Quantum Mechanics II

George Siopsis

Department of Physics and Astronomy

The University of Tennessee
Knoxville, TN 37996-1200
U.S.A.
e-mail: [email protected]

Lecture notes by students

Spring 2011
ii
Contents

1 Scattering 3

2 Angular momentum 33

3 Stationary perturbation theory 61

4 Fine and hyperfine structure of the hydrogen atom 85

5 Time-dependent perturbation theory 99

6 Identical particles 125

2 CONTENTS
UNIT 1

Scattering

Notes by C. Hartnett and K. Yeter

Scattering is such an important phenomenon that is used in high energy physics
to understand the forces of the nature and the properties of the materials. In
general, there are two types of scattering used in the scientific research.One
type uses colliding beams of high energy particles to produce new particles.
Looking at the by products of such collisions gives scientists insgiht into the
fundamental forces of nature. The LHC (Large Hadron Collider) is the biggest
examples for this kind of scattering, and it’s very important. The other type is
to scatter beams of particles, like neutrons, off a stationary material target.This
type of experiment is used to understand the property of the target material. A
prime example of this type is the SNS (Spallation Neutron Source) at Oak Ridge
National Laboratory. It is one-of-a-kind facility provides the most intense pulsed
neutron beams in the world for scientific research and industrial development.
Since we can consider the two colliding particle beams in the rest frame of one
of the beams, studying these kinds of scattering phenomena is not as hard as it
seems to be.
First of all, let me ask the question of whether two electrons will meet once
they are directed to each other for collision. Since electrons are fundamental
particles they are exact points and laws of classical mechanics cannot be used to
answer this question. As an answer for this question we can say that there is a
probability for these electrons to meet.Another question that begs an answer is:
if electrons do not meet how do they scatter then? The Coulomb force causes
the electrons to be deflected. On the other hand, for two particles attracting
each other we can say that there is a certain
possibility of these particles’ positions therefore there is a finite possibility for
these particles to find each other.
We are going to assume that what governs the force between each particle is
the potential which only depends on the relative position r = r1 − r2 of the
4 UNIT 1: Scattering

particles.
r /•
•O
}}}>
r1 }
}r2
}}
O·
where,
r: distance between particle m1 and m2
r1 : distance of particle m1 from the origin
r2 : distance of particle m2 from the origin

In the center-of-mass reference frame of the two particles the problem will be
reduced to the study of the scattering of a single particle by the potential V(r).
Therefore, the mass of the particles also will be reduced to a value of µ = mm11+m
m2
2
.
This µ value is called the effective or reduced mass and it is the mass of the
”relative particle”.
In the rest frame of a scattering, in which there is a particle beam directed
towards a target, we see the distribution of the particles on the detector.

1-Dimensional Case

Figure 1: 1-d potential

Figure 1 shows the incident beam interacting with an arbitrary one dimensional
potential. In case of interacting with a potential there will be transmitted and
reflected beam of particles. As you may remember from the previous semester
a square potential was given as an example for this type of scattering. For
particles whose energy greater than zero there will be a continuous spectrum
and the Schrodinger equation can be written as,

~2 00
− ϕ + V ϕ = Eϕ, E>0 (1)
2m
Since the spectrum is continuous solution of the Schrodinger equation for scat-
tering potential gives a plane and non-normalizable wavefunction. It is,

ϕinc = Aeikx , x → −∞ (2)

Scattering 5

√
where k = 2mE ~ . The solutions for reflected and transmitted beam of particles
are respectively,
ϕr = Be−ikx (3)
ϕt = Ceikx (4)
The wave function incident in region I would be the sum of the wavefunctions
which are incident and reflected.

ϕI = ϕinc + ϕr (5)

The wavefunction transmitted in region III, where the potential ends, is,

ϕIII = Ceikx , x → +∞ (6)

The wavefunction is not a physical quantity therefore we cannot measure it

in the experiments. What we measure in the experiments is the probability
i~
current. Probability current is defined as J = 2m (ψ∇ψ ∗ − ψ ∗ ∇ψ). There is
a continuity equation for probability current and current density which gives
the conservation law of probability current, J. For this scattering case the
probability current density,ρ, does not depend explicitly on time.
∂ρ ∂J
+ =0 (7)
∂t ∂x
Once we do the calculation,
i~ ∗ 00
J =− ϕ ϕ + C.C. (8)
2m
i~ ∗ 0 i~ ∗ 00
J 0 = −i ϕ ϕ − ϕ ϕ + C.C. (9)
2m 2m
As you measure the physical quantities in the scattering process you stay away
from the scattering potential therefore from the Schrodinger equation
2mE
ϕ00 = − ϕ (10)
~2

i~
J0 = − | ϕ0 |2 + C.C. = 0 (11)
2m
From equation (1.11) we can easily see that probability current, J, is conserved
- i.e. it does not depend explicitly on x and it is just a number.
As x → −∞ there will be two contributions to the wavefunction, ϕ, i.e. one from
incident beam of particles and one from reflected beam of particles. Therefore
J becomes,
i~ ~k
J =−−i (A∗ e−ikx + B ∗ eikx )ik(Aeikx − Be−ikx ) + C.C. = (|A|2 − |B|2 )
2m m
(12)
6 UNIT 1: Scattering

~k ~k
Jinc = | A |2 , Jr = | B |2 (13)
m m
As x → +∞ the only contribution comes from the transmitted probability
current density.
~k
Jt = | C |2 (14)
m
Since J is a conserved quantity no matter how complicated the Schrodinger
equation is the incident probability current is equal to the sum of the
reflected probability current and the transmitted probability current.
Jinc = Jr + Jt . (15)
If equation (1.15) is divided by Jinc then we obtain that
Jr Jt
1= + . (16)
Jinc Jinc
In the equation above the JJinc
r
value is called the reflection coefficient and de-
Jt
noted as R. The Jinc value is called the transmission coefficient and denoted as
T. Therefore,
R+T =1 (17)
The most interesting point of the scattering phenomenon is that the reflected
particles are the scattered particles from the scattering potential. The number
of reflected particles is about 1020 . Say we have a beam of particles which
consists of N number of particles. The probability density, ρ, for such a beam is
Z
2
ρ=|ϕ| and ρdx = 1 (18)

If we measure a number of particles given a certain distance then,

# of particles
= N ρ = #density (19)
length
If the N number of particles are electrons with charge e then what you measure
in the experiment is the charge density.
charge density = eN ρ (20)
The electrical current is defined as the number of particles passing through a
certain point in unit time. Therefore, the electrical current for this beam of
particles is
charge
electricalcurrent = eN J = (21)
time
Since the electrical current is directly proportional to the probability current
density, J, reflection and transmission coefficients can be written as the ratio of
the reflected and transmitted currents to the incident current.
ref lected current transmitted current
R= and T = . (22)
incident current incident current
Scattering 7

Figure 2: 3-d potential

3-Dimensional Case
In 3-d case we are going to introduce current density. To use the current density
we need to introduce the concept of flux- the number of particles per unit time
which traverse a unit surface perpendicular to direction of the beam. The flux
of the incident particles beam is Finc .
# of particles
Finc = (23)
(area)(time)
charge I
current density = J = = (24)
(area)(time) area
In Figure 2 a detector is placed far from the region under the influence of the
potential. The number of particles, dn, scattered per unit time into the solid
angle dΩ about the direction (θ, ϕ) is

dn = Finc σ(θ, ϕ)dΩ (25)

where σ(θ, ϕ) is the differential scattering cross section in the direction (θ, ϕ).
Scattering cross section is the coefficient of proportionality between dn and
Finc dΩ. In other words, the scattering cross section is
# of def lected particles/time
(26)
incident f lux
The scattering cross section has the dimension of area which is on the order of
≈ 10−24 cm2 . Since 10−24 cm2 is small the unit of barn is used.

1barn = 10−24 cm2 (27)

8 UNIT 1: Scattering

The total cross section is the integral over the solid angle, that is,
Z
dσ
σ = dΩ (28)
dΩ
The second term on the right-hand-side of the equation above is called the
differential cross section and it is rather useful.
Now, let us find a simple expression for scattering cross section, σ. As far as
the incident
√
beam is concerned the wavefunction is ϕ = Aeikz , where
2mE
k = ~ and m is the reduced mass. For simplicity we set A=1, therefore the
wavefunction for the incident beam is ϕ = eikz . On the other hand, reflected
and transmitted wavefunctions are more complicated.
To find the scattered wavefunction we shall solve the Schrodinger equation.
Since the detector is far from the scattering potential we can neglect the poten-
tial in the Schrodinger equation. Therefore,

~2 2
− ∇ ϕ = Eϕ (29)
2m
∇2 ϕ + k 2 ϕ = 0 (30)
We assume that there is a spherical symmetry and we shall use the spherical
coordinates to solve the equation above.
1
(rϕ)00 + k 2 ϕ = 0 (31)
r
Let’s denote (rϕ) as u, that is,

u00 + k 2 ϕ = 0 ⇒ u = e±ikr (32)

Since the particles are scattering we choose eikr as a solution for u. Therefore
the generalized solution for the scattered wavefunction is,
1 ikr
ϕ= e . (33)
r
Figure 2 does not have spherical symmetry therefore we hope to manipulate ϕ
in such a way as to arrive at an appropriate solution for ϕ. We therefore need
to solve an equation that is a little more general,

∇2 ϕ + k 2 ϕ = ρ (34)
Where ρ will be something that I have to figure out if it means anything at all.
If we have to solve Eq. (3) the solution is known because all one has to do is
look to electrostatics for the necessary steps. And that will give me ϕ at r, that
is,
Z ~0
eik|~r−r |
ϕ(~r) = − d3 r0 ρ(r~0 )G+ (~r, r~0 ) G± (~r, r~0 ) = − (35)
4π|~r − r~0 |
Scattering 9

G(~r, r~0 ) is a Green’s function, and ϕ(~r) is now a general solution to Eq. (3). So
this Green function is the same as ϕ, the spherically symmetric case, so it must
obey the Schrodinger equation.

∇2 G± + k 2 G± = 0 (36)
Have to be careful here because as ~r → 0 you get a singularity, so it’s zero
everywhere except exactly at the origin, therefore,

∇2 G± + k 2 G± = δ 3 (~r) (37)
Integrating Eq. (6),
Z Z
d3 r0 (∇2 G± + k 2 G± ) = d3 r0 δ 3 (~r) (38)
Z Z
~ ± · dS
∇G ~ + k2 d3 rG± = 1 (39)

And as the spherical volume element within the integral shrinks to zero Eq. (8)
becomes one as it should.

1+0=1 (40)
And now we understand how to solve Eq. (3), in general. But no matter what,
we want to go far away from the scattering potential and we want to put the
detector way out so it’s not effected by what’s happening. So we look at the
behavior as ~r → ∞.
p
|~r − r~0 | = r2 + r02 − 2rr0 cos β ≈ r − r0 cos β (41)
Where β is the angle between ~r and r~0 . If we go back to the expression for ϕ(~r)
and use the approximation for |~r − r~0 |,
Z 0 Z
eikr−ikr cos β eikr 0
ϕ(~r) = − d3 r0 ρ(r~0 ) =− d3 r0 ρ(r~0 )e−ikr cos β (42)
4πr 4πr
Notice the factor outside the integral it’s a nice spherical wave traveling out-
ward, and then there’s the integral. What is this integral a function of? After
integrating over r0 this integral is going to be a function of ~r, which has three
component in spherical coordinates f (r, θ, ϕ). But because there is no r (the
distance r not the vector ~r) f will be a function of the angles only. So we have
a nice spherical wave as it goes out but it’s modulated as you go around the
angles. We don’t quite understand ρ(~r) yet but at least we can write ϕ(~r) as,

eikr
ϕ(~r) = − f (θ, ϕ) (43)
4πr
So this is our equation for the out going scattered wave. The entire wave function
will be the combination of the incoming wave, eikz , and the outgoing wave.
10 UNIT 1: Scattering

eikr
Φ = eikz + f (θ, φ) (44)
r
The minus and factor of 4π were absorbed in the arbitrary function f (θ, φ). This
generalizes what we found in one dimension. Next, we’re going to take a look
at the probability current and try to generalize from one to three dimensions.
Recall that the probability current and the gradient in spherical coordinates are
given by,

i~ ∗ ~
J~ = − Φ ∇Φ + C.C. (45)
2m

~ = ∂ r̂ + 1 ∂ θ̂ + 1
∇
∂
φ̂ (46)
∂r r ∂θ r sin θ ∂φ
The incident current J~i = ~k
m ẑ. Now the scattered wave, or the out going wave,
it’s going to have three components.

i~ e−ikr ∗ ikeikr eikr ~k |f |2

(Jout )r = − f f[ − 2 ] + C.C. = (47)
2m r r r m r2

i~ 1 ∗ ∂f
(Jout )θ = − f + C.C. (48)
2m r3 ∂θ

i~ 1 ∂f
(Jout )φ = − f∗ + C.C. (49)
2m r3 sin θ ∂φ
The currents are very complicated. So what can we do about it? Now remember
we want to know what happens very far away from the scattering potential, as
~r → ∞. That’s where the detector is going to be and that’s where I’m going to
observe the currents. The angular components of the probability currents fall
off like r−3 and we can forget about them. The current represents particles that
come out so it has to be conserved. It’s all coming out in the radial direction.
The total out going current is given by integrating over a sphere.
Z Z
~k
Iout = Jr dS = dΩ|f |2 (50)
m
It turns out that Iout is a constant, no r dependence and only the solid angle.
Which it needs to be because probability currents need to be conserved. Now
we can build an expression for the scattering cross section.
Z
Iout
σ= = dΩ|f |2 (51)
Jin
If we care about the distribution of particles then we need the differential cross
section. You can see that the differential cross section is given precisely by the
function f . This is a very important function because this is what we’re going
to be able to measure.
Scattering 11

Z Z
2 dσ dσ
σ= dΩ|f | = dΩ → = |f |2 (52)
dΩ dΩ
Where do we even find this f (Ω)? We turn back to our old friend the Schrodinger
equation. And when it is written in this way we can see that it’s form is similar
to that of Eq. (3). Except now we can see that ρ comes from ϕ itself.
√
2m 2mE
∇2 ϕ + k 2 ϕ = 2 V ϕ k= (53)
~ ~
If we can say ρ = 2m~2 V ϕ then we already know of the solution to Eq. (22). It
is an integral equation and nobody can solve it.
Z
2m
ϕ(~r) = eikz + 2 d3 r0 G+ (~r, r~0 )V (r~0 )ϕ(r~0 ) (54)
~
But, if you can solve it you’re going to get the function f . Well, like before
we only care about what happens really far away at the detector. We have the
right ρ now so f is therefore given by,
Z
1 2m
f (θ, φ) = − d3 r0 eik~r·r̂ V (r~0 )ϕ(r~0 ) (55)
4π ~2
Now let’s calculate this. To do so we need to use the Born approximation. Let
ϕ ≈ eikz . Then for Eq. (24),
Z
1 2m 0
f (θ, φ) ≈ − d3 r0 eik~r·r̂ V (r~0 )eikz (56)
4π ~2
And if we look in the direction of r we can write kz 0 = k~i · r~0 . Which is the
definition of the Fourier transform of V.
Z
1 2m ~ ~ ~0 1 2m ~
fB = − d3 r0 e−i(ks −ki )·r V (r~0 ) = − Ṽ (ks − k~i ) (57)
4π ~2 4π ~2
We can improve our Born approximation through iterations.

ϕ0 = eikz (58)
Z
2m 0
ikz
ϕ1 = e + d3 r0 G+ (~r, r~0 )V (r~0 )eikz (59)
~
Z
2m
ϕ2 = eikz + d3 r0 G+ (~r, r~0 )V (r~0 )ϕ(r~0 ) (60)
~
Instead of having a plane wave let’s try spherical wave. How do we find all
solutions? Need to solve the equation ∇2 ϕ + k 2 ϕ = 0, and find the eigenvalues,
of which will commute with a set of operators {H, L2 , Lz }.

ϕklm = Rkl Ylm (θ, φ) (61)

12 UNIT 1: Scattering

~2 1 l(l + 1)~2
− (rR)00 + R = ER let u = rR (62)
2m r 2mr2
~2 00 l(l + 1)~2
−u + u = Eu (63)
2m 2mr2
Now as r → 0 there is asymptotic behavior.

Figure 3: asymptotic behavior

Therefore,

l(l + 1)
− u00 + u = Eu ≈ 0 (64)
r2
Solution of the differantial equation above is done by writing u = rλ instead of
u and then,
λ(λ + 1) = l(l + 1) =⇒ λ = −l or l + 1 (65)
We choose the solution of λ = l + 1 therefore the solution of this differential
equation is rl+1 for r → 0.
On the other hand, as r → ∞, u becomes that of a free particle.

~2 00
− u = Eu =⇒ u = e±ikr (66)
2m
We have found the general solution valid for a far away detector and the bound-
ary conditions examined. For instance as r→0 we have found that u also goes
to zero therefore u≈ rl+1 and as r→ ∞ u≈ e±ikr . Let us do an example about
this kind of situation. We are going to examine them for different l and now we
are going to start with the simplest case which is l=0.
• l=0
When we are far away from the potential there are two solutions and the
most general solution is
u = A0 eikr + B 0 e−ikr (67)
Scattering 13

If we Taylor expand the function u it becomes

u = A0 + B 0 + ik(A0 − B 0 )r + . . . . (68)

In the limit when r goes to zero we find that

A0 = −B 0 (69)

Using this relation between A’ and B’ we can write u as

u = A0 (eikr − e−ikr ) = A sin(kr) (70)

then the radial part of the wavefunction which is R will be

sin(kr)
R=A (71)
r
In this case it is very simple because l is zero and then to find the wave-
function we need to multiply this by spherical harmonics which is constant
for l = 0. Now it can be seen that this radial wavefunction, R, is a mixture
of incoming and outgoing waves which has to be in this way because we
want a valid solution for everywhere so as r→0 this wavefunction needs
to satisfy this condition as well. In case of the Hydrogen atom the ra-
dial part of the wavefunction needs to be normalized. But here we don’t
have normalization because the energy of the particles are greater than
zero and we have a continuous spectrum therefore to normalize this radial
wavefunction we are going to use the delta function. Let us write the inner
product for the wavefunction.
Z
I = d3 rϕ∗k0 0 (~r)ϕk0 (~r) (72)

If we separate this into radial and angular parts we find

Z ∞ Z
I= druk (r)uk (r) dΩ|Y00 |2
∗
0 (73)
0

Since spherical harmonics are normalized the second integral term will
be equal to 1. Now to calculate the integral for the radial part let us
substitute the value of u = A sin(kr) in the integral.
Z ∞
I = A∗k Ak0 dr sin(kr) sin(k 0 r) (74)
0

To solve this integral we are going to write the sine functions in terms of
exponentials.
Z ∞ 0 0
(eikr − e−ikr ) (eik r − e−ik r )
I = A∗k Ak0 dr (75)
0 2i 2i
14 UNIT 1: Scattering

Z ∞
A∗k Ak0 0 0 0 0
I= dr[ei(k+k )r − e−i(k−k )r − ei(k −k)r + e−i(k+k )r ] (76)
(2i)2 0
If we do change of variable from r→-r the first and the last terms in the
integral will be the same except for their limits. For the last term the
limit of the integral is going to be from −∞ to zero.
Z
A∗ Ak0 ∞ 0 0 A∗ Ak0
I= k 2 dr[ei(k+k )r − e−i(k−k )r ] = 2π k 2 [δ(k + k 0 ) − δ(k − k 0 )]
(2i) −∞ (2i)
(77)
Since both k, k0 are positive the first delta function goes to zero and the
integral becomes
π
I = |Ak |2 δ(k − k 0 ) (78)
2
The normalization for a continuous spectrum will be equal to the delta
function therefore the normalization constant, Ak , is
r
2
Ak = (79)
π
This normalization constant is the same for the plane waves as well because
< k|k 0 >= δ(k − k 0 ).
• l =1
To calculate the wavefunction for the case of l = 1 we are going to use the
wavefunction we obtained for l = 0 and we shall define the momentum
operator, P+ = Px + iPy . The momentum operator commutes with the
Hamiltonian which means that they have the same eigenfunction. Let
us apply the Hamiltonian and P+ operator to the eigenfunction of the
Hamiltonian.
H|ϕi = E|ϕi (80)
HP+ |ϕi = P+ H|ϕi = EP+ |ϕi (81)
Once we apply this momentum operator, P+ , to the wavefunction for the
case of l = 0 we can find the wavefunction for the case of l = 1.
sin(kr) ∂ ∂ sin(kr)
P+ = −i~( +i ) (82)
r ∂x ∂y r
The derivative of r with respect to x and y is respectively,
∂r x ∂r y
= , = (83)
∂x r ∂y r
Therefore,
sin(kr) x + iy sin(kr) 0
P+ = −i~( )( ) (84)
r r r
and
x + iy
= sin θeiϕ = (constant)Y11 (85)
r
Scattering 15

· ¸
k cos(kr) sin(kr)
ϕk11 = AY11 (θ, φ) − (86)
r r2
For l = 1, m = −1, 0, 1 therefore there are three wavefunctions and they
are
− ϕk11 = ϕk10 and − ϕk10 = ϕk1−1 (87)

• l =2
Let us apply the momentum operator, P+ , for the second time to find the
wavefunctions for the case of l = 2.
sin(kr) x + iy sin(kr) 0
P+2 = A0 P+ ( ) (88)
r r r
Here one thing that you need to realize is that P+ operator commutes
with (x+iy). Therefore,

sin(kr) 1 sin(kr) 0
P+2 = A0 (x + iy)P+ ( ) (89)
r r r
sin(kr) 1 1 sin(kr) 0
P+2 = A0 (x + iy)2 ( ) (90)
r r r r
As you look for different l’s you shall realize that there is a pattern which
gives the wavefunction for a general l.

sin(kr) 1 1 d l sin(kr)
P+l = A0 (x + iy)l ( ) = ϕkll (91)
r r r dr r
The term of (x+iy)l = rl Yll and then if we define ρ = kr then the equation
above will become,
1 d l sin ρ
jl (ρ) = (−ρ)l ( ) . (92)
ρ dρ ρ
The functions which satisfy the equation above are called the Spherical Bessel
F unctions. The most general solution for the wavefunction written in
terms of spherical Bessel functions is therefore,

ϕklm = Ajl (kr)Ylm (θ, φ) (93)

q
2
where A = k π is the same for every l.

The differential equation for the spherical Bessel function is the radial part of
the Schrodinger equation.

