SWAT 96/56

April, 1996.

The Schrödinger Wave Functional

and Vacuum States in Curved Spacetime

arXiv:hep-th/9605004v1 1 May 1996

D.V. Long1 and G.M. Shore2

Department of Physics
University of Wales, Swansea
Singleton Park
Swansea, SA2 8PP, U.K.

Abstract
The Schrödinger picture description of vacuum states in curved space-
time is further developed. General solutions for the vacuum wave func-
tional are given for both static and dynamic (Bianchi type I) spacetimes
and for conformally static spacetimes of Robertson-Walker type. The for-
malism is illustrated for simple cosmological models with time-dependent
metrics and the phenomenon of particle creation is related to a special
form of the kernel in the vacuum wave functional.

1
[email protected]
2
[email protected]
1 Introduction
Many of the unusual and potentially paradoxical phenomena which occur in quan-
tum field theories in curved spacetimes reflect the nature of the vacuum state, for
a review see for example [1, 2, 3, 4]. However, whereas in Minkowski spacetime
we have a simple and unambiguous prescription for determining the vacuum, the
specification of the vacuum state in a general curved spacetime involves many
subtleties. Indeed, it is not assured that a state satisfying the defining attributes
of the Minkowski vacuum even exists for a general spacetime.
In this and a companion paper [5], we discuss the nature of the vacuum
state for a broad class of flat and curved spacetimes and develop a formalism
of sufficient generality to allow a future application to the vacuum structure of
black hole spacetimes and the associated Hawking radiation [6, 4].
This formalism is based on the Schrödinger wave functional. The essential
reason for this choice is to avoid the conventional description of the vacuum
as the ‘no-particle’ state in a Fock space of states generated by the creation
operators of particles defined with respect to a particular mode decomposition of
the quantum field. This approach is inherently problematic in curved spacetime,
where there is no uniquely favoured mode decomposition and no guarantee that
the usual concept of a particle is a good description of the spectrum of the theory.
The Schrödinger wave functional circumvents this second problem by giving an
intrinsic description of the vacuum without reference to the spectrum of excited
states.
The wave functional that defines the vacuum state satisfies a functional Schrö-
dinger equation, which presupposes a foliation of spacetime into a series of space-
like hypersurfaces, the Schrödinger equation describing the evolution of the wave
functional between successive hypersurfaces. The wave functional therefore de-
pends on the choice of foliation.3 Our expectation is that physical quantities
should be independent of this foliation. However, intermediate quantities such
as Green functions may have a foliation dependence. This dependence should
be controlled by a functional Ward identity (somewhat analogous to the depen-
dence of Green functions on the choice of gauge in ordinary quantum field theory).
Physical quantities should depend only on the spacetime manifold and boundary
conditions.
3
Notice that the choice of foliation is quite distinct from a choice of coordinates. The wave
functional is independent of any particular coordinates used to describe either the spacetime or
the foliation (see section 3). In the Schrödinger formalism, the choice of foliation replaces the
mode decomposition in the conventional approach. One of the advantages of this formalism is
the clear separation it provides between the choice of vacuum state and the issue of coordinate
independence.

1
 In this paper, which is intended as the first step in a general programme, we
consider a class of globally hyperbolic spacetimes, viz. manifolds with the topol-
ogy M = R × Σ where Σ are spacelike hypersurfaces. Within this category, we
distinguish manifolds which are ‘static’ and ‘dynamic’, essentially implying that
the metric depends only on the spacelike and timelike coordinates respectively.
The class referred to as ‘dynamic’ is in fact the Bianchi type I spacetimes. We
also consider more general conformally static spacetimes where the conformal
factor is purely time dependent.This includes the cosmologically important class
of Robertson-Walker spacetimes.
The ultimate goal of our work is a full analysis of the physics of the vac-
uum state in the Schwarzschild and Kruskal spacetimes describing collapsed and
eternal black holes. In this and the companion paper, we develop all the nec-
essary theoretical formalism and, as a warm-up, illustrate the techniques on a
number of simple illustrative examples. The solutions for static and dynamic
spacetimes are sufficient to encompass the full Kruskal metric and we check ex-
plicitly that subtleties involving foliation dependence and the interpretation of
vacuum states are under control. The extension to black hole spacetimes is in
principle a straightforward application of these techniques, although the presence
of real event horizons separating the static and dynamic regions is an important
new feature.
A number of the results and techniques described here can be found already
elsewhere in the literature. In these cases, we have referenced the original works
as fully as possible, but have included the material as we have been concerned to
present a coherent and systematic study of the Schrödinger wave functional.

The paper is organised as follows. In section 2 we review the Schrödinger

picture formulation of quantum field theory in Minkowski spacetime and derive
the Schrödinger equation and its solutions for the vacuum and excited states for
a free scalar field. We also discuss some properties of a general Gaussian state
which will be needed in sections 4 and 5.
Then in section 3 we construct the Schrödinger wave functional equation for an
arbitrary foliation of a globally hyperbolic spacetime with topology M = R × Σ.
This makes explicit the dependence of the equation, and therefore the vacuum
wave functional, on the ‘lapse’ and ‘shift’ functions specifying the foliation. The
simplification for special coordinate choices which mirror the foliation is shown.
We find explicit solutions for the vacuum wave functional for the special cases
of static and dynamic spacetimes and a particular class of conformally static
metrics including the Robertson-Walker spacetimes. This analysis shows clearly
that while there is frequently an essentially unique vacuum state for a static
spacetime, in the dynamic case there is necessarily a one-parameter family of
vacuum solutions of the Schrödinger equation.

2
 In sections 4 and 5, these techniques are illustrated using two simple two-
dimensional ‘cosmological models’ which have been much discussed in the liter-
ature. The first represents an asymptotically static universe that undergoes a
period of expansion while the second describes a universe which expands from an
initial singularity before recollapsing. These are of particular interest as models
illustrating cosmological particle creation. We solve the Schrödinger wave func-
tional equation exactly to determine the candidate vacuum states and discuss
carefully the interpretation that particles are created by the expansion.
A number of technical results on solutions and Green functions for various
differential equations occurring in the text are given in the appendices.

2 Schrödinger Picture in Minkowski Spacetime

To introduce the Schrödinger formalism, we begin by reviewing the wave func-
tional equation and its solutions in Minkowski spacetime. (For a review see, e.g.
[7].) Consider a massive scalar field ϕ(x) with action4
Z
S= dd+1 x L (1)

where the Lagrangian density is

1 n µν o
L= η (∂µ ϕ)(∂ν ϕ) − m2 ϕ2 (2)
2
The canonical momentum conjugate to the field is defined by
∂L
π(x) = = ∂0 ϕ(x) (3)
∂(∂0 ϕ)
The Hamiltonian is constructed by a Legendre transform of the Lagrangian
Z
H(π, ϕ) = dd x [π ϕ̇ − L] (4)
1Z d n 2 o
= d x π − η ij (∂i ϕ)(∂j ϕ) + m2 ϕ2 (5)
2
The system can now be canonically quantised by treating the fields as oper-
ators and imposing appropriate commutation relations. This involves the choice
of a foliation of the spacetime into a succession of spacelike hypersurfaces. In
Minkowski spacetime, it is natural to choose these to be the hypersurfaces of
fixed t and impose the equal-time commutation relations

[ϕ(x, t), π(y, t)] = iδ d (x − y)

4
Our conventions are as follows. For generality we consider a spacetime of d + 1 dimensions.
The metric signature is (+, −, . . . , −). Coordinates for the spacetime point x are denoted either
by (x0 , xi ), where i = 1, . . . , d, or by (t, x).

3
 [ϕ(x, t), ϕ(y, t)] = [π(x, t), π(y, t)] = 0 (6)
In the Schrödinger picture, we take the basis vectors of the state vector space
to be the eigenstates of the field operator ϕ(x) on a fixed t hypersurface, with
eigenvalues φ(x),
ϕ(x)|φ(x), ti = φ(x)|φ(x), ti (7)
Notice that the set of field eigenvalues φ(x) is independent of the value of t la-
belling the hypersurfaces. This will be especially important in the curved space-
time context.
In this picture, the quantum states are explicit functions of time and are
represented by wave functionals Ψ[φ(x), t]. Operators O(π, ϕ) acting on these
states may be represented by
   δ 
O π(x), ϕ(x) ∼ O −i , φ(x) (8)
δφ(x)

consistent with the commutation relations (6). The Schrödinger equation, which
governs the evolution of the wave functional between the spacelike hypersurfaces,
is
∂Ψ[φ, t]  δ 
i = H −i , φ(x) Ψ[φ, t]
∂t δφ(x)
Z n δ2 o
1
= dd x − 2 − η ij (∂i φ)(∂j φ) + m2 φ2 Ψ[φ, t] (9)
2 δφ

To solve this equation, we make the ansatz that, up to a time-dependent

phase, the vacuum wave functional is a simple Gaussian. This is in direct analogy
with the standard solution of the harmonic oscillator in quantum mechanics. We
therefore write
 Z Z 
1
Ψ0 [φ, t] = N0 (t) exp − dd x dd y φ(x) G(x, y) φ(y) = N0 (t)ψ0 [φ] (10)
2
Substituting into (9) and comparing coefficients of [φ0 ] determines N0 (t) from
Z
d ln N0 (t) i
=− dd x G(x, x) (11)
dt 2
while comparing coefficients of [φ2 ] gives an equation for the kernel G(x, y):
Z
dd z G(x, z)G(z, y) = (−∇2 + m2 )δ d (x − y) (12)

The kernel equation can be solved by using several methods, either by working
in momentum space, or coordinate space, or using the Schwinger-DeWitt repre-
sentation of the propagator [8, 9].

