(Boston Studies in The Philosophy of Science 277) Paul Needham (Auth.), Paul Needham (Eds.) - Commentary On The Principles of Thermodynamics by Pierre Duhem-Springer Netherlands (2011)

COMMENTARY ON THE PRINCIPLES
OF THERMODYNAMICS BY PIERRE DUHEM

BOSTON STUDIES IN THE PHILOSOPHY OF SCIENCE
Editors
ROBERT S. COHEN, Boston University

JÜRGEN RENN, Max Planck Institute for the History of Science
KOSTAS GAVROGLU, University of Athens
Managing Editor
LINDY DIVARCI, Max Planck Institute for the History of Science
Editorial Board
THEODORE ARABATZIS, University of Athens

ALISA BOKULICH, Boston University
HEATHER E. DOUGLAS, University of Pittsburgh
JEAN GAYON, Université Paris 1
THOMAS F. GLICK, Boston University
HUBERT GOENNER, University of Goettingen
JOHN HEILBRON, University of California, Berkeley
DIANA KORMOS-BUCHWALD, California Institute of Technology
CHRISTOPH LEHNER, Max Planck Institute for the History of Science
PETER M C LAUGHLIN, Universität Heidelberg
AGUSTÍ NIETO-GALAN, Universitat Autònoma de Barcelona
NUCCIO ORDINE, Universitá della Calabria
ANA SIMÕES, Universidade de Lisboa
JOHN J. STACHEL, Boston University
SYLVAN S. SCHWEBER, Harvard University
BAICHUN ZHANG, Chinese Academy of Science
VOLUME 277
For further volumes:
http://www.springer.com/series/5710
COMMENTARY
ON THE PRINCIPLES
OF THERMODYNAMICS
BY PIERRE DUHEM
Edited and Translated, with an Introduction, by
Paul Needham
123
Editor
Paul Needham
University of Stockholm
Dept. Philosophy
SE 10691
Stockholm
Sweden
Paul.Needham@philosophy.su.se
ISSN 0068-0346
ISBN 978-94-007-0310-0 e-ISBN 978-94-007-0311-7
DOI 10.1007/978-94-007-0311-7
Springer Dordrecht Heidelberg London New York
Library of Congress Control Number: 2011929132
© Springer Science+Business Media B.V. 2011

No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by
any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written
permission from the Publisher, with the exception of any material supplied specifically for the purpose
of being entered and executed on a computer system, for exclusive use by the purchaser of the work.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Introduction
Pierre Maurice Marie Duhem (1861–1916) held the chair of physics (changed to
chair of theoretical physics in 1895) at Bordeaux from 1894 to his death. He estab-
lished a reputation in both the history and philosophy of science as well as in science
(physics and physical chemistry). His pioneering work in medieval science opened
up the area as a new discipline in the history of science, and his La théorie physique
(Duhem 1906) is a classic in the philosophy of science which is still read and dis-
cussed today. Although his work in these two fields is now well represented in
English with a number of translations that have appeared in recent decades (Duhem
1892b, 1903, 1902, 1905–1906, 1906, 1908, 1915, 1985, 1996), there is little of his
scientific work available in English. The original manuscript of Duhem (1898) was
translated by J. E. Trevor, one of the editors of The Journal of Physical Chemistry,
for its first issue. But his work almost invariably appeared in French. The present
volume contains translations of some of his important early work in thermody-
namics, which I hope will contribute to a more balanced picture in English of the
breadth of Duhem’s publications and provide a further source of insight into his
thought.
Duhem’s first book, Le potentiel thermodynamique (1886) furthers work initiated
by Massieu, Helmholtz and Gibbs on the application of thermodynamics in chem-
istry. Clausius had shown that the second law was associated with a function of
state which he called entropy, just as the first law was associated with a more famil-
iar function of state, energy. But he made no real use of his new concept of entropy
in his own work on the development of thermodynamics, eliminating energy and
entropy as soon as possible in favour of the original thermodynamic concepts of
heat and work (Klein 1978, pp. 331–2). It was Gibbs (1876–1878) who put the new
concept of entropy to serious use in his now familiar combined form of the first and
second laws, which characterises systems entirely in terms of functions of state. This
was the form that made perspicuous these application of the science first formulated
to articulate the principles governing the working of steam engines and seemingly
having nothing to do with chemistry.
v
vi Introduction
Duhem (1886)1 exploits the application of Euler’s theorem on homogeneous

functions, from which the association of Duhem’s name with the Gibbs-Duhem
equation derives. The following year he began to explore the foundations of ther-
modynamics on which Gibbs built his theory in his “Study of the thermodynamic
works of J. Willard Gibbs” (1887), which was the first critical study of Gibbs
(1876–1878). It is translated here because it is something of a predecessor to the
principal work translated here, Duhem’s three-part study “Commentary on the prin-
ciples of thermodynamics” (1892a, 1893a and 1894), henceforth referred to simply
as Commentary. In the 1887 essay Duhem presents a first rigorous definition of the
notion of a reversible process. This was problematic because thermodynamics pro-
vides a characterisation of matter (or radiation) at equilibrium. It is paradoxical how
any process connecting two equilibrium states could itself comprise just equilibrium
states, as thermodynamicists had supposed, because once in a state of equilibrium,
the system never changes. Real processes only occur as a result of an imbalance of
forces between a system and its environment. Duhem suggested that in the limit,
now called a quasi-static process, in which the imbalance is successively reduced,
each step is an equilibrium state, but he emphasised that the totality of such steps
cannot itself represent a process, however slow. If such a limit in the process of get-
ting from state A to state B coincides with the limit for getting from B to A, Duhem
(1887, pp. 132–4)2 defines it as a reversible change (transformation réversible),
observing that this cannot be an actual change.
Duhem introduced this conception in his physical chemistry textbooks. In his
Introduction à la mécanique chimique, he began his presentation of thermodynamics
with an account of reversible change (modification réversible), emphasising that
real changes are never reversible (1893b. pp. 93–9). And in later works he devotes
several pages to carefully describing the notion (1897, pp. 56–60; 1910, pp. 59–82),
always stressing the impossibility of any such process. Rechel (1947, p. 301) was
later to bemoan the fact that the standard English term “reversible process” “contains
a contradiction within itself . . . which text book writers are prone to ignore” but
exonerates Duhem from this charge.
Duhem sharpens the definition of a modification réversible in §8 of the second
part of his Commentary (Section 6.8 here), having argued in the preceding §7 that
a sequence of equilibrium states can be regarded as a virtual change which, as dis-
tinct from a real change, doesn’t occur in time with the independent variables on
which the state of the system depends being functions of time and having determi-
nate rates of change. As Miller (1966, 1971) notes, Duhem points out in this more
extended discussion that there are circumstances in which the limiting quasi-static
processes are not the same in each direction, so that it is a substantial restricting
assumption to speak of “systems for which all changes . . . , which are sequences of
1 Jaki (1984, pp. 50–3) argues Duhem (1886) must be identical with the doctoral thesis which he
submitted in 1884 but was not accepted.
2 The original page numbers to Duhem’s works translated here are preserved in the translation, and
references to these works in this Introduction are to the original pagination.
Introduction vii
equilibrium states, are reversible changes” (Duhem 1893a, p. 307). A more compre-
hensive development of thermodynamics in which this restriction is lifted is pursued
in Duhem (1896a). Another definition given in the 1887 paper and repeated in the
first part of Commentary is noteworthy for providing for the first time a definition
of the amount of heat in terms of energy and work.
The earlier 1887 paper professed to address the circumstance that
“Mathematicians regret that the principles of Thermodynamics should have
been developed in general with so little precision that the same proposition can
be regarded by some as a consequence, and by others as a negation, of these
principles” (p. 123). This task is more earnestly pursued in the later three-part
Commentary, where he sets about articulating the theory in axiomatic form which
provides a framework in which definitions can be properly formulated and from
which he rigorously establishes the standard results of thermodynamics such as the
existence of a function serving as the entropy (Chapter 8 here). Earlier on in the
nineteenth century, mathematicians had turned to the axiomatic method in order to
clarify the foundations of the differential and integral calculus in the real number
system. But the axiomatic foundations of Euclidean geometry and the status of the
non-Euclidean geometries were only being developed around the time Duhem was
writing, and as Miller (1971) says, his application of this approach in physics was
truly pioneering in making the assumptions (what he calls conventions) explicit and
formulating them rigorously.
Duhem had another goal in his Commentary, in addition to providing a clear
and adequate foundation for thermodynamics. He was writing at a time when many
physicists accepted the vision of all physical phenomena as essentially mechanical,
and several were actively engaged in the project of demonstrating this by reducing
physical theory to mechanics. There was some opposition to this in the develop-
ment of schools of energetics. In Germany, the two leading figures, Georg Helm
(1851–1923) and Wilhelm Ostwald (1853–1932) promoted a theory of energetics
inspired by thermodynamics at the end of the nineteenth century, but with rather
different motivations. Whereas Ostwald offered a realist conception of energy as
the fundamental thing or substance in terms of which the properties of matter were
to be reduced, Helm advocated an instrumentalist or phenomenological conception.
Their theories were heavily criticised by Boltzmann and Planck at a famous meet-
ing in Lübeck in 1895 (Deltete 1999), in view of which it is somewhat surprising
that Duhem later used the term (Duhem 1911). But Duhem cannot be criticised, as
they were, for misunderstanding the basic principles of thermodynamics. Nor did
he adopt either Ostwald’s stance of treating energy as the only ultimate real object
or Helm’s phenomenal view. Although he was in agreement with them in opposing
universal reduction to mechanics, he didn’t offer an alternative form of reduction
instead (to phenomena, like Helm, or to energy, like Ostwald), and never appeals
to their writings. The person he acknowledges for introducing the term “energetics”
is Rankine (1855) (see, for example, Duhem 1896b, p. 498 or Duhem 1911, vol. I,
p. 3). He states his conception at the end of the third part of the Commentary, where
he offers a view of science as unified by supplementing mechanical conceptions
with new ones which are incorporated into a single overarching theory adequate
viii Introduction
to deal with mechanical and non-mechanical phenomena alike. Thus he brings his
three-part study to a close by surmising
It seems to us that a general conclusion arises from this study. If the science of motion ceases
to be the first of the physical Sciences in logical order, and becomes just a particular case
of a more general science including in its formulas all the changes of bodies, the temptation
will be less, we think, of reducing all physical phenomena to the study of motion. It will be
better understood that change of position in space is not a more simple change than change
of temperature or of any other physical quality. It will then be easier to get away from what
has hitherto been the most dangerous stumbling block of theoretical physics, the search for
a mechanical explanation of the Universe. (Duhem 1894, p. 285)
The English reader might well turn to the article Duhem (1898) referred to above
after these translations. It was the first rigorous proof, without any restrictions, of
Gibbs’ phase rule, which gives a condition on the number of independent intensive
variables specifying temperature, pressure and concentration of each substance in
each phase in a heterogeneous mixture. He also extends the result, giving necessary
conditions for the masses of each phase, in what has come to be called Duhem’s
theorem.
Although Duhem would make minor changes in wording when republishing a
text, he seems to have been in the habit of writing out his manuscripts for first pub-
lication without revising. This is apparent in turgid formulations and sentences that
can be convoluted, which would naturally have been rewritten. I have resisted any
such temptation and sought to reduce adjustments to a minimum in these trans-
lations, giving as literal a translation as possible. His very long sentences have
often been broken down, however, but his division into paragraphs is retained, even
though many of them comprising just one or two lines would normally be integrated
into larger paragraphs in modern English. Page numbers of the original texts are
given in square brackets, except for the numbers of the first pages, which are given
in the reference to the original name and place of publication accompanying each
article. Together with the retention of the original paragraphs, the original pagina-
tion should facilitate comparison with the original text and following references in
the secondary literature. Remarks and additions to the text in square brackets are my
own. Whenever a change has been made to a mathematical formula, this is indicated
in the footnotes.
In the 1887 paper Duhem uses both the expressions demi-force vivre and force
vivre, and in the Commentary he uses both the expressions force vivre and énergie
cinétique. Whilst this varying usage is often indicated in the footnotes, these terms
have been uniformly rendered as “kinetic energy”. The term introduced by equation
(14) of the first Commentary paper (Duhem 1892, p. 305), which he calls force vivre,
is precisely how kinetic energy is defined in modern texts.
In the Commentary Duhem uses both oeuvre and travail, each normally trans-
lated into English as work, as distinct technical terms. I have systematically
translated oeuvre as “mechanical work” and reserved the plain “work” for travail in
order to preserve this distinction in the translation. Since oeuvre figures only in the
second chapter of the first part of the Commentary (except for a sporadic appearance
Introduction ix
in the third part), the two-word English term is confined to a relatively short portion
of the translation. Duhem distinguishes between variables α, β, . . . , λ on which the
shape and motion of the system under consideration depends, and variables a, b, . . .,
l not affecting shape and motion. The discussion of oeuvre (mechanical work) in the
second chapter of the first part of the Commentary is concerned only with the for-
mer, those affecting shape and motion, and is due to the actions of bodies external
to the system. Duhem points out at the beginning of section 3 of this second chapter
that the oeuvre (mechanical work) is the increase in total energy of the system, and
includes the kinetic energy.
Since Duhem’s symbols, which with one exception I have tried to reproduce
as nearly as possible, differ from those now in common use, I enumerate here a
glossary of his principal symbols with a page reference to an occurrence in one of
the articles (indicated by year of publication). (The exception is W, for work, which
replaces Duhem’s symbol which for typographical reasons I can’t reproduce.)
Symbol Meaning Page reference
E Mechanical equivalent of heat p. 173 (1887)

F Internal thermodynamic potential p. 207 (1894)
F Internal thermodynamic potential p. 214 (1894)
P Uncompensated transformation p. 146 (1887)
But also used for pressure
Q Amount (quantity) of heat p. 6 (1887)
Q Total calorific effect p. 298 (1893)
Rα , Rβ , Calorific coefficients p. 208 (1894)
S Entropy p. 46 (1887)
But is also used to denote a state or a series of states pp. 133–4 (1887)
1, 2 Entropy (of parts of a system) p. 234 (1894)
T Kinetic energy p. 305 (1892)
τ Work done by inertial forces p. 297 (1893)
τ Instant of time p. 301 (1893)
W work p. 312 (1893)
U Internal energy p. 3 (1887)
U Internal energy p. 212 (1894)
Finally, I would like to acknowledge the help and advice on numerous points that
I have received from Robert Deltete, Donald Miller, and, last but in Swedish alpha-
betical order, Jan Österberg for their help and advice on numerous points. Don has
gone through the translation of the 1887 article and Rob the Commentary papers
(1892, 1893 and 1894), comparing them with the orginal papers and giving me
many suggestions which have been incorporated into the final version published
here. At an earlier stage, I was able to discuss many of Duhem’s less easily pene-
trated passages with Jan. To these three gentlemen I would like to extend a heartfelt
thanks.
x Introduction
References
Deltete, R. J. (1999), “Helm and Boltzmann: Energetics at the Lübeck Naturforscherver-
sammlung”, Synthese, 119, 45–68.
Duhem, P. (1886), Le potentiel thermodynamique et ses applications à la mécanique chimique et
à l’étude des phénomènes électriques, A. Hermann, Paris.
Duhem, P. (1887), “Étude sur les travaux thermodynamiqes de M. J. Willard Gibbs”, Bulletin des
Sciences Mathématiques, 11, 122–148 and 159–176.
Duhem, P. (1892a), “Commentaire aux principes de la Thermodynamique. Première Partie: Le
principe de la conservation de l’énergie”, Journal de Mathématiques Pure et Appliquées, 8,
269–330.
Duhem, P. (1892b), “Notation atomique et hypothèses atomistiques,” Revue des questions sci-
entifiques, 31, 391–457. Translated by Paul Needham as “Atomic Notation and Atomistic
Hypotheses,” Foundations of Chemistry, 2 (2000), 127–180.
Duhem, P. (1893a), “Commentaire aux principes de la Thermodynamique. Deuxième Partie: Le
principe de Sadi Carnot et de R. Clausius”, Journal de Mathématiques Pure et Appliquées, 9,
293–359.
Duhem, P. (1893b), Introduction à la mécanique chimique, Gand, Ad. Hoste, Paris.
Duhem, P. (1894), “Commentaire aux principes de la Thermodynamique. Troisième Partie: Les
équations générales de la Thermodynamique”, Journal de Mathématiques Pure et Appliquées,
10, 207–285.
Duhem, P. (1896a), “Sur les Déformations Permanentes et l’Hysteresis”, Mémoires de l’Académie
Royale des Sciences, des Lettres et des Beaux-Arts de Belgique, 54 (Présenté à la Class des
sciences dans la séance du 3 août 1895), 3–56.
Duhem, P. (1896b), “L’évolution des théories physiques du XVIIe siècle jusqu’à nos jours”,
Revue des questions scientifiques, 40, 463–499. Translated by Paul Needham as “The Evolution
of Physical Theories from the Seventeenth Century to Our Day” in Mixture and Chemical
Combination, and Related Essays, Kluwer, Dordrecht, 2002.
Duhem, P. (1897), Traité Élémentaire de Mécanique Chimique, Tome I, A. Hermann, Paris.
Duhem, P. (1898), “On the General Problem of Chemical Statics,” Journal of Physical Chemistry,
2, 1–42; 91–115.
Duhem, P. (1902), Le mixte et la combinaison chimique: Essai sur l’évolution d’une idée, C. Naud,
Paris; reprinted Fayard, Paris, 1985. Translated by Paul Needham in Mixture and Chemical
Combination, and Related Essays, Kluwer, Dordrecht, 2002.
Duhem, P. (1903), L’Evolution de la Mécanique (A. Joanin, Paris, 1903; reprinted Vrin, Paris,
1992). Translated by Michael Cole as The Evolution of Mechanics, Sijthoff & Noordhoff,
Alphen aan den Rijn, The Netherlands, 1980.
Duhem, P. (1905–1906), Les origines de la statique, 2 vols., Hermann, Paris. Translated by Grant
F. Leneaux, Victor N. Vagliente and Guy H. Wagener as The Origins of Statics: The Sources of
Physical Theory, Boston Studies in the Philosophy of Science, Kluwer, Dordrecht, 1991.
Duhem, P. (1906), La théorie physique: son objet – sa structure, 2nd ed. 1914, Marcel Rivière &
Cie, Paris; reprinted Vrin, Paris, 1981. Translated by Philip Wiener as The Aim and Structure
of Physical Theories, Princeton University Press, Princeton, 1954.
Duhem, P. (1908), ZEN TA AINOMENA: Essai sur la notion de théorie physique de Platon
à Galilée, A. Hermann et Fils, Paris, 1908. Translated by Edmund Doland and Chaninah
Maschler as To Save the Phenomena: An Essay on the Idea of Physical Theory from Plato
to Galileo, University of Chicago Press, Chicago, 1969.
Duhem, P. (1910), Thermodynamique et chimie: Leçons élémentaires, 2nd. ed., Hermann et fils,
Paris.
Duhem, P. (1911), Traité d’énergétique ou de thermodynamique générale, Gauthier-Villers, Paris;
2 vols.
Duhem, P. (1915), La Science allemande, Hermann, Paris. Translated by John Lyon as German
Science, Open Court, La Salle, Illinois, 1991.
Introduction xi
Duhem, P. (1985), Medieval Cosmology: Theories of Infinity, Place, Time, Void, and the Plurality
of Worlds, ed. and Translated by Roger Ariew, University of Chicago Press, Chicago. Selections
from Le Système du Monde.
Duhem, P. (1996), Essays in the History and Philosophy of Science, Translated and ed. by Roger
Ariew and Peter Barker, Hackett, Indianapolis and Cambridge.
Gibbs, J. W. (1876–1878), “On the Equilibrium of Heterogeneous Substances”, Transactions of
the Connecticut Academy of Arts and Sciences, 3, 108–248 and 343–520. Reprinted in Gibbs
(1948).
Gibbs, J. W. (1948), The Collected Works of J. Willard Gibbs, vol. I, Yale University Press, New
Haven.
Jaki, S. L. (1984), Uneasy Genius: The Life and Work of Pierre Duhem, Martinus Nijhoff, The
Hague.
Klein, M. J. (1978), “The Early Papers of J. Willard Gibbs: A Transformation in
Thermodynamics”, in E. G. Forbes, ed., Human Implications of Scientific Advance. Proceedings
of the XVth International Congress of the History of Science, Edinburgh, 10–15 August 1977,
Edinburgh University Press, Edinburgh; 330–341.
Miller, D. G. (1966), “Ignored Intellect: Pierre Duhem”, Physics Today, 19 (Dec.), 47–53.
Miller, D. G. (1971), “Duhem, Pierre-Maurice-Marie”, Dictionary of Scientific Biography, vol. IV,
ed. C. C. Gillispie, Scribner, New York; pp. 225–233.
Rankine, J. Macquorn (1855), “Outlines of the Science of Energetics,” Glasgow Philosophical
Society Proceedings, 3, (no. 6, 2 May), reprinted in J. Macquorn Rankine, Miscellaneous
Scientific Papers, Charles Griffin, London 1881.
Rechel, E. E. (1947), “The Reversible Process in Thermodynamics,” Journal of Chemical
Education, 24 (6), 298–301.
This is Blank Page Integra xii
Contents
Part I Study of the Thermodynamic Works of J. Willard

Gibbs (1887)
1 An Examination of the Second Law of Thermodynamics . . . . . . 3
1.1 The Principle of the Equivalence of Heat and Work . . . . . . . 3
1.2 Clausius’ Postulate . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 The Carnot Cycle and the Postulate of Sir W. Thomson . . . . . 6
1.4 Carnot’s Theorem and Absolute Temperature . . . . . . . . . . 6
1.5 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6 Uncompensated Work, Thermodynamic Potential
and Gibbs’ Theorem . . . . . . . . . . . . . . . . . . . . . . . 20
2 Historical Background and Principal Applications
of Gibbs’ Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.1 Historical Background of Gibbs’ Theory . . . . . . . . . . . . 23
2.2 Applications of Gibbs’ Theory to Dissociation Phenomena . . . 30
2.3 Application of Gibbs’ Theory to the Voltaic Cell . . . . . . . . 32
Part II Commentary on the Principles of Thermodynamics:

The Principle of the Conservation of Energy (1892)
3 Preliminary Definitions . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 On Absolute Motion . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 On Bodies and Mixtures or Combinations . . . . . . . . . . . . 38
3.3 On Bodies Isolated in Space . . . . . . . . . . . . . . . . . . . 39
3.4 On the Variables Which Define the State and Movement
of a System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.5 On Independent Systems . . . . . . . . . . . . . . . . . . . . . 43
3.6 On Temperature . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 The Principle of the Conservation of Energy . . . . . . . . . . . . . 51
4.1 The Mechanical Work and the Energy of a System . . . . . . . 51
4.2 The Kinetic Energy and the Internal Energy . . . . . . . . . . . 56
4.3 The Principle of the Conservation of Energy . . . . . . . . . . . 63
xiii
xiv Contents
5 Work and Quantity of Heat . . . . . . . . . . . . . . . . . . . . . . 65

5.1 The Establishment of a Fundamental Equation . . . . . . . . . 65
5.2 On Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.3 On the Quantity of Heat . . . . . . . . . . . . . . . . . . . . . 73
5.4 The Classical Problem of Dynamics . . . . . . . . . . . . . . . 77
5.5 Calorimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Part III Commentary on the Principles of Thermodynamics:

The Principle of Sadi Carnot and R. Clausius (1893)
6 The Carnot Cycle and Reversible Changes . . . . . . . . . . . . . . 85
6.1 Virtual Changes . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2 Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3 A Real Cycle Can Reproduce Itself Indefinitely . . . . . . . . . 87
6.4 Adiabatic, Exothermic and Endothermic Changes . . . . . . . . 88
6.5 Isothermal Changes: The Carnot Cycle . . . . . . . . . . . . . 90
6.6 Simultaneous Independent Changes: Generalisation
of the Carnot Cycle . . . . . . . . . . . . . . . . . . . . . . . . 91
6.7 Changes Which Are a Sequence of Equilibrium States . . . . . 92
6.8 Reversible Changes . . . . . . . . . . . . . . . . . . . . . . . . 94
7 Carnot’s Theorem and Absolute Temperature . . . . . . . . . . . . 99
7.1 The Hypotheses of Clausius and Sir W. Thomson . . . . . . . . 99
7.2 Addition to the Hypotheses of Clausius and Sir W. Thomson . . 100
7.3 Various Kinds of Carnot Cycles . . . . . . . . . . . . . . . . . 100
7.4 Carnot’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 102
7.5 Carnot’s Theorem (Continued): The Use of Reversible
Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
7.6 Absolute Temperature . . . . . . . . . . . . . . . . . . . . . . 114
7.7 A Definitive Statement of Carnot’s Theorem . . . . . . . . . . 117
8 Entropy and Clausius’ Theorem . . . . . . . . . . . . . . . . . . . 121
8.1 Conditions and Hypotheses . . . . . . . . . . . . . . . . . . . . 121
8.2 Calorific Coefficients of the System at Equilibrium . . . . . . . 124
8.3 Demonstration of Clausius’ Theorem . . . . . . . . . . . . . . 126
8.4 A Property of Reversible Cycles . . . . . . . . . . . . . . . . . 135
8.5 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
8.6 On the Case Excepted from the preceding . . . . . . . . . . . . 138
Part IV Commentary on the Principles of Thermodynamics:

The General Equations of Thermodynamics (1894)
9 Properties of a System at Equilibrium . . . . . . . . . . . . . . . . 143
9.1 The Internal Thermodynamic Potential . . . . . . . . . . . . . 143
9.2 Properties of a System Formed from Several
Independent Parts with the Same Temperature . . . . . . . . . . 145
9.3 The Fundamental Hypothesis; Normal Variables . . . . . . . . 152
9.4 The Problem of Equilibrium . . . . . . . . . . . . . . . . . . . 154
Contents xv
10 Properties of a System in Motion . . . . . . . . . . . . . . . . . . . 157

10.1 Meaning of the Word “Motion” . . . . . . . . . . . . . . . . . 157
10.2 Motion of a System with the Same Temperature at All Points . . 157
10.3 Calorific Coefficients of a System in Motion . . . . . . . . . . 159
10.4 Fundamental Property of the Passive Resistances . . . . . . . . 161
10.5 Systems Without Viscosity . . . . . . . . . . . . . . . . . . . . 163
10.6 Systems Formed of Independent Parts with Different
Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
11 Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
11.1 Bilateral and Unilateral Connections . . . . . . . . . . . . . . . 171
11.2 The Internal Energy of a System with Connections . . . . . . . 173
11.3 Entropy [244] and the Thermodynamic Potential
of a System with Connections . . . . . . . . . . . . . . . . . . 175
11.4 Equilibrium of a System Subject to Bilateral Connections . . . 175
11.5 Equilibrium of a System Subject to Unilateral Connections . . . 179
11.6 Motion of a System with Uniform Temperature Subject
to Bilateral Connections . . . . . . . . . . . . . . . . . . . . . 181
11.7 Motion of a System Whose Various Parts, Taken to
Different Temperatures, Are Subject to Bilateral Connections . 184
11.8 Calorific Coefficients of a System Subject to Bilateral
Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
12 Stability and Displacement of Equilibrium . . . . . . . . . . . . . . 191
12.1 On the Total Thermodynamic Potential . . . . . . . . . . . . . 191
12.2 Stability of the Equilibrium at Constant Temperature . . . . . . 192
12.3 Displacement of Equilibrium by a Variation
of Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . 193
12.4 Isothermal Displacement of Equilibrium . . . . . . . . . . . . . 196
12.5 Isentropic Stability of the Equilibrium . . . . . . . . . . . . . . 198
12.6 Isentropic Displacement of Equilibrium . . . . . . . . . . . . . 204
12.7 Specific Heat Under Constant Actions . . . . . . . . . . . . . . 206
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Part I
Study of the Thermodynamic
Works of J. Willard Gibbs (1887)∗
In 1875 Gibbs published On the equilibrium of heterogeneous substances, a mem-

oir which introduced a new method into thermodynamics. Just as Lagrange based
all statics on the single principle of virtual velocities, so Gibbs was able to base
the entire study of equilibrium in thermodynamics on a principle analogous to that
of virtual velocities and which includes the latter as a special case. In this way,
mechanics and thermodynamics are [123] joined together more closely than in the
past, and a whole section of physical science is taken to a higher degree of unity. At
the same time, at the same hands of Gibbs, the new theory confirms its fecundity by
making considerable progress, on the one hand, in the study of dissociation, and on
the other, in the study of the voltaic cell.
In the now appreciable time since their publication, Gibbs’ ideas have stimulated
much research, some of which, like that of H. von Helmholtz, are intended to con-
firm them, while others, on the contrary, have the object of criticising either the very
principles of the new theory or some of its applications. Accordingly, many minds
have yet to be made up about the value of Gibbs’ theory. Mathematicians regret that
the principles of thermodynamics are developed in general with so little precision
that the same proposition can be regarded by some as a consequence, and by others
as a negation, of these principles. Experimentalists have but little confidence in con-
sequences of a theory whose principles are disputed in this way and resist appealing
to it for the explanation of phenomena which they observe or for the prediction of
facts that it might lead them to discover.
∗ “Étude sur les travaux thermodynamiqes de M. J. Willard Gibbs”, Bulletin des Sciences
Mathématiques, 11 (1887), 122–148, 159–176.

On the equilibrium of heterogeneous substances, 1st Part [Transactions of the Connecticut
Academie [sic.] of Arts and Sciences, vol. III, Part I, pp. 108–248 (1875–1876); 2nd Part, ibid., vol.
III, Part II, pp. 343–524 (1877–1878)]. Abstract of the same (American Journal of Sciences and
Arts, vol. XVI, p. 441; 1878). On the density of vapour (American Journal of Sciences and Arts, vol.
XVIII; 1879). Graphical methods in the thermodynamics of fluids (Transact. of the Connecticut
Academy of Arts and Sciences, vol. III, Part II, p. 310; 1873). A method of geometrical repre-
sentation of the thermodynamics [sic.] properties of surfaces by means of surfaces (ibid., p. 382;
1873).
2 Part I Study of the Thermodynamic Works of J. Willard Gibbs (1887)
Under these circumstances is seems appropriate to return to the very foundations

of the principles of this controversial theory. It is true that this task is an arduous one,
because it involves redoing afresh an exposition of the second law of thermodynam-
ics. But isn’t it useful to submit the principles of the various branches of physics
to a rigorous discussion if this science is to approach more closely the precision of
mathematical sciences?
We will, therefore, in the first place show how the first principles of thermody-
namics lead to Gibbs’ new method. In the second place, we will describe the history
of the previous attempts in the same direction, and the applications which have since
been made of this method.
Chapter 1
An Examination of the Second Law
of Thermodynamics
1.1 The Principle of the Equivalence of Heat and Work

Gibbs’ [124] method is a logical consequence of Clausius’ thoughts on Carnot’s
principle. Gibbs indicated this connection of ideas himself, and we have tried else-
where to make this more precise. But several physicists have raised doubts about
these ideas of Clausius, or rather about those which serve as the point of depar-
ture for Gibbs’ work. Since it is impossible to know what an edifice is worth
before being assured of the solidity of its foundations, we must call a halt to
the discussion of Clausius’ theories and go over them again from their point of
departure.
Thermodynamics rests on two principles: the principle of the equivalence of
heat and work, and Carnot’s principle. It is the second law which is the sub-
ject of the discussions which divide many physicists today. The first, on the
contrary, presents little difficulty. Suffice it here to recall the proposition and
deduce from it a consequence which will be indispensable in the remainder of this
study.
We will suppose, first of all, that the meanings of the terms temperature1 and
amount of heat have been precisely defined. These definitions present more than
one difficulty which it would be interesting to examine. But we cannot take up this
matter here without inordinately lengthening this article.
All systems studied in thermodynamics are defined by the temperature ϑ at each
of its points, and by a limited or unlimited number of parameters α, β, . . ., such that
in order to know the entire series of modifications that the system undergoes, it is
necessary and sufficient to know the temperature of each point and the parameters
α, β, . . . as a function of time. From [125] this definition it follows that, among the
parameters α, β, . . . the coordinates of the various points of the system can be found,
but not the velocities of these points.
1 I intend here to speak of a temperature read on an arbitrary thermometer and not the absolute
temperature which will be taken up later.
P. Needham (ed.), Commentary on the Principles of Thermodynamics 3

by Pierre Duhem, Boston Studies in the Philosophy of Science 277,
DOI 10.1007/978-94-007-0311-7_1, C Springer Science+Business Media B.V. 2011
4 1 An Examination of the Second Law of Thermodynamics
An infinitesimal change of the system considered is accompanied by the release

of an amount of heatdQ. At the same time, the kinetic energy2 of the system
increases by d mv2 2 and the external forces applied to the system do a certain
work dW.3 The first law of thermodynamics consists in assuming that
mv2
dQ + Ad = −dU + dW, (1)
2
A being a constant, the calorific equivalent of work, the inverse of which is called the
mechanical equivalent of heat,4 and U is a function of the values of ϑ at each point
of the system and of α, β, . . .. This function, introduced into thermodynamics by
Clausius, is now known by the name internal energy given to it by Sir W. Thomson.
For infinitesimal changes, we can write to a second order of approximation

dW = (Xdx + Ydy + Zdz),
X, Y and Z being the components of the exterior force which act at the coordi-
nates x, y and z, and dx, dy and dz being the components of the displacement
of this point. Therefore, from the equality (1) we can derive the following
consequence:
Designating by the symbols (0) and (1) two infinitesimally close states of the
system, we imagine passing from state (0) to state (1) in two different ways, under
the action of the same exterior forces and leading to the same change of kinetic
energy,5 each of these transformations being constituted by a limited number of
infinitesimal changes. The amount of heat released is the same in each of these two
transformations.
We will not show here the importance of this remark for the exposition of thermo-
dynamics. We will in any case have occasion to make use of it in the course of this
study. We merely observe that it explains how Laplace and Poisson, who assumed
along with all their contemporaries the erroneous hypothesis that the quantity of
heat released in a change [126] depends solely on the initial and final states of the
system, were able to obtain exact results whenever they applied this hypothesis just
to infinitesimal transformations.
Such are the notions relating to the equivalence of heat and work that it will be
necessary for us to recall. Now we move on to examine the second law, which will
detain us much longer.
2 la demi-force vive.
3 [Duhem’s symbol for work is a script “T” (for travail) which for typographical reasons is replaced
here by the symbol W usual in modern thermodynamics texts.]
4 [Denoted by E later on in the paper.]
5 force vive.
1.2 Clausius’ Postulate 5
1.2 Clausius’ Postulate

Clausius based Sadi Carnot’s proposition on a very simple postulate, obtained by
generalising the most common notions concerning conduction.
Let us imagine a metal bar whose surfaces are surrounded by a non-conducting
substance which doesn’t allow any exchange of heat between the bar’s surface and
the surroundings. One end of the bar is at a base temperature ϑ, the other is sub-
mitted to the action of a source of heat with a temperature ϑ , higher than ϑ. After
a certain time, a permanent regime is established. The temperature of each point in
the bar and the state of each element of volume of the bar then remain invariable.
If, therefore, starting from the moment that the permanent regime is established, the
bar is observed for a certain time, its internal energy will not vary, the external forces
acting on it will do no work, its kinetic energy will remain equal to 0. At the same
time, the bar will have absorbed heat at its hot end and released heat at its cold end.
This is the simple observation that Clausius6 has generalised in such a way as to
give the following fundamental postulate.
Suppose that a system undergoes a change subject to the following four
restrictions:
1. The internal energy of the system has the same value at the beginning and the
end of the change. [127]
2. The kinetic energy of the system has the same value at the beginning and the end
of the change.
3. The external forces applied to the system during the course of the change produce
as much positive as negative work.
4. The exchange of heat between the system and the surroundings takes place exclu-
sively either while all the points of the system have the same temperature ϑ, or
while all the points of the system have the same temperature ϑ higher than ϑ.
Under these conditions, it is impossible that the system has absorbed heat at the
temperature ϑ in order to release it at the temperature ϑ .
This is, in its most precise form, Clausius’ postulate. After having been discussed
for a long time, it is now accepted by all physicists. We will allow ourselves to
introduce a slight modification which, although apparently trifling, suffices in fact
to eliminate all the difficulties which can be raised against certain of Clausius’ ideas.
This modification will, moreover, have the advantage of approaching more strictly
the statement of the postulate about conduction phenomenon which suggested the
idea. The new form that we propose for Clausius’ postulate is the following:
In a change subject to the restrictions indicated, the system necessarily absorbs more heat
at the temperature ϑ than it releases at the same temperature, and necessarily releases
more heat at the temperature ϑ than it absorbs at the same temperature.
6 R.Clausius, Poggendorff’s Annalen der Physik und Chemie, vol. LXXIX; 1859. Mémoires sur la
theéorie mécanique de la chaleur. Trad. Folie, vol. I, p. 54.
1.3 The Carnot Cycle and the Postulate of Sir W. Thomson

A system is said to traverse a closed cycle when it undergoes a series of changes
which bring it back to its initial state with its initial kinetic energy.
A closed cycle is called a Carnot cycle when it satisfies the following restriction:
While traversing the cycle, the exchange of heat between the system and the
environment only takes place when all the points of the system are [128] at the same
temperature ϑ, or else when all the points of the system have the same temperature
ϑ higher than ϑ.
A steam engine in which all the points of the furnace are at the same temperature
ϑ , while all the points of the condenser are at the same temperature ϑ lower than
ϑ , in which, moreover, the cylinder is impermeable to heat, provides an image of a
system traversing a Carnot cycle.
A system which traverses a Carnot cycle in such a way that during the cycle
the external forces acting on the system do negative work provides an image of the
simplest conceivable heat engine.
Generalising our experience of the most common heat engines, Sir W. Thomson7
has stated the following postulate:
When a system describes a Carnot cycle during which the external forces acting on the
system perform a total negative work, it is impossible for the system to absorb more heat at
the temperature ϑ than it releases at the same temperature.
We will subject Sir W. Thomson’s postulate to a very slight modification,

analogous to that introduced to Clausius’ postulate, and say:
When a system describes a Carnot cycle during which the external forces acting on the
system perform a total negative work, the system releases more heat at the temperature ϑ
than it absorbs at the same temperature.
Sir W. Thomson had proposed the postulate which we have just adjusted as a
proposition equivalent to Clausius’ postulate. We will, in the following exposition,
make use of both postulates at the same time.
1.4 Carnot’s Theorem and Absolute Temperature

We propose [129] first of all to introduce a classification of Carnot cycles whose
existence is compatible with the first law of thermodynamics and with the two
postulates that we have stated.
We designate by Q the total amount of heat exchanged between the system and
the surroundings while all the points are at the temperature ϑ, and by Q the total
amount of heat exchanged between the system and the surroundings while all the
points are at the temperature ϑ . We take the two quantities Q and Q to be positive
7 W. Thomson, Transactions of the Royal Society of Edinburgh, vol. XX, p. 255; 1851.
1.4 Carnot’s Theorem and Absolute Temperature 7
when they represent heat released from the system, and negative when they represent
heat absorbed by the system.8
We designate the work done during the traversal of the cycle by the external
forces acting on the system by W.9
We then have, by virtue of the first law,
W = E(Q + Q ) (2)
From this follow the propositions:
1. If the external work is zero, the two quantities Q and Q have opposite signs and
equal absolute values.
2. If the external work is positive, then at least one of the quantities Q and Q is
positive, and if only one, then that is the one with the greater absolute value.
3. If the external work is negative, then at least one of the quantities Q and Q is
negative, and if only one, then that is the one with the greater absolute value.
On the other hand, the two postulates of Clausius and Sir W. Thomson lead to
the following proposition:
When the external work is zero or negative, the quantity Q is necessarily positive.
From these propositions, it is easy to conclude that all Carnot cycles can be
classified in the following way:
1. The work done by the external forces is zero; Q is positive, Q is negative and
equal to Q in absolute value.
2. The work done by the external forces is negative; Q is positive, Q is negative
and greater than Q in absolute value.
3. [130] The work done by the external forces is positive; three cases can then arise:
a. Q is positive, Q is negative and less than Q in absolute value.
b. Q is negative, Q is positive and greater than Q in absolute value.
c. Q and Q are positive.
Now we propose to compare, for the various kinds of cycles, the values of the
ratio
Q + Q
ρ= ,
Q
and to demonstrate that, for cycles described between the temperatures ϑ and ϑ
corresponding to a positive or zero external work, the ratio ρ is greater than
8 [It is more usual in modern thermodynamics to define Q, the heat absorbed by the system from
the surroundings, as positive.]
9 [Again, it is more usual in modern thermodynamics to define W, the work done by the system on
the surroundings, as positive.]
for cycles described between the same temperatures corresponding to a negative

external work.
This theorem is evident, except for cycles with positive external work where Q is
negative and Q is positive. In fact, for all the other cycles where the external work
is positive, the ratio ρ is positive. It is zero for cycles where the external work is
zero, and negative for cycles where the external work is negative.
Let us therefore compare a cycle described between the temperatures ϑ and ϑ
where the external work is positive, the quantity Q is negative and the quantity Q
is positive, with a cycle described between the temperatures ϑ and ϑ where the
external work is negative, the quantity Q is positive and the quantity Q is negative.
Designating the values of the quantities W, Q and Q for the first cycle by W1 , Q1
and Q 1 , and for the second cycle by W2 , Q2 and Q 2 , we wish to demonstrate that
Q 1 + Q1 Q + Q2
> 2 .
Q1 Q2
Let m1 /m2 be a commensurable number greater than or equal to W1 /|W2 |,10 where
m1 and m2 are whole numbers.
Let [131] us consider a system comprising m2 systems identical to that which
describes the first cycle and of m1 systems identical to that which describes the
second cycle. The m2 first systems are made to describe the first cycle and the m1
systems the second cycle. The total system describes a Carnot cycle in which the
exchanges of heat still take place at temperatures ϑ and ϑ . The work done by the
external forces during the course of the cycle will have the value
m2 W1 + m1 W2 .
From the way m1 and m2 were chosen, it is negative, unless the ratio W1 /|W2 |
is commensurable and m1 /m2 is identical with this ratio. In this case, the external
work in question will be zero. In every case, in view of the fundamental postulates,
the cycle in question should release heat at the temperature ϑ. We therefore have
m2 Q1 + m1 Q2 > 0.
From this inequality, we deduce
m2 m1
< .
Q2 |Q1 |
Putting
m1 W1
= + ε,
m2 |W2 |
10 The notation |W2 | signifies the absolute value of W2 .

ε will be zero or a positive quantity that can be made as small as one may wish. The
preceding inequality becomes
|W2 | W1 + ε |W2 |
< .
Q2 |Q1 |
This inequality cannot tend towards an equality when ε tends towards 0. We
therefore have
W1 W2
>
Q1 Q2
or, in virtue of the equalities
W1 = E (Q1 + Q1 )
W2 = E (Q2 + Q2 )
given [132] by the first law of thermodynamics,
Q 1 + Q1 Q + Q2
> 2 .
Q1 Q2
This is precisely the inequality we wished to demonstrate.
Summarising, if all the Carnot cycles are considered for which the exchanges of
heat take place at the same temperatures ϑ and ϑ , ϑ being greater than ϑ, then
for all those cycles that
correspond to a negative work by the external forces, the
ratio ρ = (Q + Q) Q takes a value less than a certain negative quantity A. For
those that correspond
to a positive or zero work by the external forces, the ratio
ρ = (Q + Q) Q takes a value greater than a certain negative quantity, A . A is
greater than or equal to A.
We will now demonstrate that
A = A.
In order to arrive at this important proposition, it is necessary for us to introduce

the concept of a reversible transformation. This concept, one of the most important
in thermodynamics, is at the same time one of those least easy to make precise. It
will therefore be necessary for us to dwell on this for some time.
Let us imagine a system having the same temperature at all its points, under the
action of certain external forces and subject to certain connections11 expressed by
the equalities and inequalities between the various parameters defining the state of
the system. We will say that such a system is in equilibrium in a certain state if, put
in this state without kinetic energy, it remains there perpetually.
Let (1) and (2) be two states of equilibrium of the same system. Suppose that an
infinite series of its equilibrium states can be put in a linear series beginning with
11 liaisons.
the state (1) and finishing with state (2) in such a way that the following conditions
are met:
1. The parameters which determine the state of the system undergo [133] variations
which are continuous and compatible with the connections on passing from one
state to the neighbouring state.
2. The external forces acting on the system vary continuously in magnitude and
direction.
3. The equalities and inequalities expressing the connections of the system undergo
continuous variations of form.

Such a series will be called a continuous series of equilibrium states.
If the external forces acting on the system in the state of equilibrium (1), and
the connections to which it is subjected in this state, are changed, or even if it is
simply placed in this state without its kinetic energy becoming zero, the system will
no longer be able to remain perpetually in this state. It will undergo changes.
Suppose that it is possible to arrange infinitesimal initial velocities12 of various
points, of external forces and of the connections in such a way as to form an infinity
of ways of passing from state (1) without kinetic energy to state (2) without kinetic
energy. Each of these transitions is constituted of a linear and continuous series of
states. Let S, S , . . . be these series, whose number is supposed to be unlimited.
We also suppose, analogously, that it is possible to set up an infinity of ways
of passing from state (2) to state (1), corresponding to an infinity of linear and
continuous series s, s , . . . of states of the system.
Now suppose that amongst the series S, S , . . . it is possible to choose in at least
one way an infinity of them which, along with the series , can be arranged as a
series in linear sequence possessing the following properties:
It is possible to establish a unique correspondence between a state of each of the
series and a state of each of the others and of the series .
The ensemble of states corresponding to the same state of the series forms a
linear and continuous series of states ending in the state of the series , and whose
kinetic energy tends towards 0.
Let [134] us suppose that the series s, s , . . . exhibits a property analogous to that
we have just supposed for the series S, S , . . ..
We will briefly express all the properties that we have just attributed to the series
by saying that it constitutes a reversible change allowing the passage either from
state (1) to state (2) or from state (2) to state (1).
We will say that a reversible change is isothermal when it is possible to form
two continuous sequences of isothermal changes from, on the one hand, theseries
S, S , . . ., and on the other hand, the series s, s , . . ., having the sequence as a
common limit, in the way we have indicated. It is clear that such a sequence is
12 When the initial speeds are infinitesimal, the initial kinetic energy is infinitesimal to the second
order. It can therefore be regarded as zero.
formed from equilibrium states all corresponding to one and the same temperature
of the system.
We will say that a reversible change is adiabatic13 when it is the same com-
mon limit of two sequences of adiabatic changes in oppossite directions. Taking two
neighbouring states from a reversible sequence , the two states will correspond
to the values U and U + dU of the internal energy of the system. If the system is
transported from one of these states to the other, the external forces will perform
a virtual work dW. In general, the amount of heat released in a reversible passage
from one of these states to the other will be referred to by
dQ = A dW − dU.
This quantity does not represent an actual release of heat because a reversible
change does not represent an actual change. It is only the limit of the amount
of heat
released in one of the changes S or s which have as their limit the sequence during
which the interval is crossed between the two states having as their correspondents
the two states considered in the sequence .
Following the definition just given, in the case of a reversible adiabatic change,
we have
dQ = 0.
[135] Having stated these definitions, we now return to the properties of the
Carnot cycle.
Given two temperatures, ϑ and ϑ , ϑ being greater than ϑ, let us suppose that it
is always possible to find four reversible changes of a certain system exhibiting the
following properties:
The first is a reversible isothermal change 21 corresponding to the temperature
ϑ , and taking the system from a state (1) to another state (2).
The second is a reversible adiabatic change 32 taking the system from the state
(2), where it has the temperature ϑ , to the state (3) where its temperature is ϑ.
The third is a reversible isothermal change 43 taking the system from a state (3)
to another state (4) at the same temperature ϑ.
The fourth is a reversible adiabatic change 14 returning the system from the
state (4) to the state (1).
We have thus constructed a reversible Carnot cycle described between the
temperatures ϑ and ϑ .
In accordance with the definition of reversible transformations, there are
infinitely many ways of passing from state (1) to state (2), among which are found
an infinity of isothermal changes S21 , S 21 , . . . forming a linear series whose limit is

the sequence 21 . Similarly, there are infinitely many ways of passing from state (2)
to state (1), among which are found an infinity of isothermal changes S12 , S 12 , . . .
13 An adiabatic change is a change in which heat is neither absorbed nor released at any time.

forming a linear series whose limit is the sequence 12 . Analogous considerations
3 3
apply to the three other reversible changes 2 , 2 and 14
Let us take one of the isothermal changes S21 , one of the adiabatic changes S32 , one
of the isothermal changes S43 , and one of the adiabatic changes S14 . We have a Carnot
cycle passing through the four states (1), (2), (3) and (4) and described between the
temperatures ϑ and ϑ in the direction (1), (2), (3), (4). Clearly, it is possible to form
infinitely many similar cycles, and it is possible, in infinitely many different ways,
to form from these cycles a linear series of cycles having the reversible cycle as a
limit. Let C(1,2,3,4) be such a series of Carnot cycles.
In the same way, it is possible, from the changes S41 , S34 , S23 and S12 to form [136]
an infinite linear series of Carnot cycles, described between the temperatures ϑ and
ϑ passing through the four states (1), (2), (3) and (4) in the direction (4), (3), (2),
(1), transforming continuously from the one to the other and having the reversible
cycle as a limit. Let C(4,3,2,1) be such a series of Carnot cycles.
The virtual work which the external forces would carry out if the system were
made to describe a reversible cycle changes sign without changing its absolute value
according as the cycle is described in one direction or the other. In order to fix our
ideas, we suppose it to be positive when the cycle is described in the direction (1),
(2), (3), (4), and designate it by .
The work done by the external forces in the course of traversing one of the cycles
in the series C(1,2,3,4) tends towards when the cycle tends towards the reversible
cycle. One can therefore be certain, when considering only those cycles which are
quite close to the reversible cycle, that all the external work is positive.
The work done by the external forces in the course of traversing one of the cycles
in the series C(4,3,2,1) tends towards – when the cycle tends towards the reversible
cycle. One can therefore be certain, when considering only those cycles which are
quite close to the reversible cycle, that all the external work is negative.
It follows that, for all the cycles in the series C(1,2,3,4), that
Q + Q
ρ= > A
Q
and, for all the cycles in the series C(4,3,2,1), that
Q + Q
ρ= < A,
Q
with
A ≥ A.
But if we consider the value of ρ for all the cycles in the series C(1,2,3,4), it is
clear that it has in the limit the value of ρ for the reversible cycle described in the
direction (1), (2), (3), (4). And if we consider the value of ρ for all the cycles in the
series C(4,3,2,1), it is clear that it has in the limit the value of ρ for the reversible
cycle described in the direction (4), (3), (2), (1). Referring to the definition of the
amount of heat released in the course of a reversible change, Q and Q change signs
without [137] changing their absolute value when changing the direction in which
the reversible cycle is described. Consequently, for the reversible cycle, the value of
ρ is independent of the direction in which the cycle is traversed.
From this it follows that the value of ρ for all the cycles in the series C(1,2,3,4)
and the value of ρ for all the cycles in the series C(4,3,2,1) have a common limit.
Consequently, we have, as stated above,
A = A.
The common value of the quantities A and A is the value of the quantity ρ for
any reversible cycle between the temperatures ϑ and ϑ . This quantity is necessarily
a function of the two temperatures ϑ and ϑ alone. Designating this by f (ϑ, ϑ ), we
arrive at the following result:
There exists a function f (ϑ, ϑ ) of the two temperatures ϑ and ϑ , negative when ϑ is
greater than ϑ, such that, for all Carnot cycles described between the temperatures ϑ and
ϑ , and corresponding to a positive or zero external work,
Q + Q
> f (ϑ, ϑ ),
Q
and for all Carnot cycles described between the temperatures ϑ and ϑ , and corresponding
to a negative external work,
Q + Q
< f (ϑ, ϑ ).
Q
We put
ψ(ϑ , ϑ) = f (ϑ, ϑ ) − 1
The quantity ψ(ϑ , ϑ) will be negative and greater than 1 in absolute value if ϑ
is greater than ϑ. Accordingly, for the first class of Carnot cycles,
Q
> ψ(ϑ , ϑ), (3)
Q
and for the second,
Q
< ψ(ϑ , ϑ). (4)
Q
It [138] remains, in order to demonstrate Carnot’s theorem, to show that
ψ(ϑ , ϑ) = ψ(ϑ , ϑ)ψ(ϑ , ϑ), (5)
ϑ, ϑ , ϑ being any three temperatures.

In order to establish this proposition, it suffices to observe, by means of a

reversible Carnot cycle functioning between temperatures ϑ and ϑ bringing into
play at these two temperatures the amounts of heat Q and Q , and a reversible Carnot
cycle functioning between temperatures ϑ and ϑ bringing into play at these two
temperatures the amounts of heat Q and Q , that it is possible to form a reversible
Carnot cycle functioning between temperatures ϑ and ϑ bringing into play at these
two temperatures the amounts of heat Q and Q , The deduction is too simple to
delay us.
Designating a function of ϑ, always positive and increasing with ϑ, by F(ϑ), it
follows from the equality (5) that we can put
F(ϑ )
ψ(ϑ , ϑ) = − . (6)
F(ϑ)
The function F(ϑ) is evidently determined only up to a positive constant factor.

