Quantum Effective Dynamics and Stability of Vacuum in Anti-de Sitter Spacetimes Shi-Yuan Li, Chengwu Liu Abstract We investigate the details of the canonical quantization of effective quantum field theories in anti-de Sitter spacetime, emphasizing the stability of the quantum vacuum. We take the scalar and Maxwell fields as examples. For the non-minimally coupled massless real scalar field with ξ ​ R ​ ϕ 2 \xi R\phi^{2} term in the Lagrangian (mass can be introduced by shift of ξ \xi ), only when ξ ≤ 5 / 48 \xi\leq 5/48 , the quantized Hamiltonian is spontaneously non-negative and the vacuum is well defined. For ξ > 5 / 48 \xi>5/48 , one has to assign the negative energy spectrum as that of the ghost particles, introducing anti-commutation relations to make the corresponding part of the Hamiltonian trivial, ensuring the Hamiltonian non-negative and the vacuum (and the Hilbert space) well defined. This method of ghost states is applicable once the proper radial boundary conditions guarantee the Hamiltonian self-adjoint. The resulting dynamics can be compared with those resulting from the positive self-adjoint extensions when the latter is available for ξ ≤ 9 / 48 \xi\leq 9/48 . For the Maxwell fields, the gauge invariant canonical energy momentum tensor straightforwardly leads to the gauge invariant non-negative Hamiltonian (well-defined vacuum). Hence the redundant gauge degree of freedom is irrelevant, and the 2-dimensional dynamical degrees of freedom are quantized in a concrete, e.g., temporal gauge. The energy momentum tensors for both quantized fields are renormalized to be finite at operator level, which renders the stable vacuum maximally symmetric. The backreactions to the background spacetime by excited states via the semi-classical Einstein equations are also discussed. 1 Introduction The intrinsic difficulties in constructing the ’completely quantum theory’ of all the matters as well as the gravity (spacetime) indicate that deeper physical principles or deeper understanding on the present quantum principles may be needed. On the other hand, a rather modest way out for deeper understanding on the quantum principles can be the intensive studies on various effective quantum field theories (EQFT) including those of the spacetime perturbations in various backgrounds of curved spacetimes, and reversely on the backreactions from the EQFT source to the geometry. Such studies can also shed light on dilemma in Minkowski spacetime (a recent example on dilemma from infinite momentum integrations, see [ 1 ] ), and on quantum fields interacting with background fields, a practically important topic in optics and condensate matter sustaining from the early time [ 2 ] . The general framework of algebraic quantum field theory in curved spacetime has long been setup, see, e.g., [ 3 , 4 , 5 , 6 , 7 ] for recent reviews. The present quantum principles can not be simply applied to arbitrary manifold (e.g., requiring causal structure), and a specific EQFT model in a concrete spacetime may be better considered as an effective dynamics in this spacetime. There are several special cases of EQFT in curved spacetime need much more concrete studies in details, such as those in our cosmic spacetime and those in the Anti-de Sitter (AdS) spacetime, since for both cases the EQFT and the quantum principles can confront various tests via their predictions relevant to the experimental measurements. EQFT in AdS gives predictions for physical observables with the help of the AdS/CFT correspondence [ 8 , 9 , 10 ] , based on which the (in)stability of the AdS confronting the perturbations from the matter fields is a key question. The AdS instability conjecture by Dafermos and Holzegel in 2006, [ 11 , 12 ] states that there exist arbitrarily small perturbations to AdS initial data which, under evolution by the Einstein vacuum equations for \Lambda<0 with reflecting boundary conditions on conformal infinity, lead to the formation of black holes. The numerical study of this conjecture in the simpler setting of the spherically symmetric Einstein-scalar field system was initiated by Bizon and Rostworowski [ 13 ] , followed by a vast number of numerical and heuristic works by several authors. At present, cases such as the classical Einstein-scalar [ 13 , 14 , 15 ] , pure-gravity [ 16 ] Einstein-Vlassov system [ 17 ] , Einstein-null dust [ 18 ] have been numerically and analytically studied. The seek for rigorous study with simplest dynamics [ 19 , 17 , 18 ] is with non-trivial boundary condition. On the other hand, there are also stable solutions which are named as islands of stability [ 20 , 21 ] such as geons [ 22 ] for pure gravity system and [ 23 ] for Einstein massless scalar system. At the same time, back-reaction in the framework of quantized scalar field is also studied [ 24 , 25 ] with some special boundary conditions. The investigations relevant to the (in)stability of the AdS and AdS/CFT hightlight the need for various concrete EQFT in AdS [ 26 , 27 , 28 , 29 , 30 ] as working models for the extension of the present recognitions. In this paper we study the details of the canonical quantization for EQFT in AdS. We take the scalar field and the electromagnetic field as examples, and concentrate on the consistency to construct the EQFT in this maximally symmetric spacetimes, i.e., the stability of the vacuum, the global Hilbert space and the consistent definition of the quantum (and renormalized) energy momentum tensor. These are essential for the subsequent studies such as the stability of AdS confronting the back-reactions from matter field purterbations (excited states) to the geometry as well as to calculate the correlation functions. The AdS metric is time-independent so that a global Hilbert space possibly exists [ 28 , 30 ] . This is different from the cosmic spacetime, of which the metric and hence the Hamiltonian of a quantum system are explicitly time dependent and in general the global (in time) unique Hilbert space for such quantum system can not be defined; and special Hadamard states are employed to link the global curved cosmic spacetime EQFT and the local Minkowski, where the measurement is performed. In the standard canonical quantization framework [ 31 , 32 ] , particularly, the Hamiltonian is arranged in the diagonal form (à la Bogoliubov), from which the canonical variable and commutation relations are straightforwardly read out, and the particle concept thus can be introduced. The equations of motion take the form of Heisenberg equations, which in turn gives the equations of the field operators in the similar form as the classical one. This is the standard logic of canonical quantization same as that in Minkowski spacetime, and from the diagonal Hamiltonian one can easily sees whether the Hamiltonian H is bounded from below and the vacuum state (ground state) can be defined. More important, as pointed by this paper in the following, if the vacuum can not be defined because the Hamiltonian spectrum is not bounded from below, the canonical framework is feasible for us to introduce the ghost particle concept (with the ’wrong’ commutation relations) to trivialize the negative energy spectrum and to obtain the minimum of the Hamiltonian for the definition of the vacuum. This ghost method does not change the mode functions corresponding to the ’physical’ (no-negative) spectrum, so from the Heisenberg equation one can still guarantee the equations of motion the form of the classical one. Furthermore, the ghost method can be applied to any boundary conditions such as those required to guarantee the Hamiltonian self-adjoint. This