4
 We show here the momentum space approach, which is the most convenient
in Minkowski spacetime. Introducing the Fourier transform of the kernel,
Z
dd k ik.(x−y) e
G(x, y) = e G(k) (13)
(2π)d

the kernel eq.(12) reduces to

e
G(k) 2
= k 2 + m2 (14)

The inverse of the kernel, ∆(x, y), is defined as

Z
dd z G(x, z)∆(z, y) = δ d (x − y) (15)

and is therefore given explicitly by

Z
dd k eik.(x−y)
∆(x, y) = q (16)
(2π)d k 2 + m2

The kernel is simply related to the Green functions of the theory (see appendix
B). A general Green function can be written as
Z ′
dd+1 k eik.(x−x ) − ik0 (x0 −y0 )
G(x, y) = (17)
(2π)d+1 k02 − ω 2

where ω 2 = k 2 + m2 . The k0 integral has poles at k0 = ± ω and can be evaluated

with different boundary conditions, determining how the contour goes round the
poles. In particular, the Wightman function G+ (x, y) is defined by choosing the
contour integral to close around the k0 = ω pole as in Fig(3) of appendix B. For
equal times, this Green function reduces to
Z
1 dd k eik.(x−y)
G+ (x, y)|ET = q (18)
2 (2π)d k 2 + m2
!d−1
1 m 2 2
= K d − 1 (m|x − y|) (19)
2π 2π|x − y| 2 2

The inverse kernel is therefore simply twice the Wightman function evaluated at
equal times, i.e.
∆(x, y) = 2G+ (x, y)|ET (20)
The reason for this identification is made clear in appendix C where we discuss
the representation of vacuum expectation values and two-point functions in the
Schrödinger picture.

5
 The kernel itself is
1
G(x, y) = G+ (x, y)|−1
ET (21)
2
Z
dd k q 2
= d
k + m2 eik.(x−y) (22)
(2π)
!d+1
m 2 2
= −2 K d + 1 (m|x − y|) (23)
2π|x − y| 2 2

The time dependent phase factor in the vacuum wave functional is

N0 (t) = e−iE0 t (24)

where
Z
1
E0 = dd x G(x, x) (25)
2Z
q
1
= dd k k 2 + m2 δ d (0) (26)
2
is the (divergent) ground state energy.

There are several ways to derive the excited state solutions to the Schrödinger
wave functional equation, but with a view to the curved spacetime generalisation,
we start with a method that does not presuppose a particle interpretation of the
excited states.
Making the following ansatz for the first excited state
Z Z
d
Ψ1 [φ, t] = 2 N1 (t) d x dd y φ(x)G(x, y)f (y) ψ0 [φ] = N1 (t)ψ1 [φ] (27)

the Schrödinger equation is solved with the same kernel (12) as for the ground
state, provided
N1 (t) = e−iE1 t (28)
where the energy gap ∆E1 = E1 − E0 satisfies
Z
∆E1 dd y G(x, y)f (y) = (−∇2 + m2 )f (x) (29)

This is readily solved if f (x) is an eigenfunction of the operator (−∇2 + m2 ). We

therefore choose fk (x) = eik.x , which has an eigenvalue of (k 2 + m2 ). We then
find Z
∆E1 dd y G(x, y)fk (y) = (k 2 + m2 )fk (x)
which reduces to
e
∆E1 G(k) = k 2 + m2

6
so the energy gap is q
∆E1 = k 2 + m2 = ω (30)
Clearly this admits a particle interpretation where the first excited state Ψ1 is
a one-particle
q state and E1 is its energy. This is apparent from the energy gap,
2
∆E1 = k + m2 , which is the energy of one particle with momentum k. The
wave functional carries a label k, introduced through the eigenfunction fk , which
specifies the momentum of the particle.
The Schrödinger equation can also be solved directly (as in the quantum
mechanical harmonic oscillator) for an n-particle state, giving

Ψn [φ, t] = e−iEn t Hn [φ]ψ0 [φ] (31)

where the Hn are generalised Hermite polynomials defined using the kernel and
eigenfunctions fkj contracted with the field φ as in (27). The energy of the nth
excited state is simply
n
X
En = E0 + ω(ki ) (32)
i=1

where k i is the momentum of the ith particle.

The excited states can also be described using creation and annihilation op-
erators. We define
Z "Z #
d d δ
â(k) = d x d y G(x, y)φ(y) + f ∗ (x) (33)
δφ(x) k
Z " #
δ
= dd x ωφ(x) + f ∗ (x) (34)
δφ(x) k

and
Z "Z #
† d d δ
â (k) = d x d y G(x, y)φ(y) − fk (x) (35)
δφ(x)
Z " #
d δ
= d x ωφ(x) − fk (x) (36)
δφ(x)

consistent with the commutation relations

[â(k) , ↠(p)] = 2 (2π)d ω δ d (k − p) (37)

These can be used iteratively to construct all the excited states, starting from

â(k)ψ0 = 0 ↠(k)ψ0 = ψ1 (38)

The nth excited state is labelled by the n particle momenta, viz.

ψn (k 1 , . . . , k n ) = a† (k 1 ) . . . a† (k n )ψ0 (39)

7
 Orthogonality relations between these states follow directly from the commu-
tation relations of the creation and annihilation operators. We have
n
nY
hψn (k 1 , . . . , k n )|ψm (p1 , . . . , pm )i = δmn 2ωi (2π)d δ d (k i − pi )
i=1
o
+ . . . n! perms hψ0 |ψ0 i (40)

where the inner product is defined by

Z
hΨ1 |Ψ2 i = Dφ Ψ∗1 (φ)Ψ2 (φ) (41)
q
and ωi = k 2i + m2 . The Hamiltonian and particle number operators are written
in terms of a and a† as usual as
Z
1 dd k h † †
i
H= â (k)â(k) + â(k)â (k) (42)
4 (2π)d
and Z
dd k 1 †
N= â (k)â(k) (43)
(2π)d 2ω
It will often be convenient to express the wave functional for the vacuum
e
and excited states in terms of the Fourier transform φ(k) of the field eigenvalues
defined in eq.(7). We define
Z
dd k ik.x e
φ(x) = e φ(k) (44)
(2π)d

and since φ(x) is real, φe∗ (k) = φ(−k).

e The vacuum wave functional is therefore
( )
e 1 Z dd k e 2
ψ0 [φ] = exp − d
ω |φ(k)| (45)
2 (2π)
e
In terms of φ(k), the creation and annihilation operators simplify:
" #
δ e
â(k) = ω φ(k) + (46)
δ φe∗ (k)
" #
† e∗ δ
â (k) = ω φ (k) − e (47)
δ φ(k)
e
where δ φ(p)/δ e
φ(k) = (2π)d δ d (k − p) and δ φe∗ (p)/δ φ(k)
e = (2π)d δ d (k + p) . The
excited states for particles with momenta k i are therefore given by

ψ1 = 2[ω1 φe∗ (k 1 )]ψ0 (48)

n o
ψ2 = 4[ω1 φe∗ (k 1 )][ω2 φe∗ (k 2 )] − 2ω1 (2π)d δ d (k 1 + k 2 ) ψ0 (49)

8
etc. The general n-particle wave functional is expressed in terms of generalised
Hermite polynomials with n momentum labels, viz.
e t; k , . . . , k ] = e−iEn t H [φ;
Ψn [φ, e k , . . . , k ]ψ [φ]
e
1 n n 1 n 0

n " #
[2] n
X m
Y
−iEn t m n−m d
= e (−1) 2 ω2i−1 (2π) δ(k 2i−1 + k 2i )
m=0 i=1
 
n
Y   o
n
 ωj φe∗ (k j ) + . . . (2m − 1)!! 2m
perms ψ0 (50)
j=2m+1

 
n
where there are (2m − 1)!! 2m
permutations of the pairing of momenta at each
h i
stage of the sum. The sum is from zero to n2 - which is the largest integer ≤ n2 .
Q
Products bi=a with b < a are defined to be 1.

All this analysis is based on the specific foliation of Minkowski spacetime in

which the family of spacelike hypersurfaces Σ are equal-time hypersurfaces and
∂
the Hamiltonian evolution is along the timelike Killing vector field ∂t . While this
is convenient, it is by no means a unique choice. In section 3, we set up the
Schrödinger equation for an arbitrary foliation in curved spacetime. This formal-
ism may of course be specialised to Minkowski spacetime by simply replacing the
metric gµν by ηµν throughout.

2.1 General Gaussian state

In the discussion of the cosmological models in sections 4 and 5 we encounter
Gaussian states where the kernel is time dependent and not simply equal to ω,
the eigenvalue of the wave operator. In anticipation of this, we show here how
such a state in Minkowski spacetime may be expressed as a superposition of
excited many-particle states.
Consider a solution of the Schrödinger equation of the form
( Z )
e t] = N(t) exp − 1 dd k e 2
Ψ[φ, Ω(k; t) |φ(k)| (51)
2 (2π)d

where ( )
i Z t Z dd k
N(t) = exp − dt Ω(k; t) (52)
2 (2π)d
and Ω(k; t) is a solution of

∂Ω(k; t)
i = ω 2 − Ω2 (53)
∂t

9
This can be written as a linear superposition of n-particle states as follows:
∞
" n Z #
X Y dd k i
Ψ= cn (k 1 , . . . , k n )ψn (k 1 , . . . , k n ) (54)
n=0 i=1 (2π)d

where the (time-dependent) expansion coefficients are

R e ∗ (k , . . . , k )Ψ
D φψ n 1 n
cn (k 1 , . . . , k n ) = Q (55)
n! [ ni=1 2ωi ]

and we are assuming constant normalisation factors are included such that hψ0 |ψ0 i
and hΨ|Ψi are both 1. (For the justification that this is possible for the general
Gaussian state, see section 3.) The n-particle states have been given earlier. We
therefore have
Z Z R dd k
e ∗ (k , . . . , k )Ψ = − 21 e(k)|2
(Ω+ω) |φ
D φψn 1 n D φe Hn [φ;
e k ,...,k ] e
1 n
(2π)d (56)

The odd moments of the integral vanish and therefore the coefficients cn vanish
for n odd. For even n, we find
 !
n 1 n/2
Y (2π)d δ d (k 2i−1 + k 2i ) ω2i−1 − Ω2i−1 
cn (k 1 , . . . , k n ) =
n! i=1 2ω2i−1 ω2i−1 + Ω2i−1
o
+ . . . (n − 1)!! perms . . . hψ0 |Ψi (57)

where Ωi = Ω(k i ; t) and the (n − 1)!! permutations specify different combinations

of the paired momenta.
It is particularly interesting to look at the expectation value in the general
Gaussian state of the number operator N(k), which counts the number of particles
with prescribed momentum in the expansion (54). This is
1
hΨ|N(k)|Ψi = hΨ|↠(k)â(k)|Ψi
2ω( )
1 e e 1 δ2 d d
= ωhΨ|φ(k)φ(−k)|Ψi − hΨ| e e
|Ψi − (2π) δ (0)
2 ω δ φ(k)δ φ(−k)
(58)

The expectation values can be easily evaluated using techniques described in

appendix C. For complex Ω, we find the expectation value of the number density
to be
(ω − Ω)(ω − Ω∗ )
hN (k)i = (59)
2ω(Ω + Ω∗ )
This result can also be recovered using the explicit expressions for the expansion
coefficients cn given above.

