SLAC-PUB-4991 May 1989 T/AS

THE GOOGOLPLEXUS*

IGOR ICLEBANOV
Stanford Linear Accelerator Stanford University, Stanford, and
-

Center 94309

California

LEONARD SUWUND~
Physics Department Stanford University, Stanford, California 94305

ABSTRACT
We carry out third ity. The scale factor quantization in a minisuperspace model of quantum Wormholes gravturn

of the 3-geometry

plays the role of time.

the cosmological probability for-the

constant

into a new kind of dynamical constant is typically

variable.

The a priori

cosmological

a smooth distribution.

Invited

talk presented at the Tenth and Final Workshop April 20-22, 1989, Chapel Hill, NC. 89: Superstring TX Workshop,

on Grand Unification, Also invited

talk presented at Strings

March 13-18, 1989, College Station,

* Work supported by the Department of Energy, contract DE-AC03-76SF00515. t Work supported by NSF PHY 812280

1. Introduction
This talk is based on joint work with W. Fischler and J. Po1chinski.l Recent work on quantum gravity 2-5 h as d emonstrated creation and annihilation of quantum of microscopic fairly convincingly that

baby universes leads to no observable loss fluctuations is to

coherence.6 Instead,

the effect of these topological

turn the coupling parameters relevant to physics in a macroscopic peculiar further kind of dynamical development, variables governed by a probability

universe into a In a

distribution.

Coleman4 found that the probability

to find the cosmologpath The

ical constant

s equal to zero is 1. Coleman of gravity

theory relies on the Euclidean

integral quantization unboundedness

which, at present, is only a formal technique.

of the action

(14
makes it difficult particular, to interpret the Euclidean path integral of quantum of the cosmological gravity. In Coleman s mechanism for the vanishing with constant arbitrarily to the action

relies heavily

on the apparent

instability

respect to nucleating -2/3X

large numbers of the Euclidean four-spheres each contributing (A = 16G2A/9). U n f or t unately, the same instability or even cosmically

seems to lead to a catastrophic (Res-

number of macroscopically

large wormholes in spacetime:

olutions of this problem have been proposed in Ref. 8, but were strongly in Ref. 9.) On the other hand, there are mathematical instabilities. prescriptions

criticized

which eliminate

the

Consider, for example, the path integral over conformally

flat geomeof

tries of spherical topology: wormhole-connected gravity reduces to

gij = d2S;j. This set includes Coleman networks s

spherical universes. The Euclidean

path integral for Einstein

/b#dexp(/d4x (&@V) - A44))

Clearly, this expression is formal due to the unconventional

(l-2)

sign of the kinetic term

for 4. With the Gibb ons-Hawking-Perry to be

conformal rotation lo 4 + i4, it is defined

/kGlexp (--Id42 (&@W2+W4))

(1.3)

This is just the Euclidean path integral for the stable d4 theory (we assume that the cosmological constant is positive).* Such a procedure may define a consistent 0 which

quantum theory of gravity, but it surely eliminates the divergences as X j drive Coleman mechanism. s Other prescriptions proximate
-

rotate the contour about the Euclidean saddle point or apfour-spheres. In this iDm2 in
11

saddle points associated with wormhole-connected

case a careful analysis of the modes of fluctuation front of the Baum-Coleman-Hawking each four-sphere. thp marvelous amplitude

reveals a prefactor

exp(2/3X) 12J3 associated with 4

The result is, once again, not favorable: in 4 dimensions (D = 4) b ecomes a disappointing is so ill-defined exp( - exp( 2/3X)). Evi-

exp(exp(2/3X))

dently, the Euclidean

path integral

that it can be imaginatively

used to produce vastly different answers. For these reasons it seems necessary to provide a different formulation of the theory of topological fluctuations. change in a

In this paper we present a Hilbert minisuperspace model of quantum

space analysis of topology

gravity.

This model is quite simple to work

with, but is rich enough to include any number of universes with spherical spatial geometry. We will find that, under some rather general assumptions, the average

number of large universes is O(exp(2/3X)). lo- ,

In view of the known bound X 5

the average number of universes is 2 101olzo. This is the origin of the We will show that all these universes are cold, empty and

name googolplexust uninteresting.

Their presence, much like the presence of the infrared photons in In fact, in contrast

QED, makes no effect on any of the observable probabilities.

A- This t This finite New

is not a conventional theory since it must be regulated in a conformally invariant way. term, coined by L. Susskind, is a fusion of two words. The googolplex is the largest integer with a special name. It is equal to 101oloo. Plexus means a network. (Websters Collegiate Dictionary.) 3

with Coleman result, we will find that the a priori probability s constant is a smooth function, or Coleman type. The Coleman double exponential Furthermore, a transition

for the cosmological

with no enhancement of either the Baum-Hawking

does occur, but in the form exp (-,2/3x). of a probability for X but rather i. e.,

it does not have the interpretation amplitude

from the state of the googolplexus to the out-vacuum,

it is the amplitude

to create no universes. The exp(2/3X) in electrodynamics

universes are rather like is the

the soft photons emitted

) and the factor exp (-e2i3

analogue of the soft photon factor which suppresses transitions with a finite number of soft photons.
-

to exclusive states

The

same assumptions

on the state

of the googolplexus

that

produce

O(exp(2/3X))

cold empty universes, typically

lead to no enhancement in the num-

ber of warm, liveable universes as X +

0. We will show that there is a class of warm ones. It is not yet

states which contain few cold universes but O(exp(2/3X)) clear whether a theory based on such states is sensible.

