Notes on Quantum Mechanics

W.J.Zakrzewski

October 8, 2007
Abstract
Quantum Mechanics is introduced using state vectors, linear operators and canonical quantisation
rules. Spectra of some operators are calculated algebraically. The Schr odinger Equation, is introduced
and applied to problems of a single particle in a potential eld and then to some three dimensional
problems.
Contents
1 Introduction 3
1.1 Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Quantum Mechanics and its place . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Crisis in Classical Mechanics - Quantum Phenomena . . . . . . . . . . . . . . . . . . . . . . . 4
2 Theory of Quantum Mechanics 5
2.1 States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Measurements: Physical Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Revision of Classical Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Quantum Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6 Simple Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.6.1 Position and Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.6.2 Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6.3 Energy of a Simple-Harmonic Oscillator (one dimension) . . . . . . . . . . . . . . . . . 17
3 Representation Theory 18
3.1 Position Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.1 Wave function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.2 Change of basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1.3 Finding the action of the p operator (setting up the Schr odinger representation) . . . 20
3.1.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Probabilistic Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Harmonic Oscillator Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Relation to the Algebraic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5 Momentum Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

These notes are an edited version of notes kindly provided by D.J.Smith, who in turn based them on the notes of R.C.
Johnson and on the notes I gave him.
1
4 Equation of Motion - Dynamics 28
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 The Schr odinger picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3 The Heisenberg Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4 Conserved quantities; constants of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.5 Stationary states. Time independent Schr odinger equation . . . . . . . . . . . . . . . . . . . . 31
4.5.1 Stationary states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.5.2 Schr odinger wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.5.3 Example - Free particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.6 Spreading of a wave packet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.7 Ehrenfests theorems; Energy-time uncertainty principle . . . . . . . . . . . . . . . . . . . . . 35
4.7.1 Ehrenfests theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.7.2 Time-energy uncertainty principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.8 Probability current; Conservation of probability . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5 One-Dimensional Systems 38
5.1 Preliminary (recall also 4.9) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2.1 Free Particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2.2 One Dimensional Step Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2.3 One Dimensional Potential with Rigid Walls . . . . . . . . . . . . . . . . . . . . . . . 39
5.3 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.4 Finite Square Well . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.5 Reection and Transmission Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.6 The Tunnelling Eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.7 Lessons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6 Three-Dimensional Systems 46
6.1 Square well with rigid walls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.2 3 Dimensional Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.3 Central Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.4 Separation of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.5 The Legendre Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.6 Convergence Problems - Legendre Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.7 Spherical Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.8 An Algebraic Approach to Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.9 Radial Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.9.1 Free Particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.9.2 Spherical Square Well . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.10 Two Particle Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.10.1 Isotropic Simple-Harmonic Oscillator - once again . . . . . . . . . . . . . . . . . . . . 60
6.11 Hydrogen Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7 Conclusion 62
7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.3 And Theres More . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Useful Books
Classic Texts:
L I Schi, Quantum Mechanics, McGraw-Hill (3rd edition 1968)
2
L D Landau and E M Lifshitz, Quantum Mechanics, Pergamon (2nd edition 1965)
A Messiah, Quantum Mechanics, Vols I & II, North-Holland (1961)
Modern Texts:
L E Ballentine, Quantum Mechanics, Prentice Hall (1990)
F Mandl, Quantum Mechanics, John Wiley (1992)
A Sudbery, Quantum Mechanics and the Particles of Nature, CUP (1986)
Standard Monograph:
P A M Dirac, The Principles of Quantum Mechanics, OUP (4th edition 1958)
Background:
J D Jackson, Mathematics for Quantum Mechanics, Benjamin (1962)
R P Feynman, Lectures on Physics, Vol III, Addison-Wesley (1965)
A J G Hey & P Walters, The Quantum Universe, CUP (1987)
P V Landsho & A Metherell, Simple Quantum Physics, CUP (1979)
J C Polkinghorne, The Quantum World, Longman (1984)
There are very many more.
1 Introduction
1.1 Prerequisites
This lecture course will depend very strongly on the linear algebra courses from your rst two years and parts
of the second year Analysis in Many Variables module. You will also need some knowledge of Hamiltonian
dynamics (but we will revise what is needed.) But please, do revise the linear algebra courses from the
rst two years (vector spaces, bases, expansions of vectors, linear operators, eigenvectors, eigenvalues etc.)
1.2 Quantum Mechanics and its place
Familiar classical mechanics applies to the everyday world of moderate-sized objects. Very small things like
atoms and molecules, i.e. objects of size of 10
8
m or smaller, need quantum mechanics, and the scale is set
by the (reduced) Planck constant h 10
34
Js (Joule seconds SI units of energy time.) It appears that
quantum mechanics describes the world of atomic and subatomic scales probably at least down to 10
34
m.
At scales smaller than this, strings and other objects of elementary particle physics, may be relevant and
they may require a modication of standard quantum mechanics but such questions have not been resolved
yet and are subjects of current research.
Classical mechanics and quantum mechanics are examples of mathematical models of what we (aided by
measuring instruments) see around us. A mathematical model identies measurable quantities with abstract
mathematical objects that are manipulated according to certain axioms or postulates and then interpreted to
make predictions. For instance in the simplest form of classical mechanics positions, momenta, forces etc. are
identied as vectors and Newtons Laws taken as axioms that determine them. Ideally, models are accepted
or rejected strictly according to the success or otherwise of their predictions. But consistency with models
of neighbouring sectors of the world is important and, because models using mathematics seem always to
work so well, mathematical elegance exerts a strong inuence too.
Classical mechanics, in its proper domain, succeeds beautifully. Design of tables and chairs, planes and
space ships depends on it crucially. But modern life depends also on the success of quantum mechan-
ics. Quantum Mechanics describes how atoms are put together and why they are stable; is responsible for
magnetism and chemistry. Apart from its high aesthetic value it has practical applications. Lasers, super-
conductors, atom bombs, and the ubiquitous silicon chip all work because theyre designed with the help of
3
quantum mechanics. Moreover the two models are both elegant and mutually consistent. In the limit h 0
quantum mechanics gives classical results.
In addition, Quantum Mechanics is probably the most original (challenging) theory of physical phenom-
ena. Its description is very dierent from that of classical mechanics and some of its predictions may appear
very counter-intuitive.
In advanced classical mechanics the basic variables are the generalised coordinates and their conjugate
momenta, i.e. the q, p coordinates in phase space. Measurable quantities like energy, force, velocity etc.
are constructed from them. These are assumed observable to any necessary precision, simply by looking
carefully enough. The equations of motion are Hamiltons equations, which are dierential equations for the
ps and qs as functions of time. In another, completely equivalent formulation, one uses Lagrangians, which
are functions of generalised coordinates and their time derivatives (generalised velocities) and the equations
of motion are the second order equations for those generalised coordinates.
In quantum mechanics the fundamental entity is the state vector, an element of a linear vector space.
Measurements are explicitly modelled through properties of certain linear operators in the space. The
equation of motion can be written as a dierential equation for the state vector, the Schr odinger Equation.
Its a reasonable assumption based on everyday experience that classical measurements are possible to
arbitrary accuracy. But in the quantum world of atoms and molecules normal intuition fails. In general no
matter how delicate and skillful the observer, a disturbance of the observed system is inevitable. To some
extent this is understandable, for looking at an atom means bouncing light o it and the atom is so small
that the collision is bound to upset it. All is not lost, however, for quantum theory predicts the probabilities
of alternative possible results of individual measurements. In the large-scale limit averages are observed and
classical mechanics is recovered.
The linear-space structure of the theory the superposition principle is the key to uncertainty, for
a state vector composed of other state vectors will allow a measurement to realise any of the corresponding
physical congurations. Of course the superposition principle is a familiar feature of classical waves, e.g.
sound and light. In quantum mechanics we see wave-particle duality for microscopic systems.
1.3 Crisis in Classical Mechanics - Quantum Phenomena
The quantum era may be dated from Becquerels discovery of radioactivity in 1896. But explanation as a
quantum tunnelling eect did not come until Gamow, Gurney and Condon in 1928.
Planck in 1901 produced the rst satisfactory theory of blackbody radiation with the revolutionary
idea that matter and electromagnetic radiation interchange energy in packets (quanta) with energy E
proportional to frequency . The constant of proportionality is Plancks constant, h. Einstein used Plancks
idea to explain features of the photoelectric eect (1905) (for which he got his Nobel prize) and to solve
problems with specic heats (1907).
In 1897 Thomson discovered the electron. Its charge was dicult to measure but in experiments involving
metals that were heated electrons were given o. Thomson concluded that the electron had a specic value
of charge (i.e. charge was quantised) and that its mass was a very tiny fraction of the mass of the hydrogen
atom. Then the question arose; if electrons are given o, where do they come from? What do atoms look
like? Hence Thomson proposed a plum pudding model of the atoms - involving a cloud of positively
charged material with negatively charged electrons stuck in it. However, in 1911 Rutherford showed that
this is wrong. His scattering experiments suggested that the atom consists of a cloud some 10
8
cm across of
electrons bound by Coulomb attraction to a relatively massive central nucleus about 10
13
cm across. But
according to classical notions this system is completely unstable, since the bound charges are accelerated
and therefore radiate electromagnetic waves. So they rapidly lose energy and the atom collapses. Classical
mechanics inevitably predicts unstable matter with a lifetime of typically 10
10
seconds! This is even before
understanding how such apparently insubstantial things, mostly empty space, can be arranged into solids,
liquids and gases.
Bohr (1913) introduced stability at the atomic level with the ad hoc postulate that atomic electrons
can have angular momentum equal only to an integral multiple of h h/2. Transitions between two of
the resulting discrete electronic energy-levels separated by E then conserve energy by emitting or absorbing
radiation quanta of denite frequency = E/h. This model agrees with the main features of the line spectra
4
of light emitted from simple atoms.
Explanation of ner structure in atomic spectra came after the discovery of electron spin and magnetic
moment, following the Stern-Gerlach experiments (1921). The electron behaves like a spinning charge with
angular momentum
1
2
h. Subtleties stemming from this (Fermi-Dirac Statistics) led to successful models of
multi-electron atoms and eventually (in the 50s) to understanding the stability of bulk matter.
But meanwhile Compton (1923) found that X-rays (wavelength 10
11
cm) scatter from atomic
electrons like particles (photons) moving at the speed of light c with momentum p = h/ and energy
E = hc/ = h. In 1925 de Broglie proposed that also electrons and other particles might show wave-like
behaviour, with wavelength and momentum related by = h/p. For atomic electrons, if they occupy only
circular orbits with a standing de Broglie wave, the Bohr quantisation rule follows since in such an orbit
of circumference 2r there is a whole number of wavelengths h/p. Davisson and Germer (1927) conrmed
diraction and interference of electrons scattered by metals for de Broglie wavelengths of the order of atomic
size and spacing.
During the 20s Heisenberg, Born and Jordan were introducing matrix mechanics, Schr odinger was devel-
oping wave mechanics from de Broglies idea, and Dirac (1926) discovered that both are manifestations of
a new linear theory. Schr odingers theory of the hydrogen atom (the simplest) agreed remarkably with the
Bohr model.
In quantum mechanics wave-particle duality and quantisation of energy and angular momentum come
directly from non-commutativity of linear operators that model observations. Another consequence is the
Uncertainty Principle and statistical scatter of individual observations.
2 Theory of Quantum Mechanics
2.1 States
The conguration of a classical system at any time is specied by a point in 2n dimensional phase space
i.e. by its coordinates, the set of qs and ps. Equivalent description is in terms of a point (and its velocity)
in conguration space. The mathematical objects of the theory are these coordinates as functions of time.
Other observables are constructed from them. Of course, some quantities, do not play a role; i.e. colour of
a falling ball or its internal structure. Thus they not appear in the description.
For a quantum system the qs and ps are not all simultaneously measurable with precision (compatible),
as will appear. A conguration is instead specied by a set of measurements that are mutually compatible.
The results of these are used as labels for a state vector. A state is thus described as an undisturbed motion
that is restricted by as many conditions as are theoretically possible without mutual interference.
The state vector (or state for short) is the central mathematical object and contains all information
about the system. It is written | in the Dirac notation, where stands for the set of labels needed for
unambiguous specication in the current context.
In classical mechanics the basic quantities are functions of time which thus belong to a (vector) space of
functions. Similarly, the quantum mechanical state vector also belongs to a vector space which, as we will
see, is a complex Hilbert space (a complete vector space with an inner (scalar) product).
Examples: |p could be a state of a particle of momentum p; |E, p could be a state of
a free particle of denite energy and momentum; |r
1
, r
2
could be a state of two particles at
positions r
1,2
; |E, j, m could be a state of an atom with denite energy, angular momentum
and z-component of angular momentum; |E
n
or just |n could be the state of a system with
energy E
n
, the n
th
of a discrete set of possibilities.
Examples: Other, more homely, mathematical theories may be formulated with state vectors,
and operators. For instance, models of:- travel round networks; stochastic (Markov) processes.
State vectors | belong to a linear vector space over CC, comprising all possible states of the system. The
axioms of a linear vector space involve the existence of a zero element and commutativity with numbers.
The superposition principle is very important: if | and | are in the space then so is c
1
| +c
2
| for all
complex c
1,2
.
5
Superposition in quantum mechanics embodies uncertainty, for a state may be linearly composed of other
states, each corresponding to a dierent possible outcome of a measurement.
Example: If an atom may be observed to have energies E
1
or E
2
, with corresponding states
(state vectors) |E
1
and |E
2
, then another possible state of the atom is described by the state
(vector) | = |E
1
+|E
2
, say. (Here the sign is just a convenient state label. It doesnt mean
that | has energy E
1
+E
2
nor that its energy=. As will become clear, an energy measurement
made on state | may realise either outcome, E
1
or E
2
, with equal probability.
The theory gives the same signicance to | and c| for any non-zero complex number c. Such an
equivalence corresponds to considering rays in the space of states. Often one exploits this equivalence and
considers normalised states of the original space. However, for this to be made precise one needs to dene
the norm.
With each state vector | is associated a dual vector |. Then the inner or scalar product of two
states is dened as the complex number | with the property that
| = |

where * means complex conjugate. Note that this implies that the dual to the state vector c| is c

| and
that | is a real number.
In Diracs terminology | is called a ket vector and its dual | is called a bra vector.
As usual the inner product is distributive over linear combination of states and obeys
| 0
with equality i | = 0. Then the Schwarz Inequality
|| |||
2
follows from | 0 for all complex c where | = | +c|. Also the convention is to normalise states to
|
2
= | = 1
whenever possible. This xes any multiplicative complex constant up to a phase (and in fact corresponds to
standard normalisation of probability density functions, as will appear).
Examples:
(i) If | etc. are represented by complex-valued functions

(x) etc. of the real variable

x (0, 1) then a suitable inner product | is
_
1
0

(x)

(x) dx.
(ii) With states represented by column vectors
i
etc. then an appropriate inner product is

+
, where
+
is the Hermitian conjugate of that is, its transpose with complex conjugate
elements.
Examples (i) and (ii) illustrate several ideas normalisation for instance: in (i)

=

2 sin x is
normalised by the factor

2. An unnormalised state | can always be normalised by dividing by |
1/2
.
Note that an additional factor e
i
for any real does not change the normalisation.
Two states are mutually orthogonal if their inner product is zero; the only state orthogonal to all others
is the zero vector. Examples in cases (i) and (ii) are easy to nd.
Recall the idea of linear independence of a set of vectors, remember that a spanning set can be used
to express any vector in the space as a linear combination, and recall that a basis for a linear vector space
is a linearly-independent spanning set.
In quantum mechanics it is assumed that the state space is spanned by a set of states corresponding
to all the dierent possible outcomes of measurements of relevant observable quantities momentum,
6
energy, angular momentum, position, or whatever i.e. physical completeness corresponds to mathematical
completeness. These spanning states turn out to be mutually orthogonal (or can be made so by the Gram-
Schmidt procedure), hence they are linearly independent, and so basis sets. They may be nite or innite
in number, depending on the system and on the observable. Of course because some measurements are
mutually incompatible their corresponding bases are alternatives. The usual basis is that corresponding to
a maximal set of mutually compatible measurements. This is less vague than it sounds for the simplest
quantum systems.
Note that if an orthogonal basis {|i; i = 1, . . . ; i|j = 0, i = j} is normalised: i|i = 1, then in the
basis expansion
| =

i
c
i,
|i
the coecients c
i,
are simply given by
c
i,
= i|.
If moreover | is normalised then

i
|c
i,
|
2
= 1.
Examples: In case (i) recall Fourier Series; in case (ii) there is the standard basis {u
i
} where
u
i
has 1 as its ith element and zero elsewhere.
We assume at present that the basis is discrete (countable). Innite-dimensional inner-product spaces
over CC like this are called Hilbert Spaces. Comment: note, however, that the condition of countability of
the set of basis vectors is sometimes relaxed.
2.2 Operators
All physical information contained in a systems state vector is extracted by certain linear operators acting
in the state space. Any operator

A in the space maps states to states:
| | =

A|.
A linear operator commutes with complex numbers and its operation is distributive over vector addition.
Examples: A simple linear operator is the identity

I that maps every vector to itself. Another
is the back-to-back pairing of a vector and a dual vector

A = ||
that maps c
1
| + c
2
| to the state | multiplied by the complex number c
1
| + c
2
|. In
example (i) there are dierential operators d/dx, d
2
/dx
2
etc; and in (ii) linear operators are
(complex) square matrices.
The sum of two operators is an operator. Writing the expansion of an arbitrary vector
| =

i
|ii|
in terms of any complete orthonormal (basis) set {|i; i|j =
ij
} we deduce a representation of the identity
operator;

I =

i
|ii|.
Each term |ii| is an example of a projection operator.
The product

A

B is an operator dened by

B| =

A(

B|)
for all states |.
7
Examples: For any

A we have

A

I =

I

A =

A. For projectors

P
i
= |ii| we have

P
i

P
j
=
ij

P
j
.
Operator multiplication is generally not commutative, as illustrated by operators || and ||. Non-
commutativity is signicant in discussion of compatibility of measurements, when it is useful to dene the
commutator
[

A,

B]

A

B

B

A.
We will appreciate the true signicance of the commutator later, when we discuss the compatibility of
measurements.
Note that the commutator has properties similar to those of the Poisson Bracket of classical mechanics:
[

A,

B] = [

B,

A]
[

A +

B,

C] = [

A,

C] + [

B,

C]
[c

A,

B] = c[

A,

B]
[

B,

C] =

A[

B,

C] + [

A,

C]

B
[

A, [

B,

C]] + [

C, [

A,

B]] + [

B, [

C,

A]] = 0.
The last of these is the Jacobi Identity.
Positive integer powers of operators are dened by

A
n
=

A(

A
n1
) with

A
0
=

I. The usual index laws
apply and the unique inverse

A
1
obeys

A
1
=

A
1

A =

I
if it exists. The inverse of a product of operators is easily seen to be the product of inverses in reverse order.
Functions of operators, like exp

A, can be dened by e.g. power-series expansion. Then e.g. (exp

A)
1
=
exp

A, but some convention is needed to deal with expressions like exp(

A+

B) if [

A,

B] = 0. Clearly func-
tions of operators commute i the operators themselves commute.
The inner product |(

A|) is written symmetrically as

|

A|
and called a matrix element of

A. The matrix elements of

A dene its action in the dual space by dening
the bra vector |

A for arbitrary |.
Examples: The dual-space action of

A = || is clear. In the space of example (ii) of column
vectors u

, with duals the rows u

+

and where the operators are square matrices M, the result is

just the matrix product u
+

M. In the function space (i) for dierential operators where the inner
product is an integral the corresponding dual space operation is dened by integration by parts,
with homogeneous conditions specied to make boundary terms vanish.
The vector dual to the ket

A| is the bra |

. Therefore the denition of the adjoint operator

A

uses the complex inner product property, i.e.

|

| = |

A|

for all states.