provides the freedom for any \xi value to further study various concrete problems which requires to specify the boundary conditions because of the lack of global hyperbolicity of AdS [ 26 , 28 , 27 , 32 ] . Based on the above quantization, the renormalization of the canonical energy momentum tensor operator T^{\mu\nu} is investigated. We take a BPHZ-like ’maximum’ subtraction scheme, to obtain T_{R}^{\mu\nu} finite and keep the background AdS geometry with the maximal symmetry for the quantum vacuum state. The renormalized energy momentum tensor automatically gives the correct subtraction term and the non-negative renormalized Hamiltonian with the help of the time-like Killing vector. The vacuum is stable since the (possible) classical instable modes are removed. Thus for any value of \xi a Hilbert space can be well defined by the physical particle Fock states and this provides the basics for the application of the EQFT, e.g., to calculate the correlation functions. The perturbation from the matter on the background geometry can also be described by particles states or their various coherent and thermal combinations, as input to Einstein equation via T^{\mu\nu}_{R} , to study the nonlinear (in)stability [ 13 , 14 , 15 , 23 ] . All the above investigations for real scalar field are contents of Section 2. In section 3, we study the quantization of the electromagnetic field and the gauge invariance plays the key rôle. It has already been recognized in Minkowski spacetime and with the Coulomb gauge that the independent degrees of freedom are not necessarily the correct canonical degrees of freedom which lead to the diagonalized Hamiltonian. Since the canonical T^{\mu\nu} and the corresponding Hamiltonian obtained via the help of the time-like Killing vector are both gauge invariant, we adopt the the expansion of the Maxwell filed in AdS given by the Refs. [ 33 , 34 , 35 ] and taking the temporal gauge. Two dimensions of dynamical degrees of freedom are canonically quantized. The vacuum of Maxwell field is stable since the gauge invariant Hamiltonian is definitely non-negative. The renormalization are similar as those of the scalar field. The last section, section 4, is for conclusions and discussions, which include some open questions of the renormalization and the consistent study of the backreaction from the EQFT source to the background geometry, and the interaction between fields. As is well known, though the algebraic quantum field theory takes the covariant form, the Hilbert space of states is not unique (not equivalent via a unitary transformation) under the general spacetime transformations. So we can only make use of the canonical quantization framework relying on the specific parameterization of the AdS as listed in the following. Needless to say, only by investigating all inequivalent Hilbert spaces of states, one may get a thorough understanding of the covariant quantum principles. In the presence of a negative cosmological constant \Lambda , the maximally symmetric solution of the vacuum Einstein equations R_{\mu\nu}-\frac{1}{2}Rg_{\mu\nu}+\Lambda g_{\mu\nu}=0 (1) in n+1 dimensions, is the AdS_{n+1} spacetime. The (covering) AdS metric employed in this paper (taking the example n=3 ) is g_{\mu\nu}=\mathrm{diag}\left\{-\left(1+l^{-2}r^{2}\right),\left(1+l^{-2}r^{2}\right)^{-1},r^{2}\gamma_{ij}\right\}. (2) i=1,2 is the coordinate chart on \mathbb{S}^{2} and \Lambda=-3l^{-2} with l^{-2} the AdS radius. The Lagrangian density of the real scalar field we study in this paper is \mathcal{L}=\sqrt{-g}\frac{1}{2}\left(-g^{\mu\nu}\partial_{\mu}\phi\partial_{\nu}\phi-m^{2}\phi^{2}-\xi R\phi^{2}\right), (3) with the coupling to the curvature taken into account; and that of the Maxwell field is \mathcal{L}=-\frac{1}{4}\sqrt{-g}F_{\mu\nu}F^{\mu\nu},~~~~~\>F_{\mu\nu}=\partial_{\mu}A_{\nu}-\partial_{\nu}A_{\mu}. (4) 2 Scalar Field Since the scalar curvature of AdS is a constant, the mass term in the Lagrangian density is just a shift of \xi with \xi\to\xi+\frac{m^{2}}{R} . Here we mainly consider the massless case, the mass can be trivially taken into account via the shift of \xi . To proceed the canonical quantization we employ the standard way of Legendre transformation to obtain the conjugate momentum of \phi , \Pi=\frac{\delta L}{\delta\dot{\phi}}=\frac{r^{2}\sin\theta}{1+l^{-2}r^{2}}\dot{\phi}, (5) with L the spacial integration of the Lagrangian density (Hereafter in this paper overdots or primes denote derivative with respect to t or r respectively), and the Hamiltonian \displaystyle H \displaystyle=\int drd\theta d\varphi\left(\Pi\dot{\phi}-\mathcal{L}\right) (6) \displaystyle=\int drd\theta d\varphi\frac{r^{2}\sin\theta}{2}\left(\frac{1}{1+l^{-2}r^{2}}\dot{\phi}^{2}+(1+l^{-2}r^{2})\phi^{\prime 2}+\frac{1}{r^{2}}\partial_{\theta}\phi\partial_{\theta}\phi+\frac{1}{r^{2}\sin^{2}\theta}\partial_{\varphi}\phi\partial_{\varphi}\phi+\xi R\phi^{2}\right) (7) \displaystyle=\frac{1}{2}\int drd\theta d\varphi\frac{r^{2}\sin\theta}{1+l^{-2}r^{2}}\left(\dot{\phi}^{2}-\ddot{\phi}\phi\right). (8) From Eq. ( 7 ) to Eq. ( 8 ), the equations of motion and the natural periodic boundary conditions for \theta and \varphi have been used, leaving a term corresponding to the boundary of the radial boundary \frac{1}{2}\int drd\theta d\varphi\partial_{r}\left(\sin\theta r^{2}(1+l^{-2}r^{2})\phi\phi^{\prime}\right), (9) which only depends on radial boundary conditions of \phi,\phi^{\prime} , this term will be investaged for various cases in the following. The Hamiltonian is not positive definite since the scalar curvature of AdS is negative. In Eq. ( 7 ), the last term is negative for \xi>0 , and all the other terms are positive. How the spectrum of the quantized Hamiltonian is dependent on the value of \xi , especially, the range of \xi in which the spectrum of the Hamiltonian is bounded from below and the vacuum is well-defined can not be straightforwardly read out from the above form of the Hamiltonian. For the answer we need to study the details of the quantization, and we will see the result is not \xi\leq 0 . To employ the canonical quantization formulations the key point is to find the proper canonical observables and the corresponding canonical commutation relations, with which the Hamiltonian is expressed and the spectrum is obtained. For this purpose, we should not à priori insert the canonical commutative relations on \phi and \Pi , but first to explore the mode functions on which the field can be expanded. From this expansion, and if the concrete form of the Hamiltonian can be diagonalized [ 36 ] , we can determine the canonical variables, the commutation relations, as well as the spectrum. This wisdom has been employed in the quantization of gauge theory in Minkowski spacetime (e.g., quantum electrodynamics in Coulomb gauge) and will also be applied in next section for Maxwell field with the independent degree of freedom singled out. A solution of the scalar field can be expressed as \phi_{\lambda jm}(t,r,\theta,\varphi)=\phi_{\lambda}(t)u_{\lambda j}(r)Y_{jm}(\theta,\varphi), (10) via the separation of variables, where Y_{jm}(\theta,\varphi) is the spherical harmonic function, as the result of the spatial symmetry. From the classcial equation of motion \displaystyle\ddot{\phi}_{\lambda}(t)=-\lambda\phi_{\lambda}(t), (11) \displaystyle A~u_{\lambda j}=\frac{\lambda r^{2}}{1+l^{-2}r^{2}}u_{\lambda j}, (12) with the operator A defined as [ 28 ] A~u_{\lambda j}\equiv-\left(r^{2}(1+l^{-2}r^{2})u_{\lambda j}^{\prime}\right)^{\prime}+j(j+1)u_{\lambda j}+\xi Rr^{2}u_{\lambda j}. (13) Hence for the real scalar field in AdS, the