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

Exponentially-growing Mode Instability on the Reissner–Nordström-Anti-de-Sitter black holes

Weihao Zheng [email protected] Department of Mathematics, Rutgers University, Hill Center, 110 Frelinghuysen Road, Piscataway, NJ, USA
Abstract

We construct growing mode solutions to the uncharged and charged Klein–Gordon equations on the sub-extremal Reissner–Nordström-anti-de-Sitter (AdS) spacetime under reflecting (Dirichlet or Neumann) boundary conditions. Our result applies to a range of Klein–Gordon masses above the so-called Breitenlohner–Freedman bound, notably including the conformal mass case. The mode instability of the Reissner–Nordström-AdS spacetime for some black hole parameters is in sharp contrast to the Schwarzschild-AdS spacetime, where the solution to the Klein–Gordon equation is known to decay in time. Contrary to other mode instability results on the Kerr and Kerr-AdS spacetimes, our growing mode solutions of the uncharged and weakly charged Klein–Gordon equation exist independently of the occurrence or absence of superradiance. We discover a novel mechanism leading to a growing mode solution, namely, a near-extremal instability for the Klein–Gordon equation. Our result seems to be the first rigorous mathematical realization of this instability.

1 Introduction

In this paper, we are interested in constructing growing mode solutions to the Klein–Gordon equation with parameters (M,e,Λ,q0,α)(M,e,\Lambda,q_{0},\alpha) on a sub-extremal Reissner–Nordström-anti-de-Sitter background:

gRNμνDμDνϕ=(Λ3)αϕ,Dμ=μ+iq0Aμ,g_{RN}^{\mu\nu}D_{\mu}D_{\nu}\phi=\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha\phi,\quad D_{\mu}=\nabla_{\mu}+iq_{0}A_{\mu}, (1.1)

where AA is the electromagnetic potential and the Reissner–Nordström-AdS metric takes the form of

gRN=Ω2dt2+1Ω2dr2+r2dσ2,Ω2=12Mr+e2r2+(Λ3)r2.g_{RN}=-\Omega^{2}dt^{2}+\frac{1}{\Omega^{2}}dr^{2}+r^{2}d\sigma^{2},\quad\ \Omega^{2}=1-\frac{2M}{r}+\frac{e^{2}}{r^{2}}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r^{2}. (1.2)

M>0M>0 is the mass of the black hole, ee\in\mathbb{R} is the charge of the black hole, Λ<0\Lambda<0 is the cosmological constant, q0q_{0}\in\mathbb{R} is the charge of the scalar field, and α<0\alpha<0 is the negative Klein–Gordon mass. We will consider the charged scalar field case q00q_{0}\neq 0 and the uncharged scalar field case q0=0q_{0}=0. We will assume the Reissner–Nordström-AdS spacetime is sub-extremal, i.e. Ω2\Omega^{2} admits two positive roots. Let r+:=r+(M,e,Λ)r_{+}:=r_{+}(M,e,\Lambda) be the largest root of Ω2\Omega^{2} corresponding to the area radius of the event horizon.

We will make a gauge choice for the electromagnetic potential AA of the following form

A=e(1r+1r)dt.A=-e\left(\frac{1}{r_{+}}-\frac{1}{r}\right)dt. (1.3)

In the case q0=0q_{0}=0, equation (1.1)\eqref{Klein-Gordon} is reduced to the well-known uncharged Klein–Gordon equation with the usual covariant derivative Dμ=μD_{\mu}=\nabla_{\mu}.

Due to the lack of global hyperbolicity of the asymptotically AdS spacetime, the natural formulation of the Klein–Gordon equation (1.1)\eqref{Klein-Gordon} is the initial-boundary value problem; see [37, 35]. We also refer to Section 3.1 for a detailed introduction to the geometry and boundary conditions of the asymptotically AdS spacetime. The growing mode solutions we consider in this paper take the form of ϕ(t,r)=eiωtψ(r)\phi(t,r)=e^{i\omega t}\psi(r), where ω\omega\in\mathbb{C} has a negative imaginary part and ϕ\phi is a function regular at the event horizon.

In view of the fact that (1.2)\eqref{RN metric} can be parametrized by (M,e,Λ)(M,e,\Lambda) or (M,r+,Λ)(M,r_{+},\Lambda) alternatively, we will call (M,r+,Λ)(M,r_{+},\Lambda) the sub-extremal parameters of (1.2)\eqref{RN metric} if the corresponding (M,e,Λ)(M,e,\Lambda) are sub-extremal parameters. Then under these new parameters, e=0e=0 corresponds to M=Me=0:=r+2(1+(Λ3)r+2)M=M_{e=0}:=\frac{r_{+}}{2}\left(1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}\right) while the extremality corresponds to M=M0:=r++2(Λ3)r+3M=M_{0}:=r_{+}+2\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{3}. For fixed (r+,Λ)(r_{+},\Lambda), the admissible sub-extremal range of MM is Me=0M<M0M_{e=0}\leq M<M_{0}. We refer to Section 3.2 for a detailed discussion of the parameters transform.

In the following, we will call (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) parameters of the Klein–Gordon equation (1.1)\eqref{Klein-Gordon}. Our main results are as follows:

Theorem 1.1.

[Rough version of the main result] For the Klein–Gordon equation (1.1)\eqref{Klein-Gordon}, let CDN=0C_{DN}=0 for Dirichlet boundary conditions and CDN=54C_{DN}=-\frac{5}{4} for Neumann boundary conditions respectively. Imposing reflecting boundary condition (Dirichlet or Neumann) for (1.1), we have the following three results about growing mode solutions:

  1. (1)

    (Large charge case) Assume (Mb,r+,Λ)(M_{b},r_{+},\Lambda) are given sub-extremal parameters with Me=0<Mb<M0M_{e=0}<M_{b}<M_{0}, 94<α<CDN-\frac{9}{4}<\alpha<C_{DN} is fixed, and |q0||q_{0}| is a fixed large charge (the size of |q0||q_{0}| depends on (Mb,r+,Λ)(M_{b},r_{+},\Lambda) and α\alpha). Let 𝒮(Mb,r+,Λ,α,q0)\mathcal{S}(M_{b},r_{+},\Lambda,\alpha,q_{0}) be the set of all Me=0M<MbM_{e=0}\leq M<M_{b} such that (M,r+,Λ)(M,r_{+},\Lambda) are admissible sub-extremal parameters and (1.1)\eqref{Klein-Gordon} with parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) has a growing mode solution that is regular at the event horizon. Then 𝒮(Mb,r+,Λ,α,q0)\mathcal{S}(M_{b},r_{+},\Lambda,\alpha,q_{0}) is non-empty and open.

  2. (2)

    (General fixed charge case) For any fixed (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}) satisfying the conditions

    94<α<min{CDN,32+q022(Λ3)},\displaystyle-\frac{9}{4}<\alpha<\min\left\{C_{DN},-\frac{3}{2}+\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right\}, (1.4)
    (Λ3)r+2>R0(α,q02(Λ3)),\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}>R_{0}\left(\alpha,\frac{q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right), (1.5)

    where R0R_{0} is the positive solution to the quadratic equation

    24(α+32q022(Λ3))x2+(4(α+32q022(Λ3))2q02(Λ3)+6)x+1=0.24\left(\alpha+\frac{3}{2}-\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)x^{2}+\left(4\left(\alpha+\frac{3}{2}-\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)-\frac{2q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}+6\right)x+1=0. (1.6)

    Let 𝒮0(r+,Λ,α,q0)\mathcal{S}_{0}(r_{+},\Lambda,\alpha,q_{0}) be the set of all Me=0M<M0M_{e=0}\leq M<M_{0} such that (M,r+,Λ)(M,r_{+},\Lambda) are admissile sub-extremal parameters and (1.1)\eqref{Klein-Gordon} with parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) has a growing mode solution that is regular at the event horizon. Then 𝒮0(r+,Λ,α,q0)\mathcal{S}_{0}(r_{+},\Lambda,\alpha,q_{0}) is non-empty and open.

  3. (3)

    (Weakly charged case) For any given (r+,Λ,α)(r_{+},\Lambda,\alpha) satisfies (1.4) and (1.5) with q0=0q_{0}=0, there exists ϵ>0\epsilon>0 small enough, such that for all M(M0ϵ,M0)M\in(M_{0}-\epsilon,M_{0}), there exists δ>0\delta>0 depending on MM, such that M𝒮0(r+,Λ,α,q0)M\in\mathcal{S}_{0}(r_{+},\Lambda,\alpha,q_{0}) for all |q0|δ|q_{0}|\leq\delta.

We make the following remarks.

Remark 1.2.

In the large charge case, |q0||q_{0}| above is a constant depending on the given sub-extremal parameters, and growing mode solutions are constructed for spacetimes away from the extremality (in the sense that Me=0<M<Mb<M0M_{e=0}<M<M_{b}<M_{0}). For the general fixed charge case, however, there is no lower-bound requirement for q0q_{0} and thus our result includes the uncharged case q0=0q_{0}=0. Our general strategy in proving the large charge case and general fixed charge case is to construct growing mode solutions around a stationary solution and one should think M𝒮0(r+,Λ,α,q0)M\in\mathcal{S}_{0}(r_{+},\Lambda,\alpha,q_{0}) is close to McM_{c} where a stationary solution to (1.1) exists. For the weakly charged case, we can further construct growing mode solutions not only near a stationary solution but for all MM close to M0M_{0}. For the uncharged case, the conditions (1.4) and (1.5) are reduced to

94<α<32,\displaystyle-\frac{9}{4}<\alpha<-\frac{3}{2}, (1.7)
(Λ3)r+2>14(32α),\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}>\frac{1}{4(-\frac{3}{2}-\alpha)}, (1.8)

which includes the conformal mass α=2\alpha=-2.

Remark 1.3.

The construction of growing mode solutions for the uncharged case crucially relies on the non-positivity of the energy on spacelike hypersurfaces {t=const}\{t=const\}. As will be discussed in detail in Section 1.2.2, the condition

e2r+2<1+(Λ3)r+2,\frac{e^{2}}{r_{+}^{2}}<1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}, (1.9)

ensures that the energy remains positive and bounded on each constant time slice. Consequently, the boundedness of the solution to the uncharged Klein–Gordon equation (1.1) can be established using standard arguments. Therefore, for growing mode solutions in the uncharged case, the parameters must satisfy the following necessary “bound-violating” condition:

e2r+2>1+(Λ3)r+2,\frac{e^{2}}{r_{+}^{2}}>1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}, (1.10)

which can be interpreted as a largeness requirement for the black hole charge; see Section 1.2.2 for a detailed discussion.

Remark 1.4.

We also remark that the growing mode solutions constructed in this paper for the large charge case and the general fixed charge case are obtained by perturbing stationary solutions. It remains unclear to us whether growing mode solutions exist for the general fixed charge and the black hole being close to the extremality. However, by the computation in Section 1.2.2, we can show that if

(1+4q02(Λ3))e2r+2<1+(Λ3)r+2,\left(1+\frac{4q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)\frac{e^{2}}{r_{+}^{2}}<1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}, (1.11)

then there is no stationary solution to (1.1). Hence the necessary condition the existence of stationary solutions is

(1+4q02(Λ3))e2r+2>1+(Λ3)r+2.\left(1+\frac{4q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)\frac{e^{2}}{r_{+}^{2}}>1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}. (1.12)
The role of the black hole charge ee in the existence of growing mode solutions

It is important to note that, in direct contrast to the condition (1.10), the decay results in [39, 38] for Klein–Gordon equations on Schwarzschild-AdS and Kerr-AdS imply that for any parameters (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0})

Me=0𝒮0(r+,Λ,α,q0),Me=0𝒮(Mb,r+,Λ,α,q0)M_{e=0}\notin\mathcal{S}_{0}(r_{+},\Lambda,\alpha,q_{0}),\quad M_{e=0}\notin\mathcal{S}(M_{b},r_{+},\Lambda,\alpha,q_{0})

In other words, it is impossible to construct the growing mode solutions if the black hole spacetime itself is uncharged (e=0e=0) and spherically symmetric; see Section 1.1.4 for a detailed discussion.

The role of the scalar field charge q0q_{0} in the existence of growing mode solutions

We now distinguish between the charged and uncharged cases. For the charged case, since the charge |q0||q_{0}| can be taken to be arbitrarily large, we have

infq0inf𝒮(Mb,r+,Λ,α,q0)=Me=0.\inf_{q_{0}}\inf\mathcal{S}(M_{b},r_{+},\Lambda,\alpha,q_{0})=M_{e=0}.

The presence of a (large) scalar field charge allows the black hole away from the extremality and the black hole spacetime to be weakly charged. While for the general fixed charge case, by (1.5)\eqref{large} and the method in this paper, we can show

inf(r+,Λ)supM𝒮0(r+,Λ,α,q0)dΩ2dr|r=r+=0,\inf_{(r_{+},\Lambda)}\sup_{M\in\mathcal{S}_{0}(r_{+},\Lambda,\alpha,q_{0})}\frac{d\Omega^{2}}{dr}\Bigl{|}_{r=r_{+}}=0,

which can be interpreted as follows: for a fixed charge q0q_{0}, there exist parameters

(M𝒮0(r+,Λ,α,0),r+,Λ)\left(M\in\mathcal{S}_{0}(r_{+},\Lambda,\alpha,0),r_{+},\Lambda\right)

very close to the extremal parameters. The result in the weakly charged case implies that growing mode solutions exist for all (M,r+,Λ)(M,r_{+},\Lambda) close to the extremality.

Heuristically, the existence of growing mode solutions for large charge case in Theorem 1.1 can be explained in view of the fact that the charge q0q_{0} plays a role in the so-called “effective mass” αeff(r)\alpha_{eff}(r), i.e. the coefficient of zero order term of ϕ\phi in (1.1)\eqref{Klein-Gordon}. αeff(r)\alpha_{eff}(r) violates the Breitenlohner–Freedman bound in the compact region of (r+,)(r_{+},\infty) when the charge q0q_{0} is large while keeping limrαeff(r)\lim_{r\rightarrow\infty}\alpha_{eff}(r) satisfying the Breitenlohner–Freedman bound; see [26, 27, 28, 42] for the heuristical discussions. For the general fixed charge case and weakly charged case, the situation is more subtle since no large effective mass is available in general.