Examples: If

A = || then

A

= ||. In (ii) the adjoint of square matrix M is its Hermitian

conjugate M
+
. In function space (i), for a dierential operator, integrate by parts.
It is easily seen that (

=

A, that the adjoint of a product is the product of adjoints in reverse order
and that (c

A)

= c

.
The eigenvalue problem for an operator: if

A|a = a|a
8
for some complex number a and non-zero vector |a, then |a is an eigenvector of

A belonging to eigenvalue
a. (Using a as state label is actually consistent with labelling by possible results of observation, as we will
see later, the latter turn out to be just eigenvalues).
Note that the equation dening an eigenvector is homogeneous so normalisation is arbitrary and can be
chosen for convenience.
The set of eigenvalues and eigenvectors of

A is called its spectrum. The spectrum of

A can be empty,
nite, countably innite, or continuous. Some operators in quantum mechanics have an innite spectrum,
part countable and part continuous.
Examples: If

A = || then its spectrum consists of eigenvector | with eigenvalue |,
plus any non-zero states orthogonal to |, each with eigenvalue zero. The projector |ii| is
a particular case. This illustrates the possibility of degeneracy, when two or more linearly
independent eigenvectors belong to the same eigenvalue. For instance the entire spectrum of

I is degenerate, as any non-zero state is an eigenvector with eigenvalue unity. If the vectors
in space (ii) have n components we have the familiar n n matrix eigenvalue problem, where
besides degeneracy a defective spectrum is common. In a space (i) of functions (x) on [0,1]
the operator d/dx has an empty spectrum if the space is restricted by the boundary condition
(0) = (1) = 0. However with these boundary conditions the operator d
2
/dx
2
has eigenvectors
sin nx with eigenvalues (n)
2
for n = 1, 2 . . .. If there are no boundary conditions the spectrum
of d/dx is
a
(x) = exp(ax) for any complex eigenvalue a, and that of d
2
/dx
2
is
a
(x) = exp(iax)
with eigenvalue a
2
for any a; twofold degeneracy. Note the illustration of possible dangers in
too carelessly asserting that if

A|a = a|a then

f(

A)|a = f(a)|a, although this is usually true.

Theorem: If both operators

A and

B have non-empty spectra then they have a common set
of eigenvectors if [

A,

B] = 0. The converse is true if the eigenvectors are complete in the space.
Proof: Let

A

B =

B

A and

A|a = a|a. Then

A
_

B|a
_
=

B

A|a = a
_

B|a
_
.
The inference is that either
_

B|a
_
is zero or, since it obeys the same eigenvalue equation as |a, it must
actually be |a up to a factor anyway i.e.

B|a = b|a. The inference made here is a crucial step
appearing repeatedly. It is clearly true if the eigenvalue a is unique (non-degenerate). Otherwise

B|a is
some linear combination of degenerate states with eigenvalue a. However the degenerate subspace may be
diagonalised with respect to

B and so simultaneous eigenvectors constructed. Thus the rst part is proved.
To establish the converse, if both

A|a, b = a|a, b and

B|a, b = b|a, b then

A

B|a, b = ab|a, b =

B

A|a, b.
This holds for any linear combination of eigenvectors |a, b and so we deduce that [

A,

B] = 0 if the set {|a, b}
is complete. Completeness is assumed in quantum mechanics for operators identied with observables.
Its important to realise, and worth repeating, that even if [

A,

B] = 0 then an eigenstate of

A is not
automatically an eigenstate of

B unless the spectrum of

A is non-degenerate!
Example: Commuting pairs with common eigenvectors include

A and

f(

A), but then degen-

eracy may appear. For instance when

A =
1
2
p
2
and

B = p then eigenstates of

A are generally
linear combinations of two eigenstates of

B with eigenvalues p. See also the examples of d/dx
and d
2
/dx
2
above. Some operators apply to separate spaces and commute for this reason the
momentum and spin observables of a particle for instance.
Two very important special types of operator are self-adjoint operators and unitary operators.
Operator

S is self-adjoint (or Hermitian) if

S =

S

, i.e. if
|

S| = |

S|

for all states |, |, . . . in a complete set.

9
Examples: Both

I and

P
i
= |ii| are clearly self-adjoint; for (ii) with square matrices M the
property is M = M
+
, i.e. M is an Hermitian matrix. In space (i) with (0) = (1) = 0, operator
d/dx is not self-adjoint while id/dx and d
2
/dx
2
are. (Integrate by parts).
Preceding examples illustrate the following:
Theorem: The eigenvalues of a self-adjoint operator (if it has any!) are real and the eigenvec-
tors belonging to dierent eigenvalues are orthogonal.
Firstly observe that if

S|s = s|s the dual equation is s|

S = s

s|, giving two results for s|

S|s that
imply s = s

since s|s = 0. Secondly, if

S|s
i
= s
i
|s
i
for i = 1, 2 then the case i = 1 and the dual of the
case i = 2 give two calculations of s
2
|

S|s
1
which leads at once to (s
1
s
2
)s
2
|s
1
= 0 and if s
1
= s
2
orthogonality is established. In fact this applies to the whole spectrum since degenerate eigenvectors can be
orthogonalised by the Gram-Schmidt method.
A useful spectral representation of

S =

S

(of which the resolution of

I into projectors is a special case)
is found by using its (assumed) complete set of orthonormal eigenvectors. i.e. Since

S| =

S|ss| we
can write

S =

s
|sss|.
Operator

U is unitary if

U
1
=

U

. Then

U

=

U

U =

I so that if | =

U| then | = |
i.e. have the same norm; thus

U is norm preserving. If further | = c| then we see that all (complex)
eigenvalues c of

U have modulus unity.
Clearly

I is unitary; so is any operator

U() = exp(i

S) where is real and

S =

S

. This parametrisation
is useful when

U is connected continuously to the identity

I =

U(0). Then the connection between unitary
(norm-preserving) operators and self-adjoint (observable) operators is of great signicance.
2.3 Measurements: Physical Assumptions
We make two basic assumptions which connect the mathematical formulation of the theory to the observa-
tions (i.e. measurements) of physical quantities.
The rst postulate (connection with physics) states:
To every physically observable quantity i.e. an observable A corresponds a self-adjoint linear
operator

A in the state space with a complete set of eigenvectors {|a} (and conversely).
The second postulate (the measurement postulate) states:
A single measurement of A on a system in state | gives one of the (real) eigenvalues a of

A
with probability |a||
2
. After the measurement the state is |a.
Note that results of observations are real numbers, and this statement assumes for now that the spectrum
of

A is discrete and that state | and (orthogonal) eigenstates {|a;

A|a = a|a} are normalised to unity.
Otherwise |a||
2
is a relative probability.
The set of observables includes (usually) the basic degrees of freedom of the system the classical ps
and qs and quantities derived from them.
Examples: For a particle, observables include position r, linear momentum p, plus e.g. kinetic
energy T(p), potential energy V (r) , angular momentum L = r p, as well as any intrinsic
properties like spin, electric charge, etc.
The measurement postulate refers to a result of an observation merely having a probability of realisation.
Its goodbye to classical determinism results of measurements on quantum systems are generally uncertain
even when the state vector is known.
At the heart of this is the basis expansion of a state in terms of eigenstates of the observable

A: if
| =

a
c
a,
|a then the size of the coecient c
a,
= a| determines the amount of |a present. In fact
the squared moduli are the probabilities of getting the as and normalisation of the states makes their
sum equal to 1.
10
If it happens that | = |a, an eigenstate of

A, then measurement of A is certain (probability=1) to give
result a (from normalisation) and never (probability=0) gives any other outcome (from orthogonality). The
system is in a state of denite A.
However, when the system is not in a state of denitive A the measurement changes the state:
| |a,
where |a is an eigenstate of

A. The factor is there to indicate that after the measurement the vector
is not normalised. However, as the states correspond to rays - the resultant vector can be normalised by
multiplication by an appropriate .
Note that the process of measurement is very acausal; one way to think of it is to put
| =

a
c
a,
|a |a
and we have the reduction of the wave function or the collapse of the state vector onto (just one)
eigenvector of

A.
Note that as after the measurement of

A which gave the value a the state vector is an eigenstate of

A
the successive remeasurements of the same observable (i.e. of

A) will give the same answer i.e. a. This is
how the theory describes the sudden change in the observers state of knowledge and the preparation of a
system to being in a denite state.
Exactly how a non-deterministic collapse of the state vector takes place, especially when the observer
should in principle be described by quantum mechanics too, is a feature that provokes discussion.
Making a measurement of A on each of a large number of identically-prepared copies of a system
generally gives a distribution of results, for which the ordinary mean or mathematical expectation A is
given by

a
(result)
a
(probability)
a
. Each copy of the system has by denition the same state vector | so
we have:
A =

a
a |a||
2
=

a
|aaa|
and, recognising the spectral representation of

A, we obtain the important formula
A = |

A|
for an expectation value in quantum mechanics.
If A is observed on many identical systems then its expectation value is A and the standard deviation
A is a familiar measure of scatter of results dened by (A)
2
(A A)
2
A
2
A
2
. Note that
with the formula A = |

A| we have A = 0 i | is an eigenvector of

A.
Observables are simultaneously measurable with precision (compatible) if their corresponding oper-
ators commute, for as seen above, a state can then be an eigenstate of them all. The common eigenstates
of a complete set of mutually commuting observables form an orthonormal basis for the state space they
include all possible outcomes of measurement of all compatible As and physical completeness is equated
to mathematical completeness. What constitutes the former is a physical judgement and usually includes
at least those observables from which is built the systems most important operator the Hamiltonian
operator, controlling its time-dependence (as described below).
Observables that are not compatible have non-commuting operators and the degree of the mutual
interference of their measurement is given precise meaning in terms of their commutator, as shown in the
following theorem
Theorem: Let A and B be observables with self-adjoint operators

A and

B. Then for a
quantum system in state | at a given instant
AB
1
2

|[

A,

B]|

.
11
Proof: Let

A

=

A |

A|

I, similarly

B

. Then both

A

and

B

are self-adjoint (expectation values

of self-adjoint operators are real) and their commutator is identical to [

A,

B]. Therefore
|[

A,

B]| = |

| |

| = 2iIm|

|.
Note that the product of two self-adjoint operators is not self-adjoint unless they commute, so its expectation
value is not real. We now have
1
2

|[

A,

B]|

_
|(

)
2
||(

)
2
|.
The last step uses the Schwarz inequality and the square root is just AB. So the result is proved.
The conclusion is that incompatible measurements carried out simultaneously on copies of a system
always give a scatter of results. This is a manifestation of uncertainty in quantum mechanics.
Thus, after measurement of A, a measurement of quantity B on the same system will give an uncertain
result if [

A,

B] = 0, for the system cannot now be in an eigenstate of

B, even if it was originally. The
mutual interference of incompatible observations on the same system is thus clear. This is another aspect of
quantum mechanical uncertainty.
2.4 Revision of Classical Mechanics
Newtons second law for a particle of mass m at position r and experiencing force F is
dp
dt
= F,
where p = m r. Taking the vector product with r leads to introduction of angular momentum L = r p,
which is conserved for central forces. Taking the scalar product with r and integrating with respect to time t
leads to discussion of work and of energy and its conservation, and for conservative forces to the introduction
of a potential V (r) where F(r) = V . Extended systems are treated as assemblies of interacting particles.
Advanced formulations use energy. Generalised coordinates q
i
(t) in an N-dimensional abstract cong-
uration space are used to eliminate constraint forces that do no work. They coincide with coordinates in
ordinary space only in simple cases. N counts the systems degrees of freedom. The Lagrangian L is dened
in terms of kinetic energy T(q
i
, q
j
) and potential energy V as
L(q
i
, q
j
) = T V.
The N second-order Lagrangian equations of motion
d
dt
_
L
q
i
_
=
L
q
i
then follow as necessary conditions for the action
S =
_
Ldt
(which measures mean energy interchange in the motion) to be stationary against variations in each trajectory
q
i
(t) independently.
Comments:
1. The principle of least action is equivalent to Newtons second law.
2. The principle involves certain acausality. How does the particle know which way to go? Does it
try all possibilities? In fact Quantum Mechanics solves this problem. All trajectories are used but
classically we observe only those of least action. (This leads us to Feynmans path integral formulation
of Quantum Mechanics.)
12
Generalised momentum
p
i

L
q
i
conjugate to q
i
is conserved if L/q
i
= 0. Linear momentum is conjugate to linear displacement; angular
momentum is conjugate to angular displacement.
The Hamiltonian is
H(q
i
, p
j
) = L(q
i
, q
j
) +
N

k=1
p
k
q
k
with qs eliminated in favour of ps. If (as is usual) T is bilinear in qs the Hamiltonian is total energy T +V
expressed in the p, q variables. The ps and qs are independent dynamical variables; a conguration of the
system is a point in the 2N-dimensional p q phase space. Its time development is a path traced out
according to the 2N rst-order Hamilton equations of motion
p
i
=
H
q
i
, q
i
=
H
p
i
.
Observables like position, velocity, momentum, angle, angular momentum, energy etc. are constructed
from the phase-space coordinates q, p. Any observable A depends on time t through these coordinates, as
well as possibly explicitly. Then using the chain rule and Hamiltons equations we have
dA
dt
=
A
t
+ {A, H}
where the Poisson Bracket of A and H appears. This is dened for any two functions A, B of the ps and
qs as
{A, B}
N

k=1
_
A
q
k
B
p
k

A
p
k
B
q
k
_
and is an invariant of canonical transformations, i.e. changes of variables that leave the form of Hamiltons
equations intact. The Poisson Bracket has properties:
{A, B} = {B, A}
{A+B, C} = {A, C} + {B, C}
{cA, B} = c{A, B}
{AB, C} = A{B, C} + {A, C}B
{A, {B, C}} + {C, {A, B}} + {B, {C, A}} = 0.
Comments:
1. This is another equivalent formulation of Classical Mechanics.
2. Poisson Bracket of A and B has the same properties as the commutator of

A and

B.
3. The basic Poisson Bracket is
{q
j
, p
k
} =
jk
.
4. An observable A is conserved (a constant of the motion) if both A/t = 0 and {A, H} = 0. So for
instance energy is conserved if H is not explicitly t-dependent.
Example: The simple pendulum, which is a mass-point on a light inextensible string making
small oscillations in a plane about equilibrium under gravity alone, has one degree of freedom
and is an example of a (one-dimensional) simple-harmonic oscillator. If mass is m and linear
displacement is q, then kinetic energy is T =
1
2
m q
2
and potential energy is V =
1
2
m
2
q
2
. Here
is angular frequency of oscillation, related to the length of the string and the acceleration of
gravity. Conjugate momentum is p = m q and the Hamiltonian is H =
1
2
((p
2
/m) + m
2
q
2
).
Total energy is identical to H and is conserved. The classical oscillator is important because
quite general systems near equilibrium can be decoupled into normal modes undergoing simple
harmonic motions. And in quantum mechanics the quadratic Hamiltonian is one of the few
completely soluble examples.
13
2.5 Quantum Conditions
The theory gets some content by the specication of commutators of operators. This allows the computation
of their spectra and so of the actual results of measurements. This is also the place where Plancks constant
h enters.
The canonical quantisation postulate (the correspondence principle) states that for any two ob-
servables
[

A,

B] = ih{A, B},
i.e. that the commutator of two self-adjoint operators is equal to ih(the result of calculating the Poisson
Bracket of the corresponding classical quantities), where the result is interpreted as an operator. Plainly this
is not inconsistent with the properties of self-adjoint operators, commutators and Poisson Brackets. But,
most importantly, the resulting theory agrees with the experiment. It also has the correct classical limit.
An immediate application is to the degrees of freedom q and p themselves. While [ q
i
, q
j
] = [ p
i
, p
j
] = 0
we have
[ q
j
, p
k
] = ih
jk

I.
That is, a generalised coordinate and its conjugate momentum are always incompatible.
At the same time, from the result for the dispersions the Heisenberg Uncertainty Relation follows:
AB
1
2
h|{A, B}| ,
which for the p,q coordinates reads
q
j
p
k

1
2
h
jk
.
So for a quantum system like an electron or atom or molecule with position r and linear momentum
p the x-components (say) cannot be precisely determined together and many measurements on identically
prepared systems have dispersions constrained by
xp
x

1
2
h.
Plancks constant sets a fundamental limit of smallness. Note however that e.g. y and p
x
are compatible and
can be measured together exactly.
As will be elucidated, to a particle with denite momentum p corresponds a plane wave of wavelength
= 2h/p uniformly lling space and the particle is equally likely to be found anywhere. And to a particle
with denite position corresponds a uniform Fourier superposition of plane waves of all wavelengths and so
its momentum is completely indeterminate.
Also, to see things smaller than x one needs radiation of wavelength < x. This inevitably means
transfer of momentum p 2h/ > 2h/x, consistent with the Uncertainty Principle. Schi pps. 9 11
and Polkinghorne pps. 44 50 discuss the gamma-ray microscope.
Connement of a particle to a nite region inevitably means that its momentum and so its kinetic energy
is non-zero.
Direct from the quantisation rule we get for instance that [ p
j
,

A] = ih

A/ q
j
, and likewise with p
and q interchanged and opposite sign on the right. So e.g. [ p
x
, x
2
] = 2ih x, which also follows easily from
the x, p
x
commutator, using the basic properties of [

A,

B]. Indeed, often its simplest to express observables
explicitly as functions of the ps and qs (using the classical relations) and then compute commutators from
[ q, p] = ih. A warning: problems could arise if ordering of non-commuting operators cant be xed by the
need for self-adjointness, or if the system has no classical analogue.
Examples: For a particle of mass m, position r and linear momentum p, the kinetic energy
T =
1
2
p p/m and potential energy V (r) are made quantum operators by replacing components
x. . . and p
x
. . . by x. . . and p
x
. . .. Angular momentum L = r p is quantised likewise. e.g. A
free particle has

V = 0 and a one-dimensional simple harmonic oscillator of angular frequency
has

V =
1
2
m
2
x
2
. Total energy is

E =

T +

V . Then [ p,

T] = [ r,

V ] = 0 for any components and

e.g.
[ x,

E] = [ x,

T] =
ih
m
p
x
.
A free particle can be described by states |E, p of denite energy and momentum and for a given
E there is degeneracy with respect to direction of momentum p.
14
2.6 Simple Applications
Commutators determine spectra. Here are 3 cases, including discrete (quantised) spectra of two observables
(energy, angular momentum) that classically take continuous values. Plancks constant sets the scale of
discreteness.
Each calculation incorporates a similar logical step, namely if

A|? = a|? then |? c|a, where |a is
the eigenvector of

A belonging to eigenvalue a, and c could be zero. The possibility of degeneracy of

As
spectrum is ignored in these simple cases because its irrelevant. In each example theres no other observable
or operator that can distinguish degenerate states.
2.6.1 Position and Momentum
Suppose [ x, p] = ih

I where x = x

, p = p

and x|x = x|x, and let

|? = (

I
i
h
p)|x
for real where || 1. Then
x|? = (

I x
i
h
x p)|x = ((

I
i
h
p) x +

I)|x,
using the x, p commutator. So we have
x|? = (x + )|? + O(
2
)
for any small number , implying that (no degeneracy!)
|? = c|x +
to rst order. Now to the same approximation |c| = 1, since
?|? = x|(

I +
i
h
p)(

I
i
h
p)|x = x|x + O(
2
),
and we choose c = 1. The normalisation of states |x is not specied.
An arbitrary nite displacement x x+ can be composed of a large number n of small displacements
/n:
|x + (I
i
h

n
p)
n
|x,
which in the limit n gives the unitary transformation
|x + = exp(
i
h
p)|x.
Thus, as is arbitrary, without any other constraints we have that the spectrum of a conguration-
space position operator is continuous and that unitary transformations |x |x + are generated by the
canonically conjugate self-adjoint momentum p. Translations in y and z are generated by p
y
and p
z
and
commute with those in x since e.g. [ p
x
, p
y
] = 0.
Interchanging position and momentum gives independent translations in components of momentum:
|p +k = exp(
ik
h
x)|p
where k is arbitrary unless other conditions apply.
Comments:
1. The sign is dierent due to the asymmetry of the commutator.
15
2. Similar results hold for any canonically conjugate q, p pair where [ q, p] = ih. In particular continuous
angular displacements (rotations) about an axis are generated by the conjugate angular momentum.
(Replace x by an angle and p by the operator for the component of angular momentum about the
rotation axis).
3. However, the fundamental fact that rotations about dierent axes in three dimensions dont commute
implies that components of angular momentum in quantum mechanics dont commute. Then, because
of these extra conditions, continuous changes in angular momentum cannot be generated by an angular
position operator.
2.6.2 Angular Momentum
Classically L = r p and quantising by writing each Cartesian component of r and p as a self-adjoint
operator we have

L
j
=
jkl
r
k
p
l
,
i.e.

