Thermodynamics and Statistical Physics

1
B. Zheng
Zhejiang Institute of Modern Physics, Zhejiang University,
Hangzhou 310027, P.R. China

1
e-mail: zheng@zimp.zju.edu.cn; tel: 0086-571-87952753; Fax: 0086-571-87952754

1
Chapter 1

Thermodynamics

2
Chapter 2

Classical statistical mechanics

From thermodynamics one does not understand every-

thing. Thermodynamics does not develop methods for
the computations of thermodynamic parameters. For
example, what is pressure, and in particular, what are
temperature and entropy, and how does one derive the
characteristic functions? More generally, local quanti-
ties such as spatial correlation functions are usually not
touched in thermodynamics. Thermodynamics does not
look fundamental and systematic, and all the impor-
tant laws seem isolated each other. The kinetic theory
of molecules offers certain understanding, but not suffi-
ciently general.

2.1 Postulates
Historically, thermodynamics arose before the atomic
nature of matter was understood. The idea that heat is
a form of energy was first suggested around the end of
eighteenth century, and gained acceptance in the next
half century. The kinetic theory of dilute gases and more

3
general statistical mechanics were developed around the
second half of the nineteenth century and the beginning
of the twentieth century.

Statistical mechanics
• the equilibrium state
• non-equilibrium states

Statistical mechanics in this book is mainly concerned

with the equilibrium state.
• Calculate macroscopic parameters from microstruc-
tures and interactions
• Derive thermodynamics
e.g., define and calculate
temperature
internal energy
entropy
characteristic functions
Prove the important laws in thermodynamics
• Statistical mechanics generally goes beyond ther-
modynamics
e.g., compute spatial correlation functions
The system: a classical system isolated in the sense
that the energy is conserved or nearly conserved, is com-
posed of a large number of elements, typically
N ≃ 1023 molecules

4
 The thermodynamic limit

N → ∞, V →∞

but
V
→ v (a non-zero constant)
N
Here V is measured in a microscopic unit,e.g.,

V ≃ 1023 molecular volumes

In other words, extremely dilute or dense systems are

excluded in the thermodynamic limit.

The thermodynamic properties of the system may be

tackled from three levels.
• Fundamental level
Solve the microscopic equations of motion, such as New-
ton, Hamiltonian and Heisenberg Equations.
Difficulties
∗ too many degrees of freedom
∗ microscopic initial conditions and boundary conditions
∗ irregular disturbance from environments
Here the time t and space coordinate qi are micro-
scopic.
Even if the fundamental equations could be solved,
we still have to compute the macroscopic physical pa-
rameters from the microscopic solutions, based on cer-
tain statistical theories. Therefore, statistical mechan-
ics does provide ‘fundamental’ methods for dealing with
many-body systems.

5
 • Quasi-fundamental level
Do not trace the motion of each molecule, and con-
sider only the probability distribution ρ(pi , qi , t) with
∏
ρ(pi , qi , t) d3 pi d3 qi
i
being the probability of the system lying within a vol-
∏
ume i d3 pi d3 qi of the coordinate qi and momentum pi .
Equations of motion:
∗ Liouville’s equation
∗ Boltzmann equation
These equations can be solved only in some simple
cases, such as dilute gases.
The time t and space coordinate qi are mesoscopic.
Otherwise there is no meaning for the probability dis-
tribution.
• Statistical mechanics
Do not solve any equations of motion, but assume a
form of ρ(pi , qi , t) in the equilibrium state, i.e., ρ(pi , qi , ∞) ≡
ρ(p, q),
* it can be tested by experiments
* it can be derived from equations of motion in
some special cases.

More strictly, this is the so-called ensemble theory.

The equilibrium state is in the macroscopic sense, i.e.,
the time scale is macroscopic.

For non-equilibrium states, one usually assumes cer-

tain effective equations of motion. The time scale is
usually mesoscopic.

6
 Γ space: the phase space spanned by (p, q), each point
in Γ space represents a microscopic state of the system.
An ensemble: A collection of systems, identical in
composition and macroscopic conditions, existing in (dif-
ferent or the same) microscopic states, without inter-
actions each other. A system in the ensemble can be
represented by a point in Γ space, and these systems
obey the probability distribution ρ(p, q), i.e.,
∏
ρ(p, q)dp dq ≡ ρ(pi , qi ) dpi dqi
i

is the number of systems at the volume element dpdq.

