gale 'Bicentennial publication?
ELEMENTARY PRINCIPLES IN
STATISTICAL MECHANICS
pale bicentennial publications
With the approval if tbt Prindent and FcUmn
of Tali Unrveriity, a stria of volumes has keen
prepared by a number of the Pnfesson and In
structorsj to be issued in connection with the
Bicentennial Anniversary^ as a partial indica
ttm of the character of the studies in wbicb the
University teachers are engaged.
This series of volumes is respectfully dedicated 4»
ff)r ^nurtures of tljr
ELEMENTARY PRINCIPLES
IN
STATISTICAL MECHANICS
DEVELOPED WITH ESPECIAL REFERENCE TO
THE RATIONAL FOUNDATION OF
THERMODYNAMICS
BY
J. WILLARD GIBBS
Proftuor of Matktmatual Pkyrict in YaU University
OF r
UNIVERSITY
OF
NEW YORK : CHARLES SCRIBNER'S SONS
LONDON: EDWARD ARNOLD
1902
A<>
'
Copyright, 1902,
BY CHARLES SCRIBNER'S SONS
Published, March, zgoz.
UNIVERSITY PRESS • JOHN WILSON
AND SON • CAMBRIDGE, U.S.A.
PREFACE.
THE usual point of view in the study of mechanics is that
where the attention is mainly directed to the changes which
take place in the course of time in a given system. The prin
cipal problem is the determination of the condition of the
system with respect to. configuration and velocities at any
required time, when its condition in these respects has been
given for some one time, and the fundamental equations are
those which express the changes continually taking place in
the system. Inquiries of this kind are often simplified by
taking into consideration conditions of the system other than
those through which it actually passes or is supposed to pass,
but our attention is not usually carried beyond conditions
differing infinitesimally from those which are regarded as
actual.
For some purposes, however, it is desirable to take a broader
view of the subject. We may imagine a great number of
systems of the same nature, but differing in the configura
tions and velocities which they have at a given instant, and
differing not merely infinitesimally, but it may be so as to
embrace every conceivable combination of configuration and
velocities. And here we may set the problem, not to follow
a particular system through its succession of configurations,
but to determine how the whole number of systems will be
distributed among the various conceivable configurations and
velocities at any required time, when the distribution has
been given for some one time. The fundamental equation
for this inquiry is that which gives the rate of change of the
number of systems which fall within any infinitesimal limits
of configuration and velocity.
94203
viii PREFACE.
Such inquiries have been called by Maxwell statistical.
They belong to a branch of mechanics which owes its origin to
the desire to' explain the laws of thermodynamics on mechan
ical principles, and of which Clausius, Maxwell, and Boltz
mann are to be regarded as the principal founders. The first
inquiries in this field were indeed somewhat narrower in their
scope than that which has been mentioned, being applied to
the particles of a system, rather than to independent systems.
Statistical inquiries were next directed to the phases (or con
ditions with respect to configuration and velocity) which
succeed one another in a given system in the course of time.
The explicit consideration of a great number of systems and
their distribution in phase, and of the permanence or alteration
of this distribution in the course of time is perhaps first found
in Boltzmann's paper on the " Zusammenhang zwischen den
Satzen iiber das Verhalten mehratomiger Gasmolekiile mit
Jacobi's Princip des letzten Multiplicators " (1871).
But although, as a matter of history, statistical mechanics
owes its origin to investigations in thermodynamics, it seems
eminently worthy of an independent development, both on
account of the elegance and simplicity of its principles, and
because it yields new results and places old truths in a new
light in departments quite outside of thermodynamics. More
over, the separate study of this branch of mechanics seems to
afford the best foundation for the study of rational thermody
namics and molecular mechanics.
The laws of thermodynamics, as empirically determined,
express the approximate and probable behavior of systems of
a great number of particles, or, more precisely, they express
the laws of mechanics for such systems as they appear to
beings who have not the fineness of perception to enable
them to appreciate quantities of the order of magnitude of
those which relate to single particles, and who cannot repeat
their experiments often enough to obtain any but the most
probable results. The laws of statistical mechanics apply to
conservative systems of any number of degrees of freedom,
PREFACE. ix
and are exact. This does not make them more difficult to
establish than the approximate laws for systems of a great
many degrees of freedom, or for limited classes of such
systems. The reverse is rather the case, for our attention is
not diverted from what is essential by the peculiarities of the
system considered, and we are not obliged to satisfy ourselves
that the effect of the quantities and circumstances neglected
will be negligible in the result. The laws of thermodynamics
may be easily obtained from the principles of statistical me
chanics, of which they are the incomplete expression, but
they make a somewhat blind guide in our search for those
laws. This is perhaps the principal cause of the slow progress
of rational thermodynamics, as contrasted with the rapid de
duction of the consequences of its laws as empirically estab
lished. To this must be added that the rational foundation
of thermodynamics lay in a branch of mechanics of which
the fundamental notions and principles, and the characteristic
operations, were alike unfamiliar to students of mechanics.
We may therefore confidently believe that nothing will
more conduce to the clear apprehension of the relation of
thermodynamics to rational mechanics, and to the interpreta
tion of observed phenomena with reference to their evidence
respecting the molecular constitution of bodies, than the
study of the fundamental notions and principles of that de
partment of mechanics to which thermodynamics is especially
related.
Moreover, we avoid the gravest difficulties when, giving up
the attempt to frame hypotheses concerning the constitution
of material bodies, we pursue statistical inquiries as a branch
of rational mechanics. In the present state of science, it
seems hardly possible to frame a dynamic theory of molecular
action which shall embrace the phenomena of thermody
namics, of radiation, and of the electrical manifestations
which accompany the union of atoms. Yet any theory is
obviously inadequate which does not take account of all
these phenomena. Even if we confine cur attention to the
X PREFACE.
phenomena distinctively thermodynamic, we do not escape
difficulties in as simple a matter as the number of degrees
of freedom of a diatomic gas. It is well known that while
theory would assign to the gas six degrees of freedom per
molecule, in our experiments on specific heat we cannot ac
count for more than five. Certainly, one is building on an
insecure foundation, who rests his work on hypotheses con
cerning the constitution of matter.
Difficulties of this kind have deterred the author from at
tempting to explain the mysteries of nature, and have forced
him to be contented with the more modest aim of deducing
some of the more obvious propositions relating to the statis
tical branch of mechanics. Here, there can be no mistake in
regard to the agreement of the hypotheses with the facts of
nature, for nothing is assumed in that respect. The only
error into which one can fall, is the want of agreement be
tween the premises and the conclusions, and this, with care,
one may hope, in the main, to avoid.
The matter of the present volume consists in large measure
of results which have been obtained by the investigators
mentioned above, although the point of view and the arrange
ment may be different. These results, given to the public
one by one in the order of their discovery, have necessarily,
in their original presentation, not been arranged in the most
logical manner.
In the first chapter we consider the general problem which
has been mentioned, and find what may be called the funda
mental equation of statistical mechanics. A particular case
of this equation will give the condition of statistical equi
librium, i. e., the condition which the distribution of the
systems in phase must satisfy in order that the distribution
shall be permanent. In the general case, the fundamental
equation admits an integration, which gives a principle which
may be variously expressed, according to the point of view
from which it is regarded, as the conservation of densityin
phase, or of extensioninphase, or of probability of phase.
PREFACE. xi
In the second chapter, we apply this principle of conserva
tion of probability of phase to the theory of errors in the
calculated phases of a system, when the determination of the
arbitrary constants of the integral equations are subject to
error. In this application, we do not go beyond the usual
approximations. In other words, we combine the principle
of conservation of probability of phase, which is exact, with
those approximate relations, which it is customary to assume
in the " theory of errors."
In the third chapter we apply the principle of conservation
of extensioninphase to the integration of the differential
equations of motion. This gives Jacobi's " last multiplier,"
as has been shown by Boltzmann.
In the fourth and following chapters we return to the con
sideration of statistical equilibrium, and confine our attention
to conservative systems. We consider especially ensembles
of systems in which the index (or logarithm) of probability of
phase is a linear function of the energy. This distribution,
on account of its unique importance in the theory of statisti
cal equilibrium, I have ventured to call canonical, and the
divisor of the energy, the modulus of distribution. The
moduli of ensembles have properties analogous to temperature,
in that equality of the moduli is a condition of equilibrium
with respect to exchange of energy, when such exchange is
made possible.
We find a differential equation relating to average values
in the ensemble which is identical in form with the funda
mental differential equation of thermodynamics, the average
index of probability of phase, with change of sign, correspond
ing to entropy, and the modulus to temperature.
For the average square of the anomalies of the energy, we
find an expression which vanishes in comparison with the
square of the average energy, when the number of degrees
of freedom is indefinitely increased. An ensemble of systems
in which the number of degrees of freedom is of the same
order of magnitude as the number of molecules in the bodies
xii PREFACE.
with which we experiment, if distributed canonically, would
therefore appear to human observation as an ensemble of
systems in which all have the same energy.
We meet with other quantities, in the development of the
subject, which, when the number of degrees of freedom is
very great, coincide sensibly with the modulus, and with the
average index of probability, taken negatively, in a canonical
ensemble, and which, therefore, may also be regarded as cor
responding to temperature and entropy. The correspondence
is however imperfect, when the number of degrees of freedom
is not very great, and there is nothing to recommend these
quantities except that in definition they may be regarded as
more simple than those which have been mentioned. In
Chapter XIV, this subject of thermodynamic analogies is
discussed somewhat at length.
Finally, in Chapter XV, we consider the modification of
the preceding results which is necessary when we consider
systems composed of a number of entirely similar particles,
or, it may be, of a number of particles of several kinds, all of
each kind being entirely similar to each other, and when one
of the variations to be considered is that of the numbers of
the particles of the various kinds which are contained in a
system. This supposition would naturally have been intro
duced earlier, if our object had been simply the expression of
the laws of nature. It seemed desirable, however, to separate
sharply the purely thermodynamic laws from those special
modifications which belong rather to the theoiy of the prop
erties of matter.
J. W. G.
NEW HAVEN, December, 1901.
CONTENTS.
CHAPTER I.
GENERAL NOTIONS. THE PRINCIPLE OF CONSERVATION
OF EXTENSIONINPHASE.
PAGE
Hamilton's equations of motion 35
Ensemble of systems distributed in phase 5
Extensioninphase, densityinphase 6
Fundamental equation of statistical mechanics 68
Condition of statistical equilibrium 8
Principle of conservation of densityinphase 9
Principle of conservation of extensioninphase 10
Analogy in hydrodynamics 11
Extensioninphase is an invariant 1113
Dimensions of extensioninphase 13
Various analytical expressions of the principle 1315
Coefficient and index of probability of phase 16
Principle of conservation of probability of phase 17, 18
Dimensions of coefficient of probability of phase 19
CHAPTER II.
APPLICATION OF THE PRINCIPLE OF CONSERVATION OF
EXTENSIONINPHASE TO THE THEORY OF ERRORS.
Approximate expression for the index of probability of phase . 20, 21
Application of the principle of conservation of probability of phase
to the constants of this expression 2125
CHAPTER III.
APPLICATION OF THE PRINCIPLE OF CONSERVATION OF
EXTENSIONINPHASE TO THE INTEGRATION OF THE
DIFFERENTIAL EQUATIONS OF MOTION.
Case in which the forces are function of the coordinates alone . 2629
Case in which the forces are functions of the coordinates with the
time 30, 31
xiv CONTENTS.
CHAPTER IV.
ON THE DISTRIBUTIONINPHASE CALLED CANONICAL, IN
WHICH THE INDEX OF PROBABILITY IS A LINEAR
FUNCTION OF THE ENERGY.
PAGE
Condition of statistical equilibrium 32
Other conditions which the coefficient of probability must satisfy . 33
"""" Canonical distribution — Modulus of distribution 34
^ must be finite 35
The modulus of the canonical distribution has properties analogous
to temperature 3537
Other distributions have similar properties 37
Distribution in which the index of probability is a linear function of
the energy and of the moments of momentum about three axes . 38, 39
Case in which the forces are linear functions of the displacements,
and the index is a. linear function of the separate energies relating
to the normal types of motion 3941
Differential equation relating to average values in a canonical
ensemble 4244
This is identical in form with the fundamental differential equation
of thermodynamics 44, 45
CHAPTER V.
AVERAGE VALUES IN A CANONICAL ENSEMBLE OF SYS
TEMS.
Case of v material points. Average value of kinetic energy of a
single point for a given configuration or for the whole ensemble
= f 0 46, 47
Average value of total kinetic energy for any given configuration
or for the whole ensemble = % v 0 47
System of n degrees of freedom. Average value of kinetic energy,
for any given configuration or for the whole ensemble = f 0 . 4850
Second proof of the same proposition 5052
Distribution of canonical ensemble in configuration 5254
Ensembles canonically distributed in configuration 55
Ensembles canonically distributed in velocity 56
CHAPTER VI.
EXTENSION1INCONFIGURATION AND EXTENSIONTN
VELOCITY.
Extensioninconfiguration and extensioninvelocity are invari
ants . 5759
CONTENTS. XV
PAGE
Dimensions of these quantities 60
Index and coefficient of probability of configuration 61
Index and coefficient of probability of velocity 62
Dimensions of these coefficients 63
Relation between extensioninconfiguration and extensioninvelocity 64
Definitions of extensioninphase, extensioninconfiguration, and ex
tensionin velocity, without explicit mention of coordinates . . 6567
CHAPTER VII.
FARTHER DISCUSSION OF AVERAGES IN A CANONICAL
ENSEMBLE OF SYSTEMS.
Second and third differential equations relating to average values
in a canonical ensemble 68, 69
These are identical in form with thermodynamic equations enun
ciated by Clausius 69
Average square of the anomaly of the energy — of the kinetic en
ergy— of the potential energy 7072
These anomalies are insensible to human observation and experi
ence when the number of degrees of freedom of the system is very
great 73, 74
Average values of powers of the energies 7577
Average values of powers of the anomalies of the energies . . 7780
Average values relating to forces exerted on external bodies . . 8083
General formulae relating to averages in a canonical ensemble . 8386
CHAPTER VIII.
ON CERTAIN IMPORTANT FUNCTIONS OF THE ENERGIES
OF A SYSTEM.
Definitions. V = extensioninphase below a limiting energy (e).
$ = \o«dVldc 87,88
Vq = extensioninconfiguration below a limiting value of the poten
tial energy (e?). fa = \o^dVqjdfq 89,90
Vp = extensioninvelocity below a limiting value of the kinetic energy
(*). ^p = loSdVpjd€p 90,91
Evaluation of Vp and $p 9193
Average values of functions of the kinetic energy 94, 95
Calculation of FfromF^ 95,96
Approximate formulae for large values of n 97,98
Calculation of V or <£ for whole system when given for parts ... 98
Geometrical illustration . 99
xvi CONTENTS.
CHAPTER IX.
THE FUNCTION AND THE CANONICAL DISTRIBUTION.
When n > 2, the most probable value of the energy in a canonical
ensemble is determined by d(j> j de = 1 / e 100,101
When n > 2, the average value of d$ j de in a canonical ensemble
isl/e 101
When n is large, the value of <£ corresponding to d(f>/de=l/Q
(<£o) js nearly equivalent (except for an additive constant) to
the average index of probability taken negatively (— fj) . . 101104
Approximate formulae for <£0 + fj when n is large 104106
When n is large, the distribution of a canonical ensemble in energy
follows approximately the law of errors 105
This is not peculiar to the canonical distribution 107, 108
Averages in a canonical ensemble 108114
CHAPTER X.
ON A DISTRIBUTION IN PHASE CALLED MICROCANONI
CAL IN WHICH ALL THE SYSTEMS HAVE THE SAME
ENERGY.
The microcanonical distribution denned as the limiting distribution
obtained by various processes 115, 116
Average values in the microcanonical ensemble of functions of the
kinetic and potential energies 117120
If two quantities have the same average values in every microcanon
ical ensemble, they have the same average value in every canon
ical ensemble 120
Average values in the microcanonical ensemble of functions of the
energies of parts of the system 121123
Average values of functions of the kinetic energy of a part of the
system 123, 124
Average values of the external forces in a microcanonical ensemble.
Differential equation relating to these averages, having the form
of the fundamental differential equation of thermodynamics . 124128
CHAPTER XI.
MAXIMUM AND MINIMUM PROPERTIES OF VARIOUS DIS
TRIBUTIONS IN PHASE.
Theorems I VI. Minimum properties of certain distributions . 129133
Theorem VII. The average index of the whole system compared
with the sum of the average indices of the parts 133135
CONTENTS. xvii
PAGE
Theorem VIII. The average index of the whole ensemble com
pared with the average indices of parts of the ensemble . . 135137
Theorem IX. Effect on the average index of making the distribu
tioninphase uniform within any limits 137138
CHAPTER XII.
ON THE MOTION OF SYSTEMS AND ENSEMBLES OF SYS
TEMS THROUGH LONG PERIODS OF TIME.
Under what conditions, and with what limitations, may we assume
that a system will return in the course of time to its original
phase, at least to any required degree of approximation? . . 139142
Tendency in an ensemble of isolated systems toward a state of sta
tistical equilibrium 143151
CHAPTER XIII.
EFFECT OF VARIOUS PROCESSES ON AN ENSEMBLE OF
SYSTEMS.
Variation of the external coordinates can only cause a decrease in
the average index of probability 152154
This decrease may in general be diminished by diminishing the
rapidity of the change in the external coordinates .... 154157
The mutual action of two ensembles can only diminish the sum of
their average indices of probability 158, 159
In the mutual action of two ensembles which are canonically dis
tributed, that which has the greater modulus will lose energy . 160
Repeated action between any ensemble and others which are canon
ically distributed with the same modulus will tend to distribute
the firstmentioned ensemble canonically with the same modulus 161
Process analogous to a Carnot's cycle 162,163
Analogous processes in thermodynamics 163, 164
CHAPTER XIV.
DISCUSSION OF THERMODYNAMIC ANALOGIES.
The finding in rational mechanics an a priori foundation forthermo
dynamics requires mechanical definitions of temperature and
entropy. Conditions which the quantities thus defined must
satisfy 165167
The modulus of a canonical ensemble (0), and the average index of
probability taken negatively (rj), as analogues of temperature
and entropy 167169
xviii CONTENTS.
PAGE
The functions of the energy del d log Fand log Fas analogues of
temperature and entropy 169172
The functions of the energy de / cty and ) . . . . 201
Average value of (vv)* 201,202
Comparison of indices 203206
When the number of particles in a system is to be treated as
variable, the average index of probability for phases generically
defined corresponds to entropy 206
ELEMENTARY PRINCIPLES IN
STATISTICAL MECHANICS
(( UNIVERSITY J
ELEMENTARY PRINCIPLES IN
STATISTICAL MECHANICS
CHAPTER I.
GENERAL NOTIONS. THE PRINCIPLE OF
OF EXTENSIONINPHASE.
WE shall use Hamilton's form of the equations of motion for
a system of n degrees of freedom, writing ql , . . ,qn for the
(generalized) coordinates, qi , . . . qn for the (generalized) ve
locities, and
for the moment of the forces. We shall call the quantities
Fl9...Fn the (generalized) forces, and the quantities p1 . . . pn,
defined by the equations
Pl = ^t p2 = ^, etc., (2)
dqi dq2
where ep denotes the kinetic energy of the system, the (gen
eralized) momenta. The kinetic energy is here regarded as
a function of the velocities and coordinates. We shall usually
regard it as a function of the momenta and coordinates,*
and on this account we denote it by ep. This will not pre
vent us from occasionally using formulae like (2), where it is
sufficiently evident the kinetic energy is regarded as function
of the g's and ^'s. But in expressions like dep/dq1 , where the
denominator does not determine the question, the kinetic
* The use of the momenta instead of the velocities as independent variables
is the characteristic of Hamilton's method which gives his equations of motion
their remarkable degree of simplicity. We shall find that the fundamental
notions of statistical mechanics are most easily defined, and are expressed in
the most simple form, when the momenta with the coordinates are used to
describe the state of a system.
4 HAMILTON'S EQUATIONS.
energy is always to be treated in the differentiation as function
of the p's and q*s.
We have then
* = ;fe* *l = ^ + Fl' etc> (3)
These equations will hold for any forces whatever. If the
'fetces^ &i*e £ dptt§erVative, in other words, if the expression (1)
j.stant exact differential, we may set
where eq is a function of the coordinates which we shall call
the potential energy of the system. If we write e for the
total energy, we shall have
e = €P + e«> (5)
and equatipns (3) may be written
*' = ;£' * = £' etc [I <«>
The potential energy (e3) may depend on other variables
beside the coordinates q1 . . . qn. We shall often suppose it to
depend in part on coordinates of external bodies, which we
shall denote by ax , #2 , etc. We shall then have for the com
plete value of the differential of the potential energy *
deq = — FI dql . . — Fn dqn — A1 da^ — A2 daz — etc., (7)
where A^ A%, etc., represent forces (in the generalized sense)
exerted by the system on external bodies. For the total energy
(e) we shall have
de=qldpl . . . + qndpn~Pidqi . . .
— pn dqn — Al dai — A2 daz — etc. (8)
It will be observed that the kinetic energy (e^,) in the
most general case is a quadratic function of the p's (or g's)
* It will be observed, that although we call e the potential energy of the
system which we are considering, it is really so defined as to include that
energy which might be described as mutual to that system and external
bodies.
ENSEMBLE OF SYSTEMS. 5
v
involving also the ^'s but not the a's ; that the potential energy,
when it exists, is function of the 2'.
A system which at that time has that phase will at another
time have another phase. Let the density as calculated for
this second time and phase by a third system of coordinates
be Zy. Now we may imagine a system of coordinates which
at and near the first configuration will coincide with the first
system of coordinates, and at and near the second configuration
will coincide with the third system of coordinates. This will
give Dj' — ^Y' Again we may imagine a system of coordi
nates which at and near the first configuration will coincide
with the second system of coordinates, and at and near the
* If we regard a phase as represented by a point in space of 2 n dimen
sions, the changes which take place in the course of time in our ensemble of
systems will be represented by a current in such space. This current will
be steady so long as the external coordinates are not varied. In any case
the current will satisfy a law which in its various expressions is analogous
to the hydrodynamic law which may be expressed by the phrases conserva
tion of volumes or conservation of density about a moving point, or by the equation
The analogue in statistical mechanics of this equation, viz.,
may be derived directly from equations (3) or (6), and may suggest such
theorems as have been enunciated, if indeed it is not regarded as making
them intuitively evident. The somewhat lengthy demonstrations given
above will at least serve to give precision to the notions involved, and
familiarity with their use.
12 EXTENSIONINPHASE
second configuration will coincide with the third system of
coordinates. This will give D% = Ds". We have therefore
2V = 2>J.
It follows, or it may be proved in the same way, that the
value of an extensioninphase is independent of the system
of coordinates which is used in its evaluation. This may
easily be verified directly. If g1^ . . ,qn^ Qlt . . . Qn are two
systems of coordinates, and Pi, • • • pn> P\i •  • Pn the cor
responding momenta, we have to prove that
J'...Jdp1...dpndqi...dqn=j*...fdPl...dPndQ1...dQn,(2£)
when the multiple integrals are taken within limits consisting
of the same phases. And this will be evident from the prin
ciple on which we change the variables in a multiple integral,
if we prove that
. . P., ft, . . . ft) = 1
>Pn>2i, •• • 2V)
where the first member of the equation represents a Jacobian
or functional determinant. Since all its elements of the form
dQ/dp are equal to zero, the determinant reduces to a product
of two, and we have to prove that
d(Ql9
We may transform any element of the first of these deter
minants as follows. By equations (2) and (3), and in
view of the fact that the (j's are linear functions of the !7 . . . dq^ (32)
14 CONSERVATION OF
the limiting phases being those which belong to the same
systems at the times t and If respectively. But we have
identically
/.../*,..., ,/..
for such limits. The principle of conservation of extensionin
phase may therefore be expressed in the form
• • g«) , xooN
..g.9 = 1'
This equation is easily proved directly. For we have
identically
d(Pl,...qn) _ d(Pl,...qn)
• • • g.'O <*(M • • • g.O '
where the double accents distinguish the values of the momenta
and coordinates for a time if'. If we vary t, while if and t"
remain constant, we have
d_ d(Pl, ...qn) _ d(Pl"9 . . . qn") d_ d(Pl, ...qn)
Now since the time if' is entirely arbitrary, nothing prevents
us from making if1 identical with t at the moment considered.
Then the determinant
• •  ?»")
will have unity for each of the elements on the principal
diagonal, and zero for all the other elements. Since every
term of the determinant except the product of the elements
on the principal diagonal will have two zero factors, the differen
tial of the determinant will reduce to that of the product of
these elements, i. e., to the sum of the differentials of these
elements. This gives the equation
d
_.
dt d(pj>, . . . qn») dp," ' dpn" dqj* ' dqn»
Now since t = t" , the double accents in the second member
of this equation may evidently be neglected. This will give,
in virtue of such relations as (16),
EXTENSIONINPHASE. 15
d d(plt ...
dtd(Pl»,...yn")
which substituted in (34) will give
d
_

...n _
dtd(Pl',...qn')
The determinant in this equation is therefore a constant, the
value of which may be determined at the instant when t = £',
when it is evidently unity. Equation (33) is therefore
demonstrated.