10
3 Schrödinger Picture in Curved Spacetime
We consider globally hyperbolic spacetimes of topology M = R × Σ with a
global timelike vector field. To provide a general foliation of the spacetime, we
use the formalism of embedding variables [10, 11, 12]. We therefore choose a
family of spacelike hypersurfaces Σ, with intrinsic coordinates ξ i, labelled by a
∂
‘time’ parameter s and consider evolution along the integral curves of ∂s . The
embeddings characterising the foliation are maps x : Σ → M which take a point
on the hypersurface Σ to a spacetime point xµ = xµ (s, ξ i).
First we need to construct the projections of spacetime quantities normal and
tangential to the hypersurface Σ. The tangential projections are defined using
∂xµ
pµi = (60)
∂ξ i
The normal to the surface, nµ , is timelike and is uniquely defined by

nµ pµi = 0 (61)

with g µν nµ nν = 1. Any spacetime tensor can be expanded in terms of its projec-

tions normal and tangential to Σ. We can also introduce an induced metric on
Σ,
hij = gµν pµi pνj (62)
and raise and lower hypersurface indices, e.g. define piµ = hij pνj gµν .
Now consider the deformation vector of the foliation, defined as

µ µ ∂xµ
N = ẋ ≡
∂s
= Nnµ + N i pµi (63)

where N and N i are the ‘lapse’ and ‘shift’ functions of the embeddings given by

N = nµ N µ (64)
N i = piµ N µ (65)

In terms of these functions, the metric interval can be written as

gµν dxµ dxν = (N 2 + N i Ni )ds2 + 2Ni dsdξ i + hij dξ idξ j (66)

We also need the inverse relations

∂s nµ
µ
= (67)
∂x N
∂ξ i −nµ N i
= + piµ (68)
∂xµ N

11
and the Jacobian √
h
∂xµ i N −h
det = √ (69)
∂(s, ξ i ) −g

With these preliminaries complete, we can now construct the Hamiltonian

describing the evolution appropriate to this foliation, i.e. along the integral curves
∂
of the timelike vector field ∂s . We consider a massive scalar field theory coupled
to gravity. The action is
Z h i
1 √ ∂ϕ ∂ϕ
S= d d+1 x −g g µν − (m2
+ ξR)ϕ 2
(70)
2 M ∂xµ ∂xν
and so in terms of the embedding variables,
Z Z h
1 ∂xµ i√ n µν h ∂s ∂s 2
S = ds dd ξ det −g g ϕ̇
2 R Σ ∂(s, ξ i ) ∂xµ ∂xν

∂s ∂ξ i ∂ξ i ∂ξ j i o
2 2
+2 ϕ̇(∂i ϕ) + (∂i ϕ)(∂j ϕ) −(m + ξR)ϕ (71)
∂xµ ∂xν ∂xµ ∂xν
where ϕ̇ ≡ ∂ϕ
∂s
and ∂i ϕ ≡ ∂ϕ
∂ξ i
. Using eqs.(67), (68) and (69), the action can be
written in the form Z Z
S= ds dd ξ L (72)
R Σ
where the Lagrangian density is,
1 √ n ϕ̇2 2N i h N iN j
ij
i
2 2
o
L = N −h − ϕ̇(∂i ϕ) + + h (∂i ϕ)(∂j ϕ) − (m + ξR)ϕ (73)
2 N2 N2 N2
The conjugate momentum is defined as usual to be
√
∂L −h n o
π= = ϕ̇ − N i (∂i ϕ) (74)
∂ ϕ̇ N
The Hamiltonian is given by the Legendre transform
Z
H = dd ξ [π ϕ̇ − L] (75)
Σ
Z
= dd ξ [NH + N i Hi ] (76)
Σ

where
1 √ n π2 o
H = −h − − hij (∂i ϕ)(∂j ϕ) + (m2 + ξR)ϕ2 (77)
2 h
Hi = π(∂i ϕ) (78)

Canonical quantisation is implemented by imposing equal-s commutation re-

lations for the field operators on the hypersurface Σ, viz.

[ϕ(ξ, s), π(ζ, s)] = iδ d (ξ − ζ)

12
 [ϕ(ξ, s), ϕ(ζ, s)] = [π(ξ, s), π(ζ, s)] = 0 (79)
In the Schrödinger picture, the quantum states are represented by wave function-
als Ψ[φ(ξ), s; N, N i , hij ] where the variables φ(ξ) are the eigenvalues of the field
operator on the equal-s hypersurfaces, i.e.

ϕ(x)|φ(ξ), si = φ(ξ)|φ(ξ), si (80)

As before, the commutation relations are realised by representing operators in

δ
terms of the field eigenvalues φ(ξ) and functional derivatives δφ , viz.5
   δ 
O π(x), ϕ(x) ∼ O −i , φ(x) (81)
δφ(x)
The Schrödinger wave functional equation gives the evolution of Ψ[φ(ξ), s] :
Z n1 √  1 δ2 
∂Ψ δ o
i = dd ξ N −h −hij
∂i φ∂j φ+(m2
+ξR)φ 2
−iN i
∂i φ Ψ (82)
∂s Σ 2 h δφ2 δφ

This equation holds for an arbitrary foliation (specified by N, N i and hij ) and
makes explicit the foliation dependence of the wave functional. In practical ap-
plications, however, it is often convenient to choose spacetime coordinates which
reflect the desired foliation. In this case, we simply identify the embedding vari-
ables (s, ξ) with the spacetime coordinates (t, x). The lapse and shift functions
√
are then N = g00 and N i = 0 (so that g0i = 0) while the induced metric is just
hij = gij . The Schrödinger equation reduces to
Z ng o
∂Ψ 1 √ δ2
00
i = dd x −g − g ij
(∂i φ)(∂j φ) + (m2
+ ξR)φ 2
Ψ (83)
∂t 2 g δφ2
This is the form we shall usually use in the remainder of this paper.

We can search for solutions to this Schrödinger equation in analogy to those

found in Minkowski space. For the vacuum state, we make the ansatz

Ψ0 [φ(x), t] = N0 (t)ψ0 [φ, t] (84)

where
 Z q Z q 
1 d d
ψ0 [φ, t] = exp − d x −hx d y −hy φ(x)G(x, y; t)φ(y) (85)
2
This has the same form as before, except that the kernel may now depend on the
‘time’ t since the metric is in general time dependent. The spatial integrals in-
clude the correct measure to ensure that they are invariant d dimensional volume
elements.
The functional derivative is defined so that δφ(y)/δφ(x) = δ d (x − y). The delta function
5
√ R √
density is given by δ d (x, y) = ( −hx )−1 δ d (x − y) and satisfies dd x −hx δ d (x, y)f (x) = f (y).

13
 Substituting this ansatz into the functional Schrödinger equation (83) results
in the time dependence equation
d ln N0 (t) iZ d q q
x
=− d x −hx g00 G(x, x; t) (86)
dt 2
and the kernel equation
Z q
∂ q  q q
i hx hy G(x, y; t) = dd z z
−hz g00 hx hy G(x, z; t)G(z, y; t)
∂t
q q
− x
hx hy g00 (2i + m2 + ξR)x δ d (x, y) (87)
√
where 2i = √1−g ∂i (g ij −g∂j ) is the spatial part of the Laplacian (see appendix
A). Notice that this assumes that on the boundaries of the hypersurface Σ the
fields or their derivatives vanish, since the surface terms in an integration by parts
have been omitted.

We can also mimic the Minkowski solutions for excited states, although in
the general case with explicit t dependence in all the quantities the physical
interpretation is not immediately evident. Nevertheless, we can look for solutions
of the form

Ψ1 [φ, t] = N1 (t)ψ1 [φ, t] (88)

Z q Z q
= 2N1 (t) dd x −hx dd y −hy φ(x)G(x, y; t)f(λ) (y) ψ0 [φ, t]
(89)

This solves the Schrödinger equation with the same kernel as for the vacuum
state. The analogue of the energy gap equation is
Z q q
∆E1 (t) dd y −hy G(x, y; t)f(λ) (y) = x
g00 (2i + m2 + ξR)x f(λ) (x) (90)

where
d h N1 i
ln
∆E1 (t) = i (91)
dt N0
The nature of these states depends on the functions f(λ) (x). These may be chosen
to be eigenfunctions of the operator (2i + m2 + ξR), the suffix λ denoting the
set of quantum numbers specifying the degenerate eigenfunction (the analogues
of the momenta k in Minkowski spacetime).
We can also write down analogues of the creation and annihilation operators
of section 2. If we write
Z q "Z #
q 1 δ
d d
â(λ; t) = d x −hx d y −hy G(x, y; t)φ(y) + √ f ∗ (x)
−hx δφ(x) (λ)
(92)

14
and
Z q "Z #
q 1 δ
↠(λ; t) = dd x −hx dd y −hy G(x, y; t)φ(y) − √ f(λ) (x)
−hx δφ(x)
(93)
then
â(λ; t)ψ0 [φ, t] = 0 ↠(λ; t)ψ0 [φ, t] = ψ1 [φ, t] (94)
and
Z q Z q
[â(λ; t), ↠(ρ; t)] = 2 dd x −hx dd y −hy f(λ)
∗
(x)G(x, y; t)f(ρ) (y) (95)

However, in contrast to Minkowski spacetime, there is no simple particle inter-

pretation associated with these operators for a general spacetime.

