![Arxiv:2106.00206V1 [Hep-Th] 1 Jun 2021 I.Oelo Atto Ucini Hra D Ihteten the with Ads Thermal in Function Partition One-Loop III](http://data.docslib.org/img/e44acb952d57096567170f59c23a5670-1.webp)
arXiv:2106.00206v1 [hep-th] 1 Jun 2021 I.Oelo atto ucini hra d ihteten the with AdS thermal in function partition One-loop III. I reLvlPriinFnto fBZbakhl ihEWbr ETW with hole black BTZ of Function Partition Level Tree II. .Introduction I. etrfrGaiainlPyisYkw nttt o The for Institute Physics,Yukawa Gravitational for Center n-opCreto oteASBF atto ucini t in Function Partition AdS/BCFT the to Correction One-loop .Oelo atto ucinfravco field vector a for function partition One-loop D. .BZwt esols T branes ETW tensionless with BTZ A. .Arve nha enlmethod kernel heat on review A A. .Oelo atto ucinfrasaa field scalar a for function partition One-loop C. .BZwt T rnsfrgnrltension general for branes ETW with BTZ B. .Oelo atto ucini hra d ihEWbae1 brane ETW with AdS thermal in function partition One-loop B. .Oelo atto ucinfrasmercsi- ed1 field spin-2 symmetric a for function partition One-loop E. .Oelo atto ucinfrgravity for function partition One-loop F. o h he iesoa uegaiy eas aclt th calculate also We thermal gravity. consist pure of dimensional one-l set three at the a AdS/BCFT of for of breakdown or breakdown brane the ETW the indicate for conditions may This of brane. one- fluctuations ETW the the contain at which but terms, case, unexpected non-rotating contain for ones calcula previous level the tree the with At condition. boundary Neumann the in aew ontfidacerpyia nepeainulk the unlike interpretation physical clear a find not do we case n ftewrdbae EWbae)frabtaytninan tension thermal arbitrary Euclidean of for function branes) partition partition (ETW branes world the of end ecluaetete ee atto ucino Euclidean of function partition level tree the calculate We AdS iahrkw iaeh,Skok,Koo6680,Japan 606-8502, Kyoto Sakyo-ku, Oiwakecho, Kitashirakawa 3 ihteEWbaei h iihe onaycniinadin and condition boundary Dirichlet the in brane ETW the with he iesoa ueGravity Pure Dimensional Three CONTENTS uk Suzuki Yu-ki T AdS inesEWbae9 brane ETW sionless rtclPyisKooUiest and University Physics,Kyoto oretical 3 ihatninesEWbrane ETW tensionless a with ns4 anes atto ucinin function partition e emn case. Neumann h hs ed nthe on fields ghost the inorrslsmatch results our tion oplvlorresults our level loop T lc oewith hole black BTZ loteone-loop the also d o ee tleast at level oop n boundary ent ∗ YITP-21-50 this he 16 11 13 3 7 4 5 0 9 2
IV. Physical Interpretation of the results 16 A. BCFT interpretation of the partition function 16 B. One-loop partition function with Dirichlet boundary condition 18 C. Consistency of the boundary condition 19 D. SL(2, Z) summation of the partition function 23 E. One-loop exactness of the partition function 24
V. Conclusions and Discussions 24
Acknowledgments 25
References 25
Appendix A 28
I. INTRODUCTION
There is a famous work on the calculation of 1-loop partition function in gravity almost half a century ago [1]. In the context of AdS/CFT correspondence [2], there are some works on deriving the full quantum gravity partition function of pure gravity at finite temperature [3–5]. At one-loop level they deduced the results from CFT path integral [4], and later it was directly proved using the heat kernel method [3]. Our goal is repeat the calculation but the difference is that we consider AdS/BCFT case. AdS/BCFT correspondence is an extension of AdS/CFT correspondence to the case where manifolds admit boundaries [6–9]. AdS/BCFT identifies the chiral mode with the anti-chiral mode. In this paper firstly we calculate the tree level partition function of rotating Euclidean BTZ black hole with ETW branes with arbitrary tension and the result is
πR+ 1 1+ Tπ 1+ T0 ZBTZ tree = exp + log( ) log( ) . (I.1) − 8G 8G 1 T − 1 T − π − 0 This matches with previous calculations in the non-rotating case [6]. Secondly we calculate the one-loop partition function of thermal AdS3 with a tensionless ETW brane and the result is
∞ 1 ∞ 1 ql+2ql+1 1 ql+1ql+2 Zgraviry = − − . (I.2) 1 qm l+2 l l l+2 m=2 ! l=0 p 1 q q p1 q q ! Y | − | Y − − There are two contributions: the first term correspondsp to the chiralp mode of the original result and the second term is the new effect due to the ETW brane. From the AdS/BCFT point of view we expect that only the first term survives because chiral and anti-chiral mode have the 1 1 same form 1 qm and 1 qm respectively and we can only consider the chiral mode in principle. − − The second term explicitly contains anti-chiral mode and represents the contribution from a massive ghost vector field and massless spin-2 field. However we note that the second term is also invariant under the exchange of q and q, therefore we can justify the operation q = q, which we will take. This discrepancy from our expectation may indicate the breakdown of consistent boundary conditions (henceforth we denote B.C.s) for calculating an one-loop partition function of Euclidean gravity. There are a lot of attempt on this subject, see [10, 11]. Our result indicates one more important thing. At tree level the partition function with tensionless branes become square root of the original result of no brane case. If we consider Dirichlet boundary condition on the brane [12], the result is the same as Neumann case in the tensionless case. More clearly in the tensionless case the tree-level partition function does not depend on the boundary condition. However at one-loop level we have additional modes: remnants of gauge symmetry. In other words, the partition function heavily depends on the boundary condition at one-loop level. These are unphysical modes, so we can eliminate these terms if we can impose 4 consistent boundary condition on gravity. However we can view these additional modes as edge modes. Toward the complete understanding of the edge modes it will be a future work to determine what kind of modes are localized on the brane. Next we consider summing over the modular transformation. Boundary CFT lives on the conformal boundary of a half of the solid torus [13] usual AdS/CFT case. We expect that it still admits modularity even if we insert branes because from the boundary torus point of view it still admits modular invariance and then we just consider how to locate the position of the brane in the bulk. To derive the full partition function we have to sum over SL(2, Z) transformation of the thermal AdS result [4, 14]. We discuss briefly in section 4, but we do not give explicit formula. This paper is organized as follows. In section 2 we directly calculate the partition function of rotating BTZ black hole with ETW branes. The calculation is straightforward, but may be laborious. There are already some works on this calculation for no boundary case [15–18]. In section 3 we review the heat kernel method for calculating 1-loop partition function [19– 22]and then we calculate the 1-loop partition function of the scalar, the vector, and the spin-2 fields using a method of images. In section 4 we discuss physical interpretation of the partition function and the consistency of the boundary conditions. In section 5 we make conclusions and discussions on the consistency of Neumann B.C. brane at 1-loop level and present some future directions. In appendix A we will give the detailed calculation of derivatives of chordal distance u.
