Statistical Mechanics
Victor Naden Robinson
vlnr500 3rd Year MPhys
17/2/12
Lectured by Rex Godby
Lecture 1: Probabilities
Lecture 2: Microstates for system of N harmonic oscillators
Lecture 3: More Thermodynamics, Boltzmann and Entropy
Lecture 4: Entropy
Lecture 5: Entropy and applications of statistical mechanics
Lecture 6: 2-Level System
Lecture 7: More harmonic oscillator and heat capacity
Lecture 8: Modes
Lecture 9: Debye continued
Lecture 10: The ideal gas
Lecture 11: More ideal gas
Lecture 12: Maxwell-Boltzmann distribution
Lecture 13: Summary and Gibbs approach
Lecture 14: Identical particles
Lecture 15: Bose-Einstein distribution
Lecture 16: Continuation and condensation
Lecture 17: Energy in EM modes
Lecture 18: Black body radiation
Lecture 1: Probabilities
{Note lots of graphs in this course}
Probability of obtaining score “S” on: 1 die
[Figure 1. Plotting P vs. S, dots on the P=1/6 line]
2 dice:
Total Score Average Score Number of Configurations
2 1.0 1 3 1.5 2 4 2.0 3 5 2.5 4 6 3.0 5 7 3.5 6 8 4.0 5 9 4.5 4
10 5.0 3 11 5.5 2 12 6.0 1
Total:
[Figure 2: Prob vs. avg score, 1 to 6 on x-axis, 0 to 1/36 to 1/6 on y-axis, triangle dot formation]
7 Dice:
Total Average Score Number of ways Probability
7 1.0 1 8 8/7 7 9 9/7 21*
*5x(1 dot) + 2x(2 dots):
6x(1 dot) + 1x(3dot):
[Figure 3: Prob vs. score plotted as a Gaussian between scores 1 to 6, FWHM=245660]
Scores obtained: 2.85 3.80 4.70
1024 dice:
[Figure 4: same as figure 3 but very thin (width ~10-12)]
With a large number of components we can describe properties with precision.
We can make predictions about the collective behaviour of large systems without needing to
predict the detailed motion of its components.
Equally likely elementary outcomes
1.1 Microstates
Microstates are QM eigenstate solutions to S.E. but for a whole system rather than an
elementary particle.
For an isolated system the energy is fixed but this energy has degenerate states: W (large
number) microstates with energy, E.
[Figure 4: E on y axis then lots of dashed lines which I suppose is energy levels, circle
enclosing some of the dashes labelled “W microstates”.]
From Fermi’s Golden Rule:
Let pi be the probability of the system being in state I (among the W states):
∑ ∑
(1.1)
∑
(1.2)
When a steady state has been reached,
for all states .
This can happen only if , i.e. all probabilities are equal and all equal
. This is the Principle of
Equal Equilibrium Probability (PEEP).
Refresh: Possible E of single SHM, think about E of N SHM’s.
Lecture 2: Microstates for System of N harmonic oscillators
Classically,
(2.2)
Where is classical angular frequency
Energies
where
[Figure 2.1: of parabola with lines across it showing energy levels, at
going up]
A microstate of system of N oscillators is given by the states each oscillator is in.
The energy of N oscillators is:
(
) (
) (
)
(
)
(
) (2.3)
What is the number of microstates W corresponding to a given value of Q (i.e. to a given total
energy)? Consider:
: Q dots, (N-1) partitions
The no ways W is equal to no ways of arranging the Q dots and (N-1) partitions:
(2.4)
2.1 Thermal Equilibrium
Two systems of N harmonic oscillators
The no microstates corresponding to distribution of energy is:
(2.5)
From PEEP each of these W microstates is equally likely.
In the general case rather than W1 we shall focus on the density of states, i.e. the number of
microstates per unit energy:
(2.6)
Note that where the band of energies within which E is fluctuating.
