Electromagnetic Field Quantization & Applications to the Casimir Effect

Prashanth S. Venkataram
MIT Department of Physics: 8.06
(Dated: May 6, 2013)
The quantization of the electromagnetic field is demonstrated through a two-pronged approach. The
classical harmonic oscillator is discussed in depth, and then its quantization is further discussed.
Additionally, the classical electromagnetic field Hamiltonian is derived and analogized to the classical
harmonic oscillator Hamiltonian at several points. Quantization of the electromagnetic field and its
Hamiltonian are then performed, leading to an application of the free electromagnetic field energy
spectrum to the analysis of the Casimir effect. This too is performed through two routes, which are
by analyzing the energy gradient and by analyzing the radiation pressure of the electromagnetic field
subject to certain boundary conditions; these are qualitatively different but quantitatively converge
to give the same desired result.

I. INTRODUCTION II. SIMPLE HARMONIC OSCILLATOR

A. Classical

The equation of motion for a simple harmonic oscilla-

A cornerstone of quantum mechanics is the noncommu- tor of mass m and frequency arises from the Hamilto-
tation between canonically conjugate observables, lead- nian
ing to an uncertainty relation for measurements of those
observables in a quantum state. This remains true in p2 m 2 x2
H= + (1)
both nonrelativistic quantum mechanics and quantum 2m 2
electrodynamics. In the former, the famous Heisenberg where x and p are the canonically conjugate observable
uncertainty principle states that the position and mo- position and momentum. The state of the system at a
mentum of a state cannot be determined simultaneously. time t is fully determined by the phase space coordinates
In the latter, the corresponding components of the elec- (x(t), p(t)), and time evolution is given by the Hamilton
tric and magnetic fields of a given mode cannot be deter- equations
mined simultaneously. This leads to the Casimir effect,
which is a purely quantum interaction between electri- H
p = = m 2 x (2)
cally charge-free and current-free bodies through ground- x
state fluctuations in the electromagnetic field. H p
x = + = (3)
p m
which can be collapsed into the familiar form of the sim-
Almost all of the key results of quantizing the electro- ple harmonic oscillator equation of motion
magnetic field and of analyzing the Casimir effect can be
derived by treating the electromagnetic field much like a x = 2 x. (4)
harmonic oscillator. In fact, the revered quantum field An ansatz for this comes in the form (t) = (0)eit .
theory professor Sidney Coleman once said, The career However, this is a complex quantity, making it unobserv-
of a young theoretical physicist consists of treating the able; its real and imaginary parts, however, are individ-
harmonic oscillator in ever-increasing levels of abstrac- ually observable, and can be taken respectively as x and
tion. With that in mind, the structure of this paper p. If is taken as dimensionless, then the expressions of
is as follows. First, the classical harmonic oscillator is x and p become
discussed in a way that makes apparent the alternative
physical meanings of the quantum creation and annihi-
r
~
lation operators in the oscillator; immediately following x(t) = (? (t) + (t)) (5)
2m
this is a discussion of the quantum harmonic oscillator r
and several effects that are purely quantum mechanical. ~m ?
p(t) = i ( (t) (t)) (6)
Second, the classical electromagnetic field is rewritten to 2
resemble the harmonic oscillator Hamiltonian. Third, the where ~ is for now just a constant term with dimensions
quantization of the electromagnetic field and its Hamil- of action. This can be rewritten to make the meanings
tonian along with the implications thereof are discussed. of and ? more clear:
Finally, the quantum effects arising from the electro-

 
magnetic field quantization are applied in analyzing the 1 i
(t) = mx(t) + p(t) (7)
Casimir effect. 2~ m
Electromagnetic Field Quantization & Applications to the Casimir Effect 2

with ? just being the complex conjugate of . The state

of the system in this harmonic oscillator is determined by
a point in phase space (x(t), p(t)), but every ordered pair
of numbers can be written as a complex number, and
does just that. Hence, is a compact representation
of the phase space state of the system (x, p), and ?
exists so that x and p can be formed as independent real
observable quantities. Additionally, the Hamiltonian can
be rewritten as
~
H= (? (t)(t) + (t)? (t)) . (8)
2

