THE CASIMIR EFFECT

Joseph Cugnon
University of Liège, AGO Department, allée du 6 Août 17, bât. B5,
B-4000 Liège 1, Belgium

Abstract

The Casimir effect is usually interpreted as due to the modification of the zero point
energy of QED when two perfectly conducting plates are put very close to each other,
and, consequently, as a proof of the “reality” of this zero point energy. The Dark
Energy, necessary to explain the acceleration of the expansion of the Universe is
sometimes viewed as another proof of the same reality. The usual interpretation of
the Casimir effect is however challenged by some authors who rather consider it as
a “giant” van der Waals effect. All these aspects are discussed.

1 Introduction

The Casimir effect corresponds to the force acting beetween two uncharged parallel condensor plates.
It is customarily attributed to the change in zero point energy of the electromagnetic vacuum ex-
tending between the plates with respect to the one of the vacuum contained in the same region in the
absence of plates. The zero point energy is supposed to result from the standard quantization of the
free electromagnetic field. This energy is not directly observable, but the force between the two plates
results from the change of the zero point energy contained between the plates when the latter are
moved apart from each other. The Casimir effect is generally considered as a “proof” of the reality
of the zero point energy. The dark energy, seemingly necessary to explain the observed accelerating
expansion of the Universe is sometimes advocated as another “proof” of the same reality. All these
aspects are shortly examined below.

2 The “usual” derivation of the Casimir effect

Let us consider an ideal condensor with infinite perfectly conducting plates in the x and y directions
separated by a distance d along the z direction as shown in Fig. 1.

The interaction energy between the two plates can be defined as the difference between the zero
point energies contained in the space between the planes, in the two respective configurations. This

Preprint submitted to Elsevier Science 13 July 2009

supposes that the field outside the cavity is not changed, which, of course, holds classically. For a
discussion of this point, as well as for corrections due the finiteness of the plates, see Ref. [1]. In
general, the zero point energy of the electromagnetic field is given
X ~ωk
Ecav = gs (k) , (1)
k 2

where the sum runs over the normal modes of the field, where the ωk ’s are the frequencies of these
modes and where gs (k) is the degeneracy of the mode k , due to polarization.

0 z

Fig. 1. I deal condensor with infinite extension in the x and y directions. The origin of the z axis is located
on the left plate.

Let us start with the case of the cavity. In order to identify the modes more easily, let us consider,
as asual, field configurations which are periodic in the x and y directions with “periods” Lx and Ly ,
respectively. The normal modes are determined by the boundary conditions on the surface of the
conductors. We remind that the tangential electric field and the normal component of the magnetic
field should vanish on these boundaries. These conditions are realized when the vector potential ψ~ is
given, for kz 6= 0, by

~ = ~ǫ eikx x eiky y sinkz z e−iωk t ,

ψ (2)

with
2π 2π
kx = nx , ky = ny , (3)
Lx Ly

2
where nx and ny are integer numbers and with
π
kz = nz , (4)
d

where nz is a positive integer, the vector ~ǫ is the polarisation vector and where ~k = (kx , ky , kz ). It
is perpendicular to the vector ~k : ~ǫ.~k = 0, as explained below. The quantity ψ
~ can be viewed as the
vector potential in the Coulomb gauge. The electric field is then proportional to the vector potential:
~ = iωk~ǫ eikx x eiky y sinkz z e−iωk t .
E (5)

where ~ǫ is the polarization vector. The magnetic field can be written as

~ = ~k × ~ǫ eikx x eiky y coskz z e−iωk t .
B (6)

The boundary conditions are satisfied as follows. The vanishing of the tangential electric field is
~ and the values of kz . The vanishing of the
guaranteed by the presence of the sine function in ψ
normal magnetic field requires

e~z .(~k × ~ǫ) = 0, (7)

which should hold together with

~k.~ǫ = 0, (8)

~ E=0.
the latter relation resulting from ∇. ~ The last two equations are explicitated as:

kx ǫy − ky ǫx = 0, (9)

kx ǫx + ky ǫy + kz ǫz = 0. (10)

For each value of ~k with kz 6= 0, there is an infinity of solutions and it is thus possible to select
two modes corresponding to two vectors ~ǫ satisfying the last two equations and perpendicular to
each other. The conditions are satisfied differently for the modes ~k with kz = 0. Form (2) is not
satisfactory, since ψ is then vanishing identically. However, the form
~ = ~ǫ eikx x eiky y e−iωk t ,
ψ (11)

where sin(kz z) has been replaced by cos(kz z) (equal to unity), is also a solution of the Laplace
equation in this case. The electric and magnetic fields are given by
~ = iωk~ǫ eikx x eiky y e−iωk t .
E (12)

3
and
~ = −~k × ~ǫ eikx x eiky y e−iωk t .
B (13)

The vanishing of the tangential electric field can now only be guaranteed by a normal ~ǫ vector, which
also guarantees the vanishing of the normal component of the magnetic field (13), since the vector ~k
has only tangential component. However, the vector ~ǫ does still have to fulfill Eqs. (9,10). For kz = 0,
the only solution is ~ǫ = ~ez : for these modes, there is only one possible polarisation (gs =1). These
kz = 0 modes are often forgotten in the literature (see for instance Ref. [2]) , leading to confusing
statements about regularisation procedures.

