e2

e1
+e +
2
The auxiliary quantity v can

V
es
el
0,
>
e.2
>
e3
now be replaced by an
j
-2
the relations
e 3 satisfy
auxiliary quantity u,
)*
}
u,
{ff (<*+*))
and then the inversion of the

integral gives
= p (w +
where
is
roots e 1} e2
As
p (u +
e),
a constant of integration and the function

e 3 which are
is
formed with the
g>
+ a + ft
3~(a + o)
2a
_ -ajf_a -2ft
3(a + a)
-a-2a + ft
3 (a
+ a)
in the actual
motion z evidently lies between a. and ft, it follows that

lies
between es and e2 and therefore (as we wish u to be real) the
e)
imaginary part of the constant e must be the half-period 3 the real part can
be taken to be zero, since it depends
merely on the lower limit of the
,
&>
integral
for u.
We
have therefore
-~ a-,
6
9
since
+ ft = 9
108
The equation
to determine
and
t is
(a
+ z)
dv
in terms of the auxiliary variable u.
this equation gives
Lastly, \ve have to determine the
(j)
4a
azimuth
for this
we have
kdt
kdt
[CH. TV
la (a
fr
a,,)
-e
-a+a+
v
+ o)J
du,
and therefore
4a
3 (a
a)
where
is
</>
a constant of integration, and
I is
an auxiliary constant denned by
the equation
j
ft>
(I)
--
3(
so
The equation can now be written

,
ku,
"
a
the integral of
to be
which
i(0
found (as in the problem of the spherical pendulum)
is
=e
<T
this equation expresses

the solution.
(iv)
The
in
^>
cone.
r
is
0) 3
terms of the auxiliary variable
Consider next the cone, whose equation
where a
the semi-vertical angle.
z tan
is
o,
u,
and so completes
55, 56]
Since this
is
a developable surface,
we can apply the theorem
109
of
54,
and
see that the orbit of a particle on the cone under gravity becomes, when
the cone is developed on a plane, the same as the orbit of a particle of unit
we
mass in the plane under a force of constant magnitude g cos a acting towards
a fixed centre of force (namely the point on the plane which corresponds to
the vertex of the cone).
This ( 48) is one of the known cases in which the
problem of central motion can be solved in terms of
this solution furnishes at once the solution of the
elliptic functions,
and
problem of motion on the
cone.
Example
whose axis
is
Shew
1.
that the motion of a particle under gravity on a surface of revolution

terms of elliptic functions when the surface is
vertical can also be solved in
given by any one of the following equations

3a)
z(z
(r
2
,
a2 ) 2 = a 3z.
az
(Kobb and
Example 2. Shew that the same problem can be solved

when the surface is
2 3
2
6
(a* +3/ ) + 2a = 8a% (x* -f y ).
Shew that
in
terms of
Stackel.)
elliptic
functions
(Salkowski. )
an algebraic surface of revolution is such that the equations

of its geodesies can be expressed in terms of elliptic functions of a parameter, the surface
must be such that r2 and z can be expressed as rational functions of a parameter, i.e. the
2
equation of the surface regarded as an equation between r and z is the equation of
a unicursal curve
are the cylindrical coordinates of a point oil the
where z, r,
Example
3.
if
surface.
(Kobb.)
Example
Shew that
4.
in the following cases of the
of revolution, the trajectories are all closed curves
1.
When
the surface
is
2.
When
the surface
is
motion of a particle on a surface
a sphere, and the force is directed along the tangent to the

meridian and proportional to cosec 2 #, where 6 is the angular distance from the particle to
the pole. (The trajectories are in this case sphero-conics having one focus in the pole.)
a sphere, and the force
meridian and proportional to tan 6 sec 2

having the pole as centre*.)
6.
directed along the tangent to the

(The trajectories in this case are sphero-conics
is
Joukovskys theorem.
We shall now shew how to determine the potential-energy function under
which a given family of curves on a surface can be described as the orbits of
56.
a particle constrained to
move on the
surface.
The three rectangular coordinates
of a point on the surface can be expressed

two parameters, say u and v, so that an element of arc ds on the
surface is given in terms of the increments of u and v to which it
corresponds
by an equation of the form
in terms of
ds 2
where E, F,
*
are
known
Darboux has examined the
(1877).
= Edu + ZFdudv +
2
functions of u and
Gdv*,
v.
possibility of other cases, in Bull, de la Soc.
Math, de France,
v.
110
[OH. iv
Let the family of curves which are to be the orbits under the required system
of forces be defined by an equation
= Constant,
q (u, v)
and
let
p (u,
v)
denote the family of curves which
Constant
orthogonal to these.
is
Then instead of u and v we can take p and q as the two parameters

which define the position of a point on the surface let the line-element
in this system of parameters be expressed by the equation
;
ds*
=E
df+G dp
z
,
dqdp being absent, because the curves p = Constant and

and G being known functions of
Constant cut at right angles
the term in
q
and
q.
The
kinetic energy of a particle which
moves on the surface
is
the Lagrangian equations of motion are therefore
i-T^cto
where
(G p}
[-5
V
V denotes the unknown
<f
8p
op
op
potential-energy function, which
it is
required
to determine.
These equations are to be
satisfied
2
1
Eliminating
_p
by the value q =
they then become
P=
~dq~
85"
9//v^
we have
dq
,,;
dq
we have
+ V =f(q\ where f is
an arbitrary function,

or
111
^-
and therefore
where g denotes an arbitrary function.
Now
-~~
AI (P\ the differential parameter* of the
first
order of the
function p; and thus we have a theorem enunciated by Joukovsky in 1890,

that if q = Constant is the equation of a family of curves on a surface, and
p=
Constant denotes the family of curves orthogonal to these, then the curves
Constant can be freely described by a particle under the influence of forces
q
derived from the potential-energy function
V=*
where
(p)g(p) + &
(p)ff(q)^\^--ldq,
and g are arbitrary functions, and A! denotes
the first differential
parameter.
The above equations give
dq~
dq
and hence the equation of energy in the motion
is
/
1.
particle
moves under gravity on the smooth

s
where
denotes the arc and
the angle
<f>
horizontal
shew that the motion
is
cycloid
whose equation
is
= 4a sin $,
made by
the tangent to the curve with the
periodic, the period being
4v
*/
t/
2.
particle moves in a smooth circular tube under the influence of a force directed
to a fixed point and proportional to the distance from the point.
Shew that the motion is
of the same character as in the pendulum-problem.
If the line-element
on a surface
is
d2 = E du 2 + 2Fdu dv + G dv*,
the
first differential
The
differential
variables
parameter
is
<f>
(u, v) is
).
given by the formula
a deformation-covariant of the surface, i.e. when a change of

the differential parameter transforms into the expression
variables (u v ) and the corresponding new coefficients
made from (u, v) to (u v ),

the same way with the new
is
formed in
parameter of a function
112
[CH.
3.
particle moves in a straight line under the action of two centres of repulsive
Shew that,
force of equal strength p., each varying as the inverse square of the distance.
if the centres of force are at a distance 2c apart and the particle starts from rest at
a distance
kc,
where k
<
1,
from the middle point of the
line joining
it will
them,
perform
oscillations of period
IT
Jo
(1
-F
sin 2
6^
dQ.

4.
I,
1899.)
under the action of gravity travels in a smooth curved tube, starting
particle
of the tube.
from rest at a given point
If the particle describes every arc OP in
the same time that would be taken to slide down the corresponding chord OP, shew that
the tube has the form of a lemniscate.
downwards along the concave side of the curve y 3 + ax2 =

with a velocity f (2#)2 from the origin, the axis of x being horizontal
shew that the
vertical component of the velocity is constant.
(Nicomedi.)
A particle moves in a smooth tube in the form of the curve r2 = 2a 2 cos 2$, under
6.
5.
particle is projected
the action of two attractive forces, varying inversely as the cube of the distance, towards
the two points on the initial line which are at a distance a from the pole. Prove that if
the absolute force
of
thecurveis
is
p.,
and the velocity at the node 2/^ /a, the time of describing one loop
7ra /2/ii.
I,
1898.)
particle describes a space-curve under the influence of a force whose direction

always intersects a given straight line. Shew that its velocity is inversely proportional
to the distance of the particle from the line and to the cosine of the angle which the
7.
plane through the particle and the line makes with the normal plane to the orbit.
(Dainelli.)
heavy particle is constrained to move on a straight line, which is made to

rotate with constant angular velocity co round a fixed vertical axis at given distance from
Shew that the motion is given by the equation
it.
8.
where
by the
from a fixed point on the

and A, B are constants.
r is the distance of the particle

line with the horizontal,
line, a is
the angle
(H.
am
made
Ende.)
9.
heavy particle is constrained to move on a straight line, which is made to
rotate with given variable angular velocity round a fixed horizontal axis.
Shew that the
equation of motion is
2
2
sin a sin B
r8 sin a + ad sin a,
r=+g
the angle between the line and the axis of rotation, 6 the angle made with
the vertical by the shortest distance a between the lines, and r the distance of the
where a
is
particle
from the intersection of this shortest distance with the moving
line.
(Vollhering.)
particle slides in a smooth straight tube which is made to rotate with uniform
about a vertical axis
shew that, if the particle starts from relative
angular velocity
rest from the point where the shortest distance between the axis and the tube meets the
10.
o>
tube, the distance through which the particle
moves along the tube
in
time
t is
2(7
2
-^ cot a cosec a sinh (^
where a
is
u>t
sin a),
the inclination of the tube to the vertical.
I,
1899.)
iv]
113
particle is constrained to move under no external forces in a plane circular tube

constrained to rotate uniformly about any point in its plane. Shew that the
motion of the particle in the tube is similar to that in the pendulum-problem.
11.
which
is
12.
small bead
is
strung upon a smooth circular wire of radius
a,
which
is
con
strained to rotate with uniform angular velocity

about a point on itself. The bead is
initially at the extremity of the diameter through the centre of rotation, and is projected
o>
with velocity 2& relative to the wire

shew that the position of the bead at time
is
= sn
sin
<f)
(modulus a/b)
bu>t\a
or
isin
= (bja) sn
(f)
(modulus
cot,
b/a)
or
according as a
b, $ being the angle which the radius vector to the bead makes
with the diameter of the circle through the centre of rotation.
>
<
Shew
13.
.r
+^ 3 = a 3
proportional to
is
xy (x
1900.)
that the force perpendicular to the asymptote under which the curve
can be described
14.
I,
particle is acted on
+y
}~
3
.
by a force whose components (X, Y} parallel to fixed axes

Shew that the problem of its motion is
(x, y).
are conjugate functions of the coordinates

always soluble by quadratures.
15.
If ((7) be a closed orbit described
by a particle under the action of a central force,
the centre of force,
the centre of gravity of the curve (}, G the centre of
gravity of
the curve (C) on the
supposition that the density at each point varies inversely as
the velocity, shew that the
points S, 0, G are collinear and that 2<7 = 3SO.
(Laisant.)
16.
Shew that the motion of a particle which is constrained to move in a
plane,
under a constant force directed to a
point out of the plane, can be expressed by means
of elliptic functions.
17.
Shew
that the curves
where a, b, c are arbitrary constants and

/is a given function, can be described under the
same law of central force to the
origin.
18.
Shew
a point in
19.
its
that
when a
circle is described
circumference, the law of force
is
under a central attraction directed to
the inverse
fifth
power of the distance.
particle describes the pedal of a circle, taken with

respect to any point in
plane, under the influence of a centre of force at this
Shew that the law
point.
of force is of the form
its
where
A and B
are constants.
Shew that the law

respect to a focus
is
of force is also of this form when the inverse of an

ellipse with
described under a centre of force in the focus.
(Curtis.)
w. D.
114
[en.
Prove that, if when projected from r=R, 6 = with a velocity V in a direction

an
making
angle a with the radius vector the path of a particle be/(r, 0, R, F, sina) = 0,
the path with the same initial conditions but under the action of an additional central
20.
force
^-.
is
f(r,nd,R,
where
(Coll-
21.
origin
Exam.)
A particle of unit mass describes an orbit under an attractive force

and a transverse force T perpendicular to the radius vector. Prove
to the
that,
the
differential equation of the orbit is given

<Pu
by
T du
~~
P
If the attractive force is
of angle
always
and the
zero,
<M?_
d6~
dd>
moves
particle
in
an equiangular spiral
prove that
a,
^,.2860*0-3 and
A = (^ sin
cosa)^
sec2a
.
I,
1901.)
22.
varying as the distance,
particle, acted on by a central force towards a point
so as to pass through a point Q such that OP is equal to OQ
projected from a point
shew that the least possible velocity of projection is OP(^s,mPOQ}^ where p. OP is the
force per unit mass at P.
(Camb. Math. Tripos, Part I, 1901.)
is
23.
Find a plane curve such that the curve and its pedal, with regard to some point
in the plane, can be simultaneously described by particles under central forces to that
point, in such a manner that the moving particles are always at corresponding points
of the curve and the pedal
and
find the
law of force
for the pedal curve.

24.
If /(#, y) be a
I,
1897.)
homogeneous function of one dimension, then the necessary
and
l be capable of description under accele

sufficient condition that the curve f(x,y)
ration tending to the origin and varying with the distance alone, is that f be subject
to a condition of the form
Hence shew that the only curves

r (A
Proceed
to
of this class are necessarily included in the equation
+ B sin 6 + C cos
ff)
=1
the discussion of the case wherein f(x, y)
is
dimensions.
25.
An
at a point
ellipse of centre
is
described under the influence of a centre of force
on the major axis of the

nt
where
homogeneous and of n
(Coll. Exam.)
ellipse
esin
%Tr/n is the periodic time, e is the ratio of
shew that
u,
CO
to the semi-major axis,

t from the vertex.
and u
is
the
eccentric angle of the point reached by the particle in time

26.
Two
free particles
p.
and
M move in a plane under the influence of
a central force
Shew that the
ratio of the velocity of the particle p at an arbitrary

point TO of its path, to the velocity which is possessed at TO by the central projection of
on the orbit of p., is equal to the constant ratio of the areas described in unit time by the
to a fixed point 0.
radii 0/x,
OM, multiplied by the square

ratio of OM, Om.
which expresses the
of a certain function
of the coordinates of
(Dainelli.)
TO,
115
iv]
under an attraction to the focus. Shew

particle is moving freely in a parabola
at every instant a point be taken on the tangent through the particle, at distance
from the particle, this point will describe a central orbit about
4acos0/(<9 + sin<9)
27.
that, if
where
the focus, and the rate of description of areas will be the same as in the parabola
4a is the latus rectum, and 6 the vertical angle of the particle measured from the apse
line.
;
When
a periodic comet
receives a small increment 8v.
28.
will
is
at its greatest distance from the sun, its velocity

that the comet s least distance from the sun
Shew
be increased by the quantity

48v.
3
{
(l-e)//x(l
Exam.)
(Coll.
+<?)}-.
If POP is a focal chord of an elliptic path described round the sun, shew that
29.
to P through perihelion is equal to the time of falling towards the
the time from
sun from a distance 2a to a distance a (1+ cos a), where a = Zir - (uf - u\ and u -uis the
difference of the eccentric anomalies of the points P,
(Cayley.)
3
moves in a plane under attractive
^/rV /z/rV along the
if it is projected
Shew
2d
distance
at
two
fixed
that,
apart.
r,
points
with the velocity due to a fall from rest at infinity, a possible path is a circle with regard
to which the two fixed centres are inverse points, and that, if the radius of this circle is a,
30.
radii
forces
particle
to
drawn
the periodic time
is
Exam.)
(Coll.
31.
A heavy particle is projected horizontally with a velocity v inside a smooth
sphere at an angular distance a from the vertical diameter drawn downwards shew that
it will never fall below or never rise above its initial level according as
:
v2
>
or
<ag
sin a tan
a.
(Coll.
Exam.)
A particle is projected horizontally with velocity V along the interior of a smooth

32.
sphere of radius a from a point whose angular distance from the lowest point is a. Shew
that the highest point of the spherical surface attained is at an angular distance /3 from
the lowest point, where /3 is the smaller of the values of ^, x given respectively by
the equations
2=
(3 cos -^-2 cos a) # + F
0j
,,
ivXMi. Hixam.j
9
= nf
V*
sin*
(cos
x + cos
2ag
a)
OJ
motion of a spherical pendulum of length a be wholly between the levels

ia
below
the
fa,
point of support, shew that at a time t after passing a point of greatest
depth, the depth of the bob is
If the
33.
fca{4-sn
(mod. V(?/65))
V(130/14a)}
and that a horizontal coordinate referred to the point of support as origin

by the equation
which
is
34.
a case of
Shew that
Lame s
if
equation.
a conical pendulum
projection of the bob describes an ellipse

with the angular velocity
is
(Coll.
is
determined
Exam.)
executing small oscillations, the horizontal

in the sense of the motion
whose axes turn
H
where $
I
i s the
angle of greatest deviation from the vertical,
the length of the pendulum, and g gravity.
that of least deviation,

(Resal.)
3
116
35.
particle is constrained to
fixed point
move on the
surface of a sphere, and
M on the surface of the sphere with a force that varies as r~
is
[CH. iv
attracted to a
r2 )
(o?
% where
the diameter of the sphere and r is the rectilinear distance from the particle to M.
If the position of the particle on the sphere be defined by its colatitude 6 and longitude $,
is
with
M as pole, shew that the equations of motion furnish the differential equation
where a and
b are constants
and integrate
this equation,
shewing that the orbit
is
a sphero-conic.
36.
particle of
semi-vertical angle is
angular
is
momentum
mass m moves on the inner surface of a cone of revolution whose

3
from the axis the
a, under the action of a repulsive force mp/r
;
of the particle
about the axis being
an arc of a hyperbola whose eccentricity
is
sec
-Jp tan
a,
shew that the path
a.

37.
Shew that the necessary
that the path on the cone
may
1897.)
I,
central force to the vertex of a circular cone in order
be a plane section
is
4-^.
..1
(Coll.
*
ifO
Exam.)
A particle of unit mass moves on the inner surface of a paraboloid of revolution,

rectum 4a, under the action of a repulsive force pr from the axis, where r is the
distance from the axis
shew that, if the particle is projected along the surface in a
38.
latus
direction perpendicular to the axis with velocity 2a/*2
it will
describe a parabola.
(Coll.
39.
about
A
its
Exam.)
formed by rotating the catenary s = cta,n<p

axis of symmetry, and a particle is projected along its surface from a
point
smooth surface
of revolution
is
distant b from that axis with velocity h (a 2 + 6 2 )^ /fe2

The direction of projection is such
that the component velocity perpendicular to the axis is h/b and the particle moves in
.
contact with the surface, under the influence of a force of attraction A 2
(r
+ 2a 2 )/r5
in the
Shew that, if gravity be neglected, the

direction of the perpendicular r to the axis.
projection of the path on a plane at right angles to the axis will have a polar equation
r
csmh- = a#.
40.
particle
moves on a smooth
helicoid,
z=a(f>,
(Coll.
Exam.)
under the action of a force pr
per unit mass directed at each point along the generator inwards, r being the distance
from the axis of z. The particle is projected along the surface perpendicularly to the
generator at a point where the tangent plane
Shew
velocity of projection being p*a.

the plane of xy is
2
/r
makes an angle a with the plane
that the equation of the projection of
= sec 2 a cosh 2
((f>
cos a)
Shew that the problem
42.
its
path on
1.

41.
of xy, its
I,
1896.)
motion of a particle under no forces on a ruled

surface, whose generators cut the line of striction at a constant angle, and for which the
ratio of the length of the common perpendicular to two adjacent generators to the angle
between these generators is constant, can be solved by quadratures.
(Astor.)
particle (x, y,
z}
of the
whose potential energy is (ax2 + by 2 + cz2 )

l.
Determine the motion.
(C. Neumann, Journal fur Math. LVI.
is
constrained to
move on the sphere x2 +y 2 + z 2
(1859), p. 46.)
CHAPTER V
THE DYNAMICAL SPECIFICATION OF BODIES
57.
Definitions.
Before proceeding to discuss those problems in the dynamics of rigid

bodies which can be solved by quadratures, it is convenient to introduce and
number of constants which can be assigned to a rigid body, and

which depend on its constitution it will be found that these constants
determine the dynamical behaviour of the body.
Let any rigid body be considered and let the particles of which (from
the dynamical point of view) it is constituted be typified by a particle of
calculate a
mass
axes are
situated at a point whose coordinates referred to fixed rectangular

(x, y, z).
2m (y + z
The quantity
2
),
where the symbol 2 denotes a summation extended over all the particles of
the system, is called the moment of inertia of the body about the axis Ox*.
Similarly the moment of inertia about any other line is defined to be the sum
of the masses of the particles of the body, each multiplied by the square of
These summations are evidently in
its perpendicular distance from the line.
the case of ordinary rigid bodies equivalent to integrations
f
is
equivalent to
2
I
(y
z2 )
pdxdydz, where p
unit volume, of the body at the point
The quantity
is
thus
2m (y + z )
2
ff
(x,
y,
is
the density, or mass per
z).
^mxy
body about the axes Ox, Oy and
and ^nizx are the products of inertia about
called the product of inertia of the
similarly the quantities

the other pairs of axes.
my
For the moments and products of inertia with reference to the coordinate
axes, the notation
A=2m(y + z
F = Zmyz,
2
will
B=Zm(z* + x
G = ILmzx,
),
G = 2m (x +
H = ILmxy
2
),
2
),
be generally used.
Two
bodies whose
moments
of inertia about every line in space are equal

be equimomental. It will be seen later that this
involves also the equality of the products of inertia of the bodies with respect
to each other are said to
to
any pair of orthogonal

*
Moments
(Horolog.
lines.
of inertia were first introduced by
oscill.,
1673).
The name
is
Huygens
due to Euler.
in his researches
on the pendulum
The dynamical
118
specification
of bodies
[CH.
M denotes the
mass of a body and if k is a quantity such that Mk 2 is

equal to the moment of inertia of the body about a given line, the quantity
k is called the radius of gyration of the body about the line.
If
moment
In the case of a plane body, the

dicular to its plane
is
often spoken of as the
of inertia about a line perpen

moment of inertia about the
point in which this line meets the plane.
The moments of inertia of some simple bodies*.
58.
The rectangle.
(i)
be required to find the moment of inertia of a uniform rectangular

whose sides are of lengths 2a and 26 respectively, about a line through
Let
plate,
its
it
parallel to the sides of length 2a.

Taking this line as axis Ox,
other
as axis Oy, the required
to
the
sides
parallel
through
centre
and a
line
moment
of inertia
is
or
where
cr is
the mass per unit area of the plate, or the "surface-density as it is

evaluating the integral, we have for the required moment
frequently called
of inertia
3
|<ra6
or
Mass of rectangle x ^6 2
The moment
of inertia of a uniform rod, about a line through its middle

point perpendicular to the rod, can be deduced from this result by regarding
the rod as the limiting form of a rectangle in which the length of one pair
of sides is
question
It follows that
small.
indefinitely
Mass of rod x |6 2
where 26
(ii)
26,
the
moment
of inertia in
is
is
the length of the rod.
The rectangular
block.
Consider next a uniform rectangular block whose edges are of lengths 2a,
let it be required to find the moment of inertia about an axis Ox
2c
;
and
passing through the centre
moment
of inertia
parallel to the edges of length 2a.
ra
Zm(y + ^
2
where p
is
This
is
the density.
),
rb
re
n
or
-aJ -b- -c
p(y
+ z }dzdydx,
2
Evaluating the integral, we have for the
moment
of
inertia
c
^L
(fca
+ c^
or
Mass of block x 1
(6
).
*
For practical purposes the moments of inertia of a body are determined experimentally
convenient apparatus is described by W. H. Derriman, Phil. Mag. v. (1903), p. 648, and by
W. R. Cassie, Phys. Soc. Proc. xxi. (1909), p. 497.
The dynamical
57, 58]
The
(iii)
Let
plate
it
ellipse
and
the circle.
now be required
whose equation
about the axis of

ra
I
to find the
moment
of inertia of a uniform elliptic
is
It is
x.
119
of bodies
specification
ra
a-y^dydx, where
the surface-density.
cr is
Evaluating the integral, we have for the required moment of inertia

z
or
Mass of ellipse x 6 2
\7rab ar,
.
The moment
of inertia of a circle of radius b about a diameter
Mass of
therefore
The ellipsoid and the sphere.
(iv)
The moment
equation
of inertia of a uniform solid ellipsoid of density
\p(y*
To evaluate
71
is
about the axis of x
where
x \6
circle
is
77,
is
T2
ift
?%
/T^
tv
L/
fc
r
O
p,
whose
similarly
+ z^dxdydz,
integrated throughout the ellipsoid.
this integral, write
are
new
pab
variables
3
c III
is
the integral becomes
2
?
now taken throughout
a sphere whose equation
is
Since the integrals
and
^,
are evidently equal, the required
moment
of inertia can be written in the
form
or
or
or
Trpabc
(fe
c ) [
J -i
| (1
2
-^Trpabc (b
Mass of
i
ellipsoid x
- f)
2
),
2
(6
+c
2
).
120
Tlie
The moment
is
dynamical
of bodies
specification
of inertia of a uniform sphere of radius
[CH.
a about a diameter
therefore
Mass
(v)
The
Let
it
of sphere x
â
2
.
triangle.
now be
required to find the
moment
of inertia of a
uniform
triangular plate of surface-density a, with respect to any line in its plane

the position of the line can be specified by the lengths a, ft, y of the per
pendiculars drawn to
from the vertices of the triangle.
it
to be the areal coordinates of a point of the plate, the

distance
from this point to the given line is (ax + /3y + yz), and
perpendicular
the required moment of inertia is therefore
(x, y, z}
Taking
ax
4-
fiy 4- yz) dS,
where dS denotes an element of area of the
Now
on the side
side c
plate.
Y denotes
if
the length of the perpendicular from the point (x, y, z)

of the triangle, and if
denotes the length intercepted on the
between the vertex
and the
foot of this perpendicular,
we have
Y = zb sin A
X sin A
and
Fcos
A = perpendicular
= yc
We
the form
= 387 (| -4- dXdY =

x
d(X, Y)
2A
is
A dXdY= 2A
-.
be sin
Hence the
dS,
integral \\y-dS,
where
extended over the area of the triangle, can be written in
lly-dydz, where the integration
values of y and z whose
sum
is
less
By symmetry,
the integrals
extended over
is
than unity
this
la^dS and
each have the value TV A.
ffxedS,
positive
equal to
z-dS have the same value,
and a similar calculation shews that the integrals
ffyzdS,
is
all
rr
r r
A.
on the side b
A.
denotes the area of the triangle.
the integration
or
(x, y, z)
have therefore
dydz
where
sin
from
\\xydS
The dynamical
58, 59]
Substituting these
moment
values
in
the
(ax
+ /3y + yz^dS,
the
of inertia of the triangle about the given line becomes

2
<rA
(a
+ @- + f +
i x Mass of
triangle x
+ 7 + a/3),
+ (?-?Y + f*
J(*?Y
2
2
\
(\
/
But
cr
integral
121
of bodies
specification
+ *Yl
moment
this expression evidently represents the
of inertia about the
line of three particles situated

respectively at the
middle points of the

mass of each particle being one-third the mass of
the triangle; the triangle is therefore
equimomental to this set of three
given
sides of the
triangle, the
particles.
is equimomental to a set
Example. Shew that a uniform solid tetrahedron of mass
of five particles, four of which are each of mass
and are situated at the vertices
of the tetrahedron, while the fifth
particle is at the centre of gravity of the tetrahedron
-^M
and
is
of
59.
mass
-iM.
moment of
Derivation of the
of inertia about
inertia about
a parallel axis through
the centre
any axis when
of gravity
is
the
moment
known.
The moments
of inertia found in the preceding article were for the most

taken
with
part
respect to lines specially related to the bodies concerned
these results can however be
applied to determine the moments of inertia of
:
the same bodies with

respect to other lines, by means of a theorem which will
now be
given.
Let f(x., y, z, x, y, z, x, y, z) be
any polynomial (not necessarily homo
geneous) of the second degree in the coordinates and the components of
Let (x, y, I) denote the
velocity and acceleration of a particle of mass m.
coordinates of the centre of gravity of a
body which
arid write
If now we
we obtain the
(1)
substitute these values for x, y,

following classes of terms
z,
is
formed of such
particles,
respectively, in the function/,
Terms which do not involve x y

l
,z 1
these terms together evidently
give
f(x,
(2)
Terms which do not involve x,y,z: these terms give

/(!,
(3)
y, z, x, y, z, x, y, z).
Terms which
2/j,
z 1} x lt
y l} zlt x lt y l}
z\).
are linear in x lt y lt z lt x lt ly z lt x lt
y
y lt
when the
expression ^mf(x,
formed, the summation being taken
over all the particles of the body, these terms will vanish in
consequence of
the relations
y, z, x, y, z, x, y, z) is
^mx =
l
0,
Sray!
0,
S mz =
t
0.
The dynamical
122
We have
of bodies
specification
[CH.
therefore the equation
^mf(x,
y, z, x, y, z, x, y, z)
(x lt y,, z lt
+/(>
x,, y,,
V>
z>
x>
z lt x,, y,, z,}

y>
z, 5, y,
z)
Sw,
and consequently the value of the expression 2m/, taken with

respect to
any system of coordinate axes, is equal to its value taken with respect to a
parallel set of axes
through the centre of gravity of the body, together with

the mass of the body multiplied by the value of the function f at the centre
of gravity, taken with
respect to the original system of axes.
From this it immediately follows that the moments and products of inertia,
of a body with respect to any axes are equal to the corresponding moments and
products of inertia, with respect to a set of parallel axes through the centre of
gravity of the body, together with the corresponding moments and products
of inertia, with respect to the original axes, of a particle of mass equal to that
of
the body and placed at the centre of

gravity.
As an example of this result, let it be required to determine the moment
of inertia
M and
of a straight uniform rod of mass

It follows
perpendicular to the rod.
length I about a line through one extremity

from the last article that the moment of inertia
2
;
above
result,
60.
we
see that the required
moment
of inertia
and hence, applying the
is
Connexion between moments of inertia with respect
to different sets
of
axes through the same origin.
The
moments of inertia of
a given body with respect to any set of axes, when the moments of inertia
are already known with respect to a set of axes parallel to these.
We shall
now shew how the moments of inertia of a body with respect to any set of
result of the last article enables us to find the
rectangular axes can be found when the moments of inertia are known with
respect to another set of rectangular axes having the same origin.
Let A, B,
C, F, G,
H be
to a set of axes Oxyz,
the
and
having the same origin

respect to the other will
moments and products of inertia with respect

Ox y z be another set of rectangular axes
let
the direction-cosines of either set of axes with
be supposed to be given by the scheme
^
li
The dynamical
59, 60]
moments and products
If the
are denoted by A B
A = 2m(y + z where
,
F G
C",
),
123
of bodies
specification
Ox y z
of inertia with respect to the axes
we have
the summation
extended over
is
all
the particles of
the body,
= 2m {(I x + m y + n z) + (I x + m y + n z)
= 2m [x (l + } + y (m + w ) + z (n + n/) +
2
= 2m
= 2m
[x (m,
2
(y
{I,
mn
2yz (m 2 n.2
3)
+ n, ) + if (n + ) + z (I, + m ) - Zm^yz - ZnJ.zx - Zl^

+ z ) + m* (z + x + n (x + ) - Zmj^yz - Zn^zx - Zljn.xy
I,
r/
and similarly
B = A1 + Bm.? +
On? - 2Fm.2 n2 - 2Gn 2 2 - 2Hl.2 m,
C = Al, + Bm +
Cn.?
We
l.
- 2Fm n 3
2Gn,l3
- 2Hl m
3
have also
F = 2my z
= 2m (l^ + m y + n z] (I x + m y + n z)
=
2mx + m m 2my + n n 2mz + (m. n + m n
2
12 1 3
= ^ IJ, (B + G-A) +
>n
(n
2 l3
n 3 l2 )
m C + A - B) + i n. n
+ (w + m n,) F + (n.
3 (
2 i3
2)
2mzx +
(A
2 l3
m + m
(I 2
I3
2)
2mxy
+ B - C)
n z l2 }
G + (I m +
3
or
F = Al. + Bm m + Cn. n -F(m. n

2 l3
m n - G (I n.
3
2)
+-
l.M 3 )
- H(l. m 3
2
and similarly
-G =
Al,l,
+ Bm^n, +
Cn-.n,
-H = Al,L + Bm.in, + Cn,n
- F(m. n, + m,7i ) - G (I n + n^ - H
(l^ni, + I^m
- F(m,n + m^) - G (Ln, + l^) - H
(l^t,, + l^n,).
A
ls
3 ),
The quantities A B C F G
are thus determined; these results,
combined with those of the last article, are sufficient to determine the
moments and products of inertia of a given
body with respect to any set of
,
rectangular axes when the moments and products of inertia with respect to
any other set of rectangular axes are known.
If the origin of coordinates is at the centre of
gravity of the body, prove
of inertia with respect to three
mutually orthogonal
Example.
that the
and intersecting
lines
whose coordinates are
(l\i
where A\
"h,
i>
Xj, Mi,
"i),
(1 2
A + Jf^S + rf+vft
are
0",
etc.
<i,
n 2 X 2 ^,
and
2 ),
F M (X
-
(? 3 ,
X 3 + M:
3>
/X3
%,
I,
have the sarue values as above and
X3
/x 3
i/
3)
2J, 3 ) etc.,
is
the mass of the

(Coll.
Exam.)
The dynamical
124
of bodies
specification
The principal axes of inertia ; Cauchys momental

If now we consider the quadric surface whose equation
61.
where A, B,
0, F,
G, H
Cz*
ellipsoid.
1,
-<LFyz-
body with respect to the axes of reference Oxyz,
it
is
2Gzx - 2Hxy =
are the moments and products of
+ By +
Ax-
[OH.
inertia of a given
follows from the equation
that the reciprocal of the square of any radius vector of the quadric is equal
to the moment of inertia of the body about this radius.
The quadric is
therefore the same whatever be the axes of reference provided the origin
is
unchanged, and consequently
axes
Ox y z having
where
its
equation referred to any other rectangular
the same origin
is
A a? + B f- + C z* - ZF y z - ZG z x - IR x y =
A B C F G H are the moments and products
,
of inertia with
respect to these axes.
This quadric
is
called the
momental
ellipsoid of the
body at the point
principal axes are called the principal axes of inertia of the body at
the equation of the quadric referred to these axes contains no product-terms,
its
and therefore the products of inertia with respect
them
to
are zero
and
moments
the
moments
of inertia with respect to these axes are called the principal

inertia
of the body at the point 0*.
of
The momental
regard to
its
Example.
of its base.
62.
We
ellipsoid is also called the ellipsoid of inertia
centre
is
another
ellipsoid,
which
is
its polar reciprocal with

sometimes called the ellipsoid of gyration.
The height of a solid homogeneous

Shew that its momental ellipsoid at
Calculation of the angular

shall
right circular cone
the vertex
momentum of a moving
now shew how the angular momentum
is
half the radius
a sphere.
is
of a
rigid body.
moving
about any line, at any instant of its motion, can be determined.

Let
be the mass of the body, (x, y, z) the coordinates of
rigid
its
body
centre of
gravity G, and (u, v, w) the components of velocity of the point G, at the

instant t, resolved along any (fixed or moving) rectangular axes Oxyz whose
and let (oj 1; 2

be the components of the angular
3)
body about G, resolved along axes Gx y z parallel to the axes
Oxyz and passing through G. Let m denote a typical particle of the body,
and let (x, y, z) be its coordinates and (u, v, w) its components of velocity at
the instant t and write
origin
is
fixed;

velocity of the
x
U
*
= x + xi,
U +
11^
y = y
V =
+ yi,
+V
l
=z+z
W=W+W
z
The existence of principal axes was discovered by Euler, Mem. de Berl., 1750, 1758, and by
A. Segner, Specimen Th. Turbinum, 1755. The momental ellipsoid was introduced by Cauchy
in 1827, Exerc. de math. i. p. 93.
J.
The dynamical
61, 62]
so in virtue of the properties of the centre of gravity
2m^j =
moreover since
U
it
0,
Vl
y^w^,
we have
Sm^j =
0,
we have
17)
=Z w
^myj =
125
of bodies
s2)ecification
=X
o>
2 l &)j
W =y
l
(0 l
follows that
Sm^! =
If A 3 denotes the angular

have therefore
h3
mw = 0.
Sravj s= 0,
0,
momentum
of the
= Sm (xv yu)
= 2m {(x + #,) (v + Wj) - (y + yj (u + u,)}
= 2m (xv yu) + 2w (x^ yîii)
= M (xv yu) + 2m (X^W x z a) y^z^w* + y^w^)
= M (xv yu) Gw Fw + C(D
A
C, F, G,
where A, B,
body about the axis Oz, we
H are the moments and products of inertia of
the body
with respect to the axes Gx-^y^.

Similarly the angular
are
momenta about
= M (yw
h? = M (zu
h
the axes
+ Aw
Ha)
Hw
zv}
xw}
Gco 3
n
_
+ Bwo
Ox and Oy
Fo) 3
respectively
The angular momentum about any other line through the origin can be
found ( 39) by resolving these angular momenta along the line in question.
If the body is constrained to turn round one of its
points,
fixed in space, it is unnecessary to introduce the centre of
gravity.
&) 3 ) be the
o) 2
components of the angular velocity of the body
Corollary.
which
For
is
let
(&>!,
about the fixed point with respect to any rectangular axes (fixed or
moving)
which have the fixed point as origin, and let A, B, C, F, G,
denote the
moments and products of inertia with respect to these axes. The com
ponents of velocity
coordinates are
(u, v,
w) with respect to these axes of a particle
(x, y, z) are
and the angular
in
whose
(17)
momentum
about the axis of
z,
which
is
2m
(xv
yu}, can
therefore be written in the form
Sm
or
Similarly the angular

respectively are
momenta
A(i) l
and
of the
Hw
body about the axes of x and y

Gu>
126
The dynamical
of bodies
specification
[on.
Calculation of the kinetic energy of a moving rigid

body.
63.
The kinetic energy of a rigid body which is in motion can be calculated

same way as the angular momenta. If the- general theorem obtained
in the
59 is applied to the case in which the polynomial

f(x, y, z, x, y, z, x, y, z)
has the form (x 2 + y 2 + z 2 ), we
immediately obtain the result that the kinetic
in
is equal to the kinetic

energy of a moving rigid body of mass
energy of a
mass
which
moves
with
the
centre of gravity
particle of
of the body, together
with the kinetic energy of the motion
of the body relative to its centre of
gravity.
To determine the
kinetic energy of the motion of the
body
relative to its
centre of gravity G, take any rectangular axes (whose directions

may be fixed
or moving) having their origin at G; let
be
the
1;
2
3)
components of
the angular velocity of the body about 0, relative to these axes, and let
&>
o>
(a>
denote the coordinates of a typical particle

of the body referred
The components of velocity of the particle parallel to these
axes, in the motion relative to G, are (17)
(x, y, z)
to these axes.
and therefore the kinetic energy of the motion relative

gravity
to the
centre of
is
or
2m
(Aw,-
where A, B,
C, F,
\(za>
- yo
+ Bw.?
+
+ Co)./ 2
s}
- zco^ + (y Wl xca^-},
2Fco
1&
2
(.v<o
2 (,) 3
2&>
G, H are the moments and
&>
2#&>
>,),
products of inertia relative to
the axes.
This expression may (by use of 60) be interpreted as half the

square
of the resultant angular velocity of the body in the motion relative to the
centre of gravity, multiplied by the moment of inertia of the
body about
the instantaneous axis of rotation in this motion.
If one of the points of the
body is fixed in space, it is not
For let (co 1} 2
necessary to introduce the centre of gravity.
3 ) denote the
resolved
components of angular velocity of the body about the fixed point
Corollary.
a>
&>
along any rectangular axes (fixed or moving) Oxyz which have the point
as origin, and let (x, y, z} be the coordinates of a
of the
typical particle
to
referred
these
axes.
The
of
of the
body
components
velocity
are
zo) 2
and
particle
17)
so as before
where A, B,
we
yo) 3
xais
zw^
yw
xo) z
see that the kinetic

energy of the motion
C, F, G,
H denote the
body with respect to the axes Oxyz.
is
of inertia of the
The dynamical
63, 64]
specification
of bodies
127
From
this it follows that if one of the coordinate axes

say the axis of x
the instantaneous axis of rotation of the body, the kinetic energy is
Âa)^; and hence, since the directions of the axes can be arbitrarily chosen,
the kinetic energy of any body moving about one of its points, which is fixed,
is
2
^/&) where / is the
axis of rotation, and
is
moment
&>
is
of inertia of the body about the instantaneous
the angular velocity of the body about this axis.
lamina can turn freely about a horizontal axis in its own plane, and the
Example.
axis turns about a fixed vertical line, which it intersects.
If
be the azimuth of the
horizontal axis, and
the inclination of the plane of the lamina to the vertical, shew that
(f>
the kinetic energy
is
where A, B, Zf are the moments and product of inertia of the lamina about the horizontal
and a perpendicular to it at the point of intersection with the vertical. (Coll. Exam.)
axis
64.
Independence of the motion of the centre of gravity and the motion
relative to
The
it.
shews that the kinetic energy of a moving

can
be
as
body
regarded
consisting of two parts, of which one depends on the
motion of the centre of gravity and the other is the kinetic energy of the motion
relative to the centre of gravity. We shall now shew that these two
parts of
the motion of the body can be treated quite independently of each other*.
result of the last article
Let a rigid body of mass

forces
As coordinates defining
coordinates
(x, y, z)
be in motion under the influence of

any
its position we can take the three
rectangular
of its centre of gravity G, relative to axes fixed in

space,
and the three Eulerian angles (6, 0, i/r) which specify the
position, relative
to axes fixed in direction, of any three
lines,
orthogonal
intersecting in G,
which are fixed in the body and move with it. The kinetic
energy is therefore
T=
where f(0,
0, &,
0,
M (x* + p + *) +/(B,
0,
+,
e,
0, ^),
0, ^) denotes the kinetic energy of the motion relative
to G.
Let
denote the work done on the body by the external forces in an
arbitrary dis
placement (Bx, By, Bz, B0, 80, 8^r) of the body. The Lagrangian equations of
motion are
MX = X, My =
Y,
Mz = Z,
<*L
dt
di
80
df
dt
*
Euler, Scientia navalis,
i.
(1749),
128.
The dynamical
128
The
specification
of bodies
[CH.
three of these equations shew that the motion of the centre of

gravity of
body is the same as that of a particle of mass equal to the whole
mass of the body, under the influence of forces equivalent to the total external
first
the
forces acting on the body, applied to the particle parallel to their actual
directions since the work done on such a
particle in an arbitrary displace
ment would evidently be
Bx + YBy + Z8z.
;
The second three equations shew that
the motion
of the body about
its
centre of gravity is the same as if the centre of gravity were fixed and the
body
subjected to the action of the same forces for in the motion relative to the
centre of gravity, the kinetic energy of the body is f(6,
and
-^r, 6, $, i^),
the work done by the forces in an arbitrary displacement is
;
<f>,
+
These results are evidently true also
<J>S
for impulsive motion.
If a plane rigid body (e.g. a disc of

any shape) is in motion in
its mass,
y) are the coordinates of its centre of gravity,
6 the angle made by a line fixed in the body with a line fixed in the
plane,
2
the moment of inertia of the body about its centre of
gravity, and if
(A F) are the total components parallel to the axes of the external forces
Corollary.
its plane,
and
if (x,
Mk
,
acting on the body, and L the moment of the external forces about the centre
of gravity, then the kinetic energy is
\M (x + y- + k*fr\
2
and the work done by the external
forces in a displacement (Bx,

By, 86)
is
and therefore the equations of motion of the body are
MX = X, My = Y, M&0 = L.
Example.
Obtain one of the equations of motion of a rigid body in two dimensions
in
the form
the mass of the body, / is the acceleration of its centre of

gravity, p is the
perpendicular from the origin upon this vector, J/F is the moment of inertia about the
origin, 6 is the angle made by a line fixed in the body with a line fixed in its plane, and L
is the moment about the origin of the external forces.
(Coll. Exam.)
where
is
The dynamical
64]
129
of bodies
specification
1.
homogeneous right circular cone
the length of a slant side
and that
its
moment
is
I.
Shew that
moment
its
semi-vertical angle
is
/3,
and
of inertia about its axis is
(l-f
sin 2 /3),
of inertia about a generator is
Shew that the moment
2.
mass M;
moment
of inertia about a line through its vertex perpendicular to its axis is
f
atid its
of
is
its
by the two loops
of inertia of the area enclosed
of the
lemniscate
r2
about the axis of the curve
(37r
Any number
3.
distances apart o?12

i2, ..., prove that
cos20
is
8)
-
a2
x mass of area.
of particles are in one plane; if the

the relative descriptions of area A 12 ,
...,
masses are
...,
and the
Wj,
...,
their
relative velocities-
are respectively the moment of inertia about the centre of inertia, the angular momentum
about the centre of inertia, and the kinetic energy relative to the centre of inertia.
(Coll.
Prove that the
4.
moment
of inertia of a hollow cubical box about
the centre of gravity of the box and perpendicular to one of the faces
where
is
the mass of the box and 2a the length of an edge.
The
Shew
where a
is
an axis through
is
sides of the
supposed to be thin.
5.
(Coll.
that the
moment
of inertia of
the radius of the generating

and p is the density.
an anchor-ring about
circle, c is
its
the distance of
axis
its
Exam.)
box are
Exam.)
is
centre from the axis
of the anchor-ring,
Shew how
6.
body, and
if
there
what point, if any, a given straight line is a principal axis of

such a point find the other two principal axes through it.
to find at
is
a-
A uniform square lamina is bounded by the axes of x and y and the lines x=2c, y = 2c,
and a corner is cut off it by the line xja+y/b = 2. Shew that the two principal axes at
the centre of the square which are in
given by
its
own plane
- -
ab
tan 20 = -
2c
j-
(a- 6)
7.
are inclined to the axis of
Shew that the envelope
moment
(a
+b
of lines in the plane of
w. D.
at angle?
(Coll.
Exam.)
2c)
an area about which that area has a

and hyperbolas. Hence find the
of inertia is a set of confocal ellipses

direction of the principal axes at any point.
constant
(Coll.
Exam.)
9
The dynamical
130
8.
rectum
[CH.
Find the principal moments of inertia at the vertex of a parabolic lamina, latus
4a, bounded by a line perpendicular to the axis at a distance h from the vertex.
Prove that,
is
of bodies
specification
two principal axes at the point on the parabola whose abscissa
if 15A>28,
-a + (a 2 - 4aA/5 +3A 2/7)i
are the tangent and normal.
(Coll.
Exam.)
Find how the principal axes of inertia are arranged in a plane body. Write down
the conditions that particles m^ at (x ?/,-), where z =l, 2, ...,
may be equimomental to a
given plate. Shew that the six quantities m 1 m 2 &\, x 2 y^ y 2 can be eliminated from
9.
these conditions.
If three equal particles are
formed by them
is
equimomental to a given plate, the area of the triangle

3 ^/3/2 times the product of the
principal radii of gyration at the
centre of gravity.
Exam.)
(Coll.
uniform lamina bounded by the ellipse b 2x2 + a 2y 2 =

has an elliptic hole
(semi-axes c, d) in it whose major axis lies in the line x=y, the centre being at a
distance r from the origin
prove that if one of the principal axes at the point (x, y)
makes an angle 6 with the axis of x, then
10.
aW
tan 26 =
Sabxy
ab [4
(x*
~ d^
(x
^ ~ ^ ^ ~^
(
2 - cd
y*) + a
&*]
[2
"
(x
r)
~
2
r)
(c
-2
~
(y
^} ^2
r)
Exam.)
(Coll.
system of bodies or particles is moved or deformed in any way, shew that

the sum of the products of the mass of each particle into the
square of its displacement
is equal to the product of the mass of the
system into the square of the projection in any
given direction of the displacement of the centre of gravity, together with the sum of the
products of the masses of the particles into the squares of the distances through which
11.
If a
they must be moved in order to bring them to their final positions after communicating
them a displacement equal to the projection in the given direction of the displacement
to
of the centre of gravity.
The
(Fouret.)
moments of inertia of a body at its centre of gravity are (A, B, G )

if a small mass, whose moments of inertia referred to these axes are
(A B C ), be added
to the body, shew that the moments of inertia of the
compound body about its new
principal axes at its new centre of gravity are
12.
principal
A+A
accurately to the
first
B+B
C+C
order of small quantities.
(Hoppe.)
Shew
that the principal axes of a given material system at

any point are the
normals to the three quadrics which pass through the point and
belong to a certain
confocal system.
13.
If (I, m,
X, /*, i/) be the six coordinates of a principal axis and the associated
Cartesian system be the principal axes at the centre of
gravity, then shew that
,,
A l\ + Bmp. + Cnv = 0,
and therefore
all
principal axes of a given system belong to a quadratic complex.

(Coll.
Exam.)
smoothly jointed framework is in the form of a parallelogram formed bv

attaching the ends of a pair of rods of mass m and length 2a to those of a pair of rods of
mass m and length 26. Masses
are attached to each of the four corners.
Express the
14.
angular momentum of the system about the origin of coordinates, in terms of the
coordinates (A; y) of the centre of gravity and the
between the two pairs of
angles 6 and
<
sides
and the axis of x.
(Coll.
Exam.)
CHAPTER
VI
THE SOLUBLE PROBLEMS OF RIGID DYNAMICS

The motion of systems with one degree of freedom
65.
fixed axis,
We
motion round
etc.
now proceed
to apply the principles which have been developed in

the foregoing chapters in order to determine the motion of holonomic systems
of rigid bodies in those cases which admit of solution by quadratures.
systems which have only one degree of

42) that such a system is immediately soluble by
It is natural to consider first those
We
freedom.
quadratures
have seen
when
and this principle is

Sometimes, however (e.g. when
possesses an integral of energy:
it
most cases.
which one of the surfaces or curves of con
straint is forced to move in a given manner), the problem as originally formu
lated does not possess an integral of energy, but can be reduced (e.g. by the
sufficient for the integration in

we are dealing with systems in
theorem of
when
29) to another problem for which the integral of energy holds

been performed, the problem can be integrated as
;
this reduction has
before.
The
(i)
following examples will illustrate the application of these principles.
Motion of a rigid body round a fixed
axis.
Consider the motion of a single rigid body which is free to turn about an axis, fixed in
the body and in space. Let / be the moment of inertia of the body about the axis, so that
/0 2 where 6 is the angle made by a moveable plane, passing through
its kinetic energy is
,
the axis and fixed in the body, with a plane passing through the axis and fixed in space.
Let 6 be the moment round the axis of all the external forces acting on the body, so that
$ is the work done by these forces in the infinitesimal displacement which changes 6 to
6 -\-
fid.
of motion
d_
SVT\ _ dT_
dt\ti)
then gives
which
is
dd~
IB = 0,
a differential equation of the second order for the determination of
G.
92
The Soluble Problems of Rigid Dynamics
132
If the forces are conservative,
and V
[CH. vi
denotes the potential energy, this equation
(6}
becomes
-^
80
10=
which on integration gives the equation of energy
$I6
Integrating again,
and
this relation
V(ff)
= c,
where
c is
a constant.
we have
= I\
between 9 and
being determined by the
The most important
f{2 (c
V}}
"%dd
+ constant,
determines the motion, the two constants of integration
initial conditions.
that in which gravity is the only external force, and the

let G be the centre of gravity of the body, C the foot of
The potential energy is
the perpendicular drawn from G to the axis, and let CG = h.
is the mass of the body and 6 is the angle made by CG with the
Mgh cos 0, where
case
is
In this case
axis is horizontal.
downward
vertical
and the equation
of motion is
This is the same as the equation of motion of a simple pendulum of length J/Mh, and
the motion can therefore be expressed in terms of elliptic functions as in
44, the solution
being of the form
.
sin
in the oscillatory case,
--
6
= A sn
,
and of the form
in the circulating case.
The quantity IjMh
is
called the length of the equivalent simple
pendulum.
If
be a point on the line CG such that OC=IjMh, the points
and C are called
curious result in this
respectively the centre of oscillation and the centre of suspension.
connexion is that the centre of oscillation and the centre of suspension are convertible,
is the centre of oscillation when C is the centre of suspension, then C will be
i.e. if
the centre of oscillation when
have by
is
the centre of suspension.
To prove
this result,
we
59
Moment
of inertia of the
Moment
body about
of inertia about
G + M. GO 2
Moment
of inertia of
2
2
_ I- Mh + M (I/Mh - A)
body about
IjMh - h
Distance of centre of gravity from
If therefore the
body were suspended from

v
6+
0, the equation of
motion would
still
be
Mqh sin =
0,
.
J-
which establishes the result. It is evident that the period of

same about either of the points C and 0.
oscillation
would be the
65]
(ii)
Motion of a rod on which an
133
insect is crawling.
We shall next study the motion of a straight uniform rod, of mass m .and length 2a,
whose extremities can slide on the circumference of a smooth fixed horizontal circle of
radius c an insect of mass equal to that of the rod is supposed to crawl along the rod at
;
a constant rate v relative to the rod.
Let 6 be the angle made by the rod at time t with some fixed direction, and let x be
the distance traversed by the insect from the middle point of the rod.
The kinetic
2
c2
#2
(2a35-1
and the kinetic energy of the insect
is
due to
a component of velocity {x - (c2 - a 2 ) 2 0} along the rod and a component of velocity xB

perpendicular to the rod, so the total kinetic energy of the system is
there
is
no potential energy.
Since x=vt,
(t
being measured from the epoch
T= \m (c
The coordinate
6,
which
- 2a2/3)
is
now
when x
is zero),
+ \m {v - (c 2 - a 2 )^ 0} 2 +
the only coordinate,
we have
mv*t*0\
is igriorable,
and we have therefore
= constant,
or
(c 2
0-m (c
-|- )
or
0(2c
- a 2 )^
{v
t;
- (c 2 - a
2 2
i!
)* 0}
+ mv2
= constant,
= constant.
we have
#-0
where
arctan
{vt (2c
-|a
},
and k are constants. This formula determines the position of the rod at any time.
(iii)
Motion of a cone on a perfectly rough inclined plane.
Consider now the motion of a homogeneous solid right circular cone, of mass
and
/3, which moves on a perfectly rough plane (i.e. a plane on which
only
Let I be the
rolling without sliding can take place) inclined at an angle a to the horizon.
be the angle between the generator which is
length of a slant side of the cone, and let
in contact with the plane at time t and the line of greatest slope downwards in the
plane.
semi-vertical angle
^ be the angle made by the axis of the cone with the upward vertical, ^ is one
whose vertices represent respectively the normal to the plane,
the upward vertical, and the axis of the cone ; the other two sides are a and
the
Then
if
side of a spherical triangle
angle included by these sides being
(TT
- 0}.
cos x = cos a sin
ft
We
(?),
have therefore.
sin a cos
ft
cos
but the vertical height of the centre of gravity of the cone above its vertex is I cos ft cos
^,
and the potential energy of the cone is Mg x this height if therefore we denote by V the
;
potential energy of the cone,
we have
V=
(disregarding a constant term)
^Mgl sin a
cos 2
ft
cos
0.
134
We
[CH. vi
have next to calculate the kinetic energy of the cone for this the moments of
about its axis and about a line through the vertex perpendicular to the
axis are required
these are easily found (by direct integration, regarding the cone as
;
inertia of the cone

:
^MP
sin 2 /3 and
discs perpendicular to its axis) to be
J/7 2 (cos 2 /3 + Jsin 2 /3)
moment
of
inertia
about
a
and
so
the
theorem of
60
the
is,
generator
respectively,
by
(since the direction-cosines of the generator can be taken to be sin ft, 0, cos $ with respect
composed of
to rectangular axes at the vertex, of
MV
all
z is
+ ^MP sin
Jft 2 sin 2 0(cos 2
+ i).
or
Now
which the axis of
2
2
1
(cos ft + sin ft) sin
2
points of that generator which
ft
in contact
is
the axis of the cone),
2
ft
cos 2
ft,
with the plane are instantaneously
at rest, since the motion is one of pure rolling, and therefore this generator is the
If w denotes the angular velocity of the
instantaneous axis of rotation of the cone.
cone about this generator, the kinetic energy of the cone

jj-
But
15)
MP sin
2
ft
(cos
2
ft
+ J)
w2
is
therefore
63, Corollary)
we have
a)
and substituting
this value for
o>,
=6
cot
we have
ft,
finally for the kinetic
energy
of the cone
the value
T=%MP cos
becomes therefore
is
ft
(cos
of motion
in this case
2
2
| Ml* cos ft (cos
This
ft
+ Mgl sin a cos2 ft sin 6 = 0,
the same as the equation of motion of a simple

I
cosec a (cos
2
ft
pendulum
of length
the integration can therefore be effected in terms of elliptic functions, as in

(iv)
44.
Motion of a rod on a rotating frame.
Consider next the motion of a heavy uniform rod, whose ends are constrained to move
in horizontal and vertical grooves respectively, when the framework containing the grooves
is
made
to rotate with constant angular velocity
co
about the line of the vertical groove.
Let 2a be the length of the rod,

its mass, and 6 its inclination to the vertical.
of
the
allowed for by adding to the potential energy
the
effect
rotation
be
29,
may
By
a term
where p is the density of the rod and x denotes distance measured from the end of the rod
which is in the vertical groove integrating, this term can be written
;
The term
in the potential
energy due to gravity
Mga
and the
is
cos
6,
V is therefore given by the equation

V= - Mga cos d - % Jfo*a* sin 2 6.
total potential energy
The
horizontal and vertical components of velocity of the centre of gravity of the rod
motion of the
$, so the part of the kinetic energy due to the
are a sin 6 6 and a cos 6

.
is
about
its
centre
its
centre
is
\Ma?6
z
;
inertia of the rod
The
integral of energy
therefore
is
Ma
or,
and since the moment of
the part of the kinetic energy due to the rotation of the rod about
It Ma?,
2 2
6
Ma
we have therefore for the total kinetic energy T the equation
J
centre of gravity
is
135
65]
62
Mga cos 6
2
M<o
a 2 sin 2 6 = constant,
writing
where
this constant must evidently be positive, since x 2 and

is not very large and that
suppose for definiteness that
than unity, so that x oscillates between the values 3^/4aco 2
e
denotes a constant
We
are positive.
is less
To
shall
x2 )
(1
3<7/4or
e/&>.
integrate this equation,
we write*
#=!+.
^8a
where
a new dependent
we have
o>
+
^12
Substituting this value for x in the differential
variable.
is
equation,
64
18"
where the values
correspond respectively to the values
it is
easily seen that
ei+e 2 + e 3
is
zero
and that
el
>e
2 >e 3 .
We have therefore =
+ y), where the function is formed with the roots e e 2 es
lies between e 2 and e3 for
and where y denotes a constant. Since
e^, and (P (<+y)
real values of t (since x lies between 3#/4co 2 and 3^/4aa) 2 + e/a)), the imaginary part of
the real part of y can then be taken as zero,
the constant y must be the half-period
|jf>
(<
} ,
$>
e^~>
e<{>
e/o>
o>
since
it
;!
depends only on the choice of the origin of time.
We
have therefore
and hence
this equation determines ^ in

(v)
Motion of a
disc,
terms of
<.
one of whose points
is
forced
to
move in a given manner.
Consider next the motion of a disc of mass M. resting on a perfectly smooth horizontal
when one of the points A of the disc is constrained to describe a circle of radius c
plane,
in the horizontal plane,

*
Cf.
with uniform angular velocity
Whittaker and Watson,
o>.
Course of Modern Analysis, % 20
6.
136
G
be the centre of gravity of the disc, and let A G be of length a. The acceleration
A is of magnitude cw 2 and is directed along the inward normal to the circle
therefore we impress an acceleration cw 2 directed along the outward normal to the
Let
of the point
if
[OH. vi
on all the particles of the body and suppose that A is at rest, we shall obtain the
motion relative to A. The resultant force acting on the body in this motion relative to A
circle,
is
therefore
2
J/co>
acting at
outward normal to the
in a direction parallel to the
circle.
Let 6 and
be the angles made with a fixed direction in the plane by the line AG and
the outward normal to the circle respectively then the work done by this force in a small
;
displacement 86
is
2
Mca>
sin (0
<9)
8<9,
and the kinetic energy of the body is i J/ 2# 2 where Mk2 is the moment of
The Lagrangian equation of motion is therefore
body about the point A
,
inertia of the
Mm = Macrf
But
since
=
a>,
= 0;
we have
so
^ be
if
sin (0
- 6\
written for (6
acco 2
^+-p-sm^
0),
we have
= 0.
This is the same as the equation of motion of a simple pendulum of length k2

the integration can therefore be performed by means of elliptic functions as in
44.
gjaca>
(vi)
Motion of a disc rolling on a constrained disc and linked
to
2
;
it.
Consider the motion of two equal circular discs, of radius a and mass J/, with edges
perfectly rough, which are kept in contact in a vertical plane by means of a link (in the
form of a uniform bar of mass m) which joins their centres the centre of one disc is fixed,
:
and
constrained to rotate with uniform angular acceleration a

to determine the motion of the other disc B and the link.
this disc
is
it is
required
be the angle which the link makes with the downward vertical at time t, and
Let
be the angle turned through at time t by the disc A. The angular velocity of disc A
0, and the velocities of the points of the discs which are instantaneously in contact are
let 6
is
Since the velocity of the centre of the disc B is 20, it follows that the
angular velocity of the disc B about its centre is 20-0. Since the moment of inertia of
each disc about its centre is i Ma 2 the kinetic energy of the system is
therefore each a@.
+ lM.
T=M.^V
m
and
= at +
The
f,
where
f is
a constant.
potential energy of the system
F=
is
(2M+m) ag cos 0,
and the Lagrangian equation of motion
is
(o_T\_
<ti\d<p)
dT__dV
~
80
80
Ma*0}= - (2J/ + wi) ag
sin 0.
- J/a 2a + (2 M+ m) ag sin
= 0.
J/u 2 n0
= c,
j {(6Af+f m) a 0-
or
m.~&o
*-(2$-eT- + lM.(2a?<tf +
iu
Since d = a, this equation gives

(6
Integrating,
M+ 1 m) a
we have
2
(3J/+ % m) a (fi-
(2J/+ m) ag cos
65, 66]
137
are
a constant depending on the initial conditions and as the variables t and
by a quadrature: this final integral
determines the motion.
where
c is
</>
separable, this equation can again be integrated
Example.
If the
system
at rest with the bar vertically downwards, shew
is initially
that the bar will reach the horizontal position
66.
if
The motion of systems with two degrees of freedom.
In the dynamics of rigid bodies, as in the dynamics of a particle, the

possibility of solving by quadratures a problem with two degrees of freedom
The integral
generally depends on the presence of an ignorable coordinate.
corresponding to the ignorable coordinate can often be interpreted physically
momentum. The formation and
as an integral of momentum or angular

solution of the differential equations is
by application of the
this will be shewn by the
effected
principles developed in the preceding chapters
following illustrative examples.
Hod passing
(i)
through ring.
Consider, as a first example, the motion of a uniform straight rod which passes through
a small fixed ring on a horizontal plane, being able to slide through the ring or turn in any
way about it in the plane.
rod
its
Let the distance from the ring to the middle point of the rod at time t be r, and let the
make an angle 6 with a fixed line in the plane let 21 be the length of the rod, and
mass.
The moment
is
of inertia of the rod about its middle point
is
\ MP, and the kinetic energy
therefore
there
is
no potential energy.
The coordinate
is
ignorable,
and the corresponding integral
dT =
r
is
constant,
vB
or
>2
(?
The
+ %l-) 6 = constant.
integral of energy is
;,2 _|_
,.202
_|_
^ I lfri
_ cons tant.
Dividing the second of these integrals by the square of the
first,
we have
dry
1
dft)
or
B + constant
where
{(r
+}
I )
(cr*
+ J cl 2 - 1 )}
4 dr.
1
Writing cr = s, this becomes
+ constant =
{4s (s
+ \ cP)
+ J cP - 1 )}
* ^s.
C 1S
a constant
138
If therefore
g>
denotes the Weierstrassian elliptic function with the roots
which
satisfy the relation e }
since
dr
>e
>e
3 if
positive,
constant $
The
we have
-^
is
0>(0-0
gj
)>
we have
sufficiently great initially,
(0-
s=ft>
is
[CH. vi
ei
where
is
a constant of integration
and consequently the
for real values of 0,
is real.
solution of the problem
is
therefore contained in the equation
0ê cylinder rolling on another under gravity.
(ii)
Let it now be required to determine the motion of a perfectly rough heavy solid
homogeneous cylinder of mass m and radius r, which rolls inside a hollow cylinder of mass
and radius R, which in turn is free to turn about its axis (supposed horizontal).
Let
<j)
denote the angle which the plane through the axes of the cylinders at time
makes with the downward vertical, and let 6 be the angle through which the cylinder of
mass
has turned since some fixed epoch. The angular velocities of the cylinders about
their axes are easily seen to be 6 and {(R
r}
R6}jr respectively and the moments of
inertia of the cylinders about their axes are MB? and \ mr 2 respectively
so the kinetic
energy
T of the
system
is
i m (R - r? A*
while the potential energy
is
F=
The coordinate 6
is
mg (R
clearly ignorable
dT =
r
r)
cos$.
the integral corresponding to
it is
constant,
30
MR^ -\mR{(R-r)4>-RQ} = k,
or
The
where k
is
a constant.
where h
is
a constant,
integral of energy is
T+
\MRW + }m{(R - r)
or
V=/i,
- BflY+m (R - r) 2
2
tf>
mg
(R -
r}
cos
= h.
Eliminating 6 between the two integrals, we obtain the equation
m (3J/+m)
This
is
the same as the equation of energy of a simple pendulum of length

(l
the solution can be effected by

(iii)
We
Rod moving
means
of elliptic functions as in
44.
in a free circular frame.
shall next consider the motion of a rod, whose ends can slide
freely on a smooth
vertical circular ring, the ring being free to turn about its vertical
diameter, which is fixed.
139
66]
be the mass of the ring and r

be the mass of the rod and 2a its length let
let 6 be the inclination of the rod to the horizontal, and $ the azimuth of the
ring referred to some fixed vertical plane, at any time t.
Let
its
radius
The moment of inertia of the rod about an axis through the centre of the ring
2
2
perpendicular to its plane is m(r
fa ), and the moment of inertia of the rod about the
The kinetic energy of the
vertical diameter of the ring is r{(r 2 -a 2 ) sin 2 5 + Ja 2 cos 2 0}.
system
therefore
is
T=\m (r
The
potential energy
- a2 )
+ } Mr^ + lmty (r2 sin 2 6 - a 2 sin 2 6 + J a 2 cos 2 0).
is
V
The coordinate
is
mg
evidently ignorable

or
Mr2 +
(j>
where k
is
m<j)
2
(r sin
6 - a 2 sin 2 6 + \a 2 cos 2 6)
Substituting the value of
a constant.
= k,
found from this equation in the
integral of energy
T+
V=h,
we have
lm
(r
- f a2) G 2
h + mg
(r
x.
a 2 ) a cos
-1
-
2
2
(r sin
k2
9
^ - a 2 sin 2
-- ....
&
\Mi + m
0+â2 cos
,
2~m
2
5)
In this equation the variables 6 and t are separable

a further integration will
therefore give 6 in terms of t, and so furnish the solution of the problem.
;
(iv)
Hoop and
ring.
We shall
next discuss the motion of a system consisting of a uniform smooth circular

hoop of radius a, which lies in a smooth horizontal plane, and is so constrained that it can
only move by rolling on a fixed straight line in that plane, while a small ring whose mass
is 1/X that of the hoop slides on it.
The hoop is initially at rest, and the ring is projected
from the point furthest from the
fixed line with velocity
v.
Let
denote the angle turned through by the hoop after a time t from the commence
ment of the motion, and suppose that the diameter of the hoop which passes through the
ring has then turned through an angle \^. Taking the ring to be of unit mass, so that the
</>
hoop is X, the moment of inertia of the hoop about its centre is Xa 2 and this
moves with velocity 0, while the velocity of the ring is compounded of components
a$ and a\^, whose directions are inclined to each other at an angle \^. The kinetic energy
mass
of the
centre
of the system
is
therefore
T=
The coordinate
is
<j)
is zero.
evidently ignorable, and the corresponding integral
cT
r
is
= constant,
80
or
(2X
+ 1) a- ^ + a 2 ^
cos
\/^
= the
= av.
initial
value of this expression
140

(2X + 1)
(f>
[CH. vi
we have
vt
-=the
+ sin
initial
>//
value of this expression
=0,
vt
This equation determines
The equation
in
of energy
terms of
T
and substituting
for
its
Writing
value (v/a-cos-^-
this
.r,
value = 4*2
its initial
-j=
v v2X J
=
sin\^
\^.
is
+ l)
4-)/(2X
>
in this equation,
we have
becomes
--%=
v v^2X
+
P(2X
o
In order to evaluate this integral, we introduce an auxiliary variable
defined
by the
equation
u=
X
I
(2X + .r
2
)
Jo
Write x2 = 2X/, where |
which
is
new
variable
-4(1- .r 2
dx.
the last integral becomes
equivalent to
is
where the function
$>
()
is
these roots are real and satisfy the inequality

ei for real values of u.
Now we
have
dt =
%=. (2X + #
vv/2X
JVKvdt
-
or
a
Integrating,
where
~
)
(u)
Thus
^2X
6i>e 2
)4
>e
so
is
$>(w)
real
and greater than
- a;2 ~ 4 dx.
2X
+ ^^-,
\du.
we have
denotes the Weierstrassian Zeta-function.
finally the coordinate
and
variable u by the equations
2X
the time
are expressed in terms of
an auxiliary
66, 67]
67.
141
Initial motions.
We have already explained in 32 the general principles used in finding

the initial character of the motion of a system which starts from rest at
The following examples will serve to illustrate the procedure
of
rigid bodies.
systems
a given time.
for
A particle hangs by a string of length b from a point in the circumference of a disc

(i)
The disc can turn about its axis, which is horizontal, and
of twice its mass and of radius a.
To find the
the diameter through the point of attachment of the string is initially horizontal.
initial
path of the
particle.
Let 6 denote the angle through which the disc has turned, and
string to the vertical,, at time t from the beginning of the motion
the inclination of the

let
m be the mass of the
The
horizontal and (downward) vertical coordinates of the particle with respect

to the centre of the disc are
particle.
a cos 6 + b
a sin 6 + b cos 0,
and
sin
so the square of the particle s velocity is
a2
+&
and the kinetic energy of the system
is
- 2ab sin
T= ma-6 + m&
z
-|
while the potential energy
(6
+ 0) 00,
mob sin
(6
+ 0) #0,
is
+ b cos 0).
(a sin 6
mg
of motion are
dt\d0
/dT\ _ dT_ _ 9 F
d_
dt\d(p/
2
J2
<9-a&cos
6 2 0-a6cos(0
are all zero
= 0,
these equations therefore give

initially
term gt 2/4a and that of
with a
so the expansion of 6 begins with a

hiher
suare of t. Assumin
higher than the square
Assuming
<9=gr/2a
term
30
+ 0)02 -#acos$
Initially the quantities 6, 0, 6,
and
c0
a& sin (0 + 0) = 0,
+ 0)0 + #6sin0-a6sin(0 + 0) 6 =0.
(0
= C + Dt* + Et+Ff< + ...,

t--
substituting in the
above
the coefficients A, B, C,
differential equations,
...
we thus
0-^+0.
4a
Now if x and y are the

vertical axes
through
x=a
and
ff
32a6
t,
we can evaluate
^
1920a6 2
coordinates of the particle referred to horizontal and

(downward)
its initial position,
(1
of
+...
*=
and equating powers
find
- cos 6} -
b sin
we have
= \ ad - 60 = -E2
approximately,
y = a sin 6 + b (cos 0-l)=a0=~, approximately.
142
Eliminating
and
this
is
[CH. vi
between these equations, we have
the required approximate equation of the path of the particle in the
neighbourhood of
its initial position.
A ring of mass m can slide freely on a uniform rod of mass and length 2a, which
(ii)
can turn about one end. Initially the rod is horizontal, with the ring at a distance ra from
To find the initial curvature of the path of the ring in space.
the fixed end.
Let (r, 6) denote the polar coordinates of the ring at time t, referred to the fixed end of
the rod and a horizontal initial line, 6 being measured downwards from the initial line.
For the kinetic and potential energies we have
4/7 2
T=%m(r* + rW}+%M.~6\
V=
mrg sin &
Mag sin
6.
of motion are
-._.
W~
~dtdf
dr
g sin 6 = 0,
r8 2
- Mga cos 6 - mgr cos

[f Ma*d + mr d + 2mrf0
and 6 are initially zero, we can assume expansions
z
<9
Since
r, 0,
= 0.
of the form
substituting these expansions in the differential equations, and equating coefficients of

powers of t, we find
_
~
coordinates of the particle, referred to horizontal and vertical axes at

position, are
The
x = r cos 6
x=(
or approximately
The curvature
of the path is given
and y = r sin
r
r
b^}
y=r
t*,
its initial
6,
6 2 12
by the equation
and on substituting the above values of

1
~~
This
is
62
and
Ma
(4a
9r
we have
3r
()
(Ma + mr
the required initial curvature of the path of the ring.
Two uniform rods AB, SC, of masses mi and m.2 and lengths a and b
turn round the point A, which is fixed.
are
freely hinged at B, and can
respectively,
Shew
vertical.
and
is horizontal
that, if C be released, the equation of
Initially,
nearer to G can be put in the form
of trisection of
of the
the initial
Example.
EC
AB
path
point
BC
f = 60 (1 + 2?n 2 /m
1)
abx.
I,
1896.)
67, 68]
143
The motion of systems with three degrees of freedom.
68.
The possibility of solving by quadratures the motion of a system of rigid

bodies which has three degrees of freedom depends generally (as in the case
of systems with two degrees of freedom) either on the occurrence of ignorable
coordinates, giving rise to integrals of momentum and angular momentum, or
on a disjunction of the kinetic potential into parts which depend on the
coordinates separately.
(i)
The
following examples illustrate the procedure.
Motion of a rod in a given field of force.
Consider the motion of a uniform rod, of mass

and length 2a, which is free to move
on a smooth table, when each element of the rod is attracted to a fixed line of the table
with a force proportional to

Let
(#, y)
fixed line.
its
mass and
its
distance from the line.
be the coordinates of the middle point of the rod, and 6
The
kinetic energy
its
inclination to the
is
is
11
r= ~
4<2
J -a
(y
+ rsm6} z dr,
V=pm(iy 2 + %a
or
where p
is
a constant,
$in 2 d).
of motion are therefore
#=0,
|
y
((20)
The
first
+ fi sin
w>
20 = 0.
two equations give
ct
+ d,
where c, d, /, e are constants of integration the centre of the rod therefore describes
a sine curve in the plane. The equation for 6 is of the pendulum type, and can be
;
integrated as in
(ii)
44.
Motion of a rod and cylinder on a plane.
We shall
next discuss the motion of a system consisting of a smooth solid

homogeneous
mass
and radius c, which is moveable on a smooth horizontal plane,
circular cylinder, of
mass m and length 2a, placed with its length in contact with
the cylinder, in a vertical plane perpendicular to the axis of the cylinder and
passing
through the centre of gravity of the cylinder, and with one extremity on the plane.
and a heavy straight
rail of
Let 6 be the inclination of the rail to the vertical, and x the distance traversed on the
plane by the line of contact of the cylinder and plane, at any time t. The coordinates of
the centre of the rod referred to horizontal and vertical axes, the origin
being the initial
point of contact of the cylinder and plane, are easily seen to be
ccot(\4
-)
+ asin$ and
acosd.
zj
Let
$ be the angle through which the cylinder has turned at time

of
the system is
energy
+ A ma 2 sin 2 6
d2 +
t.
The
MX
kinetic
144
The
potential energy
is
V=-
The coordinates x and
mga
ox
of x)
6.
the corresponding integrals are
constant
be interpreted as the integral of
may
cos
are evidently ignorable
<
ar =
(which
[CH. vi
momentum
of the system parallel to the axis
and
tiT
^ = constant
momentum
be interpreted as the integral of angular
(which
may
axis).
These integrals can be written
ic cosec 2 [ \4
+ a cos 6
2/
.
f-
+ Mx= constant,
2
c;
Substituting for
x and
</>
of the cylinder about its
<>
= constant.
the vahies obtained from these equations in the integral of
energy
T+ F= constant,
we have the equation
If
where d
6
,/TT
-kc cosec 2 - - -
This equation is again integrable, since the variables t and 6 are

form it gives the expression of 6 in terms of t the two
integrals found above then give x and $ in terms of t.
separable
is
;
a constant.
in its integrated
69.
Motion of a body about a fixed point under no forces.
One
of the most important problems in the dynamics of systems with
three degrees of freedom is that of determining the motion of a rigid body,

one of whose points
is fixed, when no external forces are
supposed to act*.
This problem is realised ( 64) in the motion of a rigid body relative to its
centre of gravity, under the action
through the centre of gravity.
of
any
forces
whose resultant passes
momentum of the body about every line which

the
fixed
passes through
point and is fixed in space is constant ( 40), and
the
line
consequently
through the fixed point for which this angular momen
In this system the angular
Let this line, which is called the

greatest value is fixed in space.
be
taken as axis OZ, and let
invariable line,
and
be two other axes
tum- has
its
OX
through the fixed point which are perpendicular to OZ and to each other.
The angular momenta about the axes OX and OF are zero, for if this were
not the case the resultant of the angular momenta about OX, OY, OZ would
give a line about which the angular momentum would be greater than the
*
Euler, Memoires de Berlin,
Annee
1758.
Elliptic functions were applied to the
by Eueb, Specimen inaugurale... (Utrecht, 1834)

Journal fur Math, xxxix. (1849), p. 293.
first
problem
and the solution was completed by Jacobi,
68,
momentum
angular
about OZ, which
OZ
angle 6 with
d cos
is
line
making an
through
where d denotes the angular
6,
It follows
contrary to hypothesis.
is
momentum about any
39) that the angular
145
69 J
momentum
about OZ.
is sufficiently specified by the

position of the body at any time t
knowledge of the positions at that time of its three principal axes of inertia
at the fixed point let these lines be taken as moving axes Oxyz let (6,
ty)
The
<f>,
denote the three Eulerian angles which specify the position of the axes Oxyz
with reference to the axes OXYZ, let (A, B, C) be the principal moments of
inertia of the
and
let
(&>
1( to 2
at 0, supposed arranged in descending order of magnitude,

be the three components of angular velocity of the system
body
,
&>
3)
about the axes Ox, Oy, Oz respectively, so that
&>!
Bw<,
Ca) s
or
=
=
=
10, 62)
d sin 6 cos ty,

d sin 9 sin
d cos
ty,
0,
16)
sin
<
6 cos
sin 6 cos
-v/r
cj>
-fy
sin 6 sin
(f)
Y+
;
cos
^=
-j
-v/r
"
sin 6 cos
ty,
sin 6 sin
ty,
TY
cos
"
These are really three integrals of the differential equations of motion of

the system (only one arbitrary constant however occurs, namely d, our special
set of axes being such as to make the other two constants of
integration
zero);
we can
therefore take these instead of the usual Lagrangian differ

motion in order to determine 6,
ty.
ential equations of
Solving
for 6, 0,
<f>,
we have
-\jr,
,*
rj
v
(A-B)d Cl
j.
T\ t/
if \_v*Jo
Prid
O-L1-1
-j-j
I//"
\lf
Gl Tl
olll
lf/>
\Lf
A. tj
= d
-T-
q>
Y+
cos 2
(d
Y = \n
VO
^D sin
TJ
>
AJ
d
cos r
sm
BD
,\
r cos
"
The
may
integral of energy (which is a consequence of these three equations)

be written down at once by use of 63 it is
where
c is
a constant
replacing
o>
a>i,
2,
w s by
their values in terms of 6
and
this equation can be written in either of the forms

w. D.
10
146
A-B sm
~- + B-C
Bc-d*
c s
"
A-B
sm ZQ6 sin
2
AB
.
Since
>
d}
and (cC
positive,
positive or negative
The
2
-v/r
C, the quantity
>
is
^-
2
A-G
^lc-d---
or
"
Ad
(cA
negative
- d?)
the
AC
c s
>
cos 2
B (A -B)
or
quantity
we
for definiteness
of the three differential
first
[OH. vi
shall
0.
2
&>
(.Be
+ C (A 2
may
o>
to be positive.
it
suppose
C) 3 is
be either
equations may, by use of the last
equations, be written
Bc-dr-B-C
A-G
Ac-d
AdT
This equation shews that cos 6 is a Jacobian elliptic function* of a linear

and the two preceding equations shew that sin 6 cos ty and
function of t
sin 6 sin ty are the other two Jacobian functions.
;
We
therefore write
sin 6 cos
where P, Q,
i|r
= P en u,
sin 6 sin
are constants and u
-fy
Q sn u,
cos 6
a linear function of
is
= R dn
t,
say \t
u,
-f
the
quantities P, Q, R, A, and the modulus k of the elliptic functions, are then to

be chosen so as to make the above equations coincide with the equations
& cn
A? sn 2
= - k + dn u,
1
dn u,
u=
u
~r dn u
k 2 sn u en
du
The comparison gives

A(d*-cC}
u.
B(d*-cC)
d*(B-C)
d*(A-C)
2
(A-B)(d -cC)
C(cA-ffi)
d*(A-C}
(B -C)(cA-d*)
ABC
(B-C) (Ac -d*)

The equation
for
k2 shews that k
is real,
and the equation
(A-Q(Bc-.ffi)
(B-C)(Ac-d
shews that
(1
k2 )
is
positive,
i.e.
that
k<
1.
The
quantities P, Q, R,
X,
are
also evidently real.
Now
a real
quantity a
may be
defined by
the
mutually consistent
equations
sn la
*
=i
T-TT:
cma =
The theory of elliptic functions required in this and the succeeding problems
and Watson s Modern Analysis, Chs. xx. xxn.
in Whittaker
will be
found
147
69]
_i
Since
(A)
where the theta-functions are defined by the expansions
= 1 + 2q cos 2iri/ + 2g
^ (y) = 1 - 2g cos 27Ti/ + 2q-
+ 2g
cos 47TZ/
2g
cos 4?
(y)
01
and q = e~
% 0) = 2i
cos TTV
Sn (v)
sin ^TV
7>
%<f
2q* cos Sirv
+
+
4
2? sin STTV
cos
cos GTTV
cos 5?
2^"*"
2g"
+
.
.,
.,
sin
we have
+ 2g cosh 2 7 + 2g
1 -2g cosh 2 7 +2^
cosh
4-y
,,_i
-...~
cosh 4 7
where 7 stands for Tra/2K from this equation 7 (and consequently a) may
readily be determined by successive approximation.
:
The Eulerian angles 6 and
-^ at
sin 6 cos ^r
sin 6 sin
\|r
cos 6
time
are
=
cnza
dn
ia sn (\t + e)
cnm
sn la dn (\i +
^ -
e)
or (omitting the
now given by the equations
en la
e)
The modulus k
of the elliptic functions
is
known; we can
therefore
determine the parameter q of the theta-functions by the equation
or
by the more rapidly convergent

q
where cos
= (k y.
/9
and thus the period
OZ
is
2
$ tan
/3
series
+ TV tan
10
-$fa
tan 18
/3
.,
K may then be calculated from the series

4<K/\
of the inclinations of the axes
Oxyz
to the line
determined.
102
148
If
now we
write
=7
(>jra/2K)
and
K} =
(-TrX/2
/A,
[CH. vi
we have
- 2a cosh 27 -f 2o cosh47- ...)(cos u,t +

...
(1
sin
cos \f
cos
cos
+
+
+
+
(1
(cosh 7
2yu,
4/4
.)
(fcosh 87
-2q
2#
.)
+ ...)(sin /4 o sin 3/4 + ...)
(1 + 2a cosh 27 + 2o cosh
sin u sin
cosh 87 +
cos 2/4 + 2(? cos 4^ +...)
(cosh 7 +
) (1
2
__
~ (sinh 7 y sinh 87 + ...)(!+- 2g cos
(cosh 7 4- 9 cosh 87 + ...)(1
2g cos 2/j.t + 2q* cos 4/u^ + ...)
4
i/
. .
4>y
"vj/"
. .
2<?
<?
The
quantities
motion.
Example.
q,
p,
7 may be regarded
Suppose that the body
a=l,
principal
moments
^=
Suppose that the
7r
= 2,
= 3.
of inertia are
a6c(6
+ c 2 = 20-87T,
=1677-,
of energy
= J,
C0 2
=&*.
round the principal axes are
initial velocities of rotation

ft)!
The constant
which specify the
a homogeneous ellipsoid of unit density, whose
is
three semi-axes are
The three
as the constants
=i,
W 3 =l.
2= 13
is
= A a)! 2 +
2
Ha>2
and the constant of angular momentum
f<03
3r,
is
so
The modulus
of the elliptic functions is given
by the equation
(4-5X^-c)_
whence we have
-(.g-cw-rf 2
2=i_2 = 0.760,
)"
....
= 1-0342,
#=1-68013,
K
TT
We
have also
so
A =0-6045
and
M = |^=0-5651.
The period
of the angles 6
and
is -r-
or
which has the value 11 11 8.
69]
and
In order to express
^ as trigonometric
series in
terms of
149
we must determine y.
t,
For this we have
and therefore
if
"*
be neglected we have
1
+ 2? cosh 2y_
~
1-1094
0-9337
cosh 2y = 2 -503,
gi vi ng
2y = 1-568
and hence
and
y= 0-784.
The quantity a
then given by the equation
is
2A"
a=
7T
A
zero
y = 0-8385.
limiting case of the general problem is that in which A

B, so that k reduces to
In this case the solution may
elliptic functions become circular functions.
and the
be written
^=
^~cosha
cosX<
>b
<9
sin An
sm(9sm^ = cosh
(C(Ac-d 2 )}%
where
r>.
of
cos
\- \(A-C)(Ac-d^
A 2C
j
\
<
smha=\
.,
2
(A (d
~^
cosh a =
0=tanha
^J- cC)
(A (d^
so the motion
about
its
own
is
axis of
symmetry
Another limiting case
is
Oz.
that in which
degenerate into hyperbolic functions
Example
1.
cC)}
a steady precession about the invariable line OZ, the body rotating also
rigid body
is
d 2 = cB,
so that
this is illustrated
F=l
and the
elliptic functions
by the following examples.
moving about a fixed point under no forces: shew that
ij
2
and 3 being initially
Bc, and
2 is zero when t is zero,
(in the notation used above) d
the
direction-cosines
of the B-axis at time t, referred to the initial directions of
positive, then
a>
if<a
a>i
the principal axes, are
a tanh x
y sin
/*
sech
cos
^,
/j.
y tanh x + a
sech ^,
sin
p-
sech ^,
where
_dt
f-B
_dt
((A
- B) (B C)\%
-B\~-AC~-
= (A(B-C)\
\B(A-C}\
y=
(C(A-B)}^
\B(A-C]}
To
obtain this result,

coordinate Q becomes
dt
the integral of which
we observe that when Bc=d 2 the

,
BC
Ad 2
is
cos 6 = y sech ^,
I,
1899.)
differential equation for the
AC
150
where y and x are the quantities above defined.
The equation
A-B ......
Ac-d
0sm
j^-sm
J.D
A-C cos
---^2
u4d
\//-=
then gives
sin 6 sin -^
and the equation
d
= -7
cos
= tanh ^,
2)
-=
JD
\//-.
(<
cos
cos 6 sin
<
sin
\|^
cos
<
J(7
= y sin
sin
/*)
These equations shew that the direction-cosines of the
OXYZ, which ( 10) are
gives
[CH. vi
^ + cos
cos $ sin
sin
\^,
.5-axis referred to the axes
<p
cos
<
\|f,
sin $ sin
\^,
can be written
sin
But
if
o) 10
W2o,
p sech
denote the
0)30
^w = ao? and
,
p.
W + CW =
=
tanh
sech ^,
initial directions of
so that
cos
^,
^.
the principal axes, since
d 2 = Be = B ( A on 2 + Ccos2 ),
1
3
yd, we see that the direction-cosines of
are given by the scheme
C<o

10
0)20, 0)30,
are
a sin ^ sech x + 7 tanh ^.
^,
= cB,
line
shew that the axis 6ty describes, on a sphere with the fixed
with respect to the meridians passing through the invariable
line.
(Coll.
Returning now
angle
to the general case, we

in terms of the time.
have
Exam.)
have to express the third Eulerian
We
</>
Now
COt
sn
whence
This function of
vanishes,
i.e.
-vr
,/l
=
dn ia sn
\t
dn 2 ia sn 2
vanishes with
t,
\t
and has poles when the denominator
when
sn \t
=+
~
-,
k sn la
so in one period-parallelogram (2K,
\t = ia + iK
+ sn (ia +
2iK )
it
and \t =
iK }
has poles at the points

ia
+ iK
69]
Near the former of these points, writing

only the lowest powers of e, we have
2
l2
dn2 ia/& 2 sn 2
.
"
\t
= ia + iK +
e,
151
and retaining
m
dn 2 ia
k sn- ia
e&
2 sn ia en ia
dn
A;
sn2 ia
so the residue at this pole of sin 2 ty, considered as a function of \t, is
dn
2A;
ia
sn ia en ia
2id
(A)\
-
Therefore the residue of dl-f.
A]
VD
function of \t)
-G)(Ac- &)AB}*
C
((B
sin 2
at this point (considered as a
-\lr
is
ABC
and the residue when \t/2K is regarded as the variable is consequently i\/4<K.
As we now know the zeros, poles and residues of this function, we can write
down
its
expression as a sum of logarithmic derivates of theta-functions

^ 01 (i/) has a simple zero at v ^ = iK /2K, we have
fact, since
in
&>
>
/\t
4A"
^
MOl f**!2i
^ r
r:z= Mp^
J 01
/ ^-
-J 01
<v
ia
and therefore
m\
~
/X
I
constant
AT
iN
ow S 01
/\tîa\
(
2K/X
period
motion of
line.
<,
We
ZtJ^L
of
i.e.
^,
S- 01
//
- f\
purely periodic with respect to the real
have
so the coefficient of
27TI>
in
i.e.
</>,
2(?
COS
4t7TV
S-rrq* sin kirv
-..,
.
the constant part of
.,
or the precession,
<j),
is
may be
mean
the precessional motion of the system round the invariable
= 1 25- COS
ô/ (^) = 4?r^ sin ZTTV
written
q sinh 2y
oI
A
in
so the exponential on the right-hand side gives the
^01 (*
which
~\~A
is
(2id
I
:
/\t-\-ia\
/.
/
~9~V
which form
it
**,
may be
4
+ ...
2^ sinh
4
cosh 4 7 -
- 2q cosh 2 7 + 2^
calculated readily.
4<y
152
Example 1. In the
= 3, we have
6 = 2, c
case previously discussed, of an ellipsoid \vhose semi-axes are a =
2y=l
sinh2y = 2-294,
568,
7r,
so the
mean motion
of $, which
when j 4
is
d
M
cosh
^=
0-6956.
uniform circular disc has
The
in space, and
coinciding with
at any subsequent time
LQ2
arccot
4?l 2i
LQ2 + 4% 2
where ^
is
planes
0% and
the angle between
be written
centre
its
fixed,
and moves under the
disc is given initial angular velocities Q about a diameter

n about its axis coinciding with Of in space. Shew that
X = 2 arcsin f
503,
2-y
0-5986 + 0-0970,
2.
=2
=
1,
- 2q cosh 2y
or
action of no external forces.
may
neglected
q sinh
1
2-y
,
is
Example
[CH. vi
sin {(Q 2 + 4
-tan{(Q
)^
+ 4?i 2 )i.
It]
and the axis of the disc Oz and
o>
\,
is
the angle between the
(Oz.
(Coll. Exam.)
For let OZ denote as usual the invariable line, and consider the spherical triangle
Z&,
whose vertices are the intersections of the lines OZ, 0, Oz respectively with a sphere of
In this spherical triangle we have Zz = d,
centre 0.
disc
C=<2B=2A,
Zz = (fr.
Moreover we have
for the
so
and
The equations
of motion for 6
and
therefore
become
<f>
0=0,
<j>
= d/A
In the spherical triangle Zz, we have therefore
and hence
sin
and
cot
a>
= sin Z
sin
\Zz =
= cos Zf tan 1
^=
tan
(Q
+4n 2 )*
which are the required equations.