In simple words, an ensemble is a set of microscopic
states of the system with the probability distribution
ρ(p, q), and it is introduced to replace the dynamic evo-
lution of the microscopic states.
Physically, such a description is reasonable just be-
cause a macroscopic state may correspond to many mi-
croscopic states.
Naturally, ρ(p, q) should depend on the macroscopic
environment of the system. The basic assumption in
statistical mechanics starts from an isolated system.

Postulate of equal a priori probability

When an isolated system is in thermodynamic equilib-
rium, its microscopic state is equally like to be any state
satisfying the macroscopic conditions, i.e.
{
const. if E < H(p, q) < E + ∆
ρ(p, q) =
0 otherwise

7
where E is the energy, ∆ ≪ E, H(p, q) is the Hamilto-
nian, and the ensemble described by this distribution is
the so-called microcanonical ensemble.
Here it is important that the Hamiltonian is assumed
to be known, which contains the microstructure and
interactions of the system.

• why is ∆ introduced?
Theoretically, it is convenient
Practically, isolation is not strict

• why is ρ(p, q) a constant for all microscopic states

with energy around E?
Energy is assumed to be the most important quan-
tity for the microscopic motion. If ρ(p, q) is not a
constant, another quantity should be specified.

From the view of the fundamental dynamics, one may

understand the postulate from the so-called ergodicity,
which assumes that the system experiences all the mi-
croscopic states in an infinitesimal macroscopic time. If
there is the ergodicity, it is natural that the dynamic
path of a system should be a simple loop in Γ space.
Therefore, the postulate holds.
If the path does not form a loop in an infinitesimal
macroscopic time, the macroscopic state is not in equi-
librium. If there are two loops, it is not ergodic. In fact,
the ergodicity is not strict. For example, if the walls of
a cubic container are completely smooth, molecules of
an ideal gas with parallel velocities will not change their
velocities. But the ergodicity may be assumed because

8
of

∗ the measure of non-ergodic states is negligibly small

∗ disturbance
∗ boundary conditions
∗ interactions
··· ···

In statistical mechanics, the ensemble average of a

measurable property f (p, q) is defined as
∫
dp dqf (p, q)ρ(p, q)
⟨f ⟩ ≡ ∫ ,
dp dqρ(p, q)
which is assumed to be the corresponding macroscopic
physical quantity. In other words, we should find f (p, q)
for each thermodynamic parameter, and this is partic-
ular important for thermodynamic parameters such as
temperature and entropy etc.
In statistical mechanics, it is just assumed that the
ensemble average is equivalent to the time average along the dy-
namic path of the microscopic states, although no general
proof exists.
For a large system, it is usually assumed that the
relative fluctuation is arbitrarily small
⟨f 2 ⟩ − ⟨f ⟩2
≪1 (2.1.1)
⟨f ⟩2
If Eq. (2.1.1) is not satisfied in a finite system, for ex-
ample, one may enlarge the system to reduce the fluc-
tuation. Therefore, Eq. (2.1.1) is usually reached in a
sufficiently large system. In the thermodynamic limit,

9
P

f0 f

Figure 2.1:

Figure 2.2:

10
the relative fluctuation goes to zero. This assumption is im-
portant in statistical mechanics.

On the other hand, we may introduce the probability

distribution P (f ) for the observable
∫ f (p, q). The average
of f (p, q) can be computed by df f P (f ). Obviously, we
should note that P (f ) is different from ρ(p, q).
As shown in Fig. 2.1, P (f ) is usually assumed to
be with a single sharp peak. We define —it the most
probable value f0 of f (p, q), by the maximum probability
P (f ).

The ensemble average ⟨f ⟩ and the most probable value f0 are

nearly the same under the condition of Eq. (2.1.1), and should
become identical in the thermodynamic limit. Otherwise, sta-
tistical mechanics should be questioned. But it does not
mean that statistical mechanics is necessarily invalid.
One should carefully study.

Question: in what case Eq. (2.1.1) may be not valid?

Answer: strongly correlated systems, for example,
with long-range interactions, or around continuous phase
transitions.

2.2 Microcanonical ensemble

To establish statistical mechanics, it is first important to
define all the macroscopic physical parameters through
the microscopic degrees of freedom, i.e., f (p, q), espe-
cially for those typical thermodynamic observables. The
average of the Hamiltonian H is naturally defined as the

11
internal energy. So, it is crucial how to define the en-
tropy and temperature.
Let us denote the volume in Γ space of the micro-
scopic ensemble
∫
Γ(E) ≡ dp dq
E<H(p,q)<E+∆

Γ(E) is understood to be dependent of N, V and also ∆.

