You are on page 1of 202

Chapter 1

Introduction

The 1D Schrödinger equation:


∂Ψ ~2 ∂ 2 Ψ
i~ =− +VΨ (1.1)
∂t 2m ∂x2
Ψ = Ψ(x, t): wave function, V = V (x, t): potential energy function, m: particle mass

1.1 Probability
ˆ probability density = probability of finding the particle at point x at time t:
|Ψ(x, t)|2 = Ψ∗ Ψ (1.2)
Ψ∗ : complex conjugate of Ψ
ˆ normalization of Ψ (particle must be somewhere):
Z +∞
|Ψ(x, t)|2 dx = 1 (1.3)
−∞

ˆ conservation of normalization (if Ψ is normalized at t = 0, Ψ stays normalized


for all future time):
Z +∞
d
|Ψ(x, t)|2 dx = 0 (1.4)
dt −∞

ˆ probability current = rate [/s] at which probability is flowing past point x:


i~
J ≡ (Ψ∇Ψ∗ − Ψ∗ ∇Ψ) (1.5)
2m
∂ρ
+ ∇ · J = 0, where ρ ≡ |Ψ|2 (1.6)
∂t
1
2 CHAPTER 1. 1D

1.2 Expectation value


ˆ expectation value of an observable represented by its operator Q̂(x, p):
Z +∞
hQi = Ψ∗ Q̂Ψdx (1.7)
−∞

CAVEAT: The expectation value is NOT the average of repeated measurements


on one and the same system. It is the average of repeated measurements on an
ensemble of identically prepared systems.

Z +∞ Z +∞

hxi = Ψ xΨdx = x|Ψ(x, t)|2 dx (1.8)
−∞
Z +∞   −∞
~ ∂
hpi = Ψ∗ Ψdx (1.9)
−∞ i ∂x

ˆ Ehrenfest’s theorem (expectation values obey classical laws)


 
dhpi ∂V
= − (1.10)
dt ∂x
Chapter 2

Time-Independent 1D Schrödinger
Equation

2.1 stationary states


Schrödinger Equation:

∂Ψ ~2 ∂ 2 Ψ
i~ =− +VΨ (2.1)
∂t 2m ∂x2
Ψ = Ψ(x, t) but V = V (x), independent of t.
Separation of variables:

Ψ(x, t) = ψ(x)ϕ(t), (2.2)

Eq.(2.1) becomes:

1 dϕ ~2 1 ∂ 2 ψ
i~ =− +V (2.3)
ϕ dt 2m ψ ∂x2
The left side is a function of t alone, while the right side is a function of x alone. With
a separation constant E, Eq.(2.3) breaks up into two equations:

dϕ iE
= − ϕ ⇒ ϕ(t) = e−iEt/~ (2.4)
dt ~
~2 ∂ 2 ψ
− + V ψ = Eψ (2.5)
2m ∂x2
Using a solution to Eq.(2.5), the original wavefunction Ψ(x, t) is:

Ψ(x, t) = ψ(x)e−iEt/~ (2.6)

3
4 CHAPTER 2. TIME-INDEPENDENT 1D

ˆ constant probability density


|Ψ(x, t)|2 = |ψ|2 , independent of t (2.7)

ˆ constant expectationZvalue of observable


Z
Q(x, p)

hQ̂i = Ψ∗ Q̂Ψdx = ψ ∗ Q̂ψdx, independent of t (2.8)

Hamiltonian Ĥ(x̂, p̂) (= kinetic + potential energies) is:


p̂2 ~2 ∂ 2
Ĥ(x̂, p̂) = +V =− +V (2.9)
2m 2m ∂x2
Using this, Eq.(2.5) becomes
Ĥψ = Eψ (2.10)
The expectation value of the total energy is
Z Z Z

hHi = ψ Ĥψdx = E |ψ| dx = E |Ψ|2 dx = E.
2
(2.11)

Thus E is the total energy of the stationary state.


ˆ E is definite: σ H (standard deviation of E) = 0
2
σH ≡ hH 2 i − hHi2 = E 2 − E 2 = 0 (2.12)

ˆ E is real.
ˆ ψ(x) can always be taken to be real.
ˆ If V (x) is even, ψ(x) can always be taken to be either even or odd.
ˆ E must exceed the minimum value of V (x). Otherwise ψ cannot be normal-
ized.
Generally, Eq.(2.5) has an infinite collection of solutions ψ0 , ψ1 , ψ2 , . . .. For each of them,
the wave function is
Ψ0 (x, t) = ψ0 (x)e−iE0 t/~ , Ψ1 (x, t) = ψ1 (x)e−iE1 t/~ , . . . (2.13)
The general solution is a linear combination of separable solutions:
X∞ ∞
X
−iEn t/~
Ψ(x, t) = cn ψn (x)e = cn Ψn (x, t) (2.14)
n=0 n=0

cn ’s are constant.
CAVEAT: The general solution Eq.(2.14) is NOT independent of time, whereas each sep-
arable solution Ψn (x, t) is a stationary state (i.e. its probability density and expectation
values of observables are independent of time t).
2.2. INFINITE SQUARE WELL 5

2.2 Infinite square well


Consider a potential V (x):

0, if 0 ≤ x ≤ a
V (x) = (2.15)
∞, otherwise

For 0 ≤ x ≤ a, Eq.(2.5) becomes


√ !
∂ 2ψ 2mE
= −k 2 ψ k≡ (2.16)
∂x2 ~

Solving Eq.(2.16) under the boundary conditions:

ψ(0) = ψ(a) = 0, (2.17)

provides us with stationary states, with n = 1, 2, 3, . . .:

~2 kn2 n2 π 2 ~2
En = = (2.18)
r2m 2ma2
2  nπ 
ψn (x) = sin x (2.19)
a a

ˆ ψ ’s are orthonormal (= mutually orthogonal, and normalized to 1):


n
Z

ψm (x)ψn (x)dx = δmn (2.20)

with the Kronecker delta δmn :



0, if m 6= n
δmn = (2.21)
1, if m = n

ˆ ψ ’s are complete, i.e.


n any arbitrary function f (x) can be written as a linear
combination of them:

X
f (x) = cn ψn (x) (2.22)
n=1

The coefficients cn ’s are obtained from ψn and f by:


Z
cn = ψn∗ (x)f (x)dx (2.23)
6 CHAPTER 2. TIME-INDEPENDENT 1D

ˆ ψ ’s are alternately even and odd with respect to the center of the well at
n
x = a/2.

ˆ Each successive state has one more node. — ψ 1 has 0, ψ2 has 1, ψ3 has 2,
etc.

The most general solution is:

n2 π 2 ~2
 
Ψn (x, t) = ψn (x)exp −i t (2.24)
2ma2
X∞
Ψ(x, t) = cn Ψn (x, t) (2.25)
n=1

|cn |2 is the probability that a measurement of the energy would yield the value En .


X
|cn |2 = 1 (2.26)
n=1

The expectation value of the energy is



X
hHi = |cn |2 En (2.27)
n=1

2.3 Harmonic Oscillator


For the potential:

1
V (x) = mω 2 x2 , (2.28)
2
Eq.(2.5) becomes:

~2 ∂ 2 ψ 1
− 2
+ mω 2 x2 ψ = Eψ (2.29)
2m ∂x 2

2.3.1 Analytic approach


Using a dimensionless variable ξ:
r

ξ≡ x, (2.30)
~
2.3. HARMONIC OSCILLATOR 7

Eq.(2.29) reads:

d2 ψ
= (ξ 2 − K)ψ, (2.31)
dξ 2
2E
where K ≡ . (2.32)

2 /2
ψ(ξ) ∝ e−ξ as ξ → ∞. Therefore let ψ be expressed as:
2 /2
ψ(ξ) = h(ξ)e−ξ . (2.33)

Then Eq.(2.31) becomes:

d2 h dh
− 2ξ + (K − 1)h = 0 (2.34)
dξ 2 dξ
Let h(ξ) be expressed in the form of power series:

X
h(ξ) = aj ξ j (2.35)
j=0

Putting Eq.(2.35) into (2.34) and setting the coefficient of ξ j to 0, we get the recursion
formula:
2j + 1 − K
aj+2 = aj (2.36)
(j + 1)(j + 2)

As j → ∞, Eq.(2.36) becomes:
aj+2 2
= (2.37)
aj j
2
Meanwhile the Taylor series expansion of eξ is:
2 1 4
eξ = 1 + ξ 2 + ξ + . . . + bj ξ j + bj+2 ξ j+2 + . . . (2.38)
2!
1 1 bj+2 2 2
bj = , bj+2 = ⇒ = → as j → ∞ (2.39)
(j/2)! ((j + 2)/2)! bj j+2 j
2 2
So, as j → ∞, h(ξ) ∝ eξ , and, from Eq.(2.33), ψ(ξ) ∝ eξ /2 , which makes ψ UNnor-
malizable. Therefore, the power series of h(ξ) Eq.(2.35) must stop somewhere, i.e. the
numerator of Eq.(2.36) must become 0 at a certain j = n.

2n + 1 − K = 0 ⇒ K = 2n + 1 (2.40)
8 CHAPTER 2. TIME-INDEPENDENT 1D

Putting Eq.(2.40) into Eq.(2.32), we get discrete energy levels


 
1
En = n + ~ω, for n = 0, 1, 2, . . . (2.41)
2
Using Eq.(2.40), Eq.(2.34) becomes:
d2 h dh
2
− 2ξ + 2nh = 0 (2.42)
dξ dξ
The solution to Eq.(2.42) is named the Hermite polynomials Hn (ξ) 1 . For instance,
H0 = 1, H1 = 2ξ, H2 = 4ξ 2 − 2, H3 = 8ξ 3 − 12ξ, . . . (2.48)
Using the Hermite polynomials, Eq.(2.33) becomes
2 /2
ψ(ξ) = N Hn (ξ)e−ξ , (2.49)
where N is the normalization constant. After normalization, the stationary states for the
harmonic oscillator are written as:
 mω 1/4 1 2
ψn (x) = √ Hn (ξ)e−ξ /2 . (2.50)
π~ n
2 n!

2.3.2 Operator approach


Using p̂ = (~/i)d/dx, Hamiltonian Ĥ for the potential of Eq.(2.28) can be written as:
1  2
p̂ + (mωx̂)2

Ĥ = (2.51)
2m
Introduce a pair of new operators a+ and a− :
1
â± ≡ √ (∓ip̂ + mωx̂) (2.52)
2~mω
1
n 
d 2 2
Hn (ξ) = (−1)n eξ e−ξ (2.43)

Hn+1 (ξ) = 2ξHn (ξ) − 2nHn−1 (ξ) (2.44)
dHn
= 2nHn−1 (ξ) (2.45)


2 X zn
e−z +2zξ = Hn (ξ) (2.46)
n=0
n!
Z +∞ √
2
Hn (ξ)Hm (ξ)e−ξ dξ = π2n n!δnm (2.47)
−∞
2.3. HARMONIC OSCILLATOR 9

The product â− â+ becomes:


1  2 i
p̂ + (mωx̂)2 − [x̂, p̂]

â− â+ =
2~mω 2~
1 i
= Ĥ − [x̂, p̂], (2.53)
~ω 2~

where [Â, B̂] ≡ ÂB̂ − B̂  is the commutator of  and B̂. [x̂, p̂] has the canonical
commutation relation:

[x̂, p̂] = i~ (2.54)

Putting Eq.(2.54) into Eq.(2.53), we get


 
1
Ĥ = ~ω â− â+ − (2.55)
2

Using the commutator of â− and â+ :

[â− , â+ ] = 1, (2.56)

Hamiltonian can also be written as:


 
1
Ĥ = ~ω â+ â− + . (2.57)
2

Thus, using â± , Eq.(2.29) becomes:


 
1
~ω â± â∓ ± ψ = Eψ (2.58)
2

a+ is the raising operator — a+ makes a state ψ climb up the ladder of allowed


energies:

Ĥ(â+ ψ) = (E + ~ω)(â+ ψ) (2.59)

a− is the lowering operator — a− makes a state ψ climb down the ladder of allowed
energies:

Ĥ(â− ψ) = (E − ~ω)(â− ψ) (2.60)

§
As stated in 2.1, E must exceed the minimum value of V (x), which is 0. So, there must
be a ground state ψ0 , such that

â− ψ0 = 0 (2.61)
10 CHAPTER 2. TIME-INDEPENDENT 1D

Using Eq.(2.52), Eq.(2.61) becomes:


 
1 d
√ ~ + mωx ψ0 = 0
2~mω dx
dψ0 mω
⇒ = − xψ0 (2.62)
dx ~
The normalized solution to this equation is
 mω 1/4 mω 2
ψ0 (x) = e− 2~ x (2.63)
π~
Putting Eq.(2.61) into Eq.(2.58),
1
E0 = ~ω (2.64)
2
We can get all the excited states by applying a+ repeatedly:
 
n 1
ψn (x) = An (a+ ) ψ0 (x), En = n + ~ω, (2.65)
2
where An is the normalization constant, which can be obtained as follows. a± ψn is
proportional to ψn±1 , so

a+ ψn = cn ψn+1 , a− ψn = dn ψn−1 (2.66)

We want to determine cn , and dn . Note that a∓ is the hermitian conjugate of a± :


Z +∞ Z +∞

f (a± g)dx = (a∓ f )∗ gdx. (2.67)
−∞ −∞

In particular,
Z +∞ Z +∞

(a± ψn ) (a± ψn )dx = (a∓ a± ψn )∗ ψn dx. (2.68)
−∞ −∞

Using Eq.(2.58) and Eq.(2.65),

a+ a− ψn = nψn , a− a+ ψn = (n + 1)ψn (2.69)

From Eq.(2.68) and Eq.(2.69), we get


Z +∞ Z +∞ Z +∞
∗ 2 2
(a+ ψn ) (a+ ψn )dx = |cn | |ψn+1 | dx = (n + 1) |ψn |2 dx (2.70)
−∞ −∞ −∞
Z +∞ Z +∞ Z +∞
(a− ψn )∗ (a− ψn )dx = |dn |2 |ψn−1 |2 dx = n |ψn |2 dx (2.71)
−∞ −∞ −∞
2.4. FREE PARTICLE 11

Since ψn and ψn+1 are normalized, |cn |2 = n + 1, and |dn |2 = n. So,


√ √
a+ ψn = n + 1ψn+1 , a− ψn = nψn−1 . (2.72)

Thus
1
ψn = √ (a+ )n ψ0 . (2.73)
n!

From Eq.(2.67) and Eq.(2.69), the stationary states of the harmonic oscillator are or-
thonormal:
Z +∞

ψm ψn dx = δmn . (2.74)
−∞

2.4 Free Particle


Suppose V (x) = 0 everywhere. Eq.(2.5) becomes:
√ !
~2 d2 ψ d2 ψ 2mE
− = Eψ ⇒ = −k 2 ψ k≡ (2.75)
2m dx2 dx2 ~

The general solution is

ψ(x) = Aeikx + Be−ikx . (2.76)

Include the time dependence e−iEt/~ :


~k ~k
Ψ(x, t) = ψ(x)e−iEt/~ = Aeik(x− 2m t) + Be−ik(x+ 2m t) (2.77)

The first/second term in Eq.(2.77) is a wave traveling to the right/left. Let us rewrite
Eq.(2.77):
~k2
Ψk (x, t) = Aei(kx− 2m t) (2.78)
√ 
2mE k > 0 ⇒ traveling to the right
k ≡ ± (2.79)
~ k < 0 ⇒ traveling to the left

The stationary states of the free particle are propagating waves. However, the separable
solutions are NOT normalizable:
Z +∞ Z +∞
∗ 2
Ψk Ψk dx = |A| 1dx = ∞ (2.80)
−∞ −∞
12 CHAPTER 2. TIME-INDEPENDENT 1D

Therefore, there is no such thing as a free particle with a definite energy. The general
solution to Eq.(2.75) is a linear combination of Ψk ’s. In this case, it is an integral over
the continuous variable k:
Z +∞
1 ~k2
Ψ(x, t) = √ φ(k)ei(kx− 2m t) dk (2.81)
2π −∞
Eq.(2.81) is normalizable, as long as it carries a certain range of k’s (wave packet). At
time t = 0,
Z +∞
1
Ψ(x, 0) = √ φ(k)eikx dk (2.82)
2π −∞
We can get φ(k) from Plancherel’s theorem 2 :
Z +∞
1
φ(k) = √ Ψ(x, 0)e−ikx dx (2.84)
2π −∞

[ Additional remarks on the velocity of the wave packet ]


A wave packet is a superposition of sinusoidal functions whose amplitude is modulated
by φ(k) — it consists of ”ripples” contained within an ”envelope”. The general form of
a wave packet is
Z +∞
1
Ψ(x, t) = √ φ(k)ei(kx−ωt) dk (2.85)
2π −∞
The velocity of the individual ”ripples”, which is called the phase velocity, is
ω
vphase = (2.86)
k
On the other hand, the velocity of the ”envelope”, which is called the group velocity,
is

vgroup = (2.87)
dk
In the case of the free particle,
~k 2 ~k ~k
ω= ⇒ vphase = , vgroup = (2.88)
2m 2m m
2

Z +∞ Z +∞
1 1
f (x) = √ F (k)e ikx
dk ⇐⇒ F (k) = √ f (x)e−ikx dx, (2.83)
2π −∞ 2π −∞

where F (k) is the Fourier transform of f (x), and f (x) is the inverse Fourier transform of F (k).
2.5. δ-FUNCTION POTENTIAL 13

Meanwhile, the classical velocity of a free particle with energy E is


r
1 2 2E
E = mvclassical ⇒ vclassical = (2.89)
2 m

Using Eq.(2.79), we get

~k
vclassical = (2.90)
m

Thus we find the relation

vclassical = vgroup = 2vphase . (2.91)

2.5 δ-function potential


The Schrödinger equation has two kinds of solutions: bound states (as in the infinite
square well and the harmonic oscillator), and scattering states (as in the free particle).
The distinction between the two is

E < [ V (−∞) and V (+∞) ] ⇒ bound state
(2.92)
E > [ V (−∞) or V (+∞) ] ⇒ scattering state

2.5.1 potential well


Consider a potential well

V (x) = −αδ(x), (2.93)

where α is a positive constant, and δ(x) is the Dirac delta function:


  Z +∞
0, if x 6= 0
δ(x) ≡ , with δ(x)dx = 1 (2.94)
∞, if x = 0 ∞

Eq.(2.5) reads:

~2 d2 ψ
− − αδ(x)ψ = Eψ (2.95)
2m dx2

There are both bound states (E < 0) and scattering states (E > 0).
14 CHAPTER 2. TIME-INDEPENDENT 1D

bound states (E < 0)


In the region x < 0, V (x) = 0, so

d2 ψ
 
2mE −2mE
2
= − 2 ψ = κ2 ψ, κ≡ (2.96)
dx ~ ~

The normalizable solution to Eq.(2.96) is

ψ(x) = Aeκx (x < 0) (2.97)

In the same way, in the region x > 0

ψ(x) = Be−κx (x > 0) (2.98)

The boundary conditions for ψ are:



1. ψ is always continuous.
(2.99)
2. dψ/dx is continuous except at points where the potential is infinite.

From the first condition, A = B, so

ψ(x) = Ae−κ|x| (2.100)

The second condition in (2.99) does not hold here, because V (x) is infinite at x = 0.
Instead, let us integrate Eq.(2.5) from − to +:
+ + +
~2 d2 ψ
Z Z Z
− dx + V (x)ψ(x)dx = E ψ(x)dx (2.101)
2m − dx2 − −

In the limit  → 0, the right-hand side of Eq.(2.101) is 0, so


  Z +
dψ ∂ψ ∂ψ 2m
∆ ≡ − = 2 lim V (x)ψ(x)dx (2.102)
dx ∂x + ∂x − ~ →0 −

Here, from Eq.(2.93),


Z + Z +
lim V (x)ψ(x)dx = −α lim δ(x)ψ(x)dx = −αψ(0) (2.103)
→0 − →0 −

Therefore, instead of the second boundary condition in (2.99), we get


 
dψ 2mα
∆ = − 2 ψ(0) (2.104)
dx ~
2.5. δ-FUNCTION POTENTIAL 15

From Eq.(2.100), ∆(dψ/dx) = −2Aκ, and ψ(0) = A. So,



κ= (2.105)
~2
Using Eq.(2.105) and the definition of κ in Eq.(2.96), the allowed energy of the bound
state is
mα2
E=− (2.106)
2~2
The normalized ψ becomes:

mα −mα|x|/~2
ψ(x) = e (2.107)
~
The δ-function well has exactly one bound state.

scattering states (E > 0)


The general solution to Eq.(2.95) is

ψ = Aeikx + Be−ikx (x < 0)



(2.108)
ψ = F eikx + Ge−ikx (x > 0)

The continuity of ψ(x) at x = 0:

F +G=A+B (2.109)

From Eq.(2.104):
2mα
ik(F − G − A + B) = − (A + B)
~2  mα 
⇒ F − G = A(1 + 2iβ) − B(1 − 2iβ) β≡ (2.110)
~2 k
Remembering (2.79), A and F are the amplitudes of waves traveling to the right, and B
and G are the amplitudes of waves traveling to the left. Now suppose that a particle is
fired in from the left. In this case, there is no wave coming in from the right, so

G = 0. (2.111)

A is the amplitude of the incident wave, B is the amplitude of the reflected wave,
and F is the amplitude of the transmitted wave. From Eq.(2.109) and Eq.(2.110), the
reflection coefficient is
|B|2 β2 1
R≡ 2
= 2
= (2.112)
|A| 1+β 1 + (2~ E/mα2 )
2
16 CHAPTER 2. TIME-INDEPENDENT 1D

and the transmission coefficient is


|F |2 1 1
T ≡ 2
= 2
= . (2.113)
|A| 1+β 1 + (mα2 /2~2 E)
Note that

R+T =1 (2.114)

2.5.2 potential barrier


What happens if we change the sign of α in Eq.(2.93)? In this case, there is no bound
state. For the scattering states, the reflection and transmission coefficients (R and T )
are the same as Eq.(2.112) and Eq(2.113). A particle is just as likely to pass through the
barrier as to cross over the well.

2.6 Finite square well



−V0 (−a < x < a)
V (x) = (2.115)
0 (|x| > a)

There are both bound states (E < 0) and scattering states (E > 0).

2.6.1 bound states (E < 0)


In the region |x| > a, the Schrödinger equation is the same as Eq(2.96), so the normal-
izable solution is

Beκx (x < −a)
ψ(x) = (2.116)
F e−κx (x > a)

In the region −a < x < a, Eq.(2.5) reads:


p !
d2 ψ 2m(E + V0 )
= −l2 ψ l≡ (2.117)
dx2 ~

§
Remembering 2.1, E > Vmin = −V0 , so E +V0 > 0. The general solution to this equation
is

ψ(x) = Csin(lx) + Dcos(lx) (−a < x < a) (2.118)

§
As stated in 2.1, if V (x) is even, ψ(x) is either even or odd.
2.6. FINITE SQUARE WELL 17

even solution
For −a < x < a, the even solution of Eq.(2.118) is
ψ(x) = Dcos(lx) (2.119)
The continuity of ψ and dψ/dx at x = a says
F e−κa
 
= Dcos(la)
⇒ κ = ltan(la) (2.120)
−κF e−κa = −lDsin(la)
This can be rewritten as:
p  ap 
tanz = (z0 /z)2 − 1 z ≡ la, z0 ≡ 2mV0 (2.121)
~
Solving this equation numerically, we can get E through z.
ˆ If z0 is very large (i.e. the potential is wide and deep), Eq.(2.121) has solutions just
slightly below zn = nπ/2 (n = 1, 3, 5, 7, . . .). Therefore
n2 π 2 ~2
En + V0 ∼
= (n = 1, 3, 5, 7, . . .) (2.122)
2m(2a)2
The right-hand side is exactly the infinite square well energies Eq.(2.18).
ˆ As z 0 decreases, there are fewer and fewer bound states. If z0 < π/2, there is only
one bound state. No matter how weak the well becomes, there is always one bound
state.

odd solution
The odd solution of Eq.(2.118), i.e. Csin(lx), corresponds to Eq.(2.122) with n =
2, 4, 6, 8, . . ..

2.6.2 scattering states (E > 0)


Assuming a particle is fired in from the left,

 Aeikx + Be−ikx (x < −a)
ψ(x) = Csin(lx) + Dcos(lx) (−a < x < a) (2.123)
F eikx (x > a)

The boundary conditions are:




 Continuity of ψ at x = −a : Ae−ika + Beika = −Csin(la) + Dcos(la)
Continuity of ψ at x = +a : Csin(la) + Dcos(la) = F eika

(2.124)

 Continuity of dψ/dx at x = −a : ik[Ae−ika − Beika ] = l[Ccos(la) + Dsin(la)]
Continuity of dψ/dx at x = +a : l[Ccos(la) − Dsin(la)] = ikF eika

18 CHAPTER 2. TIME-INDEPENDENT 1D

Using two equations to eliminate C and D, and using the other two to express B and F
in terms of A, we get

e−2ika A
F = 2 2 (2.125)
cos(2la) − i (k 2kl
+l )
sin(2la)
sin(2la) 2
B = i (l − k 2 )F (2.126)
2kl
The transmission coefficient is
−1
|F |2 Vo2
 
2 2a p
T = = 1+ sin 2m(E + V0 ) (2.127)
|A|2 4E(E + V0 ) ~
3
Perfect transmission occurs when the sine is 0:
2a p n2 π 2 ~2
2m(En + V0 ) = nπ ⇒ En + V0 = , (2.128)
~ 2m(2a)2

where n is any integer such that En > 0. Note that this is the same as Eq.(2.122).

3
This effect has been observed in the laboratory and is known as the Ramsauer-Townsend effect
— the probability of collision of electrons with noble gas atoms has a minimum value for a given value
of electron’s kinetic energy.
Chapter 3

Formalism

Quantum Mechanics is described by linear algebra, wave functions being abstract vec-
tors and observables being linear operators. Generally, wave functions live in infinite-
dimensional spaces. Wwave functions must be normalizable:
Z
|Ψ|2 dx = 1 (3.1)

The set of all square-integrable functions constitutes a vector space called Hilbert
space. The inner product of two functions is:
Z
hf |gi ≡ f (x)∗ g(x)dx (3.2)

Note that

hg|f i = hf |gi∗ (3.3)

A set of functions fn is orthonormal when they are normalized and mutually orthogonal:

hfm |fn i = δmn (3.4)

A set of functions is complete when any other function can be expressed as


X
f (x) = cn fn (x) (3.5)

If the functions fn are orthonormal, cn ’s are given by:

cn = hfn |f i (3.6)

The expectation value of an observable represented by an operator Q̂ is:


Z
hQ̂i = Ψ∗ Q̂Ψdx = hΨ|Q̂Ψi (3.7)

19
20 CHAPTER 3. FORMALISM

The hermitian conjugate of an operator Q̂ is the operator Q̂† such that

hf |Q̂gi = hQ̂† f |gi (3.8)

The operator that represents an observable is hermitian:

Q̂ = Q̂† (3.9)

A determinate state for the observable Q is a state such that every measurement of
Q on an ensemble of identically prepared systems, all in the same state Ψ, is certain to
return the same value q:

Q̂Ψ = qΨ (3.10)

This is the eigenvalue equation for Q̂; Ψ is an eigenfunction of Q̂, and q is the
corresponding eigenvalue. The collection of all the eigenvalues of an observable Q is
called its spectrum. When two or more linearly independent eigenfunctions share the
same eigenvalue, the spectrum is said to be degenerate.
If two operators P̂ and Q̂ commute with each other, then we can choose eigenfunctions
of Q̂ that are simultaneously eigenfunctions of P̂ :

[P̂ , Q̂] = 0 ⇒ P̂ ψ = λp ψ, Q̂ψ = λq ψ (3.11)

3.1 Eigenfunctions of observables


3.1.1 Discrete spectra
The normalizable eigenfunctions of an observable have the following properties:

ˆ Their eigenvalues are real.


ˆ Eigenfunctions belonging to distinct eigenvalues are orthogonal. Even in the pres-
ence of degeneracy, the eigenfunctions can be chosen to be orthogonal.

ˆ Eigenfunctions are complete.


3.1.2 Continuous spectra
The above-mentioned three properties still hold in the case of continuous spectra.
3.2. STATISTICAL INTERPRETATION 21

momentum
Eq.(3.10) becomes:

~ d
fp (x) = pfp (x), (3.12)
i dx
where fp (x) is the eigenfunction, and p is the eigenvalue of the momentum. The general
solution is

fp (x) = Aeipx/~ (3.13)

The momentum is an observable, so its eigenvalue p is real. Using a property of the Dirac
δ-function1 :
Z +∞ Z +∞
∗ 0
fp0 (x)fp (x)dx = |A|2
ei(p−p )x/~ dx = |A|2 2π~δ(p − p0 ) (3.14)
−∞ −∞

Setting A = 1/ 2π~, we get the orthonormality for the momentum:

1
fp (x) = √ eipx/~ (3.15)
2π~
hfp0 |fp i = δ(p − p0 ) (3.16)

The eigenfunctions Eq.(3.15) are complete, since any square-integrable function f (x) can
be written as:
Z +∞ Z +∞
1
f (x) = c(p)fp (x)dp = √ c(p)eipx/~ dp (3.17)
−∞ 2π~ −∞
The expansion coefficient is obtained as:
Z +∞ Z +∞
hfp0 |f i = c(p)hfp0 |fp idp = c(p)δ(p − p0 )dp = c(p0 ) (3.18)
−∞ −∞

3.2 Statistical Interpretation


When the spectrum of Q̂ is discrete, the probability of getting an eigenvalue qn associated
with the eigenfunction fn (x) is

|cn |2 (cn = hfn |Ψi) (3.19)


Z +∞
1
1
δ(x) = eikx dk
2π −∞
22 CHAPTER 3. FORMALISM

When the spectrum is continuous, the probability of getting a value in the range dz is

|c(z)|2 dz (c(z) = hfz |Ψi) (3.20)

The wave function can be written as a linear combination of eigenfunctions:


X
Ψ(x, t) = cn (t)fn (x) (3.21)
n
Z
cn = hfn |Ψi = fn (x)∗ Ψ(x, t)dx (3.22)

The total probability is one:


X
|cn |2 = 1 (3.23)
n

The expectation value of an observable Q is:


X
hQi = qn |cn |2 (3.24)
n

Letting f = Ψ in Eq.(3.18) and using Eq.(3.15), we get


Z +∞
1
c(p) = hfp |Ψi = √ e−ipx/~ Ψ(x, t)dx (3.25)
2π~ −∞
This is called the momentum space wave function, which is associated with the
position space wave function through the Fourier transform:
Z +∞
1
momentum space wave function: Φ(p, t) = √ e−ipx/~ Ψ(x, t) dx (3.26)
2π~ −∞
Z +∞
1
position space wave function: Ψ(x, t) = √ eipx/~ Φ(p, t) dp (3.27)
2π~ −∞
The probability that a measurement of momentum yields a result in the range dp is:

|Φ(p, t)|2 dp (3.28)

CAVEAT: Keep in mind that the wave function per se is the same, no matter which
different basis we choose to take. For instance, the wave function can be expressed in the
basis of position, momentum, or energy eigenfunctions:
Z
1 X
Ψ(x, t) = Φ(p, t) √ eipx/~ dp = cn e−iEn t/~ ψn (x), (3.29)
2π~
but the system in question is in the same state.
3.3. UNCERTAINTY PRINCIPLE 23

3.3 Uncertainty Principle


For any observable A, using Eq.(3.8) and Eq.(3.9),

hA2 i = hΨ|Â2 Ψi = hÂΨ|ÂΨi (3.30)

Using Eq.(3.30) and the fact that  − hAi is also hermitian, the standard deviation σA
for the measurements of any observable A is written as:

σA2 = h(Â − hAi)2 i = h(Â − hAi)Ψ|(Â − hAi)Ψi = hf |f i, (3.31)

where f ≡ (Â − hAi)Ψ. For any other observable B,

σB2 = hg|gi g ≡ (B̂ − hBi)Ψ (3.32)

Using the Schwarz inequality2 ,

σA2 σB2 = hf |f ihg|gi ≥ |hf |gi|2 (3.33)

For any complex number z,


2

2 1 2 ∗
|z| ≥ [Im(z)] = (z − z ) (3.34)
2i

Letting z = hf |gi and using Eq.(3.3),


 2
1
σA2 σB2 ≥ [hf |gi − hg|f i] . (3.35)
2i

Now we use the relation:

hf |gi = hÂB̂i − hAihBi


hg|f i = hB̂ Âi − hAihBi
⇒ hf |gi − hg|f i = hÂB̂i − hB̂ Âi = h[Â, B̂]i (3.36)

Then Eq.(3.35) becomes


 2
1
σA2 σB2 ≥ h[Â, B̂]i (3.37)
2i
Z sZ
b b Z b
2
f (x)∗ g(x)dx ≤ |f (x)|2 dx |g(x)|2 dx


a a a
24 CHAPTER 3. FORMALISM

3.3.1 Energy-Time Uncertainty Principle


Let us suppose that Ψ(x, t) satisfies the Schrödinger equation:
∂Ψ
i~ = HΨ (3.38)
∂t
The time derivative of the expectation value of some operator Q(x, p, t) is:
  * +  
d d ∂Ψ ∂ Q̂ ∂Ψ
hQi = hΨ|Q̂Ψi = Q̂Ψ + Ψ Ψ + Ψ Q̂
dt dt ∂t ∂t ∂t
* +
1 1 ∂ Q̂
= − hĤΨ|Q̂Ψi + hΨ|Q̂ĤΨi + . (3.39)
i~ i~ ∂t

Ĥ is hermitian, so hĤΨ|Q̂Ψi = hΨ|Ĥ Q̂Ψi and we get


* +
d i ∂ Q̂
hQi = h[Ĥ, Q̂]i + (3.40)
dt ~ ∂t

Typically the operator Q does not depend explicitly on time. In that case, the rate
of change of the expectation value is determined by the commutation relation of the
operator with the Hamiltonian. In particular, if Q commutes with H, hQi is constant,
which means that Q is a conserved quantity.
Let us pick A = H and B = Q in Eq.(3.37) and assume that Q does not depend explicitly
on time:
 2  2  2  2
2 2 1 1 ~ dhQi ~ dhQi
σH σQ ≥ h[H, Q]i = =
2i 2i i dt 2 dt

~ dhQi

⇒ σH σQ ≥ . (3.41)
2 dt
Defining
σQ
∆E ≡ σH , and ∆t ≡ , (3.42)
dhQi

dt

we get the energy-time uncertainty principle:


~
∆E ∆t ≥ . (3.43)
2
∆t represents the amount of time it takes the expectation value of Q to change by one
standard deviation. If ∆E is small, the rate of change of all observables must be gradual.
Putting it the other way around, if any observable changes rapidly, the uncertainty in
the energy must be large.
3.4. HEISENBERG EQUATION OF MOTION 25

3.4 Heisenberg Equation of Motion


In the Schrödinger picture, wave functions carry the time dependence. On the other
hand, in the Heisenberg picture, it is operators that carry the time dependence.
Let us assume that the hamiltonian H does not depend on time. The operator in the
Heisenberg picture (QH ) is related to that in the Schrödinger picture (QS ) as:
   
iHt −iHt
QH = exp QS exp (3.44)
~ ~

Note that QH = QS at t = 0. Differentiating in terms of t,


   
∂QH iH −iHt/~ iHt/~ ∂QS −iHt/~ −iH
= e iHt/~
QS e +e e +eiHt/~
QS e−iHt/~
∂t ~ ∂t ~
i ∂QS −iHt/~ i iHt/~
= HeiHt/~ QS e−iHt/~ + eiHt/~ e − e QS e−iHt/~ H (3.45)
~   ∂t ~
i ∂QS i
= HQH + − QH H. (3.46)
~ ∂t H ~

Finally we get
 
∂QH i ∂QS
= [H, QH ] + (3.47)
∂t ~ ∂t H

In most cases in which QS does not depend explicitly on time, Eq.(3.47) can be simply
written as
∂QH i
= [H, QH ]. (3.48)
∂t ~
Eq.(3.48) is called the Heisenberg equation of motion.
26 CHAPTER 3. FORMALISM
Chapter 4

3D Schrödinger Equation

The 3D Schrödinger Equation is


∂Ψ ~2 2
i~ = − ∇ Ψ + V Ψ, (4.1)
∂t 2m
∂2 ∂2 ∂2
where ∇2 ≡ + + in cartesian coordinates. (4.2)
∂x2 ∂y 2 ∂z 2
Just in the same way as the 1D case, if V is independent of time, there will be a complete
set of stationary states:

Ψn (r, t) = ψn (r)e−iEn t/~ (4.3)

And the time-independent Schrödinger equation is:


~2 2
− ∇ ψ + V ψ = Eψ. (4.4)
2m
The general solution is
X
Ψ(r, t) = cn ψn (r)e−iEn t/~ (4.5)

4.1 Spherical Coordinates


Let us assume that the potential V depends only on r: V = V (r).
     2 
2 1 ∂ 2 ∂ 1 ∂ ∂ 1 ∂
∇ = 2 r + 2 sinθ + 2 2 (4.6)
r ∂r ∂r r sinθ ∂θ ∂θ r sin θ ∂φ2
Separation of variables:

ψ(r, θ, φ) = R(r)Y (θ, φ) (4.7)

27
28 CHAPTER 4. 3D

Putting Eq.(4.6) and Eq.(4.7) into Eq.(4.4), dividing by RY , and then multiplying by
−2mr2 /~2 ,
2mr2 1 ∂ 2Y
       
1 d 2 dR 1 1 ∂ ∂Y
r − 2 [V (r) − E] + sinθ + = 0 (4.8)
R dr dr ~ Y sinθ ∂θ ∂θ sin2 θ ∂φ2
The first term depends only on r, while the second term depends only on θ and φ, so
each term must be constant. Letting the separation constant l(l + 1),
2mr2
 
1 d 2 dR
r − 2 [V (r) − E] = l(l + 1) (4.9)
R dr dr ~
1 ∂ 2Y
   
1 1 ∂ ∂Y
sinθ + = −l(l + 1) (4.10)
Y sinθ ∂θ ∂θ sin2 θ ∂φ2

4.1.1 Angular Equation


Separation of variables again:

Y (θ, φ) = Θ(θ)Φ(φ) (4.11)

Putting this into Eq.(4.10) and multiplying by sin2 θ,


1 d2 Φ
    
1 d dΘ 2
sinθ sinθ + l(l + 1)sin θ + =0 (4.12)
Θ dθ dθ Φ dφ2
The first term depends only on θ, while the second term depends only on φ, so each term
must be constant. Letting the separation constant m2 ,
  
1 d dΘ
sinθ sinθ + l(l + 1)sin2 θ = m2 (4.13)
Θ dθ dθ
1 d2 Φ
= −m2 (4.14)
Φ dφ2
Solving Eq.(4.14), we get

Φ(φ) = eimφ (4.15)

There is a rotational symmetry in the azimuthal φ direction — the system must not
change when φ advances by 2π. So we require that

Φ(φ + 2π) = Φ(φ) ⇒ e2πim = 1 (4.16)

Therefore m must be an integer:

m = 0, ±1, ±2, . . . (4.17)


4.1. SPHERICAL COORDINATES 29

The solution to Eq.(4.13) is given by1 :


Θ(θ) = APlm (cosθ), (4.18)
where Plm is the associated Legendre function:
 |m|
m 2 |m|/2 d
Pl (x) ≡ (1 − x ) Pl (x), (4.19)
dx
and Pl (x) is the l-th Legendre polynomial, defined by
 l
1 d
Pl (x) ≡ l (x2 − 1)l (4.20)
2 l! dx
For Eq.(4.20) to make any sense, l must be a nonnegative integer
l = 0, 1, 2, . . . (4.21)
Since Pl is a polynomial of degree l, if |m| > l, Eq.(4.19) says Plm = 0. Therefore, for
any given l, there are (2l + 1) possible values of m:
m = −l, −l + 1, . . . , −1, 0, 1, . . . , l − 1, l (4.22)
l is called the azimuthal quantum number, and m the magnetic quantum number.
The volume element in spherical coordinates is
d3 r = r2 sinθdrdθdφ (4.23)
Let us normalize R(r) and Y (θ, φ) separately:
Z ∞ Z 2π Z π
2 2
|R| r dr = 1 and |Y |2 sinθdθdφ = 1 (4.24)
0 0 0

The normalized angular wave functions are called spherical harmonics:


s
(2l + 1) (l − |m|)! imφ m
Ylm (θ, φ) =  e Pl (cosθ), (4.25)
4π (l + |m|)!

