Equilibrium statistical mechanics

Probability theory
Grand canonical distribution
Thermodynamic equilibrium
Fermi-Dirac distribution
Fermi energy
Probability theory reminder

• Probability of an event α OR α' : pα OR α ′ = pα + pα ′

• Probability to have any outcome: p1 OR 2 OR 3 OR... = p1 + p2 + p3 + ... = ∑ pα =1

α

• Probability to have two outcomes α AND α'

pα AND α ′ = pα pα ′

provided that those outcomes are independent

• Example: Two dices are rolled and the total number of dots is counted.
What is the probability for each outcome (2, 3, 4, …, 12)?

Solution: There are N = 36 possible outcomes.

Only one way to have a 2 ( = 1+1)  N2 = 1, p2 = 1/36.
Two ways to have a 3 (= 1+2 = 2+1)  N3 = 2, p3 = 2/36 = 1/18.
Three ways to have a 4 (= 1+3 = 2+2 = 3+1)  N4 = 3, p4 = 3/36 = 1/12.
For n ≤ 7 : pn = (n-1)/36.
For n > 7 : pn = p14-n.
Probability theory example

The degeneracy of a particular energy level g = 4. The probability for a particular

quantum state belonging to this level to be occupied by an electron is f = 0.2. What is the
probability for this level to have 0, 1, 2, 3, and 4 electrons?
Microstates

• The system = a huge collections of particles

• Microstate = the most complete information

about the system

• For a completely isolated system, knowledge

of its microstate at the initial time allows to
predict the microstate at any future time

• Classical mechanics: microstate = all positions and velocities of the system’s

particles

• Quantum mechanics: the collective wave function of all the system’s particles

• But practical determination of the microstate is impossible, because:

– The system is chaotic
– The system is not completely isolated
– The initial microstate is unknown
– The system is too big
Probabilistic description

• Label different microstates with a single letter α

• Probability of a microstate α : pα

• Macroscopic property Y in microstate α : Yα

• Averaged value Y = ∑ pαYα

α

• The postulate of equal a priori probabilities:

In thermodynamic equilibrium all microstates that are consistent with the system’s
macroscopic properties (volume, energy, number of particles) are equally probable

No mathematical proof of this statement exists, but the predictions that follow from
this postulate are confirmed by all experiments known
Grand canonical distribution

• A system of interest is in contact with a much bigger

object, which we will term the “heat bath”.
In particular, the system may be a small part of a big object

• The system and the heat bath can exchange particles

and energy. That exchange affects the number of particles and the energy of the
system much more strongly than those of the heat bath, because
Nheat bath >>> nsystem and Eheat bath >>> Esystem

• The system’s microstates are denoted as α. They depend on the # of particles in the
system

• What is the probability pnα to find the system in an n-particle state α?

• Out of equilibrium, the answer depends on the concrete details of the system and the
heat bath

• In equilibrium, the answer is general and universal

Grand canonical distribution

• Postulate of equal a priori probabilities:

if, for a fixed number of particles, two different system

states α and α' have the same energy, Enα = Enα' ,
then they are equally probable in thermodynamic equilibrium
pnα = pnα' .

• This implies that pnα may depend on n and α as pnα = p( n, Enα )

with p(n, Enα) being some universal function of the number of particles and energy.

• Consider two identical systems in contact with the same

heat bath, but not with each other. The two systems are
independent. What is the probability to find one in the state
(n α) and the other in the state (n' α')?
Answer: p( n, Enα ) p( n′, En′α ′ ) = p( n + n′, Enα + En′α ′ )
Grand canonical distribution

• So, the probability pnα = p( n, Enα ) enjoys the property

p( n, Enα ) p( n′, En′α ′ ) = p( n + n′, Enα + En′α ′ )

1
• But this is only possible if pnα = exp( An + BEnα )
Z
with constant A and B and the normalization constant

Z = ∑ exp( An + BEnα )
nα

called the partition function.

Grand canonical distribution
1
• The expression pnα = exp( An + BEnα ) is usually written in terms of
Z
thermal energy: kT , where k = 1.38·10-23 J/K = Boltzmann’s constant

chemical potential: µ
1  µn − Enα 
as pnα = exp 
Z  kT 

 µn − Enα 
• Grand partition function Z = ∑∑ exp 
n α  kT 

• Thermal energy and the chemical potential are related to the average energy and the
number of particles in the system:
n = ∑∑ n pnα ; E = ∑∑ Enα pnα
n α n α
Thermodynamic equilibrium

• The average energy and the average energy and the number of particles in one
system:

n 1 = ∑∑ n pn(1α) ; E 1 = ∑∑ Enα pn(1α)

n α n α
In the other system, which is identical to the first one:

n 2
= ∑∑ n pn( α2 ) ; E 2
= ∑∑ Enα pn( α2 )
n α n α

• In thermodynamic equilibrium we must have: n1= n 2

; E 1= E 2

• But n and E are determined by the chemical potential, μ, and the

temperature, T

• In thermodynamic equilibrium the chemical potential and temperature are the same
everywhere. Non-uniformity of these parameters results in an onset of energy and
particle currents that restore equilibrium.
Fermi-Dirac distribution

• Consider a quantum state α. It may or may not be occupied by an electron. It may

not admit more than one electron. How many electrons does it have on average?

• The state itself may be regarded as “the system”. The energy of this system is

1  µn − nEα 
Enα = nEα , n = 0, 1 pnα = exp 
Z  kT 
• Average occupation number =

1 ( µ − Eα ) / kT
nα = 0 ⋅ p0α + 1 ⋅ p1α = p1α = e
Z
Z= ∑ e n ( µ − Eα ) / kT

n =0, 1
= 1 + e ( µ − Eα ) / kT

Fermi-Dirac distribution:
e( µ − Eα ) / kT 1
nα = p1α = f ( Eα ) = ( µ − Eα ) / kT
=
1+ e 1 + e( Eα − µ ) / kT
Fermi-Dirac distribution
1
• Fermi-Dirac distribution: nα =
1 + e( Eα − µ ) / kT

• The “chemical potential” is the term preferred by statistical physicists. In solid-state

physics, it is usually called the “Fermi energy”.

µ ≡ EF
• Occupation probability of a quantum state with
energy E
1.0
T=0

1 T = 0.1 EF
f (E) = T = 0.2 EF
1 + e( E − EF ) / kT T = 0.3 EF
f (E)
0.5

• It is not probability to have energy E,

because there may be many states with
this energy. It is the probability that a particular
quantum state with energy E is occupied. 0.0
0 E1F
E
Fermi-Dirac distribution: excercise

A particular energy level E has the degeneracy 6. What should be the energy difference
ΔE = E – EF for this energy level to be empty with probability 1/3 at T = 300 K?
Determination of Fermi energy
1 T=0
f (E) =
1.0
• Fermi-Dirac distribution
1 + e( E − EF ) / kT
T = 0.1 EF
T = 0.2 EF
T = 0.3 EF

f (E)
∑
0.5

• Total # of particles: f ( Eα ) = N ⇒ EF
α
0.0
Can solve for EF numerically at any T. 0
E
E1F

0 if E > E F
• At T  0: f (E) = 
1 if E < E F

• Because f ( EF + ∆E ) + f ( E F − ∆E ) = 1

EF must be just in between the highest occupied state and

the lowest unoccupied state
Determination of Fermi energy: sample problem

A system consists of 15 non-interacting spin-1/2 particles with single-particle energies

En = ΔE·(n2 + m2), where ΔE = 1 eV and n, m = 1, 2, 3, ….
Determine EF at zero temperature.