A Brief Introduction

to Relativistic Quantum Mechanics

Hsin-Chia Cheng, U.C. Davis
1 Introduction
In Physics 215AB, you learned non-relativistic quantum mechanics, e.g., Schrodinger
equation,
E =
p
2
2m
+V,
E i

t
, p i,
i

t
=

2
2m

2
+V . (1)
Now we would like to extend quantum mechanics to the relativistic domain. The
natural thing at rst is to search for a relativistic single-particle wave equation
to replace the Schrodinger equation. It turns out that the form of the relativistic
equation depends on the spin of the particle,
spin-0 Klein-Gordon equation
spin-1/2 Dirac equation
spin-1 Proca equation
etc
It is useful to study these one-particle equations and their solutions for certain
problems. However, at certain point these one-particle relativistic quantum theory
encounter fatal inconsistencies and break down. Essentially, this is because while
energy is conserved in special relativity but mass is not. Particles with mass can
be created and destroyed in real physical processes. For example, pair annihilation
e
+
e

2, muon decay

. They cannot be described by single-particle

theory.
At that stage we are forced to abandon single-particle relativistic wave equations
and go to a many-particle theory in which particles can be created and destroyed,
that is, quantum eld theory, which is the subject of the course.
1
2 Summary of Special Relativity
An event occurs at a single point in space-time and is dened by its coordinates
x

, = 0, 1, 2, 3,
x
0
= ct, x
1
= x, x
2
= y, x
3
= z, (2)
in any given frame.
The interval between 2 events x

and x

is called s,
s
2
= c
2
(t

t)
2
(x x)
2
(y y)
2
(z z)
2
= (x
0
x
0
)
2
(x
1

x
1
)
2
(x
2
x
2
)
2
(x
3
x
3
)
2
. (3)
We dene the metric
g

=
_
_
_
_
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
_
_
_
_
, (4)
then we can write
s
2
=

,
g

(x

)(x

) = g

, (5)
where we have used the Einstein convention: repeated indices (1 upper + 1 lower)
are summed except when otherwise indicated.
Lorentz transformations
The postulates of Special Relativity tell us that the speed of light is the same in
any inertial frame. s
2
is invariant under transformations from one inertial frame
to any other. Such transformations are called Lorentz transformations. We will
only need to discuss the homogeneous Lorentz transformations (under which the
origin is not shifted) here,
x

. (6)
g

= g

= g

= g

= g

. (7)
Its convenient to use a matrix notation,
x

:
_
_
_
_
x
0
x
1
x
2
x
3
_
_
_
_
= x. (8)
2
s
2
= x
T
gx,
x

= x
g =
T
g (9)
Take the determinant,
det g = det
T
det g det , (10)
so det = 1 (+1: proper Lorentz transformations, 1: improper Lorentz trans-
formations).
Example: Rotations (proper):
x
0
= x
0
x
1
= x
1
cos +x
2
sin
x
2
= x
1
sin +x
2
cos
x
3
= x
3
(11)
=
_
_
_
_
1 0 0 0
0 cos sin 0
0 sin cos 0
0 0 0 1
_
_
_
_
(12)
Example: Boosts (proper):
t

= (t
v
c
2
x
1
) or x
0
= x
0
x
1
x
1
= (x
1
vt) = x
1
x
0
x
2
= x
2
x
3
= x
3
(13)
where
=
v
c
, =
1
_
1
2
. (14)
Its convenient to dene a quantity rapidity such that cosh = , sinh = ,
then
=
_
_
_
_
cosh sinh 0 0
sinh cosh 0 0
0 0 1 0
0 0 0 1
_
_
_
_
. (15)
One can easily check that det = cosh
2
sinh
2
= 1.
Four-vectors, tensors
A contravariant vector is a set of 4 quantities which transforms like x

under a
3
Lorentz transformation,
V

=
_
_
_
_
V
0
V
1
V
2
V
3
_
_
_
_
, V

. (16)
A covariant vector is a set of 4 quantities which transforms as
A

= A

1
_

,
1
= g
T
g. (17)
An upper index is called a contravariant index and a lower index is called a co-
variant index. Indices can be raised or lowered with the metric tensor g

and its
inverse g

= diag(1, 1, 1, 1), g

. The scalar product of a contravari-

ant vector and a covariant vector V

is invariant under Lorentz transformations.

Examples: Energy and momentum form a contravariant 4-vector,
p

= (
E
c
, p
x
, p
y
, p
z
). (18)
4- gradient,

=
_
1
c

t
,

x
,

y
,

z
_

(19)
is a covariant vector,

=
x

=
_

1
_

. (20)
One can generalize the concept to tensors,
T

(
1
)

(
1
)

. (21)
Maxwells equations in Lorentz covariant from (Heaviside-Lorentz conven-
tion)
E = (22)
B = 0 (23)
E +
1
c
B
t
= 0 (24)
B
1
c
E
t
=
1
c
J (25)
From the second equation we can dene a vector potential A such that
B = A (26)
4
Substituting it into the third equation, we have

_
E +
1
c
A
t
_
= 0, (27)
then we can dene a potential , such that
E =
1
c
A
t
. (28)
Gauge invariance: E, B are not changed under the following transformation,
A A
+
1
c

t
. (29)
(c, J) form a 4-vector J

. Charge conservation can be written in the Lorentz

covariant form,

= 0,
(, A) from a 4-vector A

(A

= (, A)), from which one can derive an antisym-

metric electromagnetic eld tensor,
F

(note:
i
=
i
=

x
i
, i = 1, 2, 3). (30)
F

=
_
_
_
_
0 E
x
E
y
E
z
E
x
0 B
z
B
y
E
y
B
z
0 B
x
E
z
B
y
B
x
0
_
_
_
_
, F

=
_
_
_
_
0 E
x
E
y
E
z
E
x
0 B
z
B
y
E
y
B
z
0 B
x
E
z
B
y
B
x
0
_
_
_
_
(31)
Maxwells equations in the covariant form:

=
1
c
J

(32)

= 0 (33)
where

1
2

, (34)

