Sie sind auf Seite 1von 10

1 Hawking radiation

It was discovered by Penrose in 1971 that energy can be extracted from a rotating (Kerr)
black hole. Inside the ergoregion of such a black hole the time-like Killing vector becomes
spacelike, so that particles with negative energy can exist inside the ergoregion. For this
reason it is possible for a particle with positive energy to enter the black hole and decay
into a particle with positive energy and a particle with negative energy. The positive
energy particle can then escape the ergoregion with more energy than the original particle,
while the black hole absorbs the negative energy particle and losses mass and angular
momentum.
By regarding the Penrose process as a form of stimulated particle emission, Hawking was
led to the supposition that a quantum field theoretical treatment of the process would
reveal the occurrence of spontaneous emission. In studying this question in the context
of a black hole forming after a gravitational collapse, Hawking discovered that even
non-rotating (Schwarzschild) black holes spontaneously radiate. The discovery was the
more surprising in light of the area theorem Hawking had previously proven: Provided
that cosmic censorship holds true and that the weak energy condition is satisfied, the
area of a future event horizon cannot decrease in an asymptotically flat spacetime. This
theorem helped to motivate the identification between the entropy of a black hole and the
area of its event horizon, an identification further motivated by the fact that the energy
extracted from a Kerr black hole with the Penrose process is maximized when the area of
the event horizon is unchanged in perfect analogy with the Carnot cycle. Now, classically,
Hawking’s area theorem prohibits Schwarzschild black holes from radiating because the
radiation would increase the mass of the black hole, and the area of the event horizon
decreases monotonically with the mass – unlike the case for the Kerr black hole, where the
event horizon can remain unaltered as the mass decreases if the rotation slows down. But
the radiation, emitted even by Schwarzschild black holes, that Hawking’s computations
revealed is a quantum phenomenon to which Hawking’s area theorem does not apply
since the weak energy condition is no longer satisfied.
The main idea behind Hawking’s computation can be explained as follows: We know that
for quantum field theory in curved spacetime there is not one preferred vacuum state –
different locally inertial observers will define their vacua differently. In the spacetime
where a Schwarzschild black hole forms after gravitation collapse, the vacuum of an
observer A at infinity in the distant past will not be a vacuum to an observer B at
infinity long after the black hole forms, and this fact we can interpret as the gravitational
field of the black hole generating particles. By computing the Bogoluyabov coeffecients
relating the mode functions of the two observers, we can find the particle spectrum seen
by observer B.

1.1 Hawking’s original 1975 computation


We consider a spacetime where a spherically symmetric star collapses to form an un-
charged, non-rotating black hole. The Penrose diagram is shown on figure 1, where J−
and J+ represent past and future null infinity and H+ the event horizon that emerges at
the collapse. Everywhere outside the star, the spacetime is described by the Schwarzschild
metric:
   −1
2 2M 2 2M
ds = − 1 − dt + 1 − dr2 + r2 (dθ2 + sin2 θdφ2 ).
r r

1
Figure 1: Penrose diagram for a collapsing star

In this spacetime we have a real, massless scalar field satisfying the covariant wave equa-
tion:
0 = ∇µ ∇µ φ = g µν ∂µ ∂ν φ.
It is convenient to expand the field in mode functions and corresponding raising and
lowering operators. We can choose a basis of mode functions {fω } that are all positive
frequency modes with respect to past null infinity, and which are normalized according
to
Z
(fω , fω )J− = −i
0 d3 x(fω fω˙ 0 ∗ − f˙ω fω∗ 0 ) = δ 3 (ω − ω 0 ).
J−

The field, then, can be expressed as


d3 ω
Z
∗ †

φ= f ω a ω + f ω ω .
a
(2π)3
But we can also choose instead to opt for a basis of modes {pω } that are positive frequency
with respect to future null infinity and also normalized with respect to J+ . However, J+
is not a Cauchy surface, that is to say, there are timelike paths that never reach J+ ,
namely those that enter the event horizon J+ . Therefore the pω modes do not form a
complete basis that can describe φ in all spacetime, and so we must complete the basis
by also including modes {qω } that are positive frequency with respect to H+ . The mode
expansion then becomes
d3 ω
Z
∗ † ∗ †

φ= p ω b ω + p ω b ω + q ω c ω + q ω c ω .
(2π)3
The exact form of the event horizon mode functions qω will not be of importance to us
in what follows, but because of the spherical symmetry of the spacetime, we pull out
the angular dependency of the mode functions by further decomposing them into partial
waves, so that the field can be expressed as:
X Z dω

φ= (fωlm aω + fωlm a†ω ),
l,m

2
where the modes can be written as
eiωv Fω (r)Ylm (θ, φ)
fωlm = √
2ω r
with v being the ingoing Eddington-Finkelstein coordinate defined by
r
v = t + r + 2M ln − 1 .