II. TREE LEVEL PARTITION FUNCTION OF BTZ BLACK HOLE WITH ETW BRANES
A. BTZ with tensionless ETW branes
Let us consider an Euclidean rotating BTZ black hole in three dimension [15]. The metric is given by
2 2 2 2 2 2 (r R+)(r + R ) 2 r 2 2 R+R 2 − − ds = − 2 dt + 2 2 2 2 dr + r (dφ 2 dt) . (II.1) r (r R+)(r + R ) − r − −
Here R is real valued and R+ is the horizon radius. The absence of conical singularity − requires the periodicity of time and rotational angle
(t, φ) (t + β,φ + θ) (II.2) ∼ 5
2πR+ 2πR θ where β = 2 2 and θ = 2 −2 . If we define new coordinate as φ′ φ t , the R++R R++R β − − ≡ − periodicity can be recast as (t, φ′) (t + β,φ′) [16]. We also rewrite the metric in terms of this ∼ coordinate as
2 2 2 2 2 (r R+)(r +R ) 2 R R+R R 2 − 2 − + r − 2 − 0 − r R+R r R+ − r R+ − − r2 gab = 0 2 2 2 2 0 . (II.3) (r R+)(r +R ) − − R 2 2 − r R+R 0 r R+ − − 2 Note that det(gij) = r . To identify the location of ETW branes we use the map to the Poincare coordinate, which is given by
1 r2 R2 2 R2 +R2 − + + η = 2 2 cos − t + R φ′ exp (R+φ′) r +R R+ − − 1 r2 R2 2 R2 +R2 − + + x = 2 2 sin − t + R φ′ exp (R+φ′) (II.4) r +R R+ − − 1 2 2 R++R 2 − z = r2+R2 exp (R+φ′) − and the metric becomes
dz2 + dx2 + dη2 ds2 = . (II.5) z2
In this coordinate the identification (t, φ′) (t + β,φ′) is trivial, but identification (t, φ′) ∼ ∼ (t, φ′ +2π) is not trivial so we should identify by hand. If we define the following complexified
2π(R++iR ) 2πR coordinate w = η +ix, then the identification is translated into (w, z) (we − , ze − ). ∼ 2 From this perspective we can take the fundamental region as 1 w + z2 e2πR+ and the ≤ | | ≤ horizon r = R is mapped to x=η=0 and 1 z e2πR+ . Firstly let us consider tensionless + ≤ ≤ branes. In Poincare coordinate a brane configuration is determined by [23]
(z α)2 +(x p)2 +(η q)2 = β2 (II.6) − − − α where tension T = β . Here ETW branes, where φ′ is constant, are just a sphere of radius 1 for
πR+ φ′ =0 and e for φ′ = π. We will calculate the action
1 1 1 k S = √g(R +2)d3x √h(K T )d2x √hKd2x + S (II.7) −16πG − 8πG − − 8πG 8πG ct ZN ZQ ZM where N is the 3-dim bulk region, Q is the ETW brane and M is the a conformal boundary placed at r = R, which we later take R . S = √h is a counterterm constructed from →∞ ct M only induced geometric quantity [24]. Firstly considerR computing the induced metric and the extrinsic curvature at r = R surface. After some computations we get 6
2 (r2 R2 )(r2+R2 ) R R R R − + 2 + 2 2 − + R − 2 − 0 − R R+R R R+ − R R+ − − hab = 0 00 (II.8) R 2 2 − R R+R 0 R R+ − − and
2 2 R++R R 2 − − R+ R+ 2 2 2 2 Kab = (R R+)(r + R ). (II.9) R − − 1 − R+ q From this we see that
2 2 2 2 det(hab)=(R R+)(R + R ) −2 2 2 − (II.10) 2R R++R − − K = 2 2 2 2 . √(R R+)(R +R ) − − Einstein-Hilbert action is just the volume integral because curvature becomes R = 6: −
R β π 2 2 1 4πβ R R+ dr dt dφ′( 4r)= − . (II.11) − 16πG − 16πG 2 ZR+ Z0 Z0 We can also evaluate GHY term at r = R:
2π t β (1 )+π 2R2 R2 +R2 1 β − β 2 2 2 2 − + dt 2π t dφ′ (R R+)(R + R ) 2 2 2 −2 8πG 0 (1 ) (R R+)(R +R ) − β − β − − − − (II.12) R R pβ 2 2 2 = 8G (2R R+ + R ). − − − Now we determine the counterterm. That should be constructed from geometric quantity of boundary surface. Here as a most simple one (k = 1) we take
2 2 2 2 2 Sct = d x√h = βπ (R R+)(R + R ). (II.13) M − − Z q We combine the above results altogether and we get:
β 2 2 2 2 2 2 S = R + R (R R+)(R + R ) . (II.14) −8G − − − − q If we take R limit this becomes →∞
2 2 2 β R+ + R πR+ π β − S = = 2 2 . (II.15) ∼ − 16G − 8G −4G(β + θ ) This is what we expected because we choose the position of ETWs so that the volume of the bulk space becomes a half of the original volume. In full AdS case this is done in [18], which is just twice of our calculation. 7
B. BTZ with ETW branes for general tension T
Next we consider general T case. It is natural to assume the rotational symmetry i.e. p = q =0 in II.6. Therefore we will consider the following equation
2 2 2 2 2 R+ + R 1 r R+ 2 ( − ) 2 exp(R φ′) α +( − )exp(2R φ′)= β . (II.16) r2 + R2 + − r2 + R2 + − − Here we have two parameters α and β , so we have to eliminate and represent only by T.
The missing point is the fact that the brane is anchored at φ′ = 0, π. If we take r , then → ∞ the brane equation becomes
2 2 α +1= β (φ′ = 0) (II.17) 2 2 α + exp(2πR+)= β (φ′ = π). We note that T has range 1 2 2 2 2 2 1 T R+ + R T R+ + R ′ − − φ = log 2 2 + 2 ( 2 2 )+1 (II.18) R+ √1 T 2 s r + R s1 T r + R ! − − − − and for the case where the brane is anchored at φ′ = π, the equation also becomes 2 2 2 2 2 1 πR+ T R+ + R T R+ + R φ′ = log e − + ( − )+1 . (II.19) 2 2 2 2 2 2 R+ ( √1 T s r + R s1 T r + R !) − − − − We can move to the calculation of the partition function. The Einstein-Hilbert action can be calculated directly 1 Rreg β S = dr dt dφ′( 4r) EH −16πG − ZR+ Z0 Z 2 2 2 2 πR T R++R T 2 R++R e + − + ( − )+1 1 Rreg β 1 √1 T 2 r2+R2 1 T 2 r2+R2 = dr dt( 4r) log − r − r − − 16πG R 2 2 2 2 − R+ 0 − + T R++R T 2 R++R Z Z − + ( − )+1 √1 T 2 r2+R2 1 T 2 r2+R2 − r − r − − β 2 2 = (R R ). 8G reg − + (II.20) This is the same result as T=0 case. It is because the branes are curved in opposite direction in the bulk, so the volume is unchanged even in T = 0 case. Next we will evaluate the brane 6 action. In three dimension extrinsic curvature K = 2T , so what we have to do is calculate induced metric. We can explicitly do this for the case where the brane is anchored at φ′ =0: 8 4 2 2 2 2 ∂φ 2 r T (R++R ) r − hrr =( ∂r ) gφφ + grr = R2 (r2+R2 )2(T 2R2 +R2 r2(T 2 1)) + 2 2 2 2 + + (r R+)(r +R ) − −− − − − 2 2 2 2 ∂φ ∂t R Tr(R+ r )√R++R − − − (II.21) htr = hrt = ∂t ∂t gφt = 2 2 2 2 2 2 2 2 R+(r +R )√r (1 T )+R+T +R − − − r2 R2 r2+R2 2 ( − +)( ) R R+R 2 htt = gtt = 2 − + − 2 − r . r R+ − r From this determinant of this induced metric is derived as 2 2 2 r (R+ + R ) − det(hab)= 2 2 2 2 2 2 . (II.22) R+(r (1 T )+ R+T + R ) − − For the case where the brane is anchored at φ′ = π , we can repeat the same thing and the result is exactly the same as above. The brane action can be written as Rreg β 2 2 2 T r (R+ + R ) − Sbrane = dr dt 2 2 2 2 2 2 . (II.23) 8πG R+ 0 sR+(r (1 T )+ R+T + R ) Z Z − − However the brane contribution for φ′ = 0, π is cancelled as we note in the calculation of Einstein-Hilbert action; the brane is curved in opposite direction with respect to the bulk space. For the other terms on conformal boundary we have just the same term as T =0 case because in r limit φ′ goes to 0 and π, so we do not suffer any modification. →∞ We combine them together and we get: β 2 2 β 2 2 2 kβ 2 2 2 2 S = 8G (Rreg R+) 8G (2Rreg R+ + R )+ 8G (Rreg R+)(Rreg + R ) − − − − − 2 2 − β(R++R ) q − . (R ) → − 16G reg →∞ (II.24) This result is trivial, so let us consider more interesting case, where two branes admit different tensions. If we define the tension of the branes as T0, Tπ for the case where the branes are anchored at φ′ =0, π respectively. For Einstein-Hilbert part the integral becomes R2 +R2 2 R2 +R2 πR+ Tπ + Tπ + e 2− + 2 ( 2− )+1 2 r2+R 1 T r2+R 1 T s s − π ! 