70.
Poinsot
kinematical representation of the motion; the polhode and
herpolhode.
An
elegant
method
of representing kinematically the motion of a

body
is the following, which is due to Poinsot*.
about a fixed point under no forces

*
Poinsot, Theorie nouvelle de hi rotation des corps, Paris, 1834.
69, 70]
The equation of the momental ellipsoid

moving axes Oxyz, is
of the
153
body at the fixed
point,
referred to the
Consider the tangent-plane to the ellipsoid which is perpendicular to the

If p denotes the perpendicular on this tangent-plane from
invariable line.
the origin,
we have
(since the direction-cosines of
= -r
which
is
are
Awjd, Bw^d,
Ca) 3/d)
constant.
d-
Since the perpendicular on the plane is constant in magnitude and

direction, the plane is fixed in space: so the momental ellipsoid always
touches a fixed plane.
if (x y z) are the coordinates of the
point of contact of the
and the plane, we have on identifying the equations
Moreover,
ellipsoid
Axx + Byy +
,11
rhp
UPC;
uiic vn
vdiiico
&>l
i//
pd
Czz
=1
_.,
and A(o x +
l
Mo
--
?/
G>2
-.
yc
Bw
pd
z?
y + Cw s z = pd
\/c
_1
i
_-
pd
yc
&>)
U>3
T~
and hence the radius vector to the point (x y z ) is the instantaneous axis
of rotation of the body.
It follows that the body moves as if it were rigidly
connected to its momental ellipsoid, and the latter body were to roll about the
,
a fixed plane perpendicular
fixed point on
to the
invariable line, without
angular velocity being proportional to the radius to the point of

contact, so that the component of angular velocity about the invariable line is
sliding
the
constant.
1.
If a body which is inoveable about a fixed point is initially at rest and
acted on continually by a couple of constant magnitude and orientation, shew that
Poinsot s construction still holds good, but that the component angular velocity about the
Example
then
is
invariable line
is
no longer constant but varies directly as the time.
For in any interval of time dt the addition of angular

about the fixed axis
system at time
t is
OZ
of the couple
Now
Nt about OZ.
principal axes of inertia Oxyz are

of inertia and
2
MS) are the
o>
(<!,
Aa>i=
momentum
(Coll.
to the
Exam.)
body
is
+Ydt
so that the resultant angular momentum of the

the components of angular momentum about the
C(o 3 where A, B, C are the principal moments

z
components of angular velocity: hence we have
A<a
JVtsin 6 cos
!//,
Bu>
Bu>.
= Nt sin
6 sin
\^,
C(o 3
= Ntcosd,
where
6, 0, -^ are the Eulerian angles which fix the position of the axes Oxyz with
reference to fixed axes OXYZ.
But these equations differ from those which occur in the
motion of a body under no forces only in the substitution of t dt for dt so the motion
be the same as in the problem of motion under no forces, except that the velocities are
;
will
multiplied by
let
whence the result
follows.
In the motion of a body, one of whose points is fixed, under no forces,

Example
a hyperboloid be rigidly connected with the body, so as to have the principal axes of
2.
154
[CH. vi
body at the point as axes, and to have the squares of its axes respectively
2
2
Ac, d
Bc, cP
Cc, where A, B, C are the moments of inertia of the
proportional to d
the
fixed
at
point, c is twice its kinetic energy, and d is the resultant angular
body
momentum. Shew that the motion of this hyperboloid can be represented by causing it
inertia of the
to roll without sliding on a circular cylinder, whose axis passes through the fixed point and
is parallel to the axis of resultant angular momentum.
(Siacci.)
The curve which
in Poinsot s construction
traced on the
is
momental
ellipsoid by the point of contact with the fixed plane is called the polkode.
Its equations, referred to the principal moments of inertia, are clearly the
equation of the ellipsoid together with the equation
p=
constant,
i.e.
they
are
Ax*
Example
Shew
I.
+ By- + Cz =
2
1,
when A =B, the polhode
that
is
circle.
^B^
C, shew that there are two kinds of polhodes, one kind

Example 2. Taking A
consisting of curves which surround the axis Oz of the momental ellipsoid, and correspond
to
cC, while the other kind consists of curves which surround the axis Ox, and
z
cB and that the limiting case between these two kinds of polhodes
correspond to cA
cB>d">
>d
>
a singular polhode which corresponds to cB

pass through the extremities of the mean axis.
is
The curve which

the moving ellipsoid
d2 = 0, and
consists of
two
ellipses
which
traced on the fixed plane by the point of contact with

called the herpolhode.
is
is
find the equation of the herpolhode, let p, % be the polar coordinates

of the point of contact, when the foot of the perpendicular from the fixed
To
If (x, y z } denote the coordinates

is taken as pole.
point on the fixed plane
of the same point referred to the moving axes Oxyz, we have
,
3/2
?.
_|_
square of radius from point of suspension to point of contact
-i
+#
""
Substituting for x
z their values as given
x =
y
{
=
=
=
=
/\/c
=
/\/c
by the equations
eoj/Vc
a).2
d sin B sin ty/B^c,
&>
sin 6 cos ty/A^/c,
d cos 6/C\/c,
we have
p-
=-
c
-=-
d?
d2
A-c
-.
sin 2
cos 2
T/T
d2
2
T sin
B-c
sin 2 -\lr
Replacing B and ^r by their values in terms of
_
cd?A*B*G*
_ (cA -
d*}
(c?
cd*AC
- cC)
t,
d2
+ TG~
this
cos 2 #.
becomes
(B-C)(A-
70, 71]
where
155
denotes the half-period corresponding to the root e^, this equation

expresses the radius vector of the herpolhode in terms of the time.
w
We
have next to find the vectorial angle
observe
that
six times the
in
terms of
t.
For
this
we
volume of the tetrahedron whose
Vcp ^/c?
vertices are the fixed point, the foot of the perpendicular from the fixed
point on the fixed plane, and two consecutive positions of the point of contact,
divided by the interval of time elapsed between these positions, and that this
is
quantity can also be expressed in the form

x,
y,
Acx \d\
Bey Id
Ccz /d2
A,
t x
/
B,
>
y/y,
the quantities involved, except %, are known functions

substituting their values in terms of t, and reducing, we have
All
of
on
A-C
which can be written in the form
_
%~
.
~B
(I
&))
This equation can be integrated in the same

way as the equation for the
Eulerian angle
and gives
c/>,
a
where ô
ta)
is a constant of
The current coordinates
integration.
herpolhode are thus expressed as functions of t.
(p,
^) of the
Example 1.
particle moves in such a way that its angular momentum round the
origin is a linear function of the square of the radius vector, while the square of its velocity
is
a quadratic function of the square of the radius vector, the coefficient of the
highest
power being negative shew that the path is the herpolhode of a Poinsot motion,
however J, JB, C are not restricted to be positive.
;
Jfxample
2.
secting on the
in
which
Discuss the cases in which the polhode consists of (a) two ellipses inter
of the momenta! ellipsoid, (/3) two parallel circles,
(y) two points
mean axis
shewing that in these cases the herpolhode becomes respectively a spiral curve (whose
equation can be expressed in terms of elementary functions), a circle, or a point.
71.
Motion of a top on a perfectly rough plane ; determination

of the
Eulerian angle
top
is
6.
defined to be a material body which is

symmetrical about an axis
the
(called
apex or vertex) at one end of
and terminates in a sharp point

the axis.
We
shall now study the motion of a

top when spinning with its apex
on
a
so
that
is
placed
perfectly rough plane,
practically a fixed point. The
156
[OH. vi
problem is essentially that of determining the motion of a solid of revolution

under the influence of gravity, when a point on its axis is fixed in
*.
space
Let (A, A, G) denote the moments of inertia of the top about

rectangular
axes Oxyz, fixed relative to the top and moving with it, the
origin being the
be
apex and the axis Oz being the axis of symmetry of the top let (6,
;
<,
\Jr)
the Eulerian angles defining the position of these axes with reference to fixed
rectangular axes OXYZ, of which OZ is directed vertically upwards.
The
kinetic energy
where w
&>
is (
63)
denote the components relative to the

moving axes of the
(
16) we have
angular velocity of the top, so that
6 sin ^r
&>j
co 2
-t-
&>s
6 sin
ty,
4>
the kinetic energy
sin 6 cos ty,
= 6 cos ty + sin
=^
cos 6
therefore
is
T = \AB- + Âft sin
+ ^C (jr +
<j>
cos 0?,
and the potential energy is V = Mgh cos 6, where

is the mass of the
top
and h is the distance of its centre of gravity from the apex.
The
kinetic potential
L=T-
is
therefore
V=$A6* + 4
The coordinates
and
<f>
+ 4C (t +
sin 2
cos
<
0^-Mghcoa B.
are evidently ignorable;
the corresponding
integrals are
dT =
dT
and
constant,
^-r
--.-
sin2
or
A<j>
where a and
momentum
b are
constant,
d-fr
d(f>
constants
about the axes
+ C (^ +
C ty +
these
OZ
fj)
cos 0) cos
- a,
cos
=b,
<9)
may be
interpreted as integrals of angular
and Oz, and so are obvious a priori from
general dynamical principles.
The modified
kinetic potential
R=L
38)
is
b\s
a(>
The term -b 2 /2C can be

equation of motion
cos
neglected, as
it
is
merely a constant; the
is
d /dR\
dR _
Lagrange, Mec. Anal. (Oeuvres, xn.
p. 251).
71]
so the variation of 6 is the
of freedom for
same
as in a dynamical system with one degree
which the kinetic energy

b cos 0)
(a
157
The connexion between d and

energy of this reduced system,
is
Â6 and
is
therefore given
is
by the integral of
namely
2
(a-bcos0)
Aa-2
\AQ = - 2A sin - - Mgh cos 6 + c,
^
where
c is
a constant.
= x, this equation becomes

A*a? = - (a - bx) - 2AMgh (x - a?) + 2 Ac (1 - x
Writing cos
The right-hand side

when x = 1, the cubic
of this equation
is
).
a cubic polynomial in x
now
some real values of 6, i.e. for some

values of x between
1 and 1, the cubic must be
positive, since the left-hand
side of the equation is positive; when x = I, the cubic is
again negative; and
when x = + oo the cubic is positive. The cubic has therefore two real roots
which lie between
1 and 1, and the remaining root is also real and is
is
negative; for
greater than unity.
Let these roots be denoted by

cos
where cos /3
The
>
cos
so that a
a,
\Mghft A$ dt =
If
we
we have
>
cos
a,
cosh
ft,
7,
/3.
now becomes
{4 (x
- cos a) (x -
cos
/3)
cosh 7 }}~*dx.
write
therefore
where the constants
+ constant =
el}
e?,
e 3 are
{4 (z
ej (z
e.2
)
e, e3 are
all real
el
and
2Ac + b
satisfy the relations
+e +e =
2
b2
~~
_Mgh
=
e-i,
(z
2Ac +
so that
(x
0,
el
>
e,
>
e3
158
The connexion between

where
therefore
t is
a constant of integration, and the function
e is
roots e 1} e2
and
fy
[CH. vi
is
formed with the
and hence we have
2A
Now in order that x may be real
for real values of t, it is evident that x must

between cos a and cos fi, i.e. $ (t + e) must lie between e2 and e3 for real
values of t and therefore the imaginary part of the constant e must be the
lie
The real part of e depends on

3 corresponding to the root e s
the epoch from which the time is measured, and so can be taken to be zero
half-period
<w
by suitably choosing
have therefore
finally
^
and
We
this epoch.
which expresses the Eulerian angle 6 in terms of
this is the equation
the time*.
Example
1.
If
the circumstances
sheiv that the value of Q at
any time
of projection of the top are such that initially
t is
Mgk
sec0=l+sech(^p/),
so that the axis of the
top continually approaches the vertical.
For
in this case
we
readily find for the constants a,
so the differential equation to determine
whence the
b,
the values
is
result follows.
solid of revolution can turn

freely about a fixed point in its axis of
acted on by forces derived from a potential-energy function p. cot 2 6, where
6 is the angle between this axis and a fixed line ; shew that the
equations of motion can be
integrated in terms of elementary functions.
2.
Example
symmetry, and
is
For proceeding as in the problem of the top on the perfectly rough plane,
the integral of energy of the reduced problem the equation
we
find for
(a -ft.
Writing cos 6 = x, this becomes
for
The quadratic on the right-hand side is negative when # = 1 and x= 1, but is positive
some values of x between
1 and +1, since the left-hand side is positive for some real
*
1,
55)
0,
It
may
when
P,
I,
be remarked that the present problem reduces to that of the spherical pendulum
k are replaced respectively by
M, C, A, h, a, b, c, cos 0,
I,
the quantities
k,
0,
h,
z/l,
<f>,
<f>,
\,
fjL.
159
71, 72]
the quadratic has therefore two real roots between

these cos a and cos /3, the equation is of the form
values of 6
\-x z = (cos a - x] (x
and
+ 1.
Calling
cos /3),
is
x = cos
a sin
2
(</2X)
+ cos /3 cos2 (Z/2X).
Determination of the remaining Eulerian angles, and of the Cayley-
72.
Klein parameters ; the spherical top.

When the Eulerian angle 6 has been obtained in terms of the time, as in
the last article, it remains to determine the other Eulerian angles
and
<f>
For
this
-\|/-.
purpose we use the two integrals corresponding to the ignorable
coordinates
when
these,
and
solved for
-\jr,
<f)
~Tsin2
give
b cos 6
_
=
b cos 0) cos 6
(a
C~
\^
Asin*0
we regard the motion
as specified by the constants of the

body
of integration (a, b, c), it is evident from
constants
and
the
(M, A, C, h)
that C does not occur except in
these equations and the equation for
for
and therefore an auxiliary top
the constant term of the expression
ty
can
be
whose moments of inertia are (A, A, A)
projected in such a way that
its axis of symmetry always occupies the same position as the axis of
symmetry
If
of the top considered, the only difference in the motion of the two tops being
that the auxiliary top has throughout the motion a constant extra spin
A)/AG about its axis of symmetry. A top such as this auxiliary top,
whose moments of inertia are all equal, is called a spherical top. It follows
therefore that the motion of any top can be simply expressed in terms of the
motion of a spherical top, and that there is no real loss of generality in
supposing any top under consideration to be spherical.
b(C
If then
we take
G= A,
_a
=
_
~
a cos
Substituting for cos
sin 2
24
<9
its
value from the equation
2A
and writing
and
<f>
i/r
a b
+b
~
- f)
2
A
(cos 0+1)
(cos
a+b
a b
=
+
1
2A
2X(cos + )
(coslT^Tj
b cos
A sin
the equations to determine
m- M9h
2
Mgh
2
12 A*
become
160
and k are known imaginary constants (being in fact the values of

and TT of 6}, the differential equations
corresponding to the values
so that
t
o>
[CH. vi
become
= Mgh(a + b)
Mgh (a - b)
Now
^ %/i (a
6)
the connexion between the function
once written
down by
substituting for
X=
6)
and its derivate

x from the equation
%>
can be at
Wh
in the equation
2
^} + 2Ac (1 - a?)
the argument of the ^-function is k, it follows from the definition of k that

1
and so the last equation gives
the corresponding value of x is
if
2
.
{2Ap (k)/Mgh\* = (a +
or
&>
Similarly
and therefore the equations
2^=*-,.
is
an
with
= iMgh (a + b)/ 2A-.
we have
>
Now
(k)
6)
= iMgh (a - b)/2A
(/)
and
for
-fy
</>
7\
2
,
...T.+
the function
whose poles in any period-parallelogram are congruent

= k, the corresponding residues being 1 and 1
3
Hence we have
zero when t + 3 = 0.
elliptic function,
t
+ w = k and t +
3
and the function
$ (t +
is
&>
o>
&>
3)
and therefore
(k}dt
v
log
&
cr (
-|-
V-
<w
A;)
/7x
bv( A;)y
~y^ + 2 ^
^<,
+ constant.
72]
The
and
-fy
tr (t
+ k) a- (t + a)
a-
integrals of the equations for
</>
161
can therefore be written in
the form
..
where
<
and
(t
&>
o) 3
a- (t
k)
o>
I)
/)
i/r
These equations lead to simple expressions for the Cayley-Klein parameters

7, 5 (
12), which define the position of the moving axes Oxyz with
reference to the fixed axes OXYZ: for by definition we have
a,
fi,
cos 1 6
e$
*& + +\
= ism $6.
{3
y = ismê.e^ ^-^,
1
= cos^.
e*
*\
{<t>
But we have
0=1 +cos0
2 cos 2
2A
~ (t +
or
cos \
Similarly
we
A \l {a- (t +
(k)
a>
a- (t
k)
Mgli
(0 3
<r
(t~+
k)
a-
k)
w3 )
(t
a) 3
k)}$
find
ft
sin
(T \
Tf
~T~
7\
^3
v ) O~
and on combining these with the expressions

we have
/4
\v
7\) i
*
I f
~T~ (J&v
for e
21 *
and
2i *
already found,
t)
o-
(0
o- (
i( ^" 0o)
7 =
<?(<
"
\M~gh
<r(l)~
tr (t
These equations express the parameters
a,
<o
ft>
3)
os )
7, 8 as
functions of the
time.
W. D.
11
162
Example
1.
M moves about a faced point
gyrostat of mass
[OH. vi
in its axis of
symmetry :
inertia about the axis of figure and a perpendicular to it through the fixed
and A respectively, and the centre of gravity is at a distance h from the fixed
gyrostat is held so that its axis makes an angle arccos l/ v/3 with the downward
moments of
the
point are
The
point.
vertical,
and
is
given an angular velocity \jAJMgh */3/C about
move about
free to
left
sin 2 6 sin
sin 2 6 cos
or
where
is the
20 = (
20 =
its axis.
If
the axis be
now
will describe the cone
it
- l/v/3)* ( - cos 6 + v /3) *,
- cos
(^3/2
+ cos 0)^,
azimuthal angle and 6 the inclination of the axis to the upward vertical.
For in this problem we have
initial
initially
= 0,
= 0,
a = - jMAghltlZ,
= 4/3 jKAgh, c= -
cos0=
and these
shew that
the fixed point,
-l/v/3,
^=
= 0,
values give
Substituting in the general differential equation for
namely
6,
we have
A6 2 sin 2
6-- Mgh (cos 6 + l/v/3)
N/3
+ 2 cos 6) - cos 6 + v/3),

(
while the equation
b cos 6
Dividing this equation by the square root of the preceding equation,
= si
I
= 3*
Now
if
_ cos 6 - l/v/3)*
v/3
+ 2 cos 6)
"
l(x-
1/^/3)* (v/3 -2a;)
(^
+ v/3)
*
(
~*
we have
~
cos 6 + ^3) i cosec 6 dd,
(1
-^)-i
ote,
where .r= -cos
we write
u = (x-
we have by
We
"
l/ N/3)
(x
+ N/3) *
(v
3/2
- x) ~ ^
differentiation
have therefore
tan
or
which
is
20 = 32
"
-cos
6- l/v/3) (-cos
equivalent to the result given above.
~
<9)
*,
72, 73]
Shew that the logarithms
2.
Example
functions of cos
6,
3.
Example
163
of the Cayley-Klein parameters, considered as
are elliptic integrals of the third kind.
Obtain the expressions found above for the Cayley-Klein parameters as

t
by shewing that they satisfy differential equations typified by
functions of the time
"
where I denotes a doubly-periodic function off, these equations being of the Hermite-Lame
type which is soluble by doubly-periodic functions of the second kind.
simple type of motion of the top is that in which the axis of symmetry
maintains a constant inclination to the vertical; in this case, which is
generally known as the steady motion of the top, 6 and 6 are permanently
zero
since
we have
b cos 0) 2
-
(a
it
follows that
d ((a
= -Ja\
n
b cos
2A
dd{
QY
sm ,a
2
COS
Performing the differentiation, and substituting

we have
A<f>sin-0,
=bcj>
+ Aft cos 6 +
/i
for (a
Mgh.
This equation gives the relation between the constants
depends on the rate of spinning of the top on

73.
We
b cos 6) its value
its axis) in
and b (which
(j), 0,
steady motion.
Motion of a top on a perfectly smooth plane.

now consider the motion of a top which
is spinning with its apex

smooth horizontal plane *. The reaction of the plane is now
the horizontal component of the velocity of the centre of gravity,
shall
in contact with a
vertical, so
G, of the top is constant we can therefore without loss of generality suppose

that this component is zero, so that the point G moves vertically in a fixed
two horizontal lines fixed in space and
line, which we shall take as axis of Z
;
perpendicular to each other will
the
(0,
be taken as axes of
X and
Y.
Let Gxyz be the principal axes of inertia of the top at G, and (A, A, C)
moments of inertia about them, Gz being the axis of symmetry and let
be the Eulerian angles defining their position with reference to the
-\/r)
:
</>,
axes of X, Y, Z.
of G above the plane is h cos 6, where h denotes the distance
from the apex of the top the part of the kinetic energy due to the
motion of G is therefore ^J/A2 sin 2 6 6 2 where
is the mass of the top
and
The height
of
so,
as in
71, the total kinetic
energy
is
cos

*
2
,
is
= Mgh cos6.
Poisson, TraitS de Mtcanique (1811), n. p. 198.
112
164
Proceeding now exactly as in

and
to the ignorable coordinates
71,
ty,
</>
sin 2 6
\A<j>
we have two
b are constants;
we obtain
coordinates
for
(A + Mh*
integrals corresponding
namely
+ G (^ +
<
cos 6) cos 6
= a,
cos 0)
= b,
C(^+(j)
where a and
[OH. vi
and on performing the process of ignoration of
the modified kinetic potential the expression
P-(
^~~? - Mgh
sin* 0)
cos
0,
is the same as in the system with one degree of freedom

which the kinetic energy is
so the variation of 6
for
(A
+ Mh* sin
0)
fr,
is
(a
b cos
^-
2 J. sin 2
The connexion between

latter system,
and
BY
~
+ Mqh
,
t is
cos
0.
given by the integral of energy of this
namely
$(A+ Mh* sin
0) ft
= - (a
"
-/I
6 C
sin;
)2
Jfyfe cos
+ c,
c/
= x, this becomes
A (A + Mh - Mtta?) x =-(a- bx} - ZAMgh (x - a?) + 2Ac (1 - x
where
c is
a constant.
Writing cos
).
are separated in this equation, so the solution can

be expressed as a quadrature but the evaluation of the integral involved
will require in general hyperelliptic functions, or automorphic functions of
genus two.
The
variables
x and
74.
Kowalevski
s top.
of the motion under gravity of a body one of whose points is

not in general soluble by quadratures and the cases considered in
69 (in which the fixed point is the centre of gravity of the body, so that
71 (in which the fixed point
influence the motion), and in
gravity does not
of
axis
lie
on
an
and the centre of gravity
symmetry of the body), were for
known to be integrable. In 1888 however Mme. S.
long the only ones
Kowalevski* shewed that the problem is also soluble when two of the
The problem
fixed
is
principal moments
=
third, so that
of inertia at the fixed point are equal and double the

= 2C, and when further the centre of gravity is situated
in the plane of the equal
moments
of inertia.
and the centre of gravity be taken

through the fixed point
the fixed
of
centre
let
the
and
as the axis Ox,
gravity be at a distance a from
Let the
line
* Acta
Math. xn. (1888),
p. 177.
73, 74]
165
be the Eulerian angles which define the position of the

principal axes of inertia Oxyz with reference to fixed rectangular axes OX YZ,
co 3 ) be the
of which the axis OZ is vertical; let
2
components along the
point
let (6,
<f>,
i/r)
&>
(<uj,
be
axes Oxyz of the angular velocity of the body, and let
kinetic and potential energies are given by the equations
T = i (Aw* + Aw? +
)
= C{fc + sin 6 + (^ +
its
mass.
The
Ca>
<j>
V=
The coordinate
is
<j>
Mga
sin
cos
</>
cos 0) 2 },
ty.
evidently ignorable, giving an integral
dT =
constant,
;r-r
Off)
or
20
where k
is
a constant
sin 2 9
+ (-^ +
<j>
and the integral of energy
= k,
cos 0) cos
is
T + V = constant,
2
or
2
<
(^ +
sin2
Mme. Kowalevski shewed
cos 0) 2
<j>
= A.
sin 6 cos -f
that another algebraic integral exists, which can
be found in the following way.
The
kinetic potential
L=C0* +
is
+ ^C(^r +
sin2
C<j>
<j>
cos 0)~
+ Mga
sin
cos
-i/r,
and the equations of motion are
^(?^_ W =
dt
\M)
dt
d
dt
the
first
of these
is
20
cos d
(rf>
- -dr) 6 sin 6 +
cos
cos
ty,
(j
and on eliminating
2
-fy
(4>
between the second and
sin 0)
rit
Adding the
2
first
sin0
-=
((j>
= - (A cos
0-<dr)
third,
0+^
of these equations multiplied
+ t 0)=i
cos
(<
sin0
-^)(<f>
we obtain
by
cos
sin
^.
to the se.cond,
+ i0)+i
we have
-^ cos 0e~^,
166
an equation which can be written

sin 6
+ i6)* +
sin
=i
cos ^
7 = (i sin
where
=~
(4 sin
i6)-
sn
~"
n
~
+ i0) +
F= (0 sin (9- i8? +

"FT"
Similarly, if
we have
form
1
- ijr)
cos
(<
or
in the
[CH. vi
(0 cos ^
sin
sin
It follows that
d7
J.
.,.
tf dt
dt
UV = constant.
or
We
have therefore the equation
+ id)
sin
(<j>
7^-sin 0ar~*K j(0 sin
i6)
sin $6**
"^~
constant,
or
2
($
2
</>
sin 2 #) 2
+ (~^r
sin 2 #-|
+ i0) +e~ ((j)sm0i0)

2
l *l
^-sin0{e
((j>sm0
i *l
2]>
= constant,
and
this is the
required third algebraic integral of the system.
The first integrals which have been found constitute a system of three
differential equations, each of the first order, for the determination of 0,
(j), ty,
and they can be regarded as replacing the original
differential equations of
does not occur explicitly in them and we can there
fore use one of the three equations in order to eliminate
from the other two:
we shall then have a system of two differential equations, each of the first
motion.
The
variable
<f>
<
order, to
determine
and
-fy.
It has
been shewn by Mme. Kowalevski that
these equations can be solved by means of hyperelliptic functions

solution reference may be made to the memoir already referred to*.
for this
Cf. also Kotter, Acta Math. xvn. (1893), p. 209

Stekloff, Gorjatscheff, and Tcbapligine,
Trav. Soc. Imp. Nat. Moscou, x. (1899) and xn. (1904) G. Dumas, Nonv. Ann. (4) iv. (1904),
p. 355; Husson, Toulouse Ann. (2) vm. (1906), p. 73; Husson, Acta Math. xxxi. (1907), p. 71;
;
N. Kowalevski, Math. Ann. LXV. (1908), p. 528; P. StackeJ, Math. Ann. LXV. (1908), p. 538;
0. Olsson, Arkiv for Mat. iv. Nr. 7 (1908); E. Marcolongo, Rom. Ace. Rend. (5) xvn. (1908),
F. de Brun, Arkiv for Mat. vi. Nr. 9 (1910) P. Burgatti, Palermo Rend. xxix. (1910),
p. 698
a
p. 396; 0. Lazzarino, Rend. d. Soc. reale di Napoli, (3 ) xvn. (1911), p. 68.
;
74, 75]
167
Let y 1 y 2 y3 denote the direction-cosines of Ox, Oy, Qz referred to OZ, and

be denned by the equations
Example.
let variables x, y, r
2o>
o>
Mgay->
I
-,
0)30)!
-2a>
0,3(0!
w2
+ a, 3 2 a, 2
Shew by use
momentum)
of Kowalevski s integral (without using the integrals of energy or angular

that the equations of motion can be written in the form
where V is a function of x and y only, so that the problem

motion of a particle in a plane conservative field of force.
is
transformed into that of the

(Kolosoff.)
E. Liouville* has stated that the only other general case in which the motion under
gravity of a rigid body with one point fixed has a third algebraic integral is that in which
The momental
The centre of
1.
2.
ellipsoid of the point of suspension is
gravity of the body
an
ellipsoid of revolution.
in the equatorial plane of the
is
momental
ellipsoid.
3.
ratio
If
2C/A
On
J, A, C) are the principal
an integer
is
this, cf.
moments
of inertia at the point of suspension, the
this integer can be arbitrarily chosen.
the memoirs cited in the footnote on the preceding page.
Example. A heavy body rotates about a fixed point 0, the principal moments of
inertia at which satisfy the relation A = B = 4C and the centre of gravity of the body lies in
the equatorial plane of the momental ellipsoid, at a distance h from 0. Shew that if the
constant of angular momentum about the vertical through vanishes, there exists an integral
:
w3
where
2
(o>!
+ co 22
) +gha>i
cos 6 = constant,
are the components of angular velocity about the principal axes Oxyz,
to the centre of gravity and hence that the problem can be
Ox being the line from
solved by quadratures, leading to hyperelliptic integrals.
(Tchapligine.)
coj,

75.
Impulsive motion.
As has been observed
in
36, the solution of problems in impulsive
motion does not depend on the integration of differential equations, and can
The following examples
generally be effected by simple algebraic methods.
illustrate various types of impulsive systems.
Two uniform rods AB, BC, each of length 2a, are smoothly jointed at B
horizontal table with their directions at right angles. An impulse is applied to
the middle point of AB, and the rods start moving as a rigid body : determine the direction
Example
and rest on a
1.
of the impulse that this

ratio
V13
may
be the case,
and prove
that the velocities of A,
1.
C will
(Coll.
be in the
Exam.)
We can without loss of generality suppose the mass of each rod to be unity. Let (x, y}
be the component velocities of B referred to fixed axes Ox, Oy parallel to the undisturbed
*
Acta Math. xx. (1897),
p. 239.
168
[CH. vi
be the angular velocities of BA and BC. The

BA, BC of the rods, and let 6,
components of velocity of the middle point of AB are (x, y + ad), and the components of
y), so the kinetic energy of the system is
velocity of the middle point of BC are (x
position
<
a<,
Let the components parallel to the axes of the impulse be
The components
/, J.
of
the displacement of the point of application of the impulse in a small displacement of the
system are (bx, By + add) and hence the equations of 36 become
;
0= -ax + â 2
while the condition that the system moves as
\x=y=lae =
Hence /=/, which shews that the
and as the components of velocity
are (x
2a<,
we have
y),
Example
lai>
=
<p.
= \I=lJ.
direction of the impulse makes an angle of 45 with BA

of A are (x, y + 2a0), and the components of velocity of
;
for the velocities of
respectively, so the velocity of
),
if rigid is
is
and of C the values \/65y and *Jby
^/13 x the velocity of C: which
is
the required result.
A framework in
2.
the ends of tivo pairs of
the form of a parallelogram is made by smoothly jointing

uniform bars of lengths 2a, 26, masses m, m and radii of gyration
,
moving without any rotation of its sides, and with velocity V, in

the direction of one of its diagonals; it impinges on a smooth fixed wall with which the sides
make angles 6, $ and the direction of the velocity V a right angle, the vertex which impinges
being brought to rest by the impact. Shew that the impulse on the wall is
k,
The parallelogram
is
(Coll.
Exam.)
Let x and y be the coordinates of the centre of the parallelogram, x being measured at
right angles to the wall and towards it. The kinetic energy is
T= (m + m
(x
+y 2 + (mk2 + m a2 2 + (mb 2 + m
contact is # + asin# + &sin0, so
The ^-coordinate of the point of

the axis of
point of contact parallel to
(&r,
By,
B6,
impulse by
80)
/,
is
2
k"
2
)
the displacement of the
x corresponding to an arbitrary displacement

The equations of motion, denoting the
S.r + acos#50 + &cos080.
are therefore
fdT_ (dT\
_ _
dx
/^T\
la cos
- Ib cos
0,
80/o
(mk
(mb
+ m a2
+m k 2
6

is zero,
we have
= 0.
.
<f>
from these equations, we have
<
[2
which
cos
169
b 2 cos 2
(m + mf)
the result stated.
The next example relates to a case of sudden fixture if one point (or line)
of a freely-moving rigid body is suddenly seized and compelled to move in a
given manner, there will be an impulsive change in the motion of the body,
;
which can be determined from the condition that the angular momentum of
the body about any line through the point seized (or about the line seized)
is unchanged
by the seizure this follows from the fact that the impulse of
;
seizure has
Example
no moment about the point (or

3.
line).
uniform circular disc is spinning with an angular velocity Q about a

P on its rim is suddenly fixed. Prove that the subsequent velocity of
J of the velocity of the point P immediately before the impact.
diameter when a point

the centre is equal to
(Coll.
Exam.)
disc, and let a be the angle between the radius to P and the
diameter about which the disc was originally spinning.
The original velocity of P is
Q.C sin a, where c is the radius of the disc.
The original angular momentum about P is
2
about an axis through P parallel to the original axis of rotation, and of magnitude
^mc Q.
and this is unchanged by the fixing of P, so when P has been fixed, the angular momentum
Let
be the mass of the
about the tangent at

tangent at
P is f me
is
rac 2 I2 sin a.
But the moment
of inertia of the disc about its
and
so the angular velocity about the tangent at

is ^ 12 sin a.
The
of the original velocity
velocity of the centre of the disc is therefore Jflcsina, which is
of P.
,
lamina in the form of a parallelogram whose mass is

has a smooth
Example 4.
It is struck at an angular
pivot at each of the middle points of two parallel sides.
point
which adheres to it after the blow.
Shew that the impulsive
by a particle of mass
reaction at one of the pivots is zero.
(Coll.
Exam.)
1.
Prove that for a disc free to turn about a horizontal axis perpendicular to its
plane
the locus on the disc of the centres of suspension for which the
simple equivalent
pendulum has a given length L consists of two circles and that, if A and B are two
;
points, one
on each
circle,
and
the length of the simple equivalent pendulum when

the middle point of AB, the radius of gyration k of the disc
is
the centre of suspension is

its centre of inertia is given
by the equation
about
where 2c
is
the length of AB.
(Coll.
Exam.)
heavy rigid body can turn about a fixed horizontal axis. If one point in the
body is given through which the horizontal axis has to pass, discuss the problem of
choosing the direction of the axis in the body in such a way that the simple equivalent
pendulum shall have a given length shewing that the axes which satisfy this condition are
2.
the generators of a quartic cone.
(Coll.
Exam.)
170
3.
sphere of radius b
rolls
without slipping down the cycloid
y = a(l
x=a(0 + sin6),
It starts
from rest with
of its centre
when
its
[CH.
costf).
centre on the horizontal line
at the lowest point is given
y = 2a.
Prove that the velocity
by
F 2 =-W (2o- 6).
Exam.)
(Coll.
rests symmetrically on two shelves

uniform smooth cube of edge 2a and mass
each of breadth b and mass m and attached to walls at a distance 2c apart. Shew that, if
one of the shelves gives w&y and begins to turn about the edge where it is attached to the
wall, the initial angular acceleration of the cube will be
4.
:-& + a) 2
where Mb and / are respectively the
of the shelf about its edge.
2
moments
of inertia of the cube about its centre
I,
and
1899.)
5.
A homogeneous rod of mass and length 2a moves on a horizontal plane, one end
being constrained to slide without friction in a fixed straight line. The rod is initially
perpendicular to the line, and is struck at the free end by a blow / parallel to the line.
Shew that after time t the perpendicular distance y of the middle point of the rod from
the line
is
(l-$ xtf(l-x )-%dx=3lt/2Ma.

2
(Coll.
Exam.)
6.
Four equal uniform
rhombus ABCD. The joint
rods, of length 2a, are smoothly jointed so as to form a

is fixed, whilst C is free to move on a smooth vertical rod
through A.
Initially C coincides with A and the whole system is rotating about the
Prove that, if in the subsequent motion 2a is the least
vertical with angular velocity
angle between the upper rods,
o>.
w 2 cos a = 3g
sin 2 a.
I,
1900.)
A disc of mass M rests on
a smooth horizontal table, and a smooth circular groove

A particle of
of radius a is cut in it, passing through the centre of gravity of the disc.
is started in the groove from the centre of gravity of the disc.
mass
Investigate the
7.
^M
Prove that if a$
round by the disc, then
motion.
Mk 2
being the
moment
is
the arc traversed by the particle and 6 the angle turned
of inertia of the disc
about a vertical
line
of gravity.
8.
velocity
motion.
latus
G>
through
(Coll.
its
centre
Exam.)
body is moving freely under the action of gravity and rotating with angular
about an axis through its centre of gravity perpendicular to the plane of its
Shew that the axis of instantaneous rotation describes a parabolic cylinder of
rigid
rectum (^/4a+
*j2g/ca}
whose vertex is at a distance vSâ/w above that of the path

where 4a is the latus rectum of the parabola
of the centre of gravity of the body

described by the centre of gravity.
(Coll.
Exam.)
smooth uniform tube which can rotate in a

A particle of mass m
9.
The system starts from rest when the tube is
vertical plane about its middle point.
If 6 is the angle the tube makes with the vertical when its angular velocity is
horizontal.
a maximum and equal to
prove that
is placed in a
o>,
4 (mr* + J/F)
4
&>
- Smgrco 2 cos 6 + mg 2 sin 2 6 = 0,
The
vi]
where
Mk z
is
the
Soluble Problems of Rigid
moment
of inertia of the tube about its centre and r the distance of the
from the centre of the tube.
particle
171
Dynamics
(Coll.
Exam.)
10.
Four uniform rods, smoothly jointed at their ends, form a parallelogram which
can move smoothly on a horizontal surface, one of the angular points being fixed.
Initially the configuration is rectangular and the framework is set in motion in such a
manner that the angular velocity of one pair of opposite sides is Q, that of the other pair
being zero. Shew that when the angle between the rods
angular velocity of the system is Q.
is
maximum
or
minimum, the
(Coll.
Exam.)
11.
Two homogeneous rough spheres of equal radii a and of masses m, m rest on a
smooth horizontal plane with m at the highest point of m. If the system is disturbed,
shew that the inclination of their common normal 6 to the vertical is given by the
equation
(l-cos<9).
12.
uniform rod
sible string of length
c.
AB
is
(Coll.
Exam.)
of length 2a and is attached at one end to a light inextenof this string is fixed at
to a point in a smooth
The other end
horizontal plane on which the rod moves.

projected without rotation with velocity
Initially
GAB
is
a straight line and the rod
Prove that the cosine of the greatest subsequent angle between the rod and string
l-/6c.
13.
them
is
in the direction perpendicular to its length.
(Coll.
is
Exam.)
To a
slides,
fixed point arc smoothly jointed two uniform rods of length 2a, and upon
by means of a smooth ring at each end, a third rod similar in all respects.
Initially the three rods are in a horizontal line
with the ends of the third rod at the

middle points of the other two and, on the application of an impulse, the rods begin to
rotate with angular velocity Q in a horizontal plane.
Shew that the third rod will slide
right off the other
two unless
J3.
(Coll.
Exam.)
hollow thin cylinder of radius a and mass

14.
is maintained at rest in a
horizontal position on a rough plane whose inclination is a, and contains an insect of mass
at rest on the line of contact with the plane.
The cylinder is released as the insect
starts off with velocity V: if this relative velocity be maintained and the cylinder roll up
shew that it will come to instantaneous

makes an angle d with the vertical given by
hill,
F 2 {1 - cos (0-a)}+ag (cos a -cos
rest
<9)
when the
radius through the insect
= (1 +M/m) ag (d-a)
sin
a.
(Coll.
Exam.)
A uniform
smooth plane tube can turn smoothly about a fixed axis of rotation
and intersecting it the moment of inertia of the tube about the axis
is 1.
Initially the tube is rotating with angular velocity Q about the axis, and a particle
of mass m is projected with velocity V within the tube from the point of intersection of
the tube with the axis. The system then moves under no external forces. Prove that,
when the particle is at a distance r from the axis, the square of its velocity relative to the
15.
lying in its plane
tube
is
1
Q2
16.
(Coll.
Exam.)
uniform straight rod of mass

is laid across two smooth horizontal
pegs so
its ends projects beyond the corresponding peg.
A second uniform rod of
and length 21 is fastened to the first at some point between the pegs by a
that each of
mass
172
[CH.
universal joint. This rod is initially held horizontal and in contact with the first rod and
then let go, so as to oscillate in the vertical plane through the first rod. Prove that if 6
:
be the angle which the second rod makes with the vertical at any instant, and
distance through which the first rod has moved from rest,
x the
(M+m) x + ml sin 6 = ml,

and
(td
--
17.
body
^>
6\
ifr
= 2g cos 6.
(Coll.
Exam.)
rotate in its plane about a fixed point, and a second plane

smooth straight groove in the first body, its motion being in
shew that the relation between the relative advance x along the groove
plane body
is free
m + M cos
is free to
to slide along a
the same plane

and the angle of rotation & (no external forces being supposed to act on the system)
is of the form
;
where
P and Q are
respectively linear
and quadratic functions of x 1
(Coll.
-.
Exam.)
pendulum is formed of a straight rod and a hollow circular bob, and fitting
bob is a smooth vertical lamina in the shape of a segment of a circle, the
distances of the centre (C) of the bob from the point of suspension (0) and from the
Prove that if M, m are the
centre of gravity (G} of the lamina being I and c respectively.
masses of the pendulum and lamina, k and 1J their respective radii of gyration about
and G, 6 and $ the angles which OC and CG make with the vertical, then twice the
work done by gravity on the system during its motion from rest is equal to
18.
inside the
2
(
+c 2
2
)
<j>
+ 2mcZcos (6-$)
6$.
(Coll.
Exam.)
is attached to the end of a fine string which passes round

19.
particle of mass
the circumference of a wheel of mass M, the other end of the string being attached to a
point in that circumference, a length I of the string being straight initially, and the wheel
(radius a and radius of gyration k) being free to move about a fixed vertical axis through
centre; the particle, which lies on a smooth horizontal plane, is projected at right
angles to the string, so that the string begins to wrap round the wheel prove that, if the
is
string eventually unwinds from the wheel, the shortest length of the straight portion
its
(P
20.
carriage
rolls straight
is
- a 2 - J/F/m)*.
(Coll.
Exam. )
the horizon and

placed on an inclined plane making an angle a with
The floor of
and
the
wheels
the
between
plane.
slipping
down without any
the carriage is parallel to the plane and a perfectly rough ball

that the acceleration of the carriage down the plane is
is
placed freely on
it.
Shew
is the mass of the carriage excluding the wheels, m the sum of the masses of the
that of the ball. The friction between the wheels
which
are uniform discs, and
wheels,
and the axes is neglected.
(Coll. Exam.)
where
A uniform rod of mass m t and length 2a is capable of rotating freely about its
upper extremity and is initially inclined at an angle of Tr/6 to the vertical. A second
to the lower end of the first and rests
rod, of mass m 2 and length 2a, is smoothly attached
Shew that if the centre
initially at an angle of 2-/3 with it and in a horizontal position.
of the lower rod commence to move in a direction making an angle -rr/G with the vertical,
21.
fixed
then 3wH = 14m 2
(Coll.
Exam.)
vi]
22.
uniform circular disc
173
symmetrically suspended by two elastic strings of
is
natural length c inclined at an angle a to the vertical, and attached to the highest point of
the disc. If one of the strings is cut, prove that the initial curvature of the path of the
centre of the disc
is
(c
where
sin 4a
b sin 2a)/6 (b
c),
b is the equilibrium length of each string.
Exam.)
(Coll.
AC
rods AC, CB of equal length 2a are freely jointed at (7, the rod
being
freely moveable about a fixed point A, and the end B of the rod CB is attached to A by
an inextensible string of length 4a/ x/3. The system being in equilibrium, the string is cut
Two
23.
shew that the radius of curvature of the
initial
path of
at
is
_!_
"
181V
24.
rod of length 2a
is
supported in a horizontal position
by two
I,
1897.)
light strings
which
pass over two smooth pegs in a horizontal line at a distance 2a apart and have at their
other extremities weights each equal to one half that of the rod. One of the strings is cut
prove that the initial curvature of the path of that end of the rod to which the cut string
;
was attached
is
27/25.
A heavy plank,
25.
Exam.)
(Coll.
straight
and very rough,
free to
is
turn in a vertical plane about
its centre of gravity is c.

A rough heavy
placed on this plank at a distance b from the axis, on the side remote from the
centre of gravity the plank being held horizontal.
The system is now left free to move.
a horizontal axis from which the distance of

sphere
is
Prove that the
radius of curvature of the path of the centre of the sphere is

Mc )j(mb + Ma ) m and
are the masses of the sphere and
110), where d = (tnb
21&#/(5
the plank, and Mob is the moment of inertia of the plank about the axis.
initial
(Coll.
A light stiff rod of length
Exam.)
two equal particles of mass m at distances k

from the centre on each side of it to each end of the rod is tied an end of an inextensible
string of length 2a on which is a ring of mass m
Initially the string and rod are in one
straight line on a smooth horizontal table with the string taut and the ring at the loop
the ring is then projected at right angles to the rod, shew that the relative motion will be
26.
2c carries
oscillatory if
c2
27.
Three equal uniform
/P
>
+ 2i/m
rods, each of length
Exam.)
(Coll.
are firmly joined to form an equilateral

a uniform bar of length 26 and weight
is freely jointed to
c,
ABC of weight W;
the triangle at C. This system rests in equilibrium in contact with the surface of a fixed
smooth sphere of radius a,
being horizontal and in contact with the sphere, and the
bar being in the vertical plane through the centre of the triangle the bar, and the centre of
triangle
AB
the triangle, are on opposite sides of the vertical line through C. Prove that the inclination
of the plane of the triangle to the horizon is the angle whose tangent is
[abfj.
where
+ 2c\ 2 ]
= a -rc -^c,
2
-T-
[np (a
2
+ 1 c2 + A V - 2a6c]
)
= 12a 2 -c2
and
n=W/W.

28.
I,
1896.)
A body, under the action of no forces, moves so that the resolved
part of its angular

constant shew that the
velocity about one of the principal axes at the centre of gravity is

angular velocity of the body must be constant, and find its resolved parts about the other
two principal axes when the moments of inertia about these axes are equal.
;
(Coll
Exam.)
174
Shew
29.
[CH.
that a herpolhode cannot have a point of inflexion.
(A simple proof of
xxxiv. (1906),
p.
this result is given
by Lecornu,
(Hess.)
Bull, de la Soc. Math, de France,
40.)
In the motion under no forces of a body one of whose points is fixed, shew that the
30.
motion of every quadric homocyclic with the momental ellipsoid relative to the fixed point,
and rigidly connected with the body, is the same as if it were made to roll without sliding
on a fixed quadric of revolution, which has its centre at the fixed point, and whose axis is
the invariable
line.
(Gebbia.)
In the motion of a body under no forces round a fixed point, shew that the three
31.
diameters of the momental ellipsoid at the fixed point and the diameter of the ellipsoid
reciprocal to the momental ellipsoid, determined respectively by the intersection of the
invariable plane with the three principal planes and with the plane perpendicular to the
instantaneous axis, describe areas proportional to the times, so that the accelerations of
their extremities are directed to the centre.
(Siacci.)
When a body moveable about a fixed point is acted on by forces whose moment
32.
round the instantaneous axis is always zero, shew that the velocity of rotation is
proportional to that radius vector of the momental ellipsoid which is in the direction
of this axis.
Shew that
this
theorem is still true if the body

on a fixed surface.
is
fixed point and

(Flye St Marie.)
moveable about a
also constrained to slide
plane lamina is initially moving with equal angular velocities Q about the
axes
of greatest and least moment of inertia at its centre of mass, and has
principal
no angular velocity about the third principal axis express the angular velocities about
these axes as elliptic functions of the time, supposing no forces to act on the lamina.
33.
If 6
be the angle between the plane of the lamina and any fixed plane, shew that

34.
A rigid
body
is
I,
1896.)
kinetically symmetrical about an axis which passes through a fixed

is set in motion in any manner
shew that in the
point above its centre of gravity and
subsequent motion, except in one case, the centre of gravity can never be vertically over
the fixed point
and
find the greatest height
it
attains.
(Coll.
Exam.)
In the motion of the top on the rough plane, shew that there exists an auxiliary
also with respect
OTJ whose motion with respect to the fixed axes OXYZ
the invariable planes being the horizontal
to the moving axes Oxyz is a Poinsot motion
35.
set of axes
&n<\
plane in the former case, and the plane perpendicular to the axis of the body in the second
case.
(Jacobi.)
uniform solid of revolution moves about a point, so that its motion may be
represented by the uniform rolling of a cone of semi-vertical angle a fixed in the body upon
an equal cone fixed in space, the axis of the former being the axis of revolution. Shew that
the couple necessary to maintain the motion is of magnitude
36.
Q 2 tan
{C+ (C- A]
cos 2a},
the resultant angular velocity and A and C the principal moments of inertia at
the point, and that the couple lies in the plane of the axes of the cones. (Coll. Exam.)
where Q
is
vi]
37.
made
vertical plane is
to rotate with
175
uniform angular velocity about a vertical
itself, and a perfectly rough cone of revolution has its vertex fixed at a point of
that axis. Shew that, if the line of contact make an angle 6 with the vertical, and /3 and y
be the extreme values of #, and a be the semi-vertical angle of the cone,
axis in
K2
A?#\ 2 _
-
-,
&Q n,
\dt /
where h
is
sin 2 a (cos & - cos /3) (cos y

cos a
cos @ cosy
- cos
the distance of the centre of gravity of the cone from
of gyration about a generator.
its vertex,
and k
I,
its
radius
1896.)
body can rotate freely about a fixed vertical axis for which its moment of
/ the body carries a second body in the form of a disc which can rotate about
a horizontal axis, fixed in the first body and intersecting the vertical axis. In the
position
of equilibrium the moments and product of inertia of the disc with regard to the vertical
and horizontal axes respectively are A, B, F. Prove that if the system start from rest
38.
inertia
is
with the plane of the disc inclined at an angle a to the vertical, the
through an angle
IF
arctan
f^sinal
-
first
body
will oscillate
Exam. )
(Coll.
}-
gyrostat consists of a heavy symmetrical flywheel freely mounted in a heavy

case
and is suspended from a fixed point by a string of length I fixed to a point
spherical
in the case.
The centres of gravity of the flywheel and case are coincident. Shew that,
39.
the whole revolve in steady motion round the vertical with angular velocity
and the axis of the gyrostat inclined at angles a, to the vertical, then
if
the strin-
i2,
Q2
/Q
and
sin
- Afl 2
j3
(I
sin a +
sin
/3
sin
cos /3=
/3
+ b cos
Mg sec a
=g tan a,
(a sin
a)
(/3
+ b cos
(/3
a)},
the mass of the gyrostat, a and b the coordinates of the point of attachment
of the string with reference to axes coinciding with, and at right angles to, the axis of the
flywheel, / the angular momentum of the flywheel about its axis and A the moment of
whore
is
about a line perpendicular to
inertia
its axis.
I,
1900.)
system consisting of any number of equal uniform rods loosely jointed and
the same straight line is struck at any point by a blow perpendicular to the rods.
Shew that if u, v, w be the initial velocities of the middle points of any three consecutive
40.
initially in
rods,
u + 4;V + w = Q.
(Coll.
Exam.)
41.
Any number of uniform rods of masses A, B, C, ..., Z are smoothly jointed to
each other in succession and laid in a straight line on a smooth table. If the end Z be
free and the end A moved with velocity F in a direction perpendicular to the line of the
rods, then the initial velocities of the joints
(AB\
(BC),
...
and the end
Z are
a, b, ...,
(Coll.
Exam.)
z,
where
0=A
(F + 2a) +
(2a + &),
= B (a
y + 22 = 0.
and
42.
points
Six equal uniform rods form a regular hexagon loosely jointed at the
angular
a blow is given at right angles to one of them at its middle point, shew that the
opposite rod begins to
move with T]o
of the velocity of the rod struck.
(Camb. Math. Tripos, 1882.)
176
[OH. vi
shew that the initial axis of

fixed, is struck
body at rest, with one point
diametral
with
to
the
momental ellipsoid at 0, of
is
the
line,
respect
body
the plane of the impulsive couple acting on the body.
43.
rotation of the
The positive octant of the ellipsoid ^72 /a 2 +3/ 2 /62 + 22 /6 2 = l has the
44.
Shew that if an impulsive couple in the plane
-
act
upon the
2 \b
origin fixed.
a/
octant, it will begin to revolve about the axis of
z.
(Coll.
Exam.)
An ellipsoid is rotating about its centre with angular velocity

45.
2 , w s) referred
the centre is free and a point (#, y, z) on the surface is suddenly
to its principal axes
(Coll. Exam.)
brought to rest. Find the impulsive reaction at that point.
&>
(<BI,
A is
equal rods AB, BC inclined at an angle a are smoothly jointed at B
prove that the initial
parallel to the external bisector of the angle A BC
angular velocities of AB, BC are in the ratio
46.
made
Two
move
to
2 + 3 sin 2 5
2t
47.
15 sin |
2*
A uniform cone is rotating with angular velocity
this generator
is
(Coll.
o>
about a generator when suddenly
the altitude, a the semi-vertical angle, and k the radius of gyration about a
is
is fixed.
is
diameter of the base.

48.
Exam.)
loosed and the diameter of the base which intersects the generator
Prove that the new angular velocity
where h
(Coll.
rough disc can turn about an axis perpendicular to
its
circular cone rests on the disc with its vertex just at the axis.
to turn with angular velocity Q,
Exam.)
plane, and a 1 ough

If the disc be made
shew that the cone takes an amount of kinetic energy
equal to
|Q
49.
One end
/{cos a/J
+sin 2
/{7}.
(Coll.
Exam.)
of an inelastic string is attached to a fixed point and the other to a point

body of mass M. The body is allowed to fall freely under gravity
in the surface of a
without rotation. Shew that just after the string becomes tight the loss of kinetic energy
dvie to the impact is
where V
is
the resolved velocity of the body in the direction of the string just before
(I, m, n, A, p, v) are
C are the principal
impact, the string only touching the body at the point of attachment,
the coordinates of the string at the instant it becomes tight, and A, B,
moments
of inertia of the
body with respect
to its principal axes at its centre of inertia.

(Coll.
Exam.)
CHAPTER
VII
THEORY OF VIBRATIONS
Vibrations about equilibrium.
76.
In Dynamics we frequently have to deal with systems for which there

an equilibrium-configuration, i.e. a configuration in which the system
exists
can remain permanently at rest thus in the case of the spherical pendulum,
the configurations in which the bob is vertically over or vertically under the
:
point of support are of this character. If (q 1} q2

of a system and L its kinetic potential, and if ( 1?
q n ) are the coordinates

...,
n ) are the values of
...,
2
the coordinates in an equilibrium-configuration, the equations of motion
d fdL\
"7-,Upat
\oqrj
must be
ji
by the
satisfied
= 0, #2=0,
...,
(r=l,
2,
...,w)
set of values
= 0, ^ = 0,
#n
dL
=
-3oq r
#2
= 0,
...,
gn
= 0,
ql
=a
,
</
or 2
...,
qn
= an
The
values of the coordinates in the various possible equilibrium-con

figurations of a system are therefore obtained by solving for q lt qz ..., qn the
,
equations
|=
in
(,.
1,2,. ..,),
which q l} q 2 ..., q n are to be replaced by zero.

In many cases, if the system is initially placed near an equilibrium-con
,
having very small initial velocities, the divergence

from the equilibrium-configuration will never become very marked, the
figuration, its particles
particles always remaining in the vicinity of their original positions and

never acquiring large velocities.
shall now study motions of this
type*;
We
they are called vibrations about an equilibrium-configuration^.

*
More
we study
in this chapter the limiting form to which this type of

divergence from a state of rest in the equilibrium-configu
ration tends to zero the study of the motions which differ by a finite, though not large, amount
from a state of rest in the equilibrium-configuration is given later in Chapter XVI the discussion
strictly speaking,
motion approximates when the
initial
of the present chapter may be regarded as a first approximation to that of Chapter XVI.
t The theory of vibrations has developed from Galileo s study of the small oscillations of a
pendulum.
In the
first
half of the eighteenth century the vibrations of a stretched cord were
investigated by Brook Taylor, D Alembert, Euler, and Daniel Bernoulli, the last-named of whom
in 1753 enunciated the principle of the resolution of all compound types of vibration into inde
pendent simple modes. The general theory of the vibrations of a dynamical system with a
number of degrees of freedom was given by Lagrange in 1762-5 (Oenvres, i. p. 520).
w. D.
finite
12
178
Theory of Vibrations
[CH.
vn
In the present work we are of course concerned only with the vibrations of systems
which have a finite number of degrees of freedom the study of the vibrations of systems
which have an infinite number of degrees of freedom, which is here excluded, will be found
in treatises on the Analytical Theory of Sound.
;
We
suppose that the system is defined by its kinetic energy T and

potential energy F, and that the position of the system is specified by the
coordinates (qlt q2 ,..,q n ) independently of the time, so that T does not
shall
its
we shall also suppose that no coordinates have been

explicitly
the
kinetic
energy T is therefore a homogeneous quadratic function
ignored
with
of
coefficients involving q lt q2 ..., q n in any way.
There
t^n,
^2.
is evidently no loss of generality in assuming that the equilibrium-con
involve
<?i,
figuration corresponds to zero values of the coordinates q lt q. ...,qn so that

are very small throughout the motion considered.
9
;
2>
<?i*
9i>
<?*>
The
and products of q 1} q 2 ..., q n in T are

all the coordinates and velocities are
the motion retain only the terms of lowest
coefficients of the squares
functions of qlt
small,
>9n
<?t>
we can
q.2
qn
...,
as
however
in approximating to
order in T, and so can replace
all these coefficients

by the constant values
which they assume when q q. ..., q n are replaced by zero. The kinetic
energy is therefore for our purposes a homogeneous quadratic function of
l
<jn
i2,
>
with constant
??i
Moreover,
2>
coefficients.
we expand the function
if
by Taylor s theorem in ascending
powers of q l q2 ..., qn the term independent of q ly q.2 ..., qn can be omitted,

since it exercises no influence on the equations of motion and there are no
,
terms linear in q 1} q z
...,
qn
dV
since if such terms existed the quantities dqr
would not be zero in the equilibrium
position, as they
must
be.
The terms
V are therefore the terms quadratic in q l} q2 ...,q n

terms of the expansion in comparison with these,
the
higher
Neglecting
we have therefore
expressed as a homogeneous quadratic form in the
variables q lt q.2 ..., q n with constant coefficients.
of
lowest
order in
Thus
problem of vibratory motions about a configuration of equilibrium

depends on the solution of Lagrangian equations of motion in which the kinetic
and potential energies are homogeneous quadratic forms in the velocities and
coordinates respectively, with constant coefficients.
77.
the
Normal
coordinates.
In order to solve the equations of motion of a vibrating system, we write

the expressions for the kinetic and potential energies in the form
T = | (a n qf + a
22
q./
+ a nn q n + 2a
2
. .
l2
q,
q2
2a ls q q 3 +...+ 2a n _
l
1>
q n -iq n ),
179
76, 77]
of these T is ( 26) a positive definite form and the determinant formed

of the quantities a rs is not zero (since if this condition is not satisfied,
The equations of
on less than n
T will
velocities).
;
independent
depend
motion are
8F
d /dT\
if
a change of variables
-o-i,
-i
2,
....),
q 2 ...,q n
made, such that the new variables
new equations of motion will be
is
(<?/,
are linear functions of (q lt q a ...,q n ), the

,
(]b
U J- \
brri
-J-.
"
\oqr j
Ck
~t
(r=l,
-;,
oq r
...,),
2,
and these equations are clearly linear combinations of the original equations.
Suppose then* that the original equations of motion are multiplied
m.2 ..., m n and added together.
respectively by undetermined constants m,,
,
The
resulting equation will be of the form
where
Q=h
provided the constants
ra 2
lt
+ h^q +
q1
,
~i~
tinn^îi
^~
These equations can coexist only

ftji A/
a.21
62i
a 22 X
b nl
n h l} h a
,
^ /\^m ^î
Gfi 2
+ hn q n
...,
h n X satisfy the equations

,
"\
if
^ji>
...,
...
^/i2^^2
II
*~
^nH***
\
n,)
^=
"\
A/t7j,.
a root of the determinantal equation
is
Oj2
~i
6 22)
>
flinA/
2ri
b.2r
Om
this equation, we can determine from the

...,
n h l h 2 ...,h n ;
2
preceding equations a possible set of ratios for
these ratios may, in certain cases, be partly indeterminate, but in all cases at
Moreover,
least
if
one function
is
any root of
m^ m
Now
let
can be obtained in this way, satisfying the equation
so
U>
= o.
a linear change of variables be effected so that the quantity Q

there will be no ambiguity in
is one of the new variables
determined
This method of proof
is
due to Jordan, Comptes Rendus, LXXIV. (1872)
p. 1395.
122
180
[OH.
vn
denoting the new variables by q 1} q 2 ..., q n we shall take q to be identical

with Q, so that the above equations are satisfied by the values h-L = \, A2 = 0,
h n = 0. Since the form T is a positive definite form, the coefficients
\
. . .
"22,
33>
of q 2 q3
,
a nn of the squares of q z q 3) ..., q n will not be zero: so instead

q n we can again take new variables respectively equal to
,
îs
+ ^12 ft. & +

.
q*
By
this
so
"
change of variables the terms in q^, q^q z

that a 21 a 31 ..., a m are zero.
q^q n are
removed from
= 0,
h n =0, a 21
Now introducing the conditions h = \, h = 0,

= in the equations which determine m m
.., a nl
/i 3
lt
we obtain the
?i-
Tin
...,
we can assume
+ **in
..,2
?i,
1*33
1*22
...,
...,
mn
h lt h 2
...
= 0,
hn
X,
values
m = l/a n
1
611
ra.2
= Xiu,
fea
=
=
0,
m = 0,
0,
= 0,
631
...,
.
m n = 0,
= 0.
6 M1
It follows that the equation

.
dt
cfia
+ \qi =
has the form
\dqj
0,
d /dT\
while the equations
ji ( o^~ )
dt \dq r J
d /dT
have the form
so that
and
dV
5~~
do not involve
q^
(**"*,
>
oq r
dV
V9^r/
T = T-a n q*,
where
=T-
3-
~*
"
(r
2, 3,
),
9^r
V = V- â^q*,
and qlf
This last system of equations may be regarded as the system of equations

1) degrees of freedom.
corresponding to a vibrational problem with (n
Treating them in the same manner, we can isolate another coordinate q2
such that
if
T"
=T - lavtf,
V"
- F - faoKqf
do not involve q 2 or
and
a%, are certain constants), then
are
determined
coordinates
the
and
...,
n
4
by the equations of
q
q3 q
q2
of
a vibrational problem with (n
freedom, in which the kinetic
2) degrees
and
and potential energies are respectively
(where X2 and
T"
V"
T"
V".
Proceeding in this way, we shall finally have the variables chosen so that
the kinetic and potential energies of the original system can be written in
terms of the new variables in the form
T = % (a n gi + 229 + + a nn q n
V = i (&i 2i + /3*q + ...+ &m? ),
2
where a n a^,
,
...,,
Ai.
^22,
*
z
Pnn are constants.
*),
181
77]
If finally we take as variables the quantities â n qi, Va^g,, .... â

instead of q l} q 2 ..., qn the kinetic and potential energies take the form
,
V = | (/ij^ +
2
where pk stands
fJ, 2
q<?
for fikkl akk-
In this reduction
immaterial whether the determinantal equation has

The final result can be
it is
or has groups of repeated roots.
its roots all distinct
expressed by the statement that if the kinetic and potential energies of a

vibrating system are given in the form
+ a nn q n + 2a ^ q +
+ b nn qn + 26 q +
2
...
V=
it is
always possible
and
12
<7,
. . .
2a,l _ 1) n q n -^ q n \
4-
2b n -i, n qn-iqn),
when expressed
in
terms of the new
form
= \ /Mi + ^2 9V +
+ Mn
s
,
^, /i.2) ..., n are constants. These new coordinates are

normal coordinates or principal coordinates of the vibrating system.
the quantities
Now
find a linear transformation of the coordinates such
to
potential energies,
coordinates, have the
called the
that the kinetic
where
12
it is
fjt
a well-known algebraical theorem that the roots of the determi
nantal equation
b ln
=0
2 A,
an i A,
are the values of
for
o n i .................. a nn A,
which the expression
- 6 n ) q* + (a, X - 6
(a n \
2
o nn
22 )
q<?
can be made to depend on
+ (a nn \ - b nn ) q n + 2 (a u \ 2
than n independent variables (which
less
will
be
linear functions of q 1} q.2 ..., q n ).

Since this is a property which persists
linear
of
variables, we see that the determinantal equation
through any
change
,
is
<?i,
invariantive,
92,
qn
i.e.
and
if g/,
if
q2
q n are any n independent linear functions of

expressed in terms of q^, q2 ..., q n take
T and V when
the form
T=\
qi
+ a* q^ +
+ 2a
12
?/ q,
...),
182
[OH. vii
=0
then the roots of the new determinantal equation a,./X &,./

same as the roots of the original determinantal equation a rg \
b r^
are the
= 0.
But when the kinetic and potential energies have been brought by the
introduction of normal coordinates to the form
the determinantal equation
is
= 0,
so its roots are
//, 2
It follows that the constants
//,.
/JL Z
p n which
,
occur as the coefficients of the squares of the normal coordinates in the potential
energy, are the n roots (distinct or repeated) of the determinantal equation
|
a rs \
b rs
= 0,
where a n a 12
,
expressions for the kinetic

Lt will
and
bn
6 12
...
are the coefficients in the original
potential energies.
be seen that the problem of reducing the kinetic and potential energies to their
expressions in terms of normal coordinates
is essentially the problem of

simultaneously
reducing each of two given homogeneous quadratic expressions in n variables to a sum of
squares of n new variables the fact that T is a function of the velocities while V is a
function of the coordinates does not affect the question, since the formulae of transforma
:
tion for the velocities j l5 j 2
coordinates qi, q 2
...,
qn
>
^n are the
same as the formulae
of transformation for the
It might be supposed from the foregoing that it is always possible to transform

simultaneously each of two given homogeneous quadratic expressions in n variables to a
sum of squares of n new variables; but this is not the case for example, it is not possible
;
to transform the
two quadratic expressions
ax2 + bxy + az 2 and

to the forms
2
where
The
may
77,
f are linear functions of x, y,
conditions which
must be
and o
z.
satisfied in order that
be simultaneously reducible to the form
two given quadratic expressions
183
77, 78]
are, in fact, that
shall be linear*.
the elementary divisors (Elementartheiler) of the determinant a rs\ brs

If however one of the two given forms is a definite form (as we saw was
\\
j|
the case with the kinetic energy in the dynamical problem), the elementary divisors are
always linear, and the simultaneous reduction to sums of squares is therefore possible
this explains the circumstance that the reduction
can always be effected in the dynamical
problem of vibrations.
The universal possibility of the reduction to normal coordinates for dynamical systems
was established by Weierstrass in 1858 1
previous writers (following Lagrange) had
supposed that in cases where the determinantal equation had repeated roots a set of
normal coordinates would not exist, and that terms involving the time otherwise than in
trigonometric and exponential functions would occur in the final solution of the equations
;
of motion.
Sylvesters theorem on the reality of the roots of the determinantal
78.
equation.
We have seen in the preceding article that by introducing new variables

which are linear functions of the original variables, it is always possible to
reduce the kinetic and potential energies of a vibrating system to the form
q
+ Ma +
The question arises as to whether this transformation is real,

the coefficients
h ly h 2 ..., h n which occur in
1} m 2
...,
n
i.e.
whether
the trans
formation are real or complex.

Sijice these coefficients are given by linear
whose
with
the
\n
coefficients,
equations
possible exception of the roots Xj A,2
,
of the determinantal equation, are certainly real, the question reduces to

investigation of the reality or otherwise of the roots of the equation
au
\b
ttojA,
a m^
it
~ b nl
a 12 X-&i2
......
amX
b ln
......
a.2n
b2n
u nn \
b nn
ooX
a n2 \
b n2
.
being known that the quantities a rs and

2
Oiiî
is
b.2l
+ a^q +
2
...+ a nn q n 2
+ 2a w g
b rs are all real,

1
ga
+ ... + 2a, _
l
an
=0;
\
and that
]>n
gn_1 g n
a positive definite form.
Let A denote| the determinant \\a rg \ b rs

and let Aj. denote the
determinant obtained from it by striking out the first row and first column
let A., denote the determinant obtained from A
by striking out the first two
,
Cf.
Muth
s treatise
on Elementartheiler (Leipzig, 1899); or Bocher
Algebra (New York, 1907).

t Cf. Weierstrass Collected Works, Vol. i. p. 233.
The following proof is due to Nanson, Mess, of Math. xxvi. (1896),
Introduction
p. 59.
to
Higher
184
rows and
first
two columns, and so
Then
on.
in
[CH. vii
any symmetrical deter
minant, say
where arg = agr
it is
known
that
9a 12 /
and hence
if
0*u
signs
thus
vanishes the quantities
we have
and 5
OOufa*
must have opposite
the result that in the series of quantities

A,
A lf A
An
...,
A n = 1),
(where
any one member of the series vanishes for a given value of X, the two
adjacent members must have opposite signs for that value of X.
if
Let A r denote the determinant formed from A r by

replacing X by unity
and each of the quantities b rs by zero, so that A r is the coefficient of the
highest power of
is
in
Ar
Since
a positive definite form, A r is positive for all values of r from

coefficients of the highest powers of X in the functions A, A 1;
Thus the
are all of the
same sign
and therefore as X increases from
oo
to
to n.
...
An
oc
these functions lose n changes of sign.
Now
since
vanishes,
it
An
not zero and
is
A r_
follows that the functions A,
A r+1
A A
1;
have opposite signs when A,.

2
..., A n cannot lose or gain a
,
change of sign except when X passes through a root of A. But as X passes

oo
from
to -f oo
the functions lose n changes of sign
and hence the
n roots of the determinant A are all real.
The transformation to normal
coordinates is therefore always a real transformation*.
,
Moreover, since a change of sign is lost in the pair A, Aj every time that X
passes through a root of A, it is evident that Aj must change sign when X
,
increases from one root of
n roots of
to the consecutive root,
(nl)
are separated by the

each of the functions A,, are separated
roots of
by
and hence that the
A^
similarly the roots of

the roots of the function A,. +1
.
has the same sign at X =

as at X =
oo
the root of the function A n _! will not be negative. If moreover A n _ 2 has
as at X =
the same sign at X =
oo
neither of the roots of A n _ 2 will be
Now A n
has no roots
and
if
A n _j
negative:
for
if
this
condition
is
satisfied,
A n_
must have either two

two negative
negative roots or no negative roots, and there cannot be

roots since there is no negative root of n _j to separate them.
Sylvester, Phil.
Mag.
(4) iv. (1852), p.
138: Coll. Papers,
i.
Similarly in
p. 378.
78, 79]
none of the functions in the series A, A ls A 2 ... A n

general the condition that
the
shall have a negative root is that each of the functions must have
,
same sign
X=
at
as at
A,
=-
in order that all the roots of
have
shall
X=
at
oo
be positive
same sign
the
to be satisfied
Hence the condition
A may
as
(- l)
is
n~r
,
that each quantity A,,

that each of the
i.e.
determinants
But these
be positive.
shall
are the well-known conditions that the quadratic
form
Ai
2
<?i
+ b nn q n + Zb^q* +
2
+ b^q? +
Hence
be a positive definite form.

a rs \ b rs =
shall
minantal equation
the quadratic
gn
roots positive is that
its
\\
form
n _ n q n _l q
b nn q
a positive
shall be
finally the condition that the deter-
shall have all
+ 2& n _ lin g n _
definite
form,
1>
i.e.
that the potential energy in the vibratory
motion shall be essentially positive.

