Stochastic quantization of scalar fields in Einstein and Rindler spacetime

This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2007 J. Phys. A: Math. Theor. 40 8545
(http://iopscience.iop.org/1751-8121/40/29/024)
Download details:
IP Address: 200.145.112.189
The article was downloaded on 28/01/2011 at 15:49
Please note that terms and conditions apply.
View the table of contents for this issue, or go to the journal homepage for more
Home Search Collections Journals About Contact us My IOPscience
IOP PUBLISHING JOURNAL OF PHYSICS A: MATHEMATICAL AND THEORETICAL
J. Phys. A: Math. Theor. 40 (2007) 85458568 doi:10.1088/1751-8113/40/29/024
Stochastic quantization of scalar elds in Einstein and
Rindler spacetime
G Menezes and N F Svaiter
Centro Brasileiro de Pesquisas FsicasCBPF, Rua Dr Xavier Sigaud 150, Rio de Janeiro,
RJ 22290-180, Brazil
E-mail: [email protected] and [email protected]
Received 16 April 2007
Published 3 July 2007
Online at stacks.iop.org/JPhysA/40/8545
Abstract
We consider the stochastic quantization method for scalar elds dened in a
curved manifold and also in a at spacetime with event horizon. The two-point
function associated with a massive self-interacting scalar eld is evaluated, up
to the rst-order level in the coupling constant , for the case of an Einstein
and also a Rindler Euclidean metric, respectively. Its value for the asymptotic
limit of the Markov parameter is exhibited. The divergences therein
are taken care of by employing a covariant stochastic regularization, where all
the symmetries of the original theory are preserved.
PACS numbers: 03.70.+k, 04.62.+v
1. Introduction
In the last century Parisi and Wu introduced the stochastic quantization method [1]. The
main idea of the stochastic quantization is that a d-dimensional quantum system is equivalent
to a (d + 1)-dimensional classical system including uctuation effects into considerations.
Some of the most important papers in the subject can be found in [2]. A brief introduction
to the stochastic quantization can be found in [3, 4] and a complete review of the stochastic
quantization is given in [5].
This program of stochastic quantization and the stochastic regularization was carried
out for generic elds dened in at, Euclidean manifolds. It is important to observe that,
since the stochastic regularization is not an action regularization, this method preserves all the
symmetries of the theory under study. In the development of this programsome authors applied
this method to the linearized Euclidean gravity and also to the non-linearized gravity [610]. It
is worth pointing out here that we cannot expect that the method will improve the perturbative
non-renormalizability of quantum gravity. Working along such lines, we may observe that
the study of a situation which lies between these two extremes is missing. Therefore, before
the attempt to implement the program in non-renormalizable models, a consistent logical step
1751-8113/07/298545+24$30.00 2007 IOP Publishing Ltd Printed in the UK 8545
8546 G Menezes and N F Svaiter
is to discuss an intermediate situation between elds in at spacetime and quantum gravity,
i.e., the semiclassical theory, which corresponds to the one-loop approximation to the totally
quantized theory. It is tempting to think that many unsolved problems, such as the loss of
information in black hole formation and the entropy of black holes, can gain new insights by
introducing the stochastic quantization.
Although there are some papers discussing the stochastic quantization in Minkowski
spacetime [11, 12], the stochastic quantization was originally introduced as an alternative
quantization method for elds dened in Euclidean space. Therefore our aimin this paper is to
discuss the stochastic quantization of scalar elds dened in a static curved manifold without
event horizon and also in a at manifold with event horizon, i.e., to analytically continue
metrics that generate real actions. We would like to stress that the stochastic quantization
is quite different from the other quantization methods; therefore, it can reveal new structural
elements of a theory which so far have gone unnoted.
The method of stochastic quantization in at spacetime with trivial topology can be
summarized by the following steps. First, starting froma eld dened in Minkowski spacetime,
after analytic continuation to imaginary time, the Euclidean counterpart, i.e., the eld dened
in an Euclidean space, is obtained. Second, a monotonically crescent Markov parameter is
introduced, which is called in the literature ctitious time and also a random noise eld
(, x), which simulates the coupling between the classical system and a heat reservoir. It
is assumed that the elds dened at the beginning in a d-dimensional Euclidean space also
depend on the Markov parameter; therefore, the eld and a random noise eld are dened in
a (d + 1)-dimensional manifold. One starts with the system out of equilibrium at an arbitrary
initial state. It is then forced into equilibrium assuming that its evolution is governed by a
Markovian Langevin equation with a white random noise eld [1315]. In fact, this evolution
is described by a process which is stationary, Gaussian and Markovian. Finally, the n-point
correlation functions of the theory in the (d +1)-dimensional space are dened by performing
averages over the random noise eld with a Gaussian distribution, that is, performing the
stochastic averages (
1
, x
1
)(
2
, x
2
) (
n
, x
n
)

. The n-point Schwinger functions of

the Euclidean d-dimensional theory are obtained, evaluating these n-point stochastic averages
(
1
, x
1
)(
2
, x
2
) (
n
, x
n
)

when the Markov parameter goes to innity ( )

and the equilibrium is reached. This can be proved in different ways for the particular case
of Euclidean scalar eld theory. One can use, for instance, the FokkerPlanck equation
[16, 17] associated with the equations describing the stochastic dynamic of the system. A
diagrammatical technique [18] has also been used to prove such equivalence.
The original method proposed by Parisi and Wu was extended to include theories with
fermions [1921]. The rst question that appears in this context is sense the Brownian problem
with anticommutating numbers makes sense. It can be shown that, for massless fermionic
elds, there will not be a convergence factor after integrating the Markovian Langevin equation.
Therefore the equilibrium is not reached. One way of avoiding this problem is to introduce a
kernel in the Langevin equation describing the evolution of two Grassmannian elds.
Usually, the ParisiWu scheme of quantization applied to bosonic and fermionic elds
converges towards non-regularized theories. In order to obtain regularized stochastic diagrams,
the original stochastic process proposed by Parisi and Wu must be modied. One way is to
replace the Markovian process by a non-Markovian one [22, 23]. This can be done by
introducing a coloured noise in the Einstein relations [2427]. Recently, the stochastic
quantization of a topological ChernSimons theory and the self-interacting
4
theory has
been investigated by the authors using the ParisiWu scheme with a non-Markovian Langevin
equation [28, 29].
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8547
Motivated by the problems that appear when one tries to implement this regularization
scheme in non-Abelian gauge theories with Zwanzigers gauge xing [3032], Bern et al
[33] derived a truly non-perturbative regularization procedure in QCD, still in the Markovian
scenario. They introduced a modied Langevin equation with a regulator multiplying the noise.
Therefore, the original Einstein relations and consequently the noise probability distribution
are maintained. This continuum regularization program was very successful in the invariant
non-perturbative regularization of all quantum eld theories [3436], including gauge theory,
gauge theory with fermions and gravity. In fewwords, the idea of any stochastic regularization
is to start from an interacting theory, then construct Langevin tree graphs, where each leg of
which ends in a regularized noise factor. Since it is possible to obtain the loops of the theory
by contracting the noise factors, one ends up with a theory where every closed loop contains
at least some power of the regulator. With this modication, one can show that the system
converges towards a regularized theory. The next step to construct a nite theory is to use, for
example, a minimal-subtraction scheme in which the ultraviolet divergent contributions are
eliminated.
The aim of this paper is to implement the stochastic quantization for the self-interacting

