LECTURES ON QUANTUM MECHANICS-II

By

A. K. Kapoor
School of Physics
University of Hyderabad
Hyderabad 500046, INDIA

Version 1.1
November 4, 2010

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Contents

1 Addition of Angular Momentum 5

§ 1 Summary of results on angular momentum . . . . . . . . . . . . . . . 5
§ 2 Statement of the Problem . . . . . . . . . . . . . . . . . . . . . . . . 6
§ 3 Outline of the method and a summary of results . . . . . . . . . . . . 8
§ 4 Tables of Clebsch Gordon coefficients . . . . . . . . . . . . . . . . . . 11
§ 5 Restrictions on CG Coefficients . . . . . . . . . . . . . . . . . . . . . 12
§ 6 Examples of Adding Angular Momenta . . . . . . . . . . . . . . . . . 15
§ 6.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
§ 6.2 Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
§ 6.3 Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2 Spin and Identical Particles 19

§ 1 Spin as a Dynamical Variable . . . . . . . . . . . . . . . . . . . . . . 19
§ 2 Spin Wave Function and Spin Operators . . . . . . . . . . . . . . . . 20
§ 2.1 Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
§ 3 Pauli Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
§ 4 Total Wave Function of a Particle with Spin . . . . . . . . . . . . . . 24
§ 5 Indistinguishability in Quantum Mechanics . . . . . . . . . . . . . . . 25
§ 6 The Symmetrisation Postulate . . . . . . . . . . . . . . . . . . . . . . 27
§ 7 Illustrative Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
§ 7.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
§ 7.2 Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
§ 7.3 Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
§ 8 A First Look at He Atom Energy Levels . . . . . . . . . . . . . . . . 30

3 Variation Method 32
§ 1 Variation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
§ 1.1 Ritz variation method . . . . . . . . . . . . . . . . . . . . . . 33

4 Time Independent Perturbation Theory 35

§ 1 Time Independent Problems . . . . . . . . . . . . . . . . . . . . . . . . 35
§ 2 Nondegenerate Case . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
§ 2.1 First Order Perturbation Corrections . . . . . . . . . . . . . . 37
§ 2.2 Second Order Nondegenerate Perturbation Theory . . . . . . . 38
§ 3 Degenerate Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
§ 3.1 First Order Energy Level Splitting . . . . . . . . . . . . . . . 41
§ 3.2 First Order Wave Function . . . . . . . . . . . . . . . . . . . . 42

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 § 3.3 Second Order Degenerate Perturbation Theory . . . . . . . . . 43
§ 4 Structure of hydrogen atom energy levels . . . . . . . . . . . . . . . . 44
§ 5 Related topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
§ 6 Examples and Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 53

5 The WKB Approximation 56

§1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
§2 Semiclassical expansion . . . . . . . . . . . . . . . . . . . . . . . . . . 56
§3 Validity of WKB Approximation . . . . . . . . . . . . . . . . . . . . 58
§4 Bound state solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
§ 4.1 Connection Formula . . . . . . . . . . . . . . . . . . . . . . . 62
§ 5 Tunnelling Through a Barrier . . . . . . . . . . . . . . . . . . . . . . . 66
§ 6 WKB approximation for radial equation . . . . . . . . . . . . . . . . . . 67
§ 7 Examples and Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 69
§ 7.1 Tie Loose Ends . . . . . . . . . . . . . . . . . . . . . . . . . . 69
§ 7.2 Learn to Apply what you Learnt . . . . . . . . . . . . . . . . 69
§ 7.3 Complete the Exercises to the Bitter End . . . . . . . . . . . . 69
§ 7.4 Establish Cross Links with Other Topics, Techniques . . . . . 71
§ 7.5 Search, Discuss and Explore . . . . . . . . . . . . . . . . . . . 72

6 Scattering Theory 73
§ 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
§ 2 Scattering in classical and quantum mechanics . . . . . . . . . . . . . 73
§ 3 Integral Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
§ 4 Integral Equation for Scattering . . . . . . . . . . . . . . . . . . . . . . 79
§ 5 Born Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
§ 6 Problems and Cross Links . . . . . . . . . . . . . . . . . . . . . . . . 85

7 The Method of Partial Waves 87

§1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
§2 Partial Wave Expansion of Plane Waves . . . . . . . . . . . . . . . . 87
§3 Radial Wave Fnction . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
§4 Relating Cross Section to Phase Shifts . . . . . . . . . . . . . . . . . 92
§5 Phase Shifts for Square Well . . . . . . . . . . . . . . . . . . . . . . . 95

8 Time Dependent Perturbation Theory 97

§ 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
§ 2 Time Dependent Perturbation Theory . . . . . . . . . . . . . . . . . 98
§ 3 Transition Amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
§ 4 Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
§ 5 Fermi Golden Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
§ 6 Probability for Resonance Transitions . . . . . . . . . . . . . . . . . . 105
§ 7 Examples and Applications . . . . . . . . . . . . . . . . . . . . . . . 109
§ 8 Tutorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 9 Semiclassical Theory of Radiation 114
§ 1 Electromagnetic Waves in Vacuum . . . . . . . . . . . . . . . . . . . 114
§ 1.1 Free E.M. waves . . . . . . . . . . . . . . . . . . . . . . . . . . 114
§ 1.2 Scalar and Vector Potential . . . . . . . . . . . . . . . . . . . 115
§ 1.3 Gauge Transformation and Gauge Invariance . . . . . . . . . . 115
§ 1.4 Gauge Condition . . . . . . . . . . . . . . . . . . . . . . . . . 116
§ 1.5 Plane Wave Solutions . . . . . . . . . . . . . . . . . . . . . . . 116
§ 2 Interaction of Charged Particle with Radiation . . . . . . . . . . . . . 116
§ 2.1 Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . . . 117
§ 3 Semiclassical Treatment of Radiation . . . . . . . . . . . . . . . . . . 117
§ 4 Einstein A and B Coefficients . . . . . . . . . . . . . . . . . . . . . . 120

Version 1.1

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 1

Addition of Angular Momentum

§1 Summary of results on angular momentum

QM-
08-
Sum
⊲
⊠1) The angular momentum commutation relations are
[Jx , Jy ] = i~Jz , (1.1)
[Jy , Jz ] = i~Jx , (1.2)
[Jz , Jx ] = i~Jy . (1.3)

⊠2) It is useful to introduce the operators J± = Jx ± iJy and these operators obey
commutation relations
[J+ , J− ] = 2~Jz , (1.4)
[Jz , J+ ] = ~J+ , (1.5)
[Jz , J− ] = −~J− . (1.6)
Also J~2 commute with all the three components of the angular momentum
Jx , Jy , Jz .
⊠3) Some useful relations are
J+ J− = J 2 − Jz2 + ~Jz , (1.7)
J− J+ = J 2 − Jz2 − ~Jz , (1.8)
1
J2 = (J+ J− + J− J+ ) + Jz2 . (1.9)
2

⊠4) One can find simultaneous eigenvectors of J~2 and any one component, Jn =
~ along a fixed direction given by the unit vector n̂.
n̂ · J,
⊠5) The eigenvalues of J~2 are j(j + 1)~2 where j can be integer of half integer.
The eigenvalues of a component of Jn take values from −j to j in step of 1.
⊠6) It is customary to denote the normalised, simultaneous,eigenvectors of J~2 and
Jz by |jmi so that
J~2 |jmi = j(j + 1)~|jmi, (1.10)
Jz |jmi = m~|jmi. (1.11)

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 ⊠7) The operators J± acting on |jmi give a ket vector proportional to |j(m ± 1)i.
The proportionality coefficient can be worked out using the relation Eq.(1.7)
and Eq.(1.8). One then gets
p
J± |jmi = (j ∓ m)(j ± m + 1) ~ |j(m ± 1)i. (1.12)

⊠8) The operators J± annihilate |j, ±ji because the Jz value cannot be increased
beyond j nor can it decrease beyond −j.
J+ |j, ji = 0 J− |j, −ji = 0. (1.13)

⊠9) The above results are applicable to operators satisfying angular momentum
commutation relations except that the half integral values are ruled out for
the orbital angular momentum, beacause of additional requirement of single
valuedness of the wave function.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-08-Sum;

§2 Statement of the Problem

QM-
Lec- ⊲
Consider two sets of angular momentum operators J~ (1) and J~ (2) such that
13001

[Jx(α) , Jy(α) ] = i~Jz(α) α = 1, 2, (1.14)

and every component of J~ (1) commutes with every component of J~ (2) ,

[J~ (1) , J~ (2) ] = 0 . (1.15)
(1) (2)
is obvious that J~ (1) , J~ (2) , Jz and Jz from a commuting set. Let us now define
2 2

total angular momentum J~ = J~ (1) + J~ (2) , then

(a) J~ satisfies angular momentum commutation relations
[Jx , Jy ] = i~Jz etc. (1.16)

(b) All the three components of J~ commute with J~12 and J~22 .

(c) It follows from (b) that J~ 2 commutes with J~12 and J~22 .
The components of total J~ satisfy angular momentum commutation relations and
hence the operators J 2 and Jz have simultaneuos eigenvectors and have eigenvalues
as shown in table below:
Angular momentum eigenvalues
Operator Eigenvalues Remarks

J2 J(J + 1)~2 J = 0, 1/2, 1, · · ·

Jz M~ values of M are from
−J to J in steps of unity

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 In fact the component of J~ along every unit vector n̂ has the same set of eigenval-
ues as Jz but eigenvectors are of course different. Given the angular momentum
commutation relations it is straight forward to see that one can have two sets of
commuting operators. Each of these two sets of operators can have simultaneous
eigenvectors. The two commuting sets of operators and the notation for their simul-
taneous eigenvectors is shown in the table below.

Commuting Set Simultaneous Eigenvectors

2 2 (1) (2)
J (1) , J (2) , Jz Jz |j1 , j2 ; m1 m2 i
2 2
J (1) , J (2) , J 2 , Jz |j1 , j2 JMi

The vectors |j1 , j2 ; m1 m2 i listed in the second column of the first row are just the
(1)
direct product of vectors |j1 m1 i and |j2 m2 i which are eigenvectors of J~ (1)2 , Jz and
(2)
of J~ (2)2 , Jz , respectively. The set of all the vectors

|j1 , j2 ; m1 m2 i ≡ |j1 m1 i|j2 m2 i (1.17)

for different m1 , m2 values form a basis in the tensor product V1 ⊗V2 of spaces V1 , V2
spanned by |j1 m1 i and |j2 m2 i.
The problem of addition of angular momenta consists of the following questions

(i) For a state |j1 j2 ; m1 m2 i, which has definite values j1 , m1 and j2 , m2 for the in-
dividual values of the square, and the z− component of the angular momenta,
what values of square of total angular momentum J~2 and Jz are allowed?

(ii) The second part of the problem is to construct the transformation matrix con-
necting the two sets of simultaneous eigenvectors, one set for each commuting
set.
The states |j1 j2 JMi can be expressed as linear combination of states
|j1 j2 ; m1 m2 i and vice versa
X
|j1 j2 m;1 m2 i = C(JM; j1 j2 m1 m2 )|j1 j2 JMi (1.18)
JM

The coefficients C(JM; j1 j2 m1 m2 ) are found by using orthogonality of states

|j1 j2 JMi for different J, M. This gives

C(JM; j1 j2 m1 m2 ) = hj1 j2 , JM|j1 j2 ; m1 m2 i (1.19)

j1 +j2
X
∴ |j1 j2 ; m1 m2 i = hj1 j2 , JM|j1 j2 ; m1 m2 i |j1 j2 JMi. (1.20)
j=|j1 −j2 |

Conversely
X
|j1 j2 JMi = hj1 j2 ; m1 m2 |j1 j2 , JMi |j1 j2 ; m1 m2 i. (1.21)
m1 ,m2
m1 +m2 =M

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The states |j1 j2 JMi are linearly independent and will form another basis set
in the tensor product space V1 ⊗ V2 . The transformation between the two o.n.
bases will be a unitary transformation preserving normalisation and scalar
products of vectors in the tensor product space.

Since the values of j1 , j2 are fixed and common for all the states, we shall frequently
suppress these labels and write the kets |j1 j2 JMi and |j1 j2 ; m1 m2 i as |JMi and
as |; m1 m2 i respectively. The coefficients hj1 j2 , JM|j1 j2 ; m1 m2 i , written in short
as hJM|j1 j2 ; m1 m2 i , are known as Clebsch Gordon coefficients and expressions are
tabulated in books. The problem of addition of angular momenta consists of finding
allowed range of values of J, M and obtaining expressions for the Clebsch Gordon
coefficients.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
13001;

§3 Outline of the method and a summary of re-

sults
QM-
Lec-
13002
⊲
The derivation of the results on angular momentum addition will carried out in
several parts outlined below. We will show that the allowed values of J, M in the
state |j1 m1 i|j2 m2 i are restricted by the following constraints.

⊠1) The z-component of the total angular momentum, M, is restricted by M =

m1 + m2

⊠2) The total angular momentum J 2 can take J(J + 1)~2 , where the minimum
value of J is |j1 − j2 | and the maximum value is (j1 + j2 ).

⊠3) The allowed values of J are all values in the range |j1 − j2 | to (j1 + j2 ) in steps
of unity.

⊠4) Each combination of J, M appears only once, no value of J, M gets repeated.

The above statements are equivalent to saying that the Clebsch Gordon coefficients
are zero when any one of the above conditions is not satisfied. In addition one must
remember the angular momentum results that value of M must always be between
−J and J, and J must be positive.

M = m1 + m2 :
(1) (2)
The vectors |; m1 m2 i are eigenvectors of Jz , Jz . We shall now show that they are
eigenvectors of Jz with eigenvalue m1 + m2 .


Jz |; m1 , m2 i = Jz(1) + Jz(2) |; m1 , m2 i = (m1 + m2 )|; m1 , m2 i. (1.22)

Hence the state |JMi is orthogonal to |; m1 , m2 i, giving

hJM|; m1 m2 i = M 6= m1 + m2 (1.23)

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 J takes values bewteen |j1 − j2 | and (j1 + j2 )
All CG coefficients can be fixed in terms of a single
coefficient hj1 j2 ; m1 m2 |JMi |m1 =j1 ,M =J , with m2 = M − m1 = J − j1 ), which is
nonzero only if J − j1 lies in between −j1 and j1 . Thus

−j2 ≤ J − j1 ≤ j2 (1.24)

By repeating the above steps with j1 and j2 interchanged we would get

−j1 ≤ J − j2 ≤ j1 (1.25)

These two conditions, Eq.(1.24) and Eq.(1.25), are equivalent to the requirement
J ≥ |j1 − j2 |. Since maximum possible value of M = m1 + m2 is j1 + j2 we must
have J ≤ j1 + j2 . Thus the total angular momentum is constrained to lie between
|j1 − j2 | and j1 + j2 . The three numbers J, j1 , j2 should be such that they satisfy
triangle inequalities
|j1 − j2 | ≤ J ≤ j1 + j2 . (1.26)
It should be remarked that Eq.(1.24)-Eq.(1.26), written as conditions on j1 and j2 ,
are equivalent to each of the following two alternate forms

|J − j2 | ≤ j1 ≤ J + j2 and |J − j1 | ≤ j2 ≤ J + j1 . (1.27)

These inequalities, known as triangle inequalities, state that the allowed combina-
tions of values of j1 , j2 and J are precisely those which can represent lengths of three
sides of a triangle.

J takes values in steps of 1

The range of J, the total angular momentum value, has been determined; the mini-
mum value being |j1 − j2 | and the maximum value is (j1 + j2 ). Are all integral, and
half integral values, allowed in this range allowed ? We must fix which values of J
are allowed and which ones are not allowed?
The state |j1 j2 ; m1 m2 i, when both m1 and m2 have their maximum allowed
values j1 and j2 , the value of M = m1 + m2 is maximum and equal to j1 + j2 . There
is only one such state and this must correspond to J = j1 + j2 .

|J = (j1 + j2 ), M = (j1 + j2 )i = |j1 j2 ; m1 = j1 m2 = j2 i (1.28)

The next value of M = j1 + j2 − 1 corresponds to two linearly independent states

corresponding to
(a) m1 = j1 − 1, m2 = j2 , and

(b) m1 = j1 , m2 = j2 − 1
What are the corresponding J values? One combination of these two states must be
the M = j1 + j2 − 1 partner of the state Eq.(1.28); and the other linear combination
can only correspond to the next value J = j1 + j2 − 1. Continuing in this way, we
now consider M = j1 + j2 − 2 which will come from three sets of m1 , m2 values, i.e.

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 (a) m1 = j1 − 2, m2 = j2

(b) m1 = j1 − 1, m2 = j2 − 1, and,

(c) m1 = j1 , m2 = j2 − 2

Two linear combinations of the three states |j1 j2 ; m1 m2 i, corresponding to the set of
above m1 , m2 values, will correspond to the J values j1 +j2 , j1 +j2 −1 already found;
a third linear combination must therefore correspond to the value J = j1 + j2 − 2.
Proceeding in this fashion we see that the successive J values differ by one. How
are we sure that all the J values have been correctly identified ?. We will now count
the number of states in two different ways to confirm the conclusion that J takes
all the values in the allowed range from |j1 − j2 | to (j1 + j2 ).

Count in two ways to cross check the results

Thus we have two set of orthonormal bases
n  o

|j1 j2 ; m1 m2 im1 = −j1 , · · · , j1 , m2 = −j2 , · · · , j2 (1.29)

and n  o

|j1 j2 ; JMiJ = j1 − j2 , · · · , j1 + j2 , M = −J, · · · , J (1.30)
It is easily seen that the total number of vectors in two bases are equal. The
number of elements in the set (1.29) is (2j1 + 1)(2j2 + 1) and in the set (1.30) is also
the same
j1 +j2
X
(2J + 1) = (2j1 + 1)(2j2 + 1). (1.31)
J=|j1 −j2 |

The two bases in Eq.(1.29) and Eq.(1.30) are orthonormal and hence the transforma-
tion connecting the two must be a unitary transformation and the Clebsch Gordon
coefficients must satisfy the relations
X
hj1 j2 ; m1 m2 |j1 j2 JMi hJM|j1 j2 ; m′1 m′2 i = δm1 m′1 δm2 m′2 (1.32)
JM
X
hJM|j1 j2 ; m1 m2 i hj1 j2 ; m1 m2 |j1 j2 J ′ M ′ i = δJJ ′ δM M ′ (1.33)
m1 ,m2

These relations can also be seen as a consequence of the completeness formula such
as X
|JMi hJM| = Iˆ (1.34)
JM

and a similar relation for the vectors |j1 j2 ; m1 m2 i.

The Clebsch Gordon coefficients for addition of angular momenta are tabulated.
Two such tables for j2 = 21 and j2 = 1 are given on the next page.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
13002;

10

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §4 Tables of Clebsch Gordon coefficients

TABLE OF CLEBSCH GORDON COEFFICIENTS

< j1 j2 = 21 ; m1 m2 |JM >

1
m2 = 2
m2 = − 21

s s
1 1
1
j1 + M + 2
j1 − M + 2
j = j1 + 2
2j1 + 1 2j1 + 1
s s
1 1
1
j1 − M + 2
j1 + M + 2
j = j1 − 2
−
2j1 + 1 2j1 + 1

< j1 j2 = 1; m1 , m2 |JM >

m2 = 1 m2 = 0 m2 = −1

s s s
(j1 + M)(j1 + M + 1) (j1 − M + 1)(j1 + M + 1) (j1 − M)(j1 − M + 1)
j = j1 + 1
(2j1 + 1)(2j1 + 2) (2j1 + 1)(j1 + 1) (2j1 + 1)(2j1 + 2)
s s
(j1 + M)(j1 − M + 1) M (j1 − M)(j1 + M + 1)
j = j1 − p
2j1 (j1 + 1) j1 (j1 + 1) 2j1 (j1 + 1)
s s s
(j1 − M)(j1 − M + 1) (j1 − M)(j1 + M) (j1 + M)(j1 + M + 1)
j = j1 − 1 −
2j1 (2j1 + 1) j1 (2j1 + 1) 2j1 (2j1 + 1)

11

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §5 Restrictions on CG Coefficients
QM-
Lec-
13003
⊲
We shall now give proof of the statement that all CG coefficients hJM|; m1 m2 i can
be determined in terms of the coefficient for the highest value M = J, m1 = j1 . In
this connection the following results from the theory of angular momentum derived
earlier will be useful.
p
J± |JMi = (J ∓ M)(J ± M + 1)|J(M ± 1)i (1.35)
(1) (2)
p
(J± + J± )|j1 j2 ; m1 m2 i = (j1 ∓ m1 )(j1 ± m1 + 1)|j1 j2 ; (m1 ± 1)m2 i
p
+ (j2 ∓ m2 )(j2 ± m2 + 1)|j1 j2 ; m1 (m2 ± 1)i(1.36)

On taking conjugate of Eq.(1.36) we get

(1) (2)
p
hj1 j2 ; m1 m2 |(J∓ + J∓ ) = hj1 j2 ; m1 ± 1m2 | (j1 ∓ m1 )(j1 ± m1 + 1)
p
+ hj1 j2 ; m1m2 ± 1| (j2 ∓ m2 )(j2 ± m2 + 1)(1.37)

which is a consequence of the angular momentum commutation relations. Consid-

ering the matrix element
(1) (2)
hj1 j2 ; m1 m2 |J± |JMi = hj1 j2 ; m1 m2 |J± + J± |JMi (1.38)

and using Eq.(1.35) and Eq.(1.37) we get two relations one for J+ and
p
(J − M)(J + M + 1)hj1 j2 ; m1 m2 |JM + 1i
p
= hj1 j2 ; m1 − 1m2 |JMi (j1 + m1 )(j1 − m1 + 1)
p
+hj1 j2 ; m1 m2 − 1|JMi (j2 + m2 )(j2 − m2 + 1) (1.39)

and a second relation for J−

p
(J + M)(J − M + 1)hj1 j2 ; m1 m2 |J(M − 1)i
p
= hj1 j2 ; (m1 + 1)m2 |JMi (j1 − m1 )(j1 + m1 + 1)
p
+hj1 j2 ; m1 (m2 + 1)|JMi (j2 − m2 )(j2 + m2 + 1) (1.40)

We will make repeated use of the results Eq.(1.39),Eq.(1.40) given above. These
equations can be used successively with M = J, J − 1, . . . to compute the Clebsch
Gordon coefficients. The restrictions ⊠1) − ⊠4), given on page 8, on the allowed
values of the total angular momentum J will be derived by considering the matrix
elements hj1 j2 ; m1 m2 |Jz |JMi,hj1 j2 ; m1 m2 |Jz |JMi and hj1 j2 ; m1 m2 |J± |JMi and by
repeated use of Eq.(1.39) and Eq.(1.40) . We shall suppress j1 j2 and use the
(1) (2)
shorthand notation |; m1 m2 i to denote the eigenvectors |j1 j2 m1 m2 i of Jz , Jz . On
the other hand the eigenvectors of Jz , |j1 j2 JMi, will be written as |JMi.

CG Coefficient hJM|j1 j2 ; m1 m2 i is zero unless M = m1 + m2

We have already seen that the vector |; m1 m2 i is also an eigenvector of Jz with
eigenvalue m1 + m2 . Since the eigenvectors of a hermitian operator for different
eigenvalues are orthogonal and since Jz is hermitian, it follows that hJM|; m1 m2 i

12

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 will be zero unless M = m1 + m2 . This result can also be seen as follows. Since
(1) (2)
Jz = Jz + Jz , taking the matrix element and using the properties

Jz |JMi = M~|JMi ⇒ hJM|Jz = hJM|M~ (1.41)

(Jz(1) + Jz(2) )|; m1 m2 i = (m1 + m2 )~|; m1 m2 i (1.42)

we obtain
hJM|Jz(1) + Jz(2) − Jz |j1 j2 ; m1 m2 i = 0 (1.43)
Therefore,
(m1 + m2 − M)hJM|j1 j2 ; m1 m2 i = 0 (1.44)
Thus if M 6= m1 + m2 , the Clebsch Gordon coefficient hJM|j1 j2 ; m1 m2 i has to be
zero. In other words, a nonzero value of hJM|j1 j2 ; m1 m2 i is possible only when

M = m1 + m2 (1.45)

Recurrence relations for CG coefficients

The results will be derived by considering the matrix elements hj1 j2 ; m1 m2 |Jz |JMi
and hj1 j2 m1 m2 |J± |JMi.
Note that there is one relation between three the variables m1 m2 , M. Hence we
need the Clebsch Gordon coefficients for all allowed values of M and m1 which vary
in the range M = −J, · · · , J and m1 = −j1 , · · · , j1 . We will now argue that these
can all be related to single coefficient hJJ|; j1 J − j1 i . They can all be related to
h; j1 (J − j1 )|JJi

Use Eq.(1.39) with M = J, m1 = j1

0 × h; m1 m2 |J(M + 1)i
p
= h; (j1 − 1)m2 |JJi 2j1
p
+h; j1 (m2 − 1)|JMi (j2 + m2 )(j2 − m2 + 1) (1.46)

Use Eq.(1.40) with M = J, m1 = j1

√
2Jh; j1 m2 |J(J − 1)i
= h; (j1 + 1)m2 |JMi × 0
p
+h; j1 (m2 + 1)|JMi (j2 − m2 )(j2 + m2 + 1) (1.47)

Use Eq.(1.39) with M = J − 1, m1 = j1

√
2Jh; j1 m2 |JJi
p
= h; (j1 − 1)m2 |J(J − 1)i 2j1
p
+h; j1 (m2 − 1)|J(J − 1)i (j2 + m2 )(j2 − m+ 1) (1.48)

13

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 UseEq.(1.40) with M = J, m1 = j1 − 1
√
2Jh; (j1 − 1)m2 |J(J − 1)i
= h; j1 m2 |JJi × 0
p
+h; (j1 − 1)m2 + 1|JMi (j2 − m2 )(j2 + m2 + 1) (1.49)

Use Eq.(1.40) with M = J − 1, m1 = j1

p
(J + M)(J − M + 1)hj1 j2 ; m1 m2 |J(J − 2)i
p
= hj1 j2 ; (j1 + 1)m2 |JMi (j1 − m1 )(j1 + m1 + 1)
p
+hj1 j2 ; j1 (m2 + 1)|JMi (j2 − m2 )(j2 + m+ 1) (1.50)

We can continue in this fashion. We see that the Clebsch Gordon coefficient for
different pairs of values of M, m1 are known in terms of a single coefficient for
M = J, m1 = ji as follows

Equation New Coefficient Known in terms of

Eq.(1.46) m1 = j1 − 1, M = J m1 = j1 , M = J

Eq.(1.47) m1 = j1 , M = J − 1 m1 = j1 , M = J

Eq.(1.48) m1 = j1 − 1, M = J − 1 m1 = j1 , M = J − 1 and m1 = j1 − 1, M = J

Eq.(1.49) m1 = j1 − 1, M = J − 1 m1 = j1 − 1, M = J

Eq.(1.50) m1 = j1 , M = J − 2 m = j1 , M = J − 1

Next we consider the state |j1 j2 ; m1 m2 i with m1 = j1 and m2 = j2 . In this state

M has the highest value j1 + j2 . All these coefficients can be fixed in terms of a
single coefficient hj1 j2 ; j1 m2 = J − j1 |JMi which is nonzero only if J − j1 lies in
between −j1 and j1 . Thus
−j2 ≤ J − j1 ≤ j2 (1.51)
By repeating the above steps with j1 and j2 interchanged we would get

−j1 ≤ J − j2 ≤ j1 (1.52)

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-Lec-

13003;

14

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §6 Examples of Adding Angular Momenta
QM-
Lec-
13004
⊲
§ 6.1 Example
We give a few short examples on addition of angular momenta.
a) If j1 = 21 , j2 = 3
2 the total angular momentum can have values J = 12 , 32

b) If We add angular momenta j1 and j2 = 1, the possible values of the total angular
momentum are j1 − 1, j1 , and j1 + 1.

c) The minimum value of j which when added to itself will give rise to a given value J
for the total angular momentum is J2 .
Explicit values of the Clebsch Gordon coefficients can be obtained by following
the procedure in § 5. This is not the only procedure available to get the values of ←
the Clebsch Gordon coefficients. In the solved problem below we show, by means
of an example of adding angular momenta j1 = 1 and j2 = 1/2, how the Clebsch
Gordon coefficients could be obtained using the properties of angular momentum
ladder operators and orthogonality of states with different values of J.

§ 6.2 Problem.
Enumerate all possible states with definite JM values for the case j1 = 1, j2 = 21 .
Using properties of J− and orthogonality, obtain different states |JMi in terms of
the states |j1 j2 ; m1 m2 i.

Solution : Let |1 12 ; m1 m2 i denote the state |1m1 i| 21 m2 i eigenstate of Jz(1) and Jz(2) and
|JM i will be used to denote the states with definite values of total angular momentum.
This completes the construction of the states |JM i for J = 32 .
1
1 Since j1 = 1 and j2 = 2 the only possible values of J are 12 , 32 .

