Thermodynamics and Kinetic Theory

Igor Boettcher

University of Alberta
Contents
1 What is thermodynamics? 3

2 Thermodynamic systems 4
2.1 Thermodynamic variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 States and equilibrium states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Equation of state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Extensive and intensive variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Thermodynamic processes and work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 First law of thermodynamics 20

3.1 Conservative and non-conservative forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Energy and heat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Differentials in thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.4 Heat capacity and calorimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.5 Application of the first law to gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.6 Adiabatic transformations of a gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4 Second law of thermodynamics 33

4.1 Perpetuum mobile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Carnot cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 Statement of the second law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.4 Heat engines and refrigerators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.5 Zeroth law of thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.6 Absolute temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.7 Clausius inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.8 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5 Thermodynamics potentials 52
5.1 Free energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Chemical potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Overview of thermodynamic potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.4 Maxwell relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5 Phase transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

6 Kinetic theory of gases 65

6.1 Brief review of classical mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.2 Boltzmann equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.3 Maxwell–Boltzmann distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.4 Equipartition and virial theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

2
1 What is thermodynamics?

Thermodynamics is mainly concerned with the transformations of heat into mechanical work and the opposite
transformations of mechanical work into heat.

But what are work and heat?

While work is a relatively well-understood concept from Newtonian mechanics and classical electrodynamics,
it is less obvious what heat really is. For a long time, scientists believed that heat was some sort of fluid
whose total amount was invariable, and heating of a body was transfer of this fluid from one body to another.
Only in the mid-1800s, when the theory of thermodynamics was developed, did it crystallize that heat is a
form of energy, and that heat and mechanical work are equivalent.
Thermodynamics is a phenomenological theory. This means that it formulates a few basic postulates based
on experimental evidence and derives conclusions from these postulates. It does not make (or need!) any
assumptions about the microscopic mechanisms underlying the phenomena observed.
Having an insight into such microscopic mechanisms, on the other hand, is often very useful and also more
satisfying, since pure thermodynamics can be rather abstract. The kinetic theory of gases explains the
equivalence of heat and mechanical work by reducing all thermal phenomena to the disordered Newtonian
motion of large numbers of atoms and molecules. More generally, the fundamental laws of thermodynamics
can be derived very satisfactorily from the theory of statistical mechanics, which also accounts for quantum
mechanical effects.

Some historical context and milestones of thermodynamics:

1760–1840 Industrial revolution: transition from hand production to machines

1710–1770 Steam engines (Watt, . . . )
1780s Electric batteries (Galvani, Volta)
1800–1850 Steam and electric locomotives, railways
> 1850 Age of oil: use of refined petroleum as fuel and in machines

1850–1900 Internal combustion engines (Otto, Diesel, . . . )

1886 Cars (Benz)

1687 Newton’s Principia

1824 Second law of thermodynamics: limitations on converting heat into mechanical work (Carnot)
1840s First law of thermodynamics: equivalence of heat and work, conservation of energy (Joule, Mayer)
1850s Definition of entropy (Rankine, Clausius)
1870s Kinetic theory and statistical mechanics (Maxwell, Boltzmann, Gibbs)
1912 Third law of thermodynamics: entropy at zero temperature, needs quantum mechanics (Nernst)

3
2 Thermodynamic systems
2.1 Thermodynamic variables
Typical systems we consider in thermodynamics are
ˆ a block of iron,

ˆ a bucket of water,

ˆ a balloon filled with Helium gas.

These systems have two important aspects in common. They

(i) occupy a volume much larger than the extension of their constituent particles (atoms and molecules),
(ii) contain a large number of these constituent particles.

Systems with qualities (i) and (ii) are called macroscopic. This is in contrast to microscopic systems, which
are typically tiny and may contain only a few particles.

In each example system, we could try to describe the physics by solving Newton’s equations of motions
for each of the N particles in the system. Complete information then requires to determine the position
xi = (xi , yi , zi ) and velocity vi = (vx,i , vy,i , vz,i ) for each particle, i = 1, . . . , N , at each instance of time.

ˆ Exercise. Estimate the number of atoms N in a 1 cm3 block of iron, with mass density ρ = 7.87g/cm3
and atomic mass of m0 = 55.845 u. (The atomic mass unit is 1u = 1.66 × 10−27 kg.)
ˆ Answer. The block has a volume of V = 1cm3 . The total mass of the block is M = ρV . One atom
weighs on average m0 , so the total number of atoms is
M ρV
N= = (2.1)
m0 m0
7.87g × 1cm3
= (2.2)
cm3 × 55.845 × 1.66 × 10−27 kg
7.87 × 10−3 kg × 1cm3
= (2.3)
cm3 × 55.845 × 1.66 × 10−27 kg
7.87 × 10−3
= (2.4)
55.845 × 1.66 × 10−27
= 8.5 × 1022 . (2.5)

Solving that many Newtonian equations, or specifying the initial conditions, or even storing and sensibly
interpreting the final outcome, is practically impossible. However, it is also absolutely unnecessary:
Thermodynamic systems consist of a very, very large number of particles N . Experiments performed on
macroscopic length and time scales are dominated by the collective behavior of particles, not individual
particles. These systems are, therefore, described by a small set of thermodynamic variables.
We now discuss the most important thermodynamic variables and how they can be measured.

Volume V . The volume is simply the interior volume of the container that contains the thermodynamic
system. If the linear extensions of the system are Lx , Ly , Lz in x, y, z direction, then V = Lx Ly Lz . If Lz is
changed by an infinitesimal amount dLz , then this corresponds to a change in volume

V → V + dV (2.6)

4
with
dLz dLz
dV = Lx Ly dLz = Lx Ly Lz =V . (2.7)
Lz Lz
Note that on the right hand side it does not matter whether we insert the original volume (V ) or the new
volume (V + dV ), because the rules of differential calculus imply

dV dLz ≈ 0 (2.8)

to leading order. The SI unit for volume is m3 or liter,

1 l = 1 dm3 = 1 × (0.1m)3 = 10−3 m3 . (2.9)

Memorize that 1 liter of water almost exactly weighs 1 kg. 1 quart (= 1 pack of chicken stock) is roughly
the same as 1 liter. Another common measure is the gallon,
(
4.5 l (UK, Canada, ...)
1 gal = = 1 big milk bottle. (2.10)
3.8 l (US,...)

To measure the volume of a liquid, one pours it into a container of known linear extensions such as a
measuring jug. To measure the volume of a (perhaps weirdly shaped) solid object, one can suspend it into
water and then measure the amount of water that has been displaced. Measuring the volume of a gas is
more tricky, because gases are compressible, i.e. their volume can be changed upon exerting pressure, and
their volume stronger depends on temperature. For gas inside a container of known volume V , one always
needs to also measure pressure and temperature.

Particle number N . The particle number N remains constant in many elementary thermodynamic processes.
(Counter-examples are chemical reactions, or biological systems with osmotic walls which allow particles to
leave or enter the system.) Note that N is a dimensionless number, say, N = 1, 999, 045, 788. However,
chemists like to use a standard unit of particle number, which is the mole.
A mole consists of

NA = 6.022 × 1023 (2.11)

particles (a dimensionless number!). Compare this to other such conventions: a pair consists of 2 entities,
a quartet consists of 4 entities, a dozen consists of 12 entities. In calculations, I recommend to use the
dimensionless unit

1 mol = NA = 6.022 × 1023 (2.12)

to keep track of the number of moles. For instance, the number of iron atoms determined earlier is
1 mol 1 mol
N = 8.5 × 1022 = 8.5 × 1022 = 8.5 × 1022 = 0.14 mol. (2.13)
NA 6.022 × 1023
The number of moles ν, i.e. the number that satisfies N = ν mol = νNA (ν = 0.14 in the example), is given
by
N
ν= . (2.14)
NA
The quantity ν is sometimes denoted n in other reference, but not in this lecture! Instead, we will use n to
denote the particle number density
N
n= . (2.15)
V

5
For physicists, the introduction of moles is not too useful, and I recommend to always work with N instead
of ν. Chemists might find ν more handy for expressing their results. Indeed, the unwieldy atomic mass unit
1 u = 1.66 × 10−27 kg corresponds to a mass per mole given by

NA 6.022 × 1023 g
1u=1u = 1.66 × 10−27 kg =1 , (2.16)
1 mol 1 mol mol
which is a decent quantity to work with. The hardly intuitive atomic mass of iron, m0 = 55.845 u, becomes
NA g
m0 = m0 = 55.845 , (2.17)
1 mol mol
and so on. For a thermodynamic system that consists of one chemical compounds with mass m0 , the total
mass of the system is

M = N m0 , (2.18)

and the mass density is given by

M
ρ= = m0 n. (2.19)
V
The total mass M of a system can be measured with scales, i.e. by comparing its mass to a reference object
whose mass is known. This allows to measure the particle number N if m0 is known.

Pressure P . The pressure P is defined quite mechanistically as

force
P = . (2.20)
area
It can be measured at any spatial position inside the thermodynamic system. When measured at the system
boundary, P measures the force exerted by the system onto the container walls. When measured inside the
bulk of the system, one should think of a small imagined planar element of area A. The force F exerted
by the system onto this plate from one side (which equals the force exerted from the other side), is related
to the pressure via P = F/A. Constant pressure P means that the amount of force F (in Newton) grows
linearly with the area A (in m2 ) of the imagined planar element.
A pressure measurement device is called manometer or barometer. They usually translate the volume
measurement of a gas into a pressure. Depending on the context, pressure is measured in many units. The
SI unit is Pascal,
N kg
1 Pa = 1 = . (2.21)
m2 m s2
Standard atmospheric pressure is

1 atm = 101.325 kPa, (2.22)

which is the Earth’s average air pressure at sea level. Another popular unit is

1 bar = 100 kPa, (2.23)

so that

1 bar ≈ 1 atm. (2.24)

Other units are Torr or psi (pounds per square inch). Typical car tires have a pressure of

34 psi = 34 × 7 kPa = 240 kPa = 2.4 atm, (2.25)

6
whereas bicycles have a wider range, and typically higher pressure.

Temperature T . Temperature is the most interesting variable here. In the kinetic picture, temperature is
defined as the average kinetic energy of a gas of particles, basically the formula
3
⟨E⟩ = N kB T, (2.26)
2
which, again, is a mechanistic definition. However, the property ”temperature” of a system can be defined
purely phenomenologically through the experimentally confirmed zeroth law of thermodynamics. For now,
we take for granted that there exist thermometers, which are devices which, when brought into contact with a
thermodynamic system, display an empirical temperature ϑ. For example, a mercury thermometer displays
ϑ through thermal expansion of a mercury column. A gas thermometer measures the temperature T via the
variation of volume or pressure of a gas. We call T the absolute temperature. Any two thermometers can
be calibrated with each other to determine the function ϑ(T ).
The absolute temperature T is measured in the SI-unit Kelvin. Note that we write (and say)

1 K, not 1◦ K. (2.27)
Temperature differences ∆T = T2 − T1 in K are identical when expressed in ◦ C. The freezing and boiling
point of water at 1 atm are

Tfreeze = 0◦ C = 273.15 K, (2.28)

◦
Tboil = 100 C = 373.15 K, (2.29)

and the absolute zero of temperature corresponds to

T = 0 K = −273.15◦ C. (2.30)

Temperatures expressed in degrees Fahrenheit can be converted to degrees Celsius through the rough rule
”minus 30, divide by two”, for instance

70◦ F ≈ 20◦ C, (2.31)

◦ ◦
80 F ≈ 25 C, (2.32)
90◦ F ≈ 30◦ C. (2.33)

Summary. The main thermodynamic variables are:

ˆ volume V

ˆ particle number N

ˆ particle number density n = N/V

ˆ pressure P

ˆ temperature T

The 6N positions and velocities of all particles in a system determine its dynamical state. The knowledge of
the thermodynamic variables of a macroscopic system are by no means sufficient to determine its dynamical
state. For any choice of admissible thermodynamic variables, there is an infinite number of states of atomic or
molecular motion that correspond to it. During macroscopic time scales, the system rapidly passes through
many of these corresponding dynamical states, while leaving the thermodynamic variables constant.

7
2.2 States and equilibrium states
The distinction between (general) thermodynamic states and (specific) equilibrium states of macroscopic
systems is a crucial aspect of thermodynamics. This important concept, for instance, needs to be mastered
to understand the difference between reversible and irreversible processes later.

A macroscopic system is in a thermodynamic state if it can be described by volume, V , and local particle
number density, pressure, and temperature, n(x), P (x), T (x), where x = (x, y, z) is any space point inside
the system. More local thermodynamic variables may be needed for more complex system.

What does, say, local T (x) mean? We only defined T for large macroscopic systems. While the concept of a
local temperature may be intuitive, we need to properly define it. For this note that the number of particles
N is usually so large that we can divide the volume V into many smaller compartments of volume V0 ≪ V ,
such that each volume V0 still contains a macroscopically large number of particles, and hence pressure and
temperature are well-defined thermodynamic variables. Furthermore, the compartments can be chosen small
enough so that pressure and temperature within each compartment are approximately constant. For any
individual compartment at position x0 inside the system, we then define
N0
n(x0 ) = , P (x0 ) = P0 , T (x0 ) = T0 , (2.34)
V0
where V0 , N0 , P0 , T0 are the volume, particle number, pressure, and temperature of that individual com-
partment. By considering all compartments at various locations, we obtain the functions n(x), P (x), T (x).

The thermodynamic variables of a system are usually controlled by external conditions. For instance, the
volume is determined by the geometry and hence the walls of the container. The temperature can be ma-
nipulated through the temperature of the environment. The particle number may be fixed by ensuring that
no particles can leave through leaks in the container walls.

An equilibrium state of a macroscopic system is a thermodynamic state whose local thermodynamic variables
do not vary as long as the external conditions remain unchanged.

The reference to time in this definition, of course, means macroscopic time scales relevant to experiment,
not time scales relevant for the microscopic motion of individual atoms or molecules. An equilibrium state
can be created in practice by
1. imposing certain external conditions,
2. waiting.
Over a sufficiently long time, the microscopic mechanisms in the system will make it equilibrate, i.e. reach
the equilibrium state.

In many cases, the external conditions do not impose any spatial variation of the thermodynamic variables
inside the system, and the equilibrium states are simply the homogeneous states, in which n(x), P (x),
T (x) are independent of x. Non-equilibrium states typically feature macroscopic currents as they appear in
hydrodynamics.

The following two examples of equilibrium and non-equilibrium states may be instructive.
ˆ Example 1. In the isothermal expansion, a gas is initially confined to the left one of two equal containers,
with a volume V , with constant n, P, T . This is an equilibrium state. The gas is coupled to a heat

8
 equilibrium 2 non-equilibrium
1

3 non-equilibrium 4 equilibrium

Figure 2.1: Isothermal expansion of a gas (Example 1)

bath that keeps it at a constant temperature.

A wall is removed and the gas can freely expand into the second container, with the total system
volume now being 2V . The temperature does not change in this process because the gas is coupled to
the heat bath. After a sufficiently long time the system reaches a new equilibrium state with volume
2V , density n/2, pressure P/2, and temperature T .
However, at any intermediate instant of time, the system with volume 2V is in a non-equilibrium state
with inhomogeneous density and pressure profile, n(x) and P (x). These density and pressure gradients
induce macroscopic currents (or flow) in the system.
It is possible to modify the setup so that the intermediate states are also equilibrium states (see below).
ˆ Example 2. Consider a homogeneous gas in a volume V with spatially constant n, P , T . If the system
is placed in Earth’s gravitational field with gravitational force F = −m0 gẑ, then the homogeneous
state is not the equilibrium state. The equilibrium state, instead, has a z-dependent density, pressure,
and temperature (barometric height formula).
Note: For most gases, m0 is much too small to produce a noticeable effect on the length scales of
typical laboratory experiments. The gravitational force can then be neglected.

These examples may give the impression that equilibrium states cannot describe thermodynamic systems
that ”move”, meaning that their thermodynamic variables vary in time. This would be bad, since, clearly,
a Diesel engine is only useful when it is running. Note, however, that our definition of equilibrium relies
on unchanged external conditions. In any motor or machine, pistons and the like are moving, so that the
external conditions themselves are time-varying.

We can now formulate some important definitions that specify the relation between a thermodynamic system
(the ”system”), assumed to be contained in a container, and the surrounding of the container (the ”environ-

9
 environment

system

Figure 2.2: System and environment

ment”). Consider the system as a compartment at position xs , with system thermodynamic variables

Ts = T (xs ), Ps = P (xs ), (2.35)

and the environment as a compartment at position xe , with environment thermodynamic variables

Te = T (xe ), Pe = P (xe ). (2.36)

We say that the system is thermally insulated or thermally isolated, if the container walls do not allow heat
to be exchanged between the system and the environment. (For instance, a double wall filled with air, which
is a bad heat conductor.) In contrast, if the walls allow for heat to be transmitted, such as a thin sheet of
metal, we say that system and environment are in thermal contact or that they can exchange heat.1 When
a system and its environment are in thermal contact, we say that the system is in thermal equilibrium with
its environment if

Ts = Te . (2.37)

When a system and its environment are separated by a movable wall, then this wall will be shifted around if
the pressures from either side do not match. In this case, we say that they are in mechanical equilibrium if

Ps = Pe . (2.38)

Quite generally, if two systems with temperatures T1 and T2 , and pressures P1 and P2 , are in equilibrium
with each other, then

T1 = T2 , (2.39)
P1 = P2 . (2.40)
This equilibrium state can only be reached when the two systems are able to ”talk to each other” (transfer
heat or do work), hence we need thermal contact and movable walls to establish equilibrium.

In the following we will always assume that there is no exchange of particles between the system and the
environment. Let us discuss, however, how such an exchange would look like. Assume the system consists of
1 Good heat conductors are heat and stone, bad heat conductors are air, wood, clothes. As a rule of thumb, if something

feels cold to the touch, then it is a good heat conductor (heat leaves your hand), whereas if it feels warm, then it is a bad heat
conductor (heat stays in your hand).

10
water, with density n(wat) , and the environment contains oil, with density n(oil) , both separated by container
walls. Then a complete set of thermodynamic variables is (n(wat) , n(oil) , T, P ). We assume Ts = Te = T . If
the container walls are impenetrable for the molecules, then

n(wat) (xs ) = ns , n(wat) (xe ) = 0, (2.41)

(oil) (oil)
n (xs ) = 0, n (xe ) = ne . (2.42)

On the other hand, if particles can pass through penetrable or porous walls, then we eventually reach a state
of chemical equilibrium with

n(wat) (xs ) = n(wat) (xe ), (2.43)

(oil) (oil)
n (xs ) = n (xe ). (2.44)

2.3 Equation of state

So far, we did not say which external condition needs to be changed to specifically modify the pressure. The
reasons is simple: There is none. The pressure can only be changed indirectly through changing T, N, V .