It can be designated by the name absolute temperature.14
By means of this equality (6) and the inequalities (3) and (4), we arrive at the
following proposition:
All Carnot cycles described between temperatures ϑ and ϑ , ϑ being higher than ϑ, are
divided into two classes.
The first class is formed from those cycles corresponding to a positive or zero work done
by the external forces. For all these cycles, it holds that
Q F(ϑ )
>− .
Q F(ϑ)
The [139] second class is formed from those cycles corresponding to a negative work done
by the external forces. For all these cycles, it holds that
Q F(ϑ )
<− .
Q F(ϑ)
These two propositions can be united in a single statement. Recalling the classi-
fication of Carnot cycles introduced at the beginning of this section, it can be seen
that, for all possible Carnot cycles, it holds that
Q Q
+ > 0. (7)
F(ϑ) F(ϑ )
This is the inequality to which we will give the name Carnot’s Theorem.
14 It will be noticed that the absolute temperature is presented here by a definition analogous to that
given by G. Lippmann (G. Lippmann, Journal de Physique théorique et appliquée, 2nd. series,
vol. III, pp. 52 and 277; 1884. Cours de Thermodyamique: 1886).
1.5 Entropy 15
1.5 Entropy
We consider two arbitrary states of the same system, designated by (1) and (2), and
we suppose that, in each of these two states, the system has the same temperature at
all of its points. This temperature needn’t necessarily be the same in state (1) and in
state (2). We will suppose, at the same time, once and for all, that it is the same in
all the states that we might have to consider.
Suppose that we can pass from state (1) to state (2) in infinitely many ways. Each
of these manners of transformation is constituted by an infinite linear series S21 of
states of the system, the system being, in each of these states, subject to the action
of certain external forces, constrained by certain connections and endowed with a
certain kinetic energy. Further, we suppose that among the series S21 , we consider
only those which satisfy the following condition: The system, being without initial
kinetic energy in any one of the states constituting S21 and remaining subject to
the same external forces, can pass into any one of the states which follow it in the
series S21 ; and, with the addition of certain appropriately chosen connections, we
can arrange that its kinetic energy, zero in the first state, will remain zero in the
second. This is a fundamental restriction which we will suppose is always fulfilled
in what follows.
Suppose, [140] in the same way, that we can pass from state (2) to state (1) in
infinitely many ways, each of these manners of transformation being constituted by
a linear series S12 of states of the system, satisfying restrictions analogous to those
imposed on the series S21 .
The states through which the system passes, either when it goes from state (1) to
state (2) or when it goes from state (2) to state (1), are not in general states of
equilibrium for the system subject to external forces really acting on it when it
passes through each of these states. But we will assume that, for each of these states,
it is possible to imagine external forces and connections such that the states become
the equilibrium states for the system. It will then be possible to suppose that there
exist reversible modifications leading from any one of these states to any one of the
others.
In order to simplify the following demonstrations, we will assume that, in each
of the series S21 and in each of the series S12 the system never passes more than once
through the same state. We will easily be able to show later that the results obtained
do not depend on this simplifying assumption.
Next we will assume that, given any series S21 and any series S12 , it is possible to
establish a unique correspondence between the states composing the one series with
the states composing the other series, this correspondence satisfying the following
conditions:
1. Two states infinitesimally close in the series S21 correspond to two states
infinitesimally close in the series S12 .
2. If, in the series S21 the state A precedes the state B, then in the series S12 the state
A corresponding to A follows the state B corresponding to B.
3. Given two corresponding states, it is possible to pass from the one to the other
by a reversible adiabatic change.
These are the rather numerous restrictions that we will assume realised by the
changes that we will study. Only after it is assured that they are fulfilled in a par-
ticular case do we have the right to apply to that case the theorems we are going to
demonstrate. Should their large number be alarming, we can only say that this large
number of restrictions to which [141] the theorems of thermodynamics are subjected
makes it all the more necessary to state these restrictions precisely, because each of
these restrictions becomes a source of errors when it is ignored, just as we ordinarily
do, under the veil of false evidence.
Let A and B be two infinitesimally close states in the series S21 . When the system
passes from the state A tothe state
B, it releases an amount of heat dQ. Its kinetic
energy15 increases by d mv2 2. Let ϑ be the temperature of the system in the
state A. We form the quotient
mv2
1
dQ + A d ,
F(ϑ) 2
and designate by
(2) mv2
1
J12 = dQ + A d (8)
(1) F(ϑ) 2
the sum of analogous quotients for all the elements of the series S21 .
Similarly, let A and B be two infinitesimally close states in the series S12 . When
the system passes from the state B to the state A , it releases an amount of heat

dQ .16 Its kinetic energy17 increases by d mv 2 /2. Let ϑ be the temperature of
the system in the state A . We form the quotient

1 mv 2

dQ + A d ,
F(ϑ ) 2
and designate by

(1) 1 mv 2
J21 =
dQ + A d (8a)
(2) F(ϑ ) 2
the sum of analogous quotients for all the elements of the series S12 .
15 demi-force vive.
16 [dQ in the original.]
17 demi-force vive.
1.5 Entropy 17
We will demonstrate that it always holds that
J21 + J21 > 0. (9)
Since it can always be supposed that the two elements AB, A B are two corre-
sponding elements of the series S21 and S12 , it suffices to demonstrate that, for two
such corresponding elements, [142] it holds that
mv2 mv 2
dQ + A d dQ + A d
2
+ 2
> 0. (10)
F(ϑ) F(ϑ )
This is quite easily done in the following way.

We have supposed that we can make the system begin in the state A without
kinetic energy and arrive in state B without kinetic energy, while allowing the
external forces applied during the transformation AB to act, but introducing cer-
tain connections where necessary. We designate this new change by A1 B1 . During
the change AB and the change A1 B1 , the external forces perform the same work and
the internal energy undergoes the same variation. Therefore, designating by dQ1 the
amount of heat released by the system during the change A1 B1 , we have
mv2
dQ1 = dQ + A d . (11)
2
Similarly, it is possible for the system to pass from state B to state A with the
kinetic energy equal to 0 in both of these two states, by preserving the external
forces acting during the transformation B A , but introducing certain connections.
Designating this new change by B 1 A 1 , and the amount of heat released during the
change B 1 A 1 by dQ 1 , we have
mv 2
dQ 1 = dQ + A d . (11a)
2
In view of the equalities (11) and (11a), the inequality to be demonstrated, (10),
becomes
dQ1 dQ 1
+ > 0. (12)
F(ϑ) F(ϑ )
It is possible, in accordance with what has been assumed, to pass from state A
to the state A by a reversible adiabatic series, which we will designate by AA .
Similarly, it is possible to pass from state B to the state B by a reversible adiabatic
series, which we will designate by BB .
This being the case, we note that there exists either a state B2 in the series BB
infinitesimally close to B in which the system has the same [143] temperature ϑ as
in the state A, or else a state A2 in the series AA infinitesimally close to A in which
the system has the same [143] temperature (ϑ + dϑ) as in the state B. In order to
fix our ideas, we suppose that the first case obtains.
Similarly, there exists either a state A 2 in the series

AA infinitesimally
close
to A in which the system has the same temperature ϑ + dϑ as in the state B ,

or else a state B 2 in the series BB infinitesimally close to B in which the system
has the same [143] temperature ϑ as in the state A . In order to fix our ideas, we
suppose that the first case obtains.
Here we introduce a new hypothesis implicitly accepted by all the treatises on
thermodynamics but which must be verified in each particular case. This hypothesis
can be formulated in the following way.
When the system is subject to the action of external forces acting during the
transformation AB or A1 B1 , it is possible to find connections which force the sys-
tem, departing without kinetic energy from state A, to arrive also without kinetic
energy in state B2 by an infinitesimal isothermal transformation A1 B2 performed at
the temperature ϑ. Further, it is possible to find connections such that the system,
leaving state B2 without initial kinetic energy, arrives also without kinetic energy in
the state B by an infinitesimal adiabatic transformation B2 B1 .
If such a hypothesis is satisfied, it follows from the remark made in Section 1.1
in connection with the first law of thermodynamics that the collection of the two
infinitesimal changes A1 B2 and B2 B1 release the same amount of heat as does the
infinitesimal change A1 B1 . Moreover, since the change B2 B1 is adiabatic, the two
changes A1 B2 and B2 B1 give rise to the same release of heat dQ1 .
Similarly, if such a hypothesis is satisfied for the states B , A and A 2 , the change
B 1 A 1 gives rise to the release of the same heat dQ 1 as an isothermal change

B 1 A 2 , carried out at the temperature (ϑ + dϑ ), and takes the system from the state
B without initial kinetic energy to the final state A 2 without final kinetic energy.
Given that the state B2 is within the reversible adiabatic change BB , the change
B2 B is itself adiabatic and reversible. Consequently, [144] it is possible to take the
system in state B2 without kinetic energy in infinitely many ways, and make it reach
B without kinetic energy by an adiabatic change. We designate one such change by
B2 B 1 .
Similarly, we designate by A 2 A1 one of the adiabatic changes by which it is pos-
sible to take the system without kinetic energy in the state A 2 and carry it without
kinetic energy into the state A.
The series of four changes A1 B1 , B2 B 1 , B 1 A 2 , A 2 A1 constitutes a Carnot
cycle described by the system. If we now apply the inequality (7), true of all Carnot
cycles, to this cycle, we have
dQ1 dQ 1
+ >0
F(ϑ) F(ϑ + dϑ )
or simply
dQ1 dQ 1
+ > 0.
F(ϑ) F(ϑ )
This inequality is precisely the inequality (12) that was to be demonstrated.
1.5 Entropy 19
We have seen that the inequality (12) is equivalent to (10), and that this entails
the inequality (9). We therefore have, in virtue of the equalities (8) and (8a),

(2) 1 mv2 (1)
1 mv 2
dQ + A d +
dQ + A d > 0.

(1) F(ϑ) 2 (2) F(ϑ ) 2
(13)
This inequality is easily transformed.
The symbol

(2) 1 mv 2

dQ + A d
(1) F(ϑ ) 2
for a transformation such as S12 which begins in the state (2) and ends in state (1) has
not, as yet, been given any sense. It is convenient to attribute to it the same sense as

(1) 1 mv 2
−
dQ + A d .
(2) F(ϑ ) 2
The [145] inequality (13) becomes

(2) 1 mv2 (2) 1 mv 2
dQ + A d >
dQ + A d (13a)
F(ϑ) 2
(1) (1) F(ϑ ) 2
and can be formulated in the following way:

The sum
(2) mv2
1
dQ + A d
(1) F(ϑ) 2
is greater for any transformation in which the system passes from state (1) to state (2) than
for any transformation in which the system passes from state (2) to state (1).

Now let us suppose that there exists a reversible transformation from state
(1) to state (2), which is the common limit of an infinity of transformations S21
and an infinity of transformations S12 . It will then be quite easy for us to demon-
strate the following proposition, which is the fundamental proposition of this part of
thermodynamics:
There exists a quantity F(1, 2) uniquely dependent on the state of the system in (1) and the
state of the system in (2) such that, for all transformations by which the system passes from
(1) to (2), we have
(2) mv2
1
dQ + A d − F(1, 2) > 0,
(1) F(ϑ) 2
and for each transformation by which the system passes from (2) to (1), we have
(2) mv2
1
dQ + A d − F(1, 2) < 0.
(1) F(ϑ) 2
This quantity F(1, 2) is the value of the integral

(2) dQ
,
(1) F(ϑ)
for [146] any reversible transformation whatsoever which connects the states (1) and (2).
By very simple considerations which are unnecessary to present here, we arrive

at the following proposition:
There exists a function S of the parameters defining the state of the system such
that, if S1 and S2 designate the values of S relative to the states (1) and (2), we have
F(1, 2) = S1 − S2
It is this function S, completely defined up to an additive constant as above, that

Clausius has proposed to call the entropy of the system. This name is now adopted
by all physicists. If, for any change whatsoever transporting the system from state
(1) to state (2), we give with Clausius the name uncompensated transformation to
the quantity
(2) mv2
1
P= dQ + A d + S2 − S1 , (14)
(1) F(ϑ) 2
then the preceding proposition becomes:

All changes correspond to a positive uncompensated transformation.
This is the great law for which thermodynamics is indebted to Clausius. Having
brought it to the fore, we will have no more than a word to say in deducing the
fundamental theorem on which Gibbs’ work rests.
1.6 Uncompensated Work, Thermodynamic Potential

and Gibbs’ Theorem
Clausius’ law gives us the means of determining the equilibrium states of the sys-
tem. If, from a state of the system, all the virtual changes that can be imagined
would engender a zero or negative uncompensated transformation, the system is
certainly at equilibrium. But this proposition leads above all to interesting results
when applied to systems whose temperature ϑ is maintained constant. In that case,
we call [147] the product
= E F(ϑ) P. (15)
the uncompensated work.

1.6 Uncompensated Work, Thermodynamic Potential and Gibbs’ Theorem 21
We will be able to state the preceding proposition in a form like this:

A system will certainly be at equilibrium in a definite state if all virtual changes from this
state imposed on the system correspond to a zero or negative uncompensated work.
This theorem is analogous, in its statement, to the principle of virtual velocities

that Gauss has formulated. Far from being a simple coincidence, this analogy arises
from the very nature of things, for it would be easy to prove that, if the restrictive
hypotheses to which mechanics subjects all the systems that it studies were intro-
duced into thermodynamics, this proposition would reduce to the statement of the
principle of virtual velocities. Despite this important analogy, the preceding propo-
sition would be of little use if the uncompensated work performed in an isothermal
change were not susceptible to a remarkable expression as we are going to show.
When the temperature is maintained constant, the equalities (14) and (15)
give us
2 mv2
=E dQ + Ad + E F(ϑ) (S2 − S1 ).
1 2
But on the other hand, if we designate the work done by the external forces during
the change under consideration by W, equality (1) gives us
2 mv2
E dQ + Ad = E(U2 − U1 ) + W.
1 2
Summarising, we have the expression
= E[U1 − F(ϑ)S1 ] − E[U2 − F(ϑ)S2 ] + W. (16)
This expression is particularly remarkable in the case where the external forces
allow, either by themselves or in virtue of the connections imposed on the sys-
tem, what is called in mechanics a potential. [148] In this case, if the potential
is designated by V,18 whose values are V1 and V2 in state (1) and state (2), then
W = V1 − V2 . (17)
Then, putting
= E[U − F(ϑ)S] + V, (18)
the equality (16) becomes
= 1 − 2 . (19)
18 [Duhem uses the symbol “W” for the potential.]

The uncompensated work will be determined by means of the quantity just

as external work W is determined by means of the potential V. The quantity can
therefore, with some justification, be called the thermodynamic potential.
This name will be even more appropriate when the form taken by the rule just
indicated for the determination of stable equilibrium states is noted. This rule can in
fact be stated thus:
A state of a system will be a stable equilibrium state if it corresponds to a minimum of the
thermodynamic potential.
This is the fundamental theorem discovered by Gibbs. It is the analogue in

thermodynamics of the proposition on the stability of equilibrium discovered by
Lagrange and demonstrated by Lejeune-Dirichlet. Suffice it to say that it consti-
tutes the origin of one of the greatest steps forward that have been made in this
science, whether in the clarification of its principles or enriching the number of its
applications.
Chapter 2
Historical Background and Principal
Applications of Gibbs’ Theory
2.1 Historical Background of Gibbs’ Theory

We have seen how the fundamental postulates of thermodynamics lead to Clausius’
ideas about Carnot’s principle, [160]∗ perhaps the most important to have been intro-
duced into natural philosophy during the second half of our century. We will not
retrace here the history of the research which has led physicists from the work of
Sadi Carnot to these ideas. Clausius has himself taken the trouble to record the
principal points of this history,1 and it would be rash to rival his competence and
impartiality. We will therefore restrict ourselves to following the path which leads
from the ideas of Clausius to the beautiful and fruitful method of Gibbs.
This path will be concerned with studies of chemical mechanics involving ther-
modynamics. The discovery of the phenomena of dissociation by H. Sainte-Claire
Deville instigated numerous important works in the domain of chemical mechanics
revealing the existence of equilibrium states closely related to those whose laws
physics traces by studying vaporisation and fusion. But they are rendered more
complex by the large number of substances2 among which such an equilibrium is
established, and by the large number of variables which can influence the state of
these substances. It was difficult to apply synthetic methods to these complex phe-
nomena, based exclusively on the employment of closed cycles, which have led to
such important propositions in the study of vaporisation and fusion.
These methods have not, however, been condemned to futility before the new
order of just discovered phenomena. J. Moutier,3 who made the first applications of
thermodynamics to dissociation phenomena, has shown that synthetic reasoning can
lead to important consequences in the domain of chemical mechanics. Applying first
∗ [The second part of the article continues later on in the journal, beginning on p. 159.]
1 R. Clausius, Zur Geschichte der mechanischen Wärmetheorie (Poggendorff’s Annalen der Physik
und Chemie, Bd. CXLV, p. 496; 1872). See also R. Clausius, Die mechanische Wärmetheorie.
Zweitte Auflage, t. I, Abschnitt XIII.
2 substances.
3 J. Moutier, Sur quelques relations entre la Physique et la Chimie (Encyclopedie chimique de
Fremy, Introduction, t. II).

24 2 Historical Background and Principal Applications of Gibbs’ Theory
the general propositions of Clausius to isothermal cycles, he gave a [161] complete

and elegant solution to the problem of the vapour pressure of the same substance4 in
two different states—a problem about which Regnault and J. Thomson had, in the
one case, only an inkling, and in the other, merely hinted at a solution.
But these applications remained limited to those phenomena of chemical
mechanics most directly analogous to the phenomena of fusion and vaporisation.
Physicists limited themselves with difficulty to the consideration of closed cycles,
which had on several occasions furnished demonstrations which, though full of
elegance, required the theorist’s ingenuity to invent an appropriate cycle for each
new investigation. They felt the need to replace this approach by analytic methods
leading by a sure and uniform road to the equations on which the solution of each
problem in thermodynamics depends.
One method, it is true, stands out immediately from Clausius’ propositions. For
any elementary reversible transformation, the excess of heat equivalent to the work
of the external forces over the heat released is the total differential of the internal
energy. The ratio of the amount of heat and the absolute temperature is, except
for sign, the total differential of the entropy. Consequently, expressing these two
combinations by means of the independent variables and writing the conditions of
integration of the expressions thus obtained is a reliable method of obtaining the
conditions of equilibrium of a system. Since 1854, in the memoir where he extended
the second law of thermodynamics to all closed cycles, Clausius5 has made use of
this method. In 1858, G. Kirchhoff 6 showed how fruitful this was by magnificent
applications to the study of changes of states—first-rate applications which it is
fitting to place alongside the discovery of the relation obtaining between the heat of
[162] dilution of a solution and vapour pressure that this solution sustains. Finally,
in 1863, Clausius7 wrote a memoir especially concerned with showing how all the
equations of thermodynamics could be established by this unique method, defying
the attempts of those who sought, 20 years later, to lay claim to this method and put
their name on it.
The analytic precision of this method is beyond question. But, despite the advan-
tages of using it in certain cases, it doesn’t entirely satisfy the need physicists
have felt since the creation of thermodynamics for something like the methods that
mechanics offers where there is an analogy between the objects studied by the two
sciences.
Two centuries of exertions, crowned by the work of Lagrange, have condensed
statics in its entirety into a single principle of virtual velocities, complemented,
for systems sustaining a potential, by the criterion of stability that Lagrange stated
4 corps.
5 R. Clausius, Sur un autre forme du second principe de la théorie mécanique de la chaleur (Théorie
mécanique de la chaleur, trans. Folie, t. I, p. 154).
6 G. Kirchoff, Ueber einen Satz der mechanischen Wärmetheorie und einige Anwendungen
desselben (Poggendorff’s Annalen der Physik und Chemie, Bd. CIII; 1858).
7 R. Clausius, Sur diverses formes des équations fondamentales de la théorie mécanique de la
chaleur, qui sont commodes dans l’application (Théorie mécanique de la chaleur, trans. Folie,
Memoir IX).
2.1 Historical Background of Gibbs’ Theory 25
and Lejeune-Dirichlet demonstrated. Which proposition in thermodynamics should

replace the principle of virtual velocities and the theorem about the stability of equi-
librium? What function should play the role of the potential? This is the statement
of the question that physicists had been led to put.
The first to have clearly posed the problem and tried to provide a solution is
Berthelot. The rule he proposed, under the name of the principle of maximum work,
is formulated as follows: All chemical reactions accompanied without the interven-
tion of external energy tend towards the production of the system which releases
most heat. This principle entails the following consequence: of two conceivable
reactions, the one the inverse of the other, such that the one releases heat while the
other absorbs heat, only the first is possible.
The heat released by a reaction which brings about no external work is the reduc-
tion undergone, as a result of this [163] change, of the internal energy. Consequently,
in accordance with the rule proposed by Berthelot, the possibility of a reaction
presupposes that this reaction results in a reduction in energy. The stability of a
chemical equilibrium is thereby assured if this equilibrium corresponds to the least
value which the energy of the system can take. Energy plays the role in chemical
statics that the potential plays in statics proper.
The rule put forward by Berthelot is easy to apply: its consequences can be imme-
diately subjected to experimental control and, in an immense number of cases, they
accord most satisfactorily with the rule. But however useful the principle of maxi-
mum work is in the study of chemical reactions, it is subject to numerous exceptions,
both among changes of physical state, and among the phenomena of solution or
allotropic change which unite physics and chemistry. Most chemists too have been
led to wish for a rule based on thermodynamics which would at the same time
account for the numerous agreements which make the principle of maximum work
so useful as well as the exceptions which prevent it being accorded an absolute
generality.
Hortsmann8 was the first to have the idea of turning to Clausius’ views on
Carnot’s principle for the solution to the problem of chemical mechanics. According
to Hortsmann, the role that thermochemistry attributed to internal energy fell upon
entropy.
“Clausius”, says Hortsmann, “by giving a mathematical form to certain ideas of
W. Thomson, has defined a quantity, the entropy, which, in all changes in nature,
always increases and which, on the contrary, cannot by any known natural force,
be reduced. Phenomena in which the entropy remains constant are possible, such as
stationary movements that we attribute to atoms of bodies of constant temperature.
According to me, in the phenomena of dissociation, the cause of the limit is
the same; it results when the entropy has [164] become as large as possible, with
all the changes that could occur. The problem is therefore resolved if it is known
under what circumstances and in what way the entropy is changed in the phenomena
concerned”.
Such is the point of departure for Hortsmann. More recently, in a Note com-
municated to the Royal Institution on 5th March, 1875, Lord Rayleigh presented
8 Hortsmann, Annalen der Chemie und Pharmacie, t. CLXX, 20 Nov. 1873.

similar thoughts. According to Lord Rayleigh, it is not the sign of the amount of heat
brought into play which determines the direction of chemical reactions; a chemical
reaction is possible only if it corresponds to an increase in entropy.
This theory cannot be accepted without restriction. Clausius has demonstrated, it
is true, that it is not possible to decrease the entropy of a system, but this demon-
stration is subject to a restriction. It applies only to a system which cannot borrow
from the environment, or give up to it, heat, kinetic energy or work. If account is
not taken of this restriction, it is possible to deduce erroneous consequences from
the theorem. It is quite certain, for example, that the entropy of a mass of water
decreases when it is vaporised at constant temperature. Nevertheless, the princi-
ple stated by Hortsmann can, when applied in the circumstances under which it
can be demonstrated, lead to consequences which are in accord with experiment.
It was in this way that Hortsmann, by studying the dissociation of ammonium car-
bamate, obtained the statement of an important law which applies to all analogous
dissociation phenomena.9
Energy cannot, therefore, play the role of potential in chemical mechanics.
Entropy cannot play the role of a function of forces. Physicists were led to look
for a function which can serve to characterise equilibria amongst other quantities
studied in thermodynamics.
In an important memoir, extracts of which appeared in [165] 1869,10 but which
were unfortunately not published in their entirety until 1876,11 Massieu published
the following result:
All the coefficients that determine the mechanical and physical properties of a
body, the internal energy, entropy, specific heat, coefficients of expansion, com-
pressibility, etc., are known when a certain function of the parameters defining the
state of the body are known. Massieu called this function the characteristic function.
The characteristic function changes according as one system or another of inde-
pendent variables determining the body is chosen. Massieu considered in particular
two cases: that in which the proper variables determining the state of the body are
taken to be the volume and the temperature, and that in which the pressure and the
temperature are taken as variables.
In the first case, if T designates the absolute temperature, S the entropy of
the body and U the internal energy, the body has as its characteristic function
the quantity12
9 The theory of Hortsmann now rejected has just been revived without essential modification, at
least in its principles, by Max Planck in a memoir entitled Ueber das Princip der Vermehrung der
Entropie (Wiedemann’s Annalen der Physik und Chemie, vol. XXX, p. 562; 1887).
10 F. Massieu, Sur les fonctions caractéristiques (Comptes rendus des séances de l’Académie des
Sciences, t. LXIX, pp. 858 and 1057; 1869).
11 F. Massieu, Mémoire sur les fonctions caractéristiques des divers fluides et sur la théorie des
vapeurs (Mémoires des Savants étrangers, t. XXII, year 1876; Journal de Physique, 1st. series,
vol. VI, p. 216; 1877).
12 [Modern notation uses A rather than Duhem’s H.]
H = T S − U.
In the second case, retaining the preceding notation, and in addition designating
the calorific equivalent of work by A, the volume of the body by v and the pressure
it supports by p, the body has as its characteristic function the quantity13
H = T S − U − A p v.
Massieu established several properties of these functions H and H , and showed

how all the equations of thermodynamics could be written in terms of just these
functions and their partial derivatives. With this, he singularly advanced and per-
fected the application of purely analytic methods to thermodynamics. He brought to
light new functions, more important than the internal energy and the entropy, [166]
since the expression of the latter two can be deduced from knowledge of the former.
These are the new functions whose use, thanks to J. Willard Gibbs, was to give
to thermodynamics a new and fruitful impetus.
The ideas which guided Gibbs are quite like those which Hortsmann wanted
for the laws of chemical equilibria. But more happily than his predecessor, Gibbs
avoided the errors into which he had fallen and finally succeeded in obtaining the
fundamental theory.
“The energy of the universe is constant”.
“The entropy of the universe tends towards a maximum”.
These are Clausius’ two propositions which Gibbs, like Hortsmann, takes as the
point of departure for his investigations and as the most important epigraph of his
memoirs. These are the propositions which lead him to the equilibrium conditions
of a system—conditions which he states, without demonstration, in two equivalent
forms as follows14 :
For the equilibrium of an isolated system it is necessary and sufficient that in all possible
variations of the state of the system which do not alter its energy, the variation of its entropy
shall either vanish or be negative.
For the equilibrium of an isolated system it is necessary and sufficient that in all possible
variations of the state of the system which do not alter its entropy, the variation of its energy
shall either vanish or be positive.
The first of these two conditions is nothing other than the condition of equi-
librium given by Hortsmann, but modified with the precaution restricting its
application to changes in which the energy doesn’t vary. Nevertheless, on the subject
of the exactitude of these statements, a reservation should be made. The conditions
of equilibrium stated by Gibbs are sufficient; they [167] are not necessary. From
what, thanks to Clausius, we know about all possible changes, we can deduce that
13 [Modern notation uses G rather than Duhem’s H .]

14 On the equilibrium, etc. (Transactions of the Connecticut Academy, vol. III, 1st part, p. 109).
it is not possible to produce change corresponding to a negative sum of the uncom-

pensated transformations. But from the fact that a change produces a positive sum
of the uncompensated transformations, we cannot conclude that it will necessarily
take place. Gibbs15 has already indicated several considerations which relate to this
principle. But it is Moutier16 we owe for having precisely stated this and in several
cases shown the necessity.
I would add that the equilibrium conditions suggested by Gibbs necessitate
another correction. They are only exact for changes which entail neither external
work nor variation in temperature. Given this restriction, it is easy to convince
oneself of their exactitude by referring back to our Equation (16) in Chapter 1.
Assuming an infinitesimal isothermal change and designating the uncompensated
work by d and the external work by dW, we have
d = E dU − E T dS + dW,
and this quantity should be positive. Restricting attention to changes in which

dW = 0, clearly dU will be positive for all of these changes for which dS equals
0 and dS will be negative for all those changes for which dU equals 0. But these
conclusions no longer hold when the change under consideration involves a positive
or negative external work.
The restrictions just indicated should also be applied to the following theorem,
stated by Gibbs as a consequence of the conditions of equilibrium17 :
If the temperature of a system is maintained constant, it is necessary and sufficient for the
system to be at equilibrium that the [168] function
F = E(U − T S),
is a minimum.18
But only the first of the two restrictions that we have indicated applies to another
theorem of Gibbs.19
When the temperature and pressure of a system are maintained uniform and constant, it is
necessary and sufficient for equilibrium of the system that the quantity
= E(U − T S) + pv,
in which p designates the pressure and v the volume of the system, is a minimum.
15 Gibbs, loc. cit., p. 111.

16 J.Moutier, Sur quelques relations entre la Physique et la Chimie (Encyclopedie chimique de
Fremy, Introduction, vol. II).
18 [As Donald Miller points out, this requires the volume to be constant.]
These two propositions of Gibbs, with the restrictions that we have introduced,
are just the statement, in two particular cases—that where the external forces do
no work, and that where the external forces amount to a uniform and constant nor-
mal pressure—of the general theorem that we have stated in Section 1.6 in the first
chapter. It is therefore quite appropriate to name this Gibbs’ theorem.
On 2 February, 1882, Hermann von Helmholtz communicated to the Berlin
Academy of Sciences a memoir On the thermodynamics of chemical phenomena,20
soon followed by two other memoirs devoted to applications of the theory developed
in the first.
This theory contains nothing not already found in the works of Massieu and
Gibbs. It suffices to quote, in this regard, the acknowledgement Helmholtz gives of
his predecessors at the beginning of his third memoir.
The employment of partial derivatives of a single integral function to represent the energy
and the entropy of a system of bodies, introduced [169] a great simplification into thermo-
dynamics. This simplification was discovered before me, as early as the year 1877,21 by
Massieu. He has fully developed their consequences, at least in the case where the state of
the body depends on a single parameter apart from temperature, but he has not deduced
any conclusions relating to chemical phenomena. He calls the integral function that he con-
siders the characteristic function of the body. This function, which he designates by the
letter H, corresponds to my function (−F). I propose to retain the name free energy which I
chose for the function F. It seems to me that this name captures very well the true physical
significance of this quantity.
The form in which Massieu presented the theorems concerned is a little more general
than that on which I have settled. It is at the same time more convenient for the execution
of certain calculations . . . .
In 1878,22 J.-W. Gibbs developed analytically and considerably extended and gener-
alised the thermodynamic relations concerning changes of physical state or chemical state
of a material system which can be composed of as many substances as you might wish
either contiguous or mixed. Gibbs recovered from this Massieu’s characteristic function
and called it the force function at constant temperature . . . .
Thus, by his own account, Helmholtz has not introduced anything new in the
theories of Massieu and Gibbs except for the name free energy for the quantity
F = E(U − T S)
and binding energy for the quantity

Z = E T S.
20 H. von Helmholtz, Zur Thermodynamik chemischer Vorgänge I Abhandl. (Sitzungsberichte der

Akademie der Wissenschaften zu Berlin, vol. I, p. 23; 1882). II. Abhandl., Untersuchungen uber
[sic] Chlorzink-Kalomel Elemente (ibid., vol. II, p. 825; 1882). III. Abhandl., Folgerungen über
galvanische Polarisation (ibid., vol. I, p. 647; 1883).
21 The works of Massieu date from 1869, as we have pointed out above.
22 The investigations of Gibbs date from 1875 and not 1878.
This leads us to examine the names which have been [170] proposed to designate
the various quantities figuring in this theory.
Massieu gave the name characteristic functions to the quantities
H =TS−U
and
H = T S − U − A p v.
This name admirably describes the role which he discovered for these functions,
because knowledge of them for a definite body yields knowledge of all the prop-
erties of this body and thus characterises the body. This name left more to be
desired as soon as it was discovered what role the functions studied by Massieu
play in the determination of equilibrium, when a name alluding to their mechanical
analogues was wanted. It was therefore with good reason that Gibbs carried over
from mechanics to thermodynamics the names of force function and potential to
designate the various functions figuring in his equations. And it is with a view to
maintaining this usage, introduced by the inventor of the theory we are now occu-
pied with, that we have called the function , defined by Equation (18) in Chapter 1,
the thermodynamic potential of the system.
For the names just mentioned, which have the double advantage of not introduc-
ing a term unfamiliar to physicists, and of nicely recalling the role of the quantities
employed, Helmholtz has suggested substituting the names free energy and bind-
ing energy as we have just mentioned above. These names are open to more than
one criticism; but, applying Gauss’ slogan: Simus faciles in verbis,23 we will not
delay matters by taking up these criticisms but continue to a swift exposition of the
principal applications of Gibbs’ theory.
2.2 Applications of Gibbs’ Theory to Dissociation Phenomena

A certain number of dissociation phenomena show a direct analogy with vapori-
sation phenomena. Everyone knows of Debray’s experiments which brought this
analogy to the fore by the study of the [171] dissociation of the carbonate of chalk.
Thermodynamics didn’t delay in seizing upon these phenomena and extending to
them considerations analogous to those it has applied quite successfully to the study
of vaporisation. But many dissociation phenomena are distinguished from those by
a circumstance which singularly complicates the study of them. The system whose
equilibrium is to be studied contains not just a single gaseous body, but two or
more which mix with one another. To this category of phenomena belong the dis-
sociation of water vapour, carbonic acid, sulphurous acid, the oxide of carbon and
hydrochloric acid—the study of which led H. Sainte-Claire Deville to the concep-
tion of dissociation phenomena—and the dissociation of sal ammoniac, which has
23 [Let’s not quibble over words.]

2.2 Applications of Gibbs’ Theory to Dissociation Phenomena 31
recently been studied with the same care by Naumann, Hortsmann and Isambert. It
is clearly of great interest to apply the considerations of thermodynamics to such
phenomena.
Further interest in this study was aroused by the following circumstance.
In 1840, Cahours observed that the density of acetic acid vapour decreased in
relation to that of air in a remarkable way when the temperature at which this den-
sity was determined was increased. Similar facts were indicated in the study of a
certain number of gaseous bodies. H. Sainte-Claire Deville and Troost studied with
great care the density variations of hyponitrous acid and sulphur vapour. In the latter
case, they found that the density increased threefold when the temperature fell from
1,000 to 500◦ . Crafts and Meier have more recently shown that the density of iodine
vapour varies twofold when the temperature falls from 1,600 to 800◦ .
Many physicists have thought that such variations should not be simply attributed
to errors which the gases under consideration show in relation to the laws of Mariotte
and Gay-Lussac. They have sought an explanation of a chemical kind. They have
thought that the variations of density were due to the splitting of polymers, a split-
ting which becomes accentuated when the temperature increases. They have in that
way reconciled [172] the variations of density with the apparent variations in the
density of gaseous phosphorus perchloride, because of the gradual decomposition
of this body into phosphorous chloride and chlorine. It is of the first importance
to develop the theory of dissociation of gaseous compounds that split into gaseous
elements, to be sure that this theory can furnish an explanation of the observed
phenomena.
It is this theory which constitutes one of the principal applications of Gibbs’
method.24
Gibbs supposes that the gases under study are brought to the ideal state where
they follow exactly the laws of Mariotte and Gay-Lussac. He can then obtain the
complete expression of their thermodynamic potential and establish the laws of
their decomposition. What is obtained is, it is true, only an approximate theory.
But however approximate are the laws of Mariotte and Gay-Lussac in the study of
the physical properties of gases, they have an incontestable utility. Shouldn’t the
same hold for a theory of dissociation of gaseous compounds subject to the same
approximations?
We do not wish to detail Gibbs’ theory here. Suffice it to say that its author
has compared the numerical results of experimental observations of phosphorous
perchloride25 and that this comparison has shown that the theory agrees with exper-
iment as well as could be hoped. An analogous comparison has shown that the
hypothesis of the dissociation of a polymer satisfactorily explained the variations of
density of the vapour of nitrogen peroxide26 and acetic acid.
24 Gibbs, On the equilibrium of heterogeneous substances (Transactions of the Connecticut

Academy of Arts and Sciences, vol. III, p. 210).
25 Gibbs, On the density of vapours, etc. (American Journal of Sciences and Arts, vol. XVIII; 1879).
26 l’acide hypo-azotique.
2.3 Application of Gibbs’ Theory to the Voltaic Cell

Gibbs has not restricted the application of the fruitful theory which he has developed
to the study of dissociation. He has also applied it to the study of capillarity and
shown that [173] thermodynamics can give a new significance to the theory under
which Gauss had subsumed this branch of physics. But is it above all in the study of
the electromotive force of a voltaic cell that Gibbs has been led, by his method, to
new and fruitful results. Let us give an idea of the problem which had to be solved
and the solution which Gibbs has put forward.
Edmond Becquerel had the first idea of relating the value of the electromotive
force of a voltaic cell to the magnitude of the thermal phenomena accompanying
the reaction taking place in the cell. Here is the law that he has thus been led to
state:
When an electric charge, equal to unity, traverses a cell, this cell is the location of a certain
chemical reaction. If this chemical reaction produces no current, it releases a quantity of
heat Q. If the electromotive force of the cell is designated by E, and E is the mechanical
equivalent of heat, then we have
E Q = E.
Unfortunately, it was not long before this simple law was contradicted by exper-
iment. Favre showed first, in 1858, that there frequently existes a quite considerable
positive or negative difference between the chemical heat Q and what he has since
called the voltaic heat, E/E. Raoult (1864–1865), Edlund (1869–1883) and F. Braun
(1878) have shown in a considerable number of experiments that the vast majority
of voltaic couples deviate from the rule proposed by Edmond Becquerel, and that
those that do follow this law are exceptions.
This discrepancy between experiment and Becquerel’s theory seemed inexpli-
cable. For as several, amongst others, Favre, have indicated, they believed that
Becquerel’s formula was a rigorous consequence of the principles of thermody-
namics. Braun had the idea that the reason for this discrepancy should be sought
in Carnot’s theorem, which would lead to the substitution of Becquerel’s rule by a
more exact rule. But the [174] rule that he proposed lacked precision and several
of his conclusions are now unacceptable. The true rule, which should be substi-
tuted for Edmond Becquerel’s rule, was given by Gibbs, at the very moment that
Braun published his ideas, in the second part of On the equilibrium of heterogeneous
substances.27 This is how this fundamental rule might be formulated:
When a unit electric charge traverses a cell, a certain chemical reaction takes place in
the cell. If this chemical reaction didn’t produce any current, it would produce a variation
(2 − 1 ) of the thermodynamic potential of the system such that
E = 1 − 2 .
27 On the equilibrium of heterogeneous substances (Transactions of the Connecticut Academy of

Arts and Sciences, vol. III, part II, p. 503).
2.3 Application of Gibbs’ Theory to the Voltaic Cell 33
This fundamental postulate became the point of departure for an entirely new
theory of the galvanic cell—a theory which experiment had already given important
confirmation. As we have said, the postulate is due to Gibbs. However, the theory
at issue usually goes under the name of Helmholtz’ theory. A work that we have
published on this question, at a time when, unfortunately, we had only an incom-
plete knowledge of Gibbs’ works, might, perhaps, have had some influence in the
introduction of this misleading name. It therefore obliges us to rectify this injustice
here and proclaim Gibbs’ incontestable priority.
In a hydro-electric couple, the electromotive force depends on the greater or
lesser concentration of the liquids in which the electrodes are immersed. In order
to make precise the influence that the concentration of liquids exerts on the elec-
tromotive force, James Moser28 undertook the study of cells in which the reaction
produced at one [175] pole is reversed at the other pole. The electromotive force
then depends solely on the concentration of the liquids. Two vessels, put in com-
munication by a siphon, contained unequal concentrations of a single metallic
salt. Two electrodes, formed from the metal of this salt, were immersed in these
vessels. This is the kind of cell in which Moser measured the electromotive
force.
At the same time as Moser published the results of his experimental investiga-
tions performed in Helmholtz’ laboratory, Helmholtz29 applied the propositions of
thermodynamics to the phenomena studied by Moser.
The solutions, enclosed in the two vessels each containing cells studied by Moser,
had different vapour pressures. Helmholtz showed that, from the variations under-
gone by the vapour pressure of these solutions when their concentrations varied, it
was possible to deduce the value of their electromotive force. Comparison of the
values thus calculated with the value determined experimentally was satisfactorily
in agreement.
In this theoretical study, it is true that Helmholtz made use of Carnot’s theorem;
but for all that, he had not renounced the law of Edmond Becquerel, as can be
concluded from the following passage30 : “. . . By virtue of the law which regulates
the other phenomena of electrolysis, the work produced in this case by the chemical
forces should act as the electromotive force. This work can be calculated by means
of the mechanical theory of heat . . .”. Thus, Gibbs became the true founder of the
new theory of the cell when he formulated the fundamental postulate which we have
just stated precisely in the same year (1878) that Helmholtz’ memoir appeared in
the Annalen der Physik und Chemie. The beautiful consequences that, several years
later, [176] Helmholtz deduced from this theory should not allow us to forget the
name of the man who created it and who proclaimed its rigour and generality, thus
28 J.Moser, Galvanische Ströme zwischen verschieden concentrirten Lösungen desselben Körpers

und Spannungsreihen (Monatsber. der Berl. Akad., 8 Nov. 1877: Wiedemann’s Annalen der Physik
und Chemie, vol. III, p. 216; 1878).
29 H. Helmholtz, Ueber galvanische Ströme verursacht durch Concentrationsunterschiede.
Folgerungen aus der mechanischen Wärmetheorie (Monatsber. der Berl. Akad., 26 Nov. 1877,
Wiedemann’s Annalen der Physik und Chemie, vol. III, p. 201; 1878).
30 Wiedemann’s Annalen der Physik und Chemie, vol. III, p. 203.
drawing a consequence of the first order from the new method which he introduced
into thermodynamics.
This is the point to which the method and its principal applications has reached
after the work of Gibbs and Helmholtz. The path had been traced out and was known
to be reliable. It remained to show that the principle of the new method could be
simply deduced from the postulates of thermodynamics, to develop the theory of
dissociation to which it led, to relate the proposition on which the theory of the cell
is based to the elementary laws of electricity, and finally, to extend the consequences
of the new doctrine into all the branches of physics. What has been the fruit of the
efforts in these various directions since the works of Gibbs and Helmholtz? It is not
for us to say.
Part II
Commentary on the Principles
of Thermodynamics: The Principle
of the Conservation of Energy (1892)∗
Introduction
Science always advances by a series of oscillations.
During some periods, the principles of science are discussed. An examination of
the hypotheses which they presuppose and the restrictions by which they are con-
strained is undertaken. Then, for a time, when the principles seem well-established,
the efforts of the theorists are directed towards the deduction of consequences;
applications multiply and experimental verifications become numerous and precise.
But this development, at first rapid and easy, eventually becomes slower and
more arduous. The over-cultivated ground loses its fertility. Obstacles emerge which
the established principles cannot surmount, contradictions which they do not suc-
ceed in resolving and problems with which they cannot grapple. At such a time, it
becomes necessary to return to the foundations on which the science is based, to
examine anew their degree of soundness, to assess exactly what they can support
without giving way. Once this work is done, it will be possible to build up the new
consequences of the theory.
The applications of thermodynamics during the last thirty years have been numer-
ous. And by common agreement of all those who are interested in this science, a
revision of its principles has become necessary. It is the attempt at such a revision
that we now submit to the readers of the Journal de Mathémathiques.
All [270] physical theory rests on a certain number of definitions and hypotheses
which are, to a certain extent, arbitrary. It is therefore possible to seek to expound
such a theory in a logical order; but to claim to have given it the only logical order
to which it is susceptible would be an unjustifiable claim. We will be careful to
make no such claim. We are convinced that the principles of thermodynamics can be
connected together in a different, but equally—perhaps more—satisfactory, manner
from that we have adopted. We dare not even hope that no lacuna exists in the
organisation that we have sought to establish.
∗ “Commentaire aux principes de la Thermodynamique. Première Partie: Le principe de la
conservation de l’énergie”, Journal de Mathématiques Pure et Appliquées, 8 (1892), 269–330.

36 Part II Commentary on the Principles of Thermodynamics: The Principle . . .
If the question that we have examined appears more philosophical than mathe-
matical, then we may be allowed to invoke, by way of justifying its introduction
in this Journal, the interest manifested quite recently by an illustrious analyst in
research into the principles of thermodynamics. This will be our excuse to the
mathematicians.
Chapter 3
Preliminary Definitions
3.1 On Absolute Motion

We will take for granted geometry and kinematics. We will borrow from these
sciences any results that we need.
Experience allows us to conclude whether one of two portions of matter is dis-
placed in relation to the other, so that the notion of relative motion is an experimental
concept. This is the concept that kinematics deals with.
But [271] this concept is not sufficient for the object that we propose to treat. The
hypotheses which we will state, and the laws which we will formulate, will involve
not only the relative movement of various quantities of matter in relation to one
another. They will involve the movements of various quantities of matter in relation
to a certain ideal triad of coordinate axes which we suppose traced somewhere.
It will frequently happen that propositions concerning movements relative to this
particular coordinate system, and which we regard as exact, become manifestly false
if the movements are supposed to be relative to some other coordinate system which
is in motion relative to the first.
We will call this particular coordinate system, to which all the motions we will
speak of are referred, the absolutely fixed coordinate system. The axes of this coor-
dinate system will be absolutely fixed axes. A movement relative to this particular
coordinate system is called an absolute motion. A portion of matter whose various
points are not kept in motion by any movement relative to this coordinate system
will be said to be at absolute rest. In particular, a coordinate system which is immo-
bile in relation to the absolutely fixed coordinate system will define a new absolutely
fixed coordinate system.
We are not able to judge whether a given coordinate system is indisputably fixed
or not. Any judgement of this kind is subordinated to the belief in the legitimacy of
some hypothesis. If we regard a certain hypothesis involving consideration of abso-
lute movements as exact, and if this hypothesis, applied to the relative movements
of a certain coordinate system, leads to inexact results, we say that this coordinate
system is not absolutely fixed. But this conclusion is only obligatory in so far as we
are bent on upholding the hypothesis which has served us as a criterion. We will
be right to regard as fixed the coordinate system at issue if we agree to reject the
hypothesis.