The energy of the cavity for the kz 6= 0 modes (indicated by the prime) is given by
+∞ +∞ ∞
" 2 2 2 #1/2
nx 2π ny 2π nz π
 
′
X X X
Ecav = ~c + + . (14)
nx =−∞ ny =−∞ nz =1 Lx Lx d

As usual, one replaces the summation on nx and ny by integration on continuous variables:

+∞ +∞ ∞
" 2 2 2 #1/2
nx 2π ny 2π nz π
Z Z  
′
X
Ecav = ~cLx Ly dnx dny + + , (15)
nz =1 Lx Lx d
−∞ −∞

which causes no difference when the limit Lx , Ly → ∞ is taken. Changing variables from nx , ny to
kx , ky , dividing the expression by Lx Ly and taking the limit of Lx and Ly tending to infinity, yields
for the energy per unit area:
′ ∞ +∞ +∞ " 2 #1/2
Ecav dkx dky 2 nπ
Z Z 
kx + ky2 +
X
= ~c . (16)
S n=1 2π 2π d
−∞ −∞

Finally, one integrates over the angle of the wave vector in the x − y plane, introduces the auxiliary
variable u = ((kx2 + ky2 )d2 )/π 2 = k⊥
2 2
d /π 2 and obtains
∞ Z ∞
′
Ecav ~cπ 2 X 
2 1/2

= du u + n . (17)
S 4d3 n=1
0

We still have to add the contribution of the kz = 0 (or n = 0) modes. One has:
∞ ∞
 
∞ Z
Ecav ~cπ 2  X 2 1/2 1 √
  Z
= 3
du u + n + du u . (18)
S 4d n=1 2
0 0

This expression is divergent, as is the similar expression for the energy of the free field. It may be
hoped that the difference between the two expressions is finite.

4
We need the value of the energy of the free field in the volume of the cavity. In general, the energy
per unit volume is given by 1

Ef ree d3~k
Z
= ~c k. (19)
V (2π)3

It is advantageous to rewrite this expression as

Ef ree ~c Z 2~ Z q
2 2
= d k ⊥ dk z k⊥ + kz , (20)
V (2π)3

or as
∞ +∞
Ef ree ~c Z Z q
2
= k ⊥ dk ⊥ dk z k⊥ + kz2 . (21)
V (2π)2
0 −∞

Taking account of the fact that the integrand is an even function of kz and introducing the auxiliary
2 2
variables u = k⊥ d /π 2 and x = kz /d lead to
Z∞ Z∞ √
Ef ree ~cπ 2
= dx du u + x2 . (22)
V 4d4
0 0

The energy of the free field in the volume of the cavity is thus obtained by multiplying this expression
by the volume V (equal to Sd). One then obtains for the change in the zero point energy per unit
surface:
Z∞ ∞ Z ∞ ∞ ∞
∆E Ecav Ef ree ~cπ 2 1 √ 2 1/2
1/2
  Z Z 
− dx du u + x2
X
= − = { du u + du u + n }. (23)
S S S 4d3 2 n=1 0
0 0 0

All terms in the rhs are divergent. We will come to this problem soon. It is remarkable that this
expression involves the difference between the integral from zero to infinity of the function f (x) =
2 1/2
R∞
0 du (u + x ) and the sum of the values of this function on the positive integers. There is a famous
theorem by Euler and McLaurin connecting these quantities, in general. It states that:
∞ Z∞
1 1 ′ 1
[f (∞) − f ′ (0)] − [f ′′′ (∞) − f ′′′ (0)] + · · · (24)
X
f (n) = f (x)dx − [f (0) + f (∞)] +
n=1 2 12 720
0

where the dots indicates similar terms for higher order odd derivatives. The coefficients in front of
the brackets are related to the Bernoulli numbers Bi : they are equal to −B2k /(2k)! for the term
involving the derivatives of order 2k − 1. See Ref. [3].
1 The ~k = 0 mode does not pose any worry, since its contribution is vanishing.

5
We can write the function f (x) mentioned above as
Z∞ Z∞ √
2 1/2
 
f (x) = du u + x = dt t. (25)
0 x2

Formally, considering the dependence upon x through the lower bound of the integral only, one has:

f ′ (x) = −2x2 , f ′′(x) = −4x, f ′′′ (x) = −4 (26)

and all higher derivatives are vanishing. So, retaining the single nondivergent term (which corresponds
to f ′′′ (0)), one finally obtains:

∆E ~cπ 2
=− . (27)
S 720d3

This is the expression of the Casimir effect, which looks universal and which depends only upon the
two fundamental constants ~ and c and on the distance d.

