Klein-Gordon equation and Dirac Equation

It is natural to assume that a particle moving in the +x direction with velocity v can
be represented by the following wave function:

y ( x, t ) = Ae - iw (t - x / v ) = Ae - iwt +ikx

Using momentum and energy parameters y = Ae -( i / h )( Et - px )

Let us differentiate the above equation and try to find its implication
¶y p ¶y E
=i y = -i y
¶x h ¶t h
In the wavefunction algebra, the
¶ ˆ ¶
momentum and energy can be pˆ = -ih (?) E = ih (?)
represented by the following operators: ¶x ¶t

Many other operators can be written in 2

pˆ 2 1 æh ¶ ö h2 ¶ 2
terms of them. For example, the kinetic KE = = ç ÷ =- 2
energy operator for the free particle is 2 m 2 m è i ¶x ø 2 m ¶x

1
One way to “derive” the Schrödinger equation is to start with the classical energy-
momentum relation
p2
+V = E
2m
Applying the quantum prescription,
h ¶
p® Ñ E ® ih
i ¶t
Letting the resulting operator act on the “wave function” (why?)
h2 2 ¶y
- Ñ y + Vy = ih
2m ¶t
The Klein-Gordon equation can be obtained in exactly the same way, beginning
with the relativistic energy-momentum relation for a free particle
E 2 - p2c 2 = m 2c 4
Its tensor form is more helpful:
p m pm - m 2 c 2 = 0

2
Converting the quantum prescription into a relativistically covariant form,

h ¶
p® Ñ E ® ih
i ¶t
E ¶ ¶
= p0 ® ih = ih 0
c ¶ ( ct ) ¶x
h ¶ ¶
- p x = p1 ® - = ih 1
i ¶x ¶x
Thus, we can introduce a four-vector form of quantum prescription.

¶
pm ® i h m = i h¶ m pm ® ih¶ m
¶x
2
2 æ 2 æ mc ö
ö
p m pm - m 2 c 2 ® - h 2¶ m ¶ m - ( mc ) = -h ç W + ç
2
÷ ÷÷
ç è h ø ø
è
According to the previous considerations, the relativistic relation between energy
and momentum is fulfilled, if the wave function obeys the differential equation

3
 æ 2 æ mc ö 2 ö
çç W + ç ÷ ÷÷y ( x ) = 0
è è h ø ø
which is the so-called Klein-Gordon equation. Note that here x means the four-
vector variable x = (x0, x1, x2, x3).

A more familiar, three vector form is

2 2
1 ¶ y ( x, t ) 2 æ mc ö
- + Ñ y ( x , t ) = ç ÷ y ( x, t )
è h ø
2 2
c ¶t

Nonrelativistic Limit of the Klein-Gordon Equation

In order to study the nonrelativistic limit, we make the ansatz,

æ i ö
y (x, t ) = f (x, t ) exp ç - mc 2t ÷
è h ø
i.e. we split off a term containing the rest mass. Let us think of a case that the
particle only has the rest mass, E = mc2. The trivial answer is (up to normalization)
æ i ö
y (x, t ) = exp ç - mc 2t ÷ f ( x, t ) = 1
h
è ø
4
 The next case to consider is the situation when E is above mc2, but much less than
the rest mass energy. This is essentially the nonrelativistic limit

E ¢ = E - mc 2 E ¢ = mc 2
We know that the time derivative of ϕ now deviates from zero, but we expect its
value to be small:
¶f
ih = mc 2f
¶t
Equipped with this information, let us calculate the derivatives

¶y æ ¶f imc 2 ö æ i 2 ö imc 2 æ i ö
=ç - f ÷ exp ç - mc t ÷ » - f exp ç - mc 2t ÷
¶t è ¶t h ø è h ø h è h ø
¶ 2y ¶ éæ ¶f imc 2 ö æ i 2 ö ù é 2imc 2
¶f m 2 4
c ù æ i 2 ö
= êç - f ÷ exp ç - mc t ÷ú ê» - - f ú exp ç - mc t÷
¶t 2
¶t ëè ¶t h ø è h øû ë h ¶t h 2
û è h ø

