Professional Documents
Culture Documents
T.D. LEE
Columbia University, New York, N Y 10027, USA
We give a new derivation for the properties of radiation appearing to an observer bounded in
a space-time domain that has a horizon; this includes the Rindler case (constant accelerating
frame) and the Schwarzschild solution (black hole). Because of the global nature of the quantum
state, which inevitably extends to space beyond the horizon, it is possible for the observer to gather
information concerning the physical state beyond his horizon, similar to the Einstein-Podolsky-
Rosen experiments. Depending on the quantum states, the radiation can appear to be either
black-body or something quite different.
I. Introduction
437
438 T.D. Lee / Are black holes black bodies
3C=½£~ [ (0 t2]
622+ ~ - ~ ] ]dX, (1.2)
1
T = - e gx sinh gt, (1.5)
g
in N, as shown in fig. 1. The horizon is defined by X = I TI, and (I) refers to its
inside. T h e corresponding lagrangian and hamiltonian in the accelerating frame Zac c
are
H = !2 oo p 2 + dx, (1.8)
\
II ~ I • X
Fig. 1. An observer in a constant acceleratingframe •acc "sees" only a quarter of the space (I): X >/ ITI )
in 2.
82q~ 82ep
OT 2 8X 2 O, (1.10)
a% a%
O. (1.11)
Ot 2 OX 2
In the Heisenberg picture, the usual expansion in terms of the annihilation and
creation operators (A K and A~ in Z, and a k and a~ in Zacc) assumes the
expression
q ~ ( X , T ) = J _,.oo
~ - ~d( K2 1 2 ) - 1/2
(AKeiKX mr + A ~ e-iKX+i~2T ) (1.a2)
~o~ d k
- ~ ' (2c°)-l/2(ake'kX-i'~t+a~e-ikx+i't)
dp(x, t) = L" ~oo (1.13)
440 T.D. Lee / Are black holes black bodies
The crucial point is that, viewed in Z, while the first expansion (1.12) is valid in the
entire X, T space, the second expansion (1.13) holds only in the region (I). In the
larger frame ~, the ground state [VAC) of the hamiltonian ~ is determined by
AKIVAC) = 0 (1.17)
for all K. Assuming that the quantum state of the system is ]VAC), we may evaluate
the expectation value of the occupation number operator a~ak in the accelerating
frame ~acc" The result is the well-known black-body radiation formula [4-5]
Although the usual derivation is quite straightforward (and is given in appendix A),
the emergence of a Bose-Einstein distribution often comes as a surprise.
In sect. 2, we give an alternative derivation which makes it transparent why such a
distribution should appear in the final expression. This derivation is based on a new
theorem proved for a quantum field theory in curved space. It also extends the
validity of (1.18) to a general class of problems with horizons, such as the
Schwarzschild case. The central idea stems from the simple observation that the field
equation in the Minkowski space is hyperbolic; its characteristics are closely
connected with the horizon. On the other hand, in the euclidean space the field
equation is elliptic; hence there is no horizon. A much better perspective on the
problem can therefore be obtained. Detailed rules of Feynman diagrams for the
euclidean quantum field theory in polar coordinates are given in sect. 3. We show
that, unlike the familiar cartesian case, there are additional diagrams* which may be
interpreted as due to an external field, one that is coherent. The result is a
full-fledged quantum field theory with a unitary S-matrix, but without any statistical
averages. The method developed is readily applicable to other problems, including
the black hole. These will be discussed in sects. 4 and 5.
In these problems, because the dynamical variables outside the horizon are
assumed to be undetected, there is always a loss of information whenever one
performs a final measurement. However, such entropy increase is common to any
* While in the literature there exist many papers on this subject [5,11,12], the necessity for such
additional diagrams seems not to have been recognized.
T.D. Lee / Are black holes black bodies 441
inclusive measurement in physics*. This does not mean that the state vector itself
(before the final measurement) has to be incoherent. It is in this sense we conclude
that radiations resulting from horizons are inherently coherent, different from the
black-body radiation in statistical mechanics.
Consider again the above example of an observer restricted to the accelerating
frame ~ . Let O be any observable in ~acc and H the corresponding hamiltonian
in the same frame. By assumption, both O and H are functions only of dynamical
variables, denoted by q, available in ~acc" Let l) be the state vector of the entire
system in 2. There are two important questions related to measurements that we
shall examine in this paper:
(i) What is the condition on the state vector that will enable us to write the
expectation value of O as a statistical ensemble average
tr e-2~H/g 0
I O I) tr e - 2,~n/~ 9• (1.19)
As we shall see, if l) = IVAC) defined by (1.17), then the above formula holds for
any observable O in 2ace. While IVAC) is a pure state, its coherence is between the
variables q in 2:acc (within the horizon) and other variables, say q', outside 2~acc
(beyond the horizon). Since the variables q' are not measured and O = O ( q ) , this
coherence does not show up in {VAC IO ( q ) I V A C ) .
(ii) For what kinds of state vectors does (1.19) fail to hold? It will be shown that
for a large class of excited states, there exist expectation values {lO(q)l ) that cannot
be written in the form of an ensemble average (1.19). For example, assume the state
vector I ) consists of IVAC), plus an additional coherent excitation outside 2~acc, say
a wave packet which lies completely outside ,~c~ and is described by the appropriate
q'. Although this wave packet is beyond the reach of an observer in the accelerating
frame, through measurements such as { l O ( q ) l ) the same observer can in fact detect
its existence. Furthermore, he can also measure the long-range correlations between
q and q' that are contained in IVAC). [See (2.38)-(2.39) below.] In this case, (1.19)
no longer holds. Perhaps we can understand this situation better if we recall the
resolution of the well-known question raised by Einstein, Podolsky and Rosen
[13-16]. A pure quantum state is defined globally; its coherence may extend over
field variables located at well-separated points on a space-like surface, e.g., like the
above q and q'. To detect such correlations the simplest method is to employ two
detectors, A and B, with A measuring q and B measuring q'. Assume that B is
sensitive to the field amplitude which, because of its quantum nature, can create or
annihilate a quantum. So far as A is concerned, the action of B is equivalent to
altering the state by the additional amplitude of a particle, or antiparticle, located
near B and therefore outside the light cone of A. Conversely, the presence of such a
coherent additional particle amplitude (described by q') in the state vector I) can
replace the necessity of having an actual detector B. In this case a single detector A
is sufficient to measure the long-range correlation between q and q'. Hence, in our
problem, the single observer bounded within ~acc may also, through ( l O ( q ) l ) ,
detect such a correlation, and with that the existence of a wave packet outside Zacc-
This is why he can gather information concerning the state beyond the horizon.
These conclusions are quite general; they can be extended to the Schwarzschild
problem, and therefore to an eternal black hole. For a realistic black hole, its
nonstatic complexity makes a complete analysis difficult. In appendix C, we examine
in some detail a model calculation. We find that, at least in principle, it is possible
for an observer within the Schwarzschild frame to infer physics inside the black hole.