(−1)m for m ≥ 0
where  = (4.26)
1 for m < 0
Spherical harmonics are orthogonal:
Z 2π Z π
0
[Ylm (θ, φ)]∗ [Ylm
0 (θ, φ)]sinθdθdφ = δll0 δmm0 (4.27)
0 0

1
Since Eq.(4.13) is a 2nd-order differential equation, there must be the other linearly independent
solution. It exists mathematically, but it is not physically acceptable because it is not normalizable.
30 CHAPTER 4. 3D

Figure 4.1: First few associated Legendre functions and spherical harmonics.
4.2. HYDROGEN ATOM 31

4.1.2 Radial Equation


Using a variable:

u(r) ≡ rR(r), (4.28)

Eq.(4.9) can be written as:

~2 d2 u ~2 l(l + 1)
 
− + V + u = Eu (4.29)
2m dr2 2m r2
The normalization condition Eq.(4.24) is:
Z ∞
|u|2 dr = 1 (4.30)
0

4.2 Hydrogen Atom


4.2.1 Energy levels and wave functions

e2 1
V (r) = − (4.31)
4π0 r
Let us see bound states (E < 0). Using these variables:

−2mE
κ ≡ (4.32)
~
me2
ρ ≡ κr, and ρ0 ≡ , (4.33)
2π0 ~2 κ
Eq.(4.29) can be rewritten as:

d2 u
 
ρ0 l(l + 1)
= 1− + u (4.34)
dρ2 ρ ρ2
As ρ → ∞, Eq.(4.34) approximately becomes:

d2 u
= u (ρ → ∞) (4.35)
dρ2
The solution to this equation is2

u(ρ) = Ae−ρ (ρ → ∞) (4.36)


2
The other solution eρ is removed because it is not physically acceptable (normalizable).
32 CHAPTER 4. 3D

On the other hand, as ρ → 0, Eq.(4.34) approximately becomes:


d2 u l(l + 1)
2
= u (ρ → 0) (4.37)
dρ ρ2
The solution to this equation is3

u(ρ) = Cρl+1 (ρ → 0) (4.38)

Thus let us introduce a new function v(ρ) such that:

u(ρ) = ρl+1 e−ρ v(ρ) (4.39)

In terms of v(ρ), Eq.(4.34) becomes:


d2 v dv
ρ 2
+ 2(l + 1 − ρ) + [ρ0 − 2(l + 1)]v = 0 (4.40)
dρ dρ
Now let us express v(ρ) as a power series in ρ:

X
v(ρ) = cj ρ j (4.41)
j=0

Putting Eq.(4.41) into Eq.(4.40), and choosing the coefficients of like powers (ρj ), we get

j(j + 1)cj+1 + 2(l + 1)(j + 1)cj+1 − 2jcj + [ρ0 − 2(l + 1)]cj = 0


 
2(j + l + 1) − ρ0
⇒ cj+1 = cj (4.42)
(j + 1)(j + 2l + 2)
As j → ∞, Eq.(4.42) becomes:
2
cj+1 ≈ cj
j+1
2j
⇒ cj = c0 (j → ∞) (4.43)
j!
Putting Eq.(4.43) into Eq.(4.41),

X 2j
v(ρ) = c0 ρj = c0 e2ρ (j → ∞) (4.44)
j=0
j!

Therefore

u(ρ) = c0 ρl+1 eρ (j → ∞) (4.45)


3
The other solution ρ−l is removed because it is not physically acceptable (normalizable).
4.2. HYDROGEN ATOM 33

Eq.(4.45) blows up as ρ → ∞, so the series in Eq.(4.41) must terminate at some point.


Letting the maximum j be jmax , from Eq.(4.42),

2(jmax + l + 1) − ρ0 = 0 (4.46)

Let us define

n ≡ jmax + l + 1 (4.47)

Then

ρ0 = 2n (4.48)

From Eq.(4.48), Eq.(4.32) and Eq.(4.33), we get the Bohr formula:


"  2 2 #
m e 1
En = − 2
(n = 1, 2, 3, . . .) (4.49)
2~ 4π0 n2

n is called the principal quantum number. The Bohr radius is defined as:
4π0 ~2
a≡ = 0.529 × 10−10 m (4.50)
me2
Putting this into Eq.(4.33) and Eq.(4.48), we get
1 r
κ= , ρ= (4.51)
an an
The spatial wave functions for hydrogen are

ψnlm (r, θ, φ) = Rnl (r)Ylm (θ, φ), (4.52)

where, referring back to Eq.(4.28) and Eq.(4.39),


1
Rnl (r) = ρl+1 e−ρ v(ρ) (4.53)
r
v(ρ) is a polynomial of degree jmax = n − l − 1 in ρ. From Eq.(4.42) and Eq.(4.48), the
coefficients are given by
 
2(j + l + 1 − n)
cj+1 = cj (4.54)
(j + 1)(j + 2l + 2)
The binding energy of the ground state (n = 1) is
"  2 2 #
m e
E1 = − = −13.6 eV (4.55)
2~2 4π0
34 CHAPTER 4. 3D

For n = 1, Eq.(4.47) and Eq.(4.22) say l = m = 0, so the wave function of the ground
state is
ψ100 (r, θ, φ) = R10 (r)Y00 (θ, φ) (4.56)
The normalized radial part is
 
c0 2
R10 (r) = e−r/a c0 = √ (4.57)
a a

And Y00 = 1/ 4π, thus
1
ψ100 (r, θ, φ) = √ e−r/a (4.58)
πa3
For arbitrary n, the allowed values of l and m are, from Eq.(4.47) and Eq.(4.22),
l = 0, 1, 2, . . . , n − 1 (4.59)
m = −l, −l + 1, . . . , −1, 0, 1, . . . , l − 1, l (4.60)
The total degeneracy of the energy level En is
n−1
X
(2l + 1) = n2 (4.61)
l=0

The polynomial v(ρ), defined by Eq.(4.54) and c0 = 2/ a, is written as a form of the
associated Laguerre polynomial:
v(ρ) = L2l+1
n−l−1 (2ρ), (4.62)
 p
p p d
Lq−p (x) ≡ (−1) Lq (x), (4.63)
dx
 q
d
Lq (x) ≡ e x
(e−x xq ), (4.64)
dx
where Lq (x) is the q-th Laguerre polynomial. The normalized hydrogen wave functions
are:
s 
3  l
2 (n − l − 1)! −r/na 2r  2l+1  m
ψnlm = e L n−l−1 (2r/na) Yl (θ, φ) (4.65)
na 2n[(n + l)!]3 na
The orthonormality of the wave functions holds:
Z

ψnlm ψn0 l0 m0 r2 sinθdrdθdφ = δnn0 δll0 δmm0 (4.66)
4.2. HYDROGEN ATOM 35

Figure 4.2: The first few radial wave functions for hydrogen, Rnl (r).
.
36 CHAPTER 4. 3D

Figure 4.3: Density plots for the hydrogen wave functions (n, l, m)
.
4.2. HYDROGEN ATOM 37

4.2.2 Spectrum of hydrogen


From Eq.(4.49), the energy of photons emitted by the transition of an electron from the
initial state to the final state is:
!
1 1
Eγ = −13.6 eV − 2 (4.67)
n2i nf

In terms of the photon wavelength, using E = 2π~c/λ,


!
1 1 1
= R − 2 , (4.68)
λ n2f ni
 2 2
m e
where R ≡ 3
= 1.097 × 107 m−1 (4.69)
4πc~ 4π0
R is called the Rydberg constant.
Lyman series (ultraviolet): transitions to the ground state (nf = 1)
Balmer series (visible): transitions to the first excited state (nf = 2)
Paschen series (infrared): transitions to the second excited state (nf = 3)

Figure 4.4: Energy levels and transitions in the spectrum of hydrogen


.
38 CHAPTER 4. 3D

4.3 Angular Momentum


The operator of the angular momentum is

L = r×p (4.70)
⇒ Lx = ypz − zpy , Ly = zpx − xpz , Lz = xpy − ypx , (4.71)

where we make the replacements: px → (~/i)∂/∂x, py → (~/i)∂/∂y and pz → (~/i)∂/∂z.

4.3.1 Eigenvalues
Lx , Ly and Lz are NOT mutually commutable:

[Lx , Ly ] = i~Lz , [Ly , Lz ] = i~Lx , [Lz , Lx ] = i~Ly (4.72)

Let us introduce the square of the total angular momentum:

L2 ≡ L2x + L2y + L2z (4.73)

L2 commutes with Lx , Ly and Lz :

[L2 , L] = 0 (4.74)

Therefore we can pick out a certain direction, define it as the z axis, and find simultaneous
eigenstates of L2 and Lz :

L2 f = λf and Lz f = µf (4.75)

Let us define the raising/lowering operator:

L± ≡ Lx ± iLy (4.76)

L± has the commutation relations:

[Lz , L± ] = ±~L± , [L2 , L± ] = 0 (4.77)

Eq.(4.77) leads to:

L2 (L± f ) = λ(L± f ), Lz (L± f ) = (µ ± ~)(L± f ) (4.78)

For a given value of λ, we have a set of states, each separated by ~ in the eigenvalue of
Lz . We can increase µ by applying L+ , but there must be the top state4 such that

L+ ft = 0 (4.79)
4
λ = hL2x i + hL2y i + µ2 ≥ µ2
4.3. ANGULAR MOMENTUM 39

Let us write the eigenvalue of Lz of the top state ft as ~l:

Lz ft = ~lft , L2 ft = λft (4.80)

Using the relation:

L2 = L± L∓ + L2z ∓ ~Lz , (4.81)

we get

L2 ft = ~2 l(l + 1)ft ⇒ λ = ~2 l(l + 1) (4.82)

There is also the bottom state such that

L− fb = 0 (4.83)

Let us write the eigenvalue of Lz of the bottom state ft as ~l:

Lz fb = ~lfb , L2 fb = λfb (4.84)

Using Eq.(4.81), we get

L2 fb = ~2 l(l − 1)fb ⇒ λ = ~2 l(l − 1) (4.85)

From Eq.(4.82), Eq.(4.85) and l ≤ l, we get

l = −l (4.86)

If we write the eigenvalues of Lz as m~, m goes from −l to +l in N integer steps, which


means l = N/2 i.e. l must be an integer or a half-integer. So, the eigenstates are
designated by l and m as:

L2 flm = ~2 l(l + 1)flm , Lz flm = ~mflm (4.87)


1 3
where l = 0, , 1, , . . . ; m = −l, −l + 1, . . . , l − 1, l (4.88)
2 2

For a given value of l, there are 2l + 1 different values of m.


Incidentally, using Eq.(4.81) and the fact that the hermitian conjugate of L± is L∓ , we
get
p
L± flm = ~ (l ∓ m)(l ± m + 1) flm±1 (4.89)
40 CHAPTER 4. 3D

4.3.2 Eigenfunctions

~
L= r×∇ (4.90)
i
In spherical coordinates,
∂ 1 ∂ 1 ∂
∇ = r̂ + θ̂ + φ̂ (4.91)
∂r r ∂θ rsinθ ∂φ
 
~ ∂ 1 ∂
L = φ̂ − θ̂ (4.92)
i ∂θ sinθ ∂φ
Lx , Ly and Lz are
 
~ ∂ ∂
Lx = −sinφ − cosφ cotθ , (4.93)
i ∂θ ∂φ
 
~ ∂ ∂
Ly = +cosφ − sinφ cotθ , (4.94)
i ∂θ ∂φ
~ ∂
Lz = (4.95)
i ∂φ
L± and L2 are
 
±iφ∂ ∂
L± = ±~e ± i cotθ (4.96)
∂θ ∂φ
1 ∂2
   
1 ∂ ∂
L2 = −~2
sinθ + (4.97)
sinθ ∂θ ∂θ sin2 θ ∂φ2

Let us define flm (θ, φ) as the simultaneous eigenfunctions of L2 and Lz with eigenvalue
~2 l(l + 1) and m~ respectively. We get flm (θ, φ) from:

1 ∂2
   
2 m 2 1 ∂ ∂
L fl = −~ sinθ + f m = ~2 l(l + 1)flm (4.98)
sinθ ∂θ ∂θ sin2 θ ∂φ2 l
~ ∂ m
Lz flm = f = ~mflm (4.99)
i ∂φ l

§
Eq.(4.88) says l takes on integer and half-integer values. For integer values of l, we already
know the answer from 4.1.1 — spherical harmonics Ylm (θ, φ) are the eigenfunctions of
L2 and Lz .

L2 Ylm (θ, φ) = ~2 l(l + 1)Ylm (θ, φ), Lz Ylm (θ, φ) = ~mYlm (θ, φ) (4.100)

The half-integer values of l are accounted for by the spin of a particle.


4.4. SPIN 41

4.4 Spin
In addition to extrinsic orbital angular momentum (L), the electron (and the other
elementary particles) carries intrinsic angular momentum (S), which has nothing to do

§
with motion in space and is not described by position variables r, θ, φ.
The algebraic theory of spin is the same as what we saw in 4.3.1. The commutation
relations are:
[Sx , Sy ] = i~Sz , [Sy , Sz ] = i~Sx , [Sz , Sx ] = i~Sy (4.101)
The eigenvectors of S 2 and Sz satisfy:
S 2 |s mi = ~2 s(s + 1)|s mi, Sz |s mi = ~m|s mi (4.102)
Letting S± ≡ Sx ± iSy ,
p
S± |s mi = ~ (s ∓ m)(s ± m + 1)|s (m ± 1)i, (4.103)
1 3
s = 0, , 1, , . . . , m = −s, −s + 1, . . . , , s − 1, s. (4.104)
2 2
CAVEAT: The quantum number l for orbital angular momentum can take on integer
values and change from one to another when the system is perturbed, whereas the quan-
tum number s for spin angular momentum can also take on half-integer values and
does NOT change due to perturbation.

4.4.1 Spin 1/2


There are just two eigenstates: spin up | 21 12 i, and spin down | 21 − 12 i. To use these as
basis vectors, let us use a two-element column matrix (called spinor):
 
1
χ+ = (4.105)
0
representing spin up, and
 
0
χ− = (4.106)
1
representing spin down. Then the general state of a spin-1/2 particle can be written as:
 
a p
χ= = aχ+ + bχ− , where |a|2 + |b|2 = 1. (4.107)
b
Let us obtain the spin operators, which are 2×2 matrices. Setting s = 1/2 in Eq.(4.102),
3 3
S 2 χ+ = ~2 χ+ , S 2 χ− = ~2 χ− . (4.108)
4 4
42 CHAPTER 4. 3D

Therefore,
 
3 2 1 0
2
S = ~ . (4.109)
4 0 1

Similarly from Eq.(4.102),


~ ~
Sz χ+ = χ+ , Sz χ− = − χ− (4.110)
2 2
Therefore
 
~ 1 0
Sz = . (4.111)
2 0 −1

From Eq.(4.103),

S+ χ− = ~χ+ , S− χ+ = ~χ− , S+ χ+ = S− χ− = 0. (4.112)

Therefore,
   
0 1 0 0
S+ = ~ , S− = ~ . (4.113)
0 0 1 0

Using Eq.(4.113) and S± = Sx ± iSy , we get


   
~ 0 1 ~ 0 −i
Sx = , Sy = . (4.114)
2 1 0 2 i 0
It is customary to write S as S = (~/2)σ, where
     
0 1 0 −i 1 0
σx ≡ , σy ≡ , σz ≡ . (4.115)
1 0 i 0 0 −1
σx , σy and σz are called the Pauli spin matrices.
CAVEAT: Sx , Sy , Sz , and S 2 are all hermitian (i.e. they are observables), whereas S+
and S− are NOT hermitian (they are not observables).
The eigenspinors of Sz are:
       
1 ~ 0 ~
χ+ = eigenvalue + , χ− = eigenvalue − . (4.116)
0 2 1 2
If we measure Sz on a particle in the general state χ in Eq.(4.107), we get +~/2 with
probability |a|2 , or −~/2 with probability |b|2 .
What if we measured Sx ?. Let us solve

Sx χ = λx χ. (4.117)
4.4. SPIN 43

The characteristic equation is



−λx ~/2 ~

~/2 −λx
=0 ⇒ λx = ± (4.118)
2
The eigenspinors of Sx are obtained by solving
    
~ 0 1 α ~ α
=± ⇒ β = ±α. (4.119)
2 1 0 β 2 β
Therefore, the normalized eigenspinors of Sx are
       
(x) 1 1 ~ (x) 1 1 ~
χ+ = √ eigenvalue + , χ− = √ eigenvalue − .(4.120)
2 1 2 2 −1 2
(x) (x)
χ+ and χ− span the space — the general state in Eq.(4.107) can be written as:
   
a+b (x) a−b (x)
χ= √ χ+ + √ χ− . (4.121)
2 2
If we measure Sx on a particle in the state Eq.(4.121), we get +~/2 with probability
(1/2)|a + b|2 , and −~/2 with probability (1/2)|a − b|2 .

4.4.2 Electron in a Magnetic Field


A spinning charged particle constitutes a magnetic dipole. Its magnetic dipole mo-
ment, µ, is proportional to its spin angular momentum, S:
µ = γS, (4.122)
where γ is called the gyromagnetic ratio. A magnetic dipole placed in a magnetic field
B feels a torque, µ × B. The energy associated with this torque is
H = −µ · B. (4.123)
The Hamiltonian of a spinning charged particle that is at rest in a magnetic field B is
H = −γB · S. (4.124)

Larmor precession
Let us place a particle of spin 1/2 at rest in a uniform magnetic field pointing in the z
direction: (Bx , By , Bz ) = (0, 0, B0 ). Using Eq.(4.124) and Eq.(4.111), the Hamiltonian
can be written as
 
γB0 ~ 1 0
H = −γB0 Sz = − (4.125)
2 0 −1
44 CHAPTER 4. 3D

The eigenstates of H are the same as those of Sz :



χ+ : with energy E+ = −(γB0 ~)/2
(4.126)
χ− : with energy E− = +(γB0 ~)/2.

Let us imagine that a particle is in the general state Eq.(4.107) at t = 0:


 
a p
χ(0) = = aχ+ + bχ− , where |a|2 + |b|2 = 1. (4.127)
b

Let us set a = cos(α/2) and b = sin(α/2). Since the Hamiltonian Eq.(4.125) is time-
independent, the state evolves with time as:
 
−iE+ t/~ −iE− t/~ cos(α/2)eiγB0 t/2
χ(t) = aχ+ e + bχ− e = . (4.128)
sin(α/2)e−iγB0 t/2

The expectation values of Sx , Sy , and Sz are

~
hSx i = χ(t)† Sx χ(t) = sin α cos(γB0 t), (4.129)
2
~
hSy i = χ(t)† Sy χ(t) = − sin α sin(γB0 t), (4.130)
2
~
hSz i = χ(t)† Sz χ(t) = cos α. (4.131)
2
hSi is tilted at a constant angle α to the z axis, and precesses about the magnetic field
at the Larmor frequency ω = γB0 .

Stern-Gerlach experiment
In an inhomogeneous magnetic field, a magnetic dipole feels a force:

F = ∇(µ · B), (4.132)

in addition to the torque µ × B. This force can be used to separate out particles with a
particular spin orientation.
Imagine a beam of neutral5 particles, traveling in the y direction, passes through a region
of inhomogeneous magnetic field of B 6 :

B = (Bx , By , Bz ) = (−αx, 0, B0 + αz), where αx  B0 , αz  B0 (4.133)


5
We assume that the particles are neutral, to avoid the deflection due to the Lorentz force.
6
We want just the z component Bz = B0 + αz, but it is impossible because it would violate the law
∇ · B = 0. The x component Bx = −αx inevitably arises.
4.4. SPIN 45

In the reference frame that moves along with the beam, the Hamiltonian can be written
as:

 0 (t < 0)
H(t) = −γ{(−αx)Sx ) + (B0 + αz)Sz } (0 ≤ t ≤ T ) (4.134)
0 (t > T )

Suppose the particles have spin 1/2. Putting Eq.(4.115) into Eq.(4.134), the matrix of
the Hamiltonian for 0 ≤ t ≤ T can be written as
 
γ~ −(B0 + αz) αx
H= (4.135)
2 αx B0 + αz

The eigenenegies can be obtained by solving:



−(γ~/2)(B0 + αz) − E± (γ~/2)αx
=0 (4.136)
(γ~/2)αx (γ~/2)(B0 + αz) − E±
γ~ p γ~
⇒ E± = ∓ (B0 + αz)2 + (αx)2 ≈ ∓ (B0 + αz) (4.137)
2 2
Letting χ± be the eigenspinors that belong to the eigenenegies E± , the time evolution of
χ(t) can be written as:

χ(t) = c+ χ+ e−iE+ t/~ + c− χ− e−iE− t/~ (0 ≤ t ≤ T ) (4.138)

Therefore, at t = T , the particle emerges in the state:

χ(T ) = c+ eiγT B0 /2 χ+ ei(αγT /2)z + c− e−iγT B0 /2 χ− e−i(αγT /2)z


 
(4.139)

Referring to Eq.(3.15), the two terms in Eq.(4.139) carry momentum in the z direction
— the spin-up component χ+ has momentum
1
pz = αγT ~, (4.140)
2
and the spin-down component has the opposite momentum, which results in the splitting
of the beam in two.

4.4.3 Addition of Angular Momenta


When we combine spin s1 with spin s2 , the total spin can be7

s = (s1 + s2 ), (s1 + s2 − 1), (s1 + s2 − 2), . . . , |s1 − s2 |. (4.141)


7
Either one or both could just as well be orbital angular momentum.
46 CHAPTER 4. 3D

system of two particles with s1 = s2 = 1/2


Suppose that we have a system that consists of two spin-1/2 particles. The state of each
particle can be expressed in the basis of spin up | ↑i (|s1,2 m1,2 i = | 21 12 i) and spin down
| ↓i (|s1,2 m1,2 i = | 21 − 12 i). The operators of the combined angular momentum are:
(1) (2)
S ≡ S (1) + S (2) , S± ≡ S± + S± . (4.142)
Let us find all the states of the two spin-1/2 particle system using these four states:
| ↑↑i, | ↑↓i, | ↓↑i, | ↓↓i. (4.143)
O Find the state with the maximum m.
1
This state has s = m = 1. It must be | ↑↑i, because, using Eq.(4.142) and
Eq.(4.102),
1 1
Sz | ↑↑i = (Sz(1) + Sz(2) )| ↑↑i = ~ + ~ = ~, ⇒ m=1 (4.144)
2 2
So we get
|s mi = |1 1i = | ↑↑i (4.145)

O Apply S
2 − to the state with m = 1 to make a state with m = 0.
Using Eq.(4.142) and Eq.(4.103),
(1) (2)
S− | ↑↑i = (S− + S− )| ↑↑i = ~| ↓↑i + ~| ↑↓i = ~(| ↑↓i + | ↓↑i) (4.146)
Normalizing Eq.(4.146), we get
1
|s mi = |1 0i = √ (| ↑↓i + | ↓↑i) (4.147)
2

O Apply S
3 − to the state with m = 0 to make a state with m = −1.

1 √
S− |1 0i = √ (~| ↓↓i + 0 + 0 + ~| ↓↓i) = 2~| ↓↓i (4.148)
2
Normalizing Eq.(4.148), we get
|s mi = |1 − 1i = | ↓↓i (4.149)

O Find a state with m = 0, which is orthogonal to |1 0i in Eq.(4.147).


4
This is the state that has s = 0:
1
|s mi = |0 0i = √ (| ↑↓i − | ↓↑i) (4.150)
2
4.4. SPIN 47

In summary, the three states with s = 1, called the triplet, are




 |1 1i = | ↑↑i
 1
s = 1 (triplet) |1 0i = √ (| ↑↓i + | ↓↑i) (4.151)

 2
 |1 − 1i = | ↓↓i,

and the state with s = 0, called the singlet, is:


1
s = 0 (singlet) : |0 0i = √ (| ↑↓i − | ↓↑i) (4.152)
2
Let us confirm |1 0i has s = 1, and |0 0i has s = 0. S 2 can be written as:

S 2 = (S (1) + S (2) ) · (S (1) + S (2) ) = (S (1) )2 + (S (2) )2 + 2S (1) · S (2) . (4.153)

Using Eq.(4.115), we get

S (1) · S (2) | ↑↓i = (Sx(1) Sx(2) + Sy(1) Sy(2) + Sz(1) Sz(2) )| ↑↓i (4.154)
~~ i~ −i~ ~ −~
= | ↓↑i + | ↓↑i + | ↑↓i (4.155)
22 2 2 2 2
~2
= (2| ↓↑i − | ↑↓i) (4.156)
4
In the same way, we get

~2
S (1) · S (2) | ↓↑i = (2| ↑↓i − | ↓↑i) (4.157)
4
Therefore,

(1) (2) ~2 1 ~2
S · S |1 0i = √ (2| ↓↑i − | ↑↓i + 2| ↑↓i − | ↓↑i) = |1 0i (4.158)
4 2 4
2
~ 1 3~2
S (1) · S (2) |0 0i = √ (2| ↓↑i − | ↑↓i − 2| ↑↓i + | ↓↑i) = − |0 0i (4.159)
4 2 4

Putting Eq.(4.158) and Eq.(4.159) into Eq.(4.153),


 2
3~2 ~2

2 3~
S |1 0i = + +2 |1 0i = 2~2 |1 0i, ⇒ s = 1, (4.160)
4 4 4
 2
3~2 3~2

2 3~
S |0 0i = + −2 |0 0i = 0. ⇒ s=0 (4.161)
4 4 4

Obviously |1 1i and |1 − 1i have s = 1.


48 CHAPTER 4. 3D

system of two particles with angular momentum l1 = l2 = 1


Each particle has three states: |l1,2 m1,2 i = |1 − 1i, |1 0i and |1 1i, so the combined system
has 3 × 3 = 9 states.

O Find the state with the maximum m.


1
For this state, m1 = m2 = 1, so the total m = 2 and the total l = 2. Let us write
this state as |l miT = |m1 m2 i:

|2 2iT = |1 1i (4.162)

O Apply L
2 − to |2 2iT to make a state with m = 1.

(1) (2)
L− |2 2iT = (L− + L− )|1 1i ∝ |0 1i + |1 0i (4.163)

After normalization we get

1
|2 1iT = √ (|0 1i + |1 0i) (4.164)
2

O Apply L
3 − to |2 1iT to make a state with m = 0.

L− |2 1iT ∝ (| − 1 1i + |0 0i) + (|0 0i + |1 − 1i) (4.165)


= | − 1 1i + 2|0 0i + |1 − 1i (4.166)

After normalization we get

1
|2 0iT = √ (| − 1 1i + 2|0 0i + |1 − 1i) (4.167)
6

O Apply L
4 − to |2 0iT to make a state with m = −1.

L− |2 0iT ∝ 0 + | − 1 0i + 2(| − 1 0i + |0 − 1i) + |0 − 1i + 0 (4.168)


∝ | − 1 0i + |0 − 1i (4.169)

After normalization we get

1
|2 − 1iT = √ (| − 1 0i + |0 − 1i) (4.170)
2
4.4. SPIN 49

O Apply L
5 − to |2 − 1iT to make a state with m = −2.

L− |2 − 1iT ∝ | − 1 − 1i (4.171)
⇒ |2 − 2iT = | − 1 − 1i (4.172)

O Find a state with m = 1, which is orthogonal to |2 1i


6
T
T
in Eq.(4.164). This
is the state for |1 1i :
1
|1 1iT = √ (|0 1i − |1 0i) (4.173)
2

O Apply L
7 − to |1 1iT to make a state with m = 0.

L− |1 1iT ∝ | − 1 1i + |0 0i − |0 0i − |1 − 1i (4.174)
= | − 1 1i − |1 − 1i (4.175)

After normalization we get


1
|1 0iT = √ (| − 1 1i − |1 − 1i) (4.176)
2

O Apply L
8 − to |1 0iT to make a state with m = −1.

L− |1 0iT ∝ 0 + | − 1 0i − |0 − 1i − 0 (4.177)
= | − 1 0i − |0 − 1i (4.178)

After normalization we get


1
|1 − 1iT = √ (| − 1 0i − |0 − 1i) (4.179)
2

O Find a state with m = 0, which is orthogonal to |2 0i


9
T T
T
in Eq.(4.167) and
|1 0i in Eq.(4.176). This is the state for |0 0i :
1
|0 0iT = √ (| − 1 1i − |0 0i + |1 − 1i). (4.180)
3

If a hydrogen atom is in the state ψnlm , the net (= spin + orbital) angular momentum
of the electron is l + 1/2 or l − 1/2. Including the spin of the proton, the atom’s total
angular momentum can be l + 1, l, or l − 1.
50 CHAPTER 4. 3D

Clebsch-Gordan coefficients
Generally, the combined state |s miT , with total spin s and z-component m, can be
written as a linear combination of the composite states |s1 m1 i|s2 m2 i:
X
|s miT = Cms1 s2 s
1 m2 m
|s1 m1 i|s2 m2 i, (4.181)
m1 +m2 =m

where the only composite states that contribute are those for which m1 +m2 = m, because
s1 s2 s
the z components simply add. The coefficients Cm 1 m2 m
are called the Clebsch-Gordan
coefficients. Some of the coefficients are listed in Fig.4.5. For example, in the 2 × 1
table, we can find
1 3 1
|3 0iT = √ |2 1i|1 − 1i + √ |2 0i|1 0i + √ |2 − 1i|1 1i (4.182)
5 5 5
Eq.(4.182) tells us that, when a system of two particles of s1 = 2 and s2 = 1 has the
(1)
total s = 3 and its z component sz = 0, then, a measurement of Sz could return ~ with
probability 1/5, or 0 with probability 3/5, or −~ with probability 1/5.
Fig.4.5 can also be used the other way round:
X
s1 s2 s
|s1 m1 i|s2 m2 i = Cm 1 m2 m
|s miT . (4.183)
s

Another example: in the 3/2 × 1 table, we can find

5 1 T 3 1 T
 r  r  r T
3 1 3 1 1 1 1
2 2 |1 0i = 5 2 2 + − . (4.184)

15 2 2 3 2 2

Eq.(4.184) tells us that, when a system of two particles of s1 = 3/2 and s2 = 1 is in the
state of m1 = 1/2 and m2 = 0, then, a measurement of the total spin s could return 5/2
with probability 3/5, or 3/2 with probability 1/15, or 1/2 with probability 1/3.
4.4. SPIN 51

Figure 4.5: Clebsch-Gordan coefficients.


52 CHAPTER 4. 3D
Chapter 5

Identical Particles

5.1 two-particle system


For a single particle, Ψ(r, t)1 is a function of the spatial coordinates r and time t. The
state of a two-particle system is a function of the coordinates of particle 1 (r1 ), the
coordinates of particle two (r2 ), and time t:
Ψ(r1 , r2 , t). (5.1)
Its time evolution is determined by the Schrödinger equation:
∂Ψ
i~ = HΨ, (5.2)
∂t
~2 2 ~2 2
where H = − ∇1 − ∇ + V (r1 , r2 , t). (5.3)
2m1 2m2 2
The probability density:
|Ψ(r1 , r2 , t)|2 d3 r1 d3 r2 (5.4)
is the probability of finding particle 1 in the volume d3 r1 and particle 2 in the volume
d3 r2 . The normalization of Ψ must hold:
Z
|Ψ(r1 , r2 , t)|2 d3 r1 d3 r2 = 1. (5.5)

For time-independent potentials, a complete set of solutions can be written as:


Ψ(r1 , r2 , t) = ψ(r1 , r2 )e−iEt/~ , (5.6)
where the spatial wave function ψ satisfies the time-independent Schrödinger equation:
~2 2 ~2 2
− ∇1 ψ − ∇ ψ + V ψ = Eψ (5.7)
2m1 2m2 2
1
We will ignore spin for the moment.

53
54 CHAPTER 5. IDENTICAL PARTICLES

5.1.1 Bosons and Fermions


Suppose that we have two electrons in a box, and that particle 1 is in the state ψa (r)
and particle 2 is in the state ψb (r). In that case, ψ(r1 , r2 ) is a simple product:
ψ(r1 , r2 ) = ψa (r1 )ψb (r2 ) (5.8)
Eq.(5.8) assumes that we can tell the particles apart, as in classical mechanics. In Quan-
tum Mechanics, however, the two electrons are utterly identical. To accomodate
the existence of particles that are indistinguishable in principle, we construct a wave
function that is indifferent to which particle is in which state. There are two ways to do
it:
ψ± (r1 , r2 ) = A[ψa (r1 )ψb (r2 ) ± ψb (r1 )ψa (r2 )] (5.9)
In Eq.(5.9), the plus sign corresponds to bosons, and the minus sign corresponds to
fermions. We know that2

particles with integer spin are bosons
(5.10)
particles with half-integer spin are fermions.
The Pauli exclusion principle: two identical fermions (e.g. two electrons) cannot
occupy the same state, because, if ψa = ψb ,
ψ− (r1 , r2 ) = A[ψa (r1 )ψa (r2 ) − ψa (r1 )ψa (r2 )] = 0, (5.11)
which means we have no wave function at all3 . More generally, let us define the exchange
operator P , which interchanges the two particles:
P f (r1 , r2 ) = f (r2 , r1 ) (5.12)
P 2 = 1, and the eigenvalues of P are ±1. Now, if the two particles are identical, the
Hamiltonian H must treat them in the same way: m1 = m2 and V (r1 , r2 ) = V (r2 , r1 )
in Eq.(5.3). Therefore, P and H are commutable:
[P, H] = 0. (5.13)
Eq.(5.13) means that we can find a complete set of functions that are simultaneous
eigenstates of both P and H — we can find solutions to the Schrödinger equation that
are either symmetric (P = 1) or antisymmetric (P = −1) under the the exchange of the
two particles3 :

ψ(r1 , r2 ) = +ψ(r2 , r1 ) for bosons
(5.14)
ψ(r1 , r2 ) = −ψ(r2 , r1 ) for fermions
2
This connection between spin and statistics is proved in relativistic quantum mechanics. In the
non-relativistic theory, we take it as an axiom.
3
Keep in mind that we are still leaving out the spin for the moment.
5.1. TWO-PARTICLE SYSTEM 55

two bosons/fermions in the infinite square well


Suppose we have two bosons or fermions in the infinite square welll. The one-particle
states are
r
2  nπ 
ψn (x) = sin x , En = n2 K, (5.15)
a a
where K ≡ π 2 ~2 /2ma2 . If the particles are distinguishable, the combined wave function
is a simple product:

ψn1 n2 (x1 , x2 ) = ψn1 (x1 )ψn2 (x2 ), En1 n2 = (n21 + n22 )K (5.16)

The ground state is


2  πx 
1
 πx 
2
ψ11 = sin sin , E11 = 2K. (5.17)
a a a
The first excited state is doubly degenerate:
 
2  πx 
1 2πx2
ψ12 = sin sin , E12 = 5K (5.18)
a a a
 
2 2πx1  πx 
2
ψ21 = sin sin , E21 = 5K. (5.19)
a a a
If the particles are bosons, the ground state is the same as Eq.(5.17), but the first excited
state is non-degenerate, with energy 5K.
√        πx 
2 πx1  2πx2 2πx1 2
sin sin + sin sin . (5.20)
a a a a a
If the particles are fermions, there is no state with energy 2K. The ground state is:
√        πx 
2 πx1  2πx2 2πx1 2
sin sin − sin sin , (5.21)
a a a a a
with energy 5K.

5.1.2 Exchange Forces


Suppose that one particle is in state ψa (x), and the other is in ψb (x), and that these
states are orthonormal. The combined wave function is

 ψ(x1 , x2 ) = ψa (x1 )ψb (x2 ) for distinguishable particles
1
ψ+ (x1 , x2 ) = 2 [ψa (x1 )ψb (x2 ) + ψb (x1 )ψa (x2 )]
√ for bosons (5.22)
1
ψ− (x1 , x2 ) = 2 [ψa (x1 )ψb (x2 ) − ψb (x1 )ψa (x2 )]
√ for fermions

56 CHAPTER 5. IDENTICAL PARTICLES

Let us calculate the expectation value of the square of the distance between the two
particles:

h(x1 − x2 )2 i = hx21 i + hx22 i − 2hx1 x2 i (5.23)

1. distinguishable particles

Z Z
hx21 i = x21 |ψa (x1 )|2 dx1 |ψb (x2 )|2 dx2 = hx2 ia ,
Z Z
hx22 i = 2
x22 |ψb (x2 )|2 dx2 = hx2 ib ,
|ψa (x1 )| dx1
Z Z
hx1 x2 i = 2
x1 |ψa (x1 )| dx1 x2 |ψb (x2 )|2 dx2 = hxia hxib

⇒ h(x1 − x2 )2 i = hx2 ia + hx2 ib − 2hxia hxib . (5.24)

2. bosons/fermions

Z Z
1
hx21 i = x21 |ψa (x1 )|2 dx1 |ψb (x2 )|2 dx2
2
Z Z
+ x21 |ψb (x1 )|2 dx1
|ψa (x2 )|2 dx2
Z Z
± x1 ψa (x1 ) ψb (x1 )dx1 ψb (x2 )∗ ψa (x2 )dx2
2 ∗

Z Z 
2 ∗ ∗
± x1 ψb (x1 ) ψa (x1 )dx1 ψa (x2 ) ψb (x2 )dx2
1 2  1
hx ia + hx2 ib ± 0 ± 0 = hx2 ia + hx2 ib .

= (5.25)
2 2

In the same way, we get

1
hx22 i = hx2 ib + hx2 ia .

(5.26)
2
5.1. TWO-PARTICLE SYSTEM 57

On the other hand,


Z Z
1
hx1 x2 i = x1 |ψa (x1 )| dx1 x2 |ψb (x2 )|2 dx2
2
2
Z Z
+ x1 |ψb (x1 )| dx1 x2 |ψa (x2 )|2 dx2
2

Z Z
± x1 ψa (x1 ) ψb (x1 )dx1 x2 ψb (x2 )∗ ψa (x2 )dx2

Z Z 
∗ ∗
± x1 ψb (x1 ) ψa (x1 )dx1 x2 ψa (x2 ) ψb (x2 )dx2
1
= (hxia hxib + hxib hxia ± hxiab hxiba ± hxiba hxiab )
2
= hxia hxib ± |hxiab |2 , (5.27)

where
Z
hxiab ≡ xψa (x)∗ ψb (x)dx. (5.28)

Therefore,

h(x1 − x2 )2 i± = hx2 ia + hx2 ib − 2hxia hxib ∓ 2|hxiab |2 . (5.29)

Compared with Eq.(5.24), Eq.(5.29) has an extra term: −2|hxiab |2 for bosons,
and +2|hxiab |2 for fermions. Bosons tend to be closer together, while fermions
tend to be farther apart4 . In other words, the system behaves as if there were an
attractive force between bosons, and a repulsive force between fermions. We call
it an exchange force5 . BUT, we have to take spin into account. The whole
state includes not only the position wave function, but also a spinor for spin:

ψ(r)χ(s), (5.30)

which, as a whole, has to be symmetric between bosons, and anti-symmetric be-


tween fermions with respect to the exchange of the two particles.

system of two electrons


Since electrons are fermions, the wave function that represents the two-electron system
must be anti-symmetric with respect to the exchange of the two particles. As for spin, the
4
Note that hxiab = 0 when ψa (x) and ψb (x) do not overlap.
5
The exchange force is not really a force. It is a purely geometrical consequence of the requirement
of the wave function Eq.(5.14).
58 CHAPTER 5. IDENTICAL PARTICLES

combined system can take the triplet states Eq.(4.151) and the singlet state Eq.(4.152).
The triplet spin states, with total spin 1, are all symmetric, so they require anti-symmetric
spatial wave function, which causes the anti-bonding exchange force. On the other hand,
the singlet state, with total spin 0, is anti-symmetric, so it requires symmetric spatial
wave function, which causes the bonding exchange force.
If the two electrons of the hydrogen molecule (H2 ) is in the singlet spin state, they tend
to get closer together toward the middle, between the two protons, and the resulting
accumulation of negative charge would attract the protons inward. This is a simple
example of what is called the covalent bond.