1 l(l + 1)
(rR)00 − R + k2 R = 0 (94)
r r2
2 l(l + 1)
j 00 + jl0 + [1 − ]jl = 0 (95)
ρ ρ2
16 UNIT 1: Scattering

The asymptotic behaviour of the spherical Bessel functions which is important

for physics as r→ ∞ is known.

sin(kr − l π2 )
r → ∞; jl (kr) ≈ (96)
r
Because of the sine function above jl is a mixture of incoming and outgoing
waves. Now, let us expand sine in exponentials
π π
ei(kr−l 2 ) − e−i(kr−l 2 )
jl (kr) ≈ (97)
r

eikr −il π e−ikr il π

jl (kr) ≈ e 2 − e 2 (98)
r r
π
If we take the first phase e−il 2 as a constant then the equation becomes

eikr e−ikr ilπ

jl (kr) ≈ − e (99)
r r
eilπ = ±1 so we obtain a phase difference between the first term in the equation
above (the outgoing wave) and the second term (the incoming wave). This phase
different is fixed by the behaviour of the wavefunction where r→ 0. Generally
near zero when there is a potential we cannot just go to zero and keep the
same Schrodinger equation. In this type of situation what changes is this phase
difference between the incoming and outgoing waves. Now we shall examine the
behaviour of the wavefunction as r→ 0 to understand the physics. We know that
as r→ 0 the wavefunction behaves like r(l+1) . On the other hand, the spherical
Bessel function’s behaviour near r→ 0 is

jl (kr) ≈ (kr)l (100)

In figure 4, as r→ 0, jl goes to zero as well. Where it starts picking up from

zero is a number that is approximately l. Therefore, the larger l is, the more
jl stays at zero. The role that case plays in physics is that if kr . l then the
wavefunction R≈ 0. This means that the probability of finding a particle there
is zero. Even if it is not exactly zero it will be very very small.
Now if we switch on a potential, for example a nuclear potential, whose range
is . kl (shown in the figure above with the dashed lines) then the particles that
you send for scattering shall not feel the potential because the probability of
them being there is very very small. √In other words, these particles will never
see the potential. We know that k = 2mE ~ therefore, it is directly proportional
to the energy of the particles. The higher the energy, the smaller the ratio of
l
k which is the reason why we want to have high energy. With high energy you
can go smaller distances.
l is a quantum number, so large l means that we are in the limit of classical
physics. Let us consider the classical limit in more detail.
Scattering 17

Figure 4: Plot of jl vs. kr

CLASSICAL PHYSICS

Say we send a classical particle for scattering which has a momentum of p~ as

shown in figure 5. The distance b shown in the figure is called the impact parameter
which is defined as the perpendicular distance between the path of a particle and
the center of the scattering potential. The angular momentum of the particle is
then,
p
L = pb = ~ l(l + 1) ≈ ~l (101)

Figure 5: Scattering from a potential

18 UNIT 1: Scattering

The momentum of the particle and the angular momentum for large l where we
are expecting to see the classical results are
p = ~k and L ≈ ~l (102)
respectively. Using these two equalities for the impact parameter we obtain
L ~l l
b= = = (103)
p ~k k
In quantum mechanical perspective we have found that if the range is . kl
then the particle does not feel the potential. On the other hand, in classical
perspective we also see that if the range of the potential is less than the ratio
l
k then the particle doesn’t see the potential. Therefore, this case makes sense
for both perspectives and this explains the physical meaning of this kind of
wavefunctions.
Now let us assume that we have a central potential, V(r). Then we still have a
general solution which is
ϕklm (~r) = Rkl (r)Ylm (θ, ϕ) (104)
ukl
We write R = r and then the wavefunction becomes
· ¸
l(l + 1) 2mV
u00 + k 2 u − + u=0 (105)
r2 ~2
We have solved this equation without any potential and we have found that
the solution was spherical Bessel functions but now we have a potential. This
equation becomes a one dimensional equation as r turns to be x. The limits for
r is 0→ ∞ but for x is −∞ → ∞. Fortunately we know the boundary condition
for r = 0. This boundary condition says us that there is a wall at r = 0. so we
can think of r from −∞ → ∞ with a wall at r = 0. Therefore the potential will
be complicated for r > 0 and infinite for r < 0. This way we can think of this
case in one dimension. From the one dimensional point of view the scattering
means reflection only. If we think of eikr as the incident wave and e−ikr as the
reflected wave then the wavefunction when you look at r→ ∞ is
u = Aeikr + Be−ikr (106)
When r < 0 then
u=0 (107)
Since the potential is infinite in r < 0 this means there is no transmission.
A 2
Hence, the transmission coefficient, T=0 and since R+T=0, R=1=| B | . This
simple argument tells us that
|A| = |B| (108)
A and B are comlex numbers so even though their norms are equal there will
be a phase difference between them. If we set A=1 then
B = eiβ (109)
Scattering 19

New form of the wavefunction u is

u ≈ eikr + eiβ e−ikr (110)

iβ iβ
u ≈ e− 2 eikr + e 2 e−ikr (111)
β
u ≈ 2 cos(kr −
) (112)
2
This is the behaviour at infinity. We don’t know β unless we solve the equation
but we know how it should be. We can write the cosine in terms of sine.
β π
u ≈ 2 sin(kr − − ) (113)
2 2
To compare this solution with that of no potential we can write

lπ
u ≈ 2 sin(kr − + δl (k)) (114)
2
In other words, the difference between having a potential and not having a
potential is just a δ which is called the phase shift and δ in general will have
an index of l and is a function of energy. It is very important to calculate the
phase shift because it gives all of the information.
We have found the solution for particles coming from all directions. In exper-
iments, instead of bombarding a target from every direction scientists send a
beam on the target in one direction. But the solutions that we have found
are the most general solutions and they form a complete set because they are
the eigenfunctions of the Hamiltonian. We can write other solutions in terms of
these general solutions’ superposition. Hence, for any experimental arrangement
you should be able to write it in terms of the eigenfunctions of Hamiltonian. In
particular, the usual setup is to send the beam in z direction. It is a plane wave
eikz . We can write this plane wave in terms of eigenfunctions of the Hamiltonian.
∞ X
X l
eikz = alm ϕklm (~r) (115)
l=0 m=−l

The coefficients alm can be written as

Z
alm = d3 rϕ∗klm (~r)eikz (116)

Now we are going to write these alm in terms of spherical coordinates to calculate
the integral.

ϕklm (~r) = Cjl (kr)Ylm (θ, ϕ) and eikz = eikr cos θ (117)
Z
∗
alm = C d3 reikr cos θ jl (kr)Ylm
∗
(θ, ϕ) (118)
20 UNIT 1: Scattering

Since the ϕ appears just in the last term of the integral above we can calculate
the ϕ part separately.
Z 2π Z 2π
∗
dϕYlm (θ, ϕ) = Plm (θ) dϕeimϕ = 0 (unless m = 0) (119)
0 0

This simplifies the calculation and then

∞
X
eikz = Cl0 jl (kr)Yl0 (θ) (120)
l=0

∞
X
eikz = Cl jl (kr)Pl (cos θ) (121)
l=0

where you don’t even need to sum till infinity since the first three or four terms
suffice. For the purpose of normalization we can calculate the Cl0 and Cl .
p
Cl0 = il 4π(2l + 1) and Cl = il (2l + 1). (122)
π
where il = ei 2 l .
It’s nice to have the wave expressed in terms of the spherical harmonics because
they are orthogonal functions. When you express the wave in terms of a sum
you can see that the full wave is just the sum of partial waves. In practice,
keeping only the first two or three terms is good enough. Depending on the
value of l the partial waves will have different names. For instance, l = 0 is
called the S wave and l = 1 is called the P wave, and so on.
So what is the wave function going to be after scattering from the potential?
The incoming wave will turn into the outgoing wave, eikz → ϕ, which is the
sum of the incoming and outgoing wave ϕscattered . The entire wave will be,

ϕ = eikz + ϕs (123)
In order to figure out the scattering cross section we need to determine the
outgoing wave ϕs . To begin we look at the behavior of the wave very far away
- because that’s where we want to put the detector. The incoming partial wave
is built with spherical Bessel functions, jl (kr), and we can expand this in the
region where r → ∞, where it becomes a sin function.
· ¸
1 eikr −il π e−ikr il π
jl (kr) ∼ e 2 − e 2 (124)
2i kr kr
If we look at the total wave ϕ in the region r → ∞ and there’s gonna be a
mixture of incoming and outgoing spherical waves again. And we introduced
the phase shift δl which distinguishes the total wave from the incoming wave.
The overall phase is arbitrary so we chose to put it with the first term.
· ¸
1 eikr −il π e2iδl e−ikr il π
ϕ∼ e 2 − e 2 (125)
2i kr kr
Scattering 21

That will be the behavior of the total wave and I know in general it’ll behave
like this. Really what I’m after is the outgoing wave, which will be the difference
between the total and incoming waves. Therefore ϕs will take the form,

eikr −il π 2iδl eikr 1 iδl

ϕs ∼ e 2 [e − 1] = e sin δl (126)
2ikr kr il
We haven’t really solved for anything yet. Just simplified our general expression
for the outgoing wave to better see the behavior. So for each l there will be a
different solution for the scattered wave. If I look at the entire wave, I just have
to replace the jl with the behavior of jl at large r by this new behavior for ϕs .
∞
eikr X p
eikz → 4π(2l + 1)eiδl sin δl Yl (θ) (127)
kr
l=0

So that’s the expression for the outgoing wave, the entire outgoing wave. If you
ikr
remember, this whole thing multiplying ekr we called it f (θ, ϕ) before. The
entire sum is then this f (θ, ϕ) function.
∞
eikr X p eikr
eikz → 4π(2l + 1)eiδl sin δl Yl (θ) = f (θ, ϕ) (128)
kr r
l=0

∞
1 Xp
⇒ f (θ) = 4π(2l + 1)eiδl sin δl Yl (θ) (129)
k
l=0

So if I somehow manage to calculate those δl ’s I will be able to figure out the

entire f (θ). Remember the cross section is given in terms of this function. Recall
that f (θ) is sort of an amplitude because of the relation to the differential cross
section. This is the quantity that I’m after because that’s the experimental
result that I’ll be comparing things with.

dσ
= |f |2 (130)
dΩ

∞
1 Xp
|f (θ)|2 = | 4π(2l + 1)eiδl sin δl Yl (θ)|2 (131)
k2
l=0

From this I can find the total cross section. I’ll have to integrate the differential
cross section over all the angles. Since the variable ϕ was constant I only need
to integrate over θ.
Z Z Z π
dσ 2
σ= dΩ = dΩ|f (θ)| = 2π dθ sin θ|f (θ)|2 (132)
dΩ 0

I want to calculate the total cross section, so I’ll plug in for |f (θ)|2 and remem-
bering to keep track of the indices for f ∗ (θ) and f (θ).
22 UNIT 1: Scattering

∞ Z
1 X p p
0 + 1)e−iδl eiδl0 sin δ sin δ 0
σ= 4π(2l + 1) 4π(2l l l dΩYl∗ (θ)Yl0 (θ0 )
k2 0
l,l =0
(133)
This integral we can do pretty easily because of the orthonormal nature of the
spherical harmonics. It’s a Kronecker delta function equal to one only when
l = l0 . So the total cross section boils down to a manageable expression in
terms of the phases δl . And calculating the first few terms is usually good
enough.
∞
1 X
σ= 2 4π(2l + 1) sin2 δl (134)
k
l=0

As an interesting aside, lets look at the function f (θ) when θ = 0 and compare
it to our expression for the total cross section when we look in the forward
direction.
r
2l + 1
Yl (0) = (135)
4π

∞ ∞
1 Xp 1X
f (θ)|θ=0 = 4π(2l + 1)eiδl sin δl Yl (0) = (2l + 1)eiδl sin δl (136)
k k
l=0 l=0

Comparing this with the total cross section we see that σ has an extra factor of
4π
k and sin δl . This is a very general result and it even has it’s own name, the
Optical Theorem.

4π
σ= Imf (0) (137)
k

Example 1: Hard Sphere

We will now move on to do an example problem, scattering from a hard sphere.
Hard sphere means anything you send in gets reflected. Modeled as an infinite
potential wall at r 6 r0 .
½
0 if r > r0
V (r) =
∞ if r 6 r0
So how do we solve this problem? We can actually solve it exactly. Remember
that when r > r0 we have a free particle and must solve the free Schrodinger
equation. We already found it and lets call it Rk,l (r), a spherical Bessel function.

Rk,l (r) = Ajl (kr) A = constant (138)

Scattering 23

Figure 6: Hard sphere potential

Since we no longer have to worry about the behavior of the function at the origin
we need to keep both pieces of the wave function. This other Bessel function
we need is nl (kr). The n stands for Neumann.

Rk,l (r) = Ajl (kr) + Bnl (kr) (139)

Next thing to do would be to apply the boundary conditions, and setting A = 1
because we arbitrarily can.

B jl (kr0 )
Rk,l (r = r0 ) = 0 ⇒ = (140)
A nl (kr0 )
So this completely determines the function but we want an expression for δl
because this is what have expressions for. To do this we have to go far away
r → ∞.

1 π
Definition of δl : Rk,l (r) ∼ sin(kr − l + δl ) (141)
kr 2

1 π
jl (kr) ∼ sin(kr − l ) (142)
kr 2

1 π
nl (kr) ∼ − cos(kr − l ) (143)
kr 2

A π B π
⇒ Rk,l (r) = Ajl (kr) + Bnl (kr) ∼ sin(kr − l ) − cos(kr − l ) (144)
kr 2 kr 2
24 UNIT 1: Scattering

That’s the behavior of my wavefunction. When we compare this expression to

that of the definition of δl I should get δl . Expanding the definition in terms
of a trigonometric identity comparing the coefficients to match, then taking the
ratio of B
A,

B sin(δl ) jl (kr0 )
− = = tan(δl ) = (145)
A cos(δl ) nl (kr0 )
That’s the solution I’m looking for. Let’s look at it and try to figure out the
behavior to find the cross section. Begin with the lowest value, l = 0. We can
look up the spherical Bessel functions j0 (ρ) = sin(ρ)
ρ and n0 (ρ) = − cos(ρ)
ρ , then
the first phase shift.

tan(δ0 ) = − tan(kr0 ) ⇒ δ0 = −kr0 (146)

Plugging this into the expression for the total cross section and keeping just the
first term,

1
σ= 4π sin2 (kr0 ) (147)
k2
When is this approximation good? Remember that k has to do with the energy
of the beam. If I had a very low energy beam then kr0 ¿ 1 then σ = 4πr02 .
So in the low energy limit the cross section becomes a constant. The lower the
energy the better the approximation.
So let’s make this problem a little more difficult. You never really have a hard
sphere but you do have a potential well of finite depth. Such is life.

Example 2: Square Well

Outside the potential we have a free particle so the equation for Rk,l (r) will be
the same as that for the hard sphere. But now we do have to consider what
happens for r < r0 . So when we √
are the region of the potential we’ll have a new
2m(E+V0 )
value for k and that’ll be k 0 = ~ . So in the region of the potential,
where the nl → 0,

Rk,l (r) = Cjl (k 0 r) (148)

And outside the potential.

Rk,l (r) = Ajl (kr) + Bnl (kr) (149)

Now we have to match the solutions at the boundary, including the derivatives.

Ajl (kr0 ) + Bnl (kr0 ) = Cjl (k 0 r0 ) (150)

Akjl0 (kr0 ) + Bkn0l (kr0 ) = Ck 0 jl0 (k 0 r0 ) (151)

Scattering 25

Figure 7: Potential step

Dividing the two equations and dividing through by A, and writing − B

A =
tan(δl ),

jl (kr0 ) − tan(δl )nl (kr0 ) jl (k 0 r0 )

0 0 = 0 0 0 (152)
kjl (kr0 ) − tan(δl )knl (kr0 ) k jl (k r0 )
Now we only have one constant, δl and if we can solve this equation that will
give me δl and that’s all we need. So let’s do the case when l = 0. Let ρ = kr0
and ρ0 = k 0 r0 , and using the Bessel functions from earlier. Plugging in and
doing some algebra will give,

cos(ρ) cos(ρ0 ) sin(ρ) sin(ρ0 ) sin(ρ0 ) cos(ρ) sin(ρ) cos(ρ0 )

[ + ] tan(δ l ) = − (153)
ρ ρ0 ρ0 ρ
√
Then going to the low energy realm, ρ ¿ 1. In this limit ρ0 ≈ r~0 2mV0 and
ρ → 0.

cos(ρ0 ) sin(ρ0 )
[ ] tan(δ0 ) = − cos(ρ0 ) (154)
ρ ρ0

tan ρ0
⇒ tan(δ0 ) = ρ[ − 1] ≈ sin(δ0 ) ≈ δ0 (155)
ρ0
Now we can look into figuring out the first terms in the cross section and plugging
in for ρ = kr0 .

4π 2 tan ρ0
σ= 2
δ0 = 4πr02 [ 0 − 1]2 (156)
k ρ
26 UNIT 1: Scattering

Again the cross section is a constant and this time it’s less the four times the
cross section. And how much less depends on the depth of the potential because
of the dependence on ρ0 . Looking at the behavior of the tangent function, it’ll
blow up as V0 increases and then ρ0 approaches π2 . Then,

π 2 ~2
V0 < (157)
8mr02
If V0 becomes as big as this number then the cross section blows up and there
is lots of scattered particles. When this happens it’s called a resonance. Is the
cross section itself infinite? No, because we have made an approximation for
the cross section when we wrote it. When ρ0 is exactly π2 a value may blow up
but δ0 will not be infinite because we take the tangent of δ0 . That tells me the
δ0 will be precisely π2 . And if it’s exactly π2 I have to calculate the cross section
again using the same formula.

4π
σ= (158)
k2
So σ is not infinite but as k → 0 the cross section becomes very, very large. The
cross section is then large for low energies. So that means I have a resonance.
Which is something we should be able to see in an experiment. Lots of particles
coming out.
Lets now increase V0 above this number. The tangent of ρ will begin to decrease
and eventually you will get down to zero. If you keep increasing the potential
there will be a point where tan(ρ0 ) = ρ0 . If you solve that equation you’ll find
that ρ0 ≈ 4.5. At this point δ0 will be exactly zero and the cross section will
get a zero contribution from the first term, which is the most significant. So in
practice, the other terms are going to contribute but they’re going to contribute
very little. If you do an experiment and you have a potential thats exactly at
that value you’ll see nothing.
This is a famous result that was obtained back in 1923 when people didn’t
know anything about quantum mechanics. They didn’t have a wave equation,
Schrodinger hadn’t come up with it yet. So all of a sudden the experiment that
they performed, which was the scattering of electrons by rare gasses, you detect
nothing. They were wondering why you see nothing, and there was no expla-
nation in terms of classical mechanics. It’s a completely quantum mechanical
effect, because it has to do with interference. It’s only with interference that
you can get absolutely nothing. This was a great mystery and is known as the
Ramsauer-Townsend effect.
For E<0 the zeroth order phase shift, which is the most important term in low
energies, ρ, and ρ0 values are respectively,

tan ρ0
tan δ0 = ρ( − 1) (159)
ρ0
√
2mE
ρ = kr0 = r0 (160)
~
Scattering 27

p
0 0 2m(E + V0 )
ρ = k r0 = r0 . (161)
~
From the formula of the phase shift we can see that the resonance occurs when
tan ρ0 becomes infinite. In other words, the resonance occurs when ρ0 takes the
following values.
π 3π 5π
ρ0 ≈ , , ,... (162)
2 2 2
At these points the cross section becomes large, not infinite. Now let us consider
the values of V0 that will cause the resonances.

Bound States
For the energy less than zero the spectrum will be discrete. For this kind of
wavefunctions the solutions of the Schrodinger equation are

(I) u00I + k 02 uI = 0 ⇒ uI = A sin(k 0 r) + B cos(k 0 r) (163)

(II) u00II + k 02 uII =0 (164)

√
−2mE
Since E<0 say, k=iκ where κ = ~ and is a real number. Therefore, uII
becomes,
u00II − κ2 uII = 0 ⇒ uII = Ceκr + De−κr (165)
As r → ∞ uII → 0 and once r = 0, uI = 0. Therefore, coefficients B and C
are zero. Using the continuity, these two solutions and their derivatives need to
match at the boundary, that is, r= r0 .

A sin(k 0 r0 ) = De−κr0 (166)

k 0 A cos(k 0 r0 ) = −κDe−κr0 (167)

To cancel the coefficients A and D we divide by the two equations above and
we obtain
1 1
0
tan(k 0 r0 ) = − (168)
k κr0
Once we solve this equation above we obtain the energy, E. We get discrete
spectrum because the solution of this equation has finite number of solutions.
To solve this equation we are going to use the graphical method and we are
going to denote k 0 r0 as x and κr0 as y. Then the equation becomes,
1 1
tan(x) = − ⇒ y = −x cot x (169)
x y

If we look at the figure we see that one more function intersects with this graph.
2mV0 2
x2 + y 2 = r (170)
~2 0
The circles in the figure changes according to the value of V0 . For low V0 values
there is no intersection on the graphs. But there is one intersection where we
28 UNIT 1: Scattering

Figure 8: Plot of k 0 r0 vs. κr0

exactly get the first bound state for which x = π2 . We can continue in first bound
state until drawing a circle intersecting with the second part of the graph and
so on. Therefore,
2mV0 2 π
x2 + y 2 = 2
r0 = ( )2 (171)
~ 2
~2 π 2
V0 < ⇒ N O BOU N D ST AT ES! (172)
8mr02
The values π2 , 3π 5π
2 , 2 ... are also the values for the resonance. This means that as
V0 changes, and there’s a resonance, a new bound state has been added. This
result relates the resonance, bound states and the cross section.

Absorption
We have been working on the elastic collisions so far but Nature does not work
this way. We need to study inelastic scattering as well.
• Elastic Collisions:
∂ρ ~ ~
+∇·J =0 ⇒ Conserved current (173)
∂t
∂ρ
=0 ⇒ N o dependence on time (174)
∂t
~ · J~ = 0
∇ ⇒ Conservation law (175)
where
~
J~ = ~ − ψ ∇ψ)
(ψ ∗ ∇ψ ~ (176)
2mi
Scattering 29

• Inelastic Collisions:
V = V1 + iV2 (177)
~2 2 ∂ψ
− ∇ ψ + V ψ = i~ (178)
2m ∂t
∂ρ ∂ψ ∗
=ψ + C.C. (179)
∂t ∂t
∂ρ 1 ~2 ∗ 2
= [− ψ ∇ ψ + V |ψ|2 + C.C. (180)
∂t i~ 2m
∂ρ ~ ∗ 2 V2
=− ψ ∇ ψ + |ψ|2 + C.C. (181)
∂t 2mi ~
∂ρ ~ ~ ∗~ ~ ∗ ) + 2V2 |ψ|2
=− ∇(ψ ∇ψ − ψ ∇ψ (182)
∂t 2mi ~
∂ρ ~ · J~ + 2V2 ρ
= −∇ (183)
∂t ~
Since there is no time dependence,

~ · J~ = 2V2 ρ
∇ (184)
~

Using divergence theorem the total current is,

Z Z Z
2V2 3
I= J~ · dS
~= ~ · Jd
∇ ~ 3r = ρd r (185)
S V ~

We are interested in this situation because this appears in absorption. If there

were no absorption then V2 would be zero and there wouldn’t be a term like
I. Flux means the number of particles passing through a surface per time but
they can both go out or come in.Therefore, I is the absorbed flux.

Fabs = −I (186)

If we measure the entire cross section for the absorption it is,

Fabs
σabs = (187)
Jinc

We have already calculated the value of Jinc for incoming plane wave.

~k
Jinc = = V elocity (188)
m
Then the absorbed cross section is calculated from
Z Z
2m m
σabs = − 2 3
V2 ρd r = − J~ · dS.
~ (189)
~ k ~k
30 UNIT 1: Scattering

To calculate this integral we will go through the same steps as the elastic scatter-
ing. First of all, we place the detector far away, that is, r → ∞. The incoming
wave is a plane wave which is eikz and it turns into the total ϕ,
∞ p
X eikr −il π 2iδl e−ikr il π
ϕ= 4π(2l + 1)[ e 2e − e 2 ]Yl0 (θ) (190)
2ikr 2ikr
l=0

As you go over the same procedure as in the elastic scattering part you get
exactly the same answer. The crucial difference is that since the potential has
the imaginary part the phase shift, δl , has an imaginary part. Hence we can
write δl = αl + iβl . Since we know ϕ we can find J and the cross section.
Z Z
m mr2
σabs = − Jr r2 dΩ = − Jr dΩ (191)
~k ~k

~ ∂ϕ
Jr =(ϕ∗ − C.C.) (192)
2mi ∂r
Let us calculate the partial differential of ϕ with respect to r.
∞
∂ϕ X p eikr eikr −il π 2iδl
= 4π(2l + 1)[( − )e 2 e + C.C.(δl = 0)]Yl0 (θ) (193)
∂r 2r 2ikr2
l=0

eikr
As r → ∞ the term 2ikr 2 may be ignored and let us call the coefficient as bl .
∂ϕ 1 X
= bl Yl0 (θ) (194)
∂r 2r
l

1
For ϕ we already have an expression and if we take 2r out and write the rest as
a coefficient, cl , then,
1 X
ϕ= cl Yl0 (θ) (195)
2r
l
Now we can calculate the absorption cross-section.
Z
mr2 ~ X 1
σabs = − dΩbl0 c∗l Yl∗0 0 (θ)Yl0 (θ) + C.C. (196)
~k 2mi 0 (2r)2
l,l

Since spherical harmonics are orthogonal this expression can be simplified.

1 X ∗
σabs = − bl cl + C.C. (197)
8ki
l

As expected you can see from the equation above the cross-section does not
depend on r. The cross-section should not depend on where you put the detector.
The flux has to be the same no matter where you are. We need to calculate the
coefficient. To do this we are going to calculate the following expression.
8π(2l + 1) −4βl
bl c∗l − C.C. = (e − 1) (198)
ik
Scattering 31

Then the absorption coefficient is,

∞
π X
σabs = (2l + 1)(1 − e−4βl ). (199)
k2
l=0

If the phase shift due to absorption is zero then it is obivious from the equation
that the absorption coefficient is zero. Remember when there is an elastic scat-
tering we had a very similar expression for the cross-section. To remind, the ϕ
and the scattering amplitude, f (θ), were

eikr
ϕ= f (θ) (200)
r

∞
1 Xp
f (θ) = 4π(2l + 1)eiδl sin δl Yl0 (θ) (201)
k
l=0

These two expressions are still the same for the inelastic scattering. But now
we need to be careful because δl is not real anymore. Hence, we can write the
terms with
1
eiδl sin δl = (e2iδl − 1) (202)
2i
Using the terms above the elastic scattering cross section becomes,
Z ∞
π X
σel = dΩ|f |2 = (2l + 1)|e2iδl − 1|2 (203)
k2
l=0

Now we don’t have sin2 δl because δl is not real. For imaginary δl ,

| e2iδl − 1 |2 =| e2iδl |2 +1 − 2Re2iδl = e−4βl + 1 − 2e−2βl cos(2αl ) (204)

The total cross-section will be the sum of these cross-sections.

