Spacetime Thermodynamics Without Hidden Degrees of Freedom
Spacetime Thermodynamics Without Hidden Degrees of Freedom
Spacetime Thermodynamics Without Hidden Degrees of Freedom
i
p
i
ln p
i
= ln n (5)
measures our ignorance of the microstate, where the
probability of every microstate is p
i
= 1/n. This is the
standard entropy of statistical mechanics. (In a similar
manner one can construct the canonical version of the
density matrix.)
b. Entanglement entropy. Consider now one of the
subsystems of a larger system formed by two parts, de-
scribed by the Hilbert spaces 1 and 1
, respectively. A
generic pure state of the composite system can be written
in the form
[ =
n,m
c
nm
[n [m
(6)
where [n and [m
n
p
n
[nn[ (7)
where
p
n
=
[c
nm
[
2
. (8)
The von Neumann entropy is
S
vN
=
n
p
n
ln p
n
(9)
and measures the amount of entanglement between the
two systems.
Therefore we see that the von Neumann entropy can
describe two (related but) distinct physical quantities: in
the rst example, it describes our ignorance of the spe-
cic microstate; in the second, it is a measure of entan-
glement. The rst describes simply a translation of the
standard classical entropy of Gibbs and Boltzmann. The
second is a genuine quantum phenomenon, which does
not exist in classical physics.
III. ENTANGLEMENT THERMODYNAMICS
When the von Neumann entropy describes statistical
ignorance, its relation with thermodynamical entropy is
manifest. In fact, the thermodynamics of quantum sys-
tems is described in all textbooks in terms of von Neu-
mann entropy. Is there a relation between, thermal quan-
tities and von Neumann entropy when it describes entan-
glement?
Entanglement entropy is dierent from standard statis-
tical entropy in some respects, but several characteristic
thermodynamical relations remain valid [6]. Let us illus-
trate this fact in a simple case. Consider a system with
two degrees of freedom, respectively described by the
Hilbert spaces 1 and
1. In the tensor product Hilbert
space consider the pure state
[
n
e
2
En
[n [ n. (10)
where is a parameter that we keep free, H[n = E
n
[n
in 1, where H is the hamiltonian, and the basis [ n in
= N
e
H
(11)
where N
[H[
= tr[
H] = H
(12)
and the von Neumann entropy is given by
S() = tr[
ln
]. (13)
Restricted to 1, the state looks thermal even if the
systems as a whole is in a pure state and does not have
many degrees of freedom. Now, consider a small variation
[
of the state [
ln
] = tr[
H] = H. (14)
where the rst equality follows from the identity tr[
] =
0 and the second from the Gibbs form of (11).
3
By using (12) and dening
T
1
, (15)
equation (14) takes the form of a thermodynamical rela-
tion
S = E/T. (16)
Equation (16) is a relation of equilibrium. However, there
is no thermodynamical limit, no large number of degrees
of freedom in the argument above. In the simple context
of a two-degrees of freedom system, T does not have an
interpretation as temperature.
On the other hand, if we have many copies of this cou-
pled system, and a thermometer coupled with them all,
then T can be interpreted as a temperature. Indeed it
determines the raising and lowering transition probabil-
ities between thermometer eigenstates that thermalizes
to the probability distribution
p e
E
(17)
where E is the energy dierence between the thermome-
ter states. This result follows from the fact that (11) is
a KMS state for the time ow, and its proof is recalled
in Appendix A.
The example above might sound articial, since the
one-parameter family of states (10) was introduced ad
hoc. But it is not so articial: in fact, its structure
exemplies that of the Minkowski vacuum [0
M
under
the split of the Fock space into the tensor product of a
Hilbert space capturing the degrees of freedom accessible
from within the horizon of a uniformly accelerating ob-
server, and those outside (Rindler wedge localization) [7].
The Minkowski state is a pure state and when restricted
to the Rindler wedge observables it turns out to be given
by the density matrix
= tr
H
[0
M
0
M
[ = Ne
K
(18)
where K is the boost generator. This is the content of
the Bisognano-Wichmann theorem [8], which can be de-
rived rigorously from quantum eld theory axioms, or
formally with a variety of simple manipulations. It is a
consequence of Lorentz invariance and positivity of the
energy.
The motion of a uniformly accelerated observer is pre-
cisely generated by K. The proper time of an observer
moving at constant acceleration a is given by = /a,
where is the boost parameter of the transformation gen-
erated by K. Therefore once rescaled by a, K generates
proper time translations on the world line of a uniformly
accelerated observer, that is
H = aK (19)
is the generator of the evolution in the proper time in
the Rindler wedge. As a consequence, the density ma-
trix that represents the outcome of measurements on the
spacetime region bounded by the horizon of an acceler-
ated observer in the pure state [0
M
can be rewritten as
= tr
H
[0
M
0
M
[ = N e
H
, (20)
where
T =
1
=
a
2
, (21)
which is the Unruh temperature associated to an observer
accelerating in the Minkowski quantum-eld vacuum [9].
The Unruh temperature is a physical temperature: it
determines the transition probabilities of a system in-
teracting with the quantum eld and moving along the
accelerated trajectory.
This shows that under appropriate conditions the
quantum mechanical entanglement can dene a temper-
ature which is the temperature measured by a physical
thermometer. In particular, this happens when the den-
sity matrix obtained by reducing the state to a subalge-
bra has the form (11) for a hamiltonian H generating a
ow that can be identied with the time evolution of the
thermometer.