5.2 Atoms
A neutral atom, of atomic number Z, consists of a heavy nucleus, with electric charge Ze,
surrounded by Z electrons of mass m and charge −e. The Hamiltonian for this system
is:
Z  Z
~2 2
 2
e2
  X
X 1 Ze 1 1
H= − ∇j − + . (5.31)
j=1
2m 4π0 rj 2 4π0 j6=k |rj − rk |

The state of the whole system is obtained by solving

Hψ(r1 , r2 , . . . , rZ ) = Eψ(r1 , r2 , . . . , rZ ). (5.32)

Since electrons are fermions, acceptable solutions are only those for which the whole state
(position + spin):

ψ(r1 , r2 , . . . , rZ )χ(s1 , s2 , . . . , sZ ), (5.33)

is anti-symmetric with respect to the interchange of any two electrons. In particular, NO


two electrons can occupy the same state.

5.2.1 Helium
The Hamiltonian for helium (Z = 2) is:
~2 2 1 2e2 ~2 2 1 2e2 e2
   
1
H= − ∇1 − + − ∇2 − + . (5.34)
2m 4π0 r1 2m 4π0 r2 4π0 |r1 − r2 |
If we ignore the last term, which represents the repulsion of the two electrons, the
Schrödinger equation separates, and the solutions can be written as products of hydrogen
wave functions:

ψ(r1 , r2 ) = ψnlm (r1 )ψn0 l0 m0 (r2 ), (5.35)


5.2. ATOMS 59

but we need to change e2 to Ze2 in the Bohr radius Eq.(4.50) and the Bohr energy
Eq.(4.49):

4π0 ~2 0.529
Bohr radius a= 2
= × 10−10 [m] (5.36)
mZe" Z #
 2 2
m Ze 1 Z2
Bohr energy En = − = −13.6 [eV] (n = 1, 2, . . .), (5.37)
2~2 4π0 n2 n2

where Z = 2 in the case of helium. The total energy would be

E = 4(En + En0 ). (5.38)

ground state
As for the ground state of helium, the energy would be E = 8(−13.6 eV) = −109 eV,
and the spatial wave function would be:
8 −2(r1 +r2 )/a
ψ(r1 , r2 ) = ψ100 (r1 )ψ100 (r2 ) = e . (5.39)
πa3
Since Eq.(5.39) is symmetric, the spin state has to be anti-symmetric, which means that
the ground state of helium should be the singlet state, with the two spins oppositely
aligned. The actual ground state is indeed a singlet, but the experimentally measured
energy is −78.975 eV, which is not in good agreement with −109 eV. If we take the
electron repulsion into account, the total energy rises up from −109 to −79 eV.

excited states
The excited states of helium consist of one electron in the ground state and the other in
an excited state6 :

ψnlm ψ100 . (5.40)

We can construct both symmetric and anti-symmetric spatial wave functions, as in


Eq.(5.9). The symmetric spatial wave functions are coupled with the singlet (anti-
symmetric) spin state, and they are called parahelium, whereas the anti-symmetric
spatial wave functions are coupled with the triplet states, and they are called ortho-
helium. The ground state is parahelium. The excited states come in both forms, and
parahelium states, which have symmetric wave functions, bring the electrons closer to-
gether, as we found in Eq.(5.29). Therefore, a higher interaction energy is expected
in parahelium than in orthohelium. Indeed, it has been experimentally confirmed that
parahelium states have higher energy than their orthohelium counterparts.
6
Remember that you need to use Eq.(5.36) and Eq.(5.37), taking Z = 2 into account.
60 CHAPTER 5. IDENTICAL PARTICLES

Figure 5.1: Energy level diagram for helium. Parahelium energies are slightly higher
than their orthohelium counterparts. The values on the vertical scale are relative to the
ground state of ionized helium (He+ ): 4×(−13.6) = −54.4 eV. So, To get the total energy
of the state, you need to subtract 54.4 eV.
5.2. ATOMS 61

5.2.2 Periodic Table

§
The ground state electron configurations for heavier atoms can be discussed in the same
way as 5.2.1. Ignoring the electron repulsion, individual electrons occupy one-particle
states (n, l, m), which are called orbitals. Because of the Pauli exclusion principle, only
two electrons can occupy any given orbital, with the singlet spin state (= one with spin
up and one with spin down). There are n2 states7 that belong to the same energy En .
Thus the n = 1 shell holds 2 electrons, the n = 2 shell holds 8, and the n-th shell can
accommodate 2n2 electrons.

screening effect

With helium, the n = 1 shell is filled. Lithium (Z = 3) has to put one electron into
the n = 2 shell. For n = 2, we have l = 0 or l = 1 — which of these does the electron
choose? Ignoring the electron repulsion, Eq.(5.37) says that the energy of (n, l) = (2, 1)
is the same as that of (n, l) = (2, 0), but the orbital (n, l) = (2, 0) is taken because of
the electron repulsion. Qualitatively, angular momentum tends to throw the electron
outward. The father out it goes, the more the inner electrons screen the attractive force
of the nucleus. Therefore, within a given shell, the lowest-energy state is l = 0, and the
energy increases as l increases. So, the third electron in lithium (Z = 3) takes the orbital
(2, 0, 0). Beryllium Z = 4 also fits into this state, but boron (Z = 5) has to use l = 1.
When we reach neon Z = 10, the n = 2 shell is filled, and at argon Z = 18, the
orbitals (n, l) = (3, 0), (3, 1) are also filled8 . After argon, potassium Z = 19 could take
(n, l) = (3, 2). By this time, however, the screening effect is so strong that the energy
of (n, l) = (4, 0) is lower than that of (n, l) = (3, 2). So, potassium Z = 19 and calcium
Z = 20 takes (n, l) = (4, 0). After that, we drop back to (n, l) = (3, 2) from scandium
Z = 21 to zinc Z = 30, and then take (n, l) = (4, 1) from gallium Z = 31 to krypton
Z = 36. After that, we again jump to (n, l) = (5, 0) instead of (n, l) = (4, 2) . . . . . ..

nomenclature

n = 1 is called the K shell, n = 2 is L, n = 3 is M , etc. in alphabetical order.


l = 0 is called s (sharp), l = 1 is p (principal), l = 2 is d (diffuse), l = 3 is f (fundamental).
After that, g, h, i, k, l etc.9
The state of a particular electron is represented by nl, with the number n giving the shell,
and the letter l giving the orbital angular momentum. The magnetic quantum number
m is not shown, but an exponent is used to indicate the number of electrons that occupy
7
Remember Eq.(4.61)
8
Note that (n, l) = (3, 1) can take 6 electrons, for m = −1, 0, 1.
9
j is not used for unknown reasons.
62 CHAPTER 5. IDENTICAL PARTICLES

the state in question. For instance, the ground state of carbon:

(1s)2 (2s)2 (2p)2 (5.41)

says that it has two electrons in (1,0,0), two in (2,0,0), and two in some combination
of the orbitals (2, 1, 1), (2, 1, 0), and (2, 1, −1). Since two electrons have l = 1, the total
angular momentum L could be 2, 1, or 0. As for total spin, the two (1s) electrons are
singlet with total spin 0, and so are the two (2s) electrons, but the two (2p) electrons
could be in the singlet or triplet configuration. So, the total spin S could be 1 or 0. The
grand total J (orbital + spin) could be 3, 2, 1, or 0. We can figure out what L, S, and
J will be for a particular atom, according to Hund’s rules10 . These values are written
in this manner:
2S+1
LJ . (5.42)

The ground state of carbon is 3 P0 , with S = 1, L = 1 and J = 0.

10
Hund’s rules:
1. All other things being equal, the state with the highest total spin S has the lowest energy.
2. For a given spin, the state with the highest total L, consistent with overall anti-symmetrization,
has the lowest energy.
3. If a subshell (n, l) is no more than half filled, the lowest energy level has J = |L − S|. If it is more
than half filled, J = L + S has the lowest energy.
5.2. ATOMS 63

Figure 5.2: Ground state electron configurations for hydrogen Z = 1 to krypton Z = 36.
64 CHAPTER 5. IDENTICAL PARTICLES

5.3 Solids
In the solid state, a few of the loosely bound outermost valence electrons in each atom
become detached, and move around throughout the material. These electrons mainly
feel the potential of the entire crystal lattice, rather than the Coulomb field of a specific
nucleus.

5.3.1 Free Electron Gas Model


Suppose the object is a rectangular box, and that an electron inside feels no forces at all:

0 if 0 < x < lx , 0 < y < ly , and 0 < z < lz
V (x, y, z) = (5.43)
∞, otherwise

The Schrödinger Equation can be solved easily, using the separation of variables ψ =
X(x)Y (y)Z(z). Based on Eq.(2.19) and Eq.(2.18), we get
s      
8 nx π ny π nz π
ψnx ny nz = sin x sin y sin z (5.44)
lx ly lz lx ly lz
~2 π 2 n2x n2y n2z ~2 k 2
 
Enx ny nz = + + = , (5.45)
2m lx2 ly2 lz2 2m

where

nx = 1, 2, 3, . . . , ny = 1, 2, 3, . . . , nz = 1, 2, 3, . . . (5.46)
 
πnx πny πnz
k ≡ (kx , ky , kz ) = , , . (5.47)
lx ly lz

k is called the wave vector.


Let us imagine a three-dimensional k space, with axes kx , ky , kz , and planes drawn in
at kx = πnx /lx , at ky = πny /ly , and at kz = πnz /lz (See Fig.5.3). Each intersection
represents a one-particle stationary state. Each block contains one state in a volume of
k-space:

π3 π3
= . (5.48)
lx ly lz V

Suppose that the object contains N atoms, and each atom contributes q free electrons11 .
Since electrons are of spin-1/2 fermions, only two electrons can occupy any given state
11
Typically, N is on the order of Avogadro’s number 6 × 1023 , while q is 1 or 2.
5.3. SOLIDS 65

Figure 5.3: Each intersection on the grid represents a stationary state. Shading indicates
one block in which there is one state.

of volume π 3 /V . kx , ky and kz are all positive, so the electrons fill up one octant (= 1/8)
of a sphere in k-space. Letting the radius of the sphere kF , we get

π3
   
1 4 3 Nq
πk = . (5.49)
8 3 F 2 V
Nq
⇒ kF = (3ρπ 2 )1/3 , where ρ ≡ . (5.50)
V

ρ is called the free electron density. The boundary in k-space separating occupied and
unoccupied states is called the Fermi surface, and the corresponding energy is called
the Fermi energy EF . For a free electron gas,

~2 kF2 ~2
EF = = (3ρπ 2 )2/3 . (5.51)
2m 2m

What is the total energy of the electron gas? One octant of a shell of thickness dk
contains a volume 81 (4πk 2 )dk. Since each volume π 3 /V contains a pair of electrons, the
number of electron states in this shell is

1
2 (4πk 2 )dk V
8 = 2 k 2 dk. (5.52)
3
π /V π
66 CHAPTER 5. IDENTICAL PARTICLES

According to Eq.(5.45), each of these states has an energy ~2 k 2 /2m, so the energy of the
shell is
~2 k 2 V 2
dE = k dk (5.53)
2m π 2
Therefore, the total energy is
kF
~2 V ~2 kF5 V ~2 (3π 2 N q)5/3 −2/3
Z
Etot = 2 k 4 dk = = V (5.54)
2π m 0 10π 2 m 10π 2 m

Using Eq.(5.51) and Eq.(5.54), we find the relation:

3
Etot = N qEF . (5.55)
5
The free electron gas exerts a pressure on the walls. If the box expands by an amount
dV , the total energy Etot decreases:

2 ~2 (3π 2 N q)5/3 −5/3 2 dV


dEtot = − 2
V dV = − Etot . (5.56)
3 10π m 3 V
And the decrease in the total energy shows up as work done on the outside by the pressure
P:

dW = P dV = −dEtot . (5.57)

Comparing Eq.(5.56) and Eq.(5.57), the pressure P is

2 Etot 2 ~2 kF5 (3π 2 )2/3 ~2 5/3


P = = = ρ (5.58)
3 V 3 10π 2 m 5m
The pressure P , called degeneracy pressure, is a stabilizing internal pressure that
keeps a cold solid object from collapsing.

5.3.2 Band Structure


Solids contain regularly spaced, positively charged, stationary nuclei. The potential
of these nuclei is periodic, and let us discuss an electron in a crude model using one-
dimensional Dirac comb:
N
X −1
V (x) = α δ(x − ja), (5.59)
j=0
5.3. SOLIDS 67

consisting of evenly spaced δ-function spikes12 . Bloch’s theorem says that for a peri-
odfical potential, the wave functions of electrons can be taken to satisfy the condition:

ψ(x + a) = eiKa ψ(x), (5.60)

where the constant K is called the Bloch wave number. In reality, the solids have edges,
which make Bloch’s theorem inapplicable. But the edge effects hardly influence the
electrons deep inside the solids, because the solids contain a huge number of atoms on
the order of Avogadro’s number. Therefore, we apply a boundary condition

ψ(x + N a) = ψ(x), (5.61)

where we are assuming that the x-axis is wrapped around in a circle and is connected
onto its tail after a large number (N ∼ 1023 ) of periods. From Eq.(5.60) and Eq.(5.61),
we get

eiN Ka ψ(x) = ψ(x), (5.62)


2πn
⇒ N Ka = 2πn ⇒ K= (n = 0, ±1, ±2, . . .). (5.63)
Na
In the region 0 < x < a, the potential is 0, so the wave function can be written as

2mE
ψ(x) = A sin(kx) + B cos(kx), where k ≡ (0 < x < a). (5.64)
~
From Eq.(5.60), the wave function in the cell immediately to the left is

ψ(x) = e−iKa [A sin k(x + a) + B cos k(x + a)] (−a < x < 0). (5.65)

At x = 0,ψ must be continuous, so

B = e−iKa [A sin(ka) + B cos(ka)]. (5.66)

The boundary condition of dψ/dx for the δ-function potential is Eq.(2.104), with the sign
of α switched:
2mα
kA − e−iKa k[A cos(ka) − B sin(ka)] = B. (5.67)
~2
From Eq.(5.66) and Eq.(5.67), we get

cos(Ka) = cos(ka) + sin(ka). (5.68)
~2 k
12
Imagine that the nuclei are located at ±a/2, ±3a/2, ±5a/2, . . ..
68 CHAPTER 5. IDENTICAL PARTICLES

Introducing the following variables:


mαa
z ≡ ka, β≡ , (5.69)
~2
Eq. (5.68) becomes
sin(z)
cos(Ka) = cos(z) + β ≡ f (z). (5.70)
z
Eq.(5.70) determines the energy of the electron E through z. | cos(Ka)| cannot be greater
than 1, but f (z) strays outside the range (−1, +1). Therefore, there are gaps that
represent forbidden energies, and these gaps are separated by bands of allowed energies.
Within a given band there are N states, which can be regarded as continuous because N
is huge.
So far we have discussed one electron in the potential. What if there are N q electrons,
where q is the number of free electrons per atom? Because of the Pauli exclusion principle,
only two electrons can ooupy a given state. So, if q = 1, they half fill the first band. If
q = 2, they entirely fill the first band. If q = 3, they half fill the second band, etc.13 If a
band is entirely filled, it takes a relatively large energy to excite an electron, because the
electron has to jump across the gap — such materials are electrical insulators. If a band
is only partly filled, it takes very little energy to excite an electron — such materials are
electrical conductors. If you dope an insulator with a few atoms, some extra electrons
are put into the next higher-energy band14 , allowing for weak electric currents to flow —
such materials are semiconductors.

5.4 Quantum statistical mechanics


Let us consider a system of a large number of particles in thermal equilibrium at temper-
ature T . We make a fundamental assumption that, in thermal equilibrium, every distinct
state with the same energy is equally probable15 . The number of particles in a particular
state with energy  is given by the following equations:
1


 (−µ)/k T Maxwell-Boltzmann
 e B


1
n() = (−µ)/k
Fermi-Dirac (5.71)
 e BT + 1

 1
Bose-Einstein,


e(−µ)/k BT − 1

13
We are talking about the ground state now.
14
In other words, some holes are created in the previously filled band.
15
The total energy of every distinct state in thermal equilibrium is fixed by conservation of energy,
though random thermal motions constantly shift energy from one particle to another and from one form
to another (kinetic, rotational, vibrational, etc.) — this continual redistribution of energy does not favor
any particular state.
5.4. QUANTUM STATISTICAL MECHANICS 69

where µ is called the chemical potential, kB is the Boltzmann constant, and T is


the temperature of the system. The Maxwell-Boltzmann distribution is for classical (=
distinguishable) particles, the Fermi-Dirac distribution is for fermions, and the Bose-
Einstein distribution is for bosons. The total number of particles N and the total energy
of the system E is written as:
Z ∞
N = n()g()d (5.72)
0
Z ∞
E =  n()g()d, (5.73)
0

where g() is the density of states — g()d is the number of states whose energies are
within  and +d. Note that the chemical potential µ, which is determined by Eq.(5.72),
does not depend on , but it does depend on T .

5.4.1 Classical (Maxwell-Boltzmann) particles


Let us consider an ideal gas that is at temperature T and consists of classical particles
obeying the Maxwell-Boltzmann distribution. From Eq.(5.52) and Eq.(5.45), the number
of states in a shell of thickness d is
1 V k2
dN = dk (5.74)
2 π2
V m3/2 1/2
= √  d ≡ A1/2 d (5.75)
2π 2 ~3
So, the density of states is
g() = A1/2 . (5.76)
Eq.(5.72) becomes:
Z ∞
N= e−(−µ)/kB T A1/2 d (5.77)
0

Letting 0 = /(kB T ), Eq.(5.77) becomes16 :


Z ∞
0
N = Ae µ/kB T
(kB T ) 3/2
e− (0 )1/2 d0 (5.79)
0

µ/kB T 3/2 π
= Ae (kB T ) (5.80)
2
16


(2n)! √
Z
1
e−x xn− 2 = π (n = 1, 2, 3, . . .) (5.78)
0 22n n!
70 CHAPTER 5. IDENTICAL PARTICLES

On the other hand, Eq.(5.73) becomes


Z ∞
E = e−(−µ)/kB T A3/2 d (5.81)
0
Z ∞
0
= Ae µ/kB T
(kB T )5/2
e− (0 )3/2 d0 (5.82)
0

3 π
= Aeµ/kB T (kB T )5/2 (5.83)
4
From Eq.(5.80) and Eq.(5.83), we get
3
E(T ) = N kB T, (5.84)
2
which is the same as the classical formula for the average kinetic energy of an atom at
temperature T . From Eq.(5.80), the chemical potential µ is given by
2π~2
    
N 3
µ(T ) = kB T ln + ln . (5.85)
V 2 mkB T

5.4.2 Fermi-Dirac particles


The Fermi-Dirac distribution in Eq.(5.71) has a simple behavior as T → 0:

1 if  < µ(0)
n() → (5.86)
0 if  > µ(0)

§
At T = 0, all states are filled, up to an energy µ(0), and none are occupied for energies
above µ(0). Referring back to 5.3.1, we can notice that the chemical potential at absolute
zero temperature is the Fermi energy:
µ(0) = EF . (5.87)
For the free electron gas model,
" 2 #
5π 2 kB T

3
E(T ) = N µ(0) 1 + + ... (5.88)
5 12 µ(0)
" 2 #
π 2 kB T

µ(T ) = µ(0) 1 − + ... . (5.89)
12 µ(0)

The specific heat at constant volume Cv at low temperature can be obtained from
Eq.(5.88):
∂E π 2 N kB 2
Cv = ≈ γT, where γ = . (5.90)
∂T 2 µ(0)
5.4. QUANTUM STATISTICAL MECHANICS 71

If we measure Cv of a metal at low temperature, we can obtain the Fermi energy µ(0) of
the metal from Eq.(5.90).
Meanwhile, the entropy S is given as:

T " 2 #
π2 π2
Z  
Cv kB T kB T
S= dT = N kB 1− − ... . (5.91)
0 T 2 µ(0) 10 µ(0)

Note that S = 0 at T = 0, which means that, at T = 0, there is only one microstate17


because all the states below the Fermi energy µ(0) are occupied by electrons.

5.4.3 Bose-Einstein particles

In Eq.(5.71), let us suppose the minimum value of  = 0. n() must be always positive,
so

e−µ/kB T ≥ 1 ⇒ µ(T ) ≤ 0 for all T. (5.92)

Let us consider the free gas model again. The density of states g() is basically the same
as Eq.(5.76), but we need to take into account the fact that most of the particles
occupy the ground state at a very low temperature. So, we modify Eq.(5.76) to:


g() = δ() + A . (5.93)

Then, Eq.(5.72) becomes

Z ∞
N = N ( = 0) + n()g()d. (5.94)
0

17
Remember Bolzmann’s principle: S = kB ln W , where W is the number of microstates that the
system in question can take.
72 CHAPTER 5. IDENTICAL PARTICLES

Letting ξ = eµ/kB T , the second term of Eq.(5.94) becomes


Z ∞ Z ∞ √

n()g()d = A d
0 0 (1/ξ)e B T − 1
/k
Z ∞
√ −/kB T
 
1
= A  ξe d
0 1 − ξe−/kB T
Z ∞ ∞
√ −/kB T X n
= A  ξe ξ n e−/kB T d
0 n=0
∞ ∞

Z X n
= A  ξ n e−/kB T d
0 n=1

X Z ∞√
= A ξn  e−n/kB T d
n=1 0
∞  3/2 Z ∞
X kB T
= A ξ n
x1/2 e−x dx
n=1
n 0
√ ∞
π X
= A (kB T )3/2 ξ n n−3/2 (5.95)
2 n=1

Using Eq.(5.75) and Eq.(5.95), Eq.(5.94) becomes


 3/2 X∞
mkB T
N = N ( = 0) + V 2
enµ/kB T n−3/2 . (5.96)
2π~ n=1

N ( = 0) is the number of particles that occupy the ground state with  = 0:


1
N ( = 0) = n(0) = −µ/k T . (5.97)
e B −1
The second term of Eq.(5.96) is the number of particles in excited states:
 3/2 X∞
mkB T
N ( > 0) ≡ V enµ/kB T n−3/2 . (5.98)
2π~2 n=1

As T → 0, N ( > 0) → 0 and N ( = 0) → N . Therefore, the chemical potential18


µ → −0. But N is very large, so e−µ/kB T ≈ 1, that is, eµ/kB T ≈ 1. In this limit, Eq.(5.98)
can be approximated as:
 3/2 X ∞
mkB T
N ( > 0) ≈ V 2
n−3/2
2π~ n=1
 3/2
mkB T
= 2.612 × V . (5.99)
2π~2
18
Remember that µ ≤ 0 for all T (Eq.(5.92)).
5.4. QUANTUM STATISTICAL MECHANICS 73

At T = 0, N ( > 0) = 0 and N ( = 0) = N — all the particles occupy the ground state.


As the temperature rises, N ( > 0) increases and N ( = 0) decreases. The temperature
T0 at which N ( = 0) becomes negligible is given by setting N ( > 0) = N in Eq.(5.99):
2/3
2π~2

N
T0 = . (5.100)
mkB 2.612V

From Eq.(5.100), the critical temperature for 4 He is 3.1 K.19 From Eq.(5.99) and Eq.(5.100),
 3/2
N ( > 0) T
= , (5.101)
N T0
"  3/2 #
T
N ( = 0) = N − N ( > 0) = N 1 − (5.102)
T0

At T > T0 , N ( = 0) is negligible compared to N , whereas N ( = 0) suddenly increases


at T < T0 . This phenomenon is called the Bose-Einstein condensation. The critical
temperature T0 is called the degeneracy temperature.
In the same way as Eq.(5.95), the total energy of the system E(T ) is calculated as20 :
Z ∞
E(T ) =  n()g()d
0
 ∞
3/2 X
3 mkB T
= kB T V enµ/kB T n−5/2 . (5.103)
2 2π~2 n=1

At T > T0 , N ( = 0) ≈ 0, so the chemical potential µ(T > T0 ) is determined by setting


N ( = 0) ≈ 0 in Eq.(5.96):
 ∞
3/2 X
mkB T
N =V enµ/kB T n−3/2 . (5.104)
2π~2 n=1

At T < T0 , µ(T < T0 ) ≈ 0, so from Eq.(5.103),


 ∞
3/2 X
3 mkB T
E(T < T0 ) = kB T V n−5/2
2 2π~2 n=1
 3/2
3 mkB T
= 1.341 × kB T V . (5.105)
2 2π~2
19
The experimental value is 2.17 K. R∞
20
δ() in Eq.(5.93) has no contribution here because 0 δ()d = 0.
74 CHAPTER 5. IDENTICAL PARTICLES

Using the relation21 :

2
pV = E, (5.106)
3

and Eq.(5.103) and Eq.(5.105), the pressure of the system is obtained as



3/2 X
 
mkB T
enµ/kB T n−5/2

 kB T 2π~2 (T > T0 )


p=  n=1 3/2 (5.107)
 mkB T
 1.341 × kB T (T < T0 )


2π~2

At T > T0 , the pressure p depends on the volume V through µ, but at T < T0 , p does
not depend on V .

5.4.4 Blackbody radiation


The black body is an ideal body that absorbs all incident electromagnetic radiation
(= photons). A black body in thermal equilibrium (= at a constant temperature)
emits electromagnetic radiation called the blackbody radiation isotropically (= in
all directions with equal intensity). The energy density contained in a blackbody at
temperature T in the range of dω, dν, or dλ, is given by22 :

~ω 3 1
I(ω) dω [J m−3 ] = 2 3
dω (5.108)
π c e ~ω/kBT − 1

16π 2 ~ν 3 1
I(ν) dν [J m−3 ] = 3 2π~ν/k
dν (5.109)
c e BT − 1

16π 2 ~c 1
I(λ) dλ [J m−3 ] = dλ (5.110)
λ5 e2π~c/kB λT − 1

Stefan-Boltzmann Law
The total energy density contained in a blackbody is:

π 2 kB
4
Z  
V
Itot −3
[J m ] = I(ω)dω = T4
0 15~3 c3
= 7.565 × 10−16 T 4 (5.111)
21
See Eq.(5.58). This relation holds for any ideal gas, whether it obeys Maxwell-Boltzmann, Fermi-
Dirac, or Bose-Einstein statistics.
22
For photons, E = hν = ~ω = hc/λ.
5.4. QUANTUM STATISTICAL MECHANICS 75

The total energy of the blackbody radiation per unit surface area per unit time is:

c ∞
Z  2 4 
S π kB
−1 −2
Itot [J s m ] = I(ω)dω = T4
4 0 60~3 c2
= 5.670 × 10−8 T 4 (5.112)

The constant σ = 5.670 × 10−8 [J s−1 m−2 T−4 ] is called the Stefan-Boltzmann con-
stant.

Approximations
In the limit of high frequency 2π~ν/kB T  1, Eq.(5.109) can be approximated as:

16π 2 ~ν 3 −2π~ν/kB T
I(ν) dν [J m−3 ] ≈ e dν, (5.113)
c3
which is called Wien’s distribution law.
In the limit of low frequency 2π~ν/kB T  1, Eq.(5.109) can be approximated as:

8πν 2
I(ν) dν [J m−3 ] ≈ kB T dν, (5.114)
c3
which is called Rayleigh-Jeans distribution law.

Wien’s displacement law


Let λm be the wavelength at which the energy density per unit wavelength (= I(λ) in
Eq.(5.110)) is at its maximum. λm is proportional to T −1 :

2π~c
λm T = b, where b = = 2.898 × 10−3 [m K] (5.115)
4.965 kB
On the other hand, letting νm be the frequency at which the energy density per unit
frequency (= I(ν) in Eq.(5.109)) is at its maximum, we get

2.821kB
νm = b0 T, where b0 = = 5.878 × 1010 [s−1 K−1 ] (5.116)
2π~
76 CHAPTER 5. IDENTICAL PARTICLES
Chapter 6

Time-Independent Perturbation

6.1 Non-degenerate case


6.1.1 General Formulation
Suppose we know the solutions to the time-independent Schrödinger equation for some
potential:

H (0) ψn(0) = En(0) ψn(0) , (6.1)


(0) (0)
where En ’s are eigenenergies, and ψn ’s are a complete set of orthonormal eigenfunc-
tions:

hψn(0) |ψm
(0)
i = δnm . (6.2)

Now let us perturb the potential:

H = H (0) + λH 0 , (6.3)

and find approximate solutions to the new eigenfunctions and eigenenergies:

(H (0) + λH 0 )ψn = En ψn . (6.4)

We expand the new En and ψn as:

En = En(0) + λEn(1) + λ2 En(2) + . . . (6.5)


ψn = ψn(0) + λψn(1) + λ2 ψn(2) + . . . . (6.6)
(1) (1)
En is the first-order correction to the n-th eigenvalue, and ψn is the first-order
(2)
correction to the n-th eigenfunction. En is the second-order correction to the n-th
(2)
eigenvalue, and ψn is the second-order correction to the n-th eigenfunction. Let us put

77
78 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

Eq.(6.5) and Eq.(6.6) into Eq.(6.4), and collect like powers of λ. To lowest order λ0 , we
(0) (0) (0)
get H (0) ψn = En ψn , which is the same as Eq.(6.1). To first order λ1 ,

H (0) ψn(1) + H 0 ψn(0) = En(0) ψn(1) + En(1) ψn(0) . (6.7)

To second order λ2 ,

H (0) ψn(2) + H 0 ψn(1) = En(0) ψn(2) + En(1) ψn(1) + En(2) ψn(0) . (6.8)

6.1.2 first-order correction


(0)
Taking the inner product of Eq.(6.7) with ψn ,

hψn(0) |H (0) |ψn(1) i + hψn(0) |H 0 |ψn(0) i = En(0) hψn(0) |ψn(1) i + En(1) hψn(0) |ψn(0) i. (6.9)

H (0) is hermitian, so

hψn(0) |H (0) ψn(1) i = hH (0) ψn(0) |ψn(1) i = En(0) hψn(0) |ψn(1) i, (6.10)
(0) (0)
and this cancels the first term on the right. And hψn |ψn i = 1, so

En(1) = hψn(0) |H 0 |ψn(0) i. (6.11)


(0)
This is the expectation value of the perturbation H 0 in the unperturbed state ψn .
(1)
To obtain ψn , let us rewrite Eq.(6.7) as

(H (0) − En(0) )ψn(1) = −(H 0 − En(1) )ψn(0) (6.12)


(0)
The unperturbed ψn ’s are complete, so any function can be expressed as a linear com-
bination of them:
X
ψn(1) = c(n) (0)
m ψm . (6.13)
m6=n

(1)
We can exclude m = n from the sum, because, if ψn satisfies Eq.(6.12), so too does
(1) (0)
ψn + αψn for any constant α. Putting Eq.(6.13) into Eq.(6.12) and using Eq.(6.1), we
get
X
(Em(0)
− En(0) )cm ψm = −(H 0 − En(1) )ψn(0) .
(n) (0)
(6.14)
m6=n

(0)
Taking the inner product with ψl ,
X (0) (0) (0) (0) (0)
(0) 0 (0)
(Em − En(0) )c(n) (1)
m hψl |ψm i = −hψl |H |ψn i + En hψl |ψn i (6.15)
m6=n
6.1. NON-DEGENERATE CASE 79

If l = n, we recover Eq.(6.11). If l 6= n,
(0) (0)
(0) (n) (0) hψm |H 0 |ψn i
(El − En(0) )cl = −hψl |H 0 |ψn(0) i ⇒ c(n)
m = (0) (0)
(6.16)
En − Em
Therefore
(0) (0)
X hψm |H 0 |ψn i
ψn(1) = (0) (0)
(0)
ψm . (6.17)
m6=n En − Em

6.1.3 second-order correction


(0)
Taking the inner product of the second-order equation Eq.(6.8) with ψn ,

hψn(0) |H (0) |ψn(2) i + hψn(0) |H 0 |ψn(1) i = En(0) hψn(0) |ψn(2) i + En(1) hψn(0) |ψn(1) i + En(2) hψn(0) |ψn(0) i.

H (0) is hermitian, so

hψn(0) |H (0) ψn(2) i = hH (0) ψn(0) |ψn(2) i = En(0) hψn(0) |ψn(2) i (6.18)
(0) (0)
And hψn |ψn i = 1, so we get

En(2) = hψn(0) |H 0 |ψn(1) i − En(1) hψn(0) |ψn(1) i (6.19)

But
X
hψn(0) |ψn(1) i = c(n) (0) (0)
m hψn |ψm i = 0. (6.20)
m6=n

So
X
En(2) = hψn(0) |H 0 |ψn(1) i = (n)
cm hψn(0) |H 0 |ψm
(0)
i
m6=n
(0) (0) (0) (0)
X hψm |H 0 |ψn ihψn |H 0 |ψm i
= (0) (0)
. (6.21)
m6=n En − Em

Finally we get the second-order correction to the energy as:


(0) 0 (0) 2

X hψ m |H |ψn i
En(2) = (0) (0)
(6.22)
m6=n En − Em

In summary, in the short-hand notation


(0)
Vmn ≡ hψm |H 0 |ψn(0) i and ∆mn ≡ Em
(0)
− En(0) , (6.23)
80 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

the first three corrections to the n-th energy are:


X |Vnm |2
En(1) = Vnn , En(2) = , (6.24)
m6=n
∆nm
X Vnl Vlm Vmn X |Vnm |2
En(3) = − Vnn
l,m6=n
∆nl ∆nm m6=n
∆2nm

6.2 Degenerate case


6.2.1 Two-Fold degeneracy
(0) (0)
Suppose that two distinct unperturbed states ψa and ψb share the same energy E (0) :
(0) (0) (0)
H (0) ψa(0) = E (0) ψa(0) , H (0) ψb = E (0) ψb , hψa(0) |ψb i = 0, (6.25)
(0) (0)
with ψa and ψb both normalized. Let us write a general degenerate unperturbed state
(0) (0)
as a linear combination of ψa and ψb :
(0)
ψ (0) = αψa(0) + βψb . (6.26)

ψ (0) is also an eigenstate of H (0) with the same eigenenergy E (0) :

H (0) ψ (0) = E (0) ψ (0) . (6.27)

Let us write the perturbed Hamiltonian as

H = H (0) + λH 0 . (6.28)

We want to solve the Schrödinger equation

Hψ = Eψ, (6.29)

where we expand E and ψ as:

E = E (0) + λE (1) + λ2 E (2) + . . . (6.30)


ψ = ψ (0) + λψ (1) + λ2 ψ (2) + . . . . (6.31)

Putting Eq.(6.28), Eq.(6.30) and Eq.(6.31) into Eq.(6.29), we get, to first order λ1 ,

H (0) ψ (1) + H 0 ψ (0) = E (0) ψ (1) + E (1) ψ (0) . (6.32)


(0)
Taking the inner product with ψa ,

hψa(0) |H (0) |ψ (1) i + hψa(0) |H 0 |ψ (0) i = E (0) hψa(0) |ψ (1) i + E (1) hψa(0) |ψ (0) i (6.33)
6.2. DEGENERATE CASE 81

(0) (0)
Since H (0) is hermitian, hψa |H (0) |ψ (1) i = E (0) hψa |ψ (1) i. And using Eq.(6.26) and the
(0) (0)
orthogonality hψa |ψb i = 0, we get
(0)
αhψa(0) |H 0 |ψa(0) i + βhψa(0) |H 0 |ψb i = αE (1) . (6.34)
Let us define
(0) (0)
Wij ≡ hψi |H 0 |ψj i, (i, j = a, b). (6.35)
Then Eq.(6.34) can be written as
αWaa + βWab = αE (1) . (6.36)
(0)
In the same way, taking the inner product of Eq.(6.32) with ψb , we get
αWba + βWbb = βE (1) . (6.37)
Multiplying Eq.(6.37) by Wab and using Eq.(6.36) to eliminate βWab , we get
α[Wab Wba − (E (1) − Waa )(E (1) − Wbb )] = 0. (6.38)
If α 6= 0, Eq.(6.38) becomes
2
E (1) − E (1) (Waa + Wbb ) + (Waa Wbb − Wab Wba ) = 0. (6.39)

Therefore, noting Wba = Wab ,
(1) 1 h p
2 2
i
E± = Waa + Wbb ± (Waa − Wbb ) + 4|Wab | . (6.40)
2
What if α = 0? In this case β = 1, and Eq.(6.36) says Wab = 0, and Eq.(6.37) says
E (1) = Wbb . This result can be obtained in the general result Eq.(6.40) when we take
the minus sign, assuming Waa > Wbb . The plus sign corresponds to α = 1, β = 0. The
answers we have got here:
(1) (1) (0) (0)
E+ = Waa = hψa(0) |H 0 |ψa(0) i, E− = Wbb = hψb |H 0 |ψb i, (6.41)
are the same as the first-order correction in the non-degenerate case Eq.(6.24). So, it
(0) (0)
makes calculation simple to bring ψa and ψb that satisfy Wab = 0 from the start,
instead of a linear combination of them. For this purpose, the following theorem is
useful:
[Theorem]
(0)
Let A be a hermitian operator (= observable) that commutes with H (0) and H 0 . If ψa
(0)
and ψb (= the degenerate eigenfunctions of H (0) ) are also eigenfunctions of A with
distinct eigenvalues, then Wab = 0:
(0) (0)
Aψa(0) = µψa(0) , Aψb = νψb (µ 6= ν) ⇒ Wab = 0. (6.42)
In summary, if we are faced with degenerate states,
82 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

1. Look for an hermitian operator A that commutes with H (0) and H 0 .


2. Choose the states that are simultaneously eigenfunctions of H (0) and A, and take
them as our ”good” unperturbed states.
3. Then use the simple first-order correction of the non-degenerate case.
4. If we cannot find the operator A, we cannot help but use Eq.(6.40).

6.2.2 Higher-order degeneracy


We can rewrite Eq.(6.36) and Eq.(6.37) in matrix form:
    
Waa Wab α (1) α
=E . (6.43)
Wba Wbb β β
Eq.(6.39) is the characteristic equation for this matrix:

Waa − E (1) W ab

= 0. (6.44)
Wba Wbb − E (1)
E (1) ’s are the eigenvalues of the W -matrix. If we take the eigenvectors of W as our
”good” unperturbed states, we can use the simple first-order correction of the non-
degenerate case to calculate E (1) .
In the case of n−fold degeneracy, we look for the eigenvalues of the n × n matrix with
elements:
(0) (0)
Wij = hψi |H 0 |ψj i. (6.45)
Finding our ”good” unperturbed states is equal to constructing a basis in the degenerate
subspace that diagonalizes W , which will be automatically diagonal if we find an op-
erator A that commutes with H (0) and H 0 , and then use the simultaneous eigenfunctions
of A and H (0) as our ”good” unperturbed states.