For example, the volume is shown for H(p, q) = p2 + q 2 in
Fig.2.2.
Please note that Γ(E) is the volume, not simply the
number of states in Γ space. To define the “number”,
one should introduce the unit volume.
From experiments, the entropy is closely related to
the number of the microscopic states. The entropy is
thus defined by

S(E, V ) ≡ k log Γ(E)

where k is a universal constant eventually shown to be

the Boltzmann’s constant. Such a definition is closely
related to the fact that the entropy must be extensive.

To justify the definition of S, we should show

(a) S is extensive
(b) S satisfies the properties of the entropy as re-
quired by the second law of thermodynamics
Proof:
(a) Let the system be divided into two subsystems:
1: N1 , V1 , H1
2: N2 , V2 , H2

12
with fixed N1 , N2 , V1 and V2 , which satisfy

N = N1 + N2
V = V1 + V2

and
H(p, q) = H1 (p1 , q1 ) + H2 (p2 , q2 )
Here (p1 , q1 ) are the coordinates of particles in the sys-
tem 1, and (p2 , q2 ) are the coordinates of particles in the
system 2.
Especially it is assumed that the interaction between
two subsystems is negligibly small, although it must
exit to allow the exchange of energy between two sub-
systems. For example, if the intermolecular potential
is finite-range, and the surface-to-volume ratio of each
subsystem is negligibly small.
If we define

S1 (E1 , V1 ) = k log Γ1 (E1 , ∆)

S2 (E2 , V2 ) = k log Γ2 (E2 , ∆)

and
S(E, V ) = k log Γ(E, 2∆)
we should show in the thermodynamic limit

S(E, V ) = S1 (E1 , V1 ) + S2 (E2 , V2 )

If E = E1 + E2 is a given decomposition of the energy,

i.e., H1 fluctuates around E1 and H2 around E2 respec-

13
tively, the volume in Γ space of the whole system is
∫
Γ(E) = dp1 dp2 dq1 dq2
∫ ∫
E1 <H1 <E1 +∆, E2 <H2 <E2 +∆

= dp1 dq1 dp2 dq2

E1 <H1 <E1 +∆ E2 <H2 <E2 +∆
= Γ1 (E1 )Γ2 (E2 )
In other words, the extensive property of S has been
proven if the two subsystems are simply added up.
The key point is that the possible decomposition of
the energy E = E1 + E2 is not unique when the system
is divided into two subsystems. In other words, both
E1 and E2 may fluctuate from zero to E although con-
strained by the energy conservation E1 + E2 = E.
Let us divide E into E/∆ intervals with Ei = i∆, i =
0, 1, · · · E/∆, then
∑
E/∆
Γ(E) = Γ1 (Ei )Γ2 (E − Ei )
i=0

We will show or have to show that only one term in the

sum is dominating. This is the only possibility to set up
the extensiveness of the entropy.

Let us assume that there uniquely exists a decomposi-

tion of energy Ē1 +Ē2 = E with the largest term Γ1 (Ē1 )Γ2 (Ē2 ).
Then
E
Γ1 (Ē1 )Γ2 (Ē2 ) ≤ Γ(E) ≤ Γ1 (Ē1 )Γ2 (Ē2 )
∆
or
E
S1 (Ē1 , V1 )+S2 (Ē2 , V2 ) ≤ S(E, V ) ≤ S1 (Ē1 , V1 )+S2 (Ē2 , V2 )+k log
∆
14
 In the thermodynamic limit, we expect or assume

log Γ1 ∝ N1 , log Γ2 ∝ N2 (2.2.2)

and
E ∝ N1 + N2 = N
In addition, both N1 and N2 should be the order of
N , otherwise it may not be meaningful. Then

S(E, V ) = S1 (Ē1 ) + S2 (Ē2 ) + O(log(N ))

In the thermodynamic equilibrium, the system is homo-

geneous, therefore
N ∝V
Thus S is extensive up to the order of N , i.e., when
omitting the terms of the order O(log(N )).
In summary, the extensive property of S is based on
that a single decomposition E = Ē1 + Ē2 of the energy is
dominating when the system is divided into two subsys-
tems, with fixed N1 , N2 and V1 , V2 . The existence of such
a dominating decomposition will be discussed below.

E ∝ N1 + N2 = N ∝ V is simply the extensive property

of E. However, log Γ1 ∝ N1 ∝ V1 in (2.2.2) indicates
already that S1 is extensive.
Therefore, the derivation with E = Ē1 + Ē2 sounds
only to prove that if S1 and S2 are extensive then S is
also extensive.
Finally, we may notice that there is still another small
parameter ∆. However, this is a macroscopic param-
eter, and not necessarily infinitesimal in microscopic

15
sense. At least, one should not mix this small parameter
with the thermodynamic limit.

