how to fall into a black holeqpt.physics.harvard.edu/simons/verlinde.pdf · de grisogono delacour...
TRANSCRIPT
Simons Symposium
February, 2013
Quantum Entanglement: From Quantum Matter to String Theory
How to Fall into a Black Hole
Princeton University
PCTS
Herman Verlinde
Tuesday, February 19, 2013
?
Can a finite temperature CFTdescribe an object falling through the horizon of an AdS black hole?
The AdS/CFT correspondence has convinced usthat black hole physics follows the rules of QM.
Tuesday, February 19, 2013
?
Can a finite temperature CFTdescribe an object falling through the horizon of an AdS black hole?
The AdS/CFT correspondence has convinced usthat black hole physics follows the rules of QM.
Tuesday, February 19, 2013
?
Can a finite temperature CFTdescribe an object falling through the horizon of an AdS black hole?
The AdS/CFT correspondence has convinced usthat black hole physics follows the rules of QM.
2/6/13 9:30 AMIkepod Horizon Black Hole RS (LNIB) | Collectors Time
Page 1 of 3http://www.collectorstime.com/products/ikepod-horizon-black-hole-rs-lnib
SEND ENQUIRY
Approx Retail: $14,000.00
Price: $9,490.00 SEND ENQUIRY
BRANDS
Antoine Preziuso
Armand Nicolet
Audemars Piguet
Azimuth
Ball
Baume & Mercier
Bell & Ross
Bertolucci
Blu
Breguet
Breitling
Bvlgari
Carl F. Bucherer
Carlo Ferrara
Cartier
Chaumet
Chopard
Chronoswiss
Concord
Corum
Cvstos
De Grisogono
DeLaCour
Dubey & Schaldenbrand
Eberhard & Co
Edox
Eterna
Fortis
Franck Muller
Girard-Perregaux
Glashutte
Graham
Hublot
Ikepod
Ikepod Horizon Black Hole RS (LNIB)HHB10
Tell a Friend Like 2
Printer-friendly version
All new watches purchased from Collectors Time areguaranteed to be original genuine articles. Our products aresourced from authorized members of the distributionchannel and each of these new products comes withoriginal box, paper, and international warranty card.
All watches from Collectors Time are professionally packedthat ensures a well secured package when it arrives at yourdoorstep. We use renowned worldwide courier service forour shipping and all shipment will be accompanied with atracking number to trace your product. All orders are shipwithin 24 hours upon receipt of certified funds.
Details Guarantee
You may also like this
Quick Search: Keywords GO
Currency: US DollarSubscribe Newsletter: * Sign Up!
NEW ARRIVALS SPECIALS HOME WOMEN ORDER CONTACT US TESTIMONIALS
Tuesday, February 19, 2013
Hawking pair
horizon
b
a
r
e
Thursday, December 13, 2012
Figure 1: The firewall argument in a nut shell: quanta r and e can escape the black hole thanks to
entanglement with Hawking partners b and a. Monogamy prevents the partners to get entangled
with other interior quanta. So after the Page time, the black hole is in a maximally mixed state,
and a and b can not escape via the normal Hawking process.
bFIREWALL
ea
a
rb
Thursday, December 13, 2012
Figure 2: When the Hawking partner b leaves the black hole, it can not be entangled with any
mode a right behind the horizon: the horizon turns into a firewall. To avoid this conclusion, one
needs a mechanism for allowing entanglement to re-establish inside the black hole.
Hawking pair
horizon
b
a
r
e
Thursday, December 13, 2012
Figure 1: The firewall argument in a nut shell: quanta r and e can escape the black hole thanks to
entanglement with Hawking partners b and a. Monogamy prevents the partners to get entangled
with other interior quanta. So after the Page time, the black hole is in a maximally mixed state,
and a and b can not escape via the normal Hawking process.
bFIREWALL
ea
a
rb
Thursday, December 13, 2012
Figure 2: When the Hawking partner b leaves the black hole, it can not be entangled with any
mode a right behind the horizon: the horizon turns into a firewall. To avoid this conclusion, one
needs a mechanism for allowing entanglement to re-establish inside the black hole.
Rn =1
√Wn
i
i
iC †
n , R j
0a=
n
Rn
jn
a
RU i
0a
0b=
i 0U
if
i∈ Hcode
0 if i
/∈ Hcode
0U=
n
√Wn
na
nb.
Rn =1
√Wn
i
i
iC †
n , R j
0a=
n
Rn
jn
a
RU i
0a
0b=
i 0U
if
i∈ Hcode
0 if i
/∈ Hcode
0U=
n
√Wn
na
nb.
Tuesday, February 19, 2013
Hawking pair
horizon
b
a
r
e
Thursday, December 13, 2012
Figure 1: The firewall argument in a nut shell: quanta r and e can escape the black hole thanks to
entanglement with Hawking partners b and a. Monogamy prevents the partners to get entangled
with other interior quanta. So after the Page time, the black hole is in a maximally mixed state,
and a and b can not escape via the normal Hawking process.
bFIREWALL
ea
a
rb
Thursday, December 13, 2012
Figure 2: When the Hawking partner b leaves the black hole, it can not be entangled with any
mode a right behind the horizon: the horizon turns into a firewall. To avoid this conclusion, one
needs a mechanism for allowing entanglement to re-establish inside the black hole.
Hawking pair
horizon
b
a
r
e
Thursday, December 13, 2012
Figure 1: The firewall argument in a nut shell: quanta r and e can escape the black hole thanks to
entanglement with Hawking partners b and a. Monogamy prevents the partners to get entangled
with other interior quanta. So after the Page time, the black hole is in a maximally mixed state,
and a and b can not escape via the normal Hawking process.
bFIREWALL
ea
a
rb
Thursday, December 13, 2012
Figure 2: When the Hawking partner b leaves the black hole, it can not be entangled with any
mode a right behind the horizon: the horizon turns into a firewall. To avoid this conclusion, one
needs a mechanism for allowing entanglement to re-establish inside the black hole.
Tuesday, February 19, 2013
“What are the characteristics of a black hole?”
Tuesday, February 19, 2013
A black hole is always in a mixed state!
i∈ HM N = dimHM = eSBH , S
BH= 4πM2.