0123
and its even permutation = +1, its odd permutation = 1.
Gauge invariance: A

. One can check F

is invariant under this

transformation.
5
3 Klein-Gordon Equation
In non-relativistic mechanics, the energy for a free particle is
E =
p
2
2m
. (35)
To get quantum mechanics, we make the following substitutions:
E i

t
, p i, (36)
and the Schrodinger equation for a free particle is

2
2m

2
= i

t
. (37)
In relativistic mechanics, the energy of a free particle is
E =
_
p
2
c
2
+m
2
c
4
. (38)
Making the same substitution we obtain

2
c
2

2
+m
2
c
2
= i

t
. (39)
Its dicult to interpret the operator on the left hand side, so instead we try
E
2
= p
2
c
2
+m
2
c
4
(40)

_
i

t
_
2
=
2
c
2

2
+m
2
c
4
, (41)
or
1
c
2
_

t
_
2

2
2 =
m
2
c
2

2
, (42)
where
2 =
1
c
2
_

t
_
2

2
=

. (43)
Plane-wave solutions are readily found by inspection,
=
1

V
exp
_
i

p x
_
exp
_

Et
_
, (44)
where E
2
= p
2
c
2
+ m
2
c
4
and thus E =
_
p
2
c
2
+m
2
c
4
. Note that there is a
negative energy solution as well as a positive energy solution for each value of p.
Navely one should just discard the negative energy solution. For a free particle
in a positive energy state, there is no mechanism for it to make a transition to
6
the negative energy state. However, if there is some external potential, the Klein-
Gordon equation is then altered by the usual replacements,
E E e, p p
e
c
A, (45)
(i
t
e)
2
= c
2
(i
e
c
A)
2
+m
2
c
4
. (46)
The solution can always be expressed as a superposition of free particle solutions,
provided that the latter form a complete set. They from a complete set only if
the negative energy components are retained, so they cannot be simply discarded.
Recall the probability density and current in Schrodinger equation. If we multiply
the Schrodinger equation by

on the left and multiply the conjugate of the

Schrodinger equation by , and then take the dierence, we obtain

2
2m
(

) = i(

2
2m
(

) = i

t
(

) (47)
Using
s
=

, j
s
=

2mi
(

), we then obtain the equation of

continuity,

s
t
+ j
s
= 0 (48)
Now we can carry out the same procedure for the free-particle Klein-Gordon equa-
tion:

2 =
m
2
c
2

=
m
2
c
2

(49)
Taking the dierence, we obtain

22

) = 0. (50)
This suggests that we can dene a probability 4-current,
j

= (

), where is a constant (51)

and its conserved,

= 0, j

= (j
0
, j). To make j agree with j
s
, is chosen to
be =

2mi
. So,
=
j
0
c
=
i
2mc
2
_

t
_
. (52)
does reduce to
s
=

in the non-relativistic limit. However, is not positive

-denite and hence can not describe a probability density for a single particle.
Pauli and Weisskopf in 1934 showed that Klein-Gordon equation describes a spin-0
(scalar) eld. and j are interpreted as charge and current density of the particles
in the eld.
7
4 Dirac Equation
To solve the negative probability density problem of the Klein-Gordon equation,
people were looking for an equation which is rst order in /t. Such an equation
is found by Dirac.
It is dicult to take the square root of
2
c
2

2
+m
2
c
4
for a single wave function.
One can take the inspiration from E&M: Maxwells equations are rst-order but
combining them gives the second order wave equations.
Imagining that consists of N components
l
,
1
c

l
t
+
3

k=1
N

n=1

k
ln

n
x
k
+
imc

n=1

ln

n
= 0, (53)
where l = 1, 2, . . . , N, and x
k
= x, y, z, k = 1 , 2, 3.
=
_
_
_
_
_

2
.
.
.

N
_
_
_
_
_
, (54)
and
k
, are N N matrices. Using the matrix notation, we can write the
equations as
1
c

t
+ +
imc

= 0, (55)
where =
1
x+
2
y+
3
z. N components of describe a new degree of freedom
just as the components of the Maxwell eld describe the polarization of the light
quantum. In this case, the new degree of freedom is the spin of the particle and
is called a spinor.
We would like to have positive-denite and conserved probability, =

, where

is the hermitian conjugate of (so is a row matrix). Taking the hermitian

conjugate of Eq. (55),
1
c

t
+

imc

= 0. (56)
Multiplying the above equation by and then adding it to

(55), we obtain
1
c
_

t
+

t

_
+

+
imc

) = 0. (57)
The continuity equation

t
(

) + j = 0 (58)
8
can be obtained if

= ,

= , then
1
c

t
(

) + (

) = 0 (59)
with
j = c

. (60)
From Eq. (55) we can obtain the Hamiltonian,
H = i

t
=
_
c

i
+mc
2
_
. (61)
One can see that H is hermitian if , are hermitian.
To derive properties of , , we multiply Eq. (55) by the conjugate operator,
_
1
c

t

imc

__
1
c

t
+ +
imc

_
= 0

_
1
c
2

2
t
2

i

j
+
m
2
c
2

2

2

imc

(
i
+
i
)
i
_
= 0 (62)
We can rewrite
i

j
as
1
2
(
i

j
+
j

i
)
i

j
. Since its a relativistic system, the
second order equation should coincide with the Klein-Gordon equation. Therefore,
we must have

j
+
j

i
= 2
ij
I (63)

i
+
i
= 0 (64)

2
= I (65)
Because

i
=
i
= (I)
i
, (66)
if we take the determinant of the above equation,
det det
i
= (1)
N
det
i
det , (67)
we nd that N must be even. Next, we can rewrite the relation as
(
i
)
1

i
= (no summation). (68)
Taking the trace,
Tr
_
(
i
)
1

= Tr
_
(
i

i
)
1

= Tr[] = Tr[], (69)

we obtain Tr[] = 0. Similarly, one can derive Tr[
i
] = 0.
9
Covariant form of the Dirac equation
Dene

0
= ,

j
=
j
, j = 1, 2, 3

= (
0
,
1
,
2
,
3
),

= g

(70)
Multiply Eq. (55) by i,
i
_
1
c

t
+ +
imc

_
= 0

_
i
0

x
0
+i
j

x
j

mc

_
=
_
i

mc

_
= 0 (71)
Using the short-hand notation:

