Primordial Black Holes:

Tunnelling vs. No Boundary Proposal∗

arXiv:gr-qc/9608009v1 2 Aug 1996

Raphael Bousso† and Stephen W. Hawking‡

Department of Applied Mathematics and
Theoretical Physics
University of Cambridge
Silver Street, Cambridge CB3 9EW

DAMTP/R-96/34

Abstract
In the inflationary era, black holes came into existence together with the
universe through the quantum process of pair creation. We calculate the pair
creation rate from the no boundary proposal for the wave function of the uni-
verse. Our results are physically sensible and fit in with other descriptions of
pair creation. The tunnelling proposal, on the other hand, predicts a catas-
trophic instability of de Sitter space to the nucleation of large black holes,
and cannot be maintained.

∗
Contribution to the Proceedings of COSMION 96 (Moscow, May 1996)
†
[email protected]
‡
[email protected]

1
1 Introduction
1.1 Primordial Black Holes
We now have good observational evidence for black holes from stellar masses up to
super-massive holes of 108 to 1010 solar masses and maybe even more. However,
one can also speculate on the possible existence of black holes of much lower mass.
These are the holes for which quantum effects can be important. Such holes could
not form from the collapse of normal baryonic matter because degeneracy pressure
will support white dwarfs or neutron stars below the Chandrasekhar limiting mass.
One can express this limiting mass as mPlanck (mPlanck /mbaryon )2 . Its value is about a
solar mass, which might seem a coincidence, but there are good anthropic principle
reasons why stars should be just on the verge of gravitational collapse.
This limiting mass applies only to the formation of black holes through the
gravitational collapse of fermions. In the case of bosons the limiting mass is given
by mPlanck (mPlanck /mboson ). To form a black hole by the gravitational collapse of
bosons, they need to have a non-zero mass and either be stable or have a fairly
long life. About the only candidate is the axion, which might have a mass of about
10−5 eV. In this case the limiting mass would be about the mass of the Earth,
which is still quite high, and too large for quantum effects to be observable. To
get black holes that are significantly smaller, one could not rely on gravitational
collapse, but would have to shoot matter together with high energies. John Wheeler
once calculated that if one made a hydrogen bomb with all the deuterium from
the oceans, the centre would implode so violently that a little black hole would be
formed. Perhaps fortunately, this experiment is unlikely to be performed. Thus the
only place where tiny black holes might be formed is the early universe.
Previous discussions of black holes formed in the early universe have concen-
trated on black holes formed by matter coming together during the radiation era or
first order phase transitions. Recent work on the critical behaviour of gravitational
collapse has shown it is possible to form black holes in these situations. However, it
is difficult because one has to arrange for matter to be fired together at high speed
and accurately focused into a small region. Yet if too much matter is fired together
it forms a closed universe on its own, with no connection with our universe. Such a
separate universe would not be a black hole in our universe.
Black holes formed by collapse, or by hurling matter together, are not really
primordial, in the sense that they do not form until a definite time after the beginning
of the universe. On the other hand, the black holes we are going to consider form

2
by the quantum process of pair creation and are truly primordial, in that they can
be considered to have existed since the beginning of the universe.

1.2 Inflation
It is generally assumed that the universe began with a period of exponential ex-
pansion called inflation. This era is characterised by the presence of an effective
cosmological constant Λeff due to the vacuum energy of a scalar field φ. In chaotic
inflation [1, 2] the effective cosmological constant typically starts out large and then
decreases slowly until inflation ends when Λeff ≈ 0. Correspondingly, these models
predict cosmic density perturbations which are proportional to the logarithm of the
scale. On scales up to the current Hubble radius Hnow −1
, this agrees well with ob-
servations of near scale invariance. However, on much larger length scales of order
Hnow
−1
exp(105 ), perturbations are predicted to be on the order of one. Of course, this
means that the perturbational treatment breaks down; but it indicates that black
holes may be created.
Linde [3, 4] noted that in the early stages of inflation, when the strong den-
sity perturbations originate, the quantum fluctuations of the inflaton field are much
larger than its classical decrease per Hubble time. He concluded that therefore there
would always be regions of the inflationary universe where the field would grow, and
so inflation would never end globally (“eternal inflation”). However, this approach
only allows for fluctuations of the field. One should also consider fluctuations which
change the topology of space-time. This topology change corresponds to the forma-
tion of a pair of black holes. The pair creation rate can be calculated using instanton
methods, which are well suited to this non-perturbative problem.

