You are on page 1of 31

SLAC-PUB-4957 UTTG-03-89 April 1989 T/AS



of the Googolplexus




Physics Department University of Texas at Austin Austin, TX 78712


Stanford Linear Accelerator Center Stanford University, Stanford, California 94309


Physics Department University of Texas at Austin Austin, TX 78712 and



Department of Physics University, Stanford, CA 94305


to Nuclear Physics B

* Work supported by the Department of Energy, contract DE-AC03-76SF00515. t Previous title: A Hilbert Space Analysis of Topology Change in Quantum Gravity. $ Research supported in part by the Robert A. Welch Foundation and NSF PHY-8605978. 5 A.P. Sloan Foundation Fellow. 1 Work supported by NSF Grant NSF-PHY-812280.


The Euclidean formulation of gravity is not a subject with firm foundations I think I

and clear rules of procedure; indeed it is more like a trackless swamp.

have threaded my way through it safely, but it is always possible that unknown to myself I am up to my neck in quicksand and sinking fast. Sidney Coleman

1. Introduction
According ordinary to some very recent speculations, the laws of nature as seen by an determined

observer inside a macroscopic universe are at least partially

by a background

of baby universe or wormhole operators which act on a Fock Th us far the most widely known formulation of this idea

space of universes [l-5].

is due to Coleman [4] and is based on a nonexistent tization.

euclidean path integral quanconclusion that the

s C o 1eman theory seems to lead to the remarkable

baby universes will always adjust themselves to exactly constant. Unfortunately, the unboundedness of the action

cancel the cosmological

IE = J makes it difficult particular, to interpret


- 16kG R, gravity. In

the euclidean path integral of quantum

Coleman mechanism for the vanishing of the cosmological s on the apparent instability with respect to nucleating -2/3X

constant arbitrarily

relies heavily

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

to the action

seems to lead to a catastrophic in spacetime [6].

number of macroscopically (Resolutions

large wormwholes

of this problem have been proposed in ref.

[7], but it is shown in

ref. [S] that these ideas do not work.) prescription which eliminate

On the other hand, there are mathematical They involve rotating the contours

the instabilities.

of path integration The simplest

for all or part of the conformal degree of freedom of the metric. clearest prescription due to Gibbons, Hawking

and mathematically

and Perry [9] rotates the entire conformal factor to an imaginary

value, thus elimi-

nating any negative terms in the action (we assume that the cosmological constant is positive). Such a procedure may define a consistent quantum theory of gravity, the divergences as X -+ 0 which drive Coleman mechas

but it surely eliminates nism.

Other approximate



the contour

about the Euclidean

saddle point four-spheres.

or In

saddle points

associated with wormhole-connected

this case a careful analysis of the modes of fluctuation in front of the Baum-Coleman-Hawking with each four-sphere. (D = 4) the marvelous Evidently, the Euclidean amplitude

[lo] reveals a prefactor [ll,



12, 41 associated in 4 dimensions exp(-exp(2/3X)).

The result is, once again, not favorable: exp(exp(2/3X)) b ecomes a disappointing

path integral is so ill-defined

that it can be imaginatively

used to prove anything. In this paper we will formulate change by returning mechanics ultimately construct a simplified the quantum mechanics of gravity with topology Hilbert space quantum We is

to the high dry ground of Hamiltonian to be derived from the Minkowski quantum

space path integral.

mechanics of a system -the googolplexus-which

rich enough to describe any number of spatially holes which mediate transitions quantization between states.

spherical universes as well as worm The procedure wave function involves a third by an operator couplings with a

which replaces the Wheeler


which acts on a Fock space of universes. and, as in refs. probability

Worm holes induce nonlinear constant results:

[4, 51, make the cosmological We find the following of universes

into a variable


(1) The mean number exp(2/3X)

which are created constant.

by the dynamics

is -

w h ere X is the cosmological

In view of the known bound

on X (A S 10-12 )

th e mean number of universes is X 1010120. This is the

origin of the name googolplexusf (2) The probability for a given X is flat at X = 0, with no enhancement or Coleman type. does occur but in the form exp (-e2j3 ). of a probability for X but i. e., it of either

the Baum-Hawking

(3) The Coleman double exponential Furthermore,

it does not have the interpretation amplitude from an initial

rather a transition

state to an out-vacuum,

* The googolplex is the largest finite integer with a special name. (Websters New Collegiate Dictionary.) 4

It is equal to 10 loOO

is the amplitude

to create no universes.

