Chap 16 Statistical Mechanics

Table of Contents 目錄內容

16.1 The Microscopic State of a System 系統的微觀狀態
16.2 Single-Molecular Partition Function for an Ideal Gas 理想氣體的分子配分函數
16.3 Translational Contributions to the Thermodynamics Properties of Ideal Gases 平移的
貢獻
16.4 Vibrational Contributions to the Thermodynamic Properties of Ideal Gases 振動的貢
獻
16.5 Rotational Contributions to the Thermodynamic Properties of Ideal Gases 轉動的貢
獻
16.6 Electronic Contributions to the Thermodynamic Properties of Ideal Gases 電子的貢獻
16.7 Thermodynamic Properties of Ideal Gases 理想氣體的熱力學性質
16.8 Direct Calculation of Equilibrium Constants for Reactions of Ideal Gases 直接計算平
衡常數
16.9 Equipartition 等配分性質
16.10 Ensembles 系综分布（微正則、正則、巨正則)
16.11 Nonideal Gases 非理想氣體
16.12 Heat Capacities of Solids 固體的熱容
16.13 Special Topic: Fluctuations of Thermodynamic Quantities 最大機率分佈的計算

2006/5/05 1
 導論

物質的平衡性質,有兩種觀點: 微觀(Microscopic)的描述與巨觀(Macroscopic)的行為.
• 微觀的描述: 由量子力學(quantum mechanics) 提供個別分子的結構與彼此的交
互作用. 亦可由光譜的測量以及波函數的理論計算得知.
• 巨觀的行為: 由熱力學(thermodynamics)的系統狀態函數量測得到的平均性質.
統計力學(Statistical Mechanics) :
• 以統計的方法和原理,機率的概念及最大機率分佈(most probable distribution)
的方法, 就是統計力學的方法.
• 由分子的微觀運動性質,如速度,動量,位置,動能,位能,轉動慣量,振動頻率,化學鍵能
等. 用統計的方法來預測系統的平均熱力學行為.
• 因此統計力學是聯結巨觀性質與微觀性質的橋樑.
所謂統計熱力學(Statistical Thermodynamics)：是指從微觀的角度出發 ,用統計力
學的方法,處理系統的大量粒子集合.推得到系統的巨觀性質. 從19世紀的
Boltzmann(Austria), Maxwell(UK), Gibbs(USA) 發展的統計熱力學理論,可以從
單一分子的轉動慣量,振動頻率,化學鍵能等資料, 計算出理想氣體的熱力學性質.

2006/5/05 2
 系統的微觀狀態 (Microscopic State of a System)

• 機械性質: (Mechanical properties): 如P,V n,U,H 直接以統計力學

推得, 未涉及熵值,與溫度無關.
• 非機械性質: (Nonmechanical properties): 如T, S, A, G, μ, 與溫
度有關,不能直接以統計力學推得. 需要熱力學的參與
• 古典力學(Classical Mechanics)的描述: 對無結構的N個粒子(如原
子), 系統由粒子位置(qx,qy,qz)與動量(px,py,pz)的6xN個自由度構成.
• 量子力學(Quantum Mechanics): 對無結構且無交互作用的粒子以
總波函數 Ψ = Ψ1*Ψ2*Ψ3*…*ΨN with 3xN quantum no.
(nx,ny,nz)描述. 對有結構或有交互作用的粒子, 需要更多的量子數
來描述如rotational, vibrational, electronic的內自由度 (internal
degree of freedom, DOF). 一般而言, N molecules 的系統, 可由
Ψ = Ψ(n1)*Ψ(n2)*Ψ(n3)*…*Ψ(nN) 完全描述. (ni)代表i分子的全
部量子數. 由於粒子的不可分辨(indistinguishable),亦可將系統各
個狀態的粒子數形成之一組數字,來描述此特定的系統組態.

2006/5/05 3
 Configuration and weights

• Configuration(系統組態):
{n 0 , n1 , n 2 ,...}

n 0 + n1 + n 2 + ... = N
• Weights of configuration W:
W = N!
n 0 ! n1! n 2 !...

• 20 particles: {1,0,3,5,10,1} Ö W=931000000

• The number of ways for a configuration increased drastically
with tiny increase in particles numbers.

2006/5/05 4
 Microstates

1 Whereas a configuration
{5, 0, 0, . . . } can be
achieved in only one way,
a configuration {3, 2,
0, . . . } can be achieved in W{3,2,0,...}=10
the ten different ways
shown here, where the
tinted blocks represent
different molecules.

