Chapter 1

Spin and Dirac Notation

The spin of the electron was discovered spectroscopically and observed

most directly in the Stern-Gerlach experiment. It is an intrinsic angular mo-
mentum that has the same commutation rules as other angular momenta and
combines additively with other angular momenta. It does not correspond
to the spatial rotation of an object, because its eigenvalues are half-integral
multiples of ~ which are not allowed for such rotations. It is a new type
of quantity that cannot be described by a wave function dependent on an
angular variable.
To describe the spin properties of a particle, we introduce the more
general quantum mechanical formalism invented by Dirac. This formalism
is more general than the wave mechanical description treated up to now in
this course.
It is sufficiently general and flexible to accomodate any quantum system
whether or not it can be described by a spatial wave function. We will
illustrate its application to the electron spin, then discuss how it can be
applied to any situation in which it is possible to visualize measurement
of a suitable set of dynamical variables. It is a very adaptable notation
which lends itself readily to approximations and intuitive descriptions, some
of which we will illustrate below. We proceed to the discussion of spin by
describing the Stern-Gerlach experiment.

1.1 Stern-Gerlach Experiment

A charged particle with angular momentum carries a current distributed
in space more or less like the electric current in an ordinary circuit. The
circulating current produces a magnetic moment that affects and is affected

1
by nearby charges and fields. To fix the scale of the magnetic moment and
its relation to the angular momentum, we can consider a classical electron
travelling in a circular orbit of radius r with velocity v. The classical ex-
pression for the magnetic moment of a circuit is µ = IA/c, where I is the
current and A is the area of the circuit. The current is the charge e divided
by the time for one orbit 2πr/v, and the area is πr 2 . Introducing the angular
momentum L = mvr, we find
~
eL
µ
~ =− . (1.1)
2mc
The magnetic moment and the angular mometum are vectors that have
opposite directions due to the negative charge of the electron. Since L
comes in units of ~, the unit of magnetic moment is

µ = e~/2mc, (1.2)

known as the Bohr magneton. A magnetic moment experiences a torque

µ
~ ×B ~ but no net force in a uniform magnetic field. The direction of the
torque is such that it tends to align the moment with the field. The energy
−~µ · B~ associated with the interaction is lowest when the two vectors are
aligned, and the moment would align itself with the field in the presence of
damping. Without damping, the moment precesses about the direction of
the field. The rate of precession is obtained from the equation of motion
for the angular momentum dL/dt ~ = µ
~ × B.~ It is known as the Larmor
frequency, and has the value

ΩL = eB/2mc (1.3)

independent of the angular momentum.

Figure 1.1: Experimental setup for the Stern-Gerlach Experiment

Stern and Gerlach used a non-uniform magnetic field to exert a net force
on the magnetic moment of an atom. A schematic sketch of their apparatus

2
is shown in Fig.(1.1). The asymmetric pole faces of the magnet produce a
field in the z direction with a gradient in the same direction. The z com-
ponent of the force on the moment is ∂(µ z Bz )/∂z = µz ∂Bz /∂z. When a
beam of atoms passes through the space between the pole faces, the force
will cause an atom to experience a deflection in the z direction proportional
to the z component of its magnetic moment. The deflection is observed by
registering the atom when it strikes a screen on the far side of the magnet.
The deflection acts as a measure of µz , and by virtue of Eq.(1.1) of Lz . The
source of atoms in the beam is an oven that can be expected to give the mo-
ments random orientations. Classical magnetic moments whose direction is
randomly distributed in space will give a uniform distribution of deflections.
Quantization of the angular momentum, however, is expected to produce
discrete spots, one for each allowed value of the z component of the angular
momentum. The orbital angular momentum with integral m would always
produce an odd number of spots. Stern and Gerlach found just two spots
when the experiment was done with silver atoms, indicating that m has the
values ±12 . This is allowed by the commutation rules for angular momen-
tum, but forbidden for orbital angular momentum. Thus they found direct
evidence for the spin of the electron attached to the atom. According to the
rules for angular momentum, the total spin angular momentum quantum
number, denoted by s in place of `, is s = 12 .
Quantization implies that every electron in the beam, whatever the ori-
entation of its spin, reaches one of the two spots corresponding to m = 12
(spin up) and m = −12 (spin down). This can be confirmed by successive
Stern-Gerlach experiments with the gradients in different directions. Spin
up electrons selected from the first magnet have their spin in the direction
of its gradient. For example, if the field gradient is in the x direction and
the spin up beam is selected, the atoms in this beam will have S x = ~/2.
When an atom with this spin is sent through a Stern-Gerlach magnet with
field gradient in the z direction, they still reach one of the same two spots
corresponding to spin up and spin down in the z direction. This is clear
indication that there are just two basis states of the spin. Other states such
as the ones coming form the oven or from the first magnet can be treated
as superpositions of these.

