Quantum Cosmology and Open Universes

D.H. Coule
School of Mathematical Sciences,
University of Portsmouth, Mercantile House,
Hampshire Terrace, Portsmouth PO1 2EG, United Kingdom.
e-mail: [email protected]

Jérôme Martin
DARC, Observatoire de Paris-CNRS UMR 8629, 92195 Meudon Cedex, France.
e-mail: [email protected]
Quantum creation of Universes with compact spacelike sections that have curvature k either closed,
flat or open, i.e. k = ±1, 0 are studied. In the flat and open cases, the superpotential of the Wheeler
De Witt equation is significantly modified, and as a result the qualitative behaviour of a typical
wavefunction differs from the traditional closed case. In the open case boundary conditions that
arXiv:gr-qc/9905056v2 8 Oct 1999

include the Tunneling ones are allowed but not the no boundary choice. Restricting ourselves to
the Tunneling boundary condition, and applying it in turn to each of these curvatures, it is shown
that quantum cosmology actually suggests that the Universe be open, k = −1. In all cases sufficient
inflation ∼ 60 e-foldings is predicted: this is an improvement over classical measures that generally
are ambiguous as to whether inflation is certain to occur.

PACS numbers: 98.80.Cq, 98.80.Hw

I. INTRODUCTION Euclidean or “forbidden region”. Starting at zero scale

factor one can envision tunneling into the classically al-
Quantum cosmology is regarded as a possible way of lowed region. Boundary conditions need to be supplied in
obtaining the initial conditions required to start the evo- order to make predictions: the two most widely used be-
lution of a classical cosmological model, for a general in- ing the Hartle-Hawking (HH) [4] and Tunneling ones [5].
troduction, see Ref. [1]. These initial conditions corre- The Tunneling boundary condition can be formulated in
spond to a number of arbitrary constants that determine a number of ways, see [6] for a recent discussion, but we
entirely, in the absence of chaos, the future evolution of will generally refer to [7] where the Tunneling condition is
the model. In the simple Friedman-Lemaı̂tre-Robertson- defined as “outgoing modes” only. Roughly speaking the
Walker (FLRW) metric these constants determine the HH condition makes use of the Euclidean nature of space
amount of matter present, and in the case of a scalar field time to smooth out singularities while the Tunneling one
source the initial balance of kinetic energy to potential corresponds to allowing only outgoing modes, making it
energy of the scalar field. This determines how the initial analogous to quantum α decay of an atom.
expansion proceeds and whether the strong-energy con- To overcome the limitation of only having closed mod-
dition is first violated to create an inflationary expansion. els quantum creation of bubbles during any subsequent
One could hope to overcome ambiguities in the predic- inflationary phase can enable locally open regions to form
tion of whether inflation occurs that result from purely [8]. The use of these so-called Hawking-Turok instantons
classical measures of probability [2]. [9] are essentially making use of the fact that in DeSit-
The spatial curvature k is also another constant to be ter space all curvatures are equivalent. Different slicings
supplied. Although with strong-energy satisfying matter of the 5-dimensional DeSitter hyperboloid correspond to
e.g. radiation, it will only dominate at large scale fac- different curvature k when considered as a 4-dimensional
tor to either force the universe to re-collapse or to keep model, see e.g. Ref. [10]. But if the final requirement is
eternally expanding. But when the strong-energy condi- simply an open or flat universe this would seem, at best,
tion is violated the curvature instead dominates at small a rather convoluted procedure for their production. Be-
scale factors. Curvature is therefore especially important cause open universes can also be compact [11] it seems
if the universe is assumed to be created in an inflationary possible to work directly with a universe of specific cur-
state. But, because closed k = 1 models have finite size vature, instead of by necessity starting with a closed uni-
they have generally been thought to be most relevant for verse that can later by quantum tunneling create locally
quantum cosmology as they will have finite action, see open regions. This will be the subject of this paper, for a
e.g. Ref. [3]. general reference to topology in cosmology see [11]. For
The archetypal model that has been considered is the some results this requirement of compactness seems un-
closed DeSitter space that has at small scale factors a necessary as one can let the volume become arbitrarily
large. Or alternately quantize the Friedmann equation

1
for the scale factor directly where the volume factor is a It is well-known that the metric does not fix the global
redundant multiplicative factor cf. [12]. topology [11]. In this article, we will generally assume
Allowing k = −1 has a drastic effect on quantum cre- that the spacelike sections are compact (i.e. they have
ation scenarios as the forbidden region that is assumed a finite volume) with constant curvature characterized
to be tunneled through is no longer present. The clas- by k = 0, ±1. Different topologies correspond to differ-
sical singularity at zero scale factor is no longer isolated ent ranges of variation for the coordinates (r, θ, ϕ). The
from the larger universe but instead classical evolution volume of the spacelike hypersurfaces is formally given
can start from arbitrarily small size. by:
The fact that curvature plays such an important role Z √
in quantum creation scenarios is worrying since the stan- vk ≡ d3 x h, (2)
dard notions of curved space, and their associated met-
rics, are likely to require extensive modifications as the
quantum gravity epoch is approached. As a first approx- where we have written the metric as: ds2 = −N 2 (t)dt2 +
imation to including the effect of quantization of the cur- a2 (t)hij dxi dxj . Then, the Einstein-Hilbert action plus
vature, we will also consider that the curvature is initially the boundary term for this minisuperspace is given by
a quantum (q) variable that is allowed to take values the following expression:
within an ensemble. With this assumption we will argue
c3
 
6k 6 ȧ
Z
that creation is in some sense “more likely” if this initial SE−H = vk dtN a3 2 − 2 ( )2 . (3)
tunneling is not required which leads one to conclude that 16πG a N a
open universes are favoured. A somewhat related work
The matter is described by a scalar field whose action
[13] has concluded that quantum tunneling is favoured if
can be written as:
the strong-energy condition is only just being violated.
This can be understood since this corresponds to making
 
vk 1 2
Z
the barrier to be tunneled through very shallow and so Sφ = − dtN a3 − φ̇ + V (φ) . (4)
c 2N 2
easily overcome. We are taking this idea a step further
by removing the barrier entirely. In order to pass to the Hamiltonian formalism, conjugate
Because the forbidden region is absent the wave- momenta must be calculated. They are given by:
functions will be purely oscillatory. This is similar to
those wavefunctions obtained when the Wheeler-DeWitt c3 vk 12aȧ a3 vk φ̇
(WDW) equation corresponding to a classical signature πa = − , πϕ = . (5)
16πG N c N
change was solved [14]. It was further found that from
notions of regularity only a boundary condition analo- ¿From the last expressions, the canonical Hamiltonian
gous to the Tunneling boundary condition was allowed. can be deduced. It reads:
This article is organized as follows: in section II, we 
16πG πa2
describe the minisuperspace model for the case where Hc = N − 3
the spacelike sections have zero, positive or negative con- c vk 24a
stant curvature and we investigate the general behaviour c πφ2 c3 vk

vk 3
of the wavefunctions in such Universes. In section III, + − 6ka + a V (φ) . (6)
vk 2a3 16πG c
we discuss the choice of the initial state by making use of
regularity requirements . In section IV, we make physical We are now in a position where the quantization à la
predictions using the obtained wavefunctions. In partic- Dirac can be carried out cf. [1,15]. It consists in replacing
ular, we compute the probability of creating a Universe the two momenta according to the rule:
with a particular spatial curvature and show that it is
most likely to be an open Universe. The probability for ∂ p ∂ ∂2
πa2 → −h̄2 a−p (a ), πφ2 → −h̄2 , (7)
inflation to occur is further obtained for any curvature. ∂a ∂a ∂φ2
We finish with the conclusions presented in section V.
where p takes into account the factor ordering ambiguity.
The second rule is that the action of the operator Ĥc on
II. DESCRIPTION OF THE MODEL the wave function Ψ(a, φ) gives zero. This leads to the
WDW equation:
We consider the quantization of the following FLRW ∂2 p ∂ 6 ∂2
metric: 2
Ψ(a, φ) + Ψ(a, φ) − 2 2 Ψ(a, φ)
∂a a ∂a κa ∂φ
2
ds2 = −N 2 (t)dt2 36v a a
− 2 2k 2 a20 ( )2 [k − ( )2 ]Ψ(a, φ) = 0, (8)
dr2 κ h̄ c a0 a0
 