The kernel equation (87) is difficult to solve in general, so we shall specialise

to three cases of particular interest – ‘static’ spacetimes, where the metric is a
function only of the space coordinates, ‘dynamic’ or Bianchi type I spacetimes,
where the metric depends only on the time coordinate, and a special class of
conformally static metrics of sufficient generality to include all Robertson-Walker
spacetimes. Before restricting to these special cases, however, it is of some interest
to give a partial solution valid in general.
This partial solution is given in terms of a solution ψ(x) to the wave equation
√
−g[2 + m2 + ξR] ψ(x) = 0 (96)

Any G(x, y; t) satisfying the equation

Z q q ∂
dd y −hy G(x, y; t)ψ(t, y) = − i gx00 ψ(t, x) (97)
∂t
is then a solution of the full kernel eq.(87).
To see this, multiply the kernel equation by ψ(t, y) and integrate with respect
to y and then substitute (97). This gives
Z q Z
∂ hq i ∂ h i
i dd y ψ(t, y) hx hy G(x, y; t) + i dd y hx hy G(x, y; t) ψ(t, y)
∂t √ ∂t
= − −gx (2i + m2 + ξR)x ψ(t, x)

and combining the two differential terms we find

Z q q 
∂ d √
i d y −hx −hy G(x, y; t)ψ(t, y) + −gx (2i + m2 + ξR)x ψ(t, x) = 0
∂t
Using eq.(97) for the kernel again now gives
!
∂ √ ∂ √
gx00 −gx ψ(t, x) + −gx (2i + m2 + ξR)x ψ(t, x) = 0
∂t ∂t

15
which is just the wave equation satisfied by ψ(t, x).
Later, when we have found explicit solutions for static or dynamic spacetimes,
we see that these may be found from eq.(97) by writing the function ψ(x) as an
appropriate Fourier transform, using eqs.(216) or (220) respectively, and using
orthonormality and completeness relations where applicable.

3.1 Static spacetimes

For static spacetimes, the kernel equation is solved with a time-independent kernel
satisfying
Z q
√
dd z −gz G(x, z) G(z, y) = g00
x
(2i + m2 + ξR)x δ d (x, y) (98)

The solutions can be expressed in terms of the eigenfunctions of (2i +m2 +ξR),
which are the Fourier transformed solutions ψe(λ) (ω, x) of the wave equation for a
static metric, viz.

(2i + m2 + ξR)ψe(λ) (ω, x) = g 00 ω 2 (λ)ψe(λ) (ω, x) (99)

The eigenfunctions are specified by a set of discrete or continuous quantum num-

bers λ which generalise the momentum label k in flat space. The eigenvalue
ω(λ) is a function of the λ. These eigenfunctions satisfy orthonormality and
completeness relations described in appendix A.
It is then straightforward to check that the kernel equation is solved by
q Z
dµ(λ)
G(x, y) = gx00 gy00 ω(λ) ψe(λ) (ω, x)ψe(λ)
∗
(ω, y) (100)
(2π)d
In passing, we also note that this expression follows by expressing ψ(x) in eq.(97)
as a Fourier transform and using the completeness relation (219). It is also clear
that the kernel is a real function.
In Minkowski spacetime, we showed that the kernel was related to the inverse
Wightman function evaluated at equal time. We can derive a similar result for a
general static metric. Introducing the inverse of the kernel, ∆(x, y), defined by
Z q
dd z −hz ∆(x, z) G(z, y) = δ d (x, y) (101)

we find using the orthonormality and completeness relations that

Z
dµ(λ) 1 e
∆(x, y) = d
ψ(λ) (ω, x)ψe(λ)
∗
(ω, y) (102)
(2π) ω(λ)
The Wightman function is evaluated in appendix B, eq.(246) and as expected we
confirm
∆(x, y) = 2 G+ (x, y)|ET (103)

16
Again the reason for this identification is given in appendix C.
The complete Schrödinger vacuum wave functional can therefore be written
as
Ψ0 [φ, t] = N0 (t)ψ0 [φ] (104)
where
(Z Z Z )
dµ(λ) ω(λ) d √ q q q
ψ0 = exp − d x −h g 00 φ(x) e
ψ (ω, x) d d
y −h gy00 φ(y)ψe(λ)
∗
(ω, y)
x x (λ) y
(2π)d 2
(105)
e
Notice how the fields φ(x) effectively project out the kernel functions ψ(λ) (ω, x)
appropriate to the boundary conditions on φ on the hypersurface Σ. We shall dis-
cuss this point in more detail in the examples which follow. The time dependent
phase factor is
N0 (t) = e−iE0 t (106)
where E0 is the divergent vacuum energy,
1Z d q q
x
E0 = d x −hx g00 G(x, x) (107)
2Z
1
= dµ(λ) ω(λ) δ d (0) (108)
2

The analysis given above of the excited states for a general spacetime may
now be applied. For a static spacetime, all the quantities are time independent,
so the states generated by the creation and annihilation operators are genuine
stationary states with respect to the chosen Hamiltonian evolution.
The wave functional for the first excited state is
Z q Z q
d
Ψ1 [φ, t] = 2N1 (t) d x −hx dd y −hy φ(x)G(x, y)f(λ) (y) ψ0 [φ] (109)

where G(x, y) is the static kernel, N1 (t) = e−iω(λ)t N0 (t) and f(λ) (y) is a particular
eigenfunction of (2i + m2 + ξR) specifying the state. The corresponding creation
and annihilation operators simplify to
Z "q #
q δ
d
â(λ) = d x −hx gx00 ω(λ)φ(x) + f ∗ (x) (110)
δφ(x) (λ)
Z "q #
q δ
↠(λ) = dd x −hx gx00 ω(λ)φ(x) − f(λ) (x) (111)
δφ(x)
and satisfy the commutation relations

[â(λ), ↠(ρ)] = 2ω(λ) (2π)d δ d (λ, ρ) (112)

17
 As an elementary example, it is straightforward to check
q that in Minkowski
e 2
spacetime choosing ψ(λ) (ω; x) = exp ik.x, where ω(k) = k + m2 and dµ(λ) =
dd k, reproduces precisely the results of section 2.

3.2 Dynamic spacetimes

By a ‘dynamic’ spacetime, we mean one where the metric depends only on the
‘time’ coordinate t. The general metric, after rescaling the t coordinate, can
therefore be written as
ds2 = dt2 − a2ij (t)dxi dxj (113)
and defines a Bianchi type I spacetime [15].
To simplify the kernel equation for the special case of a dynamic spacetime,
e
we introduce the Fourier transform G(k, t) (see appendix A for definitions),
Z
dd k ik.(x−y) e
G(x, y; t) = e G(k; t) (114)
(2π)d

The kernel equation reduces to

∂ h e i √ √
e 2 (k; t) + −gω 2 (k; t)
i hG(k; t) = h −g G (115)
∂t
where ω 2 (k; t) = −g ij ki kj + m2 + ξR.
e k) introduced
The solution6 may be expressed in terms of the functions ψ(t,
in appendix A which satisfy the Fourier transform of the wave equation, viz.
e k) = 0
(20 − g ij ki kj + m2 + ξR)ψ(t, (116)

We find
e −i ∂ e
G(k; t) = √ lnψ(t, k) (117)
−g ∂t
and thus for the kernel itself,
q Z
1 dd k ik.(x−y) ∂ e
G(x, y; t) = −i g 00 √ e lnψ(t, k) (118)
−h (2π)d ∂t

Notice that this is again consistent with the general expression (97) above.
To show this [13], we convert the non-linear first-order differential equation
e
for G(k; t) into a linear second-order equation. Define
Z
F (k; t) = exp{−i e
dt ahG(k; t)} (119)
6
The Schrödinger formalism in flat Robertson-Walker spacetime has been extensively studied
in refs.[13, 14].

18
 √
−g ∂a
where a = −h
and let ȧ = ∂t
. We then have

∂2 ∂ e − a2 h2 G
e 2 F − iȧhGF
e
F = −ia (hG) (120)
∂t2 ∂t
and using eq.(115),
∂2 ȧ ∂ √
2
F = F − a −gω 2 F (121)
∂t a ∂t
e k), since by
However, this is just the equation satisfied by the functions ψ(t,
expanding 20 in eq.(116) we see
 
ȧ √ e k) = 0
∂02 2
− ∂0 + a −gω ψ(t, (122)
a
e k) and eq.(117) follows.
We therefore identify F (k; t) with ψ(t,

The function ψ(t,e k) is the solution of a second-order ordinary differential

equation and is therefore a linear combination of two independent solutions with

arbitrary coefficients. Since the kernel itself is the derivative of the logarithm of
e k), the overall normalisation is unimportant. The kernel therefore depends
ψ(t,
on one arbitrary parameter.
This is a physically important difference from the static spacetime, where
the kernel is essentially unique. For a dynamic spacetime, the ‘vacuum’ states
described by the Gaussian solution of the Schrödinger wave functional equation
form a one-parameter family [13]. The selection of one of these as the vacuum
must then be made on further physical grounds. We shall return to this point in
the context of particular examples later.
The Schrödinger vacuum wave functional can therefore be written as

Ψ0 [φ, t] = N0 (t)ψ0 [φ, t] (123)

where
( q Z Z Z )
i √ dd k ∂ h e i
ψ0 = exp g 00 −h ln ψ(t, k) dd x dd y φ(x)eik.(x−y) φ(y)
2 (2π)d ∂t
(124)
and Rt
N0 (t) = e−i dtE0 (t)
(125)
with Z
1√ √ e
E0 (t) = g00 −h dd k G(k; t)δ d (0) (126)
2
In this case, however, E0 does not have a useful interpretation as an energy.
The analogue of the higher excited states can be found by substituting for
the kernel in the general formulae following eq.(89), where fk (x) = eik.x is an

19
eigenstate of (2i + m2 + ξR) with eigenvalue ω 2(k; t). Again, however, we em-
phasise that the physical interpretation of these states is not as straightforward
as for a static spacetime where they are genuine stationary states of the energy
associated with the chosen Hamiltonian evolution.