mode functions are completely determined when the solution of Eq. ( 12 ) are obtained. The spectrum of the Hamiltonian is also determined by the spectrum of the operator A , Eq. ( 13 ). Thanks to Wald [ 28 ] , the eigenfunctions and spectrum of the symmetric operator A have been completely investigated. The self-adjoint extension of A can be obtained corresponding to a certain radial boundary condition, which guarantees the dimensions of the deficiency subspaces n_{\pm} equals to each other, and u_{\lambda j} is orthonormal eigenfunctions of the self-adjoint operator A , i.e. \int_{0}^{\infty}\frac{u_{\omega j}u_{\omega^{\prime}j}r^{2}dr}{1+l^{-2}r^{2}}=\frac{\delta(\omega-\omega^{\prime})}{l}, (14) where u_{\lambda j}\equiv u_{\omega j} , with \omega\geq 0 for \lambda=\omega^{2}\geq 0 , and \omega<0 for \lambda=-\omega^{2}<0 . Since the second term of Eq. ( 8 ) will be infinite summation of terms as \lambda\phi_{\lambda}^{2} , the eigenvalue of the Hamiltonian can take any negative value and can not be bounded from below once \lambda can take negative value. Furthermore, for \lambda<0 , the time dependence of the mode function can be easily seen from the above Eq. ( 11 ), as the form of e^{at},\>a\in\mathbb{R} , which is called classical instability [ 37 ] . According to [ 28 ] , for \xi\leq\frac{9}{48} , A and H can be bounded from below via positive self-adjoint extensions. This will be discussed in details in sec. 2.2. For \xi>\frac{9}{48} , no positive self-adjoint extensions and the spectra of A and hence of H have no minimum value for the definition of the vacuum. However, in the framework of canonical quantization, for any \xi , the negative part of the spectrum of the Hamiltonian can be made trivial (an infinite constant) and subtracted via the renormalization scheme, leading to a spectrum with minimum. At the same time, the ’classical instability’ can be assigned into the complexity of the vacuum by the renormalization. 2.1 The Stability of Quantum Vacuum in AdS By expanding the scalar field with the above mode functions Eq. ( 10 ), we find certain regularity exists for the terms corresponding to \lambda<0 in the Hamiltonian, after diagonalized in the conventional Bogoliubov procedure. Applying the canonical anti-commutation relations to the coefficients of these negative part, one can trivialize these terms, independent from the boundary conditions of these mode functions or their concrete functional forms, only requiring that they are orthogonal to each other as well as to the mode functions corresponding to the \lambda\geq 0 spectrum. Since their ’wrong’ commutation relations, they are to be interpreted as ’ghosts’. We also demonstrate that these trivial part play trivial rôle in the energy-momentum tensor, that the trivial constant can be subtracted from the Hamiltonian as well as from the energy-momentum tensor. In such sense the ’ghost particles’ will not lead to any physical results, hence the quantum vacuum can be well defined and is stable. The real scalar field operator, which is Hermitian, can be expanded as, \begin{split}\phi(t,r,\theta,\varphi)=&\int_{0}^{+\infty}d\omega\sum_{j,m}N_{\omega}\left[a_{\omega jm}e^{-i\omega t}u_{\omega j}(r)Y_{jm}(\theta,\varphi)+a_{\omega jm}^{\dagger}e^{i\omega t}u_{\omega j}(r)Y_{jm}^{*}(\theta,\varphi)\right]\\ &+\int_{-\infty}^{0}d\omega\sum_{j,m}N_{\omega}\left[a_{\omega jm}e^{\omega t}u_{\omega j}(r)Y_{jm}(\theta,\varphi)+a_{\omega jm}^{\dagger}e^{\omega t}u_{\omega j}(r)Y_{jm}^{*}(\theta,\varphi)\right.\\ &\>\>\>\>\>\>\>\>\>\>\>\>\>\>\>\>\>\>\>\>+\left.a_{\omega jm}e^{-\omega t}u_{\omega j}(r)Y_{jm}(\theta,\varphi)+a_{\omega jm}^{\dagger}e^{-\omega t}u_{\omega j}(r)Y_{jm}^{*}(\theta,\varphi)\right],\\ \end{split} (15) where the normalized constant N_{\omega} can be real number through absorbing the phase factors into the coefficient operators a_{\omega jm} . Here we do not specify the concrete value of \xi , and for the case \xi\leq 5/48 , we can simply understand the above integral of the negative spectrum as zero. We also do not specify the boundary condition for u_{\omega j} . In some special boundary condition (see sec. 2.2) in the summation of \int_{-\infty}^{0}d\omega , some or all the u_{\omega j} could be zero. With this expansion of the field operator, the Hamiltonian can be expressed by the coefficient operators. The part corresponding to the spectrum \omega\geq 0 is straightforwardly diagonal. However, for the negative part, we have to introduce the linear transformation \begin{split}A_{\omega jm}&=\frac{1}{\sqrt{2}}a_{\omega jm}+\frac{1}{\sqrt{2}}a_{\omega j-m}^{\dagger}(-1)^{m},~\forall m\in[1,j];~A_{\omega j0}=\frac{1}{2}a_{\omega j0}+\frac{1}{2}a_{\omega j0}^{\dagger}\end{split} (16) With these A_{\omega jm} , and taking taking into account the fact A_{\omega jm}^{\dagger}=(-1)^{m}A_{\omega j-m} , We can obtain the whole Hamiltonian in the diagonal form \begin{split}H=&\int_{0}^{+\infty}d\omega\sum_{j,m}\frac{N_{\omega}^{2}\omega^{2}}{l}\left(a^{\dagger}_{\omega jm}a_{\omega jm}+a_{\omega jm}a^{\dagger}_{\omega jm}\right)-\int_{-\infty}^{0}d\omega\sum_{j,m}\frac{2N_{\omega}^{2}\omega^{2}}{l}\left(A_{\omega jm}A_{\omega jm}^{\dagger}+A_{\omega jm}^{\dagger}A_{\omega jm}\right)\\ =&\int_{0}^{+\infty}d\omega\sum_{j,m}\frac{\omega}{2}\left(a^{\dagger}_{\omega jm}a_{\omega jm}+a_{\omega jm}a^{\dagger}_{\omega jm}\right)\\ &+\int_{-\infty}^{0}d\omega\sum_{j=1}^{\infty}\sum_{m=0}^{j}2\omega\left(A_{\omega jm}A_{\omega jm}^{\dagger}+A_{\omega jm}^{\dagger}A_{\omega jm}\right),\end{split} (17) where the normalized constant is N_{\omega}=\sqrt{\frac{l}{2\left|\omega\right|}} The operators in the bracket is exactly in the same form as those in the Minkowski spacetime. As addressed in the Introduction we can read from this diagonal form of the Hamiltonian that the operators a and A can be taken as canonical operators. Since this is scalar field, we may introduce the standard canonical commutation relations. Unfortunately, this will lead to the first term (positive spectrum) in the Hamiltonian positive but the second one (negative spectrum) negative, and hence any positive or negative energy eigenvalues can be taken; thus the spectrum will not be bounded from below, the definition of vacuum lost. However, the operator form in the bracket of the second term (negative spectrum) of the Hamiltonian can be trivialized to be a constant by introducing the anti-commutation relations for the operators A ’s 1 1 1 Recall that the triviality is one of the well-known basic arguments why commutation relation (bosonic statistics) rather than anti-commutation relation should be introduced for any fields with integer spin in Minkowski spacetime. . The commutation relations for the positive spectrum, i.e., \forall\omega\geq 0 \left[a_{\omega jm},a^{\dagger}_{\omega^{\prime}j^{\prime}m^{\prime}}\right]=\delta(\omega-\omega^{\prime})\delta_{jj^{\prime}}\delta_{mm^{\prime}},\>\mathrm{others\>are\>}0, (18) and the anti-commutation relations for the negative spectrum, i.e., \forall\omega<0 \left\{A_{\omega jm},A^{\dagger}_{\omega^{\prime}j^{\prime}m^{\prime}}\right\}=\delta(\omega-\omega^{\prime})\delta_{jj^{\prime}}\delta_{mm^{\prime}}~(m,\>m^{\prime}\geq 1),\>\mathrm{others\>are\>}0, (19) together lead the Hamiltonian to be H=\int_{0}^{+\infty}d\omega\sum_{j,m}\frac{1}{2}\omega\left(a^{\dagger}_{\omega jm}a_{\omega jm}+a_{\omega jm}a^{\dagger}_{\omega jm}\right)+\int_{-\infty}^{0}d\omega\sum_{j=1}^{\infty}\sum_{m=1}^{j}2\omega\delta(0). (20) The negative part becomes a trivial constant indepedent from any operators (to be subtracted via renomalization in the