This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

institutetext: School of Mathematical Sciences and STAG Research Centre, University of Southampton, Highfield, Southampton, SO17 1BJ, UK.

Generalised proofs of the first law of entanglement entropy

Marika Taylor and Linus Too m.m.taylor@soton.ac.uk, l.h.y.too@soton.ac.uk
Abstract

In this paper we develop generalised proofs of the holographic first law of entanglement entropy using holographic renormalisation. These proofs establish the holographic first law for non-normalizable variations of the bulk metric, hence relaxing the boundary conditions imposed on variations in earlier works. Boundary and counterterm contributions to conserved charges computed via covariant phase space analysis have been explored previously. Here we discuss in detail how counterterm contributions are treated in the covariant phase approach to proving the first law. Our methodology would be applicable to generalizing other holographic information analyses to wider classes of gravitational backgrounds.

preprint:

1 Introduction

The first law of entanglement entropy states that the variation of the entanglement entropy SBS_{B} is equal to the variation of the modular energy HB\langle H_{B}\rangle,

δSB=δHB.\displaystyle\delta S_{B}=\delta\expectationvalue{H_{B}}. (1)

The holographic first law of entanglement entropy first demonstrated in Faulkner:2013ica is applicable to field theories that admit a holographic description. In the analysis of Faulkner:2013ica the righthandside of (1) is by construction finite, as it derives from the standard holographic renormalization expressions for energy deHaro2001 . The authors of Faulkner:2013ica work with a regulated entanglement entropy and restrict variations such that the left hand side of (1) has no ultraviolet divergences. The goal of this paper is to demonstrate the holographic first law in general situations, without imposing restrictions on variations, making use of the consistent renormalization procedure for the entanglement entropy developed in Taylor:2016aoi . Holographic renormalization of entanglement entropy has been discussed in a number of other works, including Taylor:2017zzo ; Anastasiou:2018mfk ; 2018aao2 ; Anastasiou:2019ldc ; Anastasiou:2020smm .

Entanglement entropy in quantum field theory is divergent due to the correlations across the boundary of the entangling region. Holographically, the entanglement entropy is captured by the area of the Ryu-Takayangi surface Ryu:2006bv , which is also divergent due to the infinite volume of the entangling surface in the bulk spacetime. In both situations the entropy can be systematically renormalized, inheriting its renormalization scheme from that for the partition function of the theory. The renormalized holographic entanglement entropy in Taylor:2016aoi can be derived from the holographically renormalized action deHaro2001 using the replica trick. We will show it is necessary to use the renormalized entanglement entropy on the left hand side of (1)(\ref{eq:first1st}) to obtain the correct finite contributions when one considers general linear perturbations.

The covariant charge formalism can be used to give an elegant discussion of the holographic first law Faulkner:2013ica . In the covariant formalism both sides of (1)(\ref{eq:first1st}) can be expressed as integrals of charge densities over entangling surfaces. We will review this approach in the following section. The charge associated with the change in modular energy used in Faulkner:2013ica was renormalized, following the earlier works of Papadimitriou:2004ap ; Papadimitriou:2005ii . However, the charge associated with the change in entropy was not renormalized; its variation was finite in Faulkner:2013ica due to constraints on the asymptotic falloff of metric perturbations. In this paper we construct a renormalized charge corresponding to the change in entropy such that the integral version of the holographic first law applies to generic metric perturbations.

At a technical level, one can understand the construction of this charge as follows. Onshell, the density of the conserved charge is defined in terms of the current density as

𝑱=d𝑸\displaystyle\boldsymbol{J}=d\boldsymbol{Q} (2)

where 𝑱\boldsymbol{J} and 𝑸\boldsymbol{Q} are differential forms. The charge density clearly has an intrinsic ambiguity: additional exact terms in 𝑸\boldsymbol{Q} will not change the current. In our context, the exact term ambiguity in the density of the conserved charge contributes to the entanglement entropy (and modular energy) at the boundary of the entangling surface. The holographic counterterms fix the ambiguity in the density of the conserved charges, with the matching of renormalization schemes for energy and entropy ensuring that the first law holds. Relative to the expressions given for the entropy in Faulkner:2013ica , our expressions have additional boundary terms. Our general expressions are applicable to variations of the entanglement entropy associated with generic variations of the bulk metric i.e. perturbations of the non-normalizable terms in the metric.

Boundary terms in the construction of charges using the covariant phase space formalism have been discussed recently in 2020HarlowWu . The boundary counterterms associated with holographic charges were constructed using Hamiltonian renormalisation methods in Papadimitriou:2005ii . There are key conceptual differences in the entropy variation that require us to generalize relative to both of these works. The vector used to construct the entropy variation is no longer Killing. In Papadimitriou:2005ii the goal was to compute conserved charges for black holes and accordingly any variations considered would preserve the non-normalizable modes of the background. In our case the non-normalizable modes are not held fixed: the metric perturbations can be such that the non-normalizable modes vary, corresponding to deforming not just the state of the dual field theory, but the theory itself. This different physical setup leads to differences in the counterterms arising in the analysis of the covariant phase space construction, which are explained in detail in Appendix C.


The structure of the paper is the following. In section 2.1 we review the holographic renormalized entanglement entropy, introducing the notion of renormalized area integral for codimension two minimal surfaces in AlAdSAlAdS that allows us to express the renormalized entanglement entropy functional in terms of certain conformal invariants. In section 2.2 we summarise the covariant formalism or Hamiltonian formalism for holographic renormalization and conserved charges and explain in 2.3 the first law of holographic entanglement entropy, explaining the constraints imposed on variations in previous works. In section 3 we explore the infinitesimal version of the first law i.e. the radius of the entangling region is infinitesimal, for general variations, explaining the differences between odd and even dimensions. In odd dimensions the variation of the renormalized entropy can be expressed elegantly in terms of the pullback of the Weyl tensor variation.

We demonstrate the integral version of the renormalized first law in section 4. We first need to introduce in 4.1 the proper definition of the conserved charges and their integrals: we demonstrate how the equivalence relations between conserved charges need to be generalized to include appropriate counterterms once one allows for generic variations. In section 4.2, we use these conserved charges and their equivalence relations to derive the renormalized first law of entanglement entropy. This general proof is illustrated using two examples in d=3d=3, 44 and 55. We end the paper with discussion of implications and applications of our results.

2 Review of renormalized entanglement entropy, holographic charges and first law

In this section we briefly review the definition of renormalized entanglement entropy and holographic charges, and describe the first law of entanglement entropy.

2.1 Renormalized Entanglement Entropy

One of the main goals of this work is to generalise first laws for holographic entanglement entropy, relaxing assumptions on boundary conditions for bulk metric perturbations. One of the tools that will be used in our analysis is renormalized entanglement entropy; this is relevant as general boundary conditions for bulk metric perturbations are associated with UV divergences in the regulated entanglement entropy. Working with quantities that are consistently renormalized allows us to work systematically with such setups.

Renormalized entanglement entropy was discussed extensively in Taylor:2016aoi , with explicit formulae for holographic renormalized entanglement entropy being derived. A convenient way to construct expressions for renormalized entanglement entropy is via the replica trick. Using the replica trick, entanglement entropy can be derived as the limit of Rényi entropy

S=αα[log𝒵(α)αlog𝒵(1)]α=1S=-\alpha\partial_{\alpha}[\log\mathcal{Z}(\alpha)-\alpha\log\mathcal{Z}(1)]_{\alpha=1} (3)

where 𝒵(α){\mathcal{Z}}(\alpha) is the partition function on the α\alpha fold cover manifold. In much of the condensed matter literature this approach is applied to UV regulated quantities, with the UV regulator being interpreted in terms of the lattice scale of the discrete condensed matter system of interest. However, from a quantum field theory perspective, it is much more natural to work directly with renormalized quantities, ie.e. 𝒵{\mathcal{Z}} is the renormalized partition function.

In holography the partition function is computed to leading order from the onshell bulk action i.e.

𝒵grav=eIgrav,\mathcal{Z}_{grav}=e^{-I_{grav}}, (4)

where IgravI_{grav} denotes the onshell gravitational action. Applying the holographic dictionary and the replica trick to the renormalized gravity action one obtains a formal definition of the renormalized holographic entanglement entropy.

S=αα[Iren(α)αIren(1)]α=1S=\alpha\partial_{\alpha}[I_{ren}(\alpha)-\alpha I_{ren}(1)]_{\alpha=1} (5)

This approach thus directly relates the renormalization scheme for the partition function (gravitational action) to the scheme for the entanglement entropy.


To obtain a finite value for the gravitational action, one needs to use holographic renormalization. The renormalized action can then be obtained by the procedure of regularization and the introduction of appropriate covariant boundary counterterms

Iren=IregIctI_{ren}=I_{reg}-I_{ct} (6)

For pure gravity with negative cosmological constant, the renormalized action in (d+1)(d+1) dimensions takes the form deHaro2001

Iren=116πGd+1\displaystyle I_{ren}=\frac{1}{16\pi G_{d+1}} zϵdd+1xg(R+d(d1))\displaystyle\int_{\mathcal{M}_{z\geq\epsilon}}d^{d+1}x\sqrt{g}(R+d(d-1)) (7)
116πGd+1\displaystyle-\frac{1}{16\pi G_{d+1}} ϵ~ddxg~[K+2(1d)+12d\displaystyle\int_{\tilde{\mathcal{M}_{\epsilon}}}d^{d}x\sqrt{\tilde{g}}\big{[}K+2(1-d)\mathcal{R}+\frac{1}{2-d}\mathcal{R}
1(d4)(d2)2(μνμνd4(d1)2)logϵa(d)+]\displaystyle-\frac{1}{(d-4)(d-2)^{2}}(\mathcal{R}_{\mu\nu}\mathcal{R}^{\mu\nu}-\frac{d}{4(d-1)}\mathcal{R}^{2})-\log\epsilon a_{(d)}+\cdots\big{]}

In these expressions the bulk manifold is regulated using a radial coordinate zϵz\geq\epsilon; RR denotes the curvature of the bulk manifold while KK and \mathcal{R} refer to the extrinsic and intrinsic curvature of the boundary manifold respectively. Here the given counterterms suffice for d5d\leq 5; expressions for the additional counterterms required for d>5d>5 can be found in deHaro2001 . Logarithmic counterterms associated with conformal anomalies arise for dd even, and explicit expressions for these can also be found in deHaro2001 .

One can then derive the renormalized entanglement entropy from the renormalized action, making use of the following expressions for the integrals of curvature invariants, expressed as series in powers of (1α)(1-\alpha) Fursaev_1995 ; Fursaev_2013 :

αdd+1xgα=αdd+1xg+4π(1α)B~dd1xγ\displaystyle\int_{\mathcal{M}_{\alpha}}d^{d+1}x\sqrt{g}\mathcal{R}_{\alpha}=\alpha\int_{\mathcal{M}}d^{d+1}x\sqrt{g}\mathcal{R}+4\pi(1-\alpha)\int_{\tilde{B}}d^{d-1}x\sqrt{\gamma} (8)
αdd+1xgα2=αdd+1xg2+8π(1α)B~dd1xγ\displaystyle\int_{\mathcal{M}_{\alpha}}d^{d+1}x\sqrt{g}\mathcal{R}_{\alpha}^{2}=\alpha\int_{\mathcal{M}}d^{d+1}x\sqrt{g}\mathcal{R}^{2}+8\pi(1-\alpha)\int_{\tilde{B}}d^{d-1}x\sqrt{\gamma}\mathcal{R}
αdd+1xgαμναμν=αdd+1xgμνμν\displaystyle\int_{\mathcal{M}_{\alpha}}d^{d+1}x\sqrt{g}{\mathcal{R}_{\alpha}}_{\mu\nu}\mathcal{R}_{\alpha}^{\mu\nu}=\alpha\int_{\mathcal{M}}d^{d+1}x\sqrt{g}\mathcal{R}_{\mu\nu}\mathcal{R}^{\mu\nu}
+4π(1α)B~dd1xγ(μνnμnν12(TrK)2)\displaystyle\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;+4\pi(1-\alpha)\int_{\tilde{B}}d^{d-1}x\sqrt{\gamma}\big{(}\mathcal{R}_{\mu\nu}n^{\mu}\cdot n^{\nu}-\frac{1}{2}(TrK)^{2}\big{)}

Using these replica curvature integrals the explicit expression for the holographic renormalized entanglement entropy becomes Taylor:2016aoi

Sren=14Gd+1B~dd1xγ14(d2)Gd+1B~dd2xγ~\displaystyle S_{ren}=\frac{1}{4G_{d+1}}\int_{\tilde{B}}d^{d-1}x\sqrt{\gamma}-\frac{1}{4(d-2)G_{d+1}}\int_{\partial\tilde{B}}d^{d-2}x\sqrt{\tilde{\gamma}} (9)
14(d2)(d4)Gd+1B~dd2xγ~(μνnμnν12(TrK)2d2(d1))\displaystyle-\frac{1}{4(d-2)(d-4)G_{d+1}}\int_{\partial\tilde{B}}d^{d-2}x\sqrt{\tilde{\gamma}}\bigg{(}\mathcal{R}_{\mu\nu}n^{\mu}\cdot n^{\nu}-\frac{1}{2}(TrK)^{2}-\frac{d}{2(d-1)}\mathcal{R}\bigg{)}

where \mathcal{R} is the Ricci scalar of the metric gμνg_{\mu\nu}, aa=a(1)aμνnaμnaν\mathcal{R}_{aa}=\sum_{a}(-1)^{a}\mathcal{R}_{\mu\nu}n^{\mu}_{a}n^{\nu}_{a} is the projection of the Ricci tensor on the subspace orthogonal to B~\partial\tilde{B} with temporal and spatial normals naμn_{a}^{\mu}, a=1,2a=1,2, γ~\widetilde{\gamma} is the determinant of the induced metric on B~\partial\tilde{B} and k2=aKaKak^{2}=\sum_{a}K_{a}K_{a} with KaK_{a} trace of the extrinsic curvature corresponding to the two normals naμn_{a}^{\mu}. Here B~\tilde{B} denotes the entangling surface with boundary B~\partial\tilde{B}. The counterterms given here are sufficient for d<6d<6, but can straightforwardly be computed for d6d\geq 6. For dd even there are logarithmic counterterms related to conformal anomalies, see Taylor:2016aoi for details.

When the CFT dimension dd is odd, the renormalized entanglement entropy can be written in terms of the Euler characteristic and other renormalized curvature invariants of the bulk entangling surface Taylor:2020uwf ,

Sren(B~)(1)n+1nχ(B~)r𝒲r(B~)pp(B~)qq(B~).\displaystyle S_{ren}(\tilde{B})\sim(-1)^{n+1}{\cal F}_{n}\;\chi(\tilde{B})-\sum_{r}{\cal W}_{r}(\tilde{B})-\sum_{p}{\cal H}_{p}(\tilde{B})-\sum_{q}{\cal I}_{q}(\tilde{B}). (10)

where 𝒲r\mathcal{W}_{r} are renormalized integral of projections of the Weyl curvature, p\mathcal{H}_{p} are renormalized integral of even powers of the extrinsic curvature and for d>5d>5 there are q{\cal I}_{q} renormalized integrals containing products of Weyl and extrinsic curvature. More explicitly the renormalized entanglement entropy proportional to the renormalized area of the bulk entangling surface B~\tilde{B}

Sren(B~)=𝒜(B~)4Gd+1\displaystyle S_{ren}(\tilde{B})=\frac{\mathcal{A}(\tilde{B})}{4G_{d+1}} (11)

and renormalized area integral in d=3d=3 is

𝒜(B~)=2πχ(B~)12B~d2xg|K|2B~d2xgW1212,\displaystyle{\cal A}(\tilde{B})=-2\pi\chi(\tilde{B})-\frac{1}{2}\int_{\tilde{B}}d^{2}x\sqrt{g}|{K}|^{2}-\int_{\tilde{B}}d^{2}x\sqrt{g}{W}_{1212}, (12)

and in d=5d=5 is

𝒜(B~)\displaystyle{\cal A}(\tilde{B}) =\displaystyle= 4π23χ(B~)16(B~)13𝒲(B~)\displaystyle\frac{4\pi^{2}}{3}\chi(\tilde{B})-\frac{1}{6}{\cal H}(\tilde{B})-\frac{1}{3}{\cal W}(\tilde{B})
124B~d4xg(H24HμνHμν+HμνρσHμνρσ\displaystyle-\frac{1}{24}\int_{\tilde{B}}d^{4}x\sqrt{g}\left(H^{2}-4H_{\mu\nu}H^{\mu\nu}+H_{\mu\nu\rho\sigma}H^{\mu\nu\rho\sigma}\right.
+4W121224WμnνnW~μnνn+WμνρσWμνρσ).\displaystyle\left.+4{W}_{1212}^{2}-4{W}_{\mu n\nu n}\widetilde{W}^{\mu n\nu n}+{W}_{\mu\nu\rho\sigma}{W}^{\mu\nu\rho\sigma}\right).

In what follows we will find these geometric expressions for renormalized entanglement entropy useful. Note in particular that these will simplify considerably in the context of first variations around AdS backgrounds.

2.2 Hamiltonian Formalism and Charges in AdS

In this section we review the description of Wald Hamiltonians Lee:1990nz ; Wald:1993nt ; Iyer:1994ys ; Iyer:1995kg ; Wald:1999wa and charges in anti-de Sitter spacetimes. Our review follows closely the work of Papadimitriou:2004ap ; Papadimitriou:2005ii , and more details may be found in these references. The Wald approach assumes that the gravitational theory is described by a diffeomorphism covariant Lagrangian d-form 𝐋(ψ){\bf{L}}(\psi), where 𝐋(ψ){\bf{L}}(\psi) will depend both on the metric and other fields, denoted collectively as ψ\psi. In the anti-de Sitter context we work with a renormalized Lagrangian

𝑳ren=𝑳d𝑩\displaystyle\boldsymbol{L}^{ren}=\boldsymbol{L}-{d}\boldsymbol{B} (14)

where 𝑳\boldsymbol{L} is the bulk Lagrangian form and 𝑩\boldsymbol{B} may be viewed as the combination of the Gibbon-Hawking term and boundary counterterms. The onshell regular Lagrangian is exact i.e.

𝑳onshell=d(𝜺anaλ)16πGd+1.\displaystyle\boldsymbol{L}^{onshell}=-\frac{{d}(\boldsymbol{\varepsilon}_{a}n^{a}\lambda)}{16\pi G_{d+1}}. (15)

where nan^{a} is the outward normal in the asymptotic radial direction,

n=dzz,\displaystyle n=-\frac{dz}{z}, (16)

𝜺\boldsymbol{\varepsilon} is the volume form and 𝜺a1an\boldsymbol{\varepsilon}_{a_{1}\dots a_{n}} is a (dn+1)(d-n+1) form

𝜺a1an=1(dn+1)!εa1anb1bdn+1dxb1dxbdn+1,\displaystyle\boldsymbol{\varepsilon}_{a_{1}\dots a_{n}}=\frac{1}{(d-n+1)!}\varepsilon_{a_{1}\dots a_{n}b_{1}\dots b_{d-n+1}}dx^{b_{1}}\wedge\cdots\wedge dx^{b_{d-n+1}}, (17)

with the orientation

εztx1=+g.\displaystyle\varepsilon_{ztx^{1}\cdots}=+\sqrt{-g}. (18)

In the Hamiltonian formalism, fields can be expanded asymptotically near the conformal boundary in series of dilatation eigenfunctions with ascending weight, see appendix B.1 for more detailed explanation. The general structure of the boundary term is then

𝑩\displaystyle\boldsymbol{B} =𝑩GH𝑩ct\displaystyle=\boldsymbol{B}^{GH}-\boldsymbol{B}^{ct}
=𝜺ana16πGd+1(K(Kctλct))\displaystyle=-\frac{\boldsymbol{\varepsilon}_{a}n^{a}}{16\pi G_{d+1}}\left(K-(K_{ct}-\lambda_{ct})\right)
=𝜺ana16πGd+1(K(d)+λct).\displaystyle=\frac{-\boldsymbol{\varepsilon}_{a}n^{a}}{16\pi G_{d+1}}\left(K_{(d)}+\lambda_{ct}\right). (19)

where typically counterterms contribute up to terms at the dthd^{th} order i.e.

Octi<dO(i).\displaystyle O_{ct}\sim\sum_{i<d}O_{(}i). (20)

Variations can then be expressed as

δ𝑳ren\displaystyle\delta\boldsymbol{L}^{ren} =δ𝑳dδ𝑩\displaystyle=\delta\boldsymbol{L}-d\delta\boldsymbol{B} (21)
=𝑬ψδψ+d𝚯[δψ]dδ𝑩\displaystyle=\boldsymbol{E}^{\psi}\delta\psi+d\boldsymbol{\Theta}[\delta\psi]-d\delta\boldsymbol{B} (22)

where 𝑬ψ\boldsymbol{E}^{\psi} denotes the equations of motion and 𝚯\boldsymbol{\Theta} is the symplectic potential. This expression can be rewritten as

δ𝑳ren=𝑬ψδψ+d𝚯ren[δψ]\displaystyle\delta\boldsymbol{L}^{ren}=\boldsymbol{E}^{\psi}\delta\psi+d\boldsymbol{\Theta}^{ren}[\delta\psi] (23)

where the renormalized symplectic potential form can be expressed as

𝚯ren[δψ]=𝜺anaπ(d)μνδγμν\displaystyle\boldsymbol{\Theta}^{ren}[\delta\psi]=\boldsymbol{\varepsilon}_{a}n^{a}\pi_{(d)}^{\mu\nu}\delta\gamma_{\mu\nu} (24)

The canonical momentum πμν\pi_{\mu\nu} can be expressed in terms of the extrinsic curvature as

πμν=116πGd+1γ(KμνKγμν).\displaystyle\pi^{\mu\nu}=-\frac{1}{16\pi G_{d+1}}\sqrt{\gamma}\left(K^{\mu\nu}-K\gamma^{\mu\nu}\right). (25)

If we expand both sides of the equality in the dilatation eigenfunction expansion, we can match the dilatation weights and obtain,

π(n)μν=116πGd+1(K(n)μνK(n)γμν).\displaystyle\pi_{(n)}^{\mu\nu}=-\frac{1}{16\pi G_{d+1}}\left(K^{\mu\nu}_{(n)}-K_{(n)}\gamma^{\mu\nu}\right). (26)

In (24) π(d)μν\pi_{(d)}^{\mu\nu} is the dthd^{th}-term in the dilatation eigenfunction series of the conjugate momentum with respect to the metric and this is in turn related to the renormalized CFT stress tensor as

2π(d)μν=116πGd+1Trenμν,\displaystyle 2\pi_{(d)}^{\mu\nu}=-\frac{1}{16\pi G_{d+1}}T^{\mu\nu}_{ren}, (27)

i.e. the first variation of the renormalized action is

δIrenonshell=132πGd+1ddxγTrenμνδγμν.\displaystyle\delta I^{onshell}_{ren}=\frac{-1}{32\pi G_{d+1}}\int_{\partial\mathcal{M}}d^{d}x\sqrt{-\gamma}T^{\mu\nu}_{ren}\delta\gamma_{\mu\nu}. (28)

Now let us consider the asymptotic behaviour of metric perturbations. Expressing the AdSd+1 metric as

ds2=dz2z2+1z2ημνdxμν\displaystyle ds^{2}=\frac{dz^{2}}{z^{2}}+\frac{1}{z^{2}}\eta_{\mu\nu}dx^{\mu\nu} (29)

where η\eta is the Minkowski metric, only the normalizable mode is allowed to vary under a Dirichlet condition and

δγμν=zd2δγ(d)μν+O(zd1),z0.\displaystyle\delta\gamma_{\mu\nu}=z^{d-2}\delta\gamma_{(d)\;\mu\nu}+O(z^{d-1}),\quad\quad z\rightarrow 0. (30)

For dd odd, using the tracelessness of the stress tensor and absence of trace anomaly, the Dirichlet boundary condition can be automatically generalized to a conformal Dirichlet boundary condition which fixes the conformal class only. In the present of a conformal anomaly, for the onshell action to be stationary under perturbations a representative of the conformal class has to be fixed.


Now let us turn to Noether charges. If the field variation is induced by a vector field ξ\xi, we can define the Noether current form as

𝑱[ξ]=𝚯[δξψ]ιξ𝑳\displaystyle\boldsymbol{J}[\xi]=\boldsymbol{\Theta}[\delta_{\xi}\psi]-\iota_{\xi}\boldsymbol{L} (31)

where ιξ\iota_{\xi} contracts ξ\xi with the first index of 𝑳\boldsymbol{L}. The exterior derivative of the Noether current is proportional to the equation of motion

d𝑱[ξ]=𝑬ψδξψ,\displaystyle d\boldsymbol{J}[\xi]=-\boldsymbol{E}^{\psi}\delta_{\xi}\psi, (32)

and thus vanishes onshell. Hence we can define the Noether charge form 𝑸[ξ]\boldsymbol{Q}[\xi] as the exact term in the Noether current

𝑱[ξ]=d𝑸[ξ]𝑵[ξ]\displaystyle\boldsymbol{J}[\xi]=d\boldsymbol{Q}[\xi]-\boldsymbol{N}[\xi] (33)

where

d𝑵[ξ]=𝑬ψδξψ.\displaystyle d\boldsymbol{N}[\xi]=\boldsymbol{E}^{\psi}\delta_{\xi}\psi. (34)

There is another conserved charge in AlAdSAlAdS induced by ξ\xi called the holographic charge. Using (27)(\ref{eq:pidTren}) and the fact that the renormalized CFT stress tensor is conserved, we can construct the total relativistic momentum of the boundary system. The holographic charge form 𝓠[ξ]\boldsymbol{\mathcal{Q}}[\xi] is defined by

𝓠[ξ]=𝜺abna2π(d)bcξc.\displaystyle\boldsymbol{\mathcal{Q}}[\xi]=-\boldsymbol{\varepsilon}_{ab}n^{a}2\pi_{(d)}^{bc}\xi_{c}. (35)

and this form is integrated over a timeslice at the boundary to obtain the holographic charge. This can be interpreted as the renormalized relativistic momentum along the ξ\xi direction.

In Lemma 4.1 in Papadimitriou:2005ii it was proved that for any asymptotically locally anti-de Sitter space {\mathcal{M}} the two definitions of charges corresponding to asymptotic conformal Killing vector, ξ\xi, on a spatial slice on the conformal boundary, C\partial\mathcal{M}\cap C, are equivalent i.e.

C𝓠[ξ]=CQfull[ξ].\displaystyle-\int_{\partial\mathcal{M}\cap C}\boldsymbol{\mathcal{Q}}[\xi]=\int_{\partial\mathcal{M}\cap C}\textbf{Q}^{full}[\xi]. (36)

where

Qfull[ξ]=Q[ξ]ιξB.\displaystyle\textbf{Q}^{full}[\xi]=\textbf{Q}[\xi]-\iota_{\xi}\textbf{B}. (37)

Note that this equivalence is defined up to exact terms since C\partial\mathcal{M}\cap C is a cycle and the asymptotic conformal Killing vector ξ\xi has the following fall off condition:

ξz=O(zd),\displaystyle\xi^{z}=O(z^{d}), ξμ=ζμ(1+O(zd+2))\displaystyle\xi^{\mu}=\zeta^{\mu}(1+O(z^{d+2})) (38)

where ζ\zeta is a boundary conformal Killing vector. We will later need to generalise this equivalence to less restrictive fall off conditions on the vector field.

2.3 Holographic First Law of Entanglement Entropy

In this section we briefly review the first law of entanglement entropy. Given a reduced density matrix ρB\rho_{B} the modular Hamiltonian HBH_{B} is given by

ρB=eHB\rho_{B}=e^{-H_{B}} (39)

Under a small variation of the entanglement entropy, we obtain the relation

δSB=δHB\displaystyle\delta S_{B}=\delta\expectationvalue{H_{B}} (40)

and the equivalence between δHB\delta\expectationvalue{H_{B}} and the change in energy δE\delta E gives the first law of entanglement entropy.

Following Casini:2011kv , we now review relevant properties of the modular Hamiltonian and modular flow for CFTs on Minkowski space. There is a symmetry group associated with the modular Hamiltonian: the modular group, a group of one-parameter transformations of the form

UB(s)=eisHB\displaystyle U_{B}(s)=e^{-isH_{B}} (41)

where s\partial_{s} is called the modular flow. For QFTs on Minkowski space, the modular flow generates a boost. In null coordinates X±X^{\pm} this is given by

X±(s)=X±e±2πs.\displaystyle X^{\pm}(s)=X^{\pm}e^{\pm 2\pi s}. (42)

For an accelerated observer in Rindler coordinates, the state is thermal in τ\tau where the longitudinal part of the metric is given by

dX+dX=ρ2R2dτ2+dρ2,\displaystyle dX^{+}dX^{-}=-\frac{\rho^{2}}{R^{2}}d\tau^{2}+d\rho^{2}, (43)

where RR relates to the imaginary time periodicity i.e. β=2πR\beta=2\pi R. The thermal density matrix of the state is

ρ=eβHτTr(eβHτ).\displaystyle\rho_{\mathcal{R}}=\frac{e^{\beta H_{\tau}}}{Tr(e^{\beta H_{\tau}})}. (44)

The modular flow generator is 2πRτ2\pi R\partial_{\tau} and the modular Hamiltonian is given by 2πRHτ+logTr(eβHτ)2\pi RH_{\tau}+\log Tr(e^{\beta H_{\tau}}).


For a spatial ball BB of radius RR centred at xi=0,t=0x^{i}=0,\,t=0 on dd-dimensional Minkowski space, we can conformally map the causal development of the spatial ball D(B)D(B) to the Rindler wedge. This conformal map XμxμX^{\mu}\rightarrow x^{\mu} can also map the modular flow (42)(\ref{eq:modflowR}) to

x±(s)=R(R+x±)e2πs(Rx±)(R+x±)+e2πs(Rx±)\displaystyle x^{\pm}(s)=R\frac{(R+x^{\pm})-e^{\mp 2\pi s}(R-x^{\pm})}{(R+x^{\pm})+e^{\mp 2\pi s}(R-x^{\pm})} (45)

and hence the modular flow generator s\partial_{s} as ζB\zeta_{B}

s\displaystyle\partial_{s} =πR((R2t2x2)t2txii)\displaystyle=\frac{\pi}{R}\left((R^{2}-t^{2}-\vec{x}^{2})\partial_{t}-2tx^{i}\partial_{i}\right) (46)
ζB\displaystyle\zeta_{B} =iπR(R2Pt+Kt)\displaystyle=\frac{i\pi}{R}(R^{2}P_{t}+K_{t}) (47)

where PtP_{t} and KtK_{t} are the time translation and special conformal transformation generators, respectively.

Since the modular Hamiltonian is the translation operator in ss, on BB we get

HB=Bdd1xTts.\displaystyle H_{B}=\int_{B}d^{d-1}xT^{ts}. (48)

We can write (48)(\ref{eq:HBs}) in covariant form as

HB=B𝑑σμTμνζBν\displaystyle H_{B}=\int_{B}d\sigma^{\mu}T_{\mu\nu}\zeta_{B}^{\nu} (49)

and the modular energy as

EB=B𝑑σμTμνζBν.\displaystyle E_{B}=\int_{B}d\sigma^{\mu}\expectationvalue{T_{\mu\nu}}\zeta_{B}^{\nu}. (50)

The entanglement entropy of region BB can be calculated holographically by the area of the corresponding bulk entangling surface B~\tilde{B}.

