Kerr–Vaidya black holes

Pravin Kumar Dahal

Department of Physics & Astronomy, Macquarie University, Sydney New South Wales 2109, Australia

Daniel R. Terno
Department of Physics & Astronomy, Macquarie University, Sydney New South Wales 2109, Australia and
Shenzhen Institute for Quantum Science and Engineering, Department of Physics,
Southern University of Science and Technology, Shenzhen 518055, Guandong, China

Kerr–Vaidya metrics are the simplest nonstationary extensions of the Kerr metric. We explore their proper-
ties and compare them with the near-horizon limits of the spherically symmetric self-consistent solutions (the
ingoing Vaidya metric with decreasing mass and the outgoing Vaidya metric with increasing mass) for the evap-
arXiv:2008.13370v3 [gr-qc] 15 Dec 2020

orating and accreting physical black holes. The Newman–Janis transformation relates the corresponding Vaidya
and Kerr-Vaidya metrics. For nonzero angular momentum, the energy-momentum tensor violates the null en-
ergy condition (NEC). However, we show that its structure differs from the standard form of the NEC-violating
tensors. The apparent horizon in the outgoing Kerr–Vaidya metric coincides with that of the Kerr black hole.
For the ingoing metric, its location is different. We derive the ordinary differential equation for this surface and
locate it numerically. A spherically symmetric accreting black hole leads to a firewall — a divergent energy
density, pressure, and flux as perceived by an infalling observer. We show that this is also true for the outgoing
Kerr–Vaidya metric.

I. INTRODUCTION hole as the final product of the gravitational collapse [4, 5].
These objects also require a violation of the energy conditions
Black holes are described both as “the most perfect macro- for their existence. Another resolution of the information loss
scopic objects in the universe” [1] and as one of the “most paradox posits that the infalling observer (Alice) does not see
mysterious concepts conceived by the human mind” [2]. a vacuum at the black hole horizon, but instead encounters a
Thanks to the successes of the gravitational wave astronomy large number of high-energy modes [16, 21], known as the
and direct observations of ultracompact objects (UCOs), the firewall.
old debate about the physical relevance of black hole solu- A self-consistent approach [13, 22–24] starts with the as-
tions [3] has been reframed as a question about the nature of sumption that physical black holes do form. Once the as-
UCOs [4–6]. sumption of formation of a singularity-free apparent horizon
Diversity of opinions about black holes and their signifi- is translated into mathematical statements, it allows to obtain
cance are matched by absence of a universally accepted defi- a number of concrete results. In spherical symmetry there
nition [7]. However, the core idea of a black hole as a space- are only two possible classes of black hole solutions, and it
time region from which nothing can escape is formalized in is possible to identify the amount of violation of the energy
the notion of a trapped region. Gravity there is so strong that conditions that they require. Accreting physical black hole
both ingoing and outgoing future-directed null geodesics orig- solutions lead to divergent energy density, pressure, and flux
inating at a spacelike two-dimensional surface with spherical as experienced by Alice, while the curvature scalars remain
topology have negative expansion [8–10]. The apparent hori- finite.
zon is its evolving outer boundary. Real astrophysical objects are rotating. Hence it is impor-
A physical black hole contains such a trapped region [12]. tant to verify that the firewall is not an artifact of the spherical
To be relevant to distant observers with a finite lifespan it symmetry. The Kerr metric is the asymptotic result of the clas-
has to be formed in finite time according to their clocks [13]. sical collapse [1, 8, 9]. The simplest models that allow for an
Otherwise, black hole solutions can have only approximate or axially symmetric variable mass distribution are given by the
asymptotic meaning. A physical black hole may possess other so-called Kerr–Vaidya metrics [25–27]. In Sec. II we review
classical features, such as an event horizon and a singularity, the relevant properties of the spherically symmetric solutions.
or be a singularity-free regular black hole. One of the issues at In Sec. III we discuss their axially symmetric counterparts,
stake is whether the observed astrophysical black hole candi- focusing on the violation of the energy conditions, location of
dates contain light-trapping regions, i.e., they are black holes the apparent horizons and presence of a firewall.
or do not, and thus they are horizonless UCOs.
Quantum effects make the black hole physics particularly
interesting [9–11, 14–17]. On the one hand, an apparent hori- II. NEAR-HORIZON REGIONS OF SPHERICALLY
zon is accessible to an observer at infinity (Bob) only if the SYMMETRIC BLACK HOLES
classical energy conditions [8, 18–20] are violated [8]. The
Hawking radiation [9, 11, 14] has precisely this property. On Working in the framework of semiclassical gravity [15, 28,
the other hand, the Hawking radiation precipitates the infa- 29] we use classical notions (horizons, trajectories, etc.), and
mous information loss paradox [16, 17]. One way to resolve describe dynamics via the Einstein equations Gµν = 8πTµν ,
the paradox is to have a horizonless UCO or a regular black where the Einstein tensor Gµν is equated to the expectation
 2