2. A theorem
r=o. (2.1)
Let ( Q [ ) be the coordinate representation of the state vector l) where the set
Q= (QK) (2.2)
c o ( X ) _- J_
f~ -di K
g e iKX (2.3)
In this section, we shall adopt the Schroedinger picture and regard the operator ~ as
T-independent. The X > 0 portion of the surface T = 0 coincides with the entire
t = 0 surface in ~acc" In that region, the same operator q~ may also be expanded as
dk 1
- ~ 2---~-qke with X = - e gx > 0. (2.4)
g
T.D. Lee / Are black holes black bodies 443
fo~ d k , 1
q , ( X ) =jo~--~qke'*X' when X = - - eg~'< 0. (2.5)
g
The ranges of x and x' are both from - o ¢ to oo; together they cover the entire
X-region. Similar to (2.2), we define the sets q and q' to be
For example, in 2; the vacuum state IVAC), defined by (1.17), can be expressed
either as a function of Qx; i.e., written as (QIVAC) in the Q-representation, or in
terms of qk and q~ as
(q, q'IVAC). (2.9)
The set q denotes the field-coordinates in -~acc. If one wishes, one may think of q' as
those in another accelerating frame ~c~ which, when viewed from ~, has an
acceleration of the same magnitude as g but in the opposite direction. (We note that
in the Heisenberg picture, (2.4) can be extended only to the region within (I) in fig.
1, and the other expansion (2.5) to the region within (II), which is outside the
horizon of "~acc but inside the horizon of ~'cc.)
For an observer confined to the accelerating frame 2~ac~, he may "pretend" that q
forms a complete set. In that case, he would use the accelerating system hamiltonian
H, given by (1.8) for his Schroedinger equation
1 0
i Ot I t ) = H I t ) '
where the quantum state It) is associated with the space-like surface
t = constant
contained entirely in ~a~. Viewed from Z, the physical meaning of such a state
vector ]t) is quite unclear. Operationally one also notes that something is defective.
The Green function e - ' H can bring the state vector only up to the future horizon
444 T.D. Lee / A re black holes black bodies
X = T > 0 when t varies towards oo, and up to the past horizon X = - T > 0 when t
moves towards - oo. By introducing interactions, it is not difficult to see that such a
theory, which is formulated solely in the accelerating frame, would at least be
considered incomplete.
On the other hand, a totally different vista is opened if instead of e - " n we
consider the corresponding euclidean Green function e - ' H with T real. Under the
analytical continuation
T = - iY, t = - i~ ; (2.10)
When 0 varies from 0 to ~r, the Green function e "'~ takes a state vector from X > 0
in region (I) to X < 0 in region (II). (See figs. 1 and 2.) There is n o horizon in the
euclidean space. Consequently, as we shall see, in contrast to the Minkowski
operator e -ill4, a skillful manipulation of the euclidean operator e - ' H can give the
complete solution of the quantum field theory. In particular, the state vector IVAC)
is given by the matrix elements of e - ~ n / g :
Theorem.
x ~ 0 -x
Fig. 2. The (X, T) plane defines the Minkowski space in X, with OH along one of the horizons in Xacc.
The corresponding euclidean space is the (X, Y) plane. When 0 varies from 0 to 7r, the point p moves
from (I) in fig. 1 to (II).
T.D. Lee / Are black holes black bodies 445
where, as in (2.9), IVAC} is the ground state of the total hamiltonian % in ~. The
numerator on the right-hand side of (2.12) refers to the matrix elements of e -'H with
r = ~r/g and H the hamiltonian operator in 2;ace,
=[e(x)2+ (d°,21Jdx (2.13)
given before by (1.8). Here, because we are in the Schroedinger picture, both ~(x)
and its conjugate momentum P(x) are independent of r.
Proof. Express the matrix elements {Q'le-r3ClQ) and {q'le-°n/glq ) in terms
of Feynman's path integrals. In both cases, the integrand is exp(action). In the
former,
with the boundary conditions that at Y = 0, the initial field configuration is given by
Q = { QK } through (2.3), and at the final Y, q~ is given by the same expression but
with Q replaced by Q ' = {Q~}; in the latter,
with the boundary condition that at 0 = 0 the initial field configuration ~ is given by
q = { q, } through (2.4) and at the final O, q is replaced by q ' = { qk }. Because the
action is a positive quadratic functional of ~, it is equal to its minimum (the classical
action) plus a quadratic function depending only on fluctuations. The result of
integration over these fluctuations is independent of the boundary values Q, Q' or q,
q'. Thus, the matrix elements of the quantum operators e- r% and e -°n/g are related
to their corresponding classical actions t2cl and Acl:
(i.e., Q ~ = 0 for all K). Since Q is related to q and q' by (2.8), the boundary
condition of ~cl described by Q at Y = 0 is identical to that described by q at 0 = 0
and q' at 0 = ~r. Thus the classical solutions of these two problems are the same, and
so are their classical actions; i.e., ffc~= Acv Consequently,
As Y ~ ,
(q'le-~n/glq) = const(VACI Q ) .
Since Q = Q(q, q'), we derive (2.12) and complete the proof. The denominator in
(2.12) insures {VACIVAC ) = 1.
The ground-state wave function can always be set to be real. In that case,
{q, q ' I V A C ) = (VACIq, q'), and therefore (2.12) can also be written as
Remarks.
(i) From this point, it is clear that the theorem can be applied without change to
any theory whose H is a quadratic function of ~; this includes massive or massless
fields in a Rindler or Schwarzschild space. Extensions to interacting fields is
straightforward; the details will be given in a separate paper.
(ii) At first sight, (2.12) may appear unusual since both the field-coordinates q
and q' occur in the ket-vector on its left side, while on the right q' is in the
bra-vector and only q in the ket-vector. A moment of reflection may reassure one
that this is not so surprising. In the Minkowski formulation, there is a clear
distinction between a space-like surface and a time-like surface. No such difference
exists in the euclidean formulation. In the Minkowski space, the field equations are
hyperbolic; hence, e.g., in the classical theory because of the characteristics of these
differential equations, special care is required whenever there is a closed surface
boundary. On the other hand, in the euclidean space the field equations are elliptic;
the appropriate boundary conditions in such classical theories are most conveniently
expressed in terms of closed surfaces. The existence of the horizon is closely
connected with the characteristics of the hyperbolic equation; therefore it is absent
in the euclidean space. The closed surface boundary condition for a classical
euclidean field, when transposed to the quantum theory, makes it possible to
T.D. Lee / Are black holes black bodies 447
associate the initial state with part of the closed surface, and the final state with the
remaining part. In the classical euclidean field theory (e.g., the Laplace equation in
electrostatics) sometimes the closed surface boundary may consist of an open surface
plus boundaries at infinity. In this case in its quantum version we can associate part
of the open surface (which is also a part of the entire closed surface boundary) with
the initial condition and the other part with the final state, and that is what occurs in
the matrix element ( q ' [ e x p ( - trH/g)[q) in (2.12).
(iii) Any observable in the accelerating flame Zacc can be represented by an
operator O(q) which depends only on qk and -iO/3qk. In the q-representation,
O(q) takes on the matrix form (qrlO(q)lqin), where qf and qin refer to the final
and initial values of q. Hence
X (qin, q ' [ V A C ) .
Substituting (2.12) into the above expression and noting that O(q) is independent of
q', we have, after carrying out the sum over q',
tre 2~H/gO(q)
(VACI O(q)[VAC ) = tre_2~n/g (2.17)
The right-hand side is precisely the canonical ensemble average in statistical me-
chanics, with a hamiltonian given by H in ~acc and a temperature =
(27r)-Xg/(Boltzmann constant). Thus, when the state vector is [VAC), the ground
state of the total hamiltonian ~ in X, observations in Xacc give the same results as if
the system were in thermal equilibrium. However, as we shall see in remarks (v) and
(vi) below, the situation can be quite different, when the state vector is not IVAC).