Let us comment on the divergence problems first. Of course, expression (23) and the Euler-MacLaurin
theorem apply when the quantities are convergent. However, one can make the final result meaningful
by regularizing the integral and the sum. The regularisation at infinity is not a problem. It is easy to
introduce a suitable integration factor. For instance, it is easy to see that all terms at infinity vanish
owing to the substitution
Z∞  1/2 2 )α
f (x) → du u + x2 e−(u+x , (28)
0

The terms corresponding to higher order derivatives (at x = 0 as well as at x = ∞) are finite and
vanish as α → 0. Similarly, the first and third derivatives at x = 0 are incremented by quantities
that vanishes as α → 0. Actually, the regularisation can be achieved by any cut-off function which
decreases sufficiently rapidly at large x (non necessarily exponentially), which goes to unity as x goes
to zero with vanishing derivatives, like the functions g(x) = 1 − exp(−a/x).

3 Which force?

The force (per unit surface) acting between the plates is given by

F ~cπ 2
=− . (29)
S 240d4

6
The negative sign corresponds to an attractive force. It is a tiny force. For instance, at d=1 µm, it
amounts to F/S= 4×10−4 N/m2 . Of course, due to the fourth power, it increases very rapidly as the
distance decreases. At d=1 nm, the force reaches F/S= 4×108 N/m2 .

Needless to say that the experimental verification of the Casimir effect has taken quite a long time.
Among the unsuccessful trials, one should mention the experiment by Sparnaay [4], which although
unsuccessful, has nevertheless identified the main difficulties: a perfect parallelism of the plates, a
lack of impurities (which may scatter the normal modes) and the elimination of the residual charges.
Let us also mention the experiment of Derjaguin et al [5], who were the first to obtain a meaningful
result, verifying the predictions at the 60% level, before the experiments by Lamoreaux [6] and Ederth
[7] who verified the theoretical value with an accuracy of ∼1%.

4 The Casimir force and the van der Waals effect

4.1 Introduction

The Casimir force may be viewed as a quantum interaction between two neutral objects. Of course,
the conducting properties of these objects should be taken into account at some point. But for the
moment, let us consider the Casimir force as the force acting between macroscopic neutral objects
and address the question whether there is some relationship with the force acting in another system
of this kind, namely the system of two neutral atoms. We will examine this question in a bit historical
perspective, which helps to understand the relationship.

4.2 The van der Waals force in the simplest approach

The van der Waals interaction has been calculated microscopically for the first time by London [8].
The hamiltonian of the system can be written as

H = H1 + H2 + H ′ ,
Z1 Z2 Z1 XZ2
′ Z1 Z2 e2 X Z2 e2 X Z1 e2 X e2
H = − − + . (30)
~1 − R~2 | i=1 |R~1 + r~i − R~2 | j=1 |R~2 + r~j − R~1 | i=1 j=1 |R
|R ~2 + r~j − R~1 − r~i |

In this equation, Z1 and Z2 are the charge numbers of the nuclei and r~i , r~j are the coordinates of the
electrons with respect to the position of the respective nuclei. Expanding H ′ up to second order in
~1 − R~2 | is large
the electron coordinates r~i , r~j (which is presumably sufficient if the distance r = |R
in comparison with the atomic sizes), one has
Z1 Z2
′ Z2 e ~ ~
X Z1 e ~ ~
X
H = −2 (R1 − R2 ). e~
ri − 2 (R2 − R1 ). e~
rj . (31)
|R~1 − R~2 |3 i=1 |R~1 − R~2 |3 j=1

7
Considering H ′ as a perturbation, the change in energy of the ground state can be calculated by
standard perturbation theory. The first order contribution vanishes. The second order contribution
writes:
6e4 X X |hk | i ~ri .~n | 0i|2 |hl | j ~rj .~n | 0i|2
P P
(2)
∆E =− 6 . (32)
r k6=0 l6=0 E1k − E10 + E2k − E20

In this equation, k (l) labels the excited states of the first (second) atom, |0i is the ground state of
the atoms (we avoided to put an indice recalling which atom is concerned, since there is no risk of
confusion) and ~n is the unit vector along the line joining the two nuclei (the direction is irrelevant).
The sums run, in principle, over all the excited states, but in practice only on those which are
connected to the ground state, by off-diagonal matrix elements of the dipole moments d~1 = ~ri or
P

d~2 = ~rj . For atoms with J = 0 ground states, i.e. with spherical shapes, the sum in Eq. 32 is
P

limited to J = 1 states, but involves a summation over the magnetic quantum numbers. Using the
Wigner-Eckart theorem to relate the matrix elements of different magnetic quantum numbers, it is
easy to see that that the quantity ∆E (2) does not depend upon the orientation of the vector ~n, as
intuitively expected.

