Sie sind auf Seite 1von 22

Chapter 1 Introduction to Equilibrium Statistical Mechanics

T. L. Hill, An Introduction to Statistical Thermodynamics, 1986.


D. A. McQuarrie, Statistical Mechanics, 2000.
1.1 Background
1.1.1
Sterling "Approximation"
Consider the following identity:
ln x ! ln1 ln 2
ln x
ln xi
i

For a very large x, the above function can be approximated by


x

ln x !

ln ydy
1

x ln x x

For numbers in the order of Avogadro's number, the Sterling "approximation" is very
accurate.
1.1.2
Sharpness of Multiplicity Function
The following results of coin tossing suggest that the outcome will be getting more definite as
the number of trials increases:

For a system of N magnets, we denotes the number of "up" and "down" states as:
1
1
N
s and N
s , where s is an integer.
2 N
2 N
Hence, the spin excess is calculated as:
N N
2s

We define the multiplicity function W(N, s) as follows:


N!
W N, s
N !N !
The total number of states is:
1N
2

1 1

2N

1N
2

For a very large N, let us consider the logarithm of the multiplicity function:
ln W ln N ! ln N ! ln N !
Using the Stirling approximation,
ln W N ln N N N ln N N
N ln

Knowing that
N
ln
ln 12 1 2 s N
N
we have
N
ln
ln 2 2s N
N

N
N

N ln

N ln N

N
N

ln 2 ln 1 2 s N

1
2

2s N .
2

Consequently,
ln W

ln 2 2s N

1
2

2s N

ln 2 2s N

1
2

2s N

2 s2
N

ln 2 e

The above result suggests that


1N
2

2N

W N, s

W N,0

1N
2

That is, when N is very large, the distribution is exceedingly sharply defined at s = 0.
We know from common experience that systems held at constant temperature usually have
well-defined properties; this stability of physical properties follows as a consequence of the
exceedingly sharp peak in the multiplicity function and of the steep variation of that function
away from the peak.
1.1.3
Method of Lagrange Undetermined Multipliers
Assume that there are n possible outcome for an experiment:
Events: 1, 2, 3, , n.

Perform the experiment for N times and denote the frequency of the outcome i as Ni:
N1, N2, , Nn
Obviously, we must have
Ni N
i

For a particular set of {N1, N2, , Nn}, the number of possible ways is
N!
W
N1 ! N 2 ! N n !
The total possible outcome is nN.
The probability of getting the distribution {N1, N2, , Nn} is:
W
p Ni
nN

Take a logarithm of W:
ln W ln N !
ln N i !
i

If N is extremely large, we can employ the Stirling approximation:


ln W

N ln N

N i ln N i

Ni

pi ln pi
i

where pi is defined as the probability of the appearance of a particular event i.


Question: As all pi vary, what would be the maximum value of W? What would be the most
probable distribution?
Consider the following function of pi
n

pi ln pi
i 1

We want to maximize R while subjecting to the following constraint:


n

pi

i 1

Define the function L:


L R
1
pi , where
0

is a constant.

As pi varies, maximum L would give maximum R:


dL dR
1
dpi
0
i

Given that
dR
1 ln pi dpi
i

ln pi

dL

dpi

In the presence of 0, which is not yet defined, we can assert that all pi are independent
variables. Thus,
dL = 0
ln pi
pi e 0
0
The value of

is defined by the condition of

pi

1.

i 1

That is,
n

ne

pi

1 n.

i 1

Consequently,
ln W
N pi ln pi
i

1.2 Ensembles and Postulates

The object of statistical mechanics is to provide the molecular theory of interpretation of


equilibrium properties of macroscopic systems the answer of "why".

