Introducing thermodynamics through energy and entropy

Rodrigo de Abreu and Vasco Guerra

Departamento de Fı́sica, Instituto Superior Técnico, Universidade Técnica de Lisboa, 1049–001 Lisboa,
Portugal
(Received 29 September 2011; accepted 12 March 2012)
We suggest a simple approach to introducing thermodynamics, beginning with the concept of
internal energy of deformable bodies. From a series of thought experiments involving ideal gases,
we show that the internal energy depends on the volume and on a second parameter, leading to the
development of the concept of entropy. By introducing entropy before the notions of temperature
and heat, the proposed approach avoids some of the major conceptual difficulties with the
traditional presentation. The relationship between mechanics and thermodynamics naturally
emerges, mechanics corresponding to isentropic thermodynamics. The questions of evolution to
equilibrium and irreversibility are studied under the light of the action of the “dynamic force” and
its dissipative character, evincing the benefits of keeping in mind the microscopic picture. VC 2012
American Association of Physics Teachers.
[http://dx.doi.org/10.1119/1.3698160]

I. INTRODUCTION Sec. IV, from generalizations to the base model. That section
also discusses the adiabatic piston as an example to illustrate
Thermodynamics is one of the key disciplines of physics the advantages of the proposed approach and discusses some
and chemistry, yet in its traditional form it continues to elicit further extensions of the model. The main findings are
intellectual resistance. Despite being well established for reviewed and discussed in Sec. V.
many years and dealing with apparently simple and intuitive
concepts, thermodynamics remains quite difficult and subtle.
We believe the difficulties with thermodynamics arise essen- II. WORK AND ENERGY: MECHANICS AND
tially from two main factors. The first one is the traditional de- THERMODYNAMICS
velopment using the variable temperature. As is well known,
the natural variables associated with the internal energy U are The concepts of work and energy were developed empiri-
the volume and the entropy. We can then Legendre-transform cally, associated first with the notions of force, gravitational
U into the other thermodynamic potentials that flank it on the potential energy, and kinetic energy. The complexity of the
“thermodynamic square,”1 and obtain temperature from these action of muscles was replaced with a simplified description
potentials via Maxwell’s relations. A formulation of thermody- of a static force, which can be measured, for instance, by a
namics that starts with U ¼ U(S, V) contributes to a neat mathe- dynamometer. This led to an understanding of weight as a
matical derivation and an easier understanding of the meaning force, and of the corresponding opposing force present in
of the different symbols. An early introduction of the entropy equilibrium. The concepts of work of the weight and work of
was also proposed by Callen.2 the force opposing weight appear naturally from here. The
A second difficulty comes from the fact that thermody- idea of gravitational potential energy arises from the design
namic quantities such as temperature and pressure are and construction of weight-lifting machines.
defined in equilibrium. The application of the formalism to These intuitive concepts are a fundamental part of any in-
non-equilibrium processes makes several variables and troductory study of physics. Among many excellent text-
notions lose their intuitive physical meaning and the mere books, we would like to recall the brilliant presentation by
use of the words denoting these variables may induce errors. Feynman,3 who introduces mechanics following the route
The difference between the static and dynamic pressures is just described. Another hypothesis, which he immediately
crucial and is thoroughly discussed in this paper. advances and was early noted by da Vinci,4 is that “there is
Herein, we suggest a way to avoid most of the traps fre- no such thing as perpetual motion with weight lifting
quently encountered in the presentation of thermodynamics. machines.” Feynman goes on to warn that “in fact, that there
To this purpose, we try to maintain a natural and strong con- is no perpetual motion at all is a general statement of the law
nection with mechanics and the microscopic view. We start of conservation of energy.”
by making the link between mechanical energy and the inter- Newton’s laws allow us to identify the increase in the ki-
nal energy of deformable bodies in Sec. II. In Sec. III, we use netic energy as the work done by the resultant or net force
a simple “base model” involving an ideal gas to show that the applied to a particle. Combining the concepts of gravitational
internal energy cannot be a function only of the deformation potential energy and kinetic energy yields the principle of
variable. However, unlike the conventional presentation of conservation of mechanical energy for a particle moving
thermodynamics in which the variable temperature is a cor- only under the effect of the gravitational force.
nerstone, here it is entropy that is immediately introduced as The generalization of these ideas to extended deformable
the additional required variable. The relationship with bodies is at the origin of the notion of internal energy. A
mechanics is then direct: mechanics is isentropic thermody- very simple example is depicted in Fig. 1(a). A pair of forces
namics. The second law and the questions of evolution to of the same magnitude and opposite directions acts on a
equilibrium and irreversibility are also addressed in Sec. III. body, deforming it. Clearly, there is work done by the forces,
Heat and the first law of thermodynamics are introduced in but the translational kinetic energy and the gravitational

627 Am. J. Phys. 80 (7), July 2012 http://aapt.org/ajp C 2012 American Association of Physics Teachers
V 627
Fig. 1. A body deforms under the action of two forces of equal magnitude: (a) macroscopic picture; (b) microscopic picture, the body being an ideal gas in a
container with movable walls.