Again, if we write a, ... h for a system of 2 n arbitrary con
stants of the integral equations of motion, pv qv etc. will be
functions of. a, ... h, and t, and we may express an extension
inphase in the form
/rd(p
"V «*(<
,, ^T da   • dh (35>
d(a, ...h)
If we suppose the limits specified by values of a, . . . ^, a
system initially at the limits will remain at the limits.
The principle of conservation of extensioninphase requires
that an extension thus bounded shall have a constant value.
This requires that the determinant under the integral sign
shall be constant, which may be written
...n
dt d(a,...h) =°* (36)
This equation, which may be regarded as expressing the prin
ciple of conservation of extensioninphase, may be derived
directly from the identity
• • gj <*(pi, ...gn) d(pi', . . . qnr)
d(a, ...h) ' d(plf, . . . qn') d(a, ... h)
in connection with equation (33).
Since the coordinates and momenta are functions of a, ... . h,
and t, the determinant in (36) must be a function of the same
variables, and since it does not vary with the time, it must
be a function of a, ... h alone. We have therefore
„...*). ' (37)
16 CONSERVATION OF
It is the relative numbers of systems which fall within dif
ferent limits, rather than the absolute numbers, with which we
are most concerned. It is indeed only with regard to relative
numbers that such discussions as the preceding will apply
with literal precision, since the nature of our reasoning implies
that the number of systems in the smallest element of space
which we consider is very great. This is evidently inconsist
ent with a finite value of the total number of systems, or of
the densityinphase. Now if the value of D is infinite, we
cannot speak of any definite number of systems within any
finite limits, since all such numbers are infinite. But the
ratios of these infinite numbers may be perfectly definite. If
we write ZVfor the total number of systems, and set
r = %. (38)
P may remain finite, when JV* and D become infinite. The
integral
" * ... dqn (39)
taken within any given limits, will evidently express the ratio
of the number of systems falling within those limits to the
whole number of systems. This is the same thing as the
probability that an unspecified system of the ensemble (i. e.
one of which we only know that it belongs to the ensemble)
will lie within the given limits. The product
PdPl...dqn (40)
expresses the probability that an unspecified system of the
ensemble will be found in the element of extensioninphase
dpi . . . dqn. We shall call P the coefficient of probability of the
phase considered. Its natural logarithm we shall call the
index of probability of the phase, and denote it by the letter 77.
If we substitute NP and Ne1 for D in equation (19), we get
and
PROBABILITY OF PHASE. 17
The condition of statistical equilibrium may be expressed
by equating to zero the second member of either of these
equations.
The same substitutions in (22) give
.,=°' (43)
(IX.... =° (44)
That is, the values of P and rj, like those of D, are constant
in time for moving systems of the ensemble. From this point
of view, the principle which otherwise regarded has been
called the principle of conservation of densityinphase or
conservation of extensioninphase, may be called the prin
ciple of conservation of the coefficient (or index) of proba
bility of a phase varying according to dynamical laws, or
more briefly, the principle of conservation of probability of
phase. It is subject to the limitation that the forces must be
functions of the coordinates of the system either alone or with
the time.
The application of this principle is not limited to cases in
which there is a formal and explicit reference to an ensemble of
systems. Yet the conception of such an ensemble may serve
to give precision to notions of probability. It is in fact cus
tomary in the discussion of probabilities to describe anything
which is imperfectly known as something taken at random
from a great number of things which are completely described.
But if we prefer to avoid any reference to an ensemble
of systems, we may observe that the probability that the
phase of a system falls within certain limits at a certain time,
is equal to the probability that at some other time the phase
will fall within the limits formed by phases corresponding to
the first. For either occurrence necessitates the other. That
is, if we write P' for the coefficient of probability of the
phase pi, • • • qn' at the time ^, and P" for that of the phase
jp/', . . . qn" at the time tf',
2
18 CONSERVATION OF
J. . . JV dtf . . . dqj =f. . . Jp" dp{' . . . dqn", (45)
where the limits in the two cases are formed by corresponding
phases. When the integrations cover infinitely small vari
ations of the momenta and coordinates, we may regard P* and
P" as constant in the integrations and write
P'f. . .fdPl> • • • <%»" =
Now the principle of the conservation of extensioninphase,
which has been proved (viz., in the second demonstration given
above) independently of any reference to an ensemble of
systems, requires that the values of the multiple integrals in
this equation shall be equal. This gives
P1' = Pf.
With reference to an important class of cases this principle
may be enunciated as follows.
When the differential equations of motion are exactly known,
but the constants of the integral equations imperfectly deter
mined, the coefficient of probability of any phase at any time is
equal to the coefficient of probability of the corresponding phase
at any other time. By corresponding phases are meant those
which are calculated for different times from the same values
of the arbitrary constants of the integral equations.
Since the sum of the probabilities of all possible cases is
necessarily unity, it is evident that we must have
all
f...fpdPl...dqn = l, (46)
phases
where the integration extends over all phases. This is indeed
only a different form of the equation
811
phases
which we may regard as defining
PROBABILITY OF PHASE. 19
The values of the coefficient and index of probability of
phase, like that of the densityinphase, are independent of the
system of coordinates which is employed to express the distri
bution in phase of a given ensemble.
In dimensions, the coefficient of probability is the reciprocal
of an extensioninphase, that is, the reciprocal of the nth
power of the product of time and energy. The index of prob
ability is therefore affected by an additive constant when we
change our units of time and energy. If the unit of time is
multiplied by ct and the unit of energy is multiplied by ce , all
indices of probability relating to systems of n degrees of
freedom will be increased by the addition of
•"
n log ct + n log c€. (47)
CHAPTER II.
APPLICATION OF THE PRINCIPLE OF CONSERVATION
OF EXTENSIONINPHASE TO THE THEORY
OF ERRORS.
LET us now proceed to combine the principle which has been
demonstrated in the preceding chapter and which in its differ
ent applications and regarded from different points of view
has been variously designated as the conservation of density
inphase, or of extensioninphase, or of probability of phase,
with those approximate relations which are generally used in
the 'theory of errors.'
We suppose that the differential equations of the motion of
a system are exactly known, but that the constants of the
integral equations are only approximately determined. It is
evident that the probability that the momenta and coordinates
at the time t' fall between the limits pj and pj + dp^ q^ and
qL + dq^ etc., may be expressed by the formula
e* dPl' . . . dqj, (48)
where rf (the index of probability for the phase in question) is
a function of the coordinates and momenta and of the time.
Let Qi, P^t etc. be the values of the coordinates and momenta
which give the maximum value to ?/, and let the general
value of rj be developed by Taylor's theorem according to
ascending powers and products of the differences p^ — P/,
Q.I ~ Ci'» Qte"> an(i let us suppose that we have a sufficient
approximation without going beyond terms of the second
degree in these differences. We may therefore set
n' = c — F', (49)
where c is independent of the differences p^ — P/, q{ — §/,
etc., and F1 is a homogeneous quadratic function of these
THEORY OF ERRORS. 21
differences. The terms of the first degree vanish in virtue
of the maximum condition, which also requires that F' must
have a positive value except when all the differences men
tioned vanish. If we set
0=ef, (50)
we may write for the probability that the phase lies within
the limits considered
dPl> . . . dqj. (51)
C is evidently the maximum value of the coefficient of proba
bility at the time considered.
In regard to the degree of approximation represented by
these formulae, it is to be observed that we suppose, as is
usual in the 'theory of errors/ that the determination (ex
plicit or implicit) of the constants of motion is of such
precision that the coefficient of probability e* or Ce~F' is
practically zero except for very small values of the differences
Pi — P1/, q^ — Ci'> e^c< For very small values of these
differences the approximation is evidently in general sufficient,
for larger values of these differences the value of Ce~F' will
be sensibly zero, as it should be, and in this sense the formula
will represent the facts.
We shall suppose that the forces to which the system is
subject are functions of the coordinates either alone or with
the time. The principle of conservation of probability of
phase will therefore apply, which requires that at any other
time (t") the maximum value of the coefficient of probability
shall be the same as at the time t\ and that the phase
(Pi', Qi') etc.) which has this greatest probabilitycoefficient,
shall be that which corresponds to the phase (P/, §/, etc.),
i. e., which is calculated from the same values of the constants
of the integral equations of motion.
We may therefore write for the probability that the phase
at the time t" falls within the limits p^1 and p:" + dp^ #/'
and #/' + cfy/', etc.,
" dpi" ...dqj', (52)
CONSERVATION OF+EXTENSIONINPHASE
where C represents the same value as in the preceding
formula, viz., the constant value of the maximum coefficient
of probability, and Fn is a quadratic function of the differences
Pi ~ pi"> i" . . . d£»" = 1, (53)
when the integration is extended over all possible phases.
It will be allowable to set ± oo for the limits of all the coor
dinates and momenta, not because these values represent the
actual limits of possible phases, but because the portions of
the integrals lying outside of the limits of all possible phases
will have sensibly the value zero. With ± oo for limits, the
equation gives
l, (64)
Vf Vf"
where/' is the discriminant * of F1, and/" that of F". This
discriminant is therefore constant in time, and like C an abso
lute invariant hi respect to the system of coordinates which
may be employed. In dimensions, like (72, it is the reciprocal
of the 2nth power of the product of energy and time.
Let us see precisely how the functions F' and F'f are related.
The principle of the conservation of the probabilitycoefficient
requires that any values of the coordinates and momenta at the
time tf shall give the function F' the same value as the corre
_ sponding coordinates and momenta at the time tn give to F".
Therefore Fn may be derived from F' by substituting for
Pi* • •  9.n their values in terms of p^', . . . (77)
d(rs, • ..r2n) d(r8, . . . r2n)
the coefficients of drl and dr% may be regarded as known func
tions of rx and r2 with the constants (78)
d(r8, . ..ran) d(r8, ...r2n)
which may be integrated by quadratures and gives V as func
tions of r1? r2 , ..., (83)
da " d(a,...h) d(rt, . . . r,n)
by which equation (82) may be reduced to the form
da =
M M
a, . . . h) d(b, ... A)
d(r2, . . .
Now we know by (71) that the coefficient of da is a func
tion of a, ... h. Therefore, as £, ... h are regarded as constant
in the equation, the first number represents the differential
AND THEORY OF INTEGRATION. 31
of a function of a, . . . h, which we may denote by a'. We
have then
da'= d(b,...h) dr^~ d(b*..K) dt> (85)
dfa, ...r2n) d(r2, ...r2n)
which may be integrated by quadratures. In this case we
may say that the principle of conservation of extensionin
phase has supplied the * multiplier '
1
d(b, ...h) (86)
d(rz, . . . rzn)
for the integration of the equation
dr, rldt = 0. (87)
The system of arbitrary constants a', 5, ... h has evidently
the same properties which were noticed in regard to the
system a, 6', ... h.
CHAPTER IV.
ON THE DISTRIBUTION IN PHASE CALLED CANONICAL,
IN WHICH THE INDEX OF PROBABILITY IS A LINEAR
FUNCTION OF THE ENERGY.
LET us now give our attention to the statistical equilibrium
of ensembles of conservation systems, especially to those cases
and properties which promise to throw light on the phenom
ena of thermodynamics.
The condition of statistical equilibrium may be expressed
in the form*
where P is the coefficient of probability, or the quotient of
the densityinphase by the whole number of systems. To
satisfy this condition, it is necessary and sufficient that P
should be a function of the p's and q*s (the momenta and
coordinates) which does not vary with the time in a moving
system. In all cases which we are now considering, the
energy, or any function of the energy, is such a function.
P = f unc. (e)
will therefore satisfy the equation, as indeed appears identi
cally if we write it in the form
... <*?» = !. (89)
phases
These considerations exclude
P = e X constant,
as well as
P = constant,
as cases to be considered.
The distribution represented by
(90)
or
where ® and i/r are constants, and % positive, seems to repre
sent the most simple case conceivable, since it has the property
that when the system consists of parts with separate energies,
the laws of the distribution in phase of the separate parts are
of the same nature, — a property which enormously simplifies
the discussion, and is the foundation of extremely important
relations to thermodynamics. The case is not rendered less
simple by the divisor ®, (a quantity of the same dimensions as
e,) but the reverse, since it makes the distribution independent
of the units employed. The negative sign of e is required by
(89), which determines also the value of ^ for any given
©, viz.,
all f
~®
=f. . .f
e dp,... dqn . (92)
phases
When an ensemble of systems is distributed in phase in the
manner described, i. e.^ when the index of probability is a
3
34 CANONICAL DISTRIBUTION
linear function of the energy, we shall say that the ensemble is
canonically distributed, and shall call the divisor of the energy
(®) the modulus of distribution.
The fractional part of an ensemble canonically distributed
which lies within any given limits of phase is therefore repre
sented by the multiple integral
9 dpl . . . dqn (93)
taken within those limits. We may express the same thing
by saying that the multiple integral expresses the probability
that an unspecified system of the ensemble (i. e., one of
which we only know that it belongs to the ensemble) falls
within the given limits.
Since the value of a multiple integral of the form (23)
(which we have called an extensioninphase) bounded by any
given phases is independent of the system of coordinates by
which it is evaluated, the same must be true of the multiple
integral in (92), as appears at once if we divide up this
integral into parts so small that the exponential factor may be
regarded as constant in each. The value of ^r is therefore in
dependent of the system of coordinates employed.
It is evident that ty might be defined as the energy for
which the coefficient of probability of phase has the value
unity. Since however this coefficient has the dimensions of
the inverse nth power of the product of energy and time,*
the energy represented by \Jr is not independent of the units
of energy and time. But when these units have been chosen,
the definition of ^ will involve the same arbitrary constant as
e, so that, while in any given case the numerical values of
^r or e will be entirely indefinite until the zero of energy has
also been fixed for the system considered, the difference ty — e
will represent a perfectly definite amount of energy, which is
entirely independent of the zero of energy which we may
choose to adopt.
* See Chapter I, p. 19.
OF AN ENSEMBLE OF SYSTEMS. 35
It is evident that the canonical distribution is entirely deter
mined by the modulus (considered as a quantity of energy)
and the nature of the system considered, since when equation
(92) is satisfied the value of the multiple integral (93) is
independent of the units and of the coordinates employed, and
of the zero chosen for the energy of the system.
In treating of the canonical distribution, we shall always
suppose the multiple integral in equation (92) to have a
finite value, as otherwise the coefficient of probability van
ishes, and the law of distribution becomes illusory. This will
exclude certain cases, but not such apparently, as will affect
the value of our results with respect to their bearing on ther
modynamics. It will exclude, for instance, cases in which the
system or parts of it can be distributed in unlimited space
(or in a space which has limits, but is still infinite in volume),
while the energy remains beneath a finite limit. It also
excludes many cases in which the energy can decrease without
limit, as when the system contains material points which
attract one another inversely as the squares of their distances.
Cases of material points attracting each other inversely as the
distances would be excluded for some values of ®, and not
for others. The investigation of such points is best left to
the particular cases. For the purposes of a general discussion,
it is sufficient to call attention to the assumption implicitly
involved in the formula (92).*
The modulus © has properties analogous to those of tem
perature in thermodynamics. Let the system A be defined as
one of an ensemble of systems of m degrees of freedom
distributed in phase with a probabilitycoefficient
*£%
e 0 ,
* It will be observed that similar limitations exist in thermodynamics. In
order that a mass of gas can be in thermodynamic equilibrium, it is necessary
that it be enclosed. There is no thermodynamic equilibrium of a (finite) mass
of gas in an infinite space. Again, that two attracting particles should be
able to do an infinite amount of work in passing from one configuration
(which is regarded as possible) to another, is a notion which, although per
fectly intelligible in a mathematical formula, is quite foreign to our ordinary
conceptions of matter.
36 CANONICAL DISTRIBUTION
and the system B as one of an ensemble of systems of n
degrees of freedom distributed in phase with a probability
coefficient
which has the same modulus. Let qv . . .qm, pv . . . pm be the
coordinates and momenta of A, and qm+l , . . . qm+n, pm+l , . . . pm+n
those of £. Now we may regard the systems A and B as
together forming a system 0, having m + n degrees of free
dom, and the coordinates and momenta q^ . . . /r B are constants, the probabilitycoefficient is of the
general form which we are considering, and the ensemble to
which it relates is in statistical equilibrium and is canonically
distributed.
This result, however, so far as statistical equilibrium is
concerned, is rather nugatory, since conceiving of separate
systems as forming a single system does not create any in
teraction between them, and if the systems combined belong to
ensembles in statistical equilibrium, to say that the ensemble
formed by such combinations as we have supposed is in statis
tical equilibrium, is only to repeat the data in different
OF AN ENSEMBLE OF SYSTEMS. 37
words. Let us therefore suppose that in forming the system
C we add certain forces acting between A and .5, and having
the forcefunction — eAB. The energy of the system C is now
€A + €B + €ABI and an ensemble of such systems distributed
with a density proportional to
(96)
would be in statistical equilibrium. Comparing this with the
probabilitycoefficient of C given above (95), we see that if
we suppose eAB (or rather the variable part of this term when
we consider all possible configurations of the systems A and B)
to be infinitely small, the actual distribution in phase of C
will differ infinitely little from one of statistical equilibrium,
which is equivalent to saying that its distribution in phase
will vary infinitely little even in a time indefinitely prolonged.*
The case would be entirely different if A and B belonged to
ensembles having different moduli, say ®A and ®5. The prob
abilitycoefficient of C would then be
which is not approximately proportional to any expression of
the form (96).
Before proceeding farther in the investigation of the dis
tribution in phase which we have called canonical, it will be
interesting to see whether the properties with respect to
* It will be observed that the above condition relating to the forces which
act between the different systems is entirely analogous to that which must
hold in the corresponding case in thermodynamics. The most simple test
of the equality of temperature of two bodies is that they remain in equilib
rium when brought into thermal contact. Direct thermal contact implies
molecular forces acting between the bodies. Now the test will fail unless
the energy of these forces can be neglected in comparison with the other
energies of the bodies. Thus, in the case of energetic chemical action be
tween the bodies, or when the number of particles affected by the forces
acting between the bodies is not negligible in comparison with the whole
number of particles (as when the bodies have the form of exceedingly thin
sheets), the contact of bodies of the same temperature may produce con
siderable thermal disturbance, and thus fail to afford a reliable criterion of
the equality of temperature.
38 OTHER DISTRIBUTIONS
statistical equilibrium which have been described are peculiar
to it, or whether other distributions may have analogous
properties.
Let rjr and 77" be the indices of probability in two independ
ent ensembles which are each in statistical equilibrium, then
rf _j_ y wni De the index in the ensemble obtained by combin
ing each system of the first ensemble with each system of the
second. This third ensemble will of course be in statistical
equilibrium, and the function of phase vf + if1 will be a con
stant of motion. Now when infinitesimal forces are added to
the compound systems, if r/ + rf1 or a function differing
infinitesimally from this is still a constant of motion, it must
be on account of the nature of the forces added, or if their action
is not entirely specified, on account of conditions to which
they are subject. Thus, in the case already considered,
V + ??" is a function of the energy of the compound system,
and the infinitesimal forces added are subject to the law of
conservation of energy.
Another natural supposition in regard to the added forces
is that they should be such as not to affect the moments of
momentum of the compound system. To get a case in which
moments of momentum of the compound system shall be
constants of motion, we may imagine material particles con
tained in two concentric spherical shells, being prevented from
passing the surfaces bounding the shells by repulsions acting
always in lines passing through the common centre of the
shells. Then, if there are no forces acting between particles in
different shells, the mass of particles in each shell will have,
besides its energy, the moments of momentum about three
axes through the centre as constants of motion.
Now let us imagine an ensemble formed by distributing in
phase the system of particles in one shell according to the
index of probability
• ^I++S+S' (98)
where e denotes the energy of the system, and ©j , o>2 , &>3 , its
three moments of momentum, and the other letters constants.
HAVE ANALOGOUS PROPERTIES. 39
In like manner let us imagine a second ensemble formed by
distributing in phase the system of particles in the other shell
according to the index
where the letters have similar significations, and O, Ox , O2 , 113
the same values as in the preceding formula. Each of the
two ensembles will evidently be in statistical equilibrium, and
therefore also the ensemble of compound systems obtained by
combining each system of the first ensemble with each of the
second. In this third ensemble the index of probability will be
k + ^!±^ + SL±^ + 2d^ + a±3L, (ioo)
vy i/j 1/2 »*a
where the four numerators represent functions of phase which
are constants of motion for the compound systems.
Now if we add in each system of this third ensemble infini
tesimal conservative forces of attraction or repulsion between
particles in different shells, determined by the same law for
all the systems, the functions o^ + &>', &>2 + o>2', and &>3 + w3'
will remain constants of motion, and a function differing in
finitely little from el + e will be a constant of motion. It
would therefore require only an infinitesimal change in the
distribution in phase of the ensemble of compound systems to
make it a case of statistical equilibrium. These properties are
entirely analogous to those of canonical ensembles.*
Again, if the relations between the forces and the coordinates
can be expressed by linear equations, there will be certain
" normal " types of vibration of which the actual motion may
be regarded as composed, and the whole energy may be divided
* It would not be possible to omit the term relating to energy in the above
indices, since without this term the condition expressed by equation (89)
cannot be satisfied.
The consideration of the above case of statistical equilibrium may be
made the foundation of the theory of the thermodynamic equilibrium of
rotating bodies, — a subject which has been treated by Maxwell in his memoir
" On Boltzmann's theorem on the average distribution of energy in a system
of material points." Cambr. Phil. Trans., vol. XII, p. 547, (1878).
40 OTHER DISTRIBUTIONS
into parts relating separately to vibrations of these different
types. These partial energies will be constants of motion,
and if such a system is distributed according to an index
which is any function of the partial energies, the ensemble will
be in statistical equilibrium. Let the index be a linear func
tion of the partial energies, say
Let us suppose that we have also a second ensemble com
posed of systems in which the forces are linear functions of
the coordinates, and distributed in phase according to an index
which is a linear function of the partial energies relating to
the normal types of vibration, say
^~i?'*'~if (102)
Since the two ensembles are both in statistical equilibrium,
the ensemble formed by combining each system of the first
with each system of the second will also be in statistical
equilibrium. Its distribution in phase will be represented by
the index
and the partial energies represented by the numerators in the
formula will be constants of motion of the compound systems
which form this third ensemble.
Now if we add to these compound systems infinitesimal
forces acting between the component systems and subject to
the same general law as those already existing, viz., that they
are conservative and linear functions of the coordinates, there
will still be n + m types of normal vibration, and n + m
partial energies which are independent constants of motion.
If all the original n + m normal types of vibration have differ
ent periods, the new types of normal vibration will differ infini
tesimally from the old, and the new partial energies, which are
constants of motion, will be nearly the same functions of
phase as the old. Therefore the distribution in phase of the
HAVE ANALOGOUS PROPERTIES. 41
ensemble of compound systems after the addition of the sup
posed infinitesimal forces will differ infinitesimally from one
which would be in statistical equilibrium.
The case is not so simple when some of the normal types of
motion have the same periods. In this case the addition of
infinitesimal forces may completely change the normal types
of motion. But the sum of the partial energies for all the
original types of vibration which have any same period, will
be nearly identical (as a function of phase, i. e., of the coordi
nates and momenta,) with the sum of the partial energies for
the normal types of vibration which have the same, or nearly
the same, period after the addition of the new forces. If,
therefore, the partial energies in the indices of the first two
ensembles (101) and (102) which relate to types of vibration
having the same periods, have the same divisors, the same will
be true of the index (103) of the ensemble of compound sys
tems, and the distribution represented will differ infinitesimally
from one which would be in statistical equilibrium after the
addition of the new forces.*
The same would be true if in the indices of each of the
original ensembles we should substitute for the term or terms
relating to any period which does not occur in the other en
semble, any function of the total energy related to that period,
subject only to the general limitation expressed by equation
(89). But in order that the ensemble of compound systems
(with the added forces) shall always be approximately in
statistical equilibrium, it is necessary that the indices of the
original ensembles should be linear functions of those partial
energies which relate to vibrations of periods common to the
two ensembles, and that the coefficients of such partial ener
gies should be the same in the two indices.f
* It is interesting to compare the above relations with the laws respecting
the exchange of energy between bodies by radiation, although the phenomena
of radiations lie entirely without the scope of the present treatise, in which
the discussion is limited to systems of a finite number of degrees of freedom.
t The above may perhaps be sufficiently illustrated by the simple case
where n = 1 in each system. If the periods are different in the two systems,
they may be distributed according to any functions of the energies : but if
42 CANONICAL DISTRIBUTION
The properties of canonically distributed ensembles of
systems with respect to the equilibrium of the new ensembles
which may be formed by combining each system of one en
semble with each system of another, are therefore not peculiar
to them in the sense that analogous properties do not belong
to some other distributions under special limitations in regard
to the systems and forces considered. Yet the canonical
distribution evidently constitutes the most simple case of the
kind, and that for which the relations described hold with the
least restrictions.