(2.7)
Lecture 3: More thermodynamics
System:
∙ ∙ ∙ ∙ ∙
E1
W1
G1
E1
W1
G1
energy
The probability of this discussion of energy:
(3.1)
For most likely partition, maximise , that is, maximise:
(3.2)
(
)
(3.3)
Thus for most likely partition of energy (i.e. equilibrium):
(3.4)
This quantity:
is therefore equalised between two systems when they can exchange energy.
[Figure 3.2: Ok I’ll try and draw it!]
Ok that went well.
Define Statistical Mechanical temperature as:
(3.5)*
*For a large system
This T is identical to the ideal gas temperature.
3.2 Boltzmann Distribution
Visa vies systems of constant T.
[Figure 3.3]
𝑙𝑛𝑔 𝜕
𝜕𝐸𝑙𝑛𝑔
𝐸
ER
gR
T
E1
energy
Heat Reservoir
Microstate 𝑖
Because the reservoir is large, T is too good an approximation slowly varying with ER.
(3.6)
By PEEP the probability of the situation shown is:
(3.7)
But:
(3.8)
Since
∑
∑
(3.9)
∑
(3.10)
Z is the sum over all microstates j of the system of the Boltzmann Factor
.
Mean energy of a system at constant T
∑
∑
∑
(3.11)
The mean energy
∑
(3.12)
Recall:
(3.13)
3.3 Entropy
For an isolated system we define:
(3.14)
For a non-isolated system exploring different microstates with probabilities pi (assumed to be
known): {ends}
Lecture 4: Entropy
(4.1)
For the general case where our system explores its microstates with given (generally equal)
probabilities pi, consider N replicas of the system, where N is arbitrarily large. Then ~Np1 systems will
be in microstate 1, ~Np2 in microstate 2, etc.
Total Energy Np1E1 + Np2E2+… is effectively fixedm thus the system is thermally isolated in its
behaviour, exploring states only with a total fixed energy.
Thus the effective W is the number of distinct permutations of Np1 identical objects, Np2 identical
objects, etc.
(4.2)
(4.3)
Using Stirling Approximation:
∑
(4.4)
So the entropy of N replicas is:
∑
(4.5)
For N=1, then:
∑
(4.6)
Entropy of an isolated system
{
(4.7)
(4.8)
Entropy of a system at constant temperature
∑
(4.9)
∑
∑
(
)
∑
∑
∑
(4.10)
Recall Helmholtz free energy:
(4.11)
Thermodynamic entropy
Consider a system thermally isolated on which work can be done. The total internal energy is:
∑
(4.12)
When the volume of the system is changed:
∑
(4.13)
The RHS (and last term) must be associated with heat.
Lecture 5: Entropy and applications of statistical mechanics
If a system is insulated (thermally isolated):
(5.1)
Now,
∑
(5.2)
For a slow change in constraints, FMG shows that no additional transitions between microstates are
induced, i.e. dpi = 0 for all
∑
(5.3)
For an insulated system:
(5.4)
Since ∑ , the other term in (4.2) must be the heat:
∑
(5.5)
At constant temperature, the probabilities, pi, are given by the Boltzmann distribution:
∑
Also (4.6): ∑
∑{
}
∑
∑ ∑
∑ {
}
∑
(5.6)
Thus statistical mechanical and thermodynamic entropy are identical.
5.1 Applications of Statistical Mechanics
Vacancies in crystals
For N atoms, how many vacancies will there be at temperature T?
Each vacancy costs an amount of energy ( ) due to not being bonded fully. Free energy F is
to be minimised:
(5.7)
So when volume and temperature are held constant, hence we minimise not
What are and for a given ?
(5.8)
Provided a system is sufficiently large, it’s energy is effectively fixed as if it were thermally isolated
so we may use:
W is the number of arrangements of N atoms, and vacancies on a total of sites:
(5.9)
{ }
{ }
{ }
(
) (
)
(5.10)
Since ,
(5.11)
Note:
At 300K,
At 3000K,
Practical crystals at room temperature are not yet at thermal equilibrium but reflect vacancy
concentrations from temperatures ( ) where the crystal first formed.