2M
Similarly we can decompose the pω mode functions:

eiωu Pω (r)Ylm (θ, φ)


pωlm = √ (1)
2ω r
where u is the outgoing Eddington-Finkelstein coordinate:
r
u = t − r − 2M ln − 1 .

2M
In the following, we will omit the l and m subscripts, as what we will have to say applies
equally to each partial wave so long as the black hole is not rotating.
Now, because we are working with quantum field theory in a curved spacetime, the mode
bases that describe particles on the flat regions of spacetime J− and J+ will be different.
The bω operators need not annihilate the vacuum state with respect to the aω operators.
Since the fω modes form a complete basis, we know that we can expand pω in terms of
these modes via Bogolyubov coeffecients:
Z ∞
pω = dω 0 (αωω0 fω0 + βωω0 fω∗0 ) .
0

And we know that the particle spectrum at J+ that the J− vacuum state gives rise to
can be determined from the βωω0 coefficients:
Z
hNω i = h0|J− bω bω |0iJ− = dω 0 |βωω0 |2 .

Our approach to finding the Bogolyubov coefficients is to consider a mode function pω


(1)
and trace it backwards from J+ to J− . A part of the mode pω scatters off the gravi-
(2)
tational field outside the star without changing frequency while another part pω enters
the collapsing star before the event horizon forms and is scattered and reflected. Corre-
spondingly, we can write
(1) (2)
αωω0 = αωω0 δωω0 + αωω0 ,
(2)
βωω0 = βωω0 .

The particle production will exclusively be due to the latter part, for which the scattering
coefficient is given by
Z  2 
0 (2) (2) 2
Γω = dω αωω0 − βωω0 . (2)

On the Penrose diagram on figure 1, ingoing lightrays with v ≥ v0 are all captured by the
black hole, while lightrays with v < v0 escape the collapsing star and reach future null

3
infinity. Ultimately, the particle creation of the collapsing star is entirely dominated by
the light rays that only just barely escape, and consequently we consider a lightray γ that
originated at J− with v0 − v small and positive and that terminates at some outgoing
Eddington-Finkelstein coordinate u on J+ .
If by nµ we denote a future-directed null vector on J+ , then −nµ will point from H+ to a
null surface of constant u. So the vector connection H+ and γ is just such a vector −nµ
that is tangent to an ingoing null curve xµ (λ). We can determine the relation between u
and the affine distance between H+ and γ by solving the geodesic equation. In outgoing
Eddington-Finkelstein coordinates the metric is given by:
 
2 2M
ds = − 1 − du2 − (dudr + drdu) + r2 (dθ2 + sin2 θ2 dφ2 )
r
Hence, we can compute the following Christoffel symbols:
M
Γuuu = − , Γuru = Γuur = Γurr = 0
r2
The geodesic equation for a radial null curve therefore tells us the following:
 2
d2 u α
u dx dx
β
d2 u M du
0 = 2 + Γαβ = 2− 2 .
dλ dλ dλ dλ r dλ
du
By defining h ≡ dλ
we can rewrite this equation as a separable first order differential
equation:
Z Z
1 M
dh 2 = 2 dλ.
h r
If we choose to set λ = 0 at the event horizon where h goes to infinity, we can integrate
this equation to obtain:
 −1
M 1 du
λ = − = − ,
r2 h dλ
which is in turn another separable first-order differential equation that it is straightfor-
ward to solve. Solving for λ and as a function of r and setting r = rS yields
M
− u u
λ = −|A|e (rS )2
= −|A|e− 4M . (3)
(2)
Now, the task of tracing pω through the star and back to past null infinity can be greatly
simplified by invoking the geometric optics approximation, which applies to waves of
the form A(x)eiS(x) where the amplitude A(x) varies much slower than the phase S(x).
Disregarding the variation of the amplitude altogether, the Klein-Gordon equation tells
us that

∇µ S∇µ S = 0.

In other words, curves of constant phase follow null geodesics. The geometric optics
approximation is valid since the wave is greatly blue-shifted near the star:
−λ
pω ∼ eiωu = e−iω4M ln( |A| ) .

