THE PRINCIPLE OF CARATHÉODORY

Author(s): R. EISENSCHITZ
Source: Science Progress (1933-) , April, 1955, Vol. 43, No. 170 (April, 1955), pp. 246-260
Published by: Sage Publications, Ltd.

Stable URL: https://www.jstor.org/stable/43423527

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to
Science Progress (1933-)

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY
BY R. EISENSCHITZ, D.SC., F.Inst.P.
Reader in Theoretical Phy»ic8 in the University of London , Queen Mary College

1. Introduction

The pioneers of thermodynamics formulated the second law

terms of the indicator diagrams of heat engines€ Cycles were us
for expressing basic principles and deriving specific results ; e
tually their use was confined to the derivation of general equatio
from which further deductions were made by mathematical meth
In this standard approach to thermodynamics natural phenome
are presented as if they were the consequence of certain experien
in engineering. In rectifying that anomaly the properties of mat
in thermal equilibrium should be deduced from laboratory ex
ments without depending on technological concepts.
Carathéodory (1909) and other authors (Born, 1921, 1948 ;
Lande, 1926) have shown that the second law can, in fact, be dedu
from simple and familiar phenomena, no appeal to cycles be
required. Their papers are somewhat encumbered by the emp
which is laid on an abstract mathematical theorem which is other-
wise of little use in physics. Their work has not yet evoked much
response. It will be better appreciated if mathematical deductions
are simplified, perhaps on lines recently developed by Buchdahl
(1949).
In the present review of the subject care is taken to avoid any
concept, argument or terminology that is too remote from established
practice in thermodynamics.

2. Fundamental Assumptions

Following the example of Clausius and Kelvin, Carathéodory

derives the second law from a principle according to which certa
virtual processes are prohibited.
Let a state of equilibrium in a thermodynamical system, briefly
a state, be determined by the values of two or more independen
246

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY 247

variables,* the volume (V) the temperature (0) as measured o

arbitrary scale (mercury in glass, thermocouple), concentra
strains and others. The pressure and other dependent varia
are determined by equations of state.
" Processes " are supposed to connect an initial and a final sta
Intermediately conditions will, in general, involve non-uni
density, pressure or temperature and cannot be fully specif
terms of the independent variables. Experience shows that
cesses of this kind are " irreversible," implying that the initial
of the system cannot be recovered or can be recovered onl
conditions in the surroundings of the system are subjected
changes. There is a special type of process in which all i
mediate stages are fully determined by the independent var
Since changes in these variables can be reversed ad libitum ,
processes are termed reversible. In reversible processes the
done by the system is
dW = P dV + Xdx + Ydy + . . .
where the pressure P and the quantities X, Y . . . are functions
of the independent variables 0, V, x, y . . .
In this article free reference is made to the concepts of heat (Q)
and internal energy (U) and the first law
dQ = dU + dW
Adiabatic processes are specified by the condition dQ = 0.
The principle of Carathéodory says : " It is impossible to reach
all states in the neighbourhood of any arbitrary initial state by an
adiabatic process."
The preference shown here for adiabatic processes is not as
arbitrary as it may appear. It is due to the fact that irreversibility
is more conspicuous in adiabatic than in other processes. Experi-
ence shows that the initial state of a system cannot be recovered by
any adiabatic process, if an adiabatic irreversible process has taken
place. In contrast, the initial state of a system can be recovered
by an isothermal process after an isothermal irreversible process ;
in this case changes in the surroundings of the system are taking place.
* In this article the term " degrees of freedom " is used for the number
of independent thermodynamical variables. For a simple liquid or gas this
number is two, for a fluid binary mixture it is three, for other systems it
may be larger. If two systems of m and n degrees of freedom axe brought
into thermal contact so that they form a composite system, the resulting
number of degrees of freedom is n + m - 1. The present definition of the
degrees of freedom is similar to the definition employed in formulating the
phase rule but, in distinction from the latter, the number of separate phases
is here not taken into account.