L
x
= y p
z
z p
y
and so on. Now

L
j
=

L

j
since [ r
k
, p
l
] = 0 for k = l; there is no operator-ordering
ambiguity. With the commutation rules for position and linear momentum or direct from Poisson brackets
we nd
[ r
2
,

L
j
] = [ p
2
,

L
j
] = 0,
consistent with lengths of vectors being unaected by rotations. At the same time we nd the fundamental
commutator for angular momentum operators:
[

L
j
,

L
k
] = ih
jkl

L
l
,
or [

L
x
,

L
y
] = ih

L
z
and so on.
Denition: any self-adjoint operators,

J
1,2,3
say, that obey these commutators are called angular mo-
mentum operators. Then as
[

J
2
,

J
k
] = 0
where

J
2
=

J
2
1
+

J
2
2
+

J
2
3
= (

J
2
)

the length of J and just one of its components (say

J
z


J
3
) can be
sharply dened together. This is in contrast to classical mechanics where all components of the vector can
be determined completely.
So there is a set of orthonormal states |a, b where simultaneously

J
2
|a, b = a|a, b

J
z
|a, b = b|a, b
for real a,b. To nd them, rst observe that a b
2
0, because a b
2
= a, b|(

J
2

J
2
z
)|a, b and this is a
sum of terms of the form
_
_
_

A|
_
_
_
2
, where

A

J
x,y
=

A

.
To deal with

J
2
x
+

J
2
y
dene

=

J
x
i

J
y
= (

,
and calculate

=

J
2
x
+

J
2
y
i[

J
x
,

J
y
] =

J
2

J
2
z
h

J
z
.
So
a, b|

|a, b = a b
2
bh 0
since this is
_
_
_

|a, b
_
_
_
2
. Now since
[

J
z
,

J

] = [

J
z
,

J
x
] i[

J
z
,

J
y
] = h

we nd that

J
z
(

|a, b) = (

J
z
and h

)|a, b = (b h)(

|a, b)
16
and conclude that either

J

|a, b = 0 or

J

|a, b = c

|a, b h where from above |c

|
2
= a b
2
bh. Since
[

J
2
,

J

] = 0 the eigenvalue a is unaected by

J

and keeps an upper limit on b

2
.
Thus from a given |a, b with b
2
a successive applications of

J
+
yield eigenvectors of

J
z
with b increasing
in steps of h until b = b

where

J
+
|a, b

= 0, when c
+
= 0 and a = b

(b

+ h). Likewise, successive applications

of

J

step down in b until c

= 0, or b = b

where a = b

(b

h). Clearly (b

)/h = n, an integer, so
solving:
b

= b

=
n
2
h and a =
n
2
(
n
2
+ 1) h
2
.
The spectra are discrete; length and z-projection of angular momentum are quantised in units of h.
Introduce angular momentum quantum numbers j and m as labels for the state |j, m where

J
2
|j, m = j(j + 1) h
2
|j, m and

J
z
|j, m = mh|j, m
for j = n/2 = 0,
1
2
, 1,
3
2
, 2, . . . with m = j, j + 1, j + 2, . . . j 1, j; i.e. (2j + 1) values and where
j, m|j

, m

=
jj

mm
. The usual phase choice is c

real and positive, giving for the ladder operators

|j, m =
_
(j m)(j m+ 1) h|j, m 1.
Since

J
x
= (

J
+
+

J

)/2 and

J
y
= (

J
+

)/2i, although the x- and y-components of angular momentum

are not sharply dened in an eigenstate |j, m we still have J
x,y
= 0.
Comments:
1. For ordinary (orbital) angular momentum of a particle L = r p only integer values of l j are
realised; all possibilities are encountered for spin angular momentum, an internal degree of freedom
of a quantum particle.
2.
a. For j =
1
2
two values m =
1
2
are possible; the a.m. vector of length
1
2

3h is either nearly
parallel (m =
1
2
) or antiparallel (m =
1
2
) to the z-axis.
b. For j = 1 the values m = 1, 0, 1 correspond to a vector of length

2h nearly parallel,
perpendicular, nearly antiparallel, to the z-axis. And so on. In each case the projection on the
x y plane is not simultaneously measurable but its average is zero.
2.6.3 Energy of a Simple-Harmonic Oscillator (one dimension)
For a particle of mass m executing SHM on the x-axis about the origin with angular frequency we have
the Hamiltonian (which is the same as the total energy operator)

H =
1
2m
p
2
+
m
2
2
x
2
=

H

where [ x, p] = ih and x = x

and p = p

, and we wish to solve

H|E = E|E. Firstly we have E = E

and, because

H is a sum of squares of self-adjoint operators, E 0. For the previous sum of squares a
factorisation into what proved to be ladder operators did the trick. So dene
a = ( p +im x)/

2m, a

= ( p im x)/

2m.
Then
aa

=

H +
i
2
[ x, p] =

H
1
2
h

I a

a =

H +
1
2
h

I.
Thus

Ha = (aa

+
1
2
h)a = a(a

a +
1
2
h) = a(

H + h),
or [

H, a] = ha. Likewise

Ha

= a

H ha

.
17
The similarity with the J
z
, J

commutator is clear and indeed

H(a|E) = (a

H + ha)|E = (E + h)(a|E),
so unless a|E = 0 we infer that
a|E = c
+
|E + h.
Similarly either a

|E = 0 or
a

|E = c

|E h.
Using the formulae for aa

and a

a and with normalised states |E we have

|c
+
|
2
= E|a

a|E = E +
1
2
h 0,
|c

|
2
= E|aa

|E = E
1
2
h 0.
So from given |E

with E = E

0 successive uses of of a

step E down by h at a time until E = E

0
where a

|E
0
= 0. Evidently this happens at E
0
=
1
2
h, when c

= 0. To avoid negative |c

| only starting
values E

=
1
2
h,
3
2
h,
5
2
h, . . . are possible. Clearly a steps up this sequence indenitely.
Summary: The one-dimensional simple-harmonic oscillator has a discrete energy spectrum, with eigen-
states of

H labelled by quantum number n:
{|n; n|n

=
nn
}
and eigenvalues
E
n
= (n +
1
2
) h,
where

H =
1
2
(aa

+a

a),
a|n =
_
(n + 1) h|n + 1, a

|n =

nh|n 1
for n = 0, 1, 2, 3, . . .. The ground state |0 obeys a

|0 = 0 and
|n =
1
_
n! ( h)
n
(a)
n
|0.
Observables p and x are proportional to a a

where the annihilation and creation operators obey

[a

, a] = h

I.
Comments:
The quantum oscillator is never at rest for its ground-state energy has zero-point value
1
2
h.
Note the appearance of a relation of form Energy = h(angular frequency), which has general
signicance in discussion of time-dependence.
3 Representation Theory
3.1 Position Representation
3.1.1 Wave function
Consider a system which classically has N degrees of freedom. Then q
1
, .. q
N
form a complete set of commuting
operators and we have
[ q
i
, q
j
] = 0, [ p
i
, p
j
] = 0, [ q
i
, p
j
] = ih
ij
.
18
Hence we can take as a basis in the space of states of the system the basis formed by all eigenstates of
q
i
i.e.
|q
1
, ..q
N
= |q,
where
q
i
|q = q
i
|q
As we already know these eigenvalues are continuous; they generalise the concept of position of a particle.
To see this consider one particle and the self-adjoint operator r that measures its position components as
generalised variables, coinciding with its coordinates in ordinary three-dimensional space. An eigenstate |r
of r is labelled by the continuously-variable real vector eigenvalue r. Orthonormality must be generalised
from the discrete case that uses the Kronecker delta, and completeness (resolution of the identity) from a
simple sum over eigenvalues.
Proceed by generalising the completeness relation. Dene

I =
_
d
3
r |rr|,
where d
3
r stands for dxdy dz and the integral extends over all space. Then the basis expansion is
| =
_
d
3
r|rr|
and from considering r

| we nd
r

|r = (r

r).
So the presence of operators with continuous spectra necessarily leads to state vectors with delta-function
normalisation, i.e. orthogonal but with a certain type of innite norm. We will see later that this is equally
so if the basis of r is replaced by another continuous basis, e.g. momentum.
In our N dimensional space we have as normalisation
q

|q = q

1
...q

N
|q
1
...q
N
=

i=1,N
(q

i
q
i
) =
N
(q

q)
and in this basis
1 =
_
d
N
q|qq|.
Moreover,
| =
_
d
N
q|qq|
The representative of | is the complex number q| that depends continuously on the real (N dimen-
sional) vector q. This is a complex function of q, which we can denote by wavefunction

(q). This function

is called the wavefunction of the state | in the basis {|q}. Note that in many texts and we adopt this
convention too this function is denoted, simply, as (q).
The inner product is
| =
_
d
N
q

(q)

(q)
and for normalisable states | the wavefunction is square-integrable:
| =
_
d
N
q |

(q)|
2
= 1.
This implies that the wavefunction decreases faster than |q|
N/2
as |q| .
When labels , , . . . are themselves continuous eigenvalues then evidently the eigenfunctions are not
square-integrable but:
|

=
_
d
N
q

(q)

(q) = C()(

),
19
if

I =
_
dC
1
||. The converse implication in fact holds too: eigenfunctions that are not square-
integrable but have -function norm imply continuous eigenvalues (and normalisable eigenfunctions imply a
discrete spectrum).
Comment:
Note that
|

(q)
denes an isomorphism between the abstract Hilbert space of states | and L
2
(R
N
), the Hilbert
space of functions, square integrable on R
N
.
How does any operator

( q) act on

(q)? Clearly

( q)|q = (q)|q.
So

( q)| =

( q)
_
d
N
q

(q)|q =
_
d
N
q

(q)(q)|q
so the wavefunction of

( q)| is ( q)

(q) and so the action of

( q) is the multiplication by (q).
Denition: The basis {|q} is said to be a position representation of the Hilbert space of states in
which

(q) is the wavefunction of |.

3.1.2 Change of basis
Note that the equation
q
i
|q = q
i
|q
denes the eigenstates up to a factor; normalisation condition denes the eigenvectors up to a factor of
modulus 1. Thus we can replace the basis {|q} by a new basis
|q, = e
i(q)
|q.
In this basis | will be represented by

(q), where

(q) = q, | = e
i(q)

(q)
and
| =
_
d
N
q

(q)|q, .
3.1.3 Finding the action of the p operator (setting up the Schr odinger representation)
Next we have to decide how the operators p act in the space of wave functions. We will argue that we can
represent their action by
p
i
= ih

q
i
in the sense that
p
i
(q) = ih

q
i
(q).
First dene

i
|
_
d
N
q

q
i
(q)|q
and then we will show that ih

i
behaves as p
i
. Note that here we are using the convention that (q) =
q|.
20
Next we calculate
|

i
| =
_
d
N
q

(q)

qi
(q) =
=
_
d
N
q

qi
[

(q)(q)]
_
d
N
q (q)

qi

(q) =
=
_
d
N1
q

(q)(q) |

__
d
N
q

(q)

qi
(q)
_

,
where in going from the rst line to the second we have integrated by parts. Assuming that the functions
vanish at innity the rst term in the last line vanishes and we have
|

i
| = (|

i
|)

and so
| ih

i
| = (| ih

i
|)

and so we see that ih

i
is Hermitian in the space of functions which vanish at innity.
Next we consider the commutator [ih

i
, ih

j
]. We have
(ih

i
)(ih

j
)| = h
2
_
d
N
q

2

qiqj
|q = (ih

j
)(ih

i
)|
thus showing that
[ih

i
, ih

j
] = 0 = [ p
i
, p
j
].
Next we consider [ih

i
, q
j
] or, more generally, [ih

i
, ( q)]. We have
(ih

i
) ( q)| = ih

i
( q)
_
d
N
q (q)|q
= ih

i
_
d
N
q(q)(q)|q = ih
_
d
N
q

qi
{(q)(q)} |q
= ih
_
d
N
q[(q)

qi
+

qi
(q)]|q = [ih ( q)

i
ih

qi
( q)]|.
So we see that
[ih

i
, ( q)] = ih

q
i
( q).
In particular
[ q
i
, ih

j
] = ih
ij
= [ q
i
, p
j
]
and so we see that
[ q
i
, p
j
+ih

j
] = 0.
Thus p
j
+ ih

j
commutes with a complete commuting set of observables ( q
i
) and so must be a function of
them. So
p
j
+ih

j
= Z
j
( q).
However, as [ p
i
, p
j
] = 0 we have
[Z
i
( q) ih

i
, Z
j
( q) ih

j
] = [Z
i
, Z
j
] + [ih

i
, Z
j
] [ih

j
, Z
i
] + [ih

i
, ih

j
] = 0.
Thus
0 = [ih

i
, Z
j
] [ih

j
, Z
i
] = ih
_
Zj
qi

Zi
qj
_
and so we see that
Z
i
=
Z
qi
21
for some Z. Thus
p
i
= ih

i
+
Z
qi
.
Finally, we now change the basis and use [|q, } and

(q). This leads to a new

i
which we denote by

. Its action is given by

|
_
d
N
q

qi
|q, =
_
d
N
q

qi
_
e
i

_
e
i
|q
=
_
d
N
q
_

qi
i

qi

_
|q =
_

i
i

qi
_
|.
So choosing =
Z
h
we have p
i
= ih

. Thus, dropping we see that we can take

p
i
= ih

i
.
The representation in which this is the case is called the Schr odinger representation and (q) is called
the Schr odinger wave function.
3.1.4 Example
Operators that are functions of p are dierential operators in wave mechanics; e.g. kinetic energy of a particle
in 3 dimensions

T = p
2
/2m becomes ( h
2
/2m)
2
and its angular momentum operator is ihr. Matrix
elements are calculated as
|

A( q, p)| =
_
d
3
q

(q)A(q, ih)

(q).
As an example let us consider a simple harmonic oscillator in 1 dimension.
With energy eigenfunctions
n
(x) x|n, n = 0, 1, 2, . . . and

H|
n
= E
n
|
n
the dierential equation

h
2
2m
d
2

n
dx
2
+
1
2
m
2
x
2

n
= E
n

n
, x|

H|
n
= E
n
x|
n

determines eigenvalues E
n
= (n +
1
2
) h if the solution is required to be normalisable i.e. to vanish as
|x| . The textbook procedure of series solution (Schi, p. 66 and see later) gives
n
as the product of
a factor exp(
1
2
(x/x
0
)
2
) and a Hermite polynomial in x/x
0
, where x
2
0
h/m (x
0
is the amplitude of a
classical oscillator of energy
1
2
h). In fact instead we can nd the energy eigenfunctions starting from the
denition of the ground-state, a

|0 = 0, which for the wavefunction

0
(x) x|0 is the dierential equation
h
d
0
dx
+mx
0
= 0.
The solution, normalised using
_

exp(x
2
) dx =

, is easily found:

0
(x) = (
m
h
)
1/4
exp(
1
2
mx
2
/h) =
1/4
x
1/2
0
exp(
1
2
(x/x
0
)
2
).
Then n applications of a = (ihd/dx +imx)/

2m gives the (normalised) eigenfunction

x|n
n
(x) =
1

n!
_
i

2
_
n
_
x
0
d
dx

x
x
0
_
n

0
(x),
showing the origin of the Hermite polynomial. Also note that
n
(x) = (1)
n

n
(x).
22
3.2 Probabilistic Interpretation
Let us consider rst one particle in 3 dimensions and as q use Cartesian coordinates x, y, z. The expectation
value of observable A in state |(t) of a particle is
A = (t)|

A|(t) =
_
dV

(r, t)

A(r, ih)

(r, t),
where dV = dxdy dz. For a quantity A independent of momentum is
A =
_
dV

A(r) ||
2
.
Recalling the formalism of elementary probability theory makes clear the standard interpretation of the
wavefunction, namely that
|

|
2
V |r|(t)|
2
V
is the relative probability of a measurement showing the particle to be in volume V = xy z at position r
at time t.
For a system with N degrees of freedom and generalised coordinates
q = (q
1
, . . . q
N
),
which by virtue of their commutators with canonically-conjugate variables have continuous spectra, |q||
2
dq
N
is interpreted as a relative probability of observing the system in a state described by a point in a small
volume d
N
q at q in the abstract N-dimensional conguration space.
Compare with the discrete case where |a|(t)|
2
is the relative probability of a measurement of A at
time t giving as outcome the eigenvalue a.
So for a single particle in a potential (r, t) ||
2
can be viewed as a density of the particle in ordinary
space.
3.3 Harmonic Oscillator Revisited
Previously we discussed the eigenstates of the harmonic oscillator using algebraic methods. Now we discuss
this problem using the wave functions and then show the relation between the two approaches.
V =
1
2
m
2
x
2
hence