Inserting this result into the Klein-Gordon equation,

2
1 é 2imc 2 ¶f m 2c 4 ù æ i 2 ö é 2 æ mc ö ù æ i 2 ö
2 ê
+ f ú exp ç - mc t =
÷ ê -Ñ f + ç ÷ ú f exp ç - mc t÷
c ë h ¶t h 2
û è h ø êë è h ø úû è h ø
5
From which follows,

1 2imc 2 ¶f ¶f h2 2
= -Ñ 2
f ® ih =- Ñf
c 2
h ¶t ¶t 2m
This is the Schrödinger equation without the potential term which is know to be
the correct nonrelativistic limit for the spin 0 particle. Thus, we may suspect that
the Klein-Gordon equation describes the relativistic motion of a spin-0 particle.

6
 Physical Problems of the Klein-Gordon Equation
The Klein-Gordon equation fulfills the laws of special relativity, but contains two
fundamental problems to be a physically meaningful equation. The first problem is
obvious when considering the solutions of the differential equations. Using the
ansatz,

y (x, t ) = aei( k×x -wt ) = a exp ( -ik × x ) = a exp ( -ik m x m ) k m = ( w / c, - k x , - k y , - k z )

Then, the Klein-Gordon equation in k-space is

2 2
æ mc ö æ mc ö
-km k m + ç ÷ =0 or k2 = ç ÷
è h ø è h ø
From which follows,
2 2
0 2 æ mc ö
2 æ mc ö
2
(k ) = k +ç
è h ø
÷
0
k = ± k +ç
è h ø
÷

This means that the Klein-Gordon equation allows negative energies as solution.
There is the problem of the physical interpretation of negative energies. (Can’t we
just ignore them? No we cannot. In the quantum world, all solutions must be
counted for the completeness of the quantum states.)
7
The second problem is less obvious. It occurs when interpreting the function ψ(x)
as probability amplitude. From our experience with nonrelativistic quantum theory,
it is tempting to set ψ*ψ as the probability density ρ so that it satisfies the
continuity equation
¶r
+ Ñ× j = 0 ¶m jm = 0 j m = ( c r , j)
¶t
For the nonrelativistic quantum theory,
ih æ ih ö * t
r NR = y y*
jNR =- éy Ñy - ( Ñy )y ù º - ç
* *
÷y Ñy
2m ë û è 2m ø
where we define
t t
A¶ i B º A ( ¶ i B ) - ( ¶ i A) B AÑB º A ( ÑB ) - ( ÑA) B

Note that saving space is not the only reason we use this notation. For clarification,
(∂iA)B may better be written as B(∂iA), but when A and B are matrices or operators,
there is no guarantee that we can exchange their order. The above notation is one
method to clarify the derivative while keeping the operator positions.

8
For the theory to be truly relativistic, we need to express all the equations using
four-vector notation, thus the current four-vector must be
m ih * t m
j = y ¶y
2m
ih t ih
m
¶m j =
2m
( * m
¶m y ¶ y = )
2m
( (
¶ m y *¶ my - ¶ my * y ) )
ih
=
2m ë
( )( ) ( ) ( ) (
é ¶ my * ¶ my + y * ¶ m ¶ my - ¶ m ¶ my * y - ¶ my * ¶ my ù )( û)
ih
=
2m ë
( ) (
éy * W2y - W2y * y ù
û )
In the second line, the first and fourth terms cancel each other. If you’re not sure,
compare all four components to check this. For the final equation, if ψ fulfills the
Klein-Gordon equation, the right hand side vanishes, and the continuity equation
holds. However, the current defined above contains the second problem:
1 0 ih * t 0 ih é *¶ ¶ *ù
r= j = y ¶y = y
êë ¶t y - y y ú
c 2mc 2mc 2 ¶t û

It can be positive or negative, depending on the values of ψ and its derivative.

