Chapter 10

Relativistic Quantum Mechanics

In this Chapter we will address the issue that the laws of physics must be formulated in a form
which is Lorentz–invariant, i.e., the description should not allow one to differentiate between frames
of reference which are moving relative to each other with a constant uniform velocity ~v . The
transformations beween such frames according to the Theory of Special Relativity are described
by Lorentz transformations. In case that ~v is oriented along the x1 –axis, i.e., ~v = v1 x̂1 , these
transformations are
x1 − v1 t 0 t − vc21 x1 0 0
x1 0 = q , t = q
, x 2 = x2 ; x 3 = x3 (10.1)
v1 2 v1 2


1 − c 1 − c

which connect space time coordinates (x1 , x2 , x3 , t) in one frame with space time coordinates
(x01 , x02 , x03 , t0 ) in another frame. Here c denotes the velocity of light. We will introduce below
Lorentz-invariant differential equations which take the place of the Schrödinger equation of a par-
ticle of mass m and charge q in an electromagnetic field [c.f. (refeq:ham2, 8.45)] described by an
electrical potential V (~r, t) and a vector potential A(~~ r, t)
  2 
∂ 1 ~ q~
i~ ψ(~r, t) = ∇ − A(~r, t) + qV (~r, t) ψ(~r, t) (10.2)
∂t 2m i c
The replacement of (10.2) by Lorentz–invariant equations will have two surprising and extremely
important consequences: some of the equations need to be formulated in a representation for which
the wave functions ψ(~r, t) are vectors of dimension larger one, the components representing the
spin attribute of particles and also representing together with a particle its anti-particle. We will
find that actually several Lorentz–invariant equations which replace (10.2) will result, any of these
equations being specific for certain classes of particles, e.g., spin–0 particles, spin– 12 particles, etc.
As mentioned, some of the equations describe a particle together with its anti-particle. It is not
possible to uncouple the equations to describe only a single type particle without affecting nega-
tively the Lorentz invariance of the equations. Furthermore, the equations need to be interpreted
as actually describing many–particle-systems: the equivalence of mass and energy in relativistic
formulations of physics allows that energy converts into particles such that any particle described
will have ‘companions’ which assume at least a virtual existence.
Obviously, it will be necessary to begin this Chapter with an investigation of the group of Lorentz
transformations and their representation in the space of position ~r and time t. The representation

287
288 Relativistic Quantum Mechanics

in Sect. 10.1 will be extended in Sect. 10.4 to cover fields, i.e., wave functions ψ(~r, t) and vectors
with functions ψ(~r, t) as components. This will provide us with a general set of Lorentz–invariant
equations which for various particles take the place of the Schrödinger equation. Before introduc-
ing these general Lorentz–invariant field equations we will provide in Sects. 10.5, 10.7 a heuristic
derivation of the two most widely used and best known Lorentz–invariant field equations, namely
the Klein–Gordon (Sect. 10.5) and the Dirac (Sect. 10.7) equation.

10.1 Natural Representation of the Lorentz Group

In this Section we consider the natural representation of the Lorentz group L, i.e. the group of
Lorentz transformations (10.1). Rather than starting from (10.1), however, we will provide a more
basic definition of the transformations. We will find that this definition will lead us back to the
transformation law (10.1), but in a setting of representation theory methods as applied in Secti. 5
to the groups SO(3) and SU(2) of rotation transformations of space coordinates and of spin.
The elements L ∈ L act on 4–dimensional vectors of position– and time–coordinates. We will
denote these vectors as follows
def
xµ = (x0 , x1 , x2 , x3 ) (10.3)

where x0 = ct describes the time coordinate and (x1 , x2 , x3 ) = ~r describes the space coordinates.
Note that the components of xµ all have the same dimension, namely that of length. We will,
henceforth, assume new units for time such that the velocity of light c becomes c = 1. This choice
implies dim(time) = dim(length).

Minkowski Space

Historically, the Lorentz transformations were formulated in a space in which the time component
of xµ was chosen as a purely imaginary number and the space components real. This space is
called the Minkowski space. The reason for this choice is that the transformations (10.1) leave the
quantity
s2 = (x0 )2 − (x1 )2 − (x2 )2 − (x3 )2 (10.4)

invariant, i.e., for the transformed space-time–cordinates x0 µ = (x0 0 , x0 1 , x0 2 , x0 3 ) holds

0 1 2 3
(x0 )2 − (x1 )2 − (x2 )2 − (x3 )2 = (x0 )2 − (x0 )2 − (x0 )2 − (x0 )2 . (10.5)
√
One can interprete the quantity −s2 as a distance in a 4–dimensional Euclidean space if one
chooses the time component purely imaginary. In such a space Lorentz transformations corre-
spond to 4-dimensional rotations. Rather than following this avenue we will introduce Lorentz
transformations within a setting which does not require real and imaginary coordinates.

The Group of Lorentz Transformations L = O(3,1)

The Lorentz transformations L describe the relationship between space-time coordinates xµ of two
reference frames which move relative to each other with uniform fixed velocity ~v and which might
be reoriented relative to each other by a rotation around a common origin. Denoting by xµ the
10.1: Natural Representation of the Lorentz Group 289

coordinates in one reference frame and by x0 µ the coordinates in the other reference frame, the
Lorentz transformations constitute a linear transformation which we denote by
3
0µ
X
x = Lµ ν xν . (10.6)
ν=0

Here Lµ ν are the elements of a 4 × 4–matrix representing the Lorentz transformation. The upper
index closer to ‘L’ denotes the first index of the matrix and the lower index ν further away from
‘L’ denotes the second index. [ A more conventional notation would be Lµν , however, the latter
notation will be used for different quantities further below.] The following possibilities exist for the
positioning of the indices µ, ν = 0, 1, 2, 3:

4-vector: xµ , xµ ; 4 × 4 tensor: Aµ ν , Aµ ν , Aµν , Aµν . (10.7)

The reason for the notation is two-fold. First, the notation in (10.6) allows us to introduce the
so-called summation conventon: any time the same index appears in an upper and a lower position,
summation over that index is assumed without explicitly noting it, i.e.,

X3 X 3 X 3
yµ x µ = yµ xµ ; Aµ ν xν = Aµ ν xν ; Aµ ν B ν ρ = Aµ ν B ν ρ . (10.8)
| {z } | {z } | {z }
µ=0
new | {z } new |ν=0 {z } new |ν=0 {z }
old old old

The summation convention allows us to write (10.6) x0 µ = Lµ ν xν . The second reason is that upper
and lower positions allow us to accomodate the expression (10.5) into scalar products. This will be
explained further below.
The Lorentz transformations are non-singular 4 × 4–matrices with real coefficients, i.e., L ∈ GL(4,
R), the latter set constituting a group. The Lorentz transformations form the subgroup of all
matrices which leave the expression (10.5) invariant. This condition can be written
µ ν
xµ gµν xν = x0 gµν x0 (10.9)

where  
1 0 0
 0 −1 0 0 
( gµν ) = 
 0 0 −1 0  = g .
 (10.10)
0 0 0 −1
Combining condition (10.9) and (10.6) yields

Lµ ρ gµν Lν σ xρ xσ = gρσ xρ xσ . (10.11)

Since this holds for any xµ it must be true

Lµ ρ gµν Lν σ = gρσ . (10.12)

This condition specifies the key property of Lorentz transformations. We will exploit this property
below to determine the general form of the Lorentz transformations. The subset of GL(4, R), the
290 Relativistic Quantum Mechanics

elements of which satisfy this condition, is called O(3,1). This set is identical with the set of all
Lorentz transformations L. We want to show now L = O(3,1) ⊂ GL(4, R) is a group.
To simplify the following proof of the key group properties we like to adopt the conventional matrix
notation for Lµ ν  0
L 0 L0 1 L0 2 L0 3

 L1 0 L1 1 L1 2 L1 3 
L = ( Lµ ν ) =   L2 0 L2 1 L2 2 L3 3  .
 (10.13)
L3 0 L3 1 L3 2 L3 3
Using the definition (10.10) of g one can rewrite the invariance property (10.12)

LT gL = g . (10.14)

From this one can obtain using

g2 = 11 (10.15)
(gLT g)L = 11 and, hence, the inverse of L

L0 0 −L1 0 −L2 0 −L3 0

 
 −L0 1 L1 1 L2 1 L3 1 
L−1 = g LT g =  0 1 2
 . (10.16)
 −L 2 L 2 L 2 L3 2 
−L0 3 L1 3 L2 3 L3 3

The corresponding expression for (LT )−1 is obviously

(LT )−1 = (L−1 )T = g L g . (10.17)

To demonstrate the group property of O(3,1), i.e., of

O(3, 1) = { L, L ∈ GL(4, R), LT gL = g } , (10.18)

we note first that the identity matrix 11 is an element of O(3,1) since it satisfies (10.14). We consider
then L1 , L2 ∈ O(3,1). For L3 = L1 L2 holds

LT3 g L3 = LT2 LT1 g L1 L2 = LT2 (LT1 gL1 ) L2 = LT2 g L2 = g , (10.19)

i.e., L3 ∈ O(3,1). One can also show that if L ∈ O(3,1), the associated inverse obeys (10.14), i.e.,
L−1 ∈ O(3,1). In fact, employing expressions (10.16, 10.17) one obtains

(L−1 )T g L−1 = gLgggLT g = gLgLT g . (10.20)

Multiplying (10.14) from the right by gLT and using (10.15) one can derive LT gLgLT = LT and
multiplying this from the left by by g(LT )−1 yields

L g LT = g (10.21)

Using this result to simplify the r.h.s. of (10.20) results in the desired property

(L−1 )T g L−1 = g , (10.22)

i.e., property (10.14) holds for the inverse of L. This stipulates that O(3,1) is, in fact, a group.
10.1: Natural Representation of the Lorentz Group 291

Classification of Lorentz Transformations

We like to classify now the elements of L = O(3,1). For this purpose we consider first the value
of det L. A statement on this value can be made on account of property (10.14). Using det AB =
det A det B and det AT = det A yields (det L)2 = 1 or
det L = ±1 . (10.23)
One can classify Lorentz transformations according to the value of the determinant into two distinct
classes.
A second class property follows from (10.14) which we employ in the formulation (10.12). Consid-
ering in (10.12) the case ρ = 0, σ = 0 yields

2
2
2
2
L0 0 − L1 0 − L2 0 − L3 0 = 1 . (10.24)
or since (L1 0 )2 + (L2 0 )2 + (L3 0 )2 ≥ 0 it holds (L0 0 )2 ≥ 1. From this we can conclude
L0 0 ≥ 1 or L0 0 ≤ −1 , (10.25)
i.e., there exist two other distinct classes. Properties (10.23) and (10.25) can be stated as follows:
The set of all Lorentz transformations L is given as the union
L = L↑+ ∪ L↓+ ∪ L↑− ∪ L↓− (10.26)

where L↑+ , L↓+ , L↑− , L↓− are disjunct sets defined as follows

L↑+ = { L, L ∈ O(3, 1), det L = 1, L0 0 ≥ 1} ; (10.27)

L↓+ = { L, L ∈ O(3, 1), det L = 1, L0 0 ≤ −1} ; (10.28)
L↑− = { L, L ∈ O(3, 1), det L = −1, L0 0 ≥ 1} ; (10.29)
L↓− = { L, L ∈ O(3, 1), det L = −1, L0 0 ≤ −1} . (10.30)
It holds g ∈ L and −11 ∈ L as one can readily verify testing for property (10.14). One can also
verify that one can write
L↑− = gL↑+ = L↑+ g ; (10.31)
L↓+ = −L↑+ ; (10.32)
L↓− = − gL↑+ = − L↑+ g (10.33)
where we used the definition aM = {M1 , ∃M2 , M2 ∈ M, M1 = a M2 }. The above shows that
the set of proper Lorentz transformations L↑+ allows one to generate all Lorentz transformations,
except for the trivial factors g and −11. It is, hence, entirely suitable to investigate first only
Lorentz transformations in L↑+ .
We start our investigation by demonstrating that L↑+ forms a group. Obviously, L↑+ contains 11.
We can also demonstrate that for A, B ∈ L↑+ holds C = AB ∈ L↑+ . For this purpose we consider
the value of C 0 0 = A0 µ B µ 0 = 3j=1 A0 j B j 0 + A0 0 B 0 0 . Schwartz’s inequality yields
P

 2
3 3 3
X
0 j 
X
0 2

X
2
 A jB 0 ≤ A j Bj 0 . (10.34)
j=1 j=1 j=1
292 Relativistic Quantum Mechanics

From (10.12) follows (B 0 0 )2 − 3j=1 (B j 0 )2 = 1 or 3j=1 (B j 0 )2 = (B 0 0 )2 − 1. Similarly, one can

P P

conclude from (10.21) 3j=1 (A0 j )2 = (A0 0 )2 − 1. (10.34) provides then the estimate
P

 2
3
X
A0 j B j 0  ≤ (A0 0 )2 − 1 (B 0 0 )2 − 1 < (A0 0 )2 (B 0 0 )2 .
  
 (10.35)
j=1

One can conclude, therefore, | 3j=1 A0 j B j 0 | < A0 0 B 0 0 . Since A0 0 ≥ 1 and B 0 0 ≥ 1, obviously

P

A0 0 B 0 0 ≥ 1. Using the above expression for C 0 0 one can state C 0 0 > 0. In fact, since the group
property of O(3,1) ascertains CT gC = g it must hold C 0 0 ≥ 1.
The next group property of L↑+ to be demonstrated is the existence of the inverse. For the inverse of
any L ∈ L↑+ holds (10.16). This relationship shows (L−1 )0 0 = L0 0 , from which one can conclude
L−1 ∈ L↑+ . We also note that the identity operator 11 has elements

11µ ν = δ µ ν (10.36)

where we defined1 
µ 1 for µ = ν
δ ν = (10.37)
0 for µ 6= ν

It holds, 110 0 = ≥ 1 and, hence, 11 ∈ L↑+ . Since the associative property holds for matrix multipli-
cation we have verified that L↑+ is indeed a subgroup of SO(3,1).
L↑+ is called the subgroup of proper, orthochronous Lorentz transformations. In the following we
will consider solely this subgroup of SO(3,1).

Infinitesimal Lorentz transformations

The transformations in L↑+ have the property that they are continously connected to the identity 11,
i.e., these transformations can be parametrized such that a continuous variation of the parameters
connects any element of L↑+ with 11. This property will be exploited now in that we consider first
transformations in a small neighborhood of 11 which we parametrize by infinitesimal parameters.
We will then employ the Lie group properties to generate all transformations in L↑+ .
Accordingly, we consider transformations

L µ ν = δ µ ν + µ ν ; µ ν small . (10.38)

For these transformations, obviously, holds L0 0 > 0 and the value of the determinant is close to
unity, i.e., if we enforce (10.14) actually L0 0 ≥ 1 and det L = 1 must hold. Property (10.14)
implies
11 + T g ( 11 +  ) = g


(10.39)
where we have employed the matrix form  defined as in (10.13). To order O(2 ) holds

T g + g  = 0 . (10.40)
1 ρ
It should be noted that according to our present definition holds δµν = gµρ δ ν and, accordingly, δ00 = 1 and
δ11 = δ22 = δ33 = −1.
10.1: Natural Representation of the Lorentz Group 293

Using (10.15) one can conclude

T = − g  g (10.41)
which reads explicitly
 0
1 0 2 0 3 0 −0 0 0 1 0 2 0 3
  
 0
 0 1 1 1 2 1 3 1 1 1 1
 1  =   0 − 1 − 2 − 3
 
  . (10.42)
 0 2 1 2 2 2 3 2   2 0 −2 1 −2 2 −2 3 
0 3 1 3 2 3 3
 3 3 0 −3 1 −3 2 −3 3

This relationship implies

µ µ = 0
0 j = j 0 , j = 1, 2, 3
k
j k = − j , j, k = 1, 2, 3 (10.43)

Inspection shows that the matrix  has 6 independent elements and can be written
 
0 −w1 −w2 −w3
 −w1 0 −ϑ3 ϑ2 
(ϑ1 , ϑ2 , ϑ3 , w1 , w2 , w3 ) = 
 −w2
 . (10.44)
ϑ3 0 −ϑ1 
−w3 −ϑ2 ϑ1 0

This result allows us now to define six generators for the Lorentz transformations(k = 1, 2, 3)

Jk = (ϑk = 1, other five parameters zero) (10.45)

Kk = (wk = 1, other five parameters zero) . (10.46)

The generators are explicitly
     
0 0 0 0 0 0 0 0 0 0 0 0
 0 0 0 0   0 0 0 1   0 0 −1 0 
J1 =   ; J2 =  ; J =  (10.47)
0  3  0 1

 0 0 0 −1   0 0 0 0 0 
0 0 1 0 0 −1 0 0 0 0 0 0
     
0 −1 0 0 0 0 −1 0 0 0 0 −1
 −1 0 0 0 
  0 0 0 0 
  0 0 0 0 
K1 = 
 0 ; K2 = 
 −1 0 ; K3 =   (10.48)
0 0 0  0 0   0 0 0 0 
0 0 0 0 0 0 0 0 −1 0 0 0
These commutators obey the following commutation relationships

[ Jk , J` ] = k`m Jm (10.49)
[ Kk , K` ] = − k`m Jm
[ J k , K` ] = k`m Km .

The operators also obey

~J · K
~ = J1 J1 + J2 J2 + J3 J3 = 0 (10.50)
294 Relativistic Quantum Mechanics

as can be readily verified.

Exercise 7.1:
Demonstrate the commutation relationships (10.49, 10.50).

The commutation relationships (10.49) define the Lie algebra associated with the Lie group L↑+ .
The commutation relationships imply that the algebra of the generators Jk , Kk , k = 1, 2, 3 is closed.
Following the treatment of the rotation group SO(3) one can express the elements of L↑+ through
the exponential operators
 
~ w)
L(ϑ, ~ = exp ϑ ~ · ~J + w ~
~ ·K ~ w
; ϑ, ~ ∈ R3 (10.51)

where we have defined ϑ ~ · ~J = P3 ϑk Jk and w ~ ·K~ = P3 wk Kk . One can readily show,

k=1 k=1
following the algebra in Chapter 5, and using the relationship
 
0 0
Jk = (10.52)
0 Lk

where the 3 × 3–matrices Lk are the generators of SO(3) defined in Chapter 5, that the transfor-
mations (10.51) for w
~ = 0 correspond to rotations of the spatial coordinates, i.e.,
 
~ 0 0
L(ϑ, w
~ = 0) = ~ . (10.53)
0 R(ϑ)

~ are the 3 × 3–rotation matrices constructed in Chapter 5. For the parameters ϑk of the
Here R(ϑ)
Lorentz transformations holds obviously

ϑk ∈ [0, 2π[ , k = 1, 2, 3 (10.54)

which, however, constitutes an overcomplete parametrization of the rotations (see Chapter 5).
We consider now the Lorentz transformations for ϑ~ = 0 which are referred to as ‘boosts’. A boost
in the x1 –direction is L = exp(w1 K1 ). To determine the explicit form of this transformation we
evaluate the exponential operator by Taylor expansion. In analogy to equation (5.35) it issufficient
to consider in the present case the 2 × 2–matrix
∞ n
w1n
   
0 0 −1 X 0 −1
L = exp w1 = (10.55)
−1 0 n! −1 0
n=0

since  
0 0
L0
 0 0 
exp (w1 K1 ) = exp   . (10.56)
 0 0 1 0 
0 0 0 1
Using the idempotence property
 2  
0 −1 1 0
= = 11 (10.57)
−1 0 0 1
10.1: Natural Representation of the Lorentz Group 295

one can carry out the Taylor expansion above:

∞ ∞
w12n w12n+1
 
X X 0 −1
L0 = 11 + (10.58)
(2n)! (2n + 1)! −1 0
n=0 n=0
   
0 −1 cosh w1 −sinh w1
= cosh w1 11 + sinh w1 = .
−1 0 −sinh w1 cosh w1

The conventional form (10.1) of the Lorentz transformations is obtained through the parameter
change
sinh w1
v1 = = tanh w1 (10.59)
cosh w1
p p
Using cosh2 w1 − sinh2 w1 = 1 one can identify sinhw1 = cosh2 w1 − 1 and coshw1 = sinh2 w1 + 1.
Correspondingly, one obtains from (10.59)
p
cosh2 w1 − 1 sinh w1
v1 = = p . (10.60)
cosh w1 sinh2 w1 + 1

These two equations yield

q q
cosh w1 = 1/ 1 − v12 ; sinh w1 = v1 / 1 − v12 , (10.61)

and (10.56, 10.59) can be written

√−v1
 1

√ 0 0
1 − v12 1 − v12
√−v1 2 1
 
 √ 0 0 
exp (w1 K1 ) = 
 1 − v1 1 − v12 
 (10.62)
 0 0 1 0 
0 0 0 1

According to (10.3, 10.6, 10.51) the explicit transformation for space–time–coordinates is then

x1 − v1 t t − v1 x1
x01 = p 2
, t0 = p 2
, x02 = x2 , x03 = x3 (10.63)
1 − v1 1 − v1

which agrees with (10.1).

