]>NUPHB14565S0550-3213(19)30033-110.1016/j.nuclphysb.2019.02.005The Author(s)High Energy Physics – TheoryFig. 1Flow in a z-β plane along a curve 4βz = y2 with fixed y.Fig. 1Holographic duality for 3d spin-3 gravity coupled to scalar fieldRyuichiNakayamanakayama@particle.sci.hokudai.ac.jpKenjiShioharak-shiohara@particle.sci.hokudai.ac.jpTomotakaSuzukit-suzuki@particle.sci.hokudai.ac.jpDivision of Physics, Graduate School of Science, Hokkaido University, Sapporo 060-0810, JapanDivision of PhysicsGraduate School of ScienceHokkaido UniversitySapporo060-0810JapanEditor: Clay CórdovaAbstractThe 3d spin-3 gravity theory is holographically dual to a 2d W3-extended CFT. In a large-c limit the symmetry algebra of the CFT reduces to SU(1,2)×SU(1,2). On ground of symmetry the dual bulk space–time will be given by an 8d group manifold SU(1,2). Hence we need to introduce five extra coordinates in addition to three ordinary ones. The 3d space–time is a 3d hyper-surface Σ embedded at constant values of the extra variables. Operators in the CFT at the boundary of Σ are expressed in terms of W descendants of the operators at the boundary of Σ0, where the extra variables vanish. In this paper it is shown that AdS/CFT correspondence for a scalar field coupled to 3d spin-3 gravity is realized in this auxiliary 8d space. A bulk-to-boundary propagator of a scalar field is found and a generating functional of boundary two-point functions of scalar W-descendant operators is obtained by using the classical action for the scalar field. Classically, the scalar field must satisfy both Klein–Gordon equation and a third-order differential equation, which are related to the quadratic and cubic Casimir operators of su(1,2). It is found that the coefficient function of the derivatives of the scalar field in the latter equation is the spin-3 gauge field, when restricted to the hypersurface. An action integral in the 8d auxiliary space for the 3d spin-3 gravity coupled to a scalar field is presented. In general, this 8d auxiliary space is a deformation of the manifold SU(1,2). An 8d local frame is introduced and the equations of motion for the 8d connections Aμ, A‾μ are solved. By restricting those solutions onto Σ, flat connections in 3d SL(3,R)×SL(3,R) Chern–Simons theory are obtained and new 3d black hole solutions with and without spin-3 charge are found by this method.1IntroductionAfter the discovery of AdS/CFT correspondence [1] this subject has been studied extensively and a lot of evidences have been accumulated until now. It is, however, still necessary to extend the range of applicability of this correspondence. One of those possible directions will be the 3d spin-3 gravity [2][3] coupled to a matter (scalar) field. This is a 3d version of Vasiliev's theory [4] with higher spins consistently truncated up to 3. In this theory the gravity and spin-3 gauge field are described by SL(3,R)×SL(3,R) Chern–Simons gauge theory. However, the action integral for a scalar field which has spin-3 charge has not been found. Hence a check of duality for the correlation functions of conformal field theory (CFT) primary operators in the boundary W3 conformal field theory (CFT) has not been carried out except for those of the currents. Therefore it is necessary to have a Lagrangian formulation of this coupled system. Study of the formulation of such a theory will also elucidate the nature of the 3d spin-3 gravity.There is also another attempt to construct 2d CFTs which are dual to 3d higher-spin gravity theories [5]. These 2d theories are based on 2d WN-minimal models which are obtained in terms of cosets of the form:(1.1)SU(N)k⊗SU(N)1SU(N)k+1. This CFT has a central charge(1.2)cN,k=(N−1)[1−N(N+1)(N+k)(N+k+1)]≤(N−1). This model has WN algebra as a symmetry algebra. This is a special case of an extended symmetry algebra W∞[μ], which has all integer spins s≥2 and which can be truncated to WN for μ=N. In the case of W3 algebra N must be set to 3. The corresponding bulk theory is based on so-called hs[μ] algebra [6] and a scalar field with a suitable mass can be consistently coupled to the higher-spin gravity at the level of equations of motion.In this paper the 3d spin-3 gravity theory [2][3] which has a spin-3 gauge field in addition to a gravity field as well as a scalar field will be considered. On the boundary of AdS3 space in the spin-3 gravity the CFT has an additional symmetry, W3 symmetry. This is a non-linear algebra [7][8].(1.3)[Lm,Ln]=(m−n)Lm+n+c12m(m2−1)δm+n,0,[Lm,Wn]=(2m−n)Wm+n,[Wm,Wn]=c36m(m2−1)(m2−4)δm+n,0+13(m−n)(2m2+2n2−mn−8)Lm+n+10β(m−n)Λm+n(m,n∈Z) Here Λm is a normal-ordered operator(1.4)Λm=∑n≤−2LnLm−n+∑n≥−1Lm−nLn−310(m+3)(m+2)Lm, and β in the last line of (1.3) is a constant related to the central charge c as β=16/(22+5c). In the semi-classical limit c→∞, β can be dropped and the sub-algebra of the wedge modes, Ln(n=0,±1) and Wm(m=0,±1,±2), survives. It is a linear su(1,2) algebra and given by(1.5)[Lm,Ln]=(m−n)Lm+n,[Lm,Wn]=(2m−n)Wm+n,[Wm,Wn]=13(m−n){2m2+2n2−mn−8}Lm+n. Actually, su(1,2) is one of the two real forms of sl(3,C), distinct from sl(3,R) [3].11Conventions for sl(3,R) and su(1,2) used in this paper are given in appendices A and B. The difference of the two is the sign of the right hand side of the commutators [Wm,Wn]. By combining the left and right sectors, the boundary field theory has global su(1,2)×su(1,2) symmetry. Then by the principle of holography the bulk spin-3 gravity theory is also expected to have su(1,2)×su(1,2) asymptotic symmetry.22In [15] it was assumed that the symmetry of the spin-3 gravity is sl(3,R)×sl(3,R)). The symmetry of the bulk space–time is, however, su(1,2)×su(1,2), while sl(3,R)×sl(3,R) is a symmetry of the local frame. Translation from sl(3,R) to su(1,2) is simply carried out by analytic continuation of some variables. See footnote 7.It was shown in [3] that the 3d higher-spin gravity theory can be formulated as a Chern–Simons gauge theory.(1.6)S=SCS[A]−SCS[A‾],(1.7)SCS[A]=k4π∫tr(A∧dA+23A∧A∧A), where k=ℓAdS/4G and ℓAdS is the AdS length.33In this paper ℓAdS will be sometimes set to 1. The gauge group acts on the local frame fields, A=ω+e and A‾=ω−e, and in the case of the spin-N gravity the gauge group is SL(N,R)×SL(N,R). The gauge connections must satisfy suitable conditions on the boundary in order for the boundary value problem to be well-posed. The usually adopted boundary condition is A−=A‾+=0. In [9][10] an asymptotically AdS3 solution which shows UV/IR interpolating behavior were found and a black hole solution with spin-3 charge was also obtained. In these solutions all components A± and A‾± do not vanish and the boundary conditions are imposed on the components of connections A±=∑a=18A±ata separately, where ta is a generator of sl(3,R), and the W3 algebra of the higher-spin currents in the CFT is realized as Ward identities of the currents in the presence of a perturbation term ∫d2zμW in the action integral. See [27].When matter fields such as scalar fields are coupled to higher-spin gravity, it is still possible to describe matter degrees of freedom by means of Wilson lines [11][12]. However, a natural description of matter degrees of freedom and their coupling to spin-3 gravity in terms of Lagrangian formalism are still missing. For matter fields which have spin-3 charges it is not possible to write down an action which is invariant under spin-3 gauge transformations as well as diffeomorphisms. It is not possible to derive boundary conformal field theory (CFT) correlation functions from on-shell action by using the standard differentiating dictionary of holography, either. The purpose of this paper is to improve this situation. The action integral for scalar fields and spin-3 gravity fields are presented in a higher-dimensional setting, which will be explained below. It is shown that when a solution to an equation of motion for a scalar field is substituted into the scalar action, a generating functional for a two-point function of scalar operators is obtained semiclassically. The source functions work as boundary conditions of the scalar fields.There is another motivation for the present work. When the symmetry of the 2d CFT is W-extended, all states of a scalar field must be reconstructed [13][14] in the bulk from W-descendants of a scalar primary state in the boundary CFT. In this paper we will introduce an 8d auxiliary space dual to W3 CFT by following our previous paper [15]. Usually in 2d (Euclidean) CFT without W3 symmetry, a representation of global Virasoro algebra is constructed as follows. In a highest-weight representation a highest-weight state |h〉 which satisfies L0h|h〉=h|h〉 and L1h|h〉=0 is introduced. Here the generators with a superscript h represent those in the hyperbolic representation [17], which is appropriate for Lorentzian Poincaré coordinates. These are defined in (2.13) and (2.18) of [15]. L±1h and L0h are global Virasoro generators. Then any states in this representation are given by linear combinations of descendants (L−1h)n|h〉 (n=0,1,2,…). These states are combined into a single state |ϕ(x)〉=exp{ixL−1h}|h〉 by introducing a coordinate x. A shift in x corresponds to a translation. In the case of a global large-c W3 algebra highest-weight representation is defined by a highest-weight state|h,q〉, which satisfies L0h|h,q〉=h|h,q〉, W0h|h,q〉=q|h,q〉, and L1h|h,q〉=W1h|h,q〉=W2h|h,q〉=0.44In [15] W3-charge was denoted as μ. In this paper it will be denoted as q instead. Later μ is used for a chemical potential. Any descendant states in this representation are linear combinations of states of a form (L−1h)n(W−1h)m(W−2h)ℓ|h,q〉. By introducing variables x, α and β, these states are combined