Stat Mech Notes

MIT OpenCourseWare
http://ocw.mit.edu
8.044 Statistical Physics I

Spring 2008
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
8.044 STATISTICAL PHYSICS I
Phase T=0 Quantum gases

Transitions
Canonical ensemble
Grand 3rd Law
Canonical
Ensemble 2nd Law
Language: Microcanonical ensemble
Probability 1st Law Transport

Processes
8.044 L1B1
PROBABILITY
Random variable (ignorance and/or QM)

Continuous, discrete, or mixed
Probability density: p(x) ↔ px(ζ)
Histogram
(normalized) p(x)
PROB(ζ ≤ x < ζ + dζ) = px(ζ)dζ
8.044 L1B2
⇒ px(ζ) ≥ 0,
∞
px(ζ)dζ = 1,
−∞
b
PROB(a ≤ x < b) = px(ζ)dζ
a
Cumulative probability:
ζ d
Px(ζ) ≡ px (ζ )dζ ⇒ px(ζ) = Px(ζ)
−∞ dζ
Either px(ζ) or Px(ζ) completely specifies the RV x.
8.044 L1B3
Example Physical adsorption of a gas
most of the time ( when hot)
z
v
n
θ
y
x
φ
small fraction of the time (when hot) leaving the surface
8.044 L1B4
p( φ)
p(v) v2
v3
e 2σ2
2σ4
1
2π
σ= kT
m
2π φ 3σ v
p( θ) p(t)
2sin( θ)cos( θ)
1 t τ
τe
π/2 θ τ t
8.044 L 1B5a
p( φ) p(v)
1.0
P(φ) 1.0
P(v)
2π φ 3σ v
p( θ) p(t)
1.0 P(θ) 1.0 P(t)
π/2 θ τ t
8.044 L1B5b
z
dΩ
PROB = p(θ)dθ p(φ)dφ
= 2 sin(θ) cos(θ)dθ(1/2π)dφ θ
dθ
dΩ = sin(θ)dθdφ y
φ
x dφ
PROB/dΩ = (1/π) cos(θ)
8.044 L1B6
Gaussian density (memorize)
p(x) h
(x−x0 )2 1
− he 2 = 0.61 h
p(x) = √ 1 e 2σ 2
2πσ 2
x0 x0 + σ x
2 parameters
8.044 L1B7
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Gaussian density (memorize)
p(x) h
(x−x0 )2 1
− he 2 = 0.61 h
p(x) = √ 1 e 2σ 2
2πσ 2
x0 x0 + σ x
2 parameters
8.044 L1B7
Example Atom escaping from a cavity
A Hole Atom escapes after nth wall

encounter
p(n) = ( AH )(1 − AH )n
A T A T
A Total n = 0, 1, 2, · · ·
8.044 L2B1
∞

pn(x) = (AH )(1 − AH )nδ(x − n)
AT A T
n=0
Called a geometric or a Bose-Einstein density
Pn(x)
p n(x)
0 2 4 x 0 2 4 x
8.044 L2B2
Example Mixed, t dependent RV
p(x)
Chemical adsorption
e- t / τ
Physical adsorption
x x
P(x)
1
Given: atom on bottom at t = 0
e- t / τ
p(x) = e−t/τ δ(x)+(1−e−t/τ ) f (x)

x
8.044 L2B3
Averages
p(x)
∞
< f (x) >≡ f (x)p(x) dx
std.
−∞
dev.
<x> x
< x > is the mean
< x2 > is the mean square
< (x− < x >)2 > = < (x2 − 2x < x > + < x >2) >
= < x2 > −2 < x >2 + < x >2
= < x2 > − < x >2 is the variance
≡ (standard deviation)2
8.044 L2B4
Gaussian
Exponential
p(t)
1 -t/τ
−(x−x 2 /2σ 2 τe
p(x) = √ 1 e 0 )
2πσ 2
<τ> t
< x >= x0 < t >= τ

Var(x) = σ 2 Var(t) = τ 2
Controlled separately Determined by same
parameter
8.044 L2B5
θ
Example Mean free path
d n L=d/cos θ
π/2
<L> = (d/ cos θ)p(θ) dθ
π/2
π/2
= (d/ cos θ) 2 sin θ cos θ dθ = 2d sin θ dθ
0 0
π/2
= 2d [0 (− cos θ) = 2d
8.044 L2B6
Poisson density Events occur randomly along a line
at a rate r per unit length
L ∆x
x
p(1) → r∆x as ∆x → 0
Events are statistically independent
1 −rL 1
n
p(n) = (rL) e = < n >n e−<n>
n! n!
8.044 L2B7
Examples of Poisson probability densities
0.6
0.25
0.5
<n> = 0.5 0.2 <n> = 2.3
0.4
0.15
0.3
0.1
0.2
0.1 0.05
1 2 3 4 5 2 4 6 8 10
0.12 0.07
0.1 <n> = 10 0.06

<n> = 30
0.05
0.08
0.04
0.06
0.03
0.04
0.02
0.02 0.01
5 10 15 20 25 30 10 20 30 40 50
8.044 L2B8
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
p x,y ( ζ,η)
p x( ζ)
0 ζ
η
ζ
8.044 L3B1
px,y (ζ, η)dζdη ≡ PROB(ζ < x ≤ ζ+dζ and η < y ≤ η+dη)
Px,y (ζ, η) ≡ PROB(x ≤ ζ and y ≤ η)

ζ η
= px,y (ζ , η )dζ dη

−∞ −∞
∂ ∂
px,y (ζ, η) = Px,y (ζ, η)
∂ζ ∂η
8.044 L3B2
∞ ∞
f (x, y) = f (ζ, η)px,y (ζ, η)dζdη
−∞ −∞
NEW CONCEPTS:
Reduction to a single variable
∞
px(ζ) = px,y (ζ, η)dη
−∞
Conditional probability density
px(ζ |y)dζ ≡ prob.(ζ < x ≤ ζ + dζ given that η = y)
8.044 L3B3
1 2
( ) = ( |y) )
( )
8.044 L3B4
∞

1 px,y (ζ, y) dζ = py (η = y)
−∞

∞

2 c px(ζ |y) dζ = c
−∞
1
⇒ c = py (η = y)
px,y (ζ, y) p(x, y)
px(ζ|y) = or
p(x|y) =
py (η = y) p(y)
Bayes’ Theorem or “fundamental law of conditional
probability”
8.044 L3B5
p(x, y) = p(x|y)p(y)
p(x, y) = p(y|x)p(x)
x and y are statistically independent if p(x, y) = p(x)p(y)
⇒ p(x |y) =

p(x)

p(y|x) = p(y)
conditioned unconditioned
8.044 L3B6
y
2 2 FO
a a -y
a x
�= = O

p(x, y) = 1/πa2 p(x, y) = 0 p(y) = 2 1 − ( y )2
πa a
8.044 L3B7
1
p(x|y) = |x| ≤ a2 − y 2
2 a2 − y 2

= 0 |x| > a2 − y 2
p(x|y)
x
a2 -y 2 a2 -y 2
8.044 L3B8
Derivation of Poisson density p(n;L)
Given: p(n = 1; ∆x) → r∆x as ∆x → 0
Start by finding p(n = 0; L)
x
0 dL L
dL → 0 ⇒ p(0) p(1) p(n > 1)

⇒ p(n = 0, dL) ≈ 1 − p(1) = 1 − r dL
8.044 L3B9

p(n = 0; L) = (1 − r dL) statistical independence

ln p(n = 0; L) = ln(1 − r dL)
L
→ (−r dL) → (−r dL) = −rL
0
p(n=0;L)
p(n = 0; L) = e−rL
0 r -1 L
8.044 L3B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Poisson Random Variable
• S.I. events r(x)
• p(1) in ∆x → r∆x
• r is constant x
p(n = 0; L) = e−rL 1st element of p(n; L)
8.044 L4B1
L L+∆L
0 x
p(n; L + ∆L) ≈ p(n; L) p(0;

∆L) +p(n − 1; L) p(1;

∆L)

(1−r∆L) (r∆L)
p(n; L + ∆L) − p(n; L) d p(n; L)

= rp(n−1; L)−rp(n; L) →
∆L dL
8.044 L4B2
d p(n; γ)
Let γ ≡ rL + p(n; γ) = p(n − 1; γ)

dγ
1st order, linear DE already known
DE: γ is the variable and n is an index

Probability: n is the variable and γ is a parameter
1 n −γ
p(n; γ) = γ e
n!
8.044 L4B3
Example Jointly Gaussian random variables
Random processes:
Noise voltage or current v(t), i(t)
Thermodynamic variable P (t), T (t), ρ(t)
Thermal radiation E(t), B(t)
Seismic background signal s(t)
8.044 L4B4
v(t) v1
v2
t1 t 2= t 1+τ
 
1 v 2
− 2ρv v + v 2
p(v1, v2) = exp − 1 2 1 2 2 2

2πσ 2 1 − ρ2 2σ (1 − ρ )
σ is constant ρ = ρ(τ ) → 1 as τ → 0
→ 0 as τ → ∞
8.044 L4B5
v2 v2 v2
ρ ≈1 ρ ≈ -1 ρ =0
v1 v1 v1
v 2 and v 2
are S.I.
8.044 L4B6
 
1 v 2
p(v1) = √ exp − 12 
2πσ 2 2σ
a zero mean gaussian with variance σ 2
 
1 )2
(v2 − ρv1 
p(v2|v1) = exp −
2πσ 2(1 − ρ2) 2σ 2(1 − ρ2)
a Gaussian with mean ρv1 and variance σ 2(1 − ρ2)
As ρ goes from 1 → 0, the mean goes from v1 → 0
and the variance goes from 0 → σ 2.
8.044 L4B7
δ(v 2 -v 1 ) when ρ =1
p(v 2 |v 1 )
v2
0
ρ
8.044 L4B8
τc T
v(t)
amplitude and phase drift
ρ( τ)
1
τ
τc
T
-1
8.044 L4B9
τc
v(t)
ρ( τ)
1
τ
τc
-1
8.044 L4B10
Averages
∞ ∞
< v 1 v2 > = v1v2 p(v1, v2) dv1 dv2
−∞ −∞
∞ ∞
= v1v2 p(v2|v1) p(v1) dv1 dv2
−∞ −∞
∞ ∞
= v1p(v1) v2p(v2|v1) dv2 dv1
−∞ −∞
conditional mean = ρ(τ )v1
∞
= ρ(τ ) v12p(v1) dv1 ⇒ ρ =< v1v2 > /σ 2
−∞
<v 2>=σ 2
8.044 L4B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Functions of a random variable
f(x)
Given: px(ζ) and f (x)

Find: pf (η)
x
8.044 L5B1
A. Sketch f (x). Find where f (x) < η
B. Integrate to find Pf (η).
C. Differentiate to find pf (η).
8.044 L5B2
Example Intensity of light
A I
aε2
I = aE 2
η
p(E ) = √ 1 exp[−E 2/2σ 2]
2πσ 2
η 0 η ε
a a
B

√
η/a
PI (η) = √ pE (ζ) dζ
− η/a
8.044 L5B3
∂g
g ∂y
dy
db
dy
da dy
dy
dy
x
a(y) b(y)
d b(y)
g(y, x) dx =
dy a(y)
db(y) da(y) b(y) ∂g(y, x)

g(y, x = b(y)) −g(y, x = a(y)) + dx
dy dy a(y) ∂y
8.044 L5B4
C In general
1 1 1
1

pI (η) = √ pE ( η/a) − (− √ ) pE (− η/a)
2 ηa 2 ηa
1 1

= √ [pE ( η/a) + pE (− η/a)]
2 ηa
5
2 π aσ2 p(I)
4
In our particular case
1 I
p(I) = √
2
exp[− 2
]
2
2πaσ I 2aσ 1 I/ a σ2
0 1 2 3 4
8.044 L5B5
Example Black Body Radiation
1 1 (ν/ν0)2
p(ν) =
2ζ(3) ν0 exp[ν/ν0] − 1

1/2.404
T
ν0 = kT /h
p( ν)
∝ ( ν/ν0 )
( ν/ν0 )
∝ ( ν/ν0 ) 2 e
8.044 L5B6
Given λ = c/ν and p(ν)
Find p(λ) λ
λ=c/ν
η
B
∞
Pλ(η) = pν (ζ) dζ 0 c/η ν
c/η
8.044 L5B7
C
In general
pλ(η) = −(−c/η 2) pν (c/η)
In our case
c 1 1 (c/ην0)2
pλ(η) = 2
η 2.404 ν0 exp[(c/ην0)] − 1
Let λ0 ≡ c/ν0, then

 −4
1 1 
η

1
pλ(η) =
2.404 λ0 λ0
exp[(η/λ0)−1] − 1
8.044 L5B8
1 −(λ/λ ) −1
As (λ/λ0)→0
exp[(λ/λ0)−1]−1
→ e 0
As (λ/λ0)→∞ 1 → (1+(λ/λ1 )−1−1) → (λ/λ0)

exp[(λ/λ0)−1]−1 0
p( λ)
∝ ( λ/λ0 ) -4 ( λ/λ0 ) -1
e
∝ ( λ/λ0 ) -3
8.044 L5B9
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Example Random number generator for programmers
p(x) p(y)
1 1
0 1 x 0 1 y
x and y are statistically independent
8.044 L6B1
z ≡ MAX(x, y) Find p(z)
A
y p(x,y)=0
1
p(x, y) = p(x) p(y) p(x,y)=1
η
Where is MAX(x, y) = η?
Where is MAX(x, y) < η?
0 η 1 x
8.044 L6B2
p(z)
2
B Pz (η) = η2
C pz (η) = 2η 0≤η≤1
0 1 z
1 1 3
<z> = 2η 2 dη = (2/3) [ 0η = 2/3
0
1 1 4
< z2 > = 2η 3 dη = (2/4) [ 0η = 1/2
Var(z) = 1
2 − 4 = 1,
9 18 S.D. = √1 = 0.24
18
8.044 L6B3
z
Example Desorbing atom
v
n
p(v, θ, φ) = p(v) p(θ) p(φ)
θ
p(v) = (1/2σ 4) v 3 exp[−v 2/2σ 2]
p(θ) = 2 sin θ cos θ
p(φ) = 1/2π
y
φ
Find p(vz ) x
leaving the surface
8.044 L6B4
A
vz = v cos θ v
v cos θ < γ
⇒ v < γ/ cos θ
γ
π/2 θ
8.044 L6B5
B
π/2 γ/ cos η
Pvz (γ) = pv (ζ) pθ (η) dζ dη
0 0
π/2
γ/ cos η
= pθ (η) pv (ζ) dζ dη
0 0
C
π/2
dPvz (γ) 1 p ( γ ) dη
pvz (γ) = = pθ (η) cos η v cos η
dγ 0
8.044 L6B6
pvz (γ) =
3
π/2 1 1 γ 1 γ2
0 (2 sin η cos η ) cos η 2σ 4 cos η exp[− 2σ 2 cos2 η ] dη
1 γ2
Let 2σ 2 cos2 η
≡ X
1 γ2
dX = − σ 2 cos3 η (− sin η)dη
η=0 ⇒ X= γ 2/2σ 2; η = π/2 ⇒ X= ∞
8.044 L6B7
γ ∞ −X γ ∞ −X
pvz (γ) = 2 2 e dX = − [ 2 2 e
σ γ /2σ 2 σ 2 γ /2σ
γ
= 2 exp[−γ 2/2σ 2] γ>0
σ
p(v z )
0 vz
8.044 L6B8
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Sums of Random Variables
n
�
Consider n RVs xi and let s ≡

xi .
i=1
If the RVs are statistically independent, then
�
<s> = < xi >
i
�
Var(s) =
Var(xi)
i
8.044 L7B1
• The individual p(xi) could be quite different
• Both continuous and discrete RVs could be present
• True for any n
• Even if one RV dominates the sum
8.044 L7B2
Results have a special meaning when
0)
1) The means are finite (=
∞)
2) The variances are finite (=

