Black holes in the quantum

universe
royalsocietypublishing.org/journal/rsta
Steven B. Giddings1,2
1 Department of Physics, University of California, Santa Barbara,

CA 93106, USA
Review 2 Theory Department, CERN, 1 Esplande des Particules, Geneva 23,

Cite this article: Giddings SB. 2019 Black CH 1211, Switzerland

holes in the quantum universe. Phil. Trans. R. SBG, 0000-0002-9504-4166
Soc. A 377: 20190029.
http://dx.doi.org/10.1098/rsta.2019.0029 A succinct summary is given of the problem of
reconciling observation of black hole-like objects with
quantum mechanics. If quantum black holes behave
Accepted: 22 June 2019 like subsystems, and also decay, their information
must be transferred to their environments.
One contribution of 14 to a discussion meeting Interactions that accomplish this with ‘minimal’
issue ‘Topological avatars of new physics’. departure from a standard description are
parametrized. Possible sensitivity of gravitational
Subject Areas: wave or very long baseline interferometric
observations to these interactions is briefly outlined.
high energy physics, quantum physics,
This article is part of a discussion meeting issue
relativity ‘Topological avatars of new physics’.

Keywords:
black holes, unitarity, quantum information
1. Introduction
Author for correspondence: While there is spectacular evidence for objects that look
Steven B. Giddings and act very much like black holes in the Universe,
e-mail: [email protected] there is no known description of their evolution that
is consistent with the principles of quantum mechanics,
which are believed to govern all physical law. This
appears to present a deep crisis in fundamental physics.
Black holes (BHs) certainly appear to exist. Evidence
includes jets shot from the centres of galaxies, like the
5000 lightyear long jet from the centre of M87, thought to
be generated by a supermassive black hole. Stellar orbits
have been observed about a highly compact object with
approximate mass 4 × 106 M at the centre of our own
galaxy, and more recently, an orbiting hotspot has been
observed at a radius just a few times the corresponding
Schwarzschild radius [1]. Gravitational waves have now
been observed by LIGO from what appear to be BH
collisions, beginning with [2], and continuing with many
more. And remarkably, since this meeting, the Event
Horizon Telescope has imaged the central object in M87,
revealing an apparent ‘shadow’ of a 6.5 × 109 M black
hole [3].

2019 The Author(s) Published by the Royal Society. All rights reserved.
 To the best of our knowledge, all physical law needs to respect the principles of quantum
2
mechanics, which are increasingly well tested. But, our present ‘standard’ description of BHs,
based on the geometrical spacetime picture of general relativity (GR), together with quantized

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
fluctuations of other fields (including the gravitational field) on a background geometry, indicates
a breakdown of quantum mechanics. So, BHs appear to directly point to a deep inconsistency in
our present quantum-mechanical laws.
Specifically, all other observed phenomena in the Universe appear to be well-described within
the framework of local quantum field theory (LQFT), on a semiclassical background geometry.
As a theoretical structure, LQFT arises from a set of basic principles: the principles of quantum
mechanics, the principles of relativity and the principle of locality. In particular, the latter two
are directly associated with the statement that we work with classical geometry and topology.
BHs appear to reveal that this framework is inconsistent, and thus that these principles are
inconsistent, and one or more of them must be modified. We first turn to a lightening review
of this crisis in physics.

2. The unitarity crisis

Figure 1 illustrates the basic phenomenon of BH formation and Hawking radiation [4]. Collapsing
or colliding matter forms a BH, with Schwarzschild radius R. We can describe evolution by
introducing a slicing of the resulting spacetime, as shown. In LQFT, fluctuations of any of the
quantum fields near the horizon get ‘pulled apart’ due to the strong gravitational field, resulting
in outgoing particles together with ingoing excitations that evolve towards r = 0. The typical
energy of the outgoing particles is given by the Hawking temperature, E ∼ TH ∼ 1/R. A ‘toy’
version of the resulting state of such a pair is

|ψ ∼ |0̂|0 + |1̂|1 (2.1)

where hatted/unhatted kets denote the state of internal/external excitations and we focus only
on one possible mode with occupation number zero or one.1 The state (2.1) is just like the Bell
state of an EPR pair, and in particular, exhibits the key feature of entanglement.
Of order one Hawking quantum is emitted per time R, building up a state of the form

|Ψ Hawking ∼ (|0̂|0 + |1̂|1)⊗n . (2.2)

Once n ∼ RM ∼ GM2 such quanta have been emitted, all the initial energy has been carried away
and the BH is expected to disappear. If it does, the internal excitations are no longer part of the
physical description, and the quantum state is found by tracing over them to find a density matrix,
 
ρHawking = Trbh |Ψ Hawking Ψ |Hawking ∼ (|00| + |11|)⊗n . (2.3)