4
theory in a Riemannian manifold and also in a at manifold with event horizon, being
more specic, the Einstein [37] and the Rindler spacetime [3841], where positive-denite
forms can be obtained which generate real actions in the ParisiWu Langevin equation. The
only article that has some overlap with our work is the Huang paper [42]. This author has
used the stochastic quantization to nd the regularized stressenergy tensor associated with a
scalar eld in an inhomogeneous spacetime. We would like to stress that although Brownian
motion on a manifold has been studied by the mathematicians, there are a few papers studying
Brownian motion in general manifolds. See, for example, [43].
This quantization method differs from the others, the canonical and the path integral eld
quantization, based on the Hamiltonian and the Lagrangian, respectively, in many aspects. The
method starts from a classical equation of motion, but not from Hamiltonian or Lagrangian,
and consequently can be used to quantize dynamical systems without canonical formalism.
Furthermore, it is useful in situations where the others methods lead to difcult problems and
can bring us new important results.
For example, a quite important point in a regularization procedure is that it must preserve
all the symmetries of the unregularized Lagrangian. Many authors have been stressed that
a priori we cannot expect that a regularization independent proof of the renormalization of
theories in a curved background exists. The presence of the Markov parameter as an extra
dimension leads us to a regularization scheme, which preserves all the symmetries of the
theory under study. Since the stochastic regularization is not an action regularization, it may
be a way to construct such a proof. As a starting point of this program, we should calculate
the two-point function up to the rst-order level in the coupling constant and apply the
continuum stochastic regularization. Our results are to be compared with the usual ones in the
literature.
The organization of the paper is the following. In section 2, we discuss the stochastic
quantization for the (
4
)
d
scalar theory in a d-dimensional Euclidean manifold. In section 3,
we use the stochastic quantization and the stochastic regularization to obtain the two-point
Schwinger function in the one-loop approximation in the Einstein manifold using the stochastic
quantization. In section 4, we repeat the method for the case of the Rindler manifold, a non-
simple connected manifold in the Euclidean version. Conclusions are given in section 5. In
the appendix we sketch the general formalism to construct the maximal analytic extension of
the vacuum Schwarzschild solution. In this paper we use h = c = k
B
= G = 1.
8548 G Menezes and N F Svaiter
2. Stochastic quantization for the (
4
)
d
scalar theory: the Euclidean case
In this section, we give a brief survey for the case of self-interacting scalar elds, implementing
the stochastic quantization and the continuum stochastic regularization theory up to the one-
loop level. Let us consider a neutral scalar eld with a (
4
) self-interaction. The Euclidean
action that usually describes a free scalar eld is
S
0
[] =
_
d
d
x
_
1
2
()
2
+
1
2
m
2
0

2
(x)
_
(1)
and the interacting part dened by the non-Gaussian contribution is
S
I
[] =
_
d
d
x

4!

4
(x). (2)
The simplest starting point of the stochastic quantization to obtain the Euclidean eld
theory is a Markovian Langevin equation. Assume a at Euclidean d-dimensional manifold,
where we choose periodic boundary conditions for a scalar eld and also a random noise. In
other words, they are dened in a d-torus T
d
. To implement the stochastic quantization
we supplement the scalar eld (x) and the random noise (x) with an extra coordinate ,
the Markov parameter, such that (x) (, x) and (x) (, x). Therefore, the elds
and the random noise are dened in a domain, T
d
R
(+)
. Let us consider that this dynamical
system is out of equilibrium, being described by the following equation of evolution:

(, x) =
S
0
(x)

(x)=(,x)
+ (, x), (3)
where is a Markov parameter, (, x) is a random noise eld and S
0
is the usual free
Euclidean action dened in equation (1). For a free scalar eld, the Langevin equation reads

(, x) =
_
+ m
2
0
_
(, x) + (, x), (4)
where is the d-dimensional Laplace operator. Equation (4) describes a OrnsteinUhlenbeck
process and we assume the Einstein relations, that is,
(, x)

= 0 (5)
and for the two-point correlation function associated with the random noise eld
(, x)(

, x

= 2(

)
d
(x x

), (6)
where

means stochastic averages. The above equation denes a delta-correlated random

process. In a generic way, the stochastic average for any functional of given by F[] is
dened by
F[]

=
_
[d]F[] exp
_

1
4
_
d
d
x
_
d
2
(, x)
_
_
[d] exp
_

1
4
_
d
d
x
_
d
2
(, x)
_ . (7)
Let us dene the retarded Green function for the diffusion problemthat we call G(

, xx

).
The retarded Green function satises G(

, x x

) = 0 if

< 0 and also

_

+
_

x
+ m
2
0
_
_
G(

, x x

) =
d
(x x

)(

). (8)
Using the retarded Green function and the initial condition (, x)|
=0
= 0, the solution for
equation (4) reads
(, x) =
_

0
d

d
d
x

G(

, x x

)(

, x

). (9)
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8549
Let us dene the Fourier transforms for the eld and the noise given by (, k) and (, k).
We have respectively
(, k) =
1
(2)
d
2
_
d
d
x e
ikx
(, x), (10)
and
(, k) =
1
(2)
d
2
_
d
d
x e
ikx
(, x). (11)
Substituting equation (10) into equation (1), the free action for the scalar eld in the (d + 1)-
dimensional space writing in terms of the Fourier coefcients reads
S
0
[(k)]|
(k)=(,k)
=
1
2
_
d
d
k (, k)
_
k
2
+ m
2
0
_
(, k). (12)
Substituting equations (10) and (11) into equation (4) we have that each Fourier coefcient
satises a Langevin equation given by

(, k) =
_
k
2
+ m
2
0
_
(, k) + (, k). (13)
In the Langevin equation the particle is subject to a uctuating force (representing a stochastic
environment), where its average properties are presumed to be known and also the friction
force. Note that the friction coefcient in equation (7) is given by
_
k
2
+ m
2
0
_
.
The solution for equation (7) reads
(, k) = exp
_

_
k
2
+ m
2
0
_

_
(0, k) +
_

0
d

exp
_

_
k
2
+ m
2
0
_
(

)
_
(

, k). (14)
Using the Einstein relation, we get that the Fourier coefcients for the random noise satisfy
(, k)

= 0 (15)
and
(, k)(

, k

= 2(

)
d
(k + k

). (16)
It is possible to show that (, k)(

, k

|
=
D(k, k

; ,

) is given by
D(k; , ) = (2)
d

d
(k + k

)
1
_
k
2
+ m
2
0
_
_
1 exp
_
2
_
k
2
+ m
2
0
__ _
. (17)
where we assume =

.
Nowlet us analyse the stochastic quantization for the (
4
)
d
self-interaction scalar theory.
In this case the Langevin equation reads

(, x) =
_
+ m
2
0
_
(, x)

3!

3
(, x) + (, x). (18)
The two-point correlation function associated with the random eld is given by the Einstein
relations, while the other connected correlation functions vanish, i.e.,
(
1
, x
1
)(
2
, x
2
) (
2k1
, x
2k1
)

= 0 (19)
and also
(
1
, x
1
) (
2k
, x
2k
)

(
1
, x
1
)(
2
, x
2
)

(
k
, x
k
)(
l
, x
l
)

, (20)
where the sum is to be taken over all the different ways in which the 2k labels can be divided
into k parts, i.e., into k pairs. Performing Gaussian averages over the white random noise, it is
possible to prove the important formulae
lim

(
1
, x
1
)(
2
, x
2
) (
n
, x
n
)

=
_
[d](x
1
)(x
2
) (x
n
) e
S()
_
[d] e
S()
, (21)
8550 G Menezes and N F Svaiter
where S() = S
0
() + S
I
() is the d-dimensional action. This result leads us to consider the
Euclidean path integral measure a stationary distribution of a stochastic process. Note that the
solution of the Langevin equation needs a given initial condition. As for example
(, x)|
=0
=
0
(x). (22)
Let us use the Langevin equation to perturbatively solve the interacting eld theory.
One way to handle equation (12) is with the method of Greens functions. We dened the
retarded Green function for the diffusion problem in equation (8). Let us assume that the
coupling constant is a small quantity. Therefore to solve the Langevin equation in the case of
a interacting theory we use a perturbative series in . Therefore we can write
(, x) =
(0)
(, x) +
(1)
(, x) +
2

(2)
(, x) + . (23)
Substituting equation (23) into equation (12), and if we equate terms of equal power in , the
resulting equations are
_

+
_

x
+ m
2
0
_
_

(0)
(, x) = (, x), (24)
_

+
_

x
+ m
2
0
_
_

(1)
(, x) =
1
3!
(
(0)
(, x))
3
, (25)
and so on. Using the retarded Green function and assuming that
(q)
(, x)|
=0
= 0, q, the
solution to the rst equation given by equation (24) can be written formally as

(0)
(, x) =
_

0
d

d
d
x

G(

, x x

)(

, x

). (26)
The second equation given by equation (25) can also be solved using the above result. We
obtain

(1)
(, x) =
1
3!
_

0
d
1
_

d
d
x
1
G(
1
, x x
1
)

__

1
0
d

d
d
x

G(
1

, x
1
x

)(

, x

)
_
3
. (27)
We have seen that we can generate all the tree diagrams with the noise-eld contributions.
We can also consider the n-point correlation function (
1
, x
1
)(
2
, x
2
) (
n
, x
n
)