3) No subset dominates the sum
4) n is large
p(s) width
∝ 1
mean n
∝ n
∝n s
8.044 L7B3
Given p(x, y), find p(s ≡ x + y)
y
A
α x+y = α
y = α−x
α x
dx
∞ α−ζ
B Ps(α) = dζ dη px,y (ζ, η)
−∞ −∞
8.044 L7B4
∞
C ps(α) = dζ px,y (ζ, α − ζ)
−∞
This is a general result; x and y need not be S.I.
Application to the Jointly Gaussian RVs in Section 2

shows that p(s) is a Gaussian with zero mean and a
Variance = 2σ 2(1 + ρ).
8.044 L7B5
In the special case that x and y are S.I.
∞ ∞
ps(α) = dζ px(ζ) py (α−ζ) = dζ px(α−ζ ) py (ζ )
−∞ −∞
The mathematical operation is called “convolution”.
∞
p⊗q ≡ p(z)q(x − z)dz = f (x).
−∞
8.044 L7B6
Example
p(z)
Given:
∝ ( z / a ) n e-z/a
1
p(z) = (z/a)n exp(−z/a)
n!a
∝ zn
q(z) = (z/a)m exp(−z/a) z

m!a
0 < z and n, m = 0, 1, 2, · · ·
Find: p ⊗ q
8.044 L7B7
q(z)
q(-z)
0 z 0 z
p(z)
q(x-z)=q(-(z-x)) q(x-z)
0 x z 0 x z
finite product
8.044 L7B8
n
1 1 x z x − z m −z/a −(x−z)/a
p⊗q = e e dz
n!m! a2 0 a a
n+m+1 x
1 1 1 −x/a
= e z n (x − z)m dz
n!m! a a 0
n+m+1 1
1 1 x −x/a
= e ζ n (1 − ζ)m dζ
n!m! a a 0
n!m!
(n+m+1)!
8.044 L7B9
n+m+1
1 1 x
p⊗q = e−x/a
(n + m + 1)! a a
a function of the same class
8.044 L7B10
Example Atomic Hydrogen Maser
RF out
flask
ν0
H* beam ν = 1.4....... GHz
ν1 −ν0 about 10 KHz

ν1
p ( t wall | n stays) = ?
cavity
8.044 L7B11
n

twall = tn, Each stay is S.I.
i=1
p(t |1) = (1/τ ) e−t/τ
p(t |2) = p(t |1) ⊗ p(t |1) = (1/τ )(t/τ ) e−t/τ
p(t |3) = p(t |2) ⊗ p(t |1) = (1/2)(1/τ )(t/τ )2 e−t/τ
8.044 L7B12
n−1
1 1 t
p(t |n) = e−t/τ
(n − 1)! τ τ
0.15
τ p(t | 12)
0.12
0.10
0.1
0.08
0.06
0.05
0.04
0.02
0
5
1100 15
2200 25
330
0
t/τ
8.044 L7B13
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Facts about sums of RVs
• Exact expressions for < s > and Var(s) if S.I.
• p(s) = p(x) ⊗ p(y) if S.I.
• p(s) slightly more complicated if not S.I.
8.044 L8B1
• ⊗ usually changes functional form
• But not always
• Fourier techniques are very useful
8.044 L8B2
Very important special case: Central Limit Theorem
• RVs are S.I.

• All have identical densities p(xi)
• Var(x) is finite but < x > could be zero
• n is large
p(s) Central Limit Theorem:

p(s) is Gaussian
∝ n
∝n s
8.044 L8B3
If x is continuous
1 −(s−<s>) 2 /2σ 2
p(s) = √ e
2πσ 2
< s >= n < x >
σ 2 = n σx2
8.044 L8B4
If x is discrete in equal steps of ∆x
∆x −(s−<s>) 2 /2σ 2
p(s) = √
2
e δ(s

−i ∆x)
i 2πσ
comb
envelope
p(s)
∆x s
8.044 L8B5
Example K.E. of an ideal gas
1 2 /2σ 2
p(vx) = √ e−v x σ= kT /m
2πσ 2
K.E.x = 12 mvx2
p (K.E.x)
mean = 12 m vx2 = 12 kT
σ2
mean sq.= ( 12 m)2 vx4
3σ 4
= 34 (kT )2
1 (kT )2 K.E.x
variance = 2
8.044 L8B6
3 directions, N atoms
E = total K.E.
U ≡ E = 3N K.E.x = 32 N kT
Variance(E) = 3N Variance(K.E.x) = 32 N (kT )2
1 (E−(3/2)N kT )2
p(E) = exp[− 2{(3/2)N (kT )2} ]
2
2π{(3/2)N (kT ) }

3 N kT
s.d. 2 1
= =
mean 3 N kT 3N
2 2
8.044 L8B7
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Some terms that must be understood
Microscopic Variable
Macroscopic Variable
8.044 L9B1
Extensive (∝ N ) f (N ))
Intensive (=
V volume P pressure
A area S surface tension
L length F tension
P polarization E electric field
M magnetization H magnetic field
········· ·········
T temperature
U internal energy
8.044 L9B2
Adiabatic Walls
Diathermic Walls
Equilibrium
Steady State
Complete Specification:
Independent and Dependent Variables
8.044 L9B3
Equation of State
P V = N kT
V = V0(1 + αT − KT P )
M = cH/(T − T0) T > T0
In Equilibrium with Each Other
8.044 L9B4
OBSERVATIONAL FACTS
" 0 th Law"
equilibrium equilibrium
if A C and B C
equilibrium
then A B
8.044 L9B5
" Law 0.5 ? " Many macroscopic states of B can be
in equilibrium with a given state of A
YA
YB
A B
1 1 YB = f (XB)
also
XA , YA
XA XB
8.044 L9B6
THEOREM A "predictor" of equilibrium h(X, Y, ...) exists
h(X, Y)
• only in equilibrium
• state variable
• many states, same h
• different systems,
different functional forms
• value the same if
systems in equilibrium Y
locus of
X constant h
8.044 L9B7
XA , YA XC , YC
XA , YA , XC , YC all free [ P A , V A , PC , V C ]
equilibrium
XA , YA XC , YC
keep same adjust
XC = f 1 (YC , XA , YA ) [ P C = PA V A / V C ]
F1 (XC , YC , XA , YA ) = 0 [ PC V C - PA V A = 0 ]
8.044 L9B8
equilibrium
XB , YB XC , YC
adjust keep same
XB = g (YB , XC , YC ) [ P B = PC V C / V B ]
F2 (XC , YC , XB , YB ) = 0 [ PC V C - PB V B = 0 ]
solve for X C
XC = f 2 (YC , XB , YB ) [ P C = PB V B / V C ]
same value as before
8.044 L9B9
f 1 (YC , XA , YA ) = XC = f 2 (YC , XB , YB )
1
[ P A V A / V C = PB V B / V C ]
equilibrium
XA , YA XB , YB
due to 0 th law
F3 (XA , YA , XB , YB ) = 0 2
1 + 2 F3 factors
YC drops out
8.044 L9B10
For this equilibrium condition
h( XA , YA ) = constant = h( XB , YB )
[ P A V A = PB V B ]
8.044 L9B11
Isotherm at T 1
YA A YB B
2
1 T2
T2
1
2
Isotherm at T 1
XA XB
8.044 L9B12
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
THEOREM A "predictor" of equilibrium h(X, Y, ...) exists
h(X, Y)
• only in equilibrium
• state variable
• many states, same h
• different systems,
different functional forms
• value the same if
systems in equilibrium Y
locus of
X constant h
8.044 L9B7
Empirical Temperature: t
low density gas
Y P PV/N = constant'
2
PV/N = constant
2
1
X V
8.044 L10B1
we could possible alternative
• Define t = c g PV/N t' = c g ' (PV/N) α
• Use to find isotherms
in other systems
• Then in a simple paramagnet

t = c m (M/H) -1 t' = c m ' (M/H) - α
Many possible choices for t
8.044 L10B2a
PV = Nkt → t = PV/Nk
P 1
0.8
t
P 0.6
0.4
0.2
V 0 0.2 0.4 0.6 0.8 1
V
PV = constant
8.044 L10B2b
t' = (PV/Nk) 2
P1
0.8
t'
P 0.6
0.4
0.2
V 0 0.2 0.4 0.6 0.8 1
V
PV = constant
8.044 L10B2c
t'' = PV/Nk
P 1
t'' 0.8
P
0.6
0.4
0.2
V 0 0.2 0.4 0.6 0.8 1
V
PV = constant
8.044 L10B2d
M = cH/t → t = cH/M
H 1
t H 0.8
0.6
0.4
0.2
M 0.2 0.4 0.6 0.8
M
1
H/M = constant
8.044 L10B2e
V = V 0 ( 1 + αt - κP) → t = (V-V 0 ) / αV 0 + κP/α
P P 1
0.8
t
0.6
0.4
0.2
(V-V 0 ) 0 0.2 0.4 0.6 0.8
(V-V 0 )
1
(V-V 0 ) / V 0 + κP = constant
8.044 L10B2f
Work
dW = differential of work done o n the system
= - (work done b y the system)
dx
Hydrostatic system
dW = -PdV F
dW = Fdx = (PA)(-dV/A) = -PdV
8.044 L10B3
F F
Wire
dW = FdL
P pushes, F pulls
dW = Fdx = (F)(dL) = FdL
Surface
dW = SdA wire
2 SL F constraint
frame
film
(2 surfaces)
dW = Fdx = (SL)(dA/L) = SdA
8.044 L10B4
Chemical Cell (battery)
dW = EEMF dZCHARGE
Electric charges Field in absence of matter

dW = EdP as set up by external
sources. Does not include
energy stored in the field
itself in the absence of
Magnetic systems the matter.
dW = HdM
8.044 L10B5
• All differentials are extensive
• Only -PdV has a negative sign
• Good only for quasistatic processes

b
• ∆W =
dW depends on the path
a
W is not a state function
8.044 L10B6
low density gas

A isotherm PV= NkT
Y P
b
P2
B a c b
P1 a
X V2 V1 V
dW = YdX
depends on Y(X)
8.044 L10B7
(a) W1→2 =
−P1(V2 − V1) = P1(V1 − V2)
(b) W1→2 =
−P2(V2 − V1) = P2(V1 − V2)
2 2 2
(c) W1→2 =
− P (V ) dV = − N kT dV = −N kT dV
1 1 V 1 V
= −N kT ln V2 = N kT ln V1 = P V ln V1
V1 V2 1 1 V2
8.044 L10B8
MATH
I) 3 variables, only 2 are independent
F (x, y, z) = 0
⇒ x = x(y, z), y = y(x, z), z = z(x, y)
    � �
� �
∂x

1
∂x
∂y
∂z
⇒ 
=
� �
,


= −1
∂y
z ∂y
∂y
z ∂z
x ∂x y
∂x
z
8.044 L10B9
� �
� � ∂x
∂z
and if W (x, y, z), then ∂x
∂y = � �w
w ∂y
∂z w
II) State function of 2 independent variables
S = S(x, y)
� �  
∂S 
∂S 
dS = dx + dy
∂x y ∂y x
� ��
A(x,y) B(x,y)
An exact differential
8.044 L10B10
  � �
∂A

∂ 2S ∂ 2S ∂B
= = =
∂y
x ∂y∂x ∂x∂y ∂x y
⇒ necessary condition, but it is also sufficient
  � �
∂A

∂B
Exact differential if and only if =

∂y
x ∂x y
�
2
Then
dS = S(x2, y2) − S(x1, y1) is independent of
1
the path.
8.044 L10B11
III) Integrating an exact differential
dS = A(x, y) dx + B(x, y) dy
1. Integrate a coefficient with respect to one variable

∂S
= A(x, y)
∂x y

S(x, y) = A(x, y) dx +f (y)

y fixed
8.044 L10B12
2. Differentiate result with respect to other variable
 
�
�
�
∂S


=
∂
d f (y)
A(x, y) dx
+
= B(x, y)
∂y
x ∂y
dy
3. Integrate again to find f (y)
 
d f (y)
 ∂
� 
= 
B(x, y) − A(x, y) dx 
dy ∂y
f (y) = {· · ·} dy
done
8.044 L10B13
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Thermodynamics focuses on state functions: P, V, M, S, . . .
Nature often gives us response functions (derivatives):

1 ∂V 1 ∂V 1 ∂V
α≡ κT ≡ − κS ≡ −
V ∂T P V ∂P T V ∂P adiabatic

∂M
χT ≡
∂H T
8.044 L11B1
Example Non-ideal gas
Given
• Gas → ideal gas for large T & V

∂P Nk
• =
∂T V V − Nb

∂P N kT 2aN 2
• =− +
∂V T (V − N b) 2 V3
Find P
8.044 L11B2

∂P ∂P
dP = dV + dT
∂V T ∂T V
∂P Nk

P = dT + f (V ) = dT + f (V )
∂T V V − Nb
N kT
= + f (V )
(V − N b)
8.044 L11B3
2
∂P N kT (V ) = − N kT 2aN
= − + f

+
3
∂V T (V − N b) 2 (V − N b) 2
V
2aN 2 aN 2
f (V ) = 3
dV = − 2
+ c
V V
N kT aN 2
P = − 2
+ c
(V − N b) V
but c = 0 since P → N kT /V
as V → ∞
8.044 L11B4
Internal Energy U
Observational fact
initial final
isolated
∆W (adiabatic)
Final state is independent of how ∆W is applied.