Solution of the differential equations ; the periods; stability.
79.
In order to express the configuration of any vibrating system in terms of

we first determine the normal coordinates of the system, and
the time,
express the kinetic and potential energies in terms of them, so that these
take the form
where
(q l} q 2 ...,q,i) are the
roots
of
(A,!,
normal coordinates, and (Xj,A, 2 ...,\ n ) are the

a rs \
b rs =
these quantities
equation
,
X>,
the
...,
determinantal
\\
||
\ n ) have been shewn
in the last article to be all real.
of motion for
any coordinate q r
= __
dt \dq
is
dqr
dqr
therefore
r
The
solution of this equation

.
rt
Ar
and
0.
is
+B
r /)
*
+ B r e~
where in each case
if X,. is
if X,. is
zero,
if X,.
negative,
rt
is
positive,
j.
denote constants of integration.
namely
186
vu
[CH.
appears from these equations that if all the normal coordinates except
one, say qr are initially zero, and if the constant \. corresponding to the
non-zero coordinate is positive, then the coordinates (qi,q2
q r -i,q r +i,
It
will
the coordinate
>
<Jn)
zero, and the system

perform vibrations in which
Moreover the configuration of the
is alone affected.
be permanently
qr
will
an interval of time 2?r/Vx,.. This is usually

that
each
expressed by saying
of the normal coordinates corresponds to an
system
will repeat itself after
independent mode of vibration of the system, provided the corresponding

constant
\r
and
is positive;
the period of this vibration is 2?r/Vx,..
Moreover, if the system be referred to any other set of coordinates which

are not normal coordinates, these coordinates are linear functions of the
normal coordinates
anji the
normal coordinates perform their vibrations
quite independently of each other thus every conceivable vibration of the

system may be regarded as the superposition of n independent normal
vibrations.
This is generally known as Daniel Bernoulli s principle of the
.
superposition of vibrations*.
If the quantities (Xj, \ 2 ...,X n ) are not all positive, it appears from the
above solution that those normal coordinates q r which correspond to the
non-positive roots X,. will not oscillate about a zero value when the system is
in its equilibrium position, but will
slightly disturbed from a state of rest
,
increase so as to invalidate the assumption made at the outset of the work,

namely that the higher powers of the coordinates can be neglected. In
this case therefore, there will not be a vibration at all, and the equilibrium
If however the initial disturbance is
configuration .is said to be unstable.
such that these normal coordinates which correspond to non-positive roots
X r are not affected, the system will perform vibrations in which the rest
of the normal coordinates oscillate about zero values.
The normal modes
of vibration,
coordinates for which the
be
stable.
as a whole
corresponding root
X,.
is
positive,
are said to
Xr
are all positive, the equilibrium-configuration

The condition for stability of the equi
said to be stable.
If the constants
is
which correspond to those normal
librium-configuration
is
therefore,
by the theorem
of the last article, that
of the vibrating system shall be a positive
definite
form.
This result might have been expected from a consideration of the integral of energy
for this integral is
T+
V=h,
where T and F are the quadratic forms which represent the kinetic and potential energies,
and where h is a constant. This constant h will be small if the initial divergence from the
7
if V is also a positive
equilibrium state is small. But T is a positive definite form and
definite form, we must have T and V each less than h, so T and V will remain small
throughout the motion the motion will therefore never differ greatly from the equilibrium;
configuration,
i.e. it
will
*
be
stable.
Histoire de V Academic de Berlin, annee 1753,
p. 147.
187
79, 80]
Examples of vibrations about equilibrium.
80.
We
shall
now
number
discuss a
of illustrative cases of vibration about
equilibrium.
To find the vibration-period of a cylinder of any
a perfectly rough fixed cylinder.
(i)
cross-section which
can
roll
on the
outside of
Let s be the arc described on the fixed cylinder by the point of contact, s being
measured from the equilibrium position let p and p be the radii of curvature of the
cross-sections of the fixed and moving cylinders respectively at the points which are in
contact in the equilibrium position p and p being supposed
positive when the cylinders
are convex to each other let
be the mass of the moving cylinder, Mk* its moment of
inertia about its centre of
gravity, and c the distance of the centre of gravity from the
;
initial position of
the point of contact in the moving cylinder.
between the common normal to the cylinders and the

the angle between the common normal at time t arid the vertical,
a + s/p + s/p is the angle made with the vertical by the line
joining the centre of curvature
of the moving cylinder with the original
point of contact in the moving cylinder, and
s/p + s/p is the angle made with the vertical by the line joining the last-named point to
If a denotes the initial angle
vertical,
then a + s/p
is
the centre of gravity of the moving cylinder.

is
The angular
velocity of the
moving cylinder
therefore
..,.,.
so its
kinetic energy is
.
The
V
\P
potential energy
Mg x height
is
of the centre of gravity of the
= Mg \(p
(p + p )cos (o+-] -p
moving cylinder above some
cos(a + - +
P
Neglecting
s3
P/
+ ccos
[\P
fixed position
+
p
this gives
PP
\ PP
of motion,
dT
fdT\
dt \di
Jf (# + c 2) (1
gives
+P/
PP
\ PP
and the vibrations are therefore given by the equation
where
X
is
A and
are constants of integration to be determined
by the
initial conditions,
and
XV9
The vibration-period
(ii)
tion of
ff
PP
-r~ 5 {(
COSa
K + C 2 (p + p
,
C[
is 27T/X.
To find the periods of the normal modes of vibration about an

equilibrium-configura
a particle moving on a fixed smooth surface under gravity.
The tangent-plane
to
equilibrium-configuration
the surface at the point occupied by the particle in the

evidently horizontal take as axes of x and y the tangents to
is
188
vn
[OH.
the lines of curvature of the surface at this point, and as axis of

upwards so that the equation to the surface is approximately
a line drawn vertically
where p
The
and p 2 denote the principal radii of curvature, measured positively upwards.

l
kinetic energy and potential energy are approximately
T=\m(xi -\-i/2
V= mgz
and
&
It is evident
(where
in is
the mass),
y*
from these expressions that x and y are the normal coordinates
the
equations of motion are

P2
Pi
and the periods
of the
normal modes of vibration are therefore

9^ZiTT
fl
\
1
I
anH
clIlLl
9
-.
zAe normal modes of vibration of a rigid body, one of whose points is fixed,
vibrating about a position of stable equilibrium under the action of any system
of conservative forces.
ô ymc?
(iii)
and which
Take
is
OXYZ the equilibrium
positions of the principal axes of

will
axes
be taken as usual to be these
the
moving
body
We shall suppose the position of the body at any instant
principal axes of inertia.
as the
9 ; we shall regard
defined by the symmetrical parameters (, ij,
77,
x) of
as fixed axes of reference
inertia of the
at the fixed point
independent coordinates of the system, x being defined in terms of them by the equation
The components
so
first
we
body about the moving axes are (16)
as small quantities of
account of the smallness of the vibration, we regard
?;, f
order x therefore differs from unity by a small quantity of the second order, and
On
the
of angular velocity of the
have, correctly to the
and the
order of small quantities,
kinetic energy of the body,
where A, B,
written
The
first
are the principal
potential energy
parameters (,
r;,
is
let it
which
is
moments
some function
of inertia at the point of suspension, can be
of the position of the body,
be denoted by V (,
77,
f ).
and therefore of the
189
80]
Since zero values of
(|,
correspond to the equilibrium position, there will be
77,
no terms linear in (, 77, f) when F is expanded in ascending powers of (, 77, f) the lowest
terms are therefore of the second order neglecting terms of higher order, we can therefore
:
write
V = a? + b^ + o? + 2fyC+ 2g& + Zhfa

where
a, b,
f, g,
c,
of determining the normal coordinates

two quadratic expressions
The problem
of reducing the
to the
form
where
(#, y, z)
Now
h are constants.
are linear functions of (,
77,
same as that
therefore the
is
).
the equation, referred to the fixed axes, of the momental ellipsoid in
librium position
consider in connexion with this the quadric whose equation
which we shall
its
equi
is
call
the
"
ellipsoid of equal potential energy
is
and determine the common
"
Y Z
be the coordinates, referred

Let (X\
set of conjugate diameters of these quadrics.
,
)
to these conjugate diameters, of a point whose coordinates referred to the fixed axes
are (X, Y,
By
Z\ and
let
the equations connecting
(X
Y Z
and (X,
Z) be
this transformation the equations of the quadrics are reduced to the
X ^+b^ + CjZ
=1,
and therefore the transformation which gives the normal coordinates

problem is
It follows that in
quantities (,
77,
a normal
f) will
But from the
mode
form
of vibration, say that in
in the
dynamical
which x alone
varies, the
be permanently in the ratio
definitions of
it is
evident that
r;,
are, to
the
first
order of small
quantities, proportional to the direction-cosines of the line about which the rotation
of the rigid body takes place, and consequently the normal mode of vibration of the
rigid
body consists of a small
oscillation
X
i.e.
about the
is
Z=li
line
2
whose equation
/,,
line
r
which
about a
one of the
common
=o,
=o,
conjugate diameters of the two quadrics.
is
190
Hence
we have the result

common conjugate
finally
oscillations about the
[CH.
vn
that the normal vibrations of the body are small
diameters of the momental ellipsoid
and the
ellipsoid
of equal potential energy.
To find the normal coordinates and the periods of normal vibration in the
(iv)
system
of three degrees of freedom for which
F= i
small in comparison with p and q ; and to shew that if such a system be let
go from rest with y and z initially zero, the vibration in x will have temporarily ceased
z
2
2
after a time irp (q -jo )/a and that there will then be a vibration of the same amplitude in
y as the original one was in x.
(Coll. Exam.)
where a
is
The form
of the kinetic
and potential energies suggests the transformation
which gives
(f7=
The
kinetic
variable
is
77
therefore a normal coordinate
to reduce the remaining terms in the
and potential energies to sums of squares, we write
and then we have
The
variables
?/,
<^>,
^ are therefore the normal coordinates.
Suppose that initially
and suppose that k

neglected.
Then
we have
z = 0,
^ = 0,
= U,
<f>
= U,
vibrations of the normal coordinates
last
= 0,
2 = 0,
z
so small that its product with other small quantities can be

to this degree of approximation we have
initially
rj
The
y = 0,
is
r,
The
x = lc,
TJ
and
f=0.
<
are therefore given by the equations
k cos pt,
=1
COS
equation can be written

4.
U cos [pt
L (l ---jL-Jl
I
or
jrW-tfU
= M cos pt cos r a7i

-
(f>
p(q -p-)
+ ~ k sin pt sin
a -t
191
80, 81]
The motion can
therefore be approximately represented initially by

T)
k cos
pt,
4>
= ^k cos pt,
or
y = 0.
x = k cos pt,
After an interval of time Trp(q
?7
)/a?,
the motion
= \k cos pt,
l;k
<f)
or
is
approximately represented by
cos pt,
y~ kcospt;
0,
which establishes the result stated.
81.
a new constraint on
Effect of
a vibrating system.
the periods of
We
shall now consider the effect produced on the periods of normal

vibration of a dynamical system about a configuration of stable equilibrium
when the number of degrees of freedom of the system is diminished by the
introduction of an additional constraint
is specified in terms of its normal
so that the kinetic and potential energies have
Suppose that the original system

coordinates (q 1 q,
the form
and
let
q n ),
...,
the additional constraint be expressed by the equation
we can expand the function / in ascending

and
retain
..., q n
q.2
only the first terms of the expansion: we
the
constraint
can thus express
by the equation
Since q
<?
>
powers of q l}
where A,,
<?n
An
...,
are small,
>
are constants.
As the
equilibrium-configuration
is
supposed
to be compatible with the constraint, there will be no constant term.

means of this equation we can eliminate q n we thus have
By
V = \ X.V +
+ ^n-i

fore the (n
/n-!
(A
9l
+A
_,
qn^y
of motion of the constrained

system are there
1) equations
Ar
qr
+ \ r *q r + f*A r =
(r
(r
l, 2, ...,
n-1),
I, 2, ...,
n-
1)
192
[CH.
vn
where
1
P = -T--.
(A & +
^l
n
+ -4
_!</_,)
+ -j-iCAtfi + ... + ^-i^n-i)

-d.
so the equations of motion of the constrained

form of the n equations
+ Vgv + pA r =
qr
where
is
//,
Now
system can be written in the

(r
1, 2,
n),
.,
undetermined.
consider a normal
mode
of vibration of the modified system, defined
by equations
(/!
cos \t,
qo
cos \t,
..,
qn
ar
(X,.
Substituting the values of a 1
equation
A^ + A
- X ) + vA r =
2
a.
2 a.2
2>
//,
v cos \t.
we have
Substituting in the equations of motion,

2
cos \t,
a.n
(r=l,
...,
a n given
...
+ A n *n =
2, ..., n).
by these equations in the

0,
we have
A*
-V
\i>
_AJ_
This equation in \- has (w

tion are evidently
interspaced
AJ_
-
-V
1) roots,
^n"
which from the form of the equa
between the quantities Xf,
A 22
...,
XH
quantities 2?r/X corresponding to these roots are the periods of the
2
:
the
normal
modes of vibration of the constrained system, and it therefore follows that

the (n
l) periods of normal vibration of the constrained system are spaced
between the n periods of the original system.
The stationary character of normal
82.
vibrations.
We shall next consider the effect of adding constraints to a dynamical

system to such an extent that only one degree of freedom is left to the
Let (q 1} q 2 .... q n } be the normal coordinates of the
system.
original
system the constraints may, as in the last article, be represented by linear

equations between these coordinates, and can therefore be expressed in the
;
form
gi
= /*i?i
& = Wl,
q,i
= p n q,
where /^ 1 /i2 ..., n are constants and q is a new variable which

may be
taken as defining the configuration of the constrained
system at time t.
Let the kinetic and potential energies of the original system be
,
fj.
V=
193
81-83]
so 27T/XJ, 27T/X 2 ..., 2TT/\ n are its periods of normal vibration: the kinetic
and potential energies of the constrained system are then
,
V = i (XV + V^ +
2
The period
where X
+ xn
>
f.
of a vibration of the constrained system
is
therefore
2-Tr/X,
is
_
If the constraints are varied, this expression has a stationary value when
n are zero this stationary value is one
1) of the quantities /i 1} /u 2
(n
of the quantities X a 2 X 22 ..., X n 2 and thus we have the theorem that when
constraints are put on the system so as to reduce its number of degrees
of
freedom to unity, the period of the constrained system has a stationary value
,
f*>
for those constraints which make the vibration
to be
a normal vibration of the
unconstrained system.
Vibrations about steady motion.
83.
type of motion which presents many analogies with the equilibriumconfiguration is that known as the steady motion of systems which possess
ignorable coordinates: this is defined to be a motion in which the nonignorable coordinates of the system have constant values, while the velocities
corresponding to the ignorable coordinates have also constant values.
One example of a steady motion is that of the top, discussed in

72 as another
example we may take the case of a particle which is free to move in a plane and is
attracted by a fixed centre of force, the potential energy
depending only on the distance
from the centre of force; for such a particle, a circular orbit described with constant
velocity is always a possible orbit, and this is a form of steady motion, since the radius
vector is constant and the angular velocity
corresponding to the ignorable coordinate 6
;
is also
In
constant.
cases, if a system is initially in a state of motion

differing only
from
a given form of steady motion, the
slightly
divergence from this form of
motion will never subsequently become very marked we shall now consider
many
motions of this kind, which are called vibrations about

steady motion.
The steady motion is said to be stable* if the
vibratory motion tends to
a certain limiting form, namely the
steady motion,
when
the initial disturb
ance from steady motion tends to zero.
Let
(>!,
j>
2,
p k ) be the ignorable and
coordinates of the system.

will
(q lt &,..., q n ) be the non-ignorable

Corresponding to the ignorable coordinates, there
be k integrals
dL
*
W. D.
This definition
is
due
to Klein
and Sommerfeld.
194
[OH.
vn
We shall suppose that these constants

where &, /3 2
ftk are constants.
in
the
value
have the same
vibratory motion as in the undisturbed steady
motion of which it is regarded as the disturbed form; this of course only
amounts to coordinating each vibratory motion to some particular steady
,
>
motion.
We
time
suppose the system conservative, with constraints independent of the

let its kinetic energy be
T=\
where the
The
2 ay fay + 2 2
Cy,
byqipj
+b 2 2
coefficients a#, by, cv are functions of q lt qz

-
...,
qn
coordinates are
integrals corresponding to the ignorable
pi
Let
n
2,
Cij
be the minor of
c y in
-
divided by this determinant;
quantities p r
+ 2b
{j
= fy
qi
= 1,2,
.,
k).
the determinant formed of the coefficients

then solving the last equations for the
we have
]j r
= 2 Cm (Ps
2 big qi).
pt in the above expression for T,
Substituting for pi, pz ...,

the properties of minors of determinants,
,
T=
i,
Now
(j
(a*,
- 2C
ls
I,
and
utilising
we have
b u bjs ) q iqj
2 Cufrfr.
I,
Let
perform the process of ignoration of coordinates.
be the
modified kinetic potential, so
R=
=
T-V-
2 (ay - 2 Cubabjg) qiqj + 2 Crs @ r b
i,
We
2 pr /3 r
I,
can without
in the steady
loss of generality
motion are
variables
qi,
R
qz
all zero.
...,
qn
q-i,
q.2 ,
- $ 2 C&&& -
and quadratic
suppose that the values of q lt q 2
If then the coefficients in
qi,
qz
...,
qn
expanded
q n are neglected in comparison with the

an expression consisting of terms
obtain for
in q l} qz
...,
qn
q l} q 2
...
qn
d fdR\
dR
(^
)-a r
dq
\dqj
-r.
and these terms can therefore be omitted.

by permanent
zero values of q^ q2
Now
...,
the terms which
qn disappear auto
(r-1,
dt
satisfied
R are
...,
and independent of q 1; q 2
q^, q z ..., q n
from
the
equations of motion
matically
are linear in
V.
l,s
...,
terms of the second degree, we

linear
qi
qn by Taylor s theorem, and all terms in the

thus obtained which are above the second degree in the
in ascending powers of
expression of
ls
l,f,s
2, ..., n),
Moreover, since the equations are

that no terms
..., q n it is evident
,
195
83, 84]
linear in q 1} q2 ,
...,
It follows that the
solution of
q n and independent of q l q 2 ..., q n can be present in R.

problem of vibrations about steady motion depends on the
Lagrangian equations of motion
homogeneous
in
which the kinetic potential
quadratic function of the velocities
and
is
coordinates, with constant
coefficients.
difference between vibrations about equilibrium and vibrations about

motion
consists in the possible presence in the latter case of terms of
steady
the type q r q s (i.e. products of a coordinate and a velocity) in the kinetic
The
These are called gyroscopic terms. The vibrations about steady

motion of a system are in fact the same thing as the vibrations about
equilibrium of the reduced or non-natural ( 38) system to which the problem
potential.
is
brought by ignoration of coordinates.
The equations
of motion for the vibrating system are therefore
dR
-5 =
d fiR\
-nlo-rdt \dq r j
where
rf!
q r qs
2 &., gr
r,s
a,,.,
but where y rs is not in general equal

expanded form are
2i</i
+ 2 jrsqr qs
</
(r,
$n =
to
jsr
/3 sr
= 1,
2,
. .
ri),
The equations
- &1 qi + 12# + (72! - Tia) - & 2 +

J2
9
13& + (731
+ (712 ~ 721) qi ~ 0*qi + && - j3vq + a, q + (y - y
a
r, s
r,s
and where
ql
2, ...,w),
R=i2a
(11
(r=l,
dq r
7l3>
3,
zi )
of motion in the
& ~ $^3
q,
-/3&q3
etc.
These are linear equations with constant
coefficients, which are of the same

character
as
the
general
corresponding equations in the case of vibrations about
equilibrium they differ only in the presence of the gyroscopic terms, which
involve the coefficients (7^.
The presence of these terms makes it
7^).
;
impossible to transform the system to normal coordinates*; but as we shall

next see, the main characteristic of vibrations about
equilibrium is retained,
namely that any vibration can be regarded as a superposition of n purely
periodic vibrations, which
vibration of the system.
shall call (as before)
the normal modes of
The integration of the equations.
84.
We
we
shall
now shew how
the nature of the vibrations can be determined,
by integration of the equations of motion.

That
formation
is
to say, impossible to transform the

system to
it is
possible to effect the transformation to
formation, and this
is
normal coordinates by a point-trans

normal coordinates by a contact-trans
actually done in Chapter XVI.
132
196
vn
[OH.
them into a system of equations

denote the modified kinetic potential of the
It will be convenient first to transform
each of the
Let
first order.
R
system, so that in the vibratory problem
Write
function of q 1} q2 ..., q n q l q z ..., q n
,
R -==
<V-1
Qn+T
<5~T~
&n
a homogeneous quadratic
so that qn +i, q n +2,
is
are linear functions of q 1} q2
...,
qn
"
>
*/>
and
vice versa:
=l
the equations of motion can be written
^ n+r
(r
n)
dq r
Now
if
an increment of a function of the variables q lt q z

due to small changes in these variables, we have
B denote
q n +i, ---,qzn,
8R =
r =i
Bq r
qn
...,
-.
&<]>,
oq r
\oqr
n
^
2/\qii-\-roqr
~T q n .^-roqr)
*
"
S^
qn +rqr
\r=1
S (q n+r &q r
}+ r=l
q r $q n +r)-
Let the quantity
R,
q n+r qr
when expressed as a function of q l} q2 ..., qzn be denoted by H,

a known homogeneous quadratic function of the variables q 1} q
,
last
so that
..-,q2n
H
,
is
the
equation can be written
of motion, which consisted originally of n equations

can
be replaced by a system of 2n equations, each of
each of the second order,
and therefore*
the equations
the first order,
namely
9
= dH
0n+r
oq n+r
the independent variables being q l
We
shall
For
q2
...,
q 2n
dH
-goqr
/no
(r- 1,2,
...,),
as the
now shew that the function H, which has replaced
of
the
sum
the
kinetic
and
of
the
function
equations, represents
determining
considered.
of
the
dynamical system
potential energies
contains terms of degrees
This transformation
Chapter X.
is really
2, 1,
and
in the velocities,
and
a case of the Hamiltonian transformation given later in
197
equivalent to twice the terms of degree two together with the terms of
degree one, by Euler s theorem it follows that H, being defined as
is
dR
v
2
qr
r=\
T?
^-R,
oqr
will be equal to the terms of degree two in the velocities in R, together with
the terms of zero degree in R with their signs changed on comparing the
:
T and R
expressions for
given on page 194,
follows that
it
H=T+V,
so
is
the total energy
variables q lt qz
...
,q.
of the dynamical system, expressed in terms of the
In the case of vibrations about an equilibrium-configuration, we have

seen that the condition for stability is that the potential as well as the
kinetic energy shall be a positive definite form we shall now make a similar
assumption for the case of vibrations about steady motion, namely that the
;
total
energy
is
a positive
definite
form
in the variables q 1} q 2
assumption we shall shew that the steady motion

that the equations of motion
this
dqr
-rr
at
= 5dff
dqn+r
-^f^
dt
oqn+r
...,
and
is stable,
gwj
in fact
dH
^
dq r
can be integrated in the following way*.

Consider the set of linear equations in the variables q lt qz
S
qnj-r
~
+ dH( qi ,q
..., q^ n
,...,q2n )
~-yr
(r- 1,2, ...,);
if
we denote the determinant
element in the Xth row and
of the system by/(s),
/ith
f(s)
the expression of q lt q2
...,
q^
and the minor of the
column by
in terms of y lt y9
(X, /*
,
...,
ym
= !,
is
2,
...,2n),
given by the
equations
and the degree of f(s) in

than (2n-l).
q\,
s is 2n,
while the degree of f(s)^
is
not greater
In order to solve the equations of motion, consider

expressions
-, q.2n of the form
q?,
The method
of integration
which follows
is
for
due to Weierstrass, Berlin. Monatsberichte, 1879.
198
[CH.
vn
is taken round a large circle C which encloses all the

roots of the equation f(s)
0.
These values of q 1} q 2 ...,q2n will satisfy
the equations of motion, provided the equations
,
,
(r=l
...
n)
dp n+r
are satisfied.
If therefore p l9 p 2 ..., p 2n are polynomials in 5 so

the expressions in brackets under the integral sign vanish when
one of the roots of the equation f(s) = 0, these equations will
since the integrands will then have no singularities within the
,
It follows that Pi,j}.2
",pzn
must be a
be
satisfied,
contour C*.
set of solutions of the equations
lie ....gn
,
~r
chosen that
s is equal to
Pr
when
s is
a root of the equation f(s)
this condition is satisfied
by the
expressions
P* (s)
where a
a2
...,
a^
The equations
<?
+o
/(5) 2M
+ a 2n f(s}.
2n>>l
(^
1, 2,
.,
2n),
by the values
of 1/5 in the Laurent expansion -|- in positive

s of the expression
and nega
powers of
{Oi/()m + Ot/(*). +
Now
of motion are therefore satisfied
= coefficient
tive
I/(*)I F
e s(t-t
t>mf(*)m,t]
-77-75
7v
(p = 1,
2, ..., 2?i).
on inspection of the determinant /(s) we see that minors of the
types
(X<H*
(#*!,
2,
...,)
are of degree (2?i

1) in s, and the other minors are of degree (2n
2) in s;
so the coefficient of 1/5 in the Laurent expansion of f(s)\n/f(s) is zero unless
\ = n + p, or fj, = n + \; in the former case it is 1, and in the latter case it
is 1.
Hence on taking
we
see that the quantities
ttj
OF-2
. . .
dzn
are respectively the values of
at the time
Whittaker and Watson, A Course of Modern Analysis,
t 76id.
5-6.
5-2.
we
If therefore
(f)
and
to
199
84]
() A/x
if ft,
any
write
= coefficient
>,
of l/s in the Laurent expansion of
are the values of
...,ft
definite value
of
ft,
...,
e*
..
respectively corresponding
22
we have
t,
^=
{ft,+a<MO,M-
it is necessary to discuss the

In order to evaluate the quantities
let ki+l, where
nature of the roots of the determinantal equation f(s) =
of
this
k and I are real and i denotes \/-l, be any root
equation; then
<f>(t)^,
the 2n equations
n +
,
(ki
7\
9j
+-
?,
^(?i>
q n+a
I)
>
&n)
(- 1.2, ....)
can be satisfied by values of q 1}
system of such values be
1
where %
gz
%zn
...,
7i
T),
rji,
dH(q
we
l>q2
o7
q^, ...,
+ M?2
q^ which
,
>
Let a
are not all zero.
&n + ^2i
^ 2 i are real quantities.
Then
if
we
write
,...,q 2n )_
-"
Wi>
<12n)n>
&>
have, on separating the last equations into their real and imaginary parts,
- ^ n+ . =
- â + Atya =
+,
+a
a
But
since
is
homogeneous and of degree two
2tf(&, |a
)= 2=
-.,
and using the
first
&,
....
=1
(a
2,
in its arguments,
fxfl d!,
-..,
W).
we have
f)A,
two of the preceding equations

,
=
=
0|
this gives
A S
(A)
2H
Similarly
(77,
77.,
=
77^)
A;
2 (. ??+ =
i; a
n+a ).
Moreover on multiplying the first of the preceding equations by t) a and the

second by r) n+a adding, and summing for values of a from 1 to n, we have
,
X=l
a=l
and similarly
2/1
71
200
[CH.
we must have
Since the left-hand sides of these equations are equal,
2
But from equations (A) we
see that, as
vn
0.
a positive definite form, neither
is
k nor
yaZn+a) can be zero;
(fa^n+a
so the equation f(s)
quantity different
We
has each of
its
we must
therefore have
roots of the
form
ik,
zero;
where k
and
a real
is
zero.
from
which the equation f(s) =

has
a J-tuple root s each of the functions f(s)\n is divisible by (s
s )i~
For let Cj, c 2 ..., c2n be a set of definite real quantities; define quanti
shall next
shew that
in the case in
ties q 1} q 2
...,
q2n
by the equations
Sq n+a
...,
((fr, q.2 ,
= Ca
<?2Tl)a
(a=l,2,
...,n)...
-sq a
we have
so that
(/=!,
Let
be any root of the equation f(s)

positive integer for which all the functions
s-ii
m
Sj i)
When
s is
(s
are finite for the value sî of
expand
+ hî) (s
let
be the smallest
taken sufficiently near
s x i,
we can
M , h^, g^, h^,

c:
c2
-.,
s i i)~ m+l
+ ((// + hp i) (s
1
sî)""
+ ...,
denote real constants; and we can suppose the

so chosen that the quantities g^ and AM are not zero.
...
<jr
quantities
c.2n
Substituting this value of q^ in equations (B),

of (s
and
q^ in a series of the form

(<7
where
s.
= Q,
2,..., 2n).
m we have
Sii)~
and equating the
coefficients
H
H (g
(gi,
l
g2
. . .
.
2>
=
gm ) n +a + sji* =
gzn)a
Sihn+a
(C),
TT /I
and on equating the
coefficients of (s
(0
/\
s 1 t )~ m+1 ,
when
we have
>
when m 1
=
[c n+a when m
[0
w* = o
la
=0
>
(a
= 1,2,... ,)
(D).
201
84]
Now
by Euler
theorem on homogeneous functions we have

Zn
2H(g g
1}
gm )
...,
= 2 yJHfr, g
A=
...,g^)\,
or
by
(C),
%H (g
l;
g2n )
gs, ...,
= s, 2
a=l
and similarly
(ga h n+a
- h a g n+a
),
(gji n +*
- h a gn+a
),
2H
. .
(/*!, 7*3,
m ) = Si 2=
a
from which
evident that
it is
=l
o.
Moreover, the
h a g n+a )
(g a h n+a
not zero.
is
two of equations (C) give
first
2n
2 hiH(gl) gi ,...,gm)i + 8i 2 (hji n+a a=l

A=l
and the
2n
2
=l
g^
(h,
A an ) A
A., ... ,
a=l
But from the
first
- g gn +) =
(g a g n +
two of equations (D), when
\=1
g2
...,g2n
1, \ve
>
(F).
have
2 h ,H(g
(E),
two of equations (C) give
last
A.
=Q
h a hn+a )
)\-s 2
l
a=l
(h a h
n+aL
-h a hn+ai )- 2
a=l
(g a h n+a
- ha gn+a ) =
(G),
and from the
two of equations (D) we have
last
Zn
gjl(&,k{>
X=l
>..,h^} ),
+s 2
l
a=l
(g a g n+a
2 (g a hn+a - h
-ga g n+a )+ a=l
g n+a ) =
(H).
Also since
.ST is
homogeneous
we
of the second degree in its arguments,
have the identities

Zn
2n
2 h^H (g,,g
2,
. . ,
A=l
gzn )\
= 2 #*# (A/, V,
X=l
(K)
^//)A
I
2w
2n
2 gJH Qh,
and
h2
A 2n ) A
A=l
From
equations (E), (H), (K)
2
o=l
(gji n+
a-h
gn+a )
a=l
(ga h n+a
2
=l
- h a gn+a = - s 2
l
-h
ai
h n+a )-s l
(L).
2
a= l
(ga g n+a
- g a g n+a
),
we have
n
(h a h n+a
- h a h n+a ) + s 2
l
a=l
a=l
Comparing these equations, we have

n
gm
n
(h a.h n+a
(F), (G), (L)
...,
we have
n
sl
<x
and from equations
= 2 h^H (g, g
A=l
(ga hn+ai
-h
n+<1
0,
(g a g n +*
~ ga gn+*).
202
which
m
m
[CH.
vn
The assumption that

contrary to what has already been proved.
which was used in obtaining equation (G), must therefore be false
must therefore be unity, and consequently when f(s) is divisible
- sîf,
by (s
>
is
1,
each of the functions f (s)^
Now let Sj
s2 ,
.,
k
by (s-sî) -\
is divisible
be the moduli of the distinct roots of the equation/(s) = 0,
sr
so that the functions

f(s)^/f(s) are infinite only for s= s^,
1
then denoting the coefficient of (s
in the Laurent
Spi)
in
/()W/(*)
powers of
(s
- sp i)
(X, /i) p
and
s r i;
expansion of
by
(X, fi)p
where
s 2 i, ...,
(X, /*)/ are real,
i (X,
fj,) p
and observing that the only poles of the

= sp i, and that these are simple poles,
function f(s)^/f(s) are the points s
we have
+ i(\
s-si
n)p
(X,
and therefore
<f>
(t)^
(X? /*)P
+ i (\
(\
rip
p =i
in powers of
(\,f*)-i\,n
the coefficient of 1/5 in the Laurent expansion of
is
<-)
M
,
Spi
rip
s
~ i (\
+ Spi
AQp
l
j
s.
But the
8 (t ~^
coefficient of I/s in the Laurent
sp i) is
expansion of e
f(s
s{t t
an(j t ^ e coefficient of
in
the
Laurent
of
e
l/s
expansion
/(s + sp i)
-<o)i
is
e~
Sf>(t
<>
)l
;
we have
therefore
{(X, /i)p
cos sp
(t
t )
- (X,
fi) p
sin sp
)},
P=I
and so
qn
finally
2 2 [&+. {(,AOP cos Sp

a=l
(*-*<>)
-(a, M)P sin 5P
p=l
- ^ a {(n + a, /i)p cos s

T/it s
motion.
p (i
- (n + a, /x)p
sin sp
(t
- ^ )}
- 1 )}]
O=
1, 2,
.,
2w).
formula constitutes the general solution of the differential equations of

Hence finally we see that when the total energy of a system vibrating
about a state of steady motion is a positive

definite form, the vibratory motion
can be expressed in terms of circular functions of t, and the steady motion is
is 1}
stable; the periods of the normal vibrations are ZTT/S!, 27r/s2 ..., where
,
are the roots of the determinantal equation

= 0, whose order in s 2
f(s)
to
the
number
coordinates
the
equal
of non-ignorable
of
system.
is.2 ,
is
...
The above
investigation
is
valid
whether the determinantal equation has
repeated roots or not.

Between the
coefficients (X,
p.} p
(X, /x) p
(^ M)P= -
and so in particular
(p-,
there exist the relations
x) P
(X,
(x, /i)p
X) p = 0.
=(M>
^)p
These relations follow from equations which

/()>
203
84, 85J
Cin virtue of their definitions) are
true for
/(*)*! namely
/(*)
-/(-),
Example. If the number of degrees of freedom of the system, after ignoration of the
ignorable coordinates, is even, shew that when the ignorable velocities are large (e.g. if
the ignorable coordinates are the angles through which certain fly-wheels have rotated,
would imply that the fly-wheels are rotating very rapidly), half the periods of
vibration are very long and the other half are very short, the one set being proportional
to the ignorable velocities and the other set being inversely proportional to these
this
velocities.
It was pointed out by Poincare* that the discussion of stability by the

method of small oscillations does not take account of some features which are
Thusf, consider a particle moveable

which
rotates with constant angular
bowl
on the inner surface of a spherical
If the bowl be perfectly smooth, the
velocity about its vertical diameter.
likely to be present in actual problems.
the
equilibrium of the particle in the lowest position is certainly stable,
But if there be the slightest
rotation of the bowl having no effect on it.
friction between the particle and the bowl, and if the angular velocity of the
bowl exceeds a certain value, the particle will work its way outwards in
a spiral path towards the position in which
bob of a conical pendulum.
85.
Examples of
A number
will
now be
(i)
of illustrative cases of vibration about a state of steady motion
considered.
= a, z =
particle is describing the circle r
2
is F=((r, 2), where r
when r=a,
If
we
we have
rotates with the bowl like the
vibrations about steady motion.
= b.
To find
write
for the kinetic
b,
in the cylindrical field of force in
= #2 4-y 2
which the potential energy

zero
it
the conditions
x=rcos8,
it being given that clV/dz

for stability of the motion.
,
is
y = rsir\8,
and potential energies of the
particle,
whose mass
will
be denoted
by m,
The integral corresponding to the ignorable coordinate 6 is mr 2 6 k, where k

The modified kinetic potential after ignoration of 6 is therefore
constant.
R=T- V-W
Acta Math. vn. (1885), p. 259.

is due to Lamb, Proc. Roy. Soc. LXXX. (1908),
t This illustration
p. 168.
is
204
[CH.
vn
For the steady motion we must have
the latter condition
is satisfied
2
3
by hypothesis, and the former gives k = ma
c(f>/da.
We
have therefore
is
z-^-.
r,
<
Writing
and neglecting terms above the second degree in p and

0aa
As no terms linear
3
-
we have
(Pa
in p or f occur, this is essentially the
about equilibrium, and the condition for stability
shall be a positive definite form,
both be positive.
is
same as a problem
of vibrations
79) that
that
i.e.
shall
f,
+-a
(f) a )
<f>bb
<t>
ab
and
(p bb
These are the required conditions
for stability of the steady
motion.
Corollary.
If a particle of unit
mass
describing a circular orbit of radius a in a
is
plane about a centre of force at the centre of the circle, the potential energy being
where r is the distance from the centre, the modified kinetic potential is
(3
<aa
where
r=a + p,
so the condition for stability
+a

and the period of a vibration about the circular motion
is
To find the period of the vibrations about steady circular motion of a particle
(ii)
moving under gravity on a surface of revolution whose axis is vertical.
Let zf(r) be the equation of the surface, where (z, r, ff) are cylindrical coordinates
with the axis of the surface as axis of z. If the particle is projected along the horizontal
tangent to the surface at any point with a suitable velocity, it will describe a horizontal
circle
on the surface with constant
velocity.
Let a be the radius of the circle
we
shall
take the mass of the particle to be unity, as this involves no loss of generality.
The
kinetic potential
is
The integral corresponding to the ignorable coordinate 6 is r2d

kinetic potential of the system after ignoration of Q is therefore
r}}
-gf(r)
- F/2r2
= k,
and the modified
205
85]
is thus reduced to that of finding the vibrations about equilibrium of the

kinetic potential. The condition for
one
degree of freedom for which R is the
system with
The problem
is
equilibrium
= 0,
/)
\or l r=a
or
V=ga\f
(a\
and this gives
R = ir* {1 +/
Writing
r=a + p,
The equation
where p
is
small,
(r)}-gf(r)
and expanding
in
2
.
powers of
p,
we have
of motion
<^/
dp
dt\df>
therefore
is
-ga?f (a)/2r
p{l
and the condition
+/
(a)}+0p{/"()
+ |/
()]
= 0,
for stability is
<
the period of a vibration being

27T
Tg
axis
Example. If the surface is a paraboloid of revolution whose
vertex downwards, shew that the vibration-period is
where
is
and
vertical
the semi-latus rectum of the paraboloid.
To determine
(iii)
is
the vibrations about steady
motion of a top on a perfectly rough
plane.
Let A denote the moment of inertia of the top about a line through its apex perpen
dicular to its axis of symmetry, and let 6 denote the angle made by the axis with the
the mass of the top, and h the distance of its centre of gravity from its apex
vertical,
that after ignoring the Eulerian angles (p and ^, the angle

determined by solving the dynamical system defined by the kinetic potential
then we have seen
where a and
Let
a,
(71)
b are constants
n be the values
we have
depending on the
of 6
and
<j>
initial
is
respectively in the
steady motion, so
72)
A n z cos a + Mgk = bn,
An sin 2 a = ab cos a.
To
e
discuss the vibratory motion of the top about this form of steady motion, we write
in ascending powers of x, neglecting
is a small quantity, and expand
= a + x where x
of the
powers of x above the second and eliminating a and b by use
we thus obtain for R the value
RÂJP-lAx* {n
sin 2 a
+ (n cos a - MghjA n) 2
}.
last
two equations
206
The equation
As
x is therefore
x + {n 2 sin 2 a + (n cos a - Mgh/ A nf] x = 0.
x is positive, the state of steady motion is
stable
and
the period
is
27T
The sleeping
(iv)
vn
of motion for
the coefficient of
of a vibration
[OH.
{* - 2Mgh cos a/ A + J*g
h*IA*n*}~t.
top.
If we consider that form of steady motion of the top in which a is zero, so that
the axis of the top is permanently directed vertically upwards, the top rotating about this
axis with a given angular velocity, the method of the preceding example must be modified,
since
now the form
of steady motion in which a
is
a small constant
is
to be regarded as a
vibration about the type of motion in which a is zero so that we may now expect to have
two independent periods of normal vibration, the analogues of which in the previous
:
example are the period of the steady motion and the period of vibration about
As
71, the kinetic
in
and
potential energies of the top are
+
T=\
V= Mgh cos 6.
2
4<9
M(/>
The
sin 2
<?+|
tf(^ +
after ignoration of
= ^(^ +
we obtain
is
<
<f)
From
and
77,
system the value
+ b$ cos 6 -Mgh cos 6.
not a small quantity throughout the motion,
the quantities
is
<9),
R=\Afr+%A<& sin 6
As
cos
In the two last terms we can replace cos 6 by (cos 6

thus added disappear from the equations of motion.
and
<9)
for the kinetic potential of the

2
cos
<j>
integral corresponding to the ignorable coordinate
and hence
it.
1),
since the terms
we take
b<j)
and Mgh
as coordinates in place of
where
these equations, neglecting terms above the second degree in
77,
rj,
we have
and so we have
&-lAt*
The equations
of motion are
"
dt
bf,-Mg
(Arj-b-Mghr,=0.
If STT/X is the period of a. normal vibration, on substituting
differential equations
and eliminating
J and
- A 2 /I - Mgh
2
or
(X
r)
Ke
in these
= 0,
ib\
- \2 A
ib\
=7
K we obtain the equation

Mg,
A + Mgh) 2 - 6 2 X 2 = 0.
The two roots of this quadratic in X2 give the values of X corresponding to the two
normal vibrations we have therefore to determine the nature of these roots.
:
85, 86]
The
solution of the quadratic
207
is
The values of X are therefore real or not according as b 2 is greater or less than 4AMgh.
In the former case the steady spinning motion round the vertical is stable in the latter
:
case, unstable.
must not be supposed, however, that
It
in the unstable case the axis
necessarily departs very far from the vertical: all that is meant by the term
is that when b 2 <.4AMgh the disturbed motion does not, as the disturbance is
of the top
"unstable"
indefinitely
diminished, tend to a limiting form coincident with the undisturbed motion.
As a matter of fact, if b -4AMgk, though negative, is very small, it is possible for the
axis of the top in its "unstable" motion to remain permanently close to the vertical but
in this case the maximum divergence from the vertical cannot be made
indefinitely small
(for a given value of 6) by making the initial disturbance indefinitely small*.
2
Vibrations of systems involving moving constraints.
86.
If a dynamical system involves a constraint which varies with the time

(e.g. if one of the particles of the system is moveable on a smooth wire or
surface which
made
to rotate uniformly about a given axis), the kinetic

potential of the system is no longer necessarily composed of terms of degrees
2 and
in the velocities
terms which are linear in the velocities may also
is
The equations which determine the
occur.
vibrations of such a system will
therefore in general include gyroscopic terms, even when the vibration is

about relative equilibrium the solution can be effected by the methods above
:
developed
example
for
The
the problem of vibrations about steady motion.
following
will illustrate this.
Example. To find the periods of the normal vibrations of a heavy particle about its
position of equilibrium at the lowest point of a surface which is rotating with constant
about a vertical axis through the point.
angular velocity
u>
of the particle, referred to axes which revolve with the

(.v, ?/, z) be the coordinates
surface, the axes of x and y being the tangents to the lines of curvature at the lowest
Let
point,
and the axis of
being vertical.
i2
z= ~
%>i
The
kinetic
Let the equation of the surface be
+ ~- + terms
of higher order.
2p 2
and potential energies
of the particle are
V= mgz.
The
kinetic potential of the vibration-problem
is
therefore
y2\
5T(y&
/Pi
*
discussion of the stability of the sleeping top

129132, 292.
(1897), pp.
is
+ j-
^Pz/
given by Klein, Ball. Amer. Math. Soc. in.
208
The equations
of motion are
._
[CH.
=
cy
Pi
normal vibration is 27T/X, we have (substituting

equations, and eliminating A and B)
If the period of a
the differential
-2wiX
or
(X
The
- g/pi)
(X
2
o>
XAe, y = Be
lX
in
=0,
- glp z } - 4X2
= 0.
roots of this quadratic in X 2 determine the periods of the
normal vibrations.
1.
particle
moves on a curve which
rotates uniformly about a fixed axis, the

the
only on its position as defined by
potential energy V () of the particle depending

Shew that the period of a vibration about a position of relative rest on the
arc s.
curve
is
dVd
where
r is the distance of the particle
from the
axis.
Determine the vibrations of a solid horizontal circular cylinder rolling inside a

2.
the
hollow horizontal circular cylinder whose axis is fixed, shewing that the length of
b
is the radius and J/the
where
is
(b -a) (3M+m)l(2M+m)
simple equivalent pendulum
the mass, of the inner cylinder.
mass, of the outer cylinder, and a is the radius and
;
(Coll.
Exam.)
on a perfectly rough
and radius a
thin hemispherical bowl of mass
3.
is in contact with the inner surface of the bowl,
horizontal plane, and a particle of mass
small oscillations, the motion of
supposed smooth. Shew that when the system performs
the particle and the centre of gravity of the bowl being in one plane, the periods of the
is
normal vibrations are 2^/^X7 and 2^/^X3, where A t and X 2 are the roots of the equation
- aA)
- - a A)
0.
(Coll. Exam. )
%
ma\g
4.
string of length 4a
respectively,
and
is
is
M=
(\g
(g
loaded at equal intervals with three weights m,
suspended from two points
and
perform small vertical vibrations, the length of the simple equivalent
M and m
Shew that
symmetrically.
pendulum
if
is
sin (a - ft} cos (a- 0)

2
cos
a + sin ft cos2 ft
sin a
a cos a cos
where a and
ft
ft
are the inclinations of the parts of the string to the vertical.

(Coll.
Exam.)
A uniform bar whose length is 2a is suspended by a short string whose length is I

that
the time of vibration is greater than if the bar were swinging about one
prove
(Coll. Exam.)
extremity in the ratio l+9/32a 1 nearly.
;
5.
209
vn]
6.
An elliptic cylinder with plane ends at right angles to its axis rests upon two fixed
smooth perpendicular planes which are each inclined at 45 to the horizon. Shew that
there are two stable configurations and one unstable, and that in the former case the
length of the equivalent
ab (a
a and
is
pendulum
2
+ 62 )/2 V2 (a - 6)
(a
+ 6),
b being the lengths of the semi-axes.
Exam.)
(Coll.
rough circular cylinder of radius a and mass m is loaded so that its centre of
at a distance h from the axis, and is placed on a board of equal mass which
can move on a smooth horizontal planei If the system is disturbed slightly when in a
7.
is
gravity
shew that the length of the simple equivalent pendulum
position of stable equilibrium,
k 2 lh+\ (a- h} 2 /k, where
mk1
the
is
moment
is
of inertia of the cylinder about a horizontal
axis through its centre of gravity.
Exam.)
(Coll.
One end
is freely jointed to a point in a

of a uniform rod of length b and mass
smooth vertical wall the other end is freely jointed to a point in the surface of a uniform
and radius a which rests against the wall. Shew that the period of the
sphere of mass
8.
vibrations about the position of equilibrium
p 2 (sin
a and
sin 2 (a
/3
- /3)
is ZTT/JO,
where
+| cos a sin (a -/3) + f sin /3 cos 2 /3} =
-?--
2
(a sin a cos a + b sin
/3
cos 2 /3),
being given by the equations
a sin a + b sin /3-a=0,

(
m + M) tan - M tan a = 0.
Exam. )
(Coll.
ft
9.
thin circular cylinder of mass
and radius b rests on a perfectly rough
horizontal plane, and inside it is placed a perfectly rough sphere of mass
and radius a.
If the system be disturbed in a
plane perpendicular to the generators of the cylinder, find
the equations of finite motion, and deduce two first integrals of them
be small, shew that the length of the simple equivalent pendulum is
and
if
the motion
l4M(b-a)/(WM+7m.).
I,
1899.)
sphere of radius c is placed upon a horizontal perfectly rough wire in the

form of an ellipse of axes 2a, 2b. Prove that the time of a vibration under gravity about
10.
the position of stable equilibrium
where
is
F=2c 2/5
that of a simple pendulum of length
and d 2 = c 2 -6 2
(Coll.
given by
Exam.)
11.
A rhombus of four equal uniform rods of length a freely jointed together is laid
on a smooth horizontal plane with one angle equal to 2a. The opposite corners are
connected by similar elastic strings of natural lengths 2a cos a, 2a sin a. Prove that if
one string be slightly extended and the rhombus left free, the periods during which
the strings are extended in the subsequent motion are in the ratio
(cos a)$
mass
(sin a)^.
(Coll.
Exam.)
attached by n equal elastic strings of natural length a to

the fixed angular points of a regular polygon of n sides, the radius of whose circumscribing
circle is c.
Shew that if the particle be slightly displaced from its equilibrium position in
12.
particle of
is
the plane of the polygon, it will execute harmonic vibrations in a straight line, the length
the simple equivalent pendulum being 2mgac/n\ (2c - a), and that for vibrations
perpendicular to the plane of the polygon, the corresponding length will be mgacjriX (c - a),
X being the modulus of each string.
of
w. D.
14
210
The energy -equation
13.
of a particle
2
f(x)x =
is
2<f)
(x)
+ constan t,
If
and a is a value of x for which
(x~) is zero.
which does not vanish for x=a, shew that the period
<<
2p )
<$>
4
v
where h
is
[en.
(i/2/>)
(2/>)
(x) is
the
first
derivative of
<j>(x)
of a vibration about the position
is
/ra
2p
h
F (1/2/3+^) (
(a)
4/30(
the value of (x- a) corresponding to the extreme displacement.
)
(Elliott.)
A cone has its centre
of gravity at a distance c from its axis, there being in other

If the cone oscillates on a horizontal
respects the usual kinetic symmetry at the vertex.
and
the
be
shew
that
the length of the simple equivalent
plane
plane
perfectly rough,
]
4.
is
pendulum
(cos a / Me)
whereas
if
2
(A sin a +
this plane be perfectly smooth, the length is

2
(cos a/ Me) (sin a/A
15.
C cos 2 a),
+ cos2 a/C).
(Coll.
Exam.)
of equal uniform rods each of length 2a are freely jointed at a common

arranged at equal angular intervals like the ribs of an umbrella. This cone
number
extremity and
of rods is put over a smooth fixed sphere of radius 6, each rod being in contact with the
Shew that, if the system be slightly disturbed so that
sphere, and rests in equilibrium.
the hinge performs vertical vibrations about the position of equilibrium, their period is
l+3sin 2 a\4
where sec
16.
where
c is
w 1+2
a sin a = a/6.
sin 2 a,
sm
a.
heavy rectangular board
I,
1896.)
symmetrically suspended in a horizontal position

by four light elastic strings attached to the corners of the board and to a fixed point
Shew that the period of the vertical vibrations is
vertically above its centre.
is
^WM)
the equilibrium distance of the board below the fixed point, a
a semi-diagonal,
=(
+ c2 )2,
and X
is
the modulus.
is
the length of
(Coll.
Exam.)
heavy lamina hangs in equilibrium in a horizontal position suspended by three

Shew that the normal vibrations are
vertical inextensible strings of unequal lengths.
(1) a rotation about either of two vertical lines in a plane through the centroid, and
17..
(2)
a horizontal swing parallel to this plane.

18.
(Coll.
Exam.)
A uniform rod of
of length b
is
length 2a is freely hinged at one end, at the other end a string

attached which is fastened at its further end to a point on the surface of a
homogeneous sphere of radius

motion of the system when
c.
If the masses of the rod
slightly disturbed
from the
and sphere are equal, find the

vertical, and shew that the
equation to determine the periods is

3 2
- 3
2a6c/x
gp? (66c + 19ca + 5a6) +g n (35a +156 + 2k ) g = 0.
1
Exam.)
upon a rough
suspended by
(Coll.
19.
A uniform
wire, in the shape of
an
ellipse of semi-axes a,
6,
rests
horizontal plane with its minor axis vertical and a particle of equal mass is
a fine string of length I attached to the highest point. If vibrations in a vertical plane
be performed, prove that their periods will be those of pendulums whose lengths are the
value of x given by the equation
{x (36
where k
is
- 2a 2 /6) + 56 2 + i?} (x-l) + 46 2 / = 0,
the radius of gyration about the centre of gravity.
(Coll. Exarn.)
211
vn]
A fine inextensible string has its ends tied to two fixed pegs in a horizontal
whose distance apart is three-quarters of the length of the string. The string
also passes through two small smooth rings which are fixed to the ends of a uniform
The rod hangs in equilibrium
straight rod whose length is half that of the string.
in a horizontal position and receives a small disturbance in the vertical
plane of the
Shew that initially its normal coordinates in terms of the time are Lcos(pt+a)
string.
and Mcosh (qt + fi), where p 2 and q 2 are the roots of the equation
20.
line
yA
^ 9a X _
2
ff*
a-*
= o.
Exam
(ColL
heavy uniform rod of length 2a, suspended from a fixed point by a string
Shew that the periods of the
slightly disturbed from its vertical position.
normal vibrations are 2nlp l and 27r/p2 where pi2 and 2 2 are the roots of the
equation
21.
of length
b, is
p>
abp*-(4a
22.
circular disc, mass M, is attached by a string from its centre C to a fixed

is fixed to the disc at a point
on the rim. Find the
particle of mass
point 0.
OC and CP
which
equations of motion on a vertical plane in terms of the angles 6 and
make with the vertical, and prove that if the system vibrates about the
position
of
equilibrium the periods in these coordinates are given by the equation
where a
is
the length
(M+ m) (p~a - g] {(M+ 2i) cp 2 - 2mg] = 2m 2cap\

of the string OC and c the radius of the disc.
Exam.)
(Coll.
23.
A hemispherical bowl of radius 2b rests on a smooth table with the plane of its
rim horizontal within it and in equilibrium lies a perfectly rough sphere of radius
b, and
mass one-quarter of that of the bowl. A slight displacement in a vertical
plane con
taining the centres of the sphere and the bowl is given prove that the periods of the
consequent vibrations are 2ir/pi and 2ir/p z where p^ and p% 2 are the roots of the
;
equation
1566%2 - 2606a^ + 75#2 = 0.
uniform circular disc of mass
Exam.)
(Coll.
and radius a is held in equilibrium on a

smooth horizontal plane by three equal elastic strings of modulus X, natural
length 1 and
24.
stretched length I.
The strings are attached to the disc at the extremities of three radii
equally inclined to one another and their other ends are attached to points of the plane
Shew that the periods of vibration of the disc are
lying on the radii produced.
and
J )}
where /i=2m^
/3X.
particle is describing a circle

varying as the nth power of the distance.
25.
be
less
than -
I,
1898.)
under the influence of a force to the centre

Shew that this state of motion is unstable if n
3.
Shew
that, if the force vary as e
the radius of the circle

26.
particle
is less
moves
2
"/r
the motion
is
stable or unstable
according as
or greater than a.
in free space
(Coll.
Exam.)
under the action of a centre of force which varies
as the inverse square of the distance and a field of constant force shew that a
circle
described uniformly is a possible state of
steady motion, but this will be stable only
provided the circle as viewed from the centre of force appears to lie on a right circular
cone whose semi-vertical angle is greater than
arccosj.
(Coll. Exam.)
:
142
212
27.
which
[CH.
under the influence of two centres of force

Prove that the motion is stable if
which a radius of the circle subtends
a circle
uniformly
particle describes
attract inversely as the square of the distance.
3 cos e cos
where 6 and
<<1,
<f>
are the angles
at the centres of force.
1889.)
I,
on the interior of a smooth cone with

heavy particle is projected horizontally
from the apex is c and the
its axis vertical and apex downwards; the initial distance
a horizontal circle should be
that
condition
the
Find
is
a.
cone
of
the
semi-vertical angle
and shew that the time of a vibration about this steady motion is that
28.
described;
of a simple
pendulum of length ^csec
(Coll.
a.
Exam.)
A circular disc has a thin rod pushed through its centre perpendicular to its
to the radius of the disc prove that the system
the
length of the rod being equal
plane,
cannot spin with the rod vertical unless the velocity of a point on the circumference
after falling from rest
of the disc is greater than the velocity acquired by a body
(Coll. Exam.)
ten times the radius of the disc.
29.
vertically through
Prove that for a symmetrical top spinning upright with sufficient angular
two types of motion, differing slightly from the steady motion
velocity for stability, the
functions of the time,
in the upright position, which are determined by simple harmonic
30.
and that
are the limits of steady motions with the axis slightly inclined to the vertical,
the period of the vibrations is the limiting value of that which corresponds to steady
motion in an inclined position when the inclination is indefinitely diminished.
Exam.)
(Coll.
One end of a uniform rod of length 2a whose radius of gyration about one
31.
end is k is compelled to describe a horizontal circle of radius c with uniform angular
Prove that when the motion is steady the rod lies in the vertical plane
velocity w.
an angle a with the vertical given by
through the centre of the circle and makes
w2
Shew that the
(FX
(k
+ ac cosec a) = ay sec a.
are 27r/\i, 27r/X 2 where X l5 X 2 are the

periods of the normal vibrations
sin a
2
a>
ac)
X2 sin a -
2
a>
ac -
2
a>
&2 sin 3 a) =
2
4a>
X 2 sin 2 a cos 2
a.

disturbed from
I,
1889.)
conical pendulum when

32.
Investigate the motion of a
harmonic oscillation of the point of support. Can the
steady motion by a small vertical
such a disturbance?
(Coll. Exam.)
steady motion be rendered unstable by
33.
it
The middle point
its state of
of one side of a uniform rectangle is fixed and the line joining

is constrained to describe a circular cone
to the middle point of the opposite side
The rectangle being otherwise free,

a with uniform angular velocity.
about the
the positions of steady motion and prove that the time of a vibration
sin
a.
divided
of
revolution
the
to
by
period
position of stable steady motion is equal
of semi -angle
find
(Coll.
Exam.)
centre of gravity
A solid of revolution, symmetrical about a plane through

34.
b which is
is suspended from a fixed point by a string of length
perpendicular to its axis,
The mass
of length 2a.
axis
this
the
of
axis
of
the
being
end
one
to
solid,
attached
moments of inertia at its centre of gravity are
and its
the solid is
its
of
M,
principal
from the state of steady motion in which the

,
is spinning on its axis with angular velocity
the
and
axis
are
and
body
vertical,
string
where p^ and p 22 are
2
shew that the periods of the normal vibrations are îrlp l and
(A, A, C}.
If the solid is slightly disturbed
27r//>
the roots of the equation

HaPgp*-
= (g- fy2 (Mag + Cnp - Ap*}.

)
213
vn]
A symmetrical top spins with its axis vertical, the tip of the peg resting in
35.
a fixed socket. A second top, also spinning, is placed on the summit of the first, the tip
Shew
of the peg resting in a small socket.
that the arrangement
is
stable provided the
equation
(Mcgx* +
has
M,
roots real
all its
CQx + A {(M c + MK) gx* + C Q x + (A + Mh?}} = J/ 2AV)

1
their masses, C,
fl,
ii
being the spins of the upper and lower tops respectively,

moments of inertia about the axis of figure, A, A about
their
perpendiculars through the pegs,

the distance between the pegs.
c,
the distances of the centroids from the pegs, and h

A homogeneous body spins on a smooth horizontal plane in stable steady motion,

36.
with angular velocity w about the vertical through the point of contact and the centre of
The body is symmetrical about each of two perpendicular planes through the
gravity.
The principal radii of curvature at the vertex on which it rests are pi p 2 the
of inertia about the principal axes through the centre of gravity (parallel to the
lines of curvature) are respectively A and
and that about the vertical is C. The
vertical.
moments
height of the centre of gravity about the vertex

weight of the body.
Shew that the
following conditions
must be
is
a = a l + p l = a.2 +p 2
satisfied
and Xw 2
is
(i)
(ii)
(iii)
The value
(A
if
of X
must not
+ B - C] \JB {M +
lie
(A
between the two values
- Cf)Y
JA
[a.2
B 4- a (E v
the two radicals in the expression are both real.
I,
1897.)
the
CHAPTER
NON-HOLONOMIC SYSTEMS.
VIII
DISSIPATIVE SYSTEMS
Lagranges equations with undetermined
87.
multipliers.
We now
proceed to the consideration of non-holonomic dynamical systems.

In these systems, as was seen in 25, the number of independent coordinates
In) required in order to specify the configuration of the
>
(<?i>
<?2,
system at
any time is greater than the number of degrees of freedom of the system,
owing to the fact that the system is subject to constraints which will be
supposed to do no work, and which are expressed by a number of nonintegrable* kinematical relations of the form
An A
where
q%,
lk
12
+ A 2k dq +
2
...,
A nm
+ A nk dq n + Tk dt =
T ..., Tm are
...
7\,
(k
= l,2,
...,
functions
given
m),
of
q 1}
t.
Qn
dq
>
familiar example of such a system is that of a body which is constrained to

without sliding on a given fixed surface the condition that no sliding takes place is
expressed by two relations of the type given above. A still simpler example is that of ,a
The most
roll
vertical wheel with a sharp edge which rolls on a horizontal sheet of paper, as in the
integraph of Abdank-Abakanowicz and the integrator of Pascal the wheel moves only in
its own instantaneous plane, the friction at the sharp edge preventing it from slipping
:
If (x, y} are the rectangular coordinates of its point of contact with the
its plane, we have in this case the non-holonomic equation
sideways.
paper, and $ the azimuth of

of condition
tan
dy
The number
dx = 0.
<p
being m, the system will have

not
possible to apply Lagrange s equations
m)
(n
of the Lagrangian equations will
an
extension
a
but
to
such
system,
directly
to
discuss
the motion of non-holonomic
enable
us
which
will
be
now
given
of kinematical relations
degrees of freedom
it is
systems in a way analogous to that previously developed
for
holonomic
systems.
Consider then a non-holonomic
system, whose configuration at any

n coordinates q l} q 2 ..., q n
let the
instant is completely specified by

kinetic energy be T, and let the kinematical conditions
holonomic constraints be expressed by the relations
A.kdq,
+ A^dq +
2
...
* If these relations were

integrable,
(<?i>
92
which
1
>
is
1 n)
(k
due
to the non-
= 1,
2,
m).
would be possible to express some of the coordinates

i Q terms of the others, and the n coordinates would therefore not be independent
contrary to our assumption.
it
Non-holonomic Systems.
87]
Now
215
Dissipative Systems
is open to us either simply to

regard the system as subject to
these kinematical conditions, or in place of these to
regard the system as
acted on by certain additional external forces, namely the forces which have
it
be exerted by the constraints in order to compel the system to fulfil the

we shall for the present take the latter point of
view.
Let
Qi 8qi+Q*Bqt +... + Qn *q*
to
kinematical conditions;
be the work done on the system by these additional forces in an

arbitrary
is now not restricted to
satisfy the
displacement (Sq^ 8q 2 ..., Sq n ) (which

kinematical conditions), and let
,
+Q
8q 1
Sq 2 +...
+ Qn Sq n
be the work done on the system by the original external forces in this dis
placement. Since the substitution of additional forces for the kinematical
relations has made the system holonomic, we can
apply the Lagrangian
we have therefore
equations
;
as the equations of motion of the

system.
The
forces
Q/>
Qz,
Qn are unknown
>
but they are such that, in any
displacement consistent with the instantaneous constraints, they do no work.

It follows that the quantity
Qi
is
zero
for
all
values
dq
of
+ Q2 dq +
...+
the
ratios
dq :dq 2
l
equations
A*dql + A ti dqa +
t
Qn dq n
...
...
+ A nk dq n =
dqn which
satisfy
the
0;
hence we must have
Qr =
\A n + \a4 + ...+ \m A rm
where the quantities \ l} \2

altogether the (n
+ m)
dT
dT\
and
these
are
+ A 2k q +
sufficient
X.
q n î,
of this set of equations*.
fc,
>
)u>
The extension
\ m are independent of
r.
1, 2,
We
n),
thus have
to
...+
A nk qn +
determine
the
The problem
is
2^
(n
(k=l,2,
+ m) unknown
...,
m),
quantities
thus reduced to the solution
of Lagrange s equations to non-holonomic

systems is due to Ferrers, Quart.
C. Neumann, Leipzig Berichte, XL.
(1888), p. 22: and Vierkandt,
Math. u. Phys. in. (1892), p. 31.
Journ. Math. xn. (1871), p. 1
Monatshefte fur
...,
equations
îfcffi
<1\,
(r
216
Equations of motion referred
88.
to
Dissipative Systems [CH. vin

axes moving in any manner.
The method given in the preceding article depends essentially on the

reduction of the non-holonomic system to a holonomic system by introducing
the forces due to the non-holonomic constraints. In practice, this is often
most conveniently done by forming separately the equations of motion of
each of the bodies of the system. It is moreover frequently advantageous
to use axes of reference which are not fixed either in space or in the body,
and we shall now find the equations of motion of a rigid body referred to
axes which have their origin at the centre of gravity of the body, and are
turning about
it
in
any manner*.
be the centre of gravity of the body, and let Gxyz be the moving
(u, v, w) be the components of velocity of the centre of gravity
resolved parallel to these axes, and let (0 l
2
#3) be the components of
Let
Let
axes.
0<
angular velocity of the system of axes Gxyz resolved along the axes them
w 2 3 ) be the components of angular velocity of the
selves; further let
&>
(&>j,
body, resolved along the same axes. Then ( 64) the motion of G is the same
as that of a particle of mass M, equal to that of the body, acted on by forces
equal to the external forces which act on the body (including all forces of
constraint, except the molecular reactions
between the constituent particles

Gxyz of
of the body) let (X, Y, Z) be the components parallel to the axes

these external forces.
;
The component of velocity of G parallel to Gx is u, and consequently

v0 + wQ2 we
17) the component of its acceleration in this direction is u
3
have therefore the equation
which can be written
where
(u, v, w,
of
denotes the kinetic energy of the body, expressed in terms of

and similar equations can be obtained for the motion
3 );
2
&>
<w
a>!,
>
parallel to the axes
Gy and
Gz.
Consider next the motion of the body relative to G, which ( 64) is

62, 63, we see that the angular
independent of the motion of G from
;
body about the axis Gx is dT/dwi,

increase of angular momentum about an axis fixed
momentum
of the
stantaneously coinciding with
Gx
in
space
and
in
is
A f^.\ -B
*
so that the rate of
In the applications of this method, the axes are usually chosen subject to the condition
moments and products of inertia of the body with respect to them do not vary but this
that the
condition
is
not essential.
Non-liolonomic Systems.
88, 89]
denote the moments

M,
have
therefore ( 40)
we
Gxyz,
If L,
217
Dissipative Systems
of the external forces about the axes
dT
dT _
=
**j
and two similar equations.
Hence
motion of the body
finally the
l*.(W\-aW
ff
d_T_
is
d_fdT^_.dT
,dT
dt
dt\du)
d.
determined by the six equations
(W\ _
dt\dw/
du
dt\dwj
dv
9o>i
3h
be observed that these are really Lagrangian equations of motion

in terms of quasi-coordinates, and could have been derived by use of the
It will
theorem of
30.
let (w 1} u 2
a)
Example. If the origin of the moving axes is not fixed in the body,
be the components of velocity of the origin of coordinates, resolved parallel to the
of angular velocity
instantaneous position of the axes; let (0 l5 2
3 ) be the components
of the system of axes, resolved along themselves; let (vi, v 2 v 3 ) be the components of
situated at the origin of
velocity of that point of the body which is instantaneously
,
coordinates; and let

referred to the
o>
(&>!,
moving
o>
axes.
3)
be the components of angular velocity of the body, also

Shew that the equations of motion can be written in
the form
tfiJ^-^W+e^X,
(VL\-fi^J+eM=
dt \c
dT - 7
dT
^
OV
+ u3
s
dT
where
(JT, F, Z, Z, J/, ^) are the

reference to the moving axes.
89.
Application
We
shall
now
dT OVi
dT
5
00)3
00)2
.
0!
dT
^
Oo>3
dT
dT =
JT-
Oo>i
dT
components and moments of the external
to special
forces with
non-holonomic problems.
consider some examples illustrative of the theory of non-
holonomic systems.
1.
Sphere rolling on a fixed sphere.
be required to determine the motion of a perfectly rough sphere of radius a and
which rolls on a fixed sphere of radius 6, the only external force being gravity.
Example
Let
mass
TO
it
218
Non-Jwlonomic Systems.
Dissipative Systems [CH.
vm
Let (b, 0,
be the polar coordinates of the point of contact, referred to the centre of
the fixed sphere, the polar axis being vertical. We take moving axes GABC, where G is
the centre of the moving sphere, GC is the prolongation of the line joining the centres of
<)
the spheres, GA is horizontal and perpendicular to GC, and

and GC, in the direction of 6 increasing.
With these axes we have,
0-2=
F,
GA
perpendicular to
si"
-<j>
03
0,
cos
0,
{
if
is
in the notation of the last article,
0i= -P,
and
and
GB
denote the components of the force at the point of contact parallel to
GB respectively, we
Y=mgsm0 + F
N=0.
M=-Fa,
X=F,
L = F a,
The equations
GA
have
,
become therefore
of motion of the last article
m (u
"*"
F=
=F =
fam
w3 +
(o>j
a>2
(o>
in (v-{-u0s)
mgsin
C/ 2
0)3
v0 3 ) =
O)l ~|~
U I O) 2
gam
i),
+ 0201s),
w.
Moreover, the components parallel to the axes GA, GB of the velocity of the point of
contact are u
and consequently the kinematical equations which express
2 and v -f ao) t
the condition of no sliding at the point of contact are
ao>
Eliminating F,
a^,
o>
2,
we have
-t70 3
|
<
v+u0 3
a02(>)
J*-
0)
The
= n,
where n is a constant; while substituting

two equations their values in terms of 0, 6, 0, we have
last
in the first
equation gives
a>
(a
(a + b]
The former
by
sin 0,
sin 0) + (a + 6)
+ b}-r
dt (0
the
anf>
= 0,
+ %an<f>sin0-%gsiu = 0.
0-(a + b)<p?cos0sin
a constant.
equation by
(/>
+ f cm cos 0=k,
and
Moreover, multiplying the second equation throughout by
and adding, we obtain an equation which can be at once
sin 0,
integrated, giving
n2 ^ +
where h
1}
and gives
is
first
cos
0<f>
for u, v,
of these equations can be integrated at once after multiplying throughout

2
(a + 6)0 sin $
where k
-f0i3=0,
g sin $ = 0,
= 0.
3
is
a constant
Eliminating
(a +
2
fe)
-2- cos
this is really the equation of
= h,
energy of the system.
between these two integral equations, we have
sin 2
&=
and on writing cos# = #,
- (k - f an cos
this equation
2
(9)
- ^-g
becomes
(a
+ b) sin
cos
+ h (a + 6) 2 sin 2
89]
219
Dissipative Systems
The cubic polynomial in x on the right-hand side of this equation is positive when
some real values of 6, i.e. for some values of x
negative when #=1, positive for
between -1 and 1, and negative when #=-1; it has therefore one root greater than
- 1 we shall denote these roots by
unity, and two roots between 1 and
x= + oc
cosh
where cos
j3>
cos a;
cos
/3,
a,
and we then have
\7 a + b)
where
cos
y,
+f =
)
- cosh
j {(x
y) (x
- cos
|8)
(x
cos a)}
e is
Writing
2
14 a+b
7A(a + 6) + â
* +
+ i(coshy + cos/3 + cosa)= 3
+
.
v(a
6)
the equation becomes
2=
or
where the function
is
e2
~14(a + 6){
these quantities e^ e2 e3 are

,
all real,
C S/3
"
30<7(a+i)
and
satisfy the relations
ei + 62 + 63=0,
el
>e
>e
cos a and cos 3

(since x is real) lies between
hence the imaginary part of the constant e in the
we shall
argument of the ^-function is the half-period corresponding to the root e3 which
denote by
the real part of e may be taken to be zero by suitably choosing the origin
of time and therefore we have finally
Now
a*
so z is real
is real
and
for real values of
lies
between
ez
and
and
e3

this integration can be effected by a procedure similar to that used ( 72) to obtain the
Eulerian angles which define the position of a top spinning on a perfectly rough plane.
Example 2. A rough sphere rolls in contact with the outside of a fixed rough sphere,
under gravity if z2 z3 be the greatest and least heights of its centre, during the motion,
and z be the height at a time t from an instant when z was equal to 2 2 prove that
;
(*2
where
el
e2 , e3
(= -
el
- e2 )
z}
[|f>
~
(0 e 2] = (*2 23)
(<?i
- e 2 ),
are real quantities in descending order of magnitude.

(Coll.
Exam.)
220
Example
vm
Sphere rolling on a moving sphere.
3.
now the motion of a rough sphere of radius a and mass m which rolls under
on
another
sphere, of radius b and mass M, the latter sphere being free to turn
gravity
about its centre 0, which is fixed.
Consider
Let (6, 0) be the polar coordinates of the point of contact referred to axes fixed in space
with the fixed centre as origin, the axis from which 6 is measured being vertical.
To obtain the equations of motion of the sphei e m, we take (as in the last example)
moving axes O ABC of which GC is the prolongation of the line OGf joining the centres of
the spheres, and GA is horizontal. Let (0
$2
3 ) denote the components of angular
velocity of the coordinate-system resolved along its own axes, and let (wj, 2 ^3) denote the
components of angular velocity of the sphere m along the same axes. Then, as in the last
example, we have
1
o>
T= \
and
if
F,
parallel to
111
U Z + V- + W2 +
((Bj
(B 2
2
0>
3 )
be the components of the force acting on the sphere
OA
and
GB
respectively,
at the point of contact
we have
L^F a,
so the equations of motion
M=-Fa,
N=Q,
become
a) 3 -
<9
an
+ 0j 0)2 =
(1)3
+ ^3 Wi) .................. (1),
(02
+ 62013) .................. (2),
...................................................... (3).
To determine the motion of the sphere M, we take moving axes parallel to the axes
(JABC, but with their origin at 0; let (Q 1? Q 2 Q 3 ) denote the components of angular
we have
Then for the sphere
velocity of the sphere resolved along these axes.
,
and
its
3)
Q 3 -(9 2 Qi +
The conditions
<9iQ
=F
=0
.............................. (5),
................................. (6).
of no sliding at the point of contact are

u-aa>2
= bQ 2
v+
aa>
= -bQ
.............................. (7).
In order to solve this set of equations we multiply equations

respectively, and add
thus, using
(7),
(3)
and
(6)
by a and
we have
a w 3 + bQ s + u&i + v6% = 0,
or
ai>
Integrating,
-f
&Q 3 = 0.
we have
a<a
+ bQ 3 =an,
where n
is
a constant.
89, 90]
Moreover, from equations
(4)
and
- f M(u Eliminating
ând
o>
#30>i
(7)
ao>
we have
- b6iQ 3 2
..
-v
(
(7),
Now
ud 3 + 0.0%
a&i
,.
(9-^
-63 v) =
Similarly from equations (5) and
Eliminating
- 6z
3v
between this and equations

(u
221
Diasipative Systems
this
<w
+ b6 2 Q 3 = F
and equations
we have
(2),
L
= and + b(M+ni)
9 sin 6,
2j/
.
M0 3 )
2M
an
d
(M+m}qsin
*
rr= ----r.r -r
_
the equations (A) and (B), from which 6 and
.4>sme
......... (B).
are to be determined in terms of
t,
are of essentially the same character as the equations found for the determination of 6
in the previous example the former equations being in fact derivable from the
and
:
present ones by making

very large compared with m.
as
in
the
former
case.
proceeds exactly
The
integration
therefore
4.
A uniform sphere rolls on a perfectly rough horizontal plane, under
whose resultant passes through its centre. Shew that the motion of its centre
the same as that of a particle acted on by the same forces reduced in the ratio 5 7.
Example
forces
is
Example 5. Form the equations of motion of a perfectly rough sphere rolling under
gravity inside a fixed right circular cylinder, the axis of which is inclined to the vertical at
an angle a; and shew that, if the sphere be such that &2 =â 2 a being its radius and k
,
the radius of gyration about any diameter, and if it be placed at rest with the axial plane
through its centre making an angle /3 with the vertical axial plane, the velocity of
the centre parallel to the axis,
{sin
where
+a
is
\6 arccosh
this angle is 6, is
(cos
6 sec
i/3)
the radius of the cylinder.
For other examples
cf.
+ cos i(9 arccos
(sin
\d cosec A/3)},
I,
1895.)
Woronetz, Math. Ann. LXX. (1911), p. 410.
Vibrations of non-holonomic systems.
90.
We
system
when
shall
:
it
next consider the small vibratory motions of a non-holonomic

appear that so far as vibrations about equilibrium are con
will
cerned, the difference between holonomic
and non-holonomic systems
is
unimportant.
For consider the vibrations about equilibrium of a non-holonomic system

with n independent coordinates and (n
m) degrees of freedom, in which
the constraints are independent of the time. Let T be the kinetic and V the
T will be supposed to be
potential energy, so that for the vibrational problem
a homogeneous quadratic function of (q lt q,
...,
q n ), and
F to be a homogeneous
Dissipatlve Systems [CH. vui
222
quadratic function of (q l
There are
constants.
m
A
...,q n ], the coefficients in both cases being

equations of the type
q.2
lk q!
+ A nk q n =
+ Aaq* +
which express the non-holonomic constraints

are
= 1,2,...,
m),
and the equations of motion
87)
dV
= -^-+\jA
n + \ A r2 +
d /?T\
}
at (^\cq r j
-r.
+ \ m A rm
...
(r=l,
2,
...,>
oq r
these equations it is evident that \ l} X 2 ..., \m are in general small

quantities of the order of the coordinates and therefore for the vibrational
From
problem only the constant parts of ^. n A 12 ..., A nm need be considered. The

vibrational motion is therefore the same as if the coefficients A n A 12 ..., A nm
were constants independent of the coordinates but in this case the equations
,
A lk q + A sk q + ... + A nk q n =
2
can be integrated
(k
I, 2,
(A
1, 2,
...,i)
in fact, they give
lk
ql
+ A^q, +... + A nk q n =
m),
the constants of integration being zero since the values

qi
= 0,
q2
0,
qn
represent a possible position of the system.
motion of the given non-holonomic system is

the same as that of the holonomic system for which the equations of con
straint are expressible in the integrated form
It follows that the vibratory
A
we can
nate
lk
ql
+ A 2k q. +...+A nk qn =
2
(k
1, 2, ....
m)
therefore determine the vibrations by using these equations to elimi

of the coordinates (q l} q z ...,q n ) from T and V; we shall then have
,
and potential
and
the corre
m) coordinates
energies being expressed in terms of (n
the
sponding velocities the vibrations of this system can be determined by
a holonomic system with (n
m) degrees
of freedom, the kinetic
usual method described in the preceding chapter.
As an example, we
shall consider the following problem*.
horizontal
heavy homogeneous hemisphere is resting in equilibrium on a perfectly rough
second heavy homogeneous hemisphere is
A
itith its spherical surface downwards.
plane
a perfectly rough plane face of the first, the point of contact
resting in the same way on
The equilibrium being slightly disturbed, it is required
being in the centre of the face.
to
find the vibrations of the system.
Take as axes
of reference
its
rectangular set of axes

(1)
centre of gravity 2
Due
to
Z>>xyz
fixed in the
upper hemisphere, the origin being
Madame Kerkhoven-Wythoff, Nieuw Archiefvoor
Wiskunde, Deel
iv.
(1899).
90]
its
Z ^^
(2)
rectangular set of axes
centre of gravity Z\
223
Dissipative Systems
fixed in the lower hemisphere, the origin being
A rectangular set of -axes Rlmn fixed in space, the origin R being the equi
(3)
librium position of the point of contact of the lower hemisphere and the plane.
We further
Z<iZ,
define these axes
Z\, and Rn
by supposing that
Zz y,
are parallel, the axes
Suppose that at time

of axes are connected
in the equilibrium position the axes

therefore coincident, while the axes
z x, Z^, Rl
being therefore also parallel.
are vertical and
Z^, and
Rm
the coordinates of a point referred to these different sets
by the equations
= a+ n
The 24 coefficients in these transformation-formulae completely specify the

position of
the system at any instant. As however the system has
only six degrees of freedom, there
must be 18 equations connecting these coefficients or their differentials. Of these, 12 are
the ordinary conditions of the types
which express the orthogonal character of the axes

contact and rolling, which we shall now find.
Let
R R2
l
the remaining 6 are the conditions of
be the radii of the lower and upper hemispheres respectively, and
the distances of the centres of gravity from their plane faces, so li = %Ri, ^2 = f #2coordinates of the point of contact of the upper hemisphere with the lower are
x 2 =-li z yi,
yi= -
Rtf*,
h-R-iyz
zz
/j,
12
The
the conditions that this point shall be at rest relative to the lower hemisphere are
= 0.
The
last of these
equations gives 7 + ^273 = 0, which
the differentiated form of the
is
equation
^-7-73^2= -RZ,
an equation which expresses the condition of contact of the two hemispheres
the first two of the equations give
a-
at
R 2 y^ - a
0!
R ayi - & RW +
and these express the condition of

equations give as a
first
R^y-i
a3
/3 3 (1 2
rolling of the
approximation
(^ 2
R-zy^
while
= 0,
- -R 2 y3 = 0,
,
upper on the lower hemisphere.
These
224
Dissipative Systems
[CH.
vm
and therefore on integration
Similarly the condition of contact of the lower hemisphere and the horizontal plane
and the conditions of
is
rolling are
We
have thus now obtained the 18 equations connecting the 24 coefficients: taking
c i as the 6 independent coordinates of the system, and solving for
y\i a
the other 18 coefficients in terms of these, we find with the necessary approximation
a 2)
/3s?
2>
^3>
-l - $ (a 22 + 713
),
=-yi-
=1-
.-/sip,
U=i-i(yi
The
+33 2
).
is
F= J/j^c + M2 g (c + c ia + c2 /3 + c3 y),
or,
retaining only small quantities of the second order,
If
express the coordinates I, m, n of any particle of the

in terms of its coordinates relative to the axes 2 xyz and
l
now we
hemisphere
and form the sum
i2m (P + m2 + h2
for each hemisphere, neglecting
upper or lower
respectively,
terms above the second
order of small quantities, and remembering that the principal moments of inertia of
and radius R at its centre of gravity are %MR 2
a hemisphere of mass
R*, we find for the kinetic energy of the system the value T, where
+ C! 2
{^ Wi + l/
The equations
(i)
in F,
is
(1J7? 2
^2 +
+f
of motion evidently separate into three distinct sets, consisting of
Equations
for the coordinates
a 2 and
and do not correspond to vibrations
not disturbed
revolution.
(ii)
(iii)
We
^^^^
a2
these coordinates give rise to no terms

in fact, the equilibrium
in the stricter sense
either of the hemispheres is turned through

can therefore neglect these equations.
if
Equations involving the coordinates
b3
and
any angle about
its axis of
3.
Equations for the coordinates c 1 and y l these are exactly the same as the
so we need consider only the latter.
/33
equations for 6 3 and
90]
The equations
and
for 6 3
j33
225
Dissipative Systems
are, in extenso,
RM ^
- 1^2^2/3
2 J/2 6 3
+g (|
3-|^3 + ^/33=0.
The corresponding determinantal equation
for X,
where
STT/V/X is a period, is
=0.
This
and
is
a quadratic equation in X
this is the condition
The
it is
easily
found that
its
roots are positive
if
for stability of the equilibrium.
vibrations of non-holonomic systems about a state of steady motion

by use of the equations of motion given
are most conveniently discussed

in
88.
The method
Example.
be illustrated by the following example.
will
solid of revolution has an equatorial plane of symmetry, and is rolling

round its axis in steady motion on a perfectly rough horizontal
with angular velocity n
This motion being slightly disturbed,
plane, the equatorial plane of the solid being vertical.

to find the
period of a vibration.
Let G be the centre of gravity of the solid, and let (C, A) be its moments of inertia
about the axis and about a line through G perpendicular to the axis. Take as moving
axes of reference Gxyz, where Gz is the axis of the solid, Gy is perpendicular to the plane
through Gz and the point of contact (so Gy is horizontal), and Gx is normal to the plane
Let F,
R be the components of the force acting on the solid at the point
Gt/z.
F
F being in the plane
,
of contact,
plane.
Let
(6 l
and
#3)
Gxz,
w2
(wj,
being parallel to Gy, and R being normal to the

denote as usual the components of angular
co 3 )
velocity of the axes and of the body respectively, and let (u, v, w) be the components of
the velocity of G, parallel to the moving axes. Further, let p be the radius of curvature
of the meridian of the solid at the equator, a the radius of its
equatorial circle, 6 the
the angle between Gy and its undisturbed
angle made by Gz with the vertical, and
direction.
Then we have
di
and the kinetic energy
=-
=
o>!
4>
sin 6,
$ 2 =co 2 = #,
is
T= $ M (u 2 + v 2 + 708) + 1 A
The equations
of
from this point on the
# 3 = $cos#,
2
(a>!
co 2 )
+\
2
<7o>
18 therefore give, if P is the point of contact, PAT the

perpendicular
and G the perpendicular from G on the horizontal plane,
axis,
M (u - w9
+ w6 2
ai-Ato<i0 3
2
.3
<7co
=Fuo6-(R- Mg) sin 6,
+ Ca s 6 Z =- F GK,
.
6^+Aê-^-F.GN-R. XP,
=F .PK.
In these equations, (rA and j!\Tare measured

positively parallel to the positive direction
of the axis of z and the horizontal projection of this direction
respectively.
w. D.
15
226
The conditions
of no sliding at
P are
GN.
sin 6
o>
sin 6
=
-^
= 0,
=o,
.O!
and the condition of contact of the body and plane
w cos 6 -11
vm
is
- OK cos + PKsin 6).
These equations determine the motion in the general case, when the disturbance from
When this latter assumption is made,
steady motion is not supposed to be small.
we have
where
x,
"&,
*)
equal to Mg.
are small; and F, F u, w, wj,

Moreover we have NP= (p - a) x-
o>
M (u + an6
3)
is nearly
are small, while
therefore become
QI, # 2 , $3
The equations
= - R + Mg,
=F
Mr,
=0,
Cw
w
+ atzr
aa>2
,77
where
o>i
Eliminating F,
=6
=^
=F a,
= 0,
=0,
QZ
W2=#2=x>
R, and replacing 6 1
62
a>
1}
o>
by their
values, the equations
become
=0,
Cw
w
=ax,
=-aw.
From
or
and
77
the third and
are constants.
of these equations we
The other three equations
fifth
and therefore the equation
Dissipative systems
We now
x + {MgA ( P -a) +
shews that the period of a vibration

2?r
91.
give,
for the determination of
A(A + Ma 2
this equation
see that
Cn*
and
are zero, and therefore
on eliminating w,
i-
(C+ Ma
)}
x=
is
2
2
(MgA (p-a) + Cn (C+Ma ))
frictional forces.
proceed to the consideration of systems for which the principle of

conservation of dynamical energy is not valid, the energy of the system being
90, 91]
continually changed into

in dynamics.
We
227
Dissipative Systems
some other form
(e.g.
heat) which
is
not recognised
shall first consider frictional systems.
If two rigid bodies which are not perfectly smooth are in contact, the
reaction between them at the point of contact may be resolved into a com
ponent along the common normal to their surfaces at the point, which is
called the normal pressure, and a
component in the common tangent-plane,
which is called the frictional force. The frictional force is determined by
the following law*, which has been established experimentally
The bodies
will not slide on each other, provided the frictional

force required for the
does
not
exceed
times
the
normal
/j,
prevention of sliding
pressure, where /A
is
a constant called
the
"
limiting coefficient of friction" ivhich depends only

on the material of which the surfaces in contact are composed. If on the
other hand the frictional force required to prevent sliding is
greater than fj,
times the normal pressure, there will be sliding at the point of contact, and
play will be /* times the normal pressure.
the frictional force called into
Painleve has pointed out that the four hypotheses
(1)
that the above laws of friction
hold, (2) that there exist rigid bodies, (3) that the normal pressure between bodies cannot
be negative, (4) that all accelerations and tensions are finite taken together lead in some
cases to contradictions of the fundamental laws of dynamics.