Since the Klein-Gordon equation is a second order partial differential equation,
one has the option to choose the two boundary conditions,
9
 ¶
y ( x, t = 0 ) and y ( x, t = 0 )
¶t
Thus it is impossible to prevent ρ, the probability density, becoming negative. This
is the problem of the indefinite probability density.
The solution of both problems has an interesting historical development, which
can briefly be summarized as follows:
1927: Pauli upgraded Schrödinger equation to include particle spin, but the theory was still
nonrelativistic.
1928: Dirac “invents” the Dirac equation. The theory is relativistic and the probability
density is positive. However, negative energies are allowed.
1930: Dirac solves the problem of negative energies via the “hole” theory. Antiparticles are
related to negative energy eigenstates.
1934 : Pauli and Weisskopf present a new interpretation of the Klein-Gordon equation as
field equation for a charged spin-0 field. ρ represents the charge density and the energy is
definitely positive.
1934: The Dirac equation acquires a field theoretical interpretation. It does no longer
determine a probability amplitude. It became the field operator for a spin 1/2 field.

We will follow the same route for the next few classes.
10
Pauli Equation
The Klein-Gordon equation was announced in 1926, but it definitely had problems.
Heisenberg’s and Schrödinger’s nonrelativistic theory was somewhat successful,
but it does not say anything about the nature of spin. In 1927, Pauli upgraded
Schrödinger equation to include particle spin, but the theory was still
nonrelativistic.
For a particle of mass m and charge q, in an electromagnetic field described by the
scalar potential ϕ and vector potential A = (Ax, Ay, Az), the Pauli equation reads:

¶ ˆ é 1 2 ù
ih y = H y = ê
¶t
( σ × ( ˆ
p - q A ) ) + qf ú y
ë 2m û
where σ = (σx, σy, σz) are the Pauli matrices collected into a vector for convenience,

æ0 1ö æ 0 -i ö æ1 0 ö
sx = ç ÷ sy = ç ÷ sz = ç ÷
è 1 0 ø è i 0 ø è 0 - 1 ø
æy + ö
and |ψ> is a two component spinor wavefunction given as y = ç ÷
èy - ø

11
Using the following identity (HW 1), ( σ × a )( σ × b ) = a × b + iσ × ( a ´ b )
2 2
( σ × ( pˆ - qA ) ) = ( pˆ - qA ) 1 + iσ × ( ( pˆ - qA ) ´ ( pˆ - qA ) )
Be careful that the latter term is not simply zero. Because p has the operator nature
pˆ ´ A y = -ihÑ ´ ( A y ) = -i h ( Ñ ´ A ) y + ihA ´ Ñ y = -ihB y - A ´ pˆ y

It suggests the following operator equation pˆ ´ A + A ´ pˆ = -ih ( Ñ ´ A ) = -ihB

Now the final result is

1 é 2
Hˆ = ( pˆ - qA ) - qhσ × B ù + qf
2m ë û
In this way, the change of physics due to spin is described as equations even with
the almost correct g-factor, 2, as described in the next slide.

The Pauli theory can be regarded as the low energy limit of the Dirac theory which
is explained in the next subsection (link).

12
In the classical picture, the magnetic moment of the orbital electron in a hydrogen
atom depends on its angular momentum L. The magnetic moment of a current
loop has the magnitude

ev 2 evr e
μ = iA = pr = =- L
2p r 2 2m
L = mrv

e
μ=- L
2m
In this regard, one may expect that the magnetic moment and potential term
corresponding to electron spin is,

e e
μ=- S U = - μ×B = S×B
2m 2m
ehBz
If electron spin (ħ/2) is in the +z direction, U =
4m
However, the Pauli equation (and Dirac equation) predicts potential twice larger.
Which one is right? In fact, it is 2.00231930436… times larger, and only QFT can
predict such a number.
13
 Dirac Equation (1928)
People still needed relativistic equation for the electron description. As we saw, the
problem of Klein-Gordon equation came from the 2nd order derivative in time.
Dirac searched for an equation consistent with the relativistic energy-momentum
formula, and yet first order in time.

Dirac’s basic strategy was to factor the energy-momentum relation, pμpμ − m2c2 =
0. This would be easy if we had only p0 (that is if p were zero):
2
( p0 ) - m2c 2 = 0 = ( p0 + mc )( p0 - mc ) ® p 0 - mc = 0 or p 0 + mc = 0

either one of which guarantees that pμpμ − m2c2 = 0. But it’s a different matter
when the other three components of pμ are included; in that case we are looking for
something of the form

p m pm - m 2 c 2 = ( b k pk + mc )( g l pl - mc )

I know students are starting to be confused. For this time, I will write out the full
components.