W{5,0,0,...}=1

2006/5/05 5
 系統分類

¾ 按照粒子統計系統的分類:分為可區分(distinguishable如晶體)與不可區
分(indistinguishable,如一般氣體) 粒子系統.
¾ 按照粒子間的相互作用的分類: 分為獨立系統(作用力可忽略, 如idea gas)
與相依系統 (作用力不可忽略) .
¾ 按照粒子系統遵循的分布規律的分類:
¾ 費米子體系(Fermi-Dirac): 遵循包立的不互容分布原理, 每個量子態
只能容納一個粒子的規律.故每一量子態非空即被獨占(例如金屬電子).
¾ 波色子體系(Bose-Einstein): 不遵循包立的不互容分布原理,每個量
子態容納的粒子數不限. (例如熱輻射光子).
¾ 馬斯威爾波茲曼體系(Maxwell-Boltzmann): 以總能量守恆,且總粒
子數守恆為邊界條件之分配.前面的兩種量子化統計分配的結果都可
在一定的條件下, 近似於Maxwell-Boltzmann 分布的統計.

2006/5/05 6
 Example

• Example 1 Putting balls in boxes

(a) In how many ways W can six distinguishable molecules be placed
in three different energy levels with three molecules in the first
level, two in the second, and one in the third, without any regard
for the energy required?
(b) Write down the different ways of distributing three balls (a, b, and
c) between two boxes.
Answer:
(a) W = 6! /(3! 2! 1!) = 60
(b) Three are two possibilities: three balls in one and none in the
other (W = 2) or two balls in one and one ball in the other (W = 6).
The eight different distributions are:
abc/-, ab/c, ac/b, bc/a, c/ab, b/ac, a/bc and -/abc.

2006/5/05 7
Distribution of Gas molecules in two boxes

3.10 (a) Equilibrium

gases system after a
hole is punched in the
diagram. (b) Highly
improbable state of
the gaseous system
after the hole has
been punched.

2006/5/05 8
 Entropy and Statistical Probability

• N = 4 個理想氣體分子，分佈在兩間氣室，可能有 5 種不同的組態分配
方式,分別表為(0,4), (1,3), (2,2), (3,1), (4,0)。其中任一種組態分配可能
出現的微組態數Wj為
N! N!
W J ( n1 ,n2 ,Lnk ) = =
n1!n2! L Π ni !
i =1..k
而全部微組態總數為(1+4+6+4+1)=16. 每個可能微組態出現的統計熱
力機率都相等,為 P=1/16= (½)4.
• 而且4個分子全部退回到A部的微組態可能性為1,機率為(½)4.
• 在上例中，<均勻分佈>這種巨觀態，相應的微觀態最多，熱力學機率最
大(P=6/16)，實際觀測到的可能性或機率最大。
• 對於1023個分子組成的巨觀系統來說，<均勻分佈>這種巨觀態的熱力學
機率與各種可能的巨觀態的熱力學機率的總和相比，此比值幾乎或實際上
為100%。

2006/5/05 9
 Stirling approximation

• If N is a large number (N >> 10) , N ! and lnN ! can be

represented in a simple way:
ln N ! ≈ N ln N - N
since N ! ≈ e-N NN = (N/e )N
lnN ! ≈ ln (N/e )N =N ln (N/e )
• An even better approach is to combine the use of logarithms
with Stirling approximation. For N<20, N! can be represented
as:
N
N   1 1 
N != 2πN   1+ + + K
 e   12N 288N 2


2006/5/05 10
 Weight of Configuration for large N

N! N !
ln W J = ln = ln = ln N ! − ln( n 0 ! n 1! n 2 !...)
Π ni ! n 0 ! n 1! n 2 !...
i =1 ..k

= ln N ! − (ln n 0 ! + ln n 1! + ln n 2 ! + ...)
= ln N ! − ∑i ln n i !
• With Stirling approximation:
ln N ! ≈ N ln N − N
lnWJ = (N lnN − N ) − ∑(ni lnni − ni )
i

= N lnN − ∑ni lnni

i

2006/5/05 11
 Entropy and the Statistical Probability

• 熵和其他熱力學性質一樣，也具有統計本性。獨立系統的熵值表示大量分
子的平均性質。平衡狀態在巨觀上有最大的熵值，而在微觀的平衡狀態則
是含有最多的統計機率。定位系統的微態數:一個由N 個可區分的粒子組成
的特定系統狀態，在量子化的 i 個能階上可以有k 種不同的組態分配方
式。在其中任一種組態分配 j下，可能出現的微組態數用Wj 表示:

W j = C N C N −n1C N −n1 −n2 L =

n1 n2 n3 N!
⋅
(N − n1 )!
⋅
(N − n1 − n2 )!
L
n1! (N − n1 )! n2! (N − n1 − n2 )! n3! (N − n1 − n2 − n3 )!
N!
=
n1! n2! L
• 得到的微組態數總和Ω為

k
 N!   N! 
Ω = ∑W j = ∑   = ∑  
j =1 j  n1! n2! L  j
 Π ni! 
 i 

2006/5/05 12
 Entropy and Statistical Probability

• 等機率假設:由統計熱力學觀點，對於 U, V 和 N 固定的某一巨觀系統，
任何一個可能出現的微觀狀態，都有相同的統計熱力機率，所以這假定又
稱為等機率原理。如果ㄧ個巨觀系統的微組態總數為 Ω=∑Wi ，則每一種
微觀狀態出現的微組態機率都相等，即 P=1/Ω。
• 例如:由可區分的獨立粒子組成的ㄧ個巨觀系統: N=4個理想氣體分子，分
佈在兩間氣室，可能有5種不同的組態分配方式,分別表為(0,4), (1,3),
(2,2), (3,1), (4,0)。其中任一種組態分配可能出現的微組態數Wj 為