It is important to check the stability of the vacuum state defined by the

time-dependent wave functional (124). Vacuum stability requires that
Z Z
DφΨ∗0 (φ, t)Ψ0 (φ, t) = |N0 (t)|2 Dφψ0∗ (φ, t)ψ0 (φ, t) (127)

should be time independent. From eqs.(125) and (126) we have

Z Z
1 dd k e 2}
|N0 (t)|2 = exp{− dd x ln |ψ| (128)
2 (2π)d

where we have used eq.(117) for the kernel. Obviously this will in general be time
dependent.
The second contribution to eq.(127) is
Z
1
Dφ |ψ0 |2 ∼ [det(h Gℜ (x, y; t))]− 2
Z Z
1 dd k e (k; t)]}
= exp{− dd x ln[h G ℜ (129)
2 (2π)d
e is the real part of the (generally complex) kernel. From eq.(117) we
where G ℜ
see
e e∗
Ge (k; t) = i W [ψ, ψ ] (130)
ℜ √ e2
2 −g |ψ|
while for the imaginary part,

e (k; t) = −1 ∂ e2
G ℑ √ e 2 ∂t
|ψ| (131)
2 −g |ψ|
e ψ
where W [ψ, e∗ ] = ψ
e( ∂ ψ
e∗ ) − ψ
e∗ ( ∂ ψ)
e is the Wronskian of the solution and its
∂t ∂t
complex conjugate. Using eqs.(114) and (130) therefore gives
Z Z Z
2 1 d dd k e 2}
Dφ |ψ0 | ∼ exp{+ d x ln |ψ| (132)
2 (2π)d

up to time-independent terms. These remaining terms are seen to be time inde-

√
pendent using Abel’s theorem, viz. that g 00 −gW is a constant, where W is the
Wronskian defined above.
The overall time dependence therefore cancels as required, confirming the
stability of the vacuum state in a dynamic spacetime.

20
3.3 Conformally static and Robertson-Walker spacetimes
A straightforward generalisation of the results of sections 3.1 and 3.2 allows us
to find the kernel for the class of (d + 1)-dimensional spacetimes with metric
ds2 = g00 (t)dt2 − a2 (t)dσ2 (133)
where dσ2 is a d-dimensional static space with metric σij (x). Let σ = detσij .
Introducing a rescaled ‘conformal time’ η, the metric can be written as
h i
ds2 = C(η) dη 2 − σij (x)dxi dxj (134)
The spacetime is therefore conformally static, but with the conformal scale factor
restricted to depend only on time. This class includes the important Robertson-
Walker spacetimes, which are special cases of eq.(133) where g00 = 1 and dσ 2 is
a 3-dimensional static space of constant curvature κ = −1, 0, +1, viz.
dσ 2 = dχ2 + f 2 (χ)[dθ2 + sin2 θdφ2 ] (135)
where 

sin χ 0 ≤ χ < 2π κ = +1
f (χ) = χ 0≤χ<∞ κ=0 (136)


sinh χ 0≤χ<∞ κ = −1
For κ = −1, 0, +1 this metric describes hyperbolic (open), flat, and spherical
(closed) spaces respectively.
Since the spatial metric is static, general theorems guarantee the existence of
orthonormal spatial modes Y(λ) (x) which play the rôle of the Fourier functions
exp ik.x in the previous section. A full discussion of the solutions of the wave
equation for these spacetimes is given in appendix A.2.
We therefore write the kernel as
Z
dµ(λ) e ∗
G(x, y; t) = G(λ; t) Y(λ) (x) Y(λ) (y) (137)
(2π)d
Then, using the solutions T (t, λ) of the transformed wave equation (see eq.(225)),
we readily find (c.f. eqs.(117) or (97))
√
e σ ∂
G(λ; t) = −i √ ln T (t, λ) (138)
−g ∂t
The kernel for a conformally static spacetime with a conformal factor which is a
function of time only is therefore
Z " #
−i dµ(λ) ∂ ∗
G(x, y; t) = √ ln T (t, λ) Y(λ) (x) Y(λ) (y) (139)
g00 ad (2π)d ∂t
The interpretation is identical to that for the dynamic metric case. Again, there is
a one-parameter family of vacuum states corresponding to the freedom in selecting
the solution T (t, λ) of the wave equation.

21
4 Cosmological Model I
To illustrate this formalism, we now consider two simple two-dimensional ‘cos-
mological models’ of Robertson-Walker type. We give an exact description of the
vacuum states, emphasising the rôle of boundary conditions, and discuss carefully
the phenomenon of particle creation due to the cosmological expansion.
The spacetime metric is chosen to be of Robertson-Walker form,

ds2 = dt2 − a(t)2 dx2 (140)

Rescaling the time coordinate such that dt = a(t)dη, the metric may be rewritten
in terms of the ‘conformal time’ η as

ds2 = C(η)(dη 2 − dx2 ) (141)

The metric is conformal to Minkowski spacetime.

A+B
C(n)

A-B

0
-infinity 0 infinity
n

Figure 1: Conformal scale factor for the cosmological model with expansion.

The first model [1] is chosen to have asymptotically static regions separated
by a period of expansion (see Fig(1)). The conformal scale factor is chosen to be

C(η) = A + B tanh(ρη) (142)

where A, B, ρ are constants with A > B > 0. In the asymptotic IN (η → −∞)

and OUT (η → ∞) regions,

C(η) → A ± B as η → ±∞ (143)

This choice allows explicit analytic solutions for the wave functionals in terms of
hypergeometric functions.
The metric is in the class we have called ‘dynamic’, so the construction of
the vacuum wave functional is a straightforward application of the techniques
described in section 3.2. The first step is to find the solutions of the wave equation
in the metric specified by eq.(142).

22
4.1 Wave equation
Taking the Fourier transform,
Z ∞ dk ikx e
ψ(η, x) = e ψ(η; k) (144)
−∞ (2π)

the wave equation becomes

e
[∂η2 + (k 2 + Am2 ) + Bm2 tanh(ρη)]ψ(η, k) = 0 (145)

Notice that the only occurrence of the conformal scale factor is with the mass
m. This is a trivial consequence of the conformal invariance of the scalar field
equation for m = 0 (we have taken ξ = 0 for simplicity). Obviously any non-
Minkowskian behaviour of the vacuum must involve the mass.
It is convenient at this point to make the definitions:

ω 2 (k; η) = k 2 + m2 C(η) (146)

2
ωin = k 2 + m2 (A − B) (147)
2
ωout = k 2 + m2 (A + B) (148)
1
ω± = (ωout ± ωin ) (149)
2
With the substitution
iw−
e
ψ(η, k) = e−iw+ η [2 cosh(ρη)]− ρ χ(η,
e k) (150)

and a change of variable z = 12 [1 + tanh(ρη)], the wave equation reduces to the

hypergeometric equation
( " # )
2
iwin iw− iw− w−
z(1 − z)∂z2 + 1− − z(2 + 2 ) ∂z − + 2 χ(z,
e k) = 0 (151)
ρ ρ ρ ρ

The properties of this equation and its solutions are discussed in [16, 17]. Here,
since −∞ < η < ∞, we are only concerned with solutions in the range 0 ≤ z ≤ 1.
Because of the singular points at z = 0 and z = 1 in the hypergeometric
equation we have to consider two sets of solutions to cover this range completely.
The first set are valid in the range 0 ≤ z < 1, while the second set are valid for
0 < z ≤ 1. The radius of convergence of the two sets of solutions is strictly less
than 1.
Within the radius of convergence of the z = 0 singular point there are two
linearly independent solutions. We can choose these so that as z → 0, one, ψe+ in
,

23
behaves as a positive frequency solution while the other, ψe−
in
, becomes a negative
frequency solution. Explicitly,
iw−
ψe+
in
(η, k) = e−iw+ η [2 cosh(ρη)]− ρ ×
 
iw− iw− iwin
2 F1 ρ
,1 + ρ
;1 − ρ
;z (152)

η→−∞
∼ e−iwin η (153)

  iwin
− 1iw− ρ
ψe−
in
(η, k) = (ψe+
in ∗
) = e−iw+ η
[2 cosh(ρη)] ρ(1 + tanh(ρη)) ×

2 
iw+
2 F1 ρ
, 1 + iwρ+ ; 1 + iwρin ; z (154)

η→−∞
∼ e+iwin η (155)

The solutions valid around the z = 1 singular point can similarly be chosen
so that they correspond to positive and negative frequency solutions in the limit
z → 1. These are
iw−
ψe+
out
(η, k) = e−iw+ η [2 cosh(ρη)]− ρ ×
 
iw− iw− iwout
2 F1 ρ
,1 + ρ
;1 + ρ
;1 −z (156)

η→∞
∼ e−iwout η (157)

 − iwout
1
−
iw− ρ
ψeout (η, k)
− = (ψeout )∗
+ = e−iw+ η
[2 cosh(ρη)] (1 − tanh(ρη))
ρ ×

2 
iw+ iw+ iwout
F
2 1 − ρ
, 1 − ρ
; 1 − ρ
; 1 − z (158)

η→∞
∼ e+iwout η (159)

In the overlap region 0 < z < 1, both sets of solutions are valid and we may
express one in terms of the other. Specifically, we have

ψe−
in
(η, k) = α(k)ψe−
out
+ β(k)ψe+
out
(160)

where
iwin
Γ(1 + ρ
)Γ( iwρout )
α(k) = (161)
Γ( iwρ+ )Γ(1 + iwρ+ )
iwin
Γ(1 + ρ
)Γ( −iwρout )
β(k) = (162)
Γ( −iw
ρ
−
)Γ(1 − iwρ− )

24
The magnitudes of these two coefficients are
 
  πw+
win sinh2 ρ
|α(k)|2 =     (163)
wout sinh πwout
sinh πwin
ρ ρ
 
  πw−
win sinh2 ρ
|β(k)|2 =     (164)
wout sinh πwout
sinh πwin
ρ ρ

and satisfy  
2 win 2
|α(k)| − |β(k)| = (165)
wout
Notice that the solution which is positive frequency in the z → 0 IN region is
a mixture of the solutions of both positive and negative frequencies in the z → 1
OUT region. As we shall see, this mixing is at the heart of the ‘particle creation’
interpretation. The α and β coefficients above are just the Bogoliubov coefficients
in conventional treatments [18, 1].