4
Since we consider a ray γ very close to H+ , the affine distance −λ is very close to zero
and the frequency correspondingly great.
Following a null geodesic, then, we can extend γ all the way back to past null infinity,
where the spacetime becomes flat and v is affine parameter for the null geodesic generators
of J− . Parallel transporting −nµ along γ back to J− , the vector will continually connect
(2)
γ with a curve where the phase of pω is constant. We therefore find that
v0 − v ∝ −λ,
which, inserting the value we found for λ above, yields the following relation between u
and v:
u
v0 − v = Ce− 4M ,
which can be inverted to give
 
v0 − v
u = −4M ln .
C
By substituting this expression for u into the expression for pω in equation (1), we find
that
v0 −v
e−iω4M ln( C ) Pω (r)Ylm (θ, φ)
 i4M ω
C 1 P (r)Ylm (θ, φ)
pω = √ = √ ω .
ω r v0 − v ω r
Now that we know pω as a function of v rather than u, we can find the Bogolyubov
coefficients by taking the inner product of pω and fω0 on past null infinity:
Z
(2)
αωω0 = (pω , fω0 )J− = −i d3 x(pω f˙ω∗0 − ṗω fω∗0 )
J−
i4M ω Z v0
(C) 0 4M ω
∝ √ dv(v0 − v)−i4M ω e−iω v (ω 0 + )
ωω 0
−∞ v0 − v
0
(C)i4M ω e−iω v0 ∞
Z
0 4M ω
= √ dx x−i4M ω eiω x (ω 0 + ) (4)
ωω 0 0 x
We are only carrying out the integration with respect to v because the remaining integra-
tions not only are quite cumbersome but also unnecessary to our agenda. The v integral
is of a form that can be carried out generically:
Z ∞   Z ∞  
B C −Dx 1 BD
dx A + x e = C+1 dy A + y C e−y
0 x D 0 y
1 
= C+1 AΓ(C + 1) + BDΓ(C)
D
1
= C+1 (AC + BD)Γ(C).
D
Consequently, if we set
A = ω0,
B = 4M ω,
C = −4iM ω,
D = −iω 0 ,

5
we find that, up to a factor of no importance to us arising from the remaining integrations,
the αωω0 coefficients are given by
0
(2) (C)i4M ω e−iω v0 1
αωω0 ∝ √ 0 1−4iM ω
(−8iM ωω 0 )Γ(−4iM ω). (5)
ωω 0 (−iω )
(2)
Rather than evaluating βωω0 completely analogously, we notice that
(2) (2)
βωω0 = −(pω , fω∗0 ) = −αω,−ω0 ,

since taking the complex conjugate of fω0 merely sends ω 0 to ω 0 . We can therefore find
(2)
the βωω0 coefficients by making the substitution

ω 0 → eiπ ω 0
(2)
in the expression for αωω0 . Performing this substitution on the various factors in equation
(5) merely results in an overall change of phase, except when the substitution is carried
out on the factor

(−iω 0 )4iM ω → e−4πM ω (−iω 0 )4iM ω .

We conclude that

(2) −4πM ω (2)
βωω0 = e αωω0 . (6)

Of course one could ask why we chose to send ω 0 to ω 0 via a phase factor iπ
of e instead

−iπ (2) 4πM ω (2)
of a phase factor of e , which would have resulted in the relation βωω0 = e αωω0 .
The reason for this choice can be realized by considering expression (4). The integral
is strictly speaking not converging. To regulate it we have to send ω 0 to ω 0 + i, with 
positive since x is positive and goes to infinity in the integration domain. If, therefore
(2) (2)
we wish to continuously deform αωω0 to αω,−ω0 in a well-defined manner, ω 0 has to have a
positive imaginary part throughout the deformation. In other words, ω 0 has to be rotated
counter-clockwise about the origin, which means that it picks up a phase of eiθ with θ
going from 0 to π.

Applying equation (6) to the transmission coefficient given in equation (2), we find that
Z  2  Z
(2) 2 (2) 2

0 (2) 8πM ω
dω 0 βωω0 .