而全部微組態總數為(1+4+6+4+1)=16. 每個可能微組態出現的統計熱
力機率都相等,為 P=1/16= (½)4.
N! N!
W J (n1, n2, L, nk ) = =
n1! n2! L Π ni !
i =1..k

2006/5/05 13
 Entropy and Statistical Probability

• 從Boltzmann 的假說，熱力學熵值是統計熱力機率S的度量。在特定溫度下物質相對於
絕對零度的的熵值可由 δqrev /T 積分而得： ：

S=k ln Ω 式中 k 表示Boltzmann Constant。當Ω=1時，S=0。

Ω'
• 狀態改變前後伴隨的微組態與其熵值的變化可表示為： ∆S = S'- S = k ln
Ω
• Ω‘/Ω等於改變前後微組態的比值。絕對零度(T=0) 的完整晶體 (Ω =1) 的熵值等於
零，可見熵值是有絕對值的，這就是絕對熵定律（熱力學第三定律)。

• 一莫耳理想氣體分子從兩間氣室重新分佈到單一氣室(Ω’=1)，過程前後的熵值改變量,
與反向的膨脹過程大小相等，但為負值
23
ΔS = k ln (½)6.022x10 = k ln (e4.174x1023)

= (1.381x10-23 J K-1) (-4.174x1023) = -5.76 J K-1

• 這個結果也印證了k = R/ NA.

2006/5/05 14
 6.1 波爾茲曼分布定律 (The Boltzmann Distribution)

¾ 基本假設: 對某一確定的(N,U,V) 巨觀狀態,其中任一微組態都有相同的出現

機率. postulate: equal A priori probabilities
¾ 對一個獨立的系統,有確定的 (N,U,V)巨觀狀態,內含 N 個可區分的粒子,分
布在可容納粒子數不限的能階 ei 上,其能量,狀態,及微組態的分佈方式為:
e1 e2 e3 e4 …… No. of microstate(微組態數)
n1 n2 n3 n4 …… W1 = N!/(n1! n2! n3! n4!) = N!/(Πi ni !)
n1’ n2’ n3’ n4’ …… W2 = N! /(Π ni’!)
n1〃 n2〃 n3〃 n4〃 …… W3 = N! /(Π ni〃!)
n1‴ n2‴ n3‴ n4‴ …… W4 = N! /(Π ni‴!)
:  N! 
Ω = ∑ 
系統分布的總組態數Ω是其微組態數總和: 
Ω = ∑jWj = W1 + W2 + W3 +W4 + … j Π n i!
 i 
2006/5/05 15
• The Boltzmann distribution applied to particles which can be
thought of as distinguishable from one another and for
which there is no restriction on the number of them which
can occupy a given energy state, for such particles, the
number of ways a specified set of population numbers (a
configuration) can be achieve is given by
N! N!
W J ( n1 ,n2 ,Lnk ) = =
n1!n2! L Π ni !
i =1..k
• This number is referred to as the statistical weight, W.
• A system in a given macroscopic state can be in any one of a
very large number of microscopic states that all have the
same macroscopic properties.

2006/5/05 16
系統狀態分布
¾ 若考慮能階εi的簡併數 gi (degeneracy):
¾ 若 gi =1,非簡併的總微狀態數
(nondegenerate, one particle at most in any cell)

 N! 
Ω = ∑ 
j
 Π ni ! 
 i 

2006/5/05 17
系統狀態分布
¾ 在gi ≠1(簡併能階) 的系統, 在能階εi的狀態有gi 個簡併度
(degeneracy),分佈有ni 個粒子, 即有 gi n i 種可能的分布.
定義所有可能的微狀態數分佈數Ω可以下式表示:
ni
N!  gi 
W j (n i ) = W Π gi =
ni
∏ g i = N ! ∏ 
ni


∏
i
n i ! i i n
 i !
¾ 對N 個可區分的粒子,總微狀態數Ωdis為
 g ni   g ni 
Ωdis = ∑W j = ∑N!  i  =N !  i
∑j  Π n ! 
j j
 Π n i!  
 i   i i 
¾ 對N 個不可區分的粒子且 gi >> ni ,修正的總微狀態數
 g ni 
Ω’ = Ωdis/N! = Σi (Π (gi ni/ni !)) Ω = ∑ i 
indis
i
 Π n i! 
 i 
2006/5/05 18
 Equilibrium configuration

The dominating configuration is the configuration

that has the largest weight in a isolated system.

The dominating configuration is what we actually observe.

All other configurations are regarded as fluctuation.