Ψ=
i∈Hcode
a i
i Φ i
↔ ρ
BH=
i∈Hcode
p i
i
i
i∈ HM N = dimHM = eSBH , S
BH= 4πM2.
Ψ=
i∈Hcode
a i
i Φ i
↔ ρ
BH=
i∈Hcode
p i
i
i
i∈ HM N = dimHM = eSBH , S
BH= 4πM2.
Ψ=
i∈Hcode
a i
i Φ i
↔ ρ
BH=
i∈Hcode
p i
i
i
i∈ HM N = dimHM = eSBH , S
BH= 4πM2.
Ψ=
i∈Hcode
a i
i Φ i
↔ ρ
BH=
i∈Hcode
p i
i
i
Tuesday, February 19, 2013
Hamiltonian evolution in Interaction Picture
Hilbert space splits up into three sectors:
with
H = HA⊗H
B⊗H
earlyR
HA= H
BH, H
B= H
lateR .
H = HBH
+HR+Hint,
H = HA⊗H
B⊗H
earlyR
HA= H
BH, H
B= H
lateR .
H = HBH
+HR+Hint,
Tuesday, February 19, 2013
Hamiltonian evolution in Interaction Picture
Hilbert space splits up into three sectors:
with
H = HA⊗H
B⊗H
earlyR
HA= H
BH, H
B= H
lateR .
H = HBH
+HR+Hint,
H = HA⊗H
B⊗H
earlyR
HA= H
BH, H
B= H
lateR .
H = HBH
+HR+Hint,
Tuesday, February 19, 2013
ρAB(0) =
i
p i
i
i ⊗
0b
0,
ρAB(τ) = U ρ(0)U † =
n,m
ρnmn
b
m
ρB =
n
Wn
nb
n ; Wn =
e−βEn
ZSB = logZ + βE
Initial mixed state
evolves into
B = Hawking atmosphere
ρAB(0) =
i
p i
i
i ⊗
0b
0,
ρAB(τ) = U ρ(0)U † =
n,m
ρnmn
b
m
ρB =
n
Wn
nb
n ; Wn =
e−βEn
ZSB = logZ + βE
ρAB(0) =
i
p i
i
i ⊗
0b
0,
ρAB(τ) = U ρ(0)U † =
n,m
ρnmn
b
m
ρB =
n
Wn
nb
n ; Wn =
e−βEn
ZSB = logZ + βE
Tuesday, February 19, 2013
Time evolution in terms of Kraus Operators C
C’s are ergodic `random’ matrices with known statistical properties!
cf. loss of coherence
of a quantum computer
U (τ) i
0b=
n
Cn
i n
.
n
C †nCn = M .
CnC†m=
1
Zδnm M−En
iC †
nCm
k= Wn δnm δ ik .
U (τ) i
0b=
n
Cn
i n
.
n
C †nCn = M .
CnC†m=
1
Zδnm M−En
iC †
nCm
k= Wn δnm δ ik .
“BAD” “GOOD”
b
Unitary:
U (τ) i
0b=
n
Cn
i n
.
n
C †nCn = M .
CnC†m=
1
Zδnm M−En
iC †
nCm
k= Wn δnm δ ik .
U (τ) i
0b=
n
Cn
i n
.
n
C †nCn = M .
CnC†m=
1
Zδnm M−En
iC †
nCm
k= Wn δnm δ ik .
Tuesday, February 19, 2013
ρ(E)2=
wE
Nc+ 2
dB(E)
NZ
ρ(E)−
dB(E)
NZ
2
M−E
S(E) = wEdB(E)
log
NZ
dB(E)
−
y
2 + ylog
1 + y +
2y + y2
y =N
Nc
e−βE
2dB(E),
young BH : SAB = logNc SA = logNc + logZ + βE
old BH : SAB = logN, SA = logN − βE
We can compute the Page curve of our model!
Two limiting cases:
ρ(E)2=
wE
Nc+ 2
dB(E)
NZ
ρ(E)−
dB(E)
NZ
2
M−E
S(E) = wEdB(E)
log
NZ
dB(E)
−
y
2 + ylog
1 + y +
2y + y2
y =N
Nc
e−βE
2dB(E),
young BH : SAB = logNc SA = logNc + logZ + βE
old BH : SAB = logN, SA = logN − βE
ρ(E)2=
wE
Nc+ 2
dB(E)
NZ
ρ(E)−
dB(E)
NZ
2
M−E
S(E) = wEdB(E)
log
NZ
dB(E)
−
y
2 + ylog
1 + y +
2y + y2
y =N
Nc
e−βE
2dB(E),
young BH : SAB = logNc SA = logNc + logZ + βE
old BH : SAB = logN, SA = logN − βE
ρ(E)2=
wE
Nc+ 2
dB(E)
NZ
ρ(E)−
dB(E)
NZ
2
M−E
S(E) = wEdB(E)
log
NZ
dB(E)
−
y
2 + ylog
1 + y +
2y + y2
y =N
Nc
e−βE
2dB(E),
young BH : SAB = logNc SA = logNc + logZ + βE
old BH : SAB = logN, SA = logN − βE
ρ(E)2=
wE
Nc+ 2
dB(E)
NZ
ρ(E)−
dB(E)
NZ
2
M−E
S(E) = wEdB(E)
log
NZ
dB(E)
−
y
2 + ylog
1 + y +
2y + y2
y =N
Nc
e−βE
2dB(E),
young BH : SAB = logNc SA = logNc + logZ + βE
old BH : SAB = logN, SA = logN − βE
Tuesday, February 19, 2013
Key Tool: Recovery Super Operator R
Produces the Unruh Vacuum!
Rn =1
√Wn
i
i
iC †
n , R j
0a=
n
Rn
jn
a
RU i
0a
0b=
i 0U
if
i∈ Hcode
0 if i
/∈ Hcode
0U=
n
√Wn
na
nb.
Rn =1
√Wn
i
i
iC †
n , R j
0a=
n
Rn
jn
a
RU i
0a
0b=
i 0U
if
i∈ Hcode
0 if i
/∈ Hcode
0U=
n
√Wn
na
nb.
Rn =1
√Wn
i
i
iC †
n , R j
0a=
n
Rn
jn
a
RU i
0a
0b=
i 0U
if
i∈ Hcode
0 if i
/∈ Hcode
0U=
n
√Wn
na
nb.