potential energy of the body remain unchanged. This heat. This approach allows a formulation of thermodynamics
straightforward example also illustrates the power of the mi- that uses from the beginning the natural variables associated
croscopic view. The simplest microscopic picture is to imag- with the internal energy, which are the volume and the en-
ine the body as an ideal gas in a container with movable tropy. Moreover, all remaining quantities are introduced in a
walls, Fig. 1(b). In this case, the work done is converted into general and clear way, which, we believe, helps to prevent
kinetic energy of the constituents of the gas, which is the the misunderstanding and errors that spread even in the sci-
same as its internal energy. entific literature. Finally, this procedure immediately pro-
With further maturation, this extension of the concepts of vides an easily identifiable connection between mechanics
mechanics to deformable bodies is at the genesis of thermody- and thermodynamics.
namics. Historically, however, thermodynamics evolved in a
rather autonomous way. Its development was based on the
concepts of heat and temperature, with properties apparently A. The need for the variable entropy
foreign to the phenomenology described by mechanics. These
two new concepts were not easily encompassed by the formal- Thermodynamics can be introduced with generality from
ism of mechanics, although there have always been “kinetic” a clear, uncomplicated, “base model,” corresponding to one
conceptualizations of the thermodynamic quantities.5 of the most typical systems in thermodynamics: a classical
The application of Newtonian physics to engineering, mili- ideal gas in a container with a movable piston on top, under
tary engineering being of particular importance, forced mechan- the action of gravity, as shown in Fig. 2. The system is sur-
ics to face the question of friction, which is always present in rounded by vacuum, so that there is no external atmospheric
practice. Friction was—and still is—treated as a nonessential pressure on the piston. There is no friction between the pis-
part of mechanics. One recurrent statement revealing the ton and the container walls. Furthermore, it is assumed that
attempt to isolate mechanics from the phenomenology of fric- the piston and the container walls do not have any internal
tion is the claim that the equations of mechanics are reversible. structure, so that all collisions between the gas particles and
Friction brings to light that mechanics and thermodynam- the piston or the containing walls are perfectly elastic. This
ics are one and the same subject. The works of Maxwell and system can be studied starting only with the notions of
Boltzmann culminate an amazing effort of many authors to
reconcile mechanics with thermodynamics, addressing the
question of evolution to equilibrium. Boltzmann, in particu-
lar, proposing a microscopic interpretation of entropy,6,7
opens the door for a reevaluation of the meaning of several
concepts, still problematic in the basic formulation of ther-
modynamics today.8,9 Let us pass through the door and travel
this road in the remainder of this paper.

III. ENERGY AND ENTROPY: THE BASE MODEL

Thermodynamics has been established for a long time and
there is a relatively consensual view on the way to present it.
Most concepts, such as work, temperature, and heat, are
introduced from the very beginning and in a very intuitive
way. Nevertheless, various difficulties arise in the interpreta-
tion of fundamental quantities such as heat and work, even
in simple situations. The subtleness of these seemingly sim-
ple and unambiguous notions leads to a search for consis-
tency in the definitions of work and heat in the formulation
of the first and second laws of thermodynamics, still very
active today.10–20
We suggest an early introduction of the variable entropy, Fig. 2. The base model: An ideal gas is contained in a cylinder with a fric-
postponing the appearance of the quantities temperature and tionless movable piston of mass M on the top.

628 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 628
weight, work done by the weight, gravitational potential version of our base model, interesting as a model of thought
energy, and internal energy. Note that a similar system was and adequate for numeric simulations, is also suggested.
used by Kivelson and Oppenheim to discuss the concept of The final equilibrium position can be calculated from the
work in irreversible processes.21 principle of conservation of energy and the equality of the gas
Let us assume that the gas initially occupies a volume V1 and piston pressures. The former implies that the decrease in
and exerts a pressure P1 on the container walls. The piston is the gravitational potential energy must correspond to an
held fixed in its place. Its total mass, M, is such that the pres- increase in the internal energy,
sure exerted by the piston as soon as it is released,
 Mgðh2  h1 Þ ¼ U2  U1 ; (4)
Mg
Pe ¼ (1) where the indexes 1 and 2 refer to the initial and final states,
A
respectively, and h is the height of the piston, h ¼ V/A; the
is larger than P1 , A denoting the area of the base of the latter is the statement
piston.
The final equilibrium position of the piston can be deter- Mg U2
¼a ; (5)
mined if we know the dependence of the gas pressure on the A V2
gas internal energy. For a classical monoatomic ideal gas, where Eq. (2) was used. We have as well
from the microscopic picture and the calculation of the aver-
age momentum transfer per unit time, it is easy to show that P1 V1
this relation is U1 ¼ ; (6)
a
2U U so that
P¼ ¼a ; (2)  
3V V a P1 A
h2 ¼ h1 1þ : (7)
where the value a ¼ 2=3 is specific to a monatomic gas, but 1þa Mga
other values of a can be used for diatomic and polyatomic
gases, and even a relativistic gas of photons (a ¼ 1=3). As a If P1 < Pe , as we have assumed, then h2 < h1 , as it should
matter of fact, for a container of dimensions L  L  L, a be. Nevertheless, Eq. (7) is valid for any relation between
particle of mass m and speed vx moving along the x direction the initial gas pressures P1 and Pe .
hits one of the walls vx =2L times per unit time, and in each Let us now assume we remove a certain mass from the pis-
collision transfers a momentum 2mvx . Thus, the momentum ton (for instance, we can imagine that the piston has several
transfer per unit time to the wall is mv2x =L. If we have N par- weights on the top of it, and we can simply remove one of
ticles moving along the x direction and with different speeds, them). The new total mass of the piston is M0 < M and the
the average momentum transfer per unit time is piston will reach a new equilibrium position, which we iden-
hDp=Dti ¼ Nmhv2x i=L, where hv2x i is the average value of v2x . tify with the subscript 3. Let us look at the case where M0 is
Further assuming that there is no privileged direction of such that the new equilibrium position, h3 , is the same as the
motion, hv2 i ¼ hv2x þ v2y þ v2z i ¼ 3hv2x i. Finally, dividing by initial position h1 . The energy conservation between posi-
the area L2 to obtain the pressure, we obtain tions 2 and 3 reads, similarly to Eq. (4),
   M0 gðh1  h2 Þ ¼ U3  U2 : (8)
1 Dp 1N
P¼ 2 ¼ mhv2 i; (3)
L Dt 3V Therefore, combining this expression with Eq. (4), we arrive
at the following very interesting result:
which is Eq. (2) identifying the internal energy with the ki-
netic energy of the gas. Note that the potential energy of the U3  U1 ¼ ðM  M0 Þgðh1  h2 Þ > 0: (9)
gas particles is negligible compared to their kinetic energy.
Thus, when the piston goes back to its initial position, the in-
Furthermore, all the main results derived below do not
ternal energy of the gas is higher than it was initially. It is
depend on this approximation.
straightforward to show that this conclusion remains valid if
It is worth noting that, despite the absence of friction, the
initially we would have P1 > Pe and hence h2 > h1 and
piston does reach a final equilibrium position and does not
M0 > M.
remain oscillating indefinitely. Or, more precisely, it will at
Equation (9) means that the internal energy cannot be
first oscillate while slowly evolving to its equilibrium posi-
solely a function of the volume, in contrast to the typical sit-
tion and, once this position is reached, the piston stays jig-
uations of mechanics (where the potential energy is only a
gling around it. What happens is that the pressure exerted by
function of position). If we assume that one further variable
the gas on the piston is a “dynamic pressure.”17–20 For the
suffices to completely determine the internal energy, then we
same volume, the dynamic pressure is higher in a compres-
can write
sion and lower in an expansion.17–19 The piston moves under
the action of the weight and of the dynamic force. Because of U ¼ UðV; SÞ; (10)
the imbalance between the two forces, the latter has a
“dissipative character,” leading the piston to the final equilib- where the new variable, S, is called entropy.
rium position.18 The situation is somewhat similar to the case
of a bullet entering at high speed in a region where there is a B. Properties of entropy and the second law
gas at room temperature. Even if all collisions between the
gas particles and the bullet are perfectly elastic, the net effect Suppose now that the piston has initially a mass M0 such
is one of slowing down the bullet, until it finally stops. We that AP1 =g ¼ M0 , which means that if the piston is released
will return to this point in Sec. V, where a one-dimensional it stays at its equilibrium position. We can now add to the