Setting O(q) = a~ak, defined by (1.13), we find
tr e- 2~'H/gatk at:
(VACIatkaktVAC) = tre_2,~H/g (2.18)
~dk.,
H=f_.~-~(akak+½)', (2.19)
we derive the Bose-Einstein distribution (1.18). Note that on account of (1.15), the
commutator of a k and a~, is a &function, which gives rise to the factor "volume" in
(1.18). In order to remove this factor, it is customary to convert the ~-function in the
commutator into a Kronecker ~-symbol by introducing a large linear box of size V
448 T.D. Lee / A re black holes black bodies
and defining
Otk = V - 1 / 2 a k , (2.20)
with a k satisfying
Because the hermitian conjugate of the operator qk satisfies the hermiticity condition
q~ = q - k , (2.26)
the q-representation of the state vector In) under a complex conjugation (which is
like a time reversal in the present case) becomes
n' k = nk . (2.28)
T.D. Lee / Are black holes black bodies 449
Combining (2.12'), (2.27) and (2.28), the result of the theorem can also be stated as
Returning to the expansions (2.4) and (2.5), written for the surface T = 0 in the
Minkowski space (which is identical to Y = 0 in the euclidean space, since T = - i Y ) ,
in the qk and q~ representation of the Hilbert space in Z, qk and Pk = - i O/Oqk
can be expressed in terms of a k and ark:
Likewise, we introduce
0
-i =-i~o(26o) 1/2(a'_k-- a'kt) . (2.31)
Oq~
Hence a;, a~,* satisfy the same commutation relation (1.15), as a k and at; further-
more a~, commutes with a k and a t. As in (2.20), we define
acting system, although ]VAC) is a coherent state the correlation is entirely between
the q u a n t u m in 2 ~ and that in X'cc. Since an observer restricted to Za~c cannot
measure the quantum in 2~cc, he would see an incoherent black-body distribution of
q u a n t a in ,~ .... in accordance with (2.18). However, the situation can be radically
different if the system is in a different state, as will be illustrated by the examples
below.
(v) Consider now the hypothetical problem that a " m i r r o r " is placed [18] at X = 0
in the Minkowski space Z ( X , T). Depending on the nature of the mirror, we may
impose the Dirichlet condition ~ = 0 at X = 0 or the N e u m a n n condition Oep/OX = 0
at X = 0 . If we stay in the Schroedinger picture, then #,=q~(X); the former
condition implies q ~ ( X ) = - 4 ~ ( - X ) while the latter q ~ ( X ) = q ~ ( - X ) . By using
(2.4)-(2.5) and (2.30)-(2.31), we see that the corresponding constraints on their
Fourier components are a k = -a'k, or a k = a'k. Thus, the vacuum state becomes, as
shown in appendix B,
oodw
F=- fo ~ e-~°~/gat,,,at_,,,, (Dirichlet),
or
F = fo ~ ~--~
dw e -~r'°/gat,~a t_~, , (Neumann). (2.36)
(VACIa,oa_~,IVAC)
where the upper sign is for the Dirichlet mirror and the lower sign for the N e u m a n
mirror. Note that l i(°'-'~)/g = exp[i ((w' - ~o)/g )In l] which in the limit l ---, 0, but at
w 4: w', carries an infinite phase; this together with F ( i ( w - w')/g) leads to the
8-function when l = 0. Therefore, when the mirror is at X = - l < 0, although an
observer in ~a¢c can never reach the mirror, yet by measuring correlation functions
within the accelerating frame, such as (VACIa,~a_~,,IVAC), the same observer can
T.D. Lee / Are black holes black bodies 451
determine the existence, the location and the nature of the mirror!
(vi) As a further illustration, let us consider the case without a mirror. The ground
state of ~ is 2: is then IVAC), given by (2.33)-(2.35). Assume that the state of
interest is*
I>=[Co+f
dk Cl( k )a,kt]l VAC ) (2.38)
If there is a genuine thermal bath of black-body radiation, then the same expectation
value would necessarily be zero. A nonzero value of (2.39) can tell the observer in
2;acc that the state is a coherent mixture of I V A C ) plus an additional amplitude of a
q u a n t u m b e y o n d the horizon; in addition, from (2.39) he can infer the BCS nature
of I V A C ) , which pair-correlates the a~ quanta in 2;ac¢ with the a~t q u a n t a in Z~¢~.
F r o m these examples, we see that the existence of a horizon is simply due to an
artificial restriction of the space-time domain influenced by a bad choice of
c o o r d i n a t e system. The q u a n t u m mechanical state vector l) is of a global character;
its p r e p a r a t i o n is not in the control of an observer restricted to staying within the
horizon. Yet, by careful experimentation the same observer can extract (from
the state vector) information b e y o n d his horizon. As mentioned in the Introduction,
the present situation is analogous to the resolution to the Einstein-Podolsky-Rosen
paradox. However, the experiments that can verify the global nature of a pure
q u a n t u m state are usually in atomic or subatomic physics, such as the decay of
p o s i t r o n i u m [16] or the K-meson [20]. Here, it happens on a macroscopic scale, and
that accentuates the unusual circumstances.
* This is representative of the physical situation when the state is known to be of low excitation, as
observed in ,~. In such a case, [) differs from IVAC) only by an amplitude consisting of a limited
number of excited quanta. Without such a mild and (in many cases) quite reasonable assumption, by
measuring only the variables q in ~ac¢, one cannot draw conclusions about the variables q', which
are outside Za¢c and commute with q. Note that, because of (2.33)-(2.34), we have a~IVAC)=
e '~'/~at_klVAC) and a'k*lVAC)=e~/ga k[VAC). Hence for any operator v = v ( q ' ) outside
X~¢c, there exists an operator u(q) inside Zac¢, so that when operating on IVAC) their effects
become identical: v(q')IVAC) = u(q)IVAC). This gives the underlying reason why an observer in
Zac¢, through measurements of observables O(q) within his horizon, can distinguish between the state
vector I) = v(q')IVAC) and IVAC), provided that v(q') is not unitary.
452 T.D. Lee / Are black holes black bodies
-OOl_
in which V(q0 can be an arbitrary function of q~, and the operators ~ ( X ) and ~ ( X )
depend only on X, not on Y. As before, they satisfy
where
* Readers who are not interested in the technical details of Feynman diagrams may wish to skip this
section.
T.D. Lee / Are black holes black bodies 453
where IVAC) is the ground state of %. Throughout our discussion, we adhere to the
well-known fact that the full content of the field theory is completely contained in
W ( j ) . In what follows, we show that the same W ( j ) can also be computed in polar
coordinates.
In terms of polar coordinates X = 0 cos 0 and Y = p sin O, the hamiltonians ~ and
~ j are replaced by
o~
Hi(O) = H +
f --00
e2gXj(x,O)q~(x)dx, (3.9)
where, as before,
1
p = - e gx. (3.10)
g
In the (polar coordinate) Schroedinger picture, q~(x) and its conjugate momentum
P ( x ) are operator functions of x, independent of O. They satisfy
where, as in (2.11),
where
O=(N+l)eg>~O,
On= neg (3.16)
tr uj(2~)
w(j) - tr u(Z~r) " (3.17)
In what follows, we shall examine the precise relationship between W(j) and w(j).
As we shall see, although different, they are closely related to each other. For
example, from w(j) we can derive W(j). Therefore, w(j) also contains all the
information of the original quantum theory in cartesian coordinates.
That they are different can be seen by considering the special case V = 0. In the
cartesian coordinates, from (3.1)-(3.7) we derive the familiar expressions:
W(j) = exp
[½fj(1)(l[@12>j(2) 1-I2 dXpdYp ], (3.18)
p=l
o~ dK x o~ dK v -1
(I[@[2>=@(X'Y)=f-~ 2~r f - ~ 2~r (K:+K:) expi(KxX+KyY ),
(3.19)
X = X 2 - X 1, Y = Y2 - Y1; (3.20)
02 02 )
@=-3(X)a(Y). (3.21)
where pp and Op refer to the polar coordinates of p. As will be shown, these two
T.D. L e e / A r e b l a c k holes black bodies 455
propagators @ and D are not the same. To evaluate (11DI2), we expand the
right-hand side of (3.17) in powers of j and extract the term quadratic in j.