value Tµν = hT̂µν iω of the renormalized energy-momentum Using the Einstein equations and the relationships between
tensor (EMT). For simplicity we consider an asymptotically components of the EMT in two coordinates systems [24] one
flat space. We do not make any specific assumptions apart can show that
from (i) the apparent horizon was formed at some finite time
of Bob, (ii) it is regular, i.e., the curvature scalars, such as C+ (v, r) = r+ (v) + w1 (v)x + . . . , (8)
T := T µµ ≡ −R/8π and T := T µν Tµν ≡ Rµν Rµν /64π 2 are h+ (v, r) = χ1 (v)x + . . . , (9)
finite at the horizon. (Here Rµν and R := Rµµ are the Ricci
tensor and the Ricci scalar, respectively). where r+ (v) is the radial coordinate of the apparent horizon,
A general spherically symmetric metric in the C+ (v, r+ ) ≡ r+ , x =: r − r+ (v).As a result, at the apparent
Schwarzschild coordinates is given by horizon both the metric corresponds to the Vaidya geometry
′
with C+ (v) = 2M ′ (v) < 0.
ds2 = −e2h(t,r)f (t, r)dt2 + f (t, r)−1 dr2 + r2 dΩ, (1) ′
If rg > 0 it is useful to switch to the retarded null coordinate
u. The near-horizon geometry is then described by the Vaidya
where r is the areal radius. The Misner-Sharp mass [10, 30] metric with C− ′
(u) > 0.
M (t, r) is invariantly defined via 1 − 2M/r := ∂µ r∂ µ r, and A static observer finds that the energy density ̺ =
f (t, r) = 1 − 2M (t, r)/r. The apparent horizon is located at Tµν uµ uν = −T tt , the pressure p = Tµν nµ nν = T rr , and
the Schwarzschild radius rg that is the largest root of f (t, r) = the flux φ := Tµν uµ nν (where uµ is the four-velocity and nµ
0 [10, 31]. is the outward-pointing radial spacelike vector), diverge at the
Only two near-horizon forms of the EMT and the metric are apparent horizon. A radially infalling Alice moves on a trajec-
consistent with the above two assumptions. Here we consider tory xµA (τ ) = (T (τ ), R(τ ), 0, 0). Horizon crossing
happens
the generic form that agrees with the ab initio calculations not only at some finite proper time τ0 , rg T (τ0 ) = R(τ0 ),
of the EMT on the background of the Schwarzschild solution but thanks to the form of the metric also at a finite time T (τ0 )
[32]. In this case the leading terms in the expansion of the of Bob.
metric functions in terms of x := r − rg (t) are However, experiences of Alice are different at the apparent
√ horizon of an evaporating and accreting physical black holes.
2M (t, r) = rg − w x + O(x), (2) For an evaporating black hole, rg′ < 0, energy density, pres-
1 x √
sure and flux are finite. For example, if we approximate the
h(t, r) = − ln + O x , (3) near-horizon geometry by a pure outgoing Vaidya metric with
2 ξ
M ′ (v) < 0, Alice’s energy density at the horizon crossing is
where the function ξ(t) is determined by the choice of the time
variable (and requires for its determination knowledge of the Υ2
̺< < <
A = pA = φA = − , (10)
full solution of the Einstein equations), and w2 := 16πΥ2 rg3 4Ṙ2
characterizes the leading behavior of the EMT [13].
In particular, in the orthonormal basis the (t̂r̂) block of the at rg = R. For an accreting black hole, rg′ > 0, Alice ex-
EMT near the apparent horizon is given by periences the divergent values of energy density, pressure and
flux,
Υ2 1 ±1
 
Tâb̂ = − . (4) 2Ṙ2 Υ2
f ±1 1 ̺> > >
A = pA = −φA = − + O(F −1 ), (11)
F2
The upper (lower) signs of Tt̂r̂ correspond to evaporation
in the vicinity of the apparent horizon, as F := f (T, R) → 0.
(growth) of the physical black hole. Consistency of the Ein-
Thus an expanding trapped region is accompanied by a
stein equations results in the relation
firewall— a region of unbounded energy density, pressure,
p √ √ and flux—that is perceived by an infalling observer. Unlike
rg′ / ξ = ±4 π Υ rg = ±w/rg . (5)
the firewall from the eponymous paradox, it appears as a con-
The null energy condition (NEC) requires Tµν lµ lν > 0 for all sequence of regularity of the expanding apparent horizon and
null vectors lµ [8, 19, 20]. It is violated by radial vectors lâ = its finite formation time. The divergent energy density leads to
(1, ∓1, 0, 0) for the evaporating and the accreting solutions, a violation [23, 24] of the inequality that bounds the amount
respectively [13] . of negative energy along a timelike trajectory in a moderately
Null coordinates allow to represent the near-horizon geom- curved spacetime [33]. As a result a physical black hole, once
etry in a simpler form. The advanced null coordinate v, formed, can only evaporate. Another possibility is that the
semiclassical physics breaks down at the horizon scale.
dt = e−h (eh+ dv − f −1 dr), (6)

is useful in the case rg′ < 0. A general spherically symmetric III. KERR–VAIDYA METRIC
metric in (v, r) coordinates is given by
  A general time-dependent axisymmetric metric contains
2 2h+ C+ seven functions of three variables (say, t, r and θ) that en-
ds = −e 1− dv 2 + 2eh+ dvdr + r2 dΩ. (7)
r ter the Einstein equations via six independent combinatons
 3

[1]. However, to verify that certain predictions of the self- vector e2̂ ∝ ∂θ and the remaining vector e3̂ is found by com-
consistent approach (such as a firewall at an expanding appar- pleting the basis, the EMT takes the form
ent horizon) are not an artefact of the spherical symmetry, it is  
enough to consider a simpler geometry. ν ν q 2̂ q 3̂
The Kerr metric can be represented using either the ingoing  ν ν q 2̂ q 3̂ 
 
