Y. D. Chong Ch.

5: Quantum Electrodynamics | Graduate Quantum Mechanics

Chapter 5: Quantum Electrodynamics

This chapter gives an introduction to quantum electrodynamics, the quantum theory
of the electromagnetic field and its interactions with electrons and other charged particles.
We begin by formulating a quantum Hamiltonian for an electron in a classical electromag-
netic field. Then we study how to quantize Maxwell’s equations, arriving at a quantum field
theory in which the elementary excitations are photons—particles of light. The final step
is to formulate a theory in which electrons and photons are treated on the same quantum
mechanical footing, as excitations of underlying quantum fields. Along the way, we will see
how relativity can be accomodated with quantum theory.
Quantum electrodynamics is an extremely rich and intricate theory, and we will leave out
many important topics. Interested readers are referred to are Dyson’s 1951 lecture notes on
quantum electrodynamics [1], and Zee’s textbook Quantum Field Theory in a Nutshell [2].

5.1. QUANTIZATION OF THE LORENTZ FORCE LAW

5.1.1. Classical Lagrangian and Hamiltonian

Consider a charged particle in an electromagnetic field. As we are mainly interested in

the case where the particle in question is an electron, we will henceforth take the electric
charge to be −e, where e = 1.602 × 10−19 C. For particles with charge q, simply perform the
substitution e → −q in the formulas appearing in this chapter.
Let us formulate the Hamiltonian governing the quantum dynamics of such a particle,
subject to some simplifying assumptions: (i) the particle has charge and mass but no other
relevant mechanical properties (e.g., we ignore spin); (ii) the electromagnetic field is treated
as a classical field, meaning that the electric and magnetic field vectors are classical field
variables, not operators; and (iii) the mechanics is non-relativistic. We will later see how to
go beyond each of these simplifications.
Classically, the electromagnetic field acts on the particle via the Lorentz force law,
 
F(r, t) = −e E(r, t) + ṙ × B(r, t) , (5.1)

where r and ṙ denote the position and velocity of the particle, t is the time, and E and B
are the electric and magnetic fields. With no other forces present, and in the non-relativistic
limit, the equation of motion can be derived using Newton’s second law:
 
mr̈ = −e E(r, t) + ṙ × B(r, t) , (5.2)

where m is the particle’s mass.

We assert that Eq. (5.2) can be derived from the Lagrangian

1 h i
L(r, ṙ, t) = m|ṙ|2 + e Φ(r, t) − ṙ · A(r, t) , (5.3)
2

90
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

where Φ and A are the scalar and vector potentials, which are related to E and B by
∂A
E(r, t) = −∇Φ(r, t) − , (5.4)
∂t
B(r, t) = ∇ × A(r, t). (5.5)
Roughly speaking, Eq. (5.3) follows the usual prescription that the Lagrangian is “the kinetic
energy minus the potential energy.” In this case, the “potential energy” consists of a term
contributed by the electric potential, −eΦ, and a term eṙ · A that seems to describe a
coupling between the electron’s velocity and the vector potential.
To show that this is the right Lagrangian, we want to verify that the equation of motion
(5.2) is equivalent to the Euler-Lagrange equations
∂L d ∂L
= , (5.6)
∂ri dt ∂ ṙi
where ri denotes the i-th component of r. From Eq. (5.3), the partial derivatives of L are
∂L h X i
= e ∂i Φ − ṙj ∂i Aj , (5.7)
∂ri j
∂L
= mṙi − eAi , (5.8)
∂ ṙi
where ∂i ≡ ∂/∂ri . Next, we want to apply the total time derivative, d/dt, to Eq. (5.8). In
doing this, note that the A field felt by the particle varies with t via r(t), as well as the
t-dependence of the field itself. By the chain rule of total derivatives,
d ∂L d
= mr̈i − e Ai (r(t), t)
dt ∂ ṙi dt
∂Ai X (5.9)
= mr̈i − e −e ṙj ∂j Ai .
∂t j

Plugging Eqs. (5.7) and (5.9) into Eq. (5.6) gives

"  X  #
∂Ai 
mr̈i = −e −∂i Φ − + ṙj ∂i Aj − ∂j Ai . (5.10)
∂t j

We want to compare this to the equation of motion (5.2). The latter is expressed in terms
of E and B, so we first re-express it using Φ and A via Eqs. (5.4) and (5.5). The E term
then matches the first term in the square brackets in Eq. (5.10). The ṙ × B term, which
produces the magnetic force, needs a bit more manipulation. Using the Levi-Cevita symbol,
X
(ṙ × B)i = εijk ṙj Bk (5.11)
jk
!
X X
= εijk ṙj εkpq ∂p Aq (5.12)
jk pq
!
X X
= εijk εpqk ṙj ∂p Aq . (5.13)
jpq k

91
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

Next, we use the identity X

εijk εpqk = δip δjq − δiq δjp . (5.14)
k

This allows us to simplify Eq. (5.13) into

X
(ṙ × B)i = (ṙj ∂i Aj − ṙj ∂j Ai ) , (5.15)
j

which matches the second term in the square brackets in Eq. (5.10), as desired.
We now proceed to Hamiltonian mechanics. Following the usual procedure, we first
defining the canonical momentum, pi = ∂L/∂ ṙi . Using the Lagrangian (5.3), we obtain

mṙ = p + eA(r, t). (5.16)

The Hamiltonian is then defined as

H(r, p) = p · ṙ − L. (5.17)

We plug Eqs. (5.3) and (5.16) into Eq. (5.17), to express H in terms of r and p, without ṙ:
|p + eA|2
   
p + eA e
H =p· − + eΦ − (p + eA) · A (5.18)
m 2m m
|p + eA(r, t)|2
= − eΦ(r, t). (5.19)
2m
Let us compare the result (5.19) to the Hamiltonian for a non-relativistic particle in a scalar
potential,
|p|2
H= + V (r, t).
2m
The −eΦ term acts like a potential energy, which is no surprise. More interestingly, the
vector potential enters into the kinetic energy term, through the substitution

p → p + eA(r, t). (5.20)

Why does the vector potential manifest as some kind of momentum shift? This can
be traced back to Eq. (5.16), which says that in the presence of a vector potential, the
quantity mṙ—which we normally think of as the “kinetic momentum” giving rise to the
particle’s kinetic energy—differs from the canonical momentum p by a shift of eA. Recall
that the concept of canonical momentum is tied to Noether’s theorem, which states that any
symmetry of a system is associated with a conservation law. For translational symmetry,
the conserved quantity is the canonical momentum. Hamilton’s equation,
dpi ∂H
= ,
dt ∂ri
implies that if H is r-independent, dp/dt = 0. However, for a particle in a vector potential,
even if A is r-independent (i.e., translationally symmetric), mṙ is not necessarily conserved!
For example, consider the r-independent potentials