The Breitenlohner–Freedman bound and local well-posedness

The bound α>94\alpha>-\frac{9}{4} is called the Breitenlohner–Freedman bound, which is crucial for local well-posedness concerns of (1.1)\eqref{Klein-Gordon}. (1.1)\eqref{Klein-Gordon} is proved to be local well-posed under Dirichlet boundary conditions [34] for 94<α<0-\frac{9}{4}<\alpha<0 and under Neumann boundary conditions [55] for 94<α<54-\frac{9}{4}<\alpha<-\frac{5}{4}; see also Section 1.1.1. The ranges of our α\alpha in all three cases in Theorem 1.1 (condition (1.4)\eqref{large1}) fit within both local well-posedness regimes for the reflecting boundary conditions. Furthermore, we note (1.7)\eqref{uncharged range} is conjectured in [42] as the optimal masses range for the uncharged Klein–Gordon equation under which the growing mode solutions exist; see also Section 1.3. However, condition (1.5)\eqref{large} is new to the best of our knowledge.

A toy model for the linearized Einstein–Maxwell equations

The negative Klein–Gordon mass α\alpha in the setting is motivated by considering the Einstein–Maxwell equations with Λ<0\Lambda<0:

Ricμν(g)=Λgμν+TμνEM,Ric_{\mu\nu}(g)=\Lambda g_{\mu\nu}+T_{\mu\nu}^{EM},

where TμνEMT_{\mu\nu}^{EM} is the energy tensor given by the Maxwell field. Informally, RicRic can be viewed as a wave operator \Box under wave harmonic coordinates. Hence a toy model for studying the (in)stability of the Reissner–Nordström-AdS spacetime is the uncharged Klein–Gordon equation with a negative mass α\alpha. Furthermore, (1.1)\eqref{Klein-Gordon} with the conformal mass α=2\alpha=-2 can be viewed as a simplification of the analogous Teukolsky equations derived in the Reissner–Nordström-AdS spacetime, which plays an important role in the study of linear stability of the Reissner–Nordström-AdS spacetime. One can refer to [22] for the analogous Teukolsky equations on the asymptotically flat Reissner–Nordström-AdS spacetime.

Construction of hairy black holes

In [39], Holzegel and Smulevici proved that any spherically symmetric solutions to the Einstein–Klein–Gordon equations with a negative cosmological constant Λ\Lambda in the vicinity of Schwarzschild-AdS converge exponentially to the Schwarzschild-AdS again. In particular, the scalar field ϕ\phi decays exponentially in time. Their result shows that the only stationary spherical black hole solutions of the Einstein–Klein–Gordon equations with a negative cosmological constant Λ\Lambda are Schwarzschild-AdS spacetimes with vanishing scalar field ϕ\phi. However, using the mode solution construction obtained in this paper, which essentially relies on the non-zero black hole charge ee, see Remark (1.3) and (1.10), we show the existence of stationary black hole with a non-trivial (charged or uncharged) scalar field ϕ\phi under the reflecting boundary conditions for the Einstein–Maxwell–Klein–Gordon equations with a negative cosmological constant Λ\Lambda in our companion work [58].

Largely charged and weakly charged instability mechanism

In [51] and [19], growing mode solutions were constructed respectively to the Klein–Gordon equation on the Kerr and Kerr-AdS spacetimes violating the Hawking–Reall bound. The mechanism behind these instabilities is of a superradiant nature for spacetimes outside of spherical symmetric. In [5], growing mode solutions to the charged Klein–Gordon equation on Kerr–Newman-dS and Reissner–Nordström-dS have also been constructed, due to the strong coupling of the black hole charge and the scalar field charge. The mechanism behind this instability is also of a superradiant nature induced by the scalar field charge [2]; see Section 1.2.1 for a detailed discussion.

As mentioned above, the instability mechanism for our large charge case is due to the effective mass violating the Breitenlohner–Freedman bound. In view of (1.11), the coupling of the black hole charge ee and the scalar field charge q0q_{0} is also strong in our large charge case, which is very similar to the case of growing mode solutions on Kerr–Newman-dS. This instability is indeed of superradiant nature [18] and is called tachyonic instability in physics literature [26, 27, 28, 42, 8]. However, since there is no superradiance for the uncharged scalar field in the Reissner–Nordström-AdS spacetime, our growing mode solution is due to a new mechanism called near-extremal instability, as already discussed in [42]; see also Section 1.2.2 for a further discussion of this mechanism.

Outline of the rest of the introduction

In Section 1.1, we discuss some previous results regarding (in)stability on the asymptotically AdS spacetime and establish a connection between growing mode solutions we get in this paper and an instability conjecture of the Reissner–Nordström-AdS spacetime. In Section 1.2.1, we review previous works on constructing growing mode solutions to the Klein–Gordon equation on different spacetimes. We also give the brief introduction to the instability mechanism of these growing mode solutions there. In Section 1.2.2, we discuss the novel instability mechanism leading to the growing mode solutions of the uncharged Klein–Gordon equation on the Reissner–Nordström-AdS spacetime. In Section 1.3, we discuss the physics motivation and heuristic argument.

1.1 Previous results on stability/instability of the asymptotically AdS spacetime and decay of the field

In the past decades, despite intensive research aimed at proving the stability of the Einstein vacuum equation in the asymptotically flat spacetimes, there have been only a few results regarding the (in)stability of the asymptotically AdS spacetime. Due to the presence of a conformal timelike boundary, the choice of boundary conditions plays a decisive role in the (in)stability issues. To address the nonlinear (in)stability problems in the asymptotically AdS spacetime, the first step is to understand the decay and boundedness properties of the linearized field equation, which will heuristically suggest the (in)stability of the spacetime. As mentioned above, a natural toy model for the Einstein equations (coupled with matter field) with a negative cosmological constant is the Klein–Gordon equation on a fixed asymptotically AdS spacetime with a negative conformal mass.

First, we give a brief definition of boundary conditions for the Klein–Gordon equation with a conformal mass α=2\alpha=-2 here; see Section 3.3 for the definition for more general Klein–Gordon masses.

Definition 1.5.

For the Klein–Gordon equation (1.1)\eqref{Klein-Gordon} on an asymptotically AdS spacetime, the solution ϕ\phi satisfies the Dirichlet, Neumann, or dissipative boundary condition if the following holds:

  1. (1)

    Dirichlet boundary condition:

    rϕ0,r.r\phi\rightarrow 0,\quad r\rightarrow\infty.
  2. (2)

    Neumann boundary condition:

    r2r(rϕ)=0,r.r^{2}\frac{\partial}{\partial r}\left(r\phi\right)=0,\quad r\rightarrow\infty.
  3. (3)

    Optimally dissipative boundary condition:

    (rϕ)t+r2(rϕ)r0,r.\frac{\partial(r\phi)}{\partial t}+r^{2}\frac{\partial(r\phi)}{\partial r}\rightarrow 0,\quad r\rightarrow\infty.

We define the reflecting boundary condition to be either the Dirichlet boundary condition or the Neumann boundary condition. Note it is also possible to consider inear combinations of the Dirichlet and the Neumann boundary conditions, which are called the Robin boundary conditions. We will not pursue results under Robin boundary conditions in this paper.

1.1.1 Local well-posedness results

The study of the Klein–Gordon equation on asymptotically AdS spacetimes has been initiated in a series of works [34, 35]. In [34], Holzegel proved the local well-posedness results for Klein–Gordon equations on general asymptotically AdS spacetimes under the Dirichlet boundary conditions with 94<α<0-\frac{9}{4}<\alpha<0. For the Neumann boundary conditions, the local well-posedness results are still available [55] for 94<α<54-\frac{9}{4}<\alpha<-\frac{5}{4}. [21] establishes the local well-posedness under the dissipative boundary conditions for some certain range of α\alpha. In [35], Holzegel and Smulevici proved the local well-posedness results for the spherically symmetric Einstein–Klein–Gordon equations with a negative cosmological constant Λ\Lambda under Dirichlet boundary conditions with 2α<0-2\leq\alpha<0.

1.1.2 Results and conjectures on the stability of the pure AdS spacetime with dissipative boundary conditions

The pure AdS spacetime takes the form of

gAdS=(1+(Λ3)r2)dt2+(1+(Λ3)r2)1dr2+r2dσ2.g_{AdS}=-\left(1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r^{2}\right)dt^{2}+\left(1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r^{2}\right)^{-1}dr^{2}+r^{2}d\sigma^{2}.

In [36], the following conjecture was made for the pure AdS with optimally dissipative boundary conditions.

Conjecture 1.6 ([36]).

Anti-de Sitter spacetime is asymptotically stable for optimally dissipative boundary conditions.

Holzegel–Luk–Smulevici–Warnick showed the standard energy boundedness and integrated decay result for the Klein–Gordon equation with α=2\alpha=-2 under optimally dissipative boundary condition on the pure AdS spacetime in [36], which is the first step toward the proof of Conjecture 1.6.

1.1.3 Results and conjectures on the instability of the pure AdS spacetime with reflecting boundary conditions

In [11, 12], the following instability conjecture about the pure AdS spacetime with reflecting boundary conditions was made by Holzegel and Dafermos.

Conjecture 1.7 ([11, 12]).

Anti-de Sitter spacetime is non-linearly unstable for reflecting (Dirichlet or Neumann) boundary conditions.

The above conjecture relies on the expectation that, for any small perturbations to pure AdS initial data, the solutions to the Einstein vacuum equations with a negative cosmological constant Λ\Lambda will form a trapped surface region. In a series of works [47, 46, 48, 49], Moschidis proved the instability of AdS for the Einstein-null dust system and Einstein-massless Vlasov system with reflecting boundary conditions under the spherically symmetric setting.

1.1.4 Linear (in)stability of AdS black hole

In [38, 40], Holzegel and Smulevici showed a sharp inverse logarithmic decay rate for the Klein–Gordon equation on the Schwarzschild-AdS spacetime 111In fact, under the spherically symmetric assumption of ϕ\phi, they can further show ϕ\phi decays exponentially in time. under the Dirichlet boundary condition. Moreover, their sharp logarithmic decay result also holds for the Kerr-AdS spacetime satisfying the Hawking–Reall bound. One should note that the Hawing–Reall bound is crucial in obtaining the decay. In fact, Dold [19] constructed growing mode solutions of the Klein–Gordon equation on the Kerr-AdS spacetime violating the Hawking–Reall bound.

Once the decay result for the Klein–Gordon equation is obtained, the next step is to study the (in)stability of linearized gravity. For the Kerr spacetime, Teukolsky derived two decoupled equations for the linearized curvature components in [53]. Most recently, Graf and Holzegel proved mode stability for the analogous Teukolsky equations on the Kerr-AdS spacetime satisfying the Hawking–Reall bound in their work [23], thus ruling out the possibility of growing mode solutions.

However, the existence of growing mode solutions to the Klein–Gordon equation on the Reissner–Nordström-AdS spacetime constructed in this paper leads to the expectation of mode instability of the analogous Teukolsky equations, in contrast to the cases of Schwarzschild-AdS and Kerr-AdS. We will come back to this problem in our future work.

1.1.5 Nonlinear (in)stability of the asymptotically AdS spacetime

Now we discuss nonlinear (in)stability results for asymptotically AdS spacetimes. In [39], Holzegel and Smulevici showed that, for any small spherical perturbation of the Schwarzschild-AdS initial data, under the Dirichlet boundary conditions, solutions of the Einstein–Klein–Gordon equations with a negative cosmological constant will converge to the Schwarzschild-AdS spacetime exponentially in time, demonstrating the spherical stability of the Schwarzschild-AdS spacetime.

Most recently, in a series of works [24, 25] establishing the linear stability of Schwarzschild-AdS, Holzegel and Graf showed that under Dirichlet-type boundary conditions, the solutions to the Teukolsky equations with perturbed Schwarzschild-AdS initial data converge to Kerr-AdS with an inverse logarithmic rate in time.

Despite decay results for the Klein–Gordon equations and the Teukolsky equations, this inverse logarithmic decay rate is believed to be too slow to ensure the nonlinear stability of the Kerr-AdS spacetime. Hence, in [38], Holzegel and Smulevici made the following conjecture:

Conjecture 1.8 ([38]).

The Kerr-AdS spacetimes are non-linearly unstable solutions to the initial-boundary value problem for the Einstein equations with Dirichlet boundary conditions.

In contrast, the growing mode solutions we construct in this paper show instability even at the linear level. We naturally expect the nonlinear instability of the Reissner–Nordström-AdS spacetimes as solutions to the Einstein–Maxwell equations with a negative cosmological constant Λ\Lambda under reflecting boundary conditions.

1.2 Comparing mechanisms for growing mode solutions to the Klein–Gordon equation on different spacetimes

1.2.1 The known instability mechanism I: Superradiant instability

Growing mode solutions on the Kerr and Kerr-AdS spacetimes violating the Hawking–Reall bound have been constructed in [51, 19] respectively. Both spacetimes exhibit superradiance induced by rotation, namely, negative energy at the event horizon, which gives rise to the growing mode solutions222For the Klein–Gordon equation on Kerr, growing mode solutions are due to the combination of the superradiance and massive character of the equation. Heuristically, the Klein–Gordon mass serves as a reflecting mirror that would reflect the superradiance and result in a “black-hole bomb”; see [51, 15, 17, 59]. Note on the contrary, the massless wave equation on the Kerr spacetime does not admit exponentially growing mode solutions [52, 56]. Moreover, the solutions to the wave equation on Kerr are bounded and decay in time [13, 14, 1].. See [51, 19] for a detailed discussion of this superradiant instability. See also [50, 15, 17, 59] for some previous heuristic discussions.

