Light

What is light?
Gordon McCabe
September 9, 2020
Abstract
The purpose of this paper is to answer the question, ‘What is light?’,
from a mathematical and foundational perspective. The paper begins
by exploring the relevance of space-time symmetries, and the nature of
polarization, before a detailed exposition of the quantized radiation field,
and the difficulties created by gauge freedom.
Attention then turns to the interaction of light with matter, beginning
with the coupled electromagnetic field, then progressing to the representa-
tion of scattering and virtual particles in quantum electrodynamics. This
provokes an analysis of the Coulomb electrostatic field, and the question
of whether longitudinal and scalar photons exist. It is argued that the lon-
gitudinal component of the electric field is associated with the state-space
of a charged matter field, not the Fock space of the free electromagnetic
field, hence the presence of electrostatic fields is consistent with the vac-
uum state of the free electromagnetic field. It is also argued that scalar
and longitudinal photons do indeed exist as links in the spacelike networks
into which the Coulomb interaction can be decomposed.
Consideration of the stimulated emission of light leads to a general
exposition and analysis of the ‘coherent states’ of the quantized radiation
field. As a by-product of this, a novel explanation is proposed for why
there is something classical rather than nothing classical. An attempt
is made to develop this into a fully-fledged universe creation scenario.
The role of fermions and the gravitational degrees of freedom in such a
scenario are discussed, and a comparison is drawn with the inflationary
cosmological scenario.
The role of coherent states in our concept of the classical world is then
critically analysed. The notion that the classical states of the radiation
field are emergent from the quantum states is rejected. In particular, it is
argued that the classical states of light do not emerge in the limit where
there are large numbers of photons, and it is pointed out that the putative
emergent classical states fail the test of reference-frame independence.
The paper concludes by expounding the implications of the nature of
light for ‘decoherence’, a way of trying to reconcile quantum theory with
the apparent nature of the macroscopic world.
1
Contents
1 Introduction 3
2 What is light? 4
2.1 Classical electromagnetic waves . . . . . . . . . . . . . . . . . . . 4
2.2 Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Quantum Field Theory 11

3.1 Fock space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 The quantum vacuum . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3 The quantized radiation field . . . . . . . . . . . . . . . . . . . . 18
4 Interactions between light and matter 22

4.1 The coupled electromagnetic field . . . . . . . . . . . . . . . . . . 22
4.2 Scattering, Feynman diagrams and virtual particles . . . . . . . . 24
4.3 The interacting field vacuum . . . . . . . . . . . . . . . . . . . . 28
4.4 Do longitudinal photons exist? . . . . . . . . . . . . . . . . . . . 31
4.5 Spontaneous and stimulated emission of light . . . . . . . . . . . 37
5 Coherent states 39
6 Why is there something classical rather than nothing classical? 46

6.1 Coherent states and the vacuum . . . . . . . . . . . . . . . . . . 46
6.2 Fermions, Supersymmetry and Grand Unified Theories . . . . . . 47
6.3 Gravity and gravitons . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4 Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7 Coherent states and the classical world 56

7.1 Are coherent states classical? . . . . . . . . . . . . . . . . . . . . 56
7.2 Coherent scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 64
8 Decoherence 65
9 Conclusions 71
2
1 Introduction
A plurality of mathematical structures exist for the representation of light: it is
represented as bundles, pencils and beams of 1-dimensional rays in geometrical
optics; as a tensor field satisfying a wave-equation in Maxwellian electromag-
netism; and as the elements of a photonic Fock space in quantum field theory.
All of which begs the question, ‘What is light?’ The purpose of this paper is to
answer that question from a mathematical and foundational perspective.
The second section begins with a concise exposition of light as a classical
electromagnetic wave. Noting that such waves can be decomposed into differ-
ent modes and polarizations, prompts the question ‘What is polarization?’ To
answer this, we digress to consider light as a representation of the space-time
symmetry group. Equipped with this perspective, we then consider the detailed
mathematical representation of polarization.
The third section of the paper begins with a brief recapitulation of the
mathematical structure of Fock space and the quantum vacuum. In particular,
the isomorphism between Fock space and the space of wavefunctionals on the
one-particle space is made explicit. The remainder of the section is devoted to
the quantized radiation field, and the problems created by gauge freedom.
The fourth section considers how interactions between light and matter are
represented, beginning with the coupled electromagnetic field in the classical
theory, and progressing to the representation of interacting fields in quantum
field theory. This provokes an analysis of the Coulomb electrostatic field, and
the question of whether longitudinal and scalar photons exist. As a special
case of the interaction between light and matter, the spontaneous emission of
light is explained. This is then contrasted with the production of laser light by
stimulated emission.
Laser light is represented mathematically by coherent states in photonic
Fock space, and the fifth section is devoted to an exposition and analysis of
these privileged states of the radiation field.
This analysis leads to a number of important philosophical points. Section
six is devoted to the first of these, a possible explanation for why there is
something classical rather than nothing classical. An attempt is made to develop
this into a fully-fledged universe creation scenario. The role of fermions and
the gravitational degrees of freedom in such a scenario are discussed, and a
comparison is drawn with the inflationary cosmological scenario.
The seventh section provides a general discussion of the implications of co-
herent states for the concept of classicality. A number of the common beliefs, in
particular the notion that the classical world emerges from the quantum world,
are shown to be false.
The paper concludes with a section on the implications of the nature of light
for ‘decoherence’, a way of trying to reconcile quantum theory with the apparent
nature of the macroscopic world.
3
2 What is light?
2.1 Classical electromagnetic waves
According to classical electromagnetism, light is a solution of the free-field
Maxwell equations. Under a Lorentz choice of gauge, the electromagnetic po-
tential is mapped to a real vector potential A satisfying ∇ · A = 0. Given a real
vector potential A, the electromagnetic field strength is F = dA. The Maxwell
equations for a vector potential obtained with a Lorentz choice of gauge are
simply:
∇ · dA = 0, ∇ · A = 0 .
These equations are equivalent to, (Derdzinski 2002, Appendix 48):
A = 0, ∇·A=0.
The first equation here is the wave equation on Minkowski space-time M,
∂2 ∂2 ∂2 ∂2
( 2 − 2 − 2 − )A = A = (Aµ )dxµ = 0 ,
∂x0 ∂x1 ∂x2 ∂x23
where is the d’Alembertian.
Such solutions can be subjected to a two-fold decomposition: (i) a Fourier
decomposition into monochromatic modes of different wavelengths; and (ii) a
decomposition of monochromatic light into different polarizations.
Both of these decompositions were part of classical optics, theoretically and
in terms of the historical experiments which explored their nature. Yet whilst
the Fourier decomposition is clearly classical, the polarization decomposition
introduces a quantum-mechanical substructure into classical theory. Moreover,
whilst numerous physicists are fond of claiming that there is no analogue of
quantum spin in classical physics, the case of polarization provides a counter-
example to such claims. To gain a better understanding of what polarization is,
the next subsection considers the role of space-time symmetries.
2.2 Symmetry
Wigner used the method of induced group representation to obtain an irre-
ducible unitary Hilbert space representation of the universal cover of the local
space-time symmetry group for every type of elementary particle. The largest
group of space-time symmetries satisfied by all the elementary particles is the
restricted Poincaré group, SO0 (3, 1)nR3,1 . This group does not contain time re-
versal operations, parity transformations, or combinations thereof. The univer-
sal cover of the restricted Poincaré group is SL(2, C) n R3,1 , and the irreducible
unitary representations of this group can be classified by two parameters, (m, s),
mass and spin, hence each elementary particle is associated with a Hilbert space
Hm,s .
The photon is a particle of mass 0 and spin 1. The Wigner approach obtains
a Hilbert space H0,1 of square-integrable cross-sections of a vector bundle E0,s +
4
over the forward light cone V0+ in energy-momentum space, with typical fibre
isomorphic to C1 .
However, the electromagnetic force respects symmetries such as parity trans-
formations. As a consequence, it is conventional to treat the interaction car-
rier of the electromagnetic force, the photon, as an irreducible representation of
O↑ (3, 1) n R3,1 , the isochronous Poincaré group, (also called the orthochronous
Poincaré group). This group consists of both the identity component of the
Poincaré group, and the component which contains the operation of parity re-
versal, P : (x0 , x1 , x2 , x3 ) 7→ (x0 , −x1 , −x2 , −x3 ).
Whilst the irreducible representations of the restricted Poincaré group are
parameterized by s ∈ 21 Z, the irreducible representations of the isochronous
Poincaré group are parameterized by t ∈ 12 Z+ . For t 6= 0, the t-representation
is a space of cross-sections of a bundle over the forward light cone with typical
fibre isomorphic to C2 . It can be decomposed into a direct sum of the s = t
and s = −t representations. In the case of a mass zero particle, the discrete
parameter s is often alternatively referred to as the ‘helicity’ or the ‘polarization’
of the particle, and the spin is treated as the absolute value |s|.
Thus, to summarise, the photon is treated as a particle of mass m = 0 and
+ +
spin t = 1. The bundle E0,t = E0,1 possesses sub-bundles of helicity s = 1
and s = −1, which correspond to the right-handed and left-handed ‘circular
+ +
polarization’ of a photon. These sub-bundles correspond to the Em,s = E0,1
+ +
and Em,s = E0,−1 bundles used in the representations of the restricted Poincaré
group.
With an understanding of how polarization enters the fray in terms of space-
time symmetries, in the next subsection we proceed to explore the nature of
polarization in greater depth.
2.3 Polarization
Explanations of the polarization of light typically begin by introducing ‘linear’
(or ‘plane’) polarization. They then proceed to define two states of opposite ‘cir-
cular polarization’ as superpositions of orthogonal states of linear polarization.
The most general polarized state is said to be one of ‘elliptical polarization’,
which can be defined as an arbitrary superposition of the states of opposite
circular polarization.
The states of definite helicity correspond to the states of circular polarization.
A general state of polarization, (elliptical polarization) will be one in which the
helicity is indefinite. Hence, a general state will belong to the direct sum of two
irreducible representations of SL(2, C) n R3,1 .
The discovery of polarization pre-dated the discovery of quantum theory,
(see Guillemin and Sternberg, 1984). Hence, the polarization of light can be
represented in both geometrical optics, and Maxwellian electromagnetism, as
well as quantum theory. The possible polarization states of a ray of light, a free
electromagnetic wave, or a photon, are represented by a 2-dimensional complex
vector space isomorphic to C2 . In the case of a wave, this polarization space is
tensored with the space of wave solutions. The real component of these complex
5
solutions corresponds to the electric field vector of a classical electromagnetic
wave.
In the case of a plane-wave with a spacelike wave-vector k, the electric field
vector is constrained to lie in the real 2-dimensional space perpendicular to
k. The complexification of this space corresponds to the space of complex
polarization vectors.
For example, start with a pair (e1 , e2 ) of spacelike orthogonal unit vectors
which span the spacelike plane perpendicular to k. If we align the x-axis with
e1 , the y-axis with e2 , and the z-axis with k, then e1 represents light which
is linearly polarized along the x-axis, and e2 represents light which is linearly
polarized along the y-axis. We can therefore denote our basis as (ex , ey ). In
terms of the electric field vector, a state of linear polarization along an axis
{ae : a ∈ R1 } is one in which the tip of the electric vector oscillates back and
forth along the line spanned by e.
The two-dimensional real vector-space generated by (ex , ey ) captures all the
possible states of linear polarization. To be precise, the set of all possible states
of linear polarization corresponds to the real projective space RP1 defined by all
the 1-dimensional subspaces through the origin of this 2-dimensional real vector
space.
Linear polarization states exist in any given direction within the real plane
perpendicular to k. Given the basis (ex , ey ), one can define an alternative basis
(ex0 , ey0 ) which is rotated by an angle θ from the original pair:
ex = cos θex0 + sin θey0

.
ey = − sin θex0 + cos θey0
If a beam of photons linearly polarized along the ex -axis impinges on a Po-
laroid sheet which is oriented along the ex0 -axis, then only a fraction cos2 θ
will be transmitted. In effect, we have the familiar quantum-mechanical sce-
nario where a system has been prepared into a state which is not an eigen-
state of the quantity being measured, and a decomposition of the state vector
in terms of its components along the eigenvectors of the measured quantity,
Ψ = c1 ψ1 + c2 ψ2 = ex = cos θex0 + sin θey0 . The probability of measuring po-
larization along the ex0 -axis is the square modulus of the coefficient c1 = cos θ,
which in this case happens to be a real coefficient.
The states of linear polarization are clearly not classical. The photons in a
beam polarized along the ex -axis are indefinite with respect to (ex0 , ey0 ) until
a measurement-like interaction takes place. All the familiar logic of quantum
theory can be found in the states of linear polarization, despite the fact that it
is only a real space.
However, whilst the space of linearly polarized states is non-classical, it does
not comprise the entire space of polarized states. This is obtained from the
complexification of the space of linearly polarized states. Define a basis (e+ , e− )
of this complex space as follows:
6
e+ = ex + iey
.
e− = ex − iey
This is the helicity basis, or equivalently, the eigenstates of circular polarization.
e+ is the state of positive helicity, or left-circular polarization, while e− is the
state of negative helicity, or right-circular polarization. In a circularly polarized
state, the tip of the electric field vector will sweep out a circle in the plane
perpendicular to k.
A general polarized state is then expressible as a superposition of the positive
and negative helicity eigenstates:
Ψ = c+ e+ + c− e− .
These are the elliptically polarized states. In general, the tip of the electric field
vector will sweep out an ellipse in the plane perpendicular to k. Both circular
polarization and linear polarization can be seen as limiting cases of elliptical
polarization.1
However, the helicity eigenstates provide only one possible basis for the
space of polarization vectors. Orthogonal linear polarization eigenstates provide
alternative bases for the entire space. Whilst the real vector space generated
by a pair of linear polarization eigenstates only spans the linear polarization
subspace, the set of complex linear combinations spans the entire space.
There is one particularly elegant way of representing the space of polariza-
tion states. The space of states corresponds to the set of complex 1-dimensional
subspaces of a 2-dimensional complex vector space. This is the complex projec-
tive line CP1 , the Riemann sphere,2 which can be coordinatized as the set of
ratios of complex number pairs, w/z.
In general, the ratio of a pair of complex numbers is determined by the
relative size of the respective amplitudes, and by the phase-difference:
w rw eiθw rw i(θw −θz )

= iθ
= e .
z rz e z rz
Orthogonal states correspond to antipodal points on the Riemann sphere. If
the positive helicity eigenstate Ψ = e+ is mapped to the North pole, and the
ratio c− /c+ = 0/1 = 0, then the negative helicity eigenstate Ψ = e− is mapped
to the South pole and the ratio c− /c+ = 1/0 = ∞.
If the helicity eigenstates are mapped to the poles, the set of linear polar-
ization states is mapped to the equator. For example,
1 1 1 1
e+ + e− = (ex + iey ) + (ex − iey ) = ex ,
2 2 2 2
and
1 As the ratio of semi-major axis and semi-minor axis of an ellipse tends to infinity, it flattens
into a line oriented in a particular direction, corresponding to the case of linear polarization.
2 The optics literature also refers to this geometrical representation as the Poincaré sphere.
7
1 1 1 1
−i e+ + i e− = −i (ex + iey ) + i (ex − iey ) = ey .
2 2 2 2
In terms of ratios of numbers, the first case ex corresponds to c− /c+ = 1, and
the orthogonal state on the equator ey corresponds to c− /c+ = −1, two complex
numbers with unit modulus and a phase difference of π.
States of elliptical polarization lie between the poles and the equator. If
the state of left-circular polarization corresponds to the North pole, then left-
elliptical states lie in the Northern hemisphere, and right-elliptical states lie
in the Southern hemisphere. Points on lines of constant latitude correspond to
ellipses of constant ellipticity, but a varying angle of the semi-major axis. Points
on lines of constant latitude correspond to a constant angle of the semi-major
axis, but a varying ellipticity.
To link the polarization states with the electric field component of a classical
free electromagnetic wave, the polarization state Ψ = c1 e1 +c2 e2 can be inserted
as the complex amplitude of a plane-wave:
Ψei(k·r−ωt) = (c1 e1 + c2 e2 )ei(k·r−ωt) .

The electric field vector is then deemed to be the real component of this complex
field:
E(r, t) = Re[Ψei(k·r−ωt) ] .
For example, if we start with a basis of linear polarization eigenstates (e1 , e2 ) =
(ex , ey ), and take the positive helicity state e+ = ex + iey , then:
Re[E0 (ex + iey )ei(k·r−ωt) ] = E0 ex cos(k · r − ωt) + E0 ey cos(k · r − ωt + π/2)

= E0 ex cos(k · r − ωt) − E0 ey sin(k · r − ωt) .
For any point r in the plane perpendicular to k, this defines the clockwise
rotation of a vector around a circle. E0 is the real magnitude of this polarization
vector. The clockwise rotation corresponds to positive helicity. Taking the
negative helicity polarization state, e− = ex − iey , then:
Re[E0 (ex − iey )ei(k·r−ωt) ] = E0 ex cos(k · r − ωt) + E0 ey cos(k · r − ωt − π/2)

= E0 ex cos(k · r − ωt) + E0 ey sin(k · r − ωt) .
This defines the anti-clockwise rotation of a vector of length E0 around a circle,

which corresponds to negative helicity.
To close this section on polarization, note that a space of mixed states can
be built from the pure polarized states defined above. Whilst the pure states
correspond to points on the surface of the Riemann sphere, the mixed states
correspond to points in the interior of the unit ball enclosed by the Riemann
sphere. These mixed states can be represented as density operators such as the
following:
D = a1 Pe1 + a2 Pe2 ,
8
where Pei is the projection operator onto the subspace spanned by the vector ei ,
and the coefficients define a convex linear combination of projection operators,
so that a1 + a2 = 1, and ai ≥ 0.
In the case where a1 = a2 , the mixed state is referred to as representing
‘unpolarized’ or ‘totally unpolarized’ light. This corresponds to the point at the
centre of the unit ball enclosed by the Riemann sphere. It can be represented as
the equally weighted convex sum of any pair of orthogonal pure states at either
end of a straight line passing through the centre. For example,
0.5Pe+ + 0.5Pe− = 0.5Pex + 0.5Pey .
Direct sunlight, or sunlight which has undergone numerous scattering interac-