2 The state, |1 12 ; 1 12 i, with highest values for m1 , m2 , corresponds to the highest value
of both J and M . Thus
| 23 32 i = |1 21 ; 1 12 i (1.53)
(1) (2)
3 Next apply J− = J− + J− on Eq.(1.53) and use Eq.(1.35) and Eq.(1.36) to get
q √ q
3 5 3 1 31 1 1 1 3 1 1 1 1
·
2 2 − · |
2 2 22 i = 1.2 − 1.0|1 2 ; 0, 2 i + 2 · 2 − 2 · (− 2 )|1 2 ; 1, − 2 i (1.54)
We thus arrive at q q
| 32 21 i = 2 1
3 |; 0, 2 i +
1 1
3 |; 1, − 2 i (1.55)
(1) (2)
4 Applying J− = J− + J− on Eq.(1.55) again we get the state with total M = − 21 .
hq q i
(1) (2)

J− | 32 12 i = J− + J− 2
3 |; 0, 1
2 i + 1
3 |; 1, − 1
2 i
hq q i
(1) 2 1 1 1
= J− 3 |; 0, 2 i + 3 |; 1, − 2 i
hq q 
(2) 2 1 1 1
+J− 3 |; 0, 2 i + 3 |; 1, − 2 i (1.56)

15

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Using Eq.(1.35) and Eq.(1.36) we get
q q √ q √
3 5 1
2 · 2 − 2 · (− 21 )| 32 , − 12 i = 2
3 1.2 − 0|; −1 1
2 i + 1 1
3 1.2 − 0|; 0, − 2 i
q q
+ 23 12 · 32 − 12 · (− 21 )|; 0, − 12 i + 0 (1.57)

where the last term in Eq.(1.56) vanishes because the minimum value of m2 is − 12 .
On simplifying we get
q q
| 32 , − 21 i = 13 |; −1, 21 i + 23 |; 0, − 12 i (1.58)

5 Repeating the above procedure of applying J− we would get

| 32 , − 32 i = |; −1, − 21 i (1.59)

The above relation can also be written down directly because for the state with J =
M = − 32 would involve only one combination of m1 , m2 , viz., m1 = −1, m2 = − 21
and hence by normalisation we must have Eq.(1.59).

6 We now consider the two states M = 12 for J = 12 . The state | 12 21 i would be linear
combination of the states with (m1 , m2 ) values given by (0, 12 ) and (1, − 21 ) and hence

| 12 12 i = α|; 0, 21 i + β|; 1, − 12 i (1.60)

The same values of (m1 , m2 ) also appear in the state J = 32 , Eq.(1.55). The states
in Eq.(1.55) and Eq.(1.60) being states with different total angular momenta should
be orthogonal. Therefore orthogonality gives
q q
2 1
3α + 3β = 0 (1.61)

This equation together with normalisation condition α2 + β 2 = 1 determines α and

β and we get q q
| 12 21 i = 13 |; 1, − 12 i − 23 |; 0, 12 i (1.62)

7 Applying J− on Eq.(1.62) gives the expression for the state | 12 , − 21 i

q q
| 21 , − 21 i =− 2 1
3 |; −1, 2 i +
1 1
3 |; 0, − 2 i (1.63)

It is easily verified that state is orthogonal to Eq.(1.58).

We shall now take up an example of constructing the states |JMi using the tables
of Clebsch Gordon coefficients. There two tables of Clebsch Gordon coefficients are
given on page 11. Note that the first table is for j2 = 21 and the second one for
j2 = 1. For j2 = 12 , the two columns correspond to the two values, 21 and − 12 , of
m2 . The two rows correspond to the tow possible values of total angular momentum
J = j1 + 12 and J = j1 − 21 . Similarly the second table, corresponding to j2 = 1, has
three columns for the three values m2 = 1, 0, −1 and the three rows correspond to
three allowed values J = j1 + 1, j1 , j1 − 1.

16

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 § 6.3 Problem.
Construct all possible states with J = 25 , 23 in terms of states with j1 = 2, j2 = 1
2
using the table of Clebsch Gordon coefficients.

1 5
Solution : Since j2 = 2the first table is needed here. The allowed values of J are 2
3 5 3
and 2. For all J = states we must use the first row of the first table and for J =
2 2
states the second row should be used.
1 The state with highest values J = 52 and M = 25 can be written down directly as
there is only one possible set of values (m1 , m2 ) = (2 12 ) is allowed. Hence

| 25 52 i = |; 2, 21 i (1.64)

5 3
2 For the next state with J = 2 and M = 2 the possible values of (m1 , m2 ) are (1, 12 )
and (2, − 12 ) and hence

| 25 , 32 i = α1 |; 1, 21 i + α2 |; 2, − 12 i (1.65)

and the coefficients α1 , α2 are to be read from the first row.

s s r
1
j1 + M + 2 2 + 32 + 12 4
α1 = = = ; (1.66)
2j1 + 1 2.2 + 1 5
s s r
j1 − M + 12 2 − 32 + 12 1
α2 = = = (1.67)
2j1 + 1 2.2 + 1 5

5 1
3 For the next state with J = 2 and M = 2 the possible values of (m1 , m2 ) are (0, 12 )
and (1, − 12 ) and hence

| 25 , 32 i = α3 |; 0, 21 i + α4 |; 1, − 12 i (1.68)

and the coefficients α1 , α2 are again to be read from the first row.
s s r
j1 + M + 12 2 + 12 + 12 3
α3 = = = ; (1.69)
2j1 + 1 2.2 + 1 5
s s r
1
j1 − M + 2 2 − 12 + 12 2
α4 = = = (1.70)
2j1 + 1 2.2 + 1 5

4 For the next state with J = 52 and M = − 21 the possible values of (m1 , m2 ) are
(−1, 12 ) and (0, − 21 ) and hence

| 52 , − 12 i = α5 |; −1, 21 i + α6 |; 0, − 12 i (1.71)

and the coefficients α1 , α2 are again to be read from the first row.
s s r
j1 + M + 12 2 − 12 + 12 2
α5 = = = ; (1.72)
2j1 + 1 2.2 + 1 5
s s r
j1 − M + 12 2 + 12 + 12 3
α6 = = = (1.73)
2j1 + 1 2.2 + 1 5

17

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 5 Next for state with J = 52 and M = − 23 the possible values of (m1 , m2 ) are (−2, 21 )
and (−1, − 12 ) and hence

| 25 , − 32 i = α7 |; −2, 21 i + α8 |; −1, − 12 i (1.74)

and the coefficients α1 , α2 are again to be read from the first row.
s s r
1
j1 + M + 2 2 − 32 + 12 1
α7 = = = ; (1.75)
2j1 + 1 2.2 + 1 5
s s r
j1 + M − 12 2 + 32 + 12 4
α8 = = = (1.76)
2j1 + 1 2.2 + 1 5

6 Finally the state | 52 , − 52 i has the highest value of J and the lowest value of M , and
easily written as
| 25 , − 52 i = |; −2, − 21 i (1.77)

3
7 Now we take up the construction of states with J = 2 and M = 32 , 12 , − 12 , − 23 we
write

| 32 , 3
2i = β1 |; 1, 21 i + β2 |; 2, − 21 i (1.78)
3 1
|2, 2i = β3 |; 0, 21 i + β4 |; 1, − 21 i (1.79)
| 2 , − 12 i
3
= β5 |; −1, 12 i + β6 |; 0, − 21 i (1.80)
| 32 , − 32 i = β7 |; −2, 12 i + β8 |; −1, − 21 i (1.81)
(1.82)

where β1 , β3 , β5 , β7 are to be taken from the entries in the second row, first column
and β2 , β4 , β6 , β8 are to read from the second row, second column of the first table
with appropriate values of M . Thus
s  r s  r
 
j1 − M + 12  1 j1 + M + 21  4
β1 = −  =− ; β2 =  = (1.83)
2j1 + 1  5 2j1 + 1  5
3 3
M= 2 M= 2
s  r s  r
 
j1 − M + 12  2 j1 + M + 21  3
β3 = −  =− ; β4 =  = (1.84)
2j1 + 1  5 2j1 + 1  5
1 1
M= 2 M= 2
s  r s  r
 
j1 − M + 12  3 j1 + M + 21  2
β5 = −  =− ; β6 =  = (1.85)
2j1 + 1  5 2j1 + 1  5
1 1
M =− 2 M =− 2
s  r s  r
1  1 
j1 + M + 2  4 j1 − M + 2  1
β7 = − =− ; β8 = = (1.86)
2j1 + 1  5 2j1 + 1  5
3 3
M =− 2 M =− 2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-Lec-

13004;

18

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 2

Spin and Identical Particles

§1 Spin as a Dynamical Variable

QM-
Lec-
14001
⊲
For a classical particle the angular momentum, given by ~r × ~p, is zero when the
particle is at rest.In order to explain anomalous Zeeman effect it was suggested by
Goudsmidt and Uhlenbeck that electron possesses angular momentum at rest whose
component in any fixed direction can take one of the two values 21 ~ or − 21 ~. An
electron has a magnetic moment, associated with spin, equal to one negative Bohr
magneton, given by
e ~
~µ = − S
mc
Many elementary particles are found to have angular momentum at rest. This
angular momentum is called spin.
Spin is an observable associated with all the fundamental particles, having the
same properties as the angular momentum. We therefore associate three hermitian
operators Sx , Sy , and Sz with spin angular momentum and assume that they satisfy
angular momentum algebra.

[Sx , Sy ] = i~Sz , (2.1)

[Sy , Sz ] = i~Sx , (2.2)
[Sz , Sx ] = i~Sy . (2.3)

This algebra imples that the operator S ~ 2 = S 2 + S 2 + S 2 commutes with all the
x y z
three components of spin. Since different components of spin do not commute,
a commuting set of operators consist of S ~ 2 and component of the spin along any
one direction; most popular choice being S ~ 2 and Sz . Because the algebra of spin
operators is assumed to be the same as the angular momentum operators the results
on angular momentum apply to the spin also.We therefore have the following results:
~ 2 are given by s(s + 1)~2 , where s is a positive integer or
• The eigenvalues of S
half integer.

• For a given value of s, the eigenvalues of Sz are s, s − 1, s − 2, . . . , −s.

• A particle will be said to have spin s if the the maximum allowed value of Sz
is s~, which is same as S~ 2 having value s(s + 1)~2 .

19

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The operators S~ 2 and Sz commute and hence they have a complete set of simultane-
~ 2 and Sz will be denoted by |smi
ous eigenvectors. A simultaneous eigenvector of S
which will have the properties
~ 2 |smi = s(s + 1)~2 |smi
S (2.4)
~z |smi = m~|smi
S (2.5)

In all there are (2s + 1) values of m ranging from −s to s and therefore (2s + 1)
eigenvectors |smi. The vector space needed to describe spin is linear span of all the
vectors |smi and is (2s + 1) dimensional.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
14001;

§2 Spin Wave Function and Spin Operators

QM-
Lec-
14002
⊲
Representation of Spin Wave Function
In order to describe the spin degrees of freedom, it is convenient to introduce a
representation. For this we need to select a ccomplete commuting set of hermitian
operators and construct an orthonormal basis from their simulataneous eignevectors.
~ 2 and Sz . In order to proceed further, we want to work
A suitable set consists of S
with an explicit representation of the spin. We arrange
 the eigenvectors |s, mi in
descending order in m to get o.n. basis |s, mim = s, s − 1, · · · , −s + 1, −s ,(
WHY are these vectors orthogonal?). An arbitrary state vector |xi is then a linear
combination of the basis elements
−s
X
|xi = cm |smi (2.6)
m=s

The interpretation of the numbers cm is that square of its modulus, |cm |2 , gives the
probability that Sz will have the corresponding value m~. Following the convention
of arranging the basis vectors in the order of decreasing values for the spin projection
m, the |χi will be represented by a (2s + 1) component column vector of the form
 
α1
 α2 
 
|χi =  ..  (2.7)
 . 
α(2s+1) .

The spin wave function of a particle with spin s is a column vector with (2s + 1)
components.

Representation of Spin Operators

As there are (2s + 1) basis vectors in the vector space needed to describe spin, the
~ will be represented by matrices with (2s + 1) rows and (2s + 1)
spin operators S

20

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 columns. First of all, the matrix for Sz will be diagonal matrix with eigenvalues of
Sz appearing along the main diagonal.
 
s 0 0 ··· ··· 0
 0 s−1 0 ··· ··· 0 
 
 0 0 s−2 ··· ··· 0 
Sz = ~   (2.8)
0 0 0 · · · · · · 
 
· · · · · · ··· ··· ··· 0 
0 0 0 · · · · · · −s

This follows from the fact that the basis vectors |smi are eigenvectors of Sz with
eigenvalues m~. (WHY?) The matrices for Sx and Sy are found by first obtaining
the matrices for S± and using Sx = 12 (S+ + S− ) and −i2
(S+ − S− ). To construct
these matrices one needs to know the matrix elements hs, m′ |S± |s, mi which can be
computed by making use of the result
p
S± |s, mi = s(s + 1) − m(m ± 1) ~ |s, m ± 1i (2.9)

We shall give the answer for spin 12 and spin 1 matrices. The spin half matrices
are related to the Pauli matrices σx , σy , σz and are given by

~ = ~ ~σ
S (2.10)
2
This result is derived in the solved problem below. The corresponding result for the
spin one matrices
 is     
0 1 0 0 −i 0 1 0 0
Sx = √~2 1 0 1 ; Sy = √~2  i 0 −i ; Sz = ~ 0 0 0  .
0 1 0 0 i 0 0 0 −1

qm-
and is left as en exercise for the reader.
ymp-
20001
⊲ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-Lec-
14002;

§ 2.1 Problem.
Construct the spin matrices for a spin half particle.

1
Solution : For a spinparticle,
n 2let |o21 i and | − 21 i denote the eigenvectors of Sz with
eigenvalues ± 12 .In the basis | 12 i, | − 12 i the matrix for Sz is
 
~ 1 0
Sz = . (2.11)
2 0 −1

( WHY?) Next we use Eq.(2.9) for s = 21 , m = ± 12 to get

S+ | 12 i = 0 ⇒ h 21 |S+ | 12 i = 0, h 21 |S+ | − 12 i = 0 (2.12)

1
S+ | − 2i = | 12 i ⇒ h− 21 |S+ | 21 i = 1, h− 12 |S+ | − 1
2i =0 (2.13)

21

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Therefore the matrix !
h 12 |S+ | 21 i h 12 |S+ | − 21 i
, (2.14)
h− 12 |S+ | 21 i h− 21 |S+ | − 12 i
for S+ , becomes  
~ 0 1
S+ = (2.15)
2 0 0
The matrix for S− is hermitian adjoint of S+ and hence we get
 
~ 0 0
S− = (2.16)
2 1 0
The matrices for, Sx , Sy , are therefore given by
 
~ 0 1
Sx = (2.17)
2 1 0
 
~ 0 −i
Sy = (2.18)
2 i 0

To summarize we get the result that the spin 21 matrices are given by ~
2 times the Pauli
matrices:
S~ = ~ ~σ . (2.19)
2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ qm-ymp-

20001;

§3 Pauli Matrices
⊲
Pauli-
Matrices
We summarize some important properties of Pauli Matrices.
1. The three Pauli matrices are given by
     
0 1 0 −i 1 0
σ1 = , σ2 = , σ3 = . (2.20)
1 0 i 0 0 −1

2. The Pauli matrices satisfy the commutation relations:

[σ1 , σ2 ] = 2iσ3 , [σ2 , σ3 ] = 2iσ1 , [σ2 , σ3 ] = 2iσ1 . (2.21)
Frequently the above relations are written in a compact form as
[σi , σj ] = 2iǫijk σk . (2.22)

3. The square of each Pauli matrix is unity, so is the square of n̂ · ~σ where n̂ is a

unit vector:
ˆ
σ12 = σ22 = σ32 = I; ˆ
(n̂ · ~σ )2 = I. (2.23)
4. Every Pauli matrix anticommutes with the other two Pauli matrices.
σi σj + σj σi = 0, i 6= j. (2.24)

There does not exist a nonzero 2 × 2 matrix which anticommutes with all the
three Pauli matrices.

22

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 5. The above relations can be written in various different forms.
σj σk − σk σj = 2iǫijk σk (2.25)
σj σk + σk σj = 2δjk (2.26)

6. The above two relations imply that

σj σk = δjk + iǫjkℓ σℓ . (2.27)

7. The above statements can be rewritten as

(a) [~a · ~σ , ~b · ~σ ] = 2i(~a × ~b) · ~σ
(b) (~a · ~σ )2 = |~a|2
(c) (~a · ~σ )(~b · ~σ ) + (~b · ~σ )(~a · ~σ ) = 2(~a · ~b)Iˆ
(d) (~a · ~σ )(~b · ~σ ) = (~a · ~b)Iˆ + i(~a × ~b) · ~σ

where ~a, ~b are two arbitrary numerical vectors.

8. The trace of each of the three matrices is zero. If we use the notation σ0 = Iˆ
we have the relation, we can write
T r(σµ σν ) = 2δµν (2.28)
where µ, ν take values 0, 1, 2, 3.
9. The above identity can be used to prove linear independence of Pauli matrices.
The four matrices σµ , µ = 0, .., 3 form a basis in the complex vector space of
all 2 × 2 matrices.
10. If complex matrix S is written as a linear combination of 2 × 2 basis matrices
σµ in the form
X3
S= Cµ σµ , (2.29)
µ=0

then the expansion coefficients are given by

1
Cµ = T r(Sσµ ). (2.30)
2
P
11. The completeness relation, n |nihn| = I, for the Pauli matrices is becomes
the identity X
(σ a )ij (σ a )kl = 2δil δjk − δij δkl . (2.31)
a

12. An important identity satisfied by the Pauli matrices is

exp(i~
α · ~σ ) = cos |~
α| + i~
α · σ sin |~
α| (2.32)
where α
~ is a vector and
q
(α1 , α2 , α3 ), α| = α12 + α22 + α22
|~ (2.33)
This result can be proved by expanding the exponential in a power series.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ Pauli-
Matrices;

23

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §4 Total Wave Function of a Particle with Spin
QM-
Lec-
14003
⊲
We have so far discussed a quantum description of the spin degrees of freedom of
a particle. The other dynamical variables of a particle are the usual coordinates,
momenta, etc. It is assumed that the spin and position are independent of each
other and hence can be measured simultaneously. Similarly, spin and momentum
can be measured simultaneously. Thus the spin operators S ~ commute with position
operators and also with the momentum operators. In such a space a representation
could be chosen in which the basis vectors are simultaneous eigenvectors of S ~ 2 , Sz ,
and ~r operators. Denoting a basis vector as |~ri|smi, an arbitrary vector will have
an expansion
XZ
|ψi = dxCm~r |~ri|smi (2.34)
m
The coefficients Cm~r give the probability amplitude of position being ~r and Sz having
a value m. These coefficients are (2s + 1) component functions of ~r; by a change in
notation we write collect them into a (2s + 1) component wave function
 
ψ1 (~r)
 
Ψ(~r) = ψ2 (~r) . (2.35)
..
.
In this reperesentation the state of a particle with spin is described by a vector
in the vector space which is tensor product of a complex vector space of dimension
(2s + 1) and the space of square integrable functions. So a particle with spin 21 , such
as an electron, is described by a two component wave function
 
ψ1 (~r)
Ψ(~r) = (2.36)
ψ2 (~r)
The interpretation of the different components of Ψ is, that |ψ1 (~r)|2 d3 r gives the
probability of spin being up and position being between ~r and ~r + d~r. Similarly,
|ψ2 (~r)|2 d3 r gives the probability of spin being down and position being between ~r
and ~r + d~r. The normalisation condition now reads
Z Z  
† 3
Ψ (~r) Ψ(~r)d ~r = |ψ1 (~r)| + |ψ2 (~r)| d3~r = 1
2 2
(2.37)

Frequently, the total wave function factorises and assumes the form
Ψ(~r) = ψ(~r) × χ (2.38)
where χ is a column vector with (2s + 1) components, so for an electron we will have
 
α
χ= . (2.39)
β
In case the wave function can be written in the product form, we shall refer to ψ(~r)
as the space part of the wave function and the column vector χ as the spin part of
the wave function. The function ψ(~r) describes the translational degrees of freedom,
as usual, and the spin degrees of freedom are described by the column vector χ.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-20-01-
L3;

24

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §5 Indistinguishability in Quantum Mechanics
QM-
Lec-
14004
⊲
In transition form classical to quantum mechanics one is forced to revise, or com-
pletely give up, many classical concepts. Some examples are

- Superposition principle is new for particles.

- Precise values for observable is given up, in general only probabilities can be
computed.

- Not all observable can be measured simultaneously. Only those represented

by a simultaneous commuting set can be measured simultaneously.

We shall see that for a system of two or more identical particles, the identity of
individual particles looses its meaning in quantum mechanics; one can only talk
about the system as a whole. We also need to introduce a new hypothesis about
the states of a quantum mechanical system of several identical particles.
Consider an electron electron scattering experiment. Classically it is always
possible to distinguish two electrons because they are point particles having well
defined positions. More importantly, this distinction can always be maintained
at all times because each particle has a well defined trajectories. By successive
measurements of its position at regular intervals very accurately the path of each
electron can, in principle, be followed without disturbing their motion. In fact
classical mechanics admits the possibility of predicting the position of a particle
at all times knowing initial conditions and interactions.classical mechanics admits
the possibility of distinguishing and maintaining the distinction between the two
electrons at all times.
For a quantum mechanical system consisting of several identical particles we note
the following.

• In the first place the particles cannot have well defined positions, only some
probabilities can be assigned to different values of the position.

• At any given time, it is certainly possible to localise each particle with great
accuracy. However, in general, the wave packets spread and it will not be
possible to maintain the localisation at later times. Thus when the particles
come very close we would not be able to tell which particle was which one.

• The trajectories of particles do not have any meaning in quantum mechanics.

One may attempt to follow the motion in quantum mechanics by measuring
their positions very accurately at short intervals; but this exercise turns out
to be useless for all practical purposes because it ’disturbs’ the motion of the
particles.

It is imporatnt to realise that inability to distinguish identical nature of particles

is not due to some technical, theoretical or experimental, reasons. Even in principle
a distinction between two identical particles cannot be maintained in any thought
experiment or any theoretical analysis. Nor is this conclusion avoidable by improv-
ing upon the measuring apparatus. This surprising conclusion is more due to the

25

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 structure of the quantum theory, rather than lack of a very accurate measuring ap-
paratus. In some sense the nature of the conclusion is very similar to the uncertainty
principle about impossibility of very precise simultaneous measurement of position
and momentum. We ,therefore, wish to conclude that the possibility of being able
to distinguish between two identical particles does not exist even in principle and
that we should accept that they are indistinguishable.
In view of the above discussion, we abandon the attempts to distinguish between
identical particles. We assume that in a quantum mechanical system of identical
particles it is not possible to distinguish between any two identical particles; the
individual particles loose their identity and we should refer to the system as a ’whole’.
This has far reaching consequences. for sake of definiteness, let us consider a sys-
tem of two electrons and let their wave function be ψ(ξ1 , ξ2 ), where ξ1 , ξ2 collectively
denote the space as well as the spin variables of the two electrons. Then ψ(ξ2 , ξ1 )
will denote the state of two electrons obtained by exchanging the two electrons. If
the two electrons are indistinguishable, this interchange of two electrons can have
no effect on the state of the system as whole and thus the wave functions ψ(ξ1 , ξ2 )
and ψ(ξ2 , ξ1 ) must represent the same state and hence we must have

ψ(ξ1 , ξ2 ) = exp(iα)ψ(ξ2 , ξ1 ) (2.40)

for some real α, as required by the first postulate. The above equation is valid for
all ξ1 , ξ2 and hence, replacing ξ1 with ξ2 and ξ2 with ξ1 we get

ψ(ξ2 , ξ1) = exp(iα)ψ(ξ1 , ξ2 ) (2.41)

= exp(iα)ψ(ξ2 , ξ3 ) (2.42)

and hence we conclude that

exp(2iα) = 1 ⇒ exp(iα) = ±1 (2.43)

Therefore, we arrive at an important conclusion that under exchange of all variables

the wave function of two identical particles must be symmetric or antisymmetric.

ψ(ξ2 , ξ1) = ±ψ(ξ1 , ξ2 ) (2.44)

The symmetric or the antisymmetric nature Eq.(2.44), if enforced at initial time,

will be preserved at all times. To see this let us introduce a permutation operator
P12 by
P12 ψ(ξ1 , ξ2 ) = ψ(ξ2 , ξ1 ) (2.45)
The Eq.(2.44) is just the statement that the state must be an eigenstate of the
permutation operator. The Hamiltonian for two identical particles will be symmet-
ric under exchange of ξ1 and ξ2 and hence commutes with P12 implying that the
permutation operator is a constant of motion. Thus the wave function will remain
an eigenfunction of P12 with the same eigenvalue at all times if it is chosen to be
eigenfunction at initial time.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
14004;

26

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §6 The Symmetrisation Postulate
QM-
Lec-
14005
⊲
For a system of two identical particles, the total wave function must be chosen to
be symmetric or anti-symmetric under exchange of the space and spin variables;
However, theoretical considerations alone, in the nonrelativistic quantum mechanics
do not help us in deciding which one, symmetric or antisymmetric wave function,
is the correct choice. for a given type of particle. An appeal to experiments does
give an answer which is contained in the statement of symmetrisation postulate.
Basically the symmetry property of the wave function is tied down to the spin of
the particle. The symmetrisation postulate states that

• For a system of two identical particles with integral spin bosons the total wave
function must be symmetric under simultaneous exchange of all the variables
such as the space and spin variables. For a system of two identical particles
of half integral spin fermions the full wave function must be anti-symmetric
under a simultaneous exchange of all the variables such as the space and spin
variables.
If ξ1 , ξ2 denote the set of all variables such as, space and spin, of two identical
particles. Then the symmetrisation postulate states that the total wave func-
tion ψ(ξ1 , ξ2 ) must be symmetric for bosons and antisymmetric for fermions
under an exchange of ξ1 and ξ2 .

ψ(ξ2 , ξ1) = +ψ(ξ1 , ξ2 ) ( boson) (2.46)

ψ(ξ2 , ξ1) = −ψ(ξ1 , ξ2) ( fermions) (2.47)

• The symmetrisation postulate for a system of n− identical particles sates that

the total wave function must be symmetric simultaneous under exchange of
variables ξj and ξk for every pair j, k, if the particles are bosons and the relation

ψ(ξ1 , ·ξj · · · , ξk , · · · , ξn ) = +ψ(ξ1 , ·ξk , · · · , ξj , · · · , ξn ) (2.48)

should hold for all pairs AK. Similarly, for a system of n− identical fermions
the total wave function must be anti-symmetric under simultaneous exchange
of variables ξj , ξk for every pair j, k

ψ(ξ1 , ·ξj · · · , ξk , · · · , ξn ) = −ψ(ξ1 , ·ξk , · · · , ξj , · · · , ξn ) (2.49)

For a system of several identical bosons, the total wave function Ψ(ξ1 , · · · , ξn )
remains unchanged under an arbitrary permutation of ξ1 , · · · , ξn ; where as for
fermions the wave function remains unchanged under an even permutation but
changes sign under an odd permutations.

We now give some explanatory remarks on the symmetrisation postulate.

• The postulate is a statement about the full wave function of the system of
identical particles under a simultaneous exchange of all the variables. For
example, there is no constraint on the space part (or the spin part) of the wave
function alone need.

27

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 • For composite systems such those consisting of both bosons and fermions, the
symmetry requirements hold for every pair of identical bosons and identical
fermions separately.

• For a system consisting of several ’particles’ which themselves could be bound

state of bosons and fermions, the postulate applies with spin interpreted to
mean the total angular momentum at rest.

• While for a system of two particles the symmetry property is restricted to

symmetry or antisymmetry alone, for a system of many identical particles
theoretical considerations allow existence of a variety of possibilities under
permutation of variables. These choices, known generally as ’para-statistics’,
do not seem to play any role for real physical systems.
Our discussion of the symmetrisation postulate will be incomplete if we don’t men-
tion that the spin statistics connection contained in the postulate was proved by
Pauli and Luders within the framework of relativistic quantum field theory under
very general assumptions such as relativistic invariance and positivity of the Hamil-
tonian.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
14005;

§7 Illustrative Examples
qm-
xmp-
20001
⊲
We shall now take up some examples of consequence of symmetrization postulate
for a system of two identical particles. We begin with two useful statements which
will be needed repeatedly in this connection.
⋆1) When we add two, equal, angular momenta j the possible resulting values are
J = 2j, 2j − 1, · · · , 0. Of these the state with the highest value, J = 2j, is
symmetric under an exchange of the two particles, the next one, with 2j − 1 is
antisymmetric; the states being alternately symmetric and antisymmetric as
J takes on the values in descending order.

