The canonical ensemble I

1
Equilibrium between a system and a heat reservoir

2
How 𝑆 𝑁, 𝑉, 𝐸 = 𝑘 ln Ω 𝑁, 𝑉, 𝐸 ?

Two noninteracting physical systems, 𝐴1 and 𝐴2 , are separately in equilibrium with 𝐴1 in

macrostate 𝑁1 , 𝑉1 , 𝐸1 and having Ω1 = Ω1 𝑁1 , 𝑉1 , 𝐸1 microstates, while 𝐴2 in macrostate
𝑁2 , 𝑉2 , 𝐸2 and having Ω2 = Ω2 𝑁2 , 𝑉2 , 𝐸2 microstates.

Now, bring 𝐴1 and 𝐴2 into thermal contact with each other, allowing an exchange of energy
between the two – e.g., by sliding in a conducting wall and removing the impervious one.
(𝐴1 and 𝐴2 are still separated by impenetrable, rigid wall, so that 𝑁1 and 𝑁2 , and 𝑉1 and 𝑉2
remain fixed.)

Neglecting the energy of interaction between 𝐴1 and 𝐴2 , if any, the total energy of the closed
and isolated composite system 𝐴 0 = 𝐴1 + 𝐴2 ,
0
𝐸 = 𝐸1 + 𝐸2 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡

3
How 𝑆 𝑁, 𝑉, 𝐸 = 𝑘 ln Ω 𝑁, 𝑉, 𝐸 ? (cont.)

The number of microstates of the composite system in equilibrium is

0 0
Ω = ෍ Ω1 𝐸 Ω2 (𝐸 − 𝐸)
𝐸
0 0
=Ω (𝐸 , 𝐸 ∗ ) ≈ Ω1 𝐸 ∗ Ω2 (𝐸 0
− 𝐸∗ )
where 𝐸 ∗ satisfies
𝜕Ω1 𝐸1 𝜕Ω2 𝐸2 𝜕 ln Ω1 𝐸1 𝜕 ln Ω2 𝐸2
Ω2 𝐸2 𝑑𝐸1 + Ω1 𝐸1 𝑑𝐸2 = 0, 𝑜𝑟 =
𝜕𝐸1 𝜕𝐸2 𝜕𝐸1 𝜕𝐸2
since 𝐸1 = 𝐸, 𝐸2 = 𝐸 0 − 𝐸1 = 𝐸 0 − 𝐸, and 𝑑𝐸1 = 𝑑𝐸, 𝑑𝐸2 = −𝑑𝐸. Inspired by the
zeroth law of thermodynamics and required by dimensional consistency,
𝜕 ln Ω 1
= ≡𝛽
𝜕𝐸 𝑁,𝑉 𝑘𝑇

4
The canonical distribution

Consider a closed and isolated composite system, 𝐴 0 = 𝐴 + 𝐴′ , made up of a system 𝐴

immersed in a very large heat reservoir 𝐴′ , with constant total energy 𝐸 0 . Here, 𝐴 can
only exchange energy with 𝐴′ .

If 𝐴 happens to be in a macrostate characterised by the energy value 𝐸𝑟 , then 𝐴′ would be in

a macrostate characterised by the energy value
0 ′ ′
𝐸𝑟 𝐸 − 𝐸 𝑟 𝐸𝑟
𝐸𝑟′ = 𝐸 0 − 𝐸𝑟 , 𝑠. 𝑡. 0 = 0
= 1 − 0
≪1
𝐸 𝐸 𝐸
Let Ω′ = Ω′ 𝐸𝑟′ be the number of distinct microstates of 𝐴′ . Then,
𝜕
ln Ω 𝐸𝑟 = ln Ω (𝐸 − 𝐸𝑟 ) = ln Ω (𝐸 ) − 𝐸𝑟 ′ ln Ω′ 𝐸𝑟′ ቤ
′ ′ ′ 0 ′ 0
+⋯
𝜕𝐸𝑟 𝐸 ′ =𝐸 0
𝑟
≈ 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 − 𝛽 ′ 𝐸𝑟
5
The canonical distribution (cont.)