Parenthetically we note that the density matrix asso-
ciated to a subsystem can generically be written in the
form
e
H
(22)
for some hermitian operator H. It suces to dene H as
minus the log of . The quantity H dened in this man-
ner is called the entanglement hamiltonian, modular
hamiltonian[7] or thermal time [10, 11] hamiltonian
in various contexts. It generates an evolution in the
Hilbert space 1, and the state is a KMS state with
respect to this ow. In general this ow does not corre-
spond to a ow in spacetime. In the case of the restriction
of vacuum to a Rindler wedge, it does.
The thermodynamical aspects of the setting consid-
ered by Jacobson can be understood in this framework.
There is a quantum eld in its vacuum seen by an accel-
erated observer. The observer interacts only with a sub
algebra of the eld observable algebra, and therefore de-
scribes the state of the eld in terms of a density matrix
(20). The entropy that quanties the incertitude, due to
entanglement, of an observers measurement, and the ex-
pectation value of the generator (19) of his proper-time
translations are related by (14), namely
S = E/T. (23)
The Unruh temperature and the entropy
S =
2
a
E (24)
can be computed directly from the form of the entangled
state . These are relations that pertain to standard
quantum eld theory and do not require any additional
4
hypothesis about underlying degrees of freedom to be
true. They depend on the local Lorentz invariance of the
vacuum and in particular on its local (short scale) conse-
quences. General Relativity plays no role so far. Equa-
tion (24) has nothing to do with the Einstein equations.
In particular, the relation S = E/T knows nothing
about the Newton constant or the Einstein equations.
How do then Einsteins equations emerge in Jacobsons
argument?
IV. THE JACOBSON RESULT
The key ingredient of Jacobsons derivation is the as-
sumption that the entropy density per unit horizon area
is not only nite, but, most importantly, universal. That
is, in the relation
S = A, (25)
it is assumed that is nite and does not depend on any
detail of the physical situation. Here, A is the area of the
2-dimensional spacelike bifurcation surface of the causal
horizon comprising the wedge.
This universality is a strong assumption that falsely
may appear to be innocent. The fact that the entropy
across a region is proportional to the area is common
when correlations are suciently short scale compared
to the geometry of the surface. However, in general the
proportionality constant will depend on the system and
on the state. Jacobson assumes it does not.
The entropy density of the density matrix (18) is in-
nite, because of the contribution of arbitrarily short
wavelength correlations across the Rindler horizon. If
we cut o the theory at some length
p
, we obtain an
entropy that scales as 1/
p
2
[1215], that is
S =
c
2
p
A (26)
or =
c
2
p
. If we change the state, the area does not
change (we are on a xed geometry) but the entropy does
change, therefore
S = (A) = () A. (27)
If one instead assumes, as Jacobson does, that the pro-
portionality constant between entropy and area is uni-
versal, then one either gets a contradiction (because one
can change the entropy by changing the state), or one
has to relax the fact that the geometry is xed and allow
it to change with a change in the state, thus allowing the
area to vary
S = A. (28)
As we have seen above the entanglement entropy across
a surface satises the relation S = E/T, where E is an
energy. Therefore we obtain
E = TA, (29)
and as a consequence if the constant is to be universal,
area and energy variations must be related. Inserting
the value of T from equation (24) into (29) leads to the
fundamental relation
E =
a
2
A. (30)
Notice that this relation has nothing thermodynamical
in it. It is a relation between energy and geometry. This
relation was derived by Frodden, Gosh and Perez [19] in a
dierent context as a consequence of the Einstein equa-
tions (see also [20]). The Frodden-Gosh-Perez relation
reads
E =
a
8G
A. (31)
and agrees with (30) identifying the universal constant
as = 1/4G.
Thus, what Jacobson has shown in his derivation
(which we recall in Appendix C) is that equation (31) is
not only a consequence of Einsteins equations, but is also
a sucient condition for the Einstein equations to hold.
If we assume the validity of (31) in any frame, then the
Einstein equations follow. In this sense, Einsteins equa-
tions are encoded in the proportionality between classical
variation of energy and horizon area, as measured by a
uniformly accelerating observer.
By itself, the fact that the Einstein equations can be
encoded in a single simple equation relating an energy
and a geometrical quantity is well known. Of course the
trick is requiring that this holds everywhere and in any
frame, thus adding general covariance as an independent
input. For instance, see the nice paper by John Baez and
Emory Bunn [21] where it is pointed out that Einsteins
equations follow entirely from the request that the second
variation of the volume of a small sphere of particles is
proportional to the sum of the energy and momentum
ow components in the centre of the sphere. Jacobsons
is an elegant and null version of Baez and Bunns timelike
result.
The important point of this discussion is that Jacob-
sons result can be split into two parts. The second part
being the step from (30) to the Einstein equations. This
is a nice piece of dierential geometry, but has nothing
to do with thermodynamics. The rst step is to get to
equation (30). As we have seen, this equation comes from
considering the entanglement entropy across the horizon
plus a universality assumption. Again, here statistical
considerations play no role, and nothing points to un-
derlying degrees of freedom, or to a reading of the Ein-
stein equations as equations of state. The key is simply
the universality of the entanglement entropy-density. We
comment on this below.
V. UNIVERSALITY: THE SPECIES PROBLEM
The interpretation of the Bekenstein entropy as entan-
glement entropy has been called into question precisely
5
due to the unclear dependence of the entanglement en-
tropy on the number of species at the cuto scale. How
can entanglement entropy not be dependent on the num-
ber of species?