While the rotation-induced superradiance phenomenon affects the Klein–Gordon equation on the Kerr/Kerr-AdS spacetimes, there is no such phenomenon on Reissner–Nordström/Reissner–Nordström-AdS since they are spherically symmetric. However, if one considers the charged Klein–Gordon equation on the charged spacetime, there is a charged analog of the superradiance induced by the coupling of the black hole charge ee and the scalar field charge q0q_{0}; see [18, 8, 16, 2] and references therein. Growing mode solutions on Kerr-Newman-dS and Reissner–Nordström-dS have been constructed in [5], for spacetimes where the product of the angular momentum and the Klein–Gordon mass is small compared to the product of the black hole charge ee and the scalar field charge q0q_{0}, which is a rigourous mathematical realization of this charge-induced superradiant instability. However, if the coupling is not strong, namely, the product of the black hole charge ee and the scalar field charge q0q_{0} is small, then the exponential decay of the local energy is obtained in [4] for the Klein–Gordon equation on Reissner–Nordström-dS.

The Breitenlohner–Freedman bound α>94\alpha>-\frac{9}{4} is crucial to provide a positive scalar field energy near the infinity. This is exploited to prove the local well-posedness of the asymptotically AdS spacetime; see [6] for the original physics argument and [35, 55] for a rigorous mathematics proof. The charged Klein–Gordon equation (1.1) takes the form of the uncharged Klein–Gordon equation with a effective mass

αeff=αq02At2(Λ3)Ω2,\alpha_{eff}=\alpha-\frac{q_{0}^{2}A_{t}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}\Omega^{2}},

for which the relevant Breitenlohner–Freedman bound αeff>94\alpha_{eff}>-\frac{9}{4} is satisfied when rr approaches infinity. Hence, we still have the local well-posedness result. However, for large charge q0q_{0}, it is possible to violate the analogous condition of the Breitenlohner–Freedman bound

αeff94\alpha_{eff}\leq-\frac{9}{4}

for a bounded value of rr. We exploit this fact in the proof of Lemma 6.2, which plays an important role in the construction. This superradiant mechanism has been identified as tachyonic instability in the context of physics literatures; see for instance [8, 29].

1.2.2 The new type of instability mechanism: Near-extremal instability

There is, however, no analogous superradiant mechanism for the uncharged Klein–Gordon equation on the Reissner–Nordström-AdS spacetime. Instead, the existence of growing mode solutions results from the black hole being too close to the extremality, which is called the near-extremal instability [29].

We try to illustrate the mathematical meaning of this new type of instability by comparing the uncharged Klein–Gordon equations on Schwarzschild-AdS and Reissner–Nordström-AdS. Using the (t,r,θ,φ)(t,r,\theta,\varphi) coordinates as in (1.2), let Σt={t=constant}\Sigma_{t}=\{t=constant\} be the foliation of spacetimes. Putting the Dirichlet boundary conditions at timelike infinity, the standard energy estimate gives

Σt1TμνXμnν𝑑vol=Σt2TμνXμnν𝑑vol,\int_{\Sigma_{t_{1}}}T_{\mu\nu}X^{\mu}n^{\nu}dvol=\int_{\Sigma_{t_{2}}}T_{\mu\nu}X^{\mu}n^{\nu}dvol,

where X=tX=\partial_{t} is the unique timelike Killing vector field on the spacetimes, nn is the normal vector field on Σt\Sigma_{t}, dvoldvol is the volume form of Σt\Sigma_{t}, and the energy momentum tensor TμνT_{\mu\nu} for the Klein–Gordon equation is

Tμν=(μϕνϕ¯)12gμν(gαβαϕβϕ¯+(Λ3)α|ϕ|2)T_{\mu\nu}=\Re\left({\nabla_{\mu}\phi\overline{\nabla_{\nu}\phi}}\right)-\frac{1}{2}g_{\mu\nu}\left(g^{\alpha\beta}\nabla_{\alpha}\phi\overline{\nabla_{\beta}\phi}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha|\phi|^{2}\right) (1.13)

The boundedness of the solution follows from a standard argument if the energy on the spacelike slice Σt=0\Sigma_{t=0} is positive. The conservation of energy shows that the quantity

r+𝕊2TttΩ2r2sinθdrdθdφ\int_{r_{+}}^{\infty}\int_{\mathbb{S}^{2}}\frac{T_{tt}}{\Omega^{2}}r^{2}\sin{\theta}drd\theta d\varphi

is independent of tt. However, since the Klein–Gordon mass α\alpha in our setting is negative, the positivity of TttT_{tt}

Ttt=12(|tϕ|2+Ω2(Ω2|rϕ|2+|∇̸ϕ|2+(Λ3)α|ϕ|2))T_{tt}=\frac{1}{2}\left(|\partial_{t}\phi|^{2}+\Omega^{2}\left(\Omega^{2}|\partial_{r}\phi|^{2}+|\not{\nabla}\phi|^{2}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha|\phi|^{2}\right)\right)

is not guaranteed at first glance. On the Schwarzschild-AdS spacetime, the integral of the negative term (Λ3)α|ϕ|2\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha|\phi|^{2} can be absorbed by the integral of the derivative term Ω2|rϕ|2\Omega^{2}|\partial_{r}\phi|^{2} using the Hardy inequality. The following Hardy inequality can indeed be obtained:

r+𝕊2|ϕ|2r2sinθdrdθdφ49r+𝕊2|rϕ|2(rr+)2(r+r++r+2r)2𝑑r𝑑θ𝑑φ.\int_{r_{+}}^{\infty}\int_{\mathbb{S}^{2}}|\phi|^{2}r^{2}\sin{\theta}drd\theta d\varphi\leq\frac{4}{9}\int_{r_{+}}^{\infty}\int_{\mathbb{S}^{2}}|\partial_{r}\phi|^{2}(r-r_{+})^{2}\left(r+r_{+}+\frac{r_{+}^{2}}{r}\right)^{2}drd\theta d\varphi. (1.14)

Using the above Hardy inequality, for the Schwarzschild-AdS spacetime, we have

r+𝕊2TttΩ2r2sinθdrdθdφ=r+𝕊2Ω2|rϕ|2r2sinθ+(Λ3)α|ϕ|2r2sinθdrdθdφ\displaystyle\int_{r_{+}}^{\infty}\int_{\mathbb{S}^{2}}\frac{T_{tt}}{\Omega^{2}}r^{2}\sin{\theta}drd\theta d\varphi=\int_{r_{+}}^{\infty}\int_{\mathbb{S}^{2}}\Omega^{2}|\partial_{r}\phi|^{2}r^{2}\sin\theta+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha|\phi|^{2}r^{2}\sin{\theta}drd\theta d\varphi (1.15)
\displaystyle\geq r+𝕊2(Λ3)(rr+)1r2(r3(Λ3)+r+3r2+r+4r+r+5)|ϕ|2sinθdrdθdφ>0.\displaystyle\int_{r_{+}}^{\infty}\int_{\mathbb{S}^{2}}\bigl{(}-\frac{\Lambda}{3}\bigr{)}(r-r_{+})\frac{1}{r^{2}}\left(\frac{r^{3}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}+r_{+}^{3}r^{2}+r_{+}^{4}r+r_{+}^{5}\right)|\phi|^{2}\sin{\theta}drd\theta d\varphi>0.

Moreover, for the Reissner–Nordström-AdS spacetime, using the argument as above, one can in fact show that if

e2r+<1+(Λ3)r+2,\frac{e^{2}}{r_{+}}<1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}, (1.16)

the energy is still positive. Hence by the standard commuting vector field and redshift methods, one can still show the boundedness of the solutions. Combining (1.16) and (1.5), we can show that growing mode solutions to the uncharged Klein–Gordon equation (1.1) can only exist if

e2r+2>1+14(32α).\frac{e^{2}}{r_{+}^{2}}>1+\frac{1}{4\left(-\frac{3}{2}-\alpha\right)}. (1.17)

In the above computation, the integral of the negative term (Λ3)αr2|ϕ|2sinθ\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha r^{2}|\phi|^{2}\sin{\theta} in the energy is controlled by the weighted L2L^{2} norm of rϕ\partial_{r}\phi by the Hardy inequality (1.14)\eqref{nav hardy}. This weight in (1.14) decays quadratically like (rr+)2(r-r_{+})^{2} towards the event horizon, whereas the weight Ω2\Omega^{2} of the L2L^{2} norm of rϕ\partial_{r}\phi in the original energy of the Schwarzschild-AdS spacetime decays at a linear rate C(rr+)C(r-r_{+}) with CC bounded away from 0. This fact is crucial in showing the absence of the negative energy. Hence, near the horizon, the integrand on the right-hand side of (1.15)\eqref{after hardy} is positive, giving the hope to absorb the negative term in the whole black hole exterior region r>r+r>r_{+}. However, on the extremal Reissner–Nordström-AdS spacetime, Ω2\Omega^{2} decays exactly quadratically like (rr+)2(r-r_{+})^{2}. A simple computation shows that the analogous integrand on the right-hand side of (1.15)\eqref{after hardy} might be negative near the event horizon. Hence, for the near-extremal Reissner–Nordström-AdS spacetime, if the initial data (ϕ,tϕ)|t=0(\phi,\partial_{t}\phi){\bigl{|}_{t=0}} is supported near the event horizon, the energy Σ0TμνXμnν\int_{\Sigma_{0}}T_{\mu\nu}X^{\mu}n^{\nu} might be negative; see Section 6.2 for the proof of the existence of negative energy in near-extremal Reissner–Nordström-AdS spacetimes. This phenomenon is the key observation allowing us to construct the growing mode solution in our main theorem.

1.3 Hairy black holes and connection to the AdS/CFT correspondence

1.3.1 Connection between growing mode solutions, oscillating mode solutions, and hairy black holes

The well-known no-hair conjecture in General Relativity asserts that all stationary black hole solutions to the Einstein (coupled with reasonable matter fields) equations are uniquely determined by the mass, charge, and angular momentum; for instance see the review [33] and references therein.

In [51], Shlapentokh-Rothman constructed solutions of the form eiωteimφSκml(θ)R(r)e^{-i\omega t}e^{im\varphi}S_{\kappa ml}(\theta)R(r) to the Klein–Gordon equation on the Kerr spacetime, which is interpreted as a “linear hair”, in violation to at least the spirit of the conjecture. He showed the existence of growing mode solutions (ω\omega has a negative imaginary part) and time-periodic solutions (ω\omega\in\mathbb{R}). Later, with Chodosh [9], they further constructed one-parameter families of the stationary axisymmetric asymptotically flat solutions (,gδ,ϕδ)(\mathcal{M},g_{\delta},\phi_{\delta}) to the Einstein–Klein–Gordon equations bifurcating off the Kerr solution. Their work gives a counter-example to the no-hair conjecture.333In fact, their result violates the no-hair conjecture to the extend that the metric is stationary but the scalar field is time-periodic. See also [32, 30, 7, 3, 31, 10] for previous physics and numerics results in this direction.

In the works [54, 43], the mathematical study of the hairy black hole interiors has been initiated. Putting the characteristic initial data (g,ϕ)(g,\phi) on the event horizon, corresponding to a stationary non-zero scalar field ϕ\phi, Van de Moortel [54] investigated the interior of the stationary Einstein–Maxwell–Klein–Gordon hairy black holes and their singular structures, assuming the scalar field is uncharged. Later, Van de Moortel and Li studied the analogous problem with the charged scalar field [43]. However, the existence of a stationary black hole exterior corresponding to the solutions in [54, 43] was open; see Open Problem iv in [54]. We resolve this problem in the companion work [58], taking advantage of the method in this paper, which gives the affirmative answer to Open Problem iv and provides examples of asymptotically AdS hairy black holes whose interior is governed by the results of [54, 43].

1.3.2 AdS/CFT correspondence, holographic superconductor, and near-horizon geometry

Asymptotically AdS black holes have also emerged as an object of interest in the celebrated AdS/CFT correspondence [44, 45, 57]. Growing efforts have been spent in finding holographic analogs of superconductors, a problem for which hairy black holes appear as natural candidates. In fact, the existence of non-trivial stationary asymptotically AdS hairy black holes has been predicted in the paper [29], using heuristics relying on the so-called near-horizon geometry. Indeed, it is possible to “map” the near-extremal Reissner–Nordström-AdS spacetime to AdS2×𝕊2AdS_{2}\times\mathbb{S}^{2} by a limiting procedure. The heuristics of [29] argue that the 22-dimensional Breitenlohner–Freedman bound is relevant to the existence of growing mode solutions to the uncharged Klein–Gordon equation on the extremal Reissner–Nordström-AdS spacetime, specifically for α\alpha in the range

94<α<32.-\frac{9}{4}<\alpha<-\frac{3}{2}.

Our result actually shows the growing mode solutions exist within this range for sub-extremal Reissner–Nordström-AdS black holes. We also identify explicitly a sufficient condition (1.5) for the growing mode solutions to exist, which is not previously identified in [29]. While our method in principle also works to construct growing mode solutions on the extremal Reissner–Nordström-AdS spacetime, we leave this question to future works.

Outline of the rest of the paper

In Section 3, we introduce the geometry of the Reissner–Nordström-AdS spacetime and discuss the near-horizon boundary conditions and near-infinity boundary conditions. In Section 4, we give precise statements of our main theorems and discuss the difficulties and new ideas. The construction of the stationary solutions of (1.1) is viewed as an important step toward the construction of the growing mode solutions. We construct the stationary solution with arbitratry boundary conditions in Section 5 and with reflecting boundary conditions in Section 6 and 7 respectively. Lastly, in section 8.1, we prove the existence of the desired growing mode solutions.

2 Acknowledgements

The author would like to thank his advisor, Maxime Van de Moortel, for his kind support, continuous encouragement, numerous inspiring discussions, and valuable suggestions for the manuscript. Additionally, the author expresses special thanks to Zheng-Chao Han for several very enlightening discussions.

