Holographic construction of CFT excited states. Kostas Skenderis

Transcription

1 Holographic construction of CFT excited states STAG CH RESEARCH ER C TE CENTER Aspects of Conformal Field Theories Thessaloniki, Greece 24 September 2015

2 Introduction According to holography, gravity in (d + 1) dimensions is related to QFT in d dimensions. A central question is how the bulk is reconstructed by QFT data. Here I will summarize what we know and show how to obtain the bulk solution corresponding to an excited CFT state.

3 References KS, Ariana Christodoulou, Holographic construction of CFT excited states, to appear Earlier relevant work: S. de Haro, S. Solodukhin, KS, Holographic reconstruction of space-time and renormalization in AdS/CFT Commun.Math.Phys. 217 (2001) hep-th/ KS, B. van Rees, Real-time gauge/gravity duality, Phys.Rev.Lett. 101 (2008) , ; Real-time gauge/gravity duality: prescription, renormalization and examples, JHEP 0905 (2009) 085,

4 Outline

5 Outline

6 QFT To define a theory we typically give the Lagrangian. Depending on context we may be interested in: vacuum correlators 0 O 1 (x 1 ) O(x n ) 0 correlators in a non-trivial states, α O 1 (x 1 ) O(x n ) α For example, α may be a state that spontaneously breaks some of the symmetries, a thermal state, a non-equilibrium state, etc. The correlators may be time-ordered, Wightman functions, advanced, retarded...

7 QFT Given a theory, a state and a set of operators there is a unique answer for their correlators. Any dual formulation should have the same properties. How does it work in holography?

8 Holography Within the gravity approximation: The dual QFT is represented by: An (asymptotically) AdS solution of (d + 1)-dimensional gravity coupled to matter. Operators are dual to bulk fields and the state is encoded in subleading terms in bulk solutions.

9 Euclidean vs If we are interested in time-independent states (vacuum state, thermal state) we may Wick rotate to. In Lorenzian signature there are different types of correlators (time-ordered, Wightman, retarded, etc.). In Euclidean QFT there is a unique type of correlators. So to simplify matters let s initially focus on. We will return to full fledged Lorentzian discussion afterwards.

10 Outline

11 Bulk scalar field Bulk scalar fields Φ of mass m 2 = ( d) are dual to operators O of dimension. The solution of the bulk field equation has an asymptotic expansion of the form [de Haro, Solodukhin, KS (2000)]: Φ(x, r) = r d φ (0) + + log r 2 ψ (2 d) + r φ (2 d) + φ (0) is the source for the dual operator and all blue terms are locally determined by it. φ (2 d) is related with the expectation value of O, O φ (2 d)

12 QFT vs gravity We argued earlier that the Lagrangian and the state uniquely specify the correlators. The counterpart of this statement on the gravity side is that (φ (0), φ (2 d) ) uniquely specify a bulk solution. One way to see this is to use a radial Hamiltonian formalism, where the radial direction plays the role of time. The renormalised radial canonical momentum is π = O φ (2 d) By a standard Hamiltonian argument, specifying the conjugate pair (φ (0), π ) uniquely picks a solution. [Papadimitriou, KS (2004)]

13 Holographic reconstruction I Given QFT data (φ (0), O ) there is a unique solution Φ(x, r) of the bulk field equations.

14 QFT vs gravity In QFT the vacuum structure is a dynamical question: in general, one cannot tune the value of O. The counterpart of this statement in gravity is that regularity in the interior selects O. A generic pair (φ (0), π ) leads to a singular solution.

15 Outline

16 New issues On the gravity side: In the bulk equations of motion do not have a unique solution given boundary conditions. One needs in addition initial conditions. There are non-trivial normalisable modes that vanish at the boundary. On the QFT side: Given a Lagrangian and operators there is more than one type of correlator one may wish to compute. One may wish to compute ttime-ordered, anti-time ordered, Wightman products, etc. on non-trivial states.

17 QFT On the QFT side one can summarise the new data needed by providing a contour in the complex time plane specifying where operators are inserted in this contour.