629 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 629
piston a mass m ¼ ðM  M0 Þ=2 and wait for the piston to
reach its new equilibrium position. Finally, we add yet
another mass m to the piston, so that its total mass is M, as in
the compression described in Sec. III A. Using the same rea-
soning as above, it is not difficult to show that the final equi-
librium position corresponds to a position h02 < h2 . This,
again, shows that the internal energy cannot be uniquely a
function of the volume. Moreover, the lower height reached
by the piston reflects the smaller imbalance between the
dynamic force exerted by the gas and the weight, compared
to the previous case.
We can further imagine a process in which the mass of the
piston is increased in N equal steps from M0 to M, for
instance, by slowly adding sand grains, one by one, on the
top of the piston. At each step, the dynamic force exerted by
the gas is nearly the same as the weight. Therefore, the dissi-
pative character of the dynamic force almost does not mani- Fig. 4. Final equilibrium height of the piston after compression and expan-
fest itself. The piston reaches the lowest possible final sion, as a function of the number N of masses used both to increase the total
mass from M0 ¼ 10 kg to M ¼ 11 kg and then to decrease it back to M0
position corresponding to a total mass M. Figure 3 illustrates
(see text).
this effect, by showing the final height of the piston as a
function of the number N of masses m ¼ ðM  M0 Þ=N added
on the top of the piston. The calculations were carried out for happens if we do the expansion slowly with a large number
M0 ¼ 10 kg, M ¼ 11 kg, and a ¼ 2=3. The final height is of steps, but the compression with few steps.
plotted as a fraction of the initial height. The reader is invited Some properties of entropy can now be easily derived.
to verify that the final result is given by Referring to this simple case treated with the base model, we
N   have
h2 Y a 1 M0 þ ðk  1Þm
¼ þ ; (11) UðV; S3 Þ  UðV; S1 Þ: (12)
h1 k¼1 1 þ a 1 þ a M0 þ km
If we arbitrarily postulate that S3  S1 , then the entropy can
and that the limiting result is ðM0 =MÞ1=ð1þaÞ . For the case only increase or remain constant. Moreover,
depicted in Fig. 3, we have ð10=11Þ3=5 ’ 0:944418.  
Interestingly enough, if the sand grains are now slowly @U
> 0: (13)
removed, one by one, we have again a near equilibrium @S V
between the dynamic force and the weight at all times.
Everything nearly returns to the initial configuration and the Clearly, the situation with no entropy change corresponds to
gas will thus have very approximately the same internal a reversible transformation, i.e., to a transformation where
energy as in the beginning. Figure 4 shows the final height the initial state of the system can be recovered. In this case,
h3 as a function of masses used both during the compression the internal energy can be calculated only from the volume,
and the expansion, normalised to the initial height h1 . as in the typical examples of mechanics. Thus, mechanics
Lastly, note that if we make the compression with a large corresponds to isentropic thermodynamics, that is, to a set of
number of steps, but the expansion quickly in few steps, we transformations where the entropy does not change.
still find the need to add some additional mass to the piston It is worth underlining two issues evidenced by the previ-
to bring it to height h1 and U3 > U1 , as in Eq. (9). The same ous discussion and Figs. 3 and 4. The first one is that for de-
formable bodies (bodies with internal structure), the
isentropic transformations can only be performed by always
keeping the system nearly in equilibrium, when the dynamic
force is always approximately balanced by the “static”
weight force. Failing to do so, either in the compression or
the expansion, or both, leads to an increase of the entropy
and to the impossibility of recovering the initial state of the
system. The latter statement exposes the second matter: it is
not necessary to look at the complete sequence, compression
plus expansion, to speak about a reversible transformation,
as its reversible character, corresponding to constant entropy
during the process, can be ascribed separately to the com-
pression and to the expansion. In short, if a transformation
leads from one state to a different one with the same entropy,
then the transformation can be reversed, and vice-versa. The
process just described of slowly adding or subtracting sand
grains approaches the reversible ideal.
Fig. 3. Final equilibrium height of the piston after the compression, as a We have thus obtained the second law of thermodynamics,
function of the number N of masses used to increase the total mass from while keeping its interpretation and understanding at a very
M0 ¼ 10 kg to M ¼ 11 kg (see text). fundamental level. Evidently, in many situations studied in

630 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 630
mechanics, a variation of entropy does take place, although Similarly, we can define the quantity T from
usually it is not taken into account. An example would be the  
treatment of the compression depicted in Fig. 1(a), consider- @U
T¼ > 0; (22)
ing the energy only as a function of the deformation. A simi- @S V
lar case is the deformation of a spring and the application of
Hooke’s law. We will further discuss this point in Sec. V. where the inequality is simply Eq. (13). It is not difficult to
later identify T with the ideal gas temperature. This determi-
C. Introduction of the formalism nation has been made by other authors.2,22 Notice, however,
that, for an irreversible transformation, P and T are defined
As shown in Secs. III A and III B, the internal energy by Eqs. (21) and (22), respectively. In a dynamic situation,
depends (at least) on two variables, volume and entropy, as when the gas has a certain volume and a certain internal
given by Eq. (10). Therefore, we can write energy, P and T are the pressure and temperature it would
    have if it were in equilibrium, with the same volume and in-
@U @U ternal energy. This is the general meaning of P and T, and
dU ¼ dV þ dS: (14) no other. Furthermore, in a dynamic situation P is not the
@V S @S V
pressure exerted by the gas. Naming P and T “pressure” and
The idea of the increase in entropy as a result of the lack of “temperature” and thinking in physical terms in these quanti-
balance between the dynamic and the static forces, advanced ties, with these designations, as defined in equilibrium, is a
and discussed in Secs. III A and III B, can be confirmed as fol- common source of mistakes and misunderstandings.
lows. The work of the resultant of the forces on the piston is Finally, we can write
equal to the variation of its kinetic energy. In differential ~ ¼ PdV þ T dS
form, dU ¼ PdV (23)