1.2 Representation of Spin States

In Dirac’s notation the label for the state is enclosed within brackets |i.
Here we use the labels u and d for the spin up and spin down states with the

3
magnet oriented in the z direction, denoting them by |ui and |di. Since any
incident spin is in one of the two states after passage through the magnet,
we draw on the principles of wave mechanics to express an arbitrary state
as a superposition of the two:

|ψi = a|ui + b|di. (1.4)

The squares of the absolute values of the coefficients a and b are the prob-
abilities for observing deflections corresponding to spin up and spin down
after passage through the magnet. To complete the quantum formalism, we
introduce the conjugate of a state and the inner product. The conjugate of
the state |ψi is denoted by

hψ| = a∗ hu| + b∗ hd|.

It can also be written as |ψi† . The inner product is obtained by putting a

conjugated state to the left of an unconjugated state, as in hψ 1 |ψ2 i. Com-
bining the states in this way produces a bracket, which prompted Dirac to
call the conjugated state a bra and the unconjugated state a ket. The defi-
nition of the conjugate is completed by assuming orthonormality conditions
for the basis states:

hu|ui = hd|di = 1,
hu|di = hd|ui = 0.

Using these relations, the inner product of two states defined as above with
coefficients a1 , b1 and a2 , b2 is

hψ1 |ψ2 i = a∗1 a2 + b∗1 b2 .

The inner product satisfies all the conditions given in Eqs.(1.5)- (1.9), which
now look like this:

hψ1 |ψ1 i is real (1.5)

hψ1 |ψ1 i ≥ 0 (1.6)
∗
hψ1 |ψ2 i = hψ2 |ψ1 i (1.7)
hψ1 |(a|ψ2 i + b|ψ3 i) = ahψ1 |ψ2 i + bhψ1 |ψ3 i (1.8)
∗ ∗
(ahψ1 | + bhψ2 )|ψ3 i = a hψ1 |ψ3 i + b hψ2 |ψ3 i (1.9)

An operator can also act on the states before the inner product is taken,
giving quantities of the form hψ1 |A|ψ2 i. When an operator acts on the

4
conjugated state, it is less awkward to express it as hψ 2 |A|ψ1 i∗ than to use
(A|ψ1 i)† |ψ2 i. Thus the Hermitian conjugate is defined by means of the
relation
hψ2 |A|ψ1 i∗ = hψ1 |A† |ψ2 i.
The fact that any state can be expressed as a superposition of the basis
states |ui and |di means that this set must be complete within the context
of spin states. The completeness relation can be expressed mathematically.
Note that the coefficients a and b in the superposition can be calculated as
inner products a = hu|ψi and b = hd|ψi. If we insert these formal expressions
into the superposition equation and rearrange it so that the inner products
in each term are to the right of the basis states, we find

|ψi = |uihu|ψi + |dihd|ψi,

= (|uihu| + |dihd|) |ψi.

Since this is valid for all ψ, the opeator in parentheses must be the unit
(identity) operator. Thus the equation

|uihu| + |dihd| = 1 (1.10)

expresses the completeness relation for the basis set |ui, |di. In general, a
basis is complete if the sum over all basis states of the product of the basis
state on the left and its conjugate on the right is equal to the identity. Note
that when the product is taken in the order bra, ket, the result is the inner
product–a pure number, while if it is taken in the order ket, bra, the result
is an operator that can act on a state to produce a new state.
While Dirac bras and kets express the underlying framework of quantum
mechanics, it is often convenient in practice to employ a matrix represen-
tation of states and operators. The matrices are constructed in exactly the
same way as they were with wave functions. Matrix elements of operators
have the form hψ1 |A|ψ2 i, with basis states hψ1 | labelling the rows and |ψ2 i
labelling the columns. The column matrix representing a ket |ψi has ele-
ments hψ1 |ψi with basis states |ψ1 i labelling the rows. Similarly hψ|ψ 1 i are
the elements of the row matrix representing the conjugated state. Here also,
taking the conjugate of a state or the Hermitian conjugate of an operator
takes the complex conjugate transpose of the corresponding matrix.
The spin S ~ itself is an operator with x, y, and z components. When
using the matrix representation, it is conventional to define dimensionless
matrices ~σ via
S~ = ~ ~σ .
2

5
The matrices ~σ are the Pauli spin matrices. They can be written out by
noting that the basis states |ui and |di are eigenstates of σ z with eigenvalues
+1 and -1, respectively. The other components are obtained by using the
raising and lowering operators Eqs.(??) and (??) to find their action on the
basis states. We find
 
0 1
σx = , (1.11)
1 0
 
0 −i
σy = , (1.12)
i 0
 
1 0
σz = . (1.13)
0 −1

As an illustration of the application of these matrices, let us outline the

determination of the spin state of the electron that emerges in the spin up
beam of a Stern Gerlach experiment in which the field and its gradient point
in an arbitrary direction defined by the polar angles θ and φ. If n̂ is a unit
vector in this direction, then σn = n̂ · ~σ is the corresponding component
of the spin. The matrix eigenvalue equation σ n ψ = +1ψ then determines
the column matrix ψ that represents the desired state. The details of the
calculation are left to the problems. For future reference, the result is
 −iφ 
1 e sin θ
ψ=p (1.14)
2(1 − cos θ) 1 − cos θ