into a single state |ϕ(x,α,β)〉=exp{ixL−1h}exp{−αW−2h}exp{iβW−1h}|h,q〉. By using the coefficients of the Taylor expansion of this state in the variables, x, α, β, any states can be obtained. Due to the left and right movers it turns out it is necessary to introduce six variables to describe states in W3 extended CFT.55Wakimoto representation of large-c W3 algebra is expressed in terms of similar variables, whose relation to x, α, β is not known, and this representation was used for calculating correlation functions in W3 extended CFT in [22]. As for the variables in the bulk it is necessary to additionally introduce a radial coordinate y (y=0 is the boundary), which corresponds to L0h+L‾0h, and another coordinate γ, corresponding to W0h+W‾0h. Hence the ‘bulk space–time’ holographically dual to the boundary CFT with W3 symmetry is 8 dimensional.In the remainder of this section we will give a review of our paper [15]. At the end of this section the content of this paper will be presented. In [15] we constructed a state in the boundary W3 CFT which represents the one of a scalar field put at one point inside the bulk. In the case of ordinary AdS3/CFT2 without spin-3 gauge field, such a state |Φ(y=1,x+=0,x−=0)〉 at the center of AdS3 (in the Poincaré coordinates) must satisfy sl(2,R) conditions [16][17][18][19]. By the action of exp{ix+L−1h}, exp{ix−L‾−1h} and yL0h+L‾0h a state at any point in the bulk is obtained.66Light-cone coordinates are defined by x±=t±x. In the case of black holes in sec. 4 x±=t±ϕ. In the case of spin-3 gravity, a state of a scalar field |Φ(x+,α+,β+,x−,α−,β−,y,γ)〉 must satisfy su(1,2) conditions.77In [15] we solved an sl(3,R) conditions for the state in the bulk. To convert the results of [15] to those appropriate for su(1,2), it is necessary to make substitutions, α±→iα±, β±→iβ±, γ→iγ, μ→−iμ. At the center of the bulk these conditions are written as(1.8)(Lnh−(−1)nL‾−nh)|Φ(y=1,other coord's=0)〉=0(n=−1,0,1),(1.9)(Wnh−(−1)nW‾−nh)|Φ(y=1,other coord's=0)〉=0(n=−2,…,2). In [15] the state of a scalar field in the boundary W3 CFT was explicitly constructed, and although this state is a formal integral expression, the existence of such a state was established. Then by using exponentials of W3 generators the state for a scalar field at any point in the bulk was also obtained. During this work an infinite-dimensional representation of W3 generators in the bulk in terms of differential operators were also obtained. This is presented in Appendix B, because some variables are redefined compared to those in [15] by analytic continuation.From the structure of the local state in the bulk it was also found that the scalar local state in the bulk satisfies a partial differential equation which is associated with the quadratic Casimir operator of su(1,2).88See Appendix A and B for conventions.(1.10)C˜2(Lh,Wh)=(L0h)2−12(Lh1L−1h+L−1hL1h)−18(W2hWh−2+W−2hW2h)+12(W1hW−1h+W−1hW1h)−34(W0h)2 By using99In replacing sl(3,R) with su(1,2) the eigenvalue μ of W0 must also be replaced by −iq. the representation (B.1)–(B.2) and(1.11)(C˜2(Lh,Wh)+C˜2(L‾h,W‾h))|OΔ,q〉=12{Δ2−8Δ−3q2}|OΔ,q〉, where |OΔ,q〉 is a CFT primary state on the boundary with eigenvalues L0h=L¯0h=Δ/2 and W0h=W‾0h=q, a differential equation for a scalar field in 8d space was derived. It takes a form(1.12)(∇2−m2)|Φ〉=0, where ∇2 is a Laplacian in a 8d space, which has a metric (1.15) defined below. m is a mass of the scalar field related to Δ and q by(1.13)m2=Δ2−8Δ−3q2. The explicit form of the equation is presented in Appendix C. Conversely, Δ is given in terms of the mass and charge as(1.14)Δ=4+m2+16+3q2. In [15] this equation was interpreted as a Klein–Gordon equation for a scalar field in the 8d space. Because the scalar field transforms non-trivially under SU(1,2)×SU(1,2), its equation of motion must be formulated in 8d. The equation (C.1) coincides with Klein–Gordon equation for a scalar field in a space–time with a metric:(1.15)ds02=gμνdxμdxν=y−2dy2−y−4(y2cosh2γ−4β+β−)dx+dx−+4y−4dα+dα−−y−2cosh2γdβ+dβ−−4y−4(β+dx+dα−+β−dx−dα+)+y−2sinh2γ(dx+dβ−+dx−dβ+)+13dγ2, where xμ=(x+,x−,y,α+,α−,β+,β−,γ). The determinant of this metric is g=detgμν=−112y−18. Metric (1.15) is that of the group manifold SU(1,2). This is invariant under SU(1,2)×SU(1,2) generated by (B.1), (B.2). The coordinates (y,x±) are those of SU(1,1)=AdS3 which is principally embedded in SU(1,2). The additional coordinates (α±,β±,γ) parametrize a coset SU(1,2)/SU(1,1). This metric plays the role of the vacuum of gravity theory.1010The metric is not sufficient to specify the vacuum. It is also necessary to identify the spin-3 field. This will be studied in sec. 2.The ordinary pure AdS3 space is the hypersurface Σ0 embedded at α±=β±=γ=0 in this 8d space. Then, it will be natural to consider other hypersurfaces Σαβγ with constant non-vanishing α±, β± and γ in the 8d space–time. All operators on the boundary of Σαβγ are transformed by exp(iα+W−2h), exp(β+W−1h) and exp(i2γW0h) compared to those on Σ0. An induced metric on this hypersurface Σαβγ is given by(1.16)ds02|Σαβγ=y−2dy2−y−4(y2cosh2γ−4β+β−)dx+dx−. Property of the hypersurface Σαβγ depends on the values of β±.•For β+=β−=0 the hypersurface is AdS3 with AdS length ℓAdS=1 (in our units).•For β+β−<0 the space–time on the hypersurface is not AdS, but asymptotically AdS. This is a solution interpolating two vacua: one corresponding to UV CFT at y=0 with a AdS length ℓAdS′=12ℓAdS=12, and the other to IR CFT at y=∞ with AdS length ℓAdS=1. Hence conformal symmetry is broken in the boundary field theory for non-zero β+β−.•When we set β+=−β−≡λ and take the limit λ→∞, the term proportional to y−4dx+dx− dominates over y−2dx+dx− and the hypersurface Σαβγ asymptotes to a new AdS3 vacuum with AdS length ℓAdS′=12ℓAdS.It was observed in [15] that the parameters β± play the role of flow parameters of renormalization group. To identify this flow in the bulk let us set β+=−β−≡β in (1.16) for simplicity.(1.17)ds02|Σαβγ=y−2dy2−y−4(y2cosh2γ+4β2)dx+dx− In general, in the bulk of an asymptotically AdS space–time AdS symmetry is broken at y≠0 and conformal symmetry is also broken on the holographic screen located at this value of y. As y gets closer to 0, then AdS symmetry will be recovered and the field theory on the holographic screen will flow in the UV to a fixed point, if it exists. Now, to describe a new type of flow on the holographic screen, we should change the radial variable y to a new one z as y=2βz. Then the metric (1.17) is transformed to(1.18)ds02|Σαβγ=14z2dz2−(cosh2γy2+14z2)dx+dx−. If y and γ are fixed, this is an asymptotically AdS metric, where z is a new radial coordinate. In this case we can consider a flow on a constant-y holographic screen by fixing the value of y and sending z→0. Then the field theory flows to a UV fixed point and the above metric flows to that of AdS3 with AdS length ℓAdS′. Along the way β goes to ∞. This flow is depicted in Fig. 1.On the CFT side this flow is associated with a change of the translation operator L−1h→−(1/4)W−2h as follows [15]. On the boundary of a chosen hypersurface Σαβγ (holomorphic) primary operators in general have a form,(1.19)eix+L−1heiα+W−2heβ+Wh−1O(0)e−β+W−1he−iα+W−2he−ix+L−1h. Let us concentrate on (global) Virasoro symmetry, because α, β and γ are fixed. For simplicity we set γ=0 here. When the correlation functions of scalar operators on the common boundary of Σαβγ are computed, they depend on β+ in addition to x, because the exponentials exp{ixL−1h} and exp{βW−1h} do not commute. They do not depend on α+, because [L−1h,W−2h]=0. For simplicity α+ will be omitted in the following discussion. Then an operator O(x+,β+) on the boundary are rewritten as(1.20)O(x+,β+)≡eix+L−1heβ+W−1hO(0)e−β+W−1he−ix+L−1h=eβ+W−1hO(x)e−β+Wh−1, where(1.21)O(x)≡eix+L−1hO(0)e−ix+L−1h,(1.22)L−1h≡e−β+W−1hLh−1eβ+W−1h=L−1h−β+W−2h. In the limit β+→∞ (1.22) asymptotes to −β+W−2h, which is proportional to the Virasoro generator Lˆ−1=−(1/4)W−2h in the W3(2) algebra [9][20]. And the central charge of W3(2) CFT is c/4. Hence up to a congruence transformation, all operators of form O(x+,β+) are mapped onto operators of form (1.21), where the translation operator is modified. In [15] by using an explicit form of the two-point function it was also shown that as β+→∞ two-point functions of scalar operators in the boundary CFT flow from that in W3 vacuum to that in W3(2) vacuum and the conformal weight changes from h to h/2. This flow is generated by a perturbation ΔS=β+W−1h to the action. This flow does not contradict the c-theorem [21], because [W1,W−1]=−2L0 and the ‘metric’ Gij(g) is not positive-definite in the Euclidean CFT.In this paper we will show that holography of 3d spin-3 gravity and the boundary W3 CFT is realized in the auxiliary 8d space. This will be done in the following steps. In the above discussion Klein–Gordon equation is associated with the quadratic Casimir operator. In sec. 2 of this paper the other equation for the scalar field, which stems from the cubic Casimir operator of su(1,2), is studied. It is shown that the coefficient functions of the third-order derivatives in this third-order differential equation coincide with the spin-3 gauge field φμνλ, when restricted to 3d hypersurface Σαβγ. In sec. 3 a bulk to boundary propagator is obtained by solving the Klein–Gordon equation and it is shown that semi-classically evaluated path integral for a free scalar field in the background (1.15) yields a generating functional for a two-point function of scalar operators on the boundary. In sec. 4 an 8d action integral for the spin-3 gravity coupled to a scalar field is proposed. In sec. 5 black hole solutions, both with and without spin-3 charge, are