a

Fermi National Accelerator

Laboratory
FERMILAB-PUB-89/141-T CERN-TH.5433/89

QUENCHING

THE COSMOLOGICAL

CONSTANT

T.R. Taylor
Fermi Nationsi Accelerator Laboratory P.O.Box 500, Batavia, IL 60510, U.S.A. ad

G. Veneziano CERN,lZll Geneva 23, Switzerland

Abstract We propose loop quartrum is ccnsirtent and try to justify gravity with comctiom a model - inspired - in wbicb needa at early by known features coamoiogicd of oneconstant suppressed

the effective

infiationary

time8 and suf6cimtiy

at later epochs.

RetCERN-TH.5433189 May 1939

by Flss..rsil AsIOc~ation contract u7c s,.t*s Inc. undmr with United D*p*Rm*, 0,Enwgy e QwS1.dn,r.rsit,.r

-l-

1. Introduction In a fundamental physical cosmological hand side of Einsteins theory constant of quantum gravity, such as Superstring Theory, the

A is best defined as the coefficient

of g,,,, on the right

equations: = -bg, + sources + higher d&v. terms,

Rw - fgwR
seen as the Euler-Lagrange The great mystery

(1)

equations

of the low-energy

effective action of the theory. by the need to have a As

about the smallness of A is reinforced

large A [say O(1) in Planck units] in the early universe in order to sustain inflation.

the universe expands and cools down, and after many changes due to all sorts of phasc transitions, smd.value symmetry breakings etc., A should eventually relax to an infinitesimally

[say 0(10-20)

in Planck units] in the present epoch. problem was started a few years sgo

A new line of approach to this long-standing by Baum [l] and Hawking and stimulating

[z], and was later developed by Coleman in a very influential by Banks [4]. Gravity

paper [3]. Related ideaa have been also put forward invoked in these papers is that of Euclidean employed in the functional

The general framework

Quantum integral

(EQG) and the specific configurations called wormholes [S].

are the so-

After a few months

of great excitement,

the wormhole

scenario appears now to

suffer from many problems, levels. In our opinion,

both at the conceptual

and at the phenomcnological in the first category are related to

the most serious objections wormboles

the fact !6,7] that standard

lower the Euclidean

action in such a way as this

to make it unbounded from below at large Euclidean unboundedness that gives Colemans

volumes V. It is precisely

result A --) 0 o V + co, and that casts, at the

-2-

same

time, great doubts on the legitimacy correspond to any stationary

of the conclusion.

In particular, wormhole-induced

A = 0 doe,, effective

not

point of the non-local,

action. In spite of this and other criticisms, abandon Probiem. the hope that EQG contains we believe that it would be too hasty to

the solution of the Cosmological Constant : The general idea that some quantum gravity corrections could become aad thus affect the present value of A (without changing

sizeable at large distances it in the early universe) A wormbole-like

does seem very appealing. using just saddle points and doing precisely this job

mechanism

was put forward in ref.[6]. It would leave a positive Ao ua.fTe.cted, thus saving itiation in the early universe. would-be final, negative Howevu, a~ the U~~V~XSC COOISand Ilo rolls down towards its wormhole term effectively renormalizes

value, the non-local quantity.

it to a very tiny positive

The problems with that proposal

are basically two:

i) wormholes should contribute

to the action with a sign opposite

to Colemans; of other

ii) the mechanism becomes ineffective

if one assumes that the contribution

universes leads to Colemans double exponential.

In this short note, we present an attempt understood effects of the wormhole nothing

which does not involve

exotic and poorly mechcorrec-

type: it is supposed to generate a similar quantum gravity radiative

anism by invoking

other than standard

tions. We should point out that this idea is not entirely past have analyzed EQG corrections an instability to Einsteins

new. Other authors [g] in the a cosmological solution term,

action with

with the goal of finding flat space-time.

of the classical (DeSitter)

in favour of

The previous

proposals, however, suffered from being either limited

-3-

to on-shell

calculations

or from being affected by the notorious to the off-shell effective action.

gauge-dependence Our approach is free enough

problem of quantum of these problems. for deciding whether itational vacuum.

corrections

However, as we shall see later on, it is not yet complete quantum effects lead to the desired rearrangement

of the grav-

We hope to convince

the reader that every effort should be made cor&tions at large distances in EQG in

to improve

our understanding

of radiative

order to find out if our proposal - or a suitable 2. A toy model that works

modification

thereof - will work.