A CFT in the vacuum state on the causal wedge D(B)D(B) can also be mapped conformally to a CFT in a thermal state on the hyperbolic cylinder. This can be easily seen from writing the Rindler metric as:

ds2=ρ2R2(dτ2+R2ρ2(dρ2+dXidXi))=ρ2R2ds×d12.\displaystyle ds^{2}=\frac{\rho^{2}}{R^{2}}\left(-d\tau^{2}+\frac{R^{2}}{\rho^{2}}(d\rho^{2}+dX^{i}dX^{i})\right)=\frac{\rho^{2}}{R^{2}}ds^{2}_{\mathbb{R}\times\mathbb{H}^{d-1}}. (51)

As discussed in Faulkner:2013ica , the first law of entanglement entropy of the CFT thermal state on hyperbolic cylinder can be related to the first law of black hole dynamics via holography. Essentially, the CFT on hyperbolic cylinder ×d1\mathbb{R}\times\mathbb{H}^{d-1} is dual to the Rindler AdSd+1AdS_{d+1} black hole exterior and the bulk entangling surface B~\tilde{B} can be viewed as the black hole horizon. The perturbation of entanglement entropy δSB\delta S_{B} is equal to the perturbation of black hole entropy calculated from the Wald functional δSWald\delta S_{Wald} where

SWald=2π𝑑σδδRcdabnabncd.\displaystyle S_{Wald}=-2\pi\int_{\mathcal{H}}d\sigma\frac{\delta\mathcal{L}}{\delta R^{ab}_{\;\;cd}}n^{ab}n_{cd}. (52)

Changing back to the interpretation in terms of a Minkowski boundary, we label Σ\Sigma as the bulk region enclosed by BB and B~\tilde{B} and the bulk causal wedge of Σ\Sigma as D(Σ)D(\Sigma). We extend the boundary modular flow to the bulk as the Killing vector

ξB=2πR(tt0)[zz+(xix0i)i]+πR[R2z2(tt0)2(xx0)2]t.\displaystyle\xi_{B}=-\frac{2\pi}{R}(t-t_{0})[z\partial_{z}+(x^{i}-x^{i}_{0})\partial_{i}]+\frac{\pi}{R}[R^{2}-z^{2}-(t-t_{0})^{2}-(\vec{x}-\vec{x}_{0})^{2}]\partial_{t}. (53)

Note that this Killing vector does not satisfy (38)(\ref{eq:falloff}) but instead has the weaker fall-off behaviour

ξBz=O(z),\displaystyle\xi^{z}_{B}=O(z), ξBμ=ζBμ(1+O(z2)).\displaystyle\xi^{\mu}_{B}=\zeta^{\mu}_{B}(1+O(z^{2})). (54)

One can check ξB\xi_{B} vanishes on B~\tilde{B}.

It was shown in Faulkner:2013ica that for metric perturbations limited to normalizable modes, δgμν=zd2hμν(d)\delta g_{\mu\nu}=z^{d-2}h^{(d)}_{\mu\nu}, one get the infinitesimal first law of entanglement entropy as

δTtt\displaystyle\delta\expectationvalue{T_{tt}\ } =d212πΩd2limR0(1RdδSB)\displaystyle=\frac{d^{2}-1}{2\pi\Omega_{d-2}}\lim_{R\rightarrow 0}\left(\frac{1}{R^{d}}\delta S_{B}\right)
δTtt\displaystyle\delta\expectationvalue{T_{tt}} =d16πGd+1htt(d)\displaystyle=\frac{d}{16\pi G_{d+1}}h^{(d)}_{tt} (55)
δTμν\displaystyle\delta\expectationvalue{T_{\mu\nu}} =d16πGd+1hμν(d)\displaystyle=\frac{d}{16\pi G_{d+1}}h^{(d)}_{\mu\nu}

where the tracelessness condition of hμν(d)h^{(d)}_{\mu\nu} is used to go from the second line to the final covariant expression. The final expression matches the holographic dictionary between stress tensor and normalizable metric coefficient found in deHaro2001 .

The covariant first law of entanglement entropy utilises the charges associated with with energy and entropy corresponding to the bulk Killing vector ξB\xi_{B} introduced in section 2.2. The entanglement entropy is

SBgrav=B~Qfull[ξB]\displaystyle S_{B}^{grav}=\int_{\tilde{B}}\textbf{Q}^{full}[\xi_{B}] (56)

and the modular energy is

EBgrav=B𝓠[ξB].\displaystyle E_{B}^{grav}=-\int_{B}\boldsymbol{\mathcal{Q}}[\xi_{B}]. (57)

Limiting to the variation involving only the normalizable mode with no boundary variation on B=B~\partial B=\partial\tilde{B} and using (36)(\ref{eq:Lemma4.1}),

Bδ𝓠[ξB]=BδQfull[ξB].\displaystyle-\int_{B}\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]=\int_{B}\delta\textbf{Q}^{full}[\xi_{B}]. (58)

The off-shell difference is expressed in terms of the Einstein equations

δEBgravδSBgrav\displaystyle\delta E_{B}^{grav}-\delta S_{B}^{grav} =Bδ𝑸full[ξB]B~δ𝑸full[ξB]\displaystyle=\int_{{B}}\delta\boldsymbol{Q}^{full}[\xi_{B}]-\int_{\tilde{B}}\delta\boldsymbol{Q}^{full}[\xi_{B}] (59)
=Σ𝑑δQfull[ξB]=ΣδJfull[ξB]=Σ2𝜺aδEabξBa,\displaystyle=\int_{\Sigma}d\delta\textbf{Q}^{full}[\xi_{B}]=\int_{\Sigma}\delta\textbf{J}^{full}[\xi_{B}]=\int_{\Sigma}-2\boldsymbol{\varepsilon}^{a}\delta E_{ab}\xi^{a}_{B},

hence recovering the first law of entanglement entropy onshell. We also obtain the version of (36)(\ref{eq:Lemma4.1})

Bδ𝓠[ξB]=B~δ𝑸full[ξB].\displaystyle-\int_{B}\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]=\int_{\tilde{B}}\delta\boldsymbol{Q}^{full}[\xi_{B}]. (60)

Note there are many caveats regarding boundary terms and fall-off condition when we allow the variation of the non-normalizable modes. We shall address them in the following sections.

3 Infinitesimal Renormalized First Law

In this section we will discuss the renormalized version of the first law of entanglement entropy in the infinitesimal limit, for =AdSd+1\mathcal{M}=AdS_{d+1} with spherical boundary entangling surfaces B~=Sd2\partial\tilde{B}=S^{d-2} in d6d\leq 6. We begin by collecting together expressions for the renormalized entanglement entropy of such spherical regions. We derive the infinitesimal renormalized first law of entanglement entropy in AlAdSd+1AlAdS_{d+1} for odd dd and explain its connection with the variation of the renormalized integral of a curvature invariant. Since the renormalized entanglement entropy in even dimensions is scheme dependent, we postpone the proof of the generalized first law in even dd to section 4.3.2 to avoid repetitions.

3.1 Spherical Entangling Region in AdS

The metric of AdSd+1AdS_{d+1} on the Poincare patch may be parameterized as

ds2=dz2z2+1z2gμνdxμdxν,ds^{2}=\frac{dz^{2}}{z^{2}}+\frac{1}{z^{2}}g_{\mu\nu}dx^{\mu}dx^{\nu}, (61)

where gμν=ημνg_{\mu\nu}=\eta_{\mu\nu} is flat Minkowski metric with signature (-,++,\cdots,++). In the case of spherical entangling regions, the (d1)(d-1)-dimensional bulk extending entangling surface B~\tilde{B} with boundary B~\partial\tilde{B} as the entangling surface of the boundary CFT can be described by

r2+z2=R2r^{2}+z^{2}=R^{2} (62)

where rr is the radial coordinate on the boundary and RR is the radius of the spherical entangling region. The induced metric on the entangling surface in AdSd+1AdS_{d+1} is then

ds2=R2z2r2dz2+r2z2dΩd22.ds^{2}=\frac{R^{2}}{z^{2}r^{2}}dz^{2}+\frac{r^{2}}{z^{2}}d\Omega_{d-2}^{2}. (63)

where dΩd22d\Omega_{d-2}^{2} is the standard unit sphere metric. Another convenient choice of coordinates (w,u)(w,\,u) are defined by

r=wcosu,\displaystyle r=w\cos u, z=wsinu\displaystyle z=w\sin u (64)

so the AdSd+1AdS_{d+1} metric can be written as

ds2=1w2sin2u(dw2dt2)+1sin2udz2+cos2usin2udΩd22\displaystyle ds^{2}=\frac{1}{w^{2}\sin^{2}u}(dw^{2}-dt^{2})+\frac{1}{\sin^{2}u}dz^{2}+\frac{\cos^{2}u}{\sin^{2}u}d\Omega_{d-2}^{2} (65)

and the induced metric on B~\tilde{B} becomes

ds2=1sin2udz2+cos2usin2udΩd22.\displaystyle ds^{2}=\frac{1}{\sin^{2}u}dz^{2}+\frac{\cos^{2}u}{\sin^{2}u}d\Omega_{d-2}^{2}. (66)

The regularised bulk contribution to the entanglement entropy for such an entangling surface is then

SBreg\displaystyle S^{reg}_{B} =14Gd+1B~ϵdd1xγ\displaystyle=\frac{1}{4G_{d+1}}\int_{\tilde{B}_{\epsilon}}d^{d-1}x\sqrt{\gamma}
=Ωd24Gd+1ϵR𝑑zRrd3zd1\displaystyle=\frac{\Omega_{d-2}}{4G_{d+1}}\int_{\epsilon}^{R}dz\,\frac{Rr^{d-3}}{z^{d-1}} (67)
=Ωd24Gd+1ξπ2𝑑ucosd2usind1u\displaystyle=\frac{\Omega_{d-2}}{4G_{d+1}}\int_{\xi}^{\frac{\pi}{2}}du\,\frac{\cos^{d-2}u}{\sin^{d-1}u} (68)

where Ωd2\Omega_{d-2} is the area of (d2d-2)-dimensional unit sphere and RsinξϵR\sin\xi\equiv\epsilon.

The divergent contributions are of the form ϵn\epsilon^{-n} except in even dd where there are extra logarithmic terms. Focussing first on odd dd, from (9)(\ref{eq:renhee}) we know the counterterms for d<6d<6 are

SBct\displaystyle S^{ct}_{B} =14(d2)Gd+1B~ϵdd2xγ~[1+1(d2)(d4)(aa12k2d2(d1))]\displaystyle=\frac{1}{4(d-2)G_{d+1}}\int_{\partial\tilde{B}_{\epsilon}}d^{d-2}x\sqrt{\widetilde{\gamma}}\left[1+\frac{1}{(d-2)(d-4)}(\mathcal{R}_{aa}-\frac{1}{2}k^{2}-\frac{d}{2(d-1)}\mathcal{R})\right] (69)
=Ωd24(d2)Gd+1rd2ϵd2[1(d2)ϵ22(d4)r2].\displaystyle=\frac{\Omega_{d-2}}{4(d-2)G_{d+1}}\frac{r^{d-2}}{\epsilon^{d-2}}\left[1-\frac{(d-2)\epsilon^{2}}{2(d-4)r^{2}}\right].

Note that the intrinsic curvature terms do not contribute here since the boundary metric is flat, but they will contribute to the variation of the entanglement entropy under metric perturbations later. For odd dd, using the definition of the entangling surface one obtains counterterm contributions

SBct=Ωd24Gd+1[Rd2(d2)ϵd2Rd4(d4)ϵd4+].\displaystyle S^{ct}_{B}=\frac{\Omega_{d-2}}{4G_{d+1}}\left[\frac{R^{d-2}}{(d-2)\epsilon^{d-2}}-\frac{R^{d-4}}{(d-4)\epsilon^{d-4}}+\cdots\right]. (70)

Combing (70)(\ref{eq:Sctoddd}) with (67)(\ref{eq:IntzSreg}) we get

d=3:Sren=π2G4\displaystyle d=3:\quad\quad\quad\quad S_{ren}=-\frac{\pi}{2G_{4}} (71)
d=5:Sren=π23G6.\displaystyle d=5:\quad\quad\quad\quad S_{ren}=\frac{\pi^{2}}{3G_{6}}.

For d=4d=4, the regularized entanglement entropy from (67)(\ref{eq:IntzSreg}) is

SBreg=Ω24G5[lnR2+ln(R2)2+R(R2ϵ2)122ϵ2+lnϵ2ln(R2+R(R2ϵ2)12)2]\displaystyle S^{reg}_{B}=\frac{\Omega_{2}}{4G_{5}}\left[-\frac{\ln R}{2}+\frac{\ln(R^{2})}{2}+\frac{R(R^{2}-\epsilon^{2})^{\frac{1}{2}}}{2\epsilon^{2}}+\frac{\ln\epsilon}{2}-\frac{\ln(R^{2}+R(R^{2}-\epsilon^{2})^{\frac{1}{2}})}{2}\right] (72)

and the corresponding the full set of counterterms, including the logarithmic counterterm, gives

SBct\displaystyle S^{ct}_{B} =18G5B~d2xγ~lnϵ16G5B~d2xγ~(aa12k223)\displaystyle=\frac{1}{8G_{5}}\int_{\partial\tilde{B}}d^{2}x\sqrt{\widetilde{\gamma}}-\frac{\ln\epsilon}{16G_{5}}\int_{\partial\tilde{B}}d^{2}x\sqrt{\widetilde{\gamma}}(\mathcal{R}_{aa}-\frac{1}{2}k^{2}-\frac{2}{3}\mathcal{R})
=Ω28G5[r2ϵ2+lnϵ].\displaystyle=\frac{\Omega_{2}}{8G_{5}}\left[\frac{r^{2}}{\epsilon^{2}}+{\ln\epsilon}\right]. (73)

Combining these we obtain the renormalized entanglement entropy in d=4d=4

SBren=Ω24G5[ln2R2+14R].\displaystyle S^{ren}_{B}=\frac{\Omega_{2}}{4G_{5}}\left[-\frac{\ln 2R}{2}+\frac{1}{4R}\right]. (74)

Since the action in this case has logarithmic counterterms there is an intrinsic scheme dependence in the renormalised entanglement entropy, which is completely determined by the scheme chosen for the renormalization of the action.

3.2 First Law and Variation of Modular Energy

We now consider the variation of the entanglement entropy under a linear perturbation of the bulk metric. We will express the perturbed metric in radial gauge so that

ds2=dz2z2+1z2(ημν+hμν)dxμdxν,ds^{2}=\frac{dz^{2}}{z^{2}}+\frac{1}{z^{2}}(\eta_{\mu\nu}+h_{\mu\nu})dx^{\mu}dx^{\nu}, (75)

A general perturbation hμνh_{\mu\nu} can be expanded near the conformal boundary as

hμν=hμν(0)+z2hμν(2)++zdhμν(d)+zdlogzh~μν(d)+h_{\mu\nu}=h^{(0)}_{\mu\nu}+z^{2}h^{(2)}_{\mu\nu}+\cdots+z^{d}h^{(d)}_{\mu\nu}+z^{d}\log z\tilde{h}^{(d)}_{\mu\nu}+\cdots (76)

where the logarithmic terms arise in even dd and all coefficients in the expansion can be expressed in terms of the pair of data (hμν(0),hμν(d))(h^{(0)}_{\mu\nu},h^{(d)}_{\mu\nu}) using the Einstein equations.

The goal of this section is to show that the change in the renormalized entropy δSBren\delta S_{B}^{ren} under such metric perturbations is equal to the change in modular energy i.e.

δEB=δSBren.\displaystyle\delta E_{B}=\delta S_{B}^{ren}. (77)

In the previous literature Faulkner:2013ica , the first law was derived by restricting the variation of metric to only normalizable modes i.e. imposing hμν(0)=0h_{\mu\nu}^{(0)}=0 with hμν(d)0h_{\mu\nu}^{(d)}\neq 0. Accordingly, the change in the entanglement entropy δSB\delta S_{B} is finite even without including the counterterms. Here we will derive the first law for general perturbations for which hμν(0)h^{(0)}_{\mu\nu} is not necessarily zero; from a QFT perspective a general bulk metric perturbation corresponds to changing the background for the dual QFT as well as the state in the theory.

We will first demonstrate the renormalized first law in the infinitesimal limit where the radius of the boundary entangling region BB tends to zero R0R\rightarrow 0. The modular energy may be approximated by

δEB\displaystyle\delta E_{B} =B𝑑σμδTμνrenξBν\displaystyle=\int_{B}d\sigma^{\mu}\delta T_{\mu\nu}^{ren}\xi_{B}^{\nu} (78)
δEB\displaystyle\delta E_{B} R02πRdΩd2d21δTttren.\displaystyle\stackrel{{\scriptstyle\scalebox{0.5}{$R\rightarrow 0$}}}{{\scalebox{1.3}{$\longrightarrow$}}}\frac{2\pi R^{d}\Omega_{d-2}}{d^{2}-1}\delta T_{tt}^{ren}.

From holographic renormalization deHaro2001 , the variation of the renormalized energy momentum tensor for odd dd is

δTμνren=d16πGd+1hμν(d).\displaystyle\delta T_{\mu\nu}^{ren}=\frac{d}{16\pi G_{d+1}}h^{(d)}_{\mu\nu}. (79)

In even dimensions the relation between the renormalized stress tensor and the coefficients of the asymptotic expansion is more complicated, capturing the conformal anomalies. For example, in d=4d=4

δTμνren=116πG5(4hμν(4)+6h~μν(4)),\displaystyle\delta T_{\mu\nu}^{ren}=\frac{1}{16\pi G_{5}}\left(4h^{(4)}_{\mu\nu}+6\tilde{h}^{(4)}_{\mu\nu}\right), (80)

i.e. there is an additional contribution associated with the coefficient of the logarithmic term h~(4)\tilde{h}^{(4)}. At linearized order we can express h~(4)\tilde{h}^{(4)} in terms of the curvature R(0)R^{(0)} of the perturbation of the QFT metric h(0)h^{(0)} as

h~μν(4)=148μνR(0)+116ρρRμν(0)196(ρρR(0))ημν.\tilde{h}^{(4)}_{\mu\nu}=-\frac{1}{48}\partial_{\mu}\partial_{\nu}R^{(0)}+\frac{1}{16}\partial^{\rho}\partial_{\rho}R^{(0)}_{\mu\nu}-\frac{1}{96}(\partial^{\rho}\partial_{\rho}R^{(0)})\eta_{\mu\nu}. (81)

The infinitesimal first law of entanglement entropy for general variation is thus equivalent to showing that the variation of renormalized entanglement entropy can be expressed in terms of the renormalized stress tensor as

δSBren=2πRdΩd2d21δTttren\displaystyle\delta S^{ren}_{B}=\frac{2\pi R^{d}\Omega_{d-2}}{d^{2}-1}\delta T_{tt}^{ren} (82)

3.3 Infinitesimal First Law for odd dd

We shall focus on odd dd. The linearized variation of regularized entanglement entropy can be expressed in Cartesian spatial coordinates as

δSBreg=R8Gd+1B~ϵdd1x1zd(hiix^ix^jhij),\displaystyle\delta S^{reg}_{B}=\frac{R}{8G_{d+1}}\int_{\tilde{B}_{\epsilon}}d^{d-1}x\frac{1}{z^{d}}(h_{ii}-\hat{x}^{i}\hat{x}^{j}h_{ij}), (83)

where ii runs over the spatial indices of the dd-dimensional Minkowski space.

To obtain the infinitesimal version of the first law, we consider the limit R0R\rightarrow 0. To evaluate the integrals explicitly it is more convenient to use the (w,u)(w,\,u) coordinates in (64)(\ref{eq:wcoord}), in terms of which the variation of regularized entanglement entropy is

δSBreg=18Gd+1ξπ/2𝑑uSd2𝑑Ωd2cosd2usind1u(δijcos2ux^ix^j)hij.\delta S^{reg}_{B}=\frac{1}{8G_{d+1}}\int_{\xi}^{\pi/2}du\int_{S^{d-2}}d\Omega_{d-2}\frac{\cos^{d-2}{u}}{\sin^{d-1}{u}}(\delta^{ij}-\cos^{2}{u}\hat{x}^{i}\hat{x}^{j})h_{ij}. (84)

For the variation of the counterterms we need the linearized variation of the spatial extrinsic curvature, which can be expressed as

δK2=(d2)z2rx^ix^jhij+z2r(hiix^ix^jhij)\displaystyle\delta K_{2}=-\frac{(d-2)z}{2r}\hat{x}^{i}\hat{x}^{j}h_{ij}+\frac{z}{2}\partial_{r}(h_{ii}-\hat{x}^{i}\hat{x}^{j}h_{ij}) (85)

and the variation of a specific combination of Ricci tensors,

δtt+δrrd2(d1)δ=(d2)(hii(2)x^ix^jhij(2)).\displaystyle-\delta\mathcal{R}_{tt}+\delta\mathcal{R}_{rr}-\frac{d}{2(d-1)}\delta\mathcal{R}=(d-2)\left(h^{(2)}_{ii}-\hat{x}^{i}\hat{x}^{j}h^{(2)}_{ij}\right). (86)

The latter equality holds at linearized level, see equation (91) below.

Substituting the above expressions into the variation of (69)(\ref{eq:IntSct}) we get the following expression for the counterterms in general d6d\leq 6 to first order:

δSBct\displaystyle\delta S^{ct}_{B} =14(d2)Gd+1Sd2dΩd2[12rd2ϵd2(hiix^ix^jhij)(d2)4(d4)rd4ϵd4(hii3x^ix^jhij)\displaystyle=\frac{1}{4(d-2)G_{d+1}}\int_{S^{d-2}}d\Omega_{d-2}\bigg{[}\frac{1}{2}\frac{r^{d-2}}{\epsilon^{d-2}}\Big{(}h_{ii}-\hat{x}^{i}\hat{x}^{j}h_{ij}\Big{)}-\frac{(d-2)}{4(d-4)}\frac{r^{d-4}}{\epsilon^{d-4}}\Big{(}h_{ii}-3\hat{x}^{i}\hat{x}^{j}h_{ij}\Big{)}
+1(d4)rd2ϵd4(hii(2)x^ix^jhij(2))12(d4)rd3ϵd4(x^kkhiix^ix^jx^kkhij)].\displaystyle+\frac{1}{(d-4)}\frac{r^{d-2}}{\epsilon^{d-4}}\Big{(}h^{(2)}_{ii}-\hat{x}^{i}\hat{x}^{j}h^{(2)}_{ij}\Big{)}-\frac{1}{2(d-4)}\frac{r^{d-3}}{\epsilon^{d-4}}\Big{(}\hat{x}^{k}\partial_{k}h_{ii}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\partial_{k}h_{ij}\Big{)}\bigg{]}. (87)

In the (w,u)(w,u) coordinate system, the area integral for the bulk entangling surface B~ϵ\tilde{B}_{\epsilon} is expressed in terms of an integral over the asymptotic angular coordinate uu and the spatial angular coordinates. We can thus evaluate the integral up to the upper limit u=π2u=\frac{\pi}{2} even when expanding around R=0R=0.

The Taylor expansion around xi=0x^{i}=0 at each order of the Fefferman-Graham expansion can be written as:

hμν(n)(xi)=\displaystyle h^{(n)}_{\mu\nu}(x^{i})= hμν(n)(0)+Rx^iihμν(n)(0)+R2x^ix^j2!ijhμν(n)(0)+R3x^ix^jx^k3!ijkhμν(n)(0)\displaystyle\;h^{(n)}_{\mu\nu}(0)+R\hat{x}^{i}\partial_{i}h^{(n)}_{\mu\nu}(0)+\frac{R^{2}\hat{x}^{i}\hat{x}^{j}}{2!}\partial_{i}\partial_{j}h^{(n)}_{\mu\nu}(0)+\frac{R^{3}\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}}{3!}\partial_{i}\partial_{j}\partial_{k}h^{(n)}_{\mu\nu}(0)
+R4x^ix^jx^kx^l4!ijklhμν(n)(0)+\displaystyle+\frac{R^{4}\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}}{4!}\partial_{i}\partial_{j}\partial_{k}\partial_{l}h^{(n)}_{\mu\nu}(0)+\dotsc (88)

The Fefferman-Graham expansion also becomes an expansion in RR,

hμν=hμν(0)+R2sin2uhμν(2)+\displaystyle h_{\mu\nu}=h_{\mu\nu}^{(0)}+R^{2}\sin^{2}{u}h_{\mu\nu}^{(2)}+\cdots (89)

We can now expand (84)(\ref{eq:IntSregwcoord}) using (89)(\ref{eq:FGu}) and (88)(\ref{eq:xi0exp}) up to RdR^{d}. The two angular integrals du,dΩd2du,\,d\Omega_{d-2} can be evaluated independently for each term in the expansion. In the appendix we give a general formula (215)(\ref{eq:OmegaxiInt}) for integrating over products of unit vectors over SnS^{n}. Together with (218)(\ref{eq:deltaspartialsh}), generic terms in the expansion after the spatial angular dΩd2d\Omega_{d-2} integral take the form

(2)mhii(n),(2)mijhij(n).\displaystyle(\partial^{2})^{m}h^{(n)}_{ii},\quad\quad\quad\quad(\partial^{2})^{m}\partial_{i}\partial_{j}h^{(n)}_{ij}. (90)

All the non-normalizable modes are related to the first term in the Fefferman-Graham expansion h(0)h^{(0)} through the Einstein equation deHaro2001 . For h(2)h^{(2)} we have

hμν(2)=1d2(δμν12(d1)δημν)\displaystyle h^{(2)}_{\mu\nu}=-\frac{1}{d-2}\Big{(}\delta\mathcal{R}_{\mu\nu}-\frac{1}{2(d-1)}\delta\mathcal{R}\eta_{\mu\nu}\Big{)} (91)

The linear variation of the Ricci tensor is,

δμν=12μνh(0)σσ+σ(μh(0)ν)σ12σσhμν(0)\delta\mathcal{R}_{\mu\nu}=-\frac{1}{2}\partial_{\mu}\partial_{\nu}{h^{(0)}}^{\sigma}_{\sigma}+\partial_{\sigma}\partial_{(\mu}{h^{(0)}}^{\sigma}_{\nu)}-\frac{1}{2}\partial^{\sigma}\partial_{\sigma}h^{(0)}_{\mu\nu} (92)

We can use the above information to express h(2)h^{(2)} in terms of derivatives of h(0)h^{(0)} and h(4)h^{(4)} in terms of derivatives of h(2)h^{(2)} as

hii(2)=12(d2)(iihjj(0)ijhij(0)).\displaystyle h^{(2)}_{ii}=\frac{1}{2(d-2)}\left(\partial_{i}\partial_{i}h^{(0)}_{jj}-\partial_{i}\partial_{j}h^{(0)}_{ij}\right). (93)

In d=5d=5 we will also require the following relation between h(4)h^{(4)} and h(2)h^{(2)},

hii(4)=14jjhii(2)14ijhij(2).\displaystyle h^{(4)}_{ii}=\frac{1}{4}\partial_{j}\partial_{j}h^{(2)}_{ii}-\frac{1}{4}\partial_{i}\partial_{j}h^{(2)}_{ij}. (94)

We can follow appendix A.2 and A.3 to obtain

δSBren=dRdΩd28(d1)(d+1)Gd+1hii(d)\displaystyle\delta S^{ren}_{B}=\frac{dR^{d}\Omega_{d-2}}{8(d-1)(d+1)G_{d+1}}h^{(d)}_{ii} (95)

Here by working with the renormalized quantities we recover the first law of entanglement entropy for general linearized variations of the metric, including both non-normalizable and normalizable modes.

3.4 Curvature Invariants Formula

The first variation of the entanglement entropy around spherical entangling regions in AdSd+1 with dd odd can be expressed in a particularly simple and elegant geometric form, using the expression for the renormalized entanglement entropy in terms of curvature and topological invariants (10). Since such variations do not change the topology of the entangling surface, the topological Euler invariant contribution does not change. All contributions from the extrinsic curvature are quadratic or higher order; since the extrinsic curvature vanishes to leading order, and this means the contributions p{\cal H}_{p} do not contribute to first variations (but do contribute to second variations). By analogous reasoning, the only contribution from the Weyl terms 𝒲r{\cal W}_{r} comes from the term that is linear in the Weyl tensor. Thus we arrive at

δSren14G2nδ𝒲\displaystyle\delta S^{ren}\propto-\frac{1}{4G_{2n}}\delta{\cal W} (96)

where G2nG_{2n} is the Newton constant (with 2n=d+12n=d+1) and

δ𝒲=Σd2(n1)xgδW1212Σd2n3xhδW1212+\displaystyle\delta{\cal W}=\int_{\Sigma}d^{2(n-1)}x\sqrt{g}\;\delta W_{1212}-\int_{\partial\Sigma}d^{2n-3}x\sqrt{h}\;\delta W_{1212}+\cdots (97)

where δW1212\delta W_{1212} is the pullback of the normal components of the bulk linearized Weyl curvature in an orthonormal frame and δW1212\delta W_{1212} is the pullback of the normal components of the boundary linearized Weyl curvature in an orthonormal frame. The boundary terms are such that δ𝒲\delta{\cal W} is a finite conformal invariant for a generic non-normalizable metric perturbation. Note that the boundary term vanishes for AdS4. The ellipses denote additional boundary terms expressed in terms of higher powers of the boundary Weyl curvature that are required for n>3n>3.