Φ(r, t) = 0, A(r, t) = Ctẑ, (5.21)

92
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

where C is some constant. Thanks to the t-dependence in the A potential, plugging these
potentials into Eqs. (5.4)–(5.5) gives a non-vanishing electric field:

E(r, t) = −C ẑ, B(r, t) = 0. (5.22)

Therefore, mṙ is not conserved:

d
(mṙ) = eC ẑ. (5.23)
dt
On the other hand, the canonical momentum defined in Eq. (5.16) is conserved:

dp d
= (mṙ − eA) = eC ẑ − eC ẑ = 0. (5.24)
dt dt

5.1.2. Quantum Hamiltonian

From the classical Hamiltonian (5.19), we can formulate the Hamiltonian operator for
a nonrelativistic quantum particle in an electromagnetic field. As usual, the quantization
procedure merely involves replacing the r and p variables with r̂ and p̂ operators:

|p̂ + eA(r̂, t)|2

Ĥ = − eΦ(r̂, t). (5.25)
2m

Both the scalar and vector potential terms are functions of the position operator r̂, so if we
want to expand the square in the first term of Eq. (5.25), it is important to note that p̂ and
A(r̂, t) need not commute:

2
Xh ih i
p̂ + eA(r̂, t) = p̂j + eAj (r̂, t) p̂j + eAj (r̂, t) (5.26)
j
X 
p̂2j + e p̂j Aj (r̂, t) + Aj (r̂, t)pj + e2 A2j (r̂, t)2 .
 
= (5.27)
j

For example, if we are working in the position representation (i.e., using wavefunctions),
then p̂ = −iℏ∇. When dealing with p̂j Aj (r, t), the gradient in the momentum operator
applies to everything on its right, i.e., both Aj and the wavefunction. By the product rule,
 
∂Aj ∂
p̂j Aj (r, t)ψ(r, t) = −iℏ + Aj ψ(r, t). (5.28)
∂rj ∂rj

Thus, in the position representation, Eq. (5.27) can also be written as

2 ℏ2 2
p̂ + eA =− ∇ − 2iℏeA · ∇ − iℏe (∇ · A) + e2 |A|2 . (5.29)
2m

93
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

5.1.3. Gauge symmetry

It is interesting that the Hamiltonian (5.25) is expressed in terms of the scalar and vector
potentials, Φ and A. In classical electromagnetism, it is known that Φ and A do not uniquely
define the E and B fields. Suppose we perform a gauge transformation on the potentials,

Φ(r, t) → Φ(r, t) − Λ̇(r, t) (5.30)

A(r, t) → A(r, t) + ∇Λ(r, t), (5.31)

where Λ(r, t) is an arbitrary scalar field. Plugging these into Eqs. (5.4) and (5.5), we find
that the E and B fields are unchanged; thus, the gauge transformation has no effect on the
dynamics of any charged particle interacting with the electromagnetic field.
In the quantum theory, this feature manifests as a subtle symmetry of the Hamiltonian,
called gauge symmetry. Let us insert the gauge transformed potentials (5.30)–(5.31) into
the Hamiltonian (5.25). This results in the transformed Hamiltonian

|p̂ + eA(r̂, t) + e∇Λ(r̂, t)|2

ĤΛ = − eΦ(r̂, t) + eΛ̇(r̂, t). (5.32)
2m
We will work in the position representation. Suppose the wavefunction ψ(r, t) is a solution
to the Schrödinger wave equation under the original Hamiltonian,
∂ψ
iℏ = Ĥψ(r, t). (5.33)
∂t
Then we assert that there is a corresponding solution for the gauge transformed Hamiltonian:
∂ψΛ
iℏ = ĤΛ ψΛ (r, t), (5.34)
∂t  
ieΛ(r, t)
ψΛ (r, t) = ψ(r, t) exp − . (5.35)
ℏ

To prove this, observe how time and space derivatives act on ψΛ :

      
∂ ieΛ ∂ψ ie ieΛ
ψ exp − = − Λ̇ ψ exp
∂t ℏ ∂t ℏ ℏ
       (5.36)
ieΛ ie ieΛ
∇ ψ exp − = ∇ψ − ∇Λ ψ exp .
ℏ ℏ ℏ
When the extra terms generated by the exp(ieΛ/ℏ) factor are slotted into the Schrödinger
equation, they cancel the gauge terms in the scalar and vector potentials. For example,
  
ieΛ
 h i 
ieΛ

− iℏ∇ + eA + e∇Λ ψ exp − = (−iℏ∇ + eA) ψ exp − (5.37)
ℏ ℏ
If we apply (−iℏ∇ + eA + e∇Λ) another time, it has a similar effect, but with the quantity
in square brackets on the right-hand side of Eq. (5.37) taking the place of ψ. Hence,
   h  
2 ieΛ 2
i ieΛ
− iℏ∇ + eA + e∇Λ ψ exp − = |−iℏ∇ + eA| ψ exp − . (5.38)
ℏ ℏ

94
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

By applying this observation to the left and right sides of Eq. (5.34), we find that it reduces
to the original Schrödinger equation (5.33).
The above result can be stated in a simpler form if the potential fields are static. In this
case, the time-independent electromagnetic Hamiltonian is

|p̂ + eA(r̂)|2
Ĥ = − eΦ(r̂). (5.39)
2m
By a time-independent gauge transformation, we can define

|p̂ + eA(r̂) + e∇Λ(r)|2

ĤΛ = − eΦ(r̂) (5.40)
2m
Then the gauge symmetry states that

ĤΛ ψΛ (r) = EψΛ (r),

Ĥψ(r) = Eψ(r) ⇔ (5.41)
 
ieΛ(r)

 ψΛ (r) = ψ(r) exp − .
ℏ

In other words, the gauge transformed Hamiltonian has the same energy spectrum, and its
eigenfunctions are the original eigenfunctions modified by a phase factor involving Λ(r).

5.1.4. The Aharonov-Bohm effect

The Hamiltonian’s gauge symmetry might tempt us to speculate that even though Ĥ is
expressed using Φ and A, maybe only E and B actually govern physical behavior, as is the
case in classical electrodynamics. But this turns out to be false: quantum electrodynamics
depends on the potentials, not just E and B. In particular, a charged particle can experience
B = 0 and yet be affected by the A field, a phenomenon called the Aharonov-Bohm effect.
A setup for realizing the Aharonov-Bohm effect is depicted below. We place a particle on a
ring, or “annulus,” of radius R and width d ≪ R. Outside the annulus, in the regions shaded
in gray, the scalar potential is Φ → −∞; this means the potential energy is −eΦ → ∞,
forcing the wavefunction to vanish. The wavefunction is thus confined to the annulus, where
we set the scalar potential to be Φ = 0.