1.3 Dynamics of Spin

If a classical spin is placed in a uniform magnetic field, it will precess at
the Larmor frequency. To see what happens in quantum theory, we must
solve the Schrödinger equation that governs the time evolution of the spin
state |ψ(t)i. In the new notation, the Schrödinger equation becomes
d|ψ(t)i
i~ = H|ψ(t)i.
dt
The role of the Hamiltonian is played by the energy of a spin in a uniform
~ The magnetic moment is correctly given by Eq.(1.1) for
field H = −~µ · B.
the moment associated with the orbital angular momentum. But there is
a difference in the relation when it concerns the spin moment. The correct
relation is written
ge ~ ge~
µ
~= S= ~σ (1.15)
2mc 4mc

6
The ‘g-factor’ in this expression was introduced empirically at first with a
value of 2. Later it was explained by Thomas as a relativistic effect, now
known as the Thomas precession. It appears automatically with the value 2
in the relativistic treatment based on the Dirac equation (treated in a later
course). For a particle other than the electron or the muon, it may have
other values. The other particles are composite, and the g-factor can be
attributed to a combination of spin and orbital moments of the constituents
(usually quarks and gluons). For example, the g-factor of the proton is 2.73
and the g-factor of the neutron is -1.78. Even for electrons and muons, there
are small but observable corrections to the g-factor due to the interaction of
these particles with the quantized electromagnetic field. In this course we
accept the g-factor of 2 as an empirical fact.
We will discuss two ways of solving the Schrödinger equation with a
particular initial state for a spin in a magnetic field. Both ways are useful
in other contexts. (i) The first way uses a superposition of eigenstates of
the Hamiltonian. The temporal behavior of an eigenstate is expressed by an
exponential factor e−iEt/~ , where E is the eigenvalue. The coefficients in the
superposition are found by matching the initial state to the superposition at
t = 0. (ii) The second way is to write the equation in matrix form, and solve
directly for the elements of the matrix representing the state, imposing the
initial conditions.
Let an electron with its spin in the z direction be placed in a magnetic
field in the x direction. We first solve this problem by superposition of
eigenstates of H (i above). The Hamiltonian in matrix form is

ge~B
H = − σx
4mc 
0 1
= −~ΩL (1.16)
1 0

We have used Eq.(1.3) and g=2 in the last equality. The eigenvalues of H
are found in the standard way to be ~Ω L and −~ΩL , with corresponding
eigenvectors
 p1 
ψ1 = p2
1
2
 p1 
ψ2 = p2
− 12

7
It can be seen by inspection that the initial state is
  q  p1  q  p1 
1
ψ(0) = = 12 p12 + 12 p2
0 2 − 12

Multiplying the eigenstates in this equation by the exponential factors that

give their time dependence yields
q  p1  q  p1 
iΩL t
ψ(t) = 1
2
p 2
1 e + 12 p2 e−iΩL t
2 − 12
 
cos ΩL t
= (1.17)
i sin ΩL t

The problem can also be solved by direct solution of Eq.(1.3) in its matrix
representation (ii above):
    
ȧ 0 1 a
i~ = −~ΩL ,
ḃ 1 0 b

with dots standing for time derivatives. This reduces to the set of equations

iȧ = −ΩL b,
iḃ = −ΩL b,

which can be solved by differentiating the first and substituting the second
to obtain ä = −Ω2L a. Solving this equation subject to the initial condition
a(0) = 1 gives a = cos ΩL t and the first equation then gives b = i sin Ω L t.
Eq.(1.17) shows that the probability of spin up is cos 2 ΩL t. This oscil-
lating probability is the quantum manifestation of precession. To observe it
we must repeatedly observe the spin after different times of exposure to the
B field. In practice, a beam of fixed velocity can be sent through a magnet
that produces a spatially constant magnetic field. The magnetic field can be
varied to change the Larmor frequency and therefore the angle of precession
during the fixed time the particle is in the field. The spin actually precesses
at twice the Larmor frequency because cos Ω L t is unity initially and again
after a time π/ΩL , indicating that the spin has made a complete rotation.
Although the precession rate is generally independent of the angular mo-
mentum, the g factor of 2 causes the spin to precess twice as fast as the
orbital angular momentum.
Another interesting feature of the rotation of the spin is that the spin
does not come back to its initial state after a complete rotation. Instead it

8
comes back to the negative of its initial state. The state picks up a phase
factor (-1) in a 2π rotation. This poses no problem in the case of spin be-
cause the angle of rotation is not a coordinate of the wave function, and
thus no discontinuity of the wave function occurs. If this were the case, as
it is for orbital angular momentum, there would be a large localized linear
momentum at the site of the discontinuity, causing disruption and spread of
the wave function. In the spin case, the rotation angle is contained in the
coefficients in the superposition that defines the state, and the 2π rotation
only causes an overall phase factor (-1) which does not affect any probabili-
ties or give rise to large linear momenta. It is only observable in a carefully
designed and conducted experiment where the spin is passed through an
interferometer and one of the split beams is exposed to a magnetic field.
This alters the relative phase between the interfering beams and thus shifts
the interference pattern when the beams are recombined. Experiments of
this type have been conducted with neutrons by Collela, Overhauser, and
Werner, verifying the change of sign in a 2π rotation.

