Foundations of Quantum Mechanics

Lecture notes

Fabio Deelan Cunden, April 23 2019

C ONTENTS

1. Tests, probabilities and quantum theory 2

2. The Old Quantum Theory 3
3. Bell’s inequalities 6
4. Two-level systems 9
5. The Schrödinger equation and wave mechanics 12
6. Exactly solvable models 18
7. Matrix Mechanics 35
8. Mathematical Foundations of Quantum Mechanics 38
9. Angular momentum and spin 49
10. Identical particles 59
References 60

Note. I created these notes for the course ACM30210 - Foundations of Quan-
tum Mechanics I taught at University College Dublin in 2018. The exposition
follows very closely the classical books listed at the end of the document. You can
help me continue to improve these notes by emailing me with any comments or
corrections you have.
2

1. Tests, probabilities and quantum theory

Quantum theory is a set of rules allowing the computation of probabilities for

the outcomes of tests which follow specified preparations.
A preparation is an experimental procedure that is completely specified, like
a recipe in a good cookbook. Preparation rules should be unambiguous, but they
may involve stochastic processes, such as thermal fluctuations, provided that the
statistical properties of the stochastic process are known. A test starts like a prepa-
ration, but it also includes a final step in which information, previously unknown,
is supplied to an observer.
In order to develop a theory, it is helpful to establish some basic notions and
terminology. A quantum system is an equivalence class of preparations. For exam-
ple, there are many equivalent macroscopic procedures for producing what we call
a photon, or a free hydrogen atom, etc. While quantum systems are somewhat elu-
sive, quantum states can be given a clear operational definition, based on the notion
of test. Consider a given preparation and a set of tests. If these tests are performed
many times, after identical preparations, we find that the statistical distribution of
outcomes of each test tends to a limit: each outcome has a definite probability.
We can then define a state as follows: a quantum state is characterised by the
probabilities of the various outcomes of every conceivable test.
This definition is highly redundant. We shall see that these probabilities are
not independent. One can specify – in many different ways – a restricted set of
tests such that, if the probabilities of the outcomes of these tests are known, it
is possible to predict the probabilities of the outcomes of every other test. (A
geometric analogy is the definition of a vector by its projections on every axis.
These projections are not independent: it is sufficient to specify a finite number of
them, on a complete, linearly independent set of axes.)
As a simple example of definition of a state, suppose that a photon is said to
have right-handed polarisation. Operationally, this means that if we subject that
photon to a specific test (namely, a quarter wave plate followed by a suitably ori-
ented calcite crystal) we can predict with certainty that the photon will exit in a
particular channel. For any other test, consisting of arbitrarily arranged calcite
crystals and miscellaneous optically active media, we can then predict probabili-
ties for the various exit channels. (These probabilities are computed in the same
way as the classical beam intensities.) Note that the word ‘state’ does not refer
to the photon by itself, but to an entire experimental setup involving macroscopic
instruments.
The essence of quantum theory is to provide a mathematical representation
of states, together with rules for computing the probabilities of the various out-
comes of any test. This set of rules can be consistently organised assuming that
 3

there is associated with each physical system a set of quantities, constituting a non-
commutative algebra in the technical mathematical sense, the elements of which
are the physical quantities themselves.

2. The Old Quantum Theory

There are two main groups of experimental phenomena which are inconsistent
with classical physics, namely:
(i) when the internal energy of an atom changes, owing to emission or ab-
sorption of light, it does not do so evenly or continuously but in ‘quan-
tised’ steps;
(ii) a beam of electrons exhibits interference phenomena similar to those of a
light wave.
(i) Quantum states. A great wealth of experimental data, chiefly derived from
spectroscopy, shows that an atom cannot exist in states of continuously varying
energy but only in different discrete states of energy, referred to as ‘discrete energy
levels’. These levels and the spacings between them are different for the various
chemical elements, but are identical for all atom of the same element. The emitted
energy (in the form of light) is confined to spectral lines of sharply defined fre-
quencies, and, moreover, the frequencies are not related to one other by integral
factors, as overtone, but instead show an interesting additive relation, expressed in
the Ritz-Rydberg combination principle. The change from level E2 to another E1
is associate with the emission if E2 > E1 (or absorption if E2 < E1 ) of light,
whose frequency ν is determined by the relation

(2.1) E2 − E1 = hν,

where h is a universal constant known as Planck’s constant. The above relation

is the Bohr frequency rule. It is important to notice that the experimentally mea-
surable quantities are the spacings (frequencies) and not the single energy levels!
Note that Bohr’s assumption is not only in agreement with the existence of sharp
spectral lines, but contains in addition the combination principle. If we order the
energy levels E0 < E1 < E2 · · · , then in accordance to (2.1) each frequency is the
difference of two terms
1
ν(i → j) = (Ei − Ej ) (i > j).
h
Consequently, generally speaking there will occur in addition to the frequency
ν(i → j), ν(j → k) the frequency

(2.2) ν(i → k) = ν(i → j) + ν(j → k)

obtained from them by addition (combination).

4

(ii) Interference of waves. It is well known that light waves when reflected
or refracted from a regularly spaced grating interfere and form what is called a
‘diffraction’ pattern visible on a screen. This phenomenon was used, for instance,
to prove the wave nature of X-rays. For this purpose the regular spacing is provided
by the atoms of a crystal, as an ordinary grating would be too coarse. From the
diffraction pattern one can determine the wavelength of the light. Davisson and
Germer (1927) carried out a similar experiment with a beam of electron, all having
the same velocity, or momentum p, passing through a crystal, etc. The diffraction
pattern so found is very similar to that produced by X-rays. This experiment shows
that waves are associate with a beam of electrons. The wavelength of the electrons
is then found to be inversely proportional to the momentum λ ∝ 1/p, the slower
the electrons the longer is the wavelength. The proportionality constant can be
measured and it turns out to be equal to Planck’s constant
h
(2.3) λ= .
p
This relation was suggested by de Broglie in his doctoral thesis (1924), and later
confirmed in the experiment by Davisson and Germer. This relation was also fun-
damental in the formulation of the most famous Heisenberg’s uncertainty relation
(2.4) ∆x∆p ≥ ~/2.

E XAMPLE 2.1. Hydrogen atoms in a discharge lamp emit a series of lines in

the visible part of the spectrum. This series is called the Balmer Series after the
Swiss teacher Johann Balmer. Here are the lines observed by Balmer (in convenient
units)
5 3 21 2
, , ,
36 16 100 9
In 1885, Balmer found by trial and error a formula to describe the wavelengths
of these lines. Balmer suggested that his formula may be more general and could
describe spectra from other elements. Then in 1889, Johannes Robert Rydberg
found several series of spectra that would fit a more general relationship, similar to
Balmer’s empirical formula. A sample of lines is
11 9 5 7 16 1 5 3 21 2 3 8 15 24 35
··· , , , , , , , , , , , , , , , ,···
|900 400 144{z 144 225 12} |36 16{z100 9} |4 9 16{z 25 36}
infrared visible ultraviolet

The actual spectral lines are given by the numbers above multiplied by the
Rydberg constant
2π 2 µe2
(2.5) R=
h3 c
where µ is the reduced mass of the electron and the nucleus. The numerical value
of the constant R agrees with values obtained from spectroscopic data. The values
 5

for hydrogen, ionized helium, and infinite nuclear mass are

RH = 109677.759 ± 0.05 cm−1

RHe = 109722.403 ± 0.05 cm−1
R∞ = 109737.42 ± 0.06 cm−1 .

E XERCISE 2.1. Try to find Balmer’s empirical formula. If you succeed, try to
find Rydberg’s generalisation to explain the above sequence. Hint: rearrange the
sequence as a table
3
4
8 5
9 36
15 3 7
16 16 144
24 21 16 9
25 100 225 400
35 2 1 5 11
36 9 12 144 900
.. ..
. .

TABLE 1. Relations between experimental interpretation and the-

oretical inferences.

Diffraction (Young 1803, Laue 1912) Electromagnetic waves

Blackbody radiation (Planck 1900) Electromagnetic quanta
Photoelectric effect (Einstein 1904)
Combination principle (Ritz-Rydberg 1908) Discrete values
Franck-Hertz experiment (1913) for physical quantities
Stern and Gerlach (1922)
Davisson and Germer (1927) Matter waves

2.1. Quantisation rules. In 1915 W. Wilson and A. Sommerfeld discovered

independently a simple method of quantisation which was soon applied in the dis-
cussion of several atomic phenomena with good success.
The first step of the method consists in solving the classical equations of mo-
tion defined by the classical Hamiltonian H = H(p, q). The assumption is then
introduced that only those classical orbits are allowed as stationary states for which
the following conditions are satisfied
I
(2.6) pk dqk = nk h, nk an integer,
6

for all k = 1, . . . , N (N being the number of degrees of freedom of the system).

These integrals, which are called action integrals, can be calculated only for inte-
grable systems or, in the language of old quantum theory, ‘conditionally periodic
systems’.

E XAMPLE 2.2 (Quantisation of the harmonic oscillator). A classical harmonic

oscillator is defined by the Hamiltonian (total energy)
1 2 1
(2.7) H(p, q) = p + mω 2 q 2 , (p, q) ∈ R2 .
2m 2
The solution of the Hamilton equations
(
dq ∂H 1
dt = ∂p = m p
(2.8) dp ∂H 2
dt = − ∂q = −mω q

are oscillations about the point (p, q) = (0, 0)

(
q(t) = A sin(ωt + φ)
(2.9)
p(t) = mωA cos(ωt + φ)

where A and φ are arbitrary constants. It is easy to see that the (p, q)-plane is
partitioned into ellipses of equation
q2 p2
+ = 1.
A2 m2 ω 2 A2
The action integral can be evaluated explicitly
I
(2.10) pdq = πmωA2 = hn.

Applying the quantisation rule, we see that the amplitude A is restricted to the
quantised values An = (hn/πmω)1/2 . The corresponding energy values are
1
(2.11) En = mω 2 A2n = (h/2π)ωn = ~ωn.
2
The energy levels predicted by the old quantum theory are integral multiples of ~ω.

3. Bell’s inequalities

3.1. An experiment with polarizers. In a beam of monochromatic (fixed

color) light we put a pair of polarizing filters, each of which can be rotated around
the axis formed by the beam. As is well known, the light emerging from both fil-
ters changes in intensity when the filters are rotated relative to each other. Starting
from the orientation where the resulting intensity is maximal, and rotating one of
the filters through an angle α, the light intensity decreases with α, vanishing for
α = π/2. If we call the intensity of the beam before the filters I0 , after the first
 7

1
I1 , and after the second I2 , then I1 = 2 I0 (we assume the original beam to be
unpolarized), and
(3.1) I2 = I1 cos2 α.
So far the phenomenon is described well by classical physics. During the last cen-
tury, however, it has been observed that for very low intensities (monochromatic)
light comes in small packages, which were called photons, whose energy depends
on the color, but not on the total intensity. So the intensity must be proportional to
the number of these photons, and formula (3.1) must be given a statistical mean-
ing: a photon coming from the first filter has probability cos2 α to pass through
the second. Thinking along the lines of classical probability, we may associate to
a polarization filter in the direction α a random variable Pα taking value 0 if the
photon is absorbed by the filter and value 1 if the photon passes through. Then,
for two filters in the direction α and β these random variables should correlated as
follows
(3.2)
1 1
EPα = EPβ = and EPα Pβ = P(Pα = 1 and Pβ = 1) = cos2 (α − β).
2 2
Here we hit on a difficulty: the function on the right hand side is not a possible cor-
relation function! To see this, consider three polarizing filters having polarization
directions α1 , α2 , and α3 respectively. They should give rise to random variables.
P1 , P2 and P3 satisfying
1
EPi Pj = cos2 (αi − αj ).
2
P ROPOSITION 3.1 (Bell’s three variable inequality). For any three 0-1-valued
random variables P1 , P2 , and P3 on the same probability space, the following
inequality holds:
P(P1 = 1, P3 = 0) ≤ P(P1 = 1, P2 = 0) + P(P2 = 1, P3 = 0)

P ROOF.
P(P1 = 1, P3 = 0) = P(P1 = 1, P2 = 0, P3 = 0) + P(P1 = 1, P2 = 1, P3 = 0)
≤ P(P1 = 1, P2 = 0) + P(P2 = 1, P3 = 0).


In our example,
1
P(Pi = 1, Pj = 0) = P(Pi = 1) − P(Pi = 1, Pj = 1) = sin2 (αi − αj ).
2
Bell’s inequality thus reads
1 1 1
sin2 (α1 − α3 ) ≤ sin2 (α1 − α2 ) + sin2 (α2 − α3 )
2 2 2
which is clearly violated for the choices α1 = 0, α2 = π/6 and α3 = π/3.
8

The above calculation could be summarized as follows: we are looking for a

family of 0-1-valued random variables (Pα )0≤α<π on the same probability space
with P(Pα = 1) = 1/2, satisfying
P(Pα 6= Pβ ) = sin2 (α − β).
Now, on the space of 0-1-valued random variables on a probability space, the func-
tion (X, Y ) 7→ P(X 6= Y ) equals the L1 -distance of X and Y . On the other
hand, the function (α, β) 7→ sin2 (α − β) does not satisfy the triangle inequality.
Therefore no family (Pα )0≤α<π exists which meets the above requirement.

3.2. An improved experiment. On closer inspection the above example is

not very convincing. Indeed, when two polarizers are arranged on the optical
bench, why should not the random variable for the second polarizer depend on
the angle of the first?
In fact we can do a better experiments using a clever technique from quantum
optics. It is possible to build a device that produces pairs of photons, such that the
members of each pair move in opposite directions and show opposite behaviour
towards parallel polarization filters: if one passes the filter, then the other is surely
absorbed. With these photon pairs, the very same experiment can be performed,
but this time the polarizers are far apart, each one acting on its own photon. (This
is the optical version of the Gedankenexperiment was proposed and discussed by
Einstein, Podolski, Rosen and Bohm.) The same outcomes are found, violating
Bell’s three variable inequality.

3.3. The decisive experiment. Advocates of classical probability could still

find serious fault with the argument given so far. So the argument has to be tight-
ened still further. This brings us to the experiment (conceptually devised by Bell
in the sixties) which was actually performed by A. Aspect and his team at Orsay,
Paris, in 1982. In this experiment a random choice out of two different polarization
measurements was performed on each side of the pair-producing device, say in the
direction α1 or α2 on the left and in the direction β1 or β2 on the right, giving rise to
four random variables P1 = P (α1 ), P2 = P (α2 ) and Q1 = Q(β1 ), Q2 = Q(β2 ),
two of which are measured and compared at each trial.

P ROPOSITION 3.2 (Bell’s four variable inequality). For any four 0-1-valued
random variables P1 , P2 , Q1 , and Q2 on the same probability space, the following
inequality holds:
P(P1 = Q1 ) ≤ P(P1 = Q2 ) + P(P2 = Q1 ) + P(P2 = Q2 ).

Quantum mechanics predicts, and the experiment of Aspect showed that

1
P(P (α) = Q(β) = 1) = P(P (α) = Q(β) = 0) = sin2 (α − β).
2
 9

Hence
P(P (α) = Q(β)) = sin2 (α − β).
So Bell’s four variable inequality reads
sin2 (α1 − β1 ) ≤ sin2 (α1 − β2 ) + sin2 (α2 − β1 ) + sin2 (α2 − β2 ),
which is clearly violated for the choices α1 = 0, α2 = π/3, β1 = π/2 and
β2 = π/6.
There does not exist, on any classical probability space, a quadruple P1 , P2 ,
Q1 , and Q2 of random variables with the correlations measured in this experiment.

