Professional Documents
Culture Documents
we include the backreaction of the negative energy Hawking flux in the interior geometry of the
collapsing star and solve the full 4-dimensional Einstein and hydrodynamical equations numerically.
We find that Hawking radiation emitted just before the star passes through its Schwarzschild radius
slows down the collapse of the star and substantially reduces its mass thus the star bounces before
reaching the horizon. The area radius starts increasing after the bounce. Beyond this point our
program breaks down due to shell crossing. We find that the star stops collapsing at a finite radius
larger than its horizon, turns around and its core explodes. This study provides a more realistic
investigation of the backreaction of Hawking radiation on the collapsing star, that was first presented
in [1].
ab
∇b Ttotal = 0, ∇b T ab = −∇b τ ab , (6)
II. MODEL
ds2 = −e2Φ(r,t) dt2 + eλ dr2 + R2 dΩ2 (1) The form of the renormalized stress energy tensor for
the Hawking radiation flux in the Unruh vacua, at fu-
where R(r, t) is the areal radius and dΩ2 = dθ2 + ture infinity and near the surface of the star were cal-
sin(θ)2 dφ2 . This form of the metric is convenient for culated in Candelas [18]. Based on energy conservation,
describing a radiating star. The set of Einstein and en- the net inward negative energy flux in the star’s inte-
ergy conservation equations, were originally derived in rior is equal and opposite to its value at infinity, shown
Misner [12]. We will use the set of equations [12] and in [3, 18, 20, 23] who also showed that the Unruh vac-
[10], jointly known as the hydrodynamic equations, by uum best approximates the state that follows gravita-
modifying them to reflect our particular system. At the tional collapse since it such that it is nearly empty in the
onset of collapse the star gradually starts producing a far past before the star collapses, but then produces a
flux of Hawking radiation. The positive energy flux trav- flux of radiation once the collapse starts. On these ba-
els outwards to future infinity and an equivalent but neg- sis the quantum mechanical stress energy tensor of the
ative energy flux travels radially inwards in the interior Hawking radiation flux in the interior is as follows
of the star towards the center. Most of the outgoing
null radiation is produced on the last stages of collapse 1 1 0 0
[9, 11, 19, 20]. We wish to include the backreaction of the L −1 −1 0 0
τa b = 2 (8)
negative energy Hawking radiation flux on the interior of R 0 0 0 0
the star and solve for its evolution. 0 0 0 0
The stress energy tensor of the Hawking radiation flux
is given by The total energy conservation is best written as a set of
ab a b two equations which contain the radiative heat transfer
τ = qH k k (2)
from the fluid to the Hawking flux, its t-component and
where k a is an outgoing null vector k a ka = 0 and qH the r-component that are obtained by contracting T ab
is the Hawking radiation energy density related to the in 6 with ua and the unit vector na orthogonal to the
luminosity of radiation by L = 4πR2 qH . 3-hypersurface Σt at t = constant as follows
The fluid of the inhomogenenous dust star with a 4-
velocity ua , normalized such that ua ua = −1, has a stress Qa = ∇b τ ab = (e−Φ nC, e−λ/2 nCr , 0, 0) (9)
energy tensor
From Eqn. 9 and the Hawking flux stress energy tensor in
T ab = ua ub , (3) the interior Eqn. 8, it is immediately clear that Cr = −C.
The latter simply reflect the fact that the star is receiving
where the energy density is expressed in terms of a spe- negative energy transfer travelling inwards in its interior.