1.4 Systems with Two Spins

The spin, being the simplest non-trivial quantum system, is a good sys-
tem for dealing with new topics, such as systems with more than one particle
When we treat a system of two spins—examples include the proton and elec-
tron spins in a hydrogen atom, the two electron spins in a helium atom, and
the proton and neutron spins in a deuterium nucleus—each spin has its own
operators. Let them be called S~1 and S~2 . The components of one spin vector
commute with all the components of the other spin vector. Starting from
basis states for each spin, a set of basis states for the two spin system can
be created by writing a product state with one factor from each basis. For
example, |ui1 |di2 would be a state in which particle 1 has spin up and par-
ticle 2 has spin down. We will omit the subscripts with the understanding
that the first factor is for particle 1 and the second for particle 2: |ui|di.
The product here is not a regular product–no multiplication is involved–but
a type of product known as a direct product. It can’t be reduced further.
When an operator such as S1z (z component of the spin of particle 1) acts
on a direct product state, it acts only on its part of the product, leaving the
part referring to the other spin unchanged. Thus S 1x |di|di = |ui|di, and
S2y |ui|ui = i|ui|di. The complete basis for the two-spin system is |ui|ui,
|ui|di, |di|ui, and |di|di. Any state of the two spin system can be written
as a linear superposition of these four states.

9
 With two spin operators, we can define the total spin S ~ = S~1 + S~2 . Each
component of S ~ is the sum of the corresponding components of the individ-
ual spins. The components of the total spin satisfy the same commutation
rules as the individual spins, and indeed as any angular momentum. This
means that all the commutation rule-based arguments used in connection
with orbital angular momentum are valid here. There are raising and low-
ering operators, the states and eigenvalues are arranged in multiplets, the
eigenvalues of Sz are of the form ~mS with mS an integer or half-integer,
the eigenvalues of S 2 are of the form ~2 s(s + 1) with s being the largest
value of mS in the multiplet. There are two distinct choices for a complete
set of commuting operators to describe the spin eigenstates, given that the
z direction is chosen as the principal direction. In the basis set obtained
from the product states, the set comprises S 12 , S22 , S1z , and S2z , the union
of the complete sets for each spin. A The second possible choice consists
of S12 , S22 , S 2 , and Sz . No other distinct spin operators can be added to
either of these sets. For example, S 2 fails to commute with S1z and S2z and
these operators cannot be added to the second set above. To employ a basis
defined by the second complete set, we proceed to find the eigenstates of S 2
and Sz in terms of the original product basis.
Since the original basis states are eigenstates of S 1z and S2z , they are also
eigenstates of the total Sz with eigenvalues ~mS with mS = 1, 0, 0, and − 1.
Since a state with mS = 1 is present, there must by a multiplet with s = 1,
since s is always the largest value of m S in a multiplet. Application of the
lowering operator to this state gives a unique state with s = 1 and m S = 0.
This leaves another independent state with m S = 0, which must belong to
an s = 0 multiplet. Note that we are able to determine the values of S 2
by simply counting the states with different eigenvalues of S z . Carrying out
the calculation explicitly leads to expressions for the states of the multiplets
in terms of the basis states. Labelling the states by |s, m S i, we calculate for
the s = 1 multiplet

|1 1i = |ui|ui,
1
|1 0i = p (S1x + S2x − iS1y − iS2y )|1 1i
~ s(s + 1) − mS (mS − 1)
q
= 1 (|ui|di + |di|ui),
2
|1 −1i = |di|di.

The spin version of Eq.(??) has been used. The highest and lowest states
are unique basis states for the given m S , while the mS = 0 state is a linear

10
combination of two independent states. The spin 0 state |0 0i can be found
either by using its orthogonality with |1 0i or by using one of the conditions
S± |0 0i = 0: q
|0 0i = 1 (|ui|di − |di|ui).
2