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 248 SCIENCE PROGRESS

It follows that the mere possibility of an adia

ition between two states is sufficient for e
reversible transition between these states.
cesses no similar statement could be made.*
Every adiabatic irreversible process accordingly provides experi-
mental evidence in favour of Carathéodory's principle. It also
provides evidence for the impossibility of reaching all states in the
neighbourhood of an initial state by an adiabatic reversible process.
An important statement regarding reversible processes accordingly
depends on evidence derived from irreversible phenomena, f
The arguments to follow are mainly concerned with reversible
processes. The greatest possible measure of neighbouring states is
arrived at by independent increments to all independent variables.
Conditions imposed on the increments, such as dU = 0, diminish
the degrees of freedom and the number of accessible states. The
condition dQ = 0 is of a different kind ; the choice of a final state
does not determine the amount of heat absorbed, for U, not Q,
is a function of the independent variables. According to Cara-
théodory's principle the condition dQ = 0 prevents accession to
some of the neighbouring states without necessarily diminishing
the degrees of freedom.
As no criterion is offered for distinguishing between admissible
and prohibited processes, the principle of Carathéodory will be
studied by attempting to violate it. The attempt may be precluded
by the first law or on account of some property of matter that has
not yet been considered. Its success would clash with the fact
that irreversible processes are known to exist.
Before starting with the main argument it may be helpful to
say a few words with regard to the mathematical method and nota-
tion. Thermodynamical quantities will be expressed as functions
of other thermodynamical quantities. P, U and other, possibly less
familiar, quantities (denoted by al9 a2 . . .) are, at first, defined

* Example . Diffusion of one liquid into another is irreversible. Assume

that the energy of the components is less than the energy of the mixture.
After isothermal diffusion has led to uniform concentration, the initial state
can be recovered by means of isothermal compression with a semi-permeable
piston. The piston can also be used for performing a reversible process of
mixing. After adiabatic diffusion, involving lowering of temperature,
separation of the components by means of the piston raises the temperature
beyond the initial value, so that the initial state is not restored ; reversible
mixing by means of the piston leads to a temperature lower than the final
temperature after the irreversible process.
t This peculiar relation between irreversible and reversible processes is
invoked also in the traditional presentations of thermodynamics.

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY 249

or given as functions of 0, V and other independent vari

Substitutions of alternative variables in these functions have
proved to be a very powerful mathematical tool and will be fre-
quently made. The pressure will, for instance, be expressed as a
function of U and V, or the quantity ax as a function of 0 and a2.
In accordance with the common usage in thermodynamics (as distinct
from mathematics) the symbols P, U . . . will be used for denoting
physical quantities rather than functional relations. In expressing
the pressure in terms of two different sets of independent variables
we put
P(0,V) = P(U,V)
using on both sides of the equation the same symbol for the pressure
although the functional relations * on the two sides of the equation
are different. Equally
= ®i(0>®2)
Of course this notation requires careful specification of partial
differential coefficients because (3P/3V)0 differs from (3P/3V)U.

3. Constraints due to Adiabatic Enclosures

In a simple fluid where 0 and V are the only independent

ables, adiabatic reversible processes have to comply wit
condition

dQ = d'J + PdV = 0
where U and P are functions of 0 and V. Substituting U for 0
as an independent variable, the differential equation

^=-P(U,V) .... (1)

is obtained ; it establishes a relation between U and V and can be
solved in terms of a function (a) of the thermodynamical variables
in the form

a(U,V) = const.
where a, originally a function of U and V, can be expressed as a
function of 0 and V, giving
cr(0,V) = const. . . . (2)
as the solution of equation (1). It follows that a c
of the temperature and the volume has the proper
* Example . For an ideal monatomic gas, 0 being ° C.,
P(0,V) = NR(0 + 273)/V ; P(U,V) = 2U/3V

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
250 SCIENCE PROGRESS

constant during adiabatic reversible proce

not affected by the fact that the function a
Borrowing a term from analytical mecha
said to be an " adiabatic invariant." The loci of constant a are
familiar and known as adiabatic lines.
On adiabatic lines da = 0 and d Q = 0. Regarding both, dQ
and da , as limits of measurable increments, the ratio da/dQ defines
a quantity (denoted by A) which will depend on the thermodynamical
variables and may, for instance, be written as A(0,V). According
to this definition