In the isolated system :

∑n = N
i
i
⇒
ni  gi  - (ε −ε ) kT
=  e i j

nj g 
∑ (n ε ) = E
i
i i
 j

2006/5/05 19
 最可能分布

• 獨立系統的熵值表示大量分子的平均性質。平衡狀態在巨
觀上有最大的熵值，而在微觀的平衡狀態則是含有最多的
統計機率。
• 最可能分布: 知道系統的分布,就可計算出微狀態數(number of
possible microstate).但在大量粒子時,分布相當多,無法一一
算出. 以系統中,出現機率最大的分布,就是微狀態數最多的分布
tmax,叫做最可能分布.

2006/5/05 20
 • There is a configuration
{ni }max with an overwhelming
probability W {ni }max .

2006/5/05 21
當粒子數很大時, 可以 ln tmax 代替 ln Ω at {ni*}max
• ln Ωdis ≈ ln tmax n *i
 gi 
t max,dis = N ! Π  * 
¾ for distinguishable: i
 n i! 
ln Ωdis ≈ ln tmax = ln N! + Σi [ ln (gini ) – ln (ni! ) ]
= (N ln N – N ) + Σi ni ln (gi ) – Σi ni ln (ni ) + Σi (ni )
  gi 
ln Ωdis = N ln N + ∑ ni ln  
i   ni  g  n*i
t max,indis = Π   i
¾ for indistinguishable: i
n ! 
*
i

ln Ωindis ≈ ln tmax = Σi [ ln (gini) – ln ni! ]

= Σi ni ln(gi) – Σi ni ln (ni ) + Σi (ni )
  gi 
ln Ωindis = ∑ ni ln  + N
i   ni 

2006/5/05 22
 Undetermined Multipliers:

ln tmax = ln Ω = f
以求極值的方法求 tmax ;
• Now d f = d ln Ω = Σi (∂ ln Ω/∂ ni ) dni = 0
(since not all dni are all independent) and ∑i
dni = 0
• 兩個限制條件 (subject to two constraints):
¾ Constant total number of particle:
∑n i
i =N

¾ Constant total energy :

∑ (ni εi ) = E
i

2006/5/05 23
 Lagrange’s method of undetermined multipliers

z=f (x,y) with c1= g (x,y), c2= h (x,y)

To find the maximum of z with constraints g and h, we

May use dz + a*dg + b*dh = 0

ni  gi  - (ε −ε ) kT
⇒ =  e i j

nj g 
 j

2006/5/05 24
 Undetermined Multipliers:

用 Lagrange 不定乘因子法:
df =d ln Ω + α dN – β dE = 0 (α,β: undetermined multipliers)
 ∂ ln Ω 
¾ dln Ω = ∑ i   dni = 0;
 ∂ni 
¾ dN =Σi dni = 0;

¾ dE =Σi (εi dni) = 0 ;

 ∂ ln Ω 
df = ∑   dni + α ∑i dni − β ∑i ε i dni = 0;
 ∂ni 
i

 ∂ ln Ω  
= ∑i   + α − βε i dni = 0;
 ∂ni  

or  ∂ ln Ω 
  + α − βε i = 0;
 ∂ni 

2006/5/05 25
 Justification

For each distinguishable ln Ωdis = N ln N + Σk [ nk ln (gk /nk) ]

 ∂ ln Ωdis

 ∂ (N ln N ) ∂
 = +
[∑ n k k ln (gk nk )]
 ∂ni ∂ni ∂ni

 ∂ ln Ωdis

 ∂ (N ln N ) ∂
 = −
[∑ n
k k ]
ln (nk )
Start with gi = 1,then  ∂ni  ∂ni ∂ni

∂ (N ln N )  ∂N   ∂ ln N   ∂ ln N
 1 ∂N 
=   ln N + N   = ln N + 1 = = 1
∂ni  ∂n N ∂n 
 ∂ni   ∂ni   i i 
1)
∂ (nk ln nk )  ∂ nk   ∂ (ln nk )   ∂ nk  
∑k ∂n = ∑k  ∂n  ln nk + nk   = ∑k  (ln nk + 1) = ln nk + 1
i  i   ∂ni   ∂ni  
∂ (ln nk ) 1  ∂ nk  ∂n k 1 if i =k
2) here =   = 
∂ni nk  ∂ni ∂n i  0 if i ≠k
3) ∂ ln Ωdis/∂ ni = - (ln ni +1) + (ln N + 1) = - ln (ni /N )

4) so – ln (ni /N ) + α – β εi = 0 Ö Pi = (ni /N ) = eα –β εi
2006/5/05 26
• with gi ≠ 1, then ln Ωdis = N ln N + Σk [ nk ln gk - nk ln nk ]
 ∂ ln Ωdis  ∂ (N ln N )  ∂ (nk ln nk )
  = − ∑k   + ln gi
 ∂ni  ∂ni  ∂ni 
1) ∂ (N ln N )  ∂N   ∂ ln N 
=   ln N + N   = ln N + 1
∂ni  ∂ni   ∂ni 
2) ∂ (nk ln nk )  ∂ nk   ∂ (ln nk )   ∂nk  
∑k ∂n = ∑k  ∂n  ln nk + nk   = ∑k  (ln nk + 1) = ln ni + 1
i  i   ∂ni   ∂ni  
here ∂ (ln nk ) 1  ∂ nk  a) if i ≠ k (∂ nk /∂ ni ) = 0
=  
∂ni nk  ∂ni  b) if i = k (∂ nk /∂ ni ) = 1
 ∂ ln Ωdis   ni 
3)   = (ln N + 1) − (ln ni + 1) + ln gi = − ln 
 ∂ni   N gi 