For a discussion on this
cf.
Comptes Rendus, CXL. (1905), pp. 635, 702, 847: ibid. CXLI. (1905), pp. 310, 401,
546; Zeitschrift fur M. u. P. LVIII. (1909), p. 186.
subject,
The
following
examples
illustrate
the
motion
of
systems
involving
frictional forces.
Motion of a particle on a rough fixed plane curve.

1.
Consider the motion of a particle which is constrained to move on a rough fixed tube
of small bore, in the form of a plane curve, under forces which
on its
depend
Example
solely
Let f(s) and g(s) denote the components of force

per unit mass
acting on the particle in direction of the tangent and normal to the tube, where s is the
distance of the particle from some fixed point of the tube, measured
along the arc in the
direction in which the particle is moving and let R be the normal reaction
per unit mass,
position in the tube.
and p the
coefficient of friction.
Since the components of acceleration of the particle along the

tangent and normal are
is the velocity of the particle and
p the radius of curvature of the
2
vdv/ds and v /p, where v
tube,
we have
S-/0-tt
Eliminating R, we have
__ + 2u
dv*
Integrating,
where
(f>
= $ds/p,
JV = 2
we have
and
c is
a constant depending on the
initial
The discovery that the friction is proportional to the normal pressure was
G. Amontons, Paris Mem., annee
1699, p. 206.
made by
152
228
The right-hand
we have
side of this equation is a

known
function of
s,
say =F(s).
vm
Then
v2
and
so the relation between s
is
t-t
t
/
ds.
{F(s)}~2
This equation represents the solxition of the problem.
Example
mass
2.
m is attached
circular hoop of mass

stands on rough ground, and a particle of
end of the horizontal diameter. To find whether the hoop will roll
to the
or slide.
Let us investigate the rolling motion, assumed possible, and so determine whether the
produce this motion is, or is not, greater than the maximum friction
friction required to
i.e. p. times the corresponding normal pressure.

Let 6 be the angle
turned through by the hoop from the commencement of the motion, and let x and y be
actually available,
the coordinates of the centre of gravity of the system, referred to horizontal and vertical
its own initial position, so that
(downward) axes through
ma
x = a6a
where a
The
a\
ti
y=
9
cos#),
f
(1
M+m^
kinetic
6.
and potential energies are
For the
so initially
initial
[2
of motion
is
- sin
0),
therefore
M+ m(l- sin 0)}] + ma ft cos

2
2
-T-
Cvt
= mga cos 0.
motion, this equation gives
we have
_
if
,,
the radius of the hoop.
is
T= Ma2 ft + ma?ft
V= mga sin 6.
But
ma
M+m sin
/J
ma
nf
_
~
ft
F be the frictioual force and R the normal pressure,

F=(X+m),
we have
R=(M+m)(
we have
so initially
The hoop
-y+g
m(M+m)
2
will therefore roll or slide according as the coefficient of friction is greater or
than
less
m (M+m)
Example 3. A particle moves under gravity on a rough cycloid whose plane is
and whose base is horizontal if
be the inclination of the tangent at any point
vertical
to the horizontal, so that the equation of the cycloid can be written

s
and
if
tan
be the coefficient of
tar
ce
where
c is
a constant.
friction,
= 4a sin
<,
shew that the motion
is
91, 92]
92.
229
Dissipative Systems
Resisting forces depending on the velocity.
different type of dissipative system is illustrated by the motion of

a projectile in the air, as the resistance of the air depends on the velocity of
No general rule can be formulated for the solution of prob

lems involving forces of this kind a case of practical interest, however,
namely the motion of a projectile under the influence of gravity arid of a
resistance varying as some power of the projectile s velocity, can be integrated
the projectile.
in the following manner.
For low velocities (below lOOft./sec.) the resistance of the air to a projectile is nearly
For high velocities (say 2000 ft. /sec.) the
proportional to the square of the velocity.
resistance is approximately a linear function of the velocity.
At time
be the velocity of the
let v
n
projectile, kv the resistance per
unit mass, 6 the inclination of the path to the horizontal, and p the radius of
The components of acceleration of the projectile
curvature of the path.
along the tangent and normal to
its
path are vdv/ds and
v-/p
and hence the

(vdv/ds
2
v /p
Dividing the
dv
d
dv
Integrating,
(1 \
-^
n
\v j
tan 6
d0
-T7j
kv n
gcos0
ti"~
+ -=
vn
we obtain
equation by the second,
first
v"+~
or
= g sin 6
= g cos 6.
sec 0)
do (n log
nk
sec
0.
we have
(I/O sec
Constant
This equation gives v in terms of
= - (nk/g)
fsec n+1 0d0.
To obtain
0.
t,
the equation
= pg cos
gives
gt
Iv sec0d0,
v is a known function of 0, this equation gives t as a function of 0.

The rectangular coordinates (x, y) of the particle can now be found from the
and as
equations
x=
The
solution of the
problem
Resisting forces proportional
Book
n.
Ivcos0dt,
is
thus reduced to quadratures.
z
and av + bv z were considered by
respectively to v, v
3.
The
case
of
a
resistance
2,
proportional to any power
,
Newton, Principia,
1,
of the velocity was then examined by John Bernoulli*
*
Opera,
i.
p. 502.
in 1711.
230
D Alembert *
shewed that
projectile, the integration
if
Dissipative Systems
gu denotes the
ratio of the resistance to the
[CH.
vm
mass of the
can be effected in the four cases
u = a + bv n
u = a + b log v,
u=a
(log v}
+ R log v + b,
a, b, n are arbitrary constants and R is another constant depending on them.

Siaccit obtained many more integrable cases, of which the following maybe mentioned
where
log
8
vdu
^c
^
where
a,
number
c,
b,
l+a(u-\}
are arbitrary constants
of terms involved being finite
du
du
-r- + C,
-icr j.
l+b(u + l) c
this equation defines

c is rational.
terms of
in
u,
the
when
Poisson pointed out in 1806J that the theory of singular solutions of differential
equations has applications in Dynamics, notably in the case of a particle under a resisting
If a particle is moving in a straight line under a resisting force varying as the
force.
square root of the velocity, the equation of motion

dv/dt =
The
initial velocity
being
c2,
the motion
v
so long as
t<2c/a,
after
which
it is
is
is
av 2.
represented by the general integral
= (c
ât)
represented by the singular solution
= 0.
Example 1. A heavy particle falls vertically from rest at the origin in a medium
whose resistance varies directly as the velocity. Shew that the distance traversed
in time
is
ff,9
-- --
9*
PHI?
where pv
is
e -*t,
K
the resistance per unit mass.
Example 2. A heavy particle falls vertically from rest at the origin in a medium
whose resistance varies as the square of the velocity shew that the distance traversed in
time t is
:
- log cosh
where
p.v
93.
(Jgii. t),
denotes the resistance per unit mass.
Rayleigtis dissipation-function.
When
a system is subject to external resisting forces which are directly

proportional to the velocities of their points of application, it is possible to
express the equations of motion of the system in general coordinates in terms
and potential energies and of a single new function.
of the kinetic
For
which
the energy lost to the system by the action of the resisting force
applied to a particle ra of the system, whose coordinates are (x, y, z),
let
is
in an arbitrary displacement
(Sac,
By, Bz)
be
kz z8z,
du mouvement des fluides,

t Comptes Rendus, cxxxn. (1901), p. 1175.
I Journal de VEcole Polyt. vi. (Cahier 13), p. 60.
Traite de Vequilibre
et
Paris, 1744.
92, 93]
where kx kv kz are functions of
x, y,
typical particle ra will therefore
be
231
Dissipative Systems
The equations
z only.
of motion of the
(mx =
<my
\mz
where X, Y,
Z are
kx x + X,
= -k,J y+Y,
= kz z + Z,
the components of the total force (external and

molecular)
on the particle, except the force of resistance.
Now
let
a function
F be
defined by the equations
where the summation
is extended over all the

particles of the system; so
called the dissipation-function, represents half the rate at
energy is being lost to the system by the action of the resisting forces
that F, which
which
and let
is
(q lt
q.2
...,
q n ) be coordinates specifying the configuration of the
system.
Multiplying the equations of motion of the particle ra by dx/dqr dy/dq r

and summing for all the particles of the
system, we have
,
dz/dqr respectively,
,
dx
dz\
dy
+ y ^+z=
..
..
>
^
=-S
dqj
dq r
dq r
/,
dy
dz\
--h kyy
~
v y ^- T k,z
dx
(kx x
dq r
-f
dqj
dq r
-^
oqr
As
in
26,
dx
;^
dqr
T is
-.
s
dq r j
we have
.
where
dqr
..
dy
.,dz\
-
oq
dqj
*
+y^-+^r
r
the kinetic energy
/dT\
-
dt \dqr j
dT
-^
dq r
and
--
oqr
Q^
=d
--
"
^,
dq rj
dq r
r>
where
+ Q2 8q2 ... Q n Sq n denotes the work done by the external
forces (excluding the
resistances) in an arbitrary infinitesimal displacement
while
we have
&+M
oq r
_dF
dqr
It follows that the
equations
ordinates (q lt q2
...,
of motion of the system in terms of the co
q n ) can be written in the
dT\
dT
dF
form
232
Example.
depend on the
If the resisting forces
relative (as
opposed to the absolute)
so that the forces acting on

velocity of their points of application,
and
(# 2
>
g z)
3/2?
vm
two
particles (x^
zt)
have the components
-kx (xi-x&
-k v (yi-fa\
-*,(*i
-**(#2-*l)>
-*(&-&)
-*(*2-*l)
-z)
and
the equations in general coordinates can be formed with the
respectively, shew that,
expression
\ 2 {kx (x,
- xtf + k y (ft - ytf + *, fr - z,) 2 }
as a dissipation-function.
Vibrations of dissipative systems.
94.
is specified by its kinetic energy function, potential

methods similar to those of
and
dissipation function,
energy function,
the nature of the small
determine
to
order
in
Chapter VII can be applied
an
vibrations of the system about
equilibrium-configuration.
If a dynamical system
system with two degrees of freedom.

As in 76, we find that for the vibrational problem the kinetic energy and
can be taken as homogeneous quadratic functions of the
dissipation function
For simplicity we
shall consider a
and the potential energy as a homogeneous quadratic function of

the coordinates, the coefficients in these functions being constants. Taking
as coordinates those variables which would be normal coordinates if there
function, we can write these three functions in the form
were no
velocities,
dissipation
where
stable if
and X2 will be supposed

there were no dissipative
The equations
drtT\dT + W + d_V =
or
q\
</
we attempt
aq
dq r
l
A<jl
dqr
(r=l,2),
9<?-
+ hq^ + \ q =0,
+ &?2 + ^2?2 = 0.
l
to find a particular solution of these equations in the form
on substituting these values in the differential equations we have
A ( p* +ap + XO + Bhp = 0,
Ahp + B(p*+bp +
from which
it
would be
of motion are
dt \dqj
If
positive, so that the equilibrium

forces.
=
X>)
0,
p must be a root of the equation

+ ap + \0 (p + bp + X 2 ) - A ? = 0.
follows that
2
(p
93, 94]
We
233
Dissipative Systems
so that
suppose the dissipative forces to be comparatively small,
on this supposition, the
b can be neglected
squares of the quantities a, h,
roots of the last equation are readily found to be
shall
the second of the equations

Corresponding to the root p^ we have, from
connecting
and B,
B _
ih
VXj
of the differential equations

particular solution
therefore given by
is
= (Xj - X ) e ^ at (cos N/X] + i sin x/Xjtf),

= h*J\ e~^ at (icos VXxi - sin Vx^),
fa
^
i in these
and a second particular solution is obtained by changing i to
of the
solutions
It follows that two independent real particular
expressions.
differential equations are

r
= (Xj - X
2)
e~
^ at
cos VXjtf
and therefore the most general

(q t
f (?!
and
-{
= (Xj - X
=
~
e
2)
^ at sin
t al nns
Vx^,
v"X .
real solution involving e Pit is
= (Xj - X ) Ae ~ $ at sin (\/M + e),

2
sn
where -4 and e are real arbitrary constants. This represents one of the normal
modes of vibration of the system. Adding to this the corresponding solution
in e p **, we have finally the general solution of the vibrational problem, namely
1
= (\ - X ) Ae
at
sin
sin
(\/M +
Vx^ +
-4,
initial
Now we
<;,
+ A x 5e~
a
sin
+ (X2 - X ) Be~
:
bt
sin (*/\2 t
+ 7),
7 are four constants which must be determined from the
where
-B,
e)
con
suppose the dissipative forces such that energy is being
to the system, so that F is a positive definite form, and therefore
tinually lost
a and 6 are positive.
The
last
equations therefore shew that the vibration

at
and e
gradually dies away, on account of the presence of the factors e~^
of
of
u.
the
normal
are
vibrations
the periods
h, 6) the
(neglecting squares
as if the dissipative forces were absent and in a normal vibration, the

amplitude of oscillation of one of the coordinates is small compared with the
same
amplitude of oscillation of the other coordinate, while the phases of the

vibration in the two coordinates at any instant differ by a quarter-period.
Non-1lolonomic Systems.
234
vm
similar analysis leads to corresponding results for systems with more

than two degrees of freedom supposing that the dissipative forces are small
and that the dissipation function and potential energy are positive definite
forms, we find that the periods of the normal vibrations are (neglecting
unaltered by the
squares of the coefficients in the dissipation function)
;
dies away
presence of the dissipative forces, but that the vibration gradually
if (q lt q 2 ..., q n } are the normal coordinates of the system when the
forces are absent, there is a normal vibration of the system when
:
and
dissipative
the dissipative forces are present, in which the amplitude of the vibrations in
of the vibration in q 1}
q-s,
q n is small compared with the amplitude
,
<?2,
and the phase of the vibrations in q2 q 3

,
...,
q n differs
by a quarter-period
from the phase of the vibration in q lf

Example. Discuss the vibrations of a system which is acted on by periodic external
which have the same period as one of the normal modes of free vibration of the
system shewing the importance of dissipative forces (even where small) in this case.
forces
95.
Impact.
Another mode in which energy may be lost* to a dynamical system is by

a collision generally
the collision of bodies which belong to the system
;
results in a decrease of dynamical energy.
The
mental
is based on the following experi

analytical discussion of collisions
When two bodies collide, the values of the relative velocity of the
law-f-.
surfaces in contact (estimated normally
to the
surfaces) at instants immediately
impact bear a definite ratio to each other

before
this ratio depends only on the material of which the bodies are composed.
and immediately
after the
This ratio will in general be denoted by

are said to be inelastic.
When
e.
e is zero,
the bodies
The general problem of impact reduces therefore to a problem in impulsive

motion in which the unknown impulsive force at the point of contact of the
bodies is to be determined by the condition that the change in relative
normal velocity of the bodies satisfies the above law.
96.
We
Loss of kinetic energy in impact.

shall
now
when two
find the loss of kinetic energy
perfectly smooth
bodies impinge on each other.
and let (u v w ) and

typify the mass of a particle of either body,
after the impact, and
and
w) denote its components of velocity before
Let
(u, v,
*
I.e. lost to
the system considered as a dynamical system
the energy
is
not annihilated, but
appears in some other manifestation, e.g. heat.

t The laws of impact were discovered in 1668 by John Wallis (Phil. Trans. No. 43,
and Christopher Wren
(ibid. p. 867).
p. 864)
94-97]
let
W)
U, V,
235
Dissipative Systems
be the components of the total impulsive force (external and

The equations of impulsive motion ( 35) give
molecular) on this particle.
m (u
ti
m(v - v ) =V,
U,
m (w - w =
W.
Multiplying these equations by (u + eu 9 ), (v + ev Q ), (w + ewQ ) respectively,

summing for all the particles of both bodies, we have
adding, and
2w {(u -
-MO)
(u
eu
+ (v
+ ev ) + (w - w } (w + ew )}
= ${U (u + eu + V(v + ev ) -f
)
(v
Now
so far as molecular impulses are concerned,
2 ( Uu + Vv + Ww) =
0,
and
W (w + ew
)}.
we have
Ww = 0,
2 ( Uu + Vv +
since the impulsive forces which correspond to each other in virtue of the law
of Action and Reaction will
give contributions to these sums which mutually
destroy each other.

Also, since the part of (u + eu ) due to the normal component of velocity
has the same value for each of the particles in contact at the
point where
the impact takes place (in virtue of the law of

impact) it follows that
the impulsive force between the bodies does not contribute to the sum
%U(u + eu,), and

+ ew
and 2TF (w
We
similarly does not contribute to
sums ^V(v+ev
the
).
have therefore
2{U(u + eu ) + V(v + ev ) +
W (w + ew
)}
0,
and consequently
2m {(u - M
(u
eu
+ (v- v ) (v + ev ) +
(w
-w)
(w
+ ew
)}
0,
or
This equation can be expressed by the statement that the kinetic

energy
impact is (1
e)/(l + e) times the kinetic energy of that motion
which would have to be compounded with the motion at the instant
before the
lost in the
impact in order
97.
to
produce the motion at the instant after the impact.
Examples of impact.
The impulsive change

rigid
bodies
in
of motion consequent on the collision of two free

can
be most simply determined by the
space
following
considerations.
The motion
of each
body before or
after
impact
is
specified
by
six
quantities (e.g. the three components of velocity of its centre of gravity and
the three components of
angular velocity of the body about axes through its
centre of gravity).
The total number of equations required to determine the
236
Dissipative Systems [OH.
vm
impulsive change of motion is therefore twelve. Of these, six are immediately

furnished by the condition that the angular momentum of each body about
any axis through the point of contact
is unchanged (since the

impulsive forces
act at this point) another equation is obtained from the condition that the
momentum of the system in the direction normal to the surfaces in contact
;
(since the normal impulsive forces on the two bodies at the

of
contact
are equal and opposite), and another by the experimental
point
law of impact.
If the bodies are perfectly smooth, the remaining four
is
unchanged
equations can be derived from the condition that the linear momentum of
each body in any direction tangential to the surfaces in contact is unchanged
is no
tangential impulse if the bodies are smooth) if on the
other hand the bodies are perfectly or imperfectly rough, the condition that
the linear momentum of the system in any direction tangential to the
(since there
is
unchanged gives two equations if the bodies are
condition
the
that the relative velocity of the bodies in
perfectly rough,
direction
after the impact is zero gives the other two while
any tangential
contact
surfaces in
the bodies are imperfectly rough, the coefficient of friction between the
surfaces in contact being y-t, the remaining two equations are given by the
if
conditions that
the relative velocity in any tangential direction is zero after the

impact, provided the tangential component of the impulse required for this
does not exceed //, times the normal component of the impulse
(a)
if
(/3)
equal to
In
//,
the last condition
is
not
satisfied, there is
a tangential impulse
times the normal impulse between the bodies.
all cases,
therefore, the required twelve equations can be found.
If the motion takes place in a plane, or if one of the bodies

procedure is still valid after making some obvious modifications.
The
following examples illustrate these principles
Example
1.
inelastic inclined
An
inelastic sphere of
plane of mass
M and
m falls
mass
angle
a,
is fixed,
V on a perfectly rough
on a smooth horizontal plane.
with velocity
which
rests
Sheiv that the vertical velocity of the centre of the sphere immediately after the impact
2
Fsin a
r-=-
Let
U be the
Exam.)
velocity of the plane after impact, u the velocity of the sphere parallel to
the angular velocity of the sphere, arid a its radius.
relative to the plane,
The equation
is
,.
(Coll.
omsin*a
and
this
<B
of horizontal
momentum
gives
TO (u cos a
The kinematical condition
U~)
= MU.
at the point of contact is
acau.
condition that the angular momentum of the sphere about the point of contact
shall be the same before and after impact is
The
m Va sin a = g
ma-o>
+ ma (u
7 cos a).
Non-liolonomic Systems.
97]
These three equations
give,
on eliminating
5
~
is
eo
and
U,
Fsin 2 a
(M+m)
+ -2m + 5m sin* a
~1M
which
237
Dissipative Systems
the result stated.
Example 2. A sphere of radius a rotating with angular velocity Q, about an axis

an angle a to the vertical and moving, in the vertical plane containing that axis,
with velocity V in a direction making an
angle a with the horizon, strikes a perfectly rough
inclined at
horizontal plane.
If the plane be tangentialltj inelastic, find the angle which the vertical
plane containing the new direction of motion makes with the old.
Take rectangular axes Oxyz, where

is the point of contact, Oz is
vertical, and yOz is
the initial plane of motion; and let
and 2 be the components of angular velocity
about Ox and Oy respectively after the impact, and
the mass of the sphere.
o>
o>!
Equating the
and
initial
final
angular
momenta about
Ma Fcos a = Ma
|-
Equating the
and
initial
final
angular
Ox,
we have
Oy,
we have
2
.
a>i
momenta about
= I J/a 2 co 2
| Ma Q sin a
2
The tangent
new plane
of the inclination of the
account of the perfect roughness of the plane)
0)2/0)!,
of motion to the plane yOz

and this is therefore equal to
is
(on
\Ma?Q. sin a
Ma V cos a
or
f a (Q/ F) tan
A perfectly rough circular

and length 2a capable of
a.
mass
and radius c impinges upon
turning freely about a pivot at its centre. If
the point of impact is distant b
from the centre of the rod, and the direction of motion of the
centre of the disc makes angles a, ft with the rod
before and after collision, shew that
Example
3.
a rod of mass
Let
F denote
its final
disc of
2
2
2
2
(3Mb + ma ) tan/3 = 3 (ema -3Mb tan a.
(Coll. %Exam.)
the initial velocity of the disc, and let v denote
its final
velocity
and Q
angular velocity.
Since there
no sliding at the point of contact, we have
is
v cos
Denoting by co the final angular velocity of the rod, and by / the normal impulse
between the rod and disc, the equation of the motion of the rod is
Ib =
The equation
\ma
2
u>.
of impulsive motion of the disc in the direction
normal to the rod
is
M(vsin(i+ ^ sin a) = 7,
and the law of impact gives the
relation
v sin
Equating the
initial
and
final
ft
angular
6o>
= e V sin a.
momenta
of the disc about the point of contact,
we have
Fcos a = v cos
Eliminating
v,
Q,
/,
o>
is
\ cO.
from these equations, we have
2 tan
which
ft
the result stated.
ft
(3Mb
+ ma2 = 3 tan a (mea 2 )
238
Dissipative Systems
[CH.
Example 4. A circular hoop, in motion without rotation in its own plane, impinges on a
rough fixed straight-edged obstacle in the plane. The velocity of the centre of the hoop
an angle a with the edge, and the coefficient
before impact is V, in a direction making
of friction
is
p..
To find
the impulsive
change of motion.
Let u and v denote the components of velocity of the centre of the hoop after the
the
impact, parallel and perpendicular to the edge, and let co be the angular velocity,
mass and a the radius
of the hoop.
Equating the angular momenta about the point of contact before and after the impact,
we have
-
Ma-<a
+ Mau = M Va cos a.
The law of impact gives the equation

is zero after the impact, provided the frictional
Since the plane is rough, u +
impulse required for this does not exceed /x times the normal impulse but if this
condition is not satisfied, the frictional impulse is p. times the normal impulse.
a<o
Let
F be the frictional and R the normal impulse
We have
and
if
u+
aa>
is zero,
we
The quantity u + aw
and
if
shall
-ft,
Ma?v=-aF.
have
will therefore be zero after the impact,
provided
7sino.
F=pM(l+e)
if
then we have
does not satisfy this inequality, we shall have
p.
Thus
while
M(u- Fcosa)= -F, M(v+ 7 sin a) =

R = M(\ + e) 7sin a,
therefore
finally, if
p.
<
/x^cota/2(l+e), the motion
cot a/2 (1
+ e),
the motion
is
is
determined by the equations
determined by the equations
7 sin a.
1.
which
A
is
of radius a is made to rotate about a vertical diameter,

perfectly rough sphere
uniform sphere of radius b is
with a constant angular velocity n.
fixed,
distant aa from the highest point: investigate the motion

placed on it at a point
and determine in any position the angular velocity of the sphere. Shew that the sphere
from
will leave the rotating sphere when the point of contact is at an angular distance 6
the vertex, where

COS
10
=^
17
^119
- -
a 2 w 2 sin 2 a
JT
(a+o)g
I,
1889.)
on the surface of a cone of revolution

2.
rough sphere of radius a rolls under gravity
with uniform angular velocity n,
axis
its
vertical
about
which is compelled to turn
if a be the semi-vertical angle of the cone, r sin a be the
vertex
its
being uppermost
distance of the centre of the sphere from the axis of the cone,
be the angle turned
vm]
239
Dissipative Systems
through, relatively to the cone, by the vertical plane containing the centre of the sphere,
3 be the rate of rotation of the sphere about the common normal,
prove that
and
o>
14 cos a
(1^
where A, B,
3.
Shew
that
6,
solid
of revolution
I,
of.
1897.)
mass
o>,
Mac
~ (a cos
A (C+Mc 2 ) - M*a?#}
with a plane circular base of

without slipping with its edge in contact with a
rough horizontal plane.
O are determined by the equations
homogeneous
radius c rolls
r-=A,
Qn}
are determinate constants.
(7
6)
- Mc*& cos
~ (a cos
= (C+ J/c2
+ C (C+Mc2
6)
o,
+ Me & + AQ? cos 2 6 - ZMacvQ. cos 6 + C+ Me 2

2
cos 6
cos 6
- Mac Ca cos 2 = 0,
= Constant,
)
(
) or +
(a sin + c cos
the inclination of the axis of the
body to the horizon, Q the angular velocity of
the vertical plane containing its axis,
the angular velocity of the
body about its axis,
A the moment of inertia of the body about a diameter of its base, C the moment of
inertia of the body about its axis and a the distance of the centre of
gravity from the
base
(-4
where 6
2%
<9
<9)
is
o>
4.
wheel with 4
spokes arranged symmetrically rolls with its axis horizontal on a

If the wheel and spokes be made of a fine
heavy wire,
perfectly rough horizontal plane.
prove that the condition for stability
is
where a
5.
is
the radius of the wheel and
body
rolls
under gravity on a
V its
velocity.
(Coll.
fixed horizontal plane.
Exam.)
If this plane be taken as
plane of yz, shew that
2m
(y -.y A )z-
(z
-z A )y} = Constant,
where
and (.r^, yA ZA ) of the point of contact,

(x, y, z) are the coordinates of a particle
and the summation is extended over i\ll the particles of the
body.
,
(Neumann.)
6.
One portion
perfectly rough.
moves with
of a horizontal plane
is
perfectly
uniform heavy ellipsoid of semi-axes
smooth
velocity v in the direction of its a-axis along the
towards the rough.
Shew
arid the other portion is
has
its 6-axis vertical and

smooth portion of the plane
(a, b, c)
that, if
the ellipsoid will return to the smooth portion, k

being the radius of gyration about
the c-axis, and that the motion will then consist of an oscillation about a
steady state of
*
motion.
In the special case a = 26, shew that after the return of the
ellipsoid to the smooth
the 6-axis can never make an
angle with the vertical which is greater than
portion,
arctan
Vf
(Coll.
Exam.)
240
Dissipative Systems
[CH.
A shell in the form of a prolate spheroid whose centre of gravity is at its centre
7.
contains a symmetrical gyrostat, which rotates with angular velocity to about its axis and
Shew that in the steady
whose centre and axis coincide with those of the spheroid.
motion of the spheroid on a perfectly rough horizontal plane, when its centre describes
a circle of radius c with angular velocity Q, the inclination a of the axis to the vertical
is
given by
{
Mbc (a cot a + 6)
A b cos a + C (a sin a + c)} O 2 + C baQ - Mgb
(a
b cot a)
=0,
the mass of the shell and gyrostat, A the moment of inertia of the shell and
to their axis, (?, C
gyrostat together about a line through their centre perpendicular
those of the shell and gyrostat respectively about the axis, a the distance measured
the shell and plane from the centre and
parallel to the axis of the point of contact of
b its distance from the axis.
where
is
uniform perfectly rough sphere of radius a starting from rest rolls down under
between
two non-intersecting straight rods at right angles to each other whose
gravity
shortest distance apart is 2c and which are equally inclined at an angle a to the vertical.
If po, po are the original distances of the points of contact from the points where the
their distances at a subsequent time when
shortest distance intersects the rods and
8.
p,
the velocity
is V,
shew that
-/
,.
i6c4
(p*-
and that
I,
1889.)
A particle moves under gravity on a rough helix whose axis is vertical. If a be the
9.
radius and y the angle of the helix, shew that the velocity v and arc described s can be
expressed in terms of a parameter 6 by the equations
_
A particle
cc y
cos
y + 2 sin
y)}
inclined
projected horizontally with velocity u so as to slide on a rough
motion.
the
Investigate
10.
plane.
Prove that
is
if
2p,cota>l,
the particle approaches asymptotically a line of greatest slope at distance
u2
g
where
p,
is
2p,
2
4/x
cos a
cos 2 a
the coefficient of friction, and a
is
- sin 2 a
the inclination of the plane.

(Coll.
Exam.)
A rough cycloidal tube has its axis vertical and vertex uppermost. If a be the
11.
radius of the generating circle and a particle be projected from the vertex with velocity
\/4ag sin
a,
shew that
it will
reach the cusp with velocity equal to
2
[4aff cos a
where
a is the angle of friction.
(1
-2
sin
a~
(Jir
~ a) tana
}]*,
(Coll.
Exam.)
Non-1lolonomic Systems.
vm]
241
Dissipative Si/stenis
A heavy rod of length 2 is moving in a vertical plane so that one end is in contact
12.
with a rough vertical wall and the other end moves along the ground supposed to be equally
rough and the coefficient of friction for each of the rough surfaces is tan e. Shew that the
;
any time
inclination of the rod to the vertical at
0(&
+ a 2 cos2f)-a 2
(9
is
given by
sin2e==a#sin(0-2f).
Exam.)
(Coll.
thin spherical shell rests upon a horizontal plane and contains a particle of finite
mass which is initially at its lowest point. The coefficient of friction between the particle
13.
and the shell is given, that between the shell and the plane being practically infinite.
Motion in two dimensions is set up by applying to the shell an impulse which gives it an
angular velocity Q. Obtain an equation for the angle through which the shell has rolled
when the
particle begins to slip.
Exam.)
(Coll.
a is placed in a vertical plane touching a uniform rough

board which can turn freely about a horizontal axis in the upper surface of the board
(ft.)
through its centre of gravity, the point of contact of the disc being at a distance b from
14.
circular disc of radius
this axis.
string, parallel to the surface of the board, is attached to the point of the
from the board and to an arm perpendicular to the board at the axis, and
The centre of gravity of the board and arm lies in the
connected
to the board.
rigidly
axis.
The system starts from rest in that position in which the centre of the disc lies in
disc furthest
Shew that slipping will take place between the disc

the horizontal plane through the axis.
and the board, when the board makes an angle Q with the vertical given by
tan
where
is
the
moment
of inertia of the board about the axis divided
disc.
by the mass of the

(Coll. Exam.)
15.
hoop is projected with velocity V down a plane of inclination a, the coefficient
of friction being p (
tan a). It has initially such a backward spin O that after a time ^ it
starts moving uphill and continues to do so for a time t 2 after which it once more
>
Shew
descends.
that, if the
motion take place
given inclined plane, then

(ti
t2
)ff
sin a
in a vertical plane at right angles to the
= aQ. -
V.
Exam.)
(Coll.
ring of radius a is fixed on a smooth horizontal table a second ring is placed

on the table inside the first and in contact with it, and is projected with velocity T 7 but
without rotation, in a direction parallel to the tangent at the point of contact. Find the
16.
time that elapses before slipping ceases between the rings if the coefficient of friction
is p, and prove that the point of contact will in this time describe an arc
between them
of length (a log
2)//*.
Discuss the motion that will ensue
if
at the
moment
slipping ceases the fixed ring be
and left free to move, and prove that during the time that the inner rinhalf round the outer one the centre of the latter will be displaced a distance
released
where m,
M are the
rin g17.
rolls
masses of the inner and outer rings and b is the radius of the inner
In the vertical motion of a heavy particle

descending in a
shew that the quantity
medium whose
resistance
varies as the square of the

velocity,
-ka.
e
where kv
is
the resistance, and a and
+e
are the distances described in two successive

equal
intervals r of time, depends only on T and is
independent of the initial velocity.
/3
(Coll.
w. D.
Exam.)
16
242
Prove that a heavy
18.
particle, let fall
Dissipative Systems
from rest
in a
medium
in
[OH.
which the resistance
varies as the square of the velocity, will acquire a velocity Uta.nh(gt/U), and describe
2
denotes the terminal velocity in the
a space
log cosh (gtjU)lg in a time t, where
medium.
Shew
also that, for the complete trajectory of a projectile in such a
between the asymptotes
is
U
where V
19.
is
the velocity
medium, the angle 6
given by
z
j
T72
when the
= arcsinh cot 6 + cot 6 cosec 0,

projectile
moves
horizontally.
(Coll.
Exam.)
Shew that the horizontal and vertical coordinates (x, y) of a particle moving under
medium of which the resistance is R satisfy the equation
gravity in a
dx3
v being the velocity
and
v* cos 3
(j)
the inclination of the tangent to the horizontal.

(Coll.
Exam.)
in a medium in which the resistance

20.
particle is moving, under gravity,
Shew that the equation of the trajectory referred to the vertical
varies as the velocity.
and a line parallel to the direction of motion when the velocity was infinite, can
asymptote
y=b
(Coll.
log (xjo).
Exam.)
Prove that in the motion of a projectile through a resisting medium which causes
21.
with
a retardation kv3, where k is very small and the particle is projected horizontally
is
the
of
(neglecting
path
velocity F, the approximate equation
2~.2-
the axis of x being in the direction of projection and the axis of y vertically downwards.
(Coll.
Exam.)
moves in a straight line under no forces in a medium whose resistance

v is the velocity and s the distance from a given point in the line.
where
(^_
Shew that the connexion between s and t is given by an equation of the form
22.
particle
v s logs)/*,
is
where a and
are constants.
A particle is moving in a resisting medium under a central attraction ; shew that,

be the retardation due to the resistance of the medium, and v the velocity, the rate of
vector to the fixed centre of force varies as
description of areas by the radius
23.
if
(Coll.
v,-f^ctt
J v
Exam.)
Prove that in a resisting medium, a particle can describe a parabola under the
24.
resistance at
action of a force to the focus which varies as the distance, provided the
vertex.
a point, where the velocity is v, be k {v(v-v )fi where v is the velocity at the
;
Determine
If
(Coll.
k.
A particle moves in a resisting medium

25.
R be the resistance, shew that
d
f 7
>
dr\
*
ds-\Pp*-T-\
\
dp)
r being the radius vector
and
under a force
-2%
2
,
the perpendicular on the tangent.
Exam.)
tending to a fixed centre.
vrn]
If
M=
l/r,
P=p.u
2
,
and
R = kv
equation to the path
differential
2
,
and we neglect
and higher powers, shew that the
is
P ecu
^2~ 3
1,7
\ctu
\~
-i
A being a certain constant.
A particle
26.
is
243
Dissipative Systems
Exam.)
(Coll.
moving under a central
force
(?)
<
repelling
it
from the
origin, in
a resisting medium which imposes a retarding force equal to k times the velocity.
that the orbit is given by the equations
~
~
A -; 3 e 2fc = ( (f),
r-6 = /ie~
r-^-kr
Prove
Jct
where A
is
a constant quantity.
Exam.)
(Coll.
A particle is moving in a circle under a force of attraction to an interior point

27.
varying as the distance the resistance of the medium is equal to its density multiplied by
Shew that the density at any point is proportional to the
the square of the velocity.
;
tangent of the angle between the lines joining

the circle.
it
to the centre of force
and the centre of

Exam.)
(Coll.
28.
A rod of length a is rotating about one extremity, which is fixed, under the
action of no forces except the resistance of the atmosphere.
Supposing the retarding
2
shew that
effect of the resistance on a small element of length dx to be Adx (velocity)
.
the angular velocity at the time
where
Mkz
is
the
moment
t is
given by
11
Aa*
of inertia about the fixed extremity,
and Q
is
a constant.
(Coll.
29.
Exam.)
disc of mass M, turning on a smooth horizontal table with

but without any translational velocity, strikes a smooth horizontal rod
Prove that the angular velocity is diminished in
at its middle point.
smooth oval
angular velocity
of
mass
the ratio
e is the coefficient of elasticity, x the distance of the centre of gravity from the
normal at the point of impact and k the radius of gyration about a vertical axis through
the centre of gravity.
(Coll. Exam.)
where
Two
and mass TO, are jointed together at their upper ends

symmetrically, with its plane vertical, on to a smooth inelastic
Just before impact the joint has a velocity V and each rod has an angular
plane.
Shew that the impulse
velocity 12, tending to increase its inclination a to the horizon.
between each rod and the plane is
30.
rods, each of length a
and the system
falls
m (P + c
where
c is
moment
sin 2 a)
F+ aQ cos a)/{ 2 + c2 +a (a - 2e) cos2 a},
the distance of the centre of gravity of each rod from the joint and
of inertia of each rod about its centre of gravity.
(Coll.
mk 2
is
the
Exam.)
31.
Three equal uniform rods AB, BC, CD, each of length 2a, and hinged at B and C,
are in one straight line and moving with a given velocity in a horizontal plane at
The ends A and
meet simultaneously two fixed inelastic
right angles to their lengths.
obstacles, reducing
triangle,
A and D
and shew that
to rest.
Determine when they
of the original
momentum
is
will
form an equilateral
destroyed by the impacts.

(Coll.
Exam.)
162
244
Dissipative Systems
[CH.
vm
32.
A smooth uniform cube is free to turn about a horizontal axis passing through
the centres of two opposite faces and is at rest with two faces horizontal an
equal and
similar cube is dropped with velocity u and without rotation so as to strike the former
along a line parallel to the fixed axis and at a distance c from the vertical plane containing
it, prove that the angular velocity imparted to the lower cube is
;
(\+e)cu
where a
is the inclination to the horizon of the lower face of the

falling cube, 2a
length of an edge, k the radius of gyration and e the coefficient of restitution.
Find also the motion of the upper cube immediately
is
the
after the impact.

(Coll.
Exam.)
and radius c impinges without rotation

perfectly elastic circular disc of mass
and length 2a which is free to turn about a pivot at its centre, the
upon a rod of mass
point of impact being at a distance b from the pivot. Prove that if the component of the
the
velocity of the centre of the disc normal to the rod be halved
Mb 2 =
the
33.
by
impact,
friction being sufficient to prevent sliding.
ma?,
(Coll.
Exam.)
34.
A perfectly rough sphere of radius a is projected horizontally with a velocity V
from a point at a height h above a horizontal plane. The
sphere has also initially
an angular velocity O about its horizontal diameter
to the
of its
perpendicular
Shew
motion.
that before
ceases to
it
bound on the plane
it
plane
passes over a horizontal
distance
where
e is
the coefficient of elasticity, and the distance
is
reckoned from the
first
point of
contact.
Compare the
35.
final
with the
initial kinetic energy.
(Coll.
Exam.)
elastic
homogeneous
sphere (coefficient of elasticity e) is projected against
a perfectly rough vertical wall so that its centre moves in a vertical plane at right
angles
to the wall.
If the initial components of the
velocity of its centre are u and v, and
angular velocity (Q) is about an axis perpendicular to the vertical plane, find
the subsequent motion after impinging on the wall, and shew that if its centre returns
to its original position the coordinates of the point of impact referred to this
point are
its initial
g
where a
is
the radius of the sphere.
(Coll.
Exam.)
CHAPTER
IX
THE PRINCIPLES OF LEAST ACTION AND LEAST CURVATURE

98.
The
The
chief
trajectories
of a dynamical system.
of
object
investigation in Dynamics is the gradual change in
time of the coordinates (q lt qa ..., q n ) which
specify the configuration of a
dynamical system. When the system has three (or less than three) degrees
,
of freedom, there
often a gain in clearness when we avail ourselves of a

geometrical representation of the problem if a point be taken whose rect
is
angular coordinates referred to fixed axes are the coordinates (q 1} q2 q3 ) of

the given dynamical system, the
path of this point in space can be regarded
as illustrating the successive states of the
In the same way when
system.
n 3 we can still regard the motion of the system as
represented by the path
of a point whose coordinates are (q lt 2 ..., n ) in
q
q
space of n dimensions;
this path is called the
of
the
and
its introduction makes
trajectory
system,
it natural to use
terms
such
as "intersection," "adjacent," etc.,
geometrical
,
>
when speaking
of the relations of different states or

types of motion in the
system.
99.
Hamilton s
principle, for conservative holonomic systems.
Consider any conservative holonomic dynamical

system whose configuration
at any instant is specified by n
independent coordinates (q1} q.2) ..., q n \ and
let L be the kinetic
Let a given
potential which characterises its motion.
arc
in space of n dimensions
of
a
of
the system,
represent part
trajectory
and let CD be part of an adjacent arc which is not
a
necessarily
trajectory
it would however of course be
sub
possible to make CD a
AB
by
be the time at which
trajectory
Let t
jecting the system to additional constraints.
the representative point (q 1} q. ..., q n )
on AB: we
occupies any position
shall suppose each point on CD correlated to some value of the
time, so
that there will be a point Q on CD (or on the arc of which CD is a
portion)
which corresponds to the same value t as
does.
As the arc CD is
2>
described, the correlated value of
be supposed to vary
continuously
in the same sense.
A moving point which describes the arc CD will there
fore pass through positions
corresponding to a continuous sequence of values
of q lt qi, ...,qn t, and
consequently to each point on CD there will correspond
,
a set of values of q 1}
q<>,
...,
qn
will
The Principles of Least Action
246
We
denote by 8 the variation by which we pass from a point of
shall
CD
to that point of
denote by
shall
which
+A
t lt t
terminal points A, B, C,
at
any point
If
[OH. ix
+ A^
AB
same value of the time, and
correlated to the
is
the values of
which correspond to the

the value of the function
LR
respectively, and by
of either arc.
now we form the
difference of the values of the integral
q n qi, q,
((& q*,
arcs AB and CD respectively,
,
taken along the
Ldt
= L&t -L&t +
Ldt
qn, t)dt,
we have
SLdt
AB
CD
-LA A +
64,
(o
r=
by Lagrange
,
at \rmi
r=l
But
if
/s
if
equations,
dt
\r=l
(Agr )B denote the increment of gr in passing from
and similarly
we have
dt
,,
to D,
we have
(Agv)^ denote the increment of q r in passing from J. to C,
and consequently
Ldt -
f
j
cx>
Ldt =
^#
Ag r
-^r-
L*=I dq?
and that
Suppose now that C coincides with A, and D coincides with B,
C and D are t and t, respectively, so that A^,
an(^ -^ then the last equation becomes
A(? n A, are zero a t ^
Ag 2
the times correlated to
>
= 0,
which shews that
the integral
Ldt has a stationary value for any part of an
actual trajectory AB, as compared with neighbouring paths CD which have

the same terminal points as the actual trajectory and for which the time has
the
same terminal
*
values.
This result
is
Hamilton, Phil. Trans. 1834,
Hamilton s principle*.
called
p.
307
ibid. 1835, p. 95.
and Least Curvature
99, 100]
247
If the kinetic potential L does not contain the time

explicitly, we can
evidently replace the condition that the time is to have the same terminal
values by the condition that the total time of description is to be the same
for
AB
n
gZ
CD, since S q r ^ - L, which represents the
r= l
r
as for
the system,
is
Helmholtz, J. filr Math.

value of
c.
(where the q s and
so that
total
Oq
in this case constant.
151,
p.
n,
...,
remarked that the conditions
ql
...,
q n)
+2
Vr
for
energy of
a stationary
(f,-*r)l dt
}
are regarded as independent variables) are
ffs
we again obtain Lagrange s equations.
The principle of Least Action for conservative holonomic systems.
100.
Suppose now that the dynamical system considered
is
such that the
kinetic potential does not involve the time explicitly, so that the
integral
of energy
4
2
rl
qr
dL - T I
L=h
dq r
AB to
be part of a trajectory and CD to be

Taking
part of
to
the
successive points of which values of the time are so
any adjacent arc,
correlated as to satisfy an equation of the form
as before
exists.
dL
r=l
where AA
is
2
=i
a small constant,
dL\
\dt-
qr
oqr J
(h
|
CD
A/&
OQV
we have
+ A/0 dt- !
J
dL\
~
}dt
.
AB\r=\
qr
dqJ
hdt
AB
!
J CD
Ldt-
Ldt
t
J
AB
coincides with
If therefore
and that
L = h + AA,
4r^~-
is
we suppose that C
zero, we shall have
and
coincides with B,
248
[on. ix
n
"bL\
dt has a stationary value for any

oq r /
part of an actual trajectory, as compared with neighbouring paths between the
same termini for which the time is correlated to the coordinates in such a way
qr
//[2
\r=i
^-r
to satisfy the same equation of energy.

Least Action, the integral
This
as
is
called the principle of
being called the Action.
In natural problems, for which L is the difference of a kinetic energy T,

homogeneous of the second degree in the velocities, and a potential energy
F, independent of the velocities, we have (41)
and the stationary integral can therefore
p.
"
in this case be written iTdt.
The Principle of Least Action originated in Maupertuis attempt (Mem. de I Acad., 1744,
417) to obtain for the corpuscular theory of light a theorem analogous to Ferrnat s
Principle of Least Time."
Maupertuis principle was established by Euler (Addit. n.
p.
309 of his Methodus inveniendi lineas curvas, 1744) for the case of a single particle under
force, and by Lagrange

more general problems.
a central
Example
1.
Shew
(Miscell. Taurin.
II.
(1760-1), Oeuvres,
I.
p.
365) for
much
that the principle of Least Action can be extended to systems for

Let the expression
exist, in the following form.
which the integral of energy does not

2
r=i
qr
~^r
vqr
be denoted by h
then the integral

n
.
2 qr
dL
+ 1 dh\
-j- }dt
,
c ?r
dt)
has a stationary value for any part of an actual trajectory, as compared with other paths
between the same terminal points for which h has the same terminal values.
Example 2. If a dynamical system which possesses an integral of energy is reduced to

a system of lower order as in
42, shew that the principle of Least Action for the
original system is identical with Hamilton s principle for the reduced system.
Extension of Hamilton
101.
principle to non-conservative dynamical
systems.
We
now extend Hamilton s principle to holonomic dynamical

which the forces are no longer supposed to be conservative.
shall
systems in
Let
denote the kinetic energy of such a system, and
let
2 Q r $q r
r=l
denote the work done on the system by the external forces in an arbitrary
displacement (Bqi, &q, ..., &q n ) the equations of motion of the system are
,
therefore
d /dT\ - dT
-
-j. U-.
dt
\dqr/
dq r
Qr
(r=l,
2, ...,
?i).
and Least Curvature
100-102]
249
Let a denote a part of a trajectory of the system, and let ft be an

adjacent
same terminals, the times correlated to the path /3 at the
terminals being the same as the values t and ^ of the time at the terminals
in the trajectory a; then if 8 denotes the variation
by which we pass from a
position on a to the contemporaneous position on /3, we have
arc having the
8r +
!,
**) dt
t,
r=i (oqr
<
0.
This result
Qr Sqr
r=l
is (like
the theorem of
Hamilton s
to
99,
which
is
dt
known
really a particular case of it)
as
principle.
102.
Extension of Hamilton
non-holonomic systems*.
principle
and
the principle of Least
Action
We
is
shall now shew that Hamilton s

principle, when suitably formulated,
true even for dynamical
which
are not holonomic.
systems
Consider a non-holonomic conservative

system, in which the variations
..., q n ) are connected by
non-integrable
kinematical equations
of the n coordinates
(q l} q2
A
where
that
Au A
,
lt
12
dq
...,
+ A&dq +
a
A nm T
,
l ,
...
...,
Tm
are given functions of q l}
denotes the kinetic potential, the motion

the n equations
if
d fdL\
dt
(dfr )
(k=l,
~ dL = ** An + ** A
<*+
is
+ ^nA rm
dq~r
together with the above kinematical equations; the
2, ...,
q.2 ,
...,q n
determined
(r
= 1,
unknown
m)
2,
so
87) by
n),
quantities
being
<?i>
Let
AB
derived from
*
Cf.
$2>
<Jn,
^-i, ^-2)
^m-
be part of a trajectory of the

system, and let CD be a path
by displacements consistent with the instantaneous kine-
AB
Holder, Gott. Nach. 1896,
p.
122,
and Voss,
Gntt.
Nach. 1900,
p. 322.

matical equations,
omitted
this path
[OH. ix
the above kinematical equations with the terms Tk dt

CD will not in general be itself a path whose continuous
i.e.
CD
description would satisfy the kinematical conditions, so
is
really a kine-
matically impossible path.

It may naturally be asked why we do not take CD to be a kiriematically possible path
the answer to which is, that in that case the displacements from
to CD would not be
:
AB
displacements consistent with the kinematical equations for in non-holonomic systems,

if two
adjacent possible configurations are given, the displacement from one to the other
is not in general a
there are infinitely more possible adjacent
possible displacement
:
positions than there are possible displacements from the given position.
Proceeding as in the proof of Hamilton

as usual a
displacement from a point of
AB
principle given in | 99, B denoting

to the
contemporaneous point on
CD, we have
Ldt-l
JAB
CD
Ldt = L B
^-L
A &t
-f
(~
Jtt r=i\oqr
Bq r
~
oqr
Bq r ] dt
/
...
+\ m A rm )Bq rldt.
)
Since the displacements obey the relations
A
it
lk
8q l
+ A 2k Bq2 +
...
+ A nk Bq n =
\ g A rs Bq r
follows that the terms of the type
0,
in the integral annul each
we have
other, so
T
TJ.
Left
>
J^
Z/<ft
T
A
T
A
= Lj
Aj - LÂô 4i
-0
L\V44
.
Bqr
Jfcr-ll0$r
fV*<
-
. -
7,
cv
dt\dq r j
Bq
1r
\<dt.
We
From this point the proof proceeds as in 99.

thus obtain the result
that Hamilton s principle applies to every dynamical system, whether holonomic
or not.
In every case the varied path considered is to be derived from the
actual orbit by displacements which do not violate the kinematical equations
representing the constraints; but it is only for holonomic systems that the
varied motion is a possible motion ; so that if we compare the actual motion
with adjacent motions which obey the kinematical equations of constraint,

Hamilton s principle is true only for holonomic systems.
to
The same remarks obviously apply to the principle of Least Action, and
Hamilton s principle as applied to non-conservative systems.
103.
Are
the stationary integrals actual
minima
Kinetic foci.
we have only shewn that the integrals which occur in Hamilton s

and
the principle of Least Action are stationary for the trajectories
principle
The question now arises, whether they
as compared with adjacent paths.
are actually maxima or minima.
So
far
and Least Curvature
102, 103]
251
We shall select for consideration the principle of Least Action, and for
convenience of exposition shall suppose the number of
degrees of freedom
in the dynamical
a kinetic
system to be two, the motion being defined
by
energy
T=â n (q
q z ) q?
a 12 (q
l ,
qs )
q,
qa
and a potential energy
The
discussion can be extended without

difficulty to Hamilton s principle,
with any number of degrees of freedom. The
principle of
Least Action, as applied to the above system, is (
100) that the integral
and
to systems
F
I
a
n<?i
2a, 2 g 1 5 2
2
22<7"
dt
has a stationary value for an actual

trajectory as compared with other paths
between the same termini for which dt is connected with the differentials of
the coordinates by the same equation of
energy
T+V=h.
This latter equation gives
or
dt
{2 (h
~2
i/r)}
so the stationary
integral can
where q2 stands
for
dq 2 /dq
(a ll dq 1
we
first
2a 2 dq 1 dq
1
+ a^dq^,
be taken to be
this integral is to be
at each of which the values of

q
Writing
and
q.2
taken between terminals,
are given.
this equation
maxima and minima (which was

by Jacobi) by a method suggested by Culverwell*.
shall discuss the discrimination of its
effected
Consider any number of paths adjacent to the actual

These
trajectory.
paths will be supposed to have the same terminals, and to be continuous,
but their directions may have abrupt
changes at any finite number of
For such a path let (q lt q2 + &q2 ) be a
points.
point corresponding to a
point (ft, ga ) on the actual trajectory; we shall
write
for
frequently
8q 2
a small constant the order of which determines the order of
is zero at the terminal
magnitude of the quantities we are dealing with, and
a<
where a
is
</>
points.
*
Proc. Lond. Math. Soc.
xxm.
(1892), p. 241.
252
[CH. ix
Let the expansion of the function
f(q 1}
let 8
/ denote the terms
When
value of
in a2
at
<
and
is
to
<
</>
Thus
1
Un
dq
(j)
degree in
and
its
terminals are fixed, the
(f>
If therefore
and R,
it
ft
follows that
be the numerically
can never exceed
<f>
sufficiently small the
can be diminished indefinitely.

very small, the
is
and as the sign of
^\2 J?
Un = ~ = (h"
most important term in
this is always the
(a n
same
+ 2a ls q + a^_q.2 2 )~ *
t
this is positive, since the kinetic
therefore a n a&
a 12
ranges the Action
Now
</>
is
is
positive.
We
is
-a
sign
maximum
2
12 ),
energy is a positive definite form and

thus have the result that for small
minimum for
consider any point
(a n a^
82I is
U n (the
as that of
of dq t is taken to be positive), we see that for small ranges,

Now
or minimum according as Uu is negative or positive.
and
For since
<.
we have
between
the range
if
and consequently by taking the range
<?i(p))/3,
ratio of
first
small,
denote the terminals.
greatest value of
(g KJZ)
<
large compared with the value of
is
.p
where
the range of integration
any point
zero at the terminals,
g2
57 denote the terms involving a in the
and
is
+a<f>,
powers of a be
in ascending
let
qa
the actual trajectory.
A on an actual trajectory, and let another actual

A making a very small angle with the first. If
trajectory be drawn through
this intersects the first trajectory again, say at a point B, then the limiting
when the angle between the trajectories diminishes

position of the point
on the first trajectory, or the
indefinitely is called the kinetic focus of
point conjugate to A.
We
shall
provided the
For
let
now shew
final
point
that for finite ranges the Action is a minimum,

not beyond the kinetic focus of the initial point.
is
and Q be the terminals
we have seen
that
if
is
very near
2
2
to P, the quantity S / is always positive and of order a compared with the
and Q. It is therefore evident that as we remove
value of / for the limits
and Least Curvature
103, 104]
253
Q further from P, the quantity 8-1 cannot become capable of a negative

value until after Q has passed through the point for which B 2I can vanish
for a suitably chosen value of
(/>.
Suppose then that PBQ is an arc of an actual trajectory, Q being the first
for which 2/ is zero
point for which it is possible to draw a varied curve
PHQ
we
shew that the varied curve PHQ must itself be a

For if
trajectory.
it is not a
trajectory between two of its own points A and C (supposed near
each other), let a trajectory ADC be drawn between these
Then the
points.
shall
ADC is less than that taken along AHC, so the integral

PADCQ is less than that along PHQ, which by hypothesis is
that along PBQ.
Hence frl along PADCQ is negative, and there
integral taken along
taken along
equal to
Q cannot be the
fore
ceases to be positive
that
is a
PAHCQ
Action
first
;
point for which, as
which
is
trajectory,
contrary to
and Q
is
we proceed from P, the
what has been proved.
the kinetic focus of P.
variation
It follows
Hence
the
a true minimum, provided that in

passing along the trajectory the
final point is reached before the kinetic focus of the initial point.
is
Lastly we shall consider the case in which the kinetic focus of the initial
point is reached before we arrive at the final point. Suppose, with the notation
P and R
PHQ and the
just used, that the initial and final points are

and
be taken, the former on the curve
and
let
latter
two points
on the arc
QR
EGF joining
along EGF is less
taken along PEGFR is less
these points being taken so close

together that the trajectory
them gives a true minimum. Since the integral taken
than that along EQF, it follows that the integral

than that along PEQR but the latter is
equal to that along
both integrals are equal from P to Q; and therefore the
;
PBQR,
integral along
since
PBQR
minimum but it is not a maximum, since the integral taken along

small
any
part of it is a minimum. Hence when the kinetic focus of the initial
point is reached before we arrive at the final point, the Action is neither a
is
not a
maximum
nor a minimum,
simple example illustrative of the results obtained in this article is furnished by the
motion of a particle under no forces on a smooth sphere. The
trajectories are greatcircles on the sphere, and the Action taken
or
along any path (whether a
trajectory
not)
proportional to the length of the path. The kinetic focus of any point A is the
diametrically opposite point A on the sphere, since any two great-circles through A
intersect again at A
The theorems of this article amount therefore in this case to the
statement that an arc of a great-circle joining any two points A and
on the sphere is
the shortest distance from A to B when (and only when) the
point A diametrically
opposite to A does not lie on the arc, i.e. when the arc iu question is less than half
is
a great-circle.
104.
Representation of the motion of dynamical systems by means of
geodesies.
The principle of Least Action leads to an interesting transformation of

the motion of natural dynamical systems with two
degrees of freedom.
254
[CH. ix
Let the kinetic energy of such a system be
and
energy be
let its potential
^(g 1}
By
q.2 ).
which the
to that family of solutions for
total
100, the orbits corresponding

energy is h are given by the
condition that
r
2
I
(an9i
Zouqiqi
+ a^q*} dt
is stationary for
any part of an actual orbit, as compared with any other arc
between the same terminals for which dt is connected with the differentials
of the coordinates
The
by the
relation
integral
r
I
tyyz (a u dq*
(h
is
But
therefore stationary.
2a 12 dq l dq 2
this integral expresses the principle of Least
Action for the motion of a particle under no forces on any surface whose
linear element is given by the equation
ds 2
and
therefore
is
Consequently
same as
= (h
1.
{a n dq^
%a iz dq^dq z
aZ2 dq./),
the defining condition of the geodesies on this surface.

of the orbits in the given dynamical system are the
the equations
the equations
to the geodesies
i|r)
of the geodesies on
Shew that the
this surface.
parabolic orbits of a free heavy projectile correspond
on a certain surface of revolution.
Example 2. Shew that the orbits described under a central attractive force
(r) in a
plane correspond to geodesies on a surface of revolution, the equation of whose meridiancurve is z=f(p), where
<p
and where
and p are connected by the
relation p 2 =r 2
<p
The least-curvature principle of Gauss and Hertz.
105.
We shall now discuss a principle which, like Hamilton s principle, can be

used to define the orbits of a dynamical system, but which does not involve
the sign of integration.
(xr
(X
In any dynamical system (whether holonomic or non-holonomic) let

y r z r } be the coordinates of a typical particle m r at time t, and
r,
Yr Zr )
,
the components of the external force which acts on the particle.
Consider the function

..
xr
X
--
--F y
m j +(y
V mj
\
r
-
/..
and Least Curvature
104, 105]
255
where the summation is extended over all the particles of the

system, and
where (x r yr zr ) refer to any kinematically possible
path for which the
coordinates and velocities at the instant considered are the same as in some
,
actual trajectory.
This function substantially represents what was called
by
Gauss the constraint and by Hertz (who however considered
primarily the
case in which the external forces are
zero) the curvature* of the kinematically
possible path considered.
We
shall
are supposed
In what follows Hertz
terminology will be used.
shew that of
to
all paths consistent with the constraints

(luhich
do no work), the actual trajectory is that which has the
least curvature^.
In the simple case of a single particle

moving on a smooth surface under no external
reduces to the statement that the curvature in
space (in the
ordinary sense of the term) of the orbit is the least which is consistent with the condition
that the particle is to remain on the surface.
forces, this result clearly
To
establish this result, let the equations which

express the constraints
to typify any one of the three coordinates of
(using
any particle) be
as r
?.xkr dxr
(k=l,2,
...,
where the
coefficients
xkr are given functions of the coordinates.
entiating these relations,
m),
Differ
we have
OTi*
xr xs =
Let x r be a typical component of acceleration in the

path considered
is
to
be
but
is
not
supposed
the
kinematically possible,
(which
necessarily
actual trajectory), and let xro be the
corresponding component of acceleration
in the actual trajectory.
Subtracting the preceding equation, considered as
relating to the actual trajectory, from the same equation, considered as
relating to the kinematically possible path, we have (since the velocities are
the same in the two paths)
Safe. (x r
r
- xn) =
(A?
1, 2,
m).
This equation shews that a small displacement of the

system, in which
the displacement 8xr of the coordinate xr is
proportional to (x r - x m ), is con
sistent with the equations of constraint, i.e. is a
possible displacement.
The components
of the forces exercised

by the constraints are typified
by
Strictly speaking, the square root of this function, and not the function
the curvature by Hertz.
itself,
was called
t Gauss, Crelle s Journal, iv. (1829), p. 232; Werke, v. p. 23. Gauss measured the
constraint
by "the sum of the masses of the particles, each multiplied by the square of its deviation from
unconstrained motion." The above analytical expression for it was first
given by H. Scheffler,
Hertz s theory is given in his Mecltanik.
Zeitschrift filr Math. in. (1858), p. 197.
256
(m r x n
work.
We
[OH. ix
and in any possible displacement the forces of constraint do no

have therefore
r)
2 (m r x ro -
- x ro = 0,
(x r
r)
an equation which may be written in the form
X- \ = ~2,m
--
..
xr
or (reverting to the use of y s
r
x ro ----
..
\*
)
... 2
2,??i,.
(x r
Xro)
and z s)
F,
=^
2,ra r
\/
<sc
x r\* +
---
(..
Fry
n ---
+ Sl {(r - ^ro) + (j/r ~ j/ro) + (^r ~ ^Vo)

2
}-
Since the terms in the last summation on the right-hand side are
positive, it follows that
.,
xr
X\
-r
mj
+ y r -- +
V mj
zr
^
>
which establishes the result
(/..
x,n
2<m r
]
r
(\
all
t,
Z
---
r\
f
>nj
Fr \
+ y n --- +
mj
/..
/.
(zn
\
stated.
Expression of the curvature of a path in terms of generalised
106.
coordinates.
Lipschitz has shewn* that the curvature of a kinematically possible path

in a holonomic dynamical system with n degrees of freedom can be expressed
in terms of the derivates of the n independent coordinates
position of the system.
which define the
q 2 ..., qn) be the coordinates; let (q 1} q2 ..., qn ) be the accelera

these coordinates in any kinematically possible path, and let
..., q no ) be the accelerations in the actual trajectory which corre
20
(ry 10
to
the same values of (q l} q 2 ..., q q 1} q. ..., q n ). Using xr to typifysponds
of
the three rectangular coordinates of any particle r and
one
r to typify
any
Let
(<?!,
of
tions
</
n>
2>
the corresponding component of force, the Gauss-Hertz curvature of the path

2
and it has been shewn in the last article that this can
is ^ni r (x r
r /m r )

(x ro
*
/m r }
"2.m
(x r
x r0 )~.
Journal far Math. LXXXII. p. 323. Cf. also Wassmuth, Wien. Sitz. civ. (1895) ; and for
further work connected with the principle of Least Curvature see Leitinger, Wien. Sitz. cxvi.
(1908), p. 1321 and Schenkl, Wien. Sitz. cxxn. (1913), p. 721.
and Least Curvature
105, 106]
The
257
summations is the same for all the paths considered, since

on
the actual trajectory we can therefore omit it without
depends only
the
whole
causing
expression to lose its minimum-property, and we can call
first
of these
it
the remaining summation 2w,. (x r
x ro ) 2 the curvature of the path.
Let the kinetic energy be
T=^Za
k
qk ql}
kl
where the quantities ak{ are given functions of (q l} q 2 ..., q n ); let D denote
the determinant formed of the quantities a
and let AM denote the minor of
ak in this determinant.
,
kt>
From
the equation
= 2 S a k iq k q
2,m r x; r2
au =
we have
av
=V
2
far
r-*
2m r
\}JUy
\jdjrp
v ~^ Xr22
+ V
~
<fc
_x
..
j_
r<)
3a?r
v
2i _
But
if
we
velocities are the
...
..
same
for all the
qko ).
(qk
oqk
write
_dsdT\
3*
at
"*
since this expression
Sk =
q k qt
*
dqrfqi
and consequently, since the coordinates and

paths considered, we have
..
oqk dqi
toq*
xr
XT
Now
is
55"
\dqk j
dT
dx
^ 5~r Xr
""
dq k
zero for the actual

trajectory,
the difference of the values of
-=-
(k
= 1,
2,
.,
n),
dq k
,.
at \dq k
we have
for the
path considered and
the actual trajectory,

or
8k = 2a w (qt -
q lo )
(k=l,2,
whence we have
qk
-qko = ^ Â S
kl
(k
...,
n),
1,2,..., n)
and consequently
The curvature,
or
2m
(x r
- a^)
raSS
^
k
I
W. D.
2
,
is
therefore
17
258
But by a well-known property
of determinants,
[CH. ix
we have
22
k
and therefore
(q\>
q2,
107.
can be expressed in terms of the coordinates
finally the curvature
qn )
and
their derivates in the form
AppelVs equations.
The Gauss-Hertz law
of Least Curvature is the basis of a form in which

has
to
write
the general differential equations of dynamics.
Appell
proposed*
This form, as will be seen, is equally applicable to holonomic and non-
holonomic systems.
Consider any dynamical system
lk
dq,
+ A 2k dq +
2
...
let
+ A nk dqn + Tk dt =
(k=l,
2, ..., ni)
be the non-integrable equations connecting the variations of the generalised

in holonomic systems these equations will of course
coordinates q l q 2
.., q n
be non-existent.
,
Let S denote the function
^m
(x k
+ zk
i/ k
2
),
where
mk typifies the mass
of a particle of the system, whose rectangular coordinates at time t are

By means of the equations which define the position of the
at
any time in terms of the coordinates (q 1} q2 ..., q n ), it is possible
particles
(%k, Vk, Zk)-
to express 8 in terms of (q lt q 2 ..., qn ) and the first

these variables with respect to the time.
Moreover,
,
of constraint
we can express
m of the velocities
and second derivates of
by use of the equations
(q l} q 2 ..., q n ) in terms of the

the coordinates corresponding to these latter be denoted by
(pi,p 2
Pn-m)By differentiating these relations we can express q 1} q2 ..., qn
in terms of the quantities p 1} p.2 ...,p n _ m p ly p2 ..., p n_ m
q lt q2 ..., q and
others
let
hence
can be expressed in terms of this
n>
last set of variables.
Now any small displacement which is consistent with the constraints

can be defined by the changes (Sp lt 8p 2 ..., Bp n _ m ) in the quantities
,
m
(Pi, Pz,
Pi-m);
let
r=l
in such a displacement.
As
r &p r
in
denote the work done by the external forces

26,
we have
Journal fUr Math. cxxi. (1900),
p. 310.
and Least Curvature

Let the equation which expresses the change in
..., p n - m ) be
ock
259
in
terms of the changes
in (PI, pa,
where
7T 2
7r n _ m ) are
known functions of the coordinates: the
...,
of
this
are
of
course non-integrable.
From this we have
equations
type
9#/9pr *B irr and so the equation which expresses xk in terms of
(TTJ,
\pi>
PZ>
pnrn)
>
be of the form
will
ir r
r=l
pr +
a,
where a denotes some function of the coordinates.

equation, we have
n-m U7r
fj
rj n
ftGC
ctxic
*
whence
-^
Differentiating
this
= 7rr =
dp r
It follows that
Pnr = ^2mt
^+
k
..
xk
k
..
xk
V
and therefore
/*e
..
yk
~+y
..
f* + zk
..
dpr
9 r
+ zk 3
equations of a dynamical system, whether holonomic or not,
can be expressed in the form
dS
=Pr
(r
dJr
where
= 1,2,..., n-m),
&m
*
2
2
denotes the function
k (x k + yk + z k ), and (p,, p 2
...,p n - m } are
coordinates equal in number to the
degrees offreedom of the system*.
^8
It is evident that the result is valid even if the

quantities
are not true coordinates, but are
quasi-coordinates.
Example.
...,
p n _m
Obtain from AppelPs equations the equations
a>i
a>2
Co>
for the
p lt
motion of a
= L,
(C A) 0)30)! =
= -Ar
(A B)
2
(
C)
a>2
(03
JJ/,
<o
<i
body one of whose points is fixed; where (o^, co 2

3 ) are
the components of angular velocity of the
body resolved along its own principal axes
of inertia at the fixed point, (A, B,
C] are the principal moments of inertia, and (L, M, N)
rigid
are the
*
Sitz.
On
moments
o>
of the external forces about the

principal axes.
the connexion of these equations with the

Principle of Least Action,
p. 933.
cf.
H.
Brell, Wien.
cxxn. (1913),
172
[OH. ix
260
Bertrand
108.
s .theorem.
theorem in impulsive motion, which belongs to the same group of

as the least-curvature principle of Gauss and Hertz, is due to
Bertrand* and may be stated thus If a given set of impulses is applied to
results
of a system (whether holonomic or non-holonomic) in motion,

the kinetic energy of the resulting motion is greater than the kinetic energy
different points
the same
of the motion which the system would acquire under the action of
to the reactions
due
constraints
and
and
additional
constraints
impulses
of any
of perfectly smooth or perfectly rough fixed surfaces, or rigid connexions

between particles of the system.
For
(u, v
let
),
be the mass of a typical particle of the system, and let (u, v, w),
before
(u ly vlf Wi) denote the components of velocity of this particle
the application of the impulses, after the application of the impulses, and in
the comparison motion, respectively.
Let (X, Y, Z) denote the components of the external impulse acting on

the particle (X Y , Z ) the components of the impulse due to the con
Y + Ylt Z + ZJ the components of
straints of the system: and (X +
lt
f
the impulse due to the constraints in the comparison motion.
The equations
m(v
m(v
Subtracting,
m(u
all
-v)=Y+Y
m(w -w) = Z+ Z
-v)=Y+Y +Y m (w,- w) = Z + Z
,
+Z
we have
u)
motion are
of impulsive
=X
m(v
)=Y
m(w
w}=Z
Multiply these last equations by u lt v 1} w 1} respectively, add, and

the particles of the system we thus have
sum
for
2m
Now
-u
|(M I
MJ
+ (v
- v ) V! + (w - w) wi] = 2 (X^ +
1
from the nature of the constraints,
Y& + Z^).
follows that
it
finite
forces
the particles of the system and proportional to the impulsive

Zj), would on the whole do no work in a displacement whose
are
proportional to the quantities (u v 1} w^): and therefore we
components
acting on
forces
all
(X lt
have
I,(X 1 u
2m
or
this equation can
{(ttj
Bertrand
+Y
v1
+ (Vj
+Z w
1
v1
+ (w
w } w^\ =
2m (u + v + w ) - 2m
2
u^
notes to Lagrange
+ v? +
s
w,
Mec. Anal.
=
;
2m
{(u
- utf +
and Liouville
- v,) + (w -
(v
Journal
(1),
vn. (1842),
w,)
},
p. 166.
and Least Curvature
108]
261
which shews that
2m (V + v + w
2
and so establishes Bertrand
The theorem may
2m (V + vf + wf),
>
theorem.
readily be extended to the case
when
the forces are not
in this case the increase of kinetic energy per

impulsive but continuous
unit of time is diminished by the introduction of fresh constraints that do
not affect the potential energy.
:
The following result, due to Lord Kelvin and generally known as Thomson s theorem*,
can easily be established by a proof of the same character as the above If any number of
points of a dynamical system are suddenly set in motion with prescribed velocities, the
kinetic energy of the resulting motion is less than that of any other kinematicalty possible
:
motion which the system can take with the prescribed velocities, the excess being the energy of
motion which must be compounded with either to produce the other.
the
Lord Rayleigh has remarked + that the theorems of Thomson and Bertrand may both
be comprehended in the statement that the introduction of fresh constraints increases the
inertia, or moment of inertia, of a system.
Example. A framework of (n-1) equal rhombuses, each with one diagonal in the
same continuous straight line, and two open ends, each of which is half of a rhombus, is
formed by 2n equal rods which are freely jointed in pairs at the corners of all the
rhombuses. Impulses P perpendicular to and towards the line of the diagonals are
shew that the initial velocity,
applied to the two free extremities of one open end
parallel to the diagonal, of the extremities of the other open end is
;
3P
sin a cos a
TO cos 2 a
where
is
the mass of each rod, and 2a
/i
is
the points of crossing.
sin 2 a
the angle between each pair of rods at

1.
If the
element
is
problem of determining the motion of a particle on a surface whose linear

ds z = Edu 2 +
ZFdu dv + Gdv\
under the action of forces such that the potential energy is V(u, v), can be solved, shew
that the problem of determining the motion of a particle on a surface whose linear
element is given by
ds*= V(u, v)(Edu 2 + 2Fdudv + Gdv 2 ),
under forces derivable from a potential energy
I/
V(u,
v),
can also be solved.

(Darboux.)
2.
two dynamical systems in which the kinetic energies are respectively

an d S&ijtjijfr, and the potential energies are respectively 7 and F, the trajectories
If
Satjtjtji
in
Thomson and
Tait s Natural Philosophy,
t Theory of Sound, Vol.
i.
p. 100.
317.
262
[CH. ix
same curves, though described with different velocities, so that the relations
between the coordinates (q\, y 2 ..., q n ) are the same in the two problems, shew that
are the
where
a,
/3,
y, 8
are constants, and that

"S,b
3.
ik
dqi dqk =(yU+ 8) 2aik dqidqk
(Painleve.)
If all the trajectories of a particle in a plane, described under forces such that the
is V(x, y\ with a value A of the constant of
energy, are
potential energy of the particle

subjected to a transformation
x=$(X, Y\
where $ and
y=+(X, Y\
are conjugate functions of (#, y\ shew that the new curves so obtained are
the trajectories of a particle acted on by forces derivable from the potential energy
with a zero value of the constant of energy.

