Quantum Field Theory in Curved Spacetime
Quantum Field Theory in Curved Spacetime
Quantum Field Theory in Curved Spacetime
Robert M. Wald
Abstract
We review the mathematically rigorous formulation of the quantum theory of
a linear eld propagating in a globally hyperbolic spacetime. This formulation is
accomplished via the algebraic approach, which, in essence, simultaneously admits
all states in all possible (unitarily inequivalent) Hilbert space constructions. The
physically nonsingular states are restricted by the requirement that their two-point
function satisfy the Hadamard condition, which insures that the ultra-violet be-
havior of the state be similar to that of the vacuum state in Minkowski spacetime,
and that the expected stress-energy tensor in the state be nite. We brie
y review
the Unruh and Hawking eects from the perspective of the theoretical framework
adopted here. A brief discussion also is given of several open issues and questions
in quantum eld theory in curved spacetime regarding the treatment of \back-
reaction", the validity of some version of the \averaged null energy condition",
and the formulation and properties of quantum eld theory in causality violating
spacetimes.
1
1 Introduction
The subject of quantum eld theory in curved spacetime is the study of the behavior
of quantum elds propagating in a classical gravitational eld. It is used to analyze
phenomena where the quantum nature of elds and the eects of gravitation are both
important, but where the quantum nature of gravity itself is assumed not to play a
crucial role, so that gravitation can be described by a classical, curved spacetime, as in
the framework of general relativity.
The main initial development of the theory occurred in the late 1960's, driven pri-
marily by the desire to analyze the phenomenon of particle creation occurring in the very
early universe. By 1969, one can nd the theory formulated in recognizably modern form
and applied to cosmology in the paper of Parker [1]. In the early 1970's, the theory was
applied to the study of particle creation near rotating and charged black holes, where
the discovery of classical \superradiant scattering" (analogous to stimulated emission)
strongly suggested that spontaneous particle creation should occur. This line of research
culminated in the analysis by Hawking of particle creation resulting from the gravita-
tional collapse of body to form a black hole [2]. It thereby was discovered that black
holes radiate as perfect black bodies at temperature T = =2, where denotes the
surface gravity of the black hole. This result solidied an undoubtedly deep relationship
between the laws of black hole physics and the laws of thermodynamics, the ramications
of which continue to be pondered today.
As a direct consequence of Hawking's remarkable discovery, there occurred in the
mid-to-late 1970's a rapid and extensive development of the theory of quantum elds
in curved spacetime and its applications to a variety of phenomena. A good summary
of this body of work can be found in the monograph of Birrell and Davies [3]. Further
important applications to cosmology were made in the early 1980's, as the methods and
results of quantum eld theory in curved spacetime were used to calculate the pertur-
bations generated by quantum eld
uctuations during in
ation. Many of these lines of
investigation begun in the late 1970's and early 1980's continue to be pursued today.
Although it would be more dicult to point to major historic landmarks, during the
2
past twenty years the theoretical framework of quantum eld theory in curved spacetime
has undergone signicant development, mainly through the incorporation of key aspects
of the algebraic approach to quantum eld theory. As a result, the theory of a linear
quantum eld propagating in a globally hyperbolic spacetime can be formulated in an
entirely mathematically rigorous manner insofar as the denition of the fundamental eld
observables is concerned.
My main goal here is to review the key developments leading to a mathematically
rigorous formulation of quantum eld theory in curved spacetime. (Much more detail
can be found in my recent book on this subject [4], to which I refer the reader for a
more comprehensive and pedagogically oriented discussion.) I will also brie
y review
some open issues and questions regarding the treatment of \back-reaction", positivity
properties of the expected stress-energy tensor, and the formulation and properties of
quantum eld theory in causality violating spacetimes. Notational conventions follow
those of [5].
form
is then independent of the choice of these coordinates.) An observable in classical
mechanics is simply a real-valued function on M.
This basic structure of the classical description of a system stands in marked contrast
3
to the corresponding quantum description. In quantum mechanics, the state of a system
at a given instant of time is described by a vector (or, more generally, a density matrix)
in an innite-dimensional, separable Hilbert space F . An observable is a self-adjoint
operator on F . Since all innite-dimensional, separable Hilbert spaces are isomorphic to
each other, the content of a quantum theory corresponding to a given classical theory
b
is completely specied by giving a map : Oc ! Oq, where Oc denotes the set of
classical observables (i.e., real valued functions on M), and Oq denotes the set of quantum
observables (i.e., self-adjoint operators on F ).