1 Rreg β 1 √ − π − − SEH = dr dt( 4r) log . (II.25) 16πG R+ 0 R+ R2 +R2 T 2 R2 +R2 − − T0 + 0 + − + ( − )+1 r2 R2 T 2 r2 R2 1 T 2 s + s 1 0 + ! R R √ − 0 − − − The integral is complicated. We use the following formula: B drrlog(C + √D + r2) = [ r2 log(C + √D + r2)]B 1 B r3dr A 2 A − 2 A (C+√D+r2)√D+r2 =R 1 (B2 C2 + D)log(C + √D + B2) 1 (A2 C2 + D)Rlog(C + √D + A2) 2 − − 2 − (II.26) + 1 (A2 B2)+ C (√B2 + D √A2 + D) 4 − 2 − B2 log(C + √D + B2) A2 log(C + √D + A2). 2 − 2 9 The calculation is straightforward. We choose the counterterm as k =1 and we get: 2 2 2 2 β R+ + R β R+ + R 1+ Tπ 1+ T0 S = − − (log log ) (II.27) − 16G − 16πGR 1 T − 1 T + − π − 0 where we omit the divergent term which is proportional to Rreg. This matches with the non-rotating BTZ result derived in [6]. III. ONE-LOOP PARTITION FUNCTION IN THERMAL ADS WITH THE TENSIONLESS ETW BRANE In this section we will use the explicit form of the heat kernel presented in [3], but we will not explicitly write here because it is complicated. Please refer to the original paper if necessary. A. A review on heat kernel method Heat kernel method is a convenient way of calculating an one-loop partition function. This section is based on [3, 19–22]. Let us consider calculating the following partition function S(φ) Z = Dφ e− . (III.1) Z Here we assume φ is an arbitrary field and a free field for simplicity. The action can be rewritten as S(φ)= d3x √gφ∆φ (III.2) ZM where we omit indices of tensorial structure. We will consider a compact space, so ∆ has a discrete set of eigenvalues λn. The one-loop partition function is 1 1 S(1) = log det(∆) = logλ . (III.3) −2 −2 n n X If we consider non-compact space, the spectrum becomes continuous and one-loop partition function is divergent and it is proportional to volume. This divergence can be absorbed by the renormalization of the Newton constant. The heat kernel is defined as λnt K(t, x, y)= e− ψn(x)ψn(y) (III.4) n X which we usually call a propagator. We can normalize the eigenfunctions as 10 3 ψn(x)ψn(y)= δ (x y) n − (III.5) P3 M d x √gψn(x)ψm(x)= δnm. n The trace of the heat kernel is givenR byP 3 λnt d x √gK(t, x, x)= e− . (III.6) M n n Z X X Using this we can compute the 1-loop partition function as an integral over t: 1 1 ∞ dt S(1) = logλ = d3x √gK(t, x, x). (III.7) −2 n 2 t n +0 M n X Z Z X We can show the above equation by differentiating with respect to λn and note that the it is an identity up to infinite constant. The point is that K satisfies the heat conduction equation (∂t + ∆x) K(t, x, y)=0 (III.8) with boundary condition at t =0 K(0,x,y)= δ(x, y). (III.9) B. One-loop partition function in thermal AdS with ETW brane In this section we apply the calculation in [3] to our ETW brane setup. Now consider Poincare AdS3 and metric is given by dy2 + dzdz ds2 = . (III.10) y2 Here an ETW brane is placed at (z)=0 and bulk region is defined as (z) > 0. We note ℜ ℜ that the ETW brane is connected in the bulk. Since the AdS space is maximally isometric, the geodesic distance r(x, x′) only depends on the chordal distance u(x, x′): r(x, x′)= arccosh (1 + u(x, x′)) where 2 2 (III.11) (y y′) + z z′ u(x, x′)= − | − | . 2yy′ Thermal AdS can be obtained from AdS by the following identification 1 1 (y, z) q − y, q− z (III.12) ∼ | | 2πiτ where q = e and τ = τ1 + iτ2. Now we consider applying the method of images on the heat kernel method. The tensionless ETW brane is inserted at (z)=0, so if we consider mirror ℜ 11 position, z and z are mapped to z and z respectively. Then one-loop partition function − − becomes 1 ∞ dt H Z H Z S(1) = d3x √g(K / (t, x, x)+ K / (t, xmirror, x)). (III.13) 2 t +0 thermal AdS n Z Z X The heat kernel on thermal AdS can be obtained by using method of images from that of AdS, so we get 1 ∞ dt H H S(1) = d3x √g(K (t, x, γnx)+ K (t, xmirror,γnx)). (III.14) 2 t +0 thermal AdS n Z Z X For later convenience we will use different coordinate: y = ρsinθ (III.15) z = ρcosθeiφ where 1 ρ e2πτ2 , 0 θ π and π φ π . In terms of this coordinate the geodesic ≤ ≤ ≤ ≤ 2 − 2 ≤ ≤ 2 distance can be given by n coshβ cos α r(x, γ x)= arccosh 2 2 sin θ − tan θ (III.16) mirror n coshβ cos(2φ α) − r(x ,γ x)= arccosh sin2θ + tan2θ where we define α =2πnτ1 and β =2πnτ2. C. One-loop partition function for a scalar field The heat kernel of a scalar field on AdS3 is given by 2 r2 (m +1)t 4t H3 e− − r K (t, r(x, x′)) = 3 . (III.17) (4πt) 2 sinh(r) Let us consider calculating the ordinary(not mirror)part. One-loop determinant can be recast as an integral (m2+1)t Z ∞ dt e− logdet∆= V ol(H3/ ) 0 t 3 − (4πt) 2 2 m2 t r (III.18) dρdθdφcosθ ( +1) 4t ∞ dt e− R − r + 0 t ρsin3θ 3 sinh(r) n=0 (4πt) 2 6 where we split into n =0 partP andR non-zeroR part. The first term can be easily regularized: 2 3 3 (m +1)t 2 2 2 2 ∞ dt e− (m + 1) ∞ 5 (m + 1) 2 k 3 = 3 dkk− e− = . (III.19) t 2 2 6π Z0 (4πt) 8π Z0 12 The second term can be directly calculated. Firstly we change variable from θ to r and we get 2 r2 dρdθdφcosθ (m +1)t 4t ∞ dt e− − r 0 t ρsin3θ 3 sinh(r) n=0 (4πt) 2 6 2 (III.20) m2 t r P R R dρdrdφ ( +1) 4t ∞ dt e− − r = 0 t ρ 3 2(coshβ cosα) . n=0 (4πt) 2 − 6 P R R Integrating over r and φ and t in order, we can reach the final answer: 2 m2 t r dρdrdφ ( +1) 4t ∞ dt e− − r 0 t ρ 3 2(coshβ cosα) n=0 (4πt) 2 − 6 (2πnτ )2 m2 t 2 P R R√πτ ( +1) 4t 2 ∞ dt e− − = 4(coshβ cosα) 0 t 3 n=0 − t 2 6 2πnτ √m2+1 P e−R 2 = 4n(coshβ cosα) (III.21) n=0 − 6 √ 2 P q n(1+ m +1) = | | 2n 1 qn 2 n=0 | − | 6 n 1+√m2+1 ∞P q ( ) = | | . n 1 qn 2 n=1 | − | P Now let us consider the mirror part. Firstly we consider n = 0 case. The strategy is that we firstly fix φ and integrate θ or r and then integrate over other variables. This leads to the divergent result even though we use a regularization by a gamma function: 2 (m2+1)t r dt dρdθdφcosθ e− − 4t r ∞ 3 3 0 t ρsin θ (4πt) 2 sinh(r) π (III.22) 2 R R τ2√m +1 2 dφ = 2 π 1+cos(2φ) . − − 2 R However this divergence can be eliminated by the renormalization of Newton constant, so we will ignore this term. Next consider n =0 case. The trick is almost the same but the situation is slightly changed. 