We shall start by producing

a sort of toy model which works and which possesses corrections are supposed to exhibit. We shall then calculations.

many of the features that radiative

try to find out how close we can get to what we need by some bona-fide The toy model
assumes

that the tree-level

action:

s = l-0 = &
is modified by quantum

/ d+

Wo - R)
by a term:

(2)

effects at large V = i dzfi

AS

l-1 = @A, Vlog(V/X)

(3)
in position

where 0 is a pure number and X is the ultra-violet

cut-off of the theory

space. We do not know of any other way of making sense of EQG other than assuming the existence candidate of a finite UV cut-off somewhat larger than Plancks length [9]. A

theory providing

just such a cut-off is, of course: String Theory, a = inverse string tension

where:

x = A, = hiz,

(4)

and

x,fi=x,=dxE,
Although

a0 = Grand Unified

coupling

constant.

(5)

we shall have in mind string theory e.athe way to regulate EQG, the reader favouritc way provided it does not spoil general covmimce. The fact that, at to Ai V is
computed too

can replace it with her/his

A few remarks are in order before we proceed to using eq.(3). one loop, the logarithm Al known [io,ll]. of the UV cut-off multiplies

a term proportional P can be reliably

In any given theory the coefficient

(see below). the i&a-red

What is much less obvious - aad which (IR) scale of the logarithm

we shall discuss later - is that Adding classical now eqs.(2) equations:

is just the volume itself.

and (3) into the effective action S + AS, we find the modified

Ri, whose solution renormalized,

f g,wR

- {As + WG universe,

A; [ log(V/A)

+ I]}

gw,

(6) where the

is again a DeSitter

R,

= Agp, V = 24x2/A,
condition:

or effective A is given by the bootstrap

A = A., -

16rPGA,[log(AX)

+ const]

(7)

If 0 > 0, A is found graphically fig.(2). Evidently,

as shown in fig.(l).

Its dependence on A0 is shown in

as long as AO is positive,

AO = A, while, when he becomes negative the value:

and not too large in Planck units, A becomes very tiny, approaching A = X- exp(-16x@G/A,/ )-I.

(8)

If the tree-level,

zero temperature
0 <

value of AO satisfies:
-GA,, < O(!O-'),

(9)

which should be easy to achieve with some supersymmetry value of A wilI be certainly compatible with experimental

protection, bounds!

the resniting

In case the reader might be worried that exactly action

about our derivation

of es.(S), we mention action by a local

the same result follows from replacing the introduction of a wormhole-like

the non-local auxiliary

through

variable

a,, [3]. One

finds that the integral cosmological constant

over 010is dominated

by a saddle point at which the effective

is precisely given by eq.(g). feature of the model discussed here is the existence of a

The most appeaiing phase transition determination eq.(9).

at A0 = 0. The magnitude of the gravitational vacuum,

of lAoI does not play any r6le in the as long as GlAo/ is small enough, scenario, in which see

One can readily

envisage an inflationary

the transition

from the expanding

universe to flat space-time is triggered

by the change of the sign sector.

of A0 due to the rearrangement 3. Real life is harder

of the vacuum state in the matter

We shall now proceed with a bona-fide review existing

attempt

at computing them.

rl.

We shall first

results and then try to go beyond

However, even before preto be fuM.Ued in order for

senting any calculations,

we stress three basic requirements and useful. It is important: gravity

the result to be both reliable i) to have a finite

theory of quantum

which preserves general mu&once; action;

ii) to work with an ofi-~-shelf background iii) to maintain

field effective

general covariance even of-shell. proof for point i), but, as we stressed, other regw of

String theory is an existence

lators, if they exist, can be used. As for point ii),>& comes from the very definition

-6-

the cosmological

constant.