Lecture 6: 2-Level System
E.g. each magnetic atom/ion in a dilute magnetic semi-conductor
If
then
∙
Because
(6.1)
Take
(6.2)
(6.3)
Low T:
High T:
So,
(
)
(6.4)
0
𝜇𝐵
𝜇𝐵
𝐵
0
1
P
p1
𝐾𝑇 𝜇𝐵
T
𝐾𝑇 𝜇𝐵
Diagram:
Heat capacity
(6.5)
Mean magnetic dipole moment
So the magnetic polarisation with n magnetic ions per unit volume of crystal is (
)
Harmonic Oscillator
(
)
[
]
𝐸
T
𝜇𝐵
𝑘𝑇 𝜇𝐵
C
T
Schottky
Anomaly
𝜇
T
𝜇
0
𝜔
𝜔
𝜔
So
(6.6)
This is a geometric series
(6.7)
Lecture 7: More Harmonic Oscillator and Heat Capacity
(7.1)
Harmonic oscillator continued,
(7.2)
(7.3)
(7.4)
This is the Planck oscillator formula (eqn. 7.4)
At low T ( );
At high T ( );
But
(7.5)
An example of the classical equipartition theorem: A mean energy of
per degree of freedom in
the total energy is proportional to position2 or velocity2.
Heat capacity of oscillator
( )
(
)
( )
(7.6)
Graph:
Vibrational energy of a crystal
First consider two atoms in one dimension
2 modes:
𝐸
T
𝜔
𝑘𝑇
𝐶
T
𝜔
𝑘
𝑘
𝑚
𝑚
𝜆
𝜔 𝜆/𝑚
And
Arbitrary motion of the atoms can be written as a linear combination of these normal modes.
Next: 3 atoms in 1D
Now 3 modes:
N atoms in 3D have 3N modes each with a well-defined frequency .
Lecture 8: Modes
For a periodic crystal the modes can be classified by their wave vectors q. Displacement of atom at
position is:
( )
(8.1)
‘Zero’ boundary conditions give a wave of zero when .
(8.2)
Cubic lattice of points in q-space, each one representing an allowed value of q
𝜔
𝜔
𝜔
𝜔
𝐿
𝐿
𝐿
𝐶
𝜋
𝐿
𝜋
𝐿
𝜋
𝐿
𝑞𝑦
𝑞𝑥
For a realistic solid:
We study two simplified models:
Einstein model
Debye model
Einstein model
3N simple harmonic oscillators
(
)
(8.3)
(8.4)
𝜔
𝑞
𝑏𝑜𝑢𝑛𝑑𝑎𝑟𝑦 𝑜𝑓 𝑠𝑠𝑡𝐵𝑟𝑖𝑙𝑙𝑜𝑢𝑖𝑛 𝑧𝑜𝑛𝑒
𝜔
𝑞
𝜔𝐸
𝑁 𝑚𝑜𝑑𝑒𝑠
𝜔
𝑞
𝑠𝑙𝑜𝑝𝑒 𝑐
𝑞𝑚𝑎𝑥
𝐸
𝑞𝑚𝑎𝑥
𝐶
𝑞𝑚𝑎𝑥
𝑁
𝜔
𝑁𝑘
At low T,
(
)
(8.5)
Einstein model good except at low T (see hand out).