2
+a2 (t) + r 2
(dθ 2
+ sin θdϕ2
) . (1)
1 − kr2
where κ ≡ 8πG/c4 and a0 ≡ [κV (φ)/3]−1/2 . This equa-
tion is not exactly soluble in its present form so we will

2
make a further assumption. Although we still let the solutions. Introducing the variable z(a; k) defined by cf.
wave function keep a scalar field dependence we will ig- [7]:
nore the 2nd derivative term w.r.t φ in the WDW equa-
tion. This should be valid during any “slow roll” regime 3 4/3 2/3 ρPl 2/3 a
z(a; k) ≡ ( ) vk ( ) [k − ( )2 ], (11)
where the scalar potential plays the role of an effective 8π ρΛ a0
cosmological constant, i.e. a0 = (3/Λ)1/2 . As a con- the general solution can be expressed in terms of Airy
sequence we find that a20 /(κ2 h̄2 c2 ) = 3ρPl /(512π 3 lPl
2
ρΛ ) functions of first and second kind [16]:
where lPl is the Planck length and ρPl the Planck energy
density, ρPl ≡ c7 /(h̄G2 ). ρΛ is the energy density of the αAi[z(a; k)] + βBi[z(a; k)] N (a; k)
effective cosmological constant, ρΛ ≡ Λ/κ. Finally, it Ψ(a; k) = ≡ , (12)
αAi[z(0; k)] + βBi[z(0; k)] D(k)
is convenient to work with a dimensionless scale factor
expressed in units of the Planck length. Therefore we re- where the coefficients α and β are arbitrary complex
define the scale factor according to: a → lPl a. Equation numbers determined by the choice of a state for the wave
(8) can then be rewritten as: function of the Universe.
We can make some comments at this point. The pres-
d2 Ψ(a) p dΨ(a) ence of the denominator in the previous equation comes
+
da2 a da from the requirement that the wave function be regular
27vk2 ρPl a 2 a everywhere in the minisuperspace. Recall that the wave
− ( ) [k − ( )2 ]Ψ(a) = 0. (9) function depends on the scalar field and the approxima-
128π 3 ρΛ a0 a0
tion made previously was only adopted for computational
The last equation determines the superpotential: convenience. Let us therefore re-consider the full WDW
equation given by formula (8). It is convenient to define
27vk2 ρPl a 2 a the quantity α ≡ ln(a/a0 ). Then the WDW equation
U (a; k) = ( ) [k − ( )2 ]. (10)
128π 3 ρΛ a0 a0 becomes:

A related WDW equation has been obtained by using a ∂ 2 Ψ(α, φ) ∂Ψ(α, φ) 6 ∂ 2 Ψ(α, φ)
+ (p − 1) −
normalization of the scale factor to remove any volume ∂α2 ∂α κ ∂φ2
factor divergences [12]. In the following figure, the su- 2 2
27vk a0 ρPl 4α
perpotentials for the three cases k = 0, ±1 are displayed. − e [k − e2α ]Ψ(α, φ) = 0. (13)
128π 3 ρΛ
Note that for k = 0, −1, the superpotentials are always
negative. This is a crucial difference in comparison with This equation is separable. If we define Ψ(α, φ) ≡
1−p
the k = 1 case . This means that when k = 0, −1 there e 2 α f (α)g(φ) and if λ is the separation constant then
is no possibility of tunneling anymore since a zero energy the solution can be written as:
system is always above the superpotential. This will have Z √κ
1−p
important consequences which are now investigated. Ψ(α, φ) = c(λ)e 2 α fλ (α)ei 6 λφ dλ, (14)

where c(λ) are a priori arbitrary coefficients and the func-

tion fλ (α) satisfies the equation:

d2 fλ (α) (1 − p)2

2
+ λ2 −
dα 4
2 2

27vk a0 ρPl 4α 2α
− e [k − e ] fλ (α) = 0. (15)
128π 3 ρΛ
Note that when considering quantum wormhole solutions
with massless scalar fields the oscillatory divergence can
be regulated by a proper choice of the coefficients c(λ)
[17–19]. But we will rather consider single wavefunc-
tion like solutions for which such a scheme is absent [20].
When the scale factor becomes small, a ≪ a0 , Eq. (15)
FIG. 1. Superpotentials for different values of k. The full can be readily solved. The solution is fλ (α) ≈ eiωλ,p α
line represents the case k = 1, the dotted line the case k = 0 where ωλ,p can be expressed as:
and the dashed line the case k = −1. r
(1 − p)2
ωλ,p = λ2 − . (16)
In order to obtain exact solutions of the WDW equa- 4
tion we choose to work with the factor ordering given In particular when p = −1, i.e. the factor
by p = −1. This permits to work with analytical exact √ ordering
adopted in this article, one has ωλ,p=−1 = λ2 − 1. We