,,

, A,
_
i ,
mc

_
= 0 (72)
From the properties of the
j
and matrices, we can derive

=
0
, (hermitian) (73)

= (
j
)

=
j

=
j
=
j
=
j
, (anti-hermitian)(74)

=
0

0
, (75)

= 2g

I. (Cliord algebra). (76)

Conjugate of the Dirac equation is given by
i

mc

= 0
i

mc

= 0 (77)
We will dene the Dirac adjoint spinor by

0
. Then
i

+
mc

= 0. (78)
The four-current is
j

c
=

=
_
,
j
c
_
,

= 0. (79)
10
Properties of the

matrices
We may form new matrices by multiplying matrices together. Because dierent
matrices anticommute, we only need to consider products of dierent s and
the order is not important. We can combine them in 2
4
1 ways. Plus the identity
we have 16 dierent matrices,
I

0
, i
1
, i
2
, i
3

1
,
0

2
,
0

3
, i
2

3
, i
3

1
, i
1

2
i
0

3
, i
0

1
, i
0

2
,
1

3
i
0

3

5
(=
5
). (80)
Denoting them by
l
, l = 1, 2, , 16, we can derive the following relations.
(a)
l

m
= a
lm

n
, a
lm
= 1 or i.
(b)
l

m
= I if and only if l = m.
(c)
l

m
=
m

l
.
(d) If
l
,= I, there always exists a
k
, such that
k

k
=
l
.
(e) Tr(
l
) = 0 for
l
,= I.
Proof:
Tr(
l
) = Tr(
k

k
) = Tr(
l

k
) = Tr(
l
)
.
(f)
l
are linearly independent:

16
k=1
x
k

k
= 0 only if x
k
= 0, k = 1, 2, , 16.
Proof:
_
16

k=1
x
k

k
_

m
= x
m
I +

k=m
x
k

m
= x
m
I +

k=m
x
k
a
km

n
= 0 (
n
,= I)
. Taking the trace, x
m
Tr(I) =

k=m
x
k
a
km
Tr(
n
) = 0 x
m
= 0. for any m.
This implies that
k
s cannot be represented by matrices smaller than 4 4. In
fact, the smallest representations of
k
s are 4 4 matrices. (Note that this 4 is
not the dimension of the space-time. the equality is accidental.)
(g) Corollary: any 44 matrix X can be written uniquely as a linear combination
of the
k
s.
X =
16

k=1
x
k

k
Tr(X
m
) = x
m
Tr(
m

m
) +

k=m
x
k
Tr(
k

m
) = x
m
Tr(I) = 4x
m
x
m
=
1
4
Tr(x
m
)
11
(h) Stronger corollary:
l

m
= a
lm

n
where
n
is a dierent
n
for each m, given
a xed l.
Proof: If it were not true and one can nd two dierent
m
,
m
such that
l

m
=
a
lm

n
,
l

m
= a
lm

n
, then we have

m
= a
lm

n
,
m
= a
lm

l

n

m
=
a
lm
a
lm

m
,
which contradicts that
k
s are linearly independent.
(i) Any matrix X that commutes with

(for all ) is a multiple of the identity.

Proof: Assume X is not a multiple of the identity. If X commutes with all

then
it commutes with all
l
s, i.e., X =
l
X
l
. We can express X in terms of the Ga
matrices,
X = x
m

m
+

k=m
x
k

k
,
m
,= I.
There exists a
i
such that
i

i
=
m
. By the hypothesis that X commutes
with this
i
, we have
X = x
m

m
+

k=m
x
k

k
=
i
X
i
= x
m

i
+

k=m
x
k

i
= x
m

m
+

k=m
x
k

k
.
Since the expansion is unique, we must have x
m
= x
m
.
m
was arbitrary except
that
m
,= I. This implies that all x
m
= 0 for
m
,= I and hence X = aI.
(j) Paulis fundamental theorem: Given two sets of 44 matrices

and

which
both satisfy

= 2g

I,
there exists a nonsingular matrix S such that

= S

S
1
.
Proof: F is an arbitrary 4 4 matrix, set
i
is constructed from

and

i
is
constructed from

. Let
S =
16

i=1

i
F
i
.

j
= a
ij

j
= a
2
ij

2
k
= a
2
ij

j
=
j

i
= a
2
ij

j
= a
3
ij

j
= a
ij

k
12
For any i,

i
S
i
=

j
F
j

i
=

j
a
4
ij

k
F
k
=

k
F
k
= S, (a
4
ij
= 1).
It remains only to prove that S is nonsingular.
S

=
16

i=1

i
G

i
, for G arbitrary.
By the same argument, we have S

=
i
S

i
.
S

S =
i
S

i
S
i
=
i
S

S
i
,
S

S commutes with
i
for any i so S

S = aI. We can choose a ,= 0 because F, G

are arbitrary, then S is nonsingular. Also, S is unique up to a constant. Otherwise
if we had S
1

S
1
1
= S
2

S
1
2
, then S
1
2
S
1

S
1
2
S
1
S
1
2
S
1
= aI.
Specic representations of the

matrices
Recall H = (c(i)+mc
2
). In the non-relativistic limit, mc
2
term dominates
the total energy, so its convenient to represent =
0
by a diagonal matrix. Recall
Tr = 0 and
2
= I, so we choose
=
_
I 0
0 I
_
where I =
_
1 0
0 1
_
. (81)

k
s anticommute with and are hermitian,

k
=
_
0 A
k
(A
k
)

0
_
, (82)
A
k
: 2 2 matrices, anticommute with each other. These properties are satised
by the Pauli matrices, so we have

k
=
_
0
k

k
0
_
,
1
=
_
0 1
1 0
_
,
2
=
_
0 i
i 0
_
,
3
=
_
1 0
0 1
_
(83)
From these we obtain

0
= =
_
I 0
0 I
_
,
i
=
i
=
_
0
i

i
0
_
,
5
= i
0

3
=
_
0 I
I 0
_
.
(84)
This is the Pauli-Dirac representation of the

matrices. Its most useful for

system with small kinetic energy, e.g., atomic physics.
Lets consider the simplest possible problem: free particle at rest. is a 4-
component wave-function with each component satisfying the Klein-Gordon equa-
tion,
= e
i