6.3 Fine Structure of Hydrogen


The fine structure of hydrogen can be analyzed, based on time-independent perturba-
tion theory. Compared to the Bohr energies Eq.(4.49), fine structure is smaller by α2 ,
where
e2 1
α≡ ≈ (6.46)
4π0 ~c 137.036
is called the fine structure constant. The Lamb shift results from the quantization of
the electric field. The hyperfine splitting comes from the magnetic interaction between
the dipole moment of the electron and the proton.
6.3. FINE STRUCTURE OF HYDROGEN 83

Bohr energies   ∝ α2 mc2


relativistic correction
Fine structure ∝ α4 mc2
spin-orbit coupling
Lamb shift ∝ α5 mc2
Hyperfine structure ∝ (m/mp )α4 mc2

6.3.1 relativistic correction


In the Schrödinger equation:
~2 2
   2 
∂Ψ p
i~ = − ∇ +V Ψ= + V Ψ ≡ (T + V )Ψ, (6.47)
∂t 2m 2m
T is the classical expression for kinetic energy. The relativistic formula is
mc2
T =p − mc2 . (6.48)
1 − (v/c)2
Let us express T in terms of the relativistic momentum:
mv
p= p (6.49)
1 − (v/c)2
Notice that
2 2 m2 v 2 c2 + m2 c4 [1 − (v/c)2 ]
2 4 m 2 c4
p c +m c = 2
= 2
= (T + mc2 )2
1 − (v/c) 1 − (v/c)
"r #
p  p 2
⇒T = p2 c2 + m2 c4 − mc2 = mc2 1+ −1 . (6.50)
mc

Let us expand T in powers of the small number (p/mc), we get


 
2 1  p 2 1  p 4
T = mc 1 + − ... − 1
2 mc 8 mc
p2 p4
= − + ... (6.51)
2m 8m3 c2
So, the first-order relativistic correction to the Hamiltonian is
p4
H0 = − . (6.52)
8m3 c2
Using Eq.(6.11) and the fact that p2 is hermitian, the first-order correction to En is
1 1
En(1) = − 3 2
hψn |p4 ψn i = − 3 2 hp2 ψn |p2 ψn i (6.53)
8m c 8m c
84 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

From the Schrödinger equation:

p2 ψn = 2m(En − V )ψn . (6.54)

Putting Eq.(6.54) into Eq.(6.53),


1 1
En(1) = − 2
h(En − V )2
i = − 2
[En2 − 2En hV i + hV 2 i] (6.55)
2mc 2mc
In the case of hydrogen, V = −(1/4π0 )e2 /r, so
"  2     2 2  #
1 e 1 e 1
En(1) = − 2
En2 + 2En + (6.56)
2mc 4π0 r 4π0 r2

To obtain h1/ri and h1/r2 i, we use the Feynman-Hellmann theorem:

Suppose the Hamiltonian H is a function of some parameter λ. Let En (λ) and ψn (λ) be
the eigenvalues and eigenfunctions of H(λ), and assume that En ’s are non-degenerate,
or, if degenerate, that ψn ’s are our ”good” linear combinations of the degenerate eigen-
functions. Then,
 
∂En ∂H
= ψn ψn (6.57)
∂λ ∂λ

From Eq.(4.29), the effective Hamiltonian for the radial wave functions is

~2 d2 ~2 l(l + 1) e2 1
Hef f = − + − , (6.58)
2m dr2 2m r2 4π0 r
and the eigenenergies can be written in terms of l as:
me4
En = − . (6.59)
32π 2 20 ~2 (jmax + l + 1)2
To get h1/ri, from Eq.(6.58) and Eq.(6.59),
∂H 2e 1 ∂En 4
=− , = En (6.60)
∂e 4π0 r ∂e e
Therefore, from Eq.(6.57) and the Bohr formula Eq.(4.49), we get
 
4 e 1
En = − (6.61)
e 2π0 r
me2 1
 
1 1
= 2 2
= 2 , (6.62)
r 4π0 ~ n na
6.3. FINE STRUCTURE OF HYDROGEN 85

where a is the Bohr radius. To get h1/r2 i,

∂H ~2
= (2l + 1) (6.63)
∂l 2mr2
∂En 2me4 2En
= 2 2 2 3
=− (6.64)
∂l 32π 0 ~ (jmax + l + 1) n
(6.65)

So, from Eq.(6.57) we get


 
1 4mEn 1
2
=− 2
= 3 , (6.66)
r n(2l + 1)~ n (l + 1/2)a2

and Eq.(6.56) becomes


" 2 #
2 2
  
1 e 1 e 1
En(1) = − 2
En2 + 2En +
2mc 4π0 n2 a 4π0 (l + 1/2)n3 a2
E2
 
4n
= − n2 −3 . (6.67)
2mc l + 1/2

6.3.2 spin-orbit coupling


In the electron’s point of view, the proton is circling around the electron — this orbiting
positive charge sets up a magnetic field B in the electron’s frame. The Hamiltonian is

H = −µ · B. (6.68)

B: magnetic field of the proton


Let us imagine the proton as a continuous current loop in the electrons’ frame. Its
magnetic field is given by the Biot-Savart law:

µ0 I
B= , (6.69)
2r
with an effective current I = e/T , where e is the charge of the proton and T is the period
of the orbit. Meanwhile, the orbital angular momentum of the electron in the proton’s
rest frame is L = rmv = 2πmr2 /T . B and L point in the same direction, so, using

c = 1/ 0 µ0 ,

1 e
B= L (6.70)
4π0 mc2 r3
86 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

µ: magnetic dipole moment of the electron


Let us derive µ in classical electrodynamics. Imagine a charge q is smeared out around
a ring of radius r, and that the ring rotates about the axis with period T . Then, the
magnetic moment of the electron is
q 2
µ= πr . (6.71)
T
If the mass of the ring is m, its angular momentum S is

S = mr2 (6.72)
T
S and µ point in the same direction (or opposite, if the charge is negative), so
 q 
µ= S. (6.73)
2m
But Eq.(6.73) is WRONG. Quantum electrodynamics says that the real µ has a
deviation from this naive classical expectation and can be written as
 q 
µ = ge S, (6.74)
2m
where ge is called the g-factor of the electron. Putting Eq.(6.74) and Eq.(6.70) into
Eq.(6.68) and setting q = −e, we get
 2 
0 e 1
H = ge S · L. (6.75)
8π0 m c2 r3
2

But this is still WRONG. We saw the system in the rest frame of the electron, but it is
not an inertial system — the electron accelerates as the electron orbits the nucleus1 . For
this reason, we have to make a kinematic correction known as the Thomas precession,
which subtracts 1 from the g-factor:
 2 
0 e 1
H = (ge − 1) S · L. (6.76)
8π0 m c2 r3
2

Let us calculate the first-order correction to the energy En , using the perturbation theory.
In the presence of spin-orbit coupling, the Hamiltonian commutes with L2 , S 2 and the
total angular momentum:

J ≡ L + S. (6.77)
1
In the lab frame, the electron is a magnetic dipole moving around the proton, and a moving magnetic
dipole acquires an electric dipole moment. In fact, the spin-orbit coupling is due to the interaction of
the electric field of the proton with the electric dipole moment of the electron.
6.3. FINE STRUCTURE OF HYDROGEN 87

So, the eigenstates of L2 , S 2 , J 2 and Jz are our ”good” unperturbed states to use. Note
that the hamiltonian neither commutes with L nor S.

J 2 = (L + S) · (L + S) = L2 + S 2 + 2L · S (6.78)
1 2
⇒L·S = (J − L2 − S 2 ). (6.79)
2
So, the eigenvalues of L · S(= S · L) are

~2
[j(j + 1) − l(l + 1) − s(s + 1)], (6.80)
2
where s = 1/2, and j = l ± 1/2. Meanwhile, putting Eq.(6.62) and Eq.(6.66) into
Kramers’ relation:
s+1 s s
2
hr i − (2s + 1)ahrs−1 i + [(2l + 1)2 − s2 ]a2 hrs−2 i = 0, (6.81)
n 4
we get
 
1 1
= . (6.82)
r3 l(l + 1/2)(l + 1)n3 a3

From Eq.(6.76), Eq.(6.80) and Eq.(6.82), the first-order correction becomes

En2 n[j(j + 1) − l(l + 1) − 3/4]


 
(1)
En = (ge − 1) (6.83)
mc2 l(l + 1/2)(l + 1)

Since the g-factor of the electron2 is almost 2, Eq.(6.83) can be written as:

En2 n[j(j + 1) − l(l + 1) − 3/4]


 
(1)
En = (6.84)
mc2 l(l + 1/2)(l + 1)
2
Quantum electrodynamics says
 α  α 2  α 3  α 4 
ge = 2 1 + c1 + c2 + c3 + c4 + ... ,
π π π π
1
where c1 = , c2 = −0.3285, c3 = 1.177, c4 = −1.911, . . . .
2
The deviation from 2 is called the anomalous magnetic moment, which is defined as:
ge − 2
ae = .
2
Theoretical calculation has provided the value:
atheory
e = 0.001 159 652 181 78(±77),
which is in exquisite agreement with the experimental value:
aexperiment
e = 0.001 159 652 180 91(±26).
88 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

Adding Eq.(6.84) and the relativistic correction Eq.(6.67) together, we get the total fine-
structure formula:
En2
 
(1) 4n
En,F S = 3− . (6.85)
2mc2 j + 1/2

Combining this with the Bohr formula, we get the grand energy levels of hydrogen with
fine structure included:
α2
  
13.6 eV n 3
Enj = − 1+ 2 − (6.86)
n2 n j + 1/2 4

The exact fine-structure formula for hydrogen is


 
 !2 −1/2 
2
 α 
Enj = mc 1+ p  −1 (6.87)

 n − (j + 1/2) + (j + 1/2)2 − α2 

Fine structure breaks the degeneracy in l, but it still preserves degeneracy in j. The
stationary states are linear combinations of states with different values of n, l, s, j, and
mj , NOT with ml or ms .

6.4 Hyperfine Structure of Hydrogen


The electron constitutes a magnetic dipole, and so does the proton3 :
e −e e
µp = gp Sp , µe = ge Se ≈ − Se . (6.88)
2mp 2me me

A dipole µ creates a magnetic field:


µ0 2µ0 3
B= 3
[3(µ · r̂)r̂ − µ] + µδ (r). (6.89)
4πr 3
So, the Hamiltonian of the electron in B is
   
0 e µ0 2µ0 3
H = − − Se · [3(µp · r̂)r̂ − µp ] + µp δ (r)
me 4πr3 3
µ0 gp e2 [3(Sp · r̂)(Se · r̂) − Sp · Se ] µ0 gp e2
= + Sp · Se δ 3 (r). (6.90)
8πmp me r3 3mp me
3
The measured value of gp is 5.59. This is far from the electron’s 2.00, because the proton is a
composite structure made up of three quarks.
6.4. HYPERFINE STRUCTURE OF HYDROGEN 89

The first-order correction of the energy En is


µ0 gp e2
 
(1) 3(Sp · r̂)(Se · r̂) − Sp · Se
En,HF =
8πmp me r3
µ0 gp e2
+ hSp · Se i|ψnlm (0)|2 . (6.91)
3mp me
Let us think about the ground state. In any state for which l = 0, including the ground
state, the wave function is spherically symmetrical. Because of this, the first term in
Eq.(6.91) vanishes. Using |ψ100 (0)|2 = 1/(πa3 ), we get

(1) µ0 gp e2
E100,HF = hSp · Se i, (6.92)
3πmp me a3
in the ground state. This is called spin-spin coupling. The total spin S ≡ Se + Sp
commutes with the Hamiltonian, so let us take the eigenvectors of S as our ”good”
unperturbed states.
1
Sp · Se = (S 2 − Se2 − Sp2 ) (6.93)
2
Both the electron and the proton have spin 1/2, so Se2 = Sp2 = (3/4)~2 . The triplet state
has the total spin 1, so S 2 = 2~2 , and the singlet state has the total spin 0, so S 2 = 0.
So, Eq.(6.92) becomes:

4gp ~4

(1) +1/4 (triplet)
E100,HF = × (6.94)
2
3mp me c a2 4 −3/4 (singlet).

Spin-spin coupling breaks the spin degeneracy of the ground state. The energy gap is:
4gp ~4
∆E = = 5.88 × 10−6 eV. (6.95)
3mp m2e c2 a4
When a hydrogen causes a transition from the triplet to the singlet state, it emits a
photon of frequency:
∆E
ν= = 1420 MHz, (6.96)
2π~
1420 MHz corresponds to the wavelength:
c
λ= = 21 cm, (6.97)
ν
which falls in the microwave region. This is the famous 21-centimeter line, which is a
ubiquitous form of radiation in the universe.
90 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

6.5 Zeeman effect


When an atom is placed in a uniform external magnetic field Bext , its energy levels
are shifted. This is called the Zeeman effect. For a single electron of the atom, the
perturbation is

H 0 = −(µl + µs ) · Bext , (6.98)


e e
where µl = − L, µs = − S. (6.99)
2m m
µl is the magnetic dipole moment associated with orbital motion, and µs is the magnetic
dipole moment associated with spin. So,
e
H0 = (L + 2S) · Bext (6.100)
2m
The atom has the internal magnetic field Bint , Eq.(6.70), which causes the spin-orbit
coupling. If Bext  Bint , fine structure dominates, and H 0 in Eq.(6.100) is a small
perturbation. If Bext  Bint , the Zeeman effect dominates, and fine structure is a small
perturbation. If Bext ∼ Bint , we need the full machinery of degenerate perturbation
theory.

6.5.1 Bext  Bint


Fine structure dominates, so our ”good” quantum numbers are n, l, j and mj , but NOT
ml and ms . The first-order Zeeman correction to the energy is
(1) e
EZ = hnljmj |H 0 |nljmj i = Bext · hL + 2Si
2m
e
= Bext · hJ + Si. (6.101)
2m
The total angular momentum J (= L + S) is constant, so L and S precess about the
direction of J , and the time average of S is equal to its projection along J :

(S · J )
Save = J. (6.102)
J2
Using Eq.(6.80), Eq.(6.101) becomes
  
(1) e (S · J )
EZ = Bext · 1+ J
2m J2
 
e j(j + 1) − l(l + 1) + 3/4
= 1+ Bext · hJ i (6.103)
2m 2j(j + 1)
6.5. ZEEMAN EFFECT 91

The term in square brackets is known as the Landé g-factor, gJ . We can choose the
z-axis to lie along Bext , then
(1)
EZ = µB gJ Bext mj , (6.104)

where
e~
µB ≡ = 5.788 × 10−5 eV/T (6.105)
2m
is called the Bohr magneton. The total energy is the sum of Eq.(6.104) and the fine
structure part Eq.(6.86):

α2
  
13.6 eV n 3
En l j mj = − 2
1+ 2 − + µB gJ Bext mj . (6.106)
n n j + 1/2 4

For example, the ground state (n = 1, l = 0, j = 1/2) splits into two levels:

−13.6 eV × (1 + α2 /4) ± µB Bext , (6.107)

with the + sign for mj = 1/2 and the − sign for mj = −1/2.

6.5.2 Bext  Bint : Paschen-Back effect


The Zeeman effect dominates. Let us set the z axis in the direction of Bext . Our ”good”
quantum numbers are n, l, ml , and ms , but NOT j and mj . The Hamiltonian is
e
H0 = Bext (Lz + 2Sz ). (6.108)
2m
The unperturbed energies are

13.6 eV
Enml ms = − + µB Bext (ml + 2ms ), (6.109)
n2
The first-order correction to Enml ms is
(1)
EZ = hnlml ms |(Hr0 + HSO
0
)|nlml ms i, (6.110)

where Hr0 represents the relativistic correction, and HSO 0


represents the spin-orbit cou-
pling. The relativistic correction is the same as before:

En2
 
0 4n
hnlml ms |Hr |nlml ms i = − −3 . (6.111)
2mc2 l + 1/2
92 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

0
HSO is the same as Eq.(6.76). Setting ge ≈ 2,
 2 
0 e 1
HSO = S · L. (6.112)
8π0 m c2 r3
2

Noting that hSx i = hSy i = hLx i = hLy i = 0 for eigenstates of Sz and Lz , we get

hS · Li = hSx ihLx i + hSy ihLy i + hSz ihLz i = hSz ihLz i = ~2 ml ms . (6.113)

Using Eq.(6.82), Eq.(6.111), Eq.(6.112) and Eq.(6.113), Eq.(6.110) becomes


  
(1) 13.6 eV 2 3 l(l + 1) − ml ms
EZ = α − (for l = 1, 2, 3, . . .). (6.114)
n3 4n l(l + 1/2)(l + 1)

Notice that, if l = 0, the term in square brackets is indeterminate.

case of l = 0
If l = 0, then j = s, mj = ms . As our ”good” unperturbed states, we can take the same
states |n ms i for both weak and strong external magnetic field. The fine structure term
is obtained by setting j = 1/2 in the second term of Eq.(6.86):

13.6 eV α2
 
(1) 3
EZ = − n−
n2 n2 4
 
13.6 eV 2 3
= α −1 (for l = 0). (6.115)
n3 4n

This is the same as Eq.(6.114) with the term in square brackets set equal to 1.

The total energy is the sum of the Zeeman part Eq.(6.109) plus the fine structure part
Eq.(6.114) or Eq.(6.115).
6.5. ZEEMAN EFFECT 93

6.5.3 Bext ∼ Bint

We have to treat the fine structure and the Zeeman effect on an equal footing. Let us
examine the case n = 2 in the basis of the states characterized by l, j and mj . Using the
notation |j mj iT = |l ml i|s ms i, we have these eight states:

|ψ1 i ≡ | 12 12 iT = |0 0i | 12 12 i

l = 0 (6.116)
|ψ2 i ≡ | 2 2 i = |0 0i | 12 −1
1 −1 T
2
i
|ψ3 i ≡ | 32 32 iT = |1 1i | 21 12 i


−3
3
|ψ4 i ≡ | 2 2 iT = |1 − 1i | 12 −1




 q 2
i q
= 3 |1 0i | 2 2 i + 13 |1 1i | 21 −1
 3 1 T 2 1 1
 |ψ5 i ≡ | 2 2 i i


 2
q q
l = 1 |ψ6 i ≡ | 1 1 iT = − 13 |1 0i | 12 12 i + 23 |1 1i | 12 −1 i (6.117)

 22 2
 q q
|ψ7 i ≡ | 32 −1 iT = 13 |1 − 1i | 12 12 i + 23 |1 0i | 12 −1 i


2 2



 q q
 |ψ i ≡ | 1 −1 iT = − 2 |1 − 1i | 1 1 i + 1 |1 0i | 1 −1 i

8 2 2 3 22 3 2 2

Setting n = 2 in Eq.(6.85), the first-order fine structure correction to the energy is

2
 
(1) En=2 4×2
En=2,F S
= 3−
2mc2 j + 1/2
 
8
≡ γ 3− , (6.118)
j + 1/2
2 2
En=1 −α2
 
En=2 En=1 13.6 eV 2
where γ = 2
= 2
= = α
2mc 32mc 32 2 64

(1)
For the states of j = 1/2 (= ψ1 , ψ2 , ψ6 , ψ8 ), En=2,F S = −5γ, and for the states of j = 3/2
(1)
(= ψ3 , ψ4 , ψ5 , ψ7 ), En=2,F S = −γ. The fine structure correction yields only the diagonal
elements.
From Eq.(6.100), the Hamiltonian for Zeeman effect is (e/2m)Bext (Lz +2Sz ). ψ1 , ψ2 , ψ3 , ψ4
are eigenstates of Lz and Sz , so there are only diagonal elements for these states:

hHZ0 i = (e/2m)Bext ~(ml + 2ms ) = µB Bext (ml + 2ms ) ≡ β(ml + 2ms )


⇒ hHZ0 i11 = β, hHZ0 i22 = −β, hHZ0 i33 = 2β, hHZ0 i44 = −2β.
94 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

For ψ5 , ψ6 , ψ7 , ψ8 , we need to operate Lz + 2Sz on each state. For ψ5 ,


r    r    
2 1 1 1 1 −1 1 −1
(Lz + 2Sz )|ψ5 i = 0 + 2 ~ |1 0i + ~ + 2~ |1 1i

3 2 22 3 2 2 2
r 
2 1 1
= ~ |1 0i (6.119)
3 22
r r r !
2 2 1
= ~ |ψ5 i − |ψ6 i (6.120)
3 3 3

2 2
= ~|ψ5 i − ~|ψ6 i. (6.121)
3 3

Thus we get the matrix elements hHZ0 i55 = (2/3)β and hHZ0 i56 = −( 2/3)β. In the same
way,
√ √
(Lz + 2Sz )|ψ6 i = − 32 ~|ψ5 i +√31 ~|ψ6 i hHZ0 i65 = −( 2/3)β, hHZ0 i66 = (1/3)β

2 0 0
(Lz + 2Sz )|ψ7 i = − 2√
3
~|ψ 7 i − 3
~|ψ8 i ⇒ hHZ i 77 = −(2/3)β,
√ hH Z i78 = −( 2/3)β
0 0
2 1
(Lz + 2Sz )|ψ8 i = − 3 ~|ψ7 i − 3 ~|ψ8 i hH i
Z 87 = −( 2/3)β, hH i
Z 88 = −(1/3)β

Therefore, with γ ≡ (α/8)2 × 13.6 eV and β ≡ µB Bext , the W -matrix is


 
−5γ + β 0 0 0 0 0 0 0

 0 −5γ − β 0 0 0 0 0 0 


 0 0 −γ + 2β 0 0 0 0 0 


 0 0 0 −γ − 2β 0 0

0 0 

2 2

 0 0 0 0 −γ √+ 3 β −3β 0 0 

2 1
0 0 0 0 − 3 β −5γ + 3 β 0 0
 

 
 2 2 
 0 0 0 0 0 0 −γ √
− 3β −3β 
0 0 0 0 0 0 − 32 β −5γ − 13 β

The first four eigenvalues are obvious; we need to find the eigenvalues of the two 2 × 2
blocks. The characteristic equation for the first of these is

11
λ2 + (6γ − β)λ + (5γ 2 − γβ) = 0. (6.122)
3
Using the quadratic formula, the eigenvalues are
p
λ± = −3γ + β/2 + 4γ 2 + (2/3)γβ + β 2 /4. (6.123)

The eigenvalues of the second block is obtained by reversing the sign of β in Eq.(6.123).
6.5. ZEEMAN EFFECT 95

1 = En=2 − 5γ + β (En=2 = −13.6/4 = −3.4 eV)


2 = En=2 − 5γ − β
3 = En=2 − γ + 2β
4 = En=2 − γ − 2β
p
5 = En=2 − 3γ + β/2 + 4γ 2 + (2/3)γβ + β 2 /4
p
6 = En=2 − 3γ + β/2 − 4γ 2 + (2/3)γβ + β 2 /4
p
7 = En=2 − 3γ − β/2 + 4γ 2 − (2/3)γβ + β 2 /4
p
8 = En=2 − 3γ − β/2 − 4γ 2 − (2/3)γβ + β 2 /4

Figure 6.1: Zeeman splitting of the n = 2 states of hydrogen in the weak, intermediate
and strong field regimes.
96 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

6.6 Stark effect


When an atom is placed in a uniform external electric field, the energy levels are shifted.
Let us study the Stark effect for the n = 1 and n = 2 states of hydrogen. Setting the z
direction in the direction of the electric field, the Hamiltonian is

H 0 = eEext z = eEext r cos θ. (6.124)

6.6.1 case of n = 1

ψ100 = (1/ πa3 )e−r/a , so the first-order correction to the energy is
Z
(1) 1
En=1 = hψ100 |H |ψ100 i = eEext 3 e−2r/a r cos θr2 sin θdrdθdφ
0
πa
Z Z Z
1 −2r/a 3
= eEext 3 e r dr cos θ sin θdθ dφ
πa
= 0, (6.125)

because 0 cos θ sin θdθ = 0. Let us obtain the first-order correction to the wave function
using Eq.(6.12):
(0) (1) (1) (0)
(H (0) − E1 )ψ1 = −(H 0 − E1 )ψ1 , (6.126)
~2 2 e2 1 ~2
 
2
where H (0) = − ∇ − =− 2
∇ + ,
2m 4π0 r 2m ar
(0) ~2 (0) 1 −r/a (1)
E1 = − 2
, ψ1 = √ e , E1 = 0
2ma πa3

(1)
We look for ψ1 that can be written as:
(1)
ψ1 = f (r)e−r/a cos θ = (A + Br + Cr2 )e−r/a cos θ. (6.127)

f e−r/a d
   
2 (1) cos θ d 2 d −r/a
 d
∇ ψ1 = r fe + 2 sin θ (cos θ)
r2 dr dr r sin θ dθ dθ
    
−r/a 00 2 0 1 0 1 1 1
= cos θ e f − f + 2f + 2 f − f − 2f 2 .
a a a r r
Putting this into Eq.(6.126), we get
 
2 0 1 1 4γ
f − f + 2f 0 − 2f 2 =
00
r, (6.128)
a r r a
meEext
where γ ≡ √ . (6.129)
2~2 πa
6.6. STARK EFFECT 97

Putting Eq.(6.127) into this, we get


2 2 2 4γ
2C − (B + 2Cr) + (B + 2Cr) − 2 (A + Br + Cr2 ) = r. (6.130)
a r r a
Collecting like powers of r, we get A = 0, B = −2aγ, C = −γ. So,
(1)
ψ1 = −γr(r + 2a)e−r/a cos θ (6.131)

Using Eq.(6.19), the second-order correction to the energy is4


(2) (0) (1)
En=1 = hψ1 |H 0 |ψ1 i
me2 Eext
2 Z Z
4 −2r/a
= − 2π r (r + 2a)e dr cos2 θ sin θdθ
2πa2 ~2
2
3eEext a2

= −m . (6.133)
2~

6.6.2 case of n = 2
Let us obtain the first-order correction to the energy. There are 4-fold degenerate states:
ψ200 , ψ211 , ψ210 , ψ21−1 :
  −r/2a
|1i = ψ 200 = √1 1
1 − r
e
2πa 2a 2a


1 1 −r/2a
 |2i = ψ = − πa 8a2 re sin θeiφ

211 √
1 1 −r/2a (6.134)

 |3i = ψ210 = √2πa 4a 2 re cos θ
−r/2a
1 1
sin θe−iφ

 |4i = ψ
21−1 = √πa 8a2 re

All matrix elements of H 0 are zero except h1|H 0 |3i and h3|H 0 |1i.
Z Z 
0 eEext 2 r  −r/a 4
h1|H |3i = (2π) cos θ sin θdθ 1 − e r dr
2πa8a3 2a
= −3aeEext . (6.135)

Noting that h3|H 0 |1i = h3|H 0 |1i∗ , we get the W -matrix as


 
0 0 1 0
0 0 0 0
W = −3aeEext  1 0 0 0 .
 (6.136)
0 0 0 0
4
Z ∞
xn e−x/a dx = n! an+1 (6.132)
0
98 CHAPTER 6. TIME-INDEPENDENT PERTURBATION

The characteristic equation is



−λ 0
1 0
0 −λ 0
0
=0 (6.137)
1
0 −λ 0
0 0 0 −λ

−λ 0 0 0 −λ 0

⇒ −λ 0 −λ 0 + 1 0 0 = 0 (6.138)
0 0 −λ 0 0 −λ
⇒ λ2 (λ2 − 1) = 0. (6.139)

Therefore, we get the eigenvalues λ = 0, 0, ±1. So, the energy levels split into three:
E2 , E2 , E2 ± 3aeEext . The eigenvectors are
   
0 0
1 0
  = ψ211 for λ = 0,   = ψ21−1 for λ = 0,
0 0
0 1
 
1
1  0 = √1 (ψ200 + ψ210 ) for λ = +1(→ E2 − 3aeEext )
√ 
2 1
 2
0
 
1
1  0  = √1 (ψ200 − ψ210 ) for λ = −1(→ E2 + 3aeEext ).
√ 
2 −1 2
0

Now, the hydrogen consists a pair of a proton and an electron, it constitutes an electric
dipole moment:

pe = −er = −e(r sin θ cos φx̂ + r sin θ sin φŷ + r cos θẑ) (6.140)

Using Eq.(6.134), the expectation value of pe for the above eigenstates are:

hψ211 |pe |ψ211 i = 0 (6.141)


hψ21−1 |pe |ψ211 i = 0 (6.142)
√ √
h(1/ 2)(ψ200 + ψ210 )|pe |(1/ 2)(ψ200 + ψ210 )i = 3aeẑ (6.143)
√ √
h(1/ 2)(ψ200 − ψ210 )|pe |(1/ 2)(ψ200 − ψ210 )i = −3aeẑ (6.144)

The two states that split from E2 carry an electric dipole moment, and it does not depend
on the applied field Eext .
6.7. VAN DEL WAALS INTERACTION 99

6.7 Van del Waals Interaction


Two atoms are placed at a distance R apart. They are electrically neutral on the whole,
so there would be no force between them. But, if they are polarizable there is a weak
attraction, called the Van del Waals interaction. Let us use a crude model in which
each atom is an electron (mass m, charge −e) attached by a spring to the nucleus (charge
+e), and assume that the nuclei are so heavy that they are motionless. The Hamiltonian
for the unperturbed system is:

1 2 1 2 1 2 1 2
H (0) = p1 + kx1 + p + kx . (6.145)
2m 2 2m 2 2 2
The Coulomb interaction between the atoms is
 2
e2 e2 e2

0 1 e
H = − − + . (6.146)
4π0 R R + x1 R − x2 R + x1 − x2

Assuming |x1 |  R and |x2 |  R,


( " 2 #)
1 e2
  x 2    x 2    
0
x 
1 1
x 
2 2 x1 − x2 x1 − x 2
H ≈ 1− 1− + − 1+ + + 1− +
4π0 R R R R R R R
1 e2 e 2 x1 x2
 
2x1 x2
= − 2 =− .
4π0 R R 2π0 R3

Let us change the variables:

1 1
x± ≡ √ (x1 ± x2 ), p± ≡ √ (p1 ± p2 ), (6.147)
2 2

Then the total Hamiltonian becomes

1 2 1 e2
H (0) + H 0 = (p1 + p22 ) + k(x21 + x22 ) − (2x1 x2 )
2m 2 4π0 R3
1 2 1 e2
= (p+ + p2− ) + k(x2+ + x2− ) − (x2 − x2− )
2m 2 4π0 R3 +
e2 e2
       
1 2 1 2 1 2 1
= p + k− x+ + p + k+ x2−
2m + 2 4π0 R3 2m − 2 4π0 R3

Let us consider the ground state. The energy for this Hamiltonian is
r
1 k ∓ (e2 /4π0 R3 )
E = ~(ω+ + ω− ), where ω± = . (6.148)
2 m
100 CHAPTER 6. TIME-INDEPENDENT PERTURBATION
p
Without the Coulomb interaction, the energy would have been E0 = ~ k/m = ~ω0 .
Assuming that k  (e2 /4π0 R3 ),
r  1/2
k e2
ω± = 1∓
m 4π0 R3 k
" 2 #
e2 e2
  
1 1
≈ ω0 1 ∓ − (6.149)
2 4π0 R3 mω02 8 4π0 R3 mω02
So, the energy shift is
1
∆E = E − E0 = ~(ω+ + ω− ) − ~ω0
2
 2 2
1 ~ e 1
= − 2 3
, (6.150)
8 m ω0 2π0 R6
which means there is an attractive potential ∝ R−6 between the two atoms. Let us
confirm this result with the perturbation theory. The first-order correction to the energy
is
e2
E (1) = h0|H 0 |0i = − hψ0 (x1 )ψ0 (x2 )|x1 x2 |ψ0 (x1 )ψ0 (x2 )i
2π0 R3
e2
= − hψ0 (x1 )|x1 |ψ0 (x1 )ihψ0 (x2 )|x2 |ψ0 (x2 )i
2π0 R3
= 0. (6.151)
The second-order correction is, from Eq.(6.24),

(2)
X |hψn |H 0 |ψ0 i2
E = (6.152)
n=1
E0 − En
where |ψ0 i = |0i|0i and |ψn i = |n1 i|n2 i. So,
2 X ∞ X ∞
e2 |hn1 |x1 |0i|2 |hn2 |x2 |0i|2

(2)
E =
2π0 R3 n =1 n =1 E00 − En1 n2
1 2
2
e2 |h1|x1 |0i|2 |h1|x2 |0i|2

=     (6.153)
2π0 R3 1 1 3 3
~ω0 + ~ω0 − ~ω0 + ~ω0
2 2 2 2
2 2
e2
     
1 ~
= 3
− (6.154)
2π0 R 2~ω0 2mω0
 2 2
~ e 1
= − 2 3 (6.155)
8m ω0 2π0 R6
This is the same as Eq.(6.150).
Chapter 7

Variational Principle

The variational principle can get us an upper bound for the ground state energy
of a system whose Schrödinger equation we cannot solve.
Any function can be expressed as a linear combination of the unknown eigenfunctions of
the Hamiltonian H:
X
ψ= cn ψn , with Hψn = En ψn . (7.1)
n

ψ is normalized:
* +
X X XX X
1 = hψ|ψi = cm ψ m cn ψn = c∗m cn hψm |ψn i = |cn |2 .


m n m n n

The expectation value of H in the state ψ is:


* +
X X XX X
hHi = cm ψm H cn ψn = c∗m En cn hψm |ψn i = En |cn |2 .


m n m n n

The ground state energy is the smallest eigenvalue, so En ≥ Egs . Then,


X X
hHi ≡ hψ|H|ψi ≥ Egs |cn |2 = Egs |cn |2 = Egs . (7.2)
n n

So, if you want to know the upper bound for the ground state energy, pick any nor-
malized function ψ, and calculate hψ|H|ψi.

7.1 1D harmonic oscillator


~2 d2 1
H=− 2
+ mω 2 x2 . (7.3)
2m dx 2
101
102 CHAPTER 7. VARIATIONAL PRINCIPLE

Let us pick the Gaussian as our trial wave function:


2
ψ(x) = Ae−bx , (7.4)
where b is a constant, and A is determined by normalization:
Z ∞ r  1/4
2 −2bx2 2 π 2b
1 = |A| e dx = |A| ⇒ A= (7.5)
−∞ 2b π
So we calculate
~2 2 
Z Z
−bx2 d 1

−bx2 2
hψ|H|ψi = − |A|2
e 2
e 2
dx + mω |A|2
e−2bx x2 dx
2m dx 2
~2 b mω 2
= + . (7.6)
2m 8b
This exceeds Egs for any b. So, to get the tightest upper bound,
d ~2 mω 2 mω
hHi = − 2
=0 ⇒ b= . (7.7)
db 2m 8b 2~
Putting this into Eq.(7.6), we find
1
hHimin = ~ω (7.8)
2

7.2 Ground State of Helium


The helium atom consists of two electrons around a nucleus containing two protons (and
neutrons). The Hamiltonian is
~2 e2
 
2 2 2 2 1
H=− (∇ + ∇2 ) − + − (7.9)
2m 1 4π0 r1 r2 |r1 − r2 |
If we ignore the electron-electron repulsion term, H splits into two hydrogen Hamiltonians
with a nuclear charge of 2e, and the exact solution is:
8 −2(r1 +r2 )/a
ψ0 (r1 , r2 ) = ψ100 (r1 )ψ100 (r2 ) = e (7.10)
πa3
But each electron forms a cloud of negative charge that somewhat shields the nucleus,
and the effective nuclear charge that the other electron actually sees should be less than
2. So, let us take a trial function like this1 :
Z 3 −Z(r1 +r2 )/a
ψ1 (r1 , r2 ) = e . (7.11)
πa3
1
For a nucleus with atomic number Z, we need to substitute En → Z 2 En and a → a/Z.
7.2. GROUND STATE OF HELIUM 103

Let us rewrite Eq.(7.9):


~2 e2 e2
   
Z Z Z −2 Z −2 1
H=− (∇21 + ∇22 ) − + + + + .
2m 4π0 r1 r2 4π0 r1 r2 |r1 − r2 |
The expectation value of H is
e2
  
2 1
hHi = 2Z E1 + 2(Z − 2) + hVee i, (7.12)
4π0 r
where E1 = −13.6 eV, and h1/ri = Z/a, from Eq.(6.62). hVee i is
 2   3 2 Z −2Z(r1 +r2 )/a
e Z e
hVee i = 3
d3 r1 d3 r2 (7.13)
4π0 πa |r1 − r2 |
Let us do the r2 integral first. Setting the z (polar) axis along r1 ,
q
|r1 − r2 | = r12 + r22 − 2r1 r2 cos θ2 . (7.14)

Therefore
e−2Zr2 /a 3 e−2Zr2 /a
Z Z
I2 ≡ d r2 = p r22 sin θ2 dr2 dθ2 dφ2 (7.15)
|r1 − r2 | 2 2
r1 + r2 − 2r1 r2 cos θ2
The φ2 integral becomes 2π. The θ2 integral is
Z π p π
sin θ2 r12 + r22 − 2r1 r2 cos θ2
p dθ2 =
r1 r2

0 r12 + r22 − 2r1 r2 cos θ2
q 0
1
q
= r12 + r22 + 2r1 r2 − r12 + r22 − 2r1 r2 (7.16)
r1 r2

1 2/r1 (r2 < r1 )
= {(r1 + r2 ) − |r1 − r2 |} = (7.17)
r1 r2 2/r2 (r2 > r1 )
So,
Z ∞
1 r1 −2Zr2 /a 2
 Z 
−2Zr2 /a
I2 = 4π e r2 dr2 + e r2 dr2 (7.18)
r1 0 r1
πa3
   
Zr1 −2Zr1 /a
= 1− 1+ e (7.19)
r1 Z 3 a
Putting Eq.(7.19) into Eq.(7.13),
 2   3  Z −2Zr1 /a    
e Z e Zr1 −2Zr1 /a 3
hVee i = 1− 1+ e d r1
4π0 πa3 r1 a
104 CHAPTER 7. VARIATIONAL PRINCIPLE

The θ and φ integrals become 4π, and the r1 integral is


Z ∞
Zr12 −4Zr1 /a 5 a2
  
−2Zr1 /a
r1 e − r1 + e dr1 = (7.20)
0 a 32 Z 2

So finally we get

e2
 3 
5 a2

Z
hVee i = 4π
4π0 πa3 32 Z 2
5Z e2 5
= = − ZE1 . (7.21)
8a 4π0 4

So, Eq.(7.12) becomes


   
2 5 2 27
hHi = 2Z − 4Z(Z − 2) − Z E1 = −2Z + Z E1 . (7.22)
4 4

For any value of Z, this value exceeds the ground state energy. Let us minimize hHi:
 
d 27 27
hHi = −4Z + E1 = 0 ⇒ Z =
dZ 4 16
6
3
⇒ hHi = 7 E1 = −77.5 eV. (7.23)
2
Here we are within 2% of −78.975 eV, which is the experimentally measured value of the
ground state energy of helium.

7.3 Hydrogen Molecule Ion


Let us think about H2+ , which consists of a single electron in the Coulomb field of two
protons.

~2 2 e2
 
1 1
H=− ∇ − + , (7.24)
2m 4π0 r1 r2

where r1 and r2 are the distances to the electron from the respective protons. If the
distance R between the two protons is far greater than the Bohr radius a, the electron
has the same probability of being associated with either proton. So, let the trial function
be
1
ψ = A[ψ0 (r1 ) + ψ0 (r2 )], where ψ0 (r) = √ e−r/a (7.25)
πa3
7.3. HYDROGEN MOLECULE ION 105

We need to normalize ψ:
Z Z Z 
2 2 3 2 3 3
1 = |A| |ψ0 (r1 )| d r + |ψ0 (r2 )| d r + 2 ψ0 (r1 )ψ0 (r2 )d r . (7.26)

The first two terms are 1. The third term is


Z Z
1
I ≡ ψ0 (r1 )ψ0 (r2 )d r = 3 e−(r1 +r2 )/a d3 r.
3
(7.27)
πa
Let us pick the coordinates in which proton 1 is at the origin and proton 2 is on the z
axis at the point R. Then

r1 = r, and r2 = r2 + R2 − 2rR cos θ

Z
1 −r/a − r2 +R2 −2rR cos θ/a 2
⇒ I = e e r sin θdrdθdφ (7.28)
πa3
The φ integral is just 2π. To do the θ integral, let us set a variable:

y ≡ r2 + R2 − 2rR cos θ
1
⇒ d(y 2 ) = 2ydy = 2y y −1 2rR sin θdθ = 2rR sin θdθ (7.29)
2
So,
Z π √
Z r+R
− r2 +R2 −2rR cos θ/a 1
e sin θdθ = e−y/a ydy
0 rR |r−R|
a  −(r+R)/a
(r + R + a) − e−|r−R|/a (|r − R| + a) .