4.
If
and
(Goursat.)
denote respectively the kinetic and potential energies of a

dynamical
system, shew that
differs
from
i
_
(/
..
.9
rv
...an 2
..
arv
by a quantity which does not involve the accelerations
and hence that
maximum when the accelerations have the values corresponding to the actual motion,
compared with all motions which are consistent with the constraints and satisfy
the same integral of energy, and which have the same values of the coordinates and
is
as
velocities at the instant considered.
(Forster.)
CHAPTER X
HAMILTONIAN SYSTEMS AND THEIR INTEGRAL-INVARIANTS
Hamilton s form of
109.
We
shall
now
of motion.
the equations
obtain for the differential equations of motion of a con
servative holonomic dynamical system a form which constitutes the basis

of most of the advanced theory of Dynamics.
Let (q 1} q2 ..., q n ) be the coordinates and L(q lt q2 ..., q n q 1} q2 ..., q t)

the kinetic potential of the system, so that the equations of motion in the
Lagrangian form are
,
n>
dL
dL\
jiiaH-a r -0
f
at \oqr/
(r=l,
2, ...,
n\
dq
O T
^=Pr
Write
(r
(r
l,2,
..., TO),
l, 2,
09V-
7"
Pr =
so that
d</
From
If 8 denote the increment in

(q lt q2
...,
qn PI,
,
p2
any function of the variables

or (q 1} q2
...,p n )
-ST
2,
r =i
ft**
-
...,
we have
to small changes in the arguments,

T =
oL
^-^
5-
oqr
+^
,
oq r
\dq r
(>
o<7
r=l
*
6
%
2<
r=l
or
n).
the former of these sets of equations we can regard either of the

..., q n ) or (p 1} p 2
..., p n ) as functions of the other set.
sets of quantities (fa, q 2 ,
due
.,
pr q r +
,
4? /
2,
r=l
2 p^r - L\ = 2
r=l
(r=l
(p r oq r
q,
q n ql} q z
,
...,
qn )
Hamiltonian Systems and
264
[CH.
Thus
if
^ p r qr
the quantity
\
when expressed
L,
r=\
Cl
Ct
Cl
*Y)
ft
f)
"/)
in terms of
be denoted by H, we have
*\
\_i
s*
^40
&H=
2
r=1
**
>f>
% = 8^
}<l
*-
or
7-
<-\
a^
p r - pr^qr)
(q r
dH
dp r =
j-
/.
.
i
*-\
ai
dpr
(1),
T*
= 1&*#2
-??
T*
Ii
i2i
*l /
* *t
dg r
motion of the dynamical system may be regarded as defined by these

the
equations, which are said to be in the Hamiltonian or canonical form
IVie
dependent variables are (q lt q 2

of 2n equations, each of the
consists of
qn
...,
1}
p.2
...,
p n ), and
the system consists
first order; whereas the

Lagrangian system
n equations, each of the second order.
The Hamiltonian form was introduced by Hamilton in 1834*. In part he had been
anticipated by the great French mathematicians: for Poisson in 1809t had taken the step
of introducing a function
2
r=l
pr^r-T
and expressing it in terms of (q l q 2 ..., q n p 1} ..., p n ), and had actually derived half of
Hamilton s equations: while Lagrange in 1810 J had obtained a particular set of equations
in the Hamiltouian form, the disturbing function taking the
(for the variation of elements)
Moreover the theory of non-linear partial differential equations
place of the function H.
of the first order had led to systems of ordinary differential equations possessing this form
in 1814-15 and by Cauchy|| in 1819 (completing the earlier
for, as was shewn by Pfaff
work of Lagrange and Monge), the equations of the characteristics of a partial differential
,
equation
f(xi,xz -..,xn ,pi,pv

,
...,j0 n )=0,
*->
where
p.
dxl
Hamilton
the time,
etc.
dxz _
~
s investigation
~_
dxn
o$
^,
~
dp l
dpn
was extended to the cases when the kinetic potential contains

in 1848-50 and by Donkin** in 1854.
by OstrogradskylF
The equation
(1)
above
equation of virtual work.

8
is
It
often called the Hamiltonian
may
form of the
be written in the more symmetrical form
I p r dq r - Hdt) = d ( 5 p r Sq r /=!
r= l
HSt),
Brit. Ass. Rep. 1834, p. 513; Phil. Trans. 1835, p. 95.

t Journal de VEcole polyt. vin. (Cahier xv), (1809), p. 266.
J Mem. de Vlnst. 1809, p. 343.
Berlin Abhand. 1814-15, p. 76.

||
Bull. soc. philomath. 1819, p. 10.
Melanges de VAcad. de St.-Pet. Oct. 1848
** Phil. Trans.
1854, p. 71.
Mem.
de VAcad. de St. -Pet.
vi. (1850), p.
385.
265
their Integral- Invariants
109, 110]
which directly suggests the importance of the
differential
form
Hdt
p r dq r
r=l
in connexion with the differential equations of dynamics
When the kinetic potential L does not involve

function
will possess
not involve
will evidently likewise
Hamiltonian
and the system
explicitly, the
explicitly,
a constant.
-- L = h,
Or o^
r=l
is
137 below.
cf.
41) an integral of energy, namely
2
where A
t
t
Oqr
This equation can be written
(q 1}
q2
...,q n ,pi,p 2
..,p n )
= h,
and this is the integral of energy, which is possessed by the dynamical system
when the function
does not involve the time explicitly. For natural problems,
it
follows at once from
41 that
sum
the
is
of the kinetic
and potential
energies of the system.
Shew
Example.
that the equations of motion of the simple
pendulum
are
__
dt~dp>
where
ff=
>
dt
-l
and where q denotes the angle made by the pendulum with the vertical at time
the length of the pendulum, and the mass of the bob is taken as
unity.
110.
t,
is
Equations arising from the Calculus of Variations.
From the preceding chapter it appears that the whole science of Dynamics
can be based on the stationary character of certain
integrals, namely those
which occur in Hamilton
s principle and the

principle of Least Action
similarly the differential equations of most physical problems can be regarded
as arising in problems of the Calculus of Variations.
Thus, the problem of finding the state of thermal equilibrium in an isotropic

conducting body, when the points of its surface are kept at given temperatures, can be
formulated as follows to find, among all functions V having given values at the surface,
:
that one which makes the value of the integral
/3F\*
T + vy/
T
M8PV
,
integrated throughout the surface, a
We
shall
/8F\2)
\//\dxdydz.
-5-
minimum.
now shew
that all the differential equations which arise from

problems in the Calculus of Variations, with one independent variable, can be
expressed in the Hamiltonian form*
.
Cf. Ostrogradsky,
Mem.
de VAcad. de St. -Pet. vi. (1850), p. 385.
266
Suppose, for clearness, that there are two dependent variables

is
equally applicable to
any number
variable
t,
(t,
y, y, y,
the proof
(n)
z,
..., y,
of variables.
(m)
Let
[CH.
z,
z,
the dependent variables
...,
be a function of the independent

and their derivates up to orders m, n,
z)
y, z,
respectively.
The conditions that the
integral
(m)
\L(t,y,
may be
tions,
stationary, can,
y, ...,
(n)
y,z,z, ...,z)dt
by the ordinary procedure of the Calculus of Varia
dt
Now
dz
dt
write
dt
dL
P.m
=
by
dL
+ (-
Pm+2
i)
dt-
Pm+n =
and write
&=
Then
(m-l)
Qm+i
y>
==
Qm+2
=z
(n-D
>
(fm+n
=z
if
+p m - q m +p m y
l
+ Pm+ni (fm+n T Pm+n z
(where
267
110, 111]
supposed expressed as a function of
is
(m)
(t,
q lt
...,
q m+n Pi,
arguments q
lt
if
qm+n
lt
\J-LJ
\JJ-J
J,
r=0
dv
...,
dv
r=m+l
is
(n)
qr+i&Pr
+ Z$p m+n
m+n
p,.8qr
+ 2
q,-8p r
r=l
expressed in terms of the variables

(t,
PI, PZ, ...,p m+n
dq r
we have
and
dz
^
rm+1
m+n
8H =
^f
r\
dL
r=l
if
dz
m+n
(n)
pr &?r+i +j?H*&* +
dL
becomes
Thus,
(_/-/-/
r\
relations
dL
this
changes in the
r=l
=0
to small
pm = vLfdy,
we have
U JJ
-^
2 pr 8q r+l + p m 8y
m-\-n
p m+n
r=l
Using the
due
8 denote an increment
...,
î
-rj-
(>,
Pm+n),
")
the quantities y and z being eliminated by use of the equations
pm+n = dL/dz), and
(n)
dH
^7=^
at
dp
the differential equations
qi,
dp r
j7
at
q2
qm + n
...,
)>
dH
= -^r~
(r=l,
2, ...,
m + n),
dq r
of the problem are thus expressed in the Hamil-
tonian form.
The systems of differential equations which arise in the problems of the

Calculus of Variations are often called isoperimetrical systems.
111.
Integral-invariants.
The nature
of Hamiltonian systems of differential equations is funda

the properties of certain expressions to which
mentally connected with

Poincare * has given the
name
integral-invariants.
Consider any system of ordinary differential equations

dx\ ~
dt
where
JTj,
...,
dxn
^* =- Y
dt
X n are given
dt
functions of
xl} x2
...,
An>
xn
t.
We may
regard
these equations as defining the motion of a point whose coordinates are

(xlf xz ..., xn } in space of n dimensions.
,
Acta Math.
xm.
(1890).
268
If
[OH.
now we
consider a group of such points, which

occupy a ^-dimensional
at
the
of
the
motion, they will at any subsequent time t
region f
beginning
A _p-tuple integral taken over f is

occupy another p-dimensional region
called an integral-invariant, if it has the same value at all times t; the
number p
called the order of the integral-invariant.
is
motion of an incompressible fluid, the integral which repre

volume of the fluid, when the integration is extended over all the
elements of fluid which were contained initially in any
given region, is an
Thus, in the
sents the
integral -in variant
since the total
volume occupied by these elements does
not vary with the time.

Example
in a plane
Consider the dynamical problem of determining the motion of a particle

let (x, y} be the coordinates of the
particle, and (u, v) its
1.
under no forces
components of
velocity.
The quantity
The equations of motion may be written

x = u, y = v, w = 0, v = 0.
7=
I(8x-t8u),
where the integration is taken, in the four-dimensional space in which (x, y, u, v)

are coordinates, along the curvilinear arc which is the locus at time t of
points which were
initially
on some given curvilinear arc in the space, is an integral-invariant.

is given by the
equations
u = a, v = b, x=at + c, y =
For the
solution of the dynamical problem
where
a, 6,
c,
are constants:
7=
and
this is
independent of
t.
In the plane motion of a particle whose coordinates are (x, y] and whose
Example
velocity-components are (u, v), under the influence of a centre of force at the origin whose
2.
attraction
is
directly proportional to the distance,
x8u)
l(u8x
is
shew that
an integral-invariant.
112.
The variational equations.
The
integral-invariants of a given system of differential equations furnish

integrals of another system of differential equations which can be derived
from these.
For
let
the given system of equations be

doc
-j
(ar. lt
x2
...,
xn
t)
1, 2, ..., w).
Let (X, x2 ..., xn } and (^ + Bx1} #2 + & 2 ..., xn + &cn ) be the values of
the dependent variables at time t in two neighbouring solutions of this set of
equations where (Sx^ 8x2 ..., xn ) are infinitesimal quantities. Then we have
,
-j-
(xr
Sxr)
= X r (x + &%!,
l
x2 + 8x2
..., a; n
+ Sa;n
t)
(r
= 1,
2, ..., n),
their Integral-Invariants
111-113]
and consequently,
dX r
3X,,
j.
9X r
(r- 1,2,
...,
n equations, together with the original n equations,

a
set of 2n equations in which (x l} x2 ..., xn 8x lt 8x2
regarded as
are the dependent variables.
These
last
Now
may be
...,
8xn )
if
denotes an integral-invariant of the original system, the

quantity
2Fr (z\
-^
a.-
xn } 8acr
-j
must, since the path of integration is quite arbitrary, be zero in virtue of

precisely this extended system of differential equations and therefore
;
HFr
(x-i
x.2 , ...,
xn } 8xr
= constant
must be an
integral of these equations so that to an integral-invariant of

order one of the original system of equations there corresponds an
integral of
the extended system of equations, and vice versa.
:
If a particular solution
x2
(a?1}
Xn) of the original equations is known,

#2 .... x n ) in the extended
(ar 1}
...,
we can substitute the corresponding values
and so obtain n linear differential equations to deter

mine (&BJ, &fc 2 ..., &&n), i.e. to determine the solutions of the original equations
which are adjacent to the known particular solution. These n equations are
,
called the variational equations.
113.
Integral-invariants of order one.
Let us now find the conditions to be

.Sx,
where
M Sx
2
satisfied in order that
+...
Mn Sxn
),
n ) are functions of (x1} #2 ..., xn t), may be an integralinvariant of order one of the system of differential
equations
(Mlt
2,
...,
dxr jdt
We
= X r (X #
,
2,
. , .
xn
(r
t)
1, 2,
. . .
n).
must have
2
where the derivates of (8x1} 8x2
...,
...
Mn Sx
n)
= 0,
8xn ) are to be determined by the
270
[OH.
extended system of differential equations introduced in the
last article
and
therefore
11
rl
/l/i
/"/
i"
2(^&v + Jf ^F)-0
at
r
j.=i \ u/c
or
Since (&EI; &r2
$xr
Xtfcv
&) are
= 0.
ft*
independent, the coefficient of each quantity
in this equation must be zero
and consequently
the conditions
for
integral-invariancy are
Corollary
If
1.
an integral of the
jFû
is
;r 2
>
differential equations, say
î
= constant,
known, we can at once determine an integral-invariant.
For we have
a
dF\
dt\dxj
k-\0&
/a^\
a^\
^
a^ az
Z r ^5+ ^
dx
fc
Afc
OBr/
k=\oxk
="^
/aF
I
-^T-
dxr \dt
a^
2t
k=ioxk
\
A;
.JL(*
dxr \dt
= 0,
and therefore
the expression
dF
/Vv
2
^
Ar-l9r
is
an
integral-invariant.
Corollary
n
2.
S ^ Sorj
r=l u xr
is
The converse
of Corollary 1
aw integral-invariant of the
is
also true,
namely that if
a given function of the variables, then an integral of the system can
For we have
dx r )
k= i
dxk \dxr j
and consequently the expression

dt
k=i dxk
k=\3xk
()x r
where
U is
be found.
271
113, 114]
a given function of (# 1} x2 ...,xn t), is independent of

this is a known quantity.
value be
(t)
which
is
let its
(x^
x2
...,
xn )
<f>
Then we have
d U/dt
r
or
and
this is
(f)
(t)
dt
= $ (t),
= constant
an integral of the system.
Relative integral-invariants.
114.
Hitherto we have only considered those integral-invariants which have

the invariantive property when the domain of the initial values, over which
the integration is taken, is quite arbitrary
these are sometimes called
absolute integral-invariants. We shall now consider
integrals which have the
;
invariantive property only when the domain over which the

integration is
taken is a closed manifold (using the language of w-dimensional
geometry)
these are called relative integral-invariants.
The theory of relative integral-invariants can be reduced to that of absolute

integral-invariants in the following way.
Let
\(M
Sx,
M 8x
,
+ Mn
xn }
be a relative integral-invariant of the equations
dx r /dt =
where
(M
lt
2,
...,
M X X
n
1}
2>
...,
Xn
(r=l,
are functions of (x lt
2, ...,
x.2 , ...,
?i),
xn
t);
so that this expression is invariable with respect to t when the

integration is
taken, in the space in which (xlt xz ..., x n ) are coordinates, round the closed
curve which is the locus at time t of points which were
situated on
,
initially
some
definite closed curve in the space.
By
Stokes theorem, this integral
is
equivalent to the integral
where the integration is now taken over a diaphragm bounded by the curve
this diaphragm can be taken to be the locus at time t of
points which were
originally situated on a definite diaphragm bounded by the initial position of
;
and since the diaphragm is not a closed surface, this integral

an absolute integral-invariant of order two of the equations.
the closed curve

is
Similarly, by a generalisation of Stokes theorem, any relative integral

invariant of order p is equivalent to an absolute
integral-invariant of
order (p + 1 ).
272
115.
which
relative integral-invariant
[en.
possessed by all Hamiltonian
is
systems.
Consider now the case in which the system of differential equations

Hamiltonian system, so that it can be written
= dH
dqr
dt
where
For
this
system
denote Hamilton
initial
...,
r=
i2..n)
dq r
qn
p p
1}
pn
t).
let
s integral, so
(!, O 2
be the
dt
a given function of (q 1} q 2
is
= _3H
dpr
dpr
is
that
>
>
a n,
is
the kinetic potential
fii, /3 2 ,
let
ftn)
values of the variables

(q l
q 2)
...,
qn ,pi, p 2
...,pn )
respectively, and let S denote the variation from a point of one orbit to the
contemporaneous point of an adjacent orbit. By 99, we have
n
oil
= 2 p r oq r
r=l
Let
(q 1} q
...,
f3 r oot r .
r=l
denote any closed curve in the space of 2p dimensions in which

..., p n ) are coordinates, and let C denote the closed
q n PI, p
,
z>
t of the
points which are initially on G
the
last
round
the
set
of trajectories which pass from
Integrating
equation
curve which
GO to C,
is
the locus at time
we have
r
2 pfBor =
/3 r &ct r
and
this equation
shews that the quantity
>
2 p r &q r
=
is
a relative integral-
invariant of any Hamiltonian system of differential equations.
On
116.
We
the last
shall
systems which possess the relative integral-invariant
next study the converse problem suggested by the result of

namely that of determining all the systems of differential
article,
equations which possess the relative

(#!>
<li>
other
half.
qn) are half the
2 p r 8q r where
integral-invariant
dependent variables, and (pi,p2
Pn) are the
273
115, 116]
Consider then a system of ordinary differential equations of order 2n in

which the variables can be separated into two sets,
q 2 ..., qn ) and
(pi, p2 ...,p n }, such that
,
(</,,
is
a relative integral-invariant of the equations, and consequently by Stokes
theorem
J
is
an absolute integral-invariant.
Let the system of
where (Q lf Q2
...,
Qn
t-fc.
P P ..., Pn
lt
(<?!,
&,
..-,
be
-P,
(r-l.
2, ..,),
are given functions of
q n ,pi,p 2
...,p n
t).
As the domain of integration of the absolute integral-invariant is of two

dimensions, we can suppose that each point in it is specified by two quantities
X and p, which do not vary with the time but are characteristic of the
trajectory on which the point in question lies.
invariant can therefore be written in the form
The absolute
integral-
v 9 (qt, ^
PJ)\ - ,
[[( ^
d\dfj,,
//(
3 ,1
JJ
\i=i
(X, (A) J
and as X and
/x
do not vary with the time, we must have

_
dt
(A,,
a,
(Qi.Pi)
or
8 (X,
Owing
to the
the choice of
_
"*"
/i)
3p t
8 (x,
8 (X, /*)
8^
8 (x,
complete arbitrariness of the domain of integration and
X and
the coefficients of
^ fi ^ ^,
d
dX
equation must vanish separately.
dpif
W. D.
8^
/*)
We
dytt
dX
d/u,
and
^
dX ^*
in this
8/u,
thus obtain
dpi
lg
274
These equations shew that a function

exists such that
Q r = dH/dp r
and thus we have the
H(q
lt
qz
P = - dH/dq
r
...,
qn
(r
[CH.
p,
p-2
- I, 2,
...
pn,
t)
...,);
result that if a system of equations
dq r
dpr_
~ p
L
~dt~ Qr
(r=l
n)
dt
possesses the relative integral-invariant

f
I
( Pi
then the equations have the Hamiltonian
dqr
dt
this is the converse of the
form
dPr
dH_
dp r
= _dH_
dt
dqr
theorem of the
last article.
If
Corollary.
is
+ Pn &g)
+ p-2 S? +
Sg,
(pi 8?i
+p
^q 2
...
+ p n fyn)
a relative integral-invariant of a system of equations
dqr /dt
where k
(qi,
q*>
greater than
is
->
qn pi,
:
P*>
is
117.
dp r /dt
in the
=P
(r
2,
k),
same way that the equations
for
~.,pn) form a Hamiltonian system
^^
qk,p n +i,
_
dt
dp r
a function of (q lt q z
(q n +i, q n +2, --,
n, it follows
dt
where
=Q
...,
q n p,,
,
dq r
pz
...,p n
only, not involving
...,pk)-
The expression of integral-invariants in terms of integrals.
If the solution of a system of differential equations
is
known, the absolute and relative integral-invariants of the system may
easily
be constructed.
Thus,
where
let
c l} c 2 , ..., c n
integral-invariants
n integrals of the system; the absolute

order one are evidently given by the formula
are constants, be
of
116-118]
where
(N1}
involve
2,
...,
and the
Nn
are any functions of (y1
y2
275
...,
y n ) which do not
relative integral-invariants of order one are

given
by the
formula
,
where
F is any function
By,
+N
Sy2 +...+
Nn Syn + SF),
r
of (xlt
the domain of integration
a?9
...,
xn
since the
t},
term SF vanishes when
is closed.
from this that any system of differential equations possesses an

number of absolute and relative integral-invariants of the first order.
It follows
infinite
The theorem of Lie and Koenigs.
118.
The preceding
results enable us to establish a theorem due to Lie* and

on
the
reduction
of any system of ordinary differential
Koenigs f
equations to
the Hamiltonian form.
Let
llf
= Zr
be the given system of equations, and
(r-1,2,...,*)
let
be any relative or absolute integral-invariant of order one of this

system,
where ,, ,, ..., j~ k are given functions of the variables we have seen in the
last article that an infinite number of such
integral-invariants exist.
:
Now
let
the differential form
be reduced to the canonical form
(PI,
,p n ,qi,
ft,
-,
qn
ty
are independent functions of (a^, a?a ..., xk \ in number not

greater than k,
be
zero
Let
u
u
.
be
a
set
of
other functions
J.
2
k
m
may
(u^
)
of (a?!, a?2 ..., a?t ), such that (u,, u 2 ..., u k _, n
q lt qz ..., qn p lt p*, ...,p n ) are
a set of k independent functions of (x 1} x.,, ..., xk ); and
suppose that the
,
and where
,
. . .
Archiv for Math, og Natur. n. (1877), p. 10.

t Comptes Eendus, cxxi. (1895), p. 875.
J The proof of the possibility of this reduction (which however requires in general the
solution of a number of ordinary differential
treatise on Pfaff s
equations) will be found in
any
problem.
182
276
[OH.
when expressed in terms of these k functions

system of differential equations,
as independent variables, becomes
dqr/dt
= Qr
dpr/dt
du,/dt
where (&, Q2
new
Qn
....
PI, P*,
(s
U,,
1}
...,
(r
= U8
Pn,
>
= Pr
= 1,
Uk ^ n )
1, 2,
2,
k-
n),
2w),
are functions of the
variables.
The expression
is
an integral-invariant
is
invariancy
since integral(relative or absolute) of this system,

as have been
transformations
such
unaffected
a property
by
performed and consequently

have the form
:
dq_r
is
= _3H^
~
dt
dpr
a function of (q lt q2
116) that the
dpr
= dH_
dt
where
follows
it
first
2n equations
=l
n)
dqr
p n t) only. The given

qn PI, p*,
thus reduced to a Hamiltonian system of
....
>
>
system of differential equations is

order 2n, together with the (k
2ri) additional equations
The Last Multiplier.
119.
Before proceeding to discuss integral-invariants of higher order than

those hitherto considered, we shall introduce the conception, introduced
Jacobi* in 1844, of the Last Multiplier of a system of equations.
by
dx
Xn X
dxn
dx2
_ dx
are given functions of the variables (xl ,xz ,...,x n x),

be a given system of equations: and suppose that (n
l) integrals of this
system are known, say
where (Xlt
2>
\Xi,
Jfr (T
<r\
n
U/ r
<Y
\i
=1
J.,
2
^,
1)
these equations let (xlt x 2 ..., xn ^} be expressed as functions of xn

then there remains only the solution of the equation of the first
From
and x
v
T
X%, ..., tCn} ICJ
,
order
dxn _ dx
TT
A-n
~Y~
**-
be effected; in which accents are used to denote that (xl} x

been replaced in
n and
by the values thus obtained.
to
Crelle
Journal, xxvn. p. 199, xxix. pp. 213, 333.
z>
...,a; n _ 1 )
have
118, 119]
We
shall
shew that
the integral of this equation is
fjf
7
X n dx) = constant,
(X dx n
M denotes any solution of the partial differential equation
where
+
anc?
277
(MX.) +
(* JO =
o,
denotes the Jacobian
The function
.M"
is
called the
as
Multiplier of the system of differential
equations.
For the proof of
this theorem,
we
shall require the following
lemma
If a system of differential equations
dxr jdt =
is
(r=l,
2, ..., ri)
(r=l,
2, ...,n),
transformed by change of variables into another system
dyr /dt=Yr
then
8
-5
r=1 c
where
denotes the Jacobian

o
To prove
this,
2
r=1
(aSi
xz
. . .
xn )
we have
xr
r=l dx r \ k=l
dy k
3
5
3
-îfc
/fc=l
_
-
v
-"
^-<
=i 3=1 k=i
y(v
oxr
In this expression the coefficient of
J.
fc
k dxA
--^Y
=
h
d* x
>-
r
^
dy.,dyk
dYk /dy
is
^ ^,
,.=i doo r
of
Yk
which
is
zero
dyk
different from, or
Also 3y,/9a?r = A rs /D,
equal to, k.
denotes the minor of dxr/dys in the determinant
so the coefficient
in the above expression, which is
or unity according as s
where
dys dyk l
is
^l rs
=l s=l
278
may
[CH.
be written
- ?
~
or
T8
r =i s ri
5(^1, #2,
dys dy
1 8Z)
y: ^
or
We
dyk
have therefore
dxr _
I
r =i
| ajj
dx r
Yk
k= i
k=l dy k
dyk
which establishes the lemma.
Now
in the original
problem write
d IT
Ci x*
/y/>
ci A
and consider the change of variables from
by the lemma, we have
so the quantity
M, which
M
satisfies
is
a solution of the equation
dt
the equation
8
/XA
All
a
which shews that the expression
the perfect differential of some function of xn and

theorem of the Last Multiplier.
is
Boltzmann and Larmor s hydrodynamical representation of
The theorem
siderations.
For
a;
the
this establishes the
Last Multiplier.
of the Last Multiplier may also be made apparent by physical con

simplicity we shall take the number of variables to be three, so that the
may
be written
dx dy
~
U
V
_ dz
W*
where
(u,
v,
279
119, 120]
w) are given functions of
z)
(#, y,
and the
last
multiplier
satisfies
the equation
+ ^(Mv} + ^
(Mw) = 0.
x
oz
3# (Mu)
dy
j-
This equation shews that in the hydrodynamical problem of the steady motion of
a fluid in which (u, v, w} are the velocity-components at the point (#, ?/, z), the equation of
is taken as the density of the fluid at the point (#, y, z}.
continuity is satisfied when
Now
let
(f)
(.v,
=C
y, z)
be an integral of the differential equations; then the flow will take place between the
thus we can consider separately the flow in the
surfaces represented by this equation
two-dimensional sheet between consecutive surfaces C and (7+8(7. The flow through the
;
gap between any two given points P and Q on C must be the same whatever be the
arc joining P and Q across which it is estimated and since the flow across arcs PR and RQ
together is the same as that across PQ, we see that the flow across an arc joining P and Q
:
must be
expressible in the form
So
f(Q) -f(P).
r the (variable) thickness of the sheet, so that r
denotes the velocity-component perpendicular to
if
Mrds
so that
is
we have
ds,
the perfect differential of a function of position.
that this expression can be written in the form
M
is
ds denotes an element of this arc, and

~
2
+ (90/9z) 2 } * 8C, and
(B</8?/)
if
= {(d^/dx) 2 +
a perfect differential
this is the
(v
dx
MSC(vdx- u
But it is easily seen

and consequently
dy)/(d<t>/dz) ;
u dy}
last multiplier for the case
theorem of the
con
sidered.
We
readily find for
ds the value
dz
$y
<f>x
so the
4>z
theorem really states that
an
is
equation
dx
dy
(f) x
dz
ID
integrating factor
= 0.
w
<>
(t>y
This, as was remarked by Appell (Comptes Rendus, CLV. (1912), p. 878),

form of the theorem of the Last Multiplier.
Derivation of an integral
120.
from two
is
a symmetrical
multipliers.
of the partial
and
Suppose now that two distinct solutions
equation of the last multiplier have been obtained, so that
/
I
**\
-rr
(X
"
f/T*
...
-rr
ulT
-i
C/-A
1
1
til*
fi
differential
o ~y
^ ~v~
QA.Q
~v~
+X n ~-U +X ^GF/ losrJ/+^

-TV
-r-r
+
+ X.,ûSC
of the
C-^- n
o ~v
J\
= 0,
+ ^r-f + ... +^-- + ^G3C
J
()
,.
(jQC
and
3
tr
d\
ÔX
HT
3^1
9^2
dX n
dX
-- =
OXn
OX
0.
280
[CH.
Subtracting these equations, we have
-3
but this
is
the condition that the equation
(M/N) =
log
shall
constant
be an integral of the system

Ct/tX/i
C&W2
X^
-3T 2
Ct t/
\AjJU
jj,
Xn
.A.
and we have therefore the theorem that the quotient of two last multipliers of
a system of differential equations is an integral of the system.
The reader who
is
acquainted with the theory of infinitesimal transformations will be
able to prove without difficulty that
if
the equation
Y f
admits the infinitesimal transformations
then the reciprocal of the determinant
A-l
At,
A. n
A.
ll
^12
ln
fen
is
a last multiplier.
Application of the last multiplier
121.
single
known
to
Hamiltonian systems
use of
integral.
If the system of differential equations considered
we have
evidently
^dXr /dxr = 0,
and consequently
is
a Hamiltonian system,
is a solution of the
M= 1
partial differential equation
which determines the
last multiplier
so the last
multiplier of a Hamiltonian system of equations is unity.

From this result we can deduce a theorem which enables us to integrate
completely any conservative holonomic dynamical system with two degrees of
freedom when one integral
is
known
in addition to the integral of energy.
Let the system be
and in addition
V(q l} q pi, pz)

follows that
2
^2i
= ^h =
djf
d_H
dpi
dp2
be known.
1
9 (Pi.
dp 2
~
fit
_dH _dH
dqj.
to the integral of
dpl
From
(dH
dq2
energy H (q
I}
q 2 PI, PZ)
,
h, let
an integral
the theorem of the last multiplier
dH
constant
it
is
another integral
where, in the integrand,
replaced by
their values in terms of q 1
integrals
and
But
V= c
if
for
281
120, 121]
be
2 are supposed to
obtained from the known
p and p
l
and
<?
V.
H= h
we suppose that the result of solving the equations

p and p is represented by the equations
and
then we have identically

"""
dpi dc
dp 2 dc
dc
dp 2 dc
and therefore
df2
dH/dp,
8/1
dc
dc~d(V,H),
_-dH/dPl
d(V, H}
d(p ,p2 )
p2 )
so the theorem of the last multiplier can be expressed by the statement that
is
an integral.
This result leads directly to the theorem already mentioned, which

be thus stated * If in the dynamical system defined by the equations
may
dq r
_dH
dt
the integral
denotes
Pidq
+p
any
2 dq 2
dpr _
dH
dt
dp r
H (q
o\
c
1} q 2
(q^., q%, PI, PZ)
p l p^ h, and if
then
the
the
not
time,
integral
expression
involving
where p-^ and p2 have the values found from these integrals,
of energy
is
other
,
of a function 6 (g n q2
integrals of the system are
is
dqr
the exact differential
7\fi
^dc
h,
c)
and
the
remaining
o/3
constant,
and
-j
dh
=t+
constant.
This amounts to saying that if any singly-infinite family of orbits is

selected (e.g. the orbits which issue from a point qi = ct 1} q2 = 2 ) which have
*
This theorem
is
really
an application
of the
well-known method for the solution of a
partial differential equation of the first order, the equations of the
dynamical system being
As a dynamical
the partial differential equation.
theorem, it was published for a simple case (motion of a single particle) by Jacobi in 1836
(Comptes Eendus, in. p. 59), and for the general case given here by Poisson in 1837 (J. de M.
the equations of the characteristics of
n. p. 317)
and Liouville
in 1840 (J. de
M.
v.
p. 351).
282
[CH.
the same energy, so that to any point (ft, q2 ) there correspond definite values
of p and p.2 (namely the values of p l and p 2 corresponding to the orbit which
passes through the point q l} q 2 and belongs to the family), then the value of
l
the integral
and
(fto, #20)
p dq
+p
and F =
dq 2 taken along any arc joining two definite points
independent of the arc chosen.
(q u , q 21 ) is
To complete the
proof,
we have on
differentiating the equations
H=h
c,
dpi 9ft
9ft
9/>
9ft
and consequently
9(F,
#)
d similarlv
iri
But
since
V= c
is
an
^-
9 gi
ft
3F.
3F
3F.
+ ^- q* + v- P =
9
2
ga )
.ST)
0,
dp,^
g2
(
9 (ft, Pi)
--r=0,
^T)
+ 3(F,
^7
,
9 (ft, PS)
M-^ =
and therefore
_PI
9pi
8(F,
or
we have
integral,
3F.
#)
9(F,
_
- 3(p
9/;
0.
oft
9ft
This equation shews that fidq +f2 dq2 is the perfect differential of some
h, c): and the result derived above from the theory of the
l
function 6 (q ly q2
last multiplier
Moreover,
shews that d&/dc
constant
is
an
integral.
we have
and therefore
3F
3F,
5
dq
-7.
3(F, //)
9
But obtaining
we have
9/1/3 A
(p a PI)
,
and df2/dh in the same way as
found,
dh
3(F,
J3
df2
9A
3(F,
9/l/9c
and
3/a/3c
were
121, 122]
dt
Consequently
or
?)f
283
PI/*
~j dq
-f
dq.2 ,
= W + constant,
JTT
on
which completes the proof of the theorem.

In the problem of two centres of gravitation ( 53), if (r, /) denote the
the angles formed by r, / with the
}
Example.
radii vectores to the centres of force, and (d, 6

line joining the centres of force, obtain the
integral
r 2 r 2 0ff
- 2c (p cos 6
+/* cos d
= constant,
and hence complete the solution by the above theorem.

122.
whose
Integral-invariants
order
is
to
equal
the
order
of the
system.
The theory of the last multiplier of a system of differential equations is

connected with that of the integral-invariants whose order is
equal to the
order of the system.
doc
Let
where (Z n
Z)
...,
k)
of (d, c2
integral-invariant,
(c l5 c 2
ck }
..., a*, t).
MQX, ox,,
6xk
where
a function of the variables.
is
be any set of constants of integration of these

equations,
..., xk ) can be expressed in terms
Then we have
.I[
>
\ Oj
G2
and therefore the condition of

integral-invariancy
~JI
dt
M 8(^,^2,
^~
"
is
...,xk y\
dt
8(c 1(
c 2) ..., ck )
M r=l
or
dt
C l , c 2 , ...
..., a
d(c1}
.,,...,
.
= i CXr
which shews that
*)
by solving the equations, (x l} xz
so that,
(r = 1 - i
fff
...
JJJ
may be an
are given functions of (xl} xz ..., xk

t), be a system
and let us find the condition which must
of ordinary differential equations

be satisfied in order that
Let
~s
M must be a last multiplier of the system of equations.
284
[CH.
This result gives immediately the theorem that for a dynamical system
is determined by the equations
whose motion
dt~dp
where
is
any function of
theorem is of importance
is
...,
qn
dq r
p p
lt
...,
pn
For a system with two degrees of freedom,
Example.
Shew
dt~
t),
the expression
This
since in this case unity is a last multiplier.
in the applications of dynamics to thermodynamics.
solved for
(q 1} q 2
dpr__dH
_?)H
dq,.
p take the form
h,
to, Pi, Ps,
when
the energy-integral
A)+_pt = 0.
which correspond
that, for trajectories
let
to the
same value of the constant
of
energy, the quantity
is
independent of
and
also of the choice of coordinates
and hence shew that the

motion
trajectories of the problem can be represented as the stream-lines in the steady

of a fluid whose density is
Redaction of differential equations
123.
Another question
applied
is
to
the following
to the
Lagrangian form.
which the theory of the last multiplier can be

To find under what conditions a given system of
ordinary differential equations of the second order
is
equivalent to a Lagrangian system
dL
d /dL\
jlla-H-5
dt
\dq r j
dq r
where
is
a function of (q lt q2
,..,
-0
q n q l} q2
,
(r=l,
...,
qn
2, ..., w),
t).
two systems are equivalent, the equations
If these
^
k =i
3- o
\oqr dq k
<ik
oqr dq k
Qk
z }
/
+ ^-5.
dq r dt
(r
1, 2,
n)
dq r
must evidently reduce to identities when the quantities qk are replaced by

the expressions fk and therefore the required condition is that a function L
;
shall exist satisfying the simultaneous partial differential equations
00
dqr ot
where
(q lt q2
...,
(r
= l,
i
2, ..., n),
dqr
q n ,qi,qz, ...,q n ,t) are regarded as the independent variables.
When
285
122-124]
= 1,
the question can be solved in terms of the last multiplier.
For the equation
satisfied
by
then
is
_
dqdq*
dq*
dq
dqdt
from which we have

_
_ ~
dq\dq
and therefore
if
we
~
f-.
/
write d 2L/dq 2
dq \dqdq
= M,
the function
dM
-but this
is
dqj
dqdt
the equation
satisfies
/%
the equation defining the last multiplier
of the system of
equations
and therefore wAen w
1,
/<e
determination of the function
reduces
to the
determination of the last multiplier of the system.
Case in which the kinetic energy is quadratic in the velocities.

the most important case is that in which each of the functions
fr
of a part Fr which is homogeneous and of the second degree in (q^, 2 ..., n ) and
j
q
124.
When
consists
n>l,
a part Gr which does not involve (j 1} j2

the equations
j0) an d
>
qr =
Fr +O
it is
required to determine whether
r =l,
2,
...,)
are equivalent to a system
dT
7
where T
homogeneous and of the second degree in (j h j2

-,?) and also involves the
2
-, ?), and (&,
..., $ ) are functions of (g- 1? j 2
.... q n ) only.
is
variables (^, ^ 2
The value
of
is
ft
clearly not dependent on (6 1

?
consider the problem in which (G 1} 6r 2

a function T such that the equations
...,
ir
6r M .)
^j
..., 6- n ),
are zero,
i.e.
and therefore we can
the problem of finding
= Fr
(r=l,2, ...,)
are equivalent to the system
dT
d /dT
The condition
for this is the existence of a function
T satisfying the
partial differential
equations
"
Since
k is
<
homogeneous, we have 2 qs dFk /dqs = 2Fk and therefore

,
8=1
(r=1
*>
286
But
[OH.
we have
since dF/dq r is homogeneous,
and therefore
k
The equations
to be satisfied
T may
by
1
--
"=\
9j r
3^
consequently be written
n v*
fiF^
tiT
k UJ
dT _f\
UJ
(PT
u
L
T?
/.
cT\
or
~*
cq r Vfc=t 9yg dqk j
s =\
k= i Oq r dqk
9^9^ + 9T\
r rr
2 5
(r=1
"
%)
and evidently these may be replaced by the equations
dFt dT
i 2 ~-^
*k=i o^-^k
,1,
dT
+ 9~ =0
(r=l,
2, ...,
).
?r
that if the system of equations

Thus, writing/,, for (Fr + G r ), we have the theorem
qr
r consists of
w/tere
a part which
part which does not involve the
is
=fr
(r=l,
homogeneous of degree two in

is
velocities,
reducible to the
2,
...,),
the velocities
and a
form
then
must
be
an
integral of the system
+ Ô
il^pl
k= idq dq
dq
(,.1,
2,
...,).
In the problem of two centres of gravitation, the distance between the centres of
and the semi-major axes of the two conies which pass through the moving
and
have their foci at the centre of force are (q l q z }. Writing
particle
1.
force is 2c,
of motion are
shew that the equations
dqr
F-
= ôH
a<
where
and
/ni
2.
JT"
and
/u 2
ojo,.
dp r =
-T
a^
(>H
d^
r=
,
l>
_.
2),
are constants.
Shew that
j f jj
extended over the \n (n-l) combinations of the indices i and j,
where the summation
is an integral-invariant of any Hamiltonian system in which (qi, q 2 ..., q nt Pi, p^, ..., p n )
is
are the variables.
(Poincare.)
287
x]
3.
In the problem defined by the equations
dqr_dff^
dpr__dl[
dt~3p r
dqr
H= q
where
pl
2
q2 p 2 a?i +
bat
f)a
shew that
dt~
bq<?,
= constant
?i
is
an integral
121 obtain the two remaining integrals
and hence by the theorem of
(q^z
constant,
(log qi = t + constant.
4.
If
a last multiplier of a system of differential equations
is
dx-i
of which the equation

zi
is
known
shew that
and
integral,
indicate that
if
_ dx n _ dx
_ d.r2 _
A2
xn x = Constant
>
an accent annexed to a function of # l5 #2 ..., acn x is used to

in the function by its values found from this integral,
,
M /(cf/cxn
is
a last multiplier of the reduced system
dxn _i
dxn
-=-,=... =
If QI
xn has been replaced
dx,
5.
An
Constant, 6 2
= Constant,
...,
dx
=-7-
= Constant
6n
(Jacobi.)
are a set of integrals of the
equations
dx
shew that
is
dxz
dx
8(^,^ 2
0(X 1 #2
dxn
...,^)
---i
*)
a last multiplier.
Let
6.
(iii,
u2
...,
u n ) be n dependent variables, and let Tj,

denned by the equations
/2
...,
/ be a
set of
linear differential expressions
k=l
If (vi,v z ,
...,
v n ) are functions of
such that
an exact differential, shew that the functions (vj, v 2 , ..., vn ) satisfy a set of n linear
differential equations, which will be called the system adjoint to the system of linear
is
7r
(r=l,
2,
...,).
If J ^. denotes the expression

1
where
is
any given function of
(q l
j2
>
>
?>
?i>
?2>
>
qn
shew that the system
linear differential equations
is
adjoint to
Shew
itself.
that the converse of this latter theorem
is
also true.
(Hirsch.)
of
CHAPTER XI
THE TRANSFORMATION-THEORY OF DYNAMICS
125.
We
Hamilton s Characteristic Function and Contact-Transformations.

have seen* that the integration of a dynamical system which is
by quadratures can generally be effected by transforming it into

another dynamical system with fewer degrees of freedom.
We shall in
the present chapter investigate the general theory which underlies this
soluble
procedure, and, indeed, underlies the solution of
all
dynamical systems.
memoir on
in
Hamilton
which
was
to
the
Irish
optics,
Royal
Academy by
presented
the
there
introduced
were
afterwards
transferred
their
1824f
principles
by
The
origin of the
method
to be found in a celebrated
is
discoverer to the field of dynamics.
In order to follow Hamilton s thought, we must refer to the connexion

between dynamics and optics a connexion which is perhaps less obvious in
our day than in his, when the corpuscular theory of light was still widely
If a ray of light traverses an optically heterogeneous but isotropic
held.
refractive index at any point (a?, y, z} being
the path of
be determined by Fermat s Principle J, namely that the integral
medium, the
a ray
may
JJL,
(x, y, z)
ds
has a stationary value when the integration is taken along the actual ray
joining two given terminal points, as compared with neighbouring paths
joining them. If on the other hand we consider the motion of a free particle
mass in a conservative field of force where its potential energy is

and its constant of energy is h, the path of the particle may be
(IK, y, z},
determined by the Principle of Least Action ( 100), which in this case asserts
of unit
<b
that the integral

ff
i*
has a stationary value for the actual trajectory as compared with neighbouring
Comparing these two statements, we
paths joining the same terminal points.

*
Cf.
Chapter
III,
38-42.
t Trans. E. Irish Acad. xv. (1828), p. 69; xvi. (1830), pp. 4, 93; xvn. (1837), p.
J Cf. ray History of the Theories of Aether and Electricity, pp. 9-10, 102-3.
1.
The Transformation-Theory of Dynamics
125]
289
see that the trajectories of the particle in the dynamical problem are the
same as the paths of the rays in the optical problem, provided a suitable
correspondence
*-<*-*
up between the potential-energy function
set
is
in the one case
and the
refractive index in the other.
In the corpuscular theory of
light, this was regarded as furnishing the

of
the
explanation
optical phenomena, the ray of light being conceived as
a procession of
But the statement in itself is
rapidly-moving corpuscles.
true whatever
supplies a
This idea
When
different
the
be
hypothesis regarding light
adopted
and therefore
it
means
is
of connecting dynamics with the undulatory

hypothesis.
the starting-point of Hamilton s theory.
the undulatory hypothesis is adopted, we have the choice of two

methods of discussing the propagation of light mathematically
:
The
to consider rays, the second is to consider

wave-fronts.
latter method, which was introduced
by Huygens in 1690, may be
first
is
thus
explained.
it
Consider a wave-front, or locus of disturbance in an optical medium, as

Each element
t, having the form of a surface
exists at a definite instant
<r.
of this wave-front
may be regarded as the source of a secondary wave,

outwards
from it
so that at a subsequent instant t
the
propagated
disturbance originating in any point (x y, z) of the original wave-front will
extend over a surface. To obtain the equation of this surface, we observe
;
medium from an arbitrary

another arbitrary point (x y z ) depends only on the six
quantities (x, y, z, x y z) let it be denoted by V(x, y, z, x, y z }.
This
function V(x, y, z, x y z } was called by Hamilton the characteristic
function
for the medium in
A disturbance which originates at a point
question.
that the time taken by light to travel through the
(x, y, z} to
point
(x,
(x,
the original wave-front at the instant t will therefore at the

over the surface whose equation in the coordinates
z} of
y,
instant
extend
is
V(x,y,z, x ,y ,z
-t
(1).
Now
according to the principle of wave-propagation laid down by

Huygens, the wave-front which represents the whole disturbance at the
instant t is the envelope of the
secondary waves which arise from the
various elements of the original wave-front.
Call this
new wave-front 2;
and denote the direction-cosines of the normal to the wave-front

at
(x, y, z) by (I, m, n), and the direction-cosines of the normal to the wavefront 2 at the corresponding
point* (x, y z } by (V, m n ): these are the
<r
from
The point
(x, y, z}
W.
D.
(x
y z ) is said to correspond to (x,
touches the envelope S at (x y z ).
,
y, z) if
the secondary wave propagated
19
[OH. xi
290
z ) and (x, y, z) respectively, since in

Then since
to the wave-front*.
normal
an isotropic medium the ray is
V
to
2 is the envelope of the surfaces
points on a, the
corresponding
direction-cosines of the rays at (x, y
equation
dV
-=- doc
dx
must be
satisfied
by
all
dV, dV,
+ -^- dy + -^- dz = Q
i
dz
dy
values
those
the
of
to
correspond to directions in the tangent-plane
relation
ldx
ratios
a-,
i.e.
dx dy
which
:
dz which
satisfy the
+ mdy + ndz = 0.
Hence we have
1^? = 1LF = 1 8F
dx
dy
(2)
n dz
of the normal to the

Moreover, since (I m, n) are the direction-cosines
have
we
z
the
),
surface F at
point (x, y,
,
=
dx
Now
dy
n dz
a ray of light which passes through the point (x, y, z) in the direction
at time t passes through the point (x, y /) in the direction (I m, n )
1
m, n)
at time
(I,
and equations
(1), (2), (3),
together with the equation
p + m +n
/a
/2
=l
................................. (4),
z
are six equations, from which we can determine the six quantities (x y
behaviour
the
these
Thus
equations
by
I
m, n) in terms of (x, y, z, I, m, n).
in the medium is completely specified in terms of the single
,
of rays of
light
function V(x,
y, z,
x, y
}.
It will
be observed that they are not
differential
but that they give directly, in the integrated form, the changes in
equations,
after a finite interval of propagation through the medium.
any system of rays
in optics depend on the deter
It is evident therefore that all problems
function
s
characteristic
Hamilton
V(x,y,z, x ,y ,z } for the
of
mination
optical
medium
or system of
media through which the rays
pass.
the point of view of Pure Mathematics, we regard the change from

of variables (x, y z I m, n ),
the set of variables (x, y, z, I, m, n) to the set
to the surfaces 2, as
surfaces
the
from
or (to express it geometrically)
be
to
is
thus
regarded as determining a
a transformation. The function
into a new surface 2.
surface
transformation of space which changes any
a
touch at
point, the corresponding
It is evident that if two surfaces a and
on this
transformed surfaces 2 and 2 also touch at the corresponding point
From
<r
<r
</
account the transformation has been called by
Thus any function V(x,
y, z,
x, y
defines
S.
Lie a contact-transformation.
a contact-transformation, which
For simplicity we are supposing that the medium, though optically heterogeneous,
medium.
Hamilton considered also the more general case of a crystalline
isotropic.
*
is
125]
transforms any wave-front

a-
by propagation
<r
into the wave-front
medium
through the
which
291
derived
is
in the interval of time
from
t.
A simple example of a contact-transformation is the well-known geometrical trans

formation known as reciprocation. In order to find the reciprocal of any given surface
with respect to a given surface, we correlate to every point (#, y, z) on
a plane, namely
o>
<r
the polar plane of (x, y\ z) with respect to the quadric. When the point (x, y, z) takes all
possible positions on the surface a, the plane envelopes a surface 2, which is the reciprocal
of a-. The transformation from o- to 2 is evidently a contact-transformation.
In this
case Hamilton s function
(
v*
linear with respect to (x, y,
is
and
z)
also with respect to
}
)
Proceeding now with Hamilton
problem, equations (2) and (3)
may be
written
Kl
dV =
ox
= *m
8F
\1H
^,
8F =
-^r-f
Ml
where K and \ are quantities not as yet determined.

For the equations may be written
readily found.
dV =
Now
increase
K(ldx
oz
oy
+ mdy + ndz) +
They can however be
+ m dy + n dz )
\(l dx
(5).
by proceeding a small distance ds along the ray at (x, y z }, we

But if the
by the time which light takes to travel along ds
,
units are so chosen that the velocity of light in free aether
is
unity, then
the velocity of light in the medium at (x, y z) is 1/X, where fjf denotes
the refractive index at this point. Thus the time taken by light to describe
,
ds
is
[ids
or p!
(I
+m +
2
-}
ds
or //
this with equation (5), we see that

denotes the refractive index at (x, y,
(I
A.
dx
//,
+ m dy + n dz
Thus Hamilton
z}.
).
Similarly K
s
Comparing
where p
JJL,
general formula
becomes
dV = p
If
we
(I
+ m dy +
dx
n dz
(Idx
/j,
+ mdy + ndz).
write
fjd
/Jim
fin
77,
yt/,7
fji
m =v
/j!n
"
this takes the form
The
quantities (f, 77, ), (

??
) were called by Hamilton the components
normal
slowness
of
of propagation of the wave at (x, y, z) and (x y z)
,
respectively.
Consider now the particular case in which the interval of time (t 1)

a- and 2 of the same wave-front is
very small
between the two positions
192
292
denote
by A. In
Write
it
the
this case
contact-transformation
is
[OH. xi
said to be
infinitesimal.
=-MtAf,
7/=77
(6)
= + wA,
v&t,
WA*
V=
Then equation
............... (7).
j
becomes
dW.At = (J; + u&t) (dx + da A) + 0? + v&t) (dy + dp A)

+ (f + wA) (ek + dy A) -vjdy= u&tdx + v&tdy + w&tdz + %dda. + rjAtd/3 +
dW = udx + vdy + wdz + %da.+ qd/3 + %dy,
d (go. + ?7/3 + 7
udx vdy wdz.
W) = otdj; + /3dtj + yd
.
<&c
%dz
%Atd<y,
or
or
Thus
if
we denote the
function fa
expressed as a function of x, y,
z,
g,
dH=adt; + ftdrj +
Now
+ rj/3 + 7
TF by iT, and suppose
we have
f,
77,
ydûdx
evidently, from (7), in the limit u
wdz
vdy
becomes
d~
dt
............... (8).
a becomes
dx
-y,
dt
etc.
Thus we have
dx
,.,
dy
dz
-..,
-.
so the rates of increase of the six variables (x, y,
dn
z,
%,
- dt
)
77,
-.
are given
by the
equations
dx_
dH
dy_
dt
dt
dH
drj
dz_
dt
d_H
/m
(Ph
dt
dx
dt
dy
dt
dz
a Hamiltonian system of equations, such as occurs in dynamics.

it may be regarded as representing an in
finitesimal contact-transformation, that is to say, the motion of a wave-front
from one position to a position indefinitely near it. The integrals of this
Hamiltonian system are the equations (1), (2), (3), (4) above they represent
a finite contact- transformation, that is to say, the motion of a wave-front
from one position to the position which it acquires after a finite interval of
and
this is
Our
investigation shews that
Thus we see how by using the ideas of the undulatory theory of light,
Hamilton was able to obtain an integrated form for the differential equations
of dynamics, depending on a single unknown function.
time.
126.
Contact-transformations in space of any number of dimensions.
The
rest of the present chapter will be concerned with the application of

Hamilton s ideas, described in the preceding article, to the general case of a
dynamical system with any number of degrees of freedom, and the connexion
125, 126]
of the results with certain theorems
due
to
293
Lagrange, Poisson,
and
Pfaff,
Jacobi.
We
shall
define a contact-transformation in w-dimensional space,
first
this purpose a generalisation of equation (6) of the last

q2 ..., q n p^ p.2 ...,p n ) be a set of 2/i variables, and let
for
using
Let (q 1}
article.
2/i other variables which are defined in terms of them

by 2n equations.
If the equations connecting the two sets of variables are such that the
be
differential
form
+ P. dQ +
PidQj
+ Pn dQ n -
...
when expressed
p^ - p dq
2
- p n dq n
...
in terms of (qlt q2 ..., q n ,pi, PZ,

p n ) and their differ
the
differential
of
a
of
function
entials,
perfect
(q 1} q 2 ..., q n p l p 2 ..., p n ),
then the change from the set of variables (q l q.2 ..., q n p 1} p2 ..., p n ) to the
is,
other set (Q 1}
It
Q2
...,
Qn
P P
1?
be observed that this
may
is
...,
Pn
called a contact-transformation.
) is
form from the definition which
different in
is
most
convenient when contact-transformations are studied with a view to their applications in

geometry and in the theory of partial differential equations the latter definition may be
:
stated thus
?2,
(?i>
a contact-transformation
qn ,Pi,Pn,
-,
-,
Pn,
another
to
z)
a transformation from a set of (2w+l) variables
is
set
(# l5
@2
...,
Qn
P P
l5
Pn
...,
Z},
for
which the equation
dZ- PdQ - P2 dQ2 -... - Pn dQn = p (dz-p dq -p2 dq 2 - ... -p n dqn

l
where p denotes some function of
is satisfied,
(q 1
q2
...,
qn
p lt p z
...,
pn
z).
If the n variables (Q ly Q2 ..., Qn ) are functions of

(q lt q2 ..., q n ) only,
the contact-transformation from the variables (q 1} q 2 ..., qn p l ..., p n ) to the
variables (Q 1 Q2
n ) is called an extended point-transformation,
Qn 1}
,
. . .
. . .
the equations which connect (q 1} q 2 .... q n ) with (Q l}

case said to define a point-transformation.
,
From
the definition
it is
...,
Q n } being
in this
clear that the result of performing two contactis to obtain a

change of variables which is itself
transformations in succession
a contact-transformation.
(?!,
>
<?2>
q n ,pi,
>p
n)
It is also evident that if the transformation

to
(Qi,Q
...,Q n PI,
,
..,
Pn
is
from
a contact-trans
P P
formation, then the transformation from (Q n Q2 ..., Qn

1}
2
n ) to
...,
is
also
a
this
is
z
n
(qi,q
,q ,pi,p2, --^Pn)
contact-transformation;
generally
,
expressed by saying that the inverse of a contact-transformation is a contactThis, together with the foregoing, shews that contact-trans
transformation.
formations possess the group-property.

Example
is
1.
Shew that the transformation
294
[CH. xi
In this case we have
PdQ -pdq = (2qfi sin p {(2q) ~^cospdq- (2qfi siupdp} -p dq

= d(q sinp cos p qp),
which
a perfect differential.
is
Example
2.
Shew
that the transformation

1
is
3.
Example
Shew that the transformation
(P= 2(1+2^ cos p) q^ sin p,

is
We
shall
now
obtain the
explicit analytical expression of a contact-
transformation.
Let the transformation from variables (q 1} q 2

q n p ..., p n ) to variables
Pn } be a contact-transformation, so that
Qn, P\,
,
(Qi) Qz,
r=l
where
dW
From
(qi, q*,
is
(P r dQ r -p r dqr ) =
a complete differential.
the equations which define (Q lt

,
",
qn ,pi
pn)
it
maybe
possible to
...,
Qn
eliminate(P 1
1}
...,
Pn in terms of
Pn p
pn
. . .
completely, so as to obtain one or more relations between the variables
(Qn Qa
let
the
number
flrtei, 2,, -., q n
Qn,
>
of such relations be
,
k,
Qn
qi,
and
let
qn )
them be denoted by
Q) =
....
(r=l,
2, ..., A)... (A).
The meaning of these relations may be illustrated by reverting to the geometrical

theory of contact-transformations in ordinary three-dimensional space, when there are
three cases to consider
:
(a)
There
may
be only a single relation between the
Q(x,y,
When
s,
new and
old coordinates, say
x ,y /)=0.
,
are given, this equation, regarded as the locus of a point (a/, y z

represents a surface so that each point (#, y, z) is transformed into a surface, which we
may call an Q-surface and any arbitrary surface is transformed into a surface 2 which
(#, y, 2)
~),
<r
the envelope of the Q-surfaces corresponding to the individual points of

125.
general case, and is the only one we considered in
is
(/3)
There
may be two relations

Q : (as, y, 2, x y
,
If (x, y,
(x, y, z)
is
er.
This
is
the
of this kind, say

z")
= 0,
Q2
y,
*,
/,
=0.
are given, these two equations in (a/, y , z ) represent a curve

transformed into a curve, which we may call a .fiT-curve and
z]
so each point
any arbitrary
surface cr is transformed into a surface 2 which

sponding to the individual points of cr.
There
(y)
y,
flj (a?,
in
295
12(>]
be three relations of this kind, say
may
z,
the envelope of the .ff-curves corre
is
x y
,
= 0,
O 2 (x,
y,
x y
z,
= 0,
Q 3 (.r,
y,
z,
x y
,
= 0,
which case each point (x, y, z) is transformed into a point (x y z }, and any arbitrary
cr is transformed into a surface 2 which is the locus of the points corresponding to
,
surface
the individual points of
cr.
Since the variations (dq lf dqz
dq n dQ lt
...,
2 (P r dQ r - p r dqr ) =
dW
...,
dQ n )
in the equation
r=l
are conditioned only
by the relations
(r=l,
2, ...,&),
we must have
dW
where (X^Xg,
q
equations to
.of (q
l}
an,
is a function
undetermined multipliers and where
are (2n + k)
and
The
n
(B)
..., Q ).
Q,, Q
equations (A)
determine the (2/i + k) quantities
...,
...,Xt ) are
qn
(Qi,
in terms of (q lt
qn
Qn,
. . .
Xi, ..., Xfc)
These equations
p n ).
Pn>
PI>
may therefore
be
regarded
as explicitly formulating the contact-transformation, in terms of the functions

(W, i( n.2 ..., n&) which characterise the transformation.
Conversely,
(<?i,
?2
>
if (
qn, Qi,
W,
-..,
n H
if
Qn),
(Qi,
-,
are denned in terms of (q l q2

,
9TF
? =-^
n4
dq r
n,
Qn, Pi,
...,
an,-
qn
...
-- Xi ân,
dq r
any (k +
and if
are
where k ^
..-,
p 1}
...,
+X
...
Pn, X 1;
1) functions of the variables
...,
fc
p n ) by the equations*
90
fc
an,
X -r-
Vr
-L
Z
>
,
)>
dq r
* These
equations were first given in Jacobi s Vorlesungen iiber Dijnamik (1866), p. 470, where
their utility in the transformation of partial differential equations of the first order (to which
of contact-transdynamical problems can be reduced) was indicated. Their place in the theory
formations was pointed out by Lie.
The Transformation- Theory of Dynamics
296
then
the
transformation
Pn
...,
1}
from
(q lt q z
Q2
(Q 1}
...,
Qn
the expression
for
to
q n Pi, ...,p w )
...,
a contact -transformation
) is
[CH. xi
2 (P r dQr -p r dq r )
r=l
dW, and
becomes, in virtue of these equations,
Q = (Zq)k~
If
Example.
TF= Q
...
- F$2 )* - q arccos
(<2qk
is
The bilinear covariant of a general
Now
let (x 1}
x2
#) be any
...,
a perfect differential.
so that the transformation from (^ p} to ($, P)
127.
is
cosp,
P= dW
where
so
set of
differential form.
and consider a
variables,
differential
form
X dx
l
where (X J}
kind
is
...,
Xn
+ X dx2 +
2
+ X n dasn
. . .
denote any functions of
(x\,oc z ,
called a Pfaff s expression* in the variables (x1}
..,#); a form of this

xz ..., xn ). Let this
,
expression be denoted by 0&, and write
where 8
8dd
is
d6$
= & (X dx + X doc + 4- X n dx n ) d (X 8^ + X 8x +
= 8X dx + + 8X n dxn + X 8dx + ... + Xn 8dxn
l
. . .
Using the
relations
dX r
-=
dx +... +
dX r
OX-^
8x1
8dxr = d8x
are independent, and replacing
X n Sxn
X d8x
l
...
Xn d8xn
which exist since the variations d and 8
r>
by
6x 1
-^
...
OX-
dX ~
+ -^-r oxn
respectively,
OXji
we have
dXn 8xn
...
dX-r
dx n
dX r 8Xr
(]Xfi
dX
d0s= 2
8dd
where
Then we have
the symbol of an independent set of increments.
y denotes the quantity dX{/dxj
Let (y-i, 2/2 ..., yn ) be a new

transformation let the
some
by
,
dXj/dxj.
set of variables derived

differential
from (x lt
x.2 , ...,
xn }
form when expressed in terms of
these variables be
Fjefyi
* Pfaff s celebrated
+ Y dy2 +
2
...
+ Yn dy n
memoir on these expressions was presented
Abhandl. Akad. der Wiss. 1814-15,
p. 76.
to the Berlin
Academy
in 1815
126-128]
297
- dJ}/3y< be denoted by by. Then since the

the quantity
86
dd
has
d
s
expression
obviously the same value whatever be the variables
in terms of which it is
expressed, we have
and
let
9P</<?$
2 S
i
=l
j=l
The expression taydxi&K)
= 1 j = i bijdyibyj.
on account of this equation, called the
is,
bilinear covariant of the form
= 2 S
ctijdtCiSxj
dxr
The conditions for a contact-transformation

expressed by means of
128.
the bilinear covariant.

(Qi, Qz,
Qn
PI,
.-.,
Pn
be variables connected with (q lt q2
...,
n
<ln,pi,
-.ipn) by a contact transformation, so that
by an exact
r=l
P dQ
r
r=i
differs
from
differential.
It is clear
from the last article that the bilinear covariant of a differential

not affected by the addition of an exact differential to the
form, since
it
depends only on the quantities dX ifoxj - bX}fixi which are all zero when
the form is an exact differential: and we have shewn that the bilinear
form
is
covariant of a form
transformed by any transformation into the bilinear

It follows that the bilinear co variants of
is
covariant of the transformed form.

n
the forms
r=l
n
r
dQ r and 2 p r dq r
r=l
are equal,
^(8P dQ,-dP
8Q r ) = 2
so that if the
transformation
(qi,q^, ...,qn ,pi,
is
i.e.
that
(8p r dqr
-dqr 8p
from
...,^)
to
(Q l} Qz
...,
Qn ,P
a contact-transformation, the expression

n
2
r=l
is
(8p r dq r
dqr 8p r )
invariant under the transformation.

Example.
we have
r );
For the transformation denned by the equations
Q = (2q)* k~
4 cos
rfP-(2g)~ijt*sin
^cospdjz
....
Pn
298
By
multiplication
[OH. xi
we have
2
dP8Q-8PdQ=-sin*p(d<28p-8qdp)+cos p(dp8g-8pdq)
= dp8q
8pdq,
and consequently the transformation
is
The conditions for a contact-transformation in terms of Lagranges
129.
bracket- expressions.
We
shall
now
give another form to the conditions that a transformation

n)
Qn 1
p n ) to variables (Qlt Q 2
from variables (q l} q 2
q n p lt
may be a contact-transformation.
,
. . .
. .
. .
If (q l q^, ...,q n ,pi, ...,p n ) are any functions of two variables (u, v) (and
possibly of any number of other variables), the expression
,
r
(dq tyr
r =i
is
called a
symbol
If
dpr
dqA
du dv )
\du dv
Lagranges bracket-expression*, and
usually denoted by the
is
[u, v].
now
(Qi, Qz,
(q 1} q%,
Qn, PI,
-,
p l}
2
r=l
we can
(dp r q r
Bp r dq r )
dp r by
replace
and similarly
for
the other quantities
2
r=l
(dpr &qr
- Bp r dq r ) = S
k,
we thus
[u k
obtain, on collecting terms,
u{] (diii8u k
- Suidu k
is
,
taken over
(&, Q
2>
...,
n>
1}
...,
Pn)
is
pairs of
all
Pn)--
if the transformation from the variables (q l} q 2
to the variables
),
where the summation on the right-hand side

variables (uk Ui) in the set (Q 1} Qz ..., Qn PI,
But
2n variables
...,p n ) are any functions of

Pn), then in the expression
qn
...,
qn, PI,
>
a contact-transformation,
Pn)
we
have
I (dp^r -
r=l
Bprdq,)
= I (dP r SQr - 8Pr dQr
),
r=l
this holds for all types of variation 8 and d of the quantities

with the above equation, we have therefore
and
i,P t ] =
0,
[Qi,
&] =
[Q*,P*]=0
[Qi,
PJ=1
=
(i,k
comparing
(i,k
= I,
2,
l,2,
...,
i$k),
(t- 1,2,
...,>
Lagrange, Mem. de VInstitut de France, annê 1808: reprinted Oeuvres,
vi. p.
..,),
713.
128-130]
These
regarded as partial differential equations which must be
be
may
by (qlt q2
satisfied
299
...,
qn
...,p n ), considered as functions of
p 1}
Q, ...,Q,
(Qi,
in order that the transformation
from
PI, ...,Pn )
one set of variables to the other
may
be
These equations represent in an explicit form the
conditions implied in the invariance of the expression
Poisson
130.
We
shall
8p r dq r ).
(dp r Sq r
r=\
s bracket-expressions.
next introduce another class of bracket-expressions which are
intimately connected with those of Lagrange.

If u and v are any two functions of a set of variables
(q 1} q 2
Pi,
Pn), the expression
...,
qn
/8w
I
is called
the Poisson
du
dv_
dp r
r =i\dq r
dp r
dv_
dqj
denoted by the symbol
...,
between the Poisson-brackets (u r u s )

,
connexion we shall
We
now
investigate.
2 (,,,)[,..]- 2 2 I
Now
(2?t*-lS
dpi
t=\i=ij=i \oqi dpi
multiply out the right-hand side, remembering that

o
t=i oqi
are each zero if
$ j and unity
ou t
if i
t=l
=j
du dPj
^
2
^
t=ioqidu
du t d
t=l dpi d
are each zero
^
S
and
pi
and that
.
the equation becomes

%?
\ r
-i
(%, w,) [n
t>
t=i
and consequently
wj
2n
2
t=i
(w t u r ) [u t , Ug]
,
when r ^
s,
2n
while
<=i
(w t
wr )
[w
<t
M,.]
have
t=\
is
some connexion
and the Lagrange-brackets [u r us ]
this
and
um ) are 2n independent functions of the

so that conversely (q1} q2
q n p lf
pn )
variables (q l} q2
g n plt ..., p n ),
are functions of (u 1} u.2 ..., u 2n ).
There will evidently be
.
v,
(u, v).
Suppose now that (u 1} u 2

,
u and
bracket-expression* of the functions
= 1.
,
Poisson, Journal d$ VEcole poly tech. vin. (Cahier 15), (1809), p. 266.
The Transformation Theory of Dynamics
300
But these are the conditions which must be
[CH. XT
satisfied in order that the
two determinants
!,
[u2
[U 2n
u^
[u 2
M,]
...
u zn \
\_U 2n
U-^ J
and
[U lt U2n ]
(U 2
U2n \
(Ûi
U2 )
...
\U2n
Uvn)
(U 2n
U2 )
may be reciprocal, i.e. that any element in the one should be equal to the
minor of the corresponding element in the other, divided by this latter
determinant the product of the two determinants being unity and thus the
connexion between the Lagrange-brackets and the Poisson-brackets is expressed
by the fact that the determinants formed from them are reciprocal.
;
Example
1.
Example
2.
of (?D
<?2,
>
If /,
-..,
where the summation
^2>
are functions of (/l5
be
...,
qz
fk
),
...,
qn p1
,
...,
p n \ shew
that
which in turn are functions
p), shew that
taken over
let
>
(Q lt
c[n,p\,
...,
Qn
-,pn)>
satisfied in order that the
may
f2
(j t
all
combinations
fr f
Now
(î)
is
any three functions of
The conditions for a contact-transformation expressed by means of
131.
Poisson
If F,
$n,Pi,
^ are
<,
lt
we
shall
(Pi, Pj)
are expressed
may
= 0,
denote 2 n functions of 2n variables
shew that
the conditions
which must be
one set of variables to the other

from
(Qi, Qj]
(, j
129 that the conditions
for
1, 2,
n),
by the equations
f
[Pi, Pj]
Hence the
Pn
transformation
For we have seen in
...,
= 0,
[Q {
Qj]
relations
=0
(i,
1, 2,
.,
n),
130-132]
of the last article
301
become
while the relations

2n
the theorem
is
(ut,
u r }[u
so that the Poisson-brackets of

variables are equal.
2.
nr]
thus established.
Example 1. If (Qlt Q2 ..., Qn Pl} ...,

by a contact-transformation, shew that
Example
If
Pn
are connected with ( gi

,q 2 ...,q n Plt
,
any two functions

*
(&,..., Q n
and
^ with respect
are given functions of
( ?1
?2>
to the
..., ?(l ,
ft
two
...,
Pn )
sets of
...,p n \
and
satisfy the partial differential equations
(Qr,
^ )=
(r,=l,
2,
...,),
shew that % other functions (Plt P2 ..., Pn ) can be found such that the transformation
from ( ?] ?2 ..., ?n Pl ;;. f
plt ..., pw ) is a contact-trans
2
pn ) to (^
..., Q n
,
formation.
132.
/Lj e
TAe sub-groups of Mathieu transformations and extended

point-
Iran sformations.
If within a group of transformations there exists a set of transformations
such that the result of
performing in succession two transformations of the
set is always equivalent to a transformation which also
belongs to the set, this
set of transformations
is said to form a
sub-group of the group.
of
the
of
contact-transformations
sub-group
general group
constituted by those transformations for which the
equation
S
r=l
is satisfied.
They
Pr dQ =
r
is
evidently
2 p r dq r
r=l
These transformations have been studied

by Mathieu*.
are essentially the
transformations in
(ft,
g- 2
same as the transformations
...,
?n Pl ,...,p n
,
)"
by
called
"
homogeneous contact-
Lie.
In this case, we see from 126 that

(&, Q 9 ..., Q n plf ..., n ) are to be
obtained by eliminating (\ l} \ 2 ..., \ k from the
)
(2n + k) equations
,
Journal de Math. xix. (1874),
p. 265.
[CH. xi
302
From the form of these equations it is evident that if (p j} p2 ...,p n ) are

each multiplied by any quantity //., the effect is to multiply (P 1} P2 ..., n )
must be homogeneous of the first
each by yu,; and therefore (Pj, P2 ...,
?l )
,
degree (though not necessarily integral) in (p 1}
pz
...,p n ).
sub-group within the group of Mathieu transformations is constituted

by those transformations for Avhich (P P2 .. P n ) are not only homogeneous
of the first degree in (p l} p2 ...,p n ) but also integral, i.e. linear, in them; so
that we have equations of the form
: ,
Pr = 2
k=\
p k frk (q 1} q2 ,..., qn )
(r=l,2,
. .
n}.
Substituting in the equation

n
P dQ
r
r=l
r=
l*
to zero the coefficient of p k
and equating
= 0,
rdgv
,
we have
(fcl,
dq k
r=l
so
(<?!,
qz
qn ) are functions of (Q 1}
...,
frk
Q2
Qn )
...,
only,
= dqk/dQr
2, ...,n),
and
(r,
k=l,2,
..., n).
kind are obtained by assigning

It follows that transformations of
n arbitrary relations connecting the variables (q l} q2 ..., qn ) with the variables
this
(Qi,
Q 2,
Qn),
and then determining (P lf
P = I
r
k=l
...,Pn )from the equations
(r=l,
Pt j2*
V^r
These transformations are extended point-transformations

n
Example.
r=l
P*S^=
133.
n
2
r=l
O0r
S hewtha,t
We
Pr dQr =
If
p r dqr
n
2, ...,n).
126).
dPr
lf*^-
"
Infinitesimal contact-transformations.
shall
now
consider transformations
in
which the new variables
Pn) differ from the original variables (q 1} q2 ... q n

n
p l} ...,p ) by quantities which are infinitesimal. Let these differences be
denoted by (&q lt Ag 2 ..., A# &p lt ..., Ap n ), where
(Qi, $2,
>
Qn, PI,
->
and A^
is
n>
an arbitrary infinitesimal constant
Qr =
qr
+ &q r =qr +
so that
<f>
132, 133]
and the transformation
Now
is
303
by the functions
specified
suppose that the transformation
Then
is
we have
r=l
where
is
some function
(P r dQ r -p r dq r ) =
of (q l}
q.2
...,
q n p lt
pn )
....
or
2
>=!
or
It is evident that the function
W=
U&t, where
?7 is
must contain
some function of
(^ :
qz
...,
qn
as a factor
...,
lt
p n ), the
writing
equation
becomes
Hence we have
M
/
r=l
r=l
= -dK(q 1}
q2
...,q n ,p ly ...,p n ) say,
and therefore
Thus
/te
mos<
general infinitesimal contact-transformation
defined by the
is
equations
where
is
an arbitrary function of
(q 1} q z
...,
qn
arbitrary infinitesimal quantity independent of (q 1}
The increment
ments
(q 1} q
...,
in
qn
...,p n }
p lt
q.2
qn
and
...,p n ),
...,
p 1}
...,
is
an
p w ).
..., q n p 1} ...., p n ) when its argu

are subjected to this transformation is
any function f(q lt q z
p 1}
_
r
dp r dq
or
on this account the Poisson-bracket (/, K) is said to be the symbol of the

most general infinitesimal transformation of the infinite group which consists
of
all
contact-transformations of the
2??,
variables (q 1}
q.2
...,
qn
p lt
...,p n ).
304
mew
The resulting new
134.
[CH. xi
of dynamics.
established in the last article enables us to extend to
The theorem
all
conservative holonomic dynamical systems, whatever be the number of degrees

125 for
of freedom, the conception which was formulated at the end of
For the motion
certain simple systems.
is
expressed
109) by equations of
the type
dqr
j-
dt
and from the
dff
__
dp r
dp r
dt
_ dH
fB
1 2
n)
dqr
we can
interpret these equations as

implying that the transformation from the values of the variables at time t
to their values at time t + dt is an infinitesimal contact-transformation.
The
last article it follows that
whole course of a dynamical system can thus be regarded as the gradual selfunfolding of a contact-transformation. This result is really a generalisation
of the statement that the paths of the rays in a pencil of light can be specified
by the gradual propagation of a wave-front. Taken in conjunction with
the group-property of contact-transformations,
it
is
the foundation of the
transformation-theory of dynamical systems.
From
this it is evident that if (q 1} q 2
in a dynamical system, and

values at some selected epoch
(a
l>
ot 2
qn
...,
...,
a.
...,p n ) are the variables

are their respective
p lf
{3 1} ..., /3 n )
t
t Q the
equations which express (q qz
qn
a
in
of
which
terms
constitute
the
@
n
...,
lt
2
n
1}
...,
j3
(a
t) (and
Pn)
PI,
solution of the differential equations of motion) express a contact-transforma
tion from (!, or2 ..., ctn &, ..., /3 n ) to (q 1} q 3 ..., q n p lf ...,p n ); in this t is
,
l ,
>
. . .
regarded merely as a parameter occurring in the equations which define the

transformation.
Helmholtz
135.
s reciprocal theorem.
Since the values of the variables (q l} q.2 ..., q n p lt ..., p n ) of a dynamical

system at time t are derivable by a contact-transformation from their values
,
(!,
cr
2)
...,
an
at time
fi l; ..., fin)
we have (128)
where the symbols A and 8 refer to increments arrived at by passages from

a given orbit to two different adjacent orbits respectively.
Now
orbit
suppose that 8 refers to the increments obtained in passing to that

is defined by the values
which
at time
which
is
(!,
and
let
...,
&,
/3 2 , ...,
refer to the
&._!,
+S@
ftr+i,
>
fin)
increment obtained in passing to that orbit
defined by the values

(q l} qt
>
...,q n ,pi,
-, PS-I, PS
&Ps>
PS+I,
Pn)
134-136]
at time
,;
305
then the above equation becomes

&ps&qs=-8/3,-ky>;
the increment in
so
/3 1; ..., /3/--1, /3 r+1
...,
qs due to an increment in /3r (when a l cr2 ..., a n

fi n are not varied) is equal to the increment (with sign
,
reversed) in a r corresponding to an increment in. p s (when qi,q 2 ...,q n ,pi,

JOg-i, PS+-L,
p n are not varied) equal to the previous increment in /3 r
,
.,
This result can for many systems be physically interpreted, as was

observed by Helmholfcz*; for a small impulse applied to a system can be
conveniently measured by the resulting change in one of the momenta
p n and
the change in a r due to a change in p g can be realised in the

reversed motion, i.e. the motion which starts from some given
position with
each of the velocities corresponding to that position changed in
sign, so that
(p-i,
...
),
the subsequent history of the system is the same as its previous

history, but
performed in reverse order. We can therefore state the theorem broadly
thus the change produced in any interval by a small initial impulse
of any
:
type in the coordinate of
any other (or of the same) type, in the direct motion,

equal
change produced in the same interval of the reversed motion in
the coordinate of the first type by an equal small initial
impulse of the
to the
is
second type\.
Example. In elliptic motion under a centre of force in the centre, if a small velocity
8v in the direction of the normal be communicated to the particle as it is
passing through
either extremity of the major axis, shew that the tangential deviation
produced after
a quarter-period
velocity
8v,
is
p.
"dtf,
communicated
where
period an equal normal deviation
136.
Jacobi
into another
It
p.~
the constant of force.
minor
Shew
axis,
also that a tangential
produces after a quarter-
dv.
(Lamb.)
theorem on the transformation of a given dynamical system
dynamical system.
appears from
116 that
equations
dq r
-nr
dt
is
is
/*
at the extremity of the
if
Hamiltonian system of
= dH
^,
dp r
dt
dp r
differential
dH
=-3dq
(r=l,
2, ...,
w)
transformed by change of variables, the system of differential equations so

still have the Hamiltonian form
obtained will
dPr__dK
dQr_dK_
dt
dPr
provided the new variables (Q l}
is
*
Journal
t Cf.
w. D.
Q2
dt
...,
Qn
"**
dQr
ly
...,
Pn
are such that
(relative or absolute) of the original system.

fiir
Lamb,
Math.
c.
(1886).
Proc. Loud. Math. Soc. xix. (1898), p. 144.
20
306
kind
transformation of this
considered,
i.e.
it
in general,
is,
the
to
special
[on. xi
problem
given Hamiltonian system into another

will not necessarily transform any other arbitrarily
transforms the
Hamiltonian system, but it

Hamiltonian system
chosen
into a Hamiltonian system.