1.3 Pair Creation

Quantum pair creation is only possible on a background that provides a force which
pulls the pair apart. In the case of a virtual electron-positron pair, for example, the
particles can only become real if they appear in an external electric field. Otherwise
they would just fall back together and annihilate. The same holds for black holes;
examples in the literature include their pair creation on a cosmic string [5], where
they are pulled apart by the string tension; or the pair creation of magnetically
charged black holes on the background of Melvin’s universe [6], where they are sep-
arated by a magnetic field. In our case, the black holes will be accelerated apart

3
by the inflationary expansion of the universe. While preventing classical gravita-
tional collapse, this expansion provides a suitable background for the quantum pair
creation of black holes.
After the end of inflation, during the radiation and matter dominated eras, the
effective cosmological constant was nearly zero. Thus the only time when black
hole pair creation was possible in our universe was during the inflationary era, when
Λeff was large. Moreover, these black holes are unique since they can be so small
that quantum effects on their evolution are important. Indeed, their evolution turns
out to be quite interesting and non-trivial [7]. Here we will only describe the cre-
ation of black holes, summarising a more rigourous treatment [8]. We focus on the
consequences for the choice of the prescription for the wave function of the universe.
In the standard semi-classical treatment of pair creation, one finds two instan-
tons: one for the background, and one for the objects to be created on the back-
ground. From the instanton actions Ibg and Iobj one calculates the pair creation rate
Γ:
Γ = exp [− (Iobj − Ibg )] , (1.1)
where we neglect a prefactor. This prescription has been very successfully used
by a number of authors recently [9, 10, 11, 12] for the pair creation of black holes
on various backgrounds. It is motivated not only by analogies in quantum me-
chanics and quantum field theory [13, 14], but also by considerations of black hole
entropy [15, 16, 17].
In this paper, however, we will obtain the pair creation rate through a somewhat
more fundamental procedure. Since we have a cosmological background, we can
apply the tools of quantum cosmology, and use the wave function of the universe to
describe black hole pair creation. Two different prescriptions have been put forward
for the calculation of this wave function: Vilenkin’s tunnelling proposal [18], and the
Hartle-Hawking no boundary proposal [19] (reviewed in Sec. 2). We will describe
the creation of an inflationary universe by a de Sitter type gravitational instanton,
which has the topology of a four-sphere, S 4 . In this picture, the universe starts
out with the spatial size of one Hubble volume. After one Hubble time, its spatial
volume will have increased by a factor of e3 ≈ 20. However, by the de Sitter
no hair theorem, we can regard each of these 20 Hubble volumes as having been
nucleated independently through gravitational instantons. With this interpretation,
we are allowing for black hole pair creation, since some of the new Hubble volumes
might have been created through a different type of instanton that has the topology
S 2 × S 2 and thus represents a pair of black holes in de Sitter space [20]. Using the

4
no boundary proposal, we assign probability measures to both instanton types. We
then estimate the fraction of inflationary Hubble volumes containing a pair of black
holes by the fraction Γ of the two probability measures. This is equivalent to saying
that Γ is the pair creation rate of black holes on a de Sitter background.
In Sec. 3 we describe the relevant instantons and calculate the pair creation rate.
The result is compared with that obtained from the tunnelling proposal in Sec. 4,
where we demonstrate that the usual description of pair creation, Eq. (1.1), arises
naturally from the no boundary proposal. We shall use units in which mP = h̄ =
c = k = 1.

2 The Wave Function of the Universe

The prescription for the wave function of the universe has long been one of the
central, and arguably one of the most disputed issues in quantum cosmology. The
two competing proposals differ in their choice of boundary conditions for the wave
function.