On attaining a state of mutual equilibrium, 𝐴 and 𝐴′ would have a common 𝑇 and

′
1
𝛽 =𝛽=
𝑘𝑇
So,
Ω′ 𝐸𝑟′ ∝ exp −𝛽𝐸𝑟
And, the probability of 𝐴 assuming the energy value 𝐸𝑟 = 𝐸 0 − 𝐸𝑟′ ,
𝑃𝑟 ∝ Ω′ 𝐸𝑟′
Therefore,
exp −𝛽𝐸𝑟
𝑃𝑟 ∝ exp −𝛽𝐸𝑟 , 𝑃𝑟 =
σ𝑟 exp −𝛽𝐸𝑟
This is the canonical distribution. Here, the summation in the denominator goes over all
microstates accessible to 𝐴.
6
The classical ideal gas

7
The 𝑁-particle partition function

Consider a classical ideal gas of 𝑁 noninteracting indistinguishable identical monatomic

molecules, each of mass 𝑚, confined to a space of finite volume 𝑉, and in equilibrium at
thermodynamic temperature 𝑇. The system has Hamiltonian,
3𝑁 𝑁
1 2 1
𝐻 = 𝐻 𝑝1 , ⋯ , 𝑝3𝑁 = ෍ 𝑝𝑖 = ෍ 𝒑2𝑖
2𝑚 2𝑚
𝑖=1 𝑖=1
And, hence, 𝑁-particle partition function,
1 𝑑3𝑁 𝑞𝑑 3𝑁 𝑝 1 𝑉𝑁 3𝑁
𝑄𝑁 𝑉, 𝑇 = න exp −𝛽𝐻 3𝑁
= 3𝑁
නexp −𝛽𝐻 𝑑 𝑝
𝑁! ℎ 𝑁! ℎ
where the integration is over the whole phase space. Clearly,
𝑁 𝑁
𝛽 𝛽 2
exp −𝛽𝐻 = exp − ෍ 𝒑2𝑖 = ෑ exp − 𝒑𝑖
2𝑚 2𝑚
𝑖=1 𝑖=1
8
The 𝑁-particle partition function (cont.)

So,
∞ 𝑁
1 𝑉𝑁 3𝑁
1 𝑉 𝑁
𝛽 2 2
𝑄𝑁 𝑉, 𝑇 = 3𝑁
නexp −𝛽𝐻 𝑑 𝑝 = 3𝑁
න exp − 𝑝 4𝜋𝑝 𝑑𝑝
𝑁! ℎ 𝑁! ℎ 0 2𝑚
where
∞ 3Τ2 3Τ2
𝛽 2 2
𝜋 2𝑚 2𝜋𝑚 3Τ2
න exp − 𝑝 4𝜋𝑝 𝑑𝑝 = 4𝜋 × = = 2𝜋𝑚𝑘𝑇
0 2𝑚 4 𝛽 𝛽
And,
𝑁
1 𝑉𝑁 3𝑁Τ2
1 𝑉 3Τ2
1 𝑁
𝑄𝑁 𝑉, 𝑇 = 2𝜋𝑚𝑘𝑇 = 2𝜋𝑚𝑘𝑇 = 𝑄1 𝑉, 𝑇
𝑁! ℎ3𝑁 𝑁! ℎ3 𝑁!
with single-particle partition function,
𝛽 2 𝑑3 𝑞𝑑3 𝑝 𝑉 ∞ 𝛽 2 2
𝑉 3Τ2
𝑄1 𝑉, 𝑇 ≡ න exp − 𝑝 3
= 3 න exp − 𝑝 4𝜋𝑝 𝑑𝑝 = 3 2𝜋𝑚𝑘𝑇
2𝑚 ℎ ℎ 0 2𝑚 ℎ 9
Density of states to single-particle partition function