The physical meaning of Eq. (32) is rather clear. Classically, two neutral objects with spherical
symmetry, even if they are locally charged have no Coulomb interactions. All their multipole moments
are vanishing. Quantum mechanically, spherical neutral atoms, have zero electric dipole moments only
on the average. They are fluctuating. There is a non vanishing probability for having the two atoms
with non zero dipole moments and therefore experiencing a Coulomb interaction. The van der Waals
force is thus a purely quantum force originating from quantum fluctuations.

It may be of interest to make two remarks. First, Eq. (32) is not valid if r is of the order of the size
of the atoms. At short distances, the interaction should be repulsive, due to the Pauli principle: the
latter forbids to put simply electrons at the top of each other; this is only possible if the electrons of
one atom are put at unoccupied orbits of the other, which requires a strong increase of the kinetic
energy. The repulsive nature of the atom-atom interaction at short distance is often embodied by
Lennard-Jones potentials. Second, it may be worthwhile to notice that the expression in Eq. (32) is
almost but not exactly proportional to the product of the electric polarisabilities of the atoms. The
electric polarisability is defined as the the ratio between the induced electric dipole acquired by an
atom in a static electric field (considered as uniform for simplicity) and the magnitude of this electric
field. It is given in second order by

|hk | e~ri .~n | 0i|2

P
i
X
α= . (33)
k6=0 Ek − E0

For a matter made of these atoms, the dielectric constant ε is given, in the dilute limit, as

ε = 1 + 4παnat , (34)

where nat is the density of atoms.

8
4.3 The van der Waals force and retardation effects

When he was working at the Philips company in Eindhoven, Casimir got interested into the behaviour
of the van der Waals interaction at large distances. Two colleagues of him, Verwey and Overbeek,
were studying experimentally colloidal suspensions. It seems that a simple model based on the van
der Waals force successfully reproduced their observations, but failed for dilute suspensions. They
interpreted their observation as due to a weakening of the van der Waals force at large distance [9–11].
They were thinking that retardation effects were the cause of this weakening. If the van der Waals
interaction is interpreted as due to the interaction between fluctuating dipoles, the fluctuation of one
dipole takes some time before influencing the other atom when the interdistance is large enough and
vice-versa. They approached Casimir and asked him whether he could calculate this effect. A little
bit later, the answer was given in paper by Casimir and Polder [12]. We are not going to enter in the
details of this complicated calculation, but we will give the general ideas and the results.

In Eq. (32), only the static (Coulomb) interactions are introduced. How to cope with retardation
effects in Quantum Mechanics? Retardation effects are linked with the perturbations of the radiation
field which propagate at finite speed. One has thus to introduce this radiation field. This is usually
done by introducing a time-dependent vector potential in the hamiltonian. Adopting the Coulomb
gauge and the minimum substitution principle, this is equivalent to replace the momentum of the
~ ri , t). With such a prescription, one has to add, along with H’, a second
electron ~pi by ~pi + ec A(~
perturbation of the form:

e X~ e2 X ~ 2
H ′′ = A(~ri , t).~pi + A (~ri , t). (35)
mc i 2mc2 i

The procedure is fairly standard. The system of the two atoms with a relative distance d is enclosed
in a cubic box with perfectly conducting walls. The vector potential is written as an expansion on
the normal modes

~ r , t) = ~ ~ (~r),
(a~k,s e−iωt + a~†k,s eiωt )X
XX
A(~ k,s (36)
~k s

where X ~ ~ is the vector field characteristic of the mode ~k, s, duly normalized. The operators a and
k,s
a† are the usual destruction and creation operators. The total hamiltonian is written as:

H = H1 + H2 + Hrad + H ′ + H ′′ = H0 + H ′ + H ′′ . (37)

where Hrad is the free radiation hamiltonian. The change in the ground state energy of the total
system, due to the perturbation H ′ and H ′′ , is then calculated to second order. Note that to be
consistent, the calculation should include second order terms in the first part of H ′′ (linear in a and
a† ) and first order terms in the second part of this operator as it is already quadratic in a and a† .
Finally, the size of the box is extended to infinity, while keeping the distance d fixed. Needless to

9
say that the calculation is rather cumbersome. For detail, we refer to the original paper. We simply
quote the results.

For small r (actually smaller than the absolute value of the non-diagonal matrix elements of the
dipole operator), the London result (Eq. 32) is recovered. For large r (in principle larger than the
above-mentioned quantities), the following simple results is obtained:

23~c
∆E (2) = − α1 α2 . (38)
4πr 7

The van der Waals interaction is weakening as the distance increases and factorises in the polaris-
abilites of the atoms.

