Você está na página 1de 25

Absolute Temperatures as a Consequence of Carnot's General Axiom

Author(s): C. Truesdell
Source: Archive for History of Exact Sciences, Vol. 20, No. 3/4 (24.VIII.1979), pp. 357-380
Published by: Springer
Stable URL: http://www.jstor.org/stable/41133546 .
Accessed: 14/06/2014 05:07

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
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.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Archive for History of Exact
Sciences.

http://www.jstor.org

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
AbsoluteTemperaturesas a Consequence
of Carnoťs GeneralAxiom
C. Truesdell

Dedicatedto JAMESSerrín

Contents
1. Program 357
Theorem
2. The GeneralCarnot-Clapeyron 360
3. ApparatusfromConceptsand Logic 360
4. Hotness and EmpiricalTemperature,I. Continuity 361
5. Hotness and EmpiricalTemperature,II. Differentiability 367
6. Kelvin's Absolute Temperatures 369
7. Consequencesof Part II of Carnoťs General Axiom 372
8. Thermometric Axiom 373
9. The PhysicalDimensionof Absolute Temperature 375
10. Conclusion 377
11. Appendix.Axioms of Classical Thermodynamics 377

1. Program

In our book, Conceptsand Logic of Classical Thermodynamics,1


Mr. BHAR-
ATHA and I developed classical thermodynamicson the basis of Part I of
Carnot's General Axiom,namely,themotivepowerofa Carnotcycleis positive
and is determinedby its operatingtemperatures and by the amountof heat it
absorbs.For a givenbody,then,thereis a function
G suchthatforanyCarnot
cycle#
L(%)= G(0+,0 ,C +(^j)>0. (1)
The domainofG is thesetofoperating temperaturesand heatsabsorbedthat
may appertain to Carnot cyclesforthe in
body question.It is part of the
ofa Carnotcyclethat6+>6~ and thatC+(^)>0. Thisdefinition
definition and
1 C. Truesdell & S.
Bharatha, ConceptsandLogic ofClassical Thermodynamics
as a
TheoryofHeat Engines,RigorouslyConstructedupontheFoundationLaid byS. Carnotand
F. Reech,N.Y. etc.,Springer-Verlag,
1977.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
358 C. Truesdell

this axiom make sense in terms of underlyingaxioms accepted tacitly or


expresslyby the pioneersregardingpiecewisesmoothprocessesundergoneby a
body whichexertspressurep and receivespositiveor negativeor null heatingQ
in responseto its temperature9, its volume V,and theirtime-ratesof change V
and Ô:
p = w(V,6), ^<0,
(2)

Q = AV(V,6)V+KV(V,8)Ô, Kv>0. (3)


The timeratesare assumedto existat all but a finitenumberof timesin the
intervalof time over whichthe processis defined.The threeconstitutive
w, Av,and Kv are definedand continuously
functions differentiable* overthe
domainQ),whichis a non-empty,
body'sconstitutive connected, opensubsetofa
The namesof w, Av, and Kv are pressure
realhalf-plane. function, latentheat
withrespectto volume,and specificheatat constantvolume. By definition,the
heatabsorbedC+ in anyprocessis thevalueoftheintegral ofQ withrespectto
timeoverthesetoftimeson whichQ>0.
For future I remark
reference herethatbysubstituting (2) into(3) we obtain
Q = Ap(V,O)p+ Kp(V90)Ô,
ÕW
A A IÕW K K A IÕW

Kp is the specificheat at constantpressure.


In our book we took 6 as beingwhatis sometimes called "the ideal-gas
temperature". Our attitudetoward it I later founddescribed well and with
evidentdisapproval Mach 2:
by
It is remarkable how muchtimepassed beforeit was understood thatto
represent hotnessby a number restedupona convention. In naturethereare
hotnesses,buttheconceptoftemperature existsonlythrough our arbitrary
whichmighthaveturnedout otherwise.
definition, Untilveryrecently those
whoworkedin thisfieldseemto havelookedmoreor lessunconsciously for
a naturalmeasureof temperature, fora real temperature, fora sort of
Platonicidea of temperature, of whichthetemperature read on a thermo-
meterwouldbe onlyan incomplete, inexactexpression.
As amongtheproponents ofthisto himreprehensible idea Mach was able to
citeClausius, I do notthinkwe wereat faultin holdingto it,especiallysince
byimputing to one idealgas a simpleproperty longknownto holdverynearly
formany natural
gases we were able to prove theorems
as notonlya traditional
"SecondLaw" butalso a traditional "FirstLaw".
2 E. Mach, Die der Wärmelehre, Historisch-kritisch
entwickelt,
Leipzig,
Prinzipien
Barth,1896.Most of the quotationsbelow are fromthe chaptercalled "Kritikdes
referred
hereinafter
Temperaturbegriffes", The quotationaboveis
to a Temperaturbegriff.
from§ 14.
* See thenoteaddedin proot,p. 380.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 359

Nevertheless, it was not necessaryforus to introducetheconceptof ideal gas


at all. Afterhavingseen Mr.Serrin's beautifulwork on hotnessand absolute
3
temperatureand afterhavingbeen stimulatedby his lettersand lecturesand by
our conversationson the subject for three years, I have perceived that to
develop the conceptof absolute temperatureupon the basis of Carnot's ideas
would be not only easy but even worthwhile.Kelvin's work4, also, I have
studiedin detail,and I give a fullaccount of it in my TragicomicalHistoryof
Thermodynamics, 1822-1854,now being polished forthe press. Here I wish to
remarkonly that Kelvin's work restsessentiallyupon Carnot's ideas alone
but is neitherclear nor explicitnor entirenor rightin all details. Clausius
attemptedto persuade his readersthathe had establishedan absolute tempera-
turein circumstancesfarmoregeneralthanwhatcan be describedin Carnot's
terms; his work, too, is described in my TragicomicalHistory,but I cannot
followtheargumentsthathe presents;indeed,I cannoteven distinguishwhathe
assumes fromwhat he claims to prove. It is my impressionthat Mr. Serrin's
theoryachieves what Clausius attempted,or, rather,the achievable part of
Clausius' inchoateprogram.The treatmentI give below seemsto me to clarify,
consolidate,confirm, correct,and completeKelvin's. Because I offerit only as
a modestprohistoricalstudy,I presentit informally, usingonlysuch mathemati-
cal conceptsas were accessible to any competentmathematicalscientistof the
1870s- concepts,however,some of which were not at the disposal of Kelvin
and Clausius in 1848-1854 and hence are not to be found in textbooksof
thermodynamics today. Use of the modern theoryof manifoldswould effect
brieferdeductionof resultsmore general than those below, but all I attempt
here is to motivatethe assumptionsand to show that elementaryarguments,
callingupon no generaltheoremsabout manifolds,would have sufficedto prove
everything that the pioneers sought in this domain. For a fullymathematical
treatmentmeetingmodernstandardsof rigorand generalitythe reader should
consultMr. Serrin's workand studiesby otherswhichI expectto be complete
soon.
3 J.B.
Serrín, Foundationsof Classical Thermodynamics,LectureNotes, Department
of Mathematics,University ofChicago, 1975,and "The conceptsofthermodynamics", pp.
411-451 of Contemporary Developmentsin ContinuumMechanics & Partial Differential
Equations(Proceedingsof the InternationalSymposiumon ContinuumMechanics and
PartialDifferential
Equations,Rio de Janeiro,August1977),editedbyG.M. de La Penha
& L.A. Medeiros, Amsterdam,North-Holland Publishing Co., 1978. Mr. Serrín
presentedfurther resultsin his Batemanlecturesat The JohnsHopkins Universityin May,
1978,and in subsequentlectureselsewhere.His workapplies to a class of materialsmore
generalthan thatconsideredhere.
4 W. Thomson
(laterLord Kelvin), "On theabsolutethermometric scale foundedon
Carnoťs theoryofthemotivepowerofheat; and calculatedfromRegnault'sobservations".
Proceedingsof the CambridgePhilosophicalSociety 1 (1843/1866),No. 5, 66-71 (1849)
= PhilosophicalMagazine (3) 33 (1849), 313-317= (withadded notes) pp. 100-112 of W.
Thomson's Mathematicaland PhysicalPapers,Volume 1, 1882.
J.P. Joule & W. Thomson, "On the thermaleffectsof fluidsin motion,Part II",
PhilosophicalTransactionsof the Royal Society(London) 144 (1854), 321-364= (withpp.
362^-364 excised), pp. 357^400 of W. Thomson's Mathematicaland Physical Papers,
Volume1,1882= pp. 247-299ofJoule's Scientific Papers,Volume2, 1887.See SectionIV of
"TheoreticalDeductions".