A recent result [4] has shown that for a low energy
perturbation of the Minkowski vacuum, the variation of
entanglement entropy S across the Rindler horizon is
independent from the number of matter species and in-
dependent from the ultraviolet cut-o, provided one in-
cludes gravitons together with matter elds, in a per-
turbative approach. What happens is that the change
in the energy of the vacuum E, expressed in terms of
the energy-momentum of matter and gravitons T
, is re-
lated to the dAlambertian of the graviton eld by means
of the perturbative Einstein equations,
h
8G(T
1
2
), (32)
from which
E =
a
8G
A (33)
follows. The change in the area of the causal horizon has
now a semiclassical interpretation. Classical background
geometry is not aected, and the perturbation of the area
is determined by the quantum gravitational eld h
and
given by the expectation value of the operator
A associ-
ated to the rst order variation of the area with respect
to its background value, as a function of h
. That is
A =
A tr[
A].
This perturbation controls the deection of light rays
related to the expansion of the area of the horizon surface,
so that equations (16), (21) and (33) together provide a
dynamical Bekenstein-Hawking area law relation,
S =
A
4G
, (34)
which relies on the entanglement of the gravitationally
dressed vacuum elds.
Importantly, equation (34) does not depend on the
physics at the UV scale, as S = E/T involves only
dierences in the low energy modes, while its universal
character has a dynamical origin: it derives from the
universal coupling of gravitons to the energy momentum
tensor.
At the perturbative level, this result provides a con-
nection between Bekenstein-Hawking entropy and entan-
glement entropy and indicates that the relation between
area and energy variations in (33) can only have a dy-
namical origin: it requires the Einstein equations.
This connection between Bekenstein-Hawking entropy
and entanglement entropy is not in terms of statistical
ignorance about mysterious underlying microscopic de-
grees of freedom, it does not imply any thermal char-
acterization of the Einstein equations.
The universality of is the eect of an active role of
gravitational dynamics at the quantum level. One can
gain an intuitive understanding of the fact that gravity
cuts the degrees of freedom o in a universal manner as
follows. The quantum gravitational regime starts when
the total energy density reaches the Planck scale. Matter
uctuations interact gravitationally, so in the presence of
more than one eld, say n elds, this regime is reached
at a larger length scale than in the presence of a sin-
gle eld. More precisely, if we introduce a cuto at the
scale , on dimensional grounds the vacuum energy den-
sity of the uctuations of a single eld is proportional to
4
. Within an
3
volume, the (negative) gravitational
potential energy of these uctuations goes as G
3
and is
proportional to n
2
because each eld interacts with any
other eld. We enter in a strong gravitational eld regime
at the scale where these two energies balance, namely at
the length scale
G, but for
n elds the cuto scale is reached at a larger scale by
a factor
n, that is, the cuto scale depends on the
number of elds as
n. (35)
The entanglement entropy of n elds is
S n
A
2
(36)
where is the cuto scale. The last two equations to-
gether indicate that S does not depend on the number of
elds [22]. It is the universality of gravity, not statistics,
that determines the entanglement entropy universality of
the entropy density.
So far, we have stayed away from the Planck scale. It
is time to ask what may happen there and complete the
picture.
VI. FINITENESS: THE JACOBSON RESULT
FROM QUANTUM GRAVITY
At the beginning of the XX century, Planck and Ein-
stein were led to the existence of a fundamental discrete-
ness in the structure of the electromagnetic eld, and in
fact to the discovery of the existence of the photon, from
the absence of the ultraviolet catastrophe, namely from
the niteness of the entropy of the electromagnetic eld.
The entropy of black body radiation, indeed, is given by
S =
4
3
4
2
V U
3
15c
3
3
, (37)
where V is the volume of the box and U the energy,
and this diverges if we take 0. Quantum theory
is the source of the Planck-size granularity that renders
phase space volumes nite, and therefore avoids the di-
vergence of the entropy. In the entropy (34), associated
to a gravitational eld, is in the denominator as well.
This indicates that it is quantum mechanics that cuts o
6
the divergence of the entropy. In fact, the formula in-
dicates that there should be a physical (real) quantum
mechanical cuto at the Planck scale.
This is precisely the conclusion that one obtains from
a full quantum gravitational treatment of the problem
in the context of loop quantum gravity, where the dis-
creteness of the geometry at the Planck scale appears
as a conventional quantization eect: the gravitational
eld determines lengths, areas and volumes; since the
gravitational eld is a quantum operator, these quanti-
ties are given by quantum operators. Planck scale dis-
creteness follows from the spectral analysis of such op-
erators, which indicates minimum non-vanishing sizes at
the Planck scale. This is the key result of the loop the-
ory and is responsible for the UV niteness of the entropy
density.
Let us see how Jacobsons result emerges from this
framework. For this, we recall some basic notions from
the full theory, relevant to the present discussion [23, 24].
In loop gravity, quantum states of the geometry are
described by SU(2) spin networks. Let us consider here
for simplicity a single link of the spin network. The cor-
responding quantum state is a function (U) on SU(2),
and U is classically interpreted as the open path holon-
omy of the gravitational Ashtekar connection along the
link. A basis of states is provided by the Wigner matrices
(Peter-Weyl theorem)
U[j, m, n = D
(j)
mn
(U). (38)
Equivalently, the state space 1 associated to the link
decomposes as
1 =
j
(1
j
1
j
) (39)
where 1
j
is the spin j irreducible representation of
SU(2). The two factors are associated to the two ends
of the links, respectively, as each transforms under the
gauge transformation at one end. The dynamics of the
theory is obtained mapping these states to unitary repre-
sentations of SL(2, C). A unitary representation (in the
principal series) [25] is labelled by a discrete spin k
1
2
N
and a continuous parameter p R
+
and the represen-
tation space is denoted 1
(p,k)
. This space decomposes
into irreducible representations of the SU(2) subgroup
as follows
1
(p,k)
=
j=k
1
j
(40)
where 1
j
is the (nite dimensional) SU(2) representation
of spin j. Therefore 1
(p,k)
admits a basis [(p, k); j, m
obtained diagonalizing the total angular momentum L
2
and the L
z
=
L z component of the SU(2) subgroup.
