Professional Documents
Culture Documents
QUANTUMTHEORY
IN STRONG G R A V ~ T A T I O N AFIELDS
L
Q U A N T U M F I E L D T H E O R Y IN C U R V E D SPACE-
TIME: AN OVERVIEW
C. J . Isham
Deparrment of Physics
Imperial College
South Kensington
London S W7 2 BZ England
1 NTRODUCTION
The problem of quantizing the gravitational field, and hence unifying quantum
theory with the general theory of relativity, remains stubbornly u n ~ o l v e d . Until
~
recently, surprisingly little work had been done on that subject whose study might
be thought i n many respects to logically precede the construction of a theory of
quantum gravity, namely the quantization of a linear matter field on a given,
fixed and unquantized, curved space background. Many aspects of such a theory
should be explicitly soluble and would, one hopes, throw valuable light on the full
quantum gravity problem. Irrespective of what this final theory might be, one
anticipates that there will, i n any event, be a semiclassical region where the influ-
ence of necessarily quantized matter on the metric tensor is described by equations
of the type
G,,,(g) = <$I T,,(g.matter) I > (1.1)
>
where I $ is some quantum state of the matter and T,,is the quantized energy-
momentum tensor. At a more technical level the background field method92-95,73
of quantizing gauge theories, when applied to the gravitational field, involves the
construction of quantum fields and their propagators in an arbitrary space-time.
A more practical motivation for studying this subject lies in the field of early-
universe cosmology. An external time-dependent gravitational field might be
expected, like any other external potential, to produce quanta of fields to which it
couples (which, by virtue of the equivalence principle, it does to everything). The
nature of this coupling suggests that particles will be produced in pairs and pre-
sumably they will react back on the space-time geometry via Equation 1 . 1 , at least
in some semiclassical limit. I f the universe started as a big-bang then near the
moment of formation there was rapid time dependence i n the metric and such pairs
could have been copiously produced. I t is hoped that this phenomenon might
explain some of the gross macroscopic features of the universe as seen today; in
particular, initial anisotropies i n the expansion rate may be damped out by this
back-reaction process. Similarly, this production may account for the residual 3K
background radiation. The general idea, that the universe was created as pure
gravitation which then produced matter in this way, could also explain other ob-
served features such as baryon dominance (via an early C P violating force).
The reason for the recent explosive interest in quantum field theory i n a curved
space-time was however none of the above. I t stemmed instead from a remarkable
result of S . W . Hawking. He discovered that if a linear field is quantized in the
1 I4
I s h a m : Q u a n t u m Field Theory I I5
presence of a black hole, then the black hole produces quanta o f this field and
radiates them exactly a s if it is a black body with a temperature Tinversely propor-
tional to its mass
kT = (8*M)-' (1.2)
This is truly astonishing. For example, black-body radiation is described quantum
mechanically by a mixed, density matrix state, not by a pure state. T h e appearance
of a mixed state is generally associated with loss of information. In the present
case the information is truly lost-down the hole! Indeed the presence (or potential
presence) of a horizon plays a crucial role in the black hole radiation process.
Hawking visualizes the production of a virtual pair near the horizon. O n e parti-
cle moves out t o a distant observer, while the other passes through the horizon
with a negative energy (as measured by the distant observer). Once inside it be-
haves locally like a normal particle and cannot escape again. Thus, the picture is
one in which the particles are produced by quantum tunnelling through the hori-
zon rather than via an explicit metrical time dependence of the type invoked above
in discussing cosmological models. Let me cautiously add that not everyone agrees
with Hawking's picture (although they usually agree with his radiation result), and
in any event it is unwise t o overemploy classical imagery when describing what is
undoubtedly a quantum process.
Hawking's results clearly have a major impact on many hitherto unrelated
branches of theoretical physics. T h e most obvious one concerns black-hole
physics itself. From conservation of energy the emission of a particle is associated
with a corresponding decrease in the mass of the black hole, and detailed calcula-
tions show that for a black hole to decay completely in this way within the lifetime
of the universe, its mass must be less than 10"g. Thus, we a r e dealing with objects
that by astrophysical standards are rather small. For such objects, Hawking's re-
sults complete the connection between black-hole physics and thermodynamics,
first developed a t the classical Icvel,".J0~174by assigning to B black hole its inverse
mass temperature (Equation 1.2) a n d an intrinsic entropy proportional to the area
of the event horizon. This allows exact analogs of the Leroth, first, second and
even possibly the third, laws of thermodynamics l'or black-hole equilibrium con-
f i g u r a t i o n ~ . ~It' appears that such analogies can be constructed whenever there is
an event horizon, such a s for example in D e Sitter space."'
As the black hole thermally radiates its mass decreases, a n d the question of the
end state arises. Normally, a space-time singularity in general relativity is shielded
from the outside world by a n event horizon (cosmic censorship). However, a s the
Hawking process proceeds, a negative influx olencrgy causes the horizon t o shrink
(it it forms a t all), and a s the last burst cf' radiation is emitted there is a distinct
possibility that a (massless!) naked singularity is left behind. This singularity has
the peculiar property of potentially carrying a large baryon number. To see this.
consider a collapsing ball of neutrons of total mass 10" g with an associated teni-
perature of IO""K (about 250 Mev). Sincc particles can only be produced when
the temperature is above their rest mass, the bulk or the radiation is initially in the
form of photons, neutrinos, a n d ultrarelativistic electrons and positrons. The
absence of massless baryons implies that baryonic charge cannot be radiated away.
(The "no hair" theorem strikes again.) Of course, a s the mass of the black hole
I16 A n n a l s N ew York A c a d e m y of Sciences
H E U R I S T IQCU A N T U F1Ei.t)
M T I I E O R ASPECTS
Y
carries its own representation of the canonical commutation relations on its own
Hilbert space and that this is a more appropriate starting point. Actually, in the
covariant approach there is no reason why the time evolution generated by the
field equations should be implementable by a unitary operator, even though the
formalism is set up on a single Hilbert space.
T o remove the arbitrariness, it is necessary to consider the remaining three con-
ditions. The importance of boundary conditions can be appreciated even in flat
space where, for example, the difference in vacuum energy between fields in a box
and fields in Minkowski space gives the famous Casimir effect.i5 T o the extent
that boundary conditions reflect the global topological structure of the space-time
they have really already been incorporated in stages 1 and 2. Indeed, it should be
emphasized that all five points are intimately connected and cannot truly be sepa-
rated. In practice it is usually in the definition of physical states that boundaries
play a major role. Thus, for example, in the case of the eternal black hole some of
the possible candidates for a vacuum state are defined in terms of boundary condi-
tions on the event horizons rather than a t spatial or null infinity.
The construction of mathematical states and observables (point 4) is not
strictly speaking a problem if one follows a Segal Haag C*-algebra approach.
There is a problem, however, in associating a specific quantum operator with a
given classical observable. This is because many such observables are nonlinear
functions of the fields (the energy momentum tensor is a good example) and are
thus formally divergent in the quantum theory. The necessary techniques of regu-
larization and renormalization will be discussed in a later section.