2.1 No Boundary Proposal

According to the no boundary proposal, the quantum state of the universe is defined
by path integrals over Euclidean metrics gµν on compact manifolds M. From this
it follows that the probability of finding a three-metric hij on a spacelike surface
Σ is given by a path integral over all gµν on M that agree with hij on Σ. If the
spacetime is simply connected (which we shall assume), the surface Σ will divide
M into two parts, M+ and M− . One can then factorise the probability of finding
hij into a product of two wave functions, Ψ+ and Ψ− . Ψ+ (Ψ− ) is given by a path
integral over all metrics gµν on the half-manifold M+ (M− ) which agree with hij
on the boundary Σ. In most situations Ψ+ equals Ψ− . We shall therefore drop
the suffixes and refer to Ψ as the wave function of the universe. Under inclusion of
matter fields, one arrives at the following prescription:
Z
Ψ[hij , ΦΣ ] = D(gµν , Φ) exp [−I(gµν , Φ)] , (2.1)

where (hij , ΦΣ ) are the 3-metric and matter fields on a spacelike boundary Σ and the
path integral is taken over all compact Euclidean four geometries gµν that have Σ
as their only boundary and matter field configurations Φ that are regular on them;

5
I(gµν , Φ) is their action. The gravitational part of the action is given by

1 1
Z Z
IE = − d4x g 1/2 (R − 2Λ) − d3x h1/2 K, (2.2)
16π M+ 8π Σ

where R is the Ricci-scalar, Λ is the cosmological constant, and K is the trace of

Kij , the second fundamental form of the boundary Σ in the metric g.
We shall calculate the wave function semi-classically, using a saddle-point ap-
proximation to the path integral; and from the wave function we shall calculate the
pair creation rate. The method can be outlined as follows. One is interested in
two types of inflationary universes: one with a pair of black holes, and one without.
They are characterised by spacelike sections of different topology. For each of these
two universes, one has to find a classical Euclidean solution to the Einstein equa-
tions (an instanton), which can be analytically continued to match a boundary Σ
of the appropriate topology. One then calculate the Euclidean actions I of the two
types of saddle-point solutions. Semiclassically, it follows from Eq. (2.1) that the
wave function is given by
Ψ = exp (−I) , (2.3)
neglecting a prefactor. One can thus assign a probability measure to each type of
universe:  
P = |Ψ|2 = exp −2I Re , (2.4)
where the superscript ‘Re’ denotes the real part. As explained in the introduction,
the ratio of the two probability measures gives the rate of black hole pair creation
on an inflationary background, Γ.
The probability measure P for the nucleation of a space-time should be propor-
tional to the number of possible quantum states it contains, eS . The entropy S of a
space-time is given by the total of its horizon areas, divided by four; it follows that
S = −2I Re in the cosmological case [17]. So Eq. (2.4) above does indeed reflect the
number of internal states. If the black hole space-time has lower entropy than the
background, one obtains Γ < 1. Then the pair creation will be suppressed, as it
should be.

2.2 Tunnelling Proposal

The tunnelling proposal places different boundary conditions on the wave function
at small geometries in the Euclidean region.

6
 The action (2.2) is in general negative for a small boundary geometry hij . Thus
Ψ = e−I is enhanced. The proponents of the tunnelling proposal feel, however,
that the wave function ought to be suppressed in the Euclidean region because it is
supposed to be forbidden. They are therefore forced to choose the
ΨTP = exp (+I) (2.5)
solution of the Wheeler-DeWitt equation as the boundary condition at small hij .
This has the obvious disadvantage that it does not reflect the entropy difference
correctly. Transitions in the direction of lower entropy are enhanced, rather than
suppressed. This will lead to absurd predictions in the context of pair creation.
In the following two sections we shall discuss the saddle-point solutions needed to
describe the pair creation of black holes on a cosmological background [8]. We shall
use only the no boundary proposal to calculate the probability measures and the
pair creation rate. The disastrous consequences of choosing the prescription (2.5),
instead, will be discussed in Sec. 4.

3 Instantons
We shall assume spherical symmetry. Before we introduce a more realistic infla-
tionary model, it is helpful to consider a simpler situation with a fixed positive
cosmological constant Λ but no matter fields. We can then generalise quite easily
to the case where an effective cosmological “constant” arises from a scalar field.