Mg and
dEkin ¼  ~
dV þ PdV; (15)
A  ðP~  PÞdV ¼ T dS: (24)
where P~ is the dynamic pressure exerted by the gas on the
piston and dEkin is the variation of the kinetic energy of the This last equation establishes that the variation of entropy is
piston. On the other hand, the variation of the potential a consequence of the difference between the dynamic and
energy of the piston is the static pressures. This difference results in the “dissipative
character” of the force, even if there is no friction, as pointed
Mg out and discussed in Sec. III A. This expression allows an
dEpot ¼ dV: (16) additional verification that S always increases. As a matter of
A
fact, if dV > 0 it must be true that P~ < P, so that dS > 0.
Consequently, The same conclusion is obtained if dV < 0, as then P~ > P.
To finish this section, let us go back to Eq. (19), to note
~
dEkin þ dEpot ¼ PdV: (17) that
ð
In addition, by conservation of energy, ~
DU ¼  PdV; (25)
dEkin þ dEpot þ dU ¼ 0: (18)
which, denoting by W the work done by the dynamic force,
As a result,
ð
~
dU ¼ PdV; (19) ~
W ¼  PdV; (26)

so that the variation of the internal energy of the gas is equal

to the work done by the dynamic pressure.18–20 takes the expected form
Among the conclusions from Sec. III B, we have seen that
a reversible transformation corresponds to a constant value DU ¼ W: (27)
of the entropy, so that dS ¼ 0. What is more, the dynamic
pressure is the same as the static pressure in this case. Hence, Moreover, it is immediate to verify that between two points
as dU ¼ Pe dV and P ¼ Pe , with Pe given by Eq. (1), from where the piston is at rest the work done by the dynamic force
Eq. (14) we have W is equal to the work of the weight. Indeed, substituting
  Eq. (16) into Eq. (18), integrating and noting that in this case
@U DEkin ¼ 0,
dU ¼ Pe dV ¼ PdV ¼ dV: (20)
@V S ð ð
Mg
DU ¼  dV ¼  Pe dV
Thus, the quantity P can be defined from A
  Mg
@U ¼ DV ¼ DEpot ¼ W: (28)
P¼ ; (21) A
@V S
This consistency check does not constitute a surprise, as all
which corresponds to the gas pressure for an equilibrium our analysis of the base model started precisely from this
point. condition. The interesting fact is that, despite Pe 6¼ P~ along

631 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 631
the transformation, between two points where the piston is at
rest we nevertheless have
ð ð
~
Pe dV ¼ PdV: (29)

If the transformation is reversible, dS ¼ 0 along the trans-

formation and, if the gas returns to the initial volume, using
Eq. (23),
þ þ
~
DU ¼  PdV ¼  PdV ¼ 0: (30)

In general, for an irreversible transformation returning to the

same volume,
þ
DU ¼  PdV ~ > 0: (31)

Thus, the work of the dynamic force is transformed into in-

ternal energy, making the internal energy increase if the sys-
tem returns to the initial volume, in accordance with the
conclusions presented in Sec. III A.
In a generalization of the concept of heat,5 we can say that
the work has been transformed into heat. Heat, however, is a
subtle concept, still often misinterpreted (cf. Sec. IV C) and
it would be better to rather say that “work has been used to
heat the gas,” i.e., to increase its internal energy. Nonethe-
less, with the intuition acquired on the notion of variation of
entropy associated with the base model, linking Newton’s
second law with the second law of thermodynamics,18 we
are now ready to infer the properties of more complex
systems.
Fig. 5. Two subsystems, A and B, which can exchange energy through a
IV. HEAT: GENERALIZATIONS OF THE BASE common border. A frictionless piston of mass M can move and modify the
MODEL volume of subsystem A.

The natural generalization to a more elaborate configura-

tion is a system comprised of two subsystems with a com- The conclusions from our base model apply to the new sys-
mon boundary, through which they can exchange energy. tem (A þ B).
We will consider various configurations. The first one con- As the complete system is surrounded by vacuum, we
sists of one subsystem of fixed volume and a second subsys- have
tem with a movable piston on top, leading to the formulation
of the first law and the study of the heat reservoir. The sec- U ¼ UA þ UB ; (32)
ond arrangement is the case of two subsystems side by side
coupled by a moving piston, which allows the analysis of the where U is the total internal energy, and UA and UB are the
celebrated “adiabatic piston” problem.3,19,20,23–25 A third ge- internal energies of subsystems A and B, respectively.
ometry is the same as the previous one, but with the two We can now write [cf. Eqs. (25) and (26)]
subsystems arranged vertically. Finally, the last setup ð
involves two subsystems side by side and a piston on the top W ¼  PdV ~ ¼ DU ¼ DUA þ DUB : (33)
of each of them, in order to study the constant pressure
calorimeter.12,13
As before, between two points where the piston is at rest the
work of the dynamic force is equal to the work of the weight.
A. Heat and the first law Rearranging this equation gives
The first geometry we want to investigate includes a sub- DUA ¼ W  DUB ; (34)
system A similar to the base model, namely, a classical ideal
gas in a container with a moving piston. However, subsystem which we can put into a more familiar form by defining
A is in now contact with another subsystem B, of fixed vol-
ume, as shown in Fig. 5. The complete system (formed by Q ¼ DUB ; (35)
gases A and B and the piston) is isolated, but subsystems A
and B can exchange energy with the piston and therefore the to obtain
system (A þ B) is not isolated. Furthermore, subsystems A
and B can also change energy through the common border. DUA ¼ W þ Q: (36)