4. Two-level systems

Consider the historic Stern-Gerlach experiment whose purpose was to de-

termine the magnetic moment of atoms, by measuring the deflection of a neutral
atomic beam by a magnetic field. If the mean magnetic field is oriented along the
e1 -axis, i.e. B̄ = Be1 , then the classical equations of motion predict a deflection
of the beam along the e1 -axis proportional to µ1 = µ · e1 . The surprising result
found by Gerlach and Stern was that µ1 could take only two values, ±µ.
This result is extremely surprising from the point of view of classical physics,
because Gerlach and Stern could have chosen different orientations for their mag-
net, for example e2 and e3 , making angles of ±120◦ with e1 to measure µ2 = µ·e2
or µ2 = µ · e2 , respectively. As the outcome of the experiment cannot be af-
fected by merely rotating the magnet, the would have found, likewise, µ2 = ±µ or
µ3 = ±µ. This creates, however an apparent contradiction because
µ1 + µ2 + µ3 = µ · (e1 + e2 + e2 ) = 0.
Obviously, µ1 , µ2 and µ3 cannot all be equal to ±µ, and also sum up to zero.
Of course, it is impossible to measure in this way the values of µ1 , µ2 and µ3
of the same atom (this is purely classical impossibility). Quantum theory tells that
this is not a defect of this particular experimental method for measuring a magnetic
moment: no experiment whatsoever can determine µ1 , µ2 and µ3 simultaneously.

E XAMPLE 4.1. Consider an experiment with sequential Stern-Gerlach (SG)

apparata. We measure the magnetic moment (spin) Sz of a neutral beam in the
direction ez . Two beams of equal intensity emerge from the apparatus.
Suppose that we block the Sz = −1/2 beam (the bottom beam say) and we
let only the 1/2 beam go into a second identical SG apparatus. This apparatus lets
those out the top output and nothing comes out in the bottom output. We say that
the states with spin +1/2 have no component along Sz = −1/2.
Now, suppose that we measure the Sx component of the Sz = +1/2 beam.
Classically an object with angular momentum along the z axis has no component
10

of angular momentum along the x axis, these are orthogonal directions. Here how-
ever, about half of the beam exits through the top Sx = +1/2, and the other half
through the bottom Sx = −1/2.
Now we block the top beam and we only have an Sx = −1/2 output. The
beam is fed into a SG along ez . We find half of the particles make it out of the
third SG with Sz = +1/2 and the other half with Sz = −1/2.

4.1. Spin and Pauli matrices. One can consider experiments with a few Stern-
Gerlach apparata in series measuring the magnetic moment (the spin) in the direc-
tions ex , ey , ez , and so on. A mathematical framework consistent with the the
outcomes of the experiments and their probabilities is as follows. The experiment
suggests that the spin state can be described using normalised vectors in C2 . In
particular we can describe a state using two basis vectors, for instance
(4.1) |↑z i and |↓z i .
The first corresponds to a spin +1/2 in the z-direction (‘spin up along z’). The
second corresponds to the state of spin −1/2 ‘spin down along z’. A magnet
measuring the spin in the z direction is an operator Sz for which |↑z i and |↓z i are
eigenvectors with eigenvalue ±1/2, respectively:
1
(4.2) Sz |↑z i = + |↑z i
2
1
(4.3) Sz |↓z i = − |↓z i .
2
It is customary to represent the two vectors in the computational basis
   
1 0
(4.4) |↑z i = , |↓z i = .
0 1
A state can be represented as a superposition (i.e. normalised linear combination)
 
a
(4.5) |ψi = a |↑z i + b |↓z i = , a, b, ∈ C, |a|2 + |b|2 = 1.
b
The probability to measure a state φ given that the system is prepared in the state
ψ is |hφ|ψi|2 . The state |↑z i entering the second Stern-Gerlach must have zero
overlap with |↓z i since no such down spins emerge. Moreover the overlap of |↑z i
with itself must be one, as all states emerge from the second Stern-Gerlach top
output. Indeed,
   
0 1
h↑z |↓z i = (1 0) = 0, h↑z |↑z i = (1 0) = 1, etc.
1 0
The operator Sz can be represented as the 2 × 2 matrix:
 
1 1 0
(4.6) Sz =
2 0 −1
 11

When we measure the z-component of a state |↑x i we get |↑z i or |↓z i with equal
probability 1/2; hence
1
(4.7) |h↑z |↑x i| = |h↓z |↑x i| = √ .
2
Therefore we can write
1 1
(4.8) |↑x i = √ |↑z i + √ eiδ1 |↓z i .
2 2
The state |↓x i must be orthogonal to |↑x i; this leads to
1 1
(4.9) |↓x i = √ |↑z i − √ eiδ1 |↓z i .
2 2
We can now give a matrix representation of Sx :
0 e−iδ1
 
1
(4.10) Sx = .
2 eiδ1 0
A similar argument with Sx replaced by Sy leads to
0 e−iδ2
 
1
(4.11) Sy = .
2 eiδ2 0
Is there any way of determining δ1 and δ2 ? We can consider a sequential Stern-
Gerlach experiment with a measure along ex followed by a measure along ey ,
leading to
1
(4.12) |h↑y |↑x i| = |h↓y |↑x i| = √ .
2
Thus we obtain
1  
i(δ1 −δ2 )  1
(4.13) 1 ± e = √
2 2
which is satisfied only if
δ2 − δ1 = ±π/2.
We see that the matrix elements of Sz , Sx and Sy cannot all be real. The standard
representation is δ1 = 0 and δ2 = π/2 so that Si = 12 σi , i = x, y, z, with
     
0 1 0 −i 1 0
(4.14) σx = , σy = , σz = .
1 0 i 0 0 −1
E XERCISE 4.1. (1) Verify the following commutation relations of the
Pauli matrices σx , σy and σz .
[σx , σy ] = iσz , [σy , σz ] = iσx , [σz , σx ] = iσy .
(2) Compute the anticommutators {σx , σy }, {σy , σz }, {σz , σx }.
(3) Compute σx2 , σy2 and σz2 .
(4) Show that S 2 = Sx2 + Sy2 + Sz2 is a multiple of the identity operator.
12

The description and analysis of spin states that point in arbitrary directions, as
specified by a unit vector n goes as follows. Let
n = (nx , ny , nz ) = (sin θ cos φ, sin θ sin φ, cos θ) , 0 ≤ θ ≤ π, 0 ≤ φ < 2π,
be a unit vector. Here θ and φ are the familiar polar and azimuthal angles. We can
consider S = (Sx , Sy , Sz ), and define the spin operator Sn along n as
Sn = n · S = nx Sx + ny Sy + nz Sz .
E XERCISE 4.2. (1) Check that, just like Sx , Sy and Sz , the eigenvalues
of Sn are ±1/2.
(2) Show that the eigenvectors |±; ni of Sn can be written as
θ θ
(4.15) |+; ni = cos |↑z i + sin eiφ |↓z i
2 2
θ θ
(4.16) |−; ni = sin |↑z i − cos eiφ |↓z i .
2 2
The unit vector n denotes a location on the unit sphere, or ‘Bloch sphere’.
Each point on the surface of the sphere represents a (pure) state, i.e. a normalised
vector |ψi ∈ C2 . For example, |↑z i corresponds to the north pole (θ = 0), etc.

5. The Schrödinger equation and wave mechanics

In the first half of 1926, E. Schrödinger showed that the rules of quantisations
can be replaced by another postulate, in which there occurs no mention of whole
numbers. Instead, the introduction of integers arises in the same way as, for exam-
ple, in a vibrating string, for which the number of nodes is integer.
Schrödinger postulated a sort of wave equation and applied it to a number of
problems, including the hydrogen atom and the harmonic oscillator.
Schrödinger’s approach to the dynamics differs from that of classical mechan-
ics in its aims as well in its method. Instead of attempting to find equations, such
as Newton’s equations, which enable a prediction to be made of the exact positions
and momenta of the particles of a system at any time, he devised a method of cal-
culating a function Ψ of the coordinates and time (and not the momenta), with the
aid of which, probabilities for the coordinates or other quantities can be predicted.
The Schrödinger equation enables to determine a certain function Ψ (of the
coordinates and time) called wavefunction or probability amplitude. The square
of the modulus of a wavefunction is interpreted as a probability measure for the
coordinates of the system in the state represented by this wavefunction.
Besides yielding the probability amplitude Ψ, the Schrödinger equation pro-
vides a method of calculating values of the energy of the stationary states of the
system. No arbitrary postulates concerning quantum numbers are required in the
calculations; instead, integers enter automatically in the process of finding satis-
factory solutions of the wave equation.
 13

For our purposes, the Schrödinger equation, the auxiliary restrictions upon the
wavefunction and its probabilistic interpretation are conveniently taken as funda-
mental postulates.

5.1. The Schrödinger equation. Consider a system with one degree of free-
dom, consisting of a particle of mass m that can move along a straight line, and let
us assume that the system is further described by a potential energy function V (x)
throughout the region −∞ < x < +∞. The classical Hamiltonian of the system
is
p2
(5.1) H(p, x) = + V (x)
2m
and by the conservation of energy the Hamiltonian is constant on solutions of the
Hamilton equations:
H(p, x) = W.
∂
Schrödinger’s idea was to replace p by the differential operator ~i ∂x and W by
~ ∂
− i ∂t , and introduce the function Ψ(x, t) on which these operators can operate
 
~ ∂
H , x Ψ(x, t) = W Ψ(x, t).
i ∂x
For this system the Schrödinger equation is

~ ∂Ψ(x, t) ~2 ∂ 2 Ψ(x, t)
(5.2) − =− + V (x)Ψ(x, t).
i ∂t 2m ∂ 2 x
R EMARK . The analogy between the terms in the wave equation and the clas-
sical energy equation has only a formal significance.

5.2. The time-indepedent Schrödinger equation. In order to solve (5.2) let

us (as usual in the analysis of partial differential equations) first study the non-
trivial solutions Ψ (if any exist) which can be expressed as product of two func-
tions, one involving time alone and the other the coordinate alone:

(5.3) Ψ(x, t) = ψ(x)ϕ(t).

On introducing this in Eq. (5.2) we get

~2 ∂ 2 ψ(x)
 
1 ~ 1 ∂ϕ(t)
− 2
+ V (x)ψ(x) = −
ψ(x) 2m ∂ x i ϕ(t) ∂t
The right-hand side of this equation is a function of time t alone and the left-hand
side a function of x alone. It is consequently necessary that the quantity to which
each side is equal to be dependent on neither x nor t; that is, that it be a constant.
14

Let us call it E. Then,

~ ∂ϕ(t)


 − = Eϕ(t)
 i ∂t

(5.4)
 2 2
− ~ ∂ ψ(x) + V (x)ψ(x) = Eψ(x).


2m ∂ 2 x
The second equation is customarily written as a spectral problem for the Schrödinger
~2 d2
operator 2m dx2
+ V (x), i.e.,
~2 d2
 
(5.5) − + V (x) ψ(x) = Eψ(x).
2m dx2
Eq. (5.5) is often called time independent Schrödinger equation, or amplitude equa-
tion, inasmuch ψ(x) determines the amplitude of Ψ(x, t).
It is found that typically the equation possesses various solutions, correspond-
ing to various values of the constant E. Let us denote these values of E by attach-
ing a subscript n, and similarly the amplitude corresponding to En as ψn (x). The
corresponding equation for ϕ(t) can be integrated at once to give
En t
(5.6) ϕn (t) = e−i ~ .
The general solution of the Schrödinger equation is a linear combination of all the
particular solutions with arbitrary coefficients
X X En t
(5.7) Ψ(x, t) = an Ψn (x, t) = an ψn (x)e−i ~ .
n n

5.3. Wave functions: discrete and continuous sets of energy values. It is

found that satisfactory solutions ψn (x) exist only for certain values of the param-
eter En . The values En are called eigenvalues of the Schrödinger operator. It will
be shown later that the physical interpretation of the wavefunction requires that the
values En represent the energy of the system in its various stationary states.
In addition to be a solution of the Schrödinger equation, the wavefunction must
be single-valued, continuous and square integrable.
For a given system, the energy levels En may occur only as a set of discrete
values, or as a set of values covering a continuous range, or as both. From anal-
ogy with spectroscopy it is often said that in these three cases the energy values
comprise a discrete spectrum, a continuous spectrum, or both. The way in which
the postulates regarding the wave equation and its acceptable solutions lead to the
selection of definite energy values may be understood by the qualitative consider-
ation of a simple example. Let us consider, for a system of one degree of freedom,
that the potential-energy function satisfies V (x) → ∞ as |x| → ∞. For a given
value of the energy parameter E, the wave equation is
~2 d2 ψ(x)
(5.8) = (V (x) − E) ψ(x).
2m dx2
 15

For sufficiently large |x|, the quantity V (x) − E will be positive. Hence in this
2
region the curvature ∂∂ 2ψx will be positive if ψ is positive, and negative if ψ is neg-
ative. Suppose that at an arbitrary point x = c the function ψ has a certain value
and a certain slope ∂ψ∂x . The behaviour of the function, as it is continued both to
the right and to the left, is completely determined by the values assigned to two
quantities. We see that, for a given value of E only by a very careful selection of
the slope of the function at the point x = c can the function be made to behave
properly for large values of x. Similarly for large negative values of x. This selec-
tion is such as to cause the wave function to approach the value zero asymptotically
with increasing |x|. In view of the sensitiveness of the curve to the parameter E, an
infinitesimal change from this satisfactory value will cause the function to behave
improperly. We conclude that the parameter E and the slope at the point x = c
(for a given value of the function itself at this point) can have only certain values
if ψ is to be an acceptable wave function. For each satisfactory value of E there
is one (or, in certain cases, more than one) satisfactory value of the slope, by the
use of which the corresponding wave function can be built up. For this system the
characteristic values En of the energy form a discrete set, and only a discrete set,
inasmuch as for every value of E, no matter how large, V (x) − E is positive for
sufficiently large values of |x|.
It is customary to number the energy levels for such a system as E0 (the low-
est), E1 (the next), and so on, corresponding to wavefunctions ψ0 (x), ψ1 (x), etc.
The integer n, which is written as a subscript in En and ψn (x), is called quantum
number.
Let us now consider a system in which the potential-energy function remains
finite as x → ∞ or as x → −∞ or at both limits. For a value of E smaller than
both V (+∞) or V (−∞) the argument presented above is valid. Consequently the
energy levels will form a discrete set for this region. If E is greater than V (+∞),
however, a similar argument shows that the curvature will be such as always to re-
turn the wave function to the x axis, about which it will oscillate. Hence any value
of E greater than V (+∞) or V (−∞) will be an allowed value, corresponding to
an acceptable wave function, and the system will have a continuous spectrum of
energy values in this region.

5.4. Probability measures. Let Ψ(x, t) be a normalised solution of the Schrödinger

equation; that is
Z +∞
Ψ∗ (x, t)Ψ(x, t)dx = 1.
−∞

For any t, the quantity Ψ∗ (x, t)Ψ(x, t)dx = |Ψ(x, t)|2 dx is a probability measure.
It can be interpreted as the probability that the system represented by the state
Ψ(x, t) is in the region between x and x + dx at time t.
16

It is also convenient to normalise the individual amplitude function ψn (x),

Z +∞
ψn∗ (x)ψn (x)dx = 1.
−∞

Moreover, the independent solutions of the amplitude equation can always be cho-
sen in such a way that they form an orthonormal system; that is
Z +∞
∗
ψm (x)ψn (x)dx = δmn .
−∞
P
Using this relation, it is found that a wavefunction Ψ(x, t) = n an Ψn (x, t) is
normalised when the coefficients an ∈ C satisfy the condition
X
a∗n an = 1.
n

5.5. Stationary states. Let us consider the probability measure Ψ∗ Ψ for a

iEn t
system in the state Ψ(x, t) = n an ψn (x)e− ~ and its conjugate Ψ∗ (x, t) =
P
P ∗ ∗ iEn t
n an ψn (x)e
~ . Then,

X X i(Em −En )t
Ψ∗ (x, t)Ψ(x, t) = a∗n an ψn∗ (x)ψn (x)+ a∗m an ψm
∗
(x)ψn (x)e− ~

n m6=n

In general, then, the probability function and hence the properties of the system
depend on the time, inasmuch as the time enters in the exponential factors of the
double sum. Only if the coefficients are zero for all except one value of En is
Ψ∗ Ψ independent of t. In such a case the wavefunction will contain only a single
iEn t
term Ψn (x, t) = ψn (x)e− ~ ; For such a state the probability measure Ψ∗ Ψ is
independent of time, and the state is called stationary state.