tions passing through cloud, is considered to be unpolarized in this sense.
A general convex sum of pure polarization states is said to represent ‘partially
polarized’ light. The ‘degree of polarization’ p can be defined as an excess of
one pure polarization state over its orthogonal state. Thus, if
D = 0.25Pe+ + 0.75Pe− ,
then the degree of polarization is the contrast,
|0.25 − 0.75|
p= = 0.5 .
0.25 + 0.75
This mixed state would be represented a point on the diameter between the
North pole and South pole of the Riemann sphere. If the state of right-circular
polarization e− is mapped to the South pole, then this mixed state would be
half-way between the centre and the South pole.3
In operational terms the state of polarization of a light-beam is characterised
by four numbers (S0 , S) = (S0 , S1 , S2 , S3 ), called the Stokes parameters. These
can be measured by the fraction of light transmitted through a sequence of four
filters. S0 , measured by the first filter, is simply the intensity of the light, and is
used to normalize the other three parameters. (S1 /S0 , S2 /S0 , S3 /S0 ) correspond
to the three Cartesian coordinates of the ‘Stokes vector’. This vector lies on or
within the unit ball enclosed by the Riemann sphere.
If one chooses the second filter to be a linear polarizer with a horizontal trans-
mission axis, then S1 /S0 provides a coordinate along the line passing through
the horizontal ex and vertical ey states of linear polarization, antipodal to each
other on the equator of the Riemann sphere. If one chooses the second filter
to be a linear polarizer with a transmission axis at 45◦ to the horizontal, then
S2 /S0 provides a coordinate along the line passing through orthogonal states of
linear polarization which are rotated by π/2 around the equator from the first
pair of linear polarization states. If the third filter is chosen to be one which
only transmits right-circularly polarized light, then S3 /S0 provides a coordinate
along the line passing through North and South poles.
3 Personal communication with Aaron Goldberg.
9
In terms of this Stokes vector, the degree of polarization is, in general, given
by: p
||S|| (S12 + S22 + S32 )
p= = .
S0 S0
The sunlight scattered from atmospheric molecules in a clear sky is partially
polarized, with the degree of polarization peaking in a circular band at an
angular separation of 90◦ from the position of the Sun. Thus, when the Sun
is at the zenith at mid-Summer, the band of maximum polarization coincides
with the 360 degrees of the horizon, where the light is horizontally polarized.
Conversely, during sunrise or sunset at the vernal or autumnal equinox, the
band of maximum polarization coincides with the North-South meridian, the
polarization axis being vertical where the meridian meets the horizon.
One might be tempted to interpret a mixed state D = a1 Pe1 + a2 Pe2 as
representing a population of photons in which a fraction a1 possess a polarization
state of e1 , and the remaining fraction a2 possess the polarization state e2 . The
probabilities might appear to express incomplete knowledge rather than intrinsic
indefiniteness. However, density operators which are not themselves projection
operators have an infinite number of possible decompositions. In particular, if
the unpolarized state of light is expressed as a mixture of linear polarization
states,
D = 0.5Pex + 0.5Pey ,
it can also be expressed as a mixture of circular polarization states:
D = 0.5Pe+ + 0.5Pe− .
Hence, the probabilities corresponding to the coefficients cannot be interpreted
as expressing merely incomplete knowledge.
So much for polarization. In the next section we turn to quantum field the-
ory, and then the quantization of the radiation field, expressed as a superposition
of various plane-wave modes and polarizations.
10
3 Quantum Field Theory
3.1 Fock space
According to modern mathematical physics, the physical world is composed of
quantum fields, and particles are merely ‘excitations’ of those fields. There
are two types of such quantum fields: matter fields and gauge force fields. All
quantum theories are obtained by applying quantization algorithms to classical
theories, and quantum field theory is the upshot of two levels of quantization:
in first-quantized relativistic quantum theory, each type of matter field or gauge
field corresponds to a Hilbert space of cross-sections of a vector bundle over
space-time, satisfying certain conditions; in second-quantized relativistic quan-
tum field theory (quantum field theory proper), Fock spaces are constructed
from these first-quantized particle spaces. In the guise of ‘one-particle states’,
these vector bundle cross-sections are, more or less, the types of thing which
are created or annihilated in second-quantized relativistic quantum field theory
(McCabe 2007).
Wigner’s classification of the projective, unitary, irreducible representations
of the Poincaré group supplies an infinite dimensional Hilbert space for a particle
with any possible combination of mass m and spin/helicity s.4 These Hilbert
spaces are the so-called single-particle Hilbert spaces from which the Fock spaces
of the second-quantized theory can be built. Given the single-particle Hilbert
space Hm,s for a bosonic system, the Fock space is
∞
M
Fm,s = F (Hm,s ) = Hm,sn ,
n=0
where Hm,s is the n-fold symmetric tensor product of Hm,s . The symmetric
n
tensor product is the image of the tensor product under the following projection
mapping:
X
Sn+ (f1 ⊗ · · · ⊗ fn ) = (n!)−1/2 fσ(1) ⊗ · · · ⊗ fσ(n) .
σ
The sum here is over all the permutations σ of the indices (1, 2, ..., n). The
symmetric tensor product is Hm,sn = Sn+ (Hm,s
⊗n
).
Given the single-particle Hilbert space Hm,s for a fermionic system, the Fock
space is
∞
M
Fm,s = F (Hm,s ) = Hm,s
∧n
,
n=0
where Hm,s
∧n
is the n-fold anti-symmetric tensor product of Hm,s . The anti-
symmetric tensor product is the image of the tensor product under the following
projection mapping:
4 The projective, unitary, irreducible representations of the restricted Poincaré group are in
bijective correspondence with the ordinary, unitary, irreducible representations of its universal
covering group, SL(2, C) n R3,1 .
11
X
Sn− (f1 ⊗ · · · ⊗ fn ) = (n!)−1/2 χ(σ)fσ(1) ⊗ · · · ⊗ fσ(n) ,
σ
where χ(σ) is the sign of the permutation. Each permutation of a set of el-
ements can be expressed as a sequence of transpositions of pairs of elements,
and, in this sense, χ(σ) = 1 for permutations obtained from an equal number of
transpositions, and χ(σ) = −1 for permutations obtained from an odd number
of transpositions. The anti-symmetric tensor product is Hm,s ∧n
= Sn− (Hm,s⊗n
).
There is an alternative, albeit mathematically equivalent approach, in which
the quantum field state space is obtained as a space K of complex-valued poly-
nomial functionals defined upon the complex infinite-dimensional single-particle
space. In this approach, the complement of the degree n − 1 polynomials in the
space of degree n polynomials corresponds to the n-particle subspace of Fock
space, (Derdzinski 2002, Section 3.6). This approach treats states in quan-
tum field theory as ‘wavefunctionals’ φ(f ), describing superpositions of different
(complexified) classical field configurations f .
The isomorphism between these two representations uses the fact that
for a collection of Hilbert spaces H1 , . . . , Hn , equipped with inner products
h·, ·i1 , . . . , h·, ·in , the inner product h·, ·i on the tensor product H1 ⊗ · · · ⊗ Hn is
defined by
n
Y
hf, gi = hfk , gk ik = hf1 , g1 i1 · · · hfn , gn in
k=1
for simple tensors,
f = f1 ⊗ · · · ⊗ fn , g = g1 ⊗ · · · ⊗ gn .
This extends to linear combinations of such simple tensors using the bilinearity
of the inner product. (Hereafter, we will drop the subscripts on inner products
to reduce the notational clutter.)
A degree p = p1 + · · · + pk monomial φ(w1 , . . . , wn ) = w1p1 · · · wkpk on Ck can
be transplanted to a degree p monomial φ(f ) on Hm,s by means of k vectors in
the 1-particle Hilbert space, {ej ∈ Hm,s : j = 1, . . . , k}, such that:
φ(f ) = he1 , f ip1 · · · hek , f ipk .

If the ej are selected from an orthonormal basis of Hm,s , then, (with appropriate
normalizing factors), the set of such monomials provide an orthonormal basis
for K, the space of polynomial functions on the 1-particle Hilbert space.
The selection of an orthonormal basis ej of the 1-particle Hilbert space de-
termines an expansion of each vector f ∈ Hm,s :
∞
X
f= cj ej = hej , f iej .
j=1
12
Hence, the polynomials on Ck can be transplanted to polynomials on Hm,s by
applying the monomial exponents to the expansion coefficients in an orthonor-
mal decomposition of each vector.
The elements of the n-particle subspace Hm,s
⊗n
map to n-degree monomials.
For example, suppose we have a third-degree monomial φ with p1 = 1, p4 = 2.
Then,
φ(f ) = he1 , f ihe4 , f i2 = he1 ⊗ e4 ⊗ e4 , f ⊗ f ⊗ f i .

The third-degree monomial maps to the element e1 ⊗ e4 ⊗ e4 in the 3-particle
Hilbert space. Conversely, a linear combination of vectors in the 3-particle
Hilbert space maps to a sum of third-degree monomials on the 1-particle Hilbert
space. For example, Ψ = c144 e1 ⊗ e4 ⊗ e4 + c444 e4 ⊗ e4 ⊗ e4 maps to:
φ(f ) = hΨ, f ⊗ f ⊗ f i
= c144 he1 ⊗ e4 ⊗ e4 , f ⊗ f ⊗ f i + c444 he4 ⊗ e4 ⊗ e4 , f ⊗ f ⊗ f i
= c144 he1 , f ihe4 , f i2 + c444 he4 , f i3 .
The mapping can, of course, be restricted to the symmetric or anti-symmetric

subspace of each n-particle Hilbert space Hm,s⊗n
, as appropriate.
Whilst the elements of the n-particle subspace Hm,s ⊗n
⊂ Fm,s map to n-
degree monomials, elements of the Fock space Fm,s which contain components
in more than one n-particle summand, map to polynomials. A ‘finite particle
vector’ is an element of Fock space ψ = ⊕∞n=0 ψn in which all but a finite number
of the ψn are zero. The value of the corresponding polynomial on an arbitrary
element f of the 1-particle subspace is given by:
∞
X
φ(f ) = hψn , f n i ,
n=0
n
where f is the n-fold tensor product of f with itself,
f n = f ⊗ · · · ⊗ f (n times) .
The alternative definition of the quantum field state space will occasionally
provide an interesting perspective, but for the most part below we will focus on
the Fock space approach.
Having defined the relevant Fock space, the next step is define creation and
annihilation operators, and from these to construct field operators. Suppose
one has a bosonic Fock space. For each n − 1-particle Hilbert space Hm,sn−1 ,
the creation of a particle with a state f ∈ Hm,s corresponds to the operator
a∗n (f ) : Hm,sn−1 → Hm,sn defined by
√
a∗n (f )(f1 ··· fn−1 ) = Sn+ ( n + 1 f ⊗ f1 ⊗ · · · ⊗ fn−1 ) ,
where Sn+ is the projection operator onto the symmetric n-particle subspace, so
that
13
√
Sn+ ( n + 1 f ⊗ f1 ⊗ · · · ⊗ fn−1 ) = f f1 ··· fn−1 .
For each n-particle Hilbert space Hm,sn the annihilation of a particle with a
state f corresponds to the adjoint operator an (f ) : Hm,sn → Hm,sn−1 , defined by
√
an (f )(f1 ··· fn ) = Sn+ ( n hf, f1 if2 ⊗ fn ) .
These creation and annihilation operators are operator-valued distributions in
the sense that they assign operators to functions f .
At various points in this paper it will be convenient to use the physicists’
‘bra-ket’ notation in association with the application of creation and annihilation
operators. In this respect, for idealised systems with only one possible mode of
excitation, an n-particle state is typically denoted as |ni, with
1
|n + 1i = √ a∗ |ni .
n+1
One particular implication of this is that:
1
|ni = √ (a∗ )n |Ωi ,
n!
where Ω is the zero-particle vacuum vector we will meet in the next sub-section.
The annihilation and creation operators on the bosonic Fock space satisfy
canonical commutation relations (CCRs):
a(f )a(g) − a(g)a(f ) = 0

a∗ (f )a∗ (g) − a∗ (g)a∗ (f ) = 0
a(f )a∗ (g) − a∗ (g)a(f ) = hf, giI .
Note that the creation operators and annihilation operators on fermionic Fock
space satisfy the canonical anti-commutation relations (CARs):
a(f )a(g) + a(g)a(f ) = 0

a (f )a∗ (g) + a∗ (g)a∗ (f ) = 0
∗
a(f )a∗ (g) + a∗ (g)a(f ) = hf, giI .

This distinction will become crucial at a later stage when we turn to the subject
of coherent states.
Given the creation and annihilation operators, one can try to define field
operators at each point x of Minkowski space-time by expressions such as
∞
X
A(x) = [fα (x)a(fα ) + f¯α (x)a∗ (f¯α )] ,
α=1
where the {f1 , f2 , ...} provide an orthonormal basis of the single-particle Hilbert
space. However, the second part of this series diverges, and, in general, quantum
14
field operators cannot be well-defined at individual points of Minkowski space-
time. Instead, one must treat the field operators as operator-valued distributions
A(f ) by ‘smearing’ them over functions f from a test-function space as follows,
Z
A(f ) = f (x)A(x)d4 x
M
Z n
X
= w-lim f (x) [fα (x)a(fα ) + f¯α (x)a∗ (f¯α )]d4 x ,
n→∞ M α=1
where both of the terms exist as weak limits5 on a dense set in Fock space as
n → ∞, (Prugovecki 1995, p155).
The test function space for the quantized radiation field is a complexified
space of electromagnetic vector potentials; for quantization in the Coulomb
gauge it is a set of divergence-free functions valued in C3 . The test-functions
f (x) are required to be infinitely differentiable, and both they, and all their
partial derivatives, must tend towards zero faster than a polynomial function
x−n , for any n ∈ N, as |x| → ∞.
Despite the mathematical necessity for smearing the field operators, it is
conventional in the physics literature to maintain the use of notation in which
field operators are assigned to points of space-time, and this paper will uphold
that tradition. Use of such notation should be taken as a shorthand to indicate
that the field operator A(x) has been smeared with a test function that either
has support in a small open neighbourhood of x, or which decays very rapidly
inside that neighbourhood. For example, one might multiply some polynomial
2
function by an exponential decay factor e−a||x|| .
5 “The weak limit of a sequence of operators A , A , ... in a Hilbert space is determined by
1 2
the weak limits of the corresponding sequences A1 f, A2 f, ... for all vectors f in their domains.
In turn, a vector h is the weak limit of a sequence of vectors h1 , h2 , ... in a Hilbert space if
hg|hi is the limit of hg|h1 i, hg|h2 i, ... for any vector g in that space,” (Prugovecki 1995, p155,
footnote 2).
15
3.2 The quantum vacuum
In both fermionic and bosonic Fock spaces the zero-particle subspace is H 0 =
C1 , the so-called vacuum sector. This subspace contains a distinguished non-
zero vector 1 ∈ C1 , called the vacuum vector. The vacuum vector is denoted as
0 or |0i in some of the quantum field theory literature, despite the fact that it
is not the zero vector in Fock space. To distinguish it from the zero vector, it is
more usefully denoted as Ω.
The Fock space representations of the canonical commutation relations are
defined by the requirement that there is a unique vector Ω which is such that
a(f )Ω = 0 for all f . The Fock space vacuum vector is cyclic with respect to the
algebra generated by the representation of the commutation relations.
The vacuum vector Ω is such that
1. Ω is the ground state of the quantum field. It is the minimum energy

eigenstate of the Hamiltonian.
2. Ω is a state with zero particles.
3. Ω is a state which is invariant under the unitary action of the Poincaré
group, hence it is a state which is shared by all inertial observers.
On each bosonic and fermionic Fock space there is an operator called the
number operator,
N = 0 ⊕ 1I ⊕ 2I ⊕ 3I ⊕ · · · .
The eigenstates of this operator are often considered to represent the states of
the second-quantized theory in which there are a definite number of particles.
The utility of Fock space is that it enables one to represent situations where there
is (i) a variable number of particles, or (ii) an indefinite number of particles.
The vacuum state of the free electromagnetic field is not an eigenstate of any
of the local field operators representing the electric field E(x)
b or the magnetic
field B(x)
b at points x of space-time. An eigenstate Ψ of a self-adjoint operator
A is defined to be ‘dispersion-free’, where ‘dispersion’ is used as a synonym for
‘variance’.
The variance of a quantity is defined to be the mean value of (A − hAi)2 ,
where hAi denotes the mean value. (In the special case of a quantity with a
mean value of zero, the variance becomes the mean value of A2 ). In quantum
theory, the variance in the state Ψ is the expectation value:
hΨ, (A b 2 Ψi .
b − hAi)
In the vacuum state, the mean values of the electromagnetic field operators are
zero:
hΩ, E(x)Ωi
b = hΩ, B(x)Ωi
b =0.
16
Hence, the variance of the electric and magnetic field is given by expectation
values of the squared fields, and these are non-zero (Rugh and Zinkernagel,
2002):
b 2 (x)Ωi =
hΩ, E 6 0
b 2 (x)Ωi =
hΩ, B 6 0.
Hence, the local field operators of the free electromagnetic field are not
dispersion-free in the vacuum state. Equivalently, the relativistic vacuum state
of the electromagnetic field is not an eigenstate of the local field operators. In
this sense, the vacuum state of the free electromagnetic field contains ‘fluctua-
tions’.
In fact, none of the eigenstates of the photon number operator are also
eigenstates of the local field operators. The photon number operator doesn’t
commute with the field operators. Hence, the electric and magnetic fields are
subject to fluctuations whenever the state contains a definite number of photons.
Whilst the vacuum state is the state of minimum energy, it is not necessarily
a state of zero energy. This purported non-zero energy of the vacuum state is
referred to as the ‘zero-point energy’.
Given that the Hamiltonian density of the electromagnetic field is H(x) =
1 2
2 (E (x) + c2 B2 (x)), the zero point energy of the electromagnetic field per unit
volume is (Rugh and Zinkernagel 2002; Aitchison 1985, p347):
Z XX
hΩ, HΩi = hΩ, 12 (E b 2 (x) + c2 B
b 2 (x))d3 x Ωi = 1
2 ~ωk, .
b
k
To obtain a finite value, the electromagnetic field has been confined to a cubical
box of finite volume V = L3 . The sum on the right is over the two independent
polarization degrees of freedom = 1, 2, and the normal modes k up to an
‘ultraviolet cut-off’ ||k||max . i.e.,
2π 3
k∈ Z , ||k|| < ||k||max .
L
However, it is conventional for the Hamiltonian to be normally ordered in quan-
tum field theory calculations, (i.e., the order of annihilation operators and cre-
ation operators are swapped wherever they appear as products with creation
operators to the right of the annihilation operators). Using the : : notation
to indicate normal ordering, this results in the disappearance of the zero-point
energy,
hΩ, : H
b : Ωi = 0 .
Note that the vacuum state defined here is the vacuum state of the free elec-
tromagnetic field. It is not the vacuum state of the interacting electromagnetic
and electron-positron fields. We will discuss the interacting field vacuum in a
later section.
17
3.3 The quantized radiation field
The basic technique for quantizing a field is to express the field as an inverse
Fourier transform, and then to substitute creation and annihilation operators in
the place of the Fourier coefficients which specify the amplitude of each mode.
These field operators then act on the Fock space constructed from the relevant
one-particle Hilbert space; i.e., the irreducible unitary representation of the
space-time symmetry group corresponding to the excited states (‘quanta’) of
the field.
In the case of the electromagnetic field, this process is typically applied first
to the electromagnetic potential Aµ (x) rather than the electric and magnetic
fields E(x), B(x). However, because an entire gauge equivalence class of poten-
tials correspond to a single pair E(x), B(x), constraints must be applied at some
stage of the process, and there are different ways of doing this.
In the Gupta-Bleuler method, the Lorenz gauge condition is applied after a
provisional quantum state space has been obtained. The approach begins by tak-
ing the space of all real vector potentials Aµ (x) which solve the wave-equation,
and complexifying it into a space of cross-sections fµ (x) of the complexified
cotangent bundle. The Fourier transform f˜µ (k) of each element in this space is
concentrated on the light cone in energy-momentum space.
This space is then equipped with an inner product. However, the inner
product is indefinite, hence this state space is not a Hilbert space proper. It
is, instead, a Krein space HK , or ‘pseudo-Hilbert’ space. It has a direct sum
decomposition (Prugovecki 1995, p237):
HK = H − ⊕ H + ,
where
H − = {f : fa = 0, a = 1, 2, 3}, H + = {f : f0 = 0}.
The inner product (and norm) is positive definite on H + , and negative definite
on H − .
This Krein space is a space of provisional one-particle states. The one-
particle states are photon states, and three types of photons are identifiable
in this state-space. If f ∈ H − , then a∗ (f ) creates a so-called ‘scalar pho-
ton’. The positive-definite subspace of the Krein space can be decomposed
into a transverse subspace, H ⊥ = {f : f3 = 0}, and a longitudinal subspace
H k = {f : f1 = f2 = 0}. If f ∈ H ⊥ , then a∗ (f ) creates a so-called ‘transverse
photon’, and if f ∈ H k , then a∗ (f ) creates a so-called ‘longitudinal photon’.
Scalar photons and longitudinal photons, however, are considered to be unphys-
ical states of the free field.
Having obtained this Krein space, it can be treated as the one-particle space
of the unconstrained theory. To obtain a physical state space, one imposes
the Lorenz gauge condition. In terms of real 4-vector potentials, this is the
stipulation that:
∂A0
∂ µ Aµ = − +∇·A=0.
∂t
18
The space of such potentials is Lorentz invariant. i.e., a Lorentz transformation
maps one Lorenz gauge 4-potential into another Lorenz gauge 4-potential.
To apply this gauge constraint, a Krein-Fock space is constructed from the
one-particle Krein space HK , and one can define operators which create b a∗µ (k, )
and annihilate baµ (k, ) one-particle states indexed by momentum k and polar-
ization . Field operators A bµ can thence be constructed from the creation and
annihilation operators:
3 Z
1 X d3 k
A
bµ (x) =
3
aµ (k, i ) + e−i(k·x−ω(k)t) b
ε(k, i )[ei(k·x−ω(k)t) b a∗µ (k, i )]
(2π) i=0 V0+ 2ω(k)
b+
=A b−
µ + Aµ ,
where {ε(0, i ) : i = 0, 1, 2, 3} is an orthonormal basis for C4 .