.
Substituting the above results into the n-point correlation function, and taking the random
averages over the white noise eld using the Wick-decomposition property dened by
equation (20) we generate the stochastic diagrams. Each of these stochastic diagrams has
the form of a Feynman diagram, apart from the fact that we have to take into account that we
are joining together two white random noise elds many times. Besides, the rules to obtain
the algebraic values of the stochastic diagrams are similar to the usual Feynman rules.
As simple examples let us show how to derive the two-point function in the zeroth order
(
1
, x
1
)(
2
, x
2
)
(0)

, and also the rst-order correction to the scalar two-point function given
by (
1
, x
1
)(
2
, x
2
)
(1)

. Using equation (9) and the Einstein relations we have

(
1
, x
1
)(
2
, x
2
)
(0)

= 2
_
min(
1
,
2
)
0
d

d
d
x

G(
1

, x
1
x

)G(
2

, x
2
x

).
(28)
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8551
For the rst-order correction we get
(X
1
)(X
2
)
(1)

=

3!
_
_
dX
3
_
dX
4
(G(X
1
X
4
)G(X
2
X
3
) + G(X
1
X
3
)G(X
2
X
4
))
(X
3
)
_
_
dX
5
G(X
4
X
5
)(X
5
)
_
3
_

. (29)
where, for simplicity, we have introduced a compact notation
_

0
d
_

d
d
x
_
dX, (30)
and also (, x) (X) and nally (, x) (X).
The process can be repeated and therefore the stochastic quantization can be used as an
alternative approach to describe scalar quantum elds. Therefore, the two-point function up
to the rst-order level in the coupling constant is given by
(
1
, x
1
)(
2
, x
2
)
(1)

= (a) + (b) + (c), (31)

where (a) is the zero-order two-point function and (b) and (c) are given, respectively, by
(b) =

d
(k
1
+ k
2
)
_
d
d
k
_

1
0
d G(k
1
;
1
)D(k; , )D(k
2
;
2
, ), (32)
(c) =

d
(k
1
+ k
2
)
_
d
d
k
_

2
0
d G(k
2
;
2
)D(k; , )D(k
1
;
1
, ). (33)
These are the contributions in rst order. A simple computation shows that we recover the
correct equilibrium result at equal asymptotic Markov parameters (
1
=
2
)
(b)|

1
=
2

d
(k
1
+ k
2
)
1
_
k
2
2
+ m
2
0
_
1
_
k
2
1
+ k
2
2
+ 2m
2
0
_
_
d
d
k
1
_
k
2
+ m
2
0
_. (34)
Obtaining the Schwinger functions in the asymptotic limit does not guarantee that we
gain a nite physical theory. The next step is to implement a suitable regularization scheme.
A crucial point to nd a satisfactory regularization scheme is to use one that preserves the
symmetries of the original model. The presence of the Markov parameter as an extra dimension
lead us to a new regularization scheme, the stochastic regularization method, which preserves
all the symmetries of the theory under study. Therefore, let us implement a continuum
regularization procedure [36]. We begin with a regularized Markovian ParisiWu Langevin
system

(, x) =
S
0
(x)

(x)=(,x)
+
_
d
d
y R
xy
()(, y). (35)
The Einstein relations, given by equations (5) and (6), are maintained. The regulator R()
that multiplies the noise is a function of the Laplacian

xy
=
_
d
d
z (

)
xz
(

)
zy
, (36)
where
(

)
xy
=
x

d
(x y). (37)
8552 G Menezes and N F Svaiter
We will be working with a heat kernel regulator with the form
R(; ) = exp
_

2
_
, (38)
where is a parameter introduced to regularize the theory. The basic restrictions on the form
of this heat kernel regulator are
R(; )|

= 1 (39)
or
R
xy
(; )|

=
d
(x y), (40)
which guarantees that the regularized process given by equation (35) reduces to the formal
process given by equation (3) in the formal regulator limit .
With this modication in the Langevin equation, it is possible to show that all the
contributions to the n-point function at all orders in the coupling constant are nite. For
instance, the contribution to the two-point function at the one-loop level given by equation (32)
is rewritten as
(b)|

1
=
2

d
(k
1
+ k
2
)
R
2
k
2
_
k
2
2
+ m
2
0
_
1
_
k
2
1
+ k
2
2
+ 2m
2
0
_
_
d
d
k
R
2
k
_
k
2
+ m
2
0
_, (41)
where R
k
is the Fourier transform of the regulator, i.e.,
R
k
() = R(; )|
=k
2 . (42)
Nowwe use this method to discuss the quantization of scalar theories with self-interaction
in a curved spacetime without event horizon and in a at manifold with event horizon. Being
more specic, we are interested in investigating the
4
theory in the Einstein and Rindler
spacetimes.
3. Stochastic quantization for the (
4
)
d
scalar theory: the Einstein case
The aim of this section is to implement the stochastic quantization and the stochastic
regularization for the self-interacting
4
theory in the one-loop level in the Einstein spacetime.
Let us consider a M
4
manifold that admits a non-vanishing timelike Killing vector eld X.
If one can always introduce coordinates t = x
0
, x
1
, x
2
, x
3
locally such that X =

t
and
the components of the metric tensor are independent of t, M
4
is stationary. If further the
distribution X

of 3-planes orthogonal to X is integrable, then M

4
is static. Each integral curve
of the Killing vector eld X =

t
is a worldline of a possible observer. Since X =

t
generates
isometries, the 3-planes X

are invariant under these isometries. For the static manifold, it is

possible to perform a Wick rotation, i.e., analytically extend the pseudo-Riemannian manifold
to the Riemannian domain without problem. Therefore for static spacetime the implementation
of the stochastic quantization is straightforward.
In the previous section, we have been working in an Euclidean space R
d
R
(+)
, where
R
d
is the usual Euclidean space and R
(+)
is the Markov sector. Now let us generalize this
to a more complicated case, i.e., let us work in a general (and, for the time being, static)
Riemannian manifold M. In other words, we will consider a classical eld theory dened in
a M R
(+)
manifold coupled with a heat reservoir. Therefore, in a four-dimensional curved
manifold, the ParisiWu Langevin equation for scalar elds reads

(, x) =
1

g
S
0
(x)

(x)=(,x)
+ (, x), (43)
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8553
where g = det g

and the classical Euclidean action S

0
is given by
S
0
=
_
d
4
x

gL. (44)
In the above equation L is given by
L =
1
2
g

+
1
2
(m
2
+ R)
2
. (45)
Note that we introduce a coupling between the scalar eld and the gravitational eld represented
by the term R
2
, where is a numerical factor and R is the Ricci scalar curvature. The
random noise eld (, x) obeys the following generalized Einstein relations:
(, x)

= 0 (46)
and
(, x)(

, x

=
2

g(x)

4
(x x

)(

). (47)
Substituting equations (44) and (45) into the Langevin equation given by equation (43), we
get

(, x) = ( + m
2
+ R)(, x) + (, x), (48)
where is the four-dimensional LaplaceBeltrami operator dened by
= g
1/2

(g
1/2
g

)
= g

, (49)
with denoting the covariant derivative.
To proceed, as in the at situation, let us introduce the retarded Green function for the
diffusion problem G(

, x, x

), which obeys
_

+ (
x
+ m
2
+ R(x))
_
G(

, x, x

) =
1

4
(x x

)(

) (50)
if

> 0 and G(

, x, x

) = 0 if

< 0.
Using the retarded Green function and the initial condition (, x)|
=0
= 0, a formal
solution to equation (48) reads
(, x) =
_