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
360 C. Truesdell

Essentialto my argumentis the apparatus providedby Conceptsand Logic.


Of course only those parts of it that survivewhen all mentionof ideal gases is
excised can be used for the purpose. Thus I have phrased(1)- (4) in termsof
some one empiricalscale of temperature, values of whichare denoted by 6, ö+,
and 0~. The rangeof thisscale is some open real interval.Until§7 all reasoning
will be local, will referonly to temperaturesin some subinterval,perhaps very
small,and only to someone body.

2. The General Carnot-Clapeyron


Theorem

Clapeyron, applyingto (1) an argumentthat Carnot had inventedbut


had used only forideal gases, determinedas followsthe workdone by a Carnot
cycle# withinfinitesimal differenceA6 of operatingtemperatures:

I4V)*IJi(0)C+(<g)AO. (5)
Here ¡a is "Carnoťs function",derivedfromG in (1) througha limitingprocess.
From (1) and (5) we mightthinkto conclude that /i>0, and the early authors,
Kelvin among them,seem to have done so. However,examinationof the limit
processshows thatonly a weakerstatementfollows:
A^O. (6)
We shall see that the distinction,altogetherunnoticedby the pioneers,bears
upon the conceptof absolute temperatures.
Theorem:
Essentiallyequivalentto (5) is the GeneralCarnot-Clapeyron

*Ar-W (7)

Hence we at once perceivethe constitutiveinequality

It followsfrom(4)3 that
Kp^Kv. (9)
We note thatKp = Kv ifand only ifAvdm/d0= 0.

3. ApparatusfromConcepts and Logic

Conceptsand Logic providesrigorousproofsof delimitedstatementsmore


preciseand more generalthan Carnot's and more specificthan those obtained
by Reech in a muchearlierand incompleteattemptto develop Carnot's ideas.
The normalset Q)nof the body representedby the constitutivequantitiesQ), m,
Av and Kv is the set of pointsP of Q) such that

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 361

1. P is arbitrarily
near to a point of Q) at whichAv^0.
2. The isothermthroughP containsa point of Q) at whichAv 4=0.
The normalset need be neitherconnectednor open; it may be empty.It is the
largestsubset of 2 for which Carnot's ideas seem to be fruitful. The set of
temperaturesthat correspondto points of Q)n is open, but it need not be
connected.It is the union of an enumerablecollectionof disjointopen intervals.
In what followsnow we shall presumethat6 lies in one of theseintervals.
Theorem 7extin Chapter 9 and Theorems 9bisand 10bisin Chapter 10 of
Conceptsand Logic tell us thaton thatintervalthereare continuouslydifferen-
tiate functionsg and h,the formerbeingan increasingfunctionand the lattera
positivefunction, such thatin a simpleCarnot cycle# in @n

C-(*)=J|^yC+(n (10)

g(0+)-g(0-)
l(v= ¿{e^ }c+(n (ii)
and thatat each pointof Q)n

'A*Jw (12)
a /Ay'_d /Ky'
ôe'hi~õv'h~)- (13)

Moreover,h is unique to withina multiplicativefactor;whenh is assigned,g is


unique to withinan additiveconstant;and

g'>0 excepton a set withemptyinterior. (14)


In particular,Carnoťs function'i in (7) has the form

"4 (i5)
it is unique and continuous;and

/¿>0 excepton a set withemptyinterior. (16)

4. Hotnessand EmpricalTemperature,I. Continuity

I do not regard the contentsof this section and the succeeding one as
anythingbut an expositionof ideas so simple that theymusthave been sensed
informallyby thepioneersof thermodynamics. Therefore, althoughI do not find
themexplicitlyin any early source,I do not specificallyattributethemto any
laterauthor.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
362 C. Truesdell

The concept of hotnessis natural and primitive,like presence,weight,and


place. We perceiveheat and cold as we do lightand dark,wet and dry,farand
near,throughour senses.As Mach wrote5
Amongthe sensationsby which,throughthe conditionsthatexcitethem,we
perceivethe bodies around us, the sensationsof heat forma special sequence
(cold, cool, tepid,warm,hot) or a special class of mutuallyrelatedelements.
. . . The essence of thisphysicalbehaviorconnectedwiththe characteristicof
sensations of heat (the totality of these reactions) we call its hotness
["Wärmezustand"]6.
Also, wroteMach 7,
The sensationsof heat, like thermoscopicvolumes,forma simple series,a
simplecontinuousmanifold.It does not follow therefrom that the hotnesses
also form such a manifold.The propertiesof a systemof signsdo not
determinethe propertiesof thatwhichtheydesignate.. . . The assumptionof
a simplecontinuousmanifoldof hotnessesis sufficient.
thislastassertion
Mach justifies onlybystating indicatesthe
thatno experience
contrary.
The conceptof "simplecontinuousmanifold"or "continuousmanifoldof
one dimension"derivesfromRiemann8 and has been made precisein more
recentworkson pure mathematics. Mach assumesthat the set Jf of all
5 Mach, derEntwicklung derThermometrie".
op.cit.,§ 1 of"Historische Übersicht
6 Mr.Serrín "hotnesslevel". Earliermodern I
authors, beingone,haveused
prefers
"temperature" loosely to mean sometimes hotness,sometimes a number representing it.For
exampleA. H. Wilson,Thermodynamics andStatistical
Mechanics, Cambridge, Cambridge
University Press,1957,writes in§ 1.21"Ifweassumethattemperature is a primary idea,the
methods ofmeasuring it mustbe to a certainextentempirical."
7 Mach, Temperaturbegriff § 5.
8 14 ofthe article F. Enriques, " derGeometrie", der
Cf.§ by Prinzipien Enzyklopädie
Mathematischen Wissenschaften, Volume 31?Heft1, 1907:
The fundamental conceptin thetheoryof thecontinuum is thatof thecontinuous
manifold ofonedimension. Regardedabstractly, thisconceptmaybe identified withthe
concept oftheline,ifthepointofthelineistakenas theelement ofthemanifold andifno
regardishadfortherelation ofthelinetotherestofspaceorforany(metric) conceptof
thelengthofsegments (or arcs)oftheline.Thusonlythoseproperties ofthelineare
considered thatareconnected withgeometrical determination ofthelinethrough the
motionofa point:thenatural orderings ofthepointsofthelineandtheircontinuity, the
segments, etc.
Thoughthe"continuous manifold ofonedimension" isthusdefined partly interms ofwhat
it is not,theexplanation is pretty clear.Enriques attributes theconceptto Riemann's
celebrated Habilitationsvortrag of 1854,firstpublishedin 1868:"Über die Hypothesen,
welcheder Geometriezu Grundeliegen",republished in all editionsof Riemann's
Gesammelte Mathematische Werke. Riemann'spresentation, conditioned by thecircum-
stances,isnotexplicit, butitsufficed togivemathematicians someessential ideaswhichthey
couldrender, in time,precise.
Fora history oftheconcept ofmanifold seethearticlebyD. M. Johnson,"Theproblem
oftheinvariance ofdimension in thegrowth ofmoderntopology", Archive forHistory of
ExactSciences, Volume20 (1979),pp.97-189.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 363