Consider first a homogeneous system with thermodynamic variables n = N/V, P, T . There exists a relation
called the equation of state which relates these variables according to

ϕ(n, P, T ) = 0. (2.45)
Let us give some meaningful examples first:

ˆ Example 1. Almost all gases at low pressures and sufficiently far above the condensation temperature
are described by the ideal gas equation of state

P V = N kB T (2.46)

with Boltzmann’s constant kB = 1.381 × 10−23 J/K. In this example, we have

ϕ(n, P, T ) = P − nkB T. (2.47)

ˆ Example 2. When temperature and pressure are such that a real gas is near condensation, deviations
from the ideal gas law are observed. A good phenomenological model is then given by the van-der-
Waals equation of state

N kB T aN 2
P = − 2 (2.48)
V − bN V
nkB T
= − an2 (2.49)
1 − bn
where a, b are parameters specific to the gas under consideration. In this case,

ϕ(n, P, T ) = (P + an2 )(1 − nb) − nkB T. (2.50)

We will discuss these two examples in more detail later. Some remarks are in order.

11
 ˆ Thermodynamics is a set of general rules and formulas. The equation of state is the only point where
we specify what substance we are actually dealing with (water, air, iron, . . . ).
ˆ We always assume that we can solve the equation ϕ(n, P, T ) = 0 to express any of the three variables
n, P, T as a function of the remaining two variables. Only two variables of n, P, T are independent.
ˆ The equation of state has been measured for many substances and is stored in databases. Explicit
formulas for ϕ exist only in a few cases. Calculation of the equation of state is one of the subjects of
statistical mechanics or quantum field theory.
ˆ Often it is assumed that N is constant and the equation of state is then written as

ψ(P, V, T ) = 0. (2.51)

For instance, ψ(P, V, T ) = P V − N kB T for ideal gases. For fixed N , the thermodynamic state
is uniquely determined by any two of the variables P, V, T .

Consider now a macroscopic system in a general thermodynamic state, i.e. it is characterized by the ther-
modynamic variables V, n(x), P (x), T (x). Then at each point x there is a local equation of state

ϕ(n(x), P (x), T (x)) = 0. (2.52)

Crucially, the function ϕ(n, P, T ) is always the same, i.e. it does not depend itself on x. The x-dependence
is only through its arguments.

For an ideal gas we can write

N
P V = N kB T = NA kB T = νRT (2.53)
NA
with universal gas constant
J
R = NA kB = 8.314 . (2.54)
K mol

2.4 Mixtures
Mixtures of several chemical compounds are straightforward to treat within thermodynamics, but due to the
increased number of thermodynamic variables, they are a little unwieldy. Here we collect some basic facts.

For a homogeneous macroscopic system that is a mixtures of X different chemical compounds, we associate
particle numbers N (j) to each of the compounds labelled by an index j = 1, . . . , X. The concentration of
component j is given by

N (j)
c(j) = , (2.55)
N
with
X
X
N (j) = N, (2.56)
j=1
X
X
c(j) = 1. (2.57)
j=1

12
 ˆ Example. The chemical composition of air is mostly nitrogen, oxygen, argon, and carbon dioxide,
hence X = 4 and j = N, O, Ar, CO2 , with concentrations

c(N) = 0.7808 = 78.08%, (2.58)

(O)
c = 0.2095 = 20.95%, (2.59)
c(Ar) = 0.0093 = 0.93%, (2.60)
(CO2 )
c = 0.0004 = 0.04%. (2.61)

We have
X
c(j) = 100%. (2.62)
j

The relative particle number density of component j is

N (j)
n(j) = = c(j) n, (2.63)
V
with n = N/V the total particle number density. The temperature of the homogeneous system shall be
T . Each component j has an equation of state ϕ(j) (n(j) , P (j) , T ) = 0 that is independent of the remaining
chemical compounds. If we assume that the interactions between the chemical components j is approximately
absent (this need not be the case), then these equations of state remain valid and the partial pressure P (j)
of the jth compound is found from solving

ϕ(j) (n(j) , P (j) , T ) = 0 for each j. (2.64)

The total pressure is

X
X
P (j) = P. (2.65)
j=1

A complete set of thermodynamic variables of the mixture is then

(n(1) , . . . , n(X) , P (1) , . . . , P (X) , T ). (2.66)

Note also that in general each component j has a different mass, m(j) , and so ρ(j) = m(j) n(j) for the mass
density.

2.5 Extensive and intensive variables

The concept of extensive and intensive variables is very simple, but extremely important and powerful.

Consider a homogeneous thermodynamic system S described by the thermodynamic variables (V, N, P, T ).

Now imagine we would duplicate the system and consider the doubled system as a new thermodynamic
system S ′ itself. The volume and particle number of this new system will be twice that of S,

V ′ = 2V, (2.67)
′
N = 2N. (2.68)

The pressure and temperature, however, remain the same regardless of the spatial extent of the system.
Hence

T ′ = T, (2.69)
′
P = P. (2.70)

13
Similarly, the particle number density n = N/V remains the same,

n′ = n. (2.71)

A thermodynamic quantity that doubles during an imagined doubling of a homogeneous system is called
extensive. Examples are V and N . A quantity that remains invariant under an imagined doubling of a
homogeneous system is called intensive. Examples are P, T, n. The ratio of two extensive quantities is
intensive, as is the case for n = N/V .

Intensive quantities are usually local quantities. Typical examples are densities. For instance, the total mass
of a system, M = m0 N , is an extensive quantity, but the mass density
M
ρ= (2.72)
V
is an intensive quantity. The mass per particle,
M
m0 = , (2.73)
N
as a ratio of two extensive variables is, of course, also intensive.

If the left-hand-side of an equation is an extensive (intensive) quantity, then the right-hand side must be
extensive (intensive) quantity. For instance, for an ideal gas,

P V = N kB T (extensive = extensive) (2.74)

P = nkB T (intensive = intensive). (2.75)

If you are uncomfortable with doubling a given system S, you may alternatively divide the system into
two macroscopic pieces, S = S1 ∪ S2 , by inserting an additional wall. Extensive quantities X then satisfy
X = X1 + X2 , whereas intensive quantities Y satisfy Y = Y1 = Y2 .

2.6 Thermodynamic processes and work

Consider a homogeneous system in a thermodynamic state with fixed N and equation of state

ψ(P, V, T ) = 0. (2.76)

Choose P, V as the independent variables. It is often convenient to represent them in a PV-diagram, with
P on the y-axis and V on the x-axis. Any point in the PV-diagram defines a state of the system. The
temperature T = T (P, V ) is a function of P and V . Points representing states of equal temperature lie on
a curve which is called an isothermal.

ˆ Example. Consider an ideal gas with P V = N kB T . For fixed N we have

PV
T = T (P, V ) = . (2.77)
N kB
Isothermal curves in the PV-diagram correspond to
N kB T const.
P = = . (2.78)
V V
They are hyperbolas having the P - and V -axes as asymptotes.

14
 P P

non-equilibrium process
(e.g. inhomogeneous
intermediate states)

V V
Figure 2.3: Left. PV-diagram and three possible transformations connecting the states (V1 , P1 ) and (V2 , P2 ).
Right. A fourth transformation that utilizes non-equilibrium, inhomogeneous intermediate states and thus
cannot be drawn in a PV-diagram, because P and V are not defined during the intermediate steps.

We define a thermodynamic transformation or process as any transformation that brings a system from an
initial state to a final state through a continuous succession of intermediate states. If the initial and final
states are represented by two points (V1 , P1 ) and (V2 , P2 ) in a PV-diagram, then any curve connecting these
points corresponds to a possible transformation. They correspond to different ways of modifying the external
conditions to get from 1 to 2. However, not every transformation connecting 1 to 2 is represented by such a
curve.

A transformation is said to be reversible when the successive states of the process differ by infinitesimals
from equilibrium states. In particular, the initial and final states of a reversible transformation need to be
states of equilibrium.

If the intermediate states are non-equilibrium states, then they are typically inhomogeneous, and so cannot
be drawn in a PV-diagram.

We revisit here the central example of reversible vs. irreversible processes. We define a heat bath or heat
reservoir as an environment of temperature T that contains so many particles that its temperature does not
change when it is coupled to a (much smaller) system.

15
 Isothermal expansion (irreversible run)

1. Consider a gas inside a container with immovable walls of volume 2V coupled to a heat bath of
temperature T , so that the system temperature remains constant.
2. Initially, at time t ≤ 0, a wall separates the left and right halves of the container. An ideal gas of N
molecules in the equilibrium state (Pi , V, T ) occupies the left half. Its pressure is Pi = N kB T /V .

3. At time t = 0, the separating wall is removed.

4. For t > 0, volume of the system is 2V . The system is now in a non-equilibrium state, having too high
pressure in the left half and almost no pressure in the right half. Rapid macroscopic motions set in to
diminish this pressure gradient, creating macroscopic currents of gas. The temperature stays constant
due to the coupling to the bath.

5. After a sufficient long time teq , the gas reaches the final equilibrium state (Pf , 2V, T ) with pressure
Pf = N kB T /(2V ) = Pi /2.

This thermodynamic transformation from (Pi , V ) to (Pf , 2V ) at constant temperature T is irreversible, since
the intermediate states are far away from equilibrium states. Practically, it is impossible to start in the final
state of the gas occupying both halves of the container and make it move into the left half of the container
without any external influence. If you were shown a video recording of the time-interval t ∈ [0, teq ], you
could tell if the video ran forward or backward. The intermediate states cannot be plotted in a PV-diagram.
However, consider now the following modification.

Isothermal expansion (reversible run)

1. As before. However, an externally controlled piston is inserted in the right half of the container,
initially touching the separating wall.
2. As before.
3. As before.

4. After the separating wall is removed, at t > 0, the piston very slowly inches to the right, by an amount
∆ℓ over a time-interval ∆t. The speed ∆ℓ/∆t is chosen slow enough so that the gas has enough time
to adjust itself to the altered external conditions through the usual molecular motion. No macroscopic
currents are generated. At each time t, the gas is infinitesimally close to an equilibrium state, with a
volume V < V (t) < 2V , temperature T , and pressure Pi > P (t) > Pf .

5. After a certain amount of time ttot , the piston opened up the whole right half of the container, and
the gas is in the final equilibrium state (Pf , 2V, T ).

This way of performing the isothermal expansion meets all criteria of the definition of a reversible process.
And, indeed, it can be reversed: Starting from the final gas occupying the volume 2V with pressure Pf , we
could slowly move the piston to the left with a negative velocity −∆ℓ/∆t, so that the gas is in equilibrium at
each time-step. After the amount of time ttot , the gas will be back in the initial state (Pi , V, T ). If you were
shown a video recording of the time-interval t ∈ [0, ttot ], you could not tell whether the video ran forward
or backward.

The fact that two equilibrium states can be connected by various sorts of processes, some reversible and
some irreversible, will be extremely important later on. In a sense, it is at the very heart of thermodynamics.
So make sure you understood this example of isothermal expansion.

16
 isothermal expansion (irreversibe run)

equilibrium non-equilibrium non-equilibrium equilibrium

isothermal expansion (reversibe run)

equilibrium equilibrium equilibrium equilibrium

Figure 2.4: Isothermal expansion: irreversible vs. reversible run

During a transformation, the system can perform positive or negative work ; that is, the system can do work
on its surroundings or the surroundings can do work on the system. As an example, we consider a fluid
enclosed in a cylinder having a movable piston of area S at one end. If P is the pressure of the fluid against
the walls of the cylinder, then F = P S is the force exerted by the fluid on the piston. If the piston is shifted
by the gas by an infinitesimal distance dh, then an infinitesimal amount of ”work done”

dL = F dh = P S dh (2.79)

is performed, since the displacement is parallel to the force. But Sdh is equal to the increase in the volume
of the system, dV . Thus, we may write

dL = P dV. (2.80)

For a finite transformation, the work done by the system is obtained by integrating this equation,
 V2
L= P dV. (2.81)
V1

Here, 1 and 2 are the initial and final states of the transformation.

Convention. The first law of thermodynamics will be about book-keeping the energy of the system—not
of the environment. Therefore, we want to denote every energy unit that increases the system’s energy as
positive, and every unit that decreases the system’s energy as negative. (Think of your bank account.) If a
system does work on its environment, it loses energy, so this is a negative contribution. This is logical, but
can be cumbersome in some arguments, as we may tend to prefer working with positive quantities.

17
 P P

V V

Figure 2.5: Work done L = −W in the PV-diagram. The right figure shows a cyclic transformation.

We will, therefore, use two different symbols: ”work ” is denoted by the symbol

W. (2.82)

If the system does work, then W < 0, but if work is done on the system so that it gains energy, then W > 0.
We also say that W is the work performed by external forces. In addition, following Fermi, we define the
”work done”, meaning the work done by the system, as

L = −W. (2.83)

If the system does work, then L > 0. (L stands for lavoro, meaning work in Italian.)
In the reversible example above, the work done was dL = P dV , whereas dW = −P dV .

For a system whose state can be represented in a PV-diagram during the transformation, the work done
during a transformation from state (V1 , P1 ) to state (V2 , P2 ) has a simple geometric representation: the
process is represented by a curve P (V ) connecting 1 and 2, the shape of which depends on the type of
process considered. The work done during the process is given by the integral
 V2
L1→2 = P (V ) dV. (2.84)
V1

This integral, and hence the work done, geometrically correspond to the area under the curve in the PV-
diagram. If the process can be run in the opposite direction, the work done is the negative of this result,
 V1
L2→1 = P (V ) dV = −L1→2 . (2.85)
V2

The work performed in a transformation between two states 1 and 2 is typically different for reversible and
irreversible runs. For reversible runs we have

dL = −dW = P dV (reversible), (2.86)

whereas for irreversible runs such a simple formula does not exist.
Consider again the example of isothermal expansion.

Work done in irreversible run. For the irreversible isothermal or ”free” expansion, no work is done:
L = 0. (2.87)

18
Indeed, after the wall has been removed in 3, while the gas expands into the void, the system volume 2V
remains constant and there are no movable walls ”to be pushed”. No work is done.

Work done in reversible run. Now compute the work done in the reversible isothermal expansion from an
initial volume V1 to a final volume V2 > V1 . At each instant of time, the system is homogeneous and can be
represented as a state in a PV-diagram. dL = P dV applies at each time. Since T is constant we have
N kB T
P = , (2.88)
V
hence
 V2
L= P dV (2.89)
V1
 V2
1
= N kB T dV (2.90)
V1 V
V2
= N kB T ln V  (2.91)

V1
 
= N kB T ln V2 − ln V1 (2.92)
V2
= N kB T ln . (2.93)
V1
Since V2 > V1 we have L = L1→2 > 0. The system does work on the piston while expanding. In the reversed
run, the piston does work on the system when compressing the gas and W2→1 > 0.

Important transformations are those for which the initial and final states are the same. These are called
cyclic processes or cycles. If the state of the system can be represented on a PV-diagram, then a cycle can
be represented by a closed curve in this diagram. The work done during one run of the cycle equals the area
enclosed by the curve (or the negative thereof, if the cycle is run in the opposite direction).

Recall that processes with dT = 0 are called isothermal. Here we extend this list of names according to the
following processes:
ˆ dT = 0 or T = const: isothermal,

ˆ dP = 0 or P = const: isobaric,

ˆ dV = 0 or V = const: isochore.

Strictly speaking, isochore refers to dL = 0, but usually this means dV = 0.

19
3 First law of thermodynamics
3.1 Conservative and non-conservative forces
We start with a digression into Newtonian mechanics.

Conservative forces. Consider a particle at position x = (x, y, z) with potential energy V (x). The force
acting on the particle at x is
 
−∂x V (x)
F(x) = −∇V (x) = −∂y V (x) . (3.1)
−∂z V (x)

For example, Earth’s gravitational field is Vgr (x) = m0 gz, with m0 the mass of the particle, and
 
0
Fgr (x) =  0  . (3.2)
−m0 g
q1 q2
The electric potential between two electric charges q1 , q2 is Vel (x) = |x| , with x = x1 − x2 the relative
coordinate between the two charges, and
 
x
q1 q2  
Fel (x) = y . (3.3)
|x|3
z

Forces Fgr and Fel are examples of conservative forces, because they can be written as the gradient of some
function −V (x). Note that F = 0 is also a conservative force.

Mechanical work. We say that F(x) is a force field, as it specifies the force on a particle at position x. In
order to move a particle from x to x + dx in a force field F(x), we need to perform the amount of work

dW = F(x) · dx. (3.4)

This work could be supplied, for instance, mechanically (say, with our muscles) or electrically. To bring the
particle from position xA to xB , we can divide the path into K sufficiently small steps {∆x(1) , . . . , ∆x(K) }
according to

x(1) = xA , (3.5)
x(2) = xA + ∆x(1) , (3.6)
..
. (3.7)
(k+1) (k) (k)
x =x + ∆x , (3.8)
..
. (3.9)
(K+1)
x = xB . (3.10)

The work required to get from x(k) to x(k+1) is

∆W (k) = F(x(k) ) · ∆x(k) . (3.11)

The total work that needs to be performed to get from A to B is

K
X K
X
(k)
W = ∆W = F(x(k) ) · ∆x(k) . (3.12)
k=1 k=1

20
We could make the steps infinitesimal and write
 xB  xB
W = dW = F(x) · dx, (3.13)
xA xA

but the discrete version is absolutely sufficient for now.

ˆ Given two points xA and xB , there are many possible ways to get from one to the other. Each such way
corresponds to a different choice of sufficiently small steps {∆x(1) , . . . , ∆x(K) }, and hence, in principle
to a different outcome of the summation that gives W . The value of W , in general, depends on the
path we take from xA and xB .
ˆ The work
 xB
W = F(x) · dx (3.14)
xA

is independent of the path from xA to xB if and only if F(x) is a conservative force.

Non-conservative forces. Typical examples of non-conservative forces are friction, water drag, or air drag.
They cannot be written as the gradient of any function. For instance, a friction force acting on a particle
may have the form

F(x, ẋ) = η ẋ (3.15)

with some friction coefficient η.