4) so – ln (ni /N gi) + α – β εi = 0 Ö Pi = (ni /N) = gi e α –β εi

2006/5/05 27
• (ni /N gi ) = e α –βεi

• ni /N = gi e α –βεi

• As N = Σi ni
= N Σ i gi e α–β εi

=Ne α Σi g i e –β εi

=Ne α q
• Here e -α = q = Σi g i eis the molecule partition function.
–βεi

ni gi
• Now the Boltzmann equation: = ( ) e -β εi
N q
a) At T →0, all particles in ground state, N = no with εo = 0
b) At T →∞, all states are equally populated with q = Σi gi
and e –β εi = 1.
2006/5/05 28
 Find the distribution with largest lnW

 ∂ lnW 
dlnW = ∑   dni = 0 ∑i dn i = 0 ∑i εidni = 0
i  ∂ni 
 ∂ lnW 
∑i  ∂n  dni + α ∑ dni − β ∑ εi dni
 i  i i

 ∂ lnW    ∂ lnW 
= ∑   + α − βεi  dnÖi  ∂ n  + α − βεi = 0
i  ∂ ni    i 

∂ lnW ∂(N ln N ) ∂(n j lnn j ) ∂ (N ln N )  ∂N  ∂N

= −∑ =   ln N + = ln N + 1
∂ni ∂ni j ∂ni ∂ni  ∂ni  ∂ni
∂ (n j ln n j )  ∂n j   ∂ ln n j   ∂n j 
∑n = N ∑j ∂n = ∑j  ∂n  ln n j + n j    = ∑  (ln n j + 1) = ln ni + 1
 ∂ni   j  ∂ni
i
i i  i  
∂ lnW ni ni ni
= −(lnn i +1) + (lnN +1) = −ln − ln + α − βε i = 0 Ö = e α − βε
∂n i N
i
N N
2006/5/05 29
 Ω for the Bosons 波色子體系

Configuration is {3,6,5,4}={n0.n1,n2,…}

18!
W J ( 3,6,5,4 ) =
3! 6! 5! 4! L
2 The 18 molecules shown here can be distributed into four
receptacles (distinguished by the three vertical lines) in 18!
different ways. However, 3! of the selections that put three
molecules in the first receptacle are equivalent, 6! that put
six molecules into the second receptacle are equivalent, and
so on. Hence the number of distinguishable arrangements is
18!/(3!6!5!4!).

2006/5/05 30
 Ω for the Bosons 波色子體系

• 設近獨立且不可分粒子組成的系統具有確定的粒子數N, 能量E和體積V﹐
若單粒子的第j 個(j=1﹐2﹐3﹐…)能級εj和對應該能級的量子態數 gj
(簡併度)分別表示。
• 對於玻色系統,由於粒子的不可分辨和每個能級態上佔據的粒子數不限 ,
則給定的 nj 個粒子分布在 gj 個量子態的方式數,等於從 (nj + gj -1)個元
素中選取 (gi -1)個元素的組合數。 (n J + g J -1) !
n J ! (g J -1) !
• gj : Number of cell, (gj -1) of wall, and there are nj identical
particles in the J th cell.
• 考慮各能級的結果﹐就得到對應粒子數分布{nj }的系統微觀狀態數Wj
(WJ : the number of ways ) (g - n -1) !
WBE = ∏ J J

J (g J -1) ! n J !
經過計算可得粒子數按能級的最可能分布為 :
 gJ +nj  g
ln
 n
 + ln A + βε j = 0 Ö
 n j = (ε -µ ) kT
j

 j  e j −1
2006/5/05 31
 Fermi-Dirac Distribution

• Statistical thermodynamics is used to describe the behavior of

large numbers of particles.
• In statistical mechanics, Fermi-Dirac statistics determines the
statistical distribution of fermions over the energy states for a
system in thermal equilibrium. In other words, it is a
probability of a given energy level to be occupied by a fermion.
Fermions are particles which are indistinguishable and obey
the Pauli exclusion principle, i.e., no more than one particle
may occupy the same quantum state at the same time.
• A collection of non-interacting fermions is called a Fermi gas.
F-D statistics was introduced in 1926 by Enrico Fermi and
Paul Dirac and applied in 1927 by Arnold Sommerfeld to
electrons in metals.

