PII - Principles of Quantum Mechanics - Horgan (2014) 87pg

Part II Principles of Quantum Mechanics
Michaelmas 2014
Prof. R.R. Horgan
November 28, 2014
Contents
1 Introduction 1
2 Dirac Formalism 2
2.1 States and Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Observables and measurements . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Time evolution and the Schrödinger Equation . . . . . . . . . . . . . . 7
2.4 Bases and Representations . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Position and momentum basis – wavefunctions . . . . . . . . . . . . . . 11
2.6 Simultaneous Measurements and Complete Commuting Sets . . . . . . 15
3 The Harmonic Oscillator 17

3.1 Analysis using annihilation, creation and number operators . . . . . . . 17
3.2 Importance of the oscillator – applications in outline . . . . . . . . . . 20
4 Pictures and Operators 22

4.1 Unitary operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2 Schrödinger and Heisenberg Pictures . . . . . . . . . . . . . . . . . . . 23
4.3 Canonical Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5 Composite Systems and Identical Particles 27

5.1 Tensor products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2 Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3 Multiparticle states: bosons and fermions . . . . . . . . . . . . . . . . . 30
5.4 Spin and statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.5 Two-particle examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.6 Pauli Exclusion Principle and Atomic Structure . . . . . . . . . . . . . 34
6 Perturbation Theory 36
6.1 The non-degenerate case . . . . . . . . . . . . . . . . . . . . . . . . . . 36
i
CONTENTS ii
6.2 Degenerate case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7 Angular Momentum 43
7.1 Recap of orbital angular momentum . . . . . . . . . . . . . . . . . . . 43
7.2 General analysis of angular momentum eigenstates . . . . . . . . . . . 45
7.3 Matrix representations . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.4 Some physical aspects of angular momentum and spin . . . . . . . . . . 50
7.5 Addition of angular momentum . . . . . . . . . . . . . . . . . . . . . . 52
8 Transformations and Symmetries 57

8.1 Introduction and an Example . . . . . . . . . . . . . . . . . . . . . . . 57
8.2 Generators and conservation laws . . . . . . . . . . . . . . . . . . . . . 59
8.3 Angular Momentum and Rotations . . . . . . . . . . . . . . . . . . . . 61
8.4 Reflections and Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
9 Time-dependent Perturbation Theory 67

9.1 The interaction picture . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
9.2 Transition rates and Fermi’s Golden Rule . . . . . . . . . . . . . . . . . 69
9.3 Interaction of radiation/EM field with atoms . . . . . . . . . . . . . . 71
9.4 Emission and absorption of photons . . . . . . . . . . . . . . . . . . . . 75
10 Quantum Basics 77
10.1 Classical and Quantum Data . . . . . . . . . . . . . . . . . . . . . . . . 77
10.2 EPR experiment and Bell’s inequality . . . . . . . . . . . . . . . . . . . 79
10.3 Density operators and hidden sectors . . . . . . . . . . . . . . . . . . . 82
CONTENTS iii
BOOKS
‚ E. Merzbacher Quantum Mechanics, 3rd edition. Wiley 1998 (various prices)
‚ B.H. Bransden and C.J. Joachain Quantum Mechanics, 2nd edition. Pearson 2000
(£50-60 on Amazon)
‚ J. Binney and D. Skinner The Physics of Quantum Mechanics, 3rd edition. Cappella
Archive 2013 (£23.71)
‚ P.A.M. Dirac The Principles of Quantum Mechanics. Oxford University Press 1967
reprinted 1999 (£23.99).
‚ C.J. Isham Lectures on Quantum Theory: Mathematical and Structural Foundations.

Imperial College Press 1995 (around £18)
‚ S. Gasiorowicz Quantum Physics, 3rd edition. Wiley International 2003 £48.71
Mot of these books are expensive new and there are a lot of pedagogic books on
quantum mechanics, so it’s good to look at those in the library since many books are
good in one area but poor in another. Other books may be recommeneded through
the course.
1 INTRODUCTION 1
1 Introduction
‹ Recall features of elementary (IB) quantum mechanics:
‹ wave-particle duality. Waves behaving like particles – e.g., light quanta, pho-
tons and vice-versa; interference of electrons passing through crystal grating and
electron microscope. To make this more precise need:
‹ wavefunction ψpxq for particle. Probability density |ψpxq|2 ; probability is in-
trinsic to the theory.
‹ 0bservables become (hermitian) operators on wavefunctions. Lack of
commutation limits simultaneous measurement – leads to precise version of un-
certainty principle.
‹ Schrödinger’s equation specifies dynamics (evolution in time) and determines
energy levels.
This is enough to understand e.g., the hydrogen atom and transcends classical
physics.
‹ Aim of this course:
‹ reformulate QM in a more powerful, abstract, flexible and useful form: Dirac

formalism. This allows a simpler analysis of known problems such as the har-
monic oscillator and is also the clearest way to understand lots of more novel
properties, for example:
‹ the spin of particles;
‹ symmetries (e.g., translations and rotations) and conservation laws;
‹ identical particles;
‹ it provides the framework for quantizing other, more general, systems e.g., EM
field, and ultimately other forces leading to the ‘Standard Model’ of elementary
particles.
‹ Will not dwell on applications in any detail, but will keep track of what the mathe-
matical formalism is for.
‹ Assume IB QM and IA Dynamics but no electromagnetism beyond Coulomb’s law
and intuitive ideas about magnetism.
Plan:
1. Dirac formalism.
2. Harmonic oscillator.
3. Pictures of quantization.
4. Composite systems and identical particles.
5. Perturbation theory.
6. Angular momentum.
7. Transformations and symmetries.
8. Time-dependent perturbation theory.
9. Quantum basics.
2 DIRAC FORMALISM 2
2 Dirac Formalism
2.1 States and Operators
A quantum state is described at each instant by a state |ψy which belongs to a complex
vector space V . Then
|ψy, |φy P V ùñ α|ψy ` β|φy P V @ α, β P C. (2.1.1)
Physically this is the superposition principle leading to wave-like behaviour (interfer-

ence). However, these states are not wavefunctions but we will see that they carry all
the quantum information that describes the state of the system concerned in a very
general way.
There are also dual or conjugate states xφ| which belong to the dual space V : . By
definition, states and duals can be combined/paired to give a complex number:
moon , loo|ψy
looxφ| moon ÞÑ xφ|ψy
loomoon or formally V:ˆV Ñ C, (2.1.2)
‘bra’ ‘ket’ ‘bra(c)ket’
with
´ ¯
xφ| α1 |ψ1 y ` α2 |ψ2 y “ α1 xφ|ψ1 y ` α2 xφ|ψ2 y ,
´ ¯
β1 xφ1 | ` β2 xφ2 | |ψy “ β1 xφ1 |ψy ` β2 xφ2 |ψy , (2.1.3)
α, β P C. This is the definition of the dual as a vector space.

The space of states V and the dual V : come with an inner-product which can be
described as a one-to-one correspondence between states and duals:
V ÐÑ V :
with |ψy ÐÑ xψ| “ p |ψyq: (use same label for corresponding states)
and α|ψy ` β|φy ÐÑ α˚ xψ| ` β ˚ xφ| .
(2.1.4)
The inner product is
V ˆV Ñ C
|φy, |ψy ÞÑ xφ|ψy “ p|φyq: |ψy , (2.1.5)
and is assumed to obey
xφ|ψy “ xψ|φy˚ hermitian

k|ψyk2 “ xψ|ψy ě 0 (real from above) (2.1.6)
with k|ψyk2 “ 0 iffi |ψy “ 0.
This means that the inner product is positive semidefinite. Note that knowing xφ|ψy
for all xφ| determines |ψy uniquely and vice-versa.
The physical content of any state is unaltered by changing |ψy Ñ α|ψy pα ‰ 0q. We
shall usually normalize states by k|ψyk2 “ 1 but still have the freedom to change
|ψy Ñ eiθ |ψy. The absolute phase of a single state never has any physical significance,
2 DIRAC FORMALISM 3
but relative phases in combination such as α|φy ` β|ψy can be significant; for example,
for interference phenomena.
The space V is complete; we assume appropriate sequences or series converge. A
complete inner product space of this kind is a Hilbert space and this term is often
used for the space V in QM. V can be either finite or infinite dimensional and we shall
see examples of both.
An operator Q is a linear map on states, V Ñ V :
|ψy ÞÑ Q|ψy, (2.1.7)
and, by definition ´ ¯
Q α|φy ` β|ψy “ αQ|φy ` βQ|ψy . (2.1.8)
The same operator can be regarded as acting ‘to the left’ on dual states, V : Ñ V : :
xφ| ÞÑ xφ|Q , (2.1.9)

´ ¯ ´ ¯
xφ|Q |ψy “ xφ| Q|ψy @ |ψy
(2.1.10)
or simply “ xφ|Q|ψy .
For any Q the hermitian conjugate or adjoint is an operator Q: defined by

´ ¯:
:
xφ|Q “ Q|φy , (2.1.11)
or, equivalently,
´ ¯:
xφ|Q: |ψy “ Q|φy |ψy
“ xψ|Q|φy˚ @ |ψy, |φy . (2.1.12)
Simple consequences are

´ ¯:
αA ` βB “ α˚ A: ` β ˚ B : ,
` ¯:
AB “ B : A: , (2.1.13)
for any A, B.1

1
´ ¯:
xψ|pABq: |φy “ pABq|ψy |φy defn of pABq:
´ ¯:
“ A|ψ 1 y |φy |ψ 1 y ” B|ψy
“ xψ 1 |A: |φy defn of A:
´ ¯: ´ ¯
“ B|ψy A: |φy
“ xψ|B : A: |φy defn of B : .
(2.1.14)
True for all |ψy, |φy and result follows.

2 DIRAC FORMALISM 4
For any operator Q call |ψy p‰ 0q an eigenstate of Q with eigenvalue λ if
Q|ψy “ λ|ψy . (2.1.15)
Equivalently (taking : of this)
xψ|Q: “ λ˚ xψ| . (2.1.16)
For general Q, λ can be complex.

Of particular importance are commutators of operators
rA, Bs “ AB ´ BA “ ´ rB, As . (2.1.17)
We have the identities

rα1 A1 ` α2 A2 , Bs “ α1 rA1 , Bs ` α2 rA2 , Bs
linearity (2.1.18)
rA, β1 B1 ` β2 B2 s “ β1 rA, B1 s ` β2 rA, B2 s
rA, B1 B2 s “ rA, B1 sB2 ` B1 rA, B2 s

Leibnitz properties (2.1.19)
rA1 A2 , Bs “ rA1 , BsA2 ` A1 rA2 , Bs
rA, rB, Css ` rB, rC, Ass ` rC, rA, Bss “ 0 Jacobi identity (2.1.20)
2.2 Observables and measurements

An operator Q is hermitian or self-adjoint if
Q: “ Q . (2.2.1)
Such operators are called observables because they correspond to physical, measur-
able, quantities e.g., position, momentum, energy, angular momentum. Key results for
any hermitian Q:
(i) All eigenvalues are real.

(ii) Eigenstates with distinct eigenvalues are orthogonal.
(iii) The eigenstates form a basis for V (and their duals for V : ). So any state can be
expanded in terms of (written as a linear combination of) the eigenstates.
We prove (i) and (ii) and assume (iii).
(i)
Q|ψy “ λ|ψy
and xψ|Q: “ λ˚ xψ|
(2.2.2)
ñ xψ|Q “ λ˚ xψ| since Q is hermitian
ñ xψ|Q|ψy “ λxψ|ψy “ λ˚ xψ|ψy .
But k|ψyk2 “ xψ|ψy ‰ 0 p|ψy ‰ 0q and so deduce
λ “ λ˚ . (2.2.3)
2 DIRAC FORMALISM 5
(ii) Let |ny be eigenstates of Q with eigenvalues λ “ qn real, with n a discrete label
possibly of infinite range.
Q|ny “ qn |ny
and Q|my “ qm |my
(2.2.4)
or xm|Q “ qm xm|
ñ xm|Q|ny “ qn xm|ny “ qm xm|ny .
So qn ‰ qm ñ xm|ny “ 0.
Combining these three properties we have
‹ For any observable Q there is an orthonormal basis of eigenstates t|nyu for the space
of states V with
Q|ny “ qn |ny ,
xm|ny “ δmn . (2.2.5)
We speak of diagonalizing Q by choosing the basis to be the eigenstates |ny of Q.

This means that a general state |ψy can be expanded as
ÿ
|ψy “ αn |ny , (2.2.6)
n
where αn “ xn|ψy.
For the state to be properly normalized
k|ψyk2 “ xψ|ψy “ 1
´ÿ ¯´ ÿ ˘ ÿ
ðñ ˚
αm xm| αn |ny “ |αn |2 “ 1 . (2.2.7)
m n n
There might be several states with the same eigenvalue λ. Define the eigenspace for
a given eigenvalue by
Vλ “ t|ψy : Q|ψy “ λ|ψyu , (2.2.8)
which has the basis t|ny : qn “ λu.
The degeneracy of λ is the number of states in this basis, or dim Vλ . We say that λ
is non-degenerate if the degeneracy is 1.
r Note that passing from our three key results to the conclusion p‹q is achieved by
choosing an orthonormal basis for each Vλ :
(ii) ensures that these spaces are mutually orthogonal;
(iii) implies that the sum of all the eigenspaces is V , the entire space of states. u
Consider a measurement of Q when the system is in state |ψy immediately before.
Then
‚ The result is an eigenvalue, λ, say.

2 DIRAC FORMALISM 6
‚ This value is obtained with the probability

ÿ
ppλq “ |αn |2 . (2.2.9)
n:qn “λ
‚ Immediately after the measurement the state is

ÿ
|φy “ c αn |ny , (2.2.10)
n:qn “λ
where c is a normalization constant chosen so that xφ|φy “ 1.
So measurement projects the system into the eigenspace Vλ .

Example. Consider a system with three orthonormal states: |1y, |2y, |3y on which Q
has eigenvalues q1 “ q2 “ 0, q3 “ 1. Let the state of the system be
1
|ψy “ ? p2|1y ` |2y ` |3yq . (2.2.11)
6
Then
ˇ 1 ˇ2
$ ˇ ˇ
’
’
’ 1 ˇ? ˇ “ 1 |3y
’
’
’
ˇ 6ˇ 6
’
’
’
’
ˇ 2 ˇ2 ˇ 1 ˇ2
ˇ ˇ ˇ ˇ
Probability
&
0 ˇ? ˇ ` ˇ? ˇ “ 5 1
? p2|1y ` |2yq (2.2.12)
of measuring ’
’ ˇ 6ˇ ˇ 6ˇ 6 5
’
’
’
’
’
’
’
’
% Ò
final states
In this example we had degeneracy: two states with eigenvalue 0. However, often have
the case with λ non-degenerate with eigenstate |ny unique up to a phase. Then
‚ ppλq “ |αn |2 “ |xn|ψy|2 .

‚ αn “ xn|ψy is called the amplitude.
‚ The state after measurement is |ny.
In general, ÿ ÿ
ppλq “ |αn |2 “ 1 , (2.2.13)
λ n
as required for a probability distribution.

The expectation value (mean) of Q in state |ψy is
ÿ ÿ
xQyψ “ xψ|Q|ψy “ λppλq “ qn |αn |2 , (2.2.14)
λ n
and the uncertainty (spread or variance) is
p∆Qq2ψ “ xpQ ´ xQyψ q2 yψ “ xQ2 yψ ´ xQy2ψ . (2.2.15)

2 DIRAC FORMALISM 7
In thinking about many repeated measurements we must be careful to prepare the

system in the same state each time.
In the case that |ψy is an eigenstate of Q with eigenvalue λ, say, then
xQyψ “ λ , and p∆Qqψ “ 0 . (2.2.16)
r The process of measurement is still a source of some deep questions about the inter-
pretation of QM u.
Quantum mechanical behaviour arises from the fact that observables do not commute
in general. In any state |ψy
1
x∆Ayψ x∆Byψ ě |xrA, Bsyψ | , (2.2.17)
2
so rA, Bs ‰ 0 means we cannot expect to measure exact values for A and B simulta-
neously. This generalized Uncertainty Principle follows from
kpA ` iλBq|ψyk2 ě 0 @ real λ . (2.2.18)
The LHS is a quadratic in λ and the condition implies that the discriminant is ď 0;
the stated Uncertainty Principle then follows.
Paradigm example: position, x̂, and momentum, p̂, in one dimension obey
rx̂, p̂s “ i~
~
ñ ∆x ∆p ě . (2.2.19)
2
In D “ 3, x̂i and p̂i obey
rx̂i , p̂j s “ i~δij , (2.2.20)
and so the uncertainty principle applies to components of position and momentum
which are not orthogonal.
2.3 Time evolution and the Schrödinger Equation

So far our discussion of quantum states has been at a fixed time, even measurement is
assumed to be an instantaneous change of state. The evolution of states |ψptqy in time
is governed by the Schrödinger equation:
B
i~ |ψptqy “ H|ψptqy , (2.3.1)
Bt
where H “ H : is the Hamiltonian. Equivalently,
B
í~ xψptq| “ xψptq|H . (2.3.2)
Bt
Note that these equations imply
B´ ¯
xψptq|ψptqy “ 0 , (2.3.3)
Bt
2 DIRAC FORMALISM 8
so that the normalization of |ψptqy, and hence the probabilistic interpretation, is pre-
served in time.
H is an observable: the energy. Consider the eigenstates
H|ny “ En |ny . (2.3.4)
Then the states

eíEn t{~ |ny (2.3.5)
are stationary state solutions of the Schrödinger Equation.
The Schrödinger Equation is first-order in t and linear in |ψptqy and so for an initial
state at t “ 0, ÿ
|ψp0qy “ αn |ny , (2.3.6)
n
we have the unique solution

ÿ
|ψptqy “ αn eíEn t{~ |ny . (2.3.7)
n
Example. Consider system with two energy eigenstates |1y, |2y with energy eigenvalues
E1 , E2 , respectively. We are interested in measuring Q defined by
Q|1y “ |2y, Q|2y “ |1y ñ rQ, Hs ‰ 0 . (2.3.8)
The eigenstates of Q are easily found to be

1 ´ ¯
|˘y “ ? |1y ˘ |2y with eigenvalues q˘ “ ˘1 . (2.3.9)
2
Let the initial state, the state at t “ 0, be |ψp0qy “ |`y. Then have
1 ´ ¯
|ψptqy “ ? eíE1 t{~ |1y ` eíE2 t{~ |2y . (2.3.10)
2
The probability of measuring Q at time t and getting ˘1 is
ˇ ´
ˇ1 ¯´ ¯ˇˇ2
2 íE1 t{~ íE2 t{~
|x˘|ψptqy| “ ˇ x1| ˘ x2| e
ˇ |1y ` e |2y ˇˇ
2
ˇ ´
ˇ 1 íE t{~ ¯ˇˇ2
íE t{~
“ ˇˇ e 1 ˘ e 2 ˇ
2 ˇ
$ ´ ¯
2 pE1 É2 qt
& cos
’
’ 2~
“ ´ ¯ (2.3.11)
’
% sin 2 pE1 É2 qt
’
2~
Note that we are assuming no time-dependence in H. This would become a much

more complex situation.
2 DIRAC FORMALISM 9
2.4 Bases and Representations

Another use of a basis is that we can choose to reduce all states and operators to,
possibility infinite, column/row vectors and to matrices as follows
ÿ
|ψy “ αn |ny ô αn “ xn|ψy
n
ÿ
|φy “ βn |ny ô βn “ xn|φy . (2.4.1)
n
The inner product is then ÿ

xφ|ψy “ βn˚ αn . (2.4.2)
n
The operation of operator A can be written as

ÿ
A|ny “ |myAmn , (2.4.3)
m
where Amn “ xm|A|ny are the matrix elements of the complex matrix representing
the operator A in this basis. Note that the entries in this matrix depend on the basis; a
familiar result in linear algebra for any linear map. In contrast, the result of operating
with A on any state is independent of the basis. Check this result
ÿ
|φy “ A|ψy
loooooomoooooon ô β m “ Amn αn . (2.4.4)
n
basis-independent
loooooooooomoooooooooon
basis-dependent
Clearly, this representation is multiplication of a vector by a matrix: β “ Aα. Also,
have that the Hermitian conjugate has the familiar matrix form:
pA: qmn “ A˚nm . (2.4.5)
If B is another operator with matrix B then

ÿ
pABqmn “ Amp Bpn . (2.4.6)
p
I.e., as expected the usual rules of matrix multiplication apply.

This, by now familiar, way of representing linear maps on a vector space by the lin-
ear algebra of matrices gives the matrix representation of Quantum Mechanics. It is
most useful when the number of basis states is finite, but can also be useful for an 8
dimensional basis as well (does need care).
We are often interested in a function f pQq of an operator Q. How should this be
defined? The answer may be obvious if f is a polynomial or a series:
Q3
Q ` , eQ , sin pQq assuming convergence . (2.4.7)
3
But what about 1{Q or logpQq?
For Q an observable and |ny an eigenbasis for Q:
Q|ny “ qn |ny xn|my “ δmn , (2.4.8)

2 DIRAC FORMALISM 10
setting
f pQq|ny “ f pqn q|ny (2.4.9)
defines f pQq provided f pqn q is defined for all n; f pQq is defined on a basis and so is
defined on any state. This is certainly true if f is a polynomial or a power series that
converges for all qn . If qn ‰ 0 @ n then can define
1
Q´1 |ny “ |ny , (2.4.10)
qn
and likewise logpQq is defined if 0 ă qn ă 8.