Conservation of energy. A mechanical system is called conservative if all forces acting on it are conservative
forces. The total energy is then the sum of the kinetic and potential energy of all the particles at any given
time. It is determined by the dynamical state. The total energy of a conservative system is conserved.
Systems that feature non-conservative forces are called dissipative. Energy is still conserved, but only for the
total system including its environment. Mechanically, the heat that is created through friction corresponds
to disordered motion of particles. This case is captured by the 1st law of thermodynamics.

3.2 Energy and heat

The first law of thermodynamics is essentially the statement of the principle of conservation of energy for
thermodynamical systems. It may be expressed by stating that the variation in energy of a system during
any transformation is equal to the amount of energy that the system receives from its environment.

Consider first a conservative mechanical system of N particles. The system energy is the sum of the potential
and the kinetic energy, hence a function of the dynamical state. Assume A and B are two successive states
of an isolated system with energies EA and EB . If no external forces act on the system, energy remains
constant, and

EB = EA . (3.16)

When external forces act on the system, this equality need no longer hold. If W is the work performed by the
external forces during a transformation from the initial state A to the final state B, then the conservation
of energy implies

EB = EA + W. (3.17)

21
Importantly, the work done during the transformation only depends on the initial and final states, A and
B, not on the particular way in which the transformation from A to B is performed. This is only true for
conservative systems.

Conservative systems: The work W = WA→B in

EB = EA + W (3.18)

only depends on the initial and final thermodynamic states.

Assume that we find this property contradicted by an experiment performed on a particular thermodynamic
system. If we do not wish to discard the principle of conservation of energy, then we must admit the existence
of other methods, besides mechanical work, by means of which energy can be exchanged between the system
and its environment. Let us consider such an example.

Example. Heating of water

We consider a system composed of a quantity of water of volume V at atmospheric pressure P . We
consider two states A and B of the system with temperatures TA and TB such that TA < TB . To a
good approximation, these states have the same volume and pressure. [Strictly speaking, B will have
a slightly different volume. The volume of water decreases upon heating from 0◦ C to 4◦ C (anomaly of
water) and then increases when heated beyond 4◦ C (as in most liquids).]
1. First way. We heat the water by placing it over a flame and raise its temperature from the initial value
TA to the final value TB . The work performed by the system during the transformation is zero, since
it does not change its volume, W = 0.
2. Second way. We raise the temperature of the water from TA to TB by heating it by means of friction.
For this, we immerse a small set of paddles attached to a central axle into the water, and churn the
water by rotating the paddles. We observe that the temperature increases as long as the paddles
continue to rotate. This is because the water offers resistance to the motion of the paddles: we must
perform positive mechanical work W > 0 in order to keep the paddles moving until TB is reached.

The amount of work W in going from A to B depends on whether we go by means of the first or second way.
Assuming that the principle of conservation of energy holds for our system, then the energy transmitted to
the water in the form of mechanical work of rotating paddles in the second way must be transmitted to the
water in the first way in a non-mechanical form called heat. We are lead to the fact that heat and mechanical
work are equivalent in this example. They are two different aspects of the same thing, namely, energy.

First law of thermodynamics. We now put all this together into a quantitative formula. We first enclose
our system in a cylindrical container with a movable piston at one end. If the container walls are non-heat-
conducting, then the system is thermally isolated, and the exchange of energy between the system and the
environment can only occur in the form of mechanical work. The amount of work performed on the system,
W , when going from state A to state B only depends on the initial and final states of the transformation.
Denoting ∆E = EB − EA , the work satisfies

∆E = W. (3.19)
(First law of thermodynamics for thermally insulated systems)
If our system is not thermally insulated, this is generally not true, because there can be an exchange of energy

22
between the system and the environment in the form of heat. We then write the more general equation

∆E + L = Q, or (3.20)
∆E = W + Q, (3.21)
(First law of thermodynamics)
where Q is equal to zero for transformations performed on thermally insulated systems and otherwise, in
general, is different from zero.
The heat Q can be interpreted physically as the amount of energy that is received by the system in forms
other than work.
The first law of thermodynamics is a precise formulation of the equivalence of heat and work.

The work W appearing in the first law could also be electric or magnetic work, but we will mostly be dealing
with situations where W is mechanical work.

For a cyclic transformation, the first law takes on a very simple form. Since the initial and final states are
the same, we have ∆E = 0, thus

L = Q. (3.22)
(First law of thermodynamics for cyclic processes)
That is, the work done by a system during a cyclic transformation is equal to the heat absorbed by the system.

The SI unit for energy is Joule,

kg m2
1J=1Nm=1 . (3.23)
s2
In the cgs system, the unit of energy is erg, which is obtained from Joule by replacing kg by g, and m by
cm, hence
g cm2
1 erg = 1 = 10−7 J. (3.24)
s2
In the context of power consumption, energy is measured in kWh, with
J
1 kWh = 103 × 3600s = 3.6 × 106 J = 3.6 MJ. (3.25)
s
We used the SI unit of power, Watt, 1 W = 1 J/s. The average electric energy consumption of a North-
American household is ∼ 30 kWh per day. Another unit, historically defined as the unit of heat, is the
calorie, 1 cal. Since we know that heat is a form of energy, this means that 1 cal must correspond to a
certain amount of Joule. Several definitions of calorie exist, but they all roughly corresponds to the amount
of heat needed to heat 1 gram of water at standard atmospheric pressure by 1◦ C. We have
1 cal = 4.2 J. (3.26)
Note that 1 kcal = 1000 cal. The ”calorie” used to specify the nutritional value of food is actually kcal.
Most adults require a diet of about 2000 kcal per day, which amounts to
2000 kcal = 8.4 MJ = 2.3 kWh. (3.27)

23
The energy E is an extensive quantity. The energy density E/V and energy per particle E/N are intensive.

3.3 Differentials in thermodynamics

In thermodynamics, we often imagine infinitesimal processes whereupon a quantity, say energy E, changes
from a state A to an infinitesimal close (but distinct!) state A′ with energy E ′ = E + dE. If we consider a
system whose state can be defined by any two of the three variables V, P, T (keeping N constant as usual),
then energy, like any other function of the state, is a function of these two chosen variables. This means
that, strictly speaking, there are three functions called ”energy” in the game,

E(P, V ), E(P, T ), E(V, T ), (3.28)

all representing the same physical observable but through different variables.

Consider a function f (x, y) of two variables x and y. A small change

x → x′ = x + dx, (3.29)
′
y → y = y + dy (3.30)

implies a change of f according to

∂f ∂f
f ′ = f (x + dx, y + dy) = f (x, y) + (x, y) · dx + (x, y) · dy +higher orders in dx or dy. (3.31)
∂x ∂y
| {z }
df

We write this as

f ′ = f + df (3.32)

with

∂f ∂f
df = dx + dy. (3.33)
∂x ∂y
So far so good.

In thermodynamics, we face the problem that we have three functions E(P, V ), E(P, T ), E(V, T ). So if we
quantify the (physically observable!) change E → E ′ = E + dE, there are three potential differentials

∂E ∂E
(i) dE = dP + dV, (3.34)
∂P ∂V
∂E ∂E
(ii) dE = dP + dT, (3.35)
∂P ∂T
∂E ∂E
(iii) dE = dV + dT. (3.36)
∂V ∂T
The differentials dE in (i)-(iii) are all identical ; they describe the same physical change in energy dE = E ′ −E,
but simply express dE in different variables. There is an ambiguity, however, if we simply write
∂E
, (3.37)
∂T
because we do not know if this expression corresponds to the one appearing in (ii) or the one in (iii). We
will later see that these two expressions are generally different.

24
To resolve the ambiguity, we enclose the partial derivative symbol in a parenthesis and place the variable
that is to be held constant at the foot of the parenthesis. Thus
 ∂E 
(3.38)
∂T V

means the derivative of E with respect to T while keeping V constant, when T and V are taken as the
independent variables. This is in general different from
 ∂E 
, (3.39)
∂T P

in which case the pressure is kept constant. The proper way to write equations (i)-(iii) is then

 ∂E   ∂E 
(i) dE = dP + dV, (3.40)
∂P V ∂V P
 ∂E   ∂E 
(ii) dE = dP + dT, (3.41)
∂P T ∂T P
 ∂E   ∂E 
(iii) dE = dV + dT. (3.42)
∂V T ∂T V

Side comment. Consider a function F (x) and its derivative

dF
F ′ (x) = (x). (3.43)
dx
A fundamental theorem of calculus states that
 x2
F ′ (x)dx = F (x2 ) − F (x1 ). (3.44)
x1

We can write this as

 x2
dF
dx = F (x2 ) − F (x1 ), (3.45)
x1 dx

or, canceling the factors dx,

 F2
dF = F2 − F1 . (3.46)
F1

An expression dF that satisfies such an equation is called an exact differential, and F is called a function of
state in the thermodynamic context. Examples of functions of state are V, T, P, E, because they all satisfy
 V2
dV = V2 − V1 , (3.47)
V1
 T2
dT = T2 − T1 , (3.48)
T1
 P2
dP = P2 − P1 , (3.49)
P1
 E2
dE = E2 − E1 . (3.50)
E1

25
Here, for instance, E1 and E2 are the energies in state 1 and 2. In contrast, dW and dQ are not exact
differentials and

2
W and Q
are not functions of state. Instead, dW and dQ are merely small quantities and
2
the integrals 1 dW and 1 dQ depend on the process or path we choose to connect two thermodynamic
states 1 and 2. We have

 2
dW ̸= W2 − W1 , (3.51)
1
 2
dQ ̸= Q2 − Q1 , (3.52)
1

because the object W1 or “work in state 1” does not exist, etc. Some authors, for this reason, introduce
other symbols like δW or δQ to emphasize this difference. We will keep using dW and dQ, always keeping
in mind what we just said.
Note also that
 E2
dE = E2 − E1 (3.53)
E1

implies that for a cyclic process along a closed loop C in the PV-plane, with E2 = E1 , we have

dE = 0. (3.54)
C

In contrast,

Wcycl = dW (3.55)
C

is the total work performed during one cycle. This is generally not zero, otherwise there would be no useful
engines.

3.4 Heat capacity and calorimeter

Consider a system described by the variables (V, P, T ). An infinitesimal transformation is a transformation
for which the two independent variables change by infinitesimal amounts. The first law of thermodynamics
for this transformation reads

dE = dW + dQ = −P dV + dQ. (3.56)
(First law of thermodynamics for infinitesimal processes)
If we choose T and V as independent variables, E changes according to
 ∂E   ∂E 
dE = dT + dV. (3.57)
∂T V ∂V T

Inserted into the first law, this yields

 ∂E   ∂E 
dT + dV = −P dV + dQ, (3.58)
∂T V ∂V T

or
 ∂E  h ∂E  i
(I) dQ = dT + + P dV. (3.59)
∂T V ∂V T

26
Similarly, taking T and P as independent variables we generally have
 ∂E   ∂E 
dE = dT + dP, (3.60)
∂T P ∂P T
 ∂V   ∂V 
dV = dT + dP (3.61)
∂T P ∂P T
which, together with the first law, yields
 ∂E   ∂E 
(II) dQ = dT + dP + P dV (3.62)
∂T P ∂P T
 ∂E   ∂E  h ∂V   ∂V  i
= dT + dP + P dT + dP (3.63)
∂T P ∂P T ∂T P ∂P T
h ∂E   ∂V  i h ∂E   ∂V  i
= +P dT + +P dP. (3.64)
∂T P ∂T P ∂P T ∂P T
Finally, taking V and P as independent variables we obtain
 ∂E  h ∂E  i
(III) dQ = dP + + P dV. (3.65)
∂P V ∂V P

Note, again, that all three expressions for dQ are identical, just expressed in terms of different variables.

Heat capacity. The heat capacity, C, of a body is defined as the ratio C = dQ , i.e. the infinitesimal
dT
amount of heat dQ absorbed by the body as its temperature is increased by the infinitesimal amount dT .
In experiments, of course, finite amounts ∆Q and ∆T are used to determine ∆Q/∆T . The unit of C is
kB ∝ J/K.

In general, the heat capacity of a body will be different according to whether the body is heated at constant
volume or constant pressure. We denote this by a subscript according to
 ∂Q 
CV = , (3.66)
∂T V
 ∂Q 
CP = . (3.67)
∂T P

At this point, equations (I) and (II) come in handy. They need to be compared with
 ∂Q   ∂Q 
(I) dQ = dT +dV, (3.68)
∂T V ∂V T
 ∂Q   ∂Q 
(II) dQ = dT + dP, (3.69)
∂T P ∂P T
which yields,

 ∂Q  (I)
 ∂E 
CV = = , (3.70)
∂T V ∂T V
 ∂Q  (II)  ∂E   ∂V 
CP = = +P . (3.71)
∂T P ∂T P ∂T P
The second term in CP accounts for the fact that when temperature is increased at constant pressure, the
system typically expands and thus performs work. Hence we need to supply additional heat to the system
to enable this expansion. In CV , this term is not present, because the volume is kept constant and so no
expansion occurs. Note that the difference between CP and CV is largest for gases, whereas liquids and
solids expand only moderately when heated.

27
Specific heat capacity. The heat capacities CV and CP are extensive quantities. To get an intensive variable,
it is common to divide by the total mass M = m0 N of the body to obtain the specific heat capacities
CV CP
cV = , cP = . (3.72)
M M
While the unit of C is J/K, the unit of c is J/(kg K). The ratio
CP cP
κ= = (3.73)
CV cV
is called heat capacity ratio. cP and cV are examples of material constants, because they encode information
about a specific substance. They are measured and tabulated for most substances. We will meet more
material constants later.

Calorimeter. A calorimeter is a perfectly thermally isolated container containing two bodies in physical
contact with each other, whose individual temperatures are monitored. The device can be used to measure
heat in general or the specific heat capacity of one body if it is known for the other body. Typically we
have a solid object, initially at temperature T1 , immersed into a caloric liquid such as water, initially at
temperature T2 . As the two bodies reach thermal equilibrium, they acquire the same temperature T3 . Let
us assume that initially T1 > T2 . Then the water is heated during equilibration from T2 to T3 . If we neglect
the expansion of water upon heating, then the water absorbs the heat
(wat)
Q = Mwat cP (T3 − T2 ), (3.74)
(wat)
where Mwat is the total mass of water and cP = 4.182 J/(gK) is the specific heat capacity of water. To
(sol)
measure the specific heat capacity cP of the solid, we neglect the volume contraction of the solid upon
cooling, so that −Q is released as heat from the solid into the water. We have
(sol)
−Q = Msol cP (T3 − T1 ), (3.75)
hence
(sol) Mwat (T3 − T2 ) (wat)
cP = c . (3.76)
Msol (T1 − T3 ) P

3.5 Application of the first law to gases

Recall that we earlier said that the expression
∂E
(3.77)
∂T
makes no sense, as we need to indicate whether V or P is held fixed. However, for an ideal gas, both
expressions actually agree,
 ∂E  ideal gas  ∂E 
CV = = . (3.78)
∂T V ∂T P

When expressed in terms of the variables (V, T ) or (P, T ), the energy of an ideal gas is a function of T alone,

E(V, T ) = E(P, T ) = E(T ). (3.79)

As usual, we consider N as a constant. Another way to write this is

dE = CV dT. (3.80)
This property of ideal gases can be derived from statistical mechanics, and is approximately still valid for
real gases. Here we derive this result from an experimental observation.

28
 T T

A B A B

Figure 3.1: Expansion of an ideal gas does not change the temperature of the water in the calorimeter.

Experiment. Expansion of an ideal gas

We consider a calorimeter, i.e. a perfectly thermally isolated container filled with water. Into it we
place another container having two chambers, A and B, connected by a tube. Chamber A is filled with
a gas and chamber B is evacuated, the two chambers are initially shut off from each other by a closed
valve in the connecting tube.
1. First we wait sufficiently long until thermal equilibrium between the chambers and the water has set
in. For this we monitor the temperature of the water with a thermometer. If the temperature remains
constant over an extended period of time, we are in equilibrium.
2. We then open the valve to connect the two chambers, thus permitting the gas to flow from A into B
until the pressure everywhere in the chambers is the same.
3. We observe only a slight change in the reading of the thermometer.

The observation made in 3 implies that there had been practically no transfer of heat from the caloric water
to the chambers or vice versa. It is assumed that if this experiment could be performed with an ideal gas,
there would be no change at all.

We now apply the first law to the transformation. Since we observe Q = 0 between the gas and the water,
the gas enclosed in the chambers behaves like a thermally isolated system and we have
∆E + L = 0. (3.81)
Since the total volume of the chambers does not change (no movable walls), the gas cannot perform any
work, and so L = 0. Therefore,
∆E = 0. (3.82)
The energy of the gas does not change.

During the transformation, however, the volume of the gas changed from A to A + B. Since there was
no variation in energy during the process, we conclude that a variation in volume at constant temperature
produces no variation in energy,
 ∂E 
= 0. (3.83)
∂V T

29
In order words, the energy of an ideal gas is a function of the temperature only and not a function of the
volume. Similarly, the experiment showed that the change in pressure during the isothermal expansion had
no effect on the energy, so
 ∂E 
= 0. (3.84)
∂P T
This completes our proof.

Implication 1. We first determine the form of the function E(T ). We make use of the additional experimental
fact that CV of a gas only slightly depends on temperature. Integrating the equation
dE = CV dT (3.85)
we obtain

E(T ) = CV T. (3.86)
We have set the integration constant E0 , which would be the energy at T = 0, to zero.

Implication 2. We next compute CP for the ideal gas. Recall that

 ∂E   ∂V 
CP = +P . (3.87)
∂T P ∂T P
The first term is identical to CV for an ideal gas. For the second term we use P V = N kB T to arrive at

 ∂E  ∂  N kB T 
CP = +P (3.88)
∂T V ∂T P
= CV + N kB . (3.89)

Using kinetic theory one can show that

3
CV = N kB for a monoatomic gas, (3.90)
2
5
CV = N kB for a diatomic gas. (3.91)
2
For an ideal gas we then have

3
E(T ) = N kB T for a monoatomic gas, (3.92)
2
5
E(T ) = N kB T for a diatomic gas. (3.93)
2
Expressed in terms of (P, V ) this becomes
3
E(P, V ) = P V for a monoatomic gas, (3.94)
2
5
E(P, V ) = P V for a diatomic gas. (3.95)
2
We further have
5
CP = N kB for a monoatomic gas, (3.96)
2
7
CP = N kB for a diatomic gas, (3.97)
2

30
and
5
κ= for a monoatomic gas, (3.98)
3
7
κ = for a diatomic gas. (3.99)
5

3.6 Adiabatic transformations of a gas

A transformation is called adiabatic if the following two conditions are satisfied:

1. The transformation is reversible.

2. The system is thermally insulated so that no heat can be exchanged between the system and the
environment.
Condition 2 implies that dQ = 0 during the transformation and that the first law reads ∆E + L = 0 for
adiabatic processes.