I t is the last condition-the interplay between physics and mathematics-that
really causes the headaches. In Minkowski space we habitually use a Fock quanti-
zation with its associated particle labels for states and observables. Unfortunately,
these definitions involve global concepts such as positive frequency classical solu-
tions, or the Poincarit group of inertial observers, and these are nor applicable in a
generic curved space as, indeed, they are not in most situations involving external
potentia~s,152.155,159.160~ 166 w
ithout the PoincarC group invariance of the vacuum
and the required positivity of the associated Hamiltonian, we lose the crucial tech-
nical tools that provided the unique conventional free quantum field theory. Quan-
tum fields may be readily constructed using various Fock spaces, but there is no
reason why thequanta i n the Hilbert space should be related in any way to physi-
cal particles and, as remarked before, there is in general no natural method of
selecting the correct representation. Indeed, it is unclear what precisely is meant
by a physical particle when an external potential is present. In the case of back-
ground space-times, the notion of frequency (which is intimately connected with
the idea of a particle) only makes sense for wavelengths smaller than the local
radius of In addition, if particles are being produced by a time-
dependent field, a simple use of the time-energy uncertainty relations leads to a
lower bound on the intrinsic ambiguity in the definition ofparticle number.
I f there is a global, hypersurface orthogonal, Killing vector (i.e., if M is static),
then in most cases a natural quantization scheme exists that can also be extendcd
in a believable way to situations where f l is mcrely stationary. The use of
Killing vectors that are only locally timelike can, however, lead to problems. For
a general space-time, one can proceed with reasonable confidence if there are in
and out regions (interpreted in the broadest sense) in which the notion of parti-
I18 A n n a l s N e w York A c a d e m y of Sciences
cle can be sensibly defined. Most of the models that have been considered belong
to this category. However, this technique is not immediately applicable in cosmo-
logical situations where the interaction cannot be turned-of and where there
may be no convenient asymptotic regions. Even i n the in/out case, it is frustrating
to be forced to keep to a pure S-matrix formalism and not ask what is going on?
in the interaction region. There have been various attempts to resolve this prob-
lem. Some have involved inventing what it is hoped are physically plausible defini-
tions of a particle. Thus, Parker has introduced an adiabatic concept in which the
particles are defined so as to minimize the rate of change of particle number with
t,me,2.3.Y 13.17 This is loosely related to an expansion of the metric as a sort of
asymptotic series around a static space-time. Another popular scheme is to define
the particle operators such that at each time (for some given choice of time co-
ordinate) the Hamiltonian can be written in the symbolic form
ff(t) = C
n
an(oan+(l)an(l) (2.1)
It is this Hamiltonian diagonalization method that leads naturally to the use of
different Hilbert spaces at different times mentioned before.4~6920~24~28
One inter-
esting approach is the extreme operationalism of Unruh,6 who constructs quan-
tum mechanical model detectors to investigate a theorys particle content.
Operationalism is of course a fundamental feature of the Copenhagen interpreta-
tion of quantum mechanics, but although most practising physicists pay lip service
to it they are not used to being forced to invoke it as a means of resolving real
problems. I t is also noteworthy that the use of these detectors focuses attention
on the world line of a single observer, rather than on one of the Cauchy surfaces
of the traditional canonical quantization approaches. However, if in using the
detector we integrate along the whole world line (which is often the case), these
two points of view are not so different. An alternative approach is to drop the
particle language as far as possible and concentrate on field-oriented observables
such as the energy-momentum tensor T,,, . This is also one of the motivating fac-
tors behind the investigations aimed at constructing the crucial Greens functions
directly (via, say, their complex analytic structure) rather than from an inter-
mediary quantum field theory. However, since it is usually the vacuum expectation
values that are being computed or constructed, there is always some particle con-
cept hidden, either implicitly or explicitly, in the theory.
There are, needless to say, other quantum field theoretic problems that can be
anticipated in addition to that of defining particles. The quantum theory is usually
required to maintain general covariance, and where divergent objects (such as T,,)
are involved this can lead to difficulties. There may also be problems with other
group invariances, for example there is quite strong evidence that conformal
invariance is not preserved at the quantum level. One can also anticipate that the
infrared divergences of a massless theory, being associated with long-range effects,
could radically change in non-Minkowskian topologies. Finally, of course, the
eternal problem remains of how to incorporate genuine quantum gravity effects.
These will certainly be important at dimensions of the Planck length (Lp -
cm), the Planck time ( l p - sec), the Planck mass ( M p - g - lo2*
eV) or the Planck temperature ( Tp - 1032K).
At this point i t might be useful to tabulate, with a few brief comments, the
I s h a m : Q u a n t u m Field T h e o r y I19
principal techniques that have been used and the models in which they have been
employed:
Techniques
I, Conventional in/out quantum field theory in Fock space. The main tool
is Bogolubov transformations. (BT)
2. Calculation of the regularized vacuum expectation value < T,,>of the
energy-momentum tensor: The divergences are the principal ditficulty. ( T P u )
3. Greens functions: Rather than constructing the Feynman Greens function
as the vacuum expectation value of a time-ordered product C,(x,y) -
>,
<O I T 4 ( x ) @ ( y )10 one attempts to construct this function ah inirio
(usually in terms of its complex analytic behavior) and then use i t to recon-
struct the underlying quantum field theory., (GF)
4. Thermal Greens functions: Since the Hawking black-hole radiation is
thermal, it is not surprising that the use of thermal Greens functions can
be very powerful. Once again, complex analytic structure plnys an impor-
tant role. (TG)
5 . Model particle detectors: This is no1 really a technical tool but as explained
already is mainly used to investigate the particle content of the theory in
some physically motivated operational way. (See Unruhs article i n this
volume.) (D)
Models In vesrigared
TRANSFORMATIONS
BOGOLUROV
Consider as an example the ubiquitous scalar field @(x). The first aim o f co-
variant quantization is to construct a field operator that satisfies (dropping any
pretence at rigor),
(V", -+ R / 6 ) # J ( x ) = 0 (3.3)
where R is the scalar curvature.
A natural bilinear form (the Klein-Gordon "inner product") can be defined on
the space of complex classical solutions to Equation 3.1 or 3 . 3 ) by:
I n Minkowski space
#J > > 0 for positive frequency solutions
<#J,
< 0 for negative frequency solutions
and as has been previously remarked, a major problem in curved space is to find
the natural analog of positive frequency.
I s h a m : Q u a n t u m Field T h e o r y I21
Suppose that ( J }is any complete set of complex solutions to Equation 3.1
with the properties
. <f;>&>
= 61J (3.6)
2. <l;,f,*>
= 0
These statements can all be made rigorous in terms of a natural Hilbert space
structure induced on the space of solutions by the set (/;I, but for the purposes of
this review the heuristic form is sufficient.
T o quantize the theory, the coefficients a, are regarded as operators satisfying
(3.10)
(3.11)
which by Equation 3.7 give the covariant commutation relations (Equation 3.2).
Alniost invariably, the Fock representation is chosen with the vacuum denoted
lO;(fl >,
indicating the dependence of the quantization on the choice of the set
If;]
aiIo;lfI> = 0 (3.12)
Thus, a,* creates a particle whose wave function is l;, although as we keep
emphasizing, these quanta may have little or nothing to d o with physical particles.
The main problem is t h e choice of the physically corrcct set of basis solutions
{ J ] .If the space-time is static with a global timelike Killing vector K, a natu-
ral choice of { A 1 is one such that for all j
(3.14)
I22 A n n a l s N e w York A c a d e m y of Sciences
H = I2
I-
Kp:T,,:du (3.15)
then
HIO;(fJ> = 0
and a,* (resp. a , ) creates (resp. annihilates) one unit of H-energy. Thus, provided
such a basis exists (one problem for example is that Klein paradox or Schiff, Sny-
der, Weinberg problems with complex w , might a r i ~ e ~ ~ *there ~ ) is a natural
quantization for a static manifold.