=− e (7.30)
rR
Thus, Eq.(7.28) becomes
 Z ∞ Z R
2 −R/a −2r/a −R/a
I = 2 −e (r + R + a)e rdr + e (R − r + a)rdr
aR 0 0
Z ∞ 
R/a −2r/a
+e (r − R + a)e rdr
R
"    2 #
R 1 R
= e−R/a 1 + + (7.31)
a 3 a

I is called the overlap integral — I → 1 as R → 0, and I → 0 as R → ∞. Now the


normalization factor A in Eq.(7.26) can be written as
1
|A|2 = . (7.32)
2(1 + I)
106 CHAPTER 7. VARIATIONAL PRINCIPLE

Let us calculate hHi in the state ψ in Eq.(7.25). Note that


~2 2 e2 1
 
− ∇ − ψ0 (r1 ) = E1 ψ0 (r1 ), (7.33)
2m 4π0 r1
where E1 = −13.6 eV, and the same equation holds for r2 , too. Then,
~2 2 e2
  
1 1
Hψ = A − ∇ − + [ψ0 (r1 ) + ψ0 (r2 )]
2m 4π0 r1 r2
 2  
e 1 1
= E1 A[ψ0 (r1 ) + ψ0 (r2 )] − A ψ0 (r1 ) + ψ0 (r2 ) (7.34)
4π0 r2 r1
So,
e2
  
2
1 1
hHi = E1 − 2|A| hψ0 (r1 ) ψ0 (r1 )i + hψ0 (r1 ) ψ0 (r2 )i ,

4π0 r2 r1
where

1 a  a  −2R/a
D ≡ a hψ0 (r1 ) ψ0 (r1 )i = − 1 + e (7.35)
r2 R R
is called the direct integral, and
 
1 R −R/a
X ≡ a hψ0 (r1 ) ψ0 (r2 )i = 1 +
e (7.36)
r1 a
is called the exchange integral. So, recalling E1 = −(e2 /4π0 )(1/2a), we get
  
D+X
hHi = 1 + 2 E1 (7.37)
1+I
We have only considered the electron so far. For the whole H2+ , we need to add the
proton-proton repulsion:
e2 1 2a
Vpp = = − E1 . (7.38)
4π0 R R
So, the total energy of the ground state Egs of H2+ is smaller than Eq.(7.37) + Eq.(7.38).
As a function of x ≡ R/a, we get
2 (1 − (2/3)x2 )e−x + (1 + x)e−2x
 
Egs
≤ −1 + ≡ F (x). (7.39)
13.6 eV x 1 + (1 + x + (1/3)x2 )e−x
If F (x) goes below −1, then H2+ can have a bonding state, because Egs < −13.6 eV
(= the energy of a free proton and the ground state energy of a neutral hydrogen atom).
Actually, F (x) reaches its minimum at R = 2.4 a = 1.3Å, where the binding energy is
1.8 eV. The experimental values are 1.06Å and 2.8 eV.
Chapter 8

WKB Approximation

The WKB1 method is a technique for obtaining approximate solutions to the time-
independent Schrödinger equation in 1D (also applicable to the radial part of the Schrödinger
equation in 3D). It is particularly useful when we want to know tunneling rates through
potential barriers as well as bound state energies.

8.1 Classical Region: E > V (x)


Let us consider the region of x such that E > V (x). The Schrödinger equation:
~ 2 d2 ψ
− + V (x)ψ = Eψ
2m dx2
d2 ψ p2 p
⇒ 2
= − 2
ψ, where p(x) ≡ 2m[E − V (x)]. (8.1)
dx ~
This region is called the classical region — a classical particle is confined to this range
of x. Let us express ψ as a general complex function:
ψ(x) = A(x)eiφ(x) , (8.2)
where A(x) and φ(x) are real functions. Doing some math:

= (A0 + iAφ0 )eiφ (8.3)
dx
d2 ψ
= [A00 + 2iA0 φ0 + iAφ00 − A(φ0 )2 ]eiφ (8.4)
dx2
and putting this into Eq.(8.1), we get
p2
A00 + 2iA0 φ0 + iAφ00 − A(φ0 )2 = − A. (8.5)
~2
1
Wentzel, Kramers, and Brillouin

107
108 CHAPTER 8. WKB APPROXIMATION

Taking the real part and imaginary part separately, we get two equations:

p2 p2
 
00 0 2 00 0 2
A − A(φ ) = − 2 A ⇒ A = A (φ ) − 2 (8.6)
~ ~
0
2A0 φ0 + Aφ00 = 0 ⇒ A2 φ0 = 0

(8.7)

From Eq.(8.7),
const.
A2 φ0 = const. ⇒ A = √ 0 , (8.8)
φ

Now let us assume that the amplitude A(x) varies very slowly, so that A00 ≈ 0. Then,
from Eq.(8.6),

p2 1 x
Z
0 2 0 p
(φ ) = 2 ⇒ φ = ± ⇒ φ(x) = ± p(x0 )dx0 . (8.9)
~ ~ ~
Putting Eq.(8.8) and Eq.(8.9) into Eq.(8.2), ψ(x) can be written as
C i
Rx
p(x0 )dx0
ψ(x) ≈ p e± ~ (E > V (x)) (8.10)
p(x)

where C is a constant. The general solution is a linear combination of these two solutions.
Note that |ψ(x)|2 ≈ |C|2 /p(x) — the probability of finding the particle at point x is small
when its (classical) momentum is large at that point.

8.1.1 Potential well with two vertical walls


Let us consider the potential:

a certain function f (x) (0 < x < a)
V (x) = , (8.11)
∞ otherwise

and assume E > V (x). Inside the well,


1 
C+ eiφ(x) + C− e−iφ(x)

ψ(x) ≈ p
p(x)
1
or ψ(x) ≈ p [C1 sin φ(x) + C2 cos φ(x)], (8.12)
p(x)

where
Z x
1
φ(x) = p(x0 )dx0 . (8.13)
~ 0
8.2. NON-CLASSICAL REGION: E < V (X) 109

Boundary condition 1: ψ(x) = 0 at x = 0. φ(0) = 0, so C2 = 0.


Boundary condition 2: ψ(x) = 0 at x = a. So, φ(a) = nπ.
Therefore,
Z a
p(x)dx = nπ~, (8.14)
0
p
where n = 1, 2, 3, . . ., because
√ p(x) = 2m[E − V (x)] > 0.
If V (x) = 0, then p(x) = 2mE. In this case, from Eq.(8.14),

n2 π 2 ~ 2
pa = nπ~ ⇒ En = . (8.15)
2ma2
The WKB approximation yields the exact answer in this case — the amplitude of the
true wave function is constant, so dropping A00 cost us nothing.

8.2 Non-classical region: E < V (x)


In the non-classical region (E < V (x)), p(x) in Eq.(8.1) is imaginary. We can simply
change Eq.(8.10) to:

C 1 x
R 0 0
ψ(x) ≈ p e± ~ |p(x )|dx (E < V (x)) (8.16)
|p(x)|

8.3 Connection formulas


Let us consider the turning point where E = V , where the classical region joins the
non-classical region. For simplicity, we set the right-hand turning point at x = 0. The
WKB approximation says
 h i R0 i 0
i
0 0 0 0
R
 √1 Be ~ x p(x )dx + Ce− ~ x p(x )dx (x < 0)
p(x)
ψ(x) ≈ 1
Rx 0 0 (8.17)
 √ 1 De− ~ 0 |p(x )|dx (x > 0),
|p(x)|

where we excluded the positive exponent in x > 0, assuming that V (x) > E for all x > 0.
We need to do splice the two WKB solutions together, using a patching wave function
ψp that straddles the turning point. Since we need ψp only in the neighborhood of the
origin, let’s approximate the potential as

V (x) ≈ E + V 0 (0)x. (8.18)


110 CHAPTER 8. WKB APPROXIMATION

Figure 8.1: Right-hand turning point.

The Schrödinger equation in the patching region is


~2 d2 ψp
− + [E + V 0 (0)x]ψp = Eψp
2m dx2
1/3
d2 ψp

2m 0
⇒ = α3 xψp , where α ≡ V (0) . (8.19)
dx2 ~2
Eq.(8.19) can be rewritten as
d2 ψp
= zψp (8.20)
dz 2
The solutions to Eq.(8.20) are Airy functions. There are two linearly independent Airy
functions, Ai(z) and Bi(z). So, ψp is a linear combination of Ai(z) and Bi(z):
ψp (x) = aAi(αx) + bBi(αx), (8.21)
where a and b are constants. In the overlap regions Eq.(8.18) holds, so
p √
p(x) ≈ 2m(E − E − V 0 (0)x) = ~α3/2 −x. (8.22)
In overlap region 2 (See Fig.8.4),
Z x
0 0
Z x √ 2
|p(x )|dx ≈ ~α 3/2
x0 dx0 = ~(αx)3/2 . (8.23)
0 0 3
So the WKB wave function Eq.(8.17) becomes
D 2 3/2
ψ(x) ≈ √ e− 3 (αx) . (8.24)
3/4
~α x 1/4
8.3. CONNECTION FORMULAS 111

Figure 8.2: Properties of Airy functions.

Figure 8.3: Graph of Airy functions.


112 CHAPTER 8. WKB APPROXIMATION

Figure 8.4: Patching region and the two overlap zones.

Using the large-z asymptotic forms of the Airy functions,

a 2 3/2 b 2 3/2
ψp (x) ≈ √ 1/4
e− 3 (αx) + √ 1/4
e 3 (αx) (8.25)
2 π(αx) π(αx)

Comparing Eq.(8.24) and Eq.(8.25), we get


r

a= D, b=0 (8.26)
α~

Now we repeat the same procedure for overlap region 1. In this region x is negative, so
Z 0
2
p(x0 )dx0 ≈ ~(−αx)3/2 (8.27)
x 3

From Eq.(8.17), the WKB wave function is

1 h 2 3/2 2 3/2
i
ψ(x) ≈ √ Bei 3 (−αx) + Ce−i 3 (−αx) (8.28)
~α3/4 (−x)1/4

Using the asymptotic form of the Airy function for large negative z, the patching function
Eq.(8.21) is
 
a 2 3/2 π
ψp (x) ≈ √ sin (−αx) +
π(−αx)1/4 3 4
a 1 h iπ/4 i 2 (−αx)3/2 −iπ/4 −i 32 (−αx)3/2
i
= √ e e 3 − e e (8.29)
π(−αx)1/4 2i
8.3. CONNECTION FORMULAS 113

Comparing this with Eq.(8.28), we get

a B −a −iπ/4 C
√ eiπ/4 = √ , and √ e =√
2i π ~α 2i π ~α
iπ/4 −iπ/4
⇒ B = −ie D, and C = ie D (8.30)

Let us express everything in terms of D and shift the turning point back from the origin
to an arbitrary point x2 , the WKB wave function Eq.(8.17) becomes
 Z x2 
2D 1 π

0 0

 p sin p(x )dx + (x < x2 )

p(x) ~ x 4
ψ(x) ≈ (8.31)
1 x
Z
D 0 0
exp − |p(x )|dx (x > x )

2

 p
|p(x)| ~ x2

8.3.1 Potential well with one vertical wall


Let us consider the potential:
( 1
mω 2 x2 (x > 0)
V (x) = 2 (8.32)
∞ (x < 0)

In this case,
r
1 2E
p q
p(x) = 2m[E − (1/2)mω x ] = mω x22 − x2 , where x2 =
2 2 (8.33)
ω m
The turning point is x2 . So,
Z x2 Z x2
π πE
q
p(x)dx = mω x22 − x2 dx = mωx22 = . (8.34)
0 0 4 2ω

We have the boundary condition ψ(0) = 0, so

1 x2
Z
π
p(x)dx + = nπ (n = 1, 2, 3, . . .)
~ 0 4
Z x2  
1
⇒ p(x)dx = n− π~. (8.35)
0 4

From Eq.(8.34) and Eq.(8.35),


   
1 3 7 11
En = 2n − ~ω = , , , . . . ~ω (8.36)
2 2 2 2
114 CHAPTER 8. WKB APPROXIMATION

8.3.2 potential well with no vertical walls


Eq.(8.31) connects the WKB wave functions at a turning point where the potential slopes
upward. In the same way, we can get the connection formulas for a downward-sloping
turning point:

D0 1 x1
  Z 
0 0


 p exp − |p(x )|dx (x < x1 )
|p(x)|  Z ~ x
ψ(x) ≈ (8.37)
2D0 1 x

0 0 π
sin p(x )dx + (x > x1 )


 p
p(x) ~ x1 4

Suppose we have a potential well such that V (x) < E for (x1 < x < x2 ). Inside the well,
the wave function can be written in two ways, from Eq.(8.31) and Eq.(8.37):

1 x2
Z
2D π
ψ(x) ≈ p sin θ2 (x), where θ2 (x) ≡ p(x0 )dx0 + (8.38)
p(x) ~ x 4
−2D0 1 x
Z
π
ψ(x) ≈ p sin θ1 (x), where θ1 (x) ≡ − p(x0 )dx0 − (8.39)
p(x) ~ x1 4

So, θ2 = θ1 + nπ, from which we get


Z x2  
1
p(x)dx = n − π~ (n = 1, 2, 3, . . .) (8.40)
x1 2

This quantization condition determines the allowed energies for a potential well with two
sloping sides. In summary,
Z x2
p(x)dx = nπ~ (for potential well with TWO vertical walls) (8.41)
x1
Z x2  
1
p(x)dx = n− π~ (for potential well with ONE vertical wall) (8.42)
x1 4
Z x2  
1
p(x)dx = n− π~ (for potential well with NO vertical walls), (8.43)
x1 2

where n = 1, 2, 3, . . . .
The WKB approximation works best in the semi-classical (large n) regime.
8.4. TUNNELING 115

8.4 Tunneling
Let us consider scattering from a barrier with sloping walls. The wave function can be
written in the form:
i x1
   Z   Z x1 
1 0 0 i 0 0
A exp − p(x )dx + B exp p(x )dx (x < x1 )


 p
p(x) ~ ~


  Z xx x
1 x
    Z 
 1 1 0 0 0 0
ψ(x) = p C exp |p(x )|dx + D exp − |p(x )|dx (x1 < x < x2 )

 |p(x)| ~ x1 ~ x1
i x
 Z 
1


F exp p(x0 )dx0 (x2 < x).



 p
p(x) ~ x2

8.4.1 Around x ' x1


At x1 , we have an upward-sloping turning point. Shifting the origin to x1 ,

i 0
  Z   Z 0 
1 0 0 i 0 0
A exp −



 p p(x )dx + B exp p(x )dx (x < 0)
p(x)  ~ x ~ x
ψ(x) = (8.44)
1 x 1 x
 Z   Z 
1 0 0 0 0
C exp |p(x )|dx + D exp − |p(x )|dx (0 < x)


 p
|p(x)| ~ 0 ~ 0

overlap region 2
From Eq.(8.24) and Eq.(8.25),

1 h 2 3/2 2 3/2
i
ψWKB (x) ≈ √ Ce 3 (αx) + De− 3 (αx)
~α3/4 x1/4
a 2 3/2 b 2 3/2
ψp (x) ≈ √ 1/4
e− 3 (αx) + √ 1/4
e 3 (αx) .
2 π(αx) π(αx)
r r
π π
⇒ a = 2D , b=C (8.45)
α~ α~

Figure 8.5: Barrier with sloping walls.


116 CHAPTER 8. WKB APPROXIMATION

overlap region 1
From Eq.(8.28) and Eq.(8.29) with b 6= 0,
1 h
−i 23 (−αx)3/2 i 32 (−αx)3/2
i
ψWKB (x) ≈ √ Ae + Be
~α3/4 (−x)1/4
   
a 2 3/2 π b 2 3/2 π
ψp (x) ≈ √ sin (−αx) + +√ cos (−αx) +
π(−αx)1/4 3 4 π(−αx)1/4 3 4
1 h 2 3/2 2 3/2
i
= √ 1/4
(−ia + b)ei 3 (−αx) eiπ/4 + (ia + b)e−i 3 (−αx) e−iπ/4
2 π(−αx)
r   r  
~α ia + b −iπ/4 ~α −ia + b iπ/4
⇒ A= e , B= e (8.46)
π 2 π 2
Putting Eq.(8.45) into Eq.(8.46),
   
C C
A= + iD e−iπ/4 , B= − iD eiπ/4 (8.47)
2 2

8.4.2 Around x ' x2


At x2 , we have a downward-sloping turning point. Let us rewrite ψ(x) for (x1 < x < x2 )
as
 Z x2
1 x
 Z 
1 1 0 0 0 0
ψ(x) = p C exp |p(x )|dx + |p(x )|dx
|p(x)| ~ x1 ~ x2
1 x2 1 x
 Z Z 
0 0 0 0
+D exp − |p(x )|dx − |p(x )|dx (8.48)
~ x1 ~ x2
Let us introduce the variables:
1 x2
Z
γ≡ |p(x)|dx, C 0 ≡ De−γ , and D0 ≡ Ceγ . (8.49)
~ x1
Then, shifting the origin to x2 ,
  Z 0
1 0
   Z 
1 0 1 0 0 0 0 0
|p(x )|dx + D exp − |p(x )|dx



 p C exp (x < 0)
|p(x)| ~ x ~ x
ψWKB (x) =
i x
 Z 
1
F exp p(x0 )dx0 (0 < x).


 p
p(x) ~

0

In the patching region,


1/3
2m|V 0 (0)|

ψp (x) = aAi(−αx) + bBi(−αx), where α ≡ (8.50)
~2

and p(x) = ~α3/2 x. (8.51)
8.4. TUNNELING 117

overlap region 1 (x < 0)

Z 0
2
|p(x0 )|dx0 = ~(−αx)3/2 (8.52)
x 3
So,
1 h 2 3/2 2 3/2
i
ψWKB (x) ≈ √ C 0 e 3 (−αx) + D0 e− 3 (−αx)
~α3/4 (−x)1/4
a 2 3/2 b 2 3/2
ψp (x) ≈ √ 1/4
e− 3 (−αx) + √ 1/4
e 3 (−αx)
2 π(−αx) π(−αx)
r r
π 0 π 0
⇒ a=2 D, b= C (8.53)
~α ~α

overlap region 2 (x > 0)


Z x
2
p(x0 )dx0 = ~(αx)3/2 (8.54)
0 3
So,
1 2 3/2
ψWKB (x) ≈ √ F ei 3 (αx)
3/4
~α x 1/4
   
a 2 3/2 π b 2 3/2 π
ψp (x) ≈ √ sin (αx) + +√ cos (αx) +
π(αx)1/4 3 4 π(αx)1/4 3 4
1 h
i 32 (αx)3/2 iπ/4 −i 32 (αx)3/2 −iπ/4
i
= √ (−ia + b)e e + (ia + b)e e
2 π(αx)1/4
r   r
~α −ia + b iπ/4 ~α iπ/4
⇒ ia + b = 0, F = e =b e
π 2 π
r r
π −iπ/4 π −iπ/4
⇒ b= e F, a = i e F (8.55)
~α ~α

From Eq.(8.55) and Eq.(8.53),


r r
0 ~α 1 ~α i
C = b = e−iπ/4 F, 0
D = a = e−iπ/4 F (8.56)
π 2 π 2
Using this and Eq.(8.49),
i
D = eγ C 0 = eγ e−iπ/4 F, C = e−γ D0 = e−γ e−iπ/4 F (8.57)
2
118 CHAPTER 8. WKB APPROXIMATION

8.4.3 tunneling probability


From Eq.(8.47) and Eq.(8.57),
     −γ 
C −iπ/4 i −γ −iπ/4 γ −iπ/4 −iπ/4 e γ
A = + iD e = e e F + ie e F e = +e F
2 4 4
So, the tunneling probability is
−2
|F |2 e−γ e−2γ

T = = + eγ ="  −γ 2 #2 , (8.58)
|A|2 4 e
1+
2
Z x2
1 p
where γ ≡ |p(x)| dx, p(x) = 2m[E − V (x)]. (8.59)
~ x1

In the case of a broad, high barrier (γ  1), the tunneling probability is T = e−2γ .

8.4.4 alpha decay


Let us see a model (Fig. 8.6) to explain alpha decay (= the spontaneous emission of an
alpha-particle by radioactive nuclei). The alpha-particle has a positive charge 2e. Let E
be the energy of the emitted alpha particle and Ze be the charge of the leftover nucleus.
The point r2 is determined by
1 2Ze2
= E. (8.60)
4π0 r2
The factor γ in Eq.(8.59) is
Z r2 s √
2mE r2 r2
  Z r
1 1 2Ze2
γ = 2m − E dr = − 1 dr
~ r1 4π0 r ~ r1 r
√   r  p 
2mE π −1 r1
= r2 − sin − r1 (r2 − r1 ) (8.61)
~ 2 r2
Typically r1  r2 , so we can approximate Eq.(8.61) as

2mE h π √ i Z p
γ ≈ r2 − 2 r1 r2 = K1 √ − K2 Zr1 , (8.62)
~ 2 E
 2  √
e π 2m
where K1 ≡ = 1.980 MeV1/2 , (8.63)
4π0 ~
 2 1/2 √
e 4 m
K2 ≡ = 1.485 fm−1/2 (8.64)
4π0 ~
8.4. TUNNELING 119

Figure 8.6: Model for the potential energy of an alpha particle in a radioactive nucleus.

Let us imagine the alpha particle rattling around inside the nucleus, with an average
velocity v. The number of collisions of the alpha particle with the wall of the nucleus
is v/2r1 [s−1 ]. The probability of escape at each collision is e−2γ , so the probability of
emission is (v/2r1 )e−2γ [s−1 ]. So, the lifetime of the parent nucleus is about
2r1 2γ
τ= e . (8.65)
v
This formula does not give us accurate values, but notice log τ ∝ E −1/2 .


Figure 8.7: Half life (τ1/2 = τ ln 2 versus 1/ E for uranium and thorium (E is the energy
of the emitted alpha particle.)
120 CHAPTER 8. WKB APPROXIMATION
Chapter 9

Adiabatic Approximation

9.1 Adiabatic Theorem


9.1.1 Adiabatic Processes
Imagine a pendulum oscillating back and forth in a vertical plane. If you move the support
of the pendulum very gently and steadily, the pendulum will continue to swing in the
same plane with the same amplitude. This gradual change of the external conditions
defines an adiabatic process. There are two characteristic times involved here: the
period of the pendulum’s oscillations, and the period of the platform’s motion if the
pendulum is mounted on a vibrating platform. We call the former the internal time Ti
(representing the motion of the system itself) and the latter the external time Te (over
which the parameters of the system change appreciably). An adiabatic process is one for
which Te  Ti .
In quantum mechanics, the essential content of the adiabatic approximation can be
expressed in a theorem. Suppose the Hamiltonian changes gradually from some initial
form H i to some final form H f . The adiabatic theorem states that if the particle
was initially in the n-th eigenstate of H i , it will be carried into the n-th eigenstate of
H f . Unlike the perturbation theory, the change in the Hamiltonian can be huge in an
adiabatic process — all we require is that the change happen slowly. Energy is not
conserved here. Whatever is changing the Hamiltonian can extract or add the energy.

9.1.2 Proof of the Adiabatic Theorem


If the Hamiltonian is independent of time, a particle that starts out in the n-th eigenstate:
Hψn = En ψn (9.1)
remains in the n-th eigenstate, only picks up a phase factor:
Ψn (t) = ψn e−iEn t/~ . (9.2)

121
122 CHAPTER 9. ADIABATIC APPROXIMATION

If the Hamiltonian changes with time, Eq.(9.1) is time-dependent:

H(t)ψn (t) = En (t)ψn (t). (9.3)

The wave functions still constitute an orthonormal set at any particular instant:

hψn (t)|ψm (t)i = δnm . (9.4)

Also they are complete, so the general solution to the time-dependent Schrödinger equa-
tion

i~ Ψ(t) = H(t)Ψ(t) (9.5)
∂t
can be expressed as a linear combination of them:
Z t
X 1
Ψ(t) = cn (t)ψn (t)e iθn (t)
, where θn (t) ≡ − En (t0 )dt0 (9.6)
n
~ 0

Note that θn (t) is the phase factor generalized to the case where En varies with time.
Plugging Eq.(9.6) into Eq.(9.5), we get
X X
i~ [ċn ψn + cn ψ̇n + icn ψn θ̇n ]eiθn = cn (Hψn )eiθn (9.7)
n n

From Eq.(9.6),
1
θ̇n = − En ⇒ − ~cn ψn θ̇n = cn En ψn = cn Hψn (9.8)
~
So, Eq.(9.7) is simplified to
X X
ċn ψn eiθn = − cn ψ̇n eiθn (9.9)
n n

Taking the inner product with ψm and using Eq.(9.4),


X X
ċn δmn eiθn = − cn hψm |ψ̇n ieiθn (9.10)
n n
X
⇒ ċm (t) = − cn hψm |ψ̇n iei(θn −θm ) . (9.11)
n

Differentiating Eq.(9.3) and taking the inner product with ψm ,

Ḣψn + H ψ̇n = Ėn ψn + En ψ̇n (9.12)


⇒ hψm |Ḣ|ψn i + hψm |H|ψ̇n i = Ėn δmn + En hψm |ψ̇n i (9.13)
9.1. ADIABATIC THEOREM 123

Since H is an hermitian operator, hψm |H|ψ̇n i = Em hψm |ψ̇n i, so


hψm |Ḣ|ψn i = (En − Em )hψm |ψ̇n i (for n 6= m) (9.14)
Putting this into Eq.(9.11)1 ,
i t
 
hψm |Ḣ|ψn i
X Z
0 0 0
ċm (t) = −cm hψm |ψ̇m i − cn exp − [En (t ) − Em (t )]dt (9.15)
n6=m
En − Em ~ 0

Now we use the adiabatic approximation — assume that Ḣ is so small that the second
term can be dropped altogether. Then,
ċm (t) = −cm hψm |ψ̇m i (9.16)
Z t 

ψm (t0 ) 0 ψm (t0 ) dt0 . (9.17)

⇒ cm (t) = cm (0)eiγm (t) , where γm (t) ≡ i
0 ∂t
 
0 ∂
0
Notice that γm (t) is real, because ψm (t ) 0 ψm (t ) is pure imaginary:
∂t
d
0= hψm |ψm i = hψm |ψ̇m i + hψ̇m |ψm i = 2Re(hψm |ψ̇m i) (9.18)
dt
If the particle starts out in the n-th eigenstate (cn (0) = 1 and cm (0) = 0 for m 6= n),
from Eq.(9.17) and Eq.(9.6),
Ψ(t) = eiθn (t) eiγn (t) ψn (t). (9.19)
This means that the particle remains in the n-th eigenstate of the evolving Hamiltonian,
picking up only a couple of phase factors.

Nearly adiabatic regime


Suppose the particle starts out in the n-th state, i.e. cm (0) = δmn . As we have just
proved, it remains in the n-th state. From Eq.(9.17),
cm (t) = δmn eiγn (t) . (9.20)
Substituting this into the right side of Eq.(9.11), we get the ”first correction” to adia-
baticity:
X
ċm = − δjn eiγn hψm |ψ̇j iei(θj −θm ) = −hψm |ψ̇n ieiγn ei(θn −θm ) (9.21)
j
Z t 
0 ∂ 0 0 0
ψm (t ) 0 ψn (t ) eiγn (t ) ei(θn (t )−θm (t )) dt0 .
0

⇒ cm (t) = cm (0) − (9.22)
0 ∂t
This formula enables us to calculate transition probabilities in the nearly adiabatic
regime.
1
We are assuming that the energies are non-degenerate.
124 CHAPTER 9. ADIABATIC APPROXIMATION

9.1.3 Infinite Square Well


Let us consider an infinite square well potential with a narrow potential barrier in it:


 +∞ x < 0
 0 0<x<L


V (x) = +∞ L < x < L +  (9.23)
0 L +  < x < 2L




+∞ 2L < x,

where   L.

Suppose that a particle of mass m is confined in the left side of the well (0 < x < L),
and that it is in the ground state:
r
π 2 ~2 2 π 
E1 = , ψ1 = sin x , (0 < x < L). (9.24)
2mL2 L L

Now, let us remove the potential barrier suddenly at t = 0, so that the particle can move
within the entire well of width 2L. Since this sudden action does not affect the form
of the wave function, the energy does not change either (i.e. the energy is conserved).
What is the probability P that the particle is in the new ground state? Let’s expand ψgs
in the new complete set of the wave functions:


X
ψ1 = c0n ψn0 , (9.25)
n=1
r
1  nπ 
where ψn0 = sin x (0 < x < 2L). (9.26)
L 2L
9.1. ADIABATIC THEOREM 125

Then, the probability P is

P = |c01 |2 = |hψ10 |ψ1 i|2 . (9.27)


√ Z L
0 2 π  π 
hψ1 |ψ1 i = sin x sin x dx (9.28)
L 0 2L L
√ Z L   
2 π  3π
= cos x − cos x dx (9.29)
2L 0 2L 2L
√   √
2 2L 2L 4 2
= + = (9.30)
2L 3π π 3π
32
∴ P = . (9.31)
9π 2
The normalized wave function of the particle at time t can be written as
∞  
X i 0
Ψ(x, t) = c0n ψn0 exp − En t , (9.32)
n=1
~
2 2 2
nπ ~ 1  nπ 
where En0 = and ψn0 = √ sin x (0 < x < 2L). (9.33)
2m(2L)2 L 2L

The coefficients c0n are


√ Z L
2  nπ  π 
c0n = sin x sin x dx (9.34)
L 2L L
 0 √
4 2  nπ 
(n 6= 2)


 sin
= (4 − n2 )π 2 (9.35)
1
 √ (n = 2)


2

For instance, the probability that the particle is found in the left half of the well (0 <
Z L
x < L) when its position is measured at time t is |Ψ(x, t)|2 dx.
0
Now, instead of removing the barrier suddenly at t = 0, let’s move the barrier very slowly
(= adiabatically) from x = L to x = 2L. According to the adiabatic theorem, the particle
continues to stay in the ground state. Thus, the expectation value of the particle’s energy
when the width of the barrier has become 2L is
π 2 ~2 E1
E10 = 2
= . (9.36)
8mL 4
This result shows that in the adiabatic process the energy is NOT conserved.
Suppose that the particle confined in the left side of the wall (0 < x < L) is a classical
126 CHAPTER 9. ADIABATIC APPROXIMATION

particle of energy E1 . Assume that the collision of the particle on the wall is perfectly
elastic, and that the collisions take place many times during the movement of the wall.
The energy loss of the particle is equal to the work that the particle does when it collides
with the wall. The number of collisions per unit time is v/(2a), so
 
∆p v 1 2 ∆a ∆a
∆E = −F ∆a = − ∆a = −(2mv) ∆a = −2 mv = −2E
∆t 2a 2 a a
∆E ∆a
⇒ = −2 ⇒ E ∝ a−2 . (9.37)
E a
So, when the width of the wall a increases from L to 2L, the energy of the particle
decreases from E1 to E1 /4.
Let us expand ψ(x, t) as
i t
X  Z 
0
ψ(x, t) = cn (t)φn (x, a) exp − en (a)dt , (9.38)
n
~ 0
where φn (x, a) and en (a) are the wave functions and their eigenenergies when the width
of the well is a, respectively. Putting Eq.(9.38) into the time-dependent Schrödinger
∂ψ
equation i~ = Hψ, taking the inner product with φm and exploiting hφm |φn i = δmn ,
∂t
we get
 Z t 
X i 0
ċm = − cn hφm |φ̇n i exp [em (a) − en (a)]dt . (9.39)
n
~ 0

Noting that
r
2  nπ  ∂φn da
φn = sin x and φ̇n = , (9.40)
a a ∂a dt
we get
1 da
hφm |φ̇n i = Jnm (9.41)
a dt

 2nm(−1)m−n
(for n 6= m)
where Jnm = 2 2 (9.42)
 0 m −n (for n = m).
So, Eq.(9.39) can be written as
X 1 da  Z t 
i 0
ċm = − cn Jnm exp [em (a) − en (a)]dt . (9.43)
n
a dt ~ 0
da
Notice that Jnm neither depends on a nor t. If  a, the process of expanding the well
dt
1 da
is adiabatic because  1 ⇒ ċm ≈ 0 ⇒ cm (t) ≈ cm (0).
a dt
9.2. BERRY’S PHASE 127

9.2 Berry’s Phase


A system that does not return to its original state when transported around a closed
loop is said to be non-holonomic. How does the final state differ from the initial state,
if the parameters in the Hamiltonian are carried adiabatically around some closed cycle?

§
A particle that starts out in the n-th eigenstate of H(0) remains in the n-th eigenstate
of H(t) under adiabatic conditions. From 9.1.2, its wave function is
Ψ(t) = eiθn (t) eiγn (t) ψn (t), (9.44)
1 t
Z
where θn (t) ≡ − En (t0 )dt0 (9.45)
~ 0
Z t 
0 ∂
ψn (t ) 0 ψn (t ) dt0 .
0

and γn (t) ≡ i (9.46)
0 ∂t
θn (t) is called the dynamic phase, and γn (t) is the geometric phase. Now, ψn (t)
depends on t because there is some parameter R(t) in the Hamiltonian that is changing
with time2 . Thus,
∂ψn ∂ψn dR
= , (9.47)
∂t ∂R  dt
Z t  Z Rf  
∂ψn dR 0 ∂ψn
⇒ γn (t) = i ψn dt = i ψn dR, (9.48)
0 ∂R dt0 Ri ∂R
where Ri and Rf are the initial and final values of R(t). If the Hamiltonian returns to
its original form after time T , so that Rf = Ri , then γn (T ) = 0.
Suppose that there are N time-dependent parameters in the Hamiltonian: R1 (t), R2 (t),
. . . , RN (t). In that case,
∂ψn ∂ψn dR1 ∂ψn dR2 ∂ψn dRN dR
= + + ... + = (∇R ψn ) · , (9.49)
∂t ∂R1 dt ∂R2 dt ∂RN dt dt
 
∂ ∂ ∂
where R ≡ (R1 , R2 , . . . , RN ) and ∇R ≡ , ,..., . Then,
∂R1 ∂R2 ∂RN
Z Rf
γn (t) = i hψn |∇R ψn i · dR. (9.50)
Ri

Now,
hψn |ψn i = 1
⇒ 0 = ∇R hψn |ψn i = h∇R ψn |ψn i + hψn |∇R ψn i = hψn |∇R ψn i∗ + hψn |∇R ψn i
∴ hψn |∇R ψn i is pure imaginary. (9.51)
2
In the case of the infinite square well, R(t) is the width of the expanding well.
128 CHAPTER 9. ADIABATIC APPROXIMATION

So, If ψn is real, hψn |∇R ψn i must be zero, and obviously γn (t) = 0 from Eq.(9.50).
If the Hamiltonian returns to its original form after a time T , the geometric phase is
expressed as a line integral around a closed loop in parameter-space:
I
γn (T ) = i hψn |∇R ψn i · dR. (9.52)

This is called Berry’s phase3 .


We have learned that the phase of the wave function is arbitrary — physical quantities
involve |Ψ|2 and the phase factor cancels out. However, if you carry the Hamiltonian
around a closed loop, the relative phase at the beginning and the end of the process is not
arbitrary, and can be measured. For instance, suppose we take a beam of particles, all in
the state Ψ, and split it in two, so that one beam passes through an adiabatically changing
potential, whereas the other beam does not. When the two beams are recombined, the
total wave function has the form
1 1
Ψ = Ψ0 + Ψ0 eiΓ , (9.53)
2 2
where Ψ0 is the wave function of the direct beam, and Γ is the extra phase (in part
dynamic, and in part geometric) acquired by the beam subjected to the varying Hamil-
tonian. In this case,
1
|Ψ|2 = |Ψ0 |2 1 + eiΓ 1 + e−iΓ
 
(9.54)
4  
1 Γ
= |Ψ0 |2 (1 + cos Γ) = |Ψ0 |2 cos2 (9.55)
2 2
So, Γ can be measured by looking for points of constructive and destructive interference
where Γ is an even or odd multiple of π, respectively.
If the parameter space is three dimensional (R = (R1 , R2 , R3 )), Eq.(9.52) reminds you of
the expression for magnetic flux in terms of the vector potential A. The magnetic flux
Φ through a surface S bounded by a curve C is4
Z Z I
Φ≡ B · da = (∇ × A) · da = A · dr. (9.56)
S S C

Berry’s phase can be regarded as the ”flux” of a ”magnetic field” B = i∇R × hψn |∇R ψn i
through the closed loop trajectory in parameter-space. In other words, in the 3D case,
Berry’s phase can be written as a surface integral:
Z
γn (T ) = i [∇R × hψn |∇R ψn i] · da. (9.57)

3
Notice that γn (T ) does not depend on the elapsed time T , whereas θn (t) depends critically on T .
4
Remember Stoke’s theorem.
9.3. AHARONOV-BOHM EFFECT 129

Figure 9.1: Magnetic flux through a surface S bounded by the closed curve C.

9.3 Aharonov-Bohm Effect


In classical electrodynamics, the scalar and vector potentials ϕ and A are related to the
electric and magnetic fields by
∂A
E = −∇ϕ − , B =∇×A (9.58)
∂t
Even if you change the potentials as:
∂Λ
ϕ → ϕ0 = ϕ − , A → A0 = A + ∇Λ, (9.59)
∂t
where Λ is any function of position and time, it has no effect on the fields. In quantum
mechanics, the Hamiltonian is
 2
1 ~
H= ∇ − qA + qϕ, (9.60)
2m i
and the theory is invariant under the transformation Eq.(9.59). But the vector potential
can affect the quantum behavior of a charged particle.
Suppose that a particle is constrained to move in a circle of radius b. Along the axis
runs a solenoid of radius a < b, carrying a steady electric current I. If the solenoid is
extremely long, the magnetic field inside the solenoid is uniform, and the field outside is
zero. However, the vector potential A outside the solenoid is NOT zero. Notice that
I Z Z
A · dl = (∇ × A) · ds = B · ds = Φ, (9.61)

where Φ = πa2 B is the flux of B through the solenoid. The vector potential is circum-
ferential, so
I
A · dl = A(2πr). (9.62)
130 CHAPTER 9. ADIABATIC APPROXIMATION

Figure 9.2: Charged bead on a circular ring through which a long solenoid passes.