Equating that with the corresponding term in (3.22), we derive when 02 > 01
where, as in (2.23), n = (nk}. By using (2.4) and (2.30) we find that (3.26) equals
1
x e- 2~r(nk+ ~)~o/g etkx, (3.27)
where
r ~ dk
(1 ID[2) = D( x, O) = j_ o~2-~ fk( O) eikx' (3.29)
in accordance with (3.24). From (3.30) we see that fk(~r + a ) = f k ( c r - ct), hence
fk (0) = fk (2 qr) where
The function fk(O) can be extended to arbitrary 0 through the periodic condition
1
at 0 = 0 +
f/,(O) = 1 fg (3.34)
at 0 = 0 - .
8 82 )
Ox------
2 + 0~.----
5 D = -~(x)~(¢). (3.35)
1 1
Xp = - egXpcos g~-p Yp= g egXpsin gTp, (3.36)
0202 ( o2 _a_2
OX----~ ' 4 - ' ~ = e -2gx2 Ox 2 + 052 ,
$ ( X ) 8 ( Y ) = e -2gx2 8 ( x ) $ ( ~ - ) , (3.37)
~, 1 --e + 2~r'~
with (3.30) and carrying out the k-integration in (3.29), we can write D(x, O) as
where the upper sign is for x > O, the lower sign for x < O, and
(3.40)
D°(x) = f o -2-(C~ O
o )S2 ~X.
= { -- (4'/r)-i ifx>O
/o -2~r
- 2to sin~x (4~r) -1
ifx<O.
(3.41)
1 ln(p2/Pl) if P2 > Pl
Do(X ) = const + 1 4~r (3.42)
G In(p2/PI) if P2 < Pl"
and therefore
1 1 1 e m~cosm8 ifx>O
1 2-~r In P2 + 2-~ m
m = 1
1 1
(1 IDI2) = D(x, O) = const - 2~r l n r + ~ - -
In(pIp2)" (3.43)
On the other hand, from (3.19) we see that the euclidean propagator in cartesian
coordinates is
1
<11@12) = ®(X, Y) = const - ~ In r. (3.44)
Hence
C(p) =0 (3.47)
~Xp
= = = • -0 4- 4-
2 I 2
Fig. 3. The decomposition of the cartesian propagator (116~12) into the polar propagator (11D [2) plus
two other diagrams, which can be interpreted as due to an "external" field coming from the horizon. [See
eq. (3.45).]
T.D. Lee / A re black holes black bodies 459
may be viewed as the decomposition of the dashed line into the sum of a solid line
plus two similarly disconnected graphs.
The inclusion of an arbitrary V(q,) is completely straightforward. By following the
standard procedures in deriving Feynman diagrams, we can expand the logarithm of
either W(j) of (3.7) or w(j) of (3.17) as a double power series in V and j. In terms
of diagrams, these two series, one for l n W ( j ) and the other for l n w ( j ) , are
otherwise identical, except that in the former we use the cartesian propagator and in
the latter the polar propagator. Substituting the decomposition (3.48) into the
expansion of l n w ( j ) , we see that the diagrams in polar coordinates can be re-
grouped as
l n w ( j ) = l n W ( j ) + SI(j) + S 2 ( j ) + . . - (3.49)
where SI(j) is linear in the external field C(p), Sz(j) is quadratic, etc.
Thus, by starting from the Minkowski space, adopting the Schroedinger picture in
the accelerating frame 2;acc of coordinates x and t and taking its hamiltonian
operator H into the euclidean space through the analytic continuation t = -i~', we
arrive at w(j) by using polar coordinates. Next, through (3.49), we can extract the
part of l n w ( j ) that is independent of the "external" field C ( p ) ; the result is
ln W ( j ) . Reverting back to the Minkowski space, we succeed in deriving the
complete S-matrix of ~ . An alternative way is to compute In W(j) directly by using
the polar propagator. This can be done as follows:
Use (3.45) and expand In W(j) around l n w ( j ) in powers of C ( p ) :
1 ~_~ 3 1 3
4-
2 x' '× X
4
4-
2 4
where st(j) is linear in C(p), s2(j) quadratic, etc. The amplitude of any scattering
process involving n quanta in 2: is determined by the part in In W(j) that is
proportional to the nth power of j. By using (3.50), we can also calculate the same
amplitude as a sum of terms, given by its right-hand side: the first refers to diagrams
without C(p), the second to those linear in C(p), etc. All these diagrams now
employ the polar propagator ~1 ]DI2 ) expressed in terms of the variables x and ~- in
Zac c, as illustrated in fig. 4.
Remarks.
(i) The "external" field C(p) is used only in the calculation of the euclidean
scattering amplitudes via the recipe (3.50). To obtain the physical S-matrix, we must
perform the analytic continuation back to the real time T on W(j). The resulting
S-matrix connecting the in- and out-states in the frame Z(X, T) is, of course,
unitary.
(ii) The presence of the "external" field C(p) in (3.45) and (3.50) is somewhat
reminiscent of the situation when there is a black-body radiation. Let B(p)
represent the amplitude of such a radiation at point p. In that case, any scattering
matrix element can also be written as a power series in B(p). There are however
several major differences; among them we may mention:
In the black-body radiation, after squaring the amplitude there is an ensemble
average, which makes any physical probability a function of IB( p)]2. Here, there is
no statistical average; the probability depends on odd as well as even powers of
C ( p ) . Furthermore, according to (3.46), C ( p ) is linear in the space coordinate xp in
Zacc; therefore it corresponds to a "constant gradient field", not a black-body
distribution. The so-called black-body radiation information is contained in the
entire polar propagator; e.g., when points 1 and 2 coincide {1]DI2 ) gives (3.32),
which is the familiar Bose-Einstein distribution formula for the average of n k + ~.
Therefore, we conclude that the apparent radiation in the accelerating system 2:acc
should not be viewed as a genuine black-body radiation.
(iii) Consider now the question of measurements in the accelerating frame ~acc.
Returning to the Minkowski space, we assume that the complete system in Z is
described by a coherent initial wave function, say at time T = - ~ . Through the
Schroedinger equation, we determine the final wave function ~ q , q ' l S ) on any
space-like surface S = S(I) + S(II) in 2:, where S(I ) is the part of the surface and q part
of the field coordinates inside 2: .... while S¢m and q' are outside. To an observer
restricted to 2: .... the outcome of his measurement on So) can be predicted through
the density matrix, obtained by squaring the final wave function and summing over
q'. This gives rise to an entropy increase that is typical of any inclusive measure-
ment. However, the system immediately before the final measurement may well be
described as a pure state; its formation is determined by the initial condition, which
in turn depends on information outside 2: .... as illustrated by the examples given at
the end of the previous section. The inability of the observer to measure q' does not
mean the state vector itself has to be incoherent.