In general there will be no preferred choice of 1x1.Different choices may
lead to the same vacuum, different vacuums in the same Hilbert space, or different
vacuums in different Hilbert spaces (in the sense that the Fock representations are
unitarily inequivalent). A typical situation that arises is the following: Suppose
that there are two regions of space-time (called in and out) in which there are
natural choices of basis sets satisfying Equation 3.6 (the positive frequency con-
dition). We further suppose that there are complete sets of solutions to the wave
equation {ell and 1x1 that are defined on all of a and which reduce t o the
appropriate positive frequency solutions i n the in and out regions, respec-
tively. Then d(x) can be expanded as
with operator coefficients a, and bJ which we assume are defined on the same Hil-
bert space.
Now
(3.17)
but in general
1; = sjj.j1
i. + Bijej* 1 (3.19)
In other words a classical solution that is positive frequency in the "out" region
is a mixture of positive and negative frequency solutions in the "in" region (and
vice versa ofcourse). I t follows from Equation 3.19 that the various operators are
related by the Bogolubov transformation
4 = S,Ia,,*a, - B,,*U,*l (3.20)
and hence that the average number of"out" particles in the mode j is
<o; IeI I b,* h, 10; {el > = sI, I pJk I ' (3.21)
More generally the "in" vacuum can be written somemhat formally in terms of
"out" states as
I o ; { e }> = exp - b*(a')-l/Yh* 10; I f ] > (3.22)
where (possibly continuous) matrix products are understood in the exponent, the
existence of which is equivalent to the unitary equivalence of the t w o representa-
tions.
ISOTROPIC C O S M O L O C I CMODELS
A~
Over the last ten years a significant literature has evolved dealing with quan-
tum field theory in cosmological backgrounds, of which I only have time to men-
tion a l i t t l e . ' ~ 2 8 ~ ' 2 y ~ ' 3 " ~ 1 4 "Fortunately,
~ 1 4 1 ~ i ~ 7 ~ 1there
4y has recently appeared an
excellent and extensive review by L. Parker," which may be consulted for further
information.
As a n example of an isotropic cosmology, consider the spatially flat Robertson-
Walker metric'" 15:
ds2 = dt' - a2(t)dx.dx (4.1)
To facilitate the definition of particles, suppose that the expansion is "statically
bounded" in the sense that
Lt a ( r ) = a2 > 0 (4.2 1
1 - r
0
, Lt a(r) = a, > (4.3)
- -
f - L
p - to-4 ( 4.8 )
Parkers detailed calculations for the statically bounded case substantiate these
rough order-of-magnitude estimates. Using the time T :
(4.9)
I s h a m : Q u a n t u m Field T h e o r y I25
where (4.12)
The positive frequency "in" solution of Equation 4. I I is used in Equation 4.10 and
is determined by the requirement
$k( 7) :'-= (20 I3 w k ( -= )) -1/2exp - ial 3 w k (-% ) 7 . (4.13)
The field can also be expanded i n terms of the "out" positive frequency fields by
(cf. Equation 3.19) evaluating the Bogolubov coefficients (Yk. pk using
$k(T)"-"(2U23Wk( x))-l'Z((Ykexp - iaz3wk( x ) T + /jk expia230k(")T (4.14)
and the amount of particle production with wave number k is indicated by the
number operator's expectation value (cf. Equation 3.21)
<o In 1 Nk'"' 10 In > = 1 pk 1 (4.15)
(There is no integral or sum because of the rather simple nature of this particular
Bogolubov transformation.)
The only potential ditficulty lies i n solving Equation 4. I 1 , this being necessary
to compute "k and pk. ParkerI4.l5 has considered the case when the radius func-
tion U ( T )is given by
a4(7) = aI4 + e ' / ' ( ( ~-~ ~aI4)(er/'
- + 1) + b)(e'/" + (4.16)
which has the advantage that Equation 4.1 I may be explicitly solved. The 7-time
over which the transition from a l to a2 takes place is determined by s, whilst h is
an unimportant constant. For typical cosmological applications, a2 > > a , and
to a very good approximation
1 p k / a k1 - e- 4 ~ 01' 1 1 (4.17)
More precisely the particles (resp. antiparticles) are produced with a Planck energy
spectrum, but there arc particle/antiparticle correlations. This must be so since
there is no horizon (unlike the black hole casc16) into which information can be
lost converting a pure state into a mixed state. In practice, various decays and sub-
sequent interactions will remove these correlations, but it should be emphasized
that the particles are produced with a Planck spectrum -no interaction type of
thermalization is invoked.
The main problem in using Equation 4.18 is the selection of a physically reason-
able value for al . The actual big-bang was presumably not statically bounded but
I26 Annals New York Academy of Sciences
s - t,/a,. (4.19)
giving
Parker also assumes that the universe evolves from al = Ct,l/ into a radiation-
dominated a ( t ) = Ct12,and he confirms (and this is an essential consistency
check) that the amount of energy produced in the particle creation process is just
enough to sustain this Friedman Ctl/ expansion if the associated energy-momen-
tum tensor (in a rough classical approximation) is placed on the right-hand side of
the Einstein equations. Allowing for the later matter-dominated evolution he
shows that the temperature today is of the order of a degree Kelvin, and the radia-
tion would be primarily in the form of photons and gravitons. This is potentially
an attractive explanation of the observed microwave background, but the treat-
ment is rather crude at this stage. Although Parker indicates that the thermal
nature of the particle production is not really dependent on the specific form of
a ( t ) in Equation 4.16, it is not yet clear exactly what class of expansion functions
would reproduce these results. More importantly, not only are the true quantum
gravity effects unknown but even the semiclassical back-reaction effects have not
been included. The importance of these can be appreciated from earlier work of
Parker and Fulling,16 who showed that if a carefully chosen state I >
was used
in a
quantization scheme, then the singularity in the Robertson Walker metric was
avoided. The regularization employed in the construction of the right-hand side
of Equation 4.21 was rather ad hoc, but recently there has been a thorough
i n v e ~ t i g a t i o n ~ ~of. the regularized stress tensor in a Robertson Walker back-
ground, and the time seems ripe for looking seriously at the whole back reaction
problem. These effects will certainly be of major importance at times or distances
of the Planck quantum gravity scale.
I s h a m : Q u a n t u m Field T h e o r y I27
ANISOTROPIC
COSMOLOGICAL
MODELS
c c
I=
3
I
PI =
1-
3
I
PI2 = I
(5.5)
The quantum field is expanded as i n Equation 4.10 but with $k obeying the dif-
ferential equation
The most detailed calculations using these techniques are those of Lukash,
I28 A n n a l s N e w York A c a d e m y of Sciences
T H E COLLAPSING
BALL
scalar field so that future and past null infinity ( I +and I - ) are appropriate sur-
faces on which to consider initial data.
In terms of the usual null coordinate on I - we can regard incoming positive
11
va,6 = o (6.1)
defined both outside and inside the matter and which on I - have the asymptotic
behavior
(6.2)
F (r)
PWXm - Y,,,,(B, 6 ) w - ? w
e-rwu
w > o (6.3)
modes, so let usjust pick one set and label them { q i } .Then the quantum field can
be expanded in terms of these solutions as
The essential technical step is the calculation o f the decomposition of the out-
going put,,, modes into sums of the incoming f u X m and &,,,* functions.
