CHAPTER 3
RELATIVISTIC QUANTUM MECHANICS
OF SPIN-} PARTICLES
3-1. PROBABILITY CONSERVATION IN RELATIVISTIC
QUANTUM MECHANICS
It is almost traditional to start an exposition of the Dirac theory of spin-} par-
ticles by discussing some of the “difficulties” of the Klein-Gordon theory. As we
shall see Jater, there is actually nothing wrong in the Klein-Gordon equation if
it is properly interpreted. We shall sketch, however, the usual arguments against
the Klein-Gordon equation since they played an important historic role in the
formulation of relativistic quantum mechanics.
In Schrédinger’s wave mechanics we associate a complex-valued wave function
ap with a single particle, such that |xp|*d?x gives the probability of finding the
particle in a volume element d'x. This interpretation is possible because the proba-
bitity density P and the flux density S given by
P=lylt>0 GD
and
S = —(ih/2m)(p* Vy —~ Vag) G2)
satisfy the continuity equation
aPjat+V-S=0 GB.3)
by virtue of the Schrddinger equation. Using Gauss’ theorem, we also see that the
integral over all space f Pd*x is a constant of the motion which can be set to unity
by appropriately normalizing yp.
If one is to construct a relativistic quantum mechanics in analogy with non-
relativistic quantum mechanics, it may appear natural to impose the following
requirements on the theory. First, with the relativistic wave function we must be
able to construct bilinear forms which can be interpreted as the probability density
and the flux density satisfying a continuity equation of type (3.3). The probability
dens'ty we form must, of course, be positive definite. In addition, the special theory
of relativity requires that P must be the fourth component of a four-vector density.
To sce this last point, we recall that d?x — d°x,/1 — (vjc)? under a Lorentz
transformation because of the well-known Lorentz contraction of the volume
element; if Pd*x is to remain invariant, it is essential that P transforms like the
fourth component of a four-vector P ~* P//T — Je). The continuity equation
(3.3) takes the following covariant form,
(/0%,)5, = 0, G4)
where
(S, icP). (3.5)76 RELATIVISTIC QUANTUM MECHANICS OF SPIN-} PARTICLES 341
Let us now see whether the relativistic quantum mechanics based on the Klei
Gordon equation satisfies the above requirements. Consider a four-vector densi
given by
se =a(or — 2 4), 3.6)
2 OXy
where @ is a solution to the free particle Klein-Gordon equation, and 4 is a mul-
tiplicative constant. The four-divergence of (3.6) vanishes,
as — 4 [28° 28 — Censor tae — 38° 2).
ox, ax, Or, Ox, x,
G.7)
by virtue of the Klein-Gordon equation, For a Klein-Gordon particle moving
at nonrelativistic velocities (E ~ mec’),
bm peer, (G8)
where yp is the corresponding Schrddinger solution (cf. Problem 1-2). The com-
ponents of s, are then given by
So = is, = Qime/h)A lh, s = AL Vb — (WH. G9)
If we set A = —ih/2en, then s and so are precisely the flux density and c times
the probability density in the Schrddinger theory. Thus we obtain a four-vector
current density from a solution of the Klein-Gordon equation with the following
properties: (i) the current density satisfies the continuity equation, and (ii) the
components of the current density coincide with the flux density and c times the
probability density in the nonrelativistic timit,
So far everything appears satisfactory. There is, however, a difficulty in inter
preting
ih_( go 0b _ ag*
P= sl 3 ~ 79) oe
as the probability density. In the Schrédinger theory, in which the time derivative
appears only linearly in the wave equation, the sign of the frequency is determined
by the eigenvalue of the Hamiltonian operator. In contrast, because the Klein-
Gordon equation is of second order in the time derivative, both u(xje""" and
u*(xje*!*"* ace equally good solutions for a given physical situation (cf. Problem
1-3). This means that P given by (3.10) can be positive or negative. We may arbi-
trarily omit alt solutions of the form u(xje”"* with E <0. But this would be
unjustified because solutions of the form u(x)e~'** with E> 0 alone do not
form a complete set. It appears that we must cither abandon the interpretation
of (3,10) as the probability density or abandon the Klein-Gordon equation al-
together.
Let us analyze the origin of this difficulty a little more closely. From the deriva~
tion of the continuity equation (3.7) we may infer that the appearance of the linear
time derivative in s. is unavoidable so long as the wave function satisfies a partial
differential equation quadratic in the time derivative. Perhaps we could avoid
this difficulty if we wrote a relativistic wave equation /inear in the time derivative.34 PROBABILITY CONSERVATION 77
In 1928, in what is undoubtedly one of the most significant papers in the physics
of the twentieth century, P.A.M. Dirac succeeded in devising a relativistic wave
equation starting with the requirement that the wave equation be linear in 3/81.
Using his equation, known to us as the Dirac equation, he was able to construct
a conserved four-vector density whose zeroth component is positive-definite. For
this reason, from 1928 until 1934 the Dirac equation was considered to be the
only correct wave equation in relativistic quantum mechanics.
In 1934 the Klein-Gordon equation was revived by W. Pauli and V. F. Weiss-
kopf. Their proposal was that, up to a proportionality factor, s, given by (3.6)
be interpreted as the charge-current density rather than as the probability-current
density. As we saw in Section 1-3, an interpretation of this kind is reasonable
in the classical field theory of a complex scalar field. The fact that the sign of sp
changes when u*(x)e!”""* is substituted for u(x)e'"”* makes good sense if the
negative-energy solution is interpreted as the wave function for a particle with
opposite electric charge (cf. Problem 1-3).
The interpretation of s, as the charge-current density is even more satisfactory
fora theory in which a solution to the Klein-Gordon equation is to be interpreted
as a quantized field operator. In analogy with what we did for the complex scalar
field in classical field theory we form a non-Hermitian field operator (3 $*)
such that
Git
o= tae, $ ite
where $, and $2 are Hermitian operators whose properties are given in Problem
2-3. Consider now a four-current operator
GAN)
j= e(gr 2 — 8" 4).
e(# x, ax, %) (G12)
It is easy to show (Problem 3-1) that the fourth component of j, has the property
JOdliods = eE (ME? — NO, (3.13)
:
where
(3.14)
with
Pe F ia, (3.15)
Physically, N‘*) and N'~? are the number operators for the Klein-Gordon particle
of charge e and for its antiparticle with charge —e. So the eigenvalue of the opera~
tor expression (3.13) is the total charge of the field, Usually the four-vector j, is
regarded as the charge-current density operator.
As emphasized in Chapter 2, quantum field theory accommodates physical
situations in which particles are created or annihilated. When there are processes
like y — * 4-2” which take place in the Coulomb field of a nucleus, what is
conserved is not the probability of finding a given particle integrated over all
space but rather the total charge of the field given by the eigenvalue of (3.13).