3
are now in a position where the regularity of the wave- should anyway include a more realistic strong-energy sat-
function in the vicinity of a = 0 can be studied. It cru- isfying matter component. As emphasized by Gott and Li
cially depends on the value of the separation constant, [24] such a matter source should be expected if only of a
see also Ref. [20] where a similar treatment has been per- size due to quantum “ zero point” fluctuations. One can
formed. If λ ∈] − 1, 1[ then the wavefunction is regular as think of this as being because the scalar potential has
the scale factor goes to zero whereas otherwise the wave- a “fuzziness” due to quantum uncertainty in the limit
function exhibits rapid oscillations. For these values of λ a → 0. For the case of radiation, given here by a pa-
the wavefunction remains finite. Nevertheless, the “wig- rameter A the WDW equation is then given by, see e.g.
gliness” of the wavefunction is also a kind of singularity [25]
and it seems reasonable not to consider this possibility .
In this paper, we restrict ourselves to the case λ = 0 for d2 Ψ(a) p dΨ(a)
+ + AΨ(a) = 0 (18)
which the wavefunction is regular at a = 0. This means da2 a da
that Ψ(a = 0, φ) does not depend on φ and it justifies the
presence of the denominator D(k) in Eq. (12), see also and now with solution,
Refs. [7]. It is clear from the previous discussion that this 1−p √ 1−p √
is not the most general case since there exists non van- Ψ(a) = c1 a 2 J p−1 ( Aa) + c2 a 2 Y p−1 ( Aa), (19)
2 2
ishing values of λ such that the wavefunction is regular
at the origin. However, to our knowledge, there does not where J(p−1)/2 and Y(p−1)/2 are Bessel functions of order
exist a general study of the boundary conditions for an (p − 1)/2 [16]. But as pointed out by [26] the Y Bessel
arbitrary value of the separation constant. Moreover, it function still diverges for p ≥ 1 and this term is part of
turns out that the two most widely discussed choices of the Tunneling boundary condition’s solution. However,
boundary conditions, i.e. Hartle-Hawking and Vilenkin the presence now of oscillatory behaviour complicates the
states, picked out the value λ = 0 as will be demonstrated adoption of the HH boundary condition. It sometimes is
below, see also Ref. [20] (in these two cases, D(k) is re- given by the prescription “outgoing” plus “ingoing”, if
sponsible for the appearance of the factors e±1/V (φ) in so it would also suffer this same divergence, although
the probability density functions). So as a first approach if it was simply the J Bessel function term it would be
it seems reasonable to consider the case λ = 0 only. Let normalizable [26]. But regardless of such considerations
us emphasize that Eq. (12) allows us to study all the for the chosen p = −1 case the no-boundary state as well
boundary conditions such that Ψ(a = 0, φ) is indepen- as the Tunneling state are immune from factor ordering
dent of the scalar field and not only the Hartle-Hawking divergences and are a priori both allowed.
and Vilenkin states. Let us now turn to the study of the behaviour of the
We should distinguish between the sort of singular- wave functions given in Eq. (12). The case k = 1 is
ity discussed above, present for example with a massless standard. We choose the spacelike sections to be spheres
scalar field case [17], which would tend to be displayed by although of course other possibilities are allowed. This
a rapid oscillation in the wavefunction as a → 0 (recall ki- means that v1 = 2π 2 . The evolution of the Universe can
netic energy ∼ “wiggliness” of wavefunction, see eg. [21]) be viewed as the motion of a fictitious particle with zero
if not regulated and those considered by Ref. [22] which energy in the potential given by Eq. (10). The particle
are related to the choice of the factor ordering. There, for starts to the left of the potential and can proceed to tun-
the closed case, it was claimed that only the no-boundary nel through the barrier. While in the region a < a0 the
state is regular in the limit a → 0 for any choice of p. It wavefunction exponentially decays. Then, when the scale
was also shown that the Tunneling solution is regular in factor is beyond the barrier a > a0 , the wavefunction be-
this limit only if p < 1. This behaviour was caused by comes oscillatory. This behaviour is illustrated in the
the wavefunction having growing and decaying, actually following figure∗ for the boundary conditions α = 1 and
Modified Bessel functions, solutions in the forbidden re- β = i, which correspond to the Tunneling wavefunction
gion. Typically the required Tunneling solution diverges [7] . The value of ρΛ is chosen such that ρPl /ρΛ = 1000.
like Ψ ∼ a(1−p)/2 in the limit a → 0. However, this sort Since we have a20 = (3/8π)ρPl /ρΛ , this corresponds to
of divergence is present universally. Solving Eq. (9) for a dimensional a0 equal to ≈ 10.9lPl. This value can be
k = Λ = 0 gives the solution thought of as being the size at which the Universe is first
created.
Ψ(a) = c1 + c2 a1−p , (17)

with c1 and c2 arbitrary constants. So the factor order-

ing divergence remains even in flat empty space. These In this article, the figures have been obtained using the
∗

divergences appear intrinsic to these models and one can Mathematica (version 3.0) and IDL software packages
conceive that they should be removed by a renormaliza-
tion scheme. It was further suggested, in the context of
quantum wormhole that such factor ordering divergences
are of no great concern [23]. In practice, as a → 0, one

4
 It can also be noticed that the oscillations in the region
a < a0 are more pronounced for the k = −1 case than
for when k = 0. This is due to the different shapes of
the superpotential. The particle can roll faster down the
steeper k = −1 superpotential compared to the k = 0
case.

FIG. 2. Real part of the Vilenkin wavefunction for k = 1

and v1 = 2π 2 . The value ρPl /ρΛ = 10 has been chosen rather
than the more realistic value ρPl /ρΛ = 1000 only for the sake
of illustration.

The cases k = 0, −1 are very different. The fictitious

particle is now always above the potential which is now
negative. Tunneling is no longer required as classical evo- FIG. 4. Real part of the Vilenkin wavefunction for k = −1,
lution is possible. As a consequence the wave function v−1 ≈ 0.94 and ρPl /ρΛ = 1000.
always exhibits oscillatory behaviour. To go further, we
need to know the topology of the spacelike sections. The The fact that the wavefunctions behave very differently
smallest three-hyperbolic manifold is not known. The for different values of k also has an impact on the choice of
two smallest spaces known are the Weeks space [27] and the quantum state. Studying this question is the purpose
the Thurston space [28]. The volume of the first one is of the next section.
≈ 0.94 and of the second one is ≈ 0.98. The results
presented here do not depend crucially on the volume of
the spacelike sections provided they are of the same or- III. MEASURE AND THE INITIAL STATE
der of magnitude. Therefore we choose to work with the
Weeks space and consequently we take v−1 ≈ 0.94. In To be able to make predictions and to calculate prob-
the flat case the volume v0 is arbitrary. For definiteness, abilities, we need a suitable measure. If one chooses a
we choose v0 = 1. The two following figures show the surface in the minisuperspace perpendicular to the “a”
wavefunction for the cases k = 0 and k = −1 respec- direction, then the component of the current associated
tively. The boundary conditions chosen are the same as with the WDW equation through this surface is given by
previously. It can be seen that the wavefunctions indeed [15]:
oscillate even before the scale factor reaches the value a0 .
i p ∗
j= a (Ψ ∂a Ψ − Ψ∂a Ψ∗ ), (20)
2
where, in our case, p = −1. As is well-known this cur-
rent is not positive definite since the signature of the
(mini)-superspace is Lorentzian. However, in the WKB
approximation, where the wave function can be written
as Ψ ∼ CeiS , this current becomes positive definite and
permits the calculation of conditional probabilities [29].
Let us first calculate the behaviour of the numerator
of Eq. (12) when the scale factor becomes large. We
have lima→+∞ z(a; k) = −∞ for any value of k. Using
the asymptotic behaviour of the Airy functions, we find:
1 
[−z(a; k)]− 4 α 2i
3
iπ
lim N (a; k) = √ ( + β)e 3 [−z(a;k)] 2 + 4
a→+∞ 2 π i
FIG. 3. Real part of the Vilenkin wavefunction for k = 0,

α − 2i
3
[−z(a;k)] 2 − iπ
v0 = 1 and ρPl /ρΛ = 1000. −( − β)e 3 4 . (21)
i

5
Let us now turn to consider the form of the denominator. the wavefunctions is in fact a sphere of unit radius. The
This time the three cases must be treated separately. Let polar coordinates (θ, ϕ) of a wave function are calculated
us start with the usual case, i.e. k = +1. The value of according to the formulas:
the function z(a; +1) when the scale factor vanishes is  
given by: −1 1
θ = 2 tan , ϕ = ψ. (29)
ρ
3 4/3 ρPl 2/3
z(0; +1) = ( ) (v1 )2/3 ( ) . (22)
8π ρΛ Therefore, the subspace of singular wavefunctions defined
by Eqns. (28) is just the great circle going through the
Semi-classical considerations are supposed to be valid north and south poles such that ϕ = 0, π. This shows
only if we are in a region where ρPl /ρΛ ≫ 1. We can that almost all the wavefunctions are regular except those
therefore work with the approximation that z(0; +1) ≫ belonging to this subspace. In the case k = 1, the situ-
1. In that case, we obtain: ation is different since all the wavefunctions are regular.
β 2 3/2
The Tunneling (Vilenkin’s) wave function is such that:
D(k = +1) ≈ √ [z(0; +1)]−1/4 e 3 [z(0;+1)] . (23)
2 π α = 1, β = i, (30)
The case k = 0 is rather simple since we have z(0; 0) = 0.
which means that its coordinates are (π/2, π/2): it is
Therefore, the denominator can be written as:
regular. On the other hand, the Hartle-Hawking state is
√ given by:
D(k = 0) = Ai(0)(α + 3β), (24)