Tuesday, February 19, 2013
Reconstruction of Interior Operators:
A acts on black hole Hilbert spacesatisfies QFT operator algebra
Π = a
0R†R
0a, Π2 = Π.
ΠU i
0b=
U
i|0b
if i
∈ Hcode
0 if i
/∈ Hcode
A = a
0R†AR
0a
φ(y) = a
0R†φ(y)R
0a.
Tuesday, February 19, 2013
U+|
R+|
A B 0U=
n
√Wn|na|nb . (79)
Acknowledgements:
References
[1]
[2]
34
2 Setting the Stage
In this section and the next, we will describe the basic setup and main assumptions. Our
guiding principle is that a black hole is an ordinary quantum mechanical system, – much
like an atom, albeit a very large one and with an enormous level density of states – that
evolves and interacts with its environment via standard hamiltonian evolution. We start
by setting up some notation that will help us model the evaporation process in a way that
keeps the rules of quantum mechanics manifestly intact. Along the way, we will state a
version of the firewal argument put forward in [1], that is most directly addressed by our
subsequent study.
2.1 Black Holes States: Old and Young
Consider a macrosopic black hole. Suppose that we have measured its mass M with some
high accuracy. For now, we will ignore the uncertainty in M , and any other macroscopic
properties such as angular momentum and charge. We associate a Hilbert space HM
to the
black hole with mass M , whose states
i∈ HM (1)
are indistinguishable from a macroscopic perspective. Each state | i describes the interior
of a black hole, with microscopic properties that are hidden from the outside observer by the
stretched horizon. The number of black holes micro-states is determined by the Bekenstein
Hawking entropy
N = dimHM = eSBH , (2)
with SBH
= 4πM2. We will assume that the time evolution of every micro state describes
a black hole that slowly evaporates via the Hawking process. So as time progresses, every
state i
∈ HM will evolve into a new state given by the product of a micro state of a black
hole with smaller mass M = M − E and a radiation state with energy E.
To describe this evaporation process we need to enlarge our Hilbert space, so that it
incorporates all possible intermediate situations. So we introduce the total Hilbert space
H, given by the tensor product
H = HBH
⊗HR (3)
5
We read off that
RU | i |0a|0b =
| i | 0U if | i ∈ Hcode
0 if | i /∈ Hcode
(42)
where | 0U is the following familiar looking state, defined on the tensor product of the
Hilbert space of the outside radiation region with the ancillary Hilbert space
| 0U =
n
√Wn |na |n b . (43)
Given that Wn are the Boltzmann weights, we recognize this state as the Unruh vacuum
state, with the ancilla states playing the role of the interior radiation region. We see explic-
itly how the R operation works: it sweeps all entanglement between the time evolved black
hole and the emitted radiation, encoded in U |i|0b, into entanglement of the radiation
with the ancilla. Thus the recovery operation restores the purity of the black hole state,
provided it started out within the code space.
We see from (42) that the recovery operator R is not a unitary map, since it projects out
all states that do not map back into the code space. This means that the operation R as
defined in (40)-(39) in fact is not really a valid superoperator. Preferrably, one would want
the recovery operation to be a unitary operation. To achieve this, consider the operator
Π = a0|R†R|0a (44)
We may also write Π =
nR†nRn =
i ,n
1Wn
Cn | i i |C †n. If R had been a valid su-
peroperator, then Π would have been equal to the identity operator. Instead it defines a
projection operator on the space of all states that can be reached by acting with the Cn
operators on the code subspace. In other words, it is the projection of all states that the
code states can evolve into as a result of the evaporation process.
A straightforward calculation, similar to the one done in eqn (41), shows that Π indeed
has the property of a projection operator:
Π2= Π (45)
To complete the recovery super operator, one thus adds one extra term to the right-hand
21
Acknowledgements
We thank Bartek Chech, Jan de Boer, Daniel Harlow, Steve Jackson,
helpful discussions. The work of HV is supported by NSF grant PHY-0756966.
Appendix A: Some calculations
b
0U †ABU |0b =
0b
U † 0a|R†AR|0aBU |0b (69)
=0b
a0|U †R†ABRU |0a|0b (70)
= 0UAB| 0U code (71)
trρ(τ)AB
=
i ,k
ρ ik i0b
U †ABU |0b|k
(72)
=
i ,k
ρ ik i0U
AB| 0U|k= 0U
AB| 0U (73)
πnm =1
√WnWm
i
Cn | i i |C †m (74)
πns πtm =1
√WnWsWtWm
i ,k
Cn| i i |C †sCt |kk|C †
m (75)
=δst√WnWm
i
Cn | i i |C †m = δst πnm (76)
32
Acknowledgements
We thank Bartek Chech, Jan de Boer, Daniel Harlow, Steve Jackson,
helpful discussions. The work of HV is supported by NSF grant PHY-0756966.
Appendix A: Some calculations
0b
U †ABU |0b =0b
U † 0a|R†AR|0aBU |0b (69)
=0b
a0|U †R†ABRU |0a|0b (70)
= 0UAB| 0U code (71)
trρ(τ)AB
=
i ,k
ρ ik i0b
U †ABU |0b|k
(72)
=
i ,k
ρ ik i0U
AB| 0U|k= 0U
AB| 0U (73)
πnm =1
√WnWm
i
Cn | i i |C †m (74)
πns πtm =1
√WnWsWtWm
i ,k
Cn| i i |C †sCt |kk|C †
m (75)
=δst√WnWm
i
Cn | i i |C †m = δst πnm (76)
32
We read off that
RU | i |0a|0b =
| i | 0U if | i ∈ Hcode
0 if | i /∈ Hcode
(42)
where | 0U is the following familiar looking state, defined on the tensor product of the
Hilbert space of the outside radiation region with the ancillary Hilbert space
| 0U =
n
√Wn |na |n b . (43)
Given that Wn are the Boltzmann weights, we recognize this state as the Unruh vacuum
state, with the ancilla states playing the role of the interior radiation region. We see explic-
itly how the R operation works: it sweeps all entanglement between the time evolved black
hole and the emitted radiation, encoded in U |i|0b, into entanglement of the radiation
with the ancilla. Thus the recovery operation restores the purity of the black hole state,
provided it started out within the code space.