5.6. Average values. If we inquire as to what average value would be ex-

pected on measurement at a given time t of the coordinate x of the system in a
physical situation represented by the wavefunction Ψ(x, t), the above interpreta-
tion of Ψ∗ Ψ leads to the answer
Z +∞
hxi = xΨ(x, t)∗ Ψ(x, t)dx.
−∞

A similar integral gives the average value predicted for x2 , x3 , or any function f (x)
of the coordinate x:
Z +∞
(5.9) hf (x)i = f (x)Ψ(x, t)∗ Ψ(x, t)dx.
−∞

In order that the same question can be answered for a more general dynamical
function f (p, x) involving the momentum p as well as the coordinate x, we assume
 17

that the average value of f (p, x) predicted for a system in the physical situation
represented by the wavefunction Ψ is given by the integral
Z +∞  
∗ ~ ∂
(5.10) hf (p, x)i = Ψ(x, t) f , x Ψ(x, t)dx.
−∞ i ∂x
~ ∂
in which the operator f obtained from f (p, x) by replacing p by i ∂x , acts on the
function Ψ(x, t).
R EMARK . In some cases further considerations are necessary in order to de-
∂


termine the exact form of the operator f ~i ∂x , x , but we shall not encounter such
difficulties.

In general, the result of a measurement of f will not be given by the average

value hf i; it is instead described by a probability distribution.
En
Even if the system is in a stationary state Ψn (x, t) = ψn (x)e−i ~ t , the vari-
ance Var(f ) = hf 2 i − hf i2 of an arbitrary observable f is, in general, not zero.
The energy of the system, corresponding to the Hamiltonian function H(p, x), has,
however, a definite value for a stationary state Ψn of the system, equal to En . To
prove this, we evaluate hHi and Var(H). The average value of H for the state Ψn
is given by
Z +∞
~2 ∂ 2
 
hHi = ψn (x)∗ − + V (x) ψn (x)dx = En .
−∞ 2m ∂x2
By a similar computation hH r i = Enr , r = 0, 1, 2, . . . . In particular, this shows
that Var(H) = 0; thus, the energy of the system in the state Ψn has a definite value
En .
R EMARK . Note that the probabilistic interpretation of Ψ(x, t) is unchanged if
we multiply the wavefunction by a complex number α of modulus |α| = 1. Thus,
a state is an equivalence class of the form eiθ Ψ(x, t), θ ∈ R. (This is especially
relevant when discussing symmetries of a quantum system. More on this later.)
E XERCISE 5.1 (Probability conservation law). Show that, if Ψ(x, t) is a solu-
tion of the Schrödinger equation
∂Ψ(x, t)
i~ = HΨ(x, t), x ∈ Rn ,
∂t
2
(H = − 2m
~
∆ + V (x)), then the probability density |Ψ(x, t)|2 satisfies the conti-
nuity equation (conservation of probability):
∂
|Ψ|2 + ∇
~ · J~ = 0,
∂t
where the probability current is
i~  ~ ∗ 
J~ = Ψ∇Ψ − Ψ∗ ∇Ψ ~ .
2m
18

E XERCISE 5.2 (Correspondence principle). Prove the Ehrenfest theorem: if

H(p, q) is the classical Hamiltonian of the systems, then the expectation values hqi
and hpi of the quantised variables satisfy the classical equation of motion
dhqi ∂H


 = +h i
dt ∂p
 dhpi = −h ∂H i.

dt ∂q
(This is an instance of the correspondence principle.)

6. Exactly solvable models

6.1. The quantum harmonic oscillator. You may be familiar with several
examples of harmonic oscillators form classical mechanics, such as particles on a
spring or the pendulum for small deviation from equilibrium, etc.
The classical Hamiltonian of a particle in a potential V (x) = 12 mω 2 x is
1 2 1
(6.1) H= p + mω 2 x.
2m 2
The corresponding quantum Hamiltonian is
~2 d2 1
(6.2) H=− 2
+ mω 2 x2 .
2m dx 2
Using dimensionless coordinates, the corresponding Schrödinger equation is
1 d2
 
1 2
(6.3) − + x ψ(x) = Eψ(x).
2 dx2 2
There is a very useful procedure to solve (6.3) based on the determination of
the form of ψ in the regions of large |x|, and the subsequent discussion, by the
introduction of a factor in the form of a power series (which later reduces to a
polynomial), of the value of ψ for generic x ∈ R. This procedure may be called
the polynomial method.
6.1.1. Eigenvalues and eigenfunctions of the harmonic oscillator. The first
steps is the asymptotics of the solution of (6.3) when |x|  1. For large |x|, E
is negligibly small relative to x2 /2, and the asymptotic form of the Schrödinger
equation becomes
d2 ψ
= x2 ψ.
dx2
2 2
This equation is asymptotically satisfied by e−x /2 and ex /2 . Of the two asymp-
totic solutions, the second tends rapidly to infinity with increasing values of |x|;
the first, however, leads to a satisfactory treatment of the problem.
We now proceed to obtain a solution of (6.3) throughout the whole configura-
tion space x ∈ R, based upon the asymptotic solution, by introducing as factor a
power series in x and determining its coefficients by substitution and recursion.
 19

2 /2
Let ψ(x) = H(x)e−x . Then
d2 ψ 2
= e−x /2 (x2 − 1)H(x) − 2xH 0 (x) + H 00 (x) .


dx 2

If ψ(x) solves the Schrödinger equation, then the get a differential equation for
H(x):

(6.4) H 00 (x) − 2xH 0 (x) + (2E − 1)H(x) = 0.

We now represent H(x) as a power series

(6.5) X X X
H(x) = ak xk , H 0 (x) = kak xk−1 H 00 (x) = k(k − 1)ak xk−2 .
k≥0 k≥0 k≥0

On substitution of these expression, Eq. (6.4) assumes the form

X
[(k + 2)(k + 1)ak+2 − 2kak + (2E − 1)ak ] xk = 0.
k≥0

In order for this series to vanish for all values of x (i.e., for H(x) to be a solution
of (6.4)), the coeffecients of the each power of x must vanish

(k + 2)(k + 1)ak+2 + (2E − 2k − 1)ak = 0

or
2(E − k − 1/2)
(6.6) ak+2 = − ak .
(k + 2)(k + 1)
This recursion formula enables to calculate successively the coefficients a2 , a3 , a4 , . . .
in terms of a0 and a1 , which are arbitrary. Note that if a0 = 0 (resp. a1 = 0) only
odd (resp. even) powers appear.
For arbitrary values of the parameter E, the above series consists of an infinite
number of terms and does not correspond to a satisfactory wavefunction. To see
this, note that for large enough values of k, we have
2
ak+2 ' ak .
k
2
But this is the same recursion of the series coefficients of ex ,
2 x4 x6 xk xk+2
e x = 1 + x2 + + + · · · + k
+ k+2
+ · · ·
2! 3! 2 ! 2 !
2
so that the higher terms of the series for H(x) differ from those of ex only by
2
a multiplicative constant. Therefore, for large values of |x|, H(x) ∼ ex , and the
x2 x2
product e− 2 H(x) will diverge like e 2 , thus making unacceptable as a wavefunc-
tion.
20

We must therefore restrict E to values which will cause the series for H(x) to
terminate, leaving a polynomial. The value of E which causes the series to break
off after the nth term is seen from (6.6) to be
1
(6.7) E =n+ .
2
It is, moreover, also necessary that the value of either a0 or of a1 be put equal to
zero, according as n is odd or even. The solutions are thus either odd or even func-
tions of x. The normalised solutions of the Schrödinger equation may be written
in the form
x2
(6.8) ψn (x) = Nn e− 2 Hn (x).

Hn (x) is the polynomial of degree nth determined by the recursion (6.6) with
E = n + 1/2, and Nn is a constant which is adjusted so that |ψn (x)|2 dx = 1
R

(this condition fixes the initial value a0 or a1 of the recursion). The polynomials
Hn (x), called Hermite polynomials, did not originate with Schrödinger’s work but
were well known to mathematicians in connection with other problems.
They can be defined using the formula
d n −x2
 
x2
(6.9) Hn (x) = e − e .
dx
Another definition makes use of the generating function
∞
X Hn (x) 2 −(y−x)2
(6.10) S(x, y) = y n = ex .
n!
n=0

E XERCISE 6.1. Show the equivalence of the two definitions (6.9)-(6.10) of

Hn (x).

It is easy to show that S(x, y) satisfies the following identities

∂S
(6.11) = −2(y − x)S
∂y
∂S
(6.12) = 2yS.
∂x
E XERCISE 6.2. Using (6.11) and (6.12) show that the Hermite polynomials
satisfy the three-term recursion

(6.13) Hn+1 (x) − 2xHn (x) + 2nHn−1 (x) = 0

(with initial conditions H0 (x) = 1, H1 (x) = 2x), and the differential equation

(6.14) Hn00 (x) − 2xHn0 (x) + 2nHn (x) = 0.

In particular, the latter is just (6.4) if we put E = n + 1/2.

 21

The functions
x2 1
(6.15) ψn (x) = Nn e− 2 Hn (x), with Nn = p √
n!2n π
are called Hermite functions. They satisfy the orthogonality relation (why?)
Z +∞
(6.16) ψn (x)ψm (x)dx = δnm .
−∞

E XAMPLE 6.1. By using the generating function S we can evaluate certain

integrals involving ψn which are of importance. For example, we may study the
integral which determines the probability of transition from the state n to the state
m. This is
Z +∞ Z +∞
2
hn |x| mi = ψn∗ (x)xψm (x)dx = Nn Nm xHn (x)Hm (x)e−x dx.
−∞ −∞

Using the generating functions S(x, y) and S(x, z) we obtain the relation
Z +∞ X 1 Z +∞
2
xS(x, y)S(x, z)dx = n m
y z xHn (x)Hm (x)e−x dx
−∞ n,m
n!m! −∞
Z +∞
2
= e2yz xe−(x−y−z) dx
−∞
2yz √
=e (y + z) π
√  22 y 3 z 2 2n y n+1 z n
= π y + 2y 2 z + + ··· + + ···
2! n!
22 y 2 z 3 2n y n z n+1 
+ z + 2yz 2 + + ··· + + ··· .
2! n!
n m
Hence, comparing the coefficients of y z , we see that hn |x| mi is zero except
for m = n ± 1, and
r r
n+1 n
hn |x| n + 1i = , hn |x| n − 1i = .
2 2
6.1.2. Creation and annihilation operators. The harmonic oscillator is of im-
portance for general theory, because it forms a cornerstone in the theory of radia-
tion. An alternative method of solution is based on the creation and annihilation
operators. They are defined as
1 1
(6.17) a∗ = √ (x − ip) = √ (x − ∂x )
2 2
1 1
(6.18) a = √ (x + ip) = √ (x + ∂x )
2 2
Note that a and a∗ do not commute
(6.19) [a, a∗ ] = 1.
22

Using the properties of the Hermite functions it is easy to verify that

√
(6.20) aψn (x) = nψn−1 (x)
√
(6.21) a∗ ψn (x) = n + 1ψn+1 (x)
We can also show that the number operator N = a∗ a satisfies
(6.22) N ψn (x) = nψn (x).
We have the following interpretations: a∗ creates a ‘quantum’ of energy, a annihi-
lates a ‘quantum’ of energy, N counts the number of quanta of energy. We write
√
(6.23) a∗ |ni = n + 1 |n + 1i
√
(6.24) a |ni = n |n − 1i
(6.25) N |ni = n |ni .

Note that H = a∗ a + 12 = N + 12 , and H |ni = (n + 1/2) |ni. The eigenstates |ni

are obtained by the repeated action of the creation operator on the ground state
1
(6.26) |ni = √ (a∗ )n |0i .
n!
E XERCISE 6.3. Prove by induction that
[a, (a∗ )n ] = n(a∗ )n−1 , for all n ∈ N.

6.2. A quantum particle in a box. The classical Hamiltonian (energy) of a

particle in a box [0, L] is
1 2
(6.27) H= p .
2m
The corresponding quantum Hamiltonian is formally
~2 d2
(6.28) H=− .
2m dx2
This corresponds to a force-free (V = 0) situation in the interior of the box. We
have to take into account the walls of the box, i.e. the boundary conditions (b.c.).
We may assume that the wavefunctions vanishes at the walls ψ(0) = 0 and ψ(L) =
0 (so-called Dirichlet b.c.). The eigenfunction of the Schrödinger operator H with
Dirichlet b.c. satisfy the problem
 ~2 d2

− ψ(x) = Eψ(x) for 0 < x < L
(6.29) 2m dx2
ψ(x) = 0 at x = 0 and x = L.


We see that a generic solution ψ is of the form

ψ(x) = Aeiαx + Be−iαx
 23

2mE
with α2 = ~2
and constants A and B. Imposing the boundary conditions
ψ(0) = 0 ⇒ A + B = 0
ψ(L) = 0 ⇒ AeiαL + Be−iαL = 0.
This imposes A = −B and αL = πk, k = 1, 2, . . . . The value of A is fixed by the
RL
normalisation condition 0 |ψ(x)|2 dx = 1. Therefore we find that the Schrödinger
equation in a box with Dirichlet b.c. has eigenfunctions
(6.30) r
π 2 ~2 2
 
2 πk
ψk = sin x with eigenvalues Ek = k , k = 1, 2, . . . .
L L 2mL2
Note that a particle in a box has a minimum energy
π 2 ~2
(6.31) E1 = > 0.
2L2 m
This is a manifestation of the uncertainty principle. If we know that a particle
is localised in [0, L], then the momentum px has uncertainty ∆px ≥ 2π~ L . This
amounts to not knowing whether the particles moves with a momentum |px | = π~ L
in the positive or negative sense. The energy corresponding to this momentum is
π 2 ~2
E = p2x /2m = 2L 2 m = E1 . This is a simple method to obtain a rough estimate of

the minimum energy of a system. (It can be used to estimate the energy of electrons
in an atom or of nucleons in a nucleus if the atomic or nuclear radius, i.e., the size
of the box, is known.)
E XAMPLE 6.2. Consider the Hamiltonian H of a free particle in a box of
length L with Dirichlet b.c.. Suppose that a particle (of mass m) is in the state
ψ(x), where
ψ(x) = Ax(L − x)
√
if 0 ≤ x ≤ L and zero otherwise. The normalisation constant is A = 30L5 .
Suppose that we measure H in the state ψ(x). The mean value of the energy is
A~2 L ∗
Z Z
∗ 5~
hHi = ψ (x)Hψ(x)dx = ψn (x)dx = .
m 0 mL2
To find the most probable outcome of the measurement we expand the wavefunc-
P
tion ψ(x) in the eigenfunctions of the Hamiltonian ψ(x) = k≥1 cn ψn (x), where
Z
cn = ψn∗ (x)ψ(x)dx.

