Sunteți pe pagina 1din 12

Thermal physics in the introductory physics course:

Why and how to teach it from a unified atomic perspective


Frederick Reifa)
Center for Innovation in Learning and Department of Physics, Carnegie Mellon University, Pittsburgh,
Pennsylvania 15213
共Received 7 April 1999; accepted 11 June 1999兲
Heat and thermodynamics are traditionally taught in the introductory physics course from a
predominantly macroscopic point of view. However, it is advantageous to adopt a more modern
approach that systematically builds on students’ knowledge of the atomic structure of matter and of
elementary mechanics. By focusing on the essential physics without requiring more than elementary
classical mechanics, this approach can be made sufficiently simple to be readily teachable during
five or six weeks of an ordinary calculus-based introductory physics course. This approach can be
highly unified, using atomic considerations to infer the properties of macroscopic systems while also
enabling thermodynamic analyses independent of specific atomic models. Furthermore, this
integrated point of view provides a deeper physical understanding of basic concepts 共such as internal
energy, heat, entropy, and absolute temperature兲 and of important phenomena 共such as equilibrium,
fluctuations, and irreversibility兲. © 1999 American Association of Physics Teachers.

I. INTRODUCTION thermal physics. The newly introduced concepts lack a


strong integrating framework and the fundamental signifi-
Thermal physics 共heat and basic thermodynamics兲 has tra- cance of absolute temperature or entropy remains obscure.
ditionally been a part of the introductory college-level phys-
ics course. Not only does this topic deal with the macro-
scopic systems most commonly observed in everyday life, B. Advantages of an atomistic approach
but it is also centrally important in the chemical, biological,
and engineering fields of primary interest to many students. An alternative instructional approach, considering macro-
Furthermore, it is increasingly relevant to present-day scien- scopic systems from a microscopic atomic point of view,
tific and practical concerns with the behavior of complex offers the following advantages.
systems ranging from novel properties of condensed matter 共1兲 Highly unified approach. It provides a highly unified
to environmental issues. approach explicitly relating the macroscopic and atomic
But how can thermal physics best be taught in an intro- points of view and showing how they are complementary. In
ductory physics course? This question is addressed in the particular, the approach can be based on only two fundamen-
present article. In particular, Sec. II points out several rea- tal ideas—共a兲 that all macroscopic systems consist of atomic
sons why it is advantageous to teach the subject from a mod- particles and 共b兲 that their properties can be understood by
ern atomistic perspective rather than from the more tradi- using some basic knowledge of mechanics and some simple
tional macroscopic point of view. Section III discusses a probability considerations. The approach leads to all the
simple approach whereby this perspective can be imple- purely macroscopic relationships and reasoning processes
mented in a way suitable for first-year students in an intro- encountered in the traditional approach. But it also provides
ductory physics course. Finally, Sec. IV summarizes the ben- much deeper insights, making clear connections to atomic
efits and limitations entailed by this approach. knowledge and indicating how such knowledge can be used
to infer the properties of macroscopic systems.
共2兲 Modern point of view. The approach reflects a more
II. MACROSCOPIC VERSUS ATOMISTIC modern point of view which builds on present-day knowl-
APPROACH edge of the atomic structure of matter and is more in tune
A. Traditional macroscopic approach with the kinds of thinking prevalent in contemporary physics
and other scientific fields.
Thermal physics, following historical lines of develop- 共3兲 Greater interest to students. The approach is more in-
ment, has traditionally been taught from a predominantly teresting to many students. Instead of dealing with the seem-
macroscopic point of view. Thus basic concepts 共such as ingly mundane questions of heat and temperature, it shows
heat and entropy兲 are introduced largely on the basis of mac- how underlying atomic processes can provide an understand-
roscopic considerations and only some scattered attempts are ing of many commonly observed macroscopic phenomena.
made to relate them to atomic processes. Furthermore, this understanding can directly build on the
This macroscopic approach is not easy for students be- knowledge of mechanics that students have previously ac-
cause it is rather abstract and provides no readily visualizable quired in their course.
mental models. Thus concepts such as heat, internal energy, 共4兲 Introduction to widely useful ideas. Instead of dealing
absolute temperature, and entropy remain abstractly descrip- mostly with thermodynamic arguments, the atomistic ap-
tive and largely phenomenological. The emphasis on under- proach introduces some important ideas useful to students in
lying mechanisms, so prevalent in the rest of the introductory their future studies. 共a兲 It introduces some basic probability
physics course, seems strangely absent in this discussion of notions that are widely relevant in statistics, in the analysis

1051 Am. J. Phys. 67 共12兲, December 1999 © 1999 American Association of Physics Teachers 1051
of experimental results, in quantum mechanics, and in many mechanics 共for example, knowledge about position and ve-
other domains. 共b兲 It not only derives and applies the model- locity, about kinetic and potential energy, and about the en-
independent laws of thermodynamics, but also relates them ergy law兲; 共b兲 some basic mathematical knowledge about
to atomic processes and statistical physics. Hence, the ap- natural logarithms and simple derivatives and integrals.
proach introduces ideas that prepare students to deal with Needless to say, the atomistic approach must be carefully
future courses in thermodynamics, statistical mechanics, formulated so that it is sufficiently simple to be comprehen-
physical chemistry, and other fields. sible to first-year students, yet provides them with significant
abilities to explain or predict important phenomena. In the
C. Instructional feasibility of an atomistic approach rest of this section I outline such a pedagogical formulation
of the atomistic approach and try to do so in sufficient detail
The central challenge is to formulate the atomistic ap- that it can be readily imitated by other instructors.4
proach in a manner that reveals its essential insights and
purely macroscopic implications, but that is sufficiently
simple to be readily accessible to students in an introductory A. Macroscopic versus atomistic descriptions
college-level physics course.
My earlier attempts1,2 to reformulate the atomistic ap- 1. Macroscopic description
proach in simpler ways presupposed a knowledge of quan- The solids, liquids, gases, and other systems observed by
tum mechanics. However, further reflection convinced me us in everyday life are large scale or ‘‘macroscopic,’’ that is,
that the introduction of thermal physics could be further sim- very large compared to the size (10⫺10 m) of a typical atom
plified, both to clarify the underlying physics and to make or molecule. Such a system can be described by relatively
the subject readily comprehensible to much less well- few macroscopically measurable quantities. 共For example, a
prepared students. In particular, it seems possible to simplify gas or liquid in a time-independent equilibrium situation can
the discussion in some of the following ways. be adequately described by its volume and pressure.兲
共1兲 Instead of requiring a prior knowledge of quantum
physics, one can build on the basic ideas of classical me-
chanics learned by students at the very beginning of their 2. Atomistic description
introductory physics course. Although a classical discussion We also know that a macroscopic system consists of a
has some limitations compared to a more correct quantum-
very large number 共of the order of 1024兲 atoms or molecules.
mechanical treatment, it is sufficient to yield almost all the
Instead of describing such a system from a macroscopic
important reasoning processes and results. Furthermore,
point of view, we may also describe it from an atomistic
most of these results require only a slight reinterpretation
point of view that focuses attention on every atomic particle
after students have become familiar with quantum mechan-
in the system. For example, we might do so by specifying the
ics.
position and velocity of every particle in the system. Corre-
共2兲 It is sufficient to introduce the basic notion of a prob-
spondingly, the atomistic description5 would be very com-
ability, but quite unnecessary to discuss binomial or Gauss-
ian distributions. plex, requiring the specification of something like 1024 quan-
共3兲 Qualitative discussions can often help to introduce tities.
more quantitative considerations. Some formal mathematical
derivations can also be replaced by more insightful physical 3. Relation between descriptions
arguments.
Instructional experience in introductory physics. These What is the relation between the macroscopic and atomis-
considerations led me, about 20 years ago, to use the atom- tic descriptions? In particular, we know that all matter con-
istic approach to teach thermal physics in the usual calculus- sists of only a few kinds of atomic particles—and we also
based introductory physics course. The first semester of that know something about the laws of mechanics describing the
course dealt mostly with mechanics and devoted the last five motions and interactions of such particles. How can this ba-
or six weeks to thermal physics. Instead of following the sic knowledge help us to explain or predict the observable
textbook by Halliday and Resnick,3 I then taught this subject properties of the diverse macroscopic systems commonly en-
from an atomistic perspective and provided students with countered by us? This is a central question that we wish to
some special instructional materials prepared by me. answer in our study of physics.
Several years of successively refined teaching efforts led
me to the conclusion that I could realize the previously men- B. Thermodynamic energy law
tioned advantages of the atomistic approach, and yet pre-
serve practically all the macroscopic results and reasoning 1. Energy law of mechanics
traditionally taught in thermal physics. Furthermore, my ex-
perience indicated that students did not find the atomistic The energy law of mechanics, applied to any system of
approach more difficult than the traditional approach or than particles, can be written as
the classical mechanics studied by them during the earlier ⌬E⫽W ext , 共1兲
part of the course.
where E⫽K⫹U is the ‘‘internal energy’’ of the system 共that
III. SIMPLE ATOMISTIC APPROACH TO is, its energy when its center of mass is at rest兲. Here K is the
THERMAL PHYSICS total kinetic energy of all the particles in the system, and U is
the total potential energy of interactions among all these par-
Students already know that matter consists of atoms and ticles. The external work W ext is the work done on the system
molecules. They need only the following other modest pre- by all external interactions not included in this potential en-
requisites: 共a兲 some rudimentary knowledge about classical ergy.