The range of the parameters wk can now be specified. vk defined in (10.59) for the case k = 1
corresponds to the relative velocity of two frames of reference. We expect that vk can only assume
values less than the velocity of light c which in the present units is c = 1. Accordingly, we can
state vk ∈ ] − 1, 1[. This property is, in fact, consistent with (10.59). From (10.59) follows, however,
for wk
wk ∈ ] − ∞, ∞[ . (10.64)
We note that the range of wk -values is not a compact set even though the range of vk -values is
compact. This property of the wk -values contrasts with the property of the parameters ϑk specifying
rotational angles which assume only values in a compact range.
296 Relativistic Quantum Mechanics

10.2 Scalars, 4–Vectors and Tensors

In this Section we define quantities according to their behaviour under Lorentz transformations.
Such quantities appear in the description of physical systems and statements about transformation
properties are often extremely helpful and usually provide important physical insight. We have
encountered examples in connection with rotational transformations, namely, scalars like r =
p
x21 + x22 + x23 , vectors like ~r = (x1 , x2 , x3 )T , spherical harmonics Y`m (r̂), total angular momentum
states of composite systems like Y`m (`1 , `2 |r̂1 , r̂2 ) and, finally, tensor operators Tkm . Some of these
quantities were actually defined with respect to representations of the rotation group in function
spaces, not in the so-called natural representation associated with the 3–dimensional Euclidean
space E3 .
Presently, we have not yet defined representations of Lorentz transformations beyond the ‘natural’
representation acting in the 4–dimensional space of position– and time–coordinates. Hence, our
definition of quantities with special properties under Lorentz transformations presently is confined
to the natural representation. Nevertheless, we will encounter an impressive example of physical
properties.

Scalars The quantities with the simplest transformation behaviour are so-called scalars f ∈ R
which are invariant under transformations, i.e.,
f0 = f . (10.65)
An example is s2 defined in (10.4), another example is the rest mass m of a particle. However, not
any physical property f ∈ R is a scalar. Counterexamples are the energy, the charge density, the
~ r, t)|2 or the scalar product ~r1 · ~r2
z–component x3 of a particle, the square of the electric field |E(~
of two particle positions. We will see below how true scalars under Lorentz transformations can be
constructed.

4-Vectors The quantities with the transformation behaviour like that of the position–time vector
xµ defined in (10.3) are the so-called 4–vectors aµ . These quantites always come as four components
(a0 , a1 , a2 , a3 )T and transform according to
µ
a0 = Lµ ν aν . (10.66)
Examples of 4-vectors beside xµ are the momentum 4-vector
m m ~v
pµ = (E, p~) , E = √ , p~ = √ (10.67)
1 − ~v 2 1 − ~v 2
the transformation behaviour of which we will demonstrate further below. A third 4-vector is the
so-called current vector
~
J µ = (ρ, J) (10.68)
~ r, t) are the charge density and the current density, respectively, of a system of
where ρ(~r, t) and J(~
charges. Another example is the potential 4-vector
~
Aµ = (V, A) (10.69)
~ r, t) are the electrical and the vector potential of an electromagnetic field. The
where V (~r, t) and A(~
4-vector character of J µ and of Aµ will be demonstrated further below.
10.2: Scalars, 4–Vectors and Tensors 297

Scalar Product 4-vectors allow one to construct scalar quantities. If aµ and bµ are 4-vectors
then
aµ gµν bν (10.70)
is a scalar. This property follows from (10.66) together with (10.12)
µ ν
a0 gµν b0 = Lµ ρ gµν Lν σ aρ bσ = aρ gρσ bσ (10.71)

Contravariant and Covariant 4-Vectors It is convenient to define a second class of 4-vectors.

The respective vectors aµ are associated with the 4-vectors aµ , the relationship being

aµ = gµν aν = (a0 , −a1 , −a2 , −a3 ) (10.72)

where aν is a vector with transformation behaviour as stated in (10.66). One calls 4-vectors aµ
covariant and 4-vectors aµ contravariant. Covariant 4-vectors transform like

a0 µ = gµν Lν ρ g ρσ aσ (10.73)

where we defined
g µν = gµν . (10.74)
We like to point out that from definition (10.72) of the covariant 4-vector follows =aµ g µν aν .
In
µν µ
fact, one can employ the tensors g and gµν to raise and lower indices of L ν as well. We do not
establish here the consistency of the ensuing notation. In any case one can express (10.73)

a0µ = Lµ σ aσ . (10.75)

Note that according to (10.17) Lµ σ is the transformation inverse to Lσ µ . In fact, one can express
[(L−1 )T ]µ ν = (L−1 )ν µ and, accordingly, (10.17) can be written

(L−1 )ν µ = Lµ ν . (10.76)

The 4-Vector ∂µ An important example of a covariant 4-vector is the differential operator

 
∂ ∂
∂µ = = ,∇ (10.77)
∂xµ ∂t
The transformed differential operator will be denoted by

def ∂
∂µ0 = . (10.78)
∂x0µ
To prove the 4-vector property of ∂µ we will show that g µν ∂ν transforms like a contravariant 4-
vector, i.e., g µν ∂ν0 = Lµ ρ g ρσ ∂σ . We start from x0 µ = Lµ ν xν . Multiplication (and summation) of
x0 µ = Lµ ν xν by Lρ σ gρµ yields, using (10.12), gσν xν = Lρ σ gρµ x0 µ and g µσ gσν = δ µ ν ,
µ
xν = g νσ Lρ σ gρµ x0 . (10.79)

This is the inverse Lorentz transformation consistent with (10.16). We have duplicated the expres-
sion for the inverse of Lµ ν to obtain the correct notation in terms of covariant, i.e., lower, and
298 Relativistic Quantum Mechanics

contravariant, i.e., upper, indices. (10.79) allows us to determine the connection between ∂µ and
∂µ0 . Using the chain rule of differential calculus we obtain

3
X ∂xν ∂
∂µ0 = = g νσ Lρ σ gρµ ∂ν = Lµ ν ∂ν (10.80)
∂x0µ ∂xν
ν=0

Multiplication by g λµ (and summation over µ) together with g λµ gρµ = δ λ ρ yields

g λµ ∂µ0 = Lλ σ g σν ∂ν , (10.81)

i.e., ∂µ does indeed transform like a covariant vector.

d’Alembert Operator We want to construct now a scalar differential operator. For this purpose
we define first the contravariant differential operator
 
µ µν ∂
∂ = g ∂ν = , −∇ . (10.82)
∂t

Then the operator

∂µ ∂ µ = ∂t2 − ∇2 (10.83)
is a scalar under Lorentz transformations. In fact, this operator is equal to the d’Alembert operator
which is known to be Lorentz-invariant.

Proof that pµ is a 4-vector We will demonstrate now that the momentum 4-vector pµ defined
in (10.67) transforms like (10.66). For this purpose we consider the scalar differential

(dτ )2 = dxµ dxµ = (dt)2 − (d~r)2 (10.84)

It holds  2
dτ
= 1 − (~v )2 (10.85)
dt
from which follows
d 1 d
= √ . (10.86)
dτ 1 − ~v 2 dt
One can write
m m dt
p0 = E = √ = √ . (10.87)
1 − ~v 2 1 − ~v dt
2

The remaining components of pµ can be written, e.g.,

m v1 m dx1
p1 = √ = √ . (10.88)
1 − ~v 2 1 − ~v 2 dt
One can express then the momentum vector

m dxµ d
pµ = √ = m xµ . (10.89)
1 − ~v 2 dt dτ
10.3: Relativistic Electrodynamics 299

d
The operator m dτ transforms like a scalar. Since xµ transforms like a contravariant 4-vector, the
r.h.s. of (10.89) alltogether transforms like a contravariant 4-vector, and, hence, pµ on the l.h.s. of
(10.89) must be a 4-vector.
The momentum 4-vector allows us to construct a scalar quantity, namely
pµ pµ = pµ gµν pν = E 2 − p~ 2 (10.90)
Evaluation of the r.h.s. yields according to (10.67)
m2 m2~v 2
E 2 − p~ 2 = − = m2 (10.91)
1 − ~v 2 1 − ~v 2
or
p µ p µ = m2 (10.92)
which, in fact, is a scalar. We like to rewrite the last result
E 2 = p~ 2 + m2 (10.93)
or p
E = ± p~ 2 + m2 . (10.94)
1
In the non-relativistic limit the rest energy m is the dominant contribution to E. Expansion in m
should then be rapidly convergent. One obtains
p~ 2 p 2 )2
 2 3
(~ (~
p )
E = ±m ± ∓ 3
+ O . (10.95)
2m 4m 4m5
p
~ 2
This obviously describes the energy of a free particle with rest energy ±m, kinetic energy ± 2m and
relativistic corrections.

10.3 Relativistic Electrodynamics

In the following we summarize the Lorentz-invariant formulation of electrodynamics and demon-
strate its connection to the conventional formulation as provided in Sect. 8.

Lorentz Gauge In our previous description of the electrodynamic field we had introduced the
~ r, t), respectively, and had chosen the so-called Coulomb
scalar and vector potential V (~r, t) and A(~
~
gauge (8.12), i.e., ∇ · A = 0, for these potentials. This gauge is not Lorentz-invariant and we will
adopt here another gauge, namely,
~ r, t) = 0 .
∂t V (~r, t) + ∇ · A(~ (10.96)
The Lorentz-invariance of this gauge, the so-called Lorentz gauge, can be demonstrated readily
using the 4-vector notation (10.69) for the electrodynamic potential and the 4-vector derivative
(10.77) which allow one to express (10.96) in the form
∂µ Aµ = 0 . (10.97)
We have proven already that ∂µ is a contravariant 4-vector. If we can show that Aµ defined in
(10.69) is, in fact, a contravariant 4-vector then the l.h.s. in (10.97) and, equivalently, in (10.96) is
a scalar and, hence, Lorentz-invariant. We will demonstrate now the 4-vector property of Aµ .
300 Relativistic Quantum Mechanics

Transformation Properties of J µ and Aµ

~ r, t) are known to obey the continuity property
The charge density ρ(~r, t) and current density J(~
~ r, t) = 0
∂t ρ(~r, t) + ∇ · J(~ (10.98)

which reflects the principle of charge conservation. This principle should hold in any frame of
reference. Equation (10.98) can be written, using (10.77) and (10.68),

∂µ J µ (xµ ) = 0 . (10.99)

Since this equation must be true in any frame of reference the right hand side must vanish in all
frames, i.e., must be a scalar. Consequently, also the l.h.s. of (10.99) must be a scalar. Since ∂µ
transforms like a covariant 4-vector, it follows that J µ , in fact, has to transform like a contravariant
4-vector.
We want to derive now the differential equations which determine the 4-potential Aµ in the Lorentz
gauge (10.97) and, thereby, prove that Aµ is, in fact, a 4-vector. The respective equation for A0 = V
can be obtained from Eq. (8.13). Using ∇ · ∂t A(~ ~ r, t) = ∂t ∇ · A(~~ r, t) together with (10.96), i.e.,
~
∇ · A(~r, t) = −∂t V (~r, t), one obtains

∂t2 V (~r, t) − ∇2 V (~r, t) = 4πρ(~r, t) . (10.100)

~ r, t) from (8.17) using the identity (8.18) and, according to (10.96),
Similarly, one obtains for A(~
~ r, t) = −∂t V (~r, t)
∇ · A(~
~ r, t) − ∇2 A(~
∂t2 A(~ ~ r, t) = 4 π J(~
~ r, t) . (10.101)
Combining equations (10.100, 10.101), using (10.83) and (10.69), yields

∂µ ∂ µ Aν (xσ ) = 4 π J ν (xσ ) . (10.102)

In this equation the r.h.s. transforms like a 4-vector. The l.h.s. must transform likewise. Since
∂µ ∂ µ transforms like a scalar one can conclude that Aν (xσ ) must transform like a 4-vector.

The Field Tensor

The electric and magnetic fields can be collected into an anti-symmetric 4×4 tensor
 
0 −Ex −Ey −Ez
 Ex 0 −Bz By 
F µν =   . (10.103)
 Ey Bz 0 −Bx 
Ez −By Bx 0

Alternatively, this can be stated

F k0 = −F 0k = E k , F mn = −mn` B ` , k, `, m, n = 1, 2, 3 (10.104)

where mn` = mn` is the totally anti-symmetric three-dimensional tensor defined in (5.32).
One can readily verify, using (8.6) and (8.9), that F µν can be expressed through the potential Aµ
in (10.69) and ∂ µ in (10.82) as follows

F µν = ∂ µ Aν − ∂ ν Aµ . (10.105)
10.3: Relativistic Electrodynamics 301

~ r, t) and B(~
The relationships (10.103, 10.104) establishe the transformation behaviour of E(~ ~ r, t).
In a new frame of reference holds
F 0µν = Lµ α Lν β F αβ (10.106)
In case that the Lorentz transformation Lµ ν is given by (10.62) or, equivalently, by (10.63), one
obtains
Ey −v1 Bz Ez +v1 By 
− √ − √

0 −Ex 2
1−v1 1−v12
 
Bz −v1 Ey By +v1 Ez 
√ √

 E x 0 − 
0µν
 1−v12 1−v12 
F =  E −v B

Bz −v1 Ey
 (10.107)
 √ 2
y 1 z
√ 2 0 −Bx


 1−v1 1−v1 
 
Ez +v1 By By +v1 Ez
√ 2 − √ 2 Bx 0
1−v1 1−v1

Comparision with
0 −Ex0 −Ey0 −Ez0
 
 Ex0 0 −Bz0 By0 
F 0µν = 
 Ey0 0 (10.108)
−Bx0 

Bz 0
Ez0 −By0 Bx0 0
~ 0 and B
yields then the expressions for the transformed fields E ~ 0 . The results can be put into the
more general form

~ + ~v × B ~
~0
E = ~k ,
E ~ 0 = E⊥
E √ (10.109)
k ⊥
1 − ~v 2
~ − ~v × E ~
~0
B = ~k ,
B ~ 0 = B⊥
B √ (10.110)
k ⊥
1 − ~v 2

where E~ k, B
~ k and E
~ ⊥, B
~ ⊥ are, respectively, the components of the fields parallel and perpendicular
to the velocity ~v which determines the Lorentz transformation. These equations show that under
Lorentz transformations electric and magnetic fields convert into one another.

Maxwell Equations in Lorentz-Invariant Form

One can express the Maxwell equations in terms of the tensor F µν in Lorentz-invariant form. Noting

∂µ F µν = ∂µ ∂ µ Aν − ∂µ ∂ ν Aµ = ∂µ ∂ µ Aν − ∂ ν ∂µ Aµ = ∂µ ∂ µ Aν , (10.111)

where we used (10.105) and (10.97), one can conclude from (10.102)

∂µ F µν = 4π J ν . (10.112)

One can readily prove that this equation is equivalent to the two inhomogeneous Maxwell equations
(8.1, 8.2). From the definition (10.105) of the tensor F µν one can conclude the property

∂ σ F µν + ∂ µ F νσ + ∂ ν F σµ = 0 (10.113)

which can be shown to be equivalent to the two homogeneous Maxwell equations (8.3, 8.4).
302 Relativistic Quantum Mechanics

Lorentz Force

One important property of the electromagnetic field is the Lorentz force acting on charged particles
moving through the field. We want to express this force through the tensor F µν . It holds for a
particle with 4-momentum pµ as given by (10.67) and charge q

dpµ q
= pν F µν (10.114)
dτ m

where d/dτ is given by (10.86). We want to √ demonstrate now that this equation is equivalent to
the equation of motion (8.5) where p~ = m~v / 1 − v 2 .
To avoid √confusion we will employ in the following for the energy of the particle the notation
E = m/ 1 − v 2 [see (10.87)] and retain the definition E ~ for the electric field. The µ = 0
component of (10.114) reads then, using (10.104),

dE q ~
= p~ · E (10.115)
dτ m

or with (10.86)
dE q ~ .
= p~ · E (10.116)
dt E
From this one can conclude, employing (10.93),

1 dE 2 1 d~p 2 ~
= = q p~ · E (10.117)
2 dt 2 dt

This equation follows, however, also from the equation of motion (8.5) taking the scalar product
with p~
d~p ~
p~ · p·E
= q~ (10.118)
dt
√
where we exploited the fact that according to p~ = m~v / 1 − v 2 holds p~ k ~v .
For the spatial components, e.g., for µ = 1, (10.114) reads using (10.103)

dpx q
= ( EEx + py Bz − pz By ) . (10.119)
dτ m
√
Employing again (10.86) and (10.67), i.e., E = m/ 1 − v 2 , yields

dpx h
~ x
i
= q Ex + (~v × B) (10.120)
dt

which is the x-component of the equation of motion (8.5). We have, hence, demonstrated that
(10.114) is, in fact, equivalent to (8.5). The term on the r.h.s. of (10.120) is referred to as the
Lorentz force. Equation (10.114), hence, provides an alternative description of the action of the
Lorentz force.
10.4: Function Space Representation of Lorentz Group 303

10.4 Function Space Representation of Lorentz Group

In the following it will be required to decribe the transformation of wave functions under Lorentz
transformations. In this section we will investigate the transformation properties of scalar functions
ψ(xµ ), ψ ∈ C∞ (4). For such functions holds in the transformed frame

ψ 0 (Lµ ν xν ) = ψ(xµ ) (10.121)

which states that the function values ψ 0 (x0µ ) at each point x0µ in the new frame are identical to the
function values ψ(xµ ) in the old frame taken at the same space–time point xµ , i.e., taken at the pairs
of points (x0µ = Lµ ν xν , xµ ). We need to emphasize that (10.121) covers solely the transformation
behaviour of scalar functions. Functions which represent 4-vectorsor other non-scalar entities, e.g.,
the charge-current density in case of Sect. 10.3 or the bi-spinor wave function of electron-positron
pairs in Sect. 10.7, obey a different transformation law.
We like to express now ψ 0 (x0µ ) in terms of the old coordinates xµ . For this purpose one replaces
xµ in (10.121) by (L−1 )µ ν xν and obtains

ψ 0 (xµ ) = ψ((L−1 )µ ν xν ) . (10.122)

This result gives rise to the definition of the function space representation ρ(Lµ ν ) of the Lorentz
group
def
(ρ(Lµ ν )ψ)(xµ ) = ψ((L−1 )µ ν xν ) . (10.123)
This definition corresponds closely to the function space representation (5.42) of SO(3). In analogy
to the situation for SO(3) we seek an expression for ρ(Lµ ν ) in terms of an exponential operator and
transformation parameters ϑ, ~ w,
~ i.e., we seek an expression which corresponds to (10.51) for the
natural representation of the Lorentz group. The resulting expression should be a generalization of
the function space representation (5.48) of SO(3), in as far as SO(3,1) is a generalization (rotation
+ boosts) of the group SO(3). We will denote the intended representation by
 
~ w) def ~ w)) ~ ~ ~ K
~
L(ϑ, ~ = ρ(Lµ ν (ϑ, ~ = ρ eϑ·J + w· (10.124)

which we present in the form

 
~
L(ϑ, w) ~ ~
~ = exp ϑ · J + w ~
~ ·K . (10.125)

In this expression J~ = (J1 , J2 , J3 ) and K ~ w)

~ = (K1 , K2 , K3 ) are the generators of L(ϑ, ~ which
correspond to the generators Jk and Kk in (10.47), and which can be constructed following the
procedure adopted for the function space representation of SO(3). However, in the present case we
exclude the factor ‘−i’ [cf. (5.48) and (10.125)]. Accordingly, one can evaluate Jk as follows

1 h  ϑk J k  i
Jk = lim ρ e − 11 (10.126)
ϑk →0 ϑ1

and Kk
1 
ρ ewk Kk − 11 .