Among these
transformations however are included transformations which have the pro
perty of conserving the Hamiltonian form of any dynamical system to which

they may be applied these may be obtained in the following way.
:
We
have seen (115) that
a relative integral-invariant of any Hamiltonian system. Let (Q l}

PL ..., n ) be a set of 2n variables obtained from (q lt q 2 ..., q n
by a contact-transformation, so that
is
dQrr=l
r=\
dW
where
may
Qn
...,
...,p n )
p r dq r =
The equations which
denotes an exact differential.
transformation
p 1}
involve the time, so that (Q 1}
...,
Q P
n
define the
...,
1}
Pn
are
but in the variation denoted by d

functions of (q l q z
., q n ,pi,
Pn,
if t is supposed to
in this equation the time is not supposed to be varied
vary, the equation becomes
,
>
dQ r - 2 p r dq r =
U denotes
where
dW+ Udt,
r=l
r=l
some function of the
variables.
Now
the variation denoted by S in the integral-invariant is a variation

from a point of one orbit to the contemporaneous point of an adjacent orbit
if
therefore
we regard the
variables as functions of (a l
a2
...,
a 2n
t),
where
of
..., a 2n ) are the constants of integration which occur in the solution
the equations of motion, the variation & is one in which (a 1} a 2 ..., a 2n ) are
varied but t is not varied we have consequently, as a special case of the last
(a 1} a 2
equation,
SQr- 2 p r Bqr =8W,

r=l
r=l
r
and therefore
i
r
&Qr
J }=!
is
so the transformed system of differential

which (Q 1( Qz ..., Qn P1} ..., Pn ) are taken as dependent
have the Hamiltonian form and can be written
a relative integral-invariant
equations,
in
variables, will
dPI ~__ dK
dQr_dK_
dt
where
is
~dPr
some function of (Q lt Q2
dt
,
...,
Qn
dQr
,
Pi,
...,
Pn
t).
136, 137]
307
Hence a contact-transformation of the variables (q lt q2 ..., qn

Pn}
of any dynamical system conserves the Hamiltotiian form of the equations of
,
In the case of an ordinary
the system*.
system, in which (Q1}

Shew
Example.
is
...,
"change
of
variables"
are functions of (q lt q 2
Qn )
in the
...,
>
T>\->
dynamical
qn ) only, the
merely an extended point-transformation.
that the contact-transformation defined by the equations

q
= (2#)4 k ~ * cos
p = 2Q)% kt
P,
sin P,
changes the system
dq__3H_
dp =
dff
di~dp
dt
dg
H = \(p* + k q*\
z
where
into the system
dQ_dK dP__dK_
~dt~dP
where
dQ
"dt~
K=kQ.
137.
Representation of a dynamical problem by a differential form.
The reason for the importance of contact-transformations in connexion

with dynamical problems is more clearly seen by the introduction of a certain
differential form which is invariantively related to the problem.
Let any
be
form with (2n
differential
we have seen
127) that
its bilinear
independent variables (X, x2
1)
...,
covariant
+ l 2w+l
ddSaS
where a tj denotes the quantity (dXi/dxj

the form.
If
we equate
dXj/dxi), is invariantively related to

to zero the coefficients of Sx 1} Sx2 ..., Bx^^, we
,
obtain the system of (2?i+l) equations
Since the determinant of the quantities a fj is skew-symmetric and of odd

and these equations are therefore mutually compatible.
They are known as the first Pfaff s system of equations corresponding to the
order, it is zero,
differential form
2
r=l
dxr and from the mode of their formation are

,
variantively connected with it

made, the new variables (yl} y2
#2n+i)>
and
if
the differential
that
any change of variables is

..., y m+1 ) being given functions of (ar,, x2
...,
form be changed by this transformation to
;
2
This important theorem was
is
r=l
*
to say, if
in-
first
Yr dy
given by Jacobi, Comptes Rendus,
v. (1837), p. 61.
202
308
2n+l
and
if
& fl cfa/;
0,
be the
b^dyi
then this system
is
b ii2n+l dyi
F ety
ii
in the
(2n+
of
dxi
now
Consider
= 0,
qn
r,
Ui Z dxi
= 0,
;,
2?i+i
cî
= 0.
the special differential form
1) variables (g 1} q 2
>
equivalent to the system
2n+l
qn
...,
p lt
...,
pn
t),
where
is
any function
quantities a^, we
Pn,
Pfaff s system of differential equations of this differential
Forming the corresponding
t).
>
>
P\>
find that the first
form
...,
t=i
(ft,
0,
Pfaff s system derived from the differential form
first
^=1
1=1
[OH. xi
is
dp r
dt
;r
(r
dH U = A
-^dt
dq r
= 1,
(r=I,
()
2,
n\
2, ..., n),
dp r
t
~^di
Of these the
last
equation
is
a consequence of the others
and therefore the
system of equations can be written
dqr
dH
dt
dpr
-*
dp r
*
^^
dH
dt
.
/
iv*
IV
7?
dq r
but these are the equations of motion of a dynamical system in which the
It follows that the dynamical system whose
Hamiltonian function is H.
is invariantively connected with the differential
Hamiltoniah function is
form
__
Pidqi
inasmuch as
variables
(x
differential
which
is
+p
the equations of
l
x2
...,
x 2n
T)
dq2
...
+p n dqn
Udt,
motion of the dynamical system, in terms of any

whatever, are
the first
Pfaff s
system of the
form
derived from the form
Pidyi+ptdqt
by the transformation
variables (xlt
x2
from
..., x^,, T).
...
Hdt
+p n dq n
the variables (q lt q2
...,
q n p\,
,
-,pn ,t}
to the
The Hamiltonian function of the transformed equations.
138.
The
309
137, 138]
result of the last article furnishes another proof of the
theorem that
the equations of dynamics
_ dH
dq r
dt
dp r
conserve the Hamiltonian form under
and moreover
>Pn),
qn>Pi>
K of the system
it
dp,.
dH
dt
dq r
119
contact-transformations of (q l ,q.2 ...,

enables us to find the Hamiltonian function
all
thus obtained,
=
~df
For
(r
~dQ r
~dT~~
dPr
= l,2,...,n).
the contact-transformation be defined by the equations
let
(r
an
-, y~\
-i
an 2
~T~
,,
A>
where (n^
From
...,
dgv
an,
Xj-=
.*.
J_
A, ..., n),
(ral,
2. .... n).
^7*
>
d^
Pr dQ
identically
dq r + ^~r
2 I-
2 Xs 2
dQr }
(-5-^
which
in
<ferH
all
the variables,
are changed)
o
j
^ D j/a
Z
p r dqr = i Jr r d(^r -f
27
p
w
dqr
V"
7/^
Pr dQ
r
7^
/
I
ac
--_
TT 7j
= 2
Hdt
JTIT
4
a
(r + z
Jj
"
a TF
V
HTT
^ ^X
2
s
-_-
The
.-.
d v/
and hence (the symbol d denoting a variation

t,
k}
an*
A,
a^,.
we have
r=l
including
n,, TF) are any functions of the variables (q 1} q
these equations
p.r dq.r
X2 ^
u(/,.
an 2
an,
X;
-T-
.-.
U*%r
(j/j.
=1
perfect differential
does not affect the
dW
&
on the right-hand side can be neglected,

and
Pfaff s system of the differential form

hence the contact-transformation transforms the si/stem of equations
since
it
first
da^ =
to the
ciH
X.
^.
system
~
7~/
dt
where
dpr^_d_H
r\
T\
dP r
K=H
^~"
^^
dt
^
ot
K being supposed expressed in terms of (Q
2 X
x= i
lt
^s /-^
dQ r
Q2
-^
dt
...,
Qn
1}
...,
Pnt
t).
>.
/fr /.
310
Transformations in which the independent variable
139.
The
the whole set of (2n
1) variables (ql} q2
PI,
Qn>
>
>
qn
...,
p l}
, . .
pn
t)to
new
variables
Pn, T} by which any Hamiltonian system
dqr _
=
~dt
is
changed.
137 also enables us to determine those transformations of
result of
(Qi, Qz,
is
[OH. xi
dH
dp r
dp r
~dt
dH
_
=
~dq^
transformed into a system of the Hamiltonian form
dK
~dT~~dQr
dPr _
dQr _dK
dT~dPr
For
same thing
this is the
which change the
as finding the transformations
form
differential
+p
Pidq-t
where the variables
(q ly q 2
dq.2
...,
p lt
n>
...+ p n dqn
pn
...,
hdt,
are
A)
t,
equation
H(qi,
qt,
>
qn Pi,
t)
"-,pn,
connected by the
0,
into the differential form
dQ
-f
dQ +
where the variables (Q 1} Q 2
+ P n dQ n + kdT +
Qn, PI, PZ,
a perfect differential,
Pn, T,
connected by
k). are
the relation
p lt
But any contact-transformations

...,p n h) to new variables (Q l} Q
,
of the (In
2
Qn
...,
-\-
T,
2) variables (q 1} q2 ..., q n t,
n k) will satisfy
lf
3
...,
,
P P
when
the transformation has been assigned, the function

obtained by substituting in the equation
this condition
q*, ...,
q n ,pi, ...,p n
t
>
is
the values of (q ly q2 ..., q t, p 1} ...,p n h} as functions of (Q lt ..., Qn T,

Pj, ...,
n k), and then solving this equation for k, so that it takes the form
,
n>
K(Q Q
l}
Qn
...,
15
...,
7l
T)
+ k = 0;
the required transformations are thereby completely determined.
New formulation
140.
We
have seen
of the integration-problem.
137) that
dynamical system
dqr
new
which
is
any change of variables
= 3H
dt
the
if
dpr
differential equations will be the

derived from
Pidq!
by the transformation.
+p
dq2 +
==
~
...
made
in the
_dH
dt
dp r
is
dq r
first
+ pn dqn
Pfaff s system of the form
Hdt
139, 140]
Supposing that a transformation
which
the
becomes
variables,
P,dQ,
dT
where
found, defined by a set of equations
such that the above differential form, when expressed in terms of
is
new
is
311
(Qi, Qz,
the
is
Pn
dQ.
...+
differential
perfect
Qn, PI,
+P
t);
the
Pn dQn - dT,
some function of the
of
corresponding
first
Pfaff s
variables
of
system
is
equations
dQr = 0,
dPr =
(r-
1, 2, .... n),
and the integrals of these equations are
Qr =
P = Constant
Constant,
(r
1, 2,
n)
so the equations
qr=<l>r(Qi>Q...,Q n
,P
...,Pn,t}}\
l
constitute the solution

of the
Qn, PI,
Pn
dynamical system, when the quantities (Q 1; Q 2

are regarded as 2n arbitrary constants
of integration.
...,
The integration-problem is thus reduced to the determination

of a trans
formation for which the last term of the differential form becomes a perfect
differential.
1.
Shew that the transformation denned by the equations
= arctan
- arctan
P =\ arctan %L
2
a contact-transformation, and that it reduces the

dynamical system whose Hamiltonian
z
z
2
z
function is ^(pi 2
to the
z + \~ qi +\whose Hamiltonian
i
is
+p
function
2.
is
Qz
If (# 1?
!i
q.
dynamical system
x2
..., x.2n )
denote any functions of (q lt
-,
? n Pi,
>
./>),
an d
if
moreover a mn denotes dXml dxn - d nfixm D denotes the determinant formed of the
quantities amn A ik denotes the minor of a ik in D, divided by D, and u and v denote
arbitrary functions of the variables, shew that
if
8v
CU
312
Shew that
3.
for
any Hamiltonian system the

f
j j ...J8q
and
lff...J8Q
8q 2 ...8q n 8p l ...8p n
8Q2
...8Qn 8P l ...8P n
4.
xi
integral -invariants
extended over corresponding domains, are equal if (qi, q 2 ..., q n

are connected by a contact-transformation.
Qn, A;
(Qit
$2>
[OH.
Pi>
Pn) and
P")
Prove that the contact-transformation denned by the equations
= Xj
"(24k)
cos î + X2~
(2^2)
cos
Ai
= -X!~4 (2i)4 cos P +\ ~^ (2$a)"oosP|,

l
=
changes the system
dq_r
o?
dpr =
= dH_
_<W
dt
9pr
dq r
where
into the system
dQr _dK
dt~dP~r
dPr _
>
tt
dt~~dQr
where
JT-X^+X^,.
the original system.
Integrate this system, and hence integrate
(r=1
2)
CHAPTER
XII
PROPERTIES OF THE INTEGRALS OF DYNAMICAL SYSTEMS

Reduction of the order of a Hamiltonian system by use of the integral
141.
of energy.
We
have shewn in
42 how the Lagrangian equations of motion of a
conservative holoiiomic system can be reduced in order by use of the integral
of energy of the system.
shall require the corresponding theorem for
We
the equations of motion in their Hamiltonian form
this
may be
obtained as
follows.
Consider a dynamical system with n degrees of freedom for which the

Hamiltonian function
does not involve the time explicitly, so that
H+h=0,
where A
is
a constant,
is
the integral of energy of the system.
Let this equation be solved
The
differential
where the variables

equation
p.2 dq 2
+p
so that it can be written
+ p^dq. +
2
...+ p n dqn
is
hdt,
(qlt q2 ..., q
PI, p. ..., p n h, t) are connected by the
the differential form can therefore be written
dq 3 +
...
where we can regard
But the
p lt
form associated with the system

Pidq^
last
the variable
for
n>
2>
+p n dq n + hdt- K (p p
2
(q
1}
...,
qn p.2
,
..,
pn
...,
h, t)
pn
dqi
d
dq l
dp r
= d^K
dh
dp r = dK
d~ ~i
r
dq
q,, ...,
as the
differential equations corresponding to this
dqr dK
-r~ = ^
q n h) dq lt
(2n+
form are
1) variables.
(
(r=2,
137)
3, ...,
ri),
dq
dk_
~
dq l
The last pair of equations can be separated from the rest of the system,
since the first
2) equations do not involve t, and h is a constant.
(2>i
314
Properties of the Integrals of
The original
[CH.
xu
can therefore be replaced by the reduced
system
dqr_dK
7
dq
which has only (n
This result
is
~\
dK
dp r
dq
dp r
">
"
/)
dq r
of freedom.
1) degrees
equivalent to that obtained in
42, as can be
shewn by
direct transformation.
Example.
Consider the system

dt
a being a constant
which
dt
~dp r
where
dqr
these are easily seen to be the equations of motion of a particle
attracted to a fixed point with a force varying as the inverse cube of the distance
and q 1 are respectively the radius vector and vectorial angle of the particle referred to
the centre of force.
is
<?2
Writing
H=
h,
and applying the theorem given above, the equations reduce
system
dq<2_dK
~~
dp.2
dq l
dqi
dp 2
to the
dK
dq 2
where
K=
Since
does not involve q lt the equation
Constant
system, and we can therefore perform the same process again
is
:
an integral of this last

- k, we have
writing
and the system reduces to the single equation
ddL
the integral of which (supposing
If follows from
equations
This
Hamilton s partial
is
is
the equation, in polar coordinates, of the orbit
differential equation.
138 that
if
a contact-transformation defined by the
Pp = dW
Pr
-dQ r
where
-- *
q./
/u<
where e is an arbitrary constant.

described by the particle.
142.
ok
dq z
^~r
~T
= dW
Tqr
denotes a given function of (q 1}
q.2 ,
(r- 1,2,
...,
qn
Qlt Q
...,
...,),
Qn
t),
performed on the variables of a dynamical system defined by the equations
dqr_m
dt~
dpr
dpr __dH
~dt
dqr
is
315
Dynamical Systems
141, 142]
the resulting system
is
dQr = dK
dP r
dt
dK
-dQ r
dPr =
>
~di
K=H + dW/dt.
where
If the function
is zero,
the system will be said to be transformed into

is a
will be zero, provided
the function
Now
the equilibrium-problem.
function such that
*\
ft
i.e.
variables (5^
provided TF, considered as a function of the
the partial differential equation
^2,
,
<?n>
0>
satisfies
dW
^t+ H( qi
dW
dW dW
q2 ,...,
qn
,^,
^-,...,
\
=
-,*J
?
associated with the

partial differential equation
in 1834*, being the
Hamilton
was
published by
given dynamical system.
extension to dynamics of the partial differential equation which he had
This
called
is
Hamilton
It
discovered ten years previously in connexion with optics.

solution con
i.e. a
Suppose that a complete integral of this equation,
is known.
additive
to
the
in
addition
constant,
constants
n
arbitrary
taining
Let (a,, 2 ..., w ) be these arbitrary constants, so that the solution can be
written
i,
8
0; and perform on the original
(q ly qz ..., q n
dynamical system the contact-transformation from the variables (q ly qz ...,q n
"
"
<*
>
Pi,...,
pn )
to variables (a x a2
,
Since
new system
so that
(a,,
. .
a n fr
,
&,
),
Hamilton s equation, the Hamiltonian function of the

and consequently the equations of the system are
satisfies
is zero,
...,
an
/8i, ..., y3 M )
are constant throughout the motion.
It
follows that if
denotes a complete integral of Hamilton s partial differential
equation, containing n arbitrary constants (a 1} a2 ..., y n ), then the equations
,
dW
^ = -^
r
*>
= dW
Wr
^=1
W
>
dynamical problem, since they express the variables

in terms of t and 2n arbitrary constants (a lt o2 ..., ctn
constitute the solution of the

>
(<7i></2>
cLn,p\,
"-,pn)
Phil. Trans. 1834, p. 247; ibid. 1835, p. 95.
316
&,

.
/3 M )*.
.
,
of freedom
is
In this
made
equation of the
It
way the
to
first
depend
[CH.
xn
solution of any dynamical system with n degrees

on the solution of a single partial differential
order in (n 4-1) independent variables.
should however be observed that the converse of this theorem
namely the theorem
that the solution of a partial differential equation such as Hamilton s depends on the
solution of a set of ordinary differential equations (the differential equations of the
characteristics), which in this case are of the Hamiltouian form, had been discovered by
Pfaff and Cauchy (completing the earlier work of Lagrange and Monge) before Hamilton
and Jacobi approached the subject from the dynamical side.
On the use that can be made of an incomplete integral of Hamilton s partial differential
equation (i.e. one containing less than n arbitrary constants besides the additive constant),
cf.
Lehmann-Filhes, Astr. Nach. CLXV. (1904),

It
may
be rioted that Hamilton
stands to non-holonomic systems
col.
209.
s partial differential
equation
is
an extension to such systems,
for
not applicable as
cf.
it
Quanjel, Palermo
Rendiconti, xxn. (1906), p. 263.
The integration of Hamilton s equation by separation of variables is discussed

F. A. Ball Acqua, Math. Aftn. LXVI. (1908), p. 398.
by
Consider the system
Example.
dq
_3H
dp ~
=
"
dt
_3H
dt
dp
~dq
where
IT
jf=fc>-,
g
and
is
fj.
The Hamilton s equation corresponding
a constant.
a complete integral of this equation
to this system is
may
be found in the following way.
Assume
W-fW+tte),
where / and
are functions of their respective arguments
0=/
(0
2
H4>
(2)}
then we have
-F/?-
This equation can be satisfied by writing
where
a is a constant
which gives
/ (0 = 1*1 a,
(?)
- (2,ia)* arcsin (gr/a)* + {^q (a - ?)/a}*

arcsin
+ {fyq (a - ?)/a}*.
TF= fMt/a + (2/xa)*
(g/a)*
solution of the original problem is therefore given by the equations /3=

=
d
p
Wfdq, where a and /3 are the two constants of integration.
The
143.
Hamilton s integral as a
solution of
Hamilton s partial
differential
equation.
There are an
infinite
*
This theorem
pp. 60, 161.
is
of complete integrals of Hamilton s partial

each one of them furnishes a contact-transformation
number
and
due to Jacobi, Crelle
J.
xxvu. (1837),
p.
97 and Liouviltfs J. in. (1837),
317
Dynamical Systems
142, 143]
from the variables (q^, q2 ...,q n ,p 1 ...,p n ) of the dynamical system to

variables (j, or 2 .., n @ lt ..., /3 n ), (the transformation involving t}, such
that the equations of motion of the system when expressed in terms of
,
(oij,
-,
>
n>
fii,
-,
the quantities (a l
i.e.
/3 W )
,
become the equations of the equilibrium-problem,
0%, ...,
an
..., /3 n )
fti,
are constants.
number of transformations thereîs one of special

in
which the quantities (a l a2 ..., a n /3 X ..., /3 n ) are
that
interest; namely
the initial values of (q 1} q z ..., q n p lt ..., p n ) respectively, i.e. their values at
this infinite
Among
taken as an epoch from which the motion is estimated. In

this case we can find in an explicit form the corresponding complete integral
a time
of
which
Hamilton
is
partial differential equation.
For consider Hamilton
integral
(99)
Ldt,
where
L denotes
the kinetic potential of the system.

Suppose that & denotes
S
small
8a
a variation due to
2
...,
n Sfti, ..., 8/3 n ) in the initial
changes (S^,
,
conditions.
Then
99)
we have
Ldt= I (p r 8q,.-l3 r S
B [
r=l
Jtv
ft
It follows that if the quantity
Ldt,
when the
integration
is
performed,
be expressed in terms of (q l} q2 ...,qn a ly ...,a n t), (we suppose this possible,

i.e. we assume that it is not
possible to eliminate (/3 1 /3 2 ..., /3n p ly ... p n )
from the relations connecting ( 1} ..., an ^, ..., fi n q ly ..., qn p l} ...,p n ), so
,
between (q1} ..., qn a lt ..., a n )) and if the function thus

obtained (which Hamilton called the Principal Function) be denoted by
W(q l} q 2 ..., q n a 1; ..., a n t), then we shall have
as to obtain relations
^__
9JF_
and therefore*
the transformation
(q lt q 2
ifs
...,
q n PI,
,
from
pn )
to
(a jt a 2
...,
an
{3 lt ..., /9 n )
a contact-transformation, and the integral of the kinetic potential
is the
determining function of the transformation.

*
Hamilton. Phil. Trans. 1834, p. 307 ibid. 1835, p. 95. In his earliest dynamical investiga
Hamilton used a "characteristic function" strictly analogous to the characteristic function
;
tions,
which he had employed with such success in optics this function being the Action integral,
expressed in terms of the final and initial coordinates. He found however that this function,
when employed in dynamics, involved the constant of energy, and so substituted for it the
"principal function" described above.
:
318
Also
[CH.
xn
we have
dW = dW
dt
LT =
or
I frWdqr
r= i dq r dt
dt
V r r
+ 2,
p q
=i
ot
or
OL
and therefore
the integral of the kinetic potential satisfies the equation
dW
M
which
is
Hamilton
Example.
(qi, ?2,
Let
?M
partial differential equation.
(a l5
Pi,
-,
a2
...,
an
ft,
dq,.
from
the
dpJ1==
relations
values (at time
_dH
dt
dpr
initial
(r
connecting
(aj,
a2
let
denote Hamilton
= l,
ft,
On,
...,
is
and
of
2, ..., w).
dqr
possible to eliminate (ft, ft, ..., ft,

-, ?n, Pi,
P)
2,
(?i,
that a number (say m) of distinct relations exist between (q lt q 2
these be solved for (a ls a 2 , ..., a,), so as to take the form
ifc
fo)
dynamical system represented by the
in the
= cH
dt
be the
ft ( )
...,
Pn) respectively,
equations
Suppose that
dW
dW
jo n
...,
...,
ft) with
...,pn) entirely, so
..., a n ) ; let
q n a l5
,
s integral
Ft
I
for the
system, expressed in terms of
(q^,
qz
qn
...,
o?n + i,
Establish the
<*n)-
equations
Pr =
where
(Xi,
X2 ,
...,
X m ) are arbitrary
9F
--
5
da
r
^ A*
fc=1
B/^
o
9a r
and shew that the function
W=V+
2 X t /t
fc=i
is
an integral of the partial
144.
TAe
connexion
of integrals
urith
transformations
infinitesimal
admitted by the system.
Let
r
dq
-
= dH-
dp r =
;
- dH
x
(r
..
1, 2, ... ,
n)
319
Dynamical Systems
143, 144]
be the equations of any dynamical system, and

0(?i,
921
q n Pi, -~,pn,
,
denote any integral of the system
we
<)
let
= Constant
shew that the knowledge of
shall
this
integral enables us to find a particular solution of the variational equations

(
H2).
For the variational equation

d
-T7
dt
oq r
= d*H
.
,.
o<fr
j.*
for 8q r is
&H
_
J
o<?
&H
djjj
*
j.
j."
j.
...
+~
&H
-
j.
j.
*"
6 n
j.~ *
but we have
8 2 zz
9 2 /i
80
dq n dp,- dp n
dqidp r dpi
3,
A;=I
ct0
9/>r
"
dp k 80
rfi
9p
80
dpidp r
dq^
ftli
fc
ct
80 \
d^ V8p r /
dt
-/z
80
3p n dp r dq n
fc=i
SÂ;/
rf<
82
^ ?)H
80 \
dqjc
",
90\
9^/
8 2 /z
80
9?* 9pt9p r
dJJ
k=idpk dqkdpr
90\
\dp r )
1 f?^
d V8pr/
and hence the variational equations
for (Sq 1} Bq2
...,
8qn ) are satisfied by the
values
where
e is
a small constant.
Similarly the variational equations for

(Spi,
Sp 2)
...,
8p B )
can be shewn to be satisfied by these values
e is
a small constant and
is
and hence
the equations
an integral of the original equations,
constitute a solution of the variational equations.
This result can evidently be stated in the form The infinitesimal contacttransformation of the variables (q^ q 3 ..., qn p lt ...,p n ), which is defined
by
the equations
:
transforms any orbit into an adjacent orbit, and therefore transforms the
whole family of orbits into itself. Adopting the
language of the grouptheory, we say that the dynamical system admits this infinitesimal contacttransformation. We have therefore the theorem that
a
integrals
of
dynamical
320
and contact-transformations which change

substantially the same thing; any integral
system,
&,
4>(qi>
"-,
q n Pi,
,
",
Pn,
[CH.
xn
the system into itself, are
Constant
corresponds to an infinitesimal transformation whose symbol

Poisson-bracket (
f ).
the
is
133)
(f>,
It will be observed that the ignoration of coordinates arises from the particular
case of this theorem in which the integral is p,. = Constant, where q r is the ignorable
coordinate the corresponding transformation is that which changes q r without changing
;
of the other variables.
any
Poisson
145.
s theorem.
The last result leads to a theorem discovered by Poisson* in 1809, by

means of which it is possible to construct from two known integrals of a
dynamical system a third expression which is constant along any trajectory of
the system, and which therefore (when it proves to be independent of the
integrals already
known) furnishes a new
Let
0(?i
and
-\Jr
?2,
(q-i,
</
>
...,
q n P\,
,
integral of the system.

,
Pn,
t)
q n PI, --^Pn,
,
Constant
= Constant
denote the two integrals which are supposed known. Consider the in
contact-transformation whose symbol is the Poisson-bracket
finitesimal
(/,
-v/r)
since ty
adjacent
is
an integral, this
144) transforms every orbit into an
orbit.
The increment
of the function
under this transformation
<
is e
(</>,
-v/r),
has constant values

is an integral,
where e is a small constant; but since
-v/r)
along the original orbit and along the adjacent orbit: the value of
must therefore be constant throughout the motion. We thus have Poisson s
<
<f>
(</>,
and ty are two integrals of the system, the Poisson-bracket

constant throughout the motion.
theorem, that if
is
-v^)
(<f),
If
A/T),
(<,
(f>
which
is
a function of the variables (q lt q 2
...,
qn
} ,
...,
pn
t),
does not reduce to merely zero or a constant, and if moreover it is not

^r and such other integrals as are already known,
expressible in terms of
(/>,
then the equation

(</>,
constitutes
The
how Poisson s theorem can be applied

when two integrals are already known.
Mec. Anal. (1853)
the extension of Poisson
cf.
to obtain
new
p. 260.
given by Bertrand in Note VII to the third edition of

Oeuvres de Lagrange, t. xi. p. 484.
discussion of this theorem
Lagrange
On
Constant
following example will shew
Journal de VEcole polyt. vin. (1809),
a new integral of the system^.
integrals of a dynamical system
-^r)
is
theorem
to
Soc. math, de France, xxxvn. (1909), p. 120.
non-holonomic systems,
cf.
Dautheville, Bull, de la
321
Dynamical Systems
144-146]
Consider the motion of a particle of unit mass, whose rectangular coordinates are
?2, &) and whose components of velocity are (p l p.2 , p 3 ), which is free to move
in space under the influence of a centre of force at the
The integrals of angular
origin.
momentum about two of the axes are
(?!>
ps ?2
- 5 3j2 = Constant,
r
i~
and
\% = Constant.
Let these be taken as the two known integrals

which is
and
\^
</>
the Poisson-bracket
(<,
\//-),
r=l
becomes in this case
PWi-fcft!
and
in fact, the equation

i
is
= Constant
another integral of the motion, being the integral of angular
momentum
about the
third axis.
The constancy of Lagrange
146.
The theorem
of Poisson has, as
might be expected, an analogue
in the
theory of Lagrange s bracket-expressions.
u r = 0r
Let
(r=l,
2, ...,
2n)
denote 2n integrals of a dynamical system with n

degrees of freedom, con
stituting the complete solution of the problem: the quantities u r being given
functions of the variables (q lt q2 ..., q n p lt
...,p n t], and the quantities ar
,
being arbitrary constants.

(<fc,
q2
...,
Lagrange
as
q n ,pi, --^Pn)
By means
functions
bracket-expressions [ar
quantities (a 1} a 2 ..., am ).
of these equations we can

express
of (a 1; az ... a2n t), and form the
,
a,],
where a r and a s are any two of the
Since the transformation from the variables

(q 1} q 2 ..., q n p 1} ...,p n ) at
t to their values at time t+ dt is a
contact-transformation, we have (128)
,
time
d
at r= i
where the symbols
and
B refer to
independent displacements from one
If now we take the

trajectory to an adjacent trajectory.
symbol
to a variation in which a t
is
the rest of the quantities
varied,
only
(a lt a2
...,
to refer
a 2n )
.remaining unchanged, and take 8 to refer to a variation in which a} only

varied, the last equation becomes
d 3 /dq r dp r
dt r= i\daidaj
w. D.
dqr dp r \
dajdaj
21
is
322
or
-^
[a {
=
,
[on.
xn
0,
cij]
which shews that the Lagrange-bracket [a;, a,-] has a constant value during the
motion along any trajectory; this theorem was given by Lagrange in 1808.
Lagrange
integrals
Lagrange
147.
s result,
unlike Poisson
we have
for
know
to
all
does not enable us to find any new

the integrals before we can form the
s,
bracket-expressions.
Involution-systems.
Let (u l} u2
...,
u r } denote r functions of 2n independent variables

(?l, ?2,
if it is possible
q n ,Pi,
,Pn)\
to express all the Poisson-brackets
(iii,
(wj.Mg, ...,u r ), the functions (u lt u2 ...,ur ) are said to form

Any function of (u lt u.2 ..., u r ) belongs to this group.
,
uk )
a,
as functions of
function-group*.
If the quantities (uit n^} are all zero, the functions (u 1} u 2

to be in involution, or to form
Now
v
...,
u r } are said
suppose that (i/ u 2 ...,w,-) are functions in involution: and let

and w =
be any two equations which are consequences of the
>
equations
we
an involution-system.
shall
shew that
For since (u lt
and
u.2 , ...,
u^
= 0,
satisfy the relation (v,
M2
= 0,
ur
w)
= 0.
u r ) are in involution, each of the equations

M,
0,
u2
0,
ur
admits each of the r infinitesimal transformations whose symbols are

(
"
I,/), (W 2 ,/),
(Mr,/)5
and consequently the equation v = 0, being a consequence of these equations,

must also admit these transformations that is to say, we have
;
)=0
("*,
(k
l, 2, ...,*),
and therefore each of the equations

ii!
= 0,
u2
0,
ur
admits the infinitesimal transformation whose symbol

equation
equation
w=
w=
is
(v,f).
Since the
a consequence of these equations, it follows that the

must also admit this transformation, and therefore we have
is
(v, iv}
= 0,
result.
*
Lie, Math. Ann.
vm.
(1875). p. 215.
323
Dynamical Systems
146-148]
Hence we
see that if(u 1} uy

v
u r ) are in involution, and the equations
...,
= 0,
v.2
0,
vr
ur
= 0,
are consequences of the equations

M,
then the functions (v lt v2
u2 =
0,
0,
v r ) are in involution.
...,
when half
Solution of a dynamical problem
148.
known.
the integrals are
for systems with two degrees of freedom

to
extended
121 can now be
in
systems with any number of degrees of
stated * If n distinct integrals
thus
The theorem may be
freedom.
The
which was established
result
<f>r(qi,q*.
where (a 1} a 2
...,
>~,p
t)=a r
(r=l,
2, ..., n),
a n ) are arbitrary constants, are known for the dynamical
system
dq r
~TT
dt
q n ,Pi,
>
= dH
^
dp,._
~
IT
dt
>
Opr
~~
dH
"-5
"
>
dq r
and if the
>Pn,t),
.,q n
any given function of (q 1; q.2
are in involution, then on solving these integrals for
functions (^,02,
as
them in the form
so
to
obtain
ipn)
(p\,lh,
where
is
,pi>
>
<f>n)
Pr=fr(qi,
and
substituting (/i,/2
-.
q-
n>
...,/n) respectively
p dq
l
+p
the latter expression becomes
dV(q lt
dqt
q^
...,
for (p
i,
a2
On,
,p.2
(r
= l,
2, ...,
n)
...,pn) in the expression
Hdt,
differential
qn
+p n dq n
...
a perfect
^2,
...,
denoting
an
it
by
t),
remaining integrals of the system are
the
=6,
^
oa
r
where
(b l} 6 2
>
follows
<f)i-a lt
a.2 ,
...,
are in involution,
<
Pifi, p^-f*,
=0
(r, S
|i-M=0
dq
dq
or
it
>
fn are
pn
and therefore
(Pr ~fr, PS ~fs)
This theorem
well-knosvn
2, ...,n),
<f>.
last article that the functions
by the
= l,
b n ) are arbitrary constants.
For since the functions

in involution,
(r
method
of the first order.
is
1, 2,
n\
(r,-l,2,...,n>
essentially the application to
Hamilton
partial differential equation of the
Complete Integral of a non-linear partial differential equation

As a dynamical theorem it is due to Liouville, Journal de Math. xx. (1855),
for finding a
p. 137.
212
324
xn
[CH.
dH _ dp r _ dfr
dt
dq r
dt
= dA + |
Mî
s-i dq s dt
dt
= dfr + | df.dH^
dt
s =i
tyr dp s
and consequently
= _d_H _ I dHdf^
dfr
dt
dq r
=idp s dq r
_dH,
dq r
where H^ stands
H when expressed in terms of the arguments
for the function

(?1,
?2
?n,
i,
ni 0-
>
The equations
_
dqr
shew that
is
/i
rfg-j
dt
dq s
+/ dq
2
...
dq r
+f
H^dt
n dqn
the perfect differential of some function V (q lt

first part of the theorem.
q^, ...,
qn
0,1,
an
t)

If
now
d V =f dq
l
+/ dq +
2
...
W da
+fn dq n - H^dt +*2 r
r.
In this equation replace the quantities a r by their values

obtain an identity in (q l} q2 ..., q n PI, p%,
,p n i), namely
,
dV
dV- 2 ^
d(f) r
=p
dq
cf) r
we thus
+p
dq 2 +
...
+p n dq n -
where on the left-hand side of the equation we suppose that in
dV
x
V with
the symbol d denote the total differential of the function

we have therefore
respect to all its arguments,
the quantities (a ly a z
...,
a n ) are replaced by their values
(<>!,
dV
and
...,
0n).
</>
\) \AJ-f
This equation shews that the differential form

Pidqi
when expressed
+p
dq^+
...
+ p n dqn
in terms of the variables
Hdt,
(q l}
q.
2>
...,
qn
,
<}>i,
<f>z,
0n,
t),
takes the form
and hence the
differential equations of the original dynamical

to
the
first Pfaff s
equivalent
system of this differential form,
(d
=
V/da r )
0,
d(f>
(r
problem are
namely
1, 2,
. .
n).
325
Dynamical Systems
148, 149]
The expressions dV/dar are therefore constant throughout the motion,

the equations
dV/da r
where
br
(r=l,
i.e.
2, ..., n),
new
(b lt
arbitrary constants, are integrals of the system
this completes the proof of the theorem.
b2
...
b n ) are
Example. In the motion of a body under no forces with one point fixed, let ($, $,
denote the three Eulerian angles which specify the position of the body relative to any
fixed axes
at the fixed point, (.4, J5, (7) the principal moments of inertia of the
body at the fixed point, a the constant of energy, a t the angular momentum about
-v|/-)
OXYZ
the fixed axis OZ, and a 2 the angular
and
plane:
let
(# n
l5
momentum about
denote 37730,
^)
dT/d<j>,
the normal to the invariable
dT/d^
respectively.
the
Obtain
equations
Q
=arctan
{(
2
i
- #i 2 ) /a i}
Hence shew that
9d
is
the perfect differential of a function
F,
and that the remaining
integrals of the
system are
oV
9^~
where
6,
37
97
^"*
a^T
"*
&i, 6 2 are arbitrary constants.
(Siacci.)
Levi-Civita s theorem.
149.
Levi-Civita* has established a connexion between the integrals of a

dynamical system and certain families of particular solutions of the equations
of motion.
Consider
Let
q
coordinates
2
(<?!,
The
a system in which some of the coordinates are ignorable.

the non-ignorable
-, q n )
m
q ) be the ignorable and (q m +i,
and let L denote the kinetic potential.
first
...,
integrals corresponding to the ignorable coordinates are
dL/dq r
Constant
(r=l,
2, ..., ra),
and corresponding to these integrals there exists a class of particular solutions

of the system, namely those steady motions ( 83) in which (q l} q 2 ..., q m )
have constant values which can be chosen arbitrarily, while (q m
qm+z,
qn)
have constant values which are determined by the equations
,
+i>
dLjdq r
there are
oo
2in
=0
(?
=m +
of these particular solutions, since the
* Bend, dell Ace. del
Lined,
x. (1901), p. 3.
1,
m + 2,
. . .
n);
constant values of
Cf. Burgatti, ibid. xi. (1902), p. 309.
326
xn
[CH.
q m ) can be arbitrarily
The theorem of Levi-Civita, to the consideration of which we
assigned.
shall now proceed, may be regarded as an extension of this result.
(q lt
...,
q m ) and the
values of
initial
dH
dq r
dH
dp r
l = ay/
Let
(qi, q-2
*"-3
-%*.
be the equations of motion of a dynamical system, the function

supposed not to involve the time explicitly.
Let
F (q ,qt,...,q n ,p ,...,pn = Q
of m relations, which when solved
r
be a system
the form
Pr=fr(qi,
q*,
(r
q n ,p m +i,
>
Pn)
for
(r
(p 1} p
= 1,
H being
m)
1, 2, ...,
2
.)
...(A)
...,p m )
2, ...,
take
m)...(A
1 ),
and which are invariant relations with respect to the Hamiltonian system,
i.e. which are such that if we differentiate the relations (Aj) with respect
to t, we obtain relations which are satisfied identically in virtue of the
These
equations and of the equations (Aj) themselves.
of
as
a
the
invariant relations include,
particular case, integrals
system
in this case, they will involve arbitrary constants.
Hamiltonian
Since the relations (A x ) are invariant relations, we have
_
dH_df
- ~T7r--$
*
o
dt
dqr
dfrdH
~
5
j= m +\opjoqj
| dfr
*
-5
dH
V
*>
*
>
"^^
j-\oqjOp)
and writing
Tf}=
(F,
S
/-*-
becomes
this
this
equation
(Pi,p2>
Moreover,
among
an identity when for each of the

substitute the corresponding function r
becomes
,pm) we
we
suppose that the relations (A) or (Aj) are in involution

This condition is expressed by the equations
shall
themselves.
!;-!
Let
by
i*"-
c--- 1.2,
K denote the function obtained from H on replacing
their values (/i,/2
...,.).
..(2).
(p 1}
...,
jo 2 ,
pm )
...,/m ), so that
3F = 8^_
and
quantities
dp r
~dp
5T
"a
8/7
.~idp,d
--- ^
^~
(r
1, 2, ...
m)
...(4).
Dynamical Systems
149J
From
(3)
we have
[H, fr \
and combining
f)ff
[K,
fr}+^~
!/, /.}
*=i wf*
now shew
shall
+ {H,fr
},
pm+2
Pm+i,
pn
q lt
d fdK\
and
d fdE\
dt(ty,J
dt(dq r )
.,
?)
(r
i.e.
have from the Hamiltonian

equations
_
(BK\ ~
/7/
57T I
dt
\oq r /
I
and
(5) gives
f
"
1
(
dq r )
-^ ^
^T
g=l oq r dq
on differentiation,
using (B),
=
5T"
>/*f
now taking account
of (B),
we have from
(3)
___
-__
9?r
and hence equations

dt
dp
(6)
(2),
*=i9p, dq
become
dps ldpr dqs
i^r^
.r, dp. ldq r dqs
dpr
>
= 1,
that
(r=m + l,m
are zero in virtue of

equations (A), (B), (1), (2), (3), (4), (5).
1, 2,
and using
invariant with respect to the Hamiltonian

equations,
We
m),
we obtain
that the system of equations
Pr =fr
is
(r
with (4) we have
this
Substituting in (1) this value ofdH/dq r

the equations
We
327
2,
. . .
[CH.
xn
or by (7),
dK\
d /dK\
.
= 0,
jrU
which proves that the system of equations (A) and (B)

respect to the Hamiltonian equations.
Now
from the equations (A) and (B),
let
invariant with
the variables
,pn, q m +i,
(PI, Pa,
is
qn )
be determined in terms of (qlt qz ..., qm ): from the invariant character of

(A) and (B) it follows that on substituting these values in the Hamiltonian
,
independent equations, namely those which

express (dqjdt, dq 2 /dt, ..., dqm /dt) in terms of (q lt q z ...,qm), the others being
and the general solution of this system, which will
identically satisfied
we
equations,
obtain
shall
oo m particular solutions of the

arbitrary constants, will give
Hamiltonian equations. The solution of this system can, by making use
of the integral of energy, be reduced to that of a system of order (m
1)
contain
and thus we obtain Levi-Civita s theorem, which can be thus stated To any
set of TO invariant relations of a Hamiltonian system, which are in involution,
m
there corresponds a family of oo
particular solutions of the Hamiltonian
:
system, whose determination depends on the integration of a system of order
(m -
1).
If the invariant relations (A) are integrals of the system, they will contain
integrals of a
arbitrary constants and hence to a set of
Hamiltonian system, which are in involution, there corresponds in general
a family of oo 2m particular solutions of the system, which are obtained by
another set of
integrating a system of order

Example.
(m
1).
For the dynamical system defined by the Hamiltonian function
H= qi pi - q p
2
- aqi 2 + bq^,
shew that the Levi-Civita particular solutions corresponding to the integral

~
= Constant
(Pi
bqz)lq\
are given by the equations

2i
where
150.
We
is
= 0,
qz
= e~ t+f
Pl
= ae~ t+e
p,
= be~ t+e
Systems which possess integrals linear in the momenta.

shall
now proceed
to the consideration
of systems
which possess
integrals of certain special kinds.
Suppose that a dynamical system, expressed by the equations

dqr
_dH
~df-dp r
dpr
dt~
dH
dqr
.!,*.....>
329
Dynamical Systems
149, 150]
has an integral which
is
fipi
where (fi,f2
+f
p n ),
...,
say
+fnp n = Constant,
p<t
..,/) are given functions of (qi,q2
and homogeneous in (p 1} p2
linear
...,qn).
Consider the system of equations

dq\
_ dq n
_dq*_
=
7T~7
which
its
is
of order
(??
"/.
!); suppose that the (n
Qr (qi,
and
let
q?,
qn )= Constant
which constitute
(r=l,
2, ...,
1);
be a function defined by the equation
Qn
where in the integrand the variables

by their values in terms of (q lt Q 1}
is
1) integrals
solution are
(q2
Q2
q3
...,
...,
q n ) are supposed replaced

before the integration
Qn-i)
carried out.
Then
if
the variables change in such a way that (Qx Q 2 ...,

varies, it follows from the above equation that
,
Q M _j) remain
constant and Q n
_ dqn _ j n
dqi_d<h_
f
/i
so that if (Q 1}
which
((?!,
q2
...,
**
*<ln
>
Jn
/2
Q n ) are regarded as a set of new variables in terms of

can
be expressed, we shall have
qn )
...,
Suppose then that we consider the contact-transformation which

extension of the point-transformation from the variables (q 1} q 2
variables (Q lt Q2
-,
2
Qn), so that the new variables (P l}
,
defined
the
qn ) to the
n ) are
...,
132) by the equations
Pr =
i^|f
O^r
(r
*=1
By
..,,
,
is
this transformation the differential equations of the

new set of Hamiltonian equations
= l,
2, ...,n).
dynamical system
are changed into a
dP r _J>K
dQr_
dt
dPr
dt
bQ r
and the known integral becomes
Pn = Constant.
Since
dPn /dt = 0, we
n explicitly
have dK/BQ n
and thus we obtain the
0, so
the function
result that
K does not involve
when a dynamical system
330
possesses an integral which is linear
[CH.
and homogeneous in (p lt p2
..,
p n ),
xn
there
a point-transformation, from the variables (q l} q.2 ..., q n ) to new variables

Qn), which is such that the transformed Hamiltonian function
does not involve Q n
The system as transformed possesses therefore an
exists
(Qi>
Qs>
>
we have the theorem that the only dynamical

which
systems
possess integrals linear in the momenta are those which possess
ignorable coordinates, or which can be transformed by an extended pointtransformation into systems which possess ignorable coordinates.
ignorable
coordinate, and
The converse
of this theorem
is
evidently true.
This result might have been foreseen from the theorem

?2,
-,
144) that
if
= Constant
q n ,Pi, --,Pn,
an integral of the system, then the differential equations of motion admit the
For when is linear and homogeneous
infinitesimal transformation wr hose symbol is
/ ).
is
(<,
(f>
if
in (p l , pi, ..., p n }, this transformation is ($ 132) an extended point-transformation
this point-transformation is transformed by change of variables so as to have the symbol
:
df/3Qn
Qn explicitly.
now
Considering
of a
Hamiltonian function of the equations after transformation
clear that the
it is
cannot involve
kinetic
in particular
energy T(q l} q2
...,
systems whose kinetic potential consists

qn
q l}
...,
q n)
which
quadratic in the
is
q n ) which is
an
we
in
order
that
see
that
of
the
velocities,
integral linear in
independent
the velocities may exist the system must possess an ignorable coordinate,
or must be transformable by a point-transformation into a system which
velocities (q l} q2
...,
and a potential energy
qn )
V(q-i, q 2
But in either case the functions T and

possesses an ignorable coordinate.
evidently admit the same infinitesimal transformation, namely the trans
formation which,
when
the coordinates are so chosen that one of
them
is
the
ignorable coordinate, consists in increasing the ignorable coordinate by a

small quantity and leaving the other coordinates and the velocities unaltered ;
and conversely, if T and
admit the same infinitesimal transformation, then
there exists an integral linear in the velocities.
Levy
This result
is
known
as
theorem, having been published by Levy* in 1878.
Example
1.
Shew that
if
the differential equations of motion of a particle admit an

momentum, the line of action of the force must
integral linear in the components of

belong to a linear complex.
(Cerruti, Collect, math, in
p.
mem. D. Chelini
cf.
P. Grossi,
Palermo Rend. xxiv.
(1907),
25.)
Example
2.
If the equations
<
"
Comptes Rendus, LXXXVI.
(,-!,*...,
2 2 a
T=% i=lk=l
where
ik
qiq k , and where (ft,
Q.2 , ...,
Q n a u a 12
,
...,
a mi ) are given functions
?)) possess an integral of the form
f
(?i>
331
Dynamical Systems
150, 151]
?2>
Cl qi + C2 q 2 + ... + Cn q n + C= Constant,
where (C l C.2 ..., Cn C) are functions of (q 1} q%, ..., qn ), shew that it
one of its positions
displace an invariable system in one direction from any
defined by the form
,
is
possible to
in the space
Sn
i=l k=l
Shew that
for this a necessary
sufficient condition is that the ds 2
and
formed in such a way that one of the variables becomes absent from the
can be trans
coefficients.
and Levy.)
(Cerruti
Determination of the forces acting on a system for which an

integral is known.
151.
Before proceeding to discuss systems which possess integrals quadratic

we shall obtain a result due to Bertrand*, namely that
in the velocities,
in the
motion of a dynamical system of given constitution,
for
which however
the acting forces are unknown (it being supposed that the forces depend
of application, and not on the
solely on the coordinates of their points
discover
the
we
can
unknown forces provided one
of
these
velocities
points),
integral
must
is
Moreover, this integral cannot be chosen at random, but
known.
satisfy certain conditions.
Let (q1} q*,, ..., q n ) be the n independent coordinates of the system, T the
kinetic energy, and (Q 1} Q2 ..., Q n ) the unknown forces, which are supposed
to depend only on (qi,q 2
qn) so the equations of motion are
,
d idT\\
dt \dq r
Let
<(?i>
?2,
be an integral of the system
qn
8T
dT
= Qr
)-^9o,
#1,
>
qn,
(r=l,
2, ..., n).
t)= Constant
on differentiating
it,
we have
Substituting in this equation for (q ly q2 ..., q n ) their values as given by

the equations of motion, we have a relation involving (Q ly Q.2 ..., Q n ) linearly.
,
This relation, as
all
of which
it
contains only the quantities (q lt q2
we can
...,
assign arbitrary independent values,
q n q i9
,
we can therefore differentiate it with respect to (q l} q2

n new equations which, likewise containing (Q1} Q2 ..., Q n )
The
suffice in general to determine these unknown quantities.
,
Journal de Math,
(i)
xvn. (1852), p. 121.
...,
qn
>
t},
to
must be an identity:
..., q n ), and so form
linearly, will
integral will
332
relate to
an actual system only when these values of (Q lf
[CH.
...,
Qn )
xn
satisfy
the equation
the cases in which the equations for the determination of (Q Q ..., Qn) are
not independent, so that (Q 1} Q2
Qn) are indeterminate, are those in
which the integral is common to several distinct dynamical problems.
} ,
to
Example. If an integral of the equations of motion of a point

two different problems, shew that it is of the form
F(<fi,
2>
>
x, y,
t}
in a plane is
common
= Constant,
coordinates and $ is the derivate with respect to t of

(x, y) are rectangular
a function $ (x, y) which, equated to a constant, represents the equations of a set
where
of straight lines.
152.
(Bertrand.)
the case
to
Application
in
possess an integral quadratic
whose equations of motion
of a particle
the velocities.
let it be required to find the

application of Bertrand s method,
in
order that the equations of
nature of the potential energy function
motion of a particle which is free to move in a plane under the action of
As an
conservative forces,
M
dv
y~
/it
da;
._
dv
___
dy
may possess an integral (other than the integral of energy) of the form
= Constant,
Ptf + Qxy + Rif + Sy+Tx+
where P,
Q,
R, 8, T,
K are functions of x and
y.
Differentiating the last equation, and substituting

equations of motion, we have
^~
n /..8F
-Q
*
I
^.
y x T~
dxj
dy
dx J
\dy
^-
.8F
.dV\
<fc
for
^
dx
8F ^8F =
dK
dK
+ -^-y-Sj--T^
^-x
r
ox
oes
dy
dy
x and y from the
-Kdy
/A
Equating
to zero the
dx
from which
integral
where
terms of the third degree in x and
dR
- =
dP
=0,
dP
0,
dy
^
dy
.................................... (A).
-_- = 0,
+ dQ
-
ox
dQ + dR =
^
ôx
y,
we have
U,
dy
in the
readily deduced that the terms of the second degree
it is
must have the form

(ay"
by
(a, 6,
c,
c)
,
x 2 + (ax2
+ b x + c } y- + (- 2axy - b y -bx + Cj) xy,
c^ are constants.
333
Dynamical Systems
151, 152]
terms of the second degree in x and y in equation
to zero the
Equating
(A), we have
(jCC
^jOC
C/7/
(J
from these equations we deduce
T = - my +
8 = mx + p,
where (m, p,
q,
q} are constants.
Equating
to zero the
terms independent of x and y in (A), we have
dV,
(mx + p)-
or
97
d
This equation shews that if (m, p, q) are different from zero, the force is
directed to a fixed centre of force, whose coordinates are
p/m and qfm we
shall exclude this simple particular case, and hence it follows that the con
stants (m, p, q) must each be zero, so that the integral contains no terms of
the first degree in x, y.
;
Equating
to zero the
terms linear in x and y in (A), we have
2P
8F
dV
---~
ox
Q
v
+r
oy
dK
-z.
ox
-0w,
Differentiating the former of these equations with respect to y, and the

latter with respect to x,
and equating the two values of
d*K
thus obtained,
we have
-
dxdy
dx dy
and replacing P,
Q,
dy*
dx dy
dxdy
dy dy
by their values as found above, we have
87
dx
Darboux* has shewn that
97
dy
this partial differential equation for the function
can be integrated in the following way.

*
dx
Archives Neerlandaises,
(ii) vi.
p.
371 (1901).
334
[CH.
xn
Excluding the particular case in which the constant a is zero, we can

always by change of axes reduce the given integral to the simpler form
2
^ (xy
yx)
+ ex +
2
K = Constant,
which amounts to supposing that
tt
if
moreover we replace
i,
= 0,
d = 0;
= 0,
by ^c the
partial differential equation for
becomes
(y
To
~x +c
integrate this equation,
dV
dV
- 3a?3-0.
+ 3y
y =.
5-31
cxdy
dx
we form the
dy
differential
equation of the
characteristics
2
xy (dy
- dx-) + (x -y - c ) dxdy = 0.
2
2
z
equation we take x and y as new variables,
equation: we thus find that its integral is
If in this
Clairaut
(7?^
+ 1 ) (mx
x2
y-
a2
-c
now
0,
a simple change of notation, we
By
<-
a2
becomes a
denotes the arbitrary constant.

can write this integral in the form
where
me =
it
This form puts in evidence the

curves of the partial differential
interesting fact that the characteristic
where the arbitrary constant
is
a.
equation are two families of confocal conies.
Taking then as new variables a and

ellipses and hyperbolas, so that
* = ?,
the parameters of the confocal
y^^-c^^-F)}*,
C
it is
/3
known from the general theory
that the partial differential equation will
take the form

d
da.
oaofi
d/3
where
variables
B aree
and
we
+ B^ =
A^
da
"
functions of a and
y3
in fact,
0,
on performing the change of
find
which can be immediately integrated, giving
are arbitrary functions of their arguments. It follows that the

the
motion of a particle in a plane, under the action of conservative
cases
only
of
where /and
<
335
Dynamical Systems
152, 153]
which possess an integral quadratic in the velocities other than the

integral of energy, are those for which the potential energy has the form
forces,
/(a)
- $ (0)
-?T05-.
where
a.
and
/3
are the parameters of confocal ellipses and hyperbolas.
Since by differentiation
the kinetic energy
we have
is
T= A (V - P)
#
2
and an inspection of the forms of T and F shews that these problems are
of
Liouvilles class ( 43), and are therefore integrable by quadratures.
153.
General dynamical systems possessing integrals quadratic in the
The complete determination
velocities.
of the explicit form of the
most general dynamical system

whose equations of motion possess an integral quadratic in the velocities (in addition
to the integral of energy) has not yet been effected.
It is obvious from
43 that all
dynamical systems which are of Liouville s type, or which are reducible to this type by a
point- transformation, possess such integrals : and several more extended types have been
determined *.
Example
Let
(&, I
(q k )
be n 2 functions depending solely on the arguments indicated, and
4>=
=1
n)
2,
. . .
(1=1,
2,
...,%)
let
W *H
fc=i
denote the determinant formed by these functions.

dynamical system is reducible to the form
but also (n
1)
is
zero, there exists
Shew
that
if
the kinetic energy of a
not only the integral of energy,
other integrals, homogeneous and of the second

degree in the velocities,
namely
1
A-^f*
where
(a 1( a 2 ,
...,
an)
are arbitrary constants:
""
(*
and that the problem
quadratures.
Example
= 2,
is
3,
...,),
soluble
by
(Stackel.)
2.
Let the equations of motion of a dynamical

system with two degrees of
freedom be
d
where
*
Cf. G. di Pirro,
Annali di Mat. xxiv. (1896),
p. 315.
336
and
are
(a, A, 6)
any functions
integral
of the coordinates (q, q 2 )
and
[on.
let this
system possess an
a q^ + 2A J! j 2 + & ?22 = Constant,
the equation of energy, where ( , A 6 ) are

quadratic in the velocities and distinct from
2
2
If A and A denote (afc-A ) and (a & -A ) respectively,
functions of the coordinates.
,
and
if
Jl
where
<?/
_A\
(o jji
+ Zh ft qj + b q^\
stands for dq r jdf, shew that the equations
,-
at \cq r /
define the
of motion,
oq r
same relations between the coordinates (g^, q?) as the original equations
and that one set of equations can be transformed to the other by the trans
formation
1.
dynamical system
denned by
is
its
kinetic energy
ni
(where $
in
denotes the determinant
which the elements of the th
$ w ), and by
its
$21
$22
nl
$2rc
......
line are functions of q k only,
and $ w denotes the minor of
potential energy
*
*
^ = ^11
where
and the quantity
\^ fc
denotes a function of q k only.
Shew
that a complete integral of the
Hamilton-Jacobi equation
i=U
where
2.
is
(a 1? a 2 , ..., a,t )
{a 1
$ + a 2 $j 2 + ...+a n
il
(Goursat.)
If
an integral of a dynamical system which possesses an integral of energy, shew that
= Constant,
-^
ct
*f
OT
= Constant,
etc.,
are also integrals.
337
Dynamical Systems
xii]
3.
system of equations
A r(qi,
?2,
= Br(3li
?2,
-jf=
q n PI,
-,
-,?*,
(r=l,
d{
such that
is
an
if
and
<
-v|/-
S n.l
2,
...,n)
jn,
l,
are any two integrals whatever, the Poisson-bracket
(<,
is
*//)
also
Shew that the equations must have the Hamiltonian form
integral.
dH
~dt~~dgr
dff
dqr
dp r _
~di-ty r
(Korkine.)
If
4.
= Constant,
at
>
g n,^!,
an
also
...,
/3 fc
= Constant,
shew that
/>),
A,,...,
is
a*-
= Constant,
/3 2
integrals of a Hamiltonian system of differential equations, the variables being
are
e any 2
?2,
...,
= Constant,
a2
& = Constant,
!)
= Constant,
A*
p^
oqx k
?A 2
O?AI
p\ k
(Laurent.)
integral.
Let the expression
5.
f
[JZi. /z.),
....
:B(^,^...,^
-n
-..
uJ
2r
)
r
-5-7
i-l<K*Ui
where HI,
...,
n are
functions of the nv variables
be called a Poisson-bracket of the nth order.

A
^n, ^12,
n 2, -,
yn, Vi-2,
,i/h V
!
denotes
all
^2t,
where
*<)
xji (j=l,
G2
If G\,
AV>
(h
...,
2, ...,
Ghv
+ k = ri),
are
and
il,
Ai>
2, ...,
i>)
functions of
if
the Poisson-brackets formed from every n functions G, shew that
represents the necessary and sufficient conditions that the functions
yt = F*t(xn,
#12,
Xkv",
2,
i,
arising from the equations

<?
shall satisfy the
-,
6.
(.r,
y)
A
is
in
P (F n
t
move
>
is
/(#, y\ the total energy being
h.
=1
obtained when
2, ...,
order
we
replace h of the
(Albeggiani.)
mass whose coordinates
in a plane
(i=l,
first
by as many y s.
particle of unit
free to
2
!>
simultaneous partial differential equations of the
where Pi(y h F} denotes the expression which

functions
(*
(i>hv)
referred to fixed rectangular axes are
under forces derivable from a potential-energy function
Shew that
if
the orthogonal trajectories of the curves
82 \
W. D.
+ 5~~9 lg {*/(*!
I
ayy
y)}
Constant
22
338
Properties of the Integrals of Dynamical Systems [en.
xn
are orbits, the differential equations of motion of the particle possess an integral linear and
homogeneous in the velocities (#, y).
7.
The equations
of motion of a free system of
^
If
particles are
=X
(s=l,
2,
...,3m).
an integral exists of the form

3m
fx -
Ct
are functions of A^,
#2
where
/i,
/2
...,
/3m
2
=1
= Constant,
>
>
^sm? an(*
C* is
a constant, shew that this
integral can be written
3n
3m
2
s=l
.r 8
2
r,
of ,.
(.,.
- xrXg)
Ct = Constant,
s=l
where the quantities ks and a rg are constants.
(Pennacchietti.)
Two particles move on a surface under the action of different forces depending
if their differential equations of motion have in
their respective positions
on
only
common an integral independent of the time, shew that the surface is applicable on
8.
a surface of revolution.
(Bertrand.)
CHAPTER
XIII
THE REDUCTION OF THE PROBLEM OF THREE BODIES

154.
Introduction.
The most celebrated

of Three Bodies, and
of
may
all dynamical
problems
be enunciated as follows
is
known
as the Problem
Three particles attract each other according to the Newtonian law, so that
between each pair of particles there is an attractive force which is proportional
product of the masses of the particles and the inverse square of their
distance apart: they are free to move in space, and are initially supposed to be
moving in any given manner ; to determine their subsequent motion.
to the
The practical importance of this problem arises from its applications to

Celestial Mechanics
the bodies which constitute the solar system attract
:
each other according to the Newtonian law, and (as they have
approximately
the form of spheres, whose dimensions are very small
compared with the
distances which separate them) it is usual to consider the
problem of deter
their
motion
in
an
ideal
in
which
the
are replaced by
bodies
form,
mining
of
masses
to
the
masses
of
the
particles
equal
respective bodies and occupying
the positions of their centres of gravity*.
The problem
of three bodies cannot be solved in finite terms
of any of the functions at

present known to analysis.
stimulated research to such an extent, that since the
by means
This difficulty has

year 1750 over 800
memoirs, many of them bearing the names of the greatest mathematicians,

have been published on the subject f
In the present chapter, we shall discuss
.
the
known
integrals of the system and their application to the reduction of

the problem to a dynamical
problem with a lesser number of degrees of
freedom.
The motions of the bodies relative to their centres of gravity (in the consideration of which
and shapes of course cannot be neglected) are discussed separately, e.g. in the Theory
Precession and Nutation.
In some cases however (e.g. in the Theory of the Satellites of the
their sizes
of
Major Planets) the oblateness of one of the bodies exercises so great an

cannot be divided in this way.
t For the history of the
effect,
that the problem
Problem of Three Bodies,
cf. A. Gautier, Essai historique sur le

1817) B. Grant, History of Physical Astronomy from the earliest
ages to the middle of the nineteenth century (London, 1852)
E. T. Whittaker, Report on the
progress of the solution of the Problem of Three Bodies (Brit. Ass. Bep. 1899, p. 121)
and
E. 0. Lovett, Quart. Journ. Math. XLII. (1911), p. 252, who discusses the memoirs of the
period
1898-1908.
probleme des
trois corps (Paris,
222
The Reduction of
340
The
155.
denote the three particles,
their
(TIB,
and
let (q lt q 2 q 3 ), (q 4
si>
^12)
tively.
T=
m,

is therefore
m m. r
is
m.2
- m.m,
The equations
m
{(q,
their masses,
and
is
w,w a
is
Tc
mmr^
l
where k2
suppose the units so chosen that
shall
m^m^r^
becomes
so that this attraction
3)
we
the force of attraction between m^ and
constant of attraction
fixed rectangular axes Oxyz,

be the coordinates of P, Q, R, respec
q 5 q6 ), (q 7 qs q 9 ),
kinetic energy of the system
,
The
m m
(ra 1;
Take any
mutual distances.
xm
of the problem.
Let P, Q,
[CH.
the
12
is
the
is
unity,
and the corresponding term in
The
&
3 ??z 1
-q
7}
(q s
- q*)* +
(q6
-q
9)
- qj* + (q - qj + (q, - q )-} *

- qJ + (q -q ) +(q -q Y}~^
"
{(q 7
{(q,
of motion of the system are
m k q = -dV/dq
r
where k denotes the integer part of ^

9 differential equations, each of the 2nd
(r
(r=l,
2).
order,
2, ...,9),
This system consists of
and the system
is
therefore
of order 18.
mk q r = p
Writing
H= r=S
and
(r
_
~
2,
9),
-g^+V,
2mjc
the equations take the Hamiltonian form
dq r
~r,
at
and these are a
_ oH
^
dp r
It
j7
at
_ ~ oE
~~r~
1
">
9h
oq r
set of 18 differential equations, each of the 1st order, for the
determination of the variables (q
which
dp r
j
l>
q.2 ,
...,q a ,pi,p2,
,^9)-
was shewn by Lagrange* that this system can be reduced to a system

That a reduction of this kind must be possible
is only of the 6th order.
may be
seen from the following considerations.
In the
first place,
since no forces act except the
mutual attractions of the
*
Eecueil des pieces qui ont remporte les prix de I Acad. de Paris, ix. (1772). Lagrange of
course did not reduce the system to the Hamiltonian form. Cf. Boblin, Kongl. Sv. Vet.-Handl.
XLII. (1907),
No.
9, for
an improved Lagrangiau reduction.
Problem of Three Bodies
155]
particles, the centre of gravity of the
uniform velocity.
This fact
is
341
system moves in a straight line with
expressed by the 6 integrals
m q - (p,
+ m q- + m q - (p +p +p
+ m. q + m q - (p +p +p
3
8)
9) t
=a
=a
where a lt a.2 ..., a6 are constants. It may be

expected that the use of these
integrals will enable us to depress the equations of motion from the 18th to
the 12th order.
,
In the second place, the

angular momentum of the three bodies round
each of the coordinate axes is constant
throughout the motion. This fact
is
analytically expressed
qiP-2
where
by the equations
q-2pi
-qp +qp q p = a,,

p
q p + q p,
q p = 03,
=a
+
p4
qp
qp
qp
q^p*
a8) a9 are constants. By use of these three

integrals we may
to
be able to depress further the
expect
equations of motion from the
12th to the 9th order. But when one of the coordinates which
define the
a-,
position of the system is taken to be the azimuth

of one of the bodies
with respect to some fixed axis
the
axis
of
and
the
other coordinates
(say
z\
define the position of the
relative
to
the
this
system
plane
<
azimuth,
having
an ignorable coordinate, and
consequently the corre
sponding integral (which is one of the integrals of angular momentum
above-mentioned) can be used to depress the order of the
system by two
units the equations of motion can
therefore, as a matter of fact, be reduced
the coordinate
<
is
in this
This
way to the 8th order.
Lagrange s memoir already cited) was
1843, and
is
Lastly,
fact
first
(though contained implicity in

explicitly noticed by Jacobi* in
generally referred to as the elimination of the
it
is
two units as in
time.
So finally
nodes."
possible again to depress the order of the equations

by
42, by using the integral of
energy and eliminating the
the equations
of motion
may
be reduced to
6th order.
*
a system of
the
Journ. fiir Math. xxvi.
From the point of view of the theory of Partial Differential

p. 115.
express the matter by saying that the integrals of angular momentum
give
rise to an
involution-system, consisting of two functions which are in involution with each other
and with
and hence the Hamilton-Jacobi partial differential
with 6
Equations, we
may
equation
independent
variables can be reduced to a

partial differential equation with 6 - 2 or 4 independent variables
this will be the Hamilton-Jacobi
partial differential equation for the reduced system.
The Reduction of
342
the
[CH.
xm
Jacobi s equation.
156.
number of free particles in space, which
Jacobi*, in considering the motion of any
Newtonian
the
to
law, has introduced the function
other
each
attract
according
1
i,
m t and m, are the masses of two typical particles of the system, r^ is the distance
is the total mass of the particles, and the summation is extended
between them at time t,
over all pairs of particles in the system. This function, which has been used in researches
of the system, will be called Jacobi s function and denoted by the
the
where
concerning
symbol
stability
*.
be the
We shall suppose the centre of gravity of the system to be at rest ; let (#,-, yf
coordinates of the particle m t referred to fixed rectangular axes with the centre of gravity
The kinetic energy of the system is
as origin.
,
r=Umi (#,*+&* +
z>)
**),
and consequently we have
= (2?H;) X 2m; (X
(2ii) x 2 nii X?
But
(^m x
t}
=2m
% (x
%)
2
,
i,3
where the summation on the right-hand side is extended over every pair of particles in the
= in virtue of the properties of the centre of gravity.
system and we have S.m xi Q,
:
Thus we have
T=
^
*-
"
2 TO^- {(*
- %) 2 + (& - &) a + & -
2
>)
i,j
^,3
where v^ denotes the velocity of the particle

In the same way we can shew that
TOJ
relative to ?ny
V being
If now F denotes the potential energy of the system, the arbitrary constant in
determined by the condition that V is to be zero when the particles are at infinitely great
distances from each other, we have
m * my
s
F=-
The equations
of motion of the particle
IJ
1,1
are
8F
8F
8F
w=-^,
**-- 5.
**-~5,
z f respectively, add them, and sum for all the

Multiply these equations by # f y {
- 1 in the variables, we thus
F
is
since
homogeneous of degree
particles of the system
,
obtain
27/1;
faSt+ytifi +
z i Zi)
=
>
2m
or
or
i (
^
This
is
called Jacobi s equation.

*
Vorlesungen
iiber
Dyn.,
p. 22.
156, 157]
Reduction
157.
343
12th order, by use of the integrals of motion of the
to the
centre of gravity.
We
now proceed
to carry out the reductions which have been

that
it is possible to retain the Hamiltonian form
appear
of the equations throughout all the transformations.
shall
described*.
It will
Taking the equations of motion of the Problem of Three Bodies in the

form obtained in
155,
dpr__W
dqr_dH
dt
we have
dt
dp r
dq r
reduce this system from the 18th to the 12th order,

by use
of the integrals of motion of the centre of
For
this
gravity.
purpose we
perform on the variables the contact-transformation defined by the equations
first to
dW
W = p.q,
where
+ p q + p.q +p q + p + p q + (p, + p + p
+ (P* + ps + Ps) qs + (p + p +p ) q
2
q<?
7)
Interpreting these equations, it is easily seen that

q2 q s
coordinates of
1 relative to
s
(?/, qs qs } are the coordinates of
to ra3 (q 7 q8 q) are the coordinates of
s
(pi, p 2 ,p 3 ) are the
(<?/,
of
momentum
of wi lf
(/>/,
p p
f
dt
where, on substitution of the
[PiP*
are the
relative
momentum
now become
= a#
"
dp r
new
of the system.
138)
,_ 1,4.
12 ..,),
rfp/__w
dt~
^qr
variables for the old,
+ P*P* + P* P* + \Pi* + \ Ps +
2
"
3>p
we have
-p
(pS
+p
- ?/
K?/
The
components
are the components of momentum of w 2 and
components of
<%
(P?, PS Pa) are the
The
q7
contact- transformation used in
157 is due to Poincare, C.R. cxxin. (1896) that used

due to the author, and was originally published in the first edition of this work
(1904).
It appears worthy of note from the fact that it is an extended
point-transformation, which shews
that the reduction could be performed on the equations in their Lagrangian
(as opposed to their
;
in
158
is
Hamiltonian) form, by pure point-transformations. The second transformation in the alternative

Another reduction of the Problem of
160) is not an extended point-transfurmation.
reduction
Three Bodies can be constructed from the standpoint of Lie s Theory of

Involution-systems and
Distinguished Functions cf. Lie, Math. Ann. vin. p. 282. Cf. also Woronetz, Bull. Univ.
:
Kief,
1907,
and
Levi-Civita, Atti del R. 1st. Veiieto, LXXIV. (1915), p. 907.
The Reduction of
344
Since
(//,
coordinates
q s q 9 are altogether absent from

the corresponding integrals are
p = Constant,
can without
loss of generality
xm
[CH.
H, they are ignorable
p = Constant,
We
the
p = Constant.
9
suppose these constants of integration
to be zero, as this only means that the centre of gravity of the system is
taken to be at rest the reduced kinetic potential obtained by ignoration of
:
coordinates will therefore be derived from the unreduced kinetic potential

by replacing p 7 p 8 p 9 by zero, and the new Harniltonian function will be
,
derived from
H in the same way.
The system of
the 12th order, to
of three bodies have
which the
now
been reduced,
equations of motion of the problem
therefore be written (suppressing the accents to the letters)
r__
r__
dt
dt
dp r
may
_
dqr
where
[q*
{(q,
- qtf + (9, - g
2
5)
+ (q 3
This system possesses an integral of energy,
H = Constant,
and three integrals of angular momentum, namely
where
A A A
l}
158.
are constants.
Reduction
momentum and
to
the
8th
order, by use of the
integrals
of angular
elimination of the nodes.
The system of the 12th order obtained in the last article must now be
reduced to the 8th order, by using the three integrals of angular momentum
and by eliminating the nodes. This may be done in the following way.
Apply
to
the
variables
equations
q
the
denned
= dW
dW
^ = Wr
/
Wr
by the
1C*
/\
(r=l>2,...,6),
where
W =Pi
(<?/
+P* (q-A
cos ql
cos qs
~ q*
-q
4
cos qt sin q s
cos g/sin q s
It is readily seen that the
follows
)
)
+p
+p
new
(g/ sin q,
(q 3 sin q s
+q
+q
cos q6 cos qs
cos g6 cos q 5
+p q
) +p qt
)
sin qt
sin qg
f
.
variables can be interpreted physically as
157, 158]
345
In addition to the fixed axes Oxyz, take a new set of moving axes Ox y z
is to be the intersection or node of the
plane Oxy with the plane of the
three bodies, Oy is to be a line perpendicular to this in the plane of the
Ox
three bodies, and Oz
is
to
Then
be normal to the plane of the three bodies.
relative to axes drawn through ??? 3 parallel

q^) are the coordinates of
to Ox Oy
(q3 q) are the coordinates of m.2 relative to the same axes
qs
(<?/,
the angle between Ox and

and p 2 are the components of
is
p and p
p and ps
3
Ox
the angle between Oz and Oz

q
momentum of m relative to the axes Ox
is
p{
Oy
are the components of momentum of

2 relative to the same axes
are the angular momenta of the system relative to the axes Oz
;
and Ox respectively.
The equations
of motion in terms of the
"
where, on substitution in
variables are
dt
dp r
dq r
H of the new variables
Pi
V*
IB
^* *
for
j.
*"
we have
the old,
* a*
./.
j.
j.*>
cosec qt
+p, qt
_o
K Pi fc )8
^V + PS ?/ -
/>/&
+j3 5 ^2 cosec g/
r
~1
PiP
L
+P
-jf>
(q-2
q3
q 5 does not occur in
corresponding integral
) tfa
cot ?/
+ ^sV}
qi q* )
- Pi qi + p
-- T-, ^
-Pa qi + Pa q* - P* q) q*
NOW
p4
q\qi)
is
138)
dp^__dj[
dq^JbH_
dt
new
qt
H, and
- pt q,
is
cot qa
qs cot q K
+p
q 4 cosec q9
+ p, q
cosec q
+p
+p
therefore an ignorable coordinate
q3
the
is
PS
= k,
where
A;
is
a constant.
The equation dq /dt = dH/dk can be integrated by a simple quadrature

when the rest of the equations of motion have been integrated the equations
a
and p 5 will therefore

system of the 10th order
for q 5
fall
=
where
is to
/>/
p =
s
(r
be replaced by the constant k wherever
We have now
(namely
out of the system, which thus reduces to the
k)
made use
and
it
1,2, 3, 4, 6),
occurs in H.
of one of the three

integrals of angular momentum
the elimination of the nodes
when the other two
:
The Redaction of
34(3
momentum
integrals of angular
the
xm
[CH.
are expressed in terms of the
new
variables,
they become
- pi q + Pt qs - p q*) sin q cosec q - k sin ?

- jpsV) cos & cosec + ^ cos ?
-pi q* +Ko
(pa qi
(pz qi
The
?"
values of the constants
axes Oxyz
we
shall choose the axis
momentum
of the system, so that
the special
a;?/-
cos
qs=Ai,
=^
sin ?
depend on the position of the fixed
J. 2
Oz
to be the line of resultant angular

and z are zero
l
69) the constants
(cf.
plane thus introduced
The two
system.
and
J.!
+ P*
+P
cot q
cot q
called the invariable plane of the
is
equations then give
last
k cos
- Pi
=p*qi
(/
p.
+^V - #V
q.2
0.
These equations determine
and p a
q<?
in terms of the other variables,
and
so can be regarded as replacing the equations
dq
_ dH
dpi _
9-ff
"
dt
"
dt
dp 6
d</
The system thus becomes
in the system.
dq r
_dH
dt
~dpr
<W
"
=
~
_d#
dt
rr-12^4^
dq r
where
H = V2m
+ 2m_
/
3
^_
(q q*
j/
-p*qi
q*
q l q,
+ ps q*
^V) cot
(?
& cosec
& cosec ^6
^ cosec
W}
t,
4
.
cot q 6
q*-Pa qi +ps q* -pt qs)
{(Pi
fl
4 /
iV - pi qi + ps q* - Piq-t) cot ^ +
and where, after

its
the derivates of
H have been formed, q
to
is
be replaced by
value found from the equation
k cos qt
Now
in
let
then
= pi qi - pfq* + pfq* - p-iq.
be the function obtained when this value of q6 is substituted

denotes any one of the variables g/, q 2 q3 q t p^, p2 p3 p4
if s
we have
=
~
ds
ds
dq6
ds
158, 159]
But
since
p =
K
0,
we have dH/dq =
ds
/>
therefore
fa
we can make the substitution
in other words,
for q6 in
H before forming the
and thus (suppressing the .accents)
the equations of motion of

H;
Three
are
Bodies
reduced
to
the
of
system of the 8th order
derivates of
Problem
the
= 0, and
347
dqr
_ dff
dt
dH
dp r _
dt
dp r
dq r
where
*"
"
T^
,-
mm
2
(q^
+ qf)
mm
+ q% )
2
(q^
the
{(<?i
qs )
+ (q
q4 J
have simple physical interpretations

quantities occurring in
is
twice
the
area of the triangle formed by the bodies: and
qiq4 )
2m
is
Many of the
thus (q z q
mm
mm
moment
(/
/I
of inertia of the three bodies about the line in which the
plane of the bodies meets the invariable
plane through their centre of
gravity.
H differs from the value of H when k
It is also to be noted that this value of

terms which do not involve the variables
p
l ,
p.2 ,
p3 p i
,
is
zero
by
these terms in k can therefore be
regarded as part of the potential energy, and we can say that the system differs from the
corresponding system for which k is zero only by certain modifications in the potential
It may easily be shewn that when k is zero the motion takes
energy.
place in a plane.
159.
Reduction
to the Qth order.
The equations of motion can now be reduced further from the 8th to the
6th order, by making use of the
integral of energy
H = Constant,
and eliminating the time. The theorem of 141 shews that in
performing
Hamiltonian form of the differential equations can be
this reduction the
conserved.
As the
be given here in
actual reduction
is
not required subsequently,
it will
not
detail.
The Hamiltonian system of
the 6th order thus obtained is, in the

present
our
the
ultimate
reduced form of the equations of motion
of
knowledge,
of
the general Problem
of Three Bodies.
state
The Reduction of
348
[CH.
xin
Alternative reduction of the problem from the 18th to the 6th order.
160.
We
the
of the general problem of three

give another reduction*
bodies to a Ilarailtonian system of the 6th order.
shall
now
Let the original Hamiltonian system of equations of motion

be transformed by the contact-transformation
155)
where
W = Pi
(q*
~ ft) + p*
nil
(ft
^2
- ft) + P*
(ft
- ft)
The integrals of motion of the centre

of the new variables, can be written
when expressed
of gravity,
= q* = q* =p? = p* = p* =
in terms
qr
and consequently the transformed system
new
pressing the accents in the

dqr
dt
>
only of the 12th order: sup
is
variables, it is
= dH
dp^
_<^_
dt
dp r
(r=l,2,...,6),
dqr
where
H =~ (p? + pf + P/) +
2~~,
+ P? + Pe2 ) -
(PS
-f
m.2 (qf
+ q, +
2
q/}
~
\2
*M
2
.
+ m,/
?4
t
i"
*t<2
variables may be interpreted physically in the following way

Then (q lt q2 q s ) are the
the
centre of gravity of m, and
be
2
The new
Let
Due
to
Kadau, Annales de
Ec. Norm. Sup.
v. (1868), p. 311.
160]
projections of m^n^ on the fixed axes, and (g4

of
Further
3 on the axes.
q5
349
qs ) are the projections
Gm
^=Pr
and
(r-1,2,3),
l*
(r4,5,6).
^=pr
The new Hamiltonian system clearly represents the equations of motion

of two particles, one of mass p at a point whose coordinates are (q lt q2
q3 \
and the other of mass /* at a point whose coordinates are (g^, q q6 ); these
,
s>
supposed to move freely in space under the action of forces

derivable from a potential energy represented by the terms in
which
are independent of the ps.
have therefore replaced the Problem of
particles being
We
Three Bodies by the problem of two bodies moving under this system of
forces.
This reduction, though substantially contained in Jacobi s* paper of
1843, was first explicitly stated by Bertrandf in 1852.
We
plane
suppose the axes so chosen that the plane of xy is the invariable

the motion of the particles fj, and fjf, i.e. so that the angular
shall
for
momentum
of these particles about
line in the plane
any
Oxy
is zero.
Let the Hamiltonian system of the 12th order be transformed by the

is defined by the equations
contact-transformation which
^=
where
W = (p
+
dW
= dW
0-^-V-X
Wr
sin qs
+p
cos g5
q^ cos q3
+ q,
sin q3 {(p 2 cos q s
-p
(p s sin qs
+p
cos qs
q2 cos qt
+q
sin qt {(p s cos q*
-p
sin q,
sin qt
)*
+p
2
3
}%
+ p^
The new
variables are easily seen to have the following physical inter

is the length of the radius vector from the
qi
pretations
origin to the
the radius from the origin to
is
particle //,, q.,
q3 is the angle between
:
<//
and the intersection (or node) of the invariable plane with the plane through
two consecutive positions of q^ (which we shall call the plane of instantaneous
motion of /A), 4 is the angle between q.2 and the node of the invariable
plane
on the plane of instantaneous motion of pf, q^ is the angle between Ox
and the former of these nodes, q^ is the angle between Ox and the latter of
</
these nodes, pS is pq\,p is ^ q^ps is the angular momentum of p round

the origin, pl is the angular momentum of /// round the origin, ps is the
angular momentum of /A round the normal at the origin to the invariable
plane,
and p s
is
the angular
The equations
momentum
of motion in their
of
/A
round the same
new form
are
line.
138)
__dH
_
dt
Journal fur Math. xxvi.
"
dp r
p. 115.
dt
dqr
f Journal de math. xvn. p. 393.
The Reduction of
350
the
[CH.
supposed expressed in terms of the new variables.

system be transformed by the contact-transformation
where
Let this
is
dW
pr=
dW
(r
**57
W"
xm
"
6)>
where
W=
q*"
- p* ) +
(p*
The equations
pi
qi
q*"p*
q*"p*
+ qt p*.
now become
of motion
dqr
+# ) +
(p*
<i
dt
dH
dPr
-_dH
"_
"
"
dt
~3q r
as may be seen either by expressing

does not involve q 6
in terms of the new variables, or by observing that q
depends on the
while
none
of
coordinates
the
other
of
the
axis
chosen
Ox,
position
arbitrarily
But
",
"
fi
depend on
"
p6
where
A;
We
this quantity.
is
=-
have therefore
dH/dq,"
= 0,
p =
"
so
k,
this is really one of the three integrals of angular

Substituting k for p6 in H, the equation
a constant
"
momentum.
q 6 "=dH/dk
can be integrated by a quadrature when the rest of the equations have been
so the equations for p 6 and q 6 can be separated from the system,
solved
which reduces to the 10th order system
"
"
"
_
~
dqr
dt
We
have
still
to use the
these, when expressed

represented by
~dp7"
two remaining integrals of angular
in terms of the
= 90,
"
25
new
is
_,
momentum
variables, are readily found to be
Jcp 5 "=p s "*-pr-;
no arbitrary constants of integration enter, owing to the
xy
"
dq r
dt"
fact that the plane of
the invariable plane.
The system may
therefore be replaced by these two equations and the
equations
"
dq r
_ dH
dfr"
3#
"
"dt
"
where, in this last
set,
qs
dt
~dpr"
dq r
can be replaced by 90 before the derivates of

"
2
2
have been formed, and p 5 is to be replaced by (p3
p4 )/k after the
denote the function derived from
Let
have been formed.
derivates of
and
let s denote any one of the
for p 5
substitution
this
making
"
"
by
variables
",
q",
#>",
qa
",
q",
p",
8^ _ 877
p",
dH
pa
",
dp,"
p"
= dH
then we have
.
.
dp,"
"
ds
ds
dp &
ds
ds
ds
= dH
ds
160, 161]
and
it is
therefore allowable to substitute for
H before the derivates of
in
p"
351
have been formed. The equations of motion are thus reduced to a system
of the 8th order, which (suppressing the accents) may be written in the form
dt
dt
dp r
dq r
where, effecting in
H the transformations which have been indicated, we have
H-
(p
(p*
1)
kz -p,?-p?
2ra 2 <j/
1
""*-
-- smg smg*
.
2m! ft g a
~m<,m s
-\q^+m
---p,--p
^--^2
Mcosftcosft
l + m^\
(*
5*-
sin ft sin ft
2p,pt
+7
**J
The equations of motion may further be reduced to a system

6th order by the method of 141, using the integral of energy
Constant
As the reduction
and eliminating the time.
of the
is
not required subsequently,
it
be given in detail here.
will not
The problem of three bodies in a plane.
161.
The motion of the three particles may be supposed to take place in a

plane, instead of in three-dimensional space; this will obviously happen if the
directions of the initial velocities of the bodies are in the plane of the bodies.
This case
is
known
as the problem of three bodies in a plane

we shall
motion to a Hamiltonian system of
:
now proceed
to reduce the equations of

the lowest possible order.
Let
(</!,
(^5.
q z ) be the coordinates of
the coordinates of
<7e)
of the motion;
^
(r + 1).
and
1}
referred to
q 4 ) the coordinates of ??? 2 and

any
Oy in the plane
(q 3
fixed axes Ox,
pr = m k qr) where k
let
The equations
denotes the greatest integer in

155)
of motion are (as in
dqr_*ff
dt
dpr
dpr__dH
dt~
l, -,..., Oj,
dq r
where
^~to +jtf)+0tf+i* H(^

f
- tfj) +
2
g3
These equations
will
- ft) ~ - M! wi a
2
(ft
{(q,
now be reduced from the 12th
- q Y + (ft - g )
3
"
2
}
to the 8th order,
by
Perform on the
using the four integrals of motion of the centre of gravity.
variables the contact-transformation defined
by the equations
dW
dW
The Reduction of
352
the
[OH.
xm
where
W=p q
l
+p
2 q*
+ pq + P*q* + (Pi+P* +P*) ft + (P* + P* +P) ti

a
lt is easily seen that

Wg parallel to the
(<//,
through
relative to
the same
are the coordinates of
ft )
and (p 6
are the components of
of the system.
in | 157, the equations for

5
q 6 p 5 pi disappear from the system
in the new variables) the equations of motion
accents
the
(suppressing
As
and
relative to axes
original axes, (pf, p.2 )

the components of momentum of
momentum
2
fixed axes, (q a qf) are the coordinates of
the
relative
to
of
coordinates
the
are
3
axes, (q s q )
1} (p 3
are the components of momentum of
p t ) are
<?
reduce to the system of the 8th order,
dq r
= 9#
dt
ty
dH
Wr
dp r _
-dl-
where
mm
2
+ ft )
2
(q/
~?
- m,m, (q? + q
)~
h*
{(?i
?s)
+ (q* ~ qtf]
"
an ignorable coordinate,
Next, we shall shew that this system possesses
which will make possible a further reduction through two units.
Perform on the system the contact-transformation defined by the equa
tions
dW
^ = Wr
dW
^Wr
where
W = ptfi
cos qt
+p
qi
sinqS+ps (qj cos q4
- q sm ^ ) + P* (ft
a
sin q t
+ ft cos g/).
of this transformation is as follows g/ is the

physical interpretation
are the projections of 2 3 on, and perpendicular to,
distance
1
s
ft and qi
m z l and the axis of x pi is the component
between
the
is
3
1
angle
q4
of momentum
of momentum of
3 m,
1 along
p* and p 3 are the components
is
the
and
to
t
angular momentum
p
of m.2 parallel and perpendicular
The
mm
mw
mm
m.^
of the system.
The
become
when expressed
"
dt
dp r
in terms of the
new
variables,
dt
where
-\f*v
~w
ik
-*7+rt~*.
an ignorable coordinate the corre

this can be interpreted as
a constant
The equation q4 = dH/dpi
of the system.
not contained in H,
= k, where k
integral is p 4
Since qt
sponding
353
161, 162]
is
is
it
is
the integral of angular momentum

can be integrated by a quadrature when the rest of the equations have been
and
disappear from the system.
integrated and thus the equations for p t
the equations can therefore
Suppressing the accents on the new variables,
be written
;
<?/
qr
dt
Pr
= __
dt
dp r
(r
= l,2,
3),
dqr
where
H-
<?,
-#)
2
-I-
9s
a system of the 6th order

it can be reduced to the 4th order by
the process of
141, making use of the integral of energy and eliminating
This
is
the time.
The
162.
restricted
problem of three
bodies.
special case of the problem of three bodies, which has occupied a

place in recent researches, is the restricted problem of three bodies ;
Another
prominent
this may be enunciated as follows
bodies $ and J revolve round their centre of gravity, 0, in circular

A third body P,
under the influence of their mutual attraction.
Two
orbits,
without mass
(i.e.
their motion),
such that
moves
of three bodies
is
to
it is
attracted by S and J, but does not influence

as 8 and
the restricted problem
same plane
in the
</;
determine the motion of the body P, which
is
generally
called the planetoid.
Let
and
Take any
be the masses of
8 and
J,
m,
m.2
SP
JP
and write
fixed rectangular axes
OX, OF, through 0, in the plane of the

be
the
motion let (X, Y)
coordinates, and (U, V) the components of velocity,
of P.
The equations of motion are
:
dt*~dX
or in the
dt
\v.
atr
d7
dH
dt
dv
H= % (U
D.
~dY
Hamiltonian form,
dX = dH
where
dt*
dV__3H
<W__dH
dt"
dx
~dt~
ar
+ 7 ) - F.
s
23
The Reduction of
354
[OH.
but also of
and
a function not only of
of
the
an
Constant is not
system.
integral
Since
H=
the
is
t,
Perform on the variables the contact-transformation which
xm
the equation
denned by
is
the equations
_3E
_^W_
V _3W
=
=
Xv _dW
dx
~3V
~dU
dy
cos
=
sin
nt
+
W V (x cos nt y sin nt) +V(ae
nt),
y
where
and n
is
the angular velocity of 8J.
dx
_dK
~
eft
8wT
du
fy
I (u
+ n (uy -vx)-F;
x and y are the coordinates of the planetoid referred

OJ as axis of as, and a line perpendicular to this through
is now a function of x and y only, so K does not involve t
at once seen that
moving
line
as axis of y.
to the
explicitly,
and
K = Constant
an integral of the system
called the Jacobian integral* of the restricted
it is
problem of three bodies.