σ = σabs + σel (205)

∞
2π X
σ= (2l + 1)[1 − e−2βl cos(2αl )] (206)
k2
l=0

The last term in the sum above is Re[1 − e2iδl ] where,

Sl = e2iδl (207)

S is called the scattering matrix and Sl is the eigenvalues of this scattering

matrix.
32 UNIT 1: Scattering

Optical Theorem
Remember that the optical theorem relates the cross-section to the scattering
amplitude in the forward direction. For the inelastic scattering case the optical
theorem is,
∞
1 X
f (0) = (2l + 1)(e2iδl − 1) (208)
2ik
l=0

k
Imf (0) = σ (209)
4π
This shows that the optical theorem is satisfied by the inelastic scattering case
as well.
UNIT 2

Angular momentum

Notes by J. Mazer, A. Holt, and M. Rezaee

Angular Momentum
The angular momentum classically is given as:
~ = ~r × p~
L (1)

When we write this as an operator is it shown that the components satisfy the
communtation relation:
[Lx , Ly ] = i~Lz (2)
If you want to understand nature, you have to be able to ask what values of L
can you get? When we call the L, J, we see the same commutation relations are
satisfied.
[Jx , Jy ] = i~Jz (3)
We can find all the possible combinations by writing:

J± = Jx ± i~Jy (4)

and if writing J 2 as a vector, it is the sum of the squares of the components:

J 2 = Jx2 + Jy2 + Jz2 (5)

J 2 commutes with all the components, in particular lets look atJz :

[J 2 , Jz ] = 0 (6)

Since these commute we can label them with common eigenvalues. Calling the
eigenvalue λ and pulling out the ~2 for dimensional reasons we get:

J 2 | λ, mi = λ2 ~2 | λ, mi (7)

Now as a state of Jz :
Jz | λ, mi = ~ | λ, mi (8)
34 UNIT 2: Angular momentum

When acting with J− our m state is lowered by 1.

J− | λ, mi ∝| λ, m − 1i (9)

In order to avoid the negative norm states we take the norm of a state and
square it as:
k J− | λ, mi k2 = [λ2 − m(m − 1)]~2 ≥ 0 (10)
Since it has to be positive we see:

m2 − m − λ2 ≤ 0 (11)

When this is thought of a polynomial of m it must satisfy the relation:

m− ≤ m ≤ m+ (12)

Solving the polynomial we get the following roots:

1 p
m± = (1 ± 1 + 4λ2 ) (13)
2

m = m− + integer (14)
Now acting with J+ we go through the same story.

J+ | λ, mi ∝| λ, m + 1i (15)

Writing the norm:

k J+ | λ, mi k2 = [λ2 − m(m + 1)]~2 (16)

It follows that:
m2 + m − λ2 ≤ 0 (17)
Where the roots to our polynomial are:
1 p
m0± = (−1 ± 1 + 4λ2 ) (18)
2

m0− ≤ m ≤ m0+ (19)

m = m0+ + integer (20)

Rearranging gives:
m0+ − m− = integer (21)
Writing what these numbers are in terms of λ we get:
1 p 1 p
(−1 + 1 + 4λ2 ) − (1 − 1 + 4λ2 ) = integer (22)
2 2
Angular momentum 35

This integer is referred to as 2j.

p
− 1 + 1 + 4λ2 = integer = 2j (23)
Then: p
1 + 4λ2 = integer = 2j + 1 (24)
1 3
for j = 0,2 , 1, 2 , ....
Squaring we get:
1 + 4λ2 = (2j + 1)2 = 4j 2 + 4j + 1 (25)
Solving we get:
λ2 = j(j + 1) (26)
1
Now we will present the first non-trivial case of j = We all consist of spin 12
2.
particles. Its not angular momentum anymore because we can never have values
of halves. This has no classical analog to explain this, for this case we have:
1
m=± (27)
2
We have two states given as:
 µ ¶

 11 1
 | 2 2i =| +i =
0
µ ¶
| j, mi = (28)

 1 0
 | − 12 i =| −i =
2 1
The particles that carry that spin are called spinors. Such examples are elec-
trons (point particles) , protons, and neutrons which are made of quarks (point
particles).

ROTATIONS

We want to understand what happens when you rotate something that is a

spinor. Under rotation the vector components of r are transformed as:
   
x x0
~r =  y  ⇒  y0  (29)
z z0
36 UNIT 2: Angular momentum

by using the transformations for rotation about the Z-axis by an angle θ we

have:

x0 = cos(θ)x − sin(θ)y
y0 = sin(θ)x + cos(θ)y (30)
z0 = z

We get the rotation matrix which is multiplied by our original vector compo-
nents:   
cos(θ) − sin(θ) 0 x
 sin(θ) cos(θ) 0  y  (31)
0 0 1 z

For small θ, we make a Taylor expansion that gives:

  
θ −θ 0 x
θ θ 0  y  (32)
0 0 1 z

We can separate the 1’s from the θ’s we can write this as
    
x 0 −1 0 x
= II  y  + θ 0 0 0  y  (33)
z 0 0 1 z

The angular momentum is held responsible for generating such rotations, to see
this we express this as:
· ¸
∂ϕ ∂ϕ
ϕ(~r) → ϕ(x − θy, y + θx, z) = ϕ(x, y, z) + θ −y +x (34)
∂x ∂y

where are coordinates are transformed in our function. Since we are rotating
about the Z-axis, the proper component would be the Z component of the
angular momentum. This is shown and then where it is converted to its operator
form. Since θ is small we need to Taylor expand this function. Expanding in
Angular momentum 37

θ we need to take the derivative, but we are taking the derivative of a variable
that depends on another variable so we need to use the chain rule.
µ ¶
∂ ∂
Lz = xpy − ypx → −i~ x −y (35)
∂y ∂x

If we act on ϕ: µ ¶
∂ϕ ∂ϕ
Lz ϕ = −i~ x −y (36)
∂y ∂x
Pulling out common factors we see:
iθ iθ θ
= ϕ(~r) + Lz ϕ + ... = (II + Lz )ϕ + ... = ei ~ Lz ϕ (37)
~ ~
This shows us that the wave equation is being operated on by an operator. If
we choose another axis for rotation we can always just redefine our axis to meet
the requirements. We can write the exponent as θn̂ · L~ = θ~ · L
~ when we have an
arbitrary direction. θ is not the angle about the Z-axis, but the angle we rotate.
Thinking of this as an operator, if we have a spinor, we can replace L by S:

~ = ~ ~σ
S (38)
2
where σ is just our familiar Pauli matrices.
We can call this matrix R, and in terms of σ we have:
i ~ ~ i ~
R = e ~ θ·S = e 2 θ·~σ (39)

When we act on a spinor we get a 2x2 matrix. This matrix operator on a spinor
rotates operator A like A → RAR−1 . We get a 2 component object.
Lets choose n̂ = ẑ to identify the Z-axis as the direction we are rotating about.
Then if θ is small, keeping just the first two terms:
i i
R = e ~ θSz = II + θSz + ... (40)
~
We now want to find out what happens to S, in particular taking the component
Sx and rotating we see:

Sx → RSx R−1
i i
= (II + θSz )Sx (II − θSz ) + ..
~ ~
i i
= Sx + θ§z Sx − θSx Sz + ...
~ ~
i
= Sx + θ [Sz , Sx ] + ... = Sx − θSy + ... (41)
~
So now, going back to the matrix, we want to see how to write it. We want
to keep n̂ in the Z-direction. R is expressed with an exponent which can be
38 UNIT 2: Angular momentum

expanded in a series to yield:

X∞ µ ¶n
i 1 iθ
R = e 2 θσz = σzn
n=0
n! 2
à µ ¶2 ! à µ ¶3 !
1 iθ iθ 1 iθ (42)
= II 1 + + .. + σz + + ..
2! 2 2 3! 2
θ θ
R = cos II + i sin σz
2 2
This last equation for R is a single matrix and is what we get when we have
rotation around the Z-axis. All the even terms are proportional to the identity
and the odd terms are proportional to σz. If we replace σz = n̂ · ~σ we get the
rotation around an arbitrary axis.
The table below shows what σ’s behavior is for different values of n.

n σzn
0 II
1 σz
2 II
3 σz
·· ··

If we rotate by θ = 2π we get:

R(2π) = cos πII + i sin πσz = −II

(43)
R(4π) = II

Therefore we see we need to rotate by 4π to get back to where we started.

Our everday experiences would suggest 2π, but our everyday experience uses
operator and observables. A → RAR−1 = (−II)A(−II) = A The physicial
consequences of what we see for R is that if we operate on a spinor by rotating
2π, we get: R(2π) | ϕi = − | ϕi. Which gives us −ϕ, but with a phase, and
as far as the probability is concerned this sign doesn’t matter, but the phase is
very important when dealing with interference effects.

Two Spinors
Two Spinors has four states: | ++i, | +−i, | −+i, | −−i. They are eigenstate
of spin operators: S~1 , and S~2 , so we have a total of six components. For the
particles we now get states: S12 , S1z , S22 , S2z .
If we operate S12 we get:

3 2
S12 | ±±i = ~ | ±±i (44)
4
Angular momentum 39

They form a basis so, S12 is 4x4 diagonalizable matrix and we get this same
result acting on any of our 4 initial states as indicated by the ± signs.
Since they form a basis we always get the same eigenvalue we get the same
result for S12 and S22 .
3
S12 = ~2 II = S22 (45)
4
If we want to write S1z , we act on the 1st component (1st particle) of our four
initial vectors and get:
 
+1 0 0 0
~  0 +1 0 0
S1z =   (46)
20 0 −1 0 
0 0 0 −1

It is diagonizable with eigenvalues 1, 1, -1, -1.

If we do the same thing for S2z , we act on the 2nd component (2nd particle) of
each of our four states.
 
+1 0 0 0
~  0 −1 0 0
S2z =  
 (47)
2 0 0 +1 0 
0 0 0 −1

TOTAL SPIN
If we want ~ = S~1 + S~2 which follows the commuation
£ to define
¤ total spin then S
relation S 2 , Sz = 0. This means that they commute. We can ask what are
their common eigenstates? To find them we write Sz in its matrix form and
again it is diagonizable.
 
1 0 0 0
0 0 0 0 
Sz = S1z + S2z = ~ 
0 0 0 0 
 (48)
0 0 0 −1

Writing S 2 , we already know S12 and S22 are the identity matrix.

S 2 = (S~1 + S~2 )
(49)
= S12 + S22 + 2S~1 · S~2

We expand this dot product and use S+ and S− ,

S± = (Sx ± iSy ) (50)

to get:

2S~1 · S~2 = 2S1x S2x + 2S1y S2y + 2S1z S2z

(51)
= S1+ S2− + S1− S2+ + 2S1z S2z
40 UNIT 2: Angular momentum

By acting on a state with S+ and S− , we see:

S+ | +i = 0
S− | +i = ~ | −i
S+ | −i = ~ | +i
S− | −i = 0 (52)

Now if we act with S 2 on the first state | ++i:

3 2 ~
S 2 | ++i = ~ | ++i + 2( )2 | ++i = 2~2 | ++i (53)
2 2
And now the second state and third states. We get the same result as the | +−i
when acting on | −+i.

S 2 | +−i = S 2 | −+i = ~2 | +−i + ~2 | −+i (54)

Acting on | −−i we get the same result as | ++i

S 2 | −−i = 2~2 | −−i (55)

Since we know how to act on the basis we now know the matrix of S 2 . We are
no longer diagonal.  
2 0 0 0

2 0 1 1 0
2
S =~   (56)
0 1 1 0
0 0 0 2
To find the common eigenstates we need to diagonalize this matrix. So we first
need to find the eigenvalues of this matrix.
 
2−λ 0 0 0
 0 1−λ 1 0 
0 = det(S 2 − λII) =  0
 (57)
1 1−λ 0 
0 0 0 2−λ

Solving for the eigenvalues by find the determinant gives:

£ ¤
= (2 − λ)2 (1 − λ)2 − 1 = (2 − λ)2 (λ2 − 2λ) = λ(λ − 2)3 (58)

with eigenvalues λ = 2, 0 The degeneracy is 2l + 1, such that λ = l(l + 1). So

for λ = 0 → l = 0 and for λ = 2 → l = 1

2 spinors
A spinor means the particle which has spin half and when we have two spinors
we have 4 states. The can be written as:

| ++i (59)
Angular momentum 41

| +−i (60)
| −+i (61)
| −−i (62)

S± = (Sx ± iSy ) (63)

~ = S~1 + S~2
S (64)
2
[S , Sz ] = 0 (65)
I want to find the common eigenstates of these two.
 
2 0 0 0
 0 1 1 0
S 2 = ~2 
0 1 1
 (66)
0
0 0 0 2
The eigenvalues are 1 and 0. The l = 1 it has degeneracy of three. The l = 0
has the degeneracy of one. We want to figure out the eigenstates and:
| ++i =| 11i (67)

| −−i =| 1 − 1i (68)
| −−i =| 1 − 1i (69)

1
| 10i = √ (| +−i+ | −+i) (70)
2
1
| 10i = √ (| +−i− | −+i) (71)
2
1 1
m1 = − , + (72)
2 2
1 1
m1 = − , + (73)
2 2
1 1
m2 = − , + (74)
2 2
The original number of states, which is 3 + 1 = 4. All three of these states have
a spin of one.

S+ | 11i =0
S− | 11i =| 10i
S− | 10i =| 1 − 1i
(75)
The states correspond to: S = 1 being all symmetric and S = 0 being antisym-
metric.
42 UNIT 2: Angular momentum

Alternative Derivation:
M = m1 + m2 (76)

| ++i =| 11i (77)

what we should do is to select highest m1 and the highest m2
1
m1 = m2 = + (78)
2
Therefore M = 1 and S must be equal to 1. How we can find the forth state?
We know the 4th state should be orthogonal to all the previous S=1 states.
Therefore:
1
| 00i = √ (| +−i− | −+i) (79)
2
So lets generalize our results: Imagine instead of two spin half particles we have
one angular momentum J1 and another angular momentum J2 . J could be the
spin or a combination of angular momentum and spin. We know the eigenstates
of the J1 which is:
| j1 m1 i (80)
and for J2 which is
| j2 m2 i (81)
and when we put the two particle in one system then we will have:

| j1 j2 ; m2 m2 i (82)
for the eigenstate. Now we want to build states which belong to total angular
momentum and we define the total angular momentum as:

J~ = J~1 + J~2 (83)

2
And the set of commuting observables {J , Jz }. So we should be able to find
states which have quantum numbers appropriate for these two and they will be:

| J; M i (84)
So we want to find the common eigenstates of J 2 and Jz :

J 2 | J; M i = ~2 J(J + 1) | J; M i (85)

Jz | J; M i = ~M | J; M i (86)
Now we want to find all the eigenvalues. We know that m1 = −j1 , ..., j1 and
m2 = −j2 , ..., j2 so we will have space with (2j1 + 1) ∗ (2j2 + 1) different eigen-
values and that is how many states we have the highest possible M is j1 + j2
and there is only one state which has the highest M?
Angular momentum 43

| j1 j2 ; j1 j2 i =| j1 + j2 ; j1 + j2 i (87)
and

M = −J, ..., J (88)

By acting J− we should be able to produce other states. Lets count the states
which we have here.
Example:
Lets choose j1 = 3 and j3 = 2 how many states do we have originally:

(2j1 + 1) × (2j2 + 1) = (2 × 3 + 1) × (2 × 2 + 1) = 7 × 5 = 35 (89)

So J = 5, 4, 3, 2, 1, 0 but can we really have all these numbers? The degeneracy
of J = 5 is 11 = 2 × 5 + 1 and for the other J:

J Degeneracy
4 9
3 7
2 5
1 3
0 1

If we add all these degeneracies we will have 11 + 9 + 7 + 5 + 3 + 1 = 36 and

this tells us that we can not have J = 0. So the Jmax = 5 and the Jmin = 1.
We can say:

Jmax = j1 + j2 (90)
Jmin = j1 − j2 (91)
By counting states
if j1 ≥ j2 and we do know that for each J we have 2J + 1 multiplicity
jX
1 +j2 jX
1 +j2 jX
1 +j2

(2J + 1) = 2 J+ 1 (92)
j1 −j2 j1 −j2 j1 −j2

jX
1 +j2 j1 −j
X 2 −1

= 2( J− J) + j1 + j2 − (j1 − j2 ) + 1 (93)
1 1

by using:
N
X N (N + 1)
J= (94)
1
2

We will have:
44 UNIT 2: Angular momentum

(j1 + j2 ) (j1 + j2 + 1) − (j1 + j2 − 1) (j1 − j2 ) + 2j2 + 1 (95)

By doing simple algebra here we will have:
2 2
= (j1 + j2 ) + j1 + j2 − (j1 − j2 ) + j1 − j2 + 2j2 + 1 (96)

So we can write it as:

= 4j1 j2 + 2j1 + 2j2 + 1 (97)
By factoring this:

= (2j1 + 1) (2j2 + 1) (98)

That is exactly the number of states we began with and tells us that our guess
was very good. So:
j1 + j2 ≥ J ≥ |j1 − j2 | (99)
This gives the angular momentum a nice vector sense and in this picture we can
visualize it as a vector. Lets construct all those eigenstates by starting from the
highest eigenstate and acting J− on that state.

Mhighest = j1 + j2 (100)

| j1 j2 ; j1 j2 i =| j1 + j2 ; j1 + j2 i (101)
The general formula for J− is given as:
p
J− | J, M i = ~ J (J + 1) − M (M − 1) | J, M − 1i (102)

acting J− on the highest state we have:

J− | j1 + j2 , j1 + j2 i = (103)
p
~ (j1 + j2 ) (j1 + j2 + 1) − (j1 + j2 − 1) | j1 + j2 , j1 + j2 − 1i (104)

p
= ~ (j1 + j2 ) | j1 + j2 , j1 + j2 − 1i (105)
So this tells us we can have new state by acting J− on the highest one the new
state:
1
| ψi =| j1 + j2 , j1 + j2 − 1i = p J− | j1 + j2 , j1 + j2 i (106)
~ 2 (j1 + j2 )

To figure out the new state we write it as:

1
= p (J1− + J2− ) | j1 j2 , j1 j2 i (107)
~ 2 (j1 + j2 )
Then we write it using J1− and J2− .
Angular momentum 45

1 ³ p
= p ~ j1 (j1 + 1) − j1 (j1 − 1) | j1 j2 ; j1 − 1j2 i
~ 2 (j1 + j2 )
p
+~ j2 (j2 + 1) − j2 (j2 − 1) | j1 j2 ; j1 j2 − 1i ) (108)
So by summarizing it we have:
s s
j1 j2
= | j1 j2 ; j1 − 1j2 i + | j1 j2 ; j1 j2 − 1i (109)
j1 + j2 j1 + j2

So this is our new state. By repeating this procedure we can have all the M
values. We started with M = j1 + j2 and we have all of them now:
M = j1 + j2 , j1 + j2 − 1, . . . , −j1 − j2 (110)
So we know two states which have the M that we want. Do we know any other
states? There are no other states and we can find that by a simple arithmetic.
Here we only have two choices and we want to have the state with highest M
and would have to be a linear combination of these two and also it has to be
orthogonal to these two states and this is:
s s
j1 j2
= | j1 j2 ; j1 − 1j2 i − | j1 j2 ; j1 j2 − 1i (111)
j1 + j2 j1 + j2

and this tells us

s s
j1 j2
| j1 + j2 + 1j1 + j2 + 1i = | j1 j2 ; j1 − 1j2 i −| j1 j2 ; j1 j2 − 1i
j1 + j2 j1 + j2
(112)
So we constructed the next state by acting J− on the highest M.
We can repeat this until we have all the states. All these has been summarized
by the Clebsch-Gordan Coefficients.

Clebsch-Gordan Coefficients
We start with:
| j1 j2 ; m1 m2 i =⇒| JM i (113)
RHS and LHS of the above form an orthonormal bases.
j1
X j2
X
| JM i = aJM m1 m2 | j1 j2 ; m1 m2 i (114)
m1 =−j1 m2 =−j2

most of the coefficients (a) are 0 and if you multiply both sides by bra of the
RHS we will have the coefficients:
aJM m1 m2 = hj1 j2 ; m1 m2 | JM i (115)
46 UNIT 2: Angular momentum

and these coefficients are all real (and remember that most of them are zero)
and so if we take the complex conjugate we would have the same number:

= hJM | j1 j2 ; m1 m2 i (116)

and we know that:

M = m1 + m2 (117)
otherwise the coefficient is zero and this reduces the number of coefficients and
also we figure out:

|j1 − j2 | ≤ J ≤ j1 + j2 (118)
and that relation also reduces the number of coefficient to.
Example: We have

hj1 j2 + j1 j2 | j1 + j2 ; j1 + j2 i = 1 (119)

So here it should be 1 and if we have:

s
j1
hj1 j2 ; j1 − 1j2 | j1 + j2 ; j1 + j2 − 1i = (120)
j1 + j2
q
j1
the coefficient is j1 +j2 and also for:
s
j2
hj1 j2 ; j1 j2 − 1 | j1 + j2 ; j1 + j2 − 1i = (121)
j1 + j2

So these are the coefficients.

j1
X j2
X
| j1 j2 ; m1 m2 ihj1 j2 ; m1 m2 | = 1 (122)
m1 =−j1 m2 =−j2

X
aJM m1 m2 aJ0M 0m1 m2 = δJJ0 δM M 0 (123)

jX
1 +j2 J
X
| JM ihJM | = 1 (124)
J=|j1 −j2 | M =−J

X
hj1 j2 m1 m2 | aJM m1 m2 aJM m01 m02 i = δm1 m01 δm2 m02 (125)

because a’s are orthonormals to each other.

Last time we spoke of Clebsch-Gordan Coefficients. You obtain these from
transforming basis.
X
|JM i = AJM j1 j2 m1 m2 |j1 j2 ; m1 m2 i (126)
Angular momentum 47

where

AJM j1 j2 m1 m2 = hj1 j2 ; m1 m2 | JM i = hJM | j1 j2 ; m1 m2 i (127)

jX
1 +j2 J
X
|j1 j2 ; m1 m2 i = AJM j1 j2 m1 m2 |JM i (128)
J=|j1 −j2 | M =−J

and
X X
|j1 j2 ; m1 m2 i hj1 j2 ; m1 m2 | = II (129)

X X
|JM i hJM | = II (130)

Recurrence Relations
Act on equation (1.1) with the J− operator.

J~ = J~1 + J~2 (131)

From last semester we know:
p
J− |JM i = ~ J(J + 1) − M (M − 1) |JM−1 i (132)

But now our state, |JM i, is made up of two states. Introduce a sum to include
the lowering operator acting on m1 and m2 .

X p
J− |JM i = AJM j1 j2 m1 m2 [~ j1 (j1 + 1) − m1 (m1 − 1) |j1 j2 ; m1 − 1m2 i
p
+~ j2 (j2 + 1) − m2 (m2 − 1) |j1 j2 ; m1 m2 − 1i]

Take the inner product:

D 0
¯
0 ¯
E
j1 j2 ; m1 m2 ¯ j1 j2 ; m1 m2 − 1 (133)

Using equation (1.3) we obtain

p
(1.8) = ~ J(J + 1) − M (M − 1)aJM −1j1 j2 m0 m0
1 2
X p
= aJM j1 j2 m1 m2 ~ j1 (j1 + 1) − m1 (m1 − 1)δm1 −1m0 δm2 m0
1 2
p
+ ~ j2 (j2 + 1) − m2 (m2 − 1)δm1 m δm2 −1m
0 0
1 2
p
= ~ j1 (j1 + 1) − m1 (m1 + 1)aJM j1 j2 m0 +1m0
1 2
p
+ ~ j2 (j2 + 1) − m2 (m2 + 1)aJM j1 j2 m0 m0 +1 (134)
1 2
48 UNIT 2: Angular momentum

Example 1: j1 = j2
(Remember we can rewrite |lmi as Ylm (θ, ψ))

If j1 = j2 = 1 and m1 , m2 = −1, 0, 1, Then I’m interested in the state

|11; m1 m2 i → Ylm
1
1
(θ, ψ)Ylm
2
2
(θ, ψ) (135)
Where J can take on three values: 0, 1, 2

First case: J = 2
Start with the highest state |22i = |11; 11i and act on it with J− .
p
J− |22i = ~ 2(2 + 1) − 2(2 − 1) |21i = 2~ |21i (136)

We can also write (1.11) as:

|21i = A |11; 01i + B |11; 10i (137)

√
By normalization we can determine the coefficients are 2~. Therefore our
normalized state is: |21i = √12 [|11; 01i + |11; 10i]

→ Repeat this process for |21i , |20i, and |2 − 1i.