The exp(2/3X)

like the soft photons emitted

in electrodynamics

universes are rather and the factor exp ( -e2/3x) exclusive

is the analogue of the soft photon factor which suppresses individual states with a finite number of soft photons. (4) The e2/3x of emitted

soft universes are cold, empty and uninteresting. The probability to find an interesting

We warm

do not live in one of them.

universe is like an inclusive probability universes. For these inclusive

which sums over the unobserved soft no Coleman or Baum-Hawking drive X -+ 0.


factor appears. Thus we find no reason to think that wormholes In fact with our present understanding random variable.

it seems that worm holes make X a

2. Second Quantization
Consider the minisuperspace spatially

of Minisuperspace
gravity. This model includes

model of quantum

spherical geometries with metric

ds2 = g


+ a2(r)d#)


where dfli constant becomes

is the metric of a unit S-sphere. We also include a number of spatially scalar fields $;(r). The E ins t ein-Hilbert action with matter coupling

I = k J d7 [-ah This action defines a hamiltonian

+ a - a3(X + V($;))

+ a(&) ]


H = &


- a2 + Xa4 + -$I$

+ a4V(&)



which is the generator is general coordinate

of translations invariant,

in the parameter

time 7. Since the theory the physical states.

this generator must annihilate


This defines the Wheeler-Dewitt verse @(a, 4;) must satisfy: -d2 [ da2


which the wave function

of the uni-

8, - a2 + a4(X + V(&)) a2 84; 84;


= 0


There is some ambiguity

in eq. (5) d ue to operator independent


but the results about operator

in Sections II and III are completely ordering.

of assumptions

s Sometimes eq. (5) is regarded as gravity ously more like gravity s with DeWitt[l3] bedding time. Klein-Gordon

Schrodinger equation,

but it is obvi-


Indeed, many authors, beginning

h ave noted the similarity In 1 + 1 dimensional gravity

between the scale factor a and an emthis correspondence is precise [14]. As

in the case of the Klein-Gordon density

equation, the lack of a positive definite probability to think of @ as a quantum field rather than

makes it more appropriate

a state amplitude tion:

[5, 15-201. Of course, there is a second, independent, space theory with spaces of arbitrary


in order to have a Hilbert

we need an operator components. We will equation, therefore

which connects sectors with different

numbers of connected


a quantum

field theory

of the Wheeler-De


guided by the analogy between eq. (5) and the Klein-Gordon


To emphasize this we rename a+t (6)

di + Xi

so that eq. (5) becomes

[ at2

a2 -_-


t2 + t4(X

+ V(Xi)) I



= 0




In order not to confuse the reader we emphasize that the xi are not ordinary spatial positions and Q(t, xi) is not an ordinary

field. @(t, z;) creates or annihilates

whole universes with scale factor t and field values z;. t and Z; are the coordinates in a new space - the googolplexus Eq. (7) is a Klein-Gordon m2(t,Lq and time dependent all functionals is a complete conjugate metric - where particles are universes. mass-squared, (8) space consists of The field @(to, 2)

equation with position-dependent = -t2 + t4X + t41/(i) )

gtt = -1, gZ, = t2.

The Hilbert

of @ at a fixed time to, set of commuting


IS) N !P(@(to,d)). the time derivative equation


is the as


Z), and the Wheeler-Dewitt

(7), interpreted

a Heisenberg field equation, determines the field at all other times in terms of these. We now have two issues to consider: of the googolplexus, what determines IU), the wave function A meta-observer, a general

and, given 1 , what are the observables? U) via arbitrary

able to couple to the googolplexus expectation value (Ul@(tl,

sources, could determine

~21). . . Q (tk, ?k)lU). observables:

We, on the other hand, are interthat a universe will

ested in single-universe have given properties -

what is the probability that the matter

for example,

fields and energy density

will have given values when the radius of the universe takes some value ao. We are using universe a particle to mean a single connected spatial component, analogy. corresponding of identifying to

in the Klein-Gordon

So, we have the problem Hilbert

these one-universe to the problem

observables in the second quantized the position of one particle

space, analogous field theory. in a field

of identifying

in quantum

Now there is a subtlety:

it is not in general possible to identify if the Hamiltonian


theory, even in the absence of interactions,

is time-dependent. of m2

It will be possible only in regions where m2 is positive in eq. (8) and the metric is the case. m2 > 0, at sufficiently large t, provided $<l, that *<1 m gxx X + V(Z) are adiabatic.

and if the variations

In the present case one verifies that this

(9) For convenience, we

is positive.

will assume the latter

to be true in this paper.