Kk = lim (10.127)
wk →0 w1
304 Relativistic Quantum Mechanics

One obtains

J1 = x3 ∂2 − x2 ∂3 ; K1 = x0 ∂1 + x1 ∂0
J2 = x1 ∂3 − x3 ∂1 ; K2 = x0 ∂2 + x2 ∂0
J3 = x2 ∂1 − x1 ∂2 ; K3 = x0 ∂3 + x3 ∂0 (10.128)

which we like to demonstrate for J1 and K1 .

In order to evaluate (10.126) for J1 we consider first
 
1 0 0 0
 −1  0 1 0 0 
eϑ1 J1 = e−ϑ1 J1 = 
 0
 (10.129)
0 cosϑ1 sinϑ1 
0 0 −sinϑ1 cosϑ1

which yields for small ϑ1

ρ eϑ1 J1 ψ(xµ ) = ψ(x0 , x1 , cosϑ1 x2 + sinϑ1 x3 , −sinϑ1 x2 + cosϑ1 x3 )

= ψ(xµ ) + ϑ1 (x3 ∂2 − x2 ∂3 ) ψ(xµ ) + O(ϑ21 ) . (10.130)

This result, obviously, reproduces the expression for J1 in (10.128).

One can determine similarly K1 starting from
 
coshw1 sinhw1 0 0

−1  sinhw1 coshw1 0 0 
ew1 K1 = e−w1 K1 =   . (10.131)
 0 0 1 0 
0 0 0 1

This yields for small w1

ρ ew1 K1 ψ(xµ ) = ψ(coshw1 x0 + sinhw1 x1 , sinhw1 x0 + coshw1 x1 , x2 , x3 )

= ψ(xµ ) + w1 (x1 ∂0 + x0 ∂1 ) ψ(xµ ) + O(w12 ) (10.132)

and, obviously, the expression for K1 in (10.126).

The generators J~ , K
~ obey the same Lie algebra (10.49) as the generators of the natural represen-
tation, i.e.

[ Jk , J` ] = k`m Jm
[ Kk , K` ] = − k`m Jm
[ Jk , K` ] = k`m Km . (10.133)

We demonstrate this for three cases, namely [J1 , J2 ] = J3 , [K1 , K2 ] = −J3 , and [J1 , K2 ] = K3 :

[ J1 , J2 ] = [x3 ∂2 − x2 ∂3 , x1 ∂3 − x3 ∂1 ]
= [x3 ∂2 , x1 ∂3 ] − [x2 ∂3 , x3 ∂1 ]
= −x1 ∂2 + x2 ∂1 = J3 , (10.134)
10.4: Function Space Representation of Lorentz Group 305

[ K1 , K2 ] = [x0 ∂1 + x1 ∂0 , x0 ∂2 + x2 ∂0 ]
= [x0 ∂1 , x2 ∂0 ] − [x1 ∂0 , x0 ∂2 ]
= −x2 ∂1 + x1 ∂2 = − J3 , (10.135)

[ J1 , K2 ] = [x3 ∂2 − x2 ∂3 , x0 ∂2 + x2 ∂0 ]
= [x3 ∂2 , x2 ∂0 ] − [x2 ∂3 , x0 ∂2 ]
= x3 ∂0 + x0 ∂3 = K3 . (10.136)

One-Dimensional Function Space Representation

The exponential operator (10.125) in the case of a one-dimensional transformation of the type

L(w3 ) = exp w3 K3 ,


(10.137)

where K3 is given in (10.128), can be simplified considerably. For this purpose one expresses K3 in
terms of hyperbolic coordinates R, Ω which are connected with x0 , x3 as follows

x0 = R coshΩ , x3 = R sinhΩ (10.138)

a relationship which can also be stated

 p 0 2
+ p(x ) − (x3 )2 if x0 ≥ 0
R = (10.139)
− (x0 )2 − (x3 )2 if x0 < 0
and
x3 x0
tanhΩ = , cothΩ = . (10.140)
x0 x3
The transformation to hyperbolic coordinates closely resembles the transformation to radial coordi-
nates for the generators of SO(3) in the function space representation [cf. Eqs. (5.85-5.87)].
p In both
cases the radial coordinatepis the quantity conserved under the transformations, i.e., x21 + x22 + x23
in the case of SO(3) and (x0 )2 − (x3 )2 in case of transformation (10.137).
In the following we consider solely the case x0 ≥ 0. The relationships (10.139, 10.140) allow one to
∂ ∂
express the derivatives ∂0 , ∂3 in terms of ∂R , ∂Ω . We note

∂R x0 ∂R x0
= , = − (10.141)
∂x0 R ∂x3 R
and
∂Ω ∂Ω ∂tanhΩ 1
= = cosh2 Ω 0
∂x3 ∂tanhΩ ∂x 3 x
∂Ω ∂Ω ∂cothΩ 1
= = − sinh2 Ω 3 . (10.142)
∂x0 ∂cothΩ ∂x 0 x
The chain rule yields then
∂R ∂ ∂Ω ∂ x0 ∂ 1 ∂
∂0 = 0
+ 0
= − sinh2 Ω 3
∂x ∂R ∂x ∂Ω R ∂R x ∂Ω
∂R ∂ ∂Ω ∂ x3 ∂ 1 ∂
∂3 = 3
+ 3
= − + cosh2 Ω 0 . (10.143)
∂x ∂R ∂x ∂Ω R ∂R x ∂Ω
306 Relativistic Quantum Mechanics

Inserting these results into the definition of K3 in (10.128) yields

∂
K3 = x0 ∂3 + x3 ∂0 = . (10.144)
∂Ω
The action of the exponential operator (10.137) on a function f (Ω) ∈ C∞ (1) is then that of a shift
operator  
3 3 ∂
L(w ) f (Ω) = exp w f (Ω) = f (Ω + w3 ) . (10.145)
∂Ω

10.5 Klein–Gordon Equation

In the following Sections we will provide a heuristic derivation of the two most widely used quan-
tum mechanical descriptions in the relativistic regime, namely the Klein–Gordon and the Dirac
equations. We will provide a ‘derivation’ of these two equations which stem from the historical de-
velopment of relativistic quantum mechanics. The historic route to these two equations, however,
is not very insightful, but certainly is short and, therefore, extremely useful. Further below we will
provide a more systematic, representation theoretic treatment.

Free Particle Case

A quantum mechanical description of a relativistic free particle results from applying the correspon-
dence principle, which allows one to replace classical observables by quantum mechanical operators
acting on wave functions. In the position representation the correspondence principle states
~
E =⇒ Ê = − ∂t
i
~
ˆ~ = ∇
p~ =⇒ p (10.146)
i
which, in 4-vector notation reads

pµ =⇒ p̂µ = i~(∂t , ∇) = i~∂µ ; pµ =⇒ p̂µ = i(∂t , −∇) = i~∂ µ . (10.147)

Applying the correspondence principle to (10.92) one obtains the wave equation

− ~2 ∂ µ ∂µ ψ(xν ) = m2 ψ(xν ) (10.148)

or
~2 ∂ µ ∂µ + m2 ψ(xν ) = 0 .


(10.149)
where ψ(xµ ) is a scalar, complex-valued function. The latter property implies that upon change of
reference frame ψ(xµ ) transforms according to (10.121, 10.122). The partial differential equation
(10.151) is called the Klein-Gordon equation.
In the following we will employ so-called natural units ~ = c = 1. In these units the quantities
energy, momentum, mass, (length)−1 , and (time)−1 all have the same dimension. In natural units
the Klein–Gordon equation (10.151) reads

∂µ ∂ µ + m2 ψ(xµ ) = 0


(10.150)
10.5: Klein–Gordon Equation 307

or
∂t2 − ∇2 + m2 ψ(xµ ) = 0 .


(10.151)
One can notice immeadiately that (10.150) is invariant under Lorentz transformations. This follows
from the fact that ∂µ ∂ µ and m2 are scalars, and that (as postulated) ψ(xµ ) is a scalar. Under
Lorentz transformations the free particle Klein–Gordon equation (10.150) becomes
∂µ0 ∂ 0µ + m2 ψ 0 (x0µ ) = 0


(10.152)
which has the same form as the Klein–Gordon equation in the original frame.

Current 4-Vector Associated with the Klein-Gordon Equation

As is well-known the Schrödinger equation of a free particle
1 2
i∂t ψ(~r, t) = − ∇ ψ(~r, t) (10.153)
2m
is associated with a conservation law for particle probability
∂t ρS (~r, t) + ∇ · ~jS (~r, t) = 0 (10.154)
where
ρS (~r, t) = ψ ∗ (~r, t) ψ(~r, t) (10.155)
describes the positive definite probability to detect a particle at position ~r at time t and where
1
~jS (~r, t) = [ ψ ∗ (~r, t)∇ψ(~r, t) − ψ(~r, t)∇ψ ∗ (~r, t) ] (10.156)
2mi
describes the current density connected with motion of the particle probability distribution. To
1
derive this conservation law one rewrites the Schrödinger equation in the form (i∂t − 2m ∇2 )ψ = 0
and considers    
∗ 1 2
Im ψ i∂t − ∇ ψ = 0 (10.157)
2m
which is equivalent to (10.154).
In order to obtain the conservation law connected with the Klein–Gordon equation (10.150) one
considers
Im ψ ∗ ∂µ ∂ µ + m2 ψ = 0


(10.158)
which yields
ψ ∗ ∂t2 ψ − ψ∂t2 ψ ∗ − ψ ∗ ∇2 ψ + ψ∇2 ψ ∗ =
∂t (ψ ∗ ∂t ψ − ψ∂t ψ∗) + ∇· (ψ∇ψ ∗ − ψ ∗ ∇ψ) = 0 (10.159)
which corresponds to
∂t ρKG (~r, t) + ∇ · ~jKG (~r, t) = 0 (10.160)
where
i
ρKG (~r, t) = ( ψ ∗ (~r, t)∂t ψ(~r, t) − ψ(~r, t)∂t ψ ∗ (~r, t) )
2m
1
~jKG (~r, t) = ( ψ ∗ (~r, t)∇ψ(~r, t) − ψ(~r, t)∇ψ ∗ (~r, t) ) . (10.161)
2mi
308 Relativistic Quantum Mechanics

This conservation law differs in one important aspect from that of the Schrödinger equation (10.154),
namely, in that the expression for ρKG is not positive definite. When the Klein-Gordon equation
had been initially suggested this lack of positive definiteness worried physicists to a degree that the
Klein–Gordon equation was rejected and the search for a Lorentz–invariant quantum mechanical
wave equation continued. Today, the Klein-Gordon equation is considered as a suitable equation
to describe spin–0 particles, for example pions. The proper interpretation of ρKG (~r, t), it had been
realized later, is actually that of a charge density, not of particle probability.

Solution of the Free Particle Klein–Gordon Equation

Solutions of the free particle Klein–Gordon equation are
µ
ψ(xµ ) = N e−ipµ x = N ei(~p0 ·~r − Eo t) . (10.162)

Inserting this into the Klein–Gordon equation (10.151) yields

Eo2 − p~02 − m2 ψ(~r, t) = 0

(10.163)

which results in the expected [see (10.93] dispersion relationship connecting E0 , p~0 , m

E02 = m2 + p~o2 . (10.164)

The corresponding energy is p

po , ±) = ±
Eo (~ m2 + p~o2 (10.165)
This result together with (10.162) shows that the solutions of the free particle Klein-Gordon e-
quation (10.150) are actually determined by p~o and by the choice of sign ±. We denote this by
summarizing the solutions as follows

∂µ ∂ µ + m2 ψo (~p, λ|xµ )


= 0 (10.166)
p
p, λ|xµ ) = Nλ,p ei(~p·~r − λEo (~p)t) Eo (~
ψo (~ p) = m2 + p~o2 , λ = ±

The spectrum of the Klein–Gordon equation (10.150) is a continuum of positive energies E ≥ m,

corresponding to λ = +, and of negative energies E ≤ −m, corresponding to λ = −. The density
ρKG (~
p, λ) associated with the corresponding wave functions ψo (~ p, λ|xµ ) according to (10.161) and
(10.166) is
Eo (~
p) ∗
ρKG (~
p, λ) = λ p, λ|xµ )ψo (~
ψo (~ p, λ|xµ ) (10.167)
m
which is positive for λ = + and negative for λ = −. The proper interpretation of the two cases is
that the Klein–Gordon equation describes particles as well as anti-particles; the anti-particles carry
a charge opposite to that of the associated particles, and the density ρKG (~ p, λ) actually describes
charge density rather than probability.

Generating a Solution Through Lorentz Transformation

A particle at rest, i.e., with p~ = 0, according to (??) is decribed by the ~r–independent wave
function
p = 0, λ|xµ ) = N e−iλmt , λ = ± .
ψo (~ (10.168)
10.6: Klein–Gordon Equation with Electromagnetic Field 309

We want to demonstrate now that the wave functions for p~ 6= 0 in (??) can be obtained through
appropriate Lorentz transformation of (10.168). For this purpose we consider the wave function for
a particle moving with momentum velocity v in the direction of the x3 –axis. Such wave function
should be generated by applying the Lorentz transformation in the function space representation
p
(10.145) choosing m = sinhw3 . This yields, in fact, for the wave function (10.168), using (10.138)
to replace t = x0 by hyperbolic coordinates R, Ω,
 
3 ∂
3
L(w )ψo (~ µ
p = 0, λ|x ) = exp w N e−iλmRcoshΩ
∂Ω
3)
= N e−iλmRcosh(Ω+w . (10.169)

The addition theorem of hyperbolic functions cosh(Ω + w3 ) = coshΩ coshw3 + sinhΩ sinhw3 allows
us to rewrite the exponent on the r.h.s. of (10.169)

−iλ ( m coshw3 ) ( R coshΩ ) − iλ ( m sinhw3 ) ( R sinhΩ ) . (10.170)

The coordinate transformation (10.138) and the relationships (10.61) yield for this expression
m mv
−iλ √ x0 − iλ √ x3 . (10.171)
1−v 2 1−v 2

One can interpret then for λ = +, i.e., for positive energy solutions,
p
p = − mv/ 1 − v 2 (10.172)

as the momentum of the particle relative to the moving frame and

r r
m m2 m2 v 2 p
√ = 2
= m2 + 2
= m2 + p2 = Eo (p) (10.173)
1 − v2 1−v 1−v
as the energy [c.f. (10.166)] of the particle. In case of λ = + one obtains finally
3 − Eo (p)x0
L(w3 )ψo (~
p = 0, λ = +|xµ ) = N ei(px (10.174)

which agrees with the expression given in (10.166). In case of λ = −, i.e., for negative energy
solutions, one has to interprete p
p = mv/ 1 − v 2 (10.175)
as the momentum of the particle and one obtains
3 + Eo (p)x0
L(w3 )ψo (~
p = 0, λ = −|xµ ) = N ei(px . (10.176)

10.6 Klein–Gordon Equation for Particles in an Electromagnetic

Field
We consider now the quantum mechanical wave equation for a spin–0 particle moving in an elec-
tromagnetic field described by the 4-vector potential
~ r, t)) ;
Aµ (xµ ) = (V (~r, t), A(~ ~ r, t))
Aµ (xµ ) = (V (~r, t), −A(~ (10.177)
310 Relativistic Quantum Mechanics

free classical free quantum

classical particle in quantum particle in
particle ~
field (V, A) particle ~
field (V, A)

energy E E − qV i∂t i∂t − qV

momentum p~ ~
p~ − q A ˆ~ = −i∇
p ˆ~ − q A
p ˆ
~ = ~π

4-vector pµ pµ − qAµ i∂µ i∂µ − qAµ = πµ

Table 10.1:
Coupling of a particle of charge q to an electromagnetic field described by the 4-vector potential
Aµ = (V, A) ~ or Aµ = (V, −A).~ According to the so-called minimum coupling principle the presence
of the field is accounted for by altering energy, momenta for classical particles and the respective
operators for quantum mechanical particles in the manner shown. See also Eq. (10.147).

To obtain the appropriate wave equation we follow the derivation of the free particle Klein–Gordon
equation above and apply again the correspondence principle to (10.93), albeit in a form, which
couples a particle of charge q to an electromagnetic field described through the potential Aµ (xν ).
According to the principle of minimal coupling [see (10.69)] one replaces the quantum mechanical
operators, i.e., i∂t and −i∇ in (10.150), according to the rules shown in Table 10.1. For this purpose
one writes the Klein-Gordon equation (10.150)

−g µν (i∂µ )(i∂ν ) + m2 ψ(xµ ) = 0 .

(10.178)

According to the replacements in Table 10.1 this becomes

g µν (i∂µ − qAµ )(i∂ν − Aν ) ψ(xµ ) = m2 ψ(xµ ) (10.179)

which can also be written

g µν πµ πν − ; m2 ) ψ(xµ ) = 0 . (10.180)
In terms of space-time derivatives this reads
 2 
2 ~ 2
(i∂t − qV (~r, t)) ψ(~r, t) = −i∇ − q A(~r, t) + m ψ(~r, t) . (10.181)

Non-Relativistic Limit of Free Particle Klein–Gordon Equation

In order to consider further the interpretation of the positive and negative energy solutions of the
Klein–Gordon equation one can consider the non-relativistic limit. For this purpose we split-off a
factor exp(−imt) which describes the oscillations of the wave function due to the rest energy, and
focus on the remaining part of the wave function, i.e., we define

ψ(~r, t) = e−imt Ψ(~r, t) , (10.182)

and seek an equation for Ψ(~r, t). We will also assume, in keeping withnthe non-relativistic limit,
that the mass m of the particle, i.e., it’s rest energy, is much larger than all other energy terms, in
10.6: Klein–Gordon Equation with Electromagnetic Field 311

particular, larger than |i∂t Ψ/Ψ| and alrger than qV , i.e.,

i∂t Ψ
| | << m , |q V | << m . (10.183)
Ψ
The term on the l.h.s. of (10.181) can then be approximated as follows:

(i∂t − qV )2 e−imt Ψ = (i∂t − qV ) (me−imt Ψ + e−imt i∂t Ψ − qV e−imt Ψ)

= m2 e−imt Ψ + me−imt i∂t Ψ − qV e−imt Ψ
+me−imt i∂t Ψ − e−imt ∂ 2 Ψ − qV e−imt i∂t Ψ
−me−imt qV Ψ − e−imt i∂t qV Ψ + q 2 V 2 e−imt Ψ
≈ m2 e−imt Ψ − 2mqV e−imt Ψ − 2me−imt i∂t Ψ (10.184)

where we neglected all terms which did not contain factors m. The approximation is justified on
the ground of the inequalities (10.183). The Klein-Gordon equation (10.181) reads then
" #
ˆ~ − q A(~
[p ~ r, t)]2
i ∂t Ψ(~r, t) = + qV (~r, t) Ψ(~r, t) (10.185)
2m

This is, however, identical to the Schrödinger equation (10.2) of a non-relativistic spin-0 particle
moving in an electromagnetic field.