A useful way to express that t|nyu is an orthonormal basis is the completeness rela-
tion or resolution of the identity:
ÿ
|nyxn| “ I , the identity operator. (2.4.11)
n
The notation is ´ ¯ ´ ¯
|nyxm| |ψy
loomoon loomoon “ loomoon loomoon .
|ny xm|ψy (2.4.12)
operator state state number
This is confirmed by applying each side to an arbitrary state
ÿ
|nyxn|ψy “ |ψy “ I|ψy . (2.4.13)
n
In the same way can resolve any operator in a similar fashion:

ÿ
Q “ qn |nyxn| ,
n
ÿ
f pQq “ f pqn q|nyxn| . (2.4.14)
n
In the case where the eigenvalues are degenerate then we can define a projection oper-
ator onto the subspace of eigenstates with eigenvalue λ by
ÿ
Pλ “ |nyxn| . (2.4.15)
n: qn “λ
The bases considered so far may be infinite but have been assumed discrete which
includes countably infinite bases. However, we can extend the index n to be continuous.
This requires some modifications in all relevant formulas:
$ ş
& |ψy “ ş dn αn |ny
ÿ ż
Ñ dn I “ ş dn |nyxn|
Q “ dn qpnq|nyxn|
n
%
xn|my “ δnm Ñ δpn ´ mq (2.4.16)
with |αn |2 “ |xn|ψy|2 . There is no longer a probability for discrete outcomes but a
probability density for the continuous range of n. We will see this below for position
and momentum operators.
2.5 Position and momentum basis – wavefunctions

Consider particle motion in one dimension. Position and momentum operators x̂, p̂
obey
rx̂, p̂s “ i~ . (2.5.1)
Let |xy be position eigenstates
x̂|xy “ x|xy , (2.5.2)
with continuous eigenvalue x and normalization
ż
1 1
xx|x y “ δpx ´ x q , dx |xyxx| “ I . (2.5.3)
In this basis, any state can be expanded as

ż
|ψy “ dx ψpxq|xy
with ψpxq “ xx|ψy a complex function . (2.5.4)
ψpxq is just the usual position wavefunction and the standard interpretation is the ob-
vious extension of the measurement postulates in section 2.2 to continuous eigenvalues:
|ψpxq|2 is the probability density for measuring position. (2.5.5)
The inner product in terms of wavefunctions becomes
ˆż ˙ ˆż ˙
˚ 1 1 1
xφ|ψy “ dx φpxq xx| dx ψpx q|x y
ż ż
“ dx dx1 φpxq˚ ψpx1 q lo
xx|x 1
omooyn
δpx ´ x1 q
ż
“ dxφpxq˚ ψpxq . (2.5.6)
So, in particular, ż
2
k|ψyk “ xψ|ψy “ dx |ψpxq|2 “ 1 (2.5.7)
for a normalized state.

Define similarly momentum eigenstates |py with
p̂|py “ p|py , (2.5.8)
and with ż
1 1
xp|p y “ δpp ´ p q , dp |pyxp| “ I . (2.5.9)
It is very important that the eigenstates of x̂ and p̂ can be chosen so that they are
related by
1
xx|py “ ? eipx{~ , (2.5.10)
2π~
1
ñ xp|xy “ ? eípx{~ . (2.5.11)
2π~
We justify this later after deducing some consequences. First find action of x̂ and p̂ in
terms of position wavefunctions:
x̂|ψy wavefunction : xx|x̂|ψy “ xxx|ψy “ xψpxq
p̂|ψy wavefunction : xx|p̂|ψy
ż
“ dp xx|p̂|pyxp|ψy [resolution of identity using p-states]
ż
“ dp pxx|pyxp|ψy
ż
B ´ ¯
“ dp ´ i~ xx|py xp|ψy
Bx
ż
B
“ í~ dp xx|pyxp|ψy
Bx
B B
“ í~ xx|ψy “ ´ i~ ψpxq . (2.5.12)
Bx Bx
So
B
xx|p̂|ψy “ ´ i~ xx|ψy, (2.5.13)
Bx
and have recover familiar results. However, also have new possibility. Can expand
states in momentum basis instead:
ż
|ψy “ dp ψ̃ppq|py ,
with ψ̃ppq “ xp|ψy , (2.5.14)

which is the momentum space wavefunction where |ψ̃ppq|2 is the probability density for
measurements of momentum. Then have
ż ż
2
k|ψyk “ dp xψ|pyxp|ψy “ dp |ψ̃ppq|2 “ 1 . (2.5.15)
As before, but with x Ø p:

,
p̂|ψy ÝÑ pψ̃ppq
momentum space
.
B (2.5.16)
x̂|ψy ÝÑ i~ ψ̃ppq - wavefunctions
Bp
The relationship between the wavefunctions follows from Eqs. (2.5.10,2.5.11):
ż
ψ̃ppq “ xp|ψy “ dx xp|xyxx|ψy
ż
1
“ ? dx eípx{~ ψpxq Fourier transform,
2π~ ż
1
and ψpxq “ ? dp eipx{~ ψ̃ppq inverse FT. (2.5.17)
2π~
Think of these as two different representations of states |ψy and the operators on
them:
ψpxq ψ̃ppq
x̂ ÝÑ x x̂ ÝÑ i~ Bp B
(2.5.18)
p̂ ÝÑ í~ Bx B p̂ ÝÑ p
N.B. „ ˆ ˙ ˆ ˙ 
B B
xx|rx̂, p̂s|ψy “ x í~ ` i~ x ψ “ i~ψ , (2.5.19)
Bx Bx
which verifies the commutation relation.
r The transforms between x and p space are familiar but here we are deriving all the
results, including the transform inversion theorem, on the assumption that t|xyu and
t|pyu are bases. u
The corresponding representations of the Hamiltonian
p̂2
Hpx̂, p̂q “ ` V px̂q (2.5.20)
2m
are
~2 B 2
on ψpxq : H ÝÑ ´ ` V pxq ,
2m Bx2
ˆ ˙
p2 B
on ψ̃ppq : H ÝÑ ` V i~ . (2.5.21)
2m Bp
It may be easy to interpret the potential term in momentum space. E.g., V pxq “
λxn ñ ˆ ˙
B Bn
V i~ “ λpi~qn n , (2.5.22)
Bp Bp
but more generally need to use first principles.
ż
xp|V px̂q|ψy “ dx xp|V px̂q|xyxx|ψy
ż ż
“ dx V pxqxp|xy dp1 xx|p1 yxp1 |ψy
ż ˆ ż ˙
1 1 ípp´p1 qx{~
“ dp dx V pxq e ψ̃pp1 q
2π~
ż
1
“ ? dp1 Ṽ pp ´ p1 qψ̃pp1 q . (2.5.23)
2π~
Thus H|ψy “ E|ψy becomes
~2 B 2 ψ
´ ` V pxq ψpxq “ E ψpxq in position space ,
2m Bx2
ż
p2 ? 1 1 1 1
2m ψ̃ppq ` 2π~ dp Ṽ pp ´ p q ψ̃pp q “ E ψ̃ppq in momentum space .
(2.5.24)
r Note that the convolution theorem derived here. u
Now return to the key condition in Eq. (2.5.10) and justify it:
1
xx|py “ ? eipx{~ . (2.5.25)
2π~
The point is that eigenstates are only ever unique up to a phase, even if normalized,
so we need to show there is a way to choose |xy and |py which makes this result true.
Doing this will involve an approach to translations to which we return later. Claim
that
|x0 ` ay “ eíap̂{~ |x0 y , (2.5.26)
which involves the translation operator
8 ˆ ˙n
ÿ 1 ía
U paq ” e íap̂{~
“ p̂n , (2.5.27)
n“0
n! ~
defines position eigenstates |xy @ x given one with x “ x0 . To check this first note
that
rx̂, p̂s “ i~ ñ rx̂, p̂n s “ i~n p̂n´1 . (2.5.28)
r Note that x̂ acts like “ i~ d{dp̂ ” inside a commutator. u Thus find
8 ˆ ˙n
ÿ 1 ía
rx̂, U paqs “ i~n p̂n´1 “ aU paq . (2.5.29)
n“1
n! ~
So
x̂ U paq|x0 y “ prx̂, U paqs ` U paqx̂q|x0 y

“ paU paq ` U paqx0 q|x0 y
“ px0 ` aqU paq|x0 y as required. (2.5.30)
Similarly,
|p0 ` by “ eibx̂{~ |p0 y , (2.5.31)
defines momentum eigenstates |py @ p given one with p “ p0 . Then
xx0 ` a|p0 ` by “ xx0 | eiap̂{~ |p0 ` by
“ eiapp0 `bq{~ xx0 |eibx̂{~ |p0 y
“ eipabàp0 `bx0 q{~ xx0 |p0 y . (2.5.32)

?
Choosing xx0 |p0 y “ 1{ 2π~ for reference values x0 “ p0 “ 0, then gives
1 1
xa|by “ ? eiab{~ ùñ xx|py “ ? eipx{~ , (2.5.33)
2π~ 2π~
by relabelling a, b as x, p, respectively. Actually, need justification for this last step.
‚ Since t|xyu is a basis we cannot have xx|p0 y “ 0 for every x, and then Eq. (2.5.26)
implies xx0 |p0 y ‰ 0, the required result, since
|x0 y “ |x ` px0 ´ xqy “ eípx0 ´xqp̂{~ |xy

ñ xx0 |p0 y “ eipx0 ´xqp0 {~ xx|p0 y ‰ 0 . (2.5.34)
‚ Now, the phase of xx0 |p0 y is a matter of convention but the modulus must be con-
sistent with
xp|p1 y “ δpp ´ p1 q , (2.5.35)
which is the desired normalization for the t|pyu basis. To check:
ż
1
xp|p y “ dx xp|xyxx|p1 y
ż
1 ipp1 ´pqx{~
“ dx e “ δpp ´ p1 q as required. (2.5.36)
2π~
‚ Similarly,
ż
1
xx|x y “ dp xx|pyxp|x1 y
ż
1 ippx´x1 q{~
“ dp e “ δpx ´ x1 q as expected. (2.5.37)
2π~
Note that the operator U paq implements translation by a on the position states.
2.6 Simultaneous Measurements and Complete Commuting

Sets
Return to idea of labelling basis states by eigenvalues of observables. If this cannot
uniquely be done with eigenvalues λ of some observable Q because some eigenvalues are
degenerate, then need at least one more observable Q1 to distinguish the corresponding
degenerate states.
Physically, we must be able to measure Q and Q1 simultaneously or
‚ first measure Q with result λ,
‚ then immediately measure Q1 with result λ1 .
‚ This second measurement must not change the value for Q and so still get result λ
if it is measured again immediately.
Mathematically, this requires a basis of joint eigenstates which simultaneously
diagonalize Q and Q1 :
Q|λ, λ1 y “ λ|λ, λ1 y
Q1 |λ, λ1 y “ λ1 |λ, λ1 y , (2.6.1)
which is only possible iffi
rQ, Q1 s “ 0 , (2.6.2)
as we now show.
‚ If there is a basis of joint eigenstates as above then
QQ1 |λ, λ1 y “ Q1 Q|λ, λ1 y “ λλ1 |λ, λ1 y , (2.6.3)
so QQ1 “ Q1 Q on these basis states and hence on all states since they can be expanded
on this basis.
‚ Conversely, if rQ, Q1 s “ 0 and |ψy belongs to the eigenspace Vλ of Q then

´ ¯ ´ ¯ ´ ¯
1 1 1
Q Q |ψy “ Q Q|ψy “ λ Q |ψy , (2.6.4)
and so Q1 |ψy also belongs to Vλ .
‚ The definition of a hermitian operator that
xφ|Q1 |ψy “ xψ|Q1 |φy˚ @ |ψy, |φy P V (2.6.5)
holds for the restriction |ψy, |φy P Vλ . Hence, D a basis for Vλ consisting of
eigenstates of Q1 . Call these |λ, λ1 y.
‚ Doing this for each Vλ gives a basis of such joint eigenstates for V .
r Note that if dim Vλ “ 1 (no degeneracy) then any state in Vλ is automatically an

eigenstate of Q1 since Q1 maps Vλ Ñ Vλ . u
Now can extend to any number of hermitian operators. Observables tQ1 , Q2 , Q3 . . . .u
are said to be a complete commuting set if any pair commute. Then there is a basis
of joint eigenstates with members
|λ1 , λ2 , λ3 , . . .y . (2.6.6)
An equivalent way to characterize a complete commuting set is as follows. If A is any

other observable with rA, Qi s “ 0 @ Qi then
A “ f pQ1 , Q2 , Q3 , . . .q (2.6.7)
for some function f . This means that
A|λ1 , λ2 , λ3 , . . .y “ f pλ1 , λ2 , λ3 , . . .q|λ1 , λ2 , λ3 , . . .y . (2.6.8)
An example is the generalization from one to three dimensions of the position and
momentum operators px̂, p̂q. These obey the commutation relations defined in terms
of their Cartesian component operators in usual notation
rx̂i , p̂j s “ i~δij . (2.6.9)
One complete commuting set is
x̂ “ px̂1 , x̂2 , x̂3 q (2.6.10)
with joint eigenstates:
x̂i |xy “ xi |xy

orthonormality: xx|x1 y “ δş p3q px ´ x1 q
basis: |ψy “ d3 x ψpxq|xy
position space wavefunction ψpxq “ xx|ψy .
An alternative complete commuting set is
p̂ “ pp̂1 , p̂2 , p̂3 q (2.6.11)

3 THE HARMONIC OSCILLATOR 17
with joint eigenstates

p̂i |py “ pi |py (2.6.12)
and momentum space wavefunction ψ̃ppq “ xp|ψy.
The relationship between these eigenstates is
1
xx|py “ 3{2
eip¨x{~ ñ
p2π~q
ż
1
ψ̃ppq “ 3{2
d3 x eíp¨x{~ ψpxq
p2π~q
ż
1
ψpxq “ 3{2
d3 p eip¨x{~ ψ̃ppq . (2.6.13)
p2π~q
There are other possibilities such as tx̂1 , x̂2 , p̂3 u leading to mixed position and momen-
tum space wavefunctions.
3 The Harmonic Oscillator

A one-dimensional harmonic oscillator of mass m and frequency ω is defined by the
Hamiltonian
1 2 1
H “ p̂ ` mω 2 x̂2 . (3.1)
2m 2
We will derive the energy levels and construct eigenstates using operator methods and
then also see how to find the wavefunctions.
3.1 Analysis using annihilation, creation and number opera-

tors
Define
´ mω ¯1{2 ˆ ip̂
˙
a “ x̂ ` ,
2~ mω
´ mω ¯1{2 ˆ ip̂
˙
:
a “ x̂ ´ . (3.1.1)
2~ mω
Note that these are dimensionless. Equivalently,

ˆ ˙1{2
~ ` ˘
x̂ “ a ` a: ,
2mω
ˆ ˙1{2
~mω ` ˘
p̂ “ i a: ´ a . (3.1.2)
2
It is easy to check that
rx̂, p̂s “ i~ ô ra, a: s “ 1 . (3.1.3)

Also have
mω 2 1 i
aa: “ x̂ ` p̂2 ´ px̂p̂ ´ p̂x̂q
2~ 2mω~ 2~
1 1
“ H ` ,
~ω 2
and similarly
1 1
a: a “ H ´ (opposite sign for commutator),
~ω 2
(3.1.4)
which confirms the commutation relations and also gives

ˆ ˙
1
H “ ~ω N ` , (3.1.5)
2
where N “ a: a is the number operator. N : “ N and finding its eigenstates and

eigenvalues is equivalent to doing this for H:
ˆ ˙
1
N |λy “ λ|λy ô H|λy “ E|λy with E “ ~ω λ ` . (3.1.6)
2
Let |λy be any such eigenstate normalized s.t. k|λyk2 “ 1. Then
λ “ xλ|N |λy “ xλ|a: a|λy “ ka|λyk2 ě 0 . (3.1.7)
Thus,
λ ě 0 all eigenvalues non-negative
(3.1.8)
“ 0 iffi a|λy “ 0 .
Next consider commutators
rN, a: s “ ra: a, a: s “ a: ra, a: s “ a: ,
(3.1.9)
rN, as “ ra: a, as “ ra: , asa “ á .
These relations imply that a: and a act on eigenstates by respectively raising and
lowering the eigenvalues by 1, provided the new states are non-zero and so actually are
eigenstates.
N pa: |λyq “ prN, a: s ` a: N q|λy

“ pa: ` a: λq|λy
“ pλ ` 1qpa: |λyq ,
N pa|λyq “ prN, as ` aN q|λy

“ pá ` aλq|λy
“ pλ ´ 1qpa|λyq . (3.1.10)
To find whether the new states are non-zero we compute their norms.
ka|λyk2 “ λ (already done above) (3.1.11)

which is only ever zero if λ “ 0.
ka: |λyk2 “ xλ|aa: |λy

“ xλ|a: a ` 1|λy
“ λ`1, (3.1.12)
which is never zero since λ ě 0. Because of these properties a: and a are called,
respectively, creation and annihilation operators.
Suppose there is an eigenstate |λy with λ not an integer. Then
a|λy, a2 |λy, . . . , am |λy, . . . are all non-zero.

eigenvalues: λ ´ 1, λ ´ 2, . . . , λ ´ m . . . .
(3.1.13)
But for m sufficiently large λ ´ m ă 0 which is a contradiction since all eigenvalues
are non-negative. By contrast if λ “ n “ 0, 1, 2, . . . then the contradiction is avoided
because
a|ny, a2 |ny, . . . , an |ny, are non-zero states
eigenvalues: n ´ 1, n ´ 2, . . . , 0,
(3.1.14)
m
but a |ny “ 0, m ą n and so the sequence terminates. Furthermore, have additional
eigenstates
a: |ny, pa: q2 |ny, . . . ,
(3.1.15)
eigenvalues: n ` 1, n ` 2, ... .
The eigenvalues of N are therefore precisely the non-negative integers and the oscil-
lator energy levels are
ˆ ˙
1
En “ ~ n ` n “ 0, 1, 2, 3, . . . . (3.1.16)
2
From calculations of norms above, we can choose normalized eigenstates |ny, xn|ny “ 1
which are then related by
? *
a: |ny “ ?n ` 1 |n ` 1y
ladder operators (3.1.17)
a|ny “ n |n ´ 1y
Starting from one state reach others by n

operating with a and a: . In particular,
starting from the ground state, |0y,
characterized by 3
2
a|0y “ 0 , (3.1.18)
a a
1
have normalized eigenstates
0
1
|ny “ ? pa: qn |0y . (3.1.19)
n! 0 state
In the absence of any internal structure can take tx̂u or tp̂u or tN u as a complete
commuting set. Then the energy levels are non-degenerate (eigenvalues of N label
them uniquely) and, in particular, |0y is completely specified by
a|0y “ 0 . (3.1.20)
If there is some internal structure then all states can carry an additional label i as-
sociated with some observable Q (or its eigenvalues) commuting with x̂, p̂, a, a: , N .
All energy levels have the same degeneracy with states |n; iy related by a, a: without
affecting i.
The analysis above is convenient for finding wavefunctions. In the position represen-
tation
|0y ù ψ0 pxq “ xx|0y
´ ¯1{2 ´ ¯ ´ ¯1{2 ´ ¯
a “ mω i p̂
x̂ ` mω ù mω ~ B
x ` mω
2~ 2~ Bx
´ ¯ (3.1.21)
a|0y “ 0 ù ~ B ψ pxq “ 0
x ` mω 0
Bx
2
´ mωx
ùñ ψ0 pxq “ N e 2~
¯1{4
´
with normalization factor N “ π~ mω .
Can also find wavefunctions for higher energy states by using Eq. (3.1.19). E.g.,
´ ¯1{2 ´ ¯
|1y “ a: |0y ù ψ1 pxq “ mω ~ B ψ pxq
x ´ mω 0
2~ Bx
(3.1.22)
´ ¯1{2
“ 2mω xψ0 pxq .
~
The correct normalization is guaranteed.
3.2 Importance of the oscillator – applications in outline

“Physics is that subset of human experience that can be reduced to coupled harmonic
oscillators.”
M. Peskin
‚ The oscillator is the simplest QM model beyond steps, wells etc. that can be solved
exactly; the hydrogen atom with a Coulomb potential is also special in this respect. It
is a very useful example to use as test case for new ideas, approaches and techniques.
‚ More importantly, many physical systems can be described, exactly or approximately,

in terms of oscillators.
‚ Consider a smooth potential V pxq with equilibrium point x0 (V 1 px0 q “ 0). For
displacements x from equilibrium
1
V px0 ` xq “ V px0 q ` V 2 px0 qx2 ` Opx3 q , (3.2.1)
2
and so if the displacements are not too large neglecting the Opx3 q contribution may
be a good approximation. Indeed, can include the effects of these anharmonic cor-
rections systematically using perturbation theory (see later). The point is that we
start with a soluble model. E.g., diatomic molecules where the quantization of vibra-
tional energies is important in understanding the internal energy and hence the heat
capacity of the gas – has macroscopic consequences. In other systems this approach
can breakdown, though.
‚ More complicated systems can be analyzed in terms of normal modes: each mode is
a coherent motion in which all degrees of freedom oscillate with common frequency ω.
This is common classically and can now quantize this motion. The general solution
for the classical oscillator is
xptq “ A eíωt ` A˚ eiωt A a complex constant. (3.2.2)
Normal modes for a system with variables xn , n “ 1 . . . N are of the form
xn ptq “ Aun eíωt ` A˚ u˚n eiωt , (3.2.3)
with un , n “ 1 . . . N a complex vector specific to each allowed normal frequency ω.