T.D. Lee / Are black holes black bodies 461
4. Generalizations
We shall now extend the above method to any quantum field theory in which the
metric g~, of the space is known to have a horizon. For example, the square of the
invariant length-element can be
as in the constant accelerating frame Xacc of the previous sections (but now in a
three space- and one time-dimensional world), or it may be the Schwarzschild
solution
where G is Newton's constant, M the mass, a the polar angle and fl the azimuthal
angle. (In this paper, effects of quantum gravity will not be included.) In either case
the horizon implied by the metric is known to be connected with a bad choice of
coordinates, unrelated to any genuine physical singularity. Label such a coordinate
frame with the horizon as X n, and the region within the horizon as (I). Starting from
(I), there are infinitely many particle-trajectories which can cross the horizon within
a finite proper time of the particle. By following these trajectories, we can easily
extend (I) and arrive at a larger frame. When a particle crosses the horizon, if it is
leaving from (I), let F be the region that it is going into; if it is coming into (I), then
call P the region that it comes from. By construction, F lies in the future light cones
of (I), and P in its past light cones.
In order to set up a quantum field theory in a space with a Minkowski signature,
we must associate each state vector 1) with a space-like surface S. The important
question is whether such an S can be contained entirely within X H. To put it another
way, write
"horizon". Label F and P the two regions outside the horizon, with F lying in the
future light cones of (I), and P in the past. As before, the region outside (I), F and P
is called (II). In a classical field theory, by restricting our interest only to the region
within (I) we can simply ignore the other regions F, P and (II). In that case, II is
irrelevant since it is space-like with respect to (I). If the retarded Green function is
used for the integration of the field equation, then the effect of P can always be cast
into the appropriate initial conditions within (I); likewise for F through time
reversal. Quantum mechanically, the situation is quite different. The specification of
a state vector requires a complete set of mutually commuting operators (representing
observables); this is why we need S(ii) as well as S(i), since the field variables on S(m
commute with those on S(I).
T o determine whether (II) exists or not, we may consider the trajectories of signals
radiated from any point in P. If they do not all reach (I), then (II) must exist.
Likewise, think of signals received in F. If these do not all emanate from (I), again
(II) must exist.
Then, there is still the admittedly slippery course of common sense. For example,
from (4.1) one can easily see that the space is flat, although the coordinates are
curved. The restriction to X > 0 (in our previous notation) seems artificial. There-
fore, it may appear self-evident that the region X < 0 should also exist, and that is
(II). In the case of a black hole, consider the situation just before its creation.
Concentrate first on the finite space-like region from r = 0 to R > 2GM. By
excluding r > R, it is not difficult to set up a new time coordinate T in which the
black hole can be formed within a finite T. At the instant T = 0 - , assume that
inside r = 2GM there is only a mass M - e where e = 0 + . At T = 0 + , the entire
mass M is inside, thus forming a black hole. Let S(I) be the space-like surface defined
by r > 2GM at T = 0. At the instant before, we know that the region r < 2GM,
denoted by S0I) - S - S(I), existed; it was nonzero and sizable. It seems unreasonable
that S(ix) should disappear altogether at T = 0 + .
In what follows, we assume that (II) exists as a physical reality. The problem is
how to construct a complete quantum field theory by knowing only the hamiltonian
operator H in (I). Because of the horizon, this is difficult if we stay in the space with
a Minkowski signature. However, by changing into the euclidean signature we can
bypass this obstacle. The method developed in the previous sections can be readily
generalized, and will be summarized below:
(i) Start from (I) and construct the hamiltonian H in 2JH in a space with a
Minkowski signature.
(ii) G o to the euclidean signature by setting time t = -i~- with ~- real. The field
equation becomes elliptic and there is no horizon.
(iii) The theorem proved in sect. 2 holds in general, and (2.12) gives the physical
vacuum state.
(iv) For interacting fields, use (3.17) to define the generating function w(j)
(constructed entirely from information available in (I). Following the discussion
T.D. Lee / Are black holes black bodies 463
given in sect. 3, we can derive W(j) which contains the full information of the
q u a n t u m field theory in the entire space (I), (II), F and P.
As an illustration, we may consider, say, a free massive, or massless, scalar field in
the four-dimensional constant accelerating frame 2~acc described by (4.1). The
hamiltonian H in 2~ac~ can be written in terms of its normal modes as
where q, and p, are the canonical variables, o~, their frequency and a , , at. the
corresponding annihilation and creation operators. Using (2.12), one sees that the
vacuum wave function in the larger frame ((I), (II), F and P) is given by
In contrast, a conventional proof of the same formula along the line given in
appendix A can become somewhat complicated in this case.
5. Black hole
x ° = t, x 1 = r, x 2 = a, x 3 = ft. (5.1)
g0o=-e2X=- 1 - , g11=e-2X= 1
r
As usual, the repeated greek indices are summed over from 0 to 3, and the roman
indices from 1 to 3.
Adopting the Schroedinger picture, q~ and ~r depend only on the 3-vector r; their
commutation relations are
Following the step (ii) outlined in sect. 4, we set t = - i¢. Introduce (in analogy to
the Kruskal-Szekeres coordinates in the space with a Minkowski signature)
\ 1/2
r 1) e r/4GM , (5.10)
P =--- 2 G M
0 -- ¢ / 4 G M . (5.11)
O,p,a, fl,
or
(5.12)
Their ranges are
0<0<2~r, 0~<p~<oc,
correspondingly,
u = p cosh(t/4GM)
v = p sinh(t/aGM) in (I), (5.15)
u = - Ocosh(t/4GM )
v = - P sinh(t/4GM) in ( n ) (5.16)
and O remains given by (5.10). The future and past regions F and P are related to t
and r by
u = o sinh( t / 4 G m )
v = ocosh(t/4GM) in F, (5.17)
{ ~ - -o sinh(t/4GM)
inP, (5.18)
- o cosh(t/4GM)
where
( r ]l/2er/4GM (5.19)
O= 1 2GM ]
in the region r > 2GM. It can be shown that the eigenvalue ~0,2 is real and positive.
Hence, we can always choose ~o, to be also real and positive. The set of all these
eigenfunctions f, is complete. From (5.21), one can readily verify that
Expand
ep(r,a, fl) = ~ f . ( r , a , fl)q.,
n
Therefore, the vacuum average of the occupation number operator (defined in (I)
only) is
tre 81rGMHatna n
(VACIa*.a.IVAC> = tre_8~GM H
= (e 8"cM'~. - 1) -1 , (5.28)
T.D. Lee / Are black holes black bodies 467
where a , and a*, are the standard annihilation and creation operators of the nth
oscillator in the sum (5.26). This distribution corresponds to a Bose-Einstein
distribution at the Hawking temperature (8rrGM) 1. Because the Schwarzschild
metric is assumed to hold in the region (I) at all time from t = - oo to t = + oo, the
"radiation" is a static one, similar to the situation in the accelerating frame
discussed before. As emphasized in the previous sections, this radiation should not a
priori be treated as incoherent, a situation quite different from the black-body
radiation in statistical mechanics.
A realistic black hole is nonstatic and therefore more complex. In appendix C we
study a soluble two-dimensional model in which the metric is a given time-depen-
dent function, simulating the formation of a black hole. From this model calculation
and from the above discussion of the static case, we are led to the conclusion that a
black hole is not a black body. At least in principle, an observer outside the black
hole can retrieve a fair amount of information about the physical state inside the
black hole, beyond his horizon.
In these discussions, we are concerned with the application of quantum field
theory to a limited space-time domain bounded by a horizon. The main question is
how to properly take into account the global nature of a quantum state, which
inevitably extends over to the space outside the horizon. As in the well-known
Einstein-Podolsky-Rosen experiments, the same global properties of the quantum
state can give correlations between particles that are well separated on a space-like
surface, therefore seemingly unrelated in any causal way. Here, because of the
macroscopic scale of the objects involved, the situation could be more striking. We
hope the central issue raised in this paper may lead to a much deeper understanding
of the relation between quantum mechanics and general relativity.