However, if this is the fundamental evolution, from what could be an initial pure state of
the matter forming the BH to a mixed state, that violates the quantum-mechanical principle
of unitarity.
This violation of unitarity can be quantified by the von Neumann entropy of the density matrix
ρ of the radiation,
SvN = −Tr(ρ log ρ) = −I (2.4)

where we think of −I as ‘missing information’. As the BH emits Hawking particles, this missing
information grows uniformly in time, until it reaches a final value S ∼ GM20 ∼ (M0 /mpl )2 , where
M0 is the initial BH mass, and mpl is the Planck mass. This is enormous for a macroscopic BH.
This failure of unitarity is the origin of the crisis in fundamental physics which we therefore
call the unitarity crisis. Various possible resolutions have been considered. The most obvious and
mundane possibility is that black holes do not completely evaporate but instead leave behind
microscopic black hole remnants. However, these remnants would have to have an unbounded
1
For a more complete description of such a state, see e.g. [5].
 black hole
3

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
r = 2GM/c2 = 2M = R
T
light

‘singularity’
cone

T
collapsing or colliding
matter r

Figure 1. A spacetime diagram of a BH, in an Eddington–Finkelstein picture, that forms through matter collapse or collision
and then begins to evaporate. Shown is a choice of slices used to describe evolution; there is considerable flexibility in the choice
of such slices. (Online version in colour.)

number of internal states, to parametrize the missing information from an arbitrarily large black
hole. This kind of spectrum leads to a disaster—unbounded production of such remnants in
general physical processes, and is thus apparently ruled out.2
A second possibility is that there is some error in reasoning in the argument that has been just
outlined for information loss. However, over 40 years of careful examination have not revealed
such an error, and so that seems extremely unlikely. (Further discussion of one newer argument
for an error, the ‘soft hair’ proposal [8–11], will appear below.)
The third possibility now seems by far the most likely: this crisis points to an error in
underlying principles and thus calls for modification of one or more of the principles of relativity,
quantum mechanics and locality. Since these principles are the cornerstones of LQFT, our current
best description of the rest of physics, this is clearly the most exciting possibility. It indicates
that BHs should help serve as guides to a more profound understanding of physics. Indeed,
it appears that the unitarity crisis may play a role analogous to the crisis of atomic instability
in classical physics; the need to explain the atom led to the conceptual revolution of quantum
mechanics.
There are additional reasons to view the unitarity crisis as a key problem for quantum gravity.
First, the quantum physics of BH formation and evaporation appears to be the generic high-
energy physics of gravity: if we consider a general scattering process and increase the CM energy,
at large enough values, we expect to enter the regime of strong gravity, where classically a BH
would form. Secondly, as has become clear, and as we will see below, the unitarity crisis appears
to require a modification of long distance, or infrared, physics. That strongly suggests it is a more
profound problem than that of nonrenormalizability, which has historically motivated a lot of
the work in quantum gravity. Nonrenormalizability was in particular a strong motivator for
supergravity and superstring theory, since those theories promised to greatly improve, or cure, the
divergences in the perturbative expansion. But, nonrenormalizability represents a short-distance
problem, anticipated to be cured by short-distance effects, e.g. at the Planck or string length scales.
Moreover, when viewed in the perturbative scattering context, the problem of nonunitarity arises

2
For further discussion, see [6,7], and references therein.
from properties of the sum over diagrams3 , and in that sense in addition to being a long-distance
4
problem, involves the nonperturbative physics of gravity.
A first possibility for a resolution arising from an error of principles was suggested

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
by Hawking [13], who initially argued for a modification of quantum mechanics, allowing
fundamentally nonunitary evolution. However, it was later argued that such a breakdown of
unitarity would lead to massive violation of energy conservation [14], in violent disagreement
with experience.
Attempts to modify quantum mechanics4 have typically led to disaster; quantum mechanics
is a remarkably rigid framework. Because of this ‘quantum rigidity,’ and for other reasons such
as its many experimental tests, I will therefore make the (radically conservative) assumption that
the Universe is governed by quantum-mechanical principles (suitably generally formulated to
incorporate gravity [16]). This means we should look for a different error of principles.
Indeed, let us turn things around. Suppose that BHs can be thought of as quantum subsystems
(this will be discussed further below). Suppose that, as Hawking argued, they also build
up information content, or more precisely entanglement with their environments, and they
radiate and disappear. And, suppose that quantum mechanics—specifically unitarity—holds.
This collection of statements implies that information must transfer out of a BH as it decays. This
violates the locality principle, which states that information does not propagate outside the light
cone, and does so not just on microscopic scales but at scales the size of an arbitrarily large black
hole. Thus black holes strongly point to a long-distance modification of conventional locality,
which also turns out to be indicated by other behaviour of gravity.
However, if this is true, a basic principle of LQFT is not valid, and we need to think about
how to consistently describe such physics. Another important question is how to reconcile this
statement with the approximate validity of LQFT on semiclassical background geometry.

3. Quantum-first gravity
To address these problems, one can try to develop a principled approach, based on the assumption
that the fundamental description of the theory should be quantum-mechanical. This is what I call
the ‘quantum-first’ approach to gravity [16–23]: instead of trying to quantize geometry, we should ask
what kind of quantum structure can approximately describe gravity.
Basic ingredients of a quantum theory are a linear space of states, H, an algebra of observables,
A, and, for states with appropriate asymptotics, such as Minkowskian, unitarity.5 However,
clearly a more mathematical structure than this is needed to describe physics.

