UNIVERSITY of CALIFORNIA

SANTA CRUZ

THE CASIMIR EFFECT AS A PHENOMENON PRODUCED BY

ZERO POINT ENERGY

A thesis submitted in partial satisfaction of the

requirements for the degree of

BACHOLER OF SCIENCE

in

APPLIED PHYSICS

by

Joshua Najmi

June 2019

The thesis of Joshua Najmi is approved by:

Professor Michael Dine

Thesis Advisor

Professor Robert Johnson

Chair, Department of Physics

 Abstract

The attraction of two uncharged conducting plates in vacuum, known as

the Casimir effect, has no classical explanation. In quantum mechanics, things

are different; even in the ground state, a system has non-zero energy. We can

model the vacuum space between the plates as a superposition of quantum

harmonic oscillators each with a discrete energy mode. Boundary conditions

imposed by the conducting plates only allow discrete values of the energy to

exist. The energies of each of these modes are different than in the absence

of the plates, leading to a force. A calculation of this force exerted by the

free electromagnetic field on the boundaries shows that the Casimir force is

proportional to the separation of the plates by s−4 .

2
Contents

1 Introduction 4

2 Background of the Classical and Quantum Harmonic Oscilla-

tor 5

2.1 The Classical Harmonic Oscillator . . . . . . . . . . . . . . . . . 5

2.2 The Quantum Harmonic Oscillator . . . . . . . . . . . . . . . . . 6

3 Quantizing the Electromagnetic Radiation Field 9

3.1 The Classical Electromagnetic Field . . . . . . . . . . . . . . . . 9

3.2 Quantum Mechanical Treatment of the Electromagnetic Field . . 13

4 The Casimir Energy 16

5 Conclusion 18

3
1 Introduction

The Casimir effect is an experimental phenomenon that showed there exists an

attraction between two uncharged, non-conducting plates in vacuum. Branching

in a new arena of physics research, Casimir physics is now a heavily studied

branch of condensed matter and nanotechnology, and it seeks to understand

quantum electrodynamics at a deeper level.

After H.B.G Casimir and D. Polder published the first paper introducing

the Casimir effect called ”The Influence of Retardation on the London-van der

Waals Forces” in 1947, the source of this energy has been a point of controversy.

The Casimir effect was widely considered to be a manifestation of the elusive

zero point energy. This paradigm stands in contrast to the original investigation

that Casimir and Polder set out to understand.

Originally, their investigations sought to calculate the Van Der Waals force

accounting for retardation effects between atoms with static polarization at

1
large distances giving a r7 dependence. They also derived an expression for the

interaction between statically polarized atoms with a perfectly conducting plate

1
and showed that it has a r4 dependence [2]. The calculation of the Casimir force

while treating it as a modified version of the Van Der Waals force by including

relativistic effects proves to be a mathematically difficult task. The results of

this approach produces similar analytic results as treating the phenomenon as

a quantum electrodynamic problem.

In this paper, we explore a calculation of the potential energy arising from

the Casimir effect as a phenomenon of quantum electrodynamics. In quantum

electrodynamics, the electromagnetic field can be treated as a collection of quan-

tum harmonic oscillators each with a discrete energy mode k. The energy mode

becomes quantized due to the boundary conditions from the setup of our model

imposed on the radiation field. Photons are manifestations of the electromag-

4
netic field’s quanization with momentum p = h̄k. Each harmonic oscillator at

level n can be thought of as describing n photons with momentum h̄k. We

derive an expression for the Hamiltonian of the radiation field by changing it

to be in terms of the strength of the electric field q and the radiation field’s

momentum p. It then acts on a wave function expressed as a summation of

the quantum harmonic oscillators. As we take the Hamiltonian of the ground

h̄ck
state wavefunction, we get E0 = 2 . Since the ground state contains energy

proportional to k, the energy mode can be integrated for each dimension over all

possible values. This value must be compared to the ground state energy that

exists without the imposed boundary conditions to see that there is a potential

known as the Casimir energy.