4.2 Vacuum wave functional

The vacuum wave functional is given by the standard formula (118) for a dynamic
spacetime. That is,
Z ∞ dk ik(x−y) e
G(x, y; η) = e G(k; η) (166)
−∞ (2π)
where
i ∂
e
G(k; η) = − ln[aψe+
in
(η, k) + bψe−
in
(η, k)] (167)
C(η) ∂η
To specify the vacuum beyond this one-parameter ambiguity, we need to im-
pose a boundary condition. We choose this so that in the IN region, the wave
functional becomes the usual positive frequency Minkowski vacuum
 Z 
e 1 ∞ dk e
ψ0IN [φ] = exp − ωin |φ(k)| 2
(168)
2 −∞ 2π
This implies we choose a = 0, b = 1 above, i.e. use the negative frequency solution
of the wave equation in the dynamic kernel.
Explicitly,
i ∂
e
G(k; η) = − [ln ψe−
in
(η, k)] (169)
C(η) ∂η

−1
= ω+ + ω− tanh(ρη) − ωin [1 − tanh(ρη)] ×
C(η)
   
i
+ ρi ω+ ; iωρin ; z 
2 F1 ω ,1
ρ +
   (170)
i iωin
2 F1 ω ,1
ρ +
+ ρi ω+ ; 1 + ρ
;z

25
which in the asymptotic limit is simply
e ωin
G(k; η → −∞) = (171)
(A − B)
as required to reproduce eq.(168). The complete vacuum wave functional is there-
fore
e η] = N (η)ψ [φ,
Ψ0 [φ, e η] (172)
0 0

where  Z ∞
e η] 1 dk e e 2
ψ0 [φ, = exp − C(η) G(k; η)|φ(k)| (173)
2 −∞ 2π

and  Z Z 
i η ∞
e
N0 (η) = exp − dηC(η) dk G(k; η)δ(0) (174)
2 −∞

This is the full analytic expression, valid for all values of the time η, for the
vacuum state chosen to satisfy a positive frequency boundary condition in the IN
region. It is stable for all times, the proof being precisely that given in section 3
for a general dynamic metric.
The next step is to examine the behaviour of this state in the asymptotic fu-
ture OUT region. For this, we use eq.(160) to reexpress the kernel in terms of the
wave equation solutions ψe+ out
and ψe−
out
which have simple asymptotic behaviour
as z → 1. This gives
 
∂ eout ∂ eout
e −i  α ∂η (ψ− ) + β ∂η (ψ+ ) 
G(k; η) = (175)
C(η) α(ψe−
out
) + β(ψe+
out
)

As η → ∞ this function does not tend to a fixed value but rather to a limit circle
in the complex plane. We find
" #
e ωout 1 − γe−2iωout η
G(k; η → ∞) = (176)
(A + B) 1 + γe−2iωout η

where γ = αβ . The field-dependent part of the wave functional is therefore

( " # )
e 1 Z ∞ dk 1 − γe−2iωout η e 2
ψ0 [φ, η → ∞] = exp − ωout |φ(k)| (177)
2 −∞ 2π 1 + γe−2iωout η

4.3 Vacuum states and particle creation

We now discuss the interpretation of this vacuum state and in particular the
phenomenon of particle creation. First, notice that eq.(177) is not the standard

Minkowski vacuum corresponding to the asymptotic limit of the metric, viz.

( Z )
e 1 ∞ dk e
ψ0OU T [φ] = exp − ωout |φ(k)| 2
(178)
2 −∞ 2π

26
 e η] chosen from the
In fact, this is the asymptotic limit of a different state Ψ′0 [φ,
one-parameter family of vacuum solutions to the Schrödinger equation, specified
by the kernel (c.f. eq.(169))

i ∂
e ′ (k; η) = −
G [ln ψe−
out
(η, k)] (179)
C(η) ∂η

Completeness of the solutions to the Schrödinger equation means that any

state can be expressed as a superposition of one of the one-parameter family
of vacuum solutions and the excited states built from it. In particular, we can
express the chosen vacuum state Ψ0 [φ, e η] as a linear combination of the state
′ e
Ψ0 [φ, η] and its excited states. In the OUT region, this amounts to expressing
e η → ∞] of eq.(177) in terms of the excited states of the
the Gaussian state ψ0 [φ,
OU T e
Minkowski state ψ0 [φ].
In the limit η → ∞, these excitations are standard Minkowski particle states
and we can characterise them by their particle number and momenta. The prob-
lem therefore reduces to the one solved in section 2.1, i.e. expressing a general,
time-dependent Gaussian state in Minkowski spacetime in terms of a superpo-
sition of many-particle states. The time-dependent kernel Ω is read off from
eq.(177), viz. " #
1 − γe−2iωout η
Ω(k; η) = ωout (180)
1 + γe−2iωout η
Transcribing results directly from section 2.1, we find that the expectation value
of the number density of particles with momentum k in the state Ψ0 [φ, e η → ∞]
is ∗
e η → ∞]|N (k)|Ψ [φ,
hΨ0 [φ, e η → ∞]i = (ωout − Ω)(ωout − Ω ) (181)
0
2ωout (Ω + Ω∗ )
Remarkably, despite the residual time-dependence in Ω, this expectation value is
independent of η. Explicitly,

|γ|2 ωout
hN (k)i = = |β|2 (182)
1 − |γ|2 ωin
 
πω−
sinh2 ρ
= 
πωout
 
πωin
 (183)
sinh ρ
sinh ρ

This is the Schrödinger picture derivation of the traditional result on particle cre-
ation in this expanding universe model. Notice that the particle number density
is proportional to |β|2 , the Bogoliubov coefficient in the conventional treatment,
which is only non-zero if there is a mixing between the positive and negative
frequency solutions of the wave equation in the respective asymptotic regions.
As expected, hN (k)i also vanishes in the conformal limit m → 0.

27
 It is also illuminating to evaluate the expectation value of the energy momen-
tum tensor in these states. In particular, hΨ0 |T00 |Ψ0 i evaluated in the asymptotic
IN and OUT regions measures the time-independent energy eigenvalue of the state
Ψ0 . In the intermediate region, of course, Ψ0 , like all the other ‘vacuum’ solu-
tions, is not an energy eigenstate. With the definition of T00 implicit in eq.(83)
R √
(recall that H = dd x −gg 00 T00 ), we find
Z  
e η]i = 1 C(η)
e η]|T |Ψ [φ,
∞ dk e 1  e2 ω2 
hΨ0 [φ, 00 0 G(k; η) 1 − e e G − (184)
2 −∞ 2π 2 GGℜ C(η)2
where we have normalised Ψ0 such that hΨ0 |Ψ0 i = 1.
In the IN region, we simply have
1 Z ∞ dk
hΨ0 |T00 |Ψ0 iIN = ωin (185)
2 −∞ 2π
This is just the usual Minkowski zero-point energy density. On the other hand,
substituting (176) into eq.(184) we find, after some calculation,
Z 
∞ dk 1
hΨ0 |T00 |Ψ0 iOU T = ωout hN (k)i + (186)
−∞ 2π 2
Notice that the η dependence from the kernel G(k; e η → ∞) has again cancelled
e
out. Ψ0 [φ, η → ∞] is therefore not the minimum energy state in the Minkowski
OUT region. Rather, eq.(186) describes the energy of a many-particle state with
hN (k)i particles of energy ωout (k) as described above.

The Schrödinger picture therefore brings a particular clarity to the issue of

particle creation in this model. It is evident that there is no state which satisfies
all the properties of a standard Minkowski vacuum state. As we have repeatedly
emphasised, there is a one-parameter family of Gaussian ‘vacuum’ solutions to
the wave functional equation for this class of dynamic spacetimes. If we choose to
e η] which in the asymptotic past is the Minkowski vacuum
consider the state Ψ0 [φ,
e and follow its evolution, we find that in the asymptotic future OUT
state ψ0IN [φ]
region it is described by the time-dependent kernel (176). At this point, there
is really no more to be said – that is a complete description of the state. It is
stable and any physical quantity may be evaluated by taking appropriate operator
matrix elements. The remaining question is simply to describe what the state
e η] looks like. As we have shown above, in the OUT region where the
Ψ0 [φ,
spacetime is again asymptotically Minkowskian, it is precisely a superposition
of many-particle states built from the minimum energy Minkowski vacuum state
e In the asymptotic OUT region, therefore, it is true to say that particles
ψ0OU T [φ].
have been created as a result of the expansion. For any intermediate value of the
time, Ψ0 [φ,e η] continues to provide a complete description of the state, although

in this case there is no simple interpretation in terms of Minkowski particles.

28
5 Cosmological Model II
In this section, we consider a second cosmological model of the same type [3],
this time with a conformal scale factor
A2
C(η) = (187)
cosh2 (ρη)
This is shown in Fig(2). This describes a universe which expands exponentially

A^2
C(n)

0
-infinity 0 infinity
n

Figure 2: Conformal scale factor for the recollapse model.

from an initial singularity before eventually recollapsing. (Since C(η) → 0 as

η → ±∞, the size of the universe tends to zero in the past and future asymptotic
limits.) Notice that although the universe is infinitely long-lived in terms of the
conformal time η, the range of the Robertson-Walker time coordinate t in eq.(140)
is only from t = 0 to t = T = Aπ/ρ. In fact, solving the relation dt2 = C(η)dη 2,
we find a(t) = A sin ρt/A, representing an oscillating universe of which we are
considering only the first cycle.
The analysis of this model follows closely that of section 4. Again we have two
asymptotically flat regions in which we can compare the vacuum solutions of the
dynamic spacetime Schrödinger equation with the equivalent Minkowski vacua.
With some additional subtleties, the results are similar and we again find particle
creation occurring as a result of the gravitational expansion and recontraction.