632 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 632
This is the usual formulation of the first law of thermody- also the same and the temperature of the heat reservoir
namics. (The symbol W is often defined with opposite sign, always remains unchanged. Equation (43) tells us that in this
i.e., as the work done by the system, so that DUA ¼ Q  W.) case a reversible transformation is an isothermal transforma-
The quantity Q is called heat or heat exchanged with subsys- tion at T ¼ TB . Moreover, since dSA ¼ dSB , using Eqs.
tem B. Its negative, –Q, is the change in the internal energy (40) and (35),
of subsystem B, or the energy transferred from A to B. The
first law is thus a particular form of writing the principle of dQ
dSA ¼ : (44)
conservation of energy. T
The introduction of heat in this simple but clear way gives
a valuable contribution to help avoiding some of the most
common traps related to this concept, as further discussed in C. The adiabatic piston
Sec. IV C. The so-called “adiabatic piston” problem concerns two ideal
gases contained in a horizontal cylinder and separated by an
insulating piston that moves without friction, as shown in Fig. 6.
B. Temperature and the heat reservoir This system is more complex than the previous ones and has
One interesting result that can be easily obtained is the been treated by many authors.3,19,20,23–27 Worth noting are the
equality of temperatures of subsystems A and B described in qualitative kinetic description by Feynman,3 the molecular dy-
Sec. IV A along reversible transformations. First note that, namics calculations by Mansour and co-workers,23,25 and the
from Eq. (32), classical thermodynamics analysis by Gislason.19
Classical thermodynamics analyses are of major interest
dU ¼ dUA þ dUB : (37) here. As pointed out in our previous work,20 a careful use of
thermodynamics must give the same final result as molecular
Because U is a function of S and V, while dS ¼ 0, we also dynamics, because the latter is a microscopic interpretation
have of the former. However, too commonly this is not the case, a
  fact that strikingly exemplifies the difficulties associated
@U with the formalism of thermodynamics. Gislason gives a
dU ¼ dS  PdV ¼ PA dVA : (38)
@S V very interesting and enlightening discussion of the prob-
lem,19 focusing on the shorter time scale, when the two gases
On the other hand, evolve to a situation of equal pressures. However, he does
not address the second phase, when the gases evolve to a sit-
dUA ¼ TA dSA  PA dVA ; (39) uation of equal temperatures, discussed qualitatively by
Feynman3 and formally derived in our previous work.20 On
and the other hand, Anacleto and Anacleto,26 just to give one
example, make a faulty investigation of the problem, claim-
dUB ¼ TB dSB ; (40) ing that the piston does not reach a final state of equilibrium,
where TA and TB are defined by Eq. (22). Combining these instead keeping oscillating indefinitely. Furthermore, they
expressions, we obtain allege that entropy remains constant, due to the absence of
friction, which is not the case.
TA dSA þ TB dSB ¼ 0: (41) The main difficulty with this problem arises from a negli-
gent use of language. As a matter of fact, the word “adiabatic”
Finally, assuming the entropy to be an extensive quantity (an is too swiftly associated with “no heat exchange” and is itself
assumption that can be motivated and discussed) and since problematic. Moreover, “heat exchange” is rarely defined
the transformation is reversible, with generality, but we immediately are led to impose the
mathematical condition dQ ¼ 0 in the calculations.
dS ¼ dSA þ dSB ¼ 0; (42) The correct solution within the framework of thermody-
namics was presented previously,20,24 and the reader should
and hence, using Eq. (41), refer to those papers for the details on the formal use of the
thermodynamic laws. Quoting from our former work,20 by
TA ¼ TB : (43) an “adiabatic piston” it is meant a piston with zero heat con-
ductivity. If the piston is held in place (for instance, if it is
This shows that temperature characterizes the state of equi-
librium of the two subsystems.
An important limiting case is where subsystem B is infi-
nite, so that it constitutes a heat reservoir. Energy exchange
with the smaller subsystem then does not change the temper-
ature of reservoir B. This intuitive notion can be verified as
follows. If the energy density is uniform, in each unit volume
i of the heat reservoir we have the same volume Vi and the
same internal energy Ui . Hence, since Ui is a function of Si
and Vi , all unit volumes have the same Si . As the system is
infinite, any finite transfer of energy will not change its
energy per unit volume. Therefore, Vi ; Ui , and Si are not
modified by a finite energy transfer to or from the heat reser- Fig. 6. Two subsystems, A and B, which can exchange energy through a
voir. All quantities being the same, the derivatives (22) are moving “adiabatic” and frictionless piston.