We assume all fields and wavefunctions to be independent of z (the out-of-plane direction).

The in-plane position can be expressed using polar coordinates (r, ϕ), such that the annulus
is centered at the coordinate origin.

95
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

Suppose we use an infinitesimally thin solenoid to thread a magnetic flux ΦB through the
origin. This can be described by the vector potential
ΦB
A(r, ϕ) = eϕ , (5.42)
2πr
where eϕ is the unit vector pointing in the azimuthal direction. With this choice of A field,
the magnetic flux through a loop of radius r around the origin is
I  
ΦB
A · dr = · (2πr) = ΦB , (5.43)
2πr
independent of r. Thus, the magnetic flux density is entirely concentrated at the origin.
Away from the origin, we have A ̸= 0 and B = 0.
Plugging this A field into the Hamiltonian (5.39), we arrive at the time-independent
Schrödinger wave equation
2
1 eΦB
−iℏ∇ + eϕ ψ(r, ϕ) = Eψ(r, ϕ). (5.44)
2m 2πr
Due to rotational symmetry, the eigenfunctions should have the form
ψ(r, ϕ) = ψ0 A(r) einϕ , n ∈ Z. (5.45)

Since the annulus is very thin (d ≪ R), let us zoom in on a representative local segment
at ϕ = 0, as shown below:

Locally, the annulus looks like a straight waveguide. Let us adopt local Cartesian coordinates
x = r − R and y = Rϕ. Now A points in the y direction, so Eq. (5.44) reduces to
" 2 #
ℏ2 ∂ 2

1 ∂ eΦB
− + −iℏ + ψ(x, y) ≈ Eψ(x, y). (5.46)
2m ∂x2 2m ∂y 2πR

Meanwhile, the ansatz (5.45) reduces to

ψ(x, y) = A(x)einy/R , (5.47)
which is a waveguide mode with transverse profile A(x) and wavenumber ky = n/R. This
allows us to substitute ∂/∂y → in/R in Eq. (5.46), giving
" 2 #
ℏ2 ∂ 2

1 nℏ eΦB
− + + A(x) = EA(x). (5.48)
2m ∂x2 2m R 2πR

96
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

Since the wavefunction vanishes on the inner and outer walls of the annulus, the transverse
profile must satisfy
A(±d/2) = 0. (5.49)
We pick the solution A(x) = cos(πx/d), which corresponds to the fundamental waveguide
mode. Applying this to Eq. (5.48) yields
" 2  2 #
1 nℏ eΦB πℏ
E= + + (5.50)
2m R 2πR d
2
e2 π 2 ℏ2

nh
= 2 Φ B + + . (5.51)
8π mR2 e 2md2

There is an infinite set of solutions, indexed by n ∈ Z. For each n, the energy varies
quadratically with the magnetic flux ΦB , with an n-dependent ΦB offset and a constant
energy offset. The spectrum is plotted against ΦB below:

Evidently, varying ΦB affects the energy eigenvalues, in spite of the fact that the electron
resides in the annulus where B = 0. This is the Aharonov-Bohm effect.
A further noteworthy feature of this spectrum is that if ΦB changes by a multiple of the
constant h/e (called the magnetic flux quantum), the spectrum is left unchanged. This
invariance property turns out to be tied to gauge symmetry. Eq. (5.42) says that if an extra
flux nh/e (where n ∈ Z) is threaded through the annulus, the vector potential changes by
∆A = (nℏ/er)eϕ . But we can undo this via the following gauge transformation:
(
nℏ ∇Λ = −(nℏ/er) eϕ
Λ(r, ϕ) = − ϕ ⇒ −ieΛ/ℏ (5.52)
e e = einϕ .

Note that Λ is not single-valued, but that’s not a problem, as both ∇Λ and exp(−ieΛ/ℏ),
which enter into the gauge symmetry relations (5.32)–(5.35), are single-valued.
Remarkably, this gauge argument does not require the electron to live on an annulus; we
can consider a confinement region of any shape, so long as it encircles the magnetic flux. In

97
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

general, the energy spectrum need not be given by Eq. (5.51), but the gauge argument says
that it must be invariant under a flux change of ∆ΦB = nh/e.
This invariance property is also closely related to a famous argument by Dirac, showing
that if there exists a magnetic monopole—a hypothetical particle that produces a magnetic
field B = (µ/4πr2 ) er —then its magnetic charge µ must be a multiple of h/e. Conversely,
if there exists a single magnetic monopole of charge µ anywhere in the universe, electric
charges must come in multiples of h/µ, which could be regarded as an explanation for the
quantization of electric charge [3].

5.2. DIRAC’S THEORY OF THE ELECTRON

5.2.1. The Dirac Hamiltonian

So far, we have been using p2 /2m-type Hamiltonians, which are limited to describing
non-relativistic particles. In 1928, Paul Dirac formulated a Hamiltonian that can describe
electrons moving close to the speed of light, thus successfully combining quantum theory
with special relativity. Another triumph of Dirac’s theory is that it accurately predicts the
magnetic moment of the electron.
Dirac’s theory begins from the time-dependent Schrödinger wave equation,
iℏ ∂t ψ(r, t) = Ĥψ(r, t). (5.53)

Note that the left side has a first-order time derivative. On the right, the Hamiltonian Ĥ
contains spatial derivatives in the form of momentum operators. We know that time and
space derivatives of the wavefunction are related to energy and momentum, respectively:
iℏ ∂t ↔ E, −iℏ ∂j ↔ pj . (5.54)
We also know that the energy and momentum of a relativistic particle are related by
3
X
E 2 = m2 c4 + p2j c2 , (5.55)
j=1

where m is the rest mass and c is the speed of light. (Following the usual practice in relativity
theory, we use Roman indices j ∈ {1, 2, 3} for the spatial coordinates {x, y, z}.)
Since the left side of Eq. (5.53) has a first-order time derivative, we guess that the Hamil-
tonian on the right should involve first-order spatial derivatives. Let us take
3
X
2
Ĥ = α0 mc + αj p̂j c, (5.56)
j=1

where p̂j ≡ −iℏ∂j . The mc2 and c factors are placed for later convenience. To determine
the dimensionless “coefficients” α0 , α1 , α2 , and α3 , consider a wavefunction with definite
energy E and momentum p:
( 3
!
Ĥψ = Eψ X
⇒ α0 mc2 + αj pj c ψ = E ψ. (5.57)
p̂i ψ = pi ψ j=1

98
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

If ψ is a scalar, this would imply that

X
α0 mc2 + α j pj c = E
j

for certain scalar coefficients {α0 , . . . , α3 }, which does not match Eq. (5.55) at all!
But we can get things to work if ψ(r, t) is a multi-component wavefunction, rather than
a scalar wavefunction, and the α’s are matrices acting on those components via the matrix-
vector product operation. In that case,