Mechanical variables (M): Properties that can be defined in purely mechanical terms such as
pressure, energy, volume, number of molecules, etc.
Non-mechanical variables: temperature, entropy, chemical potential, etc.
Ensemble: a collection of a very large number of systems, each constructed to be a replica of
a well-defined thermodynamic state.
Postulate 1: In the limit as the number of ensemble members approach , the ensemble
average of M corresponds to a parallel thermodynamic property, provided that the system of
the ensemble replicate the thermodynamic state and environment of the actual system of
interest. (Valid for all kinds of ensembles)
Postulate 2: In an ensemble of constant N, V, and E, the systems are distributed uniformly, i.e.,
with equal probability or frequency, over the possible quantum states consistent with the
specified values of N, V, and E. (Principle of equal a priori probabilities)
Postulates 1 and 2 Quantum ergodic hypothesis:
The single isolated system of interest spends equal amounts of time, over a long period of
time, in each of the available quantum states. That is, the time average and the ensemble
average of M would be identical.
1.3 Canonical Ensemble

Canonical ensemble: A huge collection of closed isothermal systems (N, V, T).


Quantum states belonging to different energy levels E will have to be considered for a
canonical ensemble.
For the following canonical ensemble, the boarder provides thermal insulation.

Each system (square box with definite N, V, T) in the canonical ensemble has the same set of
energy states, viz., E1, E2, , etc., which are determined by N and V only.
Assume there are nt systems in the ensemble.
The entire ensemble, which is insulated, is effectively an isolated system with volume ntV,
numbers of molecules ntN, and a total energy Et. Thus, it can be considered as a
thermodynamic "supersystem", to which we can apply the postulate 2.
Quantum
States
Energy
Systems

E1 = 0.1
n1

E2 = 0.2
n2

E3 = 0.3
n3

Total systems nt n1 n2
Total energy Et n1 E1 n2 E2

nnt Ent

For an "ensemble" with nt = 6 and Et = 1.5, we can have


1
2
2
2
4
4
Distribution 1: n1 = 1, n2 = 3, n4 = 2
Permutations: (1, 2, 2, 2, 4, 4), (1, 4, 2, 2, 4, 2), (4, 1, 2, 2, 2, 4), etc.
or
Application of the Postulate 2: All the permutations having the same Et are equally probable.
For a particular set of distribution n = {n1, n2, } of the energy states, the corresponding
number of possible ways (permutation) in the canonical ensemble, i.e., t n , is
t

nt !
n1 !n2 !

(1)

n1 represents the number of systems found at E1, and so on.


In general, the required probability of observing a given quantum state Ej in an arbitrary
system of a canonical ensemble is

pj

1
nt

n nj n

n
t

, where n j n

denotes the value of nj in the distribution of n.

For nt
, we can regard all other weights t (n) as negligible compared with
denotes the most probable distribution. Hence,
1
nt

pj

n* n*j

n*j

nt

(n*) which

, where n*j is the value of nj in the most probable distribution, n*.

The supersystem has a well-defined total energy at a particular temperature:


*
j

n Ej

n*j

Et

That is,
pjEj

nt

Ej

Et
nt

(2)

where
pj 1

(3)

Question: Which of all possible sets of nj's satisfying eqns (2) and (3) gives us the largest
(That is, we want to determine n*j )
From the definition of

t,

t?

we know that the problem is quite similar to what we have treated


n

in section 1.1.3, where we try to maximize the function of R

pi ln pi , but this time the


i 1

maximization is subjected to the constraints of eqns (2) and (3):


L R
1
pi
pi Ei
i

eqn 3

where

and

eqn 2

are the undetermined multipliers.

To maximize L with respect to pj, we must have


L
0
pj
As shown in section 1.1.3,

dR

1 ln pi dpi
i

Thus, we find
ln p j
Ej

R
pj

1 ln p j

After rearrangement,
E
pj e e j

As a result, the most probable distribution in a canonical ensemble is given as:


E
e j
pj
Q
Then, we can obtain the ensemble average of some mechanical variables such as energy:
E
pjEj
j

dE

E j dp j

p j dE j

dU

From the postulate 1, E

U.

Work done on or by the system may change V, on which the energy level Ej depends:
w
p j dE j
j

Ej

pj

dV

Ej

pj

p j Pj
N

Ej

Pj

PdV

where Pj is the pressure when the system is in the state Ej.