FIG. 1: Harmonic oscillator spectrum and eigenfunctions [5]

B. Quantum

In contrast to classical mechanics, where states are III. CLASSICAL ELECTROMAGNETIC FIELD
given by definite positions in phase space (x, p) and where
measurement of an observable Y = Y (x, p) is certain, in In Gaussian CGS units, the Hamiltonian density of the
quantum mechanics, the state of a system is given by a electromagnetic field is given by [2]
vector |i in a Hilbert space. Operators can transform
such vectors to other vectors within the space. The ob- 1
E2 + B2


H= (10)
servable quantities x and p are now Hermitian operators 8
in that space with commutator [x, p] = i~. x and p have and the overall field Hamiltonian is the Hamiltonian den-
their own sets of eigenstates with the eigenvalues repre- sity integrated over space:
senting the position or momentum of that eigenstate, and Z
that last commutator forbids simultaneous diagonaliza-
H = H d3 r. (11)
tion of the two operators; this means that position and
momentum can no longer be simultaneously described in
a deterministic manner, but instead there exist funda- Meanwhile, the Poynting vector
mental uncertainties x and p satisfying xp ~2 . c
Now the meaning of ~ as the scale of the fundamental S= EB (12)
4
uncertainty of states in phase space (x, p) is clear.
and ? are not real observables, so their quantized is related to the momentum density by a factor of c2 , so
operators a and a are not Hermitian. These operators the field carries total momentum given by
do not commute, either: the noncommutation of x and Z
p turns into the commutation relation [a, a ] = 1, so c p = S d3 r
2
(13)
the ordering of operators is significant and is given as
aa = a a + 1. Furthermore, this means that the clas- which it can impart to other objects upon scattering.
sical Hamiltonian given in (8) becomes quantized as the The Hamiltonian and momentum play large roles in de-
familiar expression scribing the electromagnetic field and the Casimir effect.
  The time evolution of the electromagnetic field is given
~

1
H= a a + aa = ~ a a + . (9) by the Maxwell equations. Specifically, the time evolu-
2 2 tion of the magnetic field is given by the Faraday law

There are a few key things to note here. x, p, and H all 1 B

E= , (14)
do not commute. States of definite energy are eigenstates c t
|ni only of H (or, alternatively, eigenstates of a a), where the time evolution of the electric field in the absence of
En = ~ (n + 12 ) for n {Z : n 0}. This means the current sources is given by the Ampere law
lowest-energy state is the ground state |0i, yet its energy
E0 = ~ 2 > 0, so it is not possible to have an energy 1 E
exactly at the bottom of the harmonic oscillator well, B=+ , (15)
c t
consistent with the mutual noncommutation of x, p, and
H. This also means that while the expectation values and the spatial configuration of the electromagnetic field
hxi = 0 and hpi = 0 in |0i, the fluctuations hx2 i > 0 is restricted by the Gauss laws in the absence of charge
and hp2 i > 0 even in |0i, so the uncertainties x > 0 sources
and p > 0, which are consistent with the uncertainty
E=0 (16)
principle. This too comes from the noncommutation of
x and p (or alternatively of a and a ). B = 0. (17)
Electromagnetic Field Quantization & Applications to the Casimir Effect 3