I wish to thank James Anderson and James Hartle for discussions, and Gary
Horowitz and Malcolm Perry for conversations which stimulated my interest in this
subject.
Appendix A
Since the expansion (1.12) is valid everywhere in the (X, T ) plane (including the
468 T.D. Lee / A re black holes black bodies
region (I) shown in fig. 1), it m a y be substituted into the above expression.
C o m p a r i n g b o t h sides of the new (A.1) at t = 0, and then their t-derivatives at the
s a m e instant, we find
where
( k l K ) = [_ egXdxexpi -
, (A.4)
Consequently,
k
( k l K ) = -~ <k l K ) , (A.7)
and (A.2) b e c o m e s
d ~ [ 6o \1/2
a-+°=f0 (A.8)
where the u p p e r signs are for k = o~ > 0 and the lower ones for k = - w < 0. Using
(A.8) and its hermitian conjugate, we can readily evaluate the v a c u u m expectation
value of a~a k. The result is (1.18):
massive field, the corresponding matrix elements can be more complicated. The
alternative derivation given in the text does not depend on any such explicit
calculations; it extends in a simple way the validity of (A.9) to a much wider class of
problems: the quantum may be massive, the space dimension can be arbitrary, the
underlying metric may have a nonvanishing Riemann tensor as in the Schwarzschild
case, etc.
Appendix B
~= #
.'-12[\0T1 \OX] ]
At a given T, expand ~ and its conjugate momentum in the same region as
o~d~2
¢ = fo ---~-e~(T)sinI2(X+ 1),
ood~
P= fo ---~-~(r)sinI2( X + l), (B.3)
H=fo ood~2 1
~-~-~[~2+~22~2 ] . (B.4)
:oo a¢
6~= j ldX2ff_~ sin I2(X + l ) . (B.5)
Substituting (B.5) back into (B.3), we may extend 0 from X > - l to X < - l.
470 T.D. Lee / Are black holes black bodies
Set T = 0. Rewrite (B.3), over the entire range - ~ ~< X ~< oc, as
~¢ d K einXQK, (B.6)
q)(X) = ~¢ 2~r
62(x) = f~¢
_ dSg
K e iKXI-IK (B.7)
where
Qg = - i ~ g e ml, Q ~ = i ~ z e ml
= d k e ikx' ,
1
when X = - - e g'' < 0. (B.10)
2~r qk g
1
P ( X ) = f~¢o~ ~-~
d k e -ikXp, when X = -g e gx > 0 , (B.11)
1
P(X)= f ~_ o¢2---~
d k e _ikX, p,k when X = - -g e gx' < 0, (B.12)
P ( X ) = gX62( X ) . (B.13)
Pk = f ~ d x eik~ p ( X ) . (B.15)
T.D. Lee / Are black holes black bodies 471
Substituting (B.6) into (B.14), (B.7) into (B.15), and using (B.13), we obtain
r~¢ d K
qk = J_o~ ~--ff~(klK)Q~: , (B.16)
~ dK ,
p~ = J _ ~ 5 ~ - ( k l K ) n ~ , (B.17)
where, as in (A.3)-(A.4),
(klK)
/2 e 'Kx '~Xdx, (B.18)
(klK) = f~e'~X-'k:'dX,
"0
(a.19)
dK
f_ -~--ff(klK)(k'lK)* = 2¢r~(k - k ' ) , (B.20)
q*=
f0 a t 27r -i(klf2)eim + i(k[ - f g ) e - i m ] ~ '
oo d~2
Pk= fo - ~ [i(kl~2)*e-m'-i(kl -I2)*em']~" (B.22)
Thus, when the mirror is placed at X = - l ~<0, beyond the horizon of Xacc, q, and
p , are also canonical variables. [In contrast, if X = - l > 0 the mirror is inside Gacc,
then q, and p , , defined by (B.9) and (B.11), do not satisfy (B.24) and therefore are
472 T.D. Lee / Are black holes black bodies
qk ~ (2o~)-1/2( ak + a t k ), (B.27)
In terms of the annihilation and creation operators A~2 and A~, in X, the corre-
sponding annihilation operators in Xacc become
When the (Dirichlet) mirror is placed at X = 0 (i.e., l = 0), the vacuum expectation
value of aka k, observed in the accelerating frame Sac c is
(~,)~/2
(VAC [a ~ a _ ,,, IVAC) =
(B.33)
As emphasized before, although the mirror is beyond the horizon, by staying within
the accelerating frame X .... an observer can nevertheless determine the existence
and the position of the mirror.
B.2. To understand the significance of (B.32), (B.33), let us examine the special
case when the (Dirichlet) mirror is at X = - 1 = 0. In this case, (B.29), (B.30) can
T.D. Lee / Are black holes black bodies 473
~ dw ( ~__\1/2
A a = i fo 2~r \ ~2 ) [ ( ~ a l ~ ) * a ~ ' - ( * a l a ) a - ~ ' - ( ~ l - fa)a: + (~l - fa)*a*-~']
=ii o
=dw(-~)l/2[-~i">/g(f_(~)at,+f+(w)a_,,,)
+ ~-'~/g(f*(,~)a~ +f*(,olat<~)]. (B.34)
Multiply AaIVAC ) by ~ - l / 2 7 : i ( ' / g ) d ~ and integrate from ~2 = 0 to oo, we derive,
on account of (B.34) and f~%T ('°'-~)/g d~2/12 = 21rg8(~' - ~),
(f_(o~)a t + f + (o~)a_,~)IVAC) = 0,
Defining
F- f -~ e-"'lgaLat-~"
- Jo (B.37)
we have
[ak, F ] = - e - " ° l g a t , , (B.38)
where o~ = Ikl. Because of (A.6), it follows that
f-(~)
e - ~"'/g (B.39)
f+(o~)
and therefore
Likewise
In terms of Ivac) in Zac~, the vacuum in Z is given by the first equation in (2.36):
IVAC) = const × e r l v a c ) ,
where instead of (B.37)
f ~ dt~
F=jo ~--~e-"'~/ga~at_,~. (B.41)
Appendix C
A TWO-DIMENSIONAL MODEL OF GRAVITATIONALCOLLAPSE
In gravitational collapse, one deals with a nonstatic metric. In the literature, a
two-dimensional model [4, 5] has often been used to illustrate the kind of physical
consequences that might be present in the formation of a black hole, especially if the
collapsing matter is distributed mainly on a shell. The same model will be studied in
this appendix, but with some minor variations. As we shall see, while an observer
outside the shell can detect "Hawking" radiation, it is also possible for him, through
measurements of various functions of field variables, all restricted to the same
outside region, to ascertain that the state may be quite different from an incoherent
mixture of black-body radiation quanta.
C. 1. The model. In this model the coordinates are X ° = T and X 1 = X, with the
trajectory of the shell along a given function X = L(T). The metric is assumed to be
2M d, 12
goo= - (1 - - -
X
, when X > L(T) (C.2)
2M)x 1
gxx (1
g0o - 1
gxa = 1 ' when X < L(T) (C.3)
X -- L(T)
.(x~O/~~
f / ~z_ = 2,tl-,t 0
T]
0
/ L1 /" ILo X
X = L(T)
/ 12 L [ (dL)2 (C.4)
dT] = L-2M 1+ ~ L-2M "
W e assume
with A > e > 0. Physically, this simulates a collapsing shell moving from L 0 at T = 0
to L 1 at T = T r M u c h of the analysis can be carried out without specifying the
details of h o w the shell changes from L o to L r [An example is shown in fig. 5.] Eq.