obtained. These steps are carried out by introducing 8d local frame and sl(3,R)×sl(3,R) flat connections, A=ω+e and A‾=ω−e. These are 8 × 8 matrices. By construction, on a 3d hypersurface Σαβγ, these flat connections reduce to 3×8 matrices and they yield new black hole solutions to the 3d SL(3,R)×SL(3,R) Chern–Simons theory. The flat connections on Σαβγ have properties Ax−=0 and A‾x+=0. The integrability condition for the partition function of the charged black hole is checked. The partition function coincides with that of the solution obtained in [9], although the flat connections of the two solutions satisfy distinct boundary conditions. This paper is summarized in Sec. 6. In appendix A conventions for sl(3,R) and su(1,2) algebras in this paper are presented. In appendix B a representation of W3 generators in terms of differential operators is presented. In appendices C and D the explicit forms of Klein–Gordon equation for a scalar field Φ and a spin-3 field ϕμνλ in the auxiliary space is presented. Black hole solutions are obtained by adding extra terms ψ and ψ‾ to the flat connections for su(1,2)×su(1,2) symmetric space–time. In appendix E equations for their first-order perturbations ψ(1) and ψ‾(1) are presented and the solutions to them are shown. In appendices F, G and H results for black hole solutions with spin-3 charge are presented.2Equation related to the cubic Casimir operatorAs explained in sec. 1 the local state for a scalar field in the bulk |Φ(x+,α+,β+,x−,α−,β−,y,γ)〉 satisfies an eigenstate equation for the quadratic Casimir operator. This state also satisfies an equation corresponding to the cubic Casimir:(2.1)(C˜3(Lh,Wh)+C˜3(L‾h,W‾h))|Φ〉=−3i2q((Δ−4)2+q2)|Φ〉, where h=Δ/2 and q are the conformal weight and the spin-3 charge of the boundary primary state |OΔ,q〉. C˜3 is defined in (A.14). By substituting (B.1), (B.2) into (2.1) a differential equation for a scalar field Φ in the bulk is obtained. After some calculation, this can be succinctly written in the following simple form.(2.2)ϕμνλ∇μ∇ν∇λΦ=−3i2q((Δ−4)2+q2)Φ, where ∇μ is a covariant derivative for the metric (1.15) with the Christoffel symbol, and ϕμνλ is a completely symmetric tensor. The explicit form of this field is given in Appendix D. The imaginary unit i on the right hand side of (2.2) implies that Φ is a complex function. (See the solution (3.7).)On a 3d hypersurface Σαβγ with constant α±, β± and γ, which was introduced in sec. 1, (D.1) simplifies to(2.3)ϕ|Σαβγ=34y4(β−dx−(dx+)2+β+dx+(dx−)2−ysinh2γdydx+dx−). This field breaks Lorentz symmetry on the boundary of Σαβγ, on which the values of β± are fixed. It is also found that 8d covariant derivative of ϕμνλ vanishes.(2.4)∇ρϕμνλ=0 This fact gives a geometrical meaning to the ϕμνλ field.It will soon be shown that ϕμνλ actually coincides with a spin-3 gauge field on a hypersurface Σαβγ. This suggests that both Casimir equations have geometrical meaning via the metric and spin-3 gauge field. For this purpose we introduce a vielbein field eμa. Here a=1,2,…,8 and μ=x±,α±,β±,y,γ. This is a local-frame field in 8 dimensions and is an 8 × 8 matrix. Notice that this is different from the 3 × 8 rectangular vielbein introduced in [3]. It is required that1111Throughout this paper the vielbein eμa is assumed to be invertible.(2.5)gμν=12tr(e)2=habeμaeνb.To impose more restrictions we consider sl(3,R) connections1212See appendix A for our conventions for sl(3,R) algebra. The reason for using sl(3,R) generators ta, not su(1,2) ones t˜a is that if su(1,2) generators are used, then it turns out that eμa (a=4∼8) becomes pure imaginary. If sl(3,R) generators are used instead, eμa is real. Hence the symmetry algebra of the local frame is sl(3,R)×sl(3,R), while that of the space–time is su(1,2)×su(1,2). The vielbein connects two analytically continued spaces.(2.6)A=ω+e=ta(ωμa+eμa)dxμ,(2.7)A‾=ω−e=ta(ωμa−eaμ)dxμ, and require flatness conditions on them.(2.8)F=dA+A∧A=0,(2.9)F‾=dA‾+A‾∧A‾=0. Note that these are equations in 8d, and there are no Chern–Simons actions which classically lead to (2.8) and (2.9).1313However, as will be discussed later, there exists an 8d Einstein-like action such that its equations of motion coincide with (2.8) and (2.9), provided the metricity condition and invertibility of the vielbein are assumed for the local frame fields. On a hypersurface Σαβγ with constant α±, β± and γ, however, these frame fields reduce to 3 × 8 rectangular matrices, and these equations are nothing but the equations of motion for connections in SL(3,R)×SL(3,R) Chern–Simons gauge theory.It turns out there are two distinct solutions to (2.8) and (2.9), which produce (1.15).•Solution (I)(2.10)A=1ydyt2+1y(coshγdx+−sinhγdβ+)t3+12dγt6+1y(sinhγdx+−coshγdβ+)t7−1y2(β−dx−−dα−)t8,(2.11)A‾=−1ydyt2−1y(coshγdx−−sinhγdβ−)t1−12dγt6+1y2(β+dx+−dα+)t4−1y(sinhγdx−−coshγdβ−)t5 We note that both Ax+ and Ax− are non-vanishing. Similarly A‾x±≠0. As β± increase from 0 to ∞ in this solution, the leading terms interchange between Ax+ and Ax−. When this solution is restricted to a hypersurface Σαβγ (especially for γ=0), it coincides with the interpolating solution eq. (2.27) of [9].⋄Solution (II)(2.12)A=1ydyt2+1y(coshγdx+−sinhγdβ+)t3−12dγt6+1y(sinhγdx+−coshγdβ+)t7+1y2(β+dx+−dα+)t8,(2.13)A‾=−1ydyt2+1y(−coshγdx−+sinhγdβ−)t1+12dγt6−1y2(β−dx−−dα−)t4+1y(−sinhγdx−+coshγdβ−)t5 In this solution A has no ‘– components’, Ax−=Aα−=Aβ−=0, while A‾ has no ‘+ components’. Then on a hypersurface Σαβγ, we have Ax−=0. This is a flat connection in SL(3,R)×SL(3,R) Chern–Simons gauge theory with a boundary condition Ax−=0.For solution (II), it can be checked that (2.5) with e=(1/2)(A−A‾) coincides with the metric (1.15). Furthermore spin-3 gauge field(2.14)φ=13!tr(e)3 exactly coincides with ϕ in (D.1) up to a multiplicative constant. On Σαβγ this coincides with (2.3).On the other hand in the case of solution (I), although the metric agrees with (1.15), the spin-3 gauge field (2.14) does not coincide with (D.1). For simplicity, only the result for φ on a hypersurface Σαβγ is presented here.(2.15)φsolution I|Σαβγ=−32y4{β−(dx−)3+β+(dx+)3}+32y3sinhγdydx+dx−≠ϕ|Σαβγ On Σ0 two solutions (I) and (II) coincide. Moving away from Σ0, they do not agree.To conclude, out of the two sets of flat connections (I) and (II), both of which are invariant under SU(1,2)×SU(1,2) and produce the same metric (1.15), only for flat connections (II) the coefficient function ϕμνλ (D.1) in the cubic-order differential equation (2.2) for a scalar field coincides with the spin-3 gauge field (2.14). This is in accord with the fact that the coefficient function gμν of derivatives in the Klein–Gordon equation is the metric field. The two equations for the scalar field are written in terms of the geometrical quantities. On the boundary of Σαβγ this solution satisfies boundary conditions Ax−=A‾x+=0. The connection (II) on Σαβγ is the interpolating solution between IR and UV. In sec. 5 we will construct black hole solutions by extending this solution.3Bulk-to-boundary propagatorIn this section a bulk-to-boundary propagator for a scalar field Φ propagating in the space–time (1.15) is derived and then a generating functional of a two-point function of scalar operators on the boundary is obtained semi-classically by substituting the solution to the equation of motion into the action of a scalar field.Construction of the bulk-to-boundary propagator is carried out by solving Klein–Gordon equation (C.1). First of all, in the boundary limit it should asymptote to the two-point function of scalar operators on the boundary, which was derived in [15][22].(3.1)〈OΔ,q(x1+,α1+,β1+,x1−,α1−,β1−)OΔ,q†(x2+,α2+,β2+,x2−,α2−,β2−)〉=(D12D‾12)−(Δ+3iq)/4(D12⁎D‾12⁎)−(Δ−3iq)/4, where D12, etc are defined by(3.2)D12=(x12+)2−(β12+)2−2x12+(β1++β2+)+4α12+,D‾12=(x12−)2−(β12−)2−2x12−(β1−+β2−)+4α12−,D12⁎=(x12+)2−(β12+)2+2x12+(β1++β2+)−4α12+,D‾12⁎=(x12−)2−(β12−)2+2x12−(β1−+β2−)−4α12−. Here x12+≡x1+−x2+ etc.Solution Φ=KΔ,q is obtained by power series expansion in y near the boundary y∼01414This expansion is carried out in a region, where D12,⋯≠0.:(3.3)KΔ,q(y,x1+,α1+,β1+,x1−,α1−,β1−,γ1;x2+,α2+,β2+,x2−,α2−,β2−)=yΔf0(x1+,α1+,β1+,x1−,α1−,β1−)e−iqγ1+yΔ+2f1(x1+,α1+,β1+,x1−,α1−,β1−,γ1)+yΔ+4f2(x1+,α1+,β1+,x1−,α1−,β1−,γ1)+O(yΔ+6) Here (y,x1+,α1+,β1+,x1−,α1−,β1−,γ1) is a bulk point and (x2+,α2+,β2+,x2−,α2−,β2−) a boundary point. The first term in (3.3) should solve (C.1) to the leading order and the function f0 must be chosen to be(3.4)f0=(D12D‾12)−(Δ+3iq)/4(D12⁎D‾12⁎)−(Δ−3iq)/4, due to the boundary condition. The exponential factor for the bulk point is introduced to the first term in (3.3). By substituting this solution into (C.1) an equation for f1 is obtained and it is readily solved. By repeating this procedure we get the following solutions(3.5)f1=12{(Δ+3iq)(x12++β12+)(x12−+β12−)D12D‾12e(−iq−2)γ1+(Δ−3iq)(x12+−β12+)(x12−−β12−)D12⁎D‾12⁎e(−iq+2)γ1}f0,(3.6)f2={(Δ+3iq)(Δ+3iq+4)8(x12++β12+)2(x12−+β12−)2(D12D‾12)2e−4γ1+(Δ−3iq)(Δ−3iq+4)8(x12+−β12+)2(x12−−β12−)2(D12⁎D‾12⁎)2e4γ1+Δ2−4Δ+12iq+9q2161D⁎12D‾12⁎+Δ2−4Δ−12iq+9q2161D12D‾12+Δ2+9q216(1D‾12D12⁎+1D12D‾12⁎)}f0e−iqγ1. Now up to order yΔ+4 the series (3.3) can be summed up with the following result.(3.7)KΔ,q=yΔe−iqγ1{D12D‾12−2y2(x12++β12+)(x12−+β12−)e−2γ1+y4}−(Δ+3iq)/4{D12⁎D‾12⁎−2y2(x12+−β12+)(x12−−β12−)e2γ1+y4}−(Δ−3iq)/4 It is directly checked that (3.7) solves Klein–Gordon equation (C.1) exactly. Furthermore, it is checked that this propagator also satisfies up to order yΔ+4 the equation (2.2) which is related to the cubic Casimir operator. As an independent check, we also found that conditions for su(1,2)×su(1,2) invariance of the propagator are satisfied: limy2→0(Ln(1)+Ln(2))(y2Δeiqγ2KΔ,q)=0 and similar equations containing Wn(i) are satisfied, where Ln(i) and Wn(i) are differential operators defined in (B.1), (B.2), which act on the i-th variable. Hence it is established that this bulk-to-boundary propagator is an exact solution.By using the bulk-to-boundary propagator a scalar field inside the bulk is reconstructed in terms of a boundary CFT operator. This provides a more explicit expression for the extrapolating dictionary than that of the local state for a scalar field obtained in eq. (3.9) of [15].