38 3 Preliminary Definitions
3.2 On Bodies and Mixtures or Combinations

We will understand a body to be a linearly connected space filled, in a continuous
manner, by a certain quantity of matter.
We will not discuss the question whether bodies are really [272] continuous or
formed from very small, discontinuous quantities separated by equally small empty
intervals.
In physics, it is both impossible and useless for us to know the real constitution
of matter. We seek simply to conceive an abstract system which furnishes us with
an image of the properties of bodies. In order to construct this system, we are free
to represent a body which seems to us to be continuous either as a continuous dis-
tribution of matter in a certain space, or as a discontinuous ensemble of very small
atoms. Since the first mode of representation is conducive, in all the branches of
physics, to more simple, clear and elegant theories, we adopt it in preference to the
second.
Consider two bodies A, B which, at a certain instant t, occupy spaces a, b having
no part in common. These two bodies are not always and necessarily distinct; the
quantities of matter which constitute them might, at an instant t , distinct from t,
and earlier or later than t, make up a single body C, occupying a space c, in such a
way that any element dw of the space c encloses, at instant t , a quantity of matter
which, at instant t, constituted body A, and also a quantity of matter which, at instant
t, constituted body B, the first quantity occupying, at instant t, a certain volume
element dv of the space a and the second quantity occupying, at instant t, a certain
volume element dv of the space b.
In the case just mentioned, the body C results either from the mixture or the
combination of two bodies A and B.
Many physicists do not accept the possibility of a combination or mixture of the
kind we have just defined. They regard as impossible this intimate penetration by
which the matter filling each volume element of the continuous body C results from
the union between the matter enclosed by a volume element of the continuous body
A and the matter enclosed by a volume element of the continuous body B. It is this
impossibility that they call the impenetrability of matter.
For these physicists, the words mixture and combination do not represent appear-
ances. When we believe we see the two bodies [273] unite to form a new body C, the
extremely small parts whose discontinuous collection constitutes each of these two
bodies remain, in reality, distinct. The small parts of body A are simply interposed
between the small parts of body B, without the space occupied by any one of these
parts of body A having anything in common with the space occupied by any of the
parts of body B.
Reasons analogous to those which led us to regard the matter which constitutes a
body as continuous lead us to resist this way of conceiving mixture or combination
and to adopt the definition just given.
Let us consider a body C formed by the mixing of two bodies A and B. The matter
which, at the instant t, fills the volume element dw of the body C is composed of
a part p of the matter which forms body A and a part q of the matter which forms
3.4 On the Variables Which Define the State and Movement of a System 39
body B. At a different time t , these two parts p and q are not necessarily united with
one another in the same volume element; the matter constituting the part p might fill
a volume element dw where it is either free or unified with a part q , different from
q, of the matter of body B. At the same time, the matter which constitutes the part q
might fill another volume element dw where it is either free or unified with a part
p , different from p, of the matter of body A.
Thus, when a body C is a mixture of two bodies, the matter filling each volume
element of this body is formed by the union of two different quantities, and these
two quantities may be kept in motion by different movements. In that case, at each
point of the mixture there might be occasion to consider two different velocities,
each of these velocities being relative to one of the parts of the mixture.
All that we have said about a mixture of two bodies applies equally well to a
mixture of any number of bodies.
3.3 On Bodies Isolated in Space

Experience shows us that we can remove all the bodies which surround a given body
at a given instant. The existence of such bodies therefore seems to us to have no nec-
essary connection with the existence of the given body. We thus come to conceive
of the possibility of [274] the existence of this body isolated in space unbounded in
every sense and absolutely empty.
This conception of bodies isolated in unbounded and absolutely empty space is
a pure abstraction. Experience never offers us a body without any parts contiguous
with other bodies, and physics leads us to accept that even if we were to succeed in
removing all the solid, liquid and gaseous bodies that we can directly or indirectly
lay hold on, so as to make a physical vacuum in the space surrounding a given body,
this space would still be filled with a certain matter that has been called ether. It is
therefore, I repeat, in virtue of a pure abstraction that we can conceive of a body
as existing alone in space. But I don’t believe that it is possible to build up physics
without making use of this abstraction.
3.4 On the Variables Which Define the State

and Movement of a System
Let us consider a collection of bodies isolated in space. This collection of bodies

may, from one instant to another, change position, shape, state, etc. Let us consider
it as it is at the instant t, disregarding what it was at all instants earlier than t and what
it will be at all instants later than t. At this instant t, it possesses certain properties.
In order to represent these properties, theoretical physics defines certain algebraic
and geometric magnitudes and then establishes relations between these magnitudes
which symbolise physical laws to which the system is subjected.
The magnitudes may be defined in very different ways. Geometry, for example,
tells us in what way it is possible, by the definition of certain magnitudes, be they
limited or unlimited in number, to determine the shape and position of each of the
material particles composing the system. Other magnitudes which represent physi-
cal and chemical properties are defined within various theories of which physics is
composed. In this chapter, we will show how physics defines one of these magni-
tudes which is particularly important, temperature. In the following chapter, we will
deal with another, mass.
In the study of a system, it might be of interest to consider [275] at the same time
several magnitudes whose definitions are interrelated. Thus, it is necessary to speak
of the mass1 of a body, of its density and its volume, even though by definition, the
density of a body is the ratio of its mass and its volume. Magnitudes connected to
one another by their definition are not independent.
Magnitudes representing the properties of a system at a given instant are inde-
pendent if the definition of each implies no relation between its value at the instant
t and the value of each of the others at the same instant t. Thus, the volume and
mass of a body will be two independent magnitudes. Similarly, the components of
the electric current at each point of a conductor and the electric density at each point
of the same conductor are independent magnitudes. Without contradicting either the
definition of the current density or the definition of electric current, it is possible, at
the instant t, to attribute arbitrary values to the former and to the three components
of the latter.
We have said that several magnitudes are independent if the definition of each
of them implies no relation between its value at the instant t and the value of each
of the others at the same instant t. But while the values of independent variables
may all be chosen arbitrarily at an isolated instant t, in certain cases it would not
be permissible to arbitrarily choose their values at all the instants during a certain
interval of time. Thus, for the instant t, considered in isolation, it is possible to
choose arbitrarily the electric density and the components of the electric current at
each point of a conductor. But it is not possible, without absurdity, to do this at all
the instants of the interval of time (t1 – t0 ). In fact, the very definition of electric
current shows that if, at each point of the conductor, (i) the components of electric
current at every instant of the interval of time (t1 – t0 ) are known, as well as (ii) the
electric density at a particular instant of the interval of time (t1 – t0 ), then the value
of [276] the electric density is determined at every instant of the interval (t1 – t0 ), so
that this value cannot be chosen arbitrarily.
In all that we have just said, it should be noted that when we speak of the
dependence between various magnitudes, we intended to speak only of a depen-
dence resulting from the definition of these magnitudes and not of a dependence
resulting from a physical law. So logically independent magnitudes need not be
physically independent. Giving them arbitrary values is an operation which, without
being absurd, may be contrary to the laws of nature.
1 Inorder to give examples we are, of course, obliged to anticipate, since no other magnitude
representing a physical property has been defined in the foregoing.
3.4 On the Variables Which Define the State and Movement of a System 41
Among the magnitudes, independent or not, which serve to represent a system at

an isolated instant t, the definition of some of them constrains them to have the same
value for a given system whatever the isolated instant considered. This is the case,
for example, with the mass of the system, or again with the total electric charge.
There are others which can, for the same system, have different values at different
instants. The former are said to define the nature of the system and the latter to
define the state.
Let us consider the independent magnitudes that suffice to represent completely
the properties of a system at an isolated instant t. Some, A, B, . . ., L, define the
nature of the system; others, α, β, . . . , λ, define its state.
If the quantities A, B, . . ., L conserve their values and if the variables α, β, . . . , λ
are given different values α , β , . . . , λ , this would represent a different state of the
same system.
If the quantities A, B, . . ., L conserve their values and if the variables α, β, . . . , λ
are given different values α , β , . . . , λ , this would represent a different state of the
same system.
Let us therefore imagine a continuous series of different states of the system, that
is to say, a continuous series of groups of values of the quantities α, β, . . . , λ. Let us
successively fix our attention on these various states, in an order which allows con-
tinuous passage from one state to another. By designating this a purely intellectual
operation, we are saying that we impose a virtual change on the system.
All changes realisable by a system correspond to variations of the quantities
α, β, . . . , λ compatible with the definitions of these quantities. The sequence of
states through which the system passes therefore constitutes a virtual change of the
system.
Conversely, [277] can a virtual change always be regarded as the sequence of
states that a system traverses during a real change? Remembering that the variables
α, β, . . . , λ which can, by their definitions, take arbitrary values, may be connected
to one another by physical laws, it can be seen that a virtual change may be com-
patible with the definitions of the variables appropriate for representing the various
states of the system but conflict with certain physical laws, and, consequently, not
be physically realisable.
Moreover, the values it is possible to attribute to the variables α, β, . . . , λ at an
isolated instant t are arbitrary. But such is not always the case with the values that
can be attributed to these variables at various instants of a certain temporal inter-
val. Thus, regarding the sequence of groups of values of α, β, . . . , λ, representing
the various states of the system during a virtual change as a sequence of states
which succeed one another during a certain interval of time soon leads to a con-
tradiction with the definition of the variables α, β, . . . , λ. If, for example, in the
virtual change considered, it is supposed that non-uniform currents run through
a conductor, and if, on the other hand, the charge density at each point is taken
to be invariable, a continuous sequence of states is obtained whose succession in
time would be in contradiction with the definition of electric current. Thus, the def-
initions of the variables α, β, . . . , λ can suffice to render certain virtual changes
unrealisable.
We have said that the constants A, B, . . ., L and the variables α, β, . . . , λ represent

the physical properties of the system considered at the instant t, disregarding the
properties of the system at instants preceding and following t. Now considering a
system at an isolated instant, disregarding the properties of the system at instants
preceding and following it, does not allow for a distinction between the body at rest
and in motion. The word motion only has meaning for the system in so far as it can
be envisaged during a certain lapse of time, however short. Consequently, the values
of the quantities α, β, . . . , λ, necessary and sufficient for representing the properties
of the system at the isolated instant t, do not in general suffice to tell us whether the
system is in motion and what the motion is.
We [278] say that the motion of the system at the instant t is defined if not only
the state of the system at this instant is known, but also the magnitude and direction
of the speed with which the material filling each of the elements of volume of the
system moves. In the case where a volume element is filled with a mixture of several
substances, it is necessary to know the velocity with which each of the parts of the
material which compose the mixture moves.
Consider an infinitesimal part of the material which forms a system. The coor-
dinates x, y, z, which indicate the position of one of its points at the instant t in
relation to an absolutely fixed co-ordinate system, are known when the values of the
variables α, β, . . . , λ are known at this instant. In fact, if these variables are known,
then the shape, position and properties possessed by all the parts of the system at
the instant t should be known. In order to determine x, y, z, we should therefore have
equations of the form
⎧
⎪
⎪ x = ϕ(α, β, . . . , λ),
⎨
y = ψ(α, β, . . . , λ), (1)
⎪
⎪
⎩
z = χ (α, β, . . . , λ).
ϕ, ψ and χ are three functions whose form depends on the nature of the system and
also on the particular material considered. These relations do not depend explicitly
on the time t, because if, at two different instants t and t , the variables α, β, . . . , λ
again take on the same values, as variables sufficient to determine the state of the
system, the system will return identically to itself, and the coordinates x, y, z will
take on the same values.
From Equations (1), the equations
⎧
⎪ dx ∂ϕ dα ∂ϕ dβ ∂ϕ dλ
⎪
⎪ = + + ... + ,
⎪
⎪ dt ∂α dt ∂β dt ∂λ dt
⎪
⎪
⎨ dy ∂ψ dα ∂ψ dβ ∂ψ dλ
= + + ... + , (2)
⎪
⎪ dt ∂α dt ∂β dt ∂λ dt
⎪
⎪
⎪
⎪ dz ∂χ dα ∂χ dβ ∂χ dλ
⎪
⎩ = + + ... + ,
dt ∂α dt ∂β dt ∂λ dt
3.5 On Independent Systems 43
can be deduced. The Equations (2) show us that the components of the velocity
[279] of any elementary part of the system are linear and homogeneous functions
dβ
dt , dt , . . . , dt . These functions depend, moreover, in some way on the variables
of dα dλ
α, β, . . . , λ.
Thus, in order to define the motion of a system at the instant t, it is sufficient to
dβ
add to the values of α, β, . . . , λ the values of dαdt , dt , . . . , dt . Is this also necessary?
dλ
In a large number of cases, not all the quantities α, β, . . . , λ figure in the

dβ
Equations (1), and consequently, not all the quantities dα dt , dt , . . . , dt in Equations
dλ
(2). Sometimes we have to distinguish between those of the quantities α, β, . . . , λ

which figure in the Equations (1) from those which do not. We preserve for the
former the letters α, β, . . . , λ and for the latter, we adopt the letters a, b, . . . , l.
We will say that an isolated system is at rest when the material which composes
it is immobile, its state being, moreover, able to undergo the variations with time
which allow each of the parts composing it to be in the same position. For such a
system, the quantities α, β, . . . , λ retain their values independently of time, whereas
the quantities a, b, . . . , l are able to undergo any variation with time.
We will say that an isolated system is at equilibrium if its state does not vary
with time. For such a system at equilibrium, the quantities a, b, . . . , l as well as the
quantities α, β, . . . , λ have their values independently of time.
Imagine that a system departs from a certain state characterised by certain
determinate values of the variables α, β, . . . , λ and by a certain velocity clearly
determined for each infinitesimal material part in which it might be supposed to
be divided. Further, suppose that the system might undergo a series of greater or
lesser changes, and finally, that, after a certain time, it is brought to a state identical
to the initial state, that is to say, to a state where the variables α, β, . . . , λ have the
same values as in the initial state, where the velocity of each elementary part is the
same as that in the initial state. The series of transformations [280] undergone by
the system is called a closed cycle. We will often have occasion to consider such
transformations in the course of this work.
It is understood that when a system undergoes any transformation, the velocity
of each of the elementary parts in which it may be supposed to be divided varies
with time in continuous fashion.
In most of the preceding considerations, we have supposed that the state of the
system is defined by the values of a limited number of variable parameters. It is
easily seen that this hypothesis has merely the aim of simplifying the language, but
that everything that we have said can be extended without difficulty to a system
whose definition requires knowledge of an infinity of variable parameters.
3.5 On Independent Systems

Consider a system of bodies S isolated in space where α, β, . . . , λ are the variables
which, at each isolated instant t, completely determine its state.
Suppose that the bodies forming the system can be divided into two groups,
S1 , S2 .
Suppose further that the variables α, β, . . . , λ can be divided into two groups,
α1 , β1 , . . . , λ1 and α2 , β2 , . . . , λ2 with the following properties:
1. Nothing in the definition of the variables α1 , β1 , . . . , λ1 presupposes the exis-

tence or properties of the group of bodies S2 or the variables α2 , β2 , . . . , λ2 ;
2. Nothing in the definition of the variables α2 , β2 , . . . , λ2 presupposes the exis-
tence or properties of the group of bodies S1 or the variables α1 , β1 , . . . , λ1 .
If these conditions are realised, it may be supposed that the group of bodies S1 is
isolated in space, and that, at each instant of a certain interval of time, the variables
α1 , β1 , . . . , λ1 have, for the isolated group, values identical to those they would have,
at the corresponding instant of an equal interval of time within the system S. This
hypothesis contradicts nothing in the definition of the variables α1 , β1 , . . . , λ1 .
Similarly, it can be supposed that the group of bodies S2 is isolated in space, and
that, at each instant of a certain interval of time, that the variables α2 , β2 , . . . , λ2
have, for the isolated group, values identical [281] to those they would have at
the corresponding instant of an equal interval of time within the system S. This
hypothesis contradicts nothing in the definition of the variables α2 , β2 , . . . , λ2 .
Further, the two hypotheses that we have just presented, while not contradicting
the definition of the groups of variables α1 , β1 , . . . , λ1 and α2 , β2 , . . . , λ2 , may be at
odds with certain experimental laws.
When two groups of bodies, S1 and S2 , satisfy the conditions just enumerated, we
say that these two groups constitute two material systems able to exist independently
of one another, or, more concisely, two independent systems.
Note that in this case, the variables α1 , β1 , . . . , λ1 evidently suffice to fix the
state of the system S1 without saying anything with regard to the state of system
S2 . Conversely, the variables α2 , β2 , . . . , λ2 evidently suffice to fix the state of the
system S1 without saying anything with regard to the state of system S1 . Thus, when
two systems are independent of one another, it is possible to isolate each them from
the other without changing its state. To repeat, in this statement, the word possible
designates an operation which is not in contradiction with the definitions, but is not
necessarily an operation conforming to the laws of physics.
We will illuminate what we have just said with some examples:
(1) The system S is a body whose state we regard as defined when we know the
position, density and temperature of each of the infinitesimal materiel parts which
compose it.
Without contradicting the definitions of temperature and density, we can, after
having divided the body into two parts, S1 and S2 , ignore part S2 and consider part
S1 isolated in space while preserving for each of the infinitesimal material parts
which compose it the position, density and temperature which it would have had,
at the same instant, in the system S. We can also ignore the part S1 and retain the
part S2 isolated in space while preserving for each of the infinitesimal material parts
[282] which compose it the position, density and temperature which it would have
had, at the same instant, in the system S.
3.5 On Independent Systems 45
In this case, which is of great importance, the two systems S1 and S2 are
independent.
(2) The system S is a conductor which carries certain electric charges. The def-
inition of these charges requires that the sum of the charges distributed over the
conductor S remains constant during the period considered. But if this conductor
is regarded as comprising two contiguous parts, S1 and S2 , the sum of the charges
distributed over the portion S1 might well vary from one moment to another, as may
the sum of charges distributed over the part S2 .
Suppose now that, during a certain period of time, we consider the part S2 to be
suppressed and the part S1 to be isolated in space. Can we entertain the idea, at each
instant of this period, of a distribution of electricity over this part S1 identical to that
which it would carry at the same instant, if it were incorporated in the system S?
No, because as just remarked, this distribution will, in general, have a variable value
for the total quantity of electricity carried by S1 from one instant to another. When
isolated in space, the conductor S1 should have an invariable total electric charge in
accordance with the definition of electric charge.
Thus, in the case that we have just analysed, the two bodies S1 , S2 do not form
two independent systems.
(3) The system S is a conductor carrying a certain electric charge through which
certain currents flow. In order to fix our ideas, we will suppose that these currents
are uniform. The charge distribution is then invariable, both on the surface of the
conductor S and in the interior, during the period considered.
We cut the conductor S into two parts, S1 , S2 by a surface σ. Can we ignore part
S2 and regard part S1 as isolated in space while preserving, during the period con-
sidered, a distribution of currents and charges identical to that which would obtain
in this part S1 when [283] it is incorporated in the conductor S? Clearly not, because
in the conductor S, the various points of the surface σ carry an invariable electrifi-
cation whereas if the part S1 were isolated in space, the currents running through it
would make the electrification of the surface σ vary from one instant to another.
In this case too, the two bodies S1 , S2 are not independent systems.
It is as well to note that in saying that two bodies S1 , S2 do not constitute two
independent systems, we do not wish to say that each of these two bodies cannot be
conceived as isolated in space. On the contrary, we allow that a body may always
be thought of as isolated in space (see Example 3). What we mean is that it is not
possible to accept that each of them, isolated in space, preserves, at each instant of
an interval of time, the state which it would have at the same instant as if it were
incorporated into the entire system.
Thus, in the last two examples, the body S1 can be conceived as isolated in space.
But what is contradictory is to attribute to it, during a certain time, the distribution
of electrical charges and currents that it would carry during the same time if were
united with the body S2 to form the system S.
Physics cannot decide, without exceeding the domain or its methods of legiti-
mate application, whether the universe is bounded or not. But in all questions of
physics, it is possible to reason as if the universe were formed of a certain number
of bodies enclosed within a surface of closed extension. In fact, it is allowed that,
when studying a group of bodies, it is possible, without detectable error, to consider

as non-existent all bodies which lie beyond a certain bounded distance.
We consider, then, the very large, though bounded, part of the universe that it is
necessary to take into account when we wish to study a definite group of bodies S1 .
Let S be this part of the universe, and let S2 be what remains of S when S1 has been
removed. If the two groups of bodies S1 , S2 constitute two independent systems, we
will say that the group S1 forms a material system. It is always in this well-defined
sense that we will employ the word material system.
We can extend to such a system the words rest and equilibrium without changing
the [284] definitions that we have given them at the end of Section 3.4 for an isolated
system.
3.6 On Temperature
Among the variables serving to define the state of a system, there is one which plays
a particularly important role in the course of the present work, the temperature.
We will now show how this variable can be defined. This study will, moreover,
have the advantage of showing us how a physicist brings certain magnitudes into
correspondence with physical properties of a system.
Our organs give us the sensation of warm bodies and cold bodies, and of bodies
warmer or colder than other bodies. This sensation of warmth or cold, of greater
or lesser warmth or of more or less intense cold, we regard as the sign of a certain
property possessed by bodies that they possess to a greater or lesser degree. We
accept that that a body is warm, if it is more or less warm than another, that it is cold
if it is less warm than our own body.
Our faculty of abstraction soon attributes to bodies aspects of this property which
we characterise by the words being warm, being cold, being more or less warm that
go unnoticed in sensation.
We cannot compare the degree of warmth of bodies if the bodies are too cold or
too warm. Outside certain limits, for any of the senses, our organs will be injured or
destroyed. Nevertheless, we conceive that outside these limits bodies continue to be
more or less warm than other bodies.
In comparing our sensations with those of our fellow-creatures, we see that
we sometimes find two bodies unequally warm whereas someone else finds them
equally warm, or conversely. We are thus led to accept that the sensibility of our
organs is limited and that, without being identical, the degrees of warmth of two
bodies may be so alike that we are not able to distinguish them.
A warm body can only influence our organs by the part of [285] its surface in
contact with these organs, and this surface always has a certain extension. The time
during which we touch this surface always has a certain duration. Nevertheless, we
accept that the character of being warm holds as well of the parts in the interior of the
body as of those parts at the surface, that it holds of each of the infinitesimal parts
of which the body can be considered to be decomposed and at each infinitesimal
3.6 On Temperature 47
moment of duration, that at such an instant it varies from one point to another, and
that at such a point it varies from one instant to another.
Observations of the most ordinary kind show us that, in most cases, when a warm
body is in the presence of a cold body, the cold body is warmed and the warm body
is cooled. Generalising this observation, we accept as exact the following law:
In order that an isolated2 system be at equilibrium, it is necessary that all the material parts
composing the system are equally warm.
This law again leads us to correct the evidence of our sensations. Experience
teaches us, in fact, that in certain systems that we regard as being at equilibrium,
various parts can appear to us as being far from equally warm. For example, a piece
of steel and a piece of wood, the combination of which is in equilibrium [286],
give us very different sensations of warmth. We continue, however, to regard them
as being in reality equally warm, and we admit that our sensations do not always
inform us exactly about the degree of warmth of bodies.
The words being warm therefore correspond to a property of each of the infinites-
imal parts into which a body can be considered to be divided. What is this property
in itself? Is it reducible, by virtue of its own nature, to quantitative elements? These
are questions which physics cannot resolve. As we conceive it, this property is not
quantitative. It seems to us that it is susceptible of being reproduced identically with
itself, of being augmented and diminished, but not susceptible of addition.
But we can put this non-quantitative property into correspondence with an alge-
braic magnitude which, without standing in any relation to its nature, will be a
representation of it.
We can, in fact, conceive of the existence of a magnitude which satisfies the
following conditions:
1. This magnitude has a determinate value at each point of any body.

2. It has the same value at two points which are equally warm.
3. It has different values at two unequally warm points, the greater value corre-
sponding to the warmer point.
4. If two points tend to become equally warm, the values of the magnitude
considered which corresponds to them tends towards the same limit.
Evidently, if the values taken by such a magnitude at various points of a collection

of bodies are known, then it will always be known exactly whether the degree of
2 It should be noted that this law is only exact in so far as the system to which it applies is isolated.
A metallic bar with one end immersed in steam and the other in melting ice is in equilibrium when
a constant heat flux is established. However, the various points of this bar are not equally warm.
But this bar does not constitute an isolated system. If it were to be incorporated in an isolated
system, that would contain, at the same time, the boiling water and the melting ice which are not
in equilibrium. An analogous observation is applicable to the state of equilibrium achieved by a
thermo-electric chain when constancy is established, both for the heat flux and the electric flux.
warmth varies from one part to another of the collection, and in which sense it
varies.
This magnitude, for which the various values serve, not to measure (that would,
after the foregoing, make no sense), but to represent the various degrees of warmth,
will be called temperature.
The definition of temperature leaves a high degree of arbitrariness in the choice
of this magnitude. Let us imagine that the temperature of all points has, in fact, been
determined in a first way. Let ϑ be this temperature. The magnitude = f (ϑ) could
evidently be, in [287] turn, taken as temperature if the function f (ϑ) possesses the
following three characteristics;
1. For each value that the variable ϑ can take, the function f (ϑ) takes one and only
one value;
2. The function f (ϑ) varies in continuous fashion with continuous variations in ϑ;
3. The function f (ϑ) always varies in the same sense as ϑ.
The operation by which, at a given instant, it would be known how the various
parts of a body would be classed if all the elements of the same class were equally
warm, and if all the elements in any class were warmer than the elements in the
preceding and less warm than the elements in the following class—this operation, I
say, seems logically possible to us, although our senses do not allow us to realise it
except within restricted limits and in a grossly approximate fashion.
We therefore conceive that a temperature, constant for all the elements found in
the same class and increasing from one class to another, can be chosen, although
our sensations do not furnish us with the means of realising such a choice with
any precision. That suffices for this temperature to figure in our arguments without
risk of employing words with no sense. As a matter of fact, it is this tem-
perature exclusively, conceived in abstract fashion, that will be at issue in our
theories.
But if we wish to apply to concrete systems the results to which we are led by the
abstract arguments in which temperature figures, it doesn’t suffice to know that it is
possible to construct a magnitude, called temperature, taking a determinate value at
each point of these systems. It is necessary to have an exact or approximate means
of actually constructing such a magnitude to obtain its numerical value; that is to
say, to class the parts of bodies according to their increasing degree of warmth. We
have seen that our senses, employed directly, are not adequate for this purpose.
The method employed to obtain an experimental determination of the tempera-
ture, or rather of a temperature, applies [288] only to a particular case, albeit a very
extensive one. With various special hypotheses, which we will not examine here, it
can be extended to certain other cases.
This experimental determination depends on the following law, the origin of
which we have already spoken:
For equilibrium in an isolated system it is necessary that all its parts are equally
warm.
3.6 On Temperature 49
If, as we have every right to, we use in our arguments a temperature ϑ, whose
determination is conceived in abstract fashion, and not realisable in an effective
manner, we can state the preceding law in the following form:
If an isolated system is at equilibrium, the temperature ϑ has the same value at all of its
points.
Given this law, suppose that we have a system S, isolated and at equilibrium; then
it has the same temperature at all of its points.
The system S is itself formed of two independent systems T and U.
The system U, except for the property of being independent of T, may be any
whatever.
On the other hand, it is supposed that the system T possesses, at least approxi-
mately, the following features3 :
1. For a given value of the temperature ϑ which is the same at all its points, the
system T can be at equilibrium in just one way, whatever the independent sys-
tem U to which it is attached. The various properties exhibited by this system T
at equilibrium therefore depend uniquely on the temperature. If there is among
them one which is measurable (for example, a geometric property), the number
which measures it is a function of the temperature ϑ alone.
2. [289] Among the measurable properties of the system T, there is at least one
which always increases as the system T is warmed. However the temperature ϑ
is chosen, the number which measures this property will always vary in the
same sense.
In accordance with what was said above, the number can be taken as an
appropriate measure of the temperature of system T, and consequently, of system U.
Therefore, every time the system T, chosen once and for all, can be attached to a
certain system U independent of system T and forming with it an isolated system at
equilibrium, the degree of warmth possessed by the system U under these conditions
can effectively be put into correspondence with a value of the temperature.
When the system T and the property of the system whose measure gives the
numerical value of the temperature have been defined, a choice of thermometer has
been made. When an indication of the numerical value of the temperature of the
system U is given, the thermometer which has been chosen should evidently be
mentioned in order that this indication makes sense.
We leave to the reader the task of explaining the above generalities in application
to the various thermometers ordinarily employed.
3 These features are perceived by us in more or less gross experience, such as, for example, by
allowing the direct use of our senses. Then, by way of hypotheses, we have accepted that the system
T possesses them either rigorously or with an approximation superior to our first observations.
Chapter 4
The Principle of the Conservation of Energy
4.1 The Mechanical Work1 and the Energy of a System

We can, by our efforts, produce, or assist in producing, a certain transformation in
a system: we can displace a body, start it off with a certain velocity, break it, or
deform it. On the other hand, we can employ our efforts to place an obstacle in
the way of a transformation of the system, to impede the transformation: we can
halt a moving body, slow it down, and prevent it being deformed. We [290] say
then that we have performed a certain task, done a certain [amount of] mechanical
work.
Daily experience teaches us that we can substitute for our personal action a
body or a collection of bodies capable of producing, or contributing towards the
change that we produce or to which we contribute, and of preventing the change
that we prevent. The object of physics is, in practice, in a great number of cases
precisely to know which bodies can be substituted for our personal activity in order
to facilitate or hinder a certain change, which machines can replace labour in the
accomplishment of a certain task. We regard the mechanical work that we would
have performed if we had ourselves acted on the system which is transformed as
performed by the body or collection of bodies that we have substituted for ourselves
or our fellow men.
This notion of the mechanical work done by bodies external to a system while the
system undergoes a certain change is one we carry over to the case where the change
undergone by the system is of such a nature that our personal action can neither
help nor hinder it. The mechanical work done by external bodies is considered to
represent the mechanical work which an operator with a constitution unlike ours
would perform, and which is capable of bringing to the transformation of the system
the help or hindrance brought to bear by the external bodies.
Therefore, when a system is transformed in the presence of external bodies, we
consider the external bodies as contributing to this transformation either by causing
it or by aiding it or by impeding it. It is this contribution, whose nature remains
1 oeuvre [Duhem uses both oeuvre and travail, the latter introduced in Chapter 5 below with the
sense of work as usually understood in thermodynamics.]

52 4 The Principle of the Conservation of Energy
obscure to us, that we call the mechanical work done during a transformation of a
system by the bodies external to the system.
Without seeking to penetrate the nature of this contribution, which is not the
object of physics, but of metaphysics, we will endeavour to create a mathemat-
ical expression appropriate for serving as a symbol of this contribution. To this
end, we will determine the form of an expression subject to certain conventions.
These conventions are not established by chance; we will choose them so that
they reflect the most simple and salient characteristics [291] presented by the
notion of mechanical work, or, at least, so that they easily accord with these
characteristics.
Let us suppose that a system, subjected to the action of certain external bodies,
undergoes a certain transformation. Let us then suppose that the same system under-
goes the same transformation, now with different bodies external to the system. The
mechanical work done in the first case and in the second case by the bodies exter-
nal to the system is the same, although the bodies are different. This characteristic,
that we must attribute to the notion of mechanical work, leads us to put forward the
following convention:
FIRST CONVENTION.—The magnitude representing the mechanical work done during
a transformation of a system by the bodies external to this system is determined when
the nature of the system and the transformation to which it is subjected is known. It is
independent of the bodies external to the system.
A second convention follows that it is very natural to adopt:

SECOND CONVENTION.—Let us suppose that a system were to undergo various trans-
formations 1, 2, . . . , n, successively, while the bodies external to the system perform
mechanical work represented, respectively, by the algebraic magnitudes G1 , G2 , . . . , Gn .
The totality of transformations 1, 2, . . . , n can be considered as a single transformation.
The mechanical work done by the bodies external to the system during this resultant
transformation will be represented by the magnitude (G1 + G2 + . . . + Gn ).
Now that no confusion between the mathematical symbol and the notion that it
represents need be feared, we can give the name mechanical work done during the
transformation of a system by the bodies external to the system to the algebraic
magnitude which represents this mechanical work.
Let us imagine that a system begins in a certain initial state with a certain ini-
tial motion, and that a series of modifications takes it, [192] at the end of a certain
time, to a final state identical to its initial state, with a final motion identical to its
initial motion. We will regard the mechanical work that the external bodies have
performed during a part of this transformation as having been destroyed by the
mechanical work done during the remainder of the transformation, so that the total
mechanical work will be zero. We will thus be led to put forward the following
convention:
THIRD CONVENTION.—When a system traverses a closed cycle, the mechanical work done
during the course of the cycle by the bodies external to the system is equal to 0.
This convention suggests, in the first place, that the magnitude representing the
mechanical work done will not have the same sign for every change in a system.
4.1 The Mechanical Work and the Energy of a System 53
Some of the various changes constituting a closed cycle must correspond to positive
mechanical work, and the others negative, so that the sum of positive mechanical
work is exactly compensated by the sum of negative mechanical work.
But in addition, this convention furnishes much more precise information on the
form of the magnitude G representing the mechanical work done during the change
of a system by bodies external to the system.
Consider two different states of the same system. In one of these states, which
we designate with the symbol 1, the variables defining the properties of the system
take the values α1 , β1 , . . . , λ1 and the velocities of the various particles composing
the system have the components u1 , v1 , w1 ; u 1 , v 1 , w 1 ; . . . . In the other state, which
we designate with the symbol 2, the variables defining the properties of the system
take the values α2 , β2 , . . . , λ2 , and the velocities of the various particles composing
the system have the components u2 , v2 , w2 ; u 2 , v 2 , w 2 ; . . . .
Imagine a certain number of distinct changes M12 , M12 , M12 , . . . , all of which
take the system from state 1 to state 2, and let M21 be a change in which the system
goes from state 2 to state 1.
Let
G21 , G12 , G12 , . . . , G12
be [293] the mechanical work done by the bodies external to the system during the
changes
M12 , M12 , M12 , . . . , M21 .
The two changes M12 , M21 , imposed on the system one after the other, take it in
a closed cycle. The same holds for the two changes M12 , M21 , and again the two
changes M12 , M21 , . . . .
The preceding convention then yields
G21 + G12 = 0,
G12 + G12 = 0,
G12 + G12 = 0,
..................
or
G21 = G1 2 = G1 2 = . . . .
The result that this equation expresses may be stated thus:
The mechanical work accomplished by external bodies on a system during a change of the
system depends on the state and the motion of the system at the beginning and at the end of
the change, but not on any other details characterising this change.
In compliance with this statement, we will from now on put
G21 = ψ(α1 , . . . ; u1 , v1 , w1 . . . |α2 , . . . ; u2 , v2 , w2 . . .)
We will now give the function ψ a more explicit form.

We consider the system that we are studying to be in a certain state, determined
once and for all, that we will designate by the index 0. In this state, α0 , β0 , . . . , λ0
are the values of the variables which define the properties of the system, and
u0 , v0 , w0 ; u0 , v0 , w0 , . . . are the components of the velocities with which the material
particles composing it move.
If the system passes from the state 0 to the state 1, the external bodies perform
the mechanical work
G10 = ψ(α0 , . . . ; u0 , v0 , w0 . . . |α1 , . . . ; u1 , v1 , w1 . . .)
If [294] the system passes from the state 1 to the state 2, the external bodies
perform the mechanical work
G21 = ψ(α1 , . . . ; u1 , v1 , w1 . . . |α2 , . . . ; u2 , v2 , w2 . . .)
These two transformations, effected one after the other, constitute a transforma-
tion carrying the system from the state 0 to the state 2. During such a transformation,
the external bodies perform the mechanical work
G20 = ψ(α0 , . . . ; u0 , v0 , w0 . . . | . . . α2 , . . . ; u2 , v2 , w2 . . .).
But on the other hand, the mechanical work done by the external bodies
during

this last transformation should, by the second convention, have the value G10 + G21 .
Accordingly,

G20 = G10 + G21
or
⎧
⎨ ψ(α0 , . . . ; u0 , v0 , w0 . . . |α2 , . . . ; u2 , v2 , w2 . . .)
⎪
= ψ(α0 , . . . ; u0 , v0 , w0 . . . |α1 , . . . ; u1 , v1 , w1 . . .) (1)
⎪
⎩
+ ψ(α1 , . . . ; u1 , v1 , w1 . . . |α2 , . . . ; u2 , v2 , w2 . . .).
This identity (1) will determine the form of the function ψ.

Since the state 0 has been determined once and for all, the quantities
α0 , β, . . . , λ0 , u0 , v0 , w0 , u0 , v0 , w0 , ...
are not variables, but constants, and the quantity

ψ(α0 , . . . ; u0 , v0 , w0 . . . |α, . . . ; u, v, w . . .)
is a function of the variables
4.1 The Mechanical Work and the Energy of a System 55
α, β, . . . , λ, u, v, w, u , v , w , ...
alone.
We [295] can therefore put

ψ(α0 , . . . ; u0 , v0 , w0 . . . |α, . . . ; u, v, w . . .)
(2)
= ζ (α, . . . ; u, v, w . . .)
and Equation (1) gives us

ψ(α1 , . . . ; u1 , v1 , w1 . . . |α2 , . . . ; u2 , v2 , w2 . . .)
(3)
= ζ (α2 , . . . ; u2 , v2 , w2 . . .) − ζ (α1 , . . . ; u1 , v1 , w1 . . .).
The mechanical work done by the external bodies during any change of a system is equal
to the increase, by the effect of this change, of a certain magnitude which is determined
without ambiguity when the state of the system and its motion is known.
This magnitude, defined by Equation (2), is called the energy of the system.
In order to define the energy of the system, we must chose a certain state of the
system once and for all, which we have designated the state 0. But this choice was
arbitrary. We could have argued in the same way having chosen once and for all
the state ω, different from the state 0. If we designate by αω , βω , . . . , λω the values
of the variables determining the properties of the system in the state ω, and by
uω , vω , wω ; u ω , v ω , w ω the components of the velocities moving the various parts
of the system in this state, we would obtain a new determination of the energy of
the system, defined by the equation

ψ(αω , . . . ; uω , vω , wω . . . |α, . . . ; u, v, w . . .)
(2a)
= ε(α, . . . ; u, v, w . . .).
Let us evaluate the difference between the corresponding values of these two
determinations of energy.
The Equations (2) and (2a) yield
ε(α, . . . ; u, v, w . . .) − ζ (α, . . . ; u, v, w . . .)
= ψ(αω , . . . ; uω , vω , wω . . . |α, . . . ; u, v, w . . .) −
ψ(α0 , . . . ; u0 , v0 , w0 . . . |α, . . . ; u, v, w . . .).
But [296] Equation (3) shows that we have
ψ(α0 , . . . ; u0 , v0 , w0 . . . |α, . . . ; u, v, w . . .) =
− ψ(α, . . . ; u, v, w . . . |α0 , . . . ; u0 , v0 , w0 . . .).
Accordingly,
ε(α, . . . ; u, v, w . . .) − ζ (α, . . . ; u, v, w . . .)
= ψ(αω , . . . ; uω , vω , wω . . . |α, . . . ; u, v, w . . .) +
ψ(α, . . . ; u, v, w . . . |α0 , . . . ; u0 , v0 , w0 . . .),
or, in virtue of Equation (1)2
ε(α, . . . ; u, v, w . . .) − ζ (α, . . . ; u, v, w . . .) =
ψ(αω , . . . ; uω , vω , wω . . . |α0 , . . . ; u0 , v0 , w0 . . .).
The right-hand side of this equation is a constant.
Therefore, the values of two determinations of the energy for the same state
of the system differ by a constant. This can be stated by saying that the energy is
determined up to a constant.
4.2 The Kinetic Energy3 and the Internal Energy

Every transformation undergone by a system can be decomposed into two elements.
First, there is a change of state: the variables defining the state of the system pass
from the values α1 , β1 , . . . , λ1 to the values α2 , β2 , . . . , λ2 . Second, there is a change
of motion: the components of the velocities with which the elementary parts of
the material of the system move pass from the values u1 , v1 , w1 , u 1 , v 1 , w 1 , . . .
to the values u2 , v2 , w2 , u 2 , v 2 , w 2 , . . . . The following convention, as arbitrary as
the preceding ones, rests on the distinction between these two elements of any
transformation:
FOURTH CONVENTION.—The mechanical work done by the external bodies on a system
during a transformation of the system is the sum of two terms: the one depends on the
change of state of the system [297] and not on its movement, and the other depends on the
change of movement of the system but not on its state.
This convention is expressed by the identity:

ψ(α1 , . . . ; u1 , v1 , w1 . . . |α2 , . . . ; u2 , v2 , w2 . . .)
(4)
= ϕ(α1 , . . . λ1 |α2 , . . . , λ2 ) + χ (u1 , v1 , w1 . . . | u2 , v2 , w2 . . .).
In virtue of this Equation (4), Equation (2) becomes

ζ (α, . . . ; u, v, w . . .) =
(5)
ϕ(α0 , . . . λ0 |α, . . . , λ) + χ (u0 , v0 , w0 . . . | u, v, w . . .).
2 [Where ε as it stands in the following equation replaces ζ in the original.]

3 force vivre.
4.2 The Kinetic Energy and the Internal Energy 57
The quantities α0 , β0 , . . . , λ0 are chosen once and for all, and the quantity
ϕ(α0 , . . . λ0 |α, . . . , λ)
is therefore only a function of the variables α, β, . . . , λ, so we can put
ϕ(α0 , . . . λ0 |α, . . . , λ) = U (α, . . . , λ) (6)
Similarly, the quantities u0 , v0 , w0 . . . are chosen once and for all, and the quantity
χ (u0 , v0 , w0 . . . | u, v, w . . .)
is therefore a function of the variables u, v, w, . . . only, so we can put
χ (u0 , v0 , w0 . . . | u, v, w . . .) = K (u, v, w . . .) (7)
In virtue of Equations (6) and (7), Equation (5) becomes
T
ε(α, . . . ; u, v, w . . .) = U(α, β, . . . , λ) + (8)
E
The energy of a system is the sum of two terms, the one that depends only on the state of
the system and not at all on its motion, and [298] the other, independent of the state of the
system, that is known when the velocity of each of the elementary parts of the system is
known.
The first term is called the potential energy or the internal energy; the second is
called the actual energy or the kineticenergy. We will try to determine the form of
the latter.
To begin with, we note that the velocities designated by u0 , v0 , w0 , . . . are chosen
arbitrarily. Henceforth we agree to take
u0 = 0, v0 = 0, w0 = 0, . . .
Now, in accordance with Equation (7), the function K(u, v, w, . . .) equals 0 if we

have
u = u0 , v = v0 , w = w0 , . . .
Consequently, we are henceforth assured that the function
K(u, v, w, . . .)
equals 0 when we have
u = 0, v = 0, w = 0, . . .
Second, we will build on a convention that might be regarded as unavoidable, which

we state as follows:
FIFTH CONVENTION.—When a system is formed from several parts which are independent
of one another and infinitely distant from one another, the mechanical work accomplished
during any change of the system by the bodies external to the system is the sum of the
mechanical work accomplished during the change corresponding to each of the parts by
the bodies external to these parts.
From this statement the following consequences are easily deduced:

When a system is formed from several parts which are independent [299] of one another
and infinitely distant from one another, the internal energy of the system is the sum of the
internal energies of the isolated parts, and the kinetic energy of the system is the sum of the
kinetic energy of the isolated parts.
Since the kinetic energy does not depend on the state of the system, we can,
in order to determine the form of this energy, mentally decompose the system
into infinitesimal material parts, isolate these material parts from one another, and
disperse them into space so that they become infinitely separated. This operation
changes the state of the system, but not its kinetic energy, if care is taken to preserve
for each material particle, after the operation, the velocity which it had before the
operation.
But after this operation the kinetic energy of the system is the sum of the
kinetic energies of the various parts which compose it. We are therefore reduced
to determining the form of the kinetic energy of an infinitesimal portion of matter.
Let u, v, w be the components of the velocity of an infinitesimal portion of mate-
rial. The kinetic energy will be a function, k(u, v, w), of u, v, w, where the form of
the function k depends on the nature of the material particle, but not on its state. For
a given particle, the form of this function is invariable; it is this form that we wish
to know.
SIXTH CONVENTION.—The bodies external to an infinitesimal system always perform
the same mechanical work when, beginning at rest, they impart a velocity of the same
magnitude, whatever the direction in space of this velocity.
This convention, which amounts to saying that, in absolute space, all directions
are equivalent, might be regarded as logically necessary. It immediately entails this
consequence: The function k(u, v, w) does not depend separately on the three com-
ponents u, v, w of the velocity V, but only on the magnitude of this latter velocity. We
can replace the symbol k(u, v, w) by the symbol k(V).
SEVENTH CONVENTION.—[300] The mechanical work done by the external bodies in
imparting a certain velocity to a particle originally at rest is always mechanical work with
the same sign.
We will agree to count such mechanical work as positive. The function k(V) will
then be essentially positive.
EIGHTH CONVENTION.—Let P, P be two material particles. If the bodies external to these
two particles perform the same mechanical work to impart to both a certain velocity V0 ,
then they will also perform the same mechanical work when imparting to each the same
velocity V, whatever this latter velocity might be.
In other words, the equation
k(V) = k (V)
cannot hold for a certain value V0 of V without holding identically.

Given this convention, let us consider a particle to which a function k(V) corre-
sponds. Divide it into very small parts. These parts will have as their kinetic energy
x(V), x (V), . . . , respectively. We know that the kinetic energy k(V) of the particle4
will have the value
k(V) = x(V) + x (V) + . . . . (9)
Moreover, each of the quantities x(V), x (V), . . . is positive, and each is therefore
less than k(V). Thus, if a material element is a part of another material element, then
for a given velocity it has less kinetic energy. Moreover, it is evident that the kinetic
energy of a material element for a given velocity varies continuously if the amount of
matter enclosing this element is continuously increased or decreased. By appealing
to these two propositions, the following proposition can be easily demonstrated:
Given a material particle P, it is always possible, for any integer N, to divide it
into N [301] particles which, for a given value, have the same kinetic energy. The
preceding convention shows that these N particles always have the same kinetic
energy if they all move with the same velocity, whatever this velocity may be. If this
proposition is combined with Equation (9), it is easily seen that, for a given velocity,
the kinetic energy of each of these particles is an Nth part of the kinetic energy of
the particle P.
Take any two material elements P and P whose kinetic energies are represented
by the functions k(V) and k (V). Let V0 be a particular value of V. Consider the ratio
k (V0 )/k(V0 ) and suppose first that the ratio is commensurable.
Suppose5
k (V0 ) N
= ,
k (V0 ) N
where N and N are integers having no common factor.
Let the particle P be divided into N elements , . . . having, for a given velocity,
the kinetic energies x(V), . . . equal to one another and to k(V)/N.
Let the particle P be divided into N elements , . . . having, for a given velocity,
the kinetic energies x (V),6 . . . equal to one another and to k (V)/N .
4 la première partie.
5 [The N
original reads on the right-hand side.]
N
6 [x(V) in the original.]
We have
k (V0 ) k (V0 )
x (V0 ) = , x (V0 ) = ,
N N
and consequently,
x (V0 ) = x (V0 ) .
Then, given the preceding convention, whatever V may be,
x (V) = x (V) ,
so that the equations
k (V) = Nx (V) , k (V) = N x (V)
[302] yield
k (V) N k (V0 )
= = .
k (V) N k (V0 )
The kinetic energies which correspond to the same velocity for two different material
elements stand in a ratio which is independent of this velocity.
We have just proved this proposition for the case where the ratio k (V0 )/k(V0 ) is
commensurable. But it is evidently general, because if the ratio k (V0 )/k(V0 ) were to
vary with V, it would be a continuous function of V. Therefore, for certain values of
V, it would pass through commensurable values; the preceding argument would then
show that the hypothesis according to which k (V)/k(V) varies with V is absurd.
Consider two bodies of any finite extension, C and C , and suppose that all the
points of these two bodies move at a certain velocity V. Let K(V) and K (V) be
their kinetic energies. Each of these kinetic energies is the sum of energies which,
for the same velocity, the various material elements into which each of these two
bodies might be supposed divided would have. Consequently, it is not difficult to
deduce from the preceding proposition that the ratio K (V)/K (V) is independent of
the velocity V; it depends only on the nature of the two bodies C and C .
Take a definite body , for example the standard kilogram in the Archives.
Suppose that all the points of this body move at the same velocity V. Let χ(V)
be the kinetic energy of the body in these circumstances.
Then let C be any other body, finite or infinitesimal. Suppose that all the points
of this body move at the same velocity V.
Let K(V) be the kinetic energy of the body C under these conditions, and put
K(V) = Mχ(V). (10)

Regarding [303] the number M defined by this Equation (10), we can immedi-
ately affirm the following propositions:
1. The number M does not depend on the velocity V; it only depends on the nature
of the bodies and C;
2. The number M is essentially positive;
3. The number M associated with an infinitesimal material element is infinitesimal
like this element;
4. The number M connected with a collection of bodies is equal to the sum of the
analogous numbers connected with each of the various bodies comprising this
collection;
5. For the body , the number M is equal to 1.
The number M designates the mass of the body C. The body is said to constitute
the standard mass or the unit of mass.
It can be seen that the mass of a body is proportional to the mechanical work that
must be done on it by the bodies external to it in order to take it from a state of rest
to a state of motion in which all its points move at a given velocity.
Consider any system whose various elementary parts P, P , P , . . . move at veloc-
ities V, V , V , . . . , whatever they may be. The kinetic energy of the system is the
sum of the kinetic energies of the various parts. Now, if the masses of the elements
P, P , P , . . . are designated by m, m , m , . . . , then these partial kinetic energies
would, in view of Equation (10), have the respective values
mχ(V), m χ(V ), m χ(V ), . . .
The kinetic energy of the system therefore has the value
K = mχ(V) + m χ(V ) + m χ(V ) + . . . = mχ(V). (11)
The form of this kinetic energy will therefore be completely known to us if we

determine the form of the function χ(V). In order to do this, we appeal to a new
convention.
NINTH CONVENTION.—[304] In order to impart a certain velocity in a certain direction D
on all the points of a material element, the external bodies must perform the same mechan-
ical work, whether the material element starts from rest or is already in motion with any
velocity whatever in a direction D , normal to D.
We accept this convention which, however natural it might strike us, has by no
means the character of a logical necessity.
Let V be a velocity and u, v, w its components along three rectangular coordinates
Ox, Oy, Oz. Suppose these three components are positive. Take a material element
of mass m and give it a velocity u in the direction Ox. We will perform mechanical
work of magnitude
g1 = mχ(u)
Let us now impart to this element with velocity u in the direction Ox a velocity v
in the direction Oy. In accordance with the preceding convention, the mechanical
work done in this second change is the same as if we had imparted the veloc-
ity in the direction Oy to the moving body starting at rest, and therefore is of the
magnitude
g2 = mχ(v).
To this element now moving with the two velocities u along Ox and v along Oy,
we apply mechanical work capable of imparting to it a velocity w along Oz. This
mechanical work, equal by the preceding convention to that which would impart a
velocity along Oz to a body starting at rest, takes the value
g3 = mχ(w).
These three changes together take the moving body from rest to a velocity V.
The mechanical work done in the three changes taken together, which has the value
(g1 + g2 + g3 ), must be equal to mχ(V).
We [305] therefore have the identity
χ(V) = χ(u) + χ(v) + χ(w).
Differentiating this identity with respect to u, and using the equation
u2 + v2 + w2 = V 2 ,
we find

dχ (u) du u
= .
dχ (V) dV V

The values of u and V are arbitrary. It can therefore be seen that dχ (V) dV should
be proportional to V. Noting, in addition, that χ(V) must go to zero with V yields
the conclusion:
The function χ(V) is proportional to V2 .
We put
V2
χ (V) = , (12)
2E
where E is an essentially positive quantity independent of V.
Then, from Equation (11), the kineticenergy7 of any system is given by the
formula
7 l’énergie cinétique.
4.3 The Principle of the Conservation of Energy 63
1 mV 2
K= . (13)
E 2
The quantity
mV 2
T= (14)
2
is [also] called the kinetic energy8 of the system.

The constant E is called the mechanical equivalent of heat; we will see the reason
for this later. The value of [306] this constant depends on the units of length, time,
mass and mechanical work. Equation (12) shows, in fact, that if the standard mass
at rest is given a velocity which makes it traverse a unit of length during a unit of
time, the mechanical work done should be numerically equal to 1/2E.
From the Equations (8), (13) and (14), the energy of any system is given by the
formula
T
ε(α, . . . ; u, v, w . . .) = U (α, β, . . . , λ) + . (15)
E
It remains to state one last convention concerning the internal energy
U (α, β, . . . , λ).
TENTH CONVENTION9 .—The value of the internal energy of a system does not change
when its position in absolute space alone changes, without changing any other properties
of the system.
Thus, among the variables α, β, . . . , λ there are none figuring in the expression
U (α, β, . . . , λ) which serve solely to fix the absolute position of the system in space.
4.3 The Principle of the Conservation of Energy

The conventions enumerated above have led us to define the form of a certain alge-
braic quantity appropriate to serve as the symbol for the notion of the mechanical
work done during a transformation of a system by the bodies external to the sys-
tem. This algebraic quantity is equal to the increase in the total energy of the system
brought about by the transformation.
8 force vivre. [Since (14) is how the term “kinetic energy” is now standardly used, Duhem’s archaic
term is not preserved in this translation. Of the terms introduced by (13) and (14), T is the one
Duhem uses most frequently, dividing by E as appropriate (see e.g. (15) below).]
9 This convention might seem evident to some minds. If, however, it is noted that the internal energy
of the system depends on the absolute movement of the system, it can be seen that it would be by
no means absurd to regard it as dependent on its absolute position in space.
But in the case where no external body interacts with the system, the mechanical
work done by the external bodies during a change [307] should evidently be equal
to 0. We are therefore led to state the following proposition:
When a material system, isolated in space, undergoes any transformation, the total energy
of the system remains unchanged by the effect of this transformation.
This proposition, expressed by the equation
T
U(α, β, . . . , λ) + = const., (16)
E
constitutes the principle of the conservation of energy.
It differs from everything that has been said so far, which has been of an arbitrary
character. In the final analysis, we are absolutely free to consider the quantity10
T2 T1
G = U(α2 , β2 , . . . , λ2 ) + − U(α1 , β1 , . . . , λ1 ) − ,
E E
and give it whatever name we like, for example the name mechanical work done by
the bodies external to the system. But when we state that, in every transformation of
an isolated system, there exists a quantity of the form
T
U(α, β, . . . , λ) + ,
E
which remains invariable, we state a proposition whose consequences may confirm
or be contrary to experience. It is a proposition that we cannot accept or reject as
the fancy takes us. In a word, it is a physical hypothesis, the first that we have
encountered so far. The purpose of the considerations in Sections 4.1 and 4.2 was to
introduce this hypothesis, which we have been led to formulate; they do not prove
it. It is for experience to verify its immediate and more distant consequences. [308]
10 [The third term on the right is U (α1 , β3 , . . . , λ1 ) in the original followed by the sign +.]
Chapter 5
Work and Quantity of Heat
5.1 The Establishment of a Fundamental Equation

Consider a system , isolated in space, which can be divided into two independent
systems, S and S .
Let α, β, . . . , λ, a, b, . . . , l be the variables determining the position of the system
S, isolated in space, and its state. Of these variables, α, β, . . . , λ are those which,
relative to system S, figure in the Equations (1) of Chapter 3, whilst a, b, . . . , l do
not figure there.
Similarly, let α , β , . . . , λ , a , b , . . . , l be the variables determining the position
and state of the system S , isolated in space.
If the system S is isolated in space, its internal energy will be a certain function
of α, β, . . . , λ, a, b, . . . , l, which we will designate by
U(α, β, . . . , λ, a, b, . . . , l).
Its kinetic energy will be a quadratic form of the variables

dα dβ dλ
u= , v= , ..., w =
dt dt dt
which may depend on the variables α, β, . . . , λ in some way, but not on the variables
a, b, . . . , l. Let T be this kinetic energy.
If the system S is isolated in space, its internal energy will be a certain function
of α , β , . . . , λ , a , b , . . . , l which we will designate by
U (α , β , . . . , λ , a , b , . . . , l ).
Its kinetic energy will be a quadratic form of the variables

dα dβ dλ
u = , v = , . . . , w =
dt dt dt
which [309] may depend on the variables α , β , . . . , λ in some way, but not on the
variables a , b , . . . , l . Let T be
this kinetic energy.
Let us consider the system .