We see from (42) that the recovery operator R is not a unitary map, since it projects out
all states that do not map back into the code space. This means that the operation R as
defined in (40)-(39) in fact is not really a valid superoperator. Preferrably, one would want
the recovery operation to be a unitary operation. To achieve this, consider the operator
Π = a0|R†R|0a (44)
We may also write Π =
nR†nRn =
i ,n
1Wn
Cn | i i |C †n. If R had been a valid su-
peroperator, then Π would have been equal to the identity operator. Instead it defines a
projection operator on the space of all states that can be reached by acting with the Cn
operators on the code subspace. In other words, it is the projection of all states that the
code states can evolve into as a result of the evaporation process.
A straightforward calculation, similar to the one done in eqn (41), shows that Π indeed
has the property of a projection operator:
Π2= Π (45)
To complete the recovery super operator, one thus adds one extra term to the right-hand
21
the outgoing field φout = φ. The corresponding interaction operator V(t) can be expanded
in modes Vω and V †ω, that map black hole states with mass M to states with mass M ±ω.
The matrix elements Vij(ω) = jVω
i determine the emission and absorption rates of
particles of frequency ω. Semi-classically this process is described by the creation of a
Hawking pair. Hence it is natural to conjecture that the operators V(ω) and V †(ω) have
something to do with the creating or annihilating particles behind the horizon.
In principle one can express the Cn operators in terms of theV operators as follows: one
inserts the Fock states of the radiation and commutes the annihilation operators through
the evolution operator. One obtains an expression involving products of a number of V
operators, one for every emitted particle. It is suggestive that the number of V insertions
matches the number of infalling partner modes of the Hawking pairs. This suggests that
it is natural to view the transition amplitudes Cn as creating a state with a collection of
particles behind the horizon. Our aim is to make this intuition more precise.
There exists of course a close parallel between the above discussion and the GKPW
dictionary that underlies the AdS/CFT correspondence. Indeed, we may compare the
process of falling into a black hole with entering the near-horizon geometry of a stack of
D-branes. In this comparison, the AdS space replaces the black hole interior geometry, the
CFT represents the interior Hilbert space, the value of the scalar field φ at the stretched
horizon plays the role of the asymptotic boundary value of the bulk field in AdS, and the
V operators represent the CFT operator dual to φ.
2.4 Old and Young Black Holes Revisited
As a first application of our formal set up, let us revisit the time evolution of the young
and old black hole states (6) and (8).
First consider the young black hole. At time τ it has evolved into a state of the form
ρyoungAB
= U i
| 0
b
0 i
U † (26)
which we may expand in terms of the radiation basis as in eqn (12), with
ρyoungnm
= Cn
i
iC †
m (27)
We now make the assumption that the density matrix on the radiation states takes the
15
2 Setting the Stage
In this section and the next, we will describe the basic setup and main assumptions. Our
guiding principle is that a black hole is an ordinary quantum mechanical system, – much
like an atom, albeit a very large one and with an enormous level density of states – that
evolves and interacts with its environment via standard hamiltonian evolution. We start
by setting up some notation that will help us model the evaporation process in a way that
keeps the rules of quantum mechanics manifestly intact. Along the way, we will state a
version of the firewal argument put forward in [1], that is most directly addressed by our
subsequent study.
2.1 Black Holes States: Old and Young
Consider a macrosopic black hole. Suppose that we have measured its mass M with some
high accuracy. For now, we will ignore the uncertainty in M , and any other macroscopic
properties such as angular momentum and charge. We associate a Hilbert space HM
to the
black hole with mass M , whose states
i∈ HM (1)
are indistinguishable from a macroscopic perspective. Each state | i describes the interior
of a black hole, with microscopic properties that are hidden from the outside observer by the
stretched horizon. The number of black holes micro-states is determined by the Bekenstein
Hawking entropy
N = dimHM = eSBH , (2)
with SBH
= 4πM2. We will assume that the time evolution of every micro state describes
a black hole that slowly evaporates via the Hawking process. So as time progresses, every
state i
∈ HM will evolve into a new state given by the product of a micro state of a black
hole with smaller mass M = M − E and a radiation state with energy E.
To describe this evaporation process we need to enlarge our Hilbert space, so that it
incorporates all possible intermediate situations. So we introduce the total Hilbert space
H, given by the tensor product
H = HBH
⊗HR (3)
5
R
form depends on unknown details of Planck scale physics. It looks like cheating to assume
that Alice can access these observables, without actually having such detailed knowledge.
The situation is of course similar to other holographic dualities, like AdS/CFT. From the
dual perspective, the interior space-time emerges in a mysterious way. But for Alice, the
smooth local space-time is a classical reality, relative to which the operators (52) behave
like ordinary observables. So she has no trouble using them. Note, however, that Alice
does not organize the modes in terms of their Schwarschild energy ω, but reassembles them
into local Minkowski modes, that make no reference to the precise location of the horizon.
To make this a bit more concrete, consider the two point function of local field operators,
say one behind and one in front of the horizon. The general result (58) gives that
trρφa(y)φb(x)
= 0U |φa(y)φb(x)| 0U (61)
where the subscript indicates that φa is made up from a-modes, and φb from b-modes.
What is the meaning of this expectation value? Strictly speaking, it does not yet describe
what we want. In our discussion so far, we have only included the outgoing modes that
are emitted by the black hole. The ancillary a modes play their guest role as the Hawking
partner of the outgoing b modes: they initially live on the left region in fig 1, but can be
analytically continued to live just behind the future horizon. To really get the complete
set of modes behind the horizon, we would need to extend our framework a bit, by adding
the infalling modes. Since the horizon state looks smooth, the infalling modes can be
analytically continued to the region behind the horizon. So we have indeed explained how
a QFT mode, and by extension any object, can fall into a black hole.
Note that we can also try to go beyond the past horizon, by switching time direction.
The outgoing modes behind the past horizon are obtained by analytic continuation of
outgoing modes on the right region in fig 1. The time-flipped Hawking partners of the
incoming modes live behind the past horizon, and are obtain by analytic continuation of
modes on the left region in fig 1. To define them, we would need to introduce transition
amplitudes for the ingoing modes involved in the black hole formation process.