(pxEt)
, (85)
13
where is a 4-component spinor and E
2
= p
2
c
2
+m
2
c
4
.
Free particle at rest: p = 0, is independent of x,
H = (ic +mc
2

0
) = mc
2

0
= E. (86)
In Pauli-Dirac representation,
0
= diag(1, 1, 1, 1), the 4 fundamental solutions
are

1
=
_
_
_
_
1
0
0
0
_
_
_
_
, E = mc
2
,

2
=
_
_
_
_
0
1
0
0
_
_
_
_
, E = mc
2
,

3
=
_
_
_
_
0
0
1
0
_
_
_
_
, E = mc
2
,

4
=
_
_
_
_
0
0
0
1
_
_
_
_
, E = mc
2
.
As we shall see, Dirac wavefunction describes a particle pf spin-1/2.
1
,
2
repre-
sent spin-up and spin-down respectively with E = mc
2
.
3
,
4
represent spin-up
and spin-down respectively with E = mc
2
. As in Klein-Gordon equation, we
have negative solutions and they can not be discarded.
For ultra-relativistic problems (most of this course), the Weyl representation is
more convenient.

PD
=
_
_
_
_

4
_
_
_
_
=
_

B
_
,
A
=
_

2
_
,
b
=
_

4
_
. (87)
In terms of
A
and
B
, the Dirac equation is
i

x
0

A
+i
B
=
mc

A
,
i

x
0

B
i
A
=
mc

B
. (88)
Lets dene

A
=
1

2
(
1
+
2
),
B
=
1

2
(
2

1
) (89)
14
and rewrite the Dirac equation in terms of
1
and
2
,
i

x
0

1
i
1
=
mc

2
,
i

x
0

2
+i
2
=
mc

1
. (90)
On can see that
1
and
2
are coupled only via the mass term. In ultra-relativistic
limit (or for nearly massless particle such as neutrinos), rest mass is negligible,
then
1
and
2
decouple,
i

x
0

1
i
1
= 0,
i

x
0

2
+i
2
= 0, (91)
The 4-component wavefunction in the Weyl representation is written as

Weyl
=
_

2
_
. (92)
Lets imagine that a massless spin-1/2 neutrino is described by
1
, a plane wave
state of a denite momentum p with energy E = [p[c,

1
e
i

(pxEt)
. (93)
i

x
0

1
= i
1
c

1
=
E
c

1
,
i
1
=
1

p
1
E
1
= [p[c
1
= c p
1
or
p
[p[

1
=
1
. (94)
The operator h = p/[p[ is called the helicity. Physically it refers to the
component of spin in the direction of motion.
1
describes a neutrino with helicity
1 (left-handed). Similarly,
p
[p[

2
=
2
, (h = +1, right-handed). (95)
The

s in the Weyl representation are

0
=
_
0 I
I 0
_
,
i
=
_
0
i

i
0
_
,
5
=
_
I 0
0 I
_
. (96)
Exercise: Find the S matrix which transform between the Pauli-Dirac represen-
tation and the Weyl representation and verify that the ga

matrices in the Weyl

representation are correct.
15
5 Lorentz Covariance of the Dirac Equation
We will set = c = 1 from now on.
In E&M, we write down Maxwells equations in a given inertial frame, x, t, with the
electric and magnetic elds E, B. Maxwells equations are covariant with respect
to Lorentz transformations, i.e., in a new Lorentz frame, x

, t

, the equations have

the same form, but the elds E

(x

, t

), B

(x

, t

) are dierent.
Similarly, Dirac equation is Lorentz covariant, but the wavefunction will change
when we make a Lorentz transformation. Consider a frame F with an observer O
and coordinates x

. O describes a particle by the wavefunction (x

) which obeys
_
i

m
_
(x

). (97)
In another inertial frame F

with an observer O

and coordinates x

given by
x

, (98)
O

describes the same particle by

(x

) and

(x

) satises
_
i

m
_

(x

). (99)
Lorentz covariance of the Dirac equation means that the matrices are the same
in both frames.
What is the transformation matrix S which takes to

under the Lorentz trans-

formation?

(x) = S(x). (100)

Applying S to Eq. (97),
iS

S
1

x

S(x

) mS(x

) = 0
iS

S
1

x

(x

) m

(x

) = 0. (101)
Using

=

x

, (102)
we obtain
iS

S
1

(x

) m

(x

) = 0. (103)
Comparing it with Eq. (99), we need
S

S
1

or equivalently S

S
1
=
_

1
_

. (104)
16
We will write down the form of the S matrix without proof. You are encouraged
to read the derivation in Shultens notes Chapter 10, p.319-321 and verify it by
yourself.
For an innitesimal Lorentz transformation,

. Multiplied by g

it
can be written as

= g

, (105)
where

is antisymmetric in and . Then the corresponding Lorentz transfor-

mation on the spinor wavefunction is given by
S(

) = I
i
4

, (106)
where

=
i
2
(

) =
i
2
[

]. (107)
For nite Lorentz transformation,
S = exp
_

i
4

_
. (108)
Note that one can use either the active transformation (which transforms the ob-
ject) or the passive transformation (which transforms the coordinates), but care
should be taken to maintain consistency. We will mostly use passive transforma-
tions unless explicitly noted otherwise.
Example: Rotation about z-axis by angle (passive).

12
= +
21
= , (109)

12
=
i
2
[
1
,
2
] = i
1

2
= i
_
i
3
0
0 i
3
_
=
_

3
0
0
3
_

3
(110)
S = exp
_
+
i
2

3
0
0
3
__
= I cos

2
+i
_

3
0
0
3
_
sin

2
. (111)
We can see that transforms under rotations like an spin-1/2 object. For a
rotation around a general direction n,
S = I cos

2
+i n sin

2
. (112)
17
Example: Boost in x direction (passive).