1052 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1052
2. Thermodynamic energy law roscopic motion in the presence of a macroscopic force兲 or
eliminate any absorbed heat 关for example, by covering the
This external work can be decomposed into some macro-
system with an ‘‘adiabatic’’ 共that is, thermally insulating兲
scopic work W mac and some remaining nonmacroscopic
cover兴. Thus one can do the following. 共a兲 Any arbitrary
work W nonmac . Thus Eq. 共1兲 can be expressed in the form thermometer 共for example, a mercury thermometer or elec-
⌬E⫽W mac⫹W nonmac . 共2兲 trical thermometer兲 can be used to indicate changes in the
system’s internal energy. 共b兲 Heat transfer can be eliminated
The macroscopic work W mac 共done on the system by some by covering the system with an adiabatic cover.11 Under
macroscopically measurable force moving through some these conditions one can determine the change of thermom-
macroscopically measurable distance兲 can simply be called eter indication 共‘‘temperature’’兲 corresponding to any mea-
the work W on the system. The remaining nonmacroscopic surable amount of work done on the system, that is, corre-
work W nonmac 共done, because of atomic interactions, on par- sponding to any change ⌬E of the system’s internal energy.
ticles in the system by neighboring particles outside the sys- 共c兲 After this calibration, a measured change in thermometer
tem兲 is conventionally called the heat Q absorbed by the indication can always be used to determine the correspond-
system.6 With these definitions the relation 共2兲 becomes the ing change ⌬E of the system’s internal energy, even in the
‘‘thermodynamic energy law’’ general situation where no adiabatic cover is in place. But in
⌬E⫽W⫹Q, 共3兲 this case the change ⌬E may also be due to absorbed heat
and, according to Eq. 共3兲, this heat Q⫽⌬E⫺W. Thus the
which is historically called the ‘‘first law of
absorbed heat can be found by simple subtraction.
thermodynamics.’’ 7
The macroscopic work W can be readily determined by
macroscopic measurements. But how can such macroscopic 5. Applications of the thermodynamic energy law
measurements be used to determine the internal energy E and Any system in a time-invariant macroscopic equilibrium
heat Q, quantities that are so closely related to atomic par- situation 共‘‘macrostate’’兲 has some well-specified internal
ticles and their interactions? energy. When a system changes from one situation to an-
other, its internal energy changes by a definite amount ⌬E
3. Useful analogy independent of the process producing this change. However,
These issues can be clarified by the following analogy.8 different processes 共involving different amounts of work W
The amount A of water in a lake can change either 共a兲 be- and heat Q兲 can be used to achieve a change between the
cause of the flow F of water into the lake from some streams same two situations. The thermodynamic energy law 共3兲 then
or 共b兲 because of the condensation C of water from moisture implies that the sum W⫹Q must be the same for any such
in the surrounding air.9 Thus the change in the amount of process.
lake water is equal to All applications of the thermodynamic energy law exploit
this basic conclusion. Correspondingly, students can be
⌬A⫽F⫹C. 共4兲 given all of the usual illustrative problems and examples
This relation is analogous to the thermodynamic energy law discussed in a conventional macroscopic approach to
共3兲. Here the flow F of water from streams is readily visible thermodynamics.12
and measurable. But how can one measure the other two
quantities ⌬A and C? C. Statistical description of macroscopic systems
This measurement can be done because one can separately
1. Statistical description
eliminate flow 共for example, by closing dams in the streams兲
or eliminate condensation 共for example, by covering the lake In many cases it is practically 共or theoretically兲 impossible
with a water-impermeable cover兲. Thus one can do the fol- to make definite predictions about the behavior of a system.
lowing. 共a兲 Any level indicator 共for example, some arbi- For example, one cannot definitely predict the result of toss-
trarily shaped stick inserted into the bottom of the lake兲 can ing a pair of coins or winning a jackpot, nor can one defi-
be used to indicate changes in the water level and thus cor- nitely predict when someone will die or get into an automo-
responding changes in the amount of water in the lake. 共b兲 bile accident.
Condensation can be eliminated by covering the lake with an However, in such cases one may still be able to predict the
impermeable cover.10 Under these conditions one can deter- probabilities of various events, that is, one may still be able
mine the change of water level corresponding to any mea- to make statistical predictions. Indeed, such predictions are
surable amount of water flow into the lake 共that is, corre- often of great practical utility 共for example, in gambling or
sponding to any change ⌬A in the amount of water in the the insurance business兲.
lake兲. 共c兲 After this calibration, a measured change in water A statistical description of a system requires the following
level can be used to determine the corresponding change ⌬A four basic ingredients 共all of which can be illustrated in the
in the amount of water in the lake, even in the general situ- simple case of a pair of tossed coins兲.
ation where no cover is in place. But in this case the change 共1兲 Specification of possible states. One must be able to
⌬A may also be due to condensation and, according to Eq. specify the possible results obtained by observing the sys-
共4兲, this condensation C⫽⌬A⫺F. Thus the condensation tem. 共For example, in the case of the pair of tossed coins,
can be found by simple subtraction. each coin can display either heads or tails. Hence, this sys-
tem can be in any one of four possible states.兲
共2兲 Statistical assembly of systems. Instead of considering
4. Measurability of internal energy and heat a single system, one can consider an assembly 共or ‘‘en-
Similarly, one can separately eliminate any work done on semble’’兲 of a very large number N of similar systems sub-
a macroscopic system 共for example, by suppressing any mac- ject to similar conditions. If Nr of these systems are in a