66 5 Work and Quantity of Heat
If the variables
α, β, . . . , λ, a, b, . . . , l,
α , β , . . . , λ , a , b , . . . , l ,
are known, then the position and the state of each

of the two systems S and S which
compose it are known.The state of the system is therefore known. The internal
energy of the system will be a function of these variables. We designate it by
ϒ(α, β, . . . , λ, a, b, . . . , l, α , β , . . . , λ , a , b , . . . , l ).
Clearly, we can write
ϒ = U(α, β, . . . , λ, a, b, . . . , l) + U (α , β , . . . , λ , a , b , . . . , l )
(1)
+(α, β, . . . , λ, a, b, . . . , l, α , β , . . . , λ , a , b , . . . , l ).

is clearly equal to (T + T ).
The kinetic energy of the system
The total energy of the system will have the value
1
ε=ϒ+ (T + T ). (2)
E

Let us write that the infinitesimally small change undergone by the system
during the time dt leaves the value of this energy ε invariable.
Put
da db dl
ϕ = , χ= , ..., ψ = ,
dt dt dt
da db dl
ϕ = , χ = , . . . , ψ = .
dt dt dt
In [310] accordance with Equation (1), ϒ undergoes, during dt, a variation

∂U ∂ ∂U ∂ ∂U ∂
δϒ = + u+ + v + ... + + w
∂α ∂α ∂β ∂β ∂λ ∂λ

∂U ∂ ∂U ∂ ∂U ∂
+ ϕ+ + χ + ... + + ψ
∂a ∂a ∂b ∂b ∂l ∂l

∂U ∂ ∂U ∂ ∂U ∂
+ u + + v + ... + + w
∂α ∂α ∂β ∂β ∂λ ∂λ

∂U ∂ ∂U ∂ ∂U ∂
+ ϕ + + χ + ... + + ψ
∂a ∂a ∂b ∂b ∂l ∂l
5.2 On Work 67
As for the variation undergone by the kinetic energy during the same time, a
known method allows this to be put in the form

∂T d ∂T ∂T d ∂T
δ T + T =− − u+ − v + ...
∂α dt ∂u ∂β dt ∂v

+ − w+ + u
∂λ dt ∂w ∂α dt ∂u

∂T d ∂T T d ∂T
+ + v + ... + + w dt.
∂β dt ∂v ∂λ dt ∂w
If we say that the total energy, ε, given by Equation (2), remains invariable, we
find the following equation:
⎧
⎪ ∂U ∂ 1 ∂T d ∂T
⎪
⎪ ∂α + − − u + ...
⎪
⎪ ∂α E ∂α dt ∂u
⎪
⎪
⎪
⎪ ∂U ∂ 1 ∂T d ∂T
⎪
⎪ + + − −
⎪
⎪ ∂λ ∂λ E ∂λ dt ∂w
w
⎪
⎪
⎪
⎪
⎪
⎪ ∂U ∂ ∂U ∂
⎪
⎨ + + ϕ + . . . + + ψ
∂α ∂α ∂l ∂l
(3)
⎪
⎪ ∂U ∂ 1 ∂T d ∂T
⎪
⎪ + + − − u + . . .
⎪
⎪ ∂α ∂α E ∂α dt ∂u
⎪
⎪
⎪
⎪ ∂U ∂ 1 ∂T d ∂T
⎪
⎪ + + − − w
⎪
⎪
⎪
⎪ ∂λ ∂λ E ∂λ dt ∂w
⎪
⎪
⎪
⎪ ∂U ∂ ∂U ∂
⎩+ + ϕ + ... + + ψ = 0.
∂a ∂a ∂l ∂l
This [311] fundamental equation will serve as our point of departure for the
considerations which are the object of the present chapter.
5.2 On Work1
Let us put
⎧
⎪ ∂ ∂
⎪
⎪ E = −A, E = −A,
⎪
⎪ ∂α ∂a
⎪
⎪
⎪
⎨ ∂ ∂
E = −B, E = −B, (4)
⎪ ∂β ∂b
⎪
⎪
⎪
⎪ ............, ............,
⎪
⎪ ∂ ∂
⎪
⎩E = −L, E = −L,
∂λ ∂l
1 travail.
We will say that the quantities A, B, . . . , L represent the forces exerted by the
system S on the system S, and that the quantities A, B, . . . , L represent the influ-
ences exerted by the system S on the system S. The totality of forces and influences
exerted by the system S on the system S are called the actions of the system S on
the system S.
The quantity
(Au + Bv + . . . + Lw)dt
is the work done during the time dt by the forces that the system S exerts on the
system S. The quantity
(Aϕ + Bχ + . . . + Lψ)dt
is the work done during the time dt by the influences that the system S exerts on the
system S. The sum of these two quantities is the work done during the time dt by
the actions of the system S on the system S.
Consider a virtual change of the system S; let
δα, δβ, . . . , δλ, δa, δb, . . . , δl

be the variations undergone as a result of this change by the variables
α, β, . . . , λ, a, b, . . . , l.
The [312] expressions
Aδα+Bδβ + . . . + Lδλ,
Aδa + Bδb + . . . + Lδl,

Aδα + Bδβ + . . . + Lδλ + Aδa + Bδb + . . . + Lδl,
are called, respectively,

The virtual work of the forces done by the system S on the system S;
The virtual work of the influences exerted by the system S on the system S;
The virtual work of the actions done by the system S on the system S.
The work (real or virtual) of the actions of the system S on the system S has, in
accordance with Equations (4), the value

∂ ∂ ∂ ∂
−E δα + . . . + δλ + δa + . . . + δl .
∂α ∂λ ∂a ∂l
This is not, in general, the total differential of a uniform function of the vari-
ables α, β, . . . , λ, a, b, . . . , l determining the system S. In order to transform this
expression into a total differential, it is necessary to add the term

∂ ∂ ∂ ∂
−E δα + . . . + δλ + δa + . . . + δl ,
∂α ∂λ ∂a ∂l
that is to say, the work of the actions of the system S on the system S .
5.2 On Work 69
Thus, the work of actions of the system S on the system S is not, in general, a
total differential, but the work of the mutual actions of the two systems S and S is
always the total differential
of a function which is defined in a uniform way when
the state of the system constituted by the aggregate of the two systems S and S is
known.
The function E whose total differential, changed in sign, [313] gives the work
of the mutual actions of the two systems S and S is called the potential of these
actions.
Like the function ϒ, this potential depends on the properties of the two systems
S and
S and on their relative position, but not on the absolute position that the sys-
tem occupies in space; the same holds for the mutual actions of the two systems
S and S .
This theorem can be generalised and extended to a system formed from n
independent systems S1 , S2 , . . . , Sn . In order not to complicate the notation,but
without seriously infringing on generality, we will suppose that the system is
formed of just three partial systems S1 , S2 , S3 .
Let
α1 , . . . , λ1 , a1 , . . . , l1 ,
α2 , . . . , λ2 , a2 , . . . , l2 ,
α3 , . . . , λ3 , a3 , . . . , l3
be the three systems of variables defining, respectively, the state of each of the three
systems S1 , S2 , S3 .
Let
U1 (α1 , . . . , λ1 , α1 , . . . , l1 ),
U2 (α2 , . . . , λ2 , a2 , . . . , l2 ),
U3 (α3 , . . . , λ3 , a3 , . . . , l3 )
be the internal energies of these systems

considered in isolation.
The internal energy of the system can evidently be put in the following
form:
ϒ = U1 + U2 + U3 +
(5)
(α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 , α3 , . . . , λ3 , a3 , . . . , l3 ).
In order to construct the function , we can proceed in the following way. We

envisage first the system 23 formed from the composition of the two systems S2
and S3 , whose internal energy will have the [314] form
ϒ23 = U2 + U3 + 23 (α2 , . . . , λ2 , a2 , . . . , l2 , α3 , . . . , λ3 , a3 , . . . , l3 ). (6)


Then we
combine the system 23 with the system S1 in such a way as to form
system . The internal energy of this system will be of the form
ϒ = U1 + ϒ23 +
(7)
X1 (α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 , α3 , . . . , λ3 , a3 , . . . , l3 ).
Comparison of formulas (5), (6) and (7) yields the equation
= 23 (α2 , . . . , λ2 , a2 , . . . , l2 , α3 , . . . , λ3 , a3 , . . . , l3 ) +
(8)
X1 (α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 , α3 , . . . , λ3 , a3 , . . . , l3 ).
The definitions
put forward above show us that the actions exerted by the system
S1 on the system 23 , that is, the composite of systems S2 , S3 , are determined by
the equations
⎧
⎪ ∂X1 ∂X1
⎪
⎪ A1 = −E , A1 = −E ,
⎪
⎨ ∂α1 ∂a1
............, ..............., (4a)
⎪
⎪
⎪
⎪ ∂X1 ∂X1
⎩ L1 = −E , L1 = −E ,
∂λ1 ∂l1
which allows us to write, in virtue of Equation (8),

⎧
⎪ ∂1 ∂ψ1
⎪
⎪ A1 = −E , A1 = −E ,
⎪
⎨ ∂α1 ∂a1
..............., ..............., (4b)
⎪
⎪
⎪
⎪ ∂1 ∂ψ1
⎩ L1 = −E , L1 = −E .
∂λ1 ∂l1
The actions that the system S2 undergoes on the part of the aggregate of systems S3
and S1 , and the actions that the system S3 undergoes on the part of the aggregate
[315] of systems S1 and S2 are determined analogously. The following proposition
is easily deduced:
In a complex system formed from several independent systems, each of the latter undergo
certain actions by the others in conjunction. All these actions, taken together, support a
potential.
This potential EΨ depends on the properties of the various independent systems compos-
ing the complex system, and on their relative position. It does not depend on the absolute
position that the complex system occupies in space.
Demonstrations analogous to those furnishing the Equation (8) allow us to write,

using similar notations, the equations
5.2 On Work 71
= 31 (α3 , . . . , λ3 , a3 , . . . , l3 , α1 , . . . , λ1 , a1 , . . . , l1 )
+ X2 (α2 , . . . , λ2 , a2 , . . . , l2 , α3 , . . . , λ3 , a3 , . . . , l3 , α1 , . . . , λ1 , a1 , . . . , l1 ).
= 12 (α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 )
+ X3 (α3 , . . . , λ3 , a3 , . . . , l3 , α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 ).
(8a)
The Equations (8) and (8a) will clearly be verified by putting
⎧
⎨ X1 = 31 + 12 ,
X2 = 12 + 23 , (9)
⎩
X3 = 23 + 31 ,
which entail
= 23 + 31 + 12 ,
but they do not necessarily entail the Equations (9).

Let us see to which consequences the Equations (9) lead.
The actions that the system S2 would exert on system S1 , were these [316] two
systems to exist alone, would be given by the equations
⎧
⎪ ∂12 ∂12
⎪
⎪ A12 = −E , A12 = −E ,
⎪
⎨ ∂α1 ∂a1
..............., ..................,
⎪
⎪
⎪
⎪ ∂12 ∂12
⎩ L12 = −E , L12 = −E .
∂λ1 ∂l1
The actions that the system S2 would exert on system S2 , were these [316] two
systems to exist alone, would be given by the equations
⎧
⎪ ∂13 ∂13
⎪
⎪ A13 = −E , A13 = −E ,
⎪
⎨ ∂α1 ∂a1
..............., ..................,
⎪
⎪
⎪
⎪ ∂13 ∂13
⎩ L13 = −E , L13 = −E .
∂λ1 ∂l1
These equations, together with Equations (4a) and (9), yield
A1 = A12 + A13 , A1 = A12 + A13 ,

..............., ..................,
L1 = L12 + L13 , L1 = L12 + L13 .
From these and other analogous equations which can be demonstrated in the same
way, the following theorem can be deduced:
In a complex system formed from several independent systems, the actions on each of the lat-
ter by all of the others is obtained by superposing the actions to which it would be subjected
by each of the others if each of these others were alone placed in its presence.
It can be seen that this theorem, although compatible with the definition of mutual
actions exerted between various systems, is not, however, a necessary consequence
of it. Whenever its exactitude is accepted in a particular theory, a hypothesis is
thereby introduced.
Returning to the general study of a system composed of n independent sys-
tems [317] S1 , S2 , . . . , Sn , suppose that these n systems are n bodies each one of
which occupies a variable position in space but remains in the same state. We will
then say that any change of the system is a displacement without change of state.
When the system Si alone is displaced in space, without changing its proper-
ties, the internal energy Ui of this system remains the same. Consequently, in a
displacement without change of state, each of the quantities U1 , U2 , . . . , Un remains

the same. In accordance with Equation (5), the internal energy ϒ of the system
differs only by a constant from the function . Accordingly, we have the following
theorem:
When a complex system formed from several independent systems is subjected only to dis-
placements without change of state, the totality of the actions exerted between the various
partial systems has as its potential the product of the internal energy of the complex system
and the mechanical equivalent of heat.
This theorem is used frequently in the applications of thermodynamics.

From this theorem unfolds another:
Let T be the
kinetic energy
of the system . The total energy of this system will
1
have the value ϒ + T . If the system is isolated, this total energy cannot
E
vary. The equation
1
ϒ+ T = const.
E

should be upheld in any change of the system. If the system is subject only
to displacement without change of state, this equation can be replaced by the
following
E + T = const.
or
Ed + dT = 0.
Now [318] (–Ed) represents the work done by the actions between the systems S1 ,
S2 , . . . , Sn , while the kinetic energy of the system is augmented by dT. We can
therefore state the following theorem:
Consider a complex system formed of independent, isolated systems and subject only to
displacements without change of state. In all real changes undergone by such a system,
the kinetic energy increases by an amount equal to the work accomplished by the actions
exerted between the various partial systems of which it is composed.
5.3 On the Quantity of Heat 73
We note that we have defined the forces exerted and influences exerted on a mate-
rial system by another system which is independent of it. We have not defined the
forces or the influences obtaining between two parts of the same system when the
two parts cannot be regarded as two independent systems. Such forces and influ-
ences are sometimes spoken of in physics. Thus, in electrodynamics one speaks of
the actions exerted between the various parts of the same conductor through which
currents flow. When such actions are considered, it is necessary to give a special
definition, and it will not come as a surprise if the actions thus defined do not sat-
isfy certain theorems to which actions exerted between independent systems, as
we have just defined them, are necessarily subject. This observation has important
applications in electrodynamics and electromagnetism.
5.3 On the Quantity of Heat

Return to the fundamental Equation (3).
Let
⎧
⎪
⎪ ∂U ∂T d ∂T
⎪ A−E
⎪ + − u + ...
⎪
⎪ ∂α ∂α dt ∂u
⎪
⎪
⎨ ∂U ∂T d ∂T ∂U
+ L−E + − w+ A−E ϕ + ... (10)
⎪
⎪ ∂λ ∂λ dt ∂w ∂a
⎪
⎪
⎪
⎪ ∂U
⎪
⎪
⎩ + L−E ψ dt = EdQ,
∂l
⎧
⎪
⎪ − E ∂U + ∂T − d ∂T u
⎪
⎪ A
⎪
⎪ ∂α ∂α dt ∂u
⎪
⎪
⎪
⎪
⎨ +...........................
⎪
(10a)2
⎪ + − E ∂U + ∂T − d ∂T w
⎪
⎪ L
⎪
⎪ ∂λ ∂λ dt ∂w
⎪
⎪
⎪
⎪ ∂U ∂U
⎪
⎪
⎩ + A − E ϕ + . . . + L − E ψ dt = E dQ .

∂a ∂l
[319]
Taking account of Equations (4) and the analogous equations relating to the
actions of the system S on the system S , Equation (3) becomes
dQ + dQ = 0. (11)

2 ∂U ∂U
In original, instead of .
∂α ∂α
We call the quantity dQ the quantity of heat released by the system S during
the time dt. When this quantity is negative, we say that its absolute value is the
quantity of heat absorbed by the system S during the same time. The quantity dQ
will similarly be called the quantity of heat released by the system S during the
time dt.
Thus, Equation (11) tells us that in all real changes of a complex isolated system
comprising two independent systems S and S , one of these systems S, S releases
as much heat as the other absorbs.
Consider the system S isolated in space. If the system S didn’t exist, all the terms
comprising the left-hand side of Equation (10a) would be identically zero. The same
holds for dQ and hence, in accordance with Equation (11), for dQ. From that we
have the following theorem:
When a system which is isolated in space undergoes a real change, it neither absorbs nor
releases heat.
Several theories of physics lead us to regard a body as able to release or absorb

heat even when this body seems to us to be isolated in space. This view is only
compatible with the definition we have just given of the quantity of heat released by
a system if we regard the space which seems empty to us as filled by a certain [320]
body, the ether. This observation helps in grasping the importance of the preceding
theorem.
Put
⎧
⎪ ∂U ∂T d ∂T
⎪
⎪ E − + − A = ERα ,
⎪
⎪ ∂α ∂α dt ∂u
⎨
.............................., (12)
⎪
⎪
⎪
⎪ ∂U ∂T d ∂T
⎪
⎩E − + − L = ERλ ,
∂λ ∂λ dt ∂w
⎧
⎪ ∂U
⎪
⎪ E − A = ERa ,
⎪
⎨ ∂a
⎪
.................., (12a)
⎪
⎪
⎪
⎪ ∂U
⎪
⎩E − L = ERl .
∂l
We call the quantities Rα , . . . , Rλ , Ra , . . . , Rl calorific coefficients of the system

S, subject to the action of system S with its motion at the instant t. We see that these
coefficients depend on
1. The properties of the system S;

2. The velocities and accelerations of the various points of the system;
3. The actions of the system S on S.
5.3 On the Quantity of Heat 75
We envisage a virtual change δα, . . . , δλ, δa, . . . , δl, of the system S. By defini-
tion, the quantity
dQ = −(Rα δα + . . . + Rλ δλ + Rα δa + . . . + Ri δl) (13)
is called the quantity of heat released by the system S in the virtual change under
consideration.
In can be seen that this quantity is not, in general, the total differential of a func-
tion of α, . . . , λ, a, . . . , l, because the coefficients of δα, . . . , δλ, δa, . . . , δl depend,
in general, on other variables.
We know that the quantity3
dW = Aδα + . . . + Lδλ + Aδα + . . . + Lδl (14)
represents the virtual work of the actions exerted on the system S. [321] We will call
the quantity

dτ = − δα + . . . + − δλ (15)
∂α dt ∂u ∂λ dt ∂w
the virtual work of the inertial forces applied to the system S.

Equations (12), (12a), (13), (14) and (15) then yield the following equation,
applicable to all virtual transformations of the system S,
E (dQ + dU) = dW + dτ . (16)
Multiplying the sum of the quantity of heat released by a system during a virtual change and
the variation undergone by the internal energy during the same change by the mechanical
equivalent of heat yields a certain product; this product is equal to the virtual work of the
exterior actions and the inertial forces applied to the system.
This proposition constitutes the most general statement of the law of equivalence
of heat and work.
When a real change, rather than any virtual change, is at issue, Equation (16) can
be modified. In this case, in fact, the work dW of the inertial forces becomes equal
to the variation of the kinetic energy, with sign changed, and we can write
d
dW − E dQ = (EU + T) dt. (17)
dt
It can be seen that, for any real change, the quantity
dW − E dQ
is a total differential.
3 [For typographical reasons, Duhem’s original symbol has been replaced throughout by “W”.]

Consider a system systems S1 , S2 ; let σ be
formed of two independent partial
the system formed by the collection of bodies outside .
Let [322]
U1 (α1 , . . . , λ1 , a1 , . . . , l1 ), U2 (α2 , . . . , λ2 , a2 , . . . , l2 )
be the internal energies

of the systems S1 , S2 considered in isolation.
The system will have as its internal energy the quantity
ϒ = U1 + U2 + 12 (α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 ).
Let u be the internal energy of the system σ considered

in isolation. The
internal energy of the system formed by the aggregate σ will be given by an
expression
ϒ + u + X,
where X depends on the variables defining the position and the properties of each of
the three systems S1 , S2 , σ.
Suppose that the system undergoes a virtual change
δα1 , . . . , δλ1 , δa1 , . . . , δl1 , δα2 , . . . , δλ2 , δa2 , . . . , δl2 ,

and let us find the expression for the quantity of heat dQ released by the system .
We have

∂ϒ ∂X ∂T d ∂T
E dQ = − E +E − + δα1 + . . .
∂α1 ∂α1 ∂α1 dt ∂u1

∂ϒ ∂X ∂T d ∂T
+ E +E − + δλ1
∂λ1 ∂λ1 ∂λ1 dt ∂w1

∂ϒ ∂X ∂ϒ ∂X
+ E +E δa1 + . . . + E +E δl1
∂a1 ∂a1 ∂l1 ∂l1

∂ϒ ∂X ∂T d ∂T
+ E +E − + δα2 + . . .
∂α2 ∂α2 ∂α2 dt ∂u2

∂ϒ ∂X ∂T d ∂T
+ E +E − + δλ2
∂λ2 ∂λ2 ∂λ2 dt ∂w2

∂ϒ ∂X ∂ϒ ∂X
+ E +E δa2 + . . . + E +E δl2 .
∂a2 ∂a2 ∂l2 ∂l2
5.4 The Classical Problem of Dynamics 77

But [323] the kinetic energy T of the system is the sum of the kinetic energies
T1 , T2 of the systems S1 , S2 . It is then easily seen that the preceding equation can
be written4

∂ ∂T d ∂T
E dQ = − E (U1 + 12 + X) − + δα1 + . . .
∂α1 ∂α1 dt ∂u1

∂ ∂T d ∂T
+ E (U1 + 12 + X) − + δλ1
∂λ1 ∂λ1 dt ∂w1
∂ ∂
+E (U1 + 12 + X) δa1 + . . . + E (U1 + 12 + X) δl1
∂a1 ∂l1

∂ ∂T d ∂T
+ E (U2 + 12 + X) − + δα2 + . . .
∂α2 ∂α2 dt ∂u2

∂ ∂T d ∂T
+ E (U2 + 12 + X) − + δλ2
∂λ2 ∂λ2 dt ∂w2

∂ ∂
+E (U2 + 12 + X) δa2 + . . . + E (U2 + 12 + X) δl2
∂a2 ∂l2
= E (dQ1 + dQ2 )
where dQ1 , dQ2 designate the quantities of heat released by the two systems S1 , S2
during the same virtual change. This leads us to the following theorem:
When a system is formed from several independent parts, the quantity of heat which it
releases during any virtual change is equal to the algebraic sum of the quantities of heat
that the various parts release during the same change.
This theorem will prove useful in what follows.

It would be natural to include here a reflection similar to that which suggested
the definition of work: it is not possible to speak of the quantity of heat released by
each part of a system except in so far as each of these parts can be considered as an
independent system. When the various parts of a system are not independent of one
another, the phrase ‘quantity of heat released by each of them’ has no meaning.
5.4 The Classical Problem of Dynamics

Suppose that, for a certain system, the calorific coefficients
Rα , . . . , Rλ , Rα , . . . , Rl
4 [A – sign at the beginning of the 5th line has been changed to +, and a bracket “}” at the end of
the immediately preceding line removed.]
are all identically equal to 0, or, in other words, that the quantity of heat released by
the system in any real or virtual change is identically equal to 0. Equations (12) and
(12a) become
⎧
⎪ ∂U ∂T d ∂T
⎪
⎪E − + − A = 0,
⎪
⎪ ∂α ∂α dt ∂u
⎪
⎨
..........................., (18)
⎪
⎪
⎪ ∂U
⎪ ∂T d ∂T
⎪
⎩ E ∂λ − ∂λ + dt ∂w − L = 0,
⎪
⎧
⎪ ∂U
⎪
⎪ E − A = 0,
⎪
⎨ ∂a
⎪
..............., (18a)
⎪
⎪
⎪
⎪ ∂U
⎪
⎩E − L = 0.
∂l
The equations of motion of a system in which friction is zero are easily recognisable.
In the case usually studied in mechanics, there are no variables other than those
which figure in Equations (1) of Chapter 3. There is therefore no equation of type
(18a). All the equations which govern the motion of the system have the form (18),
given, as is known, by Lagrange.
Clearly, the laws of dynamics fall within the laws of thermodynamics as par-
ticular cases. They can be deduced from the latter by assuming all the calorific
coefficients of the system are equal to 0. But in which case is this hypothesis ver-
ified? This is a question which remains to be examined, and nothing of what has
so far been said resolves the issue. In most cases, it is only solved by way of
hypothesis, directly or indirectly. Moreover, we will see later that there is another
method, distinct from this, of deriving the equations of dynamics from the equations
of thermodynamics.
5.5 Calorimetry
Imagine [325] an isolated system itself formed from three independent systems S1 ,
S2 , S3 , on which we will impose certain conditions.
Let U1 , U2 , U3 be the internal energies of the systems S1 , S2 , S3 considered in
isolation. The internal energy of the complex system formed by their composition
can be written
ϒ = U1 + U2 + U3 + .
1. Regarding the function , we assume that it has the form
= 23 + 31 + 12 ,

5.5 Calorimetry 79
the function ij depending only on the variables which characterise the state of
the two systems Si , Sj . This hypothesis entails the following consequence: The
actions on any of the three systems S1 , S2 , S3 by the composite of the remaining
two are obtained by adding the actions to which it is subject from each of the
other two taken in isolation.
2. The actions of system S3 on system S1 are zero, or are such that they do no work
in the changes of the system S1 that will be studied.
3. The actions of system S2 on system S1 are zero, and the same holds for the
actions of system S1 on system S2 .
The following consequences arise from these two hypotheses:
In the changes of the system S1 to be studied, the work of the external actions
applying to the system is always equal to 0.
The external actions applying to the system S2 are just the actions exerted by
system S3 .
4. The quantity of heat released or absorbed by the system S3 is constant and equal
to 0.
Since the quantity of heat released by the collection S1 , S2 , S3 , which forms
an isolated system, must be equal to 0, and since, on the other hand, this quantity
of heat must be equal to the sum of the quantities of heat released by each of
the three systems S1 , S2 , S3 , the quantity of heat dQ2 released during the time dt
by the system S2 will be equal and of opposite sign to the quantity of heat dQ1
released [326] by the system S1 during the same time,
dQ1 + dQ2 = 0. (19)
Regarding the quantity dQ1 , we can write, in virtue of Equation (10) and noting
that, in accordance with our hypotheses,
(A1 u1 + . . . + L1 w1 + A1 ϕ1 + . . . + L1 ψ1 )dt = 0,
the following equation

∂U1 d ∂T ∂T ∂U1 d ∂T ∂T
E + − u1 + . . . + E + − w1
∂α1 dt ∂u1 ∂α1 ∂λ1 dt ∂w1 ∂λ1

∂U1 ∂U1
+E ϕ1 + . . . + E ψ1 dt = −E dQ1 .
∂a1 ∂l1
In virtue of Equation (19), this becomes

∂U1 d ∂T ∂T
E dQ2 = E + − u1 + . . .
∂α1 dt ∂u1 ∂α1

∂U1 d ∂T ∂T ∂U1 ∂U1
+ E + − w1 + E ϕ1 + . . . + E ψ1 dt .
∂λ1 dt ∂w1 ∂λ1 ∂a1 ∂l1
This equation can be integrated immediately, yielding the following expression

for the quantity of heat Q2 released by the system S2 during any finite change
EQ2 = EU1 + T1 − EU1 − T1 , (20)
U1 , T1 represent the values of U1 , T1 at the initial instant of the change, and
U1 , T1 represent the values of U1 , T1 at the final instant of the same change.
The [327] equation of quantities of heat released by different changes produced
within different systems can already be seen from Equation (20).
Suppose, in fact, that we retain system S1 but replace the aggregate (S2 , S3 )
by various aggregates (S 2 , S 3 ), (S 2 , S 3 ), . . . while the hypotheses continue to
hold. If, during various changes of these various systems, the system S1 always
starts from the same initial state with the same initial motion and ends in the
same final state with the same final motion, the quantities of heat Q2 , Q2 , Q2 , . . . ,
released in these various changes by the systems S2 , S 2 , S 2 , . . . , subject,
respectively, to the actions of S1 , S 1 , S 1 , . . . are equal to one another.
We now impose new restrictions on the system S1 :
5. System S1 is at rest at the beginning and the end of each of these changes which
we are studying, or, if it is in motion, its absolute motion is the same in both
cases.
6. The state of system S1 is fixed by knowledge of a single variable α1 .
7. In the changes being studied, the final value of this variable α 1 differs little from
its initial value α 1 .
The fifth hypothesis gives us
T1 − T1 = 0.
The sixth and seventh allow us to write

U1 − U1 = 1 (α1 − α1 ),
where 1 depends only on the initial state of the system S1 .

Equation (20) then becomes
Q2 = 1 (α1 − α1 ).
Therefore, if we take care to ensure that the system S1 is in the same

initial state, and if, in the various changes of the aggregates (S1 , S2 ), (S 1 , S 2 ),
(S 1 , S 2 ), . . . , we observe the variations [328] (α1 − α1 ), (α1 − α1 ) , (α1 − α1 ) ,
. . . undergone by the variable defining the state of the system S1 , then we can deter-
mine the relative values of the quantities of heat Q2 , Q 2 , Q 2 , . . . , released by the
systems S2 , S 2 , S 2 , . . . in these various changes.
The system S1 just defined is called a calorimeter. The calorimeters used in prac-
tice only approximate to the ideal type of calorimeter. By various corrections, based
either on direct hypotheses or on the consequences of various physical theories, dis-
crepancies between their readings and those of the ideal calorimeter can be reduced.
We leave it to the reader to consider by what set of ideas, in any particular case, a
5.5 Calorimetry 81
real calorimeter can be seen to reasonably verify the seven hypotheses that we have
enumerated.5
The calorimeter determines the relation between two quantities of heat released
by two different systems in two different circumstances. It therefore allows us to
determine the value of a quantity of heat, measured absolutely, if the quantity taken
as unity is known.
The definition given by Equations (10), (12), (12a) and (13) of the quantity of
heat released during a real or virtual change of a system shows that a quantity of
heat is a magnitude of the same kind as one of mechanical work.6 The unit of the
quantity of heat is therefore determined when the unit of mechanical work is known.
Indeed, from Equation (10) it is easy to arrive at the following proposition:
If the mechanical work done during any change of the system by the bodies external to the
system is equal to a unit of mechanical work, and if, during this change, the actions exerted
on the system by the external bodies do not perform any work, the system [329] absorbs a
quantity of heat equal to unity during this change.
We have not as yet established any convention for the unit of mechanical work.
Only a convention for the sign of mechanical work has been established. We have
settled on this choice of sign so that the mechanical work done by putting a system
into motion without changing its state is positive. We can therefore take any amount
of mechanical work as the unit of mechanical work provided that it is positive.
We are going to prove that the work done by raising the temperature of a unit
mass of water (whose state we suppose to be uniquely defined by the temperature)
from a determinate initial temperature that we call the 0 of the centigrade scale
to another determinate temperature that we call the +1 on the centigrade scale is
positive. We can then take this mechanical work as the unit of mechanical work. The
unit of the quantity of heat will be the quantity of heat absorbed by the unit of mass
of water when, without external work, its temperature is raised from 0◦ C to +1◦ C.
To demonstrate this proposition, suppose that we have a complex and isolated
system, itself formed of two independent systems S and S . The system S is immo-
bile and its state is assumed to be uniquely defined by its temperature. System S
is a mobile body whose state is assumed invariable. These two systems S and S
exert no action on one another. If the internal energy of system S, assumed iso-
lated and with the temperature ϑ, is designated by U(ϑ), and the kinetic energy
of system S by T, then the total energy of our complex system will have the very
simple form
T
ε = U(ϑ) + .
E
5 In practice, the system S is always tied to the earth and doesn’t, rigorously speaking, have the
1
same absolute movement at the start as at the end of each change. But, given the accepted hypothe-
ses regarding the absolute movement of the earth, the absolute movement of the calorimeter in the
course of an operation undergoes variations which exert only a negligible influence on the results
of experimental determinations.
6 oeuvre.
At the beginning of the change the system S has temperature ϑ. System S is

in motion, and its kinetic energy has the value T. System S collides with system
S and rebounds. After the impact, it has a kinetic energy T and system S has a
temperature ϑ . Since the system [330] is assumed to be isolated, the total energy
has not changed. It therefore follows that
T T
U(ϑ) + = U(ϑ ) +
E E
or
T − T
= U(ϑ) − U(ϑ ) (21)
E
Experiment shows that T is less than T; therefore

U(ϑ ) − U(ϑ)
is positive.
When passing from the temperature ϑ to the temperature ϑ without external
work and without variation in the kinetic energy, the system S has absorbed a
positive quantity of heat given by the equation
Q = U(ϑ ) − U(ϑ). (22)
We therefore know of a change which absorbs a positive quantity of heat. The

calorimeter which gives us a magnitude and a sign for the ratio of the quantities
of heat released in two changes therefore allows us to determine the sign of any
quantities of heat and to prove experimentally the proposition formulated above.
Now that the unit of the quantity of heat has been determined, the calorimeter
allows us to measure any quantity of heat, in particular the quantity of heat Q given
by Equation (22). If, on the other hand, the variation in the kinetic energy (T – T )
is measured, an experimental determination of the mechanical equivalent of heat E
can be deduced from Equation (21).
G.-A. Hirn has performed a real experiment closely related to the ideal experi-
ment that we have just described.
Other methods have been employed, notably by Joule, to determine the value of
E. The principles put forward here easily lead to the justification of these methods.
Part III
of Thermodynamics: The Principle
of Sadi Carnot and R. Clausius (1893)∗
∗ “Commentaire aux principes de la Thermodynamique. Deuxième Partie: Le principe de Sadi
Carnot et de R. Clausius”, Journal de Mathématiques Pure et Appliquées, 9 (1893), 293–359.

See vol. VIII, p. 269 [i.e. Part One.]
Chapter 6
The Carnot Cycle and Reversible Changes
6.1 Virtual Changes

We first recall the notion of virtual change with a view to making it more precise
than was done in the course of the first part.
Consider a system found in a given state, as well as the bodies external to this
system. The state of this system, its motion and the external actions acting on it are
known if we know the values of the parameters
α, β, . . . , λ, a, b, . . . , l
and of the quantities

dα dβ dλ
, ,..., .
dt dt dt
In [294] order to know the inertial forces acting on the system and the calorific
coefficients of the system, it is also necessary to know in addition the values of each
of the quantities
d2 α d2 β d2λ
2
, 2
,..., 2 .
dt dt dt
When change of a system is no longer real but virtual, the ordering of the various
states of the system exists only in our understanding, and not in time. The quantities
α, β, . . . , λ cannot be considered as functions of time. It is not, therefore, possible
to speak of the quantities
dα dβ dλ
, ,..., ,
dt dt dt
d2 α d2 β d2λ
2
, 2
,..., 2 .
dt dt dt
It would thus seem, then, that it is not possible to speak of inertial forces or
calorific coefficients for a system which undergoes a virtual change.

86 6 The Carnot Cycle and Reversible Changes
In fact, these words do preserve a sense.

In the formulas which define the inertial forces and the calorific coefficients of a
system for a real transformation, we replace the quantities
dα dβ dλ
, ,..., ,
dt dt dt
d2 α d2 β d2 λ
, , . . . ,
dt2 dt2 dt2
by any magnitudes
u, v, . . . , w,
u , v , . . . , w ,
subject only to restrictions that the same [295] definition of the system might impose
on the quantities
dα dβ dλ
, ,..., ,
dt dt dt
d2 α d2 β d2λ
2
, 2
,..., 2 .
dt dt dt
We will have new expressions representing, by definition, the inertial forces and the
calorific coefficients of the system during a virtual change. We can then speak of
the work effected by the inertial forces and of the quantity of heat released by the
system in the course of a virtual change.
The quantities u, v, . . . , w must vary continuously in the course of a vir-
tual transformation; on the other hand, the quantities u , v , . . . , w may display
discontinuities.
6.2 Cycles
Recall the definition of a closed cycle, already given in the first part of this work
(Section 4.1).1
Consider that a system starts from an initial state defined by the values
α0 , β0 , . . . , λ0 , a0 , b0 , . . . , l0
of the parameters
1 [The definition is given in Chapter 3, towards the end of Section 3.4.]

6.3 A Real Cycle Can Reproduce Itself Indefinitely 87
α, β, . . . , λ, a, b, . . . , l,
and that its initial motion is defined by the values

u0 , v 0 , . . . , w 0
of the velocities
u, v, . . . , w.
This system undergoes a series of real or virtual changes during which the
parameters
α, β, . . . , λ, a, b, . . . , l
and the velocities

u, v, . . . , w
vary continuously.
It [296] reaches a state in which the variable quantities
α, β, . . . , λ, a, b, . . . , l, u, v, . . . , w
have the values

α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , u1 , v1 , . . . , w1 .
If the quantities
α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , u1 , v1 , . . . , w1
are equal, respectively, to the quantities

α0 , β0 , . . . , λ0 , a0 , b0 , . . . , l0 , u0 , v0 , . . . , w0 ,
then the system is said to have described a closed cycle, or simply a cycle.
If all the changes comprising the cycle are real changes, the cycle itself is real.
If all the changes composing the cycle, or just a part of them, are virtual, then the
cycle itself is virtual.
6.3 A Real Cycle Can Reproduce Itself Indefinitely

This is now the place to state a hypothesis which plays a fundamental role in the
constitution of thermodynamics:

HYPOTHESIS.—Let S be a system and let be the collection of bodies external to this sys-
tem. Consider two equal temporal intervals, one comprising the instants between t0 and t1 ,
and the other comprising those between the instants t0 and t1 . Suppose that the following
conditions are satisfied:

1. At any two corresponding instants of the intervals (t0 , t1 ) and (t0 , t1 ), the system is in
the same state.
2. At the instants t0 and t0 , the system S is in the same state and moves at the same
velocities.
We accept that, under these conditions, at any two corresponding instants of the intervals
(t0 , t1 ) and (t0 , t1 ), the system S is in the same state.
This [297] hypothesis can be stated in what is, perhaps, a less precise form, but
more briefly, by saying that:
The change undergone by the system S in the temporal interval (t0 , t1 ) is determined if:
1. the state of the external bodies at all instants of the interval (t0 , t1 ), and 2. the state
and the velocities of the system S at the instant t0 are known.

We add that at the instant t0 it is only possible to define the state of the systems
and S close to the variables which fix the absolute position in space of the ensemble
of the two systems.
Let us consider a real closed cycle, described
by the system S. During the course
of this closed cycle, the external bodies have passed through a succession of
states. At the end of this cycle, the system S has returned to the same state with
the same initial
velocities. When this cycle has been completed, we see to it that
the bodies once more follow the succession of states through which they passed
during the course of the cycle. In accordance with the preceding hypothesis, the
system S will describe the cycle once more.
We may therefore state the following proposition:
A real closed cycle can be exactly reproduced indefinitely PROVIDED THAT IT IS
POSSIBLE TO ARBITRARILY ARRANGE THE BODIES EXTERNAL TO THE
SYSTEM.
6.4 Adiabatic, Exothermic and Endothermic Changes

Consider a system subject to an infinitesimal change, real or virtual, under the influ-
ence of certain other systems. As an effect of this transformation, the system releases
a quantity of heat dQ, and the inertial forces (Section 5.3) perform work dτ . If the
quantity
1
dQ = dQ − dτ (1)
E
is equal to 0, the infinitesimal change under consideration is said to be
adiabatic.
A finite change is said to be adiabatic if all the infinitesimal changes which
compose it are adiabatic.
6.4 Adiabatic, Exothermic and Endothermic Changes 89
Let [298]
τ
Q=Q− , (2)
E
where Q is the quantity of heat released in a change and τ is the work of the inertial
forces during the change.
We call Q2 the total calorific effect of the change.
It is easily seen that the total calorific effect of an adiabatic change is always equal
to 0. The converse of this proposition does not hold. The total calorific effect of a
finite change may be equal to 0 without the calorific effect of each of the elementary
changes composing this finite change being equal to 0, and consequently, without
the finite change being adiabatic.
A change is said to be exothermic, athermic or endothermic depending on
whether the total calorific effect of the change is positive, zero or negative.
What we have just said applies as well to virtual changes as to real changes. For
an infinitesimal real change we have
dτ = −dT,
where T is the kinetic energy of the system. The total calorific effect of a finite and
real change is therefore defined by the equation
1
Q=Q+ (T1 − T0 ), (2a)
E
where T0 is the value of the kinetic energy of the system at the beginning of the
change and T1 the value of the same kinetic energy at the end of the change.
Consider a cycle.
The calorific effect of each of the elementary changes constituting the cycle is
defined by the equation
1
dQ = dQ − dτ ,
E
which [299] can also be written (Chapter 5, Equation (16))
E dQ = dW − E dU.
Integrating this equation for the whole cycle, and noting that the internal energy
U of the system returns to the same value at the end of the cycle that it had at the
beginning, we find
E Q = W. (3)
2 [Q in the original.]
The total calorific effect of a closed cycle, real or virtual, is equivalent to the total work
done during the course of the cycle by all the external actions acting on the system.
This theorem will be useful to us in the following chapter.
6.5 Isothermal Changes: The Carnot Cycle

One can well imagine a system which has, in each of the states it passes through in
the course of a change, the same temperature ϑ, read on any thermometer, at all of
its points. This temperature can, moreover, vary from one of the states traversed in
this change to the following state.
When the temperature is the same not only at all the points of the system taken
in each of the states constituting the real or virtual change, but also in all the states
of this sequence, the change is said to be isothermal.
Consider a cycle subject to the following conditions:
1. The cycle is comprised exclusively of adiabatic and isothermal changes;

2. The isothermal changes figuring in the cycle occur at two different temperatures,
ϑ and ϑ , where ϑ is higher than ϑ.
Such a cycle is called a Carnot cycle described between the temperatures ϑ

and ϑ .
A Carnot cycle can, in accordance with this definition, be real or virtual.
If [300] the same system successively describes several Carnot cycles, identical
or not, but between the same temperatures ϑ and ϑ , it is evident from the preceding
definition that the collection of cycles can be regarded as a single Carnot cycle
described between the temperatures ϑ and ϑ .
Let these successive Carnot cycles be C1 , C2 , . . . , Cn , and their ensemble be the
Carnot cycle .
Let Q1 , Q2 , . . . , Qn be the total calorific effect of the changes produced at
the temperature ϑ in each of these cycles C1 , C2 , . . . , Cn , and let χ be
the total calorific effect of the changes produced at the temperature ϑ in the
cycle .
Similarly, let Q1 , Q2 , . . . , Qn be the total calorific effect of the changes pro-
duced at the temperature ϑ in each of these cycles C1 , C2 , . . . , Cn , and let χ
be the total calorific effect of the changes produced at the temperature ϑ in the
cycle .
Clearly we have
⎧
⎨ χ = Q1 + Q2 + · · · + Qn
(4)
⎩ χ = Q + Q + · · · + Q .
1 2 n
6.6 Simultaneous Independent Changes: Generalisation of the Carnot Cycle 91
6.6 Simultaneous Independent Changes: Generalisation

of the Carnot Cycle

Let σ be a complex system formed from a certain system S and by the bodies
external to this system. The system σ is isolated in space. During the time between
the instants t0 , t1 , the system S undergoes a certain change M.

let σ be a complex system formed from a certain system S and the
Similarly,
bodies external to this system. The system σ is isolated in space. During the
time between the instants t0 , t1 , the system S undergoes a certain change M .
Suppose that
t1 − t0 = t1 − t0 .
Imagine now that at any instant τ 0 , the system σ is placed in space in the state
and with the velocities that it exhibited at the instant t0 , and the system σ in the
state and with the velocities that it exhibited [301] at the instant t0 ,3 the two systems
σ and σ being infinitely far apart.
We accept the following hypothesis:
HYPOTHESIS.—Each of the two systems σ, σ will be changed as if it was isolated in space.
If this hypothesis is combined with the hypothesis stated in §5, is it easily seen
that, in the interval of time
τ1 − τ0 = t1 − t0 = t1 − t0 ,
the system S will undergoe precisely the change M and the system S the change M .
We will then say that the changes M and M are effected simultaneously and
independently.
It is easily seen, in virtue of the principles put forward in the first part of this
work, that the quantity of heat released by the complex system (S, S ) during the
interval of time (τ 0 , τ 1 ) is the sum of the quantity of heat released by the system
S undergoing the change M in the absence of the system σ , and the quantity of heat
released by the system S undergoing the change M in the absence of the system σ.
Analogous propositions can be stated concerning the work done
bythe actions that
the system (S, S ) undergoes on the part of the external bodies , , the work of
the inertial forces applied to the system (S, S ), and finally, the total calorific effect
of the two changes M, M taken together.
If the two closed cycles C, C1 are described simultaneously and independently
by the systems S and S , the system (S, S ) clearly describes a closed cycle.
Suppose that the two cycles C and C1 are two Carnot cycles described between
the same temperatures ϑ and ϑ . If the isotherm described by the system S at the
temperature ϑ and the isotherm described by the system S1 at the same temperature
ϑ are simultaneous, and similarly, if the isotherms described by the systems S and
S1 at the temperature ϑ are simultaneous, the system (S, S1 ) will also describe a
3 [t in the original.]
0
Carnot cycle [302] between the temperatures ϑ and ϑ . Otherwise, the closed cycle
described by the system (S, S1 ) will not, in general, be a Carnot cycle.
Nevertheless, by an extension of the term Carnot cycle, it will be convenient to
say that a complex system (S, S1 ) describes a Carnot cycle between the tempera-
tures ϑ and ϑ if the two infinitely separated systems S, S1 composing it describe two
simultaneous and independent Carnot cycles C, C1 between the same temperatures
ϑ, ϑ .
Let Q, Q1 be the values of the total calorific effect of the changes described at the
temperature ϑ in each of the two cycles C, C1 . Similarly, let Q , Q1 be the values of
the total calorific effect of the changes described at the temperature ϑ in each of the
two cycles C, C1 .
By definition, we say that the quantity
χ = Q + Q1 (5)
is the total calorific effect of the changes produced at the temperature ϑ in the cycle
, and the quantity
χ = Q + Q1 (5a)
is the total calorific effect of the changes produced at the temperature ϑ in the
cycle .
Of course, the term ‘Carnot cycle described between the temperatures ϑ, ϑ ’4 can
also be extended to n simultaneous, independent Carnot cycles described between
the temperatures ϑ, ϑ .
6.7 Changes Which Are a Sequence of Equilibrium States

When a system is at equilibrium in a certain state, it persists indefinitely in that state
and is not transformed. It therefore seems that it is not possible to speak without
contradiction of a real change constituted by a sequence of equilibrium states. But
it is, in fact, possible to give these words a logical meaning.
A state of a system is defined, it will be recalled, from a knowledge of [303] the
quantities
α, β, . . . , λ, a, b, . . . , l
dα dβ dλ
, ,..., .
dt dt dt
Consider a system that undergoes a real change. At an instant t during this
transformation, the system is in a well-defined state. The parameters analogous to
α, β, . . . , λ, a, b, . . . , l which define the properties of the external bodies acting on
4 [Duhem doesn’t distinguish use and mention.]

6.7 Changes Which Are a Sequence of Equilibrium States 93
the system also have well-defined values. Imagine that, beginning at the instant t,
the latter variables are maintained at the same values, so that the properties of the
bodies external to the system remain indefinitely as they were at the instant t. If,
following this operation, the system also persists in the state that it exhibited at the
instant t, we say that at this instant t it was at equilibrium under the action of the
external bodies in the presence of which it was found.
Now clearly it would be impossible that the values taken by the quantities
α, β, . . . , λ, a, b, . . . , l
dα dβ dλ
, ,..., ,
dt dt dt
at the instant t still agree at all later instants unless
dα dβ dλ
= 0, = 0, . . . , = 0.
dt dt dt
Thus, in order that a certain state of a system undergoing a real transformation
can be said to be a state of equilibrium of the system, it is necessary, but not suf-
ficient, that the velocities of the various points of the system are all zero in this
state.
From this proposition we can conclude another:
Consider a system undergoing a real change. Might it happen that the state exhib-
ited by the system at each instant can be regarded as susceptible to becoming an
equilibrium state of the system under the action of the external bodies in whose
presence it is found, if [304] the latter preserve indefinitely the properties that they
possess at that instant? For that, it is necessary, but not sufficient, that during the
entire duration of the change,
dα dβ dλ
= 0, = 0, . . . , =0
dt dt dt
or, in other words,
α = const., β = const., . . . , λ = const.
This proposition can be briefly stated in the following way:

In order that a real change be a succession of equilibrium states, it is necessary, but not
sufficient, that all the points of the system remain at the same position in space during the
entire course of the change.
Now is it absurd to accept the existence of a change during which all the points
of the system remain at the same position? Clearly not; in physics we are sometimes
led to conceive of such changes. Take, for example, a container holding a mixture
of hydrogen and chlorine. The combination occurs, and a transformation, a change
of state takes place. It might well be held, however, that the matter filling each of
the volume elements of the container at the beginning of the combination remains
in the same volume element during the course of the change.
The reservations that we have just made only apply to real changes. In the course
of a virtual change, the values of α, β, . . . , λ cannot be regarded as functions of time,
in which case the preceding argument no longer applies. It is certainly true that if a
virtual change is a sequence of equilibrium states, we should have, during the entire
change,
u = 0, v = 0, . . . , w = 0;
but [305] since we do not have
dα dβ dλ
u= ,v = ,...,w = ,
dt dt dt
these equalities do not prevent the quantities α, β, . . . , λ from changing value in the
course of such a change.
Moreover, it is clear that any sequence of equilibrium states of a system, provided
it is continuous, can always be envisaged as forming a virtual change of the system.
6.8 Reversible Changes

We now come to one of the most important notions which, at the same time, poses
one of the most subtle problems of definition in thermodynamics: the notion of a
reversible transformation.
Consider, for the same system, two real or virtual transformations S and5 S1
endowed with the following properties:
1. Each state E of the transformation S corresponds to a single state E1 of the

transformation S1 .
2. During the transformations S and S1 , the system traverses corresponding states in
the same order; in particular, the initial state of transformation S corresponds to
the initial state of transformation S1 , and the final state of change S corresponds
to the final state of transformation S1 .
3. To two states E, E , infinitesimally close in the change S, there correspond two
states E1 , E1 infinitesimally close in the change S1 .
4. Two corresponding states E, E1 exhibit the following properties:
a. The parameters α, β, . . . , λ, a, b, . . . , l determining the properties of the

system in the state S are infinitesimally close to the parameters
5 [In the original, the symbol following is S1 .]

6.8 Reversible Changes 95
α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 determining the properties of the system in the

state E1 .
dβ
b. The quantities dα dt , dt , . . . , dt (or u, v, . . . , w) determining the real (or virtual)
dλ
velocities of the system in the state E differ infinitesimally from the quantities
dα1 dβ1 dλ1
dt , dt , . . . , dt (or u1 , v1 , . . . , w1 ) [306] determining the real (or virtual)
velocities of the system in the state E1 .
c. The quantities ddtα2 , ddtβ2 , . . . , ddt2λ (or u , v , . . . , w ) determining the accelera-
2 2 2
tions (real or virtual) of the system in the state E differ infinitesimally from
the quantities ddtα21 , ddtβ21 , . . . , ddtλ21 (or u1 , v1 , . . . , w1 )6 determining the real
2 2 2
(or virtual) accelerations of the system in the state E1 .

d. The parameters analogous to α, β, . . . , λ, a, b, . . . , l determining the proper-
ties of the external bodies acting on the system while it is in state E differ
infinitesimally from the parameters analogous to α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1
determining the properties of the external bodies acting on the system while
it is in state E1 .
From these conditions on the two states E, E1 , it follows that the kinetic energy,
the external actions, the inertial forces and the calorific coefficients have values
differing infinitesimally for the system whether in state E or state E1 .
The two transformations S and S1 whose properties we have just fixed constitute
two infinitesimally close transformations. We call a sequence of transformations
each of which is infinitesimally close to that which precedes it and that which
follows it a continuously variable transformation.
Let be a real or virtual transformation of a system possessing the following
properties:
1. In each of the states constituting it, the system is in equilibrium under the action
of the external bodies in whose presence it finds itself.
2. If the change is virtual, the quantities u, v, . . . , w and u , v , . . . , w are assumed
to be zero throughout.
3. Let (A) be the initial state and (B) the final state of the transformation ; it
is possible to pass from (A) to (B) by a continuously variable transformation,
realisable under each of its forms, and having the transformation , real or
virtual, as its limiting form.
4. It is possible to pass from (B) to (A) by a continuously variable transformation
S , realisable under each of its forms, [307] and having as its limiting form the
real
or virtual transformation obtained by running through the equilibrium states
in reverse order from (B) to (A).

Such a change is called a reversible transformation.
Is a reversible transformation realisable? Since a reversible transformation is a
sequence of equilibrium states, we know that it will be impossible to realise it,
6 [In the original, (ou u1 , v1 , . . . , w1 ).]

except perhaps in the particular case specified in the preceding section. Thus, in
general, a reversible transformation is an entirely virtual change.
The question might still be raised of whether the virtual changes which are
reversible are possible. We respond to this question by accepting the following
hypothesis:
FUNDAMENTAL HYPOTHESIS.—There are systems for which all changes, real or virtual,
which are series of equilibrium states, are reversible changes.
WE WILL ONLY STUDY SYSTEMS WHICH POSSESS THIS PROPERTY.
It shouldn’t be thought that these systems alone exist in nature. It is easy to prove
the contrary; cycles of magnetic hysteresis, for example, are irreversible changes
although they are continuous sequences of equilibrium states.
We will establish some properties of reversible changes that will be of use in
what follows.
If a reversible change is real, then, since it is a sequence of equilibrium states,
dα dβ dλ
= 0, = 0, . . . , =0
dt dt dt
during the entire duration of that change, and so also
d2 α d2 β d2 λ
2
= 0, 2
= 0, . . . , 2 = 0 .
dt dt dt
If [308] a reversible change is virtual, then by definition, at each instant,
u = 0, v = 0, . . . , w = 0,
u = 0, v = 0, . . . , w = 0.
Thus, in a reversible change, whether real or virtual, the kinetic energy and the
inertial forces are always zero.
Let A, B, . . . , L, A, B, . . . , L be the external actions to which the system
is sub-
ject while it is in one of the states constituting the reversible change . If the
system undergoes this change from the state (A) to the state (B), the external actions
perform work
(B)
= (Aδα + Bδβ + · · · + Lδλ + Aδa + Bδb + · · · + Lδl).
(A)
If the system undergoes the same change from (B) to (A), the external actions
perform work
(A)
= (Aδα + Bδβ + · · · + Lδλ + Aδa + Bδb + · · · + Lδl).
(B)
6.8 Reversible Changes 97

The two integrals are taken along the same path , whence clearly
+ = 0.
Now, in accordance with the definition of a continuously variable change, the

work W performed by the external actions while the system undergoes the real
state (B) has the limit when the change S tends
change S from the state (A) to the
towards the reversible change . The work W performed by the external actions
while the system undergoes the real change S from the state (B) to the
state (A) has
the limit when the change S tends towards the reversible change reversed.
We therefore arrive at the following proposition:
If S and S are two continuously variable real changes, [309] the one the inverse of the
other, which have as their common limit a certain reversible change , the work W and W
performed by the external actions while the system undergoes these changes tends towards

the same limit with opposite sign when the changes S and S tend towards the change .
It suffices to replace the external actions
A, B, . . . , L, A, B, . . . , L
in the preceding argument by the calorific coefficients
Rα , . . . , Rλ , Ra , . . . , Rl
and the words external work with the words quantity of heat released by the system
in order to obtain the following proposition:
The quantities of heat Q and Q released by the system while it undergoes the changes S and
S tend towards the same
limit with opposite sign when the changes S and S tend towards
the reversible change .
We complete these generalisations on reversible changes with the statement of

two hypotheses which will be of much use in what follows.

FIRST HYPOTHESIS.—If a reversible change is isothermal, it can be regarded as the
common limit of two continuously variable changes S and S , the one the inverse of the
other, which are always real and always isothermal.

S ECOND HYPOTHESIS.—If a reversible change is adiabatic, it can be regarded as
the common limit of two continuously variable changes S and S , the one the inverse of
the other, which are always real and always adiabatic.
Chapter 7
Carnot’s Theorem and Absolute Temperature
7.1 The Hypotheses of Clausius and Sir W. Thomson

Let A, B, . . . , L, A, B, . . . , L be the external forces and influences acting on a system
taken in one of its states within a series constituting a Carnot cycle. During one of
the elementary changes
dα, dβ, . . . , dλ, da, db, . . . , dl
into which the cycle can be decomposed, these actions perform work
dW = Adα + Bdβ + . . . + Ldλ + Ada + Bdb + . . . + Ldl,
and during the course of the whole cycle, they perform work
W = ∫(Adα + Bdβ + . . . + Ldλ + Ada + Bdb + . . . + Ldl),
the integral extending over the whole cycle.