Since the structures of the classical and quantum theories are so dierent, it is far
b
from obvious how the map is to be determined. However, a key guiding principle arises
from the comparison of the classical and quantum dynamics. In classical mechanics with
Hamiltonian H , the rate of change of an observable, f , in the \Heisenberg representation"
is given by
df=dt = ff; H g (2)
where f; g denotes the Poisson bracket, dened by
ff; gg =
abraf rbg (3)
where
ab is the inverse of
ab. On the other hand, in quantum mechanics with Hamilto-
nian H^ , the rate of change of an observable, f^, in the Heisenberg representation is given
by
^ = i[f;^ H^ ]
df=dt (4)
Consequently, there will be a close correspondence between classical and quantum dy-
b
namics (for any choice of Hamiltonian) if the map, , can be chosen so as to satisfy the
\Poisson bracket goes to commutator" rule:
d
[f;^ g^] = iff; gg (5)
In fact, it is well known that even in standard Schrodinger quantum mechanics,
b
no map exists which implements the relation (5) on all observables (see, e.g., [6]).
However, the relation (5) can be implemented on a restricted class of observables. In
particular, consider the case where M has the structure of a symplectic vector space,
4
i.e., M is a vector space and the symplectic form
ab has constant components in a
globally parallel basis, so that it may be viewed as an antisymmetric, bilinear map
is dened by eq.(1).) In this case, we will refer to the 2n-dimensional vector space
of linear functions on M as the fundamental classical observables, since any classical
observable (i.e., function on M) can be expressed as a function of the 2n elements of a
basis for the linear observables. If we restrict attention to the fundamental observables,
then the Poisson-bracket-commutator relationship (5) can be implememented. Moreover,
b
in the sense explained below, it uniquely determines a map on these observables.
Before stating this result more precisely, it is useful to note that since
is non-
degenerate, any fundamental observable (i.e., linear function) on M can be written in
the form
(y; ) for some y 2 M, where by
(y; ) we mean the function f : M ! IR
dened by f (z) =
(y; z). In this notation, the Poisson bracket of the fundamental
observables is given by
f
(y1; );
(y2; )g =
(y1; y2)1 (6)
where 1 denotes the function on M which takes the value 1 at each point. Hence, the
Poisson-bracket-commutator relationship (5) for the fundamental observables is simply
[
^ (y1; );
^ (y2; )] = i
(y1; y2)I (7)
where I denotes the identity operator on the Hilbert space.
However, there are some potential technical diculties with eq.(7) because
^ (y; )
should be an unbounded operator and, hence, can be dened only on a dense domain,
so compositions (and, hence, commutators) are not automatically well dened. These
diculties are most easily dealt with by working with the following exponentiated version
of (7): For each y, dene the classical observable W (y) by
W (y) = exp[ i
(y; )] (8)
Then eq.(7) together with the self-adjointness of
^ (y; ) is formally equivalent to the
following Weyl relations
W^ (y1)W^ (y2) = exp[i
(y1; y2)=2]W^ (y1 + y2) (9)
5
W^ y(y) = W^ ( y) (10)
We shall view the Weyl relations as providing a mathematically precise statement of the
Poisson-bracket-commutator relationship, which avoids operator domain problems.
The key uniqueness result is provided by the Stone-von Neumann theorem: For a
nite dimensional symplectic vector space (M;
), any two strongly continuous (in y ),
irreducible representations of the Weyl relations (9), (10) are unitarily equivalent. Thus,
for a classical linear system with nitely many degrees of freedom, the \Poisson bracket
goes to commutator" rule determines in a natural, canonical way a corresponding quan-
tum theory in so far as the fundamental observables are concerned. This provides the
main justication for the standard choices of the Cartesian position and momentum op-
erators for Schrodinger quantum mechanics, as found in standard texts. Note, however,
that although all classical observables can be written as functions of the fundamental
observables, \factor ordering" ambiguities generally arise when one attempts to express
an arbitrary quantum observable as a function of the quantum representatives of the fun-
damental observables. Since the \Poisson bracket goes to commutator" rule cannot be
implemented for all observables, there does not appear to be any natural way to resolve
this factor ordering ambiguity. Thus, when a suciently general class of observables is
considered, it appears that there are many quantum theories corresponding to a given
classical theory. However, our primary interest here is in the fundamental observables,
where the Stone-von Neumann theorem does provide the desired uniqueness result for a
system with nitely many degrees of freedom.