(3.8)Φ(y,x+,α+,β+,x−,α−,β−,γ)=∫d2x2d2α2d2β2KΔ,q(y,x+,α+,β+,x−,α−,β−,γ;x+2,α2+,β2+,x2−,α2−,β2−)OΔ,q(x2+,α2+,β2+,x2−,α2−,β2−) All W-descendants of the primary scalar operators correspond to the scalar field in the bulk.Now let us switch to a Euclidean space by a Wick rotation.(3.9)x+=z,x−=−z¯,β+=iξ,β−=iξ¯,α+=iζ,α−=−iζ¯ Here z¯, ξ¯ and ζ¯ are complex conjugates of z, ξ, ζ. The metric (1.15) becomes(3.10)ds02=1y2dy2+13dγ2+4y4|dζ−ξdz|2+1y2cosh(2γ)|dz−itanh(2γ)dξ|2+1y21cosh(2γ)|dξ|2 It can be shown that in the region including z12,ξ12,ζ12∼0 and in the y→0 limit KΔ,q behaves as(3.11)KΔ,μ=N(γ1)y8−Δe−iqγ1δ2(z12)δ2(ξ12)δ2(ζ12)+⋯+yΔe−iqγ1(D12ED‾E12)−(Δ+3iq)/4(D12E⁎D‾E⁎12)−(Δ−3iq)/4+⋯ Here the dots stand for terms with higher order powers of y, and D12E,… are obtained by replacing variables in D12,… according to the rule of analytic continuation (3.9). N(γ) is a function of γ:(3.12)N(γ)=∫d2zd2ξd2ζ(1+2e−2γ|z+iξ|2+|z2+ξ2−2izξ+4iζ|2)−(Δ+3iq)/4(1+2e2γ|z−iξ|2+|z2+ξ2+2izξ−4iζ|2)−(Δ−3iq)/4.We define action integral for a scalar field coupled to spin-3 gravity by(3.13)Sscalar=∫dyd2zd2ξd2ζdγg(∇Φ⁎∇Φ+m2Φ⁎Φ). When a solution to the Klein–Gordon equation(3.14)Φ(y,z,ξ,ζ,γ)=∫d2z′d2ξ′d2ζ′KΔ,q(y,z,ξ,ζ,γ;z′,ξ′,ζ′)ϕ(z′,ξ′,ζ′), where ϕ(z,ξ,ζ) is a boundary condition, is substituted into (3.13) and g=y−9/(23) is used, a generating functional for the two-point function is obtained as a surface integral on the boundary by using the standard method [23].1515Note that there is no delta function for γ in the expansion (3.11) and the γ integral factors out in (3.15).(3.15)Sscalar[ϕ,ϕ⁎]∝(∫−∞∞dγN(γ)e−iqγ)∫d2z1d2ξ1d2ζ1∫d2z2d2ξ2d2ζ2ϕ⁎(z1,ξ1,ζ1,z¯1,ξ¯1,ζ¯1)(D12ED‾12E)−(Δ+3iq)/4(D12E⁎D‾12E⁎)−(Δ−3iq)/4ϕ(z2,ξ2,ζ2z¯2,ξ¯2,ζ¯2) Hence both holographic dictionaries of AdS/CFT, the WGKP [23][24] and BDHM dictionaries [13], also hold in the case of 3d spin-3 gravity which couples to a scalar field.4Action integral for 3d spin-3 gravity coupled to a scalar fieldOne of the purpose of this paper is to find out a formulation of 3d spin-3 gravity coupled to a scalar field in terms of 8d auxiliary bulk space–time. A natural action integral for the scalar field (3.13) was found in sec. 3. To make this formulation complete, it is necessary to write down an action for the gravity sector in the 8d auxiliary space–time.In (2.6) and (2.7) we introduced 8d vielbein eμa, spin connection ωμa and gauge connections A=ω+e and A‾=ω−e. By solving flatness conditions for the connections (2.8) and (2.9) solutions to these connections are obtained. If the space–time is restricted to a hypersurface with constant α±, β± and γ, the connections also become solutions to equations of motion of 3d SL(3,R)×SL(3,R) Chern–Simons gauge theory.In 8d space–time Chern–Simons gauge theory does not exist, and it is not possible to formulate the gravity sector in a similar way. It is, however, possible to write down an 8d action which reproduces the same equations of motion for the vielbein at the semi-classical level.Let us write the flatness condition for the connections (2.8), (2.9) in terms of ωa and ea:(4.1)dea+fabcωb∧ec=0,(4.2)dωa+12fabcωb∧ωc+12fabceb∧ec=0. These are linear combinations of the flatness conditions. In terms of components the first equation is given by a torsionless condition(4.3)∇μeνa+fabcωμbeνc=∇νeμa+fabcωbνeμc. Here ∇μ is a covariant derivative for gμν=eμaeaν. If we restrict solutions to (4.3) to satisfy the vielbein postulate,(4.4)(Dμeν)a=∇μeνa+fabcωμbeνc=0, which states that the full covariant derivative of eμa should vanish,1616This is equivalent to the condition of metricity, ∇μgνλ=0. Furthermore all the spin-3 gauge field φμνλ obtained in this paper, (D.1) and those for black hole solutions, satisfy ∇μφνλρ=0. This implies that (4.4) is obeyed, because φμνλ is proportional to tr(eμ{eν,eλ}). then by using (A.9) the spin connection ωμa is expressed in terms of eμa,(4.5)ωμa=112fabceνb∇μecν≡ω˜μa(e).We define the field strength for the gauge field of local frame sl(3,R) transformation.(4.6)Raμν(ωλa)=∂μωνa−∂νωμa+fabcωμbωνc The second equation of the flatness condition (4.2) is now written as(4.7)Raμν(ω˜(e))+fabceμbeνc=0. This equation can be derived from the following action.(4.8)Sspin-3 gravity=∫d8x|e|116πG{fabcebμecνRaμν(ω˜(e))+103Λ8}, where |e|=det(eμa)=332−g is a determinant of the vielbein and Λ8 is the cosmological constant in (5.16). G is a Newton constant. Equivalence of the equations of motion can be shown by using the Jacobi's identity for fabc. Because this is written in terms of the vielbein fields, it would be better, if it could be rewritten in terms of fields without local-frame indices, such as gμν and φμνλ. Under SL(3,R) gauge transformation gμν and φμνλ transform as tensors of 8d diffeomorphism. In the similar case of SL(2,R)×SL(2,R) Chern–Simons theory the corresponding action coincides with an Einstein–Hilbert action with a cosmological constant [29][30]. Rewriting of the action (4.8) in terms of gμν and φμνλ may be achieved through a perturbation theory around the background (2.12), (2.13). This will not be attempted in this paper. On the hypersurface with constant α±, β±, γ, solutions to the equations of motion for (4.8) reduce to those in 3d spin-3 gravity represented by SL(3,R)×SL(3,R) Chern–Simons theory.Action for a free charged scalar field is given by(4.9)Sscalar=∫d8x−g{−∇μΦ⁎∇μΦ−m2Φ⁎Φ}. The equation of motion is Klein–Gordon equation. In sec. 2 it was shown that the solution to Klein–Gordon equation in the metric (1.15) also satisfies the cubic-order differential equation. Hence the Klein–Gordon equation is sufficient to determine the classical solution at least for this case. At this point it is not clear if a cubic-order differential equation exists in other space–times such as black hole space–times, where there is no symmetry algebra to yield a set of Casimir operators. As we saw in sec. 2, however, the quadratic-order differential equation is sufficient to determine the solution. The complex scalar field Φ is assumed to have spin-3 charge q. Then the scaling dimension Δ of Φ is determined by (1.13). The self coupling of scalar fields can be introduced straightforwardly. We propose that the total action(4.10)Stotal=Sspin-3 gravity+Sscalar describes the 3d spin-3 gravity theory coupled to a scalar field. In the limit of large central charge (G→0), the action for spin-3 gravity determines the geometry of the 8d space semi-classically, and Sscalar describes the scalar field in this background.5Black hole solutionsIn this section perturbations around the background solution (2.12) and (2.13) are considered and new solutions to the flatness conditions, black hole solutions, are obtained. First we will consider flat connections A (2.12) and A‾ (2.13) at y=1, which will be denoted as A0 and A‾0, respectively.(5.1)A0=(coshγdx+−sinhγdβ+)t3−12dγt6+(sinhγdx+−coshγdβ+)t7+(β+dx+−dα+)t8,(5.2)A‾0=(−coshγdx−+sinhγdβ−)t1+12dγt6−(β−dx−−dα−)t4+(−sinhγdx−+coshγdβ−)t5 Gauge connections at an arbitrary value of y are obtained by carrying out the following gauge transformations(5.3)A0=b(y)−1A0b(y)+b(y)−1db(y),(5.4)A‾0=b(y)A‾0b(y)−1+b(y)db(y)−1, where b(y)=yt2. A0 and A‾0 also satisfy the flatness conditions.(5.5)F0=dA0+A0∧A0=0,(5.6)F‾0=dA‾0+A‾0∧A‾0=0Now we add small perturbations ψ and ψ‾ to A0 and A‾0, respectively:(5.7)A=A0+ψ,(5.8)A‾=A‾0+ψ‾. Then we impose conditions that A and A‾ should satisfy the flatness conditions (5.5), (5.6). Finally, gauge transformations b(y) are used to obtain connections A and A‾.When ψ and ψ‾ are expanded as ψ=ψ(1)+ψ(2)+⋯ and similarly for ψ‾, where ψ(i) is an infinitesimal one-form at i-th order of perturbation, the flatness conditions to first order read(5.9)dψ(1)+A0∧ψ(1)+ψ(1)∧A0=0,(5.10)dψ‾(1)+A‾0∧ψ‾(1)+ψ‾(1)∧A‾0=0. These conditions will be solved explicitly. By expanding ψ(1) and A0 into a basis of sl(3,R) generators as(5.11)ψ(1)=ψ(1)ata(5.12)A0=A0ata, where summation over a=1,⋯,8 is not shown explicitly, eq. (5.9) is transformed into(5.13)dψ(1)a+fabcA0b∧ψ(1)c=0. Explicit forms of these equations are presented in (E.1) in appendix E. For example, for a=4 the equation is simple: dψ(1)4=0. This is trivially solved locally: ψ(1)4=−dQ1, where Q1 is an arbitrary function. Other equations are also solved by taking appropriate linear combinations of (5.13) which give closed forms. General solutions are presented in (E.2) and those for ψ‾(i)a in (E.3) in appendix E. These solutions contain 16 arbitrary Qn, Q‾n. Actually, these are gauge modes. However, when those modes are changed by amounts which are not single-valued on the torus, then the flat connections before and after the change are inequivalent. For static or stationary black hole solutions the functions Qn and Q‾n have to be chosen such that ψ(1)a and ψ‾(1)a are periodic in variables x±.5.1Asymptotically AdS3 black hole solutions without spin-3 chargeThe flat connections for a black hole without spin-3 charge are obtained by choosing suitable Qn and Q‾n, which yield static or stationary connections. The exact terms dQn, dQ‾n in (E.2), (E.3) are determined to make ψ(1)a and ψ‾(1)a periodic in x±. The results for Qn are presented in (F.1). The results for ψ(1)a and ψ‾(1)a are also presented in (F.2) in appendix F. In these results parameters a and a¯ are the following constants.