Hawking did this by clever use of the geonictrics optics approximation to wave
propagation. Since his original paper48,4Y is exceptionally clear, I will not g o into
any further details about this aspect. Once this decomposition is known the in-
coming vacuum 10, i n > can be expressed as a sum of outgoing particle states,
remembering of course that outgoing includes the particles falling across the
horizon. This expression will involve the unknown coefficients relating the { q r}
to If;.) and {f;*l. Now the Hilbert space o f outgoing classical solutions i s
expressible as the direct s u m
~ 0 =
Xhor @X (6.6)
in an obvious notation. This implies that the out Fock space of particle
states can be decomposed as the tensor product
TOUI - 3 hur @ 5 /+
(6.7)
130 Annals N e w York A c a d e m y of Sciences
so the general out state can be written as the sum of products of an outgoing
particle state on I + (where the notion of particle is sensible) with a particle
state on h + . The next crucial observation is that a distant observer (ix., near I + )
has no information about what happens on the horizon. In mathematical terms
if 0 is an observable formed from the operators b,,, and but,* (and hence
representing something near I+), the expectation value <0, in 16 10, in >
depends in a trivial way on the horizon operators c i , ci* and the Fhparts of
the state can be explicitly factored out. The net result is that there is an effective
density matrix p defined on 9only, such that for all such observables 6,
<0, in 16 10, i n > = Trpb (6.8)
Furthermore, all results thus obtained are independent of the choice of horizon
modes {qil.This conversion of a pure state into a mixed density matrix state is a
well-known feature of any composite system S , x S2 with quantum mechanical
Hilbert space X , @ X 2 when for example only S , observables are being con-
sidered. However, as Hawking has repeatedly emphasized, the state conversion is
an itifrinsic feature of the black hole, being a direct consequence of a distant
observers unavoidable loss of information down the event horizon. The remark-
able result is that this state p is a thermal state with a temperature T given by
kT = Xic3(8~GM)- (6.9)
where M is the mass of the black hole. I n Hawkings original paper, it was only
shown that the number of particles produced in various modes had a Planck
distribution. Wald performed a full quantum field theory62 analysis using the
Bogolubov transformation method and showed that the state was truly thermal in
all respects. This result was rederived by Hawking from a slightly different point
of view and by Parker, who simplified the analysis considerably by neglecting
the scattering of the particles by the external gravitational field. I t should be
emphasized that the precise statement of Hawkings result is that for sufficiently
late retarded times u, the collapsing system looks like a black body with tem-
perature given by Equation 6.9 and with the back-scattering effects in addition.
There are various comments worth making at this stage.
1 . The late u results are independent of the fine details of the collapse process.
2. The thermodynamical implications are discussed in this volume in the
article by D.W. Sciama. However, to get a feel for the size of the effect, note
that a solar mass Ma black hole gives a temperature of around lo- deg K . The
critical mass is that for which, if the collapse had occurred at the big-bang, the
matter would have just now evaporated. This is 10 g, corresponding to a tem-
perature of about 10 deg K or 260 MeV. Complete calculations of production
rates for various spins (Hawkings method extends to spin 4, I , 2) are quite com-
plicated because of the difficulty in computing the classical back-scattering effects.
A numerical investigation of this problem has been done by Page. 57
3. If the black hole is charged or rotating there is additional particle produc-
tion in modes tending to remove the charge and angular momentum. This is the
quantum analog of the classical superadiance phenomenon and was in fact antici-
pated before Hawkings ~ o r k . ~ ~ ~ ~ . ~ ~
Isham: Q u a n t u m Field T h e o r y 131
T H EE N E R G Y - M O M E N T TENSOR
UM
In the discussion so far, no major quantum field theory problems have been
encountered. T h e use of in/out regions and Bogolubov transformations has lead
to mathematically unambiguous particle-production rates. However, particle pro-
duction is not the only quantum effect o f a background field. W e know, f o r
example, that an external electromagnetic field can polarize the vacuum, leading
to experimentally veriliable effects. The analog in the present case will be found by
studying the quantized energy-momentum tensor T,, which is in many respects
the gravitational equivalent of the electromagnetic current. This object will be
formally divergent, leading t o the problems of regularization a n d renormaliza-
tion. Let me start by listing the main reasons for o u r interest in T,y:W-Bh
1. The particles produced by the gravitational field carry energy a n d mo-
mentum which must react back o n the field in some way. To a first approxima-
tion, this reaction is presumably described by the equations:
G,, = < T,, > (7.1)
At a deeper level, there are many physicists w h o believe that it is fundamentally
wrong to quantize the gravitational field itself. T h e system of equations, 7.1, is
then usually regarded as being exact, with the state 1 > chosen partly in some
phenomenological way and partly on the grounds of consistency. T h e feasibility
of genuinely setting u p a meaningful theory along these lines has never been
132 Annals New York Academy of Sciences
K = s
L:
K,TdS, (7.2)
It is fashionable these days to ignore such problems, but I think this is dangerous.
If for example the energy-momentum tensor is being evaluated in a Robertson
Walker universe, then either this is simply regarded a s a n intellectual exercise or
else it is seriously intended t o be used in the study of the real big-bang. In the
latter case i t is difficult t o see how such conceptual problems can be neglected.
3. The main technical problem is the divergence of T,, and its matrix elements
< >.T,,, These divergences appear in the products of fields defined a t the same
space-time point and requirc regularization. Ideally, this involves introducing a
parameter into the definition of the product rendering the answer finite, the diver-
gence being recovered when some limit is taken in the parameter value. This
enables control and isolation of the infinities a n d then, if possible, their removal
by renormalization of the gravitational equations of motion. In practice, this will
not only renormaliLe the gravitational coupling constant but will also introduce a
cosmological constant and terms on the left hand side of Einsteins equations
involving the square of the Riemann tensor. Hopefully, the finite residue
<T,, >reg will be unambiguous up to finite renormalizations. (It is usually suf-
ficient to find the correct regularized vacuum expectation value t o fix all the regu-
larized matrix elements. Of course, this does require choosing a specific vacuum
state.) Irrespective of the scheme adopted, o n e usually rcquircs that this
regularized a n d renormalized matrix element satisfy-
(7.3)
2. < T,,(x)> I c g is a tensor on the space-time manifold. T h e precise meaning
and implications of this requirement are subtler than one might think.
Normal Ordering
Strictly speaking this is not a regularization scheme, since rather than isolating
the divergent terms it just removes them wholesale. T h e justification for this lies
in the non-measurability of the absolute zero of energy in special relativity. How-
ever, in general relativity the absolute value of energy does have a meaning via
the Einstein equations, a n d the procedure is highly questionable. Even in flat space
this approach misses the Casimir effect, since it automatically gives a vacuum
energy of zero. In curved space-times the Hamiltonian diagonalization method
may be regarded as employing time-dependent normal orderings. Often Equa-
tion 7 . 3 is violated in this ~ c h e m e . ~
I34 A n n a l s N e w York A c a d e m y of Sciences
If the quantum field is decomposed into modes {1;1 as in equation 3.8, then
the matrix elements of T,, can also be written as a (divergent) sum or integral over
the index i . This divergent expression may be regularized by truncating the sum or
integral. In conventional quantum electrodynamics, an analog is the use of a
momentum space case cutoff. This procedure violates electromagnetic gauge in-
variance, and one might anticipate analogous problems with general covariance
in the gravitational case. The method has been used with some success to find the
vacuum polarization in static ~ p a c e - t i m e s ~ and
* ~ often forms a subsidiary part
of other regularization schemes.