01
=
10
= , (113)

01
=
i
2
[
0
,
1
] = i
0

1
, (114)
S = exp
_

i
4

_
= exp
_

i
2
i
0

1
_
= exp
_

1
_
= exp
_

1
_
= I cosh

2

1
sinh

2
. (115)
For a particle moving in the direction of n in the new frame, we need to boost the
frame in the n direction,
S = I cosh

2
+ nsinh

2
. (116)
6 Free Particle Solutions to the Dirac Equation
The solutions to the Dirac equation for a free particle at rest are

1
=
_
2m
V
_
_
_
_
1
0
0
0
_
_
_
_
e
imt
, E = +m,

2
=
_
2m
V
_
_
_
_
0
1
0
0
_
_
_
_
e
imt
, E = +m,

3
=
_
2m
V
_
_
_
_
0
0
1
0
_
_
_
_
e
imt
, E = m,

4
=
_
2m
V
_
_
_
_
0
0
0
1
_
_
_
_
e
imt
, E = m, (117)
where we have set = c = 1 and V is the total volume. Note that I have chosen
a particular normalization
_
d
3
x

= 2m (118)
18
for a particle at rest. This is more convenient when we learn eld theory later,
because

is not invariant under boosts. Instead, its the zeroth component of

a 4-vector, similar to E.
The solutions for a free particle moving at a constant velocity can be obtained by
a Lorentz boost,
S = I cosh

2
+ nsinh

2
. (119)
Using the Pauli-Dirac representation,

i
=
_
0
i

i
0
_
,
S =
_
cosh

2
nsinh

2
nsinh

2
cosh

2
_
, (120)
and the following relations,
cosh =

=
E

m
, sinh =

,
cosh

2
=
_
1 + cosh
2
=
_
1 +

2
=
_
m +E

2m
,
sinh

2
=
_
cosh 1
2
=
_
E

m
2m
p

= p

+
x

= p

= mt, (121)
where p

+
= [p

+
[ n is the 3-momentum of the positive energy state, we obtain

1
(x

) = S
1
(x) =
_
2m
V
_
_
_
_
cosh

2
_
1
0
_
nsinh

2
_
1
0
_
_
_
_
_
e
imt
=
1

m +E

_
_
_
_
_
1
0
_
_
E

m
E

+m
n
_
1
0
_
_
_
_
_
e
i(p

+
x

)
=
1

m +E

_
_
_
_
_
1
0
_
p

+
E

+m
_
1
0
_
_
_
_
_
e
i(p

+
x

)
(122)
where we have used
_
E

m
E

+m
=

E
2
m
2
E

+m
=
[p

+
[
E

+m
(123)
in the last line.
19

2
(x

) has the same form except that

_
1
0
_
is replaced by
_
0
1
_
.
For the negative energy solutions E

= E

=
_
[p[
2
+m
2
and p

= v nE

=
vE

n = p

+
. So we have

3
(x

) =
1

m +E

_
_
_
_

+m
_
1
0
_
_
1
0
_
_
_
_
_
e
i(p

+E

)
, (124)
and

4
(x

) is obtained by the replacement

_
1
0
_

_
0
1
_
.
Now we can drop the primes and the subscripts,

1,2
=
1

E +m
_

+,
p
E+m

+,
_
e
i(pxEt)
=
1

V
u
1,2
e
i(pxEt)
,

3,4
=
1

E +m
_

p
E+m

+,

+,
_
e
i(px+Et)
=
1

V
u
3,4
e
i(px+Et)
, (125)
where

+
=
_
1
0
_
,

=
_
0
1
_
(126)
(V is the proper volume in the frame where the particle is at rest.)
Properties of spinors u
1
, u
4
u

r
u
s
= 0 for r ,= s. (127)
u

1
u
1
= (E +m)
_

+
p
E+m
_
_

+
p
E+m

+
_
= (E +m)

+
_
1 +
( p)( p)
(E +m)
2
_

+
. (128)
Using the following identity:
( a)( b) =
i
a
i

j
b
j
= (
ij
+i
ijk

k
)a
i
b
j
= a b +i (a b), (129)
we have
u

1
u
1
= (E +m)

+
_
1 +
[p[
2
(E +m)
2
_

+
= (E +m)

+
E
2
+ 2Em +m
2
+[p[
2
(E +m)
2

+
=
+
2E
2
+ 2Em
E +m

+
= 2E

+
= 2E. (130)
20
Similarly for other u
r
we have u

r
u
s
=
rs
2E, which reects that =

is the
zeroth component of a 4-vector.
One can also check that
u
r
u
s
= 2m
rs
(131)
where + for r = 1, 2 and for r = 3, 4.
u
1
u
1
= u

0
u
1

0
=
_
I 0
0 I
_
= (E +m)

+
_
1
[p[
2
(E +m)
2
_

+
= (E +m)

+
E
2
+ 2Em +m
2
[p[
2
(E +m)
2

+
=
+
2m
2
+ 2Em
E +m

+
= 2m

+
= 2m (132)
is invariant under Lorentz transformation.
Orbital angular momentum and spin
Orbital angular momentum
L = r p or
L
i
=
ijk
r
j
p
k
. (133)
(We dont distinguish upper and lower indices when dealing with space dimensions
only.)
dL
i
dt
= i[H, L
i
]
= i[c p +mc
2
, L
i
]
= ic
n
[p
n
,
ijk
r
j
p
k
]
= ic
n

ijk
[p
n
, r
j
]p
k
= ic
n

ijk
(i
nj
)p
k
= c
ijk

j
p
k
= c(p)
i
,= 0. (134)
We nd that the orbital angular momentum of a free particle is not a constant of
the motion.
21
Consider the spin
1
2
=
1
2
_

i
0
0
i
_
,
d
i
dt
= i[H,
i
]
= i[c
j
p
j
+mc
2
,
i
]
= ic[
j
,
i
]p
j
_
using
i

5
=
_

i
0
0
i
__
0 I
I 0
_
=
_
0
i

i
0
_
=
i
=
5

i
_
= ic[
5

j
,
i
]p
j
= ic
5
[
j
,
i
]p
j
= ic
5
(2i
ijk

k
)p
j
= 2c
5

ijk

k
p
j
= 2c
ijk

k
p
j
= 2c(p)
i
. (135)
Comparing it with Eq. (134), we nd
d(L
i
+
1
2

i
)
dt
= 0, (136)
so the total angular momentum J = L+
1
2
is conserved.
7 Interactions of a Relativistic Electron with an
External Electromagnetic Field
We make the usual replacement in the presence of external potential:
E E e = i

t
e, e < 0 for electron
p p
e
c
A = i
e
c
A. (137)
In covariant form,

+
ie
c
A

+ieA

= c = 1. (138)
Dirac equation in external potential:
i

+ieA

) m = 0. (139)
Two component reduction of Dirac equation in Pauli-Dirac basis:
_
I 0
0 I
_
(E e)
_