The general solution is a superposition of normal modes. To quantize treat normal
modes as independent oscillators. E.g.,
‚ Benzene ring with 6 CH units which oscillate around the “clock face” of the ring.
They are treated as if joined by identical springs. Actually, analyzed by discrete
group theory based on the symmetries of the ring.
‚ Crystal with # atoms N „ 1023 . The forces between the atoms are approximately
elastic and in 3D there are N “ 3N independent coordinates. Each of the 3N modes
is a collective motion of the atoms and if the approximation of elastic forces is good
then interaction between normal modes is small. If you excite just a single mode then
no other mode starts up – no energy transfer between modes; they are effectively
independent oscillators.
‚ Electromagnetic field – normal mode oscillations of electric and magnetic fields at

each point in space
Epx, tq “ Aupxq eíωt ` A˚ upxq˚ eiωt . (3.2.4)
In fact,
ω
upxq “ eik¨x with polarization K k and |k| “ . (3.2.5)
c
This gives a wave solution with behaviour e˘ipk¨x´ωtq . General solution is a linear
combination of normal modes for various ω, , k – exact for EM field.
4 PICTURES AND OPERATORS 22
‚ Quantization of normal modes as independent oscillators. For each normal mode

have apωq and apωq: (ω and other labels as necessary). Then
Ground state |0y with apωq|0y “ 0 @ modes .

"
state carrying energy but also mo-
a: pωq|0y
mentum ~k – like a particle.
* "
a destroy phonons in a crystal
particles (3.2.6)
a: create photons in EM field .
Phonon modes with long wavelength are sound waves (see AQM course in Lent).
Macroscopic consequences: heat capacity of crystals, blackbody radiation.
‚ In summary, the EM field, a relativistically invariant theory, can be understood

exactly as a collection of oscillators with quantization producing photons.
The modern view of all elementary particles is that they arise by quantizing some
field (though not classically observable like the EM field) but also with interactions
between the modes causing energy transfer between them. This give rise to particle
decay etc. This is the way to build special relativity into QM and allows particles
to be created and destroyed by the action of appropriate combinations of a and a:
on the initial state. This give Quantum Field Theory and the Standard Model of
quarks, leptons, gluons, photons, W ˘ , Z, . . ..
4 Pictures and Operators

4.1 Unitary operators
Physical predictions in QM are given by probability amplitudes, eigenvalues, expecta-
tion values, etc., and hence by expressions of the form xφ|ψy, xφ|A|ψy, etc.
An operator U is called unitary if
U U : “ U :U “ 1 or U : “ U ´1 . (4.1.1)
Given such an operator we can define a map on states
|ψy Ñ
Þ |ψ 1 y “ U |ψy
xψ| Ñ Þ xψ 1 | “ xψ|U : , (4.1.2)
and on operators
A ÞÑ A1 “ U AU : , (4.1.3)
under which all physical properties are unchanged:
xφ|ψy ÞÑ xφ1 |ψ 1 y “ xφ|U : U |ψy “ xφ|ψy ,

xφ|A|ψy ÞÑ xφ1 |A1 |ψ 1 y “ xφ|U : U AU : U |ψy “ xφ|A|ψy . (4.1.4)
Furthermore
C “ AB ÞÑ C 1 “ A1 B 1
C “ rA, Bs ÞÑ C 1 “ rA1 , B 1 s for any operators.
Q hermitian ñ Q1 “ U QU : also hermitian ,

Q|ψy “ λ|ψy ñ Q1 |ψ 1 y “ λ|ψ 1 y same eigenvalue . (4.1.5)
General results for unitary operators (compare with those for hermitian operators)
(i) Eigenvalues are complex numbers of unit modulus.

(ii) Eigenstates corresponding to distinct eigenvalues are orthogonal.
(iii) Any state can be expanded as a linear combination of eigenstates.
Summary: given U there is a basis t|nyu of orthonormal eigenstates for which
U |ny “ eiθn |ny , xn|my “ δnm . (4.1.6)
We prove (i) and (ii) and assume (iii).
(i)
U |ψy “ λ|ψy
ñ xψ|U : “ λ˚ xψ|
(4.1.7)
ñ xψ|U : U |ψy “ k|ψyk2 “ |λ|2 k|ψyk2 ,
and hence |λ| “ 1 p|ψy ‰ 0q .
(ii)
U |ny “ λn |ny
and U |my “ λm |my
(4.1.8)
or xm|U : “ λ˚m xm| “ λ´1
m xm|
ñ xm|U : U |ny “ xm|ny “ λn λ´1m xm|ny .
So λn ‰ λm ñ xm|ny “ 0.
4.2 Schrödinger and Heisenberg Pictures

The solution of the Schrödinger equation
B
i~ |ψptqy “ H|ψptqy (4.2.1)
Bt
can be written
|ψptqy “ U ptq|ψp0qy, (4.2.2)
where the time evolution operator is
ˆ ˙ 8 ˆ ˙n
ítH ÿ 1 ít
U ptq “ exp “ Hn . (4.2.3)
~ n“0
n! ~
This works because

B
i~U ptq “ HU ptq , (4.2.4)
Bt
which is shown by differentiating the power series term by term (we assume H is
independent of t). Note that
U ptq: “ U p´tq “ U ptq´1 unitary pH “ H : q

and U pt1 qU pt2 q “ U pt1 ` t2 q . (4.2.5)
Thus far we have worked in the Schrödinger picture where states depend on time
and operators do not. We can use U ptq to pass to the Heisenberg picture where the
time dependence is shifted from states to operators as follows (subscript denotes the
picture)
Schrödinger Heisenberg
states |ψptqyS |ψyH “ eitH{~ |ψptqyS “ |ψp0qyS
operators AS AH ptq “ eitH{~ AS eítH{~ , AH p0q “ AS

(4.2.6)
Because the transformation is unitary,
S xφptq|AS |ψptqyS “ H xφ|AH ptq|ψyH , (4.2.7)
all physical predictions are the same in either picture. Note that HH “ HS “ H.
The Heisenberg picture makes QM look a little more like classical mechanics where
position, momentum etc. are the variables that evolve in time. To specify the dynamics
in the H-picture we now need an equation to tell us how operators evolve in time. In
the S-picture the Schrödinger equation tells us how states evolve. Now
d d ` itH{~ ˘
AH ptq “ e AS eítH{~
dt dt
iH itH{~ iH
“ e AS eítH{~ ´ eitH{~ AS eítH{~
~ ~
i
“ rH, AH ptqs . (4.2.8)
~
or
i~ d AH ptq “ rAH ptq, Hs Heisenberg equation of motion.

dt
E.g., a particle in one dimension x̂ptq, p̂ptq in Heisenberg picture (drop H subscripts).
We have that
rx̂ptq, p̂ptqs “ i~. (4.2.9)
I.e., the commutator at equal times is unchanged. Consider

p̂2
,
H “ 2m ` V px̂q /
/
/
/
/
/
.
d x̂ptq “ 1 rx̂ptq, Hs 1
“ m p̂ptq , Heisenberg equations of motion
dt i~ /
/
/
/
/
d p̂ptq “ 1 rp̂ptq, Hs “ ´V 1
px̂ptqq .
/
-
dt i~
(4.2.10)
Taking expectation values in any state |ψy (now independent of time in the Heisenberg
picture) gives
d xx̂y “ 1 xp̂y ,
,
dt m /
.
Ehrenfest’s Theorem, true in all pictures (4.2.11)
d xp̂y “ ´xV 1 px̂qy . /
-
dt
Note the similarity to classical equations of motion. For some potentials can solve
Heisenberg’s equations.
‚ V “ 0, the free particle.

d
p̂ptq “ 0 ñ p̂ptq “ p̂p0q constant operator
dt
d 1 p̂p0q
x̂ptq “ p̂p0q ñ x̂ptq “ x̂p0q ` t. (4.2.12)
dt m m
A solution just like in classical dynamics but with appearance of constant operators.
‚ V pxq “ 12 mω 2 x2 , the oscillator.
d x̂ptq “ 1 p̂ptq
, d2 x̂ ` ω 2 x̂ “ 0
dt m dt2
/
.
ñ (4.2.13)
d p̂ptq “ ´mω 2 x̂ptq / 2
- d p̂ ` ω 2 p̂ “ 0 .
dt dt2
The solution is
p̂p0q
x̂ptq “ x̂p0q cos ωt ` sin ωt
mω
p̂ptq “ p̂p0q cos ωt ´ mω x̂p0q sin ωt . (4.2.14)

Can check that the equal-time commutation relation rx̂ptq, p̂ptqs “ i~ holds @ t.
Alternatively, can write these as
ˆ ˙1{2
~
x̂ptq “ pa eíωt ` a: eiωt q
2mω
ˆ ˙1{2
~mω 1
p̂ptq “ pa eíωt ´ a: eiωt q , (4.2.15)
2 i
with a, a: defined from the Schrödinger picture operators x̂p0q, p̂p0q.
4.3 Canonical Quantization

‹ START OF NON-EXAMINABLE MATERIAL
The final step in Dirac’s systematic approach to QM: have seen how to incorporate
position and momentum wavefunctions and S and H pictures in a single logical frame-
work. But how do we pass from general classical system to its quantum version? In
particular, what are the fundamental quantum commutation relations between observ-
ables ; why rx̂, p̂s “ i~?
Any classical system can be described by a set of generalized positions xi ptq and mo-
menta pi ptq with 1 ď i ď N (may include angles, angular momentum etc.) and a
Hamiltonian Hpxi , pi q.
In classical dynamics a fundamental idea is that of the Poisson bracket of any two
functions f pxi , pi q and gpxi , pi q, say, which is defined to be
ÿ ˆ Bf Bg Bg Bf
˙
tf, gu “ ´ , (4.3.1)
i
Bxi Bpi Bxi Bpi
which is a new function of xi and pi . (pxi , pi q are coordinates on phase space and PB
is a symplectic structure.) In particular,
txi , pj u “ δij . (4.3.2)
Properties of the PB include antisymmetry, bilinearity and Jacobi identity. In this

formulation classical dynamics is given by Hamilton’s equation
df
“ tf, Hu . (4.3.3)
dt
Check this for various choices for H and see that you derive Newton’s third law in a
first order formalism (i.e. first-order in time derivative).
In canonical quantization define quantum theory by
‚
classical f, g ÝÑ quantum fˆ, ĝ (4.3.4)
functions operators
‚ Poisson brackets become commutators
rfˆ, ĝs “ i~tf,

{ gu . (4.3.5)
In particular, get
rx̂i , p̂j s “ i~δij (4.3.6)
which are the canonical commutation relations.
‚ Moreover, Hamilton’s equations then become Heisenberg’s equations

d ˆ
i~ f “ rfˆ, Ĥs . (4.3.7)
dt
5 COMPOSITE SYSTEMS AND IDENTICAL PARTICLES 27
For those taking IIC Classical Dynamics this relationship between classical and quan-
tum mechanics should be mentioned near the end of the course.
All this provides a sound basis for understanding classical mechanics as a limit of quan-
tum mechanics with ~ Ñ 0. Going the other way, turning ~ “on” is more problematic
and not guaranteed to be either unique or, in some cases, even consistent. For exam-
ple, if we carry out the procedure above it is correct to Op~q but there may be Op~2 q
ambiguities related to how operators are ordered in defining functions like f pxi , pj q:
does xi multiply pi on left or right?
Alternative approach to quantization is to use path integrals which are sums of contri-
butions from all possible trajectories or paths between initial and final configurations
in phase space.
One of these is the classical trajectory or path. It is derived from an action principle:
the path that minimizes the action associated with the path. However, the quantum
amplitude involves contributions from all trajectories. This approach has its advan-
tages but, in principle, is equivalent to canonical quantization. In general need both,
especially for complicated systems where there are constraints amongst the variables.
‹ END OF NON-EXAMINABLE MATERIAL
5 Composite Systems and Identical Particles

5.1 Tensor products
This is a general way of constructing quantum systems from simpler subsystems. Sup-
pose
|ψy P V1 , |φy P V2 , (5.1.1)
i.e., states in the spaces for two systems. The tensor product space
V “ V1 b V2 (5.1.2)
consists of all linear combinations of tensor product states |ψy b |φy (duals xψ| b xφ|)
subject to
´ ¯
|ψy ` |ψ 1 y b |φy “ |ψy b |φy ` |ψ 1 y b |φy
´ ¯
1
|ψy b |φy ` |φ y “ |ψy b |φy ` |ψy b |φ1 y
´ ¯ ´ ¯ ´ ¯
α|ψy b |φy “ |ψy b α|φy “ α |ψy b |φy , (5.1.3)
and similarly for duals. The inner product is

´ ¯´ ¯
xψ | b xφ | |ψy b |φy “ xψ 1 |ψy xφ1 |φy ,
1 1
(5.1.4)
and extend to all of V by linearity.

If we have bases t|nyu for V1 and t|myu for V2 , then t|nyu b t|myu is a basis for V . If
the bases are finite, then
dim V “ dim V1 ˆ dim V2 . (5.1.5)
Given operators A and B on V1 and V2 , respectively, define an operator A b B on V

by ¯´ ¯ ´ ¯ ´ ¯
`
A b B |ψy b |φy “ A|ψy b B|φy
and extend by linearity. In particular,
A ÐÑ A b I acting just on V1
B ÐÑ I b B acting just on V2 . (5.1.6)
Operators of this form commute for any A & B.

Common abuses of notation:
(i) |ψy b |φy written as |ψy|φy.

(ii) Leave out b I or I b for operators acting on just one subsystem.
Consider a particle in two dimensions with position operators x̂1 , x̂2 . Basis of joint
eigenstates can be constructed as
|x1 , x2 y “ |x1 y b |x2 y

x̂1 Ø x̂1 b I
x̂2 Ø I b x̂2 . (5.1.7)
This is the V “ V1 b V2 tensor product of states for two one-dimensional particles. The
wavefunction for |ψy b |φy is
´ ¯´ ¯
xx1 | b xx2 | |ψy b |φy “ xx1 |ψyxx2 |φy
“ ψpx1 qφpx2 q . (5.1.8)
tensor product separable wavefunctions:

ÐÑ
states product of 1D wavefunctions
tensor product all linear combinations
ÐÑ (5.1.9)
space of such wavefunctions
Example: two-dimensional oscillator.
1 2 1
H “ pp̂1 ` p̂22 q ` mω 2 px̂21 ` x̂22 q
2m 2
“ H1 ` H2 , (5.1.10)
with
ˆ ˙
1 2 1 2 2 1
Hi “ p̂ ` mω x̂i “ ~ω Ni ` ,
2m i 2 2
Ni “ a:i ai , rai , a:j s “ δij . (5.1.11)
Simultaneous eigenstates of N1 , N2 constructed by
|n1 , n2 y “ |n1 y b |n2 y
´ ¯ ´ ¯
with H|n1 , n2 y “ H1 |n1 y b |n2 y ` |n1 y b H2 |n2 y
“ En1 n2 |n1 , n2 y , (5.1.12)
where En1 n2 “ ~ωpn1 ` n2 ` 1q.
5.2 Spin
Experiment shows that particles generally carry an internal degree of freedom called
spin or intrinsic angular momentum. Even if the particle appears ‘elementary’ or
pointlike, its space of states will be of the form V “ Vspace b Vspin with basis
|x, ry “ |xy|ry , (5.2.1)
where r takes a finite set of values: the quantum numbers associated with spin. The
particle is not ‘structureless’: the position operators, x̂, are not a complete commut-
ing set by themselves – there are additional observables Q acting just on Vspin with
rx̂i , Qs “ 0. We will understand these operators later in the study of angular momen-
tum but for now concentrate on the states.
Each kind of particle has a definite total spin S which is a half-integer 0, 21 , 1, 32 , . . .;
this is a basic characteristic like its mass or charge. For a spin S particle of non-zero
mass there are 2S ` 1 basis states in Vspin labelled by convention r “ S, S ´ 1, . . . , ´S.
E.g.,
S basis states
0 |0y
1
| 12 y, | ´ 12 y also written Òyon, loo|mo (5.2.2)
2 loo|mo Óyon
up down
1 |1y, |0y, | ´ 1y
The existence of spin states is revealed by e.g. the Stern-Gerlach experiment
beam of
splits beam in two:
s = 1/2 in this case
particles
magnet producing different force on

inhomogeneous field different spin states
The general state |ψy in V has a p2S ` 1q-component wavefunction
ψpx, rq “ xx, r|ψy . (5.2.3)
It is also useful to write the state in mixed, or hybrid, notation as

ÿ
ψpx, rq|ry . (5.2.4)
r
5.3 Multiparticle states: bosons and fermions

Consider particles labelled by a “ 1, 2, . . . , N . Let Va be the space of states for each
particle with basis t|xa , ra yu - position and spin labels. The general multiparticle states
belong to
V1 b V2 b . . . b VN , (5.3.1)
with basis states
|x1 , r1 ; x2 , r2 ; . . . ; xN , rN y “ |x1 , r1 y b |x2 , r2 y b . . . b |xN , rN y
constructed from single particle states. If the particles are identical, Va u V , some-
thing interesting can be added.
Consider the simplest case N “ 2. Define an operator W which exchanges particles by
its action on basis states:
W |x1 , r1 ; x2 , r2 y “ |x2 , r2 ; x1 , r1 y . (5.3.2)
When the two particles are identical its action on a general 2-particle state is
W |Ψy “ η|Ψy , (5.3.3)
because |Ψy and W |Ψy must be physically equivalent if the particles are indistin-
guishable. But, given its action on the basis states
W 2 “ 1 ñ η 2 “ 1 or η “ ˘ 1 . (5.3.4)
Thus, 2-particle states do not belong merely to

V bV spanned by |ψy b |φy ,
but to pV b V qS spanned by |ψy b |φy ` |φy b |ψy ,

η “ 1 symm under W (5.3.5)
or to pV b V qA spanned by |ψy b |φy ´ |φy b |ψy .

η “ ´1 antisymm under W
Similarly, for multiparticle states with N ě 2 we can define Wpa,bq which exchanges
pxa , ra q Ø pxb , rb q by this action on the basis states. Then for a general N -particle
state
Wpa,bq |Ψy “ ηpa,bq |Ψy (5.3.6)
2
with, again, ηpa,bq “ ˘1 because Wpa,bq “ 1.
For any permutation π of t1, 2, . . . , N u define
Wπ |x1 , r1 ; x2 , r2 ;. . . ; xN , rN y
“ |xπp1q , rπp1q ; xπp2q , rπp2q ; . . . ; xπpN q , rπpN q y (5.3.7)
on the basis states. On a general state
Wπ |Ψy “ ηπ |Ψy for some ηπ . (5.3.8)
But algebra of swaps or transpositions implies ηpa,bq “ ˘ 1 with the same value for
all pairs pa, bq since any two swaps are conjugate. This makes physical sense since the
particle are identical and the initial choice for the labelling is not unique. Then, since
any π can be obtained as a sequence of swaps, we have alternative outcomes
#
1
ηπ “ (5.3.9)
sgnpπq “ p´1qp# swaps needed for πq ,
with the same alternative for all π. These correspond to two inequivalent 1-D repre-
sentations of the permutation group.
Hence, there are two fundamentally different kinds of particles:
‚ Bosons obeying Bose-Einstein statistics:

(i) interchange of identical bosons leaves the state unchanged: η “ 1;
(ii) multiparticle states in pV b V b . . . b V qS .
‚ Fermions obeying Fermi-Dirac statistics:
(i) interchange of identical fermions changes the state by a sign: η “ ´1;
(ii) multiparticle states in pV b V b . . . b V qA .
Note that this applies only to identical particles. Indistinguishability has a different
character in QM from classical physics. It is the consequence of saying that you cannot
attach a label to a given particle and uniquely identify it from any other. You can no
longer follow individual particles because of the uncertainly principle.
In addition have the remarkable ‹ Spin-statistics relation. ‹
5.4 Spin and statistics

Integral spin particles (S “ 0, 1, 2, . . .) are bosons.
1
2
-integralspin particles (S “ 12 , 23 , . . .) are fermions.
This is confirmed by experiment and can be derived in relativistic QM or QFT (quan-

tum field theory) so becomes the spin-statistics theorem; no derivation is know within
non-relativistic QM. The derivation relies, in particular, on Lorentz symmetry (sym-
metry under Lorentz transformations and rotations) and causality (no (space-time)
event can influence a future event if they are separated by a space-like space-time
interval; no light signal can connect them).
‚ Most common elementary particles are fermions: electrons, protons, neutrons, neu-
trinos, quarks, muons, τ – all spin 12 .
‚ Particles associated with forces are bosons: photons (EM), W ˘ , Z (weak nuclear),
gluons (strong nuclear) – all spin 1.
‚ Other particles such as mesons are bosons e.g., π, K are spin 0, the ρ is spin 1, and
many more have been observed with higher spin.
‚ The recently discovered Higgs boson (LHC experiments) has almost certainly spin 0
although this is still to be confirmed.
‚ The graviton has spin 2 but is yet to be observed - not likely in the near future (if
ever).
‚ The spin-statistics theorem applies even if the particles are not ‘elementary’. Indeed,
nucleons (proton etc) and mesons are made of quarks. Atoms obey the theorem, too.
5.5 Two-particle examples

Now know that the states of two identical particles belong to
pV b V qS or pV b V qA where V “ Vspace b Vspin . (5.5.1)
It is convenient to construct these spaces by first taking
p Vspace b Vspace qS,A and p Vspin b Vspin qS,A (5.5.2)
separately, and then combining to get the correct overall symmetry.