We can expand or compress a gas adiabatically by enclosing it in a cylinder with non-heat-conducting walls
(condition 2) and a movable piston on one end, and very slowly shifting the piston inward and outward.
In this case, the gas molecules have enough time to adjust to the small gradual volume changes during the
process, and we effectively are in an equilibrium state at any time, hence the transformation is reversible
(condition 1).

Now would be a good point to recall the reversible run of the isothermal expansion.

A gas that expands adiabatically does external work, so L is positive in the first law. Hence, ∆E must be
negative, and the energy of a gas decreases during an adiabatic expansion. Since the energy is related to the
temperature through E = CV T for an ideal gas (or a similar relation for real gases), a decrease in energy
means a decrease in temperature of the gas.

Using dE = CV dT for ideal gases we have

CV dT + P dV = 0. (3.100)

Using the equation of state P V = N kB T , we can eliminate the pressure to obtain

N kB T
CV dT + dV = 0, (3.101)
V
or
dT N kB dV
+ = 0. (3.102)
T CV V
Integration yields
N kB
ln T + ln V = const. (3.103)
CV
We can write this as
N kB
TV CV
= const, (3.104)

or, using CP = CV + N kB and κ = CP /CV , as

31
 T V κ−1 = const. (3.105)
(Adiabatic transformation of an ideal gas)
Using P V = N kB T we readily find that

P V κ = const., (3.106)
κ 1−κ
T P = const. (3.107)

Note that in a PV-diagram, an isothermal transformation is given by P V = const. Hence, a curve corre-
sponding to an adiabatic transformation is steeper than an isotherm in the PV-diagram due to κ > 1.

32
4 Second law of thermodynamics
4.1 Perpetuum mobile
A perpetuum mobile (Latin for ”perpetual mover”) or perpetual motion machine is an impossible machine
that operates in cycles and, once started, would run forever and do work.
The first law of thermodynamics rules out the possibility of constructing a perpetuum mobile.
For one, in a thermally isolated system, the energy does not change in a cycle, and due to the 1st law the
work done is zero. When including the environment into the setting, any machine supplied with a finite
amount of energy at a given time will eventually stop because over time energy is lost to heat due to friction.
The 1st law forbids a machine that creates energy, which it could use to compensate this friction loss or do
work.

The first law does not limit the possibility of transforming one form of energy into the other, such as trans-
forming work into heat and heat into work, as long as the total amount of energy remains constant. Indeed,
we can always transform work into heat, for instance, by using friction to turn mechanical work into an equal
amount of heat, or using electric currents to heat up a resistance element, thus transforming electrical work
into heat.
Work can, in principle, to 100% be converted into heat.

However, one experimentally observes definite limitations to the possibilities of transforming heat into work.
If this were not the case, we could construct a machine that continuously absorbs heat from its environment
and uses this heat to perform work. Since the amount of thermal energy in the soil, water, atmosphere is
practically unlimited, such a machine would effectively be operating like a perpetuum mobile. Therefore, it
is called a perpetuum mobile of the second kind.
The second law of thermodynamics rules out the possibility of constructing a perpetuum mobile of the second
kind. It implies limits on what fraction of heat can (at best) be converted into work.

Some examples of perpetuum mobiles of the second kind: Drive a ship across the ocean only by converting
heat extracted from the water into work of the propellers. Light a light bulb or heat a pot by extracting
heat from the surrounding air.

4.2 Carnot cycle

We defined a heat bath or heat reservoir as a large body of temperature T that contains so many particles
that its temperature does not change when it is coupled to a system. It is used to bring and keep the system
to the temperature T through thermal contact and heat transfer.

In principle, the heat bath could exchange both heat Q and work W with the system. If the heat bath is
such that it cannot perform work on the system, it is called a heat source.

A system may be coupled to several reservoirs, or to different reservoirs at different times. The Carnot cycle
uses two heat sources, with temperatures T1 (lower) and T2 (higher). To memorize the lower and higher
temperatures in the following, use that 2 > 1. It is a reversible cycle built from two isothermal lines and
two adiabtic lines, intersecting at four points A, B, C, D. We use variables (V, T ) to label these states.

33
P V

V T
Figure 4.1: Carnot cycle, schematic, in the PV-diagram (left) and VT-diagram (right)

Carnot cycle
Consider a system whose initial state is A, with volume VA at temperature T2 . The Carnot cycle is
the following set of transformations that are supposed to be performed reversibly.
A → B. Isothermal expansion to volume VB at fixed temperature T2 .
B → C. Adiabatic expansion to volume VC . During that, temperature decreases to T1 .
C → D. Isothermal compression to volume VD at fixed temperature T1 .

D → A. Adiabatic compression to volume VA . During that, temperature increases to T2 .

Important for the Carnot cycle is that these transformations are performed reversibly, i.e. slow enough so
that the system is infinitesimally close to an equilibrium state at each intermediate step.

Example of a Carnot cycle. Consider a fluid enclosed in a cylindrical container with isolating side wall and
a movable isolating piston on the top. Heat can only be transferred through the base of the cylinder, which
we take to be heat-conducting. There are two heat sources, with temperatures T1 and T2 , such that T2 > T1 .
1. Initially the fluid is assumed to have volume VA and temperature T2 (corresponding to some pressure
PA via the equation of state).
2. Place the system on the heat source T2 . No heat transfer will occur initially.
3. Keeping the system on the heat source, we slowly (reversibly) raise the piston until the volume reaches
VB . The heat source ensures that the fluid remains at temperature T2 .
4. Remove the system from heat source T2 and place it on an insulator. We continue to slowly raise the
piston until we reach volume VC . The fluid is thermally insulated and cannot absorb heat, hence cools
to temperature T1 during that process.
5. Now place the system on heat source T1 . No heat transfer will occur initially.
6. Slowly press down the piston until we reach volume VD . The heat source ensures that the temperature
of the fluid remains T1 .
7. Remove the system from heat source T1 and place it on an insulator. Continue to slowly compress
until volume VA is reached. Since the fluid is thermally insulated and cannot release heat, it heats up
to temperature T2 during the process.

34
 insulator insulator

Figure 4.2: Example of a Carnot cycle discussed in the text

Heat and work. During the isothermal expansion A → B, the system absorbs an amount of heat Q2 > 0
from the source T2 . Similarly, during the isothermal compression C → D, it releases an amount of heat
Q1 > 0 to the source T1 (it absorbs an amount −Q1 from source T1 ). No heat is transferred in the adiabatic
sections. Hence the total amount of heat absorbed by the system during the cycle is Q2 − Q1 . The amount
of work done L during one cycle, given by the area enclosed in the PV-diagram, follows from the 1st law for
cycles to be
L = Q2 − Q1 . (4.1)
This equation means that only a fraction of the heat absorbed from the hot source at T2 is transferred into
work L in the Carnot cycle. The part Q1 , instead of being transformed into work, is surrendered to the
source at temperature T1 . We define the efficiency of the Carnot cycle as the ratio
L
η= . (4.2)
Q2
It represents the fraction of heat that is converted into work. We have
Q1
η =1− (4.3)
Q2
for the Carnot cycle, so the portion ∝ Q1 is not transformed into work.

Since the Carnot cycle is reversible, it can be run in the opposite direction, A → D → C → B → A. It then
absorbs the work L, and absorbs the amount of heat Q1 from source the colder T1 , whereas it gives up the
amount of heat Q2 to the hotter source T2 .

4.3 Statement of the second law

We state the 2nd law in two forms and prove their equivalence.

Second law of thermodynamics (Kelvin’s formulation)

There is no transformation whose only final result is to completely transform heat extracted from a
heat source at constant temperature into work.

35
The word ”only” is important here. For one, (i) it is possible to completely transform heat from a source
at constant temperature into work, provided some other change in the state of the system is present at the
end of the process. Furthermore, (ii) if some heat is released to another body during the process, then the
conversion of heat to work is not the only final result.

Example for (i). In the reversible run of the isothermal expansion of an ideal gas, an amount of heat Q
is supplied by the heat source to keep the gas at T . Since E(V, T ) = E(T ) for a gas, its energy does not
change, ∆E = 0. Consequently L = Q. The heat is completely transformed into work. However, this is not
the only final result of the process, because at the end of the process the gas occupies a larger volume than
at the beginning.
The fact that the initial and final state differ in this (or any similar) example, makes it impossible to use
such a transformation to form a cycle that is run several times, which would then constitute a perpetuum
mobile of the second kind.

Example for (ii). The Carnot process is a cycle, so initial and final states are the same, which extracts
heat Q2 from a source and performs work L. However, it only converts a fraction of the heat into work,
L = Q2 − Q1 , while it dumps the amount of heat Q1 into another body. Hence the conversion of heat into
work is not the only final result, but there is another final result, namely the rejection of some heat Q1 .
The coefficient η between extracted heat and performed work in L = ηQ2 is precisely the efficiency. Hence
Kelvin’s statement implies η < 1 for all cyclic processes.

The word ”completely” is, strictly speaking, superfluous and implied by ”only”, see example (ii). If the
conversion of heat into work is not complete, then there must always be some rejected heat as an additional
final result of the process.

The wording is to suggest that a positive amount of heat Q > 0 is absorbed by the system and converted
into a positive work done L > 0. If L < 0, i.e. work W > 0 is performed on the system, then the system
could convert this work via friction into negative heat Q = −W < 0 that is released into the environment.
We then have a situation where the only final result is that work is completely converted into heat. No
constraint arises in this case, because work can always be completely converted into heat.

Second law of thermodynamics (Clausius’s formulation)

There is no transformation whose only result is to transfer heat from a body at a given temperature
to a body at a higher temperature.

Clausius’s statement implies that some work must be performed in order to transfer heat from a colder body
to a hotter body. This work requires an external energy source, so that the supply of energy through work is
an additional final result of the process. Refrigerators, of course, operate in this way and are not forbidden
by the 2nd law of thermodynamics.

We prove the equivalence of Kelvin’s and Clausius’s statements. We use contraposition, i.e. to prove X ⇒ Y ,
we show not-Y ⇒ not-X.

not-Kelvin ⇒ not-Clausius. Assume Kelvin’s statement is wrong. We can then perform a transformation
which extracts an amount of heat from a source ”1” at temperature T1 , and whose only final result is to
completely transform this amount of heat into work. By means of friction, we could completely transform
this work into heat again, and use this heat to heat up a another body ”2” that is at any temperature T2 .
In particular, this would also work if T2 > T1 . The only final result of this whole process would then be to

36
transfer heat from a body ”1” at temperature T1 to a body ”2” at higher temperature T2 . Hence Clausius’s
statement is wrong.

not-Clausius ⇒ not-Kelvin. Assume Clausius’s statement is wrong. We can then transfer an amount of
heat Q2 from a heat source ”1” at temperature T1 to a body ”2” at higher temperature T2 . We next use a
Carnot process to absorb the same heat Q2 from source ”2” at the high temperature (during step A → B)
and transform it into work L. During the Carnot process, some amount of heat Q1 is dumped into ”1” at
the lower temperature. Since body ”2” absorbed and released the same amount of heat, there is no overall
change to body ”2” in the process. The net amount of heat extracted from ”1” is ∆Q = Q2 − Q1 , which
is completely transformed into work L = ∆Q. Hence we constructed a process which extracts an amount
of heat ∆Q from the source ”1” at constant temperature and transformed it completely into work. Thus
Kelvin’s statement is wrong.

Note that we have only shown the equivalence of both formulations of the 2nd law, we did not ”prove” it.
Its experimental justification comes from the fact that no-one ever accomplished to construct a perpetuum
mobile of the second kind, or observed heat to flow from a cold to a hot body by itself.

4.4 Heat engines and refrigerators

A machine that uses the flow of heat from a body at higher temperature T2 to a body at lower temperature
T1 to perform work is called a heat engine. Denote the amount of heat per cycle extracted from body ”2”
by Q2 and the amount of heat per cycle submitted to body ”1” by Q1 . The work performed per cycle is
L = Q2 − Q1 > 0. We define the efficiency of a heat engine by
L Q1
η= =1− . (4.4)
Q2 Q2

Crucially, the underlying cycle need not be reversible or a Carnot cycle. An engine using a reversible cycle
will be called reversible engine.
Note that η = L/Q2 is the fraction of heat converted into work. The closer η is to 100%, the better for
applications. Kelvin’s statement implies η < 1. In the following two sections, we show that the efficiency of
any reversible engine operating between T2 and T1 is identical, and satisfies

T1
ηrev = 1 − . (4.5)
T2
Typically the body at lower temperature, into which the amount of heat Q1 > 0 is dumped and lost, is the
environment of the engine, thus we usually cannot control T1 . In practice, it would be difficult to run an
engine with very low T1 → 0, since we continuously dump heat into ”1”. To get a large efficiency, we want
T2 to be as high as possible. Of course, any actual efficiency will generally be lower than ηrev , because all
actual heat engines are far from being reversible. In fact, we will show that

η ≤ ηrev (4.6)
in general.

We first show that if L > 0, then we have Q2 > 0 and Q1 > 0. Assume first that Q1 ≤ 0. This means the
engine absorbs the positive amount of heat |Q1 | = −Q1 from heat source ”1”. We could then use the hotter
source ”2” to heat up source ”1” to exactly compensate for the lost amount of heat |Q1 |. Indeed, this would
amount to heat flowing from a hot to a cold body, which is not forbidden by the 2nd law. During one cycle,
there would then be no change to source ”1”. Furthermore, the amount of heat effectively absorbed from
source ”2” by the engine is Q2 + |Q1 |, which is fully converted into work L = Q2 − Q1 = Q2 + |Q1 |. Since

37
 Figure 4.3: Assuming Q1 ≤ 0 in heat engines leads to a contradiction.

this is the only result of the cycle, this is in contradiction to Kelvin’s formulation of the 2nd law. Hence
Q1 > 0. Since L > 0, we immediately get Q2 > 0. □

In the following, we consider two engines that operate between the same temperatures T1 and T2 . The
first engine shall be characterized by (L, Q1 , Q2 ), the second engine by (L′ , Q′1 , Q′2 ). We show the following
fundamental theorem:
A. If the first engine is reversible, then

Q2 Q′
≥ 2′ . (4.7)
Q1 Q1

B. If both engines are reversible, then

Q2 Q′
= 2′ . (4.8)
Q1 Q1

Note that B implies that ηrev is identical for all reversible engines operating between T1 and T2 , whereas A
implies η ≤ ηrev . The second engine in A need not be reversible.

Statement A implies B, because if A is true and if the second engine is also reversible, then we obtain the
Q′2
relation with both ”≤” and ”≥, hence QQ1 = Q′ in this case.
2
1

It remains to show A. For both engines we have

L = Q2 − Q1 , (4.9)
′
L = Q′2 − Q′1 . (4.10)

The ratio Q2 /Q′2 can to arbitrary precision be approximated by a rational number

Q2 N′
′ = , (∗) (4.11)
Q2 N

where N, N ′ are positive integers. Now consider a combined cycle that consists of N ′ runs of the second
engine and N runs of the reversed first engine. (This is permissible, since the first engine is reversible.)
When operated in the reverse, the first engine absorbs an amount of work L, gives up the amount of heat

38
Q2 to source ”2”, and receives the amount of heat Q1 from source ”1”. The total amount of work done by
the combined engine is
Ltotal = N ′ L′ − N L, (4.12)
the total amount of heat received from source ”2” is
Q2,total = N ′ Q′2 − N Q2 , (4.13)
and the total amount of heat given to source ”1” is
Q1,total = N ′ Q′1 − N Q1 . (4.14)
This yields
Ltotal = Q2,total − Q1,total . (4.15)
However, equation (∗) implies Q2,total = 0 and so
Ltotal = −Q1,total . (4.16)
Importantly, at this point we do not know the sign of Ltotal (which is the sign of −Q1,total ). The combined
cycle is constructed such that there is no change to source ”2” after the whole cycle. If Ltotal > 0, then the
only final result of the combined cycle is that the heat absorbed from source ”1” at constant temperature
T1 (which is −Q1,total ) is converted into the work Ltotal . This contradicts Kelvin’s statement. However, if
Ltotal ≤ 0, then work is performed on the engine, and the engine converts this work completely into the
amount of heat |Q1,total | that is dumped into source ”1”. Nothing limits the conversion of work into heat.
Hence
Ltotal ≤ 0 (4.17)
and
Q 
2 ′
0 ≤ Q1,total = N ′ Q′1 − N Q1 = N Q − Q1 . (4.18)
Q′2 1
Hence
Q2 Q′
≥ 2′ , (4.19)
Q1 Q1
which proves A. □

A refrigerator is a heat engine run in the reverse direction, which extracts an amount of heat Q1 from a
source at low temperature T1 by absorbing work W > 0, and dumps an amount of heat Q2 to a source
at higher temperature T2 . (If the emphasis is on heating the hotter source, it is equivalently called a heat
pump.) The work W > 0 is required by Clausius’ formulation of the 2nd law. The coefficient of performance
is given by
Q1
COP = > 1. (4.20)
W
For a reversible refrigerator (such as a reverse Carnot cycle) we have
T1
COPrev = (4.21)
T2 − T1
For instance, for T2 = 20◦ C we have
T1 = 3◦ C ⇒ COPrev = 16, (4.22)
◦
T1 = −18 C (freezer) ⇒ COPrev = 6.7, (4.23)
whereas real household refrigerators have COP ∼ 2.

39
4.5 Zeroth law of thermodynamics
To formally define the concept of temperature, without resorting to kinetic theory or statistical mechanics,
one may use the zeroth law of thermodynamics. It is the following postulate that is confirmed by experiment:
When two thermodynamic systems are each in thermal equilibrium with a third system, then they are in
thermal equilibrium with each other.
Note the word ”thermal” here. This postulate implies that there is a property of thermodynamic systems
which is common to all systems in thermal equilibrium. This quantity serves as an empirical temperature
T̃ .