Γω = dω αωω0 − βωω0 = e −1 (7)

At last, then, we can write down the spectrum of particles produced by a black hole:
Z
(2) 2
Γω
hNω i = dω 0 βωω0 = 8πM ω .
e −1
This is exactly the form of a thermal spectrum, with a greybody factor of Γω and a
temperature, the so-called Hawking temperature, of
1
TH = .
8πM

6
Had we considered fermions instead of bosons, the conclusion and the steps leading to it
would have been essentially the same. Exchanging commutators with anti-commutators
in deriving the normalization of Bogoluybov coefficients results in the minus sign in
equation (2) being replaced with a plus, and so the same thing happens in equations (7).
The particle spectrum therefore becomes
Γω
hNω i = 8πM ω +
,
e 1
that is, a thermal Fermi-Dirac distribution with the temperature the same as before.
For rotating and charged black holes the result is also only slightly modified. The
value of 1/4M that appears in the exponent in equation (3) is the surface gravity of
a Schwarzschild black hole. For the Kerr-Newman solution for a black hole with charge
Q and angular momentum per mass a, this value has to be replaced by the surface gravity
κ, which in general is given by:
p
M 2 − Q2 − a2
κ= p .
2M 2 − Q2 + 2M M 2 − Q2 − a2

Additionally, when the black hole is rotating, the effective frequency at the horizon will
not be ω but rather ω − mΩ where m is the axial quantum number of the incident
wave packet and Ω is the angular frequency of the black hole. For a charged black hole
the effective frequency will be ω − eΦ where e is the charge of the particles and Φ the
electrostatic potential. Combining it all together, tells us that
Γωlm
hNωlm i = 2π
(ω−mΩ−eΦ)
.
e κ ∓1
Thermodynamically, we can interpret the effects of the rotation and charge as arising
from a chemical potential that causes a preferential emission of particles with the same
sign of charge and angular momentum as the black hole.

1.2 The eternal black hole


Remarkably, Hawking’s computation of the thermal spectrum of a black hole has no
dependency on the details of the actual formation of the black hole. In fact, it was
later demonstrated by Unruh that the radiation spectrum of a black hole can be derived
without propagating the waves back all the way to past spatial infinity but by only
propagating them back to a Cauchy surface that, while still much earlier than the final
Cauchy surface, could well lie after the formation of the black hole.
Indeed, even the eternal black hole of the maximally extended Schwarzschild solution
can emit radiation though the details depend on the choice of vacuum. This choice is
guided by the fact that for some vacua there exist regions of spacetime where the energy
momentum-tensor blows up and cannot be renormalized in a well-defined manner. There
is a vacuum, the Hartle-Hawking vacuum, that is regular everywhere and for which the
mode functions are positive frequency modes at infinity. From the perspective of an
observer in this vacuum, the black hole will radiate, but the radiation exactly matches
incoming radiation from past null infinity, and we would not describe actual black holes
as being in thermal equilibrium. Another vacuum exists, the Boulware vacuum, from
which viewpoint there is neither incoming nor outgoing radiation at the event horizon,

7
but the vacuum is singular on both the future and past event horizon. Lastly, there is the
Unruh vacuum, which features only outgoing Hawking radiation, and so corresponds the
closest to Hawking’s prediction. It is singular on the past event horizon, but of course
the actual black holes in our universe do not have past event horizons unlike the idealized
eternal black holes. The consideration of these vacua reveals the importance of horizons
to the emission of Hawking radiation since singularities are not an issue to astronomical
objects without a horizon, such as neutron stars, which are described by a vacuum state
similar to the Boulware state.1
In the case of Hawking radiation from an eternal black hole, the radiation can no longer be
thought of as particles impinging on the spherical body prior to the formation of the black
hole and only escaping the vicinity of the event horizon at much later times. Rather, one
could think of the radiation as arising due to the formation of virtual pairs of particles at
the event horizon, with one particle falling into the black hole while the other escapes. But
in spite of the conflicting interpretations of Hawking radiation, the various derivations all
agree in the predicted spectrum. In fact already in 1977, Hawking himself – together with
Gibbons – provided a very different, clean derivation of the Hawking temperature of a
black hole that does not make reference to the state of the black hole in the distant past.
The derivation is founded upon the equivalence between quantum statistical mechanics
in D dimensions and Euclidean quantum field theory in D + 1 dimensions, an equivalence
rooted in the connection between inverse temperature and periodicity in imaginary time,
first observed by Felix Bloch in 1932 and in 1965 formulated precisely by Feynman and
Hibbs. They noted that the amplitude for a field to propagate from a configuration φ1
to φ2 in time t2 − t1 can be expressed either by acting on the initial state with the time
evolution operator or via the path integral:
Z φ2
−iH(t2 −t1 )/~
hφ2 , t2 |φ1 , t1 i = hφ2 | e |φ1 i = [dφ]eiS[φ]/~ .
φ1