Consider the single particle Hamiltonian Hpx̂, p̂q independent of spin with non-degenerate
energies E0 ă E1 ă E2 ă . . ., and wavefunctions ψ0 pxq, ψ1 pxq, ψ2 pxq, . . . for states in
Vspace . For two such non-interacting particles the Hamiltonian is
Hpx̂1 , p̂1 q ` Hpx̂2 , p̂2 q , (5.5.3)
with the wavefunctions for the basis of energy eigenfunctions for Vspace b Vspace
ψi px1 qψj px2 q E “ Ei ` Ej . (5.5.4)

‚ Ground state, E “ 2E0 :

ΨS0 px1 , x2 q “ ψ0 px1 qψ0 px2 q , (5.5.5)
which is automatically (S)ymmetric.
‚ 1st excited states, E “ E0 ` E1 :
ψ0 px1 qψ1 px2 q , ψ1 px1 qψ0 px2 q . (5.5.6)
These have no definite symmetry but the following linear combinations do:
1
ΨS1 px1 , x2 q “ ? pψ0 px1 qψ1 px2 q ` ψ1 px1 qψ0 px2 qq , (5.5.7)
2
1
ΨA
1 px1 , x2 q “ ? pψ0 px1 qψ1 px2 q ´ ψ1 px1 qψ0 px2 qq , (5.5.8)
2
which are (S)ymmetric and (A)ntisymmetric, respectively. They are the S and A
1-D representations of the two-particle permutation group S2 .
Spin 0: Vspin is just one state so can ignore. Spin-statistics ñ the particles are
bosons, and states are in pV b V qS which in this case is pVspace b Vspace qS . The
allowed states are then
*
Ground state ΨS0 px1 , x2 q 2E0 both non-
degenerate (5.5.9)
1st excited state ΨS1 px1 , x2 q E0 ` E1
Note that only one of the two possible 1st excited states is allowed.
Spin 21 : Vspin has basis t| Òy, | Óyu for each particle. Vspin b Vspin therefore has
the basis
t| Òy| Òy, | Òy| Óy, | Óy| Òy, | Óy| Óyu , (5.5.10)
or normalized combinations with definite symmetry are
| Òy| Òy
´ ¯ ´ ¯
?1 | Òy| Óy ` | Óy| Òy ?1 | Òy| Óy ´ | Óy| Òy
2 2
(5.5.11)
| Óy| Óy
pVspin b Vspin qS pVspin b Vspin qA

We would expect these two kinds of states, S and A, to be associated with a definite
spin quantum number. We can see what this is simply by counting since the degen-
eracy for spin S is 2S ` 1. We see that the S-states have S “ 1 and the A-state has
S “ 0. Note how state multiplets with a definite spin also have a definite symmetry.
This is no accident.
Spin-statistics ñ the particles are fermions, and the states belong to
pV b V qA “ pVspace b Vspace qA b pVspin b Vspin qS
` pVspace b Vspace qS b pVspin b Vspin qA (5.5.12)
The allowed states are (in hybrid notation, see Eq. (5.2.4))
‹ Ground state, E “ 2E0 :

1 ´ ¯
ΨS0 px1 , x2 q ? | Òy| Óy ´ | Óy| Òy (5.5.13)
2
and there is just one state. (Note that, in general, the ground state of a system
is usually non-degenerate.)
‹ 1st excited states, E “ E0 ` E1 :
1 ´ ¯
ΨS1 px1 , x2 q ? | Òy| Óy ´ | Óy| Òy (5.5.14)
2
¨ ˛
´ | Òy| Òy ¯
ΨA
1 px1 , x2 q ˝
?1 (5.5.15)
˚ ‹
2
| Òy| Óy ` | Óy| Òy ‚
| Óy| Óy
There are four states in all made up of 1 ` 3 “ 4. The spin quantum numbers
associated with these are S “ 0 and S “ 1, respectively.
We can see that representations of the two-particle permutation group, S2 , are used
extensively in the analysis. The general analysis for N identical particles is based on
the representation theory of SN , the permutation group for N particles.
5.6 Pauli Exclusion Principle and Atomic Structure

A state of N identical fermions can be specified by taking N distinct one-particle states
(each in V ) and antisymmetrizing to ensure Fermi statistics; this produces a state in
pV b . . . b V qA . The condition that the one-particle states must all be different, else the
result is zero, is the Paul Exclusion Principle. The complete set of such N -particle
antisymmetric states from pV b . . . b V qA forms a basis for the most general state of
N fermions.
The original application was to atomic structure. Consider N electrons bound to a
nucleus with Z protons. Ignoring electron-electron interactions, the Hamiltonian is
Hpx̂1 , p̂1 q ` . . . ` Hpx̂N , p̂N q , (5.6.1)
where a single electron Hamiltonian is
1 2 Ze2
Hpx̂, p̂q “ p̂ ´ . (5.6.2)
2m 4π0 |x̂|
Single electron states in Vspace are similar to hydrogen atom states (for which Z “ 1)
$
’
’ evals
H En
&
|n, l, my which are joint eigenstates of 2 2 (5.6.3)
’
’ L ~ lpl ` 1q
L3
%
~m
where
ˆ ˙2
m Ze2 1 2 Z
2
En “ ´ 2
“ ´ α 2
mc2 ,
2 4π0 ~ n 2n
n “ 1, 2, . . . principal quantum number ,
l “ 0, 1, . . . , n ´ 1 m “ 0, ˘1, ˘2, . . . , ˘l ,
looooooooomooooooooon (5.6.4)
2l ` 1 states
where α “ e2 {p4π0 ~cq « 1{137 is the fine-structure constant. So degeneracy (excluding
spin!) is
1 ` 3 ` . . . ` p2n ´ 1q “ n2 for level En . (5.6.5)
Now, including Vspin basis t| Òy, | Óyu, each electron has 2n2 states at energy En .
Structure of electrically neutral atoms with N “ Z:
‚ Fill up energy levels, starting with the lowest, using Pauli Principle.
‚ This gives a rough picture of the periodic table with some qualitative insights into
chemical properties.
‚ The states that belong to a given energy En is called a shell. Atoms with completely
filled shells are unreactive/stable elements chemically. E.g.
E1 filled for Z “ 2 Ñ He
E1 & E2 filled for Z “ 10 Ñ Ne (5.6.6)
‚ Chemical reactivity is controlled by the valency which is given by the number of

electrons in an almost empty outer shell (Ca2` ) or the number of unoccupied states
in a nearly filled outer shell (F´ ).
‚ In real life there are many interactions which perturb the energies of the individual
levels and spread the energies within a shell even to the extent that shells can overlap.
‚ Each electron has a magnetic moment because it is a spinning charge – it acts
like a bar magnet (indeed an iron magnet is simply the sum of all these small elec-
tron magnets). Should include electron-electron interactions, the interactions of the
electron magnetic moments with each other and also with magnetic fields due to the
orbiting electric charge (see later section 7).
atomic n 1 2 3 4
no. element s s p s p d s p d f
1 H 1
2 He 2
6 C 2 2 2
9 F 2 2 5
10 Ne 2 2 6
11 Na 2 2 6 1
19 K 2 2 6 2 6 1
20 Ca 2 2 6 2 6 2
26 Fe 2 2 6 2 6 6 2
28 Ni 2 2 6 2 6 8 2
29 Cu 2 2 6 2 6 10 2
6 PERTURBATION THEORY 36
Table shows some electron configurations. s,p,d,f Ñ l “ 0, 1, 2, 3. Above atomic

number 18 the states are not always filled in naive order because of interactions between
electrons. Can use Aufbau or Madelung rule to get better sequencing. For these see
(http://en.wikipedia.org/wiki/Electron configuration.)
The eigenstate of the fully interaction N -electron Hamiltonian must be expressed as a
linear combination of the basis states in tV b . . . b V qA constructed here. The most
general problem is still a subject of active research.
In the next section we discuss perturbation theory and how we may systematically
calculate the effect of adding new interactions to a solved system if they are sufficiently
weak.
6 Perturbation Theory
Few quantum mechanical systems can be solved exactly. In perturbation theory we
start from a known, soluble, system
H|ny “ En |ny , (6.1)
with t|nyu an orthonormal basis of energy eigenstates, and calculate the energies and
eigenstates for a new perturbed Hamiltonian
pH ` µV q|ψy “ E|ψy , (6.2)
order-by-order in parameter µ which is supposed, in some sense, to be small. V is

some hermitian operator and we assume that E and |ψy depend smoothly on µ. The
discussion here is for perturbations to H but the technique applies to any observable.
r Note that methods of perturbation theory are not just applicable in QM but to the
solution of perturbed PDEs in general. u
6.1 The non-degenerate case

Suppose that as µ Ñ 0
"
|ψy Ñ |ry unique eigenstate with this en-
(6.1.1)
E Ñ Er ergy, i.e., level is non-degenerate.
States t|nyu are still a basis when µ ‰ 0, and so we can always write
ÿ
|ψy “ α|ry ` βj |jy
j‰r
ÿ
“ αp|ry ` γj |jyq , (6.1.2)
j‰r
where α, βj , γj “ βj {αj are power series in µ such that
α Ñ 1, βj , γj Ñ 0 as µ Ñ 0 . (6.1.3)
Aim is to calculate the coefficients in these expansions

p1q p2q
E “ Er ` Er µ ` Er µ2 ` . . .
α “ 1 ` a1 µ ` a2 µ 2 ` . . . (6.1.4)
γj “ cj1 µ ` cj2 µ2 ` . . .
First just substitute the expression for |ψy into Eq. (6.2):
ÿ ÿ
pEr ` µV q|ry ` γj pEj ` µV q|jy “ Ep|ry ` γj |jyq , (6.1.5)
j‰r j‰r
cancelling overall factor of α. Rewrite as

ÿ ÿ
pE ´ Er q|ry ` γj pE ´ Ej q|jy “ µV |ry ` µ γj V |jy . p˚˚q (6.1.6)
j‰r j‰r
Then taking inner product of xr| with p˚˚q gives

ÿ
E ´ Er “ µxr|V |ry ` µ γj xr|V |jy , (6.1.7)
j‰r
and, so far, this is still exact (all orders in µ). Substituting in the series expansions
from Eq. (6.1.4) and keeping terms to Opµ2 q gives
ÿ
µErp1q ` µ2 Erp2q ` . . . “ µxr|V |ry ` µ2 cj1 xr|V |jy ` . . . . (6.1.8)
j‰r
Thus, already know first-order energy shift

Erp1q “ xr|V |ry . (6.1.9)
To find second-order shift we need to know cj1 . The inner product of xi| with p˚˚q
(i ‰ r) gives ÿ
γi pE ´ Ei q “ µxi|V |ry ` µ γj xi|V |jy . (6.1.10)
j‰r
Again this is exact but the 2nd term on RHS is Opµ2 q, and so to leading order in µ
µ ci1 pEr ´ Ei q “ µxi|V |ry
xi|V |ry
ñ ci1 “ (note Ei ‰ Er since states non-degenerate)
Er ´ Ei
(6.1.11)
and so substituting in Eq. (6.1.8) we find
ÿ xr|V |jyxj|V |ry ÿ |xj|V |ry|2
Erp2q “ “ . (6.1.12)
j‰r
Er ´ Ej j‰r
Er ´ Ej
In summary,
ÿ |xj|V |ry|2
2
E “ Er ` µxr|V |ry ` µ ` ... , (6.1.13)
j‰r
Er ´ Ej
˜ ¸
ÿ xj|V |ry
|ψy “ α |ry ` µ |jy ` ... , (6.1.14)
j‰r
Er ´ Ej
where α is chosen so that |ψy has unit norm. This is second order perturbation theory.
Example:
p̂2
´ ¯
1 2 2
H “ 2m ` 2 mω x̂ “ ~ω a a ` 2 : 1
´ ¯ (6.1.15)
states |ny, En “ ~ω n ` 2 . 1
~ 2
Perturb with V “ mω 2 x̂2 “
2
mω pa2 ` a: ` 2a: a ` 1q. (6.1.16)
2
m
ω
Have
1
xn|V |ny “ 2
~ωp2n ` 1q
1
? ?
xn ` 2|V |ny “ 2
~ω n ` 1 n `2
1
? ?
xn ´ 2|V |ny “ 2
~ω n n ´ 1
xm|V |ny “ 0 all other m (6.1.17)
Perturbed energy of nth level to Opµ2 q:

ÿ |xm|V |ny|2
En ` µxn|V |ny ` µ2 En ´ Em
m‰n
¯2 ˆ ˙
pn ` 1qpn ` 2q npn ´ 1q
´
“ ~ωpn ` 1
q ` µ~ωpn ` 1
q ` µ 2 ~ω ` 2~ω
2 2 2 ´2~ω
m“n`2 m“n´2
` ˘` ˘
“ ~ω n ` 21 1 ` µ ´ 12 µ2 . (6.1.18)
But this problem can be solved exactly:

a
H Ñ H ` µV equivalent to ω Ñ ω1 “ ω 1 ` 2µ . (6.1.19)
New energies are exactly

1
~ω 1 pn ` 12 q “ ~ωpn ` 12 qp1 ` 2µq 2
“ ~ωpn ` 21 qp1 ` µ ´ 12 µ2 ` Opµ3 qq . (6.1.20)
Validity/usefulness of perturbation theory

If µ dimensionless might expect µ ! 1 enough to ensure rapid convergence such as
in above example. However, this may not be enough since we still require a non-zero
radius of convergence. Really have an expansion in
ˇ ˇ
ˇ xi|V |jy ˇ
ˇµ
ˇ Ei ´ Ej ˇ i ‰ j , (6.1.21)
ˇ
and similar quantities. Condition for accuracy is therefore qualitatively
|energy shifts| ! |original energy differences| . (6.1.22)

H evals H+:V evals
Indeed, it may be that the series is only asymptotic and more sophisticated methods
are needed to estimate the energy shifts within a given accuracy.
Example. Ground state energy for Helium.
The unperturbed problem is two non-interacting electrons, charge é orbiting nucleus
with charge `2e. The Hamiltonian is
p̂2 2e2
Hpx̂1 , p̂1 q ` Hpx̂2 , p̂2 q with Hpx̂, p̂q “ ´ . (6.1.23)
2m 4π0 |x̂|
Single electron states and energies are
ˆ ˙2
|n l my 1 2e2 1 2α2 2
En “ ´ m ” ´ mc , (6.1.24)
wavefn ψnlm pxq 2 4π0 ~ n2 n2
2
where α “ 4πe ~c is the dimensionless fine structure constant: α « 1{137. The lowest
0
energy eigenstate for two electrons is
|Ψy “ |1 0 0y b |1 0 0y b |χy
1 ´ ¯
with |χy “ ? | Òy| Óy ´ | Óy| Òy the spin state . (6.1.25)
2
This state is totally antisymmetric since constrained by Fermi statistics. The two
electron wavefunction is then
Ψpx1 , x2 q “ ψ100 px1 qψ100 px2 q
ˆ ˙ 32
1 2
ψ100 pxq “ ? e´|x|{a2
π a2
ˆ ˙
1 1 ~
with pZ “ 2q a2 “ 2 “ 12 ˚ pBohr radiusq (6.1.26)
α mc
The two-electron unperturbed energy is
2E1 “ ´ 4α2 mc2 « ´ 108.8 eV . (6.1.27)
Compare with -13.6 eV for the hydrogen atom. (Note: mc2 « 500 KeV.)
Experimentally, the ground state for He is -79.0 eV. However, have neglected the
electron-electron interaction:
e2 1
. (6.1.28)
4π0 |x̂1 ´ x̂2 |
Treat this as a perturbation and write as µV px̂1 , x̂2 q with
~c
µ “ α, V px̂1 , x̂2 q “ . (6.1.29)
|x̂1 ´ x̂2 |
Then µ is dimensionless and V has energy dimensions. The first-order correction to

the energy is then
ż ż
αxΨ|V |Ψy “ α d x1 d3 x2 Ψ˚ px1 , x2 qV px1 , x2 qΨpx1 , x2 q
3
ż ż
~c
“ α d x1 d3 x2 |ψ100 px1 q|2 |ψ100 px2 q|2
3
|x1 ´ x2 |
5 2 2
“ α mc « 34.0 eV . (6.1.30)
4
The corrected ground state energy is then ´108.8 eV ` 34.0 eV « ´74.8 eV; in much
better agreement with experiment. Note that the variational principle is more efficient
(see AQM) at this level but does not help with higher-order corrections.
We might naively expect that the perturbation series is an expansion in α but we see
p1q
that E1 and the first-order correction E1 are both 9 α2 . In fact, all corrections are
9 α2 , so what is the expansion parameter? Including the 2nd-order correction the
energy can be written 2
ˆ ˙ˇ
2 2 2 51 25 1 ˇ
E “ ´ Z α mc 1 ´ ` ` . . . ˇ . (6.1.31)
8 Z 256 Z 2 ˇ
Z“2
So the expansion is in 1{Z for Z “ 2. The series “looks” convergent and gives an
answer close to experiment but to my knowledge it is not known if it actually converges.
However, treated as an asymptotic series it does give a believable answer. This is typical
of many problems in bound state systems.
In scattering theory (Quantum Electrodynamics) the expansion parameter is α but
convergence is still not provable.
6.2 Degenerate case

Consider the perturbed Hamiltonian in Eq. (6.2) but now with the possibility that
H has degenerate eigenstates. Let the degeneracy of states in Vλ be N with common
eigenvalue λ. Then use notation
E1 “ . . . “ Er “ Es “ . . . “ EN “ λ degenerate states
(6.2.1)
|1y, . . . |ry, |sy, . . . |N y reserve r, s to label these only
and suppose that as µ Ñ 0

ÿ
E Ñ λ, |ψy Ñ ar |ry . (6.2.2)
r
2
This is the correct expression for all atoms with two electrons. For example, Li` which has Z “ 3
and Be2` which has Z “ 4.
Denote other states, not in this degenerate set, by |jy and reserve j to label these only.
Then, for general µ, expand the eigenfunction of the perturbed Hamiltonian as
E “ λ ` µE p1q ` µ2 E p2q ` . . . ,
ÿ ÿ
|ψy “ αr |ry ` βj |jy .
r j
with αr “ ar ` Opµq, βj “ Opµq . (6.2.3)
Note that for µ Ñ 0, |ψy Ñ |ψ0 y P Vλ but we do not yet know the values of the ar
which determine it uniquely in this limit. We shall see that the possible choices for the
ar are determined by the perturbation itself. Substitute into Eq. (6.2) and find
˜ ¸
ÿ ÿ ÿ ÿ
αr pEr ` µV q|ry ` βj pEj ` µV q|jy “ E αr |ry ` βj |jy . (6.2.4)
r j r j
Rearranging gives
ÿ ÿ ÿ ÿ
αr pE ´ Er q|ry ` βj pE ´ Ej q|jy “ µ αr V |ry ` µ βj V |jy . p˚˚q
r j
loooooooooooooooooooooooomoooooooooooooooooooooooon r j
loooooooooooooooomoooooooooooooooon
terms containing energy shifts terms containing V
(6.2.5)
Look for energy shift E “ λ ` µE p1q ` Opµ2 q and note that the correction is no longer
associated with a single state of the unperturbed problem but with all N states in Vλ .
Take the inner product of xs| with p˚˚q to get (remember Er “ . . . “ Es . . . “ EN “ λ)
ÿ ÿ
pE ´ Es qαs “ µ αr xs|V |ry ` µ βj xs|V |jy , (6.2.6)
r j
but the second term on RHS is Opµ2 q and αr “ ar ` Opµq, and so hence to leading
order ÿ
xs|V |ry ar “ E p1q as . (6.2.7)
r
p1q
Thus, E is an eigenvalue of the matrix xs|V |ry and the ar are the components of
the corresponding eigenvector. Since there are N degenerate states this is an N ˆ N
matrix and so have N solutions for E p1q with associated eigenvectors giving the ar in
each case.
We should expect something like this:
‚ We started with N degenerate states and we end up with N perturbed states.