All of the above basic structure present in a classical system with nitely many degrees
of freedom also is present in the theory of a classical eld propagating in a globally
hyperbolic spacetime (M; gab). Consider, for deniteness, a linear, Klein-Gordon scalar
eld, , satisfying
rara m2 = 0 (11)
This equation has a well posed initial value formulation, with the initial data consisting
of the pair of functions (; ) on a Cauchy surface , where = nara, with na the
unit normal to . We have a choice of precisely what class of functions to allow in phase
space M, but a particularly convenient choice is to require and to be smooth and of
6
compact support, i.e., we dene
The Klein-Gordon Lagrangian gives rise to a well dened, conserved symplectic structure
: M M ! IR on M, given by
Z
[(1; 1); (2; 2)] = (12 21) (13)
where Ef denotes the advanced minus retarded solution with source f , and the natural
identication of phase space with the space of solutions to the Klein-Gordon equation
is implicit on the right side of this equation. Consequently, the fundamental observable
(y; ) on phase space also may be viewed as the observable on solution space dened by
the spacetime smearing of solutions with a particular test function. The corresponding
quantum observable
^ (y; ) will then have an alternative interpretation of being a space-
time smearing of the Heisenberg eld operator. Our choice of fundamental observables
thus corresponds to the collection of Heisenberg eld operators smeared with arbitrary
test functions f .
We have seen above that the theory of a linear Klein-Gordon eld on a globally
hyperbolic spacetime has a phase space structure and a choice of fundamental observables
that parallels completely the case of a system with nitely many degrees of freedom.
Thus, the only relevant dierence which occurs when we attempt to construct a quantum
7
theory of a eld arises from the fact that now dim(M) = 1. However, this dierence
is crucial because the Stone-von Neumann theorem does not hold in innite dimensions,
and there are many inequivalent ways of implementing the \Poisson bracket goes to
commutator" rule for the fundamental observables.
The class of possible quantum eld theory constructions can be restricted in a natural
way by exploiting the analogy of a quantum eld with an innite collection of harmonic
oscillators, and performing a construction analogous to the standard construction of the
quantum theory of a harmonic oscillator using annihilation and creation operators. A
mathematically precise implementation of this idea can be achieved by introducing a
real inner product 2 on M which makes
be norm preserving, i.e., we introduce a
symmetric, bilinear map : M M ! IR such that
( 1; 1) = 41 sup j
( 1; 2 )j2
(15)
2 6=0 ( 2 ; 2 )
It can be shown (see [4]) that there always exist a wide class of 's satisfying (15). Given
a satisfying (15), we complete M in the inner product 2. We then use
to dene a
complex structure, so as to convert the completion of M into a complex Hilbert space
H. We construct the symmetric Fock space FS (H) based upon H by
FS (H) = C H (H
S H) ::: (16)
8
the time independent, quantum mechanical harmonic oscillator: For y1; y2 2 M, we
dene (y1; y2) to be the real part of the Klein-Gordon inner product of the positive
frequency parts of the solutions corresponding to the initial data y1 and y2. (A more
mathematically precise and complete description of this construction can be found in
section 4.3 of [4].) Furthermore, for this construction in stationary spacetimes, states
in FS (H) then have a natural, physical interpretation in terms of \particles", with the
one-dimensional subspace C corresponding to the \vacuum", with H corresponding to
the subspace of single particle states, with H
S H corresponding to the subspace of
two-particle states, etc.
However, in a general, non-stationary spacetime, there does not appear to any math-
ematically preferred choice of nor any physically preferred denition of \particles" {
although the Hadamard condition (discussed below) does pick out a preferred unitary
equivalence class of 's for spacetimes with compact Cauchy surfaces [4]. So the question
remains: Which should one choose? Equivalently, how should one dene the notion of
\particles" in a general, curved spacetime?