Returning to the case of the canonical distribution, we
shall find other analogies with thermodynamic systems, if we
suppose, as in the preceding chapters,* that the potential
energy (eq) depends not only upon the coordinates ql . . . qn
which determine the configuration of the system, but also
upon certain coordinates «i, «2, etc. of bodies which we call
external? meaning by this simply that they are not to be re
garded as forming any part of the system, although their
positions affect the forces which act on the system. The
forces exerted by the system upon these external bodies will
be represented by — deqjdav — deqfda2, etc., while — deqjdqv
... — deq/dqn represent all the forces acting upon the bodies
of the system, including those which depend upon the position
of the external bodies, as well as those which depend only
upon the configuration of the system itself. It will be under
stood that €p depends only upon qi , . . . qn , p\ , . . . pn , in other
words, that the kinetic energy of the bodies which we call
external forms no part of the kinetic energy of the system.
It follows that we may write
although a similar equation would not hold for differentiations
relative to the internal coordinates.
the periods are the same they must be distributed canonically with same
modulus in order that the compound ensemble with additional forces may
be in statistical equilibrium.
* See especially Chapter I, p. 4.
OF AN ENSEMBLE OF SYSTEMS. 43
We always suppose these external coordinates to have the
same values for all systems of any ensemble. In the case of
a canonical distribution, i. e., when the index of probability
of phase is a linear function of the energy, it is evident that
the values of the external coordinates will affect the distribu
tion, since they affect the energy. In the equation
(105)
by which ty may be determined, the external coordinates, ax ,
02, etc., contained implicitly in e, as well as ®,^are to be re
garded as constant in the integrations indicated. The equa
tion indicates that fy is a function of these constants. If we
imagine their values varied, and the ensemble distributed
canonically according to their new values, we have by
differentiation of the equation ^
/ v aii
f i ./. \ 1 /»
0
, \
( I ^ + I «») = p
all
phases
all Jf
• • /^ e~° dPi • • • dv ~ ete> (106)
phases
t
or, multiplying by 0 e®, and setting
^=^ £=^ etc>
all
d® = ^® f. . .f
ee
phases
—
i e ® dpl . . . dqn
phases
r r
i I . . .
phases
r * (• fcf
2J ...JA2e&dpl...dqn + etc. (107)
44 CANONICAL DISTRIBUTION
Now the average value in the ensemble of any quantity
(which we shall denote in general by a horizontal line above
the proper symbol) is determined by the equation
r M C fc!
« =J • • • J u e & dPl... dqa. (108)
phases
Comparing this with the preceding equation, we have
Z2 d«2 — etc.
Moreover, since (111) gives
dty  c?e = ©cfy + ^©, (113)
we have also
dk — — ® drj — ^ ddi — A2 da2 — etc. (114)
This equation, if we neglect the sign of averages, is identi
cal in form with the thermodynamic equation
de + Alda1 + Az daz + etc.
drj= —y— , (115)
or
de = Tdrj — A! daL — Az da2 — etc., (H6)
which expresses the relation between the energy, .tempera
ture, and entropy of a body in thermodynamic equilibrium,
and the forces which it exerts on external bodies, — a relation
which is the mathematical expression of the second law of
thermodynamics for reversible changes. The modulus in the
statistical equation corresponds to temperature in the thermo
dynamic equation, and the average index of probability with
its sign reversed corresponds to entropy. But in the thermo
dynamic equation the entropy (77) is a quantity which is
OF AN ENSEMBLE OF SYSTEMS. 45
only defined by the equation itself, and incompletely defined
in that the equation only determines its differential, and the
constant of integration is arbitrary. On the other hand, the
77 in the statistical equation has been completely defined as
the average value in a canonical ensemble of systems of
the logarithm of the coefficient of probability of phase.
We may also compare equation (112) with the thermody
namic equation
A^ = — T]dT—Aldal — Azda •••« 20 . (120)
The potential energy (e3) is independent of the velocities,
and if the limits of integration for the coordinates are inde
pendent of the velocities, and the limits of the several veloci
ties are independent of each other as well as of the coordinates,
VALUES IN A CANONICAL ENSEMBLE. 47
the multiple integral may be resolved into the product of
integrals
C. . . C
mvdzv. (121)
This shows that the probability that the configuration lies
within any given limits is independent of the velocities,
and that the probability that any component velocity lies
within any given limits is independent of the other component
velocities and of the configuration.
Since
* 2
f 4V«>, <& = vz^®, (122>
I/ 00
and
J
e 2 ® m* dx! = V^Timx©8, (123>
the average value of the part of the kinetic energy due to the
velocity x19 which is expressed by the quotient of these inte
grals, is J 's, it may be divided
into parts by the formula
_ 1 ^^p I @£p /I OQ\
ENSEMBLE OF SYSTEMS. 49
where e might be written for ep in the differential coefficients
without affecting the signification. The average value of the
first of these parts, for any given configuration, is expressed
by the quotient
/+» f+» de ^r .
• • • / i*l ~fo 6 dPl ' ' • dPn
_oo J —oo api
=r (129)
e ® dpi . . . dpn
Now we have by integration by parts
tyC
r °° PI <^~^ dPl = © r 4
,/ _oo api j _
By substitution of this value, the above quotient reduces to
— , which is therefore the average value of \P\— for the
2 dpi
given configuration. Since this value is independent of the
configuration, it must also be the average for the whole
ensemble, as might easily be proved directly. (To make
the preceding proof apply directly to the whole ensemble, we
have only to write dp1 . . . dqn for dp± . . . dpn in the multiple
integrals.) This gives J n ® for the average value of the
whole kinetic energy for any given configuration, or for
the whole ensemble, as has already been proved in the case of
material points.
The mechanical significance of the several parts into which
the kinetic energy is divided in equation (128) will be appar
ent if we imagine that by the application of suitable forces
(different from those derived from eq and so much greater
that the latter may be neglected in comparison) the system
was brought from rest to the state of motion considered, so
rapidly that the configuration was not sensibly altered during
the process, and in such a manner also that the ratios of the
component velocities were constant in the process. If we
write
50 AVERAGE VALUES IN A CANONICAL
for the moment of these forces, we have for the period of their
action by equation (3)
* =(^d^ + Fl =  — + Fl
dqi dqi dqi
The work done by the force F± may be evaluated as follows :
r rd€ *
= I Pi dqt f I y—dqit
J J dq^
where the last term may be cancelled because the configuration
does not vary sensibly during the application of the forces.
(It will be observed that the other terms contain factors which
increase as the tune of the action of the forces is diminished.)
We have therefore,
f* f* n f*
\ dqi = I pi £1 dt = I qi dpt=. — I Pi dpi . (131)
For since the p's are linear functions of the q's (with coeffi
cients involving the #'s) the supposed constancy of the 's, when the in
tegrations are to cover all values of the jt?'s (for constant #'s)
once and only once, they must cover all values of the w's once
and only once, and the limits will be ± oo for all the u's.
Without the supposition of the last paragraph the upper limits
would not always be + oo , as is evident on considering the
effect of changing the sign of a u. But with the supposition
which we have made (that the determinant is always positive)
we may make the upper limits + oo and the lower — oo for all
the t*'s. Analogous considerations will apply where the in
tegrations do not cover all values of the p's and therefore of
* The reduction requires only the repeated application of the process of
'completing the square* used in the solution of quadratic equations.
52 AVERAGE VALUES IN A CANONICAL
the w's. The integrals may always be taken from a less to a
greater value of a u.
The general integral which expresses the fractional part of
the ensemble which falls within any given limits of phase is
thus reduced to the form
...<*«*«*&...%,. (134)
For the average value of the part of the kinetic energy
which is represented by ^u^ whether the average is taken
for the whole ensemble, or for a given configuration, we have
therefore
__ (135)
— '
I/
e
00
and for the average of the whole kinetic energy, JTI©, as
before.
The fractional part of the ensemble which lies within any
given limits of configuration, is found by integrating (184)
with respect to the w's from — oo to + oo . This gives
J f.
• da,
which shows that the value of the Jacobian is independent of
the manner in which 2ep is divided into a sum of squares.
We may verify this directly, and at the same tune obtain a
more convenient expression for the Jacobian, as follows.
It will be observed that since the M'S are linear functions of
the p's, and the jt?'s linear functions of the ^'s, the u's will be
linear functions of the = 2ne>
we have I — ea = n e.
\
\
56 AVERAGES IN A CANONICAL ENSEMBLE.
/„ ^pfp
•••je & **dPl...dpn, (144)
or again
r r^=^ i
I . . . / e < Ar^Ti • • • 4»i (145)
for the fractional part of the systems of any given configura
tion which lie within given limits of velocity.
When systems are distributed in velocity according to these
formulae, i. e., when the distribution in velocity is like that in
an ensemble which is canonically distributed in phase, we
shall say that they are canonically distributed in velocity.
The fractional part of the whole ensemble which falls
within any given limits of phase, which we have before
expressed in the form
. dpndqi . . . dqn, (146)
may also be expressed in the form
. . dqndql . . . dqn. (147)
CHAPTER VI.
EXTENSION IN CONFIGURATION AND EXTENSION
IN VELOCITY.
THE formulae relating to canonical ensembles in the closing
paragraphs of the last chapter suggest certain general notions
and principles, which we shall consider in this chapter, and
which are not at all limited in their application to the canon
ical law of distribution.*
We have seen in Chapter IV. that the nature of the distribu
tion which we have called canonical is independent of the
system of coordinates by which it is described, being deter
mined entirely by the modulus. It follows that the value
represented by the multiple integral (142), which is the frac
tional part of the ensemble which lies within certain limiting
configurations, is independent of the system of coordinates,
being determined entirely by the limiting configurations with
the modulus. Now tr, as we have already seen, represents
a value which is independent of the system of coordinates
by which it is defined. The same is evidently true of
typ by equation (140), and therefore, by (141), of tyg.
Hence the exponential factor in the multiple integral (142)
represents a value which is independent of the system of
coordinates. It follows that the value of a multiple integral
of the form
^ ...dgn (148)
* These notions and principles are in fact such as a more logical arrange
ment of the subject would place in connection with those of Chapter I., to
which they are closely related. The strict requirements of logical order
have been sacrificed to the natural development of the subject, and very
elementary notions have been left until they have presented themselves in
the study of the leading problems.
58 EXTENSION IN CONFIGURATION
is independent of the system of coordinates which is employed
for its evaluation, as will appear at once, if we suppose the
multiple integral to be broken up into parts so small that
the exponential factor may be regarded as constant in each.
In the same way the formulae (144) and (145) which express
the probability that a system (in a canonical ensemble) of given
configuration will fall within certain limits of velocity, show
that multiple integrals of the form
(149)
or *» **&„. 1* (150)
relating to velocities possible for a given configuration, when
the limits are formed by given velocities, have values inde
pendent of the system of coordinates employed.
These relations may easily be verified directly. It has al
ready been proved that
d(Pl9 . . . P.) <%i . . . qn) d(ql9 ...qn)
..) d(Ql9...Qn)
where ql , . . . q^ft , . . .pn and Ql , . . . Qn9 P1 , . . . Pn are two
systems of coordinates and momenta.* It follows that
i>
= r
J
* See equation (29).
AND EXTENSION IN VELOCITY. 59
and
/Cfd(Ql, ... Qn)\% JT> Jp
' ' J \d(P^ ~^P}) ' *
"'<%>!... !,.. W
The multiple integral
•
>! . . . dpndqi . . . rf^, (151)
which may also be written
£1 . . . dqndqi . . . dqn, (152)
and which, when taken within any given limits of phase, has
been shown to have a value independent of the coordinates
employed, expresses what we have called an extensionin
phase.* In like manner we may say that the multiple integral
(148) expresses an extensioninconfiguration, and that the
multiple integrals (149) and (150) express an extensionrin
velocity. We have called
dpi . . . *Y"» of which v* is of the same nature as Fi ' V*
of the same nature as V2", while VB"f satisfies the relations
that if combined either with Fi or V£ the kinetic energy of
the combined velocities is the sum of the kinetic energies of
the velocities taken separately. When all the velocities
Fg , . . . Vn have been thus decomposed, the square root of the
product of the doubled kinetic energies of the several velocities
PI> JY'» JY"» ete*' ^H be the value of the extensionin
velocity which is sought.
This method of evaluation of the extensionin velocity which
we are considering is perhaps the most simple and natural, but
the result may be expressed in a more symmetrical form. Let
us write e12 for the kinetic energy of the velocities Fx and V%
combined, diminished by the sum of the kinetic energies due
to the same velocities taken separately. This may be called
the mutual energy of the velocities V\ and F2 . Let the
mutual energy of every pair of the velocities Fj , . . . Vn be
expressed in the same way. Analogy would make en represent
the energy of twice V1 diminished by twice the energy of Fi ,
i. e.y en would represent twice the energy of Fi , although the
term mutual energy is hardly appropriate to this case. At all
events, let en have this signification, and e22 represent twice
the energy of F^, etc. The square root of the determinant
n €12 ... €i
represents the value of the extensioninvelocity determined as
above described by the velocities V\ , . . . FJ,.
The statements of the preceding paragraph may be readily
proved from the expression (157) on page 60, viz.,
A •
by which the notion of an element of extensioninvelocity was
AND EXTENSION IN VELOCITY. 67
originally defined. Since A^ in this expression represents
the determinant of which the general element is
the square of the preceding expression represents the determi
nant of which the general element is
Now we may regard the differentials of velocity dqt, d^ as
themselves infinitesimal velocities. Then the last expression
represents the mutual energy of these velocities, and
d*e
represents twice the energy due to the velocity dq{.
The case which we have considered is an extensioninveloc
ity of the simplest form. All extensionsinvelocity do not
have this form, but all may be regarded as composed of
elementary extensions of this form, in the same manner as
all volumes may be regarded as composed of elementary
parallelepipeds.
Having thus a measure of extensionin velocity founded, it
will be observed, on the dynamical notion of kinetic energy,
and not involving an explicit mention of coordinates, we may
derive from it a measure of extensioninconfiguration by the
principle connecting these quantities which has been given in
a preceding paragraph of this chapter.
The measure of extensioninphase may be obtained from
that of extensioninconfiguration and of extensionin velocity.
For to every configuration in an extensioninphase there will
belong a certain extensioninvelocity, and the integral of the
elements of extensioninconfiguration within any extension
inphase multiplied each by its extensioninvelocity is the
measure of the extensioninphase.
CHAPTER VII.
FARTHER DISCUSSION OF AVERAGES IN A CANONICAL
ENSEMBLE OF SYSTEMS.
RETURNING to the case of a canonical distribution, we have
for the index of probability of configuration
as appears on comparison of formulae (142) and (161). It
follows immediately from (142) that the average value in the
ensemble of any quantity u which depends on the configura
tion alone is given by the formula
r au ^ *
" <227>
*(228)
The average values of the powers of the anomalies of the
energies are perhaps most easily found as follows. We have
identically, since e is a function of ®, while e is a function of
the jt?'s and e».
78 AVERAGE VALUES IN A CANONICAL
or since by (218)
e)»« = e(ee)»  A <«
In precisely the same way we may obtain for the potential
energy
(63i3)^ = @2^(e3 eq^ + h(eq eq)^ ©2g. (232)
By successive applications of (231) we obtain
(e  i)2 =
(ee)8 =•
(e  e)6 = J>5e + 15DeD*e + 10(D2€)2 + 15(Z)e)8 etc.
where D represents the operator ®'2d/d®. Similar expres
sions relating to the potential energy may be derived from
(232).
For the kinetic energy we may write similar equations in
which the averages may be taken either for a single configura
tion or for the whole ensemble. But since
d€p _ n
d®~2
the general formula reduces to
(ep  ep)™ = ©2 A (€p  ep)» + ±nh& (ep  ~ep)^ (233)
or
(234)
ENSEMBLE OF SYSTEMS. 79
But since identically
the value of the corresponding expression for any index will
be independent of <*) and the formula reduces to
we have therefore
etc.1
It will be observed that when i/r or e is given as function of
O, all averages of the form e^ or (e — T)ft are thereby deter
* In the case discussed in the preceding footnotes we get easily
and
For the total energy we have in this case
l h ~
x±Tx2 i
VeJ =n'
ft — €\ ° _ 2
etc.
rurxs iar A
• ou
: .•
/
f.
J
* «»»
ENSEMBLE OP SYSTEMS.
The multiple integrals in
average rallies of the expressions In the brackets,
may therefore set equal to zero. The first gives
as already obtained. With this relation and (191) we get
from the other equations
We may add for comparison equation (205), which might be
derived from (236) by differentiating twice with respect to 8 :
The two last equations give
dl
(Al  Al)(e  e) = — (6  €)'. (245)
e?e
If i/r or e is known as function of 0, Oj, Oj, etc*, (e — e)2 may
be obtained by differentiation as function of the same variables.
And if ir, or Av or 17" is known as function of 8, O
(e — e) may be obtained by differentiation. But
(^Al — A^y and (^Al — A^) (^2 — A2) cannot be obtained in any
similar manner. We have seen that (e— e)2 is in general a
vanishing quantity for very great values of TI, which we may
regard as contained implicitly in 0 as a divisor. The same is
true of (A^ — A^) (e — e). It does not appear that we can
assert the same of (A^ — 4X)2 or (Al — A^) (^2 — 42), since
6
82 AVERAGE VALUES IN A CANONICAL
a^ may be very great. The quantities dte/da^ an
belong to the class called elasticities. The former expression
represents an elasticity measured under the condition that
while &J is varied the internal coordinates ql9 . . . qn all remain
fixed. The latter is an elasticity measured under the condi
tion that when ax is varied the ensemble remains canonically
distributed within the same modulus. This corresponds to
an elasticity in physics measured under the condition of con
stant temperature. It is evident that the former is greater
than the latter, and it may be enormously greater.
The divergences of the force Al from its average value are
due in part to the differences of energy in the systems of the
ensemble, and in part to the differences in the value of
the forces which exist in systems of the same energy. If we
write A^ for the average value of Al in systems of the
ensemble which have any same energy, it will be determined
by the equation
/ . . . J e ®
. . . dqn
where the limits of integration in both multiple integrals are
two values of the energy which differ infinitely little, say e and
fc±
e + de. This will make the factor e & constant within the
limits of integration, and it may be cancelled in the numera
tor and denominator, leaving
/•••/ £<&>! ...dqn
2H.= / / (247)
J...J*!...*.
where the integrals as before are to be taken between e and
e + de. A^\f is therefore independent of ®, being a function
of the energy and the external coordinates.
ENSEMBLE OF SYSTEMS. 83
Now we have identically
Al — Ai = (Ai — 2T)e) + (2T1 1 — 4)>
where Al — ~A^e denotes the excess of the force (tending to
increase a^ exerted by any system above the average of such
forces for systems of the same energy. Accordingly,
But the average value of (Al — A^\f) (A^\ e — A^) for systems
of the ensemble which have the same energy is zero, since for
such systems the second factor is constant. Therefore the
average for the whole ensemble is zero, and
Atf. (248)
In the same way it may be shown that
(A,  Al) (ee) = (^  AJ (e  e). (249)
It is evident that in ensembles in which the anomalies of
energy e — e may be regarded as insensible the same will be
true of the quantities represented by A^\f — A^
The properties of quantities of the form A^\€ will be
farther considered in Chapter X, which will be devoted to
ensembles of constant energy.
It may not be without interest to consider some general
formulae relating to averages in a canonical ensemble, which
embrace many of the results which have been given in this
chapter.
Let u be any function of the internal and external coordi
nates with the momenta and modulus. We have by definition
**.>,V:.fc!
uJ...Juee d^.^dq, (250)
phases
If we differentiate with respect to ®, we have
du f a r/du u u e
d®=J J (353 <#^i
phases
84 AVERAGE VALUES IN A CANONICAL
du _du utye) udif,
d®~d®  &— + ®d®'
Setting u = 1 in this equation, we get
d\f/ _ \i/ — €
d®~ 0
and substituting this value, we have
du du ue ue
If we differentiate equation (250) with respect to a (which
may represent any of the external coordinates), and write A
for the force — ^ , we get
__ ail t *.
du r r( du u dif/ u . \
3= /.../V5 + ^^+7v^)
da J J \da © da 0 /
da
phases
du du
or — = —
Setting w = 1 hi this equation, we get
Substituting this value, we have
du au uA uA
du du
or ®r®r = ^2uI=(uu)(A2). (255)
da aa
Repeated applications of the principles expressed by equa
tions (252) and (255) are perhaps best made in the particular
cases. Yet we may write (252) in this form
ENSEMBLE OF SYSTEMS. 85
(€ + D) (u  u) = 0, (256)
where D represents the operator ®2 d/d®.
Hence
(e + D)A (u  u) = 0, (257)
where h is any positive whole number. It will be observed,
that since e is not function of ®, (e + D)h may be expanded by
the binomial theorem. Or, we may write
(e + />) u = (e + D) u, (258)
whence (e + X>)* u = (e + D)h u. (259)
But the operator (e + D)*, although in some respects more
simple than the operator without the average sign on the e,
cannot be expanded by the binomial theorem, since e is a
function of ® with the external coordinates.
So from equation (254) we have
<26°)
whence (~ + J;)* («  u) = 0 ; (261)
The binomial theorem cannot be applied to these operators.
Again, if we now distinguish, as usual, the several external
coordinates by suffixes, we may apply successively to the
expression u — u any or all of the operators
,
, etc. (264)
86 AVERAGES IN A CANONICAL ENSEMBLE.
as many times as we choose, and in any order, the average
value of the result will be zero. Or, if we apply the same
operators to u, and finally take the average value, it will be the
same as the value obtained by writing the sign of average
separately as u, and on e, A± , A2 , etc., in all the operators.
If u is independent of the momenta, formulae similar to
the preceding, but having eq in place of e, may be derived
from equation (179).
CHAPTER VIII.
ON CERTAIN IMPORTANT FUNCTIONS OF THE
ENERGIES OF A SYSTEM.
IN order to consider more particularly the distribution of a
canonical ensemble in energy, and for other purposes, it will
be convenient to use the following definitions and notations.
Let us denote by J^the extensioninphase below a certain
limit of energy which we shall call e. That is, let
>x . . . dqn, (265)
the integration being extended (with constant values of the
external coordinates) over all phases for which the energy is
less than the limit e. We shall suppose that the value of this
integral is not infinite, except for an infinite value of the lim
iting energy. This will not exclude any kind of system to
which the canonical distribution is applicable. For if
>i • • • dqn
taken without limits has a finite value,* the less value repre
sented by
e
/...
u
•
taken below a limiting value of 6, and with the e before the
integral sign representing that limiting value, will also be
finite. Therefore the value of V, which differs only by a
constant factor, will also be finite, for finite e. It is a func
tion of e and the external coordinates, a continuous increasing
* This is a necessary condition of the canonical distribution. See
Chapter IV, p. 35.
88 CERTAIN IMPORTANT FUNCTIONS
function of 6, which becomes infinite with e, and vanishes
for the smallest possible value of e, or f or e = — oo, if the
energy may be diminished without limit.
Let us also set
dV
= log — • (266)
The extension in phase between any two limits of energy, ^
and e", will be represented by the integral
/ de. (267)
And in general, we may substitute e* de for dpl . . . dqn in a
2ttfold integral, reducing it to a simple integral, whenever
the limits can be expressed by the energy alone, and the other
factor under the integral sign is a function of the energy alone,
or with quantities which are constant in the integration.
In particular we observe that the probability that the energy
of an unspecified system of a canonical ensemble lies between
the limits e' and e" will be represented by the integral *
* 0ffe, (268)
and that the average value in the ensemble of any quantity
which only varies with the energy is given by the equation j
(269)
where we may regard the constant *fy as determined by the
equation $
^»
=l
6=00
—
&
e de, (270)
F=0
In regard to the lower limit in these integrals, it will be ob
served that V= 0 is equivalent to the condition that the
value of e is the least possible.
* Compare equation (93). t Compare equation (108).
J Compare equation (92).
OF THE ENERGIES OF A SYSTEM. 89
In like manner, let us denote by Vq the extensioninconfigu
ration below a certain limit of potential energy which we may
call eg. That is, let
• JV
(2T1)
the integration being extended (with constant values of the
external coordinates) over all configurations for which the
potential energy is less than eg. Vq will be a function of eq
with the external coordinates, an increasing function of e3,
which does not become infinite (in such cases as we shall con
sider *) for any finite value of eq. It vanishes for the least
possible value of e?, or for eq = — oo , if eq can be diminished
without limit. It is not always a continuous function of eg.