The laws given in (14) and (15) can be combined to yield analogies only hold for a given mode k. Furthermore,
the wave equation k is a compact representation of the electromagnetic
field (E, B) analogous to the harmonic oscillator repre-
1 2E sentation (x, p), and the former can indeed be measured
2 E = (18)
c2 t2 deterministically in classical electrodynamics.
Returning now to the fields in real space given by (20),
in a vacuum, where the definition k = kek and the dis-
the combination of the spatial functions E0 satisfying
persion relation (k) = ck will be used hereafter.
(19) and the overall field given in (20) satisfying (16)
For a single mode k, solutions of the wave equation can
yield the simplifications [4]
be separated into temporal and spatial functions k (t)
and E0 (r), respectively; general expressions for the elec- Z Z
2 3 2
tric and magnetic fields are superpositions of the mode ( E0 ) d r = k E20 (r) d3 r (26)
functions over all modes, but it is instructive to first con- Z Z
sider the electric and magnetic fields for a single mode. ( E?0 )2 d3 r = k 2 E?2 3
0 (r) d r (27)
Plugging the single mode temporal and spatial functions Z Z
back into (18) requires that k (t) = k (0)eickt as in the | E0 |2 d3 r = k 2 |E0 (r)|2 d3 r. (28)
harmonic oscillator, while E0 must satisfy the Helmholtz
equation
Furthermore, the spatial function E0 can be chosen to
2 2 have the normalization [4]
E0 + k E0 = 0. (19)
Z
These functions may be complex in general, so by con- |E0 (r)|2 d3 r = 2~ck (29)
vention E is the real part of the product of and E0
given by
without any loss of generality; this is because the normal-
E(r, t) = k? (t)E?0 (r) + k (t)E0 (r) (20) ization choice comes from a particular choice of bound-
ary conditions, and those choices matter for deriving the
analogous to how x is chosen as the real part of in the Hamiltonian and momentum of the electromagnetic field,
harmonic oscillator. but the results of those two quantities ultimately do not
To get a better sense of how here relates to its har- depend on that choice.
monic oscillator counterpart, it is helpful to take the spa- In real space, plugging (20) into (14) yields the mag-
tial Fourier transform of E. The spatial Fourier trans- netic field
form of a function f = f (r) is given by [7]
i ?
B(r, t) = ( (t) E?0 (r) k (t) E0 (r)). (30)
k k
Z
3
f(k) = (2) 2 f (r)eikr d3 r. (21)
Using the results of (26) and (29), plugging (20) and (30)
Plugging (20) into (21) simply yields the same superpo- into the Hamiltonian density in (10) and integrating as
sition of the Fourier transform of each spatial function in (11) yields
E0 , resulting in
~ck ?
Hk = (k k + k k? ) (31)
E(k, t) = k? (t)E?0 (k) + k (t)E0 (k). (22) 2
as exactly the same Hamiltonian as (8) of the simple
The spatial Fourier transform of (15) is
harmonic oscillator, without having assumed anything
E about the electromagnetic field configuration besides its
ick B = (23) existence in a vacuum.
t Finally, the linear momentum of the field mode given
and plugging (22) into (23) yields in (13) can be found in much the same was as the Hamil-
tonian in (11); however, given that the Poynting vector
ek B(k, t) = i(k? (t)E?0 (k) k (t)E0 (k)) (24) in (12) involves the cross product of the fields, it is no
longer possible to perform these manipulations in fully
which gives the spatial Fourier transform of the magnetic general terms. Instead, it is now necessary to make use
field (in a different direction). The fields given in (22) and of the fact that a single mode is being considered to lead
(24) can be combined to give to plane wave solutions

k (t)E0 (k) = E(k, t) + iek B(k, t). (25) E0 (r) = eikr ek (32)