(C.4) gives
q,(X, T) = 0 at X = 0 (C.7)
(which simulates the usual radial boundary condition, r~ = 0, when the radius r = 0
in a three-space-dimensional problem). The lagrangian and hamiltonian for a free
field are
E= - fo~ ~/Z~ff~" d X,
0q~* 0,# ]
9c= fo°°[- (- ,g,)-l~goj02 + ( - I g l ) l / 2 g 1 1 O----)(O----x d X , (C.8)
where 1" denotes the hermitian conjugate as before, % = Oeo/OX~, eO~= g~"eO~, and
02 = _ ~ S - ~
0,~*
g OO OT '
02* = - f ~ Igl goo 0,~
OT
(C.9)
are the conjugate momenta of q~ and q,t. The quantization condition is the equal-time
commutator
[02(x, V), +(X', V)] = - i S ( X - X'). (C.10)
In the Schroedinger picture, q~ and °2 are T-independent, and the state vector IT)
satisfies
0
%1 T ) = i - ~ l T ) .
0,/,
0-~ = ' [ % ' ~ ] '
o02
OT = i [ % ' 0 2 ] " (C.11)
For 3(2 given by (C.8), the Heisenberg equations (C.11) yield the familiar wave
equation
0
OX. (1/-Z-/~-*") = O, (C.12)
T.D. Lee / Are black holes black bodies 477
where Igl is the determinant of the metric tensor. By using (C.1)-(C.3) we see that ¢
satisfies
02¢ 02¢
- at---i + -bTx2 = 0 when X > L(T),
02¢ 02¢
- OT------
T + OX---5 = 0 when X < L ( T ) , (C.13)
where
x = X+ 2Mln(X- 2M), (c.a4)
so that dx = d X / [ 1 - ( 2 M / X ) ] .
C.2. S-matrix. When T is less than 0, 9C becomes T-independent on account of
(C.6). It is convenient to expand the operator ¢(X, T) in terms of the plane wave
solutions:
= -1o) (C.17)
l o = L o+ 2 M l n 3 . (C.18)
¢(X,T)= foO~do~
2~r 2 ~1 [A°~t¢'~ut(X'T)+B'° out,0~o
out( X , T ) * ] , (C.19)
l 1 = L 1 + 2 M l n t. (C.22)
in - - 0o dt
-fo ep~(X,Z)*ep~",(g,z)~/-Iglg ~--~dX=2~r~(~-~'), (C.23)
and within each set, in or out, other commutators are zero. The transformation
matrix connecting these two sets of operators is the S-matrix.
By using the wave equation (C.12), we can extend the solutions q~" and ~out to all
time T in a standard way: Introduce the light-cone coordinates
z+=x+t,
Z +-- X + r . (C.26)
According to (C.13), any c. number solution of the wave equation can be written as
Along the dividing line X = L(T) occupied by the shell, we may regard, say, Z+ as
the independent variable (instead of T), thus through (C.4)-(C.6) we can express z+
as a function of Z+:
z + = ~ + ( Z + ), (C.28)
T.D. Lee / Are black holes black bodies 479
in which the left-hand side is the value of the dependent variable z+ and, following
Unruh, 2+(Z+) denotes the 2+-function of the independent variable Z+. Con-
versely, we may regard z+ as the independent variable and write along X = L(T)
Clearly, the functions 2 _+and Z_+ are the inverse of each other; i.e., for an arbitrary
variable y,
y=2+(Z+(y))=2+(2+(y))
=2_(Z_(y))=2 (2 ( y ) ) .
Next, we apply the wave equation (C.12) to derive the continuity condition that
relates the two solutions in (C.27). On the dividing line X-- L(T), let
dL
dS 0= - - dT, dS x = - d T
dT
V~-Igl ~ d S ~ (C.31)
Thus, among the four functions in (C.27), only one is independent. For example, if
we choose the independent function to be f+ (z +)= e-~,oz+, then the corresponding
F+_(Z+_)and f _ ( z ) are ___exp[-i~£+(___Z+)] and - e x p [ - i ~ 2 + ( - 2 ( z ) ) ] . For
T negative, these functions 2+ and 2 can be determined by considering only the
480 T.D. Lee / Are black holes black bodies
Both (C.35) and (C.36) are valid at all T; so are, then, the expansions (C.15) and
(C.19). T h r o u g h (C.19) and (C.24) A °ut and B °ut can be expressed in terms of
4)(X, T ) and its derivative at any T > T 1. Substituting (C.15) into these expressions,
we find
A°"'= -~-g-([
fo ~d~°' " tO I'~t ]+A~,+
m
' ['~l ~o']*_ Bin*),
"
floor = fo ~ -doo'
- ( [ ~,0 l o ~ ' ] * A ~ Y + [w]~o' l + B ~o'
in]1, (C.37)
2~r
where
in which the integration over X is to be evaluated at any fixed time T > / T v Since A~
a n d B~ ( a = in or out) are T-independent, the transformation matrices [~0i~0'] +_ are
also i n d e p e n d e n t of T.
T o c o m p u t e [~oi~o'] +, it is convenient to take a very large T, in which case in most
of the X-integration in (C.38) the relevant 2 + and 2 + functions are determined only
b y the final position of the shell, X = L~ (which is assumed to exist from T = T~ to
oe); it is quite easy to see that as T---, oe, that portion of the integration by itself
would give
2~r8 ( ~0 - to')e i(~1-g0> (C.39)
to [,ol,o']+ and zero to [~o1~o'] . The remaining part depends on the details of
X=L(T) for T~< T 1 which, in the integral (C.38), refers to an X-region near
infinity when T ~ o0 (e.g., the region z > 2l 1 - l 0 in fig. 5). Of course, this part
c a n n o t be neglected even in the limit, since it contains all the information a b o u t
H a w k i n g radiation, as we shall see. However, the existence of (C.39) shows that one
T.D. Lee / Are black holes black bodies 481
is not going to arrive at a pure Hawking radiation by using the S-matrix. This is
hardly surprising, since ~out carries information of the whole physical domain, both
inside and outside the shell.
C.3. Hawking radiation. To derive the Hawking radiation, it is necessary to
concentrate on measurements dealing with fields only in the outside region, which
for T>~ T 1 is simply X > L v
As mentioned before, when the parameter e in (C.5) approaches 0, the shell
collapses f r o m L 0 = 2 M + A at T = 0 to L 1 = 2 M at T 1, simulating the formation of
a black hole. In this limit, on account of (C.6) and (C.22), l I approaches - oz and
t 1 >/l o - l 1 becomes ~ . Correspondingly, the allowed range of x extends from
l 1 ~ - ~ to oo. Thus, an observer restricted to the outside of the shell might think
that his outside world in x and t could be the entire physical space-time domain.
Natural basis vectors for such an observer to adopt (in this region X > L1) would be
the plane wave solution of the first equation in (C.13), e "kx ~,t), but without any
X < L 1 portion, in contrast to ~out in (C.20). To enhance the sensitivity of our
analysis, we assume the positive parameter e to be nonzero but extremely small,
<< 2M.
In the following, we shall be interested only in T = T 1 and its immediate
neighborhood. F r o m (C.35), to construct ffin at T = T1, we need to know Z _ ( z ),
especially over the range of z for which the collapsing part of the shell-trajectory
L ( T ) is important. This corresponds to the region between the two dashed curves in
fig. 5:
l1 - t1 ~ z_ ~< 10,
i.e., at t = t 1
l 1 ~< X ~ t x + lo . (C.40)
in which 4~(x, t) refers to the same operator q~(X, T). If one wishes, one may regard
(C.42) as the definitions of a k and bt_k. Eq. (C.41) is then valid only inside the
region (C.40); outside, its right-hand side becomes zero. The parameter e > 0 can be
arbitrary. N o t e that if e ~ 0, we have l I ~ - ~ and l 0 + t 1 ~ oc and therefore the
integral in (C.42) extends over the entire range of x.