cific inetrnal energy density per baryon e and the number We can now write explicitly the full set of hydrody-
density of baryons n by = n(1 + e) = nh with h the namic equations, given above. First let us define the
enthalpy. We normalize ka by the condition ka ua = 1, so proper derivates Dr , Dt by the following relation in terms
that qH denotes the magnitude of Hawking energy flux of the metric
density of the inward moving radiation in the rest frame
of the fluid. Therefore the total stress energy tensor of
Dr = e−λ/2 ∂r Dt = e−Φ ∂t . (10)
the combined system of the star’s fluid (Tab )and of the
radiation (τab ) is
The evolutionary equations for baryon number density
ab
Ttotal = T ab + τ ab (4) n(r, t) from 7, specific internal energy e from 6, the en-
ergy of null-radiation with luminosity L given from the
The equations that describe the dynamics of the in- t-component of 9 and 6, the proper velocity of the co-
terior of the star with the backreaction of the Hawking moving fluid U , areal radius of comoving fluid elements
radiation flux included are: the Einstein equations, R(r, t), and the ’acceleration’ equation obtained from the
3
r-component of 9 and the Einstein equations 5 We being the motivation of our choice C(r, t) by not-
ing that for p = 0, Eq. (11b) simplifies to ∂t e = −C eΦ .
1 1 ∂ Therefore, we can achieve a spatially constant e (i.e. spa-
Dt n + 2 (R2 U ) = 4πRqH αf −1/2 , (11a)
n R ∂R tially uniform mass evaporation) with the choice
1
Dt e = −C − pDt , (11b)
n C(r, t) = e−Φ(r,t) C̄(t), (19)
2 −2Φ 2Φ
∂U Lα
Dt L = 4πR nC − e Dr Le − 2L − , where C̄(t) is a not yet determined function of time.
∂R Rf 1/2
(11c) To determine C̄(t), we consider Eq. (11c), which relates
the luminosity L to the energy-conversion rate C. If we
Dt R = U, Dr R = f, (11d) disregard all gradient terms and any other term indepen-
∂Φ m L dent of C, then Eq. (11c) becomes ∂t L = 4πR2 eΦ nC.
Dt U = f − + . (11e) This represents the radiation creation rate in a spherical
∂R R2 R
shell at radius R. Integrating over radius with volume
Two auxiliary quantities are given by radial ODEs, element eλ/2 dr, we find a total radiation creation rate of
∂m Z rs
= 4πR2 ε − qH 1 + αU f −1/2 , (11f)
∂R Ctotal = dr 4πR2 eΦ nC eλ/2 (20)
∂Φ ∂p 0
+ nCf −1/2 ,
Z rS
(ε + p) =− (11g) αR0
∂R ∂R = C̄(t) dr 4πR2 n √ . (21)
0 f
with boundary conditions
If we disregard radiation propagation effects, i.e. assume
m = 0, L = 0, at r = 0, for all t, (12) that the star is unchanged during a light-crossing time,
Φ=0 at r = rstar , for all t. (13) then we expect that LS ≈ Ctotal . Combining this with
Eq. (18), we arrive at
Finally, ε, f , α and R0 are short-cuts for
rS
p 2 Z αR0
2m LH = C̄(t) US + fS dr 4πR2 n √ , (22)
f = 1 + U2 − , (14) 0 f
R
ε = n(1 + e), (15) from which it follows that
R0
α = 0 = signR0 , (16) LH
Z rS 0 −1
|R | 2 αR
C̄(t) = √ 2 dr 4πR n √ . (23)
∂R US + fS 0 f
R0 = . (17)
∂r
the form
0.4
mS
u(ri , t + ∆t) − u(ri , t) 1 0.3
= (R[u](ri , t) + R[u](ri , t + ∆t)) . RS/10
∆t 2 0.2
(24)
Here, R represents the right-hand-sides of Eq. (11). R 0.1
involves second order finite-difference stencils (centered 0
-4
in the interior, one-sided at the boundary). When eval- 10 |dtk - dt2|
uating R at the origin, r = R = 0, the rule of L’Hospital 10
-5
∂η 3 h i
= ε − q 1 + αU f −1/2 − η , (26)
∂R R i.e. the lines Λ(r, t)=const are null. To achieve this, Λ
must satisfy
and discretize the partial derivative directly in terms of
the areal radius R. After solving for η, we recover m = ∂Λ ∂Λ