Let us call the systems A, B, C. For concreteness we may think of C as a thermometer. The volume and
pressure of each system are (VA , PA ), (VB , PB ) and (VC , PC ).
1. We assume that A and C are in thermal contact through a heat-conducting wall. For general values
of (VA , PA , VB , PB ), they would not be in thermal equilibrium. Instead, in thermal equilibrium there
will be a constraint of the form

f1 (VA , PA , VC , PC ) = 0 (4.24)

relating the four quantities through some function f1 . We want to show that this relation is actually
of the form

T̃A (VA , PA ) = T̃ = T̃C (VC , PC ). (4.25)

We could then define T̃ as temperature read off from the thermometer C. Furthermore, the relation
T̃ = T̃A (V, P ) constitutes the equation of state for system A, similarly for C.
2. To show Eq. (4.25), we also consider B and C, whose thermal equilibrium implies a relation of the
form

f2 (VB , PB , VC , PC ) = 0 (4.26)

with some other function f2 .

3. The zeroth law states that Eqs. (4.24) and (4.26) imply a third relation of the form

f3 (VA , PA , VB , PB ) = 0 (4.27)

for all values of the thermodynamic variables.

This is a little strange: If the equations were of the sort f1 (VA , PA , x) = 0 and f2 (VB , PB , x) = 0, then
we could solve the first equation for x = x0 (VA , PA ) and insert this into the second equation to obtain
f3 (VA , PA , VB , PB ) = f2 (VB , PB , x0 (VA , PA )) = 0. In the present case, however, we have two variables
VC , PC , but only one equation f1 = 0, so we can naively only eliminate one of the two variables. The
implication (4.24,4.26) ⇒ (4.27) can only be true if there is an A- and B-independent function

T̃C (VC , PC ) (4.28)

such that

0 = f1 (VA , PA , VC , PC ) = fˆ1 (VA , PA , T̃C (VC , PC )), (4.29)

0 = f2 (VB , PB , VC , PC ) = fˆ2 (VB , PB , T̃C (VC , PC )). (4.30)

(The functions f1,2 (◦, ◦, ◦, ◦) have four arguments, fˆ1,2 (◦, ◦, ◦) have only three arguments.) We can
then solve these two equations for T̃C to obtain both

T̃C (VC , PC ) = T̃A (VA , PA ) (4.31)

40
 and

T̃C (VC , PC ) = T̃B (VB , PB ), (4.32)

where we introduced two new functions T̃A,B . This yields the desired formula

T̃A (VA , PA ) = T̃B (VB , PB ) ⇒ f3 (VA , PA , VB , PB ) := T̃A (VA , PA ) − T̃B (VB , PB ) = 0 (4.33)

4. The three systems in thermal equilibrium all have the same value of

T̃ := T̃A (VA , PA ) = T̃B (VB , PB ) = T̃C (VC , PC ). (4.34)

This defines both the empiric temperature and the equations of state for systems A, B, C.

Comment 1. If we use an ideal gas thermometer in this construction as C, then

PV
T̃ = T̃C (V, P ) = = T, (4.35)
N kB
which agrees with the absolute temperature.

Comment 2. The derivation remains valid if the systems depend on more than the two variables V, P , say,
N1 , N2 , . . . for mixtures. The equation of state obtained would be of the form

T̃ = T̃A (P, V, N1 , N2 , . . . ). (4.36)

Note that the particle number N that we tend to ignore falls into this category.

Comment 3. The zeroth law is proven fairly easily within the framework of statistical mechanics, where T
is defined through T −1 = ∂E
∂S
|V,N . Thermal equilibrium of A and C implies TA = TC and that of B with C
implies TB = TC . This implies TA = TB .

4.6 Absolute temperature

The zeroth law of thermodynamics implies the existence of empirical temperature scales such as the column
length of a mercury thermometer. In principle, there are as many empiric temperatures as there are sub-
stances. We pick any such empiric temperature scale and denote it by T̃ in the following. Another empirical
temperature is the scale defined by an ideal gas thermometer, i.e. the temperature T defined through
PV
T = (4.37)
N kB
for an ideal gas. The two empirical temperatures can be converted into each other through calibration of
thermometers, i.e. there is a function T = T (T̃ ) and its inverse.

In this section, we show that the findings on heat engines imply the existence of an absolute thermody-
namic temperature scale, denoted ϑ, that is intrinsic to thermodynamics and thus independent of the special
properties of any thermodynamic substance. All thermodynamic laws take on a simple form when this tem-
perature scale is used. We will show that it coincides, up to an arbitrary constant factor a, with the ideal
gas temperature scale, i.e.

ϑ(T̃ ) = a · T (T̃ ), (4.38)

or, since T̃ was arbitrary,

ϑ = aT. (4.39)

41
We then choose a = 1 by convention and arrive at

ϑ = T. (4.40)

We start from statement B that for all reversible engines operating between (empirical) temperatures T̃1 and
T̃2 , the ratio Q2 /Q1 is identical. Thus
Q 
2
= f (T̃2 , T̃1 ) (4.41)
Q1 rev

for one cycle, with some fundamental function f (◦, ◦). Choose an arbitrary small temperature T̃0 such that

T̃0 < T̃1 < T̃2 (4.42)

and three corresponding heat sources ”0”, ”1”, and ”2”. We consider two more reversible engines, En1 and
En2 . During one cycle, engine En1 extracts the amount of heat Q1 from ”1” and releases the amount of heat
Q0 into ”0”. We have
Q 
1
= f (T̃1 , T̃0 ). (4.43)
Q0 rev
In contrast, engine En2 extracts the amount of heat Q2 from ”2” and releases Q0 into ”0”. We have
Q 
2
= f (T̃2 , T̃0 ). (4.44)
Q0 rev
Taking the ratio of these two equations we conclude

f (T̃2 , T̃0 )
f (T̃2 , T̃1 ) = . (4.45)
f (T̃1 , T̃0 )
(More accurately, we would have to consider the combined process of running one cycle of engine En2
followed by one cycle of the reversed engine En1 . The outcome is the same.) Since T̃0 was arbitrary, we may
keep it constant in all our equations. Hence we regard f (T̃ , T̃0 ) as a function of T̃ alone. We define a new
temperature scale

ϑ(T̃ ) = bf (T̃ , T̃0 ) (4.46)

with some constant b. We then arrive at

Q  ϑ(T̃2 ) ϑ2
2
= = . (4.47)
Q1 rev ϑ(T̃1 ) ϑ1

for all reversible cycles. The temperature scale defined by the variable ϑ is called the absolute thermodynamic
temperature scale. We are free to choose the unit of this scale, since only ratios appear in this equation. The
usual choice is made by setting the difference between the boiling temperature and freezing temperature of
water at atmospheric pressure to be 100 units (#).

We finally show that ϑ = aT by computing Q2 /Q1 for the Carnot process of an ideal gas, which is a reversible
process. The outcome is

Q  T2
2
= ⇒ ϑ = aT, (4.48)
Q1 Carnot T1

with an arbitrary scale factor a. The way we have chosen the units for ϑ in (#), however, implies a = 1.

42
For the computation of Q2 /Q1 for the Carnot process we consider an ideal gas with heat capacity ratio κ.
The work done along the isothermal lines is
V 
B
LA→B = N kB T2 ln > 0, (4.49)
VA
V 
C
LC→D = −N kB T1 ln < 0. (4.50)
VD
No heat is absorbed along the adiabatic lines and we have

Q2 = QA→B > 0, (4.51)

−Q1 = QC→D < 0. (4.52)

Now now use that the operating substance is an ideal gas and so the energy E(V, T ) = E(T ) only depends
on the temperature. Thus points along isotherms have the same energy,

EA = EB , (4.53)
EC = ED . (4.54)

This implies that work equals heat along the isothermal lines,
V 
B
Q2 = LA→B = N kB T2 ln > 0, (4.55)
VA
V 
C
Q1 = −LC→D = N kB T1 ln > 0. (4.56)
VD

To relate the volumes, we use that along the adiabatic lines we have T V κ−1 = const so that
T2  V κ−1  V κ−1 VC VB
C D
= = ⇒ = . (4.57)
T1 VB VA VD VA
For the heat ratio we eventually obtain

Q2 T2 ln( VVB
A
) T2
= V
= . (4.58)
Q1 T1 ln( VD )
C T1

This concludes the derivation of ϑ = T .

4.7 Clausius inequality

We consider a system S that undergoes a cyclic transformation. We suppose that during the cycle the system
receives heat from or surrenders heat to a set of n heat sources having temperatures T1 , T2 , . . . , Tn . Denote
the amounts of heat exchanged between the system and the sources by Q1 , Q2 , . . . , Qn . We take Qi positive
if heat is received by the system and negative in the other case.

We prove that

n
X Qi
≤ 0. (4.59)
i=1
Ti
(Clausius inequality)

The equality sign holds if the cycle is reversible.

43
 Figure 4.4: Combined cycle used in the proof of Clausius’s inequality

Example. Let us consider the case of heat engines operating between two sources with T1 and T2 , i.e. n = 2.
We have shown that
T2 Q 
2 Q2
= ≥ . (4.60)
T1 |Q1 | rev |Q1 |
Note that Q1 = −|Q1 | < 0 with the sign convention used here. This implies
T2 |Q1 | ≥ T1 Q2 ⇔ 0 ≥ T1 Q2 − T2 |Q1 | = T1 Q2 + T2 Q1 . (4.61)
Diving by T1 T2 we obtain
Q1 Q2
+ ≤0 (4.62)
T1 T2
as claimed. The same derivation also shows that equality holds for reversible cycles.

Proof. Consider an additional heat source ”0” at an arbitrary temperature T0 . Between the source at T0 and
each of the n heat sources at Ti we insert a reversible Carnot engine Ci . The Carnot engine Ci is such that if
the system receives (surrenders) the amount of heat |Qi | to the source ”i”, then Ci surrenders (receives) the
amount of heat |Qi | to the source ”i”. So if Qi is the heat absorbed by the system at Ti , then −Qi is the
heat absorbed by the Carnot engines at Ti . After running the combined cycle of S followed by C1 , . . . , Cn ,
the n heat sources remain unchanged.

Each Carnot engine Ci , schematically, is a heat engine that receives a positive amount of heat Q(a) > 0 at
T (a) and releases a positive amount of heat Q(b) > 0 at T (b) , with
Q(a) T (a)
= . (4.63)
Q(b) T (b)

44
In this case, |Qi | at Ti is one of the positive numbers. The other, |Qi,0 |, is the amount of heat exchanged
between Ci and the heat source ”0”. Thus
T0
|Qi,0 | = |Qi |. (4.64)
Ti
Denote by Qi,0 the amount of heat absorbed by Ci from ”0”. If Qi > 0 (system receives heat at Ti and Ci
releases heat at Ti ), then Qi,0 > 0. If Qi < 0 (system releases heat at Ti and Ci receives heat at Ti ), then
Qi,0 < 0. Thus Qi and Qi,0 have the same sign,
T0
Qi,0 = Qi . (4.65)
Ti

Now we run the combined cycle of S followed by C1 , . . . , Cn . The source at T0 surrenders a total amount of
heat
n n
X X Qi
Q0 = Qi,0 = T0 (4.66)
i=1
T
i=1 i

to the Carnot engines. Since the total system consisting of S, C1 , . . . , Cn returns to its initial state, the 1st
law for cyclic processes applies, and the amount of heat Q0 is converted into work done given by L = Q0 .
Since this is the only final result of the cycle (the heat sources at T1 , . . . , Tn remain unchanged), we conclude
that L ≤ 0, since otherwise Kelvin’s statement would be violated. Hence Q0 ≤ 0, or
n
X Qi
≤ 0. (4.67)
i=1
Ti

If the cycle is performed reversible, we can run it in the reversed direction, which means that all Qi change
sign (but Q0 does not). We then have
n 
X Qi 
− ≤ 0, (4.68)
i=1
Ti rev
which together with the Clausius inequality implies
n
X Q  i
= 0. (4.69)
i=1
Ti rev

This completes the proof of the theorem. □

We assumed that the number of sources at T1 , . . . , TN is finite. It is important to also consider

 the case
for which the system exchanges heat with a continuous distribution of sources. Denoting by the integral
extended over a cycle, say in the PV-diagram, and by dQ the infinitesimal amount of heat received by the
system from a source at temperature T , we have

dQ
≤ 0, (4.70)
T
which is valid for all cycles, and

dQ
= 0, (4.71)
T
which is valid only for reversible cycles.

Comment. Note that T in the denominator is the temperature of the source. This coincides with the system
temperature for reversible cycles, where system and source are in equilibrium at each point. For irreversible
processes, however, heat can only flow from hot to cold bodies, so the source is hotter for dQ > 0 or colder
for dQ < 0.

45
4.8 Entropy
The Clausius (in)equality for reversible cycles reads

dQ
= 0, (4.72)
C T
where C is a reversible transformation, for instance a closed loop in the PV-diagram. Consider now two
equilibrium states A and B of a system and some reversible transformation Prev that takes the system from
A to B. The value of the integral

dQ
IA→B = (4.73)
Prev T
′
is independent of the reversible path chosen. Indeed, if Prev is another reversible path connecting A and B,
′−1
then we denote by Prev the reversed path connecting B and A. We have

′ dQ
IA→B = (4.74)
′
Prev T
and
 
′ dQ dQ ′
IB→A = =− = −IA→B . (4.75)
′−1
Prev T ′
Prev T
(This is true for any line-integral: Going along the path in one or the other direction gives an overall minus
′−1
sign.) Consider now the combined reversible transformation C = Prev ∪ Prev . This is a reversible cycle and

dQ
0=
T
C 
dQ dQ
= +
Prev T Prev T
′−1

′
= IA→B + IB→A
′
= IA→B − IA→B . (4.76)
This proof our claim.

Entropy. The property just shown allows us to define a new function of state of the system, the entropy. It
is one of the most important quantities in thermodynamics and, in fact, most areas of physics. We choose
an arbitrary equilibrium state O of our system. For any other equilibrium state A we define the entropy of
state A as

 A
dQ
SA = , (4.77)
O T
where the path connecting O and A is taken to be reversible. We have seen that the value of SA is indepen-
dent of the reversible path chosen. Since the path is reversible, its initial and final points must be equilibrium
states. Consequently, SA is only defined for equilibrium states A. If we choose another equilibrium state B,
then
 B
dQ
SB = (4.78)
O T
along some reversible path connecting O and B. Since we have
 A  O
dQ dQ
=− , (4.79)
O T A T

46
this implies that the entropy difference between states A and B is given by

 B  A  O  B  B
dQ dQ dQ dQ dQ
SB − SA = − = + = , (4.80)
O T O T A T O T A T
where any reversible path connecting A and B can be chosen. For an infinitesimal reversible transformation
we have

dQ = T dS (reversible). (4.81)

B
In the context of differentials introduced in Section 3.3, the heat dQ is not an exact differential and A
dQ
depends on the path. In contrast, the entropy
1
dS = dQ (reversible) (4.82)
T
is an exact differential and
 B
dS = SB − SA (4.83)
A

is independent of the reversible path chosen to connect A and B. Similarly, the work dL is not an exact

B
differential and A dL depends on the path, but the volume

1
dV = dL (reversible) (4.84)
P
is an exact differential and
 B
dV = VB − VA (4.85)
A

is independent of the reversible path chosen to connect A and B. One might say that the existence of the
function of state ”entropy” restores a symmetry between work and heat in these equations.

J
Comment 1. The natural unit of entropy is kB , because Q ∝ J, T ∝ K, so S ∝ K ∝ kB .

Comment 2. The entropy is an extensive variable.

Some properties of the entropy. We show the following important properties of the entropy:

47
 (1) The entropy of independent subsystems is additive. This implies that entropy is extensive.

(2) For any thermodynamic transformation connecting equilibrium states A and B we have
 B
dQ
∆S = SB − SA ≥ , (4.86)
A T

where equality holds for reversible transformations.

(3) For a thermally isolated system, we have

SB ≥ SA (4.87)

for any transformation connecting A and B.

(4) For a thermally isolated system, entropy does not change during reversible transformations, SA = SB .
(5) For a thermally isolated system, the equilibrium state is the state of maximum entropy consistent with
the external constraints.
For (1), we need to define what independent systems are. For this we assume that a thermodynamic system
S splits into several subsystems S1 , . . . , Sn which are in equilibrium each (but not necessarily with each
other). We say the subsystems are independent if they are so large that the energy of the total system can
be split into the energies of the subsystems,
E = E1 + · · · + En , (4.88)
and the work done by the system in any transformation is equal to the work done by the subsystems,
W = W1 + · · · + Wn . (4.89)
We then have Q = Q1 + . . . Qn and entropy is additive according to
 A  A  A
dQ dQ1 dQn
S= = + ··· + = S1 + · · · + Sn . (4.90)
O T O T O T
This allows us to define the entropy for certain non-equilibrium systems, for instance for inhomogeneous
systems with spatially varying thermodynamic parameters. An example where E ̸= E1 + · · · + En is, for
instance, an interface between two substances that has a significant surface energy.
For (2), consider two equilibrium states A and B. Let P be any transformation connecting A and B, and
P ′ be a reversible transformation connecting A and B. The transformation C = P ∪ P ′−1 is a cycle and we
have

dQ
0≥ (4.91)
T
C 
dQ dQ
= + (4.92)
T ′−1 T
P P
dQ dQ
= − (4.93)
T P T
′
P
dQ
= − [SB − SA ], (4.94)
P T

thus
  B
dQ dQ
SB − SA ≥ = (4.95)
P T A T
as claimed.
For (3), consider a thermally isolated system. Since it cannot exchange heat with its environment we have

48
dQ = 0 and thus SB − SA ≥ 0 as claimed.
For (4), use that ”=” applies in (3) for reversible transformations.
For (5), consider a thermally isolated system subject to some external constraints such as fixed volume
V . The system may be in one of many (possibly inhomogeneous) thermodynamic states that satisfy these
constraints.2 If the system is not in equilibrium, due to the microscopic motion of atoms or molecules, it
will, over time, undergo spontaneous thermodynamic processes that slightly change its thermodynamic state.
Any such transformation, however, needs to satisfy ∆S ≥ 0 according to (3). Consequently, if we wait long
enough for the time-independent equilibrium state to set in, then this state must be a state of maximum
entropy.

Comment 1. The entropy of a system can decrease in a thermodynamic transformation. This requires
that the system is not thermally isolated from its environment, which is often the case. However, if the
environment and the system together are thermally isolated from the rest of the universe, then the entropy
of the combined system cannot decrease.

Comment 2. From (4) and (5) it follows that the approach to equilibrium is an irreversible process. This is
intuitively clear.