B
_

_
0
0
_
(p eA)
_

B
_
m
_

B
_
= 0,
(E e)
A
(p eA)
B
m
A
= 0 (140)
(E e)
B
+ (p eA)
A
m
B
= 0 (141)
22
where E and p represent the operators i
t
and i respectively. Dene W =
E m, = p eA, then we have

B
= (W e)
A
(142)

A
= (2m +W e)
B
(143)
From Eq. (143),

B
= (2m +W e)
1

A
. (144)
Substitute it into Eq. (142),
( )( )
2m +W e

A
= (W e)
A
. (145)
In non-relativistic limit, W e m,
1
2m +W e
=
1
2m
_
1
W e
2m
+
_
. (146)
In the lowest order approximation we can keep only the leading term
1
2m
,
1
2m
( )( )
A
(W e)
A
. (147)
Using Eq. (129),
( )( )
A
= [ +i ( )]
A
. (148)
( )
A
= [(p eA) (p eA)]
A
= [eAp ep A]
A
= [+ieA+ieA]
A
= ie
A
(A)
= ieB
A
, (149)
so
1
2m
(p eA)
2

e
2m
B
A
+e
A
= W
A
. (150)
Restoring , c,
1
2m
(p
e
c
A)
2

e
2mc
B
A
+e
A
= W
A
. (151)
This is the Pauli-Schrodinger equation for a particle with the spin-magnetic
moment,
=
e
2mc
= 2
e
2mc
S. (152)
23
In comparison, the relation between the angular momentum and the magnetic
moment of a classical charged object is given by
=
Ir
2
c
= e

2
r
2
c
=
er
2
2c
=
e
2mc
mr
2
=
e
2mc
L. (153)
We can write
= g
s
e
2mc
S (154)
in general. In Dirac theory, g
s
= 2. Experimentally,
g
s
(e

) = 2 (1.0011596521859 38 10
13
). (155)
The deviation from 2 is due to radiative corrections in QED, (g 2)/2 =

2
+ .
The predicted value for g
s
2 using from the quantum Hall eect is
(g
s
2)
qH
/2 = 0.0011596521564 229 10
13
. (156)
They agree down to the 10
11
level.
There are also spin-1/2 particles with anomalous magnetic moments, e.g.,

proton
= 2.79
[e[
2m
p
c
,
neutron
= 1.91
[e[
2m
n
c
. (157)
This can be described by adding the Pauli moment term to the Dirac equation,
i

+iqA

) m +k

= 0. (158)
Recall

=
i
2
(

),

0i
= i
0

i
= i
_
I 0
0 I
__
0
i

i
0
_
= i
_
0
i

i
0
_
= i
i
,

ij
= i
i

j
=
ijk

k
=
ijk
_

k
0
0
k
_
,
F
0i
= E
i
,
F
ij
=
ijk
B
k
. (159)
Then the Pauli moment term can be written as
i

+iqA

) m + 2ik E 2k B = 0. (160)
The two component reduction gives
(E q)
A

B
m
A
+ 2ik E
B
2k B
A
= 0, (161)
(E q)
B
+
A
m
B
+ 2ik E
A
2k B
B
= 0. (162)
24
( 2ik E)
B
= (W q 2k B)
A
, (163)
( + 2ik E)
A
= (2m +W q + 2k B)
B
. (164)
Again taking the non-relativistic limit,

B

1
2m
( + 2ik E)
A
, (165)
we obtain
(W q 2k B)
A
=
1
2m
( 2ik E)( + 2ik E)
A
. (166)
Lets consider two special cases.
(a) = 0, E = 0
(W 2k B)
A
=
1
2m
( )
2

A
W
A
=
1
2m

q
2m
B
A
+ 2k B
A
=
q
2m
2k. (167)
(b) B = 0, E ,= 0 for the neutron (q = 0)
W
A
=
1
2m
(p +i
n
E) (p i
n
E)
A
=
1
2m
[(p +i
n
E) (p +i
n
E) +i (p +i
n
E) (p i
n
E)]
A
=
1
2m
_
p
2
+
2
n
E
2
+i
n
E p i
n
p E +i (i
n
p E i
n
E p)

A
=
1
2m
_
p
2
+
2
n
E
2

n
( E) + 2
n
(E p) +i
n
(E)

A
=
1
2m
_
p
2
+
2
n
E
2

n
+ 2
n
(E p)

A
. (168)
The last term is the spin-orbit interaction,
(E p) =
1
r
d
dr
(r p) =
1
r
d
dr
L. (169)
The second to last term gives an eective potential for a slow neutron moving in
the electric eld of an electron,
V =

2m
=

n
2m
(e)
3
(r). (170)
Its called Foldy potential and does exist experimentally.
25
8 Foldy-Wouthuysen Transformation
We now have the Dirac equation with interactions. For a given problem we can
solve for the spectrum and wavefunctions (ignoring the negative energy solutions
for a moment), for instance, the hydrogen atom, We can compare the solutions
to those of the schrodinger equation and nd out the relativistic corrections to
the spectrum and the wavefunctions. In fact, the problem of hydrogen atom can
be solved exactly. However, the exact solutions are problem-specic and involve
unfamiliar special functions, hence they not very illuminating. You can nd the
exact solutions in many textbooks and also in Shultens notes. Instead, in this
section we will develop a systematic approximation method to solve a system in
the non-relativistic regime (Em m). It corresponds to take the approximation
we discussed in the previous section to higher orders in a systematic way. This
allows a physical interpretation for each term in the approximation and tells us the
relative importance of various eects. Such a method has more general applications
for dierent problems.
In Foldy-Wouthuysen transformation, we look for a unitary transformation U
F
removes operators which couple the large to the small components.
Odd operators (o-diagonal in Pauli-Dirac basis):
i
,
i
,
5
,
Even operators (diagonal in Pauli-Dirac basis): 1, , ,