Understanding:
Alternatively, we may consider Ē1 and Ē2 are the av-
eraged energy of the two subsystems. In the thermody-
namic limit, the fluctuation δ of E1 around Ē1 should
be much smaller than Ē1 . Let us take δ < ∆, then only
Γ1 (Ē1 )Γ2 (Ē2 ) is dominating

∑
E/∆
Γ(E) = Γ1 (Ei )Γ2 (E − Ei )
i=1

Therefore

S(E, V ) = S1 (Ē1 , V1 ) + S2 (Ē2 , V2 )

with E = Ē1 + Ē2 .

This also explains why we can add two systems up
in equilibrium, and divide a system into two parts in
equilibrium.
In fact, following the discussions below Eq. (2.1.1),
it is natural that the decomposition of energy with the
largest term Γ1 (E1 )Γ2 (E2 ) should be identical to the de-
composition of the average energy, since the former cor-
responds to the most probable values of energy for the
two subsystems.

Question: why is it only “an understanding”?

Answer: Γ1 (Ei )Γ2 (E − Ei ) is just P (E1 = Ei ), the prob-
ability E1 takes the value Ei . When we assume δ < ∆, it
indicates already that all other terms are negligible.

16
 In fact, Γ1 (E1 )Γ2 (E − E1 ) is usually a very sharp func-
tion E1 , and such an understanding is reasonable. For
example, suppose Γ1 (E1 ) = E1N1 , Γ2 (E − E1 ) = (E − E1 )N2
and N1 = N2 , then
Γ1 (E1 )Γ2 (E − E1 ) = [E1 (E − E1 )]N1
This function has a maximum at E1 = E/2, and it is very
sharp for a large N1 .

Reading materials:
Why
∑
E/∆
Γ(E) = Γ1 (Ei )Γ2 (E − Ei )?
i=1
e.g. N =2
1 2 1 2 2
H = p + ω q
2m 2
1 2 2 1 2
H1 = ω q , H 2 = p
2 2m
∫
Γ(E) = dq dp
∫ E<H<E+2∆

Γ1 (E1 ) = dq
∫ E1 <H1 <E1 +∆

Γ2 (E − E1 ) = dp
E−E1 <H2 <E−E1 +∆

√ 1 )Γ2 (E − E1 )
Γ(E) is just the area of a circular ring. Γ1 (E
is an element of the circular ring at q = E1 . If E1 is
fixed, one may write
Γ(E) = Γ1 (E1 )Γ2 (E − E1 )

17
If not, one needs to integrate over E1
∑
E/∆
Γ(E) = Γ1 (Ei )Γ2 (E − Ei )
i=1
since E1 takes value from 0 to E, corresponding to dif-
ferent states.
End reading materials

In other words, E = Ē1 + Ē2 maximizes the function

Γ1 (E1 )Γ2 (E2 ), i.e., also log[Γ1 (E1 )Γ2 (E2 )] with the restric-
tion δE = δE1 + δE2 = 0, therefore,
δ log[Γ1 (E1 )Γ2 (E2 )] = 0, δE1 + δE2 = 0
This leads to
∂ ∂
log Γ1 (E1 ) = log Γ2 (E2 )
∂E1 ¯
E1 =E1 ∂E 2 E2 =E¯2
or
∂S1 (E1 ) ∂S2 (E2 )
=
∂E1 E=E¯1 ∂E2 E2 =E¯2
We define the temperature of any system by
∂S(E, V ) 1
≡
∂E T
Then
∂S1 ∂S
=
∂E1 ∂E2
it is simply the zeroth law
T1 = T2
T defined in this way is an intensive parameter, and
∂S/∂E = 1/T is also one of the Maxwell relations in
thermodynamics. Therefore,
if S is correctly defined, T should be also correct.

18
2.3 Thermodynamics
Let us define
∑ ∫
(E) = dp dq
H(p,q)<E
∑
∂ (E)
ω(E) =
∂E
Then
Γ(E) = ω(E)∆
∑ ∑
Γ(E) = (E + ∆) − (E)
It can be proved that up to an additive constant of the
order O(log N ), the following definitions are equivalent
S = k log Γ(E)
S = k log ω(E)
∑
S = k log (E)

Question: why?
∑
For example, suppose (E) ∝ E N , then Γ(E) ∝ N E N −1 .
∑
Or, more generally,
∑ Γ(E) < (E) < Γ(E) · E/∆. The
behaviors of log (E) and log Γ(E) look similar, up to
the order of O(N ).

