Sei sulla pagina 1di 46

SLAC-PUB-5536 April 1991 T






Stanford Linear Accelerator Center Stanford University, Stanford, California 94309




Department of Physics Stanford University, Stanford, California 94305

Two-dimensional quantum gravity coupled to conformally invariant matter gravity

with central charge c > 25 has been proposed as a toy model for quantum in higher dimensions. and unstable to forming The associated Wheeler-Dewitt equation

is non-linear

a condensate of baby universes. Small fluctuations

This will occur even in describe the

the classical c + co limit. propagation Dewitt

about this background a more conventional

of single universes and satisfy The resulting

linear Wheeler-



cosmology depends on details of

the non-linear


In particular

the existence of a large scale cosmological tachyon potential near

constant is determined its minimum.

by the behavior of a string theoretic


to Nuclear

Physics B

* Work supported by the Department of Energy, contract DE-AC03-76SF00515. t Supported in part by NSF grant PHY89-17438

1. Introduction
This is a revised and expanded version of our previous preprint Semi-Classical material conclusion gravity Limit of Quantum Gravity Isn t (SLAC-PUB-5413). possibility entitled The

The new

in Sections 7 and 8 provides an alternative that a non-zero conventional cosmological

to our previous

constant in two-dimensional geometry.

is inconsistent

with a large scale continuum view of quantum equation [l]


to the traditional

cosmology, the universe is gov-

erned by a linear Wheeler-Dewitt

The exact form- of H,, includes both gravitational grangian is constrained this conventional is determined and matter

IS) = 0.
by the field-theoretic degrees of freedom. Lagrangian


The form of this LaIn recent years proposed, in [a-12]. The

only by the principle

of general covariance.

view has been challenged and a far richer picture equation is itself quantized and non-linear

which the Wheeler-Dewitt

number of universes is indefinite,

being subject to the rules of third quantization, determined condensates of uniof baby to

and the laws of nature are given by dynamically verses. In particular,

it has been argued that the emission and absorption constants, and even forces the cosmological

universes shifts coupling zero [13-171.


In this paper we shall analyze these issues in the relatively two-dimensional gravity coupled to conformally invariant matter.

simple context of This theory has interest, and be-

received a lot of attention cause of its application on coupling continuum gravity

in recent years, both due to its intrinsic strings.

to non-critical

Most of the effort has been focused central charge D 5 1, using either in terms of matrix models of ran-

to matter

theories with

field theory or a discrete description However, the dynamics

dom surfaces. gravity

more closely resembles higher dimensional string theory with D > 25 [l&19].

when one studies non-critical

In that

case, the kinetic

energy associated with the scale factor of the metric is negative, and familiar cosmological solutions (such as de

just as it is in higher dimensions, Sitter space with positive quantum cosmology, with



are obtained. to string theory

Two-dimensional in a Lorentzian equation is of

D > 25, is identical

fields. theory

target-space the condition

with background that the string constant

The non-linear beta-functions

Wheeler-Dewitt vanish,

and the behavior

the cosmological target-space!

is replaced by the dynamics

of the tachyon

field in

Our main result is that the non-linear become important,

terms of the Wheeler-Dewitt This will correWe then constant

equation inevitably

even in the classical limit.

spond to the universe interacting address the question of whether

with a background

of baby universes.

the condensate shifts the cosmological

to zero as suggested by Coleman [13].$ This appears to be the case if the tachyon potential has a smooth minimum. minimum On the other hand, a very mild singularity at

the potential constant. possibility.

leads to a large scale behavior with arbitrary model results of Moore et al.

cosmological the latter

Recent matrix

[21] vindicate

The outline the matrix non-linear formulation

of the paper is as follows.

In Section 2 we briefly gravity

review how

model description Wheeler-Dewitt

of two-dimensional equation. gravity

leads to an essentially

In Section 3 we describe the continuum in the conformal fields. gauge, and show how it In Section 4 we consider Wheelerof

of two-dimensional

leads to an effective action for the target-space the cosmologically Dewitt interesting

case of D > 25, and derive the non-linear

equation from the effective action. In Section 5 we consider the evolution

couplings with scale, and examine the relation between the target-space of motion and the renormalization


group. In Sections 6 and 7 we discuss whether of the cosmological constant at large scale. using two

these ideas can lead to the vanishing In Section 8 we give an explicit


of the tachyon beta-function

t This approach to two-dimensional quantum cosmology was discussed by Banks in [20]. $ The Euclidean saddle point, of Baum [15], Hawking [16] and Coleman [13], has an analog in the two-dimensional theory for D > 25 [19]. 3




We conclude with quantum

speculations gravity.

about how

these ideas may carry over into four-dimensional

2. Matrix Models and the Non-linear Wheeler-Dewitt Equation

In this section we briefly Dewitt equation from a matrix review the derivation of the non-linear gravity Wheeler-

model of discrete quantum

[22]. In partic-

ular, we consider the single matrix

model described by the path integral

2 = -.$ J dMN2emN , tr CM)

where M is a Hermitian N x N matrix. Perturbation


theory generates a set of Feyn-

man diagrams with vertices depending on V. Diagrams of genus h are weighted by a factor of N-2h, of a sphere. so in the N t oo limit the surviving graphs have the topology with a sum geometries, of Euclidean

By considering

the dual graphs, (2.1) can be identified to Euclidean signature two-dimensional

over discrete approximations and in the continuum gravity. limit

this can be thought

of as a quantization

Consider next the discretization

of surfaces with a boundary amplitude

of length 1. The

sum over such geometries defines a Wheeler-Dewitt

Z(Z) which is given


Z(Z)= s

I+; J

dMN2 tr M,--Ntr V(M), N

= ;M + gM4. Similarly,

the ampli-

where we have chosen for illustration

tudes for geometries with m boundaries

of lengths II, Z2, . e.7 1m are given by


* ..,Zm) =


dMN2 NtrM

...NtrM e-NtrV(M).


It is a simple matter

to derive the Schwinger-Dyson


equations corresponding

to a

shift of M [22,23]. Th e simplest such equation is

Z(Z+1) + 4 2(1+3)

= $2

qp;z-l-p) p=o


In the large N limit, loop equation

Z( Zr, 12) - N22(Zr)2(Z2)

and (2.4) reduces to the non-linear



+ 4 Z(Z+3) = c



with boundary


Z(0) = g. The non-linearity

of this equation has a natand

ural interpretation joining.

in terms of one-dimensional

universes, or strings, splitting l/N

Notice t-hat the string coupling constant in which the non-linear

does not appear in (2.5), so Furthermore, the

there is no limit parameter an initial

terms can be ignored. cosmological

g which controls the two-dimensional condition


appears as

in (2.5). A 1inear equation for fluctuations

in 2, or equivalently

for variations fluctuations

of 2 with respect to g, can easily be derived from (2.5). It is these which should be identified with the conventional Although Wheeler-Dewitt am-

plitudes of the single universe theory. not look like a conventional fact shown that the continuum

the linearized version of (2.5) does equation, Moore

Wheeler-Dewitt limit

et al. [al] have in

describes the same physics. theory. The

In the remainder of the paper we will be considering the continuum above calculations the importance

were presented in order to emphasize from another viewpoint effects in the Wheeler-Dewitt equation.

of non-linear

3. Two-dimensional



and String
of non-critical

string theory. For the

We shall now review the continuum This will serve to fix our notation moment,

and make contact with previous work. (o, &)

assume that both the world-sheet


and the space of matter fields


i = 1,. . . , D are Euclidean,

and that D < 25. The action is

s = $

/- d2a fi


* &x

+ x0} ,


where Tab is the two-dimensional To carry out the path integral lowing steps are taken:

metric and As is the bare cosmological over metrics Tab and matter fields Xi,

constant. the fol-

Gauge fixing:

The over-counting

of metrics due to general coordinate an arbitrary background, or fiducial,


ance is removed by introducing

qab( fl) and defining a conforma/ gauge,

The remaining mode. degree of freedom, c$, is called the Liouville over metrics field or conformal over 4, with Liouville

The path integral

reduces to an integral

a d-dependent form [24],

Fadd eev-Popov determinant,

which has the familiar

Note that the couplings exactly of the conformal mode in the Liouville background. action are This is

those of a string matter

field in a non-trivial

the first indication

of the close parallel between qSand the Xi that we will

elaborate on in due course.