Final state is independent of which adiabatic path is
followed.
8.044 L11B5
⇒
a state function U such that
∆U = ∆Wadiabatic
U = U (independent variables)
= U (T, V ) or U (T, P ) or U (P, V ) for a simple fluid
8.044 L11B6
Heat
d
If the path is not adiabatic, dU = /W
/Q ≡ dU − d
d /W
/Q is the heat added to the system.
It has all the properties expected of heat.
8.044 L11B7
First Law of Thermodynamics
dU = d
/Q + d
/W
• U is a state function
• Heat is a flow of energy
• Energy is conserved
8.044 L11B8
Ordering of temperatures
dQ
T1 T2
When d
/W = 0, heat flows from high T to low T.
8.044 L11B9
Example Hydrostatic System: gas, liquid or simple
solid
Variables (with N fixed): P, V, T , U .

Only 2 are independent.
   
d
/Q  d
/Q 
CV ≡  CP ≡ 
dT V dT P
Examine these heat capacities.
8.044 L11B10
dU = d
/Q + d /Q − P dV
/W = d
/Q = dU + P dV
d
d
We want . We have dV .
dT

∂U ∂U
dU = dT + dV
∂T V ∂V T
8.044 L11B11

∂U ∂U
/Q =
d dT + + P dV
∂T V ∂V T

d
/Q ∂U ∂U dV
⇒ = + +P
dT ∂T V ∂V T dT
 
d
/Q  ∂U
CV ≡ =
dT V ∂T V
8.044 L11B12
 
d
/Q  ∂U ∂U ∂V
CP ≡  = + +P
dT P ∂T
V
∂V T ∂T
P
CV αV

∂U
CP − CV = + P αV
∂V T
The 2nd law will allow us to simplify this further.

∂U
Note that CP = .
∂T P
8.044 L11B13
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Paths Experimental conditions, not just math
fills floating bath insulation