De WittlSch~ i n g e r ~ ~ ~ ~ ~
G F ( x . Y )= i
e-12S
<x,s 1 y,O>ds (7.4)
Adiabatic Regularization
This has been extensively employed by Fulling and Parker2%3v17 in their investi-
gations of cosmological models. Using for illustrative purposes the Robertson
Walker metric of Equation 4.1, their method consists in replacing a ( t ) by a ( t / T )
and expanding in powers of T-I. (This is closely related to Parkers adiabatic
particle definition, mentioned in an earlier section). I f the first few terms i n the
expansion of T,,, in T-l are removed, then the remainder is finite. Again, there is
not complete control over the divergences, and there is also trouble with general
c ~ v a r i a n c e The
. ~ ~ technique can be related to the DeWitt/Schwinger method
above and also to the n wave regularization scheme of Starobinsky and Zeldo-
vich,26although the latter has a somewhat different conceptual grounding.
ported parallel to the point y i f covariant results are required. This is a gen-
uine regularization/scheme, with the parameter u giving full control over the
divergences. Its main drawback is that after the renormalizations have been per-
formed, the regularized remainder depends explicitly on the direction in which the
p o i n t y was split away from x. T h e exponents of this method happily d r o p these
terms, claiming to be selecting the physically meaningful residue. However,
some quantum field theorists find this procedure aesthetically disturbing and
oppose it. T h e iniquity is compounded by the fact that (in four dimensions) there
is not even a unique way of selecting the term t o be expunged. In specific models,
however, there is often a natural choice that satisfies Equation 7 . 3 and is con-
sistent with the unambiguous particle creation calculations. Indeed, many explicit
calculations of < Tpy>reg have been m a d e using this scheme, unlike most other
methods. Another prediction (in two dimensions) is the existence of conformal
a n o m a l i e ~ , ~ .a ~result which agrees both quantitatively a n d qualitatively with
*
dimensional r e g u l a r i ~ a t i o n . ~ ~ ~ ~ ~ ~ ~
An alternative t o dropping the direction-dependent terms is t o average them
out in some way. This approach has been adopted by Adler, Liebermann, and
Ng,@ who invented a scheme in which the averaging a t each space-time point
depends on the value of the Riemmann tensor and which does not give a con-
formal anomaly. O n e oddity of the method is that it requires the R @ term t o be
present in the equations of motion, even for massive scalar fields, if the conserva-
tion requirement of Equation 7.3 is to be met. T h e relation between these two
point-splitting schemes is unclear a t present.
Dimensional Regularizarion
T H EFALLING
SHELL
As a good example of the sort of results that can be obtained for the regu-
larized energy-momentum tensor, let us briefly consider the gravitational collapse
of a shell of matter. This has been considered in a two-dimensional model by
Davies, Fulling, and U n r ~ h . ~ .Inside
~ the shell the space-time is flat, whilst
outside i t has the Schwarzschild form
the interior and exterior coordinates being connected by the required arc-length
continuity across the shell. The minimally coupled massless scalar field is em-
ployed, and the regularized (point-splitting) energy momentum tensor, when there
is no collapse, is
(8.3)
u=t-r* (8.4)
r* = r + 2M$n(r/2M - I)
v = t + r *
Equivalently,
I s h a m : Q u a n t u m Field Theory I37
Thus, the static space-time curvature outside the shell produces a vacuum polariza-
tion. I f the shell is now allowed t o collapse, the only effect is the appearance of
an additional term in < T,, >, which for late retarded times u is
<T,,> = (768rM2)- (8.7)
This term is exactly that required to give the (Bogolubov transformation com-
puted) Hawking radiation a t future n u l l infinity and constitutes a most gratifying
agreement between the two methods. There a r e various comments to b e made
concerning this calculation:
I . Near I + the flux is dominated by the outward null flow of energy described
by < T,, >. Near the horizon the main effect is an inward flow of negative
energy coming from < T, >. I n a back-reaction calculation, this would cause
the horizon t o shrink and can possibly be interpreted a s supporting Hawkings
original horizon tunnelling picture of the pair creation process. T h e form of
< > T,, can also be used, however, to support the rival picture76 in which the
radiation appears to come from the surface of the body, although this does invoke
a rather a d hoc split of the energy-momentum tensor into a sum of Hawking
radiation plus vacuum polarization.s2 This surface radiation picture also arises in
Boulwares comprehensive discussionny of a shell falling in a four-dimensional
space-time and in Gerlachs totally different treatment of the same ~ y s t e m . ~
2 . Both < T, > and < > T, are regular across the event horizon.
3. An observer falling into the black hole sees very little. T h e blue-shifted
Hawking flux is cancelled by the negative energy vacuum polarization (using the
a d hoc split mentioned a b ~ v e ) . ~
4. The treatment has been extended by D a ~ i e s t o~ a general falling body, and
he finds additional exponentially damped ( e +) particle fluxes coming from the
surface and interior of the body. They may be regarded as transient effects pro-
duced by the time dependence of the metric in the collapsing matter.
Very many other models have now been investigated using these T, methods,
and some further details are contained in Davies article in this volume.
T H EE T E R N A LBLACKH O L E
'h
InRegionR U = (1-r/2M) exp(r-t)/LM
V = (1 -r/2M)'exp ( r +t ) / L M
FIGURE 3 . The Penrose diagram of the eternal black hole.
integrating over geometries of this type. It is also possible, of course, that such
metrics were produced in the big-bang.
In the regions R and L the line element is
(9.1)
and normal modes for the scalar field equation V "a,d = 0 may be written as
= (2* I w I ) - i ' 2 e - i w ' y t m ( 0d, ) x u d m ( r ) (9.2)
where
;4 = R$ = ( 2 ~ w ) - * e - ~ y ~ ~ ( O , 4 ) ~in R (9.7)
=o in L
ww RCQ
CQ
L Lq
FIGURE 4. The proper time motions of various wave packets employed in the eternal
black hole.
I40 A n n a l s New York A c a d e m y of Sciences
The different quantization schemes are usually specified by their vacuum states,
the principal ones being as follows:
The q - v a ~ u u m ~ ' ~ ~ *
This is the obvious choice which uses the modes 9.7 to 9.10 as they stand. The q
refers to the killing vector d / d t used to specify positive frequency, which i n this
case means simply w > 0. The resulting quantization scheme has the following
properties:
I . There is no particle production in R or L , which is natural i n this context, as
the metric is static in those regions.
2. Both the Feynman Green's function GF (x, y ) and the regularized energy mo-
mentum tensor (in two-dimensional models) have singularities on the future
horizon h + . This is closely related to the nullness of the Killing vector q d/dt
on this surface.
3. G F ( x , y )= 0 if x and y are in the L and R regions, respectively. In Minkowski
space the Feynman function does not vanish for spacelike separated points, so
from this point of view the horizon crossover point does possess a rather odd
property. (See, however, the original papers by B o ~ l w a r efor ~ ~a -discussion
~~ of
this).
The [-vacuum
Unruh6' has argued that i t is more natural to choose those modes that come in
from the past horizon h - to be positive frequency with respect to the canonical
affine parameter U on the null generators of h - . ( U and V are the usual null
Kruskal coordinates). Positive frequency can be usefully described in terms o f
complex analyticity o f t h e solution. Thus, if
(9.11)
the the wave packet d ( U ) contains positive frequency modes (in the sense that
$ ( w ) has its support i n w > 0) if and only if it can be analytically continued to all
complex values of U whose imaginary parts are negative.