2 Background of the Classical and Quantum Har-

monic Oscillator

2.1 The Classical Harmonic Oscillator

In the case of the classical harmonic oscillator, an object of mass m on a spring

1 2
with a spring constant k is bound by a potential energy V (x) = 2 kx with

motion given by Hookes Law,

d2 x
F = −kx = m (1)
dt2

The solution of this equation is given by sinusoidal functions of the form x(t) =
q
k
Asin(ωt) + Bcos(ωt) where the angular frequency is ω = m . The behavior

of this system is implied in its solution where the objects displacement path is

shaped as a parabola moving from a global minimum in which kinetic energy

is maximized to the particles maximum displacement when the greatest value

5
of the systems potential energy is reached. The energy of such a system can be

described by writing the Hamiltonian as

1 2
H= (p + (mωx)2 ) (2)
2m

The classical expression of the Hamiltonian gives us a defined value for the total

energy of the system which is conserved throughout time. This differs from the

quantum mechanical expression of the Hamiltonian in which observables such

as momentum p and position x are not definite quantities until acted on by its

corresponding operator. These operators work on a vector living in phase space

(x,p) to transform it to another vector defining the state of the system.

2.2 The Quantum Harmonic Oscillator

To understand the quantum mechanical analog of this model, we must only use

allowed solutions to the Schrodinger equation with potential V (x) = 12 kx2 given

to us from the classical harmonic oscillator. For our purposes, we will only find

solutions to the Time Independent Schrodinger Equation (TISE) given by

h̄2 d2 ψ
− + V ψ = Eψ (3)
2m dx2

The TISE can be rewritten in the Hamiltonian formalism as

1 2
[p̂ + (mωx̂)2 ]ψ = Ĥψ (4)
2m

d
where p = −ih̄ dx . The Hamiltonian has now changed the observables p and x

into operators that follow the canonical commutation relation, [x, p] = ih̄. The

non-zero commutation relation implies that there exists fundamental uncertain-

ties in the simultaneous measurement of x and p. This commutation relation

6
is the basis of the Heisenberg Uncertainty Principle, a fundamental axiom of
h̄
quantum mechanics, restricting hxihpi ≥ 2 [7]. To find solutions to the TISE,

we employ an algebraic technique using ladder operators defined as

1
â = √ (−ip̂ + mωx̂) (5)
2h̄mω

and its complex conjugate, also called the adjoint operator,

1
↠= √ (ip̂ + mωx̂) (6)
2h̄mω

to create solutions that are linearly proportional to the base solution ψ0 for each

consecutive energy level, shown in equation 13 and 14. Ladder operators are

not Hermitian. An operator is Hermitian if arbitrary functions f(x) and g(x)

were integrated for all space and operated on by â and ↠respectively,

Z ∞ Z ∞
∗
f (x)âg(x)dx = ↠f ∗ (x)g(x)dx (7)
−∞ −∞

Since these ladder operators do not follow this condition, we must be careful

using them. Commutation with these operators does not work the same as with

constants because operators acting on a function produce distinct values when

ordered differently. Commutation of the ladder operators ↠â and â↠yield
1 1 1 1
h̄ω Ĥ + 2 and h̄ω Ĥ − 2 respectively. With the use of these operators, equation

4 can be rewritten as
1
Ĥ = h̄ω(a† a + ) (8)
2

for lower energy levels and

1
Ĥ = h̄ω(aa† − ) (9)
2

7
for higher energy levels [6]. Thus, the Schrodinger equation for the harmonic

oscillator can be written as

1
h̄(a† a + )ψ = Eψ (10)
2

1
h̄(aa† − )ψ = Eψ (11)
2

With these ladder operators, we shall scale the ladder of discrete energy levels

that may exist for the harmonic oscillator that is under investigation. At the

bottom of the ladder, the adjoint operator acts on ψ0 (x),

1 d
√ (h̄ + mωx)ψ0 (x) = 0 (12)
2h̄mω dx

By implementing the operator through and algebraically manipulating so as to

express equation 12 as an ordinary differential equation, we solve for ψ0 (x) to

get
mω 1 −mω x2
ψ0 (x) = ( ) 4 e 2h̄ (13)
πh̄

To progress up to the next energy level, we can use the raising operator to

produce the equation for the state

ψn (x) = An (â)n ψ0 (x) (14)

where An is a normalization constant with energy En = (n + 21 )h̄ω. To solve

the TISE, we plug in equation 13 for the ground state energy, n = 0, to get

E0 = 12 h̄ω. The presence of a ground state energy produced by the conditions

the quantum harmonic oscillator must obey is consistent with the Heisenberg

Uncertainty Principle. Expectation values of position and momentum should

yield hxi = 0 and hpi = 0. Yet, the uncertainty principle suggests that fluctua-

8
tions hx2 i > 0 and hp2 i > 0 can exist even in the ground state [7].