⋆2) For a two particle system, the effect of an exchange of the positions of the
two particles is same as the parity on the wave function in the centre of mass
frame. Therefore under an exchange of the space variables the space part of
the wave function is symmetric for even ℓ and antisymmetric for odd ℓ.

§ 7.1 Example
Let us now consider two particles interacting via a spherically symmetric potential
v(r) where r is the distance between the particles. We write the total wave function
as
Ψtot (~r, ms1 , ms2 ) = ψspace (~r) ψspin (2.50)
If each particle has spin s the total spin has the values S = 2s, 2s − 1, 2s − 2, · · · 0
with symmetry properties as given by rule ⋆1). We shall see that for each value of
total spin S only odd, or even, angular momentum ℓ values will be permitted when
the particles are idemtical. We discuss four cases in the following.

28

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 a) When the two particle are not identical there is no restriction symmetry of the total
wave function. Therefore, all possible combination of ℓ and s values are permitted.

b) When the two particles are identical fermions, 2s = odd integer, the total wave
function must be antisymmetric under an exchange of space and spin variables.
Therefore, antisymmetric space wave functions ( ℓ = odd) must be chosen with
symmetric spin wave functions, (S = 2s, 2s − 2, · · · ). Also the space part of wave
function must be symmetric, (ℓ = even ) when the spin part of the wave function is
antisymmetric (S = 2s − 1, 2s − 3, · · · ).

c) When the two particles are identical bosons, s = integer, the total wave function
must be symmetric. Hence either both space and spin parts must be symmetric or
both must be antisymmetric. Thus even ℓ values will correspond to total spin S =
2s, 2s−2, · · · and odd ℓ values will correspond to the total spin S = 2s−1, 2s−3, · · · .

d) In the special case of two identical spin zero bosons, the total wave function is just
the space part alone which must be symmetric and hence ℓ must be even. This give
the statement that the spin and parity of a system of two identical spin zero bosons
must both be even.

e) All the cases of identical particles are summarized by saying that

(−1)ℓ+2S = 1 (2.51)

or ℓ + 2S must be even.

§ 7.2 Problem.
Three identical spin zero bosons exist in states described by normalized wave func-
tions χ, φ, ψ write the total wave function for the system.

Solution : The total wave function is obtained by symmetrization of the product

χ(~r1 )φ(~r2 )ψ(~r3 ). In all there will 3! terms obtained by permuting the postions ~r1 , ~r2 , ~r3 .
Thus
1 h i
Ψtot (~r1 , ~r2 , ~r3 ) = √ χ(~r1 )φ(~r2 )ψ(~r3 ) + 1 → 2 + 1 → 3 + 2 → 3 (2.52)
3!
1 n
= √ χ(~r1 )φ(~r2 )ψ(~r3 ) + χ(~r2 )φ(~r1 )ψ(~r3 ) + χ(~r3 )φ(~r2 )ψ(~r1 ()2.53)
3!
o
+χ(~r1 )φ(~r3 )ψ(~r2 ) + χ(~r3 )φ(~r1 )ψ(~r2 ) + χ(~r2 )φ(~r3 )ψ(~r1 ) (2.54)

√
The factor (1/ 6) has been put to ensure correct normalization.

§ 7.3 Problem.
For a system of three identical fermions in states ψ1 , ψ2 , ψ3 write the total wave
function.

29

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Solution : Let ξ1 , ξ2 , ξ3 collectively denote the space and spin variables of the three
particles. The total wave function should be completely antisymmetric and can be written
as a determinant, known as Slater determinant.
 
ψ (ξ ) ψ2 (ξ1 ) ψ3 (ξ1 )
1  1 1 
Ψ(ξ1 , ξ2 , ξ3 ) = √ ψ1 (ξ2 ) ψ2 (ξ2 ) ψ3 (ξ2 ) (2.55)
3! ψ (ξ ) ψ (ξ ) ψ (ξ )
1 3 2 3 3 3

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ qm-xmp-

20001;

§8 A First Look at He Atom Energy Levels

QM-
Lec-
13006
⊲
He atom : As an example of system of two identical particles we shall discuss He
atom. The Hamiltonian for He atom is given by

p21 p2 2e2 2e2 e2

H= + 2 − − +
2m 2m r1 r2 |~r1 − ~r2 |

If the electrostatic interaction, e2 /|~r1 −~r2 |, between the two electrons is neglected
as a first approximation, the hamiltonian becomes a sum of two hydrogen atom like
hamiltonians. In this approximation the electronic states are described by quan-
tum numbers (n1 , l1 , m1 ) and (n2 , l2 , m2 ) for the two electrons. Let u1 , u2 denote
corresponding H-atom wave functions. The space part of the wave function for the
two electrons will be product wave function u1 (~r1 )u2 (~r2 ), which must be properly
symmetrized or anti-symmetrized as discussed below.
In very many situations the total wave function is a product of a part describing
space properties and a spin wave function. Thus we write

Φtotal = ψspace (~r1 , ~r2 )χspin (m1 , m2 )

where m1 , m2 refer to the spin variables for the two electrons. As each electron
carries spin 1/2, the total spin can take values 1 (triplet) and 0 (singlet). The values
of total spin determines the symmetry property of spin wave function under an
exchange of spin variables. It is known that spin wave function must be symmetric
for S = 1 and antisymmetric for S = 0 states. The requirement that total wave
function be antisymmetric (for 2 electron systems) fixes the symmetry property of
the space part of the wave function as summarized in the table given below.

Spin State Total Spin Spin wave function Space wave function

Triplet S=1 Symmetric Antisymmetric

Singlet S=0 Antisymmetric Symmetric

30

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Therefore, out of the two combinations for the space wave function

1
ψ± (~r1 , ~r2 ) = √ (u1 (~r1 )u2 (~r2 ) ± u1 (~r2 )u2 (~r1 ))
2
the symmetric combination ψ+ should be used for the singlet states (S = 0) and the
antisymmetric combination ψ− should be used for triplet states (S = 1). The ground
state corresponds to n1 = n2 = 1 l1 = l2 = 0 m1 = m2 = 0 and the antisymmetric
combination ψ− vanishes. Only the symmetric combination is nonzero. Thus the
ground state is a singlet state; the same is true of all other states corresponding to
electrons having identical (n, l, m) quantum numbers.
When the two electron states correspond to different (n, l, m) quantum numbers,
both symmetric and antisymmetric combinations ψ± (r̄1 , r̄2 ) are possible. However,
the antisymmetric combination ψ± (~r1 , ~r2 ) vanishes when ~r1 = ~r2 . Therefore, the
probability that the two electrons will be found close to each other will be small
for ψ− ( for triplet states) and large for ψ+ (singlet states). Since the Coulomb
interaction between two electrons is positive and is large when their separation is
small, total Coulomb energy will be higher in singlet states as compared to its value
in the triplet state. These predictions are in accordance with the results on energy
levels of He atom derived from the spectrum of He atom.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
13006;

31

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 3

Variation Method

QM-
§1 Variation Method
21-
01- ⊲
L1 Theorem 1. Let E0 , E1 , E2 , · · · be the exact energy eigenvalues of Hamiltonian of
a system corresponding to the ground state, the first excited state, the second excited
state etc., respectively,
Huα = Eα uα (3.1)
where uα are eigenfunctions corresponding to eigenvalue Eα . Let ψ be any square
integrable function then we have the following results.
R ∗
ψ Hψd2 x
Eα ≤ R ∗ 3 (3.2)
ψ ψd x

Proof: Let ψ be expanded as X

ψ= Cn u n (3.3)
Then Z X
(ψ, ψ) = ψ ∗ ψd3 x = |Cn |2 (3.4)
n

and
Z !
X X
ψ ∗ Hψd3x = Cn u n , H Cn u m
n
XX
= Cn∗ Cm (un , Hum )
n m
XX
= Cn∗ Cm Em (un , Hum)
n m
XX
= Cn∗ Cm δmn Em
n m
X
= En |Cn |2 (3.5)
n=o

32

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Consider
Z Z ∞
X ∞
X
∗ 3 3 ∗ 2
ψ Hψd x − E0 d xψ ψ = En |Cn | − E0 |Cn |2 (3.6)
n=o n=0
∞
X
= (En − E0 )|Cn |2 (3.7)
n=1

The right hand side is positive because En − E0 > 0, for all n 6= 0. Hence
Z Z
∴ ψ Hψd x − E0 ψ ∗ ψd3 x ≥ 0 .
∗ 3
(3.8)

or Z Z
∗ 3
ψ Hψd x ≥ E0 ψ ∗ ψd3 x . (3.9)
or R ∗
ψ Hψd3 x
R ≥ E0 . (3.10)
ψ ∗ ψd3 x
Thus we have R ∗
ψ Hψd3 x
E0 ≤ R ∗ 3 (3.11)
ψ ψd x
for all square integrable ψ.

§ 1.1 Ritz variation method

The above result can be used to estimate the ground state energy as follows

• First we choose a trial wave function ψ which satisfies all general properties
of bound states in one dimension. First requirement is that the trial function
chosen must be square integrable. In addition the trial function must satisfy
usual continuity conditions. Thus, for the case of a potential which is finite
everywhere, the trial function and its derivative should be continuous. If the
potential has a infinite jump at a point, the trial wave function must be se-
lected to have a zero at that point. In accordance with the general results the
trial wave function for the ground state must not have any node. The trial
wave function must have even parity, in case the potential is an even function
of x.

R
• Normalise the trial function to unity ψ ∗ ψd3 x = 1.

• Compute Z
Eψ ≡ ψ ∗ Hψd3 x (3.12)

The trial wave function ψ will contain some unknown parameter(s). The
unknown parameters are then varied and one we demand that the average

33

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 value of energy be minimum. In case of only one unknown parameter, say α,
α is determined by the requirement that Eψ be minimum

∂Eψ
=0 (3.13)
∂α

The value α0 , for which Eψ is minimum, can now be used for calculating Eψ
which gives an estimate for the ground state energy. A good choice of trial
wave function can lead to very good estimate for the ground state energy. The
variation method has been successfully applied to the energy for ground state
of He atom. The variation method is most useful for the ground state energy
although it can be modified to estimate energies of excited states also.

• It can be easily seen that the variation method will give an upper estimate of
nth excited state, if an additional requirement that of the trial wave function
being orthogonal to the wave functions for all the states of lower energy levels
is imposed. So for example, for a potential which is an even function of x, use
of an odd function of x as a trial wave function, can be used to get an estimate
of first excited state energy. Due to this requirement thus the variation method
is not suitable for higher excited states.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-21-01-

L1;

34

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 4

Time Independent Perturbation

Theory

§1 Time Independent Problems

QM-
Lec-
16001
⊲
We shall discuss the perturbation theory method for time independent problems.
Our aim is to compute the eigenvalues and eigenfunctions of a time independent
Hamiltonian. In this approximation scheme the starting point is to split the Hamil-
tonian of the system, H, into two parts

H = H0 + H ′

in such a way that the eigenvalues and eigenfunctions of H0 can be found exactly.
We further assume that the additional part λH ′ has small in matrix elements as
compared to H0 .
The approximate solution is obtained by assuming that the exact eigenvalues
and eigenfunctions have an expansion in powers of λ and that it is sufficient to keep
first few terms of the expansion in powers of λ.
We shall be interested in first and second order perturbation theory (terms up
order λ2 ) for the following two cases when the eigenvalues of H0 are non-degenerate
or may have degeneracy. Notation: We shall write

H = H0 + λH ′ (4.1)

where λ is introduced for book keeping purposes and at the end of all computations λ
will be set equal to one. Eigenfunctions of H0 will be denoted by un and eigenvalues
by En
H0 un = En un (4.2)
Eigenfunctions of the full Hamiltonian will be denoted by ψ and the corresponding
eigenvalues by W .
Hψ = W ψ (4.3)
Further, it will be assumed that, ψ and W can be expanded in powers of λ

ψ = ψ0 + λψ1 + λ2 ψ2 + · · · (4.4)
W = W0 + λW1 + λ2 W2 + · · · (4.5)

35

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The eigenfunctions un of the unperturbed Hamiltonian will be assumed to be or-
thonormal
(un , um ) = δnm (4.6)
We start with 4.1 and substitute the expansions for ψ and W from 4.4-4.5 in the
eigenvalue equation
Hψ = W ψ (4.7)
for the full Hamiltonian giving

(H0 + λH ′)(ψ0 + λψ1 + λ2 ψ2 + · · · ) = (W0 + λW1 + λ2 W2 + · · · )(ψ0 + λψ1 + λ2 ψ2 + · · · )

(4.8)
Comparing powers of λ on both sides we get

H0 ψ0 = W0 ψ0 (4.9)
H0 ψ1 + H ′ ψ0 = W0 ψ1 + W1 ψ0 (4.10)
H0 ψ2 + H ′ ψ1 = W0 ψ2 + W1 ψ1 + W2 ψ0 (4.11)
(4.12)

These equations will be utilized and corrections to the energy levels and wave func-
tions for different cases, degenerate and nondegenerate levels, will be obtained in
the first and second orders in perturbation.

§2 Nondegenerate Case
We shall obtain corrections, (assumed to be small), to eigenvalue and eigenfunc-
tions corresponding to a fixed level n. Let un and En be the corresponding energy
eigenfunction and eigenvalue, so that

H0 un = En un (4.13)

where n is a given, fixed number throughout this discussion and refers to a particular
energy level of H0 .
In this section we assume that the energy level En is nondegenerate.
The equation Eq.(4.9) shows that ψ0 is an eigenfunction of H0 with eigenvalue
W0 . Since we are interested in finding corrections to energy and eigenfunctions of
the n-th energy level (assumed to nondegenerate), we take

W0 = En ψ0 = un (4.14)

where n is fixed number. Eq.(4.10) and Eq.(4.11), thus, take form

H0 ψ1 + H ′ un = En ψ1 + W1 un (4.15)
H0 ψ2 + H1′ ψ1 = W0 ψ2 + W1 ψ1 + W2 un (4.16)

The overall normalisation of the full eigenfunction is arbitrary and can be always be
selected to be equal to unity at the end. It will turn out to be convenient to choose
the normalisation of ψ in the intermediate steps to be

(un , ψ) = 1 (4.17)

36

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Using
ψ = ψ0 + λψ1 + λ2 ψ2 + · · · (4.18)
and noting ψ0 = un we get

1 + λ(un , ψ1 ) + λ2 (un , ψ2 ) + · · · = 1 (4.19)

Comparing coefficients of different powers of λ we get

(un , ψ1 ) = 0, (un , ψ2 ) = 0, · · · (4.20)

Thus the corrections terms in the eigenfunction will be chosen to be orthogonal to

the lowest order eigenfunction un .

§ 2.1 First Order Perturbation Corrections

Take scalar product of Eq.(4.10) with uk gives

(uk , H0 ψ1 ) + (uk , H ′un ) = En (uk , ψ1 ) + W1 (uk , un ) (4.21)

The first term in Eq.(4.21), using the hermiticity of H0 , can be written as

(uk , H0 ψ1 ) = (H0 uk , ψ1 ) = (Ek uk , ψ1 ) = Ek (uk , ψ1 ) and hence

Ek (uk , ψ1 ) + (uk , H ′ un ) = En (uk , ψ1 ) + W1 (uk , un ) (4.22)

The above equation holds for all k. We shall consider the two cases k = n and k 6= n
separately.

§ 2.1.1 Case k = n
For k = n Eq.(4.22) takes the form

En + (un , H ′un ) = En + W1 kun k2 (4.23)

The first terms on the two sides cancel, noting (un , un ) = kun k2 = 1, because un are
assumed to be normalized, we get from Eq.(4.23)

W1 = (un , H ′ un ) (4.24)

This equation gives then first order correction to the energy eigenvalue En . Thus
the energy eigenvalue upto first order becomes

W = En + (un , H ′ un ) (4.25)

§ 2.1.2 Case k 6= n
In this case (uk , un ) = 0 and the last term in Eq.(4.22) in the r.h.s. drops out and
we get
Ek (uk , ψ1 ) + (uk , H ′ un ) = En (uk , ψ1 ) (4.26)
which on a rearrangement gives

(En − Ek )(uk , ψ1 ) = (uk , H ′un ) (4.27)

37

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 which gives us an answer for (uk , ψ1 )
(uk , H ′ un )
(uk , ψ1 ) = , k 6= n. (4.28)
En − Ek
Eq.(4.29) determines ψ1 because {uk } form a complete orthonormal set, ψ1 can be
in terms of uk as
∞
X ∞
X
ψ1 = ck u k = cn u n + ck u k (4.29)
k=1 k6=n

where the expansion coefficients ck , determined by using orthonormality of the un-

perturbed eigenfunction uk are given by ck = (uk , ψ1 ). Since ψ1 is selected to be
orthogonal to un , cn = 0, and the term k = n is absent (see Eq.(4.17)-(4.20)).
Thus for k 6= n we get
(uk , H ′ un )
ck = k 6= n (4.30)
En − En
Thus Eq.(4.29) for ψ1 takes the form
X
ψ1 = ck u k (4.31)
k6=n
X < k|H ′|n >
= uk (4.32)
k6=n
En − Ek

To summarize the first order correction terms are given by

W1 =< n|H ′|n > (4.33)

∞
X ′ < k|H ′ |n >
ψ1 = uk (4.34)
En − Ek
k=1
P′
where means sum over all values of k except k = n. Note that in the first order
correction, Eq.(31), the constant cn is not determined. Upto first order in λ the
wave function becomes

ψ ≈ ψ0 + λψ1 (4.35)
X∞
< k|H ′ |n >
≈ λ uk (4.36)
k=1
E n − Ek
k6=n

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-Lec-

16001;

QM- § 2.2 Second Order Nondegenerate Perturbation Theory

23-
01- ⊲
L2
To determine the second order corrections, we start from Eq.(4.16)

H0 ψ2 + H ′ ψ1 = W0 ψ2 + W1 ψ1 + W2 ψ0 (4.37)

38

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The expansion of ψ2 (x) in terms of the unperturbed eigenfunctions takes the form
∞
X ′
ψ2 = dn un + dk u k (4.38)
k=1
P
where we have wrtitten the un term explicitly and ′ means the term k = n is
missing. Note that as a consequence of the normalisation chosen (un , ψ) = 0, we
have dn = 0 giving
∞
X ′
ψ2 = dk u k (4.39)
k=1

We also recall earlier expressions

W1 = hn|H ′ |ni (4.40)
∞
X ′ hk|H ′ |ni
ψ1 = uk (4.41)
En − Ek
k=1

Now we take scalar product of Eq.(4.37) with up and consider cases p = n and p 6= n
separately. Eq.(4.37) gives
(up , H0 ψ2 ) + (up , H ′ ψ1 ) = W0 (up , ψ2 ) + W1 (up , ψ1 ) + W2 (up , ψ0 ) (4.42)
Recall W0 = En , ψ0 = un and rewrite the first term as (up , H0 ψ1 ) = (H0 up , ψ1 ) =
Ep (up , ψ1 ). In the first term on the right hand side of (36) use W0 = En , ψ0 = un
and rewrite (36) as
Ep (up , ψ2 ) +(up , H ′ ψ1 ) = En (up , ψ2 ) +W1 (up , ψ1 ) + W2 (up , un ) (4.43)
| {z } | {z }
(1) (2)

§ 2.2.1 Case p=n

The first terms on the two sides of Eq.(4.43), marked as (1) and (2), cancel and we
get
(un , H ′ψ1 ) = W1 (un , ψ1 ) + W2 (un , un ) (4.44)
Using expressions Eq.(4.40) and Eq.(4.41) in Eq.(4.44) gives
P < k|H ′|n >
k6=n (un , H ′uk ) (4.45)
En − Ek
X < k|H ′ |n >
= W1 (un , uk ) + W2 (4.46)
k6=n
En − Ek
| {z }
(1)

where the cn terms drop out when Eq.(4.40) is used. The first term, marked (1), on
the right hand side is zero because of orthogonality (un , uk ) = 0 when k 6= n. Thus
we get
X < k|H ′|n >
W2 = < n|H ′|k > (4.47)
k6=n
E n − Ek

39

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 or
∞
X | < n|H ′ |k > |2
W2 = (4.48)
En − Ek
k=1
k6=n

§ 2.2.2 Case p 6= n
We now determine the second order corrections to the wave function. Eq.(4.43) now
takes the form

Ep (up , ψ2 ) +(up , H ′ ψ1 ) = En (up , ψ2 ) +W1 (up , ψ1 ) + W2 (up , un ) (4.49)

| {z } | {z } | {z }
(I) (II) (III)

The last term (III) is zero because (up , un ) = 0 for p 6= n. Combining the terms
marked, (I) and (II) and using (up , ψ2 ) = 0 in Eq.(4.49) gives

Ep dp + (up , H ′ψ1 ) = En dp + W1 (up , ψ1 ) (4.50)

or (Ep − En )dp = W1 (up , ψ1 ) − (up , H ′ ψ1 ) (4.51)
(4.52)

The second order correction to the wave function is given by

X′
ψ2 = dp u p (4.53)

where
(un , H ′ un )(up , ψ1 ) (up , H ′ψ1 )
dp = − (4.54)
Ep − En Ep − En
The first order correction ψ1 , appearing in the above equations is known and is given
by Eq.(4.41).
X′ hk|H ′|ni
ψ1 = uk (4.55)
k
E n − Ek

Taking scalar product with up , (p 6= n) we get

X′ hk|H ′|ni(up , uk )
(up , ψ1 ) = (4.56)
k
En − Ek

Also, applying the operator H ′ on Eq.(4.55) and taking scalar product with up gives
∞
X ′ hk|H ′ |ni
′
(up , H ψ1 ) = (up , H ′uk ) (4.57)
k=1
E n − Ek

This completes determiantion of eigenvalues and eigenvectors upto second or-

der of perturbation. The unknown constants cn , dn are fixed by demanding the
normalization condition (ψ, ψ) = 1 to every order in λ.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-23-01-
L2;

40

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM-
§3 Degenerate Case
23-
02- ⊲
L1 The total hamiltonian H is split into two parts
H = H0 + λH ′ (4.58)
Suppose we are looking for corrections to an energy eigenvalue En which is
degenerate. Without loss of generality one may assume that the degeneracy is
2, the results derived can easily be generalized for any value of degeneracy. Thus
assume that there are two linearly independent solutions uαn , α = 1, 2.
H0 u(1)
n = En u(1)
n (4.59)
(2) (2)
H0 un = En un (4.60)
We assume that exact eigenvalue, W , and eigenfunctions, ψ(x), have expansions in
powers of λ
ψ = ψ0 + λψ1 + λ2 ψ2 + · · · (4.61)
W = W0 + λW1 + λ2 W2 + · · · (4.62)

§ 3.1 First Order Energy Level Splitting

Substituting the expansions of ψ and W from Eq.(4.61) and Eq.(4.62) in Eq.(4.58)
we get
H0 ψ0 = W0 ψ0 (4.63)
H0 ψ1 + H ′ ψ0 = W0 ψ1 + W1 ψ0 (4.64)
H0 ψ2 + H ′ ψ1 = W0 ψ2 + W1 ψ1 + W2 ψ0 (4.65)
To find the corrections to the unperturbed solutions of Eq.(4.59)-Eq.(4.60) we set
W0 = En the most general expression for the unpertubed eigenfunction s ψ0 is
ψ0 (x) = α1 u(1) (2)
n + α2 un (4.66)
(1)
Taking the scalar product of Eq.(4.64) with un and using W0 = En , Eq.(4.59) we
get
(u(1) (1) ′ (1) (1)
n , H0 ψ1 ) + (un , H ψ0 ) = En (un , ψ1 ) + W1 (un , ψ0 ) (4.67)
(u(1) ′ (1)
n , H ψ0 ) = W1 (un , ψ0 ) (4.68)
Substiuting from Eq.(4.59) and Eq.(4.60) in Eq.(4.68) we get
(u(1) ′ (1) ′
n , H un (1))α1 + (un , H un (2))α2 = W1 α1 (4.69)
[(2)
Similarly, taking the scalar product of Eq.(4.64) with un gives
(u(2) ′ (2) ′
n , H un (1))α1 + (un , H un (2))α2 = W1 α2 (4.70)
Now note that Eq.(4.69) and Eq.(4.70) can be rewritten in a matrix form
    
hn1|H ′|n1i hn1|H ′ |n2i α1 α1
′ ′ = W1 (4.71)
hn1|H |n2i hn2|H |n2i α2 α2

41

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 This equation is recognised as an eigenvalue equation. Hence the first order correc-
tion, W1 , to the energy eigenvalue En is obtained by finding the eigenvalues of the
matrix appearing in the left hand side of Eq.(4.71).
Note that the eigenvlaues W1 appearing in Eq.(4.71) will be distinct if either the
off diagonal elements are nonzero
hn1|H ′ |n2i =
6 0 (4.72)
or if the diagonal elements are distinct,
hn1|H ′|n1i =
6 hn2|H ′ |n2i (4.73)
If the eigenvalues are distinct,two sets of nontrivial values, one for each eigenvalue
W1 , for the coefficients α1 , α2 will can be found and Eq.(4.66) determines corre-
sponding lowest order eigenfunctions,ψ0 . However, when the conditions in Eq.(4.72)-
Eq.(4.73) are not satisfied, the constants α1 , α2 remain undetermined and one must
go to the second order perturbation theory to find the corrections to the energy
levels and eigenfunctions.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-23-02-
L1;

QM- § 3.2 First Order Wave Function

23-
02- ⊲
L2
We take the scalar product of Eq.(4.64) with uk , with k 6= n, to get

(uk , H0 ψ1 ) + (uk , H ′ ψ0 ) = W0 (uk , ψ1 ) + W1 (uk , ψ0 ) (4.74)

For k 6= n (uk , ψ0 ) = 0 and
(uk , H0 ψ1 ) = (H0 uk , ψ1 ) (4.75)
= Ek (uk , ψ1 ) (4.76)
Making use of Eq.(4.75)-Eq.(4.76) in Eq.(4.74) gives
Ek (uk , ψ1 ) + (uk , H ′ ψ0 ) = En (uk , ψ1 ) (4.77)
(uk , H ′ ψ0 )
(uk , ψ1 ) = (4.78)
Ek − En
To determine ψ1 we expand it in terms of the uneprturbed solutions and write
X
ψ1 = dk u k (4.79)
X
= d(1) (1) (2) (2)
n u n + dn u n + dk u k (4.80)
k6=n

and the coefficients dn , n 6= k are just the constants given by dk = (uk , ψ1 ) and hence
X (uk , H ′ψ0 )
ψ1 = d(1)
n un
(1)
+ d(2)
n un
(2)
+ uk (4.81)
k6=n
Ek − En

In case the matrix appearing in Eq.(4.71) is a multiple of identity then α1 , α2 are

not determined and one has to go to the next order of perturbation theory.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-23-02-
L2;

42

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM- § 3.3 Second Order Degenerate Perturbation Theory
23-
02- ⊲
L3 In all the cases of realistic applications of degenerate perturbation theory one is
interested in the splitting of the levels due to the perturbation term. In case the
degeneracy is not removed in the first order one must go for a second order com-
putation. In such a case the lowest order calculation does not lead to information
on the wave function because α1 , α2 remain undetermined. The change in energy
W2 and the constants α1 , α2 must then be fixed by the second order perturbation
theory.
Upto the second order in the perturbation hamiltonian the wave function is

ψ = ψ0 + λψ1 + λ2 ψ2 (4.82)
= α1 u(1) (2) 2
n + α2 un + λψ1 + λ ψ2 (4.83)

Without loss of generality, we may assume that ψ1 and ψ2 are orthogonal to the
(1) (2)
unperturbed eigenfunctions un , un . Demanding this amounts to changing, as yet
unknown, parameters α1 , α2 . We start with the equation, Eq.(4.65) ,

H0 ψ2 + H ′ ψ1 = W0 ψ2 + W1 ψ1 + W2 ψ0 (4.84)
(1)
Taking scalar product with un , we get

(u(1) (1) ′ (1) (1) (1)

n , H0 ψ2 ) + (un , H ψ1 ) = W0 (un , ψ2 ) + W1 (un , ψ1 ) + W2 (un , ψ0 ) (4.85)

The first two terms on the right hand side are zero due the orthogonality of ψ1 , ψ2
(1)
to un . The first term on the left hand side also vanishes on using the hermiticity
of H0

(u(1) (1)
n , H0 ψ2 ) = (H0 un , ψ2 ) (4.86)
= En (u(1)
n , ψ2 ) (4.87)
= 0

Also Eq.(4.85) we get,

(u(1) ′ (1)
n , H ψ1 ) = W2 (un , ψ0 ) (4.88)
= α1 W2 (4.89)

On using Eq.(4.41) for ψ1 we get

X hk|H ′|ψ0 i
= W2 α1 (4.90)
k6=n
Ek − En

Substituting for ψ0 from Eq.(4.66) in Eq.(4.90) we get

X hk|H ′|n1ihn1|H ′|ki X hk|H ′ |n2ihn1|H ′|ki

α1 + α2 = W2 α1 (4.91)
k6=n
Ek − En k6=n
Ek − En

43

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 (2)
Similarly, taking scalar product with un gives another equation
X hk|H ′|n1ihn2|H ′|ki X hk|H ′ |n2ihn2|H ′|ki
α1 + α2 = W2 α2 (4.92)
Ek − En Ek − En
k6=n k6=n

The above equations can be put in a matrix form

    
P hn1|H ′|kihk|H ′|n1i P hn1|H ′ |kihk|H ′|n2i α1 α1
 k6=n Ek − En k6=n
Ek − En     
    = W2   (4.93)
P ′ ′
hn2|H |kihk|H |n1i P hn2|H |kihk|H |n2i   
′ ′  
k6=n
Ek − En k6=n
Ek − En α2 α2

The two eigenvalues, for W2 , give the second order corrections and the eigenvectors
give the values of α1 , α2 which determines the lowest order wave function ψ0 . It
must be remembered that the above formula apply only in case the degeneracy is
not removed to the first order in λ.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-23-02-
L3;

§4 Structure of hydrogen atom energy levels

QM-
Lec-
16006
⊲
Predictions of the Schrödinger theory
The first successful attempt to explain the H atom levels was by Bohr. In quantum
mechanics the results are derived by solving the Schrödinger equation. The energy
of a level with principal quantum number n is given by

α2 mc2
En = − , (4.94)
2n2
e 2
where α = 4πǫ 0
in si units, and is known as the fine structure constant. It is
dimensionless and has a value ≈ 1/137.07. The Schrödinger theory predicts that
for a fixed value of the principal quantum number n, the angular momentum can
take values ℓ = 0, 1, · · · , (n − 1) and for each ℓ the z component of the angular
momentum has (2ℓ + 1) values mℓ = −ℓ, −ℓ + 1, · · · , ℓ. For each level corresponding
to these combinations of nℓ, mℓ quantum numbers, the component of the spin of the
electron along a specified directions can have two possible values to ± ~2 . The energy
depends only on n and hence there is a degeneracy of 2n2 . For a few energy levels
we schemetically show the set of quantum numbers in figure below.