2. Third Quantization in Minisuperspace

In this section we consider the path integral in a minisuperspace model of quantum gravity. This model includes the spatially spherical geometries with metric

ds2 = g(-dT2

+ a2(T)df$)

(24

where da: spatially

is the metric matter

of a unit

three-sphere.

We also include a number of action with matter

constant

fields &(r).

The E ins t ein-Hilbert

couplings reduces to T -ait2 + a + a3

0

(2.2)

where X = 16G2A/9.

Th is action defines a hamiltonian 9

H = i which is the generator

-f

- a + a3(X + V(q$;)) + am3rf in the parameter must annihilate

P-3)

of translations

time 7. Due to general the physical states. This of the universe @(a, 4;)

covariance of the theory, the hamiltonian defines the Wheeler-De d a-P-$da Witt

equation for the wave function

82 d _ .-2 j@ - a2 + a4(X + V(h)) Lhx 8

@(a, q$) = 0 1

(2.4)

where the uncertainty in the hamiltonian Schroedinger tion.

15

in p is a part of the operator ordering ambiguities of Eq. (2.3). Sometimes Eq.

inherent

(2.4) is regarded as gravity s Klein-Gordon equa-

equation,

but it is obviously beginning

more like gravity s

Indeed many authors,

with De Wittt4

have noted the similarity is predefinite

between the scale factor a and time. In l+l cise. As in the case of the Klein-Gordon

dimensions the correspondence equation, the lack of a positive

probability

density makes the one-universe theory hard to interpret.

In any case, in prochanges.

we are not after formulating

such a theory here. Instead, we are interested connected components of geometry

cesses where the number of spatially A convenient formalism

involves a Hilbert quantization

space for universes. With Witt

this in mind,
16-21

we will carry out third

of the Wheeler-De

equation,

which We will with con-

is a step analogous to second quantization therefore a playing consider a quantum

of the Klein-Gordon

equation. Witt First

field theory of the Wheeler-De field.

equation we will

the role of time and @ being the quantum topology

sider a theory without wormholes.

changing processes. Subsequently,

we will include

One of the goals is to build a theory whose Feynman diagrams resemwormhole-connected interpretation bubbles. We will find that this theory has a from the interpretation

ble Coleman s natural

probability

which is quite different although

suggested by Coleman. similar to Coleman s,

Surprisingly,

this theory has a graph structure we measure.

the graphs do not compute the probabilities 5

To simplify

the discussion, we will first ignore the scalar fields 4; and return to

consider their effects in Section 4. Also, we will adopt the operator ordering in the Wheeler-De think Witt Eq. (2.4) w h ic h corresponds to p = -2. of the Wheeler-De Witt Then it is convenient to wave function:

of u3 = V as the argument

(
Although our results of this prescription for large V. is that If we think

d2 -- & W2
the potential

+; >

qv>

= 0

P-5)

will not depend much on the choice of p, the advantage on the left side approaches as a real coordinate, oscillator a constant then Eq.

of V as time and Q(V)

(2.5) is just th e c 1assical equation for a harmonic which varies with becomes an ordinary eqlained time. In fact, the oscillator

with a spring constant

is upside down for V < X3i2 but as V + co. As Klein-Gordon Witt

harmonic oscillator

with frequency wg = d/3

above, it is natural to think of Q(V) as a O+l dimensional terms, third quantization to solution of the minisuperspace harmonic

field. In practical equation amounts

Wheeler-De oscillator

of the time-dependent

problem

defined by Eq. (2.5). T o make this analogy yet more explicit, and replace Q(V) with X(t). The third-quantized lagrangian

we change notation is

L = ; A2 - ; w2(t)X2

- Jxqt)

(2.6)

For reasons that will be explained

t = 0. The time-evolution

later, we have included

a S-function

source at

is generated by the hamiltonian

H = ; P2 + ; w2(t)X2

+ JXS(t)

P-7)
variable: it creates or an-

We emphasize that X(t) nihilates space-the

is not an ordinary

oscillator

entire universes with spatial volume t. t plays the role of time in a new googolplexus -where particles 6 are universes. In order to introduce

universe creation and annihilation

operators,

we expand

X(t) = f(t)a + f*(t)u+

where f and f* are the incoming equation. The conjugate and outgoing is solutions of the Wheeler-De

(2.8)
Witt

momentum

P(t) = f(t)u + f*(t)u+

From the canonical skian ff* This fixes the normalization f(t of f: ---f m) = (2wo)-ue-w+) - f*f = -i commutation relations