14
 0 2 1 2 2 2 3 2
( p ) -( p ) -( p ) -( p ) -m c 2 2

= ( b p - b p - b p - b p + mc )( g
0 0 1 1 2 2 3 3 0
p 0 - g 1 p1 - g 2 p 2 - g 3 p 3 - mc )
where βκ and γλ are eight coefficients to be determined. Multiplying out the right-
hand side, we have

p m pm - m 2 c 2 = b k g l pk pl - mc ( b k - g l ) pk - m 2 c 2
The linear term must vanish, so we must choose βκ = γλ. To finish the job, we need
to find coefficients γκ such that
p m pm = g k g l pk pl
Which is to say
2 2 2 2
( p0 ) - ( p1 ) - ( p 2 ) - ( p3 )
0 2 0 2 1 2 1 2 2 2 2 2 3 2 3 2
= (g ) ( p ) + (g ) ( p ) + (g ) ( p ) + (g ) ( p )
+ (g g + g g ) p p + (g g + g g ) p p + + (g g + g g ) p p
0 1 1 0 0 1 0 2 2 0 0 2 0 3 3 0 0 3

+ (g g + g g ) p p + (g g + g g ) p p + + (g g + g g ) p p
1 2 2 1 1 2 1 3 3 1 1 3 2 3 3 2 2 3

15
You see the problem. We could pick γ0 = 1 and γ1 = γ2 = γ3 = i, but there doesn’t
seem to be any way to get rid of the cross terms. At this point Dirac had a brilliant
inspiration: what if the γ’s are matrices, instead of numbers? Since matrices don’t
commute, we just might be able to find a set such that

0 2 1 2 2 2 3 2
(g ) = 1, (g ) = (g ) = (g ) = -1
g m g n + g n g m = 0, for m ¹ n
In a more compact and elegant expression:

{g m
, g n } = 2 g mn { A, B} º AB + BA

From now on, the curly bracket represents the “anticommutator” as defined above.

There are infinitely many answers to this equation, but none are smaller than 4×4.
Let us use the standard “Bjorken and Drell” convention:
i
0 æ1 0 ö i æ 0 s ö æ1 0ö æ0 0ö
g =ç ÷ g =ç i ÷ where 1= ç ÷, 0=ç ÷
è 0 -1 ø è -s 0 ø è0 1ø è 0 0 ø
and σi (i = 1,2,3) are the Pauli matrices.
16
As a 4×4 matrix equation, then, the relativistic energy-momentum relation does
factor:
p m pm - m 2 c 2 = ( g k pk + mc )( g l pl - mc )

The conventional choice of the Dirac equation is g m pm - mc = 0

By substituting the energy-momentum four-vector to differentiation,

ihg m ¶ my - mcy = 0 Dirac Equation
æy 1 ö
ç ÷
y
Do not forget that ψ is now a four-element column matrix. y = ç 2 ÷
çy 3 ÷
ç ÷
èy 4 ø
We call it a bispinor or Dirac spinor. Although it carries four components, this
object is not a four-vector and it does not follow the ordinary Lorentz transform.

17
Solution to the Dirac Equation
Let’s now look for simple solutions to the Dirac equation. Suppose first that ψ is
independent of position:

¶y ¶y ¶y
= = =0
¶x ¶y ¶z

This describes a state with zero momentum (p = 0). The Dirac equation reduces to

ih 0 ¶y
g - mcy = 0
c ¶t
In a more explicit form,

æ 1 0 ö æ ¶y A / ¶t ö mc 2 æy A ö æy 1 ö æy 3 ö
ç ÷ ç ¶y / ¶t ÷ = -i ç ÷ yA =ç ÷ yB = ç ÷
è 0 -1 øè B ø h èy B ø èy 2 ø èy 4 ø