da/dQ = A(0,V)
Since a is a function of the thermodynamical variables whereas Q
is not, it follows that the increment of the reversible heat can be
made a perfect differential by means of the integrating factor A :
A(0,V)dQ = dor(0,V) .... (3)
da is, of course, a perfect differential also in non-adiabat
Every state of the fluid is situated on one and only on
line ; neighbouring states situated on different adiaba
not accessible by a reversible adiabatic process. The p
Carathéodory appears then to be a consequence of the
the constraint imposed by an adiabatic enclosure amo
loss of a degree of freedom. As a further consequence th
of an integrating factor for the increment in reversible h
lished ; it might even be possible to define the quanti
a way that it is equal to the entropy.
The significance of these results is, however, strict
applying to a single fluid or, perhaps, to two single fluid
completely isolated from each other. The results obtained
apply if these two fluids can exchange heat. In this case i
taken for granted that " adiabatic surfaces " exist, in
the adiabatic lines of a single fluid. A further invest
required before the scope of the above results can be esta
Consider now a composite system consisting of tw
different containers, so that they may have different
the containers are in thermal contact so that the tem
the fluids is equal. There are three independent varia
and V2. Adiabatic reversible processes have to comply
condition

dQ = dQx + dQ2 = (da1/X1) + (dcr2/A2) = 0 . . (4)

where alf a2, A1} À2 are defined in accordance with equation (3)
as functions of 0 and Vx and functions of 0 and V8 respectively.

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY 251

Introducing o1 and a2 as independent variables in plac

volumes, equation (4) is transformed to
&ai + p{<*i><*2fi)do% = 0 . (5)
where jx - ^1/^2*
Consider at first two types of processes.
Type 1. The first fluid and the second fluid are simultane
expanded (or compressed) along their adiabatic lines so that
temperatures remain equal. In this case dax = da2 = 0 so
equation (5) is automatically satisfied.
Type 2. Both fluids are subjected to an isothermal chang

Fig. 1. - Isothermals of the two-fluid system.

volume (dd - 0) on condition that the heat released by one

is absorbed by the other. A quantitative formulation of
condition is found by solving equation (5) with 0 constant.
solution is expressed in the form
^12(^1,0*2,0) = const. ; 6 = const. . . (6)
where the function a12 is related to the above function ¡u according to
{
/ r' ' ' da2Jolf8
^((Ti,(T2,0) / r' ' - T^ā~' " *
' )azte
* According to (6) da12 = (da12/dax)dax + (5ťT12/5(r2)cřflr2 = 0. Hence
- (da1/do2) is equal to the right-hand side of (7). In order to agree with
(5) this differential coefficient must be equal to - ļu> thus proving the validity
of equation (7).

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
252 SCIENCE PROGRESS

In processes of this type a2 depends on

shown in Fig. 1, where full and broken l
temperatures.
Every adiabatic reversible process in the
resolved in successive processes of these t
If one fluid is subjected to a process along
the state of the whole system is fixed by th
that fluid proceeds to a specified final st
optional, the final state of the whole system
As far as the first law is concerned, it wi
states are accessible.
A process is readily produced which is adiabatic and reversible
and connects in four stages any arbitrary pair of neighbouring
states. In Fig. 1 the initial and the final values of the variables
ax and a2 are represented by the points A and B. In the first
stage the system proceeds at the initial temperature in a process of
Type 2 along a full-line isothermal from A to C. In the second
stage the temperature is varied in a process of Type 1 ; in Fig. 1
the state of the system remains at the point C. This stage is ter-
minated at that temperature at which the isothermals are repre-
sented by the broken lines in Fig. 1. In the third stage the system
proceeds in a process of Type 2 along an isothermal from point
C to point B. In the fourth stage the temperature is varied in a
process of Type 1 to an arbitrary final value.
If this process can be carried out the principle of Carathéodory
is violated ; if the principle is to prevail it is necessary to make an
assumption on account of which the process is inhibited. Only
one assumption of this kind is possible. The isothermals can be
independent of the temperature so that, in Fig. 1, the full lines
correspond to all temperatures. Then the point B is no longer
accessible from the point A, in accordance with the principle of
Carathéodory.
It follows that the quantity a12 and, on account of equation (7),
the quantity fi are independent of the temperature, provided they are
expressed as functions of ox and o2. In consequence, equation (5) is
turned into an ordinary differential equation to be solved in the form
^12(^1^2) = const. .... (8)
a solution which - as distinct from the sugge
valid for all adiabatic reversible processes. B
variables 0, Vl9 V2 for ax and a2 , adiabatic surfa
being the loci of constant a12 :
o'i2(ö»V1,V2) = const

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY 253

c r12 is recognised to be an adiabatic invariant of the co

system.
In adiabatic reversible processes da12 = 0 and dQ = 0 ; from
this fact it follows, in the same way as for single fluids, that the
increment in reversible heat can be made a perfect differential by
means of an integrating factor [A12(0,V1,V2)]
X12dQ := d(j 22 .... (10)
do12 being a perfect differential, also in non-adiaba
These results are a consequence of the principle of
which cannot be derived from the first law. Its
shown to depend on the equation

h) =o
' 90/ alfa¡

being satisfied for all values of ax and <r2. Since ¡x depends on the
properties of two different substances this condition is far from
obvious and, as it looks like a conspiracy, certainly involves a law
that had, so far, not been taken into account.