1053 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1053
particular state r, the fraction P r ⫽Nr /N of systems in this equilibrium situation and to make qualitative predictions
state is called the ‘‘probability’’ of finding a system in this about the approach to equilibrium. The entire remaining dis-
state. cussion of macroscopic systems will be based on this postu-
Although one may not be able to predict the outcome of an late.
experiment on a single system, one may still be able to pre-
dict the probability of the outcome occurring in such a sta- D. Qualitative properties of macroscopic systems
tistical assembly of similar systems. 共For example, although
one cannot predict the outcome of tossing one particular pair 1. Equilibrium situation
of coins, one may still be able to predict the probability of
any outcome resulting from tossing many such pairs of a. Probability of an equilibrium situation. When a mac-
coins.兲 roscopic system is in a particular situation 共or ‘‘macroscopic
共3兲 Statistical postulate. To predict such probabilities, one state’’兲 A, it can be found in any one of a large number ⍀ A of
must formulate a basic assumption 共or ‘‘postulate’’兲 serving possible basic states 共compared to the total number ⍀ tot of
as the basis of all such theoretical predictions. 共For example, possible basic states available to the system in all situations兲.
in the case of the pair of tossed coins, one might assume that According to the statistical postulate, the probability P A that
all possible states occur with equal probability.兲 the system in equilibrium is found in the situation A is then
共4兲 Statistical predictions. The statistical postulate can be P A ⫽⍀ A /⍀ tot⬀⍀ A , 共5兲
used to calculate the probabilities of various outcomes. A
successful statistical theory is one for which these calculated that is, it is proportional to the number ⍀ A of basic states in
predictions agree with the observed probabilities.13 A.
b. Example of a simple gas in equilibrium. A gas, con-
sisting of N molecules in a box, can serve as an instructive
2. Statistical theory of macroscopic systems example. The position of each molecule can be adequately
A macroscopic system consists of a very large number of specified by indicating whether it is in the left or right half of
atomic particles. A detailed consideration of the motions and the box. Thus each molecule can be in two basic states. The
interactions of all these particles would be a hopelessly com- total number ⍀ tot of possible states for the entire gas of N
plex task. However, a less detailed statistical consideration molecules is then 2⫻2⫻2⫻¯⫽2 N . The probability P n of
can be relatively simple and effective. The needed ingredi- a situation, where n of these molecules are in the left half of
ents are the following 共similar to those previously illustrated the box, can then be found merely by counting the number of
with the pair of tossed coins兲. possible basic states in this situation.
共1兲 Specification of possible states. The ‘‘basic state’’ of a For example, for a gas of 4 molecules, the probabilities
system14 共that is, its state described from an atomistic point that 0, 1, 2, 3, or 4 molecules are in the left half of the box
of view兲 can be specified by the position and velocity of are 1/16, 4/16, 6/16, 4/16, and 1/16. The students can simi-
every one of its atomic particles.15 Thus one needs to specify larly be asked to count the number of possible states 共and to
an enormous number of quantities, the three position coordi- list the corresponding probabilities兲 for a gas of 5 or 6 mol-
nates 共x,y,z兲 of every particle, and its three velocity compo- ecules. Furthermore, one can show them graphs illustrating
nents ( v x , v y , v z ). 16 the probability distributions obtained for a gas consisting of a
The description can be somewhat simplified by ‘‘digitiz- larger number of molecules 共such as N⫽40兲.18
ing’’ the position coordinates and velocity components, that c. General conclusions. The preceding examples are suf-
is, by replacing these continuous quantities by discrete inte- ficient to arrive at the following general conclusions. 共a兲 The
gral values 共each of which denotes any of the values lying in most probable situation is one where there are an equal num-
a small range determined by the desired accuracy of ber of molecules in both halves of the box 共because this
description兲.17 This digitizing makes the possible basic states situation corresponds to the largest number of possible ways
of a system discrete and countable. that the molecules can be distributed between the two
共2兲 Statistical assembly of systems. Instead of focusing at- halves兲. 共b兲 Situations where the number of molecules in the
tention on a single system, consider probabilities of occur- two halves are quite different are highly improbable. 共For
rence in a statistical assembly of a large number N of similar instance, an extreme situation where no molecule is in the
such systems. 共For example, instead of focusing attention on left half of the box can be realized in only one way and is
a single container of gas, consider an assembly of many such thus exceedingly unlikely.兲 共c兲 When the total number N of
containers.兲 The aim is to predict such probabilities theoreti- molecules is large 共especially, in realistic gases where N
cally. ⬇1024兲, situations where the molecules are nearly equally
共3兲 Statistical postulate. Nothing in the laws of mechanics distributed between the two halves of the box are over-
suggests that a system should be preferentially in any one of whelmingly more probable than any others.
its possible basic states rather than in any other. Similarly,
repeated tossing of a set of coins will ultimately lead to a 2. Fluctuations in equilibrium
situation where they are equally likely to be found in any one
of their possible states. Such considerations suggest the fol- The preceding example indicates that the number n of
lowing basic assumption or ‘‘statistical postulate:’’ The molecules in one half of the box is not always exactly equal
probabilities of finding an isolated macroscopic system in its to N/2, but may slightly deviate from this value. Deviations
possible basic states change with time until the system 共‘‘fluctuations’’兲 that are appreciable compared to N itself
reaches a time-independent equilibrium situation where are, however, very improbable when the total number N of
these probabilities are equal. molecules is large.
共4兲 Statistical predictions. The statistical postulate allows Indeed, small fluctuations are always present in the case of
one to make quantitative predictions about probabilities in an any macroscopic system in equilibrium. But these fluctua-

1054 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1054
tions are ordinarily observable only if one tries to make very example, planets moving around the sun兲 and of processes
precise observations or deals with sufficiently small involving various macroscopic systems 共for example, a
systems.19 bouncing ball coming to rest or a chicken crawling out of an
egg兲.
3. Approach to equilibrium
E. Number of basic states and entropy
According to the statistical postulate of Sec. III C 2, any
isolated macroscopic system will change until it reaches an 1. Basic states and macroscopic parameters
unchanging equilibrium situation. However, the postulate Probabilities can be calculated from Eq. 共5兲 if one knows
makes no quantitative statement about the time required to the number of basic states available to a macroscopic system.
reach equilibrium. 共The calculation of such a time would But how does the number ⍀ of such basic states depend on
require much more detailed analyses.兲 the macroscopic parameters describing a system, for ex-
For example, suppose that a gas is initially in the left half ample, on its volume V or internal energy E? This question
of a box divided by a partition into two equal parts. If a hole can be readily answered in the simple case of an ideal mon-
is made in the partition, the gas will ultimately always reach atomic gas of N molecules. 共A gas is considered ‘‘ideal’’ if
an equilibrium situation where approximately half the mol- its intermolecular interactions contribute negligibly to its in-
ecules are in either part of the box. But the time required to ternal energy.兲
reach this equilibrium situation will be longer if the hole is a. Dependence of ⍀ on volume. Suppose that the volume
small than if it is large. V of the gas container were 2 times larger. Then 2 times as
many possible 共digitized兲 positions would be available to
4. Irreversibility each molecule, and hence 2⫻2⫻2⫻¯⫽2 N times as many
a. Reversible and irreversible processes. Because a mac- positions to all the N molecules. More generally, the number
roscopic system tends to approach an equilibrium situation, it ⍀ of possible basic states depends then on the volume so that
is possible to specify a preferred direction of time and thus to
⍀⬀V N . 共6兲
distinguish the future from the past. More precisely, imagine
that one makes a movie of a process. If one can distinguish b. Dependence of ⍀ on kinetic energy. Similarly, suppose
whether the movie is run through a projector forward or that the kinetic energy K of the gas were 2 times larger. On
backward, then one has identified a special direction of time average, the contribution m v 2x /2 to the kinetic energy of a
and the process is called ‘‘irreversible.’’ But if one cannot molecule’s velocity component v x would then be 2 times
tell whether the movie is run through this projector forward
larger. Hence v x would be 2 1/2 times larger so that the num-
or backward, then one has not identified a special direction
of time and the process is called ‘‘reversible.’’ ber of possible 共digitized兲 basic states corresponding to it
b. Example of irreversibility in a gas. For example, con- would also be 2 1/2 times larger. Because there are 3N such
sider the process where a gas, initially entirely confined in velocity components, the total number of basic velocity
the left half of a box, spreads throughout the entire box until states available to all the N molecules would then be 2 1/2
it reaches equilibrium where nearly an equal number of mol- ⫻2 1/2⫻2 1/2⫻¯⫽2 3N/2 times larger. More generally, the
ecules are in the two halves. If a movie of this process is number ⍀ of possible basic states depends then on the ki-
played backward through a projector, it would show the gas netic energy K of the gas so that
molecules, initially uniformly distributed throughout the en-
tire box, spontaneously all moving into the left half of the ⍀⬀K 3N/2. 共7兲
box. Such a process would never be observed in reality. c. Qualitative conclusions. Because N⬇1024, the expo-
Hence one could immediately tell that the movie is being nents in 共6兲 and 共7兲 are extremely large. Hence, the number
played backward and that the originally filmed process is of possible basic states increases exceedingly rapidly as the
irreversible.
volume V or kinetic energy K of the gas becomes larger.
c. Irreversibility and probabilities. Strictly speaking, the
reverse process 共where the gas becomes all concentrated in
the left half兲 is not impossible, just exceedingly improbable. 2. Basic entropy
For example, all the molecules in the right half of the box Because the number ⍀ of basic states is extremely large
might possibly all be moving to the left at a particular in- and rapidly increasing, it is much more convenient to work
stant. But, in the case of some 1024 molecules, such a spon- with its logarithm,
taneous fluctuation would occur with utterly negligible prob-
␴ ⫽ln ⍀, 共8兲
ability.
To show that there is no sharp distinction between irre- which is much less large and less rapidly increasing. The
versible and reversible processes, consider a gas which con- quantity ␴ may be called the ‘‘basic entropy’’ of a system.
sists of only 4 molecules in a box. A spontaneous fluctuation, The basic entropy provides a measure of a system’s ran-
where all molecules are in the left half, would then 共in equi- domness. A situation which can be realized in many different
librium兲 occur with a probability of 1/16. Hence, it would be ways is considered more random than one which can be re-
large enough to be readily observed. If a movie showed this alized in only a few ways. For example, a situation where the
gas concentrating in the left half of a box, one would then molecules of a gas are anywhere within a box can be realized
not be able to tell whether this movie is being played forward in many more ways than one where they are all in only one
or backward. Thus this process would seem reversible.20 half of the box; hence it is a more random situation. Simi-
d. Other examples. Students can be shown many other larly, a hand consisting of any 13 red cards is more random
examples. For instance, they can be asked to judge the re- than a hand consisting only of diamonds 共a situation that can
versibility of processes involving only a few particles 共for be realized in only one way兲.