following). Namely, a_{\omega jm}^{\dagger} or a_{\omega jm} is creation or annihilation operator for the physical bosonic particle and A_{\omega jm}^{\dagger} or A_{\omega jm} is creation or annihilation operator for ghost particle never appearing in the Hamiltonian. This suggests that scalar field with any non-minimal coupling (i.e., \forall\xi ) in AdS can be quantized as scalar particles. However, in the following we have to prove the consistency of the vacuum state definition, which includes the elimination of the infinities in the Hamiltonian as well as in the energy momentum tensor. We also have to show the triviality of the ghost modes. By the approvement of these facts, the ’classical unstable’ modes in fact is separated from the the stable modes by the above quantization procedures. Two different ’classical’ dynamics, coresponding to positive or negative spectrum part, thus can be investigated respectively in different state spaces. The linear property of the quantum mechanics framework, e.g., that the Hamiltonian is linear (Hermitian), plays the key rôle for the above approach. For this sake we write down here the anti-commutation relations of small a for \omega<0 which lead to those of A 2 2 2 Especially for m=0,\omega<0 A_{\omega j0}=A^{\dagger}_{\omega j0} , namely A_{\omega<0j0} are Hermitian, and A_{\omega j0}^{2}=0 due to anti-commutation relations, so A_{\omega j0}\equiv 0 . From Eq. ( 16 ), a_{\omega j0} is anti-Hermitian operators, a_{\omega j0}=-a_{\omega j0}^{\dagger} . So reads \left\{a_{\omega j0},a_{\omega^{\prime}j^{\prime}0}\right\}=-\left\{a_{\omega jm},a^{\dagger}_{\omega^{\prime}j^{\prime}0}\right\}=0 , so a_{\omega j0}\equiv 0 . Hence the ghost modes corresponding to m^{2}=0 are trivially zero. Eq. ( 13 ): \begin{split}&\left\{a_{\omega jm},a^{\dagger}_{\omega^{\prime}j^{\prime}m^{\prime}}\right\}=\delta(\omega-\omega^{\prime})\delta_{jj^{\prime}}\delta_{mm^{\prime}},\>\left\{a_{\omega jm},a_{\omega^{\prime}j^{\prime}m^{\prime}}\right\}=0\>\mathrm{for\>}m,m^{\prime}\in\mathbb{Z}/\left\{0\right\};\\ &\left\{a_{\omega j0},a_{\omega^{\prime}j^{\prime}0}\right\}=0,\>\left\{a_{\omega j0},a_{\omega^{\prime}j^{\prime}0}^{\dagger}\right\}=0,\>\mathrm{others\>are\>}0.\end{split} (21) From the Hamiltonian Eq. ( 20 ), it is obvious that the Hilbert space can be constructed by all the bosonic Fock states with non-negative \omega and arbitrary angular momentum quantum number (j,m) , whose creation and annihilation operators appear in the Hamiltonian Eq. ( 20 ). The ’bare’ vacuum state \left|0\right\rangle is defined as \begin{split}a_{\omega jm}\left|0\right\rangle&=0,~~\forall\omega\geq 0,j\geq 0\>,-j\leq m\leq+j,\\ A_{\omega jm}\left|0\right\rangle&=0,~~\forall\omega<0,j\geq 1,+1\leq m\leq+j.\end{split} (22) On the other hand, in the following we should study the ’physical’ vacuum states \left|\Psi\right\rangle=\sum_{i}c_{i}\left|\psi_{i}\right\rangle, which include any possible states relevant of ghost particles. Here \left|\psi_{i}\right\rangle=\prod_{\omega jm}(A^{\dagger}_{\omega jm})^{n^{i}_{\omega jm}}\left|0\right\rangle,\>n_{\omega jm}=0\>\mathrm{or}\>1. (23) The product is for any number of values of \omega<0 and j,m\geq 0 . If there is no restrictions between the angular quantum numbers with the energy, and when the minimum physical energy is zero, the vacuum, as the definition correponding to the minimal of Hamiltonian, could be degenerated. The degenerated states contain any number of zero energy physical bosons with arbitrary j and m . This should be the infrared problem as the case of electromagnetic field. Such degenerated vacuum can lead to anomaly of the conservation of some quantum number once the detailed balance is destroyed by nonequilibrium. We will not discuss such complexity here since such degenerated vacuum will not be affected by taking into account the ghost states, which is not different from other simple physical states. For the purpose to show that the fermionic ghost ’particle’ states, appearing in the expansion of the field operator, will not lead to any physical results, we calculate the expectation values of the energy momentum tensor for the vacuum states \left|\psi_{i}\right\rangle . For any linear operators, the conclusions can be straightforwardly applied to the linear combination of \left|\psi_{i}\right\rangle , \left|\Psi\right\rangle , so we will only study \left|\psi_{i}\right\rangle and omit the subscript i . In the following, the renormalization counterterm will be extracted, and the renormalization procedure eliminates the unphysical infinity and demonstrates the ghost particle states not making sense. The renormalized Hamiltonian is bounded from below so the vacuum is stable. In this way the consistent and regular EQFT model is set up. The canonical energy-momentum tensor can be obtained via the functional derivative of the action with respective to the metric tensor, taking into account the fact that the scalar curvature does not change under the differential homeomorphism mapping: T^{\mu\nu}=D^{\mu}\phi D^{\nu}\phi-\frac{1}{2}g^{\mu\nu}D^{\rho}\phi D_{\rho}\phi-\frac{1}{2}g^{\mu\nu}\xi R\phi^{2}. (24) By contracted with the time-like Killing vector, this energy momentum tensor gives the Hamiltonian Eq. ( 7 ). With the expansion of the scalar field Eq. ( 15 ), and (anti-)commutation relations obtained above, the vacuum expectation of the energy momentum tensor operator is calculated. As usual, divergences including those corresponding to the ’ghost mode’ appear. These should be subtracted via a reasonable renormalization scheme with the proper renormalization condition. Since the ’physical vacuum state’ corresponds to the AdS background, which is the solution of the homogenous Einstein equations, hence the vacuums satisfy the renormalized condition \left\langle\psi|T^{\mu\nu}_{R}|\psi\right\rangle=0, (25) from which the counterterms can be extracted. Once the \Lambda_{0} is renormalized to \Lambda in Eq. ( 1 ), the consistency of the vacuum is obtained. As to the cases of excited states, we leave the discussions to sec. 4. The details of the calculation taking into account the effects of any ghost states, can be found in the Appendix. The counterterms is obtained as \begin{split}&\Delta T^{\mu\nu}=\Delta T^{\mu\nu}_{1}+\Delta T^{\mu\nu}_{2}+\Delta T^{\mu\nu}_{3}\\ &\Delta T^{\mu\nu}_{1}=-\int_{0}^{\infty}d\omega\sum_{j,m}\frac{l}{2\omega}\left(g^{\nu\rho}g^{\mu\sigma}\partial_{\rho}\phi_{\omega jm}^{*}\partial_{\sigma}\phi_{\omega jm}\right)\\ &\Delta T^{\mu\nu}_{2}=g^{\mu\nu}\left[\int_{0}^{\infty}d\omega\sum_{j,m}\frac{l}{4\omega}\left(g^{\rho\sigma}\partial_{\rho}\phi^{*}_{\omega jm}\partial_{\sigma}\phi^{*}_{\omega jm}+\xi R\phi_{\omega jm}^{*}\phi_{\omega jm}\right)\right]\\ &\Delta T^{\mu\nu}_{3}=\int_{-\infty}^{0}\frac{l}{2\omega}\sum_{j=1}^{+\infty}\sum_{m=1}^{j}\left(g^{\mu\nu*\circ}_{\omega jm\omega jm}+g^{\mu\nu\circ*}_{\omega jm\omega jm}\right).\end{split} (26) The renormalization by introducing \Delta T_{1}^{\mu\nu} and \Delta T_{2}^{\mu\nu} is equivalent to taking normal ordering to the bosonic sector of T^{\mu\nu} . The last term \Delta T^{\mu\nu}_{3} ensures the cancellation of the unphysical ghost contributions, so that the renormalized energy-momentum tensor T^{\mu\nu}_{R} is independent of the ghost. The renormalized energy-momentum tensor is T^{\mu\nu}_{R}=T^{\mu\nu}+\Delta T^{\mu\nu}. (27) Since the relation of the unrenormalized Hamiltonian and the energy momentum tensor pointed above, the renormalized Hamiltonian can be obtained via H_{R}=\int drd\theta d\varphi\sqrt{-g}\xi_{\mu}T^{\mu 0}_{R}, (28) where \xi_{\mu}=(1+l^{-2}r^{2},0,0,0) is the timelike Killing vector in AdS. \begin{split}H_{R}&=\int_{0}^{+\infty}d\omega\sum_{j=0}^{+\infty}\sum_{m=-j}^{j}\frac{1}{2}\omega\left(a^{\dagger}_{\omega jm}a_{\omega jm}+a_{\omega jm}a^{\dagger}_{\omega jm}\right)+\int_{-\infty}^{0}d\omega\sum_{j=1}^{+\infty}\sum_{m=1}^{j}2\omega\delta(0)\\ &\>\>\>\>\>\>\>\>-\int drd\theta d\varphi r^{2}\sin\theta\left(1+l^{-2}r^{2}\right)\left\langle\psi|T^{00}|\psi\right\rangle.