My view is that this question has roughly the same status in quantum eld theory as
the following question in classical general relativity: Which coordinate system should one
choose in a general, curved spacetime? This latter question is very natural one to ask (and
often is asked by begining students) if one has learned special relativity via its formulation
in terms of global inertial coordinate systems. It is \answered" by formulating general
relativity in a geometrical, coordinate independent manner, so that the question becomes
manifestly irrelevant. Similarly, the issue of how to dene \particles" is a natural one
to ask by those to whom quantum eld theory was presented as though it were a theory
of \particles". It also can be answered by reformulating the theory in a manner which
makes the question manifestly irrelevant.
In order to do this, we need a framework for quantum eld theory which (initially,
at least) simultaneously admits all states occurring in all (unitarily inequivalent) Hilbert
space constructions of the theory, so that no \preferred construction" need be specied
in advance. The algebraic approach accomplishes this goal in a mathematically elegant
and straightforward manner, and I shall now brie
y describe some of the main elements
9
of the algebraic formulation.
The key idea of the algebraic approach is to reverse the logical order in which the
notion of states and observables are specied. One rst introduces a notion of observables
and provides the set of observables with the mathematical structure of an abstract *-
algebra, A. (For some purposes, it is required that A have the additional structure of
being a C*-algebra.) On then denes a state to be a linear map ! : A ! C which satises
the positivity condition
!(AA) 0 (17)
for all A 2 A, as well as the normalization condition !(I ) = 1, where I denotes the iden-
tity element of the algebra. Although this notion of a state may appear to be drastically
dierent from the usual notion of a state as a vector (or, more generally, a density matrix)
in a Hilbert space, there is, in fact, a very close relationship between these notions: If
F is a Hilbert space and A is a sub-algebra of the C*-algebra, L(F ), of bounded linear
maps on F , then any density matrix state on F gives rise to an algebraic state, !, on A
by the formula !(A) = tr(A) for all A 2 A. Conversely, if ! is an algebraic state on the
C*-algebra, A, then the GNS construction shows that there exists a Hilbert space F , a
representation : A ! L(F ), and a cyclic vector 2 F such that for all A 2 A we have
!(A) =< j(A)j >. Thus, every state in the algebraic sense corresponds to a state
in the usual sense in some Hilbert space construction of the quantum eld theory. The
key advantage of the algebraic approach is that it allows one to consider, on an equal
footing, all states arising in all unitarily inequivalent Hilbert space constructions of the
theory.
To dene the theory of a quantum eld in a curved, globally hyperbolic spacetime
via the algebraic approach, we must specify a C*-algebra structure on the class of ob-
servables that we wish to consider. If one wishes { initially, at least { merely to have
the fundamental observables dened in the theory, then a C*-algebra structure can be
specied as follows: We start with the classical phase space M, eq.(12), with symplectic
structure
given by eq.(13). We then choose an inner product on M satisfying eq.(15),
and we perform the Hilbert space quantum eld theory construction outlined above to
dene the observables
^ (y; ) as self-adjoint operators on a Fock space F . Next, we dene
10
corresponding unitary operators W^ (y) by exponentiation (see eq.(8)). These operators
satisfy the Weyl relations (9), (10). The nite linear combinations of the W^ (y)'s then
have the natural structure of a *-algebra. Completion of this algebra in the norm pro-
vided by L(F ) denes the desired C*-algebra, A. In principle, A could depend upon the
choice of . If it did, then we would be back in a situation very similar to the situation
we faced when we had many unitarily inequivalent Hilbert space constructions of the
theory; we now would have many dierent possible constructions of the algebra of fun-
damental observables and no obvious means of choosing a \preferred" one. Fortunately,
this is not the case: Even when the inner products 1 and 2 yield unitarily inequivalent
Hilbert space constructions of a quantum eld theory, the algebras A1, A2 to which they
give rise are isomorphic as abstract C*-algebras. In other words, associated with the
symplectic vector space (M;
) for a Klein-Gordon (or other linear, bosonic) eld on a
given globally hyperbolic spacetime is a unique, well dened C*-algebra of fundamental
observables { known as the Weyl algebra { constructed in the manner described above,
using any choice of satisfying (15).
The specication of the Weyl algebra, A, as the C*-algebra of observables completes
the formulation of the quantum theory of a Klein-Gordon eld in an arbitrary glob-
ally hyperbolic, curved spacetime insofar as the denition of fundamental observables
is concerned. Note that the specication of a state on A corresponds, roughly, to the
specication of the complete list of all n-point distributions < (x1); :::; (xn) >, subject
to all of the conditions arising from the positivity requirement (17).