2006/5/05 32
 Fermi-Dirac distribution

• 對於費密系統﹐ni 個不可分辨的完全相同粒子分布在gi個狀態上(每
個態上至多只能有一個粒子)的可能方式數﹐就是從 gi個元素中選取
ni 個元素的組合數﹐應該等於 gi !
ni ! (gi - ni ) !
gi !
• 則對應粒子數分布{ni }的系統微觀狀態數 WFD =∏
i ( gi - ni ) ! ni !
若任一能級εi 上的粒子數ni 均遠小於該能級的量子態數gi ,
ni / gi <<1﹐即絕大多數量子態均未被佔據﹐則可以得到
1
= n ! ∏ gi
ni
WFD ≈ WBE
∏i i
i

• 這就過渡到了玻耳茲曼系統的微觀狀態數。利用固定N, E 條件經過
計算可得粒子數按能級的最可能分布為
gi
ni =
e (εi-µ ) kT ± 1
• 此式分母中的正負號分別對應於費密系統和玻色系統。

2006/5/05 33
 Fermi-Dirac Distribution

• For F-D statistics, the expected number of particles in states

with energy εi is
gi
ni =
• where: e(εi -µ ) kT + 1
ni is the number of particles in state i
εi is the energy of state i
gi is the degeneracy of state i (the number of states with
energy εi )
μ is the chemical potential. Sometimes the Fermi energy
EF is used instead, as a low-temperature
approximation.
k is Boltzmann's constant
T is absolute temperature

2006/5/05 34
 Fermi-Dirac Distribution

系統狀態分布
¾ 若考慮能階εi的簡併數gi (degeneracy):
¾ 若gi =1,非簡併的總微狀態數
(nondegenerate, one particle at most in any cell)
Ω = Σi N! / (Π ni !)
¾ 若gi =G (Total of G cell and ni of particles in ith cell).
G!
C G
=
n j ! (G - n )!
nj
j

G!
W=∏
J n j ! (G - n j ) !

ln W = ∑ [G ln G - G - n
j
j ln n j + n j - (G - n j )ln (G - n j ) + (G - n )]
j

 G - ni 
ln  + ln A + βε j = 0 ⇒ G
 ni  ni = (ε -µ ) kT
e j +1
2006/5/05 35
 Boltzmann distribution

ni N = ∑n j = Ne
N
= e α−βε i

j
α
∑j e − βε j

1 1
e =
α β= k: Boltzmann constant
∑j e − βε j
kT

ni e − βε i
=
N ∑ei
− βε i

2006/5/05 36
 The molecular partition function
(nondegenerate case)

ni e − βε i
pi ≡ =
N q
q= ∑e
i = states
− βεi

E Ψ gJ

E5 Ψ5 g5 = 1

E4 Ψ4 g4 = 1

E3 Ψ3 g3 = 1

E2 Ψ2 g2 = 1

E1 Ψ1 g1 = 1

2006/5/05 37
 The molecular partition function
(degenerate case)

nj gj e
∑gj e
− βε j
pj ≡ = q= − βε j

N q levels j

E Ψ gj

E5 Ψ5a, Ψ5b, Ψ5c, … g5

E4 Ψ4a, Ψ4b, Ψ4c, … g4

E3 Ψ3a, Ψ3b, Ψ3c, … g3

E2 Ψ2a, Ψ2b, Ψ2c , … g2

E1 Ψ1a, Ψ1b, Ψ1c , … g1

2006/5/05 38
 Evaluation of β

• To evaluate β, compare with U = (3/2)RT

dU : differential of internal energy in a reversible process
involving only PV-work
dU = dQrev – p dV = T dS – p dV
• For system with independent particle; like idea gas,
U = Σi ni εi
dU = Σi εi dni + Σi ni dεi

2006/5/05 39
 Heat does not change energy levels

• (a) When a system is

heated, the energy
levels are unchanged
but their populations
are changed.

2006/5/05 40
 Work changes energy levels

• (b) When work is done on a

system, the energy levels
themselves are changed.
The levels in this case are
the one-dimensional
particle-in-a-box energy
levels of Chapter 12: they
depend on the size of the
container and move apart
as its length is decreased.

2006/5/05 41
As the width of a container
is increased (going from (a)
to (b)), the energy levels
become closer together (as
1/L 2), and as a result more
are thermally accessible at
a given temperature.
Consequently, the entropy
of the system rises as the
container expands. As
before, this diagram is
schematic: the separation
of levels increases with
energy.

2006/5/05 42
Σiεi dni Σi ni dεi

For simplicity in making the essential point, we have shown the

energy levels as equally spaced; in fact, their separation increases
with energy.