18 Examples t T C0 C1 T C3 C2 a b c a Vacuum-to-vacuum contour: computes correlators 0 T ( ) 0 b In-in contour: computes correlators α α c Thermal contour: computes thermal correlators

19 Examples Wightman in-in 2-point function, 0 O(x)O(y) 0 t t1 0 t2 Thermal Wightman 2-point function. t t1 0 t2

20 Real-time gauge/gravity duality [van Rees, KS (2008)] The holographic prescription is to use "piece-wise" holography: Real segments are associated with Lorentzian solutions, Imaginary segments are associated with Euclidean solutions, Solutions are matched at the corners.

21 Examples a E L E b E L L E c E L L

22 Comments In the extended space-time associated to a given contour, boundary conditions and regularity in the interior uniquely fix the bulk solution. The Euclidean caps can also be thought of as Hartle-Hawking wave functions.

23 Holographic reconstruction II: Lorentzian case Given QFT data (φ (0), O ) and a contour in the complex time plane there is a unique solution Φ(x, r) of the bulk field equations.

24 Outline

25 Excited states Excited states are obtained by acting with operators on the CFT vacuum, = O 0 Using the in-in formalism one may compute correlation functions in this state: τ 0 τ 3 a b t 1 t 2 T This means that we are considering the path integral ( Z[φ (0) ] = [Dϕ] exp i dtd d 1 x ( ) ) g (0) LQF T + φ (0) O C

26 1-point functions In preparation for the holographic discussion let us discuss the 1-point function, ( O (x) = [Dϕ]O (x) exp i dtd d 1 x ( ) ) g (0) LQF T + φ (0) O C To leading order in the sources, O (x) = 0 O (x)o (0) 0 φ (0) + φ+ (0) 0 O ( )O (x) 0 = 0 O (x) φ (0) + φ+ (0) O (x) 0 The 2-point functions here are Wightman functions, not time-ordered.

27 Remarks Setting φ + (0) = 0, φ (0) = 1 and taking x computes the norm of the state. The orthogonality of the 2-point functions implies that O i (x) = 0, i to linear order in sources. However, to higher order in the sources, O i (x) 0, i

28 Holographic construction In holography each gauge invariant operator O i is mapped to a corresponding bulk field Φ i. If we are interested in the holographic construction of the CFT state to leading order in the sources, it suffices to consider only the corresponding field Φ to linear order. To higher order one needs to consider other bulk fields as well.

29 Holographic construction The time contour and the corresponding bulk space-time are: τ 3 E b 2T τ 0 τ 3 a b t 1 t 2 T t 2 t 1 L L T τ 0 E a

30 Procedure We need to find regular solutions of the (linear) bulk equations with the prescribed boundary conditions in each part of the extended spacetime. Impose the matching conditions. I will present the results for a scalar field in AdS 3 in global coordinates. We have also carried the same computation in Poincaré coordinates.

31 Obtaining the solution: Euclidean solutions In the past cap we consider a solution with a delta function source: Φ E = rd φ (0) (τ 0, φ) +..., φ (0) (τ 0, φ) = δ(τ 0 a)δ(φ) This leads to Φ E (τ 0, r, φ) = n=0 k Z ( i 4π 2 [ θ( τ 0 a )φ (0) (ω nk, k)e ω nk τ0+ikφ + +θ(τ 0 + a )φ (0) (ω+ nk, k)e ω+ nk τ0+ikφ] + d nk e ω nk τ0+ikφ) g(ω nk, k, r) where g(ω nk, k, r) r as r 0 and ω ± nk are the frequencies of the normalizable modes. There is a similar expression at the future EAdS cap.

32 Lorentzian solution I The Lorentzian solution in the first Lorentzian part is a sum of normalisable modes Φ 1 L(t 1, r, φ) = n,k (a nk e iω+ nk t1+ikφ + a nk eiω+ nk t1 ikφ )ig(ω nk, k, r). There is a similar expression for the second Lorenzian part.