Regularization: regularized

The theory

has ultra-violet a cutoff,

divergences which need to be distance. The cutoff scale

by introducing

or shortest metric,

is defined with reference to the fiducial metric, Yab. Therefore the regularized

?&, rather than the original covariant.

theory is not manifestly


In order to define the theory at some size scale, we have and matter field fluctuations on smaller

to integrate out both gravitational scales.

The result is an effective Lagrangian,

which does not necessarily

look covariant, since the renormalization reference to the fiducial metric.

procedure is to be carried out with one might have quite large

In particular

geometries, as measured in the original metric, Yab, which nevertheless appear as short distance fluctuations illustrated in Figure 1. on a scale set by 5&. An example of this is

The requirement that the original theory be covariant can be stated as a set of conditions, that the path integral does not depend on our choice of fiducial metric, on the allowed couplings of invariant

Tab. These turn out to place quite strong restrictions the fields, Xi and 4. In particular,

the theory must be reparametrization

with respect to Tab, and if we consider variations of the conformal part of +ab, i.e. of det 9, we find that the path integral has to be conformally .,turn implies that beta-functions All this can be summarized of all couplings must vanish. as follows. We start with a generally covariant invariant, which in

theory of gravity coupled to scalar fields, X. In order to define the path integral we fix a gauge and regularize in a non-covariant manner. The resulting theory involves a scalar field, c$,in addition to the matter fields, and is in general quite complicated. The original covariance appears as a set of restrictions that all the beta-functions on the couplings, which

include the requirement way of stating


Notice that in this The Liouville field

things 4 and Xi are placed on equal footing. to an additional target-space dimension.

has been promoted the quantization

This approach to

of two-dimensional

gravity has been advocated by a number of

authors [19,20,25,26,27]. The object of interest is thus some reparametrization invariant scalar field

theory in two dimensions, with a priori

quite general couplings,


~Jd2~~{T(X)+4.)a,X ~bXYGpU(X)+2jl~(X)+...}.


There are D+l Liouville mode.

scalar fields Xfi,


both the matter kinetic

fields, Xi,

and the field has down the

In order to have a standard with Q2 = v.

term, the Liouville

been resealed to X0 = $5, terms of scaling dimensions possible couplings involving two-dimensional curvature ii.

We have only written

zero and two,* but there is an infinite more derivatives

sequence of

on the Xfi and higher powers of the

This class of theories has been extensively the action


in string theory, where spacetime invariance dimenof the

(3.4) d escribes strings in background equations, implementing

fields in D+l the conformal

sions. The beta-function two-dimensional Q(X)

theory, have the form of field equations in target-space (tachy on, dilaton and graviton fields respectively),

for T(X), along with

and GfiV(X)


fields representing higher order couplings. and creation and annihilation

These field equations describe eigenmodes of universes

the propagation

of the particle-like

strings in spacetime, or more to the point of this paper, one-dimensional containing matter fields! The tachyon field, T(X), cosmological is of primary

interest because

it controls the two-dimensional cosmological background


To see this, note that the tachyon

term in the Liouville action in target-space,

(3.3) corresponds to a particular


= XeGxo .
obtained by setting beta-functions


The string theory equations of motion, zero, are derivable from an action. theory, containing

For simplicity,

we will work within

a truncated To

only the lowest order couplings, the target-space



and GPy(X).

leading order in derivatives,

action for these fields is [28]

I = .S
where V(T)



The general

= -T2 + &T + . . . is the tachyon effective potential.

* For simplicity, we have not included the anti-symmetric

tensor field. Its presence would not qualitatively alter our cbnclusions. t We will use the string theory names for the target-space fields, but the reader should keep in mind their cosmological interpretation. 8

form of V(T) two-dimensional

is not known,

but we show how to obtain the leading terms from theory in Section 8. Our answer is, of course, not group beta-functions always depend on the reg-


universal because renormalization ularization . -

procedure used. This prescription ambiguities

dependence is believed to correspond

to field redefinition the beta-function

in target space. In fact all higher order terms in and therefore

can be arranged to involve target space derivatives, leaving only -T2

be removed from the potential object

[33,34]. On the other hand, the

to the Wheeler-Dewitt

defined in this way will probably of a one-dimensional universe.

not correspond


We will return to this point in Sections

7 and 8. The equations of motion which follow from the above action are

v2T - 2vQ . VT =

V (T), + V(T),

25-D V2@ - 2(i7@)2 = - 6


For D < 25 these equations have a simple solution, even when higher order terms in the beta-functions

which is known to be exact [29],

are included

An important

feature of this linear dilaton


is that the strength scale,

of the

string loop coupling constant is related to the two-dimensional

9 = 90,




We can only expect the effective field theory to be simple where this coupling weak.

A background function non-linear equation

tachyon field can be added to the exact solution depends on the shape of the effective potential, If we assume that the background

(3.8). Its beta-


and is

in general.

field is weak, and only

depends on X the equation can be linearized as follows, ,

. -

and has solutions


Such a homogeneous background dimensional theory

(+q)xo 2

(3.11) precisely to the twoand Kawai [31].* In


discussed by David regime, X0 + -co,

[30] and by Distler the solution

the weak-coupling exponent damped.

with the negative sign in the

is more important

to the physics, since the other one is more rapidly theory

Note also that in the D + -oo semi-classical limit of the Liouville

this choice of sign makes the cosmological term (3.11) reduce to T = Aeaxo = Xe4, as it should. An important the strength In this limit feature of the D < 25 theory is that the coupling corrections gee governing

of quantum

(string loops) tends to zero when X0 + -co. small. Furthermore, -co. we can consider a backIn the vicinity disappear. of T = 0 On the strength wall

strings are metrically

ground solution the non-linear

which tends to T = 0 as X0 + terms in the classical target-space


other hand, as X0 -+ 00, the limit

of large scale factor,

the interaction

increases. It has been argued by Polchinski which prevents the particle-like

[32] that this leads to a reflecting to large sizes.

strings from penetrating

* Note that our X0 differs, by a sign, from the resealed Liouville field of Distler and Kawai. 10

4. Quantum

C osmology

in Two Dimensions

We have so far been concentrating Polchinski . -

on theories with D < 25.t Let us now follow

[ 191 and consider the case D > 25, which is more relevant to cosmology. of the target-space equations (3.7) does not depend on the value of changes when D signature, a linear

The derivation

D. On the other hand, the nature of their solutions qualitatively

becomes greater than 25. In particular, dilaton background if G,, has Euclidean

will no longer be a solution for D > 25. However, there exists It is given by

a solution

analogous to (3.8) if the target-space is Lorentzian.