The map that gives this injection, and denes the loop
quantum gravity covariant dynamics is given by
Y
: 1
j
1
j
(41)
[j, m [(j, j); j, m.
Here R
+
is the Immirzi parameter.
6
FIG. 1. Indicative picture, to be changed. basic system: sin-
gle link with its dual facet.
topological eld theory, and denes quantum general rel-
ativity in the Spinfoam approach []
3
.
Condition (30) denes a subspace of H
p,k
, spanned by
the subspaces H
j
p,k
where
p = j, k = j; (31)
that is, the space
H
=
j
(H
j
j,j
)
H
j
j,j
. (32)
of SL(2, C) functions of the form
|i =
X
j,m,m
0
c
j,m,m
0 |j, j; (j, m), (j, m
0
)i. (33)
Among the general quantum geometrical states, the
states in (33) are the ones dynamically relevant for the
quantum description of general relativity. The space H
: V
(j)
V
((j),j)
(34)
|j, mi |(j, j); j, mi
sends L
2
[SU(2)] to H
sends (h
l
) to a generalised func-
tion of Y
(g
l
) of SL(2, C) elements. The associated
states |i dened on the single-link Hilbert space
H
j
= (H
j
j,j
)
H
j
j,j
, (38)
satisfy (30).
The tensor product decomposition of the single-link
Hilbert space is related to a local partition of space
in two chunks, associated to the source and target nodes
comprising the link. In this sense, we wish to redene
the single-link Hilbert space as
H
j
= (H
j
j,j
)
S
(H
j
j,j
)
T
= H
S
H
T
(39)
where we think about the individual subspaces as dened
by the partition of the link in two half links l
S
, l
T
, given
by the puncture on the dual facet surface. In fact, one
can think of the facet
4
associated to the link as the
boundary surface between the two space chunks.
Now, let |i H
j
T
= tr
H
S
[|ih|]. (40)
This state encodes all the information that can be ob-
tained about |i by performing measurements localised
in the target chunk of space associated to l
T
.
4
This surface is dened by ~n, once the embedding SU(2)
t
SL(2, C) is xed by the choice of t
.
FIG. 1. Single link and its dual facet.
The two spin j representations of Eq. (39) are each
mapped in this way, therefore the spin j component of
the state space lives in a subspace of
(1
j
)
T
(1
j
)
S
(42)
where we have indicated by o and T the source and tar-
get ends of the link, respectively. On the image of the
map Y
K =
L (43)
as matrix elements.
In the classical theory, equation (43) is satised by the
conjugate momentum of the Lorentz connection, which
becomes the SL(2, C) generator in the quantum theory
(see Appendix B). This relation directly parallels (31)
because (from the canonical analysis of the action) the
area-vector is related to
L by
A = 8G
L and
K is
related to the energy by (19).
The SU(2) spin-network states represent the geome-
try of a spatial section of spacetime with normal t
. If
we measure the geometry of a spacelike surface with a
certain precision, we can describe the result of our mea-
surements by tessellating with N facets f. Each facet is
going to be described by a spin network link, which we
can visualise as puncturing the surface [26], see Fig. 1.
Call [s
f
the state describing a facet of a quantum sur-
face of a given area and with normal in direction n. This
state has maximum magnetic number in direction n. In
writing it, for notational simplicity, one can restrict to
n = z, then
[s
f
= Y
to
the unitary irreducible SL(2, C) representation where the
7
relation (43) holds. For a single puncture, the relation
K
f
=
L
f
then gives
K
z
f
=
A
z
f
8G
(46)
where K
z
f
is the boost dual to the facet surface. This
relation gives immediately equation (31), which is the
basis of the second part of Jacobsons argument. From
this relation we can obtain the Einstein equations. Notice
that this is a relation between the area of a space like
surface (say in the x-y plane) and the boost hamiltonian
(in the dual t-z plane). As rst observed by Smolin [27],
this is a direct way of deriving the Einstein equations
from the covariant loop quantum dynamics.
Next, observe that the tensor product structure in (42)
is related to a local partition of space into two chunks,
associated to the source and target nodes at the two
ends of the link. In fact, the gravitational eld opera-
tors, which are given by the left invariant and right in-
variant vector elds act respectively on the source and
target end, and correspond to measurements performed
on one side or the other of the facet [23]. It therefore
makes sense to dene a half-link Hilbert space associated
to each end of the link. The short-scale correlations of
the theory are then captured by the correlations between
these two Hilbert spaces, namely by the failure of the
link states to be diagonal in this tensor decomposition.
If [ (1
j
)
(1
j
) is a pure state of the link, then its
restriction to one half-link is the density matrix
T
= tr
H
S
[[[], (47)
that encodes the results of the measurements performed
on the target side only of the facet (with an analogous
relation holding for the other half of the link). The von
Neumann entropy of this density matrix measures the
entanglement across the facet.