¶y A mc 2 ¶y B mc 2
Thus, = -i yA =i yB
¶t h ¶t h

and the solutions are y A ( t ) = e ( )y ( )y

- i mc 2 / h t i mc 2 / h t
A ( 0) y B (t ) = e B ( 0)
18
It is natural to think the factor exp(−iEt/ħ) as the characteristic time dependence of
a quantum state with energy E. For a particle at rest, E = mc2, so ψA is exactly what
we should have expected. But what about ψB? It represents a state with negative
energy bearing the problem I mentioned earlier. Even with Dirac’s technique, the
headache does not disappear.
You might be tempted to set ψB = 0 always and forget about them. Unfortunately,
in a quantum system, we need a complete set of states and the positive energy
states by themselves are not complete.
Dirac’s original interpretation was that there are unseen infinite sea of negative
energy particles (electrons), which fill up all those unwanted states. The
mathematically equivalent interpretation is that the empty hole in the sea is
equivalent to antiparticles (positrons) with positive energy. Even though the latter
is the modern interpretation, it is totally up to you.
I know you are not convinced with these statements. Do not worry too much; we
will come back to this issue later for the spin 1/2 particle quantization. One last
comment: why do we have exp(+iEt/ħ) for the antiparticle? Because (in the most
elegant interpretation) the antiparticles are not traveling to the future, they are
normal(?) particles which happen to travel to the past.
In this way, ψA describes electrons, whereas ψB describes positrons. Each is a two-
component spinor, which is just right for a system of spin 1/2. In conclusion, the
Dirac equation with p = 0 admits four independent solutions.
19
 æ1ö æ0ö æ0ö æ0ö
ç ÷ ç ÷ ç ÷ ç ÷
- i ( mc 2 / h )t 0 - i ( mc 2 / h )t 1 i ( mc 2 / h )t 0 i ( mc 2 / h )t 0
y (1) =e ç ÷ y ( 2) =e ç ÷ y ( 3) =e ç ÷ y ( 4) =e ç ÷
ç0÷ ç0÷ ç1÷ ç0÷
ç ÷ ç ÷ ç ÷ ç ÷
è0ø è0ø è0ø è1ø
They describe, respectively, an electron with z-directional spin up, an electron with
spin down, a positron with spin down, and a positron with spin up. (No, there is no
typo. A lack of particle with spin up is equivalent to an antiparticle with spin
down.)

The next easiest solution we can look for is the plane wave, which has the
following four-vector form,
- ik × x - ik m x m
y ( x ) = Ae u( k ) = Ae u(k )

We are hoping to find a four-vector kμ and an associated bispinor u(k) such that
ψ(x) satisfies the Dirac equation. Putting its derivative into the Dirac equation, we
get
¶y
¶ my = m
= -ik my hg m k my - mcy = 0
¶x
or hg m k m e - ik × x u - mce - ik × x u = 0 ( hg m
k m - mc ) u = 0
20
Notice that it is a purely algebraic equation with no derivatives. Now

m 0 0 æ1 0 ö
0 æ 0 σö
g km = g k - γ × k = k ç ÷ - k ×ç ÷
è 0 - 1 ø è -σ 0 ø
0 æ1 0 ö 1 æ 0 s1 ö 2 æ 0 s 2 ö 3 æ 0 s 3 ö
=k ç ÷-k ç 1 ÷-k ç 2 ÷-k ç 3 ÷
è 0 -1 ø è -s 0 ø è - s 0 ø è -s 0 ø

æ hk - mc
0
- hk × σ ö æ u A ö çæ ( hk 0
- mc ) u A - hk × σuB ö
so ( hg km - mc ) u = 0 = ç hk × σ -hk 0 - mc ÷ ç u ÷ = ç
m ÷
ø è B ø è hk × σu A - ( hk + mc ) uB ÷ø
0
è

where, as before, the subscript A denotes the upper two components and B stands
for the lower two. The non-trivial solution satisfies,

k ×σ k ×σ
uA = uB and uB = uA
k - mc / h
0
k + mc / h
0

uA =
(k × σ ) uA
Combining them, 0 2 2
( k ) - ( mc / h )
21
We need to find some important formula for this calculation and for the future,

æ 0 1 ö 2 æ 0 -i ö 3 æ 1 0 ö æ k 3
1 k 1 - ik 2 ö
k ×σ = k ç ÷+k ç ÷+k ç ÷=ç 1 2 ÷
è1 0ø èi 0 ø è 0 -1ø è k + ik -k 3 ø