The variation of renormalized entanglement entropy for d=3,5d=3,5 is

d=3:\displaystyle d=3: δSren(B~)=14G4δ𝒲(B~)\displaystyle\quad\quad\quad\quad\delta S^{ren}(\tilde{B})=-\frac{1}{4G_{4}}\delta{\cal W}(\tilde{B}) (98)
d=5:\displaystyle d=5: δSren(B~)=112G6δ𝒲(B~)\displaystyle\quad\quad\quad\quad\delta S^{ren}(\tilde{B})=-\frac{1}{12G_{6}}\delta{\cal W}(\tilde{B}) (99)

In Poincaré coordinates the linear variation of the Weyl tensor δWabcd\delta W_{abcd} is,

δWμνρσ\displaystyle\delta W_{\mu\nu\rho\sigma} =1z2[η+h]μνρσ+12z3(hμρηνσ+hνσημρhμσηνρhνρημσ)\displaystyle=\frac{1}{z^{2}}\mathcal{R}[\eta+h]_{\mu\nu\rho\sigma}+\frac{1}{2z^{3}}(h_{\mu\rho}^{\prime}\eta_{\nu\sigma}+h_{\nu\sigma}^{\prime}\eta_{\mu\rho}-h_{\mu\sigma}^{\prime}\eta_{\nu\rho}-h_{\nu\rho}^{\prime}\eta_{\mu\sigma}) (100)
δWμνρz\displaystyle\delta W_{\mu\nu\rho z} =12z2[μhνρνhμρ]\displaystyle=\frac{1}{2z^{2}}[\partial_{\mu}h_{\nu\rho}^{\prime}-\partial_{\nu}h_{\mu\rho}^{\prime}] (101)
δWμzνz\displaystyle\delta W_{\mu z\nu z} =12z2hμν′′+12z3hμν\displaystyle=-\frac{1}{2z^{2}}h_{\mu\nu}^{\prime\prime}+\frac{1}{2z^{3}}h_{\mu\nu}^{\prime} (102)

where [η+h]μνρσ\mathcal{R}[\eta+h]_{\mu\nu\rho\sigma} is the Riemann tensor for boundary metric ημν+hμν\eta_{\mu\nu}+h_{\mu\nu}. In Poincaré coordinates, the two unit normals are

n1\displaystyle n_{1} =zz\displaystyle=z\frac{\partial}{\partial z} (103)
n2\displaystyle n_{2} =zr2+z2(zz+rx^ixi)\displaystyle=\frac{z}{\sqrt{r^{2}+z^{2}}}\left(z\frac{\partial}{\partial z}+r\hat{x}^{i}\frac{\partial}{\partial x^{i}}\right) (104)

Then the projection of Weyl tensor onto NB~N\tilde{B}, δW1212\delta W_{1212}, is

δW1212=z4r2+z2(z2δWtztz+r2x^ix^jδWtitj+2zrx^iδWtzti)\displaystyle\delta{W}_{1212}=\frac{z^{4}}{r^{2}+z^{2}}\left(z^{2}\delta W_{tztz}+r^{2}\hat{x}^{i}\hat{x}^{j}\delta W_{titj}+2zr\hat{x}^{i}\delta W_{tzti}\right) (105)

The bulk Weyl integral becomes

B~dd1xγW1212=ξπ2𝑑uSd2𝑑Ωd2R2cosd2usind5u(z2Wtztz+r2x^ix^jWtitj+2zrx^iWtzti)\displaystyle\int_{\tilde{B}}d^{d-1}x\sqrt{\gamma}\;W_{1212}=\int^{\frac{\pi}{2}}_{\xi}du\int_{S^{d-2}}d\Omega_{d-2}\frac{R^{2}\cos^{d-2}u}{\sin^{d-5}u}\left(z^{2}W_{tztz}+r^{2}\hat{x}^{i}\hat{x}^{j}W_{titj}+2zr\hat{x}^{i}W_{tzti}\right) (106)

and the boundary Weyl integral is

B~dd2xγ~W1212=Sd2𝑑Ωd2R2cosd2usind6u(z2Wtztz+r2x^ix^jWtitj+2zrx^iWtzti)\displaystyle\int_{\partial\tilde{B}}d^{d-2}x\sqrt{\widetilde{\gamma}}\;W_{1212}=\int_{S^{d-2}}d\Omega_{d-2}\frac{R^{2}\cos^{d-2}u}{\sin^{d-6}u}\left(z^{2}W_{tztz}+r^{2}\hat{x}^{i}\hat{x}^{j}W_{titj}+2zr\hat{x}^{i}W_{tzti}\right) (107)

Substituting (100)(102)(\ref{eq:dW1})-(\ref{eq:dW3}) into the above integrals

B~dd1xγδW1212=\displaystyle\int_{\tilde{B}}d^{d-1}x\sqrt{\gamma}\;\delta W_{1212}= ξπ2duSd2dΩd2R2(cosd2usind5u[12htt′′+12Rsinuhtt]\displaystyle\int^{\frac{\pi}{2}}_{\xi}du\int_{S^{d-2}}d\Omega_{d-2}R^{2}\bigg{(}\frac{\cos^{d-2}u}{\sin^{d-5}u}\big{[}-\frac{1}{2}h_{tt}^{\prime\prime}+\frac{1}{2R\sin u}h_{tt}^{\prime}\big{]} (108)
+cosdusind3ux^ix^j[titj+12Rsinu(httηij+hijηtt)]\displaystyle+\frac{\cos^{d}u}{\sin^{d-3}u}\hat{x}^{i}\hat{x}^{j}\Big{[}\mathcal{R}_{titj}+\frac{1}{2R\sin u}(h_{tt}^{\prime}\eta_{ij}+h_{ij}^{\prime}\eta_{tt})\Big{]}
+cosd1usind4ux^i[thtiihtt])\displaystyle+\frac{\cos^{d-1}u}{\sin^{d-4}u}\hat{x}^{i}[\partial_{t}h_{ti}^{\prime}-\partial_{i}h_{tt}^{\prime}]\bigg{)}

and

B~dd2xγ~δW1212=\displaystyle\int_{\tilde{B}}d^{d-2}x\sqrt{\widetilde{\gamma}}\;\delta{W}_{1212}= Sd2dΩd2R2(cosd2usind6u[12htt′′+12Rsinuhtt]\displaystyle\int_{S^{d-2}}d\Omega_{d-2}R^{2}\bigg{(}\frac{\cos^{d-2}u}{\sin^{d-6}u}\big{[}-\frac{1}{2}h_{tt}^{\prime\prime}+\frac{1}{2R\sin u}h_{tt}^{\prime}\big{]} (109)
+cosdusind4ux^ix^j[titj+12Rsinu(httηij+hijηtt)]\displaystyle+\frac{\cos^{d}u}{\sin^{d-4}u}\hat{x}^{i}\hat{x}^{j}\Big{[}\mathcal{R}_{titj}+\frac{1}{2R\sin u}(h_{tt}^{\prime}\eta_{ij}+h_{ij}^{\prime}\eta_{tt})\Big{]}
+cosd1usind5ux^i[thtiihtt])\displaystyle+\frac{\cos^{d-1}u}{\sin^{d-5}u}\hat{x}^{i}[\partial_{t}h_{ti}^{\prime}-\partial_{i}h_{tt}^{\prime}]\bigg{)}

where =z{}^{\prime}=\frac{\partial}{\partial z} Up to order RdR^{d}, the relevant components of the integrand are obtained by Taylor expanding about the origin and eliminating the odd components as the it is integrated over Sd2S^{d-2}. In appendix A.4, we expand titj\mathcal{R}_{titj} into linear perturbation hμνh_{\mu\nu} then further relate the higher order non-normalizable modes hμν(n<d)h_{\mu\nu}^{(n<d)} to the lower order non-normalizable modes via the Einstein equation. Finally, we can see all the lower order non-normalizable modes perturbation are cancelled and the renormalized Weyl integral is

d=3:\displaystyle d=3: δ𝒲=3R3Ω116G4htt(3)\displaystyle\quad\quad\quad\quad\delta\mathcal{W}=-\frac{3R^{3}\Omega_{1}}{16G_{4}}h^{(3)}_{tt} (110)
d=5:\displaystyle d=5: δ𝒲=5R5Ω316G6htt(5).\displaystyle\quad\quad\quad\quad\delta\mathcal{W}=-\frac{5R^{5}\Omega_{3}}{16G_{6}}h^{(5)}_{tt}. (111)

Then substituting (110),(111)(\ref{eq:dWintd3}),(\ref{eq:dWintd5}) into (98),(99)(\ref{eq:dSdWd3}),(\ref{eq:dSdWd5}) to get the renormalized entanglement entropy. We recovered the infinitesimal first law of entanglement entropy in (82)(\ref{eq:hddSren}) for variation that includes perturbation of non-normalizable modes,

d=3:\displaystyle d=3: δSBren=3R3Ω148G4htt(3)\displaystyle\quad\quad\quad\quad\delta S^{ren}_{B}=\frac{3R^{3}\Omega_{1}}{48G_{4}}h^{(3)}_{tt} (112)
d=5:\displaystyle d=5: δSBren=5R5Ω3192G6htt(5)\displaystyle\quad\quad\quad\quad\delta S^{ren}_{B}=\frac{5R^{5}\Omega_{3}}{192G_{6}}h^{(5)}_{tt} (113)

3.5 Cancellation of Divergences in d=4d=4

We now turn from odd dimensional boundaries to even dimensions and show how the cancellation of divergences of the renormalized entanglement entropy works in d=4d=4. A general perturbation of the boundary metric hμνh_{\mu\nu} can be expanded around the boundary z=0z=0

hμν=hμν(0)(r,θ,ϕ)+z2hμν(2)(r,θ,ϕ)+h_{\mu\nu}=h^{(0)}_{\mu\nu}(r,\theta,\phi)+z^{2}h^{(2)}_{\mu\nu}(r,\theta,\phi)+\dotsi (114)

Since on B~\tilde{B} the coordinate rr is a function of zz. The coefficient in the expansion of the metric perturbation can be further expanded around r=Rr=R. For hμν(0)(r,θ,ϕ)h^{(0)}_{\mu\nu}(r,\theta,\phi) the expansion is

hμν(0)(r,θ,ϕ)\displaystyle h^{(0)}_{\mu\nu}(r,\theta,\phi) =hμν(0)(R,θ,ϕ)+(rR)rhμν(0)(R,θ,ϕ)+\displaystyle=h^{(0)}_{\mu\nu}(R,\theta,\phi)+(r-R)\partial_{r}h^{(0)}_{\mu\nu}(R,\theta,\phi)+\cdots (115)
=hμν(0)(R,θ,ϕ)z22Rrhμν(0)(R,θ,ϕ)+\displaystyle=h^{(0)}_{\mu\nu}(R,\theta,\phi)-\frac{z^{2}}{2R}\partial_{r}h^{(0)}_{\mu\nu}(R,\theta,\phi)+\cdots (116)

So the variation of the regularized entanglement entropy in polar coordinates for d=4d=4 is,

δSBreg=\displaystyle\delta S^{reg}_{B}= 18G5ϵRdzS2dΩ2[1z3(hθθ(0)+1sin2θhΦΦ(0))+1z(12R2hθθ(0)12R2sin2θhϕϕ(0)\displaystyle\frac{1}{8G_{5}}\int^{R}_{\epsilon}dz\int_{S^{2}}d\Omega_{2}\bigg{[}\frac{1}{z^{3}}\Big{(}h^{(0)}_{\theta\theta}+\frac{1}{\sin^{2}{\theta}}h^{(0)}_{\Phi\Phi}\Big{)}+\frac{1}{z}\Big{(}-\frac{1}{2R^{2}}h^{(0)}_{\theta\theta}-\frac{1}{2R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}
+hrr(0)+1R2hθθ(0)+1R2sin2θhϕϕ(0)12Rrhθθ(0)12Rsin2θrhϕϕ(0)+hθθ(2)+1sin2θhϕϕ(2))]\displaystyle+h^{(0)}_{rr}+\frac{1}{R^{2}}h^{(0)}_{\theta\theta}+\frac{1}{R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}-\frac{1}{2R}\partial_{r}h^{(0)}_{\theta\theta}-\frac{1}{2R\sin^{2}\theta}\partial_{r}h^{(0)}_{\phi\phi}+h^{(2)}_{\theta\theta}+\frac{1}{\sin^{2}\theta}h^{(2)}_{\phi\phi}\Big{)}\bigg{]} (117)

Evaluating the zz integral and the divergent terms are,

(δSBreg)div=\displaystyle(\delta S^{reg}_{B})^{div}= 18G5S2dΩ2[12ϵ2(hθθ(0)+1sin2θhΦΦ(0))lnϵ(12R2hθθ(0)12R2sin2θhϕϕ(0)\displaystyle\frac{1}{8G_{5}}\int_{S^{2}}d\Omega_{2}\bigg{[}\frac{1}{2\epsilon^{2}}\Big{(}h^{(0)}_{\theta\theta}+\frac{1}{\sin^{2}{\theta}}h^{(0)}_{\Phi\Phi}\Big{)}-\ln\epsilon\Big{(}-\frac{1}{2R^{2}}h^{(0)}_{\theta\theta}-\frac{1}{2R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}
+hrr(0)+1R2hθθ(0)+1R2sin2θhϕϕ(0)12Rrhθθ(0)12Rsin2θrhϕϕ(0)+hθθ(2)+1sin2θhϕϕ(2))]\displaystyle+h^{(0)}_{rr}+\frac{1}{R^{2}}h^{(0)}_{\theta\theta}+\frac{1}{R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}-\frac{1}{2R}\partial_{r}h^{(0)}_{\theta\theta}-\frac{1}{2R\sin^{2}\theta}\partial_{r}h^{(0)}_{\phi\phi}+h^{(2)}_{\theta\theta}+\frac{1}{\sin^{2}\theta}h^{(2)}_{\phi\phi}\Big{)}\bigg{]} (118)

We can see that (118)(\ref{eq:dSregdivd4}) is identical to (293)(\ref{eq:dSctdivd4}) so the divergences of the regularized entanglement entropy will be removed by the counterterms in the renormalization procedure.

(δSBreg)div=(δSBct)div.\displaystyle(\delta S^{reg}_{B})^{div}=(\delta S^{ct}_{B})^{div}. (119)

More explicitly, in Cartesian coordinate, the set of counterterms from (9)(\ref{eq:renhee}) is

δSBct\displaystyle\delta S^{ct}_{B} =116G5S2dΩ2[(hiix^ix^jhij)(r2ϵ2+lnϵ)\displaystyle=\frac{1}{16G_{5}}\int_{S^{2}}d\Omega_{2}\bigg{[}(h_{ii}-\hat{x}^{i}\hat{x}^{j}h_{ij})(\frac{r^{2}}{\epsilon^{2}}+\ln\epsilon)
+lnϵ(δtt+δrr23δ)lnϵ(2hii+1rr(r2hiixixjhij))].\displaystyle+\ln\epsilon\Big{(}-\delta\mathcal{R}_{tt}+\delta\mathcal{R}_{rr}-\frac{2}{3}\delta\mathcal{R}\Big{)}-\ln\epsilon\Big{(}-2h_{ii}+\frac{1}{r}\partial_{r}(r^{2}h_{ii}-x^{i}x^{j}h_{ij})\Big{)}\bigg{]}. (120)

Following section A.5 we get

δSBct\displaystyle\delta S^{ct}_{B} =116G5S2dΩ2r2ϵ2(hiix^ix^jhij)+lnϵ[(hii(0)x^ix^jhij(0))2R2(hii(2)x^ix^jhij(2))\displaystyle=\frac{1}{16G_{5}}\int_{S^{2}}d\Omega_{2}\frac{r^{2}}{\epsilon^{2}}\big{(}h_{ii}-\hat{x}^{i}\hat{x}^{j}h_{ij}\big{)}+\ln\epsilon\bigg{[}\big{(}h^{(0)}_{ii}-\hat{x}^{i}\hat{x}^{j}h^{(0)}_{ij}\big{)}-2R^{2}\big{(}h^{(2)}_{ii}-\hat{x}^{i}\hat{x}^{j}h^{(2)}_{ij}\big{)}
(rx^kkhii(0)+2x^ix^jhij(0)+rx^ix^jx^kkhij(0))].\displaystyle-\big{(}-r\hat{x}^{k}\partial_{k}h^{(0)}_{ii}+2\hat{x}^{i}\hat{x}^{j}h^{(0)}_{ij}+r\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\partial_{k}h^{(0)}_{ij}\big{)}\bigg{]}. (121)

Note that there are finite contributions from the first term in (121)(\ref{eq:ExdSCtd4}).

4 Integral Renormalized First Law

Under general variations of the boundary metric where both the non-normalizable and normalizable modes are not fixed, we need to modify the relation between the conserved charges (36)(\ref{eq:Lemma4.1}) and the associated first law. Since the spatial slice Σ\Sigma where the charges are defined has a boundary, we cannot neglect the total derivative terms. In fact the boundary terms capture all the divergent behaviour of the Noether charge and act as counterterms.

As mentioned in the section 2.2, the asymptotic conformal Killing vector used to define the Noether charges in Papadimitriou:2005ii has to follow the fall off condition (38)(\ref{eq:falloff}) which our modular flow generator ξB\xi_{B} in (53)(\ref{eq:xiB}) does not satisfy. We shall see later that all these extra terms are essential to match the universal divergences of the entanglement entropy.

Charges defined on asymptotic boundary B~\tilde{B} and entangling surface BB have asymptotic behaviours analogous to the entanglement entropy. In (d+1)(d+1) even spacetime dimensions, the finite charges are universal. For (d+1)(d+1) odd spacetime dimensions, the finite parts are scheme dependent, and change covariantly under changes of scheme. Hence, the first law of entanglement entropy in odd bulk dimensions requires appropriate finite counterterms.

4.1 Charges in the Entangling Region

The Noether charge form 𝑸[ξ]\boldsymbol{Q}[\xi] is the exact term in the conserved current form induced by the vector ξ\xi. For pure Einstein gravity (with or without cosmological constant), it can be expressed as

𝑸[ξ]\displaystyle\boldsymbol{Q}[\xi] =116πGNdξ\displaystyle=-\frac{1}{16\pi G_{N}}\star d\xi
=116πGN𝜺abaξb,\displaystyle=-\frac{1}{16\pi G_{N}}\boldsymbol{\varepsilon}_{ab}\nabla^{a}\xi^{b}, (122)

up to exact terms. Since the extra exact terms will introduce boundary terms in the integral over BB and B~\tilde{B} respectively, we will treat (122)(\ref{eq:NCQ}) as the definition of 𝑸[ξ]\boldsymbol{Q}[\xi] to avoid confusion. In asymptotically locally AdS, the full expression for the Noether charge form is then written as

𝑸full[ξ]=𝑸[ξ]ιξ𝑩\displaystyle\boldsymbol{Q}^{full}[\xi]=\boldsymbol{Q}[\xi]-\iota_{\xi}\boldsymbol{B} (123)

with 𝑩\boldsymbol{B} defined as

𝑩=18πGN𝜺ana(K(d)+λct)\displaystyle\boldsymbol{B}=-\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{a}n^{a}\left(K_{(d)}+\lambda_{ct}\right) (124)

where nn is the radial unit normal pointing outwards from the asymptotic boundary \partial\mathcal{M}.


The holographic charge form 𝓠[ξ]\boldsymbol{\mathcal{Q}}[\xi] is defined in terms of the dthd^{th} term in the dilatation eigenfunction expansion of the canonical momentum, π(d)bc\pi_{(d)}^{bc}, through the following expression

𝓠[ξ]=𝜺abna2π(d)bcξc.\displaystyle\boldsymbol{\mathcal{Q}}[\xi]=-\boldsymbol{\varepsilon}_{ab}n^{a}2\pi_{(d)}^{bc}\xi_{c}. (125)

In our setup, the full Noether current form 𝑱full[ξB]\boldsymbol{J}^{full}[\xi_{B}] is induced by the bulk modular flow of a bulk Killing vector ξB\xi_{B}. The full Noether charge on the spatial slice Σϵ\Sigma_{\epsilon} can be thought of as the charge captured by the surface from the current:

Qfull[ξB]\displaystyle{Q}^{full}[\xi_{B}] =Σϵ𝑱full[ξB]\displaystyle=\int_{\Sigma_{\epsilon}}\boldsymbol{J}^{full}[\xi_{B}]
=Σϵ𝚯[δξBϕ]ιξB𝑳onshell\displaystyle=\int_{\Sigma_{\epsilon}}\boldsymbol{\Theta}[\delta_{\xi_{B}}\phi]-\iota_{\xi_{B}}\boldsymbol{L}^{onshell}
=ΣϵιξB𝑳onshell\displaystyle=-\int_{\Sigma_{\epsilon}}\iota_{\xi_{B}}\boldsymbol{L}^{onshell} (126)

As shown in (33)(\ref{eq:dQ}), the onshell Noether current form is exact.

In (56,57)(\ref{eq:EntropyInt},\ref{eq:ModEInt}), we defined the bulk entanglement entropy by an integral of the Noether charge form over B~ϵ\tilde{B}_{\epsilon} and the modular energy through an integral of the holographic charge form over BϵB_{\epsilon}. In order to relate the two we need to generalize (36)(\ref{eq:Lemma4.1}) to

Bϵ𝑸full[ξ]𝚫[ξ]=Bϵ𝓠[ξ].\displaystyle\int_{B_{\epsilon}}\boldsymbol{Q}^{full}[\xi]-\boldsymbol{\Delta}[\xi]=-\int_{B_{\epsilon}}\boldsymbol{\mathcal{Q}}[\xi]. (127)

Here 𝚫\boldsymbol{\Delta} captures the counterterms associated with renormalizing the divergences of 𝑸full\boldsymbol{Q}^{full}; this term is needed as the quantity on the righthandside, 𝓠{\boldsymbol{\mathcal{Q}}}, is renormalized. We could redefine the Noether charge on the lefthandside to include these counterterms, but in what follows we keep track of the contributions separately to emphasise how the counterterm contributions arise.

The counterterms need to be included here because of our more general falloff conditions for the perturbations. This contribution vanishes in Papadimitriou:2005ii because of the stricter fall-off condition of ξ\xi which makes the radial derivative of ξ\xi vanishes and the counterterms integrate to zero as the integral is over a surface with no boundary. In Faulkner:2013ica this term vanishes due to the falloff conditions imposed on the metric perturbations.

The conserved charge forms 𝑸full\boldsymbol{Q}^{full} and 𝓠\boldsymbol{\mathcal{Q}} can be interpreted as Hamiltonian potentials, as explained in detail in appendix C. 𝚫\boldsymbol{\Delta} in this context is the difference of the counterterm contributions of the two Hamiltonian potentials. In the covariant phase space formalism, given an action with boundary terms, one can obtain the presymplectic current through variation of the Lagrangian and boundary terms. The presymplectic current maps the vector field in the configuration space to the Hamiltonian potential.

We are interested in renormalized quantities and there are two ways to see how the counterterms arise in the Hamiltonian potential. The first approach is to use the renormalized action that includes the counterterms, and then obtain the presymplectic current and Hamiltonian potential. We denote this as the full Hamiltonian potential because it is equal to the full Noether charge form when π(d)μν=0\pi_{(d)}^{\mu\nu}=0

𝑯full[ξ]\displaystyle\boldsymbol{H}^{full}[\xi] =𝑸[ξ]+𝒃GH[ξ]𝒃ct[ξ]\displaystyle=\boldsymbol{Q}[\xi]+\boldsymbol{b}^{GH}[\xi]-\boldsymbol{b}^{ct}[\xi] (128)
=𝑯GH[ξ]𝒃ct[ξ]\displaystyle=\boldsymbol{H}^{GH}[\xi]-\boldsymbol{b}^{ct}[\xi] (129)
=𝑸full[ξ],\displaystyle=\boldsymbol{Q}^{full}[\xi], (130)

where 𝒃GH\boldsymbol{b}^{GH} and 𝒃ct\boldsymbol{b}^{ct} represent the Gibbon-Hawking boundary term and counterterm contribution. Here 𝑯GH\boldsymbol{H}^{GH} is Hamiltonian potential obtained from the action that only includes the Gibbon-Hawking boundary term.

The second way to see how the counterterms arise is to renormalize the Gibbon-Hawking Hamiltonian potential 𝑯GH\boldsymbol{H}^{GH} directly by subtracting the lower order terms in the dilatation eigenvalue expansion. The renormalized Gibbon-Hawking Hamiltonian potential is given in terms of the holographic charge form when π(d)μν=0\pi_{(d)}^{\mu\nu}=0

𝑯(d)GH[ξ]\displaystyle\boldsymbol{H}^{GH}_{(d)}[\xi] =𝑯GH[ξ]𝑯ctGH[ξ]\displaystyle=\boldsymbol{H}^{GH}[\xi]-\boldsymbol{H}^{GH}_{ct}[\xi] (131)
=𝓠[ξ].\displaystyle=-\boldsymbol{\mathcal{Q}}[\xi]. (132)

Hence we can interpret 𝚫\boldsymbol{\Delta} as the difference of the two aforementioned Hamiltonian potentials

𝚫[ξ]=𝑯ctGH[ξ]𝒃ct[ξ]\displaystyle\boldsymbol{\Delta}[\xi]=\boldsymbol{H}^{GH}_{ct}[\xi]-\boldsymbol{b}^{ct}[\xi] (133)

As we will see this term in the perturbed case this is exact and represents the counterterms for the entanglement entropy.


On BϵB_{\epsilon}, for bulk Killing vector, ξB\xi_{B} the full Noether charge form is

𝑸full[ξB]|Bϵ=18πGN𝜺ztnz(zzξBt+KμtξBμK(d)ξBtλctξBt)\displaystyle\boldsymbol{Q}^{full}[\xi_{B}]|_{B{\epsilon}}=-\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}n^{z}\left(-z\partial_{z}\xi^{t}_{B}+K_{\mu}^{t}\xi^{\mu}_{B}-K_{(d)}\xi_{B}^{t}-\lambda_{ct}\xi^{t}_{B}\right) (134)

where the first term on the right hand side was neglected in Papadimitriou:2005ii due to the falloff condition (38)(\ref{eq:falloff}). The holographic charge form is

𝓠[ξB]|Bϵ\displaystyle\boldsymbol{\mathcal{Q}}[\xi_{B}]|_{B{\epsilon}} =𝜺ztnz2π(d)tcξBc\displaystyle=-\boldsymbol{\varepsilon}_{zt}n^{z}2\pi_{(d)}^{tc}\xi_{B\;c}
=18πGN𝜺ztnz(K(d)taK(d)γta)ξBa\displaystyle=\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}n^{z}\left(K^{ta}_{(d)}-K_{(d)}\gamma^{ta}\right)\xi_{B\;a} (135)

where we used (26)(\ref{eq:pin}) to express the holographic charge in terms of the dthd^{th} term in the dilatation eigenfunction expansion of the extrinsic curvature. The difference in the charges 𝚫[ξB]\boldsymbol{\Delta}[\xi_{B}] is

𝚫[ξB]|Bϵ=18πGN𝜺ztnz(zzξBt+KctμtξBμλctξBt).\displaystyle\boldsymbol{\Delta}[\xi_{B}]|_{B{\epsilon}}=-\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}n^{z}\left(-z\partial_{z}\xi^{t}_{B}+K_{ct\,\mu}^{\;\;t}\xi^{\mu}_{B}-\lambda_{ct}\xi^{t}_{B}\right). (136)

It is important to remember that this expression is only valid when ξB\xi_{B} is Killing. We shall see later the perturbed difference of the charges δΔ[ξB]\delta\Delta[\xi_{B}] admits an extra term as ξB\xi_{B} is no longer Killing. In appendix B.1, we follow Papadimitriou:2004ap and derive the explicit dilatation eigenfunction expansion for KνμK^{\mu}_{\nu} and λ\lambda. In the unperturbed setting, the boundary metric is dd-dimensional Minkowski metric ημν\eta_{\mu\nu}. Only the zeroth term in the dilatation eigenfunction expansion is non-vanishing,

K(0)νμ=δνμ,\displaystyle K^{\;\;\mu}_{(0)\,\nu}=\delta^{\mu}_{\nu}, λ(0)=1.\displaystyle\lambda_{(0)}=1. (137)

From (135)(\ref{eq:MQinK}) we know the holographic charge is zero

𝓠[ξB]|Bϵ=0.\displaystyle\boldsymbol{\mathcal{Q}}[\xi_{B}]|_{B{\epsilon}}=0. (138)

Then 𝚫[ξB]\boldsymbol{\Delta}[\xi_{B}] simply equals the Noether charge

𝚫[ξB]|Bϵ=𝑸[ξB]|Bϵ\displaystyle\boldsymbol{\Delta}[\xi_{B}]|_{B{\epsilon}}=\boldsymbol{Q}[\xi_{B}]|_{B{\epsilon}} =18πGN𝜺ztnz(zzξBt)\displaystyle=-\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}n^{z}\left(-z\partial_{z}\xi^{t}_{B}\right)
=z34GN𝜺zt\displaystyle=\frac{z^{3}}{4G_{N}}\boldsymbol{\varepsilon}_{zt} (139)

We can now turn our attention to the Noether charge form on the entangling surface B~ϵ\tilde{B}{\epsilon}. As explained in section 2.3, the integral of the Noether charge form over B~ϵ\tilde{B}_{\epsilon} can be interpreted as both the entropy of the Rindler black hole and the entanglement entropy of boundary region BϵB_{\epsilon}. Since ξB\xi_{B} vanishes on B~ϵ\tilde{B}_{\epsilon},

𝑸full[ξB]|B~ϵ\displaystyle\boldsymbol{Q}^{full}[\xi_{B}]|_{\tilde{B}_{\epsilon}} =𝑸[ξB]|B~ϵ\displaystyle=\boldsymbol{Q}[\xi_{B}]|_{\tilde{B}_{\epsilon}}
=18πGN𝜺wtwξBt\displaystyle=\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{wt}\partial^{w}\xi_{B}^{t} (140)

where we used the ww coordinate in (64)(\ref{eq:wcoord}) and the Killing condition. Integrating over B~ϵ\tilde{B}_{\epsilon},

Bϵ𝑸full[ξB]\displaystyle\int_{B_{\epsilon}}\boldsymbol{Q}^{full}[\xi_{B}] =18πGNBϵ𝑑u𝑑Ωd2cosd2usind1u2πw(z2+r2)\displaystyle=\frac{1}{8\pi G_{N}}\int_{B_{\epsilon}}du\,d\Omega_{d-2}\frac{\cos^{d-2}u}{\sin^{d-1}u}\frac{2\pi}{w}\left(z^{2}+r^{2}\right)
=R4GNBϵ𝑑u𝑑Ωd2cosd2usind1u\displaystyle=\frac{R}{4G_{N}}\int_{B_{\epsilon}}du\,d\Omega_{d-2}\frac{\cos^{d-2}u}{\sin^{d-1}u}
=SBreg\displaystyle=S^{reg}_{B} (141)

where we use (68)(\ref{eq:SIntu}) to identify the second line with the regulated entanglement entropy.

4.2 Variation of Charges

The variations of 𝑸full\boldsymbol{Q}^{full} and 𝓠\boldsymbol{\mathcal{Q}} differ from the previous literature Faulkner:2013ica when we allow variations of the non-normalizable modes. For general perturbations of γμν\gamma_{\mu\nu}, the linear variation of the Noether charge form is

δ𝑸[ξB]\displaystyle\delta\boldsymbol{Q}[\xi_{B}] =116πGNδ[𝜺abaξBb]\displaystyle=\frac{-1}{16\pi G_{N}}\delta\left[\boldsymbol{\varepsilon}_{ab}\nabla^{a}\xi_{B}^{b}\right]
=116πGN𝜺ab[δγ2aξBbδgaccξBb+gacδΓdcbξBd]\displaystyle=\frac{-1}{16\pi G_{N}}\boldsymbol{\varepsilon}_{ab}\left[\frac{\delta\gamma}{2}\nabla^{a}\xi_{B}^{b}-\delta g^{ac}\nabla_{c}\xi_{B}^{b}+g^{ac}\delta\Gamma^{b}_{dc}\xi_{B}^{d}\right]
=z28RGN[𝜺ti(xihkkxjhij(R2z2x2)thit)\displaystyle=\frac{-z^{2}}{8RG_{N}}\bigg{[}\boldsymbol{\varepsilon}_{ti}\left(x^{i}h_{kk}-x^{j}h_{ij}-(R^{2}-z^{2}-\vec{x}^{2})\partial_{t}h_{it}\right) (142)
+𝜺tz(zhkk+(R2z2x2)(2zhtt+zhtt))]\displaystyle\quad\quad\quad\quad+\boldsymbol{\varepsilon}_{tz}\left(zh_{kk}+(R^{2}-z^{2}-\vec{x}^{2})(-\frac{2}{z}h_{tt}+\partial_{z}h_{tt})\right)\bigg{]}

Using the coordinates (64)(\ref{eq:wcoord}) on B~ϵ\tilde{B}_{\epsilon} we get the integral

B~ϵδ𝑸[ξB]\displaystyle\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}] =18RGNB~ϵdd1x1zd(R2hkkxixjhij)\displaystyle=\frac{1}{8RG_{N}}\int_{\tilde{B}_{\epsilon}}d^{d-1}x\frac{1}{z^{d}}(R^{2}h_{kk}-x^{i}x^{j}h_{ij})
=18RGNB~ϵdd1x(R2x2)d2(R2hkkxixjhij)\displaystyle=\frac{1}{8RG_{N}}\int_{\tilde{B}_{\epsilon}}d^{d-1}x(R^{2}-\vec{x}^{2})^{-\frac{d}{2}}(R^{2}h_{kk}-x^{i}x^{j}h_{ij}) (143)

which is equal to the linear variation of the holographic entanglement entropy

B~ϵδ𝑸[ξB]\displaystyle\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}] =δSBreg\displaystyle=\delta S^{reg}_{B} (144)

For the variation of the full Noether charge form, we need to evaluate the boundary term δ𝑩\delta\boldsymbol{B}. This term is related to the presymplectic form 𝚯[δϕ]\boldsymbol{\Theta}[\delta\phi] by

𝚯[δϕ]\displaystyle\boldsymbol{\Theta}[\delta\phi] =𝒅𝒅𝒙8πGNδ(γλ)\displaystyle=\frac{\boldsymbol{d^{d}x}}{8\pi G_{N}}\delta\left(-\sqrt{\gamma}\lambda\right) (145)
=𝒅𝒅𝒙[18πGNδ(γK)+πμνδγμν]\displaystyle=\boldsymbol{d^{d}x}\left[\frac{1}{8\pi G_{N}}\delta\left(-\sqrt{\gamma}K\right)+\pi^{\mu\nu}\delta\gamma_{\mu\nu}\right]
=𝒅𝒅𝒙[18πGNδ(γ(K(d)+λct))+π(d)μνδγμν]\displaystyle=\boldsymbol{d^{d}x}\left[\frac{1}{8\pi G_{N}}\delta\left(-\sqrt{\gamma}\left(K_{(d)}+\lambda_{ct}\right)\right)+\pi_{(d)}^{\mu\nu}\delta\gamma_{\mu\nu}\right]
=δ𝑩+𝜺ϵπ(d)μνδγμν.\displaystyle=\delta\boldsymbol{B}+\boldsymbol{\varepsilon}_{\scriptscriptstyle\partial\mathcal{M}_{\epsilon}}\pi_{(d)}^{\mu\nu}\delta\gamma_{\mu\nu}.

where the d-form 𝒅𝒅𝒙\boldsymbol{d^{d}x} is

𝒅𝒅𝒙=1d!dx0dxd1.\displaystyle\boldsymbol{d^{d}x}=\frac{1}{d!}dx^{0}\wedge\cdots\wedge dx^{d-1}. (146)

The variation of the full Noether charge form is then

δ𝑸full[ξB]\displaystyle\delta\boldsymbol{Q}^{full}[\xi_{B}] =δ𝑸[ξB]ιξBδ𝑩\displaystyle=\delta\boldsymbol{Q}[\xi_{B}]-\iota_{\xi_{B}}\delta\boldsymbol{B} (147)
=δ𝑸[ξB]ιξB𝚯[δϕ]ιξB𝜺ϵπ(d)μνδγμν.\displaystyle=\delta\boldsymbol{Q}[\xi_{B}]-\iota_{\xi_{B}}\boldsymbol{\Theta}[\delta\phi]-\iota_{\xi_{B}}\boldsymbol{\varepsilon}_{\scriptscriptstyle\partial\mathcal{M}_{\epsilon}}\pi_{(d)}^{\mu\nu}\delta\gamma_{\mu\nu}.