3
X
2
Ĥ = α̂0 mc + α̂j p̂j c, where p̂j ≡ −iℏ ∂j , (5.58)
j=1

where the hats on {α̂0 , . . . , α̂3 } indicate that they are matrix-valued. Applying the Hamil-
tonian twice gives  2
X
2
α̂0 mc + α̂j pj c ψ = E 2 ψ. (5.59)
j
This can be satisfied if  X 2
α̂0 mc2 + α̂j pj c ˆ
= E 2 I, (5.60)
j

where Iˆ is the identity matrix. Expanding the square (and taking care of the fact that the
α̂µ matrices need not commute) yields
X X
α̂02 m2 c4 + (α̂0 α̂j + α̂j α̂0 ) mc3 pj + ˆ
α̂j α̂j ′ pj pj ′ c2 = E 2 I. (5.61)
j jj ′

This reduces to Eq. (5.55) if the α̂µ matrices satisfy

α̂µ2 = Iˆ for µ = 0, 1, 2, 3, and
(5.62)
α̂µ α̂ν + α̂ν α̂µ = 0 for µ ̸= ν.
(We use Greek symbols for indices ranging over the four spacetime coordinates {0, 1, 2, 3}.)
The above can be written more concisely using the anticommutator:
{α̂µ , α̂ν } = 2δµν , for µ, ν = 0, 1, 2, 3. (5.63)
Also, we need the α̂µ matrices to be Hermitian, so that Ĥ is Hermitian.
It turns out that the smallest possible Hermitian matrices that can satisfy Eq. (5.63) are
4 × 4 matrices. The choice of matrices (or “representation”) is not uniquely determined.
One particularly useful choice is called the Dirac representation:
Iˆ 0̂
   
0̂ σ̂1
α̂0 = , α̂1 =
0̂ −Iˆ σ̂1 0̂
    (5.64)
0̂ σ̂2 0̂ σ̂3
α̂2 = , α̂3 = ,
σ̂2 0̂ σ̂3 0̂
where {σ̂1 , σ̂2 , σ̂3 } denote the usual Pauli matrices. Since the α̂µ ’s are 4×4 matrices, the field
ψ(r), at each point r, is a four-component object, called a spinor (it’s not a vector since its
behaves differently under coordinate transformations; we won’t get into the details).

99
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

5.2.2. Eigenstates of the Dirac Hamiltonian

According to Eq. (5.55), the energy eigenvalues of the Dirac Hamiltonian are
s X
E = ± m2 c4 + p2j c2 . (5.65)
j

This is plotted below:

The energy spectrum forms two hyperbolas. The upper hyperbola matches the standard
dispersion relation for a massive relativistic particle. But what about the negative-energy
hyperbola? Nobody ordered that, and we might be tempted to simply ignore the negative-
energy states, focusing only on the positive-energy states. However, this becomes untenable
once we couple the electron to additional systems like the electromagnetic field: even if the
electron starts with E > 0, it can hop to the negative-energy hyperbola by shedding energy
into the external system. This seems like a glaring problem, but let us wait till Section 5.2.4
to discuss how to resolve it.
Notice also that for each p, the Dirac Hamiltonian reduces to a 4 × 4 matrix, so there
are four orthogonal eigenstates. Hence, both energy hyperbolas are two-fold degenerate.
The four degrees of freedom (two from the choice of positive/negative E, two from the
degeneracy) are related to the four spinor components in a somewhat subtle way. To help
understand this, let us adopt the Dirac representation (5.64), and divide the spinor as follows:
 
ψA (r, t)
ψ(r, t) = , (5.66)
ψB (r, t)

where ψA and ψB have two components each. Using the Dirac Hamiltonian (5.58), we derive
1 X
ψA = σ̂j pj ψB , (5.67)
E − mc2 j
1 X
ψB = σ̂j pj ψA . (5.68)
E + mc2 j

For |p| → 0, E approaches either mc2 or −mc2 . On the upper hyperbola (E ∼ mc2 ), the
vanishing of the denominator in Eq. (5.67) implies that ψA dominates the wavefunction,
with ψB ∼ 0. On the lower hyperbola (E ∼ −mc2 ), ψB dominates, with ψA ∼ 0. In the

100
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

next section, we will develop this further and show that the two-fold degree of freedom of
each sub-spinor, ψA and ψB , corresponds to the electron’s spin.
However, this association only holds in the non-relativistic limit! In the relativistic regime,
|p| ∼ |E| ≫ mc2 , Eqs. (5.67) and (5.68) imply that |ψA | ∼ |ψB | for both positive-energy
and negative-energy eigenstates. (There does exist a way to make the upper/lower spinor
components correspond rigorously to positive/negative E, but this requires adopting a more
complicated matrix representation [4].)

5.2.3. Dirac electron in an electromagnetic field

Let us next look at how Dirac’s electron interacts with an electromagnetic field. To
introduce the electromagnetic field into the theory, we follow the same procedure as in the
non-relativistic theory (Section 5.1.1): add −eΦ(r, t) as a scalar potential function, and add
the vector potential via the substitution
p̂ → p̂ + eA(r̂, t). (5.69)
Applying this recipe to the Dirac Hamiltonian (5.58) yields
( )
X h i
iℏ ∂t ψ = α̂0 mc2 − eΦ(r, t) + α̂j − iℏ ∂j + eAj (r, t) c ψ(r, t). (5.70)
j

You can check that this has the same gauge symmetry properties as the non-relativistic
theory discussed in Section 5.1.3.
We now take the Dirac representation (5.64). Moreover, following Eq. (5.66), let ψA
and ψB denote the the Dirac wavefunction’s upper and lower components (each having two
components). Then Eq. (5.70) reduces to
X
iℏ ∂t ψA = + mc2 − eΦ ψA +



σ̂j − iℏ∂j + eAj c ψB (5.71)
j
X
2



iℏ ∂t ψB = − mc − eΦ ψB + σ̂j − iℏ∂j + eAj c ψA , (5.72)
j

In the non-relativistic limit, we can seek solutions of the form

  2 
mc
ψA/B (r, t) = ΨA/B (r, t) exp −i t . (5.73)
ℏ
The phase factor on the right side comes from the rest energy mc2 . This is assumed to
be by far the largest energy scale in the problem, corresponding to a fast oscillation. (By
using mc2 rather than −mc2 , we are focusing on the positive-energy states.) The functions
ΨA and ΨB are “slowly-varying envelopes”, which have t-dependences much slower than the
phase factor. When p = 0 and Φ = A = 0, they reduce to constants.
Plugging this ansatz into Eqs. (5.71)–(5.72) gives
X

iℏ ∂t ΨA = −eΦ ΨA + σ̂j − iℏ∂j + eAj c ΨB (5.74)
j
X
iℏ ∂t + 2mc2 + eΦ ΨB =



σ̂j − iℏ∂j + eAj c ΨA . (5.75)
j

101
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

On the left side of Eq. (5.75), the 2mc2 term dominates over the other two, so
1 X