4. Entropy and Absolute Temperature

Continuing with the investigation of the two-fluid system

consider now the physical significance of the thermodynam
quantities a and A which were defined in the preceding sect
Aj and a1 were originally considered to be functions of 0 and
A2 and <x2 were considered to be functions of 0 and V2. Results
the preceding section suggest substitution of 0, al9 and a2 as i
pendent variables for specifying the properties of the two-f
system. Then Xx is a function of 0 and al9 Ã2 is a function of 0
<r2. A12 is a function of the three variables.
The reversible heat

dQ = dQļ -ļ- dQ2

is readily expressed in terms of the quantities A and a. Applying
equation (10) to the left-hand side and equation (3) to the right-
hand side of the last equation it is found that
^12/^12 = (^°i/^i) (^2/^2) • • (11)
According to equation (8) a12 is independent of the tempe
when it is expressed as a function of ox and <r2. The deriv
(3 o,12/3(T1)<ya and (da12/da2)0ļ are therefore also independent
temperature ; they are readily obtained from equation (11)
doi2/dcr1 = A12/Ax
dal2/da2 = A12/A2

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 254 SCIENCE PROGRESS

It follows that the three quantities A must

perature in such a way that their quotients
temperature. That is possible only if the t
Ál9 A2 and A12 by the way of a factor which
three quantities and which will be denoted
the A by 0 must result in a quantity which
temperature (to be denoted by Llf L2 an
account of the variables on which the A m
that *

Ai(0,<Ti) = 0{d),L1(a1) ; A2(0,<t2) = 0(0)

Ai2(0, 0*1,0^) = 0(0). L ^(01,0^) . . . (12)
A relation between the newly introduced quantities L is obtained
by entering the expressions of (12) into equation (11). The common
factor 0 cancels out and

d<*i2/Li2(ovo2) = 'da1/'j1(aiy' + [d<r2/L8(<r8)] • • (13)

Inspection of the right-hand side of (13) shows that it is the perfect
differential of a function of ox and a2. The left-hand side must
accordingly also be a perfect differential. This is possible only if
L12 depends upon ax and a2 by the way of aXi(al9at).
The increment in reversible heat of the composite system and
of either component is now resolved into two factors. The first
factor, depending on the temperature, is identical for the composite
system and the constituents ; anticipating results of the next
section, this factor, the quantity 0, may be said to be a universal
function of the temperature. The second factors are differentials
of expressions depending only on the adiabatic invariant of the
composite system, the first constituent and the second constituent
respectively.
In accordance with established tradition, the expression 1/0 is
called the absolute temperature and denoted by T. Alternative
conventions are of course not ruled out.
At this stage it should be remembered that the quantities av a2
and a12 are not defined uniquely ; all results obtained remain valid
if the original o is replaced by a function of a which is to some
extent arbitrary. A particular choice of the quantity a will deter-
mine a particular function L(or). These ambiguities are now
eliminated by the way of a convention. Define the entropies of

* This resolution into factors, of Xlt A2 and Aia, is only possible when these
quantities are represented as functions of the temperature and ox and <r2.
It would, in general, not be true to say that A is the product of a function
of the volume and a function of the temperature.

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF C ARATHÉOD OR Y 255

the constituents and of the composite system, S2(<r2) and

Si2(°ri2) in such a way that they obey a rule of additivity
dS12 = dSi + dS2 . . . . (14)
For this purpose one has to define
cZSļ = d(Ti/Li((Ti) j d&2 == = ^^12/^12(^12)
The increment in reversible heat is then expressed in the
dQx = TdSx ; dQ2 = TdS2 ; dQ = TdS12
Generalising, in anticipation of the next section, it is found that in
every thermodynamical system an entropy (S) can be defined as a
function of the independent variables which is an adiabatic invariant
and obeys a rule of additivity, in accordance with the equation
dQ = TdS . . . (15)
which is valid in all reversible processes.
The second law of thermodynamics
follows : " The increment in reversible h
differential by means of the integrating f
These results have been derived from the
dory.