where Ai(0) = 3−2/3 /Γ(2/3) ≈ 0.35502. Finally, we turn α = 1, β = 0, (31)

to the case k = −1. Now we have:
i.e. it is represented by the south pole of the sphere and
3 ρPl 2/3 thus belongs to the subspace of the singular wavefunc-
z(0; −1) = −( )4/3 (v−1 )2/3 ( ) . (25) tions. The divergence occurs when t = −1 (there is only
8π ρΛ
one solution since ∆ = 0), that is to say when ρΛ satisfies
Therefore, in the semi-classical regime, we have the following expression:
|z(0; −1)| ≫ 1 and z(0; −1) < 0. The presence of the mi-
nus sign turns out to be crucial. This time the asymptotic ρPl 8π 3 32π 3
expansion of the Airy functions for large negative z must = +m , (32)
ρΛ v−1 3v−1
be used contrary to the case k = 1 where the asymptotic
expansion for large positive z has been utilized. Thus, in where m is an integer. This problem is more serious that
this limit, the product D(k = −1)D∗ (k = −1) is given it might first appear because one should not think of
by: ρΛ as being strictly constant (as we have approximated
during the previous calculations). It rather slowly varies
D(k = −1)D∗ (k = −1) ≈ in time during the “slow roll” inflationary period. This
|α|2 means that ρΛ could easily pass through one of these

p (1 + ρ2 − 2ρ cos ψ)t2 + 2(1 − ρ2 )t dangerous values and so cause a divergence in the wave
2π(1 + t2 ) −z(0; −1)
 function. Note that this would be across any chosen a =
2
+1 + ρ + 2ρ cos ψ , (26) const surface, where the semi-classical analysis should be
valid, and so more serious that the previously mentioned
factor-ordering type divergences that can occur as a → 0.
where t ≡ tan{(2/3)[−z(0; −1)]3/2} and β/α ≡ ρeiψ . It is also in addition to the problems of simply having
There is now a danger of obtaining a divergence in the real wavefunctions, that do not allow the current j to
current (see the following section) when the product be interpreted directly [12]. This divergent behaviour
D(k = −1)D∗ (k = −1) vanishes. The discriminant of is like that found in the context of the quantization of
the second order polynomial in t in Eq. (26) is given by: spacetimes which admit a classical change of signature
[14].
∆ = −16ρ2 sin2 ψ. (27) To know if a divergent value is actually passed through
Therefore, the polynomial has no real roots except when requires a knowledge of the volume v−1 . For m = 0,
∆ = 0. This corresponds to the following cases: we have ρPl /ρΛ ≈ 248/v−1. Depending on the value of
v−1 it could turn out that we are no longer in a regime
ρ = 0, and/or ψ = nπ, (28) where the semi-classical approximation is valid and so
the danger could be ignored as being outside the range
where n is an integer. Each wavefunction in the min- of validity. But for bigger values of m, we certainly will be
isuperspace is characterized by the numbers (ρ, ψ) and within this region of semi-classical validity. For example,
thus belongs to a two-dimensional space. It has been if v−1 ≈ 0.94 and m = 3, one has ρPl /ρΛ ≈ 1316.
shown by Gibbons and Grishchuk [30] that this space of

6
 Therefore, it seems that the Tunneling wave function [15]. Everything is now present and the Euclidean action
can easily be generalized to describe the quantum cre- can be calculated along this solution. Then the Hartle-
ation of hyperbolic Universes with compact spacelike sec- Hawking wavefunction is given by ΨHH (a, φ) ≈ exp(−SE )
tion whereas the Hartle Hawking wavefunction leads to (we have dropped the bars in order to avoid cumbersome
important difficulties. It is interesting to see that con- notation) and can be expressed as:
sidering the quantum creation of compact Universes can 3/2 
lead to some progress in the debate Tunneling vs no
  
6v1 κV
boundary. In the following section, we will consider that ΨHH (a, φ) ≈ exp 2 1 − 1 − a2 . (40)
κ cV 3
the wavefunction is placed in the Tunneling or Vilenkin’s
quantum state. In the region where a ≪ a0 the previous wavefunction
But before ending this section, we would like to discuss can be written as ΨHH (a, φ) ≈ exp[(3v1 a2 )/(κc)] and we
the important case of the Hartle-Hawking wavefunction see that this is indeed independent of the scalar field and
in more detail. We will concentrate on the case k = −1 therefore that we have picked out the value λ = 0.
versus k = 1. The first step of the derivation consists Let us now turn to the case k = −1. The solution to
in establishing the Euclidean equations of motion. With the algebraic equation giving the lapse function can be
the help of Eqns. (3) and (4), it is easy to see that in the written as:
gauge Ṅ = 0 they read: r  r 
3 −1 κV
Nk=−1 = i sinh ā . (41)
a′′ κ N2 κV 3
= − φ′2 − κV (φ), (33)
a 3 3
a′ ∂V (φ) Interestingly enough the solution is now unique. The ex-
φ′′ + 3 φ′ − N 2 = 0, (34) plicit expression of the scale factor can be easily deduced.
a ∂φ It reads:
a′2 κ N2 κN 2
2
− φ′2 − k 2 + V (φ) = 0, (35) r
3
  r
κV
 
a 6 a 3 a(τ ) = sinh sinh −1
ā τ . (42)
κV 3
where a prime denotes derivative with respect to τ ≡ −it.
The hypersurface on which we evaluate the wavefunction In the same manner, we can evaluate the Euclidean ac-
is the hypersurface such that τ = 1 for which we have: tion and therefore find the corresponding no boundary
a(τ = 1) ≡ ā and φ(τ = 1) ≡ φ̄. The no boundary state. One obtains:
conditions are [15]:    3/2 
6v−1 κV
dφ ΨHH (a, φ) ≈ exp i 2 1 − 1 + a2 , (43)
a(τ = 0) = 0, (τ = 0) = 0. (36) κ cV 3
dτ
where we have again suppressed the bars. In the limit
In the slow roll regime, the solutions satisfying all the where the scale factor goes to zero, this wavefunction
boundary conditions are found by integrating the Eu- can be written as ΨHH (a, φ) ≈ exp[−(3iv−1 a2 )/(κc)]. It
clidean equations of motion. They are given by: confirms that we have picked out a vanishing separation
 r  constant. This also shows that in the Euclidean region,
ā κV the wavefunction is oscillatory. One could easily find the
a(τ ) =  q  sin N τ , φ(τ ) = φ̄. (37)
3 no boundary wavefunction in the classical region with the
sin N κV 3
WKB matching method. Of course, we would recover the
fact that the denominator is now a trigonometric function
The value of k does not appear explicitly in these equa- which was the main reason for arguing that the Vilenkin
tions but is in fact hidden in the algebraic equation sat- state is preferable. In fact we can go a step further and
isfied by the lapse function: say that the very concept of a no boundary wavefunction
 r is in danger by this kind of analysis. Indeed, from Eq.
κā2 V

2 κV (35) we see that the no boundary conditions (36) imply
sin N = . (38)
3 3k that:
 2
At this point, one should consider the two cases sepa- 1 da
rately. The case k = 1 is standard. The solution of the = k. (44)
N 2 dτ
previous equation can be expressed as:
r   r  In order for the metric ds2 = N 2 dτ 2 + a2 (τ )dΩ23 to be
± 3 1 −1 κV Euclidean and regular at τ = 0 we need a(τ ) ≈ N τ
Nn,k=1 = (n + )π ± cos ā . (39)
κV 2 3 which implies k = +1! In the case k = −1, it is easy to
see that we have in fact a(τ ) ≈ −iN τ . This means that
It is common to consider only the case n = 0 together ds2 ≈ N 2 (dτ 2 − τ 2 dΩ23 ). The corresponding manifold is
with the minus sign, for a fuller discussion see Ref. no longer Euclidean. In addition, the metric found above

