Spin ½ of the Dirac Particle

When studying the spin of a Dirac particle it is best to use the Heisenberg
representation, where operators are time dependent and whose time evolution is
given by
𝑑𝒪 𝑖
= 𝐻, 𝒪
𝑑𝑡 ℏ
We will also use index notation since that is more general and powerful and
generalizes to the case of 4-dimensional space-time (4-vectors, tensors and their
contractions etc.)
Hence the 𝑖-th component of the orbital angular momentum operator is given by

𝐋𝑖 = 𝐫 × 𝐩 𝑖
= 𝜖 𝑖𝑗𝑘 𝑥𝑗 𝑝𝑘 = −𝑖ℏ𝜖 𝑖𝑗𝑘 𝑥𝑗 𝜕𝑘

And the fundamental commutation relation is written as 𝑥 𝑖 , 𝑝 𝑗 = 𝑖ℏ 𝛿 𝑖𝑗

Let us first recall the proof of the conservation of angular momentum for a
spinless particle obeying the non-relativistic Schrodinger equation in a central
potential

The Hamiltonian is
𝐩2
𝐻= +𝑉 𝑟
2𝑚
where 𝑟 = 𝐫 (Here 𝑉 𝑟 is the potential energy)

To see the conservation of 𝐋 consider the commutator 𝐿𝑖 , 𝐻 :

1 𝑖 2
𝐿𝑖 , 𝐻 = 𝐿 , 𝐩 + 𝐿𝑖 , 𝑉 𝑟
2𝑚
 𝐿𝑖 , 𝐩2 = 𝐫 × 𝐩 𝑖 , 𝑝ℓ2

= 𝜖 𝑖𝑗𝑘 𝑥𝑗 𝑝𝑘 , 𝑝ℓ2

= 𝜖 𝑖𝑗𝑘 𝑥𝑗 𝑝𝑘 𝑝ℓ2 − 𝑝ℓ2 𝑥𝑗 𝑝𝑘

Since each component of 𝐩 commutes with it’s other components 𝑝𝑘 𝑝ℓ2 = 𝑝ℓ2 𝑝𝑘

∴ 𝐿𝑖 , 𝐩2 = 𝜖 𝑖𝑗𝑘 𝑥𝑗 𝑝ℓ2 − 𝑝ℓ2 𝑥𝑗 𝑝𝑘

Now adding and subtracting 𝑝ℓ 𝑥𝑗 𝑝ℓ the middle term in brackets becomes,
𝑥𝑗 𝑝ℓ2 − 𝑝ℓ2 𝑥𝑗 = 𝑥𝑗 𝑝ℓ − 𝑝ℓ 𝑥𝑗 𝑝ℓ − 𝑝ℓ 𝑝ℓ 𝑥𝑗 − 𝑥𝑗 𝑝ℓ

= 𝑖ℏ𝛿𝑗ℓ 𝑝ℓ + 𝑝ℓ 𝑖ℏ 𝛿𝑗ℓ
= 2 𝑖ℏ 𝑝𝑗
So the first commutator becomes
𝐿𝑖 , 𝐩2 = 𝜖 𝑖𝑗𝑘 2 𝑖ℏ 𝑝𝑗 𝑝𝑘

=0
Now to calculate the commutator of 𝐿𝑖 with 𝑉 𝑟 :