[34] T âb̂ =  2̂ 2̂ . (16)
 q q 0 0 

2M r

4aM r sin2 θ q 3̂ q 3̂ 0 0
ds2 = − 1 − 2 dv 2 + 2dvdr − dvdψ
ρ ρ2 Explicit expressions for the tetrad vectors and the matrix ele-
2 2 2 2 2
(r + a ) −a ∆ sin θ ments are given in Appendix B.
− 2a sin2 θdrdψ+ ρ2 dθ2 + sin2 θdψ 2 ,
ρ2 For an arbitrary null vector lâ = (−1, nâ ), nâ =
(12) (cos α, sin α cos β, sin α sin β) the NEC becomes

or the outgoing null congruences [1], ν(1 − cos α) + 2 sin α(q 2̂ cos β + q 3̂ sin β) > 0. (17)

4aM r sin2 θ
 
2M r This inequality is satisfied if and only if ν > 0 and q 2̂ =
2
ds = − 1 − 2 du2 − 2dudr − dudψ
ρ ρ2 q 3̂ = 0. The condition q 2̂ = q 3̂ = 0 holds only when a = 0,
2 2 2 2 2 so the metric reduces to its Vaidya counterpart and the EMT
(r + a ) −a ∆ sin θ
+ 2a sin2 θdψdr+ ρ2 dθ2 + 2
sin2 θdψ 2 , becomes a type II tensor. Only in this case the NEC may be
ρ
(13) satisfied.
Each type of the EMT is characterized by its Lorentz-
where ρ 2 := 2 2 2 2
r + a cos θ, ∆ := r − 2M r + a , and 2 invariant eignevlaues [18, 19]. These are the eigenvalues of
a = J/M is the angular momentum per unit mass. the matrix T âb̂ , i.e., the roots of the equation
The easiest way to obtain this result is to follow the
complex-valued Newman–Janis transformation [35] starting det(T âb̂ − λη âb̂ ) = 0, η âb̂ = diag(−1, 1, 1, 1). (18)
with the Schwartzschild metric written in the ingoing or the
outgoing Eddington–Finkelstein coordinates. The simplest The EMT of Eq. (16) has a single quadruple-degenerated
nonstationary generalizations of the Kerr metric are obtained Lorentz-invariant eigenvalue λ = 0. On the other hand, two
by introducing evolving masses M (v) and M (u). The metric of the eigenvalues of the matrix λ̃ of T âb̂ are nonzero,
of Eq. (13) with a variable M (u) is obtained from the retarded q
Vaidya metric [36, 37]. By using the advanced Vaidya metric λ̃1,2 = ν ± 2(q2̂2 + q3̂2 ) + ν 2 . (19)
of Eq. (7) as the seed metric, the metric of Eq. (12) can be
obtained following the same procedure (Appendix A).
As a result the EMT tensor (16) cannot be brought to a generic
type III form by an arbitrary similarity transformations unless
λ̃1 = −λ̃2 , which is impossible for M ′ 6= 0 (see Appendix B
A. Energy conditions
for the details).
A schematic form of the EMT in both cases is
B. Apparent horizon
Too 0 Toθ Toψ
 
 0 0 0 0 
Tµν =  , (14) The apparent horizon of the Kerr black hole coincides with
Toθ 0 0 Tθψ 
Toψ 0 Tθψ Tψψ its event horizon. It is located at the largest root of ∆ = 0,
p
where o = u, v. Using the null vector k µ = (0, 1, 0, 0) [25] r0 := M + M 2 − a2 . (20)
the EMT can be represented as
For both the ingoing and the outgoing Vaydia metrics the ap-
Tµν = Too kµ kν + qµ kν + qν kµ , (15) parent horizon is located at rg = r0 = 2M . For the metric
(13) the relation rg = r0 also holds [39], but it fails for the
where the components of Tµν and of the auxiliary vector qµ , metric (12) [27]. In this case the difference rg (v, θ) − r0 (v) is
qµ k µ = 0, for both cases are given in Appendix B. The EMT of the order |Mv |.
[for the metric Eq. (13)] was identified in Ref. [38] as belong- We now identity its location in the foliations with the hy-
ing to the type [(1, 3)] in the Segre classification [18], i.e., to persurfaces v = const. The standard approach [40, 41] for
the type III of the Hawking–Ellis classification [8, 19], indi- constructing the ordinary differential equation for the appar-
cating that the NEC is violated for any a 6= 0. ent horizon is based on exploiting properties of a spacelike
A detailed investigation reveals some interesting properties foliation. It cannot be used in this case as the foliating hyper-
of this EMT. We use a tetrad in which the null eigenvector surfaces are timelike. However, since the approximate loca-
k µ = k â eµâ has the components k â = (1, 1, 0, 0), the third tion of the apparent horizon is known, we obtain the leading
 4

z×10-5 to the families of affinely parametrized null geodesics in the

3.5 bulk.
3. One approach to calculation of the expansion uses its ge-
ometric meaning as a relative rate of change of the two-
2.5 dimensional cross section area [8, 34]. Consider an infinites-
z /zmax
′

imal geodesic triangle (xin xθ xψ ) on the surface S0 . It is de-

2. 1
0.5
fined by three vertices
1.5 θ
π/2 π xµin , xµψ = xµin + bµψ δψ, xµθ = xµin + bµθ δθ. (21)
1. -0.5
-1
0.5
The two tangent vectors