2 æ k3 k 1 - ik 2 ö æ k 3 k 1 - ik 2 ö 2
(k × σ ) =ç 1 2 ÷ç ÷= k 1
è k + ik - k 3 ø è k 1 + ik 2 3
-k ø

Now the solution we find is

2
k 0 2 2 2 2
uA =
0 2 2
uA ( k ) - ( mc / h ) =k k k m = ( mc / h )
m

( k ) - ( mc / h )
It means ħkμ must be a four-vector whose square is m2c2. Of course we know such
a quantity: the energy-momentum four-vector. However, one has to be careful that
there are actually two solutions:
k m = ± pm / h
The positive sign (time dependence exp(−ik0t) = exp(−iEt/ħ)) is associated with
particle states, and the negative sign (time dependence exp(+iEt/ħ)) with
antiparticle states.
22
Returning to the original solution, we are still obligated to construct four
independent solutions to the Dirac equation:

k ×σ k ×σ
uA = uB and uB = uA k m = ± pm / h
k - mc / h
0
k + mc / h
0

æ1ö p × σ æ1ö c æ pz ö
(1) Pick u A = ç ÷ : uB = 0 ç ÷ = 2 ç ÷
0
è ø p + mc 0
è ø E + mc p
è x + ip yø

æ 0ö p × σ æ 0ö c æ p x - ip y ö
(2) Pick u A = ç ÷ : uB = 0 =
è1ø p + mc è 1 ø E + mc 2 çè - pz ÷ø
ç ÷

For (1) and (2) we were obliged to choose the plus sign for the kμ equation because
otherwise uB solution will blow up as p → 0 (both the denominator and numerator
goes to zero but the limiting value is infinity).

In other words, we made a clever choice that (1) and (2) represent only particle
solutions excluding antiparticles. It is not an obligation, but it is a very convenient
choice with clear physical meanings.

23
 æ1ö p × σ æ1ö c æ pz ö
(3) Pick uB = ç ÷ : u A = 0 ç ÷ = 2 ç ÷
0
è ø p + mc 0
è ø E + mc è p x + ip y ø
æ 0ö p × σ æ 0ö c æ p x - ip y ö
(4) Pick uB = ç ÷ : u A = 0 ç ÷ = 2 ç ÷
1
è ø p + mc 1
è ø E + mc è - p z ø
In the same way, we were obliged to choose the minus sign for (3) and (4), and
these are antiparticle states.
There is one trick most students wouldn’t notice. In adopting ħkμ =±pμ, if k0 turns
out to be negative, p0 is positive, and −ħk0 = p0 is chosen as the “antiparticle”
energy E so that the denominators in the above equations are always positive.
In fact, this is not the most natural way to develop the algebra. (You wouldn’t
choose this method at your first try.) This cheating is justifiable because we
anticipate the emergence of antiparticles in the end. The first version of Griffith’s
book actually follow the natural (but more painful) way which, of course, will
make the same conclusion in the end.
A convenient normalization for these spinors is (though this is not the only choice)
u †u = 2 E / c
Here the dagger means the Hermitian conjugate,
24
 æa ö
ç ÷
b 2 2 2 2
u =ç ÷ u † = (a * b* g * d*) u †u = a + b + g + d
çg ÷
ç ÷
èd ø

By using the normalizing factor (see HW problem) N= ( E + mc ) / c

2

the four canonical solutions become

æ 1 ö æ 0 ö æ c ( px - ip y ) ö æ c ( pz ) ö
ç ÷ ç ÷ ç ÷ ç 2 ÷
ç 0 ÷ ç 1 ÷ ç E + mc ÷
2
ç E + mc ÷
ç c ( pz ) ÷ ç c ( p - ip ) ÷ ç c (- p ) ÷ ç c ( px + ip y ) ÷
u (1) = Nç 2 ÷, u (2) = Nç x y
÷, v (1) = Nç z
÷, v (2) = -N ç ÷
2 2
ç E + mc ÷ ç E + mc ÷ 2
ç E + mc ÷ ç E + mc ÷
ç c ( p + ip ) ÷ ç c (- p ) ÷ ç 0 ÷ ç 1 ÷
x y z
ç ÷ ç ÷ ç ÷ ç ÷
è E + mc ø 2
è E + mc ø 2
è 1 ø è 0 ø

y = ae - ip×x / hu ( particles ) , y = aeip× x / h v ( antiparticles )