container piston
~~~~~~~
∆V=0 ∆P=0 ∆T=0 ∆Q=0
8.044L12B1
∆Q = 0 could come from time considerations
Example Sound Wave
ρ(x)
too fast for heat to flow out of compressed regions
1
v =
ρκS
8.044L12B2
Example Hydrostatic system: an ideal gas, PV=NkT

∂U
New information = 0,
∂V T
3 possible sources
• Experiment
free
expansion
bath initially at T i observe T f = T i
8.044L12B3
No work done so ∆W = 0
Tf = Ti ⇒ ∆Q = 0
together ⇒ ∆U= 0 → (∂U/∂V)T = 0

here quasi-static changes
• Physics: no interactions, single particle energies

only ⇒ (∂U/∂V )T = 0
• Thermo: 2nd law + (P V = N kT ) ⇒ (∂U/∂V )T = 0
8.044 L12B4
Consequences

∂U ∂U
dU = dT + dV
∂T
V ∂V
T
CV 0
T
U = CV (T �) dT � + constant

0
set=0
In a monatomic gas one observes CV = 32 N k.

Then the above result gives U = CV T = 3
2 N kT .
8.044 L12B5

∂U ∂V
CP − CV = ( +P )
∂V
T ∂T
P
0 ∂ (N kT /P ) =N k/P
∂T P
= Nk for any ideal gas
Applying this to the monatomic gas one finds

3 5
CP = N k + N k = N k
2 2
5
γ ≡ CP /CV =
3
8.044 L12B6
Adiabatic Changes /Q = 0
d
Find the equation for the path.

Consider a hydrostatic example.

∂U ∂U
/Q =
d dT + + P dV = 0
∂T
V ∂V
T
CV (CP −CV )/αV
 
∂T CP − CV )  1 (γ − 1)
=−  =
−
∂V ∆Q=0 CV αV αV
This constraint defines the path.
8.044 L12B7
Apply this relation to an ideal gas.

1 ∂V 1 ∂ N kT 1 Nk 1V 1
α≡ = = = =
V ∂T P V ∂T P P V P VT T
Path
dT T
= −(γ − 1)
dV V
 
dT dV T V
= −(γ − 1) → ln  = −(γ − 1) ln
T V T0 V0
   −(γ−1)

T 
V 
=
T0 V0
8.044 L12B8
Adiabatic Isothermal
γ P
P V = c
adiabat
TV 1 = c
γ
PV =c P ∝ V 1
γ = 5/3 (monatomic) isotherm

5/3
dP P
P ∝V =
dV V
dP 5P V
=
dV 3V
8.044L12B9
Expansion of an ideal gas
insulation
F
rupture diaphragm slowly move piston

adiabatic ∆Q = 0 adiabatic ∆Q = 0
not quasistatic quasistatic
∆W = 0 ∆W is negative
→ ∆U = 0 → ∆U = is negative
T T adiabat
constant U
V V
8.044L12B10
Starting with a few known facts,
1st law, d
/W , and state function math,

one can find
relations between some thermodynamic quantities,
a general expression for dU ,
and the adiabatic constraint.
Adding models for the equation of state and the heat
capacity allows one to find

the internal energy U
and the adiabatic path.
8.044 L12B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
1. The System
E < energy < E+ ∆

∆ << E
Fixed:
E
V
N A complete set of
M or H independent thermodynamic
variables is fixed.
P or E
Many micro-states
satisfy the conditions.
8.044 L13B1
2. Probability Density
All accessible microscopic states are equally probable.
Classical
p({p, q}) = 1/Ω E < H({p, q}) ≤ E + ∆

= 0 elsewhere

Ω≡ {dp, dq} = Ω(E, V, N )
accessible
8.044 L13B2
Quantum
p(k) = 1/Ω E < k|H|k ≤ E + ∆

= 0 elsewhere

Ω ≡ (1) = Ω(E, V, N )
k, accessible
8.044 L13B3
Let X be a state of the system specified by a subset
{ p”,q” } of { p,q}

p(X) = p({p, q}) {dp, dq}
except {p ,q }
1

= ,q }
{dp, dq}
Ω except {p
Ω(consistent with X)
=
Ω
volume consistent with X
=
total volume of accessible phase space
8.044 L13B4
3. Quantities Related to Ω

Φ(E, V, N ) ≡ {dp, dq}
H({p,q})<E
= cumulative volume in phase space
∂Φ(E, V, N )
ω(E, V, N ) ≡
∂E
= density of states as a function of energy
⇒ Ω(E, V, N ) = ω(E, V, N )∆
8.044 L13B5
Example Ideal Monatomic Gas
qi = x, y, z in a box V = LxLy Lz
pi = mx,
˙ my,
˙ mż − ∞ < pi < ∞
3N
p2
i
N atoms H({p, q}) =
i=1 2m
8.044 L13B6

Ω = {dp, dq} = {dq} {dp}
N N N
Lx Ly Lz
= dx dy dz {dp}
0 0 0

= VN {dp}
E<H<E+∆

Φ(E, V, N ) = V N {dp}
H<E
8.044 L13B7
3N
p2 3N

E =
i ⇒ 2mE = p
2
i=1
2m i=1
This describes a 3N dimensional spherical surface in

√
the p part of phase space with a radius R = 2mE.
8.044 L13B8
Math:
• Volume of an α dimensional sphere of radius R is
π α/2 α
R
(α/2)!
• Sterling’s approximation for large M
ln(M !) ≈ M ln M − M
M
M
→ M! ≈
e
8.044 L13B9
π 3N/2
Φ(E, N, V ) = V N (2mE)3N/2
(3N/2)!
 
 3N/2
 4πemE 
≈ V N

 3N 


3N 1
ω(E, N, V ) = { }
2 E

3N ∆
Ω(E, N, V ) = { }
2 E
8.044 L13B10
Ω V N −1Ly Lz 1
p(xi) = = N
= 0 ≤ x < Lx
Ω V Lx
Ω V N −2Ly Lz LxLz 1
p(xi, yj ) = = N
= = p(xi)p(yj ) ⇒ S.I.
Ω V LxLy
Ω
p(pxi ) = p

({p,

q}){
dp , dq} =
pxi
p=
Ω
1/Ω
Note that Ω differs on each of the three lines, be-

ing a generic symbol for the reduced phase volume
consistent with some constraint.
8.044 L13B11
≡ p2
x /2m E − distributed over other variables
 (3N −1)/2
3N − 1 ∆ 4πem(E − ) 
Ω = V N 
2 E− 3N − 1
−1/2
Ω 3N − 1 E 4πem
=
Ω 3N
−

E 3
≈1 ≈1
   
3N 1
− 2 +2 3N − 1
 (N −1 )   (E − ) 2 2 
× 

3 





− 3N 3N
N 2 E 2

A B
8.044 L13B12
− 3N  − 3N

1 1 2 1 1/2 2

A= N− 1− = N− 1 + 
3 3N 3 −3N/2

ζ
but limζ→∞ 1 + xζ = ex
√
so A ≈ N e1/2
8.044 L13B13
  3N
3N 1 /
1 2 1  < > 2
B=√ 1− =√ 1 − 2 
E− E E− 3N/2
where we have used < >≡ E/3N and E = 3N < >.
so B ≈ √ 1 e−/2<>
3N <>
8.044 L13B14
 √ 
�
√ �
3
−1/2
1
p(px) =
√

e 
N e 1/2 √ e−/2<>
4πm 3N < >
= √ e−/2<>
4πm < >
Now use = p2
x
/2m and < >=< p 2
> /2m.
x
1 −p2 /2<p2 >

p(px) = � e x x
2
>
2π < px
8.044 L13B15
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Ω = Ω(E, V, N . . .) ≡ Volume of the accessible region
of phase space.
8.044 L14B1
4. Entropy

S(E, V, N ) ≡ k ln Ω(E, V, N ) 








≈ k ln Φ(E, V, N ) 
Differ only by ln N







≈ k ln ω(E, V, N ) 
• It is a state function.
• It is extensive.
• It is a logarithmic measure of the microscopic de-
generacy associated with a macroscopic (that is, ther-
modynamic) state of the system.
• k is Boltzmann’s constant, units of energy per 0K.
8.044 L14B2
5. Statistical Mechanical Definition of Temperature
1 • Total is microcanonical
dQ
• d
/W1→
2 = 0
2
• interaction between 1 &
Find the most probable E1 2 is so small that Ω can
≡ E1∗ be separated
8.044 L14B3
Ω Ω1(E1) Ω2(E − E1)
p(E1) = =
Ω Ω(E)
ln p(E1) = ln Ω1(E1) + ln Ω2(E − E1) − ln Ω(E)
1
= ( S1(E1) + S2(E − E1) − S(E) )
k
     
∂ 1   ∂S1  ∂S2  
ln p(E1) =  −  =0
∂E1 k ∂E1 d/W =0 ∂E2 d/W =0
1 2
8.044 L14B4
The condition for determining E1∗
is
   

∂S1  
∂S2

=
∂E1 d/W =0 ∂E2 d/W =0
1 2
f of 1 f of
But this also specifies the equilibrium condition. Thus
∂S 1
= f (T ) ≡ (in equilibrium)
∂E d/W =0 T
8.044 L14B5
6. Two Fundamental Inequalities E1∗
?
What if E1 =
1 ∗ as equilibrium is established.
1→
p(E1) ≤ p(E1∗ )
Ω1(E1)Ω2(E − E1) ≤ Ω1(E1∗ )Ω2(E − E1∗ )
Ω1(E1∗ ) Ω2(E − E1∗ )

1 ≤
Ω1(E1) Ω2(E − E1)
8.044 L14B6
0≤S (E ∗ ) − S (E ) + S (E − E ∗
) − S (E − E )
1 1 1 1 2 1 2 1
∆S1 ∆S2
⇒• ∆S = ∆S1 + ∆S2 increases
∆S ≥ 0
The total entropy of an isolated system always in-

creases or, at equilibrium, remains constant.
8.044 L14B7
Now assume 1 ∗ ⇒ T2 ≡ Tbath does not change.
2
dE2 d
/Q2 d
/Q1
dS2 = = =−
T2 T2 Tbath
d
/Q1
dS = dS1 + dS2 = dS1 − ≥0

Tbath
d
/Q1
⇒• dS1 ≥
Tbath
In particular, for systems in equilibrium with a bath

/Q/T .
dS = d
8.044 L14B8
Example Ideal Monatomic Gas
 
3N/2  3/2N
4πemE  4πemE 
Φ ≈ VN = V 
3N  3N 
 
 3/2
 4πemE 
→ S(E, N, V ) = kN ln V 
 3N 
3N/2
3N ∆ 4πemE
Ω ≈ VN
2 E 3N 

 4πemE 3/2
 2 1 E
→ S(E, N, V ) = kN ln V 
− ln( )
 3N  3∆N
8.044 L14B9
The Energy Relation

1 ∂S Nk 3 1 (3/2)N k
≡ = {} =
T ∂E N,V {} 2 E E
⇒ E = (3/2)N kT

Here U = E so CV = ∂U
∂T V
= (3/2)N k.
8.044 L14B10
The Adiabatic Condition
∆Q = 0 ⇒ ∆S = 0 for a quasistatic process.

 
 3/2
 4πemE 
S(E, N, V ) = kN ln V 
 3N 
Use the energy relation to eliminate E.

  3/2

 


4πem((3/2)N
/kT 
S(E, N, V ) = kN ln V 
 3N/ 
∆S |∆N =0 ⇒ V T 3/2 = constant
8.044 L14B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Microcanonical: E fixed + equal a priori probabilities
⇒ microscopic probability densities [S.M.]
Together with the definition of entropy
⇒ temperature scale and 2N D law inequalities [Ther-

modynamics]
8.044 L15B1

∆S ≥ 0 (3 · 4 · 7) 



2 N D Law


d
/Q1 
dS1 ≥ T (3 · 6 · 5) 
Quasi-static means arbitrarily close to equilibrium.
• Necessary for work differentials to apply

• Required for = in above 2N D law relations
8.044 L15B2
7. Entropy as a Thermodynamic Variable
� �
∂S 1
≡ gives us T
∂E d/W =0 T
Other derivatives give other thermodynamic variables.
 


 −P dV 

 �
/W = 
d SdA 
+ HdM + E dP + · · · ≡ Xi dxi

 

F dL i
8.044 L15B3
We chose to use the extensive external variables (a
complete set) as the constraints on Ω. Thus
S ≡ k ln Ω = S( E, V, M, · · · )
Now solve for E.
S(E, V, M, · · ·) ↔ E(S, V, M, · · ·)
We know
dE|d/W =0 = d
/Q from the 1ST law
dE |d/W =0 ≤ T dS utilizing the 2N D law
8.044 L15B4
Now include the work.
/Q + d
dE = d /W
dE ≤ T dS + d
/W
 


 −P dV 


dE ≤ T dS +  SdA 
+ HdM + EdP +
· · ·

 

F dL
The last line expresses the combined

1ST and 2N D laws of thermodynamics.
8.044 L15B5
Solve for dS.
1 P H E
dS = dE + dV − dM − dP +
· · ·
T T T T
Examine the partial derivatives of S.

∂S 1 ∂S H
= =−
∂E V,M,P T ∂M E,V,P T
 

∂S P ∂S  Xj
=
 =
−
∂V E,M,P T ∂xj E,x =x

T
i j
8.044 L15B6
INTERPRETATION
S(E,V)

∂S ∂S
dS = dE + dV
V ∂E V ∂V E
1 P
= dE + dV
T T
8.044 L15B7
UTILITY
Internal Energy
 
∂S(E, V, N )  1
 = → T (E, V, N ) ↔ E(T, V, N )
∂E V T
Equation of State
 
∂S(E, V, N )  P
 = → P (E, T, V, N ) → P (T, V, N )
∂V E T
8.044 L15B8
Example Ideal Gas
 
 3/2
 4 E 
S(E, N, V ) = k ln Φ = kN ln V πem 
 3 N 

∂S kN {} kN P
= = =
∂V E,N {} V V T
P V = N kT
8.044 L15B9
COMBINATORIAL FACTS
# different orderings (permutations) of K distinguish

-
able objects = K!
# of ways of choosing L from a set of K:
K!
if order matters
(K − L)!
K!
if order does not matter
L!(K − L)!
8.044 L15B10
EXAMPLE Dinner Table, 5 Chairs (places)
Seating, 5 people 5 · 4 · 3 · 2 · 1 = 5! = 120
Seating, 3 people 5 · 4 · 3 = 5!
2! = 60
Place settings, 3 people 5 · 4 · 3/6 = 5! 1 = 10
2! 3!
8.044 L15B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
EXAMPLE 2 Level System
Ensemble of N "independent" systems
ENERGY
|1 > ε N = N0 + N 1
E = ε N1
|0 > 0
8.044 L16B1
SURFACE MOLECULES IONS IN A CRYSTAL LOWEST LYING STATES
ENERGY
0 ε 0 ε
ε
0
E N1
NO WORK POSSIBLE (JUST HEAT FLOW)
8.044 L16B2
N!
1 when N1 = 0 or N
Ω(E) = N1!(N −N1)!

Maximum when N1 = N/2
T= (or - )
S(E)
S(E) = k ln Ω(E)
E= ε N/2 E= εN
T= 0
E
T> 0 T< 0
8.044 L16B3
ln N ! ≈ N ln N − N
S(E) = k[N ln N − N1 ln N1 − (N − N1) ln(N − N1)

− N + N1 + N − N 1 ]

1 ∂S ∂S ∂N1 k
= = = [−1 − ln N1 + 1 + ln(N − N1)]
T ∂E N ∂N1 ∂E

1/
   
k  N − N1  k N
= ln = ln − 1
N1 N1
8.044 L16B4
N
N
− 1 = e/kT → N1 = /kT
N1 e +1
N
E = N1 = /kT
e +1
1.0 N1 /N or E/ εN
~ e−ε/ k T
0.5
4
kT/ε
1 2 3
8.044 L16B5
2
∂E e/kT
C≡ = Nk
∂T kT (e/kT + 1)2

2 −/kT Nk 2
→ Nk e low T , → high T
kT 4 kT
0.5
C/Nk
0.4
0.3
0.2
0.1
1 2 3
kT/ε 4
8.044 L16B6
Ω
p(n) =? n = 0, 1 p(n) =
Ω
In Ω N →N −1 and N1 → N1 − n
(N −1)!
(N1−n)!(N −1−N1+n)!
p(n) = N!
N1!(N −N1)!
8.044 L16B7