G,,, = 17~


where q 2 = v it is not consistent Lorentzian.

and the conformal to treat

mode, X0, is time-like! as Euclidean


speaking, is

the world-sheet

if the target-space

The reason is that the two-dimensional

action is then unbounded, We should therefore of Lorentzian signa-

since the X0 kinetic reformulate

term has opposite sign to that of Xi. field theory on a world-sheet

the two-dimensional

ture in order to discuss the D > 25 case. Unfortunately to perform the steps involved in the quantization renormalization, that a consistent etc.). Our working assumption,

it is not at all clear how

of such a theory (regularization, which may be unwarranted, is



will lead to the same covariant In interpreting


space equations as the formal Euclidean calculation. of motion as a Lorentzian implicitly

the equations path integral Such configura-

field theory of strings, the two-dimensional and joining.

includes geometries which describe splitting

tions inevitably

include well-defined events at which the universe bifurcates and the

t Actually D 5 1, since the exponent in (3.11) is complex for D > 1, and the two-dimensional action is unbounded from below. $ We are free to choose the sign of the dilaton background. This corresponds, in fact, to the choice of the direction of time. The physics just tells us that small universes are weakly coupled for D < 25 and strongly coupled for D > 25. Our solutions (3.8) and (4.1) reflect the convention that the universe was small at early times. 11


metric is singular.

These will be observable, or even catastrophic, the path integral receives contributions from the background, which also inspace

for one-dimensional

observers. In addition,

from universes being absorbed or emitted volve two-dimensional . the metric singularities

(see Figure 2). By contrast,

in Euclidean

can be chosen with no singularities. methods to compute renormalization

It should be noted that we only group beta-functions, but our

use Euclidean

subsequent discussion of the cosmology takes place with Lorentzian The gravitational proper Planck-scale. + coupling However, -co in two-dimensions is dimensionless,

signature. so there is no strength gee

in the case D > 25 the coupling the theory is strongly

increases as X0 small strings. ultraviolet.

so that

coupled for sufficiently

No longer can quantum

mechanics (string loops) be ignored in the is spontaneously induced, and define

One can say that a Planck-scale

it by the point at which gse @ = 1 . The factor of go can be absorbed by a constant shift of the dilaton. The effective Planck-scale is then set by q-l, and depends on

the number of scalar fields in the theory. limit for gravitational fluctuations.

In particular,

D + 00 is a semi-classical
conditions is complicated

The question of initial

because the theory is strongly coupled at early times. The short distance physics is summarized by some unknown initial state at the Planck-time, which then evolves

in the weakly coupled theory. in target-space target-space

In a quantum

theory this means a wave-function conditions on the

and in a classical theory it corresponds to initial

fields. (4.1) th e 1inearized equation for a homogeneous tachyon

In the background field reads

and is solved by



= Xe (-f*fi)X.

The tachyons.

One of the solutions decays with time, but the other one grows exponentially. system is unstable and is likely to form a condensate of background

Now consider fluctuations background,

of some field in the exponentially Take, for example, Ic. The target-space

growing tachyon a tachyon with

TB(X )

= X e(-f+fi)XO*

some non-zero space-like momentum translation invariant.

action (3.6) is not timeit is convenient to

In order to describe physical fluctuations by a field redefinition, U(X) = e -(a(x) T(X)

absorb the em2 pre-factor

which has the form


for tachyons.

A fluctuation


= Uk(Xo) eikiX satisfies a linear equation,

a,& + (k2 + V(TB) - ;)Uk

Near the top of the potential exponential the tachyon background

= 0.
is well approximated

by the

form (4.3) and we can drop the contribution

of all but the leading terms equation, whereupon

of the potential

V(T) = -T2 + &T + . . . in the fluctuation

(4.5) b ecomes @Uk + (k2-2-;)Uk

This has the form of a Wheeler-Dewitt excitation.

l + qXe (-f+@ixouk

= 0.

equation for a universe with some matter wherey = -s+

To see that, define a scale factor a = e2 , Ix

and reexpress (4.6) in terms of a,

{ (g-J2
Up to factor-ordering

+ (6a-9

+ gJ2}U~

= 0.


ambiguities, Lagrangian

this is the Wheeler-Dewitt of two-dimensional gravity,

from the mini-superspace

L = -( $2 - $ [P - (2+$

+ $2]

and cosmological

The three terms in square brackets are the matter, constant energy densities.


* The scale factor reduces to its classical value a = en


= e412in the q -+ co limit.

It seems that we have recovered a more or less conventional description constant of large scale cosmology. In particular,


the problem of the cosmological cosmological constant, the

is the usual one. In order to obtain vanishing increasing solution for T(X )

exponentially . -

must be fine-tuned

to zero. In other We the

words, the tachyon must be delicately are ignorant initial state,

balanced at the top of the potential.

about the short distance physics, which is supposed to determine so we have no way of gauging how likely

it is to find the system initial state is

balanced at the top of this potential. not allowed in a quantum However,

At any rate, such a fine-tuned

theory, because of the uncertainty

principle. starts

this is not the whole story.

Even if the tachyon background

out near the top of the potential higher-order non-linear

it will eventually

roll into the region where the At that point it is enwill deviate in such a way

terms in V(T)

cannot be ignored.

tirely possible that the linear Wheeler-Dewitt from (4.7). In particular, the cosmological

equation for fluctuations term could be modified

that it grows less rapidly a decreasing cosmological

than u2, causing one-dimensional constant

observers to measure One such cosmo-

as their universe expands.

logical model will be presented in the following shall see in Section 6, an entirely conventional the linearized Wheeler-Dewitt particular form. different

section. On the other hand, as we cosmological constant can appear in has a

equation at all scales if the tachyon potential

As we have already mentioned, two-dimensional Since the -T2



in the

theory will lead to different evolutions for the tachyon background. term in V(T) is universal the different schemes will all agree near pictures. For The

T = 0, but away from the origin they can present very different
example, the question of whether V(T) h as a minimum key issue here is to identify sponding to Wheeler-Dewitt the definition amplitudes

is scheme-dependent.

of the tachyon field most closely correin the two-dimensional cosmology. This

is a non-trivial candidate

task which has not been carried out, but we will outline a promising

scheme at the end of Section 8.


It should be emphasized that the non-linear do not disappear and joining in the semi-classical limit

effects that we are talking


D + oo. In particular,

the splitting equations

events described by the non-linear

terms of the target-space

are unsuppressed . coupling

even at late times. This may seem surprising

because the string

is becoming weak, with e@ = e-zXo. defined in (4.4) satisfies

Indeed, the canonical tachyon field


[V2 + (2 + $)]U

= iev:xoU2.

the tachyon mass

As we move toward

the semi-classical


q -+ 00, though,

squared increases as $, so that the unstable exponential the decreasing coupling strength. Since very little

growth of U compensates

is known about higher order non-linearities

in string theory,

we can only speculate about their detailed effect on the physics. However, the very existence of string interactions, usual Liouville along with the tachyon instability, shows that the

model described by an exponentially

growing tachyon background

is not the complete theory.

5. The Running
Before delving further connection

of Coupling

cosmology, we shall clarify the and the renormalization

into the two-dimensional

between the target space equations of motion

group flow of couplings in the two-dimensional

field theory.

The equations of motion for the target-space fields are that the beta-functions of all two-dimensional couplings vanish. From this one might conclude that the

couplings seen by a two-dimensional the correct interpretation. as renormalization renormalization hence determines

observer would not run. This, however, is not fields of the

We can think of the equations for the target-space identified with the logarithm

group equations with :X0

scale. The X0 dependence of the coupling functions their evolution with scale.