7
is very reminiscent of that used by Hawking and Turok expression of the current in the case k = 0 can be estab-
[9] in their analysis. Therefore it seems that there exists lished from Eq. (24). It reads:
an interesting link between the approach advocated here
and the one of Ref. [9]. This requires further studies 1 3 2/3 ρΛ 1/3
j(k = 0) = ( )1/3 v0 ( ) . (47)
which are beyond the scope of the present article. 4πAi2 (0) π ρPl
To end this section, let us come back to the question
Since D(k = 0) is just a constant, this does change the
of the regularity of the wave function. We have just
dependence in (ρPl /ρΛ )−1/3 of Eq. (46). Finally, we
shown that the requirement |Ψ(a; k)| < ∞ everywhere
turn to the case k = −1. From Eq. (26), we have that
in the minisuperspace and for any k favours the Tun-
D(k = −1)D∗ (k = −1) = (1/π)[−z(0; −1)]−1/2. This
neling wavefunction over the Hartle Hawking state. On
time no exponential function appears as was the case for
the other hand requiring that the wavefunction be regu-
k = 1. This is due to the fact that we have chosen the
lar as a → 0 for every factor ordering leads to a Hartle
boundary condition that D(k = −1) simply be a phase.
Hawking state for the k = 1 case [22]. Therefore, the two
However, now for the case k = −1, the factor ρPl /ρΛ
requirements are strictly not compatible if the analysis
simply cancels out. As a consequence, one finds:
was extended to p > 1. However, we have previously
mentioned that in more realistic models the actual pres- 3v−1
ence of non-inflationary matter could also complicate the j(k = −1) = . (48)
4π
adoption of the HH wavefunction.
Regardless of these complications, and because factor The current turns out to be independent of ρΛ .
ordering problems, which occur in the unrealistic limit As already mentioned, the current can now be used
a → 0, appear less serious than the divergences found in to calculate conditional probabilities. For example, the
this section for k = −1 models, we will next consider the probability of having an initial value φi of the scalar field,
predictions with these Tunneling wavefunctions. greater than the value needed to solve the problems of
standard cosmology, φsuf , knowing that 0 < φ < φsup
can be allowed. In the last inequality φsup is the value
IV. PREDICTIONS at which semi-classical considerations cease to be valid,
i.e. when the potential reaches the Planck scale, ρPl =
ρΛ = V (φ) = m4Pl in the Planck system of units. Also,
In this section, we address the problem of computing
the fact that the minimal value for φ is zero is not a prob-
physical predictions from the previously obtained wave-
lem here because the wave function is the Tunneling one.
functions. It is well known that, due to the fact of ob-
This would no longer be true if the state were the HH
taining non-normalizable wavefunctions, this is a difficult
one. We will evaluate the conditional probabilities for
task. In Ref. [15], a method to overcome this problem has
the prototype chaotic inflationary scenario, with scalar
been proposed. The idea is to use the current defined by
potential of the form V (φ) = (λ/4!)φ4 . In this context,
Eq. (20) in the WKB regime. This leads to well-defined
the initial
p value of the field necessary to get N e-folds
probabilities. Using the asymptotic form of the wave-
is φi = (N + 1)/πmPl . If we consider that sufficient
function given by Eq. (21), the following expression for
inflation is obtained when N = 60 then we find that the
the current, valid for any k, can be found:
scalar field has to start at φi ≡ φsuf = 4.4mPl . The value
2/3 of φsup is given by φsup = 241/4 λ−1/4 mPl . For chaotic in-
2 3 1/3 vk ρPl −1/3
j(k) = ( ) ∗
( ) . (45) flation, one has λ ∼ 10−15 in order to reproduce the value
π 8π D(k)D (k) ρΛ of the quadrupole of the Cosmic Microwave Background
Radiation anisotropy, Qrms−PS , measured by the COBE
As expected this expression no longer depends on the
satellite, i.e. Qrms−PS ≈ 18 × 10−6 K [31]. This implies
scale factor. Going further requires a knowledge of D(k)
φsup ≈ 1.2 × 104 mPl . The probability to have sufficient
so we must now treat the three cases separately. Let us
inflation in these three cases is computed according to
first consider the case k = 1. The expression of D(k = 1)
the formula [15]:
can be deduced from Eq. (23). This leads to:
R φsup
3v1 − 3v12 ρρPl φ j(k)dφ
j(k = +1) = e 16π Λ . (46) P (k; φi > φsuf |0 < φi < φsup ) ≡ P (k) = R φsuf .
π sup
j(k)dφ
0
We recover the well-known expression for the closed case. (49)
The coefficient of proportionality is not of great interest
since the probability distribution is not normalizable and It will be more convenient in the following to rewrite
instead must be used to calculate conditional probabil- the previous definition of the conditional probabilities as
ities, see below. The ratio ρPl /ρΛ only appears in the P (k) = 1 − R(k; N, λ) where R(k; N, λ) is given by:
argument of the exponential function. This is because R φsuf
the factor [z(0; +1)]−1/2 in the D(k = 1)D∗ (k = 1) term j(k)dφ
R(k; N, λ) ≡ R 0φsup (50)
cancels exactly the term (ρPl /ρΛ )−1/3 in Eq. (45). The j(k)dφ
0

8
We are now going to calculate R in the three cases. probability is closer to one in the case k = 1 than in the
Let us start with the case k = 1. The current j(1) cases k = 0, −1.
is given by Eq. (46). Using the change of variable We can pursue this reasoning a step further by also al-
u ≡ 9m4Pl /(λφ4 ) and the formula (3.381.6) of Ref. [32] lowing k to be quantum (q) variable. Such a possibility
one can easily show that the denominator is given by could be a first step in modeling quantum fluctuations
(24/9)5/8 e−3/16 W− 85 ,− 81 (9/24) where Wµ,ν is a Whit- in the geometry. Unlike the classical case where the cur-
taker function. Since W− 58 ,− 18 (9/24) ≈ 0.52 we find that vature is simply a fixed (c) number constant we are now
the denominator is approximately equal to 0.80. In the assuming that the curvature also is in a quantum en-
same manner, the numerator is equal to: semble. The conditional probability for having a given k
knowing that k = 0, ±1 can be calculated for a fixed value