3 Quantizing the Electromagnetic Radiation Field

To bridge the model of a quantization of the electromagnetic field, we will

introduce a vector potential A to show the Hamiltonian for the electromagnetic

field as a collection of quantum harmonic oscillators.

3.1 The Classical Electromagnetic Field

In the classical study of electromagnetism, we can write Faradays law as

1 dB
∇×E =− (15)
c dt

which explains that a spatial and time varying magnetic field B induces a current

on a circuit. Following the vector identity ∇ × ∇φ = 0 where φ is a scalar

potential, we can rewrite equation 15 as

1 dA
∇ × (E + )=0 (16)
c dt

1 dA
which suggests that (E + c dt ) = ∇φ. Inversely, Amperes Law given by

4φ 1 dE
∇×B = j+ (17)
c c dt

states that a varying electric field E or current source j induces a magnetic

field. In our model, the current source and electric charges are absent making

Amperes Law reduce to

1 dE
∇×B = (18)
c dt

9
We must introduce the Coulomb gauge allowing us to measure the change of φ

by the charge distribution to impose constraints on the vector potential A from

the fact that is φ is a scalar mapping of A. In the absence of a current source

and charges, the constraint we want to impose is ∇ · A = 0 which will apply to

the vector potentials of the electric and magnetic fields to give us ∇ · E = 0 and

∇ · B = 0. The nonrelativistic Hamiltonian gives

(p − qA)2
H= + qφ (19)
2m

for a particle of mass m and charge q that interacts with a electromagnetic field

[3]. If we take into account the absence of a current source and charges and

integrated for three dimensions, the Hamiltonian can be rewritten as

Z
1
H= d3 r(E 2 + B 2 ) (20)
8π

In classical electrodynamics, the electromagnetic field is treated as a wave the

has propagation in an orthogonal direction to both the electric and magnetic

field. This direction is expressed by the Poynting vector which is given by

c
S= 4π (E × B). Due to this wavelike nature, the electromagnetic field of a free

particle follows the equation of motion for a wave is expressed as

1 d2 E
− ∇2 E = 0 (21)
c2 dt2

where the contribution to the electric field from charges and currents have been

set to zero. Solutions to this equation follow from the equation for a plane wave,

A = A0 ei(kx−ωt) where A0 is a normalization constant, k is the wavenumber,

and ω is the wave’s angular frequency given by ω = ck. The general plane wave

equation would suffice to give us a value satisfactory our wave equation, yet

10
our purpose is to quantize the radiation field by imposing boundary conditions

meaning not all values of the wavenumber k are allowed to be solutions. This

implies that a general expression of the wave equation can be written as a sum

of plane wave solutions each characterized by a unique energy mode k.

For our model of the existence of zero point energy in the cavity between a

three dimensional cube of length L made of uncharged non-conducting plates,

we impose the boundary condition that the electromagnetic wave should not

exist beyond the boundary of the cubic walls. This implies that the exponential

in the wave equation, eikx , should die out at x = 0 and x = L. This would

require the following:

kx,y,z = 2πnx,y,z (22)

nx , ny , nz = 0, ±1, ±2, ... (23)

To obey our imposed boundary conditions, the energy levels in each dimension

nx,y,z must take on only positive and negative integer values. Now, there only

exists a discrete set of solutions to our wave equation.

We must return back to the coulomb gauge, ∇ · A = 0, where the vector

potential A follows the form of the plane wave accounting for the discrete values

of k. The coulomb gauge constraint gives us k · (k, λ) = 0, where  is a unit

vector defining the polarization of the vector potential. For each discrete energy

mode k, the vector potential can be polarized in two directions both orthogonal

to the wave’s direction of propagation. For our purposes, λ will equal 1 to

indicate polarization in the x-axis and equal 2 to indicate polarization in the

y-axis, which are both orthogonal to the z-axis [7].