‚ The degeneracy of H is a reflection of some symmetry (e.g., rotational symmetry ñ
L2 , L3 are good quantum numbers) and the perturbed Hamiltonian H `µV generally
has less symmetry and so the full degeneracy is broken, although some degeneracy
amongst a subset of states can remain. The effect is to spilt apart some, or all, levels.
‚ Raising the degeneracy in this way is important in many physical phenomena. For
example, band structure for the electron levels in a crystal giving rise to delocalization
of electrons originally bound in each atom and so leading to electrical conduction
(see AQM).
Example. Particle in a box in two dimensions, 0 ď x, y ď a. The unperturbed

problem is the free particle with states |p, qy and
2 pπx qπy
wavefns ψpq px, yq “ sin sin
a a a
~2 π 2 2
energies Epq “ pp ` q 2 q p, q “ 1, 2, . . . . (6.2.8)
2ma2
2 2
‚ Lowest level, E11 “ ~ π2 , is non-degenerate.
ma
2 2
‚ Next level has E12 “ E21 “ 5~ π2 and so has degeneracy 2.
2ma
xy
Consider the perturbation V px, yq “ , and then µ has dimensions of energy.
a2
‚ The shift in the lowest energy level is given by

ża ża
xy
x11|V |11y “ dx dy 2 |ψ11 px, yq|2
0 0 a
ˆ ˙2 ż a
πx ¯2 a
ż
2 1 ´ ´ πy ¯2
“ dx x sin dy y sin
a a2 0 a 0 a
1
“ . (6.2.9)
4
(6.2.10)
„ż a 2

´ πx ¯2 a
dx x sin “
0 a 4
‚ Shifts in next level given by eigenvalues of

ˆ ˙ ˆ ˙
x12|V |12y x12|V |21y α β
“ , (6.2.11)
x21|V |12y x21|V |21y β α
with
ża ża
xy
α “ dx dy 2 |ψ12 px, yq|2
0 0 a
ˆ ˙2 ż a ˙2
πx ¯2 a
ż ˆ
2 1 ´ 2πy
“ dx x sin dy y sin
a a2 0 a 0 a
1
“ . (6.2.12)
4
7 ANGULAR MOMENTUM 43
ża ża
xy ˚
β “ dx dy 2 ψ12 px, yqψ21 px, yq
0 0 a
ˆ ˙2 ˆ ż a ˙2
2 1 πx 2πx
“ dx x sin sin
a a2 0 a a
ˆ ˙2
16
“ . (6.2.13)
9π 2
(6.2.14)
„ż a ża ˆ ˙ 
πx 2πx 1 πx 3πx 8a2
dx x sin sin “ dx x cos ´ cos “´ 2 .
0 a a 0 2 a a 9π
The eigenvalues of the matrix are α ˘ β with respective eigenvectors

1 1
a ” pa12 , a21 q “ ? p1, 1q, ? p1, ´1q (6.2.15)
2 2
Collecting results:
2 2 µ
‚ New ground state energy ~ π2 ` 4 .
ma
‚ Next two levels
5~2 π 2 µ 256
energies 2 ` ˘µ
2ma 4 81π 4
1 ´ ¯
states |ψy “ ? |12y ˘ |21y . (6.2.16)
2
Note that the state is unperturbed at first order in µ but the perturbation does
determine the choice of basis in Vλ in order that the perturbation can be carried
out systematically. This basis is the eigenstates of the matrix with elements xs|V |ry.
7 Angular Momentum
7.1 Recap of orbital angular momentum
Mainly to set the analysis to follow in some sort of context but also a few points of
special importance.
L “ x̂ ^ p̂ or Li “ εijk x̂j p̂k L 2 “ Li Li . (7.1.1)

3
Then
rx̂i , p̂j s “ i~δij ñ rLi , Lj s “ i~ εijk Lk , rL2 , Li s “ 0 . (7.1.2)
3
Use Leibnitz properties for commutators Eqn. (2.1.19):
rAB, Cs “ ArB, Cs ` rA, CsB

In addition to these fundamental commutation relations we have
rLi , x̂j s “ i~ εijk x̂k , rLi , p̂j s “ i~ εijk p̂k . (7.1.3)
Consider the action of all these operators on wavefunctions. Using spherical polar
coordinates pr, θ, φq, the operators Li only involve angular derivatives and
1 B2 1
∇2 “ 2
r ´ 2 ~2
L2 . (7.1.4)
r Br
loomoon r
loomoon
radial angular
The joint eigenstates of L2 and L3 are the spherical harmonics Ylm pθ, φq:
L2 Ylm “ ~2 lpl ` 1q Ylm

L3 Ylm “ ~m Ylm
|m|
with Ylm “ clm eimφ Pl p cos θq
dm
where Plm puq “ p1 ´ u2 qm{2 m Pl puq , mě0. (7.1.5)
du
Pl puq is a Legendre polynomial

Plm puq is an associated Legendre function .
Only well-behaved solutions of eigenvalue problems for solving differential equations

arise for
l “ 0, 1, 2, 3, . . . and m “ 0, ˘1, ˘2, . . . , ˘l . (7.1.6)
If the Schrödinger equation has a spherically symmetric potential, V pxq “ V prq then
we can look for separable solutions
ψlm pxq “ Rl prqYlm pθ, φq , (7.1.7)
where ˆ ˙
~2 1 B 2 ~2
´ prRl q ` lpl ` 1q ` V prq Rl “ ERl . (7.1.8)
2m r Br2 2mr2
Of particular importance is the behaviour of such solutions under parity: x ÞÑ ´x.
This is equivalent to
r ÞÑ r, θ ÞÑ π ´ θ, φ ÞÑ φ ` π . (7.1.9)
Then, under parity map have
Ylm pπ ´ θ, φ ` πq “ p´1ql Ylm pθ, φq

ñ ψlm p´xq “ p´1ql ψlm pxq . (7.1.10)
A very important example of everything above is a system of two particles interacting

through a potential which depends only on their relative separation
p̂21 p̂2
H “ ` 2 ` V p|x̂1 ´ x̂2 |q
2m1 2m2
2
P̂ p̂2
“ ` ` V p|x̂|q , (7.1.11)
2M 2m
where
X̂ “ m1 x̂1 M
` m2 x̂2 x̂ “ x̂1 ´ x̂2
m2 p̂1 ´ m1 p̂2
P̂ “ p̂1 ` p̂2 p̂ “ M (7.1.12)
M “ m1 ` m2 m “ mM
1 m2 reduced mass
CoM degrees of freedom relative degrees of freedom
CoM dynamics essentially trivial and relative motion is governed by a spherically sym-
metric potential. The total wavefunction is then
ΨpX, xq “ eik¨X ψlm pxq w.r.t L “ x̂ ^ p̂ . (7.1.13)
I.e., plane wave solution in the CoM variables and effective single particle dynamics in
potential V p|x̂|q with reduced mass m.
Under x ÞÑ ´x have Ψ ÞÑ p´1ql Ψ. This is particularly important if the two particles
are identical.
7.2 General analysis of angular momentum eigenstates

Consider J “ pJ1 , J2 , J3 q Ji: “ Ji
obeying rJi , Jj s “ i~εijk Jk
(7.2.1)
Define J2 “ Ji Ji hermitian
then rJ 2 , Ji s “ 0
We want to find all possible eigenstates of J 2 and J3 (say) assuming just the commuta-
tion relations above. Use J rather than L because there will turn out to be additional
possibilities beyond those allowed for orbital angular momentum. First observe that
J 2 |ψy “ λ|ψy ñ
2
λ “ xψ|J |ψy “ kJ1 |ψyk2 ` kJ2 |ψyk2 ` kJ3 |ψyk2 ě 0 . (7.2.2)
It is convenient to set λ “ ~2 jpj ` 1q with j ě 0 without loss of generality (note: λ

unchanged if j Ñ ´pj ` 1q). So label the states |j, my where
J 2 |j, my “ ~2 jpj ` 1q|j, my J3 |j, my “ ~m|j, my . (7.2.3)
So far all we know is that m and j ě 0 are real numbers. To analyze the allowed
eigenvalues we define
J˘ “ J1 ˘ iJ2 , J˘: “ J¯ , (7.2.4)
and work with these new combinations. It is easy to check that
rJ3 , J˘ s “ ˘~J˘
rJ` , J´ s “ 2~J3
rJ 2 , J˘ s “ 0 . (7.2.5)
Furthermore, we find
J` J´ “ J 2 ´ J32 ` ~J3
J´ J` “ J 2 ´ J32 ´ ~J3 (7.2.6)
and also find that

J2
"
with j unchanged
J˘ |j, my is an eigenstate of (7.2.7)
J3 with m Ñ m ˘ 1
and, strictly, provided that the new states do not vanish. These statements follow from
the commutation relations because
´ ¯ ´ ¯
J3 J˘ |j, my “ rJ3 , J˘ s ` J˘ J3 |j, my
´ ¯
“ ˘ ~J˘ ` J˘ ~m |j, my
“ ~pm ˘ 1qJ˘ |j, my . (7.2.8)
Compare with the harmonic oscillator where
rN, a: s “ a: and rN, as “ á ñ a: |ny, a|ny have eigenvalues n ˘ 1 . (7.2.9)
To find out whether the new states vanish we compute
kJ` |j, myk2 “ xj, m|J´ J` |j, my

“ xj, m|J 2 ´ J32 ´ ~J3 |j, my
` ¯
“ ~2 jpj ` 1q ´ mpm ` 1q
“ ~2 pj ´ mqpj ` m ` 1q
ě 0, (7.2.10)
and
kJ´ |j, myk2 “ xj, m|J` J´ |j, my

“ xj, m|J 2 ´ J32 ` ~J3 |j, my
` ¯
“ ~2 jpj ` 1q ´ mpm ´ 1q
“ ~2 pj ` mqpj ´ m ` 1q
ě 0. (7.2.11)
These inequalities, which follow just because the inner-product is positive-semidefinite,

imply (remember j ě 0)
j ě m ě ´pj ` 1q
(7.2.12)
j ` 1 ě m ě ´j
and the states vanish only when equality occurs. Hence, from these bounds, we deduce
j ě m ě ´j
J` |j, my “ 0 iffi m “ j , (7.2.13)
J´ |j, my “ 0 iffi m “ ´j .
These results tell us all we need to know about the possible values of both j and m.
Remember, so far only know that m and j ě 0 are real.
Start from any given state |j, my.
‚ If the states J`n |j, my are non-vanishing, they

have J3 eigenvalues ~pm ` nq but subject to the
bound j ě m ` n from above.
‚ To avoid a contradiction (c.f. Eqn. (7.2.13))

there must be an integer k with j “ m ` k and j
hence k
J`n |j, my “ 0 @ n ą k. (7.2.14) m
Only then will the sequence terminate at m “ j.

k'
‚ Similarly, if the states Jń |j, my
are non-
vanishing, they have J3 eigenvalues ~pm ´ nq -j
but subject to the bound m ´ n ě ´j .
‚ In this case must have an integer k 1 with ´j “

m ´ k 1 so that
Jń |j, my “ 0 @ n ą k 1 . (7.2.15)
Thus,
j “m`k and j “ ´m ` k 1 , k, k 1 integer
ñ 2j “ k ` k 1 , an integer. (7.2.16)
Furthermore, for a given value of j we have states
|j, my with m “ j, j ´ 1, . . . , ´j ` 1, ´j , a total of 2j ` 1 states. (7.2.17)
This general analysis has revealed two possibilities:
j integral m “ 0 , ˘1, ˘2, . . . , ˘j odd # states,
(7.2.18)
j half-integral m “ ˘ 21 , ˘ 23 , . . . , ˘j even # states.
j integral. This possibility is realized in orbital angular momentum. The states |j, my
correspond to wavefunctions ψjm pxq or Yjm pθ, φq.
j half-integral. This possibility cannot arise for orbital angular momentum since
there are no solutions of the differential equations which are well-behaved in this case.
Such states must correspond to intrinsic angular momentum or spin as introduced
earlier.
We usually write J “ S for spin. Our analysis shows that we must have j “ S integral
or half-integral for spin. Previously, we wrote |ry with S ě r ě ´S for spin-states.
Now we see that by this was meant
|ry ” |j, my with j “ S, m “ r. (7.2.19)
Analysis reveals mathematically that spin is possible with these quantum numbers but
still need to give (very brief) indication of experimental verification.
The set of states t|j, myu for fixed j is often called an angular momentum multiplet
or representation. From the analysis above (Eqns. (7.2.10, 7.2.11)) it is clear we can
choose normalized states |j, my with
a
J` |j, my “ ~ pj ´ mqpj ` m ` 1q|j, m ` 1y , (7.2.20)
a
J´ |j, my “ ~ pj ` mqpj ´ m ` 1q|j, m ´ 1y , (7.2.21)
which are key relations between the states. The whole multiplet can be defined by
‚ taking the top state, the one with maximum J3 eigenvalue: J` |j, jy “ 0,
‚ and applying Jń :
|j, j ´ ny “ Cjn Jń |j, jy , (7.2.22)
where Cjn is a constant computable using Eq. (7.2.21).
‚ Alternatively, can start with the bottom state, |j, ´jy and determine the others by
applying J`n .
Note that the choice of J3 as the member of the commuting set along with J 2 is a
convention. We could have chosen n¨J instead. n¨J has the same possible eigenvalues
as J3 but the eigenstates are linear combinations of the basis states t|j, myu. This will
be relevant when we discuss rotations.
7.3 Matrix representations

Recall from section 2.4 that given an orthonormal basis t|nyu we can regard states
as column vectors |ψy ù αn “ xn|ψy, and operators as matrices A ù Amn “
xm|A|ny with the action of operators on states to be that of matrix multiplication.
This is particularly useful when the space of states is finite-dimensional, as with angular
momentum states t|j, myu with fixed j.
For example, with j “ 1 (3-dim state space)
¨ ˛ ¨ ˛ ¨ ˛
1 0 0
|1, 1y ù ˝0‚ , |1, 0y ù ˝1‚ , |1, ´1y ù ˝0‚ (7.3.1)
0 0 1
¨ ˛ ¨ ˛ ¨ ˛
1 0 0 ? 0 1 0 ? 0 0 0
J3 ù ~ ˝0 0 0 ‚ , J` ù ~ 2 ˝0 0 1‚ , J´ ù ~ 2 ˝1 0 0‚
0 0 ´1 0 0 0 0 1 0
(7.3.2)
These follow from the formulas
?
J3 |1, 1y “ ~|1, 1y J˘ |1, ¯1y “ ~ 2|1, 0y
? (7.3.3)
J3 |1, ´1y “ ´~|1, ´1y J˘ |1, 0y “ ~ 2|1, ˘1y
as the only non-zero results.

This is used very widely for the case of spin- 12 (now use S rather than J ) with just two
states.
ˆ ˙ ˆ ˙
1 1 1 1 1 0
|2, 2y ù , |2, ´2y ù
0 1 (7.3.4)
previously: | Òy | Óy
ˆ ˙ ˆ ˙ ˆ ˙
1 0 0 1 0 0
S3 ù 12 ~ , S` ù ~ , S´ ù ~ . (7.3.5)
0 ´1 0 0 1 0
Moreover, we also write Si ù 12 ~σi where
ˆ ˙ ˆ ˙ ˆ ˙
0 1 0 í 1 0
σ1 ù , σ2 ù , σ3 ù , (7.3.6)
1 0 i 0 0 ´1
are called the Pauli matrices. Other combinations are
ˆ ˙ ˆ ˙
0 2 0 0
σ` “ σ1 ` iσ2 “ , σ´ “ . (7.3.7)
0 0 2 0
The Pauli matrices are hermitian, traceless matrices obeying
σ12 “ σ22 “ σ32 “ I , (7.3.8)
σ1 σ2 “ ´σ2 σ1 “ iσ3 and cyclic. (7.3.9)
These properties are conveniently summarized by
σi σj “ δij ` iεijk σk . (7.3.10)
Note that the antisymmetric part (in i, j) of this equation is
rσi , σj s “ 2iεijk σk , (7.3.11)
and corresponds to the fundamental commutation relation
rSi , Sj s “ i~εijk Sk , (7.3.12)
but the remaining, symmetric part, is special to spin- 21 .
The Pauli matrices are components of a vector
1
S ù 2
~σ , σ “ pσ1 , σ2 , σ3 q . (7.3.13)
If a and b are constant vectors (or at most operators which commute with S) then we
can contract ai and bj with both sides of Eqn. (7.3.10) to obtain
pa ¨ σqpb ¨ σq “ pa ¨ bqI ` ipa ^ bq ¨ σ . (7.3.14)
As a special case
pn ¨ σq2 “ I , (7.3.15)
2 1 2
where n is any unit vector. Note that this is equivalent to pn ¨ Sq “ which
4
~ I,
agrees with the fact that the eigenstates of n ¨ S are ˘ 21 ~; these are the only possible
results for measurement of spin along some direction n.
One last example of matrix properties corresponding to known properties of operators:
σ 2 “ σ12 ` σ22 ` σ32 “ 3 I , (7.3.16)
to be compared with
`1˘ `1 ˘ ~2
S 2 “ ~2 2 2
`1 I “ 3 I (7.3.17)
4
~2
which has eigenvalue 3 on any state.
4
7.4 Some physical aspects of angular momentum and spin

Our analysis of angular momentum states in section 7.2 has revealed the mathematical
possibility of half-integral spin. As asserted earlier, this is realized in nature as previ-
ously stated that each particle carries an internal space of states |ry with ´S ď r ď S.
Now identify these with |j “ S, m “ ry angular momentum states. The spin operators
obey
rSi , x̂j s “ rSi , p̂j s “ 0 , (7.4.1)
which should be contrasted with the commutators for L given in section 7.1. This is
consistent with our earlier use of basis states |xy b |ry (fixed S) because x̂, S 2 , S3 are
a commuting set.
In general, then, a given particle will have both orbital angular momentum L and spin
angular momentum S, giving a total angular momentum
J “ L`S . (7.4.2)
Note that rSi , Lj s “ 0 and hence
rLi , Lj s “ i~εijk Lk rSi , Sj s “ i~εijk Sk (7.4.3)
together imply
rJi , Jj s “ i~εijk Jk . (7.4.4)
How do we know the world works this way? Results of many experiments confirm it.
Here just mention the theoretical ideas underlying a few of them. The key idea is how
spin and angular momentum degrees of freedom enter into the Hamiltonian.
Main example for us is interaction with a background magnetic field. From classical
EM (no previous knowledge required) a moving distribution of electric charge interacts
with magnetic field Bpxq to produce energy
´µ ¨ B , (7.4.5)
where µ is the magnetic dipole moment of the distribution. Think of µ as being a

small magnet which does not disturb B. For instance, a single charge q in an orbit
with angular momentum L can be shown to have
q
µ “ L. (7.4.6)
2m
Hence, the interaction energy is of the form
´γ L ¨ B , γ some constant. (7.4.7)
This final result is all that is important for us. Now pass to quantum theory and
consider the effect of including such a term in the Hamiltonian for the electron in the
hydrogen atom (works for any spherically symmetric potential).
(i) Before we turn on any magnetic field have ψnlm pxq joint eigenstates of H, L2 , L3 with
degenerate energies E independent of m (as discussed in section 7.1).
Add a weak magnetic field along the 3-direction, B “ p0, 0, Bq and so
H Ñ H ´ γBL3 , (7.4.8)
but ψnlm pxq are still eigenstates and energies now split
E Ñ E ´ γB~m ´l ďmďl .
looooomooooon (7.4.9)
2l ` 1 levels
This is observed: spectral lines split into distinct lines. This is the Zeeman effect.
(ii) Replacing L by S we have an interaction energy of the form
´γ 1 S ¨ B , γ 1 constant ‰ γ in general. (7.4.10)
For a spin- 21 particle originally in eigenstate of a spin-independent Hamiltonian,

adding this term splits the energy states by
~
˘γ 1 B B weak and constant. (7.4.11)
2
Also observed and also called the Zeeman effect. Get appearance of two distinct
levels – direct experimental evidence for spin- 12 .
(iii) An electron in the atom has both an orbital magnetic moment µL 9L and a spin mag-
netic moment µS 9S. Like two bar magnets near each other there is an interaction
energy
µL ¨ µS 9 L ¨ S . (7.4.12)
This is called spin-orbit coupling and when included in the atomic Hamiltonian it
leads to splitting of spectral lines into doublets because of the spin- 21 of the electron.
The two yellow sodium D lines are a famous outcome. This is called atomic fine
structure.
(iv) Unlike orbital angular momentum, it is sometimes appropriate to consider spin to-
tally divorced from any space degrees of freedom. E.g., electron somehow confined
to one atomic site in a crystal. Then
H “ ´ γS ¨ B (7.4.13)
can be the complete Hamiltonian. If allow Bpzq “ p0, 0, Bpzqq to vary in space
rather than just separating energies can instead physically separate atoms according
to whether their spin states are Ò or Ó. This is due to
d dB ~ dB
force in z-dirn “ ´ p´γS ¨ Bpzqq “ ´ γS3 “ ¯γ (7.4.14)
dz dz 2 dz
This is the Stern-Gerlach experiment.
(Source: en.wikipedia.org (Theresa Knott))
(v) The proton is spin- 21 and has a magnetic moment. Even if l “ 0 for the electron in
the H atom and no external field there is a weak interaction between the proton spin
I and electron spin S 9I ¨ S. There are two narrowly spaced levels and transitions
give radiation with wavelength « 21cm which is observed from interstellar hydrogen.
(vi) In water the proton magnetic moments of the hydrogen nuclei interact with an
external B-field, 9 ´ I ¨ B to produce two levels where the transition absorbs or
radiates microwaves. This forms the basis of MRI scanners.
Essential idea is that spin or angular momentum behaves like a magnet/dipole. Actu-
ally, an iron bar magnet is magnetized because of the sum of the microscopic electron
spin magnetic dipoles in the atoms which, below the Curie temperature (TC ), prefer to
align and produce the macroscopic magnetic dipole. (For T ą TC the magnetization
vanishes; at T “ TC there is a second-order phase transition.)
The existence of an atomic magnetic moment in iron and also the preference for them
to align and produce a macroscopic magnetization are due to the Pauli Exclusion
Principle (spin-statistics theorem) – a quantum effect. Look up Hund’s rule.
7.5 Addition of angular momentum

Set ~ “ 1 in this section. Standard relations are
J 2 |j, my “ jpj ` 1q|j, my

J3 |j, my “ m|j, my
a
J˘ |j, my “ pj ¯ mqpj ˘ m ` 1q|j, m ˘ 1y . (7.5.1)
Consider two independent systems with angular momentum operators J p1q , J p2q acting
on spaces of states V p1q , V p2q each with standard basis t|ji , mi yu consisting of joint
piq
eigenstates of pJ piq q2 , J3 for i “ 1, 2.
Construct space of states for the combined system as V “ V p1q b V p2q with basis
(shorthand)
|m1 ; m2 y ” |j1 , m1 y b |j2 , m2 y ” |m1 y|m2 y j1 , j2 fixed. (7.5.2)

Sum of angular momentum operators is the total angular momentum operator

J “ J p1q b I ` I b J p2q
(7.5.3)
or simply “ J p1q ` J p2q .
Recall ji ě mi ě ´ji for i “ 1, 2 and so have total of p2j1 ` 1qp2j2 ` 1q basis states for
V.
Our aim is to find a new basis for V which consists of joint eigenstates |J, M y of J 2 and
J3 . We are seeking a change of basis for V from states with definite m1 , m2 to states
with definite J, M p“ m1 ` m2 q. Of course, either basis will do since they are both
span V , and it is usually the physics application that determines which basis is the
most appropriate to use. These bases correspond to classification by the alternative
complete commuting sets:
p1q p2q
pJ p1q 2
q , J3 , pJ p2q q2 , J3
loooooooooooooomoooooooooooooon or J 2 , J3 , pJ p1q q2 , pJ p2q q2 .
loooooooooooomoooooooooooon (7.5.4)
|j1 , m1 , j2 , m2 y |J, M, j1 , j2 y
” |j1 , m1 y|j2 , m2 y
We now suppress the j1 , j2 labels as they are common to all states, and use the compact
notation |ji , mi y Ñ |mi y.
Since
p1q p2q
J3 “ J3 ` J3 , (7.5.5)
we have
¯ ¯
` p1q ` p2q
J3 |m1 ; m2 y “ J3 |m1 y |m2 y ` |m1 y J3 |m2 y
“ M |m1 ; m2 y with M “ m1 ` m2 . (7.5.6)
So the product states are eigenstates of J3 already and its eigenvalues have range
j1 ` j2 ě M ě ´pj1 ` j2 q . (7.5.7)
Note that this means that J ď j1 ` j2 since the maximum value of M available must
be the largest value J allowed. We shall see how this works now.
It is less obvious how to find eigenstates of J 2 . Since t|m1 ; m2 yu is a basis, we must
have ÿ
J
|J, M y “ Cm 1 m2
|m1 ; m2 y . (7.5.8)
m1 `m2 “M
(Remember j1 , j2 fixed.) We want to find
(i) which values of J are allowed;