Pionic Atoms
To apply the Klein–Gordon equation (10.181) to a physical system we consider pionic atoms, i.e.,
atoms in which one or more electrons are replaced by π − mesons. This application demonstrates
that the Klein–Gordon equation describes spin zero particles, e.g., spin-0 mesons.
To ‘manufacture’ pionic atoms, π − mesons are generated through inelastic proton–proton scattering

p + p −→ p + p + π − + π + , (10.186)

then are slowed down, filtered out of the beam and finally fall as slow pions onto elements for which
a pionic variant is to be studied. The process of π − meson capture involves the so-called Auger
effect, the binding of a negative charge (typically an electron) while at the same time a lower shell
electron is being emitted

π − + atom −→ (atom − e− + π − ) + e− . (10.187)

We want to investigate in the following a description of a stationary state of a pionic atom involving
a nucleus with charge +Ze and a π − meson. A stationary state of the Klein–Gordon equation is
described by a wave function
ψ(xµ ) = ϕ(~r ) e−it . (10.188)
Inserting this into (10.181) yields (we assume now that the Klein–Gordon equation describes a
2
~ r, t) ≡ 0
particle with mass mπ and charge −e) for qV (~r, t) = − Zer and A(~
" 2 #
Ze2 2 2
+ + ∇ − mπ ϕ(~r ) = 0 . (10.189)
r
312 Relativistic Quantum Mechanics

Because of the radial symmetry of the Coulomb potential we express this equation in terms of
spherical coordinates r, θ, φ. The Laplacian is

1 2 1 1 1 L̂2
∇2 = ∂r r + 2 2 ∂θ sinθ∂θ + 2 2 ∂φ2 = ∂r2 r − 2 . (10.190)
r r sin θ r sin θ r r
Ze2 2
With this expression and after expanding ( + r ) one obtains
!
d2 L̂2 − Z 2 e4 2Ze2
− + + 2 − m2π r φ(~r) = 0 . (10.191)
dr2 r2 r

The operator L̂2 in this equation suggests to choose a solution of the type

R` (r)
ϕ(~r ) = Y`m (θ, φ) (10.192)
r

where the functions Y`m (θ, φ) are spherical harmonics, i.e., the eigenfunctions of the operator L̂2
in (10.191)
L̂2 Y`m (θ, φ) = ` (` + 1) Y`m (θ, φ) . (10.193)
(10.192) leads then to the ordinary differential equation

d2 `(` + 1) − Z 2 e4 2Ze2
 
− + + 2 − m2π R` (r) = 0 . (10.194)
dr2 r2 r

Bound state solutions can be obtained readily noticing that this equation is essentially identical to
2
that posed by the Coulomb problem (potential − Zer ) for the Schrödinger equation

d2 2mπ Ze2
 
`(` + 1)
− + + 2mπ E R` (r) = 0 (10.195)
dr2 r2 r

The latter problem leads to the well-known spectrum

mπ (Ze2 )2
En = − ; n = 1, 2, . . . ; ` = 0, 1, . . . n − 1 . (10.196)
2n2
In this expression the number n0 defined through

n0 = n − ` − 1 (10.197)

counts the number of nodes of the wave function, i.e., this quantity definitely must be an integer.
The similarity of (10.194) and (10.195) can be made complete if one determines λ such that

λ(`) (λ(`) + 1) = ` (` + 1) − Z 2 e4 . (10.198)

The suitable choice is r

1 1 2
λ(`) = − + (` + ) − Z 2 e4 (10.199)
2 2
10.6: Klein–Gordon Equation with Electromagnetic Field 313

and one can write (10.194)

d2 2Ze2
 
λ(`) ( λ(`) + 1 )
− + + 2 − m2π R` (r) = 0 . (10.200)
dr2 r2 r

The bound state solutions of this equation should correspond to  values which can be obtained
from (10.196) if one makes the replacement

2 − m2π 
E −→ , ` −→ λ(`) , e2 −→ e2 . (10.201)
2mπ mπ

One obtains
2
2 − m2π mπ Z 2 e4 m 2
π
= − . (10.202)
2mπ 2 (n0 + λ(`) + 1)2
Solving this for  (choosing the root which renders  ≤ mπ , i.e., which corresponds to a bound
state) yields
mπ
 = q ; n0 = 0, 1, . . . ; ` = 0, 1, . . . . (10.203)
Z 2 e4
1 + ( n0 + λ(`)+1 )2

EKG (n, `, m) =
Using (10.197, 10.199) and defining EKG =  results in the spectrum mπ
r
Z 2 e4
1 + 1
q
(n−`− 2
+ (`+ 12 )2 − Z 2 e4 )2

n = 1, 2, . . .
` = 0, 1, . . . , n − 1
m = −`, −` + 1, . . . , +`
(10.204)
In order to compare this result with the spectrum of the non-relativistic hydrogen-like atom we
expand in terms of the fine structure constant e2 to order O(8 ). Introducing α = Z 2 e4 and
β = ` + 12 (10.204) reads
1
q α
(10.205)
1+ √ 2 2(n − β + β −α)

and one obtains the series of approximations

1
α
q
1+ √
(n − β + β 2 −α)2
1
≈ q α
1+ (n − α
+ O(α2 ))2
2β

1
≈ q α
1+ n2 − α
n + O(α2 )
β
314 Relativistic Quantum Mechanics

1
≈ q
α α2
1+ n2
+ βn3
+ O(α3 )
1
≈ α α2 α2
1+ 2n2
+ 2βn3
− 8n4
+ O(α3 )
α α2
α2 α2
≈ 1− − + + + O(α3 ) . (10.206)
2n2 2βn3 8n4 4n4

From this results for (10.204)

" #
mZ 2 e4 mZ 4 e8 1 3
EKG (n, `, m) ≈ m − − 1 − + O(Z 6 e12 ) . (10.207)
2n2 2n3 `+ 2
4n

Here the first term represents the rest energy, the second term the non-relativistic energy, and the
third term gives the leading relativistic correction. The latter term agrees with observations of
pionic atoms, however, it does not agree with observations of the hydrogen spectrum. The latter
spectrum shows, for example, a splitting of the six n = 2, ` = 1 states into groups of two and four
degenerate states. In order to describe electron spectra one must employ the Lorentz-invariant
wave equation for spin- 21 particles, i.e., the Dirac equation introduced below.
It must be pointed out here that  does not denote energy, but in the present case rather the negative
of the energy. Also, the π − meson is a pseudoscalar particle, i.e., the wave function changes sign
under reflection.

10.7 The Dirac Equation

Historically, the Klein–Gordon equation had been rejected since it did not yield a positive-definite
probability density, a feature which is connected with the 2nd order time derivative in this equation.
This derivative, in turn, arises because the Klein–Gordon equation, through the correspondence
principle, is related to the equation E 2 = m2 + p~ 2 of the classical theory which involves a term
E 2 . A more satisfactory Lorentz–invariant wave equation, i.e., one with a positive-definite density,
would have only a first order time derivative. However, because of the equivalence of space and
time coordinates in the Minkowski space such equation necessarily can only have then first order
derivatives with respect to spatial coordinates. It should feature then a differential operator of the
type D = iγ µ ∂µ .

Heuristic Derivation Starting from the Klein-Gordon Equation

An obvious starting
p point for a Lorentz-invariant wave equation with only a first order time deriva-
tive is E = ± m + p~ 2 . Application of the correspondence principle (10.146) leads to the wave
2

equation p
i∂t Ψ(~r, t) = ± m2 − ∇2 Ψ(~r, t) . (10.208)
These two equation can be combined
 p  p 
i∂t + m2 − ∇2 i∂t − m2 − ∇2 Ψ(~r, t) (10.209)
10.7: Dirac Equation 315

which, in fact, is identical to the two equations (10.208). Equations (10.208, 10.209), however,
are unsatisfactory since expansion of the square root operator involves all powers of the Laplace
operator, but not an operator i~γ · ∇ as suggested by the principle of relativity (equivalence of
space and time). Many attempts were made by theoretical physicists to ‘linearize’ the square root
operator in (10.208, 10.209), but for a long time to no avail. Finally, Dirac succeeded. His solution
to the problem involved an ingenious step, namely, the realization that the linearization can be
carried out only if one assumes a 4-dimensional representation of the coefficients γ µ .
Initially, it was assumed that the 4-dimensional space introduced by Dirac could be linked to 4-
vectors, i.e., quantities with the transformation law (10.66). However, this was not so. Instead,
the 4-dimensionsional representation discovered by Dirac involved new physical properties, spin- 12
and anti-particles. The discovery by Dirac, achieved through a beautiful mathematical theory,
strengthens the believe of many theoretical physicists today that the properties of physical matter
ultimately derive from a, yet to be discovered, beautiful mathematical theory and that, therefore,
one route to important discoveries in physics is the creation of new mathematical descriptions of
nature, these descriptions ultimately merging with the true theory of matter.

Properties of the Dirac Matrices

Let us now trace Dirac’s steps in achieving the linearization of the ‘square root operator’ in (10.208).
Starting point is to boldly factorize, according to (10.209), the operator of the Klein–Gordon
equation
∂µ ∂ µ + m2 = −( P + m ) ( P − m ) (10.210)
where
P = iγ µ ∂µ . (10.211)
Obviously, this would lead to the two wave equations (P − m)Ψ = 0 and (P + m)Ψ = 0 which
have a first order time derivative and, therefore, are associated with a positive-definite particle
density. We seek to identify the coefficients γ µ . Inserting (10.211) into (10.210) yields
−g µν ∂µ ∂ν − m2 = (iγ µ ∂µ + m)(iγ µ ∂µ − m)
1
= −γ µ γ ν ∂µ ∂ν − m2 = − ( γ µ γ ν ∂µ ∂ν + γ ν γ µ ∂ν ∂µ ) − m2
2
1 µ ν
= − ( γ γ + γ γ ) ∂µ ∂ν − m2
ν µ
(10.212)
2
where we have changed ‘dummy’ summation indices, exploited ∂µ ∂ν = ∂ν ∂µ , but did not commute
the, so far, unspecified algebraic objects γ µ and γ ν . Comparing the left-most and the right-most
side of the equations above one can conclude the following property of γµ
γ µ γ ν + γ ν γ µ = [ γ µ , γ ν ]+ = 2 g µν (10.213)
We want to determine now the simplest algebraic realization of γ µ . It turns out that no 4-vector of
real or complex coefficients can satisfy these conditions. In fact, the quantities γ 0 , γ 1 , γ 2 , γ 3 can
only be realized by d×d–matrices requiring that the wave function Ψ(xµ ) is actually a d–dimensional
vector of functions ψ1 (xµ ), ψ2 (xµ ), . . . ψd (xµ ).
For µ = ν condition (10.213) reads

µ 2 1 µ = 0
(γ ) = . (10.214)
−1 µ = 1, 2, 3
316 Relativistic Quantum Mechanics

From this follows that γ 0 has real eigenvalues ±1 and γ j , j = 1, 2, 3 has imaginary eigenvalues ±i.
Accordingly, one can impose the condition

γ0 is hermitian ; γ j , j = 1, 2, 3 are anti-hermitian . (10.215)

For µ 6= ν (10.213) reads

γ µ γ ν = −γ ν γ µ , (10.216)
i.e., the γ µ are anti-commuting. From this one can conclude for the determinants of γ µ

det(γ µ γ ν ) = det(−γ ν γ µ ) = (−1)d det(γ ν γ µ ) = (−1)d det(γ µ γ ν ) . (10.217)

Obviously, as long as det(γ µ ) 6= 0 the dimension d of the square matrices γ µ must be even so that
(−1)d = 1.
For d = 2 there exist only three anti-commuting matrices, namely the Pauli matrices σ 1 , σ 2 , σ 3 for
which, in fact, holds

2
σj = 11 ; σ j σ k = − σ k σ j for j 6= k . (10.218)
The Pauli matrices allow one, however, to construct four matrices γ µ for the next possible dimension
d = 4. A proper choice is

σj
   
0 11 0 j 0
γ = ; γ = , (10.219)
0 −11 −σ j 0

Using property (10.218) of the Pauli matrices one can readily prove that condition (10.213) is
satisfied. We will argue further below that the choice f γ µ , except for similarity trasnformations, is
unique.

The Dirac Equation

Altogether we have shown that the Klein–Gordon equation can be factorized formally

( iγ µ ∂µ + m ) ( iγ µ ∂µ − m ) Ψ(xµ ) = 0 (10.220)

where Ψ(xµ ) represents a 4-dimensional wave function, rather than a scalar wave function. From
this equation one can conclude that also the following should hold

( iγ µ ∂µ − m ) Ψ(xµ ) = 0 (10.221)

which is the celebrated Dirac equation.

The Adjoint Dirac Equation

The adjoint equation is  
←
† µ µ †
Ψ (x ) i(γ ) ∂µ +m =0 (10.222)

←
where we have defined Ψ† = (ψ1∗ , ψ2∗ , ψ3∗ , ψ4∗ ) and where ∂µ denotes the differential operator ∂ µ
operating to the left side, rather than to the right side. One can readily show using the hermitian
10.7: Dirac Equation 317

property of the Pauli matrices (γ 0 )† = γ 0 and (γ j )† = −γ j for j = 1, 2, 3 which, in fact, is implied

by (10.215). This property can also be written

(γ µ )† = γ 0 γ µ γ 0 . (10.223)

Inserting this into (10.222) and multiplication from the right by γ 0 yields the adjoint Dirac equation
 
←
† µ 0 µ
Ψ (x ) γ iγ ∂µ + m = 0 . (10.224)

Similarity Transformations of the Dirac Equation - Chiral Representation

The Dirac equation can be subject to any similarity transformation defined through a non-singular
4 × 4–matrix S. Defining a new representation of the wave function Ψ̃(xµ )

Ψ̃(xµ ) = S Ψ(xµ ) (10.225)

leads to the ‘new’ Dirac equation

( iγ̃ µ ∂µ − m ) Ψ̃(xµ ) = 0 (10.226)

where
γ̃ µ = S γ µ S −1 (10.227)
A representation often adopted beside the one given by (10.222, 10.219) is the socalled chiral
representation defined through
 
µ µ 1 11 11
Ψ̃(x ) = S Ψ(x ) ; S = √ (10.228)
2 11 −11

and
−σ j
   
0 0 11 j 0
γ̃ = ; γ̃ = , j = 1, 2, 3 . (10.229)
11 0 σj 0
The similarity transformation (10.227) leaves the algebra of the Dirac matrices unaffected and
commutation property (10.213) still holds, i.e.,

[γ̃ µ , γ̃ ν ]+ = 2 g µν . (10.230)

Exercise 10.7.1: Derive (refeq:Dirac-intro20a) from (10.213), (10.227).

Schrödinger Form of the Dirac Equation

Another form in which the Dirac equation is used often results from multiplying (10.221) from the
left by γ 0  
~ˆ · ∇ − β̂ m Ψ(~r, t) = 0
i ∂t + i α (10.231)
318 Relativistic Quantum Mechanics

~ˆ has the three components αj , j = 1, 2, 3 and

whereα
σj
   
11 0 0
β̂ = ; α̂j = , j = 1, 2, 3 . (10.232)
0 −11 σj 0
This form of the Dirac equation is called the Schrödinger form since it can be written in analogy
to the time-dependent Schrödinger equation
i∂t Ψ(xµ ) = Ho Ψ(xµ ) ; ~ˆ · p
Ho = α ˆ~ + β̂ m . (10.233)
The eigenstates and eigenvalues of H correspond to the stationary states and energies of the particles
described by the Dirac equation.

Clifford Algebra and Dirac Matrices

The matrices defined through
dj = iγ j , j = 1, 2, 3 ; d4 = γ 0 (10.234)
satisfy the anti-commutation property

2 for j = k
dj dk + dk dj = (10.235)
0 for j 6= k
as can be readily verified from (10.213). The associative algebra generated by d1 . . . d4 is called a
Clifford algebra C4 . The three Pauli matrices also obey the property (10.235) and, hence, form a
Clifford algebra C3 . The representations of Clifford algebras Cm are well established. For example,
in case of C4 , a representation of the dj ’s is
       
0 1 N 1 0 1 0 N 0 1
d1 = ; d2 =
1 0 0 1 0 −1 1 0
       
0 i N 1 0 1 0 N 0 i
d3 = ; d4 = (10.236)
−i 0 0 1 0 −1 −i 0
where ‘⊗’ denotes the Kronecker product between matrices, i.e., the matrix elements of C = A ⊗ B
are Cjk,`m = Aj` Bkm .
The Clifford algebra C4 entails a subgroup G4 of elements
± dj1 dj2 · · · djs , j1 < j2 < · · · < js s ≤ 4 (10.237)
which are the ordered products of the operators ±11 and d1 , d2 , d3 , d4 . Obviously, any product
of the dj ’s can be brought to the form (10.237) by means of the property (10.235). There are
(including the different signs) 32 elements in G4 which we define as follows
Γ±1 = ±11
Γ±2 = ±d1 , Γ±3 = ±d2 , Γ±4 = ±d3 , Γ±5 = ±d4
Γ±6 = ±d1 d2 , Γ±7 = ±d1 d3 , Γ±8 = ±d1 d4 , Γ±9 = ±d2 d3
Γ±10 = ±d2 d4 , Γ±11 = ±d3 d4
Γ±12 = ±d1 d2 d3 , Γ±13 = ±d1 d2 d4 , Γ±14 = ±d1 d3 d4 , Γ±15 = ±d2 d3 d4
Γ±16 = ±d1 d2 d3 d4 (10.238)
10.8: Lorentz Invariance of the Dirac Equation 319

These elements form a group since obviously any product of two Γr ’s can be expressed in terms
of a third Γr . The representations of this group are given by a set of 32 4 × 4–matrices which
are equivalent with respect to similarity transformations. Since the Γj are hermitian the similarity
transformations are actually given in terms of unitary transformations. One can conclude then that
also any set of 4 × 4–matrices obeying (10.235) can differ only with respect to unitary similarity
transformations. This property extends then to 4 × 4–matrices which obey (10.213), i.e., to Dirac
matrices.
To complete the proof in this section the reader may consult Miller ‘Symmetry Groups and their
Application’ Chapter 9.6 and R.H.Good, Rev.Mod.Phys. 27, (1955), page 187. The reader may
also want to establish the unitary transformation which connects the Dirac matrices in the form
(10.236) with the Dirac representation (10.219).

Exercise 7.2:
Demonstrate the anti-commutation relationships (10.218) of the Pauli matrices σ j .
Exercise 7.3:
Demonstrate the anti-commutation relationships (10.218) of the Dirac matrices γ µ .
Exercise 7.4:
Show that from (10.214) follows that γ 0 has real eigenvalues ±1 and can be represented by a
hermitian matrix, and γ j , j = 1, 2, 3 has imaginary eigenvalues ±i and can be represented by an
anti-hermitian matrix.

10.8 Lorentz Invariance of the Dirac Equation

We want to show now that the Dirac equation is invariant under Lorentz transformations, i.e., the
form of the Dirac equation is identical in equivalent frames of reference, i.e., in frames connected by
Lorentz transformations. The latter transformations imply that coordinates transform according
to (10.6), i.e., x0 = Lµ ν xν , and derivatives according to (10.80). Multiplication and summation
of (10.80) by Lµ ρ and using (10.76) yields ∂ρ = Lν ρ ∂ 0 ν , a result one could have also obtained by
applying the chain rule to (10.6). We can, therefore, transform coordinates and derivatives of the
Dirac equation. However, we do not know yet how to transform the 4-dimensional wave function
Ψ and the Dirac matrices γ µ .

Lorentz Transformation of the Bispinor State

Actually, we will approach the proof of the Lorentz invariance of the Dirac equation by testing
if there exists a transformation of the bispinor wave function Ψ and of the Dirac matrices γ µ
which together with the transformations of coordinates and derivatives leaves the form of the Dirac
equation invariant, i.e., in a moving frame should hold
µ µ
iγ 0 ∂µ0 − m Ψ0 (x0 ) = 0 .


(10.239)
Form invariance implies that the matrices γ 0 µ should have the same properties as γ µ , namely,
(10.213, 10.215). Except for a similarity transformation, these properties determine the matrices
γ 0 µ uniquely, i.e., it must hold γ 0 µ = γ µ . Hence, in a moving frame holds
µ
iγ µ ∂µ0 − m Ψ0 (x0 ) = 0 .


(10.240)
320 Relativistic Quantum Mechanics

Infinitesimal Bispinor State Transformation

We want to show now that a suitable transformation of Ψ(xµ ) does, in fact, exist. The transfor-
mation is assumed to be linear and of the form
µ µ
Ψ0 (x0 ) = S(Lµ ν ) Ψ(xµ ) ; x0 = L µ ν xν (10.241)

where S(Lµ ν ) is a non-singular 4 × 4–matrix, the coefficients of which depend on the matrix Lµ ν
defining the Lorentz transformation in such a way that S(Lµ ν ) = 11 for Lµ ν = δ µ ν holds. Ob-
viously, the transformation (10.241) implies a similarity transformation Sγ µ S −1 . One can, hence,
state that the Dirac equation (10.221) upon Lorentz transformation yields
µ
iS(Lη ξ )γ µ S −1 (Lη ξ ) Lν µ ∂ν0 − m Ψ0 (x0 ) = 0 .