π 2 ~2
A simple calculation reveals that the most probable value is E1 = 2L2 m
with
probability |c1 |2 = 960/π 6 ' 0.998.
E XAMPLE 6.3 (Quantum fractals). Let HD be the Hamiltonian of a free parti-
cle in the box [0, L] with Dirichlet b.c.. The simplicity of the spectrum and eigen-
functions is deceptive and could lead us to think that the dynamics, which is the
24

solution to the Schrödinger equation, is very simple. In fact, this belief is false,
as showed by M. V. Berry: the dynamics is instead very intricate.
p Consider the
evolution −i~∂t ψ = HD ψ with initial condition ψ(x, 0) = 1/L (uniform initial
state). The wavefunction ψ(x, t) at time t ∈ R is given by ψ(x, t) = θ(ξ, τ ) where
ξ = Lx , τ = mL
2π~t
2 , and

X 1 2πiξ(k+ 12 )−iπτ (k+ 21 )2

θ(ξ, τ ) = const · 1 e .
k∈Z
(k + 2 )
(This is related to a Jacobi theta function.) It is easy to check that the wavefunction
is quasiperiodic θ(ξ, τ + 1) = e−iπ/4 θ(ξ, τ ). Thus, at integer times τ the wave
function comes back (up to a phase) to its initial flat form: these are the quantum
revivals. More generally, at rational values of τ the graph of |θ(ξ, τ )|2 is piecewise
constant and there is a partial reconstruction of the initial wavefunction. On the
other hand, it can be proved that at irrational times, the wavefunction is a fractal in
space and time and form a beautifully intricate quantum carpet.
E XERCISE 6.4. Different boundary conditions correspond to different Schrödinger
operators, i.e., different physical systems. Show that the stationary states ψk (x)
and energy levels Ek of a particle in a box with Neumann b.c. (ψ 0 (0) = 0 and
ψ 0 (L) = 0) and are
r r  
1 2 πk
ψ0 (x) = , ψk (x) = cos x , k = 1, 2, . . .
L L L
π 2 ~2 2
with Ek = k , k = 0, 1, 2, . . . .
2L2 m
~ d 2 2
E XERCISE 6.5. Diagonalise the Hamiltonian − 2m dx2
, 0 ≤ x ≤ L with peri-
odic b.c. ψ(0) = ψ(L) (quantum particle in a ring).
R EMARK . The strictly positivity of the energy is a property of the Dirichlet
boundary conditions. Indeed, with the Neumann ones the energy of the ground
state is 0, and, more surprisingly, it is even negative with Robin’s boundary condi-
tions.

6.3. Free evolution of a quantum particle in Rn . The evolution of the state

of a particle of mass m > 0 in n dimensions in the absence of a potential (i.e., free
evolution, V (x) = 0) is governed by the Hamiltonian
~2
(6.32) H=− ∆,
2m
The operator H is essentially self-adjoint, and we will identify it with its unique
self-adjoint extension. If ψ0 is the state at time t = 0, the state ψ(x, t) at time t is
given by the solution of the free Schrödinger equation:
∂ψ ~2
(6.33) i~ + ∆ψ = 0, x ∈ Rn , t ∈ R
∂t 2m
 25

with initial condition ψ(x, 0) = ψ0 (x) (here we can assume ψ0 (x) ∈ L2 (Rn ) ∩
L1 (Rn )).
Let us first consider the momentum of a free particle. The wavefunction in
momentum representation is the Fourier transform of the state in position repre-
sentation
Z
1
(6.34) ψ(p, t) =
b ψ(x, t)e−ipx dx.
(2π)n/2 Rn
It is not hard to show, using the Schrödinger equation, that
(6.35) b t) = e−it~|p|2 /2m ψ
ψ(p, c0 (p).
b t)|2 = |ψ
Thus, |ψ(p, c0 (p)|2 . Hence the probability density of the momentum p is
not changing at all in time. This is what we expect, since the momentum of a free
particle is a constant of motion.
The free Schrödinger equation (6.33) has an explicit solution as inverse Fourier
transform of (6.34)
Z
1 t~
|p|2 +ipx
(6.36) ψ(x, t) = n ψ
c0 (p)e 2mi dp.
(2π) 2 Rn

The above formula can be interpreted as a superposition of plane waves with am-
plitude determined by the initial condition (in momentum space) ψ c0 (p). Using the
convolution theorem for the Fourier transform, we can write
(6.37)
2m n/2
  Z Z
im 2
|x−y|
ψ(x, t) = e 2~t ψ0 (y)dy = G(x, t; y, 0)ψ0 (y)dy
4πi~t Rn Rn
We see that the solutions is given by the convolution of the free particle propaga-
tor
 n/2
0 0 m 0 2 0
(6.38) G(x, t; x , t ) = 0
eim|x−x | /[2~(t−t )]
2πi~(t − t )
and the initial condition ψ0 . It is easy to show using this explicit representaion of
the solution that for any ψ0 ∈ L2 (Rn ), kψk∞ → 0 as t → ∞. So, the probability
density |ψ(x, t)|2 of the position flattens as t → ∞. In other words, there is
no limiting probability distribution for the position of the particle; it spreads out
across the whole space.

6.4. Scattering processes. Problems for which the energy eigenvalues are
continuously distributed usually arise in connection with the collision of a parti-
cle with a force field. The method of approach is different from that employed for
systems with discrete levels. There the boundary conditions were used to deter-
mine the discrete energy levels of the particle. In a scattering problem, the energy
is specified in advance, and the behaviour of the wavefunction at great distances
(the boundary values) is found in terms of it. This asymptotic behaviour can then
26

be related to the amount of scattering of the particle by the force field. We discuss
now a couple of exact solutions for simplified models that are routinely used for
approximate calculations on more complicated systems.
Consider the one-dimensional scattering of a particle with a potential V (x).
Suppose that V (x) is nondecreasing and that

lim V (x) = V0 > 0,

x→+∞
lim V (x) = 0.
x→−∞

In classical mechanics, a particle coming from the left with energy E < V0 can
not reach +∞: the potential is a barrier. In quantum mechanics there are new
phenomena:
(i) If E < V0 the particle can penetrate the barrier. In general, there is
nonzero probability that the particle is in the classical forbidden region
{x : E < V (x)};
(ii) If E > V0 there is the possibility that the particle is reflected by the
barrier.
These properties will be shown by calculating the probabilities of transmission and
reflection.
The wavefunction of a particle moving to the right, have the asymptotic form
p
x→+∞ ik2 x 2m(E − V0 )
(6.39) ψ(x) ∼ αe , with k2 =
~
(a particle with energy E, moving in a constant potential V0 , with positive momen-
tum p.) For large negative x, the particle is approximately free. The wave function
has the asymptotic form
√
x→−∞ ik1 x −ik1 x 2mE
(6.40) ψ(x) ∼ e + βe , with k1 =
~
(a superposition of an incident wave with intensity 1 and a reflected wave) The
probability current J~ ∝ (ψ∂x ψ ∗ − ψ ∗ ∂x ψ) (see Exercise 5.1) is
~ ∝ k1
incident wave: |J|
~ ∝ |β|2 k1
reflected wave: |J|
~ ∝ |α|2 k2 .
transmitted wave: |J|

The transmission and reflection coefficients are defined as

|J~t | k2 |J~r |
T = = |α|2 , R= = |β|2 .
J~i k1 J~i
Since the probability is conserved, we must have R + T = 1.
 27

E XAMPLE 6.4. Consider a particle moving along the x-axis with energy E ≥
V0 > 0 in a potential
(
V0 if x > 0,
V (x) =
0 if x < 0.
We compute the reflection coefficient. The (exact, not asymptotic!) solution of the
Schrödinger equation is
(
αeik2 x if x > 0,
ψ(x) = ik x −ik x
e 1 + βe 1 if x < 0.

If we impose the continuity of ψ and ψ 0 at x = 0, we determine

k1 − k2
β= .
k1 + k2
Therefore
 k1 − k2 2
 
2
R = |β| = 
  .
k1 + k2 
Since k1 6= k2 , in general there is reflection, even if the energy E > V0 is large
enough for a classical particle to overcome the barrier. If E = V0 (k2 = 0), there
is total reflection. There is total reflection even if E < V0 (k2 purely imaginary).
Finally, if E  V0 , R ∼ V0 /4E 2 → 0, there is total transmission.

E XAMPLE 6.5. Consider a potential barrier: the potential energy is positive in

the interval 0 < x < a and is zero outside this region
(
V0 if 0 < x < a,
V (x) =
0 otherwise.

We examine the case E > V0 first. In this case

ik1 x + Ae−ik1 x

e
 if x < 0,
ik x
ψ(x) = Be 2 + B e 2 0 −ik x if 0 < x < a,

 ik1 x
Ce if x > 0.

Imposing the condition that ψ and ψ 0 are continuous at x = 0 and x = a we obtain

a linear systems for the coefficients A, B, B 0 , C:

1 + A = B + B0
k1 (1 − A) = k2 (B − B 0 )
Ceik1 a = Beik2 a + B 0 e−ik2 a
Ck1 eik1 a = k2 (Beik2 a − B 0 e−ik2 a ).
28

If we set κ = k1 /k2 , then

2κ(1 + κ)
B=
(1 + κ)2− (1 − κ)2 e2ik2 a
2κ(1 − κ)e2ik2 a
B0 =
(1 + κ)2 − (1 − κ)2 e2ik2 a
4κei(k2 −k1 )a
C= .
(1 + κ)2 − (1 − κ)2 e2ik2 a
It follows that the transmission coefficient is
−1
4k12 k22 V 2 sin2 (k2 a)

2
T = |C| = 2 2 = 1+ 0 .
4k1 k2 + (k12 − k22 )2 sin2 (k2 a) 4E(E − V0 )
Note a new phenomenon: the barrier becomes transparent (T = 1) if k1 a = nπ.
p
If the case E < V0 we have k2 = i 2m|E − V0 |/~2 = ix2 . Using the
previous formulae we find the transmission coefficient
−1
4k12 x22 V02 sinh2 (x2 a)

T = 2 2 = 1+ .
4k1 x2 + (k12 + x22 )2 sinh2 (x2 a) 4E(V0 − E)
In classical mechanics the particle would not penetrate the barrier. In quantum me-
chanics, the solution of the Schrödinger equation shows that, contrary to classical
expectations, there is a finite probability that an incident particles will be transmit-
ted T > 0 . This effect goes under the name of tunnel effect.
When x2 a → ∞ it is easy to find the asymptotic formula
√
16E(V0 − E) − 8m(V0 −E)
a
T ∼ e ~ .
V02
E XERCISE 6.6. Find an expression for the transmission coefficient in the limit
V0 → +∞, a → 0, with V0 a = K constant. (This corresponds to a scattering
process by a δ-potential.)

6.5. The one-dimensional crystal. A crystal consisting of a vast number of

positive ions and electrons is a paradigmatic example of many-body quantum me-
chanical system. It is possible to derive some of its salient features from a simpli-
fied model of a quantum particles moving in a periodic potential
~2 d2
(6.41) H=− + V (x), with V (x + d) = V (x), x ∈ [0, N d].
2m dx2
In words, we model the crystal lattice of ions as a crystal ring with N sites with a
periodic potential with period d (the lattice spacing).
The fact that the potential V (x) is periodic of period d does not mean that
the eigenfunctions have to be periodic with the lattice period. The Schrödinger
 29

equation remains unchanged if we replace V (x) with V (x + d), but this leaves a
constant factor in its eigenfunctions undetermined, so that
ψ(x + d) = µ1 ψ(x)
where µ is a constant. If we move over one more lattice space, the same argument
holds true. Hence
ψ(x + 2d) = µ1 µ2 ψ(x).
If we continue N times, we get back to where we where, so that
ψ(x + N d) = µ1 µ2 · · · µN ψ(x) = ψ(x) ⇒ µ1 µ2 · · · µN = 1.
Because of the symmetry of the crystal ring, it cannot make any difference at which
lattice point we started, and we conclude that all the µk ’s must be equal. Hence
ψ(x + N d) = µN ψ(x), or µ = e2πin/N , n = 0, 1, . . . , N − 1.
We have, therefore,
ψ(x + md) = e2πinm/N ψ(x).
This statement can also be expressed in the following form
2πin x
(6.42) ψ(x) = e N d u(x) = eikx u(x)
where k = 2πn/N d and u(x) is periodic of period d. Eigenfunctions of this form
are called Bloch’s waves and it can be proved that they form a basis (this is known
as Bloch’s theorem or Floquet’s theorem). Although ψ(x) is not d-periodic, it is
easy to see that the probability density |ψ(x)|2 is periodic with the lattice period.
If we want to go on with the solution of the Schrödinger equation, we have to
specify V (x). A mathematically convenient choice is the Kronig-Penney (1931)
potential (a succession of square wells potential of width a and depth v). Consid-
ering a single period of the potential, the periodic solution u(x) is
(
Aei(α−k)x + Be−i(α−k)x , α2 = 2mE ~2
if 0 ≤ x < a
(6.43) u(x) = 2m(v−E)
Cei(β−k)x + De−i(β−k)x , β 2 = ~2
if a < x ≤ d

with the periodic condition u(0) = u(d), u0 (0) = u0 (d), and the continuity con-
ditions of u(x) and u0 (x) at x = a. These conditions provide four linear ho-
mogeneous equations to determine the four constants A, B, C and D. To have a
nontrivial solution the determinant of the matrix coefficient of linear system must
vanish. For E < v we get
β 2 − α2
sinh(β(d − a)) sin(αa) + cosh(β(d − a)) cos(αa) = cos(ka).
2αβ
The problem simplifies in the limit a → d, v → ∞ with v(d − a) = K fixed (the
potential becomes a periodic array of repulsive δ-functions). In this limit, it turns
30

out that the condition for the existence of nontrivial solutions of the Schrödinger
equation is
r
2mE Km
(6.44) P sin(αd) + cos(αd) = cos(kd), with α = 2
, and P = .
~ α~2
The left-hand side (a function of E) can assume values larger than one, the right-
hand side can not (and it does not depend on E), Therefore, the relation can be
2 (αd)2
satisfied only for those values of E = ~2md 2 for which the left-hand side remains

between −1 and +1. In other words, in a crystal only certain energy bands are al-
lowed. (A particle trapped in a potential well can have only discrete energy values.
In a periodic potential these discrete energies are replaced by energy bands.)
It is customary to present the results as a plot of the allowed energies E against
kd. Note that, the larger K, i.e., the higher the potential separating the regions of
zero potential, the narrower the energy bands. For K → ∞ the crystal reduces
to a series of independent quantum boxes, and the energy bands become discrete
2 π 2 n2
energy levels E = ~2md 2 . It is instructive to compare the results with that of a

free electron. We can obtain this result by letting v = 0 (i.e. K = 0, so that

E = ~2 k 2 /(2m).

E XERCISE 6.7. Derive the relation (6.44) by considering the one-dimensional

Schrödinger equation with potential
X
V (s) = K δ(x − nd).
n∈Z

6.6. The hydrogen atom. We consider the hydrogen atom as a system of an

electron in three dimension moving in the attractive Coulomb potential generated
by a fixed nucleus. Let us, for generality, ascribe to the nucleus charge +Z > 0
and unit charge −1 to p
the electron. The classical potential energy of the system is
Z
− |x| , in which |x| = x21 + x22 + x23 is the distance between the electron an the
nucleus. For convenience we can assume that the nucleus is fixed at the origin of
the coordinate system The classical Hamiltonian is

p2 Z
(6.45) H(p, x) = − , p = (p1 , p2 , p3 ) ∈ R3 , x = (x1 , x2 , x3 ) ∈ R3 .
2m |x|
The corresponding Schrödinger operator, in suitable units, is
Z
(6.46) H = −∆ − ,
|x|
and the time-independent Schrödinger equation is
Z
(6.47) −∆ψ(x) − ψ(x) = Eψ(x).
|x|
 31

The spherical symmetry of the problem suggests to use spherical coordinates


x1 = r sin θ cos ϕ

(6.48) x2 = r sin θ sin ϕ with r > 0, −π ≤ θ ≤ π, and 0 ≤ ϕ < 2π.

x3 = r cos ϕ


E XERCISE 6.8. Show that the Laplacian in spherical coordinates is

∂2
   
1 ∂ 2 ∂ 1 ∂ ∂ 1
(6.49) ∆= 2 r + 2 sin θ + 2 2 .
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ 2 ϕ
The Schrödinger equation in spherical coordinates for a generic symmetric
potential V (x) = V (|x|) is
∂2ψ
   
1 ∂ 2 ∂ψ 1 ∂ ∂ψ 1
r + sin θ + + (E − V (r)) ψ = 0.
r2 ∂r ∂r r2 sin θ ∂θ ∂θ r2 sin2 θ ∂ 2 ϕ
The above partial differential equation can be separated into ordinary equations
using the ansatz
ψ(r, θ, ϕ) = R(r)Θ(θ)Φ(ϕ).
On introducing this into the equation and dividing by RΘΦ we get
1 ∂2Φ
     
1 ∂ 2 ∂R 2 Z 1 ∂ ∂Θ
r +r E+ + sin θ + = 0.
R ∂r ∂r r Θ sin θ ∂θ ∂θ sin2 θΦ ∂ 2 ϕ
2
On multiplying through by sin2 θ, the remaining part of the third term, Φ1 ∂∂ 2Φ
ϕ
which
could only be a function of the independent variable ϕ, is seen to be equal to terms
independent of ϕ. Hence this term must be equal to a constant, which we call −m2 :
∂2Φ
+ m2 Φ = 0.
∂2ϕ
The equation in r and θ then can be written as
m2
     
1 ∂ 2 ∂R 2 Z 1 ∂ ∂Θ
r +r E+ + sin θ − = 0.
R ∂r ∂r r Θ sin θ ∂θ ∂θ sin2 θ
If we set the θ terms equal to the constant −l(l + 1), (and consequently the r terms
equal to l(l + 1)) we eventually obtain
 2
∂ Φ
− 2 = m2 Φ


∂ ϕ








m2
  
 1 ∂ ∂
− sin θ + Θ = l(l + 1)Θ

 sin θ ∂θ ∂θ sin θ




    
1 ∂ ∂R l(l + 1)


 − 2 r2 + + V (r) R(r) = ER(r).


r ∂r ∂r r2
32

These equations are now to be solved in order to determine the allowed values of
the energy. The sequence of solution is the following: We first find that the Φ-
equation possesses acceptable solutions only for certain values of the parameter
m. Introducing these in the Θ-equation, we find that it then possesses acceptable
solutions only for certain values of l(l + 1). (The form m2 and l(l + 1) chosen
for the constants is for convenience later on.) Finally, we introduce these values of
l in the R-equation and find that it possesses acceptable solutions only for certain
values of E. These are the energy levels for the stationary states of the system.