Then at large t we can identify

universes, adiabatic

which we will refer to as out-universes. vacuum state lout)

In particular,

the late-time

is the state with no universes as t + 00,

(Q, tlout)

- exp {-+Jdim(t,;)

( m2(t,

2) -


. d)li?




for t large. Wheeler-Dewitt

The field Q(t) 2 ) is expanded equation (6))

in a complete

set of solutions

of the

The division

into raising operators (lo),

Af- and lowering


AI is made using

the out-vacuum

AIlout) The fl are orthonormal,

= 0



where the Wronskian

w(.f-,g) = / dZ f(t, z 2) - f(t, 4dt, 2) >i(t,

is time independent. Then
[AI, A:] = &J . (15)

Let 0 be an observable defined in a single universe, and let 01 be its expectation value in the universe described by solution 8 f~. Then its expectation value




isth e number of out-universes

of type I. More generally,

if 0 is not diagonal in the basis I, this would become




is the effective density matrix

for single universes. values -


nally, we will usually be interested in conditional in particular, on the presence of cosmological

expectations conditions


suitable for our existence. weight operator. field IV)

Thus, (0) N tr OWp/tr

Wp, where W is some appropriate

Next, what determines equations (7) suggest that

IU)? The M in k owski nature of the googolplexus 117) is determined at t = 0. It is possible that

is determined similar context). in spirit

at t = 0 by a smooth match on to short distance to the Hartle-Hawking idea [22] extended

physics (this is

to the many-universe

The idea is that the universes at t = 0 are the baby universes whose the local lagrangian of our world. We shall consider a number of

state determines

hypotheses for IU). H owever we will see that the main features of the physics we are interested in is largely insensitive to the details of (U) so long as IU) does not features arise from

have a fine tuned singular evolution

dependence on X. The interesting

between t = 0 and large t. We will discuss 117) in more detail in Sec. IV. We specialize to pure gravity (with X > 0) for with

We now apply these ideas. which second quantized dependent in-state frequency.

minisuperspace The striking

reduces to a harmonic


feature of the dynamics

is this:

the generic of

IU), with few or no universes at t = 0, may contain is that the frequency-squared

a large number

universes as t -+ co. The point

(8) is negative at is upside-down is a highly

small t, changing sign at t = Xm1i2. Before t = Xm1j2, the potential and the wave packet spreads. As a result,

the wavefunction



state of the Hamiltonian

in the late time,



This state

contains a large number of quanta, which are the out-universes. We see this quantatively only one value; the solution approximation to be j(t) - (4P)1/4e-i(tw2/3 +6). 7

as follows.


no matter

fields, the index I takes


as in (13)) is found in the WKB

t >



There is a single lowering operator, A = iW(f* The number of universes in the adiabatic , a) .


region is .

n = (UIAtAIU) This number is readily estimated

(20) of the Wronskian

by using the time-independence

to express A in terms of the t = 0 Heisenberg field, A = if*(O)II(O) - if*(O)@(O) .

(21) approx= 0.

For small A, the behavior of f(t) imation. For convenience, Ref(t),

is determined

near t = 0 from the WKB

we choose the phase S in eq. (18) so that Ref(0) condition

The solution

being fixed by a t = 0 boundary - Xt2)3/2/3X]

(but normalized

at large t by eq. (18)) is O(1) . exp[(l while Imf(t), with Ref(t))

in the region 1 5 t 5 X-1/2, that it is 7r/2 out of phase

being fixed at large t (by the condition

is O(1) . exp[2/3X - (1 - Xt2)3/2/3A].

In all,

f(0) = -,;l&~ f(O) =

+ic2e1/3A )

+ icle113A

of the Wronskian.

where cl and c2 are of order 1 and we have used the conservation


Then from eqs. (20), (al),

and (22),

n _N e2/3A(uJ {c2@(0)

+ clII(o)}2

IU) - e2 3x


By our assumption

about the nature of IU), th e matrix

element is non-singular O+. This is the and [la] in

in X, and so the number of universes diverges as e2/3x for X + large exponential factor of Baum [ll] and Hawking

[12]. While refs. [ll]

were in the context

of a single universe (single connected spatial


our approach this large factor is interpreted

as the average number of universes of space. In the following of the

that is, as the average number of connected components sections we will discuss the relevance of this factor cosmological After obtained Morikawa Dewitt constant. these results,

to the determination

obtaining previously

we have learned that many of them have been Kuchar [15], Rubakov [18], Hosoya and

by other authors:

[ 191 and McGuigan

[20] h ave observed the analogy between the Wheelerequation in a time dependent background, between in-universes and out-universes. In

equation and the Klein-Gordon distinction

and have noted the resulting particular, Rubakov

[18] g ave the interpretation

of the Baum-Hawking

factor as the

number of out-universes. [13], Vilenkin context.