T, @, G,, . . .

This connection second-order trolled

may appear unfamiliar

because the equations

of motion


in X0 derivatives, equations.

whereas the usual renormalization

flows are con-

by first-order

The second-order nature of the flows is a special where the scale itself is a dynamical variable.

feature of theories containing



The situation formulation

is similar to the issue of time evolution gravity.

in the Wheeler-Dewitt

of quantum

We begin with an equation occurs. Reinterpreted,


IQ) = 0 which

seems to imply that no time evolution tells us how the wave function The Wheeler-Dewitt order. limit The first-order

though, the equation

of matter evolves with the expansion of the universe.

equation, like the equations of motion for T, a GPv, is second, Schrodinger equation is only recovered in a semi-classical [36]. In our twoD -+ 00 or field equations

in which gravitational theory,


become unimportant limit corresponds

dimensional equivalently

this semi-classical

to taking

Q + oo. In this limit

we will see how the target-space group equations.

reduce to the familiar


We consider first the case of fluctuations at the top of the tachyon potential

about the linear dilaton metric.

background A field A, at

with a Jut target-space

the nth mass level in string theory will contribute

to the effective action a term


/ fiC2

{ (vA,)~ - 2(1-n)A;

+ . . a} .


Its equation of motion in a linear dilaton



d 2+q&,+k2-2(1-n) (3x0 [(-)

As we saw previously, the tachyon rolling off the top of its potential.

for large D.



this equation has unstable solutions for n = 0, which describe Note, however, that the solutions graviton and

for n 2 1 are stable for all values of Q. In other words, the dilaton, higher couplings do not become tachyonic

Now, recalling

that the scale factor is a = e:X,

we can rewrite

(5.2) as

YQ 8 2i-~~~+k2-2(1-n)




Thus when q + 00 we find the first-order


a $ A, = -(k2 + 2(n-1))


This is the usual lowest order Callan-Symanzik mension 2n. In particular

equation for a coupling of bare di-

the field h,, has anomalous dimension a trivial matter

-k2 as expected.
of several

We have so far been considering free fields. A more stringent

sector, consisting

test of the above ideas should involve an interacting free sigma-model coupled to gravity. Expe-

matter sector such as an asymptotically

rience in flat space indicates that such a model will generate a new mass-scale and renormalize the vacuum energy accordingly. To investigate this, we consider the

example of a theory in which three the target-space dimensions are compactified of to a sphere of time-dependent dinates are left flat. dependence of r(X) in the semi-classical renormalization radius r(X ). coupling The remaining constant D - 3 spatial coorand the X0

The sigma-model is just the running limit

is then l/r,

of the coupling with scale. We find that of motion (3.7) reproduce the standard

the equations

group flow (mm dX0 1 0r


To see this we insert into (3.7) the metric

ds2 = -(dX)2

+ TV

dfl; + -&dXi)2


where d@ is the line element on a unit three-sphere.


The equations

of motion

then become

Taking the semi-classical solution

q + 00 limit, we find that to leading order these have T=O,


where c is an integration scale determined the renormalization by initial



to the induced dynamical



We see that this solution

indeed satisfies

group equation

(5.5). field are

The equations of motion (5.7) g overning the target-space gravitational valid only to first order in X-derivatives. of the one-loop beta-function X0 N c, the sigma-model gravitational equations

This corresponds to the region of validity perturbation theory. As

in the usual sigma-model

becomes strongly are important.

coupled, and higher order terms in the

The solution pect the non-trivial

(5.8) h as a vanishing sigma-model

tachyon field.

On the other hand, we exvacuum


to generate a two-dimensional

energy, which manifests itself as a source term in the tachyon equation of motion. To see this consider the effect of the non-trivial logical constant. As explained matter couplings on the cosmo-

before, it is the exponential

growth of the tachyon constant

field as it rolls off the top of the hill that gives rise to the cosmological term in the Wheeler-Dewitt equation

(4.7). W e might imagine that it would be so that the tachyon stays balanced at

possible to fine tune the initial


the top, and the cosmological in which the target-space

constant would thus vanish. In our simpler examples

was flat, we saw that this could indeed be done. Now, to the two-dimensional gravity will make

however, the coupling of the sigma-model it impossible. coupling determine

To see this, consider the higher order terms in the effective action One such term will be of the general form TvvRR. we should calculate the two-loop graviton beta-function, To but

T and G,,.
it properly

for our argument

it will suffice to note that there must be some term of this form vertex. There will As the

because string theory has a non-zero graviton-graviton-tachyon thus be an extra source term VVRR three-sphere contracts, in the tachyon equation

of motion.

this will knock the tachyon from the top of the potential. conditions to make

We would therefore

have to search for new fine tuned initial

the tachyon end up balanced at the top of the potential

at large scales. This need cosmological constant

to account for the matter vacuum energy is just the familiar problem.

Note that we are not able to use these techniques to follow the system into the strongly coupled regime in which r becomes small. However, since we know that [37], we may speculate

the flat space sigma-model

contains only massive particles

that well below the induced mass scale, the sigma-model

degrees of freedom decoubecoming

ple. This would correspond to the effective central charge of the matter smaller at some point in the evolution of the universe.

6. A Vanishing



In this section and the next we will follow the evolution of the cosmological constant as the two-dimensional universe expands. Recall that the spacetime equation

of motion for the tachyon field is obtained from the tachyon beta-function

PT = v2T - 2vQ.vT

- V (T) + . . . .


In Section 8 we find that,

in a particular


the leading terms in the


are given by V(T)

= -T2 + AT3 + . . . . There will also be higher-order

and couplings to other target-space fields. back-


to the beta-function

As we discussed at the end of Section 4, the evolution ground depends on the regularization field theory.

of the tachyon

procedure used to define the two-dimensional exists for which ampli-

In what follows, we will assume that a prescription

the tachyon field continues to be closely connected to the Wheeler-Dewitt tude as T rolls off the top of the potential. as speculations about the non-linear

The discussion should be understood obtained using such a scheme.


The large scale two-dimensional tachyon potential

cosmology is very sensitive to the form of the interesting to consider the case when of V(T),


It is particularly


has a local minimum.

This is consistent

with the cubic behavior

which we have found at leading order, but the minimum higher order corrections. First

may or may not survive is smooth (See

let us assume that the minimum

Figure 3). There will then be another exact solution of the target-space given by


T To, =


where T = To is the location Unlike the solution

of the potential

minimum is stable.

and @ = We see this by considering (6.2):

at T = 0, this solution

the linearized

equations for fluctuations

about the background

k2 -r + a$- + Qdor + V(To) 7 = 0, k2y+&+b%y=o, k2 h,v + a,zh,, + tj a0 h,, = o ,

where r = T-To,


= @-<PB, hpv = Gpv-71,w, and k is the spatial momentum p



the fluctuation.

These equations have the solution

7(X0) = pe y 7 h,, o: e(-!fj/G)xo.


and graviton

For k # 0 all fluctuations have a constant metric.

are damped, whereas for k = 0 the dilaton

mode, corresponding

to a resealing of the string coupling and the damped. of the

The fluctuations

of the higher modes are more strongly

From a two-dimensional potential fluctuations Dewitt

point of view, then, the theory at the bottom

must be a field theory with no unstable fluctuations. have no exponentially will not contain growing mode,

Because the tachyon Wheeler-

the corresponding term.


a cosmological


Thus, while we observer,

do not know the-exact form of the field theory seen by a one-dimensional we can say that he measures a zero cosmological Having studied the solutions the potential constant. of the stationary

in the neighborhood

points of


we can consider a solution

that begins at or near T = 0 and dilaton background that starts

rolls toward T G To. Because of the time-dependent off as CD(X) = -ix0 and eventually becomes )


qx> = --ix0 )
there will be a varying friction term in the equation of motion for T(X ).