1
|20i = √ [|11; 1 − 1i + 2 |11; 00i + |11; −11i] (138)
6

1
|2 − 1i = √ [|11; 0 − 1i + |11; −10i] (139)
2

|2 − 2i = |11; −1 − 1i (140)
Take time to notice that the |2 − 1i and |2 − 2i states are mirror images of the
|21i and |22i states (with m = −m).

Second case: J = 1
From inspection we can deduce |11i = √12 (|11; 10i − |11; 01i).
By applying the lowering operator J− to |11i we see:
1
J− |11i = |10i = √ (|11; 1 − 1i − |11; −11i − 0 |11; 00i) (141)
2
Therefore,
1
|10i = √ (|11; 1 − 1i − |11; −11i) (142)
2
1
|1 − 1i = √ (|11; 0 − 1i − |11; 0 − 1i −) (143)
2
Angular momentum 49

Third case: J = 0
|00i = a |11; 1 − 1i + b |11; −11i − c |11; 00i (144)
We can deduce a, b, and c by exploiting orthogonality.

h00 | 10i = √1 (a − b) = 0
2
h00 | 20i = √1 (a
+ 2c − b) = 0
6
a = b → 2b + 2c = 0 → a = b = −c
Normalize: a2 + b2 + c2 = 1 → a = √13

1
|00i = √ (|11; 1 − 1i + |11; −11i − |11; 00i) (145)
3

Example 2: j1 = l1 and j2 = l2
|l1 l2 ; m1 m2 i → Ylm
1
1
(θ1 , ψ1 )Ylm
2
2
(θ2 , ψ2 )

|LM i → Φ(θ1 , ψ1 ; θ2 , ψ2 ) =
XX
hl1 l2 ; m1 m2 | LM i Ylm
1
1
(θ1 , ψ1 )Ylm
2
2
(θ2 , ψ2 )

XX
Ylm
1
1
(θ1 , ψ1 )Ylm
2
2
(θ2 , ψ2 ) = hl1 l2 ; m1 m2 | LM i Φ(θ1 , ψ1 ; θ2 , ψ2 ) (146)

Two Coincident particles

θ = θ1 = θ2 and ψ = ψ1 = ψ2

ΦLM = A(θ, ψ) represents the probablity amplitude (wavefunction).

Where |A|2 = probability of particles ”coinciding” in a given direction.

X
Think of A as a single particle: ALM = alm Ylm
Apply Lz to ALM :

∂
Lz ALM (θ, ψ) = −i~ ALM (147)
∂ψ
∂ ∂ ∂
− i~ ALM = −i~ ALM + −i~ ALM (148)
∂ψ ∂ψ1 ∂ψ2

Lz ALM (θ, ψ) = (m1 ~ + m2 ~)ALM = M ~ALM (149)

This tells us that ALM must be a single spherical harmonic and shares common
eigenstates
50 UNIT 2: Angular momentum

Therefore ALM is no longer a sum.

ALM = aLM YLM (150)

and
Lz ALM = M ~ALM (151)

L2 ALM = ~(L + 1)LALM (152)

If we apply the lowering operator to ALM we get
p
L− ALM = ~ L(L + 1) − M (M − 1)ALM −1 (153)
Which is the same as when we apply the lowering operator to a spherical har-
monic.
p
L− ALM = ~ L(L + 1) − M (M − 1)YLM −1 (154)
From this we see aLM is independent of M and we have this final expression:

AM M
L = aL YL (155)
Consider the forward direction. Set θ1 = θ2 = 0 or ψlm (0, ψ)
The sum will collapse to:

ΦLM (0, ψ; 0, ψ) = hl1 l2 00 | L0i Yl01 (0, ψ)Yl02 (0, ψ) (156)

r
0 0 2L + 1
ΦLM (0, ψ; 0, ψ) = AL (0, ψ) = aL Yl1 (0, ψ) = aL (157)
4π
s
(2l1 + 1)(2l2 + 1)
aL = hl1 l2 00 | L0i (158)
4π(2L + 1)
Set θ1 = θ2 = θ
1 X
YLM (θ, ψ) = hl1 l2 ; m1 m2 | LM i Ylm 1
(θ, ψ)Ylm 2
(θ, ψ) (159)
aL 1 2

This is known as the spherical harmonic addition relation.

Take the inner product:

Z
d2 ΩYlm 1
1
(Ω)Ylm
2
2
(Ω)Ylm
3
3
(Ω) = (160)
X Z
hl1 l2 ; m1 m2 | LM i aL Ylm 3
3
(Ω)YLM dΩ (161)

Using the relation: Ylm = (−1)m Yl−m . We see that only one term survives:
Z
d2 ΩYlm
1
1
(Ω)Ylm
2
2
(Ω)Ylm
3
3
(Ω) = (−1)m al3 hl1 l2 ; m1 m2 | l3 − m3 i (162)
Angular momentum 51

m1 + m2 = −m3 and |l1 − l2 | ≤ l3 ≤ l1 + l2

This relates to last semester when we studied the rotational eigenstates of the
ammonia molecule.
­ ¯ ¯ ®
l1 m1 ¯ Ylm
2
2 ¯
l3 m3 (163)
Where the spherical harmonic represents the electric dipole moment and the m
values represent the energy levels.

Wigner-Eckart Theorem
Scalars
What is the definition of an scalar? It is invariant under rotations or in other
word it is simply one number. An example of common scalar operators are ~r2
and p~2 .
~r · p~ + ~r · p~ (164)
Total angular momentum always generates rotation and we have which is con-
sists of Angular Momentum and Spin:

J~ = L
~ +S
~ (165)

To check to see if it is invariant under rotation we just need to check for an

infinitesimal roation. We have a convenient check at hand, by definition of a
QM scalar, if the commutator with J vanishes then we have a scalar. If the
commutator is zero then no matter how complicated the matrix is it means that
the quantity is scalar. h i
A, J~ = 0 (166)

Vectors
To see what is the definition of vector lets do the infinitesimal rotation again.
 
x
~ S,
~r =  y  , p~, L, ~ J~ (167)
z

If we do a very small roation around the Z-axis, then we would have for rotations
with θ << 1,:

x0 = x − θy
y0 = y + θx
z0 = z (168)
1
~r0 = ~r + θ [Lz , ~r]
i~

[Lz , x] = [xpy − ypx , x] = i~y (169)

52 UNIT 2: Angular momentum

[Lz , y] = [xpy − ypx , y] = −i~x (170)

[Lz , z] = [xpy − ypx , z] = 0 (171)

So in other words, the transformation of ~r is going to be the [~r, Lz ]:
 
−y
[Lz , ~r] = [xpy − ypx , ~r] = −i~  x  (172)
0
 
Vx
~ =  Vy 
V (173)
Vz
So if it is a vector the results of commuting with J should be like this.

[Ji , Vj ] = i~εijk Vk (174)

with 
1 : cyclic
εijk =  −1 : noncyclic (175)
0 : repeated
Through the cyclic permutations of the above commutator we are given 9 rela-
tions which define a vector. We know three and the other 6 can be found with
them. Lets switch back to scalars, saying A is a scalar and we want to know
the matrix elements of it. Eigenstates of A are given in the basis:

| jmi (176)

hj0m0 | A | jmi = ajm hj0m0 | jmi (177)

£ ¤
We know that: [A, Jz ] = 0 and A, J 2 = 0 and thus we see that A, Jz , J 2 form
a C.S.C.O. We know that if two operators commute so they share the same
common eigenstates.
A | jmi = aj | jmi (178)
The selection rules tell us a transition is only possible when hj0m0 | A | jmi 6= 0
only if j0 = j and m0 = m.

hj0m0 | [J+ , A] | jmi = 0 (179)

hj0m0 | J+ A | jmi = hj0m0 | AJ+ | jmi =

p
= ~ j0(j0 + 1) − m0(m0 − 1)ajm hj0m0 − 1 | jmi =
p
= ~ j(j + 1) − m(m − 1)ajm−1 hj0m0 | jm + 1i = (180)
Angular momentum 53

m = m0 − 1
ajm+1 = ajm = ajm−1 = ajm−2 = aj
j0 = j
When j is fixed then A is a matrix and we have 2j + 1 states and we can think
about A as (2j + 1) ∗ (2j + 1) matrix, | jmi, where m = −j · · · j. This matrix
is symmetric and it is identity.

A = aj II(2j+1)×(2j+1) (181)
B is also a scalar:
B | jmi = bj | jmi (182)
B = bj II
So we can see A ∝ B and it can be rewritten in the form of ⇒ A = λB
a
where λ = bjj .

Vectors
By using our definition of vectors. Instead of A here we have V which has three
separate components:

[Vz , Jz ] = 0 (183)

hj0m0 | [Vz , Jz ] | jmi = 0 (184)

hj0m0 | Vz Jz | jmi = hj0m0 | Jz Vz | jmi (185)

~mhj0m0 | Vz | jmi = ~m0hj0m0 | Vz | jmi (186)

So hj0m0 | Vz | jmi = 0 or m = m0 Lets try for other components.

V± = Vx ± iVy (187)

[Jz , V+ ] = ~V+ (188)

[Jz , V− ] = −~V− (189)

Lets see what we can do for Vx :

hj0m0 | [Jz , V± ] | jmi = ±~hj0m0 | V± | jmi

= ~mhj0m0 | V± | jmi − ~m0hj0m0 | V± | jmi (190)

So we can conclude that hj0m0 | V± | jmi = 0 unless m0 = m + 1. For V− we

have the same. So we have selection Rules which are:
54 UNIT 2: Angular momentum

Vz : 4m = 0
V± : 4m = ±1 (191)

Now if we fix j, what we can say about the matrices?

[J+ , V+ ] = [Jx + iJy , Vx + iVy ] = i(i~)Vz + i(−i~)Vz = 0 (192)

Now lets consider the matrix elements, we specialize it for simplicity of calcula-
tions:
hj0m0 | J+ V+ | jmi = hj0m0 | V+ J+ | jmi (193)

hjm + 2 | J+ V+ | jmi = hjm + 2 | V+ J+ | jmi (194)

with: X
I= | j0m0ihj0m0 | (195)
So we only can get something from that summation when m0 = m + 1 con-
tributes.

hjm + 2 | J+ | jm + 1ihjm + 1 | V+ | jmi

= hjm + 2 | V+ | jm + 1ihjm + 1 | J+ | jmi (196)

hjm + 1 | V+ | jmi hjm + 2 | V+ | jm + 1i

= = α+ (197)
hjm + 1 | J+ | jmi hjm + 2 | J+ | jm + 1i
Notice, that this is indenpendent of m; thus the ratio is a constant and that is
a very important conclusion. So we can say:

V+ = α+ J+
V− = α− J−
Vz = αJz (198)

We have to figure out all the three components.

[J− , V+ ] = i(i~)Vz − i(−i~)Vz = −2~Vz (199)

So:

hjm | (J− V+ − V+ J− ) | jmi = −2~hjm | Vz | jmi

p
= ~ j(j + 1) − m(m + 1)hjm + 1 | V+ | jmi
p
−~ j(j + 1) − m(m − 1)hjm | V+ | jm + 1i (200)

We know everything except for αpm so let us calculate them where:

Angular momentum 55

p
hjm + 1 | V+ | jmi = α+ J+ = α+ ~ j(j + 1) − m(m + 1) (201)
and
p
hjm | V+ | jm + 1i = α+ J+ = α+ ~ j(j + 1) − (m − 1)m) (202)

This gives

= α+ ~2 [j(j + 1) − m(m + 1) − j(j + 1) + m(m − 1)]

= −2mα+ ~2 = −2~Vz (203)

Now we can say:

hjm | Vz | jmi = m~α+ = m~α− (204)

α+ = α− = α

[J+ , V− ] = 2~Vz (205)

hjm | Jz | jmi = m~ (206)

Vz = αJz (207)
This gives us a generalized result and a very powerful statement:
~ = αJ~
V (208)

~ = β J~ and V
If we have: W ~ ∝W
~ , then when we fix j:

~ ∝ J~
Vk = V (209)
So if we think of those as vectors we can write:

~ ∝ J~2 = αj(j + 1)~2

J~ · V (210)
and calculate coefficient α:

hJ~ · V
~ i = αj(j + 1)~2 (211)

hJ~ · V
~i hJ~ · V
~i
α= = (212)
j(j + 1)~2 hJ~2 i

Vk = V cos(θ) (213)

J~ · V
~ = JV cos(θ) (214)
So in this geometrical view we can conclude that:
56 UNIT 2: Angular momentum

J~ · V
~
Vk = (215)
J

Vk J~ · V
~
α= = (216)
J J2
Recap from last class: Last time we spoke about vectors and found that for
~ is proportional to J~ and more importantly for a fixed j
an arbitrary vector: V
value we have:

~ = αJ~
V (217)

Today’s Notes:

The Hydrogen Atom

Remember that we found the energy levels for the hydrogen atom before

Eion
En = − (218)
n2
and we can denote the energy levels as |n, l, mi.
Say you have a complicated atom with a Hamiltonian like:

Ho = To + Vo + ... (219)

Also this complicated atom has total angular momentum L ~ and also S.
~ For all
~
the constituents, the total angular momentum, J, will be the sum of the spin
and angular momentum components:

J~ = L
~ +S
~ (220)
Which is very important becuase this can generate rotation. And we know that
~ = 0 even though we know nothing about the complicated atom. (Ho is
[Ho , J]
a scalar and invariant under rotation)

We also know: [Ho , L~2 ] = [Ho , S~2 ] = 0 and {Ho , J 2 , Jz , L2 , S 2 }

In general our energy level can be written as: |Eo , L, S, J, M i. If we fix J and
Eo , the energy level with have a fixed J and M = −J, ..., J with a degeneracy of
at least 2J + 1 and in the case of hydrogen atom it is even more becuase it does
not even depend on J. ~ To break down this degeneracy we will add a magnetic
field.
Angular momentum 57

Example 1:
~ = B ẑ.
Set B

H = Ho + H1 (221)
Where
q ~ ~ ~ · S]
~
H1 = − [(B · L) + 2B (222)
2m
Note: The coefficient 2 is due to relativistic effects.

H1 = ωL (Lz + 2Sz ) (223)

In which:
−qB
ωL = (224)
2m
and as we know it is the Larmor frequency. By fixing L,S,J we will try to figure
out what will happen to the energy level |Eo , L, S, J, M i
D E
~ · J~
L
~ =
L J~ (225)
J(J + 1)~2
D E
~ · J~
S
~=
S J~ (226)
J(J + 1)~2
~ Now we have to calculate the
and these are two operators we write based on J.
expectation values:

~ · J~ = L(
L ~ L ~ = L~2 + (L
~ + S) ~ · S)
~ (227)

~ · J~ = 1 (J~2 − L~2 − S~2 ) + L~2 = 1 (J~2 + L~2 − S~2 )

L (228)
2 2
Therefore,
1
~ = [ 2 (J(J + 1) + L(L + 1) − S(S + 1)] J~
L (229)
J(J + 1)
and
1
~ = [ 2 (J(J + 1) − L(L + 1) + S(S + 1)] J~
S (230)
J(J + 1)
ωL
H1 = [3J(J + 1) − L(L − 1) + S(S + 1)]Jz (231)
2J(J + 1)
By simplifying that we would have:

3 S(S + 1) − L(L + 1)
H1 = ωL [ + ]Jz (232)
2 2J(J + 1)
58 UNIT 2: Angular momentum

By acting this H1 on our eigenstate we have:

H1 |E0 , L, S, J, M i = ωL gJ ~M |E0 , L, S, J, M i (233)

Here gJ is the Lande factor.

Spin-Spin Coupling: 2 magnetic moments

Even if we do not have an external magnetic field; since each spin has a magnetic
moment so one magnetic moment will produce magnetic field and the other spin
can see that magnetic field so they can interact. For example in the simplest
case like a hydrogen atom we have an elctron with spin half and a proton with
spin half. So electron can see the magnetic field due to proton spin and it
generates very little splitting on electron’s energy levels which is very small.

Ho = E o I (234)

H = Ho + a(S~1 · S~2 ) (235)

The question is do we know the energy levels of this Hamiltonian? we define
the total spin as:

~ 2 = S~1 2 + S~2 2 + 2(S~1 · S~2 )

S (236)
By acting it on our eigenstate we would have:
3 3
S~2 |SM i = ( ~2 + ~2 + 2(S~1 · S~2 )) |SM i (237)
4 4

S~2 |SM i = S(S + 1)~2 |SM i (238)

S(S + 1) 3 2
(S~1 · S~2 ) |SM i = ( − )~ |SM i (239)
2 4

a~2 3
H |SM i = [Eo + (S(S + 1) − )] |SM i (240)
2 4
and this is the new energy level, the new eigenvalue.

The system can transit between these two energy levels and when it transits, it
emits a photon with frequency ~Ω which Ω is the Bohr frequency and as we can
see from the above Ω = a~.
~1 because
Here we want to understand the spin of electron and just the first spin S
usually one is going to do something and the other will not do anything. We
want to see how its expectation value evolves in time:
D E
~1
S (241)
t
Angular momentum 59

That would tell us what kind of transition to expect and what kind of radiation
we would observe. We have 4 states and we should have a 4 by 4 matrix:

|SM i = [|1, 1i , |1, 0i , |1, −1i , |0, 0i] (242)

S1z is the easist one to find so lets find it first:

 
1 0 0 0
~ 0 0 0 1
S1z |SM i =  
 (243)
2 0 0 −1 0
0 1 0 0

We need two more components so:

 
0 1 0 −1
~  0 0 1 0
S1+ |SM i = √ 

 (244)
2 0 0 0 0
0 0 1 0

To find the S1− we dont have to caculate it because we do know that:

T
S1− = S1+ (245)

So we can conclude that:

 
0 0 0 0
~  1 0 0 0
S1− |SM i = √   (246)
2  0 1 0 1
−1 0 0 0

As we can find out here:

~ = 1 J~
S (247)
2
Which is the Wigner-Eckart theorem.
What information can we extract from:

|ψ(0)i = α |0, 0i + β1 |1, 1i β0 |1, 0i + β−1 |1, −1i (248)

Which is the wavefunction that the system can be determined with.

Apply the evolution operator.

3a~2 3a~2
) ~t2 ) ~t2
|ψ(t)i = αe−i(E0 − 4 |0, 0i + e−i(E0 +
[β1 |1, 1i β0 |1, 0i + β−1 |1, −1i]
4

(249)
This is the state of the system at time t and we can write it as:
3a~2
) ~t2
|ψ(t)i = e−i(E0 − 4 [α |0, 0i + e−iΩt [β1 |1, 1i β0 |1, 0i + β−1 |1, −1i] (250)
60 UNIT 2: Angular momentum

~
hS1z i = √ [|β1 |2 + eiΩt αβ0∗ + e−iΩt α∗ β0 + |β−1 |2 ] (251)
2
~
hS1+ i = √ [β1∗ β0 + β0∗ β1 − eiΩt β1∗ α + eiΩt β−11 α∗ ] (252)
2
∗
hS1+ i = hS1− i (253)
Example: Consider the states |1, 0i → |0, 0i and here. Set β−1 = β1 = 0 and
suppose α is real so:
~
hS1z i = √ [eiΩt αβ0∗ + e−iΩt α∗ β0 ] (254)
2
Where:
β0 = |β0 |eiψ0 (255)

~
hS1z i = α|β0 |(ei(Ωt−ψ0 ) + e−i(Ωt−ψ0 ) ) (256)
2

hS1z i = ~α|β0 |cos(Ωt − ψ0 ) (257)

hS1+ i = hS1− i = 0, hS1x i = hS1y i = 0 (258)

So we can think of S as a vector and we have oscillation along the z-axis and
get linearly polarized radiation in the z-direction because of the oscillation.
Example B(for other transition): |1, 1i → |0, 0i

β−1 = β1 = 0 (259)

~2
hS1z i = |β1 |2 (260)
2

~2 ∗ iΩt
hS1+ i = − αβ1 e (261)
2

β1 = |β1 |eiψ0 (262)

~2
hS1x i = − α|β1 |cos(Ωt − ϕ1 ) (263)
2

~2
hS1y i = − α|β1 |sin(Ωt − ϕ1 ) (264)
2
Therefore it precesses in the x-y plane around the z-axis and radiate circularly
polarized light which is right handed (counterclockwise).
UNIT 3

Stationary perturbation
theory

Notes by L. Poudel, T. Papatheodore, and Y. Song

The Method
Perturbation theory applies to systems whose Hamiltonians may be expressed
in the form
H = H0 + W. (1)
H0 is called the unperturbed Hamiltonian and it is assumed to be time-independent.
We already know the solution corresponding to H0 , which is to say that we al-
ready know its eigenvalues and eigenstates.

H0 | E0,... i = E0 | E0,... i (2)

E0 is degenerate in general and the 0,... allows for the possibility of other quan-
tum numbers if there is degeneracy. W is called the “perturbation”, which
causes modifications to the energy levels and stationary states of the unper-
turbed Hamiltonian. W is assumed to be much smaller than H0 and for sta-
tionary perturbation theory it is also time-independent. In order to quantify
the “smallness” of W we assume that it is proportional to a real, dimensionless
parameter λ which is much smaller than 1:

W = λŴ , (3)

where λ ¿ 1 and Ŵ is an operator whose matrix elements are comparable to

those of H0 . Now, substitution of (3) into (1) yields

H = H0 + λŴ . (4)

From (4) we can see that as λ → 0, H(λ) → H0 , and we recover the unperturbed
Hamiltonian.
62 UNIT 3: Stationary perturbation theory

Approximate Solution of the Eigenvalue Equation

We want to find a solution to the eigenvalue equation
H(λ) | ψ(λ)i = E(λ) | ψ(λ)i (5)
In order to approximate the solutions E(λ) and | ψ(λ)i we assume that they
can be expanded in powers of λ:
E(λ) = E0 + λE1 + λ2 E2 + ... (6)
2
| ψ(λ)i = | ψ0 i + λ | ψ1 i + λ | ψ2 i + ... (7)
Substitution of (4), (6) and (7) into (5) yields
(H0 + λŴ )(| ψ0 i + λ | ψ1 i + λ2 | ψ2 i + ...) =
(8)
(E0 + λE1 + λ2 E2 + ...)(| ψ0 i + λ | ψ1 i + λ2 | ψ2 i + ...)
Multiply equation (8) out and collect like terms of λ. Now, since λ is arbitrary
we must equate the coefficients of successive powers of λ on both sides of the
equation.
0th Order Terms: λ0
H0 | ψ0 i = E0 | ψ0 i (9)
1st Order Terms: λ1
H0 | ψ1 i + Ŵ | ψ0 i = E0 | ψ1 i + E1 | ψ0 i (10)
2
2nd Order Terms: λ
H0 | ψ2 i + Ŵ | ψ1 i = E0 | ψ2 i + E1 | ψ1 i + E2 | ψ0 i (11)
Since equation (5) defines | ψ(λ)i only to within a constant factor, we can choose
its norm and phase. We assume | ψ(λ)i to be normalized and choose its phase
so that hψ0 | ψ(λ)i is real.
Now, since hψ(λ) | ψ(λ)i = 1 we obtain the following:
To 0th Order | ψ(λ)i = | ψ0 i
=⇒ hψ(λ) | ψ(λ)i = hψ0 | ψ0 i = 1 (12)
To 1st Order | ψ(λ)i = | ψ0 i + λ | ψ1 i
=⇒ hψ(λ) | ψ(λ)i = (hψ0 | +λhψ1 |)(| ψ0 i + λ | ψ1 i)
(13)
= hψ0 | ψ0 i + λhψ0 | ψ1 i + λhψ1 | ψ0 i + λ2 hψ1 | ψ1 i = 1
We can drop the last term involving λ2 since we are only concerned with 1st
order terms. Now we also know that hψ0 | ψ0 i = 1 which then implies that
hψ0 | ψ1 i + hψ1 | ψ0 i = 0. And, since hψ0 | ψ1 i is real we obtain
hψ0 | ψ1 i = hψ1 | ψ0 i = 0 (14)
2
To 2nd Order | ψ(λ)i = | ψ0 i + λ | ψ1 i + λ | ψ2 i
A similar argument for the 2nd order terms may be applied to obtain
1
hψ0 | ψ2 i = hψ2 | ψ0 i = − hψ1 | ψ1 i (15)
2
Stationary perturbation theory 63

Non-Degenerate Energy Level

We will now find the corrections to the energy levels and energy eigenstates of
a non-degenerate level. Project the 1st order equation (10) onto the state | ψ0 i:

hψ0 | H0 | ψ1 i + hψ0 | Ŵ | ψ0 i = hψ0 | E0 | ψ1 i + hψ0 | E1 | ψ0 i (16)

Now let H0 operate to the left on hψ0 | in the first term and pull the constants
out to obtain

E0 hψ0 | ψ1 i + hψ0 | Ŵ | ψ0 i = E0 hψ0 | ψ1 i + E1 hψ0 | ψ0 i (17)

From this equation we can see that since hψ0 | ψ0 i = 1 from (12) and the first
and third terms cancel we obtain the first order correction to the energy level:

E1 = hψ0 | Ŵ | ψ0 i (18)

Now we want to find the first order correction, | ψ1 i, to the eigenstate. Project
the first order equation (10) onto the state hE00 |, where hE00 | is an eigenstate
corresponding to any other energy level other than E0 . We obtain

E00 hE00 | ψ1 i + hE00 | Ŵ | ψ0 i = E0 hE00 | ψ1 i + E1 hE00 | ψ0 i (19)

The last term, hE00 | ψ0 i, is equal to zero since the two terms of the inner product
belong to different eigenvalues. Then we can solve for hE00 | ψ1 i to obtain

hE00 | Ŵ | ψ0 i
hE00 | ψ1 i = (20)
E0 − E00

Now expanding the state | ψ1 i on the | E00 i basis we conclude that the correction
to the energy eigenstate is
X
| ψ1 i = | E00 ihE00 | ψ1 i
E00
X hE00 | Ŵ | ψ0 i
= | E00 i + | E0 ihE0 | ψ1 i
E0 − E00 (21)
E00 6=E0
X hE00 | Ŵ | ψ0 i
= | E00 i ,
E0 − E00
E00 6=E0

where the last equation follows since hE0 | ψ1 i = 0 due to the fact that the two
states of the inner product belong to different eigenvalues.
Now we follow the same procedure for the 2nd order equation. In order to find
the 2nd order correction to the energy level we project the 2nd order equation
(11) onto the state | ψ0 i:

E0 hψ0 | ψ2 i + hψ0 | Ŵ | ψ1 i = E0 hψ0 | ψ2 i + E1 hψ0 | ψ1 i + E2 hψ0 | ψ0 i (22)

64 UNIT 3: Stationary perturbation theory

Now, the first terms on each side of the equation cancel, the second term on
the right side is zero since hψ0 | ψ1 i = 0, and hψ0 | ψ0 i = 1 which yields the
conclusion
E2 = hψ0 | Ŵ | ψ1 i (23)
Now, plugging in equation (21) for | ψ1 i gives the 2nd order correction to the
energy level:

X hE00 | Ŵ | ψ0 i X |hψ0 | Ŵ | E 0 i|2

0
E2 = hψ0 | Ŵ | E00 i 0 = 0 (24)
E0 − E0 E 0 − E 0
E00 6=E0 0
E0 6=E0

The second order correction to the energy level, | ψ2 i is to be found on your

own at home following the same procedure as in the first order correction.