In fact, if there is a finite extensioninconfiguration of con
stant potential energy, the corresponding value of Vq will
not include that extensioninconfiguration, but if eq be in
creased infinitesimally, the corresponding value of Vq will be
increased by that finite extensioninconfiguration.
Let us also set
(272)
The extensioninconfiguration between any two limits of
potential energy eq and eqf may be represented by the integral
(273)
whenever there is no discontinuity in the value of Vq as
function of eq between or at those limits, that is, when
ever there is no finite extensioninconfiguration of constant
potential energy between or at the limits. And hi general,
with the restriction mentioned, we may substitute e^q deq for
Aj dq1 . . . dqn in an wfold integral, reducing it to a simple
integral, when the limits are expressed by the potential energy,
and the other factor under the integral sign is a function of
* If Vq were infinite^ for finite values of e,, V would evidently be infinite
for finite values of e.
90 CERTAIN IMPORTANT FUNCTIONS
the potential energy, either alone or with quantities which are
constant in the integration.
We may often avoid the inconvenience occasioned by for
mulae becoming illusory on account of discontinuities in the
values of Vq as function of eq by substituting for the given
discontinuous function a continuous function which is practi
cally equivalent to the given function for the purposes of the
evaluations desired. It only requires infinitesimal changes of
potential energy to destroy the finite extensionsinconfigura
tion of constant potential energy which are the cause of the
difficulty.
In the case of an ensemble of systems canonically distributed
in configuration, when Vq is, or may be regarded as, a continu
ous function of eq (within the limits considered), the proba
bility that the potential energy of an unspecified system lies
between the limits eq and eq' is given by the integral
where ^ may be determined by the condition that the value of
the integral is unity, when the limits include all possible
values of eq. In the same case, the average value in the en
semble of any function of the potential energy is given by the
equation
u = / ue d€q. (275)
Vq=0
When Vq is not a continuous function of eff, we may write d Vq
for e*qdeg in these formulae.
In like manner also, for any given configuration, let us
denote by Vp the extensioninvelocity below a certain limit of
kinetic energy specified by ep. That is, let
V, = J.
(276)
OF THE ENERGIES OF A SYSTEM. 91
the integration being extended, with constant values of the
coordinates, both internal and external, over all values of the
momenta for which the kinetic energy is less than the limit ep.
Vp will evidently be a continuous increasing function of ep
which vanishes and becomes infinite with e. Let us set
The extensioninvelocity between any two limits of kinetic
energy ep and ep" may be represented by the integral
f
e*pdep. (278)
And in general, we may substitute e^p dep for A,* dpl . . . dpn
or Ag* dql . . . dqn in an wfold integral in which the coordi
nates are constant, reducing it to a simple integral, when the
limits are expressed by the kinetic energy, and the other factor
under the integral sign is a function of the kinetic energy,
either alone or with quantities which are constant in the
integration.
It is easy to express Vp and $p in terms of ep. Since A^ is
function of the coordinates alone, we have by definition
1...dpn (279)
the limits of the integral being given by ep. That is, if
ep = F(Pl,...Pa), (280)
the limits of the integral for ep = 1, are given by the equation
F(Pl,...Pa) = \, (281)
and the limits of the integral for ep — a2, are given by the
equation
=«'. (282)
But since F represents a quadratic function, this equation
may be written
1 (283)
92 CERTAIN IMPORTANT FUNCTIONS
The value of Vp may also be put in the form
r, = ***f...f*&...*%. (284)
Now we may determine Vp for ep = 1 from (279) where the
limits are expressed by (281), and FJ, for ep ,= a2 from (284)
taking the limits from (283). The two integrals thus deter
mined are evidently identical, and we have
(285)
i. e., Vv varies as e/. We may therefore set
, n
Vp=Cep*> eP = nCep* j (286)
where C is a constant, at least for fixed values of the internal
coordinates.
To determine this constant, let us consider the case of a
canonical distribution, for which we have
_
where e& = (2*®) 2.
Substituting this value, and that of e*' from (286), we get
(287)
Having thus determined the value of the constant (7, we may
OF THE ENERGIES OF A SYSTEM.  93
substitute it in the general expressions (286), and obtain the
following values, which are perfectly general :
~ *(289)
It will be observed that the values of Vp and p for any
given ep are independent of the configuration, and even of the
nature of the system considered, except with respect to its
number of degrees of freedom.
Returning to the canonical ensemble, we may express the
probability that the kinetic energy of a system of a given
configuration, but otherwise unspecified, falls within given
limits, by either member of the following equation
Since this value is independent of the coordinates it also
represents the probability that the kinetic energy of an
unspecified system of a canonical ensemble falls within the
limits. The form of the last integral also shows that the prob
ability that the ratio of the kinetic energy to the modulus
* Very similar values for Vq, <&*, V, and e* may be found in the same
way in the case discussed in the preceding footnotes (see pages 54, 72, 77, and
79), in which e3 is a quadratic function of the q's, and Aj independent of the q'a.
In this case we have
(2 ')*(«« 
P(Jn)
+ i)
94 CERTAIN IMPORTANT FUNCTIONS
falls within given limits is independent also of the value of
the modulus, being determined entirely by the number of
degrees of freedom of the system and the limiting values
of the ratio.
The average value of any function of the kinetic energy,
either for the whole ensemble, or for any particular configura
tion, is given by
€p
—•£ ?i
ue 0e,2 dep *(291)
Thus:
^®"' if m + ^>°> t(292)
* The corresponding equation for the average value of any function of
the potential energy, when this is a quadratic function of the ^'s, and A£ is
independent of the q's, is
In the same case, the average value of any function of the (total) energy is
given by the equation
Hence in this case
j .f m + n>0
and = , if
ii f vy
If n = 1, e* = 2 ir and d^jde = 0 for any value of e. If n = 2, the case is
the same with respect to 02.
t This equation has already been proved for positive integral powers of
the kinetic energy. See page 77.
OF THE ENERGIES OF A SYSTEM. 95
n n
) /o _\9 ^2 ~j if w > 1 ; (294)
if n > 2 ; (295)
= ©. (296)
If n = 2, e*p = 2 TT, and dp/dep = 0, for any value of ep.
The definitions of F, V# and F^, give
(297)
where the integrations cover all phases for which the energy
is less than the limit e, for which the value of Fis sought.
This gives
V=CvpdVq, (298)
and ,jr €9=6
e* = ~ — f e^pdVn, (299}
de j
where Vp and e^p are connected with Vq by the equation
€p + eq = constant ~ e. (300)
If n > 2, e*? vanishes at the upper limit, i. e., for ep = 0, and
we get by another differentiation
€q=€
We may also write
62= e
F= J "P;/9^, (302)
* r
°=J
(303)
96 CERTAIN IMPORTANT FUNCTIONS
etc., when Vq is a continuous function of eq commencing with
the value Vq = 0, or when we choose to attribute to Vq a
fictitious continuity commencing with the value zero, as de
scribed on page 90.
If we substitute hi these equations the values of Vp and e^p
which we have found, we get
^= r/il /^ <«  <«) <* ^« ' (304)
(305)
where e^« c?eg may be substituted for d Vq in the cases above
described. If, therefore, n is known, and Vq as function of
€p V and e^ may be found by quadratures.
It appears from these equations that F"is always a continu
ous increasing function of e, commencing with the value V=
0, even when this is not the case with respect to Vq and eq.
The same is true of e^, when n > 2, or when n = 2 if Vq in
creases continuously with eq from the value Vq = 0.
The last equation may be derived from the preceding by
differentiation with respect to e. Successive differentiations
give, if h < } n + 1,
dhVjdQ is therefore characterized by
the equation
(309)
de, de,
The values of ep and eq determined by this maximum we shall
distinguish by accents, and mark the corresponding values of
functions of ep and eq in the same way. Now we have by
Taylor's theorem
If the approximation is sufficient without going beyond the
quadratic terms, since by (300)
€P ~€P' =  (e*  «/)»
we may write
+^(d^P\'(d\}'\(^ii^
2 *.» (312>
where the limits have been made ± oo for analytical simplicity.
This is allowable when the quantity in the square brackets
has a very large negative value, since the part of the integral
7
98 CERTAIN IMPORTANT FUNCTIONS
corresponding to other than very small values of eq — eqf may
be regarded as a vanishing quantity.
This gives
> _ A/+V /ON
(313)
or
^V+^' + ilog(2,)ilog[(^)'(^)']. (3U)
From this equation, with (289), (300) and (309), we .may
determine the value of $ corresponding to any given value of
e, when q is known as function of eq.
Any two systems may be regarded as together forming a
third system. If we have F or $ given as function of e for
any two systems, we may express by quadratures J^and $ for
the system formed by combining the two. If we distinguish
by the suffixes ( )x, ( )2, ( )12 the quantities relating to
the three systems, we have easily from the definitions of these
quantities
=ff
(sis)
$12  04>*f7T7' / p^1 fj T7" / n^1 ' ^2x7 /O1 £\
«/ «/ «y
where the double integral is to be taken within the limits
Vi = 0, V2 = 0, and el + e2 = e12 ,
and the variables in the single integrals are connected by the
last of these equations, while the limits are given by the first
two, which characterize the least possible values of e1 and e2
respectively.
It will be observed that these equations are identical in
form with those by which F'and $ are derived from Vp or cf>p
and Vq or q, except that they do not admit in the general
case those transformations which result from substituting for
Vp or (f>p the particukr functions which these symbols always
represent.
OF THE ENERGIES OF A SYSTEM. 99
Similar formulae may be used to derive Vq or q for the
compound system, when one of these quantities is known.
as function of the potential energy in each of the systems
combined.
The operation represented by such an equation as
C
= I
01 02
e e
is identical with one of the fundamental operations of the
theory of errors, viz., that of finding the probability of an error
from the probabilities of partial errors of which it is made up.
It admits a simple geometrical illustration.
We may take a horizontal line as an axis of abscissas, and lay
off 61 as an abscissa measured to the right of any origin, and
erect e^i as a corresponding ordinate, thus determining a certain
curve. Again, taking a different origin, we may lay off e2 as
abscissas measured to the left, and determine a second curve by
erecting the ordinates e^. We may suppose the distance be
tween the origins to be e12, the second origin being to the right
if e12 is positive. We may determine a third curve by erecting
at every point in the line (between the least values of ei and e2)
an ordinate which represents the product of the two ordinates
belonging to the curves already described The area between
this third curve and the axis of abscissas will represent the value
of e^12. To get the value of this quantity for varying values
of 612, we may suppose the first two curves to be rigidly con
structed, and to be capable of being moved independently. We
may increase or diminish e12 by moving one of these curves to
the right or left. The third curve must be constructed anew
for each different value of e12.
CHAPTER IX.
THE FUNCTION <£ AND THE CANONICAL DISTRIBUTION.
IN this chapter we shall return to the consideration of the
canonical distribution, in order to investigate those properties
which are especially related to the function of the energy
which we have denoted by .
If we denote by JV, as usual, the total number of systems
in the ensemble,
will represent the number having energies between the limits
e and e + de. The expression
Ne
(317)
represents what may be called the densityinenergy. This
vanishes for e = GO, for otherwise the necessary equation
(318)
could not be fulfilled. For the same reason the densityin
energy will vanish for e = — co, if that is a possible value of
the energy. Generally, however, the least possible value of
the energy will be a finite value, for which, if n > 2, e* will
vanish,* and therefore the densityinenergy. Now the density
inenergy is necessarily positive, and since it vanishes for
extreme values of the energy if n > 2, it must have a maxi
mum in such cases, in which the energy may be said to have
* See page 96.
THE FUNCTION 0. 101
its most common or most probable value, and which is
determined by the equation
d(f> 1
de ©* ^ '
This value of d(f>/de is also, when n > 2, its average value
in the ensemble. For we have identically, by integration by
parts,
'''=!+4>r~
v'=o v=o
If n > 2, the expression in the brackets, which multiplied by N
would represent the densityinenergy, vanishes at the limits,
and we have by (269) and (318)
It appears, therefore, that for systems of more tfyan two degrees
of freedom, the average value of d$/de in an eiis^ri^y canpni /
cally distributed is identical with the value of the same,
ential coefficient as calculated for the most .eoavrooi'. <
in the ensemble, both values being reciprocals of the modulus.
Hitherto, in our consideration of the quantities F", V# Vp, <£,
pi we have regarded the external coordinates as constant.
It is evident, however, from their definitions that V and <£ are
in general functions of the external coordinates and the energy
(e), that Vq and $g are in general functions of the external
coordinates and the potential energy (eg). Vp and p we have
found to be functions of the kinetic energy (ep) alone. In the
equation
/
de, (322)
by which vfr may be determined, O and the external coordinates
(contained implicitly in <£) are constant in the integration.
The equation shows that ir is a function of these constants.
102 TH& FUNCTION AND
If their values are varied, we shall have by differentiation, if
n >2
v=o
+ dai f*4. e~e+* 2, there are no terms due
to the variations of the limits.) Hence by (269)
or, since — ^ (325)
©
2. Moreover, since the external coordinates have
constant values throughout the ensemble, the values of
d(p/dav d(f>Jda^ etc. vary in the ensemble only on account
of the variations of the energy (e), which, as we have seen,
may be regarded as sensibly constant throughout the en
semble, when n is very great. In this case, therefore, we may
regard the average values
<25 ~d4
5S =S etc.,
104 THE FUNCTION <£ AND
as practically equivalent to the values relating to the most
common energy
— — I j ( — j j etc.
dtti JQ \ d&z J Q
In this case also de is practically equivalent to deQ. We have
therefore, for very large values of n,
— dri — dQ (337)
approximately. That is, except for an additive constant, — 77
may be regarded as practically equivalent to <£0, when the
number of degrees of freedom of the system is very great.
It is not meant by this that the variable part of rj + <£0 is
numerically of a lower order of magnitude than unity. For
when n is very great, — 77 and $0 are very great, and we can
only conclude that the variable part of 77 + <£0 is insignifi
cant compared with the variable part of rj or of <£0, taken
separately.
Now we have already noticed a certain correspondence
between the quantities ® and 77 and those which in thermo
dynamics are called temperature and entropy. The property
just demonstrated, with those expressed by equation (336),
therefore suggests that the quantities and de/dQ may also
correspond to the thermodynamic notions of entropy and tem
perature. We leave the discussion of this point to a sub
sequent chapter, and only mention it here to justify the
somewhat detailed investigation of the relations of these
quantities.
We may get a clearer view of the limiting form of the
relations when the number of degrees of freedom is indefi
nitely increased, if we expand the function in a series
arranged according to ascending powers of e — e0. This ex
pansion may be written
(f £)
(€ ~ ^
(338)
Adding the identical equation
THE CANONICAL DISTRIBUTION. 105
\/ — 6 ^ — €Q 6 —
© © © >
(339)
Substituting this value in
which expresses the probability that the energy of an unspeci
fied system of the ensemble lies between the limits e' and e",
we get
0
**. (340)
When the number of degrees of freedom is very great, and
e — e0 in consequence very small, we may neglect the higher
powers and write*
i .
" (341)
This shows that for a very great number of degrees of
freedom the probability of deviations of energy from the most
probable value (e0) approaches the form expressed by the
'law of errors.' With this approximate law, we get
* If a higher degree of accuracy is desired than is afforded by this formula,
it may be multiplied by the series obtained from
by the ordinary formula for the expansion in series of an exponential func
tion. There would be no especial analytical difficulty in taking account of
a moderate number of terms of such a series, which would commence
106 THE FUNCTION AND
(343)
whence
(344)
Now it has been proved in Chapter VII that
7  ^ _2
(6 ~~ e) — ~ ~r~ €P '
n dep p
We have therefore
approximately. The order of magnitude of rj — <£0 is there
fore that of log n. This magnitude is mainly constant.
The order of magnitude of rj + 0 , and therefore of — 77, is that
of n.*
Equation (338) gives for the first approximation
(1^ = _£, (346)
(**>(.0 = ^ = £*, W
/ . __ , Y — (6 ~ 6o)2 = ^ ^f (348)
a€p
The members of the last equation have the order of magnitude
of n. Equation (338) gives also for the first approximation
de fi\ ~ \ ^2 / v€ eo)>
* Compare (289), (314).
THE CANONICAL DISTRIBUTION. 107
whence
This is of the order of magnitude of n.*
It should be observed that the approximate distribution of
the ensemble in energy according to the 'law of errors' is
not dependent on the particular form of the function of the
energy which we have assumed for the index of probability
(77). In any case, we must have
(351)
where e^+t is necessarily positive. This requires that it
shall vanish for e = oo , and also for e = — oo , if this is a possi
ble value. It has been shown in the last chapter that if e has
a (finite) least possible value (which is the usual case) and
n > 2, e* will vanish for that least value of e. In general
therefore 77 + <£ will have a maximum, which determines the
most probable value of the energy. If we denote this value
by e0> and distinguish the corresponding values of the func
tions of the energy by the same suffix, we shall have
a
The probability that an unspecified system of the ensemble
* We shall find hereafter that the equation
is exact for any value of n greater than 2, and that the equation
fd(f> IV __ <^0
\d* ®) ' rf?
is exact for any value of n greater than 4.
108 THE FUNCTION <£ AND
falls within any given limits of energy (e' and e") is repre
sented by
f
e^de.
If we expand 77 and <£ in ascending powers of e — e0, without
going beyond the squares, the probability that the energy falls
within the given limits takes the form of the « law of errors ' —
de. (353)
i/
This gives
We shall have a close approximation in general when the
quantities equated in (355) are very small, i. e., when
is very great. Now when n is very great, — d*$/de* is of the
same order of magnitude, and the condition that (356) shall
be very great does not restrict very much the nature of the
function 77.
We may obtain other properties pertaining to average values
in a canonical ensemble by the method used for the average of
d/de. Let u be any function of the energy, either alone or
with ® and the external coordinates. The average value of u
in the ensemble is determined by the equation
6=00 4,e
/ —  + 4>
ue e de. (357)
F=0
THE CANONICAL DISTRIBUTION. 109
Now we have identically
Therefore, by the preceding equation
If we set u = 1, (a value which need not be excluded,) the
second member of this equation vanishes, as shown on page
101, if n > 2, and we get
^ = i, (360)
as before. It is evident from the same considerations that the
second member of (359) will always vanish if n > 2, unless u
becomes infinite at one of the limits, in which case a more care
ful examination of the value of the expression will be necessary.
To facilitate the discussion of such cases, it will be convenient
to introduce a certain limitation in regard to the nature of the
system considered. We have necessarily supposed, in all our
treatment of systems canonically distributed, that the system
considered was such as to be capable of the canonical distri
bution with the given value of the modulus. We shall now
suppose that the system is such as to be capable of a canonical
distribution with any (finite) f modulus. Let us see what
cases we exclude by this last limitation.
* A more general equation, which is not limited to ensembles canonically
distributed, is
^ + M^4.M^ \uef¥*~\*=*>
df U de U de ~ I"* J F=0
where t\ denotes, as usual, the index of probability of phase.
t The term finite applied to the modulus is intended to exclude the value
zero as well as infinity.
110 THE FUNCTION 0 AND
The impossibility of a canonical distribution occurs when
the equation
e e
e = e
s* — lj0
=J e ' de (361)
F=0
fails to determine a finite value for ^. Evidently the equation
cannot make ty an infinite positive quantity, the impossibility
therefore occurs when the equation makes ty = — oo . Now
we get easily from (191)
If the canonical distribution is possible for any values of ®,
we can apply this equation so long as the canonical distribu
tion is possible. The equation shows that as ® is increased
(without becoming infinite) — ty cannot become infinite unless
6 simultaneously becomes infinite, and that as O is decreased
(without becoming zero) — ^ cannot become infinite unless
simultaneously e becomes an infinite negative quantity. The
corresponding cases in thermodynamics would be bodies which
could absorb or give out an infinite amount of heat without
passing certain limits of temperature, when no external work
is done in the positive or negative sense. Such infinite values
present no analytical difficulties, and do not contradict the
general laws of mechanics or of thermodynamics, but they
are quite foreign to our ordinary experience of nature. In
excluding such cases (which are certainly not entirely devoid
of interest) we do not exclude any which are analogous to
any actual cases in thermodynamics.
We assume then that for any finite value of ® the second
member of (361) has a finite value.
When this condition is fulfilled, the second member of
(359) will vanish for u = e~+ V. For, if we set 6' = 26,
? ___! € _ f _ ^ ¥
F = 0 V = 0
THE CANONICAL DISTRIBUTION. Ill
where tyr denotes the value of ^ for the modulus ®'. Since
the last member of this formula vanishes for e = oo , the
less value represented by the first member must also vanish
for the same value of e. Therefore the second member of
(359), which differs only by a constant factor, vanishes at
the upper limit. The case of the lower limit remains to be
considered. Now
The second member of this formula evidently vanishes for
the value of e, which gives V — 0, whether this be finite or
negative infinity. Therefore, the second member of (359)
vanishes at the lower limit also, and we have
V
or e V=®. (362)
This equation, which is subject to no restriction in regard to
the value of n, suggests a connection or analogy between the
function of the energy of a system which is represented by
iT^ V and the notion of temperature in thermodynamics. We
shall return to this subject in Chapter XIV.
If n > 2, the second member of (359) may easily be shown
to vanish for any of the following values of u viz. : , e^, e,
e"*, where m denotes any positive number. It will also
vanish, when n > 4, for u = dfyde, and when n > 2 h for
u = e* dhV/d^. When the second member of (359) van
ishes, and n > 2, we may write
We thus obtain the following equations :
If n > 2,
(364)
112
THE FUNCTION AND
or
If w > 4,
If n
®2
dhVd 1 
6 d?fc®6
e ' Tjrj — e
de1 ae
or
(368)
t(369)
(370)
whence " ^ = ^.
Giving A the values 1, 2, 3, etc., we have
as already obtained. Also
* This equation may also be obtained from equations (252) and (321).
Compare also equation (349) which was derived by an approximative method,
t Compare equation (360), obtained by an approximative method.
THE CANONICAL DISTRIBUTION. 113
If Vq is a continuous increasing function of eg, commencing
with Vq = 0, the average value in a canonical ensemble of any
function of e^, either alone or with the modulus and the exter
nal coordinates, is given by equation (275), which is identical
with (357) except that e, $, and \jr have the suffix ( )ff. The
equation may be transformed so as to give an equation iden
tical with (359) except for the suffixes. If we add the same
suffixes to equation (361), the finite value of its members will
determine the possibility of the canonical distribution.
From these data, it is easy to derive equations similar to
(360), (362)(372), except that the conditions of their valid
ity must be differently stated. The equation
requires only the condition already mentioned with respect to
Vq. This equation corresponds to (362), which is subject to
no restriction with respect to the value of n. We may ob
serve, however, that V will always satisfy a condition similar
to that mentioned with respect to Vr
If Vq satisfies the condition mentioned, and e^ a similar
condition, i. e., if e^i is a continuous increasing function of e3,
commencing with the value (^ = 0, equations will hold sim
ilar to those given for the case when n > 2, viz., similar to
(360), (364)(368). Especially important is
deq ~®'
If Vq, 6*4 (or dVq/d€q), d?Vq/de* all satisfy similar conditions,
we shall have an equation similar to (369), which was subject
to the condition n > 4. And if cPVqjdef also satisfies a
similar condition, we shall have an equation similar to (372),
for which the condition was n > 6. Finally, if Vq and h suc
cessive differential coefficients satisfy conditions of the kind
mentioned, we shall have equations like (370) and (371) for
which the condition was n > 2 h.
8
114 THE FUNCTION <£.
These conditions take the place of those given above relat
ing to n. In fact, we might give conditions relating to the
differential coefficients of F", similar to those given relating to
the differential coefficients of Vq, instead of the conditions
relating to n, for the validity of equations (360), (363)(372).
This would somewhat extend the application of the equations.
CHAPTER X.
ON A DISTRIBUTION IN PHASE CALLED MICROCANONI
CAL IN WHICH ALL THE SYSTEMS HAVE
THE SAME ENERGY.
AN important case of statistical equilibrium is that in which
all systems of the ensemble have the same energy. We may
arrive at the notion of a distribution which will satisfy the
necessary conditions by the following process. We may
suppose that an ensemble is distributed with a uniform den
sityinphase between two limiting values of the energy, e' and
e", and with density zero outside of those limits. Such an
ensemble is evidently in statistical equilibrium according to
the criterion in Chapter IV, since the densityinphase may be
regarded as a function of the energy. By diminishing the
difference of e' and e", we may diminish the differences of
energy in the ensemble. The limit of this process gives us
a permanent distribution in which the energy is constant.
We should arrive at the same result, if we should make the
density any function of the energy between the limits e' and
e", and zero outside of those limits. Thus, the limiting distri
bution obtained from the part of a canonical ensemble
between two limits of energy, when the difference of the
limiting energies is indefinitely diminished, is independent of
the modulus, being determined entirely by the energy, and
is identical with the limiting distribution obtained from a
uniform density between limits of energy approaching the
same value.