H =
1
2
p
2
m
+
1
2
m
2
x
2
.
So, the wavefunction satises

h
2
2m
d
2
(x)
dx
2
+
1
2
m
2
x
2
(x) = E(x).
To get bound states (ie normalisable states) we require that (x) 0 as x (as V ).
To do this we rewrite
(
h
m
d
2
dx
2

m
h
x
2
) =
2E
h

and change the variable x y =

_
m
h
x and introduce =
2E
h
. Then our equation becomes
d
2

dy
2
+ ( y
2
) = 0.
When y the equation becomes
d
2

dy
2
y
2
= 0 and its solutions are
e

1
2
y
2
to order O(
1
y
2
).
23
So to satisfy our boundary condition we take e

1
2
y
2
. Thus we put
= e

1
2
y
2
H(y)
and derive the equation for H(y). As
d
dy
= e

1
2
y
2

H ye

1
2
y
2
H we see that
d
2

dy
2
= e

1
2
y
2

H 2ye

1
2
y
2

H + (y
2
1)e

1
2
y
2
H
and we nd that H satises
d
2
H
dy
2
2y
dH
dy
+ ( 1)H = 0
This equation (Hermites equation) appears harder to solve. We seek its solutions by a power series
expansion around y = 0. We put
H = y

r=0
a
r
y
r
,
where a
0
= 0 and then we try to determine and all the coecients a
r
r = 0. The left hand side of Hermites
equation is now

r=0
a
r
y

[(r + )(r + 1)y

r2
2(r + )y
r
+ ( 1)y
r
]
= y
2
a
0
( 1) +y
1
a
1
( + 1) +

r=0
y
r+
[(r + + 2)(r + + 1)a
r+2
(2r + 2 + 1 )a
r
] = 0
So comparing powers of y we see that we have to require
a
0
( 1) = 0
a
1
( + 1) = 0
( +r + 2)( + r + 1)a
r+2
= [2r + 2 + 1 ]a
r
.
As a
0
= 0 we have either = 0 or = 1. If = 0 a
1
is arbitrary, but if = 1 then a
1
= 0. So, we take
= 0 and have a
0
(times a function of y
2
) + a
1
(times an odd function of y).
Note that the behaviour of H(y) as y is determined by
a
r+2
a
r

2
r
as r ,
which, unless series terminates, gives
H(y) y

1
r!
y
2r
y

e
y
2
thus leading to a non-normalisable which goes as y

e
1
2
y
2
. Hence to have a normalisable wave
function we have to require that the series terminates; i.e. that
a
2r+2
= 0
for some r = m, say. Then
(2(m+ ) + 1 = 0 i.e. = 2(m+ ) + 1.
Then H(y) is a polynomial of degree m+ and parity (1)

.
Call m+ = n. This polynomial is H
n
(y) - the Hermite polynomial.
Thus

n
=
n
(y) = H
n
(y)e

1
2
y
2
y
n
e

1
2
y
2
as y . The corresponding energy eigenvalue is = 2n + 1 which gives us
E
n
=
h
2
= (n +
1
2
) h.
Comments
24
As
m
correspond to dierent eigenvalues they are orthogonal.
_

m
(y)
n
(y)dy =
_

e
y
2
H
m
(y)H
n
(y)dy = K
n

nm
.
The normalisation constant K
n
is chosen so that
_

2
n
dx = 1 i.e.
n
=
1

1
4
1

n!2
n
H
n
(y)e

1
2
y
2
.
3.4 Relation to the Algebraic Approach
In our algebraic approach we used
a = ( p +im x)
1

2m
a

= ( p im x)
1

2m
and |0 was the ground energy state, i.e. it satised a

|0 = 0. In the Schr odinger representation

a

= ( p im x)
1

2m
= (ih

x
imx)
1

2m
= i( h

x
+mx)
1

2m
= i
_
h
2
(y +

y
).
So the Schr odinger wave function for the lowest state
y|0 =
0
(y) satises i
_
h
2
_
y +

y
_

0
(y) = 0
and so is given by Ae

1
2
y
2
.
Higher states |n (a)
n
|0 and so (..)
n
(y

y
)
n

0
(y) and so

n
(y) A
n
_
y

y
_
n

0
(y).
But note that
e
1
2
y
2
y
_
e

1
2
y
2
f
_
=
f
y
yf =
_

y
y
_
f
and so we have
y

y
= e
1
2
y
2
(

y
)e

1
2
y
2
(y

y
)
n
= e
1
2
y
2
(

y
)
n
e

1
2
y
2
.
So

n
const e
1
2
y
2
(

y
)
n
e
y
2
= const e

1
2
y
2
H
n
(y)
as H
n
(y) = e
y
2
(

y
)
n
e
y
2
.
The last result can be derived from the generating function for H
n
(y). Recall that this function is given
by
e

2
+2
=

n=0
H
n
()
n!

n
.
But as e

2
+2
= e

2
()
2
we see that H
n
() is related to the coecient of
n
in the expansion of e
()
2
.
This coecient is given by
1
n!
(

)
n
e
()
2
|
=0
=
1
n!
(

)
n
e

2
25
and so
H
n
() = e

2
(

)
n
e

2
as required.
So the two methods give the same spectrum of the Hamiltonian.
Comments:
This time, as we saw earlier, we have
1
2
h as the zero point energy.
In many theories we can treat elds as sets of harmonic oscillators (when resolved into normal modes).
Then each mode has its own zero point energy.
3.5 Momentum Representation
The commutation relations
[ q
i
, q
j
] = 0 = [ p
i
, p
j
], [ q
i
, p
j
] = ih
ij
have a symmetry
q
i
p
i
, p
j
q
j
(corresponding to a canonical transformation in phase space) and so we can take a representation in which
p
i
acts by a multiplication and q
i
acts as ih

pi
. This corresponds to choosing as our basis momentum
eigenstates {|p
i
} with
| =
_
d
N
p

(p)|p.
Then
q
i
| = ih
_
d
N
p

pi
|p.
Note that the function

(p) is the momentum representation analogue of the wave function, called momentum
wave function.
What is q|p, i.e. an eigenstate on p in the position representation {|q}?
To answer this question let us restrict our attention to one dimension. Then we have q and p which
satisfy
q p p q = ih.
Then we observe that
q| q p|q

q| p q|q

= ihq|q

= ih(q q

).
Thus
(q q

)q| p|q

= ih(q q

).
However we can use the following property of the Dirac delta function:
x

(x) = (x), x(x) = 0

and so we see that
q| p|q

=
h
i

(q q

)
and
pq|p =
_
dq

q| p|q

|p =
h
i
_
dq

q
(q q

)q

|p =
h
i
_
dq

q
(q q

)q

|p =
h
i

q
(q|p)
and so
q|p = Ae
ipq
h
,
where A is a normalisation factor.
26
Note that q|p is a wave function of a state with a well dened momentum. Naively, we may expect this
function, like any wavefunction to be normalised i.e. to satisfy
_
dq(q)

(q) = 1
but from the above we get
|A|
2
_
dq e
ipq
h
e
ipq
h
= |A|
2
_
dq = 1
and so we see that |A|
2
=
1
V
where V is the range of integration (i.e. ).
Thus we have run into the well known problem of normalisation. There are three ways of dealing with
this problem:
consider only wave packets (i.e. always have some spread in momentum)
normalise in a box of nite dimensions and later take these dimensions to .
practical - not worry about normalisation (but keep it at the back of our mind).
Returning to the case of N degrees of freedom we see that in the momentum representation
| =
_
d
N
p |pp| =
_
d
N
p |p

(p)
=
_
d
N
q |qq| =
_
d
N
q |q(q).
Thus
(q) = q| =
_
d
N
p q|pp|
=
_
d
N
p q|p

(p).
So
(q) = A
_
d
N
p e
iqp
h
(p)
and

(p) = A
_
d
N
q e
iqp
h
(q).
Note that if we choose
A =
_
1

2h
_
N
we have
| =
_
d
N
q |(q)|
2
=
_
d
N
p

(p)

2
Then the interpretation of

(p) is analogous to the interpretation of (q) and we see that going from
the momentum to the position representation wave functions involves taking Fourier transforms.
Comments:
1. Note that (q) =
_
d
N
p e
1qp
h

(p) has a meaning of a superposition of plane waves of momentum p
with weights provided by the function

(p). Thus we have a wave packet.
2. If

(p) = B(p p
0
) then the momentum of the system is not uniquely dened and we have a
spread in momentum.
27
4 Equation of Motion - Dynamics
4.1 Introduction
In classical mechanics, as we have said earlier, the time evolution (the dynamics) can be described in terms
of Hamiltons equations
q
i
=
H
p
i
, p
i
=
H
q
i
or, equivalently, in terms of Poissons brackets
dA
dt
=
A
t
+ {A, H}
etc. In Quantum Mechanics there are, similarly, also several equivalent ways of describing the dynamics.
Such descriptions are called pictures.
4.2 The Schrodinger picture
The theory involves both state vectors and certain operators that extract the information they encode. As
times progresses either or both could change. It is perhaps more natural to associate this change with the
the state vector. The observables correspond to xed operators - although the results of their action on the
state depend on time as the state changes with time. This way of describing the development of the system
is called the Schr odinger picture. Then as long as the system is left undisturbed it evolves causally and
its evolution is described by the equation of motion.
But warning; when any measurement is performed the system is perturbed and during this measurement
its change is acausal.
So what is this causal evolution?
Let us assume that at time t the state is |
t
and at t = t
0
it is |
t0
. We expect that there is an unitary
operator

U(t, t
0
) which connects these two states of the system. i.e. which acts as
|
t
=

U(t, t
0
)|
t0
.
Why? In classical mechanics the time evolution can be thought of as a continuous unfolding of canonical
transformations which preserve Hamiltons equations and Poissons brackets. So we may even expect that
the operator U has something to do with the Hamiltonian of the system.
Let us assume that the operator

U(t, t
0
) is linear, i.e. that it satises
|
t
+ |
t
=

U(t, t
0
) [|
t0
+ |
t0
]
and that the operator

U is independent of the state vector |. Let us assume further that

t
|
t
=
t0
|
t0
,
although, this may appear less clearly motivated (this is equivalent to the statement that

U is unitary, i.e.

U =

1) and

U(t
1
, t
2
)

U(t
2
, t
3
) =

U(t
1
, t
3
).
This last requirement tells us that as

U(t
0
, t
1
)

U(t
1
, t
0
) =

1,

U(t
0
, t
1
) =

U
1
(t
1
, t
0
) and so that

(t
0
, t
1
) =

U
1
(t
0
, t
1
) =

U(t
1
, t
0
).
Let us now derive the equation of motion in a dierential form. To do this dene
d
dt
|
t
= lim
0
1

(|
t+
|
t
) = lim
0
1

U(t + , t
0
)

U(t, t
0
)
_
|
t0

=
d

U(t, t
0
)
dt
|
t0
=
d

U(t, t
0
)
dt

(t, t
0
)|
t
.
28
So
d
dt
=
d

U(t, t
0
)
dt

(t, t
0
).
Dening U
t


U(t, t
0
) we see that
d
dt
_
U
t
U

t
_
= 0
gives
dU
t
dt
U

t
+U
t
dU

t
dt
= 0
which shows that the operator
dU
t
dt
U

t
is anti-Hermitian. So putting in ih we have
_
ih
dU
t
dt
U

t
_
= ih
dU
t
dt
U

t


H(t),
where

H is a Hermitian operator. Thus our equation becomes
ih
d
dt
|
t
=

H(t)|
t
.
This equation, called the Schr odinger equation species the dynamics in the Schr odinger picture.
What is

H(t)? This will become clearer after we have introduced an alternative formulation of the
dynamics - in terms of the Heisenberg picture.
At the moment let us note that if

H =

H(t), i.e.

H=const then we can solve
ih
dU
t
dt
U

t
=

H
as then
ih
dU
t
dt
=

HU
t
and so we see that
U
t
=

U(t, t
0
) = exp
_

H(t t
0
)
h
_
where we have determined the constant of integration from

U(t
0
, t
0
) =

1.
4.3 The Heisenberg Picture
In the Schr odinger picture we had
|
t
=

U(t, t
0
)|
t0

and all the operators were xed

A. However, we can perform an unitary transformation
|
t


U(t
0
, t)|
t
= |
t0
.
and transform the operators

A

A
t
=

U(t
0
, t)

U(t, t
0
)
and then we will have an equivalent description of the dynamics of the system as all matrix elements

t
|

A|
t
=
t0
|

A
t
|
t0

are unchanged (we have dened the transformation of

A so that this is the case!).
29
This way we have a second description of the dynamics - the Heisenberg picture, in which the states
are xed (coinciding with the states in the Schrodinger picture at t = t
0
, i.e. |
t0
) but all observables are
represented by time dependent operators

A
t
.
What is the equation of motion? Now this is the equation governing the time evolution of

A
t
.
To nd it let us assume, for simplicity, that

H =

H(t). Then

U(t, t
0
) = exp
_

i H(tt0)
h
_
and
ih
d

A
t
dt
= ih
d
dt
_
e
i H(tt
0
)
h
Ae
i H(tt
0
)
h
_
=

H
_
e
i H(tt
0
)
h
Ae
i H(tt
0
)
h
_
+
_
e
i H(tt
0
)
h
Ae
i H(tt
0
)
h
_

H = [

A
t
,

H].
This equation
ih
d

A
t
dt
= [

A
t
,

H]
is called the Heisenbergs equation of motion; or the equation of motion in Heisenbergs picture.
By comparing it with
dA
dt
= {A, H}
in classical mechanics and the quantisation condition
{A, H}
[

A,

H]
ih
we see that it is natural to identify

H with the Hamiltonian.
Notes:
1 If

H =

H(t) then
ih
d
dt

A
t
= [

A
t
,

H
t
(t)]
where

H
t
(t) =

U

(t, t
0
)

H(t)

U(t, t
0
) i.e. has the same expression as

A
t
.
2 If

A =

A(t) then
ih
d

A
t
dt
= [

A
t
,

H
t
] + ih

t

A
t
,
where

t

A
t


U

(t, t
0
)
d A
dt
(t)

U(t, t
0
). Note that H
t
= H if H = H(t).
4.4 Conserved quantities; constants of motion
So which picture is the most convenient? In practice, it is the Schr odinger picture. However, the Heisenberg
picture is very useful as it tells us how to nd constants of motion. To see this note that for any operator

A
its Heisenberg equation of motion is
ih
d
dt

A
t
= ih

t

A
t
+ [

A
t
,

H
t
].
So if

t

A
t
= 0, i.e.

A =

A(p, q) only, and if
[

A
t
,

H
t
] = 0, then

A
t
= const.
However,
[

A
t
,

H
t
] = 0,

[

A,

H] = 0
and so we see that this result is true in any picture. Thus, the observables, which in the Schrodinger picture
do not depend explicitly on time are conserved if they commute with the Hamiltonian. The constants of
motion are represented by time independent observables which commute with the Hamiltonian.
30
4.5 Stationary states. Time independent Schr odinger equation
4.5.1 Stationary states
Consider
ih
d
dt
|(t) =

H|(t).
Then, if

H =

H(t)
|(t) = exp(i
t
h

H)|(0).
Next insert

I =

E
|EE| (the identity resolved into energy eigenstates) immediately to the right of
exp(i
t
h

H) giving
|(t) =

E
E|(0) exp(i
Et
h
)|E.
Time dependence is exhibited in a phase factor for each component and is harmonic with angular frequency
E/h.
Note that if |(0) = |E, where

H|E = E|E
i.e. the system starts in an energy eigenstate, then
|(t) = |E(t) = exp(i
Et
h
)|E
and the energy-value remains sharp always E = 0. Also for any

A
E(t)|[

A,

H]|E(t) = 0
and so for all observables A where

A/t = 0 we have A = constant. An energy eigenstate exp(i
Et
h
)|E
or just simply |E is called a stationary state. An isolated atom in a stationary state is stable. However,
isolation is an idealisation and real atoms in excited states (energy eigenstates above the ground state)
usually decay quickly.
Thus in general, we have all terms in the expression above. If the Hamiltonian

H has both continuous
and discrete parts of the spectrum, then
|(0) =

i
a
i
|E
i
+
_
dE

a(E

)|E

and so
|(t) =

i
a
i
exp(i
E
i
t
h
)|E
i
+
_
dE

a(E

) exp(i
E

t
h
)|E

.
We see that the problem of determining the dynamics reduces to that of nding solutions of the eigenvalue
equation for

H i.e.

H|E = E|E.
This equation is called the time independent Schrodinger equation.
4.5.2 Schr odinger wave equation
Consider one particle of mass m in a potential eld V (r). The Schrodinger Equation
ih
d
dt
|
t
=

H|
t

31
where

H = p
2
/2m+

V (r) becomes the Schr odinger Wave Equation

h
2
2m

2
+V (r) = ih

t
for the wavefunction (r, t). We can think of this equation as having come from the equation above by
acting on it, from the left, with the bra vector r|. Then exploiting the fact that the potential V is a function
of r only we get the Schr odinger Wave Equation. Its solution, using the energy eigenbasis
|(t) =

E
E|(0) exp(i
Et
h
)|E,
becomes the separation-of-variables solution
(r, t) =

E
c
E
exp(i
Et
h
)
E
(r)
where
E
(r) = r|E is a normalised energy eigenfunction obeying

h
2
2m

2
+V (r) = E
(often called the time-independent Schr odinger Wave Equation this was actually the rst form guessed
by Schr odinger, in June 1926). Here c
E
= E|(0 =
_
dV

E
(r)(r, 0). The eigenfunctions
E
also obey
the eigenvalue dierential equations corresponding to mutually compatible conserved quantities such as
momentum or angular momentum.
If the spectrum of

H is continuous (or has a continuous sector) then the

E
becomes an integral over
(part of) E and the energy eigenfunctions have Dirac -function normalisation. Otherwise the solutions are
square-integrable and and vanish quickly at large |r|.
4.5.3 Example - Free particle
For a particle experiencing no force we dene V (r) = 0, (adding a constant to

H adds only an overall
phase to solutions of the Schr odinger Equation). A free particle of mass m then has

H = p
2
/2m. Since
[

H, p] = 0, momentum p is conserved and eigenstates |E, p of p are eigenstates of

H with eigenvalues
related by E = p p/2m. Note that angular momentum is conserved too, [

H,

L] = 0, but L and P are

incompatible, [

P,

L] = 0.
In wave mechanics a momentum eigenfunction
p
(r) obeys
ih = p,
which, as we have said earlier, is satised by
=
p
(r) r|p = C exp(
i
h
p r).
Here C is a constant and in the absence of other constraints components of p take any (real) values.
With continuous p the wavefunctions are unnormalisable with inner product
_
d
3
r

p
(r)
p
(r) = |C|
2
_
dxdy dz exp(
i
h
(p

p) r) = |C|
2
(2h)
3
(p

p).
Choose C = 1 so that
p|p

= (2h)
3
(p p

) and

I =
_
d
3
p
(2h)
3
|pp|,
where the momentum-space volume element d
3
p is dp
x
dp
y
dp
z
in Cartesian coordinates.
32
Box Normalisation: While -function normalisation is generally no problem, there are some
diculties: e.g. in the proof that r has a continuous spectrum and in the handling of the
parity operator, as we will see later, where innite factors are cancelled! Also simply checking
self-adjointness of p = ih is a puzzle if wavefunctions do not vanish at . One escape
device is to use box normalisation where space is given torus topology by putting the particle
into a cube of edge-length L and identifying parallel faces. Then the wavefunction
p
(r) (for
instance) obeys periodic boundary conditions which imply that Cartesian components p
i
of p are
quantised by p
i
L = 2n
i
h for n
i
= 0, 1, . . . and which ensure that p is self-adjoint. Choosing
C = L
3/2
integration over the box gives p|p

as a product of Kronecker deltas on integers n

i
and

I =

ni
|pp|. Factors of L cancel from results of observation. For more details see Schi,
p. 48.
Momentum eigenstates obey the free-particle energy eigenvalue equation

h
2
2m

2
= E
provided |p|
2
= 2mE and with this relation understood we have
r|E(p), p = exp(
i
h
p r).
For each E there is degeneracy with respect to direction of p and all directions must be counted in summing
over the complete set of energies to construct the wavefunction (r, t).
So with continuous momentum eigenstates instead of discrete energy eigenstates the sum solution to the
Schr odinger Wave Equation for a free particle becomes the Fourier Integral expression
(r, t) =
_
d
3
p
(2h)
3