(a) Localization and subsystems in quantum gravity

LQFT furnishes an example of the role of such a mathematical structure. As described in the
algebraic approach (see e.g. [24]), this mathematical structure arises from the manifold M on
which the LQFT is defined. Associated with any open set U ⊂ M, there is a subalgebra AU ⊂ A of
the observables with support only in U (e.g. fields smeared with compact-support test functions).
For spacelike separates open sets U, U , the subalgebras AU and AU commute. Moreover, these
subalgebras have nesting and intersection properties just like the open sets. So, the subalgebras
‘mirror’ the topological structure of the manifold given by the open sets, along with the causal
structure. This ‘net’ of subalgebras is the mathematical structure relevant to describing LQFT.
However, this does not appear to be the correct mathematical structure for describing a
gravitational theory. In fact, already at the perturbative level, we discover that this local algebraic
structure must fail. Consider a scalar field φ(x), coupled to gravity. While [φ(x), φ(y)] = 0 for
spacelike x–y, φ(x) is not a gauge-invariant observable. Gauge observables can be constructed
3
For further discussion, see e.g. [12], and references therein.
4
Beyond [13], see also, for example, [15].
5
For further discussion see [16].
[25]6 , by ‘dressing’ φ(x) to give a new operator Φ(x) that also creates the gravitational field
5
associated with a particle created by φ(x).
Since Φ(x) creates a gravitational field, which extends to infinity, now generically [25]

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
[Φ(x), Φ(y)] = 0 for spacelike x–y. In fact the nonvanishing commutators match well to the ‘locality
bound’ that was proposed in [28–30], to characterize the regime where LQFT fails to be an accurate
description of physics.
This shows that the fundamental gauge-invariant observables in gravity do not obey a
local algebra, and thus we must have some other mathematical structure on H, providing the
foundations of the theory. This ‘substrate’ is not clearly based on any underlying manifold
structure, despite what is commonly assumed. In LQFT, the net structure on A is directly
associated with the manifold, but for gravity it is an important question as to what replaces this.
Indeed, without some version of locality, providing a notion of separability, it is not clear how
to even proceed to describe physics. In quantum systems, a more primitive notion of locality
is that of subsystems. For finite quantum systems, subsystems are described in terms of tensor
factorizations of the Hilbert space, such as H = H1 ⊗ H2 . In LQFT, subsystems can be described
in terms of the commuting subalgebras AU .
We can ask if there is some suitable definition of subsystems for gravity. Full discussion of this
would take us too far afield; recent discussion of this subject includes [22,23,31–33], which can be
briefly summarized. Consider a state with some matter excitations in a neighbourhood U, gotten
by acting with the field operators creating these matter excitations. This state must be dressed, as
described above, and so if this state is created from the vacuum, the dressing will also create a
gravitational field in the complement Ū of the neighbourhood. This suggests that information
about the matter state in U is accessible, by measuring this gravitational field. However, the
preceding references show that, working perturbatively in Newton’s constant, one may choose
the dressing outside U to just depend on the Poincaré charges of the matter inside U. An analogous
statement arises in electrodynamics, where a charge distribution inside U may be dressed with
a field configuration outside U that just depends on the total charge Q of the distribution. This
may be shown, for example, by running all the field lines to a common point inside U, and then
outward in some standard configuration independent of the charge distribution.
This provides an outline of an argument for why the soft charge proposal [8–11] will not help
with the unitarity crisis. If the gravitational field outside a neighbourhood can be taken to be
independent of the detailed matter configuration inside the neighbourhood, then the soft charges
will be likewise independent. This indicates information can be localized in the neighbourhood,
at least perturbatively, despite nonlocal behaviour of gravity; we could just as well take such a
neighbourhood to be inside a BH.7 If so, the soft charges do not necessarily encode information
about the matter inside a BH. If, on the other hand, they had been found to always encode
or exhibit the information inside a BH, then that would have been an argument that standard
gravitational physics, studied more carefully, contained a possible resolution to the crisis—i.e this
would have indicated a resolution via identifying an error of reasoning.
Extending this approach to the general problem of subsystems in gravity, an initial proposal
for the corresponding mathematical structure has been discussed in [22,23,33], involving a
network of Hilbert space inclusions, Hu ⊗ Hu → H, where u, u are non-trivial indices including
information about Poincaré charges. This would replace the network of subalgebras of LQFT.
In such a structure, a continuum manifold does not necessarily play a fundamental role and
may only arise in the weak-gravity, or ‘correspondence,’ limit. This also raises a question
about the role of topology, although conceivably this mathematical structure also has pertinent
topological structure.
Clearly this discussion starts to become fairly abstract, and elaboration would lead us further
afield, but there are four takeaway messages. The first is that locality is not necessarily sacred in
quantum gravity—it is not even obvious how to define it. The second is that nonetheless there

6
For related previous constructions, see [26,27].
7
Modulo details of constructing dressings in non-trivial backgrounds.
appears to be an approximate notion of subsystems in gravity, and that a BH in particular can
6
behave as a subsystem. Third, we expect there to be some new mathematical structure providing
the foundation for quantum gravity, which is plausibly related to this subsystem structure, and

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
it should match onto the structure of LQFT on semiclassical geometry in the weak-gravity,
correspondence, limit. Finally, some consistent description of black hole-like objects should fit
into such a framework; indeed, as suggested above, the need to describe them is expected to give
key hints for the structure of the theory!

