Professional Documents
Culture Documents
Lecture notes
C ONTENTS
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
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.
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,
(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.
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
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
.. ..
. .
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
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
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 ).
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
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.
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.
~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
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.
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
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
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.
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.
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):
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
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
(with initial conditions H0 (x) = 1, H1 (x) = 2x), and the differential equation
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 .
−∞
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
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
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
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|
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.
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
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
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
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|).
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
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
m = 0, ±1, ±2, . . .
l = |m|, |m| + 1, |m| + 2, . . .
n = l + 1, l + 2, l + 3, . . .
7. Matrix Mechanics
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
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.
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).
7.4. 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
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 .
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.
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 ∧.
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).
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).
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.
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
ψn − n−1
P
j=0 hϕj |ψ1 i ϕj
(8.19) ϕn = Pn−1 .
ψn − j=0 hϕj |ψ1 i ϕj
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.
(8.21) M ⊕ M⊥ = H
in this situation.
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.
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 )⊥ .
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
(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 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.)
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
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
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 .
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
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
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
a = l(l + 1)~2 .
We can summarise as
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
The matrix elements for the z-component of the angular momentum in this repre-
sentation are (Ylm = |l, mi)
In a similar manner, the matrix elements of the square of the angular momentum
are
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 .
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
..
.
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,
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
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 ).
J = J1 + J2 = J1 ⊗ 1 + 1 ⊗ J2 .
56
|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 .
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
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
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.