This work W can be positive, zero or negative.
The Carnot cycles for which W is zero and the Carnot cycles for which W is
negative are the subject of two hypotheses, one due to Clausius and the other to Sir
W. Thomson.
CLAUSIUS ’ HYPOTHESIS.—If a system describes a REAL Carnot cycle between the tem-
peratures ϑ and ϑ (ϑ being higher than ϑ), and if the external actions to which the system
is subjected perform a total work equal to 0 during the course of the cycle, then it is not
possible that an isothermal change occurring at the temperature ϑ should be endothermic,
nor that an isothermal change occurring at the temperature ϑ should be exothermic.
SIR W. THOMSON ’ S HYPOTHESIS.—If a system describes a [311] REAL Carnot cycle

between the temperatures ϑ and ϑ (ϑ being higher than ϑ), and if the external actions
to which the system is subjected perform a total negative work during the course of the
cycle, then it is impossible that the isothermal change occurring at the temperature ϑ be
endothermic.

100 7 Carnot’s Theorem and Absolute Temperature
This is not the place to go into the way in which Clausius and Sir W. Thomson
were led to state their hypotheses, nor to review the discussions undertaken by
Clausius against the physicists who denied its correctness. In the classic treatises
of thermodynamics, these two hypotheses are normally presented as equivalent and
as equally capable of serving as the basis for Carnot’s demonstration. For our part,
we regard them as distinct and will employ both in our exposition.
7.2 Addition to the Hypotheses of Clausius

and Sir W. Thomson
To these two hypotheses we will make an essential addition that we state in the
following way:
ADDITIONAL HYPOTHESIS.—If a Carnot cycle described between the temperatures ϑ and
ϑ (ϑ being higher than ϑ) is REAL AND NOT REVERSIBLE and if in the course of the cycle
the external actions perform zero or negative work, then the isothermal change occurring
at the temperature ϑ cannot be athermic.
Combined with the hypotheses of Clausius and Sir W. Thomson, this additional
hypothesis yields the following proposition:
Among all the Carnot cycles described between the temperatures ϑ and ϑ (ϑ being higher
than ϑ) which are REAL AND NOT REVERSIBLE, consider those during the course of which
the external actions perform zero or negative work; for all these cycles, the isothermal
change occurring at the temperature ϑ is exothermic.
The hypotheses that we have just stated cannot be directly verified by experiment.
Combined with more or fewer other hypotheses, [312] they constitute the point of
departure of theories whose distant consequences can alone be subjected to experi-
mental control. An analogous remark applies, however, to nearly all the hypotheses
that are encountered in physics.
7.3 Various Kinds of Carnot Cycles

Take all the Carnot cycles that can be conceived and try to distinguish among
them all those whose properties are compatible, on the one hand, with the prin-
ciple of the conservation of energy, and on the other, with the three hypotheses just
mentioned.
Among all the Carnot cycles thus distinguished are surely all those which
are realisable and irreversible. The properties which belong to all the cycles
thus distinguished belong in particular to all the realisable and irreversible
cycles.
Unless otherwise stated, we exclude from our investigations real, irreversible Carnot cycles,
described between the temperatures ϑ and ϑ , corresponding to a positive work of the
external actions and for which the thermic effect produced at the temperature ϑ is equal
to 0.
7.3 Various Kinds of Carnot Cycles 101
Let ϑ and ϑ always be the temperatures between which Carnot cycles are
described, where ϑ is higher than ϑ. Let Q be the total calorific effect of the
isothermal change produced at the temperature ϑ, and Q the total calorific effect
of the isothermal change produced at the temperature ϑ . Apart from these two, the
changes comprising the cycle are adiabatic, so that the total calorific effect of the
cycle is reduced to (Q + Q ).
The external actions to which the system is subjected during the course of
the cycle perform the total work W.
The principle of the conservation of energy gives us (Chapter 6, Equation (3))
W = E(Q + Q ). (1)
From [313] that, we have the following properties:
1. If the work W is zero, the two quantities Q and Q are of opposite sign and have
the same absolute value.
2. If the work W is positive, at least one of the two quantities Q and Q is positive,
and if only one is, then it is the one of greatest absolute value.
3. If the work W is negative, at least one of the two quantities Q and Q is negative,
and if only one is, then it is the one of greatest absolute value.
These propositions result from the application of the principle of the conservation
of energy to the Carnot cycles considered.
We now apply to them the two hypotheses of Clausius and Sir W. Thomson
stated in section 1 and the additional hypothesis stated in section 2. We are led to
the following conclusion:
If the external work is zero or negative, the quantity Q is certainly positive; if
the external work is positive, the quantity Q is positive or negative, but certainly not
zero.
It is easy to conclude from these propositions that all realisable Carnot cycles
which are not just a series of equilibrium states fall within one of the categories in
the following classification:
1. The work W performed by the external actions is zero, Q is positive, Q is negative

and equal to Q in absolute value.
2. The work W performed by the external actions is negative, Q is positive, Q is
negative and greater than Q in absolute value.
3. The work W performed by the external actions is positive, and three cases can
then be distinguished.
a. Q is positive, Q is positive or zero.

b. Q is positive, Q is negative and less than Q in absolute value.
c. Q is negative, Q is positive and greater than Q in absolute value.
This classification is summarised in the following Table1 :

Cycles of the first kind W = 0 ......... Q>0 Q < 0 Q = |Q|

Cycles of the second kind W < 0 . . . . . . . . . Q > 0 Q < 0 Q > |Q|
⎧
⎪
⎪ Species a Q > 0 Q ≥ 0
⎨
Cycles of the third kind W > 0 Species b Q > 0 Q < 0 Q < |Q|
⎪
⎪
⎩ Q > |Q|
Species c Q < 0 Q > 0
We [314] now propose to compare the values taken for different cycles by the
ratio
Q + Q
ρ= .
Q
Since (Q + Q ) is zero for cycles of the first kind, negative for cycles of the second
kind and positive for cycles of the third kind, as Equation (1)2 shows, it can be seen
that the ratio ρ has, for the various categories of cycles, the following sign:
Cycles of the first kind ............... ρ=0

Cycles of the second kind . . . . . . . . . . . . . . . ρ<0
⎧
⎪
⎪ Species a . . . . . . ρ>0
⎨
Cycles of the third kind Species b . . . . . . ρ>0
⎪
⎪
⎩
Species c . . . . . . ρ<0
7.4 Carnot’s Theorem

Inspection of the preceding Table shows us, first, that the ratio ρ is smaller for any
cycle of species c of the third kind than for any cycle of the first kind, and also than
for any cycle of species a or b of the third kind.
We will now demonstrate that of two Carnot cycles described between the same
temperatures ϑ and ϑ , one of species c of the third kind and the other of the second
kind, the ratio ρ is certainly not greater for the former than it is for the latter, it
being assumed, of course, that these cycles obey the principle of the conservation of
energy and the three hypotheses previously stated.
A rigorous demonstration of this theorem requires that we distinguish cases.
Let C1 be the first cycle and C2 the second. Let T1 , T2 be the durations of these
two cycles. We will begin by distinguishing two cases depending on whether the
periods T1 , T2 are or are not commensurable.
1 Last column added by PN.]

2 [(2) in the original.]
7.4 Carnot’s Theorem 103
1. The ratio TT21 is commensurable.

We put
T2 μ2
= , (2)
T1 μ1
where [315] μ1 , μ2 are two integers with no common factor.

Let W1 , Q1 , Q 1 , ρ1 be the quantities analogous to W, Q, Q , ρ corresponding
to the cycle C1 , and let W2 , Q2 , Q 2 be the quantities analogous to W, Q, Q , ρ
corresponding to the cycle C2 . We know that
W1 > 0, Q1 < 0, Q 1 > 0,
W2 < 0, Q2 > 0, Q 2 < 0.
We distinguish two secondary cases, depending on whether the quantities W1

and |W2 | are commensurable or not.
|W2 |
A. The ratio W1 is commensurable.
We put
|W2 | m2
= , (3)
W1 m1
where m1 , m2 are two integers which we suppose have no common factor.
Put

N1 = m2 μ1 ,
(4)
N2 = m2 μ2 .
Consider N1 cycles identical to the cycle C1 , described simultaneously and

independently. Their collection will constitute a unique Carnot cycle γ 1 , of dura-
tion T1 and will be described between the temperatures ϑ and ϑ , for which the
quantities analogous to W, Q, Q will have as values (Chapter 6, Equations (5)
and (5a))
N1 W1 , N1 Q1 , N1 Q 1 .
The cycle γ 1 can be reproduced μ2 times in succession (Section 6.3).

These successive μ2 cycles γ 1 can be considered as a unique Carnot cycle
1 , described between the temperatures ϑ and ϑ , of duration
1 = μ2 T1 , (5)
for [316] which the quantities analogous to W, Q, Q have as values

μ2 N1 W1 , μ2 N1 Q1 , μ2 N1 Q 1 .
Consider, on the other hand, N2 cycles identical to the cycle C2 , described

simultaneously and independently. Their collection constitutes a unique Carnot
cycle γ 2 , described between the temperatures ϑ and ϑ and of duration T2 , for
which the quantities analogous to W, Q, Q have as values
N2 W2 , N2 Q2 , N2 Q 2 .
The cycle γ 2 can be reproduced μ1 times in succession. These successive μ1

cycles γ 2 can be considered as a unique Carnot cycle 2 , described between the
temperatures ϑ and ϑ , of duration
2 = μ1 T2 , (5a)
for which the quantities analogous to W, Q, Q have as values
μ1 N2 W2 , μ1 N2 Q2 , μ1 N2 Q 2 .
Equations (2), (5) and (5a) show us that the duration 1 of the cycle 1 is
equal to the duration 2 of cycle 2 . It can thus be assumed that the two cycles
are described simultaneously and independently. Their collection will form a
new real Carnot cycle , described between the temperatures ϑ and ϑ , for which
the quantities analogous to W, Q, Q will have as values
W = μ2 N1 W1 , +μ1 N2 W2 ,
Q = μ2 N1 Q1 , +μ1 N2 Q2 ,
Q = μ2 N1 Q 1 , +μ1 N2 Q 2 ,
In virtue of Equations (4), these equalities can be written

⎧
⎪
⎪ W = μ1 μ2 (m2 W1 + m1 W2 ) ,
⎨
Q = μ1 μ2 (m2 Q1 + m1 Q2 ) , (6)
⎪
⎪

⎩
Q = μ1 μ2 m2 Q 1 + m1 Q 2 .
In [317] virtue of Equation (3), the first of Equations (6) becomes
W = 0.
The cycle is therefore a real, irreversible cycle of the first kind. In

accordance with Clausius’ hypothesis, presented in §1, and the complementary
hypothesis stated in §2, the quantity Q must be positive. In virtue of the second
Equation (6), this latter condition becomes
m2 Q1 + m1 Q2 > 0
or
m1 m2
> .
|Q1 | Q2
But in accordance with (3), we have,3
m1 W1
= .
m2 |W2 |
Therefore,
W1 |W2 |
>
|Q1 | Q2
or
W1 W2
< .
Q1 Q2
On the other hand, Equation (1) yields
W1 = E(Q1 + Q 1 ),
W2 = E(Q2 + Q 2 ).
The preceding inequality therefore becomes
Q1 + Q 1 Q2 + Q 2
<
Q1 Q2
or
ρ1 < ρ2 . (7)
B. [318] The ratio |W 2|

W1 is incommensurable.
Let m1 , m2 be two integers with no common factor such that
m2 |W2 |
= − ε, (8)
m1 W1
where ε is a positive quantity which can be taken to be as small as you like.

The cycle is constructed as in the previous case.
3 [In W2 W1
the original, |W2 | . rather than |W2 | .]
Again, for this cycle we have

⎧
⎪
⎪ W = μ1 μ2 (m2 W1 + m1 W2 ) ,
⎨
Q = μ1 μ2 (m2 Q1 + m1 Q2 ) , (6)
⎪
⎪

⎩
Q = μ1 μ2 m2 Q 1 + m1 Q 2 .
The first Equation (6), together with Equation (8), gives
W = −εm1 μ1 μ2 W1 .
Since W is negative, the cycle is a real, irreversible Carnot cycle of the second
kind. The Sir W. Thomson’s hypothesis, stated in §1, together with the additional
hypothesis, stated in §2, shows us that the quantity Q is necessarily positive, a
condition which can be written, in virtue of the second Equation (6), as
m2 Q1 + m1 Q2 > 0
or
m1 m2
>
|Q1 | Q2
But Equation (8) yields
m2 W1 = m1 |W2 | − εm1 W1 ,
so that
W1 |W2 | W1
> −ε .
|Q1 | Q2 Q2
ε is a positive quantity which can also be taken to be as small as you [319] like. The
preceding inequality therefore requires that
W1 |W2 |
≥
|Q1 | Q2
or
W1 W2
≤ .
Q1 Q2
Equation (1) yields, moreover,
W1 = E(Q1 + Q 1 ),
W2 = E(Q2 + Q 2 ),
and it is clear that

Q1 + Q 1 Q2 + Q 2
≤
Q1 Q2
or
ρ1 ≤ ρ2 . (9)
2. The ratio TT21 is incommensurable.

Suppose that the cycle C2 is one of a kind of Carnot cycle C2 , continuously vari-
able, always real and always described between the temperatures ϑ and ϑ . Clearly,
it is always correct to make this assumption.4
When the cycle C2 varies continuously, its duration T2 remains constant or varies
continuously. Clearly, it is always possible to choose the variations of the cycle C2
in such a way that its duration T2 doesn’t remain constant.
When the cycle C2 varies continuously, the quantities W2 , Q2 , Q 2 vary continu-
ously. Since in the case of the particular cycle C2 , none of these three quantities is
equal to zero, we can always limit the variations of the cycle C2 on either side of the
form C2 in such a way that we always have
W2 < 0, Q2 < 0, Q 2 < 0.
Then the cycle C2 will always be a cycle of the second kind. [320] Moreover, in
so far as the cycle C2 varies continuously, the ratio ρ 2 varies continuously.
Now we know that if we consider one of the cycles C2 whose duration T2 is
commensurable with the duration T1 of cycle C1 , we will have
ρ2 ≥ ρ1 .
What we have just said shows that for all the forms of cycle C2 , and in particular
for the form C2 ,
ρ2 ≥ ρ1 . (10)
If, therefore, two real and irreversible Carnot cycles are both described between the same
temperatures, one of the third5 kind of species c and the other not of the third kind of species
c, the ratio ρ for the former is less than or equal to the ratio ρ for the latter.
This theorem can be restated in the following way:

Consider all the real, irreversible Carnot cycles described between the temperatures ϑ
and ϑ .
4 [It would seem that in this paragraph Duhem introduces the convention that C denotes a kind or
2
form of Carnot cycle whereas C2 stands for a particular Carnot cycle of form C2 . The fact that this
convention is not followed in the following paragraphs is perhaps best explained by printing errors
in which the two styles of letter are confused, but I have not made any such corrections.—trans.]
5 [Here and at the next occurrence of “third” the original says première (first).]
The ratio ρ for those of species c of the third kind have an upper limit A which is essentially
negative.
The ratio ρ for those which are not of species c of the third kind have a lower limit A which
is essentially negative.
We have
A ≤ A .
This is the first form of Carnot’s theorem. The employment of reversible changes
will allow us to give this theorem a more precise form.
7.5 Carnot’s Theorem (Continued): The Use of Reversible

Changes
To begin with, we will demonstrate that
A = A .
Given [321] the two temperatures ϑ and ϑ , the latter higher than the former,
suppose that it is always possible to find a system admitting four reversible changes
with the following properties:
The first is a reversible isothermal change 12 , corresponding to the temperature
ϑ and taking the system from the state 1 to the state 2.
The second is a reversible adiabatic change 23 , taking the system from state 2,
where it has the temperature ϑ , to state 3 where
it has the temperature ϑ.
The third is a reversible isothermal change 34 , corresponding to the tempera-
ture ϑ and taking the system from state 3 to state
4.
The fourth is a reversible adiabatic change 41 , returning the system from state
4 to state 1.
We have thus formed a reversible Carnot cycle, described between the tempera-
tures ϑ and ϑ .
For this cycle, we will introduce the further
hypothesis as follows: It is possible
to choose it in such a way that the change 34 is athermic.
In accordance with one of the hypotheses introduced at the end of Section 6.8,
there is a continuously variable real isothermal change S12 taking the system from
the state 1 with velocities equal to zero
to the state 2 with its velocities equal to zero,
and having as a limit the change 12 . Similarly, there is a continuously variable
real isothermal change S21 taking the system from the state 2 with velocities equal
to zero to the state 1 with its velocities equal to zero, and having as a limit the
change 21 .
The isothermal change 34 and the adiabatic changes 23 and 41 give rise to
analogous observations.
7.5 Carnot’s Theorem (Continued): The Use of Reversible Changes 109
The collection of changes

S12 , S23 , S34 , S41
form a real Carnot cycle described between the temperatures ϑ and ϑ , [322] which
is continuously variable and has as its limit the cycle

12 , 23 , 34 , 41.
Similarly, the collection of changes

S43 , S32 , S21 , S14
form a real Carnot cycle described between the temperatures ϑ and ϑ , which is
continuously variable and has as its limit the cycle

43 , 32 , 21 , 14 .

Let χ 12 , χ 34 be
the quantities
of heat released by the reversible changes 12 ,
34 . The changes 21 , 43 release the quantities of heat χ21 = −χ12 , χ43 =
−χ34 , so that for each of the two cycles

12 , 23 , 34 , 41 and 43, 32 , 21 , 14 ,
the quantity analogous to ρ has the same value
χ12 + χ34
r= .
χ34
Let Q12 , Q34 be the thermic effects of the changes S12 , S34 . For the cycle (S12 ,
S23 , S34 , S41 ) the quantity analogous to ρ has the value
Q12 + Q34
ρ1 = .
Q34
Let Q21 , Q43 be the thermic effects of the changes S21 , S43 . For the cycle (S43 ,
S32 , S21 , S14 ) the quantity analogous to ρ has the value
Q21 + Q43
ρ2 = .
Q43

By hypothesis, the change 34 is not athermic. To fix our ideas, suppose χ 34 is
negative. S34 and S43 [323] can be taken to be sufficiently close to the
The changes
changes 34 and 43 that Q34 has the same sign as χ 34 and Q43 has the same sign
as χ 43 . We will then have
Q34 < 0, Q43 > 0,

so that ρ 1 certainly corresponds to a cycle of species c of the third kind and ρ 2 does
not correspond to a cycle of species c of the third kind.
The theorem demonstrated in the preceding section will give us
ρ1 ≤ A, ρ2 ≥ A .
But on the other hand, when the two cycles
(S12 , S23 , S34 , S41 ) and (S43 , S32 , S21 , S14 )
tend towards the cycles

12 , 23 , 34 , 41 and 43 , 32 , 21 , 14 ,
respectively, the two ratios ρ 1 , ρ 2 tend to the same limit r. Since the quantities A,
A do not vary when the temperatures ϑ and ϑ are held constant, this can only be
the case if we have
A = A , (10)
as stated.
We see, moreover, that if a reversible Carnot cycle is described between the tem-
peratures ϑ and ϑ (ϑ being higher than ϑ) and, further, if the isotherm described
at the temperature ϑ is not athermic, the ratio
Q + Q
ρ=
Q
has, for this cycle, the value A.
What happens to the ratio ρ for a reversible Carnot cycle described between the
temperatures ϑ and ϑ if the isotherm at the temperature ϑ is athermic?
Designating [324] this cycle by C1 , we have, by hypothesis,
Q1 = 0.
Let C2 be another reversible Carnot cycle, described between the temperatures ϑ

and ϑ , for which the isotherm at the temperature ϑ is not athermic. For this cycle,
we have by the preceding proposition,
Q2 + Q 2
= A.
Q2
The two cycles C1 and C2 can be considered virtual. It is always possible to sup-
pose that they run simultaneously and independently. Together they will thus form
a new reversible Carnot cycle described between the temperatures ϑ and ϑ . The
isotherm described at the temperature ϑ corresponds to a thermic effect Q2 , so that
it will not be athermic. The isotherm described at the temperature ϑ corresponds to
a thermic effect (Q 1 + Q 2 ). In accordance with the preceding theorem, we have for
this cycle
Q2 + Q 1 + Q 2
= A.
Q2
This equation will only be compatible with the preceding one if
Q 1 = 0.
Whence this proposition:

If a reversible Carnot cycle is described between the temperatures ϑ and ϑ (ϑ being higher
than ϑ) and if the isothermal change described at the temperature ϑ is athermic, then the
isothermal change produced at the temperature ϑ 6 is also athermic.
In that case, the ratio
Q + Q
Q
is meaningless.
We will now demonstrate that, for any [325] realisable and irreversible cycle
described between the temperatures ϑ and ϑ , the ratio
Q + Q
ρ=
Q
never reaches the value A.
Since the value A is essentially negative, the ratio ρ cannot achieve the value A
unless the cycle belongs either to the second kind or to the species c of the third
kind. Suppose that for a cycle C1 (to fix our ideas, we take it to be of the second
kind) we have
Q1 + Q 1
ρ1 = =A
Q1

or, since Equation (1) sanctions the replacement of Q1 + Q 1 by EW1 ,
EW1 = AQ1 . (11)
We have seen that the existence of real, reversible cycles is by no means absurd.
We therefore admit that a real, reversible Carnot cycle C2 , described between the
same temperatures ϑ, ϑ as cycle C1 , has been constituted for which the quantity
W2 is not equal to zero.
6 [ϑ in the original.]
The quantity W2 will, therefore, be positive or negative. We can always suppose

it to be positive since, if it were negative, it suffices to realise the reversible cycle7
C2 in the opposite sense to that in which it is positive.
By continually varying the Carnot cycle C2 , without changing either its
reversibility or the temperatures between which it is described, we can always see
to it that the following conditions are satisfied:
1. The duration T1 of the Carnot cycle C1 and the duration T2 of the Carnot cycle
C2 are commensurable.
2. The quantities |W1 | and W2 are commensurable.
Put
T1 μ1 W2 m2
= , = (12)
T2 μ2 |W1 | m1
where μ1 , μ2 , m1 , m2 are four integers.
From [326] the cycles C1 and C2 we form a Carnot cycle in the same way as
in §4. For the cycle we have
⎧
⎪ W = μ1 μ2 (m2 W1 + m1 W2 ) ,
⎪
⎨
Q = μ1 μ2 (m2 Q1 + m1 Q2 ) , (13)
⎪
⎪

⎩
Q = μ1 μ2 m2 Q 1 + m1 Q 2 .
The second Equation (12) together with the first Equation (13) shows us that
W = 0.
The cycle is therefore a cycle of the first kind.

The quantity Q2 cannot be equal to zero. The cycle C2 is, in fact, reversible.
According to one of the propositions just demonstrated, the quantity Q2 cannot be
equal to zero unless the quantity Q 2 is too. The equation

W2 = E Q2 + Q 2
would therefore yield

W2 = 0,
contrary to our hypothesis.

Since the quantity Q2 in the reversible cycle C2 is not equal to zero, we know that
Q2 + Q 2
=A
Q2
7 [cercle, rather than cycle, in the original.]

or
EW2 = AQ2 . (14)
Equations (11) and (14), together with the second equality (13), show that, for
the cycle , we have
E
Q = μ1 μ2 (m2 W1 + m1 W2 ) = 0.
A
The cycle will therefore be a Carnot cycle of the first kind, described between
the temperatures ϑ and ϑ , in which the isotherm described at the [327] temperature
ϑ is athermic. But on the other hand, the cycle would be a realisable Carnot cycle
and it would not be exclusively composed of equilibrium states, since the cycle C2
is not exclusively composed of equilibrium states. The existence of the cycle is,
therefore, in contradiction with the hypothesis stated in §2. Since we accept that the
latter hypothesis is correct, we can formulate the following proposition:
For any realisable and irreversible cycle, the ratio
Q + Q
ρ=
Q
cannot reach the value A.
We will now combine the various propositions demonstrated in the preceding

and present sections in a single theorem:
Let ϑ and ϑ be two temperatures, with ϑ higher than ϑ. There is a negative magnitude A,
depending solely on the two temperatures ϑ and ϑ , with the following properties:
1. For all reversible Carnot cycles, whether realisable or not, described between the
temperatures ϑ and ϑ ,
Q + Q
ρ= = A. (15)
Q
2. For all Carnot cycles described between the temperatures ϑ and ϑ which are realisable
and irreversible, and of species c of the third kind,
Q + Q
ρ= < A. (16)
Q
3. For all Carnot cycles described between the temperatures ϑ and ϑ which are realisable
and irreversible, and not of species c of the third kind,
Q + Q
ρ= >A (17)
Q
From [328] this statement are excluded:
1. Realisable and irreversible cycles for which

W > 0, Q = 0.
2. The reversible cycles for which
Q = 0.
For the latter, it also holds that

Q = 0.
7.6 Absolute Temperature

The quantity A is a function of the temperatures ϑ and ϑ ; we will now study more
deeply the nature of this function, or rather, the function
ψ(ϑ , ϑ) = A − 1. (18)
This function, which is only defined for values of ϑ higher than ϑ, is negative
and greater than 1 in absolute value.
Given Equation (18), the equality and the inequalities (15), (16) and (17) can be
replaced by this equality and the inequalities
Q
= ψ(ϑ , ϑ), (15a)
Q
Q
> ψ(ϑ , ϑ), (16a)
Q
Q
< ψ(ϑ , ϑ). (17a)
Q
Let ϑ, ϑ , ϑ be three temperatures in increasing order of magnitude. We now
propose to find a relation between the three quantities
ψ(ϑ, ϑ ), ψ(ϑ , ϑ ), ψ(ϑ, ϑ ).
Consider a reversible Carnot cycle described between the temperatures ϑ and ϑ .

The cycle takes the body to the temperature ϑ , in the [329] state A . It is brought to
the state B by an isothermal transformation, and then to the state B by an adiabatic
transformation, when the temperature is ϑ. A new isothermal transformation at the
temperature ϑ brings it to the state A, and a new adiabatic transformation brings it
back to the state A .
Suppose that the calorific effect of the transformation BA, which we will
designate by Q(BA), is different from zero.
Then we will have

Q A B
= ψ(ϑ, ϑ ). (a)
Q (BA)
7.6 Absolute Temperature 115
When undergoing the transformation B B the body passes, at least once, through
the temperature ϑ . Let B be a state which has this temperature. Similarly, when
undergoing the transformation AA the body passes, at least once, through the tem-
perature ϑ . Let A be a state in which it has this temperature. We connect the two
states A , B , by a series of equilibrium states of the system all corresponding to the
temperature ϑ , and we will obtain a reversible isothermal change A B .
The cycle A B BAA will be a reversible Carnot cycle described between the
temperatures ϑ and ϑ . Since Q(BA) is not zero, we will have

Q A B
= ψ(ϑ, ϑ ). (b)
Q (BA)
Q(A B ) is certainly not zero since ψ(ϑ, ϑ ) is not zero. But clearly,

Q A B + Q B A = 0. (c)
Q(B A ) is therefore not zero. Consequently, in the reversible Carnot cycle

A B B A A , described between the temperatures ϑ and ϑ , we have

Q A B

= ψ(ϑ , ϑ ). (d)
Q BA
Equations (a), (b), (c), (d) yield
ψ(ϑ, ϑ ) = −ψ(ϑ, ϑ )ψ(ϑ , ϑ ). (19)
This [330] is the relation we wished to obtain.

The function ψ(ϑ, ϑ ) so far is meaningful only if ϑ is higher than ϑ.
When ϑ tends towards ϑ by values higher than ϑ, Q clearly tends, in a reversible
Carnot cycle described between the temperatures ϑ and ϑ , towards the limit (–Q).
Therefore, ψ(ϑ, ϑ ) tends towards –1. We will agree to put
ψ(ϑ, ϑ) = −1.
When ϑ is less than ϑ, we define ψ(ϑ, ϑ ) by the equality
ψ(ϑ, ϑ ) = −1/ψ(ϑ , ϑ).
In accordance with these definitions, the functional Equation (19), established

under the supposition that
ϑ < ϑ < ϑ ,
remains true whatever ϑ, ϑ , ϑ may be. Moreover, we are assured that the function
ψ(ϑ, ϑ ) is continuous for all values of ϑ and ϑ .
Take an arbitrary fixed temperature, ϑ 0 . The functional Equation (19) gives us
ψ(ϑ, ϑ ) = −ψ(ϑ0 , ϑ )/ψ(ϑ0 , ϑ).
Since the temperature ϑ 0 is fixed, ψ(ϑ0 , ϑ) becomes a function of the single

variable ϑ, and we put
F(ϑ) = −λψ(ϑ0 , ϑ), (20)
λ being a constant.
The preceding equation becomes
ψ(ϑ, ϑ ) = − F(ϑ )/F(ϑ). (21)
The function ψ(ϑ, ϑ ) is a continuous function of ϑ, ϑ ; when ϑ is higher than

ϑ, it is negative and greater than 1 in absolute value. The function F(ϑ) therefore
has the following properties:
It [331] is essentially positive;
It varies continuously with the temperature ϑ;
It always increases at the same time as the temperature ϑ.
Instead of the quantity ϑ it is possible, in accordance with what we have said
above (Section 3.6), to take the quantity T = F(ϑ), whose value for each tem-
perature no longer depends on the choice of thermometer, as determining the
temperature. T is called the absolute temperature.
If the absolute temperature T is adopted to represent temperature, and in all the
preceding formulas the substitution
ϑ =T
is made, then the substitution

F(ϑ) = F(T) = T
should also be made. Recalling that
ψ(ϑ0 , ϑ0 ) = −1,
it can be seen that Equation (20) yields
F(ϑ0 ) = −λ.
The absolute temperature is therefore not entirely determined; at an arbitrary

temperature ϑ 0 , it takes an arbitrary value λ. It is usual to take as the temperature
ϑ 0 the temperature which serves as 0 on the centigrade scale and, for λ, the inverse
of a certain constant encountered in the study of gases, the coefficient of expansion
of perfect gases.
7.7 A Definitive Statement of Carnot’s Theorem 117
It is said that this definition of absolute temperature was introduced by Sir

W. Thomson and set forth on several occasions by G. Lippmann.
7.7 A Definitive Statement of Carnot’s Theorem

The determination of the function ψ(ϑ, ϑ ) provided by Equation (21) will allow us
to give Carnot’s theorem a new and definitive form.
We [332] first exclude:
1. Reversible Carnot cycles described between the temperatures ϑ and ϑ in which
the change produced at the temperature ϑ is athermic.
2. Realisable, irreversible Carnot cycles of the third kind described between the
temperatures ϑ and ϑ , in which the change produced at the temperature ϑ is
athermic.
For reversible cycles, whether realisable or not, Equations (15a) and (21) yield

Q F ϑ
=−
Q F (ϑ)
and, since Q is not zero,
Q Q
+
= 0.
F (ϑ) F ϑ
For realisable, irreversible cycles of species c of the third kind, Equation (16b)
and Equation (21) yield

Q F ϑ
<−
Q F (ϑ)
But in this case, the quantity Q is negative, so that
Q Q
+
> 0.
F (ϑ) F ϑ
For the other irreversible cycles, the inequality (17b) together with Equation (21)
yields

Q F ϑ
>−
Q F (ϑ)
For all these cycles, the quantity Q is positive, so that, again,
Q Q
+
> 0.
F (ϑ) F ϑ
We now return to the two classes that we have excluded.

For all cycles of the first class, we know that not only [333] is Q = 0, but also
Q = 0. We can therefore write
Q Q
+
= 0.
F (ϑ) F ϑ
For all cycles of the second class, we have
Q = 0, W > 0.
Equation (1), which is reduced here to
W = EQ ,
then gives us
Q > 0.
We can therefore write, for all cycles of this class,
Q Q
+ > 0.
F(ϑ) F(ϑ )
Bringing together all the results that we have just obtained, we can state
CARNOT ’ S THEOREM in the following general form:
1. For all reversible Carnot cycles, whether realisable or not, described between the
temperatures ϑ and ϑ , we have
Q Q
+
= 0. (22)
F (ϑ) F ϑ
2. For all real and irreversible Carnot cycles, whether realisable or not, described between
the temperatures ϑ and ϑ , we have
Q Q
+
> 0. (23)
F (ϑ) F ϑ
These statements can be put in somewhat more explicit form by reverting to the
definition of the total calorific effect of a change or of several changes.
Let [334] Q be the sum of the quantities of heat released by the system during
this collection of changes. If these changes are reversible, we have
Q = Q,
allowing replacement of Equation (22) by
Q Q
+
= 0. (22a)
F (ϑ) F ϑ
7.7 A Definitive Statement of Carnot’s Theorem 119
If the changes under consideration are real and irreversible, let [T] designate the
sum of the increases that the kinetic energy of the system undergoes during these
changes. We have
EQ = EQ + [T],
allowing us to replace the inequality (23) by the inequality

EQ + T EQ + T
+
>0 (23a)
F (ϑ) F ϑ
These various statements are not subject to further exceptions.

Chapter 8
Entropy and Clausius’ Theorem
8.1 Conditions and Hypotheses

Carnot’s theorem has, it is said, been generalised by R. Clausius. At all events, this
generalisation is only justified provided a certain number of conditions and hypothe-
ses are satisfied, some of which concern the system studied, others the changes
undergone by the system.
It will not be useful to us in this chapter to distinguish, among the parameters
defining the state of the system, those which we have designated by the letters
a, b, . . . , l from those that we have designated by the letters α, β, . . . , λ. On the
other hand, it will be necessary for us to [335] highlight the temperature ϑ amongst
the parameters that define the state of the system. We will suppose, moreover, that
the latter has the same value at all points of the system. Thus, the state of the system
will be defined by the n parameters
α, β, . . . , λ, ϑ.
The system that we will study will not only be subject to this restriction of having
the same temperature at all of its points in each of its states, it being possible, more-
over, for this temperature to vary from one state to another. We will also suppose
that it is subject to two more fundamental restrictions:
FIRST RESTRICTION.—For any state of the system defined by the values
α, β, . . . , λ, ϑ,
of the variable parameters, we admit that it is always possible to find, in at least one way, a
system of external bodies, all at the temperature ϑ, such that the system under consideration
remains indefinitely at equilibrium in the state α, β, . . . , λ, ϑ if subjected to the action of
these external bodies, which don’t change, and if it is taken with a collection of initial
velocities equal to zero,
The statement of this restriction is not a vain precaution. Systems are often
encountered in physics which are not subject to it and consequently to which the
considerations to follow are not applicable. We cite some examples:

122 8 Entropy and Clausius’ Theorem
1. A conducting body is traversed by electric currents. There figure among the

variable parameters determining the state of the conductor:
The components u, v, w of electric flux at each point (x, y, z) of the conductor;
The solid density ρ of the electricity at each point of the mass of the
conductor;
The [336] surface density σ of the electricity at each point of the surfaces of
discontinuity which divide the conductor or which limit it.
Is it possible, by making suitably chosen external bodies act, to maintain the
conductor in question at equilibrium so that the variables just enumerated always
∗
have values independent of time? This is not possible unless the relation
∂u ∂v ∂w
+ + =0 (i)
∂x ∂y ∂z
holds at each point of the mass of the conductor, and, at each point of a surface of
discontinuity whose normal has the two orientations N1 , N2 ,

u1 cos(N1 , x) + v1 cos(N1 , y) + w1 cos(N1 , z)
(ii)
+ u2 cos(N2 , x) + v2 cos(N2 , y) + w2 cos(N2 , z) = 0 .
In fact, if these two relations are not verified, the parameters ρ and σ necessarily
vary with the time, t, in virtue the general equations
∂u ∂v ∂w ∂ρ
+ + =− ,
∂x ∂y ∂z ∂t
u1 cos(N1 , x) + v1 cos(N1 , y) + w1 cos(N1 , z)

∂σ
+ u2 cos(N2 , x) + v2 cos(N2 , y) + w2 cos(N2 , z) = −.
∂t
Relations (i) and (ii) characterise currents that are called uniform, and it is
clear that it is possible to state the following proposition, which has profound
consequences in electrodynamics1 :
A system of non-uniform currents does not satisfy the forgoing restriction.
2. A closed circuit through which a uniform current flows encloses an electrolytic

conductor. Among the variables α, β, . . . , λ, ϑ defining the state of the system figure
the intensity of the current and [337] variables which determine the chemical state of
the system. Now, in virtue of the definition of electrolytes, known as Faraday’s law,
an electrolyte traversed by a current during each interval of time undergoes a chem-
ical change proportional in magnitude to the interval of time and the intensity of the
∗ The following two equations, numbered (1) and (2) in the original, are renumbered here to avoid
confusion with Equations (1) and (2) in the next section, which are referred to later.
1 P. Duhem, Leçons sur l’Électricité et le Magnétisme, vol. III, p. 221.
8.1 Conditions and Hypotheses 123
current. It would therefore be contradictory to imagine external bodies whose action

maintained the system at equilibrium, and, consequently, rendered invariable the
current intensity and the chemical state of the circuit at the same time. Thus, a circuit
through which even uniform currents run does not satisfy the preceding restriction
if it encloses electrolytes. This conclusion is important in electrodynamics.2
The preceding restrictive hypothesis entails a remarkable corollary:
COROLLARY.—It is always possible to pass from any state (α, β, . . . , λ, ϑ) to any other
(α , β , . . . , λ , ϑ ) by an infinity of reversible changes.
In fact, it is always possible to set up a continuous series of states of the sys-

tem between the two states (α, β, . . . , λ, ϑ) and (α , β , . . . , λ , ϑ ) in infinitely many
ways. With suitably chosen external bodies, these states can be transformed into
equilibrium states. Each of these series, constituted as we have just indicated,
then forms a continuous series of equilibrium states, or, in virtue of a hypothesis
mentioned in Section 6.8, a reversible change.
SECOND RESTRICTION.—Suppose that, beginning in a certain state (α, β, . . . , λ, ϑ) of the
system, it is subject to a virtual change
δα, δβ, . . . , δλ, δϑ.
The actions of the external bodies maintaining it in equilibrium in the state

(α, β, . . . , λ, ϑ) perform the virtual work
Aδα + Bδβ + . . . + Lδλ + δϑ.
We [338] suppose that the quantities A, B, . . . , L, are finite, uniform and

continuous functions of the parameters α, β, . . . , λ, ϑ.
As with the foregoing restriction, the statement of this one is not redundant
because it is by no means evident that all systems studied in physics are subject to it.
In particular, Marcel Brillouin3 has provided a theory for permanent deformations
which assumes that these quantities A, B, . . . , L, are continuous but non-uniform
functions of the variables α, β, . . . , λ, ϑ.
We add the following hypothesis to that just stated:
The virtual work of the actions which must be applied to a given system to maintain it at
equilibrium does not depend on the absolute position that the system occupies in space nor
on the variation in this position.
First of all, it follows from this hypothesis that the quantities A, B, . . . , L, do

not depend on those of the variables among α, β, . . . , λ, ϑ which fix the absolute
position in space of the system.
2 P. Duhem, Leçons sur l’Électricité et le Magnétisme, vol. III, p. 220. Journ. de Math. (4th. series),
vol. IX.—No. III; 1893.
3 Marcel Brillouin, Déformations permanentes et Thermodynamique (Comptes rendus, 6, 13, 20
and 27 February 1888).
Second, it follows that the virtual variations δα, δβ, . . . , δλ, δϑ corresponding to
the variables fixing the absolute position of the system in space do not figure in
the sum
Aδα + Bδβ + . . . + Lδλ + δϑ.
The coefficient of each of these is equal to zero.
8.2 Calorific Coefficients of the System at Equilibrium

In order to conveniently explain what is to follow, we will employ the language of
the geometry of n dimensions, n being the number of the parameters α, β, . . . , λ, ϑ
which define the state of the system. The statements at which we will arrive
can always be represented by plane figures if there is just one of the parameters
α, β, . . . , λ, and by figures in space if there are two of them.
A [339] state of the system, corresponding to a system of values of the coordi-
nates α, β, . . . , λ, ϑ, will be represented by a point in the space of n dimensions.
A change, corresponding to a linear series of values of α, β, . . . , λ, ϑ, will be
represented by a line.
If the system is taken in a determinate state (α, β, . . . , λ, ϑ), the actions
A, B, . . . , L, which allow it to remain at equilibrium in this state are, by hypothesis,
uniform and continuous functions of the variables α, β, . . . , λ, ϑ.
⎧
⎪
⎪ A = fα (α, β, . . . , λ, ϑ),
⎪
⎪
⎪
⎪
⎪
⎪ B = fβ (α, β, . . . , λ, ϑ),
⎨
....................., (1)
⎪
⎪
⎪
⎪
⎪
⎪ L = fλ (α, β, . . . , λ, ϑ),
⎪
⎪
⎩
= fϑ (α, β, . . . , λ, ϑ).
Equations (1) are the equilibrium equations of the system. These equations will
be said to be known when the form of the n functions α, β, . . . , λ, ϑ is known.
Let U be a uniform and continuous function of α, β, . . . , λ, ϑ which represents
the internal energy of the system. Put
⎧
⎪ ∂U fα
⎪
⎪ Rα = − ,
⎪
⎪ ∂α E
⎪
⎪
⎪
⎪ ∂U f
⎪
⎪ Rβ = − ,
β
⎪
⎪
⎪
⎨ ∂β E
..............., (2)
⎪
⎪
⎪
⎪ ∂U
⎪
⎪ fλ
⎪
⎪ Rλ = − ,
⎪
⎪ ∂λ E
⎪
⎪
⎪
⎪ ∂U fϑ
⎩ C= − .
∂ϑ E
8.2 Calorific Coefficients of the System at Equilibrium 125
The n quantities Rα , Rβ , . . . , Rλ defined by these equations will be uniform func-

tions of α, β, . . . , λ, ϑ. They will be (Chapter 5, Equations (12) and (12a)) the
calorific coefficients of the system at equilibrium. Among these parameters, there
is one which will be distinguished from the others by specific properties, the coef-
ficient [340] C. We will call it the caloric capacity of the system, relative to the
variables α, β, . . . , λ, ϑ.
Like, on the one hand the quantity U (Section 4.2, tenth convention) and
on the other hand the quantities A, B, . . . , L, (this Chapter, §1), the quantities
Rα , Rβ , . . . , Rλ , C do not depend on those among the variables α, β, . . . , λ, ϑ which
fix the absolute position of the system in space. Moreover, those of the calorific
coefficients multiplying the variations of these latter variables in the expression of
dQ are all zero.
The equation
dQ = −(Rα δα + Rβ δβ + . . . + Rλ δλ + C δϑ) (3)
determines (Chapter 5, Equations (13)) the quantity of heat released by the system
while subject to the real or virtual change (δα, δβ, . . . , δλ, δϑ).
The differential equation
Rα dα + Rβ dβ + . . . + Rλ dλ + C dϑ = 0 (4)
represents, in the n-dimensional space considered, a family of spaces of (n – 1)

dimensions (of lines, in the space of two dimensions, of surfaces in the space
of three dimensions). We accept that through every point (α, β, . . . , λ, ϑ) of the
n-dimensional space there passes one and only one of these spaces of (n – 1)
dimensions, and in addition that the space thus determined passing through the
point (α, β, . . . , λ, ϑ) is displaced and deformed continuously when the point
(α, β, . . . , λ, ϑ) is displaced continuously.
We further accept that a space of reversible adiabats is never closed on itself
like a closed line or a closed surface, but always forms a simply connected space
extending to the limits of the field of values of α, β, . . . , λ, ϑ.
Consider any line traced entirely within one of these spaces of (n –1) dimensions.
This line represents a continuous series of states of the system. In each of these states
(α, β, . . . , λ, ϑ), we suppose the system to be surrounded by bodies having the same
temperature [341] as it and exerting on it actions given by Equations (1). This series
of states will be a series of equilibrium states. In accordance with the hypothesis
stated in Section 6.8, this series of equilibrium states constitutes a reversible change.
Equations (3) and (4) show that, for all elements of this reversible change, we have
dQ = 0,
so that this reversible change will be adiabatic. Thus, each line traced entirely in one
of the spaces of (n – 1) dimensions defined by Equation (4) represents a reversible
adiabatic change of the system. We call each of these (n – 1)-dimensional spaces
defined by Equation (4) a space of reversible adiabats.
Take two points, m and n, in the space of n dimensions considered. Between these
two points, m and n, we draw two lines, l and l , such that each of these has no more
than one point in common with such a space of reversible adiabats. The following
propositions, which are geometrically obvious when the space of reversible adiabats
are lines (n = 2) or surfaces (n = 3), can be stated in entirely general fashion:
The space of reversible adiabats drawn through a point a on the line l inter-
sects the line l in one and only one point a ; these two points a and a are called
corresponding points.
If the three points a , b , c of the line l correspond, respectively, to three points
a, b, c of the line l, and the point b is situated, on the line l, between the points a and
c, then the point b is situated on the line l , between the points a and c .
If two points a, b are infinitesimally close on the line l, their correspondents a ,
b are infinitesimally close on the line l .

To the point m on the line l there corresponds the same point m on the line l , and
to the point n on the line l there corresponds the same point n on the line l .
This mode of correspondence between the points of two lines will be useful to us
in the demonstration of Clausius’ theorem, to which we now proceed.
8.3 Demonstration of Clausius’ Theorem

Any [342] line traced in the n-dimensional space under consideration represents a
series of states of the system. If, in each of these states, we suppose the system
surrounded by bodies at the same temperature as it and exerting actions on it given
by Equations (1), then each of these states becomes an equilibrium state, and the
series of these states will be a reversible change. Thus, any line traced in the space
of α, β, . . . , λ, ϑ always represents, in one and only one way, a reversible change of
the system.
We will study the properties of similar transformations, but we will provision-
ally suppose that each of these reversible changes under study is subject to certain
restrictions that we will now enumerate:
1. The line representing one of these reversible changes under study has no part of
finite extension traced in its entirety in a space of reversible adiabats.
2. The line representing one of these reversible changes under study never meets the
same space of reversible adiabats more than once. This last restriction evidently
presupposes another: the line considered never passes through the same point
more than once.
Consider two reversible changes, l and l , subject to the preceding restrictions.
Both take the system in the same initial state, represented by the point m, and lead
it to the same final state, represented by the point n.
Let a, b be two infinitesimally close states of the change l, the state b coming after
the state a when the change l is followed from m towards n. Let a , b be two states
of the change l corresponding, respectively, to the two states a, b. In accordance
with what we have seen, the two states a , b are infinitesimally close on the change
8.3 Demonstration of Clausius’ Theorem 127
l and, when this change is followed from m to n, the state a is encountered before
the state b .
Let ϑ be the temperature of the system in the state a and ϑ the temperature of
the system in the state a . Let dQ be the quantity of heat released by the system
while it undergoes the infinitesimal reversible change ab and dQ the quantity of
heat released by the [343] system while it undergoes the infinitesimal reversible
change a b . We now propose to demonstrate that
dQ dQ
=
. (5)
F (ϑ) F ϑ
Through the point a, we draw an isothermal line to the point c where it meets the
space of reversible adiabats drawn through the point b. In general, the point c will be
infinitesimally close to the two points a and b. Similarly, we draw an isothermal line
through the point a to the point c where it meets the space of reversible adiabats
drawn through the point b . In general, the point c will be infinitesimally close to the
two points a and b . For the moment, we suppose that the hypotheses just mentioned
are satisfied, leaving for treatment later the case where they are not satisfied.
We join the two points c, b by an infinitesimal line drawn in the space of
reversible adiabats to which these two points belong, and similarly we join the two
points c , b by an infinitesimal line drawn in the space of reversible adiabats to
which these two points belong.
The lines ac, cb, a c , c b each represent an infinitesimal reversible change of the
system.
The two reversible changes cb, c b are adiabatic; when the system is subjected
to any one of them, it releases a quantity of heat equal to zero.
Let dQ1 be the quantity of heat released by the system while it describes the
reversible isothermal ac and dQ 1 the quantity of heat released by the system while
it describes the reversible isothermal c a .
Consider the closed and reversible cycle abca. When the system describes this
cycle, it releases a quantity of heat (dQ − dQ1 ). Applying Equation (16) (Chapter 5)
to each of the elements of this cycle, we easily obtain the equation

E (dQ − dQ1 ) = (A dα + B dβ + . . . + L dλ + dϑ).
the integral extending over the entire cycle abca.

Let [344] (α1 , β1 , . . . , λ1 , ϑ1 ) be one of the points of this cycle, and
A1 , B1 , . . . , L1 , 1 , the values taken by the functions A, B, . . . , L, at this point;
then we can write the preceding equation as

E (dQ − dQ1 ) = A1 dα + B1 dβ + . . . + L1 dλ + 1 dϑ

+ [(A − A1 ) dα + (B − B1 ) dβ + . . . + (L − L1 ) dλ + ( − 1 ) dϑ]
But since the cycle is closed, we have

dα = 0, dβ = 0, . . . , dλ = 0, dϑ = 0.
On the other hand, since, by hypothesis, the quantities A, B, . . . , L, vary

continuously with the position of the point (α, β, . . . , λ, ϑ), the quantities
(A − A1 ) , (B − B1 ) , . . . , (L − L1 ) , ( − 1 )
are infinitesimal, at least of the same order as the dimensions of the cycle. The
quantity

[(A − A1 ) dα + (B − B1 ) dβ + . . . + (L − L1 ) dλ + ( − 1 ) dϑ]
is therefore infinitesimal to a higher order than that of the dimensions of the cycle,
so that the preceding equation can be written
dQ − dQ1 = 0. (6)
An analogous demonstration, applied to the cycle a b c a , yields the equation
dQ + dQ 1 = 0. (7)
We can now describe the system by:
1. The reversible isothermal ac, relative to the temperature ϑ.