However, there are other observables besides the fundamental observables which one
may wish to consider. Most prominent among these is the stress-energy tensor, Tab, of the
quantum eld, since it is needed to describe \back-reaction" eects of the quantum eld
on the gravitational eld. In order to extend the construction of quantum eld theory
in curved spacetime to encompass the stress-energy tensor, one would like to enlarge the
Weyl algebra A to a new algebra of observables which contains elements corresponding to
the (presumably, spacetime smeared) stress-energy tensor Tab. A rst step in this regard
would be to dene T^ab as an operator-valued-distribution in Hilbert space constructions
of the theory. There are serious diculties with doing this because, formally, T^ab is
11
the product of two distributions at the same spacetime point, so some \regularization"
is needed to give it a mathematically well dened meaning. Some progress towards
dening T^ab as an operator-valued-distribution has been reported recently [7], but most
work to date has focused on the less ambitious goal of dening expectation values of the
(unsmeared) stress-energy tensor. (In a Hilbert space construction, this corresponds to
dening the stress-energy tensor as a quadratic form on the Hilbert space rather than
as an operator-valued-distribution.) Note that for an algebraic state !, a knowledge of
< Tab >! is precisely what is needed in order to determine whether the semiclassical
Einstein equation
Gab = 8 < Tab >! (18)
is satised, so a knowledge of < Tab > is all that is needed to analyze back reaction
within the context of the semiclassical approximation.
The main results of the analysis of < Tab > are the following (see [4] for more details):
(i) < Tab >! can be dened only for states, !, that satisfy the Hadamard condition,
which, in essence, states that the \ultra-violet" behavior of the state { as measured by
the short distance behavior of the two point distribution < (x)(x0) >! { is similar
in nature to the short distance behavior of the two-point distribution for the vacuum
state in Minkowski spacetime. (A precise denition of the \global Hadamard condition"
can be found in [8]; its equivalence to a \local Hadamard condition" was proven in [9].)
States which fail to satisfy the Hadamard condition are to be viewed as \physically
singular", in that their stress-energy is innite (or otherwise ill dened). The Hadamard
condition thus provides an important additional restriction on the class of states which
otherwise would be admissible when only the fundamental observables are considered.
(ii) The prescription for assigning an expected stress-energy, < Tab >! , to all Hadamard
states, !, is uniquely determined up to addition of (state independent) conserved local
curvature terms, by a list of physical properties that < Tab >! should satisfy. The
\point-splitting" regularization prescription satises these properties and thus provides
a completely satisfactory denition of < Tab >! , up to the local curvature ambiguity.
Thus, the status of quantum eld theory for a linear eld in a globally hyperbolic
spacetime, (M; gab ), may be summarized as follows: From (M; gab ) and the classical
12
symplectic structure of the eld, we can construct the Weyl algebra, A, of fundamental
observables (or a corresponding \anticommutator algebra" in the case of fermion elds).
States are then dened in the algebraic sense with respect to A. Nonsingular states
must, in addition, satisfy the Hadamard condition. For Hadamard states, < Tab > is well
dened up to the ambiguity of adding conserved local curvature terms.
We conclude this section by brie
y describing the statements of the Unruh and Hawk-
ing eects within the framework developed above. (Again, much more detail can be found
in [4].) In the Unruh eect, one considers the action of a one-parameter family of Lorentz
boosts on Minkowski spacetime. For deniteness, we normalize the boost Killing eld to
have unit norm on an orbit of acceleration a. We note that the orbits of the boost isome-
tries are timelike in the \right wedge" (as well as the \left wedge") region of Minkowski
spacetime. Furthermore, this \right wedge" region { when viewed as a spacetime in its
own right { is globally hyperbolic. Hence, there is a well dened Weyl algebra, AR, of
\right wedge" observables, which is naturally a subalgebra of the Weyl algebra, A, for
all of Minkowski spacetime. Hence, the restriction of the ordinary Minkowski vacuum
state, !0, to the subalgebra AR denes a state on the right wedge spacetime. The Unruh
eect is the assertion that this state is a thermal state at temperature T = a=2, in
the precise sense that it satises the KMS condition with respect to the notion of \time
translations" provided by the Lorentz boosts. In fact, this mathematical statement of
the Unruh eect was proven by Bisognano and Wichmann [10], independently of (and
simultaneously with) the paper of Unruh [11]. However, the physical interpretation of
this fact { namely, that an observer with acceleration a will \feel himself" to be immersed
in a thermal bath at temperature T = a=2 when the eld is in the Minkowski vacuum
state { is due to Unruh.