ΨB ≈ σ̂j − iℏ∂j + eAj ΨA . (5.76)
2mc j

Plugging this into Eq. (5.74) yields

( )
1 X


iℏ ∂t ΨA = −eΦ + σ̂j σ̂k − iℏ∂j + eAj − iℏ∂k + eAk ΨA . (5.77)
2m jk

Using the identity σ̂j σ̂k = δjk Iˆ + i i εijk σi :

P
(
1 2
iℏ ∂t ΨA = − eΦ + − iℏ∇ + eA
2m
) (5.78)
i X


+ εijk σ̂i − iℏ∂j + eAj − iℏ∂k + eAk ΨA .
2m ijk

Look carefully at the last term in the curly brackets. Expanding the square yields
i X  
εijk σ̂i − ∂j ∂k − iℏe∂j Ak − iℏe Ak ∂j + Aj ∂k + e2 Aj Ak .
 
2m ijk

Due to the antisymmetry of εijk , all terms inside the parentheses that are symmetric under
j and k cancel out when summed over. The only survivor is the second term, which gives
ℏe X ℏe
εijk σ̂i ∂j Ak = σ̂ · B(r, t), (5.79)
2m ijk 2m

where B = ∇ × A is the magnetic field. Hence,

   
1 2 ℏe
iℏ ∂t ΨA = −eΦ + − iℏ∇ + eA − − σ̂ · B ΨA . (5.80)
2m 2m
On the right side, we see the Hamiltonian for a non-relativistic electron in an electromagnetic
field, Eq. (5.25), but with an additional term of the form −µ̂ · B̂. This new term matches
the potential energy of a magnetic dipole of moment µ in a magnetic field B. Therefore,
the theory predicts that the electron has a magnetic dipole moment of
ℏe
|µ| = . (5.81)
2m
Remarkably, this matches the experimentally-observed magnetic dipole moment to about
one part in 103 . The residual mismatch between Eq. (5.81) and the actual magnetic dipole
moment of the electron is understood to arise from quantum fluctuations of the electronic
and electromagnetic quantum fields. Using the full theory of quantum electrodynamics, that
“anomalous magnetic moment” can also be calculated and matches experiment to around
one part in 109 , making it one of the most precise theoretical predictions in physics [2]!
It is noteworthy that we did not set out to include spin in the theory, yet it arose,
seemingly unavoidably, as a by-product of formulating a relativistic theory of the electron.
This is a manifestation of the principle that relativistic quantum theory is more constrained
than non-relativistic quantum theory [1]. To satisfy the relativistic symmetries, spin cannot
be an optional part of the theory—it must be built-in at a fundamental level.

102
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

5.2.4. Positrons and the Dirac Field

As noted in Section 5.2.2, the Dirac Hamiltonian has both positive-energy and negative-
energy eigenstates, and the presence of the negative-energy states threatens to destabilize
the positive-energy states. To get around this, Dirac suggested that the “vacuum” might
actually be a multi-particle state, called the Dirac sea, in which all negative-energy states
are occupied. Since electrons are fermions, the Pauli exclusion principle forbids decay into
the negative-energy states, thus stabilizing the positive-energy solutions.
At first blush, the idea seems ridiculous; how can the vacuum contain an infinite number
of particles? However, we shall see that the idea becomes more plausible if Dirac’s single-
particle Hamiltonian is reinterpreted using quantum field theory. After all, the Dirac sea
idea is an inherently multi-particle concept, and we know from Chapter 4 that quantum
field theory is the natural framework for working with such ideas.
For the single-particle Dirac Hamiltonian, let |k, +, σ⟩ denote a positive-energy state
of momentum ℏk, where σ = 1, 2 indexes the spin degree of freedom (i.e., the two-fold
degeneracy of each hyperbola). On the other hand, let |k, −, σ⟩ denote a negative-energy
state of momentum −ℏk (not ℏk; we will justify this convention later). Hence,

Ĥ|k, ±, σ⟩ = ±Ek |k, ±, σ⟩, (5.82)

where Ek = E−k > 0 comes from the relativistic energy-mass-momentum relation (5.55).
Following the second quantization procedure from Chapter 4, we introduce a fermionic
Fock space H−F , and a set of creation/annihilation operators. We will use different labels
for the positive- and negative-energy states:

b̂†kσ and b̂kσ create/annihilate |k, +, σ⟩

dˆ†kσ and dˆkσ create/annihilate |k, −, σ⟩.

These obey the fermionic anticommutation relations

{b̂kσ , b̂†k′ σ′ } = δ 3 (k − k′ ) δσσ′ , {dˆkσ , dˆ†k′ σ′ } = δ 3 (k − k′ ) δσσ′

(5.83)
{b̂kσ , b̂k′ σ′ } = {b̂kσ , dˆk′ σ′ } = {dˆkσ , dˆk′ σ′ } = 0, etc.

We can then define the multi-particle Hamiltonian

Z X  
3 † ˆ† ˆ
Ĥ = d k Ek b̂kσ b̂kσ − dkσ dkσ , (5.84)
σ

where the negative sign inside the integrand accounts for the negative energy of the d-type
electrons. We also have the vacuum state |∅⟩, which satisfies

b̂kσ |∅⟩ = dˆkσ |∅⟩ = 0. (5.85)

Next, let us define

ĉkσ = dˆ†kσ
(5.86)
ĉ† = dˆkσ .
kσ

103
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

Using these, we can rewrite the anticommutation relations (5.83) as

{b̂kσ , b̂†k′ σ′ } = δ 3 (k − k′ ) δσσ′ , {ĉkσ , ĉ†k′ σ′ } = δ 3 (k − k′ ) δσσ′

(5.87)
{b̂kσ , b̂k′ σ′ } = {b̂kσ , ĉk′ σ′ } = {ĉkσ , ĉk′ σ′ } = 0, etc.