It is well known that errors can be made if classical equaof the effective action. An example

tions are inserted too early into the calculation was recently provided by Duff [12]. Another

one, more relevant for our present probmuitipliea the curvature scalar or

lem, concerns terms in r1 in which the logarithm its square. If such terms are rewritten

using the tree-level equation eq&tions,

R,

= Aogw, one

ends up with the wrong one-loop-corrected requirement iii) is almost obvious

as it is easy to check. Finally answers; it bar

if one wants to get unambiguous gravity literature

been widely discussed in the quantum not 60 well known by the particle the crucial point is that,

[11,13,14], but is apparently As stressed by Vilkovisky background [13),

physics community.

if one does not correct the usual

field e&c(i.e.

tive action, different by first variations

gauges provide

Bnawers that differ by equations In order to determine

of motion

of the classical action). points,

the new, quantum action making do not vanish

corrected, stationary r-t&s

however, one hau to vary

the off-shell which

obtained in different

gauges differ by second variations

even on-shell. The requirements called Vilkovisky ii) and iii) can be satisfied provided (VD) background that one computes the so.

- Dewitt

field effective action

[13,14] (or unique radii f. As

effective action) rvD. We shall consider spherical backgrounds we shall discuss later on, this will be sufficient of quantum corrections

of arbitrary

to argue about the volume dependence to rw haa the

in the general ease. The one-loop correction

formal expression:

r~ = i log det{ P(S + S,, + 5%) / (6h) ) where S is the classical action and S, the ghost action. taken with respect to the metric fluctuations

log det{ SS, / 6~6~ } , The functional derivatives

(10) are

h ,-and

the ghost fields v,,. In eq.(lO),

-7-

S,, is the gauge iixing

term taken to be of the form: S(i, = & dz&V+-i;,,Vjl,,

(11)

The corresponding

ghost action is:

s, = &
Finally, S, is the additional

/ d=&iv;(term prescribed

0 - 3/+. by VD. Fortunately, for the case of

spherical backgrounds simply end Ill]. by omitting

under consideration, the VD term in eq.(lO)

the one-loop

VD action can be obtained the limit a -+ 0 at the

and by taking

The computation

of r1 is presented in detail in ref.[lS]. of a couple of important technical points.

Here, we restrict

ourselves to the discussion

It is well known that some of the metric fluctuations sign [16] to the quadratic rotation is performed part of the action;

contribute

with the wrong Wick

in these cases, the appropriate to the functional integral positive.

in order to give meting

or, if we Special

wish, in order to keep the eigenvalues of the kinetic energy operators care has to be taken for the zero eigenvalucs. those which arise from a symmetry

There are two types of zero modes: and that, as such, do not depend are in a

of the problem

on the values of r or A,,, and those which occur when these two quantities certain relationship. just to be taken incomplete It is now widely

accepted that the first type of zero modes have to residual gauge transformations, i.e. to

out since they correspond

gauge fixing.

On the other hand, it is clear that the second type of zero

modes cannot (and should not) be subtracted. The determinants from arbitrarily
dealt

involved

in eq.(lO) contain ultra-violet

(UV) divergences coming This problem can be

large eigenvalues of the kinetic energy operators. String &eory

with in several ways and, certainly,

must have its own. We shall

assume that the way String Theory proper time t integration the determinant. encountered

works is to introduce

a lower cut-off X2 on the !I?] for evaluating

in the heat kemei method

There are strong indications

that the physical results discussed here Most of the existing results cm heat

do not depend on this choice of regularization. kernels concern their small t behaviour, dependence of r,. This is not sufficient, actual dependence

which is sufficient

to determine the UV cut-off

however, for our purposes action

since we are interested

in the

of the effective

ora the radius

7. In order to have some (IR), behaviour of

handle on that, beat kernels.

we need to study also the large t, or i&a-red is controlled

The IR behaviour

by the lowest eigenvalues, [IS]. Their

which have results lend by global,

been the subject of many investigations support non-local

by mathematicians

to the idea that the IR behaviour properties of the manifold,

of hest kernels is controlled

such as the diameter

of the largest ball that can

be inscribed

in the manifold.