44

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Fig. 4.1. qm-cdr-23001;

Each line shown here in the diagram must be regarded as a bunch of 2n2 lines
nℓ and ms quantum numbers explicitly indicated in the Fig.1.(b) of the diagram. In
Fig.1.(c) we show the labels written in terms of n, ℓ, j, mj quantum numbers, where
j refers to the total angular momentum. For each ℓ there are two values j = ℓ ± 12 ,
and again mj takes (2j + 1) values from −j, −j + 1, · · · , j.

Fine structure of H atom

Bohr’s theory predicted spectral lines for hydrogen atom correctly. However, precise
experiments showed that what was seen as a single spectral line in fact had a fine
structure; most of the lines consist of several closely packed spectral lines. Translated
in terms of energy levels, the fine structure refers to a ‘fine’ splitting of energy levels,
which is smaller approximately by a factor of α2 as compared with the differences in
Bohr levels with different n. The experimental results on fine structure of hydrogen
atom energy levels can be summarised as follows.

The exact energies depend on j but do not depend on ℓ. Thus the H

atom levels with different n, j get split by small amounts. However, the
levels having same set of values for nj quantum numbers and different ℓ
remain degenerate, see Fig.4.1(d)
Of course, there is no splitting between levels with different mj (or ms )
values but same nj quantum numbers.
So, for example, the structure of lowest three levels is as follows.
• The ground state level, n = 1, having ℓ = 0, j = 1/2 only, receives corrections
but there is no splitting as there is only one j value (=1/2) corresponding to
n = 1.

45

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 • For n = 2, ℓ has values 0, 1 and corresponding values of j are 1/2, 3/2. In the
spectroscopic notation ,nlj , these levels are denoted by 2s 1 , 2p 1 , 2p 3 . The
2 2 2
levels 2s 1 , 2p 1 remain degenerate but 2p 3 receives a different correction and
2 2 2
splits from the other levels. Thus n = 2 level splits into two levels.

• Similarly, for n = 3 in Schrödinger theory the levels

3s 1 , 3p 1 , 3p 3 , 3d 3 , 3d 5
2 2 2 2 2

are all degenerate. The relativistic effects split the level into three groups as
given below.

(a) 3s 1 , 3p 1 , (b) 3p 3 , 3d 3 , (c) 3d 5 .

2 2 2 2 2

The results described above are shown in figure, not to the scale, below. As already
noted all the above splittings are small and are of the order of α4 mc2 .

Lamb shift
The Lamb shift refers to a very tiny splitting of energy levels with same nj and
different ℓ values. The Lamb shift is of the order of α5 mc2 . For example, experi-
mentally the levels 2s 1 and 2p 1 are not degenerate and the energy difference is given
2 2
by

E(2s 1 ) − E(2p 1 ) ≈ 1057 MHz.

2 2

Hyperfine structure
The energy levels of Hydrogen atom after taking into account of of the fine structure
and the Lamb shift appear as shown in figure below. Each of these levels have a
hyperfine structure; each level splits further into two levels with energy differences
m 4
∼M α mc2 where M is the proton mass.
Order of magnitudes of various observed energy level splittings can be sum-
marised as follows.

Bohr levels Fine splitting Lamb shift Hyperfine splitting

me


α2 mc2 α4 mc2 α5 mc2 mp
α4 mc2
∼ a few eV ∼ 10−4 eV ∼ 10−6 eV ∼ 10−7 eV

46

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Fig. 4.2. qm-cdr-23002;
Relativistic Dirac equation and fine structure
The fine structure is a relativistic effect and is fully explained by the Dirac theory.
The Dirac equation for the Coulomb potential can be solved exactly and makes
predictions which are in full agreement with the experiments.
The Dirac theory does not explain the Lamb shift because the electromagnetic
field is not quantised; a full treatment of Lamb shift requires use of quantum elec-
trodynamics, which is the quantum theory of electrons and photons, and the Lamb
shift is explained within experimental errors. We shall not go into details of the
Lamb shift calculations.
The non-relativistic limit of the Dirac theory gives rise to three correction terms
to the Coulomb interaction of Schrödinger theory. These corrections are

1. Corrections due to relativistic variation of mass

p4
H1 = − .
8m3 c2

2. Spin orbit coupling

e 1 dV ~ ~
H2 = − (L · S).
2m2 c2 r dr
e
where V (r) is the electric potential due to the nucleus and is equal to for
r
the hydrogen atom.

3. Darwin term
π~2
H3 = (Ze2 )δ(~r).
2m2 c2

47

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Summary of perturbative treatment
The three terms H1 , H2 , H3 , obtained from Dirac equation, can be treated as a
perturbation to the Coulomb Hamiltonian
p2 e2
H0 = − (4.95)
2m r
of the non-relativistic Schródinger theory, and the results in the first order pertur-
bation theory are adequate to explain the experimental results on the fine structure
of the energy levels. A level with principal quantum number n has degeneracy 2n2
and, in general, one has to construct a 2n2 × 2n2 matrix whose eigenvalues will give
the first order corrections to the level En . In practice, this can be avoided by a
suitable choice of basis for the degenerate level.
The three correction terms coming from the three interactions H1 , H2 , H3 are as
follows.
 Zα 2  3 n

∆E1 = −En − (4.96)
n 4 ℓ + 1/2
(
(Zα)2 −ℓ − 1, j = ℓ − 1/2
∆E2 = En × (4.97)
nℓ(ℓ + 1)(2ℓ + 1) ℓ, j = ℓ + 1/2
( 2)
−En (Zαn
, ℓ=0
∆E3 = (4.98)
0, ℓ 6= 0

Combining all the three corrections gives a final answer

 
(Zα)2 n
(∆E)nj = En − 3/4 . (4.99)
n2 j + 1/2
and the final energy upto lowest order in relativistic corrections is given by
  
(Zα)2 n
Enj = En 1 + − 3/4 . (4.100)
n2 j + 1/2
This expression can also be obtained by expanding the exact result of Dirac’s theory
in powers of Zα and keeping lowest order relativistic corrections.

Details of perturbative calculations

We will now give details of perturbative calculation for shift in energy due to the
three terms H1 , H2 and H3 .

Corrections due to H1 term

The perturbation term H1 , being invariant under rotations, commutes with the
angular momentum operators and hence it does not connect states with different
ℓm values. Being independent of spin, it also does not connect states with different
spins projections ms .Thus

hnℓ′ m′ℓ m′s |H1 |nℓmℓ ms i = 0

48

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 if ℓ 6= ℓ′ or mℓ 6= m′ℓ , or ms 6= m′s . Thus the matrix of H1 between different states
with energy En is already diagonal and the correction to a level with quantum
numbers nℓmℓ ms is given by the diagonal elements
1
(∆E)1 hnℓmℓ ms |H1 |nℓmℓ ms i = − hnℓmℓ ms |p4 |nℓmℓ ms i (4.101)
8m3 c3
1
= − 3 3 hnℓmℓ ms |(p2 )(p2 )|nℓmℓ ms i(4.102)
8m c
Noting that the state |nℓml ms i is an eigenstate of unperturbed Hamiltonian H0 =
p2 /(2m) − Ze2 /r we have
 Ze2   Ze2 
(p2 )|nℓmℓ ms i = (2m) H0 + |nℓmℓ ms i = 2m En + |nℓmℓ ms i. (4.103)
r r
So that the correction takes the form
*  2 2 +
2
1 Ze Ze
(∆E)1 = − 2
En2 + 2En + . (4.104)
2mc r r

Computing the averages of 1/r and 1/r 2 gives the first order correction (∆E)1 .

Corrections due to spin orbit term H2

The term H2 contains L ~ ·S
~ which commutes with L ~ 2 but does not commute with
Lz and Sz . Therefore its matrix elements between states with different mℓ and ms
values will be non zero. Calculation of corrections to energy, for a given ℓ, will
require diagonalisation of 2(2ℓ + 1) × 2(2ℓ + 1) martices. This diagonalisation can
be avoided and computations can be simplified by considering states |n, ℓjmi with
definite values of operators J~2 , L
~ 2, S
~ 2 , Jz , where J~ = L ~ +S ~ is the total angular
momentum. All these operators commute with H3 and hence the off diagonal terms
in this basis set will be zero and the corrections to the energy levels, (∆E)2 , will be
given by the diagonal elements hnℓjm|H2 |nℓjmi. Squaring J~ = L ~ + S,
~ we get useful
relation


L ~ = 1 J~2 − L
~ ·S ~2 − S~2 (4.105)
2
and the states |nlℓjmi are eigenstates of the operators L ~ ·S
~ with eigenvalue 1 (j(j +
2
1) − ℓ(ℓ + 1) − 1/4)~2 . Therefore we have

~2
hnℓjm|H3 |nℓjmi = (j(j + 1) − ℓ(ℓ + 1) − 1/4)hnℓjm|ξ(r)|nℓjmi, (4.106)
2
where
1  Ze2 
ξ(r) = , (4.107)
2m2 c2 r 3
and V (r) = Ze/r for the Coulomb potential due to the nucleus has been substituted.
For spin half case in hand, j can take two values ℓ ± 21 and we get
(
~2 ℓ j = ℓ + 21
hnℓjm|H3 |nℓjmi = hξ(r)i (4.108)
2 −ℓ − 1 j = ℓ − 12 .

49

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 and hξ(r)i denotes the average value of ξ(r) in state njℓml . For ℓ = 0 the spin orbit
~ ·S
term L ~ becomes zero. For ℓ 6= 0 we get
 Ze2 D 1 E
hξ(r)i = (4.109)
2m2 c2 r 3
 Ze2  Z3
= , ℓ 6= 0. (4.110)
2m2 c2 a30 n3 ℓ(ℓ + 1)(ℓ + 1/2)

This completes calculation of the shift in energy due to spin ornit term, (∆E)2 .

Contribution of Darwin term

The radial wave function for orbital angular momentum ℓ 6= 0, for small r behaves
like r ℓ and vanishes for r = 0. Due to the presence of the Dirca delta function,
δ(~r, the Darwin term contribution is proportional to |unℓm(0)|2 and becomes zero
for states with ℓ 6= 0. The correction to the energy due to the Darwin term, for n00
level is given by
 πZe2 ~2 
∆E3 = hn00|δ(~r)|n00i
2m2 c2
 πZe2 ~2 
= |un00 (0)|2
2m2 c2
(Zα)2
= −En . (4.111)
n

Remarks:
The fine structure was explained within the Schrödinger non-relativistic quantum
mechanics before Dirac equation was proposed. The presence of relativistic variation
of mass and the spin orbit interaction, but not the Darwin term, can be understood
without recourse to the Dirac equation.
The relativistic energy of a particle of mass m is given by
p p
E = p2 c2 + m2 c4 = mc2 1 + (p/mc)2 (4.112)

Expanding the square root in powers of p/F MC we get

p2 p4
E = mc2 + − +··· (4.113)
2m 8m3 c2
The first term is the rest energy, the second term is just the kinetic energy of a point
mass in non-relativistic mechanics. The third term is lowest order correction due to
relativistic variation of mass.
The spin orbit coupling term represents the interaction of spin magnetic moment,
ge ~
~µ = − 2mc S, of the electron with the magnetic field due to its orbital motion and
where g is the gyromagnetic ratio for the electron and has the g ≈ 2. In the rest
frame of the nucleus there is Coulomb potential V (r) gives an electric field. This
electric field
~ = −∇V (r) = − dV (r) ~r
E
dr r

50

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 in the rest frame of the electron, gives rise to a magnetic field. Applying a Lorentz
transformation to a frame moving with velocity v the magnetic field is found to be,

B ~ = 1E
~ = − 1 ~v × E ~ × ~v = − 1 dV (r) (~r × ~v ). (4.114)
c c cr dr
The interaction of the magnetic dipole moment of the electron due to its spin,
e ~ ~ which has the form
~µ = − mc S leads to an interaction −~µ · B
~ = e (B
−~µ · B ~ · S)
~ (4.115)
mc
e dV (r)

= − 2 ~r × ~v · S~ (4.116)
mc r dr
e dV

= − 2 2 ~r × p~ · S ~ (4.117)
m c r dr
e 1 dV ~ ~

= − 2 2 L·S . (4.118)
m c r dr
where L~ is the orbital angular momentum of the electron. a naive application of
Lorentz transformation on the electric field to obtain the magnetic field in the rest
frame is not justified as the electron’s velocity keeps changing and hence the rest
frame of the electron is not an inertial frame. Thus the spin orbit term obtained
above requires a correction, known as Thomas precision correction, which is of the
same form as the electron spin magnetic moment interaction with the magnetic field
but requires a numerical factor of 1/2 in the final expression. This gives H2′ as the
final answer for the spin orbit term. There is no way of seeing the appearance of
Darwin term, which was first derived from the Dirac equation.

Hyperfine structure
The hyperfine structure comes from the interaction of electron’s dipole moment with
the magnetic field due to the spin dipole moment of the nucleus. The proton, like
ge ~
every charged spin half particle, carries a magnetic moment given by µ~p = 2M pc
S,
g, called the gyromagnetic ratio, is found to be 5.59. The magnetic moment for the
proton is very small compared to the magnetic moment of the electron because the
mass Mp appears in the denominator. The hyperfine structure splitting works out
to be proportional to the average value of S~e · S~p . This operator has diagonal matrix
elements between states with total spin S ~ = S~e + S~p . The total spin for the electron
proton system can be 0 (singlet) or 1 (triplet) and we write
1 2

S~e · S~p = S − Se2 − Sp2 (4.119)
2
(
1
4
for triplet, S = 1
= 3
(4.120)
− 4 for singlet, S = 0
Thus the effect of hyperfine interactions is to split each level into two levels corre-
sponding to total spin S = 0, 1. Assembling all the factors, this splitting works out
to be very tiny, of the order of 10−6 eV. For the ground state of hydrogen atom, the
photon emitted in transition from triplet to singlet has the wave length of 21cm,
and is the famous 21 cm line is present everywhere in radiation in the universe.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
16006;

51

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §5 Related topics
Learn The Following Topics From Class Notes

Ground State and the Excited States of He Atom

Stark Effect of Harmonic Oscillator

52

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §6 Examples and Problems
Tutorial-I
[1] A particle moves in a one dimensional box with rigid boundary at x = 0 and
x = L walls. A small perturbation V1 given by
( L 3L
ǫV0 for 4
<x< 4
V1 (x) =
0 otherwise

is applied find the shift in the nth energy level.

[2] Use the first order non-degenerate perturbation theory to compute the correc-
tion to the ℓ = 1 level of ( use units so that ~ = 1 )

H = L2 + αLz + βLx

Use the splitting of H as

H0 = L2 + αLz , and H ′ = βLx

in terms of unperturbed hamiltonian and perturbation hamiltonian H ′ .

(a) Find the corrections to the ℓ = 1 energy level using perturbation theory
to the lowest non-vanishing order in H ′ . Compare your answer with the
exact answers.
(b) Obtain the eigenvectors of H upto lowest order in β.

[3] A particle, having charge q, moves in harmonic oscillator potential V (x) =

1
2
mω 2 x2 . Find the shift in the energy level of the nth exctited state when
uniform electric field E is applied. Compute the corrections upto lowest order
giving a nonzero value.

Tutorial-II
[1] A particle moves in two dimensional circular oscillator potential with

p2x p2y 1
H0 = + + mω 2 (x2 + y 2)
2m 2m 2
(a) What are the quantum numbers of the first excited state? Is it degenerate
or not?
(b) If a small perturbation H ′ = λxy is applied compute the lowest order
correction to the energy of the first excited state.

[2] Use the first order degenerate perturbation theory to compute the correction
to the l = 1 level of ( use units such that ~ = 1)

H = L2 + αLx + βLy

53

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 taking
H0 = L2 , and H ′ = αLx + βLy
Find the corrections to the l = 1 level using first order perturbation theory.
Compare your answer with the exact answers.

[3] Following the steps given in the class, find eigenvectors and eigenvalues of the
matrix  
1 ǫ 0
 ǫ 1 −ǫ 
0 −ǫ 2
upto the lowest non-vanishing order in ǫ.

Questions
[1] Show that the second order perturbation correction to the ground state energy
is always negative.

[2] Write the steps you would follow to find the first order correction to the energy
levels in perturbation theory when the unperturbed energy eigenvalues are
degenerate.
b commutes with X,
[3] If a hermitian operator H b and |ni is an eigenvector of H
b
with eigenvalue En . Prove that
b
hn|X|mi = 0.

if the eigenvalues Em and En are different. What is the use of this relation in
perturbation theory ? Is it necessary that Hb be a hermitian operator ?

b is replaced with the parity operator in the previous question, what does
[4] If H
one get? Interpret your answer.

[5] Give two examples of operators X, Y , such that their commutator is propor-
tional to one of them. Where does this property appear in quantum mechanics
and how is it used ?
b Yb be two operators such that X is hermitian, and their commutator is
[6] Let X,

b Yb ] = αYb
[X,

b with eigenvalue x0
If |x0 i is an eignevector of X

b 0 i = x0 |x0 i,
X|x

what can you say about the ket Ŷ |x0 i?

[7] Why is that the first order perturbation theory, for the ground state energy,
will never give a better estimate as compared to the value that can be obtained
by using the variation method ?

54

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 [8] If a perturbation λx4 is applied to a harmonic oscillator, will the correction to
the energy levels will increase with n or decrease with n, WHY ?

[9] When is the effect of finite size of nucleus to the electronic energy levels most
important for alkali atoms ? Assume hydrogen like levels and answer in terms
of the quantum numbers n, ℓ etc .

[10] A singlet excited state of He atom always has a higher energy as compared to
the corresponding triplet state. Explain why ?

55

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 5

The WKB Approximation

§1 Introduction
QM-
Lecs-
17001
⊲
The Wentzel Kramers Brillouin or the WKB approximation scheme is a method of
solving the Schrodinger equation by means of an expansion in powers of ~. The
leading term, in the limit ~ → 0 makes contact with the classical mechanics, there-
fore this method is also known as semiclassical approximation. This scheme was
developed for solving differential equations long before it was applied to quantum
mechanical problmes. For bound state problems it connects the Bohr Sommerfeld
quantisation rule with the Schrodinger theory. The Bohr Sommerfeld scheme of old
quatum theory worked for bound state problems but failed for alpha decay. The
WKB method is historically important because it was successfully used by Gamow
(1928) and by Gurney and Condon (1929) to give first explanation of alpha decay
of nuclei.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lecs-
17001;

§2 Semiclassical expansion
QM-
Lec-
17002
⊲
We are interested in solving the Schrödinger equation in one dimension
~2 d2 ψ(x)
+ V (x)ψ(x) = Eψ(x)
− (5.1)
2m dx2
by writing an ~ expansion for S related to the wave function by
ψ(x) = exp(iS/~). (5.2)
The equation satisfied by the function S is seen to be
 2  
1 dS i~ d2 S
− = E − V (x) (5.3)
2m dx 2m dx2
or  2  2 
dS dS
− i~ = 2m(E − V (x)). (5.4)
dx dx2
We expand S in powers of ~
S = S0 + (~/i)S1 + (~/i)2 S2 + · · · (5.5)

56

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 and substitute Eq.(5.5) in Eq.(5.4) to get
( 2  2 )
dS0 dS0 dS1 dS 1 dS 0 dS 2
, − 2i~ − ~2 − 2~2 +···
dx dx dx dx dx dx
 2 
d S0 d 2 S1 2
2 d S2
− i~ − i~ 2 − ~ + · · · = 2m(E − V (x)). (5.6)
dx2 dx dx2
We compare the coefficients of different powers of ~ on both sides of Eq.(5.6) to get
 2
dS0
= 2m(E − V (x)) (5.7)
dx
dS0 dS1 1 d2 S0
+ =0 (5.8)
dx dx 2 dx2
 2
dS0 dS2 dS1 d 2 S1
2 + + =0 (5.9)
dx dx dx dx2
Note that the right hand side of Eq.(5.7) is just the square of the classical momentum
function, p(x) p
p(x) = 2m(E − V (x)). (5.10)
Therefore, we get
dS0
= ±p(x). (5.11)
dx
Solving for S0 gives two solutions
Z
S0 = ± p(x)dx.. (5.12)

Solving Eq.(5.8) for S1 gives

 
dS1 1 dp d 1
=− =− log p(x) , (5.13)
dx p(x) dx dx 2
which on integration gives
1
S1 = − log p(x). (5.14)
2
Keeping terms upto first power of ~ in S gives

S = S0 − i~S1 + · · · , (5.15)

and the corresponding wave functions are as follows

ψ(x) = exp(iS/~) = exp(iS0 /~ + S1 ), (5.16)

1  iZ 
= p exp ± p(x)dx . (5.17)
p(x) ~

Therefore, the most general solution for the wave function in this approximation is
a linear combination
c1 i Z  c2  iZ 
ψ(x) = p exp p(x)dx + p exp − p(x)dx , (5.18)
p(x) ~ p(x) ~

57

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 where c1 , c2 are complex constants to be determined. Since E = p2 /2m + V (x) >
V (x), the classical motion of a particle is confined to those values of x where the
energy is greater than the potential energy. The set of points where p E > V (x) is
called the classical region. Note that in the classical region p(x) = 2m(E − V (x)
is real. Out side the classical region, E < V (x) and p(x) becomes pure imaginary
and we will write p(x) = i|p(x)|. For x outside the classical region a general WKB
solution for the wave function assumes the form
d1 1 Z  d2  1Z 
ψ(x) = p exp |p(x)|dx + p exp − |p(x)|dx . (5.19)
|p(x)| ~ |p(x)| ~

Assuming the potential energy to be a continuous function, E = V (x) for points

on the boundary of the classical region and the momentum function p(x) vanishes.
These points are known as the classical turning points. The WKB solutions,
Eq.(5.18),Eq.(5.19), for the wave function diverge at the turning points and the
approximation scheme breaks down.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
17002;

§3 Validity of WKB Approximation

QM-
Lec-
17003
⊲
We now determine when is the WKB approximation for the wave function, given
by Eq.(5.18) and Eq.(5.19) a good approximation? It isin Eq.(5.15) clear the
WKB approximation to the wave function cannot be a good approximation near
a turning point because at a turning point p(x) = 0. The expansion will be valid
away from the turning points where the second term S1 is small compared to the
first term S0 . In other words, S0 should be a good approximation to the solution of
the exact equation Eq.(5.4) . The condition for this to happen is the second term
in Eq.(5.4) should be small compared to the first term when S is replaced with S0 .
Thus we get
 dS 2   d2 S 
 0   0
  > ~ 2 . (5.20)
dx dx
dS0
On using = p(x) we get
dx  dp 
 
p2 ≫ ~ . (5.21)
dx
Writing ~/p = λ/(2π), we transform the above equation into the form
 1 dλ 
 
  ≪ 1. (5.22)
2π dx
This means that the change in wave length ∆λ over distance ∆x should be small
compared to ∆x itself:
1
∆λ ≪ ∆x. (5.23)
2π

58

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 p
For an alternate criterion of validity,use p = 2m(E − V (x) to write

dp dV (x) 1
= −m p (5.24)
dx dx 2m(E − V (x))
mF
= (5.25)
p

where F = −dV (x)/dx is the force. Therefore, the condition Eq.(5.21) for validity
of the WKB approximation can also be cast in the form

m~|F |
≪ 1. (5.26)
p3
The above condition implies that the WKB approximation is not valid at those
points where the force is large or the momentum is small. Since near a truning
point p ≈ 0, Eq.(5.26) leads us back to the conclusion that the WKB approximation
breaks down near turning points.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
17003;

§4 Bound state solutions

QM-
Lec-
17004
⊲
We have seen that the WKB solutions are valid approximate solutions away from the
turning points. For simplicity we shall discuss bound state solution for potentials,
with two turning points. The WKB treatment, when the number of turning points
is more than two, becomes very complicated and will not be discussed here. Let
the two turning points be x1 , x2 , then the WKB solutions are not valid in regions
close to the two turning points marked as R2 , R4 in Fig.5.1. The region R3 is the
classical region and the form of the WKB solutions in R3 is given by Eq.(5.18) . In
regions R1 and R5 are outside the classical region and the form of solutions is given
by Eq.(5.19) .

59

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Fig. 5.1.
qm-cdr-wkb01;Five regionsR for WKB scheme
We shall write the indefinite integral p(x) dx as a definite integral with upper
limit as x; the lower limit can be taken to be any convenient point, any other choice
amounts to changing the undetermined constants appearing in the most general
solution (such as c1 , c2 , d1, d2 etc.). So in the region R1 , away from the turning point
x1 , we write the WKB solution Eq.(5.19) as
a1 1 Z x  b1  1Z x 
ψ1 (x) = p exp |p(x)|dx + p exp − |p(x)|dx . (5.27)
|p(x)| ~ x1 |p(x)| ~ x1
Since x < x1 in region R1 , we exchange the upper and lower limits and write the
above solutions in the form
a1  1 Z x1  b1  1 Z x1 
ψ1 (x) = p exp − |p(x)|dx + p exp |p(x)|dx . (5.28)
|p(x)| ~ x |p(x)| ~ x
Rx
As the point x moves towards −∞, the range of integration in x 1 |p(x)|Rdx keeps in-
x
creasing and with the positive integrand, |p(x)|, the value of the integral x 1 |p(x)| dx
will keep increasing and tends to ∞ as x tends to −∞. For bound states the wave
function must tend zero at ±∞, we therefore must set b1 = 0. This gives
a1  1 Z x1 
ψ1 (x) = p exp − |p(x)|dx , x < x1 . (5.29)
|p(x)| ~ x
Similarly, the WKB solution in region R5 will be written as
a5 1 Z x  b5  1Z x 
ψ5 (x) = p exp |p(x)| dx + p exp − |p(x)| dx . (5.30)
|p(x)| ~ x2 |p(x)| ~ x2
Rx
Since x > x2 , the integral x2 |p(x)| dx increases with x and its value tends to ∞ as
x → ∞. Therefore we must set a5 = 0 and the above solution takes the form
b5  1Z x 
ψ5 (x) = p exp − |p(x)| dx , x > x2 . (5.31)
|p(x)| ~ x2

60

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Next, the most general form of the solution in the classical region R3 was found
to be Eq.(5.18) which we write as
i Z   iZ 
a3 b3
ψ3 (x) = p exp p(x)dx + p exp − p(x)dx , x1 < x < x2 ,
p(x) ~ p(x) ~
(5.32)
and we must demand that the wave function must be continuous. This cannot be
done by equating values of ψ1 and ψ3 at some point, because the regions R1 and R3
are non-overlapping and have no point in common. Similar comments apply to the
continuity requirement between ψ3 (x) and ψ5 (x). One way to solve this problem is
to approximate the potential by a linear potential near the turning points x1 , x2 .
With this approximation, the resulting Schrodinger equation equation for a linear
potential can be solved exactly in regions R2 and R4 . Near the turning point x1 the
solution can be written as a linear combination of the two linearly independent Airy
functions A1 (x) and A2 (x):

ψ2 (x) = a2 A1 (x) + b2 A2 (x), (5.33)

with a similar form of solutions for the WKB wave functions in region R4 near the
turning point x2 . One then demands that the Airy function solution in region R2
, ψ2 (x), matches with WKB solution ψ1 (x) for x < x1 and with ψ3 (x) for x > x1 .
This helps us in fixing the constants a3 , b3 in terms of a1 , an exercise that will be
done in the next section. We call the resulting solution ψ3 (x), for region R3 , as χ(x)
and write it as
Z
A 1 x π

χ(x) ≡ p cos p(x) dx − , x ∈ R3 . (5.34)
p(x) ~ x1 4

where the constant A is known in terms of a1 .