5. General Thermodynamical System

In justifying the final conclusions of the preceding section it is

necessary to consider thermodynamic systems of a more genera
kind.
As an example of a general trivariant system, consider an elastic
solid in which a component of strain (a?) is an independent variable
in addition to temperature and volume. In reversible adiabati
processes

dQ = dU + PdV + Xdx . . . (16)

where X is a component of stress.
Two classes of restricted processes are considered firs
cesses at constant strain and processes at constant volum
either class the existence of an adiabatic invariant o^d^Y
is readily established. They and the corresponding
grating factors À1 and À2 are then used for transforming eq
(16) to the form of equation (5). Deductions from this eq
as presented in Sections 3 and 4, apply accordingly to the
system.
In considering a system with four independent variables, use is
made of the fact that every trivariant system has an adiabatic

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
256 SCIENCE PROGRESS

invariant <x12. The condition with which

cesses have to comply can then be writte
di Tg -f~ ļudcti 2 = 0

where [i is a function of aZi o12, cr2 and 0. T

Either the function depends on a2 and 0
ables) ; then all states in the neighbourho
accessible by ań adiabatic reversible proc
of 0*3 and o12 only. Then there exists an
and an integrating factor of the increm
Only the second alternative is compatibl
Carathéodory.
The argument is then repeated for system
of freedom. The further discussion pr
Section 4 by considering the adiabatic invar
factors in any composite system and its
Section 4 are accordingly justified.

6. Calibration of the Temperature Scale

It has as yet to be shown that the abstract quantities entropy

and absolute temperature can be derived from the results o
measurements.

Usually heat is defined in terms of calorimetric measurements,

implying that the temperature of the source of heat may be different
from the temperature of the calorimeter, so that the use of irreversible
transfer of heat is unavoidable. Energy is defined in terms of
heat and work.* The first law says that the increment of energy
depends only on the initial and final state. It is invoked in estab-
lishing the mechanical equivalent of heat.
Of these premisses the significance of entropy and absolute
temperature in terms of measurements is readily established, con-
veniently by using measurements with a simple fluid and thermo-
dynamical equations which are derived from equation (15), such as

T/T0 = exp (J6 (9P/d0)v/[P + (3U/av)e]de)

and
S - S0 = (Cp/T)dT
J To

The conventional approach to the first law and Carathéodory's

derivation of the second law are, accordingly, fully compatible.
* In order to avoid circular arguments it is necessary to define
dU = dQ - d A prior to introducing the mechanical equivalent of heat.
This is possible but somewhat involved.

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY 257

The preceding argument is at variance with Carathéodor

views. He claims that the concept of heat and, by imp
calorimetry is not a necessary prerequisite to the complet
thermodynamics but is rather in need of justification by t
Energy is, accordingly, defined in terms of work supplied
reversible or irreversible processes. The first law postu
energy is a function of the independent variables. Heat
merely as a verbal term for the sum of work done and the
in energy in non-adiabatic processes.

V
Fig. 2. - Adiabatic lines and isothermals of a fluid.

For calibrating the scale of absolute temperature, calorimetry is

admitted only to the extent that work is supplied and the increase
in energy is observed. Calorimetry by electric heating is, accord-
ingly, admitted but measurement of specific heats by mixing is
excluded. Otherwise measurement of temperature on an arbitrary
scale and of mechanical quantities is supposed to be sufficient.
Using a simple fluid, 3P/30 and 3V /90 are to be measured along
adiabatic fines. Introducing an entropy on an arbitrary scale (a),
dV /da and dV /do are to be measured along isothermals. From
these measurements the calibrated entropy S(<r) and absolute tem-
perature T(0) are to be derived. Combining the 1st and 2nd law
in the form of the equation
dV = TdS - PdV .... (17)
s

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 258 SCIENCE PROGRESS

the relation is derived

|(3V/30)(9P /da) - (dV/da)(3P/dO)' = '(d8/do)(dT/dO)' . (18)

This equation is an expression of the fact that, in Fig. 2, the in-
finitesimal parallelogram ABCD has the same area as dSdT.
Using the last equation and introducing by convention a unit
of temperature, dS/da and dT/dO are found by evaluating measure-
ments along an adiabatic Une and an isothermal. Subsequent
integration yields T and S as functions of P and V. An additive
constant S0 and an additive constant T0 remain undetermined, of
which only the second is of physical significance. As the next
step the energy has to be found by successive reversible and irre-
versible supply of work to the fluid within an adiabatic enclosure ;
if in these measurements the initial state of the fluid has a sufficiently
low entropy, the energy is obtained over a wide range of the inde-
pendent variables and can be used for determining the constant T0.
Energy is found in mechanical units, without any determination
of the mechanical equivalent of heat.
The laws of thermodynamics are accordingly established with-
out using the concept of heat. Experimental tests, however, would
be incomplete without calorimetry and, in fact, without the use of
cycles. These devices have, however, been removed to the conse-
quences of the theory instead of being part of its premisses.