Debye model
The n. of q points per unit volume of q space is
(
)
(8.6)
The volume of the sphere in q-space is
The volume of the positive octant of this sphere is:
(
)
(8.7)
Lecture 9: Debye model continued
Picture of positive octant of sphere in q space with a strip on the surface
For the Debye model
(
)
We had, for a single oscillator {lecture 7}:
Adding over the modes, with
N. of q points per unit volume of q space is
Contribution to heat capacity from modes in the shell of thickness is
∙
∙ ∙
[1st term from volume of octant shell, 2nd term is n. of points per unit volume, 3rd is modes per q]
(9.1)
Looking at
( )
(9.2)
∫
(9.3)
Using:
Sub into (9.3)
∫ (
)
(9.4)
[Note the limits change too]
∫
(9.5)
[Graph showing this integrand vs. x and the tails and for low T and high T]
Therefore for high T, integrand and so integral [
]
(
)
(9.6)
As expected for 3N oscillators with high T and heat capacity k.
For low T: let since rest of tail (graph) contributes negligible amount to C. Law of
contour integration:
∫
(9.7)
(9.8)
[Graph comparing C vs. T for Debye and Einstein]
[Graph of omega vs. q showing straight line relationship and something about not being frozen]
Sometimes we prefer to use as the parameter in the Debye model not c but the Debye frequency
Lecture 10: The ideal gas
Gas molecules move around independently:
Ignore long-range Van der Waals attraction
Ignore scattering (except to allow molecules to explore all accessible microstates)
Initially we study monatomic gases. Partition function of ideal monatomic gas is required.
For one atom in a box:
∑
(10.1)
For two non-identical atoms:
∑∑
(10.2)
For two identical atoms:
(10.3)
∑∑
{
(10.4)
For three atoms:
∑∑∑
{
(10.5)
For N atoms:
∑∑
∑
{
(10.6)
In an ideal gas, most of the states are empty so as long as the temperature is not extremely low, an
atom has so many particle states to choose from that a state with more than one atom in it is
extremely unlikely.
To a good approximation we can apply the
correction factor to all cases. Thus:
∑∑
∑
(10.7)
(∑
)(∑
) (∑
)
(10.8)
Thus Z for N atoms is:
(10.9)
Schrödinger eqn:
(10.10)
Choosing boundaries for which ; the solution is:
( )
(10.11)
Where
The energy Eigen function is:
(10.12)
Volume of the octant shell of radius thickness in q space is:
(10.13)
Number of states per unit q space is:
(10.14)
∫
(10.15)
Lecture 11: More ideal gas! Continues from L10
Continuing on, define
√
√
(11.1)
∫
(
√
)
(√ )
∫
And since ∫
√
(√
)
(
)
(11.2)
So
(
)
Free energy is given by
[ (
)
]
[ (
)
(
) ]
(11.3)
Note that F is extensive: doubling both V and N simply doubles F.
Pressure of ideal gas
Recall
(
)
(11.4)
[ (
)
(
) ]
(
)
(11.5)
This establishes the equivalence of stat mech. and (ideal gas) thermo-dynamical temperatures.
Energy of ideal gas
[ (
)
(
) ]
(11.6)
I.e. energy of
per gas molecule 1 atom, as predicted by classical equipartition theorem:
per
K.E. term
(
)
Note that our “non-degenerate” approximation that led to
assumes that T is not too low,
hence we are already sufficiently hot enough for equipartition to apply.
Heat capacity of an ideal gas:
(11.7)
Maxwell-Boltzmann distribution
The probability distribution function of c: the speed of the gas molecule
(11.8)
If speed c corresponds to
then corresponds to
Lecture 12: starts next page
Stat Mech cont.
Lecture 12:
Opens with another three dimensional picture
(12.1)
From the Boltzmann distribution, probability of an atom (only one for now) being in a
particular state q is:
(12.2)
(
)
(12.3)
(
)
(12.4)
Now
(12.5)
(12.6)
Total probability that speed is in range is:
(
)
Cancelling and subbing in c
(
)
(
)
(12.7)
Leading to the Maxwell-Boltzmann distribution:
(
)
(12.8)
NB:
is the Boltzmann factor for a state; is proportional to number of
states ; the prefactor could be deduced from normalisation of the probability
distribution:
∫
(12.9)
[Two figures of f(c) vs c for various Temps]
Molecular gases: e.g.