hotnessesis a continuous manifold.Jť maywell be thefirstmanifoldto have been


suggestedby physicalexperiencenot expressedin termsof mass, position,and
time. This circumstancelets me think it worthwhileto present the simple
physical ideas that lead us to impute specializing propertiesto the hotness
manifold.
First of all thereis our sensationthat at a certaintime one body is hotter
than another.Sensations,indeed,are not always consistent.This factdoes not
make us reject all the evidence theybringus. Rather,if theysometimesbear
conflicting witness,we regardthemas being partlyin errorthen,and we do not
on that account deny the existenceof what they usually suggest.We let the
exceptionprovetherule.Guided by sensoryexperience,we conceiveJť as being
ordered.Using > to denote "hotterthan" and ^r to denote "not cooler than",
we assume that ^ is a simpleorderingupon Jť : If híeJť and h2eJť theneither
h^h2 or h2^h1, and those two relationsboth hold ifand only ifhl=h2.
If h2>hl9 theset of all h such that/z2>/î>/ï1we may call a boundedsegment
]/i1,/i2[of Jť. The segmentsof Jť consistin the sets of this kind and the sets
definedby relationsof the formsh^hx and h^h. The segmentsprovidea basis
fora topologyon Jť, in termsof which we may definelimitsand continuity.
When we restrictattentionto a bounded segment,only its subsegments,also
necessarilybounded,are needed forthispurpose.
Temperature is a numericalindicatorof hotness,just as milestonesalong a
road indicatenearnessand farnessby numbers.As Mach wrote9,
The temperature the markof the
is ... nothingelse thenthe characterisation,
hotnessby a number.This temperature numberhas simplythe propertyof an
inventoryentry,throughwhich this same hotness can be recognizedagain
and ifnecessarysoughtout and reproduced.
An empirical-temperature functionor empirical-temperature scale is an assign-
mentof co-ordinatesto a segmentJť0of Jť or to Jf itself.The value 0O of a
temperaturefunction9 at a particularhotnessh is the empiricaltemperature of
h0 on the scale 0: If h0e3ť0,then =
0o 0(ho). That is not all. As MACH10 wrote,
This [temperature]numbermakes it possible to recognizeat the same time
the order in which the indicated hotnesses follow one another and [to
recognize]betweenwhichotherhotnessesa givenhotnesslies.
That is, the mapping9 preservestheorderof hotnesses
:

h^h2 o 9(hí)^9(h2). (17)

It is to thisassumption,I believe,that Mach referredwhen he wrotethat"the


sensationsof heat ... forma simple series". He merelyput in words the silent
prejudice all the pioneers of thermometry, calorimetry,and thermodynamics
acceptedas a matterof course: Hotnessesmaybe represented by points
faithfully

9 Mach,
§ 22.
Temperaturbegriff
1u
§ 22.
Temperaturbegriff

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
364 C. Truesdell

on the real line. Were this not so, the innumerablediagramsin which a 0-axis
appears would misrepresent.
Temperatureshad been introducedand were grownfamiliarlong beforethe
abstractconcept of hotness,and it was studyof our experiencewith tempera-
turesthattaughtus, finally,to isolate and abstracthotness.Indeed,temperature
and hotnesswere confusedfora long time,and even in Maxwell's magisterial
1*
Theoryof Heat the two are not distinguishedexceptby implication :
The temperatureof a body ... is a quantitywhichindicateshow hot or how
cold the body is.
When we say that the temperatureof one body is higheror lower than
that of another,we mean that the firstbody is hotteror colder than the
second,but we also implythatwe referthe state of both bodies to a certain
scale of temperatures.
By the use, therefore,
of the word temperature, we fix
in our minds the conviction that it is possible, not only to feel,but to
measure,how hot a body is.

When we come to considerhow to measure"how hot a body is", it suffices


forpracticalpurposesto limitthe rangeof an empiricalscale to some bounded
interval]0l902[. For our present,local considerations,that interval is best
regardedas containedin one of the intervalsmentionedin §3 in connectionwith
some one body. It is naturalto suppose thatthis real intervalis the image of a
bounded segment~'h^h2'_ of Jť. Accordinglywe begin by consideringonly those
empirical-temperature scales that are one-to-one,order-preserving
mappingsof
boundedsegments~]hl,h2[ °f ^ onto boundedintervals]0l502[ of &. Thus the
function0 in (17) is a "homeomorphism":a continuoustransformation havinga
continuousinverse.It is plausible and can be proved easily that the segment
]/zl5/z2[has the "Hausdorffproperty":Any two distincthotnessesin it lie in
subsegmentswhose intersectionis empty.That factmakes ']hi,h2[_a continuous
manifoldin the modernsense of the term;an empirical-temperature functionis a
charton Jť whichpreservestheintrinsic orderof hotnesses.
Suppose that on some one bounded segmentJf0thereare two empirical-
temperature scales, 0 and 0*. Then ifhoeJťo,

0o= 0(ho), 0* = 0*(/Io)= 0*o0-1(0o). (18)

The compositemapping0*o0- 1 maps the rangeof the scale 0 onto the rangeof
the scale 0*: If/=0*o0-1, then

0S=/(0oX 0oeRange0. (19)

Because of (17) we see that/ is a one-to-one, increasingmappingof


bicontinuous,
Range 0 onto Range 0*. It was these mappings of real intervalsonto real
intervalsthat the pioneersof thermometry had to use in comparingthe results
of usingdifferentempirical-temperature scales.

11 J.C. Maxwell, ofHeat,10th


(last)edition,1891,p.2.
Theory

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
AbsoluteTemperatures
fromCarnoťsAxiom 365

If 9 and 0* are definedon bounded segmentsJ^ and Jť2whose intersection


is the segmentJť0, the foregoingargumentshows that 0 and 0* provide
compatiblechartswhichtogethercover ^uJ^. Thus J^uJ^ is a continuous
manifold.We can go on in thisway to any collectionof overlappingsegmentsof
Jť' For thetimebeing,however,we shall revertto thestudyof a singlebounded
segmentof Jť.
The homeomorphismsof a bounded segmentof Jť with a bounded real
intervalfallinto two classes: thosethatpreservethe orderon Jť7,and those that
reverseit. How do we decide whichof the two classes of homeomorphisms does
consistof empirical-temperature scales and whichdoes not?
It is customaryhereto appeal again to the directevidenceof our senses: We
feel that boiling water is much hotterthan freezingwater,etc. The mappings
that assign to boiling water a highertemperaturethan to freezingwater are
thereforeregardedas faithfulto the intrinsicorderingof Jť, and mappingsof
the other class are rejected.While of course any comparison of theorywith
experiencerests ultimatelyupon human sensations,we preferto eliminateas
many appeals to them as we can. To discern which of the two classes of
homeomorphismsrespects (17) and hence consists of empirical-temperature
scales,we may use heat enginesas theyare idealized by Carnot cycles.A Carnot
cycle absorbs heat at the higherof its two operatingtemperatures:0+ >0~. A
particularcyclethatis a Carnot cycleaccordingto thehomeomorphisms of one
class is the reverseof a Carnot cycle accordingto those of the otherclass. The
workdone by a cycleis a purelymechanicalpropertyof thatcycle,independent
of the methodsof measuringhotness.That work is positive,negative,or null. If
a cycle does positivework,its reversedoes negativework.Carnot's General
Axiom (1) asserts that every Carnot cycle does positivework.Moreover, this
statementis invariantunder the transformations (19) defined by increasing
functions/ Thus Carnot's General Axiom serves to distinguishthe class of
empirical-temperature scales fromthe class of homeomorphismsthatreversethe
intrinsicorderof hotnesses.Suppose thata body undergoa cyclesuch that

1. it absorbs and emitsheat only on two distinctisotherms,withwhichwe


have somehowassociated numbers9+ and 0~, respectively;
2. the amountof heat it absorbs is positive;
3. it does positivework.