(ii) what are the Clebsch-Gordan (CG) coefficients for addition of j1 , j2 :
J
Cm 1 m2
“ xm1 ; m2 |J, M y . (7.5.9)
The key idea is to find the top state for a given total angular momentum with
J` |J, M y “ 0 ô J “ M . (7.5.10)
Given M , which is easy to determine, we therefore know J for this state. Then the
others with the same J are found by applying J´ .
‚ To begin the process consider the largest value of M : M “ j1 ` j2 . There is a unique

state |j1 y|j2 y (in notation |m1 y|m2 y) with this eigenvalue, and so it must be a top
state (can check by applying J` ). Hence,
J M
| j1 ` j2 , j1 ` j2 y “ |j1 y|j2 y , (7.5.11)
This is called the highest weight state since it is the state with the largest value
of M possible for given J. (This terminology arises in the group theory approach.)
The phase of the state on RHS (P V p1q b V p2q ) is chosen by convention to be `1.
p1q p2q
‚ Apply J´ “ J´ ` J´ to this state on LHS and RHS, respectively. We get
p1q p2q
J´ J´ ` J´
a ? ?
2pj1 ` j2 q |j1 ` j2 , j1 ` j2 ´ 1y “ 2j1 |j1 ´ 1y|j2 y ` 2j2 |j1 y|j2 ´ 1y
(7.5.12)
using standard formulas for V on LHS and for V p1q , V p2q on RHS. The normalized
state is then
d d
j1 j2
|j1 ` j2 , j1 ` j2 ´ 1y “ |j1 ´ 1y|j2 y ` |j1 y|j2 ´ 1y . (7.5.13)
j1 ` j2 j1 ` j2
‚ Continuing to apply J´ generates an entire multiplet of states |j1 ` j2 , M y with

j1 ` j2 ě M ě ´pj1 ` j2 q. From these expressions we read off the CG coefficients.
E.g.,
d d
j1 `j2 j1 `j2 j1 j1 `j2 j2
Cj1 ,j2 “ 1 , Cj1 ´1,j2 “ , Cj1 ,j2 ´1 “ , . . . (7.5.14)
j1 ` j2 j1 ` j2
‚ But we have not yet found all |J, M y states. We started from a unique state with
M “ j1 ` j2 . At the next level down, with M “ j1 ` j2 ´ 1, there are two states:
|j1 ´ 1y|j2 y and |j1 y|j2 ´ 1y . (7.5.15)
One combination is identified above as |j1 ` j2 , j1 ` j2 ´ 1y. The other, orthogonal,

combination (unique up to a phase) must be a new top state
d d
J M j1 j2
| j1 ` j2 ´ 1, j1 ` j2 ´ 1y “ |j1 y|j2 ´1y ´ |j1 ´1y|j2 y . (7.5.16)
j1 ` j2 j1 ` j2
The overall phase on the RHS is chosen by convention.
‚ Now proceed as before:
(i) generate the multiplet with J “ j1 ` j2 ´ 1 by applying J´ ;

(ii) look for a possible new top state at level M “ j1 ` j2 ´ 2;
(iii) repeat until all states exhausted.
The general pattern can be depicted as follows.

J j1+j2 j1+j2-1 j1+j2-2 |j1-j2|

M
j1+j2
j1+j2-1
j1+j2-2
-(j1+j2)+1
-(j1+j2)
where
e top states Ó apply J´
unique state K others (7.5.17)
u other states Ñ
with same M value
Can check that top states |ψy with J3 eigenvalue M are annihilated by J` directly.
This is also guaranteed by
|ψy K J´ |J, M ` 1y @ J ą M by our construction. (7.5.18)
Thus,
xJ, M ` 1|J` |ψy “ 0 @ J ą M ô J` |ψy “ 0 . (7.5.19)
The whole process stops with J “ |j1 ´ j2 | by counting. The number of states in the
alternative bases must be the same.
j1ÿ
`j2
p2j ` 1q “ p2j1 ` 1qp2j2 ` 1q . (7.5.20)
j“|j1 ´j2 |
We now have the answers to our original questions:
(i) j1 ` j2 ě J ě |j1 ´ j2 | in integer steps. E.g.,

j1 “ 21 , j2 “ 12 ñ J“1, 0
j1 “1, j2 “ 12 ñ J“ 32 , 12 (7.5.21)
j1 “1, j2 “1 ñ J“2, 1, 0
The range of J values matches bounds for addition of classical vectors J “ J p1q `J p2q
with lengths of J, j1 , j2 .
(ii) Clebsch-Gordan coefficients are found by explicit calculation of states for given j1
and j2 .
Simplest example is for

1
j1 “ j2 “ 2
, |m1 y|m2 y mi “ ˘ 12 , 4 states. (7.5.22)
|1, 1y “ | 12 y| 12 y top state with M “ 1

p1q p2q
J´ “ J´ ` J´ gives
?
2|1, 0y “ |´ 21 y| 21 y ` | 21 y|´ 12 y ñ
¯
|1, 0y “ ?1 |´ 12 y| 21 y ` | 12 y|´ 21 y
`
2
?
2|1, ´1y “ ?1 |´ 12 y|´ 21 y ` |´ 12 y|´ 21 y
` ˘
ñ
2
|1, ´1y “ |´ 12 y|´ 21 y completes J “ 1 multiplet
(7.5.23)
Then
1 ` ¯
|0, 0y “ ? |´ 12 y| 12 y ´ | 12 y|´ 12 y (7.5.24)
2
with M “ 0 and K |1, 0y. Need a phase convention to decide overall sign of top states.
J 1 0
M Compare this results with the combination
1 of spin states in Eqns. (5.5.14) and (5.5.15)
with | 21 y “ | Òy, |´ 12 y “ | Óy which were
0 found previously by demanding definite sym-
metry; they are precisely the same.
-1
Have
3 (triplet) J “ 1 states symmetric
(7.5.25)
1 (singlet) J “ 0 state antisymmetric
Indeed, the permutation group and its representations are often central to constructing
multiplets in this way.
Tables for CG coefficients can be found in the Particle Data Group (PDG) Tables
( http://pdg.lbl.gov/2014/reviews/rpp2014-rev-clebsch-gordan-coefs.pdf). The PDG
collate all reviews and tables for properties of elementary particles.
The table for 2 b 1 giving states with M “ 1 is
8 TRANSFORMATIONS AND SYMMETRIES 57
J 3 2 1
m1 m2 M +1 +1 +1
2 -1 1/15 1/3 3/5
1 0 8/15 1/6 -3/10
0 1 2/5 -1/2 1/10
? a
Take but keep the sign. E.g., ´3{10 Ñ ´ 3{10.
8 Transformations and Symmetries

8.1 Introduction and an Example
Given a unitary operator U define a transformation of a quantum system to be
either a map of states only:
|ψy ÞÑ U |ψy xφ| ÞÑ xφ|U : , (8.1.1)
or a map of operators only:

A ÞÑ U : AU . (8.1.2)
These are equivalent. Either way, inner products of states xφ|ψy left unchanged but
matrix elements change:
xφ|A|ψy ÞÑ xφ|U : AU |ψy . (8.1.3)
Unlike the change of picture we change states or operators but not both. Such a
transformation is a symmetry of the quantum system if
U : HU “ H or, equivalently, rU, Hs “ 0 . (8.1.4)
It then follows that
‚ in S-picture if |ψptqy is a solution of the Schrödinger equation then so is U |ψptqy;
‚ in H-picture if Aptq is a solution of the Heisenberg EoM then so is U : AptqU .
Now consider a group G and transformations of a QM system U pgq for each g P G with
U pg1 qU pg2 q “ U pg1 g2 q

U p1G q “ I
U pg ´1 q “ U pgq´1 “ U pgq: . (8.1.5)
In any given case U pgq is a representation of G. Our aim is to find the unitary
operators U pgq when G is a group of translations, rotations or reflections. In these
cases we know how G acts geometrically
x ÞÑ gpxq , (8.1.6)
and similarly for momentum p. We infer the action on position eigenstates
U pgq|xy “ |gpxqy (8.1.7)

and on operators
U pgq: x̂ U pgq “ gpx̂q . (8.1.8)
These statements are equivalent. E.g., assuming the second one
´ ¯
x̂ U pgq|xy “ U pgqgpx̂q|xy “ gpxq U pgq|xy , (8.1.9)
and so confirm the first.

On a general state
|ψy ÞÑ U pgq|ψy (8.1.10)
and the position space wavefunction of the new state is
´ ¯:
xx|U pgq|ψy “ U pgq: |xy |ψy “ xg ´1 pxq|ψy . (8.1.11)
So the effect of the transformation on the wavefunction is the change
ψpxq ÞÑ ψpg ´1 pxqq . (8.1.12)
Simplest example is translation in one dimension
x ÞÑ ga pxq “ x ` a (8.1.13)
and write U paq ” U pga q to simplify notation. Then must have
U paq|xy “ |x ` ay
:
U paq x̂ U paq “ x̂ ` a , (8.1.14)
but we showed in section 2 that
U paq “ eíap̂{~ (8.1.15)
has precisely these properties. The effect of translation by a on ψpxq should be ψpxáq
which is illustrated by the picture below.
R(x) R(x-a)
x x
0 a 0 a
Confirmation: on wavefunctions
ˆ ˙
B B
p̂ Ñ ´ i~ , U paq Ñ exp á (8.1.16)
Bx Bx
and so
ˆ ˙
B
U paqψpxq “ exp á ψpxq
Bx
1 2 2
“ ψpxq ´ a ψ 1 pxq ` a ψ pxq ´ . . .
2!
“ ψpx ´ aq Taylor’s theorem, (8.1.17)
as expected. Because of its special role in U paq the momentum p̂ is called the generator
of translations.
8.2 Generators and conservation laws

The action of many transformations such as translations, already discussed above, and
rotations can be formulated in terms of groups. In the case of rotations the group
is SOp3q, the group of orthogonal 3 ˆ 3 real matrices of unit determinant. In many
cases the relevant group is non-abelian meaning that successive group operations do
not commute; SOp3q is such an example. If Rpθ, nq P SOp3q is a rotation by θ about
the axis n then in general
Rpθ, nqRpφ, mq ‰ Rpφ, mqRpθ, nq . (8.2.1)
However, for fixed n the elements Rpθ, nq do commute and form a one-parameter,
abelian, subgroup:
Rpθ, nqRpφ, nq “ Rpφ, nqRpθ, nq “ Rpθ ` φ, nq . (8.2.2)
The translation group is another example. It is with general ideas about such abelian
subgroups that we begin.
Consider some continuous group G, and let gα (α a real number) be a one parameter
family of group elements with
gα gβ “ gβ gα “ gα`β , g0 “ 1G , (8.2.3)
and assume we know
gα pAq “ A ` αf pAq ` Opα2 q (8.2.4)
for quantities of interest A (positions, momenta, etc.). Write the corresponding unitary
operators as
U pαq “ U pgα q to simplify notation, (8.2.5)
and then have
U pαq: A U pαq “ gα pAq . (8.2.6)
Define ˇ
BU pαq ˇˇ
Q “ i~ (8.2.7)
Bα ˇα“0
so that
i
U pδαq “ I ´ δα Q ` Opδα2 q
~
i
U p´δαq “ I ` δα Q ` Opδα2 q . (8.2.8)
~
Then
U p´δαq “ U pδαq´1 “ U pδαq: ñ Q “ Q: : Q is an observable. (8.2.9)
Also have
U pα ` δαq “ U pαqU pδαq “ U pδαqU pαq , (8.2.10)
so deduce
rQ, U pαqs “ 0 . (8.2.11)
From above have
BU pαq i ´ α ¯
“ ´ Q U pαq ñ U pαq “ exp í Q given U p0q “ 1. (8.2.12)
Bα ~ ~
Q is the generator of this family of transformations, the one-parameter subgroup,
within G. To find Q we equate
gα pAq “ A ` αf pAq ` Opα2 q (8.2.13)
with
ˆ ˙ ˆ ˙
: i i
U pαq A U pαq “ I ` αQ ` ... A I ´ αQ ` ...
~ ~
i
“ A ` αrQ, As ` Opα2 q (8.2.14)
~
to obtain
rQ, As “ í~f pAq . (8.2.15)
Knowing the RHS for any A determines Q. Conversely, the behaviour of any quantity
under the transformation is fixed by its commutation relation with the generator.
Now suppose the continuous family gα corresponds to a symmetry of the quantum
system
rU pαq, Hs “ 0 ô rQ, Hs “ 0 . (8.2.16)
The second equation says that the observable Q is a conserved quantity:
‚ in S-picture can choose joint eigenstates of H and Q
Q|ψp0qy “ q|ψp0qy ñ Q|ψptqy “ q|ψptqy . (8.2.17)
‚ In H-picture expect Qptq in general but here

dQ i
“ ´ rQ, Hs “ 0. (8.2.18)
dt ~
Noether’s Theorem: for each continuous transformation of a quantum system there

is an hermitian generator, and if the transformation is a symmetry this generator is
conserved; it is called the Noether charge.
Example: system of particles labelled by r, s, . . . with positions x̂prq and momenta p̂prq
which obey
prq psq
rx̂i , p̂j s “ i~ δij δ rs . (8.2.19)
The transformation is translation along the k-axis.

prq prq prq prq
ga px̂i q “ x̂i ` aδik , ga pp̂i q “ p̂i ñ
prq prq (8.2.20)
f px̂i q “ δik , f pp̂i q “ 0
for all r. Then the generators must obey

prq prq
rQ, x̂i s “ ´ i~δik , rQ, p̂i s “ 0 (8.2.21)
ÿ prq
The solution is Q “ P̂k “ p̂k , the total momentum in the k-direction.
r
In general, translations by a are generated by the total momentum P̂ with correspond-

ing unitary operator ˆ ˙
i
U paq “ exp ´ a ¨ P̂ . (8.2.22)
~
Noether’s Theorem: for particles interacting through potential depending only on
prq psq
x̂i ´ x̂i , H is translation invariant and total momentum P̂ is conserved.
r Note that there are conserved quantities in classical dynamics such as momentum for
translation-invariant systems and angular momentum for rotationally-invariant sys-
tems. u
8.3 Angular Momentum and Rotations

As stated earlier, rotations form a non-abelian group: rg1 , g2 s ‰ 0, in general. For
fixed axis n do have an abelian subgroup of rotations gθ through angle θ so can use
approach above.
For A a scalar
gθ pAq “ A unchanged ñ f pAq “ 0 . (8.3.1)
For v “ pv1 , v2 , v3 q a vector,
gθ pvq “ v ` θn ^ v ` Opθ2 q , ñ f pvq “ n ^ v . (8.3.2)
Write the corresponding unitary operator as

i
U pθnq “ I ´ θQ ` Opθ2 q . (8.3.3)
~
We need to determine Q. For a single particle both x̂ and p̂ must transform as vectors:
rQ, x̂s “ í~ n ^ x̂

rQ, p̂s “ í~ n ^ p̂ . (8.3.4)
Recall the commutators with orbital angular momentum
rLi , x̂j s “ i~ εijk x̂k

rLi , p̂j s “ i~ εijk p̂k (8.3.5)
These imply that we identify

Q“n¨L (8.3.6)
since it produces the correct commutators. E.g.,
rni Li , x̂j s “ ni i~ εijk x̂k

“ í~ pn ^ x̂qj . (8.3.7)
For a spinless particle this is what we seek. Note that L transforms as a vector:
rQ, Ls “ í~ n ^ L . (8.3.8)
For a particle with spin the generator above needs to be modified to ensure that S
also transforms as a vector
rQ, Ss “ í~ n ^ S , (8.3.9)
and this is achieved by taking
Q“n¨J with J “ L ` S the total angular momentum. (8.3.10)
(Recall that rSi , x̂j s “ rSi , p̂j s “ rSi , Lj s “ 0.)

This analysis can easily be extended to many particles. For a general quantum system,
the unitary operator for rotation through θ “ θn is
ˆ ˙
i
U pθq “ exp ´ θ ¨ J , (8.3.11)
~
where J is the total angular momentum.

Noether’s Theorem: for a rotationally invariant quantum system the total angular
momentum J is conserved. E.g., particle with Hamiltonian
1 2
H“ p̂ ` V p|x̂|q ` U p|x̂|q L ¨ S . (8.3.12)
2m
Note the spin-orbit coupling term in L ¨ S. Clearly invariant under rotations so
rJi , Hs “ 0 or J conserved. (8.3.13)
Note that L and S are not separately conserved unless U “ 0. However, L2 and S 2
are conserved as is J 2 . So we can write (J “ L ` S)
J 2 ´ L2 ´ S 2 1
L¨S “ “ ~2 pjpj ` 1q ´ lpl ` 1q ´ sps ` 1qq (8.3.14)
2 2
for eigenstates of J 2 , L2 , S 2 with eigenvalues pj, l, sq, respectively.
We now have alternative definitions of a scalar operator A and a vector operator v
in terms of commutation relations:
scalar: rJi , As “ 0
vector: rJi , vj s “ i~ εijk vk . (8.3.15)
Have x̂, p̂, L, S, J are vector operators.

Now look at action of rotations on states.
‚ On angular momentum states |j, my, fixed j, the angular momentum generators Ji
can be represented as p2j ` 1q ˆ p2j ` 1q matrices (see section 7.3).
‚ Then ˆ ˙
i
U pθq “ exp ´ θn ¨ J (8.3.16)
~
can be computed easily in some cases.
‚ j “ 0: single state |0, 0y with J |0, 0y “ 0. Hence
U pθq “ I ñ |0, 0y ÞÑ |0, 0y scalar (8.3.17)
‚ j “ 1: states |1, my with m “ 0, ˘1. Generators are 3 ˆ 3 matrices. E.g.,

¨ ˛
1 0 0
J3 “ ~ ˝0 0 0 ‚ (8.3.18)
0 0 ´1
Consider for simplicity rotations about the 3 (or z) axis – J3 is diagonal in our
representation. Then
» ¨ ˛fi ¨ íθ ˛
1 0 0 e 0 0
i
U pθk̂q “ expp´ θJ3 q “ exp íθ 0 0 0
– ˝ ‚ fl “ ˝ 0 1 0‚. (8.3.19)
~
0 0 ´1 0 0 eiθ
So under this rotation states behave as
|1, ˘1y ÑÞ e¯iθ |1, ˘1y

|1, 0y Ñ Þ |1, 0y . (8.3.20)
Deduce that the states behave as the (spherical) components of a vector. More
familiar if we compare with same rotation on coordinates in spherical basis (bear in
mind for later):
px ˘ iyq ÞÑ pxθ ˘ iyθ q “ e˘iθ px ˘ iyq , z ÞÑ z . (8.3.21)
(Remember, wavefunction transforms with ´θ: ψpxq ÞÑ ψpx´θ q.)