(N − 1)! N1 ! (N − N1)!
p(n) =
N! − n)! (N − N1 − 1 + n)!
(N1
1/N 1 n=0 N − N1 n = 0
N1 n = 1 1 n=1

N −N1 N1 
p(0) = N =1− N




p(0) + p(1) = 1


p(1) = N 1 = [e/kT + 1]−1 
N
8.044 L16B8
1
p(n)
p(0)
0.5
p(1)
0 1 n 1 2 3
kT/ε 4
N
E = (0)N p(0) + ()N p(1) = /kT
e +1
But we knew E, so we could have worked backwards

to find p(1).
8.044 L16B9
MICROCANONICAL ENSEMBLE
MODEL THE SYSTEM
FIND Ω(E,N,V ....)
THERMODYNAMIC RESULTS MICROSCOPIC INFORMATION
FIND S(E,N,V ....) P(~~~) = Ω' /Ω
∂S = 1
∂E N,V T
∂S = P
∂V E,N T
etc.
8.044 L16B10
The microcanonical ensemble is the starting point for

Statistical Mechanics.
• We will no longer use it to solve problems.
• We will develop our understanding of the 2N D law.
• We will derive the canonical ensemble, the real

workhorse of S.M.
8.044 L16B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
ENTROPY AND THE 2nd LAW
1 2
T bath
dS ≥ 0 dS1 ≥ dQ1 / Tbath
8.044 L17B1
S as a State Function
Note: adiabatic (≡ d
/Q = 0) ⇒ constant S if the
change is quasistatic. This is the origin of the sub-

script S on the adiabatic compressibility.

1 ∂V 1 ∂V
κT ≡ − κS ≡ −
V ∂P T V ∂P S
8.044 L17B2
Example A Hydrostatic System

∂S ∂S
dS = dT + dV by expansion
∂T V ∂V T
1 P
= dU + dV from dU = T dS − P dV
T T

1 ∂U 1 ∂U
= dT + + P dV
T ∂T V T ∂V T
by expansion of U
But the cross derivatives of S must be equal.
8.044 L17B3

∂ 1 ∂U 1 ∂ 2U
=
∂V T ∂T V T T ∂V ∂T

∂ 1 ∂U 1 ∂U 1 ∂ 2U1 ∂P
+P =− 2 +P + +
∂T T ∂V T V T ∂V T T ∂V ∂T T ∂T V
Equating these two expressions gives

∂U ∂P
+ P = T
∂V T ∂T V

∂U ∂P
=T −P
∂V T ∂T V
New Information! Does not contain S!
8.044 L17B4
CONSEQUENCES a) γ

∂U ∂U
/Q = dU + P dV =
d dT + + P dV
∂T
V ∂V
T

CV T ∂P
∂T V

/Q
d ∂P ∂V
≡ CP = CV + T
dT P ∂T V ∂T
P
αV

1 ∂V 1 ∂V
Use α ≡ and κT ≡ − .
V ∂T P V ∂P T
8.044 L17B5
� �
�
∂P
�
−1 (−1) ∂V
∂T α
=� � � � = � � P =
∂T V ∂T ∂V ∂V κT
∂V P ∂P T ∂P T
 
α T α 2V T α2V
CP − CV = T   αV = → γ−1=
κT κT κT CV
For an ideal gas P V = N kT ⇒ α = 1/T and κT = 1/P .
Thus
V /T
CP − CV = = N k
1/P
This holds for polyatomic as well as monatomic gases.
8.044 L17B6
CONSEQUENCES b) Ideal Gas: CV

N kT ∂U Nk
P = =T −P =P −P =0
V ∂V T V

∂U ∂U
dU = dT + dV = CV dT
∂T
V ∂V
T
CV 0
T
U = CV (T , V ) dT + constant
0
(∂U/∂V )T = 0 for all T ⇒ CV is not f (V ); CV = CV (T ).
8.044 L17B7
CONSEQUENCES c) Ideal Gas: S

∂S ∂S
dS = dT + dV
∂T V ∂V T

/Q
d ∂S
/Q = T dS
d ≡ CV = T
dT V ∂T V
dU P
dU = T dS − P dV ⇒ dS = + dV
T T

∂S 1 ∂U P
= + .
∂V T T ∂V T T

0 N k/V
8.044 L17B8
CV (T ) Nk
dS = dT + dV
T V
T CV (T ) V
S(T, V ) =
dT + N k ln( ) + S(T0, V0)
T0 T V0
For a monatomic gas CV = (3/2)N k.
T V
S(T, V ) − S(T0, V0) = (3/2)N k ln( ) + N k ln( )
T0 V0
  3/2
V  T  
= N k ln  
V0 T0
8.044 L17B9

V T 3/2 









isentropic (adiabatic) ⇒ V 2/3T 
are constant








V 5/3P 
8.044 L17B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Maxwell Relations

∂E ∂E
dE(S, V ) = dS + dV expansion
∂S V ∂V S
= T dS − P dV 1st and 2nd laws

∂T ∂P
⇒ =−
∂V S ∂S V
8.044 L18B1

∂T ∂F
dE(S, L) = T dS + F dL ⇒ =
∂L S ∂S L

∂T ∂H
dE(S, M ) = T dS + HdM ⇒ =
∂M S ∂S M
Observe:
d(T S) = T dS + SdT
d(P V ) = P dV + V dP
8.044 L18B2
Helmholtz Free Energy F ≡ E − TS

∂S ∂P
dF = −SdT − P dV ⇒ =
∂V T ∂T V
Enthalpy H ≡ E + PV

∂T ∂V
dH = T dS + V dP ⇒ =
∂P S ∂S P
8.044 L18B3
Gibbs Free Energy G ≡ E + PV − TS

∂S ∂V
dG = −SdT + V dP ⇒ − =
∂P T ∂T P
E, F , H and G are called ”thermodynamic potentials”.
8.044 L18B4
The Magic Square Mnemonic
E -V F
dE = T dS + Xdx
S -T
∂S ∂V
(−1) = (−1)(−1)
∂T
H P G ∂P T P
(+1)
X
8.044 L18B5
Homework problem 6-2, A Strange Chain
l
F
L = N l tanh(lF /kT )
For small extensions

1
α=−
T
8.044 L18B5b
Example Elastic Rod
Given F = (a

+ bT ) (L − L0) and CL = αT 3
+ for stability
Find E(T, L) and S(T, L).
 
 
∂S  ∂S 
 
dE = T dS +F dL = T
dT + T +F  dL
∂T  ∂L 
d
/Q L  T 
CL −bT (L−L0)
8.044 L18B6
E L F

S -T ∂S ∂F
(−1) = = b(L−L0)
∂L T ∂T L
H -F G
dE = αT 3dT + a(L − L0)dL
α 4
E = T + f (L)
4
a
f (L) = a(L − L0) f (L) = (L − L0)2 + c1
2
8.044 L18B7
α 4 a
E(T, L) = T + (L − L0)2 + c1
4 2

∂S ∂S
dS = dT + dL
∂T
L ∂L
T
CV /T =αT 2 −b(L−L0)
α 3
S = T + g(L)
3
b
g (L) = −b(L − L0) g(L) = − (L − L0)2 + c2
2
8.044 L18B8
α 3 b
S(T, L) = T − (L − L0)2 + c2
3 2
A rubber band has a negative thermal expansion coef.
dF = (a + bT )dL + b(L − L0)dT set = 0
+ whenextended

∂L b (L − L0)
⇒ =− <0 for rubber
∂T F (a

+ bT )
+ for stability
⇒b>0
8.044 L18B9
α 3 b
S(T, L) = T − (L − L0)2 + c2
3 2
Rapid expansion of rubber band ⇒ ∆S ∼ 0
Increase in L ⇒ increase in T .
8.044 L18B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Heat Engine
• Takes a substance around a closed cycle
• Heat is put into the substance and taken out
• Work is taken out
• Efficiency, η ≡ (work out) / (heat in)
8.044 L19B1
Closed cycle ⇒ ∆U = ∆Q + ∆W = 0 ⇒ ∆Q = −∆W

P ∆Q ≡ d
/Q
HEAT IN
(HOT RESERVOIR)
1
2 1
= /Q +
d /Q
d
2 1 2

HEAT OUT ≡|QH | ≡−|QC |
(COLD RESERVOIR)
8.044 L19B2
Most General Case
Wout = −∆W = ∆Q = |QH | − |QC |
Wout |QH | − |QC | |QC |

η≡ = =1−
|QH | |QH | |QH |
Very Special Case Example: Carnot Cycle
• Any substance
• Isothermal and adiabatic changes
8.044 L19B3
P
1
T
1 1'
1'
TH
2'
TC
2' 2 2
∆Q=0 ∆Q=0
S V
Use the second law: d

/Q < T dS
8.044 L19B4a
P
T TH
1
1 1'
∆Q=0
2' 2
1'
∆Q=0
2'
2
TC
S V
DRAWN TO SCALE FOR AN IDEAL GAS: PV=NkT

T H = 1.5 T C S HIGH - SLOW = (3/2) Nk ln2
8.044 L19B4b
1
|QH | ≤ TH dS
2 2 1
−|QC | ≤ TC dS, use dS = − dS

2 2 1
1 1 |QC | TC
≤ −TC dS ⇒ |QC | ≥ TC dS and ≥
1 1 |QH | TH
|QC | TC
η =1− ≤1−
|QH | TH
8.044 L19B5
Arbitrary Engine Cycle
/Q ≤ T dS for each element along the path.
2 2 2
/Q ≤
d T dS ≤ Tmax dS
1 1 1
|QH | positive
8.044 L19B6
1 1
/Q ≤
d T dS, both sides are negative
2 2
1 1
|QC | ≥ | T dS| ≥ Tmin | dS|
2 2
2
≥ Tmin | dS| since dS = 0
1
|QC |
Tmin
≥
|QH | Tmax
|QC | Tmin
η =1− ≤1−
|QH | Tmax
8.044 L19B7
Carnot cycle in a pure thermodynamic approach
|Q |
• Used to define temp. η = 1 − |Q C | ≡ 1 − TTC
H H
• Used to define the entropy
d
/Q d
/Q
≤0⇒ is an exact differential
T T
8.044 L19B8
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Refrigerator Run cycle backwards, extract heat at cold
end, dump it at hot end
HEAT EXTRACTED (COLD END) |QC | |QC |

= =
WORK DONE ON SUBSTANCE ∆W |QH | − |QC |
For the special case of a quasi-static Carnot cycle

TC
=
TH − TC
8.044 L20B1
• As with engine, can show Carnot cycle is optimum.
• Practical: increasingly difficult to approach T = 0.
• Philosophical: T = 0 is point at which no more

heat can be extracted.
8.044 L20B2
Heat Pump Run cycle backwards, but use the heat
dumped at hot end.
HEAT DUMPED (HOT END) |QH | |QH |

= =
WORK DONE ON SUBSTANCE ∆W |QH | − |QC |
For the special case of a quasi-static Carnot cycle

TH
=
TH − TC
8.044 L20B3
55o F subsurface temp. at 40o latitude
→ TC = 286K
70o F room temperature
→ TH = 294K
|QH | 294
≤ ∼ 37
∆W 8
8.044 L20B4
3rd law lim S = S0
T →
0
At T = 0 the entropy of a substance approaches a

constant value, independent of the other thermody-
namic variables.
• Originally a hypothesis
• Now seen as a result of quantum mechanics
Ground state degeneracy g (usually 1)

⇒ S → k ln g (usually 0)
8.044 L20B5

∂S
Consequences =0
∂x T =0
Example: A hydrostatic system

1 ∂V 1 ∂S
α≡ =− →0 as T → 0
V ∂T P V ∂P T
V T α2
CP − CV = →0 as T → 0
KT
CV (T )
T
⇒ C (T ) → 0
S(T )−S(0) = dT V as T → 0
T =0 T
8.044 L20B6
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
√
 
√
3 −1/2 1
p(px) = √
 e  N e 1/2 √ e
−/2<>
4πm 3N < >
1
= √ e−/2<>
4πm < >
Now use = p2
x /2m and < >=< p 2 > /2m.
x
1 −p2 /2<p2 >

p(px) = e x x
2π < p2
x >
8.044 L13B15
Homework problem on classical harmonic oscillators
1
p(pi, qi) = exp[−/ < >]
(2π/ω) < >
1
= exp[−p2
i /2mkT ] exp[−(mω 2/2kT )q 2]
i
(2π/ω)kT
= ···
Ensembles
•
Microcanonical: E and N fixed
Starting point for all of statistical mechanics
Difficult to obtain results for specific systems
• Canonical: N fixed, T specified; E varies
Workhorse of statistical mechanics
• Grand Canonical: T and µ specified; E and N vary
Used when the the particle number is not fixed
8.044 L21B1
2
1 IS THE SUBSYSTEM OF INTEREST.
2, MUCH LARGER, IS THE REMAINDER OR THE "BATH".
ENERGY CAN FLOW BETWEEN 1 AND 2.
THE TOTAL, 1+2, IS ISOLATED AND REPRESENTED BY A
MICROCANONICAL ENSEMBLE.
8.044 L21B2
For the entire system (microcanonical) one has
volume of accessible phase space consistent with X

p(system in state X) =
Ω(E)
In particular, for our case
p({p1, q1}) ≡ p(subsystem at {p1, q1}; remainder undetermined)
Ω1({p1, q1})Ω2(E − E1)

=
Ω(E)
8.044 L21B3
k ln p({p1, q1}) = k lnΩ1 + k ln Ω2(E

− E1) − k

ln Ω(E)

k ln 1 = 0 S2(E − E1) S(E)
∂S2(E2)
S2(E − E1) ≈ S2(E) − E1

∂E
2
evaluated at E2 = E
8.044 L21B4
H1({p1, q1})
k ln p({p1, q1}) = − + S2(E) − S(E)

T
The first term on the right depends on the specific

state of the subsystem.
The remaining terms on the right depend on the reser-

voir and the average properties of the subsystem.
8.044 L21B5
In all cases, including those where the system is too
small for thermodynamics to apply,
H1({p1, q1})
p({p1, q1}) ∝ exp[− ]
kT
H1({p1, q1})
exp[− ]
= kT
H1({p1, q1})
exp[− ]{dp1, dq1}
kT
8.044 L21B6
If thermodynamics does apply, one can go further.
S(E) = S1(< E1 >) + S2(< E2 >)

S2(E) − S(E) =
S
2
(E) − S2(< E2 >) −S1(< E1 >)
≈ (∂S2(E2)/∂E2) < E1 >=< E1 > /T
H1({p1, q1}) < E1 >
k ln p({p1, q1}) = − + − S1
T T
(< E1 > −T S1) H1({p1, q1})
p({p1, q1}) = exp[ ] exp[− ]
kT kT
≡ 1/Z
8.044 L21B7
< E1 > −T S1 = U1 − T1S1 = F1
−1 H({p, q})
p({p, q}) = Z exp[− ]
kT
Z is called the partition function.
H({p, q})
ZN (T, V ) = exp[− ]{dp, dq}
kT
(E − T S) F (T, V, N )
= exp[− ] = exp[− ]
kT kT
8.044 L21B8
In the canonical ensemble, the partition function is
the source of thermodynamic information.
F (T, V, N ) = −kT ln ZN (T, V )

∂F
S(T, V, N ) = −
∂T V,N

∂F
P (T, V, N ) = −
∂V T,N
8.044 L21B9
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Canonical Ensemble
p(E)
p(E) ∝ e−E/kT NOT!
p({p, q}) ∝ e−H({p,q})/kT
8.044 L22B1
ADVANTAGES OF CANONICAL OVER MICROCANONICAL ENSEMBLE
1) ONE INTEGRATES OVER ALL PHASE SPACE
MICROCANONICAL CANONICAL
Ω
Ζ
SURFACE OF
CONSTANT E
Φ
8.044 L22B2
2) SEPARATION
let H = Ha + Hb , then e−H/kT = e−Ha/kT e−Hb/kT
⇒ p({p, q}) = p({p, q}a) p({p, q}b) (a & b are SI)
⇒ Z = Za Zb ⇒ F = Fa + Fb ⇒ S = Sa + Sb etc.
8.044 L22B3
⇒ For N similar, non-interacting systems
Z = (Z1)N , F = N F1 , S = N S1
⇒ For N indistinguishable particles
(Z1)N
Z= , correct Boltzmann counting
N!
8.044 L22B4
Example Non-interacting classical monatomic gas