T will behave as discussed before, with T(X )

and eventually

N X e (-f+&)XO


it will settle toward To with a damped motion

T - To + p ,(-z+$%%jx
In the semi-classical limit,

equation will


the form of the Wheeler-Dewitt


evolve from

{ (;u-$
for small a toward

+ (k2-2-;)

+ ;Xa2}V

= 0


. -

Y (p&


+ (k2+V(To)-$)

+ V(To)pa-V(To)}V
inflate, constant. the quantum

= 0


for large a. Th is means that the universe will initially settle down to a behavior with vanishing cosmological Although system, we have not yet thoroughly that investigated

but will eventually

theory of the will become

it seems very probable a dispersion

the quantum

wave functional

centered at To, with corrections contrast

which will shrink with X0, since the quantum -9x0 are of order e we 2 at late times. This behavior is in marked in the D 5 1 case in which the string coupling grows become

to that encountered

with the scale factor a. In those theories the effects of higher topologies increasingly important

as a increases, whereas for D > 25 their effect diminishes. that as the universe evolves, a condensate of constant to zero. We If they

The picture

that is suggested is

baby universes interacts

with it and forces the cosmological

are then faced with the question of the size of the corresponding turn out to be macroscopic, the real world (The since the background


then this scenario loses its appeal as a toy model for Problem). Naively one might suppose that large values of

Giant Wormhole

tachyon field takes the value To for arbitrarily

4, the baby universes would form with arbitrarily relationship

large necks. In fact the correct

between the scale size and the metric is more subtle than this, since size. It is not clear to us how to

physical size depends on both $ and the fiducial correctly Whilst derive the distribution of wormhole

sizes from T(4). constant relaxes to zero at In par-

the above scenario, where the cosmological

large scales, has some appeal, there are reasons to doubt its consistency. ticular one may ask what two-dimensional minimum. Kutasov

field theory can describe the stationary and Seiberg [35,38] have argued that a

point at the potential

generic two-dimensional cal tachyons.


field theory coupled to gravity that the theory at the bottom degrees of freedom.

contains physiof the potential

This seems to imply

cannot contain the standard gravitational

7. A Non-Vanishing
In the previous


cosmological model

section we considered a two-dimensional

which showed very different

behavior at large scales from what one would naively considerations in mini-superspace. It turns out

expect based upon semi-classical that an apparently can lead to drastic potential

minor modification

of the tachyon potential

near its minimum

changes in the associated cosmology.

In fact there exists a

V(T) such that the linearized fluctuations

about the tachyon background cosmological constant term In other words, with having a

satisfy a Wheeler-Dewitt throughout the non-linear

equation with a conventional evolution

of the tachyon background. dynamics is perfectly consistent

it is possible that the background non-zero cosmological classical q + 00 limit. constant

at .a11scales. This is most easily seen in the semi-

In this case, we can ignore the effect of a non-vanishing linear dilaton background, i.e. q N t.

tachyon field on a target-space

Consider a homogeneous tachyon background.

Its equation of motion is

-a,2T - qdoT = V (T) .


In terms of the conformal

mode $ = :X0

we have


+ 2&T + V (T) = 0 .


In the q + a first-order

00 limit equation

the second-order

term can be dropped and we are left with Fluctuations about this back-

for the tachyon background.


ground will satisfy a linear equation,

28,~ + V(T)7

= 0.

= -2 + 51~4 throughout

The idea is to look for a potential the evolution

V(T) such that V(T)

of T( 4). This is in fact satisfied by

V(T) =


+ fr2

- Ti(l

- &) ln (I - $,


which has the form of an unstable tachyon potential

near T = 0 and a stationary cannot be continued

point at T = To, which is singular (V N 00). The potential past that singularity but, as we shall see, the tachyon

field never rolls beyond in (7.4), which allows a

T = To. It is this singular behavior, due to the logarithm

non-vanishing cosmological

constant at large scales. To see how this works, insert

(7.4) into the equation of motion and take the q + CXIlimit,

284T + 2To,(l - i)

ln(l - $)

= 0.


This is easily solved by writing

(1 - 6)

= es, so that


= s


from which we obtain

T = To(1 - e-i)
The cosmological constant of a two-dimensional

. universe interacting

(7.7) with this backconditions

ground is equal to the parameter

A, which is determined

by the initial

on T. The equation for S is linear so we see that this example provides a realization of the fact that tachyon field redefinition equation of motion. Wheeler-Dewitt can eliminate the non-linear terms in the to the

However, the resulting in equation

field S is no longer proportional



It should be noted once again that evolution

in spite of the apparently equation

complicated obtained from [19, also

of T given by (7.7), th e 1inear Wheeler-Dewitt Liouville

(7.3) is precisely that of mini-superspace 21, 351. It is interesting satisfies a linear equation,

theory in the q -+ 00 limit

in this context to note that the tachyon background

This suggests the alternate definition



for the canonical tachyon field (4.4),

l?(X) = evQ(x)(To - T(X)).

In the large q limit


the following

linear second order equation for 0,


is equivalent equation

to (7.8).


(7.10) d i ff ers from the k = 0 Wheeler-Dewitt of the tachyon. This is

(4.6) by the term due to the bare dimension that the SL(2,C) vacuum of string theory

the equation natural

satisfies [19], and it is state of

to identify

that state with the most symmetric universe [ 19,391. illustrate that having

or Hartle-Hawking

a one-dimensional

The above considerations minimum

a tachyon



does not rule out a two-dimensional




a very special form of potential A deeper understanding of the tachyon eventually the relation

is required, with a mild singularity

at the minimum. the actual form will

of string theory is needed to determine but it is certainly


possible that the dynamics

turn out to be as described in this section. We need to be able to identify between the tachyon field and the Wheeler-Dewitt prescription in the two-dimensional amplitudes for a

given renormalization following

theory. At the end of the

section we set up a renormalization


scheme, in which this identification

is particularly momentum


We are able to obtain

the beta-function

for zero

tachyons to all orders in T and find a potential

precisely of the form

(7.4). For D 5 1 th ere is further evidence to this effect coming from matrix models.
Consider the non-linear the background linear equation

loop equation Z(Z).

(2.5). It can be solved explicitly about this background

to obtain satisfy the



z(Z+1) + 42(1+3) = 2 c



Moore et al. [al] h ave recently shown that in the continuum of the mini-superspace vanishing cosmological Wheeler-Dewitt constant. conformal


z(Z) is a solution

equation discussed in Section 4, with non-

It is unclear what sitting

field theory, if any, corresponds to a tachyon field one. matter By the theory

at rest at T = To, but we suspect it to be a rather trivial of Kutasov and Seiberg [35,38] i t cannot be a standard


coupled to gravity. behavior.of ical constant conformal


the T = To fixed point describes the asymptotic

an expanding

universe long after all relevant scales (e.g. the cosmologdegrees of freedom are

scale) have been passed. The only remaining

matter fields from which the scale of the metric decouples. The situation

is analogous to that in QCD at very large distance scales where the only degrees of freedom are massless pions. one-matrix a matrix matrix model. potential Another closer example is provided constant by the D = 0 corresponds to

In this case a non-zero cosmological slightly off criticality.