In the same paper, Casimir and Polder investigated also the interaction of an atom with a conducting
wall. The principle is the same: put the atom in the box at a fixed distance d from a wall and let
the size of the box become infinite while keeping a wall fixed. In this case, the hamiltonian H ′ is the
Coulomb interaction between the atom and the wall, which is taken as the interaction of the dipole
moment of the atom and its image. The dipole moment is then considered as the corresponding
quantum operator. Note that the Coulomb atom-wall is neglected in the case of two atoms, since the
walls are eventually removed to infinity. It is interesting to quote the results. For small distances, the
change of energy is given by:

(2) 3 X
e~ri .~n | 0i|2 ,
X
∆Eatom−wall = − 3
|hk | (39)
8d k i

where ~n is the unit vector perpendicular to the plane. For large distances, one has

(2) ~c 3α1
∆Eatom−wall = − , (40)
8π d4

where α1 is the polarisability of the atom.

Casimir was intrigued by the simplicity of the results, especially the one of Eq. 38 and was wandering
whether they can be more general. After all, these results were derived using the standard apparatus
of perturbation theory (to second order in the fine-structure constant α, see later). He once discussed
these results with Niels Bohr, who is said to have replied [13]: “Why don’t you calculate the effect by
evaluating the difference of zero point energies in the electromagnetic field?”. Of course this requires
to calculate the normal modes in the presence of the atoms, which is very hard. Casimir realized that
the calculation could be more easily performed for the case of the cavity as it is done in Section 2
and published his result in Ref. [14].

10
5 The nature of the Casimir force

5.1 Introduction

Although the existence and magnitude of the Casimir effect is now well established, there is still some
controversy concerning its nature and its interpretation. The Casimir effect looks universal. Formula
(27) indeed solely depends upon the the constants ~ and c and upon the interdistance d. It is therefore
considered as a property of the electromagnetic vacuum, modified by the presence of the condensor
plates. On the other hand, it is tempting to interpret the Casimir effect as a generalized van der
Waals interaction between two gigantic “molecules”, the conducting planes. In this perspective, the
Casimir effect is reduced to an ordinary (though quantal) electromagnetic effect. It is then surprising
that this effect is not dependent upon the fine structure constant α. Actually, it can be shown that the
independence upon α results from the implicit hypothesis of perfectly conducting planes. When this
hypothesis is released, correcting terms in α should be added. The result (27) appears to be correct in
the the limit of very large values of α. In the following we will give simple arguments supporting this
assertion. We will also discuss the relation between the Casimir effect and the quantum fluctuations
of the electromagnetic vacuum. Since this question is still under debate, we will limit ourselves to
general considerations.

5.2 The dependence of the Casimir effect on the fine structure constant

Actual metals are not perfectly conducting. They are characterized basically by two quantities: the
plasma frequency ωpl and the skin depth δ. For frequencies above ωpl , the conductivity basically
goes to zero. The quantity δ measures the distance up to which electromagnetic waves penetrate the
metal. A perfect conductor is characterized by infinite ωpl and δ = 0. In actual metals, ωpl and 1/δ
depend upon the fine structure constant α and vanish when α → 0. We turn to the simplest model
for real metals, namely the Drude model, to describe qualitatively what is happening. We closely
follow here Ref. [15]. Basically, in the Drude model, electrons are moving independently under the
influence of the electric field and they are subject to a friction force. Let E = E0 e−iωt be the applied
electric field. The Newton equation of motion for the electron can be written as:

d2 x dx
me 2
= −eE0 e−iωt − γ . (41)
dt dt

where me is the electron mass and γ is the friction parameter. The solution is an oscillatory function
of time x(t) = x0 e−iωt , with

eE0
x0 = , (42)
me ω(ω + iγ)

11
where we have introduce the reduced friction parameter γ = γ/me . It is then easy to calculate the
induced current (j = −endx/dt). The result gives readily the conductivity (σ = j/E) under the form

e2 n 1
σ= , (43)
me γ − iω

where n is the electron density. The plasma frequency is given by

2 4πe2 n
ωpl = . (44)
me

The skin depth, which is defined by

2πω|σ|
δ −2 = . (45)
c2

in general, becomes in the Drude model

c
δ= 1 . (46)
2
1 √ωωpl 2

2 γ 2 +ω 2

In this model, indefinitely increasing ωpl automatically implies δ → 0. In practice, typical frequencies
of interest are larger than γ, and δ becomes
c
δ≈ √ . (47)
ωpl 2

The frequencies that are relevant for the Casimir effect are those with a frequency smaller than c/d.
The perfect conductor approximation requires therefore that c/d ≪ ωpl . Combining this relation with
Eq. 44 gives the following condition:
me c
α≫ . (48)
4π~nd2