5.1 Wave equation

In this case, the Fourier transformed wave equation is
e
[∂η2 + k 2 + A2 m2 cosh−2 (ρη)]ψ(η, k) = 0 (188)

Changing variables from η to z as before and making the substitution

ik
e
ψ(η, k) = [2 cosh(ρη)]− ρ χ(η,
e k) (189)

29
the wave equation reduces to the hypergeometric equation [16, 17]
( " # " #)
ik ik k 2 ik A2 m2
z(1 − z)∂z2 + 1+ − z(2 + 2 ) ∂z + 2 − + χ(z,
e k) = 0
ρ ρ ρ ρ ρ2
(190)
Again, there are two sets of solutions of interest, valid in the neighbourhood of
the singular points z = 0 and z = 1 with a radius of convergence strictly less than
1. Choosing a basis of asymptotically positive or negative frequency solutions,
we have
ik  
ψe−
in
(η, k) = [2 cosh(ρη)]− ρ 2 F1 ik
ρ
− σ, 1 + ik
ρ
+ σ; 1 + ik
ρ
;z
(191)
η→−∞
∼ e+ikη

ik ik  
ψe+
in
(η, k) = (ψe−
in ∗
) = [2 cosh(ρη)]− ρ [z]− ρ 2 F1 −σ, 1 + σ; 1 − ik
ρ
;z
(192)
η→−∞
∼ e−ikη

and
ik ik  
ψe−
out
(η, k) = [2 cosh(ρη)]− ρ [1 − z]− ρ 2 F1 −σ, 1 + σ; 1 − ik
ρ
;1 −z
(193)
η→∞
∼ e+ikη

ik  
ψe+
out
(η, k) = (ψe−
out ∗
) = [2 cosh(ρη)]− ρ 2 F1 ik
ρ
− σ, 1 + ik
ρ
+ σ; 1 + ik
ρ
;1 −z
(194)
η→∞
∼ e−ikη

where r
1 1 2 2
±
σ=− 1 + 4Aρ2m (195)
2 2
Notice that these solutions are independent of the mass m in the asymptotic
limit. The mass decouples when the size of the universe shrinks beyond the scale
of m−1 .
In the intermediate region 0 < z < 1, we can express these solutions in terms
of each other. In particular, we will need the relation

ψe−
in
(η, k) = α(k)ψe−
out
+ β(k)ψe+
out
(196)

30
where
ik
Γ(1 + ρ
)Γ( ikρ )
α(k) = (197)
Γ( ikρ − σ)Γ(1 + ikρ + σ)
ik
Γ(1 + ρ
)Γ(− ikρ )
β(k) = (198)
Γ(1 + σ)Γ(−σ)

The magnitudes of these coefficients are

sin2 (πσ) cosh2 (πk) + cos2 (πσ) sinh2 (πk)

|α(k)|2 = (199)
sinh2 (πk)

sin2 (πσ)
|β(k)|2 = (200)
sinh2 (πk)

and satisfy
|α(k)|2 − |β(k)|2 = 1 (201)

5.2 Vacuum Wave Functional and Particle Creation

The construction of the vacuum wave functional is identical to section 4. Choos-
ing boundary conditions such that the vacuum wave functional Ψ0 [φ, e η] becomes

the standard Minkowski vacuum in the asymptotic IN region, viz.

( Z )
e 1 ∞ dk e
ψ0IN [φ] = exp − k |φ(k)| 2
(202)
2 −∞ (2π)

fixes the one-parameter freedom of vacuum states. This choice determines the
kernel,
i ∂
e
G(k; η) = − [ln ψe−
in
(η, k)] (203)
C(η) ∂η
Explicitly,

e i 1
G(k; η) = − −ik tanh(ρη) + [1 − tanh2 (ρη)] ×
C(η) 2
 
( ikρ − σ)(1 + ik
+ σ) 2 F1 1 + ik
− σ, 2 + ik
+ σ; 2 + ik
;z 
ρ ρ ρ ρ
ik

ik ik ik
 (204)
(1 + ρ
) 2 F1 ρ
− σ, 1 + ρ
+ σ; 1 + ρ
;z

Rewriting eq.(203) in terms of the wave equation solutions ψe±

out
valid in the
OUT region, we have
 
∂ eout ∂ eout
e −i  α ∂η (ψ− ) + β ∂η (ψ+ ) 
G(k; η) = (205)
C(η) α(ψe−
out
) + β(ψe+
out
)

31
Then, taking the asymptotic future limit η → ∞, we find the following expression
for the φe dependent part of the vacuum wave functional:
   
 Z ik

e η → ∞] = exp − 1 ∞ dk  1 − γ(1 − z)  e ρ
ψ0 [φ, k ik |φ(k)|2 (206)
 2 (2π) 
−∞ 1 + γ(1 − z) ρ

where γ = αβ . This is to be compared with the equivalent Minkowski vacuum in

the OUT region,
( Z )
e 1 ∞ dk e
ψ0OU T [φ] = exp − k |φ(k)| 2
(207)
2 −∞ (2π)

which is of course identical to ψ0IN in this case. The same argument as in section
4.3 now shows that the asymptotic vacuum state (206) is a linear superposition
of many-particle states built from the Minkowski vacuum ψ0OU T . The expectation
value of the particle number density is

e η → ∞]|N (k)|Ψ [φ,

e η → ∞]i = |γ|2
hΨ0 [φ, 0
1 − |γ|2
sin2 (πσ)
= (208)
sinh2 (πk)

in agreement with the canonical approach [3]. Once again, therefore, we see
that the initial Minkowski vacuum has evolved in the asymptotic future to a
state described by a Gaussian wave functional with a time-dependent kernel,
which may be interpreted as a many-particle state. We conclude that despite
the recontraction following the initial exponential expansion, particles have been
created in this cosmological model as a result of the time-dependence of the
metric.
A remarkable feature of this model, however, is that for certain values of
the parameters, there is no particle creation. From eq.(208) we see that hN (k)i
vanishes when σ is an integer n, i.e.
Am q
= n(n + 1) (209)
ρ
Equivalently, there is no particle creation when the cycle period of the Robertson-
Walker metric and the mass are related by
1q
T = n(n + 1) π (210)
m

32
6 Outlook
These simple models already illustrate some of the power of the Schrödinger
picture techniques to describe vacuum states in spacetimes with time-dependent
metrics. As we have seen in the discussion of cosmological particle creation,
there is a clear advantage in such spacetimes of representing the vacuum state by
a Gaussian wave functional, entirely characterised by a simple kernel function,
rather than the usual Fock space description. This approach also makes clear
that it is only in the very special case of certain static spacetimes that ‘vacuum’
states exist which share all the properties associated with the usual Minkowski
vacuum.
In the companion paper, we extend this investigation of the Schrödinger pic-
ture to simple model spacetimes with boundaries. In this case, even when the
spacetime is static, the Unruh effect sharpens the question of defining a vacuum
state by requiring us to identify which, if any, class of observers (modelled in
the Schrödinger picture by the choice of foliation7 ) perceives the chosen state to
resemble a Minkowski vacuum.
A complete understanding of all these issues is a necessary precursor to de-
scribing vacuum states and particle creation in Schwarzschild-Kruskal black hole
spacetimes, which comprise both static and dynamic regions separated by bound-
aries which are event horizons.

Acknowledgements
One of us (GMS) would like to thank Prof. J–M. Leinaas for extensive discussions
and hospitality at the University of Oslo and the Norwegian Academy of Sciences.
We are both grateful to Dr. W. Perkins for many helpful discussions. DVL
acknowledges the financial support of a PPARC research studentship.

7
This statement requires careful qualification, given the fact that for specified boundary
conditions, physical quantities are independent of the choice of foliation. See ref.([5]) for an
analysis of the subtleties associated with observer and foliation dependence.

33
A Wave Equation
The wave equation for a massive scalar field is

(2 + m2 + ξR)ψ(x) = 0 (211)

where the Laplacian operator is

1 √
2 = √ ∂µ (g µν −g∂ν ) (212)
−g

and g = det(gµν ). The wave equation can be rewritten as

(20 + 2i + m2 + ξR)ψ(x) = 0 (213)

where
1 √
2i = √ ∂i (g ij −g∂j ) (214)
−g
1 √
20 = √ ∂0 (g 00 −g∂0 ) (215)
−g
in coordinates where the metric components g0i vanish.

A.1 Static and dynamic spacetimes

If the metric is independent of a coordinate, then there exists a Killing vector
associated with the corresponding translation symmetry. We use this to intro-
duce convenient Fourier transforms for the special cases of static and dynamic
spacetimes.

In the static case, because of the translational invariance of the wave equation
with respect to t, the solution may be written as
Z ∞ dw −iωt e
ψ(x) = e ψ(ω, x) (216)
−∞ (2π)
where
e
(2i − g 00 ω 2 + m2 + ξR) ψ(ω, x) = 0 (217)
This equation has degenerate solutions labelled by a discrete or continuous set of
quantum numbers (λ), which generalise the momentum k in flat space.
An extensive discussion of the general properties of such solutions in static
spacetimes has been given by Fulling [2, 19]. They satisfy the orthonormality
and completeness relations
Z
dd x q q
d
−hx gx00 ψe(λ)
∗
(ω, x) ψe(ρ) (ω, x) = δ(λ, ρ) (218)
(2π)

34
and Z q
dµ(λ) e∗ e (ω, y) = x d
ψ (ω, x) ψ(λ) g00 δ (x, y) (219)
(2π)d (λ)
where the measure µ(λ) and delta function δ(λ, ρ) are such that
Z
dµ(λ) δ(λ, ρ) fe(λ) = f(ρ)
e

In the dynamic case, because of the translational invariance of the wave equa-
tion with respect to xi , the solution may be written as
Z
dd k ik.x e
ψ(x) = e ψ(t, k) (220)
(2π)d

where k.x ≡ g i j ki xj , the measure is defined in terms of the covariant vector

e k) satisfies
components as dd k = dk1 . . . dkd and ψ(t,
e k) = 0
(20 − g ij ki kj + m2 + ξR)ψ(t, (221)

A.2 Conformally static and Robertson-Walker spacetimes

Now consider spacetimes with a conformally static metric of the form (133). The
Laplacian 20 becomes
q q
20 = g 00 a−d ∂0 ( g 00 ad ∂0 ) (222)

and all dependence on the spatial coordinates drops out. The Laplacian 2i is
−1 √
2i = √ ∂i (σ ij σ∂j )
σ a2
−2
= −a 2d (223)

where 2d is the Laplacian on the d-dimensional static space with metric σij . The
wave equation takes the form

[a2 20 − 2d + a2 (m2 + ξR)]ψ(x) = 0 (224)

The solutions factorise and we find

ψ(t, x) = Y(λ) (x)T (t, λ) (225)

where Y(λ) (x) are eigenfunctions of 2d with eigenvalues K 2 (λ), i.e.