It is now immediately clear that E takes the role of x and where  is a real quantity of dimensions of the electro-
B takes the role of p in the harmonic oscillator, with the magnetic field and ek is a real-valued unit vector for the
modifications that these are vector quantities and these mode k pointing in the direction of E (to be distinguished
Electromagnetic Field Quantization & Applications to the Casimir Effect 4

from the unit vector ek = kk );  should be chosen to obey B. Harmonic Oscillator Analogies
the normalization in (29). Plugging (32) into the expres-
sions (20) and (30), taking the cross product of those Much of what was said for the nonrelativistic har-
two expressions in (12), and taking the volume integral monic oscillator remains true for the electromagnetic field
in (13) yields the linear momentum Hamiltonian. However, some quantities need to be rein-
terpreted in light of the fact that E and B are not exactly
~k ?
pk = (k k + k k? ) (33) the same as x and p, respectively.
2 The most important part is that for the familiar har-
which also obeys the dispersion relation E = cp. monic oscillator, there exists a single particle

which may
have a definite energy En = ~ck n + 12 , so the num-
ber n simply becomes a shorthand for the energy of that
IV. QUANTIZED ELECTROMAGNETIC FIELD particle. By contrast, the harmonic oscillator describing
the electromagnetic field does so for a particular mode k.
A. Canonical Quantization Photons, which are the quantized excitations of the elec-
tromagnetic field, carry a definite momentum p(k) = ~k
In single-particle mechanics, canonical quantization and energy E(k) = ~ck. This runs counter to the notion
entails using vectors |i in a Hilbert space as quantum that the electromagnetic field mode of number n has a
photon in it of energy En = ~ n + 21 . Instead, be-


states, and converting observable quantities into Hermi-
tian operators. This is done to quantities like x, p, and cause photons have the same energy for a given mode, the
any Y = Y (x, p). In the case of the electromagnetic field, number n is literally the number of photons in that given
now states are given by vectors |i in a Hilbert space mode; the state |ni has n photons contributing E =
~ck
1
rather than by values of the electromagnetic field. In- and p = ~k to the total energy E n = ~ck n + 2 and
momentum pn = ~k n + 21 .


stead, those fields and associated observable quantities,
like the Hamiltonian and momentum, are now operators. This leads to another important finding. In the famil-
In a manner exactly analogous to that of the single par- iar harmonic oscillator, h0| x |0i = 0 and h0| p |0i = 0 but
ticle harmonic oscillator, the variables k and k? are h0| x2 |0i > 0 and h0| p2 |0i > 0. This leads to uncertain-
promoted to the operators ak and ak . This means that ties in the position and momentum in the ground state,
the quantum electromagnetic Hamiltonian is tied to the nontrivial ground state energy E0 = ~ 2 . Sim-
ilar statements hold for the electromagnetic field, except
that in this case, the ground state n = 0 means that no
 
1
Hk = ~ck ak ak + (34) photons are present in this mode. Hence, the vacuum has
2
a nontrivial energy E0 = ~ck 2 . Furthermore, although
which is exactly analogous to that of the single particle it is counterintuitive that electric and magnetic fields
nonrelativistic harmonic oscillator with frequency = could exist in the absence of photons, h0| E2 |0i > 0 and
ck, with the same energy eigenstates |ni. Furthermore, h0| B2 |0i > 0, so the ground state experiences nontriv-
the linear electromagnetic field momentum is now the ial random fluctuations [3] in the electromagnetic fields.
operator These fluctuating fields are intimately tied to the non-
  trivial ground state energy. Similarly, the ground-state
1 momentum p0 = ~k 2 , so vacuum fluctuations in the elec-

pk = ~k ak ak + (35)
2 tromagnetic field carry momentum that can be imparted
on other objects even in the absence of photons.
which commutes with H.
Because the operators ak and ak play the same roles
as creation and annihilation operators as in the familiar C. General Mode Superpositions
harmonic oscillator, they satisfy a very similar commu-
tation relation as well [1]:
So far, the field operators for a single mode have been
considered. In general, many different modes k will exist
[ak , ak ] = 1. (36)
in space, with the details depending on boundary condi-
Given the dependence of E and B on ak and ak , the field tions; each mode k specifies a new harmonic oscillator,
operators E(r) and B(r) can then be found to satisfy a into which any number nk {Z : nk 0} photons can be
canonical commutation relation very similar to that of x added. The electric and magnetic field operators given in
and p [4]: (20) and (30) and the Hamiltonian and momentum op-
erators in (34) and (35) with and ? replaced by a and
[Ei (r), Bj (r0 )] = i~ij 3 (r r0 ). (37) a hold for a given mode k; the overall electromagnetic
fields, Hamiltonian, and momentum are the sums of the
This also has a similar meaning to the commutator individual mode operators.
[x, p] = i~: the electric and magnetic fields can no longer To quantitatively show this, it is necessary to make fur-
be measured without uncertainty. ther use of the spatial mode function E0 (r) = eikr ek ,
Electromagnetic Field Quantization & Applications to the Casimir Effect 5