Let us now consider a particular family of states [
0
f
= e
2K
f
. (48)
where K
f
=
K
f
n is the boost generator in the direction
normal to the facet (the notation
f
=
S
=
T
indicates
the symmetry of the reduced density matrix). We call
these states Hadamard states for a reason that will be
clear below.
These states have interesting properties. They have
von Neumann entropy
S = tr[
f
ln
f
] = 2 tr[
f
K
f
] = 2 K
f
, (49)
where K
f
=
0
[K
f
[
0
is the expectation value of K
f
on [
0
. As shown in (24), a variation relates a change
in the entropy to a change in the expectation value of K
f
,
namely
S = 2 K
f
. (50)
Therefore, given K
f
= L
f
, one nds
S =
A
f
4G
, (51)
that is the area-entropy relation (34). We see here that
the niteness of the entropy density relies on the discrete-
ness of the geometry at that scale and, in particular, on
the niteness of the spectrum of the area operator.
Equation (31) is a classical relation while (46) is as-
sociated to a quantum mechanical system. To clarify
the relation between the two, we need to recall some ba-
sic structure in quantum theory. This always refers to
results of a measurement performed by an external sys-
tem. Here the quantum system is a portion of quantum
spacetime. The external observer is an accelerated ob-
server which can measure local aspects of the geometry.
The states [s
f
describe a portion of a quantum surface,
a small facet with given area A
f
and normal in direction
n. The evolution of the facet state in spacetime as seen
by an accelerated observer is generated by the unitary
operator representing a Lorentz boost
[s
f
= U()[s
f
(52)
where U() = e
i
H
, where the operator H is
H = K
f
a (53)
with a being the acceleration of the observer. This op-
erator generates the evolution of the state along a spe-
cic isometry trajectory of the boost labelled by a. This
trajectory represents the world line of an observer who
moves in Minkowski space with uniform acceleration a.
In this sense, a classical measurement of the facet state
dynamics is associated to the covariant dynamics of the
trajectory of the accelerated observer. This can be seen
either passively as a motion of an observer in spacetime
or actively as the evolution of spacetime seen by an ob-
server.
Written in terms of the accelerated observers quanti-
ties, the density matrix in (48) reads
f
= N e
2
a
H
(54)
and the expectation value of the Hamiltonian operator
H gives
E =
A
f
a
8G
(55)
where E = H has now the dimensions of an energy.
This is exactly relation (31), the core physical input from
which the Einstein equations, as shown by Jacobson, fol-
low in the classical limit.
To the ensemble of single facet states given by (54),
the observer can eectively associate an absolute tem-
perature, via the general denition
T =
E
S
=
a
2
. (56)
8
because if it interacts with a large number of these, this is
the temperature determining the transition probabilities
between its eigenstates. This is the Unruh temperature.
1
Therefore all the ingredients for Jacobsons derivation
follow (see also [27]).
We have illustrated how geometrical discreteness arises
in the formalism of loop gravity and shown that this dis-
creteness gives rise to all of Jacobsons inputs microscop-
ically. This line of thinking has exposed the possibility
that local Lorentz invariance is encoded directly in the
quantum correlations at the Planck scale and led us to
introduce quantum gravitational Hadamard states. We
further develop this idea below.
VII. THE ARCHITECTURE OF
SEMICLASSICAL SPACETIME
In covariant loop quantum gravity [23] the dynamics
is captured by a relation between the area of a spacelike
surface (say in the x-y plane) and the boost (in the t-z
plane). This is beautifully made explicit by the result
of Jacobson, which shows that (31) yields the Einstein
equations.
The dynamical input that gives the Einstein equations
in Jacobsons derivation is not the thermodynamics. It
is the relation between the boost generator and the area.
This is the input that contains the Newton constant
G and gives the dynamics. Then, the presence of the
horizon leads to the entanglement that gives rise to the
entropy, and the relation between this entropy and the
boost generator determines the KMS temperature.
In the case of a black hole, as shown in [19], the en-
tropy at innity, the energy at innity (ADM mass) and
the temperature at innity (the Hawking temperature)
are simply obtained by red shifting the local quantities
to innity. Their relation is precisely the local Clau-
sius relation redshifted to innity. This is unproblematic
standard physics (much like the Tolman law) and is well
understood.
These considerations do not put into question the tight
constraints of reciprocal consistency that gravitational
dynamics and thermodynamics of the vacuum state im-
pose on each other; a consistency that Jacobsons deriva-
tion sheds light on. It is quite remarkable, indeed, that
so much of the gravitational dynamics is constrained by
consistency with the thermodynamics of quantum elds.
But thermodynamics does not and cannot give the full
gravitational equations, as is clear for instance from the
fact that gravitational theories dierent from general rel-
ativity are also compatible with the eld thermodynam-
ics [28].
1
Notice also that for a black hole the Hawking temperature at
innity is just the red shifted version of the near horizon tem-
perature, and the asymptotic ADM energy is just the red shifted
local energy. Therefore the standard black hole thermodynamics
follows.
What is the physical meaning of the Hadamard-like
states introduced in Sec. VI? These quantum states de-
termine all the relations necessary for Jacobsons calcu-
lation. They are analogous to those obtained restricting
the Minkowsky vacuum to the Rindler wedge. Like those
states, they precisely describe the short scale correlations
and the eect of local Lorentz invariance and the positiv-
ity of the energy. Formally, a pure state on the link that
gives rise to the density matrix (48) is simply built by
having the source component of the state related to the
target one by a Lorentz transformation with imaginary
parameter i
[j, n
S
= e
j
j, n[
T
. (57)
In turn, this form of the state is formally determined
by requiring a global Lorentz invariance and positivity
of energy that allows for an analytic continuation [8].