θ ∂xµ ∂xµ
0 π/4 π/2 3π/4 π bµψ := , bµθ := , (22)
∂ψ S0 ∂θ S0

FIG. 1. Location of the apparent horizon relative to r0 for M = 1, introduce a metric two-tensor
a = 0.1, Mv = −κ/M 2 = 0.01. The equation was first solved
as a boundary value problem z(0) = 0, z ′ (π/2) = 0, resulting in
σAB := gµν bµA bνB , A, B = θ, ψ. (23)
zm := z(π/2) ≈ 3.57 × 10−5 . Solution of the initial value problem
z(π/2) = zm , z ′ (π/2) = 0 coincides with the previous one within √
the relative precision of 2 × 10−15 outside δ = 10−6 interval from The area of the triangle (xin xθ xψ ) is δA = 21 σδθδψ, where
2
the poles. σ = σ11 σ22 − σ12 is the determinant of the two-dimensional
metric. Under the geodesic flow xµ → xµ + lµ (x)dλ the
coordinates θ and ψ are comoving and thus constant for the
three vertices, but both the metric g and the vectors bA evolve
correction in Mv by using the methods of analysis of null con- with λ.
gruences and hypersurfaces [34]. Calculating the ratio of the first-order area change δA(λ +
Assume that at some advanced time v the apparent horizon dλ) − δA(λ) to the initial area of the triangle allows to obtain
S0 (v) is located at rg = r0 (v) + z(θ), where the function the expansion by using the relation
z(θ) is to be determined. Once the future-directed outgoing √
1 d σ
null geodesic congruence orthogonal the surface is identified, ϑ= √ , (24)
calculating the expansion ϑ and equating it to zero results in σ dλ
the differential equation for z(θ). There are at least two equiv-
where λ is the affine parameter. Solution of the second-order
alent ways to obtain this equation.
differential equation ϑ(z) = 0 gives the desired function z(θ).
The outward and inward-pointing null vectors lµ and N µ , An alternative derivation is based on the direct evaluation of
µ
respectively, are defined on S0 . They are orthogonal to its ϑ = l;µ on S0 (see Appendix C for the details).
tangents and can be normalized by the condition Nµ lµ = −1. If both z(θ) and its derivatives are much smaller than r0 ,
These vectors can be extended to the fields of tangent vectors we obtain a linear ordinary differential equation for z,

8r02 (a2 + r02 )2 (z ′′ + cot θz ′ ) − (r02 − a2 ) a4 + 7a2 r02 + 8r04 + a2 (r02 − a2 ) cos 2θ z − 8r03 a2 (a2 + r02 ) sin2 θMv = 0, (25)

where the regular singular term (cot θz ′ + z ′′ ) is a standard the assumption |z ′ | ≪ r0 is clearly valid, leads to
feature of the apparent horizon equations [40, 41]. Symmetry
considerations lead to the condition z(0) = z(π) = 0. The a2 Mv r0 4
zser = θ + O(θ5 ). (26)
second initial condition z ′ (π) = 0 ensures that the surface of 16(a2 + r02 )
the apparent horizon is smooth [40].
For Mv > 0 it implies that at least near the poles rg > r0 ,
i.e., at r = r0 the expansion is still negative. However, this is
impossible: at the poles the null congruence that is orthogonal
A typical result of the numerical solution is depicted in to the two-dimensional surface r = r0 [27] has ϑ > 0. More-
Fig. 1. It was obtained by imposing the boundary conditions over, using this solution to provide the initial values z(θ),
z(0) = 0 and z ′ (π/2) = 0, that enforces the equatorial sym- z ′ (θ) at some θ = ǫ ≪ 1 leads to inconsistencies.
metry. The assumption of |z ′ | ≪ r0 fails near the poles, where We investigated stability of this result in numerical exper-
z = 0. This is not an artefact of the approximation. Using the iments. For a fixed Mv = −κ/M 2 the initial value problem
series solutions of Eq. (25) with the conventional initial con- z(π/2) = z0 , z ′ (π/2) = 0, where z0 is some number, leads
ditions z(0) = 0, z ′ (0)=0 [40, 41], i.e., in the regime where to a well-behaved numerical solution. However, the condi-
 5

tions z(0) = z(π) = 0 are satisfied within a prescribed tol- It is easy to see that for a = 0 we recover the firewall of the
erance only for a very narrow range of the values z0 around outgoing Vaidya metric.
zm = z(π/2) of the numerical solution of the above boundary Violations of the NEC are bounded by quantum energy in-
value problem. We will provide a full analysis of the apparent equalities (QEIs) [20, 42]. For spacetimes of small curvature
horizon in a future work. explicit expressions that bound time-averaged energy density
for a geodesic observer were derived in Ref. [33]. For any
Hadamard state ω and a sampling function f(τ ) of compact
C. Firewall support, negativity of the expectation value of the energy den-
sity ρ = hT̂µν iω uµ uν as seen by a geodesic observer that
All components of the EMT (14) are finite at the apparent moves on a trajectory γ(τ ) is bounded by
horizon. Divergences of the comoving parameters can appear Z
only as a result of divergences in the components of the four- f2 (τ )ρdτ > −B(R, f, γ), (34)
velocity of Alice. We now show that similarly to the spher- γ
ically symmetric geometries density and pressure in Alice’s
where B > 0 is a bounded function that depends on the tra-
frame are finite if Alice crosses the apparent horizon in the
jectory, the Ricci scalar and the sampling function [33].
metric of Eq (12), but diverge for the metric of Eq. (13).
Consider a growing apparent horizon, r0′ (u) > 0. For sim-
In the spherically symmetric case Alice was a zero angular
plicity we consider a polar trajectory θ = 0. For a macro-
momentum observer (ZAMO) [9, 34]. In axially symmetric
scopic black hole the curvature at the apparent horizon is low
spacetimes the ZAMO condition results in a nontrivial angular
and thus Eq. (34) is applicable. Horizon radius (and mass, as
velocity ΨZ . We begin with the retarded Kerr–Vaidya metric,
in this model a = const), do not appreciably change while Al-
where the apparent horizon is located at rg = r0 . Alice’s four-
ice moves in its vicinity. Hence dM/dτ = M ′ (U )U̇ ≈ const
velocity is
and Ẋ ≈ Ṙ. Given Alice’s trajectory we can choose f ≈ 1
uµA = U̇ , Ṙ, Θ̇, Ψ̇Z ,