Next, a similar exercise is to be done at the second turning point x2 . We ask
the question what should be constants a3 , b3 in Eq.(5.32) so that the solution ψ3 (x)
matches with the decreasing solution ψ5 (x) of Eq.(5.31)? This leads to values of the
constants a3 , b3 in terms of b5 , see next section, and we denote the answer by φ(x)
and write it as
Z
B 1 x2 π

φ(x) ≡ p cos p(x) dx − , x ∈ R3 , (5.35)
p(x) ~ x 4

Thus we have arrived at two answers, χ(x) and φ(x), for the wave function in region
R3 , and both represent the same bound state wave function in this region. Obviously,
these two solutions themselves must agree for all x in region R3 implying that
Z Z
1 x π
1 x2 π

cos p(x) dx − = ± cos p(x) dx − (5.36)
~ x1 4 ~ x 4

for all x. This equation can be satisfied for all x only if the sum, or the difference,
of arguments of the two cosine functions is independent of x and equal to a multiple
of π. Note that the argument of the cosine on the left is an increasing function and

61

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 that on the right is a decreasing function of x. Hence the difference cannot be a
constant and we must have
Z Z
1 x 1 x2 π
p(x) dx + p(x) dx − = nπ (5.37)
~ x1 ~ x 2

giving Z x2
1
p(x) dx = (n + 1/2)π, (5.38)
~ x1

which we rewrite as Z x2
2 p(x) dx = (n + 1/2)h. (5.39)
x1

Note carefully that the right hand side of the above equation has Planck’s constant h
and not ~ = h/2π. Remembering that p(x) is the H classical momentum, the integral
in Eq.(5.39) is just the action integral J(E) = p(x) dx taken over one period, i.e.
around the classical orbit of the particle from x1 to x2 and then back to the point
x1 and the quantisation condition takes the form

J(E) = (n + 1/2)h. (5.40)

This condition, known as WKB quantisation rule for bound states, determines the
bound state energies without solving the Schrodinger equation and differs from the
Bohr Sommerfeld quantisation rule by presence of extra term h/2 in the right hand
side.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
17004;

§ 4.1 Connection Formula

QM-
Lec-
17005
⊲
Solution for a linear potential
In order to impose the requirements of continuity on WKB solutions, the potential
will be approximated by a linear term in the vicinity of a turning point and use will
be made of the fact that the Schrödinger equation can be solved exactly for a linear
potential.
Approximating the potential energy V (x) by a linear function in the neighbour-
hood of the turning point x0 ,

V (x) = V (x0 ) + (x − x0 ) + · · · . (5.41)

the resulting Schrödinger for a linear potential

~2 d 2 ψ

+ − V (x0 ) − F (x − x0 ) − E ψ=0 (5.42)
2m dx2

dV 
can be solved exactly. If x0 is a turning point, F = − dx  is the force at the turning
x0
point x0 and V (x0 ) = E, Eq.(5.42) takes the form

~2 d 2 ψ
+ F (x − x0 )ψ = 0. (5.43)
2m dx2
62

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 This equation can be solved exactly in terms of Airy function which are related
to the Bessel functions of order 2/3. Changing the independent variable to ξ =


2mF 1/3
~2
(x − x0 ), the Eq.(5.43) takes the form.
d2 ψ
+ ξψ = 0. (5.44)
dx2
The two linearly independent solutions are the Airy functions A(1) (ξ) and A(2) (ξ)
and the most general solution, to be denoted by Φ(ξ), is given by
Φ(ξ) = α1 A(1) (ξ) + α2 A(2) (ξ). (5.45)
Now we will discuss, in detail, how linear approximation to the potential near a
turning point can be used to find result following from the continuity requirements on
the WKB solutions. Let us first consider a case when the potential is impenetrable
to the left of a turning point at x0 , and the potential tends to ∞ as x → −∞.
Let R1 and R3 be the regions where the WKB approximation holds. Let R2 be the
region around the turning point x0 where the potential can be approximated by a
linear potential. In order that linear approximation can be used to derive connection
formula for the WKB solutions in R1 and R3 , it is important that R2 must overlap
with R1 and R3 as shown in 5.2(a).

Asymptotic behaviour For ξ < 0, i.e. , in region R1 as ξ → −∞ the asymptotic

expansion of the two Airy functions is given by
i|ξ|−1/4 2  1 |ξ|−1/4  2 
ξ→−∞
A(1) (ξ) −→ − √ exp |ξ|3/2 + √ exp − |ξ|3/2 (5.46)
π 3 2 π 3
−1/4 2  1 |ξ|−1/4  2 
ξ→−∞ i|ξ|
A(2) (ξ) −→ √ exp |ξ|3/2 + √ exp − |ξ|3/2 . (5.47)
π 3 2 π 3
The asymptotic behaviour of the Airy functions, as |ξ| → ∞ is given by
1  2i iπ 
ξ→−∞
A(1) (ξ) −→ √ |ξ|−1/4 exp |ξ|3/2 − (5.48)
π 3 4
1  2i iπ 
ξ→−∞
A(2) (ξ) −→ √ |ξ|−1/4 exp − |ξ|3/2 + . (5.49)
π 3 4
In order that, for the bound states, Φ given by Eq.(5.45) be well behaved for ξ → −∞
we must select α2 = α1 . Thus in the region I, to the left of the turning point in 5.1,
we get
Φ(ξ) = α1 (A(1) (ξ) + A(2) (ξ)) (5.50)
1  2 
−1/4 3/2
≈ α1 √ |ξ| exp − ξ ξ < 0. (5.51)
π 3
Also to the right of the turning point, in region R2 , the behaviour of the solution,
for α2 = α1 is given by
|ξ|−1/4   2i πi   2i πi 
Φ(ξ) ≈ α1 √ exp ξ 3/2 − + exp − ξ 3/2 + (5.52)
π 3 4 3 4
|ξ|−1/4 2 π

≈ 2α1 √ cos ξ 2/3 − , ξ>0 (5.53)
π 3 4

63

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The results obtained for linear potential can be summarised as follows. The solution
Φ(ξ) which vanishes as ξ → −∞ is given by
Φ(ξ) = α1 (A(1) (ξ) + A(2) (ξ)) (5.54)
  
α |ξ|√−1/4 exp − 2 ξ 3/2 , ξ < 0
1 π 3
≈ −1/4
(5.55)
2α1 √ cos ξ − π
|ξ| 2 2/3
ξ>0
π 3 4

The answer Eq.(5.55) obtained for a linear potential, expressed in terms of the
original variable x, denoted as ψ2 takes the form
 h √ i
√ 2α 2 2mF 3/2
4
exp − 3 ~
(x0 − x) , x < x0
2mF (x0 −x) h i
ψ2 (x) ≈ √ . (5.56)
 √4 2α cos − 2 2mF
(x − x ) 3/2
− π
x > x
2mF (x−x ) 3 ~ 0 4 0
0

Fig. 5.2. qm-cdr-wkb02;

The WKB solution to
the left of the turning point and vanishing at −∞ is
a1  1 Z x0 
ψ1 (x) = √ exp − |p(x)|dx (5.57)
|p(x)| ~ x
This solution is a good approximation in a region R1 . We now want to compare the
will now compare the above WKB solution in R1 with the Airy function solution.
For this it is important that the regions R1 and R2 overlap and in the overlap region
the use of the asymptotic formula is justified. This condition cannot be satisfied if
the potential changes very fast.
In order to Rdo compare the WKB solution with the Airy function solution, we
must compute |p(x)| dx in linear approximation. We have V (x) ≈ E − F (x − x0 )
and for x < x0
p p p
|p(x)| = | 2m(E − V (x)| = 2mF |(x − x0 )| = 2mF (x0 − x), . (5.58)
and hence
Z x0
2√
|p(x)| dx = 2mF (x0 − x)3/2 (5.59)
x 3
(5.60)
Therefore, for small |x0 − x|, the WKB solution ψ1 (x), (5.57),becomes

a1 h 2 √2mF i
3/2
ψ1 (x) = p exp − (x0 − x) , x < x0 (5.61)
4
2mF (x0 − x) 3 ~

64

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Comparing (5.61) with (5.55) for x < x0 we get a1 = α.
Similarly, the WKB solution ψ3 (x), for region R3 , in the linear approximation
for the potential becomes
a3 i Z x  b3  iZ x 
ψ3 (x) = p exp p(x)dx + p exp − p(x)dx , x ∈(5.62)
R3 ,
p(x) ~ x0 p(x) ~ x0
a3 h2 i √ i
3/2
≈ p exp 2mF (x − x0 )
4
2mF (x0 − x) 3~
b3 h 2 i√ i
3/2
+p exp − 2mF (x − x0 ) , ξ > 0 (5.63)
4
2mF (x0 − x) 3~
A comparison of (5.63) with ψ2 (x) in (5.56) for x > x0 gives
a3 = α exp(−iπ/4), b3 = α exp(iπ/4).
This leads to the following expression for ψ3 (x), to be denoted as χ(x),
Z x 
2a1 π
χ(x) = √ cos p(x)dx − (5.64)
p x0 4
If the potential tends to ∞ to the right of a turning point x0 , we must use the
solution 1 Z x 
b5
ψ5 (x) = √ exp |p(x)|dx (5.65)
|p(x)| ~ x0
Rx
which decays as x → ∞ again computing the integral x0 |p(x)| dx in the linear
approximation and comparing with Airy function solutions, as above, we are then
led to the following WKB solution to the left of the turning point
Z x0 
2b5 π
φ(x) = √ cos p(x)dx − . (5.66)
p x 4
Eq.(5.65)-(5.66) give the rules for connecting the WKB solution with classical region
on the left and an impenetrable barrier on the right.

Connection rules The results derived above can be summarised as follows. For
the case of an infinite barrier on the left of the turning at x0 , the decaying solution
on the left of turning point is connected to an oscillatory solution on the right by
the rule
 1 Z x0  Z x 
1 2 π
√ exp − |p(x)|dx → √ cos p(x)dx − . (5.67)
|p(x)| ~ x p x0 4
In the case of an impenetrable barrier to the right of the turning point, the connec-
tion rule becomes
Z x0   1Z x 
2 π 1
√ cos p(x)dx − ←√ exp − |p(x)|dx . (5.68)
p x 4 |p(x)| ~ x0
These rules can be used only in the indicated direction. In order that the use of
connection rules may be justified it is important that the two regions, in which WKB
solutions are valid and in which linear approximation for potential holds, must be
overlapping regions. Therefore, rules will fail for a potential which varies fast.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lec-
17005;

65

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §5 Tunnelling Through a Barrier
QM-
Lec-
17007
⊲
We shall briefly discuss the problem of computing the transmission coefficient
through a potential barrier using WKB scheme. Following Landau and Lifshitz
several simplifying assumptions will be made to get an approximate answer for the
transmission coefficient.

Fig. 5.3.
qm-cdr-wkb03;Tunneling Through a Barrier
In Fig. 5.3 the regions R1 and R5 denote the classical regions and R3 is nonclas-
sical region, all away from the turning points a, b, where the WKB approximation
holds. These are separated by regions R2 , R4 around the turning points where the
WKB approximation breaks down and linear approximation to the potential has
been used to derive the connection rules.
We assume the barrier to be very wide and almost impenetrable and take the
wave functions in regions R1 and R3 to be given by the WKB result:
  R 
 √2A 1 x

 cos ~ a
p(x) dx , x ∈ R1

 p(x) h i
A
R x
ψ(x) = √|p(x)| exp − ~1 a |p(x)| dx , x ∈ R3 (5.69)

 h R i

 √C exp ~i bx p(x) dx ,
 x ∈ R5 .
p(x)

For x to the right of the barrier in R5 , only transmitted wave must be present and
the form of the WKB solution above in region R3 has been chosen so as to represent
waves travelling to the right in R5 .
The above solution (5.69) in region R3 will be rewritten in the form
A h 1Z x i
ψ(x) = p exp − |p(x)|dx
|p(x)| ~ a
h Z Z i
A 1 b 1 b
= p exp − |p(x)| dx + |p(x)| dx x
|p(x)| ~ a ~ x
D h1 Z b i
= p exp |p(x)| dx ∈ R3
|p(x)| ~ x
. (5.70)

66

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 We use D to denote Z
1 b i h
D = exp − |p(x)| dx (5.71)
~ a
and write the WKB solution as
  R 
 √2A 1 x

 cos ~ a p(x) dx , x ∈ R1

 p(x) h i
D
R
1 b
ψ(x) = √|p(x)| exp − ~ x |p(x)| dx , x ∈ R3 (5.72)

 h R i

 √C exp ~i bx p(x) dx ,
 x ∈ R5 .
p(x)

It must be noticed that the connection rules, derived previously, are not applicable
to relate the coefficients C and D because the form of the above solution has a form
in region R3 and R5 different from that appearing in the connection formulae. A
connection between the above two forms for the wave function in regions R3 and R5
can be established by making use of the result that the Wronskian of two solutions
of the Schrodinger equation is independent of x.
dφ2 (x) dφ1 (x)
W (x) = φ1 (x) − φ2 (x) = constant. (5.73)
dx dx
To use this we take the wave function in Eq.(5.72) as φ1 . The WKB solution
appearing in the connection rule Eq.(5.67) , with classical region to the right of the
turning point, is taken as φ2 (x):
  
 √ 1 exp − 1 R b |p(x)|dx x ∈ R3
|p(x)| ~ x
φ2 (x) = R

2 cos x p(x)dx − π x∈R
b 4 5

We compute the Wronskian W (x) to the left and to the right of x = x1 and equate
the two values at x = x2 . Using the above results the transmission coefficient can
now be worked out and is found to be
 2  2Z b 
C 
T =   = exp − |p(x)|dx . (5.74)
A ~ a

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-22-04-

L1;

§6 WKB approximation for radial equation

QM-
Lec-
17006
⊲
For a particle moving in three dimension in a spherically symmetric potential V (r)
the wave function ψ(~r) has the form

ψ(~r) = Yℓm (θ, φ)R(r) (5.75)

where Yℓm (θφ) are spherical harmonics and R(r) is the radial wave function obeying
the radial equation
 
~2 1 d 2 d ℓ(ℓ + 1)~2
− r R(r) + V (r) + R(r) = ER(r) (5.76)
2m r 2 dr dr 2mr 2

67

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The WKB results for one dimension are not applicable directly to the radial equation
because its form is different from the one dimensional Schrödinger equation and the
range of r is from 0 to ∞ as against the range of x from −∞ to ∞.
This equation can be brought to a form similar to the Schrödinger equation in one
diemnsion introducing χ(r) by writing R(r) = χ(r)/r which satisfies the equation
 
~2 d 2 χ ℓ(ℓ + 1)~2
− + V (r) + − E R(r) = 0 (5.77)
2m dr 2 2mr 2

For a bound state the boundary conditions satisfied by χ(r) are

χ(0) = 0; χ(r) → 0 as r → 0 (5.78)

Eq.(5.77) is similar to one dimensional equation but boundary conditions Eq.(5.78)

are different from the corresponding conditions for a bound state in one dimension.
The variable r takes values in (0, ∞) instead of (−∞, ∞) as is the case in one
dimension.
By a change of variable from r to s

r = exp(s) (5.79)

and a transformation on χ(r)

s
χ(r) = e 2 φ(s) (5.80)
the equation for φ is easily found to be
 
d2 φ 2m s 2s ~2 1 2
+ 2 (E − V (e )) e − (ℓ + 2 ) φ = 0 (5.81)
ds2 ~ 2m

The range of s is seen to be −∞ to +∞ and the boundary condition on φ(s) =

r 1/2 R(r) takes the standard form

φ(s) = 0 (5.82)

as s → ± ∞. Thus we can take over the WKB result in one dimension and the
quantization condition can now be written as
Z s2  1/2
2s 2 1 2
2 2m(E − V (s))e − ~ (ℓ + 2 ) ds = (n + 12 )h (5.83)
s1

Changing the integration variable from s to r we get

Z r2  1/2
2 2m(E − V (r)) − ~2 (ℓ + 12 )2 dr = (n + 21 )h (5.84)
r1

It should be noted that the centrifugal term appears with coefficient (ℓ + 12 )2 instead
of the usual ℓ(ℓ + 1). The same result can also be derived by starting from Eq.(5.77)
and applying the boundary condition Eq.(5.78).
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-22-03-
L1;

68

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §7 Examples and Problems
QM- § 7.1 Tie Loose Ends
22-
05- ⊲
Q1 1. Relate the Schrödinger equation for a linear potential to the Bessel’s equation.

2. Relate the Schrödinger equation for xn potential to the Bessel’s equation.

3. A classical particle executes a periodic motion between two turing points x1

and x2 with velocity v which depends on its positon x. Classically does the
probability, that it will be found in region beween x and x + dx, depend on
its velocity at the point x? If yes what should be the velocity dependence of
√
this probability. Relate this to the factor 1/ p present in the semiclassical
solution

4. List the approximations, other than semi-classical approximation, made in the

derivation of the tunneling formula.

5. What substitutes the matching requirement in the derivation of the transmis-

sion formula.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-22-05-

Q1;

§ 7.2 Learn to Apply what you Learnt

[1] For a particle incident on a potential barrier
(
0 x<∞
V (x) =
V0 (1 − x/a) x > 0

find the transmission coefficient using WKB approximation.

[2] Show that the WKB approximation gives exact answer for energy eigenvalues
of a harmonic oscillator.

[3] Using WKB approximation find the energy levels of a particle in a gravitational
potential of the earth. You may assume the acceleration due to gravity to
remain constant and that the surface of the earth is a rigid boundary.

§ 7.3 Complete the Exercises to the Bitter End

[1] Does the WKB quantization rule give exact answer for a particle in one di-
mensional box?

[2] Obtain the energy quantization condition using WKB method for a particle
in one dimensional well with potential given by

V0 0 < x < a

V (x) = 0 a < x < a + b


∞ x > a or x > a + b

69

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Solve the Schrödinger equation exactly and derive the exact quantization con-
dition. Show that the WKB answer agrees with the exact condition in the
limit of large quantum number. [ Mavromatis]
[3] Find the energy levels of a particle in well
V (x) = A|x| A>0
using WKB approximation. How does your answer compare with the large n
limit
2  3π  3  A2 ~2  3
2 1
En = (n − 1/4) 3 √
2 2 m
of the exact answer. [Mavromatis]
[4] Show that the energy levels of a particle in potential well
V (x) = A|x|p A>0
using the WKB quantization rule are given by
 
1 2p/(p+2) 2π~|A|1/p
En = (n + 2 ) √ Ip
2m
where I(p) is given by
Z 1p
I(p) = 2 1 − |x|p dx
0

[Mavromatis]
[5] Find the energy levels of a particle in a potential well
V (x) = A|x|1/2
using the WKB approxmation.
 15~A2  25
En = n2/5 √
8 2m

[6] Find the energy levels of a particle moving in a one dimensional potential
1
V (x) = −A|x|− 2
using WKB approximation and argue that an infinte number of levels are
possible
2  4~|A|−2 − 3
2
−
En = −n 3 √
2m
[7] Use WKB quantization condition to argue that the number of levels for a
potential well
V0
V (x) = − 4
(x + a4 )
cannot be infinte.

70

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 [8] What should be the condition on α so that the potential
V0
V (x) = −
(x2 + a2 )α
may have infinite number of levels?
[9] Formulate the WKB quantization condition for a two dimension potential
problem in plane polar coordinates ρ, φ when the potential depends is inde-
pendent of φ. Using your result find the energy levels of the isotropic harmonic
oscillator in two dimension with
V (ρ) = 21 mω 2 ρ2

§ 7.4 Establish Cross Links with Other Topics, Techniques

[1] Solve the bound state energy eigenvalue problem and rederive the energy quan-
tization condition, obtained in QM-I, for a square well potential
(
−V0 0<x<L
V (x) =
0 otherwise

following the steps given below.

(a) Write the most general solutions of the Schrodinger equation in the three
regions as indicated below
Region I : x < 0 u1 (x) = A1 exp px + A2 exp(−px)

Region II :0 < x < L u2 (x) = B1 cos kx + B2 sin kx

Region III x > L u1 (x) = C1 exp px + C2 exp(−px)

Taking E < 0 obtain k, p, in terms of energy.

(b) Demand the boundary condition that the wave function be zero for x →
−∞ and fix the constants A1 and A2 to the extent possible.
(c) Use the matching condition that the wave function and its derivative
should be continuous at x = 0. Using these two conditions, eliminate the
constants B1 and B2 in favour of the constants in u1 . This gives an answer
for the wave function in region-II. Let this solution be called V (x).
(d) Now repeat the above two steps for the regions-III and region-II and
region III. First use the boundary condition on the wave function as
x → ∞ and find a condition on C1 , C2 . Next use the matching conditions
at x = L and using the resulting two equations to get another solution
for B1 and B2 in terms of C1 , C2 . Denote this answer as W (x).
(e) We now have two answers, V (x) and W (x), for solution in region II.
These two must agree for all x in region II. Show that this requirement
leads to the correct quantization condition for bound states for the square
well potential obtained in the last semester.

71

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 [2] Why is the square well problem being solved again? Does it help you in under-
standing square well better? What did you learn by solving this assignment?
[3] Write a one, maximum two page summary of steps in the WKB method for
bound states.

§ 7.5 Search, Discuss and Explore

[1] Consider a particle moving in dimension confined to move on half line x > 0
by a potential which is infinity for x < 0 and tends infinity as x → ∞.
Explain why the standard WKB quantization rule cannot be used and derive
the modified WKB quantization formula for the bound state energy levels.
[2] It is said that the WKB approximation fails near a turning point on when
the potential varies very fast. In your opinion is the method applicable to a
discontinuous potential such as a square well potential? Give reasons.
[3] Estimate the number of bound states a potential can have. Apply your result
to some exactly solvable models and compare with known results.
[4] Use WKB approximation and argue that a finite range potential cannot have
infinite number of bound states.
[5] A potential which rises to infinity as x → ±∞ will have infinite number of
bound states. What about a potential which does not go to infinity as x
increases? What property is necessary so that it may have infinite number of
bound states.
[6] A well known example of correspondence rule is that the frequency of H-atom
spectral line for a transition from n to n−1 level, for large n, coincides with the
classical frequency of the orbit. Formulate and prove a corresponding result
for an arbitrary potential.
[7] The scheme to compute WKB wave function can be generalized to the case
when the turning point is zero of V (x) − E of order n. Work out the steps an
derive the connection formula.
[8] What will be the connection formula if the turning point is a point where the
potential is discontinuous.
[9] What modification if any will be needed in applying the WKB scheme to a
piece wise discontinuous potential?
[10] The WKB scheme gives exact results for a few potentials, such as the harmonic
oscillator, particle in a box. Is there any other potential for which WKB, or a
slight variation, will give an exact result.
[11] Consider a periodic potential over a period (−L, L). Can I use WKB like
scheme to obtain wave functions and energies for the band edges?
[12] What information does the WKB method resonances in the transmission am-
plitude?

72

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 6

Scattering Theory

§1 Introduction
In this unit description of scattering and scheme of computation of cross section in
quantum mechanics is introduced. This is achieved by imposing a suitable bound-
ary condition on the solution of the time independent Schrödinger equation and
converting Schrödinger equation into an integral equation using the Green function
for the free particle Schrödinger equation. A perturbative solution of the integral
equation leads to the Born approximation for the scattering amplitude.

§2 Scattering in classical and quantum mechan-

ics
QM-
Lecs-
18001
⊲
Let us consider a scattering experiment in which a beam of particles is scattered
from a target at rest. The frame of reference in which the target is at rest will be
called the laboratory frame. After the scattering the particles, at large distances,
will be moving, away from the target, like free particles.
We assume the potential between an incident particle, position r~1 , and the tar-
get, at position r~2 , to be central potential V which depends only on the relative
position,~r = r~1 − r~2 , of the particle and the target. We recall that the two particle
problem can then be reduced to the problem of one particle of reduced mass moving
in a potential V (~r). The cross section calculated for a particle moving in potential
V (~r equals the scattering cross sections in the centre of mass frame and must be
transformed to the lab frame to establish contact with experiments.
Knowledge of the classical trajectory of a particle, ~r(t) is sufficient to compute
the cross section in classical mechanics. The classical physics being deterministic,
the particles going into solid angle corresponding to a cone, covering small range
θ, θ +dθ of the scattering angle, are precisely those which come from a corresponding
range of impact parameter b. Hence we only need to know the relation between the
impact parameter and the scattering angle to compute the differential cross section
σ(θ). The result is known to be
 
b  db 
σ(θ) = (6.1)
sin θ  dθ 

73

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 In quantum mechanical framework the particles do not have a well defined trajec-
tory and it is not meaningful to associate a well defined range of impact parameters
with a given range of scattering angle. All information about a system has to be
obtained from the wave function and must be extracted from the available statistical
interpretation of the wave function.
The scattering process, like any other motion, is a problem of time evolution. In a
scattering experiment, the incident particle is far off from the target and approaches
towards the target reaching a point of closest approach. After that it moves away
from the target and goes to infinity. A wave packet description the motion of a par-
ticle, in accordance with the time dependent Schrödinger equation, is the framework
for a rigorous and a complete description of the scattering problem. It turns out
that the scattering process can also the viewed as a stationary state problem and
solution of the time independent Schrödinger equation turns out to be adequate as
a first introduction for our present purpose.
Let us consider a thought experiment in which a beam of particles is incident and
getting scattered for all times from −∞ to +∞. If take a snap shot of the beam in
the experiment, it would look the same at all times. It should therefore not come as a
surprise that one treat the scattering in terms of using stationary states. This is not
something completely new, we are already used to treating motion of electrons in an
atom as a stationary process in quantum mechanics. We will, therefore, formulate
the scattering problem in terms of stationary state solutions, i.e., the solutions of
the time independent Schrodinger equation.

Boundary conditions for the wave function

Assuming a spherically symmetric potential V (r) which goes to zero for large dis-
tances, E < 0 corresponds to the bound states and the continuous energy solution
for E > 0 is needed for a discussion of to the scattering. The Schrödinger equation
~2 2
− ∇ ψ + V (r)ψ = Eψ (6.2)
2µ
has an infinite number of solutions E > 0. To understand this we look at the free
particle solutions. For a given energy the free particle solutions can be written as

ψ(~r) = N exp(ikn̂ · ~r), (6.3)

p
where k = 2µ E/~2 and n̂ is a unit vector. For a fixed energy E there are infinitely
many plane wave solutions corresponding to the direction of propagation specified
by the unit vector n̂. Alternately, the solutions can also be written in terms of
spherical waves of definite angular momentum

ψ(~r) = Cjℓ (kr)Yℓm (θ, φ) (6.4)

The most general solution will be a superpostion of the above special solutions. The
free particle behaviour will hold for the scattering solutions for a potential which
goes to zero for large distances, giving infinite number of solutions. Thus specifying
the energy alone is not sufficient to pick a unique solution, it is necessary to specify
a boundary condition suited to the scattering problem.

74

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 In the stationary description, the solution to Schrodinger equation should de-
scribe the incident beam and an outgoing scattered wave. For short range poten-
tials this will be a spherical wave with varying amplitude in different directions.
We demand that the wave function for a definite energy must satisfy the following
boundary condition in the limit r → ∞:
eikr
lim ψ(~r) → eikz + f (θ) . (6.5)
r
The above choice of the boundary condition requires an explanation. The first term
has been written for the choice of z axis along the incident beam of definite energy
E. In general, for incident beam having momentum ~k ~ i, one must replace the first
~
term by exp(−iki · ~r). The second term represents an outgoing spherical wave, note
that the time dependence is e−iEt/~. The factor f (θ)/r represents the amplitude of
the wave at large distances in the direction θ. Since the intensity of the scattered
beam decreases as 1/r 2 for large distance, the amplitude, for large r, must decrease
as 1/r. At present it is not clear that such a solution does exist, in a later subsection
we will show that a solution satisfying the boundary condition Eq.(6.5) does exist.