(b) Some postulates for quantum gravity

Before returning to black holes, it is useful to distil a set of principles on which to proceed, from
the preceding discussion. These can be stated in the form of three postulates for quantum gravity
(a fourth will appear in the next section). For a more careful statement of these, see [34].
Postulate I is that quantum gravity respects the principles of quantum mechanics. The essential
ones were outlined above; for further discussion, see [16].
Postulate II is that there is, at least approximately, a mathematical structure of subsystems on
the Hilbert space H. As indicated, a careful definition of this structure is a little subtle, but we will
use a suitable simplification of it below.
Postulate III is the correspondence postulate, which we have also noted: in weak-gravity limits,
the structure of the more fundamental theory should approximately reproduce the behaviour of
LQFT plus semiclassical geometry. We will seek a minimal departure from the latter behaviour.
A fourth postulate will be introduced below, in describing BH evolution.

4. Quantum black hole evolution

We now return to the problem of BHs, to which we will apply the preceding postulates.
Postulate II, Subsystems, is taken to, in particular, imply that a BH and its environment can
be thought of, at least approximately, as subsystems of a larger quantum system. As we have
outlined, the definition of such subsystems in gravity is somewhat subtle. But, as a simple model,
we will assume that the quantum states can be written in the form

|K; M; ψe , T. (4.1)

Here we consider a Schrödinger picture description, where states evolve in time T; in the
geometrical picture, such a description can be given by introducing a time slicing as illustrated in
figure 1. The label K describes the internal states of the BH; we expect there to be an enormous
number of them, corresponding to the BH entropy, with mass ≤ M:

N = eSbh ; (4.2)

most of these are expected to be in a range of masses (M, M − M) with M ∼ 1/R. The state of
the environment is described by ψe . Moreover, by Postulate III, Correspondence, this state should
have a good approximate description by a state of LQFT.
Postulate I, Quantum Mechanics, implies, in particular, that evolution of states (4.1) is unitary.
Its infinitesimal generator takes the form

H = Hbh + Henv + HI , (4.3)

where Hbh and Henv act on the BH and environment subsystems, respectively, and HI acts on both,
and can transfer information between them. In fact, LQFT evolution, on a time slicing, can be put
in this form [34]. However, LQFT evolution will not be unitary once the shrinkage of the BH is
taken into account, as we have argued above. In LQFT, HI only increases entanglement between
BH and the environment, either by creating Hawking pairs or by transferring entanglement from
the environment to the BH, resulting in the unacceptable situation described in §2. So, in order to
obey Postulate I, equation (4.3) must have another term that transfers entanglement from the BH
to the environment: BH quantum states must be able to influence their surroundings. Moreover,
the LQFT version of Hbh is not expected to give the correct dynamics, although we will be largely
7
agnostic about its exact form. Finally, in LQFT, the state on a slice in principle allows a description
of a much larger number of BH states than given by (4.2); in assuming that the states are described

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
as in (4.1) and (4.2), we are also assuming an important structural departure from the LQFT
description.
We will parametrize the required information-transferring interactions, working in an effective
picture, which describes perturbations of the background BH. In effect, we can think of the BH
in analogy to a big atom, albeit one with a very dense spectrum of states, and parametrize its
couplings to its environment.
The simplest kind of interaction that can transfer information from BH to the environment
is a product of operators acting on the respective subsystems. Let λA be the U(N) generators,
which give a basis for operators acting on BH states; these can be coupled to operators Ob (x)
acting on the environment, which are well approximated as operators of LQFT. We will adopt a
strategy of ‘parametrizing our ignorance’ by introducing an additional term in HI that is a general
superposition of such bilinear operators:


HI = dVλA Ob (x)GAb (x), (4.4)
Ab

where the integral is over a time slice in the environment; dV is the background volume
element. At this stage, the coefficients GAb (x) are arbitrary; they act as ‘structure functions’ of
the quantum BH states. However, we will constrain these functions by further consideration of
the postulates.
At this point, we should also note that, when viewed from the perspective of the semiclassical
BH geometry, such as illustrated in figure 1, the couplings (4.4) are nonlocal, at least on
scales comparable to R. This is apparently inevitable, in order to save quantum mechanics.
Of course, this raises a further concern, regarding consistency. In flat Minkowski space, transfer of
information outside the light cone can be converted, via a Lorentz transformation, into transfer of
information that is backwards in time; combining such transfers allows, in principle, an observer
to send a signal to their backward lightcone, raising causality paradoxes. However, a key point
is that this argument uses the Lorentz symmetry of Minkowski space. This is not a symmetry
of the BH background, on which we are considering propagation, and in this background, the
argument that nonlocality implies acausality no longer holds. Thus, it remains quite plausible
that such interactions of BH states with their surroundings do not necessarily introduce deep
inconsistencies associated with acausality [17,35].
As a first constraint on the couplings (4.4), we will introduce one further postulate.
Postulate IV is a statement of universality: we assume that the new interactions in (4.3) couple
universally to all matter and gauge fields (including perturbative gravitons). This seems like
a natural postulate given the universal nature of gravity. However, it can be even more
strongly motivated.
One motivation arises from considering Gedanken experiments, involving ‘black hole mining’
[36–39]. The simplest example comes from imagining that we thread a BH with a cosmic
string. In that case, the rate at which the BH loses mass increases, because there are additional
modes along the string in which Hawking radiation is produced. However, if we increase the
rate at which the BH decays without increasing the rate at which information transfers from
it, we return to the basic problem of trapped information at the end of evaporation, which
precipitated the crisis. This is avoided if the new interactions couple universally to all matter,
so introduction of new modes that can carry energy also means that these modes can transfer
information.
A second motivation comes from the desire to preserve, at least approximately, the beautiful
story of BH thermodynamics. If the new interactions coupled only to certain fields, for example,
photons or gravitons, that present problems for detailed balance, if one tries to bring a BH in
equilibrium with a thermal bath.
 The simplest way to implement universality is if the interactions couple to the energy-
8
momentum tensor, and so (4.4) is replaced by the specific form
 