The spin 1 multiplet is referred to more explicitly as a triplet and the spin
0 multiplet as a singlet. The spin 1 states can be thought of as having
approximately parallel alignment of the two spins and the spin 0 state as
having approximately antiparallel alignment. The uncertainties in the x and
y components make exact parallel or antiparallel alignment impossible.
We remark here, primarily for future reference, that the singlet state is
antisymmetric under exchange of particle 1 and particle 2, and the triplet
state is symmetric. Exchange of particles is equivalent to exchanging the
labels in a product state. Symmetry under exchange is important because
there is a general quantum mechanical requirement in systems that comprise
two or more identical half-integral spin particles that the total wave function
must be antisymmetric with respect to exchange of any pair. That means
simultaneous exchange of all coordinates, spatial, spin, and any other inter-
nal coordinates. This does not forbid the triplet state, but it does require
that particles in a triplet spin state must be in spatial state that is antisym-
metric under exchange. The requirement of antisymmetry leads directly
to the Pauli exclusion principle, which is crucial for understanding atomic,
nuclear, and subnuclear structure. The requirement for integral spin parti-
cles is the opposite: the wave function must be symmetric under exchange.
Half-integral spin particles are called fermions and are said to obey Fermi
statistics, and integral spin particle are bosons and obey Bose statistics.
The symmetry or antisymmetry of the wave function has profound effects
on both large and small conglomerates of particles. The symmetry require-
ment under exchange is taken as an empirical postulate in non-relativistic
quantum theory. It can be proven in the context of relativistic quantum
theory where all particles are regarded as quanta of a field somewhat as
photons are quanta of the electromagnetic field.
Let us consider an example of a system in which two spins interact. The
interaction comes about because of a term in the Hamiltonian that involves
both spins. In our example, the interaction term is taken to be λ S~1 · S~2 , and
this term constitutes the entire Hamiltonian. It is a scalar quantity with no
directional preference like the potential in the central field problem. In such
a case we expect the total angular momentum to be conserved and thus to
commute with the Hamiltonian. Direct calculation shows that S z commutes
with H. By symmetry, the other components do so as well, and it follows

11
that S 2 also commutes. The commuting operators theorem tells us that
there are simultaneous eigenstates of H, S 2 , and Sz , i. e., the states of the
singlet and the triplet above must also be eigenstates of the Hamiltonian.
To find the eigenvalues we can make use of the identity

~ 2 = S~1 2 + S~2 2 + 2S~1 · S~2 .

S (1.18)

This equation can be solved for the dot product that appears in the Hamil-
tonian in terms of the squares of spin vectors that have definite values in
the states of the multiplets. We calculate

S 2 − S12 − S22
H|s mS i = λ |s mS i
2
1 3 1 3
2 s(s + 1) − 2 · 2 − 2 · 2
= λ~ |s mS i
 1 2 2
 4 λ~ |1 mS i for s = 1
=
− 43 λ~2 |0 0i for s = 0


The spectrum of this Hamiltonian is shown in Fig.(1.2). This somewhat

artificial example is supplemented by other examples involving two and even
three spins in the problems.

Figure 1.2: Spectrum of two interacting spins. The dotted line indicates
zero energy.

12
1.5 Combining Spin and Orbital Angular Momen-
tum
An electron in an atom or a nucleon in a nucleus has both spin and
orbital angular momentum. They combine as two commuting elements of a
composite system to form the total angular momnentum

J~ = L
~ + S.
~ (1.19)

It is easily seen that L2 and S 2 commute with all the components of J, ~

2 ~ ~
and therefore also with J . Individual components of L and S, however, do
not commute with J 2 . Components of the total angular momentum satisfy
commutation rules among themselves which have the same form as for the
orbital and spin angular momenta discussed earlier. This means that there
are eigenstates of J 2 and Jz which have the same kind of multiplet structure
seen there. We can form a basis for the composite system by taking direct
products the form Y`m |ui and Y`m |di of orbital angular momentum eigen-
states with spin eigenstates. There are two essentially different complete
sets of commuting operators here as there were in the problem of two spins.
States of the product basis are eigenstates of the complete commuting set
L2 , Lz , S 2 , and Sz . The second useful complete set, containing the total
angular momentum, comprises J 2 , Jz , L2 , and S 2 . It is useful to express
the eigenstates of the second set in terms of the eigenstates of the first. If
for clarity we call the m associated with the orbital angular momentum m L
and the m belonging to the total angular momentum m J , Eq.(1.19) implies
mJ = mL + mS . As with the spin, we can deduce the number of multiplets
and their quantum numbers by counting the m J values of the states in the
product basis. For each value of mL , mS can be either +12 or -12 . The val-
ues of mJ of the basis states are `+12 , `−12 , `−12 , ..., −`+12 , −`+12 , −`−12 . This
shows that multiplets with j = `+12 and j = `−12 can be formed. Labelling
the states by the quantum numbers `, j, and m J , and observing that the
state with the largest value of mJ is unique, we can write

|`, j = `+12 , mJ = `+12 i = Y`` |ui.

Other states of this multiplet can be generated by repeated application of

the lowering operator Lx − iLy + Sx − iSy . States of the j = `−12 multiplet
can then be found by using their orthogonality with the states of the j = `+ 12
multiplet with the same mJ . This somewhat lengthy procedure leads to the
eigenstates which can be labelled by the quantum numbers `, j, and m J .

13
The spin quantum number s is always 12 and need not be included in the
label. For reference, the states are
s s
` + 1 +m ` +12 −mJ
J
|`, `+12 , mJ i = 2
Y` mJ −12 |ui + Y` mJ +12 |di (1.20)
2` + 1 2` + 1
s s
1
` +2 −mJ ` +12 +mJ
|`, `−12 , mJ i = − Y` mJ −12 |ui+ Y` mJ +12 |di (1.21)
2` + 1 2` + 1

The coefficients in this superposition are known as Clebsch-Gordon coeffi-

cients, and they are tabulated as well as internetted for your convenience.
For small quantum numbers it is cooler and shows more virtuosity to find
them from scratch.