(2.9)

it follows that [a, CL+] 1 if the Wron=

(2.10)

(2.11)

We will also fix the phase 6 by requiring

that

h(O) = 0
where f(t) = fl(t) + ifz(t). A nice property of the hamiltonian

(2.12)

in Eq. (2.7) is that

it loses its explicit

time dependence as t + 00: (2.13)

H(t -

co) = ; wo{u, u+}

The ground state at late times is simply the oscillator

vacuum (2.14)

a 1 out >= 0

The quanta annihilate

of the theory

can now be cleanly

identified:

ut and a create and

the out-universes. 7

The key question is what determines of the googolplexus . The minisuperspace negative time is physically

Q(X, t), the Schroedinger formulation has a peculiar

wave function feature that

meaningless (t refers to the volume of S-space). Therenear t = 0 by a smooth match

fore, it is sensible to assume that \-l is determined on to short distance physics. insensitive

Since the Planck scale physics is almost completely implies that the boundary condition

to the value of X, this assumption

at t = 0 has no strong dependence on X. We will consider a few such choices of the state of the googolplexus. motivated by the assumption One of them, which we denote by 1 i >, will be of symmetry under t + -t. This suggests that

i(t

= O)l i >= qt

= O)l i >= 0

(2.15)

or, in other words, Q[r;(X,O) = const A general wave function of width consistent with our assumption is initially (2.16) a wave packet we will work

w, which has no singular dependence on X. For illustration,

with Gaussians K&,(X, t = 0) = (---g4exP We will see that some of our results are insensitive boundary conditions imposed

(-5)

(2.17)

to the precise nature of the conditions our

at small t, as long as these boundary

have no fine-tuned theory with

dependence on X. However, for the purposes of comparing it will

Coleman s,

be necessary to fine tune the physics near t = 0. of this fine-tuning in terms of imposing generic of this

Alternatively, boundary

one can think

conditions

at large t. We will discuss the possible implications procedure in Section 4.

rather counter-intuitive

In order to carry out comparison the dependence of the path integral

with Coleman results, we need to calculate s on J and X for an arbitrary

8

state of the

googolplexus

1 in >. If we set J = 0, then the path integral over all X(t) amplitude between the in-state and the out-vacuum

for t > 0

is just the transition

J
This path integral which represents torus. sical grounds Coleman dependence on l/X. topology. -

[DX(t)]2

Lw L(J=o)dt =< out 1 in >

(2.18)
of the vacuum graph of a

can be thought

of as the exponential geometries

a sum over minisuperspace

with

the topology

There are no toroidal

solutions of Rij = 8nGAgij.

Therefore,

on semiclas-

did not expect the sum over tori to have an exponential

In Coleman approach, the crucial geometries have spherical s spheres into our model, it is necessary to turn the path integral

In order to introduce Treating

on the source J. is exactly

the source term as a perturbation,

given by the sum of graphs shown in Fig.

1. Each line represents a

n.+
J J

n v

- n+
J

... w z
6335Al

tb) A4-99

Figure 1. a) The sum of Feynman diagrams representing the exponential of Eq. (22). The vertical axis is the scale factor a, the horizontal is parameter time r. Each line represents G(O,O), the sum over all paths from a = 0 back to a = 0. b) G(O,O) is a sum over minisuperspace geometries with the topology of a sphere.

universe which is created by the source with zero radius, propagates, to zero size and is annihilated

shrinks back

by the source. In other words, each line stands for In analogy with Coleman sum over s

a sum over geometries of spherical topology.

9

Euclidean bubbles, the sum of graphs in Fig. 1 exponentiates. that 2 = J where

(-,(tl , t2) = < Out

To see this, we note

[Dx(t)]e&m

L(J)dt = e-t J2GW) < out

in >

(2.19)

1T(X(t1)X(t2))1

in

< out I in >

(2.20)

If we specialize to I in >= ) i >, then

G(tl < ta) = ?fi(tl>

-

f(h)

(2.21)

As expected, G(tl, t2) is an outgoing wave as a function and also satisfies -&G(tl = 0, t2) = 0

of the bigger argument t2

(2.22) associated with each line

Let us examine in more detail the factor -3 J2G(0,0) in Fig. 1. Naively

it appears that for agreement with Coleman analysis G(O,O) s It is easy to see that this is not generally the case in

must be our model.

exp(2/3X). First,

the real part of f(t)

satisfies jr(O)

= 0. Since the boundary

condition is specified at t = 0, the generic behavior of fr is to increase exponentially with

t until

it enters the oscillatory

region t > .Xm3i2, where it must oscillate techniques indicate that

fz(t),

with the amplitude fl(0) - exp(-1/3X).

fixed by Eq. (2.11). Standard WKB On th e contrary, the imaginary in the oscillatory

part off,

which is out

of phase with fl(t) the WKB

region, exponentially As a result,

grows toward t = 0. In

approximation,

f2(0) N exp(l/3X).