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
μν
HI = dV λA GA (x)Tμν (x), (4.5)
A

where the stress tensor Tμν includes perturbative gravitons. In that case, we find that the operator-
valued quantity
 μν
Hμν (x) = λA GA (x) (4.6)
A

behaves like a metric perturbation that depends on the quantum state of the BH. The Postulates
supply additional conditions on Hμν (x).
μν
Postulate III, Correspondence, tells us that the information-transferring couplings GA (x)
should be localized near the BH; they should not, for example, describe the transfer of information
from a BH to the next galaxy, which would be a more extreme departure from LQFT. We therefore
suppose that these couplings are supported only to a radius

r = Ra = R + Ra , (4.7)

and that Ra is not enormously larger than R. Moreover, if Ra is finely tuned, e.g. to a microscopic
value Ra  R, then the couplings (4.4) or (4.5) will produce high-momentum excitations, very
near the horizon. This can be thought of as a way of modelling behaviour like that of a firewall
[40–42]. In addition to requiring an unexplained fine-tuning of Ra , this also implies a dramatic
breakdown of the geometry near what would be the horizon, and in particular, also violates
the Correspondence postulate. So, we will assume that Ra grows with R, e.g. as a power;
μν
the simplest choice is Ra ∼ R. We moreover assume that the couplings GA (x) do not vary on
microscopic scales, and e.g. only vary on scales ∼ R. Finally, also to avoid high-energy excitations
near the horizon and to minimize the departure from LQFT predictions, and particularly from BH
thermodynamics, we assume that the couplings dominantly describe transitions between states
with energy difference E ∼ 1/R.
Postulate I, and specifically unitarity, implies additional conditions. The couplings (4.5) (or
(4.4)) are required to transfer information from the BH at a rate that overcomes the Hawking
entanglement growth, and so at a rate of size8 dI/dt ∼ 1 qubit/R. In this sense, the required
information transfer is a significant effect, as emphasized by [43–45].
The simplest way to achieve this rate [46] is if the metric perturbation in a typical BH state,

Hμν (x) = ψ, T|Hμν (x)|ψ, T, (4.8)

which is dimensionless, is also O(1). This follows essentially from dimensional analysis, if the
spatial and temporal variation scales of Hμν (x) are all of order R, and is further discussed in [46].
However, we can ask if Hμν (x) = O(1) is required to achieve sufficient information transfer.
Here we encounter a general problem in quantum information theory. Suppose we have two
subsystems A and B, with A much smaller than B. Suppose that the two systems evolve by
a hamiltonian
H = HA + HB + HI , (4.9)

where HA , HB act only on their respective subsystems, and HI couples the two. Suppose moreover
that the dynamics of HA and HB are sufficiently random to distribute information efficiently. Then,
how does the information transfer rate from A to B depend on the couplings in HI and on the
energy scales? In ref. [34] a rate quadratic in the couplings was conjectured; this has been checked
in simple toy models in [47].
For our current BH purposes, we will estimate the rate by noting that (4.5) (or (4.4)) cause
transitions where the BH emits a quantum. When it does so, we expect that O(1) qubit of
8
Here we can define the BH’s information I in terms of its entanglement with the environment; this information must decrease
to zero by the time the BH decays.
information can be transferred. So, the information transfer rate is expected to be approximated
9
by the rate at which transitions occur. The latter can be estimated by Fermi’s Golden Rule, so

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
dI dP
∼ = 2πρ(Ef )|HI |2 , (4.10)
dT dT
where |HI | is the typical size of a matrix element of HI . We require this rate to be the relatively
large value ∼ 1/R. However, the density of final states ρ(Ef ) in (4.10) includes a factor of the
density of BH states that we transition to, and the latter is expected to take the enormous value
∼ N of (4.2). That means that the typical matrix elements of HI can be correspondingly small,
1
|HI | ∼ √ = e−Sbh /2 (4.11)
N
—a tiny value! Moreover, if Hμν behaves like a generic operator on the space of BH states, it has
a typical expectation value that is correspondingly exponentially small,

Hμν (x) = O(e−Sbh /2 ). (4.12)