There is also some similarity [24], although

with discussion of Dewitt

[23] and St rominger

these were all in a one-universe

3. Topology
We now introduce first is to introduce topology

Change in Minisuperspace
change. We will do this in two different ways. The the method

a baby universe field cy into the action, following action is

of refs. [2, 3, 4, 17, 21, 251. The free Klein-Gordon

co S free = J


; 62(t)

+ $(t - Xt4p2(t)}



Consider an interacting


1 f &2(t) + ; [t2 - (A + a)t4]fD2(t)} SBU = -----a2 + dt J 1 2g2



Here CYis a dynamical



of t. Just as in refs. [2, 3, 4, 17, 21,

251 the Feynman graphs resulting from such an action will correspond to spacetimes in which the spherical spaces described by @ are connected by wormholes propagator). In order to keep a Hamiltonian interpretation,

(the o

we make the action

local in time by means of a Lagrange multiplier

sf3u= & cY2(0)

00 + J dt { f

i2(t) + 3 [t - (A + a(t))t4]Q2(t) + P(l)&(t)} .

The fields (Y and p are a new pair of conjugate variables, so the wavefunctions be taken to be functions is arbitrary,


of @ and o. The argument of o in the first term of eq. (26) IO).

since du = 0; this first term can be absorbed into the wavefunction \-Iu(Q, o; t) = (a, oy;tlU) invariance

Again, we assume that the wavefunction function at t = 0, determined

is some smooth

by coordinate

and Planck scale physics.

Since d! = 0, the Hilbert independent X + a!. What constant? of time.

space breaks up into sectors labeled by (Y, which is constant X,ff =

Each sector has an effective cosmological distribution

can we say about the probability

for the cosmological

As given by the ordinary

rules of quantum

mechanics, this is

we,,> = (WV + a - X,,,)lU)


By our assumption distribution


IU), th is is a smooth function constant


of X,ff:

the probability

for the cosmological

is determined

by Planck scale physics

and is not peaked at 0. This is in clear constrast Coleman[4], which has a double exponential peak

to the quantity



exp (e2/3Aeff)

of Coleman s calcula-

In the sec. 4 we will discuss the Hilbert tion.

space interpretation

Let us next consider the total number of outgoing




= Jdn -A

jL@ --oo

exp [ &)]


We see that for a smooth matrix 0 = X,ff . Thus in a certain x eff = 0. It is tempting problem.

element this has a nonintegrable sense the overwhelming number

peak at X + LY=. of universes has

to speculate

that this in itself solves the cosmological


We do not think

that this is so. for a given It

First of all the factor

e2/3(x+(y) in eq. (29) is not the probability more singular

value of X in the same sense as Coleman s

double exponential.

represents the number of created universes, given that X,ff Coleman s

has a certain value. In

theory the number of created universes is also e2/3(x+4. Coleman has an additional probability function for X + o to have a


a given value, namely, the double exponential priori probability appears in (29). the reader may be tempted

of eq. (28).

No such additional

Nevertheless, thropic

to invoke some sort of weak an= 0,

ideas to say that, since so many more universes are created with X,ff likely to be in one of these.

we are overwhelmingly


when matter

is included matter. indicates

one finds that the e2/3Xeff universes are all empty, cold and devoid of study of the theory with matter, presented in section 5,

A preliminary

that the number of habitable

universes is not peaked at X = 0 if the condition at t = 0.

state of the googolplexus

]U) is specified by a generic boundary better treatment

Let us now give a somewhat satisfactory to introduce

of wormholes.

It is not really

an independent

baby universe field CX,since the intermedidifferent, adding

ate states are just small @ universes. We therefore try something a CF3interaction [5, 17, 191:

Scubic mdt ~2(t) (t2 Xt4)~2(t)} =3 + J {

0 cm


J 0

dtdt @(t)qt dt )qt)p(t, ,t) . t

that a universe of radius assumption

We have added a term which allows for the possibility

a splits into universes of radii u and a. We now make the standard that this is negligible unless one universe is microscopic,

and that the amplitude of lowest

for emission of this small universe is proportional dimension, the cosmological


to the local operator


S;ubic = ; J{ 0



+ t2fD2(t)

- t4Q2(t) (A+gidt ).(t p(t )) } .

Here p(t)

is a function

which falls rapidly

for large t, such as eet2.

In order to we will Also,

make the theory require p(t)

simpler to analyze, we will make a slight modification: for t > t,, where t, is some cutoff radius.

to vanish identically

we replace t4 in eq. (31) with

f(t) = t40(t- tc)

so that the support of p(t) and c(t) does not overlap. make no difference, very tiny universes. large compared since it amounts to neglecting The modification the cosmological so that wormholes

(32) should constant of

One may imagine t, - (lev)- ,

which are (32) a bit

to the Planck scale are still included,

while the modification sense. With

only affects universes which are very small in a cosmological more effort one could analyze the theory without confident that the physics would be unchanged. an auxiliary

these modifications,

and we are

As before, we introduce allow a Hilbert

field to make the action local in time and

space interpretation:


S cubic =

dt + k2(t) J

f {t2 - @)(A

+ p(t))}


- s4v(~)) .