The Lorenz gauge constraint is the requirement that physical states |f i sat-
isfy the equation:
!+
∂Ab0
− +∇·A b |f i = 0 ,
∂t
where the + superscript indicates that the positive-frequency part of the oper-
ator is taken. This part of the operator is constructed from an inverse Fourier
transform of photon annihilation operators, hence it is the component of the
field capable of mapping states f which contain particles into the zero vector.
b+ is substituted into this equation it follows that:
If A µ
a3 (k, i ) − b
(b a0 (k, i ))|f i = 0 .
This entails that a state which contains a longitudinal photon of momentum

k must also contain a scalar photon of the same momentum. This eliminates
the states of negative norm. The states of zero norm are then eliminated by
defining an equivalence relationship, so that states which only differ in their
longitudinal and scalar photon counts are physically equivalent. The set of
equivalence classes duly possesses a positive-definite inner product.
There is a variation on this method of applying the Lorenz gauge constraint.
Here we extract the salient points from the exposition of Prugovecki (1995,
pp237-240). Once the one-particle Krein space is obtained, an immediate re-
striction to a subspace satisfying the Lorenz gauge is imposed. Recall that
the Lorenz gauge condition on the fµ (x) in the configuration representation is
defined by the expression:
∂ µ fµ (x) = 0 .
On the Fourier-transformed Krein space, H˜K , it is defined by the condition:
k µ f˜µ (k) = 0 .
One can choose a basis {ea (k) : a = 0, 1, 2, 3} for the fibre of the contangent
bundle over each point k of the forward light cone which is such that k µ e1µ (k) =
19
k µ e2µ (k) = 0, and such that k = k 0 (e0 (k) + e3 (k)). In this basis, f˜(k) =
f˜a (k)ea (k), and the Lorenz gauge condition is equivalent to the restriction:
H˜Lorenz = {f˜(k) ∈ H˜K : f˜0 (k) = f˜3 (k)} .
There is a further subspace H˜0 ⊂ H˜Lorenz such that:
H˜0 = {f˜(k) ∈ H˜Lorenz : f˜1 (k) = f˜2 (k) = 0} .
These are the complexified analogues of the real vector potentials which cor-
respond to a vanishing electromagnetic field Fµν = 0. This is the subspace of
elements with zero norm, the ‘null’ subspace.
The direct sum decomposition of the Krein space H˜K = H˜ − ⊕ H˜ + reduces
to the following direct sum decomposition of the subspace satisfying the Lorenz
gauge condition:
H˜Lorenz = H˜0 ⊕ H˜⊥ .
The first summand is the null subspace, and the second is the ‘transverse’ sub-
space, the subspace of H˜ + satisfying the Lorenz gauge condition:
H˜⊥ = {f˜(k) ∈ H˜ + : f˜0 (k) = f˜3 (k) = 0} .
By taking the quotient,
H˜phys = H˜Lorenz /H˜0 ,
one obtains the physical one-particle space, equipped with a positive-definite

inner product. The members of each equivalence class differ only by virtue of
their scalar f˜0 (k) and longitudinal components f˜3 (k). Each equivalence class
can be mapped to a pair of transverse components (f˜1 (k), f˜2 (k)).
The Gupta-Bleuler approach to quantization of the free electromagnetic field
is Lorentz-invariant. An alternative approach to obtaining the physical space
of transverse photons is to select the Coulomb gauge, (or ‘radiation gauge’),
defined by the following joint condition:
A = (0, A) ∇·A=0.
The first condition alone is referred to as ‘temporal gauge’. This joint condition
is not Lorentz invariant, but it is convenient for quantizing the free electromag-
netic field.
By definition, a vector field A can be decomposed into the sum A = A⊥ +Ak
of a transverse component and a longitudinal component by the conditions,
∇ · A⊥ = 0 , ∇ × Ak = 0 .
Hence, in the Coulomb gauge, the 3-vector potential is purely transverse, A =

A⊥ .
The electric field is related to the 4-vector potential as follows:
∂A
E = −∇A0 − .
∂t
20
The transverse and longitudinal components of the electric field are as follows
(Honegger and Rieckers, p1988):
∂A⊥
E⊥ = −
∂t
∂Ak
Ek = −∇A0 −
∂t
The temporal gauge condition, A0 = 0, entails that the electric field reduces to:
∂A
E=− .
∂t
Taking the divergence of this yields:
∂∇ · A
∇·E=− .
∂t
Given that ∇ · A = 0 in the Coulomb gauge, it follows that:
∇·E=0.
This is Gauss’s law in the case of a free field. Or, to put it another way, the
electric field vector of the free field is transverse, E = E⊥ .
In the Coulomb gauge, the real 3-dimensional polarization vector ε is con-
strained to be orthogonal to the propagation vector k, hence one obtains a
Fourier decomposition of the vector potential:
2 Z
1 X d3 k
A(x) = 3
ε(k, i )[ei(k·x−ω(k)t) a(k, i ) + e−i(k·x−ω(k)t) a∗ (k, i )] .
(2π) i=1 V0+ 2ω(k)
Note that ε(k, i ) denotes the real 3-dimensional polarization vector associated
with the mode (k, i ), defined by combination of wave-vector k and complex
2-dimensional polarization basis vector i .
Substituting creation and annihilation operators in place of the Fourier co-
efficients one obtains the expression for the field operator:
2 Z
1 X i(k·x−ω(k)t) −i(k·x−ω(k)t) ∗ d3 k
A(x)
b = ε(k, i )[e a(k, i ) + e a (k, i )]
(2π)3 i=1 V0+
b b
2ω(k)
2 Z 3
1 X bk, d k .
= A i
(2π)3 i=1 V0+ 2ω(k)
21
4 Interactions between light and matter
4.1 The coupled electromagnetic field
The quantized radiation field is equivalent to a quantization of the free electro-
magnetic field. In contrast, the subject of this section is the electromagnetic
field interacting with a charge-current density (ρ, j).
In terms of the 4-vector potential, the Maxwell equations become the fol-
lowing:
∂ ρ
− ∇2 A0 − (∇ · A) =
∂t 0
∂2A ∂A0 j
− ∇2 A + ∇(∇ · A − )= ,
∂t2 ∂t 0
where 0 is the permittivity of free space. From the first equation it follows that
ρ
∇·E= .
0
In the presence of a non-zero charge density, the fact that ∇ · E 6= 0 entails that
the electric field is not purely transverse. In fact, the Gauss law becomes an
equation for the longitudinal component Ek of the electric field because:
ρ
∇ · E = ∇ · (E⊥ + Ek ) = ∇ · Ek = .
0
It follows that Z
1 ρ(y, t) 3
Ek (x, t) = − ∇ d y.
4π0 R3 |x − y|
k
Hence, the longitudinal component E of the electric field at a point in space
is determined by an integral of the charge density ρ taken over all of space.
Even if the charge density is confined to a bounded open subset of space, it will
determine the longitudinal electric field throughout all of the surrounding space.
When the theory is quantized, the distribution of expectation values of the
longitudinal electric field and charge density will possess the same relationship:
hb
Z
b k (x, t)i = − 1 ∇
hE
ρ(y, t)i 3
d y.
4π0 R3 |x − y|
The longitudinal component of the electric field represents the Coulomb field.
It is typically referred to as the electrostatic Coulomb field, but note that Ek is
permitted to evolve in accordance with:
∂Ek 1
= − jk .
∂t 0
There is no violation of relativistic causality; the electric field as a whole satisfies
the following wave-equation (Honegger and Rieckers, p2068):
∂j 1
E = −µ0 − ∇ρ ,
∂t 0
22
where µ0 is the permeability of free space, and
1 ∂2
= −4.
c2 ∂t2
(Whilst we have tacitly been using ‘relativistic units’, in which c = 1, the
speed of light has been re-introduced in this expression to make the relativistic
causality explicit).
The gradient of the charge density is longitudinal, and the wave equation
reduces to separate wave equations for the transverse and longitudinal compo-
nents of the electric field, (ibid, p2069):
∂j⊥
E⊥ = −µ0
∂t
k
∂j 1
Ek = −µ0 − ∇ρ .
∂t 0
Further sources of potential confusion can arise when specific gauges are chosen,
so let’s briefly try to clarify these matters.
Consider again the two equations at the beginning of the section for the
4-vector potential. Let’s apply the Lorenz gauge:
∂A0
− +∇·A=0.
∂t
Reverting to ‘natural’ units in which both c = 1 and 0 = 1, the equations for
the 4-vector potential reduce to the following:
∂ 2 A0
− + ∇ 2 A0 = ρ
∂t2
∂2A
− 2 + ∇2 A = j ,
∂t
Hence, in the Lorenz gauge, both the scalar potential A0 and the vector potential
A satisfy a wave-equation with source. Thus, in this gauge, changes in the
charge distribution propagate through the scalar potential at the speed of light,
and changes in the current propagate through the vector potential at the speed
of light.
In the Lorenz gauge, then, the potentials behave as a physical field would.
This is deceptive, however. If, instead of taking the Lorenz gauge, we impose
one part of the Coulomb gauge condition, ∇ · A = 0, then the first equation
becomes Poisson’s equation:
∇2 A0 = −ρ ,
and the second equation becomes
∂2A ∂A0
− + ∇2 A = j − ∇ .
∂t2 ∂t
23
The condition that ∇ · A = 0 also entails that j⊥ = j − ∇∂A0 /∂t, (Grensing,
p430), hence the second equation reduces to:
∂2A
− + ∇ 2 A = j⊥ ,
∂t2
a wave-equation with the transverse component of the current as source.
The appearance of the Poisson equation entails that under the gauge condi-
tion ∇ · A = 0, if the charge density ρ changes in time, then the scalar potential
A0 changes instantaneously to match it. However, the physical fields (E, B) do
not change instantaneously. A change in charge density entails the existence of
a non-zero current, and via the evolution of the vector potential A, the physical
fields (E, B) satisfy the strictures of relativistic causality.
4.2 Scattering, Feynman diagrams and virtual particles

We now turn to the representation of interactions in quantum field theory. Al-
though the states in Fock space are the states of a free system, by defining a
scattering operator S one can calculate the transition probabilities hψout | Sψin i
between the asymptotically free incoming states ψin and outgoing states ψout
in a collision or decay process, special types of interaction in which the coupling
is transient and spatially localised.
The scattering operator is defined in terms of an interaction Hamiltonian
density operator H b I (x). The total Hamiltonian of an interacting system can
be broken up into the free Hamiltonian operator H b 0 (x), and the interaction
Hamiltonian operator H b I (x).
The scattering operator S can be expressed in a Dyson perturbation expan-
sion as
∞
(−i)n
X Z Z
4
S =1+ d x1 · · · d4 xn T [H
b I (x1 ) · · · H
b I (xn )]
n=1
n! R4 R 4
∞
X
= Sn .
n=0
H b I (x1 ) · · · H
b I (x) is the interaction Hamiltonian density operator, and T [H b I (xn )]
is a time-ordered permutation6 in the sense that
b I (x1 ) · · · H
T [H b I (xi ) · · · H
b I (xn )] = H b I (xi ) ,
1 n
with t(xi1 ) ≥ t(xi2 ) ≥ · · · ≥ t(xin ). The interaction Hamiltonian density op-

erator is expressed in terms of the field operators relevant to the interaction
in question. In the case of quantum electrodynamics (QED), for example, one
has field operators Ψ(x)
b for the second-quantized Dirac field, field operators
6 When the Hamiltonian includes fermion field operators, one must introduce the sign of
the permutation into the expression for the time-ordering.
24
Ψ(x)
b for the second-quantized conjugate Dirac field, and field operators for the
second-quantized electromagnetic field Abµ (x), (Teller 1995, p124-125).
The creation operators, annihilation operators and field operators for an
electromagnetic field are defined upon a photonic Fock space:
Fγ = F (H0,1 )
Similarly, the creation operators, annihilation operators and field operators of
a Dirac field are defined upon the electron-positron Fock space Fe,e+ . This can
be obtained either by taking the Fock space of the direct sum of the electron
and positron single-particle spaces,
Fe,e+ = F (Hm+e ,1/2 ⊕ Hm− + ,1/2 ) ,

e
or by taking the direct sum of the Fock spaces for the electron and positron,
Fe,e+ = F (Hm+e ,1/2 ) ⊕ F (Hm− + ,1/2 ) .

e
For quantum electrodynamics, one takes the tensor product Fγ ⊗Fe,e+ of these
two Fock spaces. The operators of the individual fields are extended to the entire
tensor product space in a trivial manner. For example, the Dirac field operator
Ψ(x)
b defined upon Fe,e+ is extended to the operator I ⊗ Ψ(x) b defined upon
Fγ ⊗Fe,e+ . This fact can be conveniently suppressed from most of the notation,
but is important to bear in mind given that the interaction Hamiltonian density
operator is a product of field operators from Fe,e+ and field operators from
Fγ . The interaction Hamiltonian density operator is only defined upon the
tensor product Fock space Fγ ⊗ Fe,e+ . It is obtained by substituting field
operators into the expression for the classical interaction Hamiltonian density,
and subjecting them to ‘normal ordering’, (see Teller 1995, p127-129).
In the case of a Dirac field interacting with an electromagnetic field, the
interaction term in the Lagrangian density provides the interaction Hamiltonian
density. Thus, in the case of quantum electrodynamics, and prior to normal
ordering, one obtains the following interaction Hamiltonian density operator:
µb
b I (x) = q Ψ(x)γ
H b Ψ(x)A
bµ (x) ,
where q is the charge of the electron, and γ µ is a Dirac gamma matrix.

Even with the interaction Hamiltonian density operator for quantum elec-
trodynamics inserted into the expression for the scattering operator, it is still
impossible to rigorously calculate the transition amplitudes between the asymp-
totically free incoming states ψin and the asymptotically free outgoing states
ψout of electrons, positrons and photons, because the following perturbation
series is divergent:
hψout | Sψin i = hψout | Iψin i + hψout | S1 ψin i + hψout | S2 ψin i + · · · .
Physicists respond to this difficulty by assuming, rather than proving, that this
is a divergent asymptotic series. The significance of an asymptotic series is that
25
the first few terms can approximate the ‘true value’ one is trying to calculate,
even if the series as a whole diverges. Hence, assuming the series is asymptotic,
physicists calculate only, say, the first and second-order terms.
Under these assumptions, quantum field theory can be used to calculate the
transition probabilities hψout | Sψin i between matter field states and gauge field
states of a specific energy and momentum, as well as a specific particle number.
As an illustration, let’s consider Compton scattering, the scattering of a single
electron with a single photon (Teller 1995, p132-133).
An incoming state in which the electron has 4-momentum7 (ω(pi ), pi ) and
spin ri , and the photon has 4-momentum8 (ω(ki ), ki ) and polarization λi , is
represented as ψin = b a∗λi (ki )bb∗ri (pi ) Ω, where Ω is the vacuum vector, bb∗ri (pi )
is the creation operator for an electron of 4-momentum (ω(pi ), pi ) and spin ri ,
whilst ba∗λi (ki ) is the creation operator for a photon of 4-momentum (ω(ki ), ki )
and polarization λi .9 An outgoing state in which the electron has 4-momentum
(ω(po ), po ) and spin ro , whilst the photon has 4-momentum (ω(ko ), ko ) and
polarization λo , is represented as ψout = b a∗λo (ko )b
a∗ro (po ) Ω.
In terms of the tensor product Fock space Fγ ⊗ Fe,e+ , these incoming and
outgoing states can be represented as
a∗λi (ki ) ⊗ bb∗ri (pi )(Ωγ ⊗ Ωe,e+ )

ψin = b
a∗ (ki ) Ωγ ) ⊗ (bb∗ (p ) Ωe,e+ ) ,
= (b λi ri i
and
a∗λo (ko ) ⊗ bb∗ro (po )(Ωγ ⊗ Ωe,e+ )
ψout = b
a∗ (ko ) Ωγ ) ⊗ (bb∗ (p ) Ωe,e+ ) ,
= (b λo ro o
where Ωγ is the vacuum vector of the photonic Fock space, and Ωe,e+ is the
vacuum vector of the electron-positron Fock space.
The transition probability can be written as
a∗λo (ko )bb∗ro (po )Ω | S b

hψout | Sψin i = hb a∗λi (ki )bb∗ri (pi ) Ωi
= hΩ | bbr (p )b
o o o a∗ (ki )bb∗ (p ) Ωi .
aλ (ko ) S b λi ri i
Let’s consider the second-order term:

Z Z
1
hψout | S2 ψin i = − d4 x1 d4 x2 hψout | T [H b I (x2 )]ψin i .
b I (x1 )H
2 R4 R4
The interaction Hamiltonian density operator is:

µb
H
b I (x) = q Ψ(x)γ
b Ψ(x)A
bµ (x)
b − (x) γ µ Ψ
+
=q Ψ b (x) + Ψ b − (x) A
b + (x) + Ψ b+
µ (x) + A
b −
µ (x) .
7 With ω(pi ) = +(m2e + kpi k2 )1/2

8 With ω(ki ) = +kki k
9 As a notational convenience, in the case of a photon we shall use ‘k’ rather than ‘p’ to
denote the 4-momentum.
26
Recall that a product of field operators, such as that in the interaction Hamil-
tonian density operator, must be subjected to normal ordering, (denoted as
: Hb I (x) :), so that the order of the annihilation and creation operators is
swapped in any terms where they occur as products with creation operators
to the right of the annihilation operators.
The integrand of hψout | S2 ψin i is a sum of sixty-four terms, but only four of
those are non-zero. Each of these four terms corresponds to a distinct process
in Feynman diagram terms. Each can be evaluated by ‘walking’ the annihila-
tion operators to the right of the expression. This technique follows from the
commutation relations between annihilation operators, creation operators, and
field operators. Letting i and j independently denote either a space-time index
or a momentum-space index, these commutation relations are as follows (Teller
1995, p131):
[b a∗j ]± = c(i, j) .
ai , b
This entails that
a∗j = c(i, j) ∓ b
ai b
b a∗j b
ai ,
hence
hΩ | · · · b a∗j · · · Ωi = c(i, j)hΩ | · · · Ωi ∓ hΩ | · · · b

ai b a∗j b
ai · · · Ωi .
Repeated application of this procedure can be used to either ‘walk’ the an-
nihilation operators to the right-hand side, where bai Ω = 0, or walk the cre-
ation operators to the left-hand side, where hΩ|ba∗j = 0. This process leaves
behind only a product of complex-valued functions (‘c-numbers’), which can
then be integrated over the space-time variables, which in the case of second-
order Compton scattering are x1 and x2 . Neglecting issues of time ordering,
these commutators/c-numbers coincide with what are referred to in the physics
literature as ‘contractions’:
z }| {
a∗j = [b
ai b
b a∗j ]± .
ai , b
The technique extends to field operators, so that Ψb + (x2 ), the positive-frequency

component of the field, (and therefore the inverse Fourier transform of an-
nihilation operators), is walked to the right until Ψ b + (x2 )Ω = 0. Similarly,
−
hΩ|Ψb (x1 ) = 0.
As an illustration, let’s consider one of the four non-zero terms which con-
tribute to the second-order amplitude for Compton scattering:
− −

µ b− + µ b+ + ∗ ∗
Ω bro (po )b
b aλo (ko ) T Ψ (x1 )γ Aµ (x1 )Ψ (x1 )Ψ (x2 )γ Aµ (x2 )Ψ (x2 ) b
b b b b aλi (ki )bri (pi ) Ω .
b
This reduces to:
c(ko , x1 )c(po , x1 )c(x1 , x2 )c(x2 , pi )c(x2 , ki )hΩ | Ωi .
27
Given that hΩ|Ωi = 1, this corresponds to a product of complex-valued func-
tions. In total, the process described corresponds to the annihilation of a photon
of momentum ki and an electron of momentum pi at a space-time point x2 , the
creation of a virtual electron at x2 , and its annihilation at x1 , and the creation
of a photon of momentum ko and an electron of momentum po at a space-time
point x1 .
To calculate the contribution this process makes to second-order Compton
scattering, we then need to integrate this expression over the two arbitrary
space-time points:
Z Z
4
d x1 d4 x2 c(ko , x1 )c(po , x1 )c(x1 , x2 )c(x2 , pi )c(x2 , ki ) .
The Feynman diagrams beloved of textbooks on quantum field theory, offer a

graphical mnemonic for this algorithmic procedure involved in calculating each
term in such a perturbation series. However, each term in the perturbation
series is itself a divergent integral, hence even the calculation of the first and
second-order terms requires the use of so-called ‘renormalization’ to obtain finite
results. Renormalization introduces factors into the integrands which enable the
integrals to approach a finite value as the limits of the integrals are taken to
infinity, (see Teller 1988).
Thus, as Berestetskii et al comment, “The lack of complete logical consis-
tency in this theory [QED] is shown by the occurrence of divergent expressions
when the mathematical formalism is directly applied, although there are quite
well-defined ways of eliminating those divergencies. Nevertheless, such meth-
ods remain to a considerable extent, semiempirical rules, and our confidence in
the correctness of the results is ultimately based only on their excellent agree-
ment with experiment, not on the internal consistency or logical ordering of the
fundamental principles of the theory,” (1982, p4).
4.3 The interacting field vacuum