One may wish that the last two elements are zero for particle states and the first
two elements are zero for antiparticle states. However, such a separation is not
possible. (In the non-relativistic limit, they indeed become small, and the Pauli
theory is recovered.)
25
It is customary, from here on, to use the letter v for antiparticles (and to include a
minus sign in v(2)), as indicated. Then, the particle satisfies the momentum space
Dirac equation,
(g m
pm - mc ) u = 0

Antiparticle satisfies (g m
pm + mc ) v = 0
Be careful that our original kμ satisfies the original Dirac equation for both particle
and antiparticle cases, but after interpreting −ħkμ as the energy-momentum of an
antiparticle, our pμ for antiparticle cannot satisfy the original equation.

You might guess that u(1) describes an electron with z-directional spin up, u(2) an
electron with spin down, and so on, but this is not quite the case. For Dirac
particles the spin operator in matrix form is
h æσ 0ö
S= Σ Σ=ç ÷
2 è 0 σ ø

Note that the meaning of the vector form is that for Sz, use Σz with σz. It’s not
difficult to check that u(1), for instance, is not an eigenstate of Σz. However, if we
select z axis in the direction of the motion (px = py = 0), then u(1), u(2), v(1), v(2) are
eigenspinors of Sz. u(1) and v(1) are spin up, u(2) and v(2) are spin down.

26
Once the description of the free particle is over, one may feel like to go into a
more complicated system. It is possible (but not so easy) to apply the Dirac
equation to Fermions in a potential field. One important consequence of the Dirac
equation is that it predicts the electron g-factor value g = 2, much better than g = 1,
but still there is a room for improvement.

In this way, the Dirac equation naturally describes spin 1/2 particles. This is how
the presence of antiparticles is predicted, and later(!) it was confirmed by
experiments.

Bilinear Covariants
I mentioned earlier that the components of a Dirac spinor do not transform as a
four-vector, when you go from one inertial system to another. The transformation
rule is introduced here. (for the proof, see HW problem) If we go to a primed
system moving with speed v in the x direction

y ® y ¢ = Sy
Considering the matrix nature of ψ, S must be a 4×4 matrix. The answer is

27
 æ a+ 0 0 a- ö
ç ÷
0 1 æ a+ a-s 1 ö ç 0 a+ a- 0÷
S = a+ + a-g g = ç ÷=ç
è a-s 1 a+ ø 0 a- a+ 0÷
ç ÷
è a- 0 0 a+ ø
1 v2
with a± = ± ( g ± 1) g = 1/ 1 - 2
2 c
Suppose we want to construct a Lorentz scalar quantity out of a spinor ψ.
Everyone’s first choice would be
æy 1 ö
ç ÷
y
* ç 2÷ 2 2 2 2
y y = (y 1 y 2 y 3 y 4 )
† * * *
= y1 + y 2 + y 3 + y 4
çy 3 ÷
ç ÷
èy 4 ø
Unfortunately, this is not a Lorentz scalar, as you can check by applying the
preceding transformation rule: (note (AB)†= B†A†)
æ v ö
ç 1 - s1 ÷
¢ c
†
¢†
¢
†
(y y ) = y y = ( Sy ) Sy = y † S † Sy ¹ y †y S †S = S 2 = g ç ÷
ç- vs 1 ÷÷
ç 1
è c ø
28
So, you cannot add all of them, but they are not four-vectors so inverting sign for
the first one does not help. In order to construct the correct rule, we now introduce
the adjoint spinor:
y = y †g 0 = (y 1* y 2* -y 3* -y 4* )
Then, the relativistically invariant quantity is (S†γ0S = γ0, proof omitted)
2 2 2 2
yy = y †g 0y = y 1 + y 2 - y 3 - y 4
¢ ¢† 0 † 0
(yy ) = y g y = ( Sy ) g Sy
¢ = y † † 0
S g Sy = y † 0
g y = yy
Earlier we learned to distinguish scalars and pseudoscalars, according to their
behavior under the parity transformation. It is natural to ask whether the above
quantity is the former type or the latter. First, we need to know how Dirac spinors
transform under P: (x, y, z) → (−x, −y, −z). The result is (the proof is a HW)
y ® y ¢ = g 0y
¢ † 0 0 0 † 0 0
(yy ) = y ¢ g y ¢ = (g y ) g g y
† 0 0
= y †
( ) g g y = y †g 0y = yy
g
So, it is a true scalar (we used γ0† = γ0 and (γ0)2 = 1). We can also make a
pseudoscalar out of ψ.
5 5 0 1 2æ0 1ö 3
yg y g º ig g g g = ç ÷
è1 0ø 29
It is Lorentz invariant (proof omitted). As for its behavior under parity,