Cross section in quantum Theory

Now we come to computation of cross section for scattering from a potential spherical
symmetric, finite range potential V (r). We wish to relate the differential cross
section to the scattering amplitude f (θ).
Let us consider a scattering experiment involving a total of N incident particles
sent in time T . The number of particles detected by a detector in a direction θ will
be proportional to the flux of the incident beam and the solid angle subtended by
the detector and the constant of proportionality is just the differential cross section.
For a detector placed at a distance r from the target and having an opening area
∆S, the solid angle will be ∆Ω = ∆S/r 2 .. Thus we would get

No of particle detected per sec = σ(θ) × ∆Ω × Flux of the incident beam. (6.6)
knowing the wave function, the number of particles detected per second can be
computed using the probability current density ~j. The opening of the detector is
kept perpendicular to the radius vector and usually covers only a small solid angle,
the probability of a particle entering the detector per sec is given by the surface
integral x
jr dS ≈ jr ∆S (6.7)
S
over the surface of the detector, where jr is the radial component of the probability
current. The number of particles detected per sec will be N times the expression
(6.7). Thus Eq.(6.6) becomes
N × jr ∆S = σ(θ) × ∆Ω × Flux of the incident beam = σ(θ)∆Ω × Njz . (6.8)
where jz is the z component of the probability current. Using


~j = i~ ψ ∗ ∇ψ − ψ∇∗ ψ . (6.9)
2µ

75

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 the z component of the probability current for the incident beam eikz is easily found
to be ~k/µ. Also the radial component of the of the probability current for the
scattered wave is obtained by substituting f (θ)eikr /r for ψ(r) in Eq.(6.9) and taking
the radial component. Using
d 1 d 1 d
∇ = r̂ + θ̂ + φ̂ (6.10)
dr r dθ r sin φ dφ
the most important term in the radial component of the current for the scattered
wave becomes
jr = |f (θ)|2 /r 2 + O(1/r 3). (6.11)
Using (??) and (6.11) in Eq.(6.8) gives

~k
N(|f (θ)|2 /r 2 )∆S = σ(θ) × ∆Ω × N ,. (6.12)
µ

With ∆S = r 2 ∆Ω, we get the desired relation

σ(θ) = |f (θ)|2 . (6.13)

We have ignored the terms in the current coming from the interference of the in-
cident and scattered waves. These are of the order of 1/r 2, and are proportional
to exp(ikr(1 − cos θ)). Due to the presence of large r in the exponential, this term
oscillates rapidly with θ and sums out to vanishingly small value over range of θ
values.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lecs-
18001;

§3 Integral Equation
QM- A simple example
18-
01- ⊲
L1 Very often the problem of solving a linear differential equation can be replaced with
solution of an integral equation. An important feature of the integral equation
approach is that the initial conditions to be satisfied by the solution is built into the
integral equation. The integral equation can be solved many a times by an iterative
procedure which we shall illustrate by a simple example.
Suppose we are interested in solving the differential equation
dy
= λy (6.14)
dx
subject to the boundary condition

y(x)|x=x0 = N (6.15)

where N is a constant. To convert Eq.(6.14) into an integral equation, we integrate

Eq.(6.14) to get Z
y(x) = λ y(x)dx + constant (6.16)

76

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 or to be more precise, let us write Eq.(6.16) as
Z x
y(x) = λ y(t)dt + constant (6.17)
0

The constant in the above equation is fixed by making use of the initial condition
Eq.(6.15), and we get const= N and
Z x
y(x) = N + λ y(t)dt (6.18)
0

It is to noted that the unknown function y appears inside the integral sign, hence
an equation of this type is called integral equation. This equation can be solved
iteratively giving a solution as a series in powers of λ. This method gives the exact
answer in the case of this simple example under consideration. As a first step, we
set y equal to y0 where
y0 = N (6.19)
is the solution in the zeroth order in λ and is simply taken to be equal to the first
term Eq.(6.18). Next the zeroth order ’solution’ ,y0 , is substituted in the right hand
side of Eq.(6.18) to get the solution in the first order in λ. Thus we have
Z x
y1 (x) = N + λ y0 dx (6.20)
Z0 x
= N +λ Ndx (6.21)
0

∴ y1 (x) = N(1 + λ)x (6.22)

To improve the approximation, we substitute y1 (x) for y(x) in right hand side of
Eq.(6.15) to get the next approximation y2 (x) for our solution.
Z x
y2 (x) = N + λ y1 (x)dx (6.23)
Z0 x
= N +λ N(1 + λx)dx (6.24)
0
 2

2x
= N 1 + λx + λ (6.25)
2

continuing in this fashion we get

Z x
y3 (x) = N + λ y2 (x)dx (6.26)
0
Z x 
x2
= N + Nλ 1 + λx + λ dx (6.27)
0 2
 
λ2 x2 λ3 x3
= N 1 + λx + + (6.28)
2 3!
 2 2 3 3

λ x λx λ4 x4
y4 (x) = N 1 + λx + + + (6.29)
2 3! 4!

77

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Thus we get an infinite series in powers of λ
 
λ2 x2 λ3 x3
y(x) = N 1 + λx + + + ... (6.30)
2 3!
summing the series we get
y(x) = Neλx (6.31)
Note that this is the correct solution for ordinary differential equation and satisfies
the given boundary condition y(0) = N.
Remark:
The above method of solution is equivalent to the following method. we want to
solve Z x
y(x) = N + λ y(t)dt (6.32)
0
we assume that the solution can be written as a series in λ

y(x) = α0 + λα + λ2 α2 + · · · + αn λn + · · · (6.33)

we substitute Eq.(6.33) in Eq.(6.32) and compare powers on both the sides. There-
fore we get
Z x
2 n

α0 (x)+λα1 (x)+λ α2 (x)+. . . λ αn (x)+· · · = N+λ α0 (t) + λα1 (t) + λ2 (t)α2 (t) + · · · dt
0
(6.34)
on comparing coefficients of different powers of λ we successively get

α0 (x) = N
Z x
α1 (x) = α0 (t)dt = Nx
0
Z x
x2
α2 (x) = α1 (t)dt = N
2
Z0 x
x3
α1 (x) = α2 (t)dt = N
3!
Z0 x
xk
αk (x) = αk−1 (t)dt = N (6.35)
0 k!
Therefore we get
 
λ2 2 λn n
y(x) = N 1 + λx + x + · · · + x + · · · (6.36)
2! n!

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-18-01-

L1;

QM- Green function for Poisson equation

18-
01- ⊲
L2 In electromagnetic theory the electric potential satisfies the Poisson equation
ρ
∇2 Φ = − , (6.37)
ε0

78

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 where ρ(~r) is the volume charge density. The Green function for the Poisson equation
is defined by
∇2 G(~r) = −δ 3 (~r) (6.38)
If φ0 (~r) is a solution of the Laplace equation

∇2 φ0 (~r) = 0, (6.39)

then Z
ρ(~r′ ) 3 ′
Φ(~r) = φ0 (~r) + G(~r − ~r′ ) dr (6.40)
ε0
is a solution of the Poisson equation which can be easily verified by applying ∇2 on
both sides of Eq.(6.40).
Z
2 2 2 ρ(~r′ ) 3 ′
∇ Φ(~r) = ∇ φ0 (~r) + ∇ G(~r − ~r′ ) dr (6.41)
ε0
Z
ρ(~r′ ) 3 ′
= ∇2 G(~r − ~r′ ) dr (6.42)
ε0
Z
1
= − δ 3 (~r − ~r′ )ρ(~r′ )d3~r′ (6.43)
ε0
ρ(~r)
= − (6.44)
ε0
It can be shown that one solution of Eq.(6.38) is
1
G(~r) = . (6.45)
4πr
This Green function gives the potential due to a charge distribution subject to
the condition that the potential vanishes at infinity. The exact form of the Green
function and the solution φ0 of the Laplace equation is determined by the boundary
conditions of the problem.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-18-01-
L2;

QM-
§4 Integral Equation for Scattering
18-
02- ⊲
L1 In order to convert the Schrodinger equation
 2 
−~ 2
∇ + V (r) ψ = Eψ (6.46)
2µ
into an integral equation we first rewrite it as


∇2 + k 2 ψ = U(~r)ψ(~r) (6.47)
2µE 2µ
where k 2 = 2 , U(r) = 2 V (r) and we have defined Green function G(~r) as a
~ ~
solution of

∇2 + k 2 G(~r) = −δ(~r) (6.48)

79

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 and it is found that several solutions of Eq.(6.48) exist. The solutions G(−
→
r ) for can
be written
1
G± (~r) = exp(±ikr) (6.49)
4πr
1
G0 (~r) = cos(kr) (6.50)
4πr
using the Green function we can now write down a ”formal” solution of Eq.(6.47)
as a solution of integral equation. Next we shall determine the behaviour of the
solution as r → ∞ and show that the choice
1
G(~r) = exp(ikr) (6.51)
4πr
leads to the correct boundary condition on the wave function. Using a Green func-
tion which is a solution of Eq.(6.48) a formal solution for Eq.(6.47) can be written
as Z
ψ(~r) = φ(~r) − G(|~r − ~r ′ |)U(|~r ′ |)ψ(~r ′ )d3 r ′ (6.52)

where φ(~r) is a solution of the equation

∇2 + k 2 φ(~r) = 0 (6.53)

For the scattering problem we must select

φ(~r) = exp(ik~i .~r) (6.54)

where k~i is the momentum of the incident particles.Substituting Eq.(6.51) Eq.(6.54)

in Eq.(6.52) we get the integral equation for the scattering to be
Z ~′
1 expik|~r−r | ~′
ψ(~r) = exp(ik~i .~r) − U(r )ψ(r~′ )d3 r′ (6.55)
4π |~r − r~′ |

To verify that ψ(~r) given by Eq.(6.55) does indeed have correct asymptotic property
~r
we expand |~r − r~′ | in powers of we shall assume that the potential is short range
~
r′
potential so that the contribution to integral over r~′ comes from small value of r ′ .
Expand |~r − r~′ | in powers of r ′
√
|~r − r~′| = r 2 + r ′2 − 2~r · ~r′ (6.56)
 1/2
~r · ~r′ r ′2
= r 1−2 2 + 2 (6.57)
r r

Using binomial expansion we get

 2 !
~r · ~r′ r ′2
|~r − r~′ | = r 1 − 2 + O (6.58)
r r2

80

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 1
We substitute Eq.(6.58) in the exponential and in the factor in Eq.(6.55)
|~r − r~′ |
1 1
and write ≈ to get
|~r − r~′ | r
Z  
~ 1 ~r · ~r′
ψ(~r) −→ exp(iki · ~r) − exp ikr − ik 2 U(~r′ )ψ(~r′ )d3 r ′ (6.59)
4πr r
ikr Z
e
= exp(i~ki · ~r) − exp(−ikn̂ · ~r′ )U(r ′ )ψ(r ′)d3 r ′ (6.60)
4πr
In the last step we have introduced a unit vector n̂ = ~r/r. The Eq.(6.60) gives
the probability amplitude ( wave function ) at ~r. If the particles are to reach at
a detector at ~r, the vector n̂ must be in the direction of the final momentum and
parallel to kf . Note that
|~ki| = |~kf | = k (6.61)
holds as a consequence of energy conservation and hence
k(n̂ · ~r′ ) = ~kf · ~r (6.62)
Thus Eq.(6.60) takes the form
Z
eikr
ψ(~r) ≈ exp(i~ki · ~r) − exp(−i~k · ~r′)U(r ′ )ψ(r ′ )d3 r ′ (6.63)
4πr
This asymptotic behaviour is of the for expected for large r
eikr
ψ(~r) ≈ exp(i~ki · ~r) −
f (θ, φ). (6.64)
r
Comparing Eq.(6.63) with Eq.(6.64) we see that the scattering amplitude is given
by
Z
1
f (θ, φ) = − exp(−i~kf · ~r′ )U(r ′ )ψ(r ′ )d3 r ′ (6.65)
4π
 µ Z
= − exp(−i~kf · ~r′ )V (r ′ )ψ(r ′ )d3 r ′ (6.66)
2π~2
It must be noted that the integral equation Eq.(6.55) and the expression for the
scattering amplitude in Eq.(6.66) are exact results.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-18-02-
L1;

We shall next discuss how to use Eq.(6.55) and Eq.(6.66) to obtain scattering
amplitude in the Born approximation.

QM-
§5 Born Approximation
18-
02- ⊲
L2 We have seen that the energy eigenfunctions, with a correct asymptotic behaviour
corresponding to the scattering solutions satisfy the following integral equation
Z
~ 1 exp{ik|~r − ~r ′ |}
ψ(~r) = exp(iki .~r) − U(~r ′ )ψ(~r ′ )d3 r′ (6.67)
4π |~r − ~r ′ |
where we use the notation

81

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 ~ki = momentum of the incident beam
~kf = momentum of the scattered beam

θ= scattering angle = angle between ~ki and ~kf .

V (~r) = potential due the target
µ= mass of the incident particle ( reduced mass for two body problem)
2µ
U(~r)=V (~r)
~2
An iterative solution of the integral equation can be obtained by assuming that in
the lowest order approximation ψ(~r) is equal to ψ0 (~r) given by
ψ0 (~r) = exp(i~ki · ~r) (6.68)
Using this approximation for ψ(~r) from Eq.(6.68) in the right hand side of Eq.(6.67),
we get the next order solution, denoted as ψ1 (~r), given by
Z
~ 1 exp{ik|~r − ~r′ |}
ψ1 (~r) = exp(iki · ~r) − U(~r′ ) exp(i~ki · ~r′ )d3 r′ (6.69)
4π |~r − ~r′ |
The next approximation to the solution, ψ2 (~r), is obtained by replacing ψ(~r) in
Eq.(6.67) with ψ1 (~r). Thus
Z ik|~r−~r′ |
ik~i .~
r 1 e
ψ2 (~r) = e − U(~r′ )ψ1 (~r′ )d3 r′ (6.70)
4π |~r − ~r′ |
Z ik|~r−~r′ |
i~ki ·~
r 1 e ~ ′
= e − ′
U(~r′ )e(iki ·~r ) d3 r′ (6.71)
4π |~r − ~r |
Z ik|~r−~r′ | Z ik|~r′−~r ′′ |
1 e ~ e ~ ′ ′′
+ U(r ) ′ U(~r ′′ )eiki ·(~r +~r ) d3 r ′ d3 r ′′ (6.72)
(4π) 2 ′
|~r − ~r | ~ ′
|r − r | ~ ′′

This process can be continued indefinitely and it becomes very cumbersome to com-
pute the wave function beyond first few orders.
The first order Born approximation consists in using the first, the plane wave,
term in the above series as approximate wave function in the expression, Eq.(6.66) ,
for the scattering amplitude
Z
1
f (θ, φ) = − exp(−i~kf · ~r′ )U(r ′ )ψ(r ′ )d3 r ′ (6.73)
4π
 µ Z
= − exp(−i~kf · ~r′ )V (r ′ )ψ(r ′ )d3 r ′ (6.74)
2π~2
giving Z
µ ~ ~
f (θ, φ) ≈ − ei(ki −kf )·~r V (r)d3r (6.75)
2π~2
or Z
µ
f (θ, φ) ≈ − e−i~q·~r V (r)d3 r (6.76)
2π~2
where ~q = ~ki − ~kf is the momentum transfer. The result Eq.(6.76) is the well known,
first order, Born approximation result for the scattering amplitude.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-18-02-
L2;

82

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM- Validity of Born approximation
18-
02- ⊲
L3 The integral equation for the energy eiger functions was derived to be
 µ  Z eik|~r−~r′ |
ik~i .~
r
ψ(~r) = e − V (~r′ )ψ(~r′ )d3 r′ (6.77)
2π~2 |~r − ~r′ |
In the derivation of the integral equation it is assumed that it is good approximation
to take ψ(~r) to be plane wave
ψ(~r) ≈ exp(i~ki · ~r) (6.78)
and substitute in the right hand side of Eq.(6.77) . The second term on the right
hand side of Eq.(6.77) gives a correction to plane wave form. If Eq.(6.78) is a
good approximation to the wave function, the correction must be small compared
to the plane wave term. Hence the condition, under which the Born approximation
is valid, is given by
 Z  
 µ  eik|~r−~r′ |   ik~ .~r 
 V (~r )ψ(~r )d r′ < e i 
′ ′ 3
(6.79)
 2π~2 |~r − ~r′ |
or  Z 
 µ  eik|~r−~r′ | 
 V (~r ′
) exp(i ~ki · ~r )d r′ < 1
′ 3
(6.80)
 2π~2 |~r − ~r′ | 
The effect of the potential is to distort the wave function and make it different from
the plane wave and clearly this distortion is expected to be maximum where the
potential is large. The potential is assumed to tend to zero as r → ∞, assuming
that the potential is of short range and that most significant effect comes for r ≈ 0,
we apply the condition Eq.(6.80)for r = 0 and get
 Z 
 µ  eikr′ 
 V (~
r ′
) exp(i ~ki · ~r )d r′ < 1
′ 3
(6.81)
 2π~2 r′ 
Changing the integration variable name from r ′ to r, assuming the potential to be
spherically symmetric we get
 µ  Z 2π Z π Z ∞ ikr
e


 dφ sin θdθ V (r)e r dr  < 1
ikr cos θ 2
(6.82)
2π~ 2  r
0 0 0

The φ integration gives 2π and the θ integral is

Z π Z 1
ikr cos θ
sin θdθe = exp(ikrt)dt (6.83)
0 −1
1

= eikr − e−ikr (6.84)
ikr
Substituting Eq.(6.84) in Eq.(6.82) we get
Z 
µ  ∞ 2ikr

<1
e − 1 V (r)dr (6.85)
~2 k  0 

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-18-02-
L3;

In the next subsection we shall derive, the condition for validity of the Born
approximation for the square well potential.

83

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM- Square Well potential
18-
02- ⊲
L4 For a square well potential of strength V0 and range R0 , the expression in the left
hand side Eq.(6.85) takes the form
Z   
µV0  R0 2ikr

 µV0  e2ikR0 − 1 

e − 1 dr = − R0 (6.86)
~2 k  0  ~2 k  2ik 
µV0  2ikR0 

= e − 2ikR 0 − 1 (6.87)
2~2 k 2
Using the notation ρ = 2kR0 the condition, that the Born approximation be valid,
takes the form
µV0
1
2 2
ρ2 − 2ρ sin ρ − 2 cos ρ + 2 2 << (6.88)
2~ k
we shall consider the low energy and high energy cases separately.

Low energy scattering At low energy the de Broglie wave length is much larger
than the range of the potential i.e. 2kR0 << 1. We then have

ρ2 − 2ρ sin ρ + 2 cos ρ
   
2 ρ3 ρ2 ρ4
≈ ρ − 2ρ ρ − + ···) − 2 1 − + +··· +2
6 2 24
ρ4
= (6.89)
4
Hence at low energies the Born approximation is applicable if

µV0 ρ2 µV0 R02

= << 1 (6.90)
2~2 k 2 2 ~2
The above condition implies that the potential is so weak that the bound state does
not exist.

High energy limit In the high energy limit ρ >> 1 and we get

12
ρ2 − 2ρ sin ρ − 2 cos ρ + 2 ≈ρ (6.91)

and hence the Born approximation is valid if

µV0
ρ << 1 (6.92)
2~2 k 2
or
µV0 a
<< 1 (6.93)
~v
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-18-02-
L4;

84

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §6 Problems and Cross Links
Tutorial
[1] Calculate the differential cross section in Born approximation for scattering of
particle of mass µ from a δ function potential given by

V (~r) = gδ 3 (~r)

[2] A nucleon is scattered from a heavy nucleus. The effect of the nucleus can be
represented by a square well potential

−V0 r < R
V (r) =
0 r>R
Compute the differential cross section in the Born approximation.
[3] Classically the total cross section for scattering from a hard sphere of radius
R is πR2 . What happens if you try to compute the quantum mechanical cross
section using Born approximation? Discuss and give your comments.

[4] Compute the differential cross section in first Born approximation for the gaus-
sian potential
V (r) = −V0 exp(−r 2 /a2 )
and show that the total cross section is given by
! !
σt = V02 π 2 µ2 a4 /2~4 k 2 1 − exp(−2a2 k 2 )

Verify that the total cross section has correct dimension of area.

[5] For 2 MeV neutrons hitting protons at rest, find out if the Born approximation
is a good approximation. Consider nuclear forces to be described by a square
well of depth V0 = −35MeV and range R = 2.5 fm.

Cross Links
[1] Verify that the Green function
m
GE (x) = −i exp(ik|x|)
~2 k
p
where k = 2mE/~2 satisfies the one dimensional equation
 2 2 
~ d
+ E GE (x) = δ(x) (6.94)
2m dx2

[2] Using the Green function given in Q[1]

(a) Set up the integral equation for reflection and transmission coefficients
for the one dimensional problems.

85

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 (b) Verify that the Green function gives the correct boundary conditions for
the wave function at infinity.
(c) Solve the integral equation exactly for V (x) = −γδ(x) and find the re-
flection coefficient.

[3] (a) Use the Green function as in Q[1] and find an approximate expression for
reflection coefficient for an arbitrary potential V (x).
(b) Show that for a square well potential

−V0 for |x| < a2
V (x) =
0 otherwise

approximate reflection coefficient is given by

m2 V02
R≈ 4 4
sin2 ka
~k

86

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 7

The Method of Partial Waves

§1 Introduction
We shall now take up the metho of partial waves for calculating cross section. This
method, applicable to the spherically symmetric potentials, tyrns out to be partic-
ularly useful for short range potentials at low energies. We shall, therefore, restrict
our attention to the spherically symmetric potentials V (r) which for large r go to
zero faster than 1/r 2 , i.e., r 2 V (r) → 0.

QM-
§2 Partial Wave Expansion of Plane Waves
19-
01- ⊲
L1 The free particle Schrodinger equation in three dimensions
~2 2
− ∇ ψ = Eψ (7.1)
2µ
can be solved by separating the variables in the cartesian as well as the spherical
polar coordinates. Writing the above equations as

∇2 ψ + k 2 ψ = 0, (7.2)

where
2µE
k2 = , (7.3)
~2
solution of the free particle Schrodinger equation in the cartesian coordinates is
given by
ψ~k (~r) = exp(i~k · ~r), ~k = (k1 , k2 , k3) (7.4)
Using ~n to denote the direction of ~k to write ~k = kn̂, we see that for each energy
value E > 0 the degeneracy is infinite; there being one solution for each n̂, corre-
sponding to the direction of propagation of the particle. Note that these solutions
are2 eigenvectors of a complete commuting set of operators consisiting of the three
momentum operators ~pˆ and the free particle Hamiltonian.
The free particle Schrodinger equation Eq.(7.2) can also be solved by separating
variables in spherical polar coordinates and we get the solution

ψEℓm (~r) = jℓ (kr)Yℓm(θ, φ). (7.5)

87

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 For a given energy E, again one finds that there are infinite number of solutions,
one for each value of ℓ and m, here ℓ = 0, 1, 2, · · · and m can take integral values
from −ℓ to ℓ. The eigenvectors in Eq.(7.5) are also simultaneous eigenvectors of
~ 2 and L
L ~ z which together with the free particle Hamiltonian form another set of
commuting operators.
Summarizing the above results, we see that for E > 0 there are two infinite sets
of eigenfunctions given by
Simultaneous energy, momentum eigenfunctions
n  o

ψ~k (~r) = exp(i~k · ~r)~k = kn̂, all n̂ . (7.6)

Simultaneous energy, angular momentum eigenfunctions

n  o

ψEℓm (~r) = jℓ (kr)Yℓm(θ, φ)ℓ = 0, 1, 2, · · · ; m = −ℓ, −ℓ + 1, · · · , ℓ . (7.7)

The two sets of eigenfunctions in Eq.(7.6) and Eq.(7.7) form two different bases for
in the subspace of solutions of given energy E; every function in Eq.(7.6) can be
expanded in terms of the functions in the set Eq.(7.7) and also, every function in
Eq.(7.7) can be expanded in terms of the functions in the set Eq.(7.6). Therefore, we
can write the plane wave solutions, Eq.(7.6), as linear combinations of the spherical
wave solutions Eq.(7.7).
X
exp(i~k · ~r) = Cℓm jℓm (kr)Yℓm (θ, φ). (7.8)
ℓm

We shall now consider a special case when ~k is along the z− axis ~k = (0, 0, k) and
~k · ~r = kr cos θ. The left hand side depends only on cos θ and is independent of φ.
Hence only m = 0 terms will be present in the right hand side giving Cℓm = 0 if
m 6= 0. Also noting that
r
2ℓ + 1
Yℓ0 (θ, φ) = Pℓ (cos θ), (7.9)
4π
the Eq.(7.8) takes the form
∞
X
exp(ikr cos θ) = Aℓ jℓ (kr)Pℓ (cos θ). (7.10)
ℓ=0

We determine the coefficients Aℓ by using orthogonality property of the Legendre

polynomials. Z π
2
Pℓ (cos θ)Pn (cos θ) sin θdθ = δℓn . (7.11)
0 2ℓ + 1
Therefore, we multiply Eq.(7.10) by Pn (cos θ) sin θ and integrate over θ from 0 to π
giving
Z π X∞ Z π
exp(ikr cos θ)Pn (cos θ) sin θdθ = Aℓ jℓ (kr) Pℓ (cos θ)Pn (cos θ) sin θdθ
0 ℓ=0 0

X∞
2
= Aℓ jℓ (kr) δℓm , (7.12)
ℓ=0
2ℓ + 1

88

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 or Z 
π
2 
exp(ikr cos θ)Pn (cos θ) sin θdθ = An jn (kr). (7.13)
0 2m + 1
Changing the integration variable from θ to t = cos θ in the left hand side we get
Z 1
2Am jm (kr)
exp(ikrt)Pm (t)dt = . (7.14)
−1 (2m + 1)
At this stage we should compute the integral on the left hand side and compare
with the right hand side and get the value of An . Instead of trying to compute the
integral exactly, it is suffifient to obtain its asymptotic expression for large r and
compare the answer with the large r behaviour of the right hand side, utilising the
known large r behaviour of the spherical Bessel function
1
jn (kr) −→ sin(kr − nπ/2). (7.15)
kr
The large r asymptotic expansion of the the integral in the left hand side can be
found by integrating by parts as follows.
Z 1
exp(ikrt)Pn (t)dt
−1
1 Z 1
1 ikrt  1
= e Pn (t) − eikrt Pn′ (t)dt (7.16)
ikr −1 ikr −1
1
1 1

= eikr − (−1)n e−ikr − e ikrt ′
P n (t)  + O(1/(ikr)3 ) (7.17)
ikr (ikr)2 −1

Thus as r → ∞ we get
Z 1
1

exp(ikrt)Pn (t)dt −→ eikr − (−1)n e−ikr + O(1/r 2) (7.18)
−1 ikr
einπ/2 ikr−inπ/2

= e − e−ikr+inπ/2 (7.19)
ikr
2einπ/2
= sin(kr − nπ/2). (7.20)
kr
Using Eq.(7.15) and Eq.(7.20) in Eq.(7.14) we get

An = (2n + 1)einπ/2 = (2n + 1)in (7.21)

Thus the expansion of the plane waves, Eq.(7.10) takes the form
∞
X
ikz ikr cos θ
e =e = (2ℓ + 1)iℓ jℓ (kr)Pℓ (cos θ) (7.22)
ℓ=0

A formula similar to Eq.(7.22) can be written down when the plane wave propagates
in a direction other than the z− axis. Let α, β be the polar angles of the direction
of propagation n̂ so that

n̂ = (sin α cos β, sin α sin β, cos β) (7.23)

89

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 If the angle between ~k and ~r be Φ, the relation Eq.(7.22) assumes the form
∞
X
~
eik·~r = eikr cos Φ = (2ℓ + 1)iℓ jℓ (kr)Pℓ (cos Φ) (7.24)
ℓ=0

Using the well known addition theorem for spherical harmonics

ℓ
X
2ℓ + 1 ∗
Pℓ (cos Φ) = Yℓm (α, β)Yℓm(θ, φ) (7.25)
4π
m=−ℓ

in Eq.(7.24) we get
∞ X
X ℓ
i~k·~
r
e = 4π (2ℓ + 1)iℓ jℓ (kr)Yℓm
∗
(α, β)Yℓm(θ, φ) (7.26)
ℓ=0 m=−ℓ

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-19-01-

L1;

§3 Radial Wave Fnction

QM-
Lecs-
19002
⊲
For computing the cross section, we need to solve the energy eignevalue problem
Hψ = Eψ (7.27)
for continuous energies relevent for scattering situations, there are an infinite number
of solutions we need to pick up a solution satisfying boundary condition
ikr
e
ψ(~r) −→ exp(i~k · ~z) + f (θ, φ) (7.28)
r
where f (θ, φ) is the scattering amplitude and |f (θ, φ)|2 gives the cross section. For
a spherically symmetric potential, the Schrodinger equation can be solved by sepa-
ration of variables in polar coordinates giving the energy eigenfunctions in the form
uE (r, θ, φ) = REℓ (r)Yℓm (θ, φ) (7.29)
and the most general solution for a given energy is a superpostion
∞ X
X ℓ
ψ(~r) = Cℓm REℓ (r)Yℓm(θ, φ). (7.30)
ℓ=0 m=−ℓ

For a sperically symmetric potential, the operators L ~ 2 and Lz commute with H and
are constants of motion. Thus if incident particle has a definite value of L~ 2 and Lz ,
it contunues to have the same values at all times and the problem of finding the
scattering amplitude is solved by finding the amplitude for different values of L ~2
and Lz separately. More over, it is sufficient to find the solutions for Lz = 0. This
is because if we select the z − axis parallel to the incident momentum ~ki , the initial
value of Lz will be equal to zero and will remain zero at all times. Recall that

~ˆ z = −i~ ∂
L
∂φ

90

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 and eigenvcetors with zero eignevalue for Lz must be independent of φ. Hence is
sufficient to restrict the sum over m in Eq.(7.30) to m = 0 terms only and the sum
then reduces to the following form
∞
X
ψ(~r) = Cℓ REℓ (r)Pℓ (cos θ). (7.31)
ℓ=0

It is therefore sufficient to solve the radial equation

   
~2 1 d 2 dRℓ (r) ℓ(ℓ + 1)~2
− r + E − V (r) − Rℓ (r) = 0 (7.32)
2µ r 2 dr dr 2µ
and the knowledge of the asymptotic behaviour of the solution for the radial wave
function Rℓ (r) gives the cross section. The method of partial waves thus consists of
the following steps.
1 Solve the radial equation for Rℓ (r).