(5.14)a=2G(M+J),(5.15)a¯=2G(M−J) Here M and J are mass and angular momentum, G a Newton constant.Now, although these results are first-order perturbations in the parameters a and a¯, (5.7), (5.8) with ψ=ψ(1) and ψ‾=ψ‾(1) satisfy the full flatness conditions. This is because ψ(1) and ψ‾(1) are proportional to dx+ and dx−, respectively, and ψ(1)∧ψ(1)=ψ‾(1)∧ψ‾(1)=0 hold. Hence these are exact solutions. The gauge connections A, A‾ with y components are obtained by the gauge transformations (5.3) and (5.4). The vielbein e=12(A−A‾) then yields the metric gμν=(1/2)tr(e)2. The metric tensor for the 8d space–time is presented in (F.5). It turns out this metric satisfies 8d vacuum Einstein equation(5.16)Rμν−12gμνR=−Λ8gμν with Λ8=−36.On a hypersurface Σαβγ with constant α±, β±, γ, the metric (F.5) reduces to(5.17)ds2|Σαβγ=y−2dy2−y−4(y2cosh2γ−4β+β−)dx+dx−+a(dx+)2+a¯(dx−)2+4y4{−aβ−(β+)3−a¯β+(β−)3}dx+dx−−ay2{−3(β+)2cosh2γ+4α+sinh2γ}dx+dx−−a¯y2{−3(β−)2cosh2γ+4α−sinh2γ}dx+dx−−aa¯{y2cosh2γ−6β+β−−4y4(β+β−)3+1y2{9(β+β−)2cosh2γ+16α+α−cosh2γ−12(β+)2α−sinh2γ−12(β−)2α+sinh2γ}}dx+dx−. The induced metric (5.17) is also a solution to the equation of motion of 3d SL(3,R)×SL(3,R) Chern–Simons gauge theory. So this is a new black hole solution in the 3d space–time. This black hole does not have spin-3 charge. On a hypersurface Σ0 with α=β=γ=0 this metric coincides with that of BTZ black hole [25]. This metric changes from one Σαβγ to another Σαβγ′, when the values of β±, α±, γ are changed. If β±≠0, the leading behavior of the metric near y∼0 is y−4 and the space–time is asymptotically AdS with AdS length =1/2. As for the spin-3 field we checked that as in (2.4) φ for these flat connections A, A‾ satisfy the 8d equation,(5.18)∇μφνλρ=0. Complete result for the spin-3 field will not be presented here, because it is more complicated. On the hypersurface Σ0, where α±=β±=γ=0, it vanishes.(5.19)φ|Σ0=0 Hence the hypersuface Σ0 is exactly the BTZ black hole. On other Σ's spin-3 field φ does not vanish.The Hawking temperatures of the black hole (5.17) can be obtained by holonomy conditions [26][9][28]. Let us consider the case of finite and non-vanishing a and a¯. A matrix U is defined by the flat connection A=A0+ψ as(5.20)A=U−1dU. On the hypersurface Σαβγ this reduces to Ax+=U−1∂x+U1717Notice that A=Ax+dx+ on the hypersurface and Ax+ does not depend on x+. and U is solved as U=exp(x+Ax+). On the 3d Euclidean asymptotically AdS space, which is obtained by Wick rotation from Σ, the coordinates x+=x+itE≡z and x−=x−itE≡−z¯ are identified as (z,z¯)∼(z+2πτ,z¯+2πτ¯), where τ and τ¯ are modular parameters of the boundary tori. A holonomy matrix w is defined by U(z,z¯)−1U(z+2πτ,z¯+2πτ¯)=expw. Hence w is given by(5.21)w=2πτAx+. This is computed by using (5.7), (5.1) and (F.2). Similarly A‾=A‾0+ψ‾ defines w¯.(5.22)w¯=2πτ¯A‾x−.By requiring that the flat connections are non-singular, the matrices w, w¯ should be required to have the same eigenvalues as those for the vacuum. Hence they need to satisfy the conditions, [9](5.23)detw=0,(5.24)trw2=−8π2, and similar equations for w¯. It can be shown that the first condition is trivially satisfied. The second one yields(5.25)τ=i2a,τ¯=i2a¯. Since τ is related to the inverse right and left temperatures, βR and βL, as τ=i2πβR and τ¯=i2πβL, respectively, we obtain(5.26)TR=1βR=aπ=1π2G(M+J), where M and J are mass and angular momentum of the black hole. Similarly, for the left inverse temperature we have(5.27)TL=1βL=a¯π=1π2G(M−J). Hence α±, β± and γ do not appear in the temperatures.Now let us investigate whether the β+=−β−≡λ→∞ limit of the metric (5.17) exists. Some calculation shows that even if coordinates y, x± are rescaled,1818In the case of the metric (1.16) y must also be rescaled as y=λ1/2y˜ in order to take a finite limit of the metric as λ→∞, where β+=−β−=λ. such a limit does not exist, unless a=a¯=0. Hence in the case of a black hole, it is not possible to argue existence of a fixed point along β+=−β−≡λ by this method. This issue may be settled, if two-point functions of W primary operators in the boundary CFT at finite temperature, which is dual to (5.17), can be obtained.However, if a and a¯ are also rescaled in an appropriate way, finite limits exist. Because this may produce new solutions, we will study such limits. We perform the following rescaling of variables in (5.17),1919Here the AdS length is set to 1. The metric on the hypersurface Σαβγ is a solution to the equations of motion of 3d SL(3,R)×SL(3,R) Chern–Simons theory for each value of α±, β± and γ. On Σαβγ, β± are not coordinates, but just constants. Hence the constants a, a¯ and other variables can be rescaled and made dependent on λ.(5.28)y=λρy,x±=λρx±, as well as(5.29)a=λ−2ρa,a¯→λ−2ρa¯, while α± and γ are fixed. The metric has a finite limit for the constant ρ≥1. The limit depends on (1) ρ=1 or (2) ρ>1.(1)ρ=1:In the limit λ→∞, the 3d metric (5.17) asymptotes to(5.30)ds2|Σαβγ→y−2dy2+a(dx+)2+a¯(dx−)2−[4y4(a−1)(a¯−1)+1y2(3a−1)(3a¯−1)cosh2γ+6aa¯+y2aa¯cosh2γ]dx+dx−. If (a−1)(a¯−1)>0, the signature of the metric is correct and this is an asymptotically AdS black hole solution with AdS length ℓAdS′=12. If (a−1)(a¯−1)=0, this is an asymptotically AdS black hole with the AdS length ℓAdS=1. Metric (5.30) is a solution to the equations of motion in the spin-3 gravity based on the 3d Chern–Simons theory. The 3d metric (5.30) depends on the parameter γ in addition to a and a¯, the mass and angular momentum. As for the spin-3 field, it also has a well-defined λ→∞ limit.(5.31)φ|Σαβγ→a˜y2dy2dx+−a¯y2dy2dx−+12y4[(a+1)2(a¯−1)+aa¯(3a+1)y4−ay2{(1+a)(1−3a¯)−aa¯y4}cosh2γ](dx+)2dx−+12y4[(a¯+1)2(−a+1)−aa¯(3a¯+1)y4+a¯y2{(1+a¯)(1−3a)−aa¯y4}cosh2γ](dx−)2dx+−12y3[(1−3a)(1−3a¯)−aa¯y4]sinh2γdydx+dx−.(2)ρ>1:In the limit λ→∞, the 3d metric (5.17) asymptotes to(5.32)ds2|Σαβγ→y−2dy2+a(dx+)2+a¯(dx−)2−cosh2γ[y−2+aa¯y2]dx+dx−. This coincides with the metric at α±=β±=0 There are terms which contain γ in (5.32). Due to the factor coshγ this is a deformed BTZ solution. As for the spin-3 field, it also has a well-defined λ→∞ limit.(5.33)φ|Σαβγ→12y3(−1+aa¯y4)sinh2γdx+dx−dy5.2Black hole solution with spin-3 chargesIn this subsection the functions Qn, Q‾n and connections ψ, ψ‾ for the black hole solution with spin-3 charge will be constructed. This is more difficult than the preceding black hole, because more parameters than the mass and angular momentum must be introduced and the integrability condition for the partition function needs to be taken into account [9]. The result is presented in (G.1), (G.2) in appendix G. The result includes parameters b,b¯,μ,μ¯, which are spin-3 charges and chemical potentials, in addition to a and a¯. ψ(1) and ψ‾(1) are contributions to first order of these parameters. When the black hole has spin-3 charges, however, ψ and ψ‾ do not stop at the first order. It is then necessary to expand ψ as ψ=ψ(1)+ψ(2)+⋯ and systematically solve the equation for ψ(i). The i-th-order perturbation ψ(i) needs to satisfy(5.34)dψ(i)+A0∧ψ(i)+ψ(i)∧A0=−∑k=1i−1ψ(k)∧ψ(i−k). Suppose that ψ(k) for k=1,2,⋯,i−1 has been obtained. Then the right hand side of (5.34) can be evaluated by using these results. By taking appropriate linear combinations of the a-th components (a=1,2,⋯,8) of the left hand side such that the left hand side becomes exact forms, the equations for ψ(i)a can be solved locally. Ambiguities associated with the solutions to the homogeneous equations for ψ(i)a must be used to make ψ(i)a periodic in x+ and to keep Fefferman–Graham gauge for the metric, i.e., ψ(i)2=0. This forces us to introduce new functions Qn(2) and Qn(3) at the second and third orders of perturbation, which will yield new contributions to ψ(2)a and ψ(3)a according to (E.2). It is also necessary to introduce other new terms to ψ(2)a, which are proportional to the second order infinitesimal parameters, μa, μb, but otherwise must have been included in the first order connection ψ(1)a. The results for ψ(2) and ψ‾(2) are given in (G.3). We carried out analysis to the fourth order and the results for ψ(3) are presented in (G.4). These solutions contain extra constant parameters ζi, ζ¯i (i=1,2,3), which will be determined by the condition of integrability shortly. As will be clear from the result, the solution ψ does not have – components, while ψ‾ does not have + components. The perturbation expansions do not seem to terminate at a finite order.The metric of the black hole solution with spin-3 charge up to the first order in b,b¯,μ,μ¯ is given by(5.35)ds2=ds02+ds12, where ds02 is given in (1.15) and the other terms are presented in (5.35). On Σ0 the black hole metric does not coincide with that of the black hole solution with spin-3 charge obtained in [9].Let us consider holonomy properties of this black hole. In this case the non-vanishing element of Aμ on Σαβγ is only Ax+. Then A=U−1dU is solved as U=exp(x+Ax+) and the holonomy is given as (5.21). Then up to the fourth order of perturbation the conditions (1) detw=0 and (2) trw2+8π=0 are given by the equations,(5.36)(1)27b2ζ22μ3τ3+9bτ3{3+a(−6ζ2+ζ1(8+ζ2))μ2}+a2μτ3{27+27ζ3−2aζ13μ2+9ζ1(−1+8aζ3μ2)}=(4th-order terms),(5.37)(2)3+12aτ2−12b(−4+ζ2)μτ2+4a2(ζ12+12ζ3)μ2τ2=(5th-order terms), respectively. Here ζ1, ζ2 and ζ3 are constants like a, b, μ in the solutions as mentioned above. On the right hand sides of the above equations there are terms which are fourth- and fifth-order in the expansion parameters, respectively, and depend also on α+ and β+. At the present (fourth) order of perturbation some terms on the left hand side of the first equation are of higher orders than those on the right hand side. The left hand sides do not depend on the values of α+, β+ and γ of Σαβγ. We found that at each order of the perturbation theory (from the first to the fourth order) the leading terms on the right hand sides are exactly canceled out by new contributions from the next-order perturbation and instead, new terms which are higher orders in the expansion parameters appear. Those terms on the left hand sides in the above two equations still do not change. If this pattern observed in the perturbative calculation persists to all higher orders, ultimately (5.36) and (5.37) will converge to the following.