The model flows to the trivial


model at large scales and the random surface interpretation

breaks down. in the at

If the picture classical target

presented in this section is correct then there is nothing space dynamics which favors vanishing topologies cosmological

constant this.

large scales. Higher worldsheet that the string coupling rapidly

are not likely to influence increasing


decreases with

scale in the D > 25 at small scales and its for the target space

theory. Target space quantization

is therefore only important of initial conditions

main effect will be to provide a distribution


fields, which subsequently is dynamically initial insignificant

evolve classically.

Since a small cosmological


at small scales it appears unlikely peaked at X = 0.

that it can force the \


to be infinitely

We conclude potential. probably

this section by mentioning

some consequences of a bottomless increase, and the theory is

In this case, T(X )

wr continue to rapidly 11

too sick to describe cosmology.

To see why, we note that the increasing flows

tachyon field will act as a source for all other fields, and the renormalization

of the higher order couplings, which we discussed in the Section 5, will be disturbed even at late times. This would even be true in the semi-classical limit of large q.

8. The Tachyon
In this section, we present a calculation .to the tachyon beta-function. linear terms, since individual

of the leading non-linear contribution

One might initially

suspect that there are no non-

diagrams in the loop expansion for the product of two On the other hand, we are interested since we want the theory to be exactly

normal ordered tachyon vertices are finite. in the exact renormalization independent of couplings,

of the fiducial metric.

This means that we must keep track of the nonbeta-function will

divergent cutoff dependence. When this is done, a non-vanishing indeed emerge. We shall not rigorously method originally in the strength carry out this procedure,

but will instead adopt the

used in [40] f or o p en string theory. This approach is perturbative For

of the tachyon field, but sums all orders of the loop expansion. momenta

a certain range of target-space

k;, the sum over loops introduces


distance divergences from which the beta-function the results to the region of interest analytic in the k;. reference [40] our starting

can be read off. We then extend

by assuming that the target space theory is


point is a general renormalization



equation for a set of couplings gi:

Piz =A;$ aik +j,Wg.igkgl +C Sjgkyjk, +.... ik C

Here t is the renormalization order by order in the couplings. t = 0. Then the renormalized coupling is given by


scale and A; is the anomalous dimension of the oper-

ator which carries the coupling gi. The flow equation (8.1) can easily be integrated Let g (O) be th e value of g at some infrared scale,


= eXitgi(0)

+ C j,k


$k x ,+xk-x, 3

g (0)gk(O)

+ * * .


For the time being, we are primarily will assume a flat target-space


in tachyon


so we

with vanishing

dilaton field.

In due course we will

consider the effect of a linear dilaton then be written


The tachyon term in (3.4) can

The target-space an explicit mensionless.


kp plays the role of the index i and we have included

E to make the tachyon couplings

dependence on the cutoff

T(k) di-

We will calculate the two-dimensional techniques.

effective action using standard around some classical = X,(g) +rp(a). The

background-field background, path integral

The fields Xh are expanded

which varies slowly on the cutoff scale: Xp(a) is performed over the quantum field up,

dXbl Z[Xb] = e-S

D, eXp -; x exp(-&


d2a fi
dza fi/

dD+lk T(k) ,ik.Xb ,ik.if) . (8*4)

The effective


is given by S,.f[Xb]

= -log


The two-dimensional




= -2?y

log (101-021).

(8.5) in powers of T(k), and vertices. Because

The path integral is evaluated by expanding the exponential evaluating

diagrams with ever increasing numbers of interaction form of the operator multiplied

of the exponential

by T(k), a given vertex can have set of graphs to obtain the

any number of legs, and we have to sum over an infinite

full answer at a given order in T(k) ( see Figure 4). There are divergences which do not show up at any finite order in the standard two-dimensional loop expansion,

but only appear when the diagrams are added up. To regulate these divergences, we need a well-defined cutoff procedure. In general, the exact beta-function will

depend on the choice of regulator. redefinition regulator ambiguities

This is believed to correspond

to the field is that a with

in the target space equations.

Our assumption

scheme can be found in which the target space fields are identified amplitudes. For reasons of simplicity and tractability,


we will use at finite

a hard sphere regulator separation. couplings

defined by cutting

off position

space integrals

It is not clear that this.prescription as Wheeler-Dewitt is outlined amplitudes.

allows us to view the renormalized approach that should satisfy calculations of zero-


this criterion are awkward momentum

at the end of this section. In general, explicit

in this scheme, but we able to compute the beta-function re tachyons to all orders, and that allows us to determine

the full form of

the tachyon potential. The first-order the left in Figure 4.

dDi-1 k T(k) J d&l I; ckz-2 T(k) ,ik.Xb 7 ,ik.xt, (8.6)


to the effective action comes from the graphs on



+ &


=sO[xb] + ;


where c is the hard-sphere diameter. renormalization group flow (8.2).

We want to compare this with the integrated to identify c with ewt, so that the

It is natural

cutoff is removed as t + co. The renormalized

coupling in (8.6) is simply

Tt(k) = tk2-2To(k) = e(2-k2)TO(k)

for a tachyon of

and we read off the familiar momentum



Xk = 2-k2


At second order we have to sum over the graphs on the right in Figure 4. Their contribution to z[Xb] is given by

d2al d2a2

dD+lkl dD+lk2 T(kl) T(k2)

x eikl.Xb(~l)+ikz.Xb(az) eikl.r(m)eikz.*(az) .


We are assuming that the background we can expand

field Xb varies slowly on the cutoff scale, so

in (8.8). The sub-leading of dimension malization

terms, involving


of X:,


to terms

higher than zero in the effective action, and do not affect the renorThe dimension zero piece of the effective action

of tachyon couplings.

coming from (8.8) is therefore



dDslkl dDS1k2 ,(k1+k2)2-2T(k1)T(k2)


00 2kl.k2+1 ei(kl+k2)Xb(u) dYY J

in the momenta.

(8.10) *

The y integral

is convergent for ICI . k2 < -1 and one can define its value outside continuation Using this prescription we

that region by analytic

find that the renormalized

coupling to second order is

Tt(k) = e(2-k2)t [To(lc) + $ /

dDS1kl dD+1k2~~~~~2.~~)~o(k~)~o(k2)]

. (8.11)

This is to be compared


the general solution


The denominator 2 + 21cr . k2 = This can be

contains precisely the correct combination

of anomalous dimensions, term.

Xkl + Xk, - Xkl+kz, but we appear to be missing the ecxkl +&It explained as follows.

The integral in (8.10) is convergent for ICI . k2 + 1 < 0, which as Xkl+kz > Xkl + Xkz. As the cutoff is removed, t -P co, and

can also be written exkltkz dominates

over e cXkl+Xk~It. Our expression for the renormalized


(8.11) therefore only contains the leading divergence. However, we still have enough information to read off the value of the second-order coefficient in (8.1), giving integrals

41k2 = -$Y(D+l)(kl

+ k2 - k ) . Its simple form makes the momentum

in position space

and we obtain the tachyon beta-function

@T(X) = (2 +.v2) T(X)

- $ r(X)2

+ *** .