N

Hence,
q
E j dp j
j

Thus, for a very small change of internal energy, work done means a weighted average of
change in energy levels and heat is a change of distribution of energy.
One may be tempted to express the factor of in terms of energies. However, the independent
variables of real interest here are N, V, and T. Therefore, we will invoke thermodynamic
argument to trace the connection between and T (a non-mechanical thermodynamic
variable).
From the ensemble average of energy
E
e j
E
Ej
Q
j
we obtain:
E
V

,N

Ej

Ej
V

,N

e j
Q2

Q
V

Ej
,N

Ej

Ej
V

,N

PE

E P

From the ensemble average of pressure


E
e j
P
Pj ,
Q
j
we can similarly obtain
P
EP P E
N ,V

Thus, we finally obtain the following relation:


E
P
P
V ,N
N ,V
which is very similar to the following thermodynamic equation:

U
V

1
T

T ,N

P
1T

P
N ,V

By the virtue of the postulate 1, we can associate the thermodynamic pressure and internal
energy with the ensemble average of P and E.
Hence, it can be deduced that
1
kT
where k is a constant to be determined.
From the definition of E , we obtain
dE
E j dp j
p j dE j
j

Given that p j
p jQ

Ej

Ej

Q
ln p j

, we have
ln Q

Ej

Ej

ln p j

ln Q

Substituting the expression of Ej and replacing dEj by dV:


Ej
1
ln p j ln Q dp j
dE
pj
dV
V N
j
j

p j ln p j

PdV

For a closed system (N constant) in thermodynamics,


dU TdS pdV
Hence we can obtain
S N ,V , T
k p j ln p j

(4)

It can be shown that if any two systems are in thermal contact, they will have the same and
T at equilibrium. Thus, k is a universal constant known as the Boltzmann factor. Therefore,
we can evaluate k for a judiciously chosen system:
k = 1.381 10 23 JK 1.
It was Max Planck who first gave it an experimental value based on the law of black-body
radiation.
We finally obtained the well-known expression of the Boltzmann distribution:
e

p j N ,V , T

E j kT

and Q N , V , T

E j kT

where Q is called the "canonical ensemble partition function".


When T
, Q gives the total number of accessible state and the probability of accessing
each state would be the same. At regular temperature, the Boltzmann factor will modify such
probability significantly.
Substituting the above results into eqn (4):
S
k p j E j kT ln Q
j

E
T

k ln Q

In thermodynamics, the Helmholtz free energy is related to entropy as follows:


U A
S
T T
Therefore,
A N ,V , T

kT ln Q N ,V , T .

Hence, if Q is available, we can obtain a rather complete set of thermodynamic functions from
those derivatives of the Helmholtz free energy.
Given that dA
A
P
V T ,N

SdT
kT

PdV
dN , we have
ln Q
V T ,N
10

ln Q

T
N ,V

ln Q
T

kT 2
N ,V

ln Q
T

N ,V

Fluctuation
1.3.1
In a canonical ensemble, N, V, and T are held fixed, and we can investigate fluctuations in the
energy, pressure, and related mechanical properties because these are the ones that can vary
from system to system.
The variance in the system energy is calculated as:
2
E2 E2
E
p j E 2j

E2

The above equation can be rewritten in a more convenient form by noting that
1
1
E
E
p j E 2j
E 2j e j
E je j
Q j
Q
j
j
E

ln Q

Given that
ln Q

N ,V

we have
E

p j E 2j

E2

Consequently,
E
2
kT 2
E
T
CV

kT 2CV
N ,V
2
E
2

kT
Heat capacity depends on the energy fluctuation

diamond has relatively low CV

To appreciate the significance of the energy fluctuation, we write


E

kT 2CV
E

From thermodynamics of an ideal gas, we find


11

1
N

1 for N = 1023

PE

The system of a canonical ensemble departing appreciably from E is virtually zero.


Essentially, every system has the same energy E and therefore the canonical ensemble is
equivalent to microcanonical ensemble in practice.
Note that energy fluctuation could be quite significant in nanomaterials or nano-clusters.
Also, fluctuation in density is very significant at critical temperature.
1.4 Microcanonical Ensemble

Assume that the number of quantum states with energy E is


pj

E kT

N ,V , E , we have

E kT

where

is the number of microstates with the same energy E .