482 T.D. Lee / Are black holes black bodies
Substituting the expansion (C.15), which is valid everywhere, into (C.42), we find,
setting t = t 1,
/.~ dto' .
a k = -Io -2-~-~
" ((klto'}+Ai'+ {klto'}*-Bin*)'
So far, all these formulas are exact without approximations. While (C.43)-(C.44) are
similar to (C.37)-(C.38), unlike A °ut , A~outt and B .out. . BOUtt
. these operators a k, a~
a n d b k, btk do not form a complete set of independent b o n a fide annihilation and
creation operators.
T h e function @in at T = / ' 1 depends on the details of X = L ( T ) between T = 0 and
T v A particularly simple case is when the shell collapses from L 0 = 2 M + A to
L 1 = 2 M + e at the light-velocity; i.e.,
W h e n T = T 1 = L o -L1, I is
t I = l 0 - 11 (C.46)
as illustrated in fig. 5. When z_ is > l o, we need only the initial shell position
X = L o to determine the relevant ~+ and 2 functions in (C.35); in the region (C.47)
the collapsing trajectory (C.45) plays an i m p o r t a n t role. The result is, for X > L 1
and at T = T 1 (plus its immediate neighborhood),
where
111
~([½(lo + z _ ) ] = 2 M + - e x p - - ~ ( l o + z _ ) + O ( e )
e
] (c.sa)
where the estimate O(e) is valid for - oo ~<x ~<O(1). By using (C.48)-(C.51), the
integration (C.44) can be readily evaluated. We find, neglecting the O(e) term in
(C.51),
a~=jo f~do~'(~l/2,
~-~\~-7) L(" 112)e,oA.,+(o~l
in - ~ ) -e, , m,_,,~,
n , lj ,
• in
a_,~ = tAo e iwtI , b, = t• B ,i n e i~oq , (C.52)
where
1 (,o),-
12 = 2 e M ~ A e la/4M ¢0',
,o]o
(¢o I + ~2) = 4M(4M~2)'4M~F(--i4Mco) e +-2M,~, (c.53)
The last equation is identical to (A.5)-(A.6), with g = (4M)-a. Using the commuta-
tion relation (C.25) and the orthogonality relations between (~1 + I2), established in
484 T D. Lee / A re black holes black bodies
[a~,,aL,]=[bo,,bt,]=2rrS(oa-oa'),
[ao,,a,o,]=[b,~,bo,,]=O,
[a_,~,a_,~,]=[b_,o,b ~,,] = 0,
in IVAC>
A~ = & ,in IV A C > = 0 . (C.56)
but
(VAC[at_~,,a ,~]VAC> = (VACIbt_~,,b o,]VAC> = O .
do not carry any rapidly oscillating factors e + i,otl. Their product with a,~, b~, and
T.D. Lee / Are black holes black bodies 485
a~, b~ can produce some interesting non-vanishing expectation values, even when
the state vector is IVAC). For example,
81rM 1
to~ e 8*rM'°- 1 '
(c.58)
8~rM 1
(C.59)
t0~ e 8'rM'°- 1 "
Both would be zero if the radiation were genuinely black-body. By using (C.42), the
above operator products can be expressed entirely in terms of q, in the outside
region X > L 1. Thus, it is possible for an observer outside the collapsing shell to
ascertain that the state under consideration may be quite different from a black-body
radiation.
Because of the boundary condition (C.7), the present model somewhat resembles
the situation when there is a mirror in the Rindler case that was discussed in sect. 2.
This is why even for the IVAC) state, one can observe non-black-body correlations
between quanta. [Here, (C~58)-(C.59) play a role analogous to the previous (2.37).]
(ii) As shown in (2.38)-(2.39), it is possible for an observer in the accelerating frame
to detect the existence of a wave packet beyond his horizon. In the present case,
although (strictly speaking) there is no horizon, one can pose a similar problem: for
an observer restricted to the region outside the shell (say, X > L 1 and T >/T1), is it
possible for him through field measurements to gather information about the
physical world inside the shell, X < LI? As we shall see, the answer is yes.
F r o m (C.37)-(C.38), one notes that the IVAC) state defined by (C.56) contains
excitations when expressed in terms of quanta generated by the A~utt operators.
These excitations are described by ~out of (C.20), which correlates a wave of
frequency (or wave number) o~ in the outside region with that of a much higher
frequency
p = (tl/e)l/2to (C.60)
in the inside region. [Note that when X---, oo, x ---, X on account of (C.14); hence
asymptotically wave numbers in x and X become the same.] Therefore, if the state
vector I) at T near Tx is not IVAC), but IVAC) plus an additional coherent wave
packet amplitude of frequency p inside the shell [analogous to f d k C I ( k ) a ' k t l V A C )
in (2.38)], it would be possible for the observer, still restricted to X > L~, to detect
the deviation of I) from IVAC). In other words, just as in (2.39), he would be able
to determine the existence of such a wave packet, even though it was beyond his
486 T.D. Lee / Are black holes black bodies
References
[1] S.W. Hawking, Nature 248 (1974) 30; Comm. Math. Phys. 43 (1975) 199; Phys. Rev. D14 (1976)
2460
[2] W. Rindler, Am. J. Phys. 34 (1966) 1174
[3] S. Fulling, Phys. Rev. D7 (1973) 2850
[4] W.G. Unruh, Phys. Rev. D14 (1976) 870
[5] N.D. Birrell and P.C.W. Davies, Quantum fields in curved space (Cambridge University Press, 1982)
[6] R. Wald, Comm. Math. Phys. 45 (1975) 9; Phys. Rev. D13 (1976) 3176
[7] L. Parker, Phys. Rev. D12 (1976) 1519
[8] D.G. Boulware, Phys. Rev. D l l (1975) 1404; D13 (1976) 2169
[9] W. Israel, Phys. Lett. 57A (1976) 107
[10] W. Unruh and R. Wald, Phys. Rev. D29 (1984) 1047;
R. Haag, H. Narnhofer and U. Stein, Commun. Math. Phys. 94 (1984) 219
[11] J.B. Hartle and S.W. Hawking, Phys. Rev. D13 (1976) 2188
[12] G. Gibbons and M. Perry, Proc. Roy. Soc. A358 (1978) 467
[13] A. Einstein, B. Podolsky and N. Rosen, Phys. Rev. 47 (1935) 777
[14] D. Bohm and Y. Aharonov, Phys. Rev. 108 (1957) 1070; Nuovo Cim. 17 (1960) 964
[15] J.S. Bell, Physics 1 (1964) 195
[16] L.R. Kasday, J.D. Ullman and C.S. Wu, Nuovo Cim. 25B (1975) 633
[17] S. Tomonaga, Prog. Theor. Phys. 1 (1946) 27
[18] K. Freese, C.T. Hill and M. Mueller, Covariant functional Schroedinger formalism and application
to the Hawking effect (Fermilab preprint)
[19] J. Bardeen, L.N. Cooper and J.R. Schrieffer, Phys. Rev. 108 (1957) 1175
[20] M. Goldhaber, T.D. Lee and C.N. Yang, Phys. Rev. 112 (1958) 1796
[21] M.D. Kruskal, Phys. Rev. 119 (1960) 1743
[22] G. Szekeres, Publ. Mat. Debreen 7 (1960) 285
[23] T.D. Lee and Y. Pang, in preparation