2006/5/05 43
Let dS = Σi (εi dni /T ) ; and –dV = Σi (ni dεi /p) ;
Addition of the heat: Σiεi dni =T dS will change population dni
For reversible work: Σi ni dεi = -p dV will change volume thus
energy level dεi
A system from two subsystems, the entropy of combined
system is S = S1 + S2; and the microstates of two
independent subsystems is Ω = Ω1 Ω2
Since S = f(Ω1 Ω2) = S1 + S2 = f(Ω1) + f(Ω2)
f must be f(Ω) = a ln (Ω) + b = S here, the a, b as constants

2006/5/05 44
If we compare the difference only, then set b =0, and a =k =R/NA
Then S = k ln Ω
when maximizing Ω at constant N, U, V, one has to maximized Ω.
S = Qrev/T = k ln Ω with ln Ω = f(N1, N2, N3, …Ni, …)
dS = k d ln Ω
 ∂ ln Ω 
= k ∑i   dni
 ∂ni 
  ni 
= k ∑i − ln  dni
  N gi 
= k ∑i [ln (q ) + βε i ] dni [as ni /N = (gi /q ) e –β εi]

= k ln (q )∑i dni + k ∑i βε i dni

= kβ ∑i (εi dni )
dQrev
= kβ (dQrev ) = Ö β = 1/kT
T
2006/5/05 45
• Molecular partition function q = e-α:
 −εi 
q = ∑igi exp  
 kT 
• Boltzmann distribution:
ni 1  −εi 
= gi exp  
N q  kT 
• q roughly equals to the number of states available to the
system at a given temperature. (thermally available states )
• Limitations:
¾ (1) non-interacting particles
¾ (2) occupation << degeneracy. (ni << gi )

2006/5/05 46
 16.4 Calculation of thermo function for indep particles

• Thermodynamic properties from partition function q

Start from P, U, H, Cv, and Cp, as
N  −εi 
ni = gi exp  
q  kT 
N  −εi 
- p dV = Σi ni dεi = ∑i gi exp   dε i
q  kT 
at a fixed temperature,
 −εi 
q = ∑i gi exp  1  −εi 
 kT 
 dq = −
kT
∑i gi exp kT  dεi
N
-P dV = (− kT dq ) = NkT dln q
q
 ∂ ln q 
P = NkT  
 ∂ V T

2006/5/05 47
 Internal energy U

gi e −ε i kT N  −εi 
ni = N Pi = N = g exp  
∑i gi e q  kT 
−ε i kT i

N N
Internal energy: E = ∑i niε i = ∑g εe −εi kT
= ∑ i i
g ε e − βε i

∂ e − βε i q i
i i
q i
Since = −ε ie − βε i
∂β
N  de − βε i   d∑i gi e − βε i 
E =− ∑i gi  dβ  = − N   = −N  dq 
 
q   q  dβ  q  dβ V
 

At T=0, U=U(0) Ö U − U (0) = E

∂q  −1  ∂ q   ∂ ln q 
since   = 
 
 Ö U − U (0) = NkT 
2

 ∂ T V kT  ∂ β V
2
 ∂ T V

2006/5/05 48
 Enthalpy U and entropy S

• For ideal gas : PV=NkT

• As H = U + PV  ∂ ln q 
U = NkT 
2

 ∂ T V
 dln q 
= NkT 
2
 + NkT
 dT V
  dln q  
i

= NkT T   + 1
  dT V 

2006/5/05 49
a) For a system of distinguishable E
non-interacting particles

Q = qN

b) For a system of indistinguishable

non-interacting particles E

q N
Q = N!

2006/5/05 50
 Enthalpy U and entropy S

¾ For distinguishable entropy S:

 gni i 
*

 g nii  Ωdis = N! Π  * 
ln Ωdis = ln N! + ∑ ln  
i  n i! 
i
 n i! 

= (N ln N-N ) + ∑ni lngi − ∑ni ln ni + ∑ni

i i i

= N ln N + ∑ni lngi − ∑ni ln ni i

i i

 gi 
= N ln N + ∑ni ln 
i  ni 

2006/5/05 51
• Since N  −εi  gi q  εi 
ni = gi exp   = exp  
q  kT  ni N  kT 
 gi   q  εi
ln  = ln  +
 ni   N  kT
 gi 
ln Ωdis = N ln N + ∑ni ln 
i  ni 
 q  εi 
Sdis = k ln Ωdis = k N ln N + ∑ni ln +∑ 
 i N  kT 
 q  U   ∂ ln q 
= k N ln N + N ln  +  U = NkT  2

 N
  kT   ∂ T V

  ∂ ln q  
Sdis = Nk T   + ln q 
  ∂ T V 

2006/5/05 52
 16.5 Energy and Entropy

¾ for indistinguishable entropy:

g 
n*i
Ωindis = Π  i

[
ln Ωindis = ∑ ln (g ni i ) − ln ni !] n !
i *
i
i
= ∑ni ln (gi ) − ∑ni ln (ni ) + ∑ni
i i i
 gi 
= ∑ni ln  + N  gi  q  εi
 ni  ln  = ln +
 ni  N  kT
i

q  εi
ln Ωindis = ∑ni ln +∑ +N
i N  i kT

2006/5/05 53
• Since  gi 
ln Ωindis = ∑ni ln  + N
i  ni 
S indis = k ln Ωindis
 q  εi 
= k ∑ni ln +∑ +N 
 i N  kT 
 q  U   ∂ ln q 
= k N ln + +N  U = NkT  2