Recall that the single-particle density of state,

2𝜋𝑉
𝑔1 𝜀 = 3 2𝑚 3Τ2 𝜀
ℎ
We note that
∞ ∞
2𝜋𝑉
න exp −𝛽𝜀 𝑔1 𝜀 𝑑𝜀 = 3 2𝑚 3Τ2 න 𝜀 exp −𝛽𝜀 𝑑𝜀
0 ℎ 0
2𝜋𝑉 3Τ2
𝜋 𝑉 3Τ2
= 3 2𝑚 3Τ2
= 3 2𝜋𝑚𝑘𝑇 = 𝑄1 𝑉, 𝑇
ℎ 2𝛽 ℎ

10
Density of states to 𝑁-particle partition function

Recall that the 𝑁-particle density of state,

𝑁
1 1 𝑉 3Τ2
𝑔𝑁 𝐸 = 2𝜋𝑚 𝐸 3𝑁Τ2−1
𝑁! 3𝑁 ℎ 3
− 1 !
2
We again note that
∞ 𝑁 ∞
1 1 𝑉 3Τ2 3𝑁Τ2−1
න exp −𝛽𝐸 𝑔𝑁 𝐸 𝑑𝐸 = 2𝜋𝑚 න 𝐸 exp −𝛽𝐸 𝑑𝐸
0 𝑁! 3𝑁 − 1 ! ℎ 3
0
2
𝑁
1 𝑉 3Τ2
= 3
2𝜋𝑚𝑘𝑇 = 𝑄𝑁 𝑉, 𝑇
𝑁! ℎ

11
Density of states to 𝑁-particle partition function (cont.)

We also note that

𝜕 3 1
𝑔𝑁 𝐸 exp −𝛽𝐸 = 𝑁−1 − 𝛽 𝑔𝑁 𝐸 exp −𝛽𝐸 = 0
𝜕𝐸 2 𝐸
2
𝜕2 3 1 3 1
2
𝑔𝑁 𝐸 exp −𝛽𝐸 = − 𝑁−1 2+ 𝑁−1 −𝛽 𝑔𝑁 𝐸 exp −𝛽𝐸
𝜕𝐸 2 𝐸 2 𝐸
yields 3 ∗
1 3
𝐸 = 𝑁−1 ≈ 𝑁𝑘𝑇 = 𝑈
2 𝛽 2
∗ ∗
1 3 1
𝑔𝑁 𝐸 exp −𝛽𝐸 ≈ 𝑔𝑁 𝐸 exp −𝛽𝐸 exp − 𝑁 − 1 ∗2 𝐸 − 𝐸 ∗ 2
2 2 𝐸
and hence −1
3 ∗2
2
Δ𝐸 = 𝑁−1 𝐸 ≈ 𝑈
2 3𝑁
12
Density of states to 𝑁-particle partition function (cont.)

