Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
9/6/11
1
Thanks to the organizers for bringing us together!
9/6/11
2
Horizon entropy and higher curvature equa?ons of state
Ted Jacobson University of Maryland
Plan of talk:
Horizon entropy & Einstein Equa?on of State
Higher curvature
Noetheresque approach
Based on:
TJ, ‘95 paper
Chris Eling, Raf Guedens, TJ, ’06 Non-‐equilibrium paper
Raf Guedens, TJ, Sudipta Sarkar ‘11 Noetheresque paper
9/6/11
3
Credo:
Black hole thermodynamics is a special case of causal horizon thermodynamics.
It originates from thermality of the Lorentz-‐invariant local vacuum in local Rindler wedge, at temperature rela?ve to the local boost Hamiltonian.
This thermal state has a huge entropy, somehow rendered finite by quantum gravity effects, that scales, for any local causal horizon (LCH) at leading order, like the area.
Horizon evolu?on can be viewed as a small perturba?on of a local equilibrium state, so this entropy sa?sfies the Clausius rela?on
When applied to all LCH’s, together with conserva?on of ma\er energy-‐momentum, this implies the Einstein equa?on.
TB = �/2π
δS =δQB
TB
Euclidean space Minkowski space
ds2 = dx2 + dy2 = dr2 + r2dθ2 ds2 = dt2 − dx2 = l2dη2 − dl2
RotaAon symmetry Lorentz boost symmetry
9/6/11
4
€
Lorentz invariance and energy positivity imply the Minkowsi vacuum is a thermal state when restricted to the wedge:
ρR = TrL 0 0 ∝ exp − 2πHBoost
⎛
⎝ ⎜
⎞
⎠ ⎟
Bisognano-Wichmann (1975), Davies (1975), Unruh (1976)
A uniformly accelerated observer a distance l from the horizon sees the temperature, Tlocal = a /2π = /2π l.
Davies-‐Unruh effect
L R
Accelera?on and Tlocal diverge as l goes to 0.
Vacuum entanglement entropy
S = −Tr(ρR ln ρR) ≈�
dA dl T 3local ∝
�dA dl l−3
S = α A
Hypothesis: physics regulates the divergence:
9/6/11
5
Horizon thermodynamics
€
S = α A
2. Boost energy flux across the horizon is ‘thermalized’ at the Unruh temperature.
1. The horizon system is a ‘heat bath’, with universal entropy area density. Postulate for all
such horizons
Implies focusing of light rays by space?me curvature: the causal structure must sa?sfy Einstein field equa?on, with Newton’s constant
€
G= 14α
3. Energy conserva?on (energy-‐momentum tensor divergence-‐free)
Adjust horizon so expansion [and shear]* at final point vanishes. Then null geodesic focusing eqn implies*
δQ
T≈ 2π
�
�
H
(−λTabkakb) dλ dA
Boost Killing vector that vanishes at is so
δS ≈ α
�
H
(−λRabkakb) dλ dA
Rab =2π
�αTab + Φgab
If this holds for all horizons it follows that, for some Φ
λ = 0
λ = 0
*Three op?ons: 1. thin patch 2. symmetric patch 3. Arrange vanishing expansion on whole final surface?
ξa = −λk
a + O(x3)
*Shear ok if allow viscous entropy produc?on term in Clausius rela?on with viscosity ra?o η/s = �/4π
9/6/11
6
Rab =2π
�αTab + Φgab
Ma\er energy conserva?on ∇aTab = 0 implies ∇aRab = ∇bΦ
Bianchi iden?ty implies ∇aRab =12∇bR
Φ =12R + ΛHence
is a cosmological constant, and Newton’s constant is given by Λ
G =1
4�αNota bene: implies
I conjecture the converse holds as well.
α <∞ G �= 0
Where to place the bifurca?on point of the Killing vector?
Spacelike outside the horizon.
Timelike outside the horizon.