We shall call the limiting distribution at which we arrive
by this process microcanonical.
We shall find however, in certain cases, that for certain
values of the energy, viz., for those for which e* is infinite,
116 A PERMANENT DISTRIBUTION IN WHICH
this process fails to define a limiting distribution in any such
distinct sense as for other values of the energy. The difficulty
is not in the process, but in the nature of the case, being
entirely analogous to that which we meet when we try to find
a canonical distribution in cases when ^ becomes infinite.
We have not regarded such cases as affording true examples
of the canonical distribution, and we shall not regard the cases
in which e^ is infinite as affording true examples of the micro
canonical distribution. We shall in fact find as we go on that
in such cases our most important formulae become illusory.
The use of formulae relating to a canonical ensemble which
contain e^de instead of dpl . . . dqn, as in the preceding chapters,
amounts to the consideration of the ensemble as divided into
an infinity of microcanonical elements;
From a certain point of view, the microcanonical distribution
may seem more simple than the canonical, and it has perhaps
been more studied, and been regarded as more closely related
to the fundamental notions of thermodynamics. To this last
point we shall return in a subsequent chapter. It is sufficient
here to remark that analytically the canonical distribution is
much more manageable than the microcanonical.
We may sometimes avoid difficulties which the microcanon
ical distribution presents by regarding it as the result of the
following process, which involves conceptions less simple but
more amenable to analytical treatment. We may suppose an
ensemble distributed with a density proportional to
where &> and e1 are constants, and then diminish indefinitely
the value of the constant &>. Here the density is nowhere
zero until we come to the limit, but at the limit it is zero for
all energies except e'. We thus avoid the analytical compli
cation of discontinuities in the value of the density, which
require the use of integrals with inconvenient limits.
In a microcanonical ensemble of systems the energy (e) is
constant, but the kinetic energy (e^) and the potential energy
ALL SYSTEMS HAVE THE SAME ENERGY. 117
(eq) vary in the different systems, subject of course to the con
dition
€p f eq = e = constant. (373)
Our first inquiries will relate to the division of energy into
these two parts, and to the average values of functions of ep
and eq.
We shall use the notation y\ 6 to denote an average value in
a microcanonical ensemble of energy e. An average value
in a canonical ensemble of modulus (D, which has hitherto
been denoted by M, we shall in this chapter denote by '^@, to
distinguish more clearly the two kinds of averages.
The extensioninphase within any limits which can be given
in terms of ep and eq may be expressed in the notations of the
preceding chapter by the double integral
*dVpdVq
taken within those limits. If an ensemble of systems is dis
tributed within those limits with a uniform densityinphase,
the average value in the ensemble of any function (u) of the
kinetic and potential energies will be expressed by the quotient
of integrals
/» r
udVpdVq
dVpdVq
Since d Vp = e^p dep, and dep = de when eq is constant, the
expression may be written
To get the average value of u in an ensemble distributed
microcanonically with the energy 6, we must make the in
tegrations cover the extensioninphase between the energies
e and e + de. This gives
118 A PERMANENT DISTRIBUTION IN WHICH
de\ueVpdVq
vq=o
But by (299) the value of the integral in the denominator
is e^. We have therefore
(374)
where e^p and Vq are connected by equation (373), and w, if
given as function of ep, or of ep and eq, becomes in virtue of
the same equation a function of eq alone.
We shall assume that e^ has a finite value. If n > 1, it is
evident from equation (305) that e^ is an increasing function
of e, and therefore cannot be infinite for one value of e without
being infinite for all greater values of e, which would make
— ty infinite.* When n > 1, therefore, if we assume that e^
is finite, we only exclude such cases as we found necessary
to exclude in the study of the canonical distribution. But
when n = 1, cases may occur in which the canonical distribu
tion is perfectly applicable, but in which the formulae for the
microcanonical distribution become illusory, for particular val
ues of e, on account of the infinite value of e^. Such failing
cases of the microcanonical distribution for particular values
of the energy will not prevent us from regarding the canon
ical ensemble as consisting of an infinity of microcanonical
ensembles, f
* See equation (322).
t An example of the failing case of the microcanonical distribution is
afforded by a material point, under the influence of gravity, and constrained
to remain in a vertical circle. The failing case occurs when the energy is
just sufficient to carry the material point to the highest point of the circle.
It will be observed that the difficulty is inherent in the nature of the case,
and is quite independent of the mathematical formulae. The nature of the
difficulty is at once apparent if we try to distribute a finite number of
ALL SYSTEMS HAVE THE SAME ENERGY. 119
From the last equation, with (298), we get
= e~* V. (375)
But by equations (288) and (289)
•V,?*. (376)
Therefore
e~* V— e~ P "Pjj e =  ep\e . (377)
Again, with the aid of equation (301), we get
= £» (378)
Vq=0
if n > 2. Therefore, by (289)
These results are interesting on account of the relations of
the functions e~$ V and ^ to the notion of temperature in
thermodynamics, — a subject to which we shall return here
after. They are particular cases of a general relation easily
deduced from equations (306), (374), (288) and (289). We
have
• ' ' r : , . w <
f*
=J
The equation may be written
€g=<
material points with this particular value of the energy as nearly as possible
in statistical equilibrium, or if we ask : What is the probability that a point
taken at random from an ensemble in statistical equilibrium with this value
of the energy will be found in any specified part of the circle?
120 A PERMANENT DISTRIBUTION IN WHICH
We have therefore
if h < J n + 1. For example, when w is even, we may make
A = i n, which gives, with (307),
12
(381)
Since any canonical ensemble of systems may be regarded
as composed of microcanonical ensembles, if any quantities
u and v have the same average values in every microcanonical
ensemble, they will have the same values in every canonical
ensemble. To bring equation (380) formally under this rule,
we may observe that the first member being a function of e is
a constant value in a microcanonical ensemble, and therefore
identical with its average value. We get thus the general
equation
.*£?
if h < J n + 1.* The equations
. 9 _
(383)
may be regarded as particular cases of the general equation.
The last equation is subject to the condition that n > 2.
The last two equations give for a canonical ensemble,
x if n > 2,
(l)^leV^]0l. (385)
The corresponding equations for a microcanonical ensemble
give, if n > 2,
\l 1 A 1 ' _1 ^V* /OQ£\
I1  = I W« V> = ^wTF' (386)
See equation (292).
ALL SYSTEMS HAVE THE SAME ENERGY. 121
which shows that d$ dlog V approaches the value unity
when n is very great.
If a system consists of two parts, having separate energies,
we may obtain equations similar in form to the preceding,
which relate to the system as thus divided.* We shall
distinguish quantities rekting to the parts by letters with
suffixes, the same letters without suffixes relating to the
whole system. The extensioninphase of the whole system
within any given limits of the energies may be represented by
the double integral
taken within those limits, as appears at once from the defini
tions of Chapter VIII. In an ensemble distributed with
uniform density within those limits, and zero density outside,
the average value of any function of e1 and ea is given by the
quotient
which may also be written f
If we make the limits of integration e and e + de, we get the
* If this condition is rigorously fulfilled, the parts will have no influence
on each other, and the ensemble formed by distributing the whole micro
canonically is too arbitrary a conception to have a real interest. The prin
cipal interest of the equations which we shall obtain will be in cases in
which the condition is approximately fulfilled. But for the purposes of a
theoretical discussion, it is of course convenient to make such a condition
absolute. Compare Chapter IV, pp. 35 ff., where a similar condition is con
sidered in connection with canonical ensembles.
t Where the analytical transformations are identical in form with those
on the preceding pages, it does not appear necessary to give all the steps
with the same detail.
122 A PERMANENT DISTRIBUTION IN WHICH
average value of u in an ensemble in which the whole system
is microcanonically distributed in phase, viz.,
(387)
where fa and V2 are connected by the equation
€i + €2 = constant = e, (388)
and u, if given as function of ei , or of ei and e2 , becomes in
virtue of the same equation a function of e2 alone.*
Thus
Je = e+ J F! rf F2 , (389)
(390)
This requires a similar relation for canonical averages
© = e~+ V\e = e^rje = e~+*V\* . (391)
Again
e2=e
SB =e*f ^V'rfF,. (392)
del e J del
F^O
But if w: > 2, «*i vanishes for Fj = 0,f and
. (393)
de
Hence, if n^ > 2, and w2 > 2,
d _ dfal _ dfa\ /qq ..
^e ~ ^ f ~ dez \f
* In the applications of the equation (387), we cannot obtain all the results
corresponding to those which we have obtained from equation (374), because
p is a known function of ep, while fa must be treated as an arbitrary func
tion of €j , or nearly so.
t See Chapter VIII, equations (306) and (316).
ALL SYSTEMS HAVE THE SAME ENERGY. 123
and s « 5l =^\ = ^ • (395)
© de 0 rfej J0 rfe2 e
We have compared certain functions of the energy of the
whole system with average values of similar functions of
the kinetic energy of the whole system, and with average
values of similar functions of the whole energy of a part of
the system. We may also compare the same functions with
average values of the kinetic energy of a part of the system.
We shall express the total, kinetic, and potential energies of
the whole system by e, ep, and eg, and the kinetic energies of the
parts by e^, and e2p. These kinetic energies are necessarily sep
arate : we need not make any supposition concerning potential
energies. The extensioninphase within any limits which can
be expressed in terms of eg, e^, ezp may be represented in the
notations of Chapter VIII by the triple integral
taken within those limits. And if an ensemble of systems is
distributed with a uniform density within those limits, the
average value of any function of eq, e^, e^ will be expressed
by the quotient
fffue^ded VZpd Vq
or
To get the average value of u for a microcanonical distribu
tion, we must make the limits e and e + de. The denominator
in this case becomes e^ de, and we have
C2p=C— Cq
(396)
124 A PERMANENT DISTRIBUTION IN WHICH
where 0^, V2P, and Vq are connected by the equation
€ip + €2p + eq = constant = e.
Accordingly
J VlpdV2p dVq = e* V, (397)
and we may write
;r 2 , 2 j /onON
2p6 = — e^l€ = €^e, (398)
and
O f)
r \ _ _ ^ — I __ ^ —  ('399')
Again, if wx > 2,
C9=€ (ft (ft
~* C^'jir "*"**
= e J ^dFi=« ir*
Hence, if ^ > 2, and w2 > 2,
_ 2p _ f i 1 N 1) _ /I w IN f 11
de ~de~l '* "~ '' p ^€ ~ ^ ~~ ' p '€
We cannot apply the methods employed in the preceding
pages to the microcanonical averages of the (generalized)
forces Av Ay, etc., exerted by a system on external bodies,
since these quantities are not functions of the energies, either
kinetic or potential, of the whole or any part of the system.
We may however use the method described on page 116.
ALL SYSTEMS HAVE THE SAME ENERGY. 125
Let us imagine an ensemble of systems distributed in phase
according to the index of probability
(e  c'V
where ef is any constant which is a possible value of the
energy, except only the least value which is consistent with
the values of the external coordinates, and c and o> are other
constants. We have therefore
all
c— •
e, w dpl . . . dqn — 1, (403)
phases
or e =...e dPl . . . dqn, (404)
phases
_c  g
or again e = C e ^ de. (405)
From (404) we have
all
phases
= 00
, j
^ (406)
where H7ie denotes the average value of A1 in those systems
of the ensemble which have any same energy e. (This
is the same thing as the average value of A l in a microcanoni
cal ensemble of energy e.) The validity of the transformation
is evident, if we consider separately the part of each integral
which lies between two infimtesimally differing limits of
energy. Integrating by parts, we get
126 A PERMANENT DISTRIBUTION IN WHICH
Jr=o
(*O,
•j . v — ' • • "j~Q>
F=0 ^ /
Differentiating (405), we get
€=00 (fO2 (*~O2
de* rdcj> —rf—+* _ /  ~~rf~+ dea\
T— = I £ e de—[e — ± }
^ da^ J dc^ \ ddij
where ea denotes the least value of e consistent with the exter
nal coordinates. The last term in this equation represents the
part of de~c jda^ which is due to the variation of the lower
limit of the integral. It is evident that the expression in the
brackets will vanish at the upper limit. At the lower limit,
at which ep = 0, and eq has the least value consistent with the
external coordinates, the average sign on ^]6 is superfluous,
as there is but one value of A1 which is represented by
— dea/dar Exceptions may indeed occur for particular values
of the external coordinates, at which dejda^ receive a finite
increment, and the formula becomes illusory. Such particular
values we may for the moment leave out of account. The
last term of (408) is therefore equal to the first term of the
second member of (407). (We may observe that both vanish
when n > 2 on account of the factor e$.)
We have therefore from these equations
F=0
or
That is : the average value in the ensemble of the quantity
represented by the principal parenthesis is zero. This must
ALL SYSTEMS HAVE THE SAME ENERGY. 127
be true for any value of «. If we diminish o>, the average
value of the parenthesis at the limit when « vanishes becomes
identical with the value for e = e'. But this may be any value
of the energy, except the least possible. We have therefore
unless it be for the least value of the energy consistent with
the external coordinates, or for particular values of the ex
ternal coordinates. But the value of any term of this equa
tion as determined for particular values of the energy and
of the external coordinates is not distinguishable from its
value as determined for values of the energy and external
coordinates indefinitely near those particular values. The
equation therefore holds without limitation. Multiplying
by e*, we get
= e==
The integral of this equation is
where Fl is a function of the external coordinates. We have
an equation of this form for each of the external coordinates.
This gives, with (266), for the complete value of the differen
tial of V
dV=e*de + (/Ale  ty da,, + (e+^kF^dat + etc., (413)
or
d V= £ (de + !ZTe dai + 3^]e daz + etc.) — Fldal — Fz daz — etc.
(414)
To determine the values of the functions Fl , Fz , etc., let
us suppose aL , «2 , etc. to vary arbitrarily, while e varies so
as always to have the least value consistent with the values
of the external coordinates. This will make V= 0, and
dV= 0. If 7i < 2, we shall have also e* = 0, which will
give
JF1 = 0, F2 = 0, etc. (415)
128 THE MICROCANONICAL DISTRIBUTION.
The result is the same for any value of n. For in the varia
tions considered the kinetic energy will be constantly zero,
and the potential energy will have the least value consistent
with the external coordinates. The condition of the least
possible potential energy may limit the ensemble at each in
stant to a single configuration, or it may not do so ; but in any
case the values of A1 , Av etc. will be the same at each instant
for all the systems of the ensemble,* and the equation
de + Al da^ f Az daz + etc. = 0
will hold for the variations considered. Hence the functions
F^ , F% , etc. vanish in any case, and we have the equation
d V= e*de + e* Z^d^ + e+~Z^dat + etc., (416)
de + ~A\,dal + Z^Lrfa2 + etc.
or dlogV=; _0 '6  (417)
or again
de = e~* V d log V  "27]€ dot  lje da2  etc. (418)
It will be observed that the two last equations have the form
of the fundamental differential equations of thermodynamics,
er^V corresponding to temperature and log V to entropy.
We have already observed properties of &"*> V suggestive of an
analogy with temperature, f The significance of these facts
will be discussed in another chapter.
The two last equations might be written more simply
de + 37€ dct! + Af€ daz + etc.
™ * — '  — 7  j
er4
de = e~^ d V — "37)€ da^ — ~A^\€ da2 — etc.,
and still have the form analogous to the thermodynamic
equations, but e~^ has nothing like the analogies with tempera
ture which we have observed in e~^ V.
* This statement, as mentioned before, may have exceptions for particular
values of the external coordinates. This will not invalidate the reasoning,
which has to do with varying values of the external coordinates.
t See Chapter IX, page 111 ; also this chapter, page 119.
CHAPTER XI.
MAXIMUM AND MINIMUM PROPERTIES OF VARIOUS DIS
TRIBUTIONS IN PHASE.
IN the following theorems we suppose, as always, that the
systems forming an ensemble are identical in nature and in
the values of the external coordinates, which are here regarded
as constants.
Theorem I. If an ensemble of systems is so distributed in
phase that the index of probability is a function of the energy,
the average value of the index is less than for any other distri
bution in which the distribution in energy is unaltered.
Let us write TJ for the index which is a function of the
energy, and 77 + A?? for any other which gives the same dis
tribution in energy. It is to be proved that
all all
J*. . . J* (i, + Ar,) e"**1 dPl... dqn >f. . . Jr? 6* dp,... dqn , (419)
pliases phases
where ?? is a function of the energy, and A?; a function of the
phase, which are subject to the conditions that
all all
J. . . Je^4" dp,... dqn = f. . . J> d&... dyn = 1, (420)
phases phases
and that for any value of the energy (e')
dp,... dqn =. . .fdpi ...dqn. (421)
Equation (420) expresses the general relations which 77 and
77 + AT; must satisfy in order to be indices of any distributions,
and (421) expresses the condition that they give the same
distribution in energy.
130 MAXIMUM AND MINIMUM PROPERTIES.
Since 77 is a function of the energy, and may therefore be re
garded as a constant within the limits of integration of (421),
we may multiply by T; under the integral sign hi both mem
bers, which gives
C
J.
71 dp^ . . . dqn.
U U \J
€=«' €— e'
Since this is true within the limits indicated, and for every
value of e', it will be true if the integrals are taken for all
phases. We may therefore cancel the corresponding parts of
(419), which gives
all
f A r, e1**11 dPl... dqn > 0. (422)
J
phases
But by (420) this is equivalent to
all
/. . . / (Ar;eAl7 + 1 — e^e'dpi . . . dqn > 0. (423)
tj
phases
Now AT; e^ + 1 — e^ is a decreasing function of AT; for nega
tive values of AT;, and an increasing function of AT; for positive
values of AT;. It vanishes for AT; = 0. The expression is
therefore incapable of a negative value, and can have the value
0 only for AT; = 0. The inequality (423) will hold therefore
unless AT; = 0 for all phases. The theorem is therefore
proved.
Theorem II. If an ensemble of systems is canonically dis
tributed in phase, the average index of probability is less than
in any other distribution of the ensemble having the same
average energy.
For the canonical distribution let the index be (^ — e) / ®,
and for another having the same average energy let the index
be (t/r — e)/0 + AT;, where AT; is an arbitrary function of the
phase subject only to the limitation involved in the notion of
the index, that
MAXIMUM AND MINIMUM PROPERTIES. 131
all itr— f a11 J'— €
/(* + AIJ r r —
. . .J e* dPl . . . dqn=J . . .J e & dPl . . . dqn = 1,
phases phases
(424)
and to that relating to the constant average energy, that
all — f all
J. . . Je e"^"4 * 4,, . . . <*?„ =J . . . Je e~e~ fe . . . <*?.. (425)
phases phases
It is to be proved that
phases
all
phases
Now in virtue of the first condition (424) we may cancel the
constant term ^r /® in the parentheses in (426), and in virtue
of the second condition (425) we may cancel the term e/O.
The proposition to be proved is thus reduced to
all ty~€
I A>7 e & dpi . . . dqn > 0,
phases
which may be written, in virtue of the condition (424),
all if/— e
f. . . f (Ar; eAl? + 1  /") e®~ dpi... dqn > 0. (427)
J J
phases
In this form its truth is evident for the same reasons which
applied to (423).
Theorem III. If ® is any positive constant, the average
value in an ensemble of the expression 77  e / 0 (77 denoting
as usual the index of probability and e the energy) is less when
the ensemble is distributed canonically with modulus ©, than
for any other distribution whatever.
In accordance with our usual notation let us write
(i/r — e) / ® for the index of the canonical distribution. In any
other distribution let the index be (>/r — e)/® + AT;.
132 MAXIMUM AND MINIMUM PROPERTIES.
In the canonical ensemble rj + e / © has the constant value
»r / • • Pni Of Wnicl1 ft • ' • ?i  in)***! . . . dqn > 0,
where the integrations cover all phases. Adding the equation
... 0. (442)
phases
Let
U = r1 — r}1 — r]2. (443)
The main proposition to be proved may be written
all
n > 0. (444)
phases
This is evidently true since the quantity in the parenthesis is
incapable of a negative value.* Moreover the sign = can
hold only when the quantity in the parenthesis vanishes for
all phases, i. e., when u = 0 for all phases. This makes
i) = tjl + ?72 for all phases, which is the analytical condition
which expresses that the distributions in phase of the two
parts of the system are independent.
Theorem VIII. If two or more ensembles of systems which
are identical in nature, but may be distributed differently in
phase, are united to form a single ensemble, so that the prob
abilitycoefficient of the resulting ensemble is a linear function
* See Theorem I, where this is proved of a similar expression.
136 MAXIMUM AND MINIMUM PROPERTIES.
of the probabilitycoefficients of the original ensembles, the
average index of probability of the resulting ensemble cannot
be greater than the same linear function of the average indices
of the original ensembles. It can be equal to it only when
the original ensembles are similarly distributed in phase.
Let PijP%, etc. be the probabilitycoefficients of the original
ensembles, and P that of the ensemble formed by combining
them ; and let N^ , ZV^ , etc. be the numbers of systems in the
original ensembles. It is evident that we shall have
P = elPl + c2P2 + etc. = 2 (cjPi), (445)
where Ci = =^V> c2 = —^, etc. (446)
The main proposition to be proved is that
all all
/• • ./P log PdPl . . . <*?„ ^ s pi/ • /P, log P, ^ . . . dfcTI
phases L phases •
(447)
all
f . . . f [2 (clPl log PO  P log P] dPl... dqn > 0. (448)
J J
or
J
phases
If we set
ft = P! log P!  P! log P  P! + P
Q1 will be positive, except when it vanishes for P1 = P. To
prove this, we may regard Pl and P as any positive quantities.
Then
\dPi*JP PI '
Since Q1 and dQ1/dP1 vanish for Pl — P, and the second
differential coefficient is always positive, Q1 must be positive
except when P1 = P. Therefore, if #2, etc. have similar
definitions,
2 fa ft) ^ 0. (449)
MAXIMUM AND MINIMUM PROPERTIES. 137
But since . 2 (cx Px) = P
and 2 dp, . . . dqn, (451)
where the integrations, like those which follow, are to be
taken within the given limits. The proposition to be proved
may be written
Pl... dqn > . . . ,; JdPl . . . dqn, (452)
or, since 77 is constant,
l ...dqn >. . .rjdp! . . . dqn. (453)
In (451) also we may cancel the constant factor e^, and multiply
by the constant factor (rj + 1). This gives
f. . .
The subtraction of this equation will not alter the inequality
to be proved, which may therefore be written
/. . ./(A,  1) /" dPl... dj. >/. . ./ cfc . . . dj.
138 MAXIMUM AND MINIMUM PROPERTIES.
f . . . f (AM eA"  /" + 1) dPl . . . dqn > 0. (454)
J J
or
Since the parenthesis in this expression represents a positive
value, except when it vanishes for AT; = 0, the integral will
be positive unless AT? vanishes everywhere within the limits,
which would make the difference of the two distributions
vanish. The theorem is therefore proved.
CHAPTER XII.
ON THE MOTION OF SYSTEMS AND ENSEMBLES OF SYS
TEMS THROUGH LONG PERIODS OF TIME.
AN important question which suggests itself in regard to any
case of dynamical motion is whether the system considered
will return in the course of time to its initial phase, or, if it
will not return exactly to that phase, whether it will do so to
any required degree of approximation in the course of a suffi
ciently long time. To be able to give even a partial answer
to such questions, we must know something in regard to the
dynamical nature of the system. In the following theorem,
the only assumption in this respect is such as we have found
necessary for the existence of the canonical distribution.
If we imagine an ensemble of identical systems to be
distributed with a uniform density throughout any finite
extensioninphase, the number of the systems which leave
the extensioninphase and will not return to it in the course
of time is less than any assignable fraction of the whole
number; provided, that the total extensioninphase for the
systems considered between two limiting values of the energy
is finite, these limiting values being less and greater respec
tively than any of the energies of the firstmentioned exten
sioninphase.
To prove this, we observe that at the moment which we
call initial the systems occupy the given extensioninphase.
It is evident that some systems must leave the extension
immediately, unless all remain in it forever. Those systems
which leave the extension at the first instant, we shall call
the front of the ensemble. It will be convenient to speak of
this front as generating the extensioninphase through which it
passes in the course of time, as in geometry a surface is said to
140 MOTION OF SYSTEMS AND ENSEMBLES
generate the volume through which it passes. In equal times
the front generates equal extensions in phase. This is an
immediate consequence of the principle of conservation of
extensioninphase^ unless indeed we prefer to consider it as
a slight variation in the expression of that principle. For in
two equal short intervals of time let the extensions generated
be A and B. (We make the intervals short simply to avoid
the complications in the enunciation or interpretation of the
principle which would arise when the same extensioninphase
is generated more than once in the interval considered.) Now
if we imagine that at a given instant systems are distributed
throughout the extension A, it is evident that the same
systems will after a certain tune occupy the extension B,
which is therefore equal to A in virtue of the principle cited.