With the definition

∑
S(E, V ) = k log (E)
it is easy to show that S never decreases, i.e., the second
law for a thermally isolated system in thermodynamics.
Proof: Consider a system described by the parame-
ters N, E, V . By the definition of an isolated system, N

19
and E can not change, V can not decrease. Therefore,
the second law here is simply stated as that S∑ is a non-
decreasing function of V . This is obvious, for (E) is a
non-decreasing function of V by its definition.

Hint: an isolated gas can expand only, although it

seems somewhat less clear for liquid and solid.

From the definition itself, S is a state function. For

reversible thermodynamic transformations, the entropy
should satisfy the fundamental equation of the second
law. Assuming that the system is changed slowly by
coupling the system to external environments. Then it
is a quasi-static process
( ) ( )
∂S ∂S
dS(E, V ) = dE + dV
∂E V ∂V E
( )
1 ∂S
= dE + dV
T ∂V E
Here one again assumes that there are only two inde-
pendent parameters.
Define the pressure of the system to be
( )
∂S
P ≡T
∂V E
then
1
dS = (dE + P dV )
T
or
dE = T dS − P dV
Keeping in mind that E is the internal energy, this looks
like the first law.

20
 Question: Is it reasonable to define
( )
∂S
P ≡T
∂V E
Hints:
• P is intensive
• Exercise : prove that P defined above satisfies
( )
∂E
P =−
∂V S

Up to here, the definition of P indeed is reasonable.

The definition of T is also correct if S is really entropy.

Finally, d̄Q = T dS is the heat absorbed by the system

by assuming the first law, thus S should be the entropy.
In fact, one may show that S exhibits all the properties
of the entropy analyzed in thermodynamics.

Here it is important that the first law is still an as-

sumption in statistical mechanics!

If it is possible to compute S(E, V ), and to obtain the

internal energy as E = E(S, V ), i.e., the characteristic
function, one may derive all the thermodynamic quan-
tities just by differentiation. But this can be done only
in very simple cases.

21
2.4 Equipartition Theorem
The Hamiltonian of the system is written as
H = H(p, q) = H(pj , qj ), j = 1, · · · 3N
e.g., for an ideal gas,
1 ∑ 2
H= pi
2m
or, a set of harmonic oscillators
1 ∑ 2 1 2∑ 2
H= pi + ω qi
2m 2
Here we note that negligibly small interactions between
independent degrees of freedom have been omitted.
We will derive an equipartition theorem for a general
Hamiltonian, and then apply it to a simple one.
Let xi be either pi or qi
⟨ ⟩ ∫
∂H 1 ∂H
xi = dp dq xi
∂xj Γ(E) E<H<E+∆ ∂xj
(∫ ∫ )
∆ 1 ∂H
= dp dq − dp dq xi
Γ(E) ∆ ∂xj
∫ H<E+∆ H<E
∆ ∂ ∂H
= dp dq xi
Γ(E) ∂E H<E ∂xj
The second equality is based on the definition of the
integration.
∫ ∫
∂H ∂
dp dq xi = dp dq xi (H − E)
∂xj ∂xj
∫ H<E H<E
∫
∂
= dp dq [xi (H − E)] − δij dp dq (H − E)
H<E ∂xj H<E

22
The second equality is obtained through the integration
by parts. The first term in the last line reduces to a
surface integral over the boundary of the region defined
by H < E, i.e.,
∫
the first term = dp dq xi (H − E) = 0
H=E

Question: how to describe this more convincingly?

Hint: suppose the volume H < E is enclosed by the
surface H = E, and the surface consists of the upper
and lower sheets, i.e., xj = F2 (xi , ...) and xj = F1 (xi , ...).

⟨ ⟩ ∫
∂H δij ∂
∴ xi = dp dp(E − H)
∂xj ω(E) ∂E H<E
(∫ ∫ )
δij 1
= dp dq + dp dq(E − H)
ω(E) H<E ∆ E<H<E+∆
(the second term is negligible,
at least when compared to the first term)
( ∑ )−1
δij ∑ ∂ log (E)
= (E) = δij
ω(E) ∂E
( )−1
∂S
= δij k = δij kT
∂E
If i=j ⟨ ⟩
∂H
xi = kT
∂xi
If ∑ ∑
H= Ai Pi2 + Bi Q2i
i i

23
 Pi , Qi are canonical conjugate variables
Then one may easily show
∑ ( ∂H ∂H
)
Pi + Qi = 2H
i
∂P i ∂Qi

Denote the number of the degrees of freedom by f ,

and take the ensemble average of the above equation,
1
< H >= f kT
2
exercise Prove the equipartition theorem for
1 2 1 2 2
H= p + ω q
2m 2
by explicitly calculating the ensemble average.