𝐿𝑖 , 𝑉 𝑟 = 𝜖 𝑖𝑗𝑘 𝑥𝑗 𝑝𝑘 𝑉 𝑟 − 𝑉 𝑟 𝑥𝑗 𝑝𝑘

= −𝑖ℏ𝜖 𝑖𝑗𝑘 𝑥𝑗 𝜕𝑘 𝑉 𝑟 − 𝑉 𝑟 𝑥𝑗 𝜕𝑘

= −𝑖ℏ𝜖 𝑖𝑗𝑘 𝑥𝑗 𝜕𝑘 𝑉 𝑟 + 𝑥𝑗 𝑉 𝑟 𝜕𝑘 − 𝑉 𝑟 𝑥𝑗 𝜕𝑘

= −𝑖ℏ 𝜖 𝑖𝑗𝑘 𝑥𝑗 𝜕𝑘 𝑉 𝑟
 𝜕𝑉 𝑟
𝜕𝑘 𝑉 𝑟 =
𝜕𝑥 𝑘

𝜕𝑉 𝑟 𝜕𝑟
=
𝜕𝑟 𝜕𝑥 𝑘

𝜕𝑉 𝑟 𝜕 𝑥 𝑚 𝑥𝑚 1/2
=
𝜕𝑟 𝜕𝑥 𝑘

𝜕𝑉 𝑟 1
= 2𝛿𝑘𝑚 𝑥𝑚
𝜕𝑟 2 𝑟

1 𝜕𝑉 𝑟
= 𝑥𝑘
𝑟 𝜕𝑟
Thus
𝑖 𝑖𝑗𝑘
1 𝜕𝑉 𝑟
𝐿 ,𝑉 𝑟 = −𝑖ℏ 𝜖 𝑥𝑗 𝑥𝑘
𝑟 𝜕𝑟

=0

Hence 𝐿𝑖 , 𝐩2 = 𝐿𝑖 , 𝑉 𝑟 = 0 and therefore 𝐿𝑖 , 𝐻 = 0 and orbital angular

momentum is conserved in time in a spherically symmetric potential
We now perform the same calculation using the Dirac equation
The Dirac Hamiltonian in a central potential is

𝐻𝐷 = 𝑐 𝛂 ⋅ 𝐩 + 𝑚𝑐 2 𝛽 + 𝑉 𝑟

To find the time dependence of 𝐿𝑖 we need to calculate the commutator

𝐿𝑖 , 𝐻𝐷 = 𝑐 𝛼 𝑗 𝑝 𝑗 , 𝐿𝑖 + 𝑚𝑐 2 𝐿𝑖 , 𝛽 + 𝐿𝑖 , 𝑉 𝑟

We have already shown that 𝐿𝑖 , 𝑉 𝑟 =0

Since 𝛽 is a constant matrix (and the fact that they act on different spaces) gives
𝐿𝑖 , 𝛽 = 0
(Also 𝛽 acts on a different Hilbert space, the 4-dimensional spinor space, whereas,
𝐿𝑖 acts on wavefunctions defined in the coordinate space)
Thus we are left with
𝐿𝑖 , 𝐻𝐷 = 𝑐 𝐿𝑖 , 𝛼 ℓ 𝑝ℓ

= 𝑐 𝛼 ℓ 𝐿𝑖 , 𝑝ℓ

= 𝑐𝛼 ℓ 𝜖 𝑖𝑗𝑘 𝑥 𝑗 𝑝𝑘 , 𝑝ℓ

= 𝑐𝛼 ℓ 𝜖 𝑖𝑗𝑘 𝑥 𝑗 , 𝑝ℓ 𝑝𝑘

= 𝑐𝛼 ℓ 𝜖 𝑖𝑗𝑘 𝑖ℏ𝛿 𝑗ℓ 𝑝𝑘
= 𝑖ℏ 𝑐𝜖 𝑖𝑗𝑘 𝛼 𝑗 𝑝𝑘 = 𝑖 ℏ𝑐 𝛂 × 𝐩 𝑖
Thus 𝐿𝑖 does not commute with the Hamiltonian,

𝐋 , 𝐻𝐷 = 𝑖ℏ𝑐 𝛂 × 𝐩 ≠ 0

and consequently the orbital angular momentum is not a constant of motion

We suspect the presence of another contribution to the angular momentum, i.e., the
intrinsic spin of the particle