G(O,O) N A + iB

(2.23)

where A is O(exp(-2/3X))

and B is O(1).

It seems that there is an immediate

disagreement with Coleman analysis. However, as explained in Ref. 22, there is s

10

no disagreement.

In the first-quantized

theory, specified by Eq. (2.2)) G(O,O) is

the path integral over all trajectories

o(r) with a(O) = a(T) = 0 integrated over 7 . by sending r + ir, then the resulting

If one performs the Euclidean continuation Euclidean action

T

+a Xu3) JdT(ck2
0

(2.24) how

is not bounded from below. Within would one approximate

this ill-defined Euclidean path integration,

G(0, O)? In the spirit of Coleman theory, we need to s paths, including those with multiple bounces

sum over all the Euclidean stationary

off the point a = 0. (The bounces occur because the problem is defined on the half-line a > 0.) Therefore, we include all the trajectories
-

of the form (2.25) these are simply linear

u(7) = $1 wit-h duration T = nr/fi (see Fig. 2).

sin(fir)I Geometrically,

A
a 0 4-09 6335A2 + + +*a* z )

Figure 2. The euclidean trajectories of the form a(~) = Isin(&r)l/fi, with 0 < 7 < na/fi, that need to be included in the semiclassical approximation for G(0, 0). The reflections off the point a = 0 are the minisuperspace wormholes that attach to the north and south poles of the large four-spheres.

chains of n four-spheres glued at their poles. Il.= -2/3X. Th e saddle point approximation
+ pe4/3x + p2e6 3A + . . . =

The action of each four-sphere is to G(O,O) is then

,2/3x M -1. + ace-2/3A

W,

0)

M e2/3x

1 - pe2f3X

P-26)

11

where we have, of course, used an analytically

continued definition

of the geometric

sum. The relative factor of p in the subsequent bounces depends on the boundary conditions at a = 0. Semiclassically, action which typically p N eVsw where S, is the Euclidean wormhole

has only weak dependence on A. The agreement between Eq.

(2.26) and Eq. (2.23) is quite detailed. The formal sum over bounces of Eq. (2.26) is capable of reproducing in Eq. (2.23). even the exponentially suppressed term O(exp(-2/3X))

I nc 1usion of the multiple

bounces is therefore crucial to obtaining G(0, t) from the Euclidean saddle the minisuperspace model

the correct normalization point analysis.

of the Green function

It comes as a bit of a surprise that

contains a certain subset of wormhole configurations the asymmetry

-

in Coleman sense. Due to s included the wormholes that Thus, each line

of the model, we have inadvertently

can couple to the north and the south poles of the four-spheres. in Fig.

1 generally corresponds not just to one four-sphere but to a geometric

Is there a way to turn off these wormholes them off by

sum over linear chains of four-spheres. biilt into a generic minisuperspace

model ? If we succeed in turning in Fig.

effectively

setting p = 0 in Eq. (2.26)) th en each propagator

1 would be

associated with just one four-sphere and would carry a factor e2/3X . The answer to the above question is positive. can be turned off by a careful tuning The minisuperspace wormholes or the spring then

of either the boundary

conditions

spring constant w2(t) near t = O.* For instance, if we make the oscillator in this region sufficiently attractive.(say, by adding to w2(t) a term N b(t)),

we can fine tune fr (t) t o b ecome exponentially ously maintain the boundary condition

decreasing with t and simultaneon w2 (t)

(2.12). The precise requirement

is that the Wheeler-De Witt provided the wave function there exists a metastable

equation have a zero-energy bound state when X = 0, satisfies boundary condition O(exp(2/3X)). (2.12). Then, for small X Since fl(t) is normalcase,

state with lifetime

ized to oscillate with amplitude jr(O) N exp(l/SX). Substituting

I2 (2~0)~ at large t then, in the fine-tuned

this into Eq. (2.21), we find G(O,O) N exp(2/3X).

* The precise form of the Wheeler-De Witt equation is uncertain in this region due to the lack of knowledge of short-distance behavior of quantum gravity. 12

If we compare

this result

with

Eq.

(2.26), we find agreement

provided

we set

~1= exp(---SW) - exp(-2/3X).

Somehow we have succeeded in making the minisuw2(t) in the unknown model which is prob-

perspace wormholes very costly. At the expense of fine-tuning microscopic region near t = 0, we have constructed

a minisuperspace

is free of wormholes

defined in Coleman sense. This special adjustment s of physical parameters but a procedure model.

ably not a true fine-tuning remove the asymmetry adjustment Euclidean wormholes built

necessary to Without this

into a generic minisuperspace

the model would have picked two special points geometry that (the two poles of the four-sphere), to these special points.

on each stationary

and would have contained Now that these asymmetrithat

can attach

cally coupled wormholes

have been removed, we can add by hand wormholes

can create contacts between arbitrary ColemanP

pairs of points, such as those considered by

This will be done in Section 3. different, contained perhaps simpler prescription in the minisuperspace for turning For that case and of the