3 Preliminary

3.1 Geometry of Reissner–Nordström-AdS space

The Reissner–Nordström-AdS metric in (t,r,θ,φ)(t,r,\theta,\varphi) coordinates is given by

gRN\displaystyle g_{RN} =Ω2dt2+1Ω2dr2+r2dσ2,\displaystyle=-\Omega^{2}dt^{2}+\frac{1}{\Omega^{2}}dr^{2}+r^{2}d\sigma^{2},
Ω2\displaystyle\Omega^{2} =12Mr+e2r2+(Λ3)r2,\displaystyle=1-\frac{2M}{r}+\frac{e^{2}}{r^{2}}+\left(-\frac{\Lambda}{3}\right)r^{2},

where M>0M>0 here is the mass of the black hole, ee\in\mathbb{R} is the charge of the black hole, and Λ<0\Lambda<0 is the negative cosmological constant. The Reissner–Nordström-AdS spacetime is uniquely determined by parameters (M,e,Λ)(M,e,\Lambda). We refer to these parameters (M,e,Λ)(M,e,\Lambda) as sub-extremal if the function Ω2\Omega^{2} admits two distinct positive roots, denoted as 0<r<r+0<r_{-}<r_{+}. The region r>r+r>r_{+} is known as the exterior of the Reissner–Nordström-AdS black hole and the hypersurface {r=r+}\{r=r_{+}\} is called the event horizon. We can define the horizon temperature TT as:

T:=dΩ2dr(r+).T:=\frac{d\Omega^{2}}{dr}(r_{+}). (3.1)

For sub-extremal parameters, since r+r_{+} is the largest real root, we have T>0T>0. Parameters (M,e,Λ)(M,e,\Lambda) are called extremal if Ω2\Omega^{2} has two identical positive roots. For extremal Reissner–Nordström-AdS, we have T=0T=0.

Note that the Reissner–Nordström-AdS spacetime breaks down in the usual (t,r,θ,φ)(t,r,\theta,\varphi) coordinates when r=r+r=r_{+}. To extend the spacetime smoothly to the event horizon, let

r(r)=r(12Mr+e2r2+(Λ3)r2)1𝑑r.r^{*}(r)=\int_{\infty}^{r}\left(1-\frac{2M}{r}+\frac{e^{2}}{r^{2}}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r^{2}\right)^{-1}dr.

we can see that

limrr(r)\displaystyle\lim_{r\rightarrow\infty}r^{*}(r) =0,\displaystyle=0,
limrr+r(r)\displaystyle\lim_{r\rightarrow r_{+}}r^{*}(r) =.\displaystyle=-\infty.

Then we can define the outgoing Eddington–Finkelstein (EF) coordinates by

v=t+r,r=r.v=t+r^{*},\quad r=r.

The Reissner–Nordström-AdS metric in the outgoing EF coordinates (v,r,θ,φ)(v,r,\theta,\varphi) becomes

gRN=Ω2dv2+2dvdr+r2dσ2.g_{RN}=-\Omega^{2}dv^{2}+2dvdr+r^{2}d\sigma^{2}.

As we can see from the asymptotic behavior of r(r)r^{*}(r) when rr approaches infinity, the affine length of the curve with constant (t,θ,φ)(t,\theta,\varphi) is finite, contrary to the asymptotically flat Reissner–Nordström spacetime, where rr^{*} there approaches infinity. By letting g=Ω2g~g=\Omega^{2}\widetilde{g}, we can show that the boundary of (,g~)(\mathcal{M},\widetilde{g}) with coordinates (t,r=0,θ,φ)(t,r^{*}=0,\theta,\varphi) is timelike; see Figure 1 for the Penrose diagram of the Reissner–Nordström-AdS spacetime.

i+i^{+}i+i^{+}i+i^{+}i+i^{+}r=r=\inftyr=r=\inftyr=0r=0r=0r=0r=0r=0r=0r=0r=r+r=r_{+}r=r+r=r_{+}r=r+r=r_{+}r=r+r=r_{+}r=rr=r_{-}r=rr=r_{-}r=rr=r_{-}r=rr=r_{-}
Figure 1: Penrose diagram for the asymptotically AdS spacetime.

3.2 Parameters for the spherically symmetric Klein–Gordon equation

The spherically symmetric Klein–Gordon equation on the Reissner–Nordström-AdS background is

r2Ω22ϕt22iq0r2AΩ2tϕ+r(r2Ω2ϕr)=((Λ3)αq02A2Ω2)r2ϕ.-\frac{r^{2}}{\Omega^{2}}\frac{\partial^{2}\phi}{\partial t^{2}}-2iq_{0}r^{2}\frac{A}{\Omega^{2}}\partial_{t}\phi+\frac{\partial}{\partial r}\left(r^{2}\Omega^{2}\frac{\partial\phi}{\partial r}\right)=\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)r^{2}\phi. (3.2)

If we further assume ϕ\phi is stationary, then we can get the stationary Klein–Gordon equation:

ddr(r2Ω2dϕdr)=((Λ3)αq02A2Ω2)r2ϕ.\frac{d}{dr}\left(r^{2}\Omega^{2}\frac{d\phi}{dr}\right)=\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)r^{2}\phi. (3.3)

We call (M,e,Λ,α,q0)(M,e,\Lambda,\alpha,q_{0}) parameters of the Klein–Gordon equation (3.2)\eqref{Klein-Gordon on RN} and (M,e,Λ,α,q0)(M,e,\Lambda,\alpha,q_{0}) sub-extremal parameters if the corresponding (M,e,Λ)(M,e,\Lambda) are sub-extremal parameters for Reissner–Nordström-AdS spacetime. However, since the sub-extremal condition concerns the roots of Ω2\Omega^{2}, which are solutions of a quartic equation, it is hard to provide an algebraic criterion for when the parameters will satisfy the sub-extremal condition. Hence it may be inconvenient to use (M,e,Λ,α,q0)(M,e,\Lambda,\alpha,q_{0}) as parameters. In the following proposition, we prove that we can always use (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) as parameters.

Proposition 3.1.

The parameters transformation (M,e,Λ,α,q0)(M,r+,Λ,α,q0)(M,e,\Lambda,\alpha,q_{0})\rightarrow(M,r_{+},\Lambda,\alpha,q_{0}) is a regular transformation. The new parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) satisfy the sub-extremal condition if and only if

MMe=0:=r+2(1+(Λ3)r+2),\displaystyle M\geq M_{e=0}:=\frac{r_{+}}{2}\left(1+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}\right), (3.4)
M<M0:=r+(1+2(Λ3)r+2).\displaystyle M<M_{0}:=r_{+}\left(1+2\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}\right). (3.5)

In other words, for fixed (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}), the sub-extremal condition can be achieved by decreasing MM to be smaller than M0M_{0} but larger than Me=0M_{e=0}.

Proof.

By the form of Ω2\Omega^{2}, we have

e2=r+2(12Mr++(Λ3)r+2)e^{2}=-r_{+}^{2}\left(1-\frac{2M}{r_{+}}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}\right)

It is easy to check the Jacobian of the map

(M,r+,Λ,α,q0)(M,e,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0})\rightarrow(M,e,\Lambda,\alpha,q_{0})

is non-zero. Hence the transformation is regular. By condition (3.4)\eqref{subextremal1}, we have e20e^{2}\geq 0.

Calculating the horizon temperature TT, we can get

T=\displaystyle T= 2Mr+22e2r+3+2(Λ3)r+\displaystyle\frac{2M}{r_{+}^{2}}-\frac{2e^{2}}{r_{+}^{3}}+2\left(-\frac{\Lambda}{3}\right)r_{+}
=\displaystyle= 2r+2(M0M)>0\displaystyle\frac{2}{r_{+}^{2}}(M_{0}-M)>0

by condition (3.5). Since

Ω2=(Λ3)r4+r22Mr+e2r2,\Omega^{2}=\frac{\left(-\frac{\Lambda}{3}\right)r^{4}+r^{2}-2Mr+e^{2}}{r^{2}},

by Vieta Theorem, we have

i=14ri=0,\displaystyle\sum_{i=1}^{4}r_{i}=0, (3.6)
14ri=e2(Λ3)>0,\displaystyle\prod\limits_{1}^{4}r_{i}=\frac{e^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}>0, (3.7)

where rir_{i}, i=1,,4i=1,\dots,4 are the four roots of the polynomial (Λ3)r4+r22Mr+e2\left(-\frac{\Lambda}{3}\right)r^{4}+r^{2}-2Mr+e^{2}.

Given that Ω2\Omega^{2} already has a positive root r+r_{+}, either it has two real roots r1=r+r_{1}=r_{+} and r2=rr_{2}=r_{-} along with two non-real conjugate roots r3r_{3} and r4r_{4}, or it has four real roots r1>r2>r3>r4r_{1}>r_{2}>r_{3}>r_{4}. For the first case, we have r3×r4>0r_{3}\times r_{4}>0. Thus we have r1>0r_{1}>0, r2>0r_{2}>0 by (3.7)\eqref{Vieta product}. Since T>0T>0 by condition (3.5)\eqref{subextremal2}, r+r_{+} is the largest root and (M,e,Λ)(M,e,\Lambda) are sub-extremal. If Ω2\Omega^{2} has four real roots, then by (3.6)\eqref{Vieta sum} and (3.7)\eqref{Vieta product}, we have r1,r2r_{1},\ r_{2} positive and r3,r4r_{3},\ r_{4} negative. Since r+>0r_{+}>0 and T>0T>0, then r+r_{+} is the largest root and parameters are sub-extremal. ∎

3.3 Boundary conditions

In this section, we give the precise definition of Dirichlet, Neumann, and Robin boundary conditions. Let Δ:=94+α\Delta:=\sqrt{\frac{9}{4}+\alpha}.

Definition 3.2.

[55] We say a C1C^{1} function ff on \mathcal{M} obeys Dirichlet, Neumann, or Robin boundary conditions if the following holds:

  1. (1)

    Dirichlet:

    r32Δf0,r.r^{\frac{3}{2}-\Delta}f\rightarrow 0,\quad r\rightarrow\infty.
  2. (2)

    Neumann:

    r2Δ+1ddr(r32Δf)0,r.r^{2\Delta+1}\frac{d}{dr}\left(r^{\frac{3}{2}-\Delta}f\right)\rightarrow 0,\quad r\rightarrow\infty.
  3. (3)

    Robin:

    r2Δ+1ddr(r32Δϕ)+βr32Δϕ0,r,r^{2\Delta+1}\frac{d}{dr}\left(r^{\frac{3}{2}-\Delta}\phi\right)+\beta r^{\frac{3}{2}-\Delta}\phi\rightarrow 0,\quad r\rightarrow\infty,

    where β\beta is a real constant.

3.4 Mode solutions

Assume ϕ=eiωtψ(r)\phi=e^{i\omega t}\psi(r) is a mode solution of (3.2)\eqref{Klein-Gordon on RN} satisfying the Dirichlet boundary condition, we can get the equation for ψ\psi:

ddr(r2Ω2dψdr)=((Λ3)αq02A2Ω2ω2Ω22q0AωΩ2)r2ψ.\frac{d}{dr}\left(r^{2}\Omega^{2}\frac{d\psi}{dr}\right)=\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}-\frac{\omega^{2}}{\Omega^{2}}-\frac{2q_{0}A\omega}{\Omega^{2}}\right)r^{2}\psi. (3.8)

To facilitate the discussion of solutions under Neumann boundary conditions, letting ψ=r32+ΔΨ\psi=r^{-\frac{3}{2}+\Delta}\Psi, we can rewrite the equation (3.8)\eqref{Klein-Gordon equation for psi} as:

ddr(r2Δ1Ω2dΨdr)\displaystyle\frac{d}{dr}\left(r^{2\Delta-1}\Omega^{2}\frac{d\Psi}{dr}\right) (3.9)
=\displaystyle= r2Δ1((Λ3)αq02A2Ω2ω2Ω22q0AωΩ2+(32Δ)(12+Δ)Ω2r2+(32Δ)1rdΩ2dr)Ψ.\displaystyle r^{2\Delta-1}\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}-\frac{\omega^{2}}{\Omega^{2}}-\frac{2q_{0}A\omega}{\Omega^{2}}+(\frac{3}{2}-\Delta)(-\frac{1}{2}+\Delta)\frac{\Omega^{2}}{r^{2}}+(\frac{3}{2}-\Delta)\frac{1}{r}\frac{d\Omega^{2}}{dr}\right)\Psi.

From the asymptotic analysis, the behavior of ψ\psi near r=r=\infty is given by Dirichlet, Neumann, or Robin conditions. Next, we derive the asymptotic behavior of ψ\psi when rr+r\rightarrow r_{+}. Since the Reissner–Nordström-AdS metric can be extended to the event horizon in the outgoing EF coordinates, we require the solution ϕ\phi of the Klein–Gordon equation can also be extended to the event horizon. We have

ϕ=eiωtψ(r)=eiω(t+r)eiωrψ(r).\phi=e^{i\omega t}\psi(r)=e^{i\omega(t+r^{*})}e^{-i\omega r^{*}}\psi(r).

Hence to do the extension, we require that ρ(r):=eiωrψ\rho(r):=e^{-i\omega r^{*}}\psi can extended to be a smooth function on [r+,)[r_{+},\infty). We have the following relation

ψ\displaystyle\psi =eiωrρ(r),\displaystyle=e^{i\omega r^{*}}\rho(r), (3.10)
ψ\displaystyle\psi Ceiωr,rr+.\displaystyle\approx Ce^{i\omega r^{*}},\quad r\rightarrow r_{+}. (3.11)

From (3.10)\eqref{function relation}, we have

dψdr=iωψ+O(rr+).\frac{d\psi}{dr^{*}}=i\omega\psi+O(r-r_{+}). (3.12)

For growing mode solutions (ω)<0\Im(\omega)<0, we have the asymptotic decay behavior:

|ψ||rr+|C(ω).|\psi|\approx|r-r_{+}|^{-C\Im(\omega)}. (3.13)

We can prove the following result:

Proposition 3.3.

For any given sub-extremal parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}), the only real mode solution to the Klein-Gordon equation (3.8)\eqref{Klein-Gordon equation for psi} with Dirichlet or Neumann boundary conditions which can be extended smoothly to the horizon is the static solution ϕ(t,r)=ψ(r)\phi(t,r)=\psi(r) with ω=0\omega=0.

Proof.

Multiplying ψ¯\overline{\psi} and taking the imaginary part of the equation (3.8)\eqref{Klein-Gordon equation for psi}, we have

(ddr(r2Ω2dψdr)ψ¯)=r2|ψ|2((Λ3)αq02A2Ω2ω2Ω22q0AωΩ2).\Im\left(\frac{d}{dr}\left(r^{2}\Omega^{2}\frac{d\psi}{dr}\right)\overline{\psi}\right)=r^{2}|\psi|^{2}\Im\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}-\frac{\omega^{2}}{\Omega^{2}}-\frac{2q_{0}A\omega}{\Omega^{2}}\right).