Npi · pi
�
N
� (Z1)N
H
= =
Hi ⇒
Z
=
i=1 2m
i=1 N !
p2
x + p2
+ p2
y z
H1( r) =
p,
2m
−(p2 2 2
x +py +pz )/2mkT
p1( r) = e
p, /Z1
Gaussian px ⇒< p >=< p2

·p x + p2 + p2 >= 3mkT
y z
< H1 >= 3/2 kT
8.044 L22B5
dpxdpy dpz dxdydz
−(p2 2 2
x +py +pz )/2mkT
Z1 = e
h3
3/2
2πmkT
= (2πmkT )3/2LxLy Lz /h3 =V

h2
 N
3N/2  3/2
1 N 2πmkT 1  2πmkT 
 
Z(T, V, N ) = V = V 
N! h2 N!  2
h 


units of cm−3
 N
3/2
1  V 2πmkT
=  v0
N ! v0 h2
8.044 L22B6
F = −kT ln Z
  
2/3
 3N  2πmkT v0 
= −kT

−N ln N + N + N ln(V /v0) +
 
ln 



2 h2
3N
use N = N ln e = ln(e2/3)
2
  
2/3mkT v2/3
 V 3N  2πe 0 
=
−kT N

ln + ln





N v0 2
h2
8.044 L22B7
� �
∂F
1 1
N kT
P =
− = (−1)(−kT N ) V
=
∂V T,N N v0
N v0 V
� �
 
∂F
3N 1 ()

S = − = k[ ] + kT 
∂T
V,N 2 () T
  
3N
V 3N
2π(ev ) 2/3mkT
= k + k N ln +
ln


0
2


2
N v0 2
h
3
E = F + T S = N kT
2
8.044 L22B8
Find the adiabatic path, ∆S = 0.
V V0 3
ln − ln = − (ln T − ln T0)
N v0 N v0 2
 −3/2
V 3 T V T

ln = − ln ⇒ = 
V0 2 T0 V0 T0
8.044 L22B9
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Canonical Ensemble
CLASSICAL QUANTUM
p({p, q}) = e−H({p,q})/kT /Z p(state) = e−Estate/kT /Z

Z= e−H/kT {dp, dq} Z= e−Estate/kT
states
8.044 L23B1
EXAMPLE 2 LEVEL SYSTEM: STATES OF AN IMPURITY IN A SOLID
E= ε g
* EXCITED
E=0 1
GROUND
STATE
g-FOLD
DEGENERACY
LOCATION INTERNAL
ENERGY LEVELS PHYSICAL DIFFERENCE
8.044 L23B2
STATES: |0> , |1

>, ·· · |g >
E=0 E=

Z1 = e−Estate/kT = 1 × e0 + g × e−/kT = 1 + ge−/kT
states
p(state) = e−Estate/kT /Z1
1
= for |0 >
1 + ge−/kT
e−/kT
= for |i > i = 1, · · · g
1 + ge−/kT
8.044 L23B3
g
1+ g
p(E= ε)
1
1+ g
p(state = i = 0)
ε/ k T
8.044 L23B4
Assume ε
• N impurities (N 1) ε0
• = 0(V /V0)−γ
V0 V
Z = Z1N F (T, V, N ) = −kT ln Z = −N kT ln Z1
 
)e−/kT
∂F 
g( 2 
S=− = N k ln Z1 + N kT  kT
−/kT

∂T V 1 + ge
8.044 L23B5

−/kT
e
S = N k ln(1 + ge−/kT ) + gN k
kT 1 + ge−/kT
g e−/kT
U = F + TS = N −/kT
= N p(E = )
1 + ge

∂F ∂F ∂
P =− =−
∂V T,N ∂ T ∂V T
−γ
V

g
−( kT ) e−/kT γ γU
= N kT − =
1 + ge−/kT V V
8.044 L23B6
ALTERNATIVE WAY OF FINDING U
Usually (but not always) U =< H > .

If so, U = H({p, q}) p({p, q}) {dp, dq}

But Z = c e−H({p,q})β {dp, dq} β ≡ 1/kT
8.044 L23B7
 
∂Z
�


= c −H({p, q})e−H({p,q})β {dp, dq}
∂β N,V
 
1  ∂Z  �
e−H({p,q})β
−
= H({p, q}) � −H({p,q })β {dp, dq}
Z ∂β N,V e
{dp , dq }�
� ��
p({p,q})
 
1  ∂Z 
− =U
Z ∂β N,V
8.044 L23B8
Example Monatomic Gas
3N/2
1 N 2πmkT
Z= V = α β −3N/2
N! h2
 
1 3N 1
 −3N/2 3
U =− − αβ = N kT
α β −3N/2 2 β 2
8.044 L23B9
Example 2 Level System
N
Z = 1 + ge− β
−N N −1
U = − 1 + ge− β N 1 + ge− β −ge− β
gN e−/kT
=
1 + ge−/kT
8.044 L23B10
Background Classical Harmonic Oscillator
p2 1
H(p, x) = + Kx2
2m 2
1 p2 1 x2
p(p, x) = √ exp[− ]� exp[− ]
2πmkT 2mkT 2π(kT /K) 2(kT /K)
�  
m −1 
∂Z

Z = 2π kT ∝ β −1 U = = kT, CV = N k ∀T
K Z ∂β
8.044 L23B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Polyatomic Gases
Non-interacting, identical ⇒ Z = N1! Z1N Find Z1
Each molecule has # atoms ⇒ 3# position coordi-

nates
3# = 3
+ nr
+ (3#

−3 − nr )
C.M. rotation nv , vibration
8.044 L24B1
MONATOMIC DIATOMIC LINEAR TRI. NON-LINEAR TRI.
Xe HS CO2 H2 O
3 3 3 3
0 2 2 3
0
1
4
3
3
6
9
9
8.044 L24B2
C.M. Motion:
Particle in a box ∆E s kT ⇒
classical
Rotation:
(H2 νrot = 3.65 × 1012 Hz → 175 K ) ⇒ Q.M.
Vibration:
(H2 νvib = 1.32 × 1014 Hz → 6, 320 K ) ⇒ Q.M.
H = HCM + Hvib + Hrot ⇒ problem separates
8.044 L24B3
Vibration
 
nv
1 1 Ki 2

Hvib =  2
Ki a i + ȧi
2
i=1 2 2 ωi
nv 1 dimensional harmonic oscillators, use Q.M.
Hψ
ˆ n = nψn n = (n + 1 hω
n = 0, 1, 2, · · ·
2 )¯
The energy levels are non-degenerate.
8.044 L24B4
∞
�
−(n+ 1
p(n) = e 2 )h̄ω/kT / e−n/kT
ε n=0
hω ∞
� ∞ �
� �n
−(n+ 1
2 )¯ −1 −¯
e hω/kT
= e 2 ¯hω/kT hω/kT
e
n=0 n=0
hω
7
1
2
5
hω
1 � �
e− 2 h̄ω/kT /
1 − e−¯hω/kT
1
2
=
3
2 hω
1
hω
1 � �� n
p(n) = 1 − e−¯hω/kT e−¯hω/kT = (1 − b)bn
8.044 L24B5
Geometric or Bose-Einstein
p(n)
b 1
<n> = = ¯hω/kT
1−b e −1
→ e−¯hω/kT when kT ¯
hω
8.044 L24B6
1
For kT ¯
hω < n > → 2
1 + ¯kT
hω + 1 ¯
hω
2 kT ··· − 1

kT 1 kT 1 ¯
hω
= ≈ 1−2
¯
hω 1 + 1 ¯
hω ¯
hω kT
2 kT
kT
= −1
hω 2
¯
< >= (< n > + 1

2 hω → kT
)¯ kT ¯
hω (Classical)
→ 1 hω kT ¯
2 ¯ hω (Ground state)
8.044 L24B7
3 <n> < ε>
1
kT
1
2 hω
0
1 2 3 T
kT/h ω
-1
8.044 L24B8
� �
∂<> d<n>
CV = N = N h̄ω
∂T V dT
� �2
¯
hω e¯hω/kT
= Nk � �2
kT e¯hω/kT −1
�
�
h̄ω 2 −¯hω/kT
→ Nk e kT ¯
hω (energy gap behavior)
kT
→ Nk kT h̄ω
8.044 L24B9a
CV /Nk
1
ENERGY
GAP
BEHAVIOR
2 LEVEL SYSTEM
SHOWING SATURATION
1 2 3
kT/h ω
8.044 L24B9b
High and low temperature behavior without solving the
complete problem Consider first the high T limit.
e - ε/ k T
∆ contains ∆
h̄ω states
kT ε
hω
∞

Z1 = e−n/kT
n=0
∞1 −E/kT kT ∞ −y kT
≈ e dE = e dy = ∝ β −1
0 ¯
hω hω 0
¯ ¯
hω
8.044 L24B10
Zvib = Z1N ∝ β −N
 
1  ∂Z 
Uvib =− = −β N (−N )β −N −1 = N kT
Z ∂β N
Cvib = N k
Next, consider the low T limit.
e - ε/ k T
kT ⇒ consider only 2 states
1
hω 3
hω ε
2 2
8.044 L24B11
−3
2¯hω/kT
e 1 −¯
hω/kT
p(n = 1) ≈ = ≈ e
−1 ¯
hω/kT − 3¯
e¯hω/kT + 1
e 2 + e 2 hω/kT
p(n = 0) ≈ 1 − e−¯hω/kT

hω 1 − e−¯hω/kT + 32 N ¯
< E >= 12 N ¯ hωe−¯hω/kT
= 12 N ¯ hωe−¯hω/kT
hω + N ¯
2
∂<E> ¯hω −¯
hω/kT = N k ¯
hω −¯
hω/kT
CV = = N h̄ω e e
∂T kT 2 kT
8.044 L24B12
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Angular Momentum in 3 Dimensions

θ, φ)
CLASSICAL, 3 numbers: (Lx, Ly , Lz ); (|L|,
QUANTUM, 2 numbers: magnitude and 1 component
ˆ·L
ˆψ
≡ ˆ2 ψ 2ψ
L l,m L l,m = l(l + 1)¯
h l,m l = 0, 1, 2 · · ·
Lˆz ψl,m = mh̄ ψl,m m = l, l − 1,· · · − l

2l+1 values
8.044 L25B1
Specification: 2 numbers l & m → ψl,m or |l, m >
Molecular rotation
In general
1 2 1 2 1 2 I3 = 0
Hrot = L1 + L2 + L3
2I1 2I2 2I3 L3 = I3 θ̇3
= 0
For a linear molecule

1 2 2 1 I1 = I2 ≡ I⊥
Hrot = (L1 + L2) = L ·

L
2I⊥ 2I⊥
8.044 L25B2
1 2
Ĥrot = ˆ
L
2I⊥
ε/k
Ĥrot |l, m > = l |
l, m >
2 0 ΘR 9 l= 4
¯h2
= l(l + 1)|l, m >
2I⊥
1 2 ΘR 7 l= 3
l depends on l only;
it is 2l + 1 fold degenerate.
6 ΘR 5 l= 2
l = kΘR l(l + 1) 2 ΘR 3 l= 1
0 1 l= 0
h2
¯
ΘR ≡ (rotational temp.)
2I⊥
k
8.044 L25B3
1 −l(l+1)ΘR/T
p(l, m) = e
ZR

ZR = e−l(l+1)ΘR/T = (2l + 1)e−l(l+1)ΘR/T
l,m l
For T ΘR ZR ≈ 1 + 3e−2ΘR/T = 1 + 3e−2ΘRkβ
1 ∂Z 6ΘR k e−2ΘRkβ −2ΘR /T

< >= − = ≈ 6Θ R k e
Z ∂β 1 + 3e−2ΘRkβ
8.044 L25B4

∂<> 2ΘR −2ΘR/T
CV |rot =N = 6ΘR N k 2
e
∂T T
2
2ΘR
= 3N k e−2ΘR/T (energy gap behavior)
T
For T ΘR, convert the sum to an integral.
∞
ZR ≈ (2l + 1)e−l(l+1)ΘR/T dl
0
8.044 L25B5
x ≡ (l2 + l)ΘR/T dx = (2l + 1)ΘR/T dl
T ∞ −x T 1
ZR ≈ e dx = = β −1
ΘR 0 ΘR kΘR
1 ∂Z (−1)(−1)Z/β
< >= − = = β −1 = kT
Z ∂β Z
∂<>
CV |rot =N → Nk (classical result)

∂T
8.044 L25B6a
Image removed due to copyright reasons.
Figure 22-1 from

Morse, P. M. Thermal physics. 2nd ed. New York, NY: W. A. Benjamin, 1969. ISBN: 0805372024.
8.044 L25B6b
H = HCM + Hrot + Hvib
CV (T ) = CV|CM + C |
V rot
+ C |
V vib
all T appears at modest T only at highest T
8.044 L25B7a
Image removed due to copyright reasons. Image removed due to copyright reasons.
Figure 12-16 from Figure 12-2 from

F. W. Sears, and G. L. Salinger. F. W. Sears, and G. L. Salinger.
Thermodynamics, Kinectic Theory, and Statistical Physics. 3rd ed. Thermodynamics, Kinectic Theory, and Statistical Physics. 3rd ed.
Reading, MA: Addison-Wesley Pub. Co., 1975. ISBN: 020106894X. Reading, MA: Addison-Wesley Pub. Co., 1975. ISBN: 020106894X.
8.044 L25B7b
Raman Scattering
BEFORE AFTER
εi εf
νi
νf
∆ε = εf - εi = h( νi - νf )
FREQUENCY CHANGES IN THE SCATTERED LIGHT CORRESPOND TO

ENERGY LEVEL DIFFERENCES IN THE SCATTERER.
WHICH ENERGY LEVEL CHANGES OCCUR DEPEND ON SELECTION

RULES GOVERNED BY SYMMETRY AND QUANTUM MECHANICS
8.044 L25B8
Example Rotational Raman Scattering
Selection rule: ∆l = ±2
∆νl↑ = −(kΘR/h)[(l + 2)(l + 3) − l(l + 1)]

= −(4l + 6)(kΘR/h)
⇒ uniform spacing between lines of 4(kΘR/h)
Il↑ ∝ number of molecules with angular momentum l
∝ (2l + 1)e−l(l+1)ΘR/T
8.044 L25B9
ROTATIONAL RAMAN SPECTRUM OF A DIATOMIC MOLECULE
BOLTZMANN FACTOR
I( ∆ν)
LEVEL DEGENERACY
∆ν 0
4(k ΘR/ h )
-6(k ΘR/ h )
8.044 L25B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Thermal Radiation Radiation in thermal equilibrium
with its surroundings
E0
k
k·
E = E0 e r−ωt)
i( ω = c|k|
B0
= B0 ei(k·r−ωt)
B 1k × E0 /c
B0 =
8.044 L26B1
Time average energy density 0 |2
u = 12 0|E
Time average energy flux jE = (cu) 1k
Time average pressure (⊥ to k) P =u
Thermal radiation has a continuous distribution of

frequencies.
u( ν,T)
Peaks near hν = 3kB T

(h/kB ∼ 5 × 10−11 K-sec)
ν
8.044 L26B2
Spectral Region ν (Hz) T (K) Thermal Rad.
Radio 106 5 × 10−5
Microwave 1010 0.5 cosmic background
Infrared 1013 5 × 102 room temp.
Visible 1
2 × 1015 2 × 104 sun’s surface
Ultraviolet 1016 5 × 105
X ray 1018 5 × 107 black holes
γ ray 1021 5 × 1010
8.044 L26B3
ENERGY ABSORBED
ABSORPTIVITY α( ν,T) ENERGY INCIDENT

ISOTROPIC
ENERGY EMITTED
EMISSIVE POWER e( ν,T) AREA
ISOTROPIC
8.044 L26B4
THERMAL RADIATION: PROPERTIES
2 ENERGY FLUXES,
IN AND OUT OF CAVITY B
CAVITY A TB
TA
CAVITY B
FILTER: FREQUENCY
OR POLARIZATION
ASSUME TA = T B AND THERMAL EQUILIBRIUM
8.044 L26B5
CONCLUSIONS:
• u(ν, T ) is independent of shape and wall material
• u(ν, T ) is isotropic
• u(ν, T ) is unpolarized
8.044 L26B6
CONSIDER AN OBJECT IN THE CAVITY,
IN THERMAL EQUILIBRIUM
COMPUTE THE ENERGY FLUX

T
dA
c∆
t
θ
∆A
8.044 L26B7

∆E = (E in cylinder) p(θ, φ) dθ dφ

sin θ 1
= (u ∆A cos θ c∆t) dθ dφ
2 2π
π/2 cos θ sin θ 2π 1
= c u ∆A ∆t dθ dφ
0 2 0 2π

1/4 1
⇒ energy flux onto dA = 14 c u(ν, T )
8.044 L26B8
Momentum Flux
Plane wave momentum density = uc
p 1k
|∆p| = 2|p⊥| since ⊥ in = −

p p⊥ out
8.044 L26B9
2 cos θ
|∆p|ν = (E in cylinder) p(θ, φ) dθ dφ
c
π/2 2π 1
= u(ν, T ) ∆A ∆t cos2 θ sin θ dθ dφ
0 0 2π

1/3 1
= 1 u(ν, T ) ∆A ∆t
3
∞
⇒ P (T ) = 1
3 u(ν, T ) dν
0
8.044 L26B10
Apply detailed balance to the object in the cavity.
Eout = Ein
e dA = α ( 14 c u(ν, T )) dA
e(ν, T )
⇒ = 1 c u(ν, T )
4
α(ν, T )
This ratio has a universal form for all materials.

The result is known as KIRCHOFF’S LAW.
8.044 L26B11
Black Body Radiation
If α ≡ 1 ≡ “Black”
Then e(ν, T ) = 1
4 c u(ν, T )
OVEN
CAVITY Measure e(ν, T )
AT and obtain u(ν, T )
T
8.044 L26B12
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Thermodynamic Approach
∞
u(T ) ≡ u(ν, T ) dν
Then
E(T, V ) = u(T )V
P (T, V ) = 1 u(T )
3
8.044 L27B1
This is enough to allow us to find u(T ).
dE = T dS − P dV
� � � � � �
∂E ∂S ∂P
= T −P =T −P
∂V T ∂V T ∂T V
= 1 T u (T ) − 13 u(T )
3
also = u(T )
⇒ u(T ) = u(T )
u(T ) = AT 4
8.044 L27B2
Emissive Power of a Black (α = 1) Body
e(ν, T ) = 14 c u(ν, T ) ⇒ e(T ) = 1

c u(T ) = 1 AcT 4
4 4
e(T ) ≡ σT 4