In summary, we have found two scenarios for such ‘quantum halos’ around BHs, in which
interactions transfer information from BH quantum states to the BHs environment. The simplest,
with Hμν  ∼ 1 we refer to as the ‘strong/coherent’ scenario; here the perturbation in the effective
metric is substantial and behaves like a classical perturbation to the metric. The minimal, with
Hμν  ∼ exp{−Sbh /2}, we refer to as the ‘weak/incoherent’ scenario. Both of these scenarios raise
some important questions.
The first is how to understand such effects, from a more fundamental perspective. Here
there is more to learn, but the preceding discussion has hinted at a modification of how
one thinks about locality and transfer of information, compared to how it is described in a
semiclassical background geometry. Fundamental gravitational physics is likely not based on
classical spacetime, which only arises approximately, in a certain limit; taking spacetime too
literally may lead to errors in the description, which are accounted for by the effects we are
parametrizing. A deeper understanding likely requires a more intrinsically quantum view of
information localization and transfer, and of spacetime itself. If, as expected, it is true that the
fundamental description of physics does not involve spacetime and instead involves different
structure on Hilbert space, it may be that a description of the departures of this structure from
a spacetime description cannot be simply described in spacetime terms. Some initial exploration
of these questions appears in [22,23]. A related question is whether some of these ideas might be
actually realized in the AdS/CFT correspondence.
A second question is whether such unitarizing effects, since they apparently need to reach well
outside the horizon, might have any observational consequences.

5. Observational probes
(a) Weak/incoherent scenario
In the weak/incoherent scenario, the effective metric perturbations are tiny, as in (4.12). Since they
do not behave like large, classical fluctuations, that suggests that they have negligible effect for
astrophysical black holes.
However, to take a closer look, consider the scattering of an excitation, e.g. a photon, from such
a quantum halo. This scattering probability can also be estimated by Fermi’s Golden rule [48], and
so will be of the form
 2
dP  
= 2πρ(Ef )  dVK|Hμν |ψβ|Tμν |α , (5.1)
dt

where |α and |β are the initial and final states of the excitation. While the matrix element of
Hμν is again exponentially small, that is once again compensated by the enormous density of
states. Thus, this probability can also be of order 1/R – and correspondingly there can be O(1)
10
modifications to the scattering cross section.
However, we saw that the Correspondence postulate implied that Hμν has spatial and

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
temporal variation scales ∼ R, and thus the momentum transfer in such scattering is p ∼ 1/R.
Consider the situation for the electromagnetic image of M87. EHT is observing photons of
wavelength ∼ 1 mm. However, the radius R for the central object in M87 is ∼ 2 × 1010 km. So,
a momentum transfer of this size is utterly negligible for such photons.
Note, though, that in the case of gravitational radiation from BH mergers, radiation
wavelengths comparable to the BH size contribute importantly. At such wavelengths, this
scattering can be a significant effect. This suggests that the weak/coherent scenario can be probed
by investigating the effect of such modifications, e.g. to absorption cross sections, on observed
gravitational wave signals.

(b) Strong/coherent scenario

If the strong/coherent scenario is correct, its effects could be even more pronounced. This scenario
predicts fluctuations in the effective metric, of size gμν ∼ 1, extending a distance ∼ R from the
horizon, and fluctuating on spatial and temporal scales ∼ R. Such classical-like fluctuations can
have important effects on light propagation, and, in particular, on EHT images. Simulations of
the image distortions, with a simple model for such perturbations, indeed reveals the possibility
of dramatic effects for sufficiently large perturbations [49]. A particular observational target is the
temporal variation, on time scales given by the typical frequencies of the effective perturbations.
If one assumes these fluctuations should lead to a minimal departure from Hawking’s
description of BH evolution, and, in particular, should not greatly alter the thermal spectrum
of the BH emission [48,49], the frequencies of the fluctuations should be of comparable to the
thermal frequency,

1 1 − a2
ω ∼ ωT =  (5.2)
4π M 1 + 1 − a2

for spin parameter a = J/M2 . The two main targets of EHT are Sgr A∗ and M87, for which the
corresponding periods are
 

2π M 1 1
P=  0.93 +  hr
ωT 4.3 × 106 M 2 2 1 − a2
 

M 1 1
 59 +  d. (5.3)
6.5 × 109 M 2 2 1 − a2

The period for Sgr A∗ is less than the EHT image scan time of several hours, making it hard to
resolve such temporal fluctuation. This suggests the best sensitivity to such variation is from M87
observations.
The recently released [3] EHT image spans a period of 7 days, which is short as compared to the
M87 period (5.3); we therefore do not expect it to strongly bound such fluctuations, although some
modest bounds appear to be possible given that the 7-day images are close to what is predicted
from classical GR [48]. Clearly, longer duration observations will be of interest to increase
such sensitivity.

(c) General comments

While the preceding discussion has sought to provide strong motivation for a picture of BH
evolution, it is worth considering the situation even more broadly. The past few years have
seen the opening of two observational windows on a regime for which there were previously
none: the strong gravity regime, near what is classically expected to be a BH horizon. Over a
slightly longer time, the theoretical community has also reached a widespread (though still not
universal) view that to reconcile BHs with quantum mechanics, modifications of a description
11
via LQFT on semiclassical geometry must reach out to scales comparable to the horizon size.
While the preceding effective description of quantum dynamics seems perfectly reasonable, any

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
other alternative should also extend into the region near the horizon. Thus, it is worth actively
investigating possible observational constraints on any such scenario, if it is sufficiently well
formulated to start to make predictions.