\end{split} (29) Here in the above equation we explicitly demonstrate in the second line the counterterms obtained from the energy momentum renormalization Eq. ( 26 ), which exactly cancels the divergences in the in first line, and lead to the renormalized Hamiltonian, H_{R}=\int_{0}^{+\infty}d\omega\sum_{j=0}^{+\infty}\sum_{m=-j}^{j}\omega a^{\dagger}_{\omega jm}a_{\omega jm}. (30) We can see that the renormalization condition, Eq. ( 25 ), has compeletly determined the renormalization Hamiltonian H_{R} . The Hamiltonian is different from the energy momentum tensor in that it is a integrated quantity and is defined for the whole space. We have ignored the term Eq. ( 9 ) in H . It will be investigated in details with concrete boundary conditions in next subsection. Any states sharing the same bosonic sector and differing only in the ghost sector are physically equivalent. Here the physical Hilbert space can be constructed with all the Fock states (particle states) as \prod_{i}\frac{\left(a_{\omega_{1}j_{1}m_{1}}^{\dagger}\right)^{n_{1}}}{\sqrt{n_{1}!}}\cdots\left|\psi\right\rangle. (31) It is easy to show that the expectation values of both H_{R} and T_{R}^{\mu\nu} on excited states are finite. As stressed in the introduction, the quantization is dependent on the specific coordinate, so taking expectation value of a tensor operator on the states Eq. ( 31 ) can destory its tensor property, but it is easy to show the covariant derivative D_{\mu}\left\langle T^{\mu\nu}_{R}\right\rangle=\partial_{\mu}\left\langle T^{\mu\nu}_{R}\right\rangle+\Gamma_{\mu\rho}^{\mu}\left\langle T^{\rho\nu}_{R}\right\rangle+\Gamma_{\mu\rho}^{\nu}\left\langle T^{\mu\rho}_{R}\right\rangle=0, namely the self-consistency is still kept, i.e. , D_{\mu}\left(G^{\mu\nu}+\Lambda g^{\mu\nu}\right)=D_{\mu}\left\langle T^{\mu\nu}_{R}\right\rangle=0. (32) 2.2 Quantization with Positive Self-adjoint Extension According to Wald [ 28 ] , in three cases corresponding to the range of \xi namely, \xi\leq 5/48,\>5/48<\xi<9/48,\>\xi=9/48 3 3 3 The massive cases could be obtained by the similar method just substituting \xi with \xi-\frac{m^{2}l^{2}}{12} . Similar conclusions could be obtained by replacing \xi\leq\frac{5}{48},\>\frac{5}{48}<\xi<\frac{9}{48},\xi=\frac{9}{48} by \xi\leq\frac{5}{48}+\frac{m^{2}l^{2}}{12},\>\frac{5}{48}+\frac{m^{2}l^{2}}{12}<\xi<\frac{9}{48}+\frac{m^{2}l^{2}}{12},\xi=\frac{9}{48}+\frac{m^{2}l^{2}}{12} respectively. Here we do not discuss the details of the meaning of m^{2} which is a free parameter taking real value. . the positive self-adjoint extensions of A exist. There is uniquely positive self-adjoint extensions for \xi\leq\frac{5}{48} and in fact in this case always \lambda>0 . There exists a family of parameters (boundary conditions corresponding to radial direction) that make A (and Hamiltonian) is bounded from below in \frac{5}{48}<\xi<\frac{9}{48} and \xi=\frac{9}{48} , respectively. 2.2.1 \xi\leq\frac{5}{48} The operator A of Eq. ( 13 ) has discrete eigenvalues \lambda_{nj}=l^{-2}\left(2n+j+\frac{3}{2}+\sqrt{\frac{9}{4}-12\xi}\right)^{2},\>n=0,1,2,\cdots (33) which guarantees that the u_{nj} ’s are square integrable. 4 4 4 See [ 28 ] for further detials about the explicit solutions. The relation between u_{nj}(r) here and \Phi_{\omega j}(x) in Wald’s paper [ 28 ] is u_{nj}=l^{-1/2}\frac{\Phi_{\omega j}}{r} . . The normalization condition of u_{\omega j} can be written as, \int_{0}^{\infty}u_{nj}u_{n^{\prime}j}\frac{r^{2}dr}{1+l^{-2}r^{2}}=\delta_{nn^{\prime}}.\> (34) The real scalar field is expanded by the orthonormal completeness set of the mode functions, \begin{split}\phi(t,r,\theta,\varphi)=\sum_{n,j,m}&\frac{1}{\sqrt{2}\lambda_{nj}^{1/4}}\left[a_{njm}e^{-i\sqrt{\lambda}t}u_{nj}(r)Y_{jm}(\theta,\varphi)+a_{njm}^{\dagger}e^{i\sqrt{\lambda}t}u_{nj}(r)Y_{jm}^{*}(\theta,\varphi)\right].\end{split} (35) The Hamiltonian Eq. ( 8 ) is H=\sum_{njm}\frac{\sqrt{\lambda}}{2}\left(a_{njm}a^{\dagger}_{njm}+a^{\dagger}_{njm}a_{njm}\right). (36) The term Eq. ( 9 ) are explicitly calculated to be 0 . We introduce the canonical commutation relations consistent with the diagonal Hamiltonian, [a_{njm},a^{\dagger}_{n^{\prime}j^{\prime}m^{\prime}}]=\delta_{n,n^{\prime}}\delta_{jj^{\prime}}\delta_{mm^{\prime}},\>others\>are\>zero, (37) i.e., a_{njm} or a^{\dagger}_{njm} is the canonical annihilation or creation operator, respectively. This means the positive but not too large \xi ( \leq\frac{5}{48} ) values do not change the non-negative property of the Hamiltonian though the term \xi R\phi^{2} introducing negative contributions. The \phi and \Pi are canonical variables 5 5 5 We could obtain the familiar result [\phi(\mathbf{x},t),\>\Pi(\mathbf{x}^{\prime},t)]=i\delta^{3}(\mathbf{x-x}^{\prime}) if the Cartesian coordinate is employed. , \left[\phi(r,\theta,\varphi,t),\Pi(r^{\prime},\theta^{\prime},\varphi^{\prime},t)\right]=i\frac{\delta^{3}(\mathbf{r-r}^{\prime})}{r^{2}\sin\theta}=i\delta(r-r^{\prime})\delta(\theta-\theta^{\prime})\delta(\varphi-\varphi^{\prime}). (38) Hence the standard Heisenberg equations lead to the quantum equation of motion exactly the similar form as the classical one. The quantum effective Lagrangian density share the similar form as the classical one. As usual, the unique vacuum could be defined by a_{njm}\left|0\right\rangle=0,\>\forall njm . The particle states can be unambiguously defined by the creation operators acting on vacuum state and it is the eigenstate of the Hamiltonian, which construct the Fock space. \left\langle\Psi|H|\Psi\right\rangle\geq 0 with state vector \left|\Psi\right\rangle in the Fock space and the equal sign for and only for the quantum vacuum state. Hence the quantum vacuum is stable and the quantum effective dynamics of the scalar fields described with the Lagrangian (Eq. ( 3 )) with \xi\leq 5/48 is well defined in the AdS spacetime. The similar renormalization scheme with the renormalized condition as Eq. ( 25 ) in sec. 2.1 can be performed. The counter term can be calculated to be \begin{split}&\Delta T^{\mu\nu}=\Delta T^{\mu\nu}_{1}+\Delta T^{\mu\nu}_{2}\\ &\Delta T^{\mu\nu}_{1}=-\sum_{n,j,m}\frac{1}{2\sqrt{\lambda}}\left(g^{\nu\rho}g^{\mu\sigma}\partial_{\rho}\phi_{njm}^{*}\partial_{\sigma}\phi_{njm}\right)\\ &\Delta T^{\mu\nu}_{2}=g^{\mu\nu}\left[\sum_{n,j,m}\frac{1}{2}\left(g^{\rho\sigma}\partial_{\rho}\phi^{*}_{njm}\partial_{\sigma}\phi^{*}_{njm}+\xi R\phi_{njm}^{*}\phi_{njm}\right)\right],\end{split} (39) and the renormalized energy-momentum tensor is T_{R}^{\mu\nu}=T^{\mu\nu}+\Delta T^{\mu\nu}=:T^{\mu\nu}:, (40) i.e., the divergences in the energy momentum tensor can be completely removed by taking the normal order of the operators, so does the renormalized Hamtonian H_{R}=\sum_{njm}l^{-1}\left(2n+j+\frac{3}{2}+\sqrt{\frac{9}{4}-12\xi}\right)a^{\dagger}_{njm}a_{njm}=:H:. (41) From the above, the free field theory in AdS can be made finite by taking the normal order of the operators in AdS in case of ghost(negative spectrum)-free, which is similar as in Minkowski spacetime. And the concept of particle is well defined in the whole AdS. The energy spectrum is discrete for \xi\leq\frac{5}{48} . There is an energy-gap between the vacuum and the one particle states with the smallest energy ( n=0,\>j=0,\>\forall m ). The backreaction perturbations initialized by the various particle states are not infinitesimal, which is different from the classical consideration. 