In the Hawking eect, one considers a spacetime describing the gravitational collapse
of a body to form a black hole. One assumes that the black hole settles down to a
stationary nal state, with constant surface gravity . The Hawking eect [2], [12] is
the assertion that any nonsingular (i.e., Hadamard) state asymptotically approaches, at
late times, a thermal state at temperature T = =2 with respect to the subalgebra
associated with solutions which \appear to emerge from the direction of the black hole".
13
In other words, a black hole \radiates" as a perfect black body; its physical temperature
is =2.
14
local expression in the conformal factor, so the back reaction equations are considerably
more tractable than in the 4-dimensional case.
The second open issue I wish to mention concerns energy conditions which may hold
for < Tab >. Even for a Klein-Gordon eld in Minkowski spacetime, it is easy to nd
quantum states for which < Tab > violates any of the local positive energy conditions
which hold for classical elds. However, it is possible that < Tab > may still satisfy some
nontrivial global positive energy conditions. The most interesting of these conditions is
the \averaged null energy condition" (ANEC), which states that for any quantum state
and for any complete null geodesic, we have
Z <T > kakbd 0 (19)
ab
where denotes the ane parameter of the geodesic and ka denotes its tangent. The
validity of ANEC is sucient for proofs of the positive energy theorem [17] and for
\topological censorship" [18], so many of the key results of classical general relativity
continue to hold when the pointwise energy conditions are replaced by ANEC. Although
ANEC holds for Minkowski spacetime [19], [20], it is known to fail for arbitrary curved
spacetimes (in 4-dimensions) [20]. Nevertheless, some recent research has indicated that
(i) ANEC may come \close enough" to holding to exclude wormholes with curvature
everywhere much smaller than the Planck scale [21] and (ii) when the semiclassical Ein-
stein equation (18) is imposed, a version of ANEC may hold wherein one \transversely
averages" (with a suitable smearing function) over null geodesics within roughly a Planck
length of the given geodesic [14]. Thus, it is possible that violations of ANEC may be
conned to regimes where the semiclassical approximation is not applicable.
The nal topic of present research which I wish to mention concerns quantum eld
theory on spacetimes with closed causal curves. Consider, for deniteness, a causality
violating spacetime which is \globally hyperbolic outside of a compact set", i.e., a space-
time obtained by modifying the metric of a globally hyperbolic spacetime in a compact
region so as to produce closed timelike curves within that region. It seems plausible that
{ at least within a suitable subclass of such spacetimes { one will have a well dened,
deterministic dynamics for classical elds [22]. Does there exist a similarly well dened
15
quantum eld theory on such spacetimes?
For linear quantum elds, one needs only the symplectic vector space structure of the
classical solution space in order to construct the Weyl algebra, so if the classical dynamics
is well behaved there should be no diculty dening fundamental observables and states.
In particular, it should be possible to obtain a well dened, unitary S -matrix describing
scattering processes. However, the association of a solution, Ef , to every test function
f (see eq.(14)) uses global hyperbolicity, so it is not clear that the eld observables can
still be interpreted as distributions on spacetime. Even if they can, it can be shown
that the condition of \F-compatibility" [23] must fail [24] at least at some points of
the chronology horizon, so the local behavior of the eld observables must be dierent
from that occurring in globally hyperbolic spacetimes. Furthermore, the local Hadamard
condition cannot be satised everywhere on the chronology horizon [24]. Thus, although
in certain examples it is possible to choose states where the stress-energy tensor remains
nite as one approaches the chronology horizon [25], the stress-tensor always must be
singular (or ill dened) at least at some points of the chronology horizon. Thus, it seems
far from clear that the quantum theory of linear elds can be sensibly dened even on
the class of causality violating spacetimes which are \globally hyperbolic outside of a
compact set".
The situation for nonlinear elds appears to be considerably worse in that conven-
tional perturbation theory yields an S -matrix which is non-unitary (in the sense of not
conserving probability) [26]. Alternative ideas for constructing quantum eld theory
on causality violating spacetimes are currently being actively pursued by a number of
authors [27]-[30].