2.5 Classical ideal gas

The Hamiltonian of an ideal gas

1 ∑ 2
N
H = p
2m i=1 i
Then one may compute
∑ ∫
(E) = d3 p1 · · · d3 pN d3 q1 · · · d3 qN
∫
H<E

= VN d3 p1 · · · d3 pN
H<E
Let
√
R = 2mE
∑
(E) = V N Ω3N (R)

24
Ωn (R) is the volume of an n-sphere of radius R

Ωn (R) = Cn Rn
2π n/2
Cn =
Γ(n/2 + 1)
Γ(z) is the gamma function
n n n n
log Cn −−−−→ log π − log +
n→∞ 2 2 2 2
hence
∑ [ ]N
3/2
(E) = C3N V (2mE)
∑
S(E, V ) = k log (E)
[ ( )3/2 ]
4πmE 3
≃ N k log V + Nk
3N 2

( )
3N 2S
U (S, V ) ≡ E = exp −1
4πmV 2/3 3N k
( )
∂U 2U 3
T = = , U = N kT
∂S V 3N k 2
3
CV = Nk
2( )
∂U N kT
P = − =
∂V S V
From the above calculation, Ωn (R) ∝ Rn is essential,
and Cn is not much relevant.

Gibbs paradox

25
 Consider the mixture of two ideal gases at the same
temperature and the same density, with the volumes V1
and V2 , and particle numbers N1 and N2 respectively.
Initially, two gases are separated by a partition wall,
then the wall is removed and two gases mix.
If two gases are different, the entropy increases, since
both gases expand. If two gases are identical, nothing
is changed in thermodynamics, but the entropy seems
increasing in statistical mechanics as described above,
since both gases expand respectively.
To solve this problem, one ∑may introduce a Gibbs
factor in the computation of (E), i.e.,
∑ ∫
1
(E) = dN p dN q
N ! H(p,q)<E
With this correction factor, and log N ! ≈ N log N − N ,
one obtains
[ ( )3/2 ]
V 4πmE 5
S(E, V ) = N k log + Nk
N 3N 2
Thus the entropy remains unchanged after the mixture
of two identical gases. In other words, S ∼ N k log V is
unreasonable, since V is extensive.
The Gibbs factor corrects it to S ∼ N k log(V /N ). The
Gibbs factor is needed, whenever there exists a change
of particles in different volumes.
In fact, it seems that the factor one needs is N N rather
than N !. Why does one not directly introduce N N ? In
quantum mechanics, an exchange of two identical parti-
cles does not yield a new quantum state, and N ! is the
total number of such exchanges for N particles.

26
 Alternatively, when V is decomposed into V1 and V2 ,
∫ ∫ ∫
1 ∑
N
1 N!
N
d q= N1
d q dN −N1 q
N! V N! N1 !(N − N1 )! V1 V2
N1 =0

In the thermodynamic limit, however, only one term

dominates the sum, which ensures the same density for
both V1 and V2 , i.e., N1 /V1 = N2 /V2 . Therefore,
∫ ∫ ∫
1 1
dN q ≈ dN1 q dN2 q
N! V N1 !N2 ! V1 V2

The factorial number looks intrinsic here.

2.6 Maxwell-Boltzmann distribution

There are N identical molecules in volume V ,

H(p, q) = E

The system is isolated, and may be described by a mi-

crocanonical ensemble. If the molecules are quasi-independent,
∑
H= h(pi , qi )
i

where h(pi , qi ) represents the Hamiltonian of each molecule.

There could be a number of atoms in a molecule, but
the interactions between two molecules are negligible.
For example, the simplest case is
1 ∑ 2
H= p
2m i i

27
 µ space : the phase space spanned by a single-molecule
coordinates (p, q).
A microscopic state of a single molecule can be rep-
resented by a point in µ space. A microscopic state of
the system is described by a set of the points.
Since the energy of a molecule is bounded by E, the
points are confined to a finite region of µ space. We
divide the region into K elements of volume ω = d3 p d3 q,
and denote the number of the molecules in the l-th el-
ement by nl (need to draw a figure), then

∑
K ∑
K
nl = N ϵl nl = E.
l=1 l=1

Here the molecules are assumed to be quasi-independent,

ϵl is the energy level of a single particle.