Degenerate Energy Level

Now we allow for the energy E0 to be degenerate so that the corresponding
states are denoted | E0,a i, where a is a quantum number. If we suppose that
the degeneracy is 2, then a can take on two values: a = 1, 2.
We proceed in an analogous manner to the nondegenerate case except now we
must project the first order equation onto both possible states | E0,a i. Doing
so we see that

E0 hE0,a | ψ1 i + hE0,a | W | ψ0 i = E0 hE0,a | ψ1 i + E1 hE0,a | ψ0 i (25)

Notice that the first term on each side of the equation cancels leaving us with
two equations; one for each case a = 1, 2:

hE0,a | W | ψ0 i = E1 hE0,a | ψ0 i (26)

For a = 1
hE0,1 | W | ψ0 i = E1 hE0,1 | ψ0 i (27)
But generally | ψ0 i must be written as a linear combination of the states | E0,a i:

| ψ0 i = α1 | E0,1 i + α2 | E0,2 i (28)

Plugging equation (28) into (27) we obtain

α1 hE0,1 | W | E0,1 i + α2 hE0,1 | W | E0,2 i = E1 α1 (29)

For a = 2
hE0,2 | W | ψ0 i = E1 hE0,2 | ψ0 i (30)
Plugging equation (28) into (30) we obtain

α1 hE0,2 | W | E0,1 i + α2 hE0,2 | W | E0,2 i = E1 α2 (31)

Stationary perturbation theory 65

Now, the operator W may be written in matrix form in the | E0,a i basis as
· ¸
W11 W12
W21 W22

so that equations (29) and (31) may be written as the matrix equation
µ ¶ µ ¶
α1 α1
W = E1
α2 α2

The characteristic equation det(W − E1 I) = 0 may then be solved in order to

find the two eigenvalues and eigenstates.

Example: 1D Harmonic Oscillator

Here we can see the method in action by proceeding with an example that we
already know the answer to and then checking to see if our results match. The
Hamiltonian for the 1-D harmonic oscillator is given by

p2 1
H0 = + mω 2 x2 (32)
2m 2
Now, if the particle has a charge q we can turn on an electric field ~ε = εx̂ so
that we introduce a perturbation W = −qεx, and the total Hamiltonian then
becomes
p2 1
H = H0 + W = + mω 2 x2 − qεx (33)
2m 2
Recall that we have already solved this problem exactly in compliment FV where
we showed that
1 q 2 ε2
En0 = (n + )~ω − (34)
2 2mω 2
qε
ϕ0 (x) = ϕ(x − ) (35)
mω 2
We will now find the same results using perturbation theory.
1st order correction to the energy

En(1) = hn | W | ni = −qεhn | x | ni = 0 (36)

q
~
The last equality follows since x = 2mω (a† + a), and hence there is no shift
in the energy level to first order correction.
2nd order correction to the energy

X |hn0 | W | ni|2
En(2) = (37)
0
En − En0
n 6=n
66 UNIT 3: Stationary perturbation theory

r
0 0 ~
hn | W | ni = −qεhn | x | ni = −qε hn0 | a† + a | ni (38)
2mω
r
~ √ √
= −qε ( n + 1hn0 | n + 1i + nhn0 | n − 1i (39)
2mω
Therefore, the only nonzero contributions in equation (37) will come from n0 =
n + 1 and n0 = n − 1, yielding a second order correction to the energy term
~ n+1 n q 2 ε2
En(2) = q 2 ε2 (− + )=− (40)
2mω ~ω ~ω 2mω 2
Comparing equations (34) and (40) we notice that our second order correction
to the energy level matches identically with the exact solution.
We now find the correction to the energy state:
X hn0 | W | ni
| ψ1 i = | n0 i (41)
0
E n − En 0
n 6=n
r
~ 1 √ √
= −qε (− n + 1 | n + 1i + n | n − 1i) (42)
2mω ~ω
r
~ 1
= −qε (−a† + a) | ni (43)
2mω ~ω
iqε
=− p | ni (44)
m~ω 2
Equation (42) follows from (39), equation (43) follows from the definitions of
the creation and annihilation operators, and equation (44) follows from the
definition of momentum as p = i m~ω †
2 (a − a). We can now check to see if
equation (44) matches with the exact solution (35).
Taylor expand equation (35):
qε qε dϕ
ϕn (x − ) = ϕn (x) − + ... (45)
mω 2 mω 2 dx
d
Now, since p = −i~ dx , we can see that equation (44) can be written as
qε dϕ
ψ1 (x) = hx | ψ1 i = − , (46)
mω 2 dx
and we can verify that our first order correction to the eigenstate does match
the known exact solution.

Ex-II: Harmonic oscillator with quadratic poten-

tial
Consider a harmonic oscillator with hamiltonian H0 , which is given by
P2 1
H0 = + mw2 x2 (47)
2m 2
Stationary perturbation theory 67

Let us add a quadratic potential to above oscillator. So, the perturbation is

given by

1
W = λmw2 x2 (48)
2
We assume λ ¿ 1. In practice, this can be done by adding very weak spring
with small spring constant. Now, the hamiltonian becomes,

P2 1
H = H0 + W = + (λ + 1)mw2 x2 (49)
2m 2
The hamiltonian given by equation
√ (49) is also the equation of harmonic oscil-
lator with frequency ω 0 = ω 1 + λ, which can be solved exactly. So, the new
energy levels is given by

1
En0 = (n + )~ω 0 (50)
2
Where, p
ω0 = ω (1 + λ) (51)
Using taylor expansion, we get different order correction in energy level

1 λ λ2
En0 = (n + )~ω(1 + − + .............) (52)
2 2 8
But,this time we want to solve the problem using perturbation theory. First,
we solve problem in H0 , the solution of which is

H0 | ni = En | ni (53)
Where
1
En = (n + )~ω (54)
2
So, knowing that we can write expression for E’, which is general expression for
any perturbation
X |hn0 |W |ni|2
En0 = En + λhn|W |ni + λ2 + .... (55)
0
En − En0
n 6=n

The second term in equation (55) is average potential energy of harmonic oscil-
lator, and therefore is exactly half of total energy of harmonic oscillator without
perturbation. This is as expected from equation (52). We can also calculate it
by using creation and annihilation operator.
1
hn|W |ni = mω 2 hn|x2 |ni
2
1
= EN (56)
2
68 UNIT 3: Stationary perturbation theory

To calculate the third term let us first calculate hn0 |W |ni

1 ~
hn0 |W |ni = mω 2 hn0 |(a† + a)2 |ni
2 2mω
~ω 0 †2
= hn |a + a2 + a† a + aa† |ni
4
~ω p p
= [ (n + 1)(n + 2)δn0 ,n+2 + n(n − 1)δn0 ,n−2 + (2n + 1)δn0 ,n
4
(57)

The first terms contributes when n0 = n + 2, second term contributes when

n0 = n − 2. But third contributes only when n0 = n, which is excluded in our
relation
X |hn0 |W |ni|2 ~ω 1 1
0 − n)
= [− (n + 1)(n + 2) + n(n − 1)]
0
~ω(n 16 2 2
n 6=n
1 1
= − ~ω(n + ) (58)
8 2
We get exactly same term as expected from equation (52). Similarly, we can
calculate higher order terms.

EX-3: Harmonic oscillator with cubic potential

This time we add cubic potential to the same hamiltonian.

W = x3 (59)

In this case there is no way to solve the problem as we did previous example. So,
we must use perturbation theory to solve this problem. Here, the first term in
energy will be original energy without perturbation. The second term in energy
is also straightforward (i.e. zero since x3 is odd)

hn|x3 |ni = 0 (60)

The first correction in energy will be coefficient of λ2 . First we will evaluate

hn0 |W |ni. Most of the terms of hn0 |W |ni are zero. The non zero terms are,
(i) When n0 = n + 3
~ 3
hn + 3|W |ni = ( ) 2 hn + 3|(a† )3 |ni (61)
2mω

~ 3
|hn + 3|W |ni|2 = ( ) 2 (n + 1)(n + 2)(n + 3) (62)
2mω
(ii) When n0 = n − 3
~ 3
|hn + 3|W |ni|2 = ( ) 2 n(n − 1)(n − 2) (63)
2mω
Stationary perturbation theory 69

(iii) When n0 = n + 1

~ 3
|hn + 3|W |ni|2 = ( ) 2 9(n + 1)3 (64)
2mω
(iv)When n0 = n − 1

~ 3 3
|hn + 3|W |ni|2 = ( ) 2 9n (65)
2mω
Combining contributions from all these terms, we get

X |hn0 |W |ni|2 ~ 3 1 1 1
0 − n)
=( )2 [− (n+1)(n+2)(n+3)+ n(n−1)(n−2)−9(n+1)3 +9n3 ]
0
~ω(n 2mω ~ω 3 3
n 6=n
(66)
With little effort in algebra

1 ~2 1 7
=− [30(n + )2 ] + ] (67)
8 m3 ω 4 2 2
Finally, different between two successive energy level

0 15 λ2 ~2
En+1 − En0 = ~ω − (n + 1) (68)
2 m3 ω 4
In this case, separation in energy level is not equal but increases with increase in
value of n. In realistic system, we never have only quadratic potential. Actually,
if we expand potential around minimum, there will be higher order corrections.
So, these cubic terms comes to play.

Diatomic Molecule
Consider a molecule which consists of two atoms. Let us pretend that we don’t
know much about the molecule. First, we guess that when the atoms are very
far away the force between them must be attractive, otherwise no molecule could
be formed. However, at closer distances, they repel each other, since they are
restricted to be at finite distance from Heisenberg Uncertainty principle. Hence,
there must be an equilibrium distance, which is also the size of molecule. The
potential energy at this point is the dissociation energy of the molecule. Also,
we have bound states due to this potential.

Expanding potential(Figure 1) around minimum value of potential energy,

1 1
V (r) = V (r0 )+V 0 (r0 )(r−r0 )+ V 00 (r0 )(r−r0 )2 + V 000 (r0 )(r−r0 )3 +...... (69)
2 6
70 UNIT 3: Stationary perturbation theory

Figure 1: Potential of a diatomic moleculel

The higher order terms are less significant. V 0 (r0 ) will be zero, since potential
is minimum at this point. If we call r − r0 = x, then, the second order term
correpsonds to a harmonic oscillator and 61 V 000 (r0 ) is λ for a cubic potential in
our formulation. Assume the system is in ground state, which is quite common
at room temperature since thermal energy is much less than energy of harmonic
oscillator(KT ¿ ~ω) at room temperature. Now, we want to measure the pos-
sible transitions of the molecule by interacting with electromagnetic radiation.
The dipole moment D couples to electromagnetic waves,thus for two states |φi
and |ψi to be able to transition from one to another the condition hφ|D|ψi 6= 0
has to be satisfied.

When a system goes from first excited state |φ1 i to ground state |φ0 i so that
energy of system changes.

|φ1 i → |φ0 i + photon (70)

q
E1 −E0 V 00 (r0 )
the frequency of photon is given by ω1 = ~ ≈ m

This is what happens if we include only up till the second order term in the
potential energy in equation (69). In this case the states |φn i become the har-
monic oscillator states |ni. To be more accurate,we need to consider the third
order term of the potential energy in equation (69). So, the general expression
forthe ground state becomes

|φ0 i → |0i + λ|ψi + .... (71)

Stationary perturbation theory 71

where |ψi is the first order correction

X hn|W |0i|ni
|ψi = (72)
n~ω
n6=0

So, the ground state becomes,

h1|W |0i h3|W |0i

|φ0 i → |0i − λ |1i − λ |3i (73)
~ω 3~ω
The ground state of the entire Hamiltonian is not only |0i but |0i with some
higher order correction. We can think of higher states |φn i to be approximtely
states of the harmonic oscillator |ni. So, to this approximation, we can see there
can be transitions from |φ4 i,|φ2 i and |φ1 i to |φ0 i because the correpsonding
matrix elements hφn |D|φ0 i 6= 0. When transition takes place from |φ3 i → |φ0 i,
frequency of photon is given as by applying our results for the cubic potential
multiple times.

E30 − E00
ω3→0 = (74)
~
90 ~
= 3ω − λ2 3 4 (75)
2 m ω

Spin-Spin Interaction
Two spins interact because they correspond to magnetic moments which create
magnetic fields. Let us consider two spin half particle. Let us switch on magnetic
−
→
field B = B ẑ. So the Hamiltonian can be written as

H0 = ω1 S1z + ω2 S2z (76)

where ω1 and ω2 are given by equations

ω1 = −γ1 B0 (77)

ω2 = −γ2 B0 (78)
where γ is the gyromagnetic ratio. We know the eigenstates of Hamiltonian,
which are
~
H0 | + +i = (ω1 + ω2 )| + +i (79)
2

~
H0 | + −i = (ω1 − ω2 )| + −i (80)
2

~
H0 | − +i = (−ω1 + ω2 )| − +i (81)
2
72 UNIT 3: Stationary perturbation theory

~
H0 | − −i = − (ω1 + ω2 )| − −i (82)
2
we assume ω1 > ω2 . In this way, we get four different energy levels. If we
make transition from first energy level to second energy level, we get a photon
of frequency ω2 . Similarly, if we make transition between second and fourth en-
ergy level, we get photon of frequency ω1 . Also, transition between first to third
and third to fourth energy level gives photon of frequency ω1 and ω2 respectively.

We emphasize the transition with ω1 and ω2 since they are frequency they
appear in the evolution of hSx i. Now, I am going to switch on small magnetic
field B1 in x̂ direction. If We make this field oscillating with frequency ω then
we can keep tuning it to one of ω1 or ω2 to get resonance. To do this, make B1
time dependent B1 = B1 x̂cos(ωt). Then, we expect resonance when ω matches
with ω1 or ω2 .By tuning this B1 we can studying the transition of the system
via resonanaces.

Consider two spin S1 and S2 at distance R = Rn̂. The magnetic moment

corresponding to them will be µ1 = γ1 S1 and µ2 = γ2 S2 . Their interaction be-
tween these magnetic moment gives perturbation to hamiltonian which is given
by
−
→− → → −→
−
µ0 S1 .S2 − 3n̂.S1 n̂.S2 ]
W = γ1 γ2 [ ] (83)
4π R3
If we take ξ = − µ4πR
0 γ1 γ2
3

In spherical coordinates, n̂ can be written as

 
sinθcosφ
n̂ =  sinθsinφ 
cosθ
Finally,
1 1
W =ξ[−S1z S2z − S1+ S2− − S1− S2+
2 2
1 1
+ 3(S1z cosθ + e−iφ sinθS1+ + eiφ sinθS1− )
2 2
1 −iφ 1 iφ
(S2z cosθ + e sinθS2+ + e sinθS2− )]
2 2
Applying first order perturbation theory, we find the corrections to each of the
enrgy levels.
~ ~ ~2
h+ + |W | + +i = ξ[−( )2 + 3( cosθ)2 ] = ξ [−1 + 3cos2 θ] = ~Ω (84)
2 2 4
For perturbation theory to work, ~Ω ¿ ~ω1 , ~ω2 has to be satisfied. Similarly
we can determine the correction for the other states.
h− + |W | − +i = −~Ω = h+ − |W | + −i, h− − |W | − −i = ~Ω (85)
Stationary perturbation theory 73

Figure 2: Energy corrections(non-degenerate)

From this correction we can find that the transition from | + +i to | − +i now
has an energy of ~(ω1 + 2Ω), the transition from | + +i to | + −i has an energy
of ~(ω2 + 2Ω), transition from | + −i to | − −i now has an energy of ~(ω1 − 2Ω)
and transition from | − +i to | − −i now has an energy of ~(ω2 − 2Ω). We thus
observe that the original two peaks in the spectrum split into two sets of twin
peaks centered at ω1 and ω2 with the twin peaks separated by 4Ω(Figure 2 and
Figure 3).
So far we considered ω1 > ω2 , but as in some real materials such as gypsum(CaSO4 , 2H2 O),
ω1 = ω2 = ω. In such cases we have to employ degenerate perturbation theory
since the eigenvalues of H0 for | + −i and | − +i are now the same.
We have already found h+ − |W | + −i and h− + |W | − +i, it turns out the the
matrix elements h+ − |W | − +i and h− + W | − +i are also −~Ω. Thus in the
basis of | + −i and | − +i the perturbation can be written as
µ ¶
1 1
W = −~Ω (86)
1 1

Solving for the eigenvalues and eigenvectors of this system, we obtain in the
74 UNIT 3: Stationary perturbation theory

Figure 3: Split of resonance frequencies(non-degenerate)

Figure 4: Energy corrections(degenerate)

Stationary perturbation theory 75

Figure 5: Split of resonance frequencies(degenerate)

|SM i notation
µ ¶ µ ¶ µ ¶ µ ¶
1 1 1 1 1 1 1 1
|10i = √ +√ , |00i = √ −√ (87)
2 1 2 −1 2 1 2 −1

The eigenvalues of W in the | + −i and | − +i basis are 0 and −2~Ω correspond-

ing to the states |00i and |10i in the |SM i notation. The other 2 nondegenerate
states | + +i and | − −i corresponds to |11i and |1 − 1i in the |SM i notation.
Transitions only occur between states with the same L. The frequency of transi-
tion after correction from |11i to |10i is ω +3~Ω and the frequency for transition
from |10i to |1 − 1i becomes ω − 3~Ω. We can thus observe twin peaks centered
at ω that are separated by 6Ω(Figure 4 and 5).

Volume Effect
In a hydrogen atom, we typically assume to have a potential of the form V (r) =
2
− er , but with such an assumption we are treating the proton as a point which
it really isn’t. Thus this volume effect should show up in the spectrum of the
atom. Assume the atom has a radius a0 which is the Bohr radius and the proton
has a radius of r0 , satisfying r0 ¿ a0 . Also assume the charge of the proton is
uniformly distributed in the volume of the sphere, we ca find the potential for
76 UNIT 3: Stationary perturbation theory

both inside and outside the proton

( 2
− er , r > r0
V (r) = 3e2 e2 2 (88)
r0 − 2r03
r , r < r0

Contrasting this potential with the original potential for which we known the
energy levels, we obtain that our perturbation for this problem
½
0 , r > r0
W = e2 (89)
r 2
2r0 [( r0 ) + 2 rr0 − 3] , r < r0

With this correction the Hamiltonian of the hydrogen atom can be written as

H = H0 + W (90)

H0 is the hamiltonian of the hydrogen atom treating the proton as fixed and as
a point with H0 |nlmi = En |nlmi and En = − Enion 2 . Where Eion is the ionizing

engergy. The engery levels |nlmi are degenerate which means we need to employ
degenerate perturbation theory to solve this problem.
Consider the subspace for a fixed n and W in that subspace. The matrix
elements in this basis are
Z
hnlm|W |nl0 m0 i = d3 r φ∗nlm (→
−r )W (r)φnl0 m0 (−
→
r)
Z Z
∗
= dΩYlm Yl0 m0 dr r2 Rnl∗
(r)W (r)Rnl0 (r)
Z
= δll0 δmm0 dr r2 W (r)|Rnl (r)|2 (91)

Thus W in this basis has only diagonal elements which simplifies everything.
Also, W is non-zero only within r0 thus within r0 we can approximate Rnl (r)
by Rnl (0). Thus the matrix elements are
Z
hnlm|W |nlmi = dr r2 W (r)|Rnl (r)|2
Z r0
2
≈ |Rnl (0)| dr r2 W (r)
0
e2 r02
= |Rnl (0)|2 (92)
10
Thus we find the corrected energies are given by

0 e2 r02
Enl = En + |Rnl (0)|2 + ... (93)
10
Note that |Rnl (0)|2 is non-zero only when l = 0, thus only s-waves are affected
by this effect(within this approximation) which can be verified by observation.
Stationary perturbation theory 77

Variational Method
Suppose we are given a certain hamiltonian
H|ψi = E|ψi (94)
The expectation of the hamiltonian is given by
hψ|H|ψi
hHi = (95)
hψ|ψi
Now we introduce a perturbation to |ψi,
|ψi → |ψi + λ|δψi (96)
Now hHi becomes, to the first order of λ,
hHi + λ∗ hδψ|H|ψi + λhψ|H|δψi + ... N
hHi → = (97)
hψ|ψi + λ∗ hδψ|ψi + λhψ|δψi + ... D
Expand D−1 to first order of λ we obtain
hδψ|ψi hψ|δψi
D−1 = hψ|ψi−1 [1 − λ∗ −λ + ...] (98)
hψ|ψi hψ|ψi
Plug this into the expression for hHi we obtain
N 1
hHi → = hHi + {λhψ|H|δψi − λhHihψ|δψi + C.C.} + O(λ2 ) (99)
D hψ|ψi
When the original wavefunction |ψi is a eigenfunction of the Hamiltonian,
H|ψi = E|ψi, the terms in curly brackets in the above equation becomes
λEhψ|δψi − λEhψ|δψi + C.C. = 0 (100)
Thus in this case after introducing |δψi to |ψi, hHi → hHi + O(λ2 ), which
means any perturbation of first order to an eigenstate of the Hamiltonian will
not change the expectation of the Hamiltonian to the first order.
Conversely, if hHi is not changed by any |δψi to first order, we can choose
|δψi = λ∗ (H − hHi)|ψi (101)
With this choice we plug it into the terms in culry brackets in Equation (99)
0 = λhψ|H|δψi − λhHihψ|δψi + C.C. (102)
= |λ|2 hψ|(H − hHi)(H − hHi)|ψi + C.C. (103)
= 2hδψ|δψi (104)
From this result and the choice of |δψi we find that H|ψi = hHi|ψi which means
|ψi is an eigenvector. We see that the converse of our previous argument is also
true.
78 UNIT 3: Stationary perturbation theory

Example: Harmonic Oscillator

Consider a Harmonic oscillator, the hamiltonian of which is given by

P2 1
H= + mω 2 x2 (105)
2m 2
We want to figure out the ground state. We first want to guess the wave func-
2
tion. For the ground state, a good guess will be ψ = e−αx , where α is an
arbitrary parameter. Now, we need to calculate H

First, the denominator

Z ∞
2
hψ|ψi = dx e−2αx (106)
−∞

The numerator, i. e. average of hamiltonian is given by

Z ∞
~2 0 2 1
hψ|H|ψi = dx [ (ψ ) + mω 2 x2 ψ 2 ]
−∞ 2m 2
Z ∞
~2 2 1 2
= (2αx)2 e−2αx ) + mω 2 x2 e−2αx ]
dx [
−∞ 2m 2
2 2 Z ∞
2~ α 1 2
= ( + mω 2 ) dx x2 e−2αx
m 2 −∞
2 2 2 Z ∞
2~ α 1 2 x e−2αx ∞ 1 2
= ( + mω )[(−) |−∞ + dx e−2αx ](107)
m 2 4α 4α −∞

The first term goes to zero when x → ∞ and only the second remains. Also, the
integral inside the second term is exactly equal to hψ|ψi and get cancels with
the denominator. So

2~2 α2 1 1
hHi = [ + mω 2 ]
m 2 4α
~2 mω 2
= α+ (108)
2m 8α
I have free parameter α which we haven’t specified. So we find the extremum
of hHi with respect to α. i.e,

dhHi ~2 mω 2
= − =0 (109)
dα 2m 8α2
This equation fixes α, which is given by
mω
α= (110)
2~
Stationary perturbation theory 79

For this α,
~2 mω 1 2~
hHi = + mω 2
2m 2~ 8 mω
~ ~
= +
4 4
~ω
= (111)
2
Which is precisely the ground state energy. Now, we see ψ, which is the ground
state wave function, is given by
mω 2
ψ = e− 2~ x (112)
This is exactly the same wave function that we obtain by solving the Schrodinger
equation.