(27) at the horizon crossing and f → 0 within the NEC-violating
domain (as Eq. (12) can be valid only in the vicinity of the
where the ZAMO condition ξψ · uA = 0, where the Killing horizon). As the trajectory passes through X0 + rg → rg the
vector ξψ = ∂ψ , implies Ψ̇Z = −(guψ U̇ + grψ Ṙ)/gψψ . Dur- lhs of Eq. (34) behaves as
ing the fall Ṙ < 0. The velocity component U̇ > 0 is obtained
M ′ r02 Ṙ2 dτ
Z Z
from the normalization condition u2A = −1, 2
f ρA dτ ≈ − 2 2
γ γ 4πX (r0 − M )
Ṙ2 1
q
U̇ = − (r2 + a2 ) + Mτ r0 dX
Z
(∆(1 + ρ2 Θ̇2 ) + ρ2 Ṙ2 )Σ ,
∆ ∆ρ ≈ ∝ log X0 → −∞, (35)
(28) γ 8πM (r0 − M )X
where Σ
= (a2 + r2 )ρ2 + 2a2 rM sin2 θ. As X := R(τ ) −
where Mτ = M ′ U̇ and we used Ṙ ∼ const. The rhs of
r0 U (τ ) → 0 the derivative U̇ diverges as ∆−1 ,
Eq. (34) remains finite, and thus the QEI is violated. This
2r0 ṘM violation indicates that the apparent horizon cannot expand,
U̇ = − + O(X). (29) similarly to the spherically symmetric case.
X(r0 − M )
On the other hand, nothing dramatic happens to the comov-
The energy density in Alice’s frame is given then by ing density and pressure in the ingoing Kerr–Vaidya metric.
 2 ! Following the same steps we find that, e.g., the comoving en-
guψ guψ ergy density for the motion in the equatorial plane (Θ = π/2,
U̇ 2 + O ∆−1 ,


ρA = Tuu + Tψψ − 2Tuψ
gψψ gψψ Θ̇ = 0), is
(30) Tvv
resulting in ρA = + O(a2 ). (36)
4Ṙ2
(−2M ′ − (2M − r0 ) sin2 θM ′′ )r02 Ṙ2 This quantity is finite and for a = 0 reduces to Eq. (10).
ρA ≈
8πX 2 (r0 − M )2
(−2r0 M ′ − a2 sin2 θM ′′ )r0 Ṙ2
= . (31) IV. DISCUSSION
8πX 2 (r0 − M )2
We choose the spacelike direction analogously to the spheri- Extending the self-consistent approach of horizon analysis
cally symmetric case, to the axially symmetric spacetimes is difficult.Kerr–Vaidya
nA metrics are the simplest nonstationary extension of the Kerr
µ = (−Ṙ, U̇ , 0, 0). (32)
solution. All Kerr–Vaidya metrics violate classical energy
Then (after setting Θ̇ = 0), conditions. While it could have been previously considered as
a drawback, this violation is a necessary condition to describe
(−2r0 M ′ −a2 sin2 θM ′′ )r0 Ṙ2 an object with a trapped region that is accessible, even in prin-
pA = Tµν nµA nνA ≈ . (33)
8πX 2 (r0 − M )2 ciple, to a distant observer. Moreover, Kerr–Vaidya metrics
 6

are related by the Newman–Janis transformation to the pure of a pair of real [1]
Vaidya metrics that describe the geometry of physical black
holes near their apparent horizons. 1
lµ = δvµ + f (v, r)δrµ , (A1)
These simple geometries have several remarkable proper- 2
ties. The EMT of the Kerr-Vaidya metric, while violating the nµ = −δrµ (A2)
NEC for all a 6= 0 is a special case of type III form of the
EMT in the Segre–Hawking–Ellis classification. An expand- and a pair of complex-conjugate vectors
ing spherically symmetric apparent horizon leads to a fire- 1