Hence if ω\omega is real, we have

ddr(r2Ω2ψ¯dψdr)=0.\frac{d}{dr}\Im\left(r^{2}\Omega^{2}\overline{\psi}\frac{d\psi}{dr}\right)=0.

For ψ\psi satisfying the Dirichlet boundary condition, using (3.12)\eqref{near horizon for psi} and (3.10)\eqref{function relation}, we have

r+2limrr+(Ω2ψ¯dψdr)=r+2ω|ρ|2(r+)=0.r_{+}^{2}\lim_{r\rightarrow r_{+}}\Im\left(\Omega^{2}\overline{\psi}\frac{d\psi}{dr}\right)=r_{+}^{2}\omega|\rho|^{2}(r_{+})=0.

Hence ω=0\omega=0. For ψ\psi satisfying the Neumann boundary condition, using the equation (3.9)\eqref{twisted Klein-Gordon equation for psi} and reapplying the above argument, we can prove the same conclusion. ∎

Furthermore, for the uncharged Klein–Gordon equation, we have the following result about growing mode solutions.

Proposition 3.4.

All growing mode solutions to the uncharged equation (3.8) on the sub-extremal Reissner–Nordström-AdS spacetime have pure imaginary mode ω\omega.

Proof.

Assume ϕ=eiωtψ\phi=e^{i\omega t}\psi is a growing mode solution to the uncharged Klein–Gordon equation with ω=ωR+iωI\omega=\omega_{R}+i\omega_{I}. Then multiplying ψ¯\overline{\psi} and taking the imaginary part of the equation (3.8), we have

(ddr(r2Ω2dψdr)ψ¯)=r2|ψ|22ωRωIΩ2.\Im\left(\frac{d}{dr}\left(r^{2}\Omega^{2}\frac{d\psi}{dr}\right)\bar{\psi}\right)=-r^{2}|\psi|^{2}\frac{2\omega_{R}\omega_{I}}{\Omega^{2}}.

Integrating the above equations and using the boundary condition (3.13), we have

0=2ωRωIr+r2|ψ|2Ω2dr.0=-2\omega_{R}\omega_{I}\int_{r_{+}}^{\infty}\frac{r^{2}|\psi|^{2}}{\Omega^{2}}\mathrm{d}r. (3.14)

Since ωI\omega_{I} is negative, we conclude that ωR\omega_{R} is zero. ∎

4 Main results and outline of the proof

4.1 Precise statements of main results

Now we are ready to state the main results we get.

Theorem 4.1.

For Klein–Gordon equation (1.1)\eqref{Klein-Gordon} with negative mass α\alpha, let CDN=0C_{DN}=0 for Dirichlet boundary conditions and CDN=54C_{DN}=-\frac{5}{4} for Neumann boundary conditions. Imposing reflecting boundary condition for (1.1), we have

  1. (1)

    (Large charge case) For any given sub-extremal parameters (Mb,r+,Λ)(M_{b},r_{+},\Lambda) with Me=0<Mb<M0M_{e=0}<M_{b}<M_{0} and 94<α<CDN-\frac{9}{4}<\alpha<C_{DN}, there exists a q1>0q_{1}>0, such that for any |q0|>q1|q_{0}|>q_{1} and δ\delta sufficiently small, there exist real analytic functions Me=0<M(ϵ)<MbM_{e=0}<M(\epsilon)<M_{b} and ωR(ϵ)\omega_{R}(\epsilon)\in\mathbb{R} with ωR(0)=0\omega_{R}(0)=0 on δ<ϵ<δ-\delta<\epsilon<\delta so that there exists a mode solution ϕ=ei(ωR(ϵ)+iϵ)tψ(r)\phi=e^{i(\omega_{R}(\epsilon)+i\epsilon)t}\psi(r) of (1.1)\eqref{Klein-Gordon} with sub-extremal parameters (M(ϵ),r+,Λ,α,q0)\left(M(\epsilon),r_{+},\Lambda,\alpha,q_{0}\right). Moreover, ϕ\phi can be continuously extended to the event horizon r=r+r=r_{+}.

  2. (2)

    (General fixed charge case) For each fixed parameters (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}) satisfying

    94<α<{32+12q02(Λ3),CDN},\displaystyle-\frac{9}{4}<\alpha<\left\{-\frac{3}{2}+\frac{1}{2}\frac{q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}},C_{DN}\right\}, (4.1)
    (Λ3)r+2>R0,\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}>R_{0}, (4.2)

    where R0R_{0} is the positive solution to the quadratic equation (1.6). Then for any δ\delta sufficiently small, there exist real analytic functions Me=0<M(ϵ)<M0M_{e=0}<M(\epsilon)<M_{0} and ωR(ϵ)\omega_{R}(\epsilon)\in\mathbb{R} with ωR(0)=0\omega_{R}(0)=0 on δ<ϵ<δ-\delta<\epsilon<\delta, such that (M(ϵ),r+,Λ)(M(\epsilon),r_{+},\Lambda) are sub-extremal parameters and there exists a mode solution ϕ=ei(ωR(ϵ)+iϵ)t\phi=e^{i(\omega_{R}(\epsilon)+i\epsilon)t} of (1.1)\eqref{Klein-Gordon} with parameters (M(ϵ),r+,Λ,α,q0)\left(M(\epsilon),r_{+},\Lambda,\alpha,q_{0}\right). Moreover, ϕ\phi can be continuously extended to the event horizon r=r+r=r_{+} and we have

    dMdϵ(0)<0,\displaystyle\frac{dM}{d\epsilon}(0)<0, (4.3)
    q0edωRdϵ(0)<0,q0e0.\displaystyle q_{0}e\frac{d\omega_{R}}{d\epsilon}(0)<0,\quad q_{0}e\neq 0. (4.4)
  3. (3)

    (Weakly charged case) For each fixed parameters (r+,Λ,α,q0=0)(r_{+},\Lambda,\alpha,q_{0}=0) satisfying

    94<α<32,\displaystyle-\frac{9}{4}<\alpha<-\frac{3}{2}, (4.5)
    (Λ3)r+2>14(32α),\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}>\frac{1}{4\left(-\frac{3}{2}-\alpha\right)}, (4.6)

    there exists Me=0<Mc<M0M_{e=0}<M_{c}<M_{0} such that for any M(Mc,M0)M\in(M_{c},M_{0}), there exists a growing mode solution to (1.1) with parameters (M,r+,Λ,α,q0=0)(M,r_{+},\Lambda,\alpha,q_{0}=0). Furthermore, for M(Mc,M0)M\in(M_{c},M_{0}), there exists δ>0\delta>0 depending on MM, such that for all 0|q0|δ0\leq|q_{0}|\leq\delta, (1.1) with parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) has a growing mode solution.

Remark 4.2.

In both the large charge case and the general fixed charge case, when ϵ=0\epsilon=0, the solutions constructed in the theorem reduce to stationary solutions. One should think the growing mode solutions in these two cases are constructed by perturbing a stationary solution. However, in view of Proposition 3.4, one can treat the existence of growing mode solutions to the uncharged Klein–Gordon equation (3.8) as an negative eigenvalue problem. This perspective allows us to construct growing mode solutions to the uncharged Klein–Gordon equation for all M(Mc,M0)M\in(M_{c},M_{0}) without using perturbative approach. Additional, growing mode solutions for the weakly charged case can be obtained by perturbing the scalar field charge.

Remark 4.3.

In contrast to Proposition 3.4, (4.4) shows the existence of oscillating profile (ω0\Re\omega\neq 0) for growing mode solutions to the charged Klein–Gordon equations.

The key step toward proving the existence of growing mode solutions is the construction of non-trivial bounded stationary solutions to (1.1)\eqref{Klein-Gordon} with Dirichlet or Neumann boundary conditions. Under the stationary assumption, (1.1)\eqref{Klein-Gordon} is reduced to the ODE (3.3)\eqref{stationary Klein-Gordon}. Instead of solving ODE (3.3)\eqref{stationary Klein-Gordon} under given boundary conditions, we start with regular data at r=r+r=r_{+} and solve the equation (3.3)\eqref{stationary Klein-Gordon} to r=r=\infty. Considering the asymptotic behavior of the solution, we can prove the following proposition:

Proposition 4.4.

For any sub-extremal parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) satisfying the Breitenlohner–Freedman bound 94<α<0-\frac{9}{4}<\alpha<0, we can construct the static spherically symmetric solution ϕ(r)\phi(r) to the Klein–Gordon equation (1.1)\eqref{Klein-Gordon} under fixed Reissner–Nordström AdS metric gg such that

limrr+ϕ(r)=1,\displaystyle\lim_{r\rightarrow r_{+}}\phi(r)=1, (4.7)
ϕ(r)=CDuD(r)+CNuN(r).\displaystyle\phi(r)=C_{D}u_{D}(r)+C_{N}u_{N}(r). (4.8)

{uD(r),uN(r)}\{u_{D}(r),u_{N}(r)\} is the local basis of the solution of equations (3.2)\eqref{Klein-Gordon on RN}, with the following asymptotic behaviors:

limrr32+ΔuD(r)=1,limrr52+ΔduDdr=32Δ,\displaystyle\lim_{r\rightarrow\infty}r^{\frac{3}{2}+\Delta}u_{D}(r)=1,\quad\lim_{r\rightarrow\infty}r^{\frac{5}{2}+\Delta}\frac{du_{D}}{dr}=-\frac{3}{2}-\Delta, (4.9)
limrr32ΔuN(r)=1,limrr52ΔduNdr=32+Δ.\displaystyle\lim_{r\rightarrow\infty}r^{\frac{3}{2}-\Delta}u_{N}(r)=1,\quad\lim_{r\rightarrow\infty}r^{\frac{5}{2}-\Delta}\frac{du_{N}}{dr}=-\frac{3}{2}+\Delta. (4.10)

In other words, uDu_{D} is a function on \mathcal{M} satisfying the Dirichlet boundary condition and uNu_{N} is a function on \mathcal{M} satisfying the Neumann boundary condition.

Remark 4.5.

In the later discussion, we use the notation uDr32Δu_{D}\approx r^{-\frac{3}{2}-\Delta} to denote the asymptotic behavior (4.9)\eqref{Dirichlet asymptotic} and notation uNr32+Δu_{N}\approx r^{-\frac{3}{2}+\Delta} to denote the asymptotic behavior (4.10)\eqref{Neumann asymptotic}.

The above theorem holds for any sub-extremal parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}), highlighting the significance of the boundary conditions. For a given boundary condition, it’s highly non-trivial that one can prove the existence of a regular stationary solution on [r+,)[r_{+},\infty).

Furthermore, we can prove the following existence of non-trivial stationary solutions to (1.1)\eqref{Klein-Gordon} with reflecting boundary conditions.

Theorem 4.6.

Imposing the reflecting boundary condition for the Klein–Gordon equation (1.1)\eqref{Klein-Gordon} with a negative mass α(94,CDN)\alpha\in(-\frac{9}{4},C_{DN}), we have

  1. (1)

    (Large charge scalar field) For any given sub-extremal parameters (Mb,r+,Λ)(M_{b},r_{+},\Lambda) with Me=0<Mb<M0M_{e=0}<M_{b}<M_{0} and α\alpha within the above range, there exists a q1>0q_{1}>0, such that for any |q0|>q1|q_{0}|>q_{1}, there exists Me=0<M=M(Mb,r+,Λ,α,q0)<MbM_{e=0}<M=M(M_{b},r_{+},\Lambda,\alpha,q_{0})<M_{b}, such that a stationary solution to (1.1)\eqref{Klein-Gordon} with parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) exists and can be extended continuously to the event horizon {r=r+}\{r=r_{+}\}.

  2. (2)

    (General charged scalar field) For any given parameters (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}) satisfying the conditions (4.1) and (4.2), there exists Me=0<M<M0M_{e=0}<M<M_{0} such that a stationary solution ϕ\phi to (1.1)\eqref{Klein-Gordon} with parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) exists and can be extended continuously to the event horizon {r=r+}\{r=r_{+}\}.

4.2 Main difficulty and outline of the proof

In this section, we discuss the main difficulties in the proof and ideas we used by considering the uncharged Klein–Gordon equation with Dirichlet boundary conditions; see Section 7.1 for the discussion of Neumann boundary conditions. The ideas of doing the charged case are essentially the same. We mainly adopt the method outlined in [51]. The proof is divided into two main steps:

  • Prove the existence of stationary solutions to Klein–Gordon equation (1.1) as in Theorem 4.6.

  • Perturb the stationary solution and use the implicit function theorem to get growing mode solutions.

The main difficulty is the first step.To show the existence of a non-trivial stationary solution ϕ\phi with Dirichlet boundary condition, we apply the variational method to find the minimizer of the energy functional of (3.3)\eqref{stationary Klein-Gordon}:

L[f]=r+r2Ω2(dfdr)2+(Λ3)αr2f2dr.L[f]=\int_{r_{+}}^{\infty}r^{2}\Omega^{2}\left(\frac{df}{dr}\right)^{2}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha r^{2}f^{2}dr.

The asymptotic analysis of solution ϕ\phi gives that L[ϕ]L[\phi] is finite if and only if ϕ\phi is bounded and satisfies the Dirichlet boundary condition. Furthermore, the energy identity of (3.3)\eqref{stationary Klein-Gordon} implies L[ϕ]=0L[\phi]=0 for the stationary solution ϕ\phi.

To construct the desired solution, we apply the variational argument to L[f]L[f] within the function class H01H_{0}^{1} with

r+r2Ω2f2dr=1.\int_{r_{+}}^{\infty}\frac{r^{2}}{\Omega^{2}}f^{2}\mathrm{d}r=1.

By the constrained variational principle, the minimizer ϕ\phi will be the solution of the following eigenvalue equation

ddr(r2Ω2dϕdr)+(Λ3)αr2ϕ=λ(M,r+,Λ,α)r2Ω2ϕ.-\frac{d}{dr}\left(r^{2}\Omega^{2}\frac{d\phi}{dr}\right)+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha r^{2}\phi=-\lambda(M,r_{+},\Lambda,\alpha)\frac{r^{2}}{\Omega^{2}}\phi.