0
d

d
4
x

_
g(x

)G(

, x x

)(

, x

). (51)
We see that, in order to solve equation (48) we need to invert the differential operator given by
equation (49). Depending on the metric g

dened on the manifold, this could be a difcult

task. In the Euclidean case this is easily solved, since we decompose the elds in Fourier
modes (plane waves), which are solutions to the KleinGordon equation in a at metric; each
Fourier mode, as we have seen, obeys a Langevin equation. So, if the KleinGordon equation
in a general curved manifold allows us to obtain solutions which can be decomposed in modes
separated in a similar manner to the at case, we may as well decompose our elds in these
modes in a way that each of them, again, will obey a Langevin equation. Then, equation (48)
can be solved in a simple and direct way. Of course, this procedure will help us if we
admit manifolds with global Killing vectors, which, unfortunately, is not always the case. So
admitting that this is the case, we may write the mode decompositions as
(, x) =
_
d (k)
k
()u
k
(x) (52)
8554 G Menezes and N F Svaiter
and
(, x) =
_
d (k)
k
()u
k
(x), (53)
where the measure (k) depends on the metric we are interested in. For instance, in the at
case, we have that in a four-dimensional space d (k) = d
4
k and the modes u
k
(x) are given by
u
k
(x) =
1
(2)
2
e
ikx
. (54)
So we see that, in the at case, equations (52) and (53) reduce to the Fourier decomposition
given by equations (10) and (11), respectively.
Now let us apply this formalism to the Euclidean Einstein manifold. The general
RobertsonWalker line element is given by [44, 45]
ds
2
= dt
2
+ a
2
(t )
3

i,j=1
h
ij
dx
i
dx
j
, (55)
where
3

i,j=1
h
ij
dx
i
dx
j
= (1 Kr
2
)
1
dr
2
+ r
2
(d
2
+ sin
2
d
2
)
= d
2
+ f
2
()(d
2
+ sin
2
d
2
), (56)
and, for the Einstein manifold we have (K = 1), f () = r = sin , 0 2.
Equation (56) gives the line element on the spatial sections in the pseudo-Riemannian
case which are hyperbolic, at or closed depending on whether K = 1, 0, 1, respectively.
Writing C() = a
2
(t ) with the conformal time parameter given by
=
_
t
dt

a
1
(t

), (57)
the line element can be recast in the form
ds
2
= C()
_
d
2
+
3

i,j=1
h
ij
dx
i
dx
j
_
. (58)
In the simplest case, namely static spacetimes with C() = C = a
2
= constant, the scalar
curvature is given by R =
6K
C
. After the Wick rotation, we are working with the Euclidean
Einstein manifold. In this case, we may decompose the modes as
u
k
(x) = C
1/2
X

k
( x) e
ik

, (59)
with x = (r, , ) or (, , ) and X

k
being a solution of

(3)
X

k
= (q
2
K)X

k
. (60)
In equation (60),
(3)
is the Laplacian associated with the spatial metric h
ij
= h
1/2

i
(h
1/2
h
ij

j
). (61)
Further, the functions X

k
are normalized such that
_
d
3
x h
1
2
X

k
( x)X

( x) = (

k,

k

), (62)
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8555
where (

k,

k

) is the -function with respect to the measure

_
d (k

)f (

k)(

k,

k

) = f (

). (63)
The eigenfunctions X

k
of the three-dimensional Laplacian are, for K = 1
X

k
( x) =
+
qJ
()Y
MJ
(, ), (64)
with

k = (q, J, M), M = J, J + 1, . . . , J, J = 0, 1, . . . , q 1 and q = 1, 2, . . . . The
Y
MJ
are the usual spherical harmonics. The functions
+
qJ
can be dened from [46]

qJ
() =
_
1
2
q
2
(q
2
+ 1) (q
2
+ J
2
)
_
1/2
sinh
J

_
d
d cosh
_
J+1
cos q, (65)
by replacing q by iq and by i in the latter.
With these denitions, the measure (k) for the Einstein universe is dened as follows:
_
d (k) =
1
2
_
dk

q,J,M
. (66)
So, inserting the mode decomposition given by equation (59) into equation (48) we have that
each mode coefcient satises the Langevin equation given by

k
() =
1
C
(k
2
+
2
)
k
() +
k
(), (67)
where k
2
= q
2
+ k
2

and
2
= Cm
2
+ (6 1)K and K = 1. For simplicity, we redene
1
C
(k
2
+
2
) (k
2
+
2
).
The solution for equation (67), with the initial condition
k
()|
=0
= 0, reads

k
() =
_

0
d

G
k
(,

)(

, k), (68)
where
G
k
(,

) = exp((k
2
+
2
)(

))(

) (69)
is the retarded Green function for the diffusion problem.
Using the Einstein relations, we get that the mode coefcients for the randomnoise satisfy

k
()

= 0 (70)
and

k
()
k
(

= 2(

)
4
(k, k

), (71)
where
4
(k, k

) = (k

+ k

)
qq

JJ

MM
.
The two-point function D
k
(,

) can be calculated in a similar way as in the Euclidean

at case. We have
D
k
(,

) =
1
(2)

4
(k, k

)
1
(k
2
+
2
)
_
e
((k
2
+
2
)|

|)
e
((k
2
+
2
)(+

))
_
, (72)
or, in the coordinate space
D(,

; x, x

) =
_
d (k)u
k
(x)u

k
(x

)D
k
(,

)
=
_
d (k)u
k
(x)u

k
(x

)
1
(k
2
+
2
)
_
e
((k
2
+
2
)|

|)
e
((k
2
+
2
)(+

))
_
, (73)
where
_
d (k) is given by equation (66).
8556 G Menezes and N F Svaiter
Now, let us apply this method for the case of a self-interacting theory with an interaction
action given by
S
I
[] =
_
d
4
x
_
g(x)

4!

4
(x). (74)
In the same way, we can solve the equation using a perturbative series in . The two-point
function up to the one-loop level is given by
(
1
, x
1
)(
2
, x
2
)
(1)

= (a) + (b) + (c), (75)

where (a) is the zero-order two-point function given by equation (72) and (b) and (c) are
given respectively by
(b) =

4
(k
1
, k
2
)
_
d

(k)
_

1
0
d G
k
1
(
1
)D
k
(, )D
k
2
(
2
, ), (76)
(c) =

4
(k
1
, k
2
)
_
d

(k)
_

2
0
d G
k
2
(
2
)D
k
(, )D
k
1
(
1
, ), (77)
where
_
d

(q)
1
2
_
dk

q,J,M
X
q
X
q
. These are the contributions in rst order. A simple
computation shows that we obtain, at equal asymptotic Markov parameters (
1
=
2
),
(b)|

1
=
2

4
(k
1
, k
2
)
1
_
k
2
2
+
2
_
1
_
k
2
1
+ k
2
2
+ 2
2
_
_
d

(k)
1
(k
2
+
2
)
. (78)
Now we dene the quantity I as
I =
_
d

(q)
1
(k
2
+
2
)
=
1
2
_
dk

q,J,M
X
q
X
q
1
(k
2
+
2
)
=
_
dk

q
X
q
X
q
q
2
(q
2
+ b
2
)
, (79)
where b
2
= k
2

+
2
. It is easy to show that the series in this equation is divergent. So, we
need a procedure to regularize it and obtain a nite quantity for the two-point function. As we
will see now, this can be done within the covariant stochastic regularization [36].
The generalization of equation (35) to general four-dimensional spacetimes is

(, x) =
1

g
S
0
(x)

(x)=(,x)
+
_
d
4
y

g R
xy
()(, y), (80)
where, for covariant reasons, the regulator is now a function of the covariant Laplacian

xy
=
_
d
4
z(

)
xz
(

)
zy
(81)
with
(

)
xy
=
x

4
(x y). (82)
Using the mode decomposition given by equation (59), equation (80) reduces to

k
() =
1
C
(k
2
+
2
)
k
() +
k
()R
k
, (83)
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8557
where R
k
= R
xy
()|
=k
2
+K
and
R
xy
() =
_
d (k)u
k
(x)u

k
(y)R
k
. (84)
Then, the solution for the regularized Langevin equation follows

k
() =
_

0
d

G
k
(,

)(

, k), (85)
where G
k
(,

) is given by equation (69). Then, it is easy to show that the zero-order two-
point function and the regularized contribution (b)

at the rst order at the coupling constant

are given by, respectively,
D
k
(,

) =
1
(2)

4
(k, k

)
R
2
k
(k
2
+
2
)
_
e
((k
2
+
2
)|

|)
e
((k
2
+
2
)(+

))
_
(86)
and
(b)

1
=
2

4
(k
1
, k
2
)
R
2
k
2
_
k
2
2
+
2
_
1
_
k
2
1
+ k
2
2
+ 2
2
_
_
d

(k)
R
2
k
(k
2
+
2
)
. (87)
Let us isolate the part where we had problems. We have
I

=
_
d

(k)
R
2
k
(k
2
+
2
)
=
1
2
_
dk

q,J,M
X
q
X
q
R
2
k
(k
2
+
2
)
. (88)
After some tedious calculations [4749], we arrive at
I