Therefore,

Φ
A= φ̂ (r > a) (9.63)
2πr
Meanwhile, the solenoid itself is uncharged, so the scalar potential ϕ is zero. Thus the
Hamiltonian Eq.(9.60) becomes

1  2 2
−~ ∇ + q 2 A2 + 2i~qA · ∇

H= (9.64)
2m
The wave function depends only on the azimuthal angle φ (θ = π/2 and r = b), so
φ̂ d
∇→ , and the Schrödinger equation becomes
b dφ
" 2 #
~2 d2

1 qΦ ~qΦ d
− 2 2+ +i 2 ψ(φ) = Eψ(φ). (9.65)
2m b dφ 2πb πb dφ
d2 ψ dψ
⇒ 2
− 2iβ + ψ = 0 (9.66)
dφ dφ
qΦ 2mb2 E
where β ≡ and  ≡ − β 2. (9.67)
2π~ ~2
Solutions are
p b√
ψ = Aeiλφ , where λ = β ± β2 +  = β ± 2mE (9.68)
~
9.3. AHARONOV-BOHM EFFECT 131

Continuity of ψ(φ) at φ = 2π says that λ must be an integer, so


b√
β± 2mE = n (9.69)
~
2
~2


⇒ En = n− (n = 0, ±1, ±2, . . .). (9.70)
2mb2 2π~
Positive n, representing a particle traveling in the same direction as the current in the
solenoid, has a lower energy (if q > 0) than negative n, representing a particle traveling
in the opposite direction. The allowed energies depend on the field inside the solenoid,
even though the field at the location of the particle is zero.
More generally, suppose that a particle is moving through a region where B = ∇×A = 0
but A 6= 0. The time-dependent Schrödinger equation is
"  2 #
1 ~ ∂Ψ
∇ − qA + V Ψ = i~ . (9.71)
2m i ∂t

Let us rewrite Ψ as
Z r
0 q
Ψ=e Ψ, ig
where g(r) ≡ A(r 0 ) · dr 0 . (9.72)
~ O

The point O is an arbitrarily chosen reference point5 . Then,

∇Ψ = eig (i∇g)Ψ0 + eig (∇Ψ0 ). (9.73)

∇g = (q/~)A, so
 
~ ~
∇ − qA Ψ = eig ∇Ψ0 . (9.74)
i i
Therefore
 2
~
∇ − qA Ψ = −~2 eig ∇2 Ψ0 . (9.75)
i
Putting this into Eq.(9.71), we get
~2 2 0 ∂Ψ0
− ∇ Ψ + V Ψ0 = i~ , (9.76)
2m ∂t
which means that Ψ0 satisfies the Schrödinger equation without A.
Imagine an experiment in which a beam of electrons is split in two, and passed either
side of a long solenoid, before being recombined. A is given by Eq.(9.63), and the two
5
The line integral in Eq.(9.72) does not depend on the path taken from O to r, because ∇ × A = 0.
132 CHAPTER 9. ADIABATIC APPROXIMATION

Figure 9.3: The Aharonov-Bohm effect: The electron beam splits, with half passing
either side of a long solenoid.

beams are recombined with different phases:


Z Z  
q qΦ 1 qΦ
g= A · dr = φ̂ · (rφ̂dφ) = ± . (9.77)
~ 2π~ r 2~

The plus sign applies to the electrons traveling in the same direction as A, i.e. in the
same direction as the current in the solenoid. The beams arrive out of phase by the
amount qΦ/~, and this phase shift leads to measurable interference (Eq.(9.55)).
Chapter 10

Time-Dependent Perturbation

10.1 Two-level Systems


Suppose there are just two eignstates of the unperturbed Hamiltonian:

H (0) ψa = Ea ψa , H (0) ψb = Eb ψb , (10.1)

where the two states are orthogonal:

hψi |ψj i = δij (i, j = a, b). (10.2)

Let us express the state at t = 0 as:

Ψ(0) = ca ψa + cb ψb , |ca |2 + |cb |2 = 1. (10.3)

The state evolves with time as:

Ψ(t) = ca ψa e−iEa t/~ + cb ψb e−iEb t/~ . (10.4)

10.1.1 Perturbed System


Suppose a time-dependent perturbation H 0 (t) is added to the system. Since ψa and ψb
form a complete set, Ψ(t) can still be expressed as a linear combination of them, but ca
and cb now depend on time:

Ψ(t) = ca (t)ψa e−iEa t/~ + cb (t)ψb e−iEb t/~ . (10.5)

The state Ψ(t) satisfies the time-dependent Schrödinger equation:

∂Ψ
HΨ = i~ , where H = H (0) + H 0 (t). (10.6)
∂t
133
134 CHAPTER 10. TIME-DEPENDENT PERTURBATION

Putting Eq.(10.5) into Eq.(10.6) and using Eq.(10.1) and Eq.(10.2), we get

dca i 0 0 −i(Eb −Ea )t/~



= − ca Haa + cb Hab e (10.7)
dt ~
dcb i 0 0 i(Eb −Ea )t/~

= − cb Hbb + ca Hba e (10.8)
dt ~
where Hij0 ≡ hψi |H 0 |ψj i. (10.9)

10.1.2 Time-Dependent Perturbation Theory


0 0
Case of Haa = Hbb =0
Typically, the diagonal matrix elements of H 0 are zero. For instance, let us consider a
hydrogen atom placed in an electric field E = E(t)ẑ, where ẑ is the unit vector in the z
direction, and examine the transition between the ground state (n = 1) and the four-fold
degenerate first excited states (n = 2). The wave functions are

1 1  r  −r/2a
ψ100 = R10 Y00 = √ e−r/a , ψ200 = R20 Y00 = √ 1− e ,
πa3 8πa3 2a
1 r −r/2a 1 r −r/2a
ψ210 = R21 Y10 = √ e cos θ, ψ21±1 = R21 Y1±1 = ∓ √ e sin θe±iφ
32πa3 a 64πa3 a

The perturbed Hamiltonian is

H 0 = eEz. (10.10)

Notice that

r cos θ = z, r sin θe±iφ = x ± iy. (10.11)

So, |ψ|2 is an even function of z in all cases, and hence


Z
z|ψ|2 dxdydz = 0 ⇒ Hii0 = 0. (10.12)

And ψ100 , ψ200 and ψ21±1 are even in z, so Hij0 = 0 except

28
 
0
H100,210 = √ eEa. (10.13)
35 2
This means that only the state ψ210 is accessible from the ground state by the perturba-
tion Eq.(10.10). Therefore the system functions as a two-state configuration, assuming
transitions to higher excited states can be ignored.
10.1. TWO-LEVEL SYSTEMS 135

When the diagonal matrix elements of H 0 are zero, Eq.(10.7) and Eq.(10.8) simply be-
come:
dca i 0 −iω0 t dcb i 0 iω0 t
= − Hab e cb , and = − Hba e ca (10.14)
dt ~ dt ~
Eb − Ea
where ω0 ≡ (10.15)
~
If H 0 is small, Eq.(10.14) can be solved by a process of successive approximations. Sup-
pose the particle starts out in the state

ca (0) = a, cb (0) = b. (10.16)

If there were no perturbation at all, the state remains unchanged forever:


(0)
Zeroth Order : c(0)
a (t) = a, cb (t) = b. (10.17)

To get the first-order approximation, we insert the zeroth-order values on the right side
of Eq.(10.14).
ib t 0 0 −iω0 t0 0
 Z
dca i 0 −iω0 t (1)
= − Hab e b ⇒ ca (t) = a − H (t )e dt


~ Z0 ab

dt ~ (10.18)
dcb i 0 iω0 t (1) ia t 0 0 iω0 t0 0
= − Hba e a ⇒ cb (t) = b − H (t )e dt


~ 0 ba

dt ~
To get the second-order approximation, we insert the fist-order values on the right side
of Eq.(10.14).
ia t 0 0 iω0 t0 0
 Z 
dca i 0 −iω0 t
= − Hab e b− H (t )e dt
dt ~ ~ 0 ba
Z t Z t "Z 0 #
t
ib 0 a 0 00
⇒ c(2)a (t) = a − H 0 (t0 )e−iω0 t dt0 − 2 H 0 (t0 )e−iω0 t 0
Hba (t00 )eiω0 t dt00 dt0
~ 0 ab ~ 0 ab 0

(2)
To get cb (t), we just need to switch a ↔ b and change the sign of ω0 :
"Z 0 #
ia t 0 0 iω0 t0 0
Z Z t t
(2) b 0 00
cb (t) = b − H (t )e dt − 2 H 0 (t0 )eiω0 t 0
Hab (t00 )e−iω0 t dt00 dt0
~ 0 ba ~ 0 ba 0

In particular, if the particle starts out in the state ca (0) = 1 and cb (0) = 0,
Z t "Z 0 #
t
1 0 00
c(2)
a (t) = 1 − 2 H 0 (t0 )e−iω0 t 0
Hba (t00 )eiω0 t dt00 dt0 (10.19)
~ 0 ab 0
Z t
(2) i 0 0
cb (t) = − Hba (t0 )eiω0 t dt0 (10.20)
~ 0
136 CHAPTER 10. TIME-DEPENDENT PERTURBATION

0 0
Case of Haa 6= 0 and Hbb 6= 0
From Eq.(10.7), Eq.(10.8) and Eq.(10.15),
dca i 0 0 −iω0 t

= − ca Haa + cb Hab e (10.21)
dt ~
dcb i 0 0 iω0 t

= − cb Hbb + ca Hba e (10.22)
dt ~
If the particle starts out in the state ca (0) = 1 and cb (0) = 0,
(0)
Zeroth Order : c(0)
a (t) = 1, cb (t) = 0. (10.23)
Putting these values into the right side of Eq.(10.21) and Eq.(10.22),
i t 0 0 0
Z
dca i 0 (1)
= − Haa ⇒ ca (t) = 1 − H (t )dt (10.24)
dt ~ ~ 0 aa
i t 0 0 iω0 t0 0
Z
dcb i 0 iω0 t (1)
= − Hba e ⇒ cb (t) = − H (t )e dt (10.25)
dt ~ ~ 0 ba

10.1.3 Sinusoidal Perturbation


Let us consider the perturbation that has sinusoidal time dependence:
H 0 (r, t) = 2V (r) cos(ωt) (10.26)
0
⇒ Hab = 2Vab cos(ωt), where Vab ≡ hψa |V |ψb i. (10.27)
We assume the diagonal matrix elements are zero. If the particle starts out in the state
ca (0) = 1 and cb (0) = 0, from Eq.(10.20),
Vba ei(ω0 +ω)t − 1 ei(ω0 −ω)t − 1
 
cb (t) ≈ − + (10.28)
~ ω0 + ω ω0 − ω
Let us examine the case in which the driving frequency (ω) is close to the transition
frequency (ω0 ). Assuming ω0 + ω  |ω0 − ω|,
Vba ei(ω0 −ω)t/2  i(ω0 −ω)t/2
− e−i(ω0 −ω)t/2

cb (t) ≈ − e
~ ω0 − ω
Vba sin[(ω0 − ω)t/2] i(ω0 −ω)t/2
= −i e (10.29)
~ (ω0 − ω)/2
The transition probability (= the probability that a particle which started out in the
state ψa will be found in the state ψb at time t) is
|Vab |2 sin2 [(ω0 − ω)t/2]
Pa→b (t) = |cb (t)|2 ≈ (10.30)
~2 [(ω0 − ω)/2]2
10.1. TWO-LEVEL SYSTEMS 137

Figure 10.1: (Upper) Transition probability as a function of time for a sinusoidal pertur-
bation. (Lower) Transition probability as a function of driving frequency.

As a function of time, Pa→b (t) oscillates sinusoidally (see Fig.10.1), whereas as a func-
tion of frequency Pa→b (ω) has a peak of height (|Vab |t/2~)2 and width 4π/t. The peak
resembles a Dirac delta function in the limit that the frequency width 4π/t → 0, which
corresponds to t → ∞. Mathematically,

sin2 [(ω0 − ω)t/2]


lim = 2πtδ(ω0 − ω). (10.31)
t→∞ [(ω0 − ω)/2]2

In this limit, Eq.(10.30) becomes

|Vab |2 sin2 [(ω0 − ω)t/2] |Vab |2


Pa→b (t → ∞) = lim = 2πtδ(ω0 − ω)
~2 t→∞ [(ω0 − ω)/2]2 ~2
2π|Vab |2
= tδ(ω0 − ω) (10.32)
~2
138 CHAPTER 10. TIME-DEPENDENT PERTURBATION

The transition rate, defined as the transition probability per unit time, is constant with
time:
d 2π|Vab |2
Ra→b (t → ∞) = Pa→b (t → ∞) = δ(ω0 − ω). (10.33)
dt ~2
The delta function in Eq.(10.33) is the energy conserving delta function — it requires that
the quantum of energy causing the transition (e.g., the photon of a laser beam) match
the energy difference between the two states. In many practical applications, there is
a spread of final energy states rather than a discrete quantum state. For example, in
a solid, the band structure of the electronic energy levels represents a continuous range
of allowed states. In those cases, the relevant transition rate is a sum over the rates to
all accessible states. We assume that these rates are incoherent so that we can add the
rates (i.e., probabilities) rather than the amplitudes. We also assume that the spread in
energies is larger than the width of the sinc function that we turned into a delta function
but small enough that the rates to all states are the same. Let g(E) be the density of
states, such that g(E)dE is the number of energy levels between E and E + dE. Then
the total transition rate is given by the integral over all the rates:
Z Eb +
2π|Vab |2
Ra→b (t → ∞) = δ(ω0 − ω)g(E)dE (10.34)
Eb − ~2
2π|Vab |2 Eb +
Z
= δ(ω0 − ω)g(E)~dω (10.35)
~2 Eb −
2π|Vab |2
= g(Eb ) (10.36)
~
This result is referred to as Fermi’s golden rule.
10.1. TWO-LEVEL SYSTEMS 139

Rotating wave approximation


The first term in Eq.(10.28) comes from the eiωt part of cos(ωt), and the second term
comes from e−iωt . So, dropping the first term in Eq.(10.28) is formally equivalent to
writing H 0 = V e−iωt , that is,
0
Hba = Vba e−iωt , 0
Hab 0 ∗
= (Hba ) = Vab eiωt (10.37)

Again, we suppose that the particle starts out in the state ca (0) = 1 and cb (0) = 0. From
Eq.(10.14),

i i
c˙a = − Vab eiωt e−iω0 t cb , c˙b = − Vba e−iωt eiω0 t ca (10.38)
~ ~
Differentiating the latter and substituting in the former:

Vba 
i(ω0 − ω)ei(ω0 −ω)t ca + ei(ω0 −ω)t c˙a

c¨b = −i
~    
Vba i(ω0 −ω)t Vba i(ω0 −ω)t Vab −i(ω0 −ω)t
= i(ω0 − ω) −i e ca − i e −i e cb
~ ~ ~
|Vab |2
= i(ω0 − ω)c˙b − cb .
~2
|Vab |2
⇒ c¨b + i(ω − ω0 )c˙b + cb = 0. (10.39)
~2
The general solution to this equation can be written as

cb (t) = e−i(ω−ω0 )t/2 [C cos(ωr t) + D sin(ωr t)] (10.40)


s 2  2
ω − ω0 |Vab |
where ωr ≡ + . (10.41)
2 ~

ωr is called the Rabi flopping frequency. From the initial condition cb (0) = 0, C = 0,
so

cb (t) = De−i(ω−ω0 )t/2 sin(ωr t) (10.42)


   
−i(ω−ω0 )t/2 ω0 − ω
⇒ c˙b = De i sin(ωr t) + ωr cos(ωr t) (10.43)
2

Putting this into the second equation in Eq.(10.38),


   
~ i(ω−ω0 )t ~ i(ω−ω0 )t/2 ω0 − ω
ca (t) = i e c˙b = i e D i sin(ωr t) + ωr cos(ωr t) (10.44)
Vba Vba 2
140 CHAPTER 10. TIME-DEPENDENT PERTURBATION

−iVba
From the initial condition ca (0) = 1, D = . Therefore,
~ωr
   
i(ω−ω0 )t/2 ω0 − ω
ca (t) = e cos(ωr t) + i sin(ωr t) (10.45)
2ωr
i
cb (t) = − Vba ei(ω0 −ω)t/2 sin(ωr t) (10.46)
~ωr
The transition probability is
 2
2 |Vab |
Pa→b (t) = |cb (t)| = sin2 (ωr t) (10.47)
~ωr
The system returns to its initial state at ωr t = nπ ⇒ t = nπ/ωr .
If |Vab |2  ~2 (ω − ω0 )2 , then ωr ≈ (1/2)|ω − ω0 |, and
|Vab |2 sin2 [(ω0 − ω)t/2]
Pa→b (t) = , (10.48)
~2 [(ω0 − ω)/2]2
which confirms Eq.(10.30).

10.2 Emission and Absorption of Radiation


10.2.1 Electromagnetic Waves
An electromagnetic wave (i.e. light) consists of mutually-perpendicular oscillating electric
and magnetic fields, as shown in Fig.10.2. In the presence of a passing light wave, an
atom primarily responds to the electric component. Let us suppose that the wavelength
is long compared to the atomic size1 , and ignore the spatial variation in the field. Then,
assuming that the light is monochromatic at a single frequency ω and polarized along
the z direction, the atom feels an electric field:
E = E0 cos(ωt)ẑ (10.49)
The perturbing Hamiltonian is
H 0 = −qE0 z cos(ωt) (10.50)
⇒ 0
Hba = −PE0 cos(ωt), where P ≡ qhψb |z|ψa i. (10.51)
0 0
The diagonal matrix elements Haa and Hbb vanish2 . So Eq.(10.30) can be used with
Vba = −PE0 . P is the off-diagonal matrix element of the z component of the dipole
moment operator, qr. Therefore radiation governed by Eq.(10.51) is called electric
dipole radiation, which is overwhelmingly the dominant kind, at least in the visible
region.
1
λ ∼ 5000Å for visible light, while an atomic diameter is ∼ 1Å.
2
Since ψ is typically an even or odd function of z, z|ψ|2 is odd and integrates to zero.
10.2. EMISSION AND ABSORPTION OF RADIATION 141

Figure 10.2: An electromagnetic wave.

10.2.2 Absorption, Stimulated Emission and Spontaneous Emission


Let us assume Ea < Eb . If an atom starts out in the lower state ψa and you shine a
polarized monochromatic beam of light on the atom, the probability of a transition to
the upper state ψb is
2
sin2 [(ω0 − ω)t/2]

|P|E0
Pa→b (t) = . (10.52)
~ (ω0 − ω)2

In this process, the atom absorbs a photon of energy Eb − Ea = ~ω0 from the electro-
magnetic field.
If the atom starts out in the state ψb (i.e. cb (0) = 1 and ca (0) = 0), the probability of a
transition down to the lower level ψb is
2
sin2 [(ω0 − ω)t/2]

2 |P|E0
Pb→a (t) = |ca (t)| = . (10.53)
~ (ω0 − ω)2

This is called stimulated emission — If the particle is in the upper state and you shine
light on it, it can make a transition to the lower state, and the transition probability is
exactly the same as for a transition upward from the lower state3 . In this process, one
photon comes in and two photons come out (the original one that caused the transition +
another one of energy ~ω0 from the transition). The stimulated emission is the principle
behind the laser (Light Amplification by Stimulated Emission of Radiation). Note that it
is essential to get a majority of the atoms into the upper state (population inversion),
3
To derive Eq.(10.53), switch a ↔ b in Eq.(10.28), which substitutes −ω0 for ω0 , and then keep the
first term, with −ω0 + ω in the denominator.
142 CHAPTER 10. TIME-DEPENDENT PERTURBATION

because absorption competes with stimulated emission.


The third process is spontaneous emission — an atom in an excited state makes a
transition downward, with the release of a photon, but without any electromagnetic field
to initiate the process. This process is strange at first sight, because, if the atom is in a
stationary state and there is no external perturbation, the atom should stay in the state
forever. However, quantum electrodynamics says that the fields are non-zero even in the
ground state, just as the harmonic oscillator has non-zero energy in its ground state4 . It
is this ”zero-point radiation” that causes spontaneous emission.

10.2.3 Incoherent Perturbation


In general, the energy per unit volume in electromagnetic fields is
0 2 1 2
u= E + B . (10.54)
2 2µ0

For electromagnetic waves, the electric and magnetic contributions are equal, so

u = 0 E 2 = 0 E02 cos2 (ωt). (10.55)

The average over a full cycle is (1/2)0 E02 since the average of cos2 is 1/2, so the energy
density in an electromagnetic wave is
1
ū = 0 E02 (10.56)
2
So, Eq.(10.53) can be written as
2
2ū 2 sin [(ω0 − ω)t/2]
Pb→a (t) = |P| (10.57)
0 ~2 (ω0 − ω)2

This is for a monochromatic wave at a single frequency ω0 . Usually the system is exposed
to electromagnetic waves at a whole range of frequencies; in that case, replacing ū by
ρ(ω)dω 5 :
Z ∞  2 
2 2 sin [(ω0 − ω)t/2]
Pb→a (t) = |P| ρ(ω) dω. (10.58)
0 ~2 0 (ω0 − ω)2
4
We can turn out all the lights and cool the room down to absolute zero temperature, but there is
still some electromagnetic radiation present.
5
Note that here we are assuming that the perturbations at different frequencies are independent, so
that the total transition probability is a sum of the individual probabilities. If the different components
are coherent (i.e. phase-correlated), we need to add amplitudes ca (t), not probabilities |ca (t)|2 , and
there will be cross-terms. The perturbations are always incoherent for the applications we will consider.
10.2. EMISSION AND ABSORPTION OF RADIATION 143

The term {. . .} is sharply peaked about ω0 , while ρ(ω) is normally quite broad, so let us
replace ρ(ω) by ρ(ω0 ) and take it outside the integral:
Z ∞ 2
2|P|2

sin [(ω0 − ω)t/2]
Pb→a (t) ≈ ρ(ω0 ) dω. (10.59)
0 ~2 0 (ω0 − ω)2
Changing variables to x ≡ (ω0 − ω)t/2,
Z ∞ 2
t ω0 t/2 sin2 x t ∞ sin2 x

sin [(ω0 − ω)t/2]
Z Z

2
dω = 2
dx ≈ 2
dx = . (10.60)
0 (ω0 − ω) 2 −∞ x 2 −∞ x 2
So Eq.(10.59) becomes
π|P|2
Pb→a (t) ≈ ρ(ω0 )t. (10.61)
0 ~2
The transition rate (R ≡ dP/dt) is a constant:
π|P|2
Rb→a = ρ(ω0 ). (10.62)
0 ~2
So far we have assumed that the perturbing wave is coming in along the y direction and
polarized in the z direction. Now let us think about an atom bathed in radiation coming
from all directions and with all possible polarizations. In place of |P|2 , we need the
average of |P · n̂|2 , where P ≡ qhψb |r|ψa i, and the average is over both polarizations (n̂)
and over all incident directions.
To calculate |P · n̂|2 , let us choose the z axis along the direction of propagation, and the
y axis such that the yz plane includes the P vector. The polarization vector n̂ is in the
xy plane. So,

n̂ = cos φ x̂ + sin φ ŷ, P = P sin θ ŷ + P cos θ ẑ. (10.63)


⇒ P · n̂ = P sin θ sin φ. (10.64)
Z π Z 2π
2 1
⇒ |P · n̂|ave = dθ dφ |P|2 sin2 θ sin2 φ sin θdθdφ (10.65)
4π 0 0
1
= |P|2 . (10.66)
3
From Eq.(10.62) and Eq.(10.66), the transition rate for stimulated emission from state b
to state a under the influence of incoherent, unpolarized light from all directions is
π
Rb→a = 2
|P|2 ρ(ω0 ), (10.67)
30 ~
where P ≡ qhψb |r|ψa i and ρ(ω0 ) is the energy density in the fields per unit frequency at
ω0 = (Eb − Ea )/~.
144 CHAPTER 10. TIME-DEPENDENT PERTURBATION

10.3 Spontaneous Emission


10.3.1 Spontaneous Emission Rate
Imagine a box of atoms in which Na of them are in the lower state ψa and Nb of them
are in the upper state ψb . We define A as the spontaneous emission rate, so that the
number of particles leaving the upper state per unit time by spontaneous emission is
Nb A. Meanwhile, from Eq.(10.67), the stimulated emission rate can be written as
π
Rb→a = Bba ρ(ω0 ), where Bba ≡ |P|2 . (10.68)
30 ~2
So, the number of particles leaving the upper state by stimulated emission is Nb Bba ρ(ω0 ).
The absorption rate is also proportional to ρ(ω0 ) and written as Bab ρ(ω0 ). As a result,
dNb
= −Nb A − Nb Bba ρ(ω0 ) + Na Bab ρ(ω0 ). (10.69)
dt
Supose these atoms are in thermal equilibrium with the ambient field. Then dNb /dt = 0,
so
A
ρ(ω0 ) = . (10.70)
(Na /Nb )Bab − Bba
Statistical mechanics says that the number of particles with energy E in thermal equi-
librium at temperature T is proportional to the Boltzmann factor, exp(−E/kB T ),
so
Na e−Ea /kB T
= −E /k T = e~ω0 /kB T (10.71)
Nb e b B
A A/Bab
⇒ ρ(ω0 ) = ~ω0 /k T = ~ω0 /k T . (10.72)
e B Bab − Bba e B − (Bba /Bab )
We compare this result with Planck’s blackbody formula:
~ ω3
ρ(ω) = . (10.73)
π 2 c3 e~ω/kB T − 1
We can find
~ω03 A
Bba = Bab , 2 3
= (10.74)
π c Bab
Using this and Eq.(10.68), the spontaneous emission rate becomes

ω03 |P|2
A= . (10.75)
3π0 ~c3
10.3. SPONTANEOUS EMISSION 145

10.3.2 Lifetime of an Excited State


Suppose that we have pumped a large number of atoms into the excited state ψb . In
a time interval dt, a fraction Adt of them leave the excited state due to spontaneous
emission:
dNb = −Nb Adt (10.76)
⇒ Nb (t) = Nb (0)e−At . (10.77)
The time constant τ = 1/A is defined as the lifetime of the state, which is the time it
takes for Nb (t) to reach 1/e ≈ 0.368 of its initial value. So far we have assumed that
there are only two states, ψa and ψb , for the system, but Eq.(10.75) gives the transition
rate for ψb → ψa regardless of what other states may be accessible. Typically an excited
atom has many different decay modes, so there are a number of different lower-energy
states: ψa1 , ψa2 , ψa3 . . .. In that case the transition rates add, and the net lifetime is
τ = 1/(A1 + A2 + A3 + . . .).
The half-life t1/2 of an excited state is the time it would take for half the atoms to make
a transition. The relation between t1/2 and the lifetime of the state τ is
t1/2 = τ ln 2. (10.78)

10.3.3 Selection Rules


To calculate the spontaneous emission, we need to evaluate the matrix elements hψb |r|ψa i.
Let us consider a system like hydrogen for which the Hamiltonian is spherically symmet-
rical, so that the states are labeled with the usual quantum numbers n, l and m. We can
find selection rules for hn0 l0 m0 |r|nlmi to have non-zero values.

Selection rules involving m and m0


The following commutation relations hold:
[Lz , x] = i~y, [Lz , y] = −i~x, [Lz , z] = 0. (10.79)
From the third of these,
0 = hn0 l0 m0 |[Lz , z]|nlmi = hn0 l0 m0 |(Lz z − zLz )|nlmi
= hn0 l0 m0 |[(m0 ~)z − z(m~)]|nlmi = (m0 − m)~hn0 l0 m0 |z|nlmi.
⇒ Either m0 = m, or else hn0 l0 m0 |z|nlmi = 0. (10.80)
Meanwhile, from the first relation in Eq.(10.79),
hn0 l0 m0 |[Lz , x]|nlmi = hn0 l0 m0 |Lz x − xLz |nlmi
= (m0 − m)~hn0 l0 m0 |x|nlmi = i~hn0 l0 m0 |y|nlmi.
⇒ (m0 − m)hn0 l0 m0 |x|nlmi = ihn0 l0 m0 |y|nlmi. (10.81)
146 CHAPTER 10. TIME-DEPENDENT PERTURBATION

From the second relation in Eq.(10.79),


hn0 l0 m0 |[Lz , y]|nlmi = hn0 l0 m0 |Lz y − yLz |nlmi
= (m0 − m)~hn0 l0 m0 |y|nlmi = −i~hn0 l0 m0 |x|nlmi.
⇒ (m0 − m)hn0 l0 m0 |y|nlmi = −ihn0 l0 m0 |x|nlmi. (10.82)
From Eq.(10.81) and Eq.(10.82),
(m0 − m)2 hn0 l0 m0 |x|nlmi = i(m0 − m)hn0 l0 m0 |y|nlmi = hn0 l0 m0 |x|nlmi.
⇒ Either (m0 − m)2 = 1, or else hn0 l0 m0 |x|nlmi = hn0 l0 m0 |y|nlmi = 0. (10.83)
In conclusion, from Eq.(10.80) and Eq.(10.83),
No transitions occur unless ∆m = ±1 or 0.

Selection rules involving l and l0


The following commutation relation holds:
 2 2 
L , [L , r] = 2~2 (rL2 + L2 r). (10.84)
Therefore,
hn0 l0 m0 | L2 , [L2 , r] |nlmi = 2~2 hn0 l0 m0 |(rL2 + L2 r)|nlmi
 

= 2~4 [l(l + 1) + l0 (l0 + 1)]hn0 l0 m0 |r|nlmi (10.85)


Meanwhile,
hn0 l0 m0 | L2 , [L2 , r] |nlmi = hn0 l0 m0 |(L2 [L2 , r] − [L2 , r]L2 )|nlmi
 

= ~2 [l0 (l0 + 1) − l(l + 1)]hn0 l0 m0 |[L2 , r]|nlmi


= ~2 [l0 (l0 + 1) − l(l + 1)]hn0 l0 m0 |(L2 r − rL2 )|nlmi
= ~4 [l0 (l0 + 1) − l(l + 1)]2 hn0 l0 m0 |r|nlmi (10.86)
From Eq.(10.85) and Eq.(10.86),
Either 2[l(l + 1) + l0 (l0 + 1)] = [l0 (l0 + 1) − l(l + 1)]2
⇔ [(l0 + l + 1)2 − 1][(l0 − l)2 − 1] = 0, (10.87)
or else hn0 l0 m0 |r|nlmi = 0. (10.88)
In Eq.(10.87), (l0 + l + 1)2 − 1 6= 0 unless l = l0 = 0, and so (l0 − l)2 = 1.
If l = l0 = 0,
1
|n00i = Rn0 (r)Y00 (θ, φ) = √ Rn0 (r)
Z 4π
1
⇒ hn0 00|r|n00i = Rn0 0 (r)Rn0 (r)(xî + y ĵ + z k̂)dxdydz (10.89)

The integrand is odd in x, y or z, so hn0 00|r|n00i = 0. In conclusion,
10.4. FORBIDDEN TRANSITIONS 147

No transitions occur unless ∆l = ±1.


The allowed transitions for the first four Bohr levels in hydrogen is shown in Figure 10.3.

Figure 10.3: Allowed decays for the first four Bohr levels in hydrogen.

Notice that the 2S (ψ200 ) state cannot decay because there is no lower-energy state with
l = 1. It is called a metastable state, and its lifetime is much longer than that of,
for example, the 2P states (ψ21±1 and ψ210 ). Metastable states do eventually decay by
collisions, forbidden transitions, or multiphoton emission.

10.4 Forbidden transitions


In Eq.(10.49) we assumed that the atom is so small in comparison to the wavelength of
light that spatial variations in the field can be ignored. However, the true electric field
would be
E(r, t) = E0 cos(k · r − ωt). (10.90)
r
If the atom is centered at the origin, then k · r ∼ 2π  1 over the volume of the atom,
λ
so we could afford to drop this term. Now let us keep the first-order correction:
E(r, t) = E0 [cos(ωt) + (k · r) sin(ωt)]. (10.91)
The first term gives rise to allowed (electric dipole) transitions, while the second term
leads to forbidden (magnetic dipole and electric quadrupole) transitions. The
perturbing Hamiltonian is
H 0 = −qE · r = −q(E0 · r)(k · r) sin(ωt). (10.92)
148 CHAPTER 10. TIME-DEPENDENT PERTURBATION

ω
Let us write E0 = E0 n̂, where n̂ is the polarization vector, and k = k̂, where k̂ is the
c
propagation direction of the light. Then,
nω o
H 0 = −qE0 (n̂ · r)(k̂ · r) sin(ωt). (10.93)
c
This is the analog to Eq.(10.50):

H 0 = −qE0 z cos(ωt) = −qE0 (n̂ · r) cos(ωt). (10.94)

Eq.(10.93) has sin(ωt) instead of cos(ωt), but this does not have any effect on the tran-
sition rate. Therefore, the spontaneous emission rate for forbidden transitions can be
obtained easily, based on Eq.(10.75). First, we have not yet averaged over polarization
and propagation directions, so we need to replace |P|2 /3 to |P|2 (see Eq.(10.66)). And
then we make this change:
ω
P = qhψb |z|ψa i → q hψb |(n̂ · r)(k̂ · r)|ψa i (10.95)
c
Thus, the spontaneous emission rate for forbidden transitions is

ω 3  qω 2
A= |hψb |(n̂ · r)(k̂ · r)|ψa i|2 . (10.96)
π0 ~c3 c

10.4.1 Transition from l = 0 to l0 = 0


Suppose the initial state has l = 0√and the final state also has l0 = 0. Then, ψa = Rn0 Y00
and ψb = Rn0 0 Y00 , where Y00 = 1/ 4π, so the angular part of hψb |(n̂ · r)(k̂ · r)|ψa i is
Z

hψb |(n̂ · r)(k̂ · r)|ψa i = · · · (n̂ · r)(k̂ · r) sin θdθdφ = · · · (n̂ · k̂) (10.97)
3

But n̂ · k̂ = 0, since electromagnetic waves are transverse. So the spontaneous emission


rate is zero, both for allowed and forbidden transitions. For example, the 2S → 1S
transition in hydrogen is not possible even by a forbidden transition6 .

10.5 Multilevel Systems


Let us develop the time-dependent perturbation theory for a multilevel system:

H0 ψn = En ψn , hψn |ψm i = δnm . (10.98)


6
The dominant decay from 2S → 1S is by two-photon emission, and the lifetime is about a tenth of
a second.
10.5. MULTILEVEL SYSTEMS 149

At t = 0, a perturbation H 0 (t) is turned on:

H = H0 + H 0 (t). (10.99)

A state is written as
X
Ψ(t) = cn (t)ψn e−iEn t/~ . (10.100)

∂Ψ
Plugging these equations into the Schrödinger equation Hψ = i~ , we get
∂t
X X X
cn e−iEn t/~ En ψn + cn e−iEn t/~ H 0 ψn = i~ ċn e−iEn t/~ ψn
 
i X
+ i~ − cn En e−iEn t/~ ψ(10.101)
n
~
X X
⇒ cn e−iEn t/~ H 0 ψn = i~ ċn e−iEn t/~ ψn (10.102)

Taking the inner product with ψm ,


X X
cn e−iEn t/~ hψm |H 0 |ψn i = i~ ċn e−iEn t/~ hψm |ψn i. (10.103)

0
Using hψm |ψn i = δmn and defining Hmn ≡ hψm |H 0 |ψn i, we get

iX 0
ċm = − cn Hmn ei(Em −En )t/~ (10.104)
~ n

Suppose that the system starts out in the state ψN . Then,

(0)
Zeroth Order : cN (t) = 1 and c(0)
m (t) = 0 (m 6= N ). (10.105)

In first order,
Z t
dcN i i
= − HN0 N ⇒ cN (t) = 1 − HN0 N (t0 )dt0 (10.106)
dt ~ ~ 0

whereas for m 6= N ,
Z t
dcm i 0 i(Em −EN )t/~ i 0 0
= − HmN e ⇒ cm (t) = − HmN (t0 )ei(Em −EN )t /~ dt0 . (10.107)
dt ~ ~ 0
150 CHAPTER 10. TIME-DEPENDENT PERTURBATION

10.5.1 Constant H 0
If H 0 is constant,
i
cN (t) = 1 − HN0 N t, (10.108)
~ Z t
i 0 0
cM (t) = − HM N ei(EM −EN )t /~ dt0 (10.109)
~ 0
0
 
HM N i(EM −EN )t/2~ EM − EN
= − e 2i sin t . (10.110)
EM − EN 2~
So, the probability of transition from state N to state M is
0 2
 
2 4|HM N| 2 EM − EN
PN →M = |cM | = sin t . (10.111)
(EM − EN )2 2~
X
If you want to know the probability of remaining in the original state ψN , use 1 − |cm (t)|2 ,
m6=N
rather than |cN (t)|2 , because the former gives you a better approximate value.

10.5.2 Sinusoidal H 0
If H 0 = 2V cos(ωt),
Z t
2i
cN (t) = 1 − VN N cos(ωt0 )dt0 (10.112)
~ 0
2i
= 1− VN N sin(ωt). (10.113)
~ω Z
t
i 0 0
 0
cM (t) = − VM N eiωt + e−iωt ei(EM −EN )t /~ dt0 (10.114)
~ 0
0 0  t
ei(~ω+EM −EN )t /~ ei(−~ω+EM −EN )t /~

iVM N
= − + (10.115)
~ i(~ω + EM − EN )/~ i(−~ω + EM − EN )/~ 0
 i(EM −EN +~ω)t/~
− 1 ei(EM −EN −~ω)t/~ − 1

e
= −VM N + (10.116)
EM − EN + ~ω EM − EN − ~ω
If EM > EN , the second term dominates and transitions occur only for ω ≈ (EM −EN )/~
(absorption). So, transitions occur only to states with energy EM = EN + ~ω with the
transition probability:
|VM N |2 sin2 [(ω0 − ω)t/2]
PN →M = |cM |2 = (10.117)
~2 [(ω0 − ω)/2]2
EM − EN
where ω0 ≡ . (10.118)
~
10.6. MAGNETIC RESONANCE 151

Notice that Eq.(10.117) is the same as Eq.(10.30) for the two-level systems. In addition,
Eq.(10.117) also holds for stimulated emission from state M to state N :
|VM N |2 sin2 [(ω0 − ω)t/2]
PM →N = (10.119)
~2 [(ω0 − ω)/2]2

10.5.3 Incoherent perturbation


Suppose that a multilevel system is immersed in incoherent electromagnetic radiation.
Since Eq.(10.119) is the same as the one for the two-level systems, the transition rate for
stimulated emission from state M to state N under the influence of incoherent, unpolar-
ized light incident from all directions is also the same as Eq.(10.67):
π
RM →N = 2
|P|2 ρ(ω0 ), (10.120)
30 ~
where P ≡ qhψM |r|ψN i and ρ(ω0 ) is the energy density in the fields per unit frequency
at ω0 = (EM − EN )/~.

10.6 Magnetic resonance


Let us consider a spin-1/2 particle with gyromagnetic ratio γ at rest in a static magnetic
field B0 ẑ. The particle precesses at the Larmor frequency ω0 ≡ γB0 . Now, we turn on a
small transverse radiofrequency field, so that the total field is
B = Brf cos(ωt)x̂ − Brf sin(ωt)ŷ + B0 ẑ (10.121)
The Hamiltonian matrix is
 
γ~ B0 Brf eiωt
H = −γB · S = − (10.122)
2 Brf e−iωt −B0
 
a(t)
Let us write the spin state of the particle as χ(t) = . From the Schrödinger
b(t)

equation i~ = Hχ, we get
dt
da i db i
Ωeiωt b + ω0 a , Ωe−iωt a − ω0 b , where Ω ≡ γBrf (10.123)
 
= =
dt 2 dt 2
The general solution to Eq.(10.123) is
 
0 i 0
a(t) = a0 cos(ω t/2) + 0 [a0 (ω0 − ω) + b0 Ω] sin(ω t/2) eiωt/2 (10.124)
ω
 
i
b(t) = b0 cos(ω t/2) + 0 [b0 (ω − ω0 ) + a0 Ω] sin(ω t/2) e−iωt/2 , (10.125)
0 0
ω
0
p
where ω ≡ (ω − ω0 )2 + Ω2 . (10.126)
152 CHAPTER 10. TIME-DEPENDENT PERTURBATION

If the particle starts out with spin up (a0 = 1, b0 = 0), the probability of a transition to
spin down is
 2
Ω2
 
Ω 0
2
P (t) = |b(t)| = 2
sin (ω t/2) = sin2 (ω 0 t/2). (10.127)
ω0 (ω − ω0 )2 + Ω2

Ω2
P (ω) ≡ is called the resonance curve. The maximum occurs at ω = ω0 ,
(ω − ω0 )2 + Ω2
and the full width at half maximum (FWHM) is ∆ω = 2Ω.
In a nuclear magnetic resonance (NMR) experiment, we can measure the resonant
frequency ω0 and thus the g-factor of the particle, because
ω0 gp ep 2mp ω0
γ= = ⇒ gp = , (10.128)
B0 2mp ep B0

where mp and ep are the mass and charge of the particle, respectively.
Chapter 11

Scattering

11.1 Introduction
11.1.1 Classical Theory
Imagine a particle incident on some heavy scattering center. It comes in with impact
parameter b, and emerges at some scattering angle θ. The smaller the impact pa-
rameter, the greater the scattering angle. More generally, particles incident within a
cross-sectional area dσ scatter into a solid angle dΩ. The larger dσ is, the bigger dΩ is.
dσ = db bdφ = bdbdφ, and dΩ = sin θdθdφ, so

dσ b db
= (11.1)
dΩ sin θ dθ

dσ/dΩ is called the differential cross section.

Figure 11.1: Particles incident in the area dσ scatter into the solid angle dΩ.

153
154 CHAPTER 11. SCATTERING

The total cross section is


Z

σ≡ dΩ. (11.2)
dΩ

Imagine a beam of incident particles with uniform luminosity L(≡ number of incident
particles [/m2 /s]. The number of particles entering area dσ (and then scattering into
solid angle dΩ) [/s] is dN = Ldσ, so

dσ 1 dN
= (11.3)
dΩ L dΩ

Imagine a detector in the laboratory that accepts particles scattering into dΩ. We can
obtain the differential cross section by counting the number of particles [/s] recorded by
the detector, dividing by dΩ, and normalizing to L.

Hard-sphere scattering

Suppose that the target is a hard heavy ball of radius R, and the incident particle bounces
off elastically.

Figure 11.2: Particles incident in the area dσ scatter into the solid angle dΩ.
11.1. INTRODUCTION 155

b = R sin α, and θ = π − 2α, so


   
π θ θ
b = R sin − = R cos
2 2 2
R2
   
db 1 θ dσ R cos(θ/2) R sin(θ/2)
= − R sin ⇒ = = .
dθ 2 2 dΩ sin θ 2 4
Z 2
R
∴ σ = dΩ = πR2 . (11.4)
4
The total cross section represents the total area of incident beam that is scattered by the
target.