7. Irreversible Processes

By admitting irreversible in addition to reversible adiab

processes the measure of accessible states is increased. The
ciple of Carathéodory says that even then not all states are acc
In a divariant system an irreversible process shifts the i
state from one to another adiabatic line. The shift must be thr
out in the direction of decreasing or of increasing entropy, ot
wise the concept of irreversibility could not arise. Carathéo
principle leads to further consequences. It must, for instan
impossible for entropy to increase in some fluids irreversib
decrease in others. If that were so the two fluids could be combined
to form a composite system such as considered in Section 3. Making
use of irreversible processes of either kind the total entropy could be
increased or diminished so that all states could be reached in adia-
batic processes.
It follows that in all irreversible processes the entropy either
increases or decreases. A decision between these alternatives
requires an additional appeal to experience, such as recognising that

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
 THE PRINCIPLE OF CARATHÉODORY 259

conduction of heat in an adiabatic enclosure causes the entro

increase.
The above ambiguity is due to the fact that all premisses intro-
duced so far have taken no account of the direction of the irreversible
processes occurring in practice. The same ambiguity occurs in a
theory of cycles if the fundamental principle is formulated so that
all heat engines must have an equal maximum of efficiency. This
postulate provides the whole body of thermodynamics as far as
reversible processes are concerned and the alternative of increasing
or decreasing entropy in irreversible cycles (Ruark, 1925).
By establishing the irreversible increase of entropy the theory
of thermodynamics is completed.

8. Conclusions

In comparing the merits of the different approaches to thermo-

dynamics it may be noted that in the older formulations every sing
result could be checked against its virtual employment in providing
perpetual motion. The check could be made in a direct manner
by devising a suitable cycle. This advantage of the older approac
should, however, not be decisive, since checks on deductions fro
a theory can be made in many different ways. A correspondin
check in the thermodynamics of Carathéodory would have to
concerned with accessibility ; it could presumably be made in
convincing way but only indirectly, using the concept of entropy.
The second law, it may be argued, can be introduced by way
of a postulate in the form of equation (15), defining entropy an
absolute temperature simultaneously and without any precedin
appeal to empirical facts.
Specific results deduced from equation (15) would provide a
comprehensive experimental test of the second law. An approach
of this kind cannot be criticised from any point of view concerning
matters of principle. It is, in fact, in line with the attitude of
number of contemporary writers. This approach is, nevertheles
not likely to prevail, because the general principles of Clausius o
Carathéodory carry more conviction than numerous individual
results.
Alternatively, macroscopic thermodynamics might be regarde
as being obsolete and due for replacement by a dynamical or statis-
tical theory. This point of view is, for the time being, not accep
able. The complete elimination of the concepts of thermodynami
from the statistical theories is, if at all, only possible at the price o
reduced evidence for the fundamental assumptions. In addition

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms
260 SCIENCE PROGRESS

it may be noted, mechanical theories have

changes in recent decades, whereas the v
thermodynamics remains unimpaired and
the field of irreversible phenomena (cf. t
Davies, 1953).
Every approach to thermodynamics has
elusive nature of entropy which is not su
servation and cannot be measured directl
Carathéodory's approach seems to he in
entropy somewhat less elusive by introdu
as a property of matter and, for this reason
between macroscopic and statistical therm

References

Born, M. (1921), Phys. Zeitschr., 22, 218, 249, 282.

Buchdahl, H. A. (1949), Amer . Joum. Phys., 17, 41, 44, 212.

Carathéodory, C. (1909), Math. Ann., 67, 355.
Davies, R. O. (1953), Science News, 28, 41 (Harmondsworth).
Landé, A. (1926), Handbuch der Physik, 9, 281 (Berlin).
Ruark, A. E. (1925), Phil. Mag., 49, 584.

This content downloaded from

220.245.38.86 on Sat, 06 Apr 2024 06:16:57 +00:00
All use subject to https://about.jstor.org/terms