This implies that we can treat ĉ†kσ and ĉkσ as creation and annihilation operators in their
own right, taking over from dˆkσ and dˆ†kσ . The particle created by ĉ†kσ is called a positron.
By definition, the presence of a positron is equivalent to the absence of a negative-energy
electron; hence, the positron has positive energy. Indeed, by converting the d’s to c’s in
Eq. (5.84), we can rewrite the Hamiltonian as
Z X  
Ĥ = d3 k Ek b̂†kσ b̂kσ − ĉkσ ĉ†kσ (5.88)
σ
Z X  
= 3
dk Ek b̂†kσ b̂kσ + ĉ†kσ ĉkσ + constant. (5.89)
σ

(The constant is infinite, but this can be dealt with by standard regularization tricks.) Now
Ĥ describes two species of fermions—b-type electrons, which we simply call “electrons”, and
positrons—both carrying positive energy. We can also deduce, via similar reasoning, that
the positron created by ĉ†kσ has momentum ℏk (not −ℏk, due to how we defined |k, −, σ⟩),
and a positive electric charge +e.
The state |∅⟩, defined in Eq. (5.85), no longer serves as the vacuum for this reinterpreted
Hamiltonian. Its vacuum state |∅′ ⟩ should satisfy

b̂kσ |∅′ ⟩ = ĉkσ |∅′ ⟩ = 0. (5.90)

Such a state can be constructed from |∅⟩ (or vice versa):

Y Y †
|∅′ ⟩ = ĉkσ |∅⟩ = dˆkσ |∅⟩. (5.91)
kσ kσ

Thus, from the viewpoint of the b, d operators, |∅′ ⟩ is the “Dirac sea” state, formed by
occupying all the negative-energy states. From the viewpoint of the b, c operators, it is
simply a state of no particles.
The next step is to interpret the multi-particle system as a quantum field. This must
be done with some care, due to the peculiar relationship between the electron and positron
operators. When formulating bosonic field theory in Chapter 4, we defined a field operator
Z
ψ̂(r) = d3 k φk (r) âk , (5.92)

where âk is a bosonic annihilation operator, and φk (r) denotes the corresponding single-
particle wavefunction. In our case, the single-particle wavefunctions have the form
eik·r n
⟨r, n|k, +, σ⟩ = u (5.93)
(2π)3/2 kσ
e−ik·r n
⟨r, n|k, −, σ⟩ = v (5.94)
(2π)3/2 −k,σ

104
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

where n = 1, . . . , 4 indexes the spinor components. In Eq. (5.94), the sign of the exponent
corresponds to momentum −ℏk, matching our earlier definition. For each k, ukσ and vk,σ
are normalized eigenvectors of a single 4 × 4 Hamiltonian matrix, so they are orthogonal,
X X X
(unkσ )∗ ukσ′ = δσσ′ , n ∗
(vkσ ) vkσ′ = δσσ′ , (unkσ )∗ vkσ′ = 0, (5.95)
n n n

and satisfy the resolution of the identity,

Xh ′
∗
i
n′ ∗
unkσ unkσ n


+ vkσ vkσ = δnn′ . (5.96)
σ

By analogy with Eq. (5.92), we can use Eqs. (5.93)–(5.94) to define the field operator

X  eik·r e−ik·r n ˆ
Z 
3 n
ψ̂n (r) = dk u b̂kσ + v dkσ . (5.97)
σ
(2π)3/2 kσ (2π)3/2 −k,σ

Like the single-particle wavefunctions, the field operator at each point r has four components
(i.e., it consists of four distinct operators), indexed by n. Next, in the second term of the
integrand, let us replace the d operators with c operators, and change variables k → −k:

d3 k X ik·r  n
Z 
n †
ψ̂n (r) = e ukσ b̂kσ + vk,σ ĉ−k,σ . (5.98)
(2π)3/2 σ

Using the electron and positron anticommutation relations (5.87), and the spinor property
(5.96), we can now prove that
n o
ψ̂n (r) , ψ̂n† ′ (r′ )
= δnn′ δ 3 (r − r′ ), (5.99)
n o n o
ψ̂n (r) , ψ̂n′ (r′ ) = ψ̂n† (r) , ψ̂n† ′ (r′ ) = 0. (5.100)

These imply that ψ̂n (r) and ψ̂n† (r) act as creation/annihilation operators at a local point
r. However, neither field operator acts on electrons or positrons alone: ψ̂n (r) annihilates
an electron or creates a positron [Eq. (5.98)], and conversely ψ̂n† (r) creates an electron or
annihilates a positron.
The fermionic quantum field theory can thus be interpreted in two ways: (i) in terms of
electrons and positrons carrying positive energies, with |∅′ ⟩ referring to a vacuum state of
no particles, or (ii) in terms of electrons that can carry positive or negative energies, with
|∅′ ⟩ referring to a Dirac sea in which all negative-energy states are filled. The former picture
is arguably theoretically “cleaner”, whereas the latter picture has the advantage of mapping
directly to Dirac’s original single-particle theory.
There are many more details about the Dirac field theory that we will not discuss here,
including the expression of the Hamiltonian in terms of the field operators, and the issue of
how the particles transform under Lorentz boosts and other changes in coordinate system.
For such matters, the reader is referred to Ref. [1].

105
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

5.3. QUANTIZING THE ELECTROMAGNETIC FIELD

Up to now, we have treated the electromagnetic field as a classical entity not subject
to quantum mechanics. At a fundamental level, however, the electromagnetic field must
behave as a quantum field. We have already, in Chapter 4, discussed the quantization of a
simple scalar boson field: the classical field is decomposed into normal modes, each mode is
assigned a set of quantum mechanical creation and annihilation operators, and these are used
to construct quantum observables (particularly the Hamiltonian and field variables). We can
use the same approach, with only minor adjustments, to quantize the electromagnetic field.
First, consider a “source-free” electromagnetic field—i.e., with no electric charges and
currents. Without sources, Maxwell’s equations (in SI units, and in a vacuum) reduce to:
∇·E=0 (5.101)
∇·B=0 (5.102)
∂B
∇×E=− (5.103)
∂t
1 ∂E
∇×B= 2 . (5.104)
c ∂t
Once again, we introduce the scalar potential Φ and vector potential A:
∂A
E = −∇Φ − (5.105)
∂t
B = ∇ × A. (5.106)
With these relations, Eqs. (5.102) and (5.103) are satisfied automatically via vector identities.
The two remaining equations, (5.101) and (5.104), become:
∂
∇2 Φ = − ∇ · A (5.107)
∂t
2
   
2 1 ∂ 1 ∂
∇ − 2 2 A=∇ 2 Φ+∇·A . (5.108)
c ∂t c ∂t
In the next step, we choose a convenient gauge called the Coulomb gauge:
Φ = 0, ∇ · A = 0. (5.109)
(To see that we can always make such a gauge choice, suppose we start out with a scalar
potential Φ0 and
R t vector potential A0 not satisfying (5.109). Perform a gauge transformation
with Λ(r, t) = dt Φ0 (r, t′ ). The new potentials satisfy
′

Φ = Φ0 − Λ̇ = Φ0 − Φ0 = 0 (5.110)
Z t
∇ · A = ∇ · A0 + ∇2 Λ = ∇ · A0 + dt′ ∇2 Φ0 (r, t′ ). (5.111)

Plugging Eq. (5.107) into Eq. (5.111), we find that ∇ · A = 0.)