For fixed neutral parameters (r+,Λ,α,q0=0)(r_{+},\Lambda,\alpha,q_{0}=0), we denote the MM-dependence of ϕ\phi by ϕM\phi_{M}. To show λ(M)\lambda(M) can be zero for some MM, we start with M=M0M=M_{0} such that (M,r+,Λ)(M,r_{+},\Lambda) are extremal parameters for the given (r+,Λ)(r_{+},\Lambda) and show that λ(M0)\lambda(M_{0}) is negative. Then by exploiting the monotonicity and continuity of L[f]L[f] with respect to MM, we conclude the existence of M=McM=M_{c} such that λ(Mc)=0\lambda(M_{c})=0.

However, several difficulties arise when attempting to apply the above framework.

  1. 1.

    First, the energy functional is not even lower bounded at first glance, as the functions in our function class essentially have an L2L^{2} norm equal to 11 while L[f]L[f] contains the integral of r2f2r^{2}f^{2}.

  2. 2.

    Second, even if L[f]L[f] is lower bounded in our function class, since Ω2\Omega^{2} degenerates at the event horizon, we could not get the H1H^{1} boundedness of the minimizing sequence ϕM,n\phi_{M,n}. Hence, to the best of our knowledge, no (weak) convergence result we can use.

  3. 3.

    Third, the above framework requires us to start with M=M0M=M_{0}, and the corresponding LM0[f]L_{M_{0}}[f] should have a negative minimum. However, as already shown in Section 1.2.2, for small black hole charge (1.16), one can prove the positivity of the energy functional, which means the existence of the energy functional with a negative minimun is subtle in our setting.

The second difficulty is the easiest one to overcome. We can directly use the method outlined in [51] by considering the perturbed energy functional Lϵ[f]L^{\epsilon}[f], which has the same integrand but is integrated over (r++ϵ,)(r_{+}+\epsilon,\infty). Then for each fixed ϵ\epsilon, we can show that Ω2\Omega^{2} is lower bounded on (r++ϵ,)(r_{+}+\epsilon,\infty), thereby we can prove the H1H^{1} boundedness of the minimizing sequence of this perturbed energy functional. By the weak compactness of H1H^{1} and Sobolev embedding, we can show the existence of the minimizer ϕϵ\phi^{\epsilon} of Lϵ[f]L^{\epsilon}[f]; See Proposition 6.6. To get rid of ϵ\epsilon, our method is to take a step back by showing the local H1H^{1} boundedness on any fixed compact set K(r+,)K\subset(r_{+},\infty) and then achieving local convergence for each KK. Nonetheless, the limit obtained from the local convergence argument might be trivial since the energy can concentrate outside of the compact set KK despite having fL2(r++ϵ,)=1\|f\|_{L^{2}(r_{+}+\epsilon,\infty)}=1. Therefore, we still need some coercivity results.

One of the main new ideas in this paper lie in the proof of the first difficulty and coercivity mentioned above, which uses the so-called twisted derivative obtained by replacing the usual derivative ddr\frac{d}{dr} with ~r()=hdr(h1)\widetilde{\nabla}_{r}(\cdot)=h\frac{d}{r}(h^{-1}\cdot); see Section 6.1. The twisted derivative has been used to establish the local well-posedness of the Klein–Gordon equation on asymptotically AdS space with Neumann boundary conditions in [55]. This approach addresses the difficulty arising from the fact that the energy functional L[f]L[f] is infinite for functions with Neumann boundary conditions. The surprising aspect here is that, even for Dirichlet boundary conditions, we need to employ the twisted derivative. Writing the equation and energy functional in the twisted derivative form, the structure of Ω2\Omega^{2} will influence the potential term, enriching the sign structure of the potential. By a careful analysis of the potential and integrability of solutions ϕϵ\phi^{\epsilon} in Proposition 6.10 and 6.9 we can get the lower boundedness and coercivity.

Dealing with the third difficulty of achieving a negative minimum for the uncharged L[f]L[f] is particularly challenging. To overcome this obstacle, we derive a sharp near-horizon version of the Hardy inequality in Lemma 6.3 and devise a test function with a compact support to demonstrate the existence of a negative minimum in Lemma 6.5.

The non-trivial stationary solution with Neumann boundary conditions can be constructed similarly, by using a different twist function hh in the twisted derivative.

Growing mode solutions can be obtained from the stationary solution through the application of the implicit function theorem; see Section 8.1.

5 Proof of Proposition 4.4

The proof of Proposition 4.4 is a standard application of the asymptotic analysis. We provide a proof in this section for completeness.

Proof.

By the local asymptotic analysis of equations (3.3)\eqref{stationary Klein-Gordon}, locally near r=r+r=r_{+} we have

ϕ=Aϕ1+Blog(rr+)ϕ2,r+<r<r++ϵ,\phi=A\phi_{1}+B\log(r-r_{+})\phi_{2},\quad r_{+}<r<r_{+}+\epsilon,

where ϕ1\phi_{1} and ϕ2\phi_{2} are analytic functions on (r+,r++ϵ)(r_{+},r_{+}+\epsilon) and finite at r=r+r=r_{+}. If we can show when B=0B=0, the solution ϕ\phi can be extended to the whole domain (r+,)(r_{+},\infty), then by the asymptotic analysis at r=r=\infty, we have

ϕ=CDuD(r)+CNuN(r),\phi=C_{D}u_{D}(r)+C_{N}u_{N}(r),

where uDu_{D} and uNu_{N} are the local solutions of (3.3)\eqref{stationary Klein-Gordon} at r=r=\infty satisfying (4.9)\eqref{Dirichlet asymptotic} and (4.10)\eqref{Neumann asymptotic} respectively.

Note in (3.3)\eqref{stationary Klein-Gordon}, the term A2Ω2\frac{A^{2}}{\Omega^{2}} is defined everywhere on [r+,)[r_{+},\infty) since

limrr+A2Ω2(r)=e2r+4Tlimrr+(rr+)2rr+=0.\lim_{r\rightarrow r_{+}}\frac{A^{2}}{\Omega^{2}}(r)=\frac{e^{2}}{r_{+}^{4}T}\lim_{r\rightarrow r_{+}}\frac{(r-r_{+})^{2}}{r-r_{+}}=0.

Multiplying 1r2dϕdr\frac{1}{r^{2}}\frac{d\phi}{dr} on both sides of the equation (3.3)\eqref{stationary Klein-Gordon}, we have

12(Ω2(dϕdr)2+(Λ3)αϕ2+q02A2Ω2ϕ2)=(12dΩ2dr+2rΩ2)(dϕdr)2+ddr(q02A2Ω2)ϕ2.\frac{1}{2}\left(\Omega^{2}\left(\frac{d\phi}{dr}\right)^{2}+\bigl{(}\frac{\Lambda}{3}\bigr{)}\alpha\phi^{2}+\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\phi^{2}\right)=-\left(\frac{1}{2}\frac{d\Omega^{2}}{dr}+\frac{2}{r}\Omega^{2}\right)\left(\frac{d\phi}{dr}\right)^{2}+\frac{d}{dr}\left(\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)\phi^{2}. (5.1)

Integrating (5.1)\eqref{linear stimulated energy}, we have

12Ω2(dϕdr)2(r)+12(Λ3)αϕ2(r)+q02A22Ω2ϕ2(r)12(Λ3)αϕ2(0)+r+rddr(q02A2Ω2)ϕ2dr¯.\frac{1}{2}\Omega^{2}\left(\frac{d\phi}{dr}\right)^{2}(r)+\frac{1}{2}\bigl{(}\frac{\Lambda}{3}\bigr{)}\alpha\phi^{2}(r)+\frac{q_{0}^{2}A^{2}}{2\Omega^{2}}\phi^{2}(r)\leq\frac{1}{2}\bigl{(}\frac{\Lambda}{3}\bigr{)}\alpha\phi^{2}(0)+\int_{r_{+}}^{r}\frac{d}{dr}\left(\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)\phi^{2}\mathrm{d}\bar{r}. (5.2)

Then by the Gronwall inequality, we have

ϕ2(r)ϕ2(0)er+rddr(q02A2Ω2)dr¯,\phi^{2}(r)\leq\phi^{2}(0)e^{\int_{r_{+}}^{r}\frac{d}{dr}\left(\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)\mathrm{d}\bar{r}},

which means ϕ(r)\phi(r) and dϕdr\frac{d\phi}{dr} are finite on any interval [r+,R)[r_{+},R). Hence by the extension principle of the ODE, we know the solution of (3.3)\eqref{stationary Klein-Gordon} exists on (r+,)(r_{+},\infty). ∎

We can aslo prove the following Wronskian estimate of {uD,uN}\{u_{D},u_{N}\}.

Proposition 5.1.

For the local basis {uD,uN}\{u_{D},u_{N}\} of solutions of the linear Klein–Gordon equation (3.3)\eqref{stationary Klein-Gordon} in Proposition 4.4, for NN large and ϵ\epsilon small enough, we have the following bounds

|uDuNuDuN|Cr4,r>N,\displaystyle|u_{D}^{\prime}u_{N}-u_{D}u^{\prime}_{N}|\geq Cr^{-4},\quad r>N, (5.3)
|uDuNuDuN|C1rr+,r+<r<r++ϵ.\displaystyle|u_{D}^{\prime}u_{N}-u_{D}u^{\prime}_{N}|\geq C\frac{1}{r-r_{+}},\quad r_{+}<r<r_{+}+\epsilon. (5.4)
Proof.

By the asymptotic behavior of uDu_{D} and uNu_{N}, (5.3)\eqref{infty wronskian} follows trivially. Let

{ϕ1,log(rr+)ϕ2}\{\phi_{1},\log(r-r_{+})\phi_{2}\}

be the local basis of the solution of (3.3)\eqref{stationary Klein-Gordon} near r=r+r=r_{+}. We have

uD=ADϕ1+BDlog(rr+)ϕ2,\displaystyle u_{D}=A_{D}\phi_{1}+B_{D}\log(r-r_{+})\phi_{2},
uN=ANϕ1+BNlog(rr+)ϕ2.\displaystyle u_{N}=A_{N}\phi_{1}+B_{N}\log(r-r_{+})\phi_{2}.

Calculating the Wronskian, we have

|uDuNuDuN|\displaystyle|u_{D}^{\prime}u_{N}-u_{D}u^{\prime}_{N}| (5.5)
=\displaystyle= (ADϕ1+BDlog(rr+)ϕ2+BDϕ2rr+)(ANϕ1+BNlog(rr+)ϕ2)\displaystyle\left(A_{D}\phi_{1}^{\prime}+B_{D}\log(r-r_{+})\phi_{2}^{\prime}+B_{D}\frac{\phi_{2}}{r-r_{+}}\right)\left(A_{N}\phi_{1}+B_{N}\log(r-r_{+})\phi_{2}\right)
(ADϕ1+BDlog(rr+)ϕ2)(ANϕ1+BNlog(rr+)ϕ2+BNϕ2rr+)\displaystyle-\left(A_{D}\phi_{1}+B_{D}\log(r-r_{+})\phi_{2}\right)\left(A_{N}\phi_{1}^{\prime}+B_{N}\log(r-r_{+})\phi_{2}^{\prime}+B_{N}\frac{\phi_{2}}{r-r_{+}}\right)
\displaystyle\geq 12|ϕ1(r+)ϕ2(r+)|rr+|ADBNANBD|.\displaystyle\frac{1}{2}\frac{\left|\phi_{1}(r_{+})\phi_{2}(r_{+})\right|}{r-r_{+}}\left|A_{D}B_{N}-A_{N}B_{D}\right|.

Since uDu_{D} and uNu_{N} are linearly independent, then we have

|ADBNANBD|>0.|A_{D}B_{N}-A_{N}B_{D}|>0.

Hence we can prove (5.4)\eqref{horizon wronskian}. ∎

6 Proof of Theorem 4.6 for Dirichlet boundary conditions

In this section, we will prove Theorem 4.6 under Dirichlet boundary conditions. We use the notation ΩM2\Omega_{M}^{2} to emphasize the role of the parameter MM in the following argument.

6.1 Energy functional and the twisted energy functional

Recall that we can write the stationary Klein–Gordon equation (1.1)\eqref{Klein-Gordon} as

ddr(r2ΩM2dϕdr)+(Λ3)αr2ϕq02A2ΩM2r2ϕ=0,-\frac{d}{dr}(r^{2}\Omega_{M}^{2}\frac{d\phi}{dr})+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha r^{2}\phi-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}r^{2}\phi=0, (6.1)

where AA takes the form of

A=e(1r+1r).A=-e\left(\frac{1}{r_{+}}-\frac{1}{r}\right).

The corresponding energy functional is:

LM[f]=r+r2ΩM2(dfdr)2+((Λ3)αq02A2ΩM2)r2f2dr.L_{M}[f]=\int_{r_{+}}^{\infty}r^{2}\Omega_{M}^{2}\left(\frac{df}{dr}\right)^{2}+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}\mathrm{d}r. (6.2)

We call the term

VM(r):=(Λ3)αq02A2ΩM2V_{M}(r):=\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}

defined in the above energy functional the potential term. One can see VM(r)V_{M}(r) is always negative on (r+,)(r_{+},\infty). To overcome the difficulties mentioned in Section 4.2, we introduce the twisted derivative. Let rh\nabla_{r}^{h} be the twisted derivative

rhf=hddr(h1f),\nabla_{r}^{h}f=h\frac{d}{dr}(h^{-1}f),

where function hh is called the twist function. The dual operator of this twisted derivative operator is h1ddr(hf)h^{-1}\frac{d}{dr}(hf). We can rewrite the equation (6.1)\eqref{rkeq} with respect to the twisted derivative as

h1ddr(r2ΩM2h2dh1ϕdr)+(VM,h(r)q02A2ΩM2)r2ϕ=0,-h^{-1}\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}h^{2}\frac{dh^{-1}\phi}{dr}\right)+\left(V_{M,h}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi=0, (6.3)

where the twisted potential VM,h(r)V_{M,h}(r) is

VM,h(r):=(Λ3)α1r2hddr(r2ΩM2dhdr)V_{M,h}(r):=\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{1}{r^{2}h}\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}\frac{dh}{dr}\right) (6.4)

We can also define the twisted energy functional by

LM,h[f]:=r+r2ΩM2h2(dh1fdr)2+(VM,h(r)q02A2ΩM2)r2f2dr.L_{M,h}[f]:=\int_{r_{+}}^{\infty}r^{2}\Omega_{M}^{2}h^{2}\left(\frac{dh^{-1}f}{dr}\right)^{2}+\left(V_{M,h}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}dr. (6.5)

In the proof of Theorem 4.6 with Dirichlet boundary conditions, we choose h=rβh=r^{-\beta}. Let VM,βV_{M,\beta} denote the potential function and LM,βL_{M,\beta} denote the twisted energy functional when h=rβh=r^{-\beta}, then we have

VM,β(r)=βr1dΩM2dr+β(1β)r2ΩM2+(Λ3)α.V_{M,\beta}(r)=\beta r^{-1}\frac{d\Omega_{M}^{2}}{dr}+\beta(1-\beta)r^{-2}\Omega_{M}^{2}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha. (6.6)

We can prove that the twisted energy functional is equivalent to the original energy functional.