3.1 de Sitter Space

First we consider the case without black holes, a homogeneous isotropic universe.
Since Λ > 0, its spacelike sections will simply be round three-spheres. The wave
function is given by a path integral over all metrics on a four-manifold M+ bounded
by a round three-sphere Σ of radius aΣ . The corresponding saddle-point solution
is the de
q Sitter space-time. Its Euclidean metric is that of a round four-sphere of
radius 3/Λ:
ds2 = dτ 2 + a(τ )2 dΩ23 , (3.1)
where τ is Euclidean time, dΩ23 is the metric on the round three-sphere of unit radius,
and s s
3 Λ
a(τ ) = sin τ. (3.2)
Λ 3

7
 We can regard Eq. (3.2) as a function on
q the complex τ -plane. On a line parallel
to the imaginary τ -axis defined by τ Re = Λ3 π2 , we have
s s
3 Λ Im
a(τ )|τ Re =√ 3 π = cosh τ . (3.3)
Λ 2 Λ 3

This describes a Lorentzian de Sitter hyperboloid, with τ Im serving as a Lorentzian

time variable. One can thus construct a complex solution, which is the analyti-
cal continuation of the Euclidean four-sphere metric. q It is obtained by choosing a
Re 3 π
contour in the complex τ -plane from 0 to τ = Λ 2 (which describes half of the
Euclidean four-sphere) and then parallel to the imaginary τ -axis (which describes
half the Lorentzian hyperboloid). The geometry corresponding to this path is shown
in (Fig. 1).
The Lorentzian part of the metric will contribute a purely imaginary term to the
action. This will affect the phase of the wave function but not its amplitude. The
real part of the action of this complex saddle-point metric will be the action of the
Euclidean half-four-sphere:
Re 3π
Ide Sitter = − . (3.4)
2Λ
Thus the magnitude of the wave function will still be e3π/2Λ , corresponding to the
probability measure
3π
 
Pde Sitter = exp . (3.5)
Λ

3.2 Schwarzschild-de Sitter Space

We turn to the case of a universe containing a pair of black holes. Now the cross
sections Σ have topology S 2 ×S 1 . Generally, the radius of the S 2 varies along the S 1 .
This corresponds to the fact that the radius of a black hole immersed in de Sitter
space can have any value between zero and the radius of the cosmological horizon.
The minimal two-sphere corresponds to the black hole horizon, the maximal two-
sphere to the cosmological horizon. The saddle-point solution corresponding to this
topology is the Schwarzschild-de Sitter universe. However, the Euclidean section of
this spacetime typically has a conical singularity at one of its two horizons and thus
does not represent a regular instanton [7, 20]. The only regular Euclidean solution
is the degenerate case where the black hole has the maximum possible size. It is

8
 boundary 

Lorentzian
section

Euclidean section

Figure 1: The creation of a de Sitter universe. The lower region is half of a Euclidean
four{sphere, embedded in ve{dimensional Euclidean at space. The upper region
is a Lorentzian four{hyperboloid, embedded in ve{dimensional Minkowski space.

also known as the Nariai solution and given by the topological product of two round
two-spheres:
ds2 = dτ 2 + a(τ )2 dx2 + b(τ )2 dΩ22 , (3.6)
where x is identified with period 2π, dΩ22 = dθ2 + sin2 θ dϕ2 , and
s s
1 √ 1
a(τ ) = sin Λτ, b(τ ) = = const. (3.7)
Λ Λ
In this case the radius b of the S 2 is constant in the S 1 direction. The black
hole and the cosmological horizon have equal radius and no conical singularities
√ are
present. There will be no saddle-point solution unless we specify bΣ = 1/ Λ. Then

9
the only variable we are free to choose on Σ is the radius aΣ of the one-sphere. In
the Lorentzian section, the one-sphere expands rapidly,
s
1 √
a(τ )|τ Re =√ 1 π = cosh Λτ Im , (3.8)
Λ 2 Λ
while the two-sphere (and, therefore, the black hole radius) remains constant. Again
we can construct a complex saddle-point, which can be regarded as half a Euclidean
S 2 × S 2 joined to half of the Lorentzian solution. The real part of the action will be
the action of the half of a Euclidean S 2 × S 2 :
Re π
ISdS =− . (3.9)
Λ
The corresponding probability measure is
2π
 
PSdS = exp . (3.10)
Λ
We divide this by the probability measure (3.5) for a universe without black holes
to obtain the pair creation rate of black holes in de Sitter space:
PSdS π
 
Γ= = exp − . (3.11)
Pde Sitter Λ
Thus the probability for pair creation is very low, unless Λ is close to the Planck
value, Λ = 1.