In the Coulomb gauge, Eq. (5.107) is automatically satisfied. The sole remaining equation,
(5.108), simplifies to
1 ∂2
 
2
∇ − 2 2 A = 0. (5.112)
c ∂t

106
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

This has plane-wave solutions of the form

 
A(r, t) = A ei(k·r−ωt) + c.c. e, (5.113)

where A is a complex number (the mode amplitude) that specifies the magnitude and
phase of the plane wave, e is a real unit vector (the polarization vector) that specifies
which direction the vector potential points along, and “c.c.” denotes the complex conjugate
of the first term. Referring to Eq. (5.112), the angular frequency ω must satisfy

ω = c|k|. (5.114)

Moreover, since ∇ · A = 0, it must be the case that

k · e = 0. (5.115)

In other words, the polarization vector is perpendicular to the propagation direction. For
any given k, we can choose (arbitrarily) two orthogonal polarization vectors.
Now suppose we put the electromagnetic field in a box of volume V = L3 , with periodic
boundary conditions (we will take L → ∞ at the end). The k vectors form a discrete set:

2πnj
kj = , nj ∈ Z, for j = 1, 2, 3. (5.116)
L
Then the vector potential field can be decomposed as a superposition of plane waves,
X 
A(r, t) = Akλ ei(k·r−ωk t) + c.c. ekλ , where ωk = c|k|. (5.117)
kλ

Here, λ is a two-fold polarization degree of freedom indexing the two possible orthogonal
polarization vectors for each k. (We won’t need to specify how exactly these polarization
vectors are defined, so long as the definition is used consistently.)
To convert the classical field theory into a quantum field theory, for each (k, λ) we define
an independent set of creation and annihilation operators:

âkλ , â†k′ λ′ = δkk′ δλλ′ , âkλ , âk′ λ′ = â†kλ , â†k′ λ′ = 0.

     
(5.118)

Then the Hamiltonian for the electromagnetic field is

X
Ĥ = ℏωk â†kλ âkλ , where ωk = c|k|. (5.119)
kλ

The vector potential is now promoted into a Hermitian operator in the Heisenberg picture:
X  
Â(r, t) = Ckλ âkλ ei(k·r−ωk t) + h.c. ekλ . (5.120)
kλ

Here, Ckλ is a constant to be determined, and “h.c.” denotes the Hermitian conjugate.
The creation and annihilation operators in this equation are Schrödinger picture (t = 0)
operators. The particles they create/annihilate are photons—elementary particles of light.

107
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

To find Ckλ , we compare the quantum and classical energies. Suppose the electromagnetic
field is in a coherent state |α⟩ such that for any k and λ,

âkλ |α⟩ = αkλ |α⟩ (5.121)

for some αkλ ∈ C. From this and Eq. (5.120), we identify the corresponding classical field
X 
A(r, t) = Akλ ei(k·r−ωk t) + c.c. ekλ , where Ckλ αkλ = Akλ . (5.122)
kλ

For each k and λ, Eqs. (5.105)–(5.106) give the electric and magnetic fields
 
Ekλ = iωk Akλ ei(k·r−ωk t) + c.c. ekλ (5.123)
 
i(k·r−ωk t)
Bkλ = iAkλ e + c.c. k × ekλ . (5.124)

In the classical theory of electromagnetism, Poynting’s theorem tells us that the total energy
carried by a classical plane electromagnetic wave is
Z
3 ϵ0
 
2 2 2
E= dr Ekλ + c Bkλ
V 2 (5.125)
2 2
= 2 ϵ0 ωk |Akλ | V.

Here, V is the volume of the enclosing box, and we have used the fact that terms like e2ik·r
vanish when integrated over r. Hence, we make the correspondence
r
2 2 2 ℏ
2 ϵ0 ωk |Ckλ αkλ | V = ℏωk |αkλ | ⇒ Ckλ = . (5.126)
2ϵ0 ωk V

We thus arrive at the result

X
Ĥ = ℏωk â†kλ âkλ
kλ
r  
X ℏ (5.127)
Â(r, t) = âkλ ei(k·r−ωk t) + h.c. ekλ
kλ
2ϵ0 ωk V
ωk = c|k|, âkλ , â†k′ λ′ = δkk′ δλλ′ , âkλ , âk′ λ′ = 0.
   

To describe infinite free space rather than a finite-volume box, we take the L → ∞ limit
and re-normalize the creation and annihilation operators by the replacement
r
(2π)3
âkλ → âkλ . (5.128)
V
Then the sums over k become integrals over the infinite three-dimensional space:

108
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

Z X
Ĥ = d3 k ℏωk â†kλ âkλ
λ
Z r
(5.129)
X ℏ  
3 i(k·r−ωk t)
Â(r, t) = d k âkλ e + h.c. ekλ
λ
16π 3 ϵ0 ωk
ωk = c|k|, âkλ , â†k′ λ′ = δ 3 (k − k′ )δλλ′ , âkλ , âk′ λ′ = 0.
   

5.4. THE ELECTRON-PHOTON INTERACTION

Having derived fully quantum mechanical theories for both the electron and the electro-
magnetic field, we are now able to describe a range of phenomena involving the absorption
and/or emission of light by matter. As an example, we will model how an excited atom
decays by emitting a photon.
Consider a single electron orbiting an atomic nucleus. It is described by a single-particle
Hilbert space He , while the electromagnetic field has Hilbert space HEM . The combined
system has Hilbert space He ⊗ HEM , and is described by a Hamiltonian of the form

H = He + HEM + Hint . (5.130)

Here, He is the electron’s Hamiltonian in the absence of the electromagnetic field, HEM
is the Hamiltonian for the source-free electromagnetic field, and Hint is an “interaction
Hamiltonian” describing how the electron and electromagnetic field affect each other.
We assume Ĥe has a ground state of energy E0 and an excited state of energy E1 > E0 :

Ĥe |1⟩ = E1 |1⟩

(5.131)
Ĥe |0⟩ = E0 |0⟩.