Note that the Fock space vacuum is the vacuum state of a free field, and, with
respect to the Fock space number operator, a state of zero particles. The free
field vacuum of Fock space is a distinct concept from what is often called the
‘dressed vacuum’ of interacting fields.
The dressed vacuum purportedly contains an infinite number of ‘virtual’ par-
ticles; particles which, if they existed, would violate the relativistic relationship
hp, pi = −m2 between mass m and energy-momentum p.
The notion of virtual particles is often invoked to explain and justify other-
wise ad-hoc renormalization procedures, which are used to obtain finite results
from the perturbation series in the second-quantized scattering theory. However,
the free-field vacuum in Fock space is the only vacuum which is theoretically
well-defined in quantum field theory. As Prugovecki states, “the actual com-
putations performed in perturbation theory actually begin with expressions for
asymptotic states,. . . formulated in Fock space, and then progress through a
28
chain of computations dictated by Feynman rules, which have no direct bear-
ing to a mathematically rigorous realization of a non-Fock representation of the
canonical commutation relations. . . Hence, in conventional QFT [quantum field
theory] the existence of such a representation, and of a corresponding unique
and global ‘dressed vacuum’, is merely a conjecture rather than a mathematical
fact,” (1995, p198-199).
Rugh and Zinkernagel concur, arguing that the popular picture of the pro-
duction and annihilation of virtual particles in the ‘interacting’ vacuum, “is
actually misleading as no production or annihilation takes place in the vacuum.
The point is rather that, in the ground state of the full interacting field system,
the number of quanta (particles) for any of the fields is not well-defined. For
instance, the photon number operator does not commute with the Hamiltonian
for the interacting field system, hence one cannot speak of a definite number
(e.g. zero) of photons in the vacuum of the full interacting system” (2002, Note
27.)
Aitchison attempts to spell out this point-of-view in greater detail, (1985,
p352-353). He assumes that there is a free Hamiltonian H0 and a full Hamilto-
nian of perturbative form H = H0 + Hint , which are both defined on the same
state-space. He postulates that there is a complete set of eigenstates {|Ai} of
the free Hamiltonian:
H0 |Ai = EA |Ai ,
which are not eigenstates of the full Hamiltonian.10 Each eigenstate |Ai of the
full Hamiltonian is such that (H0 + Hint )|Ai = EA |Ai, and can be expanded in
the basis provided by the energy eigenstates of the free Hamiltonian:
X
|Ai = cA |Ai .
A
The interaction perturbs the state from an eigenstate of the free Hamiltonian
into an eigenstate of the full Hamiltonian, which is a superposition of all the
free Hamiltonian eigenstates, hence the system can subsequently transition into
any one of those eigenstates; in effect, the interaction permits the transition
between eigenstates of the free Hamiltonian with different energies. (In a later
section, we will how this works in the case of the spontaneous emission of light).
Aitchison argues that the interacting vacuum is simply a special case of this
logic. i.e., the ground state of the full Hamiltonian is different from the ground
state of the free Hamiltonian, and the ground state of the full Hamiltonian can
be expressed as a superposition of all the energy eigenstates of the free Hamil-
tonian. In this sense, the interacting vacuum contains an indefinite number of
free particles, and in particular, the interacting vacuum of QED contains an
indefinite number of photons. The photons from each definite particle-number
state in the superposition are the virtual photons commonly attributed to the
interacting vacuum.
10 If the interaction term were ‘trivial’, i.e., of the form H
int = λI, then the eigenstates of
the free Hamiltonian would also be eigenstates of the full Hamiltonian, with an extra λ added
to the eigenvalues. Hence, Aitchison assumes the interaction is non-trivial.
29
Aitchison acknowledges the assumption that the full Hamiltonian can be
decomposed as a perturbation H = H0 + Hint of the free Hamiltonian, and
points out that this condition will not hold in general. However, Haag’s theorem
raises a question-mark over whether the free-field vacuum and the interacting
vacuum can ever exist in the same state-space.
Haag’s theorem demonstrates that a free-field Fock space cannot directly
represent an interacting field system.11 It proves that a Fock space cannot
possess the vacuum vector of a free-field and the vacuum vector of an interacting
field.
A vacuum vector is required to be invariant under space-time translations,
and a Fock space possesses, up to phase, a unique translation-invariant vector.
The vacuum state |0i of a free field12 must be the ground state of the free-field
Hamiltonian H b 0 |0i = 0, and no vector in the ray spanned
b 0 , in the sense that H
by |0i, {c|0i : |c| = 1}, can also be the ground state of the full Hamiltonian H b
for an interacting system. If H0 |0i = 0, as required, then the requirement that
b
H|0i
b = 0 cannot also be satisfied.
For example, in the case of a self-interacting scalar field, with an interac-
tion Hamiltonian density HI (x) = P(φ(x)) = φ4 (x), when field operators are
substituted into this expression to obtain the interaction Hamiltonian density
operator, it contains, at the very least, one term with four creation operators,
which is not cancelled out by any other term. As a consequence, H b I |0i 6= 0,
and, in fact, |0i is not even an eigenstate of HI , hence one cannot render |0i
b
as the interaction vacuum by adding to H b I a term containing a finite constant,
(Fraser 2008).
Haag’s theorem entails that the free-field vacuum and the interacting vacuum
can only exist in the same state-space if one relaxes some of the conditions
defining a vacuum state. It is still quite possible for the ground state of the
free-field Hamiltonian and the ground-state of the full interacting Hamiltonian
to exist in the same space if one relaxes the requirement of invariance under
space-time translations. If one considers quantum field theory defined on a
discrete spatial lattice of finite volume, then there are only a finite number of
degrees of freedom (Duncan 2012, p369), hence there is a unique state-space
up to unitary equivalence, so the free-field ground state and interacting ground
state have to exist in the same state-space.
There are theorems in quantum field theory which can only be understood
if such an approach is adopted. For example, the Gell-Mann-Low theorem
assumes that there is a free-field vacuum |0i and an interacting vacuum |Ωi
in the same state-space, with a non-zero overlap hΩ, 0i = 6 0, and derives a
relationship between the time-ordered products of interacting field operators
11 SeeEarman and Fraser (2006) for an excellent discussion.
12 Herewe use |0i rather than |Ωi to denote the free-field vacuum state, just in case there
might also be the vacuum vector of an interacting field in the same Hilbert space.
30
and time-ordered products of free-field operators (Folland, p181-184):
RT
h0, T [φ̂01 (x1 ) · · · φ̂0n (xn ) exp(−i −T ĤI (τ )dτ )]0i
hΩ, T [φ̂1 (x1 ) · · · φ̂n (xn )]Ωi = lim RT ,
T →∞(1−i) h0, T [exp(−i −T ĤI (τ )dτ )]0i
shown here for the case of a self-interacting scalar field.

In fact, there are circumstances where the use of the Gell-Mann-Low theorem
can be justified without requiring a finite number of degrees of freedom. If the
quantum field is confined to a cubical box of finite spatial volume V = L3 , then
the spatial momenta k are restricted to a discrete lattice:
2π 3
k∈ Z .
L
Duncan (2012, Section 10.5) expounds the case of a self-interacting scalar field of
two different masses. The additional mass of the second field functions as a type
of self-interaction potential term in the Hamiltonian. The vacuum of one |0i1
provides the free field vacuum, and the vacuum of the other |0i2 is effectively an
interacting field vacuum. The restriction to discrete momenta permits one to
define the overlap 2 h0, 0i1 between the vacua of the two distinct fields. The Gell-
(2)
Man-Low formula can then be used to define the propagator i∆F (x − y) of the
(1)
‘interacting’ field in terms of the free-field propagator i∆F (x − y). The infinite
volume limit can then be restored, yielding a Poincare-invariant propagator
(2)
i∆F (x − y) for the interacting field.
This is a particularly interesting type of case because a finite volume ‘cut-off’
corresponds to an infinite lattice of spatial momenta. There is no restriction on
the magnitude of the spatial momenta, so no limit to the energy scale, and no
lower-limit to the length-scale which is being ‘probed’.
4.4 Do longitudinal photons exist?

In the quantization of the free electromagnetic field, the scalar photon states
and longitudinal photon states are considered to be fictitious by-products of
gauge-freedom. Once the set of gauge equivalence classes is taken, the only
types of physical photons are identified to be transverse photons.
However, in the treatment of scattering processes in quantum electrody-
namics, as outlined in the previous section, all four components A bµ of the
electromagnetic potential appeared. In fact, longitudinal and scalar photons
often re-surface in expositions of scattering between charged particles. They
can purportedly be found in the photon propagator DFµν (k).
Recall that, in Feynman diagram terms, the photon propagator
hΩ|T [A bν (y)]Ωi = Dµν (x − y) ,

bµ (x)A
F
corresponds to the amplitude for the creation of a virtual photon at x and its
annihilation at y if y0 > x0 , and the amplitude for the creation of a virtual
photon at y and its annihilation at x if x0 > y0 , (Folland 2008, p147). The
31
expression for the photon propagator contains a gauge-fixing parameter ξ. When
the so-called Feynman gauge ξ = 1 is chosen, the covariant momentum-space
photon propagator becomes:
−ηµν
DFµν (k) = ,
k 2 + i
where ηµν is the Minkowski metric tensor, and is a so-called ‘positive infinites-
imal’.
The momentum space photon propagator can be decomposed into a trans-
verse component, a Coulomb component, and a physically obscure residual term,
(Greiner and Reinhardt, p186-187):
DFµν (k) = DFµν(trans) (k) + DFµν(Coul) (k) + DFµν(resid) (k) .
The decomposition is both gauge-dependent, and Lorentz-frame dependent.

Switching back to the configuration representation, and choosing a Lorentz
frame defined by a timelike vector field n = ∂t , the Coulomb part can be ex-
pressed as:
δ(x0 − y0 )
DFµν(Coul) (x − y) = δµ0 δν0 .
4π|x − y|
Hence, in the case where x and y are two points with the same timelike coordi-
nates in the Lorentz frame of choice, there is an amplitude for the simultaneous
creation of a virtual photon at x and its annihilation at y, (or vice versa);
this amplitude contributes to the instantaneous Coulomb interaction between
charged particles in that Lorentz frame.
Greiner and Reinhardt conclude from this that “the Coulomb interaction
arises from the combined exchange of longitudinal and scalar photons,” (ibid.
p187), and similar commitments to the existence of longitudinal and scalar
photons can be easily found in the quantum field theory literature:
“Longitudinal and scalar photons are not observed as free par-

ticles . . . For an electromagnetic field in the presence of charges
. . . longitudinal and scalar photons play an important role as virtual
particles in intermediate states and provide a covariant description of
the instantaneous Coulomb interaction,” (Mandl and Shaw, Section
5.2).
“We know that the longitudinal part of the electric field results from
a certain combination of first-order derivatives of the longitudinal
and scalar potentials. Furthermore, the total longitudinal field en-
ergy is just the electrostatic Coulomb energy of the relevant system
of point charges. Thus, the exchange of longitudinal and scalar pho-
tons between charges must correspond to the instantaneous Coulomb
interaction between these,” (Keller, p319).
However, a commitment to the existence of longitudinal and scalar photons,

and their role in electrostatic forces, entails a commitment to the existence of
32
virtual photons. The virtual photons which propagate between pairs of vertices
in Feynman diagrams are not constrained to follow null curves; in the process,
they violate the special relativistic energy-momentum relation.
There are two possible interpretations here: either a static electric field is a
state of the electromagnetic field in which there are no photons at all, or it is a
state in which there are simply no transverse photons. If one accepts that lon-
gitudinal photons and scalar photons are simply a fictional by-product of gauge
freedom prior to the imposition of constraints, then the only types of physical
photons are transverse photons, and these only exist for free electromagnetic
waves.
But if a static electric field contains no photons at all, then by what means
are Coulomb forces exerted? In the absence of counteracting forces, an electro-
static field will impart momentum to a passing charged particle. How does this
generation of momentum occur if there are no photons exchanged between the
charged particles, and if action-at-a-distance is prohibited?
The problem can be cast in even starker terms if we consider an electrically
charged black hole. Such an object is created by electrically charged material
collapsing inside its Schwarzschild radius, or simply by electrically charged mat-
ter falling inside the event horizon of a pre-existing black hole. An electrostatic
field is created in the region outside the event horizon of the black hole. The
charged material inside the black hole is capable of exerting forces on charged
particles outside the black hole by virtue of its electrostatic field. Yet the mo-
mentum cannot be generated by the exchange of transverse photons, because
these cannot escape from inside the black hole, where the source of the field has
fallen.
Another conceptual problem is that static electric fields often provide clas-
sical states of the electromagnetic field, in which the expectation value of the
electric field is non-zero throughout a macroscopic volume of space, (for exam-
ple, between the plates of a charged capacitor). If states such as these are to be
represented on a photonic Fock space, it couldn’t be done with a photon-number
eigenstate |ni, because all such states have hn, E(x)
b ni = 0. But it also couldn’t
be done with a coherent state, because such a state represents an oscillating sys-
tem rather than a static system. And it couldn’t be done with a superposition
of coherent states because “a quantum superposition of distinct classical field
configurations is outside the classical domain,” (Field and Hughston, 1999).
This is a subtle issue, which is resolved only by re-assigning the longitudinal
component of the quantized electric field to the state-space of the matter field.
Consider the operator form of the expression which relates the electric field to
the 4-vector potential:
b0 (x, t) − ∂ A(x, t) .
b
E(x,
b t) = −∇A
∂t
Let’s follow the approach of Mari et al (2016), and Fourier transform the oper-
ators:
b0 (k, t) − ∂ A(k, t) .
b
E(k,
b t) = −ikA
∂t
33
Assuming the Coulomb gauge condition ∇ · A = 0, the operator form of the
Gauss law is
∇·Eb k (x, t) = ∇2 A
b0 (x, t) = ρb(x, t) ,
which Fourier transforms into:
k2 A
b0 (k, t) = ρb(k, t) .
This entails that:

b0 (k, t) = ρb(k, t) .
A
|k|2
If we substitute this into the expression for the electric field operator we obtain:
ik ∂ A(k,
b t)
E(k,
b t) = − 2 ρb(k, t) − .
|k| ∂t
For a time-independent electric field (i.e., an electrostatic field), the second term
vanishes, and one obtains:
b k (k) = − ik ρb(k) .
E
|k|2
The operator ρb(k) acts on the quantum state space of the matter field, hence the
same is true of E b k (k). Because the Gauss law is a condition on the longitudinal
component of the electric field, it entails that the operator for the longitudinal
component of the electric field is a function of ρb, hence it also is an operator on
the state-space of the matter field.
In terms of the Fock space of quantum electrodynamics, Fγ ⊗ Fe,e+ , the
matter density operator has the form Iγ ⊗ ρb, and the longitudinal component of
b k , where Iγ is the identity operator
the electric field is likewise of the form Iγ ⊗ E
on the photonic Fock space.
Thus, we have a component of the electromagnetic field which is represented
by an operator on the state space of the fermionic matter system, rather than
the bosonic state space of the free electromagnetic field.13
“The longitudinal (i.e. proportional to k) component of the electric

field operator is determined by the charge-density operator, and acts
on the Hilbert space of the particle alone. Therefore, even if the field
is in its vacuum state, the expectation value of the electric field is the
static Coulomb electric field generated by the expectation value of
the charge density, and hence depends on the particle wave-function.
This means that the state of the field alone does not contain all the
information on the electric field, since its longitudinal component is
encoded into the state of the particle.
13 We will consider in a later section how coherent states of fermions can be treated as
classical states. Hence, if the static electric field is determined by the degrees of freedom of
charged fermions, then the static electric field associated with a charged fermion will be in a
classical state if the fermion is.
34
“Seen from a different perspective, the longitudinal component of
the electric field is not a dynamical propagating degree of freedom,
since it vanishes in absence of external charges and is completely
determined by them, so there is no Hilbert space associated to it.
The Hilbert space of the field contains only the degrees of freedom
associated to the electromagnetic radiation, i.e. the magnetic field
and the transverse (orthogonal to k) component of the electric field.
Then in a product state with the field part in the vacuum, only
these components are in the vacuum mode, while there can be a
static electric field depending on the state of the particle,” (Mari et
al 2016, p6-7).
A static electric field is associated with the vacuum state of the photonic Fock
space, the zero-photon eigenstate. Whilst the photon is frequently proclaimed to
be the gauge boson which mediates the electromagnetic force between charged
particles, the Coulomb electrostatic interaction, upon which a large proportion
of chemistry depends, is not explained by freely propagating photons.
Let’s look at the role of the Gauss law constraint in this. When electromag-
netism is formulated as a canonical theory, each point of the classical phase-
space is pair of electric and magnetic fields, defined over 3-dimensional space
(E(x), B(x)). There is a Cauchy problem in which the fields evolve in time
(E(x, t), B(x, t)) given the specification of initial data (E(x, 0), B(x, 0)). In this
canonical approach, the Gauss law ∇ · E = ∇ · Ek = ρ is a constraint which
applies at each moment of time. If the initial data (E(x, 0), B(x, 0)) satisfy the
constraint, then it will be satisfied at every other moment of time.
The Gauss law, then, is a purely spacelike constraint on the longitudinal
component of the electric field. It doesn’t tell us how the longitudinal electric
field evolves, it tells us what it is. As Earman puts it, “relatively spacelike elec-
tromagnetic events are mutually constrained by the laws of electromagnetism,”
(1995, p126).
In the context of a discussion of particle horizons in cosmology, Earman
argues that an electrically charged particle, represented by a worldline γ passing
through a space-time point p, can feel the tug of another charged particle,
represented by a worldline δ, due to the Coulomb field, even if the causal past
of p doesn’t contain δ, J − (p) ∩ δ = ∅, and no signal has had time to propagate
between the particles since the beginning of the universe, (ibid, p126-127).
In a canonical approach to the quantization of the electromagnetic field,
one might impose the Gauss law constraint prior to quantization by restricting
the phase-space to pairs which do satisfy the constraint. Alternatively, one can
impose the constraint after quantization. This is also the natural approach when
a covariant quantization scheme is used. One has a Gauss law operator
b =∇·E
G b − ρb .
This is an operator on the state-space of the matter field. The physical states
of the matter field ψ are required to satisfy the condition:
Gψ
b =0.
35
This notion of a spacelike constraint supports the following interpretation: The
overall concept of the quantum field theory programme is to reduce classical
force fields to the exchange of gauge bosons. Disturbances in the electromag-
netic field can be represented by the propagation of transverse photons, but the
forces exerted by a static electric field cannot. The quantized electrostatic field
extends through a region of space, and exerts forces locally on other charged par-
ticles. The longitudinal component of the electric field transported by a charged
fermion, may not be an independent degree of freedom from the charged particle,
but it cannot be equated with the charged particle either. The quantized static
electric field is the intermediary which exerts forces on other charged particles.
If the quantum field theory programme to reduce classical force fields to the
exchange of gauge bosons is to be complete, then static electric fields have to be
reduced to the exchange of photons, the interaction carriers of the electromag-
netic force. The notion that there are ‘virtual’ longitudinal and scalar photons
is an unavoidable consequence of this. If the electrostatic force exerted at a
point in space is to be exerted locally, and if the programme of reducing the
electromagnetic field to the exchange of photons is to be implemented, then the
local exertion of the force has to be attributed to the emission and/or absorption
of photons at that point in space (or in a small bounded neighbourhood of it).
For the field to be static, transverse photons are unacceptable, hence the only
option is a structure which includes longitudinal and scalar photons.
Under this interpretation, one loses the capability to treat the exertion of
electrostatic forces as a process occurring in time. Instead, the electrostatic
field around a charged particle is reduced to a spacelike structure which includes
longitudinal and scalar photons, bound to the electric charge.
In general, a quantized electromagnetic field E(x, b t), B(x,
b t) exerts a force
on test particles in the neighbourhood of a point x in space, remote from the
charged sources of the field, because the field can be decomposed into space-
time networks of creation, propagation, and annihilation processes involving
photons (as well as the virtual quanta of the matter fields). The net effect of
these processes is to impart momentum to a test particle, (in the absence of
counteracting forces).
b k (x, t) associated with the Coulomb electro-
In particular, the electric field E
static force, exerts a force on test particles in the neighbourhood of a point x
in space remote from the charged source of the field, because the field can be
decomposed into spacelike networks involving longitudinal and scalar photons
(as well as the virtual quanta of the matter fields). The net effect of these net-
works is to impart momentum to a test particle, (in the absence of counteracting
forces).
Hence, while longitudinal and scalar photons cannot exist as free particles,
they do exist as links in the spacelike networks into which the electrostatic
interaction can be decomposed.
36
4.5 Spontaneous and stimulated emission of light
Let’s consider one particular case of the interaction between radiation and mat-
ter, the emission of light. Photons can be emitted from excited atoms either
by spontaneous emission or stimulated emission. Let’s start with the case of
spontaneous emission.
If an atom is treated in isolation from the electromagnetic field, then its en-
ergy eigenstates {ψEi : i = 0, 1, 2 . . .}, (i.e., the energy levels of its electrons, with
being the ground state ψE0 ) are all stationary states. Given the Hamiltonian for
the atom Hatom , the energy eigenstates are all mutually orthogonal, and there is
zero probability of a transition from one state to another, hψEi , Hatom ψEj i = 0.
However, if the full Hamiltonian of the interacting system is used, H =
Hatom + Hrad + Hint , with
Z
Hint = − bj · A b d3 x ,
where bj is the quantized charge-current density, and A

b is the quantized electro-
magnetic vector potential, then
hψEi , Hint ψEj i =

6 0.
Hence, transitions between energy states of the atom are possible.
Let’s represent this explicitly. Because the interacting system of atom and
electromagnetic field has been introduced, the state space becomes a tensor
product:
Hatom ⊗ Hrad .
The states of the joint system in which a definite number of particles are present
can be denoted as
ψEi ⊗ (nk1 ,1 , nk2 ,2 , . . .) ,
in which ψEi is the i-th energy level of the atom, and (nk1 ,1 , nk2 ,2 , . . .) denotes
the number of photons of each mode and polarization.
The presence of the interaction term in the Hamiltonian entails that there
are non-zero matrix elements of the form (Aitchison 1985, p345):
hψE0 ⊗ (1k, , 0, . . .), A

b k, ψE ⊗ (0, 0, . . .)i ,
i
where (0, 0, . . .) = Ω is the vacuum state of the free electromagnetic field, and
ψE0 is the ground state of the atom. So not only is there a non-zero probability of
a transition from an excited atomic state to the ground state, but in concert with
this there is non-zero probability that the initial state of the electromagnetic
field will transition from the vacuum state to a state with a photon present.
The spontaneous emission of photons from the excited atoms in a gas or
solid is isotropic. This corresponds to an isotropic directional distribution of
wave vectors k. Moreover, spontaneously emitted light will also be distributed
over a range of energies E = ~c|k|:
37
“Normal atomic spontaneous emission takes place in a time ∆t ∼ 10−9 s,
and the spontaneously radiated electromagnetic field has a frequency spectrum
centred on the atomic transition frequency, with a width ∼ 1/∆t of about
100MHz. There are additional sources of frequency broadening of the emitted
light spectrum. In a solid, there are crystal strains and inhomogeneities which
distort the energy levels of the emitting ions. Ions in different parts of the
crystal differ in their transition frequencies, so that light is emitted in a band
of frequencies. . . In a gas, atoms move with a distribution of velocities ν, and
radiate a field which is shifted in frequency by the Doppler effect by a fractional
shift (∆ω/ω) = ν/c. Atomic velocities are approximately 103 m s−1 so ∆ω/ω
is about 10−5 and [for optical wavelength radiation of ω ∼ 1015 Hz] the Doppler
shift ∆ω ∼ 1010 Hz. Of course, there is a distribution of velocities and a range
of emitted frequencies with a width approximately equal to ∆ω. The light from
such atoms therefore has an extremely broad bandwidth,” (Knight 1989, p291).
However, an excited atom can also be stimulated to emit photons by an
incident flux of photons at the same frequency as the atomic transition. In the
case of such ‘stimulated’ emission, the emitted radiation tends to be emitted
in the same direction as the incident radiation, with the same frequency and
polarization as the incident photons.
Laser light is produced by stimulated emission from a crystalline solid shaped
in a rod-like geometry. A population of excited atoms is created, which initially
emit photons spontaneously in all directions. Those which propagate along the
axis of the rod will tend to stimulate the emission of more photons in the same
direction, while those emitted in other directions will simply be lost. Hence, as
the process continues, a population of photons with (almost) the same wave-
vector k and polarization is created. Mirrors are typically placed at either end
of the cavity in which the crystalline rod is suspended. This produces a beam of
amplified light of a particular mode. Not only do the photons lie within a very
narrow range of frequencies, but because they are concentrated in a particular
direction, the energy density of the beam is much greater than that of isotropic
radiation.
Laser light provides an approximate physical realisation of a ‘coherent’ quan-
tum state of light, a concept to which we turn next.
38
5 Coherent states
It is sometimes claimed that laser light consists of a set of photons in the same
state. In fact, if a laser is treated as a coherent state of the quantized radiation
field, then this statement is, strictly speaking, false.
To understand why, first recall that according to quantum field theory, pho-
tons are merely the excitation quanta of the quantized free electromagnetic field.
There are states of the quantized free electromagnetic field in which there are
definite numbers of photons present, but there are also states in which there is
an indefinite number of photons. Such states can be described as non-particle
states of the quantized free electromagnetic field.
Let us begin by defining a ‘coherent’ state |αi of light in terms of a system
with a single mode. In this context a mode is defined by a particular combination
(k, ) of wave-vector k and polarization . Expressed in the ‘bra-ket’ notation,
a coherent state is defined in terms of a basis of photon number eigenstates as
follows:
∞
|α|2 X αn
|αi = e− 2 √ |ni ,
n=0 n!
where |ni represents a state with n photons of wave-vector k and polarization
. If fk, represents the 1-particle state corresponding to the mode (k, ), then
|ni = fk, ··· fk, (n times) .