3.3 Ancilla versus Black Hole Final State
Our use of the ancillary Hilbert space Ha is somewhat reminiscent of the black hole final
state proposal of [2]. However, there is a key difference: unlike the final state proposal, our
procedure is completely in accord with the conventional rules of quantum mechanics. We
have stated explicitly in what Hilbert space we are working, and time evolution is unitary.
Our internal operators involve an expectation value inside an auxilary Hilbert space Ha,
26
U
horizon
trρφa(y)φb(x)
=
0U
φa(y)φb(x) 0U
,
U+|
R+|
A B 0U=
n
√Wn|na|nb . (79)
Acknowledgements:
References
[1]
[2]
34
2 Setting the Stage
In this section and the next, we will describe the basic setup and main assumptions. Our
guiding principle is that a black hole is an ordinary quantum mechanical system, – much
like an atom, albeit a very large one and with an enormous level density of states – that
evolves and interacts with its environment via standard hamiltonian evolution. We start
by setting up some notation that will help us model the evaporation process in a way that
keeps the rules of quantum mechanics manifestly intact. Along the way, we will state a
version of the firewal argument put forward in [1], that is most directly addressed by our
subsequent study.
2.1 Black Holes States: Old and Young
Consider a macrosopic black hole. Suppose that we have measured its mass M with some
high accuracy. For now, we will ignore the uncertainty in M , and any other macroscopic
properties such as angular momentum and charge. We associate a Hilbert space HM
to the
black hole with mass M , whose states
i∈ HM (1)
are indistinguishable from a macroscopic perspective. Each state | i describes the interior
of a black hole, with microscopic properties that are hidden from the outside observer by the
stretched horizon. The number of black holes micro-states is determined by the Bekenstein
Hawking entropy
N = dimHM = eSBH , (2)
with SBH
= 4πM2. We will assume that the time evolution of every micro state describes
a black hole that slowly evaporates via the Hawking process. So as time progresses, every
state i
∈ HM will evolve into a new state given by the product of a micro state of a black
hole with smaller mass M = M − E and a radiation state with energy E.
To describe this evaporation process we need to enlarge our Hilbert space, so that it
incorporates all possible intermediate situations. So we introduce the total Hilbert space
H, given by the tensor product
H = HBH
⊗HR (3)
5
We read off that
RU | i |0a|0b =
| i | 0U if | i ∈ Hcode
0 if | i /∈ Hcode
(42)
where | 0U is the following familiar looking state, defined on the tensor product of the
Hilbert space of the outside radiation region with the ancillary Hilbert space
| 0U =
n
√Wn |na |n b . (43)
Given that Wn are the Boltzmann weights, we recognize this state as the Unruh vacuum
state, with the ancilla states playing the role of the interior radiation region. We see explic-
itly how the R operation works: it sweeps all entanglement between the time evolved black
hole and the emitted radiation, encoded in U |i|0b, into entanglement of the radiation
with the ancilla. Thus the recovery operation restores the purity of the black hole state,
provided it started out within the code space.
We see from (42) that the recovery operator R is not a unitary map, since it projects out
all states that do not map back into the code space. This means that the operation R as
defined in (40)-(39) in fact is not really a valid superoperator. Preferrably, one would want
the recovery operation to be a unitary operation. To achieve this, consider the operator
Π = a0|R†R|0a (44)
We may also write Π =
nR†nRn =
i ,n
1Wn
Cn | i i |C †n. If R had been a valid su-
peroperator, then Π would have been equal to the identity operator. Instead it defines a
projection operator on the space of all states that can be reached by acting with the Cn
operators on the code subspace. In other words, it is the projection of all states that the
code states can evolve into as a result of the evaporation process.
A straightforward calculation, similar to the one done in eqn (41), shows that Π indeed
has the property of a projection operator:
Π2= Π (45)
To complete the recovery super operator, one thus adds one extra term to the right-hand
21
Acknowledgements
We thank Bartek Chech, Jan de Boer, Daniel Harlow, Steve Jackson,
helpful discussions. The work of HV is supported by NSF grant PHY-0756966.
Appendix A: Some calculations
b
0U †ABU |0b =
0b
U † 0a|R†AR|0aBU |0b (69)
=0b
a0|U †R†ABRU |0a|0b (70)
= 0UAB| 0U code (71)
trρ(τ)AB
=
i ,k
ρ ik i0b
U †ABU |0b|k
(72)
=
i ,k
ρ ik i0U
AB| 0U|k= 0U
AB| 0U (73)
πnm =1
√WnWm
i
Cn | i i |C †m (74)
πns πtm =1
√WnWsWtWm
i ,k
Cn| i i |C †sCt |kk|C †
m (75)
=δst√WnWm
i
Cn | i i |C †m = δst πnm (76)
32
Acknowledgements
We thank Bartek Chech, Jan de Boer, Daniel Harlow, Steve Jackson,
helpful discussions. The work of HV is supported by NSF grant PHY-0756966.
Appendix A: Some calculations
0b
U †ABU |0b =0b
U † 0a|R†AR|0aBU |0b (69)
=0b
a0|U †R†ABRU |0a|0b (70)
= 0UAB| 0U code (71)
trρ(τ)AB
=
i ,k
ρ ik i0b
U †ABU |0b|k
(72)
=
i ,k
ρ ik i0U
AB| 0U|k= 0U
AB| 0U (73)
πnm =1
√WnWm
i
Cn | i i |C †m (74)
πns πtm =1
√WnWsWtWm
i ,k
Cn| i i |C †sCt |kk|C †
m (75)
=δst√WnWm
i
Cn | i i |C †m = δst πnm (76)
32
We read off that
RU | i |0a|0b =
| i | 0U if | i ∈ Hcode
0 if | i /∈ Hcode
(42)
where | 0U is the following familiar looking state, defined on the tensor product of the
Hilbert space of the outside radiation region with the ancillary Hilbert space
| 0U =
n
√Wn |na |n b . (43)
Given that Wn are the Boltzmann weights, we recognize this state as the Unruh vacuum
state, with the ancilla states playing the role of the interior radiation region. We see explic-
itly how the R operation works: it sweeps all entanglement between the time evolved black
hole and the emitted radiation, encoded in U |i|0b, into entanglement of the radiation
with the ancilla. Thus the recovery operation restores the purity of the black hole state,
provided it started out within the code space.