From eqn (4), S

ln

and thus we have the best-known equation in statistical mechanics:


S k ln

Interesting, Boltzmann himself never gave thought to the


possibility of carrying out an experiment to measure the
"Boltzmann constant".

12

For any isolated system whatever, the more quantum states available to the system, the higher
the entropy.
1.4.1
Statistical-Mechanical Basis of the Third Law
On the basis of the experimental observations, T. W. Richards and W. Nernst independently
found that for any isothermal process involving only pure phases in internal equilibrium,
lim S 0 ,
T

That is,
k ln 0 ln
where

and

(*)

indicate the degeneracy of the initial and final states at 0 K, respectively.

Furthermore, a value of in the following range


1
eN ,
would be indistinguishable from = 1.
Verification:
For a gas of N particles, the typical order of magnitude of entropy is Nk (or R for
one mole of particles). Even if the degeneracy of the ground state has the same
order of magnitude of N, practically we still can write 0 = 1:
S T S 0 k ln T k ln 0
k ln N

kN

0.

Practically, we have S 0

The entropy of each pure element or substance in a perfect crystalline form is zero at
absolute zero.
Max Planck (the third law of thermodynamics)
1.5 Molecular Partition Function

Consider the ensemble of a single polyatomic molecule. Based on the Born-Oppenheimer


approximation, the Hamiltonian can be decomposed into various degrees of freedom:
H H trans H rot H vib H elec
That is, we assume that the variables of each degree of freedom are independent. Hence, the
energy of the system is the sum of individual energies, and the wave function is a product of
the corresponding wave functions.

exp

Q
i

Eitrans
kT

exp
j

E rot
j
kT

exp
k

Ekvib
kT

exp
l

Elele
kT

Consequently, we can write the molecular partition function as:


q qtrans qrot qvib qele

13

1.5.1
Translational Partition Function
The energy states of a particle in a three-dimensional infinite well can be used to obtain the
partition function of qtrans:
h2
qtrans
nx2 n y2 nz2
exp
23
8mV kT
n x , n y , nz
h2n2
8mV 2 3 kT

exp
n 1

where n is a generic index representing nx, ny, and nz.


The above summation cannot be expressed in terms of any simple analytic function.
Fortunately, for macroscopic V and regular temperature T, the successive terms in the
summation differ by a very small amount and therefore the summation can be replaced by
integration:
exp

qtrans
0

h2 n2
dn
8mV 2 3 kT

Hence,
qtrans

2 mkT
h3

3
2

V
3

where
h
.
2 mkT

For a particle of = kT, the de Broglie wavelength is


h
h
p
2mkT
and therefore is commonly referred to as the thermal de Broglie wavelength.
1.5.2
Electronic Partition Function
By convention, we take the zero of the electronic energy to be the separated, electronically
unexcited atoms at rest:

14

The energy of the molecular ground state would become D0, which can be determined
spectroscopically (the difference between D0 and De is the zero-point energy).
Hence, the partition function should read:
D
E1
qele exp 0 g 0 g1 exp
kT
kT
where the degeneracy is denoted by gi and E1 is the energy difference between the ground
state and the first-excited state.
If we define a characteristic temperature for an electronic transition as
E1
e
k
we would have e in the order of 104 K for E1 = 1 eV.
For example, the degeneracy of the first two levels of halogen atoms are 2P3/2 and 2P1/2, the
probability of finding the atom at the first excited state is:
2 exp
E1 kT
p1
4 2 exp
E1 kT

We have E1 = 0.050 eV for fluorine and 0.94 eV for iodine. At 1000 K, we find p1 = 0.22
and 9 10 6 for F and I, respectively.
1.6 Partition Functions of Many-Body Systems

For an ideal gas system containing n particles, it is a good approximation to write


H tot H 1 H 2
Hn
There are many other problems in physics in which the Hamiltonian, by a proper and clever
selection of variables, can be written as a sum of individual quasi-particles, which
mathematically behave like independent real particles. Examples include photons, phonons,
plasmons, magnons, rotons, and other "ons."
1.6.1
Distinguishable Particles
Denote the individual energy states by

a
i

, where a denotes the particle and i denotes the

energy states. Thus, the canonical partition function is


Q N ,V , T

a
i

b
j

c
k

kT

i , j ,k ,

qa qb qc
where q(V, T) is the molecular partition function.