_ We now offer a somewhat off the asymmetric wormholes

models.

purpose, we need to elucidate the crucial difference between the fine-tuned

the general case. Let us solve for the shape of the Schroedinger wave function out-vacuum backwards Qout(X, t) at early times. in time, we obtain By integrating the Schroedinger

equation

(2.27)

where ..m Q!(t) = -f*(q Without fine-tuning, Re{cu(O)} is O(exp(-2/3X)) function while Im{cw(O)} is O(1). (2.28) This

means that Q?\IloUt(X, is an oscillating 0) cial property oscillations of the fine-tuned are eliminated.

with a broad envelope. is O(exp(-2/3X))

The speand the

case is that Im{cr(O)}

In fact, in the limit

X + 0, QIl,,t(X, 0) becomes flat and speaking, we are only interested

therefore indistinguishable

from Q;(X) 0). Strictly

13

in this limiting finitely

situation:

only in the limit

X +

0, where four-spheres

are in-

large, can we make a clean separation between two bubbles connected by a and one deformed bubble. Working condition with very small A, we will therefore condition

wormhole,

replace the boundary

of Eq. (2.16) with the boundary

%&(X,0) = %d(X, 0)
which is exponentially holes can be fulfilled close to it. Thus, the requirement by choosing the wave function

(2.29)

of the absence of wormat t = 0 of de

of the googolplexus

which evolves into the ground state at late times, and leads to no production Sitter universes.
-

In the Heisenberg picture, 1 in >=

this choice is simply I out >

(2.30)

This leads to

G(W) = f(O)f*(O)- exd2/3$

in agreement with the claim that the asymmetric wormholes

(2.31) have been turned off.

Perhaps, it is no coincidence that, in the absence of a source of small geometries, a suppression of the number of large universes amounts to a suppression of wormholes. The process of production wormhole geometry of a pair of large universes can be pictured as a

where the radius of the universe first contracts

to a Planckian

value and then reexpands. Let us now compare the sum of disconnected diagrams of Fig. 1 with Coleman s sum over disconnected find the path integral four-spheres. to be
z( J, A) = e-f J2f(o).f*(o)

Substituting

Eq. (2.31) into Eq. (2.19)) we

(2.32)

where f(O)f*(O)

- exp(2/3X).

Th us, we have reproduced

Coleman s

double expoto

nential with one important

change: each bubble carries a minus sign in addition

14

the large factor exp(2/3X).

In th e 1imit X + 0 this leads to an enormous suppresThe origin of this suppression

sion of the path integral instead of enhancement.

lies in the fact that the average number N of outgoing universes produced by the action of the current J diverges in the limit X + 0:

N =< out IeiJx(o)utue-iJx(o)l

out >= J2 f(O)f*(O)

(2.33)

It -is also easy to see that the de Sitter universes produced by the current are Poisson distributed. The path integral 2 with no insertions measures the amplitude implies to

produce no universes in the final state. The Poisson distribution

-

1212 eeN =
in-agreement with Eqs. (2.32) and (2.33). Th e entire situation

(2.34)

is similar to QED

with a time dependent external current, which is well known to produce a divergent number of soft photons. The exponential factor in the amplitude (2.32) to

produce no out-going universes is analogous to the infinite suppression of exclusive amplitudes, in which no infrared probability photons are emitted. interpretation: The analogy with QED

suggests the following

since we cannot detect the other which are in no way sen-

universes, we are interested sitive to the infrared probability

in inclusive probabilities,

divergences.

We will postpone a detailed discussion of the

interpretation

until Section 4.

3. Topology Change In Minisuperspace.

In order to complete our construction, cesses in our third-quantized the disconnected Following model. we now include topology changing prosome of

This will have the effect of turning

diagrams of Fig.

1 into diagrams connected by wormhole lines. into the action a

the method of Refs. 2, 3, 4, 18, 22, 23, we introduce

15

variable cy, which is essentially

the field for universes of Planckian

size:

S=;a2+jdt{;k2+;
0

(&-y)X2-JXb(t)}

(3.1)

The role of (Y is to create contact between arbitrary scopic universes. teraction universe. construct vertex, Th is is accomplished which through

pairs of points on the macroof the in-

a repeated application

describes creation

of a baby universe by a macroscopic is non-local. In order to

Since the scale factor is time, this interaction a hamiltonian treatment,

we will make the action local in time by forof t, and by introducing a lagrange multiplier

mally promoting
-

(Y into a function

P(t):

,,y= ;a +?dt{;22+;
0

(-&-

+;+))X -

JXG(t)+@(t)h(t)}

(3.2)
The argument in the first term is arbitrary since there is a constraint Therefore, initial & = 0. The fields (Y and ,B are a pair of conjugate can be taken to be functions *in(X) cy, variables. the wave functions of the form space a

of X and Q, with

conditions

t=

0). Since ir = 0 is obeyed as an operator equation, the Hilbert of time.

breaks up into sectors labeled by cr which is independent general state of the googolplexus definite values of AZ

Specifically,

can be expanded in a complete set of states with

19 >=

y2F~(a)p;Cy >. (3.3) Jdc& >Ia!

corresponding to 19; a > evolves in time according constant X,ff = X + go. F*(a) in function condition for different o-sectors, which is completely to assume that at t = 0. Thus, it is natural
16

The Schroedinger wave function to the hamiltonian Eq.

with a shifted cosmological

(3.3) is the weighting by the boundary

determined

FQ is a smooth function.