And if one sets t2 − t1 = −iβ and φ1 = φ2 and sums over all configurations φ1 , one finds
the following identity:
I
 −βH 
Tr e = [dφ]eiS[φ]/~ ,

where the path integral on the right-hand side is evaluated over all fields with a periodicity
of β in imaginary time. But the left-hand side we recognize as the canonical ensemble
for φ at inverse temperature β. In the presence of gravitational fields the path integral
must also include an integration over metrics, but the dominant contribution will come
from the metric and fields at which the action is stationary.
Setting t = iτ in the Schwarzschild metric we obtain the Euclidean metric

   −1
2 2M 2 2M
ds = 1− dτ + 1 − dr2 + r2 (dθ2 + sin2 θdφ2 ).
r r
Considering the case of constant angle, we can introduce a variable η through r = 2M + η
and expand the metric to leading order about the Schwarzschild radius to obtain
η 2M 2
ds2 = dτ 2 + dη .
2M η
1
This discussion of vacua is based on Carroll, pp. 414-415.

8
If we introduce yet another variable x through
s
2M dx p
= , x = 2 2M η,
η dη

the metric takes on the form:


x2
ds2 = dτ 2 + dx2 .
16M 2
which is nothing but the metric of the two-sphere provided we interpret τ as an angular
variable. If we consider a loop of constant geodesic distance d to the point x = 0, then
the length L(d) of the loop must satisfy

L(d)
lim =1
d→0 2πd

since we require spacetime to be locally flat everywhere and not exhibit a conical sin-
gularity. But for τ periodic with periodicity P , L(d) evaluates to P d/4M for a loop
of constant x. And so we conclude that the periodicity in imaginary time, and hence
also the inverse temperature β, is given by 8πM . That is, we recover the same value
T = 1/8πM for the temperature of the black that we obtained by very different means
in the previous section.
As we see, the connection between the connection between Euclidean action and the par-
tition function of a canonical ensemble provides a very powerful tool for quickly deter-
mining the thermal properties of black holes. We can also apply the method to succinctly
determine the Unruh temperature. For the Rindler metric, we recall, can be written as

ds2 = e2aξ −dη 2 + dξ 2 + dy 2 + dz 2 .




Going to imaginary time η → iτ and introducing the variable χ = a1 eaξ , we can write the
Rindler metric as:

ds2 = a2 χ2 dτ 2 + dχ2 + dy 2 + dz 2 ,

from which we directly read off the Unruh temperature T = a/2π.

1.3 Evaporation of black holes


Since a black hole radiates, it loses energy and so its mass decreases. This causes the
Hawking temperature and thereby also the radiation to increase. We therefore have a run-
away process that will continue until finally the black hole has evaporated altogether. To
roughly estimate the order of magnitude of the life-time of a black hole, we can disregard
the greybody factor and make use of the fact that blackbodies radiate power according
to the Stefan-Boltzmann law:

P = σT 4 A,
π2
where σ = 60
and A in our case is the surface area of the event horizon:

A = 4πrS2 = 16πM 2 .

9
By energy conservation we therefore find that the decrease in the mass of the black hole
per time is equal to
dM 1
= −P = − ,
dt 15360πM 2
which we can solve to find the mass at time t:
 1
t 3
M (t) = M0 1 − ,
tf

where tf is the lifetime of the black hole, given by


 2
3 G
tf = 5120πM0 .
~c4

If we set M0 equal to 1.5 solar masses, a low estimate of the Tolman-Oppenheimer-Volkoff


limit, we get a lifetime of 1068 years.

2 References
• Hawking, 1975

• Gibbons and Hawking, 1977

• Birrell and Davies: ”Quantum fields in curved spacetime”, 1982

• Mukhanov and Winitzki: ”Introduction to quantum fields in classical backgrounds”,


2004

• Carroll: ”Spacetime and geometry”, 2003

• Jacobsen, ”Introduction to quantum fields in curved spacetime and the Hawking


effect”, 2003
http://arxiv.org/abs/gr-qc/0308048

• Jennie Traschen: ”An introduction to black hole evaporation”


http://cds.cern.ch/record/468589/files/0010055.pdf

• Pierre-Henry Lambert: ”Introduction to black hole evaporation”, 2014


http://arxiv.org/pdf/1310.8312.pdf

10

Das könnte Ihnen auch gefallen