2.2.2 \>\frac{5}{48}<\xi<\frac{9}{48} and \xi=9/48 Adopting the positive self-adjoint extensions of the operator A Eq. ( 13 ), corresponding to certain boundary conditions, the H has non-negative spectrum. Here we denote continuous \lambda=\omega^{2},\>\omega\geq 0 . Similar quantization and renormalization scheme as the cases in \xi\leq\frac{5}{48} can be performed. The counterterms of energy-momentum tensor are \begin{split}&\Delta T^{\mu\nu}=\Delta T^{\mu\nu}_{1}+\Delta T^{\mu\nu}_{2}\\ &\Delta T^{\mu\nu}_{1}=-\int_{0}^{\infty}d\omega\sum_{j,m}\frac{l}{2\omega}\left(g^{\nu\rho}g^{\mu\sigma}\partial_{\rho}\phi_{\omega jm}^{*}\partial_{\sigma}\phi_{\omega jm}\right)\\ &\Delta T^{\mu\nu}_{2}=+g^{\mu\nu}\left[\int_{0}^{\infty}d\omega\sum_{j,m}\frac{l}{4\omega}\left(g^{\rho\sigma}\partial_{\rho}\phi^{*}_{\omega jm}\partial_{\sigma}\phi^{*}_{\omega jm}+\xi R\phi_{njm}^{*}\phi_{njm}\right)\right].\end{split} (42) The renormalized energy-momentum tensor and Hamiltonian is same as Eq. ( 40 ) and Eq. ( 41 ) respectively, i.e., can be obtained by taking normal ordering of the operator products. The boundary term Eq. ( 9 ) is \begin{split}R^{2}(1+l^{-2}R^{2})\int\frac{ld\omega d\omega^{\prime}}{4\sqrt{\omega\omega^{\prime}}}\sum_{jm}&\left[a_{\omega jm}a_{\omega^{\prime}j-m}e^{-i(\omega+\omega^{\prime})t}u_{\omega j}(R)u_{\omega^{\prime}j}^{\prime}(R)(-1)^{m}\right.\\ &+a_{\omega jm}^{\dagger}a_{\omega^{\prime}j-m}^{\dagger}e^{i(\omega+\omega^{\prime})t}u_{\omega j}(R)u_{\omega^{\prime}j}^{\prime}(R)(-1)^{m}\\ &+a_{\omega jm}^{\dagger}a_{\omega^{\prime}jm}e^{i(\omega-\omega^{\prime})t}u_{\omega^{\prime}j}(R)u_{\omega j}^{\prime}(R)\\ &\left.+a_{\omega jm}^{\dagger}a_{\omega^{\prime}jm}e^{i(\omega-\omega^{\prime})t}u_{\omega^{\prime}j}(R)u_{\omega j}^{\prime}(R)\right],\end{split} (43) where R=+\infty , and u_{\omega jm}(R) and u^{\prime}_{\omega jm}(R) get value from the concrete radial boundary conditions of mode functions u_{\omega jm} . The expectation of vacuum \left|0\right\rangle of T_{R}^{\mu\nu} and H_{R} , with the latter including the term Eq. ( 9 ), still be 0 , under the renormalization scheme here. This implies that the vacuum of real scalar field under the boundary conditions for the positive self-adjoint extensions can be well defined in whole AdS, and that the Einstein equation in this vacuum state is homogeneous G^{\mu\nu}+\Lambda g^{\mu\nu}=\left\langle 0|T_{R}^{\mu\nu}|0\right\rangle=0 . For the excited states, the expectation values of T_{R}^{\mu\nu} are finite and covariantly conserved; those of the Hamiltonian is the summation of Eq. ( 8 ) and Eq. ( 9 ). The extra contribution on the radial boundary r=+\infty to the energy density is 0 , i.e. , \displaystyle\text{density of the extra energy}\sim\begin{Bmatrix}\frac{O(R^{2\nu-2})}{1-\ln Rl^{-1}+o(R^{-2}l^{2})},\>\frac{5}{48}<\xi<\frac{9}{48}\\ \frac{-\ln(1+l^{-2}R^{2})}{\sqrt{R^{2}+l^{2}}}\frac{(R^{2}-l^{2})/R^{3}}{-\frac{b_{0}}{a_{0}}\ln(1+l^{-2}R^{2})+1}\frac{1}{1-\ln Rl^{-1}+o(R^{-2}l^{2})},\>\xi=\frac{9}{48}\end{Bmatrix}\stackrel{{\scriptstyle R=+\infty}}{{=}}0, (44) where V=\int drd\theta d\varphi\frac{r^{2}\sin\theta}{\sqrt{1+l^{-2}r^{2}}} is the spatial volume of AdS, a_{0} and b_{0} and the concrete formula can be obtained by the concrete radial boundary condition [ 28 ] . The equation above implies that the spatial average density from the term Eq. ( 9 ) is 0 , irrespective of any physical states. This energy existing on the spatial boundary can be considered as the trivial energy background that unaffects the energy of physical particles moving in AdS, hence particle concept is well defined in whole AdS as well. 3 Maxwell (Abelian gauge) field The Lagrangian density of the Maxwell field in AdS Eq. ( 4 ) is only dependent on the 2-form of the field vector A_{\mu} , i.e., the gauge invariance under transformation A_{\mu}\to A_{\mu}+\partial_{\mu}\phi is still respected. So similar as in Minkowski spacetime, redundant degrees of freedom leads to disasters in naïve Legendre transformation, i.e., the conjugate momentum of the degree of freedom A_{0} is just 0, since \partial_{t}A_{0} does not appear in the Lagrangian density. At the same time, via naïve Legendre transformation \begin{split}&\int drd\theta d\varphi\sqrt{-g}\left(\frac{\partial\mathcal{L}}{\partial\dot{A}_{i}}\dot{A}_{i}-\mathcal{L}\right)\\ &=\int drd\theta d\varphi\sqrt{-g}\left(\frac{1}{2}g^{ij}g^{00}\partial_{i}A_{0}\partial_{j}A_{0}-\frac{1}{2}g^{ij}g^{00}\partial_{0}A_{i}\partial_{0}A_{j}+\frac{1}{4}F_{ij}F_{i^{\prime}j^{\prime}}g^{ii^{\prime}}g^{jj^{\prime}}\right),\end{split} (45) we get a ‘Hamiltonian‘, which is neither gauge invariant nor positive definite (hence losing the definition of the vacuum). However, by fixing a gauge these problems are easy to be solved. Adopting the temporal gauge, A_{0}=0 , the independent degrees of freedom are A_{i} , and the correct Hamiltonian can be obtained, which equals to the one obtained from the gauge-invariant canonical momentum-energy tensor T^{\mu\nu}=F_{\lambda}^{\>\>\>\mu}F^{\lambda\nu}-\frac{1}{4}g^{\mu\nu}F_{\rho\sigma}F^{\rho\sigma}, (46) with the help of of the timelike Killing vector: H=\int drd\theta d\varphi\sqrt{-g}\xi_{\mu}T^{\mu 0}=\int drd\theta d\varphi\sqrt{-g}\left[-\frac{1}{2}g^{00}g^{ij}F_{i0}F_{j0}+\frac{1}{4}g^{ii^{\prime}}g^{jj^{\prime}}F_{ij}F_{i^{\prime}j^{\prime}}\right]. (47) It is gauge invariant and non-negative. In other words, the canonical quantization can be simply based on the Hamiltonian H in Eq. ( 47 ), with the practical way to employ the field vector A_{\mu} with the temporal gauge. It is easy to check that, in the AdS, the first order differential equation \partial_{t}\phi=-A_{0} has regular solution. To single out the proper dynamical degree of freedom in AdS, we adopt the expansion of the Maxwell field as [ 33 , 34 , 35 ] , and taking the A_{0}=0 gauge, A_{\mu}=\sum_{j,m}\begin{pmatrix}0\\ 0\\ \frac{a^{jm}(r,t)}{\sin\theta}\frac{\partial Y_{jm}}{\partial\varphi}\\ -a^{jm}(r,t)\sin\theta\frac{\partial Y_{jm}}{\partial\theta}\\ \end{pmatrix}+\begin{pmatrix}0\\ h^{jm}(r,t)Y_{jm}\\ k^{jm}(r,t)\frac{\partial Y_{jm}}{\partial\theta}\\ k^{jm}(r,t)\frac{\partial Y_{jm}}{\partial\varphi}\end{pmatrix}, (48) where the first column being of parity (-1)^{j} and the second (-1)^{j+1} . The Hamiltonian leads to the equations \displaystyle(g^{rr}a^{jm\>\prime})^{\prime}+g^{00}\frac{\partial^{2}a^{jm}}{\partial t^{2}}-\frac{j(j+1)}{r^{2}}a^{jm}=0 (49) \displaystyle g^{rr}b^{jm\>\prime}=\partial_{t}k^{jm} (50) \displaystyle g^{00}\partial_{t}b^{jm}=h^{jm}-k^{jm\>\prime} (51) where \frac{j(j+1)b^{jm}}{r^{2}}:=\partial_{t}h^{jm}. (52) The equation of b^{jm} is the same as a^{jm} , which can be obtained by combing Eq. ( 50 ) and Eq. ( 51 ). By separating varibles, a^{\lambda jm}(r,t)=a^{\lambda jm}(r)a^{\lambda jm}(t),\>b^{\lambda jm}(r,t)=b^{\lambda jm}(r)b^{\lambda jm}(t) , \displaystyle\frac{d^{2}a^{\lambda jm}(t)}{dt^{2}}=-\lambda a^{\lambda jm}(t) (53) \displaystyle(g^{rr}a^{\lambda jm\prime}(r))^{\prime}-\lambda g^{00}a^{\lambda jm}(r)-\frac{j(j+1)a^{\omega jm}(r)}{r^{2}}=0. (54) The same as for b^{\lambda jm}(r),\>b^{\lambda jm}(t) . Both the time-dependence of b^{\lambda jm}(t) and a^{\lambda jm}(t) are e^{\pm i\sqrt{\lambda}t} . \lambda is non-negetive since the Hamiltonain is positive definite here. b^{\lambda jm}(r) and a^{\lambda jm}(r) satisfy the hyperspherical differential equation which have been studied (see [ 38 ] ). Utilizing the results above, we can obtain h^{\omega jm}(r,t) and k^{\lambda jm}(r,t) respectively, \begin{split}&h^{\lambda jm}(r,t)=\frac{j(j+1)b^{\lambda jm}(r,t)}{-i\sqrt{\lambda}r^{2}}+g^{\lambda jm\prime}(r)\\ &k^{\lambda jm}(r,t)=\frac{(1+l^{-2}r^{2})b^{\lambda jm\>\prime}(r,t)}{-i\sqrt{\lambda}}+g^{\lambda jm}(r).