1055 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1055
one of its ⍀ ⬘ possible basic states. But because X itself can
be in any one of its ⍀ possible basic states, the total number
⍀ * of basic states available to the entire system X * is
⍀ * ⫽⍀⍀ ⬘ . 共14兲
Because the corresponding logarithms of these quantities are
additive, the entropies of these systems are also additive.
Thus
␴ *⫽ ␴ ⫹ ␴ ⬘ or S * ⫽S⫹S ⬘ . 共15兲
Fig. 1. Qualitative dependence of a system’s entropy on its internal energy.

F. Thermal interaction
3. Properties of the entropy Consider two macroscopic systems X and X ⬘ which are in
a. Dependence of entropy on internal energy. Consider a contact so that they can exchange energy in the form of heat,
system of fixed volume. If it is a monatomic ideal gas of N but do not do any macroscopic work. 共For example, the vol-
molecules, its internal energy E⫽K because its potential en- ume of each system remains fixed.兲 If the composite system
ergy U is negligible. The basic entropy of this gas is, by Eq. X * is isolated, its internal energy E * remains constant. Thus
共7兲, equal to the energies E and E ⬘ of the two systems must satisfy the
relation
␴ ideal gas⫽ln ⍀⫽constant⫹ 共 3N/2兲 ln E. 共9兲
E⫹E ⬘ ⫽E * ⫽constant. 共16兲
A more complex macroscopic system 共such as a liquid or a
solid兲 has not only kinetic energy, but also an appreciable The energy of each system can then have any value, but
potential energy due to the mutual interactions of its con- changes of the systems’ energies must be related so that
stituent atomic particles. The basic entropy ␴ of such a sys-
tem depends on its internal energy E in a way that is more ⌬E ⬘ ⫽⫺⌬E. 共17兲
complicated than in Eq. 共9兲, but also increases with its inter- 1. Thermal equilibrium
nal energy, roughly in the manner indicated by the graph in
Fig. 1. a. Probability of a particular energy distribution. Sup-
b. Entropy gradient. When a system’s internal energy pose that two systems are in equilibrium. What is the prob-
changes by a small amount dE, its basic entropy changes by ability P(E) of a situation where the system X has an energy
a corresponding small amount d ␴ . The quantity E 共while the other system X ⬘ has the remaining energy E ⬘
⫽E * ⫺E兲?
␤ ⫽d ␴ /dE 共10兲
According to Eq. 共5兲, this probability must be proportional
describes how rapidly the system’s basic entropy changes to the number ⍀ * of basic states available to the total system
with increasing energy. It can be called the ‘‘entropy gradi- X * when X has the energy E. Hence Eqs. 共14兲 and 共15兲 imply
ent’’ and specifies the slope of the graph in Fig. 1. that
c. Entropy and absolute temperature. For historical rea-
sons 共and partly for ease of measurement兲 it is convenient to P 共 E 兲 ⬀⍀ * ⫽⍀ 共 E 兲 ⍀ ⬘ 共 E ⬘ 兲 , 共18兲
introduce two other quantities closely related to the funda- or
mental quantities ␴ and ␤. Thus the quantity S, simply called
‘‘entropy,’’ is defined so that ln P 共 E 兲 ⫽constant⫹ ␴ * ⫽constant⫹ ␴ 共 E 兲 ⫹ ␴ 共 E ⬘ 兲 . 共19兲
S⫽k ␴ ⫽k ln ⍀ 共11兲 b. Most probable situation. The number of basic states is
and the quantity T is defined so that an extremely rapidly increasing function of the energy. As
the energy E increases and E ⬘ correspondingly decreases, the
T⫽1/共 k ␤ 兲 . 共12兲 factor ⍀(E) in Eq. 共18兲 increases extremely rapidly while
The quantity T is called the ‘‘absolute temperature,’’ for rea- the factor ⍀ ⬘ (E ⬘ ) decreases extremely rapidly. The result is
sons clarified in the next section. The value of the constant k an extremely sharp maximum in the probability P(E), or a
共‘‘Boltzmann’s constant’’兲 will be discussed later and de- correspondingly less sharp maximum in its logarithm.22 At
pends on international conventions about measurement stan- this maximum ln P(E) does not change if the energy E
dards. changes by an infinitesimal amount dE. Hence Eq. 共19兲 im-
d. Energy dependence of ␤ or T. The quantity ␤, speci- plies that the maximum occurs when
fying the slope of the graph in Fig. 1, is positive and de-
creases with increasing internal energy. Correspondingly, the 0⫽d ␴ /dE⫹d ␴ ⬘ /dE⫽d ␴ /dE⫺d ␴ ⬘ /dE ⬘ ⫽ ␤ ⫺ ␤ ⬘
absolute temperature T is also positive, but increases with so that
increasing internal energy.21
The definitions 共10兲 and 共12兲 imply that the quantity kT ␤⫽␤⬘ or T⫽T ⬘ . 共20兲
has the units of energy and that
In equilibrium, the most probable situation is thus one
d ␴ ⫽ ␤ dE or dS⫽dE/T. 共13兲 where the energy E of the system X has a value E eq such that
e. Entropy of a composite system. Consider a macro- the absolute temperatures T and T ⬘ of the systems are
scopic system X * consisting of two macroscopic parts X and equal.23 共As usual, very small energy fluctuations occur with
X ⬘ . If X is in a particular basic state, X ⬘ can then be in any lower probabilities around this equilibrium value of the en-

1056 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1056
ergy.兲 The equality of temperatures in thermal equilibrium consists of the kinetic energy K c.m. due to the center-of-mass
indicates that T has the properties of a temperature and mo- motions of these molecules and the intramolecular energy
tivates its name ‘‘absolute temperature.’’ E intra due to the atoms moving relative to the center of mass
2. Approach to thermal equilibrium of each molecule. Thus