R EMARK . Note that the above considerations are valid for a generic radially
symmetric potential V (x) = V (|x|).

6.6.1. The Φ-equation. The solutions of the Φ-equation are

1
(6.50) Φm (ϕ) = √ eimϕ .
2π
In order for the function to be periodic, m must be an integer, i.e., m ∈ Z. The
constant m is called magnetic number.
√
The factor 1/ 2π is chosen in order to satisfy the normalisation condition.
Actually, they form an orthonormal system
Z 2π
(6.51) Φ∗m (ϕ)Φn (ϕ) = δnm .
0

It may pointed out that except for the lowest eigenvalue m2 = 0, all eigenvalues
are twice degenerate (the two functions Φm and Φ−m satisfy the same differential
equation if m > 0).
6.6.2. The Θ-equation. Now we turn to the equation for Θ(θ). In order to
solve the equation, it is convenient for us to introduce the new independent variable
t = cos θ ∈ [−1, 1], and consider the function P (t) = Θ(θ(t)). The equation
transforms to the somewhat simpler equation
d dP (t) m2
− (1 − t2 ) − P (t) = l(l + 1)P (t).
dt dt 1 − t2
For l = 0, 1, 2, . . . , the above equation is solved by the associated Legendre poly-
nomials
dm 1 dl 2
(6.52) Plm (t) = (−1)m (1 − t2 )m/2 (x − 1)l , l ∈ Z+ , |m| ≤ l.
dtm 2l l! dtl
l is called azimuthal quantum number.
6.6.3. The R-equation. The equation in r is
   
1 ∂ 2 ∂R l(l + 1)
r + − + E + V (r) R = 0.
r2 ∂r ∂r r2
 33

For generic V (r) the problem is intractable. The special case V (r) = −Z/r is
solvable. In this case the R-equation is
   
1 ∂ 2 ∂R l(l + 1) Z
r + − +E− R = 0.
r2 ∂r ∂r r2 r
Let us consider the case E < 0. Introducing the symbols
Z
α2 = −E, λ=,
2E
and the new independent variable ρ = 2αr > 0, the wave equation for S(ρ) =
R(ρ(r)) becomes
   
1 ∂ 2 ∂S 1 l(l + 1) λ
ρ + − − + S = 0.
ρ2 ∂ρ ∂ρ 4 ρ2 ρ
As in the treatment of the harmonic oscillator, we first discuss the asymptotic equa-
tion. For ρ large, the equation approaches the form
 
1 ∂ 2 ∂S 1
2
ρ = S
ρ ∂ρ ∂ρ 4
the acceptable solution of which is S = e−ρ/2 .
We now assume that the solution of the complete equation in the whole region
ρ > 0 has the form
S(ρ) = e−ρ/2 ρs L(ρ),
in which L(ρ) is a power series in ρ with a non-vanishing constant term
X
L(ρ) = ak ρk , a0 6= 0.
k
After some calculations we find the conditions on s,
s=l
and a differential equation for L(ρ):
ρL00 + (2(l + 1) − ρ) L0 + (λ − l − 1)L = 0.
We now introduce the series representation of L and obtain the recursion
(2(k + 1)(l + 1) + k(k + 1)) ak+1 + (λ − l − 1 − k) ak .
It can be shown by an argument similar to that used for the harmonic oscillator
that for any values of λ and l the series whose coefficients are determined by this
formula leads to a function S unacceptable as a wavefunction, unless it breaks off.
0
The condition that it breaks off after the term in ρn is
λ − l − 1 − n0 = 0
or
λ = n, where n = n0 + l + 1.
34

n0 is called radial quantum number, and can assume values n0 = 0, 1, 2, . . . . n

is called total quantum number.
6.6.4. Energy levels and eigenfunctions. Summarising, if we introduce the
quantum numbers n, l, and m as subscript (using n in preference of n0 ), the wave-
functions we have found as acceptable solutions may be written as

(6.53) ψnlm (r, θ, ϕ) = Rnl (r)Θlm (θ)Φm (ϕ),

with
1
(6.54) Φm (ϕ) = √ eimϕ ,
2π

s
(2l + 1)!(l − |m|)! |m|
(6.55) Θlm (θ) = Pl (cos θ),
2(l + |m|)!

s 3  l  
Z (n − l − 1)! − Zr Zr Zr
(6.56) Rnl (r) = − e 2n L2l+1
n+l .
(n + l)! 2n4 n n

The wavefunctions corresponding to distinct sets of values for n, l, and m are

independent and the energy value corresponding to ψnlm is
 2
Z
(6.57) En = − .
2n
2
q E1 = −Z /4 is simple and the corresponding
In particular, the lowest eigenvalue
Z 3 −Zr/2
eigenfunction ψnlm (r, θ, ϕ) = 2 e is positive.
The allowed values of these quantum numbers we have determined to be

m = 0, ±1, ±2, . . .
l = |m|, |m| + 1, |m| + 2, . . .
n = l + 1, l + 2, l + 3, . . .

This we may rewrite as

total quantum number n = 1, 2, 3, . . .

azimuthal quantum number l = 0, 1, 2, . . . , n − 1
magnetic quantum number m = −l, . . . , −1, 0, +1, . . . , +l

There are consequently 2l + 1 independent wave functions with given values of n

and l, and n2 independent wave functions with a given value of n, that is, with the
same energy value.
 35

7. Matrix Mechanics

Heisenberg formulated and successfully attacked the problem of calculating

values of the frequencies and intensities of the spectral lines which a system could
emit or absorb; that is, of the energy levels and the electric-moment integrals which
we have been discussing. He did not use wave functions and wave equations, but
instead developed a formal mathematical method for calculating values of these
quantities.
Heisenberg invented the new type of algebra as he needed it; it was immedi-
ately pointed out by Born and Jordan, however, that in his new quantum mechanics
Heisenberg was making use of quantities called matrices, and that his newly in-
vented operations were those of matrix algebra which had been already discussed
by mathematicians.

7.1. Matrices, their relation to wavefunctions, and the rules of matrix al-
gebra. Let us consider a set of orthogonal wavefunctions Ψ0 , Ψ1 , . . . Ψ
n , . . . and

a dynamical quantity f (qi , pi ). The corresponding operator is fop = f qi , ~i ∂q∂ i .
In the foregoing sections we have often made use of integrals such as
Z
(7.1) fmn = Ψ∗ fop Ψ;

for example, we have given fnn the physical interpretation of the average value of
f when the system is in the nth stationary state. Let us now arrange the numbers
fmn (the values of the integrals) in a square array F = (fmn ) ordered according to
m and n.
We can construct similar arrays G = (gmn ), H = (hmn ) etc. for other dy-
namical quantities. It is found that the symbols F , G, H etc. representing such
arrays can be manipulated by an algebra closely related to ordinary algebra, differ-
ing from it mainly in the process of multiplication. The rules of this algebra can be
easily derived from the properties of wave functions, which we already know.
Now let us derive some rules of the new algebra. For example, the sum of two
such arrays is an array each of whose elements is the sum of the corresponding
elements of the two arrays

(7.2) {f + g}mn = {g + f }mn = fmn + gmn .

The addition of arrays is commutative: F + G = G + F .

On the other hand, multiplication is not commutative: the product F G is not
necessarily equal to the product GF . Let us evaluate the mnth element of the array
F G. It is
Z
{f g}mn = Ψ∗m fop gop Ψn .
36

Now we can express the quantity gop Ψn in terms of the functions Ψk with constant
coefficients, obtaining X
gop Ψn = gkn Ψk .
k
That the coefficients are the quantities gkn is seen on multiplying by Ψ∗k and inte-
grating. Introducing this in the integral for {f g}mn we obtain
XZ
{f g}mn = Ψ∗m fop Ψk gkn ;
k

Ψ∗m fop Ψk
R
since = fmk , this becomes
X
(7.3) {f g}mn = fmk gkn .
k
We see that the arrays F , G, etc. can be manipulated according to the rules of
matrices.
The non-commutative nature of the multiplication of matrices is of great im-
portance in matrix mechanics.

7.2. Diagonal matrices and their physical interpretation. A diagonal ma-

trix is a matrix whose elements fmn are all zero except those with m = n. The
unit matrix, 1, is a special kind of diagonal matrix, all the diagonal elements being
equal to unity. A constant matrix is equal to a constant times the unit matrix. Appli-
cation of the rule for matrix multiplication shows that the square (or any power) of
a diagonal matrix is also a diagonal matrix, its diagonal elements being the squares
(or other powers) of the corresponding elements of the original matrix.
In discussing the physical interpretation of the wave equation, we saw that
the fundamental postulate regarding physical interpretation requires a dynamical
quantity f to have a definite value for a system in the state represented by the wave
function Ψn only when fnn r is qual to (f )r , for all values of r. We can now
nn
express this in terms of matrices: if the dynamical quantity f is represented by a
diagonal matrix F , then this dynamical quantity has the definite value fnn for the
state corresponding to the wavefunction Ψn of the set Ψ0 , Ψ1 ,,. . . .
i
E XAMPLE 7.1. The solutions Ψn = ψn (x)e− ~ En t of the Schrödinger equa-
tion correspond to a diagonal energy matrix H = diag(E0 , E1 , E2 , . . . ), so that,
the system in a physical condition (i.e. a state) represented by one of these wave-
functions has a fixed value of the total energy.
In the case of a system with one degree of freedom no other quantity (except
functions of H, such as H 2 ) is represented by a diagonal matrix; with more degrees
of freedom there are other diagonal matrices.
E XAMPLE 7.2. The surface-harmonic wavefunctions Θlm (θ)Φm (ϕ) for the
hydrogen atom and any other spherical symmetric system make the matrices for
 37

the square of the total angular momentum and the z component of the angular
momentum diagonal. These quantities thus have definite values for Θlm (θ)Φm (ϕ).

7.3. Matrix elements and their physical interpretation. The principle of

interference, which applies to all quantum systems (photons, electrons, atoms, . . . ),
suggests that transition probabilities Pnm from a state ψn to a state ψm are the
squares of transition amplitudes, and that the latter combine in a linear fashion.
Likewise, for the observable f we interpret the matrix element fnm as transition
amplitudes. We will not discuss their use in this module.

7.4. The Heisenberg Uncertainty Relations. Given a state |ψi, one can give
the probabilities for the possible outcomes of a measurement of an observable A.
The probability distribution has a mean or expectation value

hAi = hψ| A |ψi

and an uncertainty or deviation about this mean

q
∆A = hψ| (A − hAi)2 |ψi.

There are states for which ∆A = 0, and these are the eigenstates of A.
If we consider two Hermitian operators A and B, they will have some uncer-
tainties ∆A and ∆B in a given state. The Heisenberg uncertainty relations provide
a lower bound on the product of uncertainties ∆A∆B. Generally the lower bound
depends not only on the operators but also on the state. Of particular interest are
those cases in which the lower bounds is independent of the state.
Let A and B be two Hermitian operators, with commutator

[A, B] = iC.

It is easy to verify that C is also Hermitian. Fix a state |ψi. Note that the pair
A0 = A − hAi and B 0 = B − hBi has the same commutator as A and B (verify
this). Then, by Schwartz inequality:
2
(∆A)2 (∆B)2 = hψ| A02 |ψi hψ| B 02 |ψi = A0 ψ A0 ψ B 0 ψ B 0 ψ ≥  A0 ψ B 0 ψ  .

 
  


We can write the above inequality in terms of commutators and anticommutators

 2
2 2
 0 0
2  1 0 0 1 0 0 
(∆A) (∆B) ≥ hψ| A B |ψi = hψ| {A , B } + [A , B ] |ψi .
  
2 2
Since [A0 , B 0 ] = iC has pure imaginary expectation value, and {A0 , B 0 } has real
expectation value, we get
1  2 1
(7.4) (∆A)2 (∆B)2 ≥ hψ| {A0 , B 0 } |ψi + |hψ| C |ψi|2 .
4 4
38

This is the general uncertainty relation between any two Hermitian operators and
is evidently state dependent. Consider now canonically conjugate operators, for
which C = ~. In this case,
1 2 1 1
(∆A)2 (∆B)2 ≥ hψ| {A0 , B 0 } |ψi + ~2 ≥ ~2 ,
4 4 4
or
(7.5) ∆A∆B ≥ ~/2.

8. Mathematical Foundations of Quantum Mechanics

In classical probability theory one assumes that all the events concerning a
statistical experiment form a Boolean σ-algebra and defines a probability measure
as a completely additive non-negative function which assigns the value unity for the
identity element of the σ-algebra. Typically, the σ-algebra is the Borel σ-algebra
BX of a nice topological space X. Under very general conditions it turns out that
all probability measures on BX constitute a convex set whose extreme points are
degenerate probability measures.
We observe that BX admits a null element, namely ∅, a unit element, namely
X, a partial order ⊂ and operations union (∪), intersection (∩) and complemen-
tation (c ). Quantum theory can be thought of as a non-commutative version of
probability theory where the σ-algebra BX of events is replaced by a lattice of pro-
jections on a Hilbert space H. Most of the computations of quantum mechanics is
done in such a lattice.
Let H be a real or complex separable Hilbert Space and let P (H) denote the set
of all orthogonal projection operators on H, where 0 denotes the null operator and
I denotes the identity operator. If P1 , P2 ∈ P (H) we say that P1 ≤ P2 if the range
of P1 is contained in the range of P2 . Then ≤ makes P (H) a partially ordered
set. For any linear operator A on H let R(A) denote its range. For two orthogonal
projections P1 , P2 , let P1 ∨ P2 be the orthogonal projection on the closed linear
span of the subspaces R(P1 ) ∪ R(P2 ). Let P1 ∧ P2 be the orthogonal projection
on R(P1 ) ∩ R(P2 ). For any P ∈ L(H), I − P is the orthogonal projection on the
orthogonal complement R(P )⊥ of the range of P .
We may compare 0, I, ≤, ∨, ∧ and the map P 7→ I − P in L(H) with ∅,
X, ⊂, ∪, ∩ and complementation A 7→ X \ A of standart probability theory on
BX . The chief distinction lies in the fact that ∪ distributes with ∩ but ∨ need not
distribute with ∧.

8.1. The postulates of quantum theory. A consistent framework of quantum

theory can be introduced through a sequence of postulates.
P1 The state of a physical system is described by a normalised vector ψ in a
separable complex Hilbert space H.
 39

P2 To each (real-valued) observable corresponds a self-adjoint operator A on

H,or on a suitable subspace of H (in fact, the observables of a physical
system are elements of a non-commutative C ∗ -algebra).
P3 If the observable A is measured, then the possible outcomes of the mea-
sure are the eigenvalues of A.
P4 Suppose we measure A on the state ψ. Then, the probability that the
observed value is λ ∈ R is
X
|hxi , ψi|2 ,
i:λi =λ

where {xi } are the normalised eigenvectors of A and {λi } the corre-
sponding eigenvalues.
P5 The state of a (closed) system evolves in time as
ψ(t) = U (t)ψ0
for some strongly continuous one-parameter unitary group that only de-
pends on the system (and not on the state).