13
𝑁-particle partition function to density of states
∞
𝑄𝑁 𝑉, 𝑇 = න 𝑔𝑁 𝐸 exp −𝛽𝐸 𝑑𝐸 ≡ 𝑄𝑁 𝛽
0
The 𝑁-particle partition function is the Laplace transform of the 𝑁-particle density of states.
It follows that the 𝑁-particle density of states is the inverse Laplace transform of the 𝑁-
particle partition function:
𝛽′ +𝑖∞
1
𝑔𝑁 𝐸 =
ณ න 𝑄𝑁 𝛽 exp 𝛽𝐸 𝑑𝛽
𝐵𝑟𝑜𝑚𝑤𝑖𝑐ℎ
2𝜋𝑖 𝛽′ −𝑖∞
𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑙
With 1 𝑉 𝑁
1 𝛽′ +𝑖∞
exp 𝛽𝐸
3Τ2
3Τ2 𝑁 𝑔𝑁 𝐸 = 3
2𝜋𝑚 න 3𝑁 Τ2
𝑑𝛽
1 𝑉 2𝜋𝑚 𝑁! ℎ 2𝜋𝑖 𝛽′ −𝑖∞ 𝛽
𝑄𝑁 𝛽 = 𝑁
𝑁! ℎ3 𝛽 1 𝑉 3Τ2
𝐸 3𝑁Τ2−1
= 2𝜋𝑚
we have 𝑁! ℎ 3 3𝑁
2 −1 ! 14
𝑁-particle partition function to density of states (cont.)
′
1 𝛽 +𝑖∞
න 𝑄𝑁 𝛽 exp 𝛽𝐸 𝑑𝛽 = ෍ 𝑟𝑒𝑠𝑖𝑑𝑢𝑒𝑠 𝑜𝑓 𝑄𝑁 𝛽 exp 𝛽𝐸 𝑖𝑛𝑐𝑙𝑢𝑑𝑒𝑑 𝑓𝑜𝑟 Re 𝛽 < 𝛽 ′
2𝜋𝑖 𝛽′ −𝑖∞
In this case,
𝛽′ +𝑖∞
1 exp 𝛽𝐸 exp 𝛽𝐸 ′
න 𝑑𝛽 = ෍ 𝑟𝑒𝑠𝑖𝑑𝑢𝑒𝑠 𝑜𝑓 𝑖𝑛𝑐𝑙𝑢𝑑𝑒𝑑 𝑓𝑜𝑟 Re 𝛽 < 𝛽
2𝜋𝑖 𝛽′ −𝑖∞ 𝛽 3𝑁Τ2 𝛽 3𝑁Τ2
The residue of a complex-valued function 𝑓 = 𝑓 𝑧 at 𝑧 = 𝑧0 is the coefficient 𝑎−1 of
𝑧 − 𝑧0 −1 in the Laurent expansion of 𝑓:
∞
𝑛
𝑓 𝑧 = ෍ 𝑎𝑛 𝑧 − 𝑧0
𝑛=−∞

15
𝑁-particle partition function to density of states (cont.)

For example,
exp 𝑎𝑧 exp 𝑎𝑧 exp 𝑎𝑧 1 1
2 2
= = −
𝑧 −𝑏 𝑧+𝑏 𝑧−𝑏 2𝑏 𝑧−𝑏 𝑧+𝑏
exp 𝑎𝑏 exp 𝑎 𝑧 − 𝑏 exp −𝑎𝑏 exp 𝑎 𝑧 + 𝑏
= −
2𝑏 𝑧−𝑏 2𝑏 𝑧+𝑏
∞ 𝑛 ∞ 𝑛
exp 𝑎𝑏 𝑎 exp −𝑎𝑏 𝑎
= ෍ 𝑧 − 𝑏 𝑛−1 − ෍ 𝑧+𝑏 𝑛−1
2𝑏 𝑛! 2𝑏 𝑛!
𝑛=0 𝑛=0
And,
𝑏 +𝑖∞
1 exp 𝑎𝑧 exp 𝑎𝑧
න 2 2
𝑑𝑧 = ෍ 𝑟𝑒𝑠𝑖𝑑𝑢𝑒𝑠 𝑜𝑓 2 2
𝑖𝑛𝑐𝑙𝑢𝑑𝑒𝑑 𝑓𝑜𝑟 Re 𝑧 < 𝑏
2𝜋𝑖 𝑏 −𝑖∞ 𝑧 − 𝑏 𝑧 −𝑏
exp 𝑎𝑏 exp −𝑎𝑏 1
= − = sinh 𝑎𝑏
2𝑏 2𝑏 𝑏 16
𝑁-particle partition function to density of states (cont.)

Indeed, the Laplace transform,

∞
1 1
න sinh 𝑎𝑏 exp −𝑎𝑧 𝑑𝑎 = 2 2
0 𝑏 𝑧 − 𝑏
In our case,
∞
exp 𝛽𝐸 𝐸 𝑛 𝑛−3𝑁Τ2 𝐸 3𝑁Τ2−1 1
3𝑁 Τ2
=෍ 𝛽 = ⋯+ +⋯
𝛽 𝑛! 3𝑁Τ2 − 1 ! 𝛽
𝑛=0
Therefore,
1 𝑖∞ exp 𝛽𝐸 exp 𝛽𝐸 𝐸 3𝑁Τ2−1
න 3𝑁 Τ2
𝑑𝛽 = ෍ 𝑟𝑒𝑠𝑖𝑑𝑢𝑒𝑠 𝑜𝑓 3𝑁Τ2 𝑖𝑛𝑐𝑙𝑢𝑑𝑒𝑑 𝑓𝑜𝑟 Re 𝛽 < 0 =
2𝜋𝑖 −𝑖∞ 𝛽 𝛽 3𝑁Τ2 − 1 !