‚ j “ 21 : states | 21 , ˘ 12 y (or | Òy and | Óy) on which
~
J“ σ. (8.3.22)
2
The property that pn ¨ σq2 “ I2 allows explicit computation of rotation operator for
any θ.
ˆ ˙ ˆ ˙
i iθ
U pθq “ exp ´ θn ¨ J “ exp ´ n ¨ σ
~ 2
8
ÿ 1 ˆ ˙ p
iθ
“ ´ pn ¨ σqp
p“0
p! 2
θ θ
“ cos ´ i n ¨ σ sin . (8.3.23)
2 2
The effect of a rotation on a general state is

ˆ ˙
θ θ
|χy ÞÑ U pθq|χy “ cos ´ i n ¨ σ sin |χy . (8.3.24)
2 2
But note the appearance of θ{2 in final result. Under any rotation by 2π we have
|χy ÞÑ U p2πnq|χy “ ´ |χy . (8.3.25)
We do not get exactly the same state vector after rotation by 2π, which may seem
surprising. However, |χy and ´|χy have same physical content. Change in sign can
be a real effect if e.g. have two particles and rotate just one; change in sign can make
the difference between constructive and destructive interference.
Under rotation by 4π get exactly the same state
|χy ÞÑ U p4πnq|χy “ |χy . (8.3.26)
Similar behaviour under 2π and 4π rotations for any states with half-integral spin.
8.4 Reflections and Parity

Simplest example of a discrete transformation is parity or spatial inversion
P : x ÞÑ ´ x . (8.4.1)
This is not part of a continuous family of isometries connected to the identity (un-
like e.g., rotations). Any reflection in a plane can be achieved by composing P with
rotations.
The corresponding unitary operator acts on position eigenstates as
U |xy “ | ´ xy
so U2 “ 1
ñ U : “ U ´1 “ U . (8.4.2)
There is no one-parameter family and so no generator, but U itself is hermitian and

thus an observable:
U |ψy “ ηψ |ψy (8.4.3)
and
U 2 “ I ñ ηψ “ ˘ 1 , the parity of the state. (8.4.4)
The corresponding action on operators is
U x̂ U “ ´x̂
U p̂ U “ ´p̂
U LU “ L pL “ x̂ ^ p̂q
USU “ S
UJU “ J. (8.4.5)
Parity is a symmetry of a quantum system if
UHU “ H, (8.4.6)
and then we can choose joint eigenstates of H and U . Recall in 1D

1 2
H“ p̂ ` V px̂q (8.4.7)
2m
and if V px̂q “ V p´x̂q can choose energy eigenstates to have definite parity: even or
odd functions. In 3D with
1 2
H“ p̂ ` V p|x̂|q (8.4.8)
2m
have definite orbital angular momentum states ψlm pxq obeying
ψlm p´xq “ p´1ql ψlm pxq

ñ ηψ “ p´1ql , (8.4.9)
which are thus states of definite parity. Note that for two-particle system P inter-
changes the spatial positions and so its eigenvalue determines the spatial symmetry
of the state: symmetric ηψ “ 1, anti-symmetric ηψ “ ´1.
There is also a notion of intrinsic parity where a particle state changes as
U |ay “ ηa |ay , ηa “ ˘1 . (8.4.10)
As with spin this is not attributable to any internal structure; it is a property of the
particle concerned. Values for some particles can be chosen by convention but then
others are determined. In general, for a system of particles,
" * " * " *
total spatial product of all
“ ˆ (8.4.11)
parity parity intrinsic parities
E.g., particles a and b with relative position x̂
x
a
in angular momentum state ψlm pxq. The total parity is
p´1ql ηa ηb . (8.4.12)
Note that parity gives rise to a multiplicative conservation law in contrast to the
additive conservation of momentum or angular momentum.
Parity is (famously) not a symmetry of all fundamental interactions; it is conserved

for electromagnetic and strong nuclear processes but not for weak nuclear processes
responsible for effects like β-decay
n Ñ p e ν̄
(8.4.13)
neutron proton, electron, anti-neutrino
If P is a symmetry we cannot tell if we are watching a physical process directly or
viewing it in a mirror. Weak processes involve the neutrino, a particle of spin- 12 which
is effectively massless and so travels close to the speed of light. A major property is
that it has -ve helicity which means that
S¨p
“ ´ 12 . (8.4.14)
|p|
It is “left-handed”
MIRROR
p -p
left-handed neutrino: right-handed neutrino:

does exist does not exist
and there is no +ve helicity, “right-handed”, neutrino. The anti-neutrino is “right-

handed” with no “left-handed” version. But under a P transformation, illustrated by
the mirror, “left-handed” Ø “right-handed”:
U S ¨ pU “ ´ S ¨ p (8.4.15)
and so in the looking-glass world Alice only sees “right-handed” neutrinos which are
unphysical for us: parity is broken.
Other discrete symmetries (not considered in detail here):
T : time reversal
C: charge conjugation - interchange of particles and anti-particles
If we apply CP to the neutrino then the mirror image above now beomes a right-handed
anti-neutrino which does exist. So CP is a symmetry of weak nuclear processes
(well almost). The combination CP T is always a symmetry of relativistic quantum
mechanics (Quantum Field Theory, QFT).
Example of use of conservation laws and intrinsic parity:
πd Ñ nn (8.4.16)
π pion, d deuteron pp n bound stateq, n neutron.
1
(8.4.17)
Sπ “ 0 Sd “ 1 Sn “ 2
9 TIME-DEPENDENT PERTURBATION THEORY 67
‚ This is a strong interaction process – total angular momentum, J , and parity con-
served.
‚ π d system:
Orbital ang. mom. l “ 0 (given), total spin S “ 1

ñ total ang. mom. j “ 1. (8.4.18)
Note that l “ 0 is the dominant state for low energy scattering (See AQM course).
‚ n n system:
Orbital ang. mom. l unknown to begin with but symmetry of spatial wavefunction
p´1ql . Total spin S “ 0 or 1.
‹ S “ 0 ñ antisymmetric spin state, so Fermi statistics requires l even. But since

S “ 0 have that j “ l, and so process forbidden since j “ 1 from J conservation.
‹ S “ 1 ñ symmetric spin state, so Fermi statistics requires l odd. But j “
l ´ 1, l, l ` 1 and have j “ 1 by J conservation and so deduce
j “ l “ S “ 1. (8.4.19)
‹ Concerning parity. The deuteron is predominantly an S-wave bound state (l “ 0)

and so ηd “ p´1q0 ηn ηp ; but ηn “ ηp (they are from the same nucleon family) and
so ηd “ 1.
p´1q0 ηπ ηd “ p´1q1 ηn2

ηd “ 1 ñ ηπ “ ´1 . (8.4.20)
9 Time-dependent Perturbation Theory

Consider a system with Hamiltonian
H “ H0 ` V ptq , (9.1)
where H0 is time-independent with known eigenstates and eigenvalues and V ptq small in
some sense (remember perturbation theory earlier). I have suppressed the dependence
of V ptq on space coordinates.
‹ Our aim: to calculate the effect of V ptq order-by-order (actually stop at first order)
and, in particular, calculate probabilities for transitions between eigenstates of H0 as
functions of time. This is a dynamical question even if V is time-independent.
9.1 The interaction picture

Start in the Schrödinger picture. States obey
B
i~ |ψptqy “ H|ψptqy “ pH0 ` V ptqq|ψptqy . (9.1.1)
Bt
The interaction picture is defined by moving the ‘known’ part of the time evolution,
which is due to H0 , from states to operators:
|ψptqy “ eiH0 t{~ |ψptqy
Aptq “ eiH0 t{~ Aptq eíH0 t{~ . (9.1.2)

The remaining time dependence of the states is then given by
ˆ ˙
B iH0 t{~ B
i~ |ψptqy “ e ´H0 ` i~ |ψptqy
Bt Bt
“ eiH0 t{~ V ptq |ψptqy
“ eiH0 t{~ V ptq eíH0 t{~ eiH0 t{~ |ψptqy

B
or i~
|ψptqy “ V ptq |ψptqy . (9.1.3)
Bt
This can be rewritten as an integral equation
i t 1
ż
|ψptqy “ |ψp0qy ´ dt V pt1 q |ψpt1 qy , (9.1.4)
~ 0
which is well suited to solving for |ψptqy as a power series in V . For instance, we can
iterate this equation to deduce
˙2 ż t ż t1
i t 1
ż ˆ
i 1
|ψptqy “ |ψp0qy ´ dt V pt1 q |ψp0qy ` ´ dt dt2 V pt1 q V pt2 q |ψpt2 qy .
~ 0 ~ 0 0
(9.1.5)
The unknown state still appears on RHS but now in a term of order V 2 . We could
continue to any desired order in V (see AQM course) but will stop here:
i t 1
ż
|ψptqy “ |ψp0qy ´ dt V pt1 q |ψp0qy ` OpV 2 q . (9.1.6)
~ 0
Let |iy and |f y be eigenstates of H0 , eigenvalues Ei and Ef , and assume xf |iy “ 0. A
transition from the state
|ψp0qy “ |ψp0qy “ |iy (9.1.7)
at time t “ 0 (when S-picture and I-picture coincide) to a state |f y at time t occurs
with probability
|xf |ψptqy|2 “ |xf |ψptqy|2 (9.1.8)
since
xf |ψptqy “ xf | eíH0 t{~ |ψptqy “ eíEf t{~ xf |ψptqy . (9.1.9)
To order V , the amplitude we need is therefore
0 i t 1
ż
dt xf | V pt1 q |iy
*

xf |ψptqy “
xf
|iy ´
~ 0
i t 1
ż
1 1
“ ´ dt xf | eiH0 t {~ V pt1 q eíH0 t {~ |iy
~ 0
i t 1 ipEf Éi qt1 {~

ż
“ ´ dt e xf |V pt1 q|iy , (9.1.10)
~ 0
and probability for transition to order V 2 is

ˇ2
1 ˇˇ t iωt1
ˇż
1
ˇ
1ˇ
2
e xf |V pt q|iy dt ˇ , ~ω “ Ef ´ Ei . (9.1.11)
~ 0ˇ
An important special case is when V pt1 q is constant in time (at least for 0 ď t1 ď tq.
The amplitude above is then
i t iωt1
ż
1
´ e xf |V |iy dt1 “ p1 ´ eiωt q xf |V |iy , (9.1.12)
~ 0 ~ω
and the probability of transition to order V 2 is

ˆ ˙2
1 sin ωt{2
P ptq “ 2 |xf |V |iy|2 . (9.1.13)
~ ω{2
For fixed ω ‰ 0 have a periodic function of t.
9.2 Transition rates and Fermi’s Golden Rule

Probability found above for transition i Ñ f to second order in V (a constant perturb-
ing potential) can be written
t
P ptq “ ft pωq |xf |V |iy|2
~2
ˆ ˙2
1 sin ωt{2
ft pωq “ . (9.2.1)
t ω{2
t
4
tw2
ft(w)
2p 4p w
available states t t
Now
sin 2 x
ż8 ż8
lim ft pωq “ t , ft pωq dω “ 2 dx “ 2π . (9.2.2)
ωÑ0 ´8 ´8 x2
As t increases ft pωq is concentrated more and more around ω “ 0 and P ptq will be
non-negligible only when
2π 2π
´ ÀωÀ . (9.2.3)
t t
So for t large enough we have
ω«0 and Ef « Ei , (9.2.4)
So we recover conservation of energy. In many cases we deal with a transition to final

states with closely packed energy levels or even a continuum of energies such as in the
decay
excited atom ÝÑ atom + photon (9.2.5)
where the photon has momentum q and energy E “ |q|c.
Similarly, a free particle of mass m and momentum p has energy E “ p2 {2m; E is
a continuous variable and the states |py form a continuum. It is often useful to
think of these states in a box of large side L with periodic boundary conditions. The
wavefunction is then (V “ L3 )
1 2π 2π 2 ~2 2
ψp pxq “ ? eip¨x , p“ pn1 , n2 , n3 q ni P Z , E“ pn ` n22 ` n23 q .
V L mL2 1
(9.2.6)
This is a discrete set which becomes a continuum as L Ñ 8. In what follows shall
assume the gap between levels is always much less than 2π~{t ùñ always choose L
sufficiently large. See diagram which illustrates this idea.
Suppose the final states have certain parameters fixed (e.g. scattering angle wrt initial
state) but with Ef variable. Define the density of states, ρpEf q, by
ρpEf q∆E “ # states with specified parameters and energy in range pEf , Ef ` ∆Eq.
Consider a transition to a number of states temporarily represented as a discrete set

as described above.
ÿ t Ef ´ Ei
P ptq “ 2
ft pωf i q|xf |V |iy|2 , ωf i “ . (9.2.7)
f
~ ~
Now consider the limit as set of states |f y becomes a continuum (L Ñ 8) and t is

large enough that ρpEq is slowly varying over the range pEf ´ 2π~{t, Ef ` 2π~{t).
Probability of transition to some state in a band of energy F can be found from the
formula for P ptq by replacing
ÿ ż
Ñ dEf ρpEf q . (9.2.8)
f F
Note, we are assuming that |xf |V |iy|2 only depends on Ef and does not depend on
other unspecified quantum numbers which distinguish the different states with energy
Ef . This assumption is easily relaxed by summing over them, too.
For t sufficiently large
ft pωq « 2πδpωq “ 2π~ δpEf ´ Ei q , (9.2.9)

and then
ż
t
P ptq “ 2 dEf ρpEf q 2π~ δpEf ´ Ei q |xf |V |iy|2
~ F
$
’
& 0 if Ef is not in F
ˇ
“ 2πt ˇ
2 ˇ (9.2.10)
% ~ ρpEi q|xf |V |iy| ˇ
’ if it is
Ef “Ei
For this to be a good approximation we need:
‚ t not so large that 1st order perturbation theory breaks down: need P ptq ! 1.
‚ t large enough that the δ-function approximation is valid. This means changes in
ρpEf q|xf |V |iy|2 are small for changes in Ef of order ~{t (see diagram above).
Under these conditions have a constant transition rate:
P ptq 2π 2
t “ ~ ρpEi q |xf |V |iy| Fermi’s Golden Rule.
9.3 Interaction of radiation/EM field with atoms

The Hamiltonian for an electron in the hydrogen atom and interacting with an external
EM field is given by the minimal substitution rule (there can, of course, be other terms
which do appear when the particle is not ‘elementary’):
1 e e2 2
H“ pp̂ ´ eApx̂, tqq2 ` V pr̂q “ H0 ´ pp̂ ¨ A ` A ¨ p̂q ` A , (9.3.1)
2m 2m 2m
where Apx̂, tq is the vector potential of the EM field and H0 is the hydrogen atom
Hamiltonian. We assume that the EM field is weak and so can ignore the last term –
it is certainly suppressed by an extra factor of e, anyway. We assume that A has some
definite frequency ω0 and wave-vector q. Then we write (using position representation)
Apx, tq “ ε cos pq ¨ x ´ ω0 tq , (9.3.2)
where ε is the polarization vector which is transverse to the direction of propagation
so that ε ¨ q “ 0 (this ensures that the EM wave has two polarizations K q). In this
case p̂ ¨ A “ A ¨ p̂ since
í~∇ ¨ ε cos pq ¨ x ´ ω0 tq “ i~ q ¨ ε sin pq ¨ x ´ ω0 tq “ 0. (9.3.3)
The interesting interaction term is then
e
VI px̂, tq “ ´ p̂ ¨ ε cos pq ¨ x̂ ´ ω0 tq . (9.3.4)
m
There is no significance to cos rather than sin here. We are considering the decay of
an excited atom H ˚ Ñ H ` γ and the reverse, absorption, process and so the energy
of the photon is
1 ~
~ω0 “ |Ef ´ Ei | „ α2 mc2 with wavelength λ „ , (9.3.5)
α2 mc
where Ei and Ef are now the energies of initial and final atomic states, respectively.
Compare λ with the size of the atom given by the Bohr radius
1 ~
a0 „ . (9.3.6)
α mc
Have |q| “ 2π{λ and x is confined to the atomic interior so |x| À a0 . Thus,
a0 1
q¨x À „ α „ . (9.3.7)
λ 137
To lowest order we therefore neglect q ¨ x and consider the interaction to be
e
VI ptq “ ´ p̂ ¨ ε cos pω0 tq . (9.3.8)
m
We now note that
p̂
rx̂, H0 s “ i~ , (9.3.9)
m
and so
ie
VI ptq “cos pω0 tqrx̂ ¨ ε, H0 s (9.3.10)
~
Now the first-order transition amplitude i Ñ f is
i t 1 ipEf Éi qt1 {~
ż
´ dt e xf |VI pt1 q|iy
~ 0
e t 1 ipEf Éi qt1 {~

ż
“ 2 dt e cos pω0 t1 q xf |rx̂ ¨ ε, H0 sq|iy
~ 0
żt
e 1
“ 2 pEi ´ Ef q dt1 eipEf Éi qt {~ cos pω0 t1 q xf |x̂ ¨ ε|iy . (9.3.11)
~ 0
This is called the dipole approximation since the operator appearing in the matrix
element has the form ex̂ ¨ ε cos pω0 tq and may be interpreted as the interaction of the
electric dipole operator d̂ “ éx̂ with a classical electric field given by E 9 ε cos pω0 tq
of the form V ptq “ ´d̂ ¨ E. However, the derivation is from minimal substitution and
there are clearly corrections higher-order in q ¨ x. Can analyze formula in two parts.
(i) Time dependence. Separate cos into exponentials and integrate gives two terms
żt
1 1
dt1 eipEf Éi qt {~ e¯iω0 t . (9.3.12)
0
Evaluating and taking |.|2 gives contributions to probability

sin 2 ppω ¯ ω0 qt{2q pEf ´ Ei q
ω“ . (9.3.13)
pω ¯ ω0 q2 {4 ~
together with a somewhat messy cross-term. As t becomes sufficiently large these
terms are sharply peaked around ω « ˘ω0 , and the cross term is small in comparison.
Thus, for t large enough we have two kinds of transitions
iÑf with Ef “ Ei ˘ ~ω0 pω0 ą 0q (9.3.14)

which occur with equal probability. These processes are absorption and emission,
respectively. Proceeding as in derivation of the Golden Rule could find precise for-
mula for transition rate. This would require computing the density of states ρpEf q
(see Statistical Physics for example of this).
(ii) Matrix element (t independent):

ε ¨ xf |x̂|iy (9.3.15)
which must be non-zero for transition to occur. Probability for transition then
proportional to
|ε|2 9 intensity 9 energy density of classical EM radiation. (9.3.16)
Consider H-atom states
|iy “ |n l my , |f y “ |n1 l1 m1 y , (9.3.17)
labelled as usual by principal and ang. mom. quantum numbers. Use parity and
angular momentum properties of operators and states to determine when xf |x̂|iy ‰ 0.
‹ Parity.
Parity operator U satisfies U 2 “ 1 and have
U |n l my “ p´1ql |n l my and U x̂U “ ´ x̂
ñ xn1 l1 m1 |x̂i |n l my “ xn1 l1 m1 |U U x̂i U U |n l my

1
“ p´1ql`l `1 xn1 l1 m1 |x̂i |n l my (9.3.18)
and hence vanishes unless
l ` l1 odd first selection rule. (9.3.19)
‹ Angular momentum.
First show that x̂i |n l my are like product states for addition of ang. mom. 1 to l.
Intuitively, x̂i is a vector and we know that l “ 1 states also behave like a vector
under rotations. Now
rLi , x̂j s “ i~εijk x̂k means x̂ behaves as a vector operator, (9.3.20)
and define
1
X1 “ ´ ? px̂1 ` ix̂2 q
2
X0 “ x̂3
1
X´1 “ ? px̂1 ´ ix̂2 q . (9.3.21)
2
It is easy to check
rL3 , Xq s “ ~qXq
a
rL˘ , Xq s “ ~ p1 ¯ qqp1 ˘ q ` 1qXq˘1 , (9.3.22)
look back at (8.3.21). Compare with standard formulas in section 7.5 and Eqs.
(7.5.1) for action of L3 , L˘ on ang. mom. states |1, qy. These formulas match
exactly implying that with respect to angular momentum
Xq |n l my behave just like product states |1, qy|l, my . (9.3.23)
Another indication that this is an exact parallel is that in the position represen-
tation using spherical polars coordinates
c c
1 4π 4π
X˘1 “ ¯ ? r sin θ e˘iφ “ r Y1 ˘1 pθ, φq , X0 “ r cos θ “ r Y1 0 pθ, φq ,
2 3 3
(9.3.24)
which are angular momentum wavefunctions for l “ 1:
Ylm pθ, φq “ xθ, φ|l, my . (9.3.25)
What this means is that
ˆ ˙ ˆ ˙
L Xq |n l my “ rL, Xq s|n l my ` Xq L|n l my (9.3.26)
corresponds exactly to
ˆ ˙ ˆ ˙ ˆ ˙
L |1, qy|l, my “ L|1, qy |l, my ` |1, qy L|l, my , (9.3.27)
which is the usual action on product states for addition of angular momentum
for two subsystems. Use the same strategy as before:
˛ note that
ˆ ˙
L` X1 |n l ly “ 0
ˆ ˙ ˆ ˙
L3 X1 |n l ly “ ~pl ` 1q X1 |n l ly . (9.3.28)
So, as far as angular momentum is concerned (i.e., ignore r dependence for

current purpose)
|l ` 1, l ` 1y 9 X1 |n l ly . (9.3.29)
˛ Then ˆ ˙
|l ` 1, ly 9 L´ X1 |n l ly etc. (9.3.30)
˛ Find orthogonal state to |l ` 1, ly and take this as the next top state |l, ly.
˛ Repeat procedure as before.
Then we find that Xq |n l my has angular momentum q. numbers
l ` 1 or l or l ´ 1 for L2 and
(9.3.31)
m`q for L3
So for xn1 l1 m1 |Xq |n l my to be non-zero need
*
l1 P pl ` 1, l, l ´ 1q
second set of selection rules. (9.3.32)
m1 “ m ` q pq “ 0, ˘1q
But parity already forbids l1 “ l.
Summary: the possible atomic transitions for dipole radiation are given by
l1 “ l ˘ 1
m1 “ m if ε3 ‰ 0 (9.3.33)
m1 “ m ˘ 1 if ε1 or ε2 ‰ 0
In addition we must have

En1 “ En ˘ ~ω0 (9.3.34)
for large enough t, and the transition rate is then proportional to |ε|2 , the strength
(intensity or energy density) of the EM field/radiation. This corresponds to absorption
(` sign) and emission (´ sign). Although there are other operators which can cause
radiative transitions with different selection rules it turns out that dipole radiation of
the nature discussed here generally has the largest transition rate.
9.4 Emission and absorption of photons

‹ START OF NON-EXAMINABLE MATERIAL
Quantization of EM field discussed in outline in section 3.2. By expanding in normal

modes we can reduce the EM field equations to a set of oscillators. Here consider just
one mode of frequency ω0 for simplicity. By comparison with results in section 3 (on
the oscillator) and 4 (on Schrödinger and Heisenberg pictures) the electric field has a
Heisenberg picture operator for each polarization of the form (c.f. Eqn. (4.2.15))
ε ¨ Epω0 q 9 a eíω0 t ` a: eiω0 t (9.4.1)
where a, a: are the S-picture operators which destroy/create a photon of frequency ω0

and the given polarization ε. Use this now to sketch how transitions between atomic
energy levels are influenced by quantum EM field; before, in section 9.3 we treated
the EM field classically in the dipole coupling approximation. The discussion there is
modified by including the photon states in the initial and final states for the reaction:
|iy “ |n l my|N y
|f y “ |n1 l1 m1 y|N 1 y (9.4.2)
where N, N 1 are the number of photons or the oscillator level in the quantum version
for the field intensity which is N ~ω0 (ignoring the oscillator zero-point energy). As
before have two possible processes corresponding to time-dependent factors
eíω0 t eiω0 t (9.4.3)
and these are now respectively accompanied by factors