Another form of the equations of motion
last
_dK
dW
K = H - -^
=
is
dv
=
dt~
dx
~di
~dv
~dt
_dK
=
~
<fy_c_K
~~
where (138)
it is
The equations become
obtained by applying to the
is
system the contact-transformation
= dW
y=
p
W>
W=q
where
The new
variables
3W
dW
-^>
(u cos qz
^^
P2
+ v sin q
= dW
2 ).
be defined directly by the equations
may
and the equations of motion become

dq-r
_ dH
~
~dt
dp,.
H=
where
Another formf
is
dW
Pr=Wr
W =p
*
dt
=12)
dqr
+ ~ -
np 2
- F.
obtained by applying to these equations the contact-
transformation
where
(pc
_ _ dH
dp,. =
q2
\P,
A u^n
-(P^-P^}
dv
"
p7*\)
Pl
Jacobi, Comptes Rendus, in. (1836), p. 59.

la Mec. Celeste.
Adopted by Poincare in his Kouvelles methodes de
^
.
-,
dW
2
/.
j
^^Wr
162]
where u denotes a current variable of integration.
355
These equations may be
written
p.
J
g/=arccos-i
and
it
is
2(7
-^ +
easily seen that
~4*
is
Q \*
- -^Vj
its
#2
=a -arccos
2
mean anomaly
the
<?/
would describe about a
it
projected from
is
2
2
which
ellipse
J)/
of the planetoid in the

body of unit mass at 0, if
fixed
instantaneous position with
its
instantaneous velocity
the longitude of the apse of this ellipse, measured from
-e )p,
2
where a
OJ; p
is a*,
q2
and
the semi-major axis and e is the

eccentricity of
this ellipse.
= Constant is an integral
does not involve t explicitly, so
of the equations of motion, which are now
is
{.(!
is
dt
If
we take the sum
denote these masses by
= aff
dq^
~dp
3ff
dp r
"
~dT~~
dq?
1}
2^
8 and J to be the unit of mass, and

and p respectively, we have
of the masses of
1
/u,
~SP
This
is
an analytic function of p^,
with the period

suppose p to be
2?r.
zero];
11
1\
Thus
which is periodic in qf and q2

p2 q, qa
Moreover, to find the term independent of fj, in H, we
since SP now becomes q lt we have
discarding the accents, the equations of motion of the restricted

may be taken in the form
finally,
dqr
_ dH
dt ~~dp r
where
while
dpr _
~dt~~
dH
~3^
H can be expanded as a power-series in p in the form
H H
lt
...
are periodic in q 1
The equations
and q 2 with
,
of this 4th order

system
may
the period 2?r.
be reduced to a Hamiltonian
= Constant and elimina
system of the second order by use of the integral

tion of the time, as in
141.
232
*The Reduction of the
356
problem of n
[CH.
xm
163.
Extension
Many
of the transformations which have been used in the present chapter

problem of three bodies can be extended so as to
to the
bodies.
in the reduction of the
apply to the general problem of n bodies which attract each other according
to the Newtonian law.
In their original form, the equations of motion of
the n bodies constitute a system of the 6?ith order this can be reduced to
;
12)th order, by using the six integrals of motion of the centre of

gravity, the three integrals of angular momentum, the integral of energy, the
elimination of the time, and the elimination of the nodes.
the (Qn
The reduction has been performed by

p.
T. L. Bennett, Mess, of Math. (2) xxxiv. (1904),
113.
1.
If in the
problem of three bodies the units are so chosen that the energy integral
2
i(vi
1111
---
+ V+*V) =
?*23
where r 12
+r +r
3l
12
Vi and v 2 and if r is a
shew that the greatest possible value of the angular momentum of the
the distance between the bodies whose velocities are
is
positive constant,
system about
its
is
centre of gravity
is
f\
2r.
1893.)
I,
2.
In the problem of three bodies, let * be Jacobi s function, let Q be the angle
between any fixed line in the invariable plane and the node of the plane of the three bodies
on the invariable plane, let i be the inclination of the plane of the three bodies to the in
be the area of the triangle formed by the three bodies. Shew that
variable plane, and let
77
dQ,
j_
"*
<fc
di
n*
dt
where k
is
the angular
momentum
of the system round the normal to the invariable
(De Gasparis.)
plane.
Let the problem of three bodies be replaced by the problem of two bodies p. and /*
160 let qi and q 2 be the distances of p. and p. from the origin let q 3 and q 4 be
the angles made by q l and 2 respectively with the intersection of the plane through the
bodies and the invariable plane let p and p 2 denote pqi and pf q-2 respectively and let p3
and
be the components of angular momentum of p. and p. respectively, in the plane
3.
as in
<?
pi
through the bodies and the
origin.
dq r
di
where
H= Constant
is
Shew
that the equations of motion
_dH
tyS
the integral of energy.
dp r _
may
be written
dff
}
~di~~dq r
(Bour.)
357
163J
Apply the contact-transformation defined by the equations
4.
qi = {(V*
92
+ (?5 - ?8) 2 + (26 - ?9) 2 }*,

2
+
2
= {(?7 -
?7)
<?i)
(<?
=m
26)
+m
=
"
&i (?i
Pr = s
A;=0
(where
stands for
V^l
and
+ iq$ + b 2 (g-4 + t
X-
(r-0,
1, 8, ..., 8),
O^r
i,
&
s>
i>
^2, b s
c l5 c 2 , c 3
are any nine constants which
satisfy the equations
a1
to the
+2 + =
3
^1
+ ^2 + ^3 = 0,
^ + 02 + 03 = 0,
Hamiltonian system of the 18th order which
a 2 &3-a 3 & 2 =l),
155) determines the motion of the
three bodies.
Shew
that the integrals of motion of the centre of gravity are

q&
Shew
zero,
further that
when the
q-i
= q = p& = PI =pa =
invariable plane
and that the integral of angular
is
taken as plane of xy, the variable p5 is

the normal to the invariable plane
momentum round
is
p i qi
where k
=k,
is
a constant.
Hence shew that the equations reduce to the 8th order system
dq,!_dff
~dt~-ty?
dpr _
~dt~
dff
fy?
where
Reduce this to a system of the 6th order, by the theorem of
141.
(Bruns.)
CHAPTER XIV
THE THEOREMS OF BRUNS AND POINCARE
Bruns theorem.
164.
Statement of the theorem.
(i)
We
155) that the problem of three bodies possesses 10 known

the
six integrals of motion of the centre of gravity, the
integrals namely
three integrals of angular momentum, and the integral of energy these are
have seen
generally called the classical integrals of the problem.

algebraic integral, i.e. is of the form
/(<7i,
<?2,
.,
q 9 ,Pi,P*>
->
Pa,
Each of them
is
an
t)= Constant,
where / is an algebraic function of the coordinates

and of t.
(q 1} q2
-,^9)
$9, PI,
Efforts have frequently been made to obtain other algebraic integrals of

the problem of three bodies independent of these (i.e. not formed by combina
tions of them), but without success; and in 1887 Bruns* shewed that no
such new algebraic integrals exist
in other words, the classical integrals are

problem of three bodies.
the only independent algebraic integrals of the
It may be remarked + that the non-existence of algebraic integrals does not necessarily
imply great complexity in a system. One of the simplest of differential equations, namely
the linear differential equation with constant coefficients
has no algebraic integral except when
î//n 2 is
-^ =
integral
a rational number, in which case the
first
Constant
can be transformed into an algebraic integral.

(ii)
Expression of an integral in terms of the essential coordinates of the
problem.
We
shall
now proceed
to a proof of
integrals which do not involve
Bruns
result, considering first those
explicitly.
* Berichte der
Kgl. Sachs. Ges. der Wiss. 1887, pp. 1, 55
Ada
Math.
Forsyth, Theory of Differential Equations, Vol. in. (1900), Ch. xvn.

t Cf. K. Bohlin, Astron. lakttagelser och Unders. a Sfockholms Observ.
xi.
p.
Cf. also
25.
ix. (1908),
1.
The Theorems of Bruns and Poincare
164]
The equations
of motion of the problem
may
359
160) be written in the
form
_dH
dqr
dH
Wr
dp r _
~dt~
W-fa
where
H=T-U,
T=
+ p + P^ +
2P
(P
5?
^ + p ^ + P^
)-*
+q
q6 )
+
\
_m
We
shall write
/,
(m + m )
2
m
yu, 5
//, 6
=
y
T=
so that
Let the coordinates of the three bodies be (g/, g/,

and let mk qr = p r where k denotes the greatest
the integrals whose existence
(qi
<
where a
(g/, q s
q6
(g/,
),
integer contained in
to
discuss
are of the form
propose
we
q*,
-,
q 9 Pi,
,
gg
(r
g/),
+ 2)
p 9 ) = a,
an algebraic function of
an arbitrary constant and

is
its
The
formulae
of
160
enable
us
to
the
variables
arguments.
express
q*,
qLp\,
,p&) as linear functions of (qlt q2 ..., q 6 p 1} ..., p 6 ): we
shall therefore, on
making these substitutions in the integral, obtain an
is
(j>
(<?/,
equation
/(?i, q2
If the integral
gravity,
/ will
<
is
evidently reduce to a constant

.
An
q,pi, ...,p 6 )=a
integral
..................... (2).
of the integrals of motion of the centre of
compounded
function of the variables (q1}

existence of integrals, such as
(iii)
-..,
if not,
We
q6 p
p s ).
of the equations
l
. . .
(2),
must involve
the
will
be an algebraic
have to enquire into the
(1).
momenta.
W e shall first shew that an integral such as (2) must involve

T
quantities p,
i.e. it
For suppose,
if
cannot be a function of
(5-^
...,
6)
some of the
only.
possible, that the integral, say
does not involve (p lt
p.2
...,
Q_
6 ).
j.
Differentiating with respect to
df
_ 4 d/jy
t,
we have
360
[CH. xiv
and therefore the equations
J^r =
(r~l,2,
...,6)
dq
must be
so
satisfied identically; that
is,
does not involve (q l} q2
...,
q6 ), and
a mere constant.
is
Only one irrationality can occur in the integral.

As the mutual distances of the bodies are irrational functions of
function of these
q6 } the function U will be an irrational
(q lt g 2j
it is easily
three
mutual
of
the
s
the
sum
variables.
distances,
Denoting by
seen that the mutual distances can be expressed as rational functions of the
(iv)
seven quantities (q l} q 2
in the
q6 s); in other words, the irrationalities involved

all capable of being expressed by means of the
...,
mutual distances are
irrationality of s
function of (q 1}
Now
we may
q.2
q6
...,
variables (q 1} ...,q s ,pi,
expressed as a rational
is
s).
the function
therefore suppose that
algebraic, but not necessarily rational, in the
is
...,
6)
let
the equation (2) be rationalised, and let

powers of a, so that it becomes
w~ 2
the resulting equation be arranged in
+ am -
am
1
(f) l
(q l ,q,, ...,qe ,pi,
...,
p )+a
6
(gi,
2
</>
</>m(?i,
9s, PI, ...,p*)
>
...,q 6 ,pi,
...,2> 6 )
...
...(3),
If this
fan are rational functions of (q lt ..., qK p lt ..., p6 ).
fa, fa,
is reducible in the variables (q lt ..., q 6 p lt ..., p 6 s}, i.e. if it can be
equation
into other equations, each of the form
where
>
decomposed
~
a + a ^(q
1
,...,q 6 ,p l ,...,p
e>
s)
...
+ î(q
,...,q 6 ,p 1 ,...,p 6 ,s)
= 0... (4),
are rational functions of (q

q p lt
p6 s), then one
of these last equations will give the value of a which corresponds to equation
where
ty z
l ,
fi ,
>|r
equation instead of (3). As the type of

the type represented by (3) as a
equation represented by (4)
be
a
to
we
shall
given by an equation of the form (4),
suppose
particular case,
and we
(2),
shall consider this
includes
irreducible in (q 1}
...,
qs p lt
...,
p6
s).
Differentiating with respect to t, and using equations

al
+ a^(^,H) + ... + (^ ,H)
l
where
(ty r
We
H}
(^H}
we have
(5),
denotes as usual the Poisson-bracket of tyr and H.
shall first
functions of (q 1}
suppose that the expressions
...,
l}
(5) have one or more
and
(1),
are not
p s),
common roots
...,
a,
(//>,
all
H}, which are rational
zero.
Then equations
and consequently equation
(4)
(4) is
and
reducible in (q 1} 2
Pe, *); but this equation is irreducible,
q&, PI,
are
all
and
the
is
this
therefore
inadmissible,
quantities (-v/rr H}
hypothesis
>
<?
zero.
This implies that
all
the coefficients (^r
in equation (4)
can be com
hence the integral
l
ifr
2>
..., tyi)
are integrals of the equations (1): and

pounded algebraically from other integrals, which are rational functions of
(?i,
...,%
Pi,
,&>*)-
164]
Expression of the integral as a quotient of two real polynomials.
(v)
We
need therefore henceforth only consider integrals of the type

?,... q,pi, ...,p,s)
=a
..................... (6),
where /is a rational function of the arguments indicated. The form of /can
be further restricted by the following observation. If in the equations of
motion we replace qr p r t by qr k2 pr k~ and tk3 respectively, where k is any
l
constant, the equations are unaltered.
made
in equation (6), this equation

be.
If therefore these substitutions are
must
still
be an integral of the system,
whatever k may
Now /
its arguments
it can therefore be ex
two functions, each of which is a polynomial in
a rational function of
is
pressed as the quotient of
When
in these polynomials we replace q r p r s

the
function / will (on multiplying its
sk-, respectively,
and
denominator
an
numerator
by
appropriate power of k) take the form
(qi, $2,
>
2
by q r k
df/dt
PI,
p r k~
where (A
is
(?e>
zero,
>P<>>
s )-
lt
...,
q)
are polynomials in (q 1}
qe
...,
p lt
...,p 6
s).
Since
we have
"fÊfr*-**
f
Now
is
/7 4
U/-iln
=
(q+p +
dt
it is
~A,
powers of k in this
/-77?
.
LLJJn
dB
dA,
dB,
dB
5,^
a
8
l) equations are equivalent to the

_
from which
and therefore
dA
"
arbitrary, so the coefficients of successive
equation must be zero
These
dt
A p ~dT
system
5g
~dt
evident that each of the quantities
rfi
362
[OH.
xiv
an integral and thus we have the result that any integral such as f can
compounded from other integrals, which are of the form
is
be
6r 1? G.z is a polynomial in its arguments, and is

a
merely multiplied by
power of k when the variables qr p r s are replaced
2
We need therefore only consider integrals of
k~
sk2
k
respectively by qr
pr
where each of the functions
this form.
It
may
further be observed that the functions
of generality, be taken to be free from imaginaries.

the real and imaginary parts of an integral
and G2 may, without loss

For if P and iQ denote
P + iQ = Constant,
dP
we have
-jat
.,
-~ = 0,
+ i.dQ
.
..
,,
identically.
ctu
Since the differential equations are free from imaginaries, it follows that
dP/dt and dQ/dt are free from imaginaries and so dP/dt and dQ/dt must be
zero separately.
Hence
and Q are themselves integrals, and every complex
:
We
integral can be compounded from real integrals.

forth assume that Gi/G z is free from imaginaries.
Derivation of integrals
(vi)
hence
numerator and denominator of the
the
from
shall therefore
quotient.
the case that the function GI is resoluble into a product of

irresoluble polynomials in (p lt p2 ...,p6 ), the coefficients in these polynomials
It
may be
q,
being rational functions of (qlt q 2
suppose that it is repeated A, times in
,
factors of
When G
G
is
1}
so that
irreducible,
G^^ X
we
T^\TT}
have GI
shall of course
The equation
be such a polynomial, and

and let x denote the remaining
Let
s).
= ty,
and % =
1.
at \Cr 2 /
1 d-y
\dilr
-r -TT H----JT
vr
dt
dt
dG 2 =
TT
JT
G2 dt
U,
dG.2
Now
(p 1} ...,p 6 ),
factor in
in
polynomial in (p lt ...,p 6 ), and -v/r is also polynomial
of order less by unity than the order of d-ty/dt.
Also, ty has no
d^fr/dt is
common with G2
Hence we
or %.
dG
\G
dt
see that
dx
dt
164]
must be a polynomial
by co then we have
in (p
...,
of order unity
p 6 ),
363
denote this polynomial
It
of
. . .
an equation of this kind.
factors of
G by
the equations they satisfy be
let
JL^-
then we have
dG
where
(ft,
Denote the various
so that
i/r",
and
be shewn in the same way that each of the other irreducible factors
may
satisfies
"
dty
<ty
"
dt
d^r"
a polynomial in (p lt ..., p e ), of order unity, and rational in

Thus G! satisfies the equation
...,q ,s).
is
&>
dG =
l
and therefore
GJGz
(since
an
is
integral),
also satisfies the equation
~dt~
As
and
l
2 satisfy the same differential equation, we shall in future use
to denote either of them: so 9 is a real
polynomial in (p 1} ...,p 6 ft, ...,q G ,s),
,
which
Now
2
2
by qr k p r k~\ sk
respectively
o><.
merely multiplied by a power of k when
is
<f>
m=
~I
9
we
9=
the equation
satisfies
~TT
ac
ft.,
pr
are replaced
since
= ^ T
1~
9 V^ft- A
^~
when
this
by
r =i
>
opr oqr /
3
substitution is made. It follows
multiplied
that &) cannot contain a term independent of (p lf ...,p 6 ), since such a term
would be multiplied by an even power of &;
is therefore of the form
see that
&>
is
&>
where each of the quantities w r

(ft,
is
homogeneous of degree
in the quantities
...,?,).
Further,
one of the terms in
let
be of order
</>
n in
(ft, ...,
order n
&
in (ft,
power of k
s),
...,
while another term

<?
s):
is
m in (p
m in
of order
lt
,p 6 ) and of order
(p l} ...,p 6 ) and of
..
since these terms are multiplied
when the above
substitution
is
by the same
made, we have
m + 2n = m + 2ri,
so
m m
is
an even number.
Hence
</>
can be arranged in the form
The Theorems of Brans and Poincare
364
[CH. xiv
denotes the terms of highest order in (p lt ..., p6 ), 2 denotes terms

by two units in (p ...,p s ) than these, and so on: and each of
where
<.
</>
of order
less
the quantities
(pi,"-,
We
and
Pe)
<f)
also in (#1
now shew
shall
a polynomial in (p l} ...,p s q l} ...,q K
is
qt
homogeneous
s),
in
s}.
that when
does not involve
s,
<
can be made into an
</>
integral by multiplying
For suppose that
by an appropriate rational function of(ql9
it
does not involve
...,
<?).
the equation
s:
</>
tty
-+ -+
or
gives,
......
wxp!
+wp
2
4- ...
ft
on equating the terms of highest degree in (p
I p,.^0=
r=l
Now
may
<^>
>
l>
...,p 6 ),
^r oq r
contain
as a factor
in order to take account of this case,
does not contain p6 as a factor, and where as a

where
k
= 0, $ =$
in the
for
have
k
case
we
Substituting p6
may
special
differential equation, it becomes
write
p<f
<f)
<f>
<j>o
<
<j>
which do not involve j; 6 equating the

Let
denote those terms in
terms which do not involve p 6 on the two sides of this equation, we have
"
<
^>
It
may be
this case
"
that
is
</>
a mere function of q lt q 2 ...,q K say equal to

,
R;
in
we have
3R
~^ =
a) r
RD
(r
= 1,
n
-\
2, ..., o)
/ardjV
or
fji r
a) r
(/n.r
d^g
Supposing next that

involve
it
"
<f>
may
= p^fa
"
where
(r
l, 2, ..., o),
1, 2, ..., 5).
_.
C/g r
a>
r)
(/*&))
(r, s
c;^
does involve some of the quantities (p 1} ...,p 5 ),

to take account of this case, we write
a factor:
as
"
<
XI
o
and therefore
1 9^2
= -^^-
"
<
becomes
r=\
does not involve p 5 as a factor.
The equation now
164]
iv
365
"
which do not involve p 5 equating the

denote the terms in $
terms which do not involve p s on the two sides of this equation, we have
Let
<
r=l
Oq r
f^r
Proceeding in this way, we ultimately arrive at the alternatives, that

either
or else a function ty exists, which

geneous in q l} ..., q 6 and also in p l}
is
mere powers of p and p2 and which

l
Now
a^
fr
dq-2
3^
/MI
let
+1
w =
l
and p 2
The
quantities a,
l+l
"^
W1 =
>
d^j
db
7
/A 2 o
0^2
last equation,
are polynomials in (q 1}
b, c, ...
on the two sides of the
da
I
--
fra
common
2>
the differential equation
= apj + bp + cp^p.* +
ty
equating coefficients of p^
we have
satisfies
polynomial in q lf ..., q 6 PI, p homo

and is free from any factors which are
q.2 , ...,
q 6 ): they
may have
polynomial factor Q, so that
a
Let
-v//
a Q,
= a pj -f
so that
= b Q,
etc.
&X + c p^ps +
-^r
Q\Jr
Then
/
ft
)
(
Q/AJ 9g 1
^j
Ji
where
Pi/*i
The
)
wz
say,
1
^ p/- + 2 -g- =
2
3^
y^
-\/
9g 2
(&>!
p +
1
2 } i/r
a>,p.
left-hand side of this equation is a polynomial in (c^, q.2 ..., q K

if a/ contains 9^ then tw/ contains a
or some factor of it, as
,
this
factor.
tut
a denominator.
But
2 j5 2
da
&>!
so
Pi,pz)
+ 6)
Hence
ty
must contain a
or
some
factor of
inconsistent with the supposition that a, b

Hence a cannot involve q
and therefore
is
is zero.
&>/
...
it,
as a factor.
have no
is zero.
common
Similarly
366
rp,
Thus
&>,
and therefore
which
dQ
dQ
^r-
-7^
to 2
^
d
the same as the alternative previously noted
is
[CH. xiv
so this equation
is
true in any case.
we can shew
Similarly
in general that
<
^-^.OM*X
and hence we may write
where
is
some
rational function of (q 1} q 2
q6 ).
...,
Thus we have
p +
a} 1
o)^p 2
...
co 6
4 pr
p = Z=
-
pr
].
>
dR
1
-p ~
"
^(?r
r = l .R
d
(/>
dt
^ = Constant.
and therefore
Thus
can 6e transformed into
<
3gV
appropriate rational function of q lt
q.2
constant,
...,
q6
by multiplying
it
namely 1/.R; which
by an
the
is
required result.
If therefore the terms
in
</>
and
of (q\,
G
^2.
G and
l
G.2
do not involve
s,
we can transform
into integrals, by multiplying them by appropriate rational functions

a nd hence, if it can be shewn that the terms
in G 1 and G2
9e)
<
>
we
have the result that any algebraic integral of

the problem of three bodies can be compounded from integrals which are
polynomial in (p lt p 2 ...,p 6 ) and rational in q 1} q2 ..., q6 s.
do not involve
s,
shall
Proof that
(vii)
The
We
<f>
case in which
shall
<
does not involve the irrationality
involves s
is
s.
not included in the above investigation.
however now proceed to shew that no
real function
<
which
satisfies
an equation
can involve
not involve
s,
and hence that the functions

</>
so that the above result
is
occurring in our problem do
quite general.
164]
367
For suppose that a function

exists, which involves s and satisfies the
above differential equation. When the 8 values of s are substituted suc
take a number of distinct values let these values be
n will
cessively in
<
<
<f>
denoted by
>
<$><>
</>
they satisfy equations of the form

"
r - Q
p
^
- - d(f)
=
2,
r=l Pr 3q r
where
<<>
a/,
fa",
p r4
2,
to
9o
o>
when the
respectively are substituted in
",
<f>
Let
4>
-5
r=l Pr fyr
are the values of
...
= a)
d(f>
>
values of s corresponding to
it.
"....
<>
Then we have
|
<f>o"
fyr
//i
&)+&>
4- ...
mQ,
where
a linear function of (pi,p 2

of
functions
(q 1} q 2 ..., q6 ).
II is
...,p6 ), the coefficients being rational
Now
(q lf q z
So we
O, from the manner of
its
formation,
is
a rational function of
...,qK ), not involving s and it is clearly a polynomial in (p lt p 2 ...,p s ).

can apply to
the results already obtained, which shew that (on
:
<l>
by some rational function of q lt q 2
multiplying
therefore that
This
is
<f>
satisfies
...,
g e)
is
zero,
and
the equation
a partial differential equation for
<&
there are 6 independent
and 5 independent solutions can at once be found, namely the
variables,
quantities
f^
V Pi
-fc&V ., (Ml
\ Pi
(**/
_^|
PK
It follows that
3>
is
a function
only of the quantities
,,P,
Now
are
the factors of

differ
used in their formation
....p..
from each other only in that different roots s

so when such a relation exists between
q 6 ) that two of these roots s become equal to each other, then

= be regarded
will become equal to each other
two factors of
hence if
as an equation in jo 1} at least two roots will become
equal to each other.
(<?!,
q2
...,
When
this relation
f(<h,
exists
<l>
between
,
...,
q-2,
...,
9)
(q 1 ,q 2 ... q6 ), we shall therefore

6 will each be zero.
3<f>/9p
have d^/dp^
= 0; and similarly
368
Since
<E>
is
Pi
equivalent to
of the equations
is
<t>
in
homogeneous

94>/8pj
H Pi o
->
so
(p 1} p 2
<&
= 0,
I-
+p
xiv
p 6 ), the equation
.,
[CH.
does not constitute an equation independent

,
d<&/dp 6
0.
If small variations are given to the variables which satisfy the equation
= 0, their increments are connected by the equation
2
r=1
but
if (q lt
?..,
?6 Pi,
,
Pe) satisfy the equations
d$>/dp r
0, this
equation
becomes
6
and
this relation
between the increments 8^ r must therefore be equivalent
to
the relation
df
2 f-Bq r =0.
r
r=i
oq
Hence the equations
are consequences of the equations
we have
(for sets of values
d<$/dp r
of q lt q2
...,
q,
and
lt
so,
...,
since
p 6 which
7) r
3<J)
g~
is zero,
satisfy these
equations)
The equations f=
r
and 2 ^
~=
are therefore algebraically derivable
r=1
from the equations 94>/3p r = 0. Now the actual values of (g^. ...,g6 ) are
of no importance in this algebraical elimination; so we can replace q r by
we see that the equations
pr) in a11 tne equations: and thus
l+
t.
...,,+,. 8,
r=1 /^ dty
are algebraical consequences of the equations
ft)
7\
3>
(q lt ...,q 8 ,p 1 , ...,p 6
)-
^T
164]
Hence the
result of eliminating
369
between the equations
AH
must be an algebraic combination of the equations
(r=l,
Now
one such algebraic combination of these equations

<(?!,
...,q 6 ,p!, ...,#,)
2, ...,6).
is
= ();
can be derived by multiplying the equations by (p l ..., ps ) in turn, and

it is the eliminant which has just been
adding them. We shall shew that
for it
mentioned.
For
let
the eliminant in question be denoted by

6
then the equation
/axir
U i .
5-
r =i
must be a combination
8g r
-
V3gr
of the equations
r=l o
and
1#a
+ i
.-
Since the latter equation involves
combination
The
for
is
<J>
and
so
8,,
Si,
we
8^ +
see that
it
S(
= 0.
cannot enter into the
we have
found
identity of these equations with those which have already been
are equivalent. Hence
and
shews that the equations
^=
<!>=
<t>=
the eliminant of the equations
and
3*
Now
which are the conditions that the

equation for s may have equal roots, can easily be written down and this
and hence, by
result enables us then to find all possible polynomials
the equations
f(q lt
q2
ê)
0,
<1>,
factorisation of
w. D.
<J>,
to find all possible polynomials
</>.
24
370
The eight
[OH. xiv
+r
roots s are the eight values of the expression
r^r
where r 1} r2 r3 denote the mutual distances: so we may have two roots

equal as a result of any one of the equations
,
?i
r2
0,
= 0,
r3
0,
The equation
rx
rz
r3
r3
r 1}
i\
r.2
r3
r.2
1\
= 0.
gives
<tf
+ q^ + q:? =
0;
and the eliminant of
+p
=
so the value of

arising in this connexion
yu-3
2
3
t\
=0
&
8-fi
fc&Y
is
7x2
this expression is not resoluble into real factors,

can arise from this source.
nomials
and therefore no
real poly
</>
similar result can be deduced in connexion with the equations r 2
and r 3
0.
Consider next the equation
r,
r3
it
-I
1,
m7T
9* M* M.
in
f/tj
2 ( 9l gr 4
or
"T~
fit s
+ ?2 ^ + ?
5
?6 )
\* ^1
^
//tj ~|
Replacing
to
5-,.
by
(q r
+ prt/f^r),
2 (q l9t
+ q*q + q
5
q 6)
+^+
q*)
= 0.
and forming the discriminant with respect
of the equation thus obtained,
^T^2 (?!
^^
we
(ft"
find
+? +
^1^4-- --^2^5-- --P

,
_,
m +m~(
l
o~r
^2
This expansion cannot be factorised into polynomials

can arise from this source.
--ipe), so no functions
</>
{p\,P-2,
^>
If
o"i~
\/j, l
pf/)
linear
in
The Theorems of Brims and Poincare

Similarly
may be shewn
it
with the equations rs
rr r
,
that no polynomials
=+
rz
</>
371
can arise in connexion
Lastly, the rationalised form of the equations
r2
1\
2
is
(r3
When
i\ is zero,
- r? +
r3
r*}*
=
4 rfa*
this case reduces to that
= 0.
which was
last discussed
since the polynomial

is not resoluble in this
special case,
resoluble in the general case.
<l>
Thus
finally,
no real polynomials
involving
s,
can
it
and
cannot be
exist.
Summarising the results obtained hitherto, we have shewn that any

algebraic integral of the differential equations, which does not involve t, is
an algebraic function of integrals
each of which can be written in the
</>,
form
<f>0
+
</>2
</>4
where
is a
homogeneous polynomial in the variables
and a homogeneous algebraic function of the variables
<
p, say of
q,
degree
say of degree
k,
I
homogeneous polynomial in the variables p, of degree (k - 2), and

a homogeneous algebraic function of the variables
q, of degree (1);
4 is
a homogeneous polynomial in the variables
- 4), and a
p, of degree (k
homogeneous algebraic function of the variables q, of degree (I 2) and
</>
is
<
so on.
Proof that
of angular momentum.
(viii)
We
is
</>
shall
a function only of
the
momenta and
the integrals
now proceed to shew that an integral

characterised by these
an algebraic function of the classical
integrals.
<f>,
properties,
is
The equation
or
g-o,
at
gives on replacing
<f>
by
<
4-
<
+$ +
_ ^
r
and equating terms of equal degree,
r
,
h 5
=\oq r Hr
**
~~
dp,
dq,.
~r
r=l
_
dqr
242
372
The
first
of these equations
is
[CH. xiv
a linear partial differential equation for
<
which can at once be solved, and gives
=/ (P P
2

<
=/
3/2
we have
/-
8<f>
pi, Pz,
p),
---
PrQl
/Ltj
fj, r
9
^
A o,
<V
(i
ft,
...
01.
..., fi\
PI,
P*
PS, ...,P s ,pi, ...,p),
r
+ 42 aF dq
d
>
._
8Pr
denotes the expression of
U
,
pi, ...,
we have
-
The terms
now seen
so the
in
^-rj
(r
>
1)
and
which may give
rise to
5^5
logarithmic terms in
to be
terms which
may be
logarithmic in
Vdqt
* +
Xdq^
are
r=2 = 2 9
Xdq^
are
The Theorems of Brims and Poincare
164]
Now V
sum
is
(A + Bq + Cq^y-
of three terms, each of the form
we have
Taking each of these terms separately,

of the last expression
Pr
9/0
for the
(7
p r ps
85
.....
Q yvcn n
V>Olll
2<7o,
<ML
Thus
for
dB
8/0
tZ- - p,
>
each of the fractions (A
+ Bq + Cq^^, we must
for
the
of these fractions,
first
have
r=2
Now
2Cq,
arcsm
fl,rr.sin
./A,
namely (q^ + q.* +
q/)~^,
we have
so the first of the three equations will be
r=2
8P2
or (since
/i x
yu. 2
and on solving
8P
yu, 3 )
this equation
Pi, P*,
-,
we
Pe, P*,
see that
P (p^
s
a function of
is
p^qt),
Since the three expressions (A + Bq1

three quantities (q, 2 + q2 2 + qs 2 ), (q.q. +
and
(p^qs
psqt).
Gq^) are linear functions of the
q^ + q.q,),
*
(qt + qf + q,*), we can for
our present purpose replace them by these three
quantities so the second
expression (A + Bq^ + Gq^) may be taken to be (q^4 + q 2 q 5 + q3 q K }, or
:
(/iP
PI
/J,p t
PP1
?i
P^q\\ ff^Ps
fftP2
I
\ pi
so for this expression
Pi
I \
Pi
/^PsQi\
,
fjfp l
we have
2
Pi
VPS
Pi*
if^Ps
I
PaQi\ /AtPfih^
1
Pi
transcendental part
arcsm
V-
373
pi
374
and the corresponding equation
is
P.
The
third expression (.4
0*i
/*
0-* 5
[OH. xiv
+ .B^ + Cijj*) may
...(B).
/*
2
be taken to be
<?
-f
qs
+ q*~
the corresponding equation proves to be the same as equation (A), and may
We have therefore only to consider equations
consequently be neglected.
(A) and (B)
simplifying (B) by
means
may be
of (A), they
- P,
+ P-2
Pi
written
+P
these equations are obviously algebraically independent and the Jacobian

conditions of existence are satisfied identically for them, since the coefficients
;
r)f
of the derivates
^ do
These two equations
not involve the quantities P.
therefore form a complete system, with 5 independent variables

2 = 3 independent solutions, and
so there must be 5
5
6
P P
,
P P P
solution will be a function of these three solutions
and of
p p
lt
any other
2
...,p
.
ti
It is easily verified that three independent solutions are

2
+P
p6
-Pp
6
+P p -P p
6
or
q3 p 2
M = q^ - q,p
where
+qp -qp
+ q p, - q,p
s
and the three equations
L=
Constant,
M = Constant, N = Constant
We
have
are the three integrals of angular momentum of the system.
therefore the result that
L,
M,
N,
2
K
ly
function of
p p ...,p only.
<
(ix)
is
<
<,
lt
a function of T, L, M, N.
when expressed in terms of

p p2 ..., ps it is clear that $
We shall write
M, N, pi, ..., p6
Since
nomial in
L,
that
Proof
..., q 6 p 1} ...,p 6 is a poly

a polynomial in its arguments
q l q2
,
is
164]
so
we have
<^.
and the equation for/2
where
p l}
F,.
,
= | |G
r =i
r
<fe
= | 8G
atf
9p
37 5
W.
r= i 9p r 9gv
is
stands for dU/dqr supposed expressed in terms of q lt

We have therefore
,
....
P,
Pi r=l &
r
=
â/h
9F
V+
,-tapr
Pi
where the symbol

expression (A
indicates
summation over the three values of the
+ Bq + Cqf).
l
Now the term %(P ..., P p ...,p ) cannot give rise to terms involving
+
(^4
Bq + Cqi ) in the denominator so the quantities multiplying each of
the expressions (A + Bq^ + Cq^ must themselves have the same character
2)
1}
as
<
(<?!>
they must be polynomial in (p lt
i-e.
$6,
<?2>
dG
]
Pi 9/i
g
W
PI,
We
,pz)-
/dG
must be a polynomial
q lt ..., J6 ).
Taking
becomes
in
dG\
(p^
first
2
or (omitting a factor
dB
dP r
/a)
6)
when expressed
in terms of
...,
A+
2gl
95
9P r
dA
9^1
"*
^gp;
4 Cgl
9P;
B ~ÂC
2
HrpidpJ
r^2\9p r
...,
see therefore that the expression
p \ when expressed
fi
%+
Cqf = q* +
q.2
in terms of
(j9 u
+q
expression
2
3
this
...,p 6
376
[CH. xiv
- dG
or
Pi oft
^^
last fraction must therefore represent a polynomial in
denominator must be a factor of the numerator.
The
so the
>.-
IS
ow n
G
(dG
U--
PiP? dG\
S*-=
an d
f^G
---
Vdp2
topiop!/
\dp s
r ua polynomiali in L,
M, N, so
is
p p
...,
are polynomials in Z, -/If, J\r and involve q lt q2 q3 only by means of L, M,

in which case the denominator
so either they contain no terms in q lt q z q s
of
a
factor
the
numerator
or
be
else
cannot
they contain some terms free
,
from q lt qz q 3
,
which case also the denominator cannot be a
in
The condition can
numerator.
8^ _ Pi Pa
p2 pi
dp 2
factor of the
therefore only be satisfied by supposing that
8^ _ Pip 3 3G =
/x 3 ^ dp
dp 3
dpj.
As might be expected from
considerations of symmetry, the conditions
arising from the other sets of values of A, B,
dG
give
dG
fi.pr
fc-l&fc~
4,5,6).
The function G
a complete
therefore satisfies these five equations, which are

evidently
system of five independent equations with six
independent
and consequently possess only one independent solution;

solution is easily found to be
variables,
=i ^,
2/,
The function G therefore involves (p

and since G is polynomial in (p lt ...,
l ,
this
or
6 ),
T.
only by means of the expression
e)
it
must
also
be polynomial in
T.
the expressions (L, M,

involve (q lt ..., q6 ) and
..., q 6 ), and also in (p l} p 2

...,p 9 ), and
are
each
linear
in (q l ..., q 6 ), while
does not
N)
is of degree 2 in
(p lt ..., p 6 ), it is clear that if T is
involved in
must be
Since
is
homogeneous
at
<f>
all, it
in (q lt q2
as a factor
= h(L M,
t
<f>
where h
(x)
is
a homogeneous polynomial in
so
N)T
its
we can
m
write
arguments.
Deduction of Brims theorem, for integrals which do not involve
The equation which determines the function fz
is
t.
164]
But we have
KWKWW
p dT
dp,
PJ.
dp,
and therefore
/ = %(P
2
...,
N)
,p lt ...,p.)-mh(L, M,
T->
U.
Thus
The
integral
namely
can therefore be compounded from two other

integrals,
<
the integral h(L, M, N}(T- U) m which

the classical integrals,
and 2 the integral $
= x (P
<
& = fc - m(
0; =
But
compounded from
where
-f
<
</
</>
and
is itself
^+
P p
6
1}
lf
+
.
. . .
p 6 ),
h (L, M,
^"
?^2)
N) T~* U*,
h
(L>
N) Tm
M>
-3
U3>
an integral of the same character as

except that its highest
of order two degrees less in (p lt ...,p6 ) than the
highest term,
is
((>
<f>,
term,
,
<f>
Now we
of
<j).
<f>
is
have shewn that
<
integrals together with the integral
<
can be compounded from the classical

can be compounded
Similarly
.
<
from the classical integrals together with an

integral
character as
but is of order less by four units than
<"
<,
Proceeding in this way,
we
see that
(f>
integrals together with an integral

(n
is of order
unity or zero. If
(n)
<
<l>M=<l>
we must evidently have & = 0;
in the variables p.
can be compounded of the classical

whose order in (p l} ...,_p6 ) is either
unity in (p lt
>
<
which has the same
<
...,
p 6 ) then
w=h(L,M, N)T
in the equation
n is
in this case, therefore,
compounded of
n
the classical integrals. If
is of order zero in
is a function
it
(p lt ..., p6 ),
of (q l
but
we
have
G
shewn
that
such integrals do not exist
q ) only
already
and so in any case
can be compounded
from the classical
>
<<
>
</><
. . .
algebraically
Hence we have Bruns theorem, that every algebraic
integrals.
integral of
the differential equations of the
problem of three bodies, which does not involve
</>
the time, can be

integrals.
compounded by purely algebraic processes from
the classical
378
Extension of Bruns result
(xi)
[CH. xiv
which involve the time.
to integrals
We now
of
proceed to consider those algebraic integrals of the problem
three bodies which involve the time explicitly.
purpose we shall take the equations of motion as a system ( 155)

of the 18th order: we have therefore to investigate integrals of the form
For
where
this
/ is
an algebraic function of
arguments, and a
its
is
a constant.
Let the last

The function
is not necessarily rational in its arguments.
it may be
that
equation be rationalised, so far as the variable t is concerned, so
arranged in the form

am + a m
(q l ..., q
^(f>
+ a m ~ ^(q^
+ m (qi,
2
p lt
...,p 9
t)
>
<t>
where the functions
are rational functions of
<
q^p,, ...,p 9
.,.,
t)
q,pi,
>#,
...
o,
<)
and algebraic functions of
may be supposed
p
p
cannot be factorised into other equations which
are of lower degree in a and are rational in t for if it is reducible, we can
factors which corresponds to
suppose it replaced by that one of its irreducible
their other
arguments
irreducible in
t, i.e.
(q
lt
such that
l}
This equation
9 ).
it
the original equation
f= a.
Differentiating with respect to

d
a^
Now
we have
t,
p + a^ ^ +
...
at
dt
0.
at
in terms of (g lf q a ..., 9
the quantities
r /dt, when expressed
the
so
that
of
t
functions
rational
are
previous equation would
p lt ...,p ,t),
be reducible in t if this equation did not vanish identically. It follows that
<?
this equation does vanish identically
The expressions
integral
d<f>
<f)
that
is,
are therefore themselves integrals: and hence the

which are rational
other integrals
can be compounded
from
<j),
and algebraic functions of (q lt ...,

be resolved into factors
Let such an integral
functions of
q^ypi,
,pa)-
linear in
<
so that
it
may
be written
where (P,^!,^,...,^,*^!,...,^*) are algebraic functions of (q 1}

Since this expression is an integral, we have
1 d-t
[~
+T
Pdt^^Â
-ri ~77
m>i
~
~ d(pi\
"+- +
dt)^
{-
niîf
*
T"
^t-<}>
k {
a<p]f\
37"
dtJ
~4
n^
-.
t-^\
IT,
...,q 9 ,pi,
~
dt)
ni
dt J
,&)
164]
Wh (<>!,
f -f
//), ..., (tyi,
if
,,
acedb y
their ralues
.....
eft
cfa
eft
are re P
must become an identity: but
this equation
H),
if
happen only
i.e.
* f -1
379
(P
*>
this can
at
each of the expressions
P,
t
-</>!,
t- $ 2
...,
t-
fa, t-yjr lt
...,
t-î
Hence am/
an integral.
which involves t can be compounded (1) of algebraic integrals which do not involve
algebraic integral of the
is
and
of the form
(2) of integrals
==
t
(f>
ivhere
is
<f>
Now
Constant,
an algebraic function of (q lt
it is
known
<>>
...,
q9 PI,
,
...,> 9 ).
that
i_^gi
fr!M? = Constant
^1+^4+^7
an integral hence any algebraic integral of the problem, which involves
can be compounded of
is
(1)
algebraic integrals which do not involve
(2)
integrals of the form
t,
_ miqi
where
<
(3)
is
an algebraic function of (q 1}
...,
q 9 p^,
,
...,
9)
and
the classical integral
_
But the
?fti9i
and (2) are algebraic integrals which do

and hence, by the result already obtained, they are
integrals in classes (1)
not involve the time
combinations of the classical integrals.
Thus
finally every algebraic integral
problem of three bodies, whether
from
it
of the differential equations of the
involves the time or not, can be
compounded
the classical integrals.
Bruns theorem has been extended by Painleve*, who has shewn that every integral of
the problem of n bodies which involves the velocities algebraically (whether the coordinates
are involved algebraically or not) is a combination of the classical integrals.
*
Bull. Astr. xv. (1898), p. 81.
380
Poincare
165.
We
shall
[CH. xiv
theorem.
next establish another theorem on the non-existence of a certain
which is in many respects

type of integrals in the problem of three bodies,
in
1889
was
discovered
and
of
Bruns,
by Poincare*.
analogous to that
The equations of motion of the
(i)
problem of three
restricted
bodies.
In the restricted problem of three bodies, the equations of motion of the

planetoid can ( 162) be written in the form
_ dH
dq r
dt
dp
H=H
where
dt
H H
l
...
2p^~
(r=l,
2),
dq r
+ p.H + fj?H +
1
and
_ _ dH
dp r
...,
UP 2
are periodic in q ly q2 with period

,
27r.
The Hessian
in the proof
evidently zero as this circumstance would prove inconvenient
of Poincare s theorem, we shall modify the form of the equations so as to obtain
is
a system for which the corresponding Hessian
Write
= K,
and
is
not zero.
let
H = h be the integral of energy
dq r
1 dK
= oi
o
~rr
dt
2h
1 dK
dp r =
~JT
~~o7^~
2h
dt
>
then we have
a\
v. ***.*)*
,
-\
oq r
dp,-
therefore, taking a new function

equal to K/2h, we can write the
differential equations of the restricted problem of three bodies in the form
and
dqr
where
for sufficiently
= _dH
dpr
= dH^
dt
dt
dp r
small values of
/*,
=l
2}
dqr
H can be expanded as a power-series
in the parameter n,
and
the Hessian of
period
is
not
now
zero,
and
(H
...)
are periodic in q lt q2 with

,
2?r.
Acta Math. xm. (1890),
p.
259; Xouv. Meth. de la Mec. Gel.
i.
(1892), p. 233.
165]
Statement of Poincare
(ii)
Let
381
s theorem.
which
is one-valued and
which
do not exceed
p
q.,,
a certain limit, and for values of p and p 2 which form a domain D, which
is periodic with
may be as small as we please and suppose that
respect to
and
the
2?r.
Under
these
conditions the function
can
q
qz having
period
denote a function of (q
<3>
p 1} p
q^,
: ,
and
regular for all real values of q 1
2,
JJL)
for values of
<J>
<f>
be expanded as a power-series in
/j,,
say
where
^, 2 ... are one-valued analytic functions of (q1} qz ,pi,p^), periodic
in ^ and q 2
Poincare s theorem is that no integral of the restricted problem
three
bodies
exists (except the Jacobian integral of energy and integrals
of


equivalent to
it),
which
of the form
is
4>
= Constant,
The proof which follows is applicable

whose
any dynamical system
equations of motion are of the same type as
those of the restricted problem of three bodies.
wliere
is
<J>
a function of
this character.
to
The necessary and

integral
is
sufficient
condition that
Constant
may be an
expressed by the vanishing of the Poisson-bracket (H,
and therefore
(H

)
so that
0,
0r ,4g+(jr,,t)-a
a
Proof that
(iii)
We
shall first
^>
**
not a function of
shew that
relation of the form
cannot be a function of
4>
O = ty (H
For suppose a
From
the equation
we have on solving for p l an equation of the form p^ = 6 (H
be a one-valued function of its arguments unless dHo/dp^
domain D.
Replacing p by
have an equation of the form
and as
its
to exist.
value 6 in the function
^>
H = H (p
,
is
(q 1}
2 ),
lt
and B
p2 )
will
zero in the
q^,p-i,p^),
we
a one-valued function of
its arguments ty will be a one-valued

but by hypothesis, the function ty depends only
on
It follows that ^r is a one-valued function of
so long as the
variables pi,pz remain in the domain D, and
is
not zero in D
provided ^H^fdp^
or more generally provided one of the derivates
and
dH /dp 2 is not
dH^/dp!
4>
is
function of (q lt q z
H ,p );
2
zero in D, a condition which
is
Since t/r is
evidently satisfied in general.
a one-valued function, the equation -/r (H) = Constant will be a one-valued
- -v/r (H) = Constant
integral of the differential equations, and therefore
will also be a one-valued
integral, and will be expansible as a power-series
<
382
in
fj,
we
it will
moreover be divisible by
//,,
since
ty
4>
(#)
[CH. xiv
If then
is zero.
write
- ^ (H) = p
<X>
the equation
the function
= Constant
<!>
be a one-valued analytic integral
be a function of
will not in general

will
writing
however
if
it is
we perform the same operation
again, thus arriving at a third

whose
one-valued analytic integral,
part independent of /A will not in general
and so on. It is evident that in this way we shall
be a function of
function of
ultimately obtain an integral which does not reduce to a function of

is a function of H, in which case the two integrals
is zero, unless
IJL
<$>
H when
H and

are not distinct.
which is one-valued and analytic

an integral
is a function of
we can
and distinct from H, but which is such that
from
it another integral, of the same character but such that
derive
always
If,
therefore, there exists
<
<l>
it
does not reduce to a function of
always suppose that

(iv)
Proof that
<&
Am
m^
co-factor of
Am
HI j,
cannot involve the variables q 1}
can therefore
<?,.
it is
periodic in
i,i
are positive or negative integers, i denotes v

1, the
are functions of p 1} p z and
represents the exponential
,
Since
does not involve q 1} q 2 we have

,
ill.-)
-Q)
/dq r
d3>
dp 2
dq,
= 2 im r A m m
m,
dp,
But we have
We
= 2 A
,
quantities
vanishes.
4>
where m, and
<
we
when p
not a function of
involves the variables q, q 2 then since

can write
If the function
these variables
is
<E>
so
%,
oq*
the equation
(H
4>
m-i
becomes
*
dPl
and
therefore (as this equation is
ij
-=-
dp,
Hence we must have
an identity)
=
+m dpj
2
0.
either
A mi ,mo =
or
>idHo/dpi
+ m dH
2
(l
fdp 2
but the latter alternative is possible only when m, and m 2 are both zero, or
It follows that all
is zero, which is not the case.
when the Hessian of
does not
and consequently
are zero, except A 0t
the coefficients A m

)OTo
involve the variables q, and q z
165]
Proof
(v)
the result of
is inconsistent ivith
in the general case.
(iii)
Consider
of a one-valued integral
that the existence
383
now
the equation
As
dp r dqr
4>
be expanded in series
and
and
C^,^
m.,
2 Bm mn m r
2
mi ,m,
OT,,m 2
or (since this equation
This equation
p and p2 which
^=i
dq
r
4
2 aâ^
r 5 -=-^
2 B
d
t,
"
m^m.,
^n mrS,
2 m,
(
\r-l
=
^)
Op
0,
r/
an identity)
p p
l}
and therefore
for values of
satisfy the equation
dH.
--^
f-
W1 2
opi
dH =
^r
0,
3jp 2
either
B m,
m2
or
>
9ô/3î
values such that
m>t
9^0/3^2
Bm ^ m becomes
m dH /dp + m dH fdp = 0.
shall say that a coefficient
is
"^
2 C
m,^)
Opr/
<1
we must have
valid for all values of
is
,2
r=l dp r dq r
(V=l
is
2 Cm
Wi>w
Bm
aâ^ =
9^
-
r =i 9jo r
becomes
We
so the equation
say,
are positive or negative integers, and the coefficients

have therefore
depend only on p lt p 2
dqr
As
^=i
We
= 2 B
! ,
dpr dqr
ei(m iqi + m , q ,)
2 B
m,
where
r !Ti
and
are periodic with respect to q 1} q2
they can
of the form
the functions
= 0.
secular
when pj,p2 have
a given function, the coefficients
Bm
>m
are given.
In the general
case of dynamical systems

expressed by differential equations of the kind we
are considering, no one of these coefficients will vanish when it becomes
secular,
and we
shall take this case first:
so that the equation
is
=
x dô/dp, + m^ d
/dp2
a consequence of the equation
l dffo/dp, + m 2 dH
/dp
Now
let
k-i,
k.2
be two integers
such that the equation
= 0.
suppose that we give to p^ and
values
384
that
1 k
+m k
z
m dH /dp + m dH /dp
expression
m-!
so the Jacobian
dH fdp
is zero,
d(H
dH jdp
and
<
)/3
0>
(p 1} p2 )
and consequently
+ m2
BOo/a^j
Comparing these two equations,
is zero.
xiv
can find an infinite number of pairs of integers m lt m., such

and for each of these systems of integers the
is zero:
We
is satisfied.
[CH.
d$>
/dp2
we have
zero for all values of
is
p p2
lt
Thus
are commensurable with each other.
in
for
which
any domain,
however small, there are an infinite number of systems of values of p lt p 2 for

which this Jacobian is zero as the Jacobian is a continuous function, it must
must be a function of
therefore vanish identically, and consequently
in
what
was
But this is contrary to
(iii), and therefore the funda
proved
:
<J>
must be erroneous
mental assumption as to the existence of the integral
that is, the Harniltonian equations possess no one-valued analytic integral
<&
other than
becomes
H=
h,
Removal of
have now
vanishes
indices
(raj
when
and (m/,
2)
becomes
it
m vanishes when
it
on the
coefficients
which at
We
secular.
least
.
mi>
one of the coefficients
two pairs of
shall say that
?n/) belong to the same class when they satisfy the

and that in this case the coefficients Bm ^ m and
belong to the same
<
wij
the restrictions
to consider the case in
mj/m/ = ra /m
relation
secular.
(vi)
We
Bm
provided no one of the coefficients
class.
,m<2
We
shew that the
shall first
result obtained in (v) as to the non-existence
true provided that in each of the classes there is at

coefficient
least one
m m which does not vanish on becoming secular. For
is zero, but the coefficient
that the coefficient
m m is not
of one- valued integrals
is
zero.
B ^
B^^
suppose
If pi,
i
have values such that
W 3H
and although the
/dp2
0,
relation n^
m-i dH ftpi + m dH /dp

2
3<&
= cannot be inferred from

/dp 2
can be inferred from the latter the proof is
/9.Pi
d3>
mj/w 2
= X-
domain,
let
lt ra 2
completely defined by the ratio of the indices
which
for
indices
of
C
the
class
be
commensurable number, and let
a class
X be any
we have
and consequently
the former of these equations, it

in other respects the same as in (v).
Now
is zero,
is
We
shall say for brevity that this class

or is in this domain, if a set of values of p l}
domain such that
dH
--
A, ^r
9//o
1-
_ nV.
G
p
2
belongs to a given
can be found in this
165]
We
shall
which
small,
which not
all
385
shew that the theorem
is still true if in
every domain B, however
contained in D, there are an infinite number of classes for
the coefficients of the class vanish when they become secular.
is
For take any
set of values of
p ,p 2 such that
l
dH
-
A.
3/f
A, is
_ U.
op*
9pi
Suppose that
we have
for these values
commensurable, and that
for the class
which corresponds
to this value of X, all the coefficients of the class do not vanish
when they
become secular the preceding reasoning then applies to this set of values,
and so for these values of p and p2 the Jacobian d(H
)/3(p 1} p. } is zero.
But, by hypothesis, there exists in every domain 8, however small, which is
:
4>
contained in D, an infinite number of such sets of values of p lt p 2

The
Jacobian consequently vanishes at all points of D, and therefore O is a func
tion of
so, as before, there exists no one- valued integral distinct from
.
Deduction of Poincare" s theorem.

In the four preceding sections, we have considered equations of the
type
(vii)
dqr_dH
dt
which
in
dpr ~_
dt
dp r
dH
dq r
H can be expanded in the form

H
H H
where the Hessian of

Q with respect to p^ and p 2 is not zero,
does not
involve q and q2 and
... are
functions
of
and
lt
2
we have
periodic
q lt q2
shewn that no integral of these equations exists which is distinct from the
equation of energy and is one-valued and regular for all real values of q and
q2 for values of /* which do not exceed a certain limit, and for values of p and
p z which form a domain D; provided that in every domain, however small,
l
contained in D, there are an infinite number of ratios

m^m^ for which not
the corresponding coefficients Bm m vanish when
they become secular.
^
all
This result can be applied at once to the restricted

problem of three
bodies for we have seen in (i) that the
of
motion
in this problem
equations
are of the character specified, and on
the
function
H^ by actual
determining
:
expansion we find that the last condition
is satisfied.
Poincare s theorem
is
thus established.
Poincare s theorem establishes the non-existence of integrals
uniform with respect to
Keplerian variables, which implies uniformity in the neighbourhood of all the tra
This however does not exclude the
jectories which have the same osculating ellipse.
the
existence of integrals which are uniform in
domains of a
different character.
Cf. Levi-
Civita, Acta Math. xxx. (1905), p. 305.
cf.
The theorem has been extended by

Nouv. Meth. de la Mec. Cel. i. p. 253
Poincare"
it
to the general problem of three bodies

,
(1900), p. 1699.
w.
D.
has also been extended by Painleve C.R. cxxx.

25
CHAPTER XV
THE GENERAL THEORY OF ORBITS
Introduction.
166.
We
shall
now
and disposition of the

pass to the study of the general form
For simplicity we shall in the present chapter
orbits of dynamical systems.
motion of a particle which is free to move in a plane

under the action of conservative forces, but many of the results obtained can
be readily extended to more general dynamical systems.
chiefly consider the
been observed ( 104) that the determination of the motion

two degrees of freedom under the action of conservative
with
of a particle
to
the problem of finding the geodesies on a surface with
forces is reducible
It has already
a given line-element; an account of the properties of geodesies might

therefore be regarded as falling within the scope of the discussion.
Many of
these properties are however of no importance for our present purpose and
as the theory of geodesies is fully treated in many works on Differential
:
which are of general

Geometry, we shall only consider those theorems
dynamical
The
orbits
interest.
have been obtained hitherto relate to periodic

principal results which
of a given orbit (especially of a periodic
to
the
stability
167-171),
with respect to small displacements from it (

172-176), and to the
i.e. the question
to
the
with
of
orbits
time,
a
of
respect
given group
stability
the lapse of a
after
character
their
orbits
the
how
far
as to
general
orbit)
preserve
very great time

167.
177-179).
Periodic solutions.
Great interest has attached in recent years to the investigation of those

motion of dynamical systems in which the same con
particular modes of
figuration of the system
is
so that the
repeated at regular intervals of time,
motion is purely periodic. Such modes of motion are called periodic solutions.
The term periodic solution is also used in cases where a relative rather than
an absolute configuration
three bodies, a solution
is
is
periodically repeated:
said to be periodic if the
thus in the problem of

mutual distances of the
The General Theory of Orbits
166-168]
387
bodies are
periodic functions of the time, although the bodies may not

have
the same orientation at the end of a period as at its
necessarily
beginning.
Considering specially the motion of a particle in a plane or on a fixed

smooth surface under the action of conservative forces, it is evident that a
family of periodic orbits will exist in the neighbourhood of each position of
stable equilibrium of the particle, namely the orbits
corresponding to the
normal vibrations of the particle about this equilibrium position. If the
position of equilibrium is unstable, it may happen that the periods of both
the modes of normal vibration are
imaginary, in which case no periodic orbits
exist in the
vibration
vicinity,
is real,
these orbits will evidently however be unstable, whereas the

neighbourhood of the position of stable equilibrium are stable.
periodic orbits
orbits in the
168.
or_that the period of one of the modes of normal

real normal vibrations give a family of
which case these
in
Poincare
normal variables for a known periodic
The equations which

pressed in a
define a periodic orbit are
form due to Poincare
orbit.
most conveniently ex
*.
Let the motion of the dynamical system considered be defined

by the
equations
dq,
dt
where the function

fc
= dH
dpr__dH
=
dt
dpr
dqr
H does not involve the time
=
0l(0>
?2=<k(0>
Pl
explicitly
P*
^l(t)>
and
let
= fa(i)
be the equations which define a known

There
periodic orbit of this system.
is clearly no loss of
if
we
generality
suppose the coordinates (q l} q2 p^ p 2 ) to
be such that after the lapse of a
period the variables q lt qz p resume their
,
initial values,
while
p.2
increases
From these equations

so that the functions 6 1}
0.2
by
2?r.
can be eliminated
let
6S have the period
the result of the elimination
2-rr.
Perform on the system the contact-transformation defined

by the equations
where
(
8*
(,)-
Nouvelles Methodes de la Mec. Cel. n. p. 369.
252
388
The equations
Q,
xv
of this transformation can be written
- 9, - *. (ft) +
The equations
[en.
{*
ft (ft))
- *3 (ft)}
fft
new
of motion of the dynamical system, in terms of the
variables, are
dPr
dt
0&
dt
and from the above equations of transformation
now
solution is
it is
evident that the periodic
Q = 0,
1
Q,
= 0,
P=
1
0,"
P =f
2
(0-
This form of the equations of the orbit will be called Poincare s
normal form.
orbits.
for the discovery of periodic
We shall now shew that the existence and position of periodic orbits can
the
be determined by a theorem* analogous to those theorems which furnish
roots of an algebraic equation by considerations depending on
position of the
We shall for simplicity
the sign of expressions connected with the equation.
considered to be that of the motion of a
suppose the dynamical problem
mass in a plane under the action of conservative forces: the
particle of unit
result can be extended to more general systems without difficulty!-
169.
criterion
of the particle at time t, referred to any fixed

y} be the coordinates
and let V(x, y) be its potential energy function,
rectangular axes in the plane,
of
the
so that
energy is
equation
Let
(as,
l(#+f)+V(a>,y)=K,
where h
The
is
the constant of energy.
differential equations of
motion of the particle form a system of the
fourth order, and their general solution consequently involves four arbitrary
One of these constants is, however, merely the constant additive
constants.
which determines the epoch in the orbit, so there are only

This triple infinity of orbits can be arranged in
distinct orbits.
to
t,
oo
really
sets,
each
those orbits for

containing a double infinity of orbits, by associating together
of oo 2 orbits
set
a
h
such
which the constant of energy has the same value
:
Cf. A. Signorini, Rend. d. Lincei,

186.
Whittaker, Monthly Notices R.A.S. LXII. (1902), p.
xxi.
Rend. d. Palermo, xxxin. (1912), p. 187; L. Tonelli, Rend. d. Lincei,
xxi. (1912), p. 36;
(1912), pp. 251, 332.

of three bodies,
f For the extension to the restricted problem
LXII. (1902), p. 346.
cf.
Monthly Notices R.A.S.
168, 169]
may
389
be defined analytically by the principle of least action
that the orbit between two given points (x

the value of the expression
and (xly
y^) is
namely
100),
make
such as to
stationary as compared with other curves joining the given terminal points
Oo,
?/ )
and
(>,,
2/ a )*.
Consider any simple closed curve C in the plane of xy and let another
simple closed curve G be drawn, enclosing C and differing only slightly from
it.
We may regard as defined by an equation of the form
;
C"
where 8p is the normal distance between the curves C and C (measured out
wards from C, and consequently always positive) and 7 is the inclination of
this normal distance to the axis of x.
Then if / be the value of the integral
when the
integration is taken round the curve C, and if / + 81 denote the

value of the same integral when the integration is taken round the curve
C"
(so that the
symbol 8 denotes an increment obtained in passing from
to
),
we have
81=
{(dxf
+ (dy)^
8 [h
V(x, y)}?
{h
V(x, y)}? 8 {(dx?
But we have
- i a rr/
{A- V(x,
i
and
8 {(dxf
where p
is
(dy}
8F
M-*/
I
cos
y}\
= 8p.dy =
^j-
9F
y-
sin
V,,
)
8p,
{(dx?
the radius of curvature of the curve
at the point (x, y).
Thus we have
S/=f{(<fe)+(dy)}*{A-r^^
J
This equation shews that
if
ox
oy }
the quantity
h-V(x,y}
-- A cos 7l dV
-- * sin 7T dV
^
-7=r-
Following Painleve, Liouville
which have the same constant
dx
s Journal, x.
(1894), it
of energy a natural family.
dy
is
customary
to call a family of orbits
390
[OH.
xv
81 is negative, and so the integral I has

negative at all points of C, then
curve
when
its value diminished
surrounding C and adjacent to C is
any
of
taken instead of C as the path
integration.
is
Now
C,
h-V
*
2
cos 7
Then, in the same way,
is positive.
diminished when any simple
taken instead of
is
3F
When,
D
we
therefore,
-,.--
D,
curve
suppose that another simple closed
all points of D the quantity
can be drawn enclosing
and such that at
it
dV\
+ sin 7 ^r
dy/
can be shewn that the integral /
closed curve
enclosed by
and adjacent
is
to
as the path of integration.
consider the aggregate of
all
simple closed curves
which is assumed to
situated in the ring-shaped space bounded by C and
that the curve
is
clear
it
function
of
the
no
contain
V(x, y)
singularity
coincide
cannot
or
be
C
and
cannot
of
I
value
D,
which furnishes the least
There exist, therefore, among the

for which the
one or more curves
simple closed curves of this aggregate,
does
value of / is less than for all other curves of the aggregate. Since
with
or
for
any part of
its
length.
it follows, that the

C or
along any part of its length,
are all members of the aggregate in question, and hence
curves adjacent to
furnishes a stationary value of / as compared with all the
that the curve
is therefore an orbit in the dynamical
The curve
to
it.
curves adjacent
theorem: // one dosed curve be
at
the
arrived
We have thus
system.
the
enclosed by another closed curve, and if
quantity
not coincide with
h-V(x,y)
v
?-
i cos
dV
7* -~dx
8F
* sin 7 -rdy
be negative at all points of the inner curve and positive at all points of the outer
the two curves there exists a periodic
curve, then in the ring-shaped space between
orbit of the
dynamical system, for which the constant of energy
quantity
h-V(x,y)
^
- i cos
.
"
dV
7 ôx
\ sin 7
is h.
As
the
dV
-5-
dy
can be calculated immediately for every point on the curves C and D, de

function and the curves
pending as it does only on the potential-energy
themselves, this result furnishes a means of detecting the presence of periodic
orbits.
170.
We
Lagranges
shall
now
three particles.
consider specially certain periodic solutions of the problem
of three bodies.
useful
summary
of theorems relating to families of periodic orbits in the restricted

K. Moulton, Proc. Inter. Cong, of Math. Cambridge,
is given by

1912,
II.
p. 182.
169, 170]
The
Proc. L.
and
relation of periodic orbits to orbits of ejection
same
the bodies occupy in the same position at the
M.
391
which two of
collision, in
studied by Moulton,
is
time,
S. (2), xi. (1912), p. 367.
category of non-plane periodic orbits in the problem of three bodies

Pavanini, Annali di Mat. (3), xin. (1906), p. 179.
is
discussed by
Let the equations of motion of the problem be taken in the reduced form
obtained in 160, and let us first enquire whether these equations admit of a
particular solution in which the mutual distances of the bodies are invariable
throughout the motion.
The mutual distances
are
--2m2 ftft/ cos q^ cos q, ---k -p --pf

sin q
m +m \
2p p,
2
11
<7i
-\Q
(*~
sin o 4
*/
k -p --pf
2m o go/
-sm ft sin aft\ +
cos
cos ft ---gW*
* + m
+ m,\ ft
)
2,
2
and
it
follows that, in the particular solution considered, the quantities
k -p-sin sin
--ft
^
^
2
and cos
ft,
ft,
ft
p<?
cos ft
must be constant, and hence the functions U, dU/dq^ dU/dq 2 must be constant,
~
where U= ^m m 2 r12
l
The equations
= ft = dH
=p
5
.
9î
shew that p and

l
p.2
p.
= 03 = dH
^=^,
2
1
,
3^2
/*
must be permanently
zero
A*
while the equations
shew that p3 and p 4 must be constant.

Moreover, the equations
A=
.
/>
a#
aff
= -^-,
0=^4 = -^
3ft
3ft
shew that the expressions

3
a
C S ?4
9ft
fj
and
are zero, so
-- __
W*2
cos ft cos o4
^
3ft \
yj^
2 ^.^
2
^J
^4
sin o, sin
2jo 3 p 4
we have
-&
= cot ft tan ft = ^Pi + ^
2
tan ^ 3 cot
ft
*ppt
and therefore
or
2
p* + ^ 4
-^ =+
^ = (/
2p 3 p 4
2
,3
p4 )
2
,
392
[en.
xv
an equation which shews that the instantaneous planes of motion of the

bodies
and // coincide with the plane through these bodies and the origin
fj,
in other words, the motion of

the
motion of
m m m
1
and
//,
//
takes place in a plane
and therefore
takes place in a plane.
It follows that, the centre of gravity

of the system being supposed at
ra
we
shall
denote by P, Q, R) must move
the
2
3
rest,
(which
particles
m^ m
in circular orbits
We
round 0.
have now to see
if
such a motion
is
possible.
One
condition which must obviously be satisfied is that the resultant

attraction of any two of the particles on the third must act in the line joining
the third particle to the centre of gravity. This condition is satisfied if the
If they are not in a
three particles are in the same straight line.
straight
line, it
gives
*ffl
sin
But
since
is
PRO = ^~-
sin
ra 2 sin
we
we have
PRO _~ sin QPR _

~ QR
QRO sin PQR P
PR = QR
combined with the preceding equation gives
this
find
similar equations.
the centre of gravity of the particles,
m
and
QRO, and two
sin
similarly
PR = PQ.
Hence
either the bodies
must
be collinear, or else the triangle
formed by
them must be equilateral.

Considering
the collinear case, let the distances of the bodies from

(measured positively in the same direction) be a l} a.,,
first
their centre of gravity
as respectively: we
which does not lessen

2
s
suppose that a l
Since the force acting on P must be that
corresponding to circular motion round 0, we have
shall
<a
<a
the generality of the discussion.
n?a
where n
n
a2
is
(a s
a 2 )~
+ k)
where & denotes the

is
(a2
+ n^ (a
these equations
rn^fc {(1
This
we
PQR
na =m
aj)~
and similarly
2
a^"
ni 3 (a s
i)~
the angular velocity of the line
From
(a s
o^)"
+m
(a 3
a 2 )~2
readily find
- 1] + ma (1 + k) (& 2
a2 )/(a 2
ratio (as
a quintic equation in
k,
1)
+ m,
[k
- (1 + k) = 0,
3
a-^.
with real
coefficients.
Since the left-hand
side of the equation is negative when k is zero, and positive when k = + oo

there is at least one positive real root; such a root determines uniquely real
values for the ratios a 1 :a 2 :a3
be completely determined.
and
if
is
given, the distances a lt a z a3 can
It follows that there are
an
infinite
number of
solutions of the problem of three bodies, in which the bodies remain always in a
straight line at constant distances from each other ; the straight line rotates
393
170]
uniformly, and when
its
angular velocity has been (arbitrarily) assigned, the
the bodies are determinate.
mutual distances of
of
Considering next the equilateral case, let a be the length of one side
Since
the triangle formed by the bodies, and let n be its angular velocity.
to a circular orbit round 0,
the force acting on
3 is that which corresponds
we have
a?
cos
PRO + ~
a
cos
QRO = n*
OR,
a condition which reduces to
relating to the motion of Q and of R reduce to the same

and
hence
a motion of the kind indicated is possible, provided n
equation:
and a are connected by this relation. Hence there are an infinite number of
The conditions
solutions of the problem of three bodies, in which the triangle formed by the
bodies remains equilateral and of constant size, and rotates uniformly in the
plane of the motion: the angular velocity of its rotation can be arbitrarily
assigned,
and
the size
of the triangle
is
then determinate.
particular types of motion which have now been found will be

Lagrange s collinear particles and Lagrange s equidistant particles
The two
called
respectively*.
after Lagrange s discovery, its interest was sup
to
be
But in 1906 a new minor planet, 588 Achilles,
theoretical.
posed
purely
was found to have a mean distance equal to that of Jupiter and it was soon
For more than a century
realised that the Sun, Jupiter, and Achilles constitute, approximately at any
rate, an example of the Lagrangian equilateral- triangular configuration.
Shortly afterwards came the discovery of three other Asteroids, 617 Patroclus,
624 Hector, and 659 Nestor, which are in the same caset- Of this "Trojan
Patroclus is in longitude Jupiter
60, and the other three in longi
group,"
tude Jupiter + 60.
Example. Shew that particular solutions of the problem of three bodies exist, in which
the bodies are always collinear or always equidistant, although the mutual distances are
not constant but are periodic functions of the time.
These are evidently periodic solutions of the problem,

as a limiting case.
*
They were discovered by Lagrange
in
1772
and include Lagrange s
Oeuvres de Lagrange,
vi.
p.
particles
229.
For
references to extensions of these results to the problem of n bodies, cf. my article iu the
Encyklopadie d. math. Wiss. vi. 2, 12, p. 529; to the papers there mentioned may be added
E. 0. Lovett, Annali di Mat. (3), xi. (1904), p. 1 W. R. Longley, Bull. Amer. Math. Soc.
(1907), p. 324, and F. R. Moulton, Annals of Math. xn. (1910), p. 1.
;
t Cf. F. J. Linders, Arkiv for Mat.
iv.
(1908), No. 20.
xm.
394
171.
Stability of
Lagrange s
particles
[CH.
xv
periodic orbits in the vicinity.
It has been observed (

167) that in the neighbourhood of any configuration of stable
equilibrium or steady motion there exists in general a family of periodic solutions, namely
the normal vibrations about the position of equilibrium or steady motion. We shall now
apply this idea to the case of the Lagrange s-particle solution of the restricted problem of
three bodies, and thereby obtain certain families of periodic orbits of the planetoid.
be the bodies of finite mass, m 1 and m 2 their masses,

their centre of
n the angular velocity of SJ, x and y the coordinates of the planetoid P when
taken as origin and OJ as axis of x.
The equations of motion of the planetoid
S and J
Let
gravity,
is
are
162)
d^_ dJK
dy_dK
du_ _dK
dv__
~di~du
~di~~dv
~di~~dx
dt~
K=\(u z + v*) + n (uy
where
Let
(a,
denote the values of
b)
(x,
sidered, so that for the collinea,r case
a = ^(m l
m2
The values
na
l/(m l +
2 ),
=\
*J%1,
vx)
_%&
8y
-mi/SP m 2 /JP.
y) in the position of relative equilibrium con

6 = 0, and for the equidistant case we have
we have
where
denotes the distance SJ, so that
46)
of u, v in the position of relative equilibrium are easily seen to be
nb and
respectively.
# = a+,
Write
where
77,
6,
,
neglecting a constant term,
we have
On expanding and retaining only terms of the second order in the small quantities,
obtain an expression for
with which the equations for the vibrations about relative
equilibrium can be formed we shall for definiteness consider vibrations about the equi
we
distant configuration
The equations
in this case the expression for
K becomes
of motion are
dK
dK
~W
~W
dK
dK
*--frj-
~3|~
Solving these equations in the manner described in Chapter VII,

is 27T/X, where X is a root of the equation
we
find that the period
of a normal vibration
X 4 - n*\* + (*i - #) n* = 0, where k =
The two values
~ ^2
(f i
with the other.

of
is
mi + m 2
by this equation will be positive provided they are real, since

2
and they will be real provided 4 (f J & 2 )
27 mj ra 2
1, or (wij + ?n 2 )
satisfied provided one of the masses S, J is sufficiently large compared
:
<
When
first
>
two families of periodic orbits

equidistant configuration of relative equilibrium : the
2
approximation, STT/XJ and 27r/X 2 where X^ and X 2 are the roots of the
this condition is satisfied, there exist
the planetoid in the vicinity
periods are, to a
2
equation in X ,
of X 2 given
s positive
a relation which
of
its
171, 172]
395
similar discussion leads to the result that the collinearLagrange s-particle configurations
are unstable; but the equation for the periods of normal modes of vibration has always one
real root, and consequently in the neighbourhood of a position of relative equilibrium of the
planetoid on the line SJ there exists a family of unstable periodic orbits*.
Example. Shew that, for one of the modes of normal vibration of the planetoid in the
in
vicinity of the equidistant configuration, the constant of relative energy is greater than
the configuration of relative equilibrium, while for the other mode the constant is less than
in the configuration of relative equilibrium.
The differential equation of the normal displacement from an
172.
We
(Charlier.)
shall
now proceed
orbit.
to consider the stability of orbits in general.
Suppose that some particular solution of the motion of a particle of unit

mass in a plane, under the action of forces derived from a given potential
energy function V, is known and consider a solution which is immediately
adjacent to this known solution, and for which the constant of energy has
;
the same value.
Let
P and
Q be
orbits respectively at time
and
arc
let
PN = NQ = u;
,
OP =
a, arc
the orbit at P.
Draw
t.
let
=
ON = so s
We shall regard the
a-
s,
orbit as specified
The
known and adjacent

QN perpendicular to the known orbit,
be a fixed origin on the known orbit let
the positions of the particle in the
and
p be the radius of curvature of

position of any point on the adjacent
by the quantities
let
(u, s).
when
T=i^+i(l +
and
its
describing the adjacent orbit
W /p)2 * 2
is
Lagrangian equations of motion are therefore

,
ou
p/ p
2
u\+- s+..
/,
p)
these equations possess a
u\
)us
p/
known
/.,
integral,
M \2
1 H
s2
z
u\us
dp = -f
as
p/ p-
3F
-JT-
os
namely the integral of energy
+ V = h,
where h
is
a constant.
The
first
Lagrangian equation and the integral become
\\dujp
dp
+
\dudajp
\du
For further work on the subject of orbits in the neighbourhood of the Lagrange s-particle
cf. the memoirs referred to on page 530 of my article in the Encyjtlopadle, and also
Lovett, Astr. Nach. CLIX. (1902), p. 281 Stromgren, Astr. Nach. CLXVIII. (1905), p. 105; Moulton,
Math. Ann. LXXIII. (1912), p. 441.
solutions,
396
[CH.
xv
Since
du
and
the two preceding equations become
Eliminating
fir),
(<T
we obtain the equation

2
f/9
or (taking s instead of
and
F\
So-
2
)
as the independent variable,
du
ds2
v ds ds
dv du
fl id
v2
this is the differential equation
and writing
v for
cr)
du2
of the adjacent orbit
From this equation we can at once deduce consequences relating to

For by Sturm s theorem*, if we have any
the stability of the known orbit.
differential equation of the form
d2u
where
range of values of t the quantity / (t) lies between two

a- and b-, then any solution u which is zero for
within the range will be zero again for some value t within the
for a certain
positive real quantities
a value
range, where
t )
(t
lies
between
ir/a
and
irjb,
provided the range
is
It follows that if the quantity

comprehend
2
2
2
2
known orbit, this orbit will
of
the
all
is
at
3v
points
/p
positive
(3 F/9w )p
it once will not diverge
which
intersects
orbit
be stable^, i.e. any adjacent
from it, but will intersect it again infinitely often. This expression
sufficiently large to
this interval.
greatly
can therefore be called the coefficient of stability for the orbit.
Cf. Darboux, Th. gen. des Surfaces, Vol. in.