 N  kT   ∂ T V
  ∂ ln q  
= k NT   + N ln q − N ln N + N 
  ∂ T V 
  ∂ ln q  q  
= Nk T   + ln  + 1
  ∂ T V N  
  ∂ ln q  qe 
S indis = Nk T   + ln 
  ∂ T V N 

2006/5/05 54
• As A = U –TS  ∂ ln q 
U = NkT  2

from   ∂ ln q    ∂ T V
Sdis = Nk T   + ln q 
  ∂ T V 
 dln q   dln q 
Adis = NkT  2
 − NkT T  − NkT ln q
 dT V  dT V
= −NKT ln q = −kT ln q N = −kT lnQ

from   ∂ ln q  q 
S indis = Nk T   + ln + 1
  ∂ T V N 

 dln q    dln q  q 
Aindis = NkT  2
 − NKT T  dT  + ln N + 1
 dT V   V 
 q 
= −NkT ln + 1 = −kT [ln q N − N ln N + N ] = −kT lnQ
 N 

2006/5/05 55
• G = A + PV  ∂ ln q 
P = NkT  
q   ∂ V T
= −NkT ln 
N 
Adis = −NKT ln q
∂ G 
µi =    q 
 ∂ ni T ,P ,{n j ≠ni } Aindis = −NkT ln + 1
 N 
∂A
=  
 ∂ ni T V, ,{n ≠n }
j i

q
= −NKT ln
N

2006/5/05 56
Alternatively, ∑ i
E e −E i kT

U − U (0) = ∑ piE i = i

i Q
 ∂ ln ∑i e −E kT   ∂ lnQ 
U (T ,V ,n1,n2 ,...) = −
i
 = −k  
 ∂ (1 kT ) ( )
  ∂ 1 T V T, ,{ni }
V ,{ni }
qN
For ideal gas: A = −kT lnQ = −kT ln = −NkT ln q + kT ln N !
N!
qe
= −NkT ln q + NkT ln N − NkT = −NkT ln
N
∂A  qe  ∂ ln q  
S = −  = Nk ln +T   
 ∂ T V ,{n } i  N  ∂T V 
∂A  ∂ ln q 
P = −  = NkT  
 ∂ V T ,{n } i
 ∂V T
∂A q 
µi =   = −kT ln 
 ∂ ni T V, ,{n ≠n } j i
N 

2006/5/05 57
And,
 qe  ∂ ln q  
S = Nk ln +T   
 N  ∂T V 

For ideal gas: qe

A = −NkT ln PV = NkT
N
 ∂ (A T )   ∂ ln q 
U =   = A +TS = NkT 
2

 ∂ (1 T ) V ,{ni }  ∂T V

  ∂ ln q  
H = U + PV = NkT T   + 1
  ∂T V 

q 
G = A + PV = −NkT ln 
N 

2006/5/05 58
 Single-molecule partition function for an ideal gas

• For ideal gas (independent particle)

• Total energy ε= ε(t) + ε(v) + ε(r) + ε(e) + ε(n)
• q = Σi Σj Σk Σl Σm [ g i(t) g j(v) g k(r) g l(e) g m(n)
exp{ -[εi(t) +εj(v) +εk(r) +εl(e) +εm(n)] /kT}
≈ q (t) q (v) q (r) q (e) (ignore nuclear part q (n) )
where q (t) = Σi gi(t) exp[-εi(t)/kT ]
• Now ln q = lnq (t) + ln q (v) + ln q (r) + ln q (e)
• U = NkT 2(∂ ln q (t)/∂T )V + NkT 2(∂ ln q (v)/∂T )V
+ NkT 2(∂ ln q (r)/∂T )V + NkT 2(∂ ln q (e)/∂T )V
= Ut + Uv + Ur + Ue
** modes of motion are independent

2006/5/05 59
 Canonical partition function

• The canonical partition function Q for ideal gas is

Q = [q (t) q (v) q (r) q (e)]N /N!

• The Helmholtz energy A is proportional to the logarithm of Q

A = - NkT lnQ
= -NkT ln [q (t) e/N] -NkT ln q (v) -NkT ln q (r) -NkT ln q (e)
= At + Av + Ar + Ae
where At = -NkT ln [q (t) e/N] and
Av = -NkT ln q (v) etc.

2006/5/05 60
Example 16.2 Showing that the partition function is a product
of partition functions.
Consider a molecule that has only two translational energy
levels and two vibrational energy levels. Assume that the
translational partition function is given by
q (t) =g1(t) exp[-ε1(t) /kT ] + g2(t) exp[-ε2(t) /kT ]=a+b
Assume that the vibrational partition function is given by
q (v) =g1(v) exp[-ε1(v) /kT ] + g2(v) exp[-ε2(v) /kT ]=c+d
Show that the molecular partition function can be written as
sum of products. For the molecular partition function is sum
of products.
q = q (t) q (v) = (a+b) (c+d) = ac + ad + bc + bd

2006/5/05 61