2. The [345] reversible adiabat cb.
3. The reversible adiabat bb .
4. The reversible adiabat b c .
5. The reversible isothermal c a , relative to the temperature ϑ .
6. The reversible adiabat a a.
The system will have traced a reversible Carnot cycle between the tempera-
tures ϑ and ϑ . At the temperature ϑ, it will have released a quantity of heat dQ1 ,
and at the temperature ϑ , a quantity of heat dQ 1 . We therefore have (Chapter 7,
Equation (22a))
dQ1 dQ 1
+
= 0
F (ϑ) F ϑ
In virtue of Equations (6) and (7), this equation yields Equation (5), which is
what we wished to demonstrate.
Considering successively all the elements of the transformation l in moving from
m to n, the elements corresponding to them in the transformation l trace once
and once only the transformation l in going from m to n . For each group of
corresponding elements, we write an equation analogous to Equation (5) and add
one after another all the results obtained. We obtain the equation

dQ dQ
=
(8)
l F (ϑ) l F ϑ
which leads us to state the following proposition:

Considering all the reversible changes (subject to the restrictions mentioned) leading the
system from the state m to the state n, for any two such states whatever, the integral

dQ
F (ϑ)
has the same value for all these changes.
It remains now to successively lift the various restrictions introduced to obtain

the foregoing demonstration.
We have supposed that from all points a of the line l it is possible to draw an
isothermal meeting the space of reversible adiabats drawn [346] through the point
b, infinitesimally close to the point a, in a point c, itself infinitesimally close to the
point a. This comes down to accepting, as is easily shown, that the isothermal line
issuing from point a is not tangent to the point a in the space of reversible adiabats
which passes through this point. Such a hypothesis is made for all points a of the
line l .
In the first place, it is easily shown that, if there is, either on the line l or on the
line l or on both, a limited number of points isolated from one another for which
the preceding hypothesis ceases to hold, the preceding theorem will certainly not be
incorrect. It could therefore only be incorrect if there exists on at least one of these
lines, the line l for example, a finite extended portion, fgh, at every point of which
the reversible isothermal would be tangent to the space of reversible adiabats.
If such conditions are fulfilled for the line fgh, two cases can be distinguished.
It is possible that, from the point f to the point h, lines fg h can be drawn
infinitesimally close to the line fgh which escape the preceding conditions.
It is possible, on the contrary, that all the points of the line fgh belong to a domain
D, drawn through the space α, β, . . . , λ, ϑ, and that at every point of this domain the
reversible isothermal is tangent to the space of reversible adiabats.
dQ
In the first case, the integral F(ϑ) over the line mfg hn will have, in accordance
with the preceding demonstration, the same value as the integral of the same form
over the line l . Its value will therefore not vary when the line mfg hn varies contin-
uously so as to tend towards the line mfghn or l. Now under these conditions, this
integral will certainly tend towards

dQ
l F (ϑ)
We therefore have once more in this case

dQ dQ
=
.
l F (ϑ) l F ϑ
Only [347] the second of the two cases that we have just distinguished is still
excluded from our investigations.
We will try to characterise this case more precisely.
A reversible isothermal satisfies the differential equation
dϑ = 0.
A space of reversible adiabats is defined by the differential equation
Rα dα + Rβ dβ + . . . + Rλ dλ + C dϑ = 0 (4)
In order that at every point (α, β, . . . , λ, ϑ) of the domain D any reversible
isothermal be tangent to the space of reversible adiabats, it is necessary and
sufficient that at each point of the domain,
Rα = 0, Rβ = 0, . . . , Rλ = 0. (9)
We therefore provisionally exclude from our investigations those systems which sat-
isfy Equations (9) at all points of a domain of finite extent belonging to the space of
α, β, . . . , λ, ϑ.
We have so far supposed that each of the two changes l and l do not meet the
same space of reversible adiabats in more than one point. We will now relinquish
this restriction and suppose that each of these two changes l and l can meet the same
space of reversible adiabats at a finite number of points, isolated from one another.
Take the space of reversible adiabats E(m) passing through the point m and the
space of reversible adiabats E(n) passing through the point n. These two spaces of
(n – 1) dimensions divide the space of n dimensions into three regions. One of these
regions is included between the spaces E(m) and E(n); the other two are exterior.
Following the straight line connecting the point m with the point n in such a way
that the point m is met before arriving at point n, I will refer to the region traversed
before reaching m as the first. Immediately after passing the point m [348] leads
to what I will call the second region. Leaving this second region at the moment of
passing point n leads to what I call the third region.4
It might happen that the line l has a point p in common with a space of reversible
adiabats E, other than the space E(m) or E(n), without passing through this space
in the meeting point. If two spaces of reversible adiabats are drawn infinitesimally
close to the space E, and situated on either side of this space, one of them will not
have any point in common with the line l infinitesimally close to the point p, while
the other will have two points in common with the line l infinitesimally close to the
point p.
4 Inthe case when the two points m and n lie in the same space of reversible adiabats, the first and
third regions coincide.
Let us take all the points on the line l analogous to p where the line l comes into
contact with a space of reversible adiabats without passing through it. Similarly, we
take all the points on the line l where the line l comes into contact with a space of
reversible adiabats without passing through it. We draw spaces of reversible adiabats
through all these points. Joined to the spaces E(m) and E(n), they divide the space
into a certain number of subregions.
All the spaces of reversible adiabats situated in the same subregion meet the line
l in the same number N of points, and the line l in the same number N of points. If
the subregion in question is a part of the second region, the two numbers N and N
are uneven; if the subregion at issue is situated in the first or the third region, N and
N are even or zero.
Let us consider a subregion contained in the second region. A space of reversible
adiabats E is drawn in this subregion, meeting the line l in (2k + 1) points. These
points are numbered in the order they occur on the line l moving from point m to
point n; let them be a1 , a2 , . . . , a2 k , a2 k+1 . The same space of reversible adiabats
meets the line l in (2 k + 1) points which are numbered in the order they occur on
the line l moving from point m to point n as a 1 , a 2 , . . . , a 2 k , a 2 k +1 .
Take a point b1 on the line l drawn from m towards n [349] infinitesimally later
than the point a1 , which is also in the same subregion as the point a1 . Through point
b1 we draw a space of reversible adiabats F. Given what we have assumed about
spaces of reversible adiabats, the space F will be infinitesimally close to the space
E without any points in common. The space F meets the line l in (2k + 1) points
b1 , b2 , . . . , b2 k close, respectively, to the points a1 , a2 , . . . , a2 k , a2 k+1 . And it meets
the line l in (2 k + 1) points b 1 , b 2 , . . . , b 2 k , b 2 k +1 close, respectively, to the
points a 1 , a 2 , . . . , a 2 k , a 2 k +1 .
The points ai and bi occur on the line l taken in the direction from m towards n,
in the following order:
a1 , b1 , b2 , a2 , a3 , b3 , . . . , b2 k , a2 k , a2 k+1 , b2 k+1 ,
which gives us (2k + 1) infinitesimal reversible changes with their representative

line between the spaces E and F. These changes are:
a1 b1 , b2 a2 , a3 b3 , . . . , b2 k a2 k, a2 k+1 b2 k+1 .
We adopt the following notations for these changes:
Change undergone by the system Initial temperature of the system Heat released by the system
a1 b1 ϑ1 dQ1
b2 a2 ϑ2 dQ2
a3 b3 ϑ3 dQ3
... ... ...
b2 k a2 k ϑ 2k dQ2 k
a2 k+1 b2 k+1 ϑ2 k+1 dQ2 k+1
The points a i and b i occur on the line l taken in the direction from m towards n
in the following order:
a 1 , b 1 , b 2 , a 2 , a 3 , b 3 , . . . , b 2 k , a 2 k , a 2 k +1 , b 2 k +1 ,
which gives us (2 k + 1) infinitesimal reversible changes [350] with their represen-

tative line between the spaces E and F. These changes are:
a 1 b 1 , b 2 a 2 , a 3 b 3 , . . . , b 2 k a 2 k , a 2 k +1 b 2 k +1
We adopt the following notations for these changes:
Change undergone by the system Initial temperature of the system Heat released by the system
a 1 b 1 ϑ 1 dQ 1
b 2 a 2 ϑ 2 dQ 2
a 3 b 3 ϑ 3 dQ 3
... ... ...
b 2 k a 2 k ϑ 2k dQ 2 k
a 2 k +1 b 2 k +1 ϑ 2 k +1 dQ 2 k +1
Consider the two changes a1 b1 and b2 a2 . An argument analogous to that which

furnished Equation (5) yields without difficulty
dQ1 dQ2
+ = 0.
F (ϑ1 ) F (ϑ2 )
Continuing in the same way up to the two changes a2 k−1 b2 k−1 and b2 k a2 k
yields
dQ2 k−1 dQ2 k
+ = 0.
F (ϑ2 k−1 ) F (ϑ2 k )
Similarly, it can be shown that
dQ 1 dQ

+
2 = 0,
F ϑ1 F ϑ2
dQ 3 dQ

+
4 = 0,
F ϑ3 F ϑ4
........................,
dQ 2 k −1 dQ
+
2 k = 0.
F ϑ2 k −1 F ϑ2 k
Finally, [351] the comparison between a2 k+1 b2 k+1 and a 2 k +1 b 2 k +1 yields the
equation
dQ2 k+1 dQ
= 2 k +1
F (ϑ2 k+1 ) F ϑ2 k +1
These several equations allow us to write
dQ1 dQ2 dQ2 k dQ2 k+1
+ + ... + + =
F (ϑ1 ) F (ϑ2 ) F (ϑ2 k ) F (ϑ2 k+1 )
dQ 1 dQ 2 dQ dQ

+
+ . . . +
2 k + 2 k +1 .
F ϑ1 F ϑ2 F ϑ 2 k F ϑ 2 k +1
An analogous equation can be written for all systems formed from two infinites-
imally close spaces of reversible adiabats drawn in the subregions considered.
Therefore, for the whole subregion considered,
dQ dQ
=
, (10)
F (ϑ) F ϑ

where the first sign ranges over all the elements of the change l situated in the
subregion considered, and the second ranges over all the elements of the change l
situated in the same subregion.
The same argument can be applied to each subregion within the second region,
furnishing an analogous equation, so that it may be supposed that Equation (10)
extends to the whole of the second region.
Consider now a subregion falling either in the first or in the third region. In this
subregion a space of reversible adiabats E is drawn meeting the line l in 2k points.
These points are numbered in the order they occur on the line l in moving from point
m to point n as a1 , a2 , . . . , a2 k .
Take a point b1 on the line l, drawn from m towards n, occurring infinitesimally
later than the point a1 , and which therefore occurs in the same subregion as the
point a1 . From point b1 a space of reversible adiabats F is drawn. In accordance
with what we have assumed [352] regarding spaces of reversible adiabats, the space
F will be everywhere infinitesimally close to the space E without intersecting it. The
space F intersects line l in 2k points b1 , b2 , . . . , b2 k−1 , b2 k , close, respectively, to the
points a1 , a2 , . . . , a2 k−1 , a2 k .
The points ai and bi occur on the line l in the following order:
a1 , b1 , b2 , a2 , a3 , b3 , . . . , a2 k−1 , b2 k−1 , b2 k , a2 k ,
which gives us 2k infinitesimal reversible changes with their representative line

between the spaces E and F. These changes are:
a1 b1 , b2 a2 , a3 b3 , . . . , a2 k−1 b2 k−1 , b2 k a2 k .
Consideration of these changes easily yields the equations
dQ1 dQ2
+ = 0,
F (ϑ1 ) F (ϑ2 )
dQ3 dQ4
+ = 0,
F (ϑ3 ) F (ϑ4 )
.....................,
dQ2 k−1 dQ2 k

+ = 0,
F (ϑ2 k−1 ) F (ϑ2 k )
which themselves give the equation
dQ1 dQ2 dQ2 k

+ + ... + = 0.
F (ϑ1 ) F (ϑ2 ) F (ϑ2 k )
An analogous equation can be written for every system formed by two infinites-
imally close spaces of reversible adiabats occurring in the same subregion under
consideration. Accordingly, for this whole subregion, we have
dQ
= 0, (11)
F (ϑ)

where the sign ranges over all the elements of the change l situated in the
subregion at issue.
For [353] each subregion falling within the first region or the third region, a simi-
lar argument can be applied furnishing an analogous equation, so that Equation (11)
can be extended to the sum of the first and the third regions.
The equation
dQ

= 0, (11a)
F ϑ

in which the sign ranges over all the elements of the change l situated in the first
or the third regions, is demonstrated similarly.
From Equations (10), (11) and (11a) it is easy to deduce the equation

dQ dQ
=
. (8)
l F (ϑ) l F ϑ
It remains to lift the last restriction.

We have accepted that neither the change l nor the change l contains a portion
of finite extent falling entirely within a space of reversible adiabats.
Suppose now that the change l does have certain finitely extended portions
λ1 , λ2 , . . . , falling entirely within the spaces of reversible adiabats E1 , E2 , . . . , and
8.4 A Property of Reversible Cycles 135
similarly, that the change l does have certain finitely extended portions λ 1 , λ 2 , . . . ,
falling entirely within the spaces of reversible adiabats E 1 , E 2 , . . ..
For all adiabatically reversible changes, it always holds that
dQ = 0.
Accordingly,

dQ dQ
= 0, = 0, . . . , (12)
λ1 F (ϑ) λ2 F (ϑ)
and also

dQ dQ

= 0,
= 0, . . . . (12a)
λ1 F ϑ λ2 F ϑ
dQ dQ
Therefore, [354] without changing the values of l F(ϑ) or of l F(ϑ )
, the por-
tions λ1 , λ2 , . . . , can be subtracted from l and the portions 1 2 , . . . , from l . It
λ , λ
then suffices to represent the spaces E1 , E2 , . . . , and the spaces E 1 , E 2 , . . . among
those which limit the subregions and then to reproduce the preceding argument to
show that
dQ dQ
=
, (13)
F (ϑ) F ϑ

where the first sign ranges over all the elements of the change l not in the portions
λ1 , λ2 , . . . , and the second ranges over all the elements of the change l not in the
portions λ 1 , λ 2 , . . . .
The Equations (12), (12a) and (13) together easily furnish the equality

dQ dQ
=
. (8)
l F (ϑ) l F ϑ
This equation is, then, not only subject to the restrictions mentioned in §1 but
also to the condition that it doesn’t hold, for all the points of a domain of finite
extension forming part of the space of α, β, . . . , λ, ϑ, that
Rα = 0, Rβ = 0, . . . , Rλ = 0. (9)
8.4 A Property of Reversible Cycles

Suppose that the reversible change l returns the system to its initial state. The state
n is therefore identical with the state m. Among the other reversible changes l
capable of leading the system from the same initial state to the same final state
may also be reckoned those cases where there is no change at all. Evidently, in
this case,

dQ

= 0.
l F ϑ
Equation [355] (8) then yields Clausius’ celebrated theorem:
For a reversible cycle,

dQ
= 0. (14)
F (ϑ)
Suppose in particular that the cycle is isothermal; the temperature ϑ remains

constant throughout the course of the cycle, the same holds for the function F(ϑ),
and Equation (14) can be written

dQ = 0,
or

Rα dα + Rβ dβ + . . . + Rλ dλ = 0.
In virtue of Equations (2), this equality becomes

∂U ∂U ∂U 1
dα + dβ + . . . + dλ − (A dα + B dβ + . . . + L dλ) = 0.
∂α ∂β ∂λ E
But for an isothermal cycle, we have

∂U ∂U ∂U
dα + dβ + . . . + dλ = 0.
∂α ∂β ∂λ
The preceding equation therefore becomes

(A dα + B dβ + . . . + L dλ) = 0. (15)
When a system traverses a reversible isothermal cycle, the external actions applied to the
system effect a total work equal to zero.
This theorem is due to J. Moutier. Clausius and J. Moutier made frequent use of
it. It may be remarked that it is exactly the same for systems for which Equations (9)
are satisfied.
8.5 Entropy 137
8.5 Entropy
[356] Equation (8) can be stated in the following way, designating the calorific
coefficients of the system at equilibrium by Rα , . . . , Rλ , C,
The curvilinear integral

1

Rα dα + Rβ dβ + . . . + Rλ dλ + C dϑ
l F (ϑ)
has a value which depends uniquely on the initial point
(α0 , β0 , . . . , λ0 , ϑ0 )
and the final point
(α1 , β1 , . . . , λ1 , ϑ1 ).
This theorem is known to be equivalent to the following:

There exists an infinity of uniform functions of the variables α, β, . . . , λ, ϑ, differing from
one another by a constant, which are such that
⎧
⎪ Rα ∂S
⎪
⎪ =− ,
⎪
⎪ F (ϑ) ∂α
⎪
⎪
⎪
⎪
⎪
⎪ Rβ ∂S
⎪
⎪ =− ,
⎪
⎪ F (ϑ) ∂β
⎪
⎨
..............., (16)
⎪
⎪
⎪
⎪
⎪
⎪ Rλ ∂S
⎪
⎪F =− ,
⎪
⎪ (ϑ) ∂λ
⎪
⎪
⎪
⎪
⎪
⎪ C ∂S
⎩ =− ,
F (ϑ) ∂ϑ
with S(α, β, . . . , λ, ϑ) designating any one of these functions.
This function has been called the entropy of the system by Clausius.
This theorem can be restated in abridged form in the following way:
For all infinitesimal reversible changes, we have
dQ
= dS. (17)
F (ϑ)
If [357] we now recall, first that the quantities
Rα , . . . , Rλ , C
do not depend on those among the variables α, β, . . . , λ, ϑ which determine the

absolute position of the system in space, and second, that those of the variations
δα, δβ, . . . , δλ, δϑ which only determine the absolute position of the system in space
do not figure in the expression
Rα δα + Rβ δβ + . . . + Rλ δλ + C δϑ,
then it is easily seen that the entropy does not depend on the absolute position in
space.
8.6 On the Case Excepted from the preceding

On the case set aside where, in all the points of a domain D constituting part of the
space of α, β, . . . , λ, ϑ it always holds that
Rα = 0, Rβ = 0, . . . , Rλ = 0 (9)
we consider whether the preceding theorems can be applied in this case.

In the first place, it is evident that they can be if only isothermal changes of the
system are considered. For then, for all reversible changes,
dQ = 0,
and consequently
dQ
= dS,
F (ϑ)
where S designates any function of the temperature.
This particular case is important because it is really this that is envisaged in
classical rational mechanics. In fact, in the classical exposition of rational mechan-
ics, the notion of temperature is omitted (this is only admissible if the temperature
is supposed invariable) and the notion of the quantity of heat [358] released
by the system is also omitted. This is admissible only if the equalities (9) are
satisfied.
In this case, Equations (2) and (9) yield5
∂U ∂U ∂U
A=E , B=E ,...,L = E . (18)
∂α ∂β ∂λ
5 [The second “E” is “A” in the original.]

8.6 On the Case Excepted from the preceding 139
This is the accepted form that is attributed in this case to the conditions of equi-
librium (1). The product EU represents what is called in mechanics the potential of
the internal forces.
But we no longer suppose that the temperature is invariable. We will show that, if
there exists a domain D where Equations (9) are satisfied, then in order for there to
exist a uniform and continuous function S of the variables α, β, . . . , λ, ϑ such that,
in every field of these variables,
⎧
⎪ Rα ∂S
⎪
⎪ =− ,
⎪
⎪ F (ϑ) ∂α
⎪
⎪
⎪
⎪
⎪
⎪ Rβ ∂S
⎪
⎪ =− ,
⎪
⎪ F (ϑ) ∂β
⎪
⎨
..............., (16)
⎪
⎪
⎪
⎪
⎪
⎪ Rλ ∂S
⎪
⎪ F (ϑ) = − ∂λ ,
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ C ∂S
⎩ =− ,
F (ϑ) ∂ϑ
it is necessary and sufficient that, in all of the domain D, the caloric capacity C is a
function of the temperature ϑ alone.
If the domain D does not occupy the entire field of the variables α, β, . . . , λ, ϑ
(which is the only case where the theorem is not evident), it is always possible to cut
the part of the field exterior to the domain D into simply connected spaces. In the
interior of each of these spaces, the preceding demonstrations are applicable. They
allow one to write equations analogous to Equations (16) which entail in their turn
equations of the type

∂ Rα ∂ Rβ
= (18)
∂β F (ϑ) ∂α F (ϑ)
and [359] equations of the type

∂ Rα ∂ C
= . (18a)
∂ϑ F (ϑ) ∂α F (ϑ)
On the other hand, in order that Equations (18) and (18a) can be written in
the domain D, where Equations (9) are satisfied, it is necessary and sufficient that
the quantity
dQ C
= dϑ
F (ϑ) F (ϑ)
be a total differential and, therefore, that the quantity C be a function of the

temperature ϑ alone.
Moreover, the field of variables α, β, . . . , λ, ϑ being supposed simply connected,
it is necessary and sufficient for the existence of a uniform and continuous function
S of the variables α, β, . . . , λ, ϑ that Equations (18) and (18a) are satisfied in the
whole field.
Thus, if we accept that, in the entire domain D where Equations (9) are satisfied,
the caloric capacity of the system is a function of the temperature alone, the theo-
rems demonstrated in the present chapter are subject only to the restrictions stated
in Section 8.1.
Part IV
of Thermodynamics: The General
Equations of Thermodynamics (1894)∗
∗ “Commentaire aux principes de la Thermodynamique. Troisième Partie: Les équations générales
de la Thermodynamique”, Journal de Mathématiques Pure et Appliquées, 10 (1894), 207–285.

See vol. VIII, p. 269, and vol. IX, p. 293 [i.e. Parts One and Two.]
Chapter 9
∗
Properties of a System at Equilibrium
9.1 The Internal Thermodynamic Potential
Let
U(α, β, . . . , λ, ϑ) and S(α, β, . . . , λ, ϑ)
be the internal energy and the entropy of a system. The system is, of course, [208]
subject to the restrictions necessary for the definition of entropy. In particular, the
temperature ϑ has the same value at all of its points.
Put
F (α, β, . . . , λ, ϑ) = E [U(α, β, . . . , λ, ϑ) − F (ϑ) S(α, β, . . . , λ, ϑ)] . (1)
The function F will, like the functions U and S, be a uniform and continuous
function of the variables α, β, . . . , λ, ϑ, and like the functions U and S, it will be
independent of any of these variables which serve solely to fix the position of the
system in absolute space.
Since the two quantities U and S are each determined up to a constant, the
function F is determined up to a function of the temperature, this function of the
temperature being of the form
C + C F(ϑ),

where C and C are two arbitrary constants.
This function F is a generalisation of one of the characteristic functions of
Massieu. It is the available energy of Gibbs and of Maxwell, and the freie Energie
of H. von Helmholtz. We call it the internal thermodynamic potential of the system.
∗
We have developed in detail the thermodynamic properties of systems at equilibrium in a recent
memoir: On the general equations of Thermodynamics (Annales de l’École Normale, 3rd series,
vol. VII, p. 231). In order to avoid as far as possible reproducing this memoir here, we have lim-
ited the present chapter to the exposition of properties necessary for understanding the following
chapters.

144 9 Properties of a System at Equilibrium
We have the following relations between the actions A, B, . . . , L, , which main-

tain the system in equilibrium, and the calorific coefficients of the system at
equilibrium, Rα , Rβ , . . . , Rλ , C (Section 8.2, Equation (2)):
⎧
⎪ ∂U A ∂U B
⎪
⎪ Rα = ∂α − E , Rβ = − ,
⎪
⎪ ∂β E
⎨
.............................., (2)
⎪
⎪
⎪
⎪
⎩ Rλ = ∂U − L ,
⎪
C=
∂U
− .
∂λ E ∂ϑ E
On [209] the other hand, we have (Chapter 8, Equations (16)),
⎧
⎪ ∂S ∂S
⎪
⎪ Rα = F (ϑ) , Rβ = F (ϑ) ,
⎪
⎪ ∂α ∂β
⎨
.............................., (3)
⎪
⎪
⎪
⎪ ∂S
⎩ Rλ = F (ϑ) ∂S ,
⎪
C = F (ϑ) .
∂λ ∂ϑ
Equations (1), (2) and (3) easily yield
⎧
⎪ ∂F
⎪
⎪ A= ,
⎪
⎪ ∂α
⎪
⎪
⎪
⎪ ∂F
⎪
⎪ B=
⎪
⎪ ∂β
,
⎪
⎨
..............., (4)
⎪
⎪
⎪
⎪ ∂F
⎪
⎪
⎪
⎪ L= ,
⎪
⎪ ∂λ
⎪
⎪
⎪
⎪ ∂F d F (ϑ)
⎩ = +ES .
∂ϑ dϑ
Necessary and sufficient conditions for the equilibrium of the system are obtained
by adding to these equations the condition that the bodies external to the system have
the same temperature as it has.
The last of Equations (4) yields

1 ∂F
ES = − (5)
d F (ϑ) dϑ ∂ϑ
Equations (1) and (5) yield

F (ϑ) ∂F
EU = F + − (6)
d F (ϑ) dϑ ∂ϑ
9.2 Properties of a System Formed from Several Independent Parts with the Same . . . 145
Equations [210] (3) and (5) yield

⎧
⎪
⎪ F (ϑ) ∂ ∂ 2F
⎪
⎪ R α = − ,
⎪
⎪ d F (ϑ) dϑ ∂α ∂ϑ∂α
⎪
⎪
⎪
⎪
⎪
⎪ F (ϑ) ∂ ∂ 2F
⎪
⎪ R = − ,
⎪
⎪ β
d F (ϑ) dϑ ∂β ∂ϑ∂β
⎪
⎪
⎨
...................................., (7)
⎪
⎪
⎪
⎪ F (ϑ) ∂ ∂ 2F
⎪
⎪
⎪
⎪ R λ = − ,
⎪
⎪ d F (ϑ) dϑ ∂λ ∂ϑ∂λ
⎪
⎪
⎪
⎪
⎪
⎪ F (ϑ) ∂ ∂ 2F F (ϑ) F (ϑ) ∂F
⎪
⎪ C = − − 2 − .
⎩ d F (ϑ) dϑ ∂ϑ ∂ϑ 2 ∂ϑ
F (ϑ)
Thus, if the following is known:
1. the expression for the internal thermodynamic potential of the system;

2. the expression
= fϑ (α, β, . . . , λ, ϑ)
for the quantity for the system at equilibrium,

then the following can be determined:
1. the internal energy and the entropy of the system in any state;
2. the necessary and sufficient conditions for equilibrium of the system;
3. the calorific coefficients of the system at equilibrium.
This is the generalisation of a proposition well known from Massieu.
9.2 Properties of a System Formed from Several Independent

Parts with the Same Temperature
In order not to unnecessarily complicate the arguments from the point of view of
generality, we will suppose that the system is formed from just two distinct parts,
independent of one another, which we designate by the indices 1 and 2. Each of
these two parts will be supposed to have the same [211] temperature at all of its
points. ϑ 1 is the temperature of part 1 and ϑ 2 the temperature of part 2. For the
moment, we will suppose these to be any temperatures whatsoever; presently, we
impose on them the condition of being equal to one another.
Let α1 , β1 , . . . , λ1 , ϑ1 be the independent variables which determine the state of
system 1, including its absolute position in space. The internal energy, the entropy
and the internal thermodynamic potential of system 1 are uniform functions of these
variables, or, at least, of those not serving solely to determine the absolute position
of the system in space. These three functions are designated, respectively, by
ϒ1 (α1 , β1 , . . . , λ1 , ϑ1 ),

(α1 , β1 , . . . , λ1 , ϑ1 ) ,
1
F1 (α1 , β1 , . . . , λ1 , ϑ1 ).
Similarly, let α2 , β2 , . . . , λ2 , ϑ2 be the independent variables which define the state

of system 2. Let
ϒ2 (α2 , β2 , . . . , λ2 , ϑ2 ),

(2 α2 , β2 , . . . , λ2 , ϑ2 ) ,
2
F2 (α2 , β2 , . . . , λ2 , ϑ2 ).
be the internal energy, the entropy and the internal thermodynamic potential of
system 2.
The state of the complex system (1, 2), formed from the aggregate of the two sys-
tems 1 and 2, and its absolute position in space, are determined when the collection
of variables
α1 , β1 , . . . , λ1 , ϑ1 , α2 , β2 , . . . , λ2 , ϑ2
is known. The internal energy of the system (1, 2) will therefore have the value
U = ϒ1 (α1 , β1 , . . . , λ1 , ϑ1 ) + ϒ2 (α2 , β2 , . . . , λ2 , ϑ2 )
(8)
+ X12 (α1 , β1 , . . . , λ1 , ϑ, α2 , β2 , . . . , λ2 , ϑ2 ),
where X12 is a uniform function of the variables as shown, or, [212] at least, of those
variables not serving solely to determine the absolute position of the system (1, 2)
in space. Further, it can be agreed that this function has the value 0 when the two
systems 1 and 2 are infinitely separated from one another.
If we designate the internal energy of the bodies external to the system (1, 2) by

U , the internal energy of the system formed from the system (1, 2) and the external
bodies will have the value
U = U + U + . (9)
The function will depend on the variables α1 , β1 , . . . , λ1 , ϑ1 ,

α2 , β2 , . . . , λ2 , ϑ2 , and also on the variables determining the state of the bodies
external to the system (1, 2).
The external actions exerted on the system (1, 2) will have the values (Chapter 5,
Equations (4))
⎧
⎪ ∂ ∂
⎪
⎪ A1 = −E , A2 = −E ,
⎪
⎪ ∂α ∂α2
⎪
⎪ 1
⎪
⎪
⎪
⎪ ∂ ∂
⎪ B1 = −E
⎪ , B2 = −E ,
⎪
⎪ ∂β ∂β2
⎨ 1
..............., ..............., (10)
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎪ L1 = −E L2 = −E
⎪
⎪ , ,
⎪
⎪ ∂λ 1 ∂λ2
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎩ W1 = −E , W2 = −E .
∂ϑ1 ∂ϑ2
On the other hand, the bodies external to the system 1 comprise:
1. The bodies external to the system (1, 2):
2. The bodies of [system] 2.
Accordingly, it is easily seen that the external actions applied to system 1 have
the values ⎧
⎪ ∂
⎪
⎪ A1 = −E (X12 + ) ,
⎪
⎪ ∂α1
⎪
⎪
⎪
⎪
⎪
⎪ ∂
⎪
⎪ B1 = −E (X12 + ) ,
⎪
⎪ ∂β1
⎨
........................, (11)
⎪
⎪
⎪
⎪ ∂
⎪
⎪
⎪ L1 = −E ∂λ1 (X12 + ) ,
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ∂
⎩ 1 = −E (X12 + ) .
∂ϑ1
Similarly, [213] the external actions applied to system 2 have the values
⎧
⎪ ∂
⎪
⎪ A2 = −E (X12 + ) ,
⎪
⎪ ∂α
⎪
⎪ 2
⎪
⎪
⎪
⎪ ∂
⎪
⎪ B2 = −E (X12 + ) ,
⎪
⎪ ∂β
⎨ 2
........................, (11a)
⎪
⎪
⎪
⎪ ∂
⎪
⎪ L2 = −E
⎪
⎪ (X12 + ) ,
⎪
⎪ ∂λ2
⎪
⎪
⎪
⎪ ∂
⎪
⎩ 2 = −E (X12 + ) .
∂ϑ2
All this is general. Suppose now that system (1, 2) is at equilibrium, and we apply
this principle, which evidently accords with the definition of equilibrium: in order
that the system (1, 2) be at equilibrium, it is necessary and sufficient that each of the
two systems 1 and 2 are at equilibrium.
In order that system 1 be at equilibrium, it is necessary and sufficient that, in the

first place,
ϑ1 = ϑ2 = ϑ, (12)
where ϑ is the temperature common to the bodies external to system (1, 2).
In the second place, in virtue of Equations (4) and (11),
⎧
⎪ ∂ ∂F1
⎪
⎪ E (X12 + ) = − ,
⎪
⎪ ∂α ∂α1
⎪
⎪ 1
⎪
⎪
⎪
⎪ ∂ ∂F1
⎪E
⎪ (X12 + ) = − ,
⎪
⎪ ∂β ∂β1
⎨ 1
..........................., (13)
⎪
⎪
⎪
⎪ ∂ ∂F1
⎪
⎪
⎪
⎪ E (X12 + ) = − ,
⎪
⎪ ∂λ ∂λ1
⎪
⎪
1
⎪
⎪ ∂ ∂F1
⎪
⎩E (X12 + ) = − − E F (ϑ1 ) 1 .
∂ϑ1 ∂ϑ1
In order that system 2 is at equilibrium, it is necessary and sufficient that:
In the first place, Equations (12) hold;
In [214] the second place, in virtue of Equations (4) and (11a),
⎧
⎪ ∂ ∂F2
⎪
⎪ E (X12 + ) = − ,
⎪
⎪ ∂α ∂α2
⎪
⎪ 2
⎪
⎪
⎪
⎪ ∂ ∂F2
⎪E
⎪ (X12 + ) = − ,
⎪
⎪ ∂β ∂β2
⎨ 2
..........................., (13a)
⎪
⎪
⎪
⎪ ∂ ∂F2
⎪
⎪
⎪E
⎪ (X12 + ) = − ,
⎪
⎪ ∂λ ∂λ2
⎪
⎪
2
⎪
⎪ ∂ ∂F2
⎪
⎩E (X12 + ) = − − E F (ϑ2 ) 2 .
∂ϑ2 ∂ϑ2
When Equations (12) are satisfied, the system (1, 2) has an internal ther-
modynamic potential that we designate by F and an entropy that we designate
by S.
In general, the virtual work of the external forces applied to this system is
expressed by1
A1 δα1 + B1 δβ1 + . . . + L1 δλ1 + W1 δϑ1
+ A2 δα2 + B2 δβ2 + . . . + L2 δλ2 + W2 δϑ2 ,
1 [A and L2 are, respectively, A1 and L1 in the original.]

2
where A1 , B1 , . . . , L1 , W1 , A2 , B2 , . . . , L2 , W2 , are given by the Equations (10). In

the particular case where the Equations (12) are constantly satisfied, the virtual work
becomes
A1 δα1 + B1 δβ1 + . . . + L1 δλ1
+ A2 δα2 + B2 δβ2 + . . . + L2 δλ2 + Wδϑ,
where W is defined by the equality

∂ ∂
W = W1 + W2 = −E + . (14)
∂ϑ1 ∂ϑ2 ϑ1 =ϑ2 =ϑ
Then, for equilibrium of the system (1, 2), it is necessary and sufficient that:
In the first place, Equations (12) hold;
In [215] the second place, in virtue of Equations (4), (10) and (14),
⎧
⎪ ∂ ∂F ∂ ∂F
⎪
⎪ E =− , E =− ,
⎪
⎪ ∂α1 ∂α1 ∂α2 ∂α2
⎪
⎪
⎪
⎪ ∂ ∂F ∂ ∂F
⎪
⎪
⎪
⎪ E =− , E =− ,
⎪
⎪ ∂β ∂β ∂β ∂β
⎨ 1 1 2 2
..............., ..............., (15)
⎪
⎪
⎪
⎪ ∂ ∂F ∂ ∂F
⎪
⎪
⎪
⎪ E =− , E =− ,
⎪
⎪ ∂λ 1 ∂λ 1 ∂λ 2 ∂λ 2
⎪
⎪
⎪
⎪ ∂ ∂ ∂F
⎪
⎩E + =− − E F (ϑ) S.
∂ϑ1 ∂ϑ2 ∂ϑ
The principle just stated requires that, given Equations (12), the collection of
Equations (15) is equivalent to the collection of equalities (13) and (13a). This
condition is expressed by the equations
⎧
⎪ ∂
∂

⎪
⎪ F − F1 − E X12 = 0, F − F2 − E X12 = 0,
⎪
⎪ ∂α1 ∂α2
⎪
⎪
⎪
⎪
∂

⎨ ∂ F − F1 − E X12 = 0,
⎪
F − F2 − E X12 = 0,
∂β1 ∂β2 (16)
⎪
⎪
⎪
⎪ ..........................., ........................,
⎪
⎪
⎪
⎪ ∂

⎪
⎪ ∂

⎩ F − F1 − E X12 = 0, F − F2 − E X12 = 0,
∂λ1 ∂λ2
⎧
⎨ ∂F − ∂ (F1 + E X12 ) − ∂ (F2 + E X12 )
⎪
∂ϑ ∂ϑ1 ∂ϑ2 (17)
⎪
⎩
+E F (ϑ) (S − 1 − 2 ) = 0.
The function F1 does not depend on the variables α2 , β2 , . . . , λ2 , ϑ2 , and the func-
tion F2 does not depend on the variables α1 , β1 , . . . , λ1 , ϑ1 . The equalities (16) can
therefore be written2 .
⎧ ∂
⎪
⎪ (F1 + F2 + E X12 − F) = 0,
⎪
⎪ ∂α1
⎪
⎪
⎪
⎪
⎪ ∂
⎪
⎪
⎪ (F1 + F2 + E X12 − F) = 0,
⎪
⎪ ∂β1
⎪
⎪
⎪
⎪
⎪
⎪ .................................,
⎪
⎪
⎨ ∂
(F + F2 + E X12 − F) = 0, (18)
⎪ ∂λ1 1
⎪
⎪
⎪
⎪
⎪ ∂
⎪
⎪ (F1 + F2 + E X12 − F) = 0,
⎪
⎪
⎪
⎪ ∂α2
⎪
⎪
⎪
⎪ .................................,
⎪
⎪
⎪
⎪
⎪
⎩ ∂ (F1 + F2 + E X12 − F) = 0.
⎪
∂λ2
In [216] virtue of the Equations (12), it holds that
∂F1 ∂F1 ∂F2 ∂F2

= , =
∂ϑ1 ∂ϑ ∂ϑ2 ∂ϑ
In virtue of the same equations, X12 can be regarded as a function of the variables
α1 , β1 , . . . , λ1 , α2 , β2 , . . . , λ2 , ϑ,
and so
∂X12 ∂X12 ∂X12
= + .
∂ϑ ∂ϑ1 ∂ϑ2
Equation (17) can therefore be written
∂
(F1 + F2 + E X12 − F) + E F (ϑ) (1 + 2 − S) = 0. (19)
∂ϑ
Equations (18) yield
F = F1 + F2 + E X12 + f(ϑ) , (20)
where f(ϑ) is an arbitrary function of the temperature ϑ. Equation (19) then yields
f (ϑ)
S = 1 + 2 − . (21)
E F (ϑ)
2 [“λ
1” is written “λ2 ” in the original.]
If it is agreed to take as the internal thermodynamic potential of a system formed from

two parts at the same temperature, infinitely separated from one another, the sum of the
internal thermodynamic potentials of the two parts, then it always holds, in accordance
with Equation (20), that
f (ϑ) = 0.
Equations (20) and (21) will therefore become
F = F1 + F2 + E X12 , (20a)
S = 1 + 2 . (21a)
When [217] a system is formed from several independent parts all at the same temperature,
the internal thermodynamic potential of the system is obtained by taking the sum of the
internal thermodynamic potential of the parts, and adding the determination of the potential
of the mutual actions of these parts which cancels when the parts are separated infinitely
far from one another.
The entropy of the system is equal to the sum of the entropies of the various parts.
These theorems are applied frequently in thermodynamics. They can be com-
pared with an analogous theorem concerning the internal energy demonstrated in
the 1st Part (Section 5.2). But the latter was general, whereas the theorems that
we have just demonstrated presuppose that the various parts of the system are at
the same temperature, and that each of them satisfies the conditions mentioned in
Chapter 8.
Equation (21a) entails an important new consequence:
The calorific coefficients of system 1 at equilibrium are
Rα1 , Rβ1 , . . . , Rλ1 , C1 .
The calorific coefficients of system 2 at equilibrium are
Rα2 , Rβ2 , . . . , Rλ2 , C2 .
The calorific coefficients of system (1, 2) at equilibrium are
ρα1 , ρβ1 , . . . , ρλ1 , ρα2 , ρβ2 , . . . , ρλ2 , γ .
We have, in virtue of Equations (3),
∂S ∂S
ρα1 = F (ϑ) , ρα2 = F (ϑ) ,
∂α1 ∂α2
..............., ...............,
∂S ∂S
ρλ1 = F (ϑ) , ρλ2 = F (ϑ) ,
∂λ1 ∂λ2
∂S
γ = F (ϑ) .
∂ϑ
Taking [218] into account Equation (21a) and noting that

∂ 1 ∂ 2
= 0, = 0,
∂α2 ∂α1
........., .........,

∂ 1 ∂ 2
= 0, = 0,
∂λ2 ∂λ1
we can write
∂ 1 ∂ 2
ρα1 = F (ϑ) , ρα2 = F (ϑ) ,
∂α1 ∂α2
.................., ..................,

∂ 1 ∂ 2
ρλ1 = F (ϑ) , ρλ2 = F (ϑ) ,
∂λ1 ∂λ2

∂ 1 ∂ 2
γ = F (ϑ) + ,
∂ϑ ∂ϑ
or, in virtue of Equations (3),
⎧
⎪
⎪ ρα1 = Rα1 , ρα2 = Rα2 ,
⎪
⎪
⎪
⎨........., ...............,
(22)
⎪
⎪ ρλ1 = Rλ1 , ρλ2 = Rλ2 ,
⎪
⎪
⎪
⎩
γ = C1 + C 2 .
The remarkable Equations (22) are subject to the same restrictions as

Equation (21a).
9.3 The Fundamental Hypothesis; Normal Variables

We will now invoke a fundamental hypothesis which is generally accepted implicitly
in treatises on thermodynamics.
Let an independent system 1, with the same temperature at all its points, be
defined by the variables α1 , β1 , . . . , λ1 , ϑ1 . Then let any other independent system 2,
without necessarily having the same temperature at all its points and being defined
by the variables α2 , β2 , . . . , λ2 , ϑ2 , constitute the bodies external to system 1.
The [219] internal energy U of the system (1, 2) can be written
U = ϒ1 (α1 , β1 , . . . , λ1 , ϑ1 ) + ϒ2 (α2 , β2 , . . . , λ2 )
+ (α1 , β1 , . . . , λ1 , ϑ1 , α2 , β2 , . . . , λ2 ),
where ϒ 1 is the internal energy of system 1 and ϒ 2 is the internal energy of
system 2.
9.3 The Fundamental Hypothesis; Normal Variables 153
The hypothesis that we wish to state is the following:

The variables α1 , β1 , . . . , λ1 , ϑ1 can be chosen so that:
1. When ϑ 1 varies, and α1 , β1 , . . . , λ1 retain constant values, no material point

of system 1 is displaced in space. The kinetic energy, T1 , of system 1 is then
independent of ϑ 1 and of dϑ1 /dt;
2. the function does not depend on the variable ϑ 1 , and whatever bodies compose
system 2.
When the variables α1 , β1 , . . . , λ1 have been chosen in that way, we will say that
the system α1 , β1 , . . . , λ1 constitutes a system of normal variables defining system 1.
The virtual work of the external actions to which system 1 is subject is an
expression of the form
A1 δα1 + B1 δβ1 + . . . + L1 δλ1 + 1 δϑ1 ,
In accordance with Equations (4) of Chapter 5,
∂
1 = −E .
∂ϑ1
If, therefore, the system of variables α1 , β1 , . . . , λ1 is a system of normal
variables, then
1 = 0.
If a system with the same temperature at all of its points is defined by normal variables,
then whatever the bodies external to the system, the actions they exert on the system do not
perform any work when the temperature alone varies.
This [220] proposition does not presuppose that the system is at equilibrium.
What are the properties of a system at equilibrium when it is defined by normal
variables?
For such a system, the quantity
= fϑ (α, β, . . . , λ, ϑ)
is always zero.
Consequently, Equations (4), (5), (6) and (7) become
∂F ∂F ∂F
A= , B= , ..., L= (4a)
∂α ∂β ∂λ
∂F
E F (ϑ) S = − (5a)
∂ϑ
F (ϑ) ∂F
EU = F+ (6a)
F (ϑ) ∂ϑ
⎧
⎪
⎪ F (ϑ) ∂ 2 F
⎪
⎪ R α = − ,
⎪
⎪
⎪ E F (ϑ) ∂α∂β
⎪
⎪
⎪
⎪ F (ϑ) ∂ 2 F
⎪
⎪ Rβ = −
⎪
⎪ E F (ϑ) ∂β∂ϑ
,
⎪
⎪
⎨
........................, (7a)
⎪
⎪
⎪
⎪ F (ϑ) ∂ 2 F
⎪
⎪ = −
⎪
⎪ R λ ,
⎪
⎪ E F (ϑ) ∂λ∂ϑ
⎪
⎪
⎪
⎪ F (ϑ) ∂ 2 F F (ϑ) F (ϑ) ∂F
⎪
⎪
⎪C = −
⎩ + 2 .
E F (ϑ) ∂ϑ 2 E F (ϑ) ∂ϑ
Thus, when a system is defined by normal variables, it suffices to know the internal
thermodynamic potential of the system in order to be able to determine the equations
of equilibrium, the energy, the entropy and the calorific coefficients. This is a well-
known proposition of F. Massieu.
Henceforth, when we consider a system all the points of which are at the same
temperature, we will suppose it to be defined by normal variables.
9.4 The Problem of Equilibrium

How, in general, is the problem [221] of equilibrium of a system posed when the
expression for the internal thermodynamic potential of the system is known?
The state of the bodies external to this system is supposed given. These bodies
have a uniform temperature θ , and it is proposed to determine the state taken by the
system under the action of these bodies.
The function , considered in the preceding section, becomes a function of
the variables α, β, . . . , λ alone when the state of the bodies external to the sys-
tem is determined, where α, β, . . . , λ, ϑ designate the normal variables defining the
system.
In virtue of Equations (4) (Chapter 5) and Equations (4a) of the present chapter,
⎧
⎪
⎪ ∂ ∂F
⎪
⎪ E + = 0,
⎪
⎪ ∂α ∂α
⎪
⎪
⎪
⎨ E ∂ + ∂F = 0,
⎪
∂β ∂β (23)
⎪
⎪
⎪
⎪ ...............,
⎪
⎪
⎪
⎪
⎪
⎪ ∂ ∂F
⎩E + = 0.
∂λ ∂λ
9.4 The Problem of Equilibrium 155
If the equation
ϑ = θ, (24)
is added, expressing the equality of the temperature of the system with that of the
external bodies, a number of equations is obtained equal to the number of the vari-
ables α, β, . . . , λ, ϑ whose values we wish to determine. These equations therefore
allow the determination of the state of equilibrium of the system, provided that the
expression for the quantities
∂ ∂ ∂
, ,..., ,
∂α ∂β ∂λ
is known as a function of α, β, . . . , λ, that is, provided that it is known how the
external actions applied to the system vary with the state of the system. Moreover,
if the variables α, β, . . . , λ, ϑ are normal variables, these actions do not depend on
the temperature of the system.
Once [222] the state of the system is known, Equations (5a), (6a) and (7a) tell us
the energy, the entropy and the calorific coefficients.
Chapter 10
Properties of a System in Motion
10.1 Meaning of the Word “Motion”

In this chapter, we take the word motion to designate not only a change of position
in space, but also any change of state, even if not accompanied by any displace-
ment. Thus, there would be motion if only the variables that we have designated
by a, b, . . . , l (Section 5.4) vary, the variables α, β, . . . , λ retaining fixed values.
Consequently, the word motion does not stand in opposition to the word rest, but to
the word equilibrium.
10.2 Motion of a System with the Same Temperature

at All Points
Imagine a system with the same temperature at all of its points, this temperature
not necessarily being the same as that of the external bodies. This system is not at
equilibrium; it remains to determine the laws of its motion.
At each instant t, the state of the external bodies is supposed given. The external
actions which affect the system are, in general, functions of the variables α, β, . . . , λ
defining the system and the variables defining the external bodies. The latter are
given functions of t, and the external actions are, to be precise, given functions of
α, β, . . . , λ, t which we designate by
A (α, β, . . . , λ, t),
B (α, β, . . . , λ, t),
...............,
L (α, β, . . . , λ, t).
For [223] brevity, we put
∂α ∂β ∂λ
α = , β = , . . . , λ = .
∂t ∂t ∂t

158 10 Properties of a System in Motion
When the system is at equilibrium, we have (Chapter 9, Equation (23))
∂F ∂F ∂F
A − = 0, B − = 0, . . . , L − = 0.
∂α ∂β ∂λ
On the other hand, it holds identically that
∂T d ∂T ∂T d ∂T ∂T d ∂T
− = 0, − = 0, . . . , − = 0.
∂α dt ∂α ∂β dt ∂β ∂λ dt ∂λ
Therefore, in the case when the system is at equilibrium,

⎧
⎪ ∂F ∂T d ∂T
⎪
⎪ A − + − = 0,
⎪
⎪ ∂α ∂α dt ∂α
⎪
⎪
⎪
⎨ B − ∂F + ∂T − d ∂T = 0,
⎪
∂β ∂β dt ∂β (1)
⎪
⎪
⎪
⎪ ...........................,
⎪
⎪
⎪
⎪
⎩ L − ∂F + ∂T − d ∂T = 0.
⎪
∂λ ∂λ dt ∂λ
When the system is not at equilibrium, it is not certain that these equations are
satisfied. The left-hand sides could have values differing from 0; we designate these
values by −fα , −fβ , . . . , −fλ , and can write, without any hypothesis,
⎧
⎪ ∂(T − F) d ∂T
⎪
⎪ − + A + fα = 0,
⎪
⎪ ∂α dt ∂α
⎪
⎪
⎪
⎨ ∂(T − F) − d ∂T + B + fβ = 0,
⎪
∂β dt ∂β (2)
⎪
⎪
⎪
⎪ ..............................,
⎪
⎪
⎪
⎪
⎩ ∂(T − F) − d ∂T + L + f = 0.
⎪
λ
∂λ dt ∂λ
The quantities fα , fβ , . . . , fλ , will be called the passive resistances [224] that the
system has to overcome. The quantity
fα dα + fβ dβ + . . . + fλ dλ
is called the elementary work of the passive resistances.

So far, we have said, the Equations (2) are in no way hypothetical. They only
take on a hypothetical character when we assign a particular form to the passive
resistances. With this in mind, we now make some assumptions:
FIRST CONVENTION.—The passive resistances fα , fβ , . . . , fλ depend uniquely on the
variables
10.3 Calorific Coefficients of a System in Motion 159
α, β, . . . , λ, ϑ,
∂α ∂β ∂λ
α = , β = , . . . , λ =
∂t ∂t ∂t
relative to the system, and the analogous variables relative to the external bodies.
SECOND CONVENTION.—The passive resistances do not change their values if the absolute
position or the absolute motion of the complex system formed by the system under study and
the bodies external to it change. They depend only on the relative position of the parts of the
system and the external bodies, and on the relative motion of the parts of the system and the
external bodies.
Like all the hypotheses we have made, these assumptions, although very natural,
are not logically necessary. They can only be verified by noting the agreement of
the consequences of Equations (2) with the facts of experience.
When the state of the external bodies is given at each instant t, the passive
resistances become functions of the variables
α, β, . . . , λ, ϑ, α, β , . . . , λ , t.
Equations (2) then become second order differential equations which would
determine the values of the variables α, β, . . . , λ, ϑ [225] as a function of t and,
therefore, the motion of the system, if they were sufficiently many. But the number
of variables whose value must be determined at each instant exceeds by one the num-
ber of equations of motion provided by thermodynamics. It is therefore necessary, in
order to complete the formulation of the equations of the problem, to borrow a last
equation from physical theory foreign to thermodynamics, such as, for example, the
equation
ϑ = ϕ(t),
which tells us the temperature of the system at each instant.
10.3 Calorific Coefficients of a System in Motion

The calorific coefficients of a system whatever its state of motion are defined by the
equations (Chapter 5, Equations (12))
⎧
⎪ ∂U ∂T d ∂T
⎪
⎪
⎪ E − + − A = ERα ,
⎪
⎪ ∂α ∂α dt ∂α
⎪
⎪
⎪
⎨....................................,
∂U ∂T d ∂T (3)
⎪
⎪ E − + − L = ERλ ,
⎪
⎪ ∂λ ∂λ ∂λ
⎪
⎪ dt
⎪
⎪
⎩ E ∂U = EC .
⎪
∂ϑ
But we have (Chapter 9, Equations (1))
E U = F + E F (ϑ) S.
Equations (3) can therefore be written

⎧
⎪ ∂F ∂T d ∂T ∂S
⎪
⎪
⎪ ∂α
−
∂α
+
∂α + E F (ϑ) ∂ϑ − A = ERα ,
⎪
⎪ dt
⎪
⎪
⎪
⎨.............................................,
⎪ ∂F ∂T d ∂T ∂S
⎪
⎪ − + + E F (ϑ) ∂ϑ − L = ERλ ,
⎪
⎪ ∂λ ∂λ dt ∂λ
⎪
⎪
⎪
⎩ ∂F + E F (ϑ) ∂S + E F (ϑ) S = EC ,
⎪
∂ϑ ∂ϑ
or, [226] taking into account Equations (2) of the present chapter and the
Equation (5a) of Chapter 9,
⎧
⎪ ∂S
⎪
⎪ ERα = E F (ϑ) + fα ,
⎪
⎪ ∂ϑ
⎪
⎪
⎪
⎪
⎨........................,
⎪ ∂S (4)
⎪
⎪ ERλ = E F (ϑ) + fλ ,
⎪
⎪ ∂ϑ
⎪
⎪
⎪
⎪ ∂S
⎩ EC = E F (ϑ) .
∂ϑ
Let Rα , . . . , Rλ , C be the calorific coefficients of the system at equilibrium in the
state (α, β, . . . , λ, ϑ). Comparing the Equations (4) that we have just written with
Equations (3) of Chapter 9, we have1
⎧
⎪ fα
⎪
⎪ R α = R α + ,
⎪
⎪ E
⎪
⎪
⎨.........,
(5)1
⎪
⎪ fλ
⎪
⎪
⎪ R λ = Rλ + ,
⎪
⎪ E
⎩
C = C.
The specific heat of a system taken in a determinate state is the same whether the
system is at equilibrium or in motion. If a variable other than the temperature is con-
sidered, the calorific coefficient corresponding to this variable does not have the same
value, depending on whether the system is at equilibrium or in motion; the second value
exceeds the first by a quantity equivalent to the passive resistance corresponding to that
variable.
1 [“f
λ” is “fβ ” in the original.]
10.4 Fundamental Property of the Passive Resistances 161
10.4 Fundamental Property of the Passive Resistances

Let A, B, . . . , L be external actions maintaining the system at equilibrium in the
state (α, β, . . . , λ, ϑ). These forces are given by Equations (4a) of Chapter 9.
Equations [227] (2) can be written
⎧
⎪ ∂T d ∂T
⎪
⎪ A − A = − + − fα ,
⎪
⎪ ∂α dt ∂α
⎪
⎪
⎪
⎨ B − B = − ∂T + d ∂T − fβ ,
⎪
∂β dt ∂β (6)
⎪
⎪
⎪
⎪ ...........................,
⎪
⎪
⎪
⎪
⎩ L − L = − ∂T + d ∂T − f .
⎪
λ
∂λ dt ∂λ
We call the quantities (A − A), (B − B), . . . , (L − L) the effective actions exerted
on the system, and the quantity
(A − A)dα + (B − B)dβ + . . . + (L − L)dλ
is called the effective work.