The photon number eigenstates are also energy eigenstates, given that they
represent states with n photons of energy E = hν = ~c|k|. These photon
number eigenstates are often dubbed ‘Fock states’, but given the potential for
confusion with a general state in Fock space, we will avoid this usage.
The coherent state |αi is an eigenstate of the annihilation operator b
a:
a|αi = α|αi .
b
In the case of a system with a single mode, b a=b a(fk, ).
However, the annihilation operator is not self-adjoint, and the eigenvalue is
a complex number α = |α|eiθ . This complex number is considered to specify
the phase θ and amplitude A = |α| of the mode-(k, ) laser light, in the sense
that the spatial distribution of expectation values for the electric field operator,
resembles the form of a classical monochromatic wave of phase θ and amplitude
A, (Duncan 2012, p229):
hα, E(x,
b t)αi = −2CA sin(k · x − ωt + θ)ε ,
p
where C = ~ωk /2V for a field confined to a cubical box of volume V , and
ε is the real polarization vector associated with the mode (k, ). Hence, these
39
coherent states appear to be (semi-)classical states of the free electromagnetic
field.14
To be an eigenstate of an annihilation operator is, in fact, only one of the
criteria used to define coherent states. Along with Annihilation Operator Coher-
ent States (AOCS), there are Displacement Operator Coherent States (DOCS),
and Minimum Uncertainty Coherent States (MUCS). The coherent states of
light satisfy all three criteria. In a later section we will encounter coherent
states of fermions which are MUCS, but not AOCS.
Whilst the basis states |ni are eigenstates of the number operator, (i.e.,
interpretable as states in which a definite number of photons exist), a coherent
state is clearly a superposition of these states, and is not itself an eigenstate
of the number operator. Hence, there are an indefinite number of photons in a
coherent state of light.
One way of putting it is that “a Fock state [i.e., a photon-number eigenstate]
implies fluctuations in the field intensity, and a. . . coherent state implies fluctua-
tions in the photon number,” (Troup and Perlman, 1981), although it should be
noted that even in a coherent state the field operators have a non-zero variance.
If one could measure the number of photons in a pulse of laser light, one
would always measure a definite number, but one would measure different num-
bers on different occasions for an identical coherent state. Each particular coher-
ent state has an expectation value for the number operator, which corresponds
to the mean number of particles detected over repeated measurements. But a
coherent state has non-zero dispersion for the number operator. Specifically,
the probability of counting n photons of wave-vector k and polarization , is the
probability |hn, αi|2 , which is given by a Poisson distribution:
|α|2n −|α|2 µn −µ
|hn, αi|2 = e = e .
n! n!
If one subscribes to an interpretation of quantum theory in which measurements
trigger a non-unitary and non-deterministic collapse of the state, then |hn, αi|2 is
the probability that the coherent state will collapse into an n-photon eigenstate.
Textbook accounts of how optical detectors measure the amplitude of radia-
tion all tend to begin with the absorption of a photon, the actual ‘measurement-
like interaction’. What follows is then an account of how that event is amplified
up to the macroscopic level. For example, in CCDs, photomultipliers, and even
in photographic emulsions, the chain of events begins with the photoelectric
effect liberating an electron. (In the human retina, the absorption of a photon
by a rod cell causes a subset of a pigment molecule to change from a cis to a
trans configuration). Measurement of a pulse of coherent light in a state |αi
will generate n such events with probability |hn, αi|2 .
This Poisson distribution has a mean value of µ = |α|2 , and being a Poisson
distribution, the variance equals the mean, and the standard deviation σ =
14 Some authors use the term ‘quasi-classical’, and Duncan (p228) refers to coherent states as
‘ultraclassical’. Pure classical states determine a precise dispersion-free value for every physical
quantity, whereas coherent states suffer non-zero dispersion, so perhaps a more appropriate
term would be ‘infraclassical’.
40
µ1/2 = |α|. Hence, the greater the amplitude |α| of the complex number defining
the coherent state, the greater the expected number of photons, but the relative
uncertainty in the number of photons, given by the coefficient of variation σ/µ,
is inversely proportional to the amplitude |α|:
σ |α|
= = |α|−1 .
µ |α|2
Just as the particle number is indefinite in a coherent state, so is the energy.
But just like the particle number, the expectation value of the energy is finite.
If such states of indefinite particle number and energy are deemed to be ‘non-
particle’ states, then laser light is a non-particle state of the quantized free
electromagnetic field.
The set of photon number eigenstates provide a countable, orthogonal ba-
sis for Fock space. The collection of all coherent states {|αi : α ∈ C1 } is a
non-denumerable set, whose members are neither linearly independent nor or-
thogonal. Specifically,
hα, βi = exp(α∗ β − 21 |α|2 − 12 |β|2 ) .

However, the set of coherent states do still span the Fock space. (They are said
to form an ‘overcomplete’ family). The identity operator on Fock space can
be decomposed either as a sum of projection operators on the photon-number
eigensubspaces,
X
I= hn, · i|ni ,
n
or as an integral of projection operators onto the subspaces generated by the

coherent-states: Z
1
I= hα, · i|αi dα .
π
Hence, an arbitrary element |f i of Fock space can be expressed as:
Z
1
|f i = hα, f i|αi dα .
π
In particular, a photon-number eigenstate |ni can be expressed as an integral
over all the coherent states:
Z
1
|ni = hα, ni|αi dα .
π
Note that because the coherent states are not linearly independent, any
coherent state can be expressed as an integral over all the others:
Z
1
|αi = hβ, αi|βi dβ
π
Z
1
= exp(β ∗ α − 21 |α|2 − 12 |β|2 )|βi dβ .
π
41
So much for single-mode coherent states. To introduce the more general
concept, consider first the ‘wavefunctional’ approach expounded in the opening
section, in which elements of Fock space correspond to polynomial functions on
the 1-particle Hilbert space Hγ . In this approach, a coherent state is represented
as an exponential function:
φf (ψ) = ehψ,f i ,
for some f ∈ Hγ . If we recall that an exponential function can be expanded as

a power series, then we obtain:
hψ, f i2 hψ, f i3
ehψ,f i = 1 + hψ, f i + + + ··· .
2! 3!
This maps to the following element of Fock space:
(f f) (f f f)
1⊕f ⊕ ⊕ ⊕ ··· ,
2! 3!
where the symmetric tensor product has been chosen for the bosonic case
appropriate to photons.
Hence, seen from this perspective, a coherent state is the result of exponen-
tiating the action of the creation operator a∗ (f ) on Fock space. A coherent
state, interpreted literally, is the consequence of an unlimited creation process.
But because the process yields a normalized superposition over all the possible
different particle numbers, the expectation value of the energy is finite.
In this more general approach, the coherent states in photonic Fock space are
not indexed by complex numbers α ∈ C1 , as they were in the case of a system
with a single possible mode, but by the 1-particle states f ∈ Hγ . Honegger and
Rieckers (2015, p436) name these ‘Glauber vectors’:
∞
M 1
φf = exp(− 12 ||f ||2 ) √ n f,
n=0 n!
One can verify as follows that these states are normalized:

*∞ ∞
+
M 1 M 1
2
hφf , φf i = exp(−||f || ) √ n f, √ n f
n=0 n! n=0 n!
∞
X hf, f in
= exp(−||f ||2 )
n=0
n!
∞
X ||f ||2n
= exp(−||f ||2 )
n=0
n!
= exp(−||f ||2 ) · exp(||f ||2 )
=1.
The general inner product between two coherent states becomes:
42
hφf , φg i = exp(hf, gi − 12 ||f ||2 − 21 ||g||2 ) .
The vacuum vector Ω is a special coherent state, which corresponds to the zero
vector f = 0 ∈ Hγ ,
φ0 = Ω ,
The Glauber vectors are duly eigenvectors of the Fock space annihilation oper-
ators:
a(f )φg = hf, giφg .
Note that just as there was no restriction to complex numbers α of unit modulus
in the case of a system with only one possible mode, there is no restriction to
elements f ∈ Hγ of unit norm.
A fact of keen philosophical significance is that a coherent state is not or-
thogonal to the vacuum state Ω. Specifically,
hφf , Ωi = exp(− 12 ||f ||2 ) .

Hence, if the photon particle number of a coherent state is measured, there
is always a non-zero probability |hφf , Ωi|2 = exp(−||f ||2 ), of measuring zero
particles. The greater the norm of the vector f which indexes the coherent
state, the lower the probability.
The fact that hφf , Ωi =
6 0 for all coherent states φf , entails that if the initial
state is the vacuum state, then there is a non-zero probability of a transition
to any coherent state. This distinguishes the coherent states from the photon
number eigenstates, which are orthogonal to the vacuum state.
If we accept that coherent states are classical states, then it follows that
there is a non-zero probability of a transition from a vacuum state to a classical
state. Once the transition to a classical state has occurred, then measurement-
like interactions are capable of inducing a transition from that state into a
photon-number eigenstate, but a direct transition from quantum vacuum to
photon-number eigenstate is not possible.
Another way of looking at this is that the vacuum state Ω can be decomposed
in a basis consisting of coherent states. We’ve already seen that in the case of a
system with a single mode, any coherent state can be expressed as an integral
over all the other coherent states. This result can be extended to the general
case, but there are some technical complications.15
As a first step to surmounting these difficulties, consider an arbitrary m-
dimensional subspace V ⊂ Hγ of the 1-particle Hilbert space. There is a pro-
jection operator PV : Hγ → V onto each such finite-dimensional subspace. In
addition to the Fock space F (Hγ ) generated by the entire infinite-dimensional
1-particle space Hγ , there is a Fock subspace F (V) = ⊕∞ n=0 ( n V) for each
finite-dimensional V.
15 I am much indebted to Alfred Rieckers for spelling out the following mathematical mate-
rial.
43
Now suppose that we select Pm an arbitrary orthonormal basis {ek : k =
1, . . . , m} of V, so that f = k=1 αk ek for any f ∈ V. The identity operator on
F (V) can be decomposed as an integral over a finite-dimensional domain:
Z
1
IF (V) = m hα, · i|αi d2m α ,
π R2m
m
P1m, . . . , αm ) ∈ C now denotes the Glauber vector φf corre-
where each α = (α
sponding to f = k=1 αk ek .
We can write this as:
Z
IF (V) = hφf , · iφf dλV (f ) ,
V
2m m
where λV = d α/π is the scaled Lebesgue measure on the Borel sets of the
finite-dimensional space V. This measure λV is independent of the choice of
orthonormal basis.
One can define a projection operator Γ(PV ) : F (Hγ ) → F (V), such that
for any ψ ∈ F (Hγ ):
Z
Γ(PV )ψ = hφf , ψiφf dλV (f ) .
V
In particular, the vacuum vector Ω from the infinite-dimensional Fock space has
a projection onto each F (V):
Z
ΩV = hφf , Ωiφf dλV (f ) .
V
The finite-dimensional subspaces V possess the structure of a directed set I, as

do the Fock subspaces F (V). (The inclusion relationship between the subspaces
provides the binary ≥ relationship of this directed set). One can express the
identity operator on the entire Fock space as the net limit, in the strong-operator
topology,16 of all the projection operators Γ(PV ):
Z
IF (Hγ ) = s-lim hφf , · iφf dλV (f ) .
V∈I V
One can then decompose an arbitrary ψ ∈ F (Hγ ) as

Z
ψ = lim hφf , ψiφf dλV (f ) ,
V∈I V
where the limit is in the norm topology. Thus, a coherent state φg can be
expressed as the limit of a net of integrals over the other coherent states:
Z
φg = lim hφf , φg iφf dλV (f )
V∈I V
Z
= lim exp(hf, gi − 12 ||f ||2 − 21 ||g||2 )φf dλV (f ) .
V∈I V
16 A net (Ai )i∈I of bounded operators on a Hilbert space H converges strongly to A if
lim Ai x = Ax for all x ∈ H . In this event, every neighbourhood U of A is such that there is
some i0 ∈ I such that Ai ∈ U whenever i ≥ i0 .
44
In particular, one can decompose the vacuum vector Ω as
Z
Ω = lim hφf , Ωiφf dλV (f )
V∈I V
Z
= lim exp(hf, 0i − 12 ||f ||2 − 12 ||0||2 )φf dλV (f )
V∈I V
Z
= lim exp(− 12 ||f ||2 )φf dλV (f ) .
V∈I V
45
6 Why is there something classical rather than
nothing classical?
6.1 Coherent states and the vacuum
If one were to invoke the Rmany-worlds interpretation of quantum theory, then
each term of Ω = limV∈I V exp(− 12 ||f ||2 )φf dλV (f ) would correspond to a dif-
ferent ‘branch’ of the universe. Because each term corresponds to a different
coherent state, each branch is classical, yet the overall state is the quantum
vacuum.
If there is a quantum field theory for the universe as a whole, so that there
is a well-defined vacuum state for the entire universe, and if there are coherent
states for the entire universe which span the state-space, then this logic could
be generalised from the quantized radiation field to the universe as a whole. If
so, it would offer a joint explanation for why the world appears to be classical,
and why there is something (classical) rather than nothing (classical).17
Indeed, there is at least one precedent for such a proposal in the philosophy
of physics literature:
“The relativistic vacua. . . are eigenstates only of global observables

(observables like the total energy, the total charge, [the particle num-
ber] and so on) and not of any local ones.
“States which entail, say, that there is a table across the room,
and states wherein the world appears roughly as it appears to us
(full of approximately localized objects, full of systems which are
changing with time), are not necessarily orthogonal to [relativistic]
vacua. . . observers such as ourselves cannot establish, by any prac-
tical means, that our experience is not merely a constituent, merely
a branch, of that vacuum,” (Albert 1988, p128-129).
Needless to say, the quantum vacuum cannot be equated with nothing; it is

merely the quantum field state in which there are no particle-like excitations.
Even if the idea worked, one would still be left without an explanation of why
there is a quantum vacuum rather than nothing. Moreover, for the idea to
actually work, there are at least two technical hurdles which need to be cleared:
1. Fermions cannot possess coherent states in the sense that bosons can.
2. There is no quantum theory of gravity.

We next consider these problems in turn.
17 Generalising from the quantized radiation field, one might be tempted to declare ‘Let
there be light!’.
46
6.2 Fermions, Supersymmetry and Grand Unified Theo-
ries
The first problem is clearly defined by Weingard:
“It is. . . true that Fermi fields can exhibit classical, macroscopic be-
haviour. The He3 superfluid and superconductivity are two exam-
ples. But the basic particles of these phenomena are composite
bosons. In elementary particle physics (and chemistry), however,
what we see are not composite bosons, but fermions engaging in the
basic interactions. These are the quanta of fermion fields which do
not admit of ‘macroscopic’ coherent states,” (1991, p215).
Now, as we will see in the final section, there are fermion states widely
referred to as ‘coherent’. These are Minimum Uncertainty Coherent States
(MUCS) rather than Annihilation Operator Coherent States (AOCS). Specifi-
cally, in the non-relativistic quantum mechanics of a simple harmonic oscillator,
there are fermion states which minimise the product of the position and momen-
tum variance, and whose mean values follow the classical equations of motion.
These states form an overcomplete family, spanning the state-space of the har-
monic oscillator, (Gazeau 2009, p4).
However, in quantum field theory the problem is that the fermions in a
composite system cannot exist in the same state, by virtue of the Pauli exclusion
principle. This fact is encoded in the canonical anti-commutation relations for
fermion creation operators:
b∗ (f )b∗ (g) + b∗ (g)b∗ (f ) = 0 .

It follows that
b∗ (f )b∗ (f ) + b∗ (f )b∗ (f ) = 0 ,
which entails that b∗ (f )2 = 0. Hence, b∗ (f )|0i = |1i, but b∗ (f )|1i = b∗ (f )2 |0i =
0. For any fermion mode, the occupation number is either zero or one. Hence,
apart from α|0i + β|1i, there cannot be coherent fermion states with an indefi-
nite number of fermions of the same mode.18 The consequence is that fermionic
coherent states do not span the state-space, and in particular we cannot de-
compose the vacuum vector of the fermionic matter fields as a sum of coherent
states.
One might wonder if it is possible to use Supersymmetry to mitigate this
problem, given that the supersymmetry ‘supercharge’ operator Q transforms
bosons into their fermionic partners and vice versa. Recall that Supersymmetry
represents the state space of each system as a direct sum of a bosonic space and a
fermionic space. The supercharge operator Q can be decomposed as Q = q + q ∗ ,
18 Note that ‘Grassmann coherent states’ have been defined for fermions. These are eigen-
states of the fermion annihilation operators with Grassmann eigenvalues. The Grassmann
eigenvalues, however, lack a physical interpretation.
47
where q annihilates a fermion and creates a boson, while q ∗ annihilates a boson
and creates a fermion. Baker (2018) provides a simple example for a system in
which there are no momenta, so that particles can only contribute their mass to
the total energy: if (a, a∗ ) denote the boson annihilation and creation operators,
and (b, b∗ ) denote the fermionic counterparts for particles of the same mass m,
then:
p
q = (m)a∗ b ,
annihilates a fermion and creates a boson, while
p
q ∗ = (m)b∗ a ,
annihilates a boson and creates a fermion, without changing the total energy of
the system.
Unfortunately, as Baker notes, only one boson of a particular mode can
be swapped for a fermion. “(q ∗ )2 = 0, which means that we cannot violate the
exclusion principle by replacing multiple bosons with fermions.” Conversely, the
coherent boson states which span the bosonic state space are not the image of a
fermionic state under the action of q. So even supersymmetry cannot eradicate
the special status of bosonic coherent states.
Despite this shortfall, is it possible to devise a universe creation scenario
using just the properties of the bosonic vacuum? There may be an opportunity
to do just that, because according to certain Grand Unified Theories (GUTs),
the massive fermions of the Standard Model are the decay products of massive
gauge bosons. Hence, it could be postulated that the initial state of the universe
was the vacuum state of a bosonic field.
A universe creation scenario might proceed by postulating the vacuum state
of gravity ΩG , tensored with the vacuum state of all the other matter fields and
gauge force fields:
Ω = ΩG ⊗ ΩGUT .
We will consider what ΩG might be in the next section. For the sake of argument,
the vacuum state of the non-gravitational fields is chosen to be ΩGUT , the
vacuum state of a non-supersymmetric Grand Unified Theory. In schematic
terms, a coherent state decomposition of ΩG entails:
Ω = ΩG ⊗ ΩGUT
Z
= lim c(g) φg dλ(g) ⊗ ΩGUT
V∈I V⊂H
G
Z
= lim c(g) φg ⊗ ΩGUT dλ(g) ,
V∈I V⊂HG
where c(g) are the complex coefficients. The global vacuum state would de-
compose into a linear combination of every possible classical coherent state φg
of 3-dimensional space, tensored with the vacuum state ΩGUT of all the non-
gravitational fields.
48
The latter could then be split into the non-gravitational bosonic and
fermionic degrees of freedom:
ΩGUT = Ωbos ⊗ Ωferm .
A coherent state decomposition of the bosonic vacuum would then entail:
ΩGUT = Ωbos ⊗ Ωferm

Z
= lim exp(− 12 ||f ||2 ) φf dλ(f ) ⊗ Ωferm
V∈I V⊂H
bos
Z
= lim exp(− 12 ||f ||2 ) φf ⊗ Ωferm dλ(f ) .
V∈I V⊂Hbos
Now assume that we have a GUT which predicts the existence of X-bosons. We
define an X-boson in generic terms as a GUT gauge boson which decays into
quarks and leptons. In each classical coherent branch of this primordial vacuum
state there would be a non-zero expectation value for the number of X-bosons
per unit volume. Because these GUT bosons decay into quarks and leptons, all
the particles of familiar big-bang cosmology would then be available.
Unfortunately, even this would not be sufficient to provide a starting point
for big-bang cosmology. For that, one needs: (i) a plasma of quarks, leptons
and gauge bosons to be in a high temperature thermal state; and (ii) expansion
of the spatial geometry.
Without the expansion of space, the thermal state, as an equilibrium state,
would be time-invariant. Only with the expansion of space can the material
system be driven out thermal equilibrium, setting the scene for the subsequent
formation of luminous stars and galaxies, and the far-from-equilibrium open
systems they support.
6.3 Gravity and gravitons