\end{split} (55) g^{\lambda jm}(r) represents the background degrees of freedom and satisfies the equation, j(j+1)g^{\lambda jm}(r)=g^{rr}(r^{2}g^{\lambda jm\prime}(r))^{\prime}, (56) which does not propagate and is not quantized. As a matter of fact, the contribution of g^{\lambda jm} to A_{\mu} is A_{\mu}^{background}(r,\theta,\varphi)=\int d\lambda\sum_{jm}c_{\lambda jm}\partial_{\mu}(g^{\lambda jm}Y_{jm})+h.c.. (57) Hence, it does not contribute to F_{\mu\nu} neither to the Hamiltonian. Hence, the Hamiltonian can be obtained from Eq. ( 47 ), H=\int_{0}^{\infty}d\lambda\sum_{jm,i=a,b}\left|N_{\lambda}\right|^{2}\lambda j(j+1)\left(c_{i}^{\lambda jm}c_{i}^{\lambda jm\dagger}+c_{i}^{\lambda jm\dagger}c_{i}^{\lambda jm}\right), (58) where N_{\lambda} is normalization factor to guarantee the creation and annihilation operators satisfy the canonical form. The dynamical degrees of freedom of Maxwell field in AdS 4 are a^{\lambda jm},b^{\lambda jm} with the parity (-1)^{j} and (-1)^{j+1} respectively. The renormalization condition is \left\langle 0|T^{\mu\nu}_{R}|0\right\rangle=0 , which is similar as the scalar case, Eq. ( 25 ), and the counterterm is, \begin{split}\Delta T^{\mu\nu}=-\left\langle 0|T^{\mu\nu}|0\right\rangle.\end{split} (59) H is positive definite, hence ghost sector is not needed, so the counterterm is equivalent to taking normal ordering. The renormalized Hamiltonian is obtained from Eq. ( 59 ) with the help of Killing vector, H_{R}=\int_{0}^{\infty}d\lambda\sum_{jm,i=a,b}2\left|N_{\lambda}\right|^{2}\lambda j(j+1)c_{i}^{\lambda jm\dagger}c_{i}^{\lambda jm}=:H:, (60) which eliminates the infinity vacuum energy in Eq. ( 58 ). 4 Conclusions and Discussions The AdS spacetime is the maximally symmetric solution of the vacuum Einstein equations with a negative cosmological constant. In this paper, we have studied the canonical quantization of two EQFT models, non-minimally coupled real scalar field and Maxwell field, in (covering) AdS spacetime with the concrete parameterization Eq. ( 2 ). These two EQFT dynamics both have the stable vacuum state consistent with the background AdS geometry, and the global Hilbert space defined by the physical particle Fock states. To get these results, two points are crucial: 1) The symmetric radial operator A Eq. ( 13 ) is self-adjoint under proper radial boundary conditions for any \xi value in Eq. ( 3 ). The orthonormal mode functions Eq. ( 10 ) can be used to expand the field and the Hamiltonian is diagonalized. So the canonical variable is read out and the particle concept can be introduced with the proper commutation relations. Particularly for the case of the Hamiltonian with negative spectrum, the diagonalized negative part of the Hamiltonian can be trivialized by introducing the ’wrong’ anti-commutation relations which can be explained as ’ghost’. 2) After the above quantization, the energy-momentum tensor is renormalized to remove the divergences in the boson sector as well as the ghost sector via the BPHZ-like way. The possible classical instability [ 37 ] only corresponding to the negative spectrum is removed incidentally. Counterterms are extracted employing the renormalization condition \left\langle\psi|T^{\mu\nu}_{R}|\psi\right\rangle=0 . For the positive spectrum, the effects of all the counter terms is equivalent to normal ordering. Extra counterterms only arise when the ghost states exist. From this finite T^{\mu\nu}_{R} , contracted with the time-like Killing vector, one gets the renormalized Hamiltonian bounded from below, and the global stable vacuums are obtained. The whole energy ( H_{R} and the term Eq. ( 9 )) of all matter fields is 0 in the vacuum state, rendering the Einstein equations homogeneous. Once the constant of the Lagrangian density is renormalized as \Lambda_{R}=\Lambda_{0}-\delta\Lambda=\Lambda , to get \Lambda g_{\mu\nu} term in Einstein equations, the AdS (background geometry) is the solution of the classical homogeneous Einstein Equations. The self-consistent Effective dynamics is thus set up. In the above sections we have demonstrated that the ghost sector does not affect the physical observables, e.g., the renormalized energy-momentum tensor, in free field theory. In the case of taking into account any interaction, the conclusion still holds via a simple argument in perturbation framework, where matrix elements of observables on physical states reduce to those of free field operators. The key point is that the physical states can always be treated as the vacuum state for any ghost modes. In this way any ghost effects introduced by the ghost part of the field operators in the interaction terms appear only as vacuum-vacuum transitions which factor out uniformly from all matrix elements. Particularly, the S-matrix defined in the Hilbert space of all the physical states keeps unitary. The quantization procedure of our paper aligns with the present works such as [ 26 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 37 , 46 , 47 , 24 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 ] with positive self-ajoint extensions of A under the proper boundary conditions. The renormalization on the energy momentum tensor has already been intensively investigated in literature [ 24 , 26 , 39 , 40 , 45 , 48 , 49 , 50 , 51 , 37 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67 , 68 , 69 , 70 , 71 , 72 , 73 , 74 , 75 , 76 , 77 , 78 , 79 , 80 , 81 ] . For general spacetime, due to the Hilbert space and vacuum not uniquely determined, one have difficulty to specify the concrete physical renormalization condition. This shares some similarities as the case lack of asymptotic state in Minkowski spacetime for dynamics with asymptotic freedom. Here for the specified background AdS spacetime, we can fix the renormalization condition e.g., Eq. ( 25 ), (see also [ 24 , 25 , 82 ] ). In the above we have shown the renormalization condition Eq. ( 25 ) removes the divergences not only for vacuum state, but also for any excited states. Vacuum keeps as the self-consistent maximally symmetric solution of the homogenous Einstein equations and the covariant conservation Eq. ( 32 ) keeps. These are not dependent on any boundary condition, which is the consequence of the quantization we employed here. The expectation values of T^{\mu\nu}_{R} on thermal and excited states can break the maximal symmetry, which leads modifications to background geometry. One of the most used scheme to study the backreaction is reparameterization of AdS metric on which quantization of matter fields is performed, then one can get the solution of the metric and the configurations of fields [ 83 , 24 , 82 , 84 ] . The other way is to study the evolution of Einstein-scalar system by parameterizing the spacetime and specifying initial data as the anti-de Sitter spacetime along with the expectation value of T_{R}^{\mu\nu} [ 13 ] . Perhaps the simplest way is to take into account the minimal quantum effects of EQFT by employing coherent states. This way is possible only because the global Hilbert space is well defined in the above. So the the expectation value of the field operator on coherent states can be used as the ’classical’ field to calculate \left\langle T_{R}^{\mu\nu}\right\rangle_{\text{coh}} to be the inhomogeneous term of Einstein