In general, the out-state

is

lout >=

daFo,t(c+ut;~

> IQ >

(3.4)
is assumed to from the amplitude

where lout; (x > is the out-vacuum be smooth. in-state Then the path integral, is given by

in each o-sector

and F,,t(a)

which is the transition

to the out-state,

J
is to be computed

daefr i2F,f,t(a)F;,(cy)Z(X

+ ga).

(3.5)

2(X + ga) =< out; alin; a > in a theory with a shifted cosmological summation constant

(3.6) X + go. This

expression closely resembles the wormhole Coleman claimed: Z(X) N exp(exp(2/3X)),

formula

of Refs. 2-4. If, as by

then the path integral is dominated constant

the value of o such that the observed cosmological after setting in>=e we have found Coleman result! s is strongly Z(X) N exp(-=w)I

is zero. Recall that,

out > *

P-7) sign from constant

exp(2/3X)),

which- differs by a crucial

It could then appear that the zero of the cosmological Before we jump to such a drastic conclusion,

suppressed.

however, we the issue

need to analyze perhaps the most confusing issue of quantum of probability interpretation interpretation.

gravity:

As we will argue in the next section, the probability quantization is quite different from the lore

suggested by the third quantum gravity.

of the Euclidean
*

There is no discrepancy with Eq. (2.30). We have simply absorbed the effects of the source at early times into the definition of the in-state. t In 4 dimensions our sign agrees with the sign found by Polchinski (Ref. 11). 17

4. The Probability

Interpretation.
the magical effect of

As shown by Coleman2 and Giddings and Stromingera 4 the wormhole sum is to introduce integral. Coleman went further probability integration

over coupling constants into the path the weight in the integration as

and interpreted

the a priori

distribution

for the coupling

constantsP This formed the

basis for his claim that the observed cosmological constant is zero in the wormhole theory. Implicit in these arguments is the assumption, path integral with insertions first advocated by Hawking, values in

that the Euclidean quantum gravity.

computes expectation

We will see, however, that this assumption

does not generally

hold in the theory of the googolplexus. Our methods quantized hamiltonian rely on thinking of the scale factor by Eq. as the time in the third(2.7), the third-quantized

dynamics. has explicit

Then, as demonstrated

time dependence. In such a theory, the path integral with matrix element of the from

operator insertion

0; is a transition

<

Out

1Oij

in >

(4.1)

On the other hand, the expectation is given by

value of Oi in a state of the googolplexus

1 in >

< in IOil in > which cannot in general be calculated stance, the probability
co PUeff) = 9 J
-KJ

P-2) path integrals. For in-

using conventional

distribution

for the cosmological

constant is given by

Lx

x =glP,, (4*3) 9i, XT - x ,t=O)/1 (AefiA)I eff 9 I (

is a smooth function. The infrared factor 2(X,,,) which

which, strongly

by assumption,

depends on the cosmological

constant does not appear in Eq. (4.3). The

18

wormhole

calculus, which is intimately

connected with the third-quantized of a priori probabilities.

Feynthe

man rules, is not suited for the calculation calculation of a priori probabilities in QED.

Instead,

bears a close analogy to the calculation definition

of inappears large

clusive cross-sections to be related infrared priori

On the other hand, Coleman s which are well-known of p(X,,,)

to exclusive

cross-sections,

to contain

suppression factors. probability

In fact, the properties constant,

in Eq. (4.3), the a on the unknown which

for the cosmological gravity

depend entirely

mechanisms

of quantum

at Planck scale. Coleman s

precognition,

provides the connection our model. Equations

between the Planckian

and cosmic physics, is absent from

(4.1) and (4.2) show that the path integral computes the expectation I out >. This corresponds to setting

values only if I in >=

J = 0in Eq.

(3.7).

However, as explained above, this state of the googolplexus leads to no sharp peaks in-the probability distribution, just like almost any other state. There,

Where does the example of the previous section fit in this discussion? in order to introduce with the state of Eq. a diagrammatic (3.7). structure into the path integral by Eq. X + (2.33),

we worked

As evidenced universes with

this state contains

a large number subscript produce

of de Sitter

0 (from which

here on we drop the is the amplitude to

on X,ff). no outgoing

As a result, quanta,

the path integral, suppressed.

is strongly

On the other hand, the a

priori probability obtained

distribution

for the cosmological constant in such a googolplexus,

by substituting

Eq. (3.7) into Eq. (4.3), displays no sharp suppression

or enhancement

of any value. constants we measure are actually distributed according to

If the coupling the a priori

probabilities

defined by Eq.