(5.38)(1)27b2ζ22μ3τ3+9bτ3{3+a(−6ζ2+ζ1(8+ζ2))μ2}+a2μτ3{27+27ζ3−2aζ13μ2+9ζ1(−1+8aζ3μ2)}=0,(5.39)(2)3+12aτ2−12b(−4+ζ2)μτ2+4a2(ζ12+12ζ3)μ2τ2=0 Henceforth we will assume that this is valid.Then (5.39) is solved as(5.40)b=1+4τ2a+(1/12)(ζ12+12ζ3)ν2a2(ζ2−4)ντ. Here ν is defined by(5.41)ν=4μτ. By substituting this into (5.38) and differentiating the result with respect to τ and ν, respectively, and we solve the differentiated equations for (∂a/∂ν)τ and (∂b/∂τ)ν. Further we require that these two differential coefficients should satisfy(5.42)(∂a∂ν)τ=(∂b∂τ)ν. This equation (5.42) is the integrability condition for the partition function of the black hole [9],(5.43)Z=Treiπ2Gτ(a+μb)e−iπ2Gτ¯(a¯+μ¯b¯). U=a/4G is the energy, q=−b/G the spin-3 charge and μ the chemical potential. These conditions determine ζn (n=1,2,3) uniquely.(5.44)ζ1=−2,ζ2=−8,ζ3=1 Then the right-mover temperature is given by TR=i/(2πτ). τ and ν(=4μτ) can be determined from (5.38) and (5.39) as functions of a and b. A similar analysis for the left mover can also be carried out. The conditions for w¯ similar to (5.38) and (5.39) with bars on the parameters ensure the integrability of the partition function, if the following values of the parameters are chosen.(5.45)ζ¯1=−2,ζ¯2=8,ζ¯3=1 We need to set ν¯=4μ¯τ¯. Then the left-mover temperature is given by TL=−i/(2πτ¯). Entropy of the black hole S=SR+SL can be obtained by the method used in [9]. The right-moving part SR is given by SR=(πℓAdS/2G)af(27b2/2a3), where f(y)=cosθ, θ=arctan[y(2−y)/6(1−y)]. The entropy and partition function do not depend on α and β. A scalar field operator in this black hole background will be dual to all W-descendants of some scalar operator in a W3-extended CFT at finite temperature.5.3Black hole solution with spin-3 charge on Σ0As mentioned above, although the flat connections are already complicated even at the fourth order of perturbation, miraculous cancellation occurs in the holonomy conditions, (5.38) and (5.39). The variables α±, β± and γ corresponding to Σαβγ do not appear in these conditions. This situation is similar to that in the black hole solution without spin-3 charge, which was observed in subsec. 4.1. So, let us study the holonomy conditions for the flat connections on the hypersurface Σ0, where α±=β±=γ=0. From the results in appendix D, after setting α±=β±=γ=0 the flat connections on Σ0 are given by(5.46)A|Σ0=((−a+ζ2bμ)t1+t3−(b+ζ3a2μ)t4−ζ1aμt6−μt8)dx+,(5.47)A‾|Σ0=(−t1+(a¯+ζ¯2b¯μ¯)t3−μ¯t4−ζ¯1a¯μ¯t6−(b¯+ζ¯3a¯2μ¯)t8)dx−. These are solutions to the equations of motion in 3d Chern–Simons theory with the boundary conditions, A−=0 and A‾+=0. Because these connections are proportional to dx+ and dx−, respectively, the flatness condition is trivially satisfied and they do not determine the parameters. Only the conditions of holonomy (5.38) and (5.39) determine these parameters and the results agree with (5.44) and (5.45). These are the only consistent conditions. Hence this may support the expectation that the conditions (5.38) and (5.39) will remain valid, even if higher-order terms which depend on α, β, γ are included in ψ and ψ‾. From (5.46)–(5.47) the metric is obtained as follows.(5.48)ds2|Σ0=1y2dy2+(a+12bμ+163a2μ2)(dx+)2+(a¯+12b¯μ¯+163a¯2μ¯2)(dx−)2−13y4{3y2+12μμ¯+8aa¯μμ¯y4+12y8(b+a2μ)(b¯+a¯2μ¯)+3y6(a+8bμ)(a¯+8b¯μ¯)}dx+dx−General black hole solutions with spin-3 charge will be obtained by making a, b and μ in A|Σ0 depend on x+. Similarly, a¯, b¯ and μ¯ in A‾|Σ0 are replaced by functions of x−.(5.49)A|Σ0={−(a(x+)+8b(x+)μ(x+))t1+t3−(b(x+)+a(x+)2μ(x+))t4+2a(x+)μ(x+)t6−μ(x+)t8}dx+,A‾|Σ0={−t1+(a¯(x−)+8b¯(x−)μ¯(x−))t3−μ¯(x−)t4+2a¯(x−)μ¯(x−)t6−(b¯(x−)+a¯(x−)2μ¯(x−))t8}dx−. These connections are still flat. These connections are further required to satisfy the holonomy conditions detw=0 and trw2=−8π2, which are similar to (5.38) and (5.39) but more complicated. Here the holonomy matrix w is defined as before now by using a path-ordered exponential U(x+)=Pexp{−∫x+x0+A+(x′+)dx′+}. It is known that when b=μ=b¯=μ¯=0, the metric constructed from these connections is the most general BTZ metric in the Fefferman–Graham gauge [25]. When A is transformed as A→U−1AU+U−1dU with U=exptaλa(x+) by restricting infinitesimal parameters λa to keep the form of A and imposing δμ=0, transformations δa and δb are obtained. For μ=0 these transformations generate the W3 algebra [3].Next we consider connections which do not satisfy A−,A‾+=0.(5.50)A′=(t3−at1−bt4)dx++μ(8bt1+a2t4−2at6+t8)dx−,(5.51)A‾′=−(t1−a¯t3−b¯t8)dx−+μ¯(8b¯t3+t4−2a¯t6+a¯2t8)dx+ It can be shown that these connections are flat and up to change of notations these coincide with eq. (5.1) of [9]. Then the metric derived from these connections is given as follows and does not coincide with (5.48).(5.52)ds′2=1y2dy2+13{3a(1+8b¯μ¯y2)+4a¯2μ¯(3by4+4μ¯)}(dx+)2+13{3a¯(1+8bμy2)+4a2μ(3b¯y4+4μ)}(dx−)2−{1y2+4μμ¯y4+43(9bμ+9b¯μ¯+2aa¯μμ¯)+4y4(bb¯+a2a¯2μμ¯)+y2(aa¯+64bb¯μμ¯)}dx+dx− It is checked that the spin-3 field obtained from (5.46)–(5.47) are also different from that computed from (5.50)–(5.51). The conditions of holonomy can be obtained by w=2π(τA+′−τ¯A−′) and its barred counterpart. The conditions (1) detw=0 and (2) trw2+8π=0 read(5.53)(1)27b2ν3+18a2ντ2−2a3ν3+27abν2τ+27bτ3=0,(5.54)(2)3+12aτ2+36bντ+4a2ν2=0, where ν=4τ¯μ. These exactly coincide with (5.38)–(5.39) when (5.41) and (5.44) are substituted. Similar conditions for the barred quantities also coincide.2020For barred quantities we must set ν¯=4τμ¯. Then it can be checked that the integrability condition (5.42) also holds in the gravity theory defined by the flat connections (5.50) and (5.51). Hence although the asymptotic behaviors of (5.46)–(5.47) and (5.50)–(5.51) as y→0 are different, and the 3d flat connections are not gauge equivalent, these two sets of flat connections define two distinct bulk gravity geometries which have the same partition function, when the parameters are appropriately identified. This means that the partition functions of the would-be CFT's dual to each background will coincide. These 8d flat connections might be related by large gauge transformations.Let us study the boundary conditions for the connections. The variation of the CS action (1.7) is given by(5.55)δSCS[A′]=k2π∫MtrδA′∧(dA′+A′∧A′)+k4π∫∂Mtr(A+′δA−′−A−′δA+′)d2x. To make this vanish after the bulk equation of motion is used, δA′ must satisfy tr(A+′δA−′−A−′δA+′)=0 on the boundary. Usually, the boundary condition A−′a=0 or A+′a=0 is imposed as (5.49). The connection (5.50), however, satisfies the following conditions, one for each canonical pair,(5.56)A+′3=y−1,A−′8=μy−2,A−′2=A′3−=A−′7=A+′6=A+′7=A+′8=0, where A′ is obtained from A′ by the transformation (5.3). These determine the Dirichlet conditions. Then a variation of the action vanishes after addition of extra local terms (k/4π)∫y=ϵ(4y−1A−′1+16μy−2A′4+)d2x to the action (1.7). Here ϵ is a UV cutoff and finally a limit ϵ→0 must be taken. Under this variation μ should not be changed. Hence there exist appropriate boundary conditions.Remaining problem is how to compute partition functions in an explicit manner. It can also be shown that the metric (5.48) is in a wormhole gauge as the black hole solution in [9]. Note, however, that in the limit a, b, μ→0 only the flat connections (5.46)–(5.47) correspond to the 8d vielbein which reproduces the spin-3 field (D.1) and the coefficient function of the cubic equation for the scalar field.6Summary and discussionsIn this paper a formulation of 3d spin-3 gravity coupled to a scalar field is studied from the view point of a realization of W3 symmetry in the bulk space–time, not in the local frame. It is shown that this formulation is possible in the extended 8d space. In the most symmetric case this is a group manifold SU(1,2). In this 8d space holographic duality between the bulk and the boundary is explicitly realized. The ordinary 3d bulk is obtained by restricting the space–time to hypersurfaces Σαβγ which have constant values of α±, β± and γ. The action for both spin-3 gravity and matter fields are obtained explicitly as 8d integrals. To the leading order of 1/c expansion the 8d vielbein field eμa is obtained as a solution to the problem of flat connections in 8d space–time, and new black hole solutions are obtained. In this case the 8d space is a deformation of the SU(1,2) manifold. The black hole solutions with and without spin-3 charge are found and their flat connections satisfy Ax−=A‾x+=0. The partition function of the black hole with spin-3 charge on Σ0 is found to coincide with that of the black hole solution with different boundary condition obtained before in [9]. Further investigation of the black hole solution (5.46)–(5.47) is necessary.In sec. 1 it is shown that the scaling dimension Δ of a scalar operator on the boundary is related to the scalar mass m and spin-3 charge q by (1.14). This is similar to the ordinary dictionary for the simple AdS3 gravity, Δ=1+1+m2, but slightly modified. In the case of 3d higher-spin gauge theory dual to WN minimal model it was shown that a scalar field can be consistently coupled to hs[μ] theory, if