Thus, these states can be viewed as implementing the
local properties that render a semiclassical geometry lo-
cally Lorentz invariant. If every link of a spin network
carries the correlations of these states, then the corre-
sponding quantum state has everywhere the short scale
correlations of a conventional quantum eld theory state.
These are therefore the states that may build up the
architecture of a classical spacetime geometry [5]. In
quantum eld theory on curved spacetime attention is
restricted to states having an appropriate short scale be-
havior, which is determined by it being the same as that
of a at Lorentz invariant quantum eld theory. States
with this local property are called Hadamard states. This
is why, by analogy, we call the states of Sec. VI Hadamard
states: they behave like quantum eld states in the small.
Notice that the relation here is not at the innitesimal
level, but at the nite Planck-scale, namely across each
single link of a spin network. Quantum gravity is a theory
without innities.
VIII. CONCLUSION
Jacobsons derivation of the Einstein equations admits
an interpretation dierent from the common one where it
is assumed to indicate the existence of microscopic states
for which the Einstein equations would be an equation of
state. The alternative interpretation is based on the iden-
tication of the relevant dynamics simply as the quantum
version of Einsteins dynamics.
The relevant microscopic degrees of freedom are the
quanta of the gravitational eld. These are discrete at
the Planck scale, and this is the source of the niteness
of the entropy. The interaction of any system with the
gravitational eld in a region of spacetime is described
by the eld operators of the quantum theory. As for the
electromagnetic eld, some of the quantum states of the
gravitational eld have a semiclassical behavior. These
appear as continuous geometries when probed at large
scales: the mean value of the eld operators determines
9
the value of the geometrical observables. The approxi-
mation holds only in the semiclassical limit, namely at
scales larger than the Planck scale. These states present
short scale correlations, like all the nite energy states
of conventional quantum eld theory. But these correla-
tions do not involve modes of arbitrarily high frequency,
because of the Planck scale quantum discreteness.
Because of the universality of the gravitational cou-
pling, and thanks to a mechanism that can be intuitively
understood as explained in Section V, the entropy den-
sity measuring these correlations is universal. An exter-
nal observer accelerating in the semiclassical geometry
interacts only with observables on one side of its causal
horizon and can therefore associate an entropy density to
the horizon, which is nite and universal. The dynamics
ties any change in the area with an energy ow, there-
fore, because of the universality, the change in the area is
tied by the dynamics to a change in entropy, with a uni-
versal temperature T. This temperature is determined
by the entropy-energy relation and therefore, ultimately,
simply by the local Lorentz invariance and the positiv-
ity of the energy of the local states showing short scale
correlations. In spite of the fact that the source of the
entropy is entanglement, this temperature behaves as a
true temperature because the density matrix is a KMS
state with respect to the ow that describes the evolu-
tion of an observer accelerating in the mean geometry,
and because this observer interacts with a large number
of degrees of freedom, all in this KMS state. It follows
from this that T is detected by a thermometer moving
with the observer.
The relation (31) between area and energy is deter-
mined by the quantum dynamics of general relativity:
it is true both in the quantum and the classical theory.
The universal relation between area and entropy is de-
termined by the short scale structure of the semiclassical
states, their local Lorentz invariance and the positivity
of the energy, as in the Bisognano-Wichmann theorem.
In the deep ultraviolet this picture is conrmed by loop
quantum gravity, where all the input formulas of Jacob-
sons derivation can be recovered starting from the ele-
mentary quantum dynamics of the theory, for appropri-
ate states that have semiclassical properties, which we
have called Hadamard states.
The entropy in Jacobsons calculation is therefore en-
tanglement entropy across the horizon. It is a property
of a pure state, when restricted to the sub algebra of ob-
servables lying on one side of the causal horizon. It is not
related to any mysterious underlying degrees of freedom.
The entropy balance relation does not imply statistical
uncertainty and in general holds also for entanglement
entropy.
What Jacobson has shown is (i) that the area-energy
relation (31), combined with general covariance, deter-
mines the Einstein equations, a result in line with simi-
lar previous derivations of the Einstein equations from a
single local equation, e.g. [21]; and (ii) that this dynam-
ics not only implies the universality of the entropy/area
ratio, but is also implied by it.
This does not diminish the relevance of Jacobsons re-
sult. To the contrary, it puts it in an even more inter-
esting light. Jacobsons remarkable discovery is not that
there are mysterious microstates beyond the quantized
gravitational eld. It is that the dynamics of gravity
and the Einstein equations can be recovered from a sim-
ple relation between boost generator and area of its dual
surface, and these quantities are related to the entangle-
ment entropy in semiclassical states. This is a step ahead
in unraveling the spectacular beauty and the simplicity
of general relativity, not an indication of its limits.
f
AB
() = e
f
AB
(). (A4)
The canonical example is provided by the case where
/ is realized by operators on a Hilbert space 1 where a
Hamiltonian operator H with eigenstates [n and eigen-
values E
n
is dened, the ow and the state are dened
in terms the evolution operator U(t) = e
iHt
and the
density matrix = e
H
=
n
e
En
[nn[, by
t
(A) = U
1
(t)AU(t) (A5)
and
(A) = tr[A] (A6)
respectively. In this case, it is straightforward to verify
that equation (A1) and (A2) follow.