= U
F
= e
iS
, S = hermitian (171)
First consider the case of a free particle, H = p +m not time-dependent.
i

t
= e
iS
H = e
iS
He
iS

= H

(172)
We want to nd S such that H

contains no odd operators. We can try

e
iS
= e
p
= cos + psin , where p = p/[p[. (173)
H

= (cos + psin ) ( p +m) (cos psin )

= ( p +m) (cos psin )
2
= ( p +m) exp (2 p)
= ( p)
_
cos 2
m
[p[
sin 2
_
+ (mcos 2 +[p[ sin 2) . (174)
To eliminate ( p) term we choose tan 2 = [p[/m, then
H

=
_
m
2
+[p[
2
. (175)
This is the same as the rst Hamiltonian we tried except for the factor which
also gives rise to negative energy solutions. In practice, we need to expand the
Hamiltonian for [p[ m.
26
General case:
H = (p eA) +m +e
= m +O +c, (176)
O = (p eA), c = e, O = O, c = c (177)
H time-dependent S time-dependent
We can only construct S with a non-relativistic expansion of the transformed
Hamiltonian H

in a power series in 1/m.

Well expand to
p
4
m
3
and
p(E, B)
m
2
.
H = i

t
_
e
iS

_
= e
iS
i

t
+
_
i

t
e
iS
_

t
=
_
e
iS
_
H i

t
_
e
iS
_

= H

(178)
S is expanded in powers of 1/m and is small in the non-relativistic limit.
e
iS
He
iS
= H +i[S, H] +
i
2
2!
[S, [S, H]] + +
i
n
n!
[S, [S, [S, H]]]. (179)
S = O(
1
m
) to the desired order of accuracy
H

= H +i[S, H]
1
2
[S, [S, H]]
i
6
[S, [S, [S, H]]] +
1
24
[S, [S, [S, [S, m]]]]

S
i
2
[S,

S] +
1
6
[S, [S,

S]] (180)
We will eliminate the odd operators order by order in 1/m and repeat until the
desired order is reached.
First order [O(1)]:
H

= m +c +O +i[S, ]m. (181)

To cancel O, we choose S =
iO
2m
,
i[S, H] = O +

2m
[O, c] +
1
m
O
2
(182)
i
2
2
[S, [S, H]] =
O
2
2m

1
8m
2
[O, [O, c]]
1
2m
2
O
3
(183)
i
3
3!
[S, [S, [S, H]]] =
O
3
6m
2

1
6m
3
O
4
(184)
i
4
4!
[S, [S, [S, [S, H]]]] =
O
4
24m
3
(185)

S =
i

O
2m
(186)

i
2
[S,

S] =
i
8m
2
[O,

O] (187)
27
Collecting everything,
H

=
_
m +
O
2
2m

O
4
8m
3
_
+c
1
8m
2
[O, [O, c]]
i
8m
2
[O,

O] (188)
+

2m
[O, c]
O
3
3m
2
+
i

O
2m
= m +c

+O

(189)
Now O

is O(
1
m
), we can transform H

by S

to cancel O

,
S

=
i
2m
O

=
i
2m
_

2m
[O, c]
O
3
3m
2
+
i

O
2m
_
(190)
After transformation with S

,
H

= e
iS

_
H

i

t
_
e
iS

= m +c

+

2m
[O

, c

] +
i

O

2m
(191)
= m +c

+O

, (192)
where O

is O(
1
m
2
), which can be cancelled by a third transformation, S

=
iO

2m
H

= e
iS

_
H

i

t
_
e
iS

= m +c

(193)
=
_
m +
O
2
2m

O
4
8m
3
_
+c
1
8m
2
[O, [O, c]]
i
8m
2
[O,

O] (194)
Evaluating the operator products to the desired order of accuracy,
O
2
2m
=
( (p eA))
2
2m
=
(p eA)
2
2m

e
2m
B (195)
1
8m
2
_
[O, c] +i

O
_
=
e
8m
2
(i i

A) =
ie
8m
2
E (196)
_
O,
ie
8m
2
E
_
=
ie
8m
2
[ p, E]
=
ie
8m
2

i,j

j
_
i
E
j
x
i
_
+
e
4m
2
E p (197)
=
e
8m
2
( E) +
ie
8m
2
(E) +
e
4m
2
E p
So, the eective Hamiltonian to the desired order is
H

=
_
m +
(p eA)
2
2m

p
4
8m
3
_
+e
e
2m
B

ie
8m
2
(E)
e
4m
2
E p
e
8m
2
( E) (198)
28
The individual terms have a direct physical interpretation.
The rst term in the parentheses is the expansion of
_
(p eA)
2
+m
2
(199)
and p
4
/(8m
3
) is the leading relativistic corrections to the kinetic energy.
The two terms

ie
8m
2
(E)
e
4m
2
E p (200)
together are the spin-orbit energy. In a spherically symmetric static potential, they
take a very familiar form. In this case E = 0,
E p =
1
r

r
r p =
1
r

r
L, (201)
and this term reduces to
H
spinorbit
=
e
4m
2
1
r

r
L. (202)
The last term is known as the Darwin term. In a coulomb potential of a nucleus
with charge Z[e[, it takes the form

e
8m
2
( E) =
e
8m
2
Z[e[
3
(r) =
Ze
2
8m
2

3
(r) =
Z
2m
2

3
(r), (203)
so it can only aect the S (l = 0) states whose wavefunctions are nonzero at the
origin.
For a Hydrogen-like (single electron) atom,
e =
Ze
2
4r
, A = 0. (204)
The shifts in energies of various states due to these correction terms can be com-
puted by taking the expectation values of these terms with the corresponding
wavefunctions.
Darwin term (only for S (l = 0) states):
_

ns

Z
2m
2

3
(r)

ns
_
=
Z
2m
2
[
ns
(0)[
2
=
Z
4

4
m
2n
3
. (205)
Spin-orbit term (nonzero only for l ,= 0):
_
Z
4m
2
1
r
3
r p
_
=
Z
4