We see from (42) that the recovery operator R is not a unitary map, since it projects out
all states that do not map back into the code space. This means that the operation R as
defined in (40)-(39) in fact is not really a valid superoperator. Preferrably, one would want
the recovery operation to be a unitary operation. To achieve this, consider the operator
Π = a0|R†R|0a (44)
We may also write Π =
nR†nRn =
i ,n
1Wn
Cn | i i |C †n. If R had been a valid su-
peroperator, then Π would have been equal to the identity operator. Instead it defines a
projection operator on the space of all states that can be reached by acting with the Cn
operators on the code subspace. In other words, it is the projection of all states that the
code states can evolve into as a result of the evaporation process.
A straightforward calculation, similar to the one done in eqn (41), shows that Π indeed
has the property of a projection operator:
Π2= Π (45)
To complete the recovery super operator, one thus adds one extra term to the right-hand
21
the outgoing field φout = φ. The corresponding interaction operator V(t) can be expanded
in modes Vω and V †ω, that map black hole states with mass M to states with mass M ±ω.
The matrix elements Vij(ω) = jVω
i determine the emission and absorption rates of
particles of frequency ω. Semi-classically this process is described by the creation of a
Hawking pair. Hence it is natural to conjecture that the operators V(ω) and V †(ω) have
something to do with the creating or annihilating particles behind the horizon.
In principle one can express the Cn operators in terms of theV operators as follows: one
inserts the Fock states of the radiation and commutes the annihilation operators through
the evolution operator. One obtains an expression involving products of a number of V
operators, one for every emitted particle. It is suggestive that the number of V insertions
matches the number of infalling partner modes of the Hawking pairs. This suggests that
it is natural to view the transition amplitudes Cn as creating a state with a collection of
particles behind the horizon. Our aim is to make this intuition more precise.
There exists of course a close parallel between the above discussion and the GKPW
dictionary that underlies the AdS/CFT correspondence. Indeed, we may compare the
process of falling into a black hole with entering the near-horizon geometry of a stack of
D-branes. In this comparison, the AdS space replaces the black hole interior geometry, the
CFT represents the interior Hilbert space, the value of the scalar field φ at the stretched
horizon plays the role of the asymptotic boundary value of the bulk field in AdS, and the
V operators represent the CFT operator dual to φ.
2.4 Old and Young Black Holes Revisited
As a first application of our formal set up, let us revisit the time evolution of the young
and old black hole states (6) and (8).
First consider the young black hole. At time τ it has evolved into a state of the form
ρyoungAB
= U i
| 0
b
0 i
U † (26)
which we may expand in terms of the radiation basis as in eqn (12), with
ρyoungnm
= Cn
i
iC †
m (27)
We now make the assumption that the density matrix on the radiation states takes the
15
2 Setting the Stage
In this section and the next, we will describe the basic setup and main assumptions. Our
guiding principle is that a black hole is an ordinary quantum mechanical system, – much
like an atom, albeit a very large one and with an enormous level density of states – that
evolves and interacts with its environment via standard hamiltonian evolution. We start
by setting up some notation that will help us model the evaporation process in a way that
keeps the rules of quantum mechanics manifestly intact. Along the way, we will state a
version of the firewal argument put forward in [1], that is most directly addressed by our
subsequent study.
2.1 Black Holes States: Old and Young
Consider a macrosopic black hole. Suppose that we have measured its mass M with some
high accuracy. For now, we will ignore the uncertainty in M , and any other macroscopic
properties such as angular momentum and charge. We associate a Hilbert space HM
to the
black hole with mass M , whose states
i∈ HM (1)
are indistinguishable from a macroscopic perspective. Each state | i describes the interior
of a black hole, with microscopic properties that are hidden from the outside observer by the
stretched horizon. The number of black holes micro-states is determined by the Bekenstein
Hawking entropy
N = dimHM = eSBH , (2)
with SBH
= 4πM2. We will assume that the time evolution of every micro state describes
a black hole that slowly evaporates via the Hawking process. So as time progresses, every
state i
∈ HM will evolve into a new state given by the product of a micro state of a black
hole with smaller mass M = M − E and a radiation state with energy E.
To describe this evaporation process we need to enlarge our Hilbert space, so that it
incorporates all possible intermediate situations. So we introduce the total Hilbert space
H, given by the tensor product
H = HBH
⊗HR (3)
5
R
form depends on unknown details of Planck scale physics. It looks like cheating to assume
that Alice can access these observables, without actually having such detailed knowledge.
The situation is of course similar to other holographic dualities, like AdS/CFT. From the
dual perspective, the interior space-time emerges in a mysterious way. But for Alice, the
smooth local space-time is a classical reality, relative to which the operators (52) behave
like ordinary observables. So she has no trouble using them. Note, however, that Alice
does not organize the modes in terms of their Schwarschild energy ω, but reassembles them
into local Minkowski modes, that make no reference to the precise location of the horizon.
To make this a bit more concrete, consider the two point function of local field operators,
say one behind and one in front of the horizon. The general result (58) gives that
trρφa(y)φb(x)
= 0U |φa(y)φb(x)| 0U (61)
where the subscript indicates that φa is made up from a-modes, and φb from b-modes.
What is the meaning of this expectation value? Strictly speaking, it does not yet describe
what we want. In our discussion so far, we have only included the outgoing modes that
are emitted by the black hole. The ancillary a modes play their guest role as the Hawking
partner of the outgoing b modes: they initially live on the left region in fig 1, but can be
analytically continued to live just behind the future horizon. To really get the complete
set of modes behind the horizon, we would need to extend our framework a bit, by adding
the infalling modes. Since the horizon state looks smooth, the infalling modes can be
analytically continued to the region behind the horizon. So we have indeed explained how
a QFT mode, and by extension any object, can fall into a black hole.
Note that we can also try to go beyond the past horizon, by switching time direction.
The outgoing modes behind the past horizon are obtained by analytic continuation of
outgoing modes on the right region in fig 1. The time-flipped Hawking partners of the
incoming modes live behind the past horizon, and are obtain by analytic continuation of
modes on the left region in fig 1. To define them, we would need to introduce transition
amplitudes for the ingoing modes involved in the black hole formation process.