Consider the 4 × 4 matrix 𝚺, defined by

𝝈 0
𝚺≡
0 𝝈
Also define another matrix, which we will suggestively call 𝐒, as

1
𝐒= ℏ𝚺
2

Now consider the commutator of 𝐒 𝑖 with 𝐻𝐷

1
𝐒 𝑖 , 𝐻𝐷 = ℏ Σ 𝑖 , 𝐻𝐷
2

1
= ℏ[Σ 𝑖 , 𝑐 𝛂 ⋅ 𝐩 + 𝑚𝑐 2 𝛽 + 𝑉 𝑟 ]
2
 1
Neglecting the ℏ for now, the commutator of 𝚺 with 𝐻𝐷 is
2

[Σ 𝑖 , 𝑐 𝛂 ⋅ 𝐩 + 𝑚𝑐 2 𝛽 + 𝑉 𝑟 ]

= Σ 𝑖 , 𝑐 α𝑗 𝑝 𝑗 + 𝑚𝑐 2 Σ 𝑖 , 𝛽 + Σ 𝑖 , 𝑉 𝑟

Now 𝚺 is a 4 × 4 matrix acting in the spinor space and has no position and
derivative dependence. Therefore,
Σ𝑖 , 𝑉 𝑟 = 0
The 2nd term is
𝑖 1 0 1 0 𝑖
𝑖
Σ ,𝛽 = 𝜎 0 − 𝜎 0
0 𝜎 𝑖 0 −1 0 −1 0 𝜎 𝑖
 𝜎 𝑖 0 − 𝜎𝑖 0 =0
=
0 −𝜎 𝑖 0 −𝜎 𝑖
Hence we are left with

Σ 𝑖 , 𝑐 α𝑗 𝑝 𝑗 = 𝑐 Σ 𝑖 , α𝑗 𝑝 𝑗

(Again remember that the derivative momentum operator and matrices 𝚺, 𝛂 act on
different spaces and thus commute)
𝑖 𝑗 𝑗 𝑖
𝑖 𝑗
Σ ,α = 𝜎 0 0 𝜎 − 0 𝜎 𝜎 0
0 𝜎𝑖 𝜎𝑗 0 𝜎𝑗 0 0 𝜎𝑖

= 𝑖0 𝑗 𝜎𝑖 𝜎𝑗 − 0 𝜎𝑗 𝜎𝑖
𝜎𝜎 0 𝜎𝑗 𝜎𝑖 0
Now 𝜎 𝑖 𝜎 𝑗 − 𝜎 𝑗 𝜎 𝑖 = 2𝑖𝜖 𝑖𝑗𝑘 𝜎 𝑘
∴ 𝑖
Σ ,α 𝑗
= 2𝑖𝜖 𝑖𝑗𝑘 0 𝜎𝑘
𝜎𝑘 0

= 2𝑖𝜖 𝑖𝑗𝑘 𝛼 𝑘
And consequently
𝑐 Σ 𝑖 , α𝑗 𝑝 𝑗 = 2𝑖𝑐 𝜖 𝑖𝑗𝑘 𝑝 𝑗 𝛼 𝑘

= 2𝑖𝑐 𝐩 × 𝛂 𝑖

In vector notation
𝚺, 𝑐 𝛂 ⋅ 𝐩 = 2𝑖𝑐 𝐩 × 𝛂

= −2𝑖𝑐 𝛂 × 𝐩
Thus finally,
1
𝐒, 𝐻𝐷 = ℏ 𝚺, 𝐻𝐷 = 𝑖ℏ𝑐 𝐩 × 𝛂
2
= −𝑖ℏ𝑐 𝛂 × 𝐩
Recall the non-conservation of orbital angular momentum:
𝐋 , 𝐻𝐷 = 𝑖ℏ𝑐 𝛂 × 𝐩

= − 𝐒, 𝐻𝐷
We are naturally led to define the total angular momentum operator
𝐉=𝐋+𝐒
So that 𝐉 is conserved