and only if the mass of the scalar field satisfies m2=μ2−1 [6][5]. In the case of the spin-3 theory it turns out μ=N=3 and hence m2=8 is singled out. The mass formula (1.14), however, does not restrict the value of m. This is not a contradiction, because the two models of spin-3 gravity are distinct. Let us note that the central charge c of the algebra (1.3) is arbitrary in principle, and it is taken to be infinite in the semi-classical treatment, while the central charge (1.2) for N=3 is finite. The scalar field Φ in our model has W0 charge q.There are several questions which are not considered in this paper. In the case of the bulk space–time with the full su(1,2)×su(1,2) symmetry, a scalar field satisfies the cubic-order differential equation (2.2) in addition to Klein–Gordon equation (C.1). Then, in the case of black hole solutions, or in the case of more general asymptotically AdS space–times, does a cubic-order differential equation, which is compatible with Klein–Gordon equation, exist and is it related to the spin-3 gauge field as in (2.2)? When there is no matter field, the SL(2,R)×SL(2,R) Chern–Simons theory of 3d gravity is renormalizable [30]. When there is no matter field in our 8d formulation of spin-3 gravity, the classical equations of motion are also conditions of flat connections and there is no physical degrees of freedom inside the bulk. When the action integral is rewritten into the metric-like formalism, is there still no graviton in the bulk? In the case of the ordinary 3d gravity Riemann tensor can be expressed in terms of Ricci tensor. Is there a similar identity in spin-3 gravity? In this paper a bulk-to-boundary propagator for a scalar field is calculated. If the bulk-to-bulk propagator is obtained, conformal blocks of W3 extended CFT may be studied by using the methods of holography.Finally, we had to introduce 8d space–time, which is a deformation of SU(1,2), to realize holographic duality of W3 CFT and spin-3 gravity. In [31][32][33][34] it was argued that the higher spin gauge theory might appear as some special limiting case of string field theory. Recently, there are also works on higher-spin gauge theory from the point of view of string theory and the tensionless limit is studied [35][36][37][38]. These are related to W∞ algebra. It is not clear how to embed the result of the present paper in the string theory. We hope to report on this issue elsewhere.Appendix AConventionsConventions for generators of sl(3,R) and su(1,2) algebras are summarized.A.1sl(3,R) algebraGenerators of sℓ(3,R) in the fundamental representation are given [3] by(A.1)t1=(000100010),t2=(10000000−1),t3=(0−2000−2000),t4=(000000200),t5=(0001000−10),t6=23(1000−20001),t7=(0−20002000),t8=(008000000) The structure constants fabc are defined by(A.2)[ta,tb]=fabctc, and a Killing metric hab is given by(A.3)hab=12tr(tatb), and its nonzero components are h22=1,h13=−2,h48=8,h57=−2,h66=4/3. Indices of the local frame are raised and lowered by hab and its inverse hab. Then fabc≡fabdhdc is completely anti-symmetric. Non-vanishing structure constants are given by(A.4)f123=−2,f158=8,f167=−4,f248=−16,f257=2,f347=8,f356=−4The completely symmetric invariant tensor dabc is defined by(A.5)dabc=12tr{ta,tb}tc. These constants are given by(A.6)d127=d235=−2,d136=d226=d567=43,d118=d334=8,d468=323,d477=d558=−8,d666=−169.Casimir operators are given by(A.7)C2=habTaTb,(A.8)C3=dabcTaTbTc, where Ta is some irreducible representation of sl(3,R). For the adjoint representation (Ta)bc=−fbac the following relations are obtained.(A.9)facdfbcd=−12hab,(A.10)dcdefcaffdfegbfg=0A.2su(1,2) algebraGenerators of su(1,2) algebra are obtained by replacing the matrices of sl(3,R) generators (A.1) as ta→t˜a, where(A.11)t˜j=tj(j=1,2,3),t˜α=itα(α=4,…8). These new matrices satisfy the relation (t˜a)†η+ηt˜a=0, where dagger stands for hermitian conjugation and η is the following matrix.(A.12)η=(0010−10100) The matrices t˜a generate the W3 wedge algebra (1.5) after an identification t˜1=L1, t˜2=L0, t˜3=L−1, t˜4=W2, t˜5=W1, t˜6=W0, t˜7=W−1 and t˜8=W−2.Killing metric h˜ab=(1/2)tr(t˜at˜b) is given by h˜22=1, h˜13=−2, h˜48=−8, h˜57=2, h˜66=−4/3. The structure constants are given by f˜123=−2, f˜158=−8, f˜167=−4, f˜248=16, f˜257=−2, f˜347=−8, f˜356=4. The constants d˜abc are given by d˜127=d˜235=−2i, d˜136=d˜226=−d˜567=43i, d˜118=d˜334=8i, d˜468=−323i, d˜477=d˜558=8i, d˜666=169i. Casimir operators are given by(A.13)C˜2=h˜abt˜at˜b,(A.14)C˜3=d˜abct˜at˜bt˜c.Appendix BRepresentation of W3 generators by differential operatorsHere the infinite-dimensional representation of su(1,2)×su(1,2) algebra in the hyperbolic representation [17] is presented. This is a representation for the generators of transformations in the bulk. This is obtained from eqs. (4.3) and (4.4) of [15] by replacements α±→iα±, β±→iβ±, γ→iγ.(B.1)L−1h=i∂x+,L0h=−x+∂x+−2α+∂α+−β+∂β+−12y∂y,L1h=−i[(x+)2+3(β+)2]∂x+−ix+y∂y−3iβ+∂γ−i[2(β+)3+4x+α+]∂α+−i[2x+β++4α+]∂β+−iy2cosh(2γ)∂x−−iy2cosh(2γ)β−∂α−−iy2sinh(2γ)∂β−,(B.2)W−2h=i∂α+,W−1h=−x+∂α+−∂β+,W0h=−2iβ+∂x+−i[(x+)2+(β+)2]∂α+−2ix+∂β+−i∂γ,W1h=3x+∂γ+[−4α++6x+β+]∂x++[(x+)3+3x+(β+)2]∂α++[3(x+)2+(β+)2]∂β++β+y∂y+y2sinh(2γ)∂x−+β−y2sinh(2γ)∂α−+y2cosh(2γ)∂β−,W2h=−i[3(β+)4−(x+)4+16(α+)2−6(x+)2(β+)2]∂α++i[y4−4y2β−β+cosh(2γ)+4y2β−x+sinh(2γ)]∂α−+i[−16α+β++4x+(β+)2+4(x+)3]∂β++iy2[−4β+sinh(2γ)+4x+cosh(2γ)]∂β−−i[4(β+)3−12β+(x+)2+16x+α+]∂x++iy2[−4β+cosh(2γ)+4x+sinh(2γ)]∂x−+i[−6(β+)2+6(x+)2]∂γ+i[−8α++4x+β+]y∂y These generators satisfy the algebra (1.5). Generators L‾nh, W‾nh are obtained from the above by interchanges, x+↔x−, α+↔α−, β+↔β−.Appendix CKlein–Gordon equation for a scalar field in the 8d space–timeHere an explicit form of the Klein–Gordon equation for |Φ〉, (1.12), is presented.(C.1)[y2∂y2−7y∂y+3∂γ2−4y2cosh2γ∂x+∂x−−4y2cosh2γ(β−∂x+∂α−+β+∂x−∂α+)−4y2sinh2γ(∂x+∂β−+∂x−∂β+)−4y2sinh2γ(β+∂α+∂β−+β−∂α−∂β+)+(y4−4β+β−y2cosh2γ)∂α+∂α−−4y2cosh2γ∂β+∂β−−m2]|Φ〉=0Appendix DSpin-3 field ϕμνλ in (2.2)An explicit form of the spin-3 field in 8d space appearing in the equation (2.2) for the cubic Casimir operator is given by2121The indices of ϕμνλ are lowered by gμν.(D.1)ϕ≡ϕμνλdxμdxμdxλ=−136y4{−27β−dx−(dx+)2−27β+dx+(dx−)2+27ysinh2γdydx+dx−+27(dx+)2dα−+27(dx−)2dα+−27dα+(dβ−)2−27dα−(dβ+)2+72dα+dα−dγ+18y2dy2dγ−2y4dγ3+27β−dx−(dβ+)2+27β+dx+(dβ−)2−72β−dx−dα+dγ−72β+dx+dα−dγ+72β+β−dx+dx−dγ+(−27ydx+dβ−dy−27ydx−dβ+dy+9y2dx+dx−dγ+9y2dβ+dβ−dγ)cosh2γ+(27ydβ+dβ−dy−9y2dx+dβ−dγ−9y2dx−dβ+dγ)sinh2γ}.Appendix EEquations for ψ(1)a and ψ‾(1)aBy substituting (5.1) and (5.2) into (5.13) and a similar equation for ψ‾, explicit equations for ψ(1)a and ψ‾(1)a are obtained. These are listed below.(E.1)•dψ(1)1+4(sinhγdx+−coshγdβ+)∧ψ(1)4+dγ∧ψ(1)5=0,•dψ(1)2+2(−coshγdx++sinhγdβ+)∧ψ(1)1+16(β+dx+−dα+)∧ψ(1)4+2(−sinhγdx++coshγdβ+)∧ψ(1)5=0,•dψ(1)3−(coshγdx+−sinhγdβ+)∧ψ(1)2+4(β+dx+−dα+)∧ψ(1)5−dγ∧ψ(1)7+2(−sinhγdx++coshγdβ+)∧ψ(1)6=0,•dψ(1)4=0,•dψ(1)5+dγ∧ψ1−4(coshγdx+−sinhγdβ+)∧ψ(1)4=0,•dψ(1)6+3(−sinhγdx++coshγdβ+)∧ψ(1)1−3(coshγdx+−sinhγdβ+)∧ψ(1)5=0,•dψ(1)7+4(−β+dx++dα+)∧ψ(1)1+(−sinhγdx++coshγdβ+)∧ψ(1)2−dγ∧ψ(1)3−2(coshγdx+−sinhγdβ+)∧ψ(1)6=0,•dψ(1)8+2(−β+dx++dα+)∧ψ(1)2+(sinhγdx+−coshγdβ+)∧ψ(1)3−(coshγdx+−sinhγdβ+)∧ψ(1)7=0 The equation for ψ‾(1)a is obtained by the following replacement.x+→x−,α+→α−,β+→β−,ψ(1)1→ψ‾(1)3,ψ(1)2→ψ‾(1)2,ψ(1)3→ψ‾(1)1,ψ(1)4→ψ‾(1)8,ψ(1)5→ψ‾(1)7,ψ(1)6→ψ‾(1)6,ψ(1)7→ψ‾(1)5,ψ(1)8→ψ‾(1)4(E.2)ψ(1)1=sinhγ{−4Q1dx++dQ3}+coshγ{dQ2+4Q1dβ+},ψ(1)2=(−16β+Q1−2Q2)dx++16Q1dα+−2Q3dβ++dQ4,ψ(1)3=coshγ{(−4β+Q3−Q4)dx+−2Q5dβ++4Q3dα++dQ6}+sinhγ{(−4β+Q2+2Q5)dx++Q4dβ++4Q2dα+−dQ7},ψ(1)4=−dQ1,ψ(1)5=−sinhγ(4Q1dβ++dQ2)+coshγ(4Q1dx+−dQ3),ψ(1)6=3Q3dx++3Q2dβ+−dQ5,ψ(1)7=coshγ{(−4β+Q2+2Q5)dx++Q4dβ++4Q2dα+−dQ7}+sinhγ{(−Q4−4β+Q3)dx+−2Q5dβ++4Q3dα++dQ6},ψ(1)8=(−2β+Q4+Q7)dx++2Q4dα+−Q6dβ+−dQ8. Here Qn (n=1,…,8) are arbitrary functions.Similarly, general solution to (5.10) is given by(E.3)ψ‾(1)1=coshγ{−4β−Q‾3dx−−Q‾4dx−−2Q‾5dβ−+4Q‾3dα−+dQ‾6}+sinhγ{−4β−Q‾2dx−+2Q‾5dx−+Q‾4dβ−+4Q‾2dα−−dQ‾7},ψ‾(1)2=−16β−Q‾1dx−−2Q‾2dx−+16Q‾1dα−−2Q‾3dβ−+dQ‾4,ψ‾(1)3=coshγ{4Q‾1dβ−+dQ‾2}+sinhγ{−4Q‾1dx−+dQ‾3},ψ‾(1)4=(−2β−Q‾4+Q‾7)dx−+2Q‾4dα−−Q‾6dβ−−dQ‾8,ψ‾(1)5=sinhγ{−4β−Q‾3dx−−Q‾4dx−−2Q‾5dβ−+4Q‾3dα−+dQ‾6}−coshγ{4β−Q‾2dx−−2Q‾5dx−−Q‾4dβ−−4Q‾2dα−+dQ‾7},ψ‾(1)6=3Q‾2dβ−+3Q‾3dx−−dQ‾5,ψ‾(1)7=−sinhγ{4Q‾1dβ−+dQ‾2}+coshγ{4Q‾1dx−−dQ‾3},ψ‾(1)8=−dQ‾1. Hence there are 16 perturbative modes Qn, Q‾n in the classical solutions. Actually, these are gauge modes. However, when those modes are changed by amounts which are not single-valued functions on the torus, then the flat connections before and after the change are inequivalent. For static or stationary black hole solutions the functions Qn and Q‾n have to be chosen such that ψ(1)a and ψ‾(1)a are periodic in x±.Appendix FBlack hole solution without spin-3 chargeThe flat connections for a black hole without spin-3 charge are obtained by choosing suitable Qn and Q‾n in (E.2) and (E.3), which yield static or stationary connections.(F.1)Q1=Q3=0,Q2=−ax+,Q4=−a(x+)2,Q5=−3ax+β+,Q6=−13a(x+)3−3ax+(β+)2,Q7=−a(x+)2β+−4ax+α+,Q8=13a(x+)3β+−2a(x+)2α++ax+(β+)3. For simplicity the variable γ is not included in Qn. Q‾n is obtained by the following replacement.Qn→Q‾n,x+→x−,α+→α−,β+→β−,a→−a¯The first order perturbation ψ(i)a for the black hole solution without spin-3 charge which are obtained by substituting (F.1) into (E.2) is given as follows. Then ψ(1)a and ψ‾(1)a are given by(F.2)ψ(1)1=−acoshγdx+,ψ(1)2=ψ(1)4=0,ψ(1)3=−a{3(β+)2coshγ−4α+sinhγ}dx+,ψ(1)5=asinhγdx+,ψ(1)6=3aβ+dx+,ψ(1)7=a{4α+coshγ−3(β+)2sinhγ}dx+,ψ(1)8=−a(β+)3dx+, ψ‾(1)a is obtained by the following replacement.x+→x−,α+→α−,β+→β−,a→−a¯,ψ(1)1→ψ‾(1)3,ψ(1)2→ψ‾(1)2,ψ(1)3→ψ‾(1)1,ψ(1)4→ψ‾(1)8,ψ(1)5→ψ‾(1)7,ψ(1)6→ψ‾(1)6,ψ(1)7→ψ‾(1)5,ψ(1)8→ψ‾(1)4 In the above equations parameters a and a¯ are the following constants.