" ? " ?
1 N N1 “ N ´ 1 1 : N `1 N1 “ N ` 1
xN |a|N y “ xN |a |N y “
0 otherwise 0 otherwise
(9.4.4)
2
while the atomic matrix elements are unmodified. On taking |.| to get probabilities
find rates for absorption and emission of photon are equal except for the factors N and
N ` 1. Interpretation is as follows.
‚ Absorption. N 1 “ N ´ 1 and N is the intensity in photon units (~ω0 ) of the EM

radiation from which one photon is absorbed. Classically this is 9 |ε|2 . This is what
we expect.
‚ Emission. Look at two cases
‹ N “ 0, N 1 “ 1. This is spontaneous emission. There is no photon in the

initial state which then decays spontaneously emitting the photon. A prediction
of QM – not a classical phenomenon.
‹ N ą 0. Rate 9 N ` 1. We still associate the “1” here with spontaneous emission
but now the rate contains a term 9 N . I.e., the atom is induced/stimulated to
decay at rate proportional to the intensity of the radiation in which it is bathed. In
addition, the emitted photon is in the same state as those inducing its emission
– it is in phase with them and adds to the intensity of the radiation coherently.
If N is very large so that the excited atom is bathed in high-intensity coherent
radiation then the stimulated emission of a photon is overwhelmingly more likely
than spontaneous emission.
Thus for general atomic energy levels Ei ą Ej and EM radiation of frequency ω0 have
transitions when Ei ´ Ej “ ˘~ω0 .
Ei
absorption emission
Ej
Transition rates are of the form
loAomoon `
iÑj BiÑj ¨ pintensityq
looooooooomooooooooon or B jÑi ¨ pintensityq
looooooooomooooooooon
spontaneous stimulated spontaneous (9.4.5)
emission rate absorption rate
The quantities above are called the Einstein A and B coefficients. (There is a nice
original treatment using thermal equilibrium, the Boltzmann distribution for atomic
excitations and the Bose-Einstein distribution appropriate for indistinguishable pho-
tons.) Always have BiÑj “ BjÑi . Also, with intensity measured in natural units of
photon number then have
AiÑj “ BiÑj “ BjÑi . (9.4.6)
Stimulated emission is the basis of operation of the laser.
‹ END OF NON-EXAMINABLE MATERIAL
10 QUANTUM BASICS 77
10 Quantum Basics
There are many aspects of QM which still seem rather mysterious, particularly in
relation to measurements and how they have been axiomatized. For example,
‚ QM is inherently probabilistic whereas classical mechanics is deterministic.
‚ QM gives special status to (macroscopic) observer as distinct from (microscopic)

system.
‚ In QM measurements always interfere with the state of the system.
Obvious question raised by the first point:
Could probabilities in QM arise from ignorance of some hidden, deterministic (i.e.

inherently classical) degrees of freedom?
The main aim of this last section is to see how such questions may be addressed, and we
will also look briefly at some loosely related topics to do with information in classical
and quantum systems.
10.1 Classical and Quantum Data

Consider the simplest possible system: one measurable quantity s with observable
values 0 and 1.
‚ Classically, specifying s is one bit of information.
‚ In QM, s becomes an operator with eigenstates |0y and |1y corresponding to observed
values. The general normalized state of the system is
|ψy “ α|0y ` β|1y , k|ψyk2 “ |α|2 ` |β|2 “ 1 . (10.1.1)
So 8 (uncountable) many complex coefficients α, β (parametrizing S 3 ). This is called

a quantum bit or qubit.
It would seem that a qubit contains vastly more information than a classical bit, but
we must beware: the information is not easily accessed.
‚ Classically we can inspect s to find 0 or 1 without disturbing the system.
‚ In QM, measuring s irrevocably changes |ψy since it is projected onto |0y or |1y in
the process. We could recover the values of |α|2 and |β|2 to some degree of accuracy
if we had a large number of copies of |ψy, by measuring s many times. This raises
the question of whether we can faithfully copy some given state |ψy in a quantum
system. It turns out that this is forbidden by the
No-cloning Theorem.
We must be careful about what we mean. I can prepare an ensemble of systems each of
which is in the given state |ψy. For example, use a Stern-Gerlach apparatus to produce
a beam (the ensemble) of spin- 12 particles each with eigenvalue of σz “ 1: a “spin up”
state. Then clearly I have a lot of systems (i.e. the particles) in state
|ψy “ | Òy , (10.1.2)
Now choose the qubit to be an eigenstate of

ˆ ˙
cos θ sin θ
σθ “ n ¨ σ “ with n “ p sin θ, 0, cos θq . (10.1.3)
sin θ ´ cos θ
Then σθ has eigenstates
¨ ˛ evals
cos 2θ
| Ò θy “ cos 2θ | Òy ` sin 2θ | Óy ù ˝ ‚ `1
θ
sin 2
(10.1.4)
¨ ˛
´ sin 2θ
| Ó θy “ ´ sin 2θ | Òy ` cos 2θ | Óy ù ˝ ‚ ´1
θ
cos 2
And so the Stern-Gerlach prepared state can be written

θ θ
|ψy “ | Òy “ cos | Ò θy ´ sin | Ó θy . (10.1.5)
2 2
Thus, I can measure σθ many times and find |α|2 and |β|2 which are the respective
probabilities for measuring σθ “ ˘1. Alternatively, I can measure xσθ y. Here have
θ
|α|2 “ |xÒ θ| Òy|2 “ cos 2
2
θ
|β|2 “ |xÓ θ| Òy|2 “ sin 2
2
2θ 2θ
xσθ y “ cos ´ sin “ cos θ . (10.1.6)
2 2
So if I have an ensemble prepared in this way, there is no problem in interrogating the
ensemble to find detailed properties of the state |ψy. However, this is the only way I
can do it – I need an ensemble over which to average.
Suppose now that have access to just one system prepared in the state |ψy. For example,
I have one particle taken from my original ensemble. Can I now clone this state and
generate an ensemble like the one we just discussed? Can I take one particle from the
Stern-Gerlach apparatus, clone it many times and then deduce properties of |ψy from
the resulting ensemble? The no-cloning theorem states that this is not possible.
There are various versions of the theorem depending on assumptions and we will look
at the simplest. Suppose we have a quantum copier; a machine that can copy any
given state |ψy onto a ‘blank state’ |by. We assume all states are normalized and that
the copying is implemented by a linear, unitary, operator, U . Note that U must be

independent of |ψy since it must be able to copy any state.
U
|ψy b |by ÝÑ |ψy b |ψy . (10.1.7)
Can copy another state |φy in the same way:
U
|φy b |by ÝÑ |φy b |φy . (10.1.8)
Since U is unitary inner-products are preserved. So
xφ|ψyxb|by “ xφ|ψyxφ|ψy
ñ xφ|ψy “ 0 or 1 . (10.1.9)
This shows that copying can never work for general states in a space with dimen-
sion ą 1. This is a vital result for the success of the quantum encryption algorithm
where it is central to the idea that the encryption key for a classical algorithm can
be transmitted in such a way that a spy cannot intercept the message and decode it.
Interception irrevocably destroys the information and the interception is known to the
sender/receiver.
10.2 EPR experiment and Bell’s inequality

The EPR (Einstein-Podolsky-Rosen) thought experiment in modern language: two
spin- 12 particles are produced in an (entangled) spin-0 state
ˆ ˙
1
|ψy “ ? | Òy| Óy ´ | Óy| Òy , (10.2.1)
2
and components of spin σ A , σ B (leave out 12 ~ factor) of first and second particle are
subsequently measured by experimenters Alice (A) and Bob (B) in some order. How-
ever, they carry the particles far apart in space without altering the spin state |ψy
before measuring anything. Until the first measurement is made there is no definite
value for e.g. σzA or σzB .
Suppose Alice makes the first measurement of σzA and gets the result `1, say (prob-
ability 21 ). This projects |ψy onto the new state | Òy| Óy and so if Bob now measures
σzB he gets result ´1 with probability 1. He does not know this (unless Alice has sent
him a signal to tell him) but Alice knows with certainty what he will get! Einstein
was troubled by these kinds of non-local correlations but there is actually no loss of
causality here. Alice cannot use the collapse of the wave-packet
collapse
|ψy ÝÑ | Òy| Óy (Alice measured σzA “ `1q (10.2.2)
to send a message to Bob superluminally. Actually, Bob does not even know whether
or not Alice has made a measurement, since if the experiment is repeated many times
with an ensemble of pairs, Bob with get σzB “ `1 and ´1 with 50:50 probability
whether or not Alice has made a prior measurement.
One way to explain such correlations and idea of non-locality is to ask whether the
spin components could actually take definite values, which are correlated, at the
instant the particles were created, i.e. at their point of origin. We still need a classical
probability distribution to describe how these values arise, but this ultimately might be
explained by some deterministic hidden variables. Can such a distribution reproduce
the predictions of QM. Or, if not, which possibility is supported by experiment?
A simple example is suppose I make two identical parcels one containing a green ball
and the other a red one. I give one to Alice and one to Bob who then travel far apart.
When Alice opens her parcel and sees, say, a green ball she knows instantly the Bob
has a red ball. This ”learning at a distance” is mundane and not controversial; there
is no perception of non-locality. Can repeat many times and, indeed, Bob has a 50:50
change of getting red and green. However, in this case each ball had a definite colour
before starting out – there is no analogue of ”collapse of the wavepacket”. Can the
spins in an entangled state be thought of as having definite spin components in a like
manner?
It is useful to think about measuring different components of σ A and σ B and use the
results above for eigenstates of σθ in Eqns. (10.1.4) and (10.1.6) to consider first what
happens to a single particle in conventional QM. Have
θ
|xÒ θ| Òy|2 “ |xÓ θ| Óy|2 “ cos 2
2
θ
|xÒ θ| Óy|2 “ |xÓ θ| Òy|2 “ sin 2 . (10.2.3)
2
These results clearly depend on the angle between k̂ and n defined earlier. Also, for a
component along a new direction φ we have
ˆ ˙
2 2 2 θ´φ
|xÒ θ| Ò φy| “ |xÓ θ| Ó φy| “ cos
2
ˆ ˙
2 2 2 θ´φ
|xÒ θ| Ó φy| “ |xÓ θ| Ò φy| “ sin . (10.2.4)
2
Now return to the two particle system |ψy and note that the spin-0 state is
ˆ ˙
1
|ψy “ ? | Òy| Óy ´ | Óy| Òy
2
ˆ ˙
1
“ ? | Ò θy| Ó θy ´ | Ó θy| Ò θy . (10.2.5)
2
This can be checked directly but also follows because spin-0 is a rotationally invariant
state – it cannot depend on the choice of the z-axis.
We want to consider probabilities for various outcomes when
Alice measures one of Bob measures one of
then (10.2.6)
σzA , σθA , σφA σzB , σθB , σφB
E.g.,
σzA “ 1 new state σθB “ ´1
then θ (10.2.7)
|ψy : prob: 1
| Òy| Óy prob: |xÓ θ| Óy|2 “ cos 2
2 2
Total probabilities for this and some other measurement outcomes are similarly
θ
P pσzA “ `1, σθB “ ´1q “ 1
2
cos 2
2
φ
P pσzA “ ´1, σφB “ `1q “ 21 cos 2
2
ˆ ˙
θ´φ
P pσθA “ `1, σφB “ `1q “ 1
2
sin 2
. (10.2.8)
2
Also, know components along same direction are exactly anticorrelated. So, e.g.
P pσzA “ 1, σzB “ ´1q “ 1 etc. (10.2.9)
All is standard QM.

Now ask could there exist classical variables
sA A A
z , sθ , sφ and sB B B
z , sθ , sφ (10.2.10)
for each particle, all taking values ˘1 according to some probability distribution, with
values assigned as the particles are created (c.f. the classical red and green ball discus-
sion), yet still reproducing the QM results?
‚ To ensure anticorrelation of spins along same direction probability must be zero

unless sB A
z “ ´sz etc. So we can write the distributions as functions of Alice’s
variables ppsA A A
z , sθ , sφ q alone.
‚ Then, summing over variables that are not specified, we can write
P psA A
θ “ `1,sφ “ ´1q
“ pp1, 1, ´1q ` pp´1, 1, ´1q

“ ‰ “ ‰
ď pp1, 1, ´1q ` pp1, 1, 1q ` pp´1, 1, ´1q ` pp´1, ´1, ´1q
“ P psA A
z “ `1, sθ “ `1q ` P psA A
z “ ´1, sφ “ ´1q (10.2.11)
‚ Specifying outcomes in terms of one measurement for A and one for B instead (using
exact anticorrelation) we get
P psA B A B A B
θ “ `1, sφ “ `1q ď P psz “ `1, sθ “ ´1q ` P psz “ ´1, sφ “ `1q . (10.2.12)
This is Bell’s inequality of which there are many versions.
‚ Can such a probability distribution reproduce the results of QM? If so, then the
expressions derived for the QM probabilities above must satisfy Bell’s inequality.
This means ˆ ˙
2 θ´φ θ φ
sin ď cos 2 ` cos 2 (10.2.13)
2 2 2
for any θ and φ. But this is false. E.g., θ “ 3π{4, φ “ 3π{2 with this inequality
would imply
ˆ ˙
2 3π 2 3π 3π
´ cos ´ sin ď cos 2
8 8
looooooooooooomooooooooooooon 4
cos 3π{4
1 1
ñ ? ď FALSE! (10.2.14)
2 2
The fact that this approach cannot reproduce QM means that we can distinguish these
alternatives experimentally, and it is QM which is correct. With these assumptions,
the indeterminate nature of the quantum state (no definite values of observables until
measurement is made) seems inescapable. There is a non-locality to the notion of
collapse of the wavefunction.
10.3 Density operators and hidden sectors

Next topic is still about conventional QM but raises questions about ‘hidden’ or unob-
served parts of the system.
‚ Observations usually confined to some subset of physical degrees of freedom, either by

choice or necessity. What effect will the rest of the system have on our measurements?
‚ There is loss of information. Expect some uncertainties and probability distribution

to emerge, but not like fundamental probabilities which are part of measurement
axioms.
‚ Need a general formalism to handle this. Useful in e.g., statistical mechanics where
‘hidden’ degrees of freedom are all the fine details of the system which are not of
interest. I.e., we care about pressure, temperature etc. but not the dynamics of
individual particles.
‚ Develop idea of density or state operator.
Consider a space of states of the form
V “ U b W (10.3.1)
where U is observed but W is the hidden/unobserved sector. The general state is then
ÿ
|Ψy “ αia |ψi y b |φa y , (10.3.2)
ia
with t|ψi yu and t|φa yu orthonormal bases for U and W , respectively. Assume
ÿ
k|Ψyk2 “ |αia |2 “ 1. (10.3.3)
ia
What we really mean by distinction between U and W is that observables Q act just
on U , i.e. Q acts on V as the operator Q b I. Then
ÿÿ
˚
xQyΨ “ αjb αia xψj |Q|ψi y loomoon
xφb |φa y
ia j b
δab
ÿ
“ βij xψj |Q|ψi y . (10.3.4)
ij
ÿ
˚
The matrix βij “ αia αja is hermitian and positive-definite, and
a
ÿ ÿ
βii “ Trpβq “ |αia |2 “ 1 : trace or sum of eigenvalues . (10.3.5)
i ia
Hence, D a unitary matrix S diagonalizing β with

ÿ
S : βS “ diagppi q where evals pi ě 0 and pi “ 1 . (10.3.6)
i
Define a new orthonormal basis for U

ÿ
|χi y “ |ψj ySji , (10.3.7)
j
and then find

ÿ
xQyΨ “ pi xQyχi with xQyχi “ xχi |Q|χi y . (10.3.8)
i
‚ Working just with U , the effect of the hidden sector W is to produce a probability
distribution tpi u for a particular set of states |χi y in the observed sector.
‚ If there is only one non-zero probability we say we have a pure state for U and |Ψy
can then be written as a single tensor product
|Ψy “ |χ1 y b |φ1 y ñ β “ diagp1, 0, . . . , 0q (10.3.9)
where I have chosen p1 “ 1, pi “ 0 i ą 1.
‚ If there is more than one non-zero probability we say we have a mixed state for U ,
and |Ψy cannot be written as a single tensor product. In this case we also say that
we have an entangled state of the subsystems U and W .
Density operators or matrices provide a powerful way of re-expressing results about
quantum measurements. Recall first that the trace of an operator A on a space V is
defined by
ÿ
Tr V pAq “ xn|A|ny , t|nyu any orthonormal basis for V. (10.3.10)
n
We can express all aspects of measurements in terms of traces of operators rather than
inner-products of states. So, with notation as before,
on V “ U b W
define ρ “ |ΨyxΨ| , the density operator corresponding to |Ψy .
(10.3.11)
Then
ÿ ÿ ˆ ˙
xQyΨ “ xΨ|Q|Ψy “ xΨ|nyxn|Q|Ψy “ xn| Q|ΨyxΨ| |ny “ Tr V pQ ρq (10.3.12)
n n
with Tr V pρq “ x1y “ 1.

When V has the tensor product structure given above we have a convenient way of
restricting the calculation to U by using the reduced density matrix operator
ÿ ÿ ÿ
˚
ρ̄ “ αia αja |ψi yxψj | “ βij |ψi yxψj | “ pi |χi yxχi | . (10.3.13)
ij a ij i
For an observable Q acting just on U we have
xQy “ Tr U pQ ρ̄q (10.3.14)
To show this choose t|nyu “ t|χi yu as the basis to evaluate the trace. Then
ÿ ÿ
Tr U pQ ρ̄q “ xχj |Q|χi y looomooon
xχi |ρ̄|χj y “ pi x|Q|yχi “ xQy . (10.3.15)
i,j i
pi δij
so everything is written in terms of the ‘observable’ sector. Other results:
‚ Provided the states |χi y obey the time-dependent Schrödinger equation with Hamil-
tonian H, the EoM for ρptq is (see Q4.8(a))
d
i~ ρ “ rH, ρs . (10.3.16)
dt
‚ For a pure state have
ρ̄2 “ ρ̄ and Trpρ̄2 q “ Trpρ̄q . (10.3.17)
I.e., ρ̄ is a projection operator on U (see Q4.8(b)).

‚ For an entangled state
ρ̄ 2 ‰ ρ̄ and Trpρ̄2 q ă Trpρ̄q . (10.3.18)

PII - Principles of Quantum Mechanics - Horgan (2014) 87pg

Uploaded by

Copyright:

Available Formats

PII - Principles of Quantum Mechanics - Horgan (2014) 87pg

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PII - Principles of Quantum Mechanics - Horgan (2014) 87pg

Uploaded by

Copyright:

Available Formats

What are the main topics covered in the document?

What are the main topics covered in the document?

What is the difference between pure and mixed states?

What is the difference between pure and mixed states?

Part II Principles of Quantum Mechanics

November 28, 2014

3 The Harmonic Oscillator 17

4 Pictures and Operators 22

5 Composite Systems and Identical Particles 27

6.2 Degenerate case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

8 Transformations and Symmetries 57

9 Time-dependent Perturbation Theory 67

‚ E. Merzbacher Quantum Mechanics, 3rd edition. Wiley 1998 (various prices)

‚ C.J. Isham Lectures on Quantum Theory: Mathematical and Structural Foundations.

‚ S. Gasiorowicz Quantum Physics, 3rd edition. Wiley International 2003 £48.71

‹ reformulate QM in a more powerful, abstract, flexible and useful form: Dirac

|ψy, |φy P V ùñ α|ψy ` β|φy P V @ α, β P C. (2.1.1)

Physically this is the superposition principle leading to wave-like behaviour (interfer-

α, β P C. This is the definition of the dual as a vector space.

and is assumed to obey

xφ|ψy “ xψ|φy˚ hermitian

|ψy ÞÑ Q|ψy, (2.1.7)

xφ| ÞÑ xφ|Q , (2.1.9)

For any Q the hermitian conjugate or adjoint is an operator Q: defined by

Simple consequences are

for any A, B.1

True for all |ψy, |φy and result follows.

For any operator Q call |ψy p‰ 0q an eigenstate of Q with eigenvalue λ if

Q|ψy “ λ|ψy . (2.1.15)

Equivalently (taking : of this)

xψ|Q: “ λ˚ xψ| . (2.1.16)

For general Q, λ can be complex.

rA, Bs “ AB ´ BA “ ´ rB, As . (2.1.17)

We have the identities

rA, B1 B2 s “ rA, B1 sB2 ` B1 rA, B2 s

2.2 Observables and measurements

(i) All eigenvalues are real.

We prove (i) and (ii) and assume (iii).

Combining these three properties we have

We speak of diagonalizing Q by choosing the basis to be the eigenstates |ny of Q.

For the state to be properly normalized

‚ The result is an eigenvalue, λ, say.

‚ This value is obtained with the probability

‚ Immediately after the measurement the state is

where c is a normalization constant chosen so that xφ|φy “ 1.

So measurement projects the system into the eigenspace Vλ .

‚ ppλq “ |αn |2 “ |xn|ψy|2 .

as required for a probability distribution.

and the uncertainty (spread or variance) is

p∆Qq2ψ “ xpQ ´ xQyψ q2 yψ “ xQ2 yψ ´ xQy2ψ . (2.2.15)

In thinking about many repeated measurements we must be careful to prepare the

xQyψ “ λ , and p∆Qqψ “ 0 . (2.2.16)

kpA ` iλBq|ψyk2 ě 0 @ real λ . (2.2.18)

2.3 Time evolution and the Schrödinger Equation

H|ny “ En |ny . (2.3.4)

Then the states

we have the unique solution

Q|1y “ |2y, Q|2y “ |1y ñ rQ, Hs ‰ 0 . (2.3.8)

The eigenstates of Q are easily found to be

Note that we are assuming no time-dependence in H. This would become a much