1.6 Spin-Orbit Coupling

When an electron moves in the electric field of the nucleus of an atom, it
experiences a magnetic field in its own rest frame. This magnetic field inter-
acts with the electron magnetic moment to produce spin-orbit coupling. As
we argue below, the spin-orbit interaction is represented by a term propor-
tional to L~ ·S~ in the Hamiltonian. This new term does not commute with
Lz , so the our initial treatment of the atom is no longer valid. The total
angular momentum J 2 does commute with both the initial Hamiltonian and
with the spin-orbit term. The energy eigenstates can now be chosen to be
simultaneous eigenstates of H, L2 , S 2 , J 2 , and Jz , but not of Lz and Sz .
Therefore the angular and spin part of the wave function now has the form
of the states (1.20) and (1.21) discussed in the last section. We deal qual-
itatively in this section with the changes in the hydrogen atom spectrum
brought about the spin-orbit coupling term.
In the rest frame of an electron orbiting in a hydrogen atom, the proton
moves with the same orbital speed v that the electron has in the labora-
tory frame. According to the Biot-Savart law, this produces a magnetic
field ev/cr 2 at the location of the electron. The magnitude of the field can
be expressed in term of the orbital angular momentum of the electron by
expressing v as L/mr. This field interacts with a magnetic moment of mag-
nitude geS/2mc. To get the correct sign and directional dependence, refer
to Fig.(1.3). Part (a) shows the electron and its angular momentum vector
in the laboratory frame. In (b), we see the motion of the proton in the
electron rest frame and the direction of the magnetic field it produces at the
location of the electron. This field has the same direction as the angular

14
momentum of the electron in (a). The interaction energy of the magnetic
field with the magnetic moment of the electron is -~µ · B~ and µ
~ is equal to
~
−(ge/2mc)S. Inserting the direction and magnitude of B, we obtain the
interaction energy
ge2 ~ ~
HSO = L · S. (1.22)
2m2 c2 r 3
The positive sign shows that the lowest energy occurs when L ~ and S ~ are
1
anti-parallel, i.e., in the state with j = `− 2 . The energy is of the form
~ · S.
VSO (r)L ~ The dot product can be evaluated in the same way it was for
spin, using the square of J~ = L~ + S.
~ In terms of the quantum numbers that
characterize the state,

` for j = `+12

HSO = VSO (r) (1.23)
−` − 1 for j = `−12

Figure 1.3: The spin-orbit interaction. (a.) In the rest frame of the proton,
the electron travels with velocity v and the proton produces only an electric
field. In the configuration shown, the orbital angular momentum vector of
the electron is directed out of the page. (b.) In the rest frame of the electron,
the proton moves with velocity v and produces a magnetic field also directed
out of the page at the location of the electron. In the lowest energy state,
the electron magnetic moment is parallel to the field. The spin is opposite
to the magnetic moment because of the negative charge of the electron, and
therefore is opposite to the orbital angular momentum in the lowest energy
state.

15
The effect of this term in the Hamiltonian is to split the degenerate states
with ` 6= 0 into a higher energy group of 2` + 2 states with j = `+ 12 and a
lower energy group of 2` states with j = `− 12 . It also adds a small term to
the radial potential which slightly modifies the radial wave equation. Since
the spin-orbit term is small compared to the Coulomb term in the potential,
the effect can be taken into account by perturbation theory, as we shall see
in Chapter 8.
The spin-orbit term binds the spin and the orbital angular momentum
together in opposite directions so that it requires energy, the energy splitting
between the j = `±12 states, to pull them apart. We will see in the chapter
on perturbation theory how an external magnetic field affects this binding
by acting differently on the spin and orbital parts of the angular momentum
vector.
Spin-orbit coupling plays a role in the nucleus also, where it is invoked
in the nuclear shell model to explain the order of levels and the spacing
between them. The sign of the nuclear spin-orbit term is negative, so that
the state with larger total angular momentum lies lower.

1.7 Model Hamiltonians

The Dirac notation is very flexible. It can be used to construct simple,
fairly easily soluble models that faithfully reproduce many features of the
quantum mechanical behavior of a system without the necessity of calculat-
ing the wave function in detail. In general, these models have parameters
that must be either calculated from wave mechanics or fitted from experi-
mental data. A simple example is provided by the case two wells with a finite
barrier between them. Let |1i and |2i be states with the particle in well 1
and well 2, respectively. Suppose the energies of the particles in the wells
in the absence of tunnelling through the barrier are  1 and 2 . If tunnelling
were impossible, these states would be eigenstates of the Hamiltonian with
these eigenvalues. How can we include the possibility of tunnelling in the
Hamiltonian? The Schrödinger equation tells us that the change in the wave
function in a short time interval is
∆t
∆|ψi = H|ψi.
i~
If tunnelling is possible, it means that a particle initially in well 1 in state |1i
will have a small admixture of the state |2i in well 2 after a short time ∆t.
Thus if |ψi = |1i initially, then H|1i on the right must have an admixture
of |2i to account for the tunnelling. This means that |1i is no longer an