(10.242)

The form invariance of the Dirac equation under this transformation implies then the condition

S(Lη ξ )γ µ S −1 (Lη ξ ) Lν µ = γ ν . (10.243)

We want to determine now the 4 × 4–matrix S(Lη ξ ) which satisfies this condition.
The proper starting point for a constructiuon of S(Lη ξ ) is actually (10.243) in a form in which the
Lorentz transformation in the form Lµν is on the r.h.s. of the equation. For this purpose we exploit
(10.12) in the form Lν µ gνσ Lσ ρ = gµρ = gρµ . Multiplication of (10.243) from the left by Lσ ρ gσν
yields
S(Lη ξ )γ µ S −1 (Lη ξ ) gρµ = Lσ ρ gσν γσ . (10.244)
from which, using gρµ γ µ = γρ , follows

S(Lη ξ )γρ S −1 (Lη ξ ) = Lσ ρ γσ . (10.245)

One can finally conclude multiplying both sides by g ρµ

S(Lη ξ )γ µ S −1 (Lη ξ ) = Lν µ γν . (10.246)

The construction of S(Lη ξ ) will proceed using the avenue of infinitesimal transformations. We had
introduced in (10.38) the infinitesimal Lorentz transformations in the form Lµ ν = δ µ ν + µ ν where
the infinitesimal operator µ ν obeyed T = − gg. Multiplication of this property by g from the
right yields (g)T = − g, i.e., g is an anti-symmetric matrix. The elements of g are, however,
µ ρ g ρν = µν and, hence, in the expression of the infinitesimal transformation

Lµν = g µν + µν (10.247)

the infinitesimal matrix µν is anti-symmetric.

The infinitesimal transformation S(Lρ σ ) which corresponds to (10.247) can be expanded
i
S(µν ) = 11 − σµν µν (10.248)
4
Here σµν denote 4 × 4–matrices operating in the 4-dimensional space of the wave functions Ψ.
S(µν ) should not change its value if one replaces in its argument µν by −νµ . It holds then
i i
S(µν ) = 11 − σµν µν = S(−νµ ) = 11 + σµν νµ (10.249)
4 4
10.8: Lorentz Invariance of the Dirac Equation 321

from which we can conclude σµν µν = −σµν νµ = −σνµ µν , i.e., it must hold

σµν = −σνµ . (10.250)

One can readily show expanding SS −1 = 11 to first order in µν that for the inverse infinitesimal
transformation holds
i
S −1 (µν ) = 11 + σµν µν (10.251)
4
Inserting (10.248, 10.251) into (10.246) results then in a condition for the generators σµν
i
− ( σαβ γ µ − γ µ σαβ ) αβ = νµ γν . (10.252)
4
Since six of the coefficients αβ can be chosen independently, this condition can actually be expressed
through six independent conditions. For this purpose one needs to express formally the r.h.s. of
(10.252) also as a sum over both indices of αβ . Furthermore, the expression on the r.h.s., like the
expression on the l.h.s., must be symmetric with respect to interchange of the indices α and β. For
this purpose we express
1 αµ 1 1 1
νµ γν =  γα + βµ γβ = αβ δ µ β γα + αβ δ µ α γβ
2 2 2 2
1 αβ µ
=  ( δ β γα − δ µ α γβ ) . (10.253)
2
Comparing this with the l.h.s. of (10.252) results in the condition for each α, β

[ σαβ , γ µ ]− = 2i ( δ µ β γα − δ µ α γβ ) . (10.254)

The proper σαβ must be anti-symmetric in the indices α, β and operate in the same space as the
Dirac matrices. In fact, a solution of condition (10.254) is
i
σαβ = [ γα , γβ ]− (10.255)
2
which can be demonstrated using the properties (10.213, 10.216) of the Dirac matrices.

Exercise 7.5:
Show that the σαβ defined through (10.255) satisfy condition (10.254).

Algebra of Generators of Bispinor Transformation

We want to construct the bispinor Lorentz transformation by exponentiating the generators σµν .
For this purpose we need to verify that the algebra of the generators involving addition and mul-
tiplication is closed. For this purpose we inspect the properties of the generators in a particular
representation, namely, the chiral representation introduced above in Eqs. (10.228, 10.229). In this
representation the Dirac matrices γ̃µ = (γ̃ 0 , −~γ̃) are

σj
   
0 11 0
γ̃0 = ; γ̃j = , j = 1, 2, 3 . (10.256)
11 0 −σ j 0
322 Relativistic Quantum Mechanics

One can readily verify that the non-vanishing generators σ̃µν are given by (note σ̃µν = −σ̃νµ , i.e.
only six generators need to be determined)
−iσ j 0
   ` 
i σ 0
σ̃0j = [γ̃0 , γ̃j ] = ; σ̃jk = [γ̃j , γ̃k ] = jk` . (10.257)
2 0 iσ j 0 σ`
Obviously, the algebra of these generators is closed under addition and multiplication, since both
operations convert block-diagonal operators
 
A 0
(10.258)
0 B
again into block-diagonal operators, and since the algebra of the Pauli matrices is closed.
We can finally note that the closedness of the algebra of the generators σµν is not affected by
similarity transformations and that, therefore, any representation of the generators, in particular,
the representation (10.255) yields a closed algebra.

Finite Bispinor Transformation

The closedness of the algebra of the generators σµν defined through (10.248) allows us to write the
transformation S for any, i.e., not necessarily infinitesimal, µν in the exponential form
 
i µν
S = exp − σµν  . (10.259)
4
We had stated before that the transformation S is actually determined through the Lorentz trans-
formation Lµ ν . One should, therefore, be able to state S in terms of the same parameters w
~ and
~ µν
ϑ as the Lorentz transformation in (10.51). In fact, one can express the tensor  through w~ and
~ using µν = µ ρ g ρν and the expression (10.44)
ϑ
 
0 −w1 −w2 −w3
 w1 0 ϑ3 −ϑ2 
µν =  w2 −ϑ3
 (10.260)
0 ϑ1 
w3 ϑ2 −ϑ1 0
Inserting this into (10.259) yields the desired connection between the Lorentz transformation (10.51)
and S.
In order to construct an explicit expression of S in terms of w ~ and ϑ~ we employ again the chiral
representation. In this representation holds
i
− 4i σ̃µν µν = − (σ̃01 01 + σ̃02 02 + σ̃03 03 + σ̃12 12 + σ̃13 13 + σ̃23 23 )
2 !
1 ~ − iϑ)
(w ~ · ~σ 0
= ~ · ~σ . (10.261)
2 0 −(w~ + iϑ)
We note that this operator is block-diagonal. Since such operator does not change its block-diagonal
form upon exponentiation the bispinor transformation (10.259) becomes in the chiral representation
!
1 ~ σ
~ − iϑ)·~
(w
~ = e 2 0
S̃(w,
~ ϑ) 1 ~ (10.262)
0 e− 2 (w~ + iϑ)·~σ
This expression allows one to transform according to (10.241) bispinor wave functions from one
frame of reference into another frame of reference.
10.8: Lorentz Invariance of the Dirac Equation 323

Current 4-Vector Associated with Dirac Equation

We like to derive now an expression for the current 4-vector j µ associated with the Dirac equation
which satisfies the conservation law
∂µ j µ = 0 . (10.263)
Starting point are the Dirac equation in the form (10.221) and the adjoint Dirac equation (10.224).
Multiplying (10.221) from the left by Ψ† (xµ )γ 0 , (10.224) from the right by Ψ(xµ ), and addition
yields  
←
Ψ† (xµ )γ 0 iγ µ ∂µ + iγ µ ∂µ Ψ(xµ ) = 0 . (10.264)

The last result can be written

∂µ Ψ† (xν )γ 0 γ µ Ψ(xν ) = 0 , (10.265)
i.e., the conservation law (10.263) does hold, in fact, for

j µ (xµ ) = (ρ, ~j) = Ψ† (xµ )γ 0 γ µ Ψ(xµ ) . (10.266)

The time-like component ρ of j µ

4
X
ρ(xµ ) = Ψ† (xµ )Ψ(xµ ) = |ψs (xµ )|2 (10.267)
s=1

has the desired property of being positive definite.

The conservation law (10.263) allows one to conclude that j µ must transform like a contravariant
4-vector as the notation implies. The reason is that the r.h.s. of (10.263) obviously is a scalar
under Lorentz transformations and that the left hand side must then also transform like a scalar.
Since ∂µ transforms like a covariant 4-vector, j µ must transform like a contravariant 4-vector. This
transformation behaviour can also be deduced from the transformation properties of the bispinor
wave function Ψ(xµ ). For this purpose we prove first the relationship

S −1 = γ 0 S † γ 0 . (10.268)

We will prove this property in the chiral representation. Obviously, the property applies then in
any representation of S.
For our proof we note first
!
− 21 (w ~ σ
~ − iϑ)·~
e 0
S̃ −1 (w, ~ = S̃(−w,
~ ϑ) ~ −ϑ)~ = 1 ~ (10.269)
0 e 2 (w~ + iϑ)·~σ

One can readily show that the same operator is obtained evaluating
!
  1
(w ~ σ
~ + iϑ)·~

0 † ~ 0 0 1
1 e 2 0 0 11
γ̃ S̃ (w,
~ ϑ)γ̃ = 1 ~ . (10.270)
11 0 0 e− 2 (w~ − iϑ)·~σ 11 0

We conclude that (10.268) holds for the bispinor Lorentz transformation.

We will now determine the relationship between the flux
µ † µ
j0 = Ψ0 (x0 )γ 0 γ µ Ψ0 (x0 ) (10.271)
324 Relativistic Quantum Mechanics

in a moving frame of reference and the flux j µ in a frame at rest. Note that we have assumed
in (10.271) that the Dirac matrices are independent of the frame of reference. One obtains using
(10.268)
µ
j 0 = Ψ† (xµ )S † γ 0 γ µ SΨ(xµ ) = Ψ† (xµ )γ 0 S −1 γ µ SΨ(xµ ) . (10.272)
With S −1 (Lη ξ ) = S((L−1 )η ξ ) one can restate (10.246)

µ
S −1 (Lη ξ )γ µ S ( Lη ξ ) = (L−1 )νµ γν = (L−1 )ν γ ν = Lµ ν γν . (10.273)

where we have employed (10.76). Combining this with (10.272) results in the expected transfor-
mation behaviour
µ
j 0 = Lµ ν j ν . (10.274)

10.9 Solutions of the Free Particle Dirac Equation

We want to determine now the wave functions of free particles described by the Dirac equation.
Like in non-relativistic quantum mechanics the free particle wave function plays a central role, not
only as the most simple demonstration of the theory, but also as providing a basis in which the wave
functions of interacting particle systems can be expanded and characterized. The solutions provide
also a complete, orthonormal basis and allows one to quantize the Dirac field Ψ(xµ ) much like the
classical electromagnetic field is quantized through creation and annihilation operators representing
free electromagnetic waves of fixed momentum and frequency.
In case of non-relativistic quantum mechanics the free particle wave function has a single component
ψ(~r, t) and is determined through the momentum p~ ∈ R3 . In relativistic quantum mechanics a
Dirac particle can also be characterized through a momentum, however, the wave function has four
components which invite further characterization of the free particle state. In the following we
want to provide this characterization, specific for the Dirac free particle.
We will start from the Dirac equation in the Schrödinger form (10.231, 10.232, 10.233)

i∂t Ψ(xµ ) = Ho Ψ(xµ ) . (10.275)

The free particle wave function is an eigenfunction of Ho , a property which leads to the energy–
momentum (dispersion) relationship of the Dirac particle. The additional degrees of freedom de-
scribed by the four components of the bispinor wave function require, as just mentioned, additional
characterizations, i.e., the identification of observables and their quantum mechanical operators, of
which the wave functions are eigenfunctions as well. As it turns out, only two degrees of freedom
of the bispinor four degrees of freedom are independent [c.f. (10.282, 10.283)]. The independent
degrees of freedom allow one to choose the states of the free Dirac particle as eigenstates of the
4-momentum operator p̂µ and of the helicity operator Γ ∼ ~σ · p ˆ~/|p
ˆ~| introduced below. These opera-
tors, as is required for the mentioned property, commute with each other. The operators commute
also with Ho in (10.233).
Like for the free particle wave functions of the non-relativistic Schrödinger and the Klein–Gordon
equations one expects that the space–time dependence is governed by a factor exp[i(~ p · ~r − t)].
As pointed out, the Dirac particles are described by 4-dimensional, bispinor wave functions and we
need to determine corresponding components of the wave function. For this purpose we consider
10.9: Solutions of the Free Particle Dirac Equation 325

the following form of the free Dirac particle wave function

φ(xµ )
   
φo
µ
Ψ(x ) = = ei(~p·~r − t) (10.276)
χ(xµ ) χo

where p~ and  together represent four real constants, later to be identified with momentum and
energy, and φo , χo each represent a constant, two-dimensional spinor state. Inserting (10.276) into
(10.231, 10.232) leads to the 4-dimensional eigenvalue problem
    
m ~σ · p~ φo φo
=  . (10.277)
~σ · p~ −m χo χo

To solve this problem we write (10.277) explicitly

( − m) 11 φo − ~σ · p~ χo = 0
− ~σ · p~ φo + ( + m) 11 χo = 0. (10.278)

Multiplication of the 1st equation by ( + m)11 and of the second equation by −~σ · p~ and subtraction
of the results yields the 2-dimensional equation
 2
( − m2 ) 11 − (~σ · p~)2 φo = 0 .

(10.279)

According to the property (5.234) of Pauli matrices holds (~σ · p~)2 = p~ 2 11. One can, hence, conclude
from (10.279) the well-known relativistic dispersion relationship

2 = m2 + p~ 2 (10.280)

which has a positive and a negative solution

p
 = ± E(~
p) , E(~
p) = m2 + p~ 2 . (10.281)

Obviously, the Dirac equation, like the Klein–Gordon equation, reproduce the classical relativistic
energy–momentum relationships (10.93, 10.94)
Equation (10.278) provides us with information about the components of the bispinor wave function
(10.276), namely φo and χo are related as follows
~σ · p~
φ0 = χ0 (10.282)
−m
~σ · p~
χo = φ0 , (10.283)
+m
where  is defined in (10.281). These two relationships are consistent with each other. In fact, one
finds using (5.234) and (10.280)

~σ · p~ (~σ · p~)2 p~ 2
φ0 = χ0 = 2 χ0 = χ0 . (10.284)
+m ( + m) ( + m)  − m2
The relationships (10.282, 10.283) imply that the bispinor part of the wave function allows only
two degrees of freedom to be chosen independently. We want to show now that these degrees of
freedom correspond to a spin-like property, the socalled helicity of the particle.
326 Relativistic Quantum Mechanics

For our further characterization we will deal with the positive and negative energy solutions [cf.
(10.281)] separately. For the positive energy solution, i.e., the solution for  = +E(~
p), we present
φo through the normalized vector
 
u1
φo = = u ∈ C2 , u† u = |u1 |2 + |u2 |2 = 1 . (10.285)
u2
The corresponding free Dirac particle is then described through the wave function
!
u
Ψ(~p, +|xµ ) = N+ (~p) σ ·~
~ p ei(~p·~r − t) ,  = +E(~
p) . (10.286)
p) + m u
E(~

Here N+ (~
p) is a constant which will be chosen to satisfy the normalization condition
Ψ† (~
p, +) γ 0 Ψ† (~
p, +) = 1 , (10.287)
the form of which will be justified further below. Similarly, we present the negative energy solution,
i.e., the solution for  = −E(~p), through χo given by
 
u1
χo = = u ∈ C2 , u† u = |u1 |2 + |u2 |2 = 1 . (10.288)
u2
corresponding to the wave function
!
−~ σ ·~
p
u
p, −|xµ ) = N− (~
Ψ(~ p) E(~
p) + m ei(~p·~r − t) ,  = −E(~
p) . (10.289)
u

Here N− (~
p) is a constant which will be chosen to satisfy the normalization condition
Ψ† (~
p, +) γ 0 Ψ† (~
p, +) = −1 , (10.290)
which differs from the normalization condition (10.287) in the minus sign on the r.h.s. The form
of this condition and of (10.287) will be justified now.
First, we demonstrate that the product Ψ† (~ p, ±)γ 0 Ψ(~p, ±), i.e., the l.h.s. of (10.287, 10.290), is
invariant under Lorentz transformations. One can see this as follows: Let Ψ(~ p, ±) denote the
solution of a free particle moving with momentum p~ in the laboratory frame, and let Ψ(0, ±)
denote the corresponding solution of a particle in its rest frame. The connection between the
solutions, according to (10.241), is Ψ(~p, ±) = S Ψ(0, ±) , where S is given by (10.259). Hence,
 
Ψ† (~
p, ±) γ 0 Ψ(~
p, ±) = Ψ† (0, ±) S † γ 0 S Ψ(0, ±)
= Ψ† (0, ±) γ 0 S −1 γ 0 γ 0 S Ψ(0, ±)

= Ψ† (0, ±) γ 0 Ψ(0, ±). (10.291)

Note that we have used that, according to (10.268), S −1 = γ 0 S † γ 0 and, hence, S † = γ 0 S −1 γ 0 .
We want to demonstrate now that the signs on the r.h.s. of (10.287, 10.290) should differ. For this
purpose we consider first the positive energy solution. Employing (10.286) for p~ = 0 yields, using
γ 0 as given in (10.219) and u† u = 1 [c.f. (10.285)],
 
2 u
† 0 †
Ψ (0, +) γ Ψ(0, +) = |N+ (0)| (u , 0) γ 0
= |N+ (0)|2 . (10.292)
0
10.9: Solutions of the Free Particle Dirac Equation 327

The same calculation for the negative energy wave function as given in (10.289) yields
 
2 0
† 0
Ψ (0, −) γ Ψ(0, −) = |N− (0)| (0, u ) γ† 0
= − |N− (0)|2 . (10.293)
u

Obviously, this requires the choice of a negative side on the r.h.s. of (10.290) to assign a positive
value to |N− (0)|2 . We can also conclude from our derivation

N± (0) = 1 . (10.294)

We want to determine now N± (~ p) for arbitrary p~. We consider first the positive energy solution.
Condition (10.287) written explicitly using (10.286) is
T ! !
u

~
σ · p
~
p) (u∗ )T ,
N+2 (~ u∗ γo σ ·~
~ p = 1 (10.295)
E(~
p) + m p) + m u
E(~

Evaluating the l.h.s. using γ 0 as given in (10.219) yields

(~σ · p~)2
 
2 † †
N+ (~p) u u − u u = 1. (10.296)
(E(~p) + m)2

Replacing (~σ · p~)2 by p~ 2 [c.f. (5.234)] and using the normalization of u in (10.285) results in

p~ 2
 
2
N+ (~
p) 1 − = 1 (10.297)
(E(~p) + m)2
from which follows s
( m + E(~p) )2
N+ (~
p) = . (10.298)
p) )2 − p~ 2
( m + E(~
Noting

p) )2 − p~ 2 = m2 − p~ 2 + 2mE(~
( m + E(~ p) + E 2 (~
p) = 2(m + E(~
p)) m (10.299)

the normalization coefficient (10.298) becomes

r
m + E(~
p)
N+ (~
p) = . (10.300)
2m
This result completes the expression for the wave function (10.286).