(4.2), then our conclusions

can not be

more disappointing. seem to turn

From the point of view of a low-energy parameters into random

observer, wormholes variables. However, Probability Our definition

all the fundamental

we should not be hasty to accept this unpleasant interpretation

state of affairs.

is one of the most confusing issues in quantum gravity.

19

of a priori

probabilities

is perfectly

reasonable from the point of view of an outside But how

observer who can watch entire universes being created and destroyed.

does the world appear to US, the observers confined to one universe? An analogous hypothetical participating question concerns the meaning of probabilities in a scattering process. Within observed by an electron framework such

a second-quantized

a question is confusing and may not have a unique answer. One interesting modification of the definition of probabilities was proposed in weighted and

Ref. 19. There it was suggested that we observe the a priori probabilities

by the number of created de Sitter universes. This number is infrared-sensitive seems to offer a new possibility for solution of the cosmological

constant problem.

To show this we come back to the example considered in Section 2. If the in-state is taken to be of the form (3.7), then th e source at late times. According

Jproduces

a large number of quanta as X + 0. If values of the

to Eq. (2.33), this number is 0(exp(2/3X)) of probabilities for different

this number is included in the weighting cosmological constant,

then we find a single exponential

peak at X = 0+, similar occurs for a

to the peak of the Baum-Hawking broad class of the wave functions boundary conditions

approach.12 This peak actually 13 of the googolplexus,

provided that some generic

are specified at early times. The reason is quite simple. Before is upside-down, and the wave packet spreads. As

t = Xm3i2 the h armonic oscillator

a result, at late times the wave function a large number of quanta, which rapidly consider the gaussian boundary conditions

is in a highly excited state and contains grows with a decreasing X. For instance, of Eq. (2.17). Calculating w, we find* the number

of quanta at late times as a function

of the width

N(w)

=<

I,+

al

>=

&l.f(W

+ ,IfP)l

W2

- f

(44

We see that, first pointed

*

without

any fine tuning

near t = 0, N(w)

N exp(2/3X).

(This was to

out by Rubakov.lg)

The crucial question is whether

it is justified

From here on we set J = 0. Turning on the source does not modify any of the conclusions.

20

identify

the number of produced de Sitter universes with the probability

to find a

given value of A? We believe that the answer is negative. point of view, it is necessary to add some matter The simplest periodic

In order to explain our

content to our model.

model which contains some of the crucial features includes one

scalar field 4, which is an angle ranging from 0 to 27r. If the field has then the problem splits into sectors labeled by the discrete values of conjugate to 4. The third-quantized Wheeler-De Witt field can be

no potential, the momentum expanded as

X(t,
where fk is an incoming

4) =

2 (P k-ccl

f#ak

+ Cik f,*(+t,)

(4.5)

solution of

d2 k2 gT+i$For a value of [El << A, Robert son- Walker a maximum (FRW) e

&

+ p

>

.w>

= 0

(4.6)
a Friedmannexpands to

this equation has two classical solutions: universe, which starts from a singularity, to a singularity;

volume

lkj3i2,

and recontracts

and a de Sitter the a

universe of minimum

volume x Xm3j2. For lkl > &

the barrier separating

FRW and de Sitter regions disappears, universe expanding forever.

and there is only one classical solution:

We will assume that, in this toy model, a universe similar to ours is described by a FRW solution with 1 << lkl << & (4.7)

In other words, we assume that our universe has a small positive spatial curvature which will eventually force it to recontract. We will further suppose that the

probabilities contained wrong

we observe must be weighted by the average number of the universes in the googolplexus, which resemble our universe. universes which contain
21

Then it is clearly no heat. (The

to count the de Sitter

virtually

energy of the scalar field is negligible condition (4.7) is satisfied.)

in the de Sitter region t > X-3/2 provided

Our goal, instead, will be to count the FRW universes with a given large value of k when the cosmological discussion of boundary constant is so small that the condition (4.7) holds. The

conditions

for a given value of k only concerns the oscillators mechanical system of Section 2. As we have imposed at t = 0 leads to production of

ak and ai and reduces to the quantum found, a generic boundary condition

O(exp(2/3X))

d e S er universes. However, as we argued, this number should not 1 tt of the probabilities measured by us. In order to count operators through

be included in the definition

the FRW universes, we introduce

their creation and annihilation

X(t,

4) =

g (eikd hk(t)bk + emit4 h;(t)bL) k-cm

(4.8)

where hk(t) + Generic boundary number conditions &%exp(-iklogt), t-+0 (4.9)

at t = 0 do not lead to a sharp dependence of the evolution at early times to imeffects

of FRW universes on X, since the Schroedinger

is very weakly sensitive to X. As explained pose boundary of quantum semiclassical conditions

in Section 2, it seems natural the short-distance

at t = 0 in order to mimic such boundary

gravity.