Lemma 6.1.

If fCc(r+,)f\in C^{\infty}_{c}(r_{+},\infty), then LM[f]=LMh[f]L_{M}[f]=L_{M}^{h}[f] for any smooth function hh.

Proof.

This can be proved by direct computation. We have

LM,h[f]\displaystyle L_{M,h}[f]
=\displaystyle= r+r2h2ΩM2(dh1fdr)2+(VM,hq02A2ΩM2)r2f2dr\displaystyle\int_{r_{+}}^{\infty}r^{2}h^{2}\Omega_{M}^{2}\left(\frac{dh^{-1}f}{dr}\right)^{2}+\left(V_{M,h}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}dr
=\displaystyle= r+r2ΩM2(dfdr)2+r2ΩM2h2(dhdr)2f2r2ΩM2h1dhdrdf2dr+(VM,h(r)q02A2ΩM2)r2f2dr\displaystyle\int_{r_{+}}^{\infty}r^{2}\Omega_{M}^{2}\left(\frac{df}{dr}\right)^{2}+r^{2}\Omega_{M}^{2}h^{-2}\left(\frac{dh}{dr}\right)^{2}f^{2}-r^{2}\Omega_{M}^{2}h^{-1}\frac{dh}{dr}\frac{df^{2}}{dr}+\left(V_{M,h}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}dr
=\displaystyle= r+r2ΩM2(dfdr)2+(VM,h(r)q02A2ΩM2+ΩM2h2(dhdr)2+1r2ddr(r2ΩM2h1dhdr))r2f2dr\displaystyle\int_{r_{+}}^{\infty}r^{2}\Omega_{M}^{2}\left(\frac{df}{dr}\right)^{2}+\left(V_{M,h}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}+\Omega_{M}^{2}h^{-2}\left(\frac{dh}{dr}\right)^{2}+\frac{1}{r^{2}}\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}h^{-1}\frac{dh}{dr}\right)\right)r^{2}f^{2}dr
=\displaystyle= LM[f].\displaystyle L_{M}[f].

The third identity is due to the integration by parts and the fact that ff is compactly supported. ∎

Since ΩM2\Omega_{M}^{2} degenerates at the event horizon, we define the perturbed energy functional LMϵ[f]L^{\epsilon}_{M}[f] to be

LMϵ[f]=r++ϵr2ΩM2(dfdr)2+((Λ3)αq02A2Ω2)r2f2dr.L^{\epsilon}_{M}[f]=\int_{r_{+}+\epsilon}^{\infty}r^{2}\Omega_{M}^{2}(\frac{df}{dr})^{2}+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)r^{2}f^{2}\mathrm{d}r. (6.7)

Similarly, we can define the perturbed twisted energy functional

LM,βϵ[f]=r++ϵr2β+2ΩM2(drβfdr)2+r2(VM,β(r)q02A2ΩM2)f2dr.L_{M,\beta}^{\epsilon}[f]=\int_{r_{+}+\epsilon}^{\infty}r^{-2\beta+2}\Omega_{M}^{2}\left(\frac{dr^{\beta}f}{dr}\right)^{2}+r^{2}\left(V_{M,\beta}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)f^{2}\mathrm{d}r. (6.8)

6.2 Negative energy bound state

Recall the temperature of the event horizon

T:=dΩM2dr(r+)=2r+2(M0M).T:=\frac{d\Omega_{M}^{2}}{dr}(r_{+})=\frac{2}{r_{+}^{2}}(M_{0}-M). (6.9)

For fixed parameters (r+,Λ,α)(r_{+},\Lambda,\alpha), the horizon temperature TT is determined by the black hole mass MM.

We say LM[f]L_{M}[f] has a negative energy bound state if there exists fCc(r+,)f\in C^{\infty}_{c}(r_{+},\infty) such that LM[f]<0L_{M}[f]<0. Let

𝒜s={Me=0<M<s,fCc(r+,) such that LM[f]<0}\mathcal{A}_{s}=\{M_{e=0}<M<s,\ \exists f\in C_{c}^{\infty}(r_{+},\infty)\text{ such that }L_{M}[f]<0\}

be the set of all admissible MM such that LML_{M} admits a negative energy bound state. Due to Lemma 6.1, if LM[f]L_{M}[f] has a negative energy bound state, so does LM,h[f]L_{M,h}[f] for any smooth function hh. If charge |q0||q_{0}| can be taken to be large, then the negative energy bound state follows trivially. We can prove the following lemma.

Lemma 6.2.

For any fixed sub-extremal parameters (Mb,r+,Λ,α)(M_{b},r_{+},\Lambda,\alpha) satisfying the bound 94<α<0-\frac{9}{4}<\alpha<0, there exists a q1(Mb,r+,Λ,α)>0q_{1}(M_{b},r_{+},\Lambda,\alpha)>0 such that for any |q0|>q1|q_{0}|>q_{1}, we can find the negative energy bound state for the functional LMb[f]L_{M_{b}}[f].

Proof.

Let η1\eta\equiv 1 on (r++14,r++34)(r_{+}+\frac{1}{4},r_{+}+\frac{3}{4}) be a smooth function defined on \mathbb{R} with support on (r+,r++1)(r_{+},r_{+}+1). Let

a1(Mb,r+,Λ,α):=r+r2ΩMb2(dηdr)2+(Λ3)αr2η2dr,\displaystyle a_{1}(M_{b},r_{+},\Lambda,\alpha):=\int_{r_{+}}^{\infty}r^{2}\Omega_{M_{b}}^{2}\left(\frac{d\eta}{dr}\right)^{2}+\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha r^{2}\eta^{2}\mathrm{d}r,
a2(Mb,r+,Λ,α):=r+A2ΩMb2r2η2dr>0.\displaystyle a_{2}(M_{b},r_{+},\Lambda,\alpha):=\int_{r_{+}}^{\infty}\frac{A^{2}}{\Omega_{M_{b}}^{2}}r^{2}\eta^{2}\mathrm{d}r>0.

If we set q12=a1a2q_{1}^{2}=\frac{a_{1}}{a_{2}}, then for any q02>q12q_{0}^{2}>q_{1}^{2}, we have the negative energy bound state. ∎

For the second case in Theorem 4.6 where the sufficient largeness of q0q_{0} is missing, the existence of a negative energy bound state becomes more subtle. The numerical method in physics literature [29] fails to find a negative energy bound state for LM[f]L_{M}[f] when Me=0<M<M0M_{e=0}<M<M_{0} and q0=0q_{0}=0. Next, by using a sharp Hardy-type inequality and continuity argument, we prove the existence of a negative energy bound state for the general fixed charge case, which paves the way to prove Theorem 4.6.

Lemma 6.3.

[Sharp Hardy-type Inequality] Assume fCc(r+,r++1)f\in C_{c}^{\infty}(r_{+},r_{+}+1), then we have

r+r++1f2(r)dr4r+r++1(rr+)2(f)2dr.\int_{r_{+}}^{r_{+}+1}f^{2}(r)\mathrm{d}r\leq 4\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}(f^{\prime})^{2}\mathrm{d}r.

The constant 44 in the inequality is sharp in the sense that for any δ>0\delta>0 small, we can find a fδCc(r+,r++1)f_{\delta}\in C_{c}^{\infty}(r_{+},r_{+}+1) such that

r+r++1fδ2dr(4δ)r+r++1(rr+)2(fδ)2dr.\int_{r_{+}}^{r_{+}+1}f_{\delta}^{2}\mathrm{d}r\geq(4-\delta)\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}(f_{\delta}^{\prime})^{2}\mathrm{d}r.
Proof.

By the integration by parts and Hölder inequality, we have

r+r++1f2dr\displaystyle\int_{r_{+}}^{r_{+}+1}f^{2}\mathrm{d}r =(rr+)f2(r)|r+r++1r+r++12(rr+)f(r)f(r)dr\displaystyle=(r-r_{+})f^{2}(r)\Bigl{|}_{r_{+}}^{r_{+}+1}-\int_{r_{+}}^{r_{+}+1}2(r-r_{+})f(r)f^{\prime}(r)\mathrm{d}r
2(r+r++1f2dr)12(r+r++1(rr+)2(f)2dr)12.\displaystyle\leq 2\left(\int_{r_{+}}^{r_{+}+1}f^{2}\mathrm{d}r\right)^{\frac{1}{2}}\left(\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}(f^{\prime})^{2}\mathrm{d}r\right)^{\frac{1}{2}}.

Thus we have

r+r++1f2dr4r+r++1(rr+)2(dfdr)2dr.\int_{r_{+}}^{r_{+}+1}f^{2}\mathrm{d}r\leq 4\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}\left(\frac{df}{dr}\right)^{2}\mathrm{d}r.

Next we prove the constant 44 here is sharp. For ϵ>0\epsilon>0 and 12<α1<α2<0-\frac{1}{2}<\alpha_{1}<\alpha_{2}<0, let fϵ,α1,α2f_{\epsilon,\alpha_{1},\alpha_{2}} be

fϵ,α1,α2={A1(rr+)α1+A2(rr+)α2(A1+A2),rr+>ϵ,0,0<rr+ϵ,f_{\epsilon,\alpha_{1},\alpha_{2}}=\left\{\begin{aligned} &A_{1}(r-r_{+})^{\alpha_{1}}+A_{2}(r-r_{+})^{\alpha_{2}}-(A_{1}+A_{2}),\quad r-r_{+}>\epsilon,\\ &0,\quad 0<r-r_{+}\leq\epsilon,\end{aligned}\right. (6.10)

where A1=ϵα2α1ϵα1A_{1}=\epsilon^{\alpha_{2}-\alpha_{1}}-\epsilon^{-\alpha_{1}} and A2=(1ϵα1)A_{2}=-(1-\epsilon^{-\alpha_{1}}). We have

r+r++1fϵ,α1,α22dr=r++ϵr++1fϵ,α1,α22dr\displaystyle\int_{r_{+}}^{r_{+}+1}f_{\epsilon,\alpha_{1},\alpha_{2}}^{2}\mathrm{d}r=\int_{r_{+}+\epsilon}^{r_{+}+1}f_{\epsilon,\alpha_{1},\alpha_{2}}^{2}\mathrm{d}r
=\displaystyle= A12(1ϵ2α1+12α1+121ϵα1+1α1+1+1ϵ)+A22(1ϵ2α2+12α2+121ϵα2+1α2+1+1ϵ)\displaystyle A_{1}^{2}\left(\frac{1-\epsilon^{2\alpha_{1}+1}}{2\alpha_{1}+1}-2\frac{1-\epsilon^{\alpha_{1}+1}}{\alpha_{1}+1}+1-\epsilon\right)+A_{2}^{2}\left(\frac{1-\epsilon^{2\alpha_{2}+1}}{2\alpha_{2}+1}-2\frac{1-\epsilon^{\alpha_{2}+1}}{\alpha_{2}+1}+1-\epsilon\right)
+2A1A2(1ϵα1+α2+1α1+α2+121ϵα1+1α1+121ϵα2+1α2+1+1ϵ),\displaystyle+2A_{1}A_{2}\left(\frac{1-\epsilon^{\alpha_{1}+\alpha_{2}+1}}{\alpha_{1}+\alpha_{2}+1}-2\frac{1-\epsilon^{\alpha_{1}+1}}{\alpha_{1}+1}-2\frac{1-\epsilon^{\alpha_{2}+1}}{\alpha_{2}+1}+1-\epsilon\right),

and

r+r++1(rr+)2(dfϵ,α1,α2dr)2dr\displaystyle\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}\bigl{(}\frac{df_{\epsilon,\alpha_{1},\alpha_{2}}}{dr}\bigr{)}^{2}\mathrm{d}r
=\displaystyle= A12α121ϵ2α1+12α1+1+A22α221ϵ2α2+12α2+1+2A1A2α1α21ϵα1+α2+1α1+α2+1.\displaystyle A_{1}^{2}\alpha_{1}^{2}\frac{1-\epsilon^{2\alpha_{1}+1}}{2\alpha_{1}+1}+A_{2}^{2}\alpha_{2}^{2}\frac{1-\epsilon^{2\alpha_{2}+1}}{2\alpha_{2}+1}+2A_{1}A_{2}\alpha_{1}\alpha_{2}\frac{1-\epsilon^{\alpha_{1}+\alpha_{2}+1}}{\alpha_{1}+\alpha_{2}+1}.

Hence if we choose ϵ\epsilon small and α1,α2\alpha_{1},\alpha_{2} close to 12-\frac{1}{2}, we have

r+r++1fϵ,α1,α22dr(42δ)r+r++1(rr+)2(dfϵ,α1,α2dr)2dr.\int_{r_{+}}^{r_{+}+1}f_{\epsilon,\alpha_{1},\alpha_{2}}^{2}\mathrm{d}r\geq(4-2\delta)\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}\left(\frac{df_{\epsilon,\alpha_{1},\alpha_{2}}}{dr}\right)^{2}\mathrm{d}r.