3.3 Effective Cosmological Constant

Of course the real universe does not have a large cosmological constant. However, in
inflationary cosmology it is assumed that the universe starts out with a very large
effective cosmological constant, which arises from the potential V of a scalar field φ.
The exact form of the potential is not critical. So for simplicity we chose V to be the
potential of a field with mass m, but the results would be similar for a λφ4 potential.
To account for the observed fluctuations in the microwave background [21], m has
to be on the order of 10−5 to 10−6 [22]. The wave function Ψ will now depend on
the three-metric hij and the value of φ on Σ. For φ > 1 the potential acts like an
effective cosmological constant

Λeff (φ) = 8πV (φ). (3.12)

10
There will again be complex saddle-points which can be regarded as a Euclidean so-
lution joined to a Lorentzian solution. Due to the time dependence of Λeff , however,
one cannot find a path in the τ -plane along which the Euclidean and Lorentzian
metrics will be exactly real [8]. Apart from this subtlety, the saddle point solutions
are similar to those for a fixed cosmological constant, with the time-dependent √ Λeff
replacing Λ. The radius of the pair created black holes will now be given by 1/ Λeff .
As before, the magnitude of the wave function comes from the real part of the ac-
tion, which is determined by the Euclidean part of the metric. This real part will
be
3π
ISRe3 = − (3.13)
2Λeff (φ0 )
in the case without black holes, and
π
ISRe2 ×S 1 = − (3.14)
Λeff (φ0 )
in the case with a black hole pair. Here φ0 is the value of φ in the initial Euclidean
region. Thus the pair creation rate is given by
" #
PS 2 ×S 1 π
Γ= = exp − . (3.15)
PS 3 Λeff (φ0 )

4 Tunnelling vs. No Boundary Proposal

In the previous sections we have used the no boundary proposal to calculate the pair
creation rate of black holes during inflation. Let us interpret the result, Eq. (3.15).
Since 0 < Λeff ≤ 1, we get Γ < 1, and so black hole pair creation is suppressed. In
the early stages of inflation, when Λeff ≈ 1, the suppression is week, and black holes
will be plentifully produced . However, those black holes will be very small, with a
mass on the order of the Planck mass. Larger black holes, corresponding to lower
values of Λeff at later stages of inflation, are exponentially suppressed. A detailed
analysis of their evolution [7] shows that the small black holes typically evaporate
immediately, while sufficiently large ones grow with the horizon and survive long
after inflation ends.
We now understand how the standard prescription for pair creation, Eq. (1.1),
arises from the no boundary proposal: By Eq. (2.4),
PS 2 ×S 1 h  i
Γ= = exp − 2ISRe2 ×S 1 − 2ISRe3 , (4.1)
PS 3

11
where I Re denotes the real part of the Euclidean action of a complex saddle-point
solution. But we have seen that this real part is equal to half of the action of the
complete Euclidean solution. Thus Iobj = 2ISRe2 ×S 1 and Ibg = 2ISRe3 , and we recover
Eq. (1.1).
Let us return to the tunnelling proposal and see what results it would have
produced. ΨTP is given by e+I rather than e−I . This choice of sign is inconsistent
with Eq. (1.1), as it leads to the inverse result for the pair creation rate: ΓTP = 1/Γ.
In our case, we would get ΓTP = exp(+π/Λeff ). Thus black hole pair creation would
be enhanced, rather than suppressed. This means that de Sitter space would decay:
it would be catastrophically unstable to √ the formation of black holes. Since the
radius of the black holes is given by 1/ Λeff , the black holes would be more likely
the larger they were. Clearly, the tunnelling proposal cannot be maintained. On
the other hand, Eq. (3.15), which was obtained from the no boundary proposal,
is physically very reasonable. It allows topological fluctuations near the Planckian
regime, but suppresses the formation of large black holes at low energies.
We summarise. The cosmological pair production of black holes provides an ideal
theoretical laboratory in which to examine the question of the boundary conditions
for the wave function of the universe. The results could not be more decisive. The no
boundary proposal leads to physically sensible results, while the tunnelling proposal
predicts a disastrous enhancement of black hole production.