Exercise 7.6: Show that the normalization condition

T ! !
u

02 ∗ T ~σ · p~ ∗
N+ (~p) (u ) , u σ ·~
~ p = 1 (10.301)
E(~p) + m E(~
p) + m u

yields the normalization coefficient

s
m + E(~p)
N+0 (~
p) = . (10.302)
2 E(~
p)
328 Relativistic Quantum Mechanics

We consider now the negative energy solution. Condition (10.290) written explicitly using (10.289)
is T ! !
 −~σ ·~
p
2 −~σ · p~ ∗ ∗ T o E(~
p) + m u
N− (~
p) u , (u ) γ = −1 (10.303)
E(~p) + m u
Evaluating the l.h.s. yields
(~σ · p~)2
 
† †
N−2 (~
p) u u − u u = −1 . (10.304)
p) + m)2
(E(~
This condition is, however, identical to the condition (10.296) for the normalization constant N+ (~
p)
of the positive energy solution. We can, hence, conclude
r
m + E(~ p)
N− (~
p) = (10.305)
2m
and, thereby, have completed the determination for the wave function (10.289).
The wave functions (10.286, 10.289, 10.300) have been constructed to satisfy the free particle Dirac
equation (10.275). Inserting (10.286) into (10.275) yields

p, λ|xµ ) = λ E(~
Ho Ψ(~ p, λ|xµ ) ,
p) Ψ(~ (10.306)

i.e., the wave functions constructed represent eigenstates of Ho . The wave functions are also
eigenstates of the momentum operator i∂µ , i.e.,

p, λ|xµ ) = pµ Ψ(~
i∂µ Ψ(~ p, λ|xµ ) (10.307)

where pµ = (, −~ p, λ|xµ ) in 4-vector

p). This can be verified expressing the space–time factor of Ψ(~
p · ~r − t)] = exp(ipµ xµ ).
notation, i.e., exp[i(~

Helicity
The free Dirac particle wave functions (10.286, 10.289) are not completely specified, the two com-
ponents of u indicate another degree of freedom which needs to be defined. This degree of freedom
describes a spin– 21 attribute. This attribute is the so-called helicity, defined as the component of
the particle spin along the direction of motion. The corresponding operator which measures this
observable is
1 ~p̂
Λ = σ· . (10.308)
2 |~p̂|
Note that ~p̂ represents here an operator, not a constant vector. Rather than considering the
observable (10.307) we investigate first the observable due to the simpler operator ~σ · ~p̂. We
want to show that this operator commutes with Ho and ~p̂ to ascertain that the free particle wave
function can be simultaneously an eigenvector of all three operators. The commutation property
[~σ · ~p̂, p̂j ] = 0 , j = 1, 2, 3 is fairly obvious. The property [~σ · ~p̂, Ho ] = 0 follows from (10.233) and
from the two identities
      
11 0 ~σ 0 ~ ~σ 0 ~ 11 0
· p̂ − · p̂ = 0 (10.309)
0 −11 0 ~σ 0 ~σ 0 −11
10.9: Solutions of the Free Particle Dirac Equation 329

and
       
0 ~σ ~σ 0 ~σ 0 0 ~
σ
· ~p̂ · ~p̂ − · ~p̂ · ~p̂
~σ 0 0 ~σ 0 ~σ ~σ 0
! !
0 (~σ · ~p̂)2 0 (~σ · ~p̂)2
= − = 0. (10.310)
(~σ · ~p̂)2 0 (~σ · ~p̂)2 0

We have shown altogether that the operators ~p̂, Ho and ~σ ·~p̂ commute with each other and, hence, can
be simultaneously diagonal. States which are simultaneously eigenvectors of these three operators
are also simulteneously eigenvectors of the three operators ~p̂, Ho and Λ defined in (10.308) above.
The condition that the wave functions (10.286) are eigenfunctions of Λ as well will specify now the
vectors u.
Since helicity is defined relative to the direction of motion of a particle the characterization of u
as an eigenvector of the helicity operator, in principle, is independent of the direction of motion
of the particle. We consider first the simplest case that particles move along the x3 –direction, i.e.,
p~ = (0, 0, p3 ). In this case Λ = 21 σ 3 .
We assume first particles with positive energy, i.e.,  = +E(~ p). According to the definition (5.224)
of σ3 the two u vectors (1, 0)T and (0, 1)T are eigenstates of 12 σ 3 with eigenvalues ± 12 . Therefore,
the wave functions which are eigenstates of the helicity operator, are
   
1
 0   ei(px − Ep t)
3
Ψ(pê3 , +, + 21 |~r, t) = Np 
 
 p 1 
m + Ep 0
   
0
 1   ei(px − Ep t)
3
Ψ(pê3 , +, − 12 |~r, t) = Np 
 
(10.311)
 −p 0 
m + Ep 1

where ê3 denotes the unit vector in the x3 -direction and where
r
p
2 2
m + Ep
E p = m + p ; Np = . (10.312)
2m
We assume now particles with negative energy, i.e.,  = −E(~ p). The wave functions which are
eigenfunctions of the helicity operator are in this case
   
−p 1
m + Ep
 0   ei(px + Ep t)
3
Ψ(pê3 , −, + 21 |~r, t) = Np 
 
 1 
0
   
p 0
m + Ep
 1   ei(px + Ep t)
3
Ψ(pê3 , −, − 12 |~r, t) = Np 
 
(10.313)
 0 
1

where Ep and Np are defined in (10.312).

330 Relativistic Quantum Mechanics

To obtain free particle wave functions for arbitrary directions of p~ one can employ the wave functions
(10.311, 10.313) except that the states (1, 0)T and (0, 1)T have to be replaced by eigenstates u± (~ p)
of the spin operator along the direction of p~. These eigenstates are obtained through a rotational
transformation (5.222, 5.223) as follows
  
1 i ~ 1
u(~p, + ) = exp − ϑ(~ p) · ~σ , (10.314)
2 2 0
  
1 i ~ 0
u(~p, − ) = exp − ϑ(~ p) · ~σ (10.315)
2 2 1

where
~ p) = ê3 × p~ ∠(ê3 , p~)
ϑ(~ (10.316)
|~
p|
describes a rotation which aligns the x3 –axis with the direction of p~. [One can also express the
rotation through Euler angles α, β, γ , in which case the transformation is given by (5.220).] The
corresponding free particle wave functions are then

p, + 21 )
!
u(~
Ψ(~ 1
p, +, + 2 |~r, t) = Np p 1
ei(~p·~r − Ep t) (10.317)
m + Ep u(~
p , + 2 )

p, − 12 )
!
u(~
Ψ(~ 1
p, +, − 2 |~r, t) = Np −p 1
ei(~p·~r − Ep t) (10.318)
m + Ep u(~
p , − 2 )
−p 1
!
m + E u(~
p , + 2 )
Ψ(~p, −, + 12 |~r, t) = N√ p
ei(~p·~r + Ep t) (10.319)
p, + 12 )
u(~
p
p, − 12 )
!
m + Ep u(~
Ψ(~ 1
p, −, − 2 |~r, t) = Np 1
ei(~p·~r + Ep t) (10.320)
p, − 2 )
u(~

where Ep and N are again given by (10.312).

Generating Solutions Through Lorentz Transformation

The solutions (10.311, 10.312) can be obtained also by means of the Lorentz transformation (10.262)
for the bispinor wave function and the transformation (10.123). For this purpose one starts from
the solutions of the Dirac equation in the chiral representation (10.226, 10.229), denoted by ˜, for
an ~r–independent wave function, i.e., a wave function which represents free particles at rest. The
corresponding wave functions are determined through

iγ̃ 0 ∂t − m Ψ̃(t) = 0 .


(10.321)

and are
 
1
 0  −imt
Ψ̃(p = 0, +, 21 |t) = √12 
 1 e
 ,
0
10.9: Solutions of the Free Particle Dirac Equation 331
 
0
 1  −imt
Ψ̃(p = 0, +, − 12 |t) = √12 
 0 e
 ,
1
 
1
 0  +imt
Ψ̃(p = 0, −, 12 |t) = √12 
 −1  e
 ,
0
 
0
 1  +imt
Ψ̃(p = 0, −, − 12 |t) = √12 
 0 e
 . (10.322)
−1

The reader can readily verify that transformation of these solutions to the Dirac representationsas
defined in (10.228) yields the corresponding solutions (10.311, 10.313) in the p → 0 limit. This
correspondence justifies the characterization ±, ± 12 of the wave functions stated in (10.322).
The solutions (10.322) can be written in spinor form
     
1 φo ∓imt 1 0
√ e , φo , χo ∈ , (10.323)
2 χo 0 1

Transformation (10.262) for a boost in the x3 –direction, i.e., for w

~ = (0, 0, w3 ), yields for the
exponential space–time dependence according to (10.174, 10.176)

∓imt → i ( p3 x3 ∓ Et ) (10.324)

and for the bispinor part according to (10.262)

1 3
! 1 3
!
e 2 w3 σ e 2 w3 σ φo
  
φo 0 φo
→ 1 3 = 1 3 . (10.325)
χo 0 e− 2 w3 σ χo e− 2 w3 σ χo

One should note that φo , χo are eigenstates of σ 3 with eigenvalues ±1. Applying (10.324, 10.325)
to (10.323) should yield the solutions for non-vanishing momentum p in the x3 –direction. For the
resulting wave functions in the chiral representation one can use then a notation corresponding to
that adopted in (10.311)
  
1
w 1
e 2 3
1  0 3
Ψ̃(p(w3 )ê3 , +, + 21 |~r, t)    ei(px − Ep t)

= √ 
2  1 1 
e− 2 w3
0
   
1 0
e− 2 w3
1  1 3
Ψ̃(p(w3 )ê3 , +, − 21 |~r, t)    ei(px − Ep t)

= √ 
2  1 0 
e 2 w3
1
332 Relativistic Quantum Mechanics
   
1
w 1
e2 3
1  0 3
Ψ̃(p(w3 )ê3 , −, + 12 |~r, t)    ei(px + Ep t)

= √ 
2  1 1 
−e− 2 w3
0
   
1 0
e− 2 w3
1  1 3
Ψ̃(p(w3 )ê3 , −, − 12 |~r, t)    ei(px + Ep t)

= √  (10.326)
2  1 0 
− e 2 w3
1

where according to (10.61) p(w3 ) = m sinhw3 . Transformation to the Dirac representation by

means of (10.228) yields
   
w3 1
cosh 2
 0   ei(px − Ep t)
3
Ψ(p(w3 )ê3 , +, + 21 |~r, t) = 
 
1
sinh w23
 
0
   
0
cosh w23
 1   ei(px − Ep t)
3
Ψ(p(w3 )ê3 , +, − 21 |~r, t) = 
 
0
− sinh w23
 
1
   
1
sinh w23
 0   ei(px + Ep t)
3
Ψ(p(w3 )ê3 , −, + 21 |~r, t) = 
 
1
cosh w23
 
0
   
w3 0
− sinh 2
 1   ei(px + Ep t)
3
Ψ(p(w3 )ê3 , −, − 21 |~r, t) = 
 
(10.327)
0
cosh w23
 
1

Employing the hyperbolic function properties

r r
x coshx + 1 x coshx − 1
cosh = , sinh = , (10.328)
2 2 2 2
the relationship (10.61) between the parameter w3 and boost velocity v3 , and the expression (10.311)
for Ep one obtains
vs vs
u u
1 u 1 1 u v32
cosh w23 = √ t
2 +1 = √ t 1+ +1
2 1 − v3 2 1 − v32
vr
u
2 2
u m2 + m v32 + m
u r
t 1 − v3 Ep + m
= = (10.329)
2m 2m
10.9: Invariance of Dirac Equation Revisited 333

and similarly r
w3 Ep − m p
sinh = = p (10.330)
2 2m 2m (Ep + m)
Inserting expressions (10.329, 10.330) into (10.327), indeed, reproduces the positive energy wave
functions (10.311) as well as the negative energy solutions (10.313) for −p. The change of sign for
the latter solutions had to be expected as it was already noted for the negative energy solutions of
the Klein–Gordon equation (10.168–10.176).

Invariance of Dirac Equation Revisited

At this point we like to provide a variation of the derivation of (10.243), the essential property
stating the Lorentz–invariance of the Dirac equation. Actually, we will derive this equation only
for infinitesimal transformations, which however, is sufficient since (1) it must hold then for any
finite transformation, and since (2) the calculations following (10.243) considered solely the limit
of infinitesimal transformations anyway.
The reason why we provide another derivation of (10.243) is to familiarize ourselves with a formu-
lation of Lorentz transformations of the bispinor wave finction Ψ(xµ ) which treats the spinor and
the space-time part of the wave function on the same footing. Such description will be essential for
the formal description of Lorentz invariant wave equations for arbitray spin further below.
In the new derivation we consider the particle described by the wave function transformed, but not
the observer. This transformation, refered to as the active transformation, expresses the system in
the old coordinates. The transformation is

Ψ0 (xµ ) = S(Lη ξ ) ρ(Lη ξ ) Ψ(xµ ) (10.331)

where S(Lη ξ ) denotes again the transformation acting on the bispinor character of the wave function
Ψ(xµ ) and where ρ(Lη ξ ) denotes the transformation acting on the space-time character of the
wave function Ψ(xµ ). ρ(Lη ξ ) has been defined in (10.123) above and characterized there. Such
transformation had been applied by us, of course, when we generated the solutions Ψ(~ p, λ, Λ|xµ )
from the solutions describing particles at rest Ψ(~ p = 0, λ, Λ|t). We expect, in general, that if Ψ(xµ )
0 µ
is a solution of the Dirac equation that Ψ (x ) as given in (10.331) is a solution as well. Making
this expectation a postulate allows one to derive the condition (10.243) and, thereby, the proper
transformation S(Lη ξ ).
To show this we rewrite the Dirac equation (10.221) using (10.331)

i S(Lη ξ )γ µ S −1 (Lη ξ ) ρ(Lη ξ )∂µ ρ−1 (Lη ξ ) − m Ψ0 (xµ ) = 0

(10.332)

Here we have made use of the fact that S(Lη ξ ) commutes with ∂µ and ρ(Lη ξ ) commutes with γ µ .
The fact that any such Ψ0 (xµ ) is a solution of the Dirac equation allows us to conclude

S(Lη ξ )γ µ S −1 (Lη ξ ) ρ(Lη ξ )∂µ ρ−1 (Lη ξ ) = γ ν ∂ν (10.333)

which is satisfied in case that the following conditions are met

ρ(Lη ξ )∂µ ρ−1 (Lη ξ ) = Lν µ ∂ν ;

S(Lη ξ )γ µ S −1 (Lη ξ ) Lν µ = γν . (10.334)
334 Relativistic Quantum Mechanics

We will demonstrate now that the first condition is satisfied by ρ(Lη ξ ). The second condition is
identical to (10.243) and, of course, it is met by S(Lη ξ ) as given in the chiral representation by
(10.262).
As mentioned already we will show condition (10.334) for infinitesimal Lorentz transformations
Lη ξ . We will proceed by employing the generators (10.128) to express ρ(Lη ξ ) in its infinitesimal
form and evaluate the expression
   
~ · J~ + w
11 + ϑ ~ ∂µ 11 − ϑ
~ ·K ~ · J~ − w
~ ·K~

= ∂µ + M ν µ ∂ν + O(2 ) (10.335)
The result will show that the matrix M ν µ is identical to the generators of Lν µ for the six choices ϑ ~ =
(1, 0, 0), w ~
~ = (0, 0, 0), ϑ = (0, 1, 0), w
~ = (0, 0, 0), . . . , ϑ = (0, 0, 0), w
~ = (0, 0, 1) . Inspection of
(10.335) shows that we need to demonstrate
[J` , ∂µ ] = (J` )ν µ ∂ν ; [K` , ∂µ ] = (K` )ν µ ∂ν . (10.336)
We will proceed with this task considering all six cases:


 0 µ = 0
0 µ = 1

[J1 , ∂µ ] = [x3 ∂2 − x2 ∂3 , ∂µ ] = (10.337)

 ∂3 µ = 2
−∂2 µ = 3



 0 µ = 0
−∂3 µ = 1

[J2 , ∂µ ] = [x1 ∂3 − x3 ∂1 , ∂µ ] = (10.338)

 0 µ = 2
∂1 µ = 3



 0 µ = 0
∂2 µ = 1

[J3 , ∂µ ] = [x2 ∂1 − x1 ∂2 , ∂µ ] = (10.339)

 −∂1 µ = 2
0 µ = 3



 −∂1 µ = 0
−∂0 µ = 1

[K1 , ∂µ ] = [x0 ∂1 + x1 ∂0 , ∂µ ] = (10.340)

 0 µ = 2
0 µ = 3



 −∂2 µ = 0
0 µ = 1

[K2 , ∂µ ] = [x0 ∂2 + x2 ∂0 , ∂µ ] = (10.341)
 −∂0
 µ = 2
0 µ = 3



 −∂3 µ = 0
0 µ = 1

[K3 , ∂µ ] = [x0 ∂3 + x3 ∂0 , ∂µ ] = (10.342)

 0 µ = 2
−∂0 µ = 3


One can readily convince oneself that these results are consistent with (10.336). We have demon-
strated, therefore, that any solution Ψ(xµ ) transformed according to (10.331) is again a solution of
the Dirac equation, i.e., the Dirac equation is invariant under active Lorentz transformations.
10.10: Dirac Particles in Electromagnetic Field 335

10.10 Dirac Particles in Electromagnetic Field

We like to provide now a description for particles governed by the Dirac equation which includes
the coupling to an electromagnetic field in the minimum coupling description. Following the re-
spective procedure developed for the Klein-Gordon equation in Sect. 10.6 we assume that the field
is described through the 4-vector potential Aµ and, accordingly, we replace in the Dirac equation
the momentum operator p̂µ = i∂µ by i∂µ − qAµ where q is the charge of the respective particles
(see Table 10.1 in Sect 10.6 above). Equivalently, we replace the operator ∂µ by ∂µ + iqAµ . The
Dirac equation (10.221) reads then

[ iγ µ (∂µ + iqAµ ) − m ] Ψ(xν ) = 0 (10.343)

One may also include the electromagnetic field in the Dirac equation given in the Schrödinger form
(10.233) by replacing i∂t by (see Table 10.1) i∂t − qV and pˆ~ by

ˆ = p
~π ˆ~ − q A
~. (10.344)

The Dirac equation in the Schrödinger form reads then

 
i∂t Ψ(xµ ) = α ~ˆ · ~π
ˆ + qV + β̂ m Ψ(xµ ) (10.345)

~ˆ and β̂ are defined in (10.232).

where α

Non-Relativistic Limit
We want to consider now the Dirac equation (10.345) in the so-called non-relativistic limit in which
all energies are much smaller than m, e.g., for the scalar field V in (10.345) holds

|qV | << m . (10.346)

For this purpose we choose the decomposition

φ(xµ )
 
µ
Ψ(x ) = . (10.347)
χ(xµ )

Using the notation ~σ = (σ1 , σ2 , σ3 )T one obtains then

i∂t φ = ˆ χ + qV φ + mφ
~σ · ~π (10.348)
i∂t χ = ˆ φ + qV χ − mχ .
~σ · ~π (10.349)

We want to focus on the stationary positive energy solution. This solution exhibits a time-
dependence exp[−i(m + )t] where for  holds in the non-relativistic limit || << m. Accordingly,
we define

φ(xµ ) = e−imt Φ(xµ ) (10.350)

µ −imt µ
χ(x ) = e X (x ) (10.351)
336 Relativistic Quantum Mechanics

and assume that for the time-derivative of Φ and X holds

   
 ∂t Φ   ∂t X 
 Φ  << m ,  X  << m . (10.352)
   

Using (10.350, 10.351) in (10.348, 10.349) yields

i∂t Φ = ˆ X + qV Φ
~σ · ~π (10.353)
i∂t X = ˆ Φ + qV X − 2mX .
~σ · ~π (10.354)

The properties (10.346, 10.352) allow one to approximate (10.354)

ˆ Φ − 2m X .
0 ≈ ~σ · ~π (10.355)

and, accordingly, one can replace X in (10.353) by

ˆ
~σ · ~π
X ≈ Φ (10.356)
2m
to obtain a closed equation for Φ
 2
ˆ
~σ · ~π
i∂t Φ ≈ Φ + qV Φ . (10.357)
2m
Equation (10.356), due to the m−1 factor, identifies X as the small component of the bi-spinor
wave function which, henceforth, does not need to be considered anymore.
Equation (10.357) for Φ can be reformulated by expansion of (~σ · ~πˆ )2 . For this purpose we employ
the identity (5.230), derived in Sect. 5.7, which in the present case states

(~σ · ~π ˆ 2 + i ~σ · (~π
ˆ )2 = ~π ˆ × ~π
ˆ) . (10.358)

ˆ × ~π
For the components of ~π ˆ holds
 
ˆ × ~π
~π ˆ = jk` ( πj πk − πk πj ) = jk` [πj , π` ] . (10.359)
`

We want to evaluate the latter commutator. One obtains

1 1
[πj , πk ] = [ ∂j + qAj , ∂k + qAk ]
i i
1 1 1 1
= [ ∂j , ∂k ] + q [Aj , ∂k ] + q [ ∂j , Ak ] + q 2 [Aj , Ak ]
|i {zi } i i | {z }
=0
=0
q q
= [Aj , ∂k ] + [∂j , Ak ] . (10.360)
i i
For an arbitrary function f (~r) holds

( [Aj , ∂k ] + [∂j , Ak ] ) f = ( ∂j Ak − Ak ∂j + Aj ∂k − ∂k Aj ) f . (10.361)

10.10: Dirac Particles in Electromagnetic Field 337

Using

∂j Ak f = ((∂j Ak )) f + Ak ∂j f
∂k Aj f = ((∂k Aj )) f + Aj ∂k f

where ((∂j · · ·)) denotes confinement of the differential operator to within the brackets ((· · ·)), one
obtains
( [Aj , ∂k ] + [∂j , Ak ] ) f = [ ((∂j Ak )) − ((∂k Aj )) ] f (10.362)
~
or, using (10.360) and Aµ = (V, −A),
q q  ~ jk` = − q B` jk`

[πj , πk ] = (( ∂j Ak − ∂k Aj )) = − ∇×A (10.363)
i i ` i
where we employed B(~ ~ r, t) = ∇ × A(~ ~ r, t) [see (8.6)]. Equations (10.344, 10.358, 10.359, 10.363)
allow us to write (10.357) in the final form
" #
ˆ~ − q A(~
[p ~ r, t)]2 q
i∂t Φ(~r, t) ≈ − ~ r, t) + q V (~r, t) Φ(~r, t)
~σ · B(~ (10.364)
2m 2m

which is referred to as the Pauli equation.