4 3 0 = k(Λ) = e−φ + e−λ/2 . (28)
3 πR η. ∂t ∂r
Figure 1 demonstrates the expected second order con-
vergence of this finite-difference code. Solving for ∂Λ/∂t results in an evolution equation for Λ,
√
∂Λ φ α f ∂Λ
= −e . (29)
B. Diagnostics ∂t R0 ∂r
This is an advection equation with positive advection
Besides monitoring the evolved variables and the aux- speed, i.e. the characteristics of Λ (the null rays) al-
iliary variables, we utilize two additional diagnostics. ways move toward larger values of r. This is an expected
First, we compute the expansion θ(k) of r =const sur- result for a comoving coordinate r. Because Λ is always
faces along the outgoing null-normal k µ = Dt + Dr = advected outward, we must supply a boundary condition
e−Φ ∂t + e−λ/2 ∂r . This quantity is computed as at r = 0. This boundary condition sets the value of Λ for
the null ray starting at (t, 0), and we choose
2 p
θ(k) = α f +U . (27)
R Λ(t, 0) = t. (30)
When θ(k) = 0, an apparent horizon has formed. Specif- We also need initial conditions at t = 0, where we set
ically, we will plot Θ(k) R/2, a quantity which is unity in
flat space. Λ(0, r) = −r. (31)
Second, we track during the evolution outgoing null
geodesics, to gain a deepened understanding of the causal These initial- and boundary conditions ensure that Λ ≤ 0
structure of the space-time. The null-geodesics are rep- represents null rays that intersect the t = 0 surface,
resented by the level-sets of an auxiliary variable Λ(t, r), whereas Λ ≥ 0 represent null rays that emenate from
5
1.2 RS/R0
5 R(ri, t) 0.8
0.6 θS RS/2
mS
4 0.4
0 nS
US
-0.6 3 0
13.5 13.8 14.1
0.4
1000 LS RS Oppenh.-Sny.
1000 L∞
0.2 1000 LH 2
2mS(t)
0
eS
1
-0.2
-0.4
0
0 5 10 15 0 5 10 15
t t
FIG. 2: Evolution of pressure-free collapse with Hawking FIG. 3: Evolution of pressure-free collapse with Hawking
radiation (R0 = 4, M0 = 0.4). The thick lines indicate the radiation (R0 = 4, M0 = 0.4). Shown are the area radii
evolution of quantities evaluated at the stellar surface: areal at sixteen equally-spaced comoving coordinate radii ri . The
radius RS , radial velocity US , total mass mS , number density thick red line indicates 2mS , which is scaled such that RS =
nS , expansion θS . The lower panel shows the internal energy 2mS would indicate the formation of a black hole. The thin
eS , luminosity at the stellar surface LS , luminosity at infinity dashed blue line indicates the areal radius of a Oppenheimer-
L∞ , and our target luminosity LH . For easy of plotting, the Snyder collapse.
luminosities are scaled by a factor of 1000. The thin dashed
lines in the upper panel represent the respective quantities
in a pure Oppenheimer-Snyder collapse of the same initial 0.4 U(ri, t)
conditions.
t
13.6 1.2 13.6 0.5 13.6 4.1
0.6 4.9
13.5 0.6 13.5 13.5 5.7
0.0 0.7 6.5
13.4 Areal radius R
140.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
13.4 Potential φ
140.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
13.4 Potential λ
140.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
12 r 12 r 0.0 12 r 0.7
3.6 0.1 0.1
10 3.0 10 0.2 10 0.9
2.4 1.7
8 8 0.3 8 2.5
1.8 0.4 3.3
t
t
6 6 6 4.1
1.2 0.5 4.9
4 4 0.6 4
0.6 5.7
2 0.0 2 0.7 2 6.5
0 0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
r r r
FIG. 5: Contour plots of metric quantities for the simulation with R0 = 4, M0 0 = 0.4. The lower panel show the entire
time-range on the y-axis, the upper panels zoom into the late-time phase with strong Hawking radiation.