One considerable hurdle to a well-defined universe creation scenario is the ab-
sence of a theory of quantum gravity. As Manin notes, “in any cosmogenic
scenario a description of the origin of matter must be accompanied by a de-
scription of the origin of four-dimensional space-time,” (1988, p6). For the
scheme outlined in the preceding section to be well-defined, one would need to
include gravity, as well as all the other forces and matter fields, in the defini-
tion of a global vacuum state. Nevertheless, despite the absence of a theory of
quantum gravity, there is a well-defined Fock space for the graviton (Ashtekar
and Geroch, 1974, p1250-1252), the hypothetical quantum mode of excitation
of the free gravitational field, so let’s see if anything can be gleaned from this.
The restriction to the free gravitational field is equivalent to a restriction
to regions of space-time in which the Weyl tensor is non-zero, but the Ricci
tensor vanishes. In particular, the gravitational Fock space is constructed from
the space of linearized Weyl tensors Cµνξo on flat Minkowski space-time which
satisfy the Bianchi identities. Equivalently, one can work directly with the space
49
of gauge equivalence classes of metric tensors γµν which yield such Weyl tensors.
(In this context, the Weyl tensor is the analogue of the free electromagnetic field
Fµν from Maxwellian electromagnetism, and the metric tensor is analogous to
the electromagnetic vector potential Aµ ). This linear space of real tensor fields
corresponds to a classical mass zero, spin-two field. It represents a space of
perturbations to the flat metric on Minkowski space-time.
The one-particle space for the graviton is constructed by turning this real
vector space into a complex Hilbert space HG . The bosonic gravitational Fock
space is then:
M∞
FG = HG n .
n=0
This gravitonic Fock space is equipped with a well-defined, unique, Poincare-

invariant vacuum vector ΩG , along with gravitonic creation and annihilation
operators indexed by momentum and helicity, and field operators constructed
from them.
At first sight, scattering interactions between gravitons and the quanta of
matter fields, such as the Dirac field for electrons and positrons, could be cal-
culated by introducing a tensor product Fock space,
FG ⊗ Fe,e+ ,
and by the definition of an interaction Hamiltonian on this space, which couples
the gravitational field to the matter field via the quantized stress-energy tensor.
As is well-known, the normal calculational scams of quantum-field theory prove
to be ‘perturbatively non-renormalizable’ in this case.
Nevertheless, in the absence of an adequate theory of quantum gravity, the
gravitonic Fock space is important as a conceptual guide. As a bosonic system,
there is an expectation that the quantized free gravitational field will possess
coherent states, which correspond to classical states of the free gravitational
field, i.e., gravitational waves. If one proceeded naively, one would write these
states as:
∞
M 1
φe = exp(− 21 ||e||2 ) √ ne,
n=0 n!
where e ∈ HG is any 1-graviton state.
If these coherent states are relevantly similar to those of the free electro-
magnetic field, they will span the state space, and provide a decomposition of
the vacuum vector ΩG . Hence, the state of zero gravitons could be decomposed
into branches, each one of which corresponds to an excited classical state of the
gravitational field.
However, this particular vacuum state of the gravitational field, ΩG , simply
corresponds to an absence of free field gravitational perturbations on a flat
Minkowski background. That background is still there in the vacuum state,
defining the geometry of space-time. So ΩG is not a state which defines an
absence of space-time geometry.
50
Nevertheless, if a state in which there are no gravitational perturbations
could be decomposed into classical branches, each of which is a well-defined
coherent state of the free gravitational field, it would be highly suggestive, and
one might be tempted by the following hypothesis:
“The geometry of space-time is a classical, macroscopic structure. It
can (presumably) be given a quantum mechanical account because
Bose fields have coherent states, whose expectation values are the
classical quantities,” (Weingard 1991, p215).
It’s important to remember, however, that the coherent states of the grav-
itational field and the electromagnetic field, defined in quantum field theoretic
Fock space, are only classical states of the respective free fields. The classi-
cal geometry of space-time won’t be captured by states of the free quantized
gravitational field.
Recall that the longitudinal component of the electric field, determined by
the charge density, has to be defined as an operator on the state-space of the
matter field rather than the state space of the free electromagnetic field. The
same logic applies to the longitudinal component of the gravitational field, de-
termined by the mass density.
The ‘weak’ gravitational field of a planet or main-sequence star can be rep-
resented by a perturbation to the flat Minkowski space-time background. Just
like a gravitational wave, this is represented by a perturbation to the Minkowski
metric:
gµν = ηµν + hµν .
However, this perturbation has a non-zero Ricci tensor. A general perturbation
will satisfy the linearized Einstein equation (Le Tiec and Novak, 2016, p26):
hµν = −16πGTµν ,
where hµν is a modified form of hµν with a trace of the opposite sign, and the
Lorenz gauge ∂ ν h̄µν = 0 has been used. This is clearly a wave-equation for
the metric, in which the stress-energy-momentum Tµν is the source. As noted
above, freely propagating gravitational waves are obtained when the Ricci tensor
vanishes, which is equivalent to the vanishing of the stress-energy-momentum
tensor. Hence, freely propagating gravitational waves are solutions of:
hµν = 0 .
In the case of a weak gravitational field generated by a massive body, Tµν 6= 0.
The Newtonian gravitational potential ΦG is incorporated into the g00 compo-
nent of the metric by the definition, (Frankel, p28)
p
−ΦG = 1 − −g00 (r) .
Assuming |ΦG | 1, g00 = −(1 + 2ΦG ). This enables us to reproduce the
Poisson equation of Newtonian gravitation:
√
∇2 ΦG = ∇2 −g00 = 4πGρm ,
51
√ √
where ρm is the mass density. Now, ∇2 −g00 = −R00 −g00 , where Rµν is the
Ricci tensor, hence we have an equation which links the Ricci tensor to the mass
density, (ibid. p29): √
R00 −g00 = −4πGρm .
This is the weak field gravitational equivalent of the Gauss law constraint. Once
the matter degrees of freedom are quantized, a field operator ρbm has to be
√
substituted for the mass density, and the quantized product of R00 −g00 is
simply a function of this, an operator on the state space of the matter field.
√
One can refer to the product R00 −g00 as a longitudinal degree of freedom,
in analogy with the longitudinal electric field. In the purely Newtonian theory,
the Poisson equation for the gravitational potential ∇2 ΦG = 4πGρm entails
that the gravitational field of Newtonian theory, g = −∇ΦG , satisfies ∇ · g =
−4πGρm . From the fact that g = −∇ΦG , it follows that Newtonian gravity
is a ‘conservative force’, so that ∇ × g = 0. In other words, the Newtonian
gravitational force field is a longitudinal vector field.
Gravitational waves are transverse shear waves in the geometry of space. If
they can be successfully quantized, they will be represented by coherent states
of transverse gravitons. The static gravitational force fields around massive
bodies, however, are analogous to the Coulomb fields of electromagnetism, and
correspond to longitudinal degrees of freedom.
As with electromagnetism, there is then a judgement to be made over
whether the longitudinal degrees of freedom should be relegated to functions
of the matter field, or whether longitudinal gravitons actually exist. Here is one
particularly clear expression of the latter point-of-view:
“Electromagnetic radiation contains transverse (observable) photons

which transport energy, the Coulomb field contains longitudinally
polarized (virtual) photons which do not carry energy away and can-
not be observed as free particles. Hence, there is a direct interaction
between a transverse photon and a gravitational field of a black hole,
but no gravitational interaction between a longitudinal photon and
black hole. Hence, the Coulomb field (electrostatic interaction by
‘lines of force’ and Gauss’ law) is able to cross the event horizon of
a black hole. Within the linearized theory of gravity, a similar dis-
tinction between longitudinal and transverse gravitons can be made.
Thus, a black hole can attract gravitationally matter and radiation
outside its event horizon, because the Newtonian gravitational field
is caused by longitudinal gravitons,” (Argyris and Ciubotariu 1998,
p149).
Certainly, a universe creation scenario of the type considered here will have to
make the leap of assuming that the macroscopic geometry of 3-dimensional space
corresponds to a coherent state in a yet-to-be discovered theory of quantum
gravity. A vacuum state of the gravitational field in this generalised sense would
be decomposable as a combination of all the possible coherent states, not merely
those corresponding to gravitational waves.
52
6.4 Inflation
Before moving on, let’s consider, for the sake of comparison, the type of vacuum
decay envisaged by the inflationary cosmology scenario. This postulates the
existence of self-interacting scalar field φ, minimally coupled to gravity, called
the ‘inflaton’. The scalar field corresponds to an effective perfect fluid in which
the potential V (φ) determines the equation-of-state p = wρ. The effective
pressure of the scalar field is given by:
p = φ̇/2 − V (φ) − (∇φ)2 /6R2 ,
where R is the scalar curvature. The energy density is:
ρ = φ̇/2 + V (φ) + (∇φ)2 /2R2 .
If the spatial derivatives are dominant, then
p −(∇φ)2 /6R2 1
w= ≈ =− ,
ρ (∇φ)2 /2R2 3
which would drive an expanding universe, but one with R(t) ∝ t, rather than the
exponential expansion required for the inflationary scenario (Kolb and Turner,
p277).
If instead the spatial derivatives are assumed be zero (or negligible), and the
self-interaction is zero (or negligible) V (φ) = 0, then
p φ̇/2
w= = =1,
ρ φ̇/2
which would also fail to yield inflation.

It is only when the spatial derivatives are assumed to be zero, and the time-
derivative is assumed to be negligible (as it must be for ‘slow-roll’ inflation),
and the self-interaction is non-negligible, that we arrive at the equation-of-state
for an effective Cosmological Constant:
p −V (φ)
w= = = −1 .
ρ +V (φ)
This will drive the exponential expansion required by the inflationary scenario.
To provide the initial conditions for big-bang cosmology, inflation postulates
that this very specific self-interacting scalar field was subject to two types of
decay:
1. A ‘false vacuum’ state decays to a ‘true vacuum’. The true vacuum corre-
sponds to the global minimum of the classical potential energy function.
During the decay of the false vacuum, the universe undergoes exponential
expansion at an approximately constant energy density.
53
2. At the end of inflation, the inflaton field oscillates about its global min-
imum. These oscillations correspond to the existence of scalar particles,
the quanta of the inflaton field. These scalar particles decay into other,
lighter particles, and the energy of the inflaton field oscillations ‘reheats’
the universe into a high temperature plasma appropriate as the starting
point for big-bang cosmology.
There are two immediate problems with the concept of the true vacuum here.
The true vacuum is the vacuum state of a self-interacting scalar field, hence it
will be the vacuum state of the full interaction Hamiltonian in the quantum
field theory. This is problematic, because the interaction Hamiltonian of a
self-interacting scalar field doesn’t commute with the particle number operator.
Hence, the true vacuum state of the inflaton cannot be a state with a definite
number of particles. The second problem is that the universe is in a state of
expansion, and the existence of a unique vacuum state requires a stationary
space-time. Hence, the ground state of the full interaction Hamiltonian will
change from one moment of time to the next.
Over and above these problems with the concept of the true vacuum, we
have to consider the postulated oscillations of the inflaton field. These oscilla-
tions are considered to be coherent states of a bosonic field: “When the scalar
field φ reached the steep part of the potential, it fell quickly to the minimum
and then oscillated about it. These oscillations were damped by terms in the
equations of motion which arise due to the coupling of φ to the other fields in
the theory. . . Note that the oscillations of φ describe a coherent state of zero
momentum scalar particles - the momentum is zero because the scalar field is
homogeneous. The damping of these oscillations may therefore be interpreted
as the decay of the scalar particles into lighter species,” (Blau and Guth 1987,
p560).
As we have explored in detail above, a coherent state of a quantum field is
not an eigenstate of the particle-number operator; it is a state with an indefinite
number of particles. Hence, the decay of the coherent inflaton state into a post-
inflationary plasma with a definite number of particles would requires some form
of wave-function collapse or decoherence. In the case of decoherence, it is unclear
what could provide the ‘environment’ or ‘macroscopic degrees of freedom’ which
become entangled with the coherent degrees of freedom.
In the absence of an empirically adequate Grand Unified Theory which ex-
plains the apparent existence of Cold Dark Matter, as well as all the familiar
particles and fields of the Standard Model, inflationary cosmology tends to be
non-specific about the immediate decay products of the inflaton field oscillations.
The particles observable in the universe today are considered to be merely the
end of a decay chain which started with the decay of the inflaton quanta.
In the period of time between the end of inflation and the time of reheating,
the mass-energy density of the universe was dominated by the coherent φ oscil-
lations, (Kolb and Turner 1990, p279). At the end of inflation, the fields into
which the inflaton particles decay can be thought of as being in their vacuum
state, Ωdecay . The interaction of the inflaton field with these other fields causes
54
the decay of the coherent inflaton state, and the transition from Ωdecay to non-
vacuum states which ultimately yield the quarks, leptons and gauge bosons of
the Standard Model.
However, as alluded to above, the existence of a unique vacuum state requires
a stationary space-time. Hence, even after the termination of the exponential
expansion, the oscillations of the inflaton φ make it difficult to define a unique
vacuum state for any field. In addition, there are competing requirements placed
upon the strength of the coupling between the inflaton and the other fields:
it has to be very small to facilitate an almost-flat potential energy function,
appropriate for ‘slow-roll’ inflation; yet it has to be large enough to permit the
reheating required for consistency with big-bang cosmology.19
In summary, inflationary cosmology is able to produce the starting conditions
for big-bang cosmology because it combines the decay of a coherent bosonic state
with the prior decay of a false vacuum state, to yield both a high temperature
thermal state and an expanding geometry.20
To reiterate, a universe creation scenario which proposes a decay of the
initial vacuum state, or the decay of a coherent state, is not the same thing
as suggesting that the current state of the universe is the vacuum. Hence,
the inclusion of fermionic degrees of freedom, and the need for an expanding
geometry, seems to force a universe creation scenario away from the notion that
the global state of the universe could be a vacuum state, and towards the more
familiar idea from modern cosmology that some sort of vacuum decay transpired.
19 Note that whilst most inflationary scenarios assume the lifetime of the coherent inflaton
oscillations to be extremely small compared to the current age of the universe, there are
variations on the theme which propose that the coherent φ-oscillations have only partially
decayed, and the mass-energy density of the remainder is masquerading as Cold Dark Matter.
20 There are, of course, other important putative successes of inflationary cosmology, such as
providing the spectrum of density-perturbations which seeded galaxy-formation, and ‘solving’

the so-called horizon and flatness problems.
55
7 Coherent states and the classical world
7.1 Are coherent states classical?
Coherent states are often presented as the class of pure quantum states which
most resemble the pure classical states: “A coherent state. . . is the closest quan-
tum analog to a classical electromagnetic wave. While the field is not assigned
a definite (vector) magnitude anywhere, the expected result of measuring it
varies from point to point in the same way as the magnitude of a classical
field, while individual measurement results are subject to random fluctuations
that decrease in relative proportion to this expected magnitude. Measurements
would also reveal a pretty well-defined phase at each point,” (Healey, 2013).
However, the claim that the coherent states of the radiation field are classical
states has several potential problems and questions to answer:
1. The classical states are not emergent from the quantum states.
2. Do the mixed classical states have unique decompositions?
3. The classical states are not obtained in the limit where there is a large
number of photons.
4. The classical states are not reference-frame independent.
Let’s elucidate each one of these in turn, starting with the implication that
classical states are not emergent from quantum states.
When we look at Fock space, we have one basis which consists of classi-
cal states (i.e., coherent states), and another basis which consists of photon-
number/energy-eigenstates. Any coherent state can be expressed as a linear
combination of photon-number eigenstates, and any photon-number eigenstate
can be expressed as a linear combination of the coherent states.
“A general (pure) state of the quantized electromagnetic field. . . has
a decomposition as a superposition of [photon-number] basis states,
and so a typical state vector is associated with an indefinite number
of photons. . . An arbitrary (pure) state also has a decomposition
as a superposition of coherent states, each of which approximates
a classical electromagnetic wave. It may therefore be thought of
either as composed of a (typically indefinite) number of photons from
various modes, or as partially constituted by each of a continuous
range of distinct, approximately classical, electromagnetic waves.
Since neither decomposition has priority, neither tells one what light
represented by such a state is really composed of: a photon-number
basis state may be decomposed into coherent states just as a coherent
state can be decomposed into photon-number basis states,” (Healey
2013).
If the radiation field is in a classical coherent state, then an appropriate
interaction to measure the number of photons in that state will induce a transi-
tion into a quantum state with a definite number of photons. Conversely, if the
56
radiation field is in a quantum state with a definite number of particles, then an
appropriate measurement-like interaction will induce a transition into a classical
coherent state. Far from yielding or defining the classical world, measurement-
like interactions can switch the state back and forth between the classical and
the quantum.
Furthermore, because coherent states are not orthogonal, repeated measure-
ments can induce transitions from one coherent state to another. If coherent
states are classical states, they are classical states which are unstable under
measurement-like interactions.
In defiance, then, of much of the literature exploring the relationship between
the classical world and the quantum world, the classical is not emergent here.
If we accept that photon-number eigenstates are quintessentially quantum, and
coherent states are classical, then the quantum and the classical exist here at
the same level.
In fact, one can flip the concept of classical emergence on its head, and
portray the quantum world as emergent from the interference of classical worlds.
This is perhaps clearest seen in the path-integral formulation of a quantum
theory, where the propagator K(q, t; q0 , t0 ) which determines the evolution of
the quantum state,
Z
Ψ(q, t) = K(q, t; q0 , t0 )Ψ(q0 , t0 ) dq0 ,
C
is decomposed as a sum (or integral) over kinematically-possible classical paths

or histories: Z
K(q, t; q0 , t0 ) = eiA(γ)/~ dµ(γ) .
P
The action A(γ) is a functional defined on the space of paths γ which interpolate
between the configuration q0 at time t0 , and the configuration q at time t.21
Orthogonal branches of the quantum state correspond to mutually disjoint
bunches of interfering classical histories. If the expression for the classical action
is altered, then the quantum state also changes, so the evolution of the quantum
state supervenes on the kinematically-possible classical histories.
Amongst the space of kinematically possible histories which interpolate be-
tween the configuration q0 at time t0 , and the configuration q at time t, will be
one history which is a critical point of the action-functional. This is the sin-
gle dynamically-possible classical history; i.e., it solves the classical dynamical
equations. Quantum states whose mean values track this critical-point of the
classical action can be said to be classical (or ‘semi-classical’, or ‘quasi-classical’)
states.
Hence, in this sense, classical theory can provide both the substrate and
the superstrate of quantum theory. Even the purported classical states, how-
ever, are the result of interfering bunches of classical histories. So, whilst there
is certainly a requirement to reconcile quantum theory with the nature of the
21 Rigorously defining this path-integral has proven to be difficult, a fact that physicists tend
to signal by prefacing its definition with the term ‘heuristic’.
57
macroscopic world, the macroscopic world is less classical than commonly be-
lieved: the universe is an interference pattern.
Let’s turn now to our second issue: the decomposition of mixed classical
states. First recall that in classical mechanics, the space of all states corresponds
to the set of probability measures on phase space, and this set does indeed
possess the structure of a simplex. A simplex of states is distinguished from
a convex set of states, in that each element in the interior of a simplex has a
unique decomposition as a convex combination of extremal boundary elements,
i.e., pure states.
In contrast, the space of all states of a quantum system is never a simplex
due to the non-unique decomposability of density operators. Hence a mixed
quantum state doesn’t have a unique decomposition as a convex combination
of pure states, and cannot be unambiguously interpreted as representing the
relative fraction of various pure states in a population (‘ensemble’) of systems.
However, Honegger and Rieckers demonstrate that the simplex of classical
states can be isomorphically mapped into the total space of quantum states.
Let’s briefly review how they define this in terms of Weyl algebras.
Start with a real pre-symplectic space (E, σ), which in the case of a field
system is infinite-dimensional, possibly a space of solutions to some relativistic
wave-equation. The dual E 0 of this space is the classical phase space. The
functions Φ0 (f ) : E 0 → R, with f ∈ E, are field observables, defined for F ∈ E 0
by F 7→ F (f ) = Φ0 (f )[F ]. To obtain a family of bounded functions, these can
be exponentiated:
W 0 (f ) = exp(iΦ0 (f )) : E 0 → C .
The closure of the complex linear hull of the W 0 (f ), f ∈ E, is the commutative
Weyl algebra W (E, 0).
The real-linear mapping E 3 f 7→ W 0 (f ) over a pre-symplectic space (E, σ)
is called a classical Weyl system. A canonical quantization replaces a classical
Weyl system W 0 (f ) over (E, σ), with a Weyl system W ~ (f ) over (E, ~σ). This
is a real-linear mapping E 3 f 7→ W ~ (f ) into the unitary operators on some
Hilbert space, which satisfies the Weyl relations:
W ~ (f )W ~ (g) = exp(− 2i ~σ(f, g))W ~ (f + g))

W ~ (f )∗ = W ~ (−f )
The non-commutative C*-algebra generated by the Weyl operators is called the