(p) exp(
i
h
(p r
p
2
2m
t)),
where

(p) E, p| =
_
d
3
r exp(
i
h
p r)(r, 0)
is the Fourier Transform of the initial data, its momentum-space wavefunction. This form of (r, t) is a
wavepacket solution to the Schrodinger Wave Equation.
For a particle with denite momentum p
0
at t = 0 we have

(p) = E, p|p
0
= (2h)
3
(p p
0
)
and the Schr odinger Wave Equation has plane-wave solution
(r, t) = exp(i(k r t))
with wave-vector k p
0
/h and angular frequency E/h connected by the dispersion relation (k) =
hk k/2m. This is the particles de Broglie wave, just as experiment observes, with group velocity
d
dk
=
p
0
m
,
which is the classical particle velocity.
Example: The Two-slit Experiment. This is discussed as a thought experiment in textbooks
(e.g. Schi pps. 1214; Feynman Ch. 1) but certainly has been performed in the lab). Here a
stream of electrons of denite momentum p falls normally on a screen with two narrow slits,
parallel and close-spaced, and the point of arrival of each particle on a parallel screen behind
is recorded. The electrons fall in bands, forming the diraction pattern appropriate to the
wavevector p/h. The stream looks like a wave. But the pattern builds up even if the intensity
33
is so low that only a single electron is in transit at any time. The wave is usually pictured like
a water wave, propagating from source to screen each time an electron travels. But it must be
recalled that in fact the wave is a solution of the Schrodinger Wave Equation with boundary
conditions appropriate to the physical circumstances; here an arrangement of source, screens and
slits, and it is a complex-valued function dened in an entirely abstract conguration space for
the electron. The position of arrival of each electron (its conguration at the that instant) is a
random event with probability density given by the wavefunctions squared modulus. The wave
serves to describe the possible outcomes of observation in the quantum model, just as a set of
p, q coordinates obeying Hamiltons equations would do (albeit with greater certainty!) in a
classical model. The quantum particle is no more a wave than the classical particle is a point
in phase space. If a slit is closed the diraction pattern disappears, for the boundary conditions
on the Schr odinger Wave Equation are changed and the wavefunction changes. In fact then the
position of the electron is known as it passes the rst screen. The same is true if any observation
determines the particles position in transit. A position measurement changes its wavefunction
to a position eigenfunction. With all this in mind its instructive to read accounts of the two-slit
experiment in Feynman, Secs. 1-4 to 1-11; Polkinghorne, Chap. 4.
4.6 Spreading of a wave packet
Let us how that a wave packet always spreads as time evolves. To see this let us take a (one dimensional)
wave packet which at t = 0 is centred at x = 0 with a spread a
0
. For this we can take a Gaussian

a0
(x, 0) =
1
(a
0

)
1
2
e

x
2
2a
2
0
.
Notice that this wave function happens to be real (this does not matter) and that the normalisation has
been so chosen that
_

dx||
2
= 1.
Then

a0
(p, 0) =
1

2h
_

dxe

ixp
h

a0
(x, 0) =
1

2h
1
(a
0

)
1
2
_

dxe

ixp
h
e

x
2
2a
2
0
=
1

2h
1
(a
0

)
1
2
_

dxe

_
x

2a
0
+
ia
0
p

2 h
_
2
e

a
2
0
p
2
2 h
2
=

2a
0

2h
_
a
0

a
2
0
p
2
2 h
2
_
dxe
x
2
=
_
2a
0

2h
e

a
2
0
p
2
2 h
2
.
In this calculation we have used
_

dxe
x
2
=

and assumed that we can alter the path of the
integration from the real line (in x) to the line (parallel to the real axis) in the complex x plane .
If the particle is free then E =
p
2
2m
and

a0
(p, t) =

a0
(p, 0) e
i
p
2
t
2m h
.
Thus
(x, t) =
1

2 h
_

dp e
ixp
h

a0
(p, t) =

2a0

2 h
_

dp e

ip
2
t
2m h
+
ixp
h

a
2
0
p
2
2 h
2
=

2a0

2 h
_

dp e
z(p+Rx)
2
e
zR
2
x
2
,
where z =
a
2
0
2 h
2
+
it
2m h
and R =
i
2z h
.
Again, we change the integration variable from p to (p +Rz)

z and, distorting the contour, obtain

_
2a
0

2h
_
_
a
2
0
2 h
2
+
it
2m h
_
e

x
2
4 h
2
z
=
1
()
1
4
_
a
0
+
it h
ma0
e

x
2
2
(
a
2
0
+
it h
m
)
.
34
Note that
|
a0
(x, t)| = |
a
(x, 0)|
where
a
2
= a
2
0
+
h
2
t
2
m
2
a
2
0
.
So we see that our Gaussian wave packet is spreading; its width grows as a(t). This eect is called the
spreading of the wave function. To prevent this from happening you need some non-linearity in - such
phenomena, (non-linear Schr odinger equation etc.) are discussed in a course on solitons.
4.7 Ehrenfests theorems; Energy-time uncertainty principle
4.7.1 Ehrenfests theorems
Ehrenfest showed that the mean values of observables

A behave like classical quantities; they satisfy

equations of classical mechanics.
To see this assume that

A =

A(t). Then
d
dt

A =
d
dt
|

A| =
_
d
dt
|
_

A| +|

A
_
d
dt
|
_
=
1
ih
|[

A,

H]| =
1
ih
[

A,

H].
Recall that in classical mechanics we had (Poisson brackets)
d
dt
A = {A, H}.
Thus, for a particle of mass m in one dimension, we have
d
dt
x =
1
ih
[ x,
p
2
2m
] =
p
m
.
If

H =
p
2
2m
+

V ( x)
and if

F( x) =

x

V then
d
dt
p =
1
ih
[ p,

V ( x)] =
V
x
=

F.
This last equation looks very much like the Newton law. But note that we have
V
x
and not
V
x
(x)
and so we cannot talk about the centre of the packet as moving classically; moreover, as we already know,
we have also some additional eects due to the spreading out of the packet.
4.7.2 Time-energy uncertainty principle
Recall that
d
dt

A =
1
ih
[

A,

H]
But,
A =
_

A
2

A
2
_
1
2
35
so
AB
1
2

A,

B]

.
So applying this to

A and

H we have
AH
1
2

A,

H]

=
h
2

d
dt

.
But

H = E so

A
E
h
2
,
where

A
=

d
dt

.
Here
A
is the time characteristic of the statistical evolution of the system as seen via its eects on the
observable

A; i.e. the time required for the centre

A of this distribution to be displaced by an amount

equal to its width; i.e. time necessary for the statistical distribution to be appreciably altered.
Take = min
A

A
. Then
E
h
2
.
which is the Time-Energy uncertainty principle.
Comments:
If a system is in a stationary state then
d A
dt
= 0 (as then |[

A,

H]| = 0) and so = but then
energy is well dened and so E = 0.
Note that the origin of the time-energy uncertainty principle is dierent from qp h/2. Theres
no self-adjoint time operator but instead an operator for evolution time as measured by A has been
introduced as

A/ |A/t|. Its clear that the less well-dened is energy, the more frequencies E/h
contribute to the series solution of the Schr odinger Equation and the faster the system may appear to
evolve in terms of any measurable A. Conversely a sharper denition of energy forces slower evolution.
The time-energy uncertainty relation is often invoked in picturing, say, an electromagnetic interaction
of two charged particles as exchange of a photon. It is said that violation of energy conservation in
emission and absorption can be allowed if it occurs quickly enough. Similar statements are often made
in connection with tunnelling, e.g. Polkinghorne pps. 50 52.
Examples:
1. Decay of nuclear matter (energy not well dened).
2. Superposition of 2 stationary states. Take a superposition of 2 stationary states of energies E
1
and
E
2
.
(r, t) =
1
(r)e

iE
1
t
h
+
2
(r)e

iE
2
t
h
.
Then
P(r, t) = |
1
(r)|
2
+ |
2
(r)|
2
+ 2Re
_

2
e
i(E
1
E
2
)t
h
_
.
This quantity oscillates in time between 2 extreme values (|
1
| |
2
|)
2
and (|| +|
2
|)
2
with the period
of the oscillation given by =
h
(E1E2)
.
Thus the statistical distribution of the results of measurements made at t
1
and t
2
will be practically
identical if t = |t
1
t
2
| is small compared to .
Thus, in order that the properties of the system be signicantly modied over the time period t,
t E must be at least h.
36
4.8 Probability current; Conservation of probability
Recall that |(r, t)|
2
d
3
x gives the probability (when is properly normalised) of nding the particle within
r and r +dr at time t. Thus
P

=
_

d
3
x |(r, t)|
2
=
_

d
3
x(r, t)
is the probability of nding the particle in . Note that P

1 as V ol() and RR
3
.
But P

= P

(t). So calculate
dP

dt
=
d
dt
_

d
3
x

=
_

d
3
x

t
+
_

d
3
x

t
,
if = (t).
Using Schrodinger equation this is equal to
ih
2m
_

d
3
x[

] =
ih
2m
_
S
dS(

),
where S is the surface bounding .
Then dene
j(r, t) =
ih
2m
(

) =
ih
m
Im(

) = Re(
ih
m

),
the probability current and we see that we have

t
(r, t) + j(r, t) = 0.
i.e. a continuity equation.
This has similar status to the local conservation laws that apply in uid mechanics, electromagnetism,
heat, etc. The density of the particle is not created or destroyed, it just moves about. Indeed in a normal-
isable state
_
d (r, t) = 1 independent of t, (the Schr odinger Wave Equation preserves norms).
Note: if the particle is in a stationary state then P

= P

(t) as (r, t) = (r) and (

) = 0 thus
[

(r) (r)] = 0
as the time dependent factors cancel out. This helps us to determine the boundary conditions on (r) when
we consider it in a given region .
[
h
2
2m

2
+V (r)] = ih

t
, or = E.
Boundary conditions (as this is a second order equation for (r, t) or (r)):
1. Conditions at |r|
| =
_
d
3
x||
2
= 1
so
(r) 0 as |r|
suciently fast.
Comment: Sometimes we use generalised eigenstates, i.e. use replace wave packets by plane waves
e
ipr
. Such states are not realisable physically but they simplify the calculations, For such states ||
const as |r| )
2. Points of discontinuity of V (r). We patch solutions in each region in which V is continuous. At the
discontinuity surfaces we demand that and j are continuous. This implies that and n, where
n is a unit vector normal to the boundary of the region, have to be continuous.
3. Points where V (r) = . At such points it may happen that = 0 and has a discontinuity.
(This will be discussed in the next chapter)
37
5 One-Dimensional Systems
5.1 Preliminary (recall also 4.9)
Justication: We study wave mechanics rst in one dimension for practice and to discover generalities
(e.g. tunnelling) without needing to deal with angular momentum. This is just like classical mechanics
and, just as there, in (important) circumstances where angular momentum is conserved three-dimensional
problems can be reduced to equivalent one-dimensional form.
Problem: A quantum particle of mass m moves on the x-axis in a potential eld V (x). Discuss
boundary conditions on the wavefunction.
Solution: The wavefunction (x, t) obeying the one-dimensional Schr odinger Wave Equation is a sum
or integral over E involving the energy eigenfunction
E
(x) which solves

=
2m
h
2
(V (x) E).
Localised or bound solutions with discrete E-values occur for E < min{V

} where V

V (x )
and have (x ) exp(K

|x|) where K
2

= 2m(V

E)/h
2
. If V (|x| ) then
E
0
faster at .
Scattering solutions
E
(|x| ) exp(ikx) where k
2
= K
2

occur for E > min{V

}, have continu-
ous E-values and correspond to leftward (-) and rightward (+) motion, being asymptotically eigenfunctions
of p = ihd/dx with momentum eigenvalues hk (k > 0 by denition).
Integrating the energy-eigenvalue equation we have

E
(x
1
)

E
(x
2
) =
2m
h
2
_
x2
x1
(V (x) E)
E
(x) dx
and so, using the continuity of ,

E
(x + )

E
(x )
2m
h
2

E
(x)
_
x+
x
V (x) dx
for very small ||. The right-hand side vanishes as 0 whenever V (x) is continuous or, we note, whenever
V (x) is piecewise continuous. So at a nite jump in V (x) both
E
(x) and

E
(x) are continuous. This
ensures continuity of the one dimensional probability current j = (h/m)Im(

/x).
As examples will illustrate, in the limit of an innite jump in V (x) where the integral on the right-hand
side diverges consistency demands that
E
vanishes. Then generally

E
has a nite discontinuity, and j is
continuous and zero. A wavefunction that penetrates a region of very large V is exponentially damped and
an innite jump in V is an impenetrable barrier; the particle bounces back.
If, bizarrely, V (x) has an innite jump of zero width, i.e. V (x) = g(xa)+. . ., then

has a discontinuity
at x = a:

E
(a
+
)

E
(a

) =
2m
h
2
g
E
(a)
where
E
is continuous and not necessarily zero at x = a.
5.2 Examples
5.2.1 Free Particle
For a free particle dene V = 0. A particle of mass m and energy E > 0 moving freely on the x-
axis has wavefunction (x, t) = exp(iEt/h)
E
(x) and everywhere the energy eigenfunction
E
(x) obeys

+k
2
= 0 where k = +

2mE/h for any E > 0. The solution

E
(x) = c
1
exp(ikx) +c
2
exp (ikx)
is a superposition of right- and left-moving parts, eigenstates of p = ihd/dx with eigenvalues hk. Energy
eigenstates are two-fold degenerate. The discussion of wavepackets in Sec. 3.5 simplies: r x, p p,
(2h)
3
2h.
38
5.2.2 One Dimensional Step Potential
Consider the stationary problem; i.e. the time independent Schr odinger equation corresponding to the po-
tential
V (x) =
_
0, x < 0
V
0
> 0 x > 0
and consider the case of E < V
0
. Classically the particle can be anywhere for x < 0 and the region x > 0 is
not accessible.
Quantum mechanically, we study the wave function in each region and then use the continuity conditions
to relate the functions in each region.
So we put (x) =
>
(
<
) for x > 0 (x < 0). So

H = E gives
x < 0
h
2
2m
d
2
<
dx
2
= E
<

<
= Asin(kx) +Bcos(kx)
x > 0
h
2
2m
d
2
>
dx
2
= (E V
0
)
>

>
= C exp(x) +Dexp(x),
where k =
_
2mE
h
2
, =
_
2m(V0E)
h
2
.
Next we impose our boundary conditions. At x = the function is not innite so D = 0.
At x = 0 we have

>
=
<
B = C

>
=

<
Ak = C C =
kA

.
So
>
(x) =
kA

e
x
and
<
(x) = A(sin kx
k

cos kx).
Note that as V
0
then and
>
(x) (due to the exponential) and
<
(x) Asin kx, as
mentioned before.
Note that our wavefunction cannot be normalised as
_
0

<
(x)
<
(x) < 0 is impossible to satisfy as
_
0

sin
2
(kx) dx = . So we interpret this situation as describing a beam of particles (we will come back
to this case later when we will discuss transmission and reection phenomena).
Comment: Note that quantum mechanically there is a non-zero probability of nding our particles in the
region of x > 0.
5.2.3 One Dimensional Potential with Rigid Walls
A surface for which V = is called rigid. Hence transition from V = to V = due to F = V
involves very strong forces. This would be the case if we had, say, very strong reaction forces preventing the
particle from entering such a region.
To study such a case we consider
V =
_
0, |x| < a
, |x| > a
So for |x| > a = 0 and for |x| < a we have

h
2
2m
d
2

dx
2
= E
which has as solutions (x) = Asin(kx) +Bcos(kx) with k as before.
The continuity at |x| = a gives us
Asin(ka) +Bcos(ka) = 0
Asin(ka) +Bcos(ka) = 0.
Its solutions are A = B = 0 (i.e. no state) or sin(ka) = 0 and B = 0 or cos(ka) = 0 and A = 0.
The two choices give us k =
n
a
or k =
2n+1
2

a
so we see that
k =
n
2a
, n = 1, 2, 3...
39
and so the possible values of energy are
E
n
=
h
2
k
2
2m
=

2
h
2
n
2
8ma
2
.
Comments:
We have a completely discrete spectrum of E. This is characteristic of potentials which go to as
|x| .
E
n
as n
E
1
=

2
h
2
8ma
2
. The lowest energy = 0. (The particle is never at rest; this phenomenon is called zero
point energy).
Note; we can understand why E
1
= 0 (from the uncertainty principle). As |x| a we have that
p
h
2x
. So
E =
1
2m
p
2

1
2m
(p)
2
=
h
2
8ma
2
.
If you plot a few lowest energy wave functions (say,
1
(x),
2
(x) and
3
(x)) we see that
1
has no nodes;

2
has one node,
3
has 2 nodes etc.
The larger the number of nodes - the higher the energy. This can be understood as follows; more nodes
corresponds the larger variation of the wave function and, in consequence, the larger energy and momentum.
5.3 Parity
Bound-state wavefunctions of the square well and, as we will see later, of the simple-harmonic oscillator
have denite symmetry under the parity operation x x. This is reection through the origin; in three
dimensions r r.
In quantum mechanics a parity operator

P for one particle can be dened by its eect on elements of the
Schr odinger basis, eigenstates of r:

P|r = c| r
or in position representation as

P(r) = c(r).
Requiring that two reections return to the start,

P
2
=

I and hence c
2
= 1. Choose c = 1. Since
r|

P|r = r| r = r|r then

P is unitary (skating over normalisation!) So both

P =

P
1
and

P
1
=

P

. Hence

P is self-adjoint, with (by inspection) eigenvectors (|r | r)/

2 for any r and

eigenvalues 1.
In the position representations the eigenfunctions of

P are (1

P) for any as

P(1

P) = (

P 1) = (1

P).
Note that

P does not have unique eigenvectors but simply partitions the r-basis into two parts. The
spectrum is complete, so

P is observable, even though it cannot be constructed from basic degrees of freedom.
(It is not a rotation because left- and right-handed Cartesian axes transform to each other).
Note that
(1

P)(r) = (r) (r)
so the eigenfunctions are even and odd functions of r with corresponding eigenvalues +1 and -1.
Now