633 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 633
fixed to the box by screws), then there is no “heat transfer” Secs. III–IV C, it is possible to proceed to more complicated
from one subsystem to the other. Even though, if the piston and richer systems. This paper would become too lengthy if
is released, both systems exchange energy via collisions with we would discuss them in detail here. Nevertheless, we will
the piston, as they are coupled through the conditions of con- mention a few of them briefly.
stant total volume and total energy, where the kinetic energy
of the piston has to be taken into account. The evolution to a 1. The adiabatic piston in a gravitational field
stage of mechanical equilibrium of equal pressures has noth-
ing to do with friction: it is simply a result of the dissipative A natural generalization is to consider the adiabatic piston
character of the dynamic pressure, discussed in our base from Sec. IV C, but now in a vertical configuration and under
model. There is indeed an entropy increase, as also acknowl- the effect of gravity. In this case, the work done by the
edged by Gislason in his analysis of the first phase of the dynamic force has two terms, one for the each gas. This case
problem.19 Gislason in fact provides significant physical follows very closely the adiabatic piston discussed in our
insight by identifying the damping of the piston motion as a previous work.20
result of the dynamic pressure on it, “because the pressure is The conservation of energy reads
greater when the piston is moving towards the gas than when
the piston is moving away from the gas.”19 dUA þ dUB þ dEpot þ dEkin ¼ 0; (45)
After the equalization of pressures, the coupling between
the two subsystems remains; only the kinetic energy of the where dUA and dUB are the internal energies of subsystems
piston becomes negligible. Still, the collisions between the A (bottom) and B (top), respectively, whereas dEpot and
gas particles and the piston will make the piston jiggle, dEkin are the piston gravitational potential energy and its ki-
allowing an exchange of energy between the two gases.3,20 netic energy, respectively. If the piston has mass M and area
These energy exchanges will always take place, despite the A, and noting that dEpot ¼ ðMg=AÞdVA , the reader is invited
piston being a thermal nonconductor and the absence fric- to adapt our former calculations20 and verify that the equilib-
tion, as they are simply a result of the momentum transfer in rium condition corresponds to equality of forces on the pis-
the collisions (cf. the discussion by Feynman3). And, as ton and equality of temperatures: PA ¼ PB þ Mg=A and
pointed out in Sec. IV A, these energy exchanges can be for- TA ¼ TB .
mally treated as heat exchanges. Therefore, in this second Furthermore, since dVA þ dVB ¼ 0, the work of the result-
phase, the system evolves to a situation of equal tempera- ant of the forces on the piston is [cf. Eq. (15)]
tures, with DUA þ DUB ¼ 0. In this case, if we write the first Mg
law for gas A we have Q ¼ DUB 6¼ 0 [Eq. (40)], and the ðP~A  P~B ÞdVA  dVA ¼ dEkin : (46)
condition dQ ¼ 0 cannot be imposed.20 In fact, we have A
instead dQA ¼ dQB [Eq. (16) in Ref. 20]. Notice that the Therefore, from Eqs. (45) and (46), we have
different quantities somewhat lose their intuitive interpreta-
tion, merely being a result of the mathematical formalism. ðP~A  P~B ÞdVA ¼ dEkin þ dEpot ¼ dUA  dUB : (47)
Another way of looking to the problem is to note that after
the equalization of pressures, there are configurations in the Finally,
vicinity of this mechanical equilibrium with greater global X X
entropy, and the system will move towards these configura- dUi ¼ P~i dVi
tions. As a consequence, the system will indeed access the i i
different available microscopic configurations and move as a X X
result of a blind entropic process, in accordance with Boltz- ¼ ðPi dVi þ Ti dSi Þ 6¼ ðPi dVi Þ (48)
i i
mann’s basic ideas and his microscopic interpretation of en-
tropy. The latter also furnishes an explanation on the with i ¼ fA, Bg, and
additive property of entropy. X X
From the discussion above, it is clear that during both Ti dSi ¼  ðP~i  Pi ÞdVi : (49)
phases of evolution there has to be a “heat exchange,” i i
according to the formalism of thermodynamics, no matter
whether the piston was defined as “adiabatic,” which might Exactly like the case of the adiabatic piston, the direct use of
seem a bit shocking at first. The problematic use of language the first law for one of the gases and the assignment of physi-
is easily avoided if we leave behind a formulation of the first cal meaning to the quantity Q are not straightforward, as nei-
law which to some extent still dates from the time of caloric, ther the conditions dQ ¼ 0 and dUi ¼ Pi dVi nor even
and instead keep in mind its introduction as suggested in dUi ¼ P~i dVi can be imposed.20,28
Sec. IV A and from Eq. (35). Then, there is no doubt that the
“adiabatic piston” system allows the energy exchange 2. The constant pressure calorimeter
between both subsystems. And it is by no means shocking to
assert that the internal energy of each subsystem changes One important configuration in practical applications is
due to the collisions, even for a piston with zero heat conduc- the constant pressure calorimeter.12,13 The system can be
tivity and moving without friction. idealized by adding a piston to subsystem B from Fig. 5 to
keep it at constant pressure. The new configuration is
D. Further generalizations depicted in Fig. 7.
Let us denote the pressure exerted by the piston on subsys-
From a general introduction of the first and second laws tem B by P0 . In other words, P0 ¼ MB g=AB , where MB is the
of thermodynamics and an early alert on the dangers of mass of the piston on subsystem B and AB its area. Likewise,
a blind use of the mathematical formalism, as outlined in let us define Pe ¼ MA g=AA .

634 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 634
 DUA ¼ Pe DVA  ðP0 DVB þ DUB Þ
¼ Pe DVA  DHB ; (56)
Ð
where DHB ¼ CP;B dT. If the specific heat at constant pres-
sure CP;B is constant, we can finally write

DUA ¼ Pe DVA  CP;B DTB : (57)

Thus, if we want to apply the first law of thermodynamics

(36) to gas A, the second term on the r.h.s. corresponds to an
energy exchange with subsystem B that we can identify with
the heat exchanged with subsystem B between two points of
equilibrium.

V. DISCUSSION
We have presented a simple and clear model for introduc-
ing thermodynamics, which reveals and naturally solves
some of the difficulties underlying the concepts of work and
heat in the formulation of the laws of thermodynamics.
The first step is the extension of the notions of kinetic and
potential gravitational energies to the one of internal energy,
inferred in Sec. II from the analysis of extended deformable
bodies. Subsequently, from the base model presented in
Sec. III it is shown that the internal energy depends on the
position and entropy, U ¼ U(S, V).
Fig. 7. The constant pressure calorimeter can be schematically represented It is pointed out that the dynamic force on the piston has a
by two subsystems, A (gas) and B (calorimeter), which can exchange energy
through a common border. A frictionless piston of mass MB keeps subsystem
dissipative character, even if we have only conservative
B at constant external pressure (instead of constant volume, as in Fig. 5). forces and there is no friction. The second law of thermody-
namics is then readily obtained (Secs. III A and III B). An
interesting idealized situation of our base model corresponds
Consider first a reversible transformation. In this case, the to a very simple one-dimensional picture, namely, a gas
pressure PB is always equilibrated at P0 and there is no dif- formed of N point particles of mass m moving only on the
ference between PB and P~B . Therefore, defining the enthalpy vertical direction under the action of gravity, and colliding
H as elastically with the piston of mass M. There is no friction
and the particles do not interact directly with each other.
H ¼ U þ PV (50) Even this straightforward model is enough to understand the
dissipative character of the dynamic force, the approach to
and the specific heat at constant pressure CP from equilibrium, and, thus, irreversibility. In the case where all
particles are initially exactly at the same height and have
ðdHÞP ¼ CP dT; (51) exactly the same velocity, the situation is the same as with a
one-dimensional elastic collision between two point masses
we have
(one of mass Nm and the other of mass M). Therefore, the
dHB ¼ dðUB þ PB VB Þ ¼ dUB þ PB dVB ¼ TB dSB piston remains oscillating indefinitely. The dissipative char-
acter of the dynamic force does not appear and the entropy
 CP;B dTB : (52) remains constant. The system “has no imagination,” the ac-
cessible volume in phase-space remaining very limited.
To address the general case of an irreversible transforma- However, if the masses m are not exactly “in phase,” if there
tion (for instance, if we would initially have Pe > PA ), we is a small difference in their positions or speeds, the dissipa-
note that the conservation of energy reads tive character emerges and there is an entropy increase (It
can be noted that the notion of “exactly the same height and
dUA þ dUB þ dEpot;A þ dEkin;A þ dEpot;B þ dEkin;B ¼ 0; exactly the same velocity” does not make sense in quantum
(53) mechanics. However, it is not necessary to invoke quantum
mechanics for the point we are making here). The accessible
where dEpot;A and dEkin;A are the potential and kinetic ener- volume in phase-space has now increased. The key factor
gies of piston A, given by Eqs. (15) and (16), respectively, leading the evolution to equilibrium is the interaction
and the same for piston B. Hence, we still have between the different particles, even if it is kept to a mini-
mum and only takes place indirectly through the collisions
dUA þ dUB ¼ P~A dVA  P~B dVB : (54) they experience with the piston. These ideas are in line with
the pioneering works by Ludwig Boltzmann. A somewhat
Between two points where both pistons are at rest poetic statement expressing this main result would be
(DEkin;A ¼ DEkin;B ¼ 0) we have, successively, “thermodynamics is mechanics with imagination.” A very
interesting simulation of a rather similar system is available
DUA þ DUB ¼ Pe DVA  P0 DVB ; (55) online from the NetLogo Models Library.29,30