where ek , which used to be called eE in (32), is a unit vec- as the allowed values of k are discretized by the periodic
tor dependent on k that specifies the direction of E and boundary conditions.
is not the unit vector ek in the direction of k. The nor- The vacuum state |0i can now be defined as n(k,j) = 0
malization condition only works in a finite volume V even for all (k, j). The overall energy and linear momentum
though the results are more generally true even in a vac- in that state then become
uum of infinite size, so the electric field coefficient
qcan be X ~ck
chosen from the normalization condition as  = 2~ck h0| H |0i = (45)
V . 2
k,j
The volume V is arbitrary as it does not ultimately in-
fluence the Hamiltonian and momentum, but it can be X ~k
h0| p |0i = . (46)
chosen for simplicity as a cubic box of side length l with 2
k,j
periodic boundary conditions; these conditions restrict
the allowed values of k to be k = 2 l N for Nj Z, The linear momentum in that state is zero because Nj
though the restrictions become unimportant in the limit Z so k + (k) = 0. However, the energy in that state
of l , which represents an infinitely large evacuated is not zero because (k) + (k) = 2ck > 0. In fact,
region. Furthermore, the charge-free Gauss law in (16) the energy ofP the vacuum is infinite because that sum
with E0 (r) = eikr ek yields the condition k ek = 0. diverges like x x. However, while the energy of this
This has two linearly independent solutions e(k,j) satis- vacuum configuration may be infinite, that turns out to
fying e(k,i) e(k,j) = ij for {i, j} {1, 2}; these unit vec- not matter in the analysis of the Casimir force as long
tors describe the two independent polarizations of light. as the differences between infinite energies of different
Now the electromagnetic field modes are no longer in- vacuum configurations are finite.
dexed only by the wavevector k, but must also include
the polarization j. Hence, the electromagnetic fields for
a given mode (k, j) are V. CASIMIR FORCE
r
2~ck  ikr ikr

A. Energy Gradient
E(k,j) (r, t) = e a(k,j) + e a(k,j) e(k,j)
l3
(38)
r
2~ck  ikr 
B(k,j) (r, t) = 3
e a(k,j) + eikr a(k,j) k e(k,j)
l
(39)

and the electric and magnetic field operators accounting

for all modes are
X
E(r, t) = E(k,j) (r, t) (40)
k,j
X
B(r, t) = B(k,j) (r, t). (41)
k,j

Using these in conjunction with the normalization condi-

tion leads to the fact that the Hamiltonian and momen-
tum accounting for all modes are simply the respective FIG. 2: Discrete modes between the plates overwhelmed by
sums of the individual mode Hamiltonians and momenta continuous modes outside the plates [6]
 
X X 1
H= H(k,j) = ~ck a(k,j) a(k,j) + (42) The Casimir force is the force that surfaces with
2 no charges or currents feel with respect to each other
k,j k,j