Pr

P|r = r|r = r|r for any |r and so r

P =

Pr. Thus, since [r, p] = ih, we also have

p

P =

P p. (Note that this means

P|p = c

| p with c
2
= 1, which can be an alternative denition of

P). Therefore it follows that [r

2
,

P] = [ p
2
,

P] = 0.
Indeed, under parity the self-adjoint operator corresponding to any observable A(r, p) transforms as

A(r, p)

P

A

P =

A(r, p).
If

A is unchanged (invariant) then [

P,

A] = 0 and

A and parity are compatible.
40
Examples: Kinetic energy T = p p/2m and angular momentum L = r p of a particle are
parity-invariant. Intrinsic spin is also parity-invariant.
For one particle, mass m, in a potential we have Hamiltonian

H = p
2
/2m+

V (r) and so if [

V ,

P] = 0 we
have [

H,

P] = 0 and parity is conserved. The condition on

V is equivalent to

V (r) =

V (r) i.e. that

V
depends only on r
2
.
If parity is conserved, energy eigenstates can also be also eigenstates of

P indeed they must have
denite parity if the spectrum of

H is non-degenerate.
Examples: For a free particle (V = 0) momentum eigenstates |E, p can be partitioned into
parity eigenstates (|E, p |E, p)/

2 with parity 1. For a central potential, where V =

V (|r|), parity is conserved alongside angular momentum.
In wave mechanics for a particle in a symmetrical potential V (r) = V (r), and when

H has non-
degenerate spectrum, for the wavefunction we have

E
(r) r|E = r|

P|E = r|(|E) =
E
(r),
i.e. denite symmetry under r r. If there is degeneracy then energy eigenfunctions may have denite
parity, but it must be checked explicitly. For bound states of a central potential (see Chap 6), where parity is
conserved, there is angular-momentum degeneracy. But even so, the wavefunctions turn out to have denite
parity, for [

L,

P] = 0 and

L has a non-degenerate spectrum.
In one-dimensional wave mechanics parity is conserved i V (x) = V (x), which is true, as we have seen,
for the symmetrical square-well potential and for the simple harmonic oscillator. There is no degeneracy for
bound states (as shown in the next Section) and so the energy eigenfunctions are necessarily eigenfunctions
of parity. Thus in the case of the square well our wavefunctions are eigenfunctions of

P. Note that

P
n
(x) = (1)
n+1

n
(x)
and so
n
(x) has parity (1)
n+1
. Note that the ground state wavefunction has parity +1, i.e. is an even
function of x). This is the case for all reasonable potentials (which are even).
Note that if is even then (x) = (x). Then

(x) =

(x) and so

(0) = 0. And if is odd we
have (x) = (x) and so (0) = 0.
We can exploit this observation as follows. If (x) is a solution of

H = E (with V (x) = V (x)) for
x > 0 then if

(0) = 0 (x) = (x)

is a solution for x < 0 and if
(0) = 0 then (x) = (x)
is a solution for x < 0.
Message For even potentials we may look for eigenstates of denitive parity. We do this by solving

H = E
in x > 0 subject to (0) = 0 or

(0) = 0 and extend as above.
For scattering from a symmetrical potential the asymptotic energy eigenfunctions are degenerate with
respect to momentum direction and there is no automatic even/odd symmetry of energy eigenfunctions.
Indeed incoming- or outgoing-wave boundary conditions generally exclude it.
5.4 Finite Square Well
Next we consider a nite square well potential, i.e. a system with V (x) given by
V (x) =
_
0, |x| < a
V
0
, |x| a
.
41
Classically, if E > V
0
the particle can move o to and so is not bound, but when E < V
0
- the particle
must remain in the hole and we have a bound state. This aspect of the problem is preserved in Quantum
Mechanics but is modied.
First look at bound states (i.e. consider E < V
0
). Then
x < a
h
2
2m
d
2

dx
2
= E (x) = Bsin(kx) +Acos(kx), k =
_
2mE
h
2
x a
h
2
2m
d
2

dx
2
= (E V
0
) (x) = Ce
x
+De
x
, =

2m(V
0
E)
h
2
and to stop from || as |x| we have to put D = 0
Hence for even parity states we take

(0) = 0 and so B = 0 and the continuity of and

at x = a
gives us
Acos(ka) = Ce
a
Ak sin(ka) = Ce
a
and so
k tan(ka) =
is our condition for a non-zero wavefunction. This is our eigenvalue equation for E.
For odd parity states we put
(0) = 0 A = 0
and so continuity of and

at x = a gives us
Bsin(ka) = Ce
a
Bk cos(ka) = Ce
a
and so
k cot(ka) = .
The equations for E i.e. for k can be solved only numerically or graphically.
graphical analysis
Change variables to x1 = ka and y1 = a. Then
x1
2
+y1
2
= (k
2
+
2
)a
2
=
2mV
0
h
2
a
2
= R
2
a
2
and so we see that the solutions of our equations lie on the intersection of this circle with x1 tan(x1) = y1
(for even states) and x1 cot(x1) = y1 for odd states.
We see that the number of states is nite and that there is always at least one state.
If V
0
is such that R
2
R
2
1
(where R
1
<

2
) we have only one (positive parity) state. If R R
2
, where

2
< R
2
< we have two states (one odd, one even); if R R
3
, where < R
3
<
3
2
we have three states
(two even, one odd).
In general, if
_
2mV
0
a
2
h
2
[n, (n +
1
2
)]
we have n + 1 even and n odd states while if
_
2mV
0
a
2
h
2
[(n +
1
2
), (n + 1)]
we have (n + 1) odd and (n + 1) even states.
Notes
42
If V
0
the momenta ka
n
2
and so E

2
h
2
n
2
8ma
2
as before.
Even when the particle is bound there is a small probability that the particle is outside the well. This
probability is proportional to exp(2x).
It is easy to sketch the wave functions for a few lowest states.
Next consider the continuous part of the spectrum, For this we need E > V
0
. Then
= Asin(kx) +Bcos(kx), a < x < a
and
for x > a = Ce
i(xa)
+De
i(xa)
,
where =
_
(EV0)2m
h
2
= i.
Now the condition of x gives no restriction on C and D and the states cannot be normalised.
We can still take eigenstates of denitive parity (even or odd). For even states A = 0 and Bcos(ka) =
C + D and kBsin(ka) = (C D)i. These can be solved for any E thus the spectrum is continuous.
Similarly for odd states. Thus we have two eigenstates (not normalised), one of each parity, for every value
of E > V
0
.
5.5 Reection and Transmission Phenomena
Consider the step potential as before (V = 0, x < 0; V = V
0
for x > 0). We have found eigenstates for
0 < E < V
0
. Let us now show that this problem has also eigenstates for E > V
0
. Moreover, there are no
states for E < 0 so the spectrum satises 0 < E < .
Return to the case of E < V
0
. Then for x < 0 the wave function satises
(x) =
<
(x) = A(sin(kx)

k
cos(kx))
=
A
2i
(1
ik

)e
ikx

A
2i
(1 +
ik

)e
ikx
,
where k =
_
2mE
h
2
and =
_
2m(V0E)
h
2
.
For x > 0
(x) =
>
(x) =
k

Ae
x
.
Thus in x < 0 the wavefunction is a superposition of e
ikx
and e
ikx
. But
p(e
ikx
) = hi

x
(e
ikx
) = hi(ikx)e
ikx
= hke
ikx
.
So e
ikx
is an eigenfunction of momentum with hk as its eigenvalue.
But e
ikx
is not normalisable. So it cannot represent a single particle of exact momentum hk. As

e
ikx

2
= 1 we would have probability 1 of nding the particle in a unit interval. Thus we interpret e
ikx
as representing a beam of particles, of density 1 per unit length, each of momentum hk.
Similarly Ae
ikx
gives density |A|
2
per unit length. Thus e
ikx
(e
ikx
) describes particles moving to the
right (left).
The ux of particles is density velocity so for Ae
ikx
the ux is |A|
2
hk
m
.
Thus, in our case, the ux moving to the right is
|A|
2
4
(1 +
k
2

2
)
hk
m
, x < 0
and to the left also
|A|
2
4
(1 +
k
2

2
)
hk
m
, x < 0.
In the region of x > 0 the wave is exponentially decreasing and so no particles get to x = . This
explains why the ux to the left is the same as the ux to the right - we have complete reection.
43
We can relate these observations to the continuity equation.
As is stationary j = 0 so in one dimension
d
dx
j = 0 and so j(a) = j(b) where, in one dimension
j =
ih
2m
_

d
dx

d

dx
_
.
So for = Ae
ikx
we see that j =
hk
m
|A|
2
conrming our observations.
For a more general case

<
= Ae
ikx
+ Be
ikx
,
>
= C e
x
Then, it is easy to check that
j(x < 0) =
hk
m
[|A|
2
|B|
2
], j(x > 0) = 0
So, as we have already veried, there is no ux to the right and |A| = |B|.
Note; this is exactly like in the classical case. There, if E < V
0
the particles would be stopped by the
potential at some point x

where V (x

) = E, T(x

) = 0. Their motion is then reversed by the action of the

force resulting in the complete reection of the beam.
If E > V
0
the situation is dierent. Classically, the particles emerge with kinetic energy T
1
and momentum
p
1
such that
T
1
=
p
2
1
2m
= E V
0
> 0.
And in Quantum mechanics? We have, for x < 0
=
<
,
d
2

<
dx
2
+k
2

<
= 0, k =
_
2mE
h
2

<
= Ae
ikx
+Be
ikx
while for x > 0
=
>
,
d
2

>
dx
2
+
2

>
= 0, =

2m(E V
0
)
h
2

>
= Ce
ix
+De
ix
.
The conditions on and

are not sucient to determine 3 constants (the overall constant is irrelevant
as it is related to the density of the beam).
Having A = 0 and D = 0 means that we have beams incident from both x = and x = .
Physically we expect some reection etc so if the incoming beam is from the left (i.e. from ) then we
do not expect any beam from so D = 0. By this requirement we specify that the beam is red at the
barrier from .
Then the continuity conditions give us

>
(0) =
<
(0) A +B = C

>
(0) =

<
(0) k(AB) = C
giving
C =
2k
k +
A, B =
k
k +
A.
So the incident ux is
hk
m
|A|
2
, reected ux is
hk
m
|B|
2
=
hk
m

k
k +

2
|A|
2
and the transmitted ux is
h
m
|C|
2
=
h
m
4k
2
|k + |
2
|A|
2
44
So
j(x < 0) =
hk
m
(|A|
2
|B|
2
)
j(x > 0) =
h
m
|C|
2
.
Due to j(x > 0) = j(x < 0) we see that the incident ux is a sum of the reected and transmitted uxes.
We dene R - the reection coecient
R =
reected ux
incident ux
=
(k )
2
(k + )
2
and T- the transmission coecient
T =
transmitted ux
incident ux
=
4k
(k + )
2
and we see that T +R = 1.
5.6 The Tunnelling Eect
Consider V as indicated
V (x) =
_
0, x < 0
V
0
> 0, a > x > 0
0, x > a
.
Then let us study what happens when a beam of particles, of unit density is incident from with some
particular energy E. Classically, if E < V
0
the beam will be reected and if E > V
0
it will be transmitted.
In Quantum mechanics we have reection and transmission in each case.
Study rst E < V
0
. Then
x < 0
h
2
2m
d
2

dx
2
= E, = e
ikx
+Be
ikx
, k =
_
2mE
h
2
0 < x < a
h
2
2m
d
2

dx
2
= (E V
0
), = Ce
x
+De
x
, =

2m(V
0
E)
h
2
x > a
h
2
2m
d
2

dx
2
= E, =

Ee
ikx
(we have set A = 1 for simplicity). We relate B, C, D and

E by the continuity of and

at x = 0 and
x = a. The conditions at x = 0 give us
1 +B = C +D, ik ikB = (C D)
2ik = ( +ik)C ( ik)D.
and at x = a we get
Ce
a
+D
a
=

Ee
ika
, Ce
a
De
a
= ik

Ee
ika
(ik + )Ce
a
= ( +ik)De
a
.
To solve them we put C = ( + ik)e
a
and D = ( ik)e
a
, which solves the last condition, and
nd
2ik

= ( +ik)
2
e
a
( ik)
2
e
a
.
So
=
2ik
(+ik)
2
e
a
(ik)
2
e
a
=
2ik
(
2
k
2
)(e
a
e
a
)+2ik(e
a
+e
a
)
=
ik
(k
2

2
)sh(a)+2ikch(a)
.
45
Then

E = (Ce
a
+De
a
)e
ika
= 2e
ika
=
2ike
ika
(k
2

2
)sh(a) + 2ikch(a)
And
B = C +D 1 = ( +ik)e
a
+ ( ik)e
a
1 = 2ch(a)
1 2iksh(a) =
(k
2
+
2
)sh(a)
(k
2

2
)sh(a)+2ikch(a)
.
Then R = |B|
2
and T =

2
and

2
+ |B|
2
=
4k
2

2
(k
2

2
)
2
sh
2
(a) + 4k
2

2
ch
2
(a)
+
(k
2
+
2
)
2
sh
2
(a)
(k
2

2
)
2
sh
2
(a) + 4k
2

2
ch
2
(a)
= 1
as required. The fact that we have a transmitted wave in the region that is not accessible classically is called
the tunnelling eect.
Note that if the barrier is very narrow i.e. a is small we have a 0 and so B 0,

E 1 and the whole
wave is transmitted.
This eect plays a very important role in nuclear physics, in the description of the decay of nuclei.
5.7 Lessons
If V () is nite the spectrum of

H has two sectors, scattering and bound-state. Scattering energy-values
are continuous with asymptotic double degeneracy with respect to momentum (in, out). The square-well
example shows that the linear and homogeneous continuity conditions are consistent with all E-values if both
in and out asymptotic components of the wavefunction are present, but that they restrict allowed E-values
to just a discrete set when the asymptotic form is a single decaying exponential.
Scattered particles may reect from sudden changes in V (x) and the tunnelling may occur. Conned,
localised, or bound particles have zero-point energy (a non-trivial wavefunction vanishing at both must
be curved, so that p
2
> 0).
Discrete energy levels are bound states, with square-integrable wavefunctions. They are always non-
degenerate in one dimension. This is easily seen by supposing otherwise when both
1
(x) and
2
(x) obey

=
2m
h
2
(V (x) E)
with the same E and, multiplying each equation by the other and subtracting, we have

2
= 0.
Integrating, the Wronskian of the two supposedly independent functions is a constant. But the constant is
zero, evaluating at x = where the (bound-state) s vanish, so they are in fact linearly dependent.
With a symmetric potential parity is a good quantum number. The ground state is always symmetrical
(theorem) and eigenfunctions have zeroes increasing by one going up the spectrum (theorem) and interleaving
(theorem). Refer to Messiah, Vol I, pps. 98 to 113 for proofs of these theorems, which follow from the Sturm-
Liouville nature of the Hamiltonian operator.
6 Three-Dimensional Systems
6.1 Square well with rigid walls
Now we look at more physical cases i.e. particles in 3 dimensions. We start with problems that can be
solved with the knowledge of 1-dim results.
46
Let us look at the stationary Schr odinger equation.
[
h
2
2m

2
+V (r)](r) = E(r).
Usually we can make progress by exploiting some symmetry of the problem or by separating the equation
in some set of coordinates.
Here we look at case of the potential
V (r) =
_
0 |x| < a, |y| < b, |z| < c,
otherwise
In this case it is convenient to use Cartesian coordinates.
Clearly (r) = 0 for |x| > a or |y| > b or |z| > c so the particle is conned to a box.
We seek a solution of the Schr odinger equation in the form (r) =
1
(x)
2
(y)
3
(z). Then

h
2
2m
_

1
+

2
+

3
_
= E.
Thus

h
2
2m

i
= E
i
,
where

i
E
i
= E and

1
(a) = 0,
2
(b) = 0,
3
(c) = 0
and so we have 3 one dimensional problems whose solutions are known.
As possible values of E
1
are
n
2
1

2
h
2
8a
2
m
we see that
E =

2
h
2
8m
_
n
2
1
a
2
+
n
2
2
b
2
+
n
2
3
c
2
_
The corresponding eigenfunctions are
(r) = A sin
_
n
1

2a
x +
n
1

2
_
sin
_
n
2

2b
y +
n
2

2
_
sin
_
n
3

2c
z +
n
3

2
_
,
which are complete (n
1
, n
2
, n
3
= 1, 2, 3, ..).
Note that if a
2
= b
2
= c
2
= d
2
then
E =

2
h
2
8md
2
[n
2
1
+n
2
2
+ n
2
3
]
so states like (1,2,2), (2,1,2,) and (2,2,1) have the same energy. There must be operators which allow us to
distinguish between these states; they are p
i
.
6.2 3 Dimensional Harmonic Oscillator
For the 3 dimensional Harmonic oscillator we take

H =
p
2
2m
+
1
2
m
2
r
2
,
where r
2
= x
2
+ y
2
+ z
2
. We solve it as a problem in wave mechanics i.e. in position representation. Then
p
2
= h
2

2
= h
2
(
2
x
+
2
y
+
2
z
)
and we put
(r) =
1
(x)
2
(y)
3
(z).
47
We get 3 equations
_

h
2
2m
d
2
dx
2
+
1
2
m
2
x
2
_

1
(x) = E
1

1
(x)
and similar equations for
2
(y) and
3
(z).
E = E
1
+E
2
+E
3
.
We use our previous knowledge to discuss their solutions. As one dim. SHO has eigenvalues E
1
=
(n
1
+
1
2
) h, n
1
= 0, 1, 2... with eigenfunctions

n1
(x) = H
n1
(x

)e

1
2
x
2
, x

=
_
m
h
x
we see that
E =
_
n
1
+n
2
+ n
3
+
3
2
_
h
with eigenfunctions
(r) =
n1
(x)
n2
(y)
n3
(z).
To put it dierently, we set

H =

i=x,y,z

H
i
where

H
x
=
p
2
x
2m
+
1
2
m
2
x
2
, etc.
As

H
i
commute and are Hermitian we can nd a basis of simultaneous eigenstates of all

H
i
and

H. So
we take |n
1
, n
2
, n
3
where

H
x
|n
1
, n
2
, n
3
= h
_
n
1
+
1
2
_
|n
1
, n
2
, n
3

H
y
|n
1
, n
2
, n
3
= h
_
n
2
+
1
2
_
|n
1
, n
2
, n
3

H
z
|n
1
, n
2
, n
3
= h
_
n
3
+
1
2
_
|n
1
, n
2
, n
3

and so

H|n
1
, n
2
, n
3
= h
_
n
1
+n
2
+n
3
+
3
2
_
|n
1
, n
2
, n
3
.
Thus we see that our theorem about commuting operators is closely related to the separation of variables;
in fact, the separation of variables is really just our usage of this theorem.
6.3 Central Potentials
In three-dimensional wave mechanics the wavefunction (r, t) of a particle of mass M in a potential V (r)
obeys the Schrodinger Wave Equation
h
2
2M

2
+V (r) = ih

t
and may be expressed as a superposition of energy eigenfunctions
E
(r) satisfying
h
2
2M