Then the numbers 9+ and 9~ cannot result froman empirical-temperature


functionunless 0+>0~. To give a practical example of this distinction,we
proceed now to the studyof thermometers.
A thermometer is a body used to determineempiricaltemperatures. Typically
it is a body whichexpands appreciablywhen it is heated,and its volumes are
regardedas the values of an empirical-temperature function.In Maxwell's
words,they make it "possible not to
only feel, but to measure,how hot a body
is." More than that,just because volumes are measurable they are taken as
replacingthe sometimesinconsistentevidenceof our sensations.
To this end we cannot use an arbitrarybody. Let us suppose, forexample,
thatour firstthermometric body consistsof air,and let us measurevolumesof a

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
366 C. Truesdell

body of waterat atmosphericpressurewhenthe air temperatureis near to 4° C.


Figure1 shows what we shall find:

Id p=1atm
|o /
o "> /
a; ot / |
£-g¿8- [I
5e -
^
'
I
I
V
Fig.1.6 (airthermometer) C
indegrees

As has beenknownsincethedaysoftheAccademiadel Cimento,thebodyof


waterat thepressure1atm.occupiesthesamevolumeat twodifferent tempera-
turesaccordingto theair thermometer, one above4° and one belowit.Thusa
waterthermometer operatingin thisrangeat constantpressurewouldindicate
thesametemperature at hotnesseswhichaccordingto theair thermometer are
The mappingfromthenumbers
different. determined by volumes ofwater onto
the numbersdetermined by volumesof air is not one-to-onein any open
thatcontains4 °C.
interval
Thatis notall. Supposewe tryto use as a thermometer a bodyofwaterat
atmospheric pressurein therange from 0° to 4° C. The temperatureaccordingto
thepresumed waterthermometer is thenan invertible ofthetempera-
function
tureaccordingto the air thermometer, but the temperature that is higher
accordingto thewaterthermometer willbe loweraccordingto theair thermo-
meter.Thusit is impossible thatbothoftheseallegedscalesoftemperature can
conform with(17).One mustbe rejected.
In thisregardMach12 wrote
Thosehotnesseswillbe called higherin whichbodiesgiveriseto greatervolume
indicationson the thermoscope.... Thereforewe avoid use of water as
thermoscopicsubstance. . .
Mach went too far here. There is no inherentreason for regardinggreater
volume of air to be more faithfulthan greatervolume of water. True, most
othersubstancesagree with air and disagree with water in theirbehavior,but
resortto majorityrule would seem dubious ifnot dangerouspracticein physics.
12 Mach, § 4. Modernauthorsuse theterm"anomalous"in this
Temperaturbegriff
regard.E.g.Wilson,§ 1.21ofthebookquotedin Footnote6:
"By comparing theresultsobtainedbyusingdifferentthermometers we can rejectas
unsuitablethosesubstances, suchas water,whosebehavioris anomalous.... All thatwe
requireofa substance a thermometer
to be usedin constructing is thattheproperty to be
measured ... shouldbe a strictly
increasingfunctionofthetemperature [i.e. hotness]."
The "anomalous"substancesare fewindeedin the rangeof hotnessespresently
availableto experiment, but we are leftwondering how to tell whichsubstancesare
"anomalous"and whicharenotwhennewrangesofhotnessshallbecomeaccessible.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 367

Under our interpretationof Carnot's General Axiom, if any cycle that


accordingto the air thermometer is a Carnot cyclewithoperatingtemperatures
in the intervalbetween0°C and 4°C does positivework,then the volumesof a
fixed mass of waterat atmospheric pressureare not the values of any empirical-
temperature scale for the correspondingrange of hotnesses.Experienceteaches
us that the enginewhich absorbs heat at the highertemperatureaccordingto
theair thermometer does positivework.Thus it is thesupposed waterthermom-
eterthat we rejectin the range including4°C and air temperatureslower than
that.The foregoing, I believe,is a precisereplacementforMach's vague claim.
It also justifiesuse of the term"anomalous behavior" here. Of course,in other
circumstancesthere is no reason at all to reject water as a thermoscopic
substance.

5. Hotnessand EmpiricalTemperature,II. Differentiability

So far,we have appealed onlyto the homeomorphismand intrinsicordering


requiredof empirical-temperature scales. If we are to use themin calorimetry,
we demand somethingmore. The verystatement(3) of the Doctrine of Latent
and SpecificHeats refersto the derivativeof a functionof timewhose value is
thetemperature, and the uses to whichthe constitutive functionsxu,Av, and Kv
are put in the developmentof thermodynamics also appeal to differentiation.
itselfrequiresassumptionsregardingJť. If we approach hot-
Differentiability
ness throughuse of empiricaltemperature-scales, we may referto those assump-
tions indirectlyby requiringthat the transformation of one scale to anotherbe
entiable.
differ Then in
/ (19) is a differentiablefunction,and we shall write

d0*=ff(0)d0. (20)

/'(#)^0. Because the inversetransformation


Because /is an increasingfunction,
/"Ms also the
differentiable, possibilitythatf'(9) = 0 forsome 6 is excluded.
Thus
/'>0. (21)

The requirementthat 6 exist almost always is independentof the choice of


scale.
Still restrictingattentionto a single bounded segment Jť0, let us now
investigatethe rules of transformation forthe quantitiesthat enter(2) and (3).
First, the chain rule and (20) show that

Here, as usual in workson physics,

^ = J_m(j//-i(ö*)).
J v ; (23)
de* de* v

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
368 C. Truesdell

From (22) and similar formulaefor the derivativesof the other constitutive
functionswe see thatto renderinvariantunderchange of empirical-temperature
scale the assumptionthat m be continuouslydifferentiable, it is necessaryto
assume that/' is continuous.We may sum the effectof adjoining this assump-
tionto theprecedingones by statingthatJ^ is a diffeomorph of an open interval.
Likewise,if a segmentis the union [jjťa of a finitenumber of bounded
segmentsJ^, on each of whichan empirical-temperature scale is defined,and if
each intersectionof thosesegmentssatisfiestheconditionsjust demandedof ^f0,
then (Jj^ is a diffeomorph of an open intervaland has an intrinsicordering.
More generalconclusionscan be obtainedfromthe moderntheoryof manifolds
but are not needed ifour objectiveis onlyto clear and specifythepioneers'ideas
about scales of empiricaltemperature.

In histreatment
of 1975Mr.Serrín assumedJfto be a diffeomorph ofthereal
line,equippedwithan intrinsic
ordering.Suchan assumption
seemsto be implicitin
someearlierstudiesbyothers.In hislaterworkSerrín requires
ofJf merely thatit
be a continuousmanifold.For detailsthereadershouldconsulthislectureof 1977,
citedabove in Footnote2. Mr. C.-S. Man in a master'sthesisacceptedby the
University ofHong Kongin 1975provideda farlessgeneralbutconstructiverather
thanpostulational
introductionofthehotnessmanifold.

We revertto the considerationof a singlebounded segmentof Jf, on which


an empirical-temperaturescale is defined.The quantitiesAv and Kv in (3) will
generallydepend upon the choice of empirical-temperaturescale, but Q will not.
Therefore,in an evidentnotation,

Av^(K6*)V+Kv^(K6*)Ô* = AvJV,e)V+KvJV,6)Ô. (24)


Hence A _A
/iv,e*~/iv,oi
(25)
Kv d*d6*=KVdd6.
It followsfrom(7) that
= ined6.
Lid*d9* (26)
In theseformulaethe subscript6 may be read "when the scale 6 is used", and
likewisefor0*, and the arguments0* on the left-handsides are relatedto the
arguments6 on the right-handsides through(19). In passages where only a
singleempirical-temperature scale is considered,we drop the subscript.
The transformation rules(22) and (25) show thatthesmoothness assumedfor
m, Av, and Kv and the signs of dw/dV,dw/dO,Av, and Kv are invariantunder
changeof empirical-temperature scale. In particular,the constitutiveinequalities
(2)2 and (3)2 hold forall scales ifand only iftheyhold forone scale.
In summary,the basic axioms (2) and (3) hold in termsof one empirical-
temperature scale ifand onlyiftheyhold in termsofall scales.
Since a differential
equation satisfiedby theadiabais is d6/dV= -Av/Kv, the
qualitativebehaviorof the adiabais near a point is the same forall scales. That
means that Carnot cycles definedin termsof one empirical-temperature scale

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 369

exist locally if and only if they exist locally according to all such scales,
temperatures correspondingto the same hotnessbeing understoodin the term
= O ifand onlyif/ie(0)= O,so the set of temperatures
"locally". Moreover,/iö*(0*)
at which [i vanishes on any one scale correspondto hotnessesthat make ¡x
vanish on all scales. We recall that the set of such temperatureshas empty
interior.