1
 due purely to quantum ground state fluctuations in the

X X
p= p(k,j) = ~k a(k,j) a(k,j) + . (43) electromagnetic field. This force is highly dependent
2
k,j k,j in both magnitude and direction on both the dielec-
tric/conductive properties of the materials and medium
In this, the creation and annihilation operators, which as well as on the geometrical configuration of the materi-
now literally create and annihilate photons in a mode als and medium. The simplest configuration in which to
(k, j) given that n
 (k,j) is the number of photons in that analyze the Casimir force is for two parallel perfectly con-
mode and n(k,j) is the corresponding eigenstate, now ducting plates of infinite area and finite separation in a
satisfy the commutation relations vacuum, in which case the Casimir force is attractive for
the two plates; if the plates are separated by a distance
[a(k,j) , a(k0 ,j 0 ) ] = (k,k0 ) (j,j 0 ) (44) s, then the magnitude of the Casimir force per unit area
Electromagnetic Field Quantization & Applications to the Casimir Effect 6

is proportional to s4 . One interpretation of the nature and the discretization of kx and ky were enforced, the
of the Casimir force between two parallel plates is that energy would still be infinite as all modes Ni 0 would
although the energies of the ground state of the field are contribute positively to the energy.
infinite both inside and outside of the plates, the energy
When the plates are very far separated, the discretiza-
inside is lower because the modes k are discrete inside
tion of kz begins to matter less as well, and the sum over
and continuous outside, so more modes contribute to the
N can be replaced by an integral with respect to kz nor-
energy outside. Because of this, there is a gradient in
malized by the spacing s . The energy in (47) becomes
energy from inside to out, and the electromagnetic fields
would tend to lower energies by pushing the plates to-
gether. lim E(s) = (49)
The easiest way to perform this energy-based analysis s
~cl s q 2
2 Z Z Z
is to consider a rectangular cavity with perfectly conduct-
kx + ky2 + kz2 dkx dky dkz (50)
ing walls of side length l and s; in terms of the infinite 3 0 0 0
plates of finite separation, l represents the dimensions of
the plates, while s represents the separation, so the limit
l  s will be considered. If the plates lie parallel to which is also infinitely large.
the xy-plane so that the separation is along the z-axis, The potential energy of the plate configuration U (s) =
then Ex and Ey need to vanish along the boundaries of E(s) lims E(s), which is the difference between (47)
the plates. This means that kx and ky can take on any and (49), is given by
nonnegative values as the boundaries in the planes per-
pendicular to the plates become irrelevant for arbitrarily
large l; the spacing between consecutive values of kx and
s 2
~cl2 X 0
ZZ 
ky is l , but this tends to 0 as l . However, kz be- N
U (s) = 2 ( kx2 + ky2 + dkx dky
comes discretized as kz = N s
s for N {Z : N 0} so N
that the boundary conditions at the plates are satisfied. (51)
In this analysis, the first quantity to determine is s
ZZZ q
the ground state energy over all modes (k, j) for infi- kx2 + ky2 + kz2 dkx dky dkz ) (52)

nite plates separated by s. The potential energy of the
plate configuration and separation s is given by the dif-
ference between this ground state energy at finite s and where the limits of integration on each ki are again from
the ground state energy at s . This potential en- 0 to . This
q can be rewritten in cylindrical
  coordinates
ergy still diverges unless a cutoff function is introduced [4] k =
k
kx2 + ky2 and = arctan kxy ; there is no
at some high magnitude of k = kc . Finally, perform-
ing further integration simplifications, appealing to the dependence of U (s) on , and its limits of integration
Euler-Maclaurin formula and its properties, and further are from 0 to 2 as kx 0 and ky 0 so a factor of 2
slightly restricting the cutoff function properties allows comes out front, while the integral over k goes from 0
the derivation of the Casimir force result. to as expected. This yields
To begin the quantitative analysis, it is important to
note that the frequency in the energy spectrum becomes s 2
2~cl2 X 0
q ZZ 

2 N
(kx , ky , N ) = c kx2 + ky2 + N s . The ground state U (s) = ( 2
k + k dk (53)
s
energy is then a sum over N and an integral over kx ZZZ qN
and ky ; for dimensionality, the latter integrals need to be s
normalized by the spacing l , so k2 + kz2 k dk dkz ). (54)