Note by the construction of fϵ,α1,α2f_{\epsilon,\alpha_{1},\alpha_{2}}, we know that fϵ,α1,α2f_{\epsilon,\alpha_{1},\alpha_{2}} is continuous on [r+,r++1][r_{+},r_{+}+1] and smooth on (r++ϵ,r++1)(r_{+}+\epsilon,r_{+}+1). Hence fϵ,α1,α2H01(r++ϵ,r++1)f_{\epsilon,\alpha_{1},\alpha_{2}}\in H_{0}^{1}(r_{+}+\epsilon,r_{+}+1). Then we can approximate fϵ,α1,α2f_{\epsilon,\alpha_{1},\alpha_{2}} by fδCc(r++ϵ,r++1)f_{\delta}\in C_{c}(r_{+}+\epsilon,r_{+}+1), we have fδCc(r+,r++1)f_{\delta}\in C_{c}^{\infty}(r_{+},r_{+}+1) such that

r+r++1(fδ)2dr(4δ)r+r++1(rr+)2(dfδdr)2dr.\int_{r_{+}}^{r_{+}+1}(f_{\delta})^{2}\mathrm{d}r\geq(4-\delta)\int_{r_{+}}^{r_{+}+1}(r-r_{+})^{2}\left(\frac{df_{\delta}}{dr}\right)^{2}\mathrm{d}r.

Remark 6.4.

Note the inequality we got above is scaling invariant. So we can get the same result with the same sharp constant if we change the domain in the setting to be [r+,r++a][r_{+},r_{+}+a] for any a>0a>0. In the later discussion, aa will be chosen to be a small number.

Now we are ready to prove the following lemma showing that for fixed q0q_{0} and the parameters (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}) satisfying conditions (4.1) and (4.2), the set 𝒜M0\mathcal{A}_{M_{0}} is non-empty.

Lemma 6.5.

For each fixed parameters(r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}) satisfying

94<α<min{0,32+q022(Λ3)},\displaystyle-\frac{9}{4}<\alpha<\min\left\{0,-\frac{3}{2}+\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right\}, (6.11)
(Λ3)r+2>R0,\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}>R_{0}, (6.12)

where R0R_{0} is the positive root of (1.6), 𝒜M0\mathcal{A}_{M_{0}} is non-empty.

Proof.

For fixed (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}), if we can find the negative energy bound state for LM0[f]L_{M_{0}}[f], then by continuity of LM[f]L_{M}[f] with respect to MM for fixed fCc(r+,)f\in C_{c}(r_{+},\infty), we can find the negative energy bound state for LM[f]L_{M}[f] with MM in the neighborhood of M0M_{0}.

Calculating LM[f]L_{M}[f] with M=M0M=M_{0}, we have

LM0[f]\displaystyle L_{M_{0}}[f] =r+(Λ3)(rr+)2(r2+2r+r+3r+2+1(Λ3))(dfdr)2+((Λ3)αq02A2Ω2)r2f2dr.\displaystyle=\int_{r_{+}}^{\infty}\bigl{(}-\frac{\Lambda}{3}\bigr{)}(r-r_{+})^{2}\left(r^{2}+2r_{+}r+3r_{+}^{2}+\frac{1}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)\left(\frac{df}{dr}\right)^{2}+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega^{2}}\right)r^{2}f^{2}\mathrm{d}r.

By Lemma 6.3, we can find fδCc[r+,)f_{\delta}\in C_{c}^{\infty}[r_{+},\infty) with fδ(r)=0f_{\delta}(r)=0 for any r>r++ϵr>r_{+}+\epsilon such that

r+r++ϵfδ2dr(4δ)r+r++ϵ(rr+)2(dfδdr)2dr.\int_{r_{+}}^{r_{+}+\epsilon}f_{\delta}^{2}\mathrm{d}r\geq(4-\delta)\int_{r_{+}}^{r_{+}+\epsilon}(r-r_{+})^{2}\left(\frac{df_{\delta}}{dr}\right)^{2}\mathrm{d}r.

Then we have the following estimate for LM0[fδ]L_{M_{0}}[f_{\delta}]:

LM[fδ]\displaystyle L_{M}[f_{\delta}]
<\displaystyle< (Λ3)r+r++ϵ(rr+)2(r2+2r+r+3r+2+1(Λ3))(dfδdr)2dr\displaystyle\left(-\frac{\Lambda}{3}\right)\int_{r_{+}}^{r_{+}+\epsilon}(r-r_{+})^{2}\left(r^{2}+2r_{+}r+3r_{+}^{2}+\frac{1}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)\left(\frac{df_{\delta}}{dr}\right)^{2}\mathrm{d}r
+(Λ3)(4δ)r+2r+r++ϵ(αq02(Λ3)1+3(Λ3)r+21+6(Λ3)r+2)(rr+)2(dfδdr)2dr\displaystyle+\left(-\frac{\Lambda}{3}\right)(4-\delta)r_{+}^{2}\int_{r_{+}}^{r_{+}+\epsilon}\left(\alpha-\frac{q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\frac{1+3\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}}{1+6\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}}\right)(r-r_{+})^{2}\left(\frac{df_{\delta}}{dr}\right)^{2}\mathrm{d}r
<\displaystyle< (Λ3)r+2r+r++ϵ(rr+)2(6+1(Λ3)r+2\displaystyle\left(-\frac{\Lambda}{3}\right)r_{+}^{2}\int_{r_{+}}^{r_{+}+\epsilon}(r-r_{+})^{2}\left(6+\frac{1}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}}\right.
4(αq022(Λ3)q022(Λ3)11+6(Λ3)r+2)+Cϵ+Cδ)(dfδdr)2dr\displaystyle\left.-4\left(\alpha-\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}-\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\frac{1}{1+6\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}}\right)+C\epsilon+C\delta\right)\left(\frac{df_{\delta}}{dr}\right)^{2}\mathrm{d}r
=\displaystyle= (Λ3)r+2r+r++ϵ(rr+)2(4(α32q022(Λ3))\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}\int_{r_{+}}^{r_{+}+\epsilon}(r-r_{+})^{2}\left(-4\left(\alpha-\frac{3}{2}-\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\right)\right.
+1(Λ3)r+2+2q02(Λ3)11+6(Λ3)r+2+Cϵ+Cδ)(dfδdr)2dr.\displaystyle\left.+\frac{1}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}}+\frac{2q_{0}^{2}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}}\frac{1}{1+6\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}}+C\epsilon+C\delta\right)\left(\frac{df_{\delta}}{dr}\right)^{2}\mathrm{d}r.

Then by the conditions (4.1) and (4.2), we can choose ϵ\epsilon and δ\delta small such that the right hand side of the above inequality is negative.

Now we have LM0[fδ]<0L_{M_{0}}[f_{\delta}]<0. By the continuity, we can prove 𝒜M0\mathcal{A}_{M_{0}} is non-empty. ∎

6.3 Minimizer of the energy functional LMϵ[f]L_{M}^{\epsilon}[f]

We consider the following function class \mathcal{F}:

:={fCc(r+,),r+r2f2Ω2dr=1.}.\mathcal{F}:=\left\{f\in C^{\infty}_{c}(r_{+},\infty),\quad\int_{r_{+}}^{\infty}\frac{r^{2}f^{2}}{\Omega^{2}}\mathrm{d}r=1.\right\}.

We can define the perturbed function class ϵ\mathcal{F}^{\epsilon} by changing the domain in the above definition to be (r++ϵ,)(r_{+}+\epsilon,\infty). Since fϵf\in\mathcal{F}^{\epsilon} is compactly supported, we still have LMϵ[f]=LM,βϵ[f]L_{M}^{\epsilon}[f]=L_{M,\beta}^{\epsilon}[f], similar to Lemma 6.1.

At first glimpse one may suspect whether the minimum of LMϵ[f]L_{M}^{\epsilon}[f] can be attained in ϵ\mathcal{F}^{\epsilon} since LMϵL_{M}^{\epsilon} contains the integral of r2f2r^{2}f^{2} while ϵ\mathcal{F}^{\epsilon} only makes the restriction on the L2L^{2} norm of the function near infinity. However, we can prove the following proposition.

Proposition 6.6.

For Me=0<M<M0M_{e=0}<M<M_{0}, if LM[f]L_{M}[f] has a negative energy bound state, then for any ϵ>0\epsilon>0 small enough, LMϵ[f]L_{M}^{\epsilon}[f] can attain its negative minimum in the function class ϵ\mathcal{F}^{\epsilon}.

Proof.

In this proof, we use CC to denote the constant independent of ϵ\epsilon. By Lemma 6.1 and the continuity of LM,32ϵ[f]L_{M,\frac{3}{2}}^{\epsilon}[f] with respect to ϵ\epsilon, we know that for ϵ\epsilon small enough, we can also find a negative energy bound state for LM,32ϵ[f]L_{M,\frac{3}{2}}^{\epsilon}[f]. It remains to prove that LM,32ϵL_{M,\frac{3}{2}}^{\epsilon} can attain its minimum in ϵ\mathcal{F}^{\epsilon}. We have

LM,32ϵ[f]\displaystyle L_{M,\frac{3}{2}}^{\epsilon}[f] =r++ϵr1ΩM2(dr32fdr)2+(VM,32(r)q02A2ΩM2)r2f2dr,\displaystyle=\int_{r_{+}+\epsilon}^{\infty}r^{-1}\Omega_{M}^{2}\left(\frac{dr^{\frac{3}{2}}f}{dr}\right)^{2}+\left(V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}\mathrm{d}r,
VM,32(r)\displaystyle V_{M,\frac{3}{2}}(r) =(Λ3)(94+α)34r2+9M2r315e24r4.\displaystyle=\left(-\frac{\Lambda}{3}\right)\left(\frac{9}{4}+\alpha\right)-\frac{3}{4r^{2}}+\frac{9M}{2r^{3}}-\frac{15e^{2}}{4r^{4}}.

Since LM[f]L_{M}[f] has a negative energy bound state, the term

VM,32(r)q02A2ΩM2V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}

must have negative values. However, by considering the limit of this term when rr\rightarrow\infty, we have

limrVM,32(r)q02A2ΩM2=(Λ3)(94+α)>0.\lim_{r\rightarrow\infty}V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}=\left(-\frac{\Lambda}{3}\right)\left(\frac{9}{4}+\alpha\right)>0.

Hence there exists a xvx_{v} such that VM,32q02A2ΩM2V_{M,\frac{3}{2}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}} is positive for r>xvr>x_{v}.

Then we have

LM,32ϵ[f]\displaystyle L_{M,\frac{3}{2}}^{\epsilon}[f]- r++ϵxv(VM,32(r)q02A2ΩM2)r2f2dr\displaystyle\int_{r_{+}+\epsilon}^{x_{v}}\left(V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}\mathrm{d}r (6.13)
=r++ϵr1ΩM2(dr32fdr)2dr+xv(VM,32(r)q02A2ΩM2)r2f2dr\displaystyle=\int_{r_{+}+\epsilon}^{\infty}r^{-1}\Omega_{M}^{2}\left(\frac{dr^{\frac{3}{2}}f}{dr}\right)^{2}\mathrm{d}r+\int_{x_{v}}^{\infty}\left(V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}\mathrm{d}r

Since on (r+,xv)(r_{+},x_{v})

|VM,32(r)q02A2ΩM2|C,\left|V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right|\leq C,

we have

LM,32ϵ[f]r++ϵxv(VM,32(r)q02A2ΩM2)r2f2drLM,32ϵ[f]+Cr++ϵxvr2f2dr.L_{M,\frac{3}{2}}^{\epsilon}[f]-\int_{r_{+}+\epsilon}^{x_{v}}\left(V_{M,\frac{3}{2}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}\mathrm{d}r\leq L_{M,\frac{3}{2}}^{\epsilon}[f]+C\int_{r_{+}+\epsilon}^{x_{v}}r^{2}f^{2}\mathrm{d}r.

Since fϵf\in\mathcal{F}^{\epsilon}, we have

r++ϵxvr2f2dr=r++ϵr2ΩM2ΩM2f2drC.\int_{r_{+}+\epsilon}^{x_{v}}r^{2}f^{2}\mathrm{d}r=\int_{r_{+}+\epsilon}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}\Omega_{M}^{2}f^{2}\mathrm{d}r\leq C. (6.14)

And since the right hand side of (6.13)\eqref{core equality1} is positive, we have

LM,32ϵ[f]>C.L_{M,\frac{3}{2}}^{\epsilon}[f]>-C.

Hence we have LM,32ϵL_{M,\frac{3}{2}}^{\epsilon} is lower bounded. Let fnϵϵf_{n}^{\epsilon}\in\mathcal{F}^{\epsilon} be the minimizing sequence of LM,32ϵL_{M,\frac{3}{2}}^{\epsilon}. Without loss of generality, we can assume LM,32ϵ[fnϵ]<0L_{M,\frac{3}{2}}^{\epsilon}[f_{n}^{\epsilon}]<0. Then since LM,32ϵ[fnϵ]L_{M,\frac{3}{2}}^{\epsilon}[f_{n}^{\epsilon}] is bounded, by (6.13)\eqref{core equality1} we know rfnϵrf_{n}^{\epsilon} is L2L^{2} integrable:

r++ϵr2(fnϵ)2drC,\int_{r_{+}+\epsilon}^{\infty}r^{2}(f_{n}^{\epsilon})^{2}\mathrm{d}r\leq C, (6.15)

For r>r++ϵr>r_{+}+\epsilon, we have

ΩM2(r)>Cϵr2.\Omega_{M}^{2}(r)>C\epsilon r^{2}.

Hence we have

r++ϵr(dr32fnϵdr)2drC(ϵ).\displaystyle\int_{r_{+}+\epsilon}^{\infty}r\left(\frac{dr^{\frac{3}{2}}f_{n}^{\epsilon}}{dr}\right)^{2}\mathrm{d}r\leq C(\epsilon).

Then we have fnϵf_{n}^{\epsilon} is H1H^{1} bounded. By Rellich compactness theorem, we have fnϵf_{n}^{\epsilon} weakly converges to ϕϵ\phi^{\epsilon} in H01H_{0}^{1} and