16
eigenstate of H. The eigenvalue equation is modified by the addition of a
tunnelling term as follows

H|1i = 1 |1i + λ|2i. (1.24)

Similarly, when H act on |2i it gives

H|2i = 2 |2i + λ∗ |1i. (1.25)

In other words, if a particle can tunnel between wells, there must be a term
in the Hamiltonian that is responsible. This implies that there is a tunnelling
energy defined as |λ| associated with the possibility of tunnelling. There is
a probability amplitude λ∆t/i~ for the particle to tunnel in time ∆t. The
amplitude for tunnelling from 1 to 2 must be the complex conjugate of the
amplitude for tunnelling from 2 to 1 as indicated in Eq.(1.25) because of
the Hamiltonian must be Hermitian. This is most easily seen if we write the
Hamiltonian in matrix form
 
1 λ
H= (1.26)
λ∗ 2
The tunnelling amplitudes lie on opposite sides of the diagonal, and Her-
miticity requires that they be complex conjugates of one another.
Consider first the case where 1 = 2 = . By suitably defining the phases
of the basis states, we can make λ real and postive. Then the eigenvalues of
the Hamiltonian are  ± λ and the corresponding eigenstates are
q q
|ψ1 i = 1 |1i + 1 |2i, (1.27)
2 2
q q
|ψ2 i = 1 |1i − 1 |2i. (1.28)
2 2
(1.29)

Thus the eigenstates are equal superpositions of states |1i and |2i in the
two wells. In order to have an eigenstate, which is a stationary state with
constant probabilities for the particle to be in each well, the probability
must be non-zero in both wells so that the probability for tunnelling from
1 to 2 is compensated by the probability for tunnelling from 2 to 1. The
possibility of tunnelling causes the levels, which were initially degenerate, to
split by 2λ. It is revealing to examine the fate of a particle that is initially
localized in well 1. We use the matrix representation to calculate as follows
  q  p1  q  p1 
1 1 p2 + 12 p2
= 2 1
0 2 − 12

17
 q  p1  q  p1 
−i(+λ)t/~
→ 1
2
p2 e
1 + 12 p2
1 e−i(−λ)t/~
2 − 2
 
cos(λt/~)
= e−it/~ . (1.30)
−i sin(λt/~)

This state describes a particle that oscillates between the wells. After start-
ing in well 1 with 100% probability, if finds itself in well 2 with 100% prob-
ability after a time π~/2λ. After another period of the same duration, it
is again in well 1 with complete certainty. If we take the time to move to
the other well as ∆t, then there is a time-energy uncertainty principle at
work here. The energy uncertainty is ∆E = λ, giving ∆E∆t = π~/2. If the
tunnelling amplitude λ is small, the time to get to the other well is large,
and the uncertainty principle allows the energy uncertainty, and with it the
energy splitting, to be small.
The case of unequal energies is also interesting. The eigenvalues of the
Hamiltonian Eq.(1.26) are
p
1 + 2 ± (1 − 2 )2 + 4|λ|2
.
2
As we allow tunnelling by letting |λ| increase from zero, the upper level
moves higher in energy and the lower level moves lower. This is the quantum
mechanical phenomenon popularly known as level repulsion. The possibility
of a particle making a transition between two states causes the states to
move farther apart in energy. As |λ| increases, each eigenstate becomes a
superposition in which a little of the state in the other well is added to
the non-tunnelling eigenstate. For example, if  1 is the lowest energy, the
eigenstate for small |λ| is a superposition of a large amount of |1i and a
small amount of |2i. Although adding the higher energy raises the energy
of the state per se, it allows the system to take advantage of the tunnelling
energy λ, which decreases the energy of the system by a greater amount if
the phases of the states in the superposition are right.

1.8 Dirac Notation with Coordinates and Momenta

The Dirac notation can be fruitfully applied to any quantum problem,
including the wave mechanics described earlier in this course. As in the case
of spin, we begin by defining a set of basis states as the states resulting
from a set of measurements that distinguish all the independent states of
the system, i.e., measurement of a complete set of commuting operators.

18
For a spinless particle confined to the x axis, for example, a measurement
of the position suffices. Let |xi be the state in which a measurement of the
coordinate definitely gives the result x. Since x is a continuous variable,
we should really define |xi by a limiting procedure that starts from an in-
finitesimal interval on the x axis. We omit the details of such a procedure
and proceed on an intuitive basis. The orthonormality condition for this
continuous basis is
hx|x0 i = δ(x − x0 ),
and the completeness relation is
Z
|xihx|dx = 1.