8.2. Hilbert spaces. Suppose H is a complex vector space. A map h·|·i : H ×

H → C is called a sesquilinear form if it is conjugate linear in the first argument
and linear in the second. A positive definite sesquilinear form is called scalar
product. Associated with every scalar product is a norm
p
(8.1) kψk = hψ|ψi.
The triangle inequality follows from the Cauchy-Schwarz inequality:
(8.2) |hψ|ϕi| ≤ kψkkϕk
with equality if and only if ψ and ϕ are parallel. If H is complete with respect to
the above norm, it is called a Hilbert space. It is no restriction to assume that H
is complete since one can easily replace it by its completion.
E XAMPLE 8.1. H = Cn with the usual scalar product
n
X
(8.3) hψ|ϕi = ψj∗ ϕj
j=1

is a (finite dimensional) Hilbert space. Similarly, the set of all square summable
sequences `2 (N) is a Hilbert space with scalar product
X
(8.4) hψ|ϕi = ψj∗ ϕj .
j∈N

E XAMPLE 8.2. The space H = L2 (M, dµ) is a Hilbert space with scalar
product given by
Z
(8.5) hψ|ϕi = ψ ∗ (x)ϕ(x)dµ(x).
M
40

(Note that the Example 8.1 is a special case of the this one; take M = R and µ a
sum of Dirac measures.)
R EMARK . Even though the elements of L2 (M, dµ) are, strictly speaking,
equivalence classes of functions, we will still call them functions for notational
convenience. However, note that for f ∈ L2 (M, dµ) the value f (x) is not well-
defined (unless there is a continuous representative and different continuous func-
tions are in different equivalence classes, e.g., in the case of Lebesgue measure).

A vector ψ ∈ H is called normalized or a unit vector if kψk = 1. Two vectors

ψ, ϕ ∈ H are called orthogonal (ψ ⊥ ϕ) if hψ|ϕi = 0.
Suppose ψ is a unit vector. Then the projection of ψ in the direction of ϕ is
given by and ψk defined via
(8.6) ψk = hϕ|ψi ϕ,
and ψ⊥ defined via
(8.7) ψ⊥ = ψ − hϕ|ψi ϕ,
is perpendicular to ϕ.
These results can also be generalised to more than one vector. A set of vectors
{ϕj } is called an orthonormal set if hϕi |ϕj i = δij .
E XERCISE 8.1. Show that every orthonormal set is linearly independent.
T HEOREM 8.1 (Pythagorean theorem). Suppose {ϕj }nj=0 is an orthonormal
set. Then every ψ ∈ H can be written as
n
X
(8.8) ψ = ψk + ψ⊥ , ψk = hϕj |ψi ϕj ,
j=0

where ψk and ψ⊥ are orthogonal. Moreover, hϕj |ψ⊥ i = 0 for all 1 ≤ j ≤ n. In

particular
n
X
(8.9) kψk2 = |hϕj |ψi|2 + kψ⊥ k2 .
j=0

Moreover, every ψ̂ in the span of {ϕj }nj=0 satisfies

(8.10) kψ − ψ̂k ≥ kψ⊥ k
with equality holding if and only if ψ̂ = ψk . In other words, ψk is uniquely deter-
mined as the vector in the span of {ϕj }nj=0 closest to ψ.

From (8.9) we obtain Bessel’s inequality

n
X
(8.11) |hϕj |ψi|2 ≤ kψk2
j=0
 41

with equality holding if and only if ψ lies in the span of {ϕj }nj=0 .
A scalar product can be recovered from its norm by virtue of the polarization
identity
1
kϕ + ψk2 − kϕ − ψk2 + ikϕ − iψk2 − ikϕ + iψk2 .


(8.12) hϕ|ψi =
4
A bijective linear operator U ∈ L(H1 , H2 ) is called unitary if U preserves
scalar products:
(8.13) hU ϕ|U ψi2 = hϕ|ψi1 , for all ψ, ϕ ∈ H1 .
By the polarization identity, this is the case if and only if U preserves norms:
kU ψk2 = kψk1 for all ψ ∈ H1 . The two Hilbert spaces H1 and H2 are called
unitarily equivalent in this case.

E XERCISE 8.2. The shift operator

S : `2 (N) → `2 (N), (a1 , a2 , a3 , . . . ) 7→ (0, a1 , a2 , . . . )
satisfies kSak = kak. Is it unitary?

By continuity of the scalar product we see that Theorem 8.1 can be gener-
alised to arbitrary orthonormal sets {ϕj }j∈J . Note that from Bessel’s inequality, it
follows that the map ψ 7→ ψk is continuous.
An orthonormal set which is not a proper subset of any other orthonormal set
is called an orthonormal basis due to the following result.

T HEOREM 8.2. For an orthonormal set {ϕj }j∈J , the following conditions are
equivalent:
(i) {ϕj }j∈J is a maximal orthonormal set;
(ii) For every ψ ∈ H we have
X
(8.14) ψ= hϕj |ψi ϕj ;
j∈J

(iii) For every ψ ∈ H we have Parseval’s relation

X
(8.15) kψk2 = | hϕj |ψi |2 ;
j∈J

(iv) hϕj |ψi = 0 for all j ∈ J implies ψ = 0.

E XAMPLE 8.3. The set of functions

1
(8.16) ϕm (x) = √ eimx , m ∈ Z,
2π
forms an orthonormal basis for H = L2 (0, 2π). The corresponding orthogonal
expansion is just the ordinary Fourier series.
42

A Hilbert space is separable if and only if there is a countable orthonormal

basis. In fact, if H is separable, then there exists a countable total set {ψj }Nj=0 .
Here N ∈ N if H is finite dimensional and N = ∞ otherwise. After throwing
away some vectors, we can assume that ψn+1 cannot be expressed as a linear com-
bination of the vectors ψ0 , . . . , ψn . Now we can construct an orthonormal basis as
follows: we begin by normalising ψ0 ,
ψ0
(8.17) ϕ0 = .
kψ0 k
Next we take ψ1 and remove the component parallel to ϕ0 and normalise again:
ψ1 − hϕ0 |ψ1 i ϕ0
(8.18) ϕ1 = .
kψ1 − hϕ0 |ψ1 i ϕ0 k
Proceeding like this, we define recursively

ψn − n−1
P
j=0 hϕj |ψ1 i ϕj
(8.19) ϕn = Pn−1 .
ψn − j=0 hϕj |ψ1 i ϕj

This procedure is known as Gram-Schmidt orthogonalization. The result of the

procedure is the orthonormal basis {ϕ}N
j=0 .

E XERCISE 8.3. In L2 (−1, 1), we can orthogonalise the monomials fn (x) =

xn . The resulting polynomials
p are up to a normalisation equal to the Legendre
polynomials ϕn (x) = n + 1/2Pn (x),
1
3x2 − 1 , . . .


(8.20) P (x) = 1, P (x) = x, P (x) =
2
T HEOREM 8.3. If H is separable, then every orthonormal basis is countable.

In quantum theory it is assumed that the Hilbert space associate to a quantum

system is separable.
In particular, it can be shown that L2 (M, dµ) is separable. Moreover, it turns
out that, up to unitary equivalence, there is only one separable infinite dimensional
Hilbert space: Let H be an infinite dimensional Hilbert space and let {ϕj }j∈N be
any orthogonal basis. Then the map U : H → `2 (N), ψ 7→ (hϕj |ψi)j∈N is unitary.
In particular,

T HEOREM 8.4. Any separable infinite dimensional Hilbert space is unitarily

equivalent to `2 (N).

E XERCISE 8.4. Let {ϕj } be some orthonormal basis. Show that a bounded
linear operator A is uniquely determined by its matrix elements Ajk = hϕj |Aϕk i
with respect to this basis.
 43

8.3. The projection theorem and the Riesz lemma. Let M ⊂ H be a subset.
Then M ⊥ = {ψ ∈ H : hϕ|ψi = 0, ∀ϕ ∈ M } is called the orthogonal comple-
ment of M . By continuity of the scalar product it follows that M ⊥ is a closed
⊥
linear subspace and by linearity that (span(M )) = M ⊥ . For example, we have
H⊥ = {0} since every vector in H⊥ must be in particular orthogonal to all vectors
in some orthonormal basis.

R EMARK . Note that if M ⊂ H is closed, it is a Hilbert space and has an

orthonormal basis {ϕj }j∈J .

T HEOREM 8.5 (Projection theorem). Let M be a closed linear subspace of a

Hilbert space H. Then every ψ ∈ H can be uniquely written as ψ = ψk + ψ⊥ with
ψk ∈ M and ψ⊥ ∈ M ⊥ . One writes

(8.21) M ⊕ M⊥ = H

in this situation.

Theorem 8.5 implies that to every ψ ∈ H we can assign a unique vector ψk ∈

M closest to ψ. The rest, ψ−ψk , lies in M ⊥ . The operator PM defined as P ψ = ψk
is called the orthogonal projection corresponding to M . Note that
2
(8.22) PM = PM , and hPM ψ|ϕi = hPM ϕ|ψi .

Clearly, PM ⊥ ψ = ψ − PM ψ = (I − PM )ψ.
Finally we turn to linear functionals, that is, to linear operators ` : H → C. By
the Cauchy-Schwarz inequality we know that `ϕ : ψ 7→ hϕ|ψi is a bounded linear
functional (with norm kϕk). It turns out that, in a Hilbert space, every bounded
linear functional can be written in this way.

T HEOREM 8.6 (Riesz lemma). Suppose ` is a bounded linear functional on a

Hilbert space H. Then there is a unique vector ϕ ∈ H such that `(ψ) = hϕ|ψi
for all ψ ∈ H. In other words, a Hilbert space is equivalent to its own dual space
H∗ ' H via the map ϕ 7→ `ϕ = hϕ|·i which is a conjugate linear isometric
bijection between H and H∗ .

P ROOF. If ` = 0 we can choose ϕ = 0. Otherwise Ker(`) = {ψ ∈ H : `(ψ) =

0} is a proper subspace of H and we can find a unit vector ϕ̃ ∈ Ker(`)⊥ . For every
ψ ∈ H, `(ϕ̃)ψ − `(ψ)ϕ̃ ∈ Ker(`) and hence

0 = hϕ̃|`(ϕ̃)ψ − `(ψ)ϕ̃i = `(ϕ̃) hϕ̃|ψi − `(ψ).

Therefore we can choose ϕ = `(ϕ̃)−1 ϕ̃. To see uniqueness, let ϕ1 , ϕ2 be two such
vectors. Then hϕ1 − ϕ2 |ψi = hϕ1 |ψi − hϕ2 |ψi = `(ψ) − `(ψ) = 0 for all ψ ∈ H.
Therefore ϕ1 − ϕ2 ∈ H⊥ = {0}. 
44

In view of the isomorphism H∗ ' H, onel often uses Dirac’s notation and de-
notes vectors ψ ∈ H by the ket symbol |ψi and linear functionals `ϕ = hϕ|·i ∈ H∗
using the bra symbol hϕ|. The contraction `ϕ (ψ) of a covector ϕ and a vector ψ is
denoted by the bracket (scalar product) hϕ|ψi. One writes the above isomorphism
as
H 3 |ψi 7→ hψ| ∈ H∗ .
The ket-bra |ψihϕ| denotes the rank-one linear operator |vi 7→ `ϕ (v) |ψi = hϕ|vi |ψi.
E XERCISE 8.5. Suppose U : H → H is unitary and M ⊂ H. Show that
U M ⊥ = (U M )⊥ .

8.4. The C ∗ algebra of bounded linear operators. Observables like mo-

mentum and spin component are to be represented by operators acting on states
(vectors). Almost all operators that appear in quantum theory are linear. A linear
operator A : H → H is called bounded if the operator norm
(8.23) kAk = sup kAψk
kψk=1

is finite. We write A ∈ L(H).

E XAMPLE 8.4. An orthogonal projection PM 6= 0 has norm one.
L EMMA 8.7.
(8.24) kAk = sup |hϕ|Aψi| .
kψk=kϕk=1

We start by introducing a conjugation for operators on a Hilbert space H. Let

A ∈ L(H). Then the adjoint operator A∗ is defined via
(8.25) hϕ|A∗ ψi = hAϕ|ψi .
E XAMPLE 8.5. If H = Cn , then a linear operator is represented by a matrix
A = (aij )ni,j=1 . Note that
sX
kAk = sup a∗ik aij ψk∗ ψj .
kψk=1 ijk

Then
 ∗
X X X X X
hAϕ|ψi =  aij ϕj  ψi = a∗ij ϕ∗j ψi = ϕ∗i a∗ji ψj
i j ij i j

Hence, A∗ = (a∗ji )ni,j=1 .

L EMMA 8.8. Let A, B ∈ L(H) and α ∈ C. Then
(i) (A + B)∗ = A∗ + B ∗ , (αA)∗ = α∗ A∗ ;
(ii) A∗∗ = A;
 45

(iii) (AB)∗ = B ∗ A∗ ;
(iv) kA∗ k = kAk and kAk2 = kA∗ Ak = kAA∗ k.
R EMARK . As a consequence of kA∗ k = kAk, observe that taking the adjoint
is continuous.

The algebra L(H) of bounded operators with the involution ∗ forms a C ∗ -

algebra.
E XAMPLE 8.6. The continuous functions C(I) on a bounded interval I ⊂ R
together with complex conjugation form a commutative C ∗ -algebra.

An operator A ∈ L(H) is called normal if AA∗ = A∗ A, self-adjoint if A =

A∗ , unitary if A∗ A = AA∗ = I, an (orthogonal) projection if A = A∗ = A2 , and
positive if A = BB ∗ for some B ∈ L(H). Clearly both self-adjoint and unitary
elements are normal.
E XERCISE 8.6. Consider the Pauli matrices σ1 , σ2 , σ3 ∈ L(C2 ):
     
0 1 0 −i 1 0
(8.26) σ1 = , σ2 = , σ3 = .
1 0 i 0 0 −1
They are Hermitian σi∗ = σi and also unitary σi∗ σi = σi σi∗ = σi2 = 1.
E XERCISE 8.7. Compute the adjoint of the shift operator
S : `2 (N) → `2 (N), (a1 , a2 , a3 , . . . ) 7→ (0, a1 , a2 , . . . ).

8.5. Resolvents and spectra. Let A be a (densely defined) closed operator.

The resolvent set of A is the subset of C defined by
(8.27) ρ(A) = {z ∈ C : (A − z)−1 ∈ L(H)}.
The spectrum of A is the complement of the resolvent set of A:
(8.28) σ(A) = C \ ρ(A).
In particular, λ ∈ σ(A), if and only if (A − λ)ψ = 0 for some nonzero vector ψ.
In this case ψ is called an eigenvector of A corresponding to the eigenvalue λ.
We can characterise the spectra of self-adjoint operators.
P ROPOSITION 8.9. Let A be self-adjoint. Then all eigenvalues are real and
eigenvectors corresponding to distinct eigenvalues are orthogonal.

P ROOF. If Aψj = λj ψj , j = 1, 2,, we have

λ1 kψ1 k2 = hψ1 |λ1 ψ1 i = hψ1 |Aψ1 i = hAψ1 |λ1 ψ1 i = hλψ1 |ψ1 i = λ∗1 kψ1 k2
and
(λ1 − λ2 )hψ1 |ψ2 i = hAψ1 |ψ2 i − hAψ1 |ψ2 i = 0.

46

The result does not imply that two linearly independent eigenfunctions to the
same eigenvalue λ are orthogonal. However, it is no restriction to assume that they
are since we can use Gram-Schmidt to find an orthonormal basis for Ker(A−λ). If
H is finite dimensional, we can always find an orthonormal basis of eigenvectors.
(In the infinite dimensional case this is no longer true in general.)