6. Kelvin'sAbsoluteTemperatures

Kelvin used "absolute temperature"in threesenses:


1. A scale of empiricaltemperatureindependent of the choice of thermome-
ter.
2. A scale of empiricaltemperaturesuch that the work done per unit heat
absorbed in a Carnot cycle depends upon the difference of operatingtempera-
turesonly.
3. A scale of empiricaltemperaturesuch that the ratio of heat emittedto
heat absorbed in a Carnot cycle shall equal the ratio of the lower operating
temperature to the upper.
Kelvin's absolute temperatureof 1848 satisfiesthe firstrequirement;so
does any differentiable functionofthattemperature havinga positivederivative;
one such functiondefinesKelvin's absolute temperatureof 1854. Kelvin's
second requirementis satisfiedby his firstabsolute temperatureboth in the
Caloric Theoryand in Clausius' thermodynamics. Kelvin's thirdrequirement
is satisfiedin Clausius' thermodynamics by his absolute temperatureof 1854.
When rendered concrete by referenceto the classical axioms (2) and (3),
Clausius' absolute temperatureof 1854 reducesto Kelvin's of the same year.
The historicaldetails will be found in my forthcomingTragicomicalHistory.
Here by use of theapparatusjust now assembledI will proveall of theforegoing
statementsthat referto Kelvin's work and compare the resultsappropriateto
different possibilities.
Kelvin's firstscale of "absolute temperature"is definedas follows:

T= J/id0. (27)

Since dx/dd= n, the requirement


(21) is satisfiedifand only if

fi>0 forall 6 in its domain. (28)


Henceforthwe shall assume provisionally untilthe next
that(28) holds,deferring
sectionthe analysisofjust what that means and whyassumptionssufficient for
it should be laid down. Since ¡jlis continuous,our provisionalassumption(28)
makes x an empirical-temperature function.
Moreover,from(26) we see thatwe can adjust theconstantsof integrationin
the definitionsof t0*and t0 so as to insurethat

V(0*) = t,(0). (29)

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
370 C. Truesdell

Thus T is an absolute temperature in the sense thatits value is independent


of the
empirical-temperature scale used to obtain it. The physicaldimensionsof x are
work -7-heat. The zero of x is arbitrary.
Any continuouslydifferentiable functionof x withpositivederivativesatisfies
all thesame requirements. Thus thereare infinitelymanyabsolutetemperatures in
Kelvin's firstsense, x is one of them. Kelvin's second scale of "absolute
temperature"T is definedas follows:
T=ezl' (30)
J being a constanthaving the dimensionsof work -r-heat. Values of T are
dimensionless and positive;T is absolute in Kelvin's firstsense.The value T= 1
corresponds the arbitrary0 of t.
to
All our considerationsso farare restrictedto one body and to an intervalof
empiricaltemperaturesin !3n forthat body on some one empirical-temperature
scale. The intervalmay be small, but it need not be. Withoutknowingsome-
thingabout ¡x,we cannot determinethe range of x. If ¡x is of the rightkind,the
rangeof x is the whole real line. Then T=0 correspondsto x= - oo, and T= oo
correspondsto x= oo. Thus the zero of T has a special status.Whetheror not x
as definedby (27) can have the value - oo, depends upon the natureof ¡x.It is
easy to give examples of functions'x that make x have a finitelower bound.
Then T neverattainsthe value 0. The pioneersregardedjx as a functionto be
determinedby experiment.Kelvin seems to have taken for grantedthat the
rangeof T would come out to be ]0, oo[.
Sincex and Tare empirical-temperature scales,we mayexpressthefunction 'xin
termsof them.Calling the results'xxand ¡xT,we see from(26) that
de

dx J <31>

Thus the General Carnot-ClapeyronTheorem (7) assumes the followingforms


whent and T are used:
A -^ JA
/vv,x~ » J/LVT-i-T^ . Í32Í
yóÁ)
^
Of course
/>t(Kt)= w(V96(t))9 pT(V,T)= m(V,0(T)).
Use ofan absolutetemperature Theorem
enablesus to state the Carnot-Clapeyron
withoutreferenceto Carnoťs function'x.
All theforegoingresultsfollowfromCarnot's workalone. Use oftheresultsin
Concepts and Logic enables us to go further.Firstwe note that

go*= ge* V =V (33)


function.In generalh is not.
If g' nevervanishes,g is an empirical-temperature
Because of (15) and (31),
K=C JhT= Tg'T. (34)

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
AbsoluteTemperatures
fromCarnoťsAxiom 371

Hence (10) becomes


C-(<g)j¿x-) JT(T~)_ ( '
C+(<ë)~ht(?+) hT(T+ý
(11) becomes t+ r+
ihz(x)dx J Í lhT(x)/x-]dx
L{^> _t- =_r: . Hg)
c+(tf) /it(T+) M^+)
and (13) becomes

ÔV ~StI ÔV ~Ht' ÕT [ hT I' (il)


K V
enables us to dispensealtogetherwiththe
Thus use of an absolute temperature
functiong.
In theCaloric Theoryofheatwe mayalwaystakeh as 1. Then(34)t showsthat
=
g i + const.,while(36) reducesto

The formeroftheseexpressionsshowsthatbyintroducing x Kelvin obtained,for


theCaloric Theory,a temperature"absolute" in hissecondsenseas wellas hisfirst.
Also (37) becomes , ~ ~A
õKViZ=dAy9X dKVtT=dAVtT .^
' '
ÕV or ÕV ÔT
thata heat functionshall existat least locally.
conditionsnecessaryand sufficient
Clausius laid down as his fundamental axiom the requirementthat in all
CycleS<^
H<g)= JO<g), (40)
in whichJ is a universalpositiveconstanthavingthedimensionsofheat -4-work,
whileC(#) is thenetgain ofheatin (€.Corollary10.3in§ 10 of ConceptsandLogic,
restatedaccordingto thedirectionson p. 98, makes Clausius' axiom equivalent,
fortemperatures thatcorrespondto pointsof ^n, to

g = j/i + const. (41)


Hence ifg is an empirical-temperature function,so is h.
We recallthatifg' does notvanish,g is an empirical-temperaturefunctionin all
theoriesconsistentwithCarnot's General Axiom,but h generallyis not. On the
otherhand, h is always a positivefunction.By makingh become an empirical-
temperature function,Clausius' thermodynamics providesan empiricaltempera-
turewhichis alwayspositive.
In thedefinition Ifwe choose it as
(30) of T theconstantcalled J was arbitrary.
beingtheJ of CLAUSIUS'axiom (40), from(34) we see that
h= KT, (42)
K beingan arbitrarypositiveconstant.Then (35) becomes

Π= r(t+-r-)/J=Ili (43)

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
372 C. Truesdell

Kelvin regardedthesecondofthesestatements as expressinga desirableproperty


of absolutetemperature;forClausius' theoryTis "absolute" in Kelvin's third
sense of the term,and forthatreason Kelvin adopted it in 1854. In Clausius'
theory(36) reducesto
im . --(t+-t-)
- , T- (44)
jcwr1-^ j =1"t^-
The formerevaluation of efficiency is due to Kelvin. It shows that even in
Clausius' theoryt is a temperature "absolute" in Kelvin's secondsenseas wellas
his first.The latterevaluationin (44) is due to Rankine and Clausius, moreor
less.Details maybe foundinmyTragicomical History.Finally,inCLAUSIUS'theory
(13) becomes
dKy^J_Ay^_Ay^' dKv^T d IAVtT'
dV dx J dV dT' T V
In (32)2 and (45)2 the readerwill recognizethe basic constitutiverestrictions
ofclassicalthermodynamics. We have obtainedthemhereonlyforthefunctions ro,
Av, Kv, and ¡i belongingto some one body,and onlyforan intervalof empirical
temperature whichmaybe verysmall.Thus,so far,therangeofTmay be small,and
we mightobtaindifferent scales T fordifferent
bodies.We proceednow to remove
the limitationand the variety.At the same time we returnto the generalityof
Carnot's theory.