The Kronecker deltas of the discrete basis are replaced by Dirac deltas and
the sums are replaced by integrals. A general state subjected to a measure-
ment of its coordinate will always yield some value x. It is reasonable to
assume, therefore, that any state can be represented in the form
Z
|ψi = ψ(x)|xidx, (1.31)

just as we did in Eq.(1.4) for a general spin state. The coefficient in the
superposition has been called ψ(x) because the square of its absolute value
gives the probabilty density that the result of a coordinate measurement
is x, i.e., it has the same interpretation as the wave function in our earlier
treatment. The orthonormality condition can be used to express the function
ψ(x) as an inner product
ψ(x) = hx|ψi. (1.32)
We can say that the wave function is the inner product of the state repre-
senting a particle localized at x with the state of the system.
Bases other than the position basis can be used as well. For example, in
the momentum basis, Eq.(1.31) is replaced by
X
|ψi = ck |ki, (1.33)
k

in which a sum occurs when k is quantized with periodic boundary condi-

tions in a large box. The state |ψi is the same state represented in the last
paragraph in the x basis, now expanded in terms of the k basis. The mo-
mentum space wave function ck is again an inner product ck = hk|ψi. The

19
results found√in Chapter 4 can all be obtained by using the inner product
hx|ki = eikx / ` together with the orthogonality and completeness relations

hk|k 0 i = δkk0 ,
X
|kihk| = 1.
k

For example the relation between ck to ψ(x) is derived as follows

ψ(x) = hx|ψi
X
= hx|kihk|ψi
k
X eikx
= √ hk|ψi
k
`
Xe ikx
= √ ck .
k
`

In the second line of the calculation, the completeness relation for the k basis
has been introduced. The coordinate space wave function is the Fourier
transform of the momentum space wave function. The converse can be
proved similarly.
We can calculate the inner product between two states in the coordinate
basis, obtaining the same result that was denoted in Chapter 6 by (ψ 1 , ψ2 ).
Henceforth we will use the Dirac notation for this quantity. To show the
equivalence of the two quantities, we use the completeness relation in the
coodinate basis:
Z
hψ1 |ψ2 i = dxhψ1 |xihx|ψ2 i
Z
= ψ1∗ (x)ψ2 (x)dx.
= (ψ1 , ψ2 ).

The same quantity can be expressed in the momentum basis by using the
corresponding completeness relation:
X
hψ1 |ψ2 i = hψ1 |kihk|ψ2 i
X
= c∗1k c2k .

For a real electron, we would have to measure its position and the z
component of its spin to determine its state completely. The appropriate

20
basis states would be denoted by |~r, m S i, and the wave function would be
h~r, mS |ψi also known as ψmS (~r). The inner product would involve an inte-
gration over ~r and a sum over mS :
XZ
hψ1 |ψ2 i = d3 rψ1m
∗
S
(~r)ψ2mS (~r).
mS

Our purpose here has been to introduce the Dirac notation and apply it
to some elementary cases. A full development of the formalism is left to a
later course.

1.9 Problems
1. Derive the percession rate Eq.(1.3) from the equation of motion for
the angular momentum.

2. Derive the Pauli spin matrices, Eqs.(1.11), Eqs.(1.12), and Eqs.(1.13).

3. Using the matrix representation, prove the completeness relation Eq.(1.10).

Find the eigenstates of σx and show that they satisfy the completeness
relation.

4. Show that the probability that a particle emerges from a Stern-Gerlach

magnet in a certain state is the square of the absolute value of the inner
product of the final state and the initial state.

5. Find the eigenstate of σy with eigenvalue +1. Calculate its inner

product with the state |ui.

6. Derive Eq.(1.14).

7. Find the probability that an electron with spin up in a direction with

polar angles θ and φ will be found with spin up if it is passed through
a Stern-Gerlach magnet with field and gradient in the z direction.

8. Let a beam of spin 12 atoms pass through 3 Stern-Gerlach magnets

successively. The outer magnets have field and gradient in the z di-
rection, and the middle magnet has spin in an arbitrary direction θ,
φ. Let the spin down beam be blocked in the first two magnets. What
is the probability that an atom emerges from the third magnet with
spin up as a function of θ and φ?

21
 9. Let an electron with its spin initially in the y direction be placed in
a uniform magnetic field B in the z direction. Find the probability
that its spin is in the y direction after time t. What is the precession
frequency? Does the wave function return to its initial state after
precessing thrrough 2π?

10. In the model tunnelling problem, it is possible to define a ”coordinate”

operator which has a value −a in the left well and +a in the right well.
Find the expectation value of the coordinate operator as a function of
time in the problem with equal ’s solved in Sec(8.7).

11. It is also possible to define a parity operator in this problem via P|1i =
|2i and P|2i = |1i. Show that the parity operator commutes with the
Hamiltonian, and find the parity of each eigenstate.

12. Solve the problem of three wells occupying the vertices of a triangle.
Find the energy eigenvalues if the energy in each well is  in the absence
of tunnelling, and if the tunnelling amplitude from each well to one of
its neighbors is λ.

13. A concrete example of level repulsion: consider the case in which the
energies of the isolated wells are 0 eV and 1 eV and the tunnelling
energy λ is 0.2 eV. Find the energy eigenvalues and eigenstates of the
system.

14. Prove that the quantity denoted previously by (ψ 1 , V ψ2 ) is the same as

hψ1 |V |ψ2 i when V is a function of the position. (HINT: By definition,
V |~ri = V (~r)|~ri when V acts on an eigenstate of position. Use the
completeness relation for position eigenstates between V and |ψ 2 i in
the Dirac bracket.)

22