172-176, all
f In the discussion of stability in
powers of the displacement above the
first
are
neglected in forming the differential equations of the adjacent orbits. The effect of the neglected
terms on the stability has been studied by Levi-Civita, Annali di Mat. v. (1901), p. 221, who has
found that the neglected terms give rise to instability in certain cases which appear to be stable
when only
first-order
number, where a
is
terms are considered
the characteristic exponent
this
(
happens when aT/2iri is a commensurable

and T is the period of the solution. Cf.
175)
also A. E. Cigala, Annali di Mat. xi. (1904), p. 67.
172, 173]
173.
Korteweg
Suppose now
397
theorem.
known
orbit, with respect to which the normal

a
measured, is
displacement u
periodic orbit whose perimeter is S then if
= (s + nS),
u=
(s) is the equation of an adjacent orbit, it is evident that u
that the
is
<f>
<f>
where n
any integer, is also the equation of an adjacent orbit the orbits

represented by these two equations are in fact congruent, but the corresponding
tracing points are separated by one or more periods.
In any orbit adjacent to the known orbit let u n u n+1 u n+2 (where n is an
integer) denote respectively the normal displacements in the nth, (n+ l)th,
and (n + 2)th period, at the same place in the orbit, so that we can write
is
un =
where
<f>(s
+ (n-
u=
+ nS),
(s
(s
(n
1) S),
a solution of the equation
(s) is
<f>
u n+l
1) S),
d 2 u,
ds 2
v ds ds
dv du
(1 /9-F\
2
\v*\ du /,
Since u n u n+l u n+2 are three solutions of this linear differential equation,
they must satisfy a relation of the form
,
where k and
We
are independent of
k^
shall first
s.
shew that these constants k and & are independent of the

number n, so that they will be the same
x
choice of the adjacent orbit and of the

for any other set
um=
ty (s
For um
+ (m
u m+l
1) S),
ty (s
+ mS),
u m+2
= ty (s + (m + 1) $).
a linear function of the two solutions un and u n+1

say
is
and therefore on adding periods
to the
argument
s,
we have
But from the equations

==
Un+2
"
Un+i
"
Un
cûn ^ + c u n+s = k (c
we have
^n+3
u n+l
c 2 w n+2 )
fo
Un+2
"
"a
k t (du n
^n+i
>
+ c u n+l \
2
u m+2 = ku m+1 + kû m)
and therefore
which shews that the constants occurring in the linear relation between
u m+z, u m+l
Next, we
same
are the
between u n+2 u n+l
un
those occurring in the linear relation
as
From
shall find the value of the constant k^.
d2 un
ds-
ds
F\
31
l*\8tt*/
p*)
dv du n
vds ds
1
[1 /S
dv du
~ -- n+l
n -L
/t-f-l
ds
ds
1 /9
If
~
"
~J~~
~^\
F\
n
the equations
398
//S?/
un
we have
tt
tv 7)4-1
/-/î
"1
/"/".I/
I//
-til
t-t t/
/Y J/
W/
xv
/7 II
IX/ I* *]
tlfr
[CH.
and hence, on integrating,

au n
un
as
Changing
s to s
+ S, we
au n+ i
j-
as
where
c is
a constant.
have
ds
ds
ds
^ ^71-t-l
fi i/
/ 7
cu n+l
+ k lUn )-
u n+1
di
rit i
wW/Tj.-^!
riii
IX/M/TJ,
^
,
^/c-^- .^-jg-J
dtO
a Wn+j
-
~dT
ds
We
1.
thus have the theorem* that if u n u n+l un+
denote the normal displacements in an orbit adjacent to a known periodic orbit
in three consecutive revolutions, the ratio k = (u n+2 + u n )/u n+1 has a constant
so that &i has the value
value, which is the
174.
same for
all
adjacent orbits.
The index of stability.
The constant
ratio
k= (u n +z+ u n )/un+1
where un u n+1 u n+2 are the normal

,
displacements from a periodic orbit in three consecutive revolutions, is called

the index of stability of the periodic orbit, for reasons which will now appear.
The nature
of the integral of the difference-equation

^n+2
depends, as
is
rCUn+i
~^~
Un
"
well known, on the reality or non-reality of the roots of the
quadratic equation
\*-k\ + I= 0,
2 or k
2.
depends on whether k
= 2 cosh a
Supposing first that k is positive and greater than 2, write k
a
a
then the roots of the quadratic are e and e~ and we know that two inde
pendent solutions of the difference-equation are of the form
i.e. it
>
<
as
u = e^
<f>
(s)
and
_as
"
tt (s),
are functions of s which have the period S

these functions so as to make the solutions u satisfy the equation
where
<
(s)
and
-v|r
(s)
d*u
+
TT
as
*
3
Idvdu (1/8 2 F\
~
4
j- -3-+ 11 T^
v as as
[& \ou* J
Korteweg, Wiener Sltzungsber. xcm. (1886).
choosing
173, 174]
399
(which gives linear differential equations of the second order for the functions
and -vjr), we have two independent particular solutions of the latter equation
:
(f>
the general solution is a linear combination of these particular solutions, and

consequently the general equation of the orbits adjacent to the known orbit,
when k 2, is of the form
>
a*
u-K
where
and
Similarly
if
a*
*4>(*y
+ K e~~s + 0),
t
are arbitrary constants, and
writing k
2,
<
orbits adjacent to the
known
orbit
is
where
which
and
K^
and ^(s) have the period
a,
_os
(s)
<f>
K e~^^r (s\
z
2 are arbitrary constants, and where

the
satisfy
equations
l
S.
the general equation of the

of the same form
2 cosh
as
<j>(s)
and
$*
<f>
are functions of s
Next suppose that k

2
k
2
let k = 2 cos a.
In the
2, so that
same way we now find that the general equation of orbits adjacent to the
<
<
known
orbit is
where
K and A
<
are arbitrary constants and where
and
<f>
^r are functions of s
with the period S.
the
From these results important consequences relative

known periodic orbit can be deduced. For if k
2,
>
to the
stability of
it follows from the
character of the expressions obtained for u that the

divergence from the
and ^r have real zeros, the oscillation about it) becomes
periodic orbit (or if
<
continually greater as s increases while if k

2, the normal displacement
is represented by circular functions of real
will
arguments, and
;
<
consequently
obtain the theorem that a periodic orbit
is stable or not,
according as the associated index of stability is less or greater
(in absolute value) than two.
remain within fixed
limits.
We thus
The results of the present article agree with, and may be deduced
from, the theorem
that the general solution of a differential
equation of the type
(Pu
+ ((n +
2?rs
p
is
of the form
where
a, b are
period S.
Of.
cos
-- +
u = ae $
_+
47T
cos
\
.
+ be~ ^ (s),
C8
(s}
is a definite
constant, and $ and
Chapter xix.
arbitrary constants, c
^ are
periodic with
400
[CH.
xv
Example. Discuss the transitional case in which the index of stability has one of the
2 shewing that the equation of the
adjacent orbits is of one of the forms
values
u = KI
where
$ and
{//
s\l/-
(s)}
+ K,,
\js
(s),
or satisfy the equations
and that the known orbit may be either stable or unstable.

175.
(Korteweg.)
Characteristic exponents.
The stability of types of motion of more general dynamical systems

may
be discussed by the aid of certain constants to which Poincare has
given the
name
characteristic exponents*.
Consider any set of differential equations
where (X^,
...,
2,
is
Xn
T in t
having a period
are functions of (a^, a?2 ..., xn } and possibly also of t,

and suppose that a periodic solution of these equations
,
known, defined by the equations
where
fa
(t
+ T) = fa (t)
In order to investigate solutions adjacent to
where (1,
equations
n) are
(i
this,
we
1, 2,
n).
write
supposed to be small, and are given by the variational
112)
7
fit
\Juv
^ ^
cjfc
,
K= 1
"^
rtf,
^
(^ ^
==
J-
.
>
n).
As these are linear differential equations, with coefficients periodic in the

independent variable t, it is known from the general theory of linear differential
equations that each of the variables & will be of the form
where the quantities S{k denote periodic functions of t with the period T, and
the n quantities ak are constants, which are called the characteristic exponents
of the periodic solution.
If
(fi,
*
all
&,
the characteristic exponents are purely imaginary, the functions

fn) can evidently be expressed as sums and products of purely
>
Ada
Math. xin. (1890), p. 1 ; Nouv. Meth. de la Mec. Gel. i. (1892). On the general problem
of stability the reader should consult the extensive memoir of A. Liapounoff , originally published
in 1892 by the
Toulouse
Math. Soc. of Kharkow, and translated into French by E. Davaux, Annales de
(2), ix. (1907), p.
203.
174-176]
401
periodic terms; while this is evidently not the case if the characteristic
exponents are not all purely imaginary. Hence the condition for stability
of the periodic orbit
is
that all the characteristic exponents
must
be purely
imaginary.
We shall now find the equation which determines the characteristic

exponents of a given solution.
In one of the orbits adjacent to the given periodic orbit, let ({31} /32 ... @n )
initial values of (, %%, ...,
n ) and let /3j + fa be the value of
,
denote the
,-
after the lapse of a period.
As the
quantities (fa^faz, ...,fa n ) are one- valued

functions of (&, $,, ...,&), which are zero when (&, /3 2 ..., /3 n ) are all zero, we
,
have by Taylor s theorem (neglecting squares and products of
&, j82
@n )
If a k is one of the characteristic

exponents, one of the adjacent orbits will
be defined by equations of the form
and consequently a
set of values of
&
/3 2
@n exists for which the equations
(i=l,
are satisfied
2,
...,n)
the quantity a k must therefore be a root of the

equation in a
vfai
9^1
r
O.T
a
^- + l-e
OPi
-i
dfa2
9"^
Op 2
dfa 2
^-
f)ft
aT
T/ie characteristic
exponents are therefore the roots
9-^
of
this
determinantal
equation.
176.
When
Properties of the characteristic exponents.

t is
evident that
not contained explicitly in the functions

Xi
is
(X lt
X.,, ...,
Xn\
if
fa
(t)
a solution of the equations, then

e)
W. D.
26
it is
402
is
where
also a solution,
xv
The equations
e is
[OH.
&-*(*+)
(t-1,2,...,*)
therefore define a particular solution of the variational equations; but as

a periodic function of t, it follows that the coeffi
d(j)i(t + e)/de is evidently
cient e akt reduces in this case to unity: and hence when t is not contained
characteristic
explicitly in the original differential equations, one of the
exponents of every periodic solution
is zero.
Suppose next that the system possesses an integral of the form
F (#!
where
x2
...,
a one-valued function of
is
In the notation of the
{</>>
for brevity
this equation
F (#;) is
(0)
x^ = Constant
x2
(oc^,
+ }=F
i
by
fa (0),
<j)
Jacobian
(0),
8(<f lf
,
(j>
i/r 2
W^
,
From
8
..., A/r, l )/a( /
&,
...,
zero
all
...,
=
must be
x.2 ,
the quantities (x1} cc2

these equations
etc.,
n (0).
0,
xn ).
Differentiating
dF/dx,
is
0,
is
...,
it
xn ) are to be replaced
follows that either the
zero, or else the quantities
= 0.
when
teW<
dF/dxn are
correct, we see that (since the origin of time
dF/dx,, dF/dx2
is
t.
+ &},
(0)
F (i
tez
x n ) and does not involve
we have
/3 fj
dFty + dFd+
Wi*
^m
where in dF/dx 1} dF/dx2
(</>;
written in place of
with respect to
...,
we have
last article,
F
where
If the latter alternative
arbitrary) the equations
dFldaen
of the periodic solution: this is evidently

the former alternative must be in general the
satisfied at all points
a very exceptional case, and

true one but when the Jacobian
the determinantal equation for the

aT
evidently satisfied by the value e =l, i.e. by
is zero,
characteristic exponents is
=
so that one of the characteristic exponents
:
is
zero.
Thus if
the
differential equations possess a one-valued integral, one of the characteristic
exponents
is zero.
A comparison of 173, 174 with the theory of characteristic

shews that in the motion of a particle in a plane under the
conservative forces, the characteristic exponents of any periodic
a is connected with
a), where the characteristic exponent
(0, 0, a,
of stability k and the period T by the equation
k
the orbit
is
2 cosh
aT
stable or unstable according as a
exponents
action
the index
is
of
orbit are
purely imaginary or not.
176, 177]
Example
1.
If the differential equations
403
do not involve the time
explicitly,
and
possess p one-valued integrals /\, ...,

p which do not involve t, shew that either (p+l)
characteristic exponents are zero, or that all the determinants contained in the matrix
dF
(i
9^-
= l,
2,...,^;
= 1,2,
are zero at all points of the periodic solution considered.
...,),
(Poincare.)
If the differential equations form a Hamiltonian system, shew that the

Example
characteristic exponents of any periodic solution can be arranged in pairs, the exponents
2.
of each pair being equal in
Attractive
177.
magnitude but opposite
and
in sign.
(Poincare.)
a field of force.
repellent regions of
The general character of the motion of a conservative holonomic system

by a theorem which was given by Hadamard* in 1897. For
we
shall suppose that the system consists of a
simplicity,
particle of unit
which
is
free
move
on
a
to
smooth
surface
under
forces derivable
mass,
given
is
illustrated
from a potential energy function V; a similar result will readily be seen to

hold for more complex systems.
Let (u, v) be two parameters which specify the position of the particle on
the surface, and let the line-element on the surface be given by the equation
ds*
= Edu* + 2Fdudv +
where (E, F, G) are given functions of n and

particle
Gdv*
The
v.
kinetic energy of the
is
T = \ (Eu? +
2Fui>
Gtf},
and the Lagrangian equations of motion are
_
du
du
dt\du
dt\dv
.__
dv
dv
which can be written

du
-^
dv
du
du
-^
dv
du
dv
1
1
"
We
"
"\
~^
ff
"
ff
*
dv
if)uj
^
dv
~~~
1
*
~7
-
duj
have, by differentiation,
V=
T>
dV..
w
-^
du
dV..
+ -^+
dv
-y
37.
-~-
ou
dV
dv
_
-u +2
du
.,
Journ. de Math.
(5),
v,
^- uv + .
dudv
v.
dv*
in. p. 331.
262
404
Substituting for
u and
their values from
[CH.
xv
the preceding equations,
we have
where
8F/ .,9^
+ -=^^3^
dv
^
z.
drG dG + * FF dG (*
dv
(
\
du
dv
ZF
ZG
ZG
dv
9tt
dv
dv
quantities occurring in this equation can be expressed in terms of

The principal deformation-co variants connected
deformation-covariants*.
The
with the surface whose line-element
is
= Edu- + ZFdudv +
ds*
Gdv"
are the differential parameters
f ) = (EG - F*r
A,
where
W = (EG - F*r
and
<
With
-fy
\B%&-F&& + % f}
dv dv
Bj
2F
du dv
+ G
are arbitrary functions of the variables u
this notation, the preceding equation
du
dv du/
\du dv
and
v.
becomes
Utilising the equation of energy
Eu + 2Fuv + Gv*=2(h2
V),
and observing that the expression

3>
Eu>
The
(u, v)
+ 2Fuv +
- d V/du)
(d V/dv,
E(d V/dvf 2F(d V/dv) (d V/du) + G(d V/du)
<
Gv*
definition of a deformation -covariant is given in the footnote
on page
111.
177]
contains the quantity (udV/du
r=-A
vdV/dv) as a factor, we can write
8(-V)/_
1 (F>+
Al (F)
where \ and
yu,
405
+(XU + ^)F;
contain in their denominators only the quantity
and where I v denotes the expression

3>
we
readily find that
(d V/dv,
- dV/du)/(EG - F*)
I v can be expressed
in the form
Consider, on the orbit of the particle, a point at which

has a minimum
value at such a point
is zero and
is positive
as A, ( F) is essentially
positive (since the line-element of the surface is a positive definite form), it
follows that
zero,
i.e.
As
an
at
/FÔ, the inequality becoming an equality only when A, (F)

an equilibrium-position of the particle.
F will
the particle describes any

trajectory, the function
number of successive maxima and minima (this
infinite
is
either have
the general case)

or (in exceptional cases) the function
will, after passing some point of the
orbit, vary continually in the same sense.
Suppose first that the former of
these alternatives
is
is the true one

then if we divide the given surface into
two regions, in which 7 r is positive and
negative respectively, it follows from
what has been proved above that the former of these
contains all the
:
regions
contains in general
an infinite number of distinct
of
the
each
of
finite
orbit,
parts
length whereas
in the other
region, for which I v is negative, the particle cannot remain per
points of the orbit at which
Fhas
minimum
value,
i.e. it
manently.
attractive
These two parts of the surface are on this account called the
and repellent regions. Each of these
regions exists in general, for
easily found that
any isolated point of the surface at which F is a mini

any point where stable equilibrium is possible) is in an attractive
region, and any point at which Fis a maximum is in a repellent region.
it is
mum
(i.e.
It is interesting to
compare this result with that which corresponds to it in the motion
of a particle with one degree of
freedom, e.g. a particle which is free to move on a curve
under the action of a force which
depends only on the position of the particle. In this case
the particle either
ultimately travels an indefinite distance in one direction or oscillates
about a position of stable equilibrium. The attractive
region, in motion with two degrees
of freedom, corresponds to the
position of stable equilibrium in motion with one degree of
freedom.
Consider next the alternative

supposition, namely that after some definite
F is always in the same sense. We shall suppose that
the surface has no infinite sheets and is
at all
and that
instant the variation of
regular
points,
an everywhere regular function of

position on the surface
Fis
so that, since the
406
variation of
finite limit,
[CH.
xv
V is always in the same sense, V must tend toward some definite

V and V tending to the limit zero. Considering the equation
see that if A ( F) is not very small, X and /j, are finite and the last term
on the right-hand side of the equation is infinitesimal; and
consequently
either there exist values of t as large as we
please for which I v is positive (in
which case the part of the orbit described in the attractive
is of
we
region
length
But
greater than any assignable quantity) or else Aj(F) tends to zero.
Aj ( F) can be zero only when d V/du and d V/dv are zero if therefore (as is in
general the case) the surface possesses only a finite number of equilibrium
;
positions, the particle will tend to one of these positions, with a velocity
which tends to zero.
position of equilibrium thus
approached asymptotically
must be a
versed
is
for the asymptotic motion re

position of unstable equilibrium
a motion in which the particle, being
initially near the
:
equilibrium
position with a small velocity, does not remain in the neighbourhood of the
equilibrium position and this is inconsistent with the definition of stability.
;
Thus
follows
finally
If a
we obtain Hadamard s theorem, which may be
particle is free to
move on a surface which
is
stated as
everywhere regular
and has no
infinite sheets, the potential energy function being regular at all

points of the surface arid having only a finite number of maxima and minima
on it, either the part of the orbit described in the attractive region is
of length
greater than any assignable quantity, or else the orbit tends asymptotically
one of the positions of unstable equilibrium.
to
Example. If all values of t from - x to + oo are considered, shew that the particle must
be in the attractive region.
for part of its course
178.
Application of the energy integral
to the
problem of stability.
simple criterion for determining the character of a given form of motion
of a dynamical system is often furnished by the equation of
energy of the
Considering the case of a single particle of unit mass which moves

under the influence of forces derived from a potential energy
system.
in a plane
function V(x,
the equation of energy can be written
y),
+ y ) = h2
^(x-
Now
the branches of the curve V(x, y)
which [V(x, y)
V(x,y).
=h
separate the plane into regions
respectively positive and negative; but as (&-+y*)

an
orbit
for which the total energy is h can
essentially positive,
only exist in
the regions for which V(x, y)<h.
If then the particle is at any time in the
interior of a closed branch of the curve V(x,y} = h, it must
always remain
for
h] is
is
within this region.
The word
of motion in which the
and in
stable.
this sense
stability is often applied to characterise types
moving
we may say
particle
is
confined to certain limited regions,
that the motion of the particle in question
is
177-179]
in
407
The above method has been used by Hill*, Bohlinf, and Darwin J, chiefly
connexion with the restricted problem of three bodies.
179.
Application of integral-invariants
The term
was applied
to investigations
of
stability.
to a system which, in
the lapse of time, returns infinitely often to positions indefinitely near to its original
It has been shewn by
position, the intervening oscillations being of any magnitude.
Poincare that the theory of integral-invariants may be applied to the discussion of Poisson
stability
by Poisson
in a different sense
stability.
Considering a system of .differential equations
for
which
I ... I
fta?&B
...
8x
an integral-invariant, we regard these equations as denning the trajectory in n dimen

sions of a point
whose coordinates are (xv #2 -., %)
If the trajectories have no
is
branches receding to an infinite distance from the origin, it may be shewn that if any small
region R is taken in the space, there exist trajectories which traverse II infinitely often
:
and, in fact, the probability that a trajectory issuing from a point of R does not traverse
this region infinitely often is zero, however small
may be. Poincare has given several
extensions of this method, and has shewn that under certain conditions it is applicable in
the restricted problem of three bodies.
1.
Shew
2.
that the motion of a particle in an ellipse under the influence of two fixed
Newtonian centres offeree is stable.
(Novikoff.)
is free to move in a
plane under the action of several
according to the Newtonian law of the inverse square of
denoting the resulting potential energy of the particle by V(x, y\ shew that
mass
particle of unit
centres of force which attract
the distance
it
the integral
I
1-
log {h
V (a-%
where the integration is taken over the interior of any periodic orbit for which the constant
of energy has the value h (the centres of force
being excluded from the field of integration
by small
circles of arbitrary infinitesimal radius),

enclosed by the orbit, diminished by two.
3.
equal to the number of centres of force

(Monthly Notices R.A.S. LXII. p. 186.)
Let a family of orbits in a plane be defined by a differential equation
is
where (#, y) are the current rectangular coordinates of a point on an orbit of the
family
and let 8n denote the normal distance from the point (,v, y) to some definite
orbit
;
adjacent
of the family.
Shew that 8n
Amer. J. Math.
i.
satisfies
the equation
f Acta Math.
(1878), p. 75.
+ Acta Math. xxi.

(1897), p. 99.
Poincare, Acta Math. xin. (1890), p. 67
x. (1887), p. 109.
Nouv. Meth. in. Ch. xxvu.
408
where
r
fdy\
4 1 +
/=
-/
and
I-,
M( - 3$
^ + dy S0\
^-) + {(b(x,
-f-
[CH.
y)}
a variable defined by the gquation
is
dt
*:l
\dx
(Sheepshanks Astron. Exam.)
A particle moves under the influence of a repulsive force from a fixed centre shew
4.
that the path is always of a hyperbolic character, and never surrounds the centre of force
that the asymptotes do not pass through the centre in the cases when the work, which has
:
to be done against the force in order to bring the particle to its position from an infinite
distance, has a finite value but that when this work is infinitely great, the asymptotes
;
pass through the centre, and the duration of the whole motion
may
be
finite.
(Schouten.)
5.
Shew that in the motion of a particle on a fixed smooth surface under the influence
of gravity, the curve of separation between the attractive and repellent regions of the
surface is formed by the apparent horizontal contour of the surface, together with the locus
of points at which an asymptotic direction
is
horizontal.
moves freely in space under the influence of two Newtonian centres of

shew that when its constant of energy is negative, it describes a spiral curve
round the line joining the centres, remaining within a tubular region bounded by two
ellipsoids of rotation and two hyperboloids of rotation, whose foci are the centres of force
and that when the constant of energy is zero or positive, the particle describes a spiral
path within a region which is bounded by an ellipsoid and two infinite sheets of hyper
boloids of the same confocal system.
(Bonacini.)
6.
particle
attraction
The necessary and sufficient condition in order that a two-parameter family of

7.
curves defined by a differential equation
y"=(x,y,
may
/)
be a system of orbits with the same constant of energy
shall
be a linear homogeneous perfect differential.
The
is
that
potential energy
is
then a
constant multiple of
.-*f(8-&)*
(P. Frank.)
In the motion of a particle in a plane under forces which depend only on its position,
8.
a one-parameter family of trajectories is obtained by starting particles at a given point in a
given direction with all possible velocities. Shew that the locus of the foci of the osculating
parabolas is a circle passing through the point. If the initial direction is now varied, shew
that the locus of the centres of the oo 1 circles obtained is a conic having the given point
as focus and if the forces are conservative, this degenerates into a straight line counted
:
twice.
9.
In order that a system of oc 5 space-curves, of which x 1 pass through every point

may be identifiable with the system of trajectories of a particle in an
in every direction,
arbitrary positional field of force, it

should have the following properties
is
necessary (but not sufficient) that the system
The
(a)
a pencil
that
2
curves passing through a given point form
osculating planes of the oo
the planes pass through a fixed direction.
is, all
xv]
The osculating spheres
(/3)
direction form a pencil
of the
oc
409
curves passing through a given point in a given
their centres thus lie on a straight line.
10.
Shew that the on 2 curves of a natural family which meet any surface orthogonally
are orthogonal to oc 1 surfaces, that is, form a normal congruence.
(The surfaces in
question are the surfaces of equal Action.)
(Hamilton.)
Shew
11.
10 belongs exclusively to natural
that the property referred to in Ex.
families.
In order that a family of
12.
oc 4
curves in space
two properties are necessary and
orbits,
sufficient, viz.
may
constitute a natural family of
If the osculating circles of those curves of the family which pass through a given
are constructed at that point, they have a second point
in common, and thus
(a)
point
form a bundle.
Consequently, three of the circles in such a bundle have four-point contact

with the corresponding curves.
(/3)
These three hyperosculating
circles will
be mutually orthogonal.
13.
The only point-transformations which convert every natural family into a natural
family are those belonging to the conformal group.
[Examples
8, 9, 11, 12,
13 above are taken from memoirs by E. Kasner in the Trans
actions of the Amer. Math. Soc., 1906-1909.

For further work in this direction the reader
is referred to Professor Kasner s Princeton
Colloquium lectures on Differential Geometric
Aspects of Dynamics.]
14.
Two
sets of oo
a certain conservative
particle considered as
y)
that
(x,
when the
curves in a plane, which form an orthogonal system, are orbits in

If U denote the Action at any point (x, y~] of a
field of force.
moving on one
of the first set of orbits, and

denote the Action at
moving on one of the second set of orbits, shew
particle is considered as
U and V are conjugate functions

U= constant, V constant, are identical
of x-
and y
and that the families of curves
with the orbits.

(P. G. Tait
and K. Ogura.)
CHAPTER XVI
The need for
180.
Poincares
We
which converge for
series
all
values of the
time;
series.
have already observed
32) that the differential equations of motion
of a dynamical system can be solved in terms of series of

ascending powers
of the time measured from some fixed epoch
these series converge in
;
within some definite circle of convergence in the

-plane, and consequently will not furnish the values of the coordinates
except for a limited interval of time. By means of the process of analytic
general for values of
continuation* it would be possible to derive from these series successive sets

of other power-series, which would converge for values of the time outside this
interval but the process of continuation is too cumbrous to be of much use
;
in practice, and the series thus derived give no insight into the
general
character of the motion, or indication of the remote future of the system.
The efforts of investigators have therefore been directed to the problem of
expressing the coordinates of a dynamical system by means of expansions

which converge for all values of the time. One method of achieving this
is to apply a transformation to the
-plane.
Assuming that the
motion of the system is always regular (i.e. that there are no collisions or
other discontinuities, and that the coordinates are always finite), there will be
no singularities of the system at points on the real axis in the -plane, and
result j-
the divergence of the power-series in t

1
after a certain interval of time
must therefore be due to the existence of singularities of the solution in the
finite part of the -plane but not on the real axis.
Suppose that the singu
larity
and
which is nearest to the real axis is at a distance h from the

T be a new variable defined by the equation
t
~t =
band which extends
log 1
to a distance h
-plane evidently corresponds

*
I+T
2h
7T
real axis
let
on either side of the real axis
to the interior of the circle
Whittaker and Watson, Modern Analysis, 5

t Due to Poincare, Acta Math. iv. (1884), p. 211.
Of.
5.
JT
=1
in the
in
the
411
Integration by Trigonometric Series
180, 181]
the coordinates of the dynamical system are therefore regular

r-plane
functions of T at all points in the interior of this circle, and consequently
within
they can be expressed as power-series in the variable T, convergent
for all real values of T
this circle.
These series will therefore
;
converge
between
and
1, i.e.
for all real values of
these series are valid for
between
oc
and
oc
Thus
all values of the time.
181.
The regularisation of
In the
last article
the
we made the
Problem of Three Bodies.
reservation that there are to be no collisions
or other discontinuities for real values of
t.
The importance
of collisions in
the mathematical theory of the Problem of Three Bodies was first indicated
by Painleve *, who shewed that the motion of the bodies is regular (i.e. their
coordinates are holomorphic functions of t) for all time, provided the initial
conditions are not such that after a finite interval of time two of the bodies
The
collide.
relations -which
must subsist between the
initial values of
the
may ultimately happen between two of

the three bodies have been discussed by Levi-Civitaf for the restricted problem
of three bodies (when there is one such relation) and by Bisconcini \ for the
general problem, when there are two relations these relations are analytic,
but they are expressed by somewhat complicated infinite series, and are not
variables in order that a collision
directly applicable except
and the
when the
interval of time
between the
initial instant
collision is sufficiently short.
considerable advance was
made when K.
F.
Sundman shewed
that the
a collision of two
singularity of the differential equations which corresponds to
of the bodies is not of an essential character, and that it may in fact be
removed altogether by making a suitable change of the independent variable
that is to say, it is possible to choose the variables which specify the motion,
and the independent variable, in such a way that the differential equations of
motion are regular even when two -of the three bodies occupy coincident
It is thus possible to obtain a real prolongation of the motion
positions
||.
after the collision IF: the coordinates can be specified for all values of the time t
oo to + oo
from
whether collisions take place or not and a positive lower
:
bound
can be assigned to the two greater of the mutual distances.
There
Lemons sur la theorie anal, des eq. diff ., Paris, 1897, p. 583.
f Annali di Mat. (3) ix. (1903), p. 1 ; Comptes Rendus, cxxxvi. (1903), pp. 82, 221.
* Acta Math. xxx.
Cf. also H. Block, Medd. fran Lunds Obs., Series n., No. 6
(1905), p. 49.
(1909); Arkivf. Mat. Astr. och Fys. v. (1909), No. 9.
Acta Math, xxxvi. (1912),
p.
105.
The
essential
features of the
work were
originally
published in Acta Societatis Sclent. Fennicae in 1906 and 1909.

Levi-Civita regularised the differential equations of the restricted problem of three bodies
by an elementary transformation in Acta Math. xxx. (1906), p. 306; and in a later paper, Rend,
d. Lincei, xxiv. (1915), p. 61, he extended this to the problem of three bodies in a plane.
||
If
The
variables can be expanded in ascending powers of (ti - 1)&, where

the orbits hare cusps at the point of impact.
instant of collision
ti
represents the
412
xvi
[CH.
only one case of exception, namely when all three bodies collide simul
taneously but this can happen only in a very special type of motion, in which
all the constants of
angular momentum are zero together*.
is
Disregarding this case of triple
independent variable
where r
Sundman
collision,
new
introduced a
r 1} rz denote the three mutual distances, and I is the lower bound

The coordinates of the bodies, and the time, are then
already mentioned.
holomorphic functions of
bounded by two
exists
within a band of finite breadth
2O
in the w-plane,
and on either side of it. There

a continuous one-to-one correspondence between the real values of t and
lines parallel to the real axis
the real values of w, so that

varies from
oo to + ac
when
varies from
oo
to
oo
likewise
Lastly,
Sundman
applied Poincare
transformation
I+T
= 2H,log8
,
iv
1-T
7T
in order to transform the
the plane of a
time, are
new
band
variable
now holomorphic
r.
in the w-plane into a circle of radius

unity in
The coordinates of the three bodies, and the
functions of T everywhere within the unit circle in
the r-plane and therefore they can be expanded as convergent series of powers
of T for all real values of the time, whether there are collisions or not the
case of triple collision alone being excepted.
:
182.
Trigonometric
series.
The
series discussed in the preceding articles are all

open to the objection
that they give no evident indication of the nature of the motion of the
system after the lapse of a great interval of time

the
they also throw no light on
number and character
in the
of the distinct types of motion which are possible

and the actual execution of the processes described is attended
problem
with great difficulties. Under these circumstances we are led to investigate
expansions of an altogether different type.
:
If in the solution of the problem of the simple

pendulum ( 44) we consider
the oscillatory type of motion, and replace the elliptic function by its ex
we have
pansion as a trigonometric series
,
27T
This
last fact
had been known
g4.-
~~
(2s-l)^(t-t )
to Weierstrass
cf.
Acta Math. xxxv.
p. 55.
The motion
is
then in one plane.
A simpler equation available in the restricted problem

Armellini, Comptes Rendus, CLVIII. (1914), p. 253.
t
G-.
J Cf. Whittaker and Watson, Modern Analysis,
22
6.
of three bodies
was given by
413
181-183]
where 6 denotes the inclination of the pendulum to the vertical at time t

and t may be regarded as the two arbitrary constants of the solution, and
is a definite constant, while q denotes e~ vK IK where
is the
complete
to
K.
This
each
term of which
elliptic integral complementary
expansion,
/j,
a trigonometric function of t, is valid for all time.

Moreover, when the
constant q is not large, the first few terms of the series give a close approxi
mation to the motion for all values of t.
The circulatory type of motion
is
of the
same
pendulum may be
similarly expressed
by a trigonometric
series of the
general character.
Turning now to Celestial Mechanics, we find that series of trigonometric

terms have long been recognised as the most convenient method of
expressing
members
the coordinates of the
of the solar system
type
2tfn,,n 2
...,n fc
cos (nl 6l
n.2
62
-\-
...
these series are of the
+nk
k ),
where the summation is taken over positive and negative integer values of
% 1} nz ..., n k and 6 r is of the form \ r t + e r the quantities a, X, and e
being
constants.
Delaunay* shewed in 1860 that the coordinates of the moon can
be expressed in this way; Newcombf in 1874 obtained a similar result for
the coordinates of the planets, and several later writers J have
,
designed
processes for the solution of the general Problem of Three Bodies in this
form these processes are also applicable to other dynamical
systems whose
equations of motion are of a certain type resembling those of the Problem
of Three Bodies.
In the following articles we shall give a method which
is applicable to all dynamical
systems and leads to solutions in the form of
;
trigonometric series the method consists essentially, as will be seen, in the

repeated application of contact-transformations, which ultimately reduce the
problem to the equilibrium -problem.
:
Removal of terms of the first degree from the energy function

Consider then a dynamical system, whose equations of motion are
183.
dqr_dH
dt~dp r
where the energy function
The
dp r
__ dff
dt~~
1,2, ...,w),
dfr
H does not involve the time
explicitly.
algebraic solution of the 2n simultaneous equations

8
/=
dp r
will furnish in general

for the variables (q,, q.
0,
^=
(r-1,
oqr
one or more sets of values
2>
...,
q n ,p lt ...,p n );
(a,,
az
...,
*.,,)
a n b l} 62
and each of these
..., 6 n )
sets of values
Theorie du mouvement de la lune. Paris, 1860.

J e.g. Lindstedt, Tisserand, and Poincare.
Whittaker, Proc. Land. Math. Soc. xxxiv. (1902),
f Smithsonian Contributions, 1874.

p. 206.
414
xvi
[OH.
will correspond to a form of equilibrium or (if the above equations are those
of a reduced system) steady motion of the system.
b n ) be selected
Let any one of these sets of values (a a2 ..., a n b 1} b 2
shall shew how to find expansions which represent the solution of the
problem when the motion is of a type terminated by this form of equilibrium
l
. .
we
Thus if the system considered were the simple pendulum,

or steady motion.
and the form of equilibrium chosen were that in which the pendulum hangs
which would
the motion is of the
represent the solution of the pendulum problem when
downwards
vertically
aim would be
at rest, our
to find series
oscillatory type.
Take then new
variables
equations
qr
ar
(q,
q2
qr
qn
...,
pS,
p r = b r +pr
by the
...,p n \ defined
(r
1,2,
n)
...,
the equations of motion become
dH
_
dqa r
7.
dt
and
"i
the
for sufficiently small values of
where
*>
/
)
">
ni
"/>
H can be
form
...,
denotes terms homogeneous of the kih degree in the variables
(qi, q*,
Since
variables the function
H=H + H + H. + H,+
l
dq r
new
series in the
7i
dt
dp,
expanded as a multiple power
dH
dpr
)
q n ,pi,
>
-,pn)-
does not contain any of the variables,
it
may be omitted and the

:
fact that the differential equations are satisfied when

q z ..., q n p, ..., p n )
vanish
should
are permanently zero requires that
l
identically. The expansion
which
with the terms
therefore
of
2
(suppressing the accents of
,
(<//,
the
begins
new
variables)
H.2
= ^2
may
(a rr g r
where
tt r s
+ 2b n qrp +
= asr,
c rs = c sr
2orS q r q s )
(c rr
pS
4-
2c n p r p,\
H H
If the terms
not necessarily equal to b sr
4 ,... were neglected
s
those
of a vibrational
become
would
the
with
in comparison
2
equations
but b n
is
problem (Chapter VII).

184.
We
Determination of the normal coordinates by a contact-transformation.
now apply
a contact-transformation to the system in order to

in fact, to obtain variables which correspond
2 in a simpler form *,
express
to normal coordinates for small vibrations of the system.
shall
*
In obtaining the transformation of this article a method is used which was suggested to the
author by Dr Bromwich, and which furnishes the transformation more directly than the method
originally devised.
415
183, 184]
Consider the set of 2n equations
(r
~ 8aSr
z^
^dy r
syr = a
or
rl
x +
1
a rz x.2
J
U.A
o
lr x l + o.2r xz
+
+
x
.
*>
Xn
>
y^)-
y^>
>
+ a rn xn + n y^ +
\)
= l,2,...,w)
+ brn yn\
=
+ crn yn l(r l,2,...,n).
+ cn y +
)
we obtain for s the determinantal equation
which in 84 was denoted by f(s) =
we shall suppose that 2 is a positive
definite form, and (as in
84) we shall denote the roots of the equation
is 2
is n
the quantities s lt s. ... sn are all real, and for
isi,
...,
by
simplicity we shall suppose no two of them to be equal.
sx.r
On
j.7o x
nr n
...
solving these equations,
To each
root there will correspond a set of values for the ratios of the
.., xn y lt ..., y n }; let the set which correspond to the root
quantities (xl xz
,
be denoted by (y^, r x2 ..., fXn ry^,

is r be denoted by
spond to the root
is r
that
and
..., r y n ),
_,^2
(_,.#!,
the set which corre
let
...,
-rXn - ryi,
,
...,
-r
so
yn)>
we have
Multiply these equations by kxp and k yp respectively, add them, and

respect to p we thus obtain the equation
sum
with
n
\ jcp k yp
2,
is,.
= -H
yp )
ifKp r
(?% k),
where
1
\ i
/t/ )
ll ^*** i
-,
so that
k)
(r,
is
Ic^l
<~
^**3
jx-
*~
Jb"^2
K^\ T^^f
^H
"
""
symmetrically related to r and
Interchanging r and
k,
vQb]
jfc
Vi
r* Jt^- i t 1/i /
~T~
k.
we have
is k
(kXp
rXp kyp )
yp
(r,
k\
P=I
n
and therefore
(s r
+ sk ) ^
P=I
So, unless s r
+ sk
is zero,
(^
r yp
^xp
yp ) =
0.
we have
n
^~
-^
and consequently H (r, k)

=
and therefore
k yp
\.^^p kyj)
is
zero
if
p
sr
jPs
+ sk
is
zero,
-,-!/]>,
n
1S r
p=l
\rXp
ryp
r Xp
ryp)
"
\f>
**/
we have
^p = -rXp

If
now we
n
~fr
{
7"
;2 1
new
define
V~T*1 6
>
variables
74
/ *2
((?/,
q2
q n PI,
...,
I"*//-
Pn) by the equations
/*
_-
lit
vr
Pr = tfr
and
and
if 8
J.
"i
denote any two independent modes of variation,
n).
evident
it is
2 (Sqi&p -
thai the coefficient of tar Ac* in
xvi
[CH.
2 (M - kyi - -&i
is
bqiSpi)
f yi),
1=1
which
zero
is
when
is
not equal to
terms except such as (Sq r &p,.
- ryi
(i-ici
-ri
Now
ryi)-
Thus 2 (Sqi&pi
k.
- Agv &p/), and
AqiSpi) contains no
the coefficient of this term
hitherto the actual values of
is
have not been
rXi, r yi
determined from their equations of definition
fixed, as only their ratios are
therefore choose their values so that
we may
2 O* - ryi - -M ryi) = 1
(r
i, 2,
n),
1=1
and then we
shall
have
n
&qi$pi)=
(8qi&pi
(Sq r
r=l
1=1
&p r
Agv Sp/),
Pn)
128) the transformation from the variables (ql} q 2 ...,qn PI,
to the variables (g/, q%, .... q n pi, ..., p n ) is a contact-transformation.
substitute for (q lt q 2 ..., q n plt ...,p n ) in terms of
2 we
Moreover, if in
so that
(q\>
->
q*>
qn, Pi,
Pn),
we obtain
H = r=l
2 H
2
H=
or
- r) qr pr
(r,
r=l
Now
sr q r
pr
q2
apply to the variables
transformation defined by the equations
...,
(<?/,
qn
PI,
dW
Pr
W=
where
As
all
...
variables
"
+ i ip
(p r
which gives
H H
"q
pn
the contact-
Ir
<n
^r
...,
~dqr
iis r q,
=i 2
r=l
"
(p r
+ sr
the transformations concerned
"
qr
2
).
have been
linear,
we
see that
be homogeneous polynomials of degrees 3, 4, ... in the new

and thus, omitting the accents, we have the result that the
will
equations of motion of the dynamical system have been brought
dq
dt
_m
~
dp r
dp_r__dH
~
dt
dq r
to
the
form
H= H + H + H
where
in which
417
184, 185]
...,
a homogeneous polynomial of degree r in the variables, and in
r is
particular
H H
we neglect
... in
and
S)
4
2
comparison with
the
integrate
equations, the solution obtained will be identical with that
It is clear that if
found in
185.
84.
to the
Transformation
The system
now be
will
form of H.
trigonometric
further transformed
by applying
transformation from the variables (q lt q 2 ..., q n p^,

q*,
q n Pi,
p n ), defined by the equations
,
(<?/,
pn )
it
a contact-
to
new
1, 2,
variables
dW
W=
where
...,
to
S
|
r=l
dW
q r arcsm
so that
p r = (2sr q r ^ sin p r
The
qr
r )?
~$
sr
cos
pr
(r
).
"
dt
~dt
dp r
H = s^ + s
and now
(<2q
become
where
q,
...
+H + H
sn q n
+...
an aggregate of terms which are

r denotes
homogeneous of
degree \r in the quantities qr and homogeneous of degree r in the
,
quantities cos_p/, ainp r
Since a product of powers of cos p r sin p r can be

expressed as a sum of
and cosines of angles of the form (n^pi + n 2 p.J + ... + n n n ), where
p
,
sines
n n have integer or zero values, it follows that

as the sum of a finite number of terms, each of the form
n 1} n 2
...,
"
tf^ft
where
...
and therefore
|
The
function
,
where
for
H
)H
is
.....
(n lPl
mn =
n x + nz
+n p +
2
r,
[
. .
nr
can be expressed
+ n n pn
...
^.
2m r
),
r.
thus expressed in the form
nin
/*
/*
olll
..
each term we have
% +
w. D.
L-Ub
+m +
m,
si
mn
qn
na
. . .
+ nn ^
|
2 (m,
-t-
m +
2
+ mn
),
418
[CH.
and
clearly the series is absolutely convergent for all values of p^,

q2 ..., q n do not exceed certain limits of magnitude.
provided
p2 ..., pn
From the
absolute convergence
<?/,
xvi
follows that the order of the terms can be rearranged
it
any arbitrary way we shall suppose them so ordered that all the terms
+ ... + n n p n f are collected together, so
involving the same argument
in
that
n^
takes the form
where the coefficients a and b are functions of

q2
q n and the expansion
i
n
of
contains
no terms of order
or
b niy n ,...,n n
of a n ,n ,...,nn
qi,q2 ..., ^n
powers
extend
n
and
where
the
summations
lower than
^ + 2 + ... +\nn \\;
over all positive and negative integer and zero values of n 1} n 2 ..., nn except
,
<//,
the combination
?i x
= nz =
= n n = 0.
Moreover, the expansion of a 0j0 ,...,o (which will be called the non-periodic part
of H, the rest of the expansion being called the periodic part) begins with
the terms
MI + Ma +
+s
n<ln
these are the most important terms in H,

and,
g/, q2 ...,q n are small,
since they contribute terms independent of
q2
q n to the differential
when
<?/,
equations.
For convenience we shall often speak of g/, q2 ..., q n as "small," in order

to have a definite idea of the relative importance of the terms which occur.
It will be understood that q^, q^ ..., q n are not, however, infinitesimal, and
,
magnitude except so far as

ensure the convergence of the various series which are used.
in fact are not restricted at all in
To avoid unnecessary complexity, we
is
required to
shall ignore the terms
sin (n^pi + ... + n n pn}

S^.n,,
in
the same way as the terms
in H. as they are to be treated
...,
Sof n ,,n 2 ,...,,,
cos(n
p +
1
and their presence complicates, but does not
...
in
+ n n pn),
any important respect modify,
the later developments.
which the problem has now been brought may therefore be

stated as follows (suppressing the accents in the variables) The equations of
The form
to
motion are
aqrdH
dt
where
H=u
0(0 .....
+ 2a n
dpr
dp r
,
,...,
dt
(r=1;2)
..,.),
dqr
cos (n
p +
l
n2 p2
...
+ n n pn \
a are functions of q ,q2 ...,q n only ; moreover, the periodic

a term which
small compared with the non-periodic part a
...,o /
at least
has for argument (n p 1 + n 2 p 2 + ... +n n p n ) has its coefficient a ^, ...,
and
the coefficients
part of H
is
0)0>
ni>
of order ^
% +
,
the expansion
It follows
of a
n2
+ nn
j
in the small quantities q lt q2
begins with the terms
,o, ...,o
+s
(s l q i
from this that when the variables q lt q2
vary very slowly, while the variables p 1}

to the time.
Other types of motion which lead
186.
419
185, 186]
q2
..,,
...,
qn
and
+ sn q n \
...
qn are small they
p n vary almost proportionally
...,
to
equations of the
same form.
The equations which have now been obtained have been shewn to be
applicable when the motion is of a type not far removed from a steady motion
or an equilibrium -configuration, e.g. the oscillatory motion of the
simple
pendulum, or those types of motion of the Problem of Three Bodies which
have been studied in
171.
But these equations may be shewn to be
applicable also to motion which is not of this character, and in particular to

motion such as that of the planets round the sun, or the moon round the
earth *.
For
the equations of motion of the Problem of Three Bodies be taken

160; and let the contact-transformation which is
let
in the form obtained in
Pr
= dW
~,
r-~
oqr
be applied
or
(r
dq r
= l,2,3,4)
where
to this system,
-*
+
The new
--
77,
--
^r
I-
dq.
variables can be interpreted in the

following way.
Suppose that at
the forces acting on the
particle //, cease, except a force of
magnitude m^m^q^ directed to the origin and let a be the semi-major axis
and e the eccentricity of the ellipse described after this instant then
the instant
all
= it^mâ
q,
-e
q3
Further, if the lower limits of the integrals are suitably chosen, /?/+ q 3 is
the true anomaly of /* in its ellipse, and - 3 is the mean
p
anomaly. The
variables qj, q t
p 2 p^ stand in a corresponding relation

The equations of motion now take the form
,
dqr
_dH
W-fa"
to the particle
dH
dp;
(^
///.
= 1,2,3,4);
"9ff
when the particles w 2 and ra3 are supposed to be of small mass

compared with
Wj, and are describing orbits of a planetary character about
l} it is readily
found that
can be expanded in terms of the new variables in the form
H
H=
ao,o,o,o
2a, li)W2i , l3Fn4 cos (n^pf
Delaunay, Theorie de la Lune
+ n p2 + n p + n^
Tisserand, Annales de
),
Obs. de Paris, Memoires,
xvm.
(1885).
272
[cu. xvi
420
q a q4 ) only, the summation

zero values of n lt w 2 n s ?i 4
and
extends over positive and negative integer
and the coefficient a 0)0)0)0 is much the most important part of the series. As
185, it
is of the same character as that obtained in
this expansion of
where the
coefficients
a are functions of
(qi, q,
method of solution given in
follows that the
motion of the planetary type or
either to
to
the following articles is applicable

motion of the type studied in 171.
Removal of a periodic term from H.
187.
We
effect of
which
will
the
system another contact-transformation,

this
be the removal of one of the periodic terms from
now apply
shall
to the
will further accentuate the feature already noted,
part of
is
much more important than
Let one of the periodic terms in

n,,n 2 ,...,nn
H=a
Write
so that
...,<>
R denotes the
in evidence the
Apply
...,.
cos
n n p n )-
+ n *P* +
+n
+ n ?P2 +
<XPi
rest of the periodic
arguments of which
to the
the periodic part*.
H be selected, say
cos ( n i.Pi
,,
terms of
we
suppose that
indicated.
^P^>
+R
>
when we wish to put

we shall write it
a function,

system the contact-transformation
W = qfa + q
shall
is
<*,,...,
= dW
,_dW
where
namely that the non-periodic
p2 +
+ q n p n + f (?i
...
?2
?n 0)
,
a function, as yet undetermined, of the arguments

is now expressed by the equations
is
The problem
"
dt
dt
dp r
dq r
where
8/
)
-f-
and 6 and
means
~Q
...
qn
+ w-n o/j
are supposed to be expressed in terms of the
cos ^
)
new
"T"
variables
>
by
of the equations of transformation
df
"tf
Readers familiar with Celestial Mechanics will notice the analogy of this method with that
the idea is essentially
Delaunay s lunar theory: the analysis is different from Delaunay s, but
*
of
Hi
the same.
186, 187]
The function/ is, as yet, undetermined and at our disposal. It will be

chosen so as to satisfy the condition that 6 shall identically disappear from
the expression
D, o, .... o
+ HI -
qn
nn
,
+
so that this quantity
is
n,,
...,
a function of
<?/,
a>
o,
~\f
+ MI
( qi
9a
...
,...,o(qi,qs,
g0
"r\f\
+ nn
qn
cos
j
,q n alone, say
)
></
Then the equation

9
9/
(*,
(
nn
.....
determines 9/730 in terms of
qi
<//,
3/
In
fig
<?
>
>
^n
+ n n 9A
.....
o,
o>
cos B
and cos
=a
8.
Suppose that the solution of this equation for f)f/d& is expressed in the
form of a series of cosines of multiples of 6 (which can be done, for instance,
so that
by successive approximation),
df
ov
^=
where
Now
c2
Cj,
0,0,.,.,0
is
condition that c
qi,
q-2,
>
qn
are
...
known
+ 2
c k cos kO,
k=i
functions of q^, q^,
qn
...,
a
0>0
.....
as yet undetermined, and is at our disposal.

to be zero; this determines a ,o,...,o as a
is
Impose the
function of
and, on substituting its value in the series for df/dO,
dffi8= I
ck
we have
cosk0,
k=i
c 3 ... are known functions of

q^
q2 ..., qn
Integrating
with
this equation
respect to 6, and for our purpose taking the constant of
integration to be zero, we have
where now
c 1} C 2
1 ^sin&tf.
/= t-i*
The equations denning the transformation now become
% 1 dc k
=p-r+ 2
75-7
k - lkd ^
Pr
qr =
ia \
Sin fvd
^
qr
nr
ck
kd
cos
-\i,
2
&,..., n)
ii).
k=--\
Multiply the
add them
first set
of these equations
writing
WiK +
1
we have
/V
=6+ 2
/I
-V*
*-
,-
"vie
/
I
n 2 p2
?h
5^5
+ w2
...
by n lt n z
+ nn p n =
OClf
,
...,
nn
respectively,
and
422
[OH. xvi
Reversing this
we have
series,
=&+ S
known
dk
sin
ke\
of
Substituting this
q 2 ..., qn
of
become
value of d in the equations
transformation, they
where d 1} d 2
...
are
functions
<//,
sn
COS
where
the coefficients
all
Now,
9k are
Vk>
known
functions of
before the transformation, the function
g./,
<?/,
#>/
consisted of an aggregate
of terms of the type
R = 2a Wi
)OT2>
...
>TOB
cos (w^ PJ
+ mn p n )
found for (q 1} q ..., q n p\,

,pn) are
reduced
is
series
and
the
in
this
substituted
by replacing powers
expression,
and products of trigonometric functions of
p n by cosines of sums
p.?,
values which have been
when the
/>/,
of multiples of p, p,
of terms of the type
,
...,
R = 2a mi)Ml2)
where the
coefficients
pn
...
a are
it is
>Mn
clear that
an aggregate
will consist of
+m p +
cos (m^pi
known
2>
...
+ mn p n
functions of (#/, q^,
qn
...,
),
}-
We thus have the result (omitting the accents of the new variables) that
has been effected, the system is still expressed by a set
after the transformation
of equations of the
form
dq r
dt
where
H = aô.-.o + Sa
dp r
dp r
dt
_dH
m,,
,,,...
cos
>Wn
~~
dH
i
x,
**
dq r
(m p +
l
p2 +
...
+ m n pn),
a are known functions of q 1} q2 ..., qn

Let us now review the whole effect of the transformation. The differential
as before; but from the
equations of motion have the same general form
and where
the coefficients
equation
Oo,o,...,o
a ni ,n
,...,n n
COS
(n^ +
?? 2
>
...
+ nn pn) =
,o,...,o
to its
we see that one term has been transferred from the periodic part of
is less important, in comparison
of
the
periodic part
non-periodic part
with the non-periodic part, than

188.
it
was before the transformation was made.
Removal of further periodic terms from H.
into the nonHaving now completed the absorption of this periodic term
of the new
terms
of
the
one
absorb
to
we
of
H,
periodic
proceed
periodic part
of
the same
a
the
into
repetition
non-periodic part, by
expansion of
the
enrich
can
In this way we
non-periodic part of
continually
process.
423
187-189]
at the expense of the periodic part, and ultimately, after a number of

will become so
applications of the transformation, the periodic part of
an /3 1; /32
o
Let
that
it may be neglected.
l
2
...,
/3 n ) be
(a
insignificant
the variables at which we arrive as a result of the final transformation then
H
,
>
the equations of motion are

doi r
_ dH
=
~di
where H, consisting
(
1} a.2 ,
...,
of
__
dT~~
d@r
only
We
a n ) only.
dp r
its
dH
~^r
non-periodic
is
part,
function
*~/g*
-0,
(~1,
*,...,>.
which shews that the quantities a are constants, and the quantities
the form
ft,.
the quantities
(d,
cr
an)
...,
189.
e,.
is
of
have therefore
fji r
er
where a r
dH
(r
-^
oar
are arbitrary constants, and the part of
fi r
are of
ft
1, 2, ...
n);
independent of
-ST.
Reversion
to the original coordinates.
Having now solved the equations of motion in their final form, it remains
only to express the original coordinates of the dynamical system in terms of
the ultimate coordinates (a 1 a2 ..., an /3 l5 ..., (3 n ). Remembering that the
,
result of performing
any number
of contact-transformations in succession
a contact-transformation, it is easily seen that the variables (q 1} q.2

p lt ...,pn ) used at the end of 185 can be expressed in terms of ( 1( 2
/3]
/8 n )
...
by equations of the form
mn
where the
is
..., q n
coefficients
sn
a and
6 are functions of (a 1} a 2
....
a n ).
From this it follows that the variables (q^, q 2 ..., q n ,pi, --^Pn) of 183,
in terms of which the configuration of the dynamical system was originally
expressed, are obtained in the form of trigonometric series, proceeding in
,
sines
and cosines of sums of multiples of the n angles
angles are linear functions of the time, of the form
@i, /3 2
t
fj, r
..., {3 n .
er
These
the quantities
424
er
/j, r
[CH. xvi
are n of the 2n arbitrary constants of the solution, while the quantities

are of the form
Pr
= -*.- +
S
*1, K 2
the coefficients
c kl ,*..., k n
i*
a*
n*
n
,
...
being independent of the constants of integration.
The
coefficients in the trigonometric series are functions of the arbitrary constants
(i,
n) only.
-,
The expansions thus obtained represent a family of solutions of the dynamical

system, the limiting member of the family being the position of equilibrium or
steady motion which was our starting-point.
187
189 to the
by applying the integration-process of
186, we obtain a solution of the Problem of
Three Bodies, when the motion is of the planetary type, in terms of trigono
Evidently
also,
equations of motion found in
metric series of the kind above specified.

For the further development of the theory of the present chapter, in connexion with
the Problem of Three Bodies, reference may be made to treatises on Celestial Mechanics
in particular, the second volume of Poincare s Nouvelles Methodes de la Mecanique Celeste
:
contains an account of several methods of deriving expansions, with a discussion of the

convergence of the series obtained. The most recent discussion of the subject will be found
by the present writer On the Adelphic Integral of

Nov. 1916).
in a paper
the Differential
Equations of
(Proc. Roy. Soc. Edin.,
Dynamics
denote any function of the variables
system which possesses an integral of energy
Let
1.
<f>
H (qi, ?
let
%, a 2
instant
...,
=t
tities qi,
bi, ..., b n
<?
2,
>
?n>
Pi,
>
q^,
g2
...,
qn
p) = Constant
1 ,
...,
pn
of a dynamical
a n 6 1} ..., b n be the values of q lt q 2 ..., qn ,pi,

iPn respectively at the
and let {/, g} denote the value of the Poisson-bracket (/, g] when the quan
a
occurring in it are replaced respectively by 1? 2
P\i
,
<lni
>Pn
>
>
Shew that
2.
Shew that the dynamical system whose equations
dq_dH
~
^ =
~
dt
^t
dp
H=
where
J
%p<+
W
^
of motion are
_^
dq
IW,
possesses a family of solutions represented by the expansion (retaining only terms of order
less
than a 7 )
,
and
3a
/2a\4
p=-k+ ~
where
and
3a
+ f,
INDEX OF AUTHORS QUOTED

(The numbers refer
to the
pages.}
Abdank-Abakanowicz, B. 214
Albeggiani, M. L. 337
Amontons, G. 227
Appell, P. 73, 258, 279
Clebsch, A. 311
Armellini, G. 81, 412
Curtis, A.
Bennett, T. L. 356
Bernoulli, Daniel 62, 177, 186
John 62, 229
Dainelli, V. 96, 112, 114
Bertrand,
J.
88,
260, 320, 331, 332,
349
Bessel, F.
W.
91
Bisconcini, G. 411
Block, H. 411
B6cher, M. 183
Bohlin, K. 340, 358, 407
Boltzmann, L. 41, 278

Bonacini, C. 408
Bonnet, 0. 94
Bour, E. 356
Brell, H. 259
Conway, A. W. 26
Cotes, R. 83
Culverwell, E. P. 251
H.
97, 113
D Alembert,
J. le R. 177, 230
Acqua, F. A. 316
Darboux, G. 80, 109, 261, 333, 396
Darwin, Sir G. H. 407
Dautheville, S. 320
Davaux, E. 400
Delaunay, C. 413, 419, 420
Derriman, W. H. 118
Donkin, W. F. 264
Dall
338,
Dumas, G. 166
E. B. 210
Ende, H. 112
Elliott,
am
Euler, L.
2, 8, 9, 41, 72, 93, 97, 100, 117,

124, 127, 144, 177, 248
Bromwich, T. J. 414
de Brun, F. 166
Brims, H. 357, 358
Ferrers, X.
Burgatti, P. 69, 166, 325

Burnside, W. 3, 5
Flye Sainte-Marie, C. 174

Forster, W. 262
Cailler, C.
89
W. R. 118
Cauchy, A. L. 4, 124, 264, 316
Cayley, A. 9, 12, 115
Cerruti, V. 330, 331
Charlier, C. V. L. 395
Cassie,
M. 215
Ford, L. R. 12
Forsyth, A. R. 358
Fouret, G. 130
Frank, P. 408
Galilei, G.
62, 72, 99, 177
Christoflfel, E.
Chretien, H. 90
B. 39
de Gasparis, A. 356
Gauss, C. F. 9, 255
Gautier, A. 339
Gebbia, M. 174
Cigala, A. R. 396
Clairaut, A. C. 78
Glaisher, J. W. L. 80
Gorjatscheff, D. 166
Chasles, M. 4
426
Legendre, A. M. 81
Lehmann-Filhes, R. 316
Leibnitz, G. W. von 35
Leitinger, R, 256
Goursat, E. 262, 336

Grant, R. 339
Green, G. 38
Grinwis, C. H. 89
330
Grossi, P.
Hadamard,
J.
Levi-Civita, T. 90, 325, 343, 385, 396, 411
69,
403
Halphen, G. 5, 106
Hamel, G. 41
Levy, M. 330, 331

Liapounoff, A. M. 400
Lie, S. 275, 290, 295, 301, 322,
Linders, F. J. 393
Hamilton, Sir W. R.
3, 9, 55, 79, 246, 264,
288, 290, 315, 316, 323, 409

J. N. 102
Helmholtz, H. von 45, 55, 247, 305
Hertz, H. 255
Lindstedt, A. 413
Liouville, J. 67, 281, 323
R. 167
Hazzidakis,
Heun, K. 37
Hill, G.
W.
343
Lipschitz, R. 256
Longley, W. R. 393
Lovett, E. 0. 339, 393, 395
407
Hiltebeitel, A.
M. 99
Hirsch, A. 45, 287

Holder, 0. 249
Hoppe, R. 130
Husson, E. 166
Huygens, C. 62,
72, 99, 117
MacMillan, W. D. 85
Marcolongo, R. 166
Mathieu, E. 301
Maupertuis, P. L. N. de 248
Mayer, A. 45
R. 77
Mehmke,
Monge, G. 264, 316

Jacobi, C. G. J. 104, 144, 174, 251, 276, 281,
287, 295, 307, 316, 341, 342, 349, 354
Jordan, C. 179
Moulton, F. R. 106, 390, 391, 393, 395

Muth, P. 183
Joukovsky, N. Ill
Nanson, E.
Kasner, E. 409
Neumann,
Newcomb,
Kelvin, Lord (W. Thomson) 261
Newton,
Kepler, J. 60, 90
1,
88,
413
S.
Sir
I.
27, 29, 30, 47, 48, 59, 62, 77,
78, 82, 83, 86, 90, 103,
Kerkhoven-Wythoff, A. G. 222
Klein, F. 12, 193, 207
Kobb, G. 109
Koenigs, G.
183
J.
C. 116, 215, 239
229
Nicomedi, R. 112
Nobile, V. 82
Novikoff, P. M. 407
275
Kolosoff, G. 167
Korkine, A. 337
Korteweg, D. 398, 400
Kotter, F. 166
Kowalevski, N. 166
S. 164
Oekinghaus, E. 91
Ogura, K. 409
Olsson, 0. 166
Ostrogradsky, M. 264, 265
Painleve,
P.
70,
227, 262, 379, 385, 389,
411
Lagrange,
J.
L. 34, 38, 41, 50, 62, 91, 94,
97, 104, 156, 177, 183, 248, 264, 298, 316,
322, 340, 393

Laisant, C. A. 113
Lamb,
II.
203, 305
Lambert, J. H. 91
Lame, G. 104
Larmor, Sir J. 278
Laurent, H. 337
P. A. 198
Lazzarino, 0. 166
Pascal, E. 214
Pavanini, G. 391
Pennacchietti, G. 338
Pfaff, J. F. 264, 296, 307, 316
di Pirro, G. 335
Poincare, H. 203, 267, 286, 343, 354, 380,

385, 387, 400, 403, 407, 410, 413, 424
Poinsot, L. 2, 152
Poisson, S. D. 163, 230, 264, 281, 299, 320,
407
Puiseux, V. 106
Index of Authors Quoted

Quanjel, J. 316
427
Sundman, K.
F. 411, 412
Sylvester, J. J. 184
Radau, R. 348
Rayleigh, Lord 230, 261
Resal, H. 115
Rodrigues, O.
Tait, P. G. 409
Taylor, Brook 177

Tchapligine, S. A. 166, 167
3,
Routh, E. J. 55
Rueb, A. S. 144
Thomson, W.,
Salkowski, E. 109
Scheffler, H. 255
Tonelli, L.
Schenkl, E. 256
Schoute, P. H. 85
Schouten, G. 408
Segner,
J.
see
Kelvin, Lord
Tisserand, F. 419
Tissot, A. 104
388
Vierkandt, A. 215
J. 23
Voss, A. 249
von Vieth,
A. 124
Siacci, F. 21, 24, 154, 174, 230,
325
Wallis, J. 48, 234
Signorini, A. 388
Wassmuth, A. 256
Sommerfeld, A. 193
Weber, W. 45
Weierstrass, K. 183, 197, 412
Whewell, W. 78
Stackel, P. 109, 166, 335

Stader, J. F. 81
Stekloff, V. 166
Stokes, Sir G. G. 271

Stromgren, E. 395
Sturm,
Suchar,
J.
I.
C. F.
80
396
Whittaker, E. T. 64, 339, 343, 388, 393, 407,

413, 424
Woronetz, P. 221, 343

Wren, Sir C. 48, 234
INDEX OF TERMS EMPLOYED

(The numbers refer
to
the pages,
where the term occurs for the
in the book or
is
Absolute integral-invariants, 271
Chasles theorem, 4
ChristofFel s symbol, 39
Classical integrals, 358
Admit, to
Law
of,
(a transformation),
time
defined.}
Acceleration, 14
Actio Agentis, 30
Action, Least, 248

Action and Reaction,
Adjoint systems, 287
first
29
Coefficient of friction, 227

of stability, 396
319
Collinear Lagrange s particles, 392

Collision, orbits of, 391
Angles, Eulerian, 9
Collisions,
Angular Momentum, Integral of, 59

Anomaly, true, eccentric, and mean, 89
Components of momentum, 48
Aphelion, 86
Apocentre, 85
Appell s equations, 258

Apse, 86
Attractive regions of a
of force, 403
field
Axes
of inertia, principal, 124

Axis of rotation, instantaneous, 2
234
of a vector, 14
Conjugate point, 252
Conservation of angular momentum, 59
of energy, 62
of momentum, 59
Conservative forces, 38
Constraint, 255
Contact-transformation, 290, 293
Azimuth, 19
homogeneous, 301
infinitesimal,
Bernoulli s principle, 186
Bertrand
theorem on
292,
302
determination
of
forces, 331
impulses, 260
Continuation, analytic, 410

Coordinates of a dynamical system, 32
elliptic, 97
Bilinear covariant, 297
ignorable or cyclic, 54
normal or principal, 181
Boltzmann-Larmor representation of the

Last Multiplier, 278
Bonnet s theorem, 94
Bracket-expressions, Lagrange s, 298
Poisson
s,
299
Brims theorem, 358
quasi-, 42
Cotes spirals, 83
Covariant, bilinear, 297
,,
deformation-, 111
Curvature, Least, 255
Cyclic coordinates, 54
Canonical form of equations of motion, 264
Cayley-Klein parameters, 12
Central forces, 77
Centre of rotation, instantaneous, 3
Definite quadratic form, 36
Centrifugal forces, 41
Characteristic exponents, 400
Density, 117
function, 289
Deformation-covariant, 111, 404

Degrees of freedom, 34
Differential form, 296
parameters, 111
429
Index of Terms Employed

Displacement,
Frictional systems, 227
possible, 33
Function, dissipation-, 231
Hamilton
Dissipation-function, 231
Dissipative systems, 226

Distance, mean, 87
s characteristic,
289
principal, 317
Hamiltonian, 265
Jacobi s, 342
Divisors, elementary, 183
Function-group, 322
Eccentric anomaly, 89
Ejection, orbits of, 391
Elementary
Gauss
Elimination of the nodes, 341

Ellipsoid of gyration, 124
of inertia, 124
momenta!, 124
Group, function-, 322

Group-property, 293
Gyration, ellipsoid of, 124
radius of, 118
Elliptic coordinates, 97
Energy, integral
of.
principle, 255
Geodesies, 254
Gravity, 27
divisors, 183
Gyroscopic terms, 195
62
Kinetic, 35
Hadamard s theorem, 406
Potential, 38
Equations, Appell
s,
258
Pfaffs system of, 307

Hamilton s, of motion, 263
first
partial
differential,
315
Jacobi
s,
342
Lagrangian, 37
in
quasi
coordi
nates, 43
with undetermined
multipliers, 213
Halphen s theorem,
Hamilton s partial
5
differential
equation,
315
,,
principle, 246
theorem, 79
Hamiltonian form of equations of motion,
264
function, 265
Herpolhode, 154
Hertz s principle, 255
Holonomic, 33
variational, 268
Equidistant Lagrange s particles, 393
Equilibrium-configuration, 177
Homogeneous contact-transformations, 301

Homography, 12
Equilibrium-problem, 315
Equimomental bodies, 117
Ignorable coordinates, 54
Eulerian angles, 9
Impact, 234
Impulse, 49
Impulsive motion. 48
Index of stability, 398
Inelastic bodies, 234
Inertia, ellipsoid of, 124
Exponents, characteristic, 400
Extended point-transformations, 293

External forces, 32, 37
Field of force, 30
Ignoration of coordinates, 56
moment
conservative, 38
parallel,
93
First PfaflPs system, 307

Fixity, 26
Fixtures, sudden, 169

Flux of a vector, 14
Focus, kinetic, 252

Force, 29
Forces, central, 77
arid
product
117
302
Initial motions,
45
Instability, 186, 193, 203, 396

Instantaneous axis of rotation, 2
centre of rotation, 3
Integral of a dynamical system, 53
momentum, 59
centrifugal, 41
of angular
external and molecular, 32, 37
358
of energy, 62
Jacobian, 354
Form, differential, 296

Freedom, degrees of, 34
of,
principal axes of, 124

Infinitesimal contact-transformations, 292,
classical,

Mean
Integral of momentum, 58
of a system of equations, 53
Integral-invariants, 268
absolute
and
distance, 87
motion, 88
Meridian plane, 18
Molecular forces, 32
Moment of a force, 30
,,
relative,
271
Invariable line and plane, 144, 346
of inertia, 117
,,
Invariant relations, 326

Invariants, integral, 268
Moinental
ellipsoid,
124
Momentum, 48
Inverse of a transformation, 293
angular, 59
Involution, 322
corresponding to a coordinate,
54
Isoperimetrical systems, 267
integral
Jacobi s equation, 342

function, 342
of,
58
Motion, impulsive, 48
initial, 45
Jacobian integral, 354

Joukovsky s theorem, 107
mean, 88
steady, 193
Multiplier, Last, 277
Kinematics, 1
Kinetic energy, 35
Natural family of
focus, 252
389
orbits,
systems, 57
,,
potential, 38
Kineto-statics, 37
Newton
Koenigs and Lie s theorem, 275

Korteweg s theorem, 397
Kowalevski s top, 164
Node, 349
Nodes, elimination of the, 341
Non-holonomic, 33
s theorem on revolving
Newtonian law, 86
Normal
Lagrange s bracket-expressions, 298
equations, 34
with
undetermined
multipliers, 213
of impulsive motion,
50
coordinates, 181
vibrations, 186, 195
Orbit, 78
periodic, 386
Order of an integral-invariant, 268
,,
forquasi-coordinates,
of a system, 52
Oscillation, centre of, 132
43
three particles, 393
Parallel fields of force, 93
Lagrangian function, 39
Lambert s theorem, 91
Larmor-Boltzmann
representation

Law of Action and Reaction, 29
Newtonian, 86
Least Action, 248
Curvature or Constraint, 255
Levi-Civita s theorem, 325
Levy s theorem, 330
Lie and Koenigs theorem, 275
Liouville s type, systems of, 67
Localised vectors, 15
Mass, 28
Mathieu transformations, 301

Mean anomaly, 89
Parameters, Cayley-Klein, 11
differential,
of
orbits,
the
Euler
s,
111
Particle, 27
Particles, Lagrange s, 393

Pendulum, simple, 72
spherical, 104
Perfect roughness, 31
Pericentre, 85
Perihelion, 86
Perihelion-constant, 87
Period, 73
Periodic solutions or orbits, 386
time, 87
Pfaff s expression, 296
,,
system of equations, 307
Pitch of a screw, 5
Plane, invariable, 346
83

Rodrigues and Hamilton
Rotation about a line, 1
Planetoid, 353
Poincare
Poinsot
normal variables, 387

theorem, 380
s representation,
of,
Rough, 31
Screw displacement, 5
Similarity in dynamical systems, 47
Sleeping top, 206
Smooth, 31
Spherical pendulum, 104
Kinetic, 38
,,
involving the velocities, 44
Principal axes of inertia, 124
coordinates, 181
top, 159
Spirals, Cotes , 83
Stability of equilibrium, 186

of orbits, 396, 407
of steady motion, 193
function, 317
moments of inertia, 124
Hamilton s, 246
396
398
secular, 203
Steady motion, 163, 193
Sub-group, 301
Sudden fixture, 169
coefficient of,
of Least Action, 248

of Least Curvature or Constraint,
255
,,
of,
centre
Possible displacement, 33
Potential energy, 38
,,
theorem, 3
point, 1
Polhode, 154
Principle,
instantaneous axis
152
Point-transformation, 293
Poisson s bracket-expressions, 299
stability, 407
theorem, 320
,,
431
of Relativity, 26
of Superposition of Vibrations,
186
of,
Superposition of vibrations, 186

Surface-density, 118
Problem of Three Bodies, 339

in a plane, 351
restricted,
index
353
Product of inertia, 116

Prolongation of motion after a collision, 411
Suspension, centre of, 132

Sylvester s theorem, 183
Symbol, Christoffel s, 39
of
an infinitesimal transformation,
303
Quadratures, problems soluble by, 54
Quantitas Motus, 48
Quasi-coordinates, 41
Quaternions, 9
Radius of gyration, 118

Rayleigh s dissipation-function, 230
Reaction, law of Action and, 29
Reciprocal theorem, Helmholtz s, 304
System, adjoint, 287

Systems, dissipative, 226
frictional, 227
involution-, 322
isoperimetrical, 267
Pfaff s, 307
Thomson s theorem, 261

Three Bodies, Problem
Reciprocation, 291
integral-invariants, 271
Relativity, principle of, 26
Repellent regions of field of force, 403
Resistance of air, 229
Restricted problem of three bodies, 353

Resultant of vectors, 14
Reversed
forces, 47
motion, 305
Revolving orbits, 83
Rigid, 1, 32
339
in a plane, 351
Regularisatiori, 411
Relations, invariant, 326

Relative velocity, 14
of,
restricted,
Time, 27
periodic, 87
Top, 155
Kowalevski
s,
164
sleeping, 206
spherical, 159
Trajectory, 78, 245
Transformation, contact-, 290, 293

Mathieu s, 301
Poincare s, 410
point-, 293
Translation,
353
432
Trojan group of asteroids, 393

True anomaly, 89
Velocity, corresponding to a coordinate, 33
Vibrations about equilibrium, 177

steady motion, 193
normal, 186, 195
Two
centres of gravitation, 97
Type, Liouville s, 67
of dissipative systems, 232

of non-holonomic systems, 221
Unstable, 186, 193, 203, 396
Virtual work, 264

Variational equations, 268
Vector, localised, 15
Vis Matrix, 29
Vis Viva, 35
Vectors, 13
Velocity, 14, 33
Wave-fronts, 289
Weber s law
angular, 15
relative, 14
CAMBRIDGE: PRINTED BY
of attraction, 45
Work, 30
j.
B.
PEACE, M.A., AT THE UNIVERSITY PRESS
KClUKN
TO
Astronomy Mathematics/Statistics Computer Science Library

1
00 Evans
PERIOD
4-M
ALL
642-3381
Hall
BOOKS MAY
Hr-i-Hi
BE RECALLED AFTER 7
DAYS
DUE AS STAMPED BELOW
I*?
^
GOT
1997
121996
APR121998
a
QCT
..
JUN
8 1998
is
UNIVERSITY OF CALIFORNIA, BERKELEY
FORM NO.
DD3,
/83
BERKELEY,
Berkeley
CA 94720

1917 - Treatise Analytical Dynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1917 - Treatise Analytical Dynamics

Uploaded by

Copyright:

Available Formats

A TREATISE

CAMBRIDGE UNIVERSITY PRESS

FETTER LANE, E.G.

All rights reserved

the University of Edinburgh

Second Edition 1917

PREFACE TO THE SECOND EDITION

and in some cases accounts

Dynamics which have been published by various investigators

matter, and rewritten

many sections. It is not necessary to specify these

125 sprang from a desire to do

opinion that the Jacobian functions are preferable to the Weierstrassian

should have liked

181), but thought

only such an account as might impel the reader to consult

and readable memoir.

obligations to the staff of the

and opposite rotations about parallel axes

Analytic representation of a displacement

Euler s parametric specification of rotations round a point

Velocity and acceleration

Angular velocity its vectorial character

their vectorial character

THE EQUATIONS OF MOTION.

Lagrange s form of the equations of motion of a holonomic system

The Lagrangian equations

explicit form of Lagrange s equations

the kinetic potential

Similarity in dynamical systems

PRINCIPLES AVAILABLE FOR THE INTEGRATION.

Problems which are soluble by quadratures

Special cases of ignoration

The general theorem

Reduction of a dynamical problem to a problem with fewer degrees of

THE SOLUBLE PROBLEMS OF PARTICLE DYNAMICS.

The particle with one degree of freedom the pendulum

Motion under the Newtonian law

integrable cases of central forces

problems soluble in terms of circular

two centres of gravitation

cases soluble in terms of circular and

THE DYNAMICAL SPECIFICATION OF BODIES.

about a parallel axis through the centre of gravity is known

some simple bodies

when the moment

through the same origin

principal axes of inertia

THE SOLUBLE PROBLEMS OF RIGID DYNAMICS.

of systems with one degree of freedom

motion round a fixed

Motion of a top on a perfectly rough plane

Determination of the remaining Eulerian angles, and of the Cayley-Klein

two degrees of freedom

determination of the Eulerian

Vibrations about equilibrium

Sylvester s theorem on the reality of the roots of the determinantal equation

Examples of vibrations about equilibrium

constraint on the periods of a vibrating system

integration of the equations

Vibrations of systems involving moving constraints

NON-HOLONOMIC SYSTEMS. DISSIPATIVE SYSTEMS.

s equations with undetermined multipliers

Loss of kinetic energy in impact