is an artifact of p = 0 choice in this model, the density Φ drops quickly to values around −0.5. The potential λ
n(r, t) at the surface of the star diverges ∝ (tcrit − t)−1/2 , shows also two distinct phases: During the Oppenheimer-
and the simulation crashes. The divergence in n(r, t) Snyder phase, λ remains approximately spatially con-
at finite stellar radius and mass indicating shell-crossing stant, but slowly decreases. During the Hawking radi-
of the outer layers of the star, can not be handled by ation phase, λ quickly decreases close to the surface of
our code. A more sophisticated numerical method, in the star, presumably in accord with the diverging number
combination with a fluid-description with pressure would density n close to the stellar surface.
be needed to proceed further and will be done in fu- Figure 6 explores the parameter space of different R0
ture work. Nevertheless, we consider the behavior before and M0 . Generically, we find the behavior shown in
shell-crossing as generic. Figs. 2 and 3: The stars first collapse at nearly con-
The behavior of this collapse with Hawking radiation stant mass until they reach a size close to the horizon (in
is explored further in Fig. 3. Shown are 16 areal radii Fig. 6, this represents a diagonal line at mS = 2RS ). As
(at constant comoving coordinates), as well as 2mS (t). the stars approach their horizons, luminosity increases
The main panel shows clearly the homologous collapse and the mass drops significantly, with the star remain-
at early times t . 13.5. The inset shows an enlargement ing just outside RS = 2mS . All trajectories approach the
of the late-time behavior, showing clearly that the mass same limiting behavior, thus we find the expected generic
mS is reduced such that the outer surface of the star behavior of Hawking radiation.
remains just outside its horizon. We have shown results here for very compact stars as
Figure 4 shows radial velocity U at 16 comoving radii. those are the more likely ones to collapse into a black
When Hawking radiation becomes dominant during the hole. We conducted the analysis for stars with low mass,
last ∼ 1 of the evolution, the inward motion is signifi- see the M = 0.1 data in Fig. 6. As expected these stars
cantly slowed down in the outer layers, and even reversed behave differently, and substantially evaporate and ex-
in the inner layers of the star. The radial gradient in the plode before they approach their horizon.
acceleration (with larger outward velocities toward the
center of the star) is responsible for the diverging num-
ber density in the star. To proceed further will either V. CONCLUSIONS
require a treatment of shocks, or a modification of our
model to avoid shocks. Einstein equations tell us that the final destiny of a
Figure (5), finally presents the metric quantities for gravitationally collapsing massive star is a black hole [2].
this evolution. The areal radius R is very similar to the This system satisfies all the conditions of the Penrose
standard Oppenheimer-Snyder collapse; the small devia- Hawking singularity theorem [25]. However a collaps-
tions during the Hawking radiation phase are not easily ing star has a spacetime dependent gravitational field
visible. The lapse-potential Φ clearly shows two distinct which by the theory of quantum fields on curved space-
phases: During the quasi Oppenheimer-Snyder collapse time should give rise to a flux of particles created [11].
Φ ≈ 0. When the Hawking radiation becomes significant, Hawking discovered in the early 0 700 s that this is in-
7
star shrinking in proportion, to preserve RS /mS > 2, that universally collapsing stars bounce into an expand-
cf. Figs. 3 and 6. The velocity of the collapsing mat- ing phase and probably blow up, instead of collapsing to
ter then reverses toward an expanding phase, cf. Fig. 4 a black hole. Thus ’fireworks’ should replace ’firewalls’.
with the inner layers expanding faster than the outer
layers. Unfortunately the latter behaviour, i.e. the ’un-
folding’ of the star from inside out leads numerically to
shell-crossing which our code can not handle. Once shell Acknowledgments
crossing occurs we can not follow the evolution further
to determine whether in the expanding phase, the stellar LMH is grateful to DAMTP Cambridge University
remnant explode, or simply evaporate away, despite that for their hospitality when this work was done. LMH
from the low mass cases and from the velocity results the would like to thank M.J. Perry, L. Parker, J. Bekenstein,
explosion seems to be the case. P. Spindel, and G. Ellis for useful discussions. LMH ac-
The star never crosses its horizon, so neither unitarity knowledges support from the Bahnson trust fund. HPP
nor causality are violated, thereby solving the longstand- gratefully acknowledges support from NSERC of Canada
ing information loss paradox. This investigation shows and the Canada Research Chairs Program.