(yg y )¢ = y ¢ g
5 † 0
g 5y ¢ = y †g 0g 0g 5g 0y = y †g 5g 0y = -y †g 0g 5y = - (yg 5y )
So it’s a pseudoscalar. We need to check the anticommutator as follows, (do you
still remember {γμ, γν} = 2gμν?)
3
g 5g 0 = ig 0g 1g 2g 3g 0 = ( -1) g 0 ( ig 0g 1g 2g 3 ) = -g 0g 5

All told, there are 16 products of the form ψ*iψj, since i and j run from 0 to 3.
These 16 products can be added together in various linear combinations to
construct quantities with distinct transformation behavior, as follows:
(1) yy = scalar (one component)
(2) yg 5y = pseudoscalar (one component)
(3) yg my = vector (four component)
(4) yg mg 5y = pseudovector (four component)
(5) ys mny = antisymmetric tensor (six component)
i m n
where s mn º
2
( g g - g n g m ) antisymmetric tensor
This gives 16 independent terms, so it’s all we can hope to make in this way.
30
It means, if you’re looking for a polar vector from ψ and ψ*, the above one is the
only candidate.

Another way to think of it is this: 1, γ5, γμ, γμγ5, and σμν constitute a “basis” for the
space of all 4×4 matrices; any 4×4 matrix can be written as a linear combination
of these 16. In particular, if you ever encounter a product of five γ matrices, say,
you may be sure that it can be reduced down to a product of no more than two.

The table is not so difficult in that you can immediately guess each item’s behavior.
The number of indices tells you if it is a scalar, vector, or second rank tensor, and
the presence of γ5 determines its behavior under parity. Be careful that γμ itself
does not transform when you go to a different inertial system – it’s ψ that changes.

You may wonder why do we really care the matrix γ5 which is not necessary in the
spinor description itself. Think it this way. If γ5 is included in the description of a
physical interaction, it means our world is not symmetric under parity inversion, r
→ –r.

We need to learn it because our world is NOT symmetric. At least the weak
interaction has maximally broken parity symmetry. All other forces may not break
this symmetry.

31
Homeworks (page 1 of 2)
1. Prove the identity, ( σ × a )( σ × b ) = a × b + iσ × ( a ´ b )
2. Calculate the normalization factor N of the plane wave bi-spinor solution.

3. (a) For the plane wave solution, if the z axis points along the direction of motion,
show that

æ ( E + mc ) / c ö÷
2
ç
ç 0 ÷
(1)
u =ç ÷
ç
ç
( E - mc ) / c ÷÷
2

ç 0 ÷
è ø

(b) What is the spin value of the above spinor? Find the eigenvalue.

32
Homeworks (page 2 of 2)
4. Confirm the transformation rule for bispinors. Here is the method. We want to
carry solutions to the Dirac equation in the original frame to solutions in the
primed frame:
ihg m ¶ my - mcy = 0 « ihg m ¶¢my ¢ - mcy ¢ = 0

¶ ¶xn ¶ ¶xn
where y ¢ = Sy and ¶¢m = m
= m n
= ¶
m n
¶x¢ ¶x¢ ¶x ¶x¢
n
¶x
It follows that ( S -1g m S ) ¶x¢m = g n
5. Derive the following transformation rule for parity, using the method of the
above problem.
y ® y ¢ = g 0y

6. Calculate S†S. and confirm the its shape given in this note.

33