6 The integral over φ yields 2 2 (2πnτ2) π (m +1)t 4t dφ ∞ e− − 2 0 dt2πτ2 3 π cos(2φ α)+cosh(β) n=0 (4πt) 2 − 2 | − | 6 2 2 (2πnτ2) P R (m +1)tR 4t ∞ e− − π = 0 dt 2πτ2 3 sinh(β) (III.23) n=0 (4πt) 2 6 n 1+√m2+1 P R ∞ q ( ) = | | . n 1 q 2n n=1 ( −| | ) P Note that in our calculation we pick up different Jacobian in φ integral, but the integral over r does not change from the ordinary part. This indicates that we can use the calculation in as it is. We also remark that he same situation occurs for a vector and a spin-2 field. 13 D. One-loop partition function for a vector field In this section we consider the vector field contribution to the partition function. For later convenience we define a set of new real coordinates: x = (z)= ρcosθcosφ, ℜ (III.24) η = (z)= ρcosθsinφ. ℑ The propagator has 2 indices and invariant under exchange of those indices, so we can expand the kernel by the basis of (1,1) symmetric bi-tensor [25]: Kµν′ (t, x, x′)= F (t, u)∂µ∂ν′ u + ∂µ∂ν′ S(t, u), (III.25) where r2 e− 4t r F (t, r)= 3 sinhr − (4πt) 2 r2 (III.26) 4 e− 4t 1 t(1 ξ)2 S(t, r)= 3 √t dξe− − sinh(rξ). (4π) 2 sinhr 0 Next step is to calculate the derivatives of u. InR Poincare coordinate it can be explicitly given by 1 (z w)µ (w z)ν ∂µ∂ν u = δµν + − δν 0 + − ′ δµ0 uδµ0δν 0 . (III.27) ′ − z0w0 ′ w0 ′ z0 − ′ n o Here we note that z = xorxmirror and w = γnx . We will present detailed calculation of these derivatives in appendix. Let us consider ordinary (not mirror) part. Using those results we can compute the kernel: dt 3 µρ ∂(γnx)ν′ H n ∞ d x g g ρ K (t, r(x, γ x)) 0 t √ ∂x µν′ n 2 ∞ dt 3 R d SR β P β = 0 t d x√g du2 (e coshr)(e− coshr) 2cosα(coshr coshβ) n { − − − − (III.28) R R P dS +(F +( du ))(coshr 2coshβ 2cosα) − − β}2 t (e +2+4t) ∞ 2π2τ 4t H Z ∞ dt − ∞ dt 2 e− t = V ol( / ) 0 t 3 + 0 t (coshβ cosα) 3 (2cosα + e− ). (4πt) 2 n=1 − 4π 2 √t R P R The first term can be regularized by considering massive vector field of massless field: t H Z dt (e− +2+4t) V ol( / ) ∞ 3 0 t (4πt) 2 m2t t H Z dt e− (e− +2+4t) = limV ol( / )R ∞ 3 m 0 0 t (4πt) 2 → (III.29) H Z 1 2 3 3 3 3 2 1 = limV ol( / ) 3 (m + 1) 2 Γ (R )+2m Γ ( )+4√m +1Γ ( ) m 0 (4π) 2 2 2 2 → − − − HnZ 1 4√π 5 H Z o = V ol( / ) 3 ( 3 8√π)= 6π V ol( / ). (4π) 2 − − 14 The second term can be straightforwardly calculated: β2 ∞ 2π2τ 4t ∞ dt 2 e− t 0 t (coshβ cosα) 3 (2cosα + e− ) n=1 − 4π 2 √t β P R ∞ 2cosα+e− = 2n(coshβ cosα) (III.30) n=1 − ∞ qn+qn+ q 2n P | | . n 1 qn 2 n=1 | − | If we omit the term which is proportionalP to the volume, the answer is just a half of the original result as we expected. Next we consider mirror part. The brane is inserted at x = 0, so x = x, η = η mirror − mirror and ymirror = y. The derivatives of u are also slightly modified and we present the results in appendix A. We can mimic the previous calculation and the result is dt 3 µρ ∂(γnx)ν′ H n ∞ d x g g ρ K (t, r(x ,γ x)) 0 t √ ∂x µν′ mirror n 2 ∞ dt 3 R Rd S βP β (III.31) = 0 t d x√g du2 (e coshr)(e− coshr) 2cosα(coshr coshβ) n { − − − − R R P +(F +( dS ))(coshr 2coshβ 2cosα) . du − − } This seems to be surprising because the above form is the same as the previous result, but here we have different geodesic distance: coshβ cos(2φ α) r = arccosh( + − ). (III.32) sin2θ tan2θ This distance r depends not only θ , but also φ . To move on we fix φ and integrate θ (or r) firstly. The integral over r gives the same as the original calculation because it is independent of φ : 2 ∞ dt 3 d S β β 0 t d x√g du2 (e coshr)(e− coshr) 2cosα(coshr coshβ) n { − − − − R R P dS +(F +( du ))(coshr 2coshβ 2cosα) (III.33) − − 2 } π β 4t ∞ dt 2 1 e− t = 0 t 2πτ2 π dφ 2(coshβ+cos(2φ α)) 3 (2cosα + e− ). n − 2 − 4π 2 √t R P R The integral over φ can also be evaluated: π 2 1 π dφ (coshβ+cos(2φ α)) − 2 − π 1 =R π dφ 2(coshβ+cos(φ α)) − − (III.34) 2π 1 =R 0 dφ 2(coshβ+cos(φ)) π R = sinhβ . 15 The final answer on mirror part becomes dt 3 µρ ∂(γnx)ν′ H n ∞ d x g g ρ K (t, r(x ,γ x)) 0 t √ ∂x µν′ mirror n β R R P ∞ 2cosα+e− = 2nsinhβ (III.35) n=1 ∞ qn+qn+ q 2n = P | | . n 1 q 2n n=1 ( −| | ) P E. One-loop partition function for a symmetric spin-2 field In the calculation of vector field one-loop partition function when we calculate the mirror part, we get the same form of the trace of the kernel as that in the ordinary part. This seems to be somewhat miraculous at first sight, but this is natural because the trace of the kernel can only depend on geodesic distance r although we have the different geodesic distance. The kernel of the symmetric spin-2 field can be expanded by the basis of (2,2) symmetric tensors [3, 25]. They are quite complicated and we do not write explicitly here. The one-loop determinant of symmetric traceless tensor is given by dt 3 µρ ∂(γnx)µ′ νσ ∂(γnx)ν′ H n H n logdet∆= ∞ d x g g ρ g σ (K (t, r(x, γ x)) + K (t, r(x ,γ x))). 0 t √ ∂x ∂x µµ′,νν′ µµ′,νν′ mirror − n R R P (III.36) Let us consider the calculation of the first term. The integral over r gives dt 3 µρ ∂(γnx)µ′ νσ ∂(γnx)ν′ H n ∞ d x g g ρ g σ K (t, r(x, γ x)) 0 t √ ∂x ∂x µµ′,νν′ n β2 R R 2 4t 5t ∞ dt P2π τ2 e− t 4t e− = 0 t 2(coshβ cosα) 3 (e− cos2α + e− cosα + 2 ) (III.37) n − 2π 2 √t √5β P R ∞ 1 β 2β e− = n(coshβ cosα) (e− cos2α + e− cosα + 2 ), n=1 − where we omit n =0 termP because it is proportional to the volume. Next we consider second term. We can do the integration similarly: dt 3 µρ ∂(γnx)µ′ νσ ∂(γnx)ν′ H n ∞ d x g g ρ g σ K (t, r(x ,γ x)) 0 t √ ∂x ∂x µµ′,νν′ mirror n 2 π β R R 4t 5t ∞ dt 2 P 1 e− t 4t e− = 0 t 2πτ2 π dφ 2(coshβ+cos(2φ α)) 3 (e− cos2α + e− cosα + 2 ) n − 2 − 2π 2 √t π (III.38) 5t P R ∞ Rdt 2 1 1 t 4t e− = 0 t 2πτ2 π dφ 2(1+cos(2φ)) 3 (e− + e− + 2 ) − 2 2π 2 √t β2 4t 5t R ∞ dt R π e− t 4t e− + 0 t 2πτ2 sinhβ 3 (e− cos2α + e− cosα + 2 ). n 2π 2 √t P R The first term contains a divergent integral and here we will ignore. The second term can be simplified as: 16 β2 4t 5t ∞ dt π e− t 4t e− 0 t 2πτ2 sinhβ 3 (e− cos2α + e− cosα + 2 ) n 2π 2 √t (III.39) √5β P R ∞ 1 β 2β e− = nsinhβ (e− cos2α + e− cosα + 2 ). n=1 P F. One-loop partition function for gravity Gravity partition function can be obtained by subtracting the contribution of the vector ghost field with m2 =4 and the scalar field with m2 =4, which corresponds to the trace part of the metric from the symmetric traceless part: 1 loop 1 graviton vector 1 scalar logZ − = logdet∆ + logdet∆ logdet∆ gravity − 2 − 2 ∞ 1 1 β 2β = ( 2nsinhβ + 2n(coshβ cosα) )(e− cos2α e− cosα) n=1 − − (III.40) ∞P 1 1 1 2n n 2n n ( + n )(q (1 q )+ q (1 q )) 2n qn 1 2 1 q 2 n=1 | − | −| | − − 2 ∞P log 1 qn ∞ 1 1 2n n 2n n = | − | + n (q (1 q )+ q (1 q )). 