The linear variation of the holographic charge is

δ𝓠[ξ]|Bϵ\displaystyle\delta\boldsymbol{\mathcal{Q}}[\xi]|_{B_{\epsilon}} =𝜺ztnz2δπ(d)ttξBt.\displaystyle=-\boldsymbol{\varepsilon}_{zt}n^{z}2\delta\pi_{(d)\,t}^{\;\;t}\xi_{B}^{t}. (148)

This is related to the renormalized boundary energy momentum tensor via

2δπ(d)μν=δTrenμν.\displaystyle 2\delta\pi_{(d)}^{\mu\nu}=-\delta T_{ren}^{\mu\nu}. (149)

Substituting this expression into (148)(\ref{eq:dHCQ}) the integral of the variation of holographic charge form δ𝓠[ξB]\delta\boldsymbol{\mathcal{Q}}[\xi_{B}] on the boundary ball region BϵB_{\epsilon} is equal to the variation of modular energy

Bϵδ𝓠[ξB]=δEB.\displaystyle-\int_{{B}_{\epsilon}}\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]=\delta E_{B}. (150)

To express the variation of modular energy in terms of dilatation eigenfunction expansion of extrinsic curvature we vary (25)(\ref{eq:ConMom}) to obtain

δπ(d)νμ\displaystyle\delta\pi_{(d)\,\nu}^{\;\;\mu} =116πGN(δK(d)νμδK(d)δνμ)\displaystyle=\frac{-1}{16\pi G_{N}}\left(\delta K^{\;\;\mu}_{(d)\,\nu}-\delta K_{(d)}\delta^{\mu}_{\nu}\right) (151)
δπ(d)tt\displaystyle\delta\pi_{(d)\,t}^{\;\;t} =116πGNδK(d)ii.\displaystyle=\frac{1}{16\pi G_{N}}\delta K_{(d)\,i}^{\;\;i}. (152)

Using the tracelessness of δK(d)νμ\delta K_{(d)\,\nu}^{\;\;\mu} at the linear level we can write δ𝓠[ξ]\delta\boldsymbol{\mathcal{Q}}[\xi] on BϵB_{\epsilon} as

δ𝓠[ξB]|Bϵ\displaystyle\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]|_{B_{\epsilon}} =18πGN𝜺ztnzδK(d)ttξBt.\displaystyle=\frac{1}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}n^{z}\delta K_{(d)\,t}^{\;\;t}\xi_{B}^{t}. (153)

(This expression holds for all dd, with conformal anomalies present if we write out K(d)ttK_{(d)\,t}^{\;\;t} in terms of gμν(n)g^{(n)}_{\mu\nu} and g~μν(d)\tilde{g}^{(d)}_{\mu\nu}, see for example (186)(\ref{eq:K4ind4}).) The variation of the full Noether charge form δ𝑸full\delta\boldsymbol{Q}^{full} on BϵB_{\epsilon} is

δ𝑸full[ξB]|Bϵ=z8πGN𝜺zt[\displaystyle\delta\boldsymbol{Q}^{full}[\xi_{B}]|_{B{\epsilon}}=\frac{z}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}\bigg{[} (δγ2(KμtξBμ1zzξBt)12ztξBzδγtt+ξBtδKtt)\displaystyle\left(\frac{\delta\gamma}{2}(K^{t}_{\mu}\xi_{B}^{\mu}-\frac{1}{z}\partial^{z}\xi_{B}^{t})-\frac{1}{2z}\partial_{t}\xi_{B}^{z}\delta\gamma^{tt}+\xi_{B}^{t}\delta K^{t}_{t}\right) (154)
ξBt(δγ2+δK(d)+δλct)].\displaystyle-\xi_{B}^{t}\left(\frac{\delta\gamma}{2}+\delta K_{(d)}+\delta\lambda_{ct}\right)\bigg{]}.

By using (151)(\ref{eq:dConMom}), we can obtain the relation between δ𝑸full\delta\boldsymbol{Q}^{full} and δ𝓠\delta\boldsymbol{\mathcal{Q}}.

Similarly to (127)(\ref{eq:New4.1}), this revised version of (60)(\ref{eq:dLemma4.1}) receives a contribution δ𝚫[ξB]\delta\boldsymbol{\Delta}[\xi_{B}]. We get

Bϵδ𝑸full[ξB]=Bϵδ𝓠[ξB]+δ𝚫[ξB].\displaystyle\int_{B_{\epsilon}}\delta\boldsymbol{Q}^{full}[\xi_{B}]=\int_{B_{\epsilon}}-\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]+\delta\boldsymbol{\Delta}[\xi_{B}]. (155)

The latter term takes the form

δ𝚫[ξB]=z8πGN𝜺zt[12zzξBtδγ12ztξBzδγtt+ξBt(δKttδλ)ct].\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}]=\frac{z}{8\pi G_{N}}\boldsymbol{\varepsilon}_{zt}\left[-\frac{1}{2z}\partial^{z}\xi_{B}^{t}\delta\gamma-\frac{1}{2z}\partial_{t}\xi_{B}^{z}\delta\gamma^{tt}+\xi_{B}^{t}\left(\delta K^{t}_{t}-\delta\lambda\right)_{ct}\right]. (156)

Note that we can understand why this term arises for two reasons. Firstly, ξB\xi_{B} is no longer Killing with respect to the perturbed metric and secondly ξB\xi_{B} has a weaker falloff condition (54)(\ref{eq:newfalloff}) instead of (38)(\ref{eq:falloff}). Here we use an abbreviated notation:

δγ=γμνδγμν,\displaystyle\delta\gamma=\gamma^{\mu\nu}\delta\gamma_{\mu\nu}, δKνμ=δ(γμσKσν).\displaystyle\delta K^{\mu}_{\nu}=\delta\left(\gamma^{\mu\sigma}K_{\sigma\nu}\right). (157)

In terms of Hamiltonian potentials, the δ𝚫\delta\boldsymbol{\Delta} term is

δ𝚫[ξB]=δ𝑯ctGH[ξB]δ𝒃ct[ξB].\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}]=\delta\boldsymbol{H}^{GH}_{ct}[\xi_{B}]-\delta\boldsymbol{b}^{ct}[\xi_{B}]. (158)

We further describe the origin of each term in the appendix C and expressed δ𝚫\delta\boldsymbol{\Delta} in (366)(\ref{eq:dDeltaexp}) using the formalism of 2020HarlowWu .

Substituting the unperturbed flat boundary metric and the bulk Killing vector we obtain

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =dd1xγ8πGN[πz2R(htt+hii)+πz2Rhtt+π(R2z2x2)R(δKttδλ)ct]\displaystyle=\frac{d^{d-1}x\,\sqrt{-\gamma}}{8\pi G_{N}}\left[\frac{\pi z^{2}}{R}(-h_{tt}+h_{ii})+\frac{\pi z^{2}}{R}h_{tt}+\frac{\pi(R^{2}-z^{2}-\vec{x}^{2})}{R}\left(\delta K^{t}_{t}-\delta\lambda\right)_{ct}\right]
δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =dd1xzd8RGN[z2hii+(R2z2x2)(δKttδλ)ct].\displaystyle=\frac{d^{d-1}x\,z^{-d}}{8RG_{N}}\left[z^{2}h_{ii}+(R^{2}-z^{2}-\vec{x}^{2})\left(\delta K^{t}_{t}-\delta\lambda\right)_{ct}\right]. (159)

The variation of the onshell boundary Lagrangian, δλ\delta\lambda, is related to the variation of the extrinsic curvature, δK\delta K, via the canonical momentum in (25)(\ref{eq:ConMom})

18πGNδ[γλ]\displaystyle-\frac{1}{8\pi G_{N}}\delta\left[\sqrt{\gamma}\lambda\right] =18πGNδ[γK]+πμνδγμν\displaystyle=-\frac{1}{8\pi G_{N}}\delta\left[\sqrt{\gamma}K\right]+\pi^{\mu\nu}\delta\gamma_{\mu\nu} (160)
γ8πGN(λ2δγ+δλ)\displaystyle-\frac{\sqrt{\gamma}}{8\pi G_{N}}\left(\frac{\lambda}{2}\delta\gamma+\delta\lambda\right) =γ8πGN(K2δγ+δK+KμνKγμν2δγμν)\displaystyle=-\frac{\sqrt{\gamma}}{8\pi G_{N}}\left(\frac{K}{2}\delta\gamma+\delta K+\frac{K^{\mu\nu}-K\gamma^{\mu\nu}}{2}\delta\gamma_{\mu\nu}\right)
λ2δγ+δλ\displaystyle\frac{\lambda}{2}\delta\gamma+\delta\lambda =δK+Kμν2δγμν.\displaystyle=\delta K+\frac{K^{\mu\nu}}{2}\delta\gamma_{\mu\nu}.

For flat boundary metrics we have

δλ=δK.\displaystyle\delta\lambda=\delta K. (161)

We then get the following simplified expression for all dimension

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =zd8RGN[z2hii+(R2z2x2)(δKttδK)ct]\displaystyle=\frac{z^{-d}}{8RG_{N}}\left[z^{2}h_{ii}+(R^{2}-z^{2}-\vec{x}^{2})\left(\delta K^{t}_{t}-\delta K\right)_{ct}\right]
=zd8RGN[z2hii(R2z2x2)δKctii].\displaystyle=\frac{z^{-d}}{8RG_{N}}\left[z^{2}h_{ii}-(R^{2}-z^{2}-\vec{x}^{2})\delta K_{ct\,i}^{\;\;i}\right]. (162)

The extrinsic curvature counterterm means that all the terms appear earlier in the dilatation eigenfunction expansion, i.e.

δKμν\displaystyle\delta K_{\mu\nu} =δK(0)μν+δK(2)μν++logz2δK~(d)μν+δK(d)μν+\displaystyle=\delta K_{(0)\,\mu\nu}+\delta K_{(2)\,\mu\nu}+\cdots+\log z^{2}\delta\tilde{K}_{(d)\,\mu\nu}+\delta K_{(d)\,\mu\nu}+\cdots
δKμν\displaystyle\delta K_{\mu\nu} =δKctμν+δK(d)μν+\displaystyle=\delta K_{ct\,\mu\nu}+\delta K_{(d)\,\mu\nu}+\cdots (163)

From (155)(\ref{eq:dNew4.1}) we can deduce that the divergence of the full Noether charge integral is equal to the divergence of the correction term integral,

(Bϵδ𝑸full[ξB])div=Bϵδ𝚫div[ξB].\displaystyle\left(\int_{B_{\epsilon}}\delta\boldsymbol{Q}^{full}[\xi_{B}]\right)^{div}=\int_{B_{\epsilon}}\delta\boldsymbol{\Delta}^{div}[\xi_{B}]. (164)

In order to see how this divergence is equivalent to the divergence in the entanglement entropy, we need to use the Stoke’s theorem of the full Noether charge form on Σϵ\Sigma_{\epsilon},

Bϵδ𝑸full[ξB]B~ϵδ𝑸full[ξB]=Σϵ𝑑δ𝑸full[ξB].\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{Q}^{full}[\xi_{B}]-\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}^{full}[\xi_{B}]=\int_{\Sigma_{\epsilon}}d\delta\boldsymbol{Q}^{full}[\xi_{B}]. (165)

The exterior derivative of the variation of Noether charge form can be deduced from (31)(\ref{eq:JTL}) and (33)(\ref{eq:dQ}),

dδ𝑸[ξB]\displaystyle d\delta\boldsymbol{Q}[\xi_{B}] =δ𝑱δ𝑵\displaystyle=\delta\boldsymbol{J}-\delta\boldsymbol{N} (166)
=δ𝚯[δξBψ]ιξBδ𝑳+δ𝑵\displaystyle=\delta\boldsymbol{\Theta}[\delta_{\xi_{B}}\psi]-\iota_{\xi_{B}}\delta\boldsymbol{L}+\delta\boldsymbol{N}
=δ𝚯[δξBψ]ιξBdδ𝚯[δψ]ιξB𝑬ψδψ+δ𝑵\displaystyle=\delta\boldsymbol{\Theta}[\delta_{\xi_{B}}\psi]-\iota_{\xi_{B}}d\delta\boldsymbol{\Theta}[\delta\psi]-\iota_{\xi_{B}}\boldsymbol{E}^{\psi}\delta\psi+\delta\boldsymbol{N}
=δ𝚯[δξBψ]ξB𝚯[δξBϕ]+dιξBδ𝚯[δϕ]ιξB𝑬ψδϕ+δ𝑵\displaystyle=\delta\boldsymbol{\Theta}[\delta_{\xi_{B}}\psi]-\mathcal{L}_{\xi_{B}}\boldsymbol{\Theta}[\delta_{\xi_{B}}\phi]+d\iota_{\xi_{B}}\delta\boldsymbol{\Theta}[\delta\phi]-\iota_{\xi_{B}}\boldsymbol{E}^{\psi}\delta\phi+\delta\boldsymbol{N}
=𝝎(δψ,δξBψ)+dιξBδ𝚯[δψ]ιξB𝑬ϕδψ+δ𝑵\displaystyle=\boldsymbol{\omega}(\delta\psi,\delta_{\xi_{B}}\psi)+d\iota_{\xi_{B}}\delta\boldsymbol{\Theta}[\delta\psi]-\iota_{\xi_{B}}\boldsymbol{E}^{\phi}\delta\psi+\delta\boldsymbol{N}
=dιξBδ𝚯[δψ]ιξB𝑬ϕδψ+δ𝑵\displaystyle=d\iota_{\xi_{B}}\delta\boldsymbol{\Theta}[\delta\psi]-\iota_{\xi_{B}}\boldsymbol{E}^{\phi}\delta\psi+\delta\boldsymbol{N} (167)

where 𝝎(δ1ψ,δ2ψ)\boldsymbol{\omega}(\delta_{1}\psi,\delta_{2}\psi) is the symplectic form

𝝎(δ1ψ,δ2ψ)=δ2𝚯[δ1ψ]δ1𝚯[δ2ψ]\displaystyle\boldsymbol{\omega}(\delta_{1}\psi,\delta_{2}\psi)=\delta_{2}\boldsymbol{\Theta}[\delta_{1}\psi]-\delta_{1}\boldsymbol{\Theta}[\delta_{2}\psi] (168)

and it vanishes when ξB\xi_{B} is Killing. Note the last two terms are off-shell terms. We first write out (155)(\ref{eq:dNew4.1}) as

Bϵδ𝑸[ξB]ιξB𝚯[δϕ]=Bϵδ𝓠[ξB]+δ𝚫[ξB].\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}]-\iota_{\xi_{B}}\boldsymbol{\Theta}[\delta\phi]=\int_{B_{\epsilon}}-\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]+\delta\boldsymbol{\Delta}[\xi_{B}]. (169)

Now substitute (167)(\ref{eq:ddQ}) and (169)(\ref{eq:dNewNew4.1}) into (165)(\ref{eq:dQStokes}),

Bϵδ𝑸full[ξB]B~ϵδ𝑸full[ξB]\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{Q}^{full}[\xi_{B}]-\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}^{full}[\xi_{B}] =Σϵ𝑑ιξBδ𝚯[δϕ]+δ𝑵ιξB𝑬ϕδϕdιξBδ𝑩\displaystyle=\int_{\Sigma_{\epsilon}}d\iota_{\xi_{B}}\delta\boldsymbol{\Theta}[\delta\phi]+\delta\boldsymbol{N}-\iota_{\xi_{B}}\boldsymbol{E}^{\phi}\delta\phi-d\iota_{\xi_{B}}\delta\boldsymbol{B}
Bϵδ𝑸[ξB]ιξB𝚯[δϕ]B~ϵδ𝑸[ξB]\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}]-\iota_{\xi_{B}}\boldsymbol{\Theta}[\delta\phi]-\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}] =ΣϵιξB𝑬ϕδϕ+δ𝑵\displaystyle=\int_{\Sigma_{\epsilon}}-\iota_{\xi_{B}}\boldsymbol{E}^{\phi}\delta\phi+\delta\boldsymbol{N}
Bϵδ𝓠[ξB]+δ𝚫[ξB]\displaystyle\int_{{B}_{\epsilon}}-\delta\boldsymbol{\mathcal{Q}}[\xi_{B}]+\delta\boldsymbol{\Delta}[\xi_{B}] =B~ϵδ𝑸[ξB]+ΣϵιξB𝑬ϕδϕ+δ𝑵.\displaystyle=\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}]+\int_{\Sigma_{\epsilon}}-\iota_{\xi_{B}}\boldsymbol{E}^{\phi}\delta\phi+\delta\boldsymbol{N}. (170)

Onshell we get

Bϵδ𝓠[ξB]\displaystyle\int_{{B}_{\epsilon}}-\delta\boldsymbol{\mathcal{Q}}[\xi_{B}] =B~ϵδ𝑸[ξB]Bϵδ𝚫[ξB].\displaystyle=\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}]-\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}]. (171)

Since the left hand side is manifestly finite we have

(B~ϵδ𝑸[ξB])div\displaystyle\left(\int_{\tilde{B}_{\epsilon}}\delta\boldsymbol{Q}[\xi_{B}]\right)^{div} =Bϵδ𝚫div[ξB]\displaystyle=\int_{B_{\epsilon}}\delta\boldsymbol{\Delta}^{div}[\xi_{B}] (172)
δSBdiv\displaystyle\delta S_{B}^{div} =Bϵδ𝚫div[ξB]\displaystyle=\int_{B_{\epsilon}}\delta\boldsymbol{\Delta}^{div}[\xi_{B}] (173)

Therefore the integral of δ𝚫\delta\boldsymbol{\Delta} on the boundary ball region can be thought of as the counterterm of the entanglement entropy. In the next section we will show that the finite part of δ𝚫\delta\boldsymbol{\Delta} matches with the counterterm of the entanglement entropy as well. Hence we get the integral first law of entanglement entropy:

δEB\displaystyle\delta E_{B} =δSBren.\displaystyle=\delta S_{B}^{ren}. (174)

Finite counterterms contribute only when the CFT dimension is even. This is an expected result as the finite part of the entanglement entropy is scheme dependent in even dd. Similarly, the left hand side is related to the renormalized energy momentum tensor which is also scheme dependent for even dd. For odd dd, the finite part of the renormalized entanglement entropy is universal We will see explicit examples in the following section.

The implication of δ𝚫\delta\boldsymbol{\Delta} acting as the density of the entanglement entropy counterterms is that δ𝚫\delta\boldsymbol{\Delta} is exact,

δ𝚫[ξB]=dδ𝑺Bct,\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}]=d\delta\boldsymbol{S}^{ct}_{B}, (175)

where δ𝑺Bct\delta\boldsymbol{S}^{ct}_{B} is the (d2)(d-2)-form that integrates to the entanglement entropy counterterm. This means the full Hamiltonian potential δ𝑯full\delta\boldsymbol{H}^{full} and the renormalized Gibbon-Hawking Hamiltonian potential δ𝑯(d)GH\delta\boldsymbol{H}^{GH}_{(d)} is equal up to an exact term. In the usual context of conserved quantities, this is attributed to the exact term ambiguity. Since the potential is integrated over a boundary manifold, which itself does not have boundary, the exact term ambiguity will not contribute to the conserved charge. For us, both the entanglement entropy and modular energy is defined as the integral of a manifold that does have boundary, so the exact term difference is no longer an ambiguity. Note the counterterm of the entanglement entropy is obtained systemically from the renormalized action through the replica trick, this indicates this exact term difference can be calculated from the renormalized action directly. In the Hamiltonian holographic renormalization framework, we show how to obtain δ𝚫\delta\boldsymbol{\Delta} from the counterterms contribution of the Hamiltonian potentials in appendix C.

4.3 Examples: Generalized First Law in AlAdSd+1AlAdS_{d+1}

We have shown that the variation of the modular energy δEB\delta E_{B} is equal to the integral of the holographic charge form over the boundary ball region BϵB_{\epsilon} in (150)(\ref{eq:dModE}) and the variation of the entanglement entropy δSB\delta S_{B} is equal to the integral of the Noether charge form over the bulk entangling surface B~ϵ\tilde{B}_{\epsilon} in (144)(\ref{eq:dEntropyInt}). To complete the generalized first law of entanglement entropy (174)(\ref{eq:New1stLaw}) for generic variations of the boundary metric δγμν\delta\gamma_{\mu\nu} in AlAdSd+1AlAdS_{d+1}, we only need to check that the integral of the term δ𝚫[ξB]\delta\boldsymbol{\Delta}[\xi_{B}] over BϵB_{\epsilon} is the counterterm of the entanglement entropy,

Bϵδ𝚫[ξB]=δSBct.\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}]=\delta S^{ct}_{B}. (176)

In the following subsections, we will demonstrate this equality up to dimension d=5d=5, thus implying the renormalized first law (174)(\ref{eq:New1stLaw}), with scheme dependence of renormalized entropy and energy systematically matched.

4.3.1 d=3d=3

The terms in the dilatation eigenfunction expansion of the extrinsic curvature variation are related to the Fefferman-Graham expansion of the boundary metric variation. For d=3d=3, we only need to include terms up to O(z4)O(z^{4}) as higher order terms will not contribute to calculations in the limit ϵ0\epsilon\rightarrow 0:

δK(0)νμ\displaystyle\delta K_{(0)\,\nu}^{\;\;\mu} =0\displaystyle=0 (177)
δK(2)νμ\displaystyle\delta K_{(2)\,\nu}^{\;\;\mu} =z2ημσh(2)σν+O(z4)\displaystyle=-z^{2}\eta^{\mu\sigma}h_{(2)\,\sigma\nu}+O(z^{4}) (178)
δK(3)νμ\displaystyle\delta{K}_{(3)\,\nu}^{\;\;\mu} =32z3ημσh(3)σν+O(z4).\displaystyle=-\frac{3}{2}z^{3}\eta^{\mu\sigma}h_{(3)\,\sigma\nu}+O(z^{4}). (179)

In this case, the counterterm δKctνμ\delta K_{ct\,\nu}^{\;\;\mu} is just the second term in the dilatation eigenfunction expansion δK(2)νμ\delta K_{(2)\,\nu}^{\;\;\mu}. Hence the counterterm from (159)(\ref{eq:dDelta}) gives

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =𝒅𝟐𝒙z38RGN[z2hii(R2z2x2)(z2h(2)ii)].\displaystyle=\frac{\boldsymbol{d^{2}x}z^{-3}}{8RG_{N}}\left[z^{2}h_{ii}-(R^{2}-z^{2}-\vec{x}^{2})\left(-z^{2}h_{(2)\,ii}\right)\right]. (180)

Keeping the terms up to O(z)O(z) we have,

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =𝒅𝟐𝒙8RGN[1z(h(0)ii+(R2x2)h(2)ii)].\displaystyle=\frac{\boldsymbol{d^{2}x}}{8RG_{N}}\left[\frac{1}{z}\left(h_{(0)\,ii}+(R^{2}-\vec{x}^{2})h_{(2)\,ii}\right)\right]. (181)

We see that for this example in odd dimensions, δ𝚫[ξB]\delta\boldsymbol{\Delta}[\xi_{B}] has no term of order z0z^{0} and there is as expected no finite counterterm contribution to the entanglement entropy.

To see the identification of the integral of δ𝚫[ξB]\delta\boldsymbol{\Delta}[\xi_{B}] over BϵB_{\epsilon} with the ordinary entanglement entropy counterterm in (87)(\ref{eq:ExdSCt}), we need to use (B.2)(\ref{eq:Inth0h2}) with the result

Bϵδ𝚫[ξB]\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}] =18RGNS1𝑑Ω1rϵ(h(0)iix^ix^jh(0)ij).\displaystyle=\frac{1}{8RG_{N}}\int_{S^{1}}d\Omega_{1}\frac{r}{\epsilon}\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right). (182)

This matches with the variation in the counterterm (87)(\ref{eq:ExdSCt}) exactly. Hence we have satisfied (176)(\ref{eq:IntDdSct}) confirming that the general variation of the modular energy is the variation of the renormalized entanglement entropy.

4.3.2 d=4d=4

For d=4d=4, in addition to including the logarithmic term in dilatation eigenfunction expansion, we also have to include terms up to O(z6)O(z^{6}) to evaluate both the divergent and finite contributions:

δK(0)νμ\displaystyle\delta K_{(0)\,\nu}^{\;\;\mu} =0\displaystyle=0 (183)
δK(2)νμ\displaystyle\delta K_{(2)\,\nu}^{\;\;\mu} =z2ημσh(2)σνz2ημσδD(2)σν+O(z6)\displaystyle=-z^{2}\eta^{\mu\sigma}h_{(2)\,\sigma\nu}-z^{2}\eta^{\mu\sigma}\delta D_{(2)\,\sigma\nu}+O(z^{6}) (184)
δK~(4)νμ\displaystyle\delta\tilde{K}_{(4)\,\nu}^{\;\;\mu} =2z4ημσh~(4)σν+O(z6)\displaystyle=-2z^{4}\eta^{\mu\sigma}\tilde{h}_{(4)\,\sigma\nu}+O(z^{6}) (185)
δK(4)νμ\displaystyle\delta{K}_{(4)\,\nu}^{\;\;\mu} =2z4ημσh(4)σνz4ημσh~(4)σν+z4ημσδD(2)σν+O(z6)\displaystyle=-2z^{4}\eta^{\mu\sigma}h_{(4)\,\sigma\nu}-z^{4}\eta^{\mu\sigma}\tilde{h}_{(4)\,\sigma\nu}+z^{4}\eta^{\mu\sigma}\delta D_{(2)\,\sigma\nu}+O(z^{6}) (186)

where we use the notation

δD(n)μν=δ[g(n)σρg(n)μνg(0)σρ].\displaystyle\delta D_{(n)\,\mu\nu}=\delta\left[\int g_{(n)\sigma\rho}\frac{g_{(n)\mu\nu}}{g_{(0)\sigma\rho}}\right]. (187)

It turns out at linear level that the second order term δD(2)μν\delta D_{(2)\,\mu\nu} is related to the coefficient of the logarithmic term in the Fefferman-Graham expansion as

δD(2)μν=2h~(4)μν,\displaystyle\delta D_{(2)\,\mu\nu}=-2\tilde{h}_{(4)\,\mu\nu}, (188)

and hence it is also traceless. Then the relevant terms in the dilatation eigenfunction expansion for the extrinsic curvature are

δK(0)νμ\displaystyle\delta K_{(0)\,\nu}^{\;\;\mu} =0\displaystyle=0 (189)
δK(2)νμ\displaystyle\delta K_{(2)\,\nu}^{\;\;\mu} =z2ημσh(2)σν+2z4ημσh~(4)σν+O(z6)\displaystyle=-z^{2}\eta^{\mu\sigma}h_{(2)\,\sigma\nu}+2z^{4}\eta^{\mu\sigma}\tilde{h}_{(4)\,\sigma\nu}+O(z^{6}) (190)
δK~(4)νμ\displaystyle\delta\tilde{K}_{(4)\,\nu}^{\;\;\mu} =2z4ημσh~(4)σν+O(z6)\displaystyle=-2z^{4}\eta^{\mu\sigma}\tilde{h}_{(4)\,\sigma\nu}+O(z^{6}) (191)
δK(4)νμ\displaystyle\delta{K}_{(4)\,\nu}^{\;\;\mu} =2z4ημσh(4)σν3z4ημσh~(4)σν+O(z6).\displaystyle=-2z^{4}\eta^{\mu\sigma}h_{(4)\,\sigma\nu}-3z^{4}\eta^{\mu\sigma}\tilde{h}_{(4)\,\sigma\nu}+O(z^{6}). (192)

The counterterm from (159)(\ref{eq:dDelta}) is then

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =𝒅𝟑xz48RGN[z2hii(R2z2x2)(z2h(2)ii+2z4h~(4)ii2z4logz2h~(4)ii)].\displaystyle=\frac{\boldsymbol{d^{3}}xz^{-4}}{8RG_{N}}\left[z^{2}h_{ii}-(R^{2}-z^{2}-\vec{x}^{2})\left(-z^{2}h_{(2)\,ii}+2z^{4}\tilde{h}_{(4)\,ii}-2z^{4}\log z^{2}\tilde{h}_{(4)\,ii}\right)\right]. (193)

Neglecting the O(z)O(z) terms as they vanish in the limit ϵ0\epsilon\rightarrow 0 we have

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =𝒅𝟑𝒙8RGN[1z2(h(0)ii(R2x2)h(2)ii)+2(R2x2)(1logz2)h~(4)ii].\displaystyle=\frac{\boldsymbol{d^{3}x}}{8RG_{N}}\left[\frac{1}{z^{2}}\left(h_{(0)\,ii}-(R^{2}-\vec{x}^{2})h_{(2)\,ii}\right)+2(R^{2}-\vec{x}^{2})(1-\log z^{2})\tilde{h}_{(4)\,ii}\right]. (194)

Finally we need to transform this integral on boundary ball region BϵB_{\epsilon} into a surface integral on the sphere Bϵ\partial B_{\epsilon} via the manipulation of h(n)μνh_{(n)\,\mu\nu} in appendix B.2. First we use (B.2)(\ref{eq:Inth0h2}) to turn the coefficient of ϵ2\epsilon^{-2} divergences into a surface integral

Bϵδ𝚫[ξB]\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}] =18RGN[(R2ϵ2)322ϵ2S2dΩ2(h(0)iix^ix^jh(0)ij)\displaystyle=\frac{1}{8RG_{N}}\bigg{[}\frac{(R^{2}-\epsilon^{2})^{\frac{3}{2}}}{2\epsilon^{2}}\int_{S^{2}}d\Omega_{2}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right) (195)
+Bϵd3xh(2)ii\displaystyle+\int_{{B}_{\epsilon}}d^{3}x\,h_{(2)\,ii}
2(1logϵ2)Bϵd3x(R2x2)h~(4)ii].\displaystyle-2(1-\log\epsilon^{2})\int_{{B}_{\epsilon}}d^{3}x\,(R^{2}-\vec{x}^{2})\tilde{h}_{(4)\,ii}\bigg{]}.

For the coefficient of the logarithmic divergence, we use (326)(\ref{eq:Inth4h2}) to turn the h~(4)ii\tilde{h}_{(4)\,ii} integral into integrals of h(2)iih_{(2)\,ii} then use (328)(\ref{eq:Inth2h0}) to turn the remaining volume integral into a surface integral of h(0)iih_{(0)\,ii}. The final result is

Bϵδ𝚫[ξB]\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}] =18RGNS2dΩ2[r32ϵ2(h(0)iix^ix^jh(0)ij)\displaystyle=\frac{1}{8RG_{N}}\int_{S^{2}}d\Omega_{2}\,\bigg{[}\frac{r^{3}}{2\epsilon^{2}}\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right) (196)
+logϵ2(r32(x^ix^jh(2)ijh(2)ii)\displaystyle+\log\epsilon^{2}\bigg{(}\frac{r^{3}}{2}(\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}-h_{(2)\,ii})
+r4(h(0)ii3x^ix^jh(0)ij+xjjh(0)iix^ix^jxkkh(0)ij))\displaystyle+\frac{r}{4}(h_{(0)\,ii}-3\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}+x^{j}\partial_{j}h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}x^{k}\partial_{k}h_{(0)\,ij})\bigg{)}
r32(x^ix^jh(2)ijh(2)ii)]\displaystyle-\frac{r^{3}}{2}(\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}-h_{(2)\,ii})\bigg{]}

Note that there are, as expected, finite contributions. Comparing with (121)(\ref{eq:ExdSCtd4}) we can see this term is exactly the counterterm for the entanglement entropy. Therefore in AlAdS5AlAdS_{5} we have satisfied (176)(\ref{eq:IntDdSct}). The renormalized stress tensor TμνrenT^{ren}_{\mu\nu} in (80)(\ref{eq:dTrend4}) has a scheme dependent term proportional to h~(d)μν\tilde{h}_{(d)\,\mu\nu} that originates from the variation of anomaly term in the counterterm action. Therefore the finite counterterm in the entanglement entropy is necessary to match the contribution associated with the holographic conformal anomaly.