+ PM (i(t)

In order to recover S~ubic from Sorbic, q(t) must satisfy q(0) = 0, while q(m) unconstrained. Intergrating over


then gives

fwdt> = 95(t)

= g<(t)

0 03 0

PvPD(t Pt

p(t )@(t )dt

the last equality


from the cutoffs on the support of 5 and ,LL.


The theory (33) is solvable. The Heisenberg equations of motion @qt) + {t2 - ( x + n(t))C(t))


W) &l(t)
&p(t) +

- 94t)Pw - 9Pww>

= 0 = 0
= 0 *


Note that

by (36), q(t)

is constant

for t > t,.

Thus, in the large - t analysis,

q is a constant independent

q(oo), and the Hilbert

space breaks up into q(oo) - sectors with constant X,ff = X + q(o). Thus


and with cosmological


almost identical

to the earlier CL We can form the same observables as in the cx distribution for X,ff,

model: the probability

ptAeff> = (UlS(X+ q(m) - X,ff)lU)

and the total number of universes,




The state IU) is given by

(Q, !?;OIU) = emc2>

the q-dependence following



from the remark below (33).

From (36) and (37) we have q(t) = qbL

t > t,

PM = P(O)?
and so

t < t,


+ (12 - ( x + qt4)Ctt)) g(X,ff,

t) which satisfies

- 9/4)P(O)

= 0


We define a function

-@g(X,ff, t) + (t2 - LffCtt)) g( t, = O eff,

with the same boundary condition as the function
16 f(t):


the large t behavior

(18), with function

X + f(t),


and Re S(X,,f,O)

= 0. Th is is essentially
t4 -+ c(t).

the same as the

except for the minor modification g(X,ff, jr( X,ff, 0) = -c~1,(Xeff)e-1'3xeff 0) = ic2( Xefj-)e1/3Xeff

Thus, at t = 0,

+ iCl(Xeff)e1'3xerr (44)

Also define gr,2 (t) satisfying

-&&2(t) +

t2g1,2(t) 91(O) 92(O)

= 0 = 1 h(O) = 0 (45)

= 0 42(O)

= 1


(45) and (43), one sees that

9Cxeff, t, = 9(xejf7 0)91(t) + b(&jj,O)gz(t)for

t < t,

(46) that we

We can now proceed to solve eqs. (35, 36, 37). The parts of the solution need are: from (42))

w>= w%n(t) II(cJ)g2(t) + t


0 +

dW )

{91(t )92(t)

- 92(t )g1(t)}


for Using this in (36),

q(m) = q(0) @(O)h




where hi are the constants


= 9

L 0


dt )gdt 4t )

h3 = g2

dt 0

tc dt p(t)p(t )gl(t)gl(t )sign(t t ) .




A = iW(g*(X

+ q(m), t), a(t))

- ig ] dt )g(X & t

+ q(m), t )p(O)


where the second term appears because of the source in (42). Evaluating
t = 0 gives

(49) at

A = ig*(X + q(m), 0) {n(o) - hp(O)) (50) - ii*(A + q(m), 0) {Q(O) + hzp(O)}

We now determine obtained the observables (38) and (39).


The following

using the integral representation Lemma, are useful:

of the delta function

and the Campbell-


(q = qqq + xp + y)lq = 0) = 1/x

(q = OK7 +
(Q = OlP%

P +

Y>Plcl= 0) = -Y/X2 Y)Pl!I = 0) = Y2/X3 *


+ XP +


tAeff> jfq =
and n=


-03 co
N 0

JdXeff J J
dX effe
2/3AeJf -03 iQ~O(ct2(Xeff)~

(ulS(X + q(m) - X,ff)A+AIU) co




ic;(xeff) %)2QlI(@)


where 4 txeff> = c1 (Xeff) ( 1 + !!g) + QtXeff) 2 G


= C2(Xeff)

( 1 -



As in the case of the o-model, . smooth. diverging There is one pathology as X,ff t foe.

the probability

distribution probability

for X,ff,

eq. (52), is

here: the integrated

is not normalized, wavefunction

This traces back to the non-normalizable

for q, eq. (40). W e assume that this is an artifact trust it in any case for IX,ffl model is reasonable for X,ff identical at X,ff

of our model, since we do not that the

;3 1 and we will proceed on the assumption

<< 1. The number of universes, eq. (53), is almost and has the same non-integrable singularity because of @, but is

to that found in the o-model, -+ 0. The matrix

elements in eqs. (29) and (53) differ slightly the small-t evolution

the wormhole

physics has gotten mixed up with

the physics is the same. The agreement between the Q3 model and the o-model the Hilbert wormholes space analog of the clean separation in the euclidean path integral. @ so as to eliminate to the o-model. the results:

between large spheres and small the Q3 model can also be


analyzed by shifting a model very similar

the linear ppQ term from S,!:,;,, leaving

To conclude this section, we summarize holes, the probability distribution for X,ff

in both models of worm=

is smooth, with no preference for X,ff by X,ff

0. One universe observables, on the other hand, are dominated divergence being the single exponential peak of Baum [ll]

= O+, the [la]. Again for

and Hawking

we emphasize that this is not a solution of the cosmological reasons that will be explained in sec. 5.