T HEOREM 8.10 (Spectral theorem for self-adjoint bounded operators). Let A

be bounded and self-adjoint. Then,
X
A= λk |ψk ihψk |
k

where λk ∈ R and |ψk i ∈ H are the eigenvalues and corresponding normalised

eigenvectors of A. Moreover, {|ψk i} is an orthonormal basis of H. The above
spectral decomposition of A provides a unique norm continuous, ∗ -homomorphism
from the set of all bounded measurable functions B(R) to L(H). For every func-
tion f ∈ B(R), we have
X
f (A) = f (λk ) |ψk ihψk | .
k

R EMARK . We can also characterise the spectra of unitary operators. If U is

a unitary operator, then all eigenvalues have modulus one and eigenvectors corre-
sponding to distinct eigenvalues are orthogonal.

8.6. Time evolution. Now let us turn to the time evolution of such a quantum
mechanical system. Given an initial state ψ(0) of the system, there should be a
unique ψ(t) representing the state of the system at time t ∈ R. We will write
(8.29) ψ(t) = U (t)ψ(0).
Moreover, it follows from physical experiments that superposition of states holds;
that is, U (t)(α1 ψ1 (0) + α2 ψ2 (0)) = α1 ψ1 (t) + α2 ψ2 (t). In other words, U (t)
must be a linear operator. Moreover, since ψ(t) is a normalised state kψ(t)k = 1,
we have
(8.30) kU (t)ψk = kψk.
Therefore U (t) is a unitary operator. Next, since we have assumed uniqueness of
solutions to the initial value problem, we must have
(8.31) U (0) = 1, U (t + s) = U (t)U (s).
A family of unitary operators U (t) having this property is called a one-parameter
unitary group. In addition, it is natural to assume that this group is strongly
continuous; that is,
(8.32) lim U (t)ψ = ψ, for all ψ ∈ H.
t→0
 47

Each such group has an infinitesimal generator, defined by

i
(8.33) Hψ = lim (U (t)ψ − ψ) , for all ψ ∈ D(H)
t→0 t

where
 
i
(8.34) D(H) = ψ ∈ H : lim (U (t)ψ − ψ) exists .
t→0 t

This operator is called the Hamiltonian and corresponds to the energy of the sys-
tem. If ψ(0) ∈ D(H), then ψ(t) is a solution of the Schrödinger equation (in
suitable units)

d
(8.35) i ψ(t) = Hψ(t).
dt
E XAMPLE 8.7. Suppose that H = C and consider the scalar equation

d
(8.36) i ψ(t) = Hψ(t), ψ(0) = ψ0 ∈ C
dt

where H ∈ R. It is immediate to integrate the equation and find ψ(t) = e−iHt ψ0 .

Recall that the exponential eA of a matrix A is defined by the series

X Ak
eA = .
k!
k≥0

E XAMPLE 8.8. Suppose that H = Cn and consider the vector equation

d
(8.37) i ψ(t) = Hψ(t), ψ(0) = ψ0 ∈ Cn
dt
where H ∈ Cn×n is Hermitian H ∗ = H. Using the definition of exponential of a
matrix, it is easy to verify that the solution of the problem is ψ(t) = e−iHt ψ0 .

There is a one-to-one correspondence between one-parameter strongly contin-

uous unitary groups of operators on H and self-adjoint operators on H.

T HEOREM 8.11 (Stone’s theorem). Let H be self-adjoint and let U (t) =

exp(−iAt).
(i) U (t) is a strongly continuous one-parameter unitary group.
(ii) The limit limt→0 ti (U (t)ψ − ψ) exists if an only ifψ is in the domain of
H, ψ ∈ D(H), in which case limt→0 ti (U (t)ψ − ψ) = Hψ.
(iii) U (t)D(H) = D(H) and HU (t) = U (t)H.
48

8.7. Orthogonal sums and tensor products. Given two Hilbert spaces H1
and H2 , we define their orthogonal sum H1 ⊕H2 to be the set of all pairs (ψ1 , ψ2 ) ∈
H1 × H2 together with the scalar product
(8.38) h(ϕ1 , ϕ2 )|(ψ1 , ψ2 )i = hϕ1 |ψ1 i1 + hϕ2 |ψ2 i2 .
It is left as an exercise to verify that H1 ⊕H2 is again a Hilbert space. Moreover, H1
can be identified with {(ψ1 , 0) : ψ1 ∈ H1 }, and we can regard H1 as a subspace
of H1 ⊕ H2 , and similarly for H2 . It is customary to write ψ1 + ψ2 instead of
(ψ1 , ψ2 ).
More generally, for a countable collection (Hj )∞ j=1 of Hilbert spaces, the set
 
M∞ X ∞ ∞
X 
(8.39) Hj = ψj : ψj ∈ Hj , kψj k2j < ∞ ,
 
j=1 j=1 j=1

with scalar product

*∞ 
∞ + ∞
X X X
(8.40) ϕj  ψj = hϕj |ψj i
j=1  j=1 j=1

is a Hilbert space.
L∞
E XAMPLE 8.9. j=1 C = `2 (N).

Similarly, if H and K are two Hilbert spaces, we define their tensor product
as follows: we start with the set of all finite linear combinations of elements of
H × K:
 
X n 
(8.41) F (H, K) = αj (ψj , ϕj ) : (ψ, ϕ) ∈ H × K, αj ∈ C .
 
j=1

Since we want (ψ1 +ψ2 )⊗ϕ = ψ1 ⊗ϕ+ψ2 ⊗ϕ, ψ ⊗(ϕ1 +ϕ2 ) = ψ ⊗ϕ1 +ψ ⊗ϕ2 ,
and (αψ) ⊗ ϕ = ψ ⊗ (αϕ), we consider the quotient F (H, K)/N (H, K), where
  
X n Xn n
X 
(8.42) N (H, K) = span αj βk (ψj , ϕk ) −  αj ψj , βk ϕk 
 
j,k=1 j=1 k=1

and write ψ ⊗ ϕ for the equivalence class of (ψ, ϕ).

Next, define
(8.43) hψ1 ⊗ ϕ1 |ψ2 ⊗ ϕ2 i = hψ1 |ψ2 i hϕ1 |ϕ2 i
which extends to a sesquilinear form on F (H, K)/N (H, K). It is easy to show that
this sesquilinear form is a scalar product. The completion of F (H, K)/N (H, K)
with respect to the induced norm is called tensor product H ⊗ K of H and K.
L EMMA 8.12. If {ψj } and {ϕk } are orthonormal bases for H, K, respectively,
then {ψj ⊗ ϕk } is an orthonormal basis for H ⊗ K.
 49

E XAMPLE 8.10. H ⊗ Cn = Hn . In particular, Cn ⊗ Cm = Cnm .

E XAMPLE 8.11. L2 (M, dµ) ⊗ L2 (N, dν) = L2 (M × N, dµ × dν).

9. Angular momentum and spin

In classical mechanics, the components of the angular momentum of a point

particle are related to the coordinates x, y and z and momenta px , py and pz by
Lx = ypz − zpy ,
Ly = zpx − xpz ,
Lz = xpy − ypx .
To find quantum mechanical operators for the orbital angular momentum, use is
made of the assumption that the correspondence principle must be satisfied. Thus
any relation which appears in classical mechanics must be valid after quantisa-
tion of the canonical variables. As an example, the resulting expression for the
z-component of orbital angular momentum is
Lz = x (−i~∂y ) − y (−i~∂x ) = −i~ (x∂y − y∂x ) .

E XERCISE 9.1. Verify the commutation relations

[Lz , x] = i~y, [Lz , y] = −i~x, [Lz , z] = 0,
[Lz , px ] = i~py , [Lz , py ] = −i~px , [Lz , pz ] = 0.

The commutation relations among the various components of the angular mo-
mentum can be obtained:
[Lx , Ly ] = i~Lz , [Ly , Lz ] = i~Lx , [Lz , Lx ] = i~Ly , etc.
Note that the operators for the three components of angular momentum do not com-
mute with one another (they are not simultaneously measurable). The commutation
relations can be written in a compact notation using the Levi-Civita antisymmetric
tensor
[Lk , Lk ] = i~jkl Ll , j, k, l = 1, 2, 3.
Another physical quantity of considerable interest is the square of the magnitude
of the angular momentum. The corresponding operator is defined through
L2 = L2x + L2y + L2z .
It is easy to check that L2 commutes with all three of the components:
[L2 , Lx ] = [L2 , Ly ] = [L2 , Lz ] = 0.
Since the z-component and the square of the angular momentum commute with
each other, it is possible to choose diagonalise simultaneously both operators. We
50

then have
L2 ψ = aψ
Lz ψ = bψ.
Of course hL2 i ≥ hL2z i, so that a ≥ b2 .
It is useful at this point to define two operators which play a role similar to that
of the ladder operators of the simple harmonic oscillator. They are
L± = Lx ± iLy .
It can be shown that
[Lz , L± ] = ±~L± .
This equation says that L+ and L− play the role of ladder operators with the regard
to the eigenvalues of Lz . In fact,
Lz (L+ ψ) = (b + ~)(L+ ψ).
Since L2 commutes with all three components of the angular momentum, we get
L2 (L+ ψ) = a(L+ ψ).
Thus the operator L+ operating on an eigenfunction of Lz and L2 generates a new
simultaneous eigenfunction of these two operators for which the eigenvalue of L2 is
left unchanged but for which the eigenvalue of Lz is increased by ~. The eigenvalue
b has an upper bound; otherwise the inequality a ≥ b2 would be violated. If we
assume that b is the largest eigenvalue satisfying the inequality, then it must be
L+ ψ = 0. If we multiply both sides by L− we get
L− L+ ψ = (L2 − L2z − ~Lz )ψ = 0,
so that a = b(b + ~). In a similar manner it is easy to show that Ln− ψ is an
eigenfunction of Lz :
Lz (Ln− ψ) = (b − n~)Ln− ψ.
Take n to be the largest integer for which the inequality a ≥ (b − n~)2 is satisfied.
In this case L− Ln− ψ = 0. If we multiply by L+ we get,
L+ L− Ln− ψ = (L2 − L2z + ~Lz )Ln− ψ.
From this, a = (b − n~)2 − (b − n~)~. Combining with the previous relation
between a and b we finally get
1
(9.1) b = (n − 1)~ = l~.
2
From this, l is nonnegative and either integer or half-integer (depending on whether
n is even or odd). We will show later that for the orbital angular moments, l takes
only integral values
l = 0, 1, 2, 3, . . . .
 51

The corresponding value of a is

a = l(l + 1)~2 .

We can summarise as

L2 Ylm = l(l + 1)~2 Ylm ,

Lz Ylm = m~Ylm .

l can take on nonnegative integral values, and m can take positive or negative
integral values such that l ≥ |m|. Note that these properties result directly from
the commutation relations of the angular momentum, and hence follow from only
the algebraic properties of the operators.
Note that only one component of the angular momentum may be precisely
specified at a time. Although a simultaneous knowledge of the two other compo-
nents is impossible, it is possible to say something about their expectation values.
For example, for a particle in the angular-momentum state Ylm (eigenfunction of
L2 and Lz ),
hLx i = hLy i = 0.
Also,
1 1
hL2x i = hL2y i = hL2 − L2z i = (l(l + 1) − m2 )~2 .
2 2
Note that when the angular momentum “parallel” to the z-axis (m = l), the x- and
y-components are still not zero.
It is helpful to visualise the results of this section of the aid of a geometrical
model.
p Consider the length of the angular momentum vector (Lx , Ly , Lz ) to be
l(l + 1)~. The 2l + 1 allowed projections of this on the the z-axis are given by
m~, with m = 0, ±1, ±2, . . . , ±l. Note that the projection on the z-axis never
exceeds the length of the vector. The angular momentum thus may be visualised as
lying on the surface of a cone having the z-axis for its axis and an altitude of m~.
All positions in the surface are equally likely.

9.1. Orbital angular momentum. Consider now the orbital angular momen-
tum wavefunctions which are simultaneously eigenfunctions of L2 and Lz . It is
helpful to introduce spherical coordinates in the usual fashion x = r sin θ cos φ,
y = r sin θ sin φ, z = r cos θ. The operator Lz in spherical coordinates takes the
form
∂
Lz = −i~ .
∂φ
We see that the simultaneous eigenfunctions of Lz and L2 are

(9.2) Ylm = exp(imφ)Θ(θ),

52

where ml must take integer values if the resulting function is to be single-valued.

In a similar manner, the operator L2 takes the form
1 ∂2
   
2 2 1 ∂ ∂
L = −~ sin θ + .
sin θ ∂θ ∂θ sin2 θ ∂θ2
It can be seen that the operator for L2 is essentially the angular part of the Laplacian
operator:
2
~2 ~2 1 ∂ L2

− ∆=− r + .
2m 2m r ∂r 2mr2
The wavefunctions Ylm (θ, φ) are called spherical harmonics. They form an or-
thonormal set of functions on the sphere in the sense that
Z
(9.3) hYlm |Yl0 m0 i = Ylm (θ, φ)∗ Yl0 m0 (θ, φ)dφ sin θdθ = δll0 δmm0 .

Consequently, one can expand any wavefunction ψ(r, θ, φ) as

X
ψ(r, θ, φ) = alm (r)Ylm (θ, φ).
l,m

The matrix elements for the z-component of the angular momentum in this repre-
sentation are (Ylm = |l, mi)

(Lz )lm,l0 m0 = hl, m| Lz l0 , m0 = m~δll0 δmm0 .

 
(9.4)

In a similar manner, the matrix elements of the square of the angular momentum
are

(L2 )lm,l0 m0 = hl, m| L2 l0 , m0 = l(l + 1)~2 δll0 δmm0 .

 
(9.5)

The (infinite) matrices representing Lz and L2 have thus been evaluated in the
representation in which they are diagonal

 0

−1
 0 
 +1

 
 −2 
 −1 
 0

 
 +1 
+2
Lz = ~  ,
 
 −3 
 −2 
 −1 
0
 
 
 +1 
 +2 
 +3 
..
.
 53

 0

2
 2 
 2

 
 6 
 6 
 6

 
 6 
L2 = ~2  6
.
 
 12 
 12 
 12 
12
 
 
 12 
 12 
 12 
..
.

The next problem is that of calculating the matrix elements of the operators Lx and
Ly . To do this, use is made of the ladder operators L± that satisfy

L+ Ylm = cYl,m+1 ,
L− Ylm = c0 Yl,m−1 .

c and c0 are the only nonzero matrix elements of J+ and J− :

c = hYl,m+1 |L+ Ylm i ,

c0 = hYl,m−1 |L− Ylm i .

It turns out that

p
L− Ylm = ~ l(l + 1) − m(m − 1)Yl,m−1 ,
p
L+ Ylm = ~ l(l + 1) − m(m + 1)Yl,m+1 .

In matrix form:
 0 
√
0 2 √
0 2
 
 
 0 √ 
 0 4 √

 0 6

 √ 
 0 6 √


 0 4 

0
L− = ~  0
√
6
.
√
 
0 10
 
 √ 
 0 12 √ 
 0 12 √

 0 10

 √ 
 0 6 
 0 
..
.

Note that L− and L+ are not self-adjoint.