2
+V (r) = E.
Each eigenfunction may be chosen to be simultaneously an eigenfunction of all the mutually compatible
conserved quantities.
48
Examples: Free-particle energy eigenfunctions may be eigenstates of ih with momentum
p in any direction so long as E =
1
2
|p|
2
/M. Alternatively they may be eigenstates of parity or
of angular momentum, or of both.
With a central potential V (r) = V (r), where r = |r| =

r r, all components of angular momentum

L commute with

H since they commute with both p
2
and r
2
. So angular momentum is conserved, as in
classical mechanics (and note that parity is conserved too). Then =
Em
(r) obeys simultaneously
h
2
2M

2
+V (r) = E,

L
2
= ( + 1) h
2
,

L
z
= mh,
where

L
2
=

L
2
x
+

L
2
y
+

L
2
z
and

L
x
= y p
z
z p
y
= ih
_
y

z
z

y
_
etc.
The commutators of angular momentum alone permit = 0,
1
2
, 1,
3
2
, . . . and constrain m = , +1, . . . , .
Because V depends only on radial distance r, spherical polar coordinates (r, , ) are better than Carte-
sians. Then the angular momentum operators that commute with V (r) can depend only on angular deriva-
tives. So the angular-momentum equations x the (, )-dependence of
Em
(r, , ) for any central V . Its
remaining r-dependence is determined by the energy eigenvalue equation and depends on details of V (r).
This is an eective one-dimensional problem.
6.4 Separation of Variables
Standard spherical polar coordinates (r, , ) are related to right-handed Cartesians by
(x, y, z) = r(sin cos , sin sin , cos )
and all space is covered by 0 r < , 0 (or 1 cos 1) and 0 < 2. The volume element
is
d = dxdy dz = r
2
sin dr d d = r
2
dr d
where the solid-angle element is d = d(cos ) d and
_
d = 4.
Simple calculation of derivatives (/x, /y, /z) leads to angular momentum operators in spherical
polars:

= ihe
i
_
cot

_
,

L
z
= ih

and so

L
2
= h
2
_
1
sin

_
sin

_
+
1
sin
2

2
_
.
Note that

=

L
2

L
2
z
h

L
z
.
The form of

L
z
is just as expected for the generator of rotations about the z-axis. Note that

2
=
1
r
2

r
_
r
2

r
_
+
1
r
2
_

1
h
2

L
2
_
.
(This suggests decomposition of p
2
= h
2

2
into radial and angular components as in classical mechanics:
P
2
= P
2
r
+L
2
/r
2
. But the radial part of
2
is not simply the square of the radial component of . Messiah
(p. 346) shows how to dene a self-adjoint p
r
obeying [ r,

P
r
] = ih with wavefunctions bounded at r = 0).
49
Thus in spherical polars =
Em
(r, , ) obeys

= im,

1
sin

_
sin

_
+
m
2
sin
2

= ( + 1)
and
h
2
2Mr
2
_

r
_
r
2

r
_
( + 1)
_
+V (r) = E.
These are eigenvalue equations for

L
z
,

L
2
and

H respectively with each simplied using its predecessor.
Hence this process corresponds to the separation of variables. Thus eectively we are putting
(r, , ) = f(r)()h().
The L
z
equation gives

Em
(r, , ) =
Em
(r, )e
im
and, since and + 2n are identied for all integer n, we have m = 0, 1, 2, . . . if is to be single-valued.
This implies = 0, 1, 2, . . . only. That is, orbital angular momentum L = r p of a quantum particle can
be only integer multiples of h.
6.5 The Legendre Equation
Next we look at the equation for (neglecting for the moment the fact that depends also on r as here it
still means (r, ) = f(r)()):
1
sin()
d
d
_
sin()
d
d
_

m
2

sin
2
()
+k = 0,
where k = l(l + 1),
To solve this equation it is convenient to change variables and introduce x = cos() (do not confuse it
with the original x). Then
d
d
=
d
dx
dx
d
=
d
dx
sin()
1
sin()
d
d
=
d
dx
and so
d
dx
_
sin
2
()
d
dx
_

m
2
sin
2
()
+ k = 0
ie
d
dx
_
(1 x
2
)
d
dx
_

m
2
1 x
2
+ k = 0
or
(1 x
2
)
d
2

dx
2
2x
d
dx
+ k =
m
2
1 x
2
.
This is clearly a very complicated equation.
Let us consider rst the case when m0. Then this equation becomes
(1 x
2
)
d
2

dx
2
2x
d
dx
+ k = 0
and is called the Legendre equation. Its solutions are called Legendre functions.
Let us solve our equation for by a power series expansion around x = 0 (note that x = 0 corresponds
to =

2
i.e. the equator).
50
So put
(x) =

n=0
a
n
x
n+
.
Actually, we do not need in this expression (ie we can put = 0 but ... it does not hurt to put it there
either).
Then

r=0
_
(r + )(r + 1)a
r
(1 x
2
)x
r+2
2a
r
(r + )x
r+
+ ka
r
x
r+

= 0
ie

r=0
(r + )(r + 1)a
r
x
r+2

r=0
[(r + )(r + + 1) k] a
r
x
r+
= 0.
Look at the powers of x. The two lowest ones are x

where = (2) and = (1). Their coecients

are respectively
( 1) a
0
= 0
( + 1)a
1
= 0
Then when = we have
( + 2)( + 1) a
2
[( + 1) k] a
0
= 0
and in general (for = + n)
( +n + 2)( +n + 1) a
n+2
[(n + )(n + + 1) k] a
n
= 0.
So we see that if we satisfy the rst two equations the remaining ones can be satised recurrsively; ie we
can use them to dene a
n+2
given a
n
. So we treat the last expression as the recurrence relation for a
n
.
To satisfy the rst two equations we note that if = 0 both a
0
and a
1
are arbitrary and the equations
are satised. Then, as the recurrence relation involves only a
k
with k diering by 2, a
2
can be expressed in
terms of a
0
, a
3
in terms of a
1
, a
4
in terms of a
2
and thus in terms of a
0
etc. Thus all the a
n
for n even are
expressible in terms of a
0
and all the ones with n odd, in terms of a
1
.
In both cases the recurrence relation is given by
a
n+2
=
n(n + 1) k
(n + 1)(n + 2)
a
n
.
Let us calculate a
n
for n even. We have
a
2
=
k
2
a
0
, a
4
=
23k
43
a
2
=
k(6k)
234
a
0
a
6
=
4 5 k
6 5
a
4
=
k(6k)(20k)
23456
a
0
etc
So if a
1
= 0 the solution is given by the a
0
series and is of the form
=
1
= a
0
_
1
k
2!
x
2

k(2 3 k)
4!
x
4

k(2 3 k)(4 5 k)
6!
x
6
...
_
.
The coecients of the odd series (ie the series starting with a
1
) take the form:
a
3
=
1 2 k
2 3
a
1
, a
5
=
3 4 k
4 5
a
3
=
(1 2 k)(3 4 k)
5!
a
1
itc
and so this series is given by
=
2
= a
1
x
_
1 +
1 2 k
3!
x
2
+
(1 2 k)(3 4 k)
5!
x
4
+ ...
_
.
51
The total solution is given by
=
1
+
2
,
where x = cos , and so is characterised by two arbitrary constants a
0
and a
1
.
Note:
A possible way of solving the rst two equations would involve = 1 and a
1
= 0. However, a little
thought shows that the resultant series gives again function
2
.
6.6 Convergence Problems - Legendre Polynomials
Look at the recurrence relations
a
n+2
a
n
=
(n + )(n + + 1) k
(n + + 2)(n + + 1)
so for large n
a
n+2
a
n
1
Hence the series becomes (for large n > N, for some N)
..... +a
N
(x
N
+ x
N+2
+ x
N+4
+ ....)
This series diverges at x
2
= 1 ie for x = 1.
However, x = cos() so x = 1 = 0 and x = 1 corresponds to = .
So if we want (r, ) = ()f(r) to be nite when = 0 or (ie on the z axis) we have to impose the
conditions that
lim
0,
() = lim
x1
(x) = nite.
However, we recall that k = l(l + 1), and so we see that for each value of one solution is a polynomial
and the other one diverges. So what are these polynomials? (they are polynomials as they involve only nite
series)
For = 0 (n even and the series starting with a
0
, or n odd and the series starting with a
1
) we have
a
n+2
=
n(n + 1) l(l + 1)
(n + 2)(n + 1)
a
n
so a
l+2
= 0 and we have a polynomial of degree l.
Such (polynomial) solutions of our equation are called Legendre polynomials. They are either odd or
even in x as they involve either odd (or even) powers of x.
Examples
l = 0
0
(x) = a
0
l = 1
1
(x) = a
1
x
l = 2
2
(x) = a
0
(1 3x
2
)
l = 3
3
(x) = a
1
x(1
5
3
x
2
)
We can choose a convenient normalisation - the conventional choice is
l
(1) = 1; this normalisation
xes the values of a
0
and a
1
in the expressions above. With this normalisation the polynomials are called
Legendre polynomials and are denoted by P
l
(x).
The lowest Legendre polynomials are therefore
P
0
(x) = 1
P
1
(x) = x
P
2
(x) =
1
2
(3x
2
1)
52
P
3
(x) =
1
2
(5x
3
3x)
P
4
(x) =
1
8
(35x
4
30x
2
+ 3)
P
5
(x) =
1
8
(63x
5
70x
3
+ 15x)
Recall, however, that is really also a function of r. Hence we can have (for m = 0)

E0
= f
E0
(r)P

(cos ),
where P

(x = cos ) is the familiar Legendre Polynomial of degree . (The other solution, Q

(x), as we have
argued, is singular at x = 1, or = 0, and so can be ignored).
6.7 Spherical Harmonics
Our previous results were obtained when we put m = 0 in our equation for :
(1 x
2
)
d
2

dx
2
2x
d
dx
+ k =
m
2
1 x
2

Then when we put m = 0, we eliminated dependence of on and our equation for became the Legendre
equation.
Let us now look at the case when m = 0.
To solve this equation let us recall that k = l(l + 1) and set
= (1 x
2
)
m
2
u.
Then
d
dx
= (1 x
2
)
m
2
du
dx
mx(1 x
2
)
m2
2
u
and
d
2

dx
2
= (1 x
2
)
m
2
d
2
u
dx
2
2mx(1 x
2
)
m2
2
du
dx
m(1 x
2
)
m2
2
u + m(m2)x
2
(1 x
2
)
m4
2
u
So
(1 x
2
)
d
2
u
dx
2
2mx
du
dx
mu +
m(m2)x
2
1x
2
u
2x
du
dx
+ 2mx
2 u
1x
2
+ l(l + 1)u
m
2
u
1x
2
= 0.
ie
(1 x
2
) u 2(m+ 1)x u + [l(l + 1) m(m+ 1)]u = 0.
But P
l
(x) satises
(1 x
2
)

P
l
2x

P
l
+ l(l + 1)P
l
= 0.
Dierentiate this equation once and obtain
2x
d
2
P
l
dx
2
+ (1 x
2
)
d
3
P
l
dx
3
2
dP
l
dx
2x
d
2
P
l
dx
2
+ l(l + 1)
dP
l
dx
= 0.
ie
(1 x
2
)
d
3
P
l
dx
3
4x
d
2
P
l
dx
2
+ [l(l + 1) 2]
dP
l
dx
= 0.
So we note that
dP
l
dx
solves the equation for m = 1.
53
It is easy to check that
d
2
P
l
dx
2
solves the equation for m = 2 and, in general, is given by
= A
l
(1 x
2
)
m
2
_
d
dx
_
m
P
l
(x) = A
l
P
m
l
(x),
where A
l
are some functions of r. These functions (for A
l
= 1) are called associated Legendre functions.
Special cases of associated Legendre functions:
As P
m
l
(x) = (1 x
2
)
m
2
d
m
dx
m
P
l
(x) we have
P
0
l
(x) = P
l
(x)
P
1
1
(x) = (1 x
2
)
1
2
= sin()
P
1
2
(x) = 3x(1 x
2
)
1
2
= 3 cos() sin()
so P
m
l
are polynomials in sin() and cos().
Moreover, choosing a normalisation constants of when A
l
= 1, appropriately, we have
Y
m
n
(, ) =

2n + 1
4
(n m)!
(n + m)!
P
m
n
(cos())e
im
which are called spherical harmonics. They form a complete set of functions on (, ), satisfy the orthog-
onality condition
_
2
0
d
_

0
sin() d Y
m1
n1
(, ) Y
m2
n2
(, ) =
n1n2

m1m2
and any function on (, ) can be expanded
f(, ) =

n,m
a
nm
Y
m
n
(, )
with the coecients of the expansion found from the orthogonality relations.
Returning to our problem we see that calling our A(r) = f
Em
(r)

Em
(r, , ) = f
Em
(r)P
m

(cos )e
im
.
Then f obeys the radial (energy) equation.
Writing f(r) = u(r)/r the radial equation simplies and u = u
E
(r) satises
h
2
2M
d
2
u
dr
2
+
_
V (r) +
( + 1) h
2
2Mr
2
_
u = Eu.
This is in the form of a one-dimensional problem with eective potential
V
1
= (r < 0) and V
1
= V (r) +
( + 1) h
2
2Mr
2
(r > 0),
which includes the boundary condition u(0) = 0 needed for niteness of
Em
at the origin. Notice the
additional centrifugal potential term.
The radial equation is independent of m so there is always (2 + 1)-fold degeneracy for given E and .
For a central potential the eigenfunctions of energy and angular momentum are then

Em
(r, , ) =
u
E
(r)
r
Y
m
(, ),
where the angular-momentum wavefunctions are the spherical harmonics dened before
Y
m
(, ) P
m

(cos ) e
im
.
54
Note that Y
m
satisfy

Y
m
=
_
( m)( m+ 1) hY
,m1
as well as both

L
2
Y
m
= ( + 1) h
2
Y
m
and

L
z
Y
m
= mhY
m
.
With P
0

(cos ) and conventional phase we have

Y
0
=
_
2 + 1
4
P

(cos ),
involving an ordinary Legendre Polynomial. Then Y
m
for m = 0 can be found by application of the
dierential operators

L

= ihe
i
(cot / i/).
Clearly Y
00
= 1/

4 and states with = 0 are spherically symmetric. Then from P

1
(cos ) = cos it
follows at once that
Y
10
=
_
3
4
cos and Y
11
= Y

1,1
=
_
3
8
sin e
i
.
Likewise from P
2
(cos ) = (3 cos
2
1)/2 we have Y
20
=
_
5/16(3 cos
2
1), from which

L

and

L
2

give
Y
21
= Y

2,1
=
_
15
8
sin cos e
i
and Y
22
= Y

2,2
=
_
15
32
sin
2
e
2i
.
Generally
Y
m
(, ) =

2 + 1
4
( m)!
( +m)!
P
m

(cos ) e
im
.
Moreover, it can be shown that
P
m

= P
m

.
Note also that for a given value of l the possible values of m are m = , + 1, ... 1, as for m >
d
+m
dx
+m
(x
2
1)

= 0.
The = 1, 2 examples above illustrate that Y
,m
= (1)
m
Y

m
. Note that Messiah (p. 495) and Schi
(p. 80) have unusual (and dierent) phase conventions, apparently clashing with that adopted for the ladder
operators.
The spherical harmonics are complete on the sphere; also it is useful to note the existence of a relation
cos Y
m
(, ) = A
m
Y
+1,m
(, ) +B
m
Y
1,m
(, )
where A and B are independent of (, ). For m = 0 this is just the recurrence formula for Legendre
polynomials.
Comment on Jargon;
is called the angular quantum number
m is called the magnetic quantum number (due to the coupling HL L
z
)
n; when we have discrete values of energy they can be ordered using an integer - normally denoted by
n which is then referred to as the principal quantum number.
More jargon:
The states corresponding to = 0, 1, 2, 3, 4, 5... are called s, p, d, f, g, h,.. wave states. The term
h
2
2M
(+1)
r
2
is called the centrifugal potential.
Under parity, r r, angles go and +. So cos changes sign and sin is unaected,
implying P
m

(1)
m
P
m

. With e
im
(1)
m
e
im
then we have simply Y
m
(1)

Y
m
, conrming
that angular momentum eigenstates have denite parity, equal to (1)

. Therefore, since r is unaected,

Em
(r) = (1)

Em
(r).
55
6.8 An Algebraic Approach to Angular Momentum
Previously we analysed the spectrum of angular momentum algebraically. We took

L
i
=
ijk
x
j
p
k
and showed
that
[

L
i
,

L
j
] = ih
ijk

L
k
and then having found that [

L
2
,

L
k
] = 0 we decided to use simultaneous eigenstates of

L
2
and

L
z

L
2
|, m = h
2
|, m

L
z
|, m = hm|, m.
By analysing various properties of such states we have found that = ( + 1), m = , .. where 2 is an
integer. So we have found more possibilities than we have here.
Clearly our cases = 0, 1, 2, ... m = , +1, ..0.. are the ones we have found before. So |, m are
represented, in the position representation, by our spherical harmonics Y
m
.
But what about the others?; those that correspond to half odd integer values of l? e.g. =
1
2
, m =
1
2
,
m =
1
2
or =
3
2
, m =
3
2
,
1
2
,
1
2
or
3
2
?
The non-integer values of m would make the wave function
Em
not single valued so for our orbital
angular momentum we have only integer values of . But the general theory allows for the existence of
another angular momentum - called spin. This new angular momentum is realised in nature; in fact, many
elementary particles, such as electron and proton, have it. This extra degree of freedom is not described in
terms of orbital wave functions Y
,m
(, ) but is described by the purely quantum mechanical spin operators

S
x
,

S
y
and

S
z
which satisfy
[

S
i
,

S
j
] = ih
ijk

S
k
.
For electron, proton, neutron

S
2
= s(s+1) h
2
where s =
1
2
. Thus the eigenvalues of

S
z
have two values
1
2
h;
we have two spin states of each electron. For an electron a complete set of commuting observables involves
its position operators x, y, z and

S
z
(

S
2
is xed). Its total angular momentum is

J
i
=

S
i
+

L
i
.
6.9 Radial Equation
With
Em
(r, , ) = u
E
(r)Y
m
(, )/r the radial function u = u
E
obeys
h
2
2ME
d
2
u
dr
2
+
_
V (r)
E
+
( + 1) h
2
2MEr
2
_
u = u
for r 0. The boundary condition u(0) = 0 applies and, with normalised spherical harmonics, we have
_

0
dr u

E
(r) u
E

(r) =
EE

to ensure that
_
d

Em

m
=
EE

mm
.
For continuous eigenvalues E replace
EE
(E E

).
The division by E = 0 makes plain the dimensionless variable kr where k
_
2M |E|/h. For non-
singular potentials where r
2
V (r) 0 as r 0 the centrifugal term dominates as r 0 and then
d
2
u
d(kr)
2

( + 1)
(kr)
2
u
with solutions u (kr)
+1
, (kr)

. Since = 0, 1, 2, . . . only the former is acceptable if u(0) = 0.

56
6.9.1 Free Particle
Here V = 0 and E > 0. The radial equation is
d
2
u
d(kr)
2
+
_
1
( + 1)
(kr)
2
_
u = 0.
For = 0 we have u
E0
(r) sin kr, rejecting the cosine that isnt zero at r = 0. For general the non-singular
solution is kr j

(kr), where j

is the (rst kind of) spherical Bessel function.