Hence, we look for combination of the horizon modes R $ and ,.$ that have
this analyticity in U on the surface V = 0. Now, in the vicinity of h - ( V = 0) we
have the behavior
- - (-
R&J U)rw/x when U < 0 (9.12)
L$ - (U)ylK when U > 0 (9.13)
where K = c 4 / 4 G M is the surface gravity. The correct linear combination (which
also have positive Klein-Gordon inner product) is then readily seen to be","
&, = cosh + sinh C,I.$wfy* (9.14)
(9.15)
where w > Oand tanh C, = e - r w / x (9.16)
These modes, together with R 6 a n d L & are used to define a new vacuum state de-
noted I O;[> (The [ refers to the killing vector d / d , defined on the past horizon.)
I s h a m : Q u a n t u m Field Theory 141
U = exp S
*, . -
~
Cw(RaoLa'o- RawLaw)
* (9.21)
W > O
ACCELERATED
OBSERVERS IN M I N K O W S KSPACE-TIME
I
These coordinates are well adapted for studying the present problem since an ob-
server moving with uniform acceleration u (in the sense of special relativity) travels
along the curve z = 1 / a .
We consider the scalar wave equation
(a + /2)@
= 0 (10.3)
which has the separable solutions, with positive Klein-Gordon inner product,
M@E,k = ( 2 n E ) - ' / 2 e - ' E ' e ' k xE, = (p2 + k2)'I2 > 0 (10.4)
I f @ is expanded (equation 3.8) in these modes as
@ = s
k,E>O
(M'E.k M@E,k + M'E.k M@g.k 1 (10.5)
FIGURE
5 . A uniformly accelerating observer in Minkowski space.
z-
d", ( d):
z- + (z2p2 + W * ) X = 0 (10.8)
144 A n n a l s N e w York A c a d e m y of Sciences
then the Minkowski vacuum I0;M> is precisely the vacuum associated with
Equation 10.16:
a, IO;M > = a-, 10; M > = 0 (10.17)
These results are again reminiscent of the eternal black hole. The killing vector
8 / 8 1 becomes null on the horizons t = f x (it generates a Lorentz boost) and
thus I 0;Rin >
is analogous to I 0;[ >. On the other hand, the Minkowski vac-
uum 1 0; M >
being manifestly time symmetric is similar to I 0 ; u >. The same
type of analysis using Bogolubov transformation gives
10; M > = N exp S e-"wRa:,k
w>o
1O;Rin >, (10.18)
k
Now suppose the state ofthe system is prepared as I O;M>; what does the ac-
celerating observer see? lfwe decide that he will naturally use equipment that mea-
sures or responds to the Rindler definition of particle, then equation 10.18 ex-
presses the desired decomposition of the given state in terms of Rindler states.
Because of the event horizons at f = *
x the accelerating observer in R has no
information about the region L and. just as in the case of the eternal black hole, he
can only measure observables constructed from the operators R a w , kRu:,k.
, Trac-
ing over the L operators once again leads to a thermal equilibrium state. The tem-
perature may be found by using the black hole result and noting that an observer
at z = I/a has proper time r / a and hence energy a m . Thus, the apparent tem-
perature is
(10.19)
2a
GREENS FUNCTIONS
AND ANALYTICITY
determines, via the use o f t h e LSZ formalism, all possible results of scattering ex-
periments. Indeed, by invoking a modification of the Wightman reconstruction
theorem, the complete quantum field theory may be obtained from this single func-
t i ~ n . This
~ suggests a possible approach to quantizing a linear field in a curved
background in which one tries not to find the vacuum state but rather to construct
a two-point distribution that can bc taken as the Feynman function. The quantum
field is then only a secondary construct. )
One obvious advantage of this schcme is that i t deals with C-numbers rather
than operators. There are two related ways of trying to construct a suitable distri-
bution:
146 Annals New York Academy of Sciences
( 1 1.5)
=iJ,' -I($t-M)J
<x;s Iy;o> ( 1 1.6)
where
<x;s 1 y,O > = <x I e-ls" I y > (11.7)
and the limit t 4 0 is understood. It must be emphasized that these expressions are
very heuristic. The operator (u + m 2 - it)-' would exist as the resolvent of
(0 + m 2 )i f the latter were essentially self adjoint on some Hilbert space (to which
the generalized Dirac state I x >
forinally belong^).'^^,'^' How ever, there is no
reason in general why this should be true, and the meaning assigned to equation
11.6 needs to be carefully stated.
The quantum mechanical propagator < x ; s I y ; O >
obeys the proper time
covariant Schrodinger equation
over all paths Z ( T ) connecting the point y with the point x at proper times. This
is the starting point o f t h e Hartle-Hawking quantization,lWand the main problem
is defining the Functional integral. The expression in Equation 1 I. 10 is rather
formal in the sense that dp(z) represents what would be the infinite dimensional
version of Lebesque measure if it existed, which unfortunately it doesnt! (Actually
even if it did exist there is, in the curved space case, the possibility of non-trivial
weight factors.) Generally speaking, the integrand and dp(z) need to be taken
together in attempting to define a genuine measure, and the presence of the i in
the exponent in Equation I I .I0 renders this impossible. One solution might be the
use of the theory of distribution measures developed by Cecille DeWitt176over the
last few years. Alternatively, one may observe that if the I were replaced by
minus one (by replacing T by ir, say) and if g p U had signature (+, +, +, +) then
equation 1 I .10 would be a Gaussian integral, which can be rigorously defined. This
is a standard procedure i n modern quantum field theory and suggests again the
desirability of complexifying the space-time manifold. The crucial step is finding a
subspace on which Equation I I . 10 is well defined and from which the physical
space-time can be reached by analytic continuation (s will also be complex).
Hartle and Hawking did not actually compute the path integral. They instead
observed that the Riemannian-space-based propagator would obey a diffusion
equation rather than the Schrodinger type of Equation in I I . 8 . The asymptotic be-
havior in s of a solution to such an equation is well known, and this information
enables the proper time integral to be properly defined. By an elegant series of
arguments, they show that the physical Greens function G F ( x , y ) for an eternal
black hole has the following complex analytic behavior for fixed y in the region R :
For x t h G , . ( x , y ) is analytic i n the lower half V plane
(11.11)
For x t h - G F ( x , y )is analytic in the upper half U plane
where U and V are the usual global Kruskal null coordinates. Furthermore, since
the metric is complex analytic in U and V , G F ( x , y ) is also complex analytic in
U and V (coordinates of the point x) for fixed y (except when x and y can be joined
by a null geodesic). The derivation of these results is non-trivial, and the reader
is urged to study the original paper.Im
The boundary conditions in 11.11 might have been postulated ab initio. How-
ever, the present approach demonstrates clearly the way in which the resulting
Feynman propagator is the analytic continuation of the unique Riemannian space
function. The boundary conditions in 11.11 may be interpreted as the statement
that G,(x, y ) propagates positive frequency waves (with respect to the a t h e
coordinate U on h -) forwards in time and negative frequencies (with respect to the
affine coordinate V on h + ) backwards in time. There is, of course, an analogous
statement for the usual Minkowski space propagator with the horizon boundary
conditions replaced by those at null infinity. It is fairly clear that the vacuum state
that can be reconstructed from the Hartle-Hawking function is the u-vacuum. The
quickest way of showing this is via the Gibbons/Perry thermal Greens function
approach discussed be lo^.^'^^'*' I
The Hartlz-Hawking propagator has most interesting analytic properties with
respect to the static Schwarschild coordinates 1 and r. For example, consider the
points x and y to lie in regions F and R, respectively, with y fixed. The null co-
Isham: Q u a n t u m Field Theory I49
ordinates o f x c F a r e
(11.12)
( 1 1.13)
1
-- =
a
-toz + [ I 2 + [2* + t3* + t4* ( 11.15)
invariant under the group SO(4, I ) . I t is a solution of the Einstein equations with
a repulsive cosmological constant A = 3 a 2 . This leads to the presence of event
horizons that, unlike those in the black hole, may be different for different inertial
observers. T h e Hartle-Hawking method leads to a specification of G,(x, y ) in
terms o f its analytic behavior in x on the past and future event horizons, with
y fixed.Il4 The end result is thermal radiation with a temperature of k T = a / 2 ~
which, with the known limits on A for the actual universe, is around 10-* deg K .