constant problem,


4. Correspondence

with the Path Integral

Coleman [4] f ound that the Euclidean path integral has a double exponential peak (28) as X,ff + 0. In this section we discuss the Hilbert space interpretation

of Coleman calculation. s In general, path integrals represent transition a time-independent Hamiltonian, the initial amplitudes. In the special case of

and final ground states are the same, amplitudes are also ground

and so the ground state to ground state transition state expectation values. In a time-dependent


as we have here, there

is no correspondence between expectation latter

values and the path integral (unless the way).


is generalized in a rather complicated

Thus, Coleman s The lout)


should represent something of the form (outI..*

vacuum is given on the path

by eq. (10). Th e ]in ) st a t e d ep en d s on the t = 0 boundary conditions integral; we assume that this is to be identified Given a quantum theory in Hilbert with IU).

space, it is a standard exercise to derive the In the present case, this is complicated can be done in but will

corresponding path integral representation. because a contour rotation

is necessary, and because this rotation

several ways [al]. We will therefore not discuss the path integral directly, calculate
(out I . . .

IU) expectation

values and compare with Coleman result. s now. In the case that the tri--

We need to discuss IU) somewhat more explicitly linear universe coupling9

is small, a connected matrix element (outI@(


is of order gnm2. Neglecting n > 3, since these go as positive powers of g, the wavefunction order 9-l: IU) 21 ,;(ro(o)+J n(o))lgll) where J and J are O(1) and (55) IU) is therefore a gaussian with width of order 1, centered on a value of

WI) = 0 B, E H(0) + i7jqO) .

For convenience we ignore wormhole effects and work at fixed X; afterwards,


can add the a-field

and introduce -

an arbitrary





Then (we set J = 0 for simplicity

the more general case works out the same),


= (out~~)e-JZG(o~o)~2g2



From eqs. (11) and (56)) B, = ctA + PA+ where Q = f(0) p = f (0) Then iv) = ,-PA I2~ A and
lout)( 1 - jp2/cY2 l)1 4


+ ivf(O> + +f*(O) e



G(t1, t2) = f(t>>

where t,


- ~ftk)}


is the larger of tl and t2 and t< is the smaller. part of eq. (57) is the exponential factor. This factor is given spheres:

The interesting

by the sum of graphs shown in figure 1. This is a sum over disconnected each line represents a universe which is created with propagates (figure lb)). for a while, Thus,
tout lu> N e-J2WJ9/2g2

zero radius by the source, by the source

shrinks to zero radius,

and is annihilated


represents the same graphs that give rise to the double exponential work of Coleman [4]. Naively

(28) in the

it appears that we must have G(0, 0) - e2i3 This .


is not generically

the case, however. Rather,

-if (0) G(o = j(0) + @f(O) O)

= -f,

Cl + 25

. -1 e-a/3X

c2 + i7jcl
c2 + iqcl

7jcz1 e-2/3x O(e-2'3X) .


This result, viously

which at first sight seems to disagree with


was found preIt can be

from an analysis of the minisuperspace as follows. The minisuperspace

path integral

in [al].


model, even in the free case (no (ra2 These are

or Q3 interaction) configurations,

already contains a subset of wormhole configurations. in the path integral over a(r),

in which a vanishes at one or more to the north and south poles

values of T: these are zero-size wormholes of large 4-surfaces. Thus, G(tl,



gets contributions

from all of the paths shown

in figure 2. Note that eq. (64) can be formally


G(O,O) =


) + i ye2i3



+ $1

e213 + .-*)


the series being geometric.

One sees roughly the correspondence

between eq. (65) which gives a

and figure 2, with each large arc representing contribution

a euclidean four-sphere

of order e 2/ 3x. The formal sum of these multiple

bounces is the O( 1)

result of eq. (64)[21]. However, this result is not consistent. It applies neither to the theory without

wormholes nor to the theory with all wormholes turned on. Indeed when wormholes are accounted for, let us say by introducing the worm holes connecting the wormhole field cy, we will overcount

north to south poles. wormholes? Evidently. Varying

Is there a way to turn off these minisuperspace q in the initial state changes the weighting

given to bounces at a = 0. Setting

17= &/cl,

we see from eq. (64) that

G(O,O) M c12e2/3x
(0utlU) -


- J2c12e2/3X/2g2)


This is rather the exponent

close to the double exponential differs.

of ref.