7. Consequencesof Part II of Carnot'sGeneral Axiom

Carnot assertedalso thatthemotivepowerofa Carnotcyclewas independent


of the bodythatunderwent it. This is PartII of his General Axiom. AxiomIV of
Concepts and Logic expressesit in one way. For our purposeshere a somewhat
weakeraxiom will do: Carnoťs functionfiis a universal functionin the sense that
whateverbe theconstitutive functionsw and Av used to determineit through(7),
the same function/ãresults.Hence the same absolute temperaturex is obtained
through(27). The value of Kelvin's absolute temperature x is independent not
only of the empirical-temperature scale used but also of the body employedto
determine ¡i. In otherwords,x maps hotnessesonto numberswithno reference
to any particularbody.
Our considerationsare stilllocal. Thereare manywaysto extendthemto the
wholehotnessmanifoldJf or as largea partas mayseem desirable.I willchoose
one.
In particular,let 3fťxand ¿tf2be bounded segmentsof Jf witha non-empty
intersection Let empirical-temperature
J»f12. functions91 and 92 be definedon Jťx
and Jť2,respectively. Then on Jf12 we mayinterconvert 61and 62bya rulelike(19),
restrictedby (21). Let bodies Bx and B2 have normal sets whose sets of empirical
temperaturesaccording 0l to and 02, respectively,include the ranges of those
functions.The correspondingfunctions¡i are fil and fi2,say. On Jť12we may
expressfi2as a functionof0x. Since ¡i fora givenscale is a universalfunction, ¡i2 so
expressedmustbe thesamefunction as ¡ii. Bychoiceoftheconstantofintegration in

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
AbsoluteTemperatures
fromCarnoťsAxiom 373

(27), we geton J"f12thesameabsolute-temperature functionx forbothbodies. Now


consideringonlyB2, stillon J«f12,we mayuse 62 insteadof 9l and because t is an
absolute-temperature scale again calculate t through(27). In termsof 62 the
definitionof t still makes sense on all the image of Jť2.We have proved the
followinglemma.On boundedsegments Jftand Jť2havinga non-emptyintersection
function6Xand 92; let therangesof6l and 62 be
let therebe empirical-temperature
intervals includedinthetemperature setsofthenormalsetsofbodiesBx andB2 . Then
theabsolutetemperature-function t, uniqueto withinan arbitrary existson
constant,

The lemma shows thatifa segmentof Jť is the union of a finitenumberof


bounded segmentson each of which¡i existsand is positiveforsome body, the
absolute-temperaturefunctiont exists on the whole segmentand providesa co-
ordinatesystemuponit.
The readerfamiliarwiththetheoryofdifferentiable manifoldswillsee thatitis
unnecessary to attention
restrict to a segment which is theunionofa finitenumber
of boundedsegmentsofJť.I have done so because I desireto presentan argument
usingonlyelementary ideas; I have no wishto makeKelvin's conceptualproblem
seem trivialby invokinga moderngeneraltheorem.Moreover,because experi-
mentsby theirnature can determineonly finitelymany numbers,thereis no
physicalgain in weakeningthe assumption,and ifthe partof Jf accessibleto us
throughuse of empirical-temperature scales could not be covered by a finite
numberof them,no experimentcould detectthatfact.PerhapsJť itselfcould be
coveredsimilarly,but no experimentcould establishthat.

8. Thermometric
Axiom

It remainsto considerthecondition(28),whichall theforegoing constructions


have presumedto hold. In orderto establish(28), we need a further axiom. Mr.
Serrín, workingupon a different
framework of ideas, has emphasized the
importance of such an axiom,whichhe calls thermometric.
Here I adopt a special
case of his axiom of 1975,phraseda littledifferently:
Thermometric Axiom.Jť containsa segmentJř{hwhichis theunionofafinitě
numberof boundedsegments, on each of whichan empirical-temperaturescale is
If60 is an empiricaltemperature
defined. ofa hotnessinJ^h,itliesinthetemperature
domainQ) of some body such thatat one pointon the
intervalof theconstitutive
isotherm =
0 #o
KP*KV. (46)
Let us call P thepointwhichtheThermometric
Axiom posits.Then because of(9)
we know that
Kp>Kv at P. (47)
By (4), then,

at P. (48)
Av^>0

Hence Av^0. Hence P belongs to the normal set of the body that the

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
374 C. Truesdell

Thermometric Axiomposits.Hence (7) holdsat P. Because /iis a universalfunction,


itis now determinedonce and forall at 0O;because of(48),¿i(0o)>O. But according
to theThermometric Axiom,foreach h in Jťth thereis a o0 ofthiskind.By appeal to
(29) we establishthe following
Theorem,x as definedby(27) is an empirical-temperature scale notonlylocallybut
also on all ofjťth;for each h thevalue t(ö(/i))does notdependupontheempirical-
temperature scale 0 used to calculate it.
Thus i is an absolute-temperature scale on all of'Jťth.
I havechosen(46)as thecritical condition on whichtobasetheThermometric
Axiombecausethe inequality(47) reflects commonexperimental knowledge.
Accordingto tablesof physicalquantities, nearlyall real materialssatisfy(47)
almostalways.Therearesomeexceptions; indeed, wecanseedirectly from (4)3that
Kp= Kv ifandonlyifAvdw/d6 = 0. Somefewmaterials seemto obeyveryclosely
an equationofstateoftheformp = xn(p'and forthemKp= Kv always.A few
others,likewater, exhibitcurveson whichõm/d6 = 0; Figure1 showsonepointon
sucha curve.Because/x > 0 invirtueoftheThermometric Axiom,weconcludethat
Av = 0 ifand if
only dm/õ6 0, = so the curves just mentioned could be defined
alternativelyas curves on which Av = 0. Counterexamples show thatwithout some
assumption such as the Thermometric Axiom it would be for
possible 'i to vanish
forcertainparticular temperatures. If/x(0o) = O,(7) showsthatôw/õ6= 0forall V
andforall bodieson theisotherm 9 = 6O.Such a statement wouldabundantly
contradict experience. This factprovides furthersupport theThermometric
for
Axiom13.
Intentionally I haveleftJťth unspecified. Mathematical authorsusuallywishto
havea singlescale oftemperature on all ofJť. An obviousmodification ofthe
thermometric axiomcan adjusttheforegoing theorem to thatrequirement. Then
13 The function of thethermometric axiomis to makesurethat/u(60) existsand
l¿(60)> 0 forall60inrange6.Assumptions sufficient
tothisendhavebeenmade,expressly or
tacitly,sincethebeginnings ofthermodynamics.
Carnot restricted hisattentionto idealgases,forwhichofcoursedw/d6 = p/6>0.He
statedexpressly thatAv>0. Sincehe also derivedthecorresponding specialcase ofthe
GeneralCarnot-Clapeyron Theorem (7),itwasunnecessary forhimto statethat/i>0. His
successors extendedthescope of thetheory, but as theyagreedthatfiwas a universal
function, thesameforall bodies,andas theyusedfreely theconceptofidealgases,theyhad
no needto determine thesignofjuafresh, forCarnot hadalreadydoneso. Theytookfor
granted that¿¿>0, as is shownbytheirdividing byitwhenever convenient.
Thefirst timethatthisargument failedtoremain validwasinKelvin'ssecondtheory of
absolutetemperature, 1854.He couldno longerhaverecourse to evaluating
fibyuseofan
idealgas,and hisanalysisofthebehaviorofwaterin 1853had shownhimthatforsome
substances in someconditions Av<0. As he concludedthatfi= J/T,T beinghis second
absolutetemperature, hecertainly concluded alsothat/¿>0,buthisworkintheseyearswas
sketchy. Wecannotbesurejustwhatheassumedorhowcarefully hethought outthedetails.
He wasthefirst persontopublish (4)3infullgenerality,andcertainly henoranyother
neither
earlystudentwouldhave hesitatedto assumethatKP>KV if dm/d0=1=0, forthatwas
precisely whatall availableexperimental dataindicated.Itis forthisveryreasonthatI have
chosentostatea Thermometric Axiominterms ofKpandKv. Ofcoursean axiommustgo
beyondexperimental data, mustextrapolate fromthe knowninto the unknown, for
otherwise it wouldbe uselessin prediction.
LaterauthorshaveabandonedCarnot's approach.Not havingtheGeneralCarnot-
Clapeyron Theorem(7) on whichto base theirconstructions, theyhavehad to use more