Varia?on away from a sta?onary state, like the physical process 1st law. Changes defini?on of heat current, but not the TOTAL heat flux if integrate from p0 to p: � p
p0
TabξbnewdH
b =�
dA
� 0
λ0
dλ(λ− λ0)Tkk =�
dA
� 0
λ0
dλ(−λ)Tkk
ξanew = (λ− λ0)ka
ξa = −λka
9/6/11
7
Sgrav ∼1l2P
�d4V (Λ + R + l2R2 + l�4R3 + · · · )
Can higher curvature terms be captured in the equa?on of state of vacuum thermodynamics?
If “natural”, i.e. only one length scale, the answer is NO! Let R ~ lc-‐2 .
Rela?ve ambiguity of local KV and therefore heat is O(x2/lc2).
Smallest x we can use is the Planck length lP, so minimum ambiguity of the heat is ~ lP2/lc2.
But this is the rela?ve size of the R2 term! The R3 term is even more suppressed.
What if NOT natural, i.e. if l >>lP ?? Then I claim the smallest length we can use is l, so once again the R2 term is lost in the noise. (See forthcoming paper.)
CAVEAT: The rela?ve ambiguity of the KV on the horizon can be one order smaller if one Imposes a further condi?on that it vanish on a geodesic D-‐2 surface through p…
Nevertheless, we and many others tried to include the higher derivaAve terms…
Special case: s(R). Then since dR is nonzero in general, the entropy rate of change has an order unity part, which cannot match the heat. This can be dealt with in many ways:
1. Conformal transforma?on (field redefini?on) to “Einstein frame”: s(R) √h = √h’
2. Cancel with nonzero expansion at p, allow for bulk viscous entropy produc?on. (Eling, Guedens, TJ) 3. Auxiliary scalar field (Chirco, Eling, Libera?)
Might declare victory in this special case, but what about the general case?
9/6/11
8
General higher curvature entropies?
The s(R) approaches don’t work. Need to account for the change of the Different projec?ons of curvature that occur in the entropy. Lovelock case looks sweet: e.g. Gauss-‐Bonnet gives integral of intrinsic Ricci scalar of the horizon cut. The problem: how to evaluate the evolu?on of this quan?ty??
Idea: try to mimic the way entropy density of sta?onary black holes occurs in Wald’s Noether charge approach. (Brustein & Hadad; Parikh & Sarkar; Padmanabhan; Guedens, TJ & Sarkar)
S =�
sabNabdA, sab = P abmn∇mξn + W abnξn
There are various problems and issues with the previous work. Awer two years of struggle we think we’ve sorted it all out, and the paper will appear shortly – perhaps even during this conference!
Some aspects of this Noetheresque approach:
1. Depends on KV, which is not unique. Sta?s?cal interpreta?on?
2. Evaluate change of entropy using Stoke’s theorem: two horizon slices must share a common boundary (see Fig.).
3. LKV must sa?sfy Killing equa?on and Killing iden?ty to required order. Using a null geodesic normal coordinate system adapted to the horizon we showed this can be done but only if the horizon patch is parametrically narrower than it is long.
4. We show that the symmetric part does not contribute to the entropy change.
5. Clausius rela?on implies: (The W term cancels the contribu?on from the deriva?ve of s(R) in that case.)
6. The equa?on of state is integrable for if
in which case it is the field equa?on for the Lagrangian L. We don’t know if this is the only way the integrability condi?on can be sa?sfied, but it seems it might be.
7. We must take the bifurca?on surface as the earlier surface, not the later one, If the entropy on a constant V slice is to be the area and not minus the area (in the GR case).
W a[de] +∇cPac[de] = 0
P ab(mn)
P ab[mn] = −∂L[g, Rpqrs]∂Rabmn
Φ
9/6/11
9
p
!
!0
ka
V = V0
V = 0
!
P0 !a = 0( )
! p
Conclusion:
Let’s talk about it…
9/6/11
10