We initialize the electron in the excited state |1⟩, and the electromagnetic field in the vacuum
state |∅⟩. Thus, the initial state of the combined system is

|ψi ⟩ = |1⟩ ⊗ |∅⟩. (5.132)

If Ĥint = 0, then |ψi ⟩ is an energy eigenstate, so the electron would remain in the excited
state for all time. But if Ĥint is a non-vanishing operator, the electron can decay to |0⟩ by
emitting a photon. The post-decay state of the combined system has the form
 
|ψkλ ⟩ = |0⟩ ⊗ â†kλ |∅⟩ . (5.133)

There are many such final states, of different photon wavevector k and polarization λ.
The decay rate κ can be estimated using Fermi’s Golden Rule (see Chapter 2):
2π
κ= |W|2 D, (5.134)
ℏ

109
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

where

W = (2π)3/2 ⟨ψkλ |Ĥint |ψi ⟩ (5.135)

d3 k
Z
D=2 δ(ℏω − ℏc|k|). (5.136)
(2π)3

Here, we have set the spatial dimension to d = 3, (· · · ) denotes the average over final states,
and D is the density of photon states, with the prefactor of 2 accounting for the polarization
degree of freedom. Eqs. (5.135)–(5.136) are to be evaluated at the resonance energy:

ℏω = ℏc|k| = E1 − E0 . (5.137)

To calculate the transition amplitude (5.135), we need the interaction Hamiltonian Ĥint .
Let us adopt the Coulomb gauge (see Section 5.3), so that the scalar potential vanishes, and
the vector potential acts on the electron via the substitution (see Section 5.1.1):

p̂ → p̂ + eA(r̂, t). (5.138)

Previously, we have treated the A field in Eq. (5.138) as a classical entity; although the
electron’s position operator r̂ slots into it, the field itself lacks quantum dynamics. Now, we
make this a quantum field by replacing A with the operator derived in Section 5.3:
Z r  
3
X ℏ ik·r̂
Â(r̂) = d k âkλ e + h.c. ekλ . (5.139)
λ
16π 3 ϵ0 ωk

Here, we are using the expression for infinite space and the Schrodinger picture, equivalent
to setting t = 0 in Eq. (5.129). Note that the position in Eq. (5.139) is not a vector, but the
electron’s position operator; hence, Â(r̂, t) acts nontrivially on both He and HEM .
For a non-relativistic electron, the Hamiltonian (5.25) yields the interaction Hamiltonian
e  
Hint = p̂ · Â + Â · p̂ . (5.140)
2m
Combining this with Eq. (5.139), we can calculate
Z r
e X ℏ ′
⟨ψkλ |Ĥint |ψi ⟩ = dk3 ′
⟨0|{p̂j , e−ik ·r̂ }|1⟩ ejk′ λ′ ⟨∅|âkλ â†k′ λ′ |∅⟩
2m jλ′
16π 3 ϵ0 ωk′
r
e ℏ X
= ⟨0|{p̂j , e−ik·r̂ }|1⟩ ejkλ (5.141)
2m 16π 3 ϵ0 ωk j

We can make two further simplifications. Firstly, note that for atomic transitions in the
visible light regime, the optical wavelength is around 10−6 m (i.e., |k| ∼ 106 m− 1), whereas
the size of a typical atomic orbital is around 10−9 m. Hence, we can take exp(−ik · r̂) ≈ 1.
Secondly, note that if the electron’s Hamiltonian has the generic form

Ĥe = |p̂|2 /2m + V (r), (5.142)

110
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

then [Ĥe , r̂] = −iℏp/m, and hence

im(E1 − E0 )d
⟨0|p̂j |1⟩ = − , d = ⟨0|r|1⟩. (5.143)
ℏ
We call d, which is easily calculated from the orbital wavefunctions, the transition dipole
moment. Putting these simplifications into Eq. (5.141) yields
r
ℏωk
⟨ψkλ |Ĥint |ψi ⟩ ≈ −ie d · ekλ . (5.144)
16π 3 ϵ0
According to Fermi’s Golden Rule, Eq. (5.134), we need to take the absolute square and
average over photon states. It can be shown (via a standard angular integration) that
|d|2
|d · ekλ |2 = . (5.145)
3
Hence,

|W|2 = (2π)3 |⟨ψkλ |Ĥint |ψi ⟩|2 (5.146)

e2 ℏω
= |d|2 , (5.147)
6ϵ0
where ω = (E1 − E0 )/ℏ is the resonance frequency.
Finally, we calculate the density of states, Eq. (5.136), which results in (see Exercise 3):
ω2
D(ℏω) = . (5.148)
π 2 ℏc3
Plugging Eq. (5.147) and (5.148) into Eq. (5.134), we obtain
e2 ω 3 |d|2
κ= (5.149)
3πϵ0 ℏc3
We can make this look nicer by defining the dimensionless fine-structure constant
e2 1
α= ≈ . (5.150)
4πϵ0 ℏc 137
Hence,

4 αω 3 |d|2
κ= . (5.151)
3 c2

This rate of spontaneous emission is called the Einstein A coefficient. It can be observed
directly in experiments, by measuring the line-widths of atomic spectral lines. In the figure
below, we present a comparison of the theoretical result, Eq. (5.151), to the experimentally-
observed decay rates of the simplest excited states of hydrogen, lithium, and sodium. In
theoretical result, we simply take the transition dipole moment to be |d| ≈ 10−10 m (i.e.,
the length scale of an atomic orbital), rather than doing a more precise calculation using
the actual orbital wavefunctions. Despite this rather crude approximation, it can be seen
that the decay rates produced by Fermi’s Golden Rule are within striking distance of the
experimental values.

111
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

FIG. 1. Spontaneous emission rates (Einstein A coefficients) for the 2p → 1s transition in hydrogen,
the 2p → 2s transition in lithium, and the 3p → 3s transition in sodium. Data points extracted
from the NIST Atomic Spectra Database (https://www.nist.gov/pml/atomic-spectra-database).
The dashed curve shows the decay rate based on Fermi’s Golden Rule, with |d| ≈ 10−10 m.

EXERCISES

1. In Section 5.3, we derived the vector potential operator, in an infinite volume, to be

Z r  
3
X ℏ i(k·r−ωk t)
Â(r, t) = d k âkλ e + h.c. ekλ . (5.152)
λ
16π 3 ϵ0 ωk

Since [âkλ , â†k′ λ′ ] = δ 3 (k − k′ )δλλ′ , the creation and annihilation operators each have
units of [x3/2 ]. Prove that  has the same units as the classical vector potential.
2. Repeat the spontaneous decay rate calculation from Section 5.4 using the finite-
volume versions of the creation/annihilation operators and the vector potential op-
erator (5.139). Show that it yields the same result (5.150).
3. The density of photon states at energy E is defined as
d3 k
Z
D(E) = 2 δ(E − Ek ), (5.153)
(2π)3
where Ek = ℏc|k|. Note the factor of 2 accounting for the polarizations. Prove that
E2
D(E) = . (5.154)
π 2 ℏ3 c3

FURTHER READING

[1] F. J. Dyson, 1951 Lectures on Advanced Quantum Mechanics Second Edition, arxiv:quant-
ph/0608140. [link]

112
Y. D. Chong Ch. 5: Quantum Electrodynamics | Graduate Quantum Mechanics

[2] A. Zee, Quantum Field Theory in a Nutshell (Princeton University Press, 2010).
[3] P. A. M. Dirac, Quantized singularities in the electromagnetic field, Proceedings of the
Royal Society (London) 133, 60 (1931).
[4] L. L. Foldy and S. A. Wouthuysen, On the Dirac Theory of Spin 1/2 Particles and Its
Non-Relativistic Limit, Physical Review 78, 29 (1950). [link]

113