i µ

wall and violates the quantum energy inequality that bounds mµ = √ δθµ + δψ , m̄µ = (mµ )∗ , (A3)
the amount of negative energy in spacetimes of low curva- 2r sin θ
ture. The outgoing Kerr-Vaidya metric has the same property, that satisfy the standard completeness and orthogonality rela-
showing that the firewall is not an artifact of spherical sym- tions,
metry.
The apparent horizon of the outgoing Kerr-Vaidya
√ metric lµ lµ = lµ mµ = lµ m̄µ = 0,
coincides with the event horizon r0 = M + M 2 − a2 of the nµ nµ = nµ mµ = nµ m̄µ = mµ mµ = 0, (A4)
Kerr metric, M (u) = const. For the ingoing the two surfaces
lµ nµ = −mµ m̄µ = −1.
are different. However, the difference z(θ) = rg − r0 is small
if |M ′ (v)| ≪ 1, as in this case z ∝ M ′ . However, while at The metric
the poles z(0) = z(π) = 0, a commonly used assumption
z ′ (0) = 0 does not hold. As a result, the apparent horizon is ds2 = −f (v, r)dv 2 + 2dvdr + r2 dθ2 + r2 sin2 θdψ 2 , (A5)
not a smooth surface.
The assumption a = const is incompatible with the con- where f (v, r) = 1 − 2M (v)/r, is rewritten as
tinuous eventual evaporation of a physical black hole, as for
M < a the equation ∆ = 0 has no real roots and the Hawking gµν = −lµ nν − lν nµ + mµ m̄ν + mν m̄µ . (A6)
temperature
We treat r and v is complex-valued coordinates and introduce
  a real-valued function
1 r0 − M
T = , (37) 

1 1

2π r02 + a2 1
f = 1 − M 2 (v + v )∗
+ ∗ , (A7)
r r
that is proportional to the surface gravity, goes to zero as
M → a. Moreover, the semiclassical analysis [9] shows that that coincides with f (v, r) for real values of the coordinates,
during evaporation a/M decreases faster than M [43, 44]. v = v ∗ , r = r∗ . The complex coordinate transformation
The variability of a = J/M ratio should not affect exis- x′µ = xµ − ia(δrµ + δvµ ) cos θ, (A8)
tence of the firewall for accreting PBHs, as it is exhibited
as a result ∆ → 0 effect in (vr) coordinates and holds for i. e.,
a = 0. We will drop the assumption a = const in the future
work, and will to use the-self consistent approach to match the v ′ = v − ia cos θ, θ′ = θ, (A9)
semiclassical results [43–45], as it was done in the spherically r′ = r − ia cos θ, ψ ′ = ψ, (A10)
symmetric case.
leaves M invariant and transforms the tetrad as
1
l′µ = δvµ + F(v, r, θ)δrµ , n′µ = −δrµ , (A11)
ACKNOWLEDGMENTS 2  
′µ 1 µ µ µ i µ
m = √ ia (δv + δr ) sin θ + δθ + δ ,
The work of PKD is supported by IMQRES. Useful dis- 2(r − ia cos θ) sin θ ψ
cussions with Pisin Chen, Eleni Kontou and Sebastian Murk, (A12)
and helpful comments of Luis Herrera, Joey Medved and AJ
Terno are gratefully acknowledged. where after restricting to the real-valued coordinates

F = 1 − 2M (v)r/ρ2 . (A13)

Appendix A: The Newman–Janis Substituting these explicit expressions into the transformed
transformation of the advanced Vaidya metric metric

g′µν = −l′µ n′ν − l′ν n′µ + m′µ m̄′ν + m′ν m̄′µ , (A14)
The procedure follows the Newman-Janis prescription [35,
37] that is applied to the Vaidya metric in advanced coordi- produces the Kerr-Vaidya metric in the advanced coordinates
nates as the seed metric. We use the null tetrad that consists that is given in Eq. (12).
 7

Appendix B: Energy-momentum tensor and the NEC q µ = q 2̂ eµ2̂ + q 3̂ eµ3̂ with

violation for Kerr–Vaidya metric
a2 rMv
q 2̂ = − sin 2θ, (B17)
The nonzero components of the energy-momentum tensor 8πρ5
for the Kerr–Vaidya metric in advanced coordinates are a(r2 − a2 cos2 θ)Mv
q 3̂ =− sin θ. (B18)
8πρ5
r2 (a2 + r2 ) − a4 cos2 θ sin2 θ a2 r sin2 θ
Tvv = M v − Mvv ,
4πρ6 8πρ4 A generic form [19] of a type III EMT is
(B1)  
a2 r sin θ cos θ ̺ 0 ϕ 0
Tvθ = − Mv , (B2)
4πρ4  0 −̺ ϕ 0 
 
T âb̂ =  . (B19)
r2 − a2 cos2 θ  ϕ ϕ −̺ 0 
2 2
Tvψ = −a sin θTvv − a sin θ Mv , (B3) 0 0 0 p
8πρ4
a3 r sin3 θ cos θ All four Lorentz-invariant eigenvalues are zero if and only if
Tθψ = Mv , (B4)
4πρ4 ̺ = p = 0. In this case the nonzero eigenvalues of the matrix
r2 − a2 cos2 θ T âb̂ are
Tψψ = a2 sin4 θTvv + a2 sin4 θ Mv . (B5)
4πρ4 √
λ̃1,2 = ± 2ϕ. (B20)
The nonzero components of the energy-momentum tensor for
the Kerr–Vaidya metric in retarded coordinates are
Appendix C: Apparent horizon in the outgoing
r2 (a2 + r2 ) − a4 cos2 sin2 θ a2 r sin2 θ Vaidya metric
Tuu =− M u − Muu ,
4πρ6 8πρ4
(B6) On a hypersurface v = const we introduce the surface co-
2a2 r sin θ cos θ ordinates (r̆, θ, φ) where the bulk coordinate r is expressed in
Tuθ =− Mu , (B7) terms of the coordinates r̆ and θ as
8πρ4
r2 − a2 cos2 θ r = r̆ + z(θ), (C1)
Tuψ = −a sin2 θTuu + a sin2 θ Mu , (B8)
8πρ4
2a3 r sin3 θ cos θ for some function z. Locating the apparent horizon rg (v, θ) is
Tθψ = Mu , (B9) then expressed as a problem of finding the function z(θ) such
8πρ4
that rg = r0 (v) + z(θ). While the function z also depends on
(r2 − a2 cos2 θ) v, it does not affect the derivation below and this dependence
Tψψ = a2 sin4 θTuu − a2 sin4 θ Mu .
4πρ4 is omitted.
(B10) Two spacelike vectors that are tangent to the surface r̆ =
const are
In the advanced coordinate the decomposition (15) of the
EMT is obtained with the vectors bµθ = z ′ (θ)δrµ + δθµ , bµψ = δψµ . (C2)