The front of the ensemble, therefore, goes on generating
equal extensions in equal times. But these extensions are
included in a finite extension, viz., that bounded by certain
limiting values of the energy. Sooner or later, therefore,
the front must generate phases which it has before generated.
Such second generation of the same phases must commence
with the initial phases. Therefore a portion at least of the
front must return to the original extensioninphase. The
same is of course true of the portion of the ensemble which
follows that portion of the front through the same phases at
a later time.
It remains to consider how large the portion of the ensemble
is, which will return to the original extensioninphase. There
can be no portion of the given extensioninphase, the systems
of which leave the extension and do not return. For we can
prove for any portion of the extension as for the whole, that
at least a portion of the systems leaving it will return.
We may divide the given extensioninphase into parts as
follows. There may be parts such that the systems within
them will never pass out of them. These parts may indeed
constitute the whole of the given extension. But if the given
extension is very small, these parts will in general be non
existent. There may be parts such that systems within them
THROUGH LONG PERIODS OF TIME. 141
will all pass out of the given extension and all return within
it. The whole of the given extensioninphase is made up of
parts of these two kinds. This does not exclude the possi
bility of phases on the boundaries of such parts, such that
systems starting with those phases would leave the extension
and never return. But in the supposed distribution of an
ensemble of systems with a uniform densityinphase, such
systems would not constitute any assignable fraction of the
whole number.
These distinctions may be illustrated by a very simple
example. If we consider the motion of a rigid body of
which one point is fixed, and which is subject to no forces,
we find three cases. (1) The motion is periodic. (2) The
system will never return to its original phase, but will return
infinitely near to it. (3) The system will never return either
exactly or approximately to its original phase. But if we
consider any extensioninphase, however small, a system
leaving that extension will return to it except in the case
called by Poinsot * singular,' viz., when the motion is a
rotation about an axis lying in one of two planes having
a fixed position relative to the rigid body. But all such
phases do not constitute any true extensioninphase in the
sense in which we have defined and used the term.*
In the same way it may be proved that the systems in a
canonical ensemble which at a given instant are contained
within any finite extensioninphase will in general return to
* An ensemble of systems distributed in phase is a less simple and ele
mentary conception than a single system. But by the consideration of
suitable ensembles instead of single systems, we may get rid of the incon
venience of having to consider exceptions formed by particular cases of the
integral equations of motion, these cases simply disappearing when the
ensemble is substituted for the single system as a subject of study. This
is especially true when the ensemble is distributed, as in the case called
canonical, throughout an extensioninphase. In a less degree it is true of
the microcanonical ensemble, which does not occupy any extensioninphase,
(in the sense in which we have used the term,) although it is convenient to
regard it as a limiting case with respect to ensembles which do, as we thus
gain for the subject some part of the analytical simplicity which belongs to
the theory of ensembles which occupy true extensionsinphase.
142 MOTION OF SYSTEMS AND ENSEMBLES
that extensioninphase, if they leave it, the exceptions, i. g.,
the number which pass out of the extensioninphase and do
not return to it, being less than any assignable fraction of the
whole number. In other words, the probability that a system
taken at random from the part of a canonical ensemble which
is contained within any given extensioninphase, will pass out
of that extension and not return to it, is zero.
A similar theorem may be enunciated with respect to a
roicrocanonical ensemble. Let us consider the fractional part
of such an ensemble which lies within any given limits of
phase. This fraction we shall denote by F. It is evidently
constant in time since the ensemble is in statistical equi
librium. The systems within the limits will not in general
remain the same, but some will pass out in each unit of time
while an equal number come in. Some may pass out never
to return within the limits. But the number which in any
time however long pass out of the limits never to return will
not bear any finite ratio to the number within the limits at
a given instant. For, if it were otherwise, let / denote the
fraction representing such ratio for the tune T. Then, in
the time T, the number which pass out never to return will
bear the ratio f F to the whole number in the ensemble, and
in a time exceeding T/(fF) the number which pass out of
the limits never to return would exceed the total number
of systems in the ensemble. The proposition is therefore
proved.
This proof will apply to the cases before considered, and
may be regarded as more simple than that which was given.
It may also be applied to any true case of statistical equilib
rium. By a true case of statistical equilibrium is meant such
as may be described by giving the general value of the prob
ability that an unspecified system of the ensemble is con
tained within any given limits of phase.*
* An ensemble in which the systems are material points constrained to
move in vertical circles, with just enough energy to carry them to the
highest points, cannot afford a true example of statistical equilibrium. For
any other value of the energy than the critical value mentioned, we might
THROUGH LONG PERIODS OF TIME. 143
Let us next consider whether an ensemble of isolated
systems has any tendency in the course of time toward a
state of statistical equilibrium.
There are certain functions of phase which are constant in
time. The distribution of the ensemble with respect to the
values of these functions is necessarily invariable, that is,
the number of systems within any limits which can be
specified in terms of these functions cannot vary in the course
of time. The distribution in phase which without violating
this condition gives the least value of the average index of
probability of phase (77) is unique, and is that in which the
in various ways describe an ensemble in statistical equilibrium, while the
same language applied to the critical value of the energy would fail to do
so. Thus, if we should say that the ensemble is so distributed that the
probability that a system is in any given part of the circle is proportioned
to the time which a single system spends in that part, motion in either direc
tion being equally probable, we should perfectly define a distribution in sta
tistical equilibrium for any value of the energy except the critical value
mentioned above, but for this value of the energy all the probabilities in
question would vanish unless the highest point is included in the part of the
circle considered, in which case the probability is unity, or forms one of its
limits, in which case the probability is indeterminate. Compare the footnote
on page 118.
A still more simple example is afforded by the uniform motion of a
material point in a straight line. Here the impossibility of statistical equi
librium is not limited to any particular energy, and the canonical distribu
tion as well as the microcanonical is impossible.
These examples are mentioned here in order to show the necessity of
caution in the application of the above principle, with respect to the question
whether we have to do with a true case of statistical equilibrium.
Another point in respect to which caution must be exercised is that the
part of an ensemble of which the theorem of the return of systems is asserted
should be entirely denned by limits within which it is contained, and not by
any such condition as that a certain function of phase shall have a given
value. This is necessary in order that the part of the ensemble which is
considered should be any assignable fraction of the whole. Thus, if we have
a canonical ensemble consisting of material points in vertical circles, the
theorem of the return of systems may be applied to a part of the ensemble
defined as cqntained in a given part of the circle. But it may not be applied
in all cases to a part of the ensemble defined as contained in a given part
of the circle and having a given energy. It would, in fact, express the exact
opposite of the truth when the given energy is the critical value mentioned
above.
144 MOTION OF SYSTEMS AND ENSEMBLES
index of probability (77) is a function of the functions men
tioned.* It is therefore a permanent distribution, f and the
only permanent distribution consistent with the invariability
of the distribution with respect to the functions of phase
which are constant in time.
It would seem, therefore, that we might find a sort of meas
ure of the deviation of an ensemble from statistical equilibrium
in the excess of the average index above the minimum which is
consistent with the condition of the invariability of the distri
bution with respect to the constant functions of phase. But
we have seen that the index of probability is constant in time
for each system of the ensemble. The average index is there
fore constant, and we find by this method no approach toward
statistical equilibrium in the course of time.
Yet we must here exercise great caution. One function
may approach indefinitely near to another function, while
some quantity determined by the first does not approach the
corresponding quantity determined by the second. A line
joining two points may approach indefinitely near to the
straight line joining them, while its length remains constant.
We may find a closer analogy with the case under considera
tion in the effect of stirring an incompressible liquid.^ In
space of 2 n dimensions the case might be made analyti
cally identical with that of an ensemble of systems of n
degrees of freedom, but the analogy is perfect in ordinary
space. Let us suppose the liquid to contain a certain amount
of coloring matter which does not affect its hydrodynamic
properties. Now the state in which the density of the coloring
matter is uniform, i. e., the statt, of perfect mixture, which is
a sort of state of equilibrium in this respect that the distribu
tion of the coloring matter in space is not affected by the
internal motions of the liquid, is characterized by a minimum
* See Chapter XI, Theorem IV.
t See Chapter IV, sub init.
J By liquid is here meant the continuous body of theoretical hydrody
namics, and not anything of the molecular structure and molecular motions
of real liquids.
THROUGH LONG PERIODS OF TIME. 145
value of the average square of the density of the coloring
matter. Let us suppose, however, that the coloring matter is
distributed with a variable density. If we give the liquid any
motion whatever, subject only to the hydrodynamic law of
incompressibility, — it may be a steady flux, or it may vary
with the time, — the density of the coloring matter at any
same point of the liquid will be unchanged, and the average
square of this density will therefore be unchanged. Yet no
fact is more familiar to us than that stirring tends to bring a
liquid to a state of uniform mixture, or uniform densities of
its components, which is characterized by minimum values
of the average squares of these densities. It is quite true that
in the physical experiment the result is hastened by the
process of diffusion, but the result is evidently not dependent
on that process.
The contradiction is to be traced to the notion of the density
of the coloring matter, and the process by which this quantity
is evaluated. This quantity is the limiting ratio of the
quantity of the coloring matter in an element of space to the
volume of that element. Now if we should take for our ele
ments of volume, after any amount of stirring, the spaces
occupied by the same portions of the liquid which originally
occupied any given system of elements of volume, the densi
ties of the coloring matter, thus estimated, would be identical
with the original densities as determined by the given system
of elements of volume. Moreover, if at the end of any finite
amount of stirring we should take our elements of volume in
any ordinary form but sufficiently small, the. average square
of the density of the coloring matter, as determined by such
element of volume, would approximate to any required degree
to its value before the stirring. But if we take any element
of space of fixed position and dimensions, we may continue
the stirring so long that the densities of the colored liquid
estimated for these fixed elements will approach a uniform
limit, viz.', that of perfect mixture.
The case is evidently one of those in which the limit of a
limit has different values, according to the order in which we
10
146 MOTION OF SYSTEMS AND ENSEMBLES
apply the processes of taking a limit. If treating the elements
of volume as constant, we continue the stirring indefinitely,
we get a uniform density, a result not affected by making the
elements as small as we choose ; but if treating the amount of
stirring as finite, we diminish indefinitely the elements of
volume, we get exactly the same distribution in density as
before the stirring, a result which is not affected by con
tinuing the stirring as long as we choose. The question is
largely one of language and definition. One may perhaps be
allowed to say that a finite amount of stirring will not affect
the mean square of the density of the coloring matter, but an
infinite amount of stirring may be regarded as producing a
condition in which the mean square of the density has its
minimum value, and the density is uniform. We may cer
tainly say that a sensibly uniform density of the colored com
ponent may be produced by stirring. Whether the time
required for this result would be long or short depends upon
the nature of the motion given to the liquid, and the fineness
of our method of evaluating the density.
All this may appear more distinctly if we consider a special
case of liquid motion. Let us imagine a cylindrical mass of
liquid of which one sector of 90° is black and the rest white.
Let it have a motion of rotation about the axis of the cylinder
in which the angular velocity is a function of the distance
from the axis. In the course of time the black and the white
parts would become drawn out into thin ribbons, which would
be wound spirally about the axis. The thickness of these rib
bons would diminish without limit, and the liquid would there
fore tend toward a state of perfect mixture of the black and
white portions. That is, in any given element of space, the
proportion of the black and white would approach 1 : 3 as a limit.
Yet after any finite time, the total volume would be divided
into two parts, one of which would consist of the white liquid
exclusively, and the other of the black exclusively. If the
coloring matter, instead of being distributed initially with a
uniform density throughout a section of the cylinder, were
distributed with a density represented by any arbitrary func
THROUGH LONG PERIODS OF TIME. 147
tion of the cylindrical coordinates r, 6 and 2, the effect of the
same motion continued indefinitely would be an approach to
a condition in which the density is a function of r and z alone.
In this limiting condition, the average square of the density
would be less than in the original condition, when the density
was supposed to vary with 0, although after any finite time
the average square of the density would be the same as at
first.
If we limit our attention to the motion in a single plane
perpendicular to the axis of the cylinder, we have something
which is almost identical with a diagrammatic representation
of the changes in distribution in phase of an ensemble of
systems of one degree of freedom, in which the motion is
periodic, the period varying with the energy, as in the case of
a pendulum swinging in a circular arc. If the coordinates
and momenta of the systems are represented by rectangu
lar coordinates in the diagram, the points in the diagram
representing the changing phases of moving systems, will
move about the origin in closed curves of constant energy.
The motion will be such that areas bounded by points repre
senting moving systems will be preserved. The only differ
ence between the motion of the liquid and the motion in the
diagram is that in one case the paths are circular, and in the
other they differ more or less from that form.
When the energy is proportional to p2 + q2 the curves of
constant energy are circles, and the period is independent of
the energy. There is then no tendency toward a state of sta
tistical equilibrium. The diagram turns about the origin with
out change of form. This corresponds to the case of liquid
motion, when the liquid revolves with a uniform angular
velocity like a rigid solid.
The analogy between the motion of an ensemble of systems
in an extensioninphase and a steady current in an incompres
sible liquid, and the diagrammatic representation of the case
of one degree of freedom, which appeals to our geometrical in
tuitions, may be sufficient to show how the conservation of
density in phase, which involves the conservation of the
148 MOTION OF SYSTEMS AND ENSEMBLES
average value of the index of probability of phase, is consist
ent with an approach to a limiting condition in which that
average value is less. We might perhaps fairly infer from
such considerations as have been adduced that an approach
to a limiting condition of statistical equilibrium is the general
rule, when the initial condition is not of that character. But
the subject is of such importance that it seems desirable to
give it farther consideration.
Let us suppose that the total extensioninphase for the
kind of system considered to be divided into equal elements
(D V) which are very small but not infinitely small. Let us
imagine an ensemble of systems distributed in this extension
in a manner represented by the index of probability 77, which
is an arbitrary function of the phase subject only to the re
striction expressed by equation (46) of Chapter I. We shall
suppose the elements D V to be so small that rj may in gen
eral be regarded as sensibly constant within any one of them
at the initial moment. Let the path of a system be defined as
the series of phases through which it passes.
At the initial moment (£') a certain system is in an element
of extension DVf. Subsequently, at the time £", the same
system is in the element DV". Other systems which were
at first in DV will at the time t" be in DV", but not all,
probably. The systems which were at first in DV1 will at
the time t'f occupy an extensioninphase exactly as large as at
first. But it will probably be distributed among a very great
number of the elements (DV) into which we have divided
the total extensioninphase. If it is not so, we can generally
take a later time at which it will be so. There will be excep
tions to this for particular laws of motion, but we will con
fine ourselves to what may fairly be called the general case.
Only a very small part of the systems initially in D V will
be found in DV" at the time t", and those which are found in
DV" at that time were at the initial moment distributed
among a very large number of elements D V.
What is important for our purpose is the value of 77, the
index of probability of phase in the element DV" at the time
THROUGH LONG PERIODS OF TIME. 149
t". In the part of DV" occupied by systems which at the
time if were in DV the value of 77 will be the same as its
value in D V at the time tr, which we shall call 77'. In the
parts of DV" occupied by systems which at if were in ele
ments very near to D V we may suppose the value of 77 to
vary little from T?'. We cannot assume this in regard to parts
of DV" occupied by systems which at tf were in elements
remote from DV. We want, therefore, some idea of the
nature of the extensioninphase occupied at tf by the sys
tems which at t" will occupy D V". Analytically, the prob
lem is identical with finding the extension occupied at t"
by the systems which at t1 occupied DV. Now the systems
in D V" which lie on the same path as the system first con
sidered, evidently arrived at DV" at nearly the same time,
and must have left D V1 at nearly the same time, and there
fore at if were in or near DV. We may therefore take T/ as
the value for these systems. The same essentially is true of
systems in DV" which he on paths very close to the path
already considered. But with respect to paths passing through
D V and D V", but not so close to the first path, we cannot
assume that the time required to pass from DV to D V" is
nearly the same as for the first path. The difference of the
times required may be small in comparison with £"£', but as
this interval can be as large as we choose, the difference of the
times required in the different paths has no limit to its pos
sible value. Now if the case were one of statistical equilib
rium, the value of 77 would be constant in any path, and if all
the paths which pass through DV1 also pass through or near
D V, the value of 77 throughout D V" will vary little from
?;'. But when the case is not one of statistical equilibrium,
we cannot draw any such conclusion. The only conclusion
which we can draw with respect to the phase at t1 of the sys
tems which at t" are in DV" is that they are nearly on the
same patji.
Now if we should make a new estimate of indices of prob
ability of phase at the time t", using for this purpose the
elements D V, — that is, if we should divide the number of
150 MOTION OF SYSTEMS AND ENSEMBLES
systems in JDF", for example, by the total number of systems,
and also by the extensioninphase of the element, and take
the logarithm of the quotient, we would get a number which
would be less than the average value of rj for the systems
within D V" based on the distribution in phase at the time t1.*
Hence the average value of 77 for the whole ensemble of
systems based on the distribution at t" will be less than the
average value based on the distribution at t'.
We must not forget that there are exceptions to this gen
eral rule. These exceptions are in cases in which the laws
of motion are such that systems having small differences
of phase will continue always to have small differences of
phase.
It is to be observed that if the average index of probability in
an ensemble may be said in some sense to have a less value at
one tune than at another, it is not necessarily priority in tune
which determines the greater average index. If a distribution,
which is not one of statistical equilibrium, should be given
for a time £', and the distribution at an earlier time t" should
be defined as that given by the corresponding phases, if we
increase the interval leaving t' fixed and taking ttf at an earlier
and earlier date, the distribution at t" will in general approach
a limiting distribution which is in statistical equilibrium. The
determining difference in such cases is that between a definite
distribution at a definite time and the limit of a varying dis
tribution when the moment considered is carried either forward
or backward indefinitely, f
But while the distinction of prior and subsequent events
may be immaterial with respect to mathematical fictions, it is
quite otherwise with respect to the events of the real world.
It should not be forgotten, when our ensembles are chosen to
illustrate the probabilities of events in the real world, that
* See Chapter XI, Theorem IX.
t One may compare the kinematical truism that when two points are
moving with uniform velocities, (with the single exception of the case where
the relative motion is zero,) their mutual distance at any definite time is less
than f or t = , or t = — oo .
THROUGH LONG PERIODS OF TIME. 151
while the probabilities of subsequent events may often be
determined from the probabilities of prior events, it is rarely
the case that probabilities of prior events can be determined \j
from those of subsequent events, for we are rarely justified in
excluding the consideration of the antecedent probability of
the prior events.
It is worthy of notice that to take a system at random from
an ensemble at a date chosen at random from several given
dates, t', t", etc., is practically the same thing as to take a sys
tem at random from the ensemble composed of all the systems
of the given ensemble in their phases at the time £', together
with the same systems in their phases at the time t/;, etc. By
Theorem VIII of Chapter XI this will give an ensemble in
which the average index of probability will be less than in
the given ensemble, except in the case when the distribution
in the given ensemble is the same at the times tr, t'f, etc.
Consequently, any indefiniteness in the time in which we take
a system at random from an ensemble has the practical effect
of diminishing the average index of the ensemble from which
the system may be supposed to be drawn, except when the
given ensemble is in statistical equilibrium.
CHAPTER XIII.
EFFECT OF VARIOUS PROCESSES ON AN ENSEMBLE OF
SYSTEMS.
IN the last chapter and in Chapter I we have considered the
changes which take place in the course of time in an ensemble
of isolated systems. Let us now proceed to consider the
changes which will take place in an ensemble of systems under
external influences. These external influences will be of two
kinds, the variation of the coordinates which we have called
external, and the action of other ensembles of systems. The
essential difference of the two kinds of influence consists in
this, that the bodies to which the external coordinates relate
are not distributed in phase, while in the case of interaction
of the systems of two ensembles, we have to regard the fact
that both are distributed in phase. To find the effect pro
duced on the ensemble with which we are principally con
cerned, we have therefore to consider single values of what
we have called external coordinates, but an infinity of values
of the internal coordinates of any other ensemble with which
there is interaction.
Or, — to regard the subject from another point of view, —
the action between an unspecified system of an ensemble and
the bodies represented by the external coordinates, is the
action between a system imperfectly determined with respect
to phase and one which is perfectly determined ; while the
interaction between two unspecified systems belonging to
different ensembles is the action between two systems both of
which are imperfectly determined with respect to phase.*
We shall suppose the ensembles which we consider to be
distributed in phase in the manner described in Chapter I, and
* In the development of the subject, we shall find that this distinction
corresponds to the distinction in thermodynamics between mechanical and
thermal action.
EFFECT OF VARIOUS PROCESSES. 153
represented by the notations of that chapter, especially by the
index of probability of phase (??). There are therefore 2 n
independent variations in the phases which constitute the
ensembles considered. This excludes ensembles like the
microcanonical, in which, as energy is constant, there are
only 2 n — 1 independent variations of phase. This seems
necessary for the purposes of a general discussion. For
although we may imagine a microcanonical ensemble to have
a permanent existence when isolated from external influences,
the effect of such influences would generally be to destroy the
uniformity of energy in the ensemble. Moreover, since the
microcanonical ensemble may be regarded as a limiting case of
such ensembles as are described in Chapter I, (and that in
more than one way, as shown in Chapter X,) the exclusion is
rather formal than real, since any properties which belong to
the microcanonical ensemble could easily be derived from those
of the ensembles of Chapter I, which in a certain sense may
be regarded as representing the general case.
Let us first consider the effect of variation of the external
coordinates. We have already had occasion to regard these
quantities as variable in the differentiation of certain equations
relating to ensembles distributed according to certain laws
called canonical or microcanonical. That variation of the
external coordinates was, however, only carrying the atten
tion of the mind from an ensemble with certain values of the
external coordinates, and distributed in phase according to
some general law depending upon those values, to another
ensemble with different values of the external coordinates, and
with the distribution changed to conform to these new values.
What we have now to consider is the effect which would
actually result in the course of time in an ensemble of systems
in which the external coordinates should be varied in any
arbitrary manner. Let us suppose, in the first place, that
these coordinates are varied abruptly at a given instant, being
constant both before and after that instant. By the definition
of the external coordinates it appears that this variation does
not affect the phase of any system of the ensemble at the time
154 EFFECT OF VARIOUS PROCESSES
when it takes place. Therefore it does not affect the index of
probability of phase (77) of any system, or the average value
of the index (?/)' at that time. And if these quantities are
constant in time before the variation of the external coordi
nates, and after that variation, their constancy hi time is not
interrupted by that variation. In fact, in the demonstration
of the conservation of probability of phase in Chapter I, the
variation of the external coordinates was not excluded.
But a variation of the external coordinates will in general
disturb a previously existing state of statistical equilibrium.
For, although it does not affect (at the first instant) the
distributioninphase, it does affect the condition necessary for
equilibrium. This condition, as we have seen in Chapter IV,
is that the index of probability of phase shall be a function of
phase which is constant in time for moving systems. Now
a change in the external coordinates, by changing the forces
which act on the systems, will change the nature of the
functions of phase which are constant in time. Therefore,
the distribution in phase which was one of statistical equi
librium for the old values of the external coordinates, will not
be such for the new values.
Now we have seen, in the last chapter, that when the dis
tributioninphase is not one of statistical equilibrium, an
ensemble of systems may, and in general will, after a longer or
shorter time, come to a state which may be regarded, if very
small differences of phase are neglected, as one of statistical
equilibrium, and in which consequently the average value of
the index (?;) is less than at first. It is evident, therefore,
that a variation of the external coordinates, by disturbing a
state of statistical equilibrium, may indirectly cause a diminu
tion, (in a certain sense at least,) of the value of rj.
But if the change in the external coordinates is very small,
the change in the distribution necessary for equilibrium will
in general be correspondingly small. Hence, the original dis
tribution in phase, since it differs little from one which would
be in statistical equilibrium with the new values of the ex
ternal coordinates, may be supposed to have a value of v
ON AN ENSEMBLE OF SYSTEMS. 155
which differs by a small quantity of the second order from
the minimum value which characterizes the state of statistical
equilibrium. And the diminution in the average index result
ing in the course of time from the very small change in the
external coordinates, cannot exceed this small quantity of
the second order.
Hence also, if the change in the external coordinates of an
ensemble initially in statistical equilibrium consists in suc
cessive very small changes separated by very long intervals of
time in which the disturbance of statistical equilibrium be
comes sensibly effaced, the final diminution in the average
index of probability will in general be negligible, although the
total change in the external coordinates is large. The result
will be the same if the change in the external coordinates
takes place continuously but sufficiently slowly.
Even in cases in which there is no tendency toward the
restoration of statistical equilibrium in the lapse of time, a varia
tion of external coordinates which would cause, if it took
place in a short time, a great disturbance of a previous state
of equilibrium, may, if sufficiently distributed in time, produce
no sensible disturbance of the statistical equilibrium.