If the absolute temperatures of the systems X and X ⬘ are E⫽K c.m.⫹E intra . 共22兲
initially not equal, they will not be in equilibrium when The system involving the center-of-mass motions of the mol-
brought into contact with each other. Hence, they will ex- ecules, and the system involving the motions of atoms within
change energy until they reach equilibrium. Except for very the molecules, can exchange energy with each other. Accord-
small fluctuations, they will then reach the most probable ing to Eq. 共20兲, these systems are characterized by the same
situation where the entropy S * of the entire system is maxi- absolute temperature T. Furthermore, the center-of-mass ki-
mum and where the absolute temperatures T and T ⬘ are netic energy K c.m. is related to the temperature T in the same
equal. way specified by 共21兲 for a monatomic gas without any in-
The entropy S * of the entire system thus increases in the tramolecular motions.
process of reaching equilibrium. Also Eq. 共17兲 indicates that c. Dependence of internal energy on volume. The internal
the internal energy of one system increases while the energy energy E of a gas depends ordinarily on its volume and tem-
of the other one decreases. The system which gains energy perature. In an ideal gas the molecules are sufficiently far
共that is, which absorbs positive heat兲 is called the initially apart that their mutual potential energy of interaction U con-
‘‘colder’’ system. The system which loses energy 共that is, tributes negligibly to their total energy. Hence U is unaf-
which absorbs negative heat or, equivalently, gives off posi- fected if the volume of the container is changed 共so that the
tive heat兲 is called the initially ‘‘warmer’’ system. average separation between molecules is changed兲. The in-
As pointed out in Sec. III E 3, a system’s absolute tem- tramolecular energies also do not depend on the separation
perature increases as its internal energy increases. Hence the between the molecules. Hence the internal energy E of an
system which gains energy 共the ‘‘colder’’ system兲 goes to a ideal gas does not depend on its volume, but only on its
higher absolute temperature while the system which loses temperature:
energy 共the ‘‘warmer’’ system兲 goes to a lower absolute tem- E⫽E 共 T 兲 independent of V. 共23兲
perature. Hence heat flows from the system at the higher
absolute temperature to the system at the lower absolute d. Ideal gas law. Familiar kinetic-theory arguments can
temperature.24 The absolute temperature provides, therefore, be used to calculate the pressure p produced by an ideal gas
information about the direction of heat flow occurring be- as a result of molecular collisions. Thus it can be shown that
tween two systems in thermal contact.
p⫽ 共 2/3兲 nK 1 , 共24兲
3. Mutual equilibrium among several systems where n is the number of molecules per unit volume and K 1
is the average center-of-mass kinetic energy of a molecule.
If a system X is in thermal equilibrium with another sys- Since Eq. 共21兲 implies that K 1 ⫽K/N⫽(3/2)kT, Eq. 共24兲
tem X ⬘ , then Eq. 共20兲 implies that T⫽T ⬘ . Similarly, if X is then yields the relation
in equilibrium with a third system X ⬙ , then Eq. 共20兲 implies
that T⫽T ⬙ . These results lead to the mathematical conclu- p⫽nkT. 共25兲
sion that T ⬘ ⫽T ⬙ . The corresponding physical conclusion can Because n⫽N/V, Eq. 共25兲 can also be expressed in the
be summarized by the following statement 共called the ‘‘ze- equivalent forms
roth law of thermodynamics’’兲: If two systems are each in
thermal equilibrium with a third system, they will be in ther- pV⫽NkT or pV⫽ ␯ RT, 共26兲
mal equilibrium with each other. where ␯ is the number of moles of the gas and the ‘‘gas
This statement indicates why any arbitrary thermometer constant’’ R⫽N A k 共where N A is Avogadro’s number兲. The
共such as a mercury thermometer兲 can be used to decide relation 共26兲 connecting the pressure, volume, and absolute
whether two systems will, or will not, be in equilibrium temperature is the ‘‘ideal gas law.’’
when brought into contact with each other. If the thermom-
eter readings are the same when the thermometer is applied
to the two systems, then the systems will be in equilibrium 2. Measurement of absolute temperature
when brought into contact with each other. But otherwise
they will not. a. Measurement of kT. The relation 共26兲 allows one to
measure kT, that is, the entropy gradient ␤ ⫽(kT) ⫺1 of an
G. Ideal gases and temperature measurement ideal gas. This measurement can be done by determining the
1. Properties of ideal gases pressure p of the gas, its volume V, and the number N of
molecules in it.
a. Energy of a monatomic ideal gas. In a monatomic Suppose that the gas is in equilibrium when in thermal
ideal gas of N molecules, the internal energy E of the gas is contact with some other system. According to Eq. 共20兲, the
just the total kinetic energy K. The basic entropy ␴ of the gas entropy gradient ␤ of the system is then the same as that of
depends on its energy E in the manner specified by Eq. 共9兲. the gas. Hence an ideal gas can be used as a thermometer to
Hence ␤ ⫽d ␴ /dE⫽3N/(2E) so that measure the entropy gradient ␤ of any other system.
E⫽K⫽ 共 3/2兲 N/ ␤ ⫽ 共 3/2兲 NkT. 共21兲 b. Kelvin temperature scale. It is somewhat difficult to
determine the number N of molecules in a gas. But the ratio
b. Energy of a polyatomic ideal gas. If a gas consists of of two absolute temperatures can be found without needing
N polyatomic molecules, the internal energy E of the gas to know this number. Indeed, suppose that one uses a

1057 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1057
‘‘constant-volume gas thermometer’’ where the volume of 共4兲 Thermal interaction and calorimetry.
the gas is kept fixed. Then Eq. 共26兲 implies that Energy conservation for two interacting systems:
Q⫹Q ⬘ ⫽0.
T/T ⬘ ⫽p/ p ⬘ , 共27兲
Determination of temperature from known heat capaci-
that is, the ratio of two absolute temperatures is equal to the ties. Experimental comparison of heat capacities.
readily measurable ratio of the corresponding pressures of 共5兲 Entropy determination from heat capacity.
the gas.
Hence, all absolute temperatures can be easily determined dS⫽dE/T⫽d⬘Q/T⫽CdT/T
by comparison with some easily reproducible standard tem-
perature. By international convention, this standard tempera-
ture is chosen to be the one where water is at its ‘‘triple

SB⫺SA⫽ CdT/T⫽C ln共TB /TA兲 if C⫽constant.
共6兲 Isothermal and adiabatic processes of an ideal gas.
point’’ 共that is, in the situation where water coexists in equi- pV⫽constant and pV ␥ ⫽constant 共where ␥ ⫽c p /c v 兲.
librium in its liquid, solid, and gaseous forms兲. Furthermore,
this convention defines the absolute temperature T t of water
at the triple point to be exactly I. General thermodynamic interaction
T t ⫽273.16 kelvin 共28兲 The preceding sections discussed thermal interactions that
involve no macroscopic work. Let us now examine more
by introducing the unit ‘‘kelvin’’ 共abbreviated by ‘‘K’’兲.
general interactions where such work may also be done.
This particular numerical value was chosen so that this tem-
perature scale agrees closely with earlier temperature scales. 1. Entropy change in a quasi-static process
All other temperature scales 共like the Celsius or Fahrenheit
scales兲 are nowadays defined in terms of the kelvin tempera- a. Entropy change due to bodily motion. Suppose that a
ture scale. solid object is slowly moved 共for example, that a weight is
c. Determination of Boltzmann’s constant. The value of lifted兲 so that macroscopic work may be done on it. Then the
Boltzmann’s constant k can be found by combining a mea- system’s internal motions 共relative to its center of mass兲 are
surement of ␤ ⫽(kT) ⫺1 with a measurement of the absolute unaffected and the system’s entropy remains correspond-
temperature T expressed in terms of the unit kelvin. Thus one ingly unchanged.
finds that 共approximately兲 b. Quasi-static adiabatic process. Consider now any sys-
tem 共for example, a gas兲 whose volume can change so that
k⫽1.38⫻10⫺23 J/K. 共29兲 macroscopic work can be done on it. For example, imagine
Correspondingly, the gas constant R⫽N A k has the approxi- that the system is in a container closed by a movable piston
mate value 8.31 J/共K mol兲. which can freely move up or down, but that the system is
d. Importance of the absolute temperature. Determina- thermally insulated so that it does not absorb any heat from
tions of the absolute temperature 共for example, by means of the piston or its other surroundings. Then the entropy S * of
an ideal-gas thermometer兲 are important for the following the entire isolated system is S * ⫽S⫹S ⬘ , where S is the en-
reasons. 共a兲 As pointed out in Sec. III F 2, the absolute tem- tropy of the system itself and S ⬘ is the entropy of the piston.
perature is directly related to the direction of heat flow be- If this system is in equilibrium, it is 共except for negligibly
tween systems in thermal contact. 共b兲 A measurement of the small fluctuations兲 in its most probable situation where the
absolute temperature of a system provides significant atomic entropy S * is maximum and the piston is at rest so that the
information about how the number of its basic states changes system’s volume remains constant. Now suppose that the
with its energy. 共c兲 Most properties of macroscopic systems system’s volume is changed by a small amount 共for example,
are temperature dependent and theoretical calculations of that the piston is slightly moved and thus does some work on
such properties are always expressed in terms of the absolute the system兲. If this process is carried out so slowly that the
temperature. Hence measurements of the absolute tempera- system remains extremely close to equilibrium 共that is, if the
ture are essential to use or verify theoretical predictions.25 process is ‘‘quasi-static’’兲, the entropy S * remains extremely
close to its maximum so that
H. Heat capacity dS * ⫽dS⫹dS ⬘ ⫽0. 共30兲
The discussion of heat capacity can be similar to that in a
traditional macroscopic approach to thermodynamics 共and But the entropy S ⬘ of the piston does not change as a result
can let students deal with similar kinds of applications and of its bodily motion. Hence
problems兲. The following merely lists the topics that may dS⫽0. 共31兲
usefully be discussed.
The entropy of a thermally insulated system remains, there-
共1兲 Definition of heat capacity C⫽d ⬘ Q/dT and of specific fore, unchanged in a quasi-static process 共even if work is
heats.26 done on the system兲.