Vibration: single SHO with frequency
(12.10)
( )
(12.11)
Typically
therefore we can use a low temp approximation:
(12.12)
Rotation:
[Picture of degrees of freedom for diatomic molecule]
Eigenvalues
at normal temperatures, therefore use high T approximation
From classical equipartition theorem
(12.13)
So for a gas of N such molecules, the total heat capacity is:
Where the 1st term is from translation, 2
nd is from rotation and 3
rd is vibration
(12.14)
Entropy of ideal gas (monatomic)
Get S from the free energy eqn using
So
(
)
(12.15)
There was a problem question which helps practice this part well.
As but our non-degenerate approximation breaks down at low temperatures
Lecture 13: Key Theory so far
Bit of recap with some new stuff too, aids course completeness. If you can’t remember what
these are then go back and learn them:
Principle of equal equilibrium (PEEP)
at equilibrium
Boltzmann distribution:
∑
With partition function
∑
Recall
For an isolated system the entropy is
Other it is (Gibb’s formula)
∑
At constant ,
Harmonic oscillator (Planck oscillator formula)
Now lecture moves on to new topic based on everything so far
Systems with variable numbers of particles
Chemical potential
[Figure similar to before ill copy later]
Most likely division of energy will maximise:
(13.1)
Consider
(13.2)
(13.3)
Define
(
)
(
)
(13.4)
Thus both temperature and the chemical potential are equalised in equilibrium. has
dimensions of energy. Since
For a large system, then
(
)
(
)
(13.5)
Comes from
[Now adds or amends original diagram to concern as well]
So now
(
)
(
)
(
)
(
)
(
) (
)
(13.6)
New version of eqn
(13.7)
(
)
(
)
(
)
(13.8)
(13.9)
(13.10)
(
)
(13.11)
Gibbs distribution
For a system able to exchange energy and particles with a reservoir
[Figure of a system connected to a heat reservoir]
For the reservoir
(
)
(
)
(13.12)
This gives
/
(13.13)
From PEEP:
/
Since
But thus
/
(13.14)
Absorbing constants into proportionality (Gibbs distribution)
Since ∑ :
/
∑ /
/
(13.15)
Where is the Grand Partition Function
Lecture 14: Identical particles
We ignore, for simplicity any interaction between particles (e.g. coulomb repulsion between
electrons). Specify how many particles in each 1-particle quantum state.
For Bosons (spin 0, 1, 2, ...) (e.g.) photon, hydrogen atom) there is no restriction on the
occupation.
For Fermions (spin ½, 3/2, ...) (e.g. electrons, protons, neutrons) the Pauli Exclusion Principle
prohibits two or more fermions in the same 1-particle state.
[Figure of particles in states for bosons]
[Similar figure isolating all other particles (bar one and its state) as a reservoir (for fermions)]
N.B.1-particle state means spatial wave function together with spin, e.g. 1s state with spin up
Key concept: Treat the 1-particle state as our system with the other states playing the role of
the particle reservoir (figure). The whole system is to be held at temperature T.
Apply Gibbs distribution:
(1) For fermions
N E
0 0
1
/
(14.1)
Gives Fermi-Dirac distribution:
(14.2)
Note
Where is the average number of particles in a state with energy .
(2) For bosons
/ / /
(14.3)
(14.4)
(14.5)
(
(
)
)
[I think that step is done using another geometric progression (the square term etc.)]
(14.6)
This is the Bose-Einstein distribution
(14.7)
[Two graphs describing each dist. For a range of temperatures]
NB. For the purpose of graph only the T-dependence of is ignored
Lecture 15: Bose Einstein Distribution
(15.1)
[Figure plotting with asymptote]
Asymptote at . There fore must be less than the lowest energy of the 1-particle states.
is the number of particles with energy
The total number of particles, N, in the system is known. This means that summed over
all the energies, , must yield a total of N:
∑
(15.2)
This applies for both fermions and bosons depends on (albeit weakly).