635 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 635
 The approach to equilibrium raises the question of irrever- The heat reservoir was analyzed in Sec. IV B, as a limit of
sibility. The example of compression/expansion with sand the case of energy exchanges between two subsystems,
grains illustrates that the higher the imbalance between the which can be described as heat exchanges. The first law,
dynamic and static pressures, the higher the increase in en- which should not be misinterpreted as a formal generaliza-
tropy, as also shown by Gislason.19 The importance of the tion of the principle of conservation of energy, is also
so-called “quasi-static” formulations is then easily under- derived in this context (Sec. IV A). Furthermore, the notion
stood. It is the work of the dynamic force that is equal to the that “there is no such thing as perpetual motion with weight
variation of the internal energy of the gas [Eq. (19)]. In a lifting machines”3,4 (Sec. II) can now be easily extended to
quasi-static process, the work of the dynamic force is a good account for the second law. In fact, this statement reflects the
approximation for the work of the static force during part of conservation of energy when entropy is not involved, so that
the trajectory of the piston, the variation of entropy being U ¼ U(V). In this case, W ¼ DU and, when the system
nearly zero [Eq. (24)]. For any real process, it is not possible returns to its initial position, DU ¼ 0 and hence W ¼ 0. In
to actually return to the initial conditions. Thus, it is neces- the general case, with friction or even simply the reorganiza-
sary to generalize the idea of mechanical potential energy to tion of the internal energy as a result of the action of the
the idea of gas internal energy. Whereas mechanical poten- dynamic force, U ¼ U(S, V), with DS > 0. When the system
tial energy is a function of only the configuration (or defor- returns to its initial position we have DU > 0, so that W > 0,
mation), here denoted V, the gas internal energy depends on leading to the conclusion that “there is no perpetual motion
one additional variable, S. Mechanics corresponds to isen- at all.”3 This inference is valid both when subsystem B is fi-
tropic thermodynamics, i.e., to situations where the internal nite (cf. Fig. 5) and in the limiting case of a heat reservoir, in
energy is a potential energy, U ¼ U(V), either by the nature what may be seen as a generalization of the Kelvin-Plank
of the problem or as an approximation. formulation of the second law.
The traditional development of thermodynamics defines The analysis of the adiabatic piston problem (Sec. IV C)
the internal energy first as a function of V and T. Noting has to be done with care. It is no longer possible to separate
that ð@U=@T ÞV ¼ CV > 0, we could then be led to think the energy-momentum exchanges of the particles from the
that mechanics corresponds to isothermal thermodynamics. two subsystems with the piston into quantities “work” and
However, this is not the case. During the sand-grain trans- “heat” with clear energetic meaning. The correct and com-
formation of the base model, where P ’ P, ~ we have plete solution of this problem may contribute to illustrate the
dT 6¼ 0. In particular, during the compression and the difficulty in assigning a physical meaning to these two quan-
expansion we have, respectively, dT > 0 and dT < 0. The tities, as they appear in the laws of thermodynamics.20,24 The
transformation is thus characterized by dS ¼ 0 and not by “jiggling piston” further provides a perfect bridge between
dT ¼ 0. That being so, mechanics indeed corresponds to thermodynamics and the microscopic structure of matter—
isentropic thermodynamics and not to isothermal thermody- Feynman’s atomic hypothesis, as he said to his students in
namics, reinforcing the importance of considering S the the very first chapter of his Lectures on Physics.3 Although
conjugate variable for V. the final result of equal pressures and temperatures can be
The idea that when a system returns to the initial position, obtained without referring to heat and thermodynamics,3 the
such as after the compression–expansion from our base complete analysis allows a further exploration of the micro-
model, it has a higher internal energy than at the beginning scopic interpretation of entropy.31
[Eq. (9)], is fairly counterintuitive. This comes from the fact
that physical systems are often surrounded by a thermostat,
which prevents the manifestation of the thermodynamic phe- ACKNOWLEDGMENTS
nomenon. A good example is the deformation of an elastic
material, such as a spring hanging vertically, holding a cer- The authors would like to thank the two anonymous refer-
tain mass. The situation is very much like our base model ees for their careful revision and many suggestions made,
and an analysis similar to that in Sec. III A can be made. By which decisively contributed to improve the clarity and
way of illustration, a new mass can be added to the spring, enlarge the scope of the text.
which oscillates until a new equilibrium position is found. 1
The additional mass can then be removed to the side and the K. Huang, Statistical Mechanics, 2nd ed. (Wiley, New York, 1987).
2
H. B. Callen, Thermodynamics and An Introduction to Thermostatics, 2nd
spring again oscillates until it reaches its original equilib-
ed. (Willey, New York, 1985).
rium position, i.e., it recovers its initial deformation amount. 3
R. P. Feynman, R. B. Leighton, and M. Sands, The Feynman Lectures on
We are led to think that everything happens as if the dynamic Physics, 13th printing (Addison-Wesley, Reading, MA, 1979), Vol. I.
force could be approximated by the static force and internal 4
L. da Vinci, “The Impossibility of Perpetual Motion,” In Energy: Histori-
dissipation would not exist, as apparently we return to the cal Development of the Concept (Benchmark Papers on Energy), edited by
initial state. However, the excess internal energy and the en- R. B. Lindsay (Hutchinson & Ross, Dowden, 1975), Vol. 1, p. 72.
5
S. G. Brush, The Kind of Motion We Call Heat: A History of the Kinetic
tropy variation of the spring are transferred to the surround-
Theory of Gases in the Nineteenth Century, Book 1: Physics and the Atom-
ing environment. The energy increase of the environment is ists (North-Holland, Amsterdam, 1986).
equal to the work of the dynamic force and is equal to the 6
L. Boltzmann, “Weitere studien üuber das wärmegleichgewicht unter
changes of gravitational potential energy of the masses gasmolekülen,” Wiener Berichte 66, 275–370 (1872).
7
which are now at a lower level. Exactly the same would hap- L. Boltzmann, “Über die beziehung dem zweiten Haubtsatze der mechani-
pen in the case of our base model in contact with a heat res- schen wärmetheorie und der Wahrscheinlichkeitsrechnung respektive den
ervoir. This example strengthens the idea of using entropy as Sätzen über das Wärmegleichgewicht,” Wiener Berichte 76, 373–435
(1877).
the additional variable, as an analysis based on the volume 8
U. Besson, “The distinction between heat and work: An approach based on
and temperature would hide the problem and we would be a classical mechanical model,” Eur. J. Phys. 24, 245–252 (2003).
tempted to say that both the spring and the environment had 9
J. M. Mı́nguez, “The work–energy theorem and the first law of
recovered the initial state. thermodynamics,” Int. J. Mech. Eng. Educ. 33, 77–82 (2005).