Weyl algebra W (E, ~σ).
Having fixed a Weyl algebra W (E, ~σ), different complex structures j defined
on the pre-symplectic space (E, σ), define different semi-norms || · ||j (and inner-
products (·|·)j ) on the complexified space E. These correspond to different vac-
j j
uum states ωvac on the Weyl algebra, defined by hωvac , W (f )i = exp(− ~4 ||f ||j ).
Now, suppose we have the pre-symplectic space E, and we have selected a
complex structure j, and defined a 1-particle Hilbert space Hj as the completion
of (E, (·|·)j ). Each pure state e ∈ Hj defines a character on E:
√
χe (·) = exp(i 2 Re(e|·)j ) .
58
Each such character corresponds to a pure state ωχe on the commutative Weyl
algebra W (E, 0):
hωχe ; W 0 (f )i = χe (f ), ∀f ∈ E .
The state space S(W (E, 0)) on the commutative Weyl algebra is closed under
convex combinations.
j
Now, with the selection of the vacuum vector state ωvac corresponding to
the complex structure j, we have a mapping from the Bauer22 simplex of states
on the commutative Weyl algebra into the total state space:
j
ωvac : S(W (E, 0)) → S(W (E, ~σ)) .
We can therefore map each ωχe into the total state space. Each one of these
pure states in the total state space can be regarded as a generalized Glauber
vector state, (Honegger and Rieckers, p657), i.e. a coherent state. Moreover,
the unique extremal decomposition of each element of S(W (E, 0)) is carried
j
over via ωvac into the classical states in the total state space.
Hence, Honegger and Rieckers represent the space of all classical coherent
states of the radiation field as a Bauer simplex in which “the extremal coherent
states are pure and given by Glauber vectors, which in turn are indexed by
one-photon wave-functions,” (ibid, p703).
In an earlier section we introduced the concept of unpolarized light as a
mixed quantum state which admits more than one decomposition. Unpolarized
light can, for example, be represented as a mixture of either linearly polarized
pure states, or a mixture of circularly polarized pure states; these expressions
are mixtures of photon-number eigenstates. Let’s spell this out in terms of states
on Weyl algebras.
Consider the following four modes ei ∈ Hj , each with the same wave-vector
k: Let ex denote the mode with linear polarization along the x-axis, let ey
denote the mode with linear polarization along the y-axis, let e+ denote the
mode with left-circular polarization, and let e− denote the mode with right-
circular polarization.
The four states of different polarization defined above therefore correspond
to four characters, χex , χey , χe+ , χe− , which in turn correspond to four pure
states ωχex , ωχey , ωχe+ , ωχe− on the commutative Weyl algebra.
The state space S(W (E, 0)) on the commutative Weyl algebra is closed under
convex combinations, hence the following pair of mixed states also belong to
S(W (E, 0)):
Mxy = 0.5ωχex + 0.5ωχey ,
M+− = 0.5ωχe+ + 0.5ωχe− .
These two classical mixed states do not equal each other:
Mxy 6= M+− .
22 A Bauer simplex is a simplex with a compact extremal boundary.
59
j
Now, with the selection of the vacuum vector state ωvac corresponding to
the complex structure j, we have a mapping from the Bauer simplex of states
on the commutative Weyl algebra into the total state space:
j
ωvac : S(W (E, 0)) → S(W (E, ~σ)) .
We can therefore map the four polarization states into the total state space,
j j j j
ωvac (ωχex ), ωvac (ωχey ), ωvac (ωχe+ ), ωvac (ωχe− ). To reiterate, each one of these
pure states in the total state space can be regarded as a generalized Glauber
vector state.
The unique extremal decomposition of each element of S(W (E, 0)) is carried
j
over via ωvac into the classical states in the total state space. Thus, we have
the following pair of mixed states in the total quantum state space:
j j
Dxy = 0.5 ωvac (ωχex ) + 0.5 ωvac (ωχey ) ,
j j
D+− = 0.5 ωvac (ωχe+ ) + 0.5 ωvac (ωχe− ) .
Once again, these mixed states, now representable as density operators on Fock
space, do not equal each other:
Dxy 6= D+− .
These states are mixtures of coherent states on Fock space, each of which is
indexed by 1-particle states that share the same mode k, but differ in their
polarization.
These mixtures need to be carefully distinguished from the following mixed
state. Let Pei denote the projection operator onto the ray spanned by ei in
the 1-particle subspace of Fock space. A state of unpolarized light can be
represented by a density operator with the following non-unique decomposition
into a convex combination of orthogonal pure states:
D = 0.5Pex + 0.5Pey = 0.5Pe+ + 0.5Pe− .
This is obviously not a mixture of coherent states, it is a mixture of photon-

number eigenstates.
The density operator D on the 1-particle space can be second-quantized into
a density operator dΓ(D), a mixed state on Fock space, and this mixed state
has a non-unique decomposition:
dΓ(D) = 0.5 dΓ(Pex ) + 0.5 dΓ(Pey ) = 0.5 dΓ(Pe+ ) + 0.5 dΓ(Pe− ) .
Once again, however, dΓ(D) is not a coherent state. In this respect, a state of
unpolarized light is very much a quantum rather than a classical state of light.
Note also that much of the visual wavelength light humans perceive is polar-
ized. Sunlight, and the light from incandescent bulbs, becomes linearly polarized
by virtue of undergoing Rayleigh scattering with atmospheric molecules, and by
virtue of reflecting off non-metallic surfaces. As we note in the next subsection,
60
these are ‘coherent scattering’ processes. They transform unpolarized light, a
mixed state of the radiation field, into linearly polarized light, another mixed
state, but one in which the mixture is only a convex combination of coherent
states with different wave-vectors, phases and amplitudes. The coherent scat-
tering processes with matter effectively perform a measurement-like interaction
on the light, selecting a basis of coherent states, and in particular, decomposing
the mixture in the basis of linearly polarized states. So, whilst laser light is a
polarized pure classical state of the radiation field, polarized sunlight is a mixed
classical state.
Let’s turn to our third issue: the notion that a coherent state is a classical
state of the radiation field undermines one particular popular notion that the
classical world emerges from the quantum in the limit in which the number of
particles tends to infinity. For example, Haag appears to link this to Bohr’s
‘correspondence principle’: “Classical electrodynamics is not the quantum the-
ory of a single photon but a correspondence limit of the quantum theory for
infinitely many photons,” (1996, p6).
To take another example, Sakurai’s well-known textbook on ‘advanced’
quantum theory asserts that classical electromagnetism is reliable when the
number of wavelength-λ photons per unit of volume, (λ/2π)3 , is much greater
than one.23 He estimates that for a Chicago FM radio station broadcasting at
(λ/2π) ≈ 48 cm, the number per unit volume at a distance of 5 miles from the
antenna is about 1017 :
“The classical limit of the quantum theory of radiation is achieved

when the number of photons becomes so large that the occupation
number may as well be regarded as a continuous variable. The space-
time development of the classical electromagnetic wave approximates
the dynamical behavior of trillions of photons,” (1967, p36).
On the contrary, if coherent states truly are pure classical states, and if the
Bauer simplex obtained by taking convex linear combinations of the coherent
states is a set of mixed classical states, then classical states of the radiation field
can be associated with a small, finite expected number of photons.
Troup and Perlman (1981) point out that in the limit where the expected
number of photons hni = |α|2 in a coherent state |αi tends to infinity, because
the relative uncertainty |α|/|α|2 = |α|−1 tends to zero, the distribution in photon
number tends towards that of the photon-number eigenstates |hnii. However,
23 Sakurai divides the energy density of a classical field by ~ω to estimate the number of
photons per unit volume, and assumes that the classical theory will be reliable when the
vacuum fluctuations of the field, of order ~ω per unit volume, are negligible compared to the
classical energy density. Sakurai’s simple method of estimating N , the number of photons per
unit volume, seems to be a monochromatic version of Zeldovich’s more general 1966 formula:
|Ẽ(k, t)|2 + |B̃(k, t)|2 3
Z
1
N = d k.
8π ~ω(k)
Interpreted literally, Zeldovich’s formula neglects the fact that a classical state is not a photon-
number eigenstate.
61
given that coherent states are classical states and photon-number eigenstates
are quantum states, this line of argument would only demonstrate that the
quantum emerges from the classical as the number of photons tends to infinity,
the very opposite of the claim made by Haag and Sakurai.
A classical state of the radiation field is not obtained in the limit where
there are many, many photons in a quantum state. In fact, if the state of the
radiation field is an eigenstate of the photon number operator, then no matter
how large the number of photons, (1023 , 10100 , etc), it will not correspond to a
classical state. It is the coherent states, and the mixtures of coherent states, not
the photon number eigenstates, which correspond to classical states. Moreover,
each coherent state is a superposition of photon number eigenstates. Hence, in
the case of the radiation field, the classical world emerges from superpositions.
To reconcile the theory with macroscopic phenomena, not only is it unnecessary
to eradicate superpositions, but on the contrary they are indispensable for the
emergence of classical phenomena.
The classical world is also not associated with a particular energy-scale. In
the case of Fock space, the classical coherent states are states of arbitrarily
high-energy. There is no upper limit on the expectation value of the particle
number in a coherent Fock state. Moreover, the quantum states are the states
whose energy can be as low as that associated with a single photon, or with the
vacuum itself.
There is one further problem with the notion of a coherent state as a classical
state, and in fact it is the same problem which besets the notion of the quantum
vacuum.
Every inertial (i.e., non-acceleratory) reference frame in Minkowski space-
time shares a common vacuum state, a common number operator, and a common
decomposition of the solution space to a relativistic differential equation into
positive-energy and negative-energy states. Uniformly accelerated observers,
however, have their own vacuum state, their own number operator, and their
own notion of positive-energy and negative-energy. In the inertial Minkowski
vacuum state, a uniformly accelerated particle detector will purportedly detect
a thermal flux of particles, (the so-called ‘Unruh effect’). These problems are
accentuated for quantum field theory in curved space-time, where not only does
each local freely-falling reference frame, defined by a geodesic of the curved
space-time geometry, possess its own notion of the vacuum, but the vacuum can
change from one moment to another within that fixed reference frame.
Now, the family of coherent states |αi includes the vacuum state as the spe-
cial case where α = 0, and it transpires that this entire class of states is mutable
under transition to an accelerated reference frame. Bishop and Vourdas (1986)
demonstrated that, not only do uniformly accelerated observers have their own
notion of a vacuum state, they also have their own notion of coherent states
in general. A single-mode coherent state in the reference frame of an inertial
observer, becomes a paired-mode state for a uniformly accelerated observer, a
particular type of ‘squeezed’ state.24
24 Tracing over one of the modes leaves the remaining mode in a thermal quantum state.
62
Along with the photon-number eigenstates, squeezed states are considered to
be non-classical states of light: “Squeezed states of light are nonclassical states
of light with phase-dependent quantum noise properties which can be less than
those of the vacuum state of the electromagnetic field,” (Knight 1989, p313).
The coherent states of the radiation field are sharp up to the quantum fluc-
tuations of the vacuum. This is best explained in terms of the ‘quadratures’ of
a wave. As a trigonometrical identity,
sin(x + φ) = cos(φ) sin(x) + sin(φ) cos(x)

= X1 sin(x) + X2 cos(x) .
This is a sum of the ‘in-phase’ quadrature component, and the component 90◦
out of phase with it. When φ = 0, sin(φ) = 0, X1 = 1 and X2 = 0. In terms of
the electric field, one has (Knight, p314):
E = E0 (X1 sin(ωt) + X2 cos(ωt)) .
The corresponding decomposition of the quantized electric field Eb contains op-

erators X1 and X2 , which are non-commuting. X1 represents the amplitude
b b b
and Xb2 represents the phase.25 They satisfy an uncertainty relationship:
b 1 · ∆X
b2 ≥ 1
∆X .
4
A coherent state of the radiation field is such that ∆X b2 = 1 and ∆X

b 1 · ∆X b1 =
4
1
∆X2 = 2 . A squeezed state is such that ∆Xi < 1/2 for one of the quadratures,
b b
and ∆X bj > 1/2 for the other.
When the electric field fluctuations are plotted over time as an interval either
side of the sinusoidally varying mean value, reducing the uncertainty ∆X b1 re-
duces the spread in amplitude at the expense of increasing the spread in phase,
while reducing the uncertainty ∆X b2 reduces the spread in phase at the expense
of increasing the spread in amplitude.
A sufficient condition for the non-classicality of an optical state is that the
variance of some field operator is less than that in the vacuum state, and by
virtue of this, squeezed states are deemed to be non-classical (Honegger and
Rieckers 2015, Criterion 25.2.8, p659).
Hence, if one cleaves to the notion that a coherent state is a classical state,
then the division between the quantum world and the classical world is not
reference-frame independent. Classical states in one reference frame can become
non-classical states in another.
25 There are other popular quadrature operators in the literature. For example, one often
sees the annihilation operator b

a decomposed as b a = (b
a+b a∗ )/2 + i(b
a−b a∗ )/2i = X
b 1 + iX
b2 .
a|αi = |α|eiθ |αi. Hence,
With respect to a coherent state |αi, recall that b
hα, (X b2 )αi = α = hX
b 1 + iX b1 i + ihX
b2 i = Re α + iIm α = |α| cos θ + i|α| sin θ .
63
7.2 Coherent scattering
Just as there are coherent and non-coherent states of the quantized radiation
field, scattering interactions between radiation and matter can be classified as
coherent scattering or incoherent scattering. When the incident radiation be-
haves in the manner of a classical wave, it is said to be coherent scattering. Two
such examples are Rayleigh scattering and Mie scattering.
The basic idea behind coherent scattering is that the electric field of the
incoming wave exerts forces on the electrical charges of an atom, displacing the
centroid of the negative electron cloud in one direction, and the positive nucleus
in the opposite direction, then reversing the process. This creates an oscillating
electric dipole, which emits electromagnetic radiation of its own. The result is a
superposition of the transmitted incident wave, and the scattered wave radiated
by the electric dipole.
Coherent scattering processes are elastic scattering interactions between low-
energy radiation and bound atomic electrons. The atom recoils to conserve
momentum, and the outgoing radiation has almost the same energy as the
incoming radiation, but its direction is deflected by a small angle.
Rayleigh scattering occurs when the wavelength is larger than the target par-
ticle, and Mie scattering occurs when the wavelength is similar to the target. In
the Earth’s atmosphere, Rayleigh scattering occurs between visual wavelength
radiation and gaseous molecules such as O2 , while Mie scattering occurs with
particles of dust or pollen.
When the wavelength is much smaller than the size of the target, then light
can be represented as 1-dimensional rays, and the interaction between light and
matter can be described with geometrical optics. Hence, geometrical optics is
essentially a limit of coherent scattering.
Thomson scattering is another type of coherent, elastic scattering interac-
tion, but in this case it occurs between radiation and an individual atomic
electron, or an unbound electron. The radiation energy is almost unchanged,
but it can scatter in any direction.
In contrast, Compton scattering is an incoherent scattering interaction be-
tween radiation and an atomic electron. The incoming radiation is considered to
be in a particle-like photon state, and the photon transfers some of its energy to
the electron, as well as changing direction. The incoming photon has an energy
much greater than the binding energy of atomic electrons, and whilst the total
energy-momentum of the photon and electron is conserved, the atom is ionized
in the process.
The importance of coherent scattering will become apparent in the next
section when we consider the role played by human visual perception in the
conception of the macroscopic world.
64
8 Decoherence
In this section, we will consider the role of the radiation field in ‘decoherence’, a
process purportedly crucial to understanding the relationship between quantum
theory and the macroscopic world.
First, however, it might be prudent to highlight the potentially confusing
double-use of the term ‘coherent’ within quantum physics. Particularly in the
context of discussions of the measurement problem, a superposition of pure
states is sometimes referred to as a ‘coherent superposition’. This is a different
use of the word, and doesn’t mean that the system is in a coherent state, or
even that the components of the superposition are coherent states. As Okon
and Sudarsky (2016) put it, “it is hard to see what the adjective of ‘coherent’
adds to the standard notion of a superposition.”
The basic idea of decoherence is as follows. When the joint state of a system
and measuring apparatus (or environment) is a superposition in the joint Hilbert
space, one can trace over the degrees of freedom of the measuring apparatus (or
environment) to obtain the reduced state of the system. This reduced state
is represented by a density operator ρS . The proponents of decoherence try
to interpret ρS as a mixed state. They claim that decoherence is a process by
which the interaction Hamiltonian between system and measuring apparatus (or
environment) rapidly diagonalizes the density operator, in a basis determined
by the quantity being measured.
“It is argued that ρS is, for all practical purposes the tool to use in order to
make predictions regarding all possible measurements to be carried out on the
system. And since ρS is identical to a mixed state, the results of all these possi-
ble measurements are going to be identical to those of measurements performed
on a mixed state. That is, for all practical purposes the system will behave
as a mixture. Decoherence, then, is said to lead to effectively nonunitary dy-
namics for the system, which explains the absence of interference between the
components of the superposition,” (Okon and Sudarsky, 2016).
One of the interpretational problems faced by quantum mechanics is to ex-
plain why macroscopic systems, composed entirely of interacting quantum sys-
tems, do not appear to exist in superposed states. In other words, why is there
no apparent interference between macroscopically different states?
The question, however, is posed, explicitly or implicitly, in terms of the
states of matter alone, not the states of the radiation field. Interference effects
have been apparent in the radiation field, long before the advent of quantum
theory, and these interference effects are associated with classical states of the
radiation field, represented either in Maxwellian electromagnetism, or its pre-
decessor, wave optics. Hence the radiation field potentially plays a unique role
in understanding the relationship between the classical and the quantum.
It is crucial to note that because human visual perception of the surrounding
spatial world is based upon detecting a pattern of light in the narrow range of
wavelengths 400 − 700 nm, humans perceive a pattern of radiation which has
primarily been reflected or scattered from matter by coherent scattering pro-
cesses. This bath of visual wavelength radiation can be thought of as a type of
65
environment for the material systems on the Earth, but equally one can think
of the material systems as providing an environment for the radiation. In accor-
dance with the dictums of ‘environmental decoherence’, we can hypothesize that
the continuous coherent scattering interactions between visual wavelength radi-
ation and matter diagonalizes the reduced state of the radiation field in a basis
of coherent states. The polarization of sunlight is one particular manifestation
of this.
Let’s consider a scattering scenario. Suppose we have a material system in
an initial superposition,
ψ = αψx1 + βψx2 ,
where ψx1 represents a ‘coherent’ state of mean position x1 , and ψx2 represents a
‘coherent’ state of mean position x2 , and ||x2 − x1 || is larger than the dispersion
∆xi of either coherent state.
These coherent states are assumed to be the states of a massive fermion, or
an aggregate system which contains massive fermions. We earlier rejected the
possibility of coherent fermionic states, but in that context they were Annihila-
tion Operator Coherent States (AOCS). i.e., they were field-theoretic superpo-
sitions in Fock space of particle-number eigenstates of the same mode. In this
context, the coherent fermionic states are Minimal Uncertainty Coherent States
(MUCS). Specifically, they are non-relativistic states which belong to the same
family as the coherent states of a simple harmonic oscillator originally identi-
fied by Schrödinger in 1926; i.e., they are localized in phase space, in the sense
that the product ∆x · ∆p is minimised, the dispersion of each quantity remains
fixed in time, and the mean values hxi and hpi evolve according to the classical
equations of motion.
Before fully expounding our scattering scenario, a brief digression on the
subject of the harmonic oscillator may be helpful to counter potential anxieties
about the physical prevalence of such systems. Recall that harmonic oscillator
potentials are those with the quadratic form:
1 2
V (x) = kx .
2
In such a potential, the restoring force on a system perturbed from its equilib-
rium state is linearly proportional to the displacement, mẍ = −kx. In contrast,
chemistry is dominated by the Coulomb potential, which has the form:
k
V (x) = − .
x
There are two senses, however, in which the harmonic potential has empiri-
cal relevance. First, electrostatic shielding in systems with positive and neg-
ative charges creates a variety of effective potentials. In the case of diatomic
molecules, for example, the vibrations along the internuclear axis can be repre-
sented by a harmonic potential. Second, the small perturbations of any system
near a local minimum x0 in its potential can be represented by a harmonic po-
tential, irrespective of the global form of the potential. This is easy to see with
66
a Taylor series expansion:
dV (x − x0 )2 d2 V
V (x) = V (x0 ) + (x − x0 ) + + ···
dx 2! dx2
By definition, at the local minimum the gradient vanishes, hence dV /dx = 0.
Thus, removing the offset V (x0 ), for small x one has:
1
V (x) ≈ k(x − x0 )2 .
2
So, given that harmonic potentials are physically prevalent in this sense, the
states of bound fermions close to local minima in their effective potentials, can
be approximately represented as coherent states.
It’s important to note here that these coherent fermion states must be bound
states, because the wave-packet of a free-particle in non-relativistic quantum
mechanics has a notorious tendency to spread. In the non-relativistic quantum
mechanics of a harmonic oscillator one typically has an ‘external’ classical poten-
tial, but this potential is ultimately a quantum field itself. A fermionic system
is bound by virtue of its interactions with other systems, and those interactions
are mediated by gauge bosons. A fermion can only behave as a simple harmonic
oscillator if there is a harmonic potential energy function, in which the restor-
ing force is provided by gauge bosons. Hence, a fermion in a coherent state is
implicitly a fermion which is constrained by the nature of bosonic fields.26
Returning to our scattering scenario, if the system in an initial superposition
ψ = αψx1 +βψx2 , is an aggregate large enough to be macroscopically observable,
this type of state is referred to as a ‘macroscopic superposition’. It’s worthwhile
making explicit what such a state would look like.
R If we 2assumeR that each ψxi is a normalized state of a system of mass m =
m|ψxi | dx = ρ dx, where ρ is the mass density, then each branch of this
superposition contains only a fraction of the total mass. i.e., if ψxi has support
in a bounded open region Σi of volume Vol(Σi ), then
Z
m|ψ|2 dx = |α|2 m
Σ
Z 1
m|ψ|2 dx = |β|2 m .
Σ2
This reduction in the mass density changes the optical depth τ of each volume
Σi of space occupied by the branches of the wave-function. Let κν denote the
opacity at frequency ν. The opacity κν is the sum of the absorption cross-
section σa (ν) and the scattering cross-section σs (ν), per unit mass of the target
particles. For an incident flux I0 , the flux I of light transmitted through an
object of diameter l at the specified frequency ν is given by:
26 An aggregate of fermions is automatically a boson-fermion composite. Specifically, it is a
system of fermions bound together by gauge bosons. However, this in itself is clearly insuffi-
cient to render the system classical. Neutrons, for example, are composed of fermions (quarks)
bound by gauge bosons (gluons), yet neutron behaviour is strongly quantum, as illustrated
by the wavelike beam splitting and re-combination phenomena of neutron interferometry.
67
I
= e−κν ρl = e−τ .
I0
Now, while transparent objects have an optical depth of τ ≈ 0, translucent ones
are defined to have 0 < τ ≤ 4, and opaque objects have 4 < τ , (Lynch and
Livingston, 2001, p267).
Hence, the greater the number of branches in a macroscopic superposition,
and the lower the density associated with each branch, the closer each branch
tends towards representing a translucent object. In other words, a macroscopic
superposition would be visually identifiable not merely by multiple images of
the same object, but by virtue of those multiple images becoming increasing
translucent as the number of branches increases.27
Let’s return to the initial superposition defined above. Suppose that a state
of the radiation field interacts (‘scatters’) off the target material system in a
way which is sensitive to xi . In other words, we have an incoming state φin of
the radiation field, so that the initial state of the combined system is:
ψ ⊗ φin ,
and we have an outgoing state of the radiation field φout , which is such that:
ψx1 ⊗ φin → ψx1 ⊗ φout