10
However, the scope of equations (A1) and (A2) is wider
than this canonical case. For instance, these equations
permit the treatment of thermal quantum eld theory,
where the Hamiltonian is ill-dened, because of the in-
nite energy of a thermal state in an innite space. It
is indeed important to remark that these two equations
capture immediately the physical notions of equilibrium
and temperature. For equation (A1), this is pretty ob-
vious: the state is in equilibrium with respect to a ow
of time if the expectation value of any observable is time
independent.
The direct physical interpretation of equation (A2) is
less evident: it describes the coupling of the system with
a thermometer.
To see this, consider a simple thermometer formed by
a two-state system with an energy gap , coupled to a
quantum system S by the interaction term
V = g([01[ +[10[)A. (A7)
where g is a small coupling constant. The amplitude for
the thermometer to jump up from the initial state [0
to the nal state [1, while the system moves from an
initial state [i to a nal state [f can be computed using
Fermis golden rule to rst order in g:
W
+
(t) = g
dt (1[ f[)
t
(V ) ([0 +[i)
= g
dt e
it
f[
t
(A)[i. (A8)
The probability for the thermometer to jump up is the
modulus square of the amplitude, summed over the nal
state. This is
P
+
(t) = g
2
dt
1
dt
2
e
i(t1t2)
(
t2
(A
)
t1
(A))
where we have used the algebraic notation (A) =
i[A[i. If the initial state is an equilibrium state
P
+
(t) = g
2
dt
1
dt
2
e
i(t1t2)
f
A
A
(t
1
t
2
).
and the integrand depends only on the dierence of the
times. The transition probability per unit time is then
p
+
=
dP
+
dt
= g
2
f
A
A
() (A9)
which shows that (A3) is precisely the quantity giving
the transition rate for a thermometer coupled to the sys-
tem. It is immediate to repeat the calculation for the
probability to jump down, which gives
p
= g
2
f
AA
(). (A10)
And therefore (A4) expresses precisely the fact that
the thermometer thermalizes at temperature , that is
p
+
/p
= e
(e e) F +
1
e e F
.
(B1)
where e is the (co)tetrad one-form, F the curvature of
the Lorentz connection , the star indicates the hodge
dual on internal indices and a contraction on the inter-
nal indices is implicit. On a spacelike boundary , the
quantity
=
1
8G
(e e +
1
e e) (B2)
is the derivative of the action with respect to /t, and
therefore is the momentum conjugate to the connection.
The (co)tetrad e can be chosen to (locally) map into
a spacelike 3d linear subspace of Minkowski space. The
subgroup of the Lorentz group that leaves this subspace
invariant is the SO(3) rotations subgroup, and its exis-
tence breaks the local SO(3, 1) invariance down to SO(3)
11
at the boundary. That is, the boundary allows us to pick
up a preferred Lorentz frame.
In coordinates, we can see this by writing the unit
length vector n
I
normal to all vectors tangents to , that
is
n
I
IJKL
e
J
e
K
e
L
1
x
2
x
3
(B3)
where x
IJ
and mag-
netic L
I
= n
J
()
IJ
parts, in the same manner in which
the electromagnetic tensor F
IJ
can be decomposed into
electric and magnetic parts once a Lorentz frame is cho-
sen. Since is antisymmetric, L
I
and K
I
do not have
components normal to , that is n
I
K
I
= n
I
L
I
= 0 and
are therefore three dimensional vectors in the space nor-
mal to n, which we can denote
K and
L. In the gauge
n
I
= (n, 0, 0, 0) these are simply
K
i
=
i0
, L
i
=
1
2
i
jk
jk
. (B4)
where we write
K = K
i
and
L = L
i
, with i = 1, 2, 3,
which are relations analogous to the ones that relate elec-
tric and magnetic elds to the Maxwell tensor. From the
denitions, we have
n
I
e
I
= 0 , (B5)
therefore at the boundary
n
I
()
IJ
=
1
n
I
IJ
, (B6)
which also reads
K =
L. (B7)
This is called the linear simplicity relation [31]. Decom-
pose
K = Kn and, because
L
i
=
1
8G
n
J
(e e)
iJ
(B8)
it follows
L =
1
8G
An, (B9)
where A is the area element for a surface element normal
to the direction n of the boost. Therefore
K =
A
8G
. (B10)
The boost generator K is related to the area element. An
observer with acceleration a moves along a ow generated
by H = aK, therefore his energy is
E = aK =
aA
8G
. (B11)
which is the Frodden-Gosh-Perez relation.
Appendix C: Jacobson derivation
We recall here the main steps of the second part of Ja-
cobsons derivation: from equation (31) to the Einsteins
equations. The change of the horizon area with the ex-
pansion of the null geodesics comprising it is
A =
H
d, (C1)
where is the 2-surface area element of the horizon cross-
section, s the ane parameter along the null geodesics
and is the usual expansion. For a small horizon per-
turbation, the innitesimal evolution of is given by a
linear expansion around its equilibrium value at p, up to
the rst order in ,
p
+
d
d
p
+O(
2
) . (C2)
At equilibrium, the 0th order contribution to the area
change
p
vanishes. The rst order contribution is given
by the Raychaudhuri equation,
d
d
=
1
2
2
||
2
R
, (C3)
where ||
2
stands for the squared congruence shear
and
H
(R
)
p
d. (C4)
The mean energy variation of the thermal system,
given by the boosted energy current ux of matter, can
be described in terms of heat, as it is perfectly thermal-
ized by the horizon. Thereby, one can dene the heat
ux across the horizon as
E =
H
T
, (C5)
where T
is the
local horizon approximate Killing vector, while the vol-
ume element is given by d
H
d () T
. (C6)
At this point, asking for relation (16) to hold for all
null vectors
= R
(C7)
12
and thereby, in any frame,
2
= R
+ g
, (C8)
where is an undetermined integration function.