It is important to note that a microscopic state of

the system may be described by a set of occupation
numbers {nl }, but a set of {nl } corresponds to not only
one microscopic state, e.g., interchange of two molecules
in different elements may lead to new states. That is,
a given set of {nl } corresponds to a volume in Γ space,
which is called the volume occupied by {nl }.

To describe a macroscopic state, we need to average

over the microcanonical ensemble, i.e., all possible mi-
croscopic states. In other words, we should first obtain
{< nl >}, then calculate all macroscopic quantities from
this distribution. Here < ... > represents the ensemble
average.

28
 However, it is difficult to perform this ensemble av-
erage.

We thus assume that the equilibrium state in the

thermodynamic limit is described by the most probable
distribution of {nl }, which occupies the largest volume
in Γ space.

Why may such an assumption hold?

• This is somewhat similar to the case when we cal-
culate Γ(E) ≃ Γ1 (Ē1 )Γ2 (Ē2 ) by dividing the system
into two subsystems.
• For a fixed l, if the relative fluctuation
(< n2l > − < nl >2 )/< nl >2
is still not small enough, we increase the total num-
ber N to reduce it. Finally, < nl > should be equal
to the most probable n̄l .
The procedure:
(a) Calculate the volume of {nl }
(b) Maximize it to obtain the most probable {n̄l }.
The volume of {nl }

N! ∏
K
Ω ({nl }) ∝ ∏K gknk
l=1 nl ! k=1

where gk is introduced for convenience, and may finally

be put to 1. For example, the derivative over gk yields
nk . Alternatively, gk may be understood as the volume
of the k-th element in the µ space.

29
 Understanding:
There are N ! ways of distributing N distinguishable
molecules to N positions. However, N positions form
K groups with the distribution {nl }. Inside a group,
there are nl ! ways of distributing nl molecules, which do
not lead to new states in Γ space. In other words, we
consider that molecules in a same element are in a same
state, since their contributions to the physical quantity
f (p, q) are the same.

For a large nl ,
log nl ! ≃ nl (log nl − 1) (i.e., nl ! ≃ nnl l )

∑
K
log Ω ({nl }) = N (log N − 1) − nl (log nl − 1)
l=1
∑
K
+ nl log gl + const
l=1
∑K
Now ∑ we vary {nl } under the condition of l=1 nl =
K
N and l=1 ϵl nl = E, to find the most probable {nl }.
We introduce the Lagrange multipliers α and β, and
calculate
( K )
∑ ∑
K
δ [log Ω ({nl })] − δ α nl + β ϵl nl = 0
l=1 l=1

Now we consider all nl are independent each other.

∑
K
[−(log nl ) + log gl − α − βϵl ] δnl = 0
l=1

30
 ∴ log n̄l = log gl − α − βϵl
n̄l = gl e−α−βϵl

Finally, α and β can be determined by the conservation

of the total particles and the total energy.
To prove that {n̄l } maximizes Ω ({nl }), we can simply
calculate the second variation
∑
K
1
− (δnl )2 < 0
nl
l=1

Note that
* n̄l is only the function of ϵl = ϵ(p, q), does not gen-
erally depends on (p, q). This is an important feature of
the equilibrium state.
* The molecules tends to gather in the lower energy
states.

2.7 Boltzmann statistical theory

We put gl = 1,

∑ ∑
N = n̄l = e−α−βϵl
l l
∑ ∑
E = ϵl n̄l = ϵl e−α−βϵl
l l

Let us assume
ϵl = ϵ(p, q, y)
The subscript l just labels p and q in µ space, and y =
{yk } represent macroscopic external parameters such as the

31
volume, and the spatial coordinates relevant to external
fields.
Define the partition function of a single particle
∑ ∫
−βϵl
Z(β, y) = e = dpdqe−βϵ(p,q,y)
l ϵ≤E

then
Z(β, y)
N = e−α Z(β, y) or α = log
N
∂ log Z(β, y)
E = −N
∂β
Suppose the system is changed very slowly.
∑ ∑
dE = n̄l dϵl + ϵl dn̄l
l l
Here we assume that the upper and lower bounds of the sum-
mation are fixed. The first term represents the macro-
scopic interaction with the external environment through
the parameters {yk }.

∑ ∑ ∂ϵl
n̄l dϵl = n̄l dyk
∂yk
l l,k
( )
∑ ∑ ∂ϵl
= n̄l dyk
∂yk
k l
Question: why not differentiate over p and q?
Answer: p and q are integrating variables.