1
ψx2 ⊗ φin → ψx2 ⊗ φout
2 .
φin might be a plane wave of wave number k, φout 1 might be a superposition of
a plane wave and an outgoing spherical wave centred on x1 , and φout 2 might be
a superposition of a plane wave and an outgoing spherical wave centred on x2 ,
(see Prugovecki 1981, Sections 4.2 and 4.3). The mapping above between initial
and final states defines the unitary evolution of the interacting system, hence it
implicitly specifies the interaction Hamiltonian.
Now, the linearity of unitary evolution entails that:
ψ ⊗ φin → α (ψx1 ⊗ φout out

1 ) + β (ψx2 ⊗ φ2 ) .
In this event, the state of the radiation field has become ‘entangled’ with the
state of the material system. The reduced density operator of the material
system after the interaction has the following matrix elements in the ψxi basis:
|α|2 αβ ∗ hφout out

2 , φ1 i
ρ= .
α∗ βhφout out
1 , φ2 i |β|2
When hφout out

2 , φ1 i ' 0, the off-diagonal ‘interference’ terms almost vanish due
to the individual scattering event. This is the case when k||x2 − x1 || 1. i.e.,
27 The other spatial senses would also detect macroscopic superpositions. The reduction in
density in the region of space occupied by each branch would reduce the resistance to touch,
given that the pressure of a solid ultimately derives from: (i) electrostatic Coulomb repulsion
between negatively charged electrons; and (ii) the Pauli exclusion principle.
68
when the wavelength λ of the incoming radiation state is smaller than the spa-
tial separation of the superposed coherent states. In this case, the individual
scattering event will induce a reduced state of the material system which gen-
erates the same subsequent pattern of measured values as a mixture of the two
coherent states.
When hφout out
2 , φ1 i ' 1, the individual scattering event does not destroy
the interference terms, but does make a contribution towards diagonalizing the
reduced density operator. This is the case where k||x2 − x1 || 1. i.e., when the
wavelength λ of the incoming radiation state is longer than the spatial separation
of the superposed coherent states. In this case, the systems only become weakly
entangled, and the reduced state of the material system generates a similar
subsequent pattern of measured values as a pure state superposition of the two
coherent states. However, repeated scattering events due to a flux of incoming
radiation will drive the diagonal elements towards zero.28 On the assumption of
such a flux F , expressions can be derived which show exponentially fast decay
of the interference terms, (Wallace 2012, p80; Joos 2003, p65-66):
exp(−Λt(|x2 − x1 ||)2 ) ,
where
Λ ∼ k2 F σ ,
with σ denoting the total interaction cross-section.
These expressions apply to decoherence in general, not merely that induced
by scattering with the radiation field. For a 10-micron dust particle suspended in
the atmosphere, Λ ∼ 1036 m−2 s−1 due to scattering with atmospheric molecules,
1021 m−2 s−1 due to the flux of sunlight, and 106 m−2 s−1 due to the flux of
cosmic microwave background radiation. Joos (p67) estimated a decoherence
timescale of order 10−13 s due to scattering of air molecules off such a dust par-
ticle. Hence, the scattering between boson-fermion composites in a gas or fluid
is most effective for decoherence. Radiation (or even possibly the gravitational
field) will be the prime agent for material systems not immersed in a gas or
fluid.
As Rosaler points out, “[the spreading of fermionic wave-packets] will result
in branching as the coherent superposition that results from the spreading is
decohered by interaction either with the electromagnetic field or with other
fermions. . . the factors that affect the rate of bosonic wave packet spreading are
simply those that affect fermionic wave packet spreading, since bosonic packets
do not tend to spread of their own accord but only indirectly by virtue of
their interaction with fermionic degrees of freedom (for example, a fermionic
state consisting of two widely separated wave packets for a single particle will
tend to generate a superposition of very different classical electromagnetic field
28 Nevertheless, the reduced state is a density operator rather than one of the pure coherent
states, and as Okon and Sudarsky (2016) point out, one needs to apply the Born rules to the
reduced state to obtain a predicted distribution of measured values, the very same Born rules
which the decoherence programme seeks to explain.
69
configurations, where each field configuration can be regarded as being generated
by a different one of the quasi-classical fermionic wave packets,” (p199-200)
Observers are special types of composite boson-fermion systems, hence the
states of observers can become entangled with the rest of the world. The human
concept of the macroscopic world is founded largely upon the focusing and
absorption of optical wavelength radiation, but also upon other spatial senses
such as touch and proprioception. In the visual case the total state space is:
Hmicro ⊗ Hmacro ⊗ Hrad ⊗ Hobserver .
Visual observations are those which occur by virtue of the ambient radiation
field becoming entangled with the state of composite boson-fermion systems,
and the boson-fermion observer state becoming entangled with the state of the
radiation field. In the case of the other spatial senses, the state space is:
Hmicro ⊗ Hmacro ⊗ Hobserver ,
and the perceptions are those which occur by virtue of the boson-fermion ob-
server state becoming directly entangled with the state of the surrounding
macroscopic (‘collective’) boson-fermion degrees of freedom.
According to decoherence, we don’t visually perceive macroscopic superpo-
sitions because the macroscopic boson-fermion systems (or degrees of freedom)
become entangled with the microscopic systems (or degrees of freedom), and
then the visual-wavelength radiation field becomes entangled with the macro-
scopic systems (or degrees of freedom). Hence, the coherent scattering interac-
tions between visual-wavelength radiation and material systems play a crucial
role in the perception that the world is classical at a macroscopic level. But the
ultimate import of quantum theory is to reveal that the universe is an evolving
interference pattern.
70
9 Conclusions
The division between the classical and the quantum is less clear-cut in the case
of light. Specifically, we have found the following:
• Coherent states are the best candidates for pure classical states of the
radiation field.
• Classical states of the radiation field are not emergent from the quantum
states.
• Coherent states provide a basis for the photonic Fock space, hence any
state of the quantized radiation field can be expressed as a superposition
of classical states. Conversely, any coherent state can be expressed as a
superposition of photon-number eigenstates, which are quantum states of
the radiation field.
• To reconcile the quantum theory of the radiation field with macroscopic
phenomena, not only is it unnecessary to eradicate superpositions, but on
the contrary they are indispensable for the emergence of classical phenom-
ena.
• The vacuum state of any boson field can be decomposed as a superposition
of coherent states. This potentially provides an explanation for why there
is something classical rather than nothing classical. However, the existence
of fermions implies that some form of vacuum decay took place.
• The classical states of the radiation field do not emerge in the limit as the
number of photons tends to infinity.
• The class of coherent states is transformed into non-classical ‘squeezed’
states under a transition to an accelerated reference frame. Hence, the
classicality of the radiation field is not reference-frame independent.
• Ordinary sunlight is a mixed state of the radiation field, which is polar-
ized by ‘coherent scattering’ with atmospheric molecules and non-metallic
surfaces. These coherent scattering processes select a basis of coherent
states for the decomposition of the mixed state.
• Decoherence is most effective when a massive system is immersed in a
gas or fluid of other massive systems. However, human visual perception
is dependent upon another level of environmental decoherence, in which
visual wavelength radiation becomes entangled with matter by means of
coherent scattering processes.
• A fermion in a coherent state is a fermion which is constrained by the
nature of bosonic fields. An aggregate of fermions large enough to be
macroscopically observable is automatically a boson-fermion composite, a
system bound together by gauge bosons. The apparent classicality of the
macroscopic world seems ultimately to be dependent upon boson fields.
71
References
[1] Aitchison, I.J.R. (1985). Nothing’s Plenty. The vacuum in modern quantum
field theory, Contemp. Phys., vol 26, No. 4, pp333-391.
[2] Albert, D.Z. (1988). On the Possibility That the Present Quantum State
of the Universe is the Vacuum, Proceedings of the Biennial Meeting of the
Philosophy of Science Association Vol. 1988, Volume Two: Symposia and
Invited Papers, pp127-133.
[3] Argyris, J. and Ciubotariu,C. (1998). A limit on the longitudinal graviton
mass, in Causality and locality in modern physics, Hunter G., Jeffers S.,
Vigier J.P. (eds). Springer.
[4] Ashtekar, A. and Geroch, R. (1974). Quantum Theory of Gravitation,
Rep.Prog.Phys. 37, pp1211-1255.
[5] Baker, D.J. (2018). Interpreting Supersymmetry. http://philsci-
archive.pitt.edu/15119/
[6] Berestetskii, V.B., Lifshitz, E.M., and Pitaevskii, L.P. (1982). Quantum
Electrodynamics, 2nd Edition. Oxford: Pergamon Press.
[7] Bishop, R.F. and Vourdas, A. (1986). Generalised coherent states and Bo-
goliubov transformations, Journal of Physics A: Mathematical and General,
Volume 19, Number 13, pp2525-2536.
[8] Blau, S.K. and Guth, A.H.(1987). Inflationary Cosmology. In S.Hawking
and W.Israel (eds.), 300 Years of Gravitation, (pp524-603). Cambridge Uni-
versity Press.
[9] Dautray, R., and Lions, J-L. (1988). Mathematical Analysis and Numerical
Methods for Science and Technology, Chapter IX, §1, 6-7, Springer-Verlag.
[10] Derdzinski, A. (2002). Quantum Electrodynamics and Gauge Fields for

Mathematicians, unpublished manuscript, Department of Mathematics,
The Ohio State University, September 2002.
[11] Duncan, A. (2012). The conceptual framework of quantum field theory, Ox-
ford University Press.
[12] Earman, J. (1995). Bangs, Crunches, Whimpers and Shrieks: Singularities

and Acausalities in Relativistic Spacetimes, Oxford University Press.
[13] Earman, J., Fraser, D. (2006). Haag’s theorem and its implications for the
foundations of quantum field theory, Erkenntnis Vol. 64, No. 3, pp305-344.
[14] Field, T.R. and Hughston, L.P. (1999). The geometry of coherent states,
Journal of Mathematical Physics 40, pp2568-2583.
72
[15] Folland, G.B. (2008). Quantum Field Theory - A tourist guide for mathe-
maticians, Americal Mathematical Society.
[16] Frankel, T. (1979). Gravitational curvature, W.H.Freeman.
[17] Fraser, D. (2008). The fate of ‘particles’ in quantum field theories with
interactions, Studies in History and Philosophy of Modern Physics, Volume
39, Issue 4, pp841-859.
[18] Gazeau, J-P. (2009). Coherent states in quantum physics, Wiley.
[19] Greiner, W. and Reinhardt, J. (1996). Field Quantization, Springer.
[20] Grensing, G. (2013). Structural Aspects of Quantum field theory, Volume
1, World Scientific.
[21] Guillemin, V. and Sternberg, S. (1984). Symplectic Techniques in Physics,
Cambridge University Press.
[22] Haag, R. (1996). Local quantum physics - Fields, Particles, Algebras,
Springer.
[23] Healey, R. A. (2013). Physical composition, Studies in History and Phi-
losophy of Science Part B - Studies in History and Philosophy of Modern
Physics, Volume 44, Issue 1, pp48-62.
[24] Honegger, R. and Rieckers, A. (2015). Photons in Fock space and beyond,
World Scientific.
[25] Joos, E. (2003). Decoherence Through Interaction with the Environment,
pp. 41-180, in Decoherence and the Appearance of a Classical World in
Quantum Theory, Joos, E., Zeh, H.D., Kiefer, C., Giulini, D.J.W., Kupsch,
J., Stamatescu, I.-O. Springer.
[26] Keller, O. (2014). Light - The physics of the photon. CRC Press.
[27] Knight, P. (1989). Quantum optics, pp289-315, in The New Physics, ed.
Paul Davies, Cambridge University Press.
[28] Kolb, E.W. and Turner, M.S. (1990). The early universe. Addison-Wesley.
[29] Le Tiec, A. and Novak, J. (2016). Theory of gravitational waves, to appear
in An Overview of Gravitational Waves: Theory and Detection, G. Auger
and E. Plagnol (eds). World Scientific.
[30] Lynch, D.K., and Livingston, W.C. (2001). Color and Light in Nature, 2nd
Edition. Cambridge University Press.
[31] Mandl, F. and Shaw, G. (2010). Quantum field theory, 2nd Edition. Wiley.
[32] Manin, Y.I. (1988). Gauge Field Theory and Complex Geometry, Berlin:
Springer-Verlag.
73
[33] Mari, A., de Palma, G, and Giovannetti, G. (2016). Experiments testing
macroscopic quantum superpositions must be slow - Supplementary infor-
mation, Scientific Reports volume 6, Article number: 22777.
[34] McCabe, G. (2007). The structure and interpretation of the standard model,
(Philosophy and Foundations of Physics Volume 2). Elsevier.
[35] Okon, E. and Sudarsky, D. (2016). Less Decoherence and More Coherence
in Quantum Gravity, Inflationary Cosmology and Elsewhere, Found. Phys.
46, pp852-879.
[36] Prugovecki, E. (1981). Quantum Mechanics in Hilbert space, 2nd Edition.
Academic Press.
[37] Prugovecki, E. (1995). Principles of Quantum General Relativity. World
Scientific.
[38] Reed, M., Simon, B. (1975). Methods of Modern Mathematical Physics Vol
II: Fourier Analysis, Self-Adjointness. Academic Press.
[39] Rosaler, J. (2018). Inter-Theory Relations in Physics: Case Studies from
Quantum Mechanics and Quantum Field Theory, University of Oxford
DPhil Thesis.
[40] Rugh, S.E., and Zinkernagel, H. (2002). The quantum vacuum and the cos-
mological constant problem, Studies in History and Philosophy of Modern
Physics, Volume 33, Issue 4, pp663-705.
[41] Saunders, S. (1991). The negative energy sea, in Philosophy of Vacuum, S.
Saunders and H. Brown (eds.), pp65-109. Clarendon Press.
[42] Sakurai, J.J. (1967). Advanced quantum mechanics, Addision-Wesley.
[43] Schrödinger, E. (1926). Der stetige Übergang von der Mikrozur
Makromechanik, Naturwissenschaften 14, pp664-666.
[44] Teller, P. (1988). Three problems of renormalization, pp73-89 in Philo-
sophical Foundations of Quantum Field Theory, Harvey R.Brown and Rom
Harre (eds.), Clarendon Press.
[45] Teller, P. (1995). An Interpretive Introduction to Quantum Field Theory.
Princeton University Press.
[46] Troup, G. J. and Perlman, H. S. (1981). Coherent states and Fock states at
very high average photon numbers, Australian Journal of Physics, vol. 34,
no. 4, pp357-360.
[47] Wallace, D. (2012). The Emergent Multiverse, Oxford University Press.
[48] Weingard, R. (1991). Making everying out of nothing, pp197-216, in
The Philosophy of Vacuum, Simon Saunders and Harvey R.Brown (eds.),
Clarendon Press.
74
[49] Wightman, A.S. (1973). Relativistic wave equations as singular hyperbolic
systems, in Proc. Symp. in Pure Math. Vol. XXIII, Berkeley 1971. AMS.
[50] Zeldovich, Y.B. (1966). Number of Quata as an Invariant of the Classical
Electromagnetic Field, Soviet Physics Doklady, 10, pp771-772.
75

Light

Uploaded by

Copyright:

Available Formats

Light

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Light

Uploaded by

Copyright:

Available Formats

What is light?

3 Quantum Field Theory 11

4 Interactions between light and matter 22

6 Why is there something classical rather than nothing classical? 46

7 Coherent states and the classical world 56

ex = cos θex0 + sin θey0

w rw eiθw rw i(θw −θz )

Ψei(k·r−ωt) = (c1 e1 + c2 e2 )ei(k·r−ωt) .

Re[E0 (ex + iey )ei(k·r−ωt) ] = E0 ex cos(k · r − ωt) + E0 ey cos(k · r − ωt + π/2)

Re[E0 (ex − iey )ei(k·r−ωt) ] = E0 ex cos(k · r − ωt) + E0 ey cos(k · r − ωt − π/2)

This defines the anti-clockwise rotation of a vector of length E0 around a circle,

0.5Pe+ + 0.5Pe− = 0.5Pex + 0.5Pey .

Direct sunlight, or sunlight which has undergone numerous scattering interac-

then the degree of polarization is the contrast,

for simple tensors,

φ(f ) = he1 , f ip1 · · · hek , f ipk .

φ(f ) = he1 , f ihe4 , f i2 = he1 ⊗ e4 ⊗ e4 , f ⊗ f ⊗ f i .

The mapping can, of course, be restricted to the symmetric or anti-symmetric

a(f )a(g) − a(g)a(f ) = 0

a(f )a(g) + a(g)a(f ) = 0

a(f )a∗ (g) + a∗ (g)a(f ) = hf, giI .

1. Ω is the ground state of the quantum field. It is the minimum energy

where {ε(0, i ) : i = 0, 1, 2, 3} is an orthonormal basis for C4 .

This entails that a state which contains a longitudinal photon of momentum

H˜Lorenz = {f˜(k) ∈ H˜K : f˜0 (k) = f˜3 (k)} .

There is a further subspace H˜0 ⊂ H˜Lorenz such that:

H˜0 = {f˜(k) ∈ H˜Lorenz : f˜1 (k) = f˜2 (k) = 0} .

H˜⊥ = {f˜(k) ∈ H˜ + : f˜0 (k) = f˜3 (k) = 0} .

By taking the quotient,

H˜phys = H˜Lorenz /H˜0 ,

one obtains the physical one-particle space, equipped with a positive-definite

Hence, in the Coulomb gauge, the 3-vector potential is purely transverse, A =

4.2 Scattering, Feynman diagrams and virtual particles

with t(xi1 ) ≥ t(xi2 ) ≥ · · · ≥ t(xin ). The interaction Hamiltonian density op-

the permutation into the expression for the time-ordering.

Fe,e+ = F (Hm+e ,1/2 ⊕ Hm− + ,1/2 ) ,

Fe,e+ = F (Hm+e ,1/2 ) ⊕ F (Hm− + ,1/2 ) .

where q is the charge of the electron, and γ µ is a Dirac gamma matrix.

hψout | Sψin i = hψout | Iψin i + hψout | S1 ψin i + hψout | S2 ψin i + · · · .

a∗λi (ki ) ⊗ bb∗ri (pi )(Ωγ ⊗ Ωe,e+ )

a∗λo (ko )bb∗ro (po )Ω | S b

Let’s consider the second-order term:

The interaction Hamiltonian density operator is:

7 With ω(pi ) = +(m2e + kpi k2 )1/2

denote the 4-momentum.

hΩ | · · · b a∗j · · · Ωi = c(i, j)hΩ | · · · Ωi ∓ hΩ | · · · b

The technique extends to field operators, so that Ψb + (x2 ), the positive-frequency

This reduces to:

c(ko , x1 )c(po , x1 )c(x1 , x2 )c(x2 , pi )c(x2 , ki )hΩ | Ωi .

The Feynman diagrams beloved of textbooks on quantum field theory, offer a

4.3 The interacting field vacuum

shown here for the case of a self-interacting scalar field.

4.4 Do longitudinal photons exist?

hΩ|T [A bν (y)]Ωi = Dµν (x − y) ,

DFµν (k) = DFµν(trans) (k) + DFµν(Coul) (k) + DFµν(resid) (k) .

The decomposition is both gauge-dependent, and Lorentz-frame dependent.

“Longitudinal and scalar photons are not observed as free par-

However, a commitment to the existence of longitudinal and scalar photons,

which Fourier transforms into:

where {ε(0, i ) : i = 0, 1, 2, 3} is an orthonormal basis for C4 .

hψE0 ⊗ (1k, , 0, . . .), A

|ni = fk, ··· fk, (n times) .