Multiplying Equations (6) by, respectively, dα, dβ, . . . , dλ, and adding the left-
hand sides and right-hand sides yields
(A − A)dα + (B − B)dβ + . . . + (L − L)dλ

(7)
= dT − (fα dα + fβ dβ + . . . + fλ dλ).
With this equality established, we now make the following fundamental

hypothesis:
HYPOTHESIS.—The effective work of the external actions on a system is, in any real change,
at least equal to the increase in the kinetic energy.
In accordance with this hypothesis, in any real transformation,
(A − A)dα + (B − B)dβ + . . . + (L − L)dλ ≥ dT,
and consequently, in accordance with Equation (7),
fα dα + fβ dβ + . . . + fλ dλ ≤ 0. (8)
The preceding hypothesis is therefore equivalent to the following proposition:

In any real change of a system, the work of the passive resistances is either zero or negative.
Multiplying [228] in the same way the two sides of Equations (4) by
dα, dβ, . . . , dλ, dϑ, and adding the two sides of the equations obtained yields

E Rα dα + Rβ dβ + . . . Rλ dλ + C dϑ
= E F(ϑ)dS + fα dα + fβ dβ + . . . + fλ dλ,
or, designating by dQ the quantity of heat released by the system in an elementary
transformation,

E dQ = − E F(ϑ) dS − fα dα + fβ dβ + . . . + fλ dλ . (9)
We write a similar equation for all the elements of a closed cycle and add the left-
hand and right-hand sides of these equations after dividing the two sides of each by
F(ϑ). Since the integral

dS,
taken over a closed cycle, is equal to 0, we have

dQ 1 fα dα + fβ dβ + . . . + fλ dλ
=−
F (ϑ) E F (ϑ)
Since the quantity (fα dα + fβ dβ + . . . + fλ dλ) can never be positive, it is clear
that, for any real closed cycle,

dQ
≥ 0. (10)
F (ϑ)
This famous inequality is due to Clausius.
Clausius has called the quantity
− (fα dα + fβ dβ + . . . + fλ dλ),
which is equal to the work of the passive resistances but with changed sign, and
which, therefore, is not negative in any real change of the system, the uncompen-
sated work done during this [229] change. By contrast, the quantity E F(ϑ)dS is for
him the compensated work done during this same change.
In accordance with Equations (2), we have

− fα dα + fβ dβ + . . . + fλ dλ = A dα + B dβ + . . . + L dλ

∂F ∂F ∂F
− dα + dβ + . . . + dλ − dT
∂α ∂β ∂λ
or, if dW designates the work of the actions external to the system,

∂F
− fα dα + fβ dβ + . . . + fλ dλ = dW − dT − dF + dϑ. (11)
∂ϑ
In the case where the change is isothermal, dϑ is equal to 0 and Equation (11)
becomes

− fα dα + fβ dβ + . . . + fλ dλ = dW − dT − dF. (11a)
10.5 Systems Without Viscosity 163
In order to calculate the uncompensated work done in an isothermal change, take the exter-
nal work and deduct from it the sum of the increases in the kinetic energy and the internal
thermodynamic potential. This theorem has many applications.
We apply Equation (9) to the transformations in a system isolated in space. For

such a system we have, by definition,
dQ = 0.
Equation (9) therefore becomes
dS = − .
E F (ϑ)
The right-hand side cannot be negative for any real change of the system.
Therefore, no real transformation of an isolated system can decrease the entropy
of the system. This proposition is said to be due to Clausius. He has demonstrated it
only in the case where the kinetic energy of the system is equal to 0.
10.5 Systems Without Viscosity

In [230] each particular case the hypotheses already introduced in connection with
passive resistances must be supplemented with additional hypotheses appropriate to
the determination of the form of the functions fα , fβ , . . . , fλ .
Among these hypotheses, the simplest that can be made, and the one that is
always tried, at least as a first approximation, consists in putting
fα = 0, fβ = 0, . . . , fλ = 0. (12)
When Equations (12) are assumed to be satisfied, the system is said to be without
viscosity.
In order to study the properties of systems without viscosity, we distinguish the
variables α, β, . . . , λ, on which the kinetic energy depends, from the variables a, b,
. . . , l, on which it doesn’t. The temperature, ϑ, should be added to the latter.
Given the Equations (12), Equations (2) become
⎧

⎪
⎪ ∂ T−F d ∂T
⎪
⎪ − + A (α, β, . . . , λ, a, b, . . . , l, t) = 0,
⎪
⎨ ∂α dt ∂α
......................................................, (13)
⎪
⎪

⎪
⎪∂ T − F d ∂T
⎪
⎩ − + L (α, β, . . . , λ, a, b, . . . , l, t) = 0,
∂λ dt ∂λ
⎧
⎪ ∂F
⎪
⎪ − + A (α, β, . . . , λ, a, b, . . . , l, t) = 0,
⎪
⎨ ∂α
......................................., (13a)
⎪
⎪
⎪ ∂F
⎪
⎩− + L (α, β, . . . , λ, a, b, . . . , l, t) = 0.
∂λ
These equations are fewer in number than the variables to be determined, and
determining the motion of the system requires the addition of a further equation
furnished by special hypotheses.
These equations are of two quite distinct kinds. Equations (13a), which corre-
spond to the variables a, b, . . . , l, are equations in the ordinary sense of the word.
Equations (13), on the other hand, which correspond to the variables α, β, . . . , λ,
are second-order partial differential equations of the kind given by Lagrange. [231]
The supplementary equation, which corresponds to the variable θ , can clearly be of
a third kind.
In the case where this supplementary equation makes the temperature a known
function of the time, which occurs, in particular, in the case when the temperature
is constant, the problem of the motion of the system can be treated as follows.
Equations (13a) are solved for a, b, . . . , l, as functions of α, β, . . . , λ, ϑ, t,
and since ϑ is known as a function of t, they are known as functions of
α, β, . . . , λ, t. If the values of the variables a, b, . . . , l are related in the function
F(α, β, . . . , λ, a, b, . . . , l), these will be transformed to a function of the variables
α, β, . . . , λ, t, which we designate by
G(α, β, . . . , λ, t).
Similarly, the functions
A (α, β, . . . , λ, a, b, . . . , l, t),
........................,
L (α, β, . . . , λ, a, b, . . . , l, t)
become functions of the variables α, β, . . . , λ, t, which we designate by
A (α, β, . . . , λ, t),
.........,
L (α, β, . . . , λ, t).
The variables α, β, . . . , λ will then be determined as a function of t by the
equations
⎧
⎪
⎪ ∂ d ∂T ∂T
⎪
⎪
⎪ ∂α
G (α, β, . . . , λ, t) +
∂α − ∂α = A (α, β, . . . , λ, t) ,
⎨ dt
......................................................, (14)
⎪
⎪
⎪ ∂
⎪ d ∂T ∂T
⎪
⎩ G (α, β, . . . , λ, t) + − = L (α, β, . . . , λ, t) ,
∂λ dt ∂λ ∂λ
which have the most general form of the classical equations of dynamics.
The [232] method that we have just outlined is used to study the motion of bodies
which are good electric conductors, bodies which are perfectly softly magnetic, etc.
10.5 Systems Without Viscosity 165
In the case of the classic problem of dynamics, the question is raised in a yet
more simple form. In dynamics, the variable ϑ is not introduced. It is implicitly
assumed that the temperature remains invariable. The other variables introduced all
figure in the expression for the kinetic energy. The problem is therefore reduced to
the integration of the equations
⎧ ∂ d ∂W ∂W
⎪
⎪ F (α, β, . . . , λ) + − = A (α, β, . . . , λ, t) ,
⎪
⎪
⎨ ∂α dt ∂α ∂α
..................................................., (15)
⎪
⎪
⎪
⎩ ∂ F (α, β, . . . , λ) + d ∂W − ∂W = L (α, β, . . . , λ, t) ,
⎪
∂λ dt ∂λ ∂λ
which differ from the Equations (14) in so far as the function F does not depend
explicitly on the temperature, whereas the function G does.
The classical problem of dynamics is thus logically connected by way of being
a special case of thermodynamics. We have indicated (Section 5.4) a method which
was presented for bringing about this reconciliation and we have shown the limita-
tions of this method. We now see that this method leads, in general, to incorrect
results. In fact, it would lead to the replacement in Equations (15) of the inter-
nal thermodynamic potential, F, by the product EU of the internal energy and the
mechanical equivalent of heat, and these two functions are not ordinarily equal to
one another.
There is, however, one case, particularly in theory, but virtually the only one
treated in classical dynamics, and mainly in celestial mechanics, where the two
methods lead to identical results.
Consider a system formed from a certain number of independent parts 1, 2, . . .
n, each one of which can be displaced in space, but remains in the same invariable
state. Let ϒ 1 , ϒ 2 , . . . , ϒ n be the internal energies of these various parts, which are
constant. Let F1 , F2 , . . . , Fn be the internal thermodynamic potentials of these parts,
[233] which are also constant. We know, on the one hand (Section 5.2), that
U = ϒ1 + ϒ2 + . . . + ϒn +
and, on the other hand (Section 9.2), that
F = F1 + F2 + . . . + Fn + E ,
where E is one of the determinations of the potential of mutual actions of the

parts 1, 2, . . . n, which is annulled when the parts are infinitely separated from
one another. The three quantities F, E U and E differ from one another only by
constants.
Thus, to write the equations of motion of a system formed from independent parts
which are displaced without change of state, the following three functions can be
substituted for one another in Lagrange’s equations:
The internal thermodynamic potential,

The product of the internal energy and the mechanical equivalent of heat,
The potential of the mutual actions of the various parts of the system.
Apart from this particular case, such a substitution would in general entail an
error—an error which is frequently committed.
Returning to the general properties of systems without viscosity, Equations (4)
become for such a system
⎧
⎪
⎪ ∂S ∂S
⎨ Rα = F(ϑ) , . . . , Rλ = E F (ϑ) ,
∂α ∂λ
(16)
⎪
⎪ ∂S
⎩ C = F (ϑ) .
∂ϑ
In a system without viscosity, the calorific coefficients have the same value, whether the
system is at equilibrium or in motion.
From [234] Equations (16) it can be deduced that
1

Rα dα + . . . + Rλ dλ + C dϑ = dS,
F (ϑ)
or, designating by dQ the quantity of heat released in an elementary change,
dQ
= −dS
F (ϑ)
Integrating this equation over a closed cycle leads to the following proposition:
When a system without viscosity traverses a real closed cycle, then for the entire cycle,

dQ
=0
F (ϑ)
We apply Equations (16) to a system formed of n independent parts which

are displaced in relation to one another while the state of the system remains
unchanged.
Designating
the entropies of various parts, which are constant, by
1 , 2 , . . . , n , we have (Chapter 9, Equations (21a))
S = 1 + 2 + . . . + n ;
the entropy of the system will also be constant. Equations (16) then yield for us the
When a system is formed from independent parts which are displaced without change of
state, all the calorific coefficients of the system are equal to 0.
10.6 Systems Formed of Independent Parts with Different Temperatures 167
We can see from this theorem why the method presented in the 1st Part
(Section 5.4) for connecting the classical problem of dynamics and thermodynamics
is legitimate in this special case.
In order to bring to an end the consideration of systems without viscosity, [235]
we state a CONJECTURE. It seems probable to us that the following supposition
might hold for all physical phenomena:
The passive resistances corresponding to the variables a, b, . . . , l not figuring in the
expression of the kinetic energy are always equal to 0.
In that case, it would be possible to replace Equations (2) in all circumstances

with the equalities
⎧ ∂(F − W) d ∂W
⎪
⎪ − = A + fα ,
⎪
⎪ ∂α dt ∂α
⎨
.............................., (17)
⎪
⎪
⎪
⎩ ∂(F − W) − d ∂W = L + f ,
⎪
λ
∂λ dt ∂λ
⎧ ∂F
⎪
⎪ = A ,
⎪
⎪
⎨ ∂α
........., (18)
⎪
⎪
⎪
⎪
⎩ ∂F = L .
∂λ
This will be the general form of Thermodynamic equations for a system of uni-
form temperature. We repeat that we have given this hypothesis the status of a
conjecture; the developments to follow are independent of it.
10.6 Systems Formed of Independent Parts

with Different Temperatures
We will now examine a case much more general than all the preceding ones. We will
study a system formed from any number of parts independent of one another. We
will assume that the temperature of each of these parts is uniform at each instant,
while being capable of varying from one instant to another. We will not suppose that
this temperature has the same value for all these parts at each instant.
In order not to complicate the discussion unnecessarily, we will consider just two
independent parts, which we designate with the indices 1 and 2.
Let [236] α1 , β1 , . . . , λ1 , ϑ1 be the normal variables defining the state of part 1
and its absolute position in space. Let α2 , β2 , . . . , λ2 , ϑ2 be the normal variables
defining the state of part 2 and its absolute position in space.
The internal energy of the system will have a value

U = ϒ1 (α1 , β1 , . . . , λ1 , ϑ1 ) + ϒ2 (α2 , β2 , . . . , λ2 , ϑ2 )
(19)
+ X12 (α1 , β1 , . . . , λ1 , ϑ1 , α2 , β2 , . . . , λ2 , ϑ2 ) .
Let
A1 dα1 + B1 dβ1 + . . . + L1 dλ1 + A2 dα2 + B2 dβ2 + . . . + L2 dλ2
be the virtual work of the external actions acting on the system.

Each of the two parts 1 and 2 are regarded as independent systems. The vir-
tual work of the external actions on each of the two systems is represented by the
expressions
A1 dα1 + B1 dβ1 + . . . + L1 dλ1 ,

A2 dα2 + B2 dβ2 + . . . + L2 dλ2 ,
with
⎧
⎪ ∂X12 ∂X12
⎪
⎪
⎪ A1 = A1 − E , A2 = A2 − E ,
⎪
⎨ ∂α 1 ∂α2
..............., ....................., (20)
⎪
⎪
⎪
⎪ ∂X ∂X12
⎪
⎩ L1 = L1 − E 12 , L2 = L2 − E .
∂λ1 ∂λ2
Part 1 forms an independent system with uniform temperature, this temperature

not necessarily being the same as that of the external bodies. The considerations
developed in the preceding sections can be applied to this system, which has an
internal thermodynamic potential F1 and, at each instant,
⎧

⎪ ∂ F1 − T 1 d ∂T1
⎪
⎪ + = A1 + fα1 ,
⎪
⎪ ∂α ∂α1
⎪
⎨ 1 dt
.............................., (21)
⎪
⎪ ∂
F − T
⎪
⎪
⎪
⎪ 1 1 d ∂T1
⎩ + = L1 + fλ1 .
∂λ1 dt ∂λ1
Similarly, [237] letting F2 designate the internal thermodynamic potential of

system 2,
⎧

⎪ ∂ F2 − T 2 d ∂T2
⎪
⎪ + = A2 + fα2 ,
⎪
⎪ ∂α2 dt ∂α2
⎪
⎨
.............................., (21a)
⎪
⎪

⎪
⎪ ∂ F 2 − T2
⎪
⎪ d ∂T2
⎩ + = L2 + fλ2 .
∂λ2 dt ∂λ2
10.6 Systems Formed of Independent Parts with Different Temperatures 169
Taking Equations (20) into account, we observe that F1 doesn’t depend on

α2 , β2 , . . . , λ2 , that F2 doesn’t depend on α1 , β1 , . . . , λ1 , that T1 doesn’t depend on
α2 , β2 , . . . , λ2 , α2 , β2 , . . . , λ2 , nor does T2 depend on α1 , β1 , . . . , λ1 , α1 , β1 , . . . , λ1 .
We put
T = T1 + T2 , (22)
F = F1 + F2 + EX12 , (23)
and Equations (21) and (21a) can be replaced by the equations

⎧

⎪
⎪ ∂ F−T d ∂T
⎪
⎪ + = A1 + fα1 ,
⎪
⎪ ∂α dt ∂α
⎪
⎪ 1 1
⎪
⎪
⎪
⎪ . . . . . . . . . . . . . . . . . . ............,
⎪
⎪

⎪
⎪
⎪
⎪ ∂ F−T d ∂T
⎪
⎪ + = L1 + fλ1 ,
⎨ ∂λ1 dt ∂λ1

(24)
⎪
⎪ ∂ F−T d ∂T
⎪
⎪ +
⎪ ∂α
⎪ ∂α = A2 + fα2 ,
⎪
⎪ 2 dt
⎪
⎪
2
⎪
⎪ . . . . . . . . . . . . . . . . . . ............,
⎪
⎪
⎪
⎪

⎪
⎪ ∂ F−T d ∂T
⎪
⎪ + = L2 + fλ2 .
⎩
∂λ2 dt ∂λ2
These Equations (24) are all entirely of the same form as Equations (2) which govern
the motion of a system with uniform temperature. If the system is composed of n
independent parts, then they are 2 less in number of the variables whose values must
be determined as a function of time. Two equations must therefore be added drawn
from considerations outside thermodynamics, such as, for example, [238] the two
equations
ϑ1 = ϕ1 (t), ϑ2 = ϕ2 (t), (25)
which tell us, at each instant, the temperature of each of the two independent parts
composing the system.
We now calculate the quantity of heat dQ released by the system during an
elementary change.
We have (Section 5.5)
dQ = dQ1 + dQ2 ,
where dQ1 and dQ2 are quantities of heat released by parts 1 and 2 during the same
change.
Let S1 and S2 be the entropies of parts 1 and 2. We have (Equation (9)),

E dQ1 = −E F (ϑ1 ) dS1 − fα1 dα1 + fβ1 dβ1 + . . . + fλ1 dλ1 ,

(26)
E dQ2 = −E F (ϑ2 ) dS2 − fα2 dα2 + fβ2 dβ2 + . . . + fλ2 dλ2 .
We deduce from Equations (26)

⎧
⎪ dQ1 dQ2 fα1 dα1 + fβ1 dβ1 + . . . + fλ1 dλ1
⎪
⎨ F (ϑ ) + F (ϑ ) = −d (S1 + S2 ) −
1 2 E F (ϑ1 )
(27)
⎪
⎪ f α dα + fβ2 dβ + . . . + fλ2 dλ2
⎩ − 2
2 2
.
E F (ϑ2 )
Integrating Equation (27) over a closed cycle gives us the generalisation of Clausius’
theorem due to H. Poincaré:
When a system formed from n independent parts, each with a distinct uniform temperature,
describes a closed cycle, then

dQ1 dQ2 dQn
+ + ... + ≥ 0. (28)
F (ϑ1 ) F (ϑ2 ) F (ϑn )
In the particular case where the system is without viscosity, the inequality sign disappears.
Thus, [239] we see how several properties of a system of uniform tempera-

ture can be extended to a system composed of independent parts with temperatures
different from one another.
Chapter 11
Connections
11.1 Bilateral and Unilateral Connections

Consider a system formed from separate parts which are independent of one another.
In order to simplify and make the arguments concise without detracting from their
generality, we will reduce the number of parts to two and designate them by the
indices 1 and 2.
We assume that each of these parts has uniform temperature and is defined by
normal variables; α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , ϑ1 are the normal variables defining
part 1, and α2 , β2 , . . . , λ2 , a2 , b2 , . . . , l2 , ϑ2 are the normal variables defining part 2.
Suppose that the parts are brought into contact by a continuous displacement.
While the parts approach contact, the variables
α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , ϑ1 ,
α2 , β2 , . . . , λ2 , a2 , b2 , . . . , l2 , ϑ2
tend towards limiting values, and knowledge of these limiting values suffices to
determine the state of the system at the moment when contact is established.
This amounts to making the FOLLOWING HYPOTHESIS: The state of the system
at the moment when the parts 1 and 2 are contiguous differs infinitesimally from
the state of the system when the parts are infinitesimally close to touching. What is
about to follow is not applicable to a system which does not satisfy this hypothesis.
In general, once contact has been established, the parts 1 and 2 cease to be
independent.
In [240] the first place, it is sometimes possible to impose on the system, once
contact has been established, virtual changes in which the state of each of the
parts 1 and 2 undergoes variations which would be impossible if the parts were
not contiguous.
If, for example, the parts 1 and 2 are electrified bodies, the distribution of elec-
tricity may vary on each of them, but as long as they are separated, the total charge
of each of them necessarily remains constant. On the other hand, when contact is
established between them, the total charge of the one might diminish provided that
the change on the other increases by an equal amount.

172 11 Connections
Similarly, when the parts 1 and 2 are in contact, they might mix with one another,
which would be inconceivable while they are separate.
Putting aside these cases where contact between the various parts introduces new
modes of variation into the system, we assume that in all virtual variations of the
system, both of the parts undergo a change which would still be a virtual variation of
this part if it were isolated. The two parts will, nevertheless, not be two independent
systems.
In fact, once contact is established, the virtual displacements of each of the two
parts are not absolutely arbitrary. These virtual displacements may either maintain
the contact of the parts or break the contact, whether at certain points or entirely,
and they may even establish new contacts. But since we exclude the hypothesis
of mixture, the two parts cannot begin to penetrate into one another, leading to a
portion of the one and a portion of the other being in the same place.
We seek an analytic method appropriate for excluding displacements which
would lead the two parts to penetrate one another.
Suppose that the surface S1 delimiting part 1 and the surface S2 delimiting part 2
are in contact at a certain point p. Let p1 be the material point of part 1 and p2 the
material point of part 2 which occur at p. Let N1 , N2 be the normals at p to the sur-
faces S1 and S2 , directed towards the interiors of parts 1 and 2, respectively. In order
that a displacement (δx1 , δy1 , δz1 ) of the point p1 and a displacement (δx2 , δy2 , δz2 )
of the point p2 should lead parts 1 and 2 to penetrate the [241] neighbourhood of
these points, it is necessary and sufficient that
cos(N1 , x)δx1 + cos(N1 , y)δy1 + cos(N1 , z)δz1

+ cos(N2 , x)δx2 + cos(N2 , y)δy2 + cos(N2 , z)δz2 < 0.
Displacements tending to make the two bodies penetrate one another are therefore
excluded by imposing on virtual displacements the condition
cos(N1 , x)δx1 + cos(N1 , y)δy1 + cos(N1 , z)δz1

+ cos(N2 , x)δx2 + cos(N2 , y)δy2 + cos(N2 , z)δz2 ≥ 0.
Now δx1 , δy1 , δz1 are expressed as linear and homogeneous functions of
δα1 , δβ1 , . . . , δλ1 , and δx2 , δy2 , δz2 are expressed as linear and homogeneous func-
tions of δα2 , δβ2 , . . . , δλ2 (Chapter 3, Equation (2)). The preceding condition can
therefore be written
M1 δα1 + N1 δβ1 + . . . + P1 δλ1

+ M2 δα2 + N2 δβ2 + . . . + P2 δλ2 ≥ 0,
where M1 , N1 , . . . , P1 are functions of the variables α1 , β1 , . . . , λ1 , and

M2 , N2 , . . . , P2 are functions of the variables α2 , β2 , . . . , λ2 .
Thus, the virtual displacements of a system formed from various parts in contact
are subject to a certain number of conditions of the form
11.2 The Internal Energy of a System with Connections 173
⎧
⎪ M δα + N1 δβ1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 ≥ 0,
⎨ 1 1
M1 δα1 + N1 δβ1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 ≥ 0, (1)
⎪
⎩
......................................................,
which are called connecting conditions for the system.

If just the virtual displacements which do not change the contact of parts 1 and 2
are to be retained, then the conditions (1) must be replaced by the equations
⎧
⎪ M δα + N1 δβ1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0,
⎨ 1 1
M1 δα1 + N1 δβ1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0, (2)
⎪
⎩
......................................................
When [242] the virtual displacements which break a contact are excluded and, con-
sequently, all the virtual displacements are taken to be subject to the Equations (2),
the system is said to be subject solely to bilateral connections. In that case, if
δα1 , . . . , δλ1 , δα2 , . . . , δλ2
is a virtual displacement of the system,
(−δα1 ), . . . , (−δλ1 ), (−δα2 ), . . . , (−δλ2 )
is also a virtual displacement of the system, so that all the virtual displacements of
the system are reversible.
When, on the other hand, displacements which break contacts are allowed, the
connections of the system, expressed by the conditions (1), are said to be unilateral
connections. The virtual displacements are no longer all reversible.
11.2 The Internal Energy of a System with Connections

Consider a system formed from two parts 1 and 2 which might be separated or in
contact. The internal energy of the system is designated by U when the two parts 1
and 2 are in contact.
When the two parts are separated infinitesimally, the internal energy of the
system will have the value

ϒ1 (α1 , . . . , λ1 , a1 , . . . , l1 , ϑ1 ) + ϒ2 (α2 , . . . , λ2 , a2 , . . . , l2 , ϑ2 )
(3)
+ (α1 , . . . , λ1 , a1 , . . . , l1 , α2 , . . . , λ2 , a2 , . . . , l2 ) ,
where the variables have values infinitesimally close to those which ensure contact
of the parts 1 and 2.
174 11 Connections
But, by hypothesis, the change that we have just considered changes the state
of the system infinitesimally. Consequently, the mechanical work1 performed in
this change is infinitesimal. And so the internal energy of the system also varies
infinitesimally. The quantity U is equal to the limit towards which the quantity
(3) tends when the parts 1 and 2 tend to touch one another. Thereof the following
proposition:
The internal energy of a system formed from several parts in [243] contact is equal to the
limit towards which the internal energy tends when the various parts, separated from one
another, tend to come into contact with one another.
Consider an isolated system formed from several parts in a state when the
parts 1 and 2 are infinitesimally close to being in contact. We calculate the
quantities
∂ ∂
E = −A1 , E = −A2 ,
∂α1 ∂α2
............, ............,
∂ ∂
E = −L1 , E = −L2 ,
∂λ1 ∂λ2
.
∂ ∂
E = −A1 , E = −A2 ,
∂α1 ∂α2
............, ...............,
∂ ∂
E = −L1 , E = −L2
∂λ1 ∂λ2
We calculate the limiting values towards which these quantities tend when the
parts 1 and 2 tend to touch one another. By definition, the limits of the quantities
A1 , . . . , L1 , A1 , . . . , L1 are the actions that part 2, acting against part 1, exerts on
part 1. Similarly, the limits of the quantities A2 , . . . , L2 , A2 , . . . , L2 are the actions
that part 1 exerts on part 2. The virtual work of these actions will have the
values
A1 δα1 + . . . + L1 δλ1 + A1 δa1 + . . . + L1 δl1 ,

A2 δα2 + . . . + L2 δλ2 + A2 δa2 + . . . + L2 δl2 ,
respectively, where δα1 , . . . , δλ1 , δα2 , . . . , δλ2 are, moreover, subject to the
connecting conditions.
What we have just said amounts to a definition: In fact, the mutual actions of
the two systems 1 and 2 have so far only been defined where these systems are
independent, and that is not the case here.
1 oeuvre.
11.4 Equilibrium of a System Subject to Bilateral Connections 175
11.3 Entropy [244] and the Thermodynamic Potential

of a System with Connections
Consider once more a system formed from several parts, two for example, and sup-
pose that these parts 1 and 2 are both at the same temperature. The system will then
have an entropy and an internal thermodynamic potential whether or not the parts
which compose it are in contact or not. Reasoning as in the preceding case, it will
be shown that the entropy and the internal thermodynamic potential of a system of
uniform temperature composed of several parts in contact are equal, respectively,
to the limits towards which the entropy and the internal thermodynamic potential of
the system tend when the several isolated parts tend to meet one another.
If we stick to our usual notation, this entropy S and internal thermodynamic
potential F will be the limiting values of the quantities

S = 1 (α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , ϑ) +

2 (α2 , β2 , . . . , λ2 , a2 , b2 , . . . , l2 , ϑ),
F = F1 (α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , ϑ) +
F2 (α2 , β2 , . . . , λ2 , a2 , b2 , . . . , l2 , ϑ) +
E(α1 , β1 , . . . , λ1 , a1 , b1 , . . . , l1 , α2 , β2 , . . . , λ2 , a2 , b2 , . . . , l2 , ϑ).
11.4 Equilibrium of a System Subject to Bilateral Connections

Consider a system formed of two parts 1 and 2 between which there obtains the
connecting conditions
⎧
⎪ M δα + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 ≥ 0,
⎨ 1 1
M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 ≥ 0, (1)
⎪
⎩
.............................................
Suppose that we are assured in some way that the system cannot make any
movement during which one of the inequalities
M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 > 0,

M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 > 0,
................................................
is satisfied.
Consequently, [245] when the connecting conditions (1) are replaced by the
connecting conditions (2), we can be certain that nothing changes the state of equi-
librium or motion of the system. We can therefore consider the system to be subject
solely to the bilateral connections
176 11 Connections
⎧
⎪ M δα + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0,
⎨ 1 1
M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0, (2)
⎪
⎩
................................................
Suppose that the two parts 1 and 2 are at the same temperature, ϑ. We seek the
conditions necessary and sufficient for the equilibrium of the system comprising
these two parts.
We give the system a virtual isothermal displacement compatible with
Equations (2). The external actions perform the virtual work dW. The internal ther-
modynamic potential undergoes a variation dF. In accordance with what we have
seen in Chapter 9, the necessary and sufficient conditions that we seek are obtained
by writing that, for all virtual isothermal displacements, we have
dF = dW. (4)
Now we can write
dW = A1 δα1 + . . . + L1 δλ1 + A1 δa1 + . . . + L1 δl1
+ A2 δα2 + . . . + L2 δλ2 + A2 δa2 + . . . + L2 δl2 ,
∂F1 ∂F1 ∂F1 ∂F1
δF = δα1 + . . . + δλ1 + δa1 + . . . + δl1
∂α1 ∂λ1 ∂a1 ∂l1
∂F2 ∂F2 ∂F2 ∂F2
+ δα2 + . . . + δλ2 + δa2 + . . . + δl2
∂α2 ∂λ2 ∂a2 ∂l2
∂ ∂ ∂ ∂
+E δα1 + . . . + E δλ1 + E δa1 + . . . + E δl1
∂α1 ∂λ1 ∂a1 ∂l1
∂ ∂ ∂ ∂
+E δα2 + . . . + E δλ2 + E δa2 + . . . + E δl2 .
∂α2 ∂λ2 ∂a2 ∂l2
Equation [246] (4) can therefore be written
⎧
⎪ ∂ ∂F1 ∂ ∂F1
⎪
⎪ A − E − δα + . . . + L − E − δλ1
⎪
⎪
1
∂α1 ∂α1
1 1
∂λ1 ∂λ1
⎪
⎪
⎪
⎪ ∂ ∂F1 ∂ ∂F1
⎪
⎪
⎪ + A1 − E
⎨ − δa1 + . . . + L1 − E − δl1
∂a1 ∂a1 ∂l1 ∂l1
(5)
⎪
⎪ ∂ ∂F2 ∂ ∂F2
⎪
⎪ + A2 − E ∂α2 − ∂α2 δα2 + . . . + L2 − E ∂λ2 − ∂λ2 δλ2
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ∂ ∂F2 ∂ ∂F2
⎩ + A2 − E − δa2 + . . . + L2 − E − δl2 = 0.
∂a2 ∂a2 ∂l2 ∂l2
But this equality should not be satisfied for all values of the variations
but only when Equations (2) are satisfied.

11.4 Equilibrium of a System Subject to Bilateral Connections 177
For this, it is necessary and sufficient that there exists as many factors , , . . . ,
as there are Equations (2) depending solely on the variables α1 , β1 , . . . , λ1 , α2 ,
β2 , . . . , λ2 , such that Equation (5) obtains whatever
might be when to its left-hand side is added the expression
(M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 )

+ (M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 )
+..............................................
In other words, in order that the system subject to the bilateral [246] connections
(2) be at equilibrium, it is necessary and sufficient that
⎧
⎪ ∂ ∂F1
⎪
⎪
⎪ A1 − E + M1 + M1 + . . . = ,
⎪
⎨ ∂α1 ∂α1
......................................., (6)
⎪
⎪
⎪
⎪ ∂ ∂F1
⎪
⎩ L1 − E + P1 + P1 + . . . = ,
∂λ1 ∂λ1
⎧
⎪ ∂ ∂F1
⎪ A1 − E
⎪ = ,
⎪
⎪ ∂a1 ∂a1
⎨
.................., (7)
⎪
⎪
⎪
⎪ ∂ ∂F1
⎪
⎩ L1 − E = ,
∂l1 ∂l1
⎧
⎪ ∂ ∂F2
⎪
⎪
⎪ A2 − E + M2 + M2 + . . . = ,
⎪
⎨ ∂α 2 ∂α2
......................................., (6a)
⎪
⎪
⎪
⎪ ∂ ∂F2
⎪
⎩ L2 − E + P2 + P2 + . . . = ,
∂λ2 ∂λ2
⎧
⎪ ∂ ∂F2
⎪
⎪ A2 − E = ,
⎪
⎪ ∂a2 ∂a2
⎨
.................., (7a)
⎪
⎪
⎪
⎪ ∂ ∂F2
⎪
⎩ L2 − E = ,
∂l2 ∂l2
We will examine the consequences of Equations (6), (7), (6a) and (7a).
In accordance with the definition of actions exerted by part 2 on part 1, we can
say that
178 11 Connections
∂ ∂ ∂ ∂
A1 − E , . . . , L1 − E , A1 − E , . . . , L1 − E
∂α1 ∂λ1 ∂a1 ∂l1
are the actions exerted on part 1 by the bodies external to this part. Thus, we see that
Equations (6) and (7) lead to the following important theorem:
The conditions of equilibrium of a system which, instead of being independent contain bilat-
eral connections with the bodies encircling it, are written as the conditions of equilibrium
of an independent system, [248] provided that to the external forces to which the system is
subject are associated the fictive forces
⎧
⎪
⎪ Fα = M1 + M1 + . . . ,
⎨ 1
..........................., (8)
⎪
⎪
⎩ F = P + P + . . . .
λ1 1 1
These fictive forces, called connecting forces, depend only on the variables which
fix the position relative to the system and the external bodies, and not on the variables
which change the state of the system or the external bodies without changing their relative
position.
The virtual work of the external connecting forces applied to system (1) has the
value

Fα1 δα1 + . . . + Fλ1 δλ1 = M1 + M1 + . . . δα1
+........................

+ P1 + P1 + . . . δλ1 .
Similarly, system 1 exerts on the external bodies connecting forces
⎧
⎪ F = M2 + M2 + . . . ,
⎨ α2
..........................., (8a)
⎪
⎩
Fλ2 = P2 + P2 + . . . . . . .
for which the virtual work has the value

Fα2 δα2 + . . . + Fλ2 δλ2 = M2 + M2 + . . . δα2
+.....................

+ P2 + P2 + . . . δλ2 .
The total virtual work of the connecting forces exerted between the two systems
1 and 2 has the value
11.5 Equilibrium of a System Subject to Unilateral Connections 179
Fα1 δα1 + . . . + Fλ1 δλ1 + Fα2 δα2 + . . . + Fλ2 δλ2

= (M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 )

+ M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2
+ ................................................
If [249] Equations (2) are taken into account, the following theorem can be stated:
The virtual work of all the connecting forces exerted between the two systems (1) and (2) is
zero for all virtual displacements which respect the bilateral connections existing between
the two systems.
The conditions of equilibrium of a system which is subject to bilateral connec-

tions with external bodies differ from conditions of equilibrium of an independent
system in one essential point.
Suppose that an independent system is subject to a virtual change
δα, . . . , δλ, δa, . . . , δl, while the state of the external bodies remains invariable. The
virtual work of the external actions will, in this case, be the total differential of a
function of the variables α, . . . , λ, a, . . . , l (see Section 9.4). If, on the other hand,
we consider a system subject to bilateral connections with the external bodies, the
virtual work of the external actions and the connecting forces, which should, in this
case, replace the work of just the external actions, will not, in general, be a total
differential. Since we have taken care never to assume in our demonstrations that
the virtual work of the external actions is a total differential, the results of these
demonstrations are applicable to systems subject to bilateral connections with the
external bodies.
11.5 Equilibrium of a System Subject

to Unilateral Connections
Why, when dealing with the equilibrium of a system subject to connections in the
preceding section, did we add the restriction that these connections are bilateral?
The reason is simple. We have made use of the laws of equilibrium of systems as
these were established in Chapter 9. Now, throughout the establishing of these laws,
it was supposed that the system is defined, in the neighbourhood of the state under
study, by a single system of parameters continuously varying arbitrarily. This con-
dition is not realised by a system subject to unilateral connections. Take a system
comprising various parts which, in the state under study, are in contact. Two differ-
ent systems of [250] variables must be employed to define the neighbouring states
of the system depending on whether the contact in these states is preserved or dis-
continued. It is therefore clear that our arguments will fail if the system exhibits
unilateral connections.
The preceding arguments presuppose the following proposition: The system can-
not, starting from the state at issue where the velocities of the various parts are all
equal to 0, take on a movement whose first element would represent an irreversible
180 11 Connections
virtual displacement. We therefore have necessary and sufficient conditions for equi-
librium for a system if we add to the conditions established in the preceding section
necessary and sufficient conditions for excluding any movement of this type.
Unfortunately, we cannot establish with complete rigour the necessary and suf-
ficient conditions. We can only put forward with confidence a condition which is
sufficient for rendering impossible any movement of the system beginning, without
initial velocity, with an irreversible displacement.
This condition is as follows:
If in any irreversible virtual change, the inequality

∂F ∂F
− A1 δα1 + . . . + − L1 δλ1
∂α1 ∂λ1

∂F ∂F
+ − A1 δa1 + . . . + − L1 δl1
∂a1 ∂l1

∂F ∂F
+ − A2 δα2 + . . . + − L2 δλ2
∂α2 ∂λ2

∂F ∂F
+ − A2 δa2 + . . . + − L2 δl2 > 0,
∂a2 ∂l2
holds, then it is certain that the system cannot make any movement, without an initial
velocity, beginning with an irreversible virtual displacement.
For suppose that the system could make such a movement. Let t0 be
the initial instant of this motion. Let dα1 , . . . , dλ1 , da1 , . . . , dl1 , dα2 , . . . , dλ2 ,
da2 , . . . , dl2 be the first element of [251] this movement. Suppose that

∂F ∂F
− A1 dα1 + . . . + − L2 dl2 > 0
∂α1 ∂l2
The first term of this inequality varies continuously during the motion of the
system. We could therefore determine an instant t1 , later than but quite close to
t0 , such that, between these two instants t0 and t1 , this first term always remains
positive. We would then certainly have2

t1 ∂F ∂F
− A1 dα1 + . . . + − L2 dl2 > 0.
t0 ∂α1 ∂l2
But having disrupted certain unilateral connections at the instant t0 , the system
may always be supposed to be exclusively subject to bilateral connections during
the time between t0 and the instant t1 quite close to the instant t0 . During this time,
its state will be defined by a single system of continuously but arbitrarily varying
parameters. Since its temperature is uniform the laws of motion for a system of
uniform temperature, such as those established in Chapter 10, can be applied. These
2 [The
t1
limits of integration are written “ t1 ” in the original.]
11.6 Motion of a System with Uniform Temperature Subject to Bilateral Connections 181
laws give us immediately the equation

t1 ∂F ∂F
− A1 dα1 + . . . + − L2 dl2
t0 ∂α1 ∂l2
t1

T0 − T1 + fα1 dα1 + . . . + fl2 dl2 .
t0
Since the initial velocities are, by hypothesis, all equal to 0, the same holds for
T0 , and the preceding equation gives us the inequality
t1
∂F ∂F
− A1 dα1 + . . . + − L2 dl2 < 0,
t0 ∂α1 ∂l2
which contradicts the preceding inequality.

This contradiction demonstrates the truth of the proposition stated.
Is the preceding condition, sufficient to impede any motion beginning with an
irreversible displacement, at the same time necessary for this purpose? Fourier,
Gauss, Cauchy and [252] many other geometers after them have thought so. But
it seems difficult to rigorously prove this by relying solely on what precedes.
11.6 Motion of a System with Uniform Temperature Subject

to Bilateral Connections
In order not to further lengthen the present work, we will only study the motion of
a system in as much as the connections between its various parts remain bilateral.
Consider a system formed of two parts 1 and 2 between which there exist bilateral
connections. Assume that the two parts 1 and 2 have the same temperature, ϑ. The
aggregate of these two parts then forms a system to which the general laws of motion
developed in Chapter II can be applied.
We can, without needing to know explicitly the independent variables which
define the system, state the laws of motion in the following way:
Let
dW be the virtual work of the external forces,

δF be the isothermal variation of the internal thermodynamic potential,
dτ be the virtual work of the forces of inertia,
dϕ be the virtual work of the passive resistances.
For any virtual displacement of the system, we must have
dW + dϕ − δF = dτ . (9)
182 11 Connections
Now we can write

dW = A1 δα1 + . . . + L1 δλ1 + A1 δa1 + . . . + L1 δl1
+ A2 δα2 + . . . + L2 δλ2 + A2 δa2 + . . . + L2 δl2 ,
∂F1 ∂F1 ∂F1 ∂F1
δF = δα1 + . . . + δλ1 + δa1 + . . . + δl1
∂α1 ∂λ1 ∂a1 ∂l1
∂F2 ∂F2 ∂F2 ∂F2
+ δα2 + . . . + δλ2 + δa2 + . . . + δl2
∂α2 ∂λ2 ∂a2 ∂l2
∂ ∂ ∂ ∂
+E δα1 + . . . + E δλ1 + E δa1 + . . . + E δl1
∂α1 ∂λ1 ∂a1 ∂l1
∂ ∂ ∂ ∂
+E δα2 + . . . + E δλ2 + E δa2 + . . . + E δl2 ,
∂α2 ∂λ2 ∂a2 ∂l2
[253]
dϕ = fα1 δα1 + . . . + fλ1 δλ1 + fa1 δa1 + . . . + fl1 δl1
+ fα2 δα2 + . . . + fλ2 δλ2 + fa2 δa2 + . . . + fl2 δl2 ,

∂T1 d ∂T1 ∂T1 d ∂T1
dτ = − δα + . . . + − δλ1
dt ∂α1
1
∂α1 ∂λ1 dt ∂λ1

∂T2 d ∂T2 ∂T2 d ∂T2
+ − δα2 + . . . + − δλ2 .
∂α2 dt ∂α2 ∂λ2 dt ∂λ2
Equation (9) can therefore be written

⎧
⎪ ∂F1 ∂F2
⎪
⎪ A1 δα1 + . . . + L2 δl2 − δα1 + . . . − δl2
⎪
⎪ ∂α1 ∂l2
⎪
⎪
⎨ ∂ ∂
−E δα1 − . . . − E δl2 + fα1 δα1 + . . . + fl2 δl2 (9a)
⎪
⎪ ∂α 1 ∂l2
⎪
⎪
⎪
⎪ ∂T1 d ∂T1 ∂T2 d ∂T2
⎪
⎩= − δα + . . . + − δλ2 .
dt ∂α1
1
∂α1 ∂λ2 dt ∂λ2
But in this Equation (9a) the quantities δα1 , . . . , δλ2 do not have arbitrary values.
They are subject to the equations of connection
⎧
⎪ M δα + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0,
⎨ 1 1
M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0, (2)
⎪
⎩
................................................
Then there exists factors , , . . . such that on multiplying the left-hand side of
the first Equation (2) by , multiplying the left-hand side of the second Equation (2)
by , etc., and adding the results obtained to the left-hand side of Equation (9a),
this would hold whatever the quantities δα1 , . . . , δλ2 might be.
11.6 Motion of a System with Uniform Temperature Subject to Bilateral Connections 183
The factors , , . . . depend solely on the variables α1 , . . . , λ1 , α2 , . . . , λ2 and

their first and second derivatives with respect to time and not on the variables
a1 , . . . , l1 , a2 , . . . , l2 nor their derivatives with respect to time.
It therefore holds, at each instant of the motion, that
⎧
⎪
⎪ ∂ ∂T1 d ∂T1 ∂F1
⎪
⎪ A1 − E + M1 + M1 + . . . + fα1 − + − = 0,
⎪
⎪ ∂α1 ∂α1 dt ∂α1 ∂α1
⎨
..................................................................,
⎪
⎪
⎪
⎪ ∂ ∂T1 d ∂T1 ∂F1
⎪
⎪ + P1 + P1 + . . . + fλ1 −
⎩ L1 − E + − = 0;
∂λ1 ∂λ1 dt ∂λ1 ∂λ1
(10)
[254]
⎧
⎪ ∂ ∂F 1
⎪
⎪ A1 − E + fα1 − = 0,
⎪
⎪ ∂α ∂α1
⎨ 1
.............................., (11)
⎪
⎪
⎪
⎪ ∂ ∂F 1
⎪
⎩ L1 − E + fλ1 − = 0.
∂λ1 ∂λ1
⎧
⎪
⎪ ∂ ∂T2 d ∂T2 ∂F2
⎪ A2 − E
⎪ + M2 + M2 + . . . + fα2 − + − ∂α = 0,
⎪
⎪ ∂α ∂α dt ∂α
⎪
⎨
2 2 2 2
..................................................................,
⎪
⎪
⎪
⎪
⎪
⎪ ∂ ∂T2 d ∂T2 ∂F2
⎪
⎩ L2 − E + P2 + P2 + . . . + fλ2 − + − = 0;
∂λ2 ∂λ2 dt ∂λ2 ∂λ2
(10a)
⎧
⎪ ∂ ∂F2
⎪
⎪ A2 − E + fα2 − = 0,
⎪
⎪ ∂α ∂α2
⎨ 2
..........................., (11a)
⎪
⎪
⎪
⎪ ∂ ∂F2
⎪
⎩ L2 − E + fλ2 − = 0.
∂λ2 ∂λ2
All the equations of motion of the system are obtained by adding to these equa-
tions a final equation showing how the temperature ϑ of the system varies with
time.
Equations (10) and (10a) lead to the following consequence:
It is possible to extend to a system exhibiting bilateral connections with the external bodies
all the laws of motion of a system independent of external bodies, provided that fictive
forces
184 11 Connections
Fα1 = M1 + M1 + . . . ,
........................,
Fλ 1 = P1 + P1 + . . . .
are added to the external forces.
These forces of connection depend solely on the variables α1 , . . . , λ1 , α2 , . . . , λ2 which
fix the position of the system and the external bodies, and the first and second derivatives of
these variables with respect to time.
This [255] proposition is only demonstrated, however, in so far as the aggregate

formed by the system and the external bodies has a uniform temperature at each
instant.
Consider any infinitesimal change of the system whose two parts 1 and 2, at
the same temperature, sustain bilateral connections. The total work of the forces of
connection exerted between the two parts will have the value

dα1
dλ1
M1 + M1 + . . . + . . . + P1 + P1 + . . .
dt dt

dα2

dλ2
+ M 2 + M2 + . . . + . . . + P2 + P2 + . . . dt.
dt dt
But the equations of connection (2) must be satisfied by putting
dα1 dα2
dt = δα1 , dt = δα2 ,
dt dt
............, ............,
dλ1 dλ2
dt = δλ1 , dt = δλ2 .
dt dt
The preceding work is thus equal to 0.
11.7 Motion of a System Whose Various Parts, Taken

to Different Temperatures, Are Subject to Bilateral
Connections
All that we have hitherto said on connections has been deduced from what has
been established in the two preceding chapters, without needing to invoke any new
hypothesis. This will not be the case in the present section.
In Section 10.6, we treated the motion of a system comprising several indepen-
dent parts at different temperatures. The equations obtained can be condensed in the
Assume that the system is formed from two parts 1 and 2. Let
U = ϒ1 + ϒ2 +
11.8 Calorific Coefficients of a System Subject to Bilateral Connections 185
be the internal energy of this system, and put
F = F1 + F2 + E ,
where [256] F1 and F2 are the internal thermodynamic potentials of parts 1 and 2,
assumed isolated. If we give the system any virtual displacement, we will have
dW + dϕ − δF = dτ , (12)
the various symbols having the same sense as in Equation (9).

This said, here is the hypothesis that we introduce:
HYPOTHESIS.—The preceding results remain true even in the case where the various parts
constituting the system are taken to different temperatures and exhibit bilateral connections
between one another.
Once this hypothesis is accepted, it suffices to revive the considerations of the

preceding section in order to be able to extend all the results obtained there to a
system with non-uniform temperature.
11.8 Calorific Coefficients of a System Subject

to Bilateral Connections
We only know the sense of the expression quantity of heat released by a system
in so far as this system is independent of external bodies. Where a system exhibit-
ing bilateral connections with external bodies is concerned, we cannot employ this
expression before giving a definition.
Consider, then, a system subject to bilateral connections with external bod-
ies. Let U be the internal energy of the system and Fα , . . . , Fλ the forces of
connection.
Put
⎧
⎪ ∂U ∂T d ∂T
⎪
⎪ E − + − (A + Fα ) = E Rα ,
⎪
⎪ ∂α ∂α dt ∂α
⎪
⎪
⎪
⎪
⎪
⎪ .......................................,
⎪
⎪
⎪
⎪
⎪
⎪ ∂U ∂T d ∂T
⎪
⎪
⎨ E ∂λ − ∂λ + dt ∂λ − (A + Fλ ) = E Rλ ,
(13)
⎪
⎪ ∂U
⎪
⎪ − A = E Ra ,
⎪
⎪ E
⎪
⎪ ∂a
⎪
⎪
⎪
⎪ ..................,
⎪
⎪
⎪
⎪
⎪
⎪
⎩ E ∂U − L = E Rl .
⎪
∂l
186 11 Connections
The [257] sum
− (Rα δα + . . . + Rλ δλ + Ra δa + . . . + Rl δl)
will, by definition, be the quantity of heat released by a system during the real or
virtual change δα, . . . , δλ, δa, . . . , δl.
It evidently follows from this definition, which includes as a particular case what
we have called an independent system (Section 5.5), that a real or virtual change of
an isolated system entails a release of heat equal to 0.
Consider a system formed from two parts 1 and 2. Let α1 , . . . , λ1 , a1 , . . . , l1 be
the variables defining the first part, and α2 , . . . , λ2 , a2 , . . . , l2 the variables defining
the second part.
The two parts exhibit between them contacts expressed by the equations of
connection
⎧
⎪ M δα + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0,
⎨ 1 1
M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0, (14)
⎪
⎩
..............................................
Moreover, they exhibit contacts with the external bodies expressed by the equations
of connection
⎧
⎪ m δα + . . . + p1 δλ1 + μ1 δa + . . . + 1 δl,
⎪ 1 1
⎪
⎪
⎨..........................................,
(15)
⎪
⎪ m2 δα2 + . . . + p2 δλ2 + μ2 δa + . . . + 2 δl,
⎪
⎪
⎩
..........................................,
where a, . . . , l are the variables defining the position of the external bodies.
The internal energy of the system formed by the ensemble of the two parts will
be designated by
U = ϒ1 + ϒ2 + .
The external actions applied to the system formed from two parts [258] will be
designated by A1 , . . . , L1 , A2 , . . . , L2 , A1 , . . . , L1 , A2 , . . . , L2 ,.
During any change of the system, part 1 releases a quantity of heat dQ1 given by
the equation

∂ϒ1 ∂T1 d ∂T 1
−E dQ1 = E − +
∂α1 ∂α1 dt ∂α1

∂
− A1 − E + M1 + M1 + . . . + P 1 m1 + . . . δα1
∂α1
+.........................................................

∂ϒ1 ∂T 1 d ∂T 1
+ E − +
∂λ1 ∂λ1 dt ∂λ1

∂
− L1 − E + P1 + P1 + . . . + P 1 p1 + . . . δλ1
∂λ1

∂ϒ1 ∂ ∂ϒ1 ∂
+ E +E − A1 δa1 + . . . + E +E − L1 δl1 .
∂α1 ∂α1 ∂l1 ∂l1
Similarly, part 2 releases a quantity of heat dQ2 given by the equation

∂ϒ2 ∂T2 d ∂T2
−E dQ2 = E − +
∂α2 ∂α2 dt ∂α2

∂
− A2 − E + M2 + M2 + . . . + P2 m2 + . . . δα2
∂α2
+.........................................................

∂ϒ2 ∂T2 d ∂T2
+ E − +
∂λ2 ∂λ2 dt ∂λ2

∂
− L2 − E + P2 + P2 + . . . + P2 p2 + . . . δλ2
∂λ2

∂ϒ2 ∂ ∂ϒ2 ∂
+ E +E − A2 δa2 + . . . + E +E − L2 δl2 .
∂α2 ∂α2 ∂l2 ∂l2
In virtue of Equations (14), we have
(M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 ) = 0,

M1 δα1 + . . . + P1 δλ1 + M2 δα2 + . . . + P2 δλ2 = 0,
....................................................
The [259] two preceding equations then straightforwardly yield
188 11 Connections

∂U ∂T d ∂T
−E (dQ1 + dQ2 ) = E − + − (A1 + P1 m1 + . . .) δα1
∂α1 ∂α1 dt ∂α1
+................................................