Example 1. Reversible isothermal expansion of an ideal gas. Assume the volume increases from VA to VB .
Since T does not change and E = E(T ), we have ∆E = 0 and

VB
Q = L = N kB T ln . (4.96)
VA
Consequently, for the system,
 B  B
dQ 1 Q VB
∆Ssys = = dQ = = N kB ln . (4.97)
A T T A T VA

The heat Q is supplied by the environment and we have

Q VB
∆Senv = − = −N kB ln , (4.98)
T VA
so that

∆Ssys + ∆Senv = 0. (4.99)

Example 2. Free isothermal expansion of an ideal gas. This is an irreversible process. As we have seen in
the section on heat capacity and calorimeters, no heat is supplied

by the environment and no work is done,
Q = W = 0. We may naively assume that Q = 0 implies ∆S ∼ dQ T = 0, but this is wrong. Indeed, we can
only conclude that
 B
dQ
∆Ssys ≥ = 0, (4.100)
A T
2 The typical example is a gas that occupies only one half-space of a room: a clear non-equilibrium state, although consistent

with external constraints.

49
 T

S
Figure 4.5: Carnot cycle in the TS-diagram

which does not rule out ∆S ̸= 0. In fact, since entropy is a function of state and the initial and final states,
A and B, are the same equilibrium states for both the reversible and free expansion, we have
VB
∆Ssys = SB − SA = N kB ln (4.101)
VA
in both cases. The free expansion is thus an example of a process with dQ = 0, but dS ̸= 0. This is no
contradiction since we only have dQ = T dS for reversible processes. For the free expansion, however, we
have

∆Senv = 0 (4.102)

for the environment, as it does not participate in the transformation, and so

VB
∆Ssys + ∆Senv = N kB ln > 0. (4.103)
VA

Carnot cycle in the TS-diagram. It is instructive to study the reversible Carnot cycle in the TS-diagram,
instead of the PV-diagram. During the adiabatic transformations B → C and D → A, we have dQ = 0,
and thus dS = 0. These correspond to vertical lines in the TS-diagram. The isothermal lines A → B and
C → D are horizontal lines. The net amount of heat extracted from the environment, due to dQ = T dS in
reversible processes, is given by the area under the curve. We have

L = Q2 − Q1 = Qnet = (T2 − T1 )(S2 − S1 ). (4.104)

Furthermore,

Q2 = T2 (S2 − S1 ), (4.105)

so that
L T1
η= =1− . (4.106)
Q2 T2

50
Differentials of energy and entropy. Consider a state in thermodynamic equilibrium, represented, for in-
stance, by a point in the PV-diagram. Choose any infinitesimally close equilibrium state and a reversible
transformation connecting them. The first law of thermodynamics then reads

dE = dQ + dW = T dS − P dV. (4.107)

Recall that we read this as

 ∂E   ∂E 
T = , P =− . (4.108)
∂S V ∂V S

Equivalently,
1 P
dS = dE + dV, (4.109)
T T
hence
1  ∂S  P  ∂S 
= , = . (4.110)
T ∂E V T ∂V E
These relations are true for any thermodynamic equilibrium state.

51
5 Thermodynamics potentials
5.1 Free energy
In this section, we consider a system that is in thermal contact with a heat bath (environment) of constant
temperature T . Work W may be performed on the system. We also assume for simplicity that the system
is homogeneous.

Thermal equilibrium. Assume the system with energy E and entropy S is in thermal equilibrium with the
heat bath at T , so that the system also has temperature T . We define the free energy by

F = E − T S. (5.1)
The free energy is a function of state, since E, T, S are functions of state. Consider now a thermodynamic
transformation from state A to state B and define
∆F = FB − FA . (5.2)
We show that:
The work done during the transformation satisfies

L ≤ −∆F, (5.3)

or, perhaps more memorably,

Lmax = −∆F. (5.4)

Equality holds for reversible transformations. In words: For a system in thermal equilibrium with a heat
bath at constant temperature, the work done during a reversible transformation is equal to the decrease in
free energy of the system.
For the proof, recall that
 B
dQ
≤ SB − SA , (5.5)
A T
where equality holds for reversible transformations. Since the temperature of the bath T is constant during
the transformation, we have
 B
Q= dQ ≤ T (SB − SA ). (5.6)
A
From the first law of thermodynamics we have
L = −∆E + Q (5.7)
≤ EA − EB + T (SB − SA ) (5.8)
= [EA − T SA ] − [EB − T SB ] (5.9)
= FA − FB (5.10)
= −∆F. □ (5.11)

Thermal contact only. If the system is not in thermal equilibrium with the bath, but rather in a state A of
temperature TA , then the free energy of the system is
FA = E − TA S, (5.12)

52
or, more precise,

FA = EA − TA SA . (5.13)

For any thermodynamic transformation from A to B (with temperatures TA and TB ) when coupled to a
heat bath at temperature T , the above derivation yields

Lmax = [EA − T SA ] − [EB − T SB ], (5.14)

but the right-hand side no longer coincides with −∆F .

Consider now a system in thermal contact with a heat bath at constant temperature T such that the
environment cannot perform work on the system,

W = 0. (5.15)

For homogeneous systems, this reduces to constant volume, V . For any thermodynamic transformation
between states A and B of the system we have

FB ≤ FA , (5.16)

or,

∆F ≤ 0. (5.17)

Consequently, the free energy cannot increase during any transformation and the equilibrium state is the
state of minimum free energy. The value of the equilibrium free energy, Feq , is fully determined by the
thermodynamic variables of the system, here T and V , so that

Feq (T, V ) = Fmin . (5.18)

This condition is not unlike for purely mechanical systems with kinetic and potential energy given by
m 2
2 ẋ + U (x), where the stable configuration is a time-independent solution that minimizes the potential
energy U (x). The free energy plays a role akin to the potential energy in classical mechanics, but for
thermodynamic systems in contact with a heat source.

Digression: Product rule for differentials. Consider two functions f (x) and g(x) of one variable. The product
rule of differentiation states that
d h i
f (x)g(x) = f ′ (x)g(x) + f (x)g ′ (x) (5.19)
dx
df dg
= g+f . (5.20)
dx dx
Multiplying by dx we obtain

d(f g) = g df + f dg. (5.21)

This simple equation is of great importance in thermodynamics!

To see why, recall that for reversible transformations we have

dE = T dS − P dV. (5.22)

53
Consequently,

dF = d(E − T S) (5.23)
= dE − T dS − SdT (5.24)
= [T dS − P dV ] − T dS − SdT (5.25)
= −SdT − P dV. (5.26)

Analogous to our analysis of dE and dS, we conclude that for any thermodynamic equilibrium state we have

dF = −SdT − P dV, (5.27)

which implies
 ∂F   ∂F 
S=− , P =− . (5.28)
∂T V ∂V T

5.2 Chemical potential

So far we have mostly considered the case of constant particle number N . However, in many physical
processes, the particle number can change because the system exchanges particles with the environment.
Examples:

ˆ Chemical reactions, like 2H + O ⇆ H2 O. The number N of water molecules varies in time. In

equilibrium, a certain number Neq is realized.
ˆ Gas molecules adsorbed on a solid or liquid surface (system) surrounded by the gas (environment).
Gas molecules permanently attach or detach from the surface, so the number of adsorbed particles N
varies in time. In equilibrium, a certain number Neq is realized.
ˆ Two differently colored liquids separated by a permeable membrane, for instance a coffee filter. The
number of particles of each color, N1 and N2 , on either side of the membrane vary in time.
In the following, if a system can exchange particles with its environment, we say that they are in chemical
contact. If we wait long enough so that the particle number of the system N does not change and reached
its equilibrium value Neq , we say that the system and environment are in chemical equilibrium.

We define the chemical potential as

 ∂E 
µ= . (5.29)
∂N S,V

The differential of the energy becomes

dE = T dS − P dV + µdN. (5.30)
This replaces the infinitesimal form of the 1st law of thermodynamics if N is not constant during a thermo-
dynamic transformation. Since E and S are extensive, we find that µ is intensive.

We have the following three sets of extensive and intensive thermodynamic variables:

54
 category extensive intensive
thermal S T
mechanical V P
chemical N µ

They always show up in these pairs in thermodynamic differentials. For instance, for the entropy S and free
energy F = E − T S we have

1 P µ
dS = dE + dV − dN, (5.31)
T T T
dF = −SdT − P dV + µdN. (5.32)
Exercise: Show these starting from dE.

Interpretation. What does the chemical potential mean or measure? The definition of µ implies that it is
measured in units of energy, say, Joule. It amounts to the energy costs ∆E when adding one particle to
the system at constant S and V , or the free energy cost ∆F when adding a particle at constant T and V .
Another perhaps more useful interpretation is to think of µ simply as some parameter which is such that if
µ increases, the particle number N increases. Similarly, the temperature T is some parameter such that if
T increases, the energy increases.

If two systems “1” and “2” are in thermal and mechanical equilibrium, then their temperatures T1,2 and
pressures P1,2 satisfy

thermal equilibrium: T1 = T2 , (5.33)

mechanical equilibrium: P1 = P2 . (5.34)
(5.35)

Similarly, if they are in chemical equilibrium, their chemical potentials µ1,2 satisfy

chemical equilibrium: µ1 = µ2 . (5.36)

(5.37)
This can be derived as follows: Consider the combined system “1+2” and isolate it (thermally, mechanically,
and chemically) from any other system. Then other systems cannot exchange heat or particles with the
system or do work on it. As a result, energy E, volume V , and particle number N of the combined system
are constant in time. However, it is not fixed how the total energy, volume, particle number divides among
the two parts “1” and “2” of the system. We have

E = E1 + E2 , (5.38)
V = V1 + V2 , (5.39)
N = N1 + N2 . (5.40)

The total entropy S(E, V, N ) is additive, so

S(E, V, N ) = S1 (E1 , V1 , N1 ) + S2 (E2 , V2 , N2 ) (5.41)

= S1 (E1 , V1 , N1 ) + S2 (E − E1 , V − V1 , N − N1 ). (5.42)

55
In equilibrium, the numbers E1 , V1 , and N1 will be such that the entropy of the combined system is maximal.
(The numbers E2 = E − E1 , V2 = V − V1 , N2 = N − N1 are then automatically determined by the values
of E, V, N .) We denote these values by the subscript “eq” for equilibrium. This means that the function
S(E, V, N ) is maximal with respect to the free, independent parameters E1 , V1 , N1 . The first derivatives
vanish at the maximum, hence
∂S 
! ∂S1  ∂S2  1 1
0=  =  −  = − , (5.43)
∂E1 eq ∂E1 eq ∂E2 eq T1 T2 eq
! ∂S  ∂S1  ∂S2 
 P P2 
1
0=  =  −  = − , (5.44)
∂V1 eq ∂V1 eq ∂V2 eq T1 T2 eq
! ∂S  ∂S1  ∂S2  µ1 µ2
    
0=  =  −  =− − . (5.45)
∂N1 eq ∂N1 eq ∂N2 eq T1 T2 eq
This completes the proof. □

Gibbs–Duhem relation. Consider a homogeneous system with entropy S(E, N, V ). If we double the system,
then we double the energy, volume, and particle number. Since entropy is extensive, we have

S(2E, 2V, 2N ) = 2S(E, V, N ). (5.46)

This can be seen as a special case of S(E, V, N ) = S1 (E1 , V1 , N1 ) + S2 (E2 , V2 , N2 ) for “1”=“2”. More
generally, resizing the system by a scale factor λ we have

S(λE, λV, λN ) = λS(E, V, N ). (5.47)

Taking a derivative with respect to λ and setting λ = 1 eventually, we find

∂ h i
S(E, V, N ) = λS(E, V, N ) (5.48)
∂λ
∂ 
= S(λE, λV, λN ) (5.49)

∂λ λ=1
h ∂S ∂(λE) ∂S ∂(λV ) ∂S ∂(λN ) i
= + + (5.50)
∂E ∂λ ∂V ∂λ ∂N ∂λ λ=1
∂S ∂S ∂S
= E+ V + N (5.51)
∂E ∂V ∂N
1 P µ
= E + V − N. (5.52)
T T T
Consequently,

E = T S − P V + µN (5.53)
Gibbs–Duhem relation. (5.54)
This is a very useful relation. It is true for every homogeneous thermodynamic system. By taking a differ-
ential of this relation we obtain

dE = T dS + SdT − P dV − V dP + µdN + N dµ (5.55)

= [T dS − P dV + µdN ] +SdT − V dP + N dµ, (5.56)
| {z }
dE

hence

56
 S N
dP = dT + dµ (5.57)
V V
= sdT + ndµ. (5.58)

(Here s = S/V is the entropy density.) This equation is very important: It tells us that the three intensive
variables T, P, µ are not independent, meaning they cannot be varied independently of one another. This
thermodynamic constraint is true for every homogeneous system and independent of the equation of state.

5.3 Overview of thermodynamic potentials

In the following we define the thermodynamic potentials. They all share these properties:
(1) They are functions of state.
(2) They are extensive.
(3) They depend on three thermodynamic variables, one chosen from each of the three pairs (T, S), (P, V ),
(µ, N ). One of the variables must be extensive.
(4) When the three corresponding variables are held fixed, the potential cannot change under any reversible
transformation, and under irreversible transformations, the potentials only decrease.
(5) As a result, when the three corresponding variables are fixed, the equilibrium state will be such that
the thermodynamic potential is minimal.
The three variables in (3) are chosen to reflect the thermodynamic situation. For instance, for a system
that is thermally isolated from its environment, we choose S, whereas for a system in thermal contact with
an environment at temperature T , we choose T instead of S. The reason that (T, P, µ) are not allowed
in (3) is that they are not independent variables, due to dP = sdT + ndµ. Also, it would be impossible
to construct an extensive thermodynamic potential by using intensive variables only. The total number of
possible thermodynamic potentials is thus

23 − 1 = 7. (5.59)

However, only five of them are used in practice. Here is an overview of them:

potential formula thermal mechanical chemical equilibrium state

contact to contact to contact to
environment environment environment
energy E(S, V, N ) Ö Ö Ö minimal E(S, V, N )
free energy F (T, V, N ) F = E − TS ✓ Ö Ö minimal F (T, V, N )
enthalpy H(S, P, N ) H = E + PV Ö ✓ Ö minimal H(S, P, N )
free enthalpy G(T, P, N ) G = F + PV ✓ ✓ Ö minimal G(T, P, N )
grand potential Φ(T, V, µ) Φ = F − µN ✓ Ö ✓ minimal Φ(T, V, µ)

The potential F (T, V, N ) is also called Helmholtz free energy, the potential G(T, P, N ) is also called Gibbs
free energy3 . The name “grand potential” of Φ(T, V, µ) derives from the fact that it naturally appears
in the so-called grand-canonical ensemble in statistical mechanics. Obviously, we did not use the variable
combinations (S, V, µ) and (S, P, µ).4
3 In Fermi’s book, G is called Φ.
4 The relations between the thermodynamic potentials are so-called Legendre transformations, see my lecture notes for PHYS
530, section 3.2.

57
The differentials of the thermodynamic potentials (valid for reversible transformations) read
dE = T dS − P dV + µdN, (5.60)
dF = −SdT − P dV + µdN, (5.61)
dH = T dS + V dP + µdN, (5.62)
dG = −SdT + V dP + µdN, (5.63)
dΦ = −SdT − P dV − N dµ. (5.64)
For instance,
dH = d(E + P V ) (5.65)
= dE + P dV + V dP (5.66)
= [T dS − P dV + µdN ] + P dV + V dP (5.67)
= T dS + V dP + µdN. (5.68)
Exercise: Show the remaining ones.

The grand potential must have the form

Φ(T, V, µ) = V φ(T, µ) (5.69)
with some function φ(·, ·), because V is the only extensive variables and Φ must be extensive. However, the
differential tells us that
 ∂Φ 
φ(T, µ) = = −P, (5.70)
∂V T,µ
and so
Φ = −P V. (5.71)
This also follows from the Gibbs–Duhem relation E = T S − P V + µN , because
Φ = F − µN (5.72)
= E − T S − µN (5.73)
= −P V. (5.74)
This means that for system described by Φ, the equilibrium state is the state of maximal pressure P (T, µ).

The enthalpy H(S, P, N ) and free enthalpy G(T, P, N ) find applications in chemistry, because chemical
reactions happening in a test tube are mostly performed at constant atmospheric pressure, not at fixed
volume.

Equation of state. Specifying the equation of state of a system is equivalent to specifying its corresponding
thermodynamic potential in the right variables, because all other thermodynamic parameters follow from
derivatives of the thermodynamic potential.

Example 1. Consider a system with thermal and chemical contact to a bath, that is mechanically isolated.
The equation of state is
Φ(T, V, µ) = −V P (T, µ). (5.75)

58
Indeed, we obtain all other thermodynamic functions from
 ∂Φ   ∂Φ   ∂Φ 
S=− , P =− , N =− (5.76)
∂T V,µ ∂V T,µ ∂µ T,V
and

E = T S − P V + µN. (5.77)

To obtain E, we could also integrate dE = T dS − P dV + µdN , since we know everything on the right-hand
side.

Example 2. Consider a system in thermal contact with a bath, that is mechanically and chemically isolated.
The equation of state is

F (T, V, N ). (5.78)

From this we obtain

 ∂F   ∂F   ∂F 
S=− , P =− , µ= (5.79)
∂T V,N ∂V T,N ∂N T,V
and

E = T S − P V + µN. (5.80)

5.4 Maxwell relations

Consider an infinitesimal transformation that changes two parameters x, y by dx, dy and the expression

δ(x, y) = a(x, y)dx + b(x, y)dy. (5.81)

The following two properties of δ are equivalent.

(1)

∂a ∂b
= (5.82)
∂y ∂x

(2) There exists a function of state R such that

δ = dR = adx + bdy (5.83)

and, hence,
 ∂R   ∂R 
a= , b= . (5.84)
∂x y ∂y x

Furthermore, for a thermodynamic transformation from state A to B, the integral

 B
dR = RB − RA (5.85)
A

is independent of the path.

(We do not proof this here.) Property (1) is called the integrability condition. Expressed in terms of R, it

59
simply states that second derivatives can be exchanged,

∂2R ∂2R
= . (5.86)
∂x∂y ∂y∂x

Property (2) is satisfied for all thermodynamic potentials, R = E, F, . . . , when considering reversible trans-
formations. The variables x, y can be chosen to be any of the three variables of the potential while keeping
the third one fixed. The equation (1) then yields quite nontrivial thermodynamic identities, called Maxwell
relations.