9π 2
−5/8
− 9π 2

9π 2
 of the scale factor. This probability is formally defined
e 2λ(N +1)2 W− 85 ,− 81 . (51) according to the equation:
λ(N + 1)2 λ(N + 1)2
|Ψ(a; k)|2
Since the parameter λ appears at the denominator of the P (a; k) ≡ P 2
. (56)
argument of the Whittaker function, this has a very large l=0,±1 |Ψ(a; l)|

value. Therefore, we can use the asymptotic expansion The use of this definition requires some comments. Let
of the Whittaker function for large values of its argument us first recall how the definition of Eq. (49) can be jus-
given by [32]: tified. Let M be the minisuperspace and MWKB ∈ M
  the region of the minisuperspace where the wavefunction
z 1 can be well approximated by the WKB wavefunction. In
lim Wµ,ν (z) = e− 2 z µ 1 + O( ) . (52)
|z|→∞ z non relativistic quantum mechanics, the time component
of the current, |Ψ|2 , gives the probability density func-
Thus, the value of the function R(1; N, λ) can be ex- tion. Since the wavefunction is normalizable, the proba-
pressed as: bility of finding a particle in the interval [a, b] is given by
Rb R∞
5 5 9π
− λ(N
2
a
dx|Ψ|2 / −∞ dx|Ψ|2 . In quantum cosmology, the cur-
R(1; N, λ) ≈ 0.0045λ 4 (N + 1) 2 e +1)2 . (53) rent is positive definite only in the WKBR regime. In this
regime, Ψ(a; k) is not normalizable, i.e. MWKB dq α |Ψ|2 ,
This expression is valid for any value of N and λ pro- where dq α is the
vided that λ is a small number. Putting N = 60 and R volume in M, is not finite. This does
13 not mean that M µ(q α )dq α |Ψ|2 is infinite since the mea-
λ = 10−15 , we find that R(1; 60, 10−15) ≈ 10−10 , an sure µ(q α ) is a priori not known. In the WKB regime, the
extremely small number which has its origin in the pres- wavefunction Ris by definition peaked
ence of the parameter λ in the exponential factor. We R over MWKB . There-
fore, one has M µ(q α )dq α |Ψ|2 ≈ MWKB dq α |Ψ|2 . Thus,
conclude that P (1) ≈ 1. The calculation of P (0) is eas- the same rule as in ordinary quantum mechanics says
ier. Using Eq. (47), we find that: that the probability of Rfinding the system in the region
R ∈ MWKB is given by R dq α |Ψ|2 / MWKB dq α |Ψ|2 . Eq.
R
  67
7
− 24 N +1 7
(49) is a special case of this more general formula.
R(0; N, λ) = 24 λ 12 . (54)
π The interpretation of Eq. (56) is roughly the same.
Although it is, of course, more contentious to apply this
This gives R(0; 60, 10−15) ≈ 1.5 × 10−9 leading to P (0) ≈ reasoning to obtain the geometry, here represented by
1 again. Finally, the case k = −1 is straightforward. k, compared to how we previously obtained the initial
Using Eq. (48), we can easily establish that: matter component φ. The matter calculation is a fairly
  21 straightforward adaptation, as stated above, of usual
1 N +1 1 quantum mechanics reasoning whereas the quantization
R(−1; N, λ) = 24− 4 λ4 . (55)
π of the geometry might require more extensive alterations
to quantum mechanics. We will proceed with the no-
This results in R(−1; 60, 10−15) ≈ 1.67 × 10−4 . There- tion that k is not initially fixed but is rather undefined
fore, one can say that P (−1) ≈ 1. The conclusion is that, in a quantum state with “equipartition” among all pos-
in the three cases, the probability turns out to be close to sible k. In the context of the histories approach of quan-
one. This means that sufficient inflation is a prediction tum mechanics, P (a; k) represents the probability that
of the Tunneling wavefunction whatever the value of k the Universe “choose” one of these possible histories. It
is. In this respect, the three cases are equally compatible will not come as a surprise that the Universe goes down
with there being a near definite prediction of inflation oc- the lowest potential case.
curring. This is actually an significant improvement over We would like to emphasize how P (a; k) differs from
classical notions of whether inflation will occur. Which, the notion of the probability of a change of topology once
due to an infinite divergence over an arbitrary scale fac- the Universe has been created, i.e. the probability of hav-
tor, gives an ambiguous prediction even for apparently ing, for example, k = 1 for some value of the scale factor
inflationary potentials [2]. We note however than the and then k = 0 for another (bigger) value. If topology

9
changes are required from one classical model to another
then an explicit time dependent curvature should be in-
troduced. Such a construction with dk/dt 6= 0 has been
explicitly made in Ref. [33] and it has been shown that,
in this case, a passage to a midi-superspace description
is mandatory. However in the present example we are
assuming that the curvature is first a quantum variable
and so not fixed in a particular classical state. What clas-
sical curvature state the universe first evolves to is our
present concern, not whether topology changes still occur
once this classical state is first achieved. As the universe
becomes increasingly classical, by for example gaining en-
ergy from falling down the potential, then the curvature
will no longer be in a quantum superposition but will be-
come increasingly in a specific classical state. How this
“measurement” takes place will be a rather complex pro- FIG. 6. Probabilities P (k) for ρPl /ρΛ = 100. The solid line
cess and likely to depend on quantum mechanical inter- represents the case k = 1, the dotted line is the k = 0 case
pretational questions. But for our present purposes, this and dashed line is the k = −1 case.
rough notion that the curvature will eventually “crystal-
lize out” into a classical state should suffice. Finally, the third and last figure represents the case
On the following figures, the evolution of the three ρPl /ρΛ = 1000 which corresponds to a0 ≈ 10.9lPl as al-
probabilities as a function of the scale factor are dis- ready mentioned. This case could be considered as the
played. The first figure is for the case ρPl /ρΛ = 10. most realistic one since inflation takes place at an energy
This means that the dimensional quantity a0 is equal to comparable to the GUT scale, i.e. ρΛ ≈ 1016 GeV.
a0 ≈ lPl . Let us also recall that we have taken v1 = 2π 2 ,
v0 = 1 and v−1 ≈ 0.94. The choice of the volume can in-
fluence the behaviour of the probabilities for small scale
factors but in the limit of big scale factors, they are
mainly determined by the ratio ρPl /ρΛ .

FIG. 7. Probabilities P (k) for ρPl /ρΛ = 1000. The solid

line represents the case k = 1, the dotted line is the k = 0
case and dashed line is the k = −1 case.

Let us now comment on these figures in more detail.

FIG. 5. Probabilities P (k) for ρPl /ρΛ = 10. The solid line In each case, we can point out the following features. At
represents the case k = 1, the dotted line is the k = 0 case vanishing scale factor, the probabilities are all assumed
and dashed line is the k = −1 case. equal to 1/3 for any value of k. Provided the probabili-
ties are roughly equal the actual values will not affect the
The second figure represents the case where ρPl /ρΛ = predictions significantly. In the region where the scale
100. This corresponds to a0 ≈ 3.5lPl. factor is still of order a0 , the behaviour of the probabili-
ties are rapidly evolving with a. Finally, when the scale
factor becomes large in comparison with a0 , the proba-
bilities tend to a constant value which depends on the
ratio ρPl /ρΛ . This suggests some interesting ideas. For
example, the fact that the probabilities remain similar in
the region a < a0 could mean that here the topology can
easily fluctuate due to quantum effects. On the other