11.1.2 Quantum Theory


We discuss scattering in terms of stationary states. As usual, the Schrödinger equation
is
~2 2
 
− ∇ + V (r) ψ(r) = Eψ(r). (11.5)
2m

Naturally V (r) → 0 as r → ∞. Therefore, at very large r, ψ(r) must satisfy

~2 2
− ∇ ψ(r) = Eψ(r) (at very large r). (11.6)
2m
Imagine an incident plane wave traveling in the z direction, which encounters a scat-
tering potential, producing an outgoing spherical wave. Assuming that the potential is
spherically symmetrical, we look for solutions to the Schrödinger equation of the general
form:
eikr
 
ikz
ψ(r, θ) ≈ A e + f (θ) for kr  1, (11.7)
r

2mE
where k ≡ . (11.8)
~
156 CHAPTER 11. SCATTERING

Figure 11.3: Incoming plane wave generates outgoing spherical wave.

Figure 11.4: The volume dV of incident beam that passes through area dσ in time dt.

f (θ) is called the scattering amplitude. The probability that the incident particle,
traveling at speed v, passes through area dσ in time dt is

dP = |ψincident |2 dV = |A|2 (v dt)dσ. (11.9)

This is equal to the probability that the particle scatters into the corresponding solid
angle dΩ:
|A|2 |f |2
dP = |ψscattered |2 dV = (v dt)r2 dΩ. (11.10)
r2
From Eq.(11.9) and Eq.(11.10),

= |f (θ)|2 . (11.11)
dΩ
Let us study two techniques for calculating f (θ): partial wave analysis and Born
approximation.
11.2. PARTIAL WAVE ANALYSIS 157

11.2 Partial Wave Analysis


11.2.1 partial wave amplitude
The Schrödinger equation for a spherically symmetrical potential V (r) admits the sepa-
rable solutions

ψ(r, θ, φ) = R(r)Ylm (θ, φ). (11.12)

u(r) ≡ rR(r) satisfies the radial equation:

~2 d2 u ~2 l(l + 1)
 
− + V (r) + u = Eu (11.13)
2m dr2 2m r2

At very large r (more precisely kr  1), the second term on the left side is negligible, so

d2 u
≈ −k 2 u ⇒ u(r) = Ceikr + De−ikr . (11.14)
dr2
eikr represents an outgoing wave, whereas e−ikr an incoming wave. We want an outgoing
wave, so D = 0. Therefore we get

eikr
R(r) ∼ , for kr  1 (radiation zone) (11.15)
r
In the intermediate region where V (r) is negligible1 , Eq.(11.13) becomes

d2 u l(l + 1)
2
− 2
u = −k 2 u. (11.16)
dr r
Since we want solutions in the form eikr , let us express the solutions to Eq.(11.16) using
spherical Hankel functions:
(1) (2)
u(r) = Arhl (kr) + Brhl (kr) (11.17)
(1) (2)
hl (x) ≡ jl (x) + inl (x), hl (x) ≡ jl (x) − inl (x), (11.18)

where jl (x) and nl (x) are spherical Bessel function and spherical Neumann func-
tion of order l, respectively:
 l  l
l 1 d sin x l 1 d cos x
jl (x) ≡ (−x) , nl (x) ≡ −(−x) . (11.19)
x dx x x dx x

1
The partial wave analysis is NOT applicable to the Coulomb potential, since 1/r goes to zero more
slowly than 1/r2 .
158 CHAPTER 11. SCATTERING

Figure 11.5: Scattering from a localized potential.

Figure 11.6: Spherical Hankel functions.


11.2. PARTIAL WAVE ANALYSIS 159

(1) (2)
At kr  1, hl ∝ eikr /r, while hl ∝ e−ikr /r. We are looking for a solution that
(1)
represents an outgoing wave, so R(r) ∼ hl (kr). Thus the exact wave function outside
the scattering region becomes
( )
X (1)
ψ(r, θ, φ) = A eikz + Cl,m hl (kr)Ylm (θ, φ) . (11.20)
l,m

We are assuming that V (r) is spherically symmetric, so ψ(r, θ, φ)r must not depend on
2l + 1
φ. Therefore, only terms with m = 0 survive. Using Yl0 (θ, φ) = Pl (cos θ) and
p 4π
letting Cl,0 ≡ il+1 k 4π(2l + 1)al ,
( ∞
)
X (1)
ψ(r, θ) = A eikz + k il+1 (2l + 1)al hl (kr)Pl (cos θ) , (11.21)
l=0

where Pl is the l-th Legendre polynomial, and al is called the l-th partial wave ampli-
(1)
tude. At kr  1, hl ∼ (−i)l+1 eikr /kr, so

eikr
  X
ikz
ψ(r, θ) ≈ A e + f (θ) , where f (θ) = (2l + 1)al Pl (cos θ)(11.22)
r l=0

 XX

 = |f (θ)|2 = (2l + 1)(2l0 + 1)a∗l al0 Pl (cos θ)Pl0 (cos θ),
 dΩ

l l0
⇒ X∞ (11.23)
2
 σ = 4π (2l + 1)|al |



l=0

Eq.(11.22) has cartesian coordinates for the incident wave eikz . Using Rayleigh’s for-
mula:

X
ikz
e = il (2l + 1)jl (kr)Pl (cos θ), (11.24)
l=0

Eq.(11.22) can be rewritten as



X h i
l (1)
ψ(r, θ) = A i (2l + 1) jl (kr) + ikal hl (kr) Pl (cos θ) (11.25)
l=0

To determine al , we need to solve the Schrödinger equation in the scattering region where
V 6= 0, and match this to the exterior solution Eq.(11.25), using the appropriate bound-
ary conditions.
160 CHAPTER 11. SCATTERING

[Theorem]
If the potential V is non-zero at r < a and zero outside, the partial waves that mainly
contribute to the scattering satisfy:
p
l(l + 1) < ka. (11.26)

Let us show the theorem in a semi-classical way. A particle of mass m and velocity v
is injected toward the scattering potential with the impact parameter b. The angular
momentum of the particle is L = mvb.
√ p
2mE 2m(mv 2 /2) mv L
k= = = ⇒ b= . (11.27)
~ ~ ~ ~k
If the potential affects the volume only at r ≤ a, the particle is scattered when b is smaller
than a:
L
b<a ⇒ < a. (11.28)
~k
Replacing L2 with ~2 l(l + 1), we get
p
~ l(l + 1) p
<a ⇒ l(l + 1) < ka. (11.29)
~k
Therefore, in the low-energy regime, the scattering is dominated by the l = 0 term, which
means that the differential cross section is independent of θ, as it was in the classical
mechanics.

Quantum hard-sphere scattering


Let us consider scattering by the potential:

∞ (r ≤ a)
V (r) = (11.30)
0 (r > a).

The boundary condition is ψ(a, θ) = 0, so from Eq.(11.25),



X h i
l (1)
i (2l + 1) jl (ka) + ikal hl (ka) Pl (cos θ) = 0 (11.31)
l=0

X h iZ π
l (1)
⇒ i (2l + 1) jl (ka) + ikal hl (ka) Pl (cos θ)Pl0 (cos θ) sin θdθ = 0 (11.32)
l=0 0
11.2. PARTIAL WAVE ANALYSIS 161

Exploiting the orthogonality of the Legendre polynomials:


Z π  
2
Pl (cos θ)Pl0 (cos θ) sin θdθ = δll0 , (11.33)
0 2l + 1
we get
0
h i
(1)
2il jl0 (ka) + ikal0 hl0 (ka) = 0 (11.34)
jl (ka)
⇒ al = − (1)
(11.35)
ikhl (ka)
From Eq.(11.23), the total cross section is

2
4π X jl (ka)

σ= 2 (2l + 1) (1) (11.36)
k i=0

h (ka)
l

low-energy regime: ka  1
k = 2π/λ, so a  λ in this regime, which means that the wavelength of the particle is
much greater than the radius of the sphere. For small z = ka,
jl (z) jl (z) jl (z)
(1)
= ≈ −i (11.37)
hl (z) jl (z) + inl (z) nl (z)
 l 2
2l l!z l /(2l + 1)! i 2 l!
≈ −i −l−1 l
= z 2l+1 . (11.38)
−(2l)!z /(2 l!) 2l + 1 (2l)!
So, Eq.(11.36) becomes
∞  l 4
4π X 1 2 l!
σ≈ 2 (ka)4l+2 . (11.39)
k l=0 2l + 1 (2l)!

When ka  1, the higher powers are negligible, so


σ ≈ 4πa2 for ka  1 (11.40)
high-energy regime: ka  1
For z = ka  1,
(1) 1 (2) 1
hl= jl (z) + inl (z) ≈ (−i)l+1 eiz and hl = jl (z) − inl (z) ≈ (i)l+1 e−iz
z z
jl (z) 1 (−i)l+1 eiz + (i)l+1 e−iz 1
1 + (−1)l+1 e−2iz

⇒ (1)
≈ l+1 eiz
= (11.41)
hl (z) 2 (−i) 2
j (z) 2 1 

l
1 + (−1)l+1 cos(2z)

⇒ (1) ≈ (11.42)

h (z)
l
2
162 CHAPTER 11. SCATTERING

From Eq.(11.36), the total cross section is



2π X
(2l + 1) 1 + (−1)l+1 cos(2ka)
 
σ = 2
(11.43)
k l=0

= {[1 − cos(2ka)] + 3 [1 + cos(2ka)] + 5 [1 − cos(2ka)] + . . .} (11.44)
k2

= 2
[sin2 (ka) + 3 cos2 (ka) + 5 sin2 (ka) + . . .] (11.45)
k
4π  2 2 2 2 2 2

= [(sin (ka) + cos (ka))] + 2[sin (ka) + cos (ka)] + 3[sin (ka) + cos (ka)] + . . .]
k2
l
4π X 2π
= 2
lim m = 2 lim l(l + 1). (11.46)
k l→∞ m=1 k l→∞

This blows up, but we can use thepabove-mentioned theorem. Let the maximum l be
lmax . If ka  1, then lmax  1, so lmax (lmax + 1) ∼ ak. Then,

2π 2 2
σ≈ a k = 2πa2 . (11.47)
k2
According to numerical calculations,

σ = 2πa2 [1 + 0.996(ka)−2/3 + . . .]. (11.48)

11.2.2 phase shift


We can exploit conservation of angular momentum to reduce a complex quantity al ,
which contains two real numbers, to a single real number δl . Since angular momentum is
conserved by a spherically symmetric potential, each partial wave, labelled by l, scatters
independently, with a change only in phase. First, let us suppose that there is no potential
at all. Then, the wave function is just ψ(V =0) = Aeikz , and the l-th partial wave is, from
Eq.(11.25),
(l)
ψ(V =0) = Ail (2l + 1)jl (kr)Pl (cos θ). (11.49)

From Eq.(11.18) and Fig.11.6,


1 h (1) (2)
i 1 
(−i)l+1 eix + il+1 e−ix

jl (x) = hl (x) + hl (x) ≈ (for x  1) (11.50)
2 2x
Putting Eq.(11.50) into Eq.(11.49),

(l) (2l + 1)  ikr


e − (−1)l e−ikr Pl (cos θ) (for kr  1).

ψ(V =0) ≈ A (11.51)
2ikr
11.2. PARTIAL WAVE ANALYSIS 163

Now let us introduce the scattering potential. In Eq.(11.51), the second term e−ikr
represents the incoming spherical wave, which is not changed by the scattering potential.
But the first term eikr is the outgoing wave, which picks up a phase shift δl :
(2l + 1)  i(kr+2δl )
ψ (l) ≈ A − (−1)l e−ikr Pl (cos θ) (for kr  1).

e (11.52)
2ikr
Meanwhile, putting Eq.(11.50) into Eq.(11.25) and extracting the l-th component,
 
(l) (2l + 1)  ikr l −ikr
 (2l + 1) ikr
ψ ≈ A e − (−1) e + al e Pl (cos θ)
2ikr r
(2l + 1) 
(1 + 2ikal )eikr − (−1)l e−ikr Pl (cos θ) (for kr  1). (11.53)

= A
2ikr
Comparing Eq.(11.52) and Eq.(11.53),
1  1 1 1
al = e2iδl − 1 = eiδl sin(δl ) = . (11.54)
2ik k k cot δl − i
Putting Eq.(11.54) into Eq.(11.22) and Eq.(11.23),

1X
f (θ) = (2l + 1)eiδl sin(δl )Pl (cos θ), (11.55)
k l=0

4π X
σ = (2l + 1) sin2 (δl ). (11.56)
k 2 l=0

The Legendre polynomials satisfy Pl (1) = 1. So, from Eq.(11.55),



1X
f (0) = (2l + 1)eiδl sin(δl ) (11.57)
k l=0

1X
⇒ Im [f (0)] = (2l + 1) sin2 (δl ). (11.58)
k l=0

Comparing this equation with Eq.(11.56),



σ= Im [f (0)] . (11.59)
k
This relation between σ and f (θ) is called the optical theorem.
Now, let us discuss the phase shift δl from another angle. If V (r) = 0 at r > a, the radial
part of the Schrödinger equation is
~2 d2
 
2 d l(l + 1)
− + − R = ER (r > a) (11.60)
2m dr2 r dr r2
164 CHAPTER 11. SCATTERING

And the general solution is


ul (r)
Rl (r) ≡ = Ajl (kr) + Bnl (kr). (11.61)
r
As r → ∞,
   
lπ lπ
sin kr − cos kr −
ul (r) 2 2
Rl (r) = ∼A −B (r → ∞) (11.62)
r kr kr
If V (r) = 0 everywhere, then B = 0, because of the boundary condition ul (0) = 0.
Therefore, the ratio B/A represents the strength of the scattering. Eq.(11.62) can be
rewritten as
 

sin kr − + δl
ul (r) 2
Rl (r) = ∼C (11.63)
r  kr   
C lπ lπ
= sin kr − cos δl + cos kr − sin δl (r → ∞). (11.64)
kr 2 2
Comparing Eq.(11.62) and Eq.(11.64), we get
B
tan δl = − . (11.65)
A
If V (r) = 0 everywhere, then δl = 0. At very large r, the scattering only causes the phase
shift δl in the radial wave function of the particle.

sign of phase shift


As we saw above, at very large r,
  
A lπ
sin kr − (if V (r) = 0 everywhere)


k 2

ul (r) ≡ rRl (r) =   (11.66)
C lπ
sin kr − + δl (if V (r) 6= 0 for r < a)


k 2

In the case of attractive potential (V < 0), k becomes larger at r < a, and ul (0) = 0. So,
ul (r) must oscillate more rapidly compared with the case of V = 0. As a result, ul (r)
gets attracted towards the region of r < a, and the phase shift δl becomes positive. On
the other hand, if the potential is repulsive (V > 0), ul (r) gets pushed away towards the
region of r > a, and δl becomes negative.

V < 0 (attractive potential) : δl > 0
(11.67)
V > 0 (repulsive potential) : δl < 0
11.2. PARTIAL WAVE ANALYSIS 165

The whole theory of √ scattering reduces to the problem of calculating the phase shift δl
as a function of k ≡ 2mE/~. The partial wave analysis is useful especially when the
incident particle has low energy, because in that case only the first few terms in the series
X∞
contribute significantly.
l=0

Quantum hard-sphere scattering


From Eq.(11.54) and Eq.(11.35),
 
nl (ka)
+i
iδl jl (ka) jl (ka)
e sin δl = i (1) =  2 . (11.68)
hl (ka) nl (ka)
1+
jl (ka)
Equating the real and imaginary parts:
 
nl (ka)
jl (ka) 1
cos δl sin δl =  2 , sin2 δl =  2 (11.69)
nl (ka) nl (ka)
1+ 1+
jl (ka) jl (ka)
jl (ka)
⇒ tan δl = . (11.70)
nl (ka)
low-energy regime (ka  1)
The S-wave (l = 0) dominates the scattering, and
sin(ka)
j0 (ka) ka
tan δ0 = = = − tan(ka) = tan(−ka) (11.71)
n0 (ka) cos(ka)

ka
⇒ δ0 = −ka. (11.72)
high-energy regime (ka  1)
Let us use the asymptotic expression of this form:
sin(ka − lπ/2) cos(ka − lπ/2)
jl (ka) ≈ , nl (ka) ≈ − for ka  1. (11.73)
ka ka
Then, from Eq.(11.70),
 

tan δl = − tan ka − (11.74)
2

⇒ δl = −ka + . (11.75)
2
166 CHAPTER 11. SCATTERING

11.3 Scattering by Finite Well Potential


Let us consider the scattering by the potential:

−V0 (r < a)
V (r) = (11.76)
0 (r > a)
The radial part of the Schrödinger equation is
~2 d2
 
2 d l(l + 1)
− + − R = ER (r > a) (11.77)
2m dr2 r dr r2
~2 d2
 
2 d l(l + 1)
− + − R = (E + V0 )R (r < a) (11.78)
2m dr2 r dr r2
Note that E > 0, since we are now discussing the scattering state. The general solutions
are

2mE
Rl (r) = Ajl (kr) + Bnl (kr), where k ≡ (r > a) (11.79)
p ~
2m(E + V0 )
Rl (r) = Cjl (αr), where α = (r < a). (11.80)
~
The boundary conditions of Rl and dRl /dr at r = a yield:

Ajl (ka) + Bnl (ka) = Cjl (αa)
(11.81)
Akjl0 (ka) + Bkn0l (ka) = Cαjl0 (αa),

where jl0 (r) ≡ djl (r)/dr and n0l (r) ≡ dnl (r)/dr. Multiplying the upper equation in Eq.(??)
by αjl0 (αa), and the lower equation by jl (αa), we get

αjl (ka)jl0 (αa)A + αjl0 (αa)nl (ka)B = αjl (αa)jl0 (αa)C



(11.82)
kjl (αa)jl0 (ka)A + kjl (αa)n0l (ka)B = αjl (αa)jl0 (αa)C.
B B
⇒ αjl (ka)jl0 (αa) + αjl0 (αa)nl (ka) = kjl (αa)jl0 (ka) + kjl (αa)n0l (ka) (11.83)
A A
0 0
B kj (ka)jl (αa) − αjl (ka)jl (αa)
⇒ = l . (11.84)
A αnl (ka)jl0 (αa) − kn0l (ka)jl (αa)
From this equation and Eq.(11.65),
αjl (ka)jl0 (αa) − kjl0 (ka)jl (αa)
tan δl = . (11.85)
αnl (ka)jl0 (αa) − kn0l (ka)jl (αa)

11.3.1 High-energy regime (ka  1)


V0  E, so k ≈ α and tan δl ≈ 0 ⇒ δl ≈ 0.
11.3. SCATTERING BY FINITE WELL POTENTIAL 167

11.3.2 Low-energy regime (ka  1)


Case of l = 0
The radial wave function at r > a can be written as:
sin(kr) cos(kr)
R0 (r) = Aj0 (kr) + Bn0 (kr) = A −B (11.86)
kr kr
sin(kr + δ0 )
= C (for r > a), (11.87)
kr
while at r < a
sin(αr)
R0 (r) = D (for r < a). (11.88)
αr
The boundary conditions of R0 and dR0 /dr at r = a yield
α cot(αa) = k cot(ka + δ0 ). (11.89)
Since α cot(αa) is non-zero, cot(ka + δ0 ) → ∞ (i.e. sin(ka + δ0 ) → 0) as k → 0. Thus,
in the limit of k → 0, sin(ka + δ0 ) ∼ ka + δ0 . Therefore,
 
k tan(αa)
α cot(αa) ≈ ⇒ δ0 = ka −1 (11.90)
ka + δ0 αa
2
sin2 δ0

dσ 2 2 tan(αa)
∴ = |f (θ)| = =a −1 (11.91)
dΩ k2 αa
 2
2 tan(αa)
∴ σ = 4πa −1 . (11.92)
αa
If the energy of the injected particle satisfies tan(αa) = αa, then σ = 0.
On the other hand, if αa = (1/2 + n) π (n = 0, 1, 2, . . .), tan(αa) → ∞ and σ → ∞.
In general, when the cross section becomes especially large on a certain condition, we
call it the resonance scattering. In Eq.(11.89), cot(αa) = 0, so cos(ka + δ0 ) = 0 and
sin(ka + δ0 ) = 1. Thus, sin δ0 ≈ 1 because ka  1. Therefore, with respect to the
S-wave, resonance scattering occurs when αa = (1/2 + n) π with the total cross section
σ = 4π/k 2 . The total cross section becomes extremely large as the energy of the particle
becomes very small. This phenomenon is called the zero-energy resonance. Typically
resonance scattering occurs when the energy of the injected particle is equal to the energy
level of a bound state. Actually, the potential Eq.(11.76) has a bound state with E = 0;
the energies of the bound states are determined by the equation:
r
α0 a 2
− cot(αa) = − 1, (11.93)
αa
p √
2m(E + V0 ) 2mV0
where α ≡ and α0 ≡ . (11.94)
~ ~
168 CHAPTER 11. SCATTERING

Letting αb be a solution to Eq.(11.93), the energy of the bound state is obtained by


~2 2
Eb = (α − α02 ). (11.95)
2m b
 
1
If α0 a = + n π, where n = 0, 1, 2, . . ., then there is a solution:αb = α0 , so Eb = 0.
2

Case of l ≥ 1
We use the asymptotic expressions:
2l l! (2l)!
jl (ka) ≈ (ka)l , nl (ka) ≈ − (ka)−l−1 (for ka  1) (11.96)
(2l + 1)! 2l l!
Putting Eq.(11.96) into Eq.(11.85), we get
2
2l l! ljl (αa) − αajl0 (αa)

tan δl ≈ (2l + 1)(ka)2l+1 . (11.97)
(2l + 1)! (l + 1)jl (αa) + αajl0 (αa)
Now let us suppose that the potential is deep enough: αa  1. Resonance scattering
occurs when cot δl = 0 (i.e. tan δl = ∞), that is,
(l + 1)jl (αa) + αajl0 (αa) = 0. (11.98)
Using the asymptotic expression:
sin (αa − lπ/2)
jl (αa) ≈ , (11.99)
αa
Eq.(11.98) becomes:
   
l+1 lπ lπ
sin αa − + cos αa − = 0 (11.100)
αa 2 2
 
(l + 1)π l+1
⇒ tan αa − = (11.101)
2 αa
l+1
 1, so
αa
(l + 1)π l+1
αa − ≈ nπ + . (11.102)
2 αa
|E|  V0 , so
p √  
2m(E + V0 ) 2mV0 E
α= ≈ 1+ . (11.103)
~ ~ 2V0
11.3. SCATTERING BY FINITE WELL POTENTIAL 169

l+1
And ≈ 0, so Eq.(11.102) becomes
αa

~2
 
E l+1
= −1 + n+ π. (11.104)
2V0 2mV0 a2 2

When the energy of the injected particle is equal to E given by Eq.(11.104), the injected
particle is temporarily trapped inside the potential, forming a quasi-bound state (i.e.
resonance state). After a certain lifetime, the trapped particle escapes out of the poten-
tial via the tunneling effect. This is the qualitative description of resonance scattering.
Although we discussed resonance scattering in the low-energy regime, it occurs generally
when an attractive potential with a finite range is involved.

11.3.3 Breit-Wigner’s formula


Letting the resonance energy be ER , let us expand cot δl as
 
d cot δl
cot δl = cot δl |E=ER + (E − ER ) + O[(E − ER )2 ] (11.105)
dE
E=ER
2
= − (E − ER ) + O[(E − ER )2 ], (11.106)
Γ

2
where Γ ≡ −   . Putting this into Eq.(11.54),
d cot δl
dE
E=ER

Γ/2
al = − (11.107)
k[E − ER + iΓ/2]

Putting this into Eq.(11.23), the partial cross section can be written as
 2
Γ
4π 2
σl = 4π(2l + 1)|al |2 = (2l + 1)  2 . (11.108)
k2 Γ
(E − ER )2 +
2

This is called Breit-Wigner’s formula. When E = ER ± Γ/2, σl drops to half of the


peak value. Γ is called the resonance width.
170 CHAPTER 11. SCATTERING

11.4 The Born Approximation


11.4.1 Integral form of the Schrödinger equation
Let us rewrite the time-independent Schrödinger equation.
~2 2
− ∇ ψ(r) + V (r)ψ(r) = Eψ(r) (11.109)
2m
⇒ (∇2 + k 2 )ψ(r) = Q(r), (11.110)

2mE 2m
where k ≡ , and Q(r) ≡ 2 V (r)ψ(r) (11.111)
~ ~
This is the inhomogeneous Helmholtz equation, but the inhomogeneous term Q de-
pends on ψ itself.
Let us look for a function G(r) that is a solution to the equation:
(∇2 + k 2 )G(r) = δ 3 (r). (11.112)
Then, ψ can be written as an integral:
Z
ψ(r) = G(r − r0 )Q(r0 )d3 r0 , (11.113)

because Eq.(11.113) satisfies the Schrödinger equation Eq.(11.110) in this way:


Z
2 2
 2
(∇ + k 2 )G(r − r0 ) Q(r0 )d3 r0

(∇ + k )ψ(r) =
Z
= δ 3 (r − r0 )Q(r0 )d3 r0 = Q(r) (11.114)

G(r) is called the Green’s function for the Helmholtz equation. To solve Eq.(11.112),
we take the Fourier transform:
Z
1
G(r) = eis·r g(s)d3 s. (11.115)
(2π)3/2
Then the left side of Eq.(11.112) becomes
Z
2 2 1  2
(∇ + k 2 )eis·r g(s)d3 s

(∇ + k )G(r) = 3/2
(2π)
Z
1
(−s2 + k 2 )eis·r g(s)d3 s
 
= 3/2
(11.116)
(2π)
Meanwhile the Fourier transform of the delta function is
Z
3 1
δ (r) = eis·r d3 s. (11.117)
(2π)3
11.4. THE BORN APPROXIMATION 171

Eq.(11.112) says that Eq.(11.116) is equal to Eq.(11.117):


Z Z
1 2 2 is·r 3 1
eis·r d3 s.
 
(−s + k )e g(s)d s =
(2π)3/2 (2π)3
Z  
2 2 1
⇒ (−s + k )g(s) − eis·r d3 s = 0. (11.118)
(2π)3/2
Applying the Plancherel’s theorem2 to Eq.(11.118), we get
1
g(s) = (11.120)
(2π)3/2 (k 2 − s2 )
Putting this equation back into Eq.(11.115),
Z
1 1
G(r) = 3
eis·r 2 d3 s. (11.121)
(2π) (k − s2 )
Let us carry out the s integration in spherical coordinates (s, θ, φ). Setting the polar axis
along r, we get s · r = s r cos θ. So,
Z ∞ Z π Z 2π
1 1
G(r) = 3
ds dθ dφ s2 sin θ eisr cos θ 2 . (11.122)
(2π) 0 0 0 (k − s2 )
The φ integral just gives 2π, and the θ integral is
Z π π
isr cos θ eisr cos θ 2 sin(sr)
e sin θdθ = − = , (11.123)
0 isr 0 sr
so G(r) becomes
Z ∞ Z ∞
1 2 s sin(sr) 1 s sin(sr)
G(r) = ds = 2 ds. (11.124)
(2π)2 r 0
2
k −s 2 4π r −∞ k 2 − s2
Using sin(sr) = (1/2i)(eisr − e−isr ),
Z ∞ Z ∞
seisr se−isr

i
G(r) = ds − as
8π 2 r −∞ (s − k)(s + k) −∞ (s − k)(s + k)
i
≡ (I1 − I2 ). (11.125)
8πI2 r 
seisr
  isr 
1 se
I1 = + ds = 2πi = iπeikr (11.126)
s+k s−k s + k s=k
I  −isr   −isr 
se 1 se
I2 = − ds = −2πi = −iπeikr (11.127)
s−k s+k s−k s=−k
2
Z Z
f (r) = F (s)eis·r d3 s ⇔ F (s) = f (r)e−is·r d3 r (11.119)
172 CHAPTER 11. SCATTERING

Therefore, the solution to Eq.(11.112) is


i  ikr
 ikr
 eikr
G(r) = iπe − −iπe = − . (11.128)
8π 2 r 4πr
But we can add to G(r) any function G0 (r) that satisfies the homogeneous Helmholtz
equation:
(∇2 + k 2 )G0 (r) = 0, (11.129)
because (G+G0 ) also satisfies Eq.(11.112). Therefore, the general solution to Eq.(11.110)
is
Z ik|r−r0 |
m e
ψ(r) = ψ0 (r) − 2
V (r0 )ψ(r0 )d3 r0 , (11.130)
2π~ |r − r0 |
where ψ0 (r) satisfies
(∇2 + k 2 )ψ0 (r) = 0. (11.131)
Eq.(11.130) is the integral form of the Schrödinger equation.

11.4.2 First Born Approximation


Suppose V (r0 ) is localized around r0 = 0 and that we want to obtain ψ(r) at points far
away from the scattering potential. Then we can assume |r|  |r0 | in Eq.(11.130), so
 r · r0 
|r − r0 |2 = r2 + r02 − 2r · r0 ≈ r2 1 − 2 2
r
 r · r0 1/2  r · r0 
⇒ |r − r0 | ≈ r 1 − 2 2 ≈r 1− 2 = r − r̂ · r0 . (11.132)
r r
Let k ≡ kr̂. Then
eik|r−r0 | ≈ eik(r−r̂·r0 ) = eikr e−ik·r0
eik|r−r0 | eikr −ik·r0
⇒ ≈ e (11.133)
|r − r0 | r
We are now talking about scattering, so ψ0 (r) = Aeikz . Putting Eq.(11.133) into
Eq.(11.130),
m eikr
Z
ikz
ψ(r) ≈ Ae − e−ik·r0 V (r0 )ψ(r0 )d3 r0 (for large r). (11.134)
2π~2 r
Comparing this equation with Eq.(11.7),
Z
m
f (θ, φ) = − 2
e−ik·r0 V (r0 )ψ(r0 )d3 r0 . (11.135)
2π~ A
11.4. THE BORN APPROXIMATION 173

Now the Born approximation — we assume that the potential is weak compared to the
incident particle’s energy, so that the deflection caused by the scattering is very small:
0
ψ(r0 ) ≈ ψ0 (r0 ) = Aeikz0 = Aeik ·r0 , where k0 = kẑ. (11.136)

Then Eq.(11.135) becomes


Z
m 0
f (θ, φ) ≈ − ei(k −k)·r0 V (r0 )d3 r0 . (11.137)
2π~2

Note that k points to the detector while k0 points in the direction of the incident beam.
~(k −k0 ) is the momentum transfer in the scattering process. Let us define κ ≡ k0 −k
and set the polar axis for the r0 integral along κ, so that

(k0 − k) · r0 = κr0 cos θ0 . (11.138)

For a spherically symmetrical potential, V (r) = V (r). In this case, Eq.(11.137) becomes
Z
m
f (θ) ≈ − 2
eiκr0 cos θ0 V (r0 )r02 sin θ0 dr0 dθ0 dφ0 . (11.139)
2π~

The φ integral just gives 2π, and the θ integral is


π
π
eiκr0 cos θ0
Z
iκr0 cos θ0 2 sin(κr0 )
e sin θ0 dθ0 = − = , (11.140)
0 iκr0 0 κr0

so, Eq.(11.139) can be written as


Z ∞  
2m θ
f (θ) ≈ − 2 r0 V (r0 ) sin(κr0 )dr0 where κ = 2k sin , (11.141)
~κ 0 2

for a spherically symmetrical potential.

11.4.3 Limit of application of Born approximation


The Born approximation is applicable when the potential is weak compared to the inci-
dent particle’s energy, so that the deflection caused by the scattering is very small, that
is, in Eq.(11.130), ψ(r) is almost the same as ψ0 (r) = eikz within the region of V 6= 0.
We can express this condition as

m eikr0
Z
ikz 3

|e | = 1  2
V (r0 )ψ(r0 )d r0 . (11.142)
2π~ r0
174 CHAPTER 11. SCATTERING

Finite well potential


Let us estimate the condition Eq.(11.142) for the potential:

−V0 (r < a)
V (r) = (11.143)
0 (r > a).

Replacing ψ(r0 ) with ψ0 (r0 ) = eikr0 cos θ0 in Eq.(11.142),

eikr0
Z
m 2

ikr0 cos θ0
1  r0 sin θ0 dr0 dθ 0 dφ 0 (−V 0 )e (11.144)
2π~2 r
Z a Z π 0 
m
= 2
(2π)V0 reikr eikr cos θ sin θdθ dr (11.145)
2π~ 0 0
mV0 2ika
= |e − 1 − 2ika| (11.146)
2~2 k 2
low-energy case: ka  1
1
Using the Taylor expansion: e2ika ' 1 + 2ika + (2ika)2 , Eq.(11.146) becomes
2
mV0 2 2 mV0 a2
1 (2k a ) = . (11.147)
2~2 k 2 ~2
Remember that there is no bound state if

2mV0 π mV0 a2 π2
a< ⇒ < . (11.148)
~ 2 ~2 8
So, when the potential well is shallow enough to satisfy Eq.(11.147), there is no bound
state.
high-energy case: ka  1
Eq.(11.146) becomes |e2ika − 1 − 2ika| ≈ 2ka, so

mV0 ma2 V0 1
1 2ka = . (11.149)
2~2 k 2 ~2 ka
Born approximation is always valid in the high-energy regime.

11.4.4 Yukawa scattering


The Yukawa potential is a crude model for the binding force in an atomic nucleus:

e−µr
V (r) = β , (11.150)
r
11.4. THE BORN APPROXIMATION 175

where β and µ are constants. Putting this into Eq.(11.141),


2mβ ∞ −µr
Z
2mβ
f (θ) ≈ − 2 e sin(κr)dr = − 2 2 (11.151)
~κ 0 ~ (µ + κ2 )
Z  2
4mβ 1
⇒ σ = |f (θ)|2 sin θdθdφ = π 2
. (11.152)
µ~ (µ~) + 8mE
Replacing ψ(r0 ) with ψ0 (r0 ) = eikr0 cos θ0 in Eq.(11.142), the condition for the validity of
Born approximation is
ikr −µr
Z
m 3 e e ikr cos θ

1  2
dr β e (11.153)
2π~ r r
 
m|β| 2ik
= ln 1 − . (11.154)
~2 k µ
low-energy
 case
2ik 2ik
ln 1 − ≈− , so
µ µ

m|β| 2ik 2m|β|
1 2 = . (11.155)
~k µ µ~2
high-energy case
       
2ik 2ik 2k π 2k
ln 1 − ≈ ln − = ln (−i) + ln = −i + ln . (11.156)
µ µ µ 2 µ
Therefore,
 
m|β| π 2k
1  −i + ln (11.157)
~2 k 2 µ
 
m|β| 2k
⇒ 1  2
ln . (11.158)
~k µ
This condition is always satisfied in the limit of k → ∞.

Rutherford scattering: Classical Coulomb scattering


Suppose that an incident particle of charge q1 and kinetic energy E scatters off a heavy
stationary particle of charge q2 . Let β = q1 q2 /4π0 and µ = 0 in Eq.(11.150), the Yukawa
potential reduces to the Coulomb potential. So,
2mq1 q2 q1 q2
f (θ) ≈ − = − (11.159)
4π0 ~2 κ2 16π0 E sin2 (θ/2)
Z  2 Z π
2 q1 q 2 sin θ
σ = |f (θ)| sin θdθdφ = 2π 4 dθ. (11.160)
16π0 E 0 sin (θ/2)
176 CHAPTER 11. SCATTERING

The integral in Eq.(11.160) does NOT convergeZ  — near θ = 0, we have sin θ ≈ θ and

sin(θ/2) ≈ θ/2, so the integral goes like 16 θ−3 dθ = −8θ−2 0 → ∞. This result means
0
that the 1/r potential has infinite range.

11.4.5 Born approximation for phase shift

To obtain the phase shift δl (k), let us get back to the radial part of the Schrödinger
equation:

d2
 
l(l + 1) 2
− − U (r) + k ul (r) = 0, (11.161)
dr2 r2
2m 2mE
where U = 2 V (r), k 2 = . (11.162)
~ ~2

Letting χl (r) be the solution to Eq.(11.161), and ϕl (r) be the solution to Eq.(11.161) in
the case of V (r) = 0 everywhere, that is,

d2
 
l(l + 1) 2
U (r)χl (r) = − + k χl (11.163)
dr2 r2
 2 
d l(l + 1) 2
0= − + k ϕl . (11.164)
dr2 r2

The boundary condition at r = 0 says χl (0) = ϕl (0) = 0.


Eq.(11.163) ×ϕl − Eq.(11.164) ×χl yields

d2 χl d2 ϕl
U χl ϕl = ϕl 2 − χl 2 . (11.165)
dr dr

Recalling Eq.(11.66), at very large r,

  
A lπ
 ϕl = sin kr − (if V (r) = 0 everywhere)


k 2
  (11.166)
C lπ
 χl = sin kr − + δl (if V (r) 6= 0 for r < a)


k 2
11.4. THE BORN APPROXIMATION 177

So, integrating Eq.(11.165) in terms of r,


Z ∞ Z ∞
d2 χ l d2 ϕl

U χl ϕl dr = ϕ l 2 − χl 2 (11.167)
0 0 dr dr
 ∞
dχl dϕl
= ϕl − χl (11.168)
dr dr
  0   
AC lπ lπ
= lim sin kr − k cos kr − + δl (11.169)
k 2 r→∞ 2 2
   
lπ lπ
− sin kr − + δl k cos kr − (11.170)
2 2
AC
= − sin δl . (11.171)
k
ϕl can be written as Arjl (kr). And when U (r) is very small, χl ≈ ϕl and δl is very small.
So,
Z ∞
A 2mk ∞ 2 2
Z
k 2 2 2
sin δl ≈ δl = − U A r jl (kr)dr = − V r jl (kr)dr. (11.172)
AC 0 C ~2 0
2l l!
Using the asymptotic form jl (kr) ∼ (kr)l at kr  1, we get
(2l + 1)!

δl (k) ∝ k 2l+1 (for kr  1), (11.173)

which shows that in the low-energy regime the S-wave (i.e. l = 0) dominates in the
scattering by a potential with a finite range.

11.4.6 Born series


Let us rewrite Eq.(11.130) as

m eikr
Z
ψ(r) = ψ0 (r) + g(r − r0 )V (r0 )ψ(r0 )d3 r0 , where g(r) ≡ − . (11.174)
2π~2 r
ψ0 (r) is the incident wave, g(r) is the Green’s function, and V is the scattering potential.
Schematically,
Z
ψ = ψ0 + g V ψ
Z Z
= ψ0 + g V (ψ0 + g V ψ)
Z Z Z
= ψ0 + g V ψ0 + g V g V ψ. (11.175)
178 CHAPTER 11. SCATTERING

Iterating this procedure,


Z Z Z Z Z Z
ψ = ψ0 + g V ψ0 + g V g V ψ0 + g V g V g V ψ0 + . . . (11.176)

The first Born approximation truncates the series after the second term, but one can
calculate the higher-order corrections according to Eq.(11.176). Fig.11.7 is the interpre-
tation of the Born series in diagrams. The Green’s function g is called the propagator,
and the scattering potential V is called the vertex factor. This diagrammatic interpre-
tation of scattering leads to the Feynman diagram of relativistic quantum mechanics.

Figure 11.7: Interpretation of the Born series in diagrams.

11.5 Quantum Coulomb Scattering


The scattering amplitude of Coulomb Scattering can be obtained by solving the Schrödinger
equation directly. Suppose that a particle of charge Z1 e is interacting with another par-
ticle of charge Z2 e in the Coulomb potential. The Schrödinger equation for the relative
motion can be written as
 2
~ 2 Z1 Z2 e2 1

− ∇ + ψ(r) = Eψ(r), (11.177)
2µ 4π0 r
m1 m2
where µ ≡ , (11.178)
m1 + m2
Eq.(11.177) can be rewritten as
 
2 2 2γk
∇ +k − ψ(r) = 0, (11.179)
r
~2 k 2 1 Z1 Z2 e2
where E = = µv 2 , γ= . (11.180)
2µ 2 4π0 ~v
Note that E > 0, since we are discussing the scattering state.
11.5. QUANTUM COULOMB SCATTERING 179

Figure 11.8: The xz-plane in parabolic coordinates. The dashed lines represent lines with
η = const., while solid lines represent lines with ξ = const. The dashed lines and solid
lines are perpendicular to each other. By definition, ξ ≥ 0 and η ≥ 0.