We can now write a generalized solution to the wave equation expressed as

X eikr−ωt e−ikr−ωt
A(r, t) = (ck,λ (k, λ) √ + c∗k,λ (k, λ) √ ) (24)
k,λ
V V

11
where c∗k,λ and ck,λ are amplitude values for each plane wave with energy mode

k and V is the volume of the wave’s probability [5]. We can rewrite equation

25 to represent the electric and magnetic fields separately as they both take on

discrete values of k. These will look like: For the electric part,

X eikr−ωt ∗ e−ikr−ωt
E(r, t) = (Ek,λ (k, λ) √ + Ek,λ (k, λ) √ ) (25)
k,λ
V V

and we must plug equation 26 into equation 15 to get the magnetic part,

iX eikr−ωt ∗ e−ikr−ωt
B(r, t) = (Ek,λ (k, λ)∇ × √ − Ek,λ (k, λ)∇ × √ ) (26)
k V V
k,λ

[7]. Now, our goal is to express the Hamiltonian given in equation 20 with

our new expressions for E and B. When substituting equation 25 and 26 into

equation 20, we must recognize that k and λ values are summed over twice

because both the electric and magnetic fields are squared leaving a double sum.

k and k 0 will be used to make the distinction. To work through this integral,

we can use an orthonormality relation given by

Z
1 0
√ eikr e−ik r d3 r = δk,k0 (27)
V

The Kronecker Delta function helps us sum these distinct wave equations in-

dependently. This also applies to summation for each mode k with its unique

polarization denoted by λ and λ0 . There exists a commutation relation for λ

of the same form as for the energy mode k. Using these commutation relations

and integrating for cubic length, we get the Hamiltonian of the electromagnetic

field expressed for a closed system,

1 X ω2 ∗
H= E Ek,λ (28)
2π c2 k,λ
k,λ

12
[7]. We can rewrite the Hamiltonian to generalize it for any vector potential,

giving us
1 X ω2 ∗
H= c ck,λ (29)
2π c2 k,λ
k,λ

With employing the change of variables used by Dirac where the coefficient c∗k,λ

and ck,λ can be rewritten to be functions of charge q and momentum p. This

gives us
1
qk,λ = √ (ck,λ + c∗k,λ ) (30)
c 4π

for the charge and

−iω
pk,λ = √ (ck,λ − c∗k,λ ) (31)
c 4π

for the momentum. By manipulating equation 29 and 30 to solve for c∗k,λ and

ck,λ , we can plug the results back into equation 28, multiply through, and cancel

out terms to derive a new expression for the Hamiltonian,

1X 2
H= (pk,λ + ω 2 qk,λ
2
) (32)
2
k,λ

This result is the Hamiltonian of the quantum harmonic oscillator for each

discrete energy mode k with polarization λ. We shall use the solution we derived

for the quantum harmonic oscillator to model the electromagnetic radiation.

3.2 Quantum Mechanical Treatment of the Electromag-

netic Field

The use of operators is the bridge between a classical analysis of a system to

the quantum mechanical approach. These operators act on vectors that exist

in Hilbert space, a complete inner product space in any defined basis, which

allows for operators to produce observable quantities. In Dirac notation, states

of a system are defined by abstract vectors called ”ket” vectors, |ψi. For our

13
representation of the electromagnetic field as a quantum harmonic oscillator, we

must translate the observable quantities in the classical treatment to operators.

The Hamiltonian itself must become an operator that acts on a vector |ψi.