Gibbons and H a w k i n g i i 4regard this as a particular case of a general phenomenon
associated with any horizon; namely, the intrinsic loss of information behind the
horizon generates entropy, which manifests itself in the presence of thermal
radiation.
The alternate quantizations based on the SO(4, I ) group d o not predict such
radiation, although it is debatable whether in any genuine operational sense any
of these different quantization schemes can be distinguished from each
other,104~IOX.11j.116
A N D FUTURE
CONCLUSIONS DEVFIDPMENTS
tempting to unify general relativity and quantum mechanics. At the technical level,
we can perhaps be more constructive. There are a number of possible approaches
to quantizing the gravitational field, usually divided into canonical and co-
variant. Most of the work on quantum field theory in an external gravitational
field has been performed in a covariant framework with the fields defined on the
full space-time manifold. For this reason it does not really add much to our
understanding of the canonical schemes based, as they invariably are, on a three-
dimensional spatial manifold. On the other hand, there clearly are connections
with the covariant approaches, especially those based on a Feynman functional
integral over field histories. Thus, for example, if we were perturbing about an
eternal black hole background, the Hartle Hawking work suggests that the
u-vacuum propagator should be used in constructing the interacting Feynman
diagrams. However, the conventional particle physics approach would un-
doubtedly give the 0-vacuum results. It is sometimes suggested that it is this
neglect of the thermal graviton radiation that is responsible for the apparent non-
renormalizablity of the theory. Personally, I think this i s unlikely, as the com-
pactification of the time coordinate associated with the use of thermal Greens
functions does not dramatically change the high energy behavior, although it will
certainly have an interesting effect on the infrared divergcnce~.~However, to
the best of my knowledge no one has checked this in detail, which would certainly
be worth doing.
For the specific case of an eternal black hole, there is no difficulty in imple-
menting the radiation results-the thermal Greens function can simply be inserted
by hand. However, for a general multihorizon manifold, this would be very difi-
cult, and in particular perturbation theory about Minkowski space cannot be
used in attempting to set the boundary conditions on the horizons (which are
global objects). One appealing answer to this problem can be found i n the con-
siderations of the previous section; namely, define the functional integral
o n a suitable subspace of a complex manifold and only continue to physi-
cal space-time at the end of the calculations. This procedure should, by analogy
with the Hartle-Hawking analysis, provide just the correct boundary conditions
on the horizons (which reappear when the continuation is made). I n modern
constructive quantum field theory such a scheme has already been used with great
success. Of course, i n the case we are discussing, the problems involved in finding
a suitable complexification of the space-time manifold will becomc of paramount
importance. I t is tempting to postulate that a specific complex manifold should be
chosen with a specified real physical subspace and that the measure in the func-
tional integral be defined on the space of holomorphic metrics only. This is a non-
trivial restriction since many real metrics (like many real analytic functions) are
not complex analytic. Thus, it is encouraging to note that on complex manifolds,
the space of holomorphic functions is naturally equipped with a nuclear topologi-
cal vector space structure. Thus, its dual space of analytic functionals can carry
measures i n exactly the same way as the usual spaces of distributions employed in
conventional quantum field theory.
I f euclideanization can be pushed through, i t opens up many interesting
possibilities, ranging from the abstractions of Markoff fields to the utilization of
Heaven, Penroses nonlinear gravitons, and the not unconnected currently
fashionable topic of instantons.
I52 Annals New York Academy of Sciences
These remarks are, of course, speculative, but one thing is certain-the explo-
sion of interest over the last three years in quantum field theory in curved space-
times will continue unabated in the next.
A CK N O w L E D G M E N T
R EF E R E N C E S
Isotropic Cosmologies
31. BOWERS,R . L. & R . L. ZIMMERMAN. 1973. Phys. Rev. D 7(2): 296 325.
32. C A R T ~ R B.,, G. W. GIBBONS, P. N.C. LIN & M . J . P ~ R R Y1976. . Astron. Astrophys.
52: 427.
33. DAMOUR, T. & R. RUFFINI. 1975. Phys. Rev. Lett. 35(7): 463-466.
34. DAVIES, P. C . W . & J. G . TAYI-OR. 1974. Nature 250: 37-38.
35. DAVIES, P. C. W. 1975. J . Phys. A 8(4): 609-616.
36. DAVIES, P. C. W. 1976. Mon. Not. R. Astron. Soc. 177: 179-190.
37. D A V I E S . P. C. W . 1976. The Thermodynamic Theory of Black Holes. King's College,
London. Preprint. Proc. R . Soc. In press.
38. FORIJ, L. 1975. Phys. R e v D I I (12): 3370 ~3377.
39. FoI.l.IN(i, S. A. 1976. Alternativc vacuum states i n stiilic space-times with horirons.
King's Collegc, London. Preprint.
40. GEKI.ACH, U . H . 1975. Why is a black hole hot? Ohio State University. Preprint.
41. GFKLACH. U . H . 1976. Phys. Rev. D 14(6): 1479 1.508.
42. G E R L A C t I . u . H . 1976. Statistical mechanics of radiation reacting against an isolated
black hole. Ohio State University, Columbus. Preprint.
43. GIBBONS, G . W . 1975. Commun. Math. Phys. 44: 245-264.
44. GIBBONS, G. W. 1977. Black hole dyons need not explode. Phys. Rev. I n prcss.
45. GIBBONS, G . W. 1977. Black holes, magnetic fields and particle creation. Mon. Not.
R. Astron. Soc. In press.
46. GIBBONS, G. W. 1975. Commun. Math. Phys. 45: 191-202.
47. H A W K I N GS., W. 1974. Nature 248: 30-31.
48. HAWKING, S. W. 1975. Commun. Math. Phys. 43:199 220.
49. H A W K I N GS., W. 1975. In Quantum Gravity- An Oxford Symposium. C . J . Isham.
D . W . Sciama & R. Pcnrose, Eds.: 219-267. Oxford University Press. Oxford,
England.
50. H A W K I N GS., W. 1976. Breakdown of predictability in gravitational collapse. Cali-
fornia Institute of Technology, Pasadena. Preprint.
51. HAWKING, S. W. 1976. Phys. Rev. D .
52. DYSON,F. J . 1976. Breakdown of physics in gravitational collapse. IAS Princeton.
Preprint.