Thus, in the case of three degrees of freedom, let the systems
be heavy points suspended by elastic massless cords, and let the
ensemble be distributed in phase with a density proportioned
to some function of the energy, and therefore in statistical equi
librium. For a change in the external coordinates, we may
take a horizontal motion of the point of suspension. If this
is moved a given distance, the resulting disturbance of the
statistical equilibrium may evidently be diminished indefi
nitely by diminishing the velocity of the point of suspension.
This will be true if the law of elasticity of the string is such
that the period of vibration is independent of the energy, in
which case there is no tendency in the course of time toward
a state of statistical equilibrium, as well as in the more general
case, in which there is a tendency toward statistical equilibrium.
That something of this kind will be true in general, the
following considerations will tend to show.
156 EFFECT OF VARIOUS PROCESSES
We define a path as the series of phases through which a
system passes in the course of time when the external co
ordinates have fixed values. When the external coordinates
are varied, paths are changed. The path of a phase is the
path to which that phase belongs. With reference to any
ensemble of systems we shall denote by 27p the average value
of the densityinphase in a path. This implies that we have
a measure for comparing different portions of the path. We
shall suppose the time required to traverse any portion of a
path to be its measure for the purpose of determining this
average.
With this understanding, let us suppose that a certain en
semble is in statistical equilibrium. In every element of
extensioninphase, therefore, the densityinphase D is equal
to its pathaverage 27]p. Let a sudden small change be made
in the external coordinates. The statistical equilibrium will be
disturbed and we shall no longer have D — ~D\P everywhere.
This is not because D is changed, but because ~D\p is changed,
the paths being changed. It is evident that if D > I)]p in
a part of a path, we shall have D < ~D\p in other parts of the
same path.
Now, if we should imagine a further change in the external
coordinates of the same kind, we should expect it to produce
an effect of the same kind. But the manner in which the
second effect will be superposed on the first will be different,
according as it occurs immediately after the first change or
after an interval of time. If it occurs immediately after the
first change, then in any element of phase in which the first
change produced a positive value of D  2JP the second change
will add a positive value to the first positive value, and where
D  1)\p was negative, the second change will add a negative
value to the first negative value.
But if we wait a sufficient time before making the second
change in the external coordinates, so that systems have
passed from elements of phase in which D  ~D\P was origi
nally positive to elements in which it was originally negative,
and vice versa, (the systems carrying with them the values
ON AN ENSEMBLE OF SYSTEMS. 157
of D  1J\p ,) the positive values of D  U\p caused by the
second change will be in part superposed on negative values
due to the first change, and vice versa.
The disturbance of statistical equilibrium, therefore, pro
duced by a given change in the values of the external co
ordinates may be very much diminished by dividing the
change into two parts separated by a sufficient interval of
tune, and a sufficient interval of time for this purpose is one
in which the phases of the individual systems are entirely
unlike the first, so that any individual system is differently
affected by the change, although the whole ensemble is af
fected in nearly the same way. Since there is no limit to the
diminution of the disturbance of equilibrium by division of
the change in the external coordinates, we may suppose as
a general rule that by diminishing the velocity of the changes
in the external coordinates, a given change may be made to
produce a very small disturbance of statistical equilibrium.
If we write r[ for the value of the average index of probability
before the variation of the external coordinates, and iff' for the
value after this variation, we shall have in any case
as the simple result of the variation of the external coordi
nates. This may be compared with the thermodynamic the
orem that the entropy of a body cannot be diminished by
mechanical (as distinguished from thermal) action.*
If we have (approximate) statistical equilibrium between
the times if and if' (corresponding to rf and ??"), we shall have
approximately
which may be compared with the thermodynamic theorem that
the entropy of a body is not (sensibly) affected by mechanical
action, during which the body is at each instant (sensibly) in
a state of thermodynamic equilibrium.
Approximate statistical equilibrium may usually be attained
* The correspondences to which the reader's attention is called are between
— t\ and entropy, and between 0 and temperature.
158 EFFECT OF VARIOUS PROCESSES
by a sufficiently slow variation of the external coordinates,
just as approximate thermodynamic equilibrium may usually
be attained by sufficient slowness in the mechanical operations
to which the body is subject.
We now pass to the consideration of the effect on an en
semble of systems which is produced by the action of other
ensembles with which it is brought into dynamical connec
tion. In a previous chapter * we have imagined a dynamical
connection arbitrarily created between the systems of two
ensembles. We shall now regard the action between the
systems of the two ensembles as a result of the variation
of the external coordinates, which causes such variations
of the internal coordinates as to bring the systems of the
two ensembles within the range of each other's action.
Initially, we suppose that we have two separate ensembles
of systems, E± and Ez. The numbers of degrees of freedom
of the systems in the two ensembles will be denoted by n^ and
n2 respectively, and the probabilitycoefficients by e^ and e"*,
Now we may regard any system of the first ensemble com
bined with any system of the second as forming a single
system of ^ + nz degrees of freedom. Let us consider the
ensemble ( J?12) obtained by thus combining each system of the
first ensemble with each of the second.
At the initial moment, which may be specified by a single
accent, the probabilitycoefficient of any phase of the combined
systems is evidently the product of the probabilitycoefficients
of the phases of which it is made up. This may be expressed
by the equation,
ew = 6V ev, (455)
or n* = in' + ^ (456)
which gives r^z = ij/ + iya' (457)
The forces tending to vary the internal coordinates of the
combined systems, together with those exerted by either
system upon the bodies represented by the coordinates called
* See Chapter IV, page 37.
ON AN ENSEMBLE OF SYSTEMS. 159
external, may be derived from a single forcefunction, which,
taken negatively, we shall call the potential energy of the
combined systems and denote by e12. But we suppose that
initially none of the systems of the two ensembles EI and
E% come within range of each other's action, so that the
potential energy of the combined system falls into two parts
relating separately to the systems which are combined. The
same is obviously true of the kinetic energy of the combined
compound system, and therefore of its total energy. This
may be expressed by the equation
€„'=€/ + €,', (458)
which gives e12' = i/ + e2'. (459)
Let us now suppose that in the course of tune, owing to the
motion of the bodies represented by the coordinates called
external, the forces acting on the systems and consequently
their positions are so altered, that the systems of the ensembles
El and E% are brought within range of each other's action,
and after such mutual influence has lasted for a time, by a
further change in the external coordinates, perhaps a return
to their original values, the systems of the two original en
sembles are brought again out of range of each other's action.
Finally, then, at a time specified by double accents, we shall
have as at first
€«" = e/' + ia". (460)
But for the indices of probability we must write *
W + W ^ W' (461)
The considerations adduced in the last chapter show that it
is safe to write
W 5 W (462)
We have therefore
5i" + i" < ^ + i', (463)
which may be compared with the thermodynamic theorem that
* See Chapter XI, Theorem VII.
160 EFFECT OF VARIOUS PROCESSES
the thermal contact of two bodies may increase but cannot
diminish the sum of their entropies.
Let us especially consider the case in which the two original
ensembles were both canonically distributed in phase with the
respective moduli ®j and ©2. We have then, by Theorem III
of Chapter XI,
nJ + ^ < r?i" +  (464)
^' + !'<^" + ^ (465)
Whence with (463) we have
_
If we write W for the average work done by the combined
systems on the external bodies, we have by the principle of
the conservation of energy
W = €„'  €M" = €/  €X" + e2'  e2". (468)
Now if TFis negligible, we have
e/' _ e/ =  (e""  €?) (469)
and (467) shows that the ensemble which has the greater
modulus must lose energy. This result may be compared to
the thermodynamic principle, that when two bodies of differ
ent temperatures are brought together, that which has the
higher temperature will lose energy.
Let us next suppose that the ensemble E% is originally
canonically distributed with the modulus @2 , but leave the
distribution of the other arbitrary. We have, to determine
the result of a similar process,
ON AN ENSEMBLE OF SYSTEMS. 161
Hence ^" + '=^' + C (470)
which may be written
%'V^^^ (471)
This may be compared with the thennodynamic principle that
when a body (which need not be in thermal equilibrium) is
brought into thermal contact with another of a given tempera
ture, the increase of entropy of the first cannot be less (alge
braically) than the loss_of heat by the second divided by its
temperature. Where W is negligible, we may write
V' + ^' +  . (472)
Now, by Theorem III of Chapter XI, the quantity
! . * +  (473)
has a minimum value when the ensemble to which ^ and ex
relate is distributed canonically with the modulus ®2. If the
ensemble had originally this distribution, the sign < in (472)
would be impossible. In fact, in this case, it would be easy to
show that the preceding formulae on which (472) is founded
would all have the sign = . But when the two ensembles are
not both originally distributed canonically with the same
modulus, the formulae indicate that the quantity (473) may
be diminished by bringing the ensemble to which ea and yl
relate into connection with another which is canonically dis
tributed with modulus ®2, and therefore, that by repeated
operations of this kind the ensemble of which the original dis
tribution was entirely arbitrary might be brought approxi
mately into a state of canonical distribution with the modulus
^0" + i" + ^2" + etc. (474)
But by Theorem III of Chapter XI,
ft" + Z ft1 + ' (476)
*" + g > .7 + {£ (477)
etc.
or, since €
we shall not have in general
deiz del dez
dlog F12 ~~ dlog Fi ~ dlog F2'
as analogy with temperature would require. In fact, we have
seen that
d log F12 d log FitM ~~ d log Fj
172 THERMODYNAMIC ANALOGIES.
where the second and third members of the equation denote
average values in an ensemble in which the compound system
is microcanonically distributed in phase. Let us suppose the
two original systems to be identical in nature. Then
The equation in question would require that
i. e., that we get the same result, whether we take the value
of del/dlog V} determined for the average value of e1 in the
ensemble, or take the average value of de^dlog F"r This
will be the case where de^dlog V^ is a linear function of er
Evidently this does not constitute the most general case.
Therefore the equation in question cannot be true in general.
It is true, however, in some very important particular cases, as
when the energy is a quadratic function of the p's and ^'s, or
of the p's alone.* When the equation holds, the case is anal
ogous to that of bodies in thermodynamics for which the
specific heat for constant volume is constant.
Another quantity which is closely related to temperature is
dcfr/de. It has been shown in Chapter IX that in a canonical
ensemble, if n > 2, the average value of d(f>fde is I/®, and
that the most common value of the energy in the ensemble is
that for which d$/de = I/®. The first of these properties
may be compared with that of de/dlog V, which has been
seen to have the average value ® in a canonical ensemble,
without restriction in regard to the number of degrees of
freedom.
With respect to microcanonical ensembles also, dfyjde has
a property similar to what has been mentioned with respect to
de/d log V. That is, if a system microcanonically distributed
in phase consists of two parts with separate energies, and each
* This last case is important on account of its relation to the theory of
gases, although it must in strictness be regarded as a limit of possible cases,
rather than as a case which is itself possible.
THERMODYNAMIC ANALOGIES. 173
with more than two degrees of freedom, the average values in
the ensemble of d(f)/de for the two parts are equal to one
another and to the value of same expression for the whole.
In our usual notations
"^12
de2 L2 ~" delz
if TIX > 2, and n2 > 2.
This analogy with temperature has the same incompleteness
which was noticed with respect to de/dlog V, viz., if two sys
tems have such energies (ej and e2) that
and they are combined to form a third system with energy
*ia = €1 + €2,
we shall not have in general
c?012 _ dfa __ dz
d€i2 deL dez '
Thus, if the energy is a quadratic function of the p's and 2 HI + n% — 2
del ~ de2 " e1 + ez
If the energy is a quadratic function of the p's alone, the case
would be the same except that we should have J n^ , J w2 , J w12 ,
instead of wx , w2 , w12. In these particular cases, the analogy
* See footnote on page 93. We have here made the least value of the
energy consistent with the values of the external coordinates zero instead
of ea, as is evidently allowable when the external coordinates are supposed
invariable.
174 THERMODYNAMIC ANALOGIES.
between de/d log V and temperature would be complete, as has
already been remarked. We should have
del e^ c?62 _ e2
n'9 dlo V~'
_=
MM rflog F! dlogV2'
when the energy is a quadratic function of the p's and #'s, and
similar equations with £ % , J ra2 ,  w12 , instead of ^ , w2 , w12 ,
when the energy is a quadratic function of the £>'s alone.
More characteristic of dcf>/de are its properties relating to
most probable values of energy. If a system having two parts
with separate energies and each with more than two degrees
of freedom is microcanonically distributed in phase, the most
probable division of energy between the parts, in a system
taken at random from the ensemble, satisfies the equation
^ = ^, (488)
del de2
which corresponds to the thermodynamic theorem that the
distribution of energy between the parts of a system, in case of
thermal equilibrium, is such that the temperatures of the parts
are equal.
To prove the theorem, we observe that the fractional part
of the whole number of systems which have the energy of one
part (ej) between the limits e/ and e/ is expressed by
r*f. ****>,
T i
where the variables are connected by the equation
€j  €2 = constant = ei2 .
The greatest value of this expression, for a constant infinitesi
mal value of the difference ex" — e/, determines a value of e1 ,
which we may call its most probable value. This depends on
the greatest possible value of fa + fa. Now if n^ > 2, and
w2 > 2, we shall have fa = — oo for the least possible value of
THERMODYNAMIC ANALOGIES. 175
6j , and 2 = — QO for the least possible value of e2. Between
these limits (/>x and <£2 will be finite and continuous. Hence
$! + <£2 will have a maximum satisfying the equation (488).
But if n^ < 2, or w2 < 2, d(f)1/d€l or d$2/de2 may be nega
tive, or zero, for all values of e1 or e2, and can hardly be
regarded as having properties analogous to temperature.
It is also worthy of notice that if a system which is micro
canonically distributed in phase has three parts with separate
energies, and each with more than two degrees of freedom, the
most probable division of energy between these parts satisfies
the equation
That is, this equation gives the most probable set of values
of ej, 62, and e3. But it does not give the most probable
value of el , or of e2 , or of e3. Thus, if the energies are quad
ratic functions of the p9s and as corresponding to
temperature, <£ will correspond to entropy. It has been denned
as log (d V/de). In the considerations on which its definition
is founded, it is therefore very similar to log F". We have
seen that d(j>/dlogV approaches the value unity when n is
very great.* .
To form a differential equation on the model of the thermo
dynamic equation (482), in which de/dcf) shall take the place
of temperature, and <£ of entropy, we may write
da* + etc> (489>
or /de) has relations
to the most probable values of energy in parts of a microca
nonical ensemble. That (del da^)^, etc., have properties
somewhat analogous, may be shown as follows.
In a physical experiment, we measure a force by balancing it
against another. If we should ask what force applied to in
crease or diminish &x would balance the action of the systems,
it would be one which varies with the different systems. But
we may ask what single force will make a given value of a^
the most probable, and we shall find that under certain condi
tions (de/da^Q, a represents that force.
* See Chapter X, pages 120, 121.
t See Chapter IX, equations (321), (327).
THERMODYNAMIC ANALOGIES. 177
To make the problem definite, let us consider a system con
sisting of the original system together with another having
the coordinates a^ , a2 , etc., and forces AJ, A<£, etc., tending
to increase those coordinates. These are in addition to the
forces Av Av etc., exerted by the original system, and are de
rived from a forcefunction (— eg') by the equations
^;_ &J A, _^L etc
Al ' ~d^> da2'
For the energy of the whole system we may write
E = e + ej + •Jm1a'12 + im2a22 + etc.,
and for the extensioninphase of the whole system within any
limits
I ... I dpi . . . dqn da,i mi da± daz mz da2 . . .
or I . . . I e$ de da^ m1 da^ daz m2 da2 . . . ,
or again I . . . / e^ d& dat mx dai da2 m2 da2 . . . ,
since de = c?E, when ax, ax, a2, «2, etc., are constant. If the
limits are expressed by E and E + c?E, a^ and a^ + da^ , a1 and
«j + da^ , etc., the integral reduces to
The values of ^ , ax , «2 , a. ete" «i. etc" ore' 
etc., flj, etc., which are closely related to ensembles of constant
energy, and to average and most probable values in such
ensembles, and most of which are defined without reference
to any ensemble, may appear the most natural analogues of
the thermodynamic quantities.
In regard to the naturalness of seeking analogies with the
thermodynamic behavior of bodies in canonical or microca
nonical ensembles of systems, much will depend upon how we
approach the subject, especially upon the question whether we
regard energy or temperature as an independent variable.
It is very natural to take energy for an independent variable
rather than temperature, because ordinary mechanics furnishes
us with a perfectly defined conception of energy, whereas the
idea of something relating to a mechanical system and corre
THERMODYNAMIC ANALOGIES. 179
spending to temperature is a notion but vaguely denned. Now
if the state of a system is given by its energy and the external
coordinates, it is incompletely denned, although its partial defi
nition is perfectly clear as far as it goes. The ensemble of
phases microcanonically distributed, with the given values of
the energy and the external coordinates, will represent the im
perfectly defined system better than any other ensemble or
single phase. When we approach the subject from this side,
our theorems will naturally relate to average values, or most
probable values, in such ensembles.
In this case, the choice between the variables of (485) or of
(489) will be determined partly by the relative importance
which is attached to average and probable values. It would
seem that in general average values are the most important, and
that they lend themselves better to analytical transformations.
This consideration would give the preference to the system of
variables in which log V is the analogue of entropy. Moreover,
if we make the analogue of entropy, we are embarrassed by
the necessity of making numerous exceptions for systems of
one or two degrees of freedom.
On the other hand, the definition of <£ may be regarded as a
little more simple than that of log F", and if our choice is deter
mined by the simplicity of the definitions of the analogues of
entropy and temperature, it would seem that the <£ system
should have the preference. In our definition of these quanti
ties, V was defined first, and e^ derived from V by differen
tiation. This gives the relation of the quantities in the most
simple analytical form. Yet so far as the notions are con
cerned, it is perhaps more natural to regard Fas derived from
C* by integration. At all events, e* may be defined inde
pendently of F", and its definition niay be regarded as more
simple as not requiring the determination of the zero from
which V is measured, which sometimes involves questions
of a delicate nature. In fact, the quantity e* may exist,
when the definition of V becomes illusory for practical pur
poses, as the integral by which it is determined becomes infinite.
The case is entirely different, when we regard the tempera
180 THERMODYNAMIC ANALOGIES.
ture as an independent variable, and we have to consider a
system which is described as having a certain temperature and
certain values for the external coordinates. Here also the
state of the system is not completely denned, and will be
better represented by an ensemble of phases than by any single
phase. What is the nature of such an ensemble as will best
represent the imperfectly defined state ?
When we wish to give a body a certain temperature, we
place it in a bath of the proper temperature, and when we
regard what we call thermal equilibrium as established, we say
that the body has the same temperature as the bath. Per
haps we place a second body of standard character, which we
call a thermometer, in the bath, and say that the first body,
the bath, and the thermometer, have all the same temperature.
But the body under such circumstances, as well as the
bath, and the thermometer, even if they were entirely isolated
from external influences (which it is convenient to suppose
in a theoretical discussion), would be continually changing in
phase, and in energy as well as in other respects, although
our means of observation are not fine enough to perceive
these variations.
The series of phases through which the whole system runs
in the course of time may not be entirely determined by the
energy, but may depend on the initial phase in other respects.
In such cases the ensemble obtained by the microcanonical
distribution of the whole system, which includes all possible
timeensembles combined in the proportion which seems least
arbitrary, will represent better than any one timeensemble
the effect of the bath. Indeed a single timeensemble, when
it is not also a microcanonical ensemble, is too illdefined a
notion to serve the purposes of a general discussion. We
will therefore direct our attention, when we suppose the body
placed in a bath, to the microcanonical ensemble of phases
thus obtained.
If we now suppose the quantity of the substance forming
the bath to be increased, the anomalies of the separate ener
gies of the body and of the thermometer in the microcanonical
THERMODYNAMIC ANALOGIES. 181
ensemble will be increased, but not without limit. The anom
alies of the energy of the bath, considered in comparison with
its whole energy, diminish indefinitely as the quantity of the
bath is increased, and become in a sense negligible, when
the quantity of the bath is sufficiently increased. The
ensemble of phases of the body, and of the thermometer,
approach a standard form as the quantity of the bath is in
definitely increased. This limiting form is easily shown to be
what we have described as the canonical distribution.
Let us write e for the energy of the whole system consisting
of the body first mentioned, the bath, and the thermometer
(if any), an4 let us first suppose this system to be distributed
canonically with the modulus ©. We have by (205)
and since ep = •=
de _ n de
H®~~2dep'
If we write Ae for the anomaly of mean square, we have
d®
If we set
A® will represent approximately the increase of ® which
would produce an increase in the average value of the energy
equal to its anomaly of mean square. Now these equations
give
(A©)* = 
n
which shows that we may diminish A ® indefinitely by increas
ing the quantity of the bath.
Now our canonical ensemble consists of an infinity of micro
canonical ensembles, which differ only in consequence of the
different values of the energy which is constant in each. If
we consider separately the phases of the first body which
182 THERMODYNAMIC ANALOGIES.
occur in the canonical ensemble of the whole system, these
phases will form a canonical ensemble of the same modulus.
This canonical ensemble of phases of the first body will con
sist of parts which belong to the different microcanonical
ensembles into which the canonical ensemble of the whole
system is divided.
Let us now imagine that the modulus of the principal ca
nonical ensemble is increased by 2 A (8), and its average energy
by 2Ae. The modulus of the canonical ensemble of the
phases of the first body considered separately will be increased
by 2 A ®. We may regard the infinity of microcanonical en
sembles into which we have divided the principal canonical
ensemble as each having its energy increased by 2Ae. Let
us see how the ensembles of phases of the first body con
tained in these microcanonical ensembles are affected. We
may assume that they will all be affected in about the same
way, as all the differences which come into account may be
treated as small. Therefore, the canonical ensemble formed by
taking them together will also be affected in the same way.
But we know how this is affected. It is by the increase of
its modulus by 2 A®, a quantity which vanishes when the
quantity of the bath is indefinitely increased.
In the case of an infinite bath, therefore, the increase of the
energy of one of the microcanonical ensembles by 2Ae, pro
duces a vanishing effect on the distribution in energy of the
phases of the first body which it contains. But 2Ae is more
than the average difference of energy between the micro
canonical ensembles. The distribution in energy of these
phases is therefore the same in the different microcanonical
ensembles, and must therefore be canonical, like that of the
ensemble which they form when taken together.*
* In order to appreciate the above reasoning, it should be understood that
the differences of energy which occur in the canonical ensemble of phases of
the first body are not here regarded as vanishing quantities. To fix one's
ideas, one may imagine that he has the fineness of perception to make these
differences seem large. The difference between the part of these phases
which belong to one microcanonical ensemble of the whole system and the
part which belongs to another would still be imperceptible, when the quan
tity of the bath is sufficiently increased.
THERMODYNAMIC ANALOGIES. 183
As a general theorem, the conclusion may be expressed in
the words : — If a system of a great number of degrees of
freedom is microcanonically distributed in phase, any very
small part of it may be regarded as canonically distributed.*
It would seem, therefore, that a canonical ensemble of
phases is what best represents, with the precision necessary
for exact mathematical reasoning, the notion of a body with
a given temperature, if we conceive of the temperature as the
state produced by such processes as we actually use in physics
to produce a given temperature. Since the anomalies of the
body increase with the quantity of the bath, we can only get
rid of all that is arbitrary in the ensemble of phases which is
to represent the notion of a body of a given temperature by
making the bath infinite, which brings us to the canonical
distribution.
A comparison of temperature and entropy with their ana
logues in statistical mechanics would be incomplete without a
consideration of their differences with respect to units and
zeros, and the numbers used for their numerical specification.
If we apply the notions of statistical mechanics to such bodies
as we usually consider in thermodynamics, for which the
kinetic energy is of the same order of magnitude as the unit
of energy, but the number of degrees of freedom is enormous,
the values of B, de/dlogV, and de/d will be of the same
order of magnitude as 1/w, and the variable part of ?;, log V,
and will be of the same order of magnitude as w.f If these
quantities, therefore, represent in any sense the notions of tem
perature and entropy, they will nevertheless not be measured
in units of the usual order of magnitude, — a fact which must
be borne in mind in determining what magnitudes may be
regarded as insensible to human observation.
Now nothing prevents our supposing energy and time in
our statistical formulae to be measured in such units as may
* It is assumed — and without this assumption the theorem would have
no distinct meaning — that the part of the ensemble considered may be
regarded as having separate energy.
t See equations (124), (288), (289), and (314) ; also page 106.
184 THERMODYNAMIC ANALOGIES.
be convenient for physical purposes. But when these units
have been chosen, the numerical values of ®, de/dlogV,
de/d, 7), log FJ <£, are entirely determined,* and in order to
compare them with temperature and entropy, the numerical
values of which depend upon an arbitrary unit, we must mul
tiply all values of ®, de/dlogV, de',d^ by a constant (7T),
and divide all values of 77, log FJ and by the same constant.