54

E XERCISE 9.2. Prove the above formulae. (Hint: Use the fact that L+ and L−
are Hermitian adjoints L− = L∗+ , and the identity L2 = L+ L− + L2z − ~Lz to
compute first hYl,m |L+ L− Ylm i = [l(l + 1) − m(m − 1)] ~2 .)
L+ +L− L+ −L−
From the definition Lx = 2 and Ly = 2i we get the nonzero matrix
elements of Lx and Ly :
p ~
hYl,m−1 |Lx Ylm i = (l + m)(l − m + 1) ,
2
p ~
hYl,m+1 |Lx Ylm i = (l + m + 1)(l − m) ,
2
p i~
hYl,m−1 |Ly Ylm i = (l + m)(l − m + 1) ,
2
p ~
hYl,m+1 |Ly Ylm i = (l + m)(l − m + 1) .
2i
9.2. Spin. Thus far, only orbital angular momentum has been dealt with ex-
plicitly. However, it was seen that the formalism based on the commutation rela-
tions permitted either half-integral or integral values for l: the restriction to integral
l-values resulted from the explicit form of the operators Lz = xpy − yp x, etc., and
the requirement of a single-values wavefunction.
Experiments confirmed that particles have an internal degree of freedom that
obey the commutation relations of an angular moments but they are not associated
to the ‘motion’ of the particles. This degree of freedom is called spin angular
momentum of the particle. For spin angular momentum, it is found empirically
that the quantum number may take on either integral or half-integral values. The
relations obtained for the spin can be derived form the commutation relations. We
therefore have the result that the simultaneous eigenfunctions of the square of the
spin, denoted by S 2 , and the z-component of the spin, denoted by Sz , are given by
S 2 ψsms = s(s + 1)~2 ψsms ,
Sz ψsms = ms ~ψsms ,
where s may take on either integral or half-integral values depending on the nature
of the particle and ms = −s, −s + 1, . . . , +s − 1, +s. It is also found that for
a given particle, s is fixed (a number characterizing the particle), while L2 can
assume any value l(l + 1)~2 .
Note that the spin is a purely quantum observable. This can be seen heuristi-
cally: as ~ → 0, S 2 → 0. In the classical limit, ~ → 0, but l can tend to infinity in
such a way that l(l + 1)~ remains finite; this is not possible for the spin since s is
fixed.

9.3. Total angular momentum. Consider tho problem of the addition of two
kinds of angular momenta, the orbital angular momentum and the spin of a particle.
 55

The relations which we will obtain, however, are valid for any two commuting
angular momenta (e.g., total spin of a system of two particles).
The total angular momentum J can be written as the sum of the orbital and the
spin angular momenta:

J = L + S,

where J has components

Jx = Lx + Sx , Jy = Ly + Sy , Jz = Lz + Sz .

Since the angular and spin momenta commute, [Lj , Sk ] = 0, the total angular
momentum J satisfies the same commutation relations of L and S. Therefore J 2
and Jz commute

J 2 ψjmj = j(j + 1)~2 ψjmj ,

Jz ψjmj = mj ~2 ψjmj ,

Consider first the eigenvalues of

Jz = Lz + Sz .

They are given by the sum mj = ml +ms of the eigenvalues of Lz and Sz (because
[Lz , Sz ] = 0). L2 and S 2 have eigenvalues l(l + 1)~2 and s(s + 1)~2 , respectively,
and the square of the total angular momentum is

J 2 = L2 + S 2 + 2(Lx Sx + Ly Sy + Lz Sz ).

We want to determine the eigenvalues of J 2 . First, note that J 2 , L2 , S 2 , Jz all

commute with one another. It is also clear that L2 , Lz , S 2 , Sz form another mu-
tually commuting set of observables. Therefore we have (at least) two alternative
sets of four operators which are mutually commuting. If we fix the values of l and
s, the corresponding subspace has dimension

(2l + 1) × (2s + 1).

(This is the number of possible ‘orientations’ of ml and ms .)

9.4. Sum of angular momenta. Consider a system characterised by two an-

gular moments J1 and J2 with [J1 , J2 ] = 0. We want to determine the eigenvalues
of

J = J1 + J2 = J1 ⊗ 1 + 1 ⊗ J2 .
56

From the previous considerations we have that

J1z has eigenvalues ~m1
J2z has eigenvalues ~m2
Jz has eigenvalues ~m = ~(m1 + m2 )
J12 has eigenvalues ~2 j1 (j1 + 1)
J22 has eigenvalues ~2 j2 (j2 + 1)
J 2 has eigenvalues ~2 j(j + 1).
A complete orthonormal set of the Hilbert space H = H1 ⊗ H2 of the system is
{|j1 , j2 ; m1 , m2 i} = {|j1 , m1 i ⊗ |j1 , m2 i}
Consider the subspace of H generated by the system
Sj1 j2 = {|j1 , j2 ; m1 , m2 i : |m1 | ≤ j1 , |m2 | ≤ j2 }.
for fixed values of j1 and j2 . The dimension of this subspace if (2j1 + 1)(2j2 + 1).
Consider another basis in this subspace, namely
Tj1 j2 = {|j1 , j2 ; j, mi}.
Note that j and m are defined by the conditions that ~2 j(j + 1) is an eigenvalue of
J 2 , and ~m is an eigenvalue of Jz . The eigenvalue m assume the 2j + 1 values,
from −j to +j. It remains to find the range of values of j.
We first compute the largest value of jmax of j. Suppose that j = jmax and
m = j; then m = mmax = (m1 + m2 )max = (m1 )max + (m2 )max = j1 + j2 .
Hence,
jmax = j1 + j2
and
|j1 , j2 ; m1 = j1 , m2 = j2 i = |j1 , j2 ; j = j1 + j2 , m = ji .
We show now that the second largest value of j is j = j1 + j2 − 1. Consider the
eigenvalue ~m = ~(m1 + m2 − 1). There are two corresponding eigenstates in the
set Sj1 j2 :
|j1 , j2 ; m1 = j − 1, m2 = ji and |j1 , j2 ; m1 = j, m2 = j − 1i .
In other words, the subspace generated by the eigenstates of Jz with eigenvalue
~m = ~(j1 + j2 − 1) has dimension 2. Therefore, there are two eigenstates of Jz
in the set Tj1 j2 . One is
|j1 , j2 ; j = j1 + j2 , m = j − 1i ;
the other must be an eigenstate of J 2 corresponding to another eigenvalue and the
only possibility is
|j1 , j2 ; j = j1 + j2 − 1, m = ji .
 57

Thus, j1 + j2 − 1 is an eigenvalue of J 2 . This reasoning can be reiterated: j can

assume values j1 + j2 , j1 + j2 − 1, j1 + j2 − 2, etc. until a minimum value jmin .
To determine jmin we reason as follows. The subspace generated by Sj1 j2 has
dimension (2j1 + 1)(2j2 + 1). The number of vectors in the second basis Tj1 j2
must be the same. Therefore
jX
1 +j2

(2j1 + 1)(2j2 + 1) = (2j + 1).

j=jmin

This is an equation in the unknown jmin whose solution is

jmin = |j1 − j2 |.
We conclude that for the sum of two angular momenta, the eigenvalues of J 2 =
(J1 + J2 )2 has eigenvalues ~2 j(j + 1) where j = j1 + j2 , j1 + j2 − 1, . . . , |j1 − j2 |.
9.4.1. Clebsch-Gordan coefficients. We found that
Sj1 j2 = {|j1 , j2 ; m1 , m2 i : |m1 | ≤ j1 , |m2 | ≤ j2 }.
and
Tj1 j2 = {|j1 , j2 ; j, mi : j = j1 + j2 , j1 + j2 − 1, . . . , |j1 − j2 |, |m| ≤ j}.
are two orthonormal sets of the same Hilbert space describing two angular mo-
menta (the angular momenta of two particles, or the spin and orbital angular mo-
mentum of the same particle, etc.). Therefore, any vector in Tj1 j2 can be written as
linear combination of vectors is Sj1 j2 :
X
(9.6) |j1 , j2 ; j, mi = |j1 , j2 ; m1 , m2 i hj1 , j2 ; m1 , m2 |j1 , j2 ; j, mi
m1 ,m2

The coefficients hj1 , j2 ; m1 , m2 |j1 , j2 ; j, mi of this expansion are called Clebsch-

Gordan (CG) coefficients. We can suppress the label j1 , j2 from the ket for con-
venience.
Suppose without loss of generality that j1 ≥ j2 . From the properties of addi-
tion of angular momenta we have
hj1 , j2 ; m1 , m2 |j, mi =
6 0 ⇐⇒ j1 − j2 ≤ j ≤ j1 + j2
hj1 , j2 ; m1 , m2 |j, mi =
6 0 ⇐⇒ m = m1 + m2 .
(The first condition is called triangle condition, for geometrically it means that we
must be able to form a triangle with sides j1 , j2 and j.) Moreover we can always
choose the CG coefficients so that
hj1 , j2 ; m1 , m2 |j, mi ∈ R
hj1 , j2 ; j1 , j − j1 |j, ji > 0.
Another useful property is
hj1 , j2 ; m1 , m2 |j, mi = (−1)j1 +j2 −j hj1 , j2 ; −m1 , −m2 |j, −mi .
58

This relation halves the work in the computations.

The calculations of the CG coefficients proceed as follows. First, from the
identity

|j1 , j2 ; m1 = j1 , m2 = j2 i = |j1 , j2 ; j = j1 + j2 , m = ji

we have
hj1 , j2 ; j1 , j2 |j1 + j2 , j1 + j2 i = 1.
Then we apply the operator J− = J1,− ⊗ 1 + 1 ⊗ J2,− to both sides and we recall
that
p
J− |j, mi = ~ j(j + 1) − m(m − 1) |j, m − 1i
p
J1,− |j1 , m1 i = ~ j1 (j1 + 1) − m1 (m1 − 1) |j1 , m1 − 1i
p
J2,− |j2 , m2 i = ~ j2 (j2 + 1) − m2 (m2 − 1) |j2 , m2 − 1i .

The other coefficients can be found by symmetry.

If we arrange the CG coefficients into a matrix, we find it is unitary. This
follows from the fact that it relates one orthonormal set to another. Since the CG
are all real, this matrix is orthogonal.
Note that the relation (9.6) can be inverted (using the orthogonality of the CG
coefficients matrix). We can write
X
(9.7) |j1 , j2 ; m1 , m2 i = ξj |j1 , j2 ; j, m = m1 + m2 i ,
j

where ξj = hj1 , j2 ; j, m = m1 + m2 |j1 , j2 ; m1 , m2 i. In this way, the tensor prod-

uct |j1 , m1 i⊗|j2 , m2 i can be written as linear combination of the vectors |j, m = m1 + m2 i,
where j = j1 + j2 , . . . , |j1 − j2 |.
When j1 = j2 = 1/2, we write symbolically
1 1
⊗ = 1 ⊕ 0;
2 2
when j1 = 1, j2 = 1/2:
1 3 1
1⊗ = ⊕ ;
2 2 2
if j1 = 1, j2 = 1:
1 ⊗ 1 = 2 ⊕ 1 ⊕ 0,
and so on.

E XAMPLE 9.1. Let J~ = J~1 + J~2 be the sum of two (commuting) angular
momenta. Suppose that j1 = 1/2 and j2 = 1/2. If we measure J 2 , the possible
 59

outcomes are ~2 j(j + 1) with j = 0, 1. The nonzero CG coefficients are

  
1 1 1 1 
, ; + , + 1, 1 = 1
2 2 2 2
  
1 1 1 1  1
, ; + , − 1, 0 = + √
2 2 2 2 2
  
1 1 1 1  1
, ; − , + 1, 0 = + √
2 2 2 2 2
  
1 1 1 1  1
, ; + , − 0, 0 = + √
2 2 2 2 2
  
1 1 1 1  1
, ; − , + 0, 0 = − √ .
2 2 2 2 2
The CG coefficients can be represented in a square matrix:
  1 1 
|1, 1i
  
1 0√ 0√ 0 + 2 , + 2 
 |1, 0i  0 1/ 2 1/ 2 0 + 1 , − 1 
  2 2  .
|1, −1i = 0
  
0√ 0√ 1 − 21 , + 12 
|0, 0i 0 1/ 2 −1/ 2 0 − 1 , − 1
2 2

10. Identical particles

Although the definition of identical particles is the same classically and quan-
tum mechanically, the implications are different in the two cases. When a system
contains a number of particles of the same kind, e.g. a number of electrons, the par-
ticles are indistinguishable one from another. No observable change is made when
two of them are interchanged. This circumstance gives rise to some curious phe-
nomena in quantum mechanics having no analogue in the classical theory, which
arise from the fact that in quantum mechanics a transition may occur resulting in
merely the interchange of two identical particles, which transition then could not
be detected by any observational means. A satisfactory theory ought, of course, to
count two indistinguishable states as the same state and to deny that any transition
does occur when two identical particles are swapped.
Consider, for simplicity, two quantum particles. Their state is represented by
the wavefunction ψ(x1 , x2 ). Consider the operator P that swap the two particles
P ψ(x1 , x2 ) = ψ(x2 , x1 ). If the particles are indistinguishable, then ψ(x2 , x1 ) =
eiα ψ(x1 , x2 ). If we swap again the particles we get

(10.1) P 2 ψ(x1 , x2 ) = e2iα ψ(x1 , x2 ) = ψ(x1 , x2 ),

so that eiα = ±1. This can be extended to an arbitrary number N of indistinguish-

able particles. The state is ψ(x1 , . . . , xN ). If we permute the i-th and the j-th
60

particles
Pij ψ(x1 , . . . , xi . . . , xj , . . . , xN ) = ψ(x1 , . . . , xj . . . , xi , . . . , xN )
= e2iα ψ(x1 , . . . , xi . . . , xj , . . . , xN ).
Again, the only possibilities (if we assume that the wavefunction is single-valued)
are e2iα = ±1. In other words, the state of N indistinguishable particles is totally
symmetric (e2iα = +1) or totally antisymmetric (e2iα = −1) under permutations.
Indistinguishable particles with symmetric states are called bosons; particles with
antisymmetric states are called fermions.

10.1. Bosonic and Fermionic Hilbert Spaces. Two identical bosons will al-
ways have symmetric state vectors and two identical fermions will always have
antisymmetric state vector. Let us call the Hilbert space of symmetric vectors VS
and the Hilbert space of antisymmetric vectors VA . We now examine the relations
between these two spaces and the tensor product V1 ⊗ V2 . The space V1 ⊗ V2
consists (in finite dimension) of all linear combinations of vectors of the form
|v1 v2 i = |v1 i ⊗ |v2 i. Suppose that the vectors {|ei i} is a basis of V1 (and of
V2 ). To each pair of vectors |ei ej i and |ej ei i (i 6= j) there is one bosonic vector
√1 (|ei ej i + |ej ei i), and one fermionic vector √1 (|ei ej i + |ej ei i). If ei = ej ,
2 2
the vector |ei ei i is already symmetric, i.e. bosonic. There is no corresponding
fermionic vector (the Pauli principle). We express this relation as
(10.2) V1 ⊗ V2 = VS ⊕ VA .
The case of N = 2 particles lacks one feature that is found at larger N . For every
N ! product vectors, we get only two acceptable (bosonic or fermionic) vectors.
Hence, the tensor product V ⊗N for N ≥ 3 is larger (in dimensionality) than VS ⊗
VA .

References
Here are the main sources I used during the preparation of these lecture notes.

[1] L. Pauling and E. B. Wilson, Introdution to quantum mechanics, McGraw-Hill Book Company,
1935.
[2] A. Peres, Quantum Theory: Concepts and Methods, Fundamental Theories of Physics 72,
Kluwer Academic Publishers, 2002.
[3] H. Maassen, Quantum Probability and Quantum Information Theory, In F. Benatti, M. Fannes,
R. Floreanini, and D. Petritis, editors, Quantum Information, Computation and Cryptography,
808, 65-108. Springer, Berlin, 2010.
[4] G. Nardulli, Meccanica Quantistica I, Principi, Collana di fisica e scienze esatte diretta da
Sergio Ratti, FrancoAngeli, 2001.
[5] K. R. Parthasarathy, Mathematical Foundations of Quantum Mechanics, Texts and readings in
Mathematics 35, Hindustan Book Agency, New Delhi, 2011.
[6] L. I. Schiff, Quantum mechanics, 3rd edition, McGraw-Hill, 1968.
[7] J. J. Sakurai, Modern Quantum Mechanics, Addison-Wesley, 1994.
 61

[8] R. Shankar, Principles of Quantum Mechanics, 2nd edition, Kluwer Academics, 1994.
[9] G. Teschl, Mathematical Methods in Quantum Mechanics With Applications to Schrödinger
Operators, 2nd edition, Graduate Studies in Mathematics Volume 157, American Mathematical
Society Providence, Rhode Island, 2010.
[10] H. Weyl, The theory of groups and quantum mechanics, Dover Publications, 1950.