Comparision of (10.364) governing a two-dimensional wave function Φ ∈ C2 with the corresponding
non-relativistic Schrödinger equation (10.2) governing a one-dimensional wave function ψ ∈ C,
reveals a stunning feature: the Pauli equation does justice to its two-dimensional character; while
agreeing in all other respects with the non-relativistic Schrödinger equation (10.2) it introduces the
~ Φ which describes the well-known interaction of a spin- 1 particle with a magnetic
extra term q~σ · B 2
~ In other words, the spin- 1 which emerged in the Lorentz-invariant theory as an algebraic
field B. 2
necessity, does not leave the theory again when one takes the non-relativistic limit, but rather
remains as a steady “guest” of non-relativistic physics with the proper interaction term.
Let us consider briefly the consequences of the interaction of a spin- 12 with the magnetic field. For
this purpose we disregard the spatial degrees of freedom and assume the Schrödinger equation
~ Φ(t) .
i∂t Φ(t) = q ~σ · B (10.365)

The formal solution of this equation is

~
Φ(t) = e−iqtB·~σ Φ(0) . (10.366)

Comparision of this expression with (5.222, 5.223) shows that the propagator in (10.366) can be
~ by an angle qtB, i.e., the interaction q~σ · B
interpreted as a rotation around the field B ~ induces a
1
precession of the spin- 2 around the magnetic field.

Dirac Particle in Coulomb Field - Spectrum

We want to describe now the spectrum of a relativistic electron (q = −e) in the Coulomb field of
a nucleus with charge Ze. The respective bispinor wave function Ψ(xµ ) ∈ C4 is described as the
stationary solution of the Dirac equation (10.343) for the vector potential
Ze2
Aµ = (− , 0, 0, 0) . (10.367)
r
338 Relativistic Quantum Mechanics

For the purpose of the solution we assume the chiral representation, i.e, we solve

[ iγ̃ µ (∂µ + iqAµ ) − m ] Ψ̃(xµ ) = 0 (10.368)

where Ψ̃(xµ ) and γ̃ µ are defined in (10.228) and in (10.229), respectively. Employing πµ as defined
in Table 10.1 one can write (10.368)

( γ̃ µ πµ − m ) Ψ̃(xµ ) = 0 . (10.369)

For our solution we will adopt presently a strategy which follows closely that for the spectrum of
pionic atoms in Sect. 10.6. For this purpose we ‘square’ the Dirac equation, multiplying (10.369)
from the left by γ ν πν + m. This yields

[ iγ̃ µ (∂µ + iqAµ ) + m ] [ iγ̃ µ (∂µ + iqAµ ) − m ] Ψ̃(xµ )

= (γ̃ µ π̂µ γ̃ ν π̂ν − m2 ) Ψ̃(xµ ) = 0 . (10.370)

Any solution of (10.368) is also a solution of (10.370), but the converse is not necessarily true.
However, once a solution Ψ̃(xµ ) of (10.370) is obtained then

[ iγ̃ µ (∂µ + iqAµ ) + m ] Ψ̃(xµ ) (10.371)

is a solution of (10.369). This follows from

[ iγ̃ µ (∂µ + iqAµ ) + m ] [ iγ̃ µ (∂µ + iqAµ ) − m ]

= [ iγ̃ µ (∂µ + iqAµ ) − m ] [ iγ̃ µ (∂µ + iqAµ ) + m ] (10.372)

according to which follows from (10.370)

[ iγ̃ µ (∂µ + iqAµ ) − m ] [ iγ̃ µ (∂µ + iqAµ ) + m ] Ψ̃(xµ ) = 0 (10.373)

such that we can conclude that (10.371), indeed, is a solution of (10.369).

Equation (10.370) resembles closely the Klein-Gordon equation (10.180), but differs from it in an
essential way. The difference arises from the term γ̃ µ π̂µ γ̃ ν π̂ν in (10.370) for which holds

3
X X
µ ν
γ̃ π̂µ γ̃ π̂ν = (γ̃ µ )2 π̂µ2 + γ̃ µ γ̃ ν π̂µ π̂ν . (10.374)
µ=0 µ,ν=1
µ 6=ν

The first term on the r.h.s. can be rewritten using, according to (10.230), (γ̃ 0 )2 = 11 and (γ̃ j )2 =
−11, j = 1, 2, 3,
3
ˆ2 .
X
(γ̃ µ )2 π̂µ2 = π̂02 − ~π (10.375)
µ=0

Following the algebra that connected Eqs. (5.231), (5.232) in Sect. 5.7 one can write the second
term in (10.374), noting from (10.230) γ̃ µ γ̃ ν = − γ̃ ν γ̃ µ , µ 6= ν and altering ‘dummy’ summation
10.10: Dirac Particles in Electromagnetic Field 339

indices,
X 1 X
γ̃ µ γ̃ ν π̂µ π̂ν = ( γ̃ µ γ̃ ν π̂µ π̂ν + γ̃ ν γ̃ µ π̂ν π̂µ )
2
µ,ν=1 µ,ν=1
µ 6=ν µ 6=ν
1 X
= ( γ̃ µ γ̃ ν π̂µ π̂ν − γ̃ ν γ̃ µ π̂µ π̂ν + γ̃ ν γ̃ µ π̂ν π̂µ − γ̃ µ γ̃ ν π̂ν π̂µ )
4
µ,ν=1
µ 6=ν
1 X µ ν
= [γ̃ , γ̃ ] [π̂µ , π̂ν ] (10.376)
4
µ,ν=1
µ 6=ν

~ = 0. Since
This expression can be simplified due to the special form (10.367) of Aµ , i.e., due to A

[π̂µ , π̂ν ] = 0 for µ, ν = 1, 2, 3 (10.377)

which follows readily from the definition (10.344), it holds

1 X µ ν
[γ̃ , γ̃ ] [π̂µ , π̂ν ]
4
µ,ν=1
µ 6=ν
3 3
1 X 0 j 1 X j 0
= [γ̃ , γ̃ ] [π̂0 , π̂j ] + [γ̃ , γ̃ ] [π̂j , π̂0 ]
4 4
j=1 j=1
3
1 X
= [γ̃ 0 , γ̃ j ] [π̂0 , π̂j ] . (10.378)
2
j=1

According to the definition (10.229), the commutators [γ̃ 0 , γ̃ j ] are

0 −σ j 0 −σ j
     
0 j 0 11 0 11
[γ̃ , γ̃ ] = −
11 0 σj 0 σj 0 11 0
 j 
σ 0
= 2 (10.379)
0 −σ j
The commutators [π̂0 , π̂j ] in (10.378) can be evaluated using (10.367) and the definition (10.344)

[π̂0 , π̂j ] = (−i∂t + qA0 , −i∂j ] = − [ (∂t + iqA0 ) ∂j − ∂j (∂t + iqA0 ) ] f

= i ((∂j qA0 )) f (10.380)

where f = f (~r, t) is a suitable test function and where ((· · ·)) denotes the range to which the
derivative is limited. Altogether, one can summarize (10.376–10.380)
 
X
µ ν ~σ 0
γ̃ γ̃ π̂µ π̂ν = i · ((∇qA0 )) (10.381)
0 −~σ
µ,ν=1
µ 6=ν

According to (10.367) holds

Ze2
∇qA0 = r̂ . (10.382)
r2
340 Relativistic Quantum Mechanics

where r̂ = ~r/|~r| is a unit vector. Combining this result with (10.381), (10.374), (10.375) the
‘squared’ Dirac equation (10.368) reads
"  2 #
Ze2 Ze2
 
2 ~σ · r̂ 0 2
− ∂t − i + ∇ + i − m Ψ(xµ ) = 0 (10.383)
r 0 −~σ · r̂ r2

We seek stationary solutions of this equation. Such solutions are of the form

Ψ̃(xµ ) = Φ̃(~r) e−it . (10.384)

 can be interpreted as the energy of the stationary state and, hence, it is this quantity that we
want to determine. Insertion of (10.384) into (10.383) yields the purely spatial four-dimensional
differential equation
" 2 #
Ze2 Ze2
 
2 ~σ · r̂ 0 2
+ + ∇ + i − m Φ(~r) = 0 . (10.385)
r 0 −~σ · r̂ r2

We split the wave function into two spin- 12 components

 
φ̃+ (~r)
Ψ̃(~r) = (10.386)
φ̃− (~r)

and obtain for the separate components φ± (~r)

" 2 #
Ze2 Ze 2
+ + ∇2 ± i ~σ · r̂ 2 − m2 φ± (~r) = 0 . (10.387)
r r

The expression (10.189) for the Laplacian and expansion of the term (· · ·)2 result in the two-
dimensional equation
" #
2 L̂2 − Z 2 e4 ∓ i~σ · r̂ Ze2 2Ze2  2 2
∂r − + +  − m r φ± (~r) = 0 . (10.388)
r2 r

Except for the term i~σ · r̂ this equation is identical to that posed by the one-dimensional Klein-
Gordon equation for pionic atoms (10.191) solved in Sect. 10.6. In the latter case, a solution of
the form ∼ Y`m (r̂) can be obtained. The term i~σ · r̂, however, is genuinely two-dimensional and, in
fact, couples the orbital angular momentum of the electron to its spin- 12 . Accordingly, we express
the solution of (10.388) in terms of states introduced in Sect. 6.5 which describe the coupling of
orbital angular momentum and spin

{( Yjm (j − 21 , 12 |r̂), Yjm (j + 12 , 12 |r̂) ) ,

j = 12 , 3
2
. . . ; m = −j, −j + 1, . . . + j } (10.389)

According to the results in Sect. 6.5 the operator i~σ · r̂ is block-diagonal in this basis such that only
the states for identical j, m values are coupled, i.e., only the two states {Yjm (j − 12 , 12 |r̂), Yjm (j +
, |r̂)} as given in (6.147, 6.148). We note that these states are also eigenstates of the angular
1 1
2 2
10.10: Dirac Particles in Electromagnetic Field 341

momentum operator L̂2 [cf. (6.151]. We select, therefore, a specific pair of total spin-orbital angular
momentum quantum numbers j, m and expand

h± (r) g± (r)
φ± (~r) = Yjm (j − 12 , 12 |r̂) + Yjm (j + 12 , 12 |r̂) (10.390)
r r
Using
~σ · r̂ Yjm (j ± 12 , 12 |r̂) = − Yjm (j ∓ 12 , 12 |r̂) (10.391)
derived in Sect. 6.5 [c.f. (6.186)], property (6.151), which states that the states Yjm (j ± 21 , 12 |r̂) are
eigenfunctions of L̂2 , together with the orthonormality of these two states leads to the coupled
differential equation
" !
2 1 0

2Ze
∂r2 + + 2 − m (10.392)
r 0 1
!# 
(j − 12 )(j + 21 ) − Z 2 e4 ±iZe2

1 h± (r)
− 2 = 0.
r ±i Ze2 (j + 21 )(j + 32 ) − Z 2 e4 g± (r)

We seek to bring (10.392) into diagonal form. Any similarity transformation leaves the first term in
(10.392), involving the 2 × 2 unit matrix, unaltered. However, such transformation can be chosen
as to diagonalize the second term. Since, in the present treatment, we want to determine solely the
spectrum, not the wave functions, we require only the eigenvalues of the matrices
!
(j − 12 )(j + 12 ) − Z 2 e4 ±iZe2
B± = , (10.393)
±i Ze2 (j + 12 )(j + 32 ) − Z 2 e4

but do not explicitly consider further the wavefunctions. Obviously, the eigenvalues are independent
of m. The two eigenvalues of both matrices are identical and can be written in the form

λ1 (j) [λ1 (j) + 1] and λ2 (j) [λ2 (j) + 1] (10.394)

where
q
λ1 (j) = (j + 12 )2 − Z 2 e4 (10.395)
q
λ2 (j) = (j + 12 )2 − Z 2 e4 − 1 (10.396)

Equation (10.392) reads then in the diagonal representation

2Ze2
 
2 λ1,2 (j)[λ1,2 (j) + 1) 2 2
∂r − + + −m f1,2 (r) = 0 (10.397)
r2 r

This equation is identical to the Klein-Gordon equation for pionic atoms written in the form
(10.200), except for the slight difference in the expression of λ1,2 (j) as given by (10.395, 10.396)
and (10.199), namely, the missing additive term − 12 , the values of the argument of λ1,2 (j) being
j = 21 , 32 , . . . rather than ` = 0, 1, . . . as in the case of pionic atoms, and except for the fact that
we have two sets of values for λ1,2 (j), namely, λ1 (j) and λ2 (j).. We can, hence, conclude that the
342 Relativistic Quantum Mechanics

spectrum of (10.397) is again given by eq. (10.203), albeit with some modifications. Using (10.395,
10.396) we obtain, accordingly,
m
1 = r
2 4
; (10.398)
1 + qZ e
( n0 + 1 + (j+ 12 )2 − Z 2 e4 )2
m
2 = r
2 4
; (10.399)
1 + qZ e
( n0 + (j+ 12 )2 − Z 2 e4 )2
0
n = 0, 1, 2, . . . , j = 12 , 32 , . . . , m = −j, −j + 1, . . . , j

where 1 corresponds to λ1 (j) as given in (10.395) and 2 corresponds to λ2 (j) as given in (10.396).
For a given value of n0 the energies 1 and 2 for identical j-values correspond to mixtures of states
with orbital angular momentum ` = j − 21 and ` = j + 12 . The magnitude of the relativistic effect
is determined by Z 2 e4 . Expanding the energies in terms of this parameter allows one to identify
the relationship between the energies 1 and 2 and the non-relativistic spectrum. One obtains in
case of (10.398, 10.399)

mZ 2 e4
1 ≈ m − + O(Z 4 e8 ) (10.400)
2 (n0 + j + 32 )2
mZ 2 e4
2 ≈ m − + O(Z 4 e8 ) (10.401)
2 (n0 + j + 12 )2
n0 = 0, 1, 2, . . . , j = 12 , 32 , . . . , m = −j, −j + 1, . . . , j .

These expressions can be equated with the non-relativistic spectrum. Obviously, the second term on
the r.h.s. of these equations describe the binding energy. In case of non-relativistic hydrogen-type
atoms, including spin- 21 , the stationary states have binding energies

mZ 2 e4 n = 1, 2, . . . ` = 0, 1, . . . , n − 1
E = − , . (10.402)
2n2 m = −`, −` + 1, . . . , ` ms = ± 12

In this expression n is the so-called main quantum number. It is given by n = n0 + ` + 1 where `

is the orbital angular momentum quantum number and n0 = 0, 1, . . . counts the nodes of the wave
function. One can equate (10.402) with (10.400) and (10.401) if one attributes to the respective
states the angular momentum quantum numbers ` = j + 12 and ` = j − 21 . One may also state
this in the following way: (10.400) corresponds to a non-relativistic state with quantum numbers
n, ` and spin-orbital angular momentum j = ` − 21 ; (10.401) corresponds to a non-relativistic state
with quantum numbers n, ` and spin-orbital angular momentum j = ` + 12 . These considerations
are summarized in the following equations
m
ED (n, `, j = ` − 21 , m) = r ; (10.403)
1 + √ Z 2 e4 2
(n−`+ (`+1) − Z 2 e4 )2
m
ED (n, `, j = ` + 21 , m) = q ; (10.404)
Z 2√e4
1 + ( n − ` − 1 + `2 − Z 2 e4 )2

n = 1, 2, . . . ; ` = 0, 1, . . . , n − 1 ; m = −j, −j + 1, . . . , j
10.10: Dirac Particles in Electromagnetic Field 343

main orbital spin- non-rel. rel.

spectr. quantum angular orbital binding binding
notation number mom. ang. mom. energy / eV energy / eV
n ` j Eq. (10.402) Eq. (10.405 )
1
1s 1 1 0 2 -13.60583 -13.60601
2
1
2s 1 2 0 2 -3.40146 -3.40151
2
1
2p 1 2 1 2 ⇑ ⇑
2
3
2p 3 2 1 2 ⇑ - 3.40147
2
1
3s 1 3 0 2 -1.51176 -1.551178
2
1
3p 1 3 1 2 ⇑ ⇑
2
3
3p 3 3 1 2 ⇑ - 1.551177
2
3
3d 3 3 2 2 ⇑ ⇑
2
5
3d 5 3 2 2 ⇑ - 1.551176
2

Table 10.2:
Binding energies for the hydrogen (Z = 1) atom. Degeneracies are denoted by ⇑. The energies
were evaluated with m = 511.0041 keV and e2 = 1/137.036 by means of Eqs. (10.402, 10.405).

ED (n, `, j, m) =
One can combine the expressions (10.403, 10.404) finally into the single formula m
r
Z 2 e4
1 + q
(n−j − 12 + (j+ 12 )2 −

n = 1, 2, . . .
` = 0, 1,1 . . . , n − 1
2
for ` = 0
j =
` ± 2 otherwise
1

m = −j, −j + 1, . . . , j
(10.405)
In order to demonstrate relativistic effects in the spectrum of the hydrogen atom we compare
in Table 10.2 the non-relativistic [cf. (10.402)] and the relativistic [cf. (10.405)] spectrum of the
hydrogen atom. The table entries demonstrate that the energies as given by the expression (10.405)
in terms of the non-relativistic quantum numbers n, ` relate closely to the corresponding non-
relativistic states, in fact, the non-relativistic and relativistic energies are hardly discernible. The
reason is that the mean kinetic energy of the electron in the hydrogen atom, is in the range of
10 eV, i.e., much less than the rest mass of the electron (511 keV). However, in case of heavier
nuclei the kinetic energy of bound electrons in the ground state scales with the nuclear charge Z
like Z 2 such that in case Z = 100 the kinetic energy is of the order of the rest mass and relativistic
effects become important. This is clearly demonstrated by the comparision of non-relativistic and
relativistic spectra of a hydrogen-type atom with Z = 100 in Table 10.3.
344 Relativistic Quantum Mechanics

main orbital spin- non-rel. rel.

spectr. quantum angular orbital binding binding
notation number mom. ang. mom. energy / keV energy / keV
n ` j Eq. (10.402) Eq. (10.405 )
1
1s 1 1 0 2 -136.1 -161.6
2
1
2s 1 2 0 2 -34.0 -42.1
2
1
2p 1 2 1 2 ⇑ ⇑
2
3
2p 3 2 1 2 ⇑ - 35.2
2
1
3s 1 3 0 2 -15.1 -17.9
2
1
3p 1 3 1 2 ⇑ ⇑
2
3
3p 3 3 1 2 ⇑ - 15.8
2
3
3d 3 3 2 2 ⇑ ⇑
2
5
3d 5 3 2 2 ⇑ - 15.3
2

Table 10.3:
Binding energies for the hydrogen-type (Z = 100) atom. Degeneracies are denoted by ⇑. The
energies were evaluated with m = 511.0041 keV and e2 = 1/137.036 by means of Eqs. (10.402,
10.405).