4
m
4n
3
[j(j + 1) l(l + 1) s(s + 1)]
l(l + 1)(l +
1
2
)
. (206)
29
Relativistic corrections:
_

p
4
8m
3
_
=
Z
4

4
m
2n
4
_
3
4

n
l +
1
2
_
. (207)
We nd
E(l = 0) =
Z
4

4
m
2n
4
_
3
4
n
_
(208)
= E(l = 1, j =
1
2
), (209)
so 2S
1/2
and 2P
1/2
remain degenerate at this level. They are split by Lamb shift
(2S
1/2
> 2P
1/2
) which can be calculated after you learn radiative corrections in
QED. The 2P
1/2
and 2P
3/2
are split by the spin-orbit interaction (ne structure)
which you should have seen before.
E(l = 1, j =
3
2
) E(l = 1, j =
1
2
) =
Z
4

4
m
4n
3
(210)
9 Klein Paradox and the Hole Theory
So far we have ignored the negative solutions. However, the negative energy solu-
tions are required together with the positive energy solutions to form a complete
set. If we try to localize an electron by forming a wave packet, the wavefunc-
tion will be composed of some negative energy components. There will be more
negative energy components if the electron is more localized by the uncertainty
relation xp . The negative energy components can not be ignored if the
electron is localized to distances comparable to its Compton wavelength /mc,
and we will encounter many paradoxes and dilemmas. An example is the Klein
paradox described below.
In order to localize electrons, we must introduce strong external forces conning
them to the desired region. Lets consider a simplied situation that we want to
conne a free electron of energy E to the region z < 0 by a one-dimensional step-
function potential of height V as shown in Fig. 1. Now in the z < 0 half space
there is an incident positive energy plan wave of momentum k > 0 along the z
axis,

inc
(z) = e
ikz
_
_
_
_
1
0
k
E+m
0
_
_
_
_
, (spin-up). (211)
30
Figure 1: Electrostatic potential idealized with a sharp boundary, with an incident
free electron wave moving to the right in region I.
The reected wave in z < 0 region has the form

ref
(z) = a e
ikz
_
_
_
_
1
0
k
E+m
0
_
_
_
_
+b e
ikz
_
_
_
_
0
1
0
k
E+m
_
_
_
_
, (212)
and the transmitted wave in the z > 0 region (in the presence of the constant
potential V ) has a similar form

trans
(z) = c e
iqz
_
_
_
_
1
0
q
EV +m
0
_
_
_
_
+d e
iqz
_
_
_
_
0
1
0
q
EV +m
_
_
_
_
, (213)
with an eective momentum q of
q =
_
(E V )
2
m
2
. (214)
The total wavefunction is
(z) = (z)[
inc
(z) +
ref
(z)] +(z)
trans
(z). (215)
Requiring the continuity of (z) at z = 0,
inc
(0) +
ref
(0) =
trans
(0), we obtain
1 +a = c (216)
b = d (217)
(1 a)
k
E +m
= c
q
E V +m
(218)
b
k
E +m
= d
q
E V +m
(219)
31
From these equations we can see
b = d = 0 (no spin-ip) (220)
1 +a = c (221)
1 a = rc where r =
q
k
E +m
E V +m
(222)
c =
2
1 +r
, a =
1 r
1 +r
. (223)
As long as [EV [ < m, q is imaginary and the transmitted wave decays exponen-
tially. However, when V E+m the transmitted wave becomes oscillatory again.
The probability currents j =

3
z, for the incident, transmitted, and
reected waves are
j
inc
= 2
k
E +M
,
j
trans
= 2c
2
q
E V +m
,
j
ref
= 2a
2
k
E +m
. (224)
we nd
j
trans
j
inc
= c
2
r =
4r
(1 +r)
2
(< 0 for V E +m),
j
ref
j
inc
= a
2
=
_
1 r
1 +r
_
2
(> 1 for V E +m). (225)
Although the conservation of the probabilities looks satised: j
inc
= j
trans
+ j
ref
,
but we get the paradox that the reected ux is larger than the incident one!
There is also a problem of causality violation of the single particle theory which
you can read in Prof. Gunions notes, p.14p.15.
Hole Theory
In spite of the success of the Dirac equation, we must face the diculties from
the negative energy solutions. By their very existence they require a massive
reinterpretation of the Dirac theory in order to prevent atomic electrons from
making radiative transitions into negative-energy states. The transition rate for
an electron in the ground state of a hydrogen atom to fall into a negative-energy
state may be calculated by applying semi-classical radiation theory. The rate for
the electron to make a transition into the energy interval mc
2
to 2mc
2
is

2
6

mc
2

10
8
sec
1
(226)
and it blows up if all the negative-energy states are included, which clearly makes
no sense.
32
A solution was proposed by Dirac as early as 1930 in terms of a many-particle
theory. (This shall not be the nal standpoint as it does not apply to scalar
particle, for instance.) He assumed that all negative energy levels are lled up in
the vacuum state. According to the Pauli exclusion principle, this prevents any
electron from falling into these negative energy states, and thereby insures the
stability of positive energy physical states. In turn, an electron of the negative
energy sea may be excited to a positive energy state. It then leaves a hole in
the sea. This hole in the negative energy, negatively charged states appears as a
positive energy positively charged particlethe positron. Besides the properties
of the positron, its charge [e[ = e > 0 and its rest mass m
e
, this theory also
predicts new observable phenomena:
The annihilation of an electron-positron pair. A positive energy electron falls
into a hole in the negative energy sea with the emission of radiation. From energy
momentum conservation at least two photons are emitted, unless a nucleus is
present to absorb energy and momentum.
Conversely, an electron-positron pair may be created from the vacuum by an
incident photon beam in the presence of a target to balance energy and momentum.
This is the process mentioned above: a hole is created while the excited electron
acquires a positive energy.
Thus the theory predicts the existence of positrons which were in fact observed
in 1932. Since positrons and electrons may annihilate, we must abandon the
interpretation of the Dirac equation as a wave equation. Also, the reason for
discarding the Klein-Gordon equation no longer hold and it actually describes
spin-0 particles, such as pions. However, the hole interpretation is not satisfactory
for bosons, since there is no Pauli exclusion principle for bosons.
Even for fermions, the concept of an innitely charged unobservable sea looks
rather queer. We have instead to construct a true many-body theory to accom-
modate particles and antiparticles in a consistent way. This is achieved in the
quantum theory of elds which will be the subject of the rest of this course.
33