Q⫽ C dT⫽C共TB⫺TA兲 if C⫽constant.
c. General quasi-static process. A more general infini-
tesimal quasi-static process 共where work d ⬘ W is done on a
Definition of heat reservoir 共C very large, T constant兲. system and heat d ⬘ Q is also absorbed by it兲 can be decom-
共2兲 Heat capacity at constant volume: C v ⫽dE/dT. posed into two successive quasi-static processes. 共a兲 A pro-
For a monatomic ideal gas, c v ⫽(3/2)R per mol. cess where no heat is absorbed, but work d ⬘ W is done. By
共3兲 Heat capacity at constant pressure. Eq. 共31兲 the system’s resulting entropy change is then dS
Relation between molar specific heats of an ideal gas: ⫽0. 共b兲 A process where no work is done on the system, but
c p ⫺c v ⫽R. heat d ⬘ Q is absorbed. By 共13兲 the resulting entropy change

1058 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1058
of the system at the absolute temperature T is then dS infer the macroscopic properties of specific systems 共for ex-
⫽d ⬘ Q/T. The system’s net entropy change is the sum of ample, the pressure or heat capacity of an ideal gas兲. Further-
these two entropy changes. Hence in any quasistatic infini- more, we have been able to derive the following general
tesimal process, principles applicable to all macroscopic systems and com-
pletely independent of any detailed knowledge about their
dS⫽d ⬘ Q/T. 共32兲 atomic constituents. 共These principles, called ‘‘laws of ther-
d. Measurement of entropy change. Suppose that a sys- modynamics,’’ were actually discovered in the nineteenth
tem is brought from a situation A to another situation B by century before the atomic nature of matter was well under-
means of a quasi-static process passing through a sequence stood.兲
of slightly differing successive equilibrium situations. Then a. Zeroth law of thermodynamics. This law summarizes
Eq. 共32兲 implies that the result obtained in Sec. III F 3: If two systems are each in
thermal equilibrium with a third system, they will be in ther-
S B ⫺S A ⫽ 冕
A
B
d ⬘ Q/T. 共33兲
mal equilibrium with each other.
b. First law of thermodynamics. This is the thermody-
namic energy law 共3兲,
Hence, the difference between the entropies of any two situ-
ations A and B can be determined by using any quasi-static ⌬E⫽W⫹Q. 共40兲
process connecting these situations. One only needs to mea- c. Second law of thermodynamics. This law specifies that
sure the successive small heats d ⬘ Q absorbed at correspond- the entropy of an isolated system tends to increase as the
ing absolute temperatures T.27 system approaches its most probable equilibrium situation. It
also specifies, according to 共32兲, how the entropy changes in
2. Implications of the entropy–heat relation any quasi-static process. Thus the law can be summarized by
the following relations:
a. Entropy related to system parameters. Relation 共32兲
For any isolated system,
can be combined with the thermodynamic energy law 共3兲 to
yield ⌬S⭓0. 共41兲
dS⫽ 共 dE⫺d ⬘ W 兲 /T⫽ 共 dE⫹pdV 兲 /T. 共34兲 For any infinitesimal quasi-static process,
Useful information can, therefore, be obtained if one knows dS⫽d ⬘ Q/T. 共42兲
how the entropy depends on the macroscopic parameters de-
scribing a system. To be specific, Eq. 共34兲 implies that28 d. Statistical significance of the entropy. The probability
P A , that a system in equilibrium is found in a situation A, is
if V⫽constant, dS/dE⫽1/T, 共35兲 related to its entropy S A according to 共5兲. Thus
if E⫽constant, dS/dV⫽p/T. 共36兲 P A ⬀exp共 S A /k 兲 . 共43兲
b. Applications to an ideal gas. The entropy S of an ideal Although this principle transcends the classical laws of ther-
gas consists of two parts so that modynamics, it is also completely macroscopic and indepen-
S⫽S V ⫹S E . 共37兲 dent of any specific knowledge about atoms.
e. Entropy and specific atomic knowledge. The relation
The part S V , due to the possible positions of the molecules
S⫽k ln ⍀, specified in Eq. 共11兲, connects the entropy to
throughout the volume V of the gas, depends on the volume
atomic knowledge about the basic states of any specific sys-
V according to Eq. 共6兲 so that tem. Hence this relation allows one to calculate the macro-
S V ⫽constant⫹kN ln V. 共38兲 scopic properties of a system from a detailed knowledge of
its atomic constituents.
The remaining part S E of the entropy depends on the internal
energy E 共that is, on the possible center-of-mass velocities of J. Some important applications
these molecules and on any intramolecular motions兲, but
does not depend on the volume V. 1. Heat engines and biological systems
Relation 共35兲 applied to the entropy 共37兲 of an ideal gas a. Converting randomness into order. The entropy of an
yields dS E /dE⫽1/T. Because the left-hand side does not isolated system increases as the system tends to approach a
depend on the volume, this relation implies that the internal more random situation. How then can one obtain a less ran-
energy E of an ideal gas does not depend on its volume, but dom situation 共for example, convert the random energy of
only on its absolute temperature. Thus one regains the pre- the molecules of gasoline into the nonrandom energy of for-
viously obtained result 共23兲. ward motion of a car兲?
Relation 共36兲, applied to 共37兲 and 共38兲, yields This conversion can be done if the system X is not iso-
kN/V⫽p/T or pV⫽NkT. 共39兲 lated, but interacts with its environment X ⬘ . The entropy
change of the total isolated system
Thus one regains the ideal gas law of 共26兲, but in a very
simple way that does not require examining how the pressure ⌬S * ⫽⌬S⫹⌬S ⬘ 共44兲
is produced by molecular collisions. can then be positive even if X becomes less random 共so that
its entropy S decreases and ⌬S is negative兲. This change is
3. Summary of statistical thermodynamics possible if the entropy S ⬘ of the environment increases by
By starting from the knowledge that any macroscopic sys- more than a compensating amount 共that is, if the entropy
tem consists of many atomic particles, we have been able to change ⌬S ⬘ is positive and sufficiently large兲.