33 Matching conditions The matching conditions are continuity of fields and momenta along the contour: Φ E τ0=0 = Φ 1 L, t1=0 τ 0 Φ E = i τ0=0 t 1 Φ 1 L t1=0 Φ 1 L t1=t = Φ 2 L, t2=t t 1 Φ 1 L = t1=t t 2 Φ 2 L t2=t Φ 2 L t2=2t = Φ + E, τ3=0 t 2 Φ 2 L = i t2=2t τ 3 Φ + E. τ3=0 The matching conditions fix all constants in terms of the sources. In particular, they imply a nk φ (0) (ω+ nk, k), a nk φ+ (0) (ω nk, k)

34 The solution across the matching surface Past Euclidean Lorentzian Lorentzian Future Euclidean τ 0 a 0 t 1 T t 2 2T a τ 3 Amplitude of a single mode as a function of (Euclidean and then Lorentzian) time for fixed r, φ.

35 Holographic 1-point function The 1-point function is extracted from the asymptotic, Φ(x, r) = + r φ (2 d) +, O φ (2 d) In our case the solution is normalizable, so O = lim r Φ 1 r 0 L(r, t, φ) ( lim φ r 0 (0) (ω+ nk, k)e iω+ nk t+ikφ + φ + (0) (ω nk, t ikφ) k)eiω+ nk ig(ω nk, k, r) n,k The limit and the sums can be evaluated exactly [van Rees, KS (2008)] leading to O (x) = 0 O (x)o (0) 0 φ (0) + φ+ (0) 0 O ( )O (x) 0 in exact agreement with our earlier QFT computation!

36 Summary Bulk excitations (normalizable modes) are dual to excited CFT states.

37 The extrapolation dictionary An a priori different dictionary was suggested in the early days of AdS/CFT [Banks, Douglas, Horowitz, Martinec (1998)]. Boundary correlation functions can be obtained as a limit of bulk correlation functions: O (x 1 ) O (x n ) QF T = lim r 0 r n Φ(x 1, r) Φ(x n, r) Bulk This prescription is equivalent to the standard one for scalars in a fixed background [Harlow, Stanford (2011)]. It also agrees with the real-time prescription for non-equilibrium scalar 2-point functions [Keranen, Kleinert (2014)].

38 Bulk reconstruction The extrapolate dictionary suggests the following map between boundary and bulk operators: Φ(x, r) = d d x K(x r, x)o (x ) K is called the smearing function. Bulk correlation functions are then related to boundary correlation functions by Φ(x 1, r 1 ) Φ(x n, r n ) Bulk = d d x 1...d d x nk(x 1 r 1, x 1 )... K(x n r n, x n ) O (x 1) O (x n) QF T [Banks, Douglas, Horowitz, Martinec (1998)],... [Hamilton, Kabat, Lifschytz, Lowe (2006)]...

39 Smearing function In our discussion we reconstructed the bulk solution from O (x ) One can also follow similar steps to those in [Hamilton, Kabat, Lifschytz, Lowe (2006)] to show that Φ L (x, r) = d d x K(x r, x) O (x ) In our derivation the smearing function is automatically convergent due to iɛ s that originate from the Euclidean caps. (In Hamilton et al these insertions were added by hand.)

40 Remarks In our derivation the relation was between classical fields in the bulk and expectation values of boundary operators in a specific state. This appears different than a map between bulk and boundary operators. If we consider a quantized bulk field ˆΦ L (t 1, r, φ) = n,k (â nk e iω+ nk t1+ikφ + â nk eiω+ nk t1 ikφ )ig(ω nk, k, r) where now a nk, a nk are creation and annihilation operators, then our results would imply that φ ± (0) are quantum operators. However, from the QFT perspective φ ± (0) are classical sources that couple to quantum operators.

41 Beyond linear fields This boundary-to-bulk map cannot hold in this form in complete generality. If the QFT contains operators with dimensions, O 1, O 2, O 3, with 3 = Then [KS, Taylor (2006)], O 3 = π 3 + c 12 π 1 π 2 where c 12 is a constant. This suggests Φ 3 (x, r) = d d x K(x r, x) O 3 (x ) + d d x K 1 (x r, x) O 1 (x ) O 2 (x ) Locality also suggest non-linear terms should be added to the map [Kabat, Lifschytz, Lowe (2011)]...[Kabat, Lifschytz (2015)].

42 Outline

43 I discussed how QFT data reconstruct classical fields in the bulk, both in Euclidean and. The map works in full generality. I explicitly worked out the dual of the CFT excited state at the linearized level. It is straightforward (but tedious) to work out the dual at the non-linear order. The dual solution would involve many bulk fields.