This is known as the STEFAN-BOLTZMANN LAW.
σ = 56.7 × 10−9 watts/m2K 4
8.044 L27B3
Statistical Mechanical Approach
H?
Single normal mode (plane standing wave) in a
rectangular conducting cavity.
Ex
z
0 L

E (
r, t) = E(t) sin(nπz/L)
1x

0,0,n,1x

B (
r, t) = (nπc 2/L)−1Ė(t) cos(nπz/L)
1y
0,0,n,
y1
8.044 L27B4
Energy density = 12 0E +11B
·E ·B
[no r or t average]
2µ 0
 
V 1
H =  1 E 2(t)
2 0 + 1
2 (nπc2/L)−2Ė 2(t)
2 µ0
V 0 2 −2 2

= E (t) + (nπc/L) Ė (t)
2 2
⇒ Each mode corresponds to a harmonic oscillator.
8.044 L27B5
Count the modes.
nx,ny ,nz = |E |j sin(nxπx/L) sin(ny πy/L) sin(nz πz/L)eiωt

E
The unit polarization vector j has 2 possible orthog-

onal directions and ni = 1, 2, 3 · · ·.
2
∂ 2E πc
− c2 ∇
2E =0 ⇒ ω2 = (n2
x + n2 + n2
)
y z
∂t2 L
8.044 L27B6
If the radian frequency < ω
nz

L
R R= n2
x + n2
y + n2
z = ω
πc
ny # modes (freq. < ω)
GRID SPACING =2×1

8 × 4
πR3
3
1 UNIT
3
nx = π3 πc
L ω 3
8.044 L27B7

d# L 3 2 V
D(ω) = =π ω = 2 3 ω 2
dω πc π c
D( ω)
ω
8.044 L27B8
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Normal modes of the radiation field in a rectangular
cavity with conducting walls

πc

E nx,ny ,nz , (
r, t) ω= n2
x + n2 + n2
y z
• Harmonic oscillators
• D(ω) = π 2Vc3 ω 2, ω≥0
8.044 L28B1
Classical Statistical Mechanics
D(ω) kB T 2
< (ω) >= kB T ⇒ u(ω, T ) =< (ω) > = 2 3ω
V π c
∞
u(T ) = u(ω, T ) dω = ∞
0
CLASSICAL
u( ω, T)
MEASURED
ω
8.044 L28B2
Quantum Statistical Mechanics
hω
< (ω) >= ¯hω/kT +¯
hω/2
e −1
D(ω) h̄ ω3
u(ω, T ) =< (ω) > = 2 3 ¯
+ z. p. term
V π c e hω/kT −1
du(ω, T )
To find the location of the maximum, set = 0.
dω
The maximum occurs at h̄ω/kT ≈ 2.82.
8.044 L28B3

Z= Zi Zi = e−¯hω/2kT (1 − e−¯
hω/kT )−1
states i
The first factor in the expression for Zi comes from

the zero-point energy.

F (V, T ) = −kT ln Z = −kT ln Zi
states i
∞
= −kT D(ω) [ln Zi] dω
0
8.044 L28B4
∞
F (V, T ) = −kT D(ω) − ln(1 − e−¯hω/kT ) dω + · · ·
0
kT V ∞ 2
= 2 3 ω ln(1 − e−¯hω/kT ) dω
π c 0
V ∞
= 2 3 3 (kT )4 x2 ln(1 − e−x) dx
π c ¯
h 0
π 4
− 45
1 π2 4V
= − (kT )
h3
45 c3¯
8.044 L28B5

∂F 1 π 2
4
P = − = (kT )
∂V T h)3
45 (c¯

∂F 4 π2 4T 3 V
S = − = k
∂T V h)3
45 (c¯

1 4 1 π2 4 V
E = F + TS = − + (· · ·) = (kT )
45 45 h)3
15 (c¯
Note: P = 1
3 E/V = 1 u(T ) independent of V .
3
8.044 L28B6
NOTE: THE ADIABATIC PATH IS T3 V=CONSTANT
T ADIABATIC
T / T 0 = (V/V 0 ) - 1 / 3
8.044 L28B7
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Wavefunctions, One Particle
r and s are the variables.

Hamiltonian H(
ˆ ˆ
r, ˆ
ˆ
p, s)
n is a state index and could
Wavefunction ψn(
r, s)
have several parts.
For an e− in hydrogen ψ = ψn,l,ml,ms (

r, s)
H(
ˆ ˆ
r, ˆ
ˆ
p, s) ψn(
r, s) = En ψn(
r,
s)
8.044 L29B1
ψn(
r, s) often factors into space and spin parts.
space spin
ψn(
r, s) = ψn (
r) ψn (
s)
−αx 2 /2 √
space
ψn (x) ∝ e Hn( α x) H.O. in 1 dimension
k·
space
ψn r) ∝
( e r
i free particle in 3 dimensions
8.044 L29B2
spin
ψn (
s)
Spin is an angular momentum so for a given value of

the magnitude S there are 2S + 1 values of mS .
For the case of S = 1/2 the eigenfunctions of the z

component of s are φ1/2(s) and φ−1/2(s)
h̄
ˆz φ1/2(s) =
S φ1/2(s)
2
h̄
ˆz φ−1/2(s) = − φ−1/2(s)
S
2
8.044 L29B3
spin
ψn (
s) is not necessarily an eigenfunction of Ŝz . For
example one might have
spin
ψn s) = √1 φ1/2(s) + √1 φ−1/2(s)
(
2 2
In some cases ψn(

r, s) may not factor into space and
spin parts. For example one may find

ψn(x, s) = f (x) φ1/2(s) + g(x) φ−1/2(s)
8.044 L29B4
Many Distinguishable Particles, Same Potential,
No Interaction
Lump space and spin variables together

r1, s1 → 1
r2, s2 → 2 etc.
H(1,
ˆ ˆ0(2) + · · · H
ˆ0(1) + H
2, · · · N ) = H ˆ0(N )
In this expression the single particle Hamltonians all

have the same functional form but each has arguments
for a different particle.
8.044 L29B5
The same set of single particle energy eigenstates is
available to every particle, but each may be in a dif-
ferent one of them. The energy eigenfunctions of the
system can be represented as products of the single
particle energy eigenfunctions.
ψ{n}(1, 2, · · · N ) = ψn1 (1)ψn2 (2) · · · ψnN (N )
{n} ≡ {n1, n2, · · · nN }. There are N #s, but each ni

could have an infinite range.
Ĥ(1, 2, · · · N ) ψ{n}(1, 2, · · · N ) = E{n} ψ{n}(1, 2, · · · N )
8.044 L29B6
Many Distinguishable Particles, Same Potential,
Pairwise Interaction
N

H(1,
ˆ 2, · · · N ) = H
ˆ0(i) + 1 H înt(i, j)
2
i=1
i=j
The ψ{n}(1, 2, · · ·
N ) are no longer energy eigenfunc-
tions; however, they could form a very useful basis set

for the expansion of the true energy eigenfunctins.
8.044 L29B7
Indistinguishable Particles
Pîj f (· · · i · · · j · · ·) ≡ f (· · · j · · · i · · ·)
(Pîj )2 = Iˆ ⇒ eigenvalues of Pˆ
ij are + 1, −1
It is possible to construct many-particle wavefunctions

which are symmetric or anti-symmetric under this in-
terchange of two particles.
Pîj ψ (+) = ψ (+) Pîj ψ (−) = − ψ (−)

8.044 L29B8
Identical ⇒ no physical operation distinguishes be-
tween particle i and particle j. Mathematically, this
means that for all physical operators Ô
[O,
ˆ Pîj ] = 0
⇒ eigenfunctions of Ô must also be eigenfunctions of

P̂ij .
(+)
⇒ energy eigenfunctions ψE must be either ψE or
(−)
ψE .
8.044 L29B9
⇒ states differing only by the interchange of the spa-
tial and spin coordinates of two particles are the same
state.
Relativistic quantum mechanics requires
integer spin ↔ ψE (+) [Bosons]

half-integer spin ↔ ψE (−) [Fermions]
8.044 L29B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Composite Particles
• Composite Fermions and Composite Bosons
• Count the number of sign changes as all the con-

stituents are interchanged
• Well defined statistics (F-D or B-E) as long as the

internal degrees of freedom are not excited
8.044 L30B1
The constitutents of nuclei and atoms are e, p & n.
Each has S = 1/2.
N even ⇒ even # of exchanges.
ψ → (+)ψ ⇒
B-E
also N even ⇒ integer spin
N odd ⇒ odd # of exchanges.
ψ → (−)ψ ⇒
F-D
also N odd ⇒ half-integer spin
8.044 L30B2
Particle Nuclear Spin Electrons Statistics
H (H1) 1
2 1 B-E
D (H2) 1 1 F-D
T (H3) 1
2 1 B-E
He3 1
2 2 F-D
He4 0 2 B-E
Li6 1 3 F-D
Li7 3
2 3 B-E
H2 0 or 1 2 B-E
x2 integer ()×2 B-E
8.044 L30B3
Let α(
r, s), β( s), · · · be single particle wavefunctions.
r,
A product many-particle wavefunction, α(1)β(2), does
not work.
Instead, use a sum of all possible permutations:
(+)
Ψ2 = √1 (α(1)β(2) + α(2)β(1))
2
(+)
ΨN
= √1 √1 (α(1)β(2)γ(3) · · ·)
N! n ! permutations
α α
8.044 L30B4
The antisymmetric version results in a familiar form,
a determinant.
(−)
Ψ2 = √1 (α(1)β(2) − α(2)β(1))
2
→ states

α(1) β(1) γ(1) · · ·

α(2) β(2) γ(2) · · ·
(−) 1
ΨN = √ ↓ particles
N ! α(3) β(3) γ(3) · · ·

..
.. ..
8.044 L30B5
(−)
•
ΨN
= 0 if 2 states are the same since 2 columns
are equal: Pauli Principle.
(−)
• ΨN = 0 if 2 particles have the same
r and s since
2 rows are equal.
• Specification: indicate which s.p. ψs are used.

{nα, nβ , nγ , · · ·}
An ∞ # of entries, each ranging

from 0 to N but with α nα = N .
8.044 L30B6
|1, 0, 1, 1, 0, 0, · · · Fermi-Dirac
|2, 0, 1, 3, 6, 1, · · · Bose-Einstein

αnα = E Prime indicates
nα =
N
α
α
Example Atomic configurations
(1S)2(2S)2(2P )6 ↔ Ne
(1S)2(2S)2(2P )6(3S)1 ↔
Na
(1S)1(2S)1 ↔ He*
8.044 L30B7
Statistical Mechanics Try Canonical Ensemble

Z(N, V, T ) = e−E(state)/kT
states

= e−E({nα})/kT
{nα}

= e−αnα/kT
{nα} α
This can not be carried out. One can not interchange

the over occupation numbers and the over states
because the occupation numbers are not independent

( nα = N ).
8.044 L30B8
T=0 LOWEST POSSIBLE TOTAL ENERGY
BOSE: ALL N PARTICLES IN LOWEST ε SINGLE PARTICLE STATE
nα( ε)
Nδ( ε)
ε
FERMI: LOWEST N SINGLE PARTICLE STATES EACH USED ONCE
ε < εF, εF CALLED THE FERMI ENERGY
nα( ε)
1
εF ε
8.044 L30B9
Free Particle in a Box
r, s) =
ψ( √1
k·
e r ψ spin(s)
i k = h2 k 2
¯ =0
if B
V 2m
Use periodic boundary conditions.
r + nxLxx
ψ( ˆ + ny Ly yˆ + nz Lz z,
ˆ s) = ψ(
r, s)
⇒ k = mx 2π
Lx ˆ
x + m 2π
y
y Ly ˆ + m 2π ẑ
z Lz
8.044 L30B10
2 π/ L x
ky −1 ( 2π )−1 ( 2π )−1
D(k) = ( 2π
Lx ) Ly Lz
2 π/ L y
V
= (2π) 3
wave vectors only!

kx
kz
radius k F
N = 2 × D(k) × 4 πk
3 F
3
1/3 ky
kF = 3π 2(N/V ) kx
8.044 L30B11
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
2 π/ L x
ky −1 ( 2π )−1 ( 2π )−1
D(k) = ( 2π
Lx ) Ly Lz
2 π/ L y
V
= (2π) 3
wave vectors only!

kx
kz
radius k F
N = 2 × D(k) × 4 πk
3 F
3
1/3 ky
kF = 3π 2(N/V ) kx
8.044 L30B11
Fermions: Non-interacting, free, spin 1/2, T = 0
Fermi sea
kz
Fermi surface
kF
Fermi wave vector
1/3
ky 2
kF = 3π (N/V )
Fermi energy
kx
h 2 kF
F = ¯ 2 /2m
8.044 L31B1
1/2
4 3 2m
#() = 2 × D(k) × πk () k() =
3 h2
¯
3/2
8π V 2m
= 3/2
h2
3 (2π)3 ¯

V 2m 3/2 1/2
D() = 2 2
2π ¯
h
D( ε)
h 2 kF
¯ 2
F = 2m ∝ (N/V )2/3
εF ε
8.044 L31B2
D() = a 1/2

F 3/2
N = D() d = 2 a
3 F
0

F 5/2
E= D() d = 2 a
5 F = 3N
5 F ∝ N (N/V )2/3
0
8.044 L31B3
Consequences: Motion at T = 0
h
= ¯k
p h F
F = ¯k
p ¯
hk
vF = m F
Copper, one valence electron beyond a filled d shell
N/V = 8.45 × 1022 atoms-cm−3

kF = 1.36 × 108 cm−1 vF = 1.57 × 108 cm-s−1
pF = 1.43 × 10−19 g-cm-s−1 F /kB = 81, 000K
8.044 L31B4
Consequences: P (T = 0)
E=3
5 N F F ∝ (N/V )2/3

3 ∂F
∂E
P = − ∂V = −5N =2
5 (N/V )F ∝ (N/V ) 5/3
N,S
∂V N
F
−2
3 V

∂P
= −
5
3 (P/V ) at T = 0
∂V N,T
8.044 L31B5

1 ∂V 31 3 1
KT = − = =
V ∂P N,T 5P 2 (N/V )F
For potassium
F = 2.46 × 104K = 3.39 × 10−12 ergs
(N/V )conduction = 1.40 × 1022 cm−3
KT = 1.40×10221.5
= 31.6 × 10−12 cm3-ergs−1
×3.39×10−12
The measued value is 31 × 10−12 !
8.044 L31B6
Temperature Dependence of < nk,m >
< nk,m >= f () only, as in the Canonical Ensemble
<n>
~ kT
1
T=0
T << εF/ k B
εF ε
8.044 L31B7
Estimate CV
Classical Quantum
# electrons N N
# electrons influenced N ∼ N × kT
F
∆ 3 kT ∼ kT
2
3 N kT (kT )2
∆E 2 ∼N F
CV 3
N k ∼ 2N k kT
2 F
Exact result: CV = π 2 N k kT
2 F
8.044 L31B8
Magnetic Susceptibility
ε

= Hẑ
H εF
k,↑ = (k) + µeH
k,↓ = (k) − µeH
µe H
D ( ε) D ( ε)
8.044 L31B9
   
N D(F )   N D(F )

N↓ − N↑ =  + µe H − − µe H
2 2 2 2
= µeHD(F ) ⇒ M = µ2
e HD(F ) ≡ χH
χ = µ2
e D(F )
This expression holds as long as kT F , so χ is

temperature independent in this region. This is not
Curie law behavior. It is called Pauli paramagnetism.
8.044 L31B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Image from Astronomy Picture of the Day (http://antwrp.gsfc.nasa.gov/apod/). Courtesy of NASA.
Stellar Configurations
• Self gravitating
• Self-consistent solution needed
• Different processes resist collapse
8.044 L32B1
Planets
• Gravity weak because of small M
• Atomic forces provide balancing pressure