1059 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1059
b. Heat engines. It would be very useful to have an en- 共c兲 At some particular intermediate temperature, the two ef-
gine that could absorb some positive heat q from a heat res- fects cancel so that melting results in no net change in the
ervoir X at some temperature T and convert this heat into an total entropy S * . At this temperature, ice and liquid water
equivalent amount of macroscopic work done on some ex- can, therefore, coexist in equilibrium.30
ternal system. But this is impossible because the entropy of
the reservoir 共and thus of the entire system兲 would then de- 3. Boltzmann distribution and molecular speeds
crease.
Suppose, however, that the engine converts only a part of What is the probability P r that a system X is in any one
the absorbed heat q into work w and rejects the remaining particular basic state r, of energy E r , when it is in thermal
heat q ⬘ ⫽q⫺w into some other heat reservoir X ⬘ . If the tem- contact with a large heat reservoir X ⬘ at some temperature
perature T ⬘ of X ⬘ is sufficiently low, the entropy S ⬘ of X ⬘ T ⬘ ? The number of basic states available to the entire iso-
can then increase by such a large amount that the entropy S * lated system X * 共consisting of X and X ⬘ 兲 is the number of
of the entire system does not decrease. basic states ⍀ ⬘ (E ⬘ ) available to the reservoir X ⬘ when its
The requirement that S * does not decrease implies that energy E ⬘ ⫽E * ⫺E r 共where E * is the constant total energy
⌬S * ⫽⌬S⫹⌬S ⬘ ⫽ 共 ⫺q/T 兲 ⫹ 共 q⫺w 兲 /T ⬘ ⭓0 共45兲 of X * 兲. According to Eq. 共5兲, the probability P r is then
or P r ⬀⍀ ⬘ 共 E * ⫺E r 兲 . 共48兲
w/q⭐ 共 T⫺T ⬘ 兲 /T. 共46兲 If X is in a state r of higher energy, the energy E ⬘ of the
reservoir 共and hence the number of basic states available to
The fraction of work obtainable from the engine 共its ‘‘effi- it兲 is correspondingly smaller. Hence the probability that X is
ciency’’兲 is, therefore, limited by the absolute temperatures in a state of higher energy is smaller than the probability that
of the reservoirs.29 it is in a state of lower energy.
c. Biological systems. Similarly, biological systems can This conclusion can be expressed quantitatively by consid-
grow 共and thus become more orderly兲 only by increasing the
ering the logarithm of 共48兲 and remembering that E r is much
randomness of their environment by more than a compensat-
ing amount. For example, an isolated kitten in a box would smaller than the total energy E * . Thus, to lowest order,
quickly die so that its body would decompose. However, if ln P r ⫽constant⫹ ␴ ⬘ 共 E * ⫺E r 兲
the box also contains food, then the kitten could grow into a
cat, but the food in the kitten’s environment would be con- ⫽constant⫹ 关 ␴ ⬘ 共 E * 兲 ⫺ ␤ ⬘ E r 兴 , 共49兲
verted into excrement. where ␤ ⬘ ⫽d ␴ ⬘ /dE ⬘ is the entropy gradient of the reservoir
Biological processes require individual molecules to be X ⬘ . Hence Eq. 共49兲 implies the following result 共known as
assembled into more orderly polymers 共for example, amino
the ‘‘Boltzmann distribution’’兲:
acids into proteins兲. Again this assembly can be done only by
coupled chemical reactions in which other molecules 共such P r ⬀exp共 ⫺ ␤ ⬘ E r 兲 or P r ⬀exp共 ⫺E r /kT ⬘ 兲 , 共50兲
as ATP兲 are transformed into more random forms.
where T ⬘ is the absolute temperature of the reservoir.
Application to molecular speeds in a gas. A single mol-
2. Phase transformations ecule in a gas can be considered as a system in thermal
contact with a heat reservoir consisting of all the other gas
Why is it that solids change into liquids, or liquids into molecules and the environment of the gas 共all at some abso-
gases, at certain well-defined temperatures? For example, if
lute temperature T ⬘ 兲. Hence the probability P(v) that such a
some ice turns into water, its entropy S increases because the
molecule, of mass m, is in a particular basic state where its
water molecules move about in much more random ways.
digitized velocity is v 共and its corresponding kinetic energy
However, the internal energy E of these molecules corre-
is m v 2 /2兲 is
spondingly also increases so that heat Q⫽⌬E must be ab-
sorbed from the environment 共which is a large heat reservoir P 共 v兲 ⬀exp共 ⫺m v 2 /2kT ⬘ 兲 . 共51兲
at some temperature T ⬘ 兲. Hence, the energy E ⬘ of the envi- The probability that such a molecule has a speed in the range
ronment must decrease by an amount ⌬E ⬘ ⫽⫺Q and its en- between v and v ⫹d v can then be found by multiplying 共51兲
tropy S ⬘ must correspondingly decrease by an amount ⌬S ⬘ by 4 ␲ v 2 d v 共because this factor is proportional to the num-
⫽⌬E ⬘ /T ⬘ ⫽(⫺Q)/T ⬘ . ber of velocities in this range兲.
If the ice is to melt, the entropy S * of the entire isolated
system 共ice and environment兲 must increase or remain un-
changed so that IV. CONCLUDING REMARKS

⌬S * ⫽⌬S⫹⌬S ⬘ ⫽⌬S⫺Q/T ⬘ ⭓0. 共47兲 The preceding sections have outlined an approach
whereby thermal physics can be taught from an atomistic
共a兲 If the temperature T ⬘ is sufficiently low, melting of the point of view, but in a way that is sufficiently simple to be
ice would lead to an increase ⌬S of its entropy, but would suitable for students in an introductory calculus-based phys-
lead to a more than compensating decrease ⌬S ⬘ ⫽⫺Q/T ⬘ of ics course.
the entropy of the environment. Hence it is overwhelmingly
more probable that the ice remains in its frozen state. 共b兲 At A. Integration of microscopic and macroscopic
sufficiently high temperatures, the heat Q required for melt- considerations
ing results in a much smaller entropy decrease ⌬S ⬘ of the The approach starts from the knowledge that all matter
environment. Hence the total entropy S * is increased if the consists of atomic particles. Using some elementary ideas of
ice melts, so that this occurs with overwhelming probability. mechanics and some primitive probability considerations, it