11.5.1 Parabolic coordinates


To solve Eq.(11.179), we use parabolic coordinates (ξ, η, φ), which is related to spherical
coordinates as

ξ = r(1 − cos θ), η = r(1 + cos θ), φ = φ. (11.181)

The transformation of derivatives are


 
∂ 2 ∂ ∂
= ξ +η , (11.182)
∂r ξ+η ∂ξ ∂η
  r s
∂ p ∂ ∂ η ∂ ξ ∂
= ξη − = ξ − η . (11.183)
∂θ ∂ξ ∂η ξ ∂ξ η ∂η

In parabolic coordinates, Eq.(11.179) can be written as

4 1 ∂ 2ψ
    
4 ∂ ∂ψ ∂ ∂ψ 4kγ
ξ + η + 2
+ k2ψ − ψ = 0. (11.184)
ξ + η ∂ξ ∂ξ ∂η ∂η ξ + η ξη ∂φ ξ+η

The Coulomb potential is spherically symmetric. Let us set the z axis along the direction
of the injected particle at very large r, so that the scattering is symmetric about the z
180 CHAPTER 11. SCATTERING

∂ 2ψ
axis and ψ does not depend on φ. Then, the term in Eq.(11.184) vanishes. Using
∂φ2
the separation of variables:

ψ(ξ, η) = g(ξ)h(η), (11.185)

and multiplying Eq.(11.184) by (ξ + η)/ψ,

k2 k2
       
1 d d 1 d d
ξ + ξ g+ η + η h = kγ. (11.186)
g dξ dξ 4 h dη dη 4

Finally, we get

k2
   
d d
ξ + ξ − c1 g(ξ) = 0 (11.187)
dξ dξ 4
k2
   
d d
η + η − c2 h(η) = 0 (11.188)
dη dη 4
c1 + c2 = kγ. (11.189)

11.5.2 Confluent hypergeometric function


We can naturally assume

ψ → eikz (z → −∞). (11.190)

In parabolic coodinates, z → −∞ corresponds to ξ → ∞ regardless of η.


Using z = (−ξ + η)/2, the asymptotic form of ψ can be expressed as

ψ(ξ, η) → eikz = eikη/2 e−ikξ/2 (ξ → ∞). (11.191)

Since Eq.(11.191) must hold for any η, h(η) is determined as

h(η) = eikη/2 . (11.192)

Let us rewrite g(ξ) as

g(ξ) = e−ikξ/2 w(ξ) (11.193)


⇒ ψ(ξ, η) = eikη/2 e−ikξ/2 w(ξ). (11.194)

Putting Eq.(11.193) into Eq.(11.187),

d2 w dw
ξ + (1 − ikξ) − γkw = 0. (11.195)
dξ 2 dξ
11.5. QUANTUM COULOMB SCATTERING 181

Changing the variables ξ 0 = ikξ, Eq.(11.195) becomes

d2 w dw
ξ0 02
+ (1 − ξ 0 ) 0 + iγw = 0. (11.196)
dξ dξ

We compare this equation with the confluent hypergeometric differential equation:

d2 F dF
z + (b − z) − aF = 0. (11.197)
dz 2 dz
A solution to Eq.(11.197) that becomes a constant at z = 0 is called the confluent
hypergeometric function, and is given in the form of series:

1a 1 a(a + 1) 2 1 a(a + 1)(a + 2) 3


F (a, b, z) = 1 + z+ z + z + . . . (11.198)
1! b 2! b(b + 1) 3! b(b + 1)(b + 2)

X 1 Γ(a + s) Γ(b) s
= z . (11.199)
s=0
Γ(1 + s) Γ(a) Γ(b + s)

So, the solution to Eq.(11.196) is

w(ξ) = CF (−iγ, 1, ikξ), (11.200)

where C is a normalization constant.

11.5.3 Asymptotic expansion


To get the scattering amplitude, we need to derive the asymptotic form of ψ as r → ∞.
Let us use the asymptotic expansion of the confluent hypergeometric function as |z| → ∞:

Γ(b) X Γ(n + a) Γ(n + a − b + 1) 1
F (a, b, z) ∼ (−z)−n−a
Γ(b − a) n=0 Γ(a) Γ(a − b + 1) n!

Γ(b) z X Γ(n + b − a) Γ(n + 1 − a) 1 −n+a−b
+ e z . (11.201)
Γ(a) n=0 Γ(b − a) Γ(1 − a) n!

Putting a = −iγ, b = 1 and z = ikξ into this equation, we get


∞  2
1 X Γ(n − iγ) 1
w(ξ) ∼ C (−ikξ)−n+iγ
Γ(1 + iγ) n=0 Γ(−iγ) n!
∞  2
1 X Γ(n + 1 + iγ) 1
+C e ikξ
(ikξ)−n−iγ−1 . (11.202)
Γ(−iγ) n=0
Γ(1 + iγ) n!
182 CHAPTER 11. SCATTERING

With k > 0 and ξ ≥ 0 in mind,


h  π i
(−ikξ)−n+iγ = exp (−n + iγ) ln kξ − i (11.203)
2h
h  nπ i γπ i
= exp i γ ln kξ + exp −n ln kξ +
2 2
−n−iγ−1
h π i
(ikξ) = exp (−n − 1 − iγ) ln kξ + i (11.204)
  2
(n + 1)π h γπ i
= exp −i γ ln kξ + exp −(n + 1) ln kξ + .
2 2

Putting these expressions into Eq.(11.202),

eγπ/2 iγ 2 γ 2 (i + γ)2
 
(1st term) = C exp (iγ ln kξ) 1 + − + . . . (11.205)
Γ(1 + iγ) kξ 2(kξ)2
eγπ/2 (i + iγ)2
 
i
(2nd term) = C exp (ikξ − iγ ln kξ) + + ... (11.206)
Γ(−iγ) kξ (kξ)2

Thus,

ψ(ξ, η) = eikz w(ξ)


eγπ/2 iγ 2
 
∼ C exp (ikz + iγ ln kξ) 1 + + ...
Γ(1 + iγ) kξ
eγπ/2 i
−C exp (ikz + ikξ − iγ ln kξ) . (11.207)
Γ(−iγ) kξ
In the second term,

exp(ikz + ikξ) = exp(ikz + ik(r − z)) = exp(ikr), (11.208)

which represents the outgoing spherical wave. And ξ = r(1 − cos θ) = 2r sin2 (θ/2), so
   
2 θ 2 θ
ln kξ = ln 2kr sin = ln(2kr) + ln sin . (11.209)
2 2

Finally Eq.(11.207) becomes

eγπ/2 iγ 2
 
ψ(ξ, η) ∼ C exp [i (kz + γ ln k(r − z))] 1 + + ...
Γ(1 + iγ) k(r − z)
  
2 θ
exp −iγ ln sin
eγπ/2 1 2
+C exp [i(kr − γ ln 2kr)] . (11.210)
Γ(−iγ) r 2 θ
2ik sin
2
11.5. QUANTUM COULOMB SCATTERING 183

Letting C = Γ(1 + iγ)e−γπ/2 ,

iγ 2
 
ψ(ξ, η) ∼ exp [i (kz + γ ln k(r − z))] 1 + + ...
k(r − z)
1
+f (θ) exp [i(kr − γ ln 2kr)] , (11.211)
r   
2 θ
exp −iγ ln sin
Γ(1 + iγ) 2
where f (θ) = . (11.212)
Γ(−iγ) 2 θ
2ik sin
2
In the right-hand side of ψ(ξ, η), the first term represents the injected wave (∝ eikz ) while
the second term represents the scattered wave (∝ eikr /r).

11.5.4 Rutherford’s formula


To calculate the differential cross section, we use
r
π
Γ(1 + z) = zΓ(z), |Γ(iy)| = (y ∈ R) (11.213)
y sinh(πy)

Then
r r
γπ π
|Γ(1 + iγ)| = , |Γ(−iγ)| = (11.214)
sinh γπ γ sinh γπ

Γ(1 + iγ)
⇒ = γ. (11.215)
Γ(−iγ)

Recalling Eq.(11.180),

γ 1 ~ Z1 Z2 e2 Z1 Z2 α~c
= = , (11.216)
2k 2 µv 4π0 ~v 2µv 2
e2 1
where α = = . (11.217)
4π0 ~c 137
So,
 2
dσ Z1 Z2 α~c 1
= |f (θ)|2 = . (11.218)
dΩ 2µv 2 θ4
sin
2
This is exactly the same as Rutherford’s formula from classical mechanics.
184 CHAPTER 11. SCATTERING

11.5.5 Probability current


The probability current is given by
 
i~ ∗ ∗ 1 ∗~
j(r) = (ψ∇ψ − ψ ∇ψ) = Re ψ ∇ψ . (11.219)
2m m i

In Eq.(11.211), the first term includes eikz and is considered to represent the injected
wave, but it has a logarithmic distortion γ ln k(r − z) even infinitely far away from the
scattering potential. Nevertheless, the probability current of the first term becomes

~k
jin = ẑ = vẑ (z → −∞), (11.220)
m

which ensures that the first term corresponds to the injected wave. Meanwhile, the
second term also has a logarithmic distortion −iγ ln 2kr even infinitely far away from the
scattering potential. But the probability current of the second term becomes

|f (θ)|2 |f (θ)|2
jsc = v r̂ = jin r̂, (11.221)
r2 r2

which ensures that the second term corresponds to the scattered wave.

11.6 Scattering between identical particles


In the case of the scattering between two identical particles, the wave function of the
whole system must be symmetric or anti-symmetric with respect to particle exchange,
according to whether the particles have integer spin (i.e. bosons) or half-integer spin (i.e.
fermions). In the center of mass system, particle exchange corresponds to the operation:
(r, θ, φ) → (r, π − θ, π + φ). Thus, for the asymptotic form of the wave function for
kr  1, we need to replace Eq.(11.7) with

eikr
 
ikz −ikz
ψ(r, θ) ≈ A e ±e + [f (θ) ± f (π − θ)] . (11.222)
r

Note that the + and − signs do not necessarily correspond to bosons and fermions,
respectively — the symmetrization/anti-symmetrization must be done with respect to
the overall wave function including not only the orbital part but also the spin part.
Eq.(11.222) only takes into account the orbital part. Using Eq.(11.59), Eq.(11.54) and
11.6. SCATTERING BETWEEN IDENTICAL PARTICLES 185

Eq.(11.55), the total cross section is


σI = Im[f (0) ± f (π)] (11.223)
k " ∞
#
4π 1 X
= Im (2l + 1)(e2iδl − 1) (Pl (cos(0)) ± Pl (cos(π))) (11.224)
k 2ik l=0
" ∞
#
4π 1 X
(2l + 1)(e2iδl − 1) 1 ± (−1)l

= Im (11.225)
k 2ik l=0

π X ∗
(2l + 1) 1 ± (−1)l (2 − e2iδl − e−2iδl ).

= 2 (11.226)
k l=0


So, the partial cross section attributed to the l-wave has a maximum (2l + 1). On the
k2
other hand, if the two particles are distinguishable particles, the term ±(−1)l vanishes

and the maximum of the partial cross section is 2 (2l + 1).
k
Let us discuss the following three cases for instance.

11.6.1 Elastic scattering between two spin-0 particles


There are two indistinguishable processes as shown in Fig.11.9. The total scattering

Figure 11.9: elastic scattering between two identical particles


186 CHAPTER 11. SCATTERING

amplitude is the sum of the amplitudes for (a) and (b):

f I (θ) = f (θ) + f (π − θ), (11.227)


 I

= |f (θ) + f (π − θ)|2 (11.228)
dΩ
= |f (θ)|2 + |f (π − θ)|2 + 2Re[f ∗ (θ)f (π − θ)]. (11.229)
 I

is symmetrical with respect to θ = 90◦ in the center of mass system. If the two
dΩ

particles were distinguishable, would be just |f (θ)|2 + |f (π − θ)|2 . The last term
dΩ
represents the identity of the two particles. The total cross section can be obtained by
integrating Eq.(11.229) with respect to the total solid angle, but we need to divide by 2
because, doing the integration, we end up counting each state twice. So,
Z
I 1
σ = |f (θ) + f (π − θ)|2 dΩ. (11.230)
2 4π

11.6.2 Elastic scattering between two spin-1/2 particles


The total spin of the system is 0 (singlet) or 1 (triplet). The potential between the
particles depends on the spin. For the spin singlet, the scattering amplitude is f s (θ) +
f s (π −θ), since the orbital part of the wave function must be symmetrical with respect to
particle exchange. For the spin triplet, the scattering amplitude is f t (θ) − f t (π − θ), since
the orbital part of the wave function must be anti-symmetrical with respect to particle
exchange. If the injected particle is not polarized, the singlet has a statistical factor of
1/4 while the triplet has 3/4. Thus,
dσ 1 3
= |f s (θ) + f s (π − θ)|2 + |f t (θ) − f t (π − θ)|2 . (11.231)
dΩ 4 4
For the total cross section we need to include a factor of 1/2:
Z  
1 1 s s 2 3 t t 2
σ= |f (θ) + f (π − θ)| + |f (θ) − f (π − θ)| dΩ (11.232)
2 4π 4 4

11.6.3 Elastic scattering between two electrons


The total spin of the system is 0 (singlet) or 1 (triplet), but in the non-relativistic limit
we can ignore the spin effect in the potential and suppose that the potential is described
by the Coulomb potential alone. For f (θ), we replace k in Eq.(11.212) with
µv me v
k= = . (11.233)
~ 2~
11.6. SCATTERING BETWEEN IDENTICAL PARTICLES 187

Then,
  
2 θ
exp −iγ ln sin
s t ~ 2 Γ(1 + iγ)
fC (θ) = f (θ) = f (θ) = (11.234)
ime v 2 θ Γ(−iγ)
sin
2
The differential cross section is
dσ 1 3
= |fC (θ) + fC (π − θ)|2 + |fC (θ) − fC (π − θ)|2 (11.235)
dΩ 4 4
= |fC (θ)|2 + |fC (π − θ)|2 − Re [fC∗ (θ)fC (π − θ)] (11.236)
 
 2  2  
α~c  1 1 1 e 2 θ
= + − cos ln tan .

4 θ θ 2 θ θ
me v 2 
~v 2
sin cos4 sin cos 2
2 2 2 2

11.6.4 Elastic scattering v.s. Inelastic scattering


Elastic scattering is a process in which the kinetic energy of the system is conserved.
From Eq.(11.54) and Eq.(11.55), the scattering amplitude for elastic scattering can be
written as

1 X
f (θ) = (2l + 1)(e2iδl − 1)Pl (cos θ). (11.237)
2ik l=0

The asymptotic form of the wave function is


eikr
 
ikz
ψ(r) = A e + f (θ) (for r → ∞) (11.238)
r
The probability current becomes
 
1 ∗ ~
j(r) = Re ψ (r) ∇ψ(r) (11.239)
m i
2
 
~k|A| 2 1
∼ ẑ + |f (θ)| 2 r̂ + . . . (11.240)
m r
The first term represents the injected wave jin , and the second term represents the
scattered wave jsc . The total cross section can be obtained as
Z Z
lim jsc (rdθ)(r sin θdφ) lim r2 jsc dΩ
r→∞ r→∞
σ = = = 4π|f (θ)|2 (11.241)
jin jin

π X
= 2 (2l + 1)|e2iδl − 1|2 . (11.242)
k l=0
188 CHAPTER 11. SCATTERING

On the other hand, if scattering causes excitation or fragmentation etc. in the target
particle, the kinetic energy is not conserved — such a process is called inelastic scat-
tering. If the scattering is inelastic, part of the injected wave is absorbed during the
process, and in general the phase shift δl becomes a complex number. Let us write the
scattering amplitude for inelastic scattering as:

1 X
f (θ) = (2l + 1)(Sl − 1)Pl (cos θ). (11.243)
2ik l=0

The total cross section is the sum of that of the elastic scattering and the inelastic
scattering (= absorption):

σtot = σel + σinel . (11.244)

To obtain σel , we just need to replace e2iδl with Sl in Eq.(11.242):



π X
σel = 2 (2l + 1)|Sl − 1|2 . (11.245)
k l=0

To obtain σinel , let us write ψ(r) as the sum of the incoming and outgoing waves (see
Eq.(11.52)):

X (2l + 1)  ikr
Sl e − (−1)l e−ikr Pl (cos θ) (for kr  1).

ψ(r) ≈ A (11.246)
l=0
2ikr

Putting Eq.(11.246) into the probability current in the radial direction:


 
1 ∗ ~ ∂
jr = Re ψ (r) ψ(r) , (11.247)
m i ∂r
we get

~k|A|2 π X
Z
2
P ≡ lim r jr dΩ = − (1 − |Sl |2 )(2l + 1). (11.248)
r→∞ m k 2 l=0

The incoming wave ∝ e−ikr makes a negative (−) contribution to P , while the outgoing
wave ∝ eikr makes a positive (+) contribution to P . In the case of elastic scattering,
|Sl | = |e2iδl | = 1, so P = 0. But if absorption occurs, then P 6= 0 and the probability of
absorption is −P . So, the cross section for absorption is
Z
− lim r2 jr dΩ ∞
r→∞ π X
σinel = = 2 (2l + 1)(1 − |Sl |2 ). (11.249)
jin k l=0
11.6. SCATTERING BETWEEN IDENTICAL PARTICLES 189

If Sl = 0, complete absorption occurs. Let us estimate σel and σinel when particles
are injected towards a complete absorber of radius R, assuming that the de Broglie
wavelength of the particles λ is much smaller than R (semi-classical regime), which means

λ=  R ⇒ kR  1. The particles with impact parameter b < R are all absorbed
k
(i.e. Sl = 0), while the particles with impact parameter b > R are not affected by the
absorber at all (i.e. Sl = 1). From Eq.(11.27), the condition R > b can be expressed as
p p
L ~ l(l + 1) l(l + 1) l
R<b= = = ≈ . (11.250)
~k ~k k k
So, from Eq.(11.245) and Eq.(11.249)
∞ kR
π X 2 π X
σel = (2l + 1)|Sl − 1| = (2l + 1) (11.251)
k 2 l=0 k 2 l=0
π π
= 2 [kR(kR + 1) + kR] ≈ 2 (k 2 R2 ) = πR2 . (11.252)
k k
∞ kR
π X 2 π X
σinel = 2 (2l + 1)(1 − |Sl | ) = 2 (2l + 1) ≈ πR2 . (11.253)
k l=0 k l=0
∴ σtot = σel + σinel = 2πR2 . (11.254)
190 CHAPTER 11. SCATTERING
Chapter 12

Additional topics

The essence of quantum mechanics lies in the indeterminacy associated with the sta-
tistical interpretation of the wave function. A quantum state Ψ does not uniquely
determine the outcome of a measurement. All Ψ provides is the statistical distribution of
possible results. The physical system does not have the attribute in question prior to the
measurement. Rather, the act of measurement itself creates the property, limited only
by the statistical constraint imposed by the wave function. An immediately repeated
measurement collapses the wave function and forces the system to take a stand, helping
to create an attribute that was not there previously.

12.1 EPR paradox


Let us consider the decay of π 0 into an electron and a positron:

π 0 → e− + e+ . (12.1)

Assuming π 0 was at rest, e− and e+ fly off in opposite directions. π 0 has spin 0, so
the conservation of angular momentum requires that e− and e+ are in the singlet spin
configuration:
1
√ (↑− ↓+ − ↓− ↑+ ). (12.2)
2
If e− has spin up, e+ has spin down, and vice versa. We cannot tell which combination
we will get in any particular π 0 decay, but we can tell that we will get each combination
half the time on average.
Suppose that e− and e+ fly way off, say 100 light years, and you measure the spin of e−
and get spin up. Then, you can say immediately that someone 100 light years away will
get spin down, if he or she measures the spin of e+ . If you think that e− really had spin
up (and so e+ spin down) from the moment they were created, then you are WRONG.

191
192 CHAPTER 12. ADDITIONAL TOPICS

Neither particle had either spin up or spin down until the act of measurement collapsed
the e− wave function to spin-up, and instantaneously produced the spin-down state of
e+ 100 light years away.
This is what takes place in reality, though in 1935, Einstein, Podolsky and Rosen claimed
that this phenomenon was contradictory to the principle of locality — no influence
can propagate faster than the speed of light.

12.2 Bell’s inequality


To solve the EPR paradox, Einstein, Podolsky and Rosen assumed that, in addition to
the wave function Ψ, some hidden variable λ is needed to characterize the state of a
system fully. A number of theories concerning λ were proposed until 1964 when Bell
proved that any hidden variable theory is incompatible with quantum mechanics. Bell

Figure 12.1: Bell’s version of the EPR experiment: detectors independently oriented in
directions a and b.
suggested a generalized EPR experiment — instead of orienting the e− and e+ detectors
along the same direction, he allowed them to be rotated independently. The first detector
measures the component of the e− spin in the direction of a unit vector a, and the second
detector measures the component of the e+ spin in the direction of a unit vector b. Let
us record the spins in units of ~/2, then each detector registers +1 for spin up or −1 for
spin down along the direction in question. Suppose now that we calculate the average
value of the product of the spins, for a given set of detector orientations. We call this
average P (a, b). If b = a (i.e. the detectors are parallel), then the detector configuration
is the same as in the EPR paradox. In that case, one particle is spin up while the other
is spin down; the product is always −1, so
P (a, a) = −1. (12.3)
If b = −a (i.e. the detectors are anti-parallel), then
P (a, −a) = +1. (12.4)
12.2. BELL’S INEQUALITY 193

For arbitrary detector orientations, we get1

P (a, b) = −a · b. (12.5)

However, Bell proved that this result is incompatible with any local hidden variable
theory, based on the following argument.
Suppose that the state of the e+ e− system is characterized by the hidden variable(s) λ.
Then there exists some function A(a, λ) which gives the result of the e− measurement,
and B(b, λ) for the e+ measurement. These functions can only take on the values ±1:

A(a, λ) = ±1, B(b, λ) = ±1. (12.6)

If the two detectors are aligned, the results are perfectly anti-correlated:

A(a, λ) = −B(a, λ), for all λ. (12.7)

The average of the product of the spin measurements is


Z
P (a, b) = ρ(λ)A(a, λ)B(b, λ)dλ, (12.8)
Z
where ρ(λ) is the probability density for λ, and satisfies ρ(λ)dλ = 1. Using Eq.(12.7),
Eq.(12.8) can be written as
Z
P (a, b) = − ρ(λ)A(a, λ)A(b, λ)dλ. (12.9)

Letting c be any other unit vector,


Z
P (a, b) − P (a, c) = − ρ(λ)[A(a, λ)A(b, λ) − A(a, λ)A(c, λ)]dλ. (12.10)

Since [A(b, λ)]2 = 1,


Z
P (a, b) − P (a, c) = − ρ(λ)[1 − A(b, λ)A(c, λ)]A(a, λ)A(b, λ)dλ. (12.11)

Noting that

−1 ≤ A(a, λ)A(b, λ) ≤ +1, ρ(λ)[1 − A(b, λ)A(c, λ)] ≥ 0, (12.12)


1

When two spin-1/2 particles are in the singlet configuration |s ms i = |0 0i = (1/ 2)(↑↓ − ↓↑), then
(2) ~2 (1)
hSa(1) Sb i = − cos θ, where Sa is the component of the spin angular momentum of particle number 1
4
(2)
along the unit vector â, Sb be the component of the spin angular momentum of particle number 2
along the unit vector b̂, and θ is the angle between â and b̂.
194 CHAPTER 12. ADDITIONAL TOPICS

we can derive from Eq.(12.11) that


Z
|P (a, b) − P (a, c)| ≤ ρ(λ)[1 − A(b, λ)A(c, λ)]dλ. (12.13)

Using A(c, λ) = −B(c, λ) and Eq.(12.8), we get


|P (a, b) − P (a, c)| ≤ 1 + P (b, c). (12.14)
This equation is the Bell inequality. Now suppose that a, b and c lie in a plane and c
makes a 45◦ angle with a and b, quantum mechanics says1
1
P (a, b) = 0, P (a, c) = P (b, c) = − √ = −0.707. (12.15)
2
These values are incompatible with Eq.(12.14):
0.707 6≤ 1 − 0.707 = 0.293. (12.16)
This result means that there exists no hidden variable that accommodates quantum me-
chanics with the principle of locality.
Actually there are many things that travel faster than light — if a bug flies across the
beam of a movie projector, the speed of its shadow is proportional to the distance to the
screen. The distance can be as large as you like, so the shadow can move at arbitrar-
ily high velocity. However, the shadow does not carry any energy, nor can it transmit
a message from one point to another. We call this type of influence ethereal. On the
other hand, a causal influence cannot propagate faster than light. Special Relativity says
that there exist inertial frames in which a signal traveling faster than light propagates
backward in time. In that case the effect precedes the cause, which leads to inescapable
logical anomalies.
In the EPR paradox, the measurement of e− does influence the outcome of the e+ mea-
surement (i.e. anti-correlation between the two spins), but the measurement of e− does
not cause a particular outcome for e+ . The person at the e− detector cannot control the
outcome of his/her own measurement, that is, the person cannot make a given e− come
out spin-up. All he/she can do is to decide whether to make a measurement at all, and
the person at the e+ detector cannot tell whether e− was measured or not, because the
lists of data at the two detectors, considered separately, are completely random. Only
when we compare the two lists of data later do we discover the anti-correlations. In
another reference frame the e+ measurements take place before the e− measurements,
but this does not result in any logical anomaly. Therefore, the anti-correlation of the two
spins is a delicate influence of ethereal type.
Thus, we distinguish two types of influence:
ˆ ”Causal” kind: an influence that produces actual changes in some physical property
of the receiver, detectable by measurements on that subsystem alone.
12.3. NO-CLONE THEOREM 195

ˆ ”Ethereal” kind: an influence that does not transmit energy or information, and
for which the only evidence is a correlation in the data taken on the two separate
subsystems — a correlation which cannot be detected by examining either list alone.

Causal influences cannot propagate faster than light, but ethereal ones can. The influ-
ences associated with the collapse of the wave function is an ethereal kind.

12.3 No-clone Theorem


Measurements are destructive in quantum mechanics — they collapse wave functions and
alter the state of the system. Why don’t we just create clones (i.e. identical copies) of
the original state and measure those clones, leaving the original system itself unaffected
by the measurements? Suppose we were able to create such a ”quantum Xerox machine”.
Schematically, the machine takes as input a particle in state |ψi (the original one to be
copied) plus a second particle in state |Xi (a ”blank sheet of paper”), and spit out two
particles in the state |ψi:

|ψi|Xi → |ψi|ψi. (12.17)

This machine could clone |ψ1 i and |ψ2 i separately2 :

|ψ1 i|Xi → |ψ1 i|ψ1 i, |ψ2 i|Xi → |ψ2 i|ψ2 i. (12.18)

But this machine cannot clone a linear combination of |ψ1 i and |ψ2 i, because

|ψi(α|ψ1 i + β|ψ2 i) → α|ψ1 i|ψ1 i + β|ψ2 i|ψ2 i (12.19)


6= (α|ψ1 i + β|ψ2 i)(α|ψ1 i + β|ψ2 i). (12.20)

Unless the machine can clone α|ψ1 i + β|ψ2 i, it cannot clone any state, because any state
can be expressed as a linear combination of states by changing the bases3 .

12.4 Schrödinger’s cat


What exactly is a measurement? How can we tell when a measurement has occurred?
To pose this essential question, Schrödinger posed a cat paradox.
A cat is placed in a chamber, together with a Geiger counter that contains a tiny amount
of radioactive substance. If one of the atoms decays, the counter triggers and activates a
hammer which breaks a container of cyanide and kills the cat. Suppose we left the entire
2
|ψ1 i and |ψ2 i could be spin up and spin down, if the particle is an electron.
3 (z) 1 (x) (x)
For instance, recall χ+ = √ (χ+ + χ− ).
2
196 CHAPTER 12. ADDITIONAL TOPICS

system, say, for an hour. If no atom has decayed, the cat is living. If the first decay has
occurred, the cat is dead. We do not know whether the cat is alive or dead. So, at the
end of the hour, the wave function of the cat can be written as
1
ψ = √ (ψalive + ψdead ), (12.21)
2
which means that the cat is neither alive nor dead until you open the door of the chamber
and observe the cat inside. Your observation forces the cat to ”take a stand”: dead or
alive, and if you find the cat to be dead, then it’s really you who killed him by looking
in the window.
Obviously this is nonsense. There is something absurd about the idea of a macroscopic
object being in a linear combination of two different states. The answer is that the
triggering of the Geiger counter constitutes the ”measurement”, not the intervention of a
human observer. The measurement occurs at the moment when the microscopic system,
described by quantum mechanics, interacts with the macroscopic system, described by
classical mechanics, in such a way as to leave a permanent record. The macroscopic
system itself does not occupy a linear combination of distinct states. Keep in mind that
the measurement does not necessarily entail human participation.

12.5 Quantum Zeno effect


A measurement collapses the wave function and makes it ”take a stand”. An immediately
repeated measurement reproduces the same value, and this fact carries directly observable
consequences. Suppose we take an unstable system (e.g. an atom in an excited state),
and subject it to repeated measurements. Each observation collapses the wave function,
and it is possible to delay the transition to the lower state. This phenomenon is called
the quantum Zeno effect.
Let us imagine a system starts out in the excited state ψ2 , which has a natural lifetime τ
for transition to the ground state ψ1 . For times substantially less than τ , the probability
of transition is proportional to t. Since the transition rate is 1/τ ,
t
P2→1 = . (12.22)
τ
If we make a measurement after a time t, the probability that the system is still in the
upper (ψ2 ) state is
t
P2 (t) = 1 − . (12.23)
τ
Suppose we do find the atom to be in the upper state. In that case, the wave func-
tion collapses back to ψ2 , and the process starts all over again. If we make a second
12.6. NEUTRINO OSCILLATION 197

measurement at time 2t,


 2
t 2t
P2 (2t) = 1− ≈1− , (12.24)
τ τ

which is the same as it would have been if we had never made the first measurement at
time t.
However, for extremely short times, the probability of a transition is not proportional to
t. If t is extremely small, Eq.(10.58) becomes
Z ∞  2 
2 2 sin [(ω0 − ω)t/2]
Pb→a (t) = |P| ρ(ω) dω (12.25)
0 ~2 0 (ω0 − ω)2
Z ∞
[(ω0 − ω)t/2]2
 
2 2
≈ |P| ρ(ω) dω (12.26)
0 ~2 0 (ω0 − ω)2
2 Z ∞
2 2t
= 2
|P| ρ(ω)dω ∝ t2 . (12.27)
0 ~ 4 0

So, the probability of a transition can be written as P2→1 = αt2 , and the probability that
the system is still in the upper state after the two successive measurements is

P2 (2t) = (1 − αt2 )2 ≈ 1 − 2αt2 , (12.28)

whereas it would have been 1 − α(2t)2 ≈ 1 − 4αt2 if we had never made the first mea-
surement. If we repeatedly examine the system at time T /n, 2T /n, 3T /n, . . ., T , then
(  2 )n
T α
P2 (T ) = 1 − α ≈ 1 − T 2, (12.29)
n n

which goes to 1 in the limit n → ∞ — a continuously observed unstable system never


decays at all. In fact, the experiment is impractical for spontaneous transitions, but in
can be done using induced transitions, and the quantum Zeno effect has been confirmed
by actual experiments.

12.6 Neutrino oscillation


There exit three kinds of neutrinos: electron neutrino (νe ), muon neutrino (νµ ) and tau
neutrino (ντ ). These are the states of neutrinos produced by particle interactions, and
called the flavor eigenstates. On the other hand, the eigenstates of Hamiltonian are
called the mass eigenstates:

H|ν1 i = E1 |ν1 i, H|ν2 i = E2 |ν2 i, H|ν3 i = E3 |ν3 i. (12.30)


198 CHAPTER 12. ADDITIONAL TOPICS

The point is that the flavor eigenstates are not the mass eigenstates, but a mixture of
them. For simplicity, let us consider νe and νµ alone, and express them as:
    
|νe i cos θ sin θ |ν1 i
= . (12.31)
|νµ i − sin θ cos θ |ν2 i

Imagine a neutrino is produced at t = 0 as a νµ :

|ν(0)i = |νµ i = − sin θ|ν1 i + cos θ|ν2 i. (12.32)

At time t, the mass eigenstates pick up phase factors:

|ν(t)i = − sin θ|ν1 ie−iE1 t/~ + cos θ|ν2 ie−iE2 t/~ . (12.33)

The probability that the neutrino is observed as a νe at time t is:

Pνe →νµ = |hνe |ν(t)i|2 (12.34)


−iE1 t/~ −iE2 t/~ 2
= |(cos θhν1 | + sin θhν2 |)(− sin θ|ν1 ie + cos θ|ν2 ie )| . (12.35)

Using hν1 |ν1 i = hν2 |ν2 i = 1 and hν1 |ν2 i = hν2 |ν1 i = 0,

Pνe →νµ = | − sin θ cos θe−iE1 t/~ + sin θ cos θe−iE2 t/~ |2 (12.36)
2 2 −i(E1 −E2 )t/~ i(E1 −E2 )t/~
= sin θ cos θ(1 − e −e + 1) (12.37)
  
2 2 (E1 − E2 )t
= sin θ cos θ 2 − 2 cos (12.38)
~
 
(E1 − E2 )t
= sin2 (2θ) sin2 (12.39)
2~

Meanwhile, special relativity says


s
p m 2 c2
E = p2 c2 + m2 c4 = pc 1+ . (12.40)
p2

Neutrino mass is very small (m2 c4  p2 c2 ), so

1 m 2 c2 m2 c3
 
E ≈ pc 1 + = pc + . (12.41)
2 p2 2p

Also, E ≈ pc holds although the neutrino may change its flavors, so

∆m2
E1 − E2 = , where ∆m2 ≡ (m1 c2 )2 − (m2 c2 )2 . (12.42)
2E
12.7. BLACKBODY RADIATION 199

Moreover, v ≈ c, so t ≈ L/c, where L is the flight distance of the neutrino. Eq.(12.39)


becomes
∆m2 L
 
2 2
Pνe →νµ = sin (2θ) sin . (12.43)
4E~c
∆m2 [eV2 ]
 
2 2
= sin (2θ) sin 1.27 L[m] (12.44)
E[MeV]

The probability that the flavor of the neutrino changes from νµ to νe oscillates according to
the flight distance of the neutrino. This phenomenon is called the neutrino oscillation.
The discovery of neutrino oscillation is the evidence for the non-zero mass of the neutrino.
When all the three kinds of neutrinos are involved, the mixing matrix is called the MNS
matrix4 and written as

0 s13 e−iδ13
      
|νe i 1 0 0 c13 c12 s12 0 |ν1 i
|νµ i = 0 c23 s23   0 1 0  −s12 c12 0 |ν2 i(12.45)
iδ13
|ντ i 0 −s23 c23 −s13 e 0 c13 0 0 1 |ν3 i

where sij and cij represent sin θij and cos θij , respectively, and δ13 is the CP (Charge-
Parity) violating phase.
Also in the quark sector there is a similar mixing of the d (down), s (strange), and b
(bottom) quarks, and the mixing matrix is named the CKM5 matrix.

12.7 Blackbody radiation


§
In 5.4.4 we learned that the energy density of the black body is described by Planck
distribution:
8πhν 3 1
I(ν) dν [J m−3 ] = 3 hν/k T −1
dν. (12.46)
c e B

Let us derive this equation in two ways — One is Planck’s original approach, and the
other is a modern approach using statistical mechanics as well as quantum mechanics.

12.7.1 Density of states


First, let us derive the density of states6 for electromagnetic radiation (i.e. light waves)
confined in a cube of side length L (cavity radiation). The derivation is similar to the
4
Maki-Nakagawa-Sakata matrix
5
Cabibbo-Kobayashi-Maskawa matrix
6
g(ν)dν is the number of states whose frequencies are within ν and ν + dν.
200 CHAPTER 12. ADDITIONAL TOPICS

§
free electron gas model that we learned in 5.3.1. Due to the boundary condition at the
walls of the cube, the light waves have the amplitudes proportional to:
n π  n π  n π 
x y z
sin x sin y sin z , (12.47)
L L L
where nx = 1, 2, 3, . . . , ny = 1, 2, 3, . . . , nz = 1, 2, 3, . . . . (12.48)

The wave vector is


 
nx π ny π nx π
k ≡ (kx , ky , kz ) = , , . (12.49)
lx ly lx

In the 3D k space, each block of volume π 3 /L3 contains two states, since light waves are
transverse and have two degrees of freedom with respect to their polarization. Noting
that the states in the space with nx > 0, ny > 0, nz > 0 are occupied, the number of
states contained in one octant of a shell of thickness dk is
1
4πk 2 dk × 2 k 2 L3
8 = dk. (12.50)
π3 π2
L3
Using k = 2πν/c, Eq.(12.50) becomes

4π 2 ν 2 2π L3 8πν 2 3
π dν = L dν. (12.51)
c2 c π3 c3
Thus, the density of states per unit volume is

8πν 2
g(ν)dν = dν. (12.52)
c3

12.7.2 Planck’s quantum hypothesis


In 1900, Planck postulated that light energy can only be emitted and absorbed in discrete
bundles called quanta7 — the energy of light of frequency ν is determined by8

E = n hν (n = 0, 1, 2, 3, . . .). (12.53)

According to Maxwell-Boltzmann distribution law, the probability that a particle in


a thermal equilibrium at temperature T has energy  is proportional to exp(−/(kB T )).
7
Today we call them photons.
8
h is called Planck constant today and known to have the value of 6.626 × 10−34 [m2 kg s−1 ].
12.7. BLACKBODY RADIATION 201

So, the average energy of light of frequency ν is


∞  
X nhν
nhν exp −
n=0
kB T
Ē = ∞   . (12.54)
X nhν
exp −
n=0
kB T

The denominator becomes


∞  
X nhν 1
exp − = −hν/(k
(12.55)
kB T 1−e BT )
n=0

Letting ξ = 1/(kB T ), the numerator becomes


∞ ∞
X
−nhνξ d X −nhνξ d −1 hνe−hνξ
nhνe =− e =− 1 − e−hνξ = (12.56)
n=0
dξ n=0 dξ (1 − e−hνξ )2

Putting Eq.(12.55) and Eq.(12.56) into Eq.(12.54),

hνe−hν/(kB T ) hν
Ē = −hν/(k T )
= hν/(k T ) . (12.57)
1−e B e B −1
Multiplying Eq.(12.57) by Eq.(12.52), we get Planck distribution Eq.(12.46).

12.7.3 Modern approach


The energy levels of a 1D harmonic oscillator is given by
   
1 1
n = n + ~ω = n + hν (n = 0, 1, 2, . . .). (12.58)
2 2

In statistical mechanics, when a closed system is in thermal contact with a heat bath and
only the thermal energy exchange occurs between them, the probability distribution of
each constituent partcle being in a state with energy n is
  ∞  
1 n X n
Pn = exp − , where Z ≡ exp − . (12.59)
Z kB T n=0
kB T

Z is called the canonical partition function. For Eq.(12.58), Z becomes


∞   
X hν 1 exp(−hν/(2kB T ))
Z= exp − n+ = . (12.60)
n=0
kB T 2 1 − exp(−hν/(kB T ))
202 CHAPTER 12. ADDITIONAL TOPICS

The Helmholtz free energy is


  
1 hν
φ = −kB T ln Z = hν + kB T ln 1 − exp − . (12.61)
2 kB T

The total free energy of the whole system is obtained by the superposition of individual
particles represented by harmonic oscillators:
X 1  
hνi

F = hνi + kB T ln 1 − exp − (12.62)
i
2 kB T

The internal energy of the system is


  X 
2 d F 1 hνi
E = −T = hνi + (12.63)
dT T i
2 exp(hνi /(kB T )) − 1)

The first term in Eq.(12.63) represents the zero-point energy that exists even at ab-
solute zero temperature. Using Eq.(12.52), the energy of the system relative to the
zero-point energy can be written as
Z ∞

E = g(ν)dν (12.64)
0 exp(hν/(kB T )) − 1
Z ∞
8πh ν3
= dν (12.65)
0 c3 exp(hν/(kB T )) − 1

The integrand is exactly Planck distribution Eq.(12.46).

You might also like