53. , ISRAEL., W. 1976. Phys. Lett. A 57(2): 107 I IO.
54. KUNDT,W . 1976. Nature 259: 30-31.
55. PAGE,D. N . 1976. Phys. Rev. D 13(2): 198-206.
56. PAGF,D. N . 1977. Particle emission rates from a black hole. 11. Massless particles
from a rotating hole. Phys. Rev. I n press.
57. PAGE,D. N . 1977. Particle emission rates from a black hole. 111. Charged leptons
from a nonrotating hole. Phys. Rcv. I n press.
58. PARKFR, L. 1976. Phys. Rev. D 12(6): 1519- 1525.
59. UNRuti, W. G. 1974. Phys. Rev. D Io(10): 3194 3204.
60. U N R U HW. , G. 1974. Proc. R. Soc. A338:517 525.
61. UNRuii, W. G. 1976. Phys. Rev. D 14(4):870-892.
62. WAI.D,R. M . 1975. Commun. Math. Phys. 45:9-34.
63. WALD,R. M . 1975. Phys. Rev. D 13(12): 3176 3182.
69. CmisrENSEN, s. M . 1975. Covariant coordinate space methods for calculations in the
quantum theory of gravity. Ph.D. Thesis. University of Texas, Austin.
70. CHRISTENSEN, S. M. 1977. The vacuum expectation value of the stress tensor in an
arbitrary curved background: The covariant point-splitting method. Phys. Rev.
I n press.
71. CHRISTENSEN, S. M . & S . A. FULLING. 1976. Trace anomalies and the Hawking effect.
Kings College, London. Preprint.
72. DAVIES, P. C. W . & s. A. FULLING. 1976. Quantum vacuum energy in two dimen-
sional space-times. Kings College, London. Preprint.
73. DAVIES, P. C . W . & S. A. FULLING. 1976. Radiation from a moving mirror and black
hole evaporation. Kings College, London. Preprint.
74. DAVIES, P. C. W. & W. G . U N R U H1976.. Neutrino stress tensor regularization in two
dimensional space-time. Kings College, London. Preprint.
75. DAVIES, P. C. W., S . A. F U L L I N&G W. G . UNRUH.1976. Phys. Rev. D 13(10): 2720-
2723.
76. D A V I E SP., C . W . 1976. Proc. R . Soc. A351: 129-139.
77. DAVIES, P. C. W . , S. A. FIXLING, S. M. CHRISTENSEN & T. S . BUNCH.1976. Energy-
momentum tensor of a massless scalar quantum field in a Robertson-Walker
universe. Kings College, London. Preprint.
78. DESER, S., M. J . D U F F &C . J. ISHAM.1976. Nucl. Phys. B 111:45-55.
79. DUFF,M . J . 1975. In Quantum Gravity-An Oxford Symposium. C . J. Isham, D . W .
Sciama & R. Penrose, Eds.: 78-135. Oxford University Press.
80. BROWN,L. S. 1977. Stress tensor trace anomaly in a gravitational metric: scalar
fields. Phys. Rev. 15(6): 1469- 1493.
81. BROWN,L. S . & J. P. CASSIDY. 1977. Stress tensor trace anomaly in a gravitational
metric: general theory, Maxwell field. Phys. Rev. In press.
82. FULLING, S. A. 1976. Radiation and vacuum polarisation near a black hole.
Kings College, London. Preprint.
83. FULLING, S. A. & P. C . W. DAVIES. 1976. Proc. R. Soc. A 348: 393-414.
84. UTIYAMA. R. & B. S. DEWITT.1962. Phys. Rev. 3(4): 608-618.
85. D U N C A NA. , 1976. Conformal anomalies in curved space-time. IAS Princeton, Pre-
print.
86. WAI.D,R. M . 1977. The back reaction effect in particle creation in curved space-time.
Enrico Fernii Institute. Preprint.
162. SCI~IFF, L. J., H . SNYDER & J . W E I N B E R G . 1940. O n the existence of stationary stales
of the rnesotron field. Phys. Rev. 57: 315 318.
163. S C I I R O E R , B., R. S E l l E R & K . SWII!CA. 1970. Problems of stability for quantum fields
in external time dependent potentials. Phys. Rev. D2: 2927 2937.
164. SCIiROER, B. & K . SWIE(,A. 1970. Indefinite metric and stationary external interactions
of quantized fields. Phys. Rev. D Z(12): 2938 2944.
165. S E l l . t R . R . 1972. Quantum theory of particles with spin Lero and one half in external
fields. Commun. Math. Phys. 25:127 151.
166. SE IL . E R , R. 1971. Remarks on the theory of particles with spin up to one i n external
fields. In Troubles in the external field problem for invariant wave equations. A. S.
Wightman, Ed.: 23 43. Gordon and Btach, N w York.
167. SEGAI., I . E. 1959. Foundations o l t h e theory of dynamical systems of infinitely many
degrees of freedom, I . Mat. Fys. Medd. Dansk. Viel. Selsk. 31(12): I~ 39.
168. StciAi, I . E. 1961. Foundations of the theory of dyiiamical systems of infinitely many
degreesoffreedoin. 11. Can. J . Math. 13: I - I X .
169. SFGAI.. I . E. 1962. Foundations o f t h c theory o l dynamical systems of infinitely many
degrees of freedom. I l l . Illinois J. Math. 6: 500b.523.
170. S F a i . , I . E. 1967. Representations of canonical commutation relations. Ifr Applica-
tions of mathematics t o problems in theoretical physics. F. Luriat. Ed.: 107 170.
Gordon and Breach. N K WYork.
171. S O M M ~ R F I EC. I DM, . 1974. Quantiiation on space-time hypcrboloids. A n n . Phys. 84:
285 302.
172. AB E RSE., S. & B. W . LEF. 1973. Gauge theories. Phys. Rep. 9C(I): I 141.
173. BEIUISSARI), J. 1975. Quantized fields in interaction with extcrnal fields. Coiiiniuii.
Math. Phys. 41: 235-266.
174. BROWN,L. S . & G . J. M A C L A Y1969. . Phys. Rev. 184: 1272-1279.
175. C A P R I , A . Z . , G . L A H O N T D. ~ , M E N O N& A. S H A M A L Y1971.
. The reconstruction
theorem with external sources. Nuovo Ciniento B 3(2): 233 241.
176. DEWITT, C. 1972. Fcynnian's path integral. Dcfinition without limiting procedure.
Commun. Math. Phys. 28: 47-67.
177. E N G l t R T , F., c. TRUFFING & R. CASTMANS. 1976. Conformal invariance in quantum
gravity. University of Leuvcn, Belgium. Preprint.
178. SckiwiNCER, J . 1954. The theory of quantized fields. V . Phys. Rev. 93(3): 615 ~ 6 2 X .
179. W E I N B k R G , S. 1976. Critical phenomena for field theorists. Lecture notes. Harvard.
Other Papers
180. CARTER, B. 1973. I n Black Holes. Les Houches 1.ccturc.s. C. DeWitt. Ed. Gordon and
Beach 1973.
181. S A I . A M , A . & J . STRATllDkE. 1976. Hndronic temperature a n d black solitons. ICTP
Preprint IC/76/107.
182. Coi.t.rss, J . C. & M. J . P E R R YNeutrons
. or asyn1plotic;ill) frrc quarks. 1975. Phys.
Rev. Lett. 34: 1353- 1357.