kµ = (1, 0, 0, −a sin2 θ), (B11) We obtain the outward- and inward-pointing future-directed
null vectors l+ ≡ l and l− ≡ N by using the orthogonality
and condition lµ± bµA = 0. Before the rescaling lv = 1 and the
normalization N · l = −1 the two null vectors are given by
r2 − a2 cos2 θ
 
qµ = 0, 0, Tvθ , −a sin2 θ M v . (B12)
8πρ4 lµ± ∝ (−1, ℓ± , −ℓ± z ′ (θ), 0). (C3)

The orthonormal tetrad with where k µ = eµ1̂ + eµ0̂ is given by The two values of ℓ± are obtained from the null condition
l± · l± = 0,
eµ0̂ = (−1, rM/ρ2 , 0, 0), (B13) 1  2
ℓ± = r + a2
eµ1̂ 2
= (1, 1 − rM/ρ , 0, 0) (B14) ∆ + z ′2
q 
eµ2̂ = (0, 0, 1/ρ, 0) (B15) ± 2a2 rM sin2 θ + ρ2 (a2 + r2 ) − a2 z ′2 sin2 θ .
1 (C4)
eµ3̂ = a sin θ, a sin θ, 0, csc θ .


(B16)
ρ
After setting lv = 1 the leading order components of the
Hence the EMT is given by Eq. (16) with ν = Tvv and future-directed outward-pointing null vector orthogonal to the
 8

two-surface r = r0 + z(θ) are respectively, the cross section tangents evolve as

lv = 1, (C5)
bµψ → bµψ , bµθ → bµθ + ∂θ lµ (xin )dλ. (C13)
(r02
− a )z2 ′
lr = , (C6)
2r0 (r02 + a2 ) √
The area differential dδA ∝ (d σ/dλ)dλ is evaluated by
z′ p
lθ = − 2 , (C7) subtracting σ(xin ) from the first-order expansion in dλ
r0 + a2 p
of σ(xfn ). The desired Eq. (25) is obtained by setting
a
lψ = 2 dδA = 0.
r0 + a2
An alternative derivation is based on extending the vector
a(a4 − 7a2 r02 − 10r04 − a2 (r02 − a2 ) cos 2θ)z field lµ from the hypersurface v = const to the bulk in such
+ , (C8)
4r0 (r02 + a2 ) a way that the new field ˚ lµ satisfies the geodesic equation
˚µ ˚ν
l;ν l = 0. In fact, this needs to be done only on the hyper-
where we assume that |z| ≪ r0 and |z ′ | ≪ r0 .
surface itself, where it is realized by setting
We now consider the change in the two-dimensional area
after one infinitesimal step δλ of the evolution xµin → xµfn ,
where ˚
lµ := lµ , (C14)

xµin = (v, r0 + z(θ), θ, 0), (C9)

and thus ˚
l0 = l0 = 1,
xµfn = xµin + lµ (xµin )dλ, (C10)

and λ is the affine parameter. ˚µ

l;m := lµ;m , (C15)
The determinant of the two-dimensional metric σAB is
given by Eq. (23). To obtain the initial area the Kerr-Vaidya for m = 1, 2, 3, and setting the covariant derivative over v as
metric is evaluated at xin and the vectors bA are given by
Eq. (C2). To calculate the final area we evaluate the four-
˚
lµ;0 := −lµ;m lm , (C16)
dimensional metric at the point xfn . In addition, since the
points xψ and xθ that are defined by Eq. (21) evolve with the
vectors
For the affinely parametrized geodesic congruence ϑ = ˚
lµ;µ ,
lµ (xψ ) = lµ (xin ), (C11) and Eq. (25) follows from

and ϑ = −l0;m lm + l;m

m
= 0. (C17)
lµ (xθ ) ≈ lµ (xin ) + ∂θ lµ (xin )δθ, (C12)