For instance, choosing R = F (T, V, N ) and x, y ∈ {T, V, N }, there are

 
3
=3 (5.87)
2

different relations to be obtained. We start from

dF = −SdT − P dV + µdN (5.88)

and obtain the Maxwell relations

∂2F ∂2F ∂ ∂  ∂S   ∂P 
= ⇒ “ (−S) = (−P )” ⇒ = , (5.89)
∂V ∂T ∂T ∂V ∂V ∂T ∂V T,N ∂T V,N
∂2F ∂2F ∂ ∂  ∂S   ∂µ 
= ⇒ “ (−S) = (µ)” ⇒ − = , (5.90)
∂N ∂T ∂T ∂N ∂N ∂T ∂N T,V ∂T V,N
∂2F ∂2F ∂ ∂  ∂P   ∂µ 
= ⇒ “ (−P ) = (µ)” ⇒ − = . (5.91)
∂N ∂V ∂V ∂N ∂N ∂V ∂N T,V ∂V T,N
The Maxwell relations are valid for every thermodynamic system. Using a different thermodynamic potential
R, we get three different equations, since they necessarily differ in the set of variables. Not all of these
equations may be equally useful in practice.

Example. Compressing a metal block. A typical application of Maxwell relations is to convert non-intuitive
thermodynamic derivatives into so-called material constants, yielding useful formulas for experiment and
engineering. As an example, consider a block of iron. Its specific properties are characterized by material
constants such as heat capacities,
 ∂S 
CV = T , (5.92)
∂T V,N
 ∂S 
CP = T , (5.93)
∂T P,N
isothermal and adiabatic compressiblities,
1  ∂V 
κT = − , (5.94)
V ∂P T,N
1  ∂V 
κS = − , (5.95)
V ∂P S,N

60
thermal expansion coefficient
1  ∂V 
α= , (5.96)
V ∂T P,N
or thermal pressure coefficient
1  ∂P 
β= . (5.97)
P ∂T V,N
(Careful: different authors use different labels.) These quantities can be measured and are tabulated for
many materials. Assume now we compress the block of iron reversibly and at constant temperature from a
pressure P1 to a higher pressure P2 > P1 . How much heat is released during this process? We have

dQ = T dS, (5.98)

so we need dS. Using the thermodynamic independent variables T, P, N we have

 ∂S   ∂S   ∂S 
dS = dT + dP + dN. (5.99)
∂T P,N ∂P T,N ∂N T,P
Since dT = dN = 0, we have
 ∂S 
dQ = T dP. (5.100)
∂P T,N

We next use a Maxwell relation to rewrite the derivative. Since the variables T, P, N are involved, we might
want to look at

dG = −SdT + V dP + µdN. (5.101)

Indeed, this yields

∂2G ∂2G ∂ ∂  ∂S   ∂V 
= ⇒ “ (−S) = (V )” ⇒ =− = −V α. (5.102)
∂P ∂T ∂T ∂P ∂P ∂T ∂P T,N ∂T P,N
Consequently,

dQ = −T V αdP (5.103)

and
 P2
Q = −T α V dP. (5.104)
P1

Since the volume of the metal block will not change much when compressed, we can assume V ≈ const and
so arrive at

Q = −T V α(P2 − P1 ). (5.105)

Heat is released from the body, so Q < 0.

Comment. The material constants are not all independent. For instance, using thermodynamic (but not
Maxwell) relations, one can show that
CV κS
= (5.106)
CP κT
for any substance.

61
5.5 Phase transitions
Definition of a phase. When a system consists of only a single homogeneous substance, it is said to consist
of only one phase. If a heterogeneous system is composed of several parts, each of which is homogeneous in
itself, the system is said to consist of several phases.

Comment. The important property of each part that comprises a phase is the spatial homogeneity. They
can by themselves be mixtures of several chemical compounds, see the examples.

Examples. Systems composed of only one phase:

(1) a homogeneous liquid (not necessarily a chemically pure substance: solutions may also be considered,
for instance a homogeneous salt-water solution)
(2) a homogeneous solid
(3) a homogeneous gas
Systems that consist of two phases:
(4) an ice block floating in water (phase 1: solid ice, phase 2: liquid water)
(5) a system composed of water and water vapor (phase 1: liquid water, phase 2: gaseous vapor)
(6) a saturated solution of salt in water with some of the solid salt present (phase 1: liquid salt-water
solution, phase 2: solid salt)
(7) a system composed of two immiscible liquids, for instance oil on top of water (phase 1: liquid 1, phase
2: liquid 2)

The thermodynamic properties of different phases are often quite distinctive, which allows us to tell them
apart macroscopically. For instance, n is much higher for solids than for liquids or gases, or whereas cP ≈ cV
for solids and liquids, we have cP > cV for gases.

Assume a system is in one phase. If this system changes its phase through an external change of thermody-
namic variables, it is said to undergo a phase transition. Some common phase transitions are:

liquid
va
po
g

co
tin

riz
nd
g
el

at
in
m

en

io
ez

n
sa
fre

tio
n

sublimation
solid gas
deposition

If a system consists of two or more phases, we say that these phases coexist. The coexistence of phases is
not generic in equilibrium, but rather requires special thermodynamic parameters; most systems are in one
phase. In fact, at a phase transition point between two phases 1 and 2, there is a coexistence of phases 1
and 2. A division of the PT-plane (or any other choice of variables) into regions of mutually distinct phases
and their coexistence regions is called the phase diagram of the system.

62
 liquid

solid

gas

Figure 5.1: Left. Phase diagram of water: AB—coexistence curve of liquid and gas phases, AC—coexistence
curve of solid and gas phase, AD—coexistence curve of solid and liquid phase, A—triple point, f—freezing
point at P = 1 atm, b—boiling point at P = 1 atm. The slope of the melting curve AD represents the
anomaly of water ( dP
dT |melt < 0). Right. Phase diagram of a generic substance (without anomaly in the
melting curve).

At the triple point of water located at

Ttr = 0.0075◦ C, (5.107)
Ptr = 0.00602 atm, (5.108)
the solid, liquid, and gas phases of water coexist in equilibrium. One can show that such a coexistence
region is necessarily an isolated point in the PT-plane. This and similar geometric statements follow from
the so-called Gibbs phase rule.

Latent heat. Consider a homogeneous system of water in its solid phase. We add heat Q to the system to
increase its temperature at constant pressure P = 1 atm. The system undergoes a melting and vaporization
transition at temperatures Tf (P ) = 0◦ C and Tv (P ) = 100◦ C. The diagram of temperature vs. heat supplied,
looks like this:

T vapor

liquid

latent heat
ice of vaporization

latent heat
of melting

heat supplied Q

63
When the system turns from a solid to a liquid, the temperature remains constant and the heat supplied
is used within the substance to break some chemical bonds and modify its internal structure. This costs
energy that cannot be used to increase temperature. The amount of heat Qlat which does not lead to a
temperature increase is called latent heat. Similarly, when the system turns from a liquid to a gas, latent
heat of vaporization needs to be supplied.

First and second order transitions. A phase transition with nonzero latent heat is called a first-order phase
transition. Whereas the thermodynamic potentials such as F (T, V, N ) or G(T, P, N ) remain continuous at
the transition or “critical” temperature T0 , the entropy S = −( ∂F ∂G
∂T )V,N = −( ∂T )P,N jumps by an amount

Qlat
∆S = (5.109)
T0
upon increasing temperature through the transition. A phase transition where the entropy remains contin-
uous, and hence there is no latent heat, is called a second-order phase transition. Crucially, we see that
derivatives of thermodynamic potentials are not always continuous at a phase transition—they may have
kinks and jumps.

In water, all phase transitions are of first order, except for the so-called critical point, which is the end-point
of the condensation curve AB located at

Tc = 374◦ C, (5.110)
Pc = 218 atm. (5.111)

For P > Pc , pressures are so high that liquid water and vapor are no longer distinct phases. Exactly at
the critical point, water and vapor coexist and show the phenomenon of critical opalescence, which is an
enhanced scattering of light that makes the substance appear milky. This is an example of a larger class of
so-called critical phenomena, which universally appear at second-order phase transitions.

64
6 Kinetic theory of gases
6.1 Brief review of classical mechanics
This section is to review some notion of classical mechanics and fix our notation.

Notation. We consider systems of particles in three dimensions. Vector quantities, such as momentum and
position, will be indicated by an arrow,
       
p1 px x1 x
p⃗ = p2  = py  , ⃗x = x2  = y  . (6.1)
p3 pz x3 z

The set of all values of p⃗ and ⃗x is called momentum and position space, respectively. The three components
of a vector are labelled by indices i, j, · · · ∈ {1, 2, 3}, i.e. pi and xi . We denote
q
p := |⃗ p| = p2x + p2y + p2z , (6.2)
p
r := |⃗x| = x2 + y 2 + z 2 . (6.3)

We define the scalar product

⃗a · ⃗b = a1 b1 + a2 b2 + a3 b3 . (6.4)

If ⃗a = ⃗b, we write

⃗a2 = ⃗a · ⃗a. (6.5)

This implies that

p √
p= p⃗2 , r = ⃗x2 . (6.6)

A derivative with respect to pi or xi is denoted

∂ ∂
= ∂pi , = ∂xi . (6.7)
∂pi ∂xi
The nabla operator is defined as
     
∂p1 ∂x1 ∂x
∇p⃗ = ∂p2  , ∇⃗x = ∂x2  = ∂y  . (6.8)
∂p3 ∂x3 ∂z
It is also customary to write
∂ ∂
= ∇p⃗ , = ∇⃗x . (6.9)
∂⃗
p ∂⃗x
These formulas are to be understood as acting on some function f (⃗x, p⃗), for instance
 
∂f
∂f  ∂x 
∇⃗x f = =  ∂f
∂y  . (6.10)
∂⃗x ∂f
∂z

A volume integral in three spatial dimensions is

   
dx dy dz f (⃗x, p⃗) = d3 x f (⃗x, p⃗), (6.11)

65
where we introduce the volume element

d3 x := dxdydz. (6.12)

Similarly,

d3 p = dp1 dp2 dp3 . (6.13)

The integration boundaries for each of the integrals depend on the context and need to be specified. If we
are integrating over a volume V in position space, we also write

d3 x f (⃗x, p⃗). (6.14)
V

For instance, if V = L3 is a cube of side lengths L, we have

  L/2  L/2  L/2
d3 x f (⃗x, p⃗) = dx dy dz f (⃗x, p⃗). (6.15)
V −L/2 −L/2 −L/2

If no volume or boundaries are specified, it is usually assumed that we integrate over all possible values, i.e.
R3 ,
   ∞  ∞  ∞
d3 p f (⃗x, p⃗) = d3 p f (⃗x, p⃗) = dpx dpy dpz f (⃗x, p⃗). (6.16)
R3 −∞ −∞ −∞

For such an integral to be finite or meaningful, the function f (⃗x, p⃗) needs to decay sufficiently fast as each
of the components pi approaches ±∞. The three most important coordinate systems are
  ∞  ∞  ∞
 
x
Cartesian: ⃗x = y  , d3 x f (⃗x) = dx dy dz f (⃗x), (6.17)
z −∞ −∞ −∞

  ∞  2π  ∞
 
r⊥ cos ϕ
cylindrical: ⃗x =  r⊥ sin ϕ  , d3 x f (⃗x) = dr⊥ dϕ dz r⊥ f (⃗x), (6.18)
z 0 0 −∞

  ∞  2π  π
 
r cos ϕ sin θ
spherical: ⃗x =  r sin ϕ sin θ  , d3 x f (⃗x) = dr dϕ dθ r2 sin θf (⃗x). (6.19)
r cos θ 0 0 0

Exercise: Compute the volumes of a cube, cylinder, and sphere using a suitable coordinate system.

Example for volume integral. Consider the so-called Gaussian distribution function
2
f (⃗x, p⃗) = e−A(⃗p−⃗p0 ) , (6.20)

with A > 0, p⃗0 some constants. We want to compute the integral

 
I= d3 x d3 p f (⃗x, p⃗). (6.21)
V

Since f (⃗x, p⃗) does not depend on ⃗x, we have

  
I= d3 x 1 d3 p f (⃗x, p⃗) = V d3 p f (⃗x, p⃗). (6.22)
V
| {z }
V

66
Writing p⃗0 = (p0x , p0y , p0z ) we have
2
+(py −p0y )2 +(pz −p0z )2 ] 2 2 2
f (⃗x, p⃗) = e−A[(px −p0x ) = e−A(px −p0x ) e−A(py −p0y ) e−A(pz −p0z ) , (6.23)

and so
 ∞  ∞  ∞ 
−A(px −p0x )2 −A(py −p0y )2 2
I=V dpx e dpy e dpz e−A(pz −p0z ) . (6.24)
−∞ −∞ −∞

We introduce the shifted variables pi → p′i = pi − p0i and see that the integrals becomes independent of p⃗0 ,
 ∞
′ 2
 ∞
′ 2
 ∞
′ 2

I=V dp′x e−A(px ) dp′y e−A(py ) dp′z e−A(pz ) (6.25)
−∞ −∞ −∞
3
=: V J . (6.26)

Each of the three integrals has the same value J. It can be computed as follows:
 ∞
2
J= dx e−Ax (6.27)
−∞
 ∞  ∞
2
+y 2 )
J2 = dx dy e−A(x (6.28)
−∞ −∞
 ∞  2π    
2 x r cos ϕ
= dr r dϕ e−Ar , = (6.29)
0 0 y r sin ϕ
 ∞
2
= 2π dr re−Ar (6.30)
0
−Ar 2 ∞
e
= 2π (6.31)

−2A

r=0
π
= , (6.32)
A
so

 ∞
r
2 π
J= dx e−Ax = . (6.33)
−∞ A

Comment: J is called a Gaussian integral.

We eventually arrive at
 π 3/2
I=V . (6.34)
A

N -particle systems. We consider a system of N particles in three dimensions. The particles are labelled by
the indices a, b, · · · ∈ {1, . . . , N }. The vectors corresponding to each of the N particles are p⃗a and ⃗xa . The
ith component of the momentum of the ath particle is indicated by a double index, pai . A dynamical state
of the system is specified by the positions and momenta of all particles. The total energy of a dynamical
state is given by the Hamiltonian function

H = H(⃗x1 , . . . , ⃗xN , p⃗1 , . . . , p⃗N ). (6.35)

This splits into kinetic and potential energy according to

H = Hkin + Hpot (6.36)

67
with
N
X p⃗2a
Hkin = , (6.37)
a=1
2m
N N
X 1 X
Hpot = Vex (⃗xa ) + Vint (|⃗xa − ⃗xb |). (6.38)
a=1
2
a,b=1

Here Vex (⃗x) is an external potential and Vint (r) is the interaction energy between two particles at distance
r. For instance, Vex (⃗x) could be Earth’s gravitational field,

Vex (⃗x) = mgz, (6.39)

and Vint (r) could be a molecular Lennard–Jones type, electrostatic Coulomb, or gravitational Newtonian
potential,

σ τ
 r12 − r6

1 q1 q2
Vint (r) = 4πε r
(6.40)
m1 m2

−G r


The force acting on particle a is

∂Hpot
F⃗a = − = −∇⃗xa Hpot . (6.41)
∂⃗xa
The velocity is

p⃗a
⃗va = = ⃗x˙ a , (6.42)
m
and Newton’s equation of motion reads

⃗
¨a = Fa .
⃗x (6.43)
m
A non-interacting (= ideal) system has Vint (r) = 0, and so
N
X
H= H1 (⃗xa , p⃗a ) (6.44)
a=1

with single-particle Hamiltonian

p⃗2
H1 (⃗x, p⃗) = + Vex (⃗x). (6.45)
2m

6.2 Boltzmann equation

Kinetic theory explains all macroscopic properties of gases through the motion of atoms and their collisions—
with each other and with the walls of the container. Its success in explaining thermodynamic phenomena
significantly contributed to the acceptance of the atomic hypothesis.

For a thermodynamic system with fixed parameters such as (T, V, N ), we do not attempt to trace the
positions and momenta of all individual particles. Instead, we pick an arbitrary particle and denote by

f (⃗x, p⃗, t) (6.46)

68
a distribution function that is proportional to the probability that this particle is at position ⃗x with mo-
mentum p⃗ at time t. Obviously, if we know f (⃗x, p⃗, t), we know how the system behaves as a collection of
particles. More precisely, we choose a normalization such that

dN = f (⃗x, p⃗, t) d3 x d3 p (6.47)

is the number of particles located inside a small cube of volume d3 x around the position ⃗x, and a small
cube d3 p around p⃗ in momentum space. One should think of d3 x and d3 p as containing a macroscopic
number of particles. For instance, for a gas at room temperature we have n ∼ 1019 cm−3 , so a volume of
d3 x = (1 µm)3 = 10−18 m3 = 10−12 cm3 would appear pointlike to us, although it still contains 107 particles.
The total number of particles is
  
N= dN = d3 x d3 p f (⃗x, p⃗, t) (6.48)
V

at any time t. The local number density of particles at position ⃗x (with any momentum p⃗) is

n(⃗x) = d3 p f (⃗x, p⃗, t), (6.49)

and

N= d3 x n(⃗x). (6.50)
V

The momentum distribution of particles with momentum p⃗ (at any position ⃗x) is

ñ(⃗
p) = d3 x f (⃗x, p⃗, t), (6.51)
V

with

N= d3 p ñ(⃗
p). (6.52)

Boltzmann equation. The Boltzmann equation is the equation of motion for f (⃗x, p⃗, t) in time. For this,
consider a particle with (⃗x, p⃗) at time t. At a later time t′ = t + δt, with δt infinitesimal, it will have
coordinates
p⃗
⃗x′ = ⃗x + ⃗v δt = ⃗x + δt, (6.53)
m
p⃗′ = p⃗ + F⃗ δt, (6.54)

where ⃗v is the velocity of the particle and F⃗ = −∇⃗x Vex the external force acting on the particle.5 If there
are no collisions (interactions) among the particles, we have

f (⃗x, p⃗, t) = f (⃗x′ , p⃗′ , t′ ), (6.55)

because the probability to find the particle ”moves” along with the particle’s trajectory, or
 p⃗ 
f ⃗x + δt, p⃗ + F⃗ δt, t + δt − f (⃗x, p⃗, t) = 0. (6.56)
m
In the limit δt → 0 we obtain
5 In ⃗ that we will ignore in the following.
the presence of interactions, Vint ̸= 0, there is a mean-field contribution to F

69
  p⃗ 
∂t + · ∇⃗x + F⃗ · ∇p⃗ f (⃗x, p⃗, t) = 0. (6.57)
m
Boltzmann equation for non-interacting particles (6.58)

Example. Consider non-interacting particles without external forces. The solution to the Boltzmann equation
 
∂t + ⃗v · ∇⃗x f (⃗x, p⃗, t) = 0 (6.59)

is any function of the form

f (⃗x, p⃗, t) = φ(⃗x − ⃗v t). (6.60)

This is a traveling wave. Since the initial distribution is

f (⃗x, p⃗, 0) = φ(⃗x), (6.61)

we have

f (⃗x, p⃗, t) = f (⃗x − ⃗v t, p⃗, 0). (6.62)

Collision integral. In the presence of interactions,

f (⃗x, p⃗, t) ̸= f (⃗x′ , p⃗′ , t′ ), (6.63)

because a particle that started at (⃗x, p⃗) at t might be scattered away and not end up at (⃗x′ , p⃗′ ) at t′ , or
scattered particles that did not start at (⃗x, p⃗) at t might appear at (⃗x′ , p⃗′ ) at t′ . We define the collision
integral as the difference,

f (⃗x + ⃗v δt, p⃗ + F⃗ δt, t + δt) − f (⃗x, p⃗, t) =: Icoll δt, (6.64)

hence

 p⃗ 
∂t + · ∇⃗x + F⃗ · ∇p⃗ f (⃗x, p⃗, t) = Icoll . (6.65)
m
Boltzmann equation (6.66)
For simplicity assume that f (⃗x, p⃗, t) does not depend on ⃗x. Consider then a collision where our particle with
momentum p⃗ scatters elastically with a particle of momentum p⃗2 into two particles with final momentum p⃗3
and p⃗4 . Similarly, consider two particles with initial momenta p⃗3 and p⃗4 that scatter elastically into particles
with momenta p⃗ and p⃗1 :

The number of events of the first kind (which decrease f ) is proportional to f (⃗p, t)f (⃗
p2 , t), which is the
probability to find this initial configuration.6 Similarly, the number of events of the second kind (which
6 We assume molecular chaos here, where the momenta of the two particles are uncorrelated and the probability factorizes.