The free-electron metal
is given by solving the Schrödinger equation:
[Cubic box of lengths L]
(15.3)
The density of these single-particle quantum states is:
(15.4)
Number of states:
1st term is the positive octant of spherical shell in q-space, 2
nd term is number of q-points per
unit volume of q-space)
(15.5)
(15.6)
/ /
/ √
(15.7)
BUT: for each solution to the S.E. eqn there are 2 modes (spin)
(15.8)
[Two graphs showing a step function and a rising graph]
∫
(15.9)
This fixes at any temperature. At , f is a step function:
∫
∫ /
/
(15.10)
(
)
(
)
(15.11)
This value of at is called the Fermi Energy,
[Figure displaying Fermi energy on a Fermi-Dirac graph]
Lecture 16: Continues really
Roughly half of the electrons within below the Fermi energy, , have been raised in
energy by ~ . The number of such electrons is:
(16.1)
In fact careful integration gives
1st term is the electrons, 2
nd is vibrational
[Graph showing these two terms plotted]
Bose-Einstein condensation
[Figure of moving for BE distribution; for details, see Mandl.]
Below what critical temperature are a significant fraction of the bosons in the very lowest
quantum state?
For particles in a box:
√ /
/
(16.2)
Where the factor of 2 {sure a half?} undoes spin degeneracy
For all states but the lowest
so the total of particles in all the states but the
lowest is:
∫
∫
∫ /
(16.3)
Approximate done as the step function has width of
(16.4)
For condensation, this is , the total of particles
(16.5)
(
)
(16.6)
Careful integration (some method, Canto?) gives:
[Figure with two graphs showing proportion of occupied states with temperature, with
being the fraction of bosons in very lowest state]
Blackbody radiation
[Cube again with lengths L]
Electric field
At boundary so
(16.7)
Mode has
Each mode is a simple harmonic oscillator we can apply Planck oscillator formula:
Lecture 17: Energy in EM modes
Useful in Electrons in solids to know these parts well – actually S.M. ties into most modules
Using all this
∫
(
)
(17.1)
Integral of 1st term gives infinity, but this is the zero point energy and is independent of T
with no observable consequences.
( )
(
)
∫
(17.2)
∫
( )
(17.3)
Spectral density
Exclude
(17.4)
Since
(17.5)
Low :
High
Lecture 18: Blackbody Radiation Continued
Energy density:
( )
(18.1)
Spectral density:
( )
(18.2)
Absorption and emission
[Figure of temperature being absorbed and emitted in a square, also a figure of a semi circle
with various parts labelled in polar co-ordinates]
Blackbody: Zero reflection, absorbs all incident radiation (and subsequently re-emits)
Consider EM radiation arriving at angle range between and to the normal. The
fraction of the total modes considered is:
(18.3)
[Another figure with some light hitting a surface and reflecting possibly, the angles and
lengths are labelled. In retrospect the previous diagram was also describing light hitting a
surface, see audio lecture for description until diagrams drawn here]
Energy arriving on area A (figure) per unit time is:
(18.4)
Total for all modes (therefore all angles)
∫
/
(18.5)
The energy flux is then
And is called the Stefan-Boltzmann constant
This flux is also the energy flux EMITTED by a blackbody at temperature T.
Classical statistical mechanics
Phase Space: For N particles in D dimensions there are DN position variables and DN
momentum coordinates, giving 2DN dimensional phase space.
For a 1D particle in an infinite well:
[Figure with lengths 0 to L and then P vs x plot related to it]
QM shows that the volume between successive quantised states is .
Boltzmann and Gibbs and co. assumed one can integrate over phase space to achieve what in
QM is given by a sum over microstates. This was later justified by the above result.
Fin