6. Conclusion
The problem of quantum BH evolution suggests that there is something deeply wrong with
a semiclassical, geometrical description of BHs, and that BHs are likely intrinsically quantum
objects at horizon scales. Indeed, in an inherently quantum-mechanical approach to physics,
spacetime may not play a fundamental role, and may be only approximately correct, and the
strong gravity regime of a BH may begin to particularly reveal this problem. If we adopt quantum
mechanics as a postulate, and demand unitary evolution, we seem to be directly led to a picture
where a BH must have interactions with its environment that are outside the standard description
of BHs. Interestingly, the necessary new couplings can be much smaller than naively expected
and still save quantum mechanics. Whether large or small, they potentially have observational
consequences, and it is of great interest to investigate these further.
Data accessibility. This article has no additional data.
Competing interests. I declare I have no competing interests.
Funding. This material is based in part upon work supported in part by the U.S. Department of Energy,
Office of Science, under Award no. DE-SC0011702.
Acknowledgements. The author thanks the CERN theory group, where this work was carried out, for its
hospitality.

References
1. GRAVITY Collaboration, Abuter R et al. 2018 Detection of orbital motions near the last
stable circular orbit of the massive black hole Sgr A*. A & A 618, L10. (doi:10.1051/
0004-6361/201834294)
2. LIGO Scientific and Virgo Collaboration, Abbott BP et al. 2016 Observation of gravitational
waves from a binary black hole merger. Phys. Rev. Lett. 116, 061102. (doi:10.1103/
PhysRevLett.116.061102)
3. Event Horizon Telescope Collaboration, Akiyama K et al. 2019 First M87 event horizon
telescope results. I. The shadow of the supermassive black hole. Astrophys. J. 875, L1.
(doi:10.3847/2041-8213/ab0ec7)
4. Hawking SW. 1975 Particle creation by black holes. Commun. Math. Phys. 43, 199–220.
doi:10.1007/BF02345020)[Erratum, Commun. Math. Phys. 46 (1976) 206].
5. Giddings SB, Nelson WM. 1992 Quantum emission from two-dimensional black holes. Phys.
Rev. D 46, 2486–2496. (doi:10.1103/PhysRevD.46.2486)
6. Giddings SB. 1995 Why aren’t black holes infinitely produced? Phys. Rev. D 51, 6860–6869.
(doi:10.1103/PhysRevD.51.6860)
7. Susskind L. Trouble for remnants. (http://arxiv.org/abs/hep-th/9501106)
8. Hawking SW. The information paradox for black holes. (http://arxiv.org/abs/1509.01147)
9. Hawking SW, Perry MJ, Strominger A. 2016 Soft hair on black holes. Phys. Rev. Lett. 116,
231301. (doi:10.1103/PhysRevLett.116.231301)
10. Hawking SW, Perry MJ, Strominger A. 2017 Superrotation charge and supertranslation hair
on black holes. JHEP 5, 161. (doi:10.1007/JHEP05(2017)161)
11. Haco S, Hawking SW, Perry MJ, Strominger A. 2018 Black hole entropy and soft hair. JHEP
12, 098. (doi:10.1007/JHEP12(2018)098)
12. Giddings SB. 2013 The gravitational S-matrix: Erice lectures. Subnucl. Ser. 48, 93–147.
(doi:10.1142/9789814522489_0005)
13. Hawking SW. 1976 Breakdown of predictability in gravitational collapse. Phys. Rev. D 14,
2460–2473. (doi:10.1103/PhysRevD.14.2460)
14. Banks T, Susskind L, Peskin ME. 1984 Difficulties for the evolution of pure states into mixed
12
states. Nucl. Phys. B 244, 125–134. (doi:10.1016/0550-3213(84)90184-6)
15. Weinberg S. 1989 Testing quantum mechanics. Ann. Phys. 194, 336–386. (doi:10.1016/