4.3.3 d=5d=5

The d=5d=5 case is very similar to the above example but without the logarithmic terms. The dilatation eigenfunction expansion for the variation of the extrinsic curvature is

δK(0)νμ\displaystyle\delta K_{(0)\,\nu}^{\;\;\mu} =0\displaystyle=0 (197)
δK(2)νμ\displaystyle\delta K_{(2)\,\nu}^{\;\;\mu} =z2ημσh(2)σνz2ημσδD(2)σν+O(z6)\displaystyle=-z^{2}\eta^{\mu\sigma}h_{(2)\,\sigma\nu}-z^{2}\eta^{\mu\sigma}\delta D_{(2)\,\sigma\nu}+O(z^{6}) (198)
δK(4)νμ\displaystyle\delta{K}_{(4)\,\nu}^{\;\;\mu} =2z4ημσh(4)σν+z4ημσδD(2)σν+O(z6)\displaystyle=-2z^{4}\eta^{\mu\sigma}h_{(4)\,\sigma\nu}+z^{4}\eta^{\mu\sigma}\delta D_{(2)\,\sigma\nu}+O(z^{6}) (199)

where at linear level we have

δD(2)ii=23h(4)ii.\displaystyle\delta D_{(2)\,ii}=\frac{2}{3}h_{(4)\,ii}. (200)

Then the relevant dilatation eigenfunction expansion terms, up to O(z6)O(z^{6}), are

δK(0)ii\displaystyle\delta K_{(0)\,i}^{\;\;i} =0\displaystyle=0 (201)
δK(2)ii\displaystyle\delta K_{(2)\,i}^{\;\;i} =z2h(2)ii23z2h(4)ii+O(z6)\displaystyle=-z^{2}h_{(2)\,ii}-\frac{2}{3}z^{2}h_{(4)\,ii}+O(z^{6}) (202)
δK(4)ii\displaystyle\delta{K}_{(4)\,i}^{\;\;i} =43z4h(4)ii+O(z6).\displaystyle=-\frac{4}{3}z^{4}h_{(4)\,ii}+O(z^{6}). (203)

The counterterm from (159)(\ref{eq:dDelta}) gives

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =𝒅𝟒𝒙z58RG6[z2hii(R2z2x2)(z2h(2)ii2z4h(4)ii)].\displaystyle=\frac{\boldsymbol{d^{4}x}z^{-5}}{8RG_{6}}\left[z^{2}h_{ii}-(R^{2}-z^{2}-\vec{x}^{2})\left(-z^{2}h_{(2)\,ii}-2z^{4}{h}_{(4)\,ii}\right)\right]. (204)

Neglecting the O(z)O(z) terms as they vanish in the limit ϵ0\epsilon\rightarrow 0 we have,

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =𝒅𝟒𝒙8RG6[1z3(h(0)ii(R2x2)h(2)ii)+2z(R2x2)h(4)ii].\displaystyle=\frac{\boldsymbol{d^{4}x}}{8RG_{6}}\left[\frac{1}{z^{3}}\left(h_{(0)\,ii}-(R^{2}-\vec{x}^{2})h_{(2)\,ii}\right)+\frac{2}{z}(R^{2}-\vec{x}^{2}){h}_{(4)\,ii}\right]. (205)

Now evaluate the integral of the correction following in appendix B.2. We use (B.2)(\ref{eq:Inth0h2}) and (331)(\ref{eq:Inh4h2}) to get

Bϵδ𝚫[ξB]=\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}]= 18RG6[(R2ϵ2)23ϵ3S3dΩ3(h(0)iix^ix^jh(0)ij)+1ϵBϵd4xh(2)ii\displaystyle\frac{1}{8RG_{6}}\bigg{[}\frac{(R^{2}-\epsilon^{2})^{2}}{3\epsilon^{3}}\int_{S^{3}}d\Omega_{3}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right)+\frac{1}{\epsilon}\int_{{B}_{\epsilon}}d^{4}x\,h_{(2)\,ii} (206)
+(R2ϵ2)2ϵS3dΩ3(h(2)iix^ix^jh(2)ij)3ϵBϵd4xh(2)ii].\displaystyle\quad\quad+\frac{(R^{2}-\epsilon^{2})^{2}}{\epsilon}\int_{S^{3}}d\Omega_{3}\,\left(h_{(2)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}\right)-\frac{3}{\epsilon}\int_{{B}_{\epsilon}}d^{4}x\,h_{(2)\,ii}\bigg{]}.

The remaining volume integral of h(2)iih_{(2)\,ii} can be converted to surface integral via (333)(\ref{eq:Inh2h0}),

Bϵδ𝚫[ξB]=\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}]= 18RG6S3dΩ3[R43ϵ3(h(0)iix^ix^jh(0)ij)2R23ϵ(h(0)iix^ix^jh(0)ij)\displaystyle\frac{1}{8RG_{6}}\int_{S^{3}}d\Omega_{3}\bigg{[}\frac{R^{4}}{3\epsilon^{3}}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right)-\frac{2R^{2}}{3\epsilon}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right) (207)
+R4ϵ(h(2)iix^ix^jh(2)ij)R23ϵ(h(0)ii4x^ix^jh(0)ij+xjjh(0)iix^ix^jxkkh(0)ij)].\displaystyle+\frac{R^{4}}{\epsilon}\,\left(h_{(2)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}\right)-\frac{R^{2}}{3\epsilon}\left(h_{(0)\,ii}-4\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}+x^{j}\partial_{j}h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}x^{k}\partial_{k}h_{(0)\,ij}\right)\bigg{]}.

After rearranging we arrive at the final expression of the correction term

Bϵδ𝚫[ξB]=\displaystyle\int_{{B}_{\epsilon}}\delta\boldsymbol{\Delta}[\xi_{B}]= 18RG6S3dΩ3[R43ϵ3(h(0)iix^ix^jh(0)ij)R2ϵ(h(0)ii2x^ix^jh(0)ij)\displaystyle\frac{1}{8RG_{6}}\int_{S^{3}}d\Omega_{3}\bigg{[}\frac{R^{4}}{3\epsilon^{3}}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right)-\frac{R^{2}}{\epsilon}\,\left(h_{(0)\,ii}-2\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right) (208)
+R4ϵ(h(2)iix^ix^jh(2)ij)R23ϵ(xjjh(0)iix^ix^jxkkh(0)ij)].\displaystyle+\frac{R^{4}}{\epsilon}\,\left(h_{(2)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}\right)-\frac{R^{2}}{3\epsilon}\left(x^{j}\partial_{j}h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}x^{k}\partial_{k}h_{(0)\,ij}\right)\bigg{]}.

This is in fact identical to the counterterm in (87)(\ref{eq:ExdSCt}) when taking the limit ϵ0\epsilon\rightarrow 0 and satisfying (176)(\ref{eq:IntDdSct}). Note that in (87)(\ref{eq:ExdSCt}) one has to expand r=R2ϵ2r=\sqrt{R^{2}-\epsilon^{2}} to arrive at (208)(\ref{eq:FinaldDd5}). Since dd is odd, there is no finite counterterm and the renormalized first law is scheme independent.

5 Conclusions and Outlook

In this paper we have proven the renormalized first law of holographic entanglement entropy, in both infinitesimal and covariant versions, for generic variations of the metric. The original proofs of the first law of holographic entanglement entropy assumed that only normalisable modes of the metric were varied, corresponding to changing the state in the dual conformal field theory. Our proof extends to non-normalisable variations of the metric, corresponding to changing the background metric for the dual conformal theory.

When the boundary dimension dd is odd, both the renormalized stressed tensor and renormalized area of the entangling surface are scheme independent and the holographic conformal anomaly is absent. When the boundary dimension dd is even, there are finite contributions from counterterms and one needs to ensure that the same renormalization scheme is used for the stress tensor and entanglement entropy; this follows immediately from the approach taken in Taylor:2016aoi because the counterterms for the entanglement entropy are derived from the counterterms for the action given in deHaro2001 using the replica trick. In our setup the background about which we are perturbing is conformally flat and thus there are no explicit contributions from the conformal anomaly at linear order.

The first law can also be derived using the covariant phase space approach, building on Faulkner:2013ica , as well discussions of the covariant phase space formalism in the presence of boundaries 2020HarlowWu and boundary counterterm contributions to conserved charges Papadimitriou:2005ii . The generalization to non-normalizable variations of the bulk metric, corresponding to deforming the background metric for the dual CFT, induces specific counterterms in the covariant phase space construction. We explain in detail how these relate to the boundary terms in 2020HarlowWu . Note that in the context of the laws of black holes one would fix the non-normalizable modes and therefore the our analysis differs from the renormalized black hole charge analysis of Papadimitriou:2005ii .


While the focus of this paper has been on proving the holographic first law of entanglement entropy for non-normalisable bulk metric variations, our methodology could be extended to many analyses within holographic information theory. One could clearly explore perturbations of the surface itself, following Klebanov:2012yf ; Allais:2014ata ; Rosenhaus:2014zza ; Mezei:2014zla . The extension to higher derivative gravity theories would be straightforward in principle although one may need to resolve analogous technical ambiguities to those encountered in 2014jc ; 2014xd . Analyses of local reconstruction in the bulk from boundary entanglement such as Lashkari:2013koa ; 2015LinOoguriStocia assume normalizable fall offs of metric perturbations (corresponding to CFT states), but our approach facilitates the discussion of marginal and indeed even irrelevant deformations. To include the latter, one would simply add in the bulk field corresponding to the irrelevant operator, and compute renormalized quantities perturbatively in the irrelevant deformation. Other analyses where our methodology would be useful to extend the class of theories/states under consideration include discussions of subregion complexity and the first law of complexity Jang:2020cbm ; Bernamonti2020 as well as analyses of the relation of holographic entanglement entropy to inverse mean curvature flow 2017fw .


Finally, let us consider the expression for the variation of the entanglement entropy in terms of the Weyl tensor (96)(\ref{eq:dSdW}). This relation could have been anticipated from the known relationship between the Einstein sector of conformal (Weyl) gravity and Einstein gravity maldacena2011einstein ; 2016ao . Up to a topological term the renormalized action for Einstein gravity is proportional to the Weyl squared term Anderson:2001vr ; 2016ao ; Anastasiou_2021 . Accordingly, the Wald entropy functional for the AdS Rindler black hole on the black hole horizon H~ϵ\tilde{H}_{\epsilon} gives

SWaldH~ϵWabcdnabncd\displaystyle S_{\scriptscriptstyle Wald}\propto\int_{\tilde{H}_{\epsilon}}W^{abcd}n_{ab}n_{cd} (209)

where nabn_{ab} is the binormal for the codimension two surface H~ϵ\tilde{H}_{\epsilon}. Using the standard Casini, Huerta and Myers approach Casini:2011kv we can then map this entropy to the entanglement entropy for a spherical region in a flat background. The computations in this paper relate to the first variation of this entropy under bulk metric variations and using the CHM map we immediately obtain the first term of the Weyl integral in (97)(\ref{eq:dWintW})

δSrenB~ϵδW1212.\displaystyle\delta S^{\rm ren}\propto\int_{\tilde{B}_{\epsilon}}\delta W_{1212}. (210)

This relation holds in all even bulk spacetime dimensions, even though the expressions for the renormalized entanglement entropy become increasingly complex expressions of the Euler characteristic and curvature invariants of the entangling surface in higher dimensions Taylor:2020uwf . The variation manifestly simplifies to just this one term for linear variations of a spherical surface around a background with zero Weyl curvature. Working to higher order in the variations, and in more general setups, one should make use of the full form of the renormalized area in terms of Euler characteristic and curvature invariants in Taylor:2020uwf to understand the underlying geometric structure.

Acknowledgements

This work is funded by the STFC grant ST/P000711/1. This project has received funding and support from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 690575. LT would like to thank A. Poole and F. Capone for relevant discussions.

Appendix A Infinitesimal first law

A.1 Useful Identities

In this appendix, we provide some useful identities that are used in section 3. First we give angular integrals of the unit vectors,

Sd𝑑Ωdx^odd=0\displaystyle\int_{S^{d}}d\Omega_{d}\hat{x}^{odd}=0 (211)
Sd𝑑Ωdx^ix^j=Ωdd+1δij\displaystyle\int_{S^{d}}d\Omega_{d}\hat{x}^{i}\hat{x}^{j}=\frac{\Omega_{d}}{d+1}\delta^{ij} (212)
Sd𝑑Ωdx^ix^jx^kx^l=Ωd(d+3)(d+1)(δijδkl+δikδjl+δilδjk)\displaystyle\int_{S^{d}}d\Omega_{d}\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}=\frac{\Omega_{d}}{(d+3)(d+1)}(\delta^{ij}\delta^{kl}+\delta^{ik}\delta^{jl}+\delta^{il}\delta^{jk}) (213)
Sd𝑑Ωdx^ix^jx^kx^lx^px^q=15Ωd(d+5)(d+3)(d+1)δ(ijδklδpq)\displaystyle\int_{S^{d}}d\Omega_{d}\,\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}\hat{x}^{p}\hat{x}^{q}=\frac{15\Omega_{d}}{(d+5)(d+3)(d+1)}\delta^{(ij}\delta^{kl}\delta^{pq)} (214)
Sd𝑑Ωdx^i1x^i2x^i2n1x^i2n=Ωdr=1n2r1(d+2r1)δ(i1i2δi2n1i2n).\displaystyle\int_{S^{d}}d\Omega_{d}\,\hat{x}^{i_{1}}\hat{x}^{i_{2}}\cdots\hat{x}^{i_{2n-1}}\hat{x}^{i_{2n}}=\Omega_{d}\prod_{r=1}^{n}\frac{2r-1}{(d+2r-1)}\delta^{(i_{1}i_{2}}\cdots\delta^{i_{2n-1}i_{2n})}. (215)

Since the angular integral of unit vectors is expressed as symmetrized Kronecker deltas, it is also useful to have the expression of the symmetrized Kronecker deltas contracted with derivatives of the metric perturbation:

δ(ijδkl)klhij=13kkhii+23ijhij\displaystyle\delta^{(ij}\delta^{kl)}\partial_{k}\partial_{l}h_{ij}=\frac{1}{3}\partial_{k}\partial_{k}h_{ii}+\frac{2}{3}\partial_{i}\partial_{j}h_{ij} (216)
δ(ijδklδpq)klpqhij=15kkllhii+45kkijhij\displaystyle\delta^{(ij}\delta^{kl}\delta^{pq)}\partial_{k}\partial_{l}\partial_{p}\partial_{q}h_{ij}=\frac{1}{5}\partial_{k}\partial_{k}\partial_{l}\partial_{l}h_{ii}+\frac{4}{5}\partial_{k}\partial_{k}\partial_{i}\partial_{j}h_{ij} (217)
δ(i1i2δi2n1i2n)i3i4i2n1i2nhi1i2=(2)n22n1[2hi1i1+(2n2)i1i2hi1i2].\displaystyle\delta^{(i_{1}i_{2}}\cdots\delta^{i_{2n-1}i_{2n})}\partial_{i_{3}}\partial_{i_{4}}\cdots\partial_{i_{2n-1}}\partial_{i_{2n}}h_{i_{1}i_{2}}=\frac{(\partial^{2})^{n-2}}{2n-1}\big{[}\partial^{2}h_{i_{1}i_{1}}+(2n-2)\partial_{i_{1}}\partial_{i_{2}}h_{i_{1}i_{2}}\big{]}. (218)

A.2 Explicit Variation in d=3d=3

In this section we will show the procedure used to calculate the variation of the variation of regularized entanglement entropy and variation of the counterterms in d=3d=3. Here we continue the calculation from (84)(\ref{eq:IntSregwcoord}). First we consider the leading order in the Taylor expansion which has no derivatives and perform the angular integrals (212)(\ref{eq:AInt2}) to get

δSBreg(0)\displaystyle\delta S^{reg}_{B}(\partial^{0}) =Ω18G4ξπ/2𝑑ucos(u)sin2u(1cos2u2)hii(z,0,0),\displaystyle=\frac{\Omega_{1}}{8G_{4}}\int_{\xi}^{\pi/2}du\frac{\cos{u}}{\sin^{2}{u}}(1-\frac{\cos^{2}{u}}{2})h_{ii}(z,0,0), (219)

where

hii(z)\displaystyle h_{ii}(z) =hii(0)+R2sin2uhii(2)+R3sin3uhii(3).\displaystyle=h^{(0)}_{ii}+R^{2}\sin^{2}{u}h^{(2)}_{ii}+R^{3}\sin^{3}{u}h^{(3)}_{ii}. (220)

After performing the uu integrals we get

δSBreg(0)\displaystyle\delta S^{reg}_{B}(\partial^{0}) =Ω18G4[12(1sin(ξ)sin(ξ))hii(0)+(23sin(ξ)2sin3ξ6)R2hii(2)\displaystyle=\frac{\Omega_{1}}{8G_{4}}\bigg{[}\frac{1}{2}\Big{(}\frac{1}{\sin{\xi}}-\sin{\xi}\Big{)}h^{(0)}_{ii}+\Big{(}\frac{2}{3}-\frac{\sin{\xi}}{2}-\frac{\sin^{3}{\xi}}{6}\Big{)}R^{2}h^{(2)}_{ii}
+(38sin2ξ4sin4ξ8)R3hii(3)]\displaystyle\;+\Big{(}\frac{3}{8}-\frac{\sin^{2}{\xi}}{4}-\frac{\sin^{4}{\xi}}{8}\Big{)}R^{3}h^{(3)}_{ii}\bigg{]} (221)

We also need to evaluate the higher derivative terms in the Taylor expansion. For our purposes we need only the Taylor expansion of hij(0)(x,y)h^{(0)}_{ij}(x,y). The contribution of the one derivative term of the variation is

δSBreg(1)=R8G4ξπ/2𝑑u02π𝑑ϕcos2usin2u(δijcos2ux^ix^j)x^kkhij(0).\delta S^{reg}_{B}(\partial^{1})=\frac{R}{8G_{4}}\int_{\xi}^{\pi/2}du\int^{2\pi}_{0}d\phi\frac{\cos^{2}{u}}{\sin^{2}{u}}(\delta^{ij}-\cos^{2}{u}\hat{x}^{i}\hat{x}^{j})\hat{x}^{k}\partial_{k}h^{(0)}_{ij}. (222)

Using the angular integrals (211)(\ref{eq:AIntodd}) we can deduce δSBreg(1)\delta S^{reg}_{B}(\partial^{1}) vanishes.

The contribution of the leading two derivative terms of the variation is

δSBreg(2)=R28G4ξπ/2𝑑u02π𝑑ϕcos3usin2u(δijcos2ux^ix^j)x^kx^l2!klhij(0).\delta S^{reg}_{B}(\partial^{2})=\frac{R^{2}}{8G_{4}}\int_{\xi}^{\pi/2}du\int^{2\pi}_{0}d\phi\frac{\cos^{3}{u}}{\sin^{2}{u}}(\delta^{ij}-\cos^{2}{u}\hat{x}^{i}\hat{x}^{j})\frac{\hat{x}^{k}\hat{x}^{l}}{2!}\partial_{k}\partial_{l}h^{(0)}_{ij}. (223)

We evaluate the angular integrals by substituting the results from (212)(\ref{eq:AInt2}) and (213)(\ref{eq:AInt4}),

δSBreg(2)=R2Ω18G4ξπ/2𝑑ucos3usin2u14jjhii(0)cos5usin2u116(jjhii(0)+2ijhij(0)).\displaystyle\delta S^{reg}_{B}(\partial^{2})=\frac{R^{2}\Omega_{1}}{8G_{4}}\int_{\xi}^{\pi/2}du\frac{\cos^{3}{u}}{\sin^{2}{u}}\frac{1}{4}\partial_{j}\partial_{j}h^{(0)}_{ii}-\frac{\cos^{5}{u}}{\sin^{2}{u}}\frac{1}{16}\Big{(}\partial_{j}\partial_{j}h^{(0)}_{ii}+2\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)}. (224)

Evaluating the uu integral and rearranging the derivatives of metric variation we get

δSBreg(2)=R2Ω18G4[14(2+1sin(ξ)+sin(ξ))jjhii(0)\displaystyle\delta S^{reg}_{B}(\partial^{2})=\frac{R^{2}\Omega_{1}}{8G_{4}}\bigg{[}\frac{1}{4}\Big{(}-2+\frac{1}{\sin{\xi}}+\sin{\xi}\Big{)}\partial_{j}\partial_{j}h^{(0)}_{ii}
116(83+1sin(ξ)+2sin(ξ)+sin3ξ3)(jjhii(0)+2ijhij(0))]\displaystyle\quad\quad\quad\quad-\frac{1}{16}\Big{(}-\frac{8}{3}+\frac{1}{\sin{\xi}}+2\sin{\xi}+\frac{\sin^{3}{\xi}}{3}\Big{)}\Big{(}\partial_{j}\partial_{j}h^{(0)}_{ii}+2\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)}\bigg{]}
δSBref(2)=R2Ω132G4[(43+34sin(ξ)+sin(ξ)2sin3ξ12)jjhii(0)\displaystyle\delta S^{ref}_{B}(\partial^{2})=\frac{R^{2}\Omega_{1}}{32G_{4}}\bigg{[}\Big{(}-\frac{4}{3}+\frac{3}{4\sin{\xi}}+\frac{\sin{\xi}}{2}-\frac{\sin^{3}{\xi}}{12}\Big{)}\partial_{j}\partial_{j}h^{(0)}_{ii}
(4312sin(ξ)sin(ξ)sin3ξ6)ijhij(0)]\displaystyle\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\Big{(}\frac{4}{3}-\frac{1}{2\sin{\xi}}-\sin{\xi}-\frac{\sin^{3}{\xi}}{6}\Big{)}\partial_{i}\partial_{j}h^{(0)}_{ij}\bigg{]} (225)

We would like to take the limit of ξ0\xi\rightarrow 0 so we need to check the divergences are cancelled out by counterterms in (87)(\ref{eq:ExdSCt}). We first evaluate the leading order terms in the Taylor series with no derivatives:

δSBct(0)=18G402π𝑑zϕcos(ξ)sin(ξ)(hiix^ix^jhij).\delta S^{ct}_{B}(\partial^{0})=\frac{1}{8G_{4}}\int^{2\pi}_{0}dz\phi\frac{\cos{\xi}}{\sin{\xi}}(h_{ii}-\hat{x}^{i}\hat{x}^{j}h_{ij}). (226)

Evaluating the angular integral and expanding around ξ=0\xi=0 we get

δSBct(0)=Ω116G4[1sin(ξ)sin(ξ)2]×[hii(0)+R2sin2ξhii(2)+R3sin3ξhii(3)]\delta S^{ct}_{B}(\partial^{0})=-\frac{\Omega_{1}}{16G_{4}}\bigg{[}\frac{1}{\sin{\xi}}-\frac{\sin{\xi}}{2}\bigg{]}\times\bigg{[}h^{(0)}_{ii}+R^{2}\sin^{2}{\xi}h^{(2)}_{ii}+R^{3}\sin^{3}{\xi}h^{(3)}_{ii}\bigg{]} (227)

We now evaluate contributions from the subleading one derivative terms in the Taylor expansion of hij(0)(z)h^{(0)}_{ij}(z)

δSBct(1)=R8G402πcos2ξsin(ξ)[x^kkhii(0)x^ix^jx^kkhij(0)].\delta S^{ct}_{B}(\partial^{1})=-\frac{R}{8G_{4}}\int^{2\pi}_{0}\frac{\cos^{2}{\xi}}{\sin{\xi}}\bigg{[}\hat{x}^{k}\partial_{k}h^{(0)}_{ii}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\partial_{k}h^{(0)}_{ij}\bigg{]}. (228)

Using the angular integrals (211)(\ref{eq:AIntodd}) we can deduce δSBct(1)\delta S^{ct}_{B}(\partial^{1}) vanishes. The next leading two derivative contribution is

δSBct(2)=R28G402πcos3ξsin(ξ)[x^kx^lklhii(0)x^ix^jx^kx^lklhij(0)]\displaystyle\delta S^{ct}_{B}(\partial^{2})=-\frac{R^{2}}{8G_{4}}\int^{2\pi}_{0}\frac{\cos^{3}{\xi}}{\sin{\xi}}\bigg{[}\hat{x}^{k}\hat{x}^{l}\partial_{k}\partial_{l}h^{(0)}_{ii}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}\partial_{k}\partial_{l}h^{(0)}_{ij}\bigg{]} (229)

After evaluating the angular integrals we obtain

δSBct(2)=R2Ω1128G4cos3ξsin(ξ)[3jjhii(0)2ijhij(0)].\displaystyle\delta S^{ct}_{B}(\partial^{2})=-\frac{R^{2}\Omega_{1}}{128G_{4}}\frac{\cos^{3}{\xi}}{\sin{\xi}}\bigg{[}3\partial_{j}\partial_{j}h^{(0)}_{ii}-2\partial_{i}\partial_{j}h^{(0)}_{ij}\bigg{]}. (230)

Combining the variations of the regularized entanglement entropy and the variation of the counterterms we get the following. For 0\partial^{0} terms we have

δSBreg(0)+δSBct(0)\displaystyle\delta S^{reg}_{B}(\partial^{0})+\delta S^{ct}_{B}(\partial^{0}) =Ω18G4[12sin(ξ)hii(0)+23R2hii(2)+38R3hii(3)]\displaystyle=\frac{\Omega_{1}}{8G_{4}}\bigg{[}\frac{1}{2\sin{\xi}}h^{(0)}_{ii}+\frac{2}{3}R^{2}h^{(2)}_{ii}+\frac{3}{8}R^{3}h^{(3)}_{ii}\bigg{]}
Ω116G41sin(ξ)hii(0),\displaystyle-\frac{\Omega_{1}}{16G_{4}}\frac{1}{\sin{\xi}}h^{(0)}_{ii}, (231)

and for 2\partial^{2} terms have

δSBreg(2)+δSBct2)\displaystyle\delta S^{reg}_{B}(\partial^{2})+\delta S^{ct}_{B}\partial^{2}) =R2Ω132G4[(43+34sin(ξ))jjhii(0)+(4312sin(ξ))ijhij(0)]\displaystyle=\frac{R^{2}\Omega_{1}}{32G_{4}}\bigg{[}\Big{(}-\frac{4}{3}+\frac{3}{4\sin{\xi}}\Big{)}\partial_{j}\partial_{j}h^{(0)}_{ii}+\Big{(}\frac{4}{3}-\frac{1}{2\sin{\xi}}\Big{)}\partial_{i}\partial_{j}h^{(0)}_{ij}\bigg{]}
R2Ω1128G41sin(ξ)[3jjhii(0)2ijhij(0)].\displaystyle-\frac{R^{2}\Omega_{1}}{128G_{4}}\frac{1}{\sin{\xi}}\bigg{[}3\partial_{j}\partial_{j}h^{(0)}_{ii}-2\partial_{i}\partial_{j}h^{(0)}_{ij}\bigg{]}. (232)

Gathering all terms together we obtain the variation of renormalized entanglement entropy,

δSBren=Ω18G4[R23jjhii(0)+R23ijhij(0)+23R2hii(2)+38R3hii(3)].\displaystyle\delta S^{ren}_{B}=\frac{\Omega_{1}}{8G_{4}}\bigg{[}-\frac{R^{2}}{3}\partial_{j}\partial_{j}h^{(0)}_{ii}+\frac{R^{2}}{3}\partial_{i}\partial_{j}h^{(0)}_{ij}+\frac{2}{3}R^{2}h^{(2)}_{ii}+\frac{3}{8}R^{3}h^{(3)}_{ii}\bigg{]}. (233)

Then from (93)(\ref{eq:h2toh0}) we can express h(2)h^{(2)} in terms of h(0)h^{(0)} and the variation of the renormalized entanglement entropy becomes

δSBren\displaystyle\delta S^{ren}_{B} =Ω18G4[R23jjhii(0)+R23ijhij(0)+23R2(12jjhii(0)12ijhij(0))+38R3hii(3)]\displaystyle=\frac{\Omega_{1}}{8G_{4}}\bigg{[}-\frac{R^{2}}{3}\partial_{j}\partial_{j}h^{(0)}_{ii}+\frac{R^{2}}{3}\partial_{i}\partial_{j}h^{(0)}_{ij}+\frac{2}{3}R^{2}\Big{(}\frac{1}{2}\partial_{j}\partial_{j}h^{(0)}_{ii}-\frac{1}{2}\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)}+\frac{3}{8}R^{3}h^{(3)}_{ii}\bigg{]}
=3R3Ω164G4hii(3),\displaystyle=\frac{3R^{3}\Omega_{1}}{64G_{4}}h^{(3)}_{ii}, (234)

which is the result stated in (95)(\ref{eq:dSrenoddd}) for d=3d=3.

A.3 Explicit Variation in d=5d=5

Following the same approaches as in the section above, we continue the calculation from (84)(\ref{eq:IntSregwcoord}) for d=5d=5. The variation of the regularized entanglement entropy to leading order of the near boundary approximation, the zero derivative terms in the Taylor expansion give

δSBreg(0)=Ω38G6ξπ/2𝑑u[cos3usin4u(1cos2u4)hii].\displaystyle\delta S^{reg}_{B}(\partial^{0})=\frac{\Omega_{3}}{8G_{6}}\int_{\xi}^{\pi/2}du\bigg{[}\frac{\cos^{3}{u}}{\sin^{4}{u}}\bigg{(}1-\frac{\cos^{2}{u}}{4}\bigg{)}h_{ii}\bigg{]}. (235)

Using the Fefferman Graham expansion and evaluating the uu integral we get

δSBreg(0)=\displaystyle\delta S^{reg}_{B}(\partial^{0})= Ω38G6[(14sin3ξ12sinξ)hii(0)+(34sinξ43)R2hii(2)+815R4hii(4)+524R5hii(5)].\displaystyle\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}\frac{1}{4\sin^{3}\xi}-\frac{1}{2\sin\xi}\bigg{)}h^{(0)}_{ii}+\bigg{(}\frac{3}{4\sin\xi}-\frac{4}{3}\bigg{)}R^{2}h^{(2)}_{ii}+\frac{8}{15}R^{4}h^{(4)}_{ii}+\frac{5}{24}R^{5}h^{(5)}_{ii}\bigg{]}. (236)

The two derivative terms give

δSBreg(2)=\displaystyle\delta S^{reg}_{B}(\partial^{2})= R2Ω38G6ξπ/2𝑑ucos5u2!sin4u[14kkhiicos2u24(kkhii+2ijhij)]\displaystyle\frac{R^{2}\Omega_{3}}{8G_{6}}\int_{\xi}^{\pi/2}du\frac{\cos^{5}u}{2!\sin^{4}u}\bigg{[}\frac{1}{4}\partial_{k}\partial_{k}h_{ii}-\frac{\cos^{2}u}{24}\big{(}\partial_{k}\partial_{k}h_{ii}+2\partial_{i}\partial_{j}h_{ij}\big{)}\bigg{]} (237)

Using the Fefferman Graham expansion and evaluating the uu integral we get

δSBreg(2)=\displaystyle\delta S^{reg}_{B}(\partial^{2})= Ω38G6[(5144sin3ξ316sinξ+29)R2kkhii(0)+(172sin3ξ+18sinξ29)R2ijhij(0)\displaystyle\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}\frac{5}{144\sin^{3}\xi}-\frac{3}{16\sin\xi}+\frac{2}{9}\bigg{)}R^{2}\partial_{k}\partial_{k}h_{ii}^{(0)}+\bigg{(}-\frac{1}{72\sin^{3}\xi}+\frac{1}{8\sin\xi}-\frac{2}{9}\bigg{)}R^{2}\partial_{i}\partial_{j}h_{ij}^{(0)}
(548sinξ415)R4kkhii(2)+(124sinξ+215)R4ijhij(2)].\displaystyle-\bigg{(}\frac{5}{48\sin\xi}-\frac{4}{15}\bigg{)}R^{4}\partial_{k}\partial_{k}h_{ii}^{(2)}+\bigg{(}-\frac{1}{24\sin\xi}+\frac{2}{15}\bigg{)}R^{4}\partial_{i}\partial_{j}h_{ij}^{(2)}\bigg{]}. (238)

The four derivative terms are

δSBreg(4)=\displaystyle\delta S^{reg}_{B}(\partial^{4})= R4Ω38G6ξπ/2𝑑ucos7u4!sin4u[18kkllhiicos2u64(kkllhii+4kkijhij)].\displaystyle\frac{R^{4}\Omega_{3}}{8G_{6}}\int_{\xi}^{\pi/2}du\frac{\cos^{7}u}{4!\sin^{4}u}\bigg{[}\frac{1}{8}\partial_{k}\partial_{k}\partial_{l}\partial_{l}h_{ii}-\frac{\cos^{2}u}{64}\big{(}\partial_{k}\partial_{k}\partial_{l}\partial_{l}h_{ii}+4\partial_{k}\partial_{k}\partial_{i}\partial_{j}h_{ij}\big{)}\bigg{]}. (239)

Using the Fefferman Graham expansion and evaluating the uu integral we get

δSBreg(4)=\displaystyle\delta S^{reg}_{B}(\partial^{4})= Ω38G6[(74608sin3ξ5384sinξ+145)R4kkllhii(0)\displaystyle\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}\frac{7}{4608\sin^{3}\xi}-\frac{5}{384\sin\xi}+\frac{1}{45}\bigg{)}R^{4}\partial_{k}\partial_{k}\partial_{l}\partial_{l}h_{ii}^{(0)}
+(11152sin3ξ+196sinξ145)R4kkijhij(0)].\displaystyle+\bigg{(}-\frac{1}{1152\sin^{3}\xi}+\frac{1}{96\sin\xi}-\frac{1}{45}\bigg{)}R^{4}\partial_{k}\partial_{k}\partial_{i}\partial_{j}h_{ij}^{(0)}\bigg{]}. (240)

We have thus obtained all the divergent and finite terms for the variation of regularized entanglement entropy up to R5R^{5}. Note that for the δSBreg\delta S^{reg}_{B} only even derivatives survive the angular integrals. This is no longer the case for the counterterms δSBct\delta S^{ct}_{B} as some terms in (87)(\ref{eq:ExdSCt}) contain an odd number of directional vectors x^\hat{x}.