X 0 ZZ ~(kx , ky , N ) dkx dky
E(s) = 2 (47)
2 l l The problem is that this integral still diverges as all
N
s modes k and kz contribute positively to the energies.
2
~cl2 X 0
Z Z 
2 2
N The solution is to renormalize by introducing a cutoff
= 2 kx + ky + dkx dky
0 0 s function f = f (k) where k 2 = k2 = k2 + kz2 that satisfies
N =0
limkkc f (k) = 1 and limkkc f (k) = 0. The purpose of
(48)
this is that at high photon energies, the walls may not
where in the original sum, the extra factor of 2 comes be perfectly conducting boundaries anymore; the cutoff
from the two different polarizations j having the same kc may be chosen to be the reciprocal of the Bohr ra-
energy for the same k and the prime on the sum indi- dius [4], as radiation of energy E = ~ckc will be able to
cates that a factor of 21 should be prepended to the term probe that length scale and find that the boundary is not
for which N = 0 as that only has one polarization [4]. perfectly conducting at the atomic scale. That said, the
This energy, despite being contained in a finite separa- particular form of the cutoff turns out to be irrelevant in
tion s, is infinitely large; in fact, even if l were finite the end. With the integrals now weighted by the cutoff,
Electromagnetic Field Quantization & Applications to the Casimir Effect 7

the potential energy in (53) becomes cavity because both modes k and k contribute to the
overall momentum, only those modes with k directed to-
2~cl2 ward a given plate impart momentum on that plate, so
U (s) = (55)
s only those will be counted. As more modes exist outside
the plates than inside, more momentum is imparted from
s  2  2
X 0 ZZ N N
( 2
k + f 2
k + k dk outside than from inside, so the net force, summing all
s s those impulses over infinitesimal time intervals, is attrac-
N
(56) tive for the plates.
ZZZ q In general, pressure is a force per unit area. At normal
s q 
k2 + kz2 f k2 + kz2 k dk dkz ). (57) incidence, if a photon is absorbed by a plate, then the
plate acquires the full photon momentum, and the radia-
It is convenient to define new integration variables tion pressure is the impulse divided by the product of the
k2 s2 kz s area and the time interval. These are doubled if the pho-
2 and and rewrite some of the integrands as ton is reflected, as that can be considered an absorption
Z p p  and then reemission out of phase each with the same
g() = + 2f + 2 d (58) impulse. If the pressure is taken to be constant, the force
0 s per unit area is just the same as the energy per unit vol-
so the potential energy in (53) then becomes ume, which is the volume energy density. Off normal inci-
dence for an angle , the effective area is increased by the
S
1
factor cos() , so S 0 = cos() ; meanwhile, the effective force
Z !
2 ~cl2 g(0) X
U (s) = + g(N ) g() d (59) is reduced by the factor cos(), so F 0 = F cos(). Fur-
4s3 2 0
N =1 thermore, modes (k, j) and (k, j 0 ) contribute the same
momentum and therefore the same pressure, so another
because between the plates, N = kz s {Z : N 0} so factor of 2 needs to be added. That said, the energy den-
= N there. sity has equal contributions from forward- and reverse-
The Euler-Maclaurin formula [4] says that propagating waves, so the pressure needs to be reduced
Z by a factor of 2. Hence the pressure for a given mode k
X 2
g(N ) g() d = (60) becomes pk = 2E(k)Vcos () .
0
N =1 For the ground state, E(k) = ~(k) 2 . The pressure for
g(0) 1 dg() 1 d3 g() a given mode k accounting for both possible values of j
|=0 + |=0 (61)
2 12 d 720 d 3 then becomes
n
assuming that the cutoff function satisfies dd nf = 0 at ~ckz2
pk = (64)
= 0 for all n 1 and assuming that g satisfies Vk
lim g() = 0. In (53), the term involving g(0) can-
cels; meanwhile, g 0 (0) = 0 and g 000 (0) = 4. This finally as (k) = ck and cos() = kkz . The pressure from inside
yields the potential energy only has contributions from discrete modes kz = N s for
N {Z : N 1}. Also, the volume of the rectangular
2 ~cl2 cavity is V = l2 s. Hence, that pressure then becomes [4]
U (s) = (62)
720s3 Z N 2
~c X
Z