[4]: only the minus sign in

It has already been noted [lo] that the plus sign assumed in but the result (66) does not agree with ref. [lo] either: which happens to be -1 in d = 4, It seems that we are on the but some

ref. [4] was not justified,

the phase found for the exponent was (-i)d+2, but the minus sign in eq. right track in identifying

(66) is d-independent.

the path integral with (out1 . . . IU) amplitudes,

details remain to be sorted out. How are we to interpret the fine-tuning needed to produce the double expo-

nential form (66)? We offer three possibilities: (i) It is possible that the correct answer for G(O,O) is eq. gravity (64)) of order 1,

and that the path integral spherical rotations

for euclidean quantum action.

does not pass over the there are contour the large

saddle of large negative which make the euclidean In this interpretation

This is plausible:

action positive the fine-tuning

[al] and eliminate

saddle point. to correspond

which gives eq. (66) happens

to a contour which does pass over the saddle, but this finely tuned

object is uninteresting. (ii) tifacts The condition q = icz/cl may be necessary to eliminate with minisuperspace ar-

and give the correct correspondence in part the state IU). way. It corresponds

the full theory.

Thus, this de-

termines in another

W e note that this choice of 77 can be described to the boundary With condition i(O) = c2@(O)/cr in the Wheeler-

the second quantized Dewitt (iii) equation One problem

path integral.

this boundary at X = 0.


has a normalizable


with a purely imaginary 23

7 is that the state IU) is not normal-

izable. This can be cured by adding an exponentially

small real part:

7 =


+ O(e-2/3X)

which sets ,B = 0 in eq. as the source J is turned universes. In

An example of this kind of fine tuning (60). According off, the fine-tuned

is 77= if*(O)/f(O)

to eq. (61)) this results in 1~) = lout): state of the googolplex boundary condition

IU) contains no outgoing at t = 0 is equivalent

this case a fine-tuned boundary Further condition analysis,

to a natural

at late times. probably going beyond minisuperspace, is needed to decide

between these alternatives.

5. Matter

and Heat

Let us return to eq. (7) which d escribes universes with some number of matter fields and restrict V(x) it to the case of one scalar field x. For simplicity invariance. let us choose

= 0 in which case we have x-translation

As usual, we can Fourier

expand Q(x) and find that the dynamics of the Fourier modes &( Ic) decouple. Thus a given universe Dewitt can be characterized by a wave number k and the Wheeler

equation for such a universe is

&(t; k) = 0


We see that for k # 0 there are two oscillatory the de Sitter region of large t
(t ;L l/a),

regions of t.

In addition


a second classically

allowed region expanding

appears with t ;S fi. and recollapsing

Th is is the region of Friedman-Robertson-Walker These universes have enough matter as outgoing


to cause them to

recollapse, never appearing these are the interesting

universes at large t. We will assume that

universes which are similar to our own.


In the FRW region creation and annihilation To do this we find two oscillatory behave like

operators can be introduced Witt


solutions of the Wheeler-De

equation which

and expand 6(t; k) as

k) = 3



&(t; k) = B(k)h+

+ B+(k)hand creation operators

for FRW

B(k) and B+(k) can be regarded as annihilation

universes. Since the FRW universes typically evolution of the wave function

collapse before X becomes important, will not significantly


from t = 0 to t z &


on X as X + 0. Therefore, function of the googolplexus

no Baum-Hawking


will occur if the wave condition at t = 0.

is specified by a generic boundary that the probability

We are thus led to the negative conclusion universe is not peaked as X,ff + 0.

for an interesting

6. Conclusion
We have given a Hilbert space analysis of quantum gravity with topology

change, using a second quantized are wormhole-connected pretation constant.


model whose Feynman diagrams probability inter-


This theory has a natural distribution

which results in a smooth probability

for the cosmological

Path integrals in this theory represents transition values. In particular, the double exponential
(o&l. --


not expectation

of Coleman appears only in ampliamplitude: the amplitude of out-

tudes of the form for IU) to contain universes.


This is an exclusive


zero, (or, with insertions,

some small number)

There is a close analogy with the soft-photon


divergence of QED. With

a small photon mass m, an external



will produce a numamplitude then

ber ny of soft photons which diverges as m + vanishes as e-7.