2 2n 1 q 2 n=2 − n=1 −| | − − The first term is justP a half of the originalP theory and the second term comes from the mirror contribution. IV. PHYSICAL INTERPRETATION OF THE RESULTS A. BCFT interpretation of the partition function In this section we summarize the results of the partition function in previous section and give the physical interpretation from BCFT point of view. Let us consider scalar part. The partition function of the free scalar field with Neumann boundary condition on ETW brane is given by ∞ ∞ 1 ∞ 1 ZScalar = . (IV.1) l+h l′+h m+h m+h 1 q q ! m=0 1 q q ! Yl=0lY′=0 − Y − The first term is the ordinary partp and the second one isp the mirror part. The first term has clear interpretation: it is from primary field with conformal dimension h and summation over its descendants. The square root is due to the fact that we take the volume of the space to be a half of the original AdS space. The second term is coming from mirror effect of the original field and looks reasonable because summation of the left and right is not independent, which reflects the fact that BCFT take only chiral part of the original CFT. Here we have encountered an important question; in the bulk we have both chiral and anti-chiral modes, but in BCFT we 17 should only have the chiral mode. This discrepancy arises because we use the method of images in the bulk calculation such as heat kernel, which makes the both modes survive. However it indeed admits anti-chiral part (q), we expect that we should set q = q physically. Then we get ∞ ∞ 1 ∞ 1 ZScalar = . (IV.2) l+l′+2h 2(m+h) 1 q ! m=0 1 q ! Yl=0lY′=0 − Y − Next we move to the vector field partitionp function. It isp given by ∞ ∞ 1 1 Zvector = √1 ql+1ql′ √1 qlql′+1 √1 ql+hql′+h l=0l′=0 − − − (IV.3) ∞ Q Q 1 1 . · √1 ql+1ql√1 qlql+1 √1 ql+hql+h l=0 − − − As is the case before the first termQ is ordinary part and the second one is mirror part. In both lines the second term is the contribution from longitudinal scalar mode, so we do not treat it here. The rest is coming from the transverse vector mode. In the mirror part the left and right mode is not independent, so it is reasonable. The summation is over L 1 and L 1 plus − − descendants contribution, which represents massless spin 1 particle. For a massive vector field we can replace 1 with 1+ h. If we identify q and q, then the partition function yields ∞ ∞ 1 1 ∞ 1 1 Zvector = l+l +1 2l+1 . (IV.4) 1 q ′ 1 ql+l′+2h ! 1 q 1 q2(l+h) ! Yl=0lY′=0 − − Yl=0 − − Finally let us consider gravity partitionp function. After some calculationp we get ∞ 1 ∞ 1 ql+2ql+1 1 ql+1ql+2 Zgraviry = − − . (IV.5) 1 qm l+2 l l l+2 m=2 ! l=0 p 1 q q p1 q q ! Y | − | Y − − The first part gives the summation over vacuump and its chiralp Virasoro descendants. The second term looks complicated, but actually has important physical meaning: the numerator represents the massive vector field with h =1 and the denominator represents massless spin-2 field. From AdS3/CFT2 case it is well known that the bulk field with spin l and mass M can be related with conformal dimension 2h via M 2 = ∆(∆ 2) + l2 − (IV.6) ∆=2h + l. Therefore the h = 1 massive vector field has a mass M 2 = 4 in the bulk, which is exactly matches what appears as a ghost vector field when we fix the gauge redundancy of the gravity. On this point we revisit in sub-section 4-C. From the viewpoint of open string or chiral mode, it is natural to identify left and right-moving mode, so if we set q = q , we get 18 ∞ 1 Zgraviry = . (IV.7) (1 q2l+2)2 Yl=0 − There appears unusual exponent 2l +2 from original result [3]. B. One-loop partition function with Dirichlet boundary condition So far we impose the Neumann boundary condition on the ETW brane. However in general we have to consider mixed boundary condition. Therefore in this section we repeat the calcu- lation as we did in previous section but with Dirichlet boundary condition on the ETW brane. We note that in AdS/BCFT we usually impose the Neumann boundary condition on the ETW brane, but it may be interesting to impose the Dirichlet boundary condition as a generalization of holography, see [12]. The procedure is very simple: just flipping the sign of the mirror part. We can easily deduce the result: ∞ 1 √1 qm+hqm+h m=0 − ZScalar = (IV.8) ∞Q∞ 1 √1 ql+hql′+h l=0l′=0 − Q Q ∞ 1 1 √1 ql+1ql√1 qlql+1 √1 ql+hql+h l=0 − − − Zvector = (IV.9) ∞ Q∞ 1 1 √1 ql+1ql′ √1 qlql′+1 √1 ql+hql′+h l=0l′=0 − − − Q Q ∞ 1 ∞ 1 ql+2ql 1 qlql+2 Zgraviry = − − . (IV.10) 1 qm l+2 l+1 l+1 l+2 m=2 ! l=0 p1 q q p1 q q ! Y | − | Y − − If we set q = q , we get p p ∞ 1 ZScalar = (IV.11) l+h l′+h l=l 1 q q Y6 ′ − p ∞ transverse 1 Zvector = (IV.12) 1 ql+l′+1 l=l Y6 ′ − ∞ 1 Z = (IV.13) graviry (1 q2l+3)2 Yl=0 − where we omit longitudinal mode in vector field partition function. The crucial point is that the result is the multiples of non-diagonal which means that only the l = l′ terms survive in 6 the vector and the scalar part and for the gravity part the exponent is odd integer, unlike the Neumann case. 19 C. Consistency of the boundary condition In this section we will be more careful about boundary conditions. Usually in gravity we impose Neumann boundary condition (henceforth Neumann B.C.) or Dirichlet boundary con- dition(Dirichlet B.C.) on the ETW brane. At tree level these conditions do not harm anything in calculating physical quantities such as the partition function and the extrinsic curvature. However the situation changes when we consider at one-loop level. At the linearized level of the metric we need a gauge fixing term and a ghost field. The boundary conditions for these fields affect one-loop partition function and what is more, the gauge transformation for the metric may not be compatible with the boundary condition. That is, the boundary operator B defines the gauge invariant