=
1
4
e
2K
C
2
+
2
C
2

2
_

dx(x
2

2
)
1/2
erfc(

x), (89)
where erfc(x) is the complementary error function. At the expense of considerable labour, it
is possible to extend our results to the four-point function.
A quite important point is that this regularization procedure preserves all the symmetries
of the unregularized Lagrangian, since it is not an action regularization. The next step would
be to isolate the parts that go to innity in the limit and remove them with a
suitable redenition of the constants of the theory, i.e., carry out the renormalization program.
A natural question now would be if we can actually renormalize all the n-point functions
at all orders at the coupling constant . Birrel [50] has given arguments that a priori we
cannot expect that a regularization independent proof of the renormalizability of the
4
theory in a curved background exists. One attempt of general proof of renormalizability of

4
theory dened in a spacetime which can be analytically continued to Euclidean situation
was given by Bunch [51]. Using the EpsteinGlaser method, Brunetti and Fredenhagen [52]
presented a perturbative construction of this theory on a smooth globally hyperbolic curved
spacetime.
Our derivation shows that the stochastic regularization may be an attempt in a direction
of such a regularization-independent proof, even though we are still restricted to the same
situation studied by Bunch. Indeed, we knowthat, in this case of the Einstein universe, the free
two-point function presents itself as a sum of Minkowski space positive frequency two-point
functions, so it is natural to expect that, as it is in the Minkowski case, a regularization-
independent proof of the renormalizability of such an interacting theory does exist, at least for
the Einstein universe.
8558 G Menezes and N F Svaiter
4. Stochastic quantization for the (
4
)
d
scalar theory: the Rindler case
Now let us consider the stochastic quantization in a spacetime with an event horizon. The
Schwarzschild black hole is the most familiar solution to the vacuum Einstein equations,
which has an event horizon. At the origin, there is a curvature singularity and at R = 2M
we have a singularity due to the bad behaviour of this particular coordinate system. Although
it is not possible to extend the spacetime across the singularity, it is possible to analytically
continue the manifold from r > 2M to the region r < 2M, (r = 0). The maximal extension
of the manifold described by the Schwarzschild line element with 2M < r < is the usual
Kruskal extension [53]. The Kruskal spacetime denes two outer asymptotically at regions
and also two regions inside the event horizon, bounded by the past and future singularities
(see appendix A). Since we know that, close to the horizons, the Schwarzschild coordinates
t and r behave as Rindlers spacetime coordinates, we shall investigate the stochastic
quantization in the Rindler metric. The background material relevant for us can be found
in the text book [54] and also in the review articles [5558].
Let us consider a d-dimensional at Minkowski spacetime where we are using the usual
Cartesian coordinates y

= (y
0
, y
1
, . . . , y
d1
). It is possible to nd a curvilinear coordinate
system x

= (t, x
1
, . . . , x
d2
, z) called the Rindler coordinate system and it can be shown
that this coordinate system is one naturally adapted to an observer with constant proper
acceleration. It is important to stress that this coordinate system with the respective coordinate
transformation covers only the region y
d1
> |y
0
|. Since this coordinate system does not
cover the whole Minkowski spacetime, we can dene three coordinate transformations with
the respective coordinate systems dened in different regions of the Minkowski spacetime.
These regions are known in the literature as Rindlers L, Milne F and nally Milne P.
The four coordinate transformations and the coordinate systems together cover all the
Minkowski spacetime. The coordinate system that covers the region inside the light cone
is the d-dimensional Milne spacetime.
The implementation of the canonical quantization in Rindlers spacetime is very simple,
since if the spacetime has a stationary geometry, there is a time-like killing vector eld K
that generates one-parametric Lie group of isometries and the orthogonal modes which satisfy
L
K
u
n
(x) = iu
n
, where L
K
is the Lie derivative with respect to K and u
n
(x) are the
positive frequency modes. In this situation, there is a natural way to dene positive and
negative frequency modes. Note that the Rindler line element is t independent, and
consequently there is a straightforward way to dene positive and negative frequency modes
in order to impose the canonical quantization in Rindlers spacetime.
It has long been recognized that the Rindlers vacuum |0, R that appears in the Fulling
canonical quantization of a scalar eld in the Rindlers spacetime is not unitarily equivalent to
the Minkowski vacuum |0, M. Let us dene the following generating functionals, Z(h) and
Z
R
(h), i.e.,
Z(h) =
0, M, out|0, M, in
h
0, M, out|0, M, in
h=0
(90)
and
Z
R
(h) =
0, R, out|0, R, in
h
0, R, out|0, R, in
h=0
(91)
where |0, M, in, |0, M, out, |0, R, in and nally |0, R, out are the IN and OUT vacuum
states for the theory with the Minkowski and Rindler Hamiltonian density, respectively.
Starting from Z(h) it is possible after analytic continuation and imposing periodicity in
the Euclidean time, to dene Z(; h), i.e., the nite-temperature Schwinger functional. It
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8559
is clear that lim

Z(; h) = Z(; h) Z(h), where Z(; h) is the zero-temperature

Euclidean functional which generates the Schwinger functions in the whole Euclidean space.
At the same way it is possible to dene Z
R
(; h), i.e. the nite-temperature Euclidean Rindler
functional. Again we have lim

Z
R
(; h) = Z
R
(; h) Z(h), where Z
R
(; h) is the
zero-temperature Euclidean functional which generates the Schwinger functions in the analytic
extended Rindler manifold.
From the above denitions, let us dene the following Schwinger functions. We are
following the discussion of Christensen and Duff [59]:
(i) G

(x, x

) two-point function, i.e., the Schwinger function obtained from the nite-
temperature Schwinger functional.
(ii) G

(x, x

) two-point function, i.e., the Schwinger function obtained from the zero-
temperature Schwinger functional.
Now, let us analyse the two-point function that we are able to construct in the Euclidean
manifold. Remember that the Rindler Euclidean metric possesses a coordinate singularity at
the origin where z = 0. The point z = 0 is a conical point of the Euclidean manifold but is a
regular point if the Euclidean time is periodic.
(i) G
(1)
(x, x

) two-point function which is periodic in the Euclidean Rindlers time. This is

the usual scalar two-point Schwinger function. It is a function of the geodesic distance
between two points in the manifold. Since all paths around the origin are topologically
equivalent, therefore G
(1)
(x, x

) should be periodic in Euclidean Rindlers time. Therefore

these Greens functions can be generated by functional derivatives in the Z(; h), i.e., the
nite-temperature Schwinger functional.
(ii) G
(0)
(x, x

) two-point function. This two-point function appears in the situation where

the Euclidean space has a hole in the origin. In the case where the two points are on
the circular world line it is possible to relate the two two-point functions G
(1)
(x, x

) and
G
(0)
(x, x

). This can be done using a simple trigonometric identity

G
(1)
(,

) =

n=
G
0
_
,

+
2n
a
_
. (92)
Since the nite-temperature Schwinger function must be periodic in the Euclidean time
with periodicity , making the identication =
2
a
it is clear that G
(1)
(x, x

) =
G

(x, x

). Also we have G
(0)
(x, x

) = G

(x, x

). Therefore these Greens functions

can be generated by functional derivatives in the Z(; h) which is the zero-temperature
Euclidean functional which generates the Schwinger functions in the whole Euclidean
space.
Now, after a Wick rotation, we should apply the stochastic quantization for the four-
dimensional Rindler space, which becomes a multiple connected manifold, with the Euclidean
metric
ds
2
= z
2
dt
2
+ dx
2
1
+ dx
2
2
+ dz
2
. (93)
In the Euclidean Rindler space, the Langevin equation for the eld (x) reads

(, x) =

g
S
0
(x)

(x)=(,x)
+ (, x), (94)
where S
0
is the Euclidean action for the scalar-free elds. Note that this expression is quite
different from the Langevin equation for the Einstein metric, equation (43), since we have an
8560 G Menezes and N F Svaiter
inverse power of the determinant of the metric multiplied by
S
0
(x)
. So, with the Rindler metric
given above, we have

(, x) = ( + z
2
m
2
)(, x) + (, x), (95)
where the operator is dened by
=
2
t
+ z
2
_

2
x
1
+
2
x
2
+
2
z
_
+ z
z
. (96)
To proceed with the implementation of the stochastic quantization, we have to use the
generalized Einstein relations, given by equations (46) and (47).
From the modes presented by [60, 61], we can obtain the following Euclidean modes:
u
k
( x, z) =
1
2
2
( sinh())
1/2
e
i

k x
K
i
(z), (97)
where

k x = k
0
t + k
1
x
1
+ k
2
x
2
, =
_

k
2
+ m
2
and K

(x) is the Macdonald function. This

are normalized such as
1

2
_
dz
z
K
i
(z)K
i
(z) =
(,

)
( +

)(sinh() sinh(

))
1/2
. (98)
Now, we use the general mode decomposition, given by equations (52) and (53), with the
following measure:
_
d (k) =
_