4 Summary
The theory of linear quantum elds propagating on a xed, globally hyperbolic, curved
spacetime is completely well posed mathematically insofar as the denition of the fun-
damental observables (or, equivalently, the smeared eld operators) are concerned, and
< Tab > is well dened (for Hadamard states) up to local curvature term ambiguities.
16
This provides the necessary tools to investigate many phenomena of interest involving
quantum eects occurring in strong gravitational elds. Nevertheless, some interesting
and important issues remain open, such as the ones mentioned in the previous section.
One may hope that some of the insights obtained from the study of quantum eld
theory in curved spacetime { particularly, the close relationship between Killing horizons
and thermal states as seen in the Unruh and Hawking eects { will be helpful in guiding
the development of a quantum theory of gravity.
Acknowledgment
This research was supported in part by National Science Foundation grant PHY-
9220644 to the University of Chicago.
References
[1] L. Parker, Phys. Rev. 183, 1057 (1969).
[2] S.W. Hawking, Commun. Math. Phys. 43, 199 (1975).
[3] N.D. Birrell and P.C.W. Davies, Quantum Fields in Curved Space (Cambridge Uni-
versity Press, Cambridge, 1982).
[4] R.M. Wald, Quantum Field Theory in Curved Spacetime and Black Hole Thermo-
dynamics (University of Chicago Press, Chicago, 1994).
[5] R.M. Wald, General Relativity (University of Chicago Press, Chicago, 1984).
[6] P.R. Cherno, Hadronic Journal 4, 879 (1981).
[7] M. Kohler, Class. Quant. Grav. 12, 1413 (1995).
[8] B.S. Kay and R.M. Wald, Phys. Rep. 207, 49 (1991).
[9] M.J. Radzikowski, \The Hadamard Condition and Kay's Conjecture in (Axiomatic)
Quantum Field Theory on Curved Spacetime" Ph.D. thesis, Princeton University
(1992).
[10] J.J. Bisognano and E.H. Wichmann, J. Math. Phys. 17, 303 (1976).
[11] W.G. Unruh, Phys. Rev. D14, 870 (1976).
17
[12] K. Fredenhagan and R. Haag, Commun. Math. Phys. 127, 273 (1990).
[13] J.Z. Simon, Phys. Rev. D41, 3720 (1990).
[14] E.E. Flanagan and R.M. Wald, \Does Backreaction Enforce the Averaged Null En-
ergy Condition in Semiclassical Gravity?", to be published.
[15] C.G. Callen, S.B. Giddings, J.A. Harvey, and A. Strominger, Phys. Rev. D45, 1005
(1992).
[16] J.G. Russo, L. Susskind, and L. Thorlacius, Phys. Rev. D46, 3444 (1992).
[17] R. Penrose, R.D. Sorkin, and E. Woolgar, \A Positive Mass Theorem Based on the
Focusing and Retardation of Null Geodesics".
[18] J.L. Friedman, K. Schleich, and D.M. Witt, Phys. Rev. Lett. 71, 1486 (1993).
[19] G. Klinkhammer, Phys. Rev. D43, 2542 (1991).
[20] R.M. Wald and U. Yurtsever, Phys. Rev. D44, 403 (1991).
[21] L.H. Ford and T. Roman, to be published.
[22] J.L. Friedman and M.S. Morris, Commun. Math. Phys. (in press).
[23] B.S. Kay, Rev. in Math. Phys. (special issue), 167 (1992).
[24] B.S. Kay, M.J. Radzikowski, and R.M. Wald, to be published.
[25] S.V. Krasnikov, \On the Quantum Stability of the Time Machine", gr-qc/9508038,
to be published.
[26] J.L. Friedman, N.J. Papastamatiou, and J.Z. Simon, Phys. Rev. D46, 4456 (1992).
[27] H.D. Politzer, Phys. Rev. D46, 4470 (1992); Phys. Rev. D49, 3981 (1994).
[28] J.B. Hartle, Phys. Rev. D49, 6543 (1994).
[29] A. Anderson, Phys. Rev. D51, 5707 (1995).
[30] S.W. Hawking, \Quantum Coherence and Closed Timelike Curves", gr-qc/9502017,
to be published.
18