boundary condition Bφ =0, (IV.14) |∂M if and only if there exists boundary conditions for corresponding gauge parameter ξ Bξξ =0 (IV.15) such that Bδξφ =0. (IV.16) This condition ensures validity of 1-loop calculation of gauge invariance with Faddeev-Popov trick. Now we revisit our problem. In our calculation we impose Neumann or Dirichlet condition for all fields including ghost fields. Let us check that whether this is reasonable or not. Firstly we consider the Neumann case. Note that we now split the metric into the background AdS Poincare metric gµν and hµν representing the metric fluctuation. Of course we should impose Neumann B.C. on hij ∂ h =0 (IV.17) x ij |x=0 where i, j represents the tangential direction along the brane. The symmetric traceless tensor φµν and trace part φ are given by 1 φµν = hµν gµνφ − 3 (IV.18) ρ φ = hρ. Therefore we should also impose Neumann B.C. for these fields: 20 ∂xφij x=0=0 | (IV.19) ∂ φ =0. x |x=0 To see the condition for ghost fields we remember that the ghost generates gauge transformation for the metric: h h + η + η . (IV.20) µν → µν ∇µ ν ∇ν µ Let us check the gauge invariance of Neumann B.C. The condition for the ghost field is ∂ ( η + η )=0. (IV.21) x ∇i j ∇j i Christoffel symbols are function of y, so we can satisfy this equation if we impose Neumann B.C. for the ghost vector field ηi and the anti-ghost vector field ηi. Next we move to the components in normal direction. If we allow the boundary to fluctuate infinitesimally along x direction, then we should impose Neumann boundary condition for ηx and ηx: ∂xηx x=0=0 | (IV.22) ∂ η =0. x x |x=0 In addition, the BRST variation for ηµ is given by 1 δη = νh ∂ φ. (IV.23) µ ∇ µν − 2 µ This variation should also satisfy IV.22, therefore we can get the boundary condition of metric along the normal direction: ∂ hµν =0. (IV.24) x∇ν |x=0 Christoffel symbols are only function of y, so we can satisfy the above equation if we impose ∂ h =0. (IV.25) x µν |x=0 This is what we explicitly calculated in section 3. However we can consider one more possible boundary conditions. If we strictly fix the boundary to be on x = 0, then we should impose the Dirichlet boundary condition for ηx and ηx: ηx x=0=0 | (IV.26) η =0. x |x=0 Correspondingly BRST variation of the ghost field changes as 21 1 δη = νh ∂ φ =0 (IV.27) µ ∇ µν − 2 µ at the boundary x =0. This equation determines the boundary condition for the metric along x direction. Next we consider the Dirichlet case. For the tangential direction we should impose Dirichlet B.C. hij =0 φij = φ =0 (IV.28) ηi = ηi =0 at the boundary x = 0. If we require that the gauge transformation should vanish at x = 0, then we get ηx x=0=0 | (IV.29) η =0. x |x=0 The BRST variation gives additional constraints on metric 1 δη = νh ∂ φ =0. (IV.30) µ ∇ µν − 2 µ This result is previously discussed in [26] and is somewhat remarkable; in Neumann case we show that we can impose Neumann condition for all fields along x direction at the boundary, but in Dirichlet case we have to Neumann like boundary condition for metric as in IV.30. Therefore our calculation in section 3 for Dirichlet case is not BRST invariant because we previously imposed Dirichlet boundary condition for all fields and for all components at the boundary. The equation IV.30 states that the fluctuation of the metric for the x direction is not zero, but this seems to be inconsistent with the previous assumption that the gauge transformation should vanish at the boundary. For the completion of classifying the boundary condition we consider one more case. If we allow the boundary to fluctuate along the x direction, we get ∂xηx = ∂xηx =0 (IV.31) at the boundary. Then BRST variation gives 1 ∂ hµν ∂ν φ =0 (IV.32) x ∇ν − 2 at the boundary. However we should consider one more problem: ellipticity of the differential operators [26]. 22 The authors discuss that Euclidean linearized gravity is not elliptic and hence perturbatively ill-defined. Our problem also comes from this fact that we have yet found the suitable set of boundary conditions Euclidean gravity at one-loop level with manifold respecting BRST invariance and ellipticity. In this sense we can say that AdS/BCFT does not hold at one-loop level in pure gravity. This problem has a long history: a small selection being [10, 11, 22, 26]. Previously we derive ∞ 1 ∞ 1 ql+2ql+1 1 ql+1ql+2 Zgraviry = − − (IV.33) 1 qm l+2 l l l+2 m=2 ! l=0 p 1 q q p1 q q ! Y | − | Y − − p p and we see that for a vector field with h =1 has mass M 2 =4 in the bulk. The point is that the second term factorizes from the first one, which means that it is independent of the bulk mode. Therefore these modes live on the brane and represent fluctuations onto both tangential and normal direction. These fluctuation will indeed change the position of the ETW brane and if we do not allow the fluctuations along the tangential direction, we expect the following following result ∞ 1 Z = (IV.34) graviry 1 qm m=2 ! Y | − | because this is the contribution from only chiral modes of the original results obtained in [4]. The complete calculation for other sets of boundary conditions is computationally hard, so it will be our future problem. For Dirichlet B.C., the interpretation of the one-loop partition function seems hard. The numerator and the denominator is exchanged from Neumann B.C. case, so there seem to be the contribution of spin-2 massless ghost field and massive spin-1 vector field. Here we do not identify the physical meaning of this term because the Dirichlet B.C. for metric on ETW brane is yet well-understood. We summarize the above discussion. Firstly in both Neumann and Dirichlet case we have two sets of boundary conditions for the normal component of the vector fields like in IV.22 and IV.26. Therefore our calculation in section 3 will be consistent with BRST invariance for the Neumann boundary condition on all the fields and all the components, but for the Dirichlet case it does not consistent. Secondly in the Dirichlet case the differential operator is not elliptic, so it is perturbatively ill-defined [26], but in the Neumann case we do not know whether it is elliptic or not. 23 D. SL(2, Z) summation of the partition function We derive the one-loop partition function in section 3 and consider taking SL(2, Z) summa- tion as in [4, 14]. At first sight this seems to be hard because the first