This non-locality

of the effective action was pointed

out long ago by Dewitt

[19] and, later on, discussed to some extent by Vilkovisky what we want in order to play

[13]. This feature of the effective action is precisely the game described earlier for the Vlog V toy model. The final result [15] for the one-loop ViUrovisky A&-

- Dewitt

effective action is:

r1

= 2 [-&A--

+ 2h,? log(XM)]

f P [y A- - 16A&(&%4)] where: Id= The following remarks are in order: =4 lAoI ll,l/r }

+ O(logr),

(13)

(14)

-9-

i) One has generated, to the cut-off. cosmoiogical

as expected,

a large

cosmological

constant

proportional

This contradicts constant.

the assumption

that

Aa was the Ltree-level assumed that some must

In other words, we have implicitly protection

(supersymmetric,

for instance)

is at work. The same protection to ,I\- in eq.(13).

necessarily wash out the term proportional ii) There is a finite renormalization

of G: a small one if X > X,, and one proporThe latter could be relevant if large logarithms

tional to & times the logarithm. are present. iii)

There is a pure log T term corresponding

to a renormalization

of R-type

terms,

which will play no r6le in the search for large volume solutions. iv) Finally, infinite there are logarithms (with just a constant in front) that can become

at special values of r and Ao (i.e. when a zero mode occurs), which we for soiutions far from these singularities. we shall now interpret

neglect since we will be looking

In accordance with the general discussion we made earlier, the factors r occurring manifold which controls under the logarithms

as V1f4, since it is the global size of the For X 2~ X,, the term dominant at

the small eigenvalues.

large V in the effective action of eq.(13) is:

rl

Ao Vlog(XM).

We notice that PI falls just short of giving our toy models action of eq.(3). The two main differences are:

a) The argument volume,

of the logarithm

which multiplies

A,, V is not a power of the unfortunately, pre-

but M z max{ IAol*,

I-

}, se: _eq.( 14). This,

-lO-

vents the logarithm of a logarithm

from becoming

very large.

On the other hand, the scale calculation. Higher

is often difficult

to determine

by a one-loop

loops might change the scale and replace it, for instance, where the final A CI V-i b) The sign of the logarithm toy model. Actually, that,

by max{ Al/l,

V-W ),

appears. This would be n&dent under discussion is the opposite

for our purposes. of the one of the one.

As such, it would rather

af%ct a positive

Ao than a negative

after an easy graphical

check [similar

to the one of fig.(l)], and that the one with

one finda A z A0

for positive

Aor there are two solutions This conclusion

has lower action. conformal

depends on the integration

contours over the

factor of the metric and could be reversed if some reason forces one

to change it.

We thus conclude that our toy model almost comes out of a bona-fide one-loop calculation. It is not at all excluded and realistic that higher loop effects or the use of a more complicated broken supersymmetry, for instance) could just

theory (with dynamically Unfortunately

lead to our toy model. either case.

we do not know, at present, how to approach

One can also take II different

attitude

towards our toy model and regard S + AS theory. constant In superterm comes [ZO]

of eqs.(Z) and (3) as the effective string scenarios, for instance, together with (local) and in the gravitino persymmetry and contain breaking logarithms

action of B more fundamental

the appearance of a cosmological breaking.

supersymmetry condensation

Both in the gluino

condensation

[21] scenarios the effective

actions

below the su[22]

scale resemble those of supersymmetric of fields induding

gauge theories

those related to the size of various spaces could yieid 8 correction proportional to AS.

[9,23]. In principle,

such a mechanism

-12-

References
[I] E. Bum, Phys. Lett. 133B (1983) 185. Phys. Lett. 134B (1984) 403.

[2] S.W. Hawking,

[3] S. Coleman, Nucl. Phys. B310 (1988) 643.. [4] T. Banks, Nuci. Phys. B309 (1988) 493. [.5] S.W. Hawking, S.B. Giddings [6] G. Veneziano, [7] W.G. stant, Unruh, Phys. Rev. D37 (1988) 904; and A. Strominger, Nud. Phys. B306 (1988) 890.

Mod. Phys. Lett. A4 (1989) 695. Quantum Coherence, preprint Fii Wormholes, NSF-ITP-88-168 The Constants and the Cosmological (1988); Of Nature?, talk given at Con-

ITP Santa Barbara

S.W. Hawking, Fermiiab

Do Wormholes (1989).