Setting this equal to zero gives an equation of motion for the tachyon background. Near the mass-shell our results essentially agree with those of Das and Sathiapalan [41]* and those of Brustein The above calculation treatment et al. [42]. only considered tachyons, whereas a more comprehensive couplings such as gravitons and dilatons. but for our present purposes it Cp= -%X0,

would include higher-dimension

Such a general approach would be quite complicated is sufficient

to consider the simple case of a linear dilaton background, The tachyon beta-function

in flat target-space.

is obtained in much the same way as action looks particularly simple if

before. The dilaton term in the two-dimensional we write the fiducial metric as a conformal

factor times the flat two-dimensional

+ The non-linear term in the tachyon equation of motion obtained in [41] has the opposite sign to ours, but the important feature, that this term has no derivatives, agrees. 31


+aa = et&,*. Then 1 Sail = G d2m[q-X, in two dimensions (8.13) and qp =

where I denotes the flat

space scalar Laplacian

tq, 0, * - * ,O). This term is linear in Xfi and its effect in the path integral is taken into account by including graphs with external legs which carry a factor of qp (see
Figure 5). The contribution of each such external leg is given by

where G(a-~0)

d2a [(a) ik . qo,G(a--ao)


= ; k . q[(ao) ,

(8.14) of

is the two-dimensional

(8.5) and as is the position

the tachyon vertex from which the leg emanates. To leading order in T(k) the effective action is given by the graphs on the left in Figure 5,
f-2 &j&G]= So[&,] J d2c dD+lk + & .J
,+ J((C) T(k) ,ik.xb .


The anomalous dimension of the renormalized properties with as the fiducial conformal metric, factor

coupling is determined [ is varied. The cutoff

by its scaling c is defined

reference to the fiducial dimension

so it scales as E N eg, and we find the The condition that this vanish is the

anomalous linearized

XI, = 2-ikaq-k2.

tachyon equation

The second-order contribution The calculation the renormalized

comes from the graphs on the right in Figure 5. theory and one finds that

involves the same steps as in the critical coupling to second order is now

*t(k) = e(2-ik q-k2)t [To(k) + i /d+ kl




the only effect of the linear dilaton background

is to modify the anomain (8.16) and

lous dimensions.

The q-dependence cancels out of the denominator


we read off the same value of the second-order beta-function


as before.

Thus we find that the leading terms in the tachyon beta-function string theory are

in non-critical


= (2 - 2vfD.v + v2) T(X)

- ; T(x)2

+ *- * )

in agreement with (3.7). Note that the non-linear vanishing term has no derivatives potential and corresponds to a noncon-

cubic term in the target-space


This has important

sequences. In particular

it implies that a uniform tachyon field will have non-trivial group flow of tachyon fluctuations since, according a c-number, with non-zero

influence on the renormalization target-space momentum.

This is surprising

to (3.4), a constant

tachyon background dimensional action.

field only contributes

& Jd2a Jr, to the two-

This is perhaps too simple a view to take. In fact, the excluded integrals introduces function non-trivial effects even

-volume of the hard sphere regularized when T is constant.

In this case the partition

T f = -m,

reduces to that of a simple

hard sphere gas, with fugacity

O f Z=C;Er J d2al.. n=o lCTi-ajl>C

f2 = 1 + f $ #l

. d2gn

- ac2) + . . . )

where the integrals are over unit area and the thermodynamic the limit E+ 0. The free energy is F = c2 log 2. The running the partition function


is obtained in T

of the coupling

with E is defined by requiring scale. This implies

to be independent

of the cutoff

0 = $F(T) (8.19) = -$F(T) + $F/(T)$f .


We define the beta-function

in the usual way,

P(T) = $

= 2%.


To second order in T we find

,6(T) = 2T - $T2,

in agreement with (8.17). W e expect this agreement between analytic and the hard sphere gas to hold to all orders in T. Another point of interest is that statistical


systems of this type typically


Lee-Yang edge type singularities functions

at negative fugacity

and therefore vanishing beta-

at some positive values of T. However, the behavior of the beta-function at these points, which suggests that cusps in V(T) of the

is usually not analytic type encountered

in Section 7 can occur. outline a regularization into the renormalcoordinate space by the

To conclude this section we would like to schematically scheme in which the Wheeler-Dewitt ization with process. Begin by introducing amplitudes a lattice

enter directly on the fiducial


spacing E. In each cell we define an amplitude over the two-dimensional fields in the interior

on the boundary of that cell, fixing


values on the boundary. edges. The remaining full path integral.

This defines an effective theory that lives on the lattice over the boundary values of the fields yields the


The integrand

of the effective theory is given by the product Schematically,

over all cells of the cell amplitudes.

The amplitudes target-space



4 are by construction



To introduce

fields we can expand $($boundary) in terms of string modes. Let 2 be


the zero-mode part of X on the boundary.


1c,= (1 -T(k)+



where Go is the free theory amplitude. be independent

By requiring the long wavelength behavior to for the target space fields calculations tachyon

of the cutoff we can define beta-functions


This is certainly

not a convenient scheme for beta-function

in the presence of general couplings,

but in the special case of a constant function is simply

field, we can obtain the full answer. Then the partition


Jn (1- obO(h)


= (l-


= 0),
to $. Requiring 2 to be inde-

where the total number of cells N is proportional pendent of E gives the beta-function

/3(T) = -2(l

- T) ln(1 - T) .



precisely agrees with the zero-momentum


obtained from the

singular potential

(7.4) (with To = 1) in Section 7.

9. Conclusion
We conclude gravity. with some observations and speculations about 4-dimensional gauge unless

Unfortunately, our attention

in this case we can not choose the conformal to conformally flat geometries.

we restrict

Let us therefore consider Weyl curvature. Letting In

such a restricted

path integral over geometries with vanishing action to a simple form.


this case we can reduce the Einstein

g,, = 42qPV

we find that the action has the form

This is just ordinary ory, thus requiring case, target-space

d*a: { -(3Fd)2

+ Ad4 + matter}

term. The original field the-

$4 theory, but with

a negative kinetic

covariance requires the path integral to result in a conformally the vanishing of all beta-functions. by allowing


As in the two-dimensional A and all other couplings

fields can be introduced

to depend on 4. The vanishing in target-space. Of particular

of beta-functions interest

defines the equations of motion

is the Ad4 term which would be replaced to conformally flat

by the more general expression T( 4). R emoving the restriction geometries makes things more complicated, regulating,

but still we expect that the general target space

logic of gauge-fixing,

and restoring covariance by requiring

field equations to be satisfied will make sense. Can one expect the non-linearities roe in four dimensions of the target-space theory to play a similar be

as in two ? More specifically,

can the non-linearities

attempt relevant at large scale factor ? To schematic non-linear Wheeler-Dewitt

to answer this we can consider a very


where H,, geometry,

is a Wheeler-Dewitt and


!P is a functional

of matter


is some coupling strength

which depends on the scale size of the

geometry. In the two-dimensional case, f is replaced by the coupling strength ecp = -9x0 =a-q e2 . Because of tachyonic instabilities, the Wheeler-Dewitt amplitude in (9.2) increases like ez 9xo. Thus the background to overcome the decrease of the coupling strength. In four dimensions, timated the coupling for universes splitting considerations. and joining can be esfield increases at a sufficient rate

from dimensional

Consider a geometry grounds,

with three bound-

aries, each with

scale size N a. On dimensional


the action for such a


is N iVia2

where IMP is the Planck mass.

Thus the amplitude

for a

universe splitting depending

is likely to be of order eVcM~ where c is a numerical features of the geometry.


on dimensionless

The rapid decrease of the quenched with consider the the

coupling strength scale size.

might suggest that non-linear this is not necessarily equation

effects are quickly


the case.

To see this,

linear part of the Wheeler-Dewitt mini-superspace Wheeler-Dewitt


= 0. In four dimensions,

equation has the form

For large a the solution Dewitt strength amplitude is 8 N eza2.