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
AbsoluteTemperatures
fromCarnoťsAxiom 375

each hotnessliesinthedomainofsomeempirical-temperature function.WhileI see


no objectionto such an assumption,it seems both daringand unnecessary.We
mightbe contentto thinkof J^h as the set of all hotnessesso faraccessible to
experiment.That set may well grow as timegoes on.

9. The PhysicalDimensionof AbsoluteTemperature

A value of Kelvin's secondabsolute-temperature scale is a dimensionless


number. We are accustomed to temperatures thatbeara dimension called"the
dimension oftemperature". Lookingback at an empirical-temperature scale 0,
we mayifwe likethinkofit as bearingthephysicaldimension whichwe may
call "thedimension ofthescale 0". Physicaldimensions suchas thoseofmass
and lengthand timeare assignedonce and forall, and onlythemass,length,
and time-intervalto whichthevalue 1 is assignedmaybe chosenat will.All
scalesofmassareproportional. Suchis nottrueofempirical-temperature scales.
We mayindeeddividethemintoequivalenceclasses,all members ofanyone of
whichareproportional. Thena physicalunitappertains to a givenclassbuthas
no meaningwithrespectto any otherclass. Thus thereare infinitely many
distinctphysicaldimensions ofempirical temperature.
Kelvin's absolutetemperatures t and T are conceivedin sucha wayas to
correct thisunsatisfactorystateof Theyannulthedifferences
affairs. betweenthe
physical dimensions ofall possibleempirical scales.The values of anyone ofthe
scalesT are dimensionless numbers whichdo notdependuponthechoiceofthe
empirical-temperature scale0 usedto obtain¡i and thento calculatet and T. If
we can proveall scalesT thatcorrespond to a givenJ to be proportional, we
can thinkof thesepure numbersas coordinateson one and the same 1-
dimensional vectorspace,thespaceofabsolute-temperature vectors. It is justthe
samethingto saythatabsolutetemperature has itsownphysicaldimension, the
dimension of absolutetemperature. Only the choice of unitof absolute tempera-
turedistinguishesone absolutescalefromanother.
elaborate reasoning.Theyalsohaverecourse toa "FirstLaw",whichI wishexpressly
freely
to avoid.
Theearliest explicitthermometric axiomI havefoundis Postulate V (thermometers)in
the paper of J.B. Boyling "An axiomaticapproachto classicalthermodynamics",
Proceedings oftheRoyalSociety (London)A 329(1972),35-70.Thatpostulate readsinpart:
"Thereexistsa class of simplesystemscalled thermometers withthe following
properties:...
(e) For everythermometer M ... therestrictionof [theheatform]'jj to an arbitrary
isotherm of M is everywhere non-zero (on thatisotherm)."
In mynotationtheaxiomjust quotedassertsthaton each sufficiently shortisothermal
segment /lK=t=0forsomebody.
To Mr.Serrín weowea deeperunderstanding oftheroleofa Thermometric Axiom;
mostofwhatI knowaboutthematter I havelearntfrom him.Inhisnotesof1975heassumes
thatforeach h thereis one bodysuch thatAvôm/ô6^0locallyon some empirical-
temperature scalecovering h.In theframework ofCarnot's ideas,hisaxiomandtheoneI
proposehereareequivalent. Inhislecture of1977heassumesonlythatforeachhthere isone
bodysuchthatAv=t= 0 locally.To within technicaldetailsthisaxiomseemstobe thesameas
partofBoyling's.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
376 C. Truesdell

Corningto the details, we firstrenderexplicitthe role of the constantof


integrationin (27) and so express(30) in the form

¡ He(x)dx),
' A=Qxp-y>0,
J
(49)
do

t0 being the value assigned to t when 6 = 6O. If to = 0, then A = l, and the


absolute temperature1 is assignedto the arbitraryempiricaltemperature60. If,
on the otherhand,we make the choice of 60 and t0 depend upon some physical
phenomenon,we may obtain a value otherthan 1 forA. By varyingthe positive
numberA we obtain all possibleabsolute-temperature scales T correspondingto
a givenconstantJ. It is to the equivalence class so obtained that we assign the
physicaldimensionof absolute temperature.Choice of A definesthe unit of a
particularscale of absolute temperatureT. Internationalconventionnow takes
60 as an empiricaltemperatureof the triplepointof waterand assignsthe value
273.16K to A.
Kelvin himselfpreferredto assign a particulardifference of temperatures
ratherthan particulartemperatures.If Tx and T2 are the values of T that
correspondto the empiricaltemperatures6Xand 02, then(49) shows us that

1 -
T ÍT T *
~'12 1') ' T /T t •>

exP yi Vo(x)dx)
di -
-(T
~^2 T)
ji> T •
JYT2
exply j>e(x)dx)-l
When 01 and 62 are given,to assignA is one and the same thingas to assign T2

For example, we may choose for the empirical-temperature scale 6 the


Celsius scale: that whichis providedby the air thermometer with0° and 100°,
respectively,assigned to the hotnessesat whichwater at standardatmospheric
pressurefreezesand boils. If we wish an absolute scale which preservesthe
differenceof 100° in the temperatures
assignedto thesehotnesses,we simplyput
T2- Tx= 100 in (50) and so obtain

100exp(-J^(x)dx)
°
T= 1Oo r-A (51)
exp íj j fie(x)dx'-l

This formula converts degrees Celsius on the air thermometerto degrees


absolute with the same differenceof temperatures, namely 100°, between the
boilingpoint and the freezingpoint of water.
It was suggestedby Helmholtz and Joule that¡iQvaried inverselyas the
temperature above absolute cold. That is, on the Celsius scale

^eh: (52)

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 377

- 9a being the Celsius temperatureof absolute cold. If this formula were


rigorouslycorrect,(51) would reduceto
T = 6 + 6a. (53)
The experimentsof Joule & Kelvin on the porous plug are interpretedas
showingthat(52) is verynearlybut not exactlytrue.Thereforetheabsolute scale
determinedby (51) is verynearly(53), which is the Celsius scale with its zero
shiftedto absolute cold. This conclusion requiresexperimentaldetermination
not only of ¿¿0but also of 9a. It has long been known that 6a is verynearly
-273°C.
The foregoingtreatmentrefersonly to temperatures, not to hotnesses.We
of
maythink each possible choice of the scale T as correspondingto a particular
hotnessh0 selectedby some rule. If we write Tho the scale that assignsto h0
for
the value 1,
Tho(h0)= l, (54)

underchange of unit
thenit is easy to demonstratethe rule of transformation
hotness:
Th2(h)= Th2(h1)Thi(h). (55)

Any two absolute scales T that correspond to one choice of the positive
constant J are proportionalto each other, and if we select two particular
hotnessesh2 and hu the constantT^hJ plays the role of a unitof temperature:

(56)
w=Thí(h)-ThÁhr
The resultspresentedat the beginningof thisappendixcan be derivedfrom(55)
and (56), providedfibe of the rightkind.I thinkthisapproach is cleareras well
as neater.For example,it shows thatif(52) held strictly,
thenthe absolute scale
definedby the hotnesshf at whichwater freezeswould be given as followsin
termsof the Celsius scale 6:
9(h)+ 8a

10. Conclusion
The conceptof absolute temperaturehas always been inherentin Carnot's
generaltheory.To constructtheabsolutetemperaturesintroducedby Kelvin, we
requireneitherthe "First Law" nor the "Second Law" of thermodynamics.