The new Hamiltonian should be a quantum analog of that for classical three

dimensional oscillator expressed as

X1
Ĥ = (p̂2n + mω 2 n̂2 ) (33)
n
2

where n = x,y,z for each dimension. The ladder operators given by equation 5

and 6 will help us transmute momentum p and charge q from equation 32 into

operators. For each value k and its polarization λ, we can use these operators to

scale through the energy levels to produce values for the desired observables. By

rewriting equation 5 and 6 in terms of momentum and charge and rearranging

terms, we get
r
ω i
â†k,λ = (q̂k,λ − p̂k,λ ) (34)
2h̄ ω

for raising the energy level and

r
ω i
âk,λ = (q̂k,λ + p̂k,λ ) (35)
2h̄ ω

for lowering the energy level. The coefficients ck,λ and c∗k,λ in Equation 29 that

gives us the generalized Hamiltonian must be redefined as

r r
2πh̄ 2πh̄ †
c âk,λ c â (36)
ω ω k,λ

respectively. By rearranging equation 34 and 35 to solve for p and q and using

the relationships above, we can rewrite equation 24 as

r
2πh̄ eikr−ωt e−ikr−ωt
+ â†k,λ (k, λ) √
X
Â(r, t) = c (âk,λ (k, λ) √ ) (37)
ω V V
k,λ

14
To derive the Hamiltonian, we can do the same treatment we did with the

classical form. By plugging equation 37 into equation 20 as the vector potential

that E and B derive from and integrating for cubic length, the new Hamiltonian

becomes
1
h̄ω(a†k,λ ak,λ + )
X
Ĥ = (38)
2
k,λ

[6]. The derivation of the Hamiltonian follows the commutation relations of the

ladder operators given by

[ak,λ , â†k,λ ] = δk,k0 δλ,λ0 (39)

Before using the Hamiltonian, we must recognise that the electric field E and the

magnetic field B do not exactly equal the position x and momentum p operators

in equation 33. In the normal quantum harmonic oscillator, the energy for a

single particle is expressed as En = (n + 12 )h̄ω where n is the discrete energy

state taking on only non-negative integer values. In our new expression for the

Hamiltonian, the ladder operators now work as “annihilation” and “creation”

operators. The operators â and ↠count the number of photons with energy

E = h̄ck. The number of photons with each discrete excitation mode k will give

n its value. We assume that there are no photons (n = 0) that exists in vacuum

implying that |ψi = |0i. Taking the expectation value of the Hamiltonian with

the ground state vector |0i we get

X h̄ck
h0|H|0i = (40)
2
k,λ

[7]. As stated in section 2.2, this result supports our model that energy fluctu-

ations in vacuum can exist.

15
4 The Casimir Energy

To understand the energy that arises from ground state fluctuations producing

the Casimir effect, we will consider how two parallel uncharged conducting plates

with dimensions L2 . The plates will be parallel to the xy plane separated by a

distance s on the z-axis resulting in the Casimir force having s−4 dependence.

With this setup, Ex and Ey should vanish at the boundary of the plates requiring

a quanitzation of energy described by kx and ky . We now recognize that kx and

π
ky can take on any positive integer value with a difference of L in order to

satisfy the wave equation. At limL→∞ , we assume that the modes of kx and ky

nearly become infinite allowing them to become continuous variables that can

be integrated over. While in the z direction, the axis normal to the parallel to
πnz
the plates, kz is constrained only to have values equal to kz = s . We will

handle the discretization of kz by summing through values of nz which become

exceedingly large tending to infinity at separation s.

We will first start by writing an expression for the ground state energy within
h̄ck
the cavity. Since we know that the ground state energy is E0 = 2 , we can

rewrite this for all possible modes as

∞ Z ∞Z ∞
h̄c X
E0cavity = k dkx dky (41)
2 n =0 0 0
z

q
where k = kx2 + ky2 + ( πns z )2 . To account for the polarization of vector (k, λ)

where  can point in two different directions for each energy mode k, we will
π 2
multiply our integrand by 2. We must also divide by ( L ) to get integer values

for kx and ky . Now, equation 40 becomes

∞ Z ∞Z ∞r
h̄c L 2 X πnz 2
E0cavity = ( ) 2 kx2 + ky2 + ( ) dkx dky (42)
2 π n =0 0 0 s
z

16
To calculate the potential contributed by the Casimir energy, we must compare

the difference of the ground state energy with and without the cavity. The

latter is given if we assume that the separation s of the plates is infinitely large

allowing us to integrate kz . This is expressed as

∞ Z ∞Z ∞Z ∞q
h̄c L 2 s X
E0nocavity = ( ) 2 kx2 + ky2 + kz2 dkx dky dkz (43)
2 π π n =0 0 0 0
z

The potential arising from the Casimir energy can now be calculated as U (s) =

E0cavity − E0nocavity .