For typical cases (copper plates separated by a micrometer), the rhs is of the order of 10−5 . Condition
(48) is comfortably satisfied by the physical value of α. The standard Casimir result can then be
regarded as the α → ∞ limit of the true result which is dependent on the nature of the metal. For
large α, one expects corrections to the Casimir result which could be put in series of negative powers
of α. This result may be obtained very roughly by saying that for real metals, the limits of the cavity
become somewhat transparent to the electromagnetic field and that the effective width of the cavity
becomes d + 2δ. One thus expects, instead of the relation (27)

~cπ 2 ~cπ 2
!
∆E 6δ
≈− ≈ 1 − + ... . (49)
S 720(d + 2δ)3 720d3 d

12
In this equation, the dots indicate higher order
√ powers in 2δ/d. Owing to Eqs. 47,44, it is easily seen
that the latter ratio is proportional to 1/ α. The corrections thus disappear as α → ∞. It is also
interesting to verify that the second term in the parenthesis of Eq. 49 becomes negligible compared
to unity when condition (48) is fulfilled.

It is also interesting to look at the α → 0 limit. This limit is a little bit tricky as the typical size of
atoms, the Bohr radius ~2 /me e2 , scales as 1/α. Therefore, n scales as α3 , ωpl scales as α2 and δ goes
as 1/α2 . At very low α, the plates become transparent to the radiation and the Casimir effect goes
away as α → 0. At low α, the Casimir effect is expected to be put in a series of increasing positive
powers of α.

As all ordinary electromagnetic effects, the Casimir effect goes away when the fine structure constant
goes to zero. The distinctive feature of the Casimir effect is that it reaches a finite value as α → ∞.

5.3 The Casimir effect: vacuum property or interaction between neutral objects?

The Casimir effect is often pointed as an evidence of the reality of quantum fluctuations of fields in
vacuum. Just to quote a typical example, Weinberg in his introduction of the cosmological constant
problem states [16]:

Perhaps surprinsingly, it was a long time before particle physicists began seriously to worry about
[quantum zero point fluctuation contribution to Λ] despite the demonstration in the Casimir effect
of the reality of zero-point energies.

There are more recent quotations of this type. In his book “Particle Astrophysics”, Perkins says [17]:

That this concept [the vacuum energy] is not a figment of the physicist’s imagination was already
demonstrated many years ago, when Casimir predicted that by modifying the boundary condi-
tions on the vacuum state, the change in the vacuum energy would lead to a measurable force,
subsequently detected and measured by...

This kind of statements should be appraised after a close examination of the meaning of the expres-
sions “reality” and “quantum fluctuations”. Let us start with the second one. The vacuum can be
considered as the quantum ground state of a field, say the electromagnetic field in the case of our
discussion. There is little doubt that there are quantum fluctuations of observables associated to this
field in the ground state, as it is for any observable which does not commute with the hamiltonian.
The simplest observables are the electric and magnetic fields themselves. The expression “quantum
fluctuations” is also used to denote the zero point energy of the vacuum which is interpreted as the
energy “generated” by the quantum fluctuations. Had the electromagnetic field been vanishing with
certainty, would it be natural to expect a vanishing energy. When one relates the Casimir effect to
the “quantum fluctuations”, it is to the second meaning of these words that one refers.

When boundaries are imposed to the electromagnetic field, the latter is changed. The question arises
whether the ground state (the vacuum) of the electromagnetic field is changed. Physicists have been

13
reluctant for a long time to admit that the energy of the vacuum is changed (advocating that the
zero point energy is infinite and thus that its physical meaning is suspicious). The experimental
measurements of the Casimir effect have given support to the idea that the zero point energy is
perhaps unphysical, because it cannot be measured directly, but its variations when the geometry is
changed are physical since they are observed. Before discussing this point, let us mention that nobody
questions the change in the fluctating properties of the electromagnetic field when boundaries are
introduced. We will come to this question later.

Let us examine the reality of the change in the zero point energy as revealed by the Casimir effect.
The experiments are realized with condenser plates which, even in the limit of perfect conductors,
are not merely a “mathematical” device serving to confine the electromagnetic field in a restricted
region of space. They are composed of atoms or molecules which interact with the electromagnetic
field. The force which is measured is in fact the force between the material plates. In some sense,
it can be viewed as a van der Waals interaction between two gigantic molecules. It is the point of
view adopted by many physicists, who consider that the usual calculation, based on the change in
~ω, is heuristic [15]. In other words, it is an accident that it gives the expression of the force
P

between two conductors. A similar example is provided by the energy of a smooth charge in classical
R 3 R 3 ′
electrostatics which is given by W = 1/2 d ~r d ~r ρ(~r)ρ(~r′ )/|~r − ~r′ | or, also, by the energy of the
electric field W = 1/(8π) d3~r|E(~
R
~ r )|2 . This second expression cannot be viewed as an evidence of
the “reality” of the electric field but as an alternative expression of the self-interaction of the charges
which, heuristically, gives the correct magnitude of this self-interaction energy. The reality of the field
and its extension outside of the sources cannot be proven by the action on a test charge as it is often
stated in elementary courses. By looking at the effect of a test charge, one probes the interaction
between the original source and the test charge. Bringing a test charge “changes the nature of the
problem”. The reality of the electromagnetic field exists, but is revealed by other kinds of phenomena
like the Hertz experiment or the pair creation in a Coulomb field.