17
Physical significance of the 𝑁-particle partition function

18
Thermodynamic relations

Recall the first law of thermodynamics,

𝑑𝑈 = 𝑇𝑑𝑆 − 𝑝𝑑𝑉 + 𝜇𝑑𝑁
With Helmholtz free energy, 𝐴 ≡ 𝑈 − 𝑇𝑆, we have
𝑑𝐴 = 𝑑𝑈 − 𝑇𝑑𝑆 − 𝑆𝑑𝑇 = −𝑆𝑑𝑇 − 𝑝𝑑𝑉 + 𝜇𝑑𝑁
That is, Helmholtz free energy 𝐴 = 𝐴 𝑁, 𝑉, 𝑇 , just like the 𝑁-particle partition function.
Suppose
2 3Τ2
𝑁 ℎ
𝐴 𝑁, 𝑉, 𝑇 = −𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇 ≈ −𝑁𝑘𝑇 + 𝑁𝑘𝑇 ln
𝑉 2𝜋𝑚𝑘𝑇
Then, pressure 𝑝 is given by
𝜕𝐴 𝑁𝑘𝑇
𝑝=− =
𝜕𝑉 𝑇,𝑁
𝑉
19
Thermodynamic relations (cont.)

Chemical potential 𝜇 is given by

2 3Τ2
𝜕𝐴 𝑁 ℎ
𝜇= = −𝑘𝑇 + 𝑘𝑇 ln + 𝑘𝑇
𝜕𝑁 𝑉,𝑇
𝑉 2𝜋𝑚𝑘𝑇
2 3Τ2
𝑁 ℎ
= 𝑘𝑇 ln
𝑉 2𝜋𝑚𝑘𝑇
Entropy 𝑆 is given by
3Τ2
𝜕𝐴 5 𝑉 2𝜋𝑚𝑘𝑇
𝑆=− = 𝑁𝑘 + 𝑁𝑘 ln
𝜕𝑇 𝑁,𝑉
2 𝑁 ℎ2
In terms of 𝑄𝑁 𝑉, 𝑇 ,
𝜕𝐴 𝜕
𝑆=− = 𝑘 ln 𝑄𝑁 𝑉, 𝑇 + 𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇
𝜕𝑇 𝑁,𝑉
𝜕𝑇 20
Thermodynamic relations (cont.)

And, 𝐴 = 𝑈 − 𝑇𝑆 yields
𝜕
2
−𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇 = 𝑈 − 𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇 − 𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇
𝜕𝑇
or internal energy,
𝜕 2
𝜕𝑇 𝜕 𝜕
𝑈 = 𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇 = − ln 𝑄𝑁 𝑉, 𝑇 = − ln 𝑄𝑁 𝑉, 𝛽
𝜕𝑇 𝜕𝛽 𝜕𝑇 𝜕𝛽
Indeed, 2
𝜕 3
𝑈 = 𝑘𝑇 ln 𝑄𝑁 𝑉, 𝑇 = 𝑁𝑘𝑇
𝜕𝑇 2
And,
𝜕 1 1 𝑑 3𝑁 𝑞𝑑3𝑁 𝑝
𝑈=− ln 𝑄𝑁 𝑉, 𝛽 = න 𝐻 exp −𝛽𝐻
𝜕𝛽 𝑄𝑁 𝑉, 𝛽 𝑁! ℎ3𝑁
−1
= නexp −𝛽𝐻 𝑑3𝑁 𝑞𝑑 3𝑁 𝑝 න𝐻 exp −𝛽𝐻 𝑑 3𝑁 𝑞𝑑 3𝑁 𝑝
21
Thermodynamic relations (cont.)