For the variation of the counterterms, (87)(\ref{eq:ExdSCt}), in the near boundary approximation, the leading order zero derivative terms are

δSBct(0)\displaystyle\delta S^{ct}_{B}(\partial^{0}) =Ω38G6[(14sin3ξ+12sinξ)hii(0)34sinξR2hii(2)].\displaystyle=\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}-\frac{1}{4\sin^{3}\xi}+\frac{1}{2\sin\xi}\bigg{)}h_{ii}^{(0)}-\frac{3}{4\sin\xi}R^{2}h_{ii}^{(2)}\bigg{]}. (241)

The one derivative terms comes from the variation of the extrinsic curvature, corresponding to the last terms in (87)(\ref{eq:ExdSCt}). Note that one derivative means up to and including the first derivative terms in the Taylor expansion:

δSBct(1)\displaystyle\delta S^{ct}_{B}(\partial^{1}) =124G6S3𝑑Ω3r3z[x^kx^lklhiix^ix^jx^kx^lklhij].\displaystyle=\frac{1}{24G_{6}}\int_{S^{3}}d\Omega_{3}\frac{r^{3}}{z}\bigg{[}\hat{x}^{k}\hat{x}^{l}\partial_{k}\partial_{l}h_{ii}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}\partial_{k}\partial_{l}h_{ij}\bigg{]}. (242)

After the integration only the following terms remain

δSBct(1)\displaystyle\delta S^{ct}_{B}(\partial^{1}) =Ω38G6[572sinξR2kkhii(0)136sinξR2ijhij(0)].\displaystyle=\frac{\Omega_{3}}{8G_{6}}\bigg{[}\frac{5}{72\sin\xi}R^{2}\partial_{k}\partial_{k}h_{ii}^{(0)}-\frac{1}{36\sin\xi}R^{2}\partial_{i}\partial_{j}h_{ij}^{(0)}\bigg{]}. (243)

Similar procedures are used for higher derivative terms. The order two derivative terms are

δSBct(2)=\displaystyle\delta S^{ct}_{B}(\partial^{2})= Ω38G6[(5144sin3ξ+17144sinξ)R2kkhii(0)\displaystyle\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}-\frac{5}{144\sin^{3}\xi}+\frac{17}{144\sin\xi}\bigg{)}R^{2}\partial_{k}\partial_{k}h_{ii}^{(0)} (244)
+(+172sin3ξ772sinξ)R2ijhij(0)548sinξR4kkhii(2)+124sinξR4ijhij(2)].\displaystyle+\bigg{(}+\frac{1}{72\sin^{3}\xi}-\frac{7}{72\sin\xi}\bigg{)}R^{2}\partial_{i}\partial_{j}h_{ij}^{(0)}-\frac{5}{48\sin\xi}R^{4}\partial_{k}\partial_{k}h_{ii}^{(2)}+\frac{1}{24\sin\xi}R^{4}\partial_{i}\partial_{j}h_{ij}^{(2)}\bigg{]}.

The order three derivative terms are

δSBct(3)\displaystyle\delta S^{ct}_{B}(\partial^{3}) =124G6S3𝑑Ω3r53!z[x^kx^lx^px^qklpqhiix^ix^jx^kx^lx^px^qklpqhij]\displaystyle=\frac{1}{24G_{6}}\int_{S^{3}}d\Omega_{3}\frac{r^{5}}{3!z}\bigg{[}\hat{x}^{k}\hat{x}^{l}\hat{x}^{p}\hat{x}^{q}\partial_{k}\partial_{l}\partial_{p}\partial_{q}h_{ii}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}\hat{x}^{p}\hat{x}^{q}\partial_{k}\partial_{l}\partial_{p}\partial_{q}h_{ij}\bigg{]} (245)

Since only integrals with even directional vectors are non vanishing, there is no term of the form 3h(0)\partial^{3}h^{(0)}. The remaining relevant terms are

δSBct(3)\displaystyle\delta S^{ct}_{B}(\partial^{3}) =Ω38G6[71152sinξR4kkllhii(0)1288sinξR4kkijhij(0)].\displaystyle=\frac{\Omega_{3}}{8G_{6}}\bigg{[}\frac{7}{1152\sin\xi}R^{4}\partial_{k}\partial_{k}\partial_{l}\partial_{l}h_{ii}^{(0)}-\frac{1}{288\sin\xi}R^{4}\partial_{k}\partial_{k}\partial_{i}\partial_{j}h_{ij}^{(0)}\bigg{]}. (246)

and the four derivative terms

δSBct(4)\displaystyle\delta S^{ct}_{B}(\partial^{4}) =Ω38G6[(74608sin3ξ+174608sinξ)R4kkllhii(0)\displaystyle=-\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}-\frac{7}{4608\sin^{3}\xi}+\frac{17}{4608\sin\xi}\bigg{)}R^{4}\partial_{k}\partial_{k}\partial_{l}\partial_{l}h_{ii}^{(0)}
+(11152sin3ξ1144sinξ)R4kkijhij(0)].\displaystyle+\bigg{(}\frac{1}{1152\sin^{3}\xi}-\frac{1}{144\sin\xi}\bigg{)}R^{4}\partial_{k}\partial_{k}\partial_{i}\partial_{j}h_{ij}^{(0)}\bigg{]}. (247)

We have thence obtained all the relevant counterterms.

For notational simplicity we express hii=hh_{ii}=h, kk=2\partial_{k}\partial_{k}=\partial^{2} and ijhij=h(2)\partial_{i}\partial_{j}h_{ij}=h(\partial^{2}). To compute the renormalized entanglement entropy we arrange all the relevant terms at order RnR^{n}:

Order R0R^{0}

δSren(R0)=\displaystyle\delta S_{ren}(R^{0})= Ω38G6[(14sin3ξ12sinξ)h(0)+(14sin3ξ+12sinξ)h(0)]\displaystyle\frac{\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}\frac{1}{4\sin^{3}\xi}-\frac{1}{2\sin\xi}\bigg{)}h^{(0)}+\bigg{(}-\frac{1}{4\sin^{3}\xi}+\frac{1}{2\sin\xi}\bigg{)}h^{(0)}\bigg{]} (248)

Order R2R^{2}

δSren(R2)=R2Ω38G6[(34sinξ43)h(2)+(5144sin3ξ316sinξ+29)2h(0)\displaystyle\delta S_{ren}(R^{2})=\frac{R^{2}\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}\frac{3}{4\sin\xi}-\frac{4}{3}\bigg{)}h^{(2)}+\bigg{(}\frac{5}{144\sin^{3}\xi}-\frac{3}{16\sin\xi}+\frac{2}{9}\bigg{)}\partial^{2}h^{(0)}
+(172sin3ξ+18sinξ29)h(0)(2)34sinξh(2)+572sinξ2h(0)136sinξh(0)(2)\displaystyle+\bigg{(}-\frac{1}{72\sin^{3}\xi}+\frac{1}{8\sin\xi}-\frac{2}{9}\bigg{)}h^{(0)}(\partial^{2})-\frac{3}{4\sin\xi}h^{(2)}+\frac{5}{72\sin\xi}\partial^{2}h^{(0)}-\frac{1}{36\sin\xi}h^{(0)}(\partial^{2})
+(5144sin3ξ+17144sinξ)2h(0)+(+172sin3ξ772sinξ)h(0)(2)]\displaystyle+\bigg{(}-\frac{5}{144\sin^{3}\xi}+\frac{17}{144\sin\xi}\bigg{)}\partial^{2}h^{(0)}+\bigg{(}+\frac{1}{72\sin^{3}\xi}-\frac{7}{72\sin\xi}\bigg{)}h^{(0)}(\partial^{2})\bigg{]} (249)

Order R4R^{4}

δSren(R4)=R4Ω38G6[815h(4)+(548sinξ415)2h(2)+(124sinξ+215)h(2)(2)\displaystyle\delta S_{ren}(R^{4})=\frac{R^{4}\Omega_{3}}{8G_{6}}\bigg{[}\frac{8}{15}h^{(4)}+\bigg{(}\frac{5}{48\sin\xi}-\frac{4}{15}\bigg{)}\partial^{2}h^{(2)}+\bigg{(}-\frac{1}{24\sin\xi}+\frac{2}{15}\bigg{)}h^{(2)}(\partial^{2})
+(74608sin3ξ5384sinξ+145)22h(0)+(11152sin3ξ+196sinξ145)2h(0)(2)\displaystyle+\bigg{(}\frac{7}{4608\sin^{3}\xi}-\frac{5}{384\sin\xi}+\frac{1}{45}\bigg{)}\partial^{2}\partial^{2}h^{(0)}+\bigg{(}-\frac{1}{1152\sin^{3}\xi}+\frac{1}{96\sin\xi}-\frac{1}{45}\bigg{)}\partial^{2}h^{(0)}(\partial^{2})
548sinξ2h(2)+124sinξh(2)(2)+71152sinξ22h(0)1288sinξ2h(0)(2)\displaystyle-\frac{5}{48\sin\xi}\partial^{2}h^{(2)}+\frac{1}{24\sin\xi}h^{(2)}(\partial^{2})+\frac{7}{1152\sin\xi}\partial^{2}\partial^{2}h^{(0)}-\frac{1}{288\sin\xi}\partial^{2}h^{(0)}(\partial^{2})
+(74608sin3ξ+144sinξ)22h(0)+(11152sin3ξ1144sinξ)2h(0)(2)]\displaystyle+\bigg{(}-\frac{7}{4608\sin^{3}\xi}+\frac{1}{44\sin\xi}\bigg{)}\partial^{2}\partial^{2}h^{(0)}+\bigg{(}\frac{1}{1152\sin^{3}\xi}-\frac{1}{144\sin\xi}\bigg{)}\partial^{2}h^{(0)}(\partial^{2})\bigg{]} (250)

Order R5R^{5}

δSren(0)=\displaystyle\delta S_{ren}(\partial^{0})= Ω38G6524R5hii(5)\displaystyle\frac{\Omega_{3}}{8G_{6}}\frac{5}{24}R^{5}h^{(5)}_{ii} (251)

Using (93)(\ref{eq:h2toh0}) and(94)(\ref{eq:h4toh2}) to express all the higher order term in the Fefferman-Graham expansion in terms of lower order ones, we find that below order R5R^{5} the variation of renormalized entanglement entropy is zero. More explicitly for each orders we have

Order R2R^{2}

δSren(R2)=R2Ω38G6[(18sinξ29)(2h(0)h(0)(2))+(5144sin3ξ316sinξ+29)2h(0)\displaystyle\delta S_{ren}(R^{2})=\frac{R^{2}\Omega_{3}}{8G_{6}}\bigg{[}\bigg{(}\frac{1}{8\sin\xi}-\frac{2}{9}\bigg{)}\big{(}\partial^{2}h^{(0)}-h^{(0)}(\partial^{2})\big{)}+\bigg{(}\frac{5}{144\sin^{3}\xi}-\frac{3}{16\sin\xi}+\frac{2}{9}\bigg{)}\partial^{2}h^{(0)}
+(172sin3ξ+18sinξ29)h(0)(2)18sinξ(2h(0)h(0)(2))+572sinξ2h(0)\displaystyle+\bigg{(}-\frac{1}{72\sin^{3}\xi}+\frac{1}{8\sin\xi}-\frac{2}{9}\bigg{)}h^{(0)}(\partial^{2})-\frac{1}{8\sin\xi}\big{(}\partial^{2}h^{(0)}-h^{(0)}(\partial^{2})\big{)}+\frac{5}{72\sin\xi}\partial^{2}h^{(0)}
136sinξh(0)(2)+(5144sin3ξ+17144sinξ)2h(0)+(+172sin3ξ772sinξ)h(0)(2)]\displaystyle-\frac{1}{36\sin\xi}h^{(0)}(\partial^{2})+\bigg{(}-\frac{5}{144\sin^{3}\xi}+\frac{17}{144\sin\xi}\bigg{)}\partial^{2}h^{(0)}+\bigg{(}+\frac{1}{72\sin^{3}\xi}-\frac{7}{72\sin\xi}\bigg{)}h^{(0)}(\partial^{2})\bigg{]}
δSren(R2)=0\displaystyle\delta S_{ren}(R^{2})=0 (252)

Order R4R^{4}

δSren(R4)=R4Ω38G6[215(2h(2)h(2)(2))(548sinξ415)2h(2)+(124sinξ+215)h(2)(2)\displaystyle\delta S_{ren}(R^{4})=\frac{R^{4}\Omega_{3}}{8G_{6}}\bigg{[}\frac{2}{15}\big{(}\partial^{2}h^{(2)}-h^{(2)}(\partial^{2})\big{)}-\bigg{(}\frac{5}{48\sin\xi}-\frac{4}{15}\bigg{)}\partial^{2}h^{(2)}+\bigg{(}-\frac{1}{24\sin\xi}+\frac{2}{15}\bigg{)}h^{(2)}(\partial^{2})
+(74608sin3ξ5384sinξ+145)22h(0)+(11152sin3ξ+196sinξ145)2h(0)(2)\displaystyle+\bigg{(}\frac{7}{4608\sin^{3}\xi}-\frac{5}{384\sin\xi}+\frac{1}{45}\bigg{)}\partial^{2}\partial^{2}h^{(0)}+\bigg{(}-\frac{1}{1152\sin^{3}\xi}+\frac{1}{96\sin\xi}-\frac{1}{45}\bigg{)}\partial^{2}h^{(0)}(\partial^{2})
548sinξ2h(2)+124sinξh(2)(2)+71152sinξ22h(0)1288sinξ2h(0)(2)\displaystyle-\frac{5}{48\sin\xi}\partial^{2}h^{(2)}+\frac{1}{24\sin\xi}h^{(2)}(\partial^{2})+\frac{7}{1152\sin\xi}\partial^{2}\partial^{2}h^{(0)}-\frac{1}{288\sin\xi}\partial^{2}h^{(0)}(\partial^{2})
+(74608sin3ξ+1144sinξ)22h(0)+(11152sin3ξ1144sinξ)2h(0)(2)],\displaystyle+\bigg{(}-\frac{7}{4608\sin^{3}\xi}+\frac{1}{144\sin\xi}\bigg{)}\partial^{2}\partial^{2}h^{(0)}+\bigg{(}\frac{1}{1152\sin^{3}\xi}-\frac{1}{144\sin\xi}\bigg{)}\partial^{2}h^{(0)}(\partial^{2})\bigg{]}, (253)

gathering all the terms, it simplifies to

δSren(R4)=\displaystyle\delta S_{ren}(R^{4})= R4Ω38G6[1452(2h(0)h(0)(2))\displaystyle\frac{R^{4}\Omega_{3}}{8G_{6}}\bigg{[}-\frac{1}{45}\partial^{2}\big{(}\partial^{2}h^{(0)}-h^{(0)}(\partial^{2})\big{)}
14522h(0)1452h(0)(2)]\displaystyle\frac{1}{45}\partial^{2}\partial^{2}h^{(0)}-\frac{1}{45}\partial^{2}h^{(0)}(\partial^{2})\bigg{]}
δSren(R4)=\displaystyle\delta S_{ren}(R^{4})= 0\displaystyle 0 (254)

Order R5R^{5}

This is the only order 5\leq 5 that is non-vanishing,

δSren(0)=\displaystyle\delta S_{ren}(\partial^{0})= Ω38G6524R5hii(5),\displaystyle\frac{\Omega_{3}}{8G_{6}}\frac{5}{24}R^{5}h^{(5)}_{ii}, (255)

which matches with (95)(\ref{eq:dSrenoddd}) for d=5d=5.

A.4 Renormalized Weyl Integrals

This appendix provides the calculation details for section 3.4. In (108)(\ref{eq:IntW}) and (109)(\ref{eq:IntdW}), the Weyl integrals are given in terms of the Riemann tensor of the boundary of AdS, μνρσ\mathcal{R}_{\mu\nu\rho\sigma}, and we need to expand μνρσ\mathcal{R}_{\mu\nu\rho\sigma} into linear perturbation hμνh_{\mu\nu}. For titj\mathcal{R}_{titj}, we have the following expression

titj\displaystyle\mathcal{R}_{titj} =12(tjhti+tihtjtthijijhtt)\displaystyle=\frac{1}{2}\left(\partial_{t}\partial_{j}h_{ti}+\partial_{t}\partial_{i}h_{tj}-\partial_{t}\partial_{t}h_{ij}-\partial_{i}\partial_{j}h_{tt}\right) (256)
titj\displaystyle\mathcal{R}_{titj} =(1+r2x^kx^l2kl)12(tjhti(0)+tihtj(0)tthij(0)ijhtt(0))\displaystyle=\left(1+\frac{r^{2}\hat{x}^{k}\hat{x}^{l}}{2}\partial_{k}\partial_{l}\right)\frac{1}{2}\left(\partial_{t}\partial_{j}h^{(0)}_{ti}+\partial_{t}\partial_{i}h^{(0)}_{tj}-\partial_{t}\partial_{t}h^{(0)}_{ij}-\partial_{i}\partial_{j}h^{(0)}_{tt}\right)
+z22(tjhti(2)+tihtj(2)tthij(2)ijhtt(2)).\displaystyle+\frac{z^{2}}{2}\left(\partial_{t}\partial_{j}h^{(2)}_{ti}+\partial_{t}\partial_{i}h^{(2)}_{tj}-\partial_{t}\partial_{t}h^{(2)}_{ij}-\partial_{i}\partial_{j}h^{(2)}_{tt}\right). (257)

The d=3d=3 integral

For d=3d=3, we do not need the subleading term in the Taylor expansion of the metric perturbation as in (88)(\ref{eq:xi0exp}). Also in d=3d=3 the boundary integral (109)(\ref{eq:IntdW}) is vanishing in the limit of ξ0\xi\rightarrow 0. After we substitute (257)(\ref{eq:Rbar}) into (108)(\ref{eq:IntW}) the renormalized Weyl integral 𝒲\mathcal{W} is mixed with different orders in the Fefferman-Graham expansion. Explicitly we have

𝒲\displaystyle\mathcal{W} =ξπ2duS1dΩ1[3R3cosusin3u2htt(3)+3R3cos3usinux^ix^j2(htt(3)ηijhij(3))\displaystyle=\int_{\xi}^{\frac{\pi}{2}}du\int_{S^{1}}d\Omega_{1}\bigg{[}-\frac{3R^{3}\cos u\sin^{3}u}{2}h^{(3)}_{tt}+\frac{3R^{3}\cos^{3}u\sin u\hat{x}^{i}\hat{x}^{j}}{2}\left(h^{(3)}_{tt}\eta_{ij}-h^{(3)}_{ij}\right) (258)
+R2cos3ux^ix^j2(tjhti(0)+tihtj(0)tthij(0)ijhtt(0)+2htt(2)ηij2hij(2))].\displaystyle+\frac{R^{2}\cos^{3}u\hat{x}^{i}\hat{x}^{j}}{2}\left(\partial_{t}\partial_{j}h^{(0)}_{ti}+\partial_{t}\partial_{i}h^{(0)}_{tj}-\partial_{t}\partial_{t}h^{(0)}_{ij}-\partial_{i}\partial_{j}h^{(0)}_{tt}+2h^{(2)}_{tt}\eta_{ij}-2h^{(2)}_{ij}\right)\bigg{]}.

After integrating over the circle we obtain

𝒲\displaystyle\mathcal{W} =Ω1[3R38htt(3)+3R316(2htt(3)hii(3))\displaystyle=\Omega_{1}\bigg{[}-\frac{3R^{3}}{8}h^{(3)}_{tt}+\frac{3R^{3}}{16}\left(2h^{(3)}_{tt}-h^{(3)}_{ii}\right) (259)
+R26(2tihti(0)tthii(0)iihtt(0)+4htt(2)hii(2))].\displaystyle+\frac{R^{2}}{6}\left(2\partial_{t}\partial_{i}h^{(0)}_{ti}-\partial_{t}\partial_{t}h^{(0)}_{ii}-\partial_{i}\partial_{i}h^{(0)}_{tt}+4h^{(2)}_{tt}-h^{(2)}_{ii}\right)\bigg{]}.

By solving the Einstein equation order by order in the Fefferman-Graham expansion, we can deduced h(n)h^{(n)} for n<dn<d from h(0)h^{(0)}. This gives

hii(2)=\displaystyle h^{(2)}_{ii}= 12(kkhii(0)ijhij(0))\displaystyle\frac{1}{2}\Big{(}\partial_{k}\partial_{k}h^{(0)}_{ii}-\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)} (260)
htt(2)=\displaystyle h^{(2)}_{tt}= 14(kkhii(0)ijhij(0)+tthii(0)+kkhtt(0)2tihti(0)).\displaystyle\frac{1}{4}\Big{(}\partial_{k}\partial_{k}h^{(0)}_{ii}-\partial_{i}\partial_{j}h^{(0)}_{ij}+\partial_{t}\partial_{t}h^{(0)}_{ii}+\partial_{k}\partial_{k}h^{(0)}_{tt}-2\partial_{t}\partial_{i}h^{(0)}_{ti}\Big{)}. (261)

Using the above two expression sfor hμν(2)h^{(2)}_{\mu\nu}, we can easily simplify the renormalized Weyl integral as

𝒲\displaystyle\mathcal{W} =3R3Ω116htt(3)\displaystyle=-\frac{3R^{3}\Omega_{1}}{16}h^{(3)}_{tt} (262)

which is the result stated in section 3.4.

The d=5d=5 integral

For d=5d=5, we need the subleading term in the Taylor expansion of the metric perturbation as in (88)(\ref{eq:xi0exp}). The relevant metric perturbation derivatives are

htt′′\displaystyle h_{tt}^{\prime\prime} =(1+r2x^kx^l2kl)2htt(2)+12z2htt(4)+20z3htt(5)\displaystyle=\left(1+\frac{r^{2}\hat{x}^{k}\hat{x}^{l}}{2}\partial_{k}\partial_{l}\right)2h^{(2)}_{tt}+12z^{2}h^{(4)}_{tt}+20z^{3}h^{(5)}_{tt} (263)
hμν\displaystyle h_{\mu\nu}^{\prime} =(1+r2x^kx^l2kl)2zhμν(2)+4z3hμν(4)+5z4hμν(5)\displaystyle=\left(1+\frac{r^{2}\hat{x}^{k}\hat{x}^{l}}{2}\partial_{k}\partial_{l}\right)2zh^{(2)}_{\mu\nu}+4z^{3}h^{(4)}_{\mu\nu}+5z^{4}h^{(5)}_{\mu\nu} (264)
μhνρ\displaystyle\partial_{\mu}h_{\nu\rho}^{\prime} =2zrx^kkμhνρ(2),\displaystyle=2zr\hat{x}^{k}\partial_{k}\partial_{\mu}h_{\nu\rho}^{(2)}, (265)

where represent the radial derivative z\partial_{z}. The renormalized Weyl integral 𝒲\mathcal{W} becomes

𝒲\displaystyle\mathcal{W} =S3dΩ3[8R415htt(4)15R524htt(5)\displaystyle=\int_{S^{3}}d\Omega_{3}\bigg{[}-\frac{8R^{4}}{15}h^{(4)}_{tt}-\frac{15R^{5}}{24}h^{(5)}_{tt} (266)
+(4R23x^ix^j4R25x^ix^jx^kx^lkl)\displaystyle+\left(-\frac{4R^{2}}{3}\hat{x}^{i}\hat{x}^{j}-\frac{4R^{2}}{5}\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\hat{x}^{l}\partial_{k}\partial_{l}\right)
×(tjhti(0)+tihtj(0)tthij(0)ijhtt(0)+2htt(2)ηij+2hij(2)ηtt)\displaystyle\quad\quad\times\left(\partial_{t}\partial_{j}h^{(0)}_{ti}+\partial_{t}\partial_{i}h^{(0)}_{tj}-\partial_{t}\partial_{t}h^{(0)}_{ij}-\partial_{i}\partial_{j}h^{(0)}_{tt}+2h^{(2)}_{tt}\eta_{ij}+2h^{(2)}_{ij}\eta_{tt}\right)
+4R415x^ix^j(tjhti(2)+tihtj(2)tthij(2)ijhtt(2)+4htt(4)ηij+4hij(4)ηtt)\displaystyle+\frac{4R^{4}}{15}\hat{x}^{i}\hat{x}^{j}\left(\partial_{t}\partial_{j}h^{(2)}_{ti}+\partial_{t}\partial_{i}h^{(2)}_{tj}-\partial_{t}\partial_{t}h^{(2)}_{ij}-\partial_{i}\partial_{j}h^{(2)}_{tt}+4h^{(4)}_{tt}\eta_{ij}+4h^{(4)}_{ij}\eta_{tt}\right)
+5R512x^ix^j(htt(5)ηij+hij(5)ηtt)\displaystyle+\frac{5R^{5}}{12}\hat{x}^{i}\hat{x}^{j}\left(h^{(5)}_{tt}\eta_{ij}+h^{(5)}_{ij}\eta_{tt}\right)
+16R415x^ix^kk(thti(2)ihtt(2))].\displaystyle+\frac{16R^{4}}{15}\hat{x}^{i}\hat{x}^{k}\partial_{k}\left(\partial_{t}h^{(2)}_{ti}-\partial_{i}h^{(2)}_{tt}\right)\bigg{]}.

After integrating this over the S3S^{3} using (215)(\ref{eq:OmegaxiInt}) we get

𝒲=Ω3[R23(\displaystyle\mathcal{W}=\Omega_{3}\bigg{[}\frac{R^{2}}{3}\Big{(} 2tihti(0)+tthii(0)+iihtt(0)8htt(2)+2hii(2))\displaystyle-2\partial_{t}\partial_{i}h^{(0)}_{ti}+\partial_{t}\partial_{t}h^{(0)}_{ii}+\partial_{i}\partial_{i}h^{(0)}_{tt}-8h^{(2)}_{tt}+2h^{(2)}_{ii}\Big{)} (267)
+R430(\displaystyle+\frac{R^{4}}{30}\Big{(} 6tkkihti(0)+ttkkhii(0)+2ttijhij(0)+3kkllhtt(0)\displaystyle-6\partial_{t}\partial_{k}\partial_{k}\partial_{i}h^{(0)}_{ti}+\partial_{t}\partial_{t}\partial_{k}\partial_{k}h^{(0)}_{ii}+2\partial_{t}\partial_{t}\partial_{i}\partial_{j}h^{(0)}_{ij}+3\partial_{k}\partial_{k}\partial_{l}\partial_{l}h^{(0)}_{tt}
22kkhtt(2)2tthii(2)+2kkhii(2)+4ijhij(2)+12tihti(2)\displaystyle-22\partial_{k}\partial_{k}h^{(2)}_{tt}-2\partial_{t}\partial_{t}h^{(2)}_{ii}+2\partial_{k}\partial_{k}h^{(2)}_{ii}+4\partial_{i}\partial_{j}h^{(2)}_{ij}+12\partial_{t}\partial_{i}h^{(2)}_{ti}
+16htt(4)8hii(4))\displaystyle+16h^{(4)}_{tt}-8h^{(4)}_{ii}\Big{)}
+R548(\displaystyle+\frac{R^{5}}{48}\Big{(} 10htt(5)5hii(5))].\displaystyle-10h^{(5)}_{tt}-5h^{(5)}_{ii}\Big{)}\bigg{]}.

Following the lower dimensional case, we need to related the terms of different orders in Fefferman-Grahm expansion to see the cancellation between divergent pieces. By solving the Einstein equations order by order in the Fefferman-Graham expansion, we can deduced h(n)h^{(n)} for n<dn<d from h(0)h^{(0)}. Hence,

hii(2)=\displaystyle h^{(2)}_{ii}= 16(kkhii(0)ijhij(0))\displaystyle\frac{1}{6}\Big{(}\partial_{k}\partial_{k}h^{(0)}_{ii}-\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)} (268)
htt(2)=\displaystyle h^{(2)}_{tt}= 124(kkhii(0)ijhij(0)+3tthii(0)+3kkhtt(0)6tihti(0))\displaystyle\frac{1}{24}\Big{(}\partial_{k}\partial_{k}h^{(0)}_{ii}-\partial_{i}\partial_{j}h^{(0)}_{ij}+3\partial_{t}\partial_{t}h^{(0)}_{ii}+3\partial_{k}\partial_{k}h^{(0)}_{tt}-6\partial_{t}\partial_{i}h^{(0)}_{ti}\Big{)} (269)
hti(2)=\displaystyle h^{(2)}_{ti}= 16(tihjj(0)tjhij(0)ijhtj(0)+kkhti(0))\displaystyle\frac{1}{6}\Big{(}\partial_{t}\partial_{i}h^{(0)}_{jj}-\partial_{t}\partial_{j}h^{(0)}_{ij}-\partial_{i}\partial_{j}h^{(0)}_{tj}+\partial_{k}\partial_{k}h^{(0)}_{ti}\Big{)} (270)
hii(4)=\displaystyle h^{(4)}_{ii}= 14(kkhii(2)ijhij(2))\displaystyle\frac{1}{4}\Big{(}\partial_{k}\partial_{k}h^{(2)}_{ii}-\partial_{i}\partial_{j}h^{(2)}_{ij}\Big{)} (271)
htt(4)=\displaystyle h^{(4)}_{tt}= 14(kkhtt(2)tthii(2))\displaystyle\frac{1}{4}\Big{(}\partial_{k}\partial_{k}h^{(2)}_{tt}-\partial_{t}\partial_{t}h^{(2)}_{ii}\Big{)} (272)
ijhij(2)=\displaystyle\partial_{i}\partial_{j}h^{(2)}_{ij}= 124(3kkllhii(0)3kkijhij(0)+ttkkhii(0)\displaystyle\frac{1}{24}\Big{(}3\partial_{k}\partial_{k}\partial_{l}\partial_{l}h^{(0)}_{ii}-3\partial_{k}\partial_{k}\partial_{i}\partial_{j}h^{(0)}_{ij}+\partial_{t}\partial_{t}\partial_{k}\partial_{k}h^{(0)}_{ii} (273)
3kkllhtt(0)+63tkkihti(0)4ttijhij(0))\displaystyle\quad-3\partial_{k}\partial_{k}\partial_{l}\partial_{l}h^{(0)}_{tt}+63\partial_{t}\partial_{k}\partial_{k}\partial_{i}h^{(0)}_{ti}-4\partial_{t}\partial_{t}\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)}
tihit(2)=\displaystyle\partial_{t}\partial_{i}h^{(2)}_{it}= 16(ttkkhii(0)ttijhij(0))\displaystyle\frac{1}{6}\Big{(}\partial_{t}\partial_{t}\partial_{k}\partial_{k}h^{(0)}_{ii}-\partial_{t}\partial_{t}\partial_{i}\partial_{j}h^{(0)}_{ij}\Big{)} (274)

Substituting the above expressions for hμν(n)h^{(n)}_{\mu\nu}, we can easily simplify the renormalized Weyl integral as

𝒲\displaystyle\mathcal{W} =5R5Ω316htt(5)\displaystyle=-\frac{5R^{5}\Omega_{3}}{16}h^{(5)}_{tt} (275)

which is the result stated in section 3.4.