If all particles are identical, we have


Q N ,V , T

q V ,T

(non-interacting and distinguishable particles)

15

1.6.2
Indistinguishable Particles
In QM, all known particles fall into two classes:
Fermions - wave function antisymmetric under the operation of interchanging two identical
particles
Bosons - wave function symmetric under the operation of interchanging two identical
particles

Consider again the partition function


Q N ,V , T

kT

i , j ,k ,

For bosons, the terms ( 1 + 1 + 3 + ) and ( 1 + 3 + 1 + ) are identical and should not be
counted twice. The corresponding distribution is known as the Bose-Einstein statistics.
For fermions, terms in which two or more indices are the same cannot be included in the
summation. The corresponding distribution is known as the Fermi-Dirac statistics.
That is, the canonical partition functions of bosons and fermions cannot be written as qN.
1.6.3
Boltzmann Statistics
Consider the energy states of a single particle in a three-dimensional infinite well:
h2
nx2 n y2 nz2 ,
nx , n y , nz
2
8ma
where nx, ny, nz = 1, 2, 3,

Let us define
nx2 n y2 nz2

8ma 2
h2

R2

For very large R, the number of molecular


can be well
quantum states with energy
approximated by the number of lattice points of the octant:
3

14
1 4 8ma 2 2
8m
R3
2
h
8 3
8 3
6 h2
where a3 is the volume V.

3
2

Calculation done for one particle in a cube of one liter would give

an order of

30

magnitude of 10 .
Thus, the number of molecular quantum states available to a molecule at room temperature is
much greater than the number of molecules in the system. Alternatively, we can state that the

16

orbital occupancy is small in comparison with unity. This regime, which is favored by large
mass, high temperature, and low density, is known as the classical limit.
In the classical limit, the terms in each summation is much larger than N:
qN

a
i

kT

b
j

kT

c
k

kT

k
N terms

(1 + 2 + 3)(1 + 2 + 3)(1 + 2 + 3):


(1 1 1), (2 2 2), (3 3
(1 2 3), (1 3 2), (2 1
(1 1 2), (1 2 1), (2 1
(1 1 3), (1 3 1), (3 1
(2 2 1), (2 1 2), (1 2
(2 2 3), (2 3 2), (3 2
(3 3 1), (3 1 3), (1 3
(3 3 2), (3 2 3), (2 3
(1 + 2 + 3 + 4)(1 + 2 + 3 + 4):
(1 1), (2 2), (3 3), (4
(1 2), (1 3), (1 4), (2
(4 3)

3),
3), (2
1),
1),
2),
2),
3),
3),

4)
1), (2

1), (3

2), (3

3), (2

4), (3

1), (3

2), (3

1),

4), (4

1), (4

2),

As such, qN will be dominated by terms which have different indices. Hence, the canonical
partition function of fermions or bosons will be given identically as
qN
Q N ,V , T
N!
where we have corrected the sum by N! for N different indices.
The corresponding distribution is known as the Boltzmann statistics.
To assure the validity of the Boltzmann statistics, we have stated that
3

8m 2
V
N
6 h2
Hence, we can rewrite the condition as
8mkT
h2
6

3
2

V N

(*)

Because (V/N)1/3 is a distance of the order of the average nearest-neighbor distance between
molecules, eqn (*) asserts that quantum effects (requiring the Bose-Einstein or Fermi-Dirac

17

statistics) will be absent if neighboring molecules are far apart relative to the "thermal" de
Broglie wavelength.
1.7 Ideal Gas

We can safely apply the Boltzmann statistics for gaseous systems with ideal gas behavior.
1.7.1
Ideal Monatomic Gas
We can obtain the Helmholtz energy as
N

q q
kT ln Q
kT ln trans ele
N!
kT N ln qtrans qele N ln N

For brevity, we assume that qele = 1 (i.e. no electronic excitation and degeneracy),
A
NkT ln qtrans ln N 1
e 2 mk
kT N ln
Nh3

3
2

3
ln T
2

ln V

ln Q

Thus, for one mole of gas particles:


A
NkT
P
V T ,N
V
PV = RT
U

ln Q
T

kT 2
CV

kT 2
N ,V

3N
2T

3
NkT
2

3
RT
2

3
R
2

1.7.2
Chemical Equilibrium in Ideal Gas Mixture
Consider an ideal gas mixture made up of NX molecules of type X and NY of type Y in a closed
container with fixed V and T. We assume no chemical reaction takes place between X and Y.