Of particular interest is the effect of spin-orbit coupling which removes, for example, the non-
relativistic degeneracy for the six 2p states of the hydrogen atom: in the present, i.e., relativistic,
case these six states are split into energetically different 2p 1 and 2p 3 states. The 2p 1 states with
2 2 2
1
j = 2 involve two degenerate states corresponding to Y 1 m (1, 12 |r̂) for m = ± 12 , the 2p 3 states with
2 2
j = 3
2
involve four degenerate states corresponding to Y 3 m (1, 12 |r̂) for m = ± 12 , ± 32 .
2
In order to investigate further the deviation between relativistic and non-relativistic spectra of
hydrogen-type atoms we expand the expression (10.405) to order O(Z 4 e8 ). Introducing α = Z 2 e4
and β = j + 21 (10.405) reads
1
q α
(10.406)
1+ √
(n − β + β 2 −α)2

The expansion (10.206) provides in the present case

" #
mZ 2 e4 mZ 4 e8 1 3
ED (n, `, j, m) ≈ m − − 1 − + O(Z 6 e12 ) . (10.407)
2n2 2n3 j+ 2
4n

This expression allows one, for example, to estimate the difference between the energies of the
states 2p 3 and 2p 1 (cf. Tables 10.2,10.3). It holds for n = 2 and j = 32 , 12
2 2

mZ 4 e8 mZ 4 e8
 
1
E (2p 3 ) − E (2p 1 ) ≈ − −1 = . (10.408)
2 2 2 · 23 2 32
10.10: Dirac Particles in Electromagnetic Field 345

Radial Dirac Equation

We want to determine now the wave functions for the stationary states of a Dirac particle in a
4-vector potential
Aµ = (V (r), 0, 0, 0) (10.409)

where V (r) is spherically symmetric. An example for such potential is the Coulomb potential
V (r) = −Ze2 /r considered further below. We assume for the wave function the stationary state
form  
µ −it Φ(~r)
Ψ(x ) = e , (10.410)
X (~r)

where Φ(~r), X (~r) ∈ C2 describe the spatial and spin- 21 degrees of freedom, but are time-independent.
The Dirac equation reads then, according to (10.232, 10.345),

~σ · pˆ~ X + m Φ + V (r) Φ = Φ (10.411)

~σ · pˆ~ Φ − mX + V (r) X = X (10.412)

In this equation a coupling between the wave functions Φ(~r) and X (~r) arises due to the term ~σ · p ˆ~.
This term has been discussed in detail in Sect. 6.5 [see, in particular, pp. 168]: the term is a scalar
(rank zero tensor) in the space of the spin-angular momentum states Yjm (j ± 12 , 12 |r̂) introduced in
Sect. 6.5, i.e., the term is block-diagonal in the space spanned by the states Yjm (j ± 12 , 12 |r̂) and does
not couple states with different j, m-values; ~σ · pˆ~ has odd parity and it holds [c.f. (6.197, 6.198)]
" #
3
ˆ~ f (r) Yjm (j + 1 , 1 |r̂) j+ 2
~σ · p 2 2
= i ∂r + f (r) Yjm (j − 12 , 12 |r̂)
r
(10.413)
" #
1
ˆ~ g(r) Yjm (j − 1 , 1 |r̂) 2 −j
~σ · p 2 2
= i ∂r + g(r) Yjm (j + 12 , 12 |r̂) .
r
(10.414)

These equations can be brought into a more symmetric form using

1 1
∂r + = ∂r r
r r

which allows one to write (10.413, 10.414)

" #
1
ˆ~ rf (r) Yjm (j + 1 , 1 |r̂) j+ 2
~σ · p 2 2
= i ∂r + r f (r) Yjm (j − 12 , 12 |r̂)
r
(10.415)
" #
1
ˆ~ rg(r) Yjm (j − 1 , 1 |r̂) j+ 2
~σ · p 2 2
= i ∂r − r g(r) Yjm (j + 12 , 12 |r̂) .
r
(10.416)
346 Relativistic Quantum Mechanics

The differential equations (10.411, 10.412) are four-dimensional with ~r-dependent wave functions.
The arguments above allow one to eliminate the angular dependence by expanding Φ(~r) and X (~r)
in terms of Yjm (j + 12 , 12 |r̂) and Yjm (j − 12 , 12 |r̂), i.e.,
 
  a(r) b(r)
Φ(~r) Yjm (j + 2 , 2 |r̂) +
1 1
Yjm (j − 2 , 2 |r̂) 
1 1

 =  r r

  . (10.417)
 c(r) d(r)
X (~r)

Yjm (j + 12 , 12 |r̂) + Yjm (j − 12 , 12 |r̂)
r r
In general, such expansion must include states with all possible j, m values. Presently, we consider
the case that only states for one specific j, m pair contribute. Inserting (10.417) into (10.411,
10.412), using (10.415, 10.416), the orthonormality property (6.157), and multiplying by r results
in the following two independent pairs of coupled differential equations
" #
j + 21
i ∂r − d(r) + [ m + V (r) −  ] a(r) = 0
r
" #
j + 21
i ∂r + a(r) + [ −m + V (r) −  ] d(r) = 0 (10.418)
r
and
" #
j + 12
i ∂r + c(r) + [ m + V (r) −  ] b(r) = 0
r
" #
j + 12
i ∂r − b(r) + [ −m + V (r) −  ] c(r) = 0. (10.419)
r

Obviously, only a(r), d(r) are coupled and b(r), c(r) are coupled. Accordingly, there exist two
independent solutions (10.417) of the form
 
  f1 (r)
 i r Yjm (j + 2 , 2 |r̂) 
1 1
Φ(~r)
  =   (10.420)
X (~r)
 g1 (r) 
− Yjm (j − 2 , 2 |r̂)
1 1
r
 
f2 (r)
 i r Yjm (j − 2 , 2 |r̂) 
! 1 1
Φ(~r)
=   (10.421)
X (~r)  g2 (r) 
− Yjm (j + 2 , 2 |r̂)
1 1
r
where the factors i and −1 have been introduced for convenience. According to (10.418) holds for
f1 (r), g1 (r)
" #
j + 21
∂r − g1 (r) + [  − m − V (r) ] f1 (r) = 0
r
" #
j + 12
∂r + f1 (r) − [  + m − V (r) ] g1 (r) = 0 (10.422)
r
10.10: Dirac Particles in Electromagnetic Field 347

and for f2 (r), g2 (r)

" #
1
j+ 2
∂r + g2 (r) + [  − m − V (r) ] f2 (r) = 0
r
" #
1
j+ 2
∂r − f2 (r) − [  + m − V (r) ] g2 (r) = 0 (10.423)
r

Equations (10.422) and (10.423) are identical, except for the opposite sign of the term (j + 21 ); the
equations determine, together with the appropriate boundary conditions at r = 0 and r → ∞,
the radial wave functions for Dirac particles in the potential (10.409).

Dirac Particle in Coulomb Field - Wave Functions

We want to determine now the wave functions of the stationary states of hydrogen-type atoms
which correspond to the energy levels (10.405). We assume the 4-vector potential of pure Coulomb
type (10.367) which is spherically symmetric such that equations (10.422, 10.423) apply for V (r) =
−Ze2 /r. Equation (10.422) determines solutions of the form (10.420). In the non-relativistic limit,
Φ in (10.420) is the large component and X is the small component. Hence, (10.422) corresponds
to states
 
f1 (r)
i Yjm (j + 2 , 2 |r̂) 
1 1
Ψ(xµ ) ≈  r , (10.424)
0

i.e., to states with angular momentum ` = j + 12 . According to the discussion of the spectrum
(10.405) of the relativistic hydrogen atom the corresponding states have quantum numbers n =
1, 2, . . . , ` = 0, 1, . . . , n − 1. Hence, (10.422) describes the states 2p 1 , 3p 1 , 3d 3 , etc. Similarly,
2 2 2
(10.423), determining wave functions of the type (10.421), i.e., in the non-relativistic limit wave
functions
 
f2 (r)
i Yjm (j − 2 , 2 |r̂) 
1 1
Ψ(xµ ) ≈  r , (10.425)
0

covers states with angular momentum ` = j− 21 and, correspodingly the states 1s 1 , 2s 1 , 2p 3 , 3s 1 , 3p 3 , 3d 5 ,

2 2 2 2 2 2
etc.
We consider first the solution of (10.422). The solution of (10.422) follows in this case from the same
procedure as that adopted for the radial wave function of the non-relativistic hydrogen-type atom.
According to this procedure, one demonstrates first that the wave function at r → 0 behaves as rγ
for some suitable γ, one demonstrates then that the wave functions for r → ∞ behaves as exp(−µr)
for some suitable µ, and obtains finally a polynomial function p(r) such that rγ exp(−µr)p(r) solves
(10.422); enforcing the polynomial to be of finite order leads to discrete eigenvalues , namely, the
ones given in (10.405).
348 Relativistic Quantum Mechanics

Behaviour at r → 0
We consider first the behaviour of the solutions f1 (r) and g1 (r) of (10.422) near r = 0. We note
that (10.422), for small r, can be written
" #
j + 21 Ze2
∂r − g1 (r) + f1 (r) = 0
r r
" #
j + 12 Ze2
∂r + f1 (r) − g1 (r) = 0 . (10.426)
r r

Setting
f1 (r) ∼ a rγ , g1 (r) ∼ b rγ (10.427)
→0 →0

yields

γ b rγ−1 − (j + 12 ) b rγ−1 + Ze2 a rγ−1 = 0

γ a rγ−1 + (j + 12 ) a rγ−1 − Ze2 b rγ−1 = 0. (10.428)

or
γ + (j + 12 ) −Ze2
  
a
= 0. (10.429)
Ze 2 γ − (j + 12 ) b
This equation
q poses an eigenvalue problem (eigenvalue −γ) for proper γ values. One obtains
γ = ± (j + 12 )2 − Z 2 e4 . The assumed r-dependence in (10.427) makes only the positive solution
possible. We have, hence, determined that the solutions f1 (r) and g1 (r), for small r, assume the
r-dependence in (10.427) with r
1
γ = (j + )2 − Z 2 e4 . (10.430)
2
Note that the exponent in (10.427), in case j + 12 )2 < Ze2 , becomes imaginary. Such r-dependence
would make the expectation value of the potential
Z
1
r2 dr ρ(~r) (10.431)
r

infinite since, according to (10.266, 10.267, 10.420), for the particle density holds then

1
ρ(~r) ∼ |rγ−1 |2 = . (10.432)
r2

Behaviour at r → ∞
For very large r values (10.422) becomes

∂r g1 (r) = − (  − m, ) f1 (r)
∂r f1 (r) = (  + m ) g1 (r) (10.433)
10.10: Dirac Particles in Electromagnetic Field 349

Iterating this equation once yields

∂r2 g1 (r) = ( m2 − 2 ) g1 (r)

∂r2 f1 (r) ( m2 − 2 ) f1 (r)
= (10.434)
√
The solutions of these equations are f1 , g1 ∼ exp(± m2 − 2 r). Only the exponentially decaying
solution is admissable and, hence, we conclude
p
f1 (r) ∼ e−µr , g1 (r) ∼ e−µr , µ = m2 − 2 (10.435)
→∞ →∞

For bound states holds  < m and, hence, µ is real. Let us consider then for the solution of (10.434)
√ √
f1 (r) = m +  a e−µ r , g1 (r) = − m −  a e−µr . (10.436)

Insertion into (10.434) results in

√ √
(m − ) m +  a − (m − ) m +  a = 0
√ √
(m + ) m −  a − (m + ) m −  a = 0 (10.437)

which is obviously correct.

Solution of the Radial Dirac Equation for a Coulomb Potential

To solve (10.422) for the Coulomb potential V (r) = −Ze2 /r We assume a form for the solution
which is adopted to the asymptotic solution (10.436). Accordingly, we set
√
f1 (r) = m +  e−µr f˜1 (r) (10.438)
√ −µr
g1 (r) = − m −  e g̃1 (r) (10.439)

where µ is given in (10.435 ). Equation (10.422 ) leads to

√ j + 12 √ Ze2 ˜
− m −  [∂r − ] g̃1 + m +  f1 +
√r r√
(m − ) m +  g̃1 − (m − ) m +  f˜1 = 0 (10.440)
√ j + 12 ˜ √ Ze2
m +  [∂r + ] f1 + m −  g̃1 −
r
√ r√
(m + ) m − f˜1 + (m + ) m − g̃1 = 0 (10.441)

The last two terms on the l.h.s. of both (10.440) and (10.441) correspond to (10.437) where they
cancelled in case f˜1 = g̃1 = a. In the present case the functions f˜1 and g̃1 cannot be chosen
identical due to the terms in the differential equations contributing for finite r. However, without
loss of generality we can choose

f˜1 (r) = φ1 (r) + φ2 (r) , g̃1 (r) = φ1 (r) − φ2 (r) (10.442)

which leads to a partial cancellation of the asymptotically dominant terms. We also introduce the
new variable
ρ = 2µ r . (10.443)
350 Relativistic Quantum Mechanics

From this results after a little algebra

j + 21 m +  Ze2
r
[∂ρ − ] (φ1 − φ2 ) − (φ1 + φ2 ) + φ2 = 0 (10.444)
ρ m− ρ
j + 12 m −  Ze2
r
[∂ρ + ] (φ1 + φ2 ) + (φ1 − φ2 ) − φ2 = 0 . (10.445)
ρ m+ ρ
Addition and subtraction of these equations leads finally to the following two coupled differential
equations for φ1 and φ2
1
j+ 2 Ze2 mZe2
∂ρ φ1 + φ2 − √ φ1 − √ φ2 = 0 (10.446)
ρ m 2 − 2 ρ m2 − 2 ρ
1
j+ 2 mZe2 Ze2
∂ρ φ2 + φ1 + √ φ1 + √ φ2 − φ2 = 0 (10.447)
ρ m 2 − 2 ρ m 2 − 2 ρ
We seek solutions of (10.446 , 10.447) of the form
n0
X
γ
φ1 (ρ) = ρ αs ρ2 (10.448)
s=0
n0
X
γ
φ2 (ρ) = ρ βs ρ2 (10.449)
s=0

for γ given in (10.430) which conform to the proper r → 0 behaviour determined above [c.f.
(10.426–10.430)]. Inserting (10.448, 10.449) into (10.446, 10.447) leads to
X Ze2
(s + γ) αs + (j + 12 ) βs − √ αs
s m 2 − 2
mZe2

−√ βs ρs+γ−1 = 0 (10.450)
2
m − 2

mZe2
X 
(s + γ) β2 + (j + 12 ) αs + √ αs
s m2 − 2
Ze2

+√ βs − βs−1 = 0 (10.451)
m2 − 2
From (10.450) follows
2
√mZe − (j + 12 )
αs m2 −2
= Ze2
. (10.452)
βs s + γ − √m 2 −2

From (10.451) follows

m2 Z 2 e4
Ze2 − (j + 12 )2
 
m2 −2
βs−1 = s+γ + √ βs + 2 βs
m2 − 2 s+γ − √ Ze
m2 −2
(s + γ)2 + Z 2 e4 − (j + 12 )2
= 2 βs . (10.453)
s+γ − √ Ze
m2 −2
10.10: Dirac Particles in Electromagnetic Field 351

Using (10.430) one can write this

2
s+γ − √ Ze
m2 −2
βs = βs−1 . (10.454)
s (s + 2γ)

Defining
Ze2
so = √ −γ (10.455)
m2 − 2
one obtains
s − so
βs = βs−1
s(s + 2γ)
(s − 1 − so )(s − so )
= βs−2
(s − 1)s (s − 1 + 2γ)(s + 2γ)
..
.
(1 − so )(2 − so ) . . . (s − so )
= β0 (10.456)
s! (2γ + 1)(2γ + 2) · · · (2γ + s)

From (10.452) follows

1 2
j+ − √mZe
2 m2 −2 (−so )(1 − so )(2 − so ) . . . (s − so )
αs = β0 (10.457)
so s! (2γ + 1)(2γ + 2) · · · (2γ + s)

One can relate the polynomials φ1 (ρ) and φ2 (ρ) defined through (10.448, 10.449) and (10.456,
10.457) with the confluent hypergeometric functions

a a(a + 1) x2
F (a, c; x) = 1 + x + + ... (10.458)
c c(c + 1) 2!

or, equivalently, with the associated Laguerre polynomials

L(α)
n = F (−n, α + 1, x) . (10.459)

It holds
1 2
j+ − √mZe
2 m −2
2
φ1 (ρ) = β0 ργ F (−so , 2γ + 1; ρ) (10.460)
so
φ2 (ρ) = β0 ργ F (1 − so , 2γ + 1; ρ) . (10.461)

In order that the wave functions remain normalizable the power series (10.448, 10.449) must be of
finite order. This requires that all coefficients αs and βs must vanish for s ≥ n0 for some n0 ∈ N.
The expressions (10.456) and (10.457) for βs and αs imply that so must then be an integer, i.e.,
so = n0 . According to the definitions (10.430, 10.455) this confinement of so implies discrete values
for , namely,
m
(n0 ) = r 2 e4
, n0 = 0, 1, 2, . . . (10.462)
Z
1 + q
1
(n0 + (j+ 2 )2 − Z 2 e4 )2
352 Relativistic Quantum Mechanics

This expression agrees with the spectrum of relativistic hydrogen-type atoms derived above and
given by (10.405). Comparision with (10.405) allows one to identify n0 = n − j + 12 which, in fact,
is an integer. For example, for the states 2p 1 , 3p 1 , 3d 3 holds n0 = 1, 2, 1. We can, hence, conclude
2 2 2
that the polynomials in (10.461 for  values given by (10.405) and the ensuing so values ( 10.455)
are finite.
Altogether we have determined the stationary states of the type (10.421) with radial wave functions
f1 (r), g1 (r) determined by (10.438, 10.439), (10.442), and (10.460, 10.461). The coefficients β0 in
(10.460, 10.461) are to be chosen to satisfy a normalization condition and to assign an overall
phase. Due to the form (10.410) of the stationary state wave function the density ρ(xµ ) of the
states under consideration, given by expression (10.267), is time-independent. The normalization
integral is then
Z ∞ Z π Z 2π
2
r dr sin θdθ dφ (|Φ(~r)|2 + |X (~r)|2 ) = 1 (10.463)
0 0 0

where Φ and X , as in (10.421), are two-dimensional vectors determined through the explicit form
of the spin-orbital angular momentum states Yjm (j ± 12 , 12 |r̂) in (6.147, 6.148). The orthonormality
properties (6.157, 6.158) of the latter states absorb the angular integral in (10.463) and yield [note
the 1/r factor in (10.421)]
Z ∞
dr (|f1 (r)|2 + |g1 (r)|2 ) = 1 (10.464)
0

The evaluation of the integrals, which involve the confluent hypergeometric functions in (10.460,
10.461), can follow the procedure adopted for the wave functions of the non-relativistic hydrogen
atom and will not be carried out here.
The wave functions (10.421) correspond to non-relativistic states with orbital angular momentum
` = j + 12 . They are described through quantum numbers n, j, ` = j + 21 , m. The complete wave
function is given by the following set of formulas

iF1 (r) Yj,m (j + 12 , 12 |r̂)

!
1 µ −it
Ψ(n, j, ` = j + 2 , m|x ) = e (10.465)
G1 (r) Yj,m (j − 12 , 12 |r̂)

1
F1 (r) = F− (κ|r) , G1 (r) = F+ (κ|r) , κ = j+ 2 (10.466)

where2

(n0 + γ)m
 
γ−1 −µr
F± (κ|r) = ∓ N (2µr) e − κ F (−n0 , 2γ + 1; 2µr)


0 0
± n F (1 − n , 2γ + 1; 2µr) (10.467)

3
v
(2µ) 2 u m ∓ )Γ(2γ + n0 + 1)
N = (10.468)
u  
Γ(2γ + 1) 4m (n0 +γ)m (n0 +γ)m − κ n0 !
t
 

2
This formula has been adapted from ”Relativistic Quantum Mechanics” by W. Greiner, (Springer, Berlin, 1990),
Sect. 9.6.
10.10: Dirac Particles in Electromagnetic Field 353

and
p
µ = (m − )(m + )
q
γ = (j + 12 )2 − Z 2 e4
n0 = n − j − 12
m
 = q
2 e4
. (10.469)
1 + (nZ0 +γ) 2

We want to consider now the stationary states of the type (10.421) which, in the non-relativistic
limit, become  
f2 (r)
i Yjm (j − 2 , 2 |r̂) 
1 1
Ψ(xµ ) ≈ e−1t  r . (10.470)
0

Obviously, this wavefunction has an orbital angular momentum quantum number ` = j − 12 and,
accordingly, describes the complementary set of states 1s 1 , 2s 1 , 2p 3 , 3s 1 , 3p 3 , 3d 5 , etc. not not
2 2 2 2 2 2
covered by the wave functions given by (10.465–10.469). The radial wave functions f2 (r) and g2 (r)
in (10.421) are governed by the radial Dirac equation (10.423) which differs from the radial Dirac
equation for f1 (r) and g1 (r) solely by the sign of the terms (j + 21 )/r. One can verify, tracing all
steps which lead from (10.422) to (10.469) that the following wave functions result

iF2 (r) Yj,m (j − 12 , 12 |r̂)

!
1 µ −it
Ψ(n, j, ` = j − 2 , m|x ) = e (10.471)
G2 (r) Yj,m (j + 12 , 12 |r̂)
1
F2 (r) = F− (κ|r) , G2 (r) = F+ (κ|r) , κ = −j − 2 (10.472)
where F± (κ|r) are as given in (10.467–10.469). We have, hence, obtained closed expressions for the
wave functions of all the stationary bound states of relativistic hydrogen-type atoms.
354 Relativistic Quantum Mechanics