0
d
_
d

k =
_

0
d
_
dk
1
_
dk
2
_
dk
0
2
. (99)
Remembering equation (97) for the modes, we will have again that each mode coefcient
obeys a Langevin equation of the form

k
() =
_
k
2
0
+
2
_

k
() +
k
(). (100)
The solution for equation (100), with the initial condition
k
()|
=0
= 0, reads

k
() =
_

0
d

G
k
(,

)(

, k), (101)
where
G
k
(,

) = exp
_

_
k
2
0
+
2
_
(

)
_
(

) (102)
is the retarded Green function for the diffusion problem. Using the Einstein relations, we get
that the mode coefcients for the random noise satisfy

k
()

= 0 (103)
and

k
()
k
(

= 2(

)(k, k

)(,

), (104)
where (k, k

) = (k
0
+ k

0
)(

k +

k

).
The two-point function D
0
k
(,

) can be calculated in a similar way as in the Einstein

case. We have
D
0
k
(,

) = (k, k

)(,

)
1
_
k
2
0
+
2
_
_
e
((k
2
0
+
2
)|

|)
e
((k
2
0
+
2
)(+

))
_
, (105)
or, in the coordinate space
D
0
(x, x

; ,

) =
1
4
4
_

0
d
_
dk sinh()K
i
(z)K
i
(z

e
i

k( x

)
k
2
0
+
2
_
e
((k
2
0
+
2
)|

|)
e
((k
2
0
+
2
)(+

))
_
. (106)
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8561
In the limit =

, and proceeding with similar steps as in [60], one may prove that we
have obtained the usual result for the free two-point function, i.e., the Schwinger two-point
function for a (straight-line) geodesic distance between the points x and x

. As is well known
[59], this function is determined uniquely by the requirement that, in the absence of any holes
in space, one identies the angle t (which is the Euclidean time) with the angle t + 2n
(n = integer) because all paths around the origin are topologically equivalent. Thus, as
expected, D
0
should be periodic in t with period 2. But we know that, when we do
the analytical continuation t it in the Rindler metric, the event horizon, which in
Minkwoski space is represented as an impenetrable barrier to the accelerating observer,
translates in Euclidean language into the statement that closed paths around the origin cannot
be continuously shrunk to a point. An inertial observer sees a space with the usual topology,
R
4
, but the topology seen by the accelerating observer is different, i.e., his space has a hole in
it. Then, paths winding around the origin cannot all be deformed into each other, they fall into
topologically distinct classes labelled by the winding number n. Now, we do not identify t
with t +2n. This two-point function, which we shall call D
1
, exhibits an innite periodicity.
To obtain it, we should work with the following modes [62]:
u
k
( x, z) =
1
2
2
( sinh())
1/2
e
i(k
1
x
1
+k
2
x
2
)k
0
|t |
K
i
(z). (107)
With this prescription, we obtain the same results as found in the literature [63]
D
0
(x, x

; ,

) =
n=+

n=
D
1
(x, x

; ,

)
=
1
4
4
n=+

n=
_

0
d
_
dk sinh()K
i
(z)K
i
(z

e
i(k
1
x
1
+k
2
x
2
)k
0
|t +2n|
k
2
0
+
2
_
e
((k
2
0
+
2
)|

|)
e
((k
2
0
+
2
)(+

))
_
, (108)
where the geodesic distance between the points x and x

is an arc length in the function D

1
.
It is interesting to note that, using the modes given by equation (107) (see [62]), and then
summing up all congurations with a winding number n, we reach the same result as using
the modes given by equation (97) (see [60]).
Let us repeat the method for the case of self-interacting theory with the interaction action
given by
S
I
[] =
_
d
4
x
_
g(x)

4!

4
(x). (109)
In this case, we have similar equations to the Einstein case. Equation (32) reads, in the Rindler
space
(b) =

2
(k
1
, k
2
)(
1
,
2
)
_
d

(k)
_

1
0
d G
k
1

1
(
1
)D
0
k
(, )D
0
k
2

2
(
2
, ), (110)
which, in the limit =

, becomes
(b)|

1
=
2

2
(k
1
, k
2
)(
1
,
2
)
1
_
k
2
2
+
2
2
_
1
_
k
2
1
+ k
2
2
+
2
1
+
2
2
_
_
d

(k)
1
(k
2
+
2
)
(111)
with similar divergences and
_
d

_
d
_
dk sinh()K
i
(z)K
i
(z). Now, we may
apply the continuum regularization as before [36], with a regulator that is a function of the
8562 G Menezes and N F Svaiter
following operator:

= z
2
_

2
x
1
+
2
x
2
+
2
z
_
+ z
z
z
2
m
2
. (112)
The regularized Langevin equation reads

k
() =
_
k
2
0
+
2
_

k
() + R
k

k
(), (113)
with the solution

k
() =
_

0
d

G
k
(,

)R
k
(

, k), (114)
where R
k
() = R(

; )|

=
2 . Then, it is easy to show that the zero-order two-point
function and the regularized contribution (b)

at the one-loop level are given by, respectively,

D

k
(,

) = (k, k

)(,

)
R
2
k
_
k
2
0
+
2
_
_
e
((k
2
0
+
2
)|

|)
e
((k
2
0
+
2
)(+

))
_
(115)
and
(b)|

1
=
2

2
(k
1
, k
2
)(
1
,
2
)
R
2
k
2

2
_
k
2
2
+
2
2
_
1
_
k
2
1
+ k
2
2
+
2
1
+
2
2
_
_
d

(k)
R
2
k
(k
2
+
2
)
. (116)
So, the regularized part of the contribution above reads
I =
1
4
4
_

0
d
_

0
d

k
_
dk
0
2
sinh()K
i
(z)K
i
(z)
e
2
2

2
k
2
0
+
2
. (117)
With some tedious manipulations [4749], we arrive at
I =
_

2
_
1/2
m
16
e

2
8
f (, z), (118)
where the function f (, z) is given by
f =
_
+

K
1
(m ) e

2
8
_
e
i
2
4
erfc
_

2
(i + )
_
e

i
2
4
erfc
_

2
(i )
__
, (119)
where
2
= 2z
2
(1 + cosh ) and erfc(x) is the complementary error function, which satises
the identity
erfc(x) = 2 erfc(x). (120)
It is not possible to present the solution of the integral in equation (119) in terms of known
functions. In spite of this inconvenience, it is easy to see that this function f (, z) has strong
convergence, proving that we indeed regularize the contribution for the two-point function at
the one-loop level. Similar calculations can be carried out for the four-point function. From
the discussions it should be clear that it is a simple matter to derive the Schwinger functions
of the theory at least in the one-loop level.
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8563
There are two rather subtle points that we have to investigate. The rst one is related to
the behaviour of the random noise near the event horizon and the second one is the validity of
the program beyond the Euclidean signature.
5. Conclusions and perspectives
The stochastic quantization method was used to study self-interacting elds in manifolds
which can be analytically continued to the Euclidean situation, i.e., static Riemannian
manifolds, namely the Einstein manifold and the Rindler manifold. First, we have solved
a Langevin equation for the mode coefcients of the eld, then we exhibit the two-point
function at the one-loop level. It was shown that it diverges and we have used a covariant
stochastic regularization to regularize it. It was shown that, indeed, the two-point function is
regularized.
A natural question that arises, when we work in Rindler space, is that what happens
to the noise eld correlation function, given by equation (47), near a horizon. From this
equation, we see that, whenever we have g = det g

= 0, this correlation function diverges,

and, therefore, all n-point correlation functions (
1
, x
1
)(
2
, x
2
) (
n
, x
n
)

will have
meaningless values, in virtue of the solution of the Langevin equation. We may implement
a brick-wall-like model [64, 65] in order to account for these effects; in other words, we
may impose a boundary condition on solutions of the Langevin equation at a point near the
horizon. On the other hand, in the limit g = det g

, all the correlation functions

vanish. We would like to point out that another powerful way to understand the behaviour
of the correlation functions in the limit g = det g