2 Find the asymptotic behaviour of the radial wave function Rℓ (r).

3 Relate the large r behaviour of the radial wave function to the scattering
amplitude and hence to the cross section.
QM-
Lecs-
19003
⊲ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-Lecs-
19002;

We assume that the potential has a finite range and that the potential V (r) → 0,
as r → ∞, faster than r12 , i.e., r 2 V (r) → 0 as r → ∞. Thus for large r, V (r) can
be neglected as compared to ℓ(ℓ + 1)~2 /2 µr 2 term in the radial equation
   
~2 1 d 2 dRℓ (r) ℓ(ℓ + 1)~2
− r + E − V (r) − Rℓ (r) = 0 (7.33)
2µ r 2 dr dr 2µr 2
which for large r assumes the form
   
~2 1 d 2 dRℓ (r) ℓ(ℓ + 1)~2
r + E− Rℓ (r) ≈ 0 (7.34)
2µ r 2 dr dr 2µr 2
or    
1 d 2 dRℓ (r) 2 ℓ(ℓ + 1)
r + k − Rℓ (r) ≈ 0 (7.35)
r 2 dr dr r2
The most general solution of this equation is a linear combination of the spherical
Bessel functions jℓ (kr) and nℓ (kr)
r→∞
Rℓ (r) ≈ Aℓ jℓ (kr) + Bℓ nℓ (kr) (7.36)
Note even though the radial equation looks like a free radial particle equation, the
combination Eq.(7.36) is approximate solution for large r only, where as for the free
particle solution is jℓ (kr) for all r, a term nℓ (kr) is absent in the free particle solution.
Since we interested only in the large r behaviour of the radial wave function, we use
the following asymptotic expansions for spherical the Bessel functions:
1
jℓ (kr) ≈sin(kr − ℓπ/2) (7.37)
kr
1
nℓ (kr) ≈ − cos(kr − ℓπ/2) (7.38)
kr
91

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Using Eq.(7.37) and Eq.(7.38) we get the following results for the asymptotic form
of the radial wave functions.

Free Particle Solution

Rℓ (r) = Cℓ jℓ (kr) (7.39)

Cℓ
≈ sin(kr − ℓπ/2) (7.40)
kr

Scattering Solution for V (r) 6= 0

Rℓ (r) = Aℓ jℓ (kr) + Bℓ nℓ (kr) (7.41)

1
≈ (Aℓ sin(kr − ℓπ/2) − Bℓ cos(kr − ℓπ/2) (7.42)
kr
A′ℓ
= sin(kr − ℓπ/2 + δℓ ), (7.43)
kr
where we have defined
q
A′ℓ = A2ℓ + Bℓ2 , tan δℓ = −Bℓ /Aℓ . (7.44)

Comparing the scattering solution for the potential Eq.(7.44) with the free
particle solution Eq.(7.40), we find that the Eq.(7.44) is shifted by a phase
δℓ given by Eq.(7.44). This quantity δℓ is called the phase shift for the ℓth
partial wave. The phase shift is a function of energy and of course the angular
momentum ℓ and carries all the information about the scattering for angular
momentum ℓ.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-Lecs-

19003;

§4 Relating Cross Section to Phase Shifts

92

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM-
Lecs- ⊲ In order to derive an expression for the scattering amplitude in terms of the
19005
phase shifts, we substitute the partial wave expansion for the plane waves
∞
X
ikz ikr cos θ
e =e = (2ℓ + 1)iℓ jℓ (kr)Pℓ (cos θ), (7.45)
ℓ=0

and that of the wave function corresponding to the scattering solution

∞
X
ψ(~r) = Cℓ Rℓ (r)Pℓ (cos θ) (7.46)
ℓ=0

with Rℓ (r) having an asymptotic expansion of the form

A′ℓ
Rℓ (r) ≈ sin(kr − ℓπ/2 + δℓ ) (7.47)
kr
in the boundary condition to be required of the wave function for large r

eikr
ψ(~r) −→ exp(ikz) + f (θ, φ) (7.48)
r
and collect the coefficients of exp(±ikr) in the the sides of Eq.(7.48). This gives us
the following expression for the left and the right hand sides of Eq.(7.48).

Left hand side of Eq.(7.48) The wave function is a superposition of the radial
solution for different partial waves and we have
∞
X
ψ(~r) = Cℓ Rℓ (r)Pℓ (cos θ) (7.49)
ℓ=0
X∞
1
≈ A′ℓ Pℓ (cos θ) sin(kr − ℓπ/2 + δ)ℓ ) (7.50)
ℓ=0
kr
X∞
1  i(kr−ℓπ/2+δℓ
= A′ℓ Pℓ (cos θ) e − e−i(kr−ℓπ/2+δℓ (7.51)
ℓ=0
2ikr
∞
1 X ′ 
= Aℓ Pℓ (cos θ)e−iℓπ/2 ei(kr+δℓ − (−1)ℓ e−ikr−δℓ (7.52)
2ikr ℓ=0
eikr  X ′ 
∞
−iℓπ/2 iδℓ
= A Pℓ (cos θ)e e (7.53)
2ikr ℓ=0 ℓ
e−ikr  X ′ 
∞
− Aℓ Pℓ (cos θ)(−1)(ℓ+1) e−iδℓ (7.54)
2ikr
ℓ=0

Right hand side of Eq.(7.48) We make use of the plane wave expansion in the
right hand side to get

93

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 ∞
X eikr
(2ℓ + 1)iℓ jℓ (kr)Pℓ (cos θ) + f (θ) (7.55)
ℓ=0
kr
∞
1 X eikr
= (2ℓ + 1)iℓ sin(kr − ℓπ/2)Pℓ (cos θ) + f (θ) (7.56)
kr ℓ=0 kr
∞
1 X ℓ
 i(kr−ℓπ/2) −i(kr−ℓπ/2)
eikr
= (2ℓ + 1)i e −e Pℓ (cos θ) + f (θ) (7.57)
2ikr kr
ℓ=0

eikr  
X∞
= 2ikf (θ) + (2ℓ + 1)iℓ Pℓ (cos θ)e−iℓπ/2 (7.58)
2ikr ℓ=0
−ikr  X
∞ 
e
+ (2ℓ + 1)iℓ Pℓ (cos θ)(−1)eiℓπ/2 (7.59)
2ikr ℓ=0

Since the exponentials eikr and e−ikr are linearly independent functions, their coef-
ficients in Eq.(7.48) must be equal giving
∞
X ∞
X
ℓ −iℓπ/2
2ikf (θ) + (2ℓ + 1)i Pℓ (cos θ)e = A′ℓ Pℓ (cos θ)e−iℓπ/2 eiδℓ (7.60)
ℓ=0 ℓ=0

and
∞
X ∞
X
ℓ iℓπ/2
(2ℓ + 1)i Pℓ (cos θ)(−1)e = A′ℓ Pℓ (cos θ)(−1)(ℓ+1) e−iδℓ (7.61)
ℓ=0 ℓ=0

Because the Legendre polynomials Pℓ (cos θ) are linearly independent their coeffi-
cients in the two sides of Eq.(7.61) must be equal. This gives
A′ℓ e−iℓπ/2 (−1)ℓ+1 e−iδℓ = (2ℓ + 1)iℓ eiℓπ/2 (7.62)
Hence we get the result
A′ℓ = (2ℓ + 1)iℓ δℓ (7.63)
ℓ iℓπ/2
Substituting Eq.(7.63) in Eq.(7.60) and noting i = e , we get
∞
X ∞
X
2ikf (θ) = − (2ℓ + 1)Pℓ (cos θ) + (2ℓ + 1)e2iδℓ Pℓ (cos θ) (7.64)
ℓ=0 ℓ=0
∞
X

= (2ℓ + 1)Pℓ (cos θ) e2iδℓ − 1 (7.65)
ℓ=0

The differential cross section is given by

dσ
= |f (θ)|2 (7.66)
dΩ
and the total cross section is
Z
dσ
σtotal = dΩ (7.67)
dΩ
Z π
= |f (θ)|2 2π sin θ dθ (7.68)
0

94

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The θ integration can be completed using orethogonality of the Legendre polynomi-
als and one gets
 X ∞
4π
σtotal = (2ℓ + 1) sin2 δℓ (7.69)
k2 ℓ=0

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-Lecs-

19005;

QM-
§5 Phase Shifts for Square Well
19-
03- ⊲
L1 As an example of calculation of phase shifts, we will compute the phase shifts for a
square well potential. (
−V0 if r < R0
V (r) = (7.70)
0 if r > R0
The radial equation for a spherically symmetric potential V (r) is
   
~2 1 d 2 dRℓ (r) ℓ(ℓ + 1)~2
− r + E − V (r) − Rℓ (r) = 0 (7.71)
2µ r 2 dr dr 2µr 2

The potential for r < R0 is −V0 and the most general solution of the radial equation
is
(I)
Rℓ (r) = αjℓ (qr) + βnℓ (qr) r < R0 (7.72)
where q 2 = 2µ(E + V0 )/~2 and for r > R0 the potential is zero and the most general
solution is
(II)
Rℓ (r) = α′ jℓ (kr) + β ′ nℓ (kr), r > R0 , (7.73)
where k 2 = 2µE/~2. Next we must impose regularity conditions on the solutions.
Since nℓ (ρ) blows up as ρ → 0, we demand that β in Eq.(7.72) must be zero.
Next the radial solution and its first derivative must be continuous at all points, in
particular at r = R0 . These two conditions give the restrictions

R(I) (r)|r=R0 = R(II) (r)|r=R0 (7.74)

 
dR(I) (r)  dR(II) (r) 
= (7.75)
dr r=R0 dr 
r=R0

Taking into account of the form of the solution we get

αjℓ (qR0 ) = α′ jℓ (kR0 ) + β ′ nℓ (kR0 ) (7.76)

αqjℓ′ (qR0 ) = α′ kjℓ′ (kR0 ) + β ′ kn′ℓ (kR0 ) (7.77)

Dividing Eq.(7.77) by Eq.(7.76) we get

qjℓ′ (qR0 ) α′ kjℓ′ (kR0 ) + β ′ kn′ℓ (kR0 )

= (7.78)
jℓ (qR0 ) α′ jℓ (kR0 ) + β ′ nℓ (kR0 )
o
qjℓ′ (qR0 ) kj ′ (kR0 ) + k(β ′/α′ )n′ℓ (kR0 )
= ℓ (7.79)
jℓ (qR0 ) jℓ (kR0 ) + (β ′/α′ )nℓ (kR0 )

95

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Noting that the phase shift is given by tan δℓ = −(β ′ /α′ ), we solve for (β ′ /α′ ) to get
 ′ 
k jℓ (kR0 ) jℓ (qR0 ) kℓ (kR0 )
q nℓ (kR0 jℓ′ (qR0 )
− nℓ (kR0 )
tan δℓ =  n ′ (kR )
 (7.80)
1 − kq jjℓ′ (qR
(qR0 ) ℓ
0 ) nℓ (kR0 )
0

For ℓ = 0, using the expressions for the spherical Bessel functions in terms of sine
and cosine functions, for s− wave one can easily obtain
k
tan (δ0 (k) + kR0 ) = tan(qR0 ) (7.81)
q

Threshold behaviour of the phase shifts We can now derive the behaviour of
the phase shifts, for a square well potential, for small k. When kR0 ≪ 1 we have
1
jℓ (kR0 ) ∼ (kR0 )ℓ nℓ ∼ (7.82)
(kR0 )lℓ+1
1
jℓ′ (kR0 ) ∼ (kR0 )ℓ−1 n′ℓ ∼ (7.83)
(kR0 )lℓ+2

This gives
δℓ (k) ≃ (kR0 )2ℓ+1 (7.84)
Therefore, as momentum k goes to zero, the ℓth partial wave phase shift goes to
zero as (2ℓ + 1) power of k. This result is generally true for finite range potentials.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-19-03-
L1;

96

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 8

Time Dependent Perturbation

Theory

QM-
§1 Introduction
24-
01- ⊲
L1 Not all problems can be solved exactly. Therefore, we need approximation methods.
We have so far discussed approximation methods for energy eigen values and eigen
functions. These methods give approximate solution of the eigenvalue problem

Hψn = En ψn . (8.1)

The next class of problems, we are interested, require us to solve the time dependent
Schrödinger equation
∂ψ
i~ = Ĥψ. (8.2)
∂t
The exact solution of Eq.(8.2) can be written as
 −iĤ(t − t ) 
0
ψ = exp ψ(x, t0 ) (8.3)
~

This is valid when Ĥ is independent of time. Eq.(8.3) is equivalent to

X  −iE (t − t ) 
n 0
ψ(x, t) = Cn exp ψn (x), (8.4)
n
~

where ψn are eigenfunctions of the full Hamiltonian and Cn are the expansion coef-
ficients which are computed from the knowledge of the wave function at an initial
time t = t0 . Thus
X
ψ(x, t0 ) = Cn ψn (x) (8.5)
Zn
Cn = ψn∗ (x)ψ(x, t0 )dx. (8.6)

This method, outlined above in Eq.(8.3)-Eq.(8.6) can be used only if Ĥ is inde-

pendent of time and the solution of time independent Eq.(8.1) can be found. We

97

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 need a different approach if the Hamiltonian Ĥ depends on time, or if Ĥ is time
independent but the energy eigenvalues and eigenfunctions cannot be found exactly.
There are three important methods to solve time dependent Schrodinger equation
approximately.

• In the [time dependent perturbation theory] we split the total Hamiltonain as

H = H0 + H ′ (8.7)

where H0 , called unperturbed Hamiltonian, is independent of time and is such

that the time dependent Schrodinger equation for H0
∂ψ
i~ = H0 ψ (8.8)
∂t
can be solved exactly. The perturbing Hamiltonian H ′ may or may not depend
on time. Both cases fall under time dependent perturbation theory when we
need to solve the time dependent Schrodinger equation.

• If the time dependence of the Hamiltonian is small, it varies slowly with time,
the approximation scheme is known as adiabatic approximation. It is assumed
that the instantaneous eigenvalues and eigenvectors of the Hamiltonian are
known at ecah instant of time.

• In sudden approximation it is assumed that the time variation of the Hamil-

tonian is very fast and the Hamiltonian changes with time over only a very
short interval of time. For example, consider a particle in a box of size L. If
the walls of the box move and the size of the box is suddenly doubled, we may
apply sudden approximation.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-24-01-

L1;

QM-
§2 Time Dependent Perturbation Theory
24-
02- ⊲
L1 In the perturbation theory, one assumes that the total Hamiltonian can be spilt as

H = H0 + H ′ (8.9)

where H0 is independent of time and is such that its eigenvalues and eigen functions,
i.e. the solutions of the equation

H0 un = En un (8.10)

are known exactly. H ′ is a perturbation and its matrix elements are assumed to be
small compared to the matrix elements of H0 . H ′ may or may not, depend on time,
but H0 must be independent of time. We shall write

H = H0 + λH ′ (8.11)

for intermediate steps and set λ = 1 in the end.

98

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Note that the states represented by un are stationary states only when H ′ = 0.
If the full Hamiltonian H is different from H0 , (H ′ 6= 0), the states represented by
un will not be stationary states. Thus if, at some initial time t0 , the system is in
the state represented by one of the eigenfunctions ui, it will not remain in the same
state afterwards. We would like to compute the probability of system being found
in a state of uf at a later time.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-24-02-
L1;

QM-
§3 Transition Amplitude
24-
03- ⊲
L1 We shall at first derive an exact equation for amplitude for transition from an initial
state ui to a final state uf .
We shall start with the expression
H = H0 + λH ′ (8.12)
where λ to be set equal to 1 in final answer. The Schrodinger equation is
∂ψ
= Hψ i~ (8.13)
∂t
We assume that at time t = t0 the wave function of the system is given to be φ0 (~r)
and we want to compute (approx) wave function at time t.
It is assumed that the eigenfunction and corresponding eigenvalues of H0 are
known exactly and will be denoted as un and En :
H0 u0 = En un . (8.14)
To solve Eq.(8.13) we expand ψ(x, t) as a sum in terms of the stationary states of
H0 X En (t−t0 )
ψ(x, t) = Cn (t)e−i ~ un (x), (8.15)
where Cn , in general, depends on time, because the total Hamiltonian H is different
from H0 . At time t = t0 the wave function is given to be φ0 (~r). Therefore, setting
t = t0 , in Eq.(8.15) we get
X
φ0 (~r) = Cn (t0 )un (~r). (8.16)
n

Therefore assuming un (~r) to be normalized, Cn (t0 ) are computed using

Cn (t0 ) = (un , φ0 ). (8.17)
Substituting Eq.(8.15) in Eq.(8.13) we shall derive an equation for Cn (t) which will
solved by perturbation theory. Therefore Eq.(8.15) used in Eq.(8.13) gives
X dCn (t) −iEn (t−t0 ) X “
−iEn (t−t0 )
”
i~ e ~ un (x) + Cn (t)En e ~
un (x) (8.18)
n
dt n
X
= (H0 + H ′ ) Cn (t)e−iEn (t−t0 )/~un (x) (8.19)
n
X X
−iEn (t−t0 )/~
= Cn (t)En e un (x) + Cn (t)e−iEn (t−t0 )/~H ′ un (x)(8.20)
n n

99

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Therefore,
X  dC (t)  X
n
i~ e−iEn (t−t0 )/~un (x) = Cn (t)e−iEn (t−t0 )/~H ′ un (x) (8.21)
n
dt n

Take the scalar product with um (x) and use the orthogonality property of u’s to get,
 dC  X
m
i~ e−iEm (t−t0 )/~ = e−iEn (t−t0 )/~ (um , H ′ un ) Cn (t) (8.22)
dt n

or
dCm X i(Em −En )(t−t0 )/~
i~ = e (um , H ′un )Cn (t). (8.23)
dt n
(Em −En )
Define ωmn = ~
and write Eq.(8.23) in the form
 dC  X
m
i~ = eiωmn (t−t0 ) (um , H ′un )Cn (8.24)
dt n

This is an exact equation. The value of Cn (t) at t = t0 is given by Eq.(8.17).

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-24-03-
L1;

QM-
§4 Perturbation Theory
24-
03- ⊲
L2 We recall some of the equations from the previous section.

H = H0 + H ′ (8.25)
H0 un = En un (8.26)


ψ(x, t) = φ0 (x) (8.27)
t=t0
X
ψ(x, t) = Cn (t) exp(−iEn (t − t0 )/~)un (x) (8.28)
n

where the coefficients Cn are functions of time and satisfy an exact equation given
by
d X
i~ Cn (t) = exp(iωmn (t − t0 ))hm|H ′ |niCn (t) (8.29)
dt n

and the initial condition Eq.(8.27) on ψ(x, t) implies

Cn (0) = (un , φ0 ). (8.30)

It should be noted that the time dependence of the coefficients Cn is controlled

by the interaction term H ′ only. In absence any perturbation, i.e. , H ′ = 0 the
coefficients Cn become constants independent of time. This is because Cn are the
expansion coefficients in the expansion of the wave function in terms of the solu-
tions exp(−iEn (t − t0 )/~)un (x) of the time dependent equation for the unperturbed
Hamiltonian H0 . To obtain the perturbative solution of Eq.(8.29) we replace H ′

100

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 with λH ′ and expand Cn in powers of λ. These steps are similar to the steps leading
to the Born approximation.
(0) (1)
Cm (t) = Cm (t) + λCm (t) + λ2 Cm
(2)
(t) + · · · (8.31)
Substituting (8.31) in Eq.(8.29) and comparing the coefficients of different powers
of λ we successively get
(0)
dCm
i~ = 0 (8.32)
dt
(1) X
dCm
i~ = exp(iωmn (t − t0 ))hm|H ′ |niCn(0) (t) (8.33)
dt n
(2) X
dCm
i~ = exp(iωmn (t − t0 ))hm|H ′ |niCn(1) (t) (8.34)
dt n

In many problems one is interested in computing the transition probability from an

initial state ui to a final state uf under the action of the perturbation term H ′. For
such a problem we have
φ0 (~r) = ui (~r), (8.35)
Cm (0) = (um , φi) = (um , ui) = δim , (8.36)
(
1 m=i
Cm (0) = (8.37)
0 m 6= i.
These equations can be solved successively to obtain the coefficients Cm to a desired
order of perturbation theory. The first order result is easily found to be
Z
(1) 1 t
Cf (t) = hf |H ′|ii exp(iωf i (t′ − t0 ))dt′ (8.38)
i~ t0
The above result is the basic result from which can be applied to actual problem of
interest. Further progress can be made if time variation of H ′ is known. A simple and
commonly encountered situation is the case of periodic perturbation with a single
frequency. Assuming that H ′ varies periodically with time with a single frequency
ω we write
H ′ = F e−iωt + F † eiωt (8.39)
where F is an operator which does not depend on t explicitly. Let us substitute
Eq.(8.39) in Eq.(8.38) and integrate to get
 i(ωf i −ω)t   −i(ωf i +ω)t 
(1) e −1 † e −1
Cf (t) = hf |F |ii + hf |F |ii (8.40)
−~(ωf i − ω) ~(ωf i + ω)
The above results have been derived assuming that the final state corresponds to
a discrete energy level. If the final state f corresponds to a state in continuum, the
quantity of interest is the transition rate, or the transition probability per unit time.
Further discussion and a derivation of the result for the transition probability per
unit time, the Fermi Golden rule, will be presented in § 5. Note that when ωf i = ±ω ←
(1)
the coefficient Cf diverges, signalling breakdown of the perturbation theory due a
resonance. This case will be dicsussed in detail in § 6 ←
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-24-03-
L2;

101

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM-
§5 Fermi Golden Rule
24-
03- ⊲
L3 In this section we assume that the perturbation is either independent of time, or
varies periodically with a single frequency and that the energy of the final states lies
in continuum. In this case the quantity of interest is the transition probability per
unit time and we will derive the Fermi Golden rule for this transition probability
per unit time.
We shall start from Eq.(8.40) with ω = 0 and a similar treatment can be for
the case ω 6= 0.
When the perturbation term is independent of time, the probability amplitude,
upto first order,(setting t0 = 0) is given by
 
(1) ′ exp(iωf i t) − 1
Cf (t) = hf |H |ii (8.41)
−~ωf i

and hence one has

(1) 4 sin2 (ωf i t/2)
|Cf (t)|2 = |hf |H ′|ii|2 . (8.42)
~2 ωf2i
(1) (1)
We plot |Cf (t)|2 in the figure below. Note that |Cf (t)|2 is large for ωf i ≈ 0,i.e.
for Ei ≈ Ef . Only a small range of energy ∆E values

∆E ≈ 2π(~/t) (8.43)

have an appreciable transition probability. As t → ∞, ∆E → 0 and one recovers

conservation of energy. The Eq.(8.43) suggests that if a measurement is made after
time ∆t, the accuracy in E will be of the order of ∆E ≈ h/∆t which a form of
statement of time energy uncertainty relation.

Fig. 8.1.
QM-24-F02;Golden-Rule

Note that the area under the peak increases as t. Thus if we compute the transition
probability at time t, given by
Z E+∆E
(1)
|Cf (t)|2 |dE, (8.44)
E−∆E

to a set of states in the energy range E and E ± ∆E, the answer will be proportional
to time. In the case of transitions to a state in continuum, the quantity of interest
is the rate of transitions to a group of final states having the energy in the range

102

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 E ± ∆E, and hence one needs to compute the transition probability per unit time.
So we compute
d (1) 2 2  
′ 2 sin ωf i t
|C (t)| = |hf |H |ii| (8.45)
dt f ~ ωf i
and for large t this expression tend to
2π 2π
2
|hf |H ′|ii|2δ(ωf i ) = |hf |H ′|ii|2 δ(Ef − Ei ) (8.46)
~ ~
where use has been made of the standard results
sin kx
lim = πδ(x) (8.47)
x→∞ x
1
δ(ax) = δ(x) (8.48)
|a|
for the Dirac δ function. Hence the required transition probability per unit time to
the group of final states, obtained by differentiating Eq.(8.44) w.r.t. t and denoted
by wi→f , is given by
2π X
wi→f = |hf |H ′|ii|2 δ(Ef − Ei ) (8.49)
~
final states
The presence of Dirac delta function implies energy conservation and allows us to
do the energy integral. Introducing the density of final states of energy ρ(E) by
means of X
δ(Ef − E)(·) = ρ(E) (8.50)
final states
gives the Golden rule
2π
wi→f = |hf |H ′|ii|2 ρ(E) (8.51)
~
where we have set Ef = Ei = E. This result, derived by Dirac, was named Golden
Rule by Fermi.
When the perturbation varies harmonically with time, we must analyse
Eq.(8.40) and the result is

2π
wi→f = |hf |H ′|ii|2 ρ(Ef ) (8.52)
~

The analysis proceeds by keeping only one of the two terms in Eq.(8.40) and
showing that the other term is not important. The final energy Ef can have only
one of the two values Ei + ~ω or Ei − ~ω only.

Density of states
If we write the sum over final states as an integral over energy and a sum over
remaining variables, such as angles,
X X′ Z
(.) = dEf (.). (8.53)
final states final

103

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 P′
Here denotes the sum over other variables. Thus, we would get
X′ Z X′
ρ(E) = dEf δ(Ef − E) = 1 (8.54)
final final

which states that the density of states equals the number of states with energy in
the range E, E + dE is ρ(E)dE.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-24-03-
L3;

QM- Application to scattering

24-
03- ⊲
L6 The Fermi Golden rule can be applied to scattering and gives the first Born ap-
proximation result for the scattering amplitude. Let the incident beam be described
by plane waves of momentum ~~ki. We wish compute the differential cross section
σ(θ, φ).
We therefore seek the rate of transtion into solid angle dΩ in the direction of ~kf .
The Fermi Golden rule gives the transition probabilty per unit time w to be
2π
wi→f = ρ(kf )|hk~f |H ′|~ki i|2 , (8.55)
~

where the potential energy is taken to be the perturbation Hamiltonian H ′ = V (~(r)).