3.3 Ancilla versus Black Hole Final State
Our use of the ancillary Hilbert space Ha is somewhat reminiscent of the black hole final
state proposal of [2]. However, there is a key difference: unlike the final state proposal, our
procedure is completely in accord with the conventional rules of quantum mechanics. We
have stated explicitly in what Hilbert space we are working, and time evolution is unitary.
Our internal operators involve an expectation value inside an auxilary Hilbert space Ha,
26
U
horizon
trρφa(y)φb(x)
=
0U
φa(y)φb(x) 0U
,
Tuesday, February 19, 2013
ΠU i
0b= U
i0
b+ error.
error = EU i
0b, with E =
n
Ncode
NnM−En.
Error of the Recovery Operation:
Reconstruction breaksdown for maximally
mixed states!
Firewall ParadoxAMPS
ΠU i
0b= U
i0
b+ error.
error = EU i
0b, with E =
n
Ncode
NnM−En.
Tuesday, February 19, 2013
ΠU i
0b= U
i0
b+ error.
error = EU i
0b, with E =
n
Ncode
NnM−En.
Error of the Recovery Operation:
Reconstruction breaksdown for maximally
mixed states!
Firewall ParadoxAMPS
ΠU i
0b= U
i0
b+ error.
error = EU i
0b, with E =
n
Ncode
NnM−En.
Reconstruction worksfor all states |Psi) that fit inside a code space!
?
Tuesday, February 19, 2013
Decoherence
In practice this means that not all of the quantum information in the early state can be
extracted in real time, since it is unsuitable for delicate quantum measurements involving
interference experiments, etc. In other words, in it is unavoidable that part of the informa-
tion contained decoheres and becomes, as far as the infalling observer is concerned, classical
information. An essential difference between classical and quantum information is that the
latter is delicate and exclusive – it can not be cloned, and while it can be highly non-local,
it can not be simultaneously accessed at different location first can be duplicated – once
it is known in some location it can be broadcast throughout space and become knowledge
that is shared by different semi-classical observers.
OBH
=
c
O(c)BH
⊗Ωc
Ωc
(10)
6
Solving the problem:
Does this preserve linearity and locality?
Tuesday, February 19, 2013
Decoherence
In practice this means that not all of the quantum information in the early state can be
extracted in real time, since it is unsuitable for delicate quantum measurements involving
interference experiments, etc. In other words, in it is unavoidable that part of the informa-
tion contained decoheres and becomes, as far as the infalling observer is concerned, classical
information. An essential difference between classical and quantum information is that the
latter is delicate and exclusive – it can not be cloned, and while it can be highly non-local,
it can not be simultaneously accessed at different location first can be duplicated – once
it is known in some location it can be broadcast throughout space and become knowledge
that is shared by different semi-classical observers.
OBH
=
c
O(c)BH
⊗Ωc
Ωc
(10)
6
Solving the problem:
Does this preserve linearity and locality?
Tuesday, February 19, 2013
Using QECC, we’ve given a general construction of localQFT operators relative to which the horizon is smooth.*
Generic BH states describe a semi-classical space-time with an interior geometry and a smooth event horizon. *
Some conclusions:
Maximally mixed BH state = sum of semi-classical states Einstein locality + general logic => old BHs have no firewall*
Formulate of the rules of observer complementarity
Some outstanding challenges:
Trace the flow of quantum information and entanglement
Quantify violations of locality and limitations of effective QFT.
Tuesday, February 19, 2013
Hawking pair
horizon
b
a
r
e
Thursday, December 13, 2012
bFIREWALL
ea
a
rb
Thursday, December 13, 2012
Tuesday, February 19, 2013
Code bits
Measurement:entangles qubits with
environment, produces classical information. r
a
b
e
horizon
Thursday, December 13, 2012
Tuesday, February 19, 2013
ba
Code bits
horizon
Measurement:enables the quantum
teleportation of Hawkingquanta back into the black hole interior
assists QT protocol
r
a
b
e
Tuesday, January 22, 2013
Tuesday, February 19, 2013
Via the Einstein equations, this will result in correspondingly large quantum fluctu-ations in the local background geometry on the Cauchy surface Σ. For a given cut-offand Cauchy surface Σ, one can in principle calculate these by integrating the cumulativegravitational effect of the local stress-energy fluctuations. It is clear that when thesegeometry fluctuations become too large, relative to the cut-off scale, the semi-classical de-scription of the Hilbert space as the space of states on a given Cauchy slice breaks down.We will call a cut-off violating this bound super-critical, and we will call it sub-critical orsemi-classical if the fluctuations of the geometry are controlled in this sense. It is clearfrom the above discussion that semi-classical cut-offs have a minimal size.
The question arises, however, how such a semi-classical truncation of the Hilbert spacemust be interpreted at a fundamental level? On the one hand, any short distance cut-offwould appear to constitute an unacceptable violation of Lorentz invariance (= frame-independence), since different observers will in general be inclined to truncate the Hilbertspace in different ways. On the other hand, there is no correspondence principle that tellsus that the Hilbert space must extend into this super-critical regime. On the contrary: ifwe would assume it does extend into the super-critical regime, we would need to explainwhy there are no large space-time fluctuations at scales much larger than the Planck scale.
Thus it seems we are faced with a dilemma: apparently we must either give up strictLorentz invariance, or find a way to deal with a Hilbert space supported on field config-urations with arbitrarily high frequencies. A way out of this dilemma is provided by theblack hole complementarity principle proposed in [3, 6, 5].
We now propose a new formulation of this principle, which we name space-time com-plementarity, because its formulation and possible consequences are in principle not re-stricted to the black hole context. The spirit in which this principle is meant is as aproposal for a reasonable effective description of some underlying, consistent theory ofquantum gravity, such as the one provided by string theory.† Clearly, however, a muchmore detailed knowledge of this underlying fundamental theory is necessary to verify andquantify this effective description.
Space-time complementarity: A variable cut-off scale (x) on a Cauchy surface
Σ provides a permissible semi-classical description of the second quantized Hilbert
space, only when the quantum fluctuations of the local background geometry induced
by the corresponding stress-energy fluctuations do not exceed the cut-off scale itself.
All critical cut-off scales that saturate this requirement provide complete, comple-
mentary descriptions of the Hilbert space.