10
hand, when the scale factor is such that a ≫ a0 , the the drastically different change in the behaviour of mat-
probabilities differ significantly and are almost constant ter: from being a cosmological constant to becoming like
as a function of a. This can be easily understood if one dust. Any small residual cosmological constant would
looks at the asymptotic behaviour of |Ψ(a; k)|2 when a also come to dominate at later times. Recent develop-
becomes large. Using Eq. (12), one finds that: ment, although still contentious, have suggested such a Λ
term is necessary to explain the Supernova data [34], but
4 3 1/2 ρPl 1/2 − 3v12 ρρPl such a value is still extremely small ρΛ ∼ 10−120 . Under-
|Ψ(a; k = +1)|2 ≈ ( ) ( ) e 16π Λ , (57)
a 8π ρΛ standing such fine detail while quantum cosmology takes
1 8π 1 ρPl 1/6 a rather “broad brush” approach to calculating the vari-
|Ψ(a; k = 0)|2 ≈ ( )1/6 1/3 ( ) , (58) ous quantities remains a difficulty. We would just remind
4πAi2 (0) 3 av ρΛ
0
readers that the wormhole, and related mechanisms, that
1 3 1/2 ρPl 1/2
|Ψ(a; k = −1)|2 ≈ ( ) ( ) . (59) were suggested should predict Λ = 0 exactly [35], could
a 8π ρΛ instead give other values which are only approximately
zero -see e.g. [36].
In each case |Ψ(a; k)|2 behaves as 1/a and therefore
P (a; k) becomes independent of the scale factor when
a ≫ a0 . Fluctuations in topology are more likely, re- V. CONCLUSIONS
gardless of the scale factor, provided the energy density
is near Planck values. But as the value of ρΛ reduces this
rapidly becomes less likely. When ρPl /ρΛ ≫ 1, it is clear In general relativity the global topology has to be im-
from the figures that a definite prediction can be made posed as an initial condition. In quantum cosmology
since one of the probabilities becomes equal to one: most studies have concentrated on elliptic space which for
the FLRW model is simply a closed universe. However,
P (−1) ≈ 1, P (0) ≈ 0, P (+1) ≈ 0. (60) hyperbolic and flat spaces are mathematically speaking
more numerous and many are also compact. These give
The fact that the greater the ratio ρPl /ρΛ is, the better rise to geometries that are locally described by the FLRW
the prediction is [i.e. the closer to one P (−1) is] means metric with k = −1, 0. For the simple DeSitter model the
that when inflation takes place at energy well below the k = −1, 0 cases are no longer distinguished by the pres-
Planck scale (typically the GUT scale seems to be the ence of a forbidden or Euclidean region at small scale
most physical case), topology changes become strictly factors.
forbidden as the scale factor becomes large. In practice Because the Euclidean nature of the model is now ab-
as the scalar field rolls down the potential the effective sent it might seem that the smooth geometric picture
cosmological constant ρΛ reduces and topology changes of the Hartle-Hawking no-boundary proposal is a serious
become increasingly unlikely. loss. But a major weakness of “quantum creation of the
Finally, let us comment on the prediction itself. It universe” ideas is that the forbidden region is anyway
says, since an inflationary phase is expected to give a fi- eroded by the presence of strong-energy satisfying mat-
>
nal a ∼ 1030 lPl , that our Universe is most likely to be ter. Such matter will generally prevent the Euclidean na-
open, or at worse flat, a quite interesting statement in- ture of the model for small scale factors. But on general
deed. Bearing in mind that we have not tried to allow for grounds such matter should be present,if only because of
the more numerous hyperbolic topologies corresponding “zero-point” fluctuations. A Lorentzian region will tend
to k = −1, we have probably still underestimated this to form anyway as a → 0.
overwhelming probability of obtaining an open universe. With a scalar field source a possible problem still re-
However, it does not say that Ω0 is necessarily much less mains since a singularity caused by the kinetic energy
than unity and so could actually be distinguished from of the scalar field blowing up could occur. The kinetic
Ω0 = 1 by observations. This is because we have seen energy of the scalar field behaves as a ‘stiff’ equation of
that even in the open case, sufficient inflation is a predic- state. This causes an effective ∼ −a−2 term in the WDW
tion of the Tunneling wavefunction. Thus, inflation will potential that would tend to create an infinite ‘wiggli-
probably drive Ωini < 1 to a value very close to one even ness’ in the wavefunction as a → 0. By ensuring that the
at the present age of the universe. The actual form of wavefunction be independent of the matter as a → 0 this
the scalar potential is crucial for determining such prop- possible problem is avoided. But one must bear in mind
erties, a quantum description alone does not mandate that one is removing the singularity by fiat, one should
such values for Ω0 . Once the inflationary phase finishes not claim that the process of quantization alone is achiev-
the matter behaves effectively like that of dust [2] so now ing this, as many studies erroneously assume. Likewise
obeying the strong-energy condition. Eventually the cur- in the usual closed models the Hartle Hawking boundary
vature term will start again to dominate the dynamics, condition only gives a Euclidean region by imposing that
with the open model expanding infinitely into the fu- the matter fields are not allowed to dominate as a → 0.
ture. Curvature can play an important role at both the Once it is accepted that a Lorentzian region is any-
beginning and end of the universe’s evolution because of way present for the smallest scale factors it is no longer

11
a major fundamental difference whether one then has a mechanics should be used to predict. We have assumed
forbidden region away from the origin. In the flat and that the universe is free to take the path of least resis-
open cases the forbidden region indeed goes away and tance i.e. follow the open case, but maybe there is actu-
we have argued that these cases are in a sense more ally no freedom in this choice and it is pre-ordained what
favourable. Further, in the open case the requirement curvature the universe should take before the universe
that the wavefunction should remain finite at arbitrary a comes into existence. This is a rather deep problem that
allows boundary conditions that include Vilenkin’s “out- affects quantum cosmology in general, what variables are
going only” but not the no boundary choice. free to roam and which are fixed externally imposed con-
For the specific Tunneling boundary condition, com- stants ? Is there a classical scaffolding surrounding the
parisons between models that only differ by their value of initial universe or is every variable initially a quantum
k can be made. If the initial curvature is given by a quan- variable? Quantum cosmology seems ambiguous why cer-
tum ensemble, so allowing any possible k, then one can tain variables (e.g. φ) are given by distribution functions,
conclude that open universes are more likely. Although although constrained by the chosen boundary conditions,
the models considered have only a cosmological constant while others (e.g. k) are imposed with no longer appar-
as their matter source. This seem to contradict the no- ent quantum uncertainty. Recall that all classical notions
tion that for DeSitter space “all curvatures are equiva- seem suspect as the Planck epoch is approached cf. [40].
lent”. But because we consider the evolution from a = 0+ But in the meantime it seems that, in admittedly sim-
and regularized the wavefunction around this point we plistic models, different curvatures can be considered and
have derived “propagators” to go to arbitrarily large scale an argument made that the open case is favoured. If the
factors. The universe is given a choice from its conception curvature is initially fixed and so not subject to quan-
which path to follow. In the closed model one can think of tum uncertainty, then one can still argue that open uni-
the universe being “held up” waiting to tunnel through verses are as possible as closed ones with Tunneling like
the barrier. While the open models gain a “push” in boundary conditions. Understanding better the measure
falling down a steeper WDW potential compared to the of possible topologies would seem the next helpful step to
flat case. Once the scale factor becomes sufficiently large see if arguments can be made to favour a specific choice
the possibility of topology change depends entirely on the cf. [41]
energy density of the scalar field driving inflation. As this In summary, we have considered quantum cosmologi-
reduces below Planck values the prediction rapidly gives cal models with arbitrary curvature. Although the mod-
that an open k = −1 universe is favoured. Interestingly els are all compact they are all possible candidates for
at small scale factors and/or Planck energy densities the quantum creation of the universe models. Unlike always
universe might “pin ball” between various possibilities starting from closed models which have a forbidden re-
before settling into one final curvature. The “no hair” gion at small scale factors, one can work directly with
property of DeSitter space, of having finite causal hori- the curvature of one’s choice. Compared to recent in-
zon, might also allow various regions to develop different stanton methods of creating an open universe from an
curvatures. Although the open case would still domi- initial closed one, one cuts out the unnecessary closed
nate pockets of closed curvature could also be created. stage. One has the further prediction that open universes
This could depend on quantum interpretations, whether are favoured followed by flat ones provided in some sense
“many worlds” or “single-history” quantum theories are the universe has the choice of deciding its curvature. The
possible cf. [37]. closed universe case is strongly suppressed in comparison.
Regardless of the actual value of curvature inflation One can still obtain Ω0 ≈ 1 in all cases since a long pe-
is predicted in all cases. This resolves an ambiguity in riod of inflation is strongly predicted. Unfortunately this
classical measures of inflation. If quantum cosmology did large inflationary period would appear to wipe out any
nothing else but gave a definite predictions for inflation to interesting “multiple images” due to topological effects
occur it would be very significant. The only drawback is that are presently being searched for. Only for lesser
that the scalar potential must be chosen to be of the cor- amount of inflation would such effects be apparent in
rect shape, just as the potential of the Hydrogen atom the patterns of Cosmic Background radiation or multiple
has to be provided before one does any quantum me- galaxy images -see [11]. But interestingly with an infla-
chanics. The potential is further constrained by the need tionary matter source, the curvature still dominates the
to create sufficiently small fluctuations and gravitational dynamics both at the beginning and the end of the uni-
waves: roughly speaking inflation should occur at GUT verse, this seems reasonable that the end of the universe
∼ 1014 GeV energy scales [38,39]. Ultimately inflation is should reflect its origins. This is unlike the conventional
a classical phenomena that can not be driven by quan- big bang model where the curvature only dominates in
tum notions alone. We have modeled curvature as being the far future and “only matter matters” during its initial
described by a quantum variable and subject to notion of phase.
probability analogous to the usual initial matter distri-
bution calculation. There is the worry that in trying to
determine the curvature, which is also a part of the po-
tential, we are going beyond the scope of what quantum