We now need to compute the density of states ρ(~k) for the final states. For the initial
and final states, we will work with the plane wave solutions with periodic boundary
conditions given by
1 1
ui(~r) = exp(i~ki · ~r), uf (~r) = exp(i~kf · ~r). (8.56)
L3/2 L3/2

The allowed values of ~k in a box are kx = 2πnx /L, ky = 2πny /L, kz = 2πnz /L etc.
where nx , ny , nz are positive integers. There will be (L/2π)3 dkx dky dkz states for the
propagation vector in the range ~k and ~k + d~k. The range dk is related to the range
dE of energy given by
~2 k 2 dE ~2 k
E= ⇒ = . (8.57)
2µ dk µ
and the number of states is
L3
ρ(E)dE = dkx dky dkz (8.58)
(2π)3

where ρ(E) is the density of states. For differential cross section we need to compute
the transtion probability to states with final propagation vector in a small range of
angles θ, φ. The number of states with the direction ~k in the solid angle dΩ =
sin θ dθ dφ and magnitude in a small range dk is given by k 2 dΩ. Using Eq.(8.57), we
get

L3 2 L3 µ k
ρ(E)dE = k dk dΩ ⇒ ρ(k) = dΩ. (8.59)
(2π)3 8π 3 ~2

104

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Using the density of states (8.58), wave functions (8.56), the Golden rule gives the
transition probability w
2π
w = ρ(E)|hk~f |H ′|~ki i|2 (8.60)
~
µL3 k ~ ′ ~ 2
= |hkf |H |ki i| dΩ. (8.61)
(4π 2 ~3 )
Let us now recall the definition of differential cross section. Let a scattering
experiment be performed with a total of N particles. The number of particles
scattered into solid angle dΩ per unit time is proportional to the solid angle and
to the incident flux. The constant of proportionality is just the differential cross
section. The number of particles scattered into the solid angle is just the transition
probability w times N. The incident flux is N times the probability current for the
initial state and equals Nµ~k . . Therefore, using (8.61 for w, and (8.56) for the wave
functions

N × w = σ(θ, φ) × dΩ × N × Flux (8.62)

µkL3 ~k
N× 2 3
dΩ = |hk~f |H ′|~ki i|2 dΩ = σ(θ, φ) × dΩ × (8.63)
(4π ~ ) µ
2 3
µl
∴ σ(θ, φ) = |hk~f |H ′ |~kii|2 . (8.64)
4π 2 ~4
Hence the differential cross section is given by the Born approximation result
 µ 2 Z 2

σ(θ, φ) =  ei~q·~r V (~r) d3r  , (8.65)
2π~2  

and ~q = (~ki − ~kf ) is the momentum transfer.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-24-03-
L6;

QM-
§6 Probability for Resonance Transitions
24-
03- ⊲
L4 The case of periodic perturbation with a single frequency is an important one for
many physical situations including the interaction of radiation with matter. Assum-
ing that H ′ varies periodically with time with a single frequency ω we write

H ′ = F eiωt + F ∗ e−iωt (8.66)

where F is an operator which does not depend on t explicitly. Let us substitute

Eq.(8.66) in Eq.(8.38) and integrate to get
   −i(ωf i +ω)t 
(1) ei(ωf i −ω)t − 1 † e −1
Cf (t) = hf |F |ii + hf |F |ii . (8.67)
−~(ωf i − ω) −~(ωf i + ω)
In this section we discuss the case of resonance transition from an initial discrete
level i to a final discrete level f when the applied perturbation varies harmonically in
time. Here the term level refers to an energy level of H0 . The first order perturbation

105

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 result for a transitions between two discrete levels is given by Eq..(8.67) . When
the frequency ~ω is close to one of the two differences Ei − Ef , or Ef − Ei , the above
result blows up and the perturbation theory breaks down. In this case we must
get back to the exact equations and analyze them again making a different kind of
approximation. We will do so and solve the resulting approximate equations exactly.
We start with Eq..(8.24) after substituting

H ′ = F eiωt + F1 e−iωt (8.68)

we get
dCm (t) X

i~ = exp i(ωmn + ω)t hm|F |niCn (t)
dt n
X

+ exp i(ωmn − ω)t hm|F † |niCn (t). (8.69)
n

In the perturbation approximation after integration, the large coefficients came from
those terms which were multiplied with an exponential with a small argument. For a
given ω when there are two energy levels i and f such that |Ef −E −i| matches with
~ω, we need to retain all the terms involving the two coefficients Ci and Cf in the
summation in the right hand side of Eq..(8.69) . Thus the resulting approximate
equations to be solved assume the form
dCf (t)

i~ = exp i(ωf i + ω)t hf |F |iiCi(t)
dt

+ exp i(ωf i − ω)t hf |F †|iiCi (t). (8.70)
and
dCi (t) X

i~ = exp i(ωif + ω)t hi|F |f iCf (t)
dt n


+ exp i(ωif − ω)t hi|F † |f iCf (t). (8.71)
. In these equations we retain only those exponentials which have small arguments.
Taking ~ω ≈ (Ef − Ei ), using the notation ν ≡ ωf i − ω, and therefore writing
ωif + ω = −ν, we get
dCf (t)
i~ = hf |F † |iieiνt Ci (t) (8.72)
dt
dCi(t)
i~ = hi|F |f ie−iνt Cf (t) (8.73)
dt
Next we solve these equations exactly with the initial conditions Ci (0) = 1, Cf (0) =
0. The probability of transition from the initial level Ei to the final level Ef at time
t is then given by
2|hf |F |ii|2
Pi→f (t) = | {1 − cos Ωt} , (8.74)
~2 Ω2
where
~2 ν 2 + 4|hf |F |ii|2
Ω2 = . (8.75)
~2
106

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 It is to noted that the transition probability is periodic in time with the period 2π/Ω
and it varies from 0 to a maximum value
2|hf |F |ii|2
(8.76)
~2 ν 2 + 4|hf |F |ii|2
Ef −Ei −~ω
For the exact resonance ν = ~
= 0 and we get the transition probability to
be
1

Pi→f (t) = 1 − cos 2|hf |F |ii|t/~ , (8.77)
2
and the system makes periodic transitions between the levels i and f with the period
π~/|hf |F |ii|.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-24-03-
L4;

QM-
Details of Solution
24-
03- ⊲
L5
For the resonance transitions the equations satisfied by the coefficients Ci and Cf ,
Eq..(8.72) and Eq..(8.73) , are

dCf (t)
i~ = hf |F † |iieiνt Ci (t) (8.78)
dt
dCi(t)
i~ = hi|F |f ie−iνt Cf (t) (8.79)
dt
In this section we solve these equations exactly and obtain expressions for Ci (t) and
Cf (t). To solve we define
bf = Cf exp(−iǫt) (8.80)
so that
d d  iǫt 
Cf (t) = bf e (8.81)
dt dt
d 
= bf + iǫbf eiǫt (8.82)
dt
Eliminating Cf Eq..(8.78) and Eq..(8.79) , using Eq..(8.80) and Eq..(8.81) , we
get
1
Ċi = hi|F |f ibf (8.83)
i~
and
1
ḃf + iǫbf = hf |F †|iiCi
i~
1
= hf |F |ii∗Ci (8.84)
i~
Eliminating Ci from Eq..(8.82) and Eq..(8.84) we get

1
b̈f + iǫḃf = hf |F †|iiĊi
i~
|Ff i |2
= − 2 bf (8.85)
~
107

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Therefore, we have
|Fif |2
b̈f + iǫḃf + ḃf = 0 (8.86)
~2
This is a linear differential equation with constant coefficient and can be solved
exactly. The solutions of Eq..(8.86) have the form

bf (t) = exp(iαt) (8.87)

where α satisfies the equation

|Fif |2
α2 + ǫα − (8.88)
~2
The two roots of this equation are α± where
ǫ
α± = − ± ∆ (8.89)
2
where ∆ is given by
ǫ2 |Fif |2
∆2 = + (8.90)
4 ~2
Substituting back in Eq..(8.87) the general solution for bf becomes

bf (t) = A exp(iα+ t) + B exp(iα− t) (8.91)

and hence h i
Cf (t) = A exp(iα+ t) + B exp(iα− t) exp(iǫt) (8.92)

We then get, from Eq..(8.79) ,

!
i~ 
Ci (t) = iAα+ exp(iα+ t) + iα− B exp(iα− t)
Fif∗

+iǫA exp(iα+ t) + iǫB exp(iα− t) (8.93)

At time t = 0, the initial conditions are Ci (0) = 1 and Cf (0) = 0 giving

iAα+ + iBα− + iǫ(A + B) = Fif∗ /~ (8.94)

A+B = 0 (8.95)

Using Eq..(8.93) and Eq..(8.94) we get

Fif∗
A(ǫ/2 + ∆) + B(ǫ/2 − ∆) = − (8.96)
~
or
Fif
2A∆ = − (8.97)
~
Fif
A = − (8.98)
2∆~
Fif∗
B = (8.99)
2~∆
108

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Rearranging Eq..(8.93) and using B = −A we get
!
i~  
Ci (t) = iA(ǫ + α+ ) exp(iα + t) + iB(ǫ + α− ) exp(iα− t) (8.100)
Fif∗
!
i~  
= ∗
(iA) (ǫ + α+ ) exp(iα+ t) − (ǫ + α− ) exp(iα− t) (8.101)
Fif

Substituting for α± from Eq..(8.89) we get

! 
i~ −iFif∗
Ci (t) = × exp(−iǫt/2)
Fif∗ 2∆~
h i
× (ǫ/2 + ∆)eiα+ t − (ǫ/2 + ∆)eiα− t (8.102)
1 h i
= exp(−iǫt/2) 2∆ cos(∆t) + iǫ sin(∆t) (8.103)
2∆
and Ci (t) is given by
 ǫ 
−i ǫt
Ci(t) = e 2 cos ∆t + i sin ∆t (8.104)
2∆
Also Eq..(8.92) with B = −A gives
 
Cf (t) = A exp(iǫt) exp(iα+ t) − exp(iα− t) (8.105)
= 2iA exp(iǫt/2) sin ∆t (8.106)
 iF ∗ 
if
= − sin ∆t (8.107)
∆~
Hence the probability o finding the system in the state f at time t is
|Fif |2
|Cf (t)|2 = sin2 ∆t (8.108)
∆2 ~2
|Fif |2  2 |F |2 1/2
2 ǫ if
= 2 2
sin + t (8.109)
∆~ 4 ~2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-24-03-

L5;

§7 Examples and Applications

Several important applications will now be taken up briefly where the results of time
dependent perturbation are used.

QM-
I-Transition to a discrete level
24-
02- ⊲
L2
We assume that under the action of perturbation the transition takes place to an-
other state of system which corresponds to a discrete level. In this case Eq.(8.38)
is the basic formula, and |Cf i (t)|2 gives the probability of a transition at time t from
the initial state i to the final state f . We list a few cases of interest.

109

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Interaction switched on adiabatically
We assume that the interaction is switched on at some time and later switched
off adiabatically. As an example consider a charged harmonic oscillator moving
in a potential 12 mω 2 x2 .
P2 mω 2 2
H0 = + x (8.110)
2m 2
Suppose an electric field varying with with time is switched on. This would
introduce an extra term in Hamiltonian given by
H′ = −qE(t)x (8.111)
corresponding to the situation when the electric field is independent of x but
may change with time. For example, one may have
−t
E = E0 (1 − e t0 ) t>0 (8.112)
with the field being switched on at t = 0 and increasing to E0 . The energy
levels (n + 21 )~ω of the harmonic oscillator are no longer stationarty states
under the action of this electric field. If the system is in the state Ei at time
t = 0, it has a non zero probability of being in some other state Ef at a later
time. The method of time dependent perturbation allows us to compute this
probability when the field is weak.
Harmonic Perturbation
Our next case of interest is when the perturbation varies with time with a single
frequency. In this case the result Eq.(8.25) provides the perturbation theory
answer for the transition amplitude. This case has application to interaction of
electromagnetic radiation with matter. The single frequency case corresponds
to a monochromatic radiation.
Resonance Case
An important application of the time dependent perturbation theory is when
the perturbation Hamiltonian H′ varies with time as
H ′ = 2H0′ sin(ωt) (8.113)
E −E
when the frequency ω matches with a difference f ~ i for some final level the
probability of transition to Ef becomes very large and is small for all other
energy levels. In this case the basic perturbation theory result, Eq.(8.38) ,
breaksdown. In this case the problem is simplified by working in two level ap-
proximation i.e. by neglecting effect of all other energy levels and the inifinite
set of coupled equations reduce to two coupled linear equations which can be
solved exactly.
H ′ (t) has a continuous range of frequencies
In this class of examples the perturbation is a superpostion of a continuous
range of frequencies. For a pair of specified initial and final states, several
frequencies close to the resonance freqeuncy, (Ef − Ei )/~, will contribute ap-
preciably and all such contributions must be added up. In this case the result
is very similar to that given under Fermi Golden rule. An example of this case
is interaction of atoms with white light.

110

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-24-02-
L2;

QM-
II-Transitions to a set of final states in continuum
24-
02- ⊲
L3
The cases of interest are further grouped according to the time dependence of the
perturbation Hamiltonian H ′ (t)
Interaction H ′ (t) switched on adiabatically
In the previous case we have considered an example in which a charged har-
monic oscillator is subjected to perturbation 21 mω 2 x2
H ′ = V0 exp(−t/t0 ) (8.114)
Suppose we place a hydrogen atom between the plates of charged capacitor
which is subsequently discharged through a resistance. We may then again ask
a question: what is the probability that if at time t = 0 H-atom is in an initial
state n, after some time t, it will be found in some final, excited, state m. The
question is similar to the above example of charged harmonic oscillator and
falls under the class of problems mentioned in I.
However, the H- atom offers one more possibility. Under the action of external
perturbation, the atom may get ionized and the final energy then does not
correspond to a discrete energy level. The transition in this case takes place
to a level in the continuum.
Perturbation H ′ (t) is independent of time
This is, for example, the case for scattering; we are interested in knowing the
probability per unit time that a particle gets scattered into solid and dΩ. The
differential scattering cross section, σ(θ, φ) is related to such a probability.
prob per unit time of particle getting scattered in solid angle d Ω
σ(θ, φ) =
prob per unit time of a particle in incident beam crossing a unit area
(8.115)
In case of scattering
P2
H = H0 + V (r) H0 =
(8.116)
2m
and V (r) is to be treated as perturbation (H′ ≡ V (r)). Here the initial state
corresponds to plane waves with momentum p~i and final states corresponds to
momentum ~pf . The initial and final states are eigenstates of the unperturbed
Hamiltonian H0 .
In the presence of the potential, the momentum is not conserved and hence
the possibility of momentum changing to some final value p~f after some time.
Note that here (H′ = V (r)) is not dependent on time, but we are still using
time dependent method because the question concerns, time evolution of eigen
states of H0 under the action of full Hamiltonian H.
H ′ (t) varies harmonically with time
An example of interest here is ionization of an atom in presence of an external
periodic electric field which varies with a single frequency.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-24-02-
L3;

111

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 QM- III-Periodic perturbation– resonant transition
24-
02- ⊲
L4 An important application of the time dependent perturbation theory is when the
perturbation Hamiltonian H′ varies with time as

H ′ = 2H0′ sin(ωt) (8.117)

E −E
when the frequency ω matches with a difference f ~ i for some final level the prob-
ability of transition to Ef becomes very large and is small for all other energy levels.
In this case problem can be simplified by working in two level approximation i.e. by
neglecting effect of all other energy levels.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-24-02-
L4;

112

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 §8 Tutorial
[1] An arbitrary quantum mechanical system is initially in the state |0 >. At time
t = 0 a perturbation of the form H ′ = H0 exp(−t/T ) is swithced on. Show
that at large times the probability of the system being in the state |1 > is
given by

|h0|H0|1i|2
(~/T )2 + (∆E)2
where ∆E is the energy difference betwen the states |0i and |1i.

[2] A particle of charge e is confined to a cubical box of side 2b. An electric field
~ given below is applied to the system.
E

~ = (0 t<0
E
E~0 exp(−αt) t > 0)

where α > 0, The vector E0 is perpendicular to one of the surfaces of the box.
To the lowest order in E0 calculate the probability that the charged particle,
in the ground state at time t = 0, is excited to the first state at time t = ∞.

[3] A particle of charge q moving in one dimension is initially bound to a delta

function potential at the origin. From time t = 0 to t = τ it is exposed to a
constant electric field E in the x direction.

(a) Assume that the continuous energy wave functions may be approximated
by the free particle wave functions, find the density of states as ρ(E) as
function of energy E
(b) Assuming that the electric field may be treated as a perturbation find the
probability that the particle will be found in a continuous energy state
with energy in the range E and E + dE if at time t = 0 it was known to
be in the bound state.Sss

You may assume that the normalized bound state wave function for the delta
function potential V (x) = −γδ(x) is given by,
r
mγ
u(x) = 2
exp(− mγ
~2
|x|)
~

113

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Chapter 9

Semiclassical Theory of Radiation

QM-
§1 Electromagnetic Waves in Vacuum
25-
01- ⊲
L1 We shall first discuss the interaction of a charged particle with electromagentic waves
in the classical Maxwell’s theory. Classically the electromagnetic fields are described
by the four Maxwell’s equations which written in cgs units take the form

~ = ρ
∇·E (9.1)
ǫ0
~
~ = − dB
∇×E (9.2)
dt
~
∇·B = 0 (9.3)
~
~ = µ0 J~ + µ0 ǫ0 dE
∇×B (9.4)
dt

Here ρ is the charged density and J~ is the current density which satisfy the continuity
equation
dρ
+ ∇ · J~ = 0 (9.5)
dt

§ 1.1 Free E.M. waves

The Maxwell’s equations admit freely propagating wave solutions even in absence
of external charges and currents. Setting ρ = 0, J~ = 0 in Eq..(9.1) -(9.4), we get

~ = 0
∇·E (9.6)
~
dB
~ = −
∇×E (9.7)
dt
~ = 0
∇·B (9.8)
~
~ = µ0 ǫ0 dE
∇×B (9.9)
dt

114

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 and the wave equations for the electric and magnetic fields follow from these equa-
tions. The equations are

1 dE~
~ =
∇2 E (9.10)
c2 dt
~
~ = 1 dB
∇2 B (9.11)
c2 dt

§ 1.2 Scalar and Vector Potential

~ such that the magnetic
The Eq..(9.3) suggests that there exists a vector potential A
field can be written as
B~ =∇×A ~ (9.12)
Eq..(9.2) and 9.12 then imply that

 ~
∇× E~ + dA = 0 (9.13)
dt
This shows that there exists a scalar function φ such that
~
~ + dA = −∇φ
E (9.14)
dt
Thus the electric field can be written as
~
~ = −∇φ − dA
E (9.15)
dt

§ 1.3 Gauge Transformation and Gauge Invariance

It is easy to see that the potentials φ and A ~ are not uniquely determined by the
~ ~ ~ and φ′ , A
fields E and B. In fact two sets of potentials φ, A ~ ′ related by

~
dΛ
φ′ = φ − (9.16)
dt
~ ′ ~
A = A + ∇Λ (9.17)

give the same answers for the electric and the magnetic fields where Λ is an arbitrary
function of space time. Since all observable quantities are functions of the electric
and magnetic fields, they are also the same for the two sets of potentials. Thus for
example the Lorentz force

F~ = q E~ + ~v × B
~ (9.18)
experienced by a charged particle in electric and magnetic field does not depend
on the choice of the function Λ in Eq..(9.16) and (9.17). The transformation
~ φ) → (A
(A, ~ ′ , φ′), given by (9.16) is called a gauge transformation and we say
that the observable quantities are invariant under gauge transformation.

115

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 § 1.4 Gauge Condition
For a given set of electric and magnetic fields one has infinitely many possible answers
for the potentails, corresponding to the different choices of the gauge parameter
Λ. The choice of potentials can be restricted by requiring them to satisfy further
condition(s), called gauge condition. A change in gauge condition amounts to
making a gauge transformation and is of no physical consequence.
When describing free e.m. waves, ρ = 0, ~j = 0 it turns out to be convenient to
choose the gauge condition

φ = 0, ~=0
∇·A (9.19)

For such a choice of gauge condition the two wave equations Eq..(9.10) -(9.11)
become equivalent to the wave equation

1 dA~
~=
∇2 A (9.20)
c2 dt

§ 1.5 Plane Wave Solutions

Taking Eq..(9.19) as the gauge condition, the plane wave solutions of the wave
equation (9.17) can be written as

A(~ ~ 0 (ω) cos(~k · ~r − ωt),

~ r, t) = A (9.21)

where ~k is the propagation vector, ω = ck is the frequency and the amplitude vector
~ 0 ( satisfies the transversality condition
A

~k · A
~ 0 = 0. (9.22)

The electric and the magnetic fields are then given by

~ = −ωA0~e sin(~k · ~r − ωt),
E ~ = −A0~k(×~e) sin(~k · ~r − ωt).
B (9.23)


The Poynting vector N~ = 1 E ~ ×B ~ relates the amplitude A0 (ω) to the intensity
µ0
per unit frequency. The intensity in the frequency range ω, ω + ∆ω is in fact given
by I(ω)∆ω, where
ǫ0 ω 2 c ~ 2
I(ω) = |A0 |. (9.24)
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : .................................. ◭ QM-25-01-
L1;

§2 Interaction of Charged Particle with Radia-

QM- tion
25-
01- ⊲
L2
The expression for force on a point charge q at a position ~r is


F~ = q E
~ + ~v × B~ (9.25)

116

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 The classical mechanics equations of motion can then be set up using Newton’s sec-
ond law of motion. These equations can be derived from the Lagrangian formalism
if we take the Lagrangian to be
1 ~ · ~r − qφ(~r)
L = m~r2 + q A (9.26)
2
The corresponding Hamiltonian is given by
1  2
~ + qφ(~r)
H= p~ − q A (9.27)
2m
Thus we can set up the Lagrangian or the Hamiltonian formalism as needed. It
must be noted that the electric and magnetic fields do not enter the Lagrangian
or the Hamiltonian, these contain only the potentials which are subject to gauge
transformations and one must discuss the gauge invariance of the formulation. In
fact it is easy to see that the equations of motion do not change.

§ 2.1 Schrödinger Equation

The Hamiltonian operator for a charged particle in external e.m. fields is obtained
from the classical Hamiltonian, Eq..(9.27) , by the replacement

p~ −→ b
p~ = −i~∇ (9.28)

Thus the Schrodinger equation takes the form

dψ b
i~ = Hψ (9.29)
dt
where the Hamiltonian operator is given by
 2
b= 1 b
H ~ + qφ(~r)
p~ − q A (9.30)
2m
Here again we must ask if our quantum mechanical description of the charged particle
is gauge invariant? Note that the potentials A, ~ φ, appearing in the Schrödinger
equation, are subject to gauge transformations and a gauge transformation should
not have any effect on the observable quantities. One can in fact prove that the
description is indeed gauge invariant.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-25-01-
L2;

QM-
§3 Semiclassical Treatment of Radiation
25-
02- ⊲
L1 We will describe the interaction of an electron in an atom with e.m. waves in
a semiclassical fashion. The electron will will be teated quantum mechanically,
whereas the e.m. waves will be treated classically. We shall ignore the effect of
motion of the electron on the e.m. waves. We will interested in finding out what
are allowed transitions of an electron that can take place in presence of e.m. waves
and we want to compute the corresponding transition probabilities.The effect of

117

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 other electrons will be represented by a potential V (~r) while the scalar and vector
~ r ) will describe the effect of the e.m. waves. Thus we shall take
potentials φ(~r), A(~
the Hamiltonian of the electron in an atom to be given by
1  2
~ + qφ(~r) + V (~r)
H= p~ − q A (9.31)
2m
We expand the first term and rewrite the Hamiltonian as
1  ~
2
H = p~ − q A + qφ(~r) + V (~r) (9.32)
2m
~p2 q ~

~ + q A
2
~ 2 + qφ(~r) + V (~r)
= − A · ~p + ~p · A (9.33)
2m 2m 2m
~p2 q ~ ~

q2 ~ 2
= − 2A · p~ + −i~∇A + A + qφ(~r) + V (~r) (9.34)
2m 2m 2m
where in the last step use has been made of the commutator
~ − A~
p~ · A ~ p = −i~∇ · A.
~ (9.35)

We shall work in the gauge

φ=0 ~ = 0.
∇·A (9.36)

In this gauge the Hamiltonian H takes the form

H = H0 + H ′ (9.37)
~p2
H0 = + V (~r) (9.38)
2m
q ~ q2 ~ 2
H′ = A · p~ + A (9.39)
m 2m
where H0 describe the electron in the atom, and H ′ gives the effect of e.m. waves
QM-
on the electron and will be treated as a perturbation.
25-
02- ⊲ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-25-02-
L1;
L2

Weak field approximation

We shall split the interaction Hamiltonian H ′ as H ′ = H1 + H2 with
q ~ q2 ~ 2
H1 = A · p~, H2 = A . (9.40)
m 2m
The transition amplitude in the first order perturbation is given by
Z
1 t
Cf (t) = hf |H ′|ii exp(iωf i t) dt. (9.41)
i~ 0
We will make weak field approximation and neglect H2 term, the matrix element
appearing in (9.40) can then be written as

hf |H1|ii = A0 hf | cos(i~k · ~r)ê · p~|ii. (9.42)

118

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 Dipole approximation
For atomic transitions, to the lowest order, it is a good approximation to replace
cos(i~k · ~r) by 1. Thus the matrix element becomes
hf |H1|ii = A0 hf |ê · p~|ii (9.43)
Next we note the commutator of ê · p~ is proportional to the commutator of ê · ~r with
~2
p
the Hamiltonian H0 = 2m + V (~r)
1
ê · ~rH0 − ê · ~rH0 = ê · p~ (9.44)
m
Taking matrix element between the two states |ii and |f i we get
1
hf |ê · ~rH0 |ii − hf |ê · ~rH0 |ii = i~ hf |ê · p~|f i (9.45)
m
Using the fact that |ii and |f i are eigenvectors of H0 with eigenvalues Ei and Ef
we get
1
i~ hf |ê · p~|f i = (Ei − Ef )hf |ê · ~r|ii (9.46)
m
Therefore,
m
hf |ê · ~p|f i = i ωf i hf |ê · ~r|ii (9.47)
~
The operator appearing in the matrix element is the dipole operator, hence this
approximation is called the dipole approximation.

Perturbation theory
The amplitude of transition from an initial state |ii to a final state |f i is given by
2 1 2 1
2 ′ 2 sin 2 (ωf i − ω)t ′ 2 sin 2
(ωf i + ω)t
|Cf i (t)| = 4|hf |H |ii| + 4|hf |H |ii| + Cross terms
(9.48)
~2 (ωf i − ω)2 ~ (ωf i + ω)2
2

where ωf i = (Ef − Ei )/~. The case Ef > Ei , with ωf i − ω ≈ 0 the first term is most
important and this corresponds to absorption. The second case when, Ef < Ei and
ωf i + ω ≈ 0, hold correspond to emission of radiation. Both emission and absorption
can take place only in presence of radiation. The emission process in presence of
external radiation is called induced emission.
In cases when the incident radiation is an incoherent superposition of radiation
of several frequencies, we can add probabilities for different frequencies, and the
probabilty of transition after time t will be given by
Z ∞
Pf i (t) = |Cf (t)|2 dω. (9.49)
0

We now repeat the same steps as in the derivation of Fermi Golden rule. To compute
the transition probability per unit time, we take the time derivative of (9.49), and
the limit t → ∞ gives Dirac δ function which can be used to do the ω integral to
give the transition rate
πq 2
wf →i = I(ω)|hf |ê · ~r|ii|2, (9.50)
3~2 cǫ0

119

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 where an averaging over different direction has been done.
Given a pair of states |mi and |ni, probabilities for the two transitions - induced
emission and induced absorption - given by one of the two terms in Eq..(9.41) -
are equal.
The semiclassical theory however does not explain why an atom in higher state
emits radiation and go to a state with lower energy even in absence of external
radiation- a process known as spontaneous emission. This can be computed only
when the radiation field is quantised. However, Einstein showed how an answer can
be obtained using Planck’s law of black body radiation.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-25-02-
L2;

QM-
§4 Einstein A and B Coefficients
25-
03- ⊲
L1 The semiclassical theory gives a reasonably good description of atomic transitions
induced by a light incident on an atom. It fails to give an explanation for ob-
served spontaneous emission of radiation when an atom is in an excited state. Ein-
stein showed how to compute the transition probability for induced emission using
Planck’s law of black body radiation.
We assume that atoms are placed inside a cavity. When the atoms are in an
equilibrium with radiation at temperature T , the number of photons of a particular
frequency emitted per sec must be the same as the number of photons absorbed
per sec by the atoms. Note that the probability of absorption per unit time as
well as the probability of induced emission per unit time is given by the same
expression, Eq.(9.50) , let’s call it A. Let N1 and N2 denote the number of atoms
in levels E1 and E2 with E2 > E1 . When in equilibrium at temperature T , the
number of atoms in the level with energy Ek , will be proportional to the Boltzmann
factor exp(−Ek /kB t). Let B denote the transition probability per unit time for the
induced emission to take place in the state E2 . We therefore have
The total number of transitions from E1 to E2 (absorption)
= N1 × Probability per unit time for induced absorption

The number of transitions from E2 to E1 ( emission)

= N2 ×( probability per unit time of induced emission
+ probability per unit time of spontaneous emission)
This gives
exp(−E1 /kB T ) × A = exp(−E2 /kB T )(A + B) (9.51)
substituting Eq.(9.50) for A and using Planck’s Law for the frequency distribution
I(ω),
~ω 3
I(ω) = 
, (9.52)
π 2 c2 exp ~ω/kB T − 1
where ω = (E2 − E1 )/hbar. Solving for B gives
q2ω3
B= |h~ri|2 (9.53)
3πǫ0 ~c3

120

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -

 which is the required transition probability per unit time for spontaneous emission.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Status : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ◭ QM-25-03-
L1;

121

DRAFT -- DRAFT -- DRAFT -- DRAFT -- DRAFT -