= 0 may be the stochastic quantization

of the two-dimensional Schwinger model in a curved background. This can also be a way to
study the most pertinent form of the Langevin equation to be used in curved spacetimes. This
subject and also the link between our results and the program to implement the perturbative
renormalization in domains where the translational invariance is broken [6669] is under
investigation by the authors.
We discussed also that the presence of the Markov parameter as an extra dimension
leads us to a regularization scheme, which preserves all the symmetries of the theory under
study. This is a quite important point in a regularization procedure. It must preserve all the
symmetries of the unregularized Lagrangian. Although many authors have stressed that we
cannot expect that a regularization-independent proof of the renormalization of theories in a
curved background exists, since the stochastic regularization is not an action regularization,
may be a way to construct such a proof. We are aware of the fact that the stochastic quantization
program can be implemented without problems (modulo event horizon problems, etc), if it is
possible to perform the Wick rotation, obtaining a real Euclidean action.
The picture that emerges from the discussions of the paper is that the implementation of
the stochastic quantization in a curved background is related to the following fact. For the
static manifold, it is possible to perform a Wick rotation, i.e., analytically extend the pseudo-
Riemannian manifold to the Riemannian domain without a problem. Nevertheless, for non-
static curved manifolds we have to extend the formalismbeyond the Euclidean signature, i.e., to
formulate the stochastic quantization in pseudo-Riemannian manifold, not in the Riemannian
space (as in the original Euclidean space) as was originally formulated. See, for example,
[11, 12]. In [11], the authors proposed a modication of the original ParisiWu scheme,
introducing a complex drift term in the Langevin equation, to implement the stochastic
quantization in Minkowski spacetime. Gozzi [12] studied the spectrum of the non-self-
adjoint FokkerPlanck Hamiltonian to justify this program. See also the [70, 71]. Of course,
8564 G Menezes and N F Svaiter
this situation is a special case of ordinary Euclidean formulation for systems with complex
actions.
The main difference between the original Euclidean and the implementation of the
stochastic quantization in the Minkowski spacetime is the fact that in the usual case, a
real Euclidean action, the approach to the equilibrium state is a stationary solution of the
FockerPlanck equation. In the Minkowski formulation, the Hamiltonian is non-Hermitian
and the eigenvalues of such Hamiltonian are in general complex value. The real part of
such eigenvalues are important to the asymptotic behaviour at a large Markov time. The
approach to the equilibrium is achieved only if we can show its positive semi-deniteness.
The fundamental question is: what happens if the Langevin equation describes diffusion
around complex action? Some authors claim that it is possible to obtain meaningful results
out of Langevin equation diffusion processes around complex action. See, for example, the
seminal paper of Parisi [72]. Klauder et al [73] investigated the complex Langevin equation,
where some numerical simulations in one-dimensional systems was presented. See also the
[74, 75]. Finally we would like to mention the approach developed by Okamoto et al [76],
where the role of the kernel in the complex Langevin equation was studied.
We would like to remark that there are many examples where Euclidean action is complex.
The stochastic quantization in the Minkowski spacetime, as we discussed; system with
chemical potential as for example QCD with a non-vanishing chemical potential at nite
temperature; for SU(N) theories with N > 2, the fermion determinant becomes complex
and also the effective action. Complex terms can also appear in the Langevin equation for
fermions, but a suitable kernel can circumvent this problem. Another quite instructive case
that deserves our attention is the stochastic quantization of topological eld theory. One of
the peculiar features within this kind of theory is the appearance of a factor i in front of the
topological action in the Euclidean space. In a topological theory, the path integral measure
weighing remains e
iS
, even after the Wick rotation. An attempt to use a Markovian Langevin
equation with a white noise to quantize the theory, fails since the Langevin equation will not
tend to any equilibrium at a large Markov parameter. There are different proposals to solve
the above-mentioned problem. In a pure topological ChernSimons theory, Ferrari et al [77]
introduced a non-trivial kernel in the Langevin equation. Other approach was developed by the
Menezes et al [28]. These authors showed that using a non-Markovian Langevin equation with
a coloured random noise the convergence problem can be solved. These authors proved that it
is possible to obtain convergence towards equilibrium even with an imaginary ChernSimons
coefcient. We conclude by saying that several alternative methods have been proposed
to deal with interesting physical systems where the Euclidean action is complex. These
methods do not suggest any general way to solve the particular difculties that arise in each
situation.
The program of using the complex Langevin equation for an investigation of quantum
eld theories in the Minkowski spacetime is not yet established and is still under discussion.
It is clear that until the problem that we discussed above is solved, one cannot say for sure that
the implementation of the programof stochastic quantization in non-static pseudo-Riemannian
manifolds can be successfully implemented.
Acknowledgments
We would like to thank S Jofly and G Flores-Hidalgo for enlightening discussions. This
paper was supported by Conselho Nacional de Desenvolvimento Cientico e Tecnol ogico do
Brazil (CNPq).
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8565
Appendix. The Kruskal extension of the Schwarzschild spacetime
The spacetime of a spherically symmetric non-rotating and eletrically neutral body is described
by the Schwarzschild metric
ds
2
=
_
1
2M
r
_
dt
2
+
_
1
2M
r
_
1
dr
2
+ r
2
d
2
, (A.1)
where d
2
= d
2
+ sin
2
d is the metric of the unit 2-sphere. The coordinate system
(t, r, , ) provides a frame in which the metric components are time independent. Since,
it is possible to nd a Killing vector eld which is hypersurface orthogonal to the family of
spacelike hypersurfaces t = cte, the Schwarzschild solution is also static. The line element
given by equation (A.1) describes the external gravitational eld generated by any spherical
mass, whatever its radius [78]. This includes several interesting situations, as a spherically
symmetric star which undergoes radial pulsations or even a radial spherically symmetric
gravitational collapse. In the extreme case of a complete gravitational collapse, we have to
consider the Schwarzschild line element as an empty spacetime solution for all values of r.
Since the line element has two singularities, one at r = 0 and another at r = 2M, it represents
only one of the patches 0 < r < 2M or 2M < r < . It is not difcult to show that although
the Schwarzschild line element is singular at r = 2M, all the invariants constructed with the
Riemann tensor and its contractions are well behaved at r = 2M, and this point is a singularity
due to an inappropriate choice of coordinates.
On the other hand, choosing the Schwarzschild line element to describe the patch
0 < r < 2M, we have that the curvature scalar diverges at r = 0, since we have
R

=
48M
2
r
6
. Therefore this point is a real spacetime singularity. Thus, although
it is not possible to extend the spacetime across the singularity, it is possible to analytically
continue the manifold from r > 2M to the region r < 2M(r = 0). The maximal extension of
the manifold described by the Schwarzschild line element with 2M < r < was obtained
by Kruskal.
From the condition for null geodesics it is possible to show that the radial null geodesics
in the Schwarzschild spacetime are given by
_
dt
dr
_
2
=
_
r
r 2M
_
2
, (A.2)
and allow us to dene r

as
r

=
_
dr
_
1
2M
r
_ = r + 2M ln
_
r
2M
1
_
, (A.3)
and it is clear that the radial geodesics must satisfy t = r

+ cte or t = r

+ cte.
The next step is to dene the null coordinates u and v, where u = t r

and v = t + r

.
Obviously, it is possible to write the Schwarzschild line element in terms of the null coordinates
u and v, and we have
ds
2
=
2Me

r
2M
r
e
(vu)
4M
du dv + r
2
d
2
, (A.4)
where the metric of the unit 2-sphere is unaltered. To go further, let us dene new coordinates
U and V, where
U = exp
_
u
4M
_
(A.5)
and
V = exp
_
v
4M
_
. (A.6)
8566 G Menezes and N F Svaiter
Using the coordinates U and V, the metric can be written as
ds
2
=
32M
3
e

r
2M
r
dU dV + r
2
d
2
. (A.7)
The value r = 2M is no longer a singularity, since it corresponds to U = 0 or V = 0. Finally,
to obtain the Schwarzschild metric in the Kruskal form we have only to dene the coordinates
T and X, by choosing T =
1
2
(U +V) and X =
1
2
(V U). The nal form of the Schwarzschild
metric in terms of the Kruskal coordinates (T, X, , ) is
ds
2
=
32M
3
e

r
2M
r
(dT
2
+ dX
2
) + r
2
d
2
. (A.8)
Note that the coordinate transformation between the original coordinates (t, r) and the Kruskal
coordinates (T, X) is given by
X
2
T
2
=
_
r
2M
1
_
exp
_
r
2M
_
(A.9)
and
t = 4M tanh
1
_
T
X
_
. (A.10)
Therefore, we show how it is possible to analytically continue the manifold fromr > 2M
to the region r < 2M, and the event horizon, a sphere of radius r = 2M is only a coordinate
singularity, which can be removed by a suitable coordinate transformation. Although the
apparent singularity at the horizon has disappeared, there are true singularities in the Kruskal
extension of the Schwarzschild spacetime. The physical singularity at r = 0 corresponds to the
values X =