A.5 Variations in AdS5AdS_{5}

Here we will fill in the computational details of section 3.5 to show that the divergences of the variation of regularized entanglement entropy and variation of the counterterms match. In (117)(\ref{eq:dSregd4}), the variation of regularized entanglement entropy was given in terms of both hμν(0)h^{(0)}_{\mu\nu} and hμν(2)h^{(2)}_{\mu\nu}. In order to compare with the counterterm we will first express hμν(2)h^{(2)}_{\mu\nu} as function of hμν(0)h^{(0)}_{\mu\nu}.

Since the perturbed metric of AdS5AdS_{5} satisfies the Einstein equation, the metric perturbation can be expanded and solved order by order in an asymptotic series. Using the results in deHaro2001 ,

gμν(2)=12(μν[g(0)]16[g(0)]gμν(0)).g^{(2)}_{\mu\nu}=-\frac{1}{2}\left(\mathcal{R}_{\mu\nu}[g^{(0)}]-\frac{1}{6}\mathcal{R}[g^{(0)}]g^{(0)}_{\mu\nu}\right). (276)

In d=4d=4, we only need to consider terms of order up to z2z^{2}, hence we have

hμν(2)=12(μν[η+h(0)]16[η+h(0)](ημν+hμν(0))).h^{(2)}_{\mu\nu}=-\frac{1}{2}\left(\mathcal{R}_{\mu\nu}[\eta+h^{(0)}]-\frac{1}{6}\mathcal{R}[\eta+h^{(0)}](\eta_{\mu\nu}+h^{(0)}_{\mu\nu})\right). (277)

Since the Ricci tensor of ημν\eta_{\mu\nu} vanishes, to first order of hh the Ricci tensor of g(0)g^{(0)} is just the first order variation. For our interests the relevant terms then become

sinθzhθθ(2)\displaystyle\frac{\sin\theta}{z}h^{(2)}_{\theta\theta} =sinθ2zδθθ+R2sinθ12zδ\displaystyle=-\frac{\sin\theta}{2z}\delta\mathcal{R}_{\theta\theta}+\frac{R^{2}\sin\theta}{12z}\delta\mathcal{R} (278)
1sinθzhϕϕ(2)\displaystyle\frac{1}{\sin\theta z}h^{(2)}_{\phi\phi} =12sinθzδϕϕ+R2sinθ12zδ\displaystyle=-\frac{1}{2\sin\theta z}\delta\mathcal{R}_{\phi\phi}+\frac{R^{2}\sin\theta}{12z}\delta\mathcal{R} (279)

Using this expression, we can write the divergent term of the regularized entanglement entropy in (118)(\ref{eq:dSregdivd4}) in terms of hμν(0)h^{(0)}_{\mu\nu}.

Now we need to evaluate the variation of the counter terms and check all the divergences are cancelled. The induced metric γ~\widetilde{\gamma} of the regularised entangling surface B~ϵ=B~|z=ϵ\partial\tilde{B}_{\epsilon}=\tilde{B}|_{z=\epsilon} is

ds2=R2ϵ2ϵ2(dθ2+sin2θdϕ2).ds^{2}=\frac{R^{2}-\epsilon^{2}}{\epsilon^{2}}\big{(}d\theta^{2}+\sin^{2}\theta d\phi^{2}\big{)}. (280)

Then the variation of the volume form is

δγ~\displaystyle\delta\sqrt{\widetilde{\gamma}} =12γ~γ~ijδγ~ij\displaystyle=\frac{1}{2}\sqrt{\widetilde{\gamma}}\widetilde{\gamma}^{ij}\delta\widetilde{\gamma}_{ij} (281)
=sinθ2(1ϵ2hθθ(0)+1ϵ2sin2θhϕϕ(0))\displaystyle=\frac{\sin\theta}{2}\bigg{(}\frac{1}{\epsilon^{2}}h^{(0)}_{\theta\theta}+\frac{1}{\epsilon^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}\bigg{)}

To calculate the variation of the counterterms we need to embed (B~,γ~)\big{(}\partial\tilde{B},\widetilde{\gamma}\big{)} into (AdS5|z=ϵ,G~)\big{(}AdS_{5}|_{z=\epsilon},\widetilde{G}\big{)} and find its unit normals which are

n1=dtϵ,n2=drϵ.\displaystyle n_{1}=\frac{dt}{\epsilon},\quad\quad\quad\quad n_{2}=\frac{dr}{\epsilon}. (282)

The extrinsic curvature KμνK_{\mu\nu} is defined by 12nγ~μν\frac{1}{2}\mathcal{L}_{n}\widetilde{\gamma}_{\mu\nu}. The trace of the extrinsic curvature is then

K=G~μνKμν=12γ~ijnγ~ij=nlnγ~\displaystyle K=\widetilde{G}^{\mu\nu}K_{\mu\nu}=\frac{1}{2}\widetilde{\gamma}^{ij}\mathcal{L}_{n}\widetilde{\gamma}_{ij}=\mathcal{L}_{n}\ln\sqrt{\widetilde{\gamma}} (283)

In time independent situations, K1K_{1} vanishes. The extrinsic curvature corresponding to the radial normal is

K2=ϵrlnr2sinθϵ2=2ϵr\displaystyle K_{2}=\epsilon\partial_{r}\ln\frac{r^{2}\sin\theta}{\epsilon^{2}}=\frac{2\epsilon}{r} (284)

Although we are only taking linear order of metric variation which leaves the direction of the normals unchanged, the coefficients of unit normals nan_{a} vary. Specifically for n2n_{2}

δn2=δ(nrr)=δ(1G~rrr)=δG~rr2G~rr32r\displaystyle\delta n_{2}=\delta(n^{r}\partial_{r})=\delta\left(\sqrt{\frac{1}{\widetilde{G}_{rr}}}\partial_{r}\right)=-\frac{\delta\widetilde{G}_{rr}}{2{\widetilde{G}_{rr}}^{\frac{3}{2}}}\partial_{r} (285)

The variation of K2K_{2} can be related to the variation of the metric gg as

δK2\displaystyle\delta K_{2} =δn2(lnγ~)+n2δ(lnγ~)\displaystyle=\delta n_{2}\big{(}\ln\sqrt{\widetilde{\gamma}}\big{)}+n_{2}\delta\big{(}\ln\sqrt{\widetilde{\gamma}}\big{)} (286)
=δG~rr2G~rr32rlnγ~+ϵr(12γ~ijδγ~ij)\displaystyle=-\frac{\delta\widetilde{G}_{rr}}{2{\widetilde{G}_{rr}}^{\frac{3}{2}}}\partial_{r}\ln\sqrt{\widetilde{\gamma}}+\epsilon\partial_{r}\bigg{(}\frac{1}{2}\widetilde{\gamma}^{ij}\delta\widetilde{\gamma}_{ij}\bigg{)}
=ϵrhrr(0)ϵr3hθθ(0)ϵr3sin2θhϕϕ(0)+ϵ2r2rhθθ(0)+ϵ2r2sin2θrhϕϕ(0)\displaystyle=-\frac{\epsilon}{r}h^{(0)}_{rr}-\frac{\epsilon}{r^{3}}h^{(0)}_{\theta\theta}-\frac{\epsilon}{r^{3}\sin^{2}\theta}h^{(0)}_{\phi\phi}+\frac{\epsilon}{2r^{2}}\partial_{r}h^{(0)}_{\theta\theta}+\frac{\epsilon}{2r^{2}\sin^{2}\theta}\partial_{r}h^{(0)}_{\phi\phi}

Keeping only the divergence, the structure of the variation of the third term in (69)(\ref{eq:IntSct}) is

δ(γ~k2)\displaystyle\delta\big{(}\sqrt{\widetilde{\gamma}}k^{2}\big{)} =δ(γ~)K22+γ~δ(K22)\displaystyle=\delta\big{(}\sqrt{\widetilde{\gamma}}\big{)}K_{2}^{2}+\sqrt{\widetilde{\gamma}}\delta\big{(}K_{2}^{2}\big{)} (287)

Separating the terms in (287)(\ref{eq:K226}),

δ(γ~)K22=\displaystyle\delta\big{(}\sqrt{\widetilde{\gamma}}\big{)}K_{2}^{2}=\ sinθ(2R2hθθ(0)+2R2sin2θhϕϕ(0))\displaystyle\sin\theta\Big{(}\frac{2}{R^{2}}h^{(0)}_{\theta\theta}+\frac{2}{R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}\Big{)} (288)
2γ~K2δK2=\displaystyle 2\sqrt{\widetilde{\gamma}}K_{2}\delta K_{2}=\ sinθ(4hrr(0)4R2hθθ(0)4R2sin2θhϕϕ(0)\displaystyle\sin\theta\Big{(}-4h^{(0)}_{rr}-\frac{4}{R^{2}}h^{(0)}_{\theta\theta}-\frac{4}{R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}
+2Rrhθθ(0)+2Rsin2θrhϕϕ(0))\displaystyle+\frac{2}{R}\partial_{r}h^{(0)}_{\theta\theta}+\frac{2}{R\sin^{2}\theta}\partial_{r}h^{(0)}_{\phi\phi}\Big{)} (289)

The remaining terms are the variation of Ricci scalar and projected Ricci tensor. Note that aa\mathcal{R}_{aa} in Fursaev_1995 ; Fursaev_2013 was given in a Euclidean setting. After Wick rotating the normal direction back to Lorentzian signature, we obtain

aa\displaystyle\mathcal{R}_{aa} =μν(in1μ)(in1ν)+μνn2μn2ν\displaystyle=\mathcal{R}_{\mu\nu}(in^{\mu}_{1})(in^{\nu}_{1})+\mathcal{R}_{\mu\nu}n^{\mu}_{2}n^{\nu}_{2}
aa\displaystyle\mathcal{R}_{aa} =z2(tt+rr).\displaystyle=z^{2}(-\mathcal{R}_{tt}+\mathcal{R}_{rr}). (290)

Again we use the fact that our unperturbed spacetime is flat so the variation of these terms is

δ(aa23)\displaystyle\delta\big{(}\mathcal{R}_{aa}-\frac{2}{3}\mathcal{R}\big{)} =δμνnaμnaν23δ\displaystyle=\delta\mathcal{R}_{\mu\nu}n_{a}^{\mu}n_{a}^{\nu}-\frac{2}{3}\delta\mathcal{R}
=z23δtt+z23δrr2z23(1r2δθθ+1r2sin2θδϕϕ)\displaystyle=-\frac{z^{2}}{3}\delta\mathcal{R}_{tt}+\frac{z^{2}}{3}\delta\mathcal{R}_{rr}-\frac{2z^{2}}{3}\bigg{(}\frac{1}{r^{2}}\delta\mathcal{R}_{\theta\theta}+\frac{1}{r^{2}\sin^{2}\theta}\delta\mathcal{R}_{\phi\phi}\bigg{)}
=z23δz2r2δθθz2r2sin2θδϕϕ\displaystyle=\frac{z^{2}}{3}\delta\mathcal{R}-\frac{z^{2}}{r^{2}}\delta\mathcal{R}_{\theta\theta}-\frac{z^{2}}{r^{2}\sin^{2}\theta}\delta\mathcal{R}_{\phi\phi} (291)

notice there is an abuse of notation where in the first line δ=G~μνδμν\delta\mathcal{R}={\widetilde{G}}^{\mu\nu}\delta\mathcal{R}_{\mu\nu} and in the last line δ=g(0)μνδμν\delta\mathcal{R}={g^{(0)}}^{\mu\nu}\delta\mathcal{R}_{\mu\nu}. Using (277)(\ref{eq:h2Rd4}) we can write (291)(\ref{eq:dRd4}) in terms of h(2)h^{(2)},

δμνnaμnaν23δ=2z2r2hθθ(2)+2z2r2sin2θhϕϕ(2)\displaystyle\delta\mathcal{R}_{\mu\nu}n_{a}^{\mu}n_{a}^{\nu}-\frac{2}{3}\delta\mathcal{R}=\frac{2z^{2}}{r^{2}}h^{(2)}_{\theta\theta}+\frac{2z^{2}}{r^{2}\sin^{2}\theta}h^{(2)}_{\phi\phi} (292)

The divergent contributions to the counterterms are

(δSBct)div=\displaystyle(\delta S^{ct}_{B})^{div}= 18G5S2dΩ2[12ϵ(hθθ(0)+1sin2θhϕϕ(0))+lnϵ2(1R2hθθ(0)+1R2sin2θhϕϕ(0)\displaystyle\frac{1}{8G_{5}}\int_{S^{2}}d\Omega_{2}\bigg{[}\frac{1}{2\epsilon}\Big{(}h^{(0)}_{\theta\theta}+\frac{1}{\sin^{2}\theta}h^{(0)}_{\phi\phi}\Big{)}+\frac{\ln\epsilon}{2}\Big{(}\frac{1}{R^{2}}h^{(0)}_{\theta\theta}+\frac{1}{R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}
2hrr(0)2R2hθθ(0)2R2sin2θhϕϕ(0)+1Rrhθθ(0)+1Rsin2θrhϕϕ(0)\displaystyle-2h^{(0)}_{rr}-\frac{2}{R^{2}}h^{(0)}_{\theta\theta}-\frac{2}{R^{2}\sin^{2}\theta}h^{(0)}_{\phi\phi}+\frac{1}{R}\partial_{r}h^{(0)}_{\theta\theta}+\frac{1}{R\sin^{2}\theta}\partial_{r}h^{(0)}_{\phi\phi}
2hθθ(2)2sin2θhϕϕ(2))]\displaystyle-2h^{(2)}_{\theta\theta}-\frac{2}{\sin^{2}\theta}h^{(2)}_{\phi\phi}\Big{)}\bigg{]} (293)

which matches with (118)(\ref{eq:dSregdivd4}).

Appendix B Asymptotic expansions and integrals

B.1 Dilatation Eigenfunction Expansion

Under dilatation transformation xμΩxμx^{\mu}\rightarrow\Omega x^{\mu}, the boundary metric transforms as

γμνΩ2γμν.\displaystyle\gamma_{\mu\nu}\rightarrow\Omega^{2}\gamma_{\mu\nu}. (294)

In terms of infinitesimal operator

γμν(1+ϵδD)γμν\displaystyle\gamma_{\mu\nu}\rightarrow(1+\epsilon\delta_{D})\gamma_{\mu\nu} (295)

where 1+ϵ=Ω1+\epsilon=\Omega. The dilatation operator for the boundary metric γ\gamma is then

δD=2ddxγμνδδγμν\displaystyle\delta_{D}=2\int d^{d}x\gamma_{\mu\nu}\frac{\delta}{\delta\gamma_{\mu\nu}} (296)

which replaces γμν\gamma_{\mu\nu} with 2γμν2\gamma_{\mu\nu} as the dilatation weight of the metric is 22. The dilatation operator in general contains all fields that transform non-trivially under dilatation. For our purposes we will actually only consider pure gravitational systems so the dilatation operator only contains the metric γμν\gamma_{\mu\nu}. In the radial gauge the extrinsic curvature depends only on γμν\gamma_{\mu\nu} and it curvature can be expanded in Fefferman-Graham coefficients as

Kμν[γ]\displaystyle K_{\mu\nu}[\gamma] =z2zγμν\displaystyle=-\frac{z}{2}\partial_{z}\gamma_{\mu\nu}
=z2gμν(0)z2gμν(4)++2d2g~μν(d)logz2g~μν(d)+2d2gμν(d)+\displaystyle=z^{-2}g_{\mu\nu}^{(0)}-z^{2}g_{\mu\nu}^{(4)}+\cdots+\frac{2-d}{2}\tilde{g}_{\mu\nu}^{(d)}\log z^{2}-\tilde{g}_{\mu\nu}^{(d)}+\frac{2-d}{2}g_{\mu\nu}^{(d)}+\cdots (297)

and in dilatation eigenfunction expansion

Kμν[γ]=K(0)μν[γ]+K(2)μν[γ]++K~(d)μν[γ]logz2+K(d)μν[γ]+\displaystyle K_{\mu\nu}[\gamma]=K_{(0)\;\mu\nu}[\gamma]+K_{(2)\;\mu\nu}[\gamma]+\cdots+\tilde{K}_{(d)\;\mu\nu}[\gamma]\log z^{2}+K_{(d)\;\mu\nu}[\gamma]+\cdots (298)

where the logarithmic terms are only present for even dd. The dilatation eigenfunctions transform according to their order: we have homogenous transformations for K(n<d)μνK_{(n<d)\;\mu\nu} and K~(d)μν\tilde{K}_{(d)\;\mu\nu} ,

δDK(n)μν=(2n)K(n)μν\displaystyle\delta_{D}K_{(n)\;\mu\nu}=(2-n)K_{(n)\;\mu\nu} (299)

and inhomogenous transformations for K(d)μνK_{(d)\;\mu\nu},

δDK(d)μν=(2d)K(d)μν2K~(d)μν.\displaystyle\delta_{D}K_{(d)\;\mu\nu}=(2-d)K_{(d)\;\mu\nu}-2\tilde{K}_{(d)\;\mu\nu}. (300)

The origin of the inhomogenous transformation will become obvious when we relate the two expansions. To do that we need to express the radial derivative in terms of functional derivative of γμν\gamma_{\mu\nu}

zz=zz|γμν=const+ddx2Kμν[γ]δδγμν.\displaystyle-z\partial_{z}=-z\partial_{z}|_{\gamma_{\mu\nu}=const}+\int d^{d}x2K_{\mu\nu}[\gamma]\frac{\delta}{\delta\gamma_{\mu\nu}}. (301)

Let us drop the first term as we are considering field that does not depend on zz explicitly. We know from (299)(\ref{eq:dKhom}) that the zeroth term in the dilatation eigenfunction expansion K(0)μν[γ]K_{(0)\;\mu\nu}[\gamma] is proportional to γμν\gamma_{\mu\nu} then comparing with the leading term in (297)(\ref{eq:KFGE}) we can deduce

K(0)μν[γ]=γμν.\displaystyle K_{(0)\;\mu\nu}[\gamma]=\gamma_{\mu\nu}. (302)

We see that expanding the extrinsic curvature in (301)(\ref{eq:radderK}) the radial derivative is related to the dilatation operator by

zz=δD+δ(2)+\displaystyle-z\partial_{z}=\delta_{D}+\delta_{(2)}+\cdots (303)

where

δ(n)=ddx2K(n)μν[γ]δδγμν.\displaystyle\delta_{(n)}=\int d^{d}x2K_{(n)\;\mu\nu}[\gamma]\frac{\delta}{\delta\gamma_{\mu\nu}}. (304)

Taylor expanding the K(n)μν[γ]K_{(n)\;\mu\nu}[\gamma] about z2gμν(0)z^{-2}g^{(0)}_{\mu\nu}

K(n)μν[γ]=K(n)μν[z2g(0)]+gρσ(2)δK(n)μνδγμν|γ=z2g(0)+\displaystyle K_{(n)\;\mu\nu}[\gamma]=K_{(n)\;\mu\nu}[z^{-2}g^{(0)}]+\int g^{(2)}_{\rho\sigma}\frac{\delta K_{(n)\;\mu\nu}}{\delta\gamma_{\mu\nu}}|_{\gamma=z^{-2}g^{(0)}}+\cdots (305)

Since K(n)μν[z2g(0)]K_{(n)\mu\nu}[z^{-2}g^{(0)}] are also dilatation eigenfunctions, we can rescale the metric to get rid of the implicit zz dependence. Using the integrated transformation of (299)(\ref{eq:dKhom}) for K(n<d)μνK_{(n<d)\;\mu\nu} and K~(d)μν\tilde{K}_{(d)\;\mu\nu},

K(n)μν[z2g(0)]=zn2K(n)μν[g(0)].\displaystyle K_{(n)\;\mu\nu}[z^{-2}g^{(0)}]=z^{n-2}K_{(n)\;\mu\nu}[g^{(0)}]. (306)

Now that we know at the leading order we can write the dilatation operator in terms of the radial derivative δDzz\delta_{D}\sim-z\partial_{z} for implicit zz dependence terms then,

zz(K~(d)μν[γ]logz2+K(d)μν[γ])\displaystyle-z\partial_{z}\left(\tilde{K}_{(d)\;\mu\nu}[\gamma]\log z^{2}+K_{(d)\;\mu\nu}[\gamma]\right) δD(K~(d)μν[γ]logz2+K(d)μν[γ]).\displaystyle\sim\delta_{D}\left(\tilde{K}_{(d)\;\mu\nu}[\gamma]\log z^{2}+K_{(d)\;\mu\nu}[\gamma]\right). (307)

Note the bracket term depends on zz through γ\gamma only because of the diffeomorphism invariance of the bulk action. Expanding the bracket we get

zzK~(d)μνlogz22K~(d)μνzzK(d)μν\displaystyle-z\partial_{z}\tilde{K}_{(d)\;\mu\nu}\log z^{2}-2\tilde{K}_{(d)\;\mu\nu}-z\partial_{z}K_{(d)\;\mu\nu} δDK~(d)μνlogz2+δDK(d)μν\displaystyle\sim\delta_{D}\tilde{K}_{(d)\;\mu\nu}\log z^{2}+\delta_{D}K_{(d)\;\mu\nu} (308)

and for all nn at leading order of zz we have

zzK(n)μν[z2g(0)](2n)K(n)μν[z2g(0)].\displaystyle-z\partial_{z}K_{(n)\;\mu\nu}[z^{-2}g^{(0)}]\sim(2-n)K_{(n)\;\mu\nu}[z^{-2}g^{(0)}]. (309)

Hence matching the leading order terms in (308)(\ref{eq:dzKddDKd}) we get back the inhomogenous transformation in (300)(\ref{eq:dKinhom}). After all the steps above we arrive at the zz expansion of the dilatation eigenfunctions,

K(0)μν[γ]\displaystyle K_{(0)\;\mu\nu}[\gamma] =z2gμν(0)+gμν(2)+\displaystyle=z^{-2}g^{(0)}_{\mu\nu}+g^{(2)}_{\mu\nu}+\cdots (310)
K(2)μν[γ]\displaystyle K_{(2)\;\mu\nu}[\gamma] =K(2)μν[g(0)]+z2gρσ(2)δK(2)μνδgμν(0)+\displaystyle=K_{(2)\;\mu\nu}[g^{(0)}]+z^{2}\int g^{(2)}_{\rho\sigma}\frac{\delta K_{(2)\;\mu\nu}}{\delta g^{(0)}_{\mu\nu}}+\cdots (311)

and so on. The final steps to relate the Fefferman-Graham coefficients to the dilatation eigenfunctions is to express K(n)μν[g(0)]K_{(n)\;\mu\nu}[g^{(0)}] in terms of gμν(m)g^{(m)}_{\mu\nu}. In general K(n)μν[g(0)]K_{(n)\;\mu\nu}[g^{(0)}] are obtained by comparing with the zn2z^{n-2} in (297)(\ref{eq:KFGE}), i.e. for d>4d>4

z0:Kμν[γ]\displaystyle z^{0}:\quad K_{\mu\nu}[\gamma] =gμν(2)+K(2)μν[g(0)]\displaystyle=g^{(2)}_{\mu\nu}+K_{(2)\;\mu\nu}[g^{(0)}] (312)
=0\displaystyle=0
z2:Kμν[γ]\displaystyle z^{2}:\quad K_{\mu\nu}[\gamma] =z2gμν(4)+gρσ(2)δK(2)μνδgμν(0)+z2K(4)μν[g(0)]\displaystyle=z^{2}g^{(4)}_{\mu\nu}+\int g^{(2)}_{\rho\sigma}\frac{\delta K_{(2)\;\mu\nu}}{\delta g^{(0)}_{\mu\nu}}+z^{2}K_{(4)\;\mu\nu}[g^{(0)}] (313)
=z2gμν(4)\displaystyle=-z^{2}g^{(4)}_{\mu\nu}

so we get

K(2)μν[g(0)]\displaystyle K_{(2)\;\mu\nu}[g^{(0)}] =gμν(2)\displaystyle=-g^{(2)}_{\mu\nu} (314)
K(4)μν[g(0)]\displaystyle K_{(4)\;\mu\nu}[g^{(0)}] =2gμν(4)+gρσ(2)δgμν(2)δgμν(0).\displaystyle=-2g^{(4)}_{\mu\nu}+\int g^{(2)}_{\rho\sigma}\frac{\delta g^{(2)}_{\mu\nu}}{\delta g^{(0)}_{\mu\nu}}. (315)

For larger nn, there will be functional derivative terms coming from the Taylor expansion in (305)(\ref{eq:KnTaylor}) at the znz^{n} order, for example

zn2g(n2)δK(2)δg(0),,zn2()m(g(p1)g(pm))(δδg(0)δδg(0))K(q),\displaystyle z^{n-2}\int g^{(n-2)}\cdot\frac{\delta K_{(2)}}{\delta g^{(0)}},\cdots,z^{n-2}\left(\int\cdots\int\right)^{m}(g^{(p_{1})}\cdots g^{(p_{m})})\cdot(\frac{\delta}{\delta g^{(0)}}\cdots\frac{\delta}{\delta g^{(0)}})K_{(q)},\cdots (316)

where q+p1++pm=nq+p_{1}+\cdots+p_{m}=n. Of course when onshell all gμνn<dg^{n<d}_{\mu\nu} and g~μν(d)\tilde{g}^{(d)}_{\mu\nu} are functions of gμν(0)g^{(0)}_{\mu\nu}. Order by order, we can write all the dilatation eigenfunctions in terms of the terms Fefferman-Graham expansion.

B.2 Volume Integrals of h(n)h_{(n)}

This appendix will address some technical steps omitted in section 4.3. In those examples, the integral term in (176)(\ref{eq:IntDdSct}) is given by a volume integral over BϵB_{\epsilon}. We know the counterterm is given by surface integral over the regulated boundary of the entangling surface B~ϵ\partial\tilde{B}_{\epsilon}. Since B~ϵ=Bϵ\partial\tilde{B}_{\epsilon}=\partial{B}_{\epsilon}, we need to express the integral term as a surface integral over Bϵ\partial{B}_{\epsilon}. In the following we will show the relation between volume and surface integrals of the terms in the Fefferman-Graham expansion.

The leading term in the Fefferman-Graham expansion, h(0)μνh_{(0)\,\mu\nu}, is part of the boundary data hence should be treated as independent variable. Nonetheless, we can express them as combination of total derivatives and moment density of the derivatives of h(0)μνh_{(0)\,\mu\nu}. For the spatial trace h(0)iih_{(0)\,ii} we have

(d2)h(0)ii\displaystyle(d-2)h_{(0)\,ii} =i(xih(0)jjxjh(0)ijx22(ih(0)jjih(0)ij))\displaystyle=\partial_{i}\left({x}^{i}h_{(0)\,jj}-{x}^{j}h_{(0)\,ij}-\frac{\vec{x}^{2}}{2}\left(\partial_{i}h_{(0)\,jj}-\partial_{i}h_{(0)\,ij}\right)\right) (317)
+x22(iih(0)jjijh(0)ij).\displaystyle+\frac{\vec{x}^{2}}{2}\left(\partial_{i}\partial_{i}h_{(0)\,jj}-\partial_{i}\partial_{j}h_{(0)\,ij}\right).

From the Einstein equation the last bracket above is related to h(2)iih_{(2)\,ii} by (93)(\ref{eq:h2toh0}) and we get

h(0)ii\displaystyle h_{(0)\,ii} =1d2i(xih(0)jjxjh(0)ijx22(ih(0)jjih(0)ij))+x2h(2)ii.\displaystyle=\frac{1}{d-2}\partial_{i}\left({x}^{i}h_{(0)\,jj}-{x}^{j}h_{(0)\,ij}-\frac{\vec{x}^{2}}{2}\left(\partial_{i}h_{(0)\,jj}-\partial_{i}h_{(0)\,ij}\right)\right)+\vec{x}^{2}h_{(2)\,ii}. (318)

Integrating over BϵB_{\epsilon}, we obtain a surface integral and a second moment of h(2)iih_{(2)\,ii} over BϵB_{\epsilon},

Bϵdd1xh(0)ii\displaystyle\int_{{B}_{\epsilon}}d^{d-1}x\,h_{(0)\,ii} =1d2Bϵdd2xx^i(xih(0)jjxjh(0)ijx22(ih(0)jjih(0)ij))\displaystyle=\frac{1}{d-2}\int_{\partial{B}_{\epsilon}}d^{d-2}x\hat{{x}}^{i}\left({x}^{i}h_{(0)\,jj}-{x}^{j}h_{(0)\,ij}-\frac{\vec{x}^{2}}{2}\left(\partial_{i}h_{(0)\,jj}-\partial_{i}h_{(0)\,ij}\right)\right) (319)
+Bϵdd1xx2h(2)ii.\displaystyle+\int_{{B}_{\epsilon}}d^{d-1}x\,\vec{x}^{2}h_{(2)\,ii}.

Since Bϵ\partial B_{\epsilon} is a sphere of radius x2=R2ϵ2\vec{x}^{2}=R^{2}-\epsilon^{2}, we can reverse the surface integral for the last terms in the first line to get back a volume integral of h(2)iih_{(2)\,ii} over BϵB_{\epsilon}.

Bϵdd1xh(0)ii\displaystyle\int_{{B}_{\epsilon}}d^{d-1}x\,h_{(0)\,ii} =(R2ϵ2)d12d2Bϵ𝑑Ωd2(h(0)iix^ix^jh(0)ij)\displaystyle=\frac{(R^{2}-\epsilon^{2})^{\frac{d-1}{2}}}{d-2}\int_{\partial{B}_{\epsilon}}d\Omega_{d-2}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right) (320)
Bϵdd1x(R2ϵ2x2)h(2)ii.\displaystyle-\int_{{B}_{\epsilon}}d^{d-1}x\,(R^{2}-\epsilon^{2}-\vec{x}^{2})h_{(2)\,ii}.

Gathering the terms that appear in the integral correction terms we get

Bϵdd1x(h(0)ii+(R2x2)h(2)ii)\displaystyle\int_{{B}_{\epsilon}}d^{d-1}x\,\left(h_{(0)\,ii}+(R^{2}-\vec{x}^{2})h_{(2)\,ii}\right) =(R2ϵ2)d12d2Sd2𝑑Ωd2(h(0)iix^ix^jh(0)ij)\displaystyle=\frac{(R^{2}-\epsilon^{2})^{\frac{d-1}{2}}}{d-2}\int_{S^{d-2}}d\Omega_{d-2}\,\left(h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij}\right) (321)
+Bϵdd1xϵ2h(2)ii.\displaystyle+\int_{{B}_{\epsilon}}d^{d-1}x\,\epsilon^{2}h_{(2)\,ii}.

For d>3d>3, the integral correction term contains higher order terms in the Fefferman-Graham expansion. In general, the nthn^{th} order terms are second derivative of (n2)th(n-2)^{th}. The following expressions for evaluating volume integral of a generic second derivative of a tensor will be useful later on. First the second moment of such a derivative is

Bϵdd1xx2ijAij=Bϵdd1xi(x2jAij2xjAij)+2Aii\displaystyle\int_{{B}_{\epsilon}}d^{d-1}x\,\vec{x}^{2}\partial_{i}\partial_{j}A_{ij}=\int_{{B}_{\epsilon}}d^{d-1}x\,\partial_{i}\left(\vec{x}^{2}\partial_{j}A_{ij}-2\vec{x}^{j}A_{ij}\right)+2A_{ii} (322)

then the shifted second moment is

Bϵdd1x(R2x2)ijAij\displaystyle\int_{{B}_{\epsilon}}d^{d-1}x\,(R^{2}-\vec{x}^{2})\partial_{i}\partial_{j}A_{ij} =Bϵdd1xi(R2jAijx2jAij+2xjAij)2Aii\displaystyle=\int_{{B}_{\epsilon}}d^{d-1}x\,\partial_{i}\left(R^{2}\partial_{j}A_{ij}-\vec{x}^{2}\partial_{j}A_{ij}+2\vec{x}^{j}A_{ij}\right)-2A_{ii}
=ϵ2Bϵdd2xx^ijAij+Bϵdd2x 2x^ixjAijBϵdd1x 2Aii\displaystyle=\epsilon^{2}\int_{\partial{B}_{\epsilon}}d^{d-2}x\,\hat{x}^{i}\partial_{j}A_{ij}+\int_{\partial{B}_{\epsilon}}d^{d-2}x\,2\hat{x}^{i}x^{j}A_{ij}-\int_{{B}_{\epsilon}}d^{d-1}x\,2A_{ii}
=2rd1Sd2𝑑Ωd2x^ix^jAij2Bϵdd1xAii\displaystyle=2r^{d-1}\int_{S^{d-2}}d\Omega_{d-2}\,\hat{x}^{i}\hat{x}^{j}A_{ij}-2\int_{{B}_{\epsilon}}d^{d-1}x\,A_{ii} (323)
+ϵ2rd2Sd2𝑑Ωd2x^ijAij.\displaystyle+\epsilon^{2}r^{d-2}\int_{S^{d-2}}d\Omega_{d-2}\,\hat{x}^{i}\partial_{j}A_{ij}.