References
[1] A. D. Linde: Chaotic inflation. Phys. Lett. 129B, 177 (1983).

[2] S. W. Hawking: The quantum state of the Universe. Nucl. Phys. B239, 257
(1984).

[3] A. D. Linde: Eternally existing selfreproducing chaotic inflationary universe.

Phys. Lett. B175, 395 (1986).

[4] A. D. Linde: Eternal chaotic inflation. Mod. Phys. Lett. A1, 81 (1986).

[5] S. W. Hawking and S. F. Ross: Pair production of black holes on cosmic strings.
Phys. Rev. Lett. 75, 3382 (1995), gr–qc/9506020.

[6] G. W. Gibbons: Quantized flux tubes in Einstein–Maxwell theory and noncom-

pact internal spaces. In Fields and Geometry 1986: Proceedings of the 22nd

12
 Karpacz Winter School of Theoretical Physics, edited by A. Jadczyk, World
Scientific, Singapore (1986).

[7] R. Bousso and S. W. Hawking: Pair creation of black holes during inflation.
Preprint no. DAMTP/R–96/33, gr–qc/9606052.

[8] R. Bousso and S. W. Hawking: The probability for primordial black holes. Phys.
Rev. D 52, 5659 (1995), gr–qc/9506047.

[9] F. Dowker, J. P. Gauntlett, D. A. Kastor and J. Traschen: Pair creation of

dilaton black holes. Phys. Rev. D 49, 2909 (1994), hep–th/9309075.

[10] R. B. Mann and S. F. Ross: Cosmological production of charged black hole pairs.
Phys. Rev. D 52, 2254 (1995), gr–qc/9504015.

[11] S. W. Hawking and S. F. Ross: Duality between electric and magnetic black
holes. Phys. Rev. D 52, 5865 (1995), hep–th/9504019.

[12] R. R. Caldwell, A. Chamblin and G. W. Gibbons: Pair creation of black holes

by domain walls. Phys. Rev. D 53, 7103 (1996), hep–th/9602126.

[13] S. Coleman: The fate of the false vacuum. 1. Semiclassical theory. Phys. Rev.
D 15, 2929 (1977).

[14] C. Callan and S. Coleman: The fate of the false vacuum. 2. First quantum
corrections. Phys. Rev. D 16, 1762 (1977).

[15] D. Garfinkle, S. B. Giddings and A. Strominger: Entropy in black hole pair

production. Phys. Rev. D 49, 958 (1994), gr–qc/9306023.

[16] F. Dowker, J. P. Gauntlett, S. B. Giddings and G. T. Horowitz: On pair creation

of extremal black holes and Kaluza–Klein monopoles. Phys. Rev. D 50, 2662
(1994), hep–th/9312172.

[17] S. W. Hawking, G. T. Horowitz and S. F. Ross: Entropy, area, and black hole
pairs. Phys. Rev. D 51, 4302 (1995), gr–qc/9409013.

[18] A. Vilenkin: Boundary conditions in quantum cosmology. Phys. Rev. D 33,

3560 (1986).

13
[19] J. B. Hartle and S. W. Hawking: Wave function of the Universe. Phys. Rev. D
28, 2960 (1983).

[20] P. Ginsparg and M. J. Perry: Semiclassical perdurance of de Sitter space. Nucl.

Phys. B222, 245 (1983).

[21] G. F. Smoot et al.: Structure in the COBE DMR first year maps. Astrophys.
J. 396, L1 (1992).

[22] J. J. Halliwell and S. W. Hawking: The origin of structure in the universe.

Phys. Rev. D 31, 1777 (1985).

14