Example II: Harmonic oscillator

Let us guess ψ which has peak at x → 0 and goes to zero smoothly at x → ∞,
1
as ground state of harmonic oscillator. For that we choose ψ = (x2 +α2 ) . In the

same way as before, we calculate the denominator.

Z ∞
1
hψ|ψi = dx 2 (113)
−∞ (x + α2 )2
substitute x = α tanθ, we get,
Z π2
2
hψ|ψi = dθ cos2 θ
α3 − π2
π
= (114)
2α3
Now, the numerator is
Z ∞ 2
~2 1 0 1 1
hψ|H|ψi = dx [ ( 2 2
) + mω 2 x2 2 ] (115)
−∞ 2m (x + α ) 2 (x + α2 )2
making substitution x = tan θ, we get,
π~2 πmω 2
hψ|H|ψi = + (116)
8mα5 4α
Therefore,
hψ|H|ψi ~2 mω 2 α2
hHi = = + (117)
hψ|ψi 4mα2 2
Taking dhHi
dα = 0, we obtain

~2
− + mω 2 α = 0 (118)
2mα3
80 UNIT 3: Stationary perturbation theory

Figure 6: H2+ ion

or,

~2
α4 = (119)
2m2 ω 2
Looking at energy

~ω
hHi = √
2
≈ 0.7 ~ω (120)

But, the correct answer is 0.5 ~ω

Chemical Bond of H2+ Ion

In this section we will study the “real life” example of the chemical bond in a
H2+ ion which consists of 2 protons which share an electron.
The geometry of the situation is shown in Figure 6 where r~1 is the distance from
proton 1 to the electron, r~2 is the distance from proton 2 to the electron, and
~ is the distance from proton 1 to proton 2. The Hamiltonian for the system is
R
given by
p2 e2 e2 e2
H= − − + (121)
2µ r1 r2 R
where the last three terms in the Hamiltonian are the potential energies due
to the particle interactions. Now consider the system when the electron is very
close to proton 1. Referring to Figure 6 we see that r2 ≈ R which implies that
Stationary perturbation theory 81

the last two terms in the Hamiltonian cancel leaving us with

p2 e2
H≈ − (122)
2µ r1

Recent work with the hydrogen atom allows us to recognize ground state solu-
tions of the form
−r1
1 −r1
ϕ(r~1 ) = Ae a0 = p 3 e a0 (123)
πa0
where the constant A was determined by normalization. We can make the same
argument for the electron being very close to proton 2 which yields a solution
1 −r2
ϕ(r~2 ) = p 3
e a
0 (124)
πa0

The Variational Approach

Because the electron moves around in the real situation we cannot use these
specific solutions to generally describe the motion. However, we may use a
linear combination of the two solutions as our guess at the variational method

ψ = c1 ϕ(r~1 ) + c2 ϕ(r~2 ) (125)

So we want to solve the equation

Hψ = Eψ (126)

in order to find the eigenvalues and eigenfunctions. We use our guess and write
the eigenvalue equation in the form

H | ψi = E | ψi (127)

Projecting Equation (125) onto the state | 1i and using the expansion

| ψi = c1 | 1i + c2 | 2i (128)

we obtain

c1 h1 | H | 1i + c2 h1 | H | 2i = Ec1 h1 | 1i + Ec2 h1 | 2i (129)

Projecting Equation (125) onto the state | 2i and using Equation (126) we
obtain
c1 h2 | H | 1i + c2 h2 | H | 2i = Ec1 h2 | 1i + Ec2 h2 | 2i (130)
These two equations may be written in matrix form as
µ ¶µ ¶ µ ¶µ ¶
H11 H12 c1 S11 S12 c1
=E (131)
H21 H22 c2 S21 S22 c2
82 UNIT 3: Stationary perturbation theory

where

hi | H | ji = Hij (132)
hi | ji = Sij (133)

and i,j = 1,2. Our job now is to analyze the equation

det(H − ES) = 0 (134)

We begin by constructing the matrix elements of H and S from Equation (131)

For the matrix S we have
S11 = h1 | 1i (135)

S12 = h1 | 2i (136)

S21 = h2 | 1i (137)

S22 = h2 | 2i (138)
We can simplify this by noting that ϕ1 and ϕ2 are both normalized so that
S11 = S22 = 1. We also notice that since our two states are real S12 = S21 = s.
Then we see that
Z Z
s = h1 | 2i = h2 | 1i = d3 r1 ϕ1 (r1 )ϕ2 (r2 ) = d3 r1 ϕ1 (r1 )ϕ2 (r1 − R) (139)

where the last equality comes from the relation between positions r2 = r1 − R.
We obtain
Z ∞ Z π Z 2π
1 −r1 −|r1 −R|
s= 3 r12 cosθdr1 dθdφe a0 e a0 (140)
πa0 0 0 0
−R R 1 R2
= e a0 [1 + + ] (141)
a0 3 a20

For the matrix H we have

p2 e2 e2 e2
H11 = h1 | H | 1i = h1 | − − + | 1i (142)
2µ r1 r2 R

The first two terms are simply the Hamiltonian corresponding to the hydrogen
atom with r = r1 where H 0 | 1i = E 0 | 1i. Since this is the familiar ionization
energy we may rewrite the equation as

e2 e2
H11 = −Eion − h1 | | 1i + (143)
r2 R
Stationary perturbation theory 83

e2
Now we determine h1 | r2 | 1i as follows
Z Z −2r1
e2 e2 1 e2
a0
h1 | | 1i = d3 r1 ϕ2 (r1 ) = d3 r1 e (144)
r2 | r1 − R | πa30 | r1 − R |
2a0 −2R R
= Eion [1 − e a0 (1 + )] (145)
R a0
The matrix element H12 is found in a similar manner to be
−R R
H12 = Eion 2e a0 (1 + ) (146)
a0
But H11 = H22 since it does not matter whether you integrate over r1 or r2 and
H12 = H21 since H is Hermitian so we have constructed the two matrices S and
H.

Solutions:
Now we find the determinant to obtain the eigenvalues.

det(H − ES) = (H11 − E)2 − (H12 − E)2 = 0 (147)

which yields the two solutions
H11 + H12
E+ = (148)
1+S

H11 − H12
E− = (149)
1−S
If the protons fly away from each other (R → ∞) what happens to these ener-
gies? Well, since both are dependent only on H11 , H12 , and S we can look at
the behavior of these quantities as R → ∞ to find out. Equations (143) and
(145) imply that
lim H11 → −Eion (150)
R→∞

and equations (138) and (144) show that

lim S → 0 (151)
R→∞

lim H12 → 0 (152)

R→∞

Referring back to Equations (148) and (148) we see that

lim E+ → −Eion (153)
R→∞

and
lim E− → −Eion (154)
R→∞
So that the equations for E+ and E− come together asymptotically as shown
in Figure 7.
84 UNIT 3: Stationary perturbation theory

Figure 7: E+ and E− converges to −Eion

UNIT 4

Fine and hyperfine

structure of the hydrogen
atom

Notes by N. Sirica and R. Van Wesep

Previously, we solved the time-independent Schödinger equation for the Hydro-
gen atom, described by the Hamiltonian

p2 e2
H= − (1)
2m r
Where e is the electron’s charge in your favorite units. However, this is not
really the Hamiltonian for the Hydrogen atom. It is non-relativistic and it does
not contain spin. In order to completely describe the Hydrogen we would need
to use the Dirac equation. We will not introduce that equation here, but we
will say a few words about the most important energy level of the relativistic
Hydrogen atom, namely the rest mass energy.

E = mc2 ≈ 0.5MeV (2)

We can compare this to the ground state (ionization) energy we found for the
Hamiltonian in 1.

E = −Eion ≈ 13.6eV ¿ mc2 (3)

Even though the rest mass energy is so much larger, it appears constant in
the non-relativistic regime. Since differences in energy are important, we could
ignore it before. The rest mass energy may be large, but it does not enter the
world of everyday experience. Regardless, it is fruitful to investigate the relative
size of the ionization energy to the rest mass energy.

Eion me4 1 e4 α2
= = = (4)
mc2 2~2 mc2 2c2 ~2 2
86 UNIT 4: Fine and hyperfine structure of the hydrogen atom

2
e 1
Where α = ~c = 137 is a fundamental constant of nature. No one understands
it, but it is important that it is small. This means that calculations, such
as those in perturbation theory, are possible and gives the orders successive
corrections. Let us move to these corrections in the case of the Hydrogen atom.

Relativistic Energy Correction

We begin with the relativistically correct expression for the energy.

r
p p2
2
E= p2 c2 + m2 c4 = mc 1 + 2 2
m c
µ 2 4
¶
2 p p
= mc 1 + − + ...
2m2 c2 8m4 c4
p2 p4
= mc2 + − + ... (5)
2m 8m3 c4
The second term is the non-relativistic kinetic energy that we are familiar with in
everyday life. The third term is the first relativistic correction to the energy. We
shall define this as the first perturbation to the Hamiltonian W1 = −p4 /8m3 c4 .
Let us investigate the relative strength of this perturbation. First we recognize
p2 p4 2
that 2m ∼ Eion so m 2 ∼ Eion . Thus,

2
Eion
W1 ∼ (6)
mc2
W1 Eion
∼ ∼ α2 (7)
Eion mc2
So once again we see the the fine structure constant. We expect to see a cor-
rection in the spectrum that is ∼ α2 . This number is small enough that we can
use perturbation theory.

Spin-Orbit Coupling
From relativity, a particle moving in an electric field feels a small magnetic field
in its reference frame. For a particle with velocity ~v = p~/m moving in an electric
field E~ this magnetic field is given by:

~ = − 1 ~v × E
B ~ + O(v 2 /c2 ) (8)
c2
In the Hydrogen we have an electric field due to the Coulomb potential.

~ = ∇ e = −e ~r
E (9)
r r3
We shall define our next perturbation as the energy of a dipole moment in a
magnetic field.
Fine and hyperfine structure of the hydrogen atom 87

W2 = −~ ~
µ·B (10)
In this case the magnetic dipole moment is related to the electron spin.
e ~
µ
~= S (11)
m
Putting this, the magnetic field given in 8 and the electric field given in 9 into
10 we obtain

e ~ ³ ~
´
W2 = S · p
~ × E
m2 c2 µ ¶
e2 ~ ~r
= − 2 2S · p~ × 3 (12)
m c r

~ = ~r × p~ we can write the spin-orbit

Since the angular momentum is given by L
term as

e2 1 ~ ~
W2 = S·L (13)
m2 c2 r3
Unfortunately, this answer is wrong. If we had performed a completely rela-
tivistic calculation we would have found the correct expression, one describing
Thomas precession.

e2 1 ~ ~
W2 = S·L (14)
2m2 c2 r3
As we did for the previous correction, let us look at the relative order of the
spin-orbit coupling energy. Recognizing that r ∼ a0 = ~2 /me2 , S~ ∼ ~, L~ ∼~
we find

W2 e2 ~2 m3 e6 ~2
∼ 2 2 6
Eion m c ~ me4
e4
∼ 2 2 = α2 (15)
c ~
Once again, the order of this correction is related to the fine structure constant.

Darwin Term
The next correction is a bit more subtle than the previous two. It has to
do with the relativistic effects that arise from the Schrödinger equation being
an approximation of the Dirac equation. The order of these relative effects is
governed by the ratio v/c = p/mc which defines an upper limit to the electron
momentum mc. This ratio is often written λ/λ0 where the deBroglie wavelength
is given by p = h/λ and the Compton wavelength is given by mc = h/λ0 . Since
the Compton wavelength is the minimum wavelength that the electron can have,
88 UNIT 4: Fine and hyperfine structure of the hydrogen atom

measurements of the position of the electron can only be accurate up to ∼ λ0 .

So, rather than being able to exactly determine the electron position, we can
only determine a fuzzy ball about the size of λ0 . This is illustrated in Figure 1.

Figure 1: Setup of the Darwin correction

In order to quantify this correction we shall define an effective potential that is

the average of the potential over the fuzzy ball.
Z
1
Vef f = V (~r + ~²) d3 ² (16)
Volume
The effective potential is almost the Coulomb potential because the sphere we
shall use for the integration is very small: |~²| ≤ λ0 . We shall perform the
¡ ~ ¢3
integration over a sphere of Volume = 34 π mc where we have used ~ instead
of h in the Compton wavelength for convenience. Since ~² is small we shall Taylor
expand:

X 3
1 ∂ ∂
V (~r + ~²) = V (~r) + ~² · ∇V + ²i ²j V + ··· (17)
i,j=1
2 ∂xi ∂xj

Let us perform the integration term by term for clarity.

Z
V (~r) d3 ² = V (~r) × Volume (18)
Z Z
3
~² · ∇V d ² = ∇V · ~²d3 ² = 0 (19)

Since there is no preferred direction and we are integrating over all directions.
This is also the reason why
Fine and hyperfine structure of the hydrogen atom 89

Z
²i ²j d3 ² = Aδij (20)

Where A is an undetermined constant. In order to find A we shall perform the

integration

Z
¡ 2 ¢
²x + ²2y + ²2z d3 ² = 3A
Z λ0
4π 5
4π ²4 d² = λ
0 5 0
4π 5
A= λ (21)
15 0

Putting 20 and 21 into the average of the third term in 17 we obtain

Z
3m3 c3 ∂ ∂ 1 4π 5 3m3 c3 2
²i ²j V d3 ² = λ ∇ V
4π~3 ∂xi ∂xj 2 15 0 4π~3
µ ¶2
1 ~
= ∇2 V (22)
10 mc

So in the end we have obtained an expression for the effective potential.

µ ¶2
1 ~
Vef f = V (~r) + ∇2 V + · · · (23)
10 mc
Unfortunately, once again we are slightly off. A full relativistic calculation gives
us our third perturbation to the Hamiltonian in 1.
µ ¶2
1 ~
W3 = ∇2 V (24)
8 mc
Let us once again examine the relative strength of this term for the Hydrogen
atom. In this case

e2
∇2 V = ∇2 = −4πe2 δ 3 (~r) (25)
r
In order to compare this W3 to the ionization energy, we must specify the state
and take an expectation value. Let us use the ground state.

1
ϕ= p e−r/a0 (26)
πa30
Then the order of the expectation value of W3 can be obtained.
90 UNIT 4: Fine and hyperfine structure of the hydrogen atom

Z
~2 e2
hW3 i ∼ 2 2 3 e−r/a0 δ 3 (~r) d3 r
m c πa0
~2 e8 m3
∼ 2 2 6 (27)
m c ~
From this we can find the relative order of the Darwin correction.

hW3 i e8 m ~ 2 e4
∼ 2 4 4
= 2 2 = α2 (28)
Eion c ~ me ~ c
Not surprisingly the correction is again ∼ α2 .

Fine Structure: n = 2 Level

The n = 2 level of the Hydrogen is the first level that will exhibit significant
changes, so we shall examine the effects of the above perturbations in this level.
The n = 2 level admits angular momentum numbers l = 0, 1 These momentum
states are termed 2s and 2p respectively. When we consider the spin degree of
freedom the degeneracy of the 2s and 2p states is 2 and 8 respectively. Luckily,
each term in the perturbation W = W1 + W2 + W3 is£rotationally ¤ invariant
(they only involve the operators L,~ p2 , r and S),~ so W, L2 = 0. Hence,
the perturbation will not mix the 2s and 2p subsets and we can treat them
separately.

2s State
In the 2s state the electron eigenfunction is given by h~r|200i = ϕ (~r) = R20 (r)Y00 (Ω)
where
µ ¶
2 r
R20 = 3/2
1− e−r/2a0 (29)
(2a0 ) 2a0

In addition to the orbital wave equation, we much include the spin degree of
freedom, so our eigenstates are |200±i. In order to find the first order correction
to the n = 2 energy level, we must find the matrix elements of W between these
eigenstates.

1
δE1 = h200 ± |W1 |200±i = − h200|p4 |200ih±|±i (30)
8m3 c2
Since h±|±i yields the elements of the identity matrix, we only have to worry
about the inner product h200|p4 |200i. We can rewrite W1 into a more convenient
form.
µ ¶
p2 e2 e2
H0 = − =⇒ p2 = 2m H0 + (31)
2m r r
Fine and hyperfine structure of the hydrogen atom 91

µ ¶2 µ ¶2
1 2 e2 1 e2
W1 = − 4m H 0 + = − H 0 + (32)
8m3 c2 r 2mc2 r
Using this form and recognizing that |200i is an eigenstate of H0 makes taking
the necessary inner product easier.

µ ¶2
1 e2
δE1 = − h200| H 0 + |200i
2mc2 r
µ ¿ 2 À ¿ 4 À¶
1 2 e e
=− 2
E 2 + 2E 2 +
2mc r r2
13
=− mc2 α4 (33)
128
Since L~ only has to do with the angles and ϕ200 has no angular dependence it
~
is easy to see that h200|L|200i = 0. This means that
Since ∇2 1r = −4πδ 3 (~r) the change in energy level due to the Darwin correction
is simple to compute.

~2 ¯ ¯2
2¯ ~0)¯¯
δE3 = hW3 i = 4πe ¯ ϕ 200 (
8m2 c2
~2 e2 1
= 2 2 3 = mc2 α4 (34)
16m c a0 16
Combining 33 and 34, we obtain the first order energy correction to the 2s state.
5
δE = hW i = − mc2 α4 (35)
128
This is a precise number that we can compare to experiment.

Fine Structure of the 2p State

Last time, using the framework of time independent perturbation theory, we had
discussed the three corrective terms to the hydrogen atom Hamiltonian. Using
this, a first order energy correction to the |200i state was then constructed.
However, if instead of considering this non-degenerate first excited state, we
turn our attention to the degenerate 2p subshell. Computing first the matrix
elements due to the relativistic correction of the Hamiltonian.

P4
h21ml |W1 |21ml0 i = h21ml | − |21ml0 i (36)
8m3 c2
Writing P 4 in terms of the unperturbed Hamiltonian of the hydrogen atom,
P2 2
H0 = 2m − er , one then notices for |21ml i being an eigenstate of H0 with
eigenvalue E2 that the matrix elements may be written separately in terms of
their angular and radial components. In doing so,
92 UNIT 4: Fine and hyperfine structure of the hydrogen atom

1 e2
− 2
h21ml |(H0 + )2 |21ml0 i (37)
2mc r
Z ∞
1 e2 2
= h1ml |1ml0 i drr2 |R21 (r)|2 [− (E2 + ) ] (38)
0 2mc2 r
r
where as we know R21 = √3(2a a0
3/2
exp(− 2ar 0 ). To evaluate this integral, begin
0)
first with considering each term independently, namely,

Z ∞
1 2
− δ 1,1 δm l , m l 0 [E 2 drr2 |R21 (r)|2
2mc2 0
Z ∞ Z ∞
2
+ 2E2 e drr|R21 (r)| + e4 2
dr|R21 (r)|2 ] (39)
0 0

After computing these integrals, the diagonal matrix elements of W1 are given
by

1 2 e2 e4
− δ m ,m [E2 + 2E2 + ]
12a20
0
2mc2 l l
4a0
7
=− δm ,m mc2 α4 (40)
384 l l0
To account then for the additional corrective terms, let’s consider first the
~2 2
Darwin term, W3 = 10m 2 c2 ∇ V . Now, in the case of a Coulomb potential,
2 e2 2
∇ V = −∇ · ∇ r = e δ(r). Thus in evaluating the matrix element, it follows
Z
h21ml |W3 |21ml0 i ∝ drr2 |R21 |2 δ(r) = 0 (41)

Meaning the Darwin term makes no contribution to the fine structure of the
2p state. Thus, in lastly accounting for spin orbit coupling, we consider from
the basis common to the C.S.C.O. H0 , L2 , S 2 , Lz , SZ the eigenstate |21ml ms i.
Here, just as with W1 , we may separate the angular from the radial part in
computing the matrix elements.

h21ml ms |W2 |21ml0 ms0 i

e2 1
= h21ml ms | 3 L · S|21ml0 ms0 i (42)
2m2 c2 r
Z ∞
e2 1
= 2 2
dr |R21 (r)|2 h1ml ms |L · S|1ml0 ms0 i
2m c 0 r

Where the integral then solves to 3(2a10 )3 . To evaluate the angular expression, it
is best to perform a change of basis to a common eigenstate of H0 , J 2 , Jz , L2 , Lz
where J of course denotes the total angular momentum J = L + S. In working
Fine and hyperfine structure of the hydrogen atom 93

then in the |Jmj i basis one defines through the triangle rule, |l − s| ≤ J ≤ l + s,
j = 3/2, with a degeneracy of 4 and j = 1/2 with a degeneracy 2. The benefit
of this change of basis stems from the fact that hL · Si may be expressed as
h 12 (J 2 − L2 − S 2 )i. Hence, for l = 1 and s = 1/2 the expectation value of
2
hL · Si becomes ~2 (j(j + 1) − 2 − 34 ). Thus the first order energy correction due
to spin-orbit coupling within the 2p state is given as

~2 e2 11
hW2 i = (j(j + 1) − )
96m2 c2 a30 4
1 1
= − mc2 α4 if j = (43)
48 2
1 2 4 3
= mc α if j = (44)
96 4
2
e
where α denotes the fine structure constant ~c . From these two expressions we
see the degeneracy of the 2p state is partially lifted. Namely, the six eigenstates
of the |21ml ms i are separated into the four upper |2 34 mj i and two lower |2 12 mj i
states.