8.044 L32B2
Normal Stars
• Gravitational energy starts process
• Fusion then supplies energy
•
Plasma of electrons and nuclei
• Kinetic pressure, P = nkT
• Radiation pressure, P = 1
3 u(T ), helps and domi-

nates above about 10M
8.044 L32B3
FOUR POSSIBLE END STATES OF STARS
DISPERSED GAS
NORMAL DWARF STAR INCREASING

STAR MASS
NEUTRON STAR
BLACK HOLE
8.044 L32B4
White Dwarf
• Fusion has stopped
• Collapses to a small size, nuclear spacing ∼ 1/100

that of a solid
• Electron degeneracy pressure supports it, P ∝ m1e n

5/3
• White → gray → brown (dead, cold)

8.044 L32B5
Assume uniform density of α++ and e−
2/3
EK =
(α) (e)
+EK = 3N = 3N ¯ h2 3π 2(N
E 5 e F 5 e 2me e/V )
K

small
V =4
3 πR 3 M ≈ Nαmα = (Ne/2)mα ⇒ Ne = 2M/mα
2/3 2/3
EK = 3 9π h2
¯ M 1 EP = 3 M
−5 G R
2
5 2 me mα R2
8.044 L32B6
ET = EK + EP
ET
¯
h 6
EK 3 = 2(9π)2
M R0
R0
G3m3 5
e mα
R
= 0.74 × 1051 kg-m3
EP
R0 ∝ 1/M 1/3 Stable for any M .
8.044 L32B7
Sirius B: M = 2.1 × 1030 kg
R
observed 5.6 × 106 m
our model 7.1 × 106 m (good)
better model 8.6 × 106 m (⇒ a problem)
8.044 L32B8
Our model of Sirius B implies
ne = 8.6 × 1029 cm−3
F = 4.7 × 10−7 ergs → 3.4 × 109K

(Tsurface ∼ 2 × 107 K)
But mec2 = 8.2 × 10−7 ergs ⇒ relativity needed
8.044 L32B9
Homework problem examines extreme relativistic gas
and finds
•
Softer equation of state, P ∝ m1e n4/3
• Potential for collapse if M is too large
Chandrasekhar limit (∼1935) MCh = 1.4M

8.044 L32B10
Neutron Star
• p+ + e− → n to lower coulomb energy
• Degeneracy pressure of neutrons
3∝¯
M R0 h6/G3m8
n ⇒ R0 ∼ 15 km if M = 1.4M

• Nuclear forces also contribute to P
• Rotating neutron stars seen as pulsars
• Also subject to stability limit, M ∼ 2M

8.044 L32B11
STELLAR CONFIGURATIONS
DISPERSED GAS
GRAVITY
NORMAL DWARF NEUTRON BLACK
STAR STAR STAR HOLE
GAS PRESSURE ELECTRON NEUTRON

(+ RADIATION ) DEGENERACY DEGENERACY
+ NUCLEAR
FORCES
1 g/cm3 106 g/cm3 1014 g/cm3
8.044 L32B12

MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
PARAMAGNETS
Spin 1/2 Paramagnet, Microcanonical Ensemble
Assume the moment is parallel to the spin. Take n to
be the state variable. µ is the magnetic moment of
the particle.
n = Hẑ
H
µH
(↑), N↑ = N − n, µ z, ↑ = −µH
↑ = µˆ
N-n
- µH ↓ = −µˆ
(↓), N↓ = n, µ z, ↓ = µH
8.044 L34B1
M = µ(N↑ − N↓) = µ(N − 2n) = M (n)
E ≡ N = N↑↑ + N↓↓ = −µH(N − 2n) = −HM

(n)
H &n
 
N!
S = k ln   = S(n)
n!(N − n)!
≈ k [ N ln N − n ln n − (N − n) ln(N − n) ]
8.044 L34B2

dS n
= k [−1 − ln n + 1 + ln(N − n)] = −k ln
dn N −n
S = S(n) and M = M (n) ⇒ S = S(M )

1 ∂S
= does not exist!
T ∂E M
 
H ∂S dS ∂n n −1 
− = = = −k ln 
T ∂M E dn ∂M E N −n 2µ
8.044 L34B3
n −2µH/kT 1
=e → n = N
2µH/kT
N −n e +1
M = µ(N − 2n) = µN tanh(µH/kT )
n
N/2
0 µH/k T
8.044 L34B4
MAGNETIZATION IN SPIN 1/2 PARAMAGNET
M / µN M / µN
1 1
0.8 0.8
CONSTANT H
0.6 0.6
CONSTANT T
0.4 0.4
0.2 0.2
1 2 3 4 5 6 1 2 3
k T / µH µH / kT
Nµ2 H
CURIE LAW BEHAVIOIR: M =
kT
∝ H/T
8.044 L34B5
Internal Energy
dU = d /W = T dS + HdM
/Q + d
dU ≡ adiabatic (d
/Q = 0) work
/Q = T dS,
d /Q = 0 ⇒ dS = 0 ⇒ dM = 0 ⇒ dU = 0
dU = 0 for any change: U = 0 for this model
But E ≡ N < >= −HM =

0 !!
8.044 L34B6
Energy = energy to create H field �
1
+ energy to assemble M �
2
+ energy to move M into H �

3
8.044 L34B7

1 does not appear when using d
/W = HdM . It is
/W = −M dH.
included in U when using d

2 We did not create the µ
. They do not interact.
⇒ U = 0 in the current example.

3 ·M
−H , energy of macroscopic moment in H

equals N < > in the current example.
8.044 L34B8
 
d
/Q  ∂S
CM ≡  =T = 0 since S = S(M )
dT M ∂T M
 
d
/Q  1 ∂M
CH ≡  =
−HdM )H = −H
(dU
dT H dT 0 ∂T H
2
2µH e2µH/kT
= Nk
kT (e2µH/kT + 1)2
CH
ENERGY CURIE LAW

GAP ∝ 1/T2
0 T
µH/k
8.044 L34B9
Spin 1/2 Paramagnet, Canonical Ensemble
Z1 = e+µH/kT + e−µH/kT
e+µH/kT e−µH/kT
µH p↑ = p↓ =
Z1 Z1
- µH
M = N µ(p↑ − p↓)
√
= µN tanh(µH/kT )
8.044 L34B10
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Simple Quantum Paramagnet, Canonical Ensemble
Origin of magnetic moments:
Electron spin and orbital angular momentum

+L
S ≡ J µ = gJ µB J µB ≡ e¯
h/2mec
Nuclear angular momentum
= g µ I µ ≡ e¯
I µ I N h/2mpc
8.044 L35B1
m = −gµB H m
2J + 1
m = J, J − 1, · · · − J g µBH
J
�
J
�
sinh[(J + 1 )η]
Z1(T, H) = e−m/kT
=
(eη )m
=
2
1 η]
m=−J m=−J sinh[ 2
gµB H level spacing

η≡ =
kT kT
Note Z1 = Z1(η) Z = Z1(η)N = Z(η) at fixed N
8.044 L35B2
p(m) = e−m/kT /Z1 = eη m/Z1
 
�
(gµB m)eη m

1 ∂Z1

<µ> = = gµB
m
Z1
Z1 ∂η
≡ gµB JBJ (η) M = N < µ >= gµB N JBJ (η)

 
1 
1 ∂Z1

BJ (η) =
J Z1 ∂η
1� 1 1 1 1
�
= (J + ) coth[(J + ) η] − coth[ η]
J 2 2 2 2
This is called the “Brillouin Function”.

8.044 L35B3
1 x J +1
coth x → + x 1 lim BJ (η) = η

x 3 η →0 3
e−η
coth x → 1 + 2e−2x x 1 lim B (η) = 1 −
η→∞
J J
(gµB )2J(J + 1) H
M → N High T (Curie Law)
3 kT
� �
1 −η
→ N gµB J 1 − e Low T (Energy Gap)
J
8.044 L35B4
2J
∂M ∂η (gµ )
χT ≡ = N gµB J BJ (η) =N B
BJ (η)
∂H T ∂H T kT
gµB H
Note: T and H enter only through η ≡
kT

∂η η ∂η η
= =−
∂H T H ∂T H T
We now show that this ⇒ U = 0.
8.044 L35B5
dU = T dS + HdM

∂S ∂S ∂M ∂M
= T dT + dH + H dT + dH
∂T H ∂H T ∂T H ∂H T

∂S ∂M ∂S ∂M
= T +H dT + T +H dH
∂T H ∂T H ∂H T ∂H T
0 0
= 0 for all paths ⇒ U = 0
8.044 L35B6

∂S ∂M ∂S ∂S
T +H = T +H
∂T H ∂T H ∂T
H ∂H
T
S (η)(−η/T ) S (η)(η/H)
= −ηS (η) + ηS (η)
= 0
A similar expansion shows that the other term is also

zero.
8.044 L35B7
    � �
d
/Q

1 ∂M
CH ≡

= 
dU −HdM

= −H
� ��
dT
H dT 0
H ∂T H
= N kJη 2BJ (η)
8.044 L35B8
MAGNETIZATION OF A QUANTUM PARAMAGNET
0.8
0.6
0.4
0.2
1 2 3 4
0.8
0.6
0.4
yc
0.2
1 x 2 3 4
8.044 L35B9
SUSCEPTIBILITY OF A QUANTUM PARAMAGNET
0.4
0.3
0.2
0.1
1 2 3 4
1.4
1.2
0.8
0.6
0.4
0.2
1 2 3 4
8.044 L35B10
HEAT CAPACITY OF A QUANTUM PARAMAGNET
0.8
0.6
0.4
0.2
1 2 3 4
0.8
0.6
0.4
0.2
1 2 3 4
8.044 L35B11
ENTROPY OF A QUANTUM PARAMAGNET
1.5
0.5
1 2 3 4
1.5
0.5
1 2 3 4
8.044 L35B12
MIT OpenCourseWare
http://ocw.mit.edu

Spring 2008
Review Paramagnet: a collection of non-interacting
magnetic moments
Quantum Mechanics
= gJ µB J
µ
⇒ 2J + 1 levels
The topic was saved until the last because it is subtle.
First J = 1/2, Microcanonical: S(M ) → M (T, H)
8.044 L36B1
Second J = 1/2, Canonical: p(mJ ) →< µ >→ M (T, H)
M
µN
SATURATION
CURIE LAW: M ∝ 1 / T
Third J > 1/2, Canonical: < µ >→ M (T, H)
Z = Z(η = gµB H/kT )
8.044 L36B2
So what’s the problem?
< H >S. M. =
U Thermo
·M
= U assembly + (−H )position (for d
/W = HdM )
 
1  ∂Z  1 dZ
 
∂η  √
< H >= − =−  = −HM
Z ∂β H Z dη

∂β H

M/gµB gµ H
B
8.044 L36B3
Entropy of a Quantum Paramagnet
• When is −kT ln Z = F ?
• How is a paramagnet like a sponge?
8.044 L36B4
HIGH AND LOW TEMPERATURE BEHAVIOR
OF A QUANTUM PARAMAGNET
kT >> g µBH
kT << g µBH
g µBH
ENERGY LEVELS ALMOST MOMENT ALMOST SATURATED,

EQUALLY POPULATED, ENERGY GAP BEHAVIOR
CURIE LAW BEHAVIOR
8.044 L36B5
S(kT gµB H) → N k ln(1) = 0
S(kT gµB H) → N k ln(2J + 1)
J

Z1 = (eη )m η ≡ gµB H/kT
m=−J
Try
−kT ln Z = F =
U −T S ⇒ S = k ln Z = N k ln Z1
0
8.044 L36B6
S(kT gµB H) → N k ln(1) = 0
S(kT gµB H) → N k ln(2J + 1) N k ln(2J + 1) O.K.
J

m=−J
Try
U −T S ⇒ S = k ln Z = N k ln Z1
−kT ln Z = F =
0
8.044 L36B6
S(kT gµB H) → N k ln(1) = 0 N kJ(gµB H/kT ) wrong!
S(kT gµB H) → N k ln(2J + 1) N k ln(2J + 1) O.K.
J

m=−J
Try
−kT ln Z U −T S ⇒ S = k ln Z = N
= F =
ln Z1
k
0
wrong ⇐ wrong
8.044 L36B6
In the derivation of the canonical ensemble we found
−kT ln Z =< E1 > −T S1 where < E1 >=< H({p, q}) >
Then we set < E1 >= U . But in the paramagnet
< E1 >= U − HM , thus
−kT ln Z = U − HM − T S = G(T, H) for our model.
⇒ S = k ln Z − HM/T = N k ln Z1(η) − N kJ ηBJ (η)
8.044 L36B7
ADIABATIC DEMAGNETIZATION
(MAGNETIC COOLING)
INITIAL COOLING
STAGE
T initial
THERMAL LINK 1
H = Hinitial
1
1 SM ~ 0
Stotal
1
SS high
⊗
MAGNET
1
TS = T initial
SAMPLE
kT << g µBH
8.044 L36B8
ADIABATIC DEMAGNETIZATION
(MAGNETIC COOLING)
INITIAL COOLING
STAGE
T initial
H~0
2J+1 ⊗ SM ~ Nk ln(2J+1)
Stotal
SS low
kT >> g µBH
SAMPLE
TS << T initial
8.044 L36B9
Electronic Example, CMN
Cerium Mangesium Nitrate
2Ce(NO3)3 · 3Mg(NO3)2 · 24H2O
Ce+++ J=5/2 Tordering ∼ 1.9 mK
Cools 3He and samples therein to ∼ 2 mK.
8.044 L36B10
Nuclear Example, Cu
Metallic Copper
Cu I=3/2 Tordering ∼ 1 µK
Cools Cu electrons and lattice to ∼ 10 µK.
8.044 L36B11

Stat Mech Notes

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Stat Mech Notes

Transféré par

Droits d'auteur :

Formats disponibles

MIT OpenCourseWare

8.044 Statistical Physics I

Phase T=0 Quantum gases

Language: Microcanonical ensemble

Probability 1st Law Transport

Random variable (ignorance and/or QM)

Probability density: p(x) ↔ px(ζ)

PROB(ζ ≤ x < ζ + dζ) = px(ζ)dζ

Either px(ζ) or Px(ζ) completely speciﬁes the RV x.

most of the time ( when hot)

PROB/dΩ = (1/π) cos(θ)

8.044 Statistical Physics I

A Hole Atom escapes after nth wall

Called a geometric or a Bose-Einstein density

p(x) = e−t/τ δ(x)+(1−e−t/τ ) f (x)

< x >= x0 < t >= τ

Events are statistically independent

0.1 <n> = 10 0.06

8.044 Statistical Physics I

Px,y (ζ, η) ≡ PROB(x ≤ ζ and y ≤ η)

= px,y (ζ , η )dζ dη

Conditional probability density

px(ζ |y)dζ ≡ prob.(ζ < x ≤ ζ + dζ given that η = y)

px,y (ζ, y) p(x, y)

x and y are statistically independent if p(x, y) = p(x)p(y)

dL → 0 ⇒ p(0) p(1) p(n > 1)

8.044 Statistical Physics I

• S.I. events r(x)

p(n = 0; L) = e−rL 1st element of p(n; L)

p(n; L + ∆L) ≈ p(n; L) p(0;

p(n; L + ∆L) − p(n; L) d p(n; L)

Let γ ≡ rL + p(n; γ) = p(n − 1; γ)

1st order, linear DE already known

DE: γ is the variable and n is an index

a zero mean gaussian with variance σ 2

a Gaussian with mean ρv1 and variance σ 2(1 − ρ2)

As ρ goes from 1 → 0, the mean goes from v1 → 0

and the variance goes from 0 → σ 2.

amplitude and phase drift

8.044 Statistical Physics I

Given: px(ζ) and f (x)

A. Sketch f (x). Find where f (x) < η

B. Integrate to ﬁnd Pf (η).

C. Diﬀerentiate to ﬁnd pf (η).

db(y) da(y) b(y) ∂g(y, x)

pλ(η) = −(−c/η 2) pν (c/η)

Let λ0 ≡ c/ν0, then

As (λ/λ0)→∞ 1 → (1+(λ/λ1 )−1−1) → (λ/λ0)

8.044 Statistical Physics I

x and y are statistically independent

p(θ) = 2 sin θ cos θ

η=0 ⇒ X= γ 2/2σ 2; η = π/2 ⇒ X= ∞

8.044 Statistical Physics I

Consider n RVs xi and let s ≡

If the RVs are statistically independent, then

• Both continuous and discrete RVs could be present

• True for any n

• Even if one RV dominates the sum

2) The variances are ﬁnite (=

Application to the Jointly Gaussian RVs in Section 2

The mathematical operation is called “convolution”.

q(z) = (z/a)m exp(−z/a) z

a function of the same class

ν1 −ν0 about 10 KHz