The canonical ensemble partition function for this binary system is


q XN X qYNY
Q N X , NY , V , T
N X ! NY !
and the Helmholtz free energy is
A
kT ln Q
Suppose a catalyst is added, making possible the reaction
X
2Y

18

Because the system is closed and has constant V and T, the second law states that the system
Helmholtz free energy will decrease as the reaction proceeds. At equilibrium, A will be at
minimum, i.e., Q is at maximum.
For convenience, we state that each X contains two Y so that we have
2 N X NY N
where N is a constant.
Consequently, we obtain
ln Q N X ln q X
N 2 N X ln qY
Hence,
ln Q
NX

N X ln N X

N 2 N X ln N 2 N X

NX

0
N ,V ,T

Knowing that
3

2 mkT 2
qtrans
V,
h3
one can remove the dependence on V by considering the concentration:
Y
X

NY V
NX V

qY V
qX V

2 kT 2 mY2 2
K eq
h3
mX
That is, we have Keq a function of temperature only.
It is also an important point to note that
K eq

qY V
qX V

1.8 Principle of Detailed Balance

1.8.1
State Functions in Thermodynamics
Consider the following reaction:

19

In principle, we could reach


d A
dt

k1 A

d A
dt

0 by having

But then by adding a catalyst, we could affect "[Aeq]". Thus, the steady state

d A
dt

cannot lead to true equilibrium.


According to the principle of detailed balance, true equilibrium can be obtained only having
balance in all steps:
B
C
A
K1 ,
K3 ,
K2 .
A
B
C
That is,
K1 K 2 K 3 1
RT ln K1 ln K 2 ln K 3

RT ln K1 RT ln K 2

RT ln K 3

As discussed in thermodynamics,
G10
RT ln K1
G20

RT ln K 2

0
3

RT ln K 3

Thus,
G10

G20

G30

Gibb's free energy is a state function.

1.8.2
Boltzmann Distribution
Chemical species is a kinetic collection of the microscopic states of a molecular system.

For the chemical reaction of


A
B,
we use the running index i to indicate all the energy levels belonging to A and j to B.
Let ni be the population per unit volume at energy level i, and so on.
At equilibrium, the forward rate of each step is equal to the reverse rate of that step (principle
of detailed balance):
ki j ni k j i n j
Divide both sides by the total population:
n
n
ki j i k j i j
nt
nt

20

ki

k j i p j , where pi and pj are the probabilities of finding the molecule at levels i

pi

and j, respectively.
That is,
p j ki
pi

kj

Because the quotient of reaction rates represents some kind of equilibrium, in analogy to the
relation of G and equilibrium constant, we can therefore write
pj
kT
e j i
pi
Thus, we obtain
1 i kT
pi
e
Q
From the normalization that

pi

1 , we have

kT

We have obtained the Boltzmann distribution from the principle of detailed balance.
(Courtesy of CY Mou)
1.8.3
Statistical Effects on Chemical Equilibrium
Consider the following reaction of isomerization:
A
B
Suppose that some of the energy levels of A are lower than those of B, but that the levels of B
are closer together:

In statistical mechanics, we can take the point of view that a given molecule has accessible to
it the full set of energy states indicated by A + B, with the partition function of
21

A
i

exp

B
j

exp

kT

kT

j
qA

qB

There is a single Boltzmann distribution of molecules among all the levels A + B. Therefore,
the fraction of molecules in all levels belonging to the subgroup A is
A
exp
kT
NA
qA
i
NA

NB

At equilibrium, we have
B
exp
j kT
N B qB
j
A
N A qA
exp
kT
i

Kc

Equilibrium constant can be understood as a ratio of the assessable states of the products and
reactants.

22

Das könnte Ihnen auch gefallen