The d=4d=4 examples in section 4.3.2, we have integral of the form of (B.2)(\ref{eq:IntddA}) where

h~(4)ii=ijAij\displaystyle\tilde{h}_{(4)\,ii}=\partial_{i}\partial_{j}A_{ij} (324)

and

Aij=18(h(2)ijδijh(2)kk).\displaystyle A_{ij}=\frac{1}{8}\left(h_{(2)\,ij}-\delta_{ij}h_{(2)\,kk}\right). (325)

Neglecting the O(ϵ2)O(\epsilon^{2}) term since they are irrelevant in (4.3.2)(\ref{eq:IntBdDd4}) we get,

Bϵd3x(R2x2)h~(4)ii\displaystyle\int_{{B}_{\epsilon}}d^{3}x\,(R^{2}-\vec{x}^{2})\tilde{h}_{(4)\,ii} =(R2ϵ2)324S2𝑑Ω2(x^ix^jh(2)ijh(2)ii)\displaystyle=\frac{(R^{2}-\epsilon^{2})^{\frac{3}{2}}}{4}\int_{S^{2}}d\Omega_{2}\,(\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}-h_{(2)\,ii}) (326)
+12Bϵd3xh(2)ii.\displaystyle+\frac{1}{2}\int_{{B}_{\epsilon}}d^{3}x\,h_{(2)\,ii}.

As seen in (93)(\ref{eq:h2toh0}), h(2)iih_{(2)\,ii} is the second derivative of h(0)ijh_{(0)\,ij}, the last volume integral can be easily turned into surface integral,

Bϵd3xh(2)ii\displaystyle\int_{{B}_{\epsilon}}d^{3}x\,h_{(2)\,ii} =(R2ϵ2)124S2𝑑Ω2x^i(ih(0)jjjh(0)ij)\displaystyle=\frac{(R^{2}-\epsilon^{2})^{\frac{1}{2}}}{4}\int_{S^{2}}d\Omega_{2}\hat{x}^{i}(\partial_{i}h_{(0)\,jj}-\partial_{j}h_{(0)\,ij})
=r4S2dΩ2x^i(ih(0)jjrh(0)ir\displaystyle=\frac{r}{4}\int_{S^{2}}d\Omega_{2}\hat{x}^{i}\bigg{(}\partial_{i}h_{(0)\,jj}-\partial_{r}h_{(0)\,ir}
1r2θh(0)θi1r2sin2θϕh(0)ϕicosθr2sinθh(0)θi2rh(0)ri)\displaystyle-\frac{1}{r^{2}}\partial_{\theta}h_{(0)\,\theta i}-\frac{1}{r^{2}\sin^{2}\theta}\partial_{\phi}h_{(0)\,\phi i}-\frac{\cos\theta}{r^{2}\sin\theta}h_{(0)\,\theta i}-\frac{2}{r}h_{(0)\,ri}\bigg{)}
=r4S2dϕdθsinθ(x^iih(0)jjx^ix^jx^kkh(0)ij\displaystyle=\frac{r}{4}\int_{S^{2}}d\phi d\theta\sin\theta\bigg{(}\hat{x}^{i}\partial_{i}h_{(0)\,jj}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\partial_{k}h_{(0)\,ij} (327)
+1r3h(0)θθ+1r3sin2θh(0)ϕϕ2x^ix^jrh(0)ij)\displaystyle+\frac{1}{r^{3}}h_{(0)\,\theta\theta}+\frac{1}{r^{3}\sin^{2}\theta}h_{(0)\,\phi\phi}-\frac{2\hat{x}^{i}\hat{x}^{j}}{r}h_{(0)\,ij}\bigg{)}

where we went from the first line to the second line by evaluating jh(0)ij\partial_{j}h_{(0)\,ij} in polar coordinates. From the second line to the third line we integrate by parts and we transform rr coordinate to Cartesian. Finally we can transform the angular coordinate into Cartesian coordinates,

Bϵd3xh(2)ii=r4S2𝑑Ω2\displaystyle\int_{{B}_{\epsilon}}d^{3}x\,h_{(2)\,ii}=\frac{r}{4}\int_{S^{2}}d\Omega_{2} [h(0)ii3x^ix^jh(0)ij\displaystyle\bigg{[}h_{(0)\,ii}-3\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij} (328)
+xjjh(0)iix^ix^jxkkh(0)ij].\displaystyle+x^{j}\partial_{j}h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}x^{k}\partial_{k}h_{(0)\,ij}\bigg{]}.

For the d=4d=4 example in section 4.3.3, we have integrals of the form of (B.2)(\ref{eq:IntddA}) where

h(4)ii=ijBij\displaystyle{h}_{(4)\,ii}=\partial_{i}\partial_{j}B_{ij} (329)

and

Bij=14(δijh(2)kkh(2)ij).\displaystyle B_{ij}=\frac{1}{4}\left(\delta_{ij}h_{(2)\,kk}-h_{(2)\,ij}\right). (330)

Neglecting the O(ϵ2)O(\epsilon^{2}) term since they are irrelevant in (205)(\ref{eq:IntBdDd5}) we get

Bϵd4x(R2x2)h(4)ii\displaystyle\int_{{B}_{\epsilon}}d^{4}x\,(R^{2}-\vec{x}^{2}){h}_{(4)\,ii} =(R2ϵ2)522S3𝑑Ω3(h(2)iix^ix^jh(2)ij)\displaystyle=\frac{(R^{2}-\epsilon^{2})^{\frac{5}{2}}}{2}\int_{S^{3}}d\Omega_{3}\,(h_{(2)\,ii}-\hat{x}^{i}\hat{x}^{j}h_{(2)\,ij}) (331)
32Bϵd4xh(2)ii\displaystyle-\frac{3}{2}\int_{{B}_{\epsilon}}d^{4}x\,h_{(2)\,ii}

Following the steps in (327)(\ref{eq:Inth2d4}) we can evaluate the volume integral of h(2)iih_{(2)\,ii},

Bϵd4xh(2)ii\displaystyle\int_{{B}_{\epsilon}}d^{4}x\,h_{(2)\,ii} =(R2ϵ2)326S3𝑑Ω3x^i(ih(0)jjjh(0)ij)\displaystyle=\frac{(R^{2}-\epsilon^{2})^{\frac{3}{2}}}{6}\int_{S^{3}}d\Omega_{3}\hat{x}^{i}(\partial_{i}h_{(0)\,jj}-\partial_{j}h_{(0)\,ij}) (332)
=r36S3dΩ3x^i(ih(0)jjx^jx^kkh(0)ij\displaystyle=\frac{r^{3}}{6}\int_{S^{3}}d\Omega_{3}\hat{x}^{i}\bigg{(}\partial_{i}h_{(0)\,jj}-\hat{x}^{j}\hat{x}^{k}\partial_{k}h_{(0)\,ij}
1r2θ1h(0)θ1i1r2sin2θ1θ2h(0)θ2i1r2sin2θ1sin2θ2ϕh(0)ϕi\displaystyle-\frac{1}{r^{2}}\partial_{\theta_{1}}h_{(0)\,\theta_{1}i}-\frac{1}{r^{2}\sin^{2}\theta_{1}}\partial_{\theta_{2}}h_{(0)\,\theta_{2}i}-\frac{1}{r^{2}\sin^{2}\theta_{1}\sin^{2}\theta_{2}}\partial_{\phi}h_{(0)\,\phi i}
2cosθ1r2sinθ1h(0)θ1icosθ2r2sin2θ1sinθ2h(0)θ2i3rh(0)ri)\displaystyle-\frac{2\cos\theta_{1}}{r^{2}\sin\theta_{1}}h_{(0)\,\theta_{1}i}-\frac{\cos\theta_{2}}{r^{2}\sin^{2}\theta_{1}\sin\theta_{2}}h_{(0)\,\theta_{2}i}-\frac{3}{r}h_{(0)\,ri}\bigg{)}
=r36S3dϕdθ1dθ2sin2θ1sinθ2(x^iih(0)jjx^ix^jx^kkh(0)ij\displaystyle=\frac{r^{3}}{6}\int_{S^{3}}d\phi d\theta_{1}d\theta_{2}\sin^{2}\theta_{1}\sin\theta_{2}\bigg{(}\hat{x}^{i}\partial_{i}h_{(0)\,jj}-\hat{x}^{i}\hat{x}^{j}\hat{x}^{k}\partial_{k}h_{(0)\,ij}
+1r3h(0)θ1θ1+1r3sin2θ1h(0)θ2θ2+1r3sin2θ1sin2θ2h(0)ϕϕ3x^ix^jrh(0)ij)\displaystyle+\frac{1}{r^{3}}h_{(0)\,\theta_{1}\theta_{1}}+\frac{1}{r^{3}\sin^{2}\theta_{1}}h_{(0)\,\theta_{2}\theta_{2}}+\frac{1}{r^{3}\sin^{2}\theta_{1}\sin^{2}\theta_{2}}h_{(0)\,\phi\phi}-\frac{3\hat{x}^{i}\hat{x}^{j}}{r}h_{(0)\,ij}\bigg{)}

Finally transforming into Cartesian coordinate we get

Bϵd4xh(2)ii=r36S3𝑑Ω3\displaystyle\int_{{B}_{\epsilon}}d^{4}x\,h_{(2)\,ii}=\frac{r^{3}}{6}\int_{S^{3}}d\Omega_{3} [h(0)ii4x^ix^jh(0)ij\displaystyle\bigg{[}h_{(0)\,ii}-4\hat{x}^{i}\hat{x}^{j}h_{(0)\,ij} (333)
+xjjh(0)iix^ix^jxkkh(0)ij].\displaystyle+x^{j}\partial_{j}h_{(0)\,ii}-\hat{x}^{i}\hat{x}^{j}x^{k}\partial_{k}h_{(0)\,ij}\bigg{]}.

Appendix C Covariant Phase Space Hamiltonian

In this section we follow the formalism in 2020HarlowWu but here we consider the renormalized action, as well as different conditions on the vector. The variational problem of a Lagrangian theory with bulk and boundary terms requires the variation of both the bulk and boundary terms to be zero onshell. Therefore the sum of the presymplectic potential and the variation of the boundary terms should be exact on the boundary of the manifold

𝚯[δϕ]δ𝑩=d𝑪[δϕ].\displaystyle\boldsymbol{\Theta}[\delta\phi]-\delta\boldsymbol{B}=d\boldsymbol{C}[\delta\phi]. (334)

The presymplectic current can be expressed as

𝝎[δ1ϕ,δ2ϕ]=δ1(𝚯[δ2ϕ]d𝑪[δ2ϕ])\displaystyle\boldsymbol{\omega}[\delta_{1}\phi,\delta_{2}\phi]=\delta_{1}\left(\boldsymbol{\Theta}[\delta_{2}\phi]-d\boldsymbol{C}[\delta_{2}\phi]\right) (335)

where δ\delta is the exterior derivative on the configuration space. In Einstein gravity with cosmological constant and Gibbons-Hawking boundary term, without imposing any boundary condition, we get

𝚯[δg]δ𝑩GH=d𝑪GH[δg]+𝝅δg.\displaystyle\boldsymbol{\Theta}[\delta g]-\delta\boldsymbol{B}^{GH}=d\boldsymbol{C}^{GH}[\delta g]+\boldsymbol{\pi}\cdot\delta g. (336)

The exact contribution 𝑪GH[δg]\boldsymbol{C}^{GH}[\delta g] captures the variation of the metric in the normal direction and the canonical momentum term captures the usual variation of the induced metric. Hence one can eliminate this term by imposing a radial gauge condition. However, as we will see later the variation of 𝑪GH\boldsymbol{C}^{GH} will have a non zero contribution. On BϵB_{\epsilon} we get

𝑪GH[δg]\displaystyle\boldsymbol{C}^{GH}[\delta g] =𝜺μν16πGNγνσnμnρδgσρ\displaystyle=-\frac{\boldsymbol{\varepsilon}_{\mu\nu}}{16\pi G_{N}}\gamma^{\nu\sigma}n^{\mu}n^{\rho}\delta g_{\sigma\rho} (337)
=𝜺zt16πGNδgtz.\displaystyle=-\frac{\boldsymbol{\varepsilon}_{zt}}{16\pi G_{N}}\delta g^{tz}. (338)

The variation of the Hamiltonian along the vector field ξ\xi can be constructed from the presymplectic form Ω~\tilde{\Omega}

δH[ξ]=ιXξΩ~\displaystyle\delta H[\xi]=-\iota_{X_{\xi}}\tilde{\Omega} (339)

where XξX_{\xi} is the configuration space vector that takes the one form in configuration space to the Lie derivative in configuration space

Xξ(δϕ)=Xξϕ.\displaystyle X_{\xi}(\delta\phi)=\mathcal{L}_{X_{\xi}}\phi. (340)

The Lie derivative in configuration space only varies the dynamical fields along ξ\xi direction and the Lie derivative in spacetime varies both the dynamical fields and background fields along the ξ\xi direction. Any tensor is called covariant under the diffeomorphism induced by ξ\xi if the two Lie derivatives coincide

XξT=ξT.\displaystyle\mathcal{L}_{X_{\xi}}T=\mathcal{L}_{{\xi}}T. (341)

In general, the normal is constructed from a background function,

ndf,\displaystyle n\propto df, (342)

such that the level sets of the function define a foliation. Anything that distinguishes the normal direction from other directions is not covariant unless we impose an extra condition on ξ\xi,

ξf=ξ(f)=0\displaystyle\mathcal{L}_{\xi}f=\xi(f)=0 (343)

which implies the normal direction of ξ\xi vanishes. We label the difference between the two Lie derivatives of 𝑪\boldsymbol{C} along generic ξ\xi by

𝑫[ξ]=ξ𝑪Xξ𝑪.\displaystyle\boldsymbol{D}[\xi]=\mathcal{L}_{{\xi}}\boldsymbol{C}-\mathcal{L}_{X_{\xi}}\boldsymbol{C}. (344)

Since the presymplectic form is given by the integral of the presymplectic form, ω\omega, on the Cauchy surface 𝒞\mathcal{C}, we can express the variation of the Hamiltonian as

δH[ξ]=𝒞ιXξ𝝎.\displaystyle\delta H[\xi]=\int_{\mathcal{C}}-\iota_{X_{\xi}}\boldsymbol{\omega}. (345)

Through some algebra in a generic theory onshell we get

Xξ𝝎=d(δ𝑸[ξ]ιξδ𝑩διXξ𝑪ιξ𝝅δϕ𝑫[ξ]+dιξ𝑪).\displaystyle-X_{\xi}\cdot\boldsymbol{\omega}=d\left(\delta\boldsymbol{Q}[\xi]-\iota_{\xi}\delta\boldsymbol{B}-\delta\iota_{X_{\xi}}\boldsymbol{C}-\iota_{\xi}\boldsymbol{\pi}\cdot\delta\phi-\boldsymbol{D}[\xi]+d\iota_{\xi}\boldsymbol{C}\right). (346)

Let us define the Hamiltonian potential as the density over 𝒞\partial\mathcal{C} so

δ𝑯[ξ]=δ𝑸[ξ]ιξδ𝑩διXξ𝑪ιξ𝝅δϕ𝑫[ξ]+dιξ𝑪.\displaystyle\delta\boldsymbol{H}[\xi]=\delta\boldsymbol{Q}[\xi]-\iota_{\xi}\delta\boldsymbol{B}-\delta\iota_{X_{\xi}}\boldsymbol{C}-\iota_{\xi}\boldsymbol{\pi}\cdot\delta\phi-\boldsymbol{D}[\xi]+d\iota_{\xi}\boldsymbol{C}. (347)

We can see that the Hamiltonian potential has an exact term ambiguity because the Hamiltonian is defined to be the integral of the Hamiltonian form over a manifold with no boundary. We will now show that the full Noether charge form is a well defined Hamiltonian potential of the renormalized action. Since we have found that the holographic charge form is equal to the full Noether charge form up to an exact term, the Hamiltonian defined through holographic charge form is the full Noether charge. In the context of the first law of entanglement entropy, neither the entanglement entropy nor the modular energy is a Hamiltonian or a conserved charge, and hence the exact term difference matters. Here we will derive an expression for δ𝚫[ξB]\delta\boldsymbol{\Delta}[\xi_{B}] in terms of the quantities defined above.

In 2020HarlowWu , the case of Einstein gravity with cosmological constant and Gibbons-Hawking boundary term was considered. The boundary condition imposed was

𝝅δg=0\displaystyle\boldsymbol{\pi}\cdot\delta g=0 (348)

and restricting normal direction of ξ\xi to be identically zero. Under these conditions, the variation of the Hamiltonian potential is

δ𝑯BY[ξ]\displaystyle\delta\boldsymbol{H}^{BY}[\xi] =δ(𝜺μνnμTσνξσ)\displaystyle=\delta\left(\boldsymbol{\varepsilon}_{\mu\nu}n^{\mu}{T}^{\nu}_{\sigma}\xi^{\sigma}\right) (349)
=δ(𝜺μνnμ2πσνξσ)\displaystyle=\delta\left(\boldsymbol{\varepsilon}_{\mu\nu}n^{\mu}2{\pi}^{\nu}_{\sigma}\xi^{\sigma}\right) (350)
=δ𝓠[ξ]\displaystyle=-\delta\boldsymbol{\mathcal{Q}}[\xi] (351)

where TμνT_{\mu\nu} is the Brown York stress tensor given by

Tμν=18πG(KμνγμνK).\displaystyle T^{\mu\nu}=-\frac{1}{8\pi G}\left(K^{\mu\nu}-\gamma^{\mu\nu}K\right). (352)

In our case, not only we do not impose the boundary condition (348)(\ref{eq:Pidg0}), we also need to use the vector field ξB\xi_{B} which will introduce a term relating to the normal component of ξB\xi_{B}.

The Hamiltonian potential from Einstein gravity with Gibbons-Hawking boundary term is

δ𝑯GH[ξB]\displaystyle\delta\boldsymbol{H}^{GH}[\xi_{B}] =δ𝑸[ξB]ιξBδ𝑩GHδιXξB𝑪GHιξB𝝅δg𝑫GH[ξB]+dιξB𝑪GH\displaystyle=\delta\boldsymbol{Q}[\xi_{B}]-\iota_{\xi_{B}}\delta\boldsymbol{B}^{GH}-\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{GH}-\iota_{\xi_{B}}\boldsymbol{\pi}\cdot\delta g-\boldsymbol{D}^{GH}[\xi_{B}]+d\iota_{\xi_{B}}\boldsymbol{C}^{GH}
=δ𝑸[ξB]ιξBδ𝑩GHδιXξB𝑪GHιξB𝝅δg\displaystyle=\delta\boldsymbol{Q}[\xi_{B}]-\iota_{\xi_{B}}\delta\boldsymbol{B}^{GH}-\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{GH}-\iota_{\xi_{B}}\boldsymbol{\pi}\cdot\delta g
=δ𝑯BY[ξB]𝜺μνnμτν16πGNδγταα(ξBβnβ)ιξB𝝅δg\displaystyle=\delta\boldsymbol{H}^{BY}[\xi_{B}]-\frac{\boldsymbol{\varepsilon}_{\mu\nu}n^{\mu}\tau^{\nu}}{16\pi G_{N}}\delta\gamma\tau^{\alpha}\partial_{\alpha}(\xi_{B}^{\beta}n_{\beta})-\iota_{\xi_{B}}\boldsymbol{\pi}\cdot\delta g (353)

where τ\tau is the future pointing timelike normal vector. To get to the last line we also used the following properties for the Killing vector ξB\xi_{B} and in radial gauge,

ιξB𝑪GH=𝑫GH[ξB]=0.\displaystyle\iota_{\xi_{B}}\boldsymbol{C}^{GH}=\boldsymbol{D}^{GH}[\xi_{B}]=0. (354)

When we consider the renormalized action there are additional counterterms in the full Hamiltonian potential

δ𝑯full[ξ]=δ𝑸[ξ]\displaystyle\delta\boldsymbol{H}^{full}[\xi]=\delta\boldsymbol{Q}[\xi] ιξδ𝑩GHδιXξ𝑪GHιξ𝝅δg𝑫GH[ξ]+dιξ𝑪GH\displaystyle-\iota_{\xi}\delta\boldsymbol{B}^{GH}-\delta\iota_{X_{\xi}}\boldsymbol{C}^{GH}-\iota_{\xi}\boldsymbol{\pi}\cdot\delta g-\boldsymbol{D}^{GH}[\xi]+d\iota_{\xi}\boldsymbol{C}^{GH} (355)
+ιξδ𝑩ct+διXξ𝑪ct+ιξ𝝅ctδg+𝑫ct[ξ]dιξ𝑪ct.\displaystyle+\iota_{\xi}\delta\boldsymbol{B}^{ct}+\delta\iota_{X_{\xi}}\boldsymbol{C}^{ct}+\iota_{\xi}\boldsymbol{\pi}^{ct}\cdot\delta g+\boldsymbol{D}^{ct}[\xi]-d\iota_{\xi}\boldsymbol{C}^{ct}.

Simplifying the above equation by gathering the boundary terms we get

δ𝑯full[ξ]=δ𝑸[ξ]+δ𝒃GH[ξ]δ𝒃ct[ξ].\displaystyle\delta\boldsymbol{H}^{full}[\xi]=\delta\boldsymbol{Q}[\xi]+\delta\boldsymbol{b}^{GH}[\xi]-\delta\boldsymbol{b}^{ct}[\xi]. (356)

Hence the Gibbon-Hawking Hamiltonian potential is related to the full Hamiltonian potential by

δ𝑯GH[ξ]=δ𝑯full[ξ]+δ𝒃ct[ξ].\displaystyle\delta\boldsymbol{H}^{GH}[\xi]=\delta\boldsymbol{H}^{full}[\xi]+\delta\boldsymbol{b}^{ct}[\xi]. (357)

From (337)(\ref{eq:CGHdg}) and (344)(\ref{eq:D2C}) we can deduce

𝑪GH=𝑪ct,\displaystyle\boldsymbol{C}^{GH}=\boldsymbol{C}^{ct}, 𝑫GH=𝑫ct\displaystyle\boldsymbol{D}^{GH}=\boldsymbol{D}^{ct} (358)

then we have

δ𝑯full[ξ]=δ𝑸full[ξ]ιξ𝝅(d)δg.\displaystyle\delta\boldsymbol{H}^{full}[\xi]=\delta\boldsymbol{Q}^{full}[\xi]-\iota_{\xi}\boldsymbol{\pi}_{(d)}\cdot\delta g. (359)

The full Hamiltonian potential is equal to the full Noether charge when the last term is zero. For a conformal Killing vector we can apply the tracelessness condition on 𝝅(d)μν\boldsymbol{\pi}^{\mu\nu}_{(d)}. In our case, the unperturbed 𝝅(d)μν\boldsymbol{\pi}^{\mu\nu}_{(d)} is zero by itself, so we can relax all boundary condition on δgμν\delta g_{\mu\nu}.

By inspecting the dilatation eigenvalue expansion of (353)(\ref{eq:dHGH}), the renormalized Brown-York Hamiltonian potential δ𝑯(d)BY[ξB]\delta\boldsymbol{H}^{BY}_{(d)}[\xi_{B}] can be expressed in terms of δ𝑯GH[ξB]\delta\boldsymbol{H}^{GH}[\xi_{B}] and its counterterm,

δ𝑯(d)BY[ξB]ιξB𝝅(d)δg=δ𝑯GH[ξB]δ𝑯ctGH[ξB].\displaystyle\delta\boldsymbol{H}^{BY}_{(d)}[\xi_{B}]-\iota_{\xi_{B}}\boldsymbol{\pi}_{(d)}\cdot\delta g=\delta\boldsymbol{H}^{GH}[\xi_{B}]-\delta\boldsymbol{H}^{GH}_{ct}[\xi_{B}]. (360)

In our setting 𝝅(d)μν=0\boldsymbol{\pi}^{\mu\nu}_{(d)}=0 so the renormalized Brown-York Hamiltonian potential is obtained by subtracting the lower order terms in the dilatation eigenvalue expansion of the Gibbons-Hawking Hamiltonian potential. This should be distinguished from the full Hamiltonian that is constructed form the renormalized Lagrangian or action. These two procedures of obtaining the Hamiltonian are equivalent if the difference between Hamiltonian potentials is exact. We will see in the following how the two renormalization procedures differ in the context of entanglement entropy and modular energy.

First we use (353)(\ref{eq:dHGH}) and (355)(\ref{eq:dHfull}), to relate the two Hamiltonian potentials, δ𝑯BY[ξB]\delta\boldsymbol{H}^{BY}[\xi_{B}] and δ𝑯full[ξB]\delta\boldsymbol{H}^{full}[\xi_{B}], by

δ𝑯BY[ξB]=δ𝑯full[ξB]ιξBδ𝑩ctδιXξB𝑪ct+𝜺μνnμτν16πGNδγταα(ξBβnβ)\displaystyle\delta\boldsymbol{H}^{BY}[\xi_{B}]=\delta\boldsymbol{H}^{full}[\xi_{B}]-\iota_{\xi_{B}}\delta\boldsymbol{B}^{ct}-\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{ct}+\frac{\boldsymbol{\varepsilon}_{\mu\nu}n^{\mu}\tau^{\nu}}{16\pi G_{N}}\delta\gamma\tau^{\alpha}\partial_{\alpha}(\xi_{B}^{\beta}n_{\beta}) (361)

The difference between the two Hamiltonian potentials is not exact, and this implies δ𝑯BY[ξB]\delta\boldsymbol{H}^{BY}[\xi_{B}] is not a proper Hamiltonian potential that integrates to give the Hamiltonian induced by ξB\xi_{B}. However, we shall see that the renormalized Brown-York Hamiltonian potential or the holographic charge form is an appropriate Hamiltonian potential. Let us first express it in terms of the full Hamiltonian potential and all the counterterms,

δ𝑯(d)BY[ξB]\displaystyle\delta\boldsymbol{H}^{BY}_{(d)}[\xi_{B}] =δ𝑯full[ξB]ιξBδ𝑩ctδιXξB𝑪ctιξB𝝅ctδgδ𝑯ctGH[ξB]\displaystyle=\delta\boldsymbol{H}^{full}[\xi_{B}]-\iota_{\xi_{B}}\delta\boldsymbol{B}^{ct}-\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{ct}-\iota_{\xi_{B}}\boldsymbol{\pi}_{ct}\cdot\delta g-\delta\boldsymbol{H}^{GH}_{ct}[\xi_{B}] (362)
δ𝑯(d)BY[ξB]\displaystyle\delta\boldsymbol{H}^{BY}_{(d)}[\xi_{B}] =δ𝑯full[ξB]δ𝚫[ξB].\displaystyle=\delta\boldsymbol{H}^{full}[\xi_{B}]-\delta\boldsymbol{\Delta}[\xi_{B}]. (363)

The difference in the Hamiltonian potentials is non zero in general.

We can express the difference in Hamiltonian potentials as

δ𝚫[ξB]\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}] =δ𝑯ctGH[ξB]+ιξBδ𝑩ct+διXξB𝑪ct+ιξB𝝅ctδg\displaystyle=\delta\boldsymbol{H}^{GH}_{ct}[\xi_{B}]+\iota_{\xi_{B}}\delta\boldsymbol{B}^{ct}+\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{ct}+\iota_{\xi_{B}}\boldsymbol{\pi}_{ct}\cdot\delta g (364)
=δ𝑯ctGH[ξB]δ𝒃ct[ξB].\displaystyle=\delta\boldsymbol{H}^{GH}_{ct}[\xi_{B}]-\delta\boldsymbol{b}^{ct}[\xi_{B}]. (365)

Hence, the physical interpretation of δ𝚫\delta\boldsymbol{\Delta} is the difference of counterterms in the two renormalization procedure where δ𝒃ct\delta\boldsymbol{b}^{ct} is the counterterms contribution of the Hamiltonian potential derived from the renormalized action and δ𝑯ctGH\delta\boldsymbol{H}^{GH}_{ct} is the counterterm of the Hamiltonian potential derived from the bare action. More explicitly the we have the expression that matches with (159)(\ref{eq:dDelta}),

δ𝚫[ξB]=𝜺μνnμτν16πGNδγταα(ξBβnβ)+διXξB𝑪ct+ιξBδ𝑩ct+δ(𝜺μνnμ2πctσνξσ),\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}]=-\frac{\boldsymbol{\varepsilon}_{\mu\nu}n^{\mu}\tau^{\nu}}{16\pi G_{N}}\delta\gamma\tau^{\alpha}\partial_{\alpha}(\xi_{B}^{\beta}n_{\beta})+\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{ct}+\iota_{\xi_{B}}\delta\boldsymbol{B}^{ct}+\delta\left({\boldsymbol{\varepsilon}_{\mu\nu}n^{\mu}}{2\pi}^{\nu}_{ct\,\sigma}\xi^{\sigma}\right), (366)

with

διXξB𝑪ct=z𝜺zt16πGNδgttzξBt.\displaystyle\delta\iota_{X_{\xi_{B}}}\boldsymbol{C}^{ct}=\frac{z\boldsymbol{\varepsilon}_{zt}}{16\pi G_{N}}\delta g^{t}_{t}\partial_{z}\xi_{B}^{t}. (367)

Let us now dissect (366)(\ref{eq:dDeltaexp}) term by term.

The first term captures the non-covariant variation of the normal direction. In 2020HarlowWu this term is absent as they restrict the diffeomorphism generator to preserve covariance of the normal. In Papadimitriou:2005ii , this term is absent as a stronger fall-off condition is imposed.

The second term captures the variation of the diffeomorphism of the metric in the normal direction. This term is non-vanishing because ξB\xi_{B} is no longer Killing in the perturbed metric. Hence we do not see the equivalent of this term in the unperturbed 𝚫[ξB]\boldsymbol{\Delta}[\xi_{B}] from (136)(\ref{eq:Deltaex}). The last two terms are the standard counterterm contributions from the full Hamiltonian potential and Brown-York Hamiltonian potential. The non trivial result we found is that this difference is exact, the exterior derivative of the density of the entanglement entropy counterterms,

δ𝚫[ξB]=dδ𝑺Bct.\displaystyle\delta\boldsymbol{\Delta}[\xi_{B}]=d\delta\boldsymbol{S}^{ct}_{B}. (368)

Then the Hamiltonian defined by the renormalized Brown-York Hamiltonian potential is the same as the full Hamiltonian potential,

δ[ξB]\displaystyle\delta\mathcal{H}[\xi_{B}] =𝒞δ𝑯(d)BY[ξB]\displaystyle=\int_{\partial\mathcal{C}}\delta\boldsymbol{H}^{BY}_{(d)}[\xi_{B}] (369)
=𝒞δ𝑯full[ξB]dδ𝑺Bct\displaystyle=\int_{\partial\mathcal{C}}\delta\boldsymbol{H}^{full}[\xi_{B}]-d\delta\boldsymbol{S}^{ct}_{B}
=𝒞δ𝑯full[ξB]\displaystyle=\int_{\partial\mathcal{C}}\delta\boldsymbol{H}^{full}[\xi_{B}]
δ[ξB]\displaystyle\delta\mathcal{H}[\xi_{B}] =δHfull[ξB].\displaystyle=\delta H^{full}[\xi_{B}]. (370)

For entanglement entropy and modular energy this difference matters because the integral is over a manifold with boundary that turns the exact term into the appropriate counterterm for the entanglement entropy. This analysis establishes the first law of renormalized entanglement entropy.

References