These functions can be shown to be given by:
j

(x) = x

1
x
d
dx
_

sin x
x
.
Clearly j
0
(x) = sin x/x and we can prove by induction that xj
l
, for l = 0, satises the equation above:
To see this dene
u
l
= x
l+1
_
1
x
d
dx
_
l
sinx
x
.
Then
u
l+1
=
du
l
dx

(l + 1)
x
u
l
.
So
du
l+1
dx
=
(l + 1)
x
du
l
dx
+
(l + 1)
x
2
u
l
+
d
2
u
l
dx
2
and
d
2
u
l+1
dx
2
=
d
3
u
l
dx
3

(l + 1)
x
d
2
u
l
dx
2
+ 2
(l + 1)
x
2
du
l
dx
2
(l + 1)
x
3
u
l
.
Putting all this into the equation we nd that
d
3
u
l
dx
3
+
_
1
l(l + 1)
x
2
_
du
l
dx
+ 2
l(l + 1)
x
3
u
l
However this expression vanishes as can be seen from dierentiating the equation for u
l
.
See Jackson (pps. 84-87) for more details.
A free-particle solution with denite momentum p = hk is (Sec 2.4)
E,p
(r) = expik r, which must
be expressible as a superposition of free angular-momentum eigenstates
Em
(r, , ) = j

(kr)Y
m
(, ).
Choosing z-axis along k there is no dependence and the link is through the well-known formula
exp(ikr cos ) =

=0
(2 + 1)i

(kr)P

(cos ).
6.9.2 Spherical Square Well
Here V = V
0
(r a).
The radial equation now takes the form
h
2
2M
d
2
u
dr
2
+
( + 1) h
2
2Mr
2
u = Eu 0 < r < a
and
h
2
2M
d
2
u
dr
2
+
( + 1) h
2
2Mr
2
u = (E V
0
)u r > a
The boundary conditions
u 0 r 0
57
1. u 0 r for bound states
2. u const r for cont. states
Consider rst the s wave states; i.e. = 0. For E < V
0
we have
for 0 < r < a
(
d
2
dr
2
+
2
)u = 0, =
_
2ME
h
2
i.e.
u = Acos(r) + Bsin(r)
while for r > a we have
(
d
2
dr
2

2
)u = 0, =

2M(V
0
E)
h
2
and so
u = Ce
r
+De
r
.
The boundary conditions impose D = 0 and A = 0 and we see that we have only odd solutions of the
corresponding one dimensional problem (where we had odd and even solutions).
Comments:
Of course r > 0 so the fact that we have odd solutions (for r < 0) is irrelevant
Recall that the number of bound states depends on
2MV0a
2
h
2
. Since we are permitting only odd states
there need not be any bound states (if V
0
is small enough) in contradistinction to the one-dimensional
case where there is always at least one state. For E > V
0
we have the continuous spectrum as in the
one dimensional case).
Next we look at the case of general . We put = r and note that our equation becomes
d
2
d
2
(f) + [1
( + 1)

2
]f = 0,
where 0 < <
a

and, as before, =
_
2ME
h
2
.
However, this is exactly like the free case discussed before; and its solutions are the spherical Bessels
functions j

() and n

(). The function j

() was dened in the free case; n

() is singular at the origin and

is more complicated. So the boundary condition at = 0 gives us
f = Aj

(r) for 0 < r < a

and for r > a we have
f = C j

(ir) + Dn

(ir) for r > a,

where the constants C and D are related by the requirement that f as r . This gives us one
condition between C and D. The continuity of f and its derivative at r = a gives us two further conditions
which together not only x C and D but also determine the energy levels. Thus in particular, it can be
shown that there is at least one p wave bound state if

_
2MV
0
a
2
h
2
< 2
and two if
2
_
2MV
0
a
2
h
2
< 3. etc
Comparing this result with the condition for an s wave we see that the minimum value of
2MV0a
2
h
2
for the
p wave is higher. This is because of the additional repulsion present due to the angular momentum barrier
(+1)
r
2
. As increases V
0
a
2
has to get larger for the system to have bound states for angular momentum .
58
6.10 Two Particle Systems
So far we have been looking at systems involving only one particle under the inuence of a xed potential.
However, our discussion generalises to two particles moving under the inuence of mutual forces.
A system of two particle can be described by two position vectors r
1
and r
2
and momenta p
1
and p
2
.
Let us assume that the particles have masses m
1
and m
2
, respectively. So
H =
1
2m
1
(p
1
)
2
+
1
2m
2
(p
2
)
2
+V (r
1
r
2
).
Such a system is invariant under translations:
r
i
r
i
+a
and also under rotations, if V (r
1
r
2
) = V (|r
1
r
2
|). In classical mechanics we can treat such a system as
corresponding to a one particle in a xed potential. We do this by introducing centre of mass, and relative
coordinates. To do this we dene
r = r
1
r
2
MR = m
1
r
1
+m
2
r
2
, M = m
1
+m
2
We can introduce similar variables in quantum mechanics.
We have
_

h
2
2m
1
(
1
)
2

h
2
2m
2
(
2
)
2
+V (|r
1
r
2
)|)
_
(r
1
, r
2
) = E(r
1
, r
2
),
in the Schr odinger representation. Here, (
i
)
2
=

2
x
(i)
k
x
(i)
k
summed over k.
But

x
(1)
i
=
m
1
M

R
i
+

x
i

x
(2)
i
=
m
2
M

R
i

x
i
we see that
1
m
1
(
1
)
2
+
1
m
2
(
2
)
2
=
1
m
1
_
m
1
M

R
i
+

x
i
_
2
+
1
m
2
_
m
2
M

R
i

x
i
_
2
=
1
M
(
c
)
2
+
1

2
where (
c
)
2
=

2
RiRi
, (
2
) =

2
xixi
and =
m1m2
m1+m2
is the reduced mass.
So the Schr odinger equation becomes
_

h
2
2M
(
c
)
2

h
2
2

2
+V (r)

= E
and so it is convenient to separate in R and r.
We write =
c
(R)(r) and obtain

h
2
2M

2
c

c
(R) = E
c

c
(R)
i.e. like the equation of a free particle of mass M = m
1
+m
2
and
[
h
2
2

2
+V (r)](r) = E

(r)
59
like a particle of reduced mass in a xed potential.
Note that the solutions of the free equation are

c
= Ae
i
PR
h
of energy E
c
=
P
2
2M
and we are left with having to solve the relative position problem and then the total
energy is the sum of the two.
Comment: Our procedure (of separating variables), once again, can be regarded as choosing an appro-
priate set of commuting observables. For if we dene P = p
1
+p
2
and p =
m1
M
p
1

m2
M
p
2
we have
[x
i
, P
j
] = 0, [x
i
, p
j
] = ih
ij
[R
i
, p
j
] = 0, [R
i
, P
j
] = ih
ij
and so we can choose as our basis the simultaneous eigenstates of

P and
h
2
2

2
+V (r).
6.10.1 Isotropic Simple-Harmonic Oscillator - once again
Here V =
1
2
M
2
r
2
. The spectrum is of bound states for E > 0 and the radial function u
E
(r) obeys

d
2
u
d(kr)
2
+
_
_
kr

_
2
+
( + 1)
(kr)
2
_
u = u,
where
2
2Ek
2
/M
2
, i.e. E = h/2. We seek a solution
u = (kr)
+1
exp(
(kr)
2
2
)F(kr),
where acceptable behaviour as r 0 and r is made explicit. A solution with F a polynomial
(and therefore harmless at 0, ) is found by the usual series method when = 2n + 3 for principal
quantum number n = 0, 1, 2, . . .. The degree of F is 0, 2, 4, . . ., and is equal to n . So energy levels are
E = E
n
= (n +
3
2
) h and there is degeneracy with respect to .
For n = 0 only = 0 is allowed and for n = 1 only = 1; in each case F = constant. For n = 2 the
possibilities are = 0 (F is quadratic) and = 2 (F = constant), and for n = 3 there is = 1 (F quadratic)
and = 3 (F = constant). Remembering m, the n = 0, 1, 2, 3 levels have degeneracy 1, 3, 6, 10 respectively;
for general n degeneracy is
1
2
(n + 1)(n + 2).
This oscillator is isotropic because r
2
= x
2
+ y
2
+ z
2
and so in Cartesians

H = ( h
2
/2M)
2
+ V is
a sum of terms for three independent one-dimensional simple-harmonic oscillators with the same frequency.
Each contributes (n
i
+
1
2
) h to E and n = n
x
+n
y
+n
z
. Counting degeneracy is straightforward: n = 0 is
all n
i
= 0; n = 1 = 1 +0 +0 (3 ways); n = 2 is 2 +0 +0 (3 ways) and 1 +1 + 0 (3 ways); n = 3 is 3 + 0 +0
(3) and 2 + 1 + 0 (6) and 1 + 1 + 1 (1) in complete agreement with our discussion at the beginning of this
chapter.
6.11 Hydrogen Atom
The simplest model of a one-electron atom is a point electron of charge e and mass M moving non-
relativistically in the inverse-square electrostatic (Coulomb) attraction of a point nucleus of charge +Ze
xed at the origin. Hydrogen has atomic number Z = 1 and so we have a three-dimensional potential
problem with V (r) = e
2
/r. There is a scattering sector to the energy spectrum for E > 0 and a bound-
state sector when E < 0. Here we deal with the latter.
With E = |E| and k
2
= 2M |E| /h
2
the radial wavefunction u = u
E
(r) obeys
d
2
u
d(kr)
2
+
_

kr

( + 1)
(kr)
2
_
u = u,
60
where

e
2
k
|E|
=
e
2
h

2M
|E|
.
Writing
u(r) = (kr)
+1
e
kr
F(kr),
with required behaviour as r 0 and r extracted (note u

u for large r), we nd that F(x) obeys

x
d
2
F
dx
2
+ 2[( + 1) x]
dF
dx
+ [ 2( + 1)]F = 0.
A routine series expansion nds (Laguerre) polynomial solutions (harmless as r 0 and r ) of degree
N = 0, 1, 2, . . . if = 2(N + +1). Therefore the bound-state energies for this model of the hydrogen atom
are
E
n
=
Me
4
2h
2
n
2
,
labelled by principal quantum number n = N + + 1 = 1, 2, 3, . . .. The ground state is at E
1
13.6 eV
and the innite set of levels packs closer as n increases, blending into the continuum at E = 0.
The unique ground state is n = 1, where N = = m = 0. All higher (excited) states are degenerate.
For n = 2 there is N = 1, = m = 0 plus N = 0, = 1 with m = 1, 0, +1. Indeed for every n
degeneracy is 1 + 3 + 5 + . . . + (2n 1) = n
2
. This high degree of accidental or dynamic degeneracy is
special to the Coulomb potential, although the isotropic simple-harmonic oscillator shows something similar.
It is symptomatic of conserved quantities beyond angular momentum and parity so far identied. For the
inverse-square force law there is the (Laplace-)Runge-Lenz vector, as described by H. Goldstein, Classical
Mechanics, Addison Wesley, 2nd Ed., 1980, p. 102 etc.
The lowest few wavefunctions
nm
(r, , ) = u
n
(r)Y
m
(, )/r are

100
=
1
_
a
3
0
exp(
r
a
0
),

200
=
1
_
8a
3
0
_
2
r
a
0
_
exp(
r
2a
0
)
and

21m
=
1
_
8a
3
0
r
a
0
exp(
r
2a
0
)
_
_
_
1

2
sine
i
cos

2
sin e
i
with m = 1, 0, +1 respectively, and where a
0
= h
2
/Me
2
is the Bohr radius.
The formula for the energy levels agrees quite well with experiment, being the same as that given by
the Bohr model. However n, the principal quantum number, is no longer the electrons orbital angular
momentum in units of h and has nothing to do with tting de Broglie waves into a circle. And while a
0
is
the radius of the lowest orbit of the Bohr atom, now we have r =
3
2
a
0
in the ground state.
Transitions E E

conserve energy by absorption and emission of radiation of characteristic frequency

= |E E

| /2h and wavelength

1
nn
= R

n
2
n
2

. The latter formula was discovered empirically

by spectroscopists before 1900. The Rydberg Constant is
R

Me
4
4h
3
c
= 10973731.571m
1
,
according to recent measurements of fundamental constants (see Physics Letters B, vol. 239, 12 Apr 1990,
page III.1).
A simple renement of this model treats the atom as two moving and interacting bodies. Then r
is a relative coordinate and otherwise results are identical except that M is replaced by reduced mass
Mm
0
/(M +m
0
) where m
0
is the mass of the nucleus. (See e.g. Schi pps. 88-90). For hydrogen corrections
are order M/m
0
1/2000.
61
Another renement recognises the intrinsic spin angular momentum
1
2
h of the electron. As a classical
spinning charge e its consequent magnetic moment is eh/2Mc. This interacts with magnetic elds, adding
terms to

H that tend to lift the degeneracy of its eigenvalues. Magnetic elds are provided by the spatial
motion of the charged nucleus relative to the electron (giving ne structure to the spectrum) and may be
provided by any intrinsic spin of the charged nucleus (hyperne structure). Hydrogen, with one proton of
spin
1
2
, shows both eects.
External static magnetic elds couple to the eective electric current of the orbital electron (producing
the (normal) Zeeman eect) and couple to its intrinsic magnetic moment to give the anomalous Zeeman
eect.
Relatively small eects can be treated perturbatively.
7 Conclusion
7.1 Summary
Observable quantities are associated with (self-adjoint) operators that work in a linear state space with
inner product. A measurement gives a (real) eigenvalue of the appropriate operator. State vectors are
linear combinations of its (orthogonal, complete) eigenvectors. A measurement is a random realisation of
an outcome (eigenvalue) and the relative probability of each is the squared modulus of the coecient of the
corresponding eigenvector in the basis expansion. After a measurement the state vector is changed to the
eigenvector belonging to the eigenvalue realised.
Quantisation is discreteness of eigenvalues and is a consequence of commutation rules for operators, where
h enters. Commutation rules and positivity also lead to the Uncertainty Principle, constraining statistical
scatter of results of mutually incompatible measurements on systems with identical state vectors.
Time-dependence of state vectors between measurements is determined by the Schr odinger Equation as
a unitary transformation. Again h enters and unitarity keeps consistency with measurement axioms. The
Schr odinger Equation and commutation rules give Ehrenfests theorems, which ensure the proper classical
limit.
States of denite energy are stable and external time-dependent perturbations cause transitions, with en-
ergy interchange. Transitions between quantised atomic and molecular levels involve emission and absorption
of energy as electromagnetic radiation of characteristic frequencies.
Wave mechanics represents a state vector in a basis of eigenstates of an operator measuring the systems
conguration-space coordinates. The Schr odinger Equation becomes the Schr odinger Wave Equation and the
squared modulus of its solution (wavefunction) gives a probability density for measurements of conguration-
space coordinates. For one particle these may coincide with coordinates of ordinary spatial position; this is
wave-particle duality.
The formalism gives predictions that agree with experiment although important questions of principle
remain, connected with measurement and the collapse of the state vector.
7.2 Measurement
Measurement gives an eigenvalue of a self-adjoint operator at random according to a certain probability
density and involves thereupon changing the state vector used to describe a quantum system from a general
superposition of eigenvectors to the single eigenvector belonging to the eigenvalue realised. In the 2-slit
experiment when one slit is closed or when an electron arrives at the detecting screen its position is recorded
and the state vector becomes a position eigenstate. This is called collapse of the state vector and is a
change not described by the electrons Schr odinger Equation.
But the electrons detection involves its interaction with a detector, colliding with constituent atoms
which are excited and then decay, emitting photons which typically travel to and interact with atoms of a
camera lm. Later this is developed and xed by chemicals, then later still photons bounce o the lm and
into your eye, where they interact with the retina to cause electrical impulses to travel via optic nerve to
brain. You see the electrons position.
62
We think at rst of the quantum system as just the electron and the detector as being a separate
classical recording device. But the chain of events involved in the observation of the electrons position is
clearly, from the description above, a set of physical processes at the microscopic level, each of which should
(if we had the computational power) be described by quantum mechanics.
So perhaps, more correctly, part at least of the chain of detection apparatus should be included in a
vastly more complicated Schrodinger Equation for the experiment, when the borderline between quantum
and classical, and so the wavefunction collapse, occurs at a much later stage.
But where? At the camera? When the lm is developed? When light bounces o it? When it enters
your eye? When youre conscious of it? When you tell me about it? The answer is not obvious.
The dilemma is sharpened by the Schrodingers Cat thought experiment. A cat is shut in a box for a
set time with a radioactive nucleus which has exactly a 50:50 chance of decaying during that period. If (and
only if) it does decay then a bulb of cyanide is broken and the cat dies. When you open the box you see
with equal probability either a cooling corpse or a frisking feline (Polkinghorne, p. 62).
The state vector of the unobserved quantum system is an equal mixture of two eigenstates decayed
and undecayed, or equivalently dead and alive. When the box is opened uncertainty has gone. But surely
the cat alone is competent to tell at least whether its dead or alive? Surely without being seen the bulb is
either broken or intact? Just where in the chain of events does observation occur and collapse of the state
vector happen?
One possible answer is that collapse occurs at a point where the system becomes so complex that irre-
versible phenomena become important. This tries to identify the quantum/classical interface and is fashion-
able now that deterministic chaos and properties of cellular automata have come to the fore.
But this is so far only a vague idea, and doesnt answer other questions that we think have some meaning:
Whats the electron really doing before it hits the screen? How is it determined which position eigenstate
it collapses to? How is it xed when the nucleus decays? Whats happening in the world while were not
looking?
We assume that an objective world indeed exists thats the simplest hypothesis consistent with every-
day experience, after all, and it seems to work in everyday aairs. But it may be that at a submicroscopic
level things are arranged so that with the tools we can make from the physical processes available we simply
cant resolve whatever mechanism may guide state vector collapse, any more than without a microscope you
can see the cells of the skin on your hand. Maybe down there eectively theres no real world to be seen
because theres no way of seeing it!
There are substantial issues here. Einstein talked of God playing dice and these questions are in focus
again with technological advances that allow clear-cut experiments. Aspect has done a version of the EPR
thought experiment (Phys. Rev. Letters 49, 91 & 1804(1982)) and for example Nagourney et al. (Phys.
Rev. Letters 56, 2797(1986)) have watched a single calcium ion making quantum jumps.
More modern texts like Ballentine, and Sudbery, give some attention to these issues, and the book by
Polkinghorne and those he cites on his p. 97 explain more and describe alternative theories for whatever may
underlie the fall of the dice.
7.3 And Theres More . . .
Only the simplest atom (hydrogen) has been treated, in the simplest model. But the theory clearly predicts
that such typical atoms, in their ground state, are stable. The most obvious classical catastrophe of the
Rutherford atom is avoided.
But most atoms have many electrons around their tiny nucleus, and would seem to be mostly empty
space. The next step is to understand how multiple-electron atoms are arranged and how aggregates of
atoms in bulk matter keep themselves apart why dont they interpenetrate and collapse? To answer this
needs development of the quantum mechanics of many-body systems and in particular Fermi-Dirac Statistics
and the Exclusion Principle. And if you want to learn more about this and other related concepts do come
to the course on Advanced Quantum Theory.
63