In terms of
1
𝑃𝑟 = exp −𝛽𝐸𝑟 , 𝑄𝑁 = ෍ exp −𝛽𝐸𝑟
𝑄𝑁
𝑟
We have
𝜕 1
𝑈=− ln 𝑄𝑁 = ෍ 𝐸𝑟 exp −𝛽𝐸𝑟 = ෍ 𝑃𝑟 𝐸𝑟
𝜕𝛽 𝑄𝑁
𝑟 𝑟
and
𝜕 𝜕 𝑘
𝑆 = 𝑘 ln 𝑄𝑁 + 𝑘𝑇 ln 𝑄𝑁 = 𝑘 ln 𝑄𝑁 − 𝑘𝛽 ln 𝑄𝑁 = 𝑘 ln 𝑄𝑁 + ෍ 𝛽𝐸𝑟 exp −𝛽𝐸𝑟
𝜕𝑇 𝜕𝛽 𝑄𝑁
𝑟

= 𝑘 ln 𝑄𝑁 + 𝑘 ෍ 𝑃𝑟 𝛽𝐸𝑟 = 𝑘 ෍ 𝑃𝑟 ln 𝑄𝑁 + 𝛽𝐸𝑟 = −𝑘 ෍ 𝑃𝑟 ln 𝑃𝑟
𝑟 𝑟 𝑟
22
Density of states
𝑁
1 1 𝑉 3Τ2
𝑔𝑁 𝐸 = 2𝜋𝑚 𝐸 3𝑁Τ2−1
𝑁! 3𝑁 3
−1 ! ℎ
2
Note that
3𝑁 3𝑁 3𝑁 𝑉 3Τ2
3𝑁
ln 𝑔𝑁 𝐸 ≈ −𝑁 ln 𝑁 + 𝑁 − ln + + 𝑁 ln 3 2𝜋𝑚 + ln 𝐸
2 2 2 ℎ 2
Τ
5 𝑉 4𝜋𝑚𝐸 3 2
= 𝑁 + 𝑁 ln
2 𝑁 3ℎ2 𝑁
Compare with when 𝐸 = 𝑈 = 3𝑁𝑘𝑇Τ2. And,
3Τ2 𝜕 3𝑁 1
5 𝑉 2𝜋𝑚𝑘𝑇 ln 𝑔𝑁 𝐸 = =
𝑆 = 𝑁𝑘 + 𝑁𝑘 ln 𝜕𝐸 2𝐸 𝑘𝑇
2 𝑁 ℎ2
when 𝐸 = 𝑈 = 3𝑁𝑘𝑇Τ2.
we have
𝑘 ln 𝑔𝑁 𝐸 = 𝑆 23
Energy fluctuations in the canonical ensemble
– correspondence with the microcanonical ensemble

24
Internal energy 𝑈 and variance in 𝐸

1
𝑈 = 𝐸 ≡ ෍ 𝑃𝑟 𝐸𝑟 = ෍ 𝐸𝑟 exp −𝛽𝐸𝑟 , 𝑄𝑁 = ෍ exp −𝛽𝐸𝑟
𝑄𝑁
𝑟 𝑟 𝑟
It follows that
𝜕𝑈 1 𝜕𝑄𝑁 1
=− 2 ෍ 𝐸𝑟 exp −𝛽𝐸𝑟 − ෍ 𝐸𝑟2 exp −𝛽𝐸𝑟
𝜕𝛽 𝑄𝑁 𝜕𝛽 𝑄𝑁
𝑟 𝑟
2

= ෍ 𝑃𝑟 𝐸𝑟 − ෍ 𝑃𝑟 𝐸𝑟2 = − 𝐸 2 + 𝐸 2

𝑟 𝑟
with we have
𝜕𝑈 𝜕𝑇 𝜕𝑈 Δ2 ≡ 𝐸 2
− 𝐸 2 = 𝑘𝑇 2 𝑐𝑉
= = −𝑘𝑇 2 𝑐𝑉
𝜕𝛽 𝜕𝛽 𝜕𝑇 and
Δ 1
=𝒪
𝐸 𝑁 25
Density of states 𝑔𝑁 𝐸 and Boltzmann factor exp −𝛽𝐸