Hence it is possible that

the Wheeler-

grows sufficiently


to overcome the decrease in coupling

in four dimensions

also. On the other hand, we have seen in Section 7 the usual conclusions of the

that such non-linearities .conventional

do not necessarily invalidate cosmology. of such, non-linearities


What the implications matter of speculation.

are for large scale cosmology is a

It should be stressed, however, that they are not connected over space-time topologies. Such sums imply quantiza-

with the issue of summing tion of the target-space level. Third is large. quantization

theory, while our considerations is important

have been at the classical

at small scales where the coupling strength at small distances would provide

It seems likely that third


an ensemble of classical trajectories fraction constant inevitably

at large scale and that only an infinitesimal

of these would remain at the unstable point T = 0 where the cosmological is naively zero and the non-linearities region. W. Fisare small. The remainder would

flow into the non-linear

Acknowledgements: chler, D. Kutasov,

We would like to thank T. Banks, S. Ben-Menahem, J. Polchinski,

S. Shenker and N. Seiberg for useful discussions.


1. B. S. Dewitt, Phys. Rev. 160 (1967), 1113; and J. A. Wheeler

J. A. Wheeler, in Batelle Recontres, eds. C. M. Dewitt (W. A. Benjamin, 2. S. W. Hawking, S. W. Hawking, inc., New York, 1968).

D. N. Page and C. N. Pope, Nucl. Phys. B170

(1980), 283;

Comm. Math. Phys. 87 (1982), 395.

3. K. Kuchar, J. Math. Phys. 22 (1981), 2640; also in Quantum Gravity, Vol. 2, C. J. Isham, R. Penrose and D. W. Sciama eds., (Clarendon 1981).

4. A. Jevicki, in Frontiers in Particle Physics Dj. Sijacki et uZ.eds., (World $3, Scientific 1984).

5. A. Strominger,

Phys. Rev. Lett. 52 (1984), 1733; also in Proceedings of the

1988 Theoretical Advanced Study Institute, A. Jevicki and C. I. Tan eds.,

(World Scientific 1989). (1984), 349.

6. D. Gross, Nucl. Phys. B236 7. N. Caderni and M. Martellini,

J. Theor. Phys. 23 (1984), 23. H. J. de Vega and

8. I. Moss, in Field Theory, Quantum Gravity and Strings,

N. Sanchez eds., (Springer 1987). 9. V. A. Rubakov, Phys. Lett. B214 G. V. Lavrelashvili, (1988), 503;

V. A. Rubakov and P. G. Tinyakov, (1988), 757.

JETP Lett. 46

(1987) 167, Nucl. Phys. B299 10. M. McGuigan,

Phys. Rev. D38 (1988), 3031, Phys. Rev. D39 (1989), 2229.
(1989), 334, Nucl. Phys. B321 (1989), 48.

11. A. Anderson, Phys. Lett. B212 12. A. Hosoya and M. Morikawa, 13. S. Coleman, Nucl. Phys. B307 14. T. Banks, Nucl. Phys. B309 15. E. Baum, Phys. Lett. 133B

Phys. Rev. D39 (1989), 1123.

(1988), 867., Nucl. Phys. B310 (1988), 643.

(1988), 493. (1983), 185.


16. S. W. Hawking, 17. S. Giddings 18. C. Teitelboim, 19. J. Polchinski,

Phys. Lett. 134B (1984), 403.

Nucl. Phys. B307 (1983), 41. (1989), 123. (1988), 854.

and A. Strominger,

Phys. Lett. B126

Nucl. Phys. B324

20. T. Banks, Physic&u Birthday 21. G. Moore,

Magazine, vol 12, Special Issue in Honor of the 60th


of R. Brout,(Gent

N. Seiberg and M. Staudacher,

From Loops to States in 2-D


Quuntnm Gravity, Rutgers preprint,

22. E. Brezin, (1978) 35 C. Itzykson,


G. Parisi and J.B. Zuber,

Comm. Math. Phys. 59

23. S. R. Wadia, Phys. Rev. D24 24. A. M. Polyakov,

(1981), 970.

Phys. Lett. 163B (1981), 207.

25. S. R. Das, S. Naik and S. R. Wadia, Mod. Phys. Lett. A4 (1989), 1033; S. R. Das, A. Dhar and S. R. Wadia, Mod. Phys. Lett. A5 (1990), 799. 26. T. Banks and J. Lykken, 27. A. A. Tseytlin,

Nucl. Phys. B331

(1990), 173.

Jour. Mod. Phys. A5 (1990), 1833. and M. J. Perry, Nucl. Phys. B262

28. C. G. Callan, D. Friedan, E. J. Martinet (1985), 593; E. S. Fradkin and A. A. Tseytlin,

Nucl. Phys. B271

(1986), 561;

C. G. Callan and Z. Gan, Nucl. Phys. B272 29. A. Chodos and C. B. Thorn, T. L. Curtright Nucl. Phys. B72

(1986), 647. (1974), 509;

and C. B. Thorn,

Phys. Rev. Lett. 48 (1982), 1309;

(1988), 371;

R. C. Myers, Phys. Lett. 199B I. Antoniades, (1988), 393; S. P. deAlwis, 449. J. Polchinski

C. Bachas, J. Ellis and D. Nanopoulos,

Phys. Lett. 211B

and R. Schimmrigk,

Phys. Lett. 218B



30. F. David, Mod. Phys. Lett. A3 (1988), 1651. 31. J. Distler and H. Kawai, Nucl. Phys. B321 Nucl. Phys. B346 (1989), 509.

32. J. Polchinski, 33. T. Banks

(1990), 253. RU-

The Tuchyon Potential in String Theory, Rutgers preprint,

91-08, 1991. 34. A.A. Tseytlin

On The Tuchyonic Terms in the String Efective Action, Johns

JHU-TIPAC-91004, 1991.

Hopkins preprint,

35. N. Seiberg, Notes on Quantum Liouville Rutgers preprint, RU-90-29, 1990. (1985), 332.

Theory and Quantum Gravity,

36. T. Banks, Nucl. Phys. B249 37. A.B. Zamdlodchikov

and A.B. Zamolodchikov,

Annal.Phys. 120 (253), 1979.

38. D. Kutasov and N. Seiberg, Number of Degrees of Freedom, Density of States

and Tuchyons in String Theory and CFT, Rutgers preprint,

39. D. Birmingham 40. I. Klebanov and C.G. Torre,, Phys.Lett.

RU-90-60, 1990.

B 194 (1987), 49 (1988), 446.

and L. Susskind, Phys. Lett. 200B

41. S. R. Das and B. Sathiapalan, 42. R. Brustein, (1988), 224. D. Nemeschansky

Phys. Rev. Lett. 56 (1986), 2664.

and S. Yankielowicz, Nucl.

Phys. B301


1) Once a fiducial

the short

metric has been chosen, and a cutoff introduced,

distance fluctuations

will still include large physical geometries, which will

affect the renormalization. 2) A Lorentzian geometry with singular points corresponding to the splitting

of universes, and the emission and absorption background. 3) A tachyon potential

of baby universes from the

with a smooth local minimum

at T = To.

4) Leading order graphs which contribute critical string theory.

to the tachyon effective action in

5) Leading or der graphs in the calculation linear dilaton background.

of the tachyon beta function

in a


/ / / /

/ /

/ / / //

/ /

/ / /

/A/ /



1-91 6607Al

Fig. 1



Fig. 2

1-91 6607A5

Fig. 3

. .




l-91 6807A3

Fig. 5