11. Appendix.Axiomsof Classical Thermodynamics


In Conceptsand Logic, Appendix to Chapter 15, Mr. BHARATHA& I have
pointed out one way to modifyour presentationso as to deliver classical

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
378 C. Truesdell

thermodynamics quickly: Adopt Clausius' "FirstLaw" (40). In §6 of thisnote I


haveshownthatthecorresponding specialchoiceofJ in(30) thengivesto Kelvin's
secondabsolutetemperature T exactlythepropertyhe desiredforit,namely(43)2,
a propertywhichit most certainlydoes not have forotherchoices of J.
However,I thinkthe basic programof Conceptsand Logic is more elegant:
Proveboththe"FirstLaw" and the"Second Law" as theorems based upon axioms
whichrenderformalCarnot's own assumptions,of course not restrictedby his
subsidiaryand undesirableadherenceto theCaloric Theoryofheat. In Concepts
andLogic we attainedthisgoal byourAxiom V: Thereis a bodyofidealgas suchthat
the whole V-6 quadrant is a thermodynamic part, and both specificheats are
constant.There we used the conceptof ideal-gas temperature.
WhileinthisnoteI haveexcisedall use ofideal gasesinthesenseofthepioneers,
ifwe startfromKelvin's second absolute temperatureT as definedin termsof
an arbitrary positiveconstantJwecanbymodifying a littletheaxiomaticstructureof
Conceptsand Logic uphold its program.By usingabsolute temperatures only we
can againproveboththeFirstLaw and theSecond Law as theorems.I now present
the details.
We replace the primitiveconceptof temperatureby the primitiveconceptof
hotnessand lay down
Axiom0. Thesetofall hotnessesis a dijfeomorph
ofa real interval,
equippedwith
an intrinsic
ordering.
Definition0. An empirical-temperature
functionis an order-preservingcharton a
boundedsegmentof thehotnessmanifold.
AxiomsI- III of Conceptsand Logic are thento be understoodas referring to
someone empirical-temperature scale and henceto all suchscales on one bounded
segment.
Axiom IV shouldbe modifiedslightly:Anytwofluidbodieswhichmayundergo
Carnot cycles withthe same operatingtemperatures do in thosecycles the same
amountof workper unitof heat absorbed.
Nextwe laydowntheThermometric Axiomstatedabove in§ 8. On thatbasis we
may constructin of
terms anypositive constantJ theabsolute-temperature scale T
on all ofJťth.
We may thenconvertall thelocal restrictions shownto be valid for
someone empirical-temperature scale intostatements valid forall T. Most ofthese
have been listedabove in § 6.
Definition.A bodyof ideal gas is a bodysuch that

, „ rr ťpV = RT, /*= const.>0, v '


(58)
for all T.
Thetemperature measured bythevolumesofsucha bodyat constant pressure provides"a
naturalmeasureof temperature, ... a real temperature,... a sortof Platonicidea of
temperature,"ofwhichthetemperature readonan airthermometer is"onlyan incomplete,
The resultsofJoule & Kelvin's experiment
inexactexpression". withtheporousplug
showthattheinexactness oftheairthermometer is negligibleformostpurposes. Thuswe
arenotastonished to learnthattheidea "absolutecold" or "absolutezero"is 150years
olderthantheidea "absolutetemperature". The relations(31)l9(32)l9(34),and (35) were
obtainedbyuse ofthe"Platonicidea" ofan idealgas.
all first

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
fromCarnoťsAxiom
AbsoluteTemperatures 379

Appealingto (32) and (4)3, we see thatthespecificheatsofan ideal gas referred


to the scale T satisfythe relation
=
J(KPtT-KVtT) R, (59)

J being the positiveconstantused to defineT This relation,like (58) itself,is


familiarbecause of its counterpartwhen ideal-gas temperatureis used, but the
dimensionsofthequantitiesoccurringin thetwoare notthesame.As theabsolute
temperatureT is dimensionless, R has thedimensionsof work,and Kv T has the
dimensionsof heat. For each choice of the arbitraryconstantJ we obtain an
absolutetemperature T throughthedefinition (30), so thedefinition (58) of"ideal
gas" dependsupon J. A substancethat is an ideal gas for one choice ofJ is not an
ideal gas foranother. For any choice of J we see that (59) holds. Thus withany
choice of J the two specifcheats are distinct; all of the V-T quadrant that
to is
corresponds Jťth normal; and both specificheats are constant ifand only if
one of themis. The ruleoftransformation (25)2 when applied to (30) fordifferent
choices Jxand J2of J shows that
= Jf±Kv<Tl.
Kv,Tl (60)
The exampleoftheCaloric Theoryofheat showsthatthereneed be no choice ofJ
such as to make KpT and Kv T constant(Conceptsand Logic,HistoricalScholion
afterDefinition13 in Chapter6).
We are now readyto lay down a finalaxiom similarin formto AxiomV of
Conceptsand Logic but conceptuallyquite different:
AxiomV. Thereis a J suchthatforone ofthecorresponding
bodiesofideal gas
Kv T= const.
To sucha gas,whichAxiomV makescompatiblewiththegeneraltheory,we may
applythereasoningthatleads to Theorem15 in Chapter15 of ConceptsandLogic.
Hence the functionsgT and hT occurringin (34)2 have the forms

gT= JhT+ const,


hT= MT,
and (37)2 reduces to (45)2, which is the local "Second Law" of classical
thermodynamics. The otherclassicalconstitutive namely(32)2,we have
restriction,
already for all choices of J, so it holds for the choice providedby Axiom V.
Thustheentireformalstructure of classical thermodynamics results.In particular,
one demonstratedconsequence of the axioms here is the "First Law" in
Clausius' form (40), and the constant J that Axiom V providesturnsout to be
the uniformand universalmechanical equivalent of a unit of heat in cyclic
processes.
The Thermometric Axiom of§ 8 has enabled us to constructKelvin's second
absolute-temperature upon thegeneralframework
scale ofCarnot's ideas. Using
it,we have shownthatthoseideas,dulymodifiedby AxiomV, suffice to construct
classical thermodynamics withoutuse of ideal-gastemperature, the "First Law",
or the"Second Law".

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions
380 C. Truesdell

Acknowledgment. I am gratefulnot onlyto Mr. Serrín but also to Messrs.C.-S. Man


and M. Pitteri forlongand patientdiscussionofthesubjectand forscrupulouscriticismof
thisnote. The workit presentswas partiallysupportedby grantsfromthe U.S. National
Science Foundation's Programsin the Historyand Philosophyof Science, in Applied
Mathematics,and in Solid Mechanics.

The JohnsHopkins University


Baltimore,Maryland

(Received December75, 1978)

Note addedin proof.The demonstrations of thetheoremsin Conceptsand


Logic on which this paper drawsdo notrequirethatdKv/d6exist;(25)2shows
thatits existenceis not invariantunderall changesof empirical-temperature
scale. Accordingly, in this paper I have supposedof Kv only that it be
continuous and thatdKv/õVexistand be continuous.

This content downloaded from 188.72.127.79 on Sat, 14 Jun 2014 05:07:01 AM


All use subject to JSTOR Terms and Conditions

Você também pode gostar