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
0003-4916(89)90276-5)
16. Giddings SB. 2008 Universal quantum mechanics. Phys. Rev. D 78, 084004. (doi:10.1103/
PhysRevD.78.084004)
17. Giddings SB. 2012 Black holes, quantum information, and unitary evolution. Phys. Rev. D 85,
124063. (doi:10.1103/PhysRevD.85.124063)
18. Giddings SB. 2015 Hilbert space structure in quantum gravity: an algebraic perspective. JHEP
12, 1–21. (doi:10.1007/JHEP12(2015)099)
19. Cao C, Carroll SM, Michalakis S. 2017 Space from hilbert space: recovering geometry from
bulk entanglement. Phys. Rev. D 95, 024031. (doi:10.1103/PhysRevD.95.024031)
20. Cao C, Carroll SM. 2018 Bulk entanglement gravity without a boundary: towards
finding Einstein’s equation in Hilbert space. Phys. Rev. D 97, 086003. (doi:10.1103/
PhysRevD.97.086003)
21. Carroll SM, Singh A. Mad-Dog everettianism: quantum mechanics at its most minimal.
(http://arxiv.org/abs/1801.08132)
22. Giddings SB. 2019 Quantum-first gravity. Found. Phys. 49, 177–190. (doi:10.1007/s10701-019-
00239-1)
23. Giddings SB. 2018 Quantum gravity: a quantum-first approach. LHEP 1, 1–3. (doi:10.31526/
LHEP)
24. Haag R. 1992 Local quantum physics: fields, particles, algebras. Texts and Monographs in Physics.
Berlin, Germany: Springer.
25. Donnelly W, Giddings SB. 2016 Diffeomorphism-invariant observables and their nonlocal
algebra. Phys. Rev. D 93, 024 030. (doi:10.1103/PhysRevD.93.024030)
26. Heemskerk I. 2012 Construction of bulk fields with gauge redundancy. JHEP 1209, 106.
(doi:10.1007/JHEP09(2012)106)
27. Kabat D, Lifschytz G. 2014 Decoding the hologram: scalar fields interacting with gravity. Phys.
Rev. D 89, 066 010. (doi:10.1103/PhysRevD.89.066010)
28. Giddings SB, Lippert M. 2002 Precursors, black holes, and a locality bound. Phys. Rev. D 65,
024 006. (doi:10.1103/PhysRevD.65.024006)
29. Giddings SB, Lippert M. 2004 The Information paradox and the locality bound. Phys. Rev. D
69, 124 019. (doi:10.1103/PhysRevD.69.124019)
30. Giddings SB. 2006 Locality in quantum gravity and string theory. Phys. Rev. D 74, 106 006.
(doi:10.1103/PhysRevD.74.106006)
31. Donnelly W, Giddings SB. 2017 How is quantum information localized in gravity? Phys. Rev.
D 96, 086 013. (doi:10.1103/PhysRevD.96.086013)
32. Donnelly W, Giddings SB. 2018 Gravitational splitting at first order: quantum information
localization in gravity. Phys. Rev. D 98, 086 006. (doi:10.1103/PhysRevD.98.086006)
33. Giddings SB. Gravitational dressing, soft charges, and perturbative gravitational splitting.
(http://arxiv.org/abs/1903.06160)
34. Giddings SB. 2017 Nonviolent unitarization: basic postulates to soft quantum structure of
black holes. JHEP 12, 047. (doi:10.1007/JHEP12(2017)047)
35. Giddings SB. 2011 Nonlocality versus complementarity: a Conservative approach to
the information problem. Class. Quant. Grav. 28, 025002. (doi:10.1088/0264-9381/28/2/
025002)
36. Unruh WG, Wald RM. 1983 How to mine energy from a black hole. Gen. Relat. Gravit. 15,
195–199. (doi:10.1007/BF00759206)
37. Lawrence AE, Martinec EJ. 1994 Black hole evaporation along macroscopic strings. Phys. Rev.
D 50, 2680–2691. (doi:10.1103/PhysRevD.50.2680)
38. Frolov VP, Fursaev D. 2001 Mining energy from a black hole by strings. Phys. Rev. D 63,
124010. (doi:10.1103/PhysRevD.63.124010)
39. Frolov VP. 2002 Cosmic strings and energy mining from black holes. Int. J. Mod. Phys. A 17,
2673–2676. (doi:10.1142/S0217751X0201159X)
40. Giddings SB. 1994 Quantum mechanics of black holes. In Proc. Summer School in High-Energy
Physics and Cosmology: Trieste, Italy, June 13–July 29, 1994, pp. 0530–574. (http://arxiv.org/abs/
hep-th/9412138)
41. Braunstein SL, Pirandola S, Życzkowski K. Better late than never: information retrieval from
13
black holes. Phys. Rev. Lett. 110, 101301. (doi:10.1103/PhysRevLett.110.101301)
42. Almheiri A, Marolf D, Polchinski J, Sully J. 2013 Black holes: complementarity or firewalls?

royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 377: 20190029

................................................................
JHEP 2, 062. (doi:10.1007/JHEP02(2013)062)
43. Page DN. 1993 Average entropy of a subsystem. Phys. Rev. Lett. 71, 1291–1294.
(doi:10.1103/PhysRevLett.71.1291)
44. Page DN. 1993 Information in black hole radiation. Phys. Rev. Lett. 71, 3743–3746.
(doi:10.1103/PhysRevLett.71.3743)
45. Mathur SD. 2009 The Information paradox: a pedagogical introduction. Class. Quant. Grav. 26,
224001. (doi:10.1088/0264-9381/26/22/224001)
46. Giddings SB. 2014 Modulated Hawking radiation and a nonviolent channel for information
release. Phys. Lett. B 738, 92–96. (doi:10.1016/j.physletb.2014.08.070)
47. Giddings SB, Rota M. 2018 Quantum information or entanglement transfer between
subsystems. Phys. Rev. A 98, 062329. (doi:10.1103/PhysRevA.98.062329)
48. Giddings SB. Searching for quantum black hole structure with the Event Horizon Telescope.
Universe 5, 201. (doi:10.3390/universe5090201)
49. Giddings SB, Psaltis D. 2018 Event horizon telescope observations as probes for
quantum structure of astrophysical black holes. Phys. Rev. D 97, 084035. (doi:10.1103/
PhysRevD.97.084035)