2d Y(λ) (x) = −K 2 (λ)Y(λ) (x) (226)

35
and T (t, λ) is a solution of the transformed wave equation

[a2 20 + K 2 (λ) + a2 (m2 + ξR)]T (t, λ) = 0 (227)

All the analysis above for static spacetimes applies to eq.(226). General theorems
assure the existence of a complete, orthonormal set of solutions Y(λ) (x) satisfying
Z
dd x √ ∗
σ Y(λ) (x) Y(ρ) (x) = δ(λ, ρ) (228)
(2π)d
Z
dµ(λ) ∗ δ d (x − z)
d
Y(λ) (x) Y(λ) (z) = √ (229)
(2π) σ
Robertson-Walker spacetimes are a special case of this class of conformally
static metrics. The solutions of the wave equation for κ = −1, 0, +1 are therefore

ψ(t, x) = T (t, λ) Yλ (χ, θ, φ) (230)

where Yλ are eigenfunctions of the 3-dimensional Laplacian 23 ,

23 Yλ (χ, θ, φ) = −K 2 (λ)Yλ(χ, θ, φ) (231)

and T (t, λ) satisfies

[a−1 ∂t (a3 ∂t ) + K 2 (λ) + a2 (m2 + ξR)]T (t, λ) = 0 (232)

Here, K 2 (λ) = k 2 −κ, where (λ) = (k, j, m) and the eigenfunctions Y(λ) (x),[20, 21]
are explicitly
 3


 (−) 




(2π) 2 Πkj (χ) Yjm (θ, φ) κ = −1 



 


 

 3q 
Y(λ) (x) = (2π) 2 χk Jj+ 1 (kχ) Yjm (θ, φ) κ=0  (233)

 2 

 


 


 3 


 (+) 

(2π) 2 Πkj (χ) Yjm (θ, φ) κ = +1

The Yjm are spherical harmonics and the function Π(−) is defined by
!
(−) 1 1 dj+1 cos(kχ)
Πkj (χ) = [ πk 2 (k 2 + 1) . . . (k 2 + j 2 )]− 2 sinhj χ (234)
2 d(cosh χ)j+1
(+) (−)
The function Πkj is obtained from Πkj by replacing k with −ik and χ with −iχ.
Alternative forms of this decomposition are given in [3, 22].
The orthonormality and completeness relations are given by eqs.(228) and
(229), viz.
Z
d3 x √ ∗
σ Y(λ) (x) Y(ρ) (x) = δ(λ, ρ) (235)
(2π)3
Z
dµ(λ) ∗ δ 3 (x − y)
3
Y(λ) (x) Y(λ) (y) = √ (236)
(2π) σ

36
 √
where σ = f 2 (χ) sin θ. The integration measures for the different types of
Robertson-Walker metric are
 R∞ P∞ Pj 
Z  0 dk
 j=0 m=−j κ = −1, 0 

dµ(λ) = (237)

 P∞ Pk−1 Pj 

k=1 j=0 m=−j κ=1

which also specifies the form and range of the quantum numbers k, j, and m.

For the spatially flat κ = 0 metrics, it is of course frequently more convenient

to choose the eigenfunctions in Cartesian form

Y(λ) (x) = exp(ikj xj ) (238)

R R
where (λ) = (k1 , k2 , k3 ) and K 2 (λ) = k 2 . The measure is just dµ(λ) = d3 k.

B Green Functions
The Green function is given by the equation
√
−g(2 + m2 + ξR)G(x, y) = −δ (d+1) (x − y) (239)

As in the previous section, if the equation is translationally invariant with respect

to a particular coordinate then the Green function may be expressed as a Fourier
transform.
In the static case,
Z ∞ dω −iω(x0 −y0 ) e
G(x, y) = e G(ω; x, y) (240)
−∞ (2π)
where
√ e
−g(2i − g 00 ω 2 + m2 + ξR)G(ω; x, y) = −δ d (x − y) (241)
For dynamic spacetimes,
Z
dd k ik.(x−y) e
G(x, y) = e G(k; x0 , y 0 ) (242)
(2π)d
where
√ e
−g(20 − g ij ki kj + m2 + ξR)G(k; x0 , y 0) = −δ(x0 − y 0 ) (243)
We cannot say more about the dynamic Green function in general. However, in
the static case, we can give a general solution. Using the orthonormality relations
of solutions to the wave equation, we find
Z Z
∞ dω dµ(λ) e−iω(x0 −y0 ) ψe(λ) (ν, x) ψe(λ)
∗
(ν, y)
G(x, y) = (244)
−∞ (2π) (2π)d (ω 2 − ν(λ)2 )

37
plus solutions of the homogeneous equation. The ω integral is along the real axis,
but there are poles at w = ±ν. The choice of contour (iǫ prescription) selects
one of a variety of different Green functions [1, 2].
For example, the contours giving the Wightman function are shown in Fig(3).
Evaluating the integral gives

G+ (x, y) = h0| ϕ(x) ϕ(y) |0i (245)

Z
1 dµ(λ) 1 −iω(x0 −y0 ) e
= e ψ(λ) (ω, x) ψe(λ)
∗
(ω, y) (246)
2 (2π)d ω

As the Wightman function is just the difference between two Green functions, it
satisfies the homogeneous wave equation.

ω
-iG - iG+

−ν +ν

Figure 3: The omega contour integral for the Wightman functions.

C Expectation values and two-point functions

The vacuum expectation value of an operator O(ϕ, π) (a function of the field ϕ
and conjugate momentum π) in the Schrödinger picture is
Z
δ
h0| O(ϕ, π) |0i = Dφ Ψ∗0 O(φ, −i ) Ψ0 (247)
δφ
The action of the momentum is simply
δ
π(x)Ψ0 = −i Ψ0
δφ(x)
Z q
d
= i d y hy hx φ(y)G(y, x; t) Ψ0 (248)

To evaluate the action of the field, it is convenient to modify the wave functional
by formally introducing a ‘source’, viz.
1Z d q
ΨJ = Ψ0 exp { d x −hx J(x) φ(x)} (249)
2

38
Omitting field independent terms for clarity, the vacuum amplitude is then
Z R d R d √ R d √
h0|0iJ ∼ Dφe− d x d y hx hy φ(x)Gℜ (x,y;t)φ(y) + d x −hx J(x)φ(x) (250)
n Z Z q o
∼ exp 1
4
dd x dd y hx hy J(x)∆ℜ (x, y; t)J(y) (251)

Notice that the functional integral involves just the real part Gℜ of the kernel.
Also note that ∆ℜ is defined here as the inverse of Gℜ (not the real part of the
inverse kernel). Expectation values of products of fields can then be found by
repeated differentiation with respect to the source, e.g.
Z 

h0|ϕ(x)ϕ(y)|0i = Dφ Ψ∗J φ(x)φ(y) ΨJ 
J=0
1 δ 1 δ 
= √ q h0|0iJ 
−hx δJ(x) −hy δJ(y) J=0

1
= ∆ℜ (x, y; t) (252)
2
This is the equal-time Wightman function, confirming the identifications (20)
and (103) in the text.
Other useful two-point functions are
1q n
h0| π(x) π(y) |0i = hx hy Gℜ (x, y; t)
Z 2 Z q o
+ dd u dd v hu hv Gℑ (x, u; t)∆ℜ (u, v; t)Gℑ (v, y; t) (253)

and

h0| [ϕ(x), π(y)] |0i = iδ d (x − y) (254)

Z q
h0| {ϕ(x), π(y)} |0i = − dd u hx hu Gℑ (x, u; t)∆ℜ (u, y; t) (255)

39
References
[1] N. D. Birrel and P. C. W. Davies, Quantum Fields in Curved Space (Cam-
bridge University Press, 1982).
[2] S. A. Fulling, Aspects of Quantum Field Theory in Curved Space-time
(Cambridge University Press, 1989).
[3] A. A. Grib, S. G. Mamayev and V. M. Mostepanenko, Vacuum Quantum
Effects in Strong Fields (Friedmann Laboratory Publishing, St. Petersburg,
1994).
[4] I. Moss, Quantum Theory, Black Holes and Inflation (John Wiley and Sons
Ltd, 1996).
[5] D. V. Long and G. M. Shore, in preparation.
[6] S. W. Hawking, Comm. Math. Phys. 43 (1975) 199.
[7] R. Jackiw, in Field Theory and Particle Physics, 5th Jorge Swieca Summer
School, Brazil (World Scientific, 1989).
[8] J. Schwinger, Phys. Rev. 82 (1951) 664.
[9] B. S. DeWitt, in Relativity, Groups and Topology, eds. B. S. DeWitt and
C. DeWitt (Gordon and Breach, 1964).
[10] K. Kuchař, J. Math. Phys. 17 (1976) 777, 792 and 801.
[11] K. Freese, C. T. Hill and M. Mueller, Nucl. Phys. B255 (1985) 693.
[12] J. J. Halliwell, Phys. Rev. D43 (1991) 2590.
[13] J. Guven, B. Lieberman and C. T. Hill, Phys. Rev. D39 (1989) 438.
[14] O. Éboli, S. Pi and M. Samiullah, Ann. Phys. 193 (1989) 102.
[15] L. P. Hughston and K. P. Tod, An Introduction to General Relativity (Cam-
bridge University Press, 1990).
[16] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products
(Academic Press, New York, 1980).
[17] W. Magnus, F. Oberhettinger and R. P. Soni, Formulas and Theorems for
Special Functions of Mathematical Physics (Springer-Verlag, Berlin, 1966).
[18] C. Bernard and A. Duncan, Ann. Phys. 107 (1977) 201.
[19] S. A. Fulling, Phys. Rev. D7 (1973) 2850.
[20] L. Parker and S. A. Fulling, Phys. Rev. D9 (1974) 341.
[21] E. M. Lifshitz and I. M. Khalatnikov, Adv. Phys. 12 (1963) 185.
[22] L. H. Ford, Phys. Rev. D14 (1976) 3304.

40