The key step here is that, while different observers may under certain circumstances be
† L. Susskind has also advocated that a complementarity principle of the type formulatedin this section may be realized in string theory. There indeed exist several indications that,compared to local field theory, string theory has drastically fewer degrees of freedom at shortdistances.
5
Space-Time Complementarity:
Kiem, EV & HV, 1995
Via the Einstein equations, this will result in correspondingly large quantum fluctu-ations in the local background geometry on the Cauchy surface Σ. For a given cut-offand Cauchy surface Σ, one can in principle calculate these by integrating the cumulativegravitational effect of the local stress-energy fluctuations. It is clear that when thesegeometry fluctuations become too large, relative to the cut-off scale, the semi-classical de-scription of the Hilbert space as the space of states on a given Cauchy slice breaks down.We will call a cut-off violating this bound super-critical, and we will call it sub-critical orsemi-classical if the fluctuations of the geometry are controlled in this sense. It is clearfrom the above discussion that semi-classical cut-offs have a minimal size.
The question arises, however, how such a semi-classical truncation of the Hilbert spacemust be interpreted at a fundamental level? On the one hand, any short distance cut-offwould appear to constitute an unacceptable violation of Lorentz invariance (= frame-independence), since different observers will in general be inclined to truncate the Hilbertspace in different ways. On the other hand, there is no correspondence principle that tellsus that the Hilbert space must extend into this super-critical regime. On the contrary: ifwe would assume it does extend into the super-critical regime, we would need to explainwhy there are no large space-time fluctuations at scales much larger than the Planck scale.
Thus it seems we are faced with a dilemma: apparently we must either give up strictLorentz invariance, or find a way to deal with a Hilbert space supported on field config-urations with arbitrarily high frequencies. A way out of this dilemma is provided by theblack hole complementarity principle proposed in [3, 6, 5].
We now propose a new formulation of this principle, which we name space-time com-plementarity, because its formulation and possible consequences are in principle not re-stricted to the black hole context. The spirit in which this principle is meant is as aproposal for a reasonable effective description of some underlying, consistent theory ofquantum gravity, such as the one provided by string theory.† Clearly, however, a muchmore detailed knowledge of this underlying fundamental theory is necessary to verify andquantify this effective description.
Space-time complementarity: A variable cut-off scale (x) on a Cauchy surface
Σ provides a permissible semi-classical description of the second quantized Hilbert
space, only when the quantum fluctuations of the local background geometry induced
by the corresponding stress-energy fluctuations do not exceed the cut-off scale itself.
All critical cut-off scales that saturate this requirement provide complete, comple-
mentary descriptions of the Hilbert space.
The key step here is that, while different observers may under certain circumstances be
† L. Susskind has also advocated that a complementarity principle of the type formulatedin this section may be realized in string theory. There indeed exist several indications that,compared to local field theory, string theory has drastically fewer degrees of freedom at shortdistances.
5
Tuesday, February 19, 2013
inclined to use very di!erent cut-o!s, and thus very di!erent bases of observables for doingmeasurements, the Hilbert spaces spanned by these di!erent bases is assumed to be thesame! The consequences of this assumption are particularly striking in a situation withtwo di!erent observers whose reference frames are related by very large red- or blue-shifts,such as the infalling and outside observer on a black hole background.
To illustrate this, let us consider the simultaneous measurement of an outside observ-able O!out
, supported on free field configurations of typical wavelength !out, and an insideobservable O!in
near or just behind the horizon, of typical wavelength !in. (See fig. 1b.)One would expect that such a simultaneous measurement should indeed be possible, sincethe two observables on "final are space-like separated.
! out"
2e
out"
in" cm
in"
t
/4Mt#E ~
#
HorizonEvent
final
Fig 1b. This figure shows two space-like separated observables defined on !final of typical wave-
length !in and !out. The field modes associated with these observables have collided in the past
with a center of mass energy that grows as exp("t/8M). The proposed complementarity prin-
ciple states that observables for which this collision energy exceeds some (possibly macroscopic)
critical value do not simultaneously exist as mutually commuting operators.
To accurately compute transition amplitudes involving these operators, however, wemust consider their past history. In first instance, we can try to compute this pasthistory using the free field propagation of the modes contained in these observables.We then discover that the past history of these observables in fact contains an ultra-high energy collision very close to the horizon, with a center of mass energy that growsexponentially in the out-going time! While it is true that this interaction takes placebetween virtual instead of real particle excitations, the absurd magnitude of the collisionenergy nevertheless indicates that the classical geometry is no longer the appropriatesetting for considering the simultaneous measurement by these two observables.
6
inclined to use very di!erent cut-o!s, and thus very di!erent bases of observables for doingmeasurements, the Hilbert spaces spanned by these di!erent bases is assumed to be thesame! The consequences of this assumption are particularly striking in a situation withtwo di!erent observers whose reference frames are related by very large red- or blue-shifts,such as the infalling and outside observer on a black hole background.
To illustrate this, let us consider the simultaneous measurement of an outside observ-able O!out
, supported on free field configurations of typical wavelength !out, and an insideobservable O!in
near or just behind the horizon, of typical wavelength !in. (See fig. 1b.)One would expect that such a simultaneous measurement should indeed be possible, sincethe two observables on "final are space-like separated.
! out"
2e
out"
in" cm
in"
t
/4Mt#E ~
#
HorizonEvent
final
Fig 1b. This figure shows two space-like separated observables defined on !final of typical wave-
length !in and !out. The field modes associated with these observables have collided in the past
with a center of mass energy that grows as exp("t/8M). The proposed complementarity prin-
ciple states that observables for which this collision energy exceeds some (possibly macroscopic)
critical value do not simultaneously exist as mutually commuting operators.
To accurately compute transition amplitudes involving these operators, however, wemust consider their past history. In first instance, we can try to compute this pasthistory using the free field propagation of the modes contained in these observables.We then discover that the past history of these observables in fact contains an ultra-high energy collision very close to the horizon, with a center of mass energy that growsexponentially in the out-going time! While it is true that this interaction takes placebetween virtual instead of real particle excitations, the absurd magnitude of the collisionenergy nevertheless indicates that the classical geometry is no longer the appropriatesetting for considering the simultaneous measurement by these two observables.
6
Tuesday, February 19, 2013