term in IV.5 cannot be expanded as a polynomial of q and q. However physically we remember that we can set q = q , so we will use IV.7 and IV.13. Firstly let us consider Neumann case. In section 1 we calculated tree-level partition function of the BTZ black hole. To derive the tree-level partition function of the thermal AdS, we can change the modular parameter as τ 1 . Therefore partition function of the thermal AdS is → − τ given by k 1 ql+2ql+1 1 ql+1ql+2 2 ∞ 1 ∞ √ √ Z0,1(τ)= qq − m − − | | 1 q √1 ql+2ql√1 qlql+2 m=2 | − | l=0 − − (IV.35) Q k ∞ Q1 = q− 2 1 q2l+2 l=0 ( − ) 1 Q where k = 16G . The whole classical Einstein solution with the boundary torus is obtained by implementing modular transformation: aτ + b τ γτ = (IV.36) → cτ + d where γ SL(2, Z)/ 1 . Hence here we can take c > 0 and sum over (c,d), which are ∈ {± } relatively prime integers. The full partition function can be written as k ∞ 1 Z(τ)= q− 2 1 q2l+2 (c,d) l=0 ( − ) k+ 1 P q−Q 6 = η(2τ)2 (c,d) (IV.37) 1 P k+ 1 = (2τ)q− 6 γ √ (2τ)η(2τ)2 ℑ (c,d) ℑ | PE(τ; kp1 ,0) = 2 − 12 √ (2τ)η(2τ)2 ℑ where we define E(τ; n, m) = ( (2τ)qnqm). Here we use the modular invariance (c,d) ℑ (2τ)η(2τ)2. P p ℑ p Next we consider Dirichlet case. As in Neumann case the partition function becomes 1 η(2τ)2 k 12 2 Z(τ)= q− − η(τ)2 (1 q) (c,d) − 2 (IV.38) η(2τ) √ (2τ) P k 1 2 √ (τ) = ℑ (q− − 12 (1 q) ℑ ) γ . η(τ)2√ (τ) √ (2τ) ℑ (c,d) − ℑ | We can continue the calculation as in [4],P but we stop here and discuss briefly how to treat this partition function. Firstly this Poincare series is divergent, so we need some regularization. One possible way is that we firstly consider the following convergent series 24 s k+ 1 ( (2τ)) q− 6 . (IV.39) ℑ X(c,d) This series is convergent for (s) > 1. However as is presented in [4], we can take analytic ℜ continuation to (s) 1 and especially at s = 1 this series is regular. We expect that the ℜ ≤ 2 spectrum has negative densities of states as is the case in pure gravity [27, 28]. It will be an interesting direction to specify the black-hole microstate using modularity of the theory. Note that here we have naturally assumed that we still admit modular invariance even inserting the ETW brane because from the boundary torus point of view it admits modularity and we just consider how to locate the position of the ETW brane in the solid torus. E. One-loop exactness of the partition function One-loop exactness of the partition function is an important problem as in any other theories. Pure gravity in three dimension is one-loop exact because the bulk diffeomorphism is governed by Virasoro symmetry, hence the partition function does not suffer any quantum correction other than Virasoro descendants and is one-loop exact. We expect the same thing in our case, though we have not explicitly shown. In BCFT, we can use the double trick as explained in [29], so we have Ln = L n. (IV.40) − Therefore half of the Virasoro symmetry or only the chiral mode survives. This also guarantees that if we can properly calculate respecting BRST invariance, then the partition function of our case will be one-loop exact. V. CONCLUSIONS AND DISCUSSIONS We calculated the tree level partition function of BTZ black hole with the end of the world brane for arbitrary tension and the result matches with non-rotating previous results. At one- loop level we calculated the thermal AdS partition function with a tensionless brane and the result heavily depends on the choice of sets of boundary conditions. We explicitly calculated the partition function in the case where the all fields satisfy Neumann boundary condition and in another case where Dirichlet boundary condition. For the Neumann case we expect that the system will be consistent with BRST quantization, but the partition function actually contains unphysical mode. For Dirichlet case we also encounter unphysical modes. There are other possible sets of boundary conditions which are sets of mixed boundary conditions for 25 each component, but in this paper we have not calculated due to computational difficulty. This will be one of our future problems. In addition, as was explained in [26], the ellipticity of the differential operator is broken if we consider the linearized Euclidean gravity on a manifold with boundary. This means that the theory is perturbatively ill-defined. In this sense we can say that AdS/BCFT does not hold at one-loop level at least for three dimensional pure gravity if we combine our result. Of course there are some possible models so that we can restore the consistency by adding a suitable set of matter fields. For example in the supergravity case we can determine using supersymmetry transformations [11]. Therefore it will be interesting to specify the minimal set of boundary conditions and matter fields for well-defined gravity. For another future direction it will be most important to specify what kind of modes localize on the brane, or decoupling from the bulk. As a bypath for formulating as a gravity, it may be interesting to understand our problem in Chern-Simons formulation. We expect that in this set up we can understand the edge mode more clearly because we only consider the boundary conditions for gauge fields. In AdS/BCFT case we identify the chiral and anti-chiral mode and for the clear physical meaning here we should take the identification for what we have calculated such as partition functions. Then we get similar expression IV.7 and IV.13 as in for one-loop partition function. Toward deriving full partition function we should sum over all possible geometries with a given boundary condition [30]. In our case the bulk geometry is a half of the solid torus, therefore we can sum over torus moduli SL(2, Z). As we discussed briefly this in section 4, the computation is quite similar as the one in [4]. We do not explicitly calculate, hence it will be an interesting direction to analyze the spectrum using modular Bootstrap in BCFT case as in pure gravity case [27, 28]. ACKNOWLEDGMENTS We thank T.Takayanagi for correcting some errors in manuscripts and encouraging me. We are also grateful to T.Nishioka, T.Ugajin, and D.Vassilevich for valuable comments and discussions. [1] Gerard ’t Hooft and M. J. G. Veltman. One loop divergencies in the theory of gravitation. Ann. Inst. H. Poincare Phys. Theor. A, 20:69–94, 1974. [2] Juan Martin Maldacena. The Large N limit of superconformal field theories and supergravity. Adv. Theor. Math. Phys., 2:231–252, 1998. arXiv:hep-th/9711200, doi:10.1023/A:1026654312961. 