Unfortunately,

conditions

do not lead to a

solution of the cosmological results follow

constant problem. if, as in the example of Section 2, we (4.7)

Much more interesting impose boundary conditions

at late times, such as in Eq. (2.30). If condition

holds, then there is a thick barrier separating

the de Sitter region from the FRW which tends to the N exp(2/3X) The same

region. As shown in Eqs. (2.27) and (2.28), th e wave function

ground state gaussian at late times must be a wave packet of width at early times. Such a state contains O(exp(2/3X)) FRW universes!

conclusion follows if the wave function

at late times is not in the ground state, but

22

in any finitely specified

excited

state.

Thus, if we work with generic boundary

conditions

at late times, we find that the number of FRW universes which reach maximum volume V,,, is O(exp(2/3X)) as X + 0. This sharp

some specified

peak is relevant if we weigh the probability are similar cosmological to ours. constant Such a theory problem.

for X by the number of universes which for solution of the

may offer new prospects

REFERENCES
1. W. Fischler, I. Klebanov, J. Polchinski and L. Susskind, submitted Quantum Me-

chanics of the Googolplexus, B 2. S. Coleman, -3. S. Giddings 4. S. Coleman,

SLAC-PUB-4957,

to Nucl. Phys.

Nucl. Phys. B307 (1988) 864. and A. Strominger, Nucl. Phys. B307 (1899) 854.

Nucl. Phys. B310 (1988) 643.

5. T. Banks, Nucl. Phys. 309 (19SS) 493. 6. S. W. Hawking, S. W. Hawking, D. N. Page, and C. N. Pope, Nucl. Phys. B170 (1980) 283; Comm. Math. Phys. 87 (1982) 395; A. Strominger, V. A. Rubakov, Phys.

Rev. Lett. 52 (1984) 1773; G. V. L arrelashvili, Tinyakov, JETP

and P. G.

Lett 46 (1987) 167; Nucl. Phys. B299 (1988) 757; D. Gross,

Nucl. Phys. B236 (1984) 349. 7. V. Kaplunovsky, (1989) 48 8. J. Preskill, preprint Wormholes in Spacetime and the Constants of Nature, Caltech unpublished; W. Fischler and L. Susskind, Phys. Lett. B217

CALT-68-1521

(1988); S. Coleman and K. Lee, Escape From the Harvard preprint HUTP-SS/A002 (1989).

Menace of the Giant Wormholes, 9. J. Polchinski, Wormholes, Decoupling University

Versus Excluded Volume, or, Return of the Giant of Texas preprint

23

UTTG-06-89

(1989).

10. G. Gibbons, 11. J. Polchinski,

S. Hawking

and M. Perry, Nucl. Phys. B138 (1978) 141.

The Phase of the Sum Over Spheres, to appear in Phys.

Lett. B. (1989). 12. E. Baum, Phys. Lett. B133 (1983) 185. 13. S. W. Hawking, 14. B. S. Dewitt, 15. J. Polchinski, Phys. Lett. B134 (1984) 403. Phys. Rev. 160 (1967) 113. A Two-Dimensional UTTG-02-89 Model for Quantum Gravity, University

of Texas preprint 16. K. Kuchar,

(1989). Gravity 2, eds.

J. Math. Phys. 22 (1981) 2640; also, in Quantum

C. J. Isham, R. Penrose, and D. W. Sciama, (Clareadon, 17. A. Jevicki, Scientific, Frontiers in Particle

1981).

Physics 83, Dj. Sijacki, et. al., eds. (World Int. Jour. Theor. and Strings

Singapore, 1984); N. C a d erni and M. Martellini,

Phys. 23 (1984) 23; I. Moss, in Field Theory, Quantzm 11, eds. H. J. de Vega and N. Sanchez (Springer, Changing preprint Topology and Non-Trivial Berlin,

Gravity,

1987); A. Anderson, of Maryland

Homotopy,

University

88-230 (1988). and A. Strominger, Baby Universes, Third preprint Quantization, and

18. S. Giddings

the Cosmological Strominger, Baby

Constant , Harvard Universes,

HUTP-SS/A036

(1988); A. of the 1988

to appear in the Proceedings

TASI Summer School. 19. V. Rubakov, Phys. Lett. B214 (1988) 503. Quantum Field Theory of Universe, Hi-

20. A. Hosoya and M. Morikawa, roshima University 21. M. McGuigan, tion, preprint

RKK

88-20 (1988). of the Wheeler-Dewitt Equa-

On the Second Quantization preprint

Rockefeller

DOE/ER/40325-3S-TASK-B

(1988); Universe DOE/ER/40325-

Creation from the Third Quantized Vacuum, Rockefeller preprint 53-TASK-B (1988).
24

22. I. Klebanov, Constant,

L. Susskind and T. Banks, Wormholes SLAC-PUB-4705

and the Cosmological

(1988), to appear in Nucl. Phys. B Derivation of Coleman Vanishing s RKK 88-28 (1988). Cosmological

23. A. Hosoya, A Diagramatic Constant, Hiroshima

University

preprint

25