equations. No matter what ways to study the backreaction, the cases \xi\leq\frac{5}{48} need more careful investigations because of the discrete spectrum. Appendix: The Calculation of \left\langle\psi\right|T^{\mu\nu}\left|\psi\right\rangle We show some details in calculation of the expactation of T^{\mu\nu} here. Applying the expansion of \phi Eq. ( 15 ) to Eq. ( 24 ), we could obtain \begin{split}&T^{\mu\nu}=\int_{0}^{+\infty}\int_{0}^{+\infty}\frac{d\omega d\omega^{\prime}l}{\sqrt{2\omega 2\omega^{\prime}}}\sum_{jj^{\prime}mm^{\prime}}\left[\left(f^{\mu\nu(0\circ)(0\circ)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}a_{\omega^{\prime}j^{\prime}m^{\prime}}+\left(f^{\mu\nu(0*)(0*)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}^{\dagger}a_{\omega^{\prime}j^{\prime}m^{\prime}}^{\dagger}\right]\\ &\>\>\>\>\>\>\>\>+\int_{0}^{+\infty}\int_{0}^{+\infty}\frac{d\omega d\omega^{\prime}l}{\sqrt{2\omega 2\omega^{\prime}}}\sum_{jj^{\prime}mm^{\prime}}\left[\left(f^{\mu\nu(0*)(0\circ)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}^{\dagger}a_{\omega^{\prime}j^{\prime}m^{\prime}}+\left(f^{\mu\nu(0\circ)(0*)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}a_{\omega^{\prime}j^{\prime}m^{\prime}}^{\dagger}\right]\\ &\>\>\>\>\>\>\>\>+\int_{-\infty}^{0}\int_{-\infty}^{0}\frac{d\omega d\omega^{\prime}l}{\sqrt{2\omega 2\omega^{\prime}}}\sum_{jj^{\prime}mm^{\prime}}^{m,m^{\prime}\neq 0}\left[\left(g^{\mu\nu\circ\circ}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}a_{\omega^{\prime}j^{\prime}m^{\prime}}+\left(g^{\mu\nu**}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}^{\dagger}a_{\omega^{\prime}j^{\prime}m^{\prime}}^{\dagger}\right]\\ &\>\>\>\>\>\>\>\>+\int_{-\infty}^{0}\int_{-\infty}^{0}\frac{d\omega d\omega^{\prime}l}{\sqrt{2\omega 2\omega^{\prime}}}\sum_{jj^{\prime}mm^{\prime}}^{m,m^{\prime}\neq 0}\left[\left(g^{\mu\nu*\circ}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}^{\dagger}a_{\omega^{\prime}j^{\prime}m^{\prime}}+\left(g^{\mu\nu\circ*}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\right)a_{\omega jm}a_{\omega^{\prime}j^{\prime}m^{\prime}}^{\dagger}\right]\\ &\>\>\>\>\>\>\>\>+\mathrm{mixing\>terms},\end{split} (61) where the mixing terms include the mixing between positive and negative frequency terms. The definition of f^{\mu\nu(i\circ)(k\circ)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}} is 6 6 6 For example, f^{\mu\nu(0*)(0\circ)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}=g^{\mu\rho}g^{\nu\sigma}\partial_{\rho}\phi_{\omega jm}^{0*}\partial_{\sigma}\phi_{\omega^{\prime}j^{\prime}m^{\prime}}^{0}-\frac{g^{\mu\nu}}{2}\left(g^{\rho\sigma}\partial_{\rho}\phi_{\omega jm}^{0*}\partial_{\sigma}\phi_{\omega^{\prime}j^{\prime}m^{\prime}}^{0}+\xi R\phi_{\omega jm}^{0*}\phi_{\omega^{\prime}j^{\prime}m^{\prime}}^{0}\right). g^{\mu\rho}g^{\nu\sigma}\partial_{\rho}\phi_{\omega jm}^{i}\partial_{\sigma}\phi_{\omega^{\prime}j^{\prime}m^{\prime}}^{k}-\frac{g^{\mu\nu}}{2}\left(g^{\rho\sigma}\partial_{\rho}\phi_{\omega jm}^{i}\partial_{\sigma}\phi_{\omega^{\prime}j^{\prime}m^{\prime}}^{k}+\xi R\phi_{\omega jm}^{i}\phi_{\omega^{\prime}j^{\prime}m^{\prime}}^{k}\right), where \phi^{i}_{\omega jm}(x)\left\{\begin{matrix}e^{-i\omega t}u_{\omega j}(r)Y_{jm}(\theta,\varphi)&i=0\\ e^{\omega t}u_{\omega j}(r)Y_{jm}(\theta,\varphi)&i=1\\ e^{-\omega t}u_{\omega j}(r)Y_{jm}(\theta,\varphi)&i=2,\end{matrix}\right. and g^{\mu\nu\circ\circ}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}} is g^{\mu\nu\circ\circ}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}=\sum_{ii^{\prime}=1,2}f^{\mu\nu(i\circ)(i^{\prime}\circ)}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}. It suffices to consider the example \left|\psi\right\rangle=\left|\underbrace{0}_{\text{Boson Sector}};\underbrace{0^{\omega_{1}<0,j_{1},m_{1}\geq 1},1^{\omega_{2}<0,j_{2},m_{2}\geq 1},0^{\omega_{3}<0,j_{3},m_{3}\geq 1},1^{\omega_{4}<0,j_{4},m_{4}\geq 1},\cdots}_{\text{Ghost Sector}}\right\rangle, (62) where ellipsis refers to possible ghost modes containing one and zero ghost particle. The expectation of the mixing terms for Eq. ( 62 ) apparently vanishes. Then the expectation of the first four terms of Eq. ( 61 ) is \int_{0}^{+\infty}d\omega\frac{l}{2\omega}\sum_{j=0}^{+\infty}\sum_{m=-j}^{j}f^{\mu\nu(0*)(0\circ)}_{\omega jm\omega jm}. The expectation of fifth term, \int_{-\infty}^{0}\int_{-\infty}^{0}\frac{d\omega d\omega^{\prime}l}{\sqrt{2\omega 2\omega^{\prime}}}\sum_{j=1}^{+\infty}\sum_{j^{\prime}=1}^{+\infty}\sum_{m\neq 0}\sum_{m^{\prime}\neq 0}\left(g^{\mu\nu\circ\circ}_{\omega jm\omega^{\prime}j^{\prime}m^{\prime}}\left\langle\psi\right|a_{\omega jm}a_{\omega^{\prime}j^{\prime}m^{\prime}}\left|\psi\right\rangle\right), can be simplified into \begin{split}\int_{-\infty}^{0}\frac{d\omega l}{-2\omega}\sum_{j=1}^{+\infty}&\left(\sum_{m=1,m\neq m_{2},m_{4},\cdots}^{j}g^{\mu\nu\circ*}_{\omega jm\omega jm}(-1)^{m}\left\langle\psi\right|a_{\omega jm}a_{\omega j-m}\left|\psi\right\rangle\right.\\ &\left.+\sum_{m=1,m\neq m_{2},m_{4},\cdots}^{j}g^{\mu\nu*\circ}_{\omega jm\omega jm}(-1)^{m}\left\langle\psi\right|a_{\omega j-m}a_{\omega jm}\left|\psi\right\rangle\right),\end{split} where ellipsis corresponds to the possible ghost particles, since the integrand vanishes unless m=-m^{\prime} and m,m^{\prime}\neq m_{2},m_{4},\cdots are satisfied. Hence, we could obtain that \text{the fifth term}=\int_{-\infty}^{0}\sum_{j=1}^{+\infty}\sum_{m=1,\neq m_{2},m_{4},\cdots}^{j}\frac{1}{2}\left(g^{\mu\nu\circ*}_{\omega jm\omega jm}-g^{\mu\nu*\circ}_{\omega jm\omega jm}\right), due to the facts \left\langle 0_{\omega<0,j,m}\right|a_{\omega jm}a_{\omega j-m}\left|0_{\omega<0,j,m}\right\rangle=\frac{1}{2}(-1)^{m},\>\left\langle 0_{\omega<0,j,m}\right|a_{\omega j-m}a_{\omega jm}\left|0_{\omega<0,j,m}\right\rangle=-\frac{1}{2}(-1)^{m}. Hence, the summation of the fifth and sixth term of Eq. ( 61 ) is 0 because the fifth term of Eq. ( 61 ) is pure imaginary number. Similarly, using the expectation values \begin{split}&\left\langle 0_{\omega<0,j,m}\right|a_{\omega jm}^{\dagger}a_{\omega jm}\left|0_{\omega<0,j,m}\right\rangle=\frac{1}{2},\left\langle 0_{\omega<0,j,m}\right|a_{\omega j-m}^{\dagger}a_{\omega j-m}\left|0_{\omega<0,j,m}\right\rangle=\frac{1}{2},\\ &\left\langle 1_{\omega<0,j,m}\right|a_{\omega jm}^{\dagger}a_{\omega jm}\left|1_{\omega<0,j,m}\right\rangle=1,\left\langle 1_{\omega<0,j,m}\right|a_{\omega j-m}^{\dagger}a_{\omega j-m}\left|1_{\omega<0,j,m}\right\rangle=0,\end{split} we obtain the expectation of seventh term of Eq. ( 61 ) \begin{split}\int_{-\infty}^{0}\frac{d\omega l}{-2\omega}\sum_{j=1}^{+\infty}\left(\frac{1}{2}\sum_{m=1,m\neq m_{2},m_{4},\cdots}^{j}\left(g^{\mu\nu\circ*}_{\omega jm\omega jm}+g^{\mu\nu*\circ}_{\omega jm\omega jm}\right)+g^{\mu\nu\circ*}_{\omega jm_{2}\omega jm_{2}}+g^{\mu\nu\circ*}_{\omega jm_{4}\omega jm_{4}}+\cdots\right),\end{split} where ellipses denote to the possible ghost particles. In addition, the expectation of eighth term equals to the conjugate of the equation above. As a consequence, the expectation of energy-momentum tensor Eq. ( 61 ) in vacuum state \left|\psi\right\rangle is obtained by summing the contributions from first to eighth term, namely \left\langle\psi|T^{\mu\nu}|\psi\right\rangle=\int_{0}^{+\infty}d\omega\frac{l}{2\omega}\sum_{j=0}^{+\infty}\sum_{m=-j}^{j}f^{\mu\nu(0*)(0\circ)}_{\omega jm\omega jm}+\int_{-\infty}^{0}\frac{l}{-\omega}\sum_{j=1}^{+\infty}\sum_{m=1}^{j}\left(g^{\mu\nu*\circ}_{\omega jm\omega jm}+g^{\mu\nu\circ*}_{\omega jm\omega jm}\right), (63) where \left|\psi\right\rangle is involved with the boson vacuum and arbitrary mutiple ghost particles.