In general, the density of states 𝑔𝑁 𝐸 increases monotonically with 𝐸, while the Boltzmann
factor exp −𝛽𝐸 decreases monotonically with 𝐸. Consequently, 𝑔𝑁 𝐸 exp −𝛽𝐸 has an
extremum at say 𝐸 = 𝐸 ∗ , that satisfies
𝜕𝑔𝑁 1 𝜕𝑔𝑁 𝜕 ln 𝑔𝑁
− 𝛽𝑔𝑁 = 0, 𝑜𝑟 = =𝛽
𝜕𝐸 𝑔𝑁 𝜕𝐸 𝜕𝐸
That is,
𝜕 1
𝑘 ln 𝑔𝑁 ቤ =
𝜕𝐸 𝐸=𝐸 ∗
𝑇
Compare with we have
𝜕 1
𝑆 𝐸 ቤ = 𝐸∗ = 𝑈
𝜕𝐸 𝐸=𝑈
𝑇

26
Density of states 𝑔𝑁 𝐸 and Boltzmann factor exp −𝛽𝐸 (cont.)

Let
𝑓 𝐸 ≡ ln 𝑔𝑁 𝐸 exp −𝛽𝐸 = ln 𝑔𝑁 𝐸 − 𝛽𝐸
Then,
𝜕 𝜕 1 𝜕
𝑓 𝐸 = ln 𝑔𝑁 𝐸 − 𝛽 = 𝑆 𝐸 − 𝛽 = 0, 𝑤ℎ𝑒𝑛 𝐸 = 𝐸 ∗ = 𝑈
𝜕𝐸 𝜕𝐸 𝑘 𝜕𝐸
and
𝜕2 𝜕2 𝜕𝑇 𝜕 1 𝜕 1 ∗
2
𝑓 𝐸 = 2
ln 𝑔𝑁 𝐸 = 𝑆 𝐸 = − 2
, 𝑤ℎ𝑒𝑛 𝐸 = 𝐸 =𝑈
𝜕𝐸 𝜕𝐸 𝜕𝐸 𝜕𝑇 𝑘 𝜕𝐸 𝑐𝑉 𝑘𝑇
Therefore,
1 2
1 2
𝑓 𝐸 ≈𝑓 𝑈 − 𝐸 − 𝑈 = −𝛽 𝑈 − 𝑇𝑆 − 𝐸 − 𝑈
2𝑐𝑉 𝑘𝑇 2 2𝑐𝑉 𝑘𝑇 2
or 1 2
𝑔𝑁 𝐸 exp −𝛽𝐸 ≈ exp −𝛽 𝑈 − 𝑇𝑆 exp − 𝐸 − 𝑈
2𝑐𝑉 𝑘𝑇 2 27
𝑁-particle partition function

With
1 2
𝑔𝑁 𝐸 exp −𝛽𝐸 ≈ exp −𝛽 𝑈 − 𝑇𝑆 exp − 𝐸 − 𝑈
2𝑐𝑉 𝑘𝑇 2
the 𝑁-particle partition function,
∞ ∞
1 2
𝑄𝑁 𝛽 = න 𝑔𝑁 𝐸 exp −𝛽𝐸 𝑑𝐸 ≈ exp −𝛽 𝑈 − 𝑇𝑆 න exp − 2
𝐸−𝑈 𝑑𝐸
0 0 2𝑐𝑉 𝑘𝑇
≈ exp −𝛽 𝑈 − 𝑇𝑆 2𝜋𝑐𝑉 𝑘𝑇 2
Therefore,
1
−𝑘𝑇 ln 𝑄𝑁 𝛽 = 𝑈 − 𝑇𝑆 − 𝑘𝑇 ln 2𝜋𝑐𝑉 𝑘𝑇 2
2
= 𝐴 + 𝒪 ln 𝑁

28