12
 ACKNOWLEDGMENTS [16] M. Abramowitz and I. A. Stegun, Handbook of Mathe-
matical Functions, Dover 1965.
It is a pleasure to thank N. Pinto Neto for useful ex- [17] S. W. Hawking and D. N. Page, Phys. Rev. D 42, 2655
changes and comments. We are also grateful to D. Wilt- (1990).
shire for various helpful remarks concerning the factor [18] L. G. Garay, Phys. Rev. D 48, 1710 (1993).
[19] A. Carlini, D. H. Coule and D. M. Solomons, Int. J. of
ordering issue.
Mod. Phys. A 12, 3517 (1997).
[20] M. B. Mijić, M. S. Morris and W. Suen, Phys. Rev. D
39, 1496 (1989).
[21] R. W. Robinett, Quantum mechanics , Oxford University
Press, Oxford 1997.
[22] N. Kontoleon and D. L. Wiltshire, preprint
gr-qc/9807075.
[1] E. W. Kolb and M. S. Turner, The Early Universe [23] S. P. Kim, Phys. Rev. D 46, 3939 (1992).
Addison-Wesley 1990; A. D. Linde, Particle Physics and
[24] R. R. Gott and Li-Xin Li, Phys. Rev. D 58, 023501
Inflationary Cosmology Harwood 1990; D. Atkatz, Am.
(1998).
J. Phys. 62, 619 (1994). [25] J. H. Kung, Gen. Rel. Grav. 27, 35 (1995);
[2] G. W. Gibbons, S. W. Hawking and J. M. Stewart, Nucl. M. L. Fil’chenkov, Phys. Lett. B 354, 208 (1995).
Phys. B 281, 736 (1987); S. W. Hawking and D. N. Page,
[26] D. L. Wiltshire, preprint gr-qc/ 9905090.
Nucl. Phys. B 298, 789 (1988); D. H. Coule, Class. Quan- [27] J. Weeks, http://www.geom.umn.edu:80/software.
tum Grav. 12, 455 (1995). [28] W. P. Thurston, Bull. Am. Math. Soc. 6, 357 (1982).
[3] D. Atkatz and H. Pagels, Phys. Rev. D 25, 2065 (1982).
[29] A. Vilenkin, Phys. Rev. D 39, 1116 (1989).
[4] J. B. Hartle and S. W. Hawking, Phys. Rev. D 28, 2960 [30] G. W. Gibbons and L. P. Grishchuk, Nucl. Phys. B 313,
(1983). 736 (1989).
[5] A. Vilenkin, Phys. Rev. D 30, 509 (1984); A. D. Linde,
[31] G. F. Smoot et al., Astrophys. J. 396, L1 (1992);
Sov. Phys. JEPT 60, 211 (1984); V. A. Rubakov, C. L. Bennett et al., Astrophys. J. 464, L1 (1996).
Phys. Lett. B 148, 280 (1984); Y. B. Zeldovich and [32] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals,
A. A. Starobinsky, Sov. Astron. Lett. 10, 135 (1984).
Series and Products, Academic, New York, 1980.
[6] A. Vilenkin, The quantum cosmology debate, preprint [33] V. De Lorenzi, J. Martin, N. Pinto Neto and I. Soares,
gr-qc/9812027. Phys. Rev. D 56, 3329 (1997).
[7] A. Vilenkin, Phys. Rev. D 33, 3560 (1986); ibid 37, 888
[34] S. Perlmutter et al, Astrophys. J. 516 (in the press), also
(1988). as preprint astro-ph/9812133; A. G. Riess et al, Astron.
[8] S. Coleman and F. De Luccia, Phys. Rev. D 21, 3305 J. 116, 1009 (1998).
(1980); J. R. Gott, Nature 295, 304 (1982); M. Bucher,
[35] S. Coleman, Nucl. Phys. B 310, 643 (1988).
A. S. Goldhaber and N. Turok, Phys. Rev. D 52, 3314 [36] J. M. Cline, Phys. Lett. B 224, 53 (1989); M. B Mijić,
(1995); K. Yamamoto, M. Sasaki and T. Tanaka, Astro- Phys. Lett. B 241, 492 (1990); A. Strominger, Nucl.
phys. J. 455, 412 (1995).
Phys. B 319, 722 (1989) ; D. H. Coule, Mod. Phys. Lett.
[9] S. W. Hawking and N. Turok, Phys. Lett. B 425, 25 A 10 , 1989 (1995).
(1998). [37] D. N. Page, Observational Consequences of many-worlds
[10] N. D. Birrell and P. C. W. Davies, Quantum Fields in
quantum theory, preprint quant-ph/9904004.
curved space, Cambridge University press, Cambridge [38] V. A. Rubakov, M. V. Sazhin and A. V. Veryaskin, Phys.
1982. Lett. B 115, 189 (1982).
[11] J. P. Luminet, Phys. Report 254, 135 (1995) and Topol-
[39] L. F. Abott and M. B. Wise, Nucl. Phys. B 244, 541
ogy of the Universe conference, Clevelend 1997, special (1984).
issue of Class. Quant. Grav. 15, 2529 (1998); for a sim- [40] C. Schiller, Does matter differ from vacuum, preprint
pler introduction to compact manifolds in cosmology see:
gr-qc/9610066.
W. P. Thurston and J. R. Weeks, Sci. Amer. 251, 108 [41] J. B. Hartle, Scientific knowledge from the perspective of
(1984); and more recently: J. P. Luminet, G. D. Stark- quantum cosmology, preprint gr-qc/9601046.
man and J. R. Weeks, Sci. Amer., April (1999).
[12] S. W. Hawking and D. N. Page, Nucl. Phys. B 264, 185
(1986); D. N. Page, J. Math. Phys. 32, 3427 (1991).
[13] M. A. Jafarizadeh, F. Darabi, A. Rezaei-Aghdam and
A. R. Rastegar, Mod. Phys. Lett. A 13, 3213 (1998).
also preprint gr-qc/9811081.
[14] J. Martin, Phys. Rev. D 49, 5086 (1994); A. Carlini,
D. H. Coule and D. M. Solomons, Mod. Phys. Lett. A
11, 1453 (1996).
[15] J. J. Halliwell in Quantum Cosmology and Baby Uni-
verses. Proceedings of the 1989 Jerusalem Winter School
for Theoretical Physics, eds. S. Coleman, J. H. Hartle,
T. Piran and S. Weinberg, World Scientific , Singapore.

13