636 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 636
10 21
E. A. Gislason and N. C. Craig, “General definitions of work and heat in D. Kivelson and I. Oppenheim, “Work in irreversible expansions,”
thermodynamic processes,” J. Chem. Educ. 64, 660–668 (1987). J. Chem. Educ. 43, 233–235 (1966).
11 22
G. M. Barrow, “Thermodynamics should be built on energy-not on heat J. De Heer, Phenomenological Thermodynamics with Applications to
and work,” J. Chem. Educ. 65, 122–125 (1988). Chemistry (Prentice-Hall, Englewood Cliffs, NJ, 1986).
12 23
E. A. Gislason and N. C. Craig, “Cementing the foundations of thermody- E. Kestemont, C. Van den Broeck, and M. Malek Mansour, “The ‘adia-
namics: Comparison of system-based and surroundings-based definitions batic’ piston: and yet it moves,” Europhys. Lett. 49, 143–149 (2000).
24
of work and heat,” J. Chem. Thermodyn. 37, 954–966 (2005). R. de Abreu, “The first principle of thermodynamics and the non-
13
E. A. Gislason and N. C. Craig, “Pressure–volume integral expressions for separability of the quantities ‘work’ and ‘heat’: The adiabatic piston con-
work in irreversible processes,” J. Chem. Educ. 84, 499–503 (2007). troversy,” hhttp://arxiv.org/abs/cond-mat/0205566i (2002).
14 25
S. G. Canagaratna, “Zeroth law, entropy, equilibrium, and all that,” M. Malek Mansour, A. L. Garcia, and F. Baras, “Hydrodynamic descrip-
J. Chem. Educ. 85, 732–736 (2008). tion of the adiabatic piston,” Phys. Rev. E 73, 016121 (2006).
15 26
J. P. Bizarro, Entropy production in irreversible processes with friction,” J. Anacleto and J. A. C. Anacleto, “Thermodynamical interactions: subtle-
Phys. Rev. E 78, 021137 (2008). ties of heat and work concepts,” Eur. J. Phys. 29, 555–566 (2008).
16 27
J. P. Bizarro, “Thermodynamics with friction I. The Clausius inequality J. J. Brey and N. Khalil, “Critical behavior of two freely evolving granular
revisited,” J. Appl. Phys. 108, 054907 (2010). gases separated by an adiabatic piston,” Phys. Rev. E 82, 051301 (2010).
17 28
M. de Abreu Faro and R. de Abreu, “A One-Dimensional Model of Irre- R. de Abreu and V. Guerra, “Comment on ’A close examination of the
versibility,” in EPS 10 Trends in Physics (10th General Conference of the motion of an adiabatic piston’ by Eric A. Gislason [Am. J. Phys. 78,
European Physical Society) (Sevilla, Spain, 1996), p. 314. 995–1001 (2010)],” hhttp://arxiv.org/abs/1012.4918i (2010).
18 29
R. de Abreu, “Análise dinâmica da tendência para o equilı́brio num mod- U. Wilensky, Center for Connected Learning and Computer-Based Model-
elo simples: a 2a lei de Newton f ¼ ma e a 2 a lei da termodinâmica ing, Northwestern University, Evanston, IL, hhttp://ccl.northwestern.edu/
dS  0,” (in Portuguese) hhttp://arxiv.org/pdf/physics/0210084i (2002). netlogo/models/GasLabAdiabaticPistoni (1997).
19 30
E. A. Gislason, “A close examination of the motion of an adiabatic U. Wilensky, Center for Connected Learning and Computer-Based Model-
piston,” Am. J. Phys. 78, 995–1001 (2010). ing, Northwestern University, Evanston, IL, hhttp://ccl.northwestern.edu/
20
R. de Abreu and V. Guerra, “Comment on ‘A close examination of the netlogoi (1997).
31
motion of an adiabatic piston,’ by Eric A. Gislason [Am. J. Phys. 78 (10), J. Arnaud, L. Chusseau, and F. Philippe, “Démocrite et la puissance
995–1001 (2010)],” Am. J. Phys. 79, 684–685 (2011). motrice du feu” (in french), hhttp://arxiv.org/abs/1104.0836i (2011).

637 Am. J. Phys., Vol. 80, No. 7, July 2012 R. de Abreu and V. Guerra 637