The Zeeman effect

As seen on a number of occasions, perhaps one of the most apparent ways
in which one may remove the degeneracy of a state is through switching on
a magnetic field, say for instance, B ~ = B ẑ. By modelling an electron as an
ideal magnetic dipole, this field will exert a torque on it. Associated with this
torque is an energy given then by H = −~ ~ As previously found, the magnetic
µ · B.
dipole of the electron is given by the sum of the dipole due to the orbital angular
momentum and intrinsic spin. Provided B ~ = B ẑ to be small, we may treat it
as a perturbation defined by

Wz = −~ ~ = ωL (Lz + 2Sz )
µ·B (45)

with ωL denoting the Larmor frequency. In considering once again the 2s state,
by acting with Wz ,

Wz |200±i = ±~ωL |200±i (46)

Hence, for the doubly degenerate eigenstate of the Hamiltonian H0 + Wf s the
inclusion of Wz serves to lift the degeneracy, separating these states by ~ωL .
Having sufficiently warmed up now, let us move on now to the more complicated
2p state. As one may recall, the degeneracy of the six 2p states was partially
lifted through the effect of spin-orbit coupling. However, there is a problem with
the operation of Wz on the 2p states, as such states were diagonalized within
an eigenstate common to H0 , L2 , S 2 , J 2 , Jz not Lz and Sz . Thus, in considering
the fact that Wf s was diagonalized in the |Jmj i basis for the 2p state, we must
94 UNIT 4: Fine and hyperfine structure of the hydrogen atom

perform a change basis to the |ml ms i basis, operate with Wz and then revert
back to |Jmj i. Namely, as found using a table of clebsch-gordan coefficients

33 1
| i = |1 i
22 2 r
31 1 1 2 1
| i = √ |1 − i + |0 i
22 3 2 3 2
r r
3 1 2 1 1 1
| − i= |0 − i + |−1 i (47)
2 2 3 2 3 2
3 3 1
| − i = |−1 − i
2 2 2 r
11 1 1 2 1
| i = − √ |0 i + |1 − i
22 3 2 3 2
r
1 1 2 1 1 1
| − i=− |−1 i + √ |0 − i
2 2 3 2 3 2

Acting with Wz on these states then entails

33 1 33
Wz | i = Wz |1 i = 2~ωL | i
22 r 2 22
√
31 2 1 2 31 2 11
Wz | i = ~ωL |0 i = ~ωL [ | i − | i]
22 3 2 3 22 3 22
r √
3 1 2 1 2 3 1 2 1 1
Wz | − i = − ~ωL |0 − i = −~ωL [ | − i + | − i] (48)
2 2 3 2 3 2 2 3 2 2
3 3 3 3
Wz | − i = −2~ωL | − i
2 2 √ 2 2
11 2 31 1 11
Wz | i = −~ωL [ | i − | i]
22 3 √2 2 3 22
1 1 2 3 1 1 1 1
Wz | − i = −~ωL [ | − i + | − i]
2 2 3 2 2 3 2 2

Writing then a sub-matrix for J = 3/2,

   
2 0 0 0 3/2 0 0 0
 0 2/3 0 0  4  0 
~ωL   = ~ωL  0 1/2 0  (49)
0 0 −2/3 0  3  0 0 −1/2 0 
0 0 0 −2 0 0 0 −3/2

Which as one may recognize is 34 ~ωL JZ . While the result obtained in (14) by
all means correct, the calculation proved to be somewhat lengthy. Fortunately
there exists a more expedient means to obtain this same result through the
exploitation of the Wigner-Eckert Theorem. According to this theorem, for a
~ That is,
fixed J, one may express any vector in terms of its projection along J.
the quantity L~ · 2S
~ is proportional to J~ by the constant of proportionality gj
Fine and hyperfine structure of the hydrogen atom 95

defined by
~ · Ji
hL ~ ~ · Ji
hS ~
gj = + 2 (50)
|J|2 |J|2
Thus for any fixed J we may then express

Wz = gj ωL Jz (51)

As a final note, one may recall that the results in this section are based upon
the magnetic field being sufficiently weak such that it may be treated as a time
independent perturbation. If the magnetic field is sufficiently strong then the
above analysis does not hold. Rather one must revert back to the |2lml msi
basis, where the action of Wz gives

Wz |2lml ms i = ~ωL (ml + 2ms ) |2lml ms i (52)

Stark Effect
In the next application of perturbation theory to the Hydrogen atom, we will
switch on an electric field E~ = E ẑ. The dipole moment of a proton and electron
separated by a vector R ~ is p~ = q R.
~ Our perturbation will be the energy of this
dipole in the above electric field.

p · E~ = −q E~ · R
Ws = −~ ~ (53)
In order to apply perturbation theory, it must be true that Ws ¿ H0 where H0
is the Hamiltonian of the unperturbed atom. In practice, unlike the magnetic
field, the electric field is almost never strong enough for this condition not to
hold. Furthermore, we shall assume that we are in a strong field regime so that
Ws À W1 + W2 + W3 where the perturbations on the right hand side are the
fine structure corrections.

Stark Effect: n = 1
We shall start by considering the n = 1 level. In this case l = 0 and m = 0 and
the zeroth order eigenstate is |100i. In this case the first order correction to the
energy is given by ∆E = h100 |Ws | 100i. This can be seen to vanish without
calculation by observing that |100i has even parity while Ws depends linearly
on R~ and so has odd parity.
Since the first order correction is zero, we must move on to the second order
correction given by
X |h100|Ws |nlmi|2
∆E = (54)
E1 − En
nlm
n6=1

Before calculating we know that ∆E ∝ E 2 since Ws ∝ E. Also, we know that

∆E < 0 since E1 , En < 0 and |E1 | > |En |. So, there is an energy shift, but the
question is: where does the dipole moment come from? We know that
96 UNIT 4: Fine and hyperfine structure of the hydrogen atom

h~ ~ = qh100|R|100i
pi = qhRi ~ =0 (55)
Since symmetry dictates that R ~ cannot choose a direction in which to point. So
at first glance, there doesn’t seem to be a dipole moment. The answer lies in
the change of state due to the perturbation.
X hnlm|Ws |100i
|100i → |100i + |nlmi (56)
E1 − En
nlm
n6=1

So the dipole moment is given approximately by

X hnlm|Ws |100i
h~
pi = hnlm|~
p|100i + c.c. (57)
E1 − En
nlm
n6=1

Before calculating anything we notice that h~pi ∝ E~ since the electric field is the
only thing in the problem that defines a direction in which the dipole moment
can point. This implies that hpx i = hpy i = 0. So, we only have to calculate
hpz i = qhzi.
In order to do this we shall use that
r
4π
z = r cos θ = rY10 (58)
3
Where Y10 is a spherical harmonic. This will greatly simplify the integral.
r Z Z
4π ∞
h100|z|nlmi = drr2 rR10 Rnl dΩY00 ∗
Y10 Ylm (59)
3 0
∗
√
Since Y00 = 1/ 4π = const. the angular integral simply gives δl1 δm0 . Since
Ws = −pz E, the dipole moment is given by

X∞
|h100|z|n10i|2
hpz i = −2Eq 2 = χE (60)
n=2
E1 − En

Where χ is the electric susceptibility.

Stark Effect: n = 2
Now l = 1, 0 where the degeneracy is 3 and 1 respectively. So, we must consider
4 states: |21mi and |200i where m = 0, ±1. We can see without calculation
that h200 |Ws | 200i = 0 by the same parity argument we used in the n = 1 case.
By a similar argument h21m |Ws | 21m0 i = 0 because Ws and the states |21mi
all have odd parity. So the only non-vanishing matrix elements are given by
r Z Z
4π ∞
h21m |Ws | 200i = − drr2 qErR20 R21 dΩY00 Y10 Y1m∗
(61)
3 0
Fine and hyperfine structure of the hydrogen atom 97

The angular integral yields a δm0 , so there are two non-vanishing matrix ele-
ments given by h210 |Ws | 200i = γE where γ is real and can be calculated from
the radial integral. The matrix representation of Ws can be constructed by
setting
       
1 0 0 0
0 1  0 0
|211i =        
0 , |210i = 0 , |21−1i = 1 , |200i = 0 (62)
0 0 0 1
Then the matrix representation of the perturbation becomes
 
0 0 0 0
0 0 0 γE 
Ws = 0 0 0 0 
 (63)
0 γE 0 0
From this matrix we can instantly see that |211i and |21−1i have eigenvalue
zero, i.e. there is no shift associated with them. This leaves the submatrix
µ ¶
0 γE
(64)
γE 0
Diagonalizing yields the energy shifts ∆E = ±γE and the zeroth order eigen-
states
µ ¶ µ ¶
1 1 1 1
|+i = √ , |−i = √ (65)
2 1 2 −1
98 UNIT 4: Fine and hyperfine structure of the hydrogen atom
UNIT 5

Time-dependent
perturbation theory

Notes by S. Kyle, A. Sunghoon, and T. Weisong

Introduction and Method
In the former chapter, we talked about the Time-independent Perturbation
Theory. In this chapter, we will work on a more complicated problem, the
Time-dependent Perturbation Theory. Now let’s consider the same problem as
we did in the chapter three: a system whose Hamiltonian can be expressed in
the form
H0 |n >= En |n >, (1)
H0 is the unperturbed Hamiltonian like what we did before. However, the
perturbation in this case is λW(t) , which depends on time. So if we want to
solve the Schrodinger equation, we will have the equation:

H = H0 + λW(t) (2)

If we want to find the eigenvalue of H, which is time dependent, we need to

solve time-depended Schrodinger equation:
d
H|ψ(t) >= i~ |ψ(t) > (3)
dt
In this case, we have the initial condition t = 0, and we have the eigenstate of
the original Hamiltonian:
|ψ(0) >= |i > (4)
where i is some values of n. After we measure the system at time t > 0, the
system will be in a different state of Hamiltonian |f >, where the f is another
number of n. If we are given the initial state |i >, the probability of the
transition to |f > is:
Pif (t) = | < f |ψ(t) > |2 (5)
100 UNIT 5: Time-dependent perturbation theory

So, Let’s calculate the probability and solve the Schrodinger equation. First
we need to split the function ψ into some states, which is the eigenstate of the
Hamiltonian. Therefore we can have
X
|ψ(t) >= cn (t)|n > (6)
n

where we have the coefficients cn are:

Cn (t) =< n|ψ(t) > (7)

Then we go back tho the Schrodinger equation, and split the Hamiltonian H.
We will have:
d
H0 |ψ > +λw|ψ >= i~ |ψ > (8)
dt
Since the equation 6 defined the |ψ > before, so can we have
X X X dcn
cn En (t)|n > +λ cn w|n >= i~ |n > (9)
n n n
dt

Next, we can select one state of n, which we can write as |k >, to multiply the
equation 9, so we can have:
X dcn
Ck Ek + λ cn < k|w|n >= i~ (10)
n
dt

In this case, the w is a matrix which we can write as wk n. let’ s simplified

the complex equation. Firstly, let assume λ = 0, which means there is no
perturbation, so we have:
dck
i~ = Ek ck (11)
dt
We can solve this equation seriately. Then we have:
Ek
Ck (t) = Ck (0)e−i ~ t
(12)

Next, we calculate the condition that λ 6= 0. Then we have

Ek
Ck (t) = bk (t)e−i ~ t
(13)

In this case, we got a complex function. However, we can separate the complex
function to two parts. The first part is bk (t), which we don’t know, while the
Ek
second part e−i ~ t that we have already talked about in former chapters. We
substitute the equation 13 into 10, then we can get:
Ek X Ek
Ck (t) = bk (t)e−i ~ t Ek + λ Wkn bn e−i ~ t (14)
n

which equals to
db k −i Ek t iEk −i Ek t
i~[ e ~ + bk (−i )e ~ (15)
dt ~
Time-dependent perturbation theory 101

and then we can get

X En −Ek dbk i
bk Ek + λ Wkn bn e−i ~ t
= i~[ − Ek bk ] (16)
n
dt ~

so we can get:
dbk λ X Ek −En
= Wkn e−i ~ t bn (17)
dt i~ n

This is the exact equation which we didn’t use any approximation, and this
equation is equivalent to the Schrodinger equation. In addition, if we write
Ek −En
~ = ωkn , then we can have the frequency ωk n, which we get from the
transition between state k to state n. We call the ωkn Born Frequency.

Approximate Solution Using Perturbation The-

ory
In last section, we calculate the exact solution of the Schrodinger equation
for a time-dependent perturbation question. However, the equation is very
complicated and we can not solve it in many cases. In this section, we find
the approximation solution through perturbation theory. Firstly, under the
perturbation theory, we will writer the equation bn (t) in the form:

bn (t) = b0n (t) + λb1n (t) + λ2 b2n (t)... (18)

First, we have to look the terms in the equation when there is no λ, then we
can have the λ0 , which means the 0th order, term:

db0k
=0 (19)
dt
Then we know that b0k is a constant. In this case, there is only one term which
we write as b0i equal to 1. Other b0k where k 6= i will be 0. So if we write equation
of ψ at t = 0, then we have:
X
|ψ(0) >= b0n |n >= |i > (20)
n

Now we consider the first order, we can get:

db1k
λ1 = i~ = eiWki t Wki (21)
dt
Form the former equation, we can integrate the b1k (t) as
Z t
1 0
b1k (t) = dt0 eiWki t Wki (t0 ) (22)
i~ 0
102 UNIT 5: Time-dependent perturbation theory

Somehow, in this case, when t = 0, we have b1k (t) = 0. Because b0i equal to
1, Other b0k where k 6= i will be 0. That is also the reason why all the other
coefficients will be vanish at t = 0. Since we have get the bn , then we can
calculate the cn and function |ψ >.
iEn t
|ψ(t) >= Σn bn (t)e− ~ |n > (23)

Then we can substitute the λ0 and λ1 we get before into this equation, we will
have: iEn t iEn t
|ψ(t) >= e− ~ |i > +λΣn b1n (t)e− ~ |n > ... (24)
So, if we measure the system at t > 0, I will find the system will stay at the
state |f >. Then we have probability of the transition from |i > to |f > is
Pif (t) = | < f |ψ(t) > |2 |, and it equals to
iEn t iEf t
Pif (t) = λ2 |Σn b1n (t)e− ~ < f |n > |2 = λ2 |b1f (t)e− ~ |2 (25)

As what we did for the Born frequency, we can write the probability as:
Z
λ2 t 0 iWif t0
Pif (t) = 2 dt e Wif (t0 ) (26)
~ 0
E −E
where we have Wf i = f ~ i . That is the probability for the first order. Now
we will work on some examples.
Example 1:W (t) = A sin(ωt)
In this case, we will have the probability of Pif (t) as:
Z t
|A|2 0
Pf i (t) = | dt0 eiWif t sin(ωt0 )|2
~2 0
2 i(ω+ωf i )t
|A| e e−i(ω−ωf i )t 2
= 2
|[ + ]|
~ 2(ω + ωf i ) 2(ω − ωf i )
|A|2 ei(ω+ωf i )t − 1 e−i(ω−ωf i )t − 1 2
= |[ − ]| (27)
4~2 ω + ωf i ω − ωf i

Example 2:W (t) = A cos(ωt)

This example is very familiar with the example 1, so we will have:

|A|2 ei(ω+ωf i )t − 1 ei(ω−ωf i )t − 1 2

Pf i (t) = |[ + ]| (28)
4~2 ω + ωf i ω + ωf i

Example 3:W (t) = A

In this case, the perturbation term equals to a constant. We will have:
iωf i t iωf i t
|A|2 eiωf i t − 1 eiωf i t − 1 2 |A|2 eiωf i t − 1 2 |A|2 e 2 − e− 2
Pf i (t) = |[ + ]| = |[ | = |[ ]|2
4~2 ωf i ωf i ~2 ωf i ~2 ωf i
(29)
Time-dependent perturbation theory 103

2ω t
fi
4|A|2 sin ( 2 )
so we can have Pif = ~2 ωf2 i
|A|2 ei(ω+ωf i )t −1
In the example 2, we have the probability of the transition is Pf i (t) = 4~2 |[ ω+ωf i +
2
ei(ω−ωf i )t −1 2
ω+ωf i which we can write as Pf i = |A|
]| 4~2 |A+ + A− |
2

So if we ignore the second term (A− ), we can calculate the rest part through
the same way we used in the example 3. Then we can get:
ω +ω
|A|2 2 |A|2 ei(ω+ωf i )t − 1 2 |A|2 sin2 f i2
P+ = |A + | = | | = (30)
4~2 4~2 ω + ωf i 4~2 (ωf i + ω)2

Next, if we ignore the fist term A+ , then we will have

ω −ω
|A|2 2 |A|2 sin2 f i2
P− = |A − | = (31)
4~2 4~2 (ωf i − ω)2

In this case, if we assume the state is near resonance, which means the distance
from the resonance is much smaller than the resonance, we will have |ω −ωf i | ¿
ωf i , then we get P+ ≈ 0 and Pif = P− . The width ∆ω = 4π t ¿ 2ωf i , so we
have t À ω2π fi
≈ 1
ω . This means we need to wait long enough, which is much
longer than a couple of oscillations, to the resonance. However, on the other
hand, I can not wait too long, for the probability can not bigger than 1. So we
have Pif ¿ 1, and because |A|t ~
2~ ¿ 1, then we can get t ¿ |A| . At last, we can
get ω1f i ¿ t ¿ |A|~
. If we write it in difference of energy levels, we will have
Ef − Ei = ~ωf i À |A|.

Secular Approximation
In last section, we discussed the situation near the resonance, which means
~ ~
t ¿ |A| . On the other hand, if we have t À |A| , the schrodinger equation will
be:
dbk X
i~ = eiWkn t Wkn bn (32)
dt n

in this case, we assume λ = 1 in the equation. Then we the example we did

in last section again. We assume W = A sin(ωt), then Wkn = Akn sin(ωt), in
which Akn =< k|A|n >. If we go near the resonance, we will have the condition
ω ≈ ωf i . We assume the initial and final state as |i > and |f >, and we only
consider the situation when the system in these two states. Then we will have:

dbi
i~ = Aii bi sin(ωt) + e−iWf i t Aif bf sin(ωt) (33)
dt
Another equation is:

dbf
i~ = eiωf i t Af i bi sin(ωt) + Af f bf sin(ωt) (34)
dt
104 UNIT 5: Time-dependent perturbation theory

eiωt −e−iωt
Because we have sin(ωt) = 2i , so we can have equations:

dbi 1
i~ = e−i(ωf i−ω)t Aif bf (35)
dt 2i
dbf 1
i~ = − ei(ωf i −ω)t Af i bi (36)
dt 2i
If we calculate the second derivative, then we can get:

d2 bf 1 1 dbi
i~ = (ωf i − ω)ei(ωf i −ω)t Af i bi − ei(ωf i −ω)t Af i
dt2 2 2i dt
dbf 1 2
= ~(ωf i − ω) − |Af i | bf (37)
dt 4~
so we can transform the equation to:

d2 bf dbf 1
+ i(ωf i − ω)t + 2 |Af i |2 bf = 0 (38)
dt2 dt 4~

If we assume bf = eλt , and we plug it into the former equation, then we have:

|λif |2
λ2 − i(ωf i − ω)λ + =0 (39)
4~2
Then we can solve the equation to get the λ is:
r
1 |Af i |2
λ = λ± = [i(ωf i − ω) ± i (ωf i − ω)2 + ] (40)
2 ~2
It is important to note that λ is pure imaginary, so that there is no damping.
We have also seen that
bf = eλ± t (41)
so the most general solution will be a linear combination of the two.

bf (t) = Aeλ+ t + Beλ− t (42)

To fix the coefficients A and B we look at the initial conditions, t = 0. At t = 0

the initial state is the state |i > so that bi (0) = 1 and bf (0) = 0. The condition
that bf (0) = 0 tells us that A = −B. The condition bi (0) = 1 is useful if we
know bi (0). From our equations above

dbi 1
i~ = ei(ω−ωf i )t Aif bf (43)
dt 2i
and using (42) we can show that

2~
bi (0) = (Aλ+ + Bλ− ) = 1 (44)
Af i
Time-dependent perturbation theory 105

and
Af i Af i
A= = q (45)
2~(λ+ − λ− ) |Aif |2
2i~ (ω − ωf i )2 + ~2

This allows us to calculate the probability of a transition from the state |i > to
the state |f >. It is given by
Pif (t) = |bf (t)|2 = |A|2 |eλ+ t − eλ− t |2
q
|Aif |2
|Aif | 2
2
(ω − ωf i )2 + ~2 t
= sin ( ) (46)
~2 (ω − ωf i )2 + |Aif |2 2
Since we are near the resonance the terms (ω − ωf i ) ≈ 0 so we can ignore them.
This gives
|Aif |
Pif (t) = sin2 ( t) (47)
2~
This is Rabi’s formula, which we studied last semester.

Interaction of an atom with light

When considering the interaction between an atom and an electromagnetic wave,
we first choose a gauge. It is always possible to choose a gauge such that
φ = A0 = 0, (48)
The vector potential is
−
→→ −→ −
→→
A (−
r , t) = Az sin( k .−
r − ωt) (49)
−
→
ω = c| k | (50)
−
→
Az = Az ẑ (51)
−
→
k = kŷ (52)
The electric field is given by
−
→
→
− −
→ dA
E = − ∇φ − = ωAz cos(ky − ωt)ẑ = E0 cos(ky − ωt)ẑ (53)
dt
and the magnetic field by
−
→ − → − →
B = ∇ ∗ A = kAz cos(ky − ωt)x̂ = B0 cos(ky − ωt)x̂ (54)
so that
E0 ω
= =c (55)
B0 k
The time averaged Poynting vector is given by
−
→ c²0 E02
< S >= ŷ (56)
2
106 UNIT 5: Time-dependent perturbation theory

Interaction of an atom with light

When considering the interaction between an atom and an electromagnetic wave,
we first choose a gauge. It is always possible to choose a gauge such that

φ = A0 = 0, (57)

The vector potential is

→
− − −
→ −
→→
A (→
r , t) = Az sin( k .−
r − ωt) (58)
−
→
ω = c| k | (59)
−
→
Az = Az ẑ (60)
−
→
k = kŷ (61)
The electric field is given by
−
→
−
→ →
− dA
E = − ∇φ − = ωAz cos(ky − ωt)ẑ = E0 cos(ky − ωt)ẑ (62)
dt
and the magnetic field by
→
− →
− − →
B = ∇ ∗ A = kAz cos(ky − ωt)x̂ = B0 cos(ky − ωt)x̂ (63)

so that
E0 ω
= =c (64)
B0 k
The time averaged Poynting vector is given by
−
→ c²0 E02
< S >= ŷ (65)
2

Interaction with an Hydrogen Atom: Low-Intensity

Limit
The hamiltonian for the electron in a Hydrogen atom interacting with an elec-
tromagnetic wave is:
−
→
(−
→
p − q A )2 q−→− →
H= + V (r) − S . B − qφ (66)
2m m
This can be rewritten as
q q q 2 A2z
H = H0 − pz Az − Sx B0 + (67)
m m 2m
where H0 is the atomic hamiltonian and
q q 2 A2z q
W (t) = W1 + W2 + W3 = − pz Az + − Sx B 0 (68)
m 2m m
Time-dependent perturbation theory 107

is the perturbation. We will consider W1 here since W1 is much larger than W3

or W2 . This can be seen by considering the order of the ratio

W3 ~k 1
≈ ≈ (69)
W1 p 1000

for everyday light, and noting that at low intensity the quadratic terms of Az
are insignificant. So,
q−→ −
→
H ≈ H0 + W 1 ≈ H0 − p .A (70)
m

The Electric Dipole Hamiltonian

Now let us consider a gauge transformation given by
−
→0 −→ → −
A = A + ∇χ (71)

dχ
φ0 = φ − (72)
dt
If we let
χ = zA0 sin(ωt) (73)

then
φ0 = −zA0 ωcos(ωt) (74)

A0x = A0y = 0 (75)

A0z = A0 sin(ky − ωt) + A0 sin(ωt) (76)

In the electric dipole approximation we have ky ≈ 0 so that

Az ≈ 0 (77)

In this case the interaction hamiltonian is

→
− − → q−→− →
H = H0 − qφ0 = H0 − qE0 zcos(ωt) = H0 − d . E − S .B (78)
m
Again, as before, we ignore the last term because its contribution is insignificant
and we have
−
→− →
W1 = − d . E (79)

W1 is called the electric dipole hamiltonian. The gauge transformation used

here illustrates that the W1 we found before((70)) is, in fact, the electric dipole
hamiltonian.
108 UNIT 5: Time-dependent perturbation theory

The Matrix Elements of the Electric Dipole Hamil-

tonian
In order to describe a transition from some initial state, i, to some final state, f,

|i > −− > |f > (80)

we need to calculate the matrix elements

qE0
< f |W1 |i >= sin(ωt) < f |pz |i > (81)
mω
In order to calculate the matrix elements we first consider the commutator
between H0 and Z. Using
p2
H0 = z + V (r) (82)
2m
[z, p2z ] = pz [z, pz ] + [z, pz ]pz = 2i~pz (83)
we get
pz
[Z, H0 ] = i~ (84)
m
On the other hand,

< f |[z, H0 ]|i >=< f |zH0 |i > − < f |H0 z|i >= −(Ef − Ei ) < f |z|i > (85)

Therefore
ωf i
< f |W1 |i >= iq E0 sin(ωt) < f |z|i > (86)
ω
ωf i = (Ef − Ei )/~ (87)
We see that the matrix elements of the electric dipole hamiltonian are propor-
tional to those of z.