Let us now examine the relation between the van der Waals effect and the Casimir effect at the light
of fluctuations of the vacuum, in the sense of fluctuations of operators in the vacuum. The van der
Waals interaction between two atoms is the change of the energy in the system of the two atoms
due to presence of the Coulomb interaction, which couples to the fluctuating dipoles of the atoms
(this is translated as the interaction to the second-order perturbation term). The Casimir-Polder
interaction is the change of the energy of the system made of the two atoms and the ground state
of the electromagnetic field, due to the presence of the Coulomb interaction, which couples to the
fluctuating dipoles of the atoms as before, but also due to the interaction between the electrons and
the electromagnetic field, which couples to the fluctuations of the latter in the vacuum. When one of
the atoms is replaced by a conducting plate, the fluctuations of this “atom” somehow disappear, but
the interaction is given by the change in energy of the whole system (atom, wall and electromagnetic
field), due to the same causes. The change in dimension of the system is reflected through the change
in the power in the dependence of the effect upon the distance. When the second atom is replaced
by the second plate, the interaction is solely due to the interaction of the plates with the fluctuating
electromagnetic field.

It is interesting to note that the interaction between two conducting plates can be constructed without
reference to zero point energy. First, we want to mention the method invented by Lifshitz [18,19] to

14
calculate the interaction between dielectrics. The starting point is the quantum fluctuations of the
electromagnetic field in large bodies. It is first argued that the fluctuations of the field are linked to
the fluctuations of the polarisation density, that the fluctuations at different points (understood as
involving scale larger than the atomic size) are uncorrelated and that the mean square of the fields at
any given point is fixed by the change of energy implied by the appearance of a dielectric constant.
The Maxwell stress tensor can be calculated and forces may be derived by taking derivatives. The
formalism is cumbersome and has not been published totally (which has hampered interest in this
kind of approach). Let us just quote some results. Lifshitz obtained an explicit expression for the
force between two infinite dielectric bodies of dielectric constant ε with plane surfaces facing each
other at a distance d (expression (90.1), p. 369 of Ref. [20]). From this expression, several limits
can be obtained. The limit ε → 1 corresponds to the dilute limit and the interaction between two
molecules can be obtained. The perfect conductor limit is obtained as ε → ∞ (this corresponds to
the vanishing of the electric field inside the bodies). All the results of Section 3 and the Casimir
formula can be obtained this way.

Let us also mention that the Quantum Field Theory can be formulated without any reference to
zero point energy. For instance, the Casimir effect may be calculated (perturbatively) in terms of
Feynmann diagrams with external legs, i.e. in terms of S-matrix elements making no reference to the
vacuum. We refer to Ref. [15] for more information.

In conclusion there is a strong division between physicists regarding the interpretation of the Casimir
effect. For many of them, the latter is a manisfestation of the vacuum energy. For many others, the
Casimir effect is the interaction between two large polarisable bodies. It does not tell upon the
quantum fluctuations more than any one-loop effect in quantum electrodynamics, like the vacuum
polarisation of the Lamb shift.

6 The Casimir effect in Cosmology

It has been suggested that the Casimir effect or rather the vacuum energy could account for dark
energy. The first suggestion dates to Einstein who introduced a cosmological constant in his funda-
mental equations of general relativity coupling the structure of space-time and the energy content:
1
Rµν − gµν R = 8πG(Tµν + Ev gµν ). (50)
2

Einstein introduced the last term in the rhs, or equivalently Λgµν , with the cosmological constant
Λ = 8πGEv , in order to manage a static solution. After the discovery of the expansion of the
Universe, he dropped this term, that he considered as “the greatest mistake in [his] life”. Later, with
the development of quantum field theory, the possibility of the existence of a vacuum energy was
taken seriously (it seems that Zeldovich was the first person to formulate such a possibility [21]) ,
especially after the experimental verification of the Casimir effect. It was even taken more seriously
when it was realized that the expansion of the Universe seems to be accelerating [22,23]. Recent
observations require a value of Λ = (2.14 ± 0.13 × 10−3 eV )4 at the present time [24]. The name “dark

15
energy” has been dubbed for this energy. The zero point energy of the electromagnetic field has been
proposed as a candidate for the dark energy. We have given serious reservations on the reality of the
zero point energy, but this suggestion meets difficulty beyond these reservations. First of all, the zero
point energy density is in principle infinite. Of course, one may admit that the contribution of the
very high fequencies should be cut somewhere, rendering the energy density finite. The natural cut
should come from gravity and can be taken as the Planck scale. One then has
k<k
d3~k
Z cut
4
ρv = ~c k = ~cπkcut . (51)
(2π)3