70
increase f ) is proportional to f (⃗
p3 , t)f (⃗
p4 , t). We have
   h i
3 3
Icoll = d p2 d p3 d3 p4 M(⃗
p, p⃗2 ; p⃗3 , p⃗4 ) −f (⃗
p, t)f (⃗
p2 , t) + f (⃗
p3 , t)f (⃗
p4 , t) . (6.67)

In an elastic collision, energy and momentum are conserved, thus the momenta are constrained by
(
p⃗ + p⃗2 = p⃗3 + p⃗4 ,
(∗) ⃗2
p ⃗22
p ⃗23
p ⃗24
p . (6.68)
2m + 2m = 2m + 2m

The function M encodes the details of the interaction, and is not important for us here. A typical form is
 p⃗2 p⃗2 p⃗2 p⃗2 
M(⃗ p, p⃗2 ; p⃗3 , p⃗4 ) = Ṽint (⃗q)δ + 2 − 3 − 4 δ (3) (⃗
p + p⃗2 − p⃗3 − p⃗4 ), (6.69)
2m 2m 2m 2m
with momentum transfer ⃗q = p⃗3 − p⃗ and Ṽint (⃗q) the Fourier transform of the interaction potential Vint (r).

6.3 Maxwell–Boltzmann distribution

Our goal is to determine the equilibrium distribution function f0 (⃗x, p⃗) defined as the solution of the Boltzmann
equation that is independent of time. For large times, any solution f (⃗x, p⃗, t) will behave as
t→∞
f (⃗x, p⃗, t) −→ f0 (⃗x, p⃗), (6.70)
which we associate with the equilibration process. For a homogeneous system, i.e. in the absence of external
potentials, Vex (⃗x) = 0, the system is translation invariant and so f0 cannot depend on ⃗x. Consequently,
ñ0 (p)
f0 (⃗x, p⃗) = f0 (⃗
p) = . (6.71)
V
Furthermore, F⃗ = −∇⃗x Vex = 0 and so the Boltzmann equation reads
p, t) = Icoll .
∂t f0 (⃗ (6.72)
The equilibrium distribution satisfying ∂t f0 = 0 is then determined from

Icoll = 0. (6.73)
A sufficient (and, one can show, necessary) condition for this to hold is
f0 (⃗
p)f0 (⃗
p2 ) = f0 (⃗
p3 )f0 (⃗
p4 ), (6.74)
or, taking the logarithm
ln f0 (⃗
p) + ln f0 (⃗
p1 ) = ln f0 (⃗
p3 ) + ln f0 (⃗
p4 ). (6.75)
This has the form of a conservation law, since (⃗ p, p⃗2 ) and (⃗
p3 , p⃗4 ) can be any momenta that satisfy momentum
and energy conservation. If χ(⃗ p) is any quantity associated to a particle with momentum p⃗, so that χ(⃗ p1 ) +
χ(⃗p2 ) is conserved in an elastic collision between two particles with p⃗1 and p⃗2 , then the most general solution
is
ln f0 (⃗
p) = χ1 (⃗
p) + χ2 (⃗
p) + . . . , (6.76)
where χ1 (⃗
p), χ2 (⃗
p), . . . is a list of all independently conserved quantities. For point-particles, these are the
energy and momentum of the particle, see Eqs. (∗), and, of course, a constant. Hence the most general form
is
p⃗2 ⃗ ′ · p⃗ + C ′
p ) = A′
ln f0 (⃗ +B (6.77)
2m
= −A(⃗p − p⃗0 )2 + C (6.78)

71
 ⃗ ′ = 2A⃗
with suitable constants A′ = −2mA, B p0 , C ′ = C − A⃗
p20 . We conclude that the stable equilibrium
distribution is given by

ñ0 (⃗
p) 1 2
f0 (⃗
p) = = e−A(⃗p−⃗p0 ) (6.79)
V N
where N is a normalization constant. We define the expectation value of a function g = g(⃗x, p⃗) in equilibrium
as
 
3
⟨g⟩ = d x d3 p g(⃗x, p⃗)f0 (⃗
p). (6.80)
V


Note that ⟨g⟩ is extensive because of the volume integral V
d3 x.

Fixing the constants. We first normalize f0 such that

  
! V 2 V  π 3/2
N= 3
d x 3
d p f0 (⃗
p) = d3 p e−A(⃗p−⃗p0 ) = , (6.81)
V N N A

thus
V  π 3/2
N = . (6.82)
N A
The average momentum of a gas particle is
 
⟨⃗
ptot ⟩ 1
⟨⃗
p⟩ = = d3 x d3 p p⃗ f0 (⃗
p) (6.83)
N N V
 
1 1 2
= d3 x d3 p p⃗ e−A(⃗p−⃗p0 ) (6.84)
NN V
   
11 2 1 2

= d3 x d3 p p⃗0 e−A(⃗p−⃗p0 ) + p − p⃗0 ) e−A(⃗p−⃗p0 )
d3 x d3 p (⃗ (6.85)
N N V N
| V {z }
0
 
p⃗0 3 3
= d x d p f0 (⃗
p) (6.86)
N V
| {z }
N
= p⃗0 . (6.87)

We used that for an even function η(⃗x) = η(−⃗x) we have

   
   
∞x 0
∞ ∞
d3 x ⃗x η(⃗x) = dx dy dz y  η(⃗x) = 0 . (6.88)
−∞ −∞ −∞ z 0

So p⃗0 is the average momentum of particles. For a system that does not move as a whole, we must therefore
choose

p⃗0 = 0. (6.89)

Indeed, for all particles with p⃗, there should be an equal number of particles with −⃗
p. At last we determine
A. We write
β
A= (6.90)
2m

72
and
 β 3/2 2
f0 (⃗
p) = n e−βp /(2m) . (6.91)
2πm
We determine β from the pressure. Consider a container wall in the yz-plane and gas particles impinging on
it from the left with positive px > 0. The number of particles hitting an area A of the wall in a time-interval
∆t, i.e. the number of collisions, is given by
1 3 1 1
dNcoll = p) A∆x
f0 (⃗ p)Avx ∆td3 p =
| {z } d p = 2 f0 (⃗ p)Apx ∆td3 p.
f0 (⃗ (6.92)
2 2m
∆V

The factor 21 accounts for the fact that we only consider half the particles due to px > 0. The momentum
transfer onto the wall per collision is

∆px = 2px , (6.93)

so the force onto the wall per collision is

∆px 2px
Fx = = . (6.94)
∆t ∆t
Hence

⟨Fx ⟩ 1 2px
P = = dNcoll (6.95)
A A ∆t

1
= d3 p p2x f0 (⃗
p) (6.96)
m
   ∞   ∞
1 1  ∞ 2 −Ap2x −Ap2y 2

= dpx px e dpy e dpz e−Apz (6.97)
mN −∞ −∞ −∞
√π √π √π
| {z }| {z }| {z }
1
2A A A A

1  A 3/2 1  π 3/2
= n × (6.98)
m π 2A A
n n
= = . (6.99)
2mA β
We used that
 ∞  ∞
π 1/2
r
2 ∂ 2 ∂ π
dx x2 e−Ax = − dx e−Ax = − = , (6.100)
−∞ ∂A −∞ ∂A A 2A3/2

which is another Gaussian integral. Since the absolute temperature is defined such that

P = nkB T (6.101)

for an ideal gas, which also holds for very weakly interacting gases, we conclude that

1
β= . (6.102)
kB T
The equilibrium distribution becomes

n 2
f0 (⃗
p) = 3/2
e−p /(2mkB T ) (6.103)
(2πmkB T )
Maxwell–Boltzmann distribution. (6.104)

73
 p), which only depends on the magnitude p = |⃗
Note that the rotation-invariant or isotropic form of f0 (⃗ p|,
implies that
⟨p2x ⟩ = ⟨p2y ⟩ = ⟨p2z ⟩. (6.105)
Furthermore,
 
3
2
⟨p ⟩ = d x d3 p(p2x + p2y + p2z )f0 (⃗
p) = 3⟨p2x ⟩. (6.106)
V

The average energy per particle is

 
⟨E⟩ 1 p2
= d3 x d3 p f0 (⃗
p) (6.107)
N N V 2m
1
= ⟨p2 ⟩ (6.108)
2mN
3
= ⟨p2 ⟩ (6.109)
2mN x
3
= V × mP (6.110)
2mN
3P V 3
= = kB T. (6.111)
2N 2
Hence we obtained from kinetic theory that

3
E= P V, (6.112)
2
3
E = N kB T (6.113)
2
for point-like particles, i.e. monoatomic gases. The equilibrium distribution h0 (p) of magnitudes p (instead
of p⃗), defined from
  ∞  ∞
d3 p f0 (⃗
p) = 4π dp p2 f0 (⃗
p) =: dp h0 (⃗
p), (6.114)
0 0

is given by

 8 3/2 2
h0 (p) = 4πp2 f0 (⃗
p) = n p2 e−p /(2mkB T ) . (6.115)
π 3 mkB T
For small momenta we have
h0 (p) ∼ p2 , (6.116)
for large momenta we have
2
h0 (p) ∼ p2 e−βp /(2m)
. (6.117)

Exercise: Show that the most probable momentum, p̄, defined as the momentum such that
h0 (p) = 4πp2 f0 (⃗
p) (6.118)
is maximal, is given by
p
p̄ = 2mkB T , (6.119)

74
whereas the root mean square momentum, defined through
 
⟨p2 ⟩ 1 3
p2rms = = d x d3 p p2 f0 (⃗
p), (6.120)
N N V

is given by
p
prms = 3mkB T . (6.121)

Example. For a gas of molecular nitrogen N2 (m = 2 × 14 u) at room temperature (T = 293 K), we have
r
2kB T m
v̄ = = 417 , (6.122)
m s
r
3kB T m
vrms = = 511 . (6.123)
m s

6.4 Equipartition and virial theorem

For an ideal monoatomic gas of pointlike particles with Hamiltonian
N N
X p⃗2a X 1  2 
H= = pax + p2ay + p2az , (6.124)
a=1
2m a=1 2m

the three momentum components p⃗ = (px , py , pz ) are the only degrees of freedom of a particle. We define

⟨. . . ⟩
⟨. . . ⟩1 = (6.125)
N
as the average per particle. We have

⟨p2 ⟩1
⟨p2x ⟩1 = = mkB T (6.126)
3
D p2 E 1
x
⇒ = kB T (6.127)
2m 1 2
per particle, which yields
N D 2 E N D 2 E
X p X px 1
⟨E⟩ = =3 = 3N × kB T. (6.128)
a=1
2m 1
a=1
2m 1 2

If the gas consists of diatomic molecules, each particle has five degrees of freedom. Indeed, besides the
momentum p⃗ there are two rotational degrees of freedom, associated to rotations about the x- and y-axis,
where we assumed that the bond connecting both atoms is along the z-axis. This means that a rotation
about the z-axis leaves the molecule invariant and would therefore not amount to an actual physical motion
of the molecule. (Similarly, a pointlike particle has zero rotational degrees of freedom.) If I is the moment
of inertia of the molecule and ωi the angular velocity for rotation about the i-axis, then
N h
X 1  2  I i
H= pax + p2ay + p2az + ωx2 + ωy2 . (6.129)
a=1
2m 2

75
For such a diatomic gas we have
1
⟨E⟩ = 5N × kB T (6.130)
2
in equilibrium. More generally, for a gas of molecules with g degrees of freedom per molecule, we have

1
⟨E⟩ = g × N × kB T (6.131)
2
equipartition theorem (6.132)
in equilibrium. This implies, of course that
CV g
= kB . (6.133)
N 2
The equipartition theorem is only valid at high temperatures, where quantum statistics can be neglected.

To prove the equipartition theorem, we write

N
X
H= H1 (⃗xa , p⃗a ) + interactions, (6.134)
a=1

where H1 (⃗x, p⃗) is the single-particle Hamiltonian and the interactions are assumed to be present, but weak.
An analysis of the Boltzmann equation and the vanishing of the collision integral then shows that ln f0 is a
sum of all conserved quantities. For a system at rest this yields

ln f0 (⃗x, p⃗) = βH1 (⃗x, p⃗) + C, (6.135)

because energy is the only conserved quantity that is nonzero. Hence

1 −βH1 (⃗x,⃗p)
f0 (⃗x, p⃗) = e , (6.136)
N
where N is some normalization constant. For any function A(⃗x, p⃗), we have
 
∂  
0= d3 x d3 p
A(⃗x, p⃗)f0 (⃗x, p⃗) (6.137)
∂pi
V   ∂A ∂f0 
= d3 x d3 p f0 + A (6.138)
∂pi ∂pi
V   ∂A ∂H 1

= d3 x d3 p − βA f0 (6.139)
V ∂pi ∂pi
D ∂A E D ∂H E
1
= −β A . (6.140)
∂pi ∂pi
The integral vanishes if Af0 decays sufficiently fast for large momenta. Hence
D ∂H E D ∂A E
1
A = kB T , (6.141)
∂pi ∂pi
and, similarly,
D ∂H E D ∂A E
1
A = kB T . (6.142)
∂xi ∂xi

76
Inserting, for instance, A(⃗x, p⃗) = pi or A(⃗x, p⃗) = xi this yields
D ∂H E
1
pi = kB T, (6.143)
∂pi 1
D ∂H E
1
xi = kB T. (6.144)
∂xi 1
More generally, if the full H depends on any coordinate Z (position, momentum, angular velocity, . . . ), then
D ∂H E
Z = kB T. (6.145)
∂Z 1

So if the Hamiltonian contains a term

A 2
H= Z + ..., (6.146)
2
∂H1
then ∂Z = AZ and
DA E 1 D ∂H E 1
Z2 = Z = kB T. (6.147)
2 1 2 ∂Z 1 2
We associate each such quadratic term with a degree of freedom. This completes the proof.

Example. Dulong–Petit law. Consider a solid as a three-dimensional crystal lattice of atoms or ions. The
positions of the N ions are labelled by xai , their momenta by pai . The Hamiltonian reads
N
X p⃗2a
H= + Hpot (⃗x1 , . . . , ⃗xN ). (6.148)
a=1
2m

In the harmonic approximation, we assume that the positions ⃗xa only slightly deviate from their equilibrium
positions ⃗x0,a defined as the minima of Hpot . The Hamiltonian can then be approximated by
N  2
X p⃗a k 
H= + ⃗u2a , (6.149)
a=1
2m 2

where ⃗ua = ⃗xa − ⃗x0,a is the displacement of the ath ion. This a three-dimensional harmonic oscillator with
g = 6 degrees of freedom per ion. We conclude that

CV = 3N kB (6.150)
Dulong–Petit law, (6.151)
which is a good approximation for CV ≈ CP of many solids at high temperatures.

Virial theorem. Consider a general Hamiltonian of the form

H = Hkin (⃗
p1 , . . . , p⃗N ) + Hpot (⃗x1 , . . . , ⃗xN ). (6.152)

For
N
X p⃗2a
Hkin = (6.153)
a=1
2m

77
we have
N
∂Hkin ∂ X p⃗2b
p⃗a · = p⃗a · (6.154)
∂⃗
pa ∂⃗
pa 2m
b=1
N
X p⃗b p⃗2a
= p⃗a · δab = . (6.155)
m m
b=1

Consequently,
N D N
X ∂Hkin E XD p⃗2a E
p⃗a · = = 2⟨Hkin ⟩. (6.156)
a=1
∂⃗
pa a=1
m

Since
D ∂Hkin E D ∂Hpot E
p⃗a · = ⃗xa · (6.157)
∂⃗
pa ∂⃗xa
we conclude that

N D
X ∂Hpot E
2⟨Hkin ⟩ = ⃗xa · (6.158)
a=1
∂⃗xa
Virial theorem. (6.159)
More generally, one can show that if

Hpot (⃗x1 , . . . , ⃗xN ) (6.160)

is a homogeneous function of degree n, i.e.

Hpot (λ⃗x1 , . . . , λ⃗xN ) = λn Hpot (⃗x1 , . . . , ⃗xN ), (6.161)

then
N
X ∂Hpot
⃗xa · = nHpot . (6.162)
a=1
∂⃗xa

The kinetic energy is always homogeneous of degree two. For a potential of the form
N
X
Hpot (⃗x1 , . . . , ⃗xN ) = ca |⃗xa |n (6.163)
a=1

we then obtain

2⟨Hkin ⟩ = n⟨Hpot ⟩. (6.164)

This formula finds applications, for instance, in solids (harmonic potential, n = 2) and astrophysics (gravi-
tational potential, n = −1).

78