Here
∑ ∂ϵl N ∂ log Z(β, y)
Yk ≡ − n̄l =
∂yk β ∂yk
l

32
is the force acting on the environment by the system. This
definition is similar to the pressure P = −(∂E/∂V ∑ )S in
the ensemble theory. Therefore, the term − l n̄l dϵl just
represents the work done by the system.
Proof: since
N
e−α =
Z(β, y)
therefore
N ∂ log Z(β, y) N ∂Z(β, y)
=
β ∂yk βZ(β, y) ∂yk
e−α ∂ ∑ −βϵl
= e
β ∂yk
l
∑ ∂ϵl
= − e−α−βϵl
∂yk
l
∑ ∂ϵl
= − n̄l
∂yk
l

Note that for yk = V , one should rescale q such that

the bounds of the summation or integration over q are
not dependent of V . Or, the dependence of the bounds
on V should be taken into account.
Then the force is just the pressure
N ∂ log Z(β, V )
P =
β ∂V

Question: how to prove it???

Hint: in the integration
∫ ∞ ∫ V
3
dp d3 qe−βϵ(p,q) ,
−∞

33
one may perform the transformation
q → qV 1/3 , p → pV −1/3 .
The parameter V is then transformed into ϵ(p, q). For a
free particle, for example, ϵ(p, q) = p2 /2m is transformed
into ϵ(p, q)V −2/3 .

If dϵl = 0, one identifies the heat by the first law

∑
dQ = ϵl dn̄l
l

which induces the change of the particle distribution.

Again, one notes that α is a function of β and yk , and
the differentiation here acts only on β and yk , not on p
and q. Direct computation of dQ may be complicated,
but in general
∑ ∑
dQ = ϵl dn̄l = dE + Yk dyk
l
( k
)
∂ log Z(β, y) N ∑ ∂ log Z(β, y)
= −N d + dyk
∂β β ∂yk
k

Since
∂ log Z(β, y) ∑ ∂ log Z(β, y)
d log Z(β, y) = dβ + dyk
∂β ∂yk
k

hence
( )
N ∂ log Z(β, y)
dQ = d log Z(β, y) − β
β ∂β
Define the temperature by
1
β=
kT
34
and the entropy by
( )
∂ log Z(β, y)
dS = N kd log Z(β, y) − β
∂β
( )
∂ log Z(β, y)
S = N k log Z(β, y) − β + const
∂β
Then, dQ = T dS, i.e., assuming the first law, we may
prove the second law. In fact, one may rewrite

T S = kT N log Z(β, y) + E + const

The term −kT N log Z(β, y) will be identified as the free

energy A(V, T ), as shown in the canonical ensemble.

Question: how to prove S never decreases in any ther-

modynamic processes?

Exercise: Perform the calculations for an ideal gas

with
1 ∑ 2
H= p
2m i i
Hint: Since ϵl ≪ E ∝ N , we neglect the bound of E in
calculating ∑
Z(β, y) = e−βϵl
l

Question: What is the relation between the Boltz-

mann theory and the microcanonical theory?
Answer: Since
1 ∑ 1
Γ(E) = Ω ({nl }) ≈ Ω ({n̄l })
N! N!
{nl }

35
and n! ≈ nn ,
∑
log Ω ({n̄l }) = − n̄l log n̄l
∑l
= (α + βϵl )e−α−βϵl
l
∑ ∑
= α e−α−βϵl + β ϵl e−α−βϵl
l l
Z(β, y) ∂ log Z(β, y)
= N log − Nβ
N ∂β
( )
Z(β, y) ∂ log Z(β, y)
∴ S = N k log −β
N ∂β
The difference comes from the Gibbs factor.
Fix y,
∂ log Z(β, y) ∂ log Z(β, y)
dS = N k dβ − N k dβ + kβdE
∂β ∂β
= kβdE

( )
1 ∂S
∴ = = kβ
T ∂E y

2.8 Summary
Ergodicity
→ Postulate of equal a prior probability for the mi-
crocanonical ensemble
A special case: the most probable distribution of
quasi-independent particles.
→ T, S, P and thermodynamics

36
 The drawback is that the calculation of observables
is clumsy because of the restriction of the energy.

Exercise :

• Assuming ( ) ( )
∂S ∂S 1
P ≡T , = ,
∂V E ∂E V T
prove ( )
∂E
P =− .
∂V S

• Prove the equipartition theorem for

1 2 1 2 2
H= p + ω q
2m 2
by explicitly calculating the ensemble average.

• Perform the calculations of the Boltzmann statistical theory for an ideal

gas with
1 ∑ 2
H= p
2m i i

• Exercises 6.1 and 6.3 in the textbook.

• Discussion
Negative temperature

37