𝐉, 𝐻𝐷 = 𝐋 , 𝐻𝐷 + 𝐒 , 𝐻𝐷 = 0
Thus the total (conserved) angular momentum operator for a spin ½ particle is

𝐉=𝐋+𝐒

1 𝝈 0
=𝐫×𝐩+ ℏ
2 0 𝝈

1 𝝈 0
= −𝑖ℏ 𝐫 × 𝛁 + ℏ
2 0 𝝈

The spin operator is block diagonal and is the familiar two dimensional spin-1/2
operator replicated twice so that it can act on 4-component spinors
As mentioned, the operators 𝐋 = −iℏ 𝐫 × 𝛁 and the spin operator 𝐒 act on
different spaces.
𝐋 being a differential operator acts on the space of infinite dimensional
wavefunctions which has coordinate dependence
𝐒 acts on the 4-dimensional spinor space, i.e., on the 4-component spinors
Hence to be really precise an equation of the form ‘𝐉 = 𝐋 + 𝐒’ should be written as
a sum of a (direct) tensor product
𝐉 = 𝐋 ⊗ 𝟏4×4 + 𝐈 ⊗ 𝐒
where 𝟏4×4 is the 4-dimensional identity matrix acting on the spinor space and 𝐈 is
the identity operator acting on wavefunctions
The state of a Dirac particle is thus a tensor product of the form

Ψ ∼ 𝜓 𝑡, 𝐱 ⊗ 4 Column Spinor
Each operator acts on their respective spaces:

𝐋 on the coordinate dependent 𝜓 𝑡, 𝐱 (which contains all the information about

position, momentum, energy, angular momentum etc)

𝐒 on the 4-component spinor (which contains the information of the spin state of
the particle and whether it is a particle or antiparticle etc.)

Since 𝐉 commutes with the Hamiltonian and obeys the commutation rules of the
usual angular momentum operators 𝐉 𝟐 , J𝑧 and 𝐻𝐷 can be simultaneously
diagonalized, as in non-relativistic quantum mechanics

Of course the components of 𝐉 do not commute with each other

While 𝑆𝑧 can be used to parametrize the spin states, it is not Lorentz invariant

Since spin is an intrinsic property of a particle, it seems reasonable that the

magnitude of the spin be invariant under Lorentz transformations and indeed 𝐒 2
being proportional to the identity operator is invariant

For 𝐩 ≠ 0 𝐒 is no longer a suitable operator since it fails to commute with even the
free particle Dirac Hamiltonian c 𝛂 ⋅ 𝐩 + 𝑚𝑐 2 𝛽 and hence it cannot be a given a
definite value at the same time as the energy

Another drawback to the rest frame spin eigenstates is that they cannot be used to
describe massless particles, since it is not possible to transform to the rest frame for
a massless particle
Still, there are two independent states for each energy, whether the particle is in
motion or not

Hence there must be some other operator which commutes with the energy
operator 𝐻𝐷 and momentum operator −𝑖ℏ𝛁 and whose eigenvalues can be used to
distinguish the two states

As discussed in class, such an operator is helicity (the choice of such an operator is

not unique)

The helicity operator ℎ is defined to the projection of the particle’s spin vector
along (or against) it’s momentum:
1 1 𝐩
ℎ =𝐒⋅𝐩 ෝ = ℏ𝚺⋅𝐩 ෝ= ℏ𝚺⋅
2 2 𝐩
Here 𝐩
ෝ is the unit vector along the direction of the particle’s momentum, not the
momentum operator

A short calculation done in class shows that the helicity commutes with
Hamiltonian ℎ, 𝐻𝐷 = 0 and thus can be used to distinguish the two states of the
same energy

1
The possible values of ℎ are ± and can be used to label the states and is therefore
2
a ‘good’ quantum number

ℎ is manifestly rotationally invariant and also transforms simply under Lorentz

transformations