Eventually, by assuming the local energy conservation,
that is
=
1
2
= R
1
2
Rg
, (C9)
where is some arbitrary integration constant. Once the
condition
=
1
4G
(C10)
is imposed, (C9) gives the Einstein equations. The equiv-
alence principle implies that the above construction can
be done at any spacetime point p, hence equation (C9)
holds in any point of a pseudo-riemannian spacetime [1].
[1] T. Jacobson, Thermodynamics of space-time: The Ein-
stein equation of state, Phys. Rev. Lett. 75 (1995) 1260
1263, arXiv:9504004 [gr-qc].
[2] T. Padmanabhan, Thermodynamical Aspects of Grav-
ity: New insights, Rept. Prog. Phys. 73 (2009) 6901,
arXiv:0911.5004 [gr-qc].
[3] E. Verlinde, On the origin of gravity and the laws of
Newton, Journal of High Energy Physics (2011) 29,
arXiv:1001.0785 [hep-th].
[4] E. Bianchi, Horizon entanglement entropy and univer-
sality of the graviton coupling, arXiv:1211.0522.
[5] E. Bianchi and R. C. Myers, On the Architecture of
Spacetime Geometry, arXiv:1212.5183 [hep-th].
[6] J. M. Deutsch, H. Li, and A. Sharma, Microscopic origin
of thermodynamic entropy in isolated systems, Phys.
Rev. E 87 (2013) 042135, arXiv:1202.2403 [quant-ph].
[7] R. Haag, Local Quantum Physics: Fields, Particles, Al-
gebras. Springer, Aug., 1996.
[8] J. J. Bisognano and E. H. Wichmann, On the Dual-
ity Condition for Quantum Fields, J. Math. Phys. 17
(1976) 303321.
[9] W. G. Unruh, Notes on black hole evaporation, Phys.
Rev. D14 (1976) 870.
[10] C. Rovelli, Statistical mechanics of gravity and the ther-
modynamical origin of time, Class. Quant. Grav. 10
(1993) 15491566.
[11] A. Connes and C. Rovelli, Von Neumann algebra au-
tomorphisms and time thermodynamics relation in gen-
eral covariant quantum theories, Class. Quant. Grav. 11
(1994) 28992918, arXiv:9406019 [gr-qc].
[12] R. D. Sorkin, On the Entropy of the Vacuum Outside
a Horizon, General Relativity and Gravitation 1 (1983)
734.
[13] L. Bombelli, R. K. Koul, J. Lee, and R. D. Sorkin, Quan-
tum source of entropy for black holes, Phys. Rev. D 34
(1986) 373383.
[14] M. Srednicki, Entropy and Area, Phys. Rev. Lett. 71
(1993) 666669, arXiv:hep-th/9303048.
[15] S. N. Solodukhin, Entanglement entropy of black holes,
Living Rev. Relativity 14 (2011) 8, arXiv:1104.3712
[hep-th].
[16] T. Jacobson, R. Parentani, Horizon Entropy, Found.
Phys. 33 (2003) 323-348, arXiv:gr-qc/0302099.
[17] C. Eling, R. Guedens, T. Jacobson, Non-equilibrium
Thermodynamics of Spacetime, Phys. Rev. Lett. 96
(2006) 121301, arXiv:gr-qc/0602001.
[18] G. Chirco, S. Liberati, Non-equilibrium Thermodynam-
ics of Spacetime: the Role of Gravitational Dissipa-
tion, Phys. Rev. D 81 (2010) 024016, arXiv:0909.4194
[gr-qc].
[19] E. Frodden, A. Ghosh, and A. Perez, A local rst law
for black hole thermodynamics, arXiv:1110.4055.
[20] R. Wald, Black Hole Entropy is Noether Charge, Phys.
Rev. D 48 (1993) 3427-3431, arXiv:gr-qc/9307038.
[21] J. C. Baez and E. F. Bunn, The Meaning of Ein-
steins equation, Am. J. Phys. 73 (2005) 644652,
arXiv:0103044 [gr-qc].
[22] E. Bianchi, Private communication, 2013.
[23] C. Rovelli, Zakopane lectures on loop gravity, PoS
QGQGS2011 (2011) 3, arXiv:1102.3660.
[24] C. Rovelli and F. Vidotto, Introduction to covariant loop
quantum gravity. Cambridge University Press, to appear.
[25] W. Ruhl, The Lorentz group and harmonic analysis.
W.A. Benjamin, Inc, New York, 1970.
[26] E. Bianchi, Entropy of Non-Extremal Black Holes from
Loop Gravity, arXiv:1204.5122.
[27] L. Smolin, General relativity as the equation of state of
spin foam,. arXiv:1205.5529 [gr-qc].
[28] T. Jacobson, Gravitation and vacuum entanglement
entropy, Int. J. Mod. Phys. D21 (2012) 1242006,
arXiv:1204.6349.
[29] R. Hojman, C. Mukku, and W. A. Sayed, Parity vio-
lation in metric torsion theories of gravity, Phys. Rev.
D22 (1980) 19151921.
[30] S. Holst, Barberos Hamiltonian derived from a gener-
alized Hilbert-Palatini action, Phys. Rev. D 53 (1996)
59665969, arXiv:9511026 [gr-qc].
[31] J. Engle, E. Livine, R. Pereira, and C. Rovelli, LQG
vertex with nite Immirzi parameter, Nucl. Phys. B799
(2008) 136149, arXiv:0711.0146.