[1] S. Chandrasekhar, The Mathematical Theory of Black Holes (Springer, Heidelberg, 2015).
(Oxford University Press, Oxford, England, 1992). [11] S. W. Hawking, Nature (London) 248, 30 (1974).
[2] Astrophysical Black Holes, edited by F. Haardt, V. Gorini, U. [12] V. P. Frolov, arXiv:1411.6981.
Moschella, A. Trever, and M. Colpi (Springer, Heidelberg, [13] V. Baccetti, R. B. Mann, S. Murk, and D. R. Terno,
2016). Phys. Rev. D 99, 124014 (2019).
[3] W. Israel, in Three Hundred Years of Gravitation, edited by S. [14] N. D. Birrel and P. C. W. Davies, Quantum Fields in Curved
W. Hawking and W. Israel (Cambridge University Press, Cam- Space (Cambridge University Press, Cambridge, England,
bridge, England, 1987), p. 199. 1986).
[4] V. Cardoso and P. Pani, Nat. Astron. 1, 586 (2017); [15] B.-L. Hu and E. Verdaguer, Semiclassical and Stochastic Grav-
Living Rev. Relativity 22, 4 (2019). ity: Quantum Field Effects on Curved Spacetime (Cambridge
[5] R. Carballo-Rubio, F. Di Filippo, S. Liberati, and M. Visser, University Press, Cambridge, England 2020).
Phys. Rev. D 98, 124009 (2018). [16] R. B. Mann, Black Holes: Thermodynamics, Information,
[6] L. Barack, V. Cardoso, S. Nissanke, and T. P. Sotiriou, and Firewalls (Springer, New York, 2015); D. Harlow,
Classical Quantum Gravity 36, 143001 (2019). Rev. Mod Phys. 88, 015002 (2016).
[7] E. Curiel, Nat. Astron. 3, 27 (2019). [17] W. G. Unruh and R. M. Wald,
[8] S. W. Hawking and G. F. R. Ellis, The Large Scale Structure Rep. Prog. Phys. 80 092002 (2017); D. Marolf
of Space-Time (Cambridge University Press, Cambridge, Eng- Rep. Prog. Phys. 80, 092001 (2017).
land, 1973). [18] H. Stephani, D. Kramer. M. MacCallum, C. Hoenselaers, and
[9] V. P. Frolov and I. D. Novikov, Black Holes: Basic Concepts E. Herlt, Exact Solutions to Einstein’s Field Equations (Cam-
and New Developments (Kluwer, Dordrecht, 1998). bridge University Press, Cambridge, England, 2003).
[10] V. Faraoni, Cosmological and Black Hole Apparent Horizons, [19] P. Martı́n-Moruno and M. Visser, Classical and semi-classical
 9

energy conditions, in Wormholes, Warp Drives and Energy [33] E.-A. Kontou and K. D. Olum,
Conditions, edited by F. N. S. Lobo (Springer, New York, Phys. Rev. D 91, 104005 (2015).
2017), p. 193. [34] E. Poisson, A Relativist’s Toolkit (Cambridge University Press,
[20] E.-A. Kontou and K. Sanders, Cambridge, England, 2004).
Classical Quantum Gravity 37, 193001 (2020). [35] E. T. Newman and A. I. Janis,
[21] A. Almheiri, D. Marolf, J. Polchinski, and J. Sully, J. Math. Phys. (N.Y.) 6, 915 (1965).
J. High Energy Phys. 02 (2013) 062. [36] C. González, L. Herrera, and J. Jiménez,
[22] V. Baccetti, S. Murk, and D. R. Terno, J. Math. Phys. (N.Y.) 20, 837 (1979); L. Herrera, and J.
Phys. Rev. D 100, 064054 (2019). Jiménez, J. Math. Phys. (N.Y.) 23, 2339 (1982).
[23] D. R. Terno, Phys. Rev. D 100, 124025 (2019). [37] S. G. Ghosh and S. D. Maharaj, Eur. Phys. J. C 75, 7 (2015).
[24] D. R. Terno, Phys. Rev. D 101, 124053 (2020). [38] Th. Christoulakis, Th. Grammenos and Ch. Kolassis,
[25] M. Murenbeeld and J. R. Trollope, Phys. Lett. 149A, 354 (1990).
Phys. Rev. D 1, 3220 (1970). [39] D.-Y. Xu, Classical Quantum Gravity 16, 343 (1999).
[26] M. Carmeli and M. Kaye, Ann. Phys. 103, 97 (1977). [40] M. Shibata, Numerical Relativity (World Scientific, New Jersey,
[27] J. M. M. Senovilla and R. Torres, 2016).
Classical Quantum Gravity 32, 189501 (2015). [41] J. Thornburg, Living Rev. Relativity 10, 3 (2007).
[28] A. Paranjape and T. Padmanabhan, [42] C. J. Fewster, Quantum Energy Conditions, in Wormholes,
Phys. Rev. D 80, 044011 (2009). Warp Drives and Energy Conditions, edited by F. N. S. Lobo
[29] V. Baccetti, R. B. Mann, and D. R. Terno, (Springer, New York, 2017), p. 215.
Classical Quantum Gravity 35, 185005 (2018). [43] R. Dong, W. H. Kinney and D. Stojkovic,
[30] C. W. Misner and D. H. Sharp, Phys. Rev. 136 B571 (1964). J. Cosmol. Astropart. Phys. 10 (2016) 034.
[31] V. Faraoni, G. F. R. Ellis, J. T. Firouzjaee, A. Helou, and I. [44] A. Arbey, J. Auffinger, and J. Silk,
Musco, Phys. Rev. D 95, 024008 (2017). Mon. Not. R. Astron. Soc. 494, 1257 (2020).
[32] A. Levi and A. Ori, Phys. Rev. Lett. 117, 231101 (2016). [45] A. Levi, E. Eilon, A. Ori, and M. van de Meent,
Phys. Rev. Lett. 118, 141102 (2017).