T
2
1 and X =

T
2
1. The original Schwarzschild solution for r > 2M
corresponds to the region where observers can obtain information. Since there are two event
horizon, the future event horizon and the past event horizon, the Kruskal spacetime denes
two outer asymptotically at regions and also two regions inside the event horizon, bounded
by the past and future singularities. The black hole is the maximal analytic extension of the
vacuum Schwarzschild solution. It is well known that close to the horizon, the Schwarzschild
coordinates t and r behave as Rindlers spacetime coordinates.
References
[1] Parisi G and Wu Y S 1981 Sci. Sin. 24 483
[2] Damgaard P and Heiffel H 1988 Stochastic Quantization (Singapore: World Scientic)
[3] Namiki M 1993 Prog. Theor. Phys. Suppl. 111 1
[4] Sakita B 1985 Quantum Theory of Many Variable Systems and Fields (Singapore: World Scientic)
[5] Damgaard P and Heiffel H 1987 Phys. Rep. 152 227
[6] Rumpf H 1986 Phys. Rev. D 33 942
[7] Halpern M B 1987 Phys. Lett. B 185 111
[8] Halpern M B 1987 Ann. Phys. 178 272
[9] Chan H S and Halpern M B 1987 Z. Phys. C 36 669
[10] Rumpf H 1993 Prog. Theor. Phys. Suppl. 111 63
[11] Huffel H and Rumpf H 1984 Phys. Lett. 148B 104
[12] Gozzi E 1985 Phys. Lett. 150B 119
[13] Kac M and Logan J 1974 Fluctuations Fluctuation Phenomena ed E W Montroll and J L Lebowitz (Amsterdam:
North-Holland)
[14] Zwanzig R 2001 Nonequilibrium Statistical Mechanics (New York: Oxford University Press)
[15] Kubo R, Toda M and Hashitsume N 1991 Statistical Physics (Heidelberg: Springer)
[16] Parisi G and Sourlas N 1982 Nucl. Phys. B 206 32
[17] Floratos E and Iliopoulos J 1983 Nucl. Phys. B 214 392
[18] Grimus W 1983 Z. Phys. C 18 129
Stochastic quantization of scalar elds in Einstein and Rindler spacetime 8567
[19] Fukai T, Nakazato H, Ohba J, Okano K and Yamanaka Y 1983 Prog. Theor. Phys. 69 1600
[20] Damgaard P H and Tsokos K 1984 Nucl. Phys. B 235 75
[21] Damgaard P H 1993 Prog. Theor. Phys. Suppl. 111 43
[22] Fox R F 1977 J. Stat. Phys. 16 259
[23] Fox R F 1977 J. Math. Phys. 18 2331
[24] Breit J D, Gupta S and Zaks A 1984 Nucl. Phys. B 233 61
[25] Bern Z 1985 Nucl. Phys. B 251 633
[26] Alfaro J 1985 Nucl. Phys. B 253 464
[27] Iengo R and Pugneti S 1988 Nucl. Phys. B 300 128
[28] Menezes G and Svaiter N F 2006 J. Math. Phys. 47 073507
[29] Menezes G and Svaiter N F 2007 Physica A 374 617
[30] Zwanziger D 1981 Nucl. Phys. B 192 259
[31] Bern Z and Halpern M B 1986 Phys. Rev. D 33 1184
[32] Sakamoto J 1986 Prog. Theor. Phys. 76 966
[33] Bern Z, Halpern M B, Sadun L and Taubes C 1985 Phys. Lett. B 165 151
[34] Bern Z, Halpern M B, Sadun L and Taubes C 1987 Nucl. Phys. B 284 1
[35] Bern Z, Halpern M B, Sadun L and Taubes C 1987 Nucl. Phys. B 284 35
[36] Halpern M B 1993 Prog. Theor. Phys. Suppl. 111 163
[37] Eddington A S 1975 The Mathematical Theory of Relativity (New York: Chelsia)
[38] Rindler W 1966 Am. J. Phys. 34 1117
[39] Fulling S A1972 PhDThesis Scalar QuantumField Thory in a Closed Universe of Constant Curvature Princeton
University, Princeton
[40] Fulling S A 1973 Phys. Rev. D 10 2850
[41] Sciama P W, Candelas P and Deutsch D 1981 Adv. Phys. 30 327
[42] Huang W-H 1992 Phys. Rev. D 46 2560
[43] van Kampen N G 1986 J. Stat. Phys. 44 1
[44] Birrel N D and Davies P C W 1982 Quantum Fields in Curved Space (Cambridge: Cambridge University Press)
[45] Grib A A, Mamayev S G and Mostepanenko V M 1994 Vacuum Quantum Effects in Strong Fields (St Petesburg:
Friedman Laboratory)
[46] Bander M and Itzykson C I 1966 Rev. Mod. Phys. 38 346
[47] Prudnikov A P, Brychkov Yu A and Marichev O I 1986 Integrals and Series vols 1 and 2 (London: Gordon and
Breach)
[48] Gradshteyn I S and Ryshik I M 1980 Table of Integrals and Products ed A Jeffrey (New York: Academic)
[49] Bateman H 1954 Tables of Integral Transformations vol 1 (California: California Institute of Technology)
[50] Birrel N D 1981 Quantum Gravity 2 ed C J Isham, R Penrose and D W Sciama (Oxford: Oxford University
Press)
[51] Bunch T S 1981 Ann. Phys. 131 118
[52] Brunetti Rand Fredenhagen K1997 Interacting quantumelds in curved space: renormalizability of
4
Preprint
gr-qc/9701048
[53] Kruskal M D 1960 Phys. Rev. 119 1743
[54] Wald R 1994 Quantum Field Theory in Curved Spacetime and Black Hole Thermodynamics (Chicago, IL:
University of Chicago Press)
[55] Strominger A 1994 Lectures on Black Holes Le Houches, Fluctuating Geometries in Statistical Mechanics and
Field Theory ed F David, P Ginsparg and J Zinn-Justin (Amsterdam: Elsevier)
[56] Brout R, Massar S, Parentani R and Spindel Ph 1995 Phys. Rep. 260 329
[57] Ford L H 1997 Quantum eld theory in curved spacetime Preprint gr-qc/9707062
[58] Padmanabhan T 2003 Gravity and the thermodynamic of horizon Preprint gr-qc/0311036
[59] Christensen S M and Duff M J 1978 Nucl. Phys. B 146 1178
[60] Candelas P and Raine D J 1976 J. Math. Phys. 17 2101
[61] Unruh G W 1976 Phys. Rev. D 14 870
[62] Fulling S A and Ruijsenaars S N M 1987 Phys. Rep. 152 135
[63] Zarro C A D 2006 MA Thesis Teoria Qu antica de Campos na Presenca de Horizonte de Eventos, Centro
Brasileiro de Pesquisas FsicasCBPF
[64] t Hooft G 1985 Nucl. Phys. B 256 727
[65] t Hooft G 1988 Acta Phys. Pol. B 19 187
[66] Fosco C D and Svaiter N F 2001 J. Math. Phys. 42 5185
[67] Caicedo M I and Svaiter N F 2004 J. Math. Phys. 45 179
[68] Svaiter N F 2004 J. Math. Phys. 45 4524
8568 G Menezes and N F Svaiter
[69] Aparicio Alcalde M J, Flores G H and Svaiter N F 2006 J. Math. Phys. 47 052303
[70] Huffel H and Landshoff P V 1985 Nucl. Phys. B 260 545
[71] Callaway D J, Cooper F, Klauder J R and Rose H A 1985 Nucl. Phys. B 262 19
[72] Parisi G 1983 Phys. Lett. B 131 393
[73] Klauder J R and Peterson W P 1985 J. Stat. Phys. 39 53
[74] Klauder J R 1984 Phys. Rev. A 29 2036
[75] Ambjorn J and Yang S K 1985 Phys. Lett. 165 140
[76] Okamoto H, Okano K, Schulke L and Tanaka S 1989 Nucl. Phys. B 342 684
[77] Ferrari F and Hufel H 1991 Phys. Lett. B 261 47
[78] Birkhoff G 1923 Relativity and Modern Physics (Cambridge, MA: Havard University Press)