(F.3)a=2G(M+J),(F.4)a¯=2G(M−J) Here M and J are mass and angular momentum, G a Newton constant.As explained in sec. 5 the perturbation series stops at the first order and the above results give exact solutions. The gauge connections A, A‾ with y components are obtained by the gauge transformations (5.3) and (5.4). The vielbein e=12(A−A‾) then yields the metric gμν=(1/2)tr(e)2.(F.5)ds2=ds02+a(dx+)2+a¯(dx−)2−2aβ+dγdx+−2a¯β−dγdx−+4y4{−aβ−(β+)3−a¯β+(β−)3}dx+dx−−ay2{−3(β+)2cosh2γ+4α+sinh2γ}dx+dx−−a¯y2{−3(β−)2cosh2γ+4α−sinh2γ}dx+dx−−ay2{3(β+)2sinh2γ−4α+cosh2γ}dx+dβ−−a¯y2{3(β−)2sinh2γ−4α−cosh2γ}dx−dβ++4ay4(β+)3dα−dx++4a¯y4(β−)3dα+dx−−aa¯{y2cosh2γ−6β+β−−4y4(β+β−)3+1y2{9(β+β−)2cosh2γ+16α+α−cosh2γ−12(β+)2α−sinh2γ−12(β−)2α+sinh2γ}}dx+dx− where ds02 is the metric (1.15).Appendix GBlack hole solution with spin-3 charge up to third orderFunctions Qn and Q‾n for static or stationary black holes with spin-3 charge are given as follows. It turned out it is necessary to include higher order corrections to ψa and ψ‾a. Higher-order corrections ψ(i)a, ψ‾(i)a are also presented.(G.1)Q1=bx+,Q2=−ax+−4bx+β+−8bα+,Q3=2b(x+)2,Q4=−a(x+)2+4b(x+)2β+−16bx+α+,Q5=−3ax+β++2b(x+)3−6bx+(β+)2−24bα+β+,Q6=−13a(x+)3−3ax+(β+)2−8b(x+)2α++4b(x+)3β+−4bx+(β+)3−24bα+(β+)2,Q7=−a(x+)2β+−4ax+α+−16bx+α+β+−16b(α+)2+2b(x+)2(β+)2+b(x+)4,Q8=13a(x+)3β+−2a(x+)2α++ax+(β+)3+μx++8b(x+)2α+β+−2b(x+)3(β+)2−16bx+(α+)2+b5(x+)5+bx+(β+)4+8bα+(β+)3, Q‾n is obtained by the following replacement.Qn→Q‾n,x+→x−,α+→α−,β+→β−,a→−a¯,b→b¯,μ→μ¯First-order perturbations for flat connections ψ(1)a, ψ‾(1)a are given as follows.(G.2)ψ(1)1=−acoshγdx+−4bβ+coshγdx+−8bcoshγdα+,ψ(1)2=0,ψ(1)3=−a{3(β+)2coshγ−4α+sinhγ}dx+−4b(β+)3coshγdx+−24b(β+)2coshγdα+,ψ(1)4=−bdx+,ψ(1)5=asinhγdx++4bβ+sinhγdx++8bsinhγdα+,ψ(1)6=3aβ+dx++6b(β+)2dx++24bβ+dα+,ψ(1)7=a{4α+coshγ−3(β+)2sinhγ}dx+−4b(β+)3sinhγdx+−24b(β+)2sinhγdα+,ψ(1)8=−a(β+)3dx+−μdx+−b(β+)4dx+−8b(β+)3dα+, ψ‾(1)a is obtained by the following replacement.x+→x−,α+→α−,β+→β−,a→−a¯,b→b¯,μ→μ¯,ψ(1)1→ψ‾(1)3,ψ(1)2→ψ‾(1)2,ψ(1)3→ψ‾(1)1,ψ(1)4→ψ‾(1)8,ψ(1)5→ψ‾(1)7,ψ(1)6→ψ‾(1)6,ψ(1)7→ψ‾(1)5,ψ(1)8→ψ‾(1)4Second-order perturbations for flat connections are given as follows.(G.3)ψ(2)1=ζ2bμsinhγdβ++ζ2bμcoshγdx+,ψ(2)2=0,ψ(2)3=−96ab(α+)2β+coshγdx+−32bμα+sinhγdx++2ζ1aμβ+coshγdx++(−2ζ2bμ(β+)2dx++4ζ2bμβ+dα+)coshγ−4ζ2bμα+sinhγdx+,ψ(2)4=0,ψ(2)5=−ζ2bμsinhγdx+−ζ2bμcoshγdβ+,ψ(2)6=48ab(α+)2dx+−ζ1aμdx+,ψ(2)7=−96ab(α+)2β+sinhγdx+−32bμα+coshγdx++2ζ1aμβ+sinhγdx+−4ζ2bμα+coshγdx++(−2ζ2bμ(β+)2dx++4ζ2bμβ+dα+)sinhγ,ψ(2)8=−48ab(α+β+)2dx++ζ1aμ(β+)2dx++2ζ2bμ(β+)2dα+−23ζ2bμ(β+)3dx+, ψ‾(2)a is obtained by the following replacement.x+→x−,α+→α−,β+→β−,a→−a¯,b→b¯,μ→μ¯,ψ(2)1→−ψ‾(2)3,ψ(2)2→−ψ‾(2)2,ψ(2)3→−ψ‾(2)1,ψ(2)4→−ψ‾(2)8,ψ(2)5→−ψ‾(2)7,ψ(2)6→−ψ‾(2)6,ψ(2)7→−ψ‾(2)5,ψ(2)8→−ψ‾(2)4ζ1→ζ¯1,ζ2→−ζ¯2Finally, third-order perturbations for flat connections are given as follows.(G.4)ψ(3)1=coshγ(H1+dQ2(3))−sinhγH2−4ζ3a2μβ+coshγdx+−8ζ3a2μcoshγdα+,ψ(3)2=0,ψ(3)3=coshγH3+sinhγ(−H4−4β+Q2(3)dx++4Q2(3)dα+)−4ζ3a2μ(β+)3coshγdx+−24ζ3a2μ(β+)2coshγdα+,ψ(3)4=ζ2abμβ+dx++2ζ2b2μ(β+)2dx++8ζ2b2μβ+dα+−ζ3μa2dx+,ψ(3)5=coshγH2−sinhγ(H1+dQ2(3))+4ζ3a2μβ+sinhγdx++8ζ3a2μsinhγdα+,ψ(3)6=−48b2μ(8+ζ2)(α+)2dx+−8abμζ2(β+)3dx+−8b2μζ2(β+)4dx+−64b2μζ2(β+)3dα++3Q2(3)dβ++6ζ3a2μ(β+)2dx++24ζ3a2μβ+dα+,ψ(3)7=sinhγH3+coshγ(−H4−4β+Q2(3)dx−+4Q2(3)dα+)−4ζ3a2μ(β+)3sinhγdx+−24ζ3a2μ(β+)2sinhγdα+,ψ(3)8=48b2μ(ζ2+8)(α+)2(β+)2dx++256ab2(α+)4β+dx+−32abμζ1(α+)2β+dx++23b2μζ2(β+)6dx++2bμ2ζ2(β+)2dx++abμζ2(β+)5dx++8b2μζ2(β+)5dα+−ζ3a2μ(β+)4dx+−8ζ3a2μ(β+)3dα+ Here one-forms Hi (i=1,2,3,4) and the function Q2(3) are as follows.(G.5)H1=5abμζ2(β+)2dx++203b2μζ2(β+)3dx++40b2μζ2(β+)2dα+,H2=256ab2(α+)3dx+−16abμζ1α+dx+,H3=96b2μ(ζ2+8)(α+)2β+dx++256ab2(α+)4dx+−32abμζ1(α+)2dx++4b2μζ2(β+)5dx++5abμζ2(β+)4dx++4bμ2ζ2β+dx++40b2μζ2(β+)4dα+,H4=−256ab2(α+)3(β+)2dx++4abμ(4ζ1+ζ2)α+(β+)2dx+,Q2(3)=−256ab2(α+)3β++4abμ(4ζ1+ζ2)α+β+ ψ‾(3)a is obtained by the following replacement of the above.x+→x−,α+→α−,β+→β−,a→−a¯,b→b¯,μ→μ¯,ψ(3)1→ψ‾(3)3,ψ(3)2→ψ‾(3)2,ψ(3)3→ψ‾(3)1,ψ(3)4→ψ‾(3)8,ψ(3)5→ψ‾(3)7,ψ(3)6→ψ‾(3)6,ψ(3)7→ψ‾(3)5,ψ(3)8→ψ‾(3)4ζ1→ζ¯1,ζ2→−ζ¯2,ζ3→ζ¯3 ζi and ζ‾i are constants to be determined at (5.44) and (5.45) in the main text by the integrability conditions for the partition function.Appendix HMetric for black hole solution with spin-3 charge up to first-order perturbationUp to the first order of perturbation the metric of the black hole solution with spin-3 charge is given by(H.1)ds2=ds02+ds12, where ds02 is given in (1.15) and the other term is given as follows.(H.2)ds12=a(dx+)2+a¯(dx−)2+[4−a(β+)3β−−a¯β+(β−)3−b(β+)4β−+b¯β+(β−)4+β+μ¯−β−μy4+3a(β+)2+3a¯(β−)2+4by2(β+)3−4b¯y2(β−)3y2coshγ−4aα++4a¯α−y2sinhγ]dx+dx−+12bdx+dα+−12b¯dx−dα−+2β+(−a−2bβ+)dx+dγ−2β−(a¯−2b¯β−)dx−dγ+4[a(β+)3+(β+)4b+2(β−)2b¯(4β+β−−3y2cosh2γ)+μ]y4dx+dα−+[4aα+cosh2γ−(β+)2(3a+4bβ+)sinh2γ]y2dx+dβ−+4[a¯(β−)3−(β−)4b¯−2(β+)2b(4β−β+−3y2cosh2γ)−μ¯]y4dx−dα++[4a¯α−cosh2γ−(β−)2(3a¯−4b¯β−)sinh2γ]y2dx−dβ++32[(β+)3b−(β−)3b¯]y4dα+dα−−16bβ+dα+dγ+16b¯β−dα−dγ−24y2b(β+)2sinh2γdα+dβ−+24y2b¯(β−)2sinh2γdα−dβ+.Parameters a, a¯ are related to the mass and the angular momentum as in (5.14), (5.15), and b and b¯ to the spin-3 charges. μ and μ¯ are chemical potentials for the charges. The result for the spin-3 gauge field is not presented. It is also checked that φ satisfies the equation, ∇μφνλρ=0.References[1]J.MaldacenaThe large-N limit of superconformal field theories and supergravityInt. J. Theor. Phys.3819991113[2]M.P.BlencoweA consistent interacting massless higher spin field theory in D=(2+1)Class. Quantum Gravity61989443[3]A.CampoleoniS.FredenhagenS.PfenningerS.TheisenAsymptotic symmetries of three-dimensional gravity coupled to higher-spin fieldsarXiv:1008.4744 [hep-th][4]M.A.VasilievMore on equations of motion for interacting massless fields of all spins in (3+1)-dimensionsPhys. Lett. B2851992225[5]M.R.GabberdielR.GopakumarMinimal model holographyarXiv:1207.6697 [hep-th][6]S.ProkushkinM.VasilievHigher-spin gauge interactions for massive matter fields in 3d AdS space–timearXiv:hep-th/98062363D higher-spin gauge theories with matterarXiv:hep-th/9812242[7]A.B.ZamolodchikovInfinite additional symmetries in two dimensional conformal quantum field theoryTheor. Math. Phys.6519851205[8]P.BouwknegtK.SchoutensW-symmetry in conformal field theoryarXiv:hep-th/9210010[9]P.M.AmmonM.GutperleP.KrausE.PerlmutterSpacetime geometry in higher spin gravityarXiv:1106.4788 [hep-th][10]M.GutperleP.KrausHigher spin black holesJ. High Energy Phys.11052011022arXiv:1103.4304 [hep-th][11]A.L.FitzpatrickJ.KaplanD.LiJ.WangExact Virasoro blocks from Wilson lines and background-independent operatorsarXiv:1612.06385 [hep-th][12]A.CastroN.IqbalE.LlabrésWilson lines and Ishibashi states in AdS3/CFT2arXiv:1805.05398 [hep-th][13]T.BanksM.R.DouglasG.T.HorowitzE.J.MartinecAdS dynamics from conformal field theoryarXiv:hep-th/9808016[14]A.HamiltonD.KabatG.LifschytzHolographic representation of local bulk operatorsPhys. Rev. D742006066009arXiv:hepth/0606141[15]R.NakayamaT.SuzukiA bulk localized state and new holographic renormalization group flow in 3D spin-3 gravityInt. J. Mod. Phys. A331220181850061arXiv:1712.04678 [hep-th][16]M.MiyajiT.NumasawaN.ShibaT.TakayanagiK.Watanabec-Mera as surface/state correspondence in AdS/CFTarXiv:1506.01353 [hep-th][17]K.GotoT.TakayanagiCFT descriptions of bulk local states in the AdS black holesarXiv:1704.00053 [hep-th][18]H.VerlindePoking holes in AdS/CFT: bulk fields from boundary statesarXiv:1505.05069 [hep-th][19]Y.NakayamaH.OoguriBulk locality and boundary creating operatorsarXiv:1507.04130 [hep-th][20]A.CastroE.Hijano CubelosA.Lepage-JutierA.MaloneyBlack holes and singularity resolution in higher spin gravityarXiv:1110.4117 [hep-th][21]A.B.ZamolodchikovIrreversibility of the flux of renormalization group in a 2D field theoryJETP Lett.431986730[22]V.FateevS.RibaultThe large central charge limit of conformal blockarXiv:1109.6764 [hep-th][23]E.WittenAnti-de Sitter space and holographyAdv. Theor. Math. Phys.21998253[24]S.S.GubserI.R.KlebanovA.M.PolyakovGauge theory correlators from noncritical string theoryPhys. Lett. B4281998105arXiv:hep-th/9802109[25]M.BañadosC.TeitelboimJ.ZanelliThe black hole in three-dimensional space-timePhys. Rev. Lett.6919921849arXiv:hep-th/9204099M.BanadosM.HenneauxC.TeitelboimJ.ZanelliGeometry of the (2+1) black holePhys. Rev. D4819931506arXiv:9r-qc/9302012M.BañadosThree-dimensional quantum geometry and black holesarXiv:hep-th/9901148[26]D.CangemiM.LeblancR.B.MannGauge formulation of the spinning black hole in (2+1)-dimensional anti-de Sitter spacePhys. Rev. D4819933606[27]M.BañadosR.CaroHolographic ward identities: examples from 2+1 gravityJ. High Energy Phys.04122004036arXiv:hep-th/0411060[28]P.KrausLectures on black holes and the AdS3/CFT2 correspondencearXiv:hep-th/0609074[29]A.AchucarroP.TownsendA Chern–Simons action for three-dimensional anti-de Sitter supergravity theoriesPhys. Lett. B180198689[30]E.Witten(2+1)-dimensional gravity as an exactly soluble systemNucl. Phys. B311198846[31]D.GrossHigh energy symmetries of string theoryPhys. Rev. Lett.6019881229[32]E.Wittentalk at the John Schwarz 60-th birthday symposium (Nov. 2001)http://theory.caltech.edu/jhs60/witten/1.html[33]B.SundborgStringy gravity, interacting tensionless strings and massless higher spinsNucl. Phys. B, Proc. Suppl.1022001113arXiv:hep-th/0103247[34]A.MarikovNotes on higher spin symmetriesarXiv:heo-th/0201019[35]M.GabberdielR.GopakumarC.HullStringy AdS3 from the world sheetarXiv:1704.08665 [hep-th][36]K.FerreiraM.GabberdielJ.I.JottarHigher spins on AdS3 from the worldsheetarXiv:1704.08667 [hep-th][37]M.GabberdielR.GopakumarTensionless string spectra on AdS3arXiv:1803.04423 [hep-th][38]L.EberhardtM.GabberdielR.GopakumarThe worldsheet dual of the symmetric product CFTarXiv:1812.01007 [hep-th]