Electric Dipole Selection Rules

Now let us consider transitions in an hydrogen atom. The wavefunctions will
be
φnlm (−
→r ) = Rnl (r)Ylm (θ, φ) (88)
The matrix elements can then be represented in integral form as
Z
m
< f |z|i >= I dΩYlf f ∗ (θ, φ)Y10 (θ, φ)Ylm
i
i
(θ, φ) (89)

where r Z ∞
4π
I= drr2 Rnf l (r)rRni li (r) (90)
3 0 f
Time-dependent perturbation theory 109

The angular integral, which was calculated in our discussion of “Addition of

Spherical Harmonics”(see Cohen-Tannoudji Cx ), is nonzero if and only if

∆l = lf − li = ±1 (91)

mf = mi (92)
The equations (91) (92) are the selection rules for the elecric dipole transition.
We can also consider the case where the polarization of the elecric field is in the
x-direction. In this case, the matrix elements are proportional to the integral
Z
m
dΩYlf f ∗ (θ, φ)[Y11 (θ, φ) − Y1−1 (θ, φ)]Ylm
i
i
(θ, φ) (93)

and the selection rules are

∆l = lf − li = ±1 (94)

∆m = mf − mi = ±1 (95)
We can generalize this to any direction of polarization. In this case the direction
will be some linear combination of x,y, and z. Our angular integral will then
have some combination of the corresponding spherical harmonics. Evaluation of
this integral will give us, as was shown previously, the selection rules for electric
dipole transitions

Transition Probability Associated with Natural

Light
Let us now consider an hydrogen atom placed in an electromagnetic field with
a spectrum of angular frequencies corresponding to natural light. In this case
our expression for E0 is a function of ω. If we consider the Poynting vector

−
→ c²0 E02
< S >= (96)
2
−
→
< S >= I(ω)∆ω (97)
This gives us a relation between E0 and I(ω), explicitly

2I(ω)∆ω
E02 = (98)
c²0

The probability of a transition, corresponding to emission or absorption, is given

by
ω −ω
q 2 ωf i 2 2 sin[( f i2 t)
Pif = 2 ( ) E0 | < f |Z|i > |2 (99)
~ ω (ωf i − ω)2
110 UNIT 5: Time-dependent perturbation theory

For natural light we replace E02 using equation 42 and get

Z ∞ ω −ω
q 2 ωf i 2 2I(ω)δω 2 sin[(
fi
2 t)
Pif = ( ) | < f |Z|i > | (100)
0 ~2 ω c²0 (ωf i − ω)2

To solve this integral we first cast it in the form

Z ∞
2q 2 2 ωf i 2
Pif = | < f |Z|i > | δω( ) I(ω)f (ω) (101)
c²0 ~2 0 ω
ω −ω
sin2 [( f i2 t)]
f (ω) = (102)
(ωf i − ω)2
If we plotI(ω) and f (ω) against ω we see that the largest contribution to the
ω
product I(ω)f (ω) comes from ω ≈ ωf i . Replacing I(ω) with I(ωf i ) and ( ωf i )
with 1 and noting that integrating from −∞ to 0 adds alomst nothing to the
integral, we can express the transition probability as
Z −∞
2q 2 2
Pif ≈ | < f |z|i > | I(ωf i ) δωf (ω) (103)
c²0 ~2 ∞

Using substitution and an integral table or solving with a computer we find

πq 2 t
Pif = | < f |z|i > |2 I(ωf i ) (104)
c²0 ~2
or
Pif πq 2
= | < f |z|i > |2 I(ωf i ) (105)
t c²0 ~2
When considering natural light we also need to average over all polarizations.
Our electric dipole matrix elements are then given by
1 1
< W1 >= (| < f |x|i > |2 + | < f |y|i > |2 + | < f |z|i > |2 ) = | < f |−
→
r |i > |2
3 3
(106)
giving
Pif πq 2
= | < f |−
→
r |i > |2 I(ωf i ) = B (107)
t c²0 ~2
This is known as Einstein’s B coefficient, it is the probability per unit time for
a transition corresponding to absorption or induced emission.

Einstein’s A and B coefficients

For spontaneous emission Einstein calculated his A coefficient using a rather
clever thought experiment. He considered a volume in which a finite number of
atoms are confined. He assumed that if the rate of absorption is given byNi B,
the rate of induced emission by Nf B, and the rate of spontaneous emission by
Time-dependent perturbation theory 111

Nf A, where Ni and Nf are the number of particles in the initial or final state
resectively, then the followng equality must hold

Nf B + Nf A = Ni B (108)

which gives
Ni Ef −Ei
A=( − 1)B = (e kT − 1)B (109)
Nf
In this case, we can get the rate of absorption can be written as

induced or
rate of absorption = ( ) emmision
stimulated

Einstein’s A and B coefficients are defined by

³ ~ωk ´
A = e kT − 1 B (110)
2
πq
B= |hf | ~r | ii|2 I (ωf i ) (111)
3²0 c~2

In case of black body radiation, we have the intensity of the photons I (ω) as

~2 ω 3
I (ω) = ³ ~ω ´ (112)
π 2 c2 e kT − 1

Thus, we get

πq 2 ~2 ωf3 i
A= |hf | ~r | ii|2
3²0 c~2 π 2 c2
q 2 ωf3 i
= |hf | ~r | ii|2 (113)
3π²0 c3 ~

Let’s look at asymptotic behavior of I (ω) in detail. For small ω, we can expand
the exponential term as
~ω ~ω
e kT − 1 = 1 + . (114)
kT
So, we get I (ω) as
~ω 3 ω2
I (ω) ≈ ¡ ~ω ¢ = 2 2 kT (115)
π 2 c2 kT π c
This result is well known from Maxwell equation.
However, if the radiation comes in ”packet” (such as photons with ~ω), we must
identify how many photons are in a certain E. We know that
nE
− nE e− kT
N (nE) ≈ e kT , P (n) = Pn = (116)
Z
112 UNIT 5: Time-dependent perturbation theory

, where
∞
X nE
Z= e− kT
n=0
1
= nE
1 − e− kT
Thus, we find hni as
X ∞
1 X − nE
hni = nP (n) = ne kT (117)
Z n=0
∞
X
E
Let’s define α as α = e− kT . Then, we get Z = αn . Hence, we can rewrite
n=0
hni as
∞
1 X n
hni = nα (118)
Z n=0
,as well as
∞
X
dZ
= nαn−1
dα n=0
∞
X
dZ
=⇒ α = nαn
dα n=0
α dZ
=⇒ hni =
Z dα
1
= α (1 − α) 2
(1 − α)
α
=
1−α
1
= E (119)
e kT − 1
3
Average # of photons with ~k = 2 hni (2π)
d k
3 ,whose 2 is the number of polarization.

Thus, the average # of photons with ω can be written as

Z Z
d3 k dkk 2 dΩ
2hni 3 = 2hni 3
(2π) (2π)
8π 2
= 3 hnidkk
(2π)
1
= 2 hnidkk 2
π
1
= 3 2 hnidωω 2 , f rom ω = ck (120)
c π
Time-dependent perturbation theory 113

Finally, we get the total energy Etotal as

1
Etotal = E hnidωω 2
c3 π 2
= I (ω) dω = cEtotal

Also, we get I (ω) dω as

1
=⇒ I (ω) dω = E hniω 2 dω
c2 π 2
1 1
=⇒ I (ω) dω = ~ω 2 2 ~ω ω2
c π e kT − 1
~ω 3 1
= 2 2 ~ω
c π e kT − 1

Dipole moment
Let’s consider an electron subject to a restoring force directed towards the origin
and proportiional to the displacement. | Ψ (0)i can be written as

| Ψ (0)i =| ii =| 0i (ground state)

and we have | Ψ (t)i as

iE0 t X iEn t
| Ψ (t)i = e− ~ | 0i + λ b(1)
n e
− ~ | ni + · · ·
n6=0

,where

Z t
Wf 0 0
λb(1)
n (t) = dt0 eiωf 0 t sin (ωt0 )
~ 0
· ¸
Wn0 ei(ωn0 +ω)t − 1 ei(ωn0 −ω)t − 1
= −
2i~ ωn0 + ω ωn0 − ω

and
qE0
Wn0 = hn | pz | 0i
mω
ωn0
= iqE0 hn | z | 0i
ω
Hence, we can rewrite | Ψ (t)i as
 
iE t X
− ~0 | 0i + λ
| Ψ (t)i = e b(1)
n e
−iωn0 t
| ni + · · ·  (121)
n6=0
114 UNIT 5: Time-dependent perturbation theory

,where
En − E0
ωn0 =
~
However, we know that e−iωn0 t goes to zero in nature due to the damping factor
e−Γt , in such that

e−iωn0 t = e−i(ωn0 −iΓ)t = e−Γt e−iωn0 t → 0, as t → ∞ (122)

Thus, we get
 
X qE0 ωn0 · iωt −iωt
¸
iE t
− ~0 | 0i + e e 
| Ψ (t)i = e hn | z | 0i | ni −
2~ ω ωn0 + ω ωn0 − ω
n6=0
(123)

~ (= q~z) becomes
Therefore, the projection of the dipole moment D

Dz (t) = hΨ (t) | qz | Ψ (t)i

X qE0 ωn0 · iωt ¸
e e−iωt
= h0 | qz | 0i + hn | z | 0i − h0 | qz | ni + C.C.
2~ ω ωn0 + ω ωn0 − ω
n6=0
R
The first term becomes zero due to even ∗ odd ∗ even, and we get
X q 2 E0 ωn0 · iωt ¸
e e−iωt
Dz (t) = | hn | z | 0i |2 − + C.C.
2~ ω ωn0 + ω ωn0 − ω
n6=0
X q 2 E0 ωn0 · ¸
2 1 1
= | hn | z | 0i | − cos (ωt)
2~ ω ωn0 + ω ωn0 − ω
n6=0
X q 2 E0 ωn0 · ¸
2 2ω
= | hn | z | 0i | − 2 cos (ωt)
2~ ω ωn0 − ω 2
n6=0
X q 2 E0 · ¸
2ω
= ωn0 | hn | z | 0i |2 − 2 cos (ωt) (124)
2~ ωn0 − ω 2
n6=0

Finally, we find the electric susceptibility Xe as,

X 2q 2 ωn0 | hn | z | 0i |2
Xe = − 2 − ω2 (125)
~ ωn0
n6=0

Spring Problem
For the spring problem, the differential equation can be written as

d2 z
m = −mω02 z + qE0 cos (ωt) . (126)
dt2
Time-dependent perturbation theory 115

And, the ansatz of this equation can be written as

z = Acos (ωt) (127)

By inserting this ansatz to find the coefficient A, we get

−mω 2 A = −mω02 A + qE0

qE0
⇒A= (128)
m (ω02 − ω 2 )

Hence, we find the projection of the dipole moment dz (= qz) as

q 2 E0
dz = cos (ωt) (129)
m (ω02 − ω 2 )

Also, we find Xe as
q2
Xe = (130)
m (ω02 − ω2 )

Let’s compare two results. We found two Xe as

X 2q 2 ωn0 | hn | z | 0i |2
Xe = − 2 − ω2
~ ωn0
n6=0

q2
Xe =
m (ω02 − ω 2 )
X 2mωn0 | hn | z | 0i |2 q2
⇒ Xe = 2 − ω2 )
~ m (ωn0
n6=0
X q2
= fn0 2 (131)
m (ωn0 − ω2 )
n6=0

,where
2mωn0 | hn | z | 0i |2
fn0 =
~
P
It turns out fn0 = 1, which means the sum of the probability is equal to 1.
Let’s prove it.
Since we find that
1
ωn0 hn | z | 0i = hn | [H0 , z] | 0i
~
i
= − hn | pz | 0i,
m
P
We can calculate fn0 as
X 2m X i
fn0 = − hn | pz | 0ih0 | z | ni (132)
~ m
116 UNIT 5: Time-dependent perturbation theory

1
Since we know C = 2 (C + C ∗ ) when C is a complex number,
X 2m X i
fn0 = − hn | pz | 0ih0 | z | ni + C.C.
~ 2m
iX
=− [h0 | z | nihn | pz | 0i − h0 | pz | nihn | z | 0i]
~
i
= − h0 | [z, pz ] | 0i
~
= 1 (q.e.d.) (133)

Blackbody Radiation
In case of the blackbody radiation, we have hn (E)i of the photon as
1
hn (E)i = E (134)
e kT eE/kT −1
The average number with ~k is, then,
d3 k
hn (E)i ∗ 3 (135)
(2π)

Since we can write ~k as

~k = (kx , ky , kz )
and using periodic conditions of the boundary condition, we find
Ψ(~k, ~r) = eikx x eiky y eikz z (136)
and
eikx L = 1
⇒ kx L = 2πL
2π
⇒ kx = n
L
2π
⇒ ∆kx = (137)
L
Thus, we get
L L L L3 3
hn (E)i( ∆kx )( ∆ky )( ∆kz ) = 3d k (138)
2π 2π 2π (2π)
So, the energy E can be found by
Z
L3 3
E = ~ω hn(E)i d k
(2π)3
V ~ω 3 dω
= 2 3 ~ω/kT (139)
π c (e − 1)
Time-dependent perturbation theory 117

And, we find the energy density du(= E/V ) by

~ω 3 dω
du = (140)
π 2 c3 (e~ω/kT − 1)

Therefore, we find I(ω) as

du ~ω 3
I(ω) = c = 2 2 ~ω/kT (141)
dω π c (e − 1)

Continuous Spectrum: | ii →| f i
We will consider the case that has a transition from discrete states to continuous
states. Let’s say the energy of the final states can be written as

p2f
Ef = . (142)
2m
In case of continuous spectrum, we may write the probability of the transition
as Pi (Ef ) instead of Pif , where Ef < E < Ef + ∆Ef . Also, the probability
can be written as
P robability = Pi (Ef ) ∆Ef ρ(Ef ) (143)
, where the density of states ρ(E) is

# of states
ρ(E) =
∆E
In order to find ρ(E), we will introduce a projection operator | p~ih~
p |.
p2
In the range of Ef < E(= 2m ) < Ef + ∆Ef , we find
Z Z Z
d3 p | p~ih~
p | = dpp2 dΩ | p~ih~p|
Z Z
m dE
= √ 2mE dΩ | p~ih~p|
2 mE
Z Z √
= dE dΩ m 2mE | p~ih~ p|
Z Z
= dE dΩ ρ(E, Ω) | p~ih~p| (144)

,where
# of states
ρ(E, Ω) = .
∆E∆Ω
We know that
| f i =| p~i =| E, Ωi (145)
So,
Pi (Ef ) =| hΨ(t) | f i |2 =| hΨ(t) | E, Ωi |2 (146)
118 UNIT 5: Time-dependent perturbation theory

Let’s look at examples.

Example 0i To measure the energy of electrons.

Assuming E = 10eV, with σE = 1eV , we have the probability as
Z
P robability = P (E, Ω)ρ(E, Ω)dλdE (147)

Example 1i Constant perburbation.

(1)
P (E, Ω) =| hE, Ω | Ψ(t)i |2 =| bE (t) |2
Z t
1 E−Ei 0
= 2 | dt0 ei ~ t | hE, Ω | W | ii |2
~ 0
| hE, Ω | W | ii |2 sin2 E−E
2~ t
i

= (148)
~2 ( E−E i 2
2~ )

Since we know Z
P robability
= dEρ(E, Ω)P (E, Ω)
∆Ω
,and by assuming ρ(E, Ω) is approximately constant, which is ρ(Ei , Ω), we find
Z
P robability ∞
| hE, Ω | W | ii |2 sin2 E−E
2~ t
i

' ρ(Ei , Ω) dE 2 E−E

∆Ω 0 ~ ( 2~ i )2
Z
| hE, Ω | W | ii |2 ∞ sin2 E−E
2~ t
i

= ρ(Ei , Ω) 2
dE E−E
(149)
~ 0 ( 2~ )i 2

For E < 0, the integral is very small(⇒ 0 → -∞.) By defining a new variable u
as
E − Ei dE
u≡ t, du = t,
2~ 2~
we find
Z
P robability | hE, Ω | W | ii |2 ∞ 2~ sin2 ut2
= ρ(Ei , Ω) 2
du
∆Ω ~ −∞ t u2
2πt
= ρ(Ei , Ω) | hEi , Ω | W | ii |2 (150)
~
Hence, we get w as
P robability 2π
w≡ = ρ(Ei , Ω) | hEi , Ω | W | ii |2 = constant!! (151)
∆Ω∆t ~
This result, w = constant, is so called ”Fermi’s Golden Rule.”

Now, let’s consider the scattering theory to describe with the time perturbation
theory. We have an electron e− for incoming particle having a mass m and a
Time-dependent perturbation theory 119

momentum p~i , as well as a heavy target having a potential V. Due to this po-
tential V, the momentum of the electron is deflected to p~f . From this situation,
we get
p2 p2f
Ei = i , Ef = , Ei = Ef , p2i = p2f (152)
2m 2m
The hamiltonian H can be written as
p2
H= + V = H0 + W (153)
2m
, where
p2
H0 =
, W = V.
2m
We define an initial state and a final state as

| ii =| p~i i, | f i =| p~f i =| Ei , Ωi (154)

Then, w becomes
2π √
w= m 2mE | hp~f | V | p~i i |2 (155)
~
We need to find the form of | p~i i. Since we know that

p | p~0 i ∝ δ 3 (~
| p~i = C ei~p¦~r/~ and h~ p − p~0 ),

we need to find the coefficient C. Let’s work on it.

We have an inner product h~ p | p~0 i as
Z
~0
p | p~0 i = C 2 d3 rei(p −~p)¦~r/~
h~
Z ∞
0
= dxei(px −px )x/~
−∞
Z ∞ Z 0
0 0
= dxei(px −px )x/~ + dxei(px −px )x/~
0 −∞
Z ∞ Z ∞
i(p0x −px )x/~ 0
= dxe + dxe−i(px −px )x/~ (156)
0 0

By multiplying e−²x with ² > 0 but ² ∼ 0, we get

1 1
p | p~0 i =
h~ (p0 −p )
+ (p0 −p )
² ²
~ − i x~ x ~ + i x~ x
2²~
= (157)
²2 + (p0x − px )2

Let’s define f (q) as

2²~
f (q) = (158)
²2 + q 2
120 UNIT 5: Time-dependent perturbation theory

When ² → 0, f (q) behaves in such that

f (q) → 0, if q 6= 0 (159)
f (q) → ∞, if q = 0 (160)

Thus, we find that Z Z

∞ ∞
2²~
f (q)dq = dq (161)
−∞ −∞ ²2 + q 2
²
Let q = ²tanθ. Then, dq = cos2 θ dθ. So, we get
Z ∞ Z ∞
2²~ ²
f (q)dq = 2
cos2 θ dθ
−∞ −∞ ² cos2 θ
Z π/2
= 2~ dθ = 2π~ (162)
−π/2

Thus, we find f (q) as

f (q) = 2π~δ(q) (163)
And, we get
Z ∞
0
dxei(px −px )x/~ = 2π~δ(p0x − px )
−∞
Z ∞
~0
=⇒ d3 r ei(p −~p)¦~r/~ = (2π~)3 δ 3 (p~0 − p~) (164)
−∞

p | p~0 i, we finally find C as

By recalling the result of h~

p | p~0 i = C 2 (2π~)3 δ 3 (p~0 − p~)

h~
1
=⇒ C = (165)
(2π~)3/2

Therefore, we find w(~

pi , p~f ) as
Z
2π √ 1
w(~
pi , p~f ) = m 2mE | d3 r V (r)ei(p~i −p~f )¦~r/~ |2 (166)
~ (2π~)6

We can write the probability current density Ji as

~
Ji = (Ψ∗ ∂z Ψ − C.C.)
2mi
pi 1
= (167)
m (2π~)3

,where
1
Ψ= eipi z/~
(2π~)3/2
Time-dependent perturbation theory 121

Hence, the differential cross section dσ/dΩ can be found

Z
dσ w 2π √ m 1
= = m 2mE | d3 r V (r)ei(p~i −p~f )¦~r/~ |2
dΩ Ji ~ pi (wπ~)3
Z
m2
= | d3 r V (r)ei(p~i −p~f )¦~r/~ |2 (168)
4π 2 ~4
= same as Born cross section!!

Example 2i Sinusoidal Perturbation.

We can state the transition probability Pif (t, ω) as

| wf i |2 sin2 { (E−E2~i −~ω) t}

Pif (t, ω) =| hE, Ω | Ψ(t)i |2 = = (169)
4~2 ( (E−E2~i −~ω) )2

Since we have a condition in such that

E ∼ Ei + ~ω,

we get w as
π
w= | hEi + ~ω, Ω | W | ii |2 ρ(Ω, Ei + ~ω). (170)
2~

Example 3i Decay; nucleus → α + nucleus0 .

Again, in this problem we also have a transition from a discrete state(nucleus,
| ii) to two continuous states(α and nucleus0 , | E, Ωi). We will talk about the
decay rate Γ at this time, which is defined as
Z
decay rate Γ ≡ probability/time = dΩw(Ω). (171)

We also define the probability of no decay Pii as

Pii = 1 − Γt. (172)

In order to make it true, we must have a condition in such that

1
Γt ¿ 1 ⇒ t ¿ . (173)
Γ
Also, if you look at this Γ as a function of E, then
2π
Γ(E) = | hE, Ω | W | ii |2 ρ(Ω, E) (174)
~
Since we have a majority of Γ(E) in the range of the small width centered by a
peak, we have another condition in such that
1
t À (175)
∆Γ
122 UNIT 5: Time-dependent perturbation theory

Therefore, we have a condition,

1 1
¿ t ¿ (176)
∆Γ Γ

Example 4i Beyond perturbation theory; t ≥ 1/Γ.

We can consider a Schrodinger equation, in such that

dbk X
i~ = eiωkn t Wkn bn (177)
dt n

, where

Ei − E
ωkn =
~
and Wkn = hi | W | E, Ωi

In the case of the continuous state and k = i, we can rewrite this equation as
Z
dbi (Ei −E)
i~ = dΩdEρ(Ω, E)ei ~ t hi | W | E, Ωib(Ω, E) (178)
dt

When k = k(Ω, E), it becomes

db(Ω, E) (Ei −E)

i~ = ei ~ t hE, Ω | W | iibi (179)
dt
At t = 0, bi = 0 and b(Ω, E) = 0. So,
Z
dbi 1 (Ei −E)
=− 2 dΩdEρ(Ω, E)ei ~ t hi | W | E, Ωi
dt ~
Z t
(E−Ei ) 0
× dt0 ei ~ t hΩ, E | W | iibi (t0 )
0
Z Z t
1 (Ei −E) 0
=− 2 dΩdEρ(Ω, E) dt0 | hi | W | E, Ωi |2 ei ~ (t−t ) bi (t0 )
~ 0
Z Z t
1 ~ (Ei −E) 0
=− 2 dEΓ(E) dt0 ei ~ (t−t ) bi (t0 ) (180)
~ 2π 0

When (t − t0 ) is very large, we get

Z
(Ei −E) 0 E(t − t0 )
dEΓ(E)ei ~ (t−t ) ∝ sin
~
E(t − t0 )
⇒ =π
~
π~
⇒E= (181)
(t − t0 )
Time-dependent perturbation theory 123

Thus, we have a rapid oscillation and it leads to

Z t
dt0 · · · = 0!! (182)
0

When (t − t0 ) is very small, we get

Z t
dt0 · · · 6= 0, (183)
0

1
compared to ∆Γ .
0
For (t − t ), we have a condition that

1
large : t − t0 À
∆Γ
1
small : t − t0 ≤
∆Γ
1
=⇒ (t − ) < t0 < t only f or a reasonable contribution.
∆Γ
=⇒ bi (t0 ) ∼ bi (t) ∼ bi

Finally, we find
· Z Z t ¸
dbi 1 (Ei −E) 0
=− dE dt0 Γ(E)ei ~ (t−t ) bi
dt 2π~ 0
" Z (Ei −E)
#
1 ei ~ t − 1
=− dE Γ(E) bi
2π~ i Ei~−E
· Z ¸
1
=− dE Γ(E)f (E) bi (184)
2π~

, where
(Ei −E)
ei ~ t
−1
f (E) =
i Ei~−E

Note that | f (E) |2 becomes

sin2 Ei2~
−E
t
| f (E) |2 = . (185)
( Ei~−E )2

Since Γ(E) is approximately constant, we can rewrite dbi /dt as

· Z ¸
dbi 1
=− Γ(Ei ) dE f (E) bi (186)
dt 2π~
R
In addition, we want to look at dE f (E) in detail.
124 UNIT 5: Time-dependent perturbation theory

R
For the real part of dE f (E), we find
Z Z (Ei −E) (Ei −E)
1 ∗ 1 ei ~ t − e−i ~ t
dE [f (E) + F (E)] = dE
2 2 i Ei~−E
Z ∞
sin Ei~−E t
= dE Ei −E (187)
−∞ ~
E−Ei
Let’s define u as u = ~ t to solve this integral.
Z Z ∞
1 sinu
dE [f (E) + F ∗ (E)] = ~ du
2 −∞ u
= π~ (188)
R
On the other hand, for the imaginary part of dEf (E), we find
Z
1
dE [f (E) − F ∗ (E)] = constant ∼ δE. (189)
2i
Thus, we find
· ¸
dbi 1 δE
= − Γ(Ei ) + i bi
dt 2 ~
Γ δE
=⇒ bi (t) = e−( 2 +i ~ )t (190)
In case of no decay, we have
Pii =| bi (t) |2 = e−Γt = 1 − Γt + · · · (191)
In the time dependent perburbation theory, we had
Pii = 1 − Γt (192)
Thus, we can see good agreement between the exact solution and the perturba-
tion theory.
For the decay probability, we have
Z
1 t 0 i E−Ei t0
b(Ω, E) = dt e ~ hΩE | W | iibi (t0 )
i~ 0
Z t
1 Γt E−Ei −δE 0
= hΩE | W | ii dt0 e− 2 ei ~ t
i~ 0
Γt E−Ei −δE
1 − e− 2 ei ~ t
= hΩE | W | ii (193)
E − Ei − δE + i~ Γ2
Since the probability density is defined by | b |2 , we finally get
1
| b |2 ∼ 2 (194)
(E − Ei − δE)2 + ~2 Γ4
We can find the uncertainty principle, in such that
~
∆E ≥ ~Γ = ⇒ τ ∆E ≥ ~. (195)
τ
UNIT 6

Identical particles