0. Any exclusive

The large cold universes produced

by the mechanism

of sec. 2

are much like the soft photons of QED, with ny + e2i3 . However, these soft universes are as uninteresting as the divergent cloud of

soft photons produced by an accelerated charge in QED. When matter is included, say by eq. (7)) then warm excited universes will be like hard photons. question will go something like this: for each value of X,ff, A meaningful

what is the number summed By

of universes with a given amount of heat (and other relevant properties) over all possible numbers of the unobserved cold empty de Sitter


analogy with QED, the suppression factor (66) will not appear in this expression. In conclusion, we seem to be left in the following unhappy predicament. Worm as

holes do influence coupling constants and give rise to a probability claimed by Coleman. Hawking or Coleman However this probability function


is not given by either the Baumshort distance physics

but is defined by unknown

which has no reason to prefer X,ff

= 0. Thus, even if X were tuned to zero, worm a probabilistic quantity with no peak at

holes, if they occur, would still make X,ff x eff = OIs there an escape.? One possibility and X = 0 for other reasons. second possibility of the googolplexus

is that

worm holes do not exist at all new to say about this option. conditions A

We have nothing

is that we have the boundary all wrong.

on the wave function let us

To see how this might affect the conclusion,

suppose that the boundary Specifically vacuum x eff.



on IU) are given at t -+ oo instead of t -+ 0. at late times is the out

we assume that, for each k, the wave function or any other wave packet whose width the argument

has no sharp dependence on the Fock space of


in section 2 we then find that

FRW universes must be highly excited. FRW universes is N exp(2/3X,ff). Such a speculation

Indeed, for each k the average number of

represents a radical departure from the usual thinking




We usually

assume that

the laws of nature

are specified

at small we see no

distances (early times).

Here we have done the reverse. At the moment

way to decide if this is reasonable. Acknowledgements We are indebted which was entirely thank Captain to Sidney Coleman for his inspiring responsible for stimulating for providing work on topology change We also

our interest in this subject. working


a stimulating


1. 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. 2. S. Coleman, 3. S. Giddings 4. S. Coleman, Nucl. Phys. B307 (1988) 864. and A. Strominger, Nucl. Phys. B307 (1899) 854.

Nucl. Phys. B310 (1988) 643.

5. T. Banks, Nucl. Phys. 309 (1988) 493. 6. V. Kaplunovsky, (1989) 48 7. J. Preskill, preprint Wormholes in Spacetime and the Constants of Nature, Caltech unpublished; W. Fischler and L. Susskind, Phys. Lett. B217


(1988); S. C o 1eman and K. Lee, Escape From the Harvard preprint HUTP-89/A002 (1989).

Menace of the Giant Wormholes, 8. J. Polchinski, Wormholes, 9. G. Gibbons, Decoupling University S. Hawking

Versus Excluded Volume, or, Return of the Giant of Texas preprint UTTG-06-89 (1989).

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


10. J. Polchinski,

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

Lett. B. (1989). 11. E. Baum, Phys. Lett. B133 (1983) 185. 12. S. W. Hawking, 13. B. S. Dewitt, 14. 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 15. 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, 16. A. Jevicki, Scientific, Frontiers in Particle


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

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

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


1987); A. Anderson, of Maryland



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

17. S. Giddings

the Cosmological Strominger, Baby



(1988); A. of the 1988


to appear in the Proceedings

TASI Summer School. 18. V. Rubakov, Phys. Lett. B214 (1988) 503. Quantum RKK Field Theory of Universe, Hi-

19. A. Hosoya and M. Morikawa, roshima University 20. M. McGuigan, Equation, preprint

88-20 (1988). of the Wheeler-Dewitt (1988); Unipreprint

On the Second Quantization preprint


DOE/ER/40325-38-TASK-B Quantized Vacuum,

verse Creation

from the Third




21. I. Klebanov, Constant,

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

and the Cosmological

(1988)) to appear in Nucl. Phys. B Phys. Rev. D28 (1983) 2960; S. W. Hawking,

22. J. B. Hartle and S. W. Hawking, Nucl. Phys. B239 (1984) 257. 23. A. Vilenkin, function Phys. Rev. lXX

(1988) 888; The Interpretation (1988). of the Cosmological

of the Wave-

of the Universe, Tufts preprint A Lorentzian Analysis (1988). Deviation

24. A. Strominger,



lem, Santa Barbara preprint 25. A. Hosoya, A Diagramatic Constant , Hiroshima

of Coleman Vanishing s RKK 88-28 (1988).





the exponential of eq. (63). time 7.

1) a) The sum of Feynman diagrams representing The vertical axis is the scale factor

a, the horizontal

is parameter

Each line represents G(0, 0)) th e sum over all paths from a = 0 back to a = 0. b) G(O,O) is a sum over minisuperspace sphere. 2) The euclidean T < rzr/fi, for G(O,O). wormholes trajectories that of the form u(r) = 1sin(fir)l/fi, with 0 < geometries with the topology of a

need to be included

in the semiclassical


The reflections

off the barrier

at a = 0 are the minisuperspace

that attach to the north and south poles of the large four-spheres.




Fig. 1


4-89 6335A2 z

Fig. 2