This constant is the same for all bodies, and depends only on
the units of temperature and energy which we employ. For
ordinary units it is of the same order of magnitude as the
numbers of atoms in ordinary bodies.
We are not able to determine the numerical value of K>
as it depends on the number of molecules in the bodies with
which we experiment. To fix our ideas, however, we may
seek an expression for this value, based upon very probable
assumptions, which will show how we would naturally pro
ceed to its evaluation, if our powers of observation were fine
enough to take cognizance of individual molecules.
If the unit of mass of a monatomic gas contains v atoms,
and it may be treated as a system of 3 v degrees of free
dom, which seems to be the case, we have for canonical
distribution
If we write T for temperature, and cv for the specific heat of
the gas for constant volume (or rather the limit toward
which this specific heat tends, as rarefaction is indefinitely
increased), we have
since we may regard the energy as entirely kinetic. We may
set the ep of this equation equal to the ep of the preceding,
* The unit of time only affects the last three quantities, and these only
by an additive constant, which disappears (with the additive constant of
entropy), when differences of entropy are compared with their statistical
analogues. See page 19.
THERMODYNAMIC ANALOGIES. 185
where indeed the individual values of which the average is
taken would appear to human observation as identical. This
gives
d® 2cv
whence =' <493)
a value recognized by physicists as a constant independent of
the kind of monatomic gas considered.
We may also express the value of K in a somewhat different
form, which corresponds to the indirect method by which
physicists are accustomed to determine the quantity cv. The
kinetic energy due to the motions of the centers of mass of
the molecules of a mass of gas sufficiently expanded is easily
shown to be equal to
where p and v denote the pressure and volume. The average
value of the same energy in a canonical ensemble of such
a mass of gas is
J0v,
where v denotes the number of molecules in the gas. Equat
ing these values, we have
pv = ®v, (494)
whence J£~~T~^' (495)
Now the laws of Boyle, Charles, and Avogadro may be ex
pressed by the equation
pv — AvT, (496)
where A is a constant depending only on the units hi which
energy and temperature are measured. 1 / K, therefore, might
be called the constant of the law of Boyle, Charles, and
Avogadro as expressed with reference to the true number of
molecules in a gaseous body.
Since such numbers are unknown to us, it is more conven
ient to express the law with reference to relative values. If
we denote by M the socalled molecular weight of a gas, that
186 THERMODYNAMIC ANALOGIES.
is, a number taken from a table of numbers proportional to
the weights of various molecules and atoms, but having one
of the values, perhaps the atomic weight of hydrogen, arbi
trarily made unity, the law of Boyle, Charles, and Avogadro
may be written in the more practical form
pv = A'T^, (497)
JXL
where A' is a constant and m the weight of gas considered.
It is evident that 1 K is equal to the product of the constant
of the law in this form and the (true) weight of an atom of
hydrogen, or such other atom or molecule as may be given
the value unity in the table of molecular weights.
In the following chapter we shall consider the necessary
modifications in the theory of equilibrium, when the quantity
of matter contained in a system is to be regarded as variable,
or, if the system contains more than one kind of matter,
when the quantities of the several kinds of matter in the
system are to be regarded as independently variable. This
will give us yet another set of variables in the statistical
equation, corresponding to those of the amplified form of
the thennodynamic equation.
CHAPTER XV.
SYSTEMS COMPOSED OF MOLECULES.
THE nature of material bodies is such, that especial interest
attaches to the dynamics of systems composed of a great
number of entirely similar particles, or, it may be, of a great
number of particles of several kinds, all of each kind being
entirely similar to each other. We shall therefore proceed to
consider systems composed of such particles, whether in great
numbers or otherwise, and especially to consider the statistical
equilibrium of ensembles of such systems. One of the varia
tions to be considered in regard to such systems is a variation
in the numbers of the particles of the various kinds which it
contains, and the question of statistical equilibrium between
two ensembles of such systems relates in part to the tendencies
of the various kinds of particles to pass from the one to the
other.
First of all, we must define precisely what is meant by
statistical equilibrium of such an ensemble of systems. The
essence of statistical equilibrium is the permanence of the
number of systems which fall within any given limits with
respect to phase. We have therefore to define how the term
" phase " is to be understood in such cases. If two phases differ
only in that certain entirely similar particles have changed
places with one another, are they to be regarded as identical
or different phases? If the particles are regarded as indis
tinguishable, it seems in accordance with the spirit of the
statistical method to regard the phases as identical. In fact,
it might be urged that in such an ensemble of systems as we
are considering no identity is possible between the particles
of different systems except that of qualities, and if v particles
of one system are described as entirely similar to one another
and to v of another system, nothing remains on which to base
188 SYSTEMS COMPOSED OF MOLECULES.
the indentification of any particular particle of the first system
with any particular particle of the second. And this would
be true, if the ensemble of systems had a simultaneous
objective existence. But it hardly applies to the creations
of the imagination. In the cases which we have been con
sidering, and in those which we shall consider, it is not only
possible to conceive of the motion of an ensemble of similar
systems simply as possible cases of the motion of a single
system, but it is actually in large measure for the sake of
representing more clearly the possible cases of the motion of
a single system that we use the conception of an ensemble
of systems. The perfect similarity of several particles of a
system will not in the least interfere with the identification
of a particular particle in one case with a particular particle
in another. The question is one to be decided in accordance
with the requirements of practical convenience in the discus
sion of the problems with which we are engaged.
Our present purpose will often require us to use the terms
phase, densityinphase, statistical equilibrium, and other con
nected terms on the supposition that phases are not altered
by the exchange of places between similar particles. Some
of the most important questions with which we are concerned
have reference to phases thus defined. We shall call them
phases determined by generic definitions, or briefly, generic
phases. But we shall also be obliged to discuss phases de
fined by the narrower definition (so that exchange of position
between similar particles is regarded as changing the phase),
which will be called phases determined by specific definitions,
or briefly, specific phases. For the analytical description of
a specific phase is more simple than that of a generic phase.
And it is a more simple matter to make a multiple integral
extend over all possible specific phases than to make one extend
without repetition over all possible generic phases.
It is evident that if i>i, vz . . . vh, are the numbers of the dif
ferent kinds of molecules in any system, the number of specific
phases embraced in one generic phase is represented by the
continued product [z^ [^ • • • ]^ and the coefficient of probabil
SYSTEMS COMPOSED OF MOLECULES. 189
ity of a generic phase is the sum of the probabilitycoefficients
of the specific phases which it represents. When these are
equal among themselves, the probabilitycoefficient of the gen
eric phase is equal to that of the specific phase multiplied by
[z/i 1 1>2 . . . \vg It is also evident that statistical equilibrium
may subsist with respect to generic phases without statistical
equilibrium with respect to specific phases, but not vice versa.
Similar questions arise where one particle is capable of
several equivalent positions. Does the change from one of
these positions to another change the phase? It would be
most natural and logical to make it affect the specific phase,
but not the generic. The number of specific phases contained
in a generic phase would then be \v± /e/1 . . . z^ /ch\ where
KV . . . Kh denote the numbers of equivalent positions belong
ing to the several kinds of particles. The case in which a K is
infinite would then require especial attention. It does not
appear that the resulting complications in the formulae would
be compensated by any real advantage. The reason of this is
that in problems of real interest equivalent positions of a
particle will always be equally probable. In this respect,
equivalent positions of the same particle are entirely unlike
the [^different ways in which v particles may be distributed
in v different positions. Let it therefore be understood that
in spite of the physical equivalence of different positions of
the same particle they are to be considered as constituting a
difference of generic phase as well as of specific. The number
of specific phases contained in a generic phase is therefore
always given by the product \v^\v^ • » . [iy
Instead of considering, as in the preceding chapters, en
sembles of systems differing only in phase, we shall now
suppose that the systems constituting an ensemble are com
posed of particles of various kinds, and that they differ not
only in phase but also in the numbers of these particles which
they contain. The external coordinates of all the systems in
the ensemble are supposed, as heretofore, to have the same
value, and when they vary, to vary together. For distinction,
we may call such an ensemble a grand ensemble, and one in
190 SYSTEMS COMPOSED OF MOLECULES.
which the systems differ only in phase a petit ensemble. A
grand ensemble is therefore composed of a multitude of petit
ensembles. The ensembles which we have hitherto discussed
are petit ensembles.
Let i>j, . . . vh9 etc., denote the numbers of the different
kinds of particles in a system, e its energy, and ql1 . . . qn,
pl , . . . pn its coordinates and momenta. If the particles are of
the nature of material points, the number of coordinates (n)
of the system will be equal to 3 vl . . . + 3 vh. But if the parti
cles are less simple in their nature, if they are to be treated
as rigid solids, the orientation of which must be regarded, or
if they consist each of several atoms, so as to have more than
three degrees of freedom, the number of coordinates of the
system will be equal to the sum of vlt i>2, etc., multiplied
each by the number of degrees of freedom of the kind of
particle to which it relates.
Let us consider an ensemble in which the number of
systems having v19 . . . vh particles of the several kinds, and
having values of their coordinates and momenta lying between
the limits ql and q^ + dq1 , p1 and pl + dpl , etc., is represented
by the expression
(498)
where IV, O, ®, /^ , . . . ph are constants, N denoting the total
number of systems in the ensemble. The expression
Qf Wi 
Ne ® (499)
[vi."h
evidently represents the densityinphase of the ensemble
within the limits described, that is, for a phase specifically
defined. The expression
e * (500)
SYSTEMS COMPOSED OF MOLECULES. 191
is therefore the probabilitycoefficient for a phase specifically
defined. This has evidently the same value for all the
[iY . . . \vh phases obtained by interchanging the phases of
particles of the same kind. The probabilitycoefficient for a
generic phase will be \vi_. . . [z^ times as great, viz.,
e . (501)
We shall say that such an ensemble as has been described
is canonically distributed, and shall call the constant © its
modulus. It is evidently what we have called a grand ensem
ble. The petit ensembles of which it is composed are
canonically distributed, according to the definitions of Chapter
IV, since the expression
(502)
is constant for each petit ensemble. The grand ensemble,
therefore, is in statistical equilibrium with respect to specific
phases.
If an ensemble, whether grand or petit, is identical so far
as generic phases are concerned with one canonically distrib
uted, we shall say that its distribution is canonical with
respect to generic phases. Such an ensemble is evidently in
statistical equilibrium with respect to generic phases, although
it may not be so with respect to specific phases.
If we write H for the index of probability of a generic phase
in a grand ensemble, we have for the case of canonical
distribution
H = 0 + M.n — + >*»*« _ (503)
It will be observed that the H is a linear function of e and
vv . . . vh ; also that whenever the index of probability of
generic phases in a grand ensemble is a linear function of
e, j/j, . . . vhi the ensemble is canonically distributed with
respect to generic phases.
192 SYSTEMS COMPOSED OF MOLECULES.
The constant Ii we may regard as determined by the
equation
/C]\TP ®
• / ^n  i  dp,... dqn, (504)
phases J ln'b_
or
[1/1 . . . [
' — phases
(505)
where the multiple sum indicated by 2Vl . . . 2rft includes all
terms obtained by giving to each of the symbols vi . . . vh all
integral values from zero upward, and the multiple integral
(which is to be evaluated separately for each term of the
multiple sum) is to be extended over all the (specific) phases
of the system having the specified numbers of particles of the
various kinds. The multiple integral hi the last equation is
JL
what we have represented by e 0. See equation (92). We
may therefore write
It should be observed that the summation includes a term
in which all the symbols vl . . . vh have the value zero. We
must therefore recognize in a certain sense a system consisting
of no particles, which, although a barren subject of study in
itself, cannot well be excluded as a particular case of a system
of a variable number of particles. In this case e is constant,
and there are no integrations to be performed. We have
therefore*
_4 _1
e ® = e ®, i. e.y \j/ = e.
* This conclusion may appear a little strained. The original definition
of ^ may not be regarded as fairly applying to systems of no degrees of
freedom. We may therefore prefer to regard these equations as defining
4/ in this case.
SYSTEMS COMPOSED OF MOLECULES. 193
The value of ep is of course zero in this case. But the
value of eq contains an arbitrary constant, which is generally
determined by considerations of convenience, so that eg and e
do not necessarily vanish with v^ , . . . vh.
Unless — II has a finite value, our formulae become illusory.
We have already, in considering petit ensembles canonically
distributed, found it necessary to exclude cases in which — ty
has not a finite value.* The same exclusion would here
make — ^r finite for any finite values of vl . . . vh. This does
not necessarily make a multiple series of the form (506) finite.
We may observe, however, that if for all values of vl . . . vh
\l/ ^. CQ + ^1 Vl) • • • 4" Ch Vht (507)
where £0, cv . . . ch are constants or functions of ®,
CoMMl+CjK . . . K/*A+CAVA
e
^
_n £p
& e
O. c_0
0^0
. . . e
£+• e .. + « e • (508)
The value of — II will therefore be finite, when the condition
(507) is satisfied. If therefore we assume that — fl is finite,
we do not appear to exclude any cases which are analogous to
those of nature.f
The interest of the ensemble which has been described lies
in the fact that it may be in statistical equilbrium, both in
* See Chapter IV, page 35.
t If the external coordinates determine a certain volume within which the
system is" confined, the contrary of (507) would imply that we could obtain
an infinite amount of work by crowding an infinite quantity of matter into a
finite volume.
13
194 SYSTEMS COMPOSED OF MOLECULES.
respect to exchange of energy and exchange of particles, with
other grand ensembles canonically distributed and having the
same values of ® and of the coefficients pv ^2, etc., when the
circumstances are such that exchange of energy and of
particles are possible, and when equilibrium would not sub
sist, were it not for equal values of these constants in the two
ensembles.
With respect to the exchange of energy, the case is exactly
the same as that of the petit ensembles considered in Chapter
IV, and needs no especial discussion. The question of ex
change of particles is to a certain extent analogous, and may
be treated in a somewhat similar manner. Let us suppose
that we have two grand ensembles canonically distributed
with respect to specific phases, with the same value of the
modulus and of the coefficients ^ . . . fih , and let us consider
the ensemble of all the systems obtained by combining each
system of the first ensemble with each of the second.
The probabilitycoefficient of a generic phase in the first
ensemble may be expressed by
e & (509)
The probabilitycoefficient of a specific phase will then be
expressed by
(510)
since each generic phase comprises \v^ . . . [z^ specific phases.
In the second ensemble the probabilitycoefficients of the
generic and specific phases will be
SYSTEMS COMPOSED OF MOLECULES. 195
The probabilitycoefficient of a generic phase in the third
ensemble, which consists of systems obtained by regarding
each system of the first ensemble combined with each of the
second as forming a system, will be the product of the proba
bilitycoefficients of the generic phases of the systems com
bined, and will therefore be represented by the formula
e (513)
where ft"' = ft' + ft", e'" = e' + e", vi'" = vj + z>i", etc. It
will be observed that i//", etc., represent the numbers of
particles of the various kinds in the third ensemble, and e'"
its energy ; also that ft'" is a constant. The third ensemble
is therefore canonically distributed with respect to generic
phases.
If all the systems in the same generic phase in the third
ensemble were equably distributed among the zV" • • •  vjj" spe
cific phases which are comprised in the generic phase, the prob
abilitycoefficient of a specific phase would be
In fact, however, the probabilitycoefficient of any specific
phase which occurs in the third ensemble is
which we get by multiplying the probabilitycoefficients of
specific phases in the first and second ensembles. The differ
ence between the formulae (514) and (515) is due to the fact
that the generic phases to which (513) relates include not
only the specific phases occurring in the third ensemble and
having the probabilitycoefficient (515), but also all the
specifier phases obtained from these by interchange of similar
particles between two combined systems. Of these the proba
196 SYSTEMS COMPOSED OF MOLECULES.
bilitycoefficient is evidently zero, as they do not occur in the
ensemble.
Now this third ensemble is in statistical equilibrium, with
respect both to specific and generic phases, since the ensembles
from which it is formed are so. This statistical equilibrium
is not dependent on the equality of the modulus and the coeffi
cients /Aj , . . . fxh in the first and second ensembles. It depends
only on the fact that the two original ensembles were separ
ately in statistical equilibrium, and that there is no interaction
between them, the combining of the two ensembles to form a
third being purely nominal, and involving no physical connec
tion. This independence of the systems, determined physically
by forces which prevent particles from passing from one sys
tem to the other, or coming within range of each other's action,
is represented mathematically by infinite values of the energy
for particles in a space dividing the systems. Such a space
may be called a diaphragm.
If we now suppose that, when we combine the systems of
the two original ensembles, the forces are so modified that the
energy is nc longer infinite for particles in all the space form
ing the diaphragm, but is diminished in a part of this space,
so that it is possible for particles to pass from one system
to the other, this will involve a change in the function e;//
which represents the energy of the combined systems, and the
equation e"f — ef + eff will no longer hold. Now if the co
efficient of probability in the third ensemble were represented
by (513) with this new function e;//, we should have statistical
equilibrium, with respect to generic phases, although not to
specific. But this need involve only a trifling change in the
distribution of the third ensemble,* a change represented by
the addition of comparatively few systems in which the trans
ference of particles is taking place to the immense number
* It will be observed that, so far as the distribution is concerned, very
large and infinite values of e (for certain phases) amount to nearly the same
thing, — one representing the total and the other the nearly total exclusion
of the phases in question. An infinite change, therefore, in the value of e
(for certain phases) may represent a vanishing change in the distribution.
SYSTEMS COMPOSED OF MOLECULES. 197
obtained by combining the two original ensembles. The
difference between the ensemble which would be in statistical
equilibrium, and that obtained by combining the two original
ensembles may be diminished without limit, while it is still
possible for particles to pass from one system to another. In 
this sense we may say that the ensemble formed by combining
the two given ensembles may still be regarded as in a state of
(approximate) statistical equilibrium with respect to generic
phases, when it has been made possible for particles to pass
between the systems combined, and when statistical equilibrium
for specific phases has therefore entirely ceased to exist, and
when the equilibrium for generic phases would also have
entirely ceased to exist, if the given ensembles had not been
canonically distributed, with respect to generic phases, with
the same values of @ and fiv . . . ph.
It is evident also that considerations of this kind will apply j
separately to the several kinds of particles. We may diminish '
the energy in the space forming the diaphragm for one kind of
particle and not for another. This is the mathematical ex
pression for a " semipermeable" diaphragm. The condition
necessary for statistical equilibrium where the diaphragm is
permeable only to particles to which the suffix ( )x relates
will be fulfilled when /^ and ® have the same values in the
two ensembles, although the other coefficients /*2, etc., may be
different.
This important property of grand ensembles with canonical
distribution will supply the motive for a more particular ex
amination of the nature of such ensembles, and especially of
the comparative numbers of systems in the several petit en
sembles which make up a grand ensemble, and of the average
values in the grand ensemble of some of the most important
quantities, and of the average squares of the deviations from
these average values.
The probability that a system taken at random from a
grand ensemble canonically distributed will have exactly
i/j, . . . vh particles of the various kinds is expressed by the
multiple integral
198 SYSTEMS COMPOSED OF MOLECULES.
phases
or « . (517)
[vi . . . [1/5
This may be called the probability of the petit ensemble
0>i, ... vh). The sum of all such probabilities is evidently
unity. That is,
(518)
which agrees with (506).
The average value in the grand ensemble of any quantity
u, is given by the formula
phases
If w is a function of i/x, . . . i/A alone, i. e., if it has the same
value in all systems of any same petit ensemble, the formula
reduces to
8 ue
= X"'
Again, if we write ^grand an(i w] petit to distinguish averages in
the grand and petit ensembles, we shall have
In this chapter, in which we are treating of grand en
sembles, u will always denote the average for a grand en
semble. In the preceding chapters, u has always denoted
the average for a petit ensemble.
SYSTEMS COMPOSED OF MOLECULES. 199
Equation (505), which we repeat in a slightly different
form, viz.,
phases
shows that O is a function of ® and pv . . . fj,h ; also of the
external coordinates «1? a2, etc., which are involved implicitly
in e. If we differentiate the equation regarding all these
quantities as variable, we have
la
phases
phases
+ etc.
all
de e
phases
 etc. (523)
5
If we multiply this equation by e9, and set as usual Av Av
etc., for — de/da^ — delda^, etc., we get in virtue of the law
expressed by equation (519),
dto O _ d® 
200 SYSTEMS COMPOSED OF MOLECULES.
that is,
da = O + PI* — ft*? ^ _ a  fa _ s 2i dai (525)
Since equation (503) gives
the preceding equation may be written
dQ, — ILd® — 2 vid/xi — 2 2l dalt (527)
Again, equation (526) gives
c?Q + Sjiie^i + S vi cZ/*! — de = ©dH + Hc2®. (528)
Eliminating <#fl from these equations, we get
de = — ©rfH + 2/x^i  S^j rfoj. (529)
If we set * = e + © H, (530)
d* = de + © dH + H d®, (531)
we have d* = H d® + 2 ^ d^  S ^ e?^. (532)
The corresponding thermodynamic equations are
de = Tdy + 5 ^dmi — S 4i ^ , (533)
(534)
/xi cZm! — SA cZ/xi , (543)
and therefore
/I 1 \
> m"  Hi' (544)
The difference /*/' — ^ is therefore numerically a very small
quantity. To form an idea of the importance of such a
difference, we should observe that in formula (498) ^ is
multiplied by v1 and the product subtracted from the energy.
A very small difference in the value of /^ may therefore be im
portant. But since v 7gen> (549)
which corresponds to the equation
we have i/^ = $ + ® log [v,
and H^^n + ^yi@+°. (551)
This will have a maximum when *
Distinguishing values corresponding to this maximum by
accents, we have approximately, when vl , . . . vh are of the
same order of magnitude as the numbers of molecules in ordi
nary bodies,
Q + /*iVi . . • + ^ftVft — Igen
 
©
2© \dvidv ©
(553)
2© Vc?!'!^/ © \log(2ff®),
where D =
V
/ dVn.V
Wviflfow
that is, D =
(558)
(559)
'(560)
(561)
Now, by (553), we have for the first approximation
H  ^gen = C = 1 log D   log (2ir0), (562)
and if we divide by the constant JT,* to reduce these quanti
ties to the usual unit of entropy,
H  ^gen = ^g J> ~ h log (27T@)
.ST 2 JL
* See page 184186.
206 SYSTEMS COMPOSED OF MOLECULES.
This is evidently a negligible quantity, since K is of the same
order of magnitude as the number of molecules in ordinary
bodies. It is to be observed that ?7gen is here the average in
the grand ensemble, whereas the quantity which we wish to
compare with H is the average in a petit ensemble. But as we
have seen that in the case considered the grand ensemble would
appear to human observation as a petit ensemble, this dis
tinction may be neglected.
The differences therefore, in the case considered, between the
quantities which may be represented by the notations *
H*en [grand » ^*en (grand ' ^^ Ipetit
are not sensible to human faculties. The difference
and is therefore constant, so long as the numbers z>1? . . . vh
are constant. For constant values of these numbers, therefore,
it is immaterial whether we use the average of rjgen or of 77 for
entropy, since this only affects the arbitrary constant of in
tegration which is added to entropy. But when the numbers
vv . . . vh are varied, it is no longer possible to use the index
for specific phases. For the principle that the entropy of any
body has an arbitrary additive constant is subject to limi
tation, when different quantities of the same substance are
concerned. In this case, the constant being determined for
one quantity of a substance, is thereby determined for all
quantities of the same substance.
To fix our ideas, let us suppose that we have two identical
fluid masses in contiguous chambers. The entropy of the
whole is equal to the sum of the entropies of the parts, and
double that of one part. Suppose a valve is now opened,
making a communication between the chambers. We do not
regard this as making any change in the entropy, although
the masses of gas or liquid diffuse into one another, and al
though the same process of diffusion would increase the
* In this paragraph, for greater distinctness, Hgengrand and %p^lpetit have
been written for the quantities which elsewhere are denoted by H and rf.
SYSTEMS COMPOSED OF MOLECULES. 207
entropy, if the masses of fluid were different. It is evident,
therefore, that it is equilibrium with respect to generic phases,
and not with respect to specific, with which we have to do in
the evaluation of entropy, and therefore, that we must use
the average of H or of 7;gen , and not that of 77, as the equiva
lent of entropy, except in the thermodynamics of bodies in
which the number of molecules of the various kinds is
constant.
RETURN TO the circulation desk of any
University of California Library
or to the
NORTHERN REGIONAL LIBRARY FACILITY
Bldg. 400, Richmond Field Station
University of California
Richmond, CA 948044698
ALL BOOKS MAY BE RECALLED AFTER 7 DAYS
• 2month loans may be renewed by calling
(510)6426753
• 1year loans may be recharged by bringing
books to NRLF
• Renewals and recharges may be made
4 days prior to due date
DUE AS STAMPED BELOW
APR 2 1 2005
2 6 2005
DD20 6M 903
General Library .
University of Calif ormz
Berkeley
VC 11402
* a
UNIVERSITY OF CALIFORNIA UBRARY
.
,v\v <