If kcut is taken as the inverse of Planck length λP L = 1.2 × 1019 GeV, one obtains a vacuum energy
of the order of 10121 GeV fm−3 . This should be compared to the critical energy density ρc ≈ 5 GeV
fm−3 and the part of about 75% taken by the dark energy. The supposed electromagnetic vacuum
energy is thus enormously too large and there is no clear mechanism to reduce it. Furthermore, one
should add in principle the contribution of the vacuum energy for the other fundamental fields. This
has led to the crisis of the cosmological constant and to a serious questioning about the reality of
the vacuum energy of the fields. Furthermore, there are plenty of condensates in the standard model
which also contribute to dark energy in principle. The conclusion is that the (perturbative) vacuum
energy (of the fields) has probably no real physical meaning or at least that our understanding of its
properties, especially concerning its coupling to gravity, has to be clarified.

7 Conclusion

This short review aimed at presenting the main current ideas about the Casimir effect, its interpreta-
tion and its relevance to Cosmology. It does not reflect the increasing activity linked with Casimir-like
effects in the nanoworld. See Ref. [25] for an introduction.

8 Appendix. Regularisation of Eq.23

As an example, we chose the regularisation provided by substitution (28) or

Z∞ √
f (x) = dt te−αt . (52)
x2

The first three derivatives are given by

2
f ′ (x) = −2x2 e−αx , (53)

16
 2
f ′′ (x) = (−4x + 4αx3 )e−αx (54)

and
2
f ′′′ (x) = (−4 + 20αx2 − 8α2 x4 )e−αx , (55)

respectively. All derivatives at infinity are vanishing. It is also easy to verify that f ′ (x) vanishes at
x = 0 and that f ′′′ (0) = −4 and that higher order derivatives at x = 0 are either vanishing or are
proportional to positive powers of α. The first term in the curly braket of Eq. (14) being equal to
f (0), formula (27) results directly from the application of the Euler-McLaurin formula to the function
(52), at the limit of vanishing α.

17
References

[1] C. Itzykson and J. B. Zuber, “Quantum Field Theory”, McGraw-Hill, New York, (1985)ch.3.

[2] T. J. Torres, http://web.mit.edu/ttorres/Public/The”Casimir”Effect.pdf

[3] M. Abramowitz and I. A. Stegun, “Handbook of Mathematical Functions”, Dover, New York, (1970)
886.

[4] M. J. Sparnaay, Physica 24 (1958) 751.

[5] B. V. Derjaguin, N. V. Churaev and V. M. Muller, “Surface Forces”, Plenum, New York, (1987) 886.

[6] S. K. Lamoreaux, Phys. Rev. Lett. 78 (1997) 5.

[7] T. Ederth, Phys. Rev. A 62 (2000) 062104.

[8] F. London, Trans. Faraday Soc. 33 (1937) 19.

[9] E. J. W. Verwey and J. T. G. Overbeek, Trans. Faraday Soc. 42B (1946) 117.

[10] E. J. W. Verwey, J. Phys. and Colloid Chem. 51 (1947) 631.

[11] E. J. W. Verwey, J. T. G. Overbeek and K. van Nes, “Theory of the Stability of Lyophobic Colloids”,
Elsevier, Amsterdam, (1948).

[12] H. B. G. Casimir and D. Polder, Phys. Rev. 73 (1948) 360.

[13] P. W. Milonni, “The Quantum Vacuum”, Academic Press, New York, (1994).

[14] H. B. G. Casimir, Kon. Ned. Akad. Wetensch. Proc. 51 (1948) 793.

[15] R. L. Jaffe, Phys. Rev. D72 (2005) 021301.

[16] S. Weinberg, Rev. Mod. Phys. 61 (1989) 1.

[17] D. Perkins, “Particle Astrophysics”, Oxford University Press, Oxford, (2003).

[18] E. M. Lifshitz, Zhurnal éksperimental’noǐ i teoreticheskoǐ fiziki 29 (1955) 9

[19] E. M. Lifshitz, Soviet Journal JETP 2 (1956) 73.

[20] L. D. Landau and E. M. Lifshitz, “Electrodynamics of continuous media”, Pergamon Press, London,
(1960).

[21] Y. B. Zeldovich, JETP Lett. 6 (1967) 316.

[22] N. Straumann, [arXiv:astro-ph/0203330].

[23] N. Straumann, “Dark Energy”, Lect. Notes Phys.721 (2007) 327.

[24] M. Tegmark et al [SDSS Collaboration], Phys. Rev. D69 (2004) 103501 [arXiv: astro-ph/0310723].

[25] P. W. Milonni, http://www.phys.nyu.edu/LarrySpruch/Milonni.pdf

18