1060 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1060
provides a seamless integration of microscopic and macro- tempted to transform my own detailed instructional materials
scopic considerations while preserving their distinctive vir- into a textbook, but did not do so because of uncertainty as to
tues. whether there would be sufficient interest to make the effort
共1兲 The approach yields the classical laws of thermody- cost-effective.
namics in their full generality 共see Sec. III I 3兲 and shows
how these laws can be used to make purely macroscopic
inferences independent of specific atomic models 共for ex- ACKNOWLEDGMENTS
ample, to infer relations between specific heats or to discuss
heat engines兲. Furthermore, it transcends the classical laws I am indebted to Jill H. Larkin for many useful discussions
because it also shows how the probabilities of macroscopic and to Leonardo Hsu for some helpful comments.
situations can be related to their corresponding entropies.
共2兲 It provides valuable physical insights about atomic a兲
Electronic mail: freif⫹@andrew.cmu.edu
mechanisms underlying observable macroscopic concepts
1
F. Reif, Fundamentals of Statistical and Thermal Physics 共McGraw–Hill,
New York, 1965兲.
and phenomena 共such as internal energy, heat, entropy, ab- 2
F. Reif, Statistical Physics, Berkeley Physics Course, Vol. 5 共McGraw–
solute temperature, the approach to equilibrium, irreversibil- Hill, New York, 1967兲.
ity, and fluctuations in equilibrium兲. 3
D. Halliday and R. Resnick, Physics 共Wiley, New York, 1978兲, 3rd ed.
共3兲 It enables one to infer the properties of macroscopic 4
There are, of course, other ways that thermal physics can be taught from a
systems from specific knowledge about their atomic constitu- microscopic point of view in an introductory physics course. For example,
ents 共for example, to predict the pressure or heat capacity of another such recent attempt is described in T. A. Moore and D. V.
an ideal gas兲. It also allows one to predict the behavior of Schroeder, ‘‘A different approach to introducing statistical mechanics,’’
Am. J. Phys. 65, 26–36 共1997兲. Also T. A. Moore, Six Ideas That Shaped
atomic particles 共for example, the velocity distribution of Physics (Unit T: Some Processes are Irreversible) 共McGraw–Hill, New
molecules in a gas兲. York, 1997兲.
Hence, the suggested approach has many advantages 5
It is wise to avoid use of the word ‘‘microscopic’’ because students are
while not omitting any significant physics ordinarily dis- likely to confuse it with the similarly sounding word ‘‘macroscopic.’’
6
cussed from a traditional macroscopic point of view. Negative work done on the system corresponds to positive work done by
the system. Similarly, negative heat absorbed by the system corresponds
to positive heat given off by the system.
B. Limitation of the approach 7
The form 共3兲 of the law ensures consistency with mechanics where W is
used to denote the work done on a system. It also clearly exhibits the
In its attempt to keep the discussion readily accessible to similarity between the work done on a system and the heat absorbed by it.
first-year college students, the approach has eschewed 共These advantages are not shared by a historical convention that uses W to
quantum-mechanical ideas that would not yet be familiar to denote the work done by a system.兲
8
such students. This limitation could easily be overcome if The analogy is due to H. B. Callen, Thermodynamics and an Introduction
students already know some quantum physics, but is not of to Thermostatistics 共Wiley, New York, 1985兲, 2nd ed.
9
major importance in an introduction to thermal physics. In- Negative flow into the lake corresponds to positive flow out of the lake.
Similarly, negative condensation into the lake corresponds to positive
deed, practically all the reasoning processes and results evaporation from the lake.
would remain unchanged in a more correct quantum- 10
One can check that condensation has been eliminated by also eliminating
mechanical treatment. Hence the suggested approach can the flow and checking that the water level then remains unchanged.
provide students with a very good basis for further learning 11
One can check that heat transfer has been eliminated by also eliminating
after they have become familiar with quantum mechanics. all macroscopic work and checking that the thermometer indication then
The major modifications entailed by the more correct remains unchanged.
12
The work W in Eq. 共3兲 has been defined 共as in mechanics兲 as work done on
quantum-mechanical treatment would be the following. 共a兲
a system. Hence the infinitesimal work d ⬘ W done on a system by a pres-
Basic states would be well-defined discrete quantum states sure p in a small volume change dV is ⫺pdV 共that is, it is negative if the
共rather than digitized states specified by position and veloc- volume increases兲.
ity兲. 共b兲 Correspondingly, the entropy would be unambigu- 13
All the preceding basic ideas can be illustrated by giving students some
ously defined and would approach zero when T→0 共so that simple problems dealing with tossed coins or dice. The aim should merely
a system approaches its quantum-mechanical ground state兲. be to let students exemplify the definition of probability in some very
simple cases. However, there is no need to introduce students to more
Thus one would also obtain the third law of thermodynam-
complex knowledge about probabilities.
ics. 共c兲 Last, the proper quantum-mechanical description of 14
The term ‘‘basic state,’’ rather than ‘‘microstate,’’ is used here to avoid
atomic systems leads to more correct predictions of macro- confusion with the similarly sounding word ‘‘macrostate.’’
scopic properties 共for example, of heat capacities兲, especially 15
It is advisable to warn students that this specification of a basic state,
at lower absolute temperatures. although adequate for present purposes, must ultimately be replaced by a
more correct quantum-mechanical specification.
16
There is no need to distinguish this specification in terms of velocities
C. Pedagogical issues from the more formal specification in terms of momenta.
17
My experience indicates that the suggested approach is no Many of today’s students already know the utility of digitizing informa-
tion. For example, computers use such digitizing to specify possible posi-
more difficult 共and in some respects even easier兲 than a tions on a display by discrete positions 共‘‘pixels’’兲 or to specify possible
largely macroscopic treatment of thermal physics. This ap- sound intensities by discrete possible values.
proach can, therefore, be used to teach thermal physics in a 18
Such graphs are sufficient to show what happens when the number N of
five- or six-week period of an otherwise traditional calculus- molecules becomes larger. Thus there is no real need to teach students the
based introductory physics course. It might also be useful to binomial distribution.
students in introductory physical chemistry courses.
19
Instead of observing fluctuations in an assembly of N systems, one may
observe a single system at a large number N of successive instants 共for
The suggested atomistic approach outlined in the preced- example, by taking movie pictures every second兲. In a time-independent
ing pages can be readily expanded into more complete in- equilibrium situation the resultant pictures then constitute a statistical as-
structional materials 共expository text and problems兲 suitable sembly equivalent to N different systems observed at any one time.
for the introductory physics course. Indeed, I had been Equivalent probability statements can then be made about these pictures,

1061 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1061
that is, about the fluctuations in a single system observed in the course of thermal physics 共for example, problems examining applications of the
time. ideal gas laws, the temperature change caused by the free expansion of a
20
It is instructive to show students successive computer-generated movie gas, the dependence of molecular speeds on temperature or masses of the
frames 共or actual running movies兲 displaying simulations of gas molecules molecules, etc.兲.
spreading out throughout a box 共or the reverse process where they con- 26
The notation d ⬘ Q 共with a prime ornamenting the d兲 is here used to denote
centrate themselves into one half of the box兲. In particular, this reverse an infinitesimal quantity which is not necessarily a small difference.
process looks strikingly different for 40 molecules or for 4 molecules. 27
It is instructive to let students contrast a sudden adiabatic expansion of a
关Such movie frames are shown on pp. 22–25 of volume 5 of the Berkeley gas 共where its entropy increases兲 with a quasi-static adiabatic expansion
Physics Course 共see the reference in Note 2兲.兴 共where this entropy remains unchanged兲. It can be pointed out that the
21
The absolute temperature can be negative if one considers special systems latter case involves two competing effects: 共a兲 an increase of the entropy
that have no kinetic energy. For example, as the internal energy of a
because of the increased gas volume and 共b兲 a decrease of the entropy
system of spins in a magnetic field increases, the system’s entropy first
because of the decreased internal energy 共caused by the negative work
increases and then decreases. In this case ␤, as well as T, can be negative.
22 done on the gas during its slow expansion兲.
Some examples or exercises with small numbers, describing unrealistically 28
There is no need to express these relations in terms of partial derivatives
small systems, can help to clarify these considerations.
23
For very simple systems, such as ideal gases, Eq. 共20兲 is equivalent to the 关for example, to write ( ⳵ S/ ⳵ E) V ⫽1/T兴 because beginning students may
condition that the average energy per particle is the same in each system. still be unfamiliar with this notation.
29
24
More generally, heat flows from the system with lower ␤ to the system Similar entropy arguments apply to refrigerators designed to transfer heat
with higher ␤. This is true even in the case of a system 共of the kind from a reservoir at a lower absolute temperature to another one at a higher
mentioned in Ref. 21兲 where ␤ or T can be negative. Unlike ␤, the abso- temperature.
30
lute temperature T is then an inconvenient concept because a negative To deal also with pressure effects, one may consider a system in contact
absolute temperature describes a very warm system whose temperature with an environment at constant temperature T ⬘ and pressure p ⬘ so that
goes through infinity as the system cools down. ⌬S * ⫽⌬S⫺(⌬E⫹p ⬘ ⌬V)/T ⬘ . In this case phase transformations may
25
The topics discussed in Sec. III G can be accompanied by many of the again be discussed by simple entropy considerations without resorting to
exercises and problems commonly used in conventional approaches to the Gibbs free energy.

CHRISTMAS DINNER
At Christmas we were invited to Lawrence’s home for dinner; oddly enough, he took the
opportunity of telling me on this occasion that the lab was short of money, and that if necessary
he would ask me to return part of my salary.
In 1939, $300 a month was a good salary, and it got me out of the woods for some time, but
after six months, in July 1939, Lawrence, who by then must have realized my situation, asked me
if I could return to Palermo. I answered by telling him the truth, and he immediately interjected:
‘‘But then why should I pay you $300 per month? From now on I will give you $116.’’

Emilio Segrè, A Mind Always in Motion—The Autobiography of Emilio Segrè 共University of California Press, Berkeley,
1993兲, p. 147.

1062 Am. J. Phys., Vol. 67, No. 12, December 1999 Frederick Reif 1062

S-ar putea să vă placă și