26 [3] Simone Giombi, Alexander Maloney, and Xi Yin. One-loop Partition Functions of 3D Gravity. JHEP, 08:007, 2008. arXiv:0804.1773, doi:10.1088/1126-6708/2008/08/007. [4] Alexander Maloney and Edward Witten. Quantum Gravity Partition Functions in Three Dimen- sions. JHEP, 02:029, 2010. arXiv:0712.0155, doi:10.1007/JHEP02(2010)029. [5] Xi Yin. Partition Functions of Three-Dimensional Pure Gravity. Commun. Num. Theor. Phys., 2:285–324, 2008. arXiv:0710.2129, doi:10.4310/CNTP.2008.v2.n2.a1. [6] Mitsutoshi Fujita, Tadashi Takayanagi, and Erik Tonni. Aspects of AdS/BCFT. JHEP, 11:043, 2011. arXiv:1108.5152, doi:10.1007/JHEP11(2011)043. [7] Masahiro Nozaki, Tadashi Takayanagi, and Tomonori Ugajin. Central Charges for BCFTs and Holography. JHEP, 06:066, 2012. arXiv:1205.1573, doi:10.1007/JHEP06(2012)066. [8] Tadashi Takayanagi. Holographic Dual of BCFT. Phys. Rev. Lett., 107:101602, 2011. arXiv:1105.5165, doi:10.1103/PhysRevLett.107.101602. [9] Andreas Karch and Lisa Randall. Open and closed string interpretation of SUSY CFT’s on branes with boundaries. JHEP, 06:063, 2001. arXiv:hep-th/0105132, doi:10.1088/1126-6708/2001/06/063. [10] Ian G. Moss and Pedro J. Silva. BRST invariant boundary conditions for gauge theories. Phys. Rev. D, 55:1072–1078, 1997. arXiv:gr-qc/9610023, doi:10.1103/PhysRevD.55.1072. [11] Peter van Nieuwenhuizen and Dmitri V. Vassilevich. Consistent boundary conditions for supergravity. Class. Quant. Grav., 22:5029–5051, 2005. arXiv:hep-th/0507172, doi:10.1088/0264-9381/22/23/008. [12] Rong-Xin Miao. Holographic BCFT with Dirichlet Boundary Condition. JHEP, 02:025, 2019. arXiv:1806.10777, doi:10.1007/JHEP02(2019)025. [13] Per Kraus. Lectures on black holes and the AdS(3) / CFT(2) correspondence. Lect. Notes Phys., 755:193–247, 2008. arXiv:hep-th/0609074. [14] Robbert Dijkgraaf, Juan Martin Maldacena, Gregory W. Moore, and Erik P. Verlinde. A Black hole Farey tail. 5 2000. arXiv:hep-th/0005003. [15] Maximo Banados, Marc Henneaux, Claudio Teitelboim, and Jorge Zanelli. Geometry of the (2+1) black hole. Phys. Rev. D, 48:1506–1525, 1993. arXiv:gr-qc/9302012, doi:10.1103/PhysRevD.48.1506. [16] Steven Carlip and Claudio Teitelboim. Aspects of black hole quantum mechanics and ther- modynamics in (2+1)-dimensions. Phys. Rev. D, 51:622–631, 1995. arXiv:gr-qc/9405070, doi:10.1103/PhysRevD.51.622. [17] G. W. Gibbons and S. W. Hawking. Action Integrals and Partition Functions in Quantum Gravity. Phys. Rev. D, 15:2752–2756, 1977. doi:10.1103/PhysRevD.15.2752. 27 [18] Juan Martin Maldacena and Andrew Strominger. AdS(3) black holes and a stringy exclusion principle. JHEP, 12:005, 1998. arXiv:hep-th/9804085, doi:10.1088/1126-6708/1998/12/005. [19] Justin R. David, Matthias R Gaberdiel, and Rajesh Gopakumar. The Heat Kernel on AdS(3) and its Applications. JHEP, 04:125, 2010. arXiv:0911.5085, doi:10.1007/JHEP04(2010)125. [20] Rajesh Gopakumar, Rajesh Kumar Gupta, and Shailesh Lal. The Heat Kernel on AdS. JHEP, 11:010, 2011. arXiv:1103.3627, doi:10.1007/JHEP11(2011)010. [21] Robert B. Mann and Sergei N. Solodukhin. Quantum scalar field on three-dimensional (BTZ) black hole instanton: Heat kernel, effective action and thermodynamics. Phys. Rev. D, 55:3622– 3632, 1997. arXiv:hep-th/9609085, doi:10.1103/PhysRevD.55.3622. [22] D. V. Vassilevich. Heat kernel expansion: User’s manual. Phys. Rept., 388:279–360, 2003. arXiv:hep-th/0306138, doi:10.1016/j.physrep.2003.09.002. [23] Ibrahim Akal, Yuya Kusuki, Tadashi Takayanagi, and Zixia Wei. Codimension two holography for wedges. Phys. Rev. D, 102(12):126007, 2020. arXiv:2007.06800, doi:10.1103/PhysRevD.102.126007. [24] Vijay Balasubramanian and Per Kraus. A Stress tensor for Anti-de Sitter gravity. Commun. Math. Phys., 208:413–428, 1999. arXiv:hep-th/9902121, doi:10.1007/s002200050764. [25] Eric D’Hoker, Daniel Z. Freedman, Samir D. Mathur, Alec Matusis, and Leonardo Rastelli. Graviton and gauge boson propagators in AdS(d+1). Nucl. Phys. B, 562:330–352, 1999. arXiv:hep-th/9902042, doi:10.1016/S0550-3213(99)00524-6. [26] Edward Witten. A Note On Boundary Conditions In Euclidean Gravity. 5 2018. arXiv:1805.11559. [27] Nathan Benjamin, Hirosi Ooguri, Shu-Heng Shao, and Yifan Wang. Light-cone modu- lar bootstrap and pure gravity. Phys. Rev. D, 100(6):066029, 2019. arXiv:1906.04184, doi:10.1103/PhysRevD.100.066029. [28] Nathan Benjamin, Scott Collier, and Alexander Maloney. Pure Gravity and Conical Defects. JHEP, 09:034, 2020. arXiv:2004.14428, doi:10.1007/JHEP09(2020)034. [29] J. Polchinski. String theory. Vol. 1: An introduction to the bosonic string. Cam- bridge Monographs on Mathematical Physics. Cambridge University Press, 12 2007. doi:10.1017/CBO9780511816079. [30] Juan Martin Maldacena and Liat Maoz. Wormholes in AdS. JHEP, 02:053, 2004. arXiv:hep-th/0401024, doi:10.1088/1126-6708/2004/02/053. 28 APPENDIX A Here we present the calculation of the derivatives of u. For ordinary part (not mirror) we get cos(θ)(eβ cos(φ α) cos(φ)) ∂ ∂ u = − − y x′ − ρ2eβsin3(θ) cos(θ)(eβ sin(φ α) sin(φ)) ∂ ∂ u = − − y η′ − ρ2eβsin3(θ) cos(θ)( eβ cos(φ α)+cos(φ)) ∂ ∂ u = − − x y′ − ρ2e2β sin3(θ) cos(θ)( eβ sin(φ α)+sin(φ)) ∂ ∂ u = − − η y′ − ρ2e2β sin3(θ) (2coshβ coshr) ∂ ∂ u = − y y′ − ρ2eβsin2(θ) ∂ ∂ u = 1 x x′ − ρ2eβ sin2(θ) 1 ∂η∂η u = 2 β 2 ′ − ρ e sin (θ) (V.1) ∂x∂η′ u = ∂η∂x′ u =0 1 β ∂ u = (e− coshr) y ρsinθ − cos(θ)(eβ cos(φ α) cos(φ)) ∂ u = − − x − ρeβsin2(θ) cos(θ)(eβ sin(φ α) sin(φ)) ∂ u = − − η − ρeβsin2(θ) ∂ u = 1 (eβ coshr) y′ ρsinθeβ − cos(θ)(eβ cos(φ α) cos(φ)) − − ∂x′ u = ρeβ sin2(θ) cos(θ)(eβ sin(φ α) sin(φ)) − − ∂η′ u = ρeβsin2(θ) . We also present those of mirror part cos(θ)(eβ cos(φ α)+cos(φ)) ∂ ∂ u = − y x′ − ρ2eβsin3(θ) cos(θ)(eβ sin(φ α) sin(φ)) ∂ ∂ u = − − y η′ − ρ2eβsin3(θ) cos(θ)(eβ cos(φ α)+cos(φ)) − ∂x∂y′ u = ρ2e2β sin3(θ) cos(θ)( eβ sin(φ α)+sin(φ)) ∂ ∂ u = − − η y′ − ρ2e2β sin3(θ) (2coshβ coshr) ∂ ∂ u = − y y′ − ρ2eβsin2(θ) ∂ ∂ u = 1 x x′ − ρ2eβ sin2(θ) 1 ∂η∂η u = 2 β 2 ′ − ρ e sin (θ) (V.2) ∂x∂η′ u = ∂η∂x′ u =0 1 β ∂ u = (e− coshr) y ρsinθ − cos(θ)(eβ cos(φ α)+cos(φ)) ∂ u = − x − ρeβsin2(θ) cos(θ)(eβ sin(φ α) sin(φ)) ∂ u = − − η − ρeβsin2(θ) ∂ u = 1 (eβ coshr) y′ ρsinθeβ − cos(θ)(eβ cos(φ α)+cos(φ)) − ∂x′ u = ρeβ sin2(θ) cos(θ)(eβ sin(φ α) sin(φ)) − − ∂η′ u = ρeβsin2(θ) .