∞ Z ∞Z ∞r
h̄c L 2 X πnz 2
U (s) = ( ) 2 kx2 + ky2 + ( ) dkx dky
2 π nz =0 0 0 s
Z ∞Z ∞Z ∞q (44)
h̄c L s
− ( )2 2 kx2 + ky2 + kz2 dkx dky dkz
2 π π 0 0 0

[1]. This integral diverges and the potential will be infinitely large. In the

original calculation for the Casimir energy, Casimir and Polder introduced a

cutoff function for high values of k causing limk→kc f (k) = 0, where kc is the

highest energy mode considered. Physically, this means that the function would

exclude high energy photons such as x-rays that would not be bounded by

the conducting plates’s boundary conditions. By translating equation 43 to

cylindrical coordinates to make integration easier, Casimir and Polder go onto

employing the Euler-Maclaurin formula to evaluate the integral. Casimir and

Polder calculated the potential to be given by

π 2 h̄cL2
U (s) = − (45)
720s3

The force per unit area acting on the plates is

1 ∂U π 2 h̄c
F (s) = − = − (46)
L2 ∂s 240s4

17
[1].

5 Conclusion

By treatment of the electromagnetic radiation field existing in vacuum as a set of

quantum harmonic oscillators, we arrive at a finite solution. These calculations

only work for the specific model that was investigated with assumptions that the

plates act as perfect conductors where the electromagnetic wavefunction would

vanish and the area of the plates parallel to the xy plane are nearly infinite

in size. An experiment done by S. K. Lamoreaux in 1996 is recognized to be

the first successful measurement of the Casimir force. He sought to confirm the

theoretical result for the force that Casimir and Polder predicted. Lamoreaux

constructed his experiment with a flat conducting surface and a conducting

sphere so that trouble in maintaining parallelism between the objects would

be avoided. Due to the geometrical change from parallel plates, the force is

approximated by the proximity force theorem giving F = 2πRE where E is the

potential energy per unit area and R is the radius of curvature of the spherical

conductor. Using the new expression of the force, it can be rewritten as

1 π 2 h̄c
F (s) = 2πR( ) (47)
3 240 s3

With Lamoreaux’s setup, the force is undetectable at around a separation of

10µm and increases non-linearly at distances less than 0.5µm. These results

were obtained to a precision of 5% and strongly support the validity Casimir

and Polder’s prediction [4].

18
References

[1] Casimir, H. (1948). On the Attraction Between Two Perfectly Conducting

Plates. Frontiers in Physiology, 61-63.

[2] Casimir, H.B.G. Polder, D. (1948). The Influence of Retardation on the

London-van der Waals Forces. APS Journals. 73. 10.1103/physrev.73.360.

[3] Kobe, D.H., Wen, E.C.-T. (1999). Gauge invariance in quantum mechanics:

Charged harmonic oscillator in an electromagnetic field. Journal of Physics

A: Mathematical and General. 15. 787. 10.1088/0305-4470/15/3/018.

[4] Lamoreaux, S.K. (1997). Demonstration of the Casimir Force in the 0.6

to 6m Range. Physical Review Letters - PHYS REV LETT. 78. 5-8.

10.1103/PhysRevLett.78.5.

[5] Passante, R., Rizzuto, L., Spagnolo, S., Tanaka, S., Petrosky, T. (2012).

Harmonic oscillator model for the atom-surface Casimir-Polder interaction

energy. Physical Review A. 85. 062109. 10.1103/PhysRevA.85.062109.

[6] Tong-Zhong, Li. (2004). New Vacuum State of the Electromagnetic Field-

Matter Coupling System and the Physical Interpretation of Casimir Effect.

Chin. Phys. Lett. 21 73-75

[7] Venkataram, Prashanth. (2013). Electromagnetic Field Quantization Ap-

plications to the Casimir Effect. MIT Department of Physics: 8.06.

19