Wormshop

[8] I. Antoniadis, I. Antoniadis preprint

J. Iliopoulos and E. Mottola,

and T.N. Tomaraa, Phys. Rev. Lett. 56 (1986) 1319; Graviton (1986); Nud. Phys. B278 (1986) 694; Fluctuations in DeSitter Space, CERN

CERN-TH-4605/86

P. Mazur and E. Mottola,

B. AlIen, Phys. Rev. D34 (1986) 3670; B. Allen and M. Turyn, Nucl. Phys. B292 (1987) 813. Phys. Lett. B212 (1988) 147.

[9] T.R. Taylor and G. Venetian+ [lo] S.M. Christensen E.S. Fradkin

and M.J. Duff, Nud. Phys. B170 (FSI] (1980) 480; Nucl. Phys..B201 (1982) 469.

and A.A. Tseytlin,

-13-

[II]

E.S. Fradkin

and A.A. Tseytiin,

Nucl. Phys. B234 (1984) 509. is possibly zero, but the proof is probably

[12] M.J. Duff, The cosmoiogical wrong, Texas A&M in: preprint Quantum

constant

CTP TAMU Theory

16-89 (1989). ed. SM. Christensen (Ad-

[I31 G. Vilkovisky, Hilger, Bristol,

of Gravity,

1984) p. 169; Nucl. Phys. B234 (1984) 125. in: Architecture of Fundamental ed. P. Ramond Interactions at Short Distances, Science Pub.

1141 B.S. Dewitt,

Les Houches Session XLIV, Co., 1987), p. 1023. [15] T.R. Taylor and

and R. Store. (Elsevier

G. Veneziano, (1989).

Quantum

Gravity

at Large

Distances,

CERN/Fermilab [16] G.W. Gibbons, J. Polchinski,

preprint

S.W. Hawking

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

Phys. Lett. B219 (1989) 251. iMath. Phys. 64 (1979) 233. Geometry, ed. S.-T. Yau (Princeton University Press,

[17] A.S. Schwarz, Commun. [la] Seminar on Differential Princeton, 1191 B.S. Dewitt, 1982).

Phys. Rev. 162 (1967) 1239. and H.P. Nilles, Phys. Lett. 125B (1983) 457; Phys. Lett. 155B (1985) 65; Phys. Lett. 156B (1985) 55. Nucl. Phys. B309 (1988) 201; Instanton-induced Super-

j20] S. Ferrara, L. Girardello J.-P. Derendinger,

L.E. Ibtiez

and H.P. Nilies,

M. Dine, R. Rohm, S. Seiberg and E. Witten, [21] K. Konishi, Generation N. Magnoli and H. Panagopoulos,

of .Mass Hierarchies

and Gravitational

-14-

symmetry

Breaking,

Univ. Geneva preprint Phys. L&t.

GEF TH 89/2 (1989);

M. Mangum [22] G. Venaiano

and M. Porrati,

B215 (1988) 317. 113B (1982) 321; Nucl. Phys. B218 (1983) 493.

and S. Yankielowie~,

Phys. L&t.

T.R. Taylor, G. Venezimo

and S. Yankielowicz,

;23] T.R. Taylor, Phys. Lett. 164B (1985) 43; G. Venaiano, unpublished (1985).

Figures
4rh.d.

(7)

------%.

A :;_/ ----

----

--

__________----------43
Fig.1) Graphical solutions of the bootstrap

--

eq.(7); the upper and lower curves cmof A in the cases of positive and

respond to the r.h.s. of eq.(7) as the functions negative Ao, respectively.

Fig.2) The effective

cosmoiogicd

constant

A a~ the function

of the tree-level

cosmo-

logical constant

Ao.

-ll-

It is straightforward mological unrelated

to repeat the self-consistent

determination

of the effective

cm and

constant in our toy model, with the coefficient to Ai. One finds that A remains positive and exponentially

of Vlog V kept positive

and large as far as A0 stays posiand large such

tive, and becomes positive compared to the coefficient a string-based

small if Ao becomes negative

of Vlog V times G. It remains to be checked whether A more radi&l at this point, alternative,

model is feasible.

which we are reluctant principle,

to take into serious consideration and postulate

is to abandon the locality action of EQG.

S + AS as the fundamental we have given arguments

claasical

In conclusion,

to support

the idea that quantum which

correc-

tions to the classical action might large Euclidean volume and/or

be accompanied

by logarithms constant.

blow up at that,

small cosmological

It is not excluded

just as in our toy models effective action,

ground state of the theory

these logarithms

drive, at late epochs, the configuration, into the Big

away from its naive, tree-level

Universe of today. We are grateful to B. Allen, W.A. Bardeen, I. Batalin, discussions. E. Fradkin, R. K&ash, A.

Lindc and T.N. Tomams for interesting