∂U ∂T d ∂T
+ E − + − (L + P p + . . .) δλ2
dt ∂λ2
2 2 2
∂λ2 ∂λ2

∂U ∂U
+ E − A2 δa1 + . . . + E − L2 δl2 .
∂a1 ∂l2
If dQ designates the quantity of heat released in the change considered by the sys-
tem comprising the two parts 1 and 2, the right-hand side of the preceding equation
represents the quantity E dQ, and so
dQ = dQ1 + dQ2 .
If, therefore, a system which can exhibit bilateral connections with neighbouring bodies is
formed from parts which can exhibit bilateral connections between one another, then the
quantity of heat released by the system in any change is the algebraic sum of the quantities
of heat released in the same change by the various parts.
Bringing this proposition together with the preceding one, we obtain the
following:
When an isolated system is formed of two parts between which there are bilateral connec-
tions, the quantity of heat released by the one part, in a real or virtual change of the system,
is equal to the quantity of heat absorbed by the other.
We have studied (Section 5.3) calorimetric procedures. We have considered three

systems S1 , S2 , S3 that were assumed independent. The preceding propositions
allow the theory of calorimetry to be extended to the case where the systems S1 ,
S2 , S3 exhibit bilateral connections between one another. This is the case which is,
in general, encountered in practice.
From Equations (13), the following consequence is easily deduced:
For a real or virtual change of a system [260] exhibiting bilateral connections
with external bodies, let dW be the work of the external actions and the external
forces of connection, and dτ the work of the forces of inertia. We have
E δU + dτ = dW + dQ.
In the case of a real change, this equation becomes
E δU + δT = dW + dQ.
These equations express the principle of the equivalence of heat and work for a
system subjected to bilateral connections with neighbouring bodies.
All that we have just said remains true even when the temperature of the system
is not uniform. Suppose now that this temperature is uniform. The equations
⎧
⎪ ∂U ∂T d ∂T
⎪
⎪ E − + − (A + Fα ) = E Rα ,
⎪
⎪ ∂α ∂α dt ∂α
⎪
⎪
⎪
⎪ .......................................,
⎪
⎪
⎪
⎪
⎨ ∂U ∂T d ∂T
E − + − (L + Fλ ) = E Rλ , (13a)
⎪ ∂λ ∂λ dt ∂λ
⎪
⎪
⎪
⎪ ∂U ∂U
⎪
⎪ E − A = E Ra , . . . , E − L = E Rl
⎪
⎪ ∂a ∂l
⎪
⎪
⎪
⎪ ∂U
⎩ E = E C,
∂ϑ
which define the calorific coefficients of the system, together with the equations of
motion
⎧
⎪ ∂F ∂T d ∂T
⎪
⎪ E − + − (A + Fα + fα ) = 0,
⎪
⎪ ∂α ∂α dt ∂α
⎪
⎪
⎪
⎪
⎨..........................................,
∂F ∂T d ∂T (10b)
⎪
⎪ E − + − (L + Fλ + fλ ) = 0,
⎪
⎪ ∂λ ∂λ dt ∂λ
⎪
⎪
⎪
⎪ ∂F ∂F
⎪
⎩E − (A + fλ ) = 0, . . . , E − (L + fl ) = 0,
∂a ∂l
and the identities
F = E [U − F (ϑ) S]
∂U ∂S
= F (ϑ) ,
∂ϑ ∂ϑ
[261] yield
⎧ ∂S
⎪
⎪ E Rα = E F (ϑ) ∂α + fα ,
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ........................,
⎪
⎪
⎪
⎪
⎪
⎪ ∂S
⎪
⎪ E Rλ = E F (ϑ) + fλ ,
⎪
⎪ ∂λ
⎪
⎪
⎪
⎪
⎨
∂S
E Ra = E F (ϑ) + fa , (16)
⎪
⎪ ∂a
⎪
⎪
⎪
⎪
⎪
⎪ ........................,
⎪
⎪
⎪
⎪
⎪
⎪ ∂S
⎪
⎪ E Rl = E F (ϑ) + fl ,
⎪
⎪ ∂l
⎪
⎪
⎪
⎪
⎪
⎪
⎩ E C = E F (ϑ) ∂S ,
∂ϑ
which are equations similar to those obtained for an independent system.
190 11 Connections
From this it can be deduced that, for any real change,
E dQ fα dα + . . . + fl dl
= −E dS − . (17)
F (ϑ) F (ϑ)
Consider a system formed from n independent parts 1, 2, . . . , n exhibiting
bilateral connections between one another. Each of these parts has a uniform temper-
ature, but this temperature is not necessarily the same for all the parts. The system
itself can exhibit bilateral connections with neighbouring bodies.
For each of these parts, we write an equation analogous to Equation (17) and add
term by term these various equations. If we put
S = S1 + S2 + . . . + Sn , (18)
we will find

dQ1 dQ2 dQn fα dα1 + . . . + fl1 dl1
E + + ... + = −E dS − 1
F (ϑ1 ) F (ϑ2 ) F (ϑn ) F (ϑ1 )
fα2 dα2 + . . . + fl2 dl2 fα dαn + . . . + fln dln

− − ... − n .
F (ϑ2 ) F (ϑn )
Integrating [262] this equation over a closed cycle, and noting that

dS = 0
fα1 dα1 + . . . + fl1 dl1 ≤ 0
fα2 dα2 + . . . + fl2 dl2 ≤ 0
...........................,
fαn dαn + . . . + fln dln ≤ 0.
we will arrive at the inequality

dQ1 dQ2 dQn
+ + ... + ≥ 0. (19)
F (ϑ1 ) F (ϑ2 ) F (ϑn )
This inequality is thus extended to systems formed from parts at different

temperatures exhibiting bilateral connections between one another.
Chapter 12
Stability and Displacement of Equilibrium
12.1 On the Total Thermodynamic Potential

The virtual work of the external actions on a system (including, if necessary, the
forces of the external connections),
Aδα + Bδβ + . . . + Lδλ + Aδa + Bδb + . . . + Lδl,
is not, in general, the total differential of a uniform function of the normal variables
α, β, . . . , λ, a, b, . . . , l. However, if it is not so in general, it is in a large number of
important particular cases.
When there is a uniform function such that
Adα + Bdβ + . . . + Ldλ + Ada + Bdb + . . . + Ldl = −d, (1)
it is said that the external actions of the system sustain a potential [263] and this
potential is . When it exists, this potential is evidently determined up to a constant.
When the external actions of the system sustain a potential we call the sum
= F + d (2)
of the internal thermodynamic potential and the potential of the external actions, the
total thermodynamic potential of the system.
Here is an important case where the system sustains a total thermodynamic
potential:
Suppose the real or imaginary variations of the external bodies to be tied to
the real or virtual variations of the system in such a way that the external actions
A, B, . . . , L, A, B, . . . , L maintain fixed values. It is then possible to write
Adα + Bdβ + . . . + Ldλ + Aa + Bb + . . . + Ll

= −d(Aα + Bβ + . . . + Lλ + Aa + Bb + . . . + Ll)
and the system sustains the total thermodynamic potential

192 12 Stability and Displacement of Equilibrium
= F − (Aα + Bβ + . . . + Lλ + Ada + Bdb + . . . + Ldl). (3)
This quantity, which is often considered in applications, is called the thermody-

namic potential under constant actions.
12.2 Stability of the Equilibrium at Constant Temperature

From now on there will be no point in distinguishing the variables a, b, . . . , l from
the variables α, β, . . . , λ. We therefore consider a system whose temperature is
uniform and which is defined by the normal variables α, β, . . . , λ, ϑ. For any real
motion of the system, we have (Chapter 10, Equation (11))
∂F
dF − dϑ − (Adα + Bdβ + . . . + Ldλ)
∂ϑ

= −dT + fα dα + fβ dβ + . . . + fλ dλ .
If [264] the external actions sustain a potential , this equation, together with
Equations (1) and (2), becomes
∂F
d − dϑ = −dT + fα dα + fβ dβ + . . . + fλ dλ.
∂ϑ
If the change is isothermal, this equation becomes
= −dT + fα dα + fβ dβ + . . . + fλ dλ. (4)
Integrating between the initial state 0 and a final state 1 of the system, it becomes
1

0 − 1 = −T1 − T0 + fα dα + fβ dβ + . . . + fλ dλ . (4a)
0
It then suffices to note that the quantity

1

0
can never be positive, and to repeat almost word-for-word the demonstration of

Lejeune-Dirichlet1 that has now become classic, to arrive at the following theorem:
A system whose temperature is maintained constant is definitely in stable equilibrium in
a state in which the total thermodynamic potential takes on a minimum value amongst all
those that it could take at the same temperature.
1 Lejeune-Dirichlet, Ueber die Stabilität des Gleichgewichts (Journal de Crelle, vol. 22, p. 85;
1846).
12.3 Displacement of Equilibrium by a Variation of Temperature 193
It is not at all certain that there are not states of stable equilibrium of a system
other than those corresponding to a minimum of the total thermodynamic potential.
In what follows, when we speak of a system at equilibrium, we will understand
it to designate, for the sake of brevity, a system [265] whose total thermodynamic
potential has a minimum value amongst all those that it can take at the same
temperature.
12.3 Displacement of Equilibrium by a Variation

of Temperature
A system is at equilibrium at a given temperature ϑ when it is subject to actions

A, B, . . . , L. The normal variables α, β, . . . , λ which, along with the temperature ϑ,
completely determine the system, have, in this state, values given by the equations
∂F ∂F ∂F
= A, = B, . . . , = L.
∂α ∂β ∂λ
.
We assume that the actions A, B, . . . , L sustain a potential , which will be a func-
tion of α, β, . . . , λ, but not of ϑ (Section 9.5), and in conformity with Equation (1),
we put = F + d. The preceding equations can be written
∂ ∂ ∂
= 0, = 0, ..., = 0. (5)
∂α ∂β ∂λ
Solving for α, β, . . . , λ, these equations become
α = α(ϑ), β = β(ϑ), ..., λ = λ(ϑ).
At the temperature (ϑ + dϑ), under the actions of the same external bodies, a
new state of equilibrium will be established in which the normal variables defining
the state of the system will have the new values
dα (ϑ)
α = α (ϑ) + dϑ,
dϑ
dβ (ϑ)
β = β (ϑ) + dϑ,
dϑ
........................,
dλ (ϑ)
λ = λ (ϑ) + dϑ .
dϑ
Differentiating the equations of equilibrium (5) with respect to ϑ yields [266] the
system of equations
⎧
⎪
⎪ ∂ 2 ∂ 2 dα ∂ 2 dβ ∂ 2 dλ
⎪
⎪ + + + . . . + = 0,
⎪
⎪ ∂α ∂ϑ ∂α 2 dϑ ∂α ∂β dϑ ∂α ∂λ dϑ
⎪
⎪
⎪
⎪
⎨ ∂ + ∂ dα + ∂ dβ + . . . + ∂ dλ = 0,
⎪ 2 2 2 2
∂β ∂ϑ ∂β ∂α dϑ ∂β 2 dϑ ∂β ∂λ dϑ (6)
⎪
⎪
⎪
⎪ ...................................................,
⎪
⎪
⎪
⎪
⎪
⎪ ∂ 2 ∂ 2 dα ∂ 2 dλ ∂ 2 dλ
⎪
⎩ + + + ... + 2 = 0.
∂λ ∂ϑ ∂λ ∂α dϑ ∂λ ∂β dϑ ∂λ dϑ

We multiply the two
sides of the first of these equations by dα dϑ, the two sides
of the second by dβ dϑ, . . . , and the two sides of the last by dλ dϑ. Adding the
results obtained yields the following equation
⎧ 2
⎪
⎪ ∂ dα ∂ 2 dβ ∂ 2 dλ ∂ 2 dα 2
⎨ ∂α ∂ϑ dϑ + ∂β ∂ϑ dϑ + . . . + ∂λ ∂ϑ dϑ + ∂α 2 dϑ
⎪
(7)
⎪
⎪ ∂ 2 dβ 2 ∂ 2 λ 2 ∂ 2 dλ dμ
⎪
⎩ + + . . . + + 2 = 0.
∂β 2 dϑ ∂β 2 dϑ ∂μ ∂ν dϑ dϑ
∂ 2 dλ dμ
In this equation, the symbol should be attributed the following
∂μ ∂ν dϑ dϑ
meaning:
∂ 2 dλ dμ
All the values distinct from one another of the quantity that can
∂μ ∂ν dϑ dϑ
be obtained by replacing μ and ν by two letters, distinct from one another, taken
from the collection α, β, . . . , λ, are considered and their sum taken.
By hypothesis, the values of α, β, . . . , λ defined by Equations (5) give to a min-
imum value amongst all those which it can take at the temperature ϑ. Consequently,
if these values are ascribed to α, β, . . . , λ, then the quadratic form
∂ 2 2 ∂ 2 2 ∂ 2 ∂ 2
a + b + . . . + 2 l2 + 2 mn
∂α 2 ∂β 2 ∂λ ∂μ ∂ν
is [267] positive, whatever the values, different from 0, that are attributed to the
quantities a, b, . . . , l.
This will be the case, in particular, if
dα dβ dλ
a= , b= ,...,l =
dϑ dϑ dϑ
Therefore,
2 2 2 ∂ 2 dλ dμ
∂ 2 dα ∂ 2 dβ ∂ 2 λ
+ + ... + +2 >0
∂α 2 dϑ ∂β 2 dϑ ∂λ2 dϑ ∂μ ∂ν dϑ dϑ
12.3 Displacement of Equilibrium by a Variation of Temperature 195
This equation, together with Equation (7), yields the inequality
∂ 2 dα ∂ 2 dβ ∂ 2 dλ
+ + ... + < 0.
∂α ∂ϑ dϑ ∂β ∂ϑ dϑ ∂λ ∂ϑ dϑ
We put
dα dβ dλ
dα = dϑ, dβ = dϑ, . . . , dλ = dϑ.
dϑ dϑ dϑ
The inequality that we have just obtained can be stated in the following way.
The quantity
∂ 2 ∂ 2 ∂ 2
dα + dβ + . . . + dλ
∂α ∂ϑ ∂β ∂ϑ ∂λ ∂ϑ
has the opposite sign to that of dϑ.
The external actions A, B, . . . , L equal, except for sign, to the partial derivatives
of the function with respect to α, β, . . . , λ, are independent of the temperature ϑ.
It therefore follows that
∂ 2 ∂ 2 ∂ 2
= 0, = 0, ..., = 0.
∂α ∂ϑ ∂β ∂ϑ ∂λ ∂ϑ
Consequently, in virtue of Equation (1), the result that we have just obtained can
be stated thus:
The quantity
∂ 2F ∂ 2F ∂ 2F
dα + dβ + . . . + dλ (8)
∂α ∂ϑ ∂β ∂ϑ ∂λ ∂ϑ
has the opposite sign to that of dϑ.
In [268] virtue of Equation (5a) of Chapter 9, the quantity (8) can be written

∂S ∂S ∂S
− E F (ϑ) dα + dβ + . . . + dλ
∂α ∂β ∂λ
In virtue of Equations (3) of the same chapter, this becomes
E F (ϑ)

− Rα dα + Rβ dβ + . . . + Rλ dλ
F (ϑ)

Since the quantity F (ϑ) F (ϑ) is certainly positive, the quantity
Rα dα + Rβ dβ + . . . + Rλ dλ
clearly has the same sign as dϑ, which is expressed by the following law:
Consider a system subject to actions which sustain a potential. The system is at equilibrium
at a given temperature. If this temperature is raised, it establishes an equilibrium different
from the first; the normal variables which characterise the state of the system undergo
certain variations; if they undergo the same variations at constant temperature, the virtual
change imposed on the system entails an absorption of heat.
This law, which was glimpsed by Lavoisier and Laplace, was stated by J.-H.
van ’t Hoff.
12.4 Isothermal Displacement of Equilibrium

A system is subject to the constant actions A, B, . . . , L and has a temperature ϑ. It
adopts a certain stable equilibrium state which renders the thermodynamic potential
a minimum under the constant actions A, B, . . . , L. The normal variables α, β, . . . , λ
have values defined as functions of A, B, . . . , L by the equations
∂F ∂F ∂F
= A, = B, ..., = L. (9)
∂α ∂β ∂λ
Letting the temperature have the value ϑ and giving the actions [269] A, B, . . . , L
new values close to the preceding ones, (A +dA), (B+dB), . . . , (C +dC), the system
adopts a new state of equilibrium defined by the values (α + dα), (β + dβ), . . . , (λ +
dλ) of the normal variables α, β, . . . , λ, putting for the sake of brevity
∂α ∂α ∂α
dα = dA + dB + . . . + dL,
∂A ∂B ∂L
∂β ∂β ∂β
dβ = dA + dB + . . . + dL,
∂A ∂B ∂L
....................................,
∂λ ∂λ ∂λ
dλ = dA + dB + . . . + dL.
∂A ∂B ∂L
Differentiating Equations (9) gives us
∂ 2F ∂ 2F ∂ 2F
dα + dβ + . . . + dλ = dA,
∂α 2 ∂α ∂β ∂α ∂λ
∂ 2F ∂ 2F ∂ 2F
dα + dβ + . . . + dλ = dB,
∂β ∂α ∂β 2 ∂β ∂λ
.............................................,
∂ 2F ∂ 2F ∂ 2F
dα + dβ + . . . + 2 dλ = dL.
∂λ ∂α ∂λ ∂β ∂λ
Multiplying the first of these equations by dα, the second by dβ, . . . , the last by
dλ, and adding the results obtained, we find
12.4 Isothermal Displacement of Equilibrium 197
⎧
∂ 2F
⎨ ∂ F (dα)2 + ∂ F (dβ)2 + . . . + ∂ F (dλ)2 + 2
⎪ 2 2 2
dμ dν
∂α 2 ∂β 2 ∂λ2 ∂μ ∂ν (10)
⎪
⎩
= dA dα + dB dβ + . . . + dL dλ.
We now propose to determine the sign of the left-hand side of Equation (10).
Consider the thermodynamic potential under the constant actions A, B, . . . , L.
= F(α, β, . . . , λ, ϑ) − (Aα + Bβ + . . . + Lλ). (11)
By [270] hypothesis, the values of α, β, . . . , λ defined by the Equations (9) lead

to a minimum value of this potential taken amongst all those that it can take at the
temperature ϑ. Therefore, if α, β, . . . , λ are given these values, the quadratic form
∂ 2 2 ∂ 2 2 ∂ 2 2 ∂ 2
a + b + . . . + l + 2 mn
∂α 2 ∂β 2 ∂λ2 ∂μ ∂ν
will be positive whatever the non-zero values that are attributed to the quantities
a, b, . . . , l. It will be, in particular, if we put
a = dα, b = dβ, . . . , l = dλ.
We will therefore have2
∂ 2 ∂ 2 ∂ 2 ∂ 2
(dα)2
+ (dβ) 2
+ . . . + (dλ)2
+ 2 dm dn > 0
∂α 2 ∂β 2 ∂λ2 ∂μ ∂ν
But since the actions A, B, . . . , L figuring in the expression (11) of are

constants, we have
∂ 2 ∂ 2F ∂ 2 ∂ 2F ∂ 2 ∂ 2F ∂ 2 ∂ 2F
= , = , ..., = 2, =
∂α 2 ∂α 2 ∂β 2 ∂β 2 ∂λ 2 ∂λ ∂μ ∂ν ∂μ ∂ν
and consequently,
∂ 2F ∂ 2F ∂ 2F ∂ 2F
(dα)2
+ (dβ) 2
+ . . . + (dλ)2
+ 2 dμ dν > 0.
∂α 2 ∂β 2 ∂λ2 ∂μ ∂ν
Equation (10) therefore leads us to the inequality
dA dα + dB dβ + . . . + dL dλ > 0,
which is expressed by the following proposition:
2 [Inits original form, the following equation has (da)2 instead of (dα)2 , (db)2 instead of (dβ)2 ,
and (dl)2 instead of (dλ)2 .]
A system is at equilibrium at a given temperature under certain external actions. To these

external actions certain infinitesimal perturbational actions are added while the temper-
ature remains the same, the initial equilibrium is disturbed, and a new [271] state of
equilibrium is established. In the passage from the old state of equilibrium to the new, the
perturbational actions always effect positive work.
A special case of this law has been stated by H. Le Chatelier.
12.5 Isentropic Stability of the Equilibrium

A change in which the entropy of the system remains the same is called, following
Gibbs, an isentropic change. During such a change the variations dα, dβ, . . . , dλ, dϑ
of the normal variables are connected by the relation
dS dS dS dS
dα + dβ + . . . + dλ + dϑ = 0. (12)
dα dβ dλ dϑ
Isentropic changes are interesting to consider because if the system describing

an isentropic change is at equilibrium at each instant under the action of external
bodies, the isentropic change that it describes is what we have called a reversible
adiabatic change. Conversely, a reversible adiabatic change is isentropic.
We consider a system subject to external actions which sustains a potential ,
and suppose that the system is at equilibrium. We have
∂(F + ) ∂(F + ) ∂(F + )

= 0, = 0, . . . , = 0.
∂α ∂β ∂λ
These equalities can be written
⎧
⎪ ∂U ∂S ∂
⎪
⎪ E + E F (ϑ) + = 0,
⎪
⎪ ∂α ∂α ∂α
⎪
⎪
⎪
⎨ E ∂U + E F (ϑ) ∂S + ∂ = 0,
⎪
∂β ∂β ∂β (13)
⎪
⎪
⎪
⎪ . . . . . . . . . . . . . . . . . . . . . . . .......,
⎪
⎪
⎪
⎪
⎩ E ∂U + E F (ϑ) ∂S + ∂ = 0.
⎪
∂λ ∂λ ∂λ
We add [272] the identity
∂U ∂S
− F (ϑ) =0
∂ϑ ∂ϑ
which, since can always be chosen independently of ϑ, can also be written as
∂U ∂S ∂
E − E F (ϑ) + = 0. (13a)
∂ϑ ∂ϑ ∂α
12.5 Isentropic Stability of the Equilibrium 199
Multiplying the two sides of the first Equation of (13) by dα, the two sides
of the second by dβ, . . . , the two sides of the last by dλ, and the two sides of
the Equation (13a) by dϑ, then dα, dβ, . . . , dλ, dϑ define an isentropic change.
Adding the left and right hand sides of the results obtained and taking into account
Equation (12) yields
d(E U + ) = 0. (14)
Thus, all isentropic changes undergone by a system starting in a state of equilib-

rium conform to Equation (14). Moreover, this equation simply expresses that the
change at issue is adiabatic.
Recall the general equality (Chapter 10, Equation (11))
∂F
dF − dϑ + d = −dT + fα dα + fβ dβ + . . . + fλ dλ.
∂ϑ
It can be written

∂U ∂U ∂U
E dα + dβ + . . . + dλ
∂α ∂β ∂λ

∂S ∂S ∂S
−E F (ϑ) dα + dβ + . . . + dλ + d
∂α ∂β ∂λ
= −dT + fα dα + fβ dβ + . . . + fλ dλ
and the identity
∂U ∂S
− F (ϑ) =0
∂ϑ ∂ϑ
allows [273] it to be transformed into
d (E U + ) − E F (ϑ) dS = −dT + fα dα + fβ dβ + . . . + fλ dλ.
For an isentropic change, this equality reduces to
d (E U + ) = −dT + fα dα + fβ dβ + . . . + fλ dλ. (15)
Since the quantity

can never be positive, a demonstration like that of Lejeune-Dirichlet can be

constructed from Equations (14) and (15) to obtain the following theorem:
The equilibrium of a system is certainly stable for any isentropic changes it may be subject
to if the quantity d(E U +) has a minimum value amongst all those that it can take without
changing the value of the entropy.
This theorem has been stated by J.-W. Gibbs. It can be said with him that if
the quantity (F + ) is the thermodynamic potential at constant temperature, the
quantity (EU + ) is the thermodynamic potential at constant entropy.
Consider a system at equilibrium. If the equilibrium of this system is stable for
all the isothermal changes that can be imposed on it, is it also stable for all isentropic
changes? In other words, if it is stable when the system is supposed encompassed
by sources of heat which maintain the temperature constant, is it still stable when
the system is supposed surrounded by a wall impermeable to heat?
In order to answer this question, it is necessary for us to invoke a fundamental
postulate which we will call Helmholtz’s Postulate, H. von Helmholtz being the
only physicist, to our knowledge, who has explicitly stated it.3 This postulate is the
following:
HELMHOLTZ’S P OSTULATE.—When [274] a system is described with normal variables,
the calorific capacity of the system is positive.
This postulate is expressed by the inequality
C > 0, (16)
which the equalities
∂U
C= ,
∂ϑ
∂S
C = F (ϑ) ,
∂ϑ
F (ϑ) F (ϑ) ∂F F (ϑ) ∂ 2 F
C= 2 −
E F (ϑ) ∂ϑ E F (ϑ) ∂ϑ 2
allow us to rewrite in any of the following three forms

⎧ ∂U
⎪
⎪ > 0,
⎪
⎪ ∂ϑ
⎪
⎪
⎨ ∂S
> 0, (16a)
⎪
⎪ ∂ϑ
⎪
⎪
⎪
⎩ F (ϑ) ∂F − ∂ F > 0.
2
⎪
F (ϑ) ∂ϑ ∂ϑ 2
It should be noted that this postulate may well cease to agree with experiment if
the system is not defined by normal variables. Thus, the calorific capacity of water
vapour under the pressure of the saturated vapour is negative. This remark shows,
3 H.
von Helmholtz, Zur Thermodynamik chemischer Vorgänge. I. (Sitzungsberichte der Berliner
Akademie, 1st. semester, pp. 12 and 19; 1882).
moreover, that the preceding postulate, however natural it may appear, is neither
obvious nor necessary.
Given this postulate, we will show that the isothermal stability of equilibrium
entails isentropic stability.
We reserve the symbol d to designate isentropic variations and use the symbol δ
to designate isothermal variations. The proposition to be deduced evidently reduces
to this: Knowing that, for a system at equilibrium, it holds that
δ 2 (F + ) > 0, (17)
it [275] follows that
d2 (EU + ) > 0. (18)
The inequality (17) can be written more explicitly as

⎧ 2
⎪ ∂ (F + )
⎪
⎨ (δα)2 + . . .
∂α 2
(17a)
⎪
⎪ ∂ 2 (F + ) ∂ 2 (F + )
⎩ + (δλ) 2
+ 2 δμ δν > 0
∂λ2 ∂μ ∂ν
During an isothermal change, the quantities δα, δβ, . . . , δλ are absolutely arbi-
trary. The preceding inequality (17a) will therefore still hold if δα, δβ, . . . , δλ are
replaced by the variations dα, dβ, . . . , dλ corresponding to an isentropic change.
Consequently, we will have, for any isentropic change,
⎧ 2
⎪ ∂ (F + )
⎪
⎨ (dα)2 + . . .
∂α 2
(19)
⎪
⎪ ∂ 2 (F + ) ∂ 2 (F + )
⎩ + (dλ) 2
+ 2 dμ dν > 0.
∂λ2 ∂μ ∂ν
On the other hand, the identity
dS dS dS
dα + . . . + dλ + dϑ = 0, (12)
dα dλ dϑ
which defines an isentropic change, gives by differentiation
⎧ 2 ∂ 2S
⎪
⎪ ∂ S ∂ 2S
⎪
⎪ (dα) 2
+ . . . + (dλ) 2
+ 2 dμ dν
⎪
⎪ ∂α 2 ∂λ2 ∂μ ∂ν
⎪
⎨ dS 2 dS 2
+ d α + ... + d λ (20)
⎪
⎪ dα dλ
⎪
⎪
⎪
⎪ ∂ 2S ∂ 2S ∂ 2S ∂S 2
⎪
⎩+ 2 dα + . . . + 2 dλ + dϑ dϑ + d ϑ = 0.
∂α∂ϑ ∂λ∂ϑ ∂ϑ 2 ∂ϑ
Adding left and right sides of the inequality (19) and the equality (20), having
first multiplied the latter by E F(ϑ), and taking into account the equation
F + EF(ϑ)S = EU, (21)
yields [276] the inequality

⎧ 2
⎪
⎪ ∂ (E U + ) 2 ∂ 2 (E U + )
⎪
⎪ (dα) + . . . + (dλ)2
⎪
⎪ ∂α 2 ∂λ 2
⎪
⎪
⎪
⎪ ∂ 2 (E U + ) dS 2 dS 2
⎪
⎪
⎨ +2 dμ dν + E F (ϑ) d α + ... + d λ
∂μ ∂ν dα dλ
(22)
⎪
⎪ ∂ 2S ∂ 2S ∂ 2S
⎪
⎪ +E (ϑ) + . . . + +
⎪
⎪ F 2 dα 2 dλ dϑ dϑ
⎪
⎪ ∂α∂ϑ ∂λ∂ϑ ∂ϑ 2
⎪
⎪
⎪
⎪ ∂S 2
⎩ +E F (ϑ) d ϑ > 0.
∂ϑ
But when the system is at equilibrium,
∂(F + ) ∂ (F + )
= 0, . . . , =0
∂α ∂λ
and, consequently,
d(F + ) 2 d(F + ) 2
d α + ... + d λ=0
dα dλ
which, in virtue of Equation (21), can be written
∂(E U + ) 2 ∂(E U + ) 2
d α + ... + d λ
∂α ∂λ

∂S 2 ∂S 2
= E F (ϑ) d α + ... + d λ .
∂α ∂λ
Given this equation, the inequality (22) becomes
⎧
⎪
⎪ ∂ 2 (E U + ) ∂ 2 (E U + )
⎪
⎪ (dα)2 + . . . + (dλ)2
⎪
⎪ ∂α 2 ∂λ 2
⎪
⎪
⎪
⎪ ∂ 2 (E U + ) ∂(E U + ) 2 ∂(E U + ) 2
⎪
⎪
⎨ +2 dμ dν + d α + ... + d λ
∂μ ∂ν ∂α ∂λ
⎪ (23)
⎪
⎪ ∂ 2S ∂ 2S ∂ 2S
⎪
⎪ +E F (ϑ) 2 dα + . . . + 2 dλ + dϑ dϑ
⎪
⎪ ∂α∂ϑ ∂λ∂ϑ ∂ϑ 2
⎪
⎪
⎪
⎪ ∂S 2
⎪
⎩ +E F (ϑ) d ϑ > 0.
∂ϑ
Noting [277] that is independent of ϑ, the identity
∂U ∂S
= F(ϑ)
∂ϑ ∂ϑ
can be written.
∂S ∂(E U + )
E F(ϑ) dϑ = dϑ.
∂ϑ ∂ϑ
Differentiating yields

∂ 2S ∂ 2S ∂ 2S
E F(ϑ) dα + . . . + dλ + dϑ dϑ
∂α∂ϑ ∂λ∂ϑ ∂ϑ 2
∂S ∂S 2
+ E F (ϑ) (dϑ)2 + E F (ϑ) d ϑ=
∂ϑ ∂ϑ
2
∂ (E U + ) ∂ 2 (E U + ) ∂ 2 (E U + )
dα + . . . + dλ + dϑ
∂α∂ϑ ∂λ∂ϑ ∂ϑ 2
∂(E U + ) 2
dϑ + d ϑ
∂ϑ
Comparing with the inequality (23), this equation allows us to write
⎧
⎪
⎪ ∂ 2 (E U + ) 2 ∂ 2 (E U + )
⎪
⎪ (dα) + . . . + (dλ)2
⎪
⎪ ∂α 2 ∂λ2
⎪
⎪
⎪
⎪ ∂ 2 (E U + )
⎪
⎪ +2 dμ dν
⎪
⎪ ∂μ ∂ν
⎪
⎪
⎪
⎪
⎪
⎪ ∂ 2 (E U + ) ∂ 2 (E U + ) ∂ 2 (E U + )
⎪
⎨+ dα + . . . + dλ + dϑ dϑ
∂α∂ϑ ∂λ∂ϑ ∂ϑ 2 (24)
⎪
⎪ ∂(E U + ) 2 ∂(E U + ) 2 ∂(E U + ) 2
⎪
⎪ + d α + ... + d λ+ d ϑ
⎪
⎪
⎪
⎪ ∂α ∂λ ∂ϑ
⎪
⎪ 2
⎪
⎪ ∂ S ∂ 2S
⎪
⎪ +E F(ϑ) dα + . . . + dλ dϑ
⎪
⎪ ∂α∂ϑ ∂λ∂ϑ
⎪
⎪
⎪
⎪
⎪
⎩ ∂S
− E F (ϑ) (dϑ)2 > 0.
∂ϑ
The identity
∂U ∂S
= F(ϑ) ,
∂ϑ ∂ϑ
which [278] can be written
∂(E U + ) ∂S
= E F(ϑ)
∂ϑ ∂ϑ
yields the identities
⎧ 2
⎪
⎪ ∂ (E U + ) ∂ 2S
⎪
⎪ = E F (ϑ) ,
⎪
⎨ ∂α∂ϑ ∂α∂ϑ
................................., (25)
⎪
⎪
⎪
⎪ ∂ 2 (E U + ) ∂ 2S
⎪
⎩ = E F (ϑ) .
∂λ∂ϑ ∂λ∂ϑ
Given Equations (25), the inequality (24) becomes
∂ 2 (E U + ) ∂ 2 (E U + ) ∂ 2 (E U + )
(dα)2 + . . . + (dλ)2 + (dϑ)2
∂α 2 ∂λ 2 ∂ϑ 2
∂ 2 (E U + ) 2
∂ (E U + ) ∂ 2 (E U + )
+2 dμ dν +2 dα +. . .+ dλ dϑ
∂μ ∂ν ∂α∂ϑ ∂λ∂ϑ
∂(E U + ) 2 ∂(E U + ) 2 ∂(E U + ) 2
+ d α + ... + d λ+ d ϑ
∂α ∂λ ∂ϑ
∂S
− E F (ϑ) (dϑ)2 > 0 ,
∂ϑ
or
F (ϑ)
d2 (E U + ) − E C (dϑ)2 > 0.
F(ϑ)

The quantities E, F(ϑ) and F (ϑ) are essentially positive, and according to
Helmholtz’s postulate, the same holds for C. The preceding inequality therefore
requires that we have
d2 (E U + ) > 0.
This is the inequality (18) that we wished to demonstrate.
12.6 Isentropic Displacement of Equilibrium

Consider a system at stable equilibrium for isothermal changes, under the constant
actions A, B, . . . , L. The quantity
= F(α, β, . . . , λ, ϑ) − (Aα + Bβ + . . . + Lλ),
in [279] which the quantities A, B, . . . , L are independent of the variables

α, β, . . . , λ, has a minimum value amongst those that it can take at the same
temperature ϑ. We have, then, in the first place
12.6 Isentropic Displacement of Equilibrium 205
∂F ∂F
− A = 0, ..., −L=0 (26)
∂α ∂λ
and in the second place,
∂ 2F ∂ 2F ∂ 2F
(δα)2
+ . . . + (δλ) 2
+ 2 δμ δν > 0 (27)
∂α 2 ∂λ2 ∂μ ∂ν
This latter inequality holds whatever the values, not all zero, of δα, δβ, . . . , δλ.
We give the actions A, B, . . . , L infinitesimal variations dA, dB, . . . , dL while
the system exchanges no heat with the outside. The system undergoes an isentropic
change and reaches a new equilibrium state. In this state the normal variables which
define the system will have the new values
α + dα, β + dβ, . . . , λ + dλ, ϑ + dϑ.
Since the change is isentropic, we must have the identity

dS dS dS
dα + . . . + dλ + dϑ = 0.
dα dλ dϑ
which the identity
dF
E F (ϑ) S = −
dϑ
allows us to transform into
∂ 2F ∂ 2F ∂ 2F F (ϑ) ∂F
dα + . . . + dλ + dϑ − = 0. (28)
∂α∂ϑ ∂λ∂ϑ ∂ϑ 2 F (ϑ) ∂ϑ
On the other hand, Equations (26) yield, by differentiation,

⎧ 2
⎪
⎪ ∂ F ∂ 2F ∂ 2F
⎪
⎪ dα + . . . + dλ + dϑ = dA,
⎪
⎨ ∂α
2 ∂α∂λ ∂α∂ϑ
.........................................., (29)
⎪
⎪
⎪
⎪
⎩ ∂ F dα + . . . + ∂ F dλ + ∂ F dϑ = dL.
2 2 2
⎪
∂λ∂α ∂λ 2 ∂λ∂ϑ
Multiplying [280] the first Equation (29) by dα, . . . , the last by dλ, and adding
the left and right sides of the results obtained, we find
∂ 2F ∂ 2F ∂ 2F
(dα) 2
+ . . . + (dλ)2
+ 2 δμ δν
∂α 2 ∂λ2 ∂μ ∂ν
2
∂ F ∂ 2F
+ dα + . . . + dλ dϑ = dA dα + . . . + dL dλ ,
∂α∂ϑ ∂λ∂ϑ
an equation which the relation (28) transforms into

∂ 2F ∂ 2F ∂ 2F
(dα) 2
+ . . . + (dλ)2
+ 2 δμ δν
∂α 2 ∂λ2 ∂μ ∂ν
2
∂ F F (ϑ) ∂F
− − (dϑ)2 = dA dα + . . . + dL dλ ,
∂ϑ 2 F (ϑ) ∂ϑ
or into
⎧
⎪
⎪ dA dα + . . . + dL dλ =
⎨
⎪ E F (ϑ) ∂ 2F ∂ 2F ∂ 2F (30)
⎪
⎩ C (δϑ)2 2 (dα)2 + . . . + 2 (dλ)2 + 2 δμ δν .
F (ϑ) ∂α ∂λ ∂μ ∂ν
This inequality (27) holds whatever δα, δβ, . . . , δλ might be if we put
dα = δα, . . . , dλ = δλ.
We therefore have
∂ 2F ∂ 2F ∂ 2F
(dα) 2
+ . . . + (dλ)2
+ 2 δμ δν > 0
∂α 2 ∂λ2 ∂μ ∂ν
On the other hand, Helmholtz’ postulate gives
E F (ϑ)
C (δϑ)2 > 0
F(ϑ)
We therefore have
dA dα + . . . + dL dλ > 0. (31)
This [281] inequality (31) is equivalent to the following theorem:

A system is at equilibrium under certain external actions and can only undergo isentropic
changes. If infinitesimal perturbations are added to the external actions, the system achieves
a new state of equilibrium, and in the passage from the old equilibrium state to the new, the
perturbational actions always perform positive work.
This principle of isentropic displacement from equilibrium is employed in the

discussion of certain problems concerning the expansion of vapours.
12.7 Specific Heat Under Constant Actions

Consider a system maintained at equilibrium by constant external actions
A, B, . . . , L. It passes from a state defined by the values α, β, . . . , λ, ϑ of the normal
variables to a new state defined by the values (α + Dα), (α + Dβ), . . . , (λ + Dλ),
(ϑ + Dϑ) of the normal variables.
12.7 Specific Heat Under Constant Actions 207
Differentiating the equations of equilibrium
∂F ∂F
− A = 0, ..., −L=0
∂α ∂λ
yields
⎧ 2
⎪
⎪ ∂ F ∂ 2F ∂ 2F
⎪
⎪ Dα + . . . + Dλ + Dϑ = 0,
⎪
⎪ ∂α 2 ∂α∂λ ∂α∂ϑ
⎨
................................................, (32)
⎪
⎪
⎪ 2
⎪
⎪
⎩ ∂ F Dα + . . . + ∂ F Dλ + ∂ F Dϑ = 0.
2 2
⎪
∂λ∂α ∂λ 2 ∂λ∂ϑ
These equations give Dα, Dβ, . . . , Dλ values proportional to Dϑ, the coefficient
of proportionality depending on the variables α, β, . . . , λ, ϑ.
Consequently, the quantity of heat absorbed in the preceding change,
Rα Dα + . . . + Rλ Dλ + CDϑ,
can be written Dϑ, being a function of α, β, . . . , λ, ϑ.

[282] is what we will call the calorific capacity under the constant actions
A, B, . . . , L. It is the calorific capacity relative to the system of non-normal variables4
A, B, . . . , L, ϑ.
By definition, we have
( − C)Dϑ = Rα Dα + . . . + Rλ Dλ,
or (Chapter 9, Equations (7a))

F(ϑ) ∂ 2F ∂ 2F
( − C) Dϑ = − Dα + . . . + Dλ (33)
E F (ϑ) ∂α∂ϑ ∂λ∂ϑ
But from Equations (32) it is easy to deduce the equation
∂ 2F ∂ 2F ∂ 2F
(Dα) 2
+ . . . + (Dλ) 2
+ 2 Dμ Dν
∂α 2 ∂λ2 ∂μ ∂ν

∂ 2F ∂ 2F
+ Dα + . . . + Dλ Dϑ = 0 .
∂α∂ϑ ∂α∂ϑ
4 Concerning
the system of variables, see our Mémoire sur les équations générales de la
Thermodynamique, Ch. V (Annales de l’École Normale, 3rd. series, vol. VII, 1891).
Comparing this equation with Equation (33) yields

⎧
⎪
⎪ ( − C) (Dϑ)2 =
⎨
2 ∂ 2F (34)
⎪
⎪ F(ϑ) ∂ F ∂ 2F
⎩ − (Dα) + . . . + 2 (Dλ) + 2
2 2
Dμ Dν .
E F (ϑ) ∂α 2 ∂λ ∂μ ∂ν
It the system is in stable equilibrium when the external actions A, B, . . . , L are
held constant, the quantity
∂ 2F ∂ 2F ∂ 2F
(δα)2 + . . . + 2 (δλ)2 + 2 δμ δν
∂α 2 ∂λ ∂μ ∂ν
is, as we have seen in §4, positive whatever δα, δβ, . . . , δλ may be. Equation (34)
therefore gives us the inequality
− C > 0.
Consequently, [283] if a system is in stable equilibrium when the external actions

affecting it are held constant, its calorific capacity is greater when the actions A, B,
. . . , L are held constant than when the normal variables α, β, . . . , λ are held
constant.
Suppose that the variations Dα, Dβ, . . . , Dλ are imposed on the variables
α, β, . . . , λ while the temperature ϑ is held constant. It would be necessary to add
to the external actions A, B, . . . , L the perturbing actions δA, δB, . . . , δL. We would
have, in accordance with Equation (10),
⎧
⎨ δA Dα + . . . + δL Dλ =
⎪
∂ 2F ∂ 2F ∂ 2F (35)
⎪
⎩ (Dα) 2
+ . . . + (Dλ) 2
+ 2 Dμ Dν.
∂α 2 ∂λ 2 ∂μ ∂ν
Suppose that the same variations Dα, Dβ, . . . , Dλ were imposed on the variables
α, β, . . . , λ by an isentropic change. The temperature will then vary by an amount
dϑ defined by the equation
∂S ∂S ∂S
Dα + . . . + Dλ + dϑ = 0. (36)
∂α ∂λ ∂ϑ
In order to accomplish this isentropic transformation, the perturbations dA,
dB, . . . , dL must be added to the external actions and we have, in accordance with
Equation (30),
⎧
⎪
⎪ F(ϑ) ∂ 2F
⎪ dA Dα + . . . + dL Dλ =
⎨ C (dϑ) 2
+ (Dα)2
E F (ϑ) ∂α 2
(37)
⎪
⎪ ∂ 2F ∂ 2F
⎪
⎩ + . . . + 2 (Dλ)2 + 2 Dμ Dν .
∂λ ∂μ ∂ν
Equations (34), (35) and (37) yield
12.7 Specific Heat Under Constant Actions 209
E F (ϑ)
( − C) (Dϑ)2 = δA Dα + . . . + δL Dλ,
F(ϑ)
E F (ϑ)
C(dϑ)2 = (dA − δA) Dα + . . . + (dL − δL) Dλ.
F(ϑ)
From [284] this it can be deduced that
2
−C Dϑ δA Dα + . . . + δL Dλ
= (38)
C dϑ (dA − δA) Dα + . . . + (dL − δL) Dλ
This relation, where Dϑ satisfies Equations (32) and dϑ satisfies Equations (36),
is the generalisation of that by which Laplace explained the experiment of Désormes
and Clément.
Conclusion
Physicists who have treated thermodynamics have placed this science into two
distinct positions vis-à-vis dynamics.
The founders of thermodynamics have nearly all been inclined to make this sci-
ence an application of dynamics. Regarding heat as a very small and rapid motion
of the particles constituting a body, the temperature as the average kinetic energy
of this motion, and changes of physical state as changes of characteristic elements
of this motion, they have tried to deduce the theorems of thermodynamics from the
theorems of rational mechanics. Their efforts were readily crowned with success in
the domain of the principle of the conservation of energy. They were less successful
when they grappled with Carnot’s principle. Despite the bold attempts of Clausius,
Boltzmann and H. von Helmholtz, Carnot’s principle has not so far been deduced in
a completely satisfactory manner from the propositions of dynamics.
Many physicists have sought to render thermodynamics independent of all
hypotheses about the nature of heat. They have tried the establish it, not on theo-
rems obtained from rational mechanics, but on principles of its own. Clausius was
guided in the edition of his very beautiful Memoirs by the desire to make thermody-
namics an independent science. G. Kirchhoff, as witnessed by his recently published
Leçons, has shown in his teaching that this wish could be realised. G. Lippmann, a
student of Kirchhoff, has recommended [285] this approach in France, and today it
dominates in the teaching in our Faculties.
We have tried in the present work to suggest a third position of dynamics in
relation to thermodynamics. We have made dynamics a particular case of thermo-
dynamics, or rather, we have constituted, under the name thermodynamics, a science
which covers in shared principles all the changes of state of bodies, including both
changes of position and changes in physical qualities.
The principles of this science are the experimental laws that Sadi Carnot, Mayer,
Joule, Clausius, W. Thomson and Helmholtz have established or made clearer. We
have put the equations of this science, first outlined by Clausius, and perfected by
Massieu, Gibbs and Helmholtz, into an analytic form like that Lagrange gave to
mechanics. A continuity of tradition is thus maintained in the course of the evolution
of Science which assures progress.

DOI 10.1007/978-94-007-0311-7, C Springer Science+Business Media B.V. 2011
212 Conclusion
It seems to us that a general conclusion arises from this study. If the science of
motion ceases to be the first of the physical Sciences in logical order, and becomes
just a particular case of a more general science including in its formulas all the
changes of bodies, the temptation will be less, we think, to reduce all physical phe-
nomena to the study of motion. It will be better understood that change of position
in space is not a more simple change than change of temperature or of any other
physical quality. It will then be easier to get away from what has hitherto been the
most dangerous stumbling block of theoretical physics, the search for a mechanical
explanation of the Universe.
Index
A Centigrade scale, 116

Absolute temperature, 14 Change(s)
Actions, 68, 144 adiabatic, 10, 89
Additional hypothesis, 100 allotropic, 25
Ammonium carbamate, 26 athermic, 89, 108
Atoms, 38 endothermic, 89
Available energy, 143 exothermic, 89
irreversible, 96
B isentropic, 208
Becquerel, 32 isothermal, 10, 90, 181
Berthelot, 25 of motion, 56
Bilateral connections, 173 physically realisable, 41
Bodies, 39 of position, 157, 212
enclosed within closed surface, 45 real, 41, 87, 92
isolated, 39 realisable, 41
system of, 43 reversible, 94, 108
Boltzmann, 211 adiabatic, 198
Braun, 32
real, 96
Brillouin, 123
realisable, 95
virtual, 95
C sequence of equilibrium states, 92
Cahours, 31 simultaneous and independent, 91
Calcium carbonate, 30 of state, 56, 157
Caloric capacity, 125
of temperature, 212
coefficients, 125, 144, 159, 166, 189
unrealisable, 41
under the constant actions, 207
virtual, 41, 85, 94
equivalent of work, 4
Characteristic function, 26, 30, 143
Calorimeter, 80
Calorimetry, 188 Chemical mechanics, 23
Carnot cycles, 6, 100, 102, 107 Chemical statics, 25
real and irreversible, 113 Classical mechanics, 138
realisable and irreversible, 113 Clausius, 5, 20, 23, 121, 162, 211
reversible, 9, 108 hypothesis, 99, 104
simultaneous and independent, 92 postulate, 5
Carnot, 5 theorem, 136
principle, 211 Closed cycle, 6, 43
theorem, 108, 117, 118, 121 Combination, 38, 94
Cauchy, 181 Compensated work, 162
Celestial mechanics, 165 Continuously variable transformation, 95
213
214 Index
Cooccupancy, 38, 93, 172 H

Coordinate system, 37 Heat, 4
Cycle quantity of, 75
closed, 86 unit of, 81
indefinitely reproducable, 87 Helmholtz, 1, 29, 33, 211
real, 87 Helmholtz’s Postulate, 200
virtual, 87 Hirn, 82
Hortsmann, 25, 31
D
Debray, 30 I
Désormes and Clément, 209 Impenetrability of matter, 38
Dissociation, 23, 25, 30 Independent magnitudes, 40
Dynamics, 211 Independent systems, 44
classical equations of, 164 Influences, 68
classical problem of, 165 Internal energy, 4
Isambert, 31
E
Edlund, 32 J
Joule, 82, 211
Effective actions, 161
Effective work, 161
Electrodynamics, 123 K
Electromotive force, 32 Kinetic energy, 163
Endothermic, 89 Kirchhoff, 24, 211
Energy, 55
internal, 57 L
kinetic, 57, 62 Lagrange, 1, 22, 78, 164, 211
potential, 57 Laplace, 4, 196
Entropy, 20, 25, 138 Lavoisier and Laplace, 196
Equations of motion, 188 Laws of dynamics, 78
Equilibrium, 9, 154, 157, 193 Le Chatelier, 198
necessary and sufficient conditions Lejeune-Dirichlet, 22, 192, 199
for, 144 Lippmann, 117, 211
stable, 208
subject to bilateral connections, 177 M
Equivalence of heat and work, 75, 188 Magnetic hysteresis, 96
Ether, 39, 74 Magnitudes, 39
dependence between, 40
independent, 40
F
Massieu, 26, 30, 143, 145, 154, 211
Faraday’s law, 122
Material system, 46
Favre, 32
Mayer, 211
First law of thermodynamics, 4 Mechanical equivalent of heat, 4, 32, 63
Forces, 68 Mechanical explanation, 212
Fourier, 181 Mechanical work, 51
Free energy, 29 Mechanics, 21, 211
Freie Energie, 143 Metaphysics, 52
Mixture, 38, 42, 172
G Moser, 33
Gauss, 21, 181 Motion, 157
Gibbs, 1, 20, 27, 29, 33, 198, 200, 211 absolute, 37
fundamental theorem, 20 relative, 37
theorem, 29 Moutier, 23, 28, 136
Index 215
N material, 46
Naumann, 31 motion of, 42
Non-normal variables, 207 nature of, 41
Normal variables, 153, 167, 171, 193 with non-uniform temperature, 185
at rest, 43
P stable, 200
Parts state of, 41
contiguous, 171 without viscosity, 163
independent, 145, 165–167, 190
non-independent, 172 T
at same temperature, 151 Temperature, 46, 157, 159
Passive resistances, 158 absolute, 116
Poincaré, 170 average kinetic energy, 211
Poisson, 4 Thermochemistry, 25
Principle Thermodynamic potential, 22, 30
of isentropic displacement from under constant actions, 192
equilibrium, 206 at constant entropy, 200
of maximum work, 25 at constant temperature, 200
of virtual velocities, 21 internal, 144, 175
total, 191
Q Thermometer, 49, 116
Quantity of heat Thomson, J., 24
system exhibiting bilateral Thomson’s
connections, 185 hypothesis, 99, 106
postulate, 6
R Thomson, W., 6, 211
Raoult, 32 Total calorific effect, 89, 118
Rayleigh, 25 Troost, 31
Regnault, 24
Rest, 157 U
Reversible Uncompensated transformation, 20
carnot cycle, 11 Uncompensated work, 20, 162
transformation, 9, 94 Unilateral connections, 173
virtual displacements, 173 Universe
bounded or not?, 45
S
Sainte-Claire Deville, 23, 31 V
Specific heat, 160 Van ’t Hoff, 196
Stability Virtual change, 41
of equilibrium, 25 Voltaic cell, 32
isentropic, 201
isothermal, 201 W
Statics, 1 Work
System(s) of inertial forces, 85
energy of, 55, 57 mechanical, 81
equilibrium, 43 virtual, of the inertial forces, 75
independent, 44 Work done, 68

(Boston Studies in The Philosophy of Science 277) Paul Needham (Auth.), Paul Needham (Eds.) - Commentary On The Principles of Thermodynamics by Pierre Duhem-Springer Netherlands (2011)

Uploaded by

Copyright:

Available Formats

(Boston Studies in The Philosophy of Science 277) Paul Needham (Auth.), Paul Needham (Eds.) - Commentary On The Principles of Thermodynamics by Pierre Duhem-Springer Netherlands (2011)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Boston Studies in The Philosophy of Science 277) Paul Needham (Auth.), Paul Needham (Eds.) - Commentary On The Principles of Thermodynamics by Pierre Duhem-Springer Netherlands (2011)

Uploaded by

Copyright:

Available Formats

COMMENTARY ON THE PRINCIPLES

OF THERMODYNAMICS BY PIERRE DUHEM

ROBERT S. COHEN, Boston University

LINDY DIVARCI, Max Planck Institute for the History of Science

THEODORE ARABATZIS, University of Athens

Edited and Translated, with an Introduction, by

Library of Congress Control Number: 2011929132

© Springer Science+Business Media B.V. 2011

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Duhem (1886)1 exploits the application of Euler’s theorem on homogeneous

Symbol Meaning Page reference

E Mechanical equivalent of heat p. 173 (1887)

Part I Study of the Thermodynamic Works of J. Willard

Part II Commentary on the Principles of Thermodynamics:

5 Work and Quantity of Heat . . . . . . . . . . . . . . . . . . . . . . 65

Part III Commentary on the Principles of Thermodynamics:

Part IV Commentary on the Principles of Thermodynamics:

10 Properties of a System in Motion . . . . . . . . . . . . . . . . . . . 157

In 1875 Gibbs published On the equilibrium of heterogeneous substances, a mem-

Mathématiques, 11 (1887), 122–148, 159–176.

Under these circumstances is seems appropriate to return to the very foundations

1.1 The Principle of the Equivalence of Heat and Work

P. Needham (ed.), Commentary on the Principles of Thermodynamics 3

An infinitesimal change of the system considered is accompanied by the release

1.2 Clausius’ Postulate

1.3 The Carnot Cycle and the Postulate of Sir W. Thomson

We will subject Sir W. Thomson’s postulate to a very slight modification,

1.4 Carnot’s Theorem and Absolute Temperature

From this follow the propositions:

for cycles described between the same temperatures corresponding to a negative

From this inequality, we deduce

10 The notation |W2 | signifies the absolute value of W2 .

In order to arrive at this important proposition, it is necessary for us to introduce

and for the second,

It [138] remains, in order to demonstrate Carnot’s theorem, to show that

ψ(ϑ , ϑ) = ψ(ϑ , ϑ)ψ(ϑ , ϑ), (5)

ϑ, ϑ , ϑ being any three temperatures.

In order to establish this proposition, it suffices to observe, by means of a

The function F(ϑ) is evidently determined only up to a positive constant factor.

We will demonstrate that it always holds that

J21 + J21 > 0. (9)

This is quite easily done in the following way.

Similarly, there exists either a state A 2 in the series

The [145] inequality (13) becomes

and can be formulated in the following way:

This quantity F(1, 2) is the value of the integral

By very simple considerations which are unnecessary to present here, we arrive

It is this function S, completely defined up to an additive constant as above, that

then the preceding proposition becomes:

1.6 Uncompensated Work, Thermodynamic Potential

the uncompensated work.

We will be able to state the preceding proposition in a form like this:

This theorem is analogous, in its statement, to the principle of virtual velocities

Summarising, we have the expression

= E[U1 − F(ϑ)S1 ] − E[U2 − F(ϑ)S2 ] + W. (16)

 = E[U − F(ϑ)S] + V, (18)

the equality (16) becomes

18 [Duhem uses the symbol “W” for the potential.]

= E[U − F(ϑ)S] + V, (18)

The uncompensated work will be determined by means of the quantity just