s
for which the Casimir force per unit area pin = dkx dky
2 s
q
N 2
0 0


N =1 kx2 + ky2 + s
1 U 2 ~c
p(s) = = (63) (65)
2
l s 240s4 as the integrals over kx and ky are each normalized by
is indeed attractive and depends on the separation as the spacing l . Meanwhile, the pressure from outside has
s4 . contributions from continuous modes kz 0, so that sum
becomes an integral normalized by the spacing s , so [4]

kz2
Z Z Z
B. Momentum & Radiation Pressure ~c
pout = dkx dky dkz .
3
q
0 0 0 kx2 + ky2 + kz2
The momentum in (13) is intimately tied with the ra- (66)
diation pressure cp = hSi. Photons carry momentum The pressure difference is then just
p = ~k, so when they strike an object like a perfectly
conducting plate, that differential impulse over an in- p(s) = pin pout (67)
finitesimal time interval is a force, and that force per
unit area of the plate is the radiation pressure. Although and using the exact same integration techniques, cut-
it was previously stated that h0| p |0i = 0 in a whole off function, and Euler-Maclaurin formula as before, the
Electromagnetic Field Quantization & Applications to the Casimir Effect 8

Casimir pressure is directly found to be discussed. Finally, these were all combined to discuss the
Casimir effect on two parallel neutral conducting plates
1 U 2 ~c from the perspectives of energy gradients and radiation
p(s) = = (68)
l2 s 240s4 momentum & pressure; in both cases, the pressure was
found to depend on the fourth power of the reciprocal of
matching the Casimir force calculation from energy gra- the plate separation.
dient considerations.

VI. CONCLUSIONS
Acknowledgments
In preparation for discussing the electromagnetic field,
the nonrelativistic harmonic oscillator was discussed in I would like to thank Rishi N. Patel and Ethan Dyer for
depth in both classical and quantum mechanics. The har- their extremely helpful comments and suggestions during
monic oscillator Hamiltonian was then derived for elec- the editing process. I would further like to thank Prof.
tromagnetic fields in a vacuum, and a similar expression Jesse Thaler for inspiring me to investigate a topic that
was observed for the linear momentum of the electromag- might help me further explore available undergraduate
netic field. These fields, Hamiltonian, and momentum research opportunities, as has been the case for this pa-
were all quantized as operators, and further analogies per. Finally, I would like to dedicate this paper to the
were made with the quantum harmonic oscillator. Con- memory of the honorable MIT police officer the late Sean
siderations of general superpositions of field modes were Collier.

[1] David Tong, Quantum Field Theory: University of Cam- [4] Peter W. Milonni, The Quantum Vacuum: An Introduc-
bridge Part III Mathematical Tripos. 2006. tion to Quantum Electrodynamics. Academic Press, 1994.
[2] Kaustubh Agashe, University of Maryland Physics 624: [5] Allen McCloud, Wavefunctions of a quantum harmonic
Homework 1 (Solutions). 2010. oscillator. 2010.
[3] Alejandro W. Rodriguez, M. T. Homer Reid, Steven [6] Emok, Illustration of the Casimir effect. 2008.
G. Johnson, Fluctuation-induced phenomena in nanoscale [7] W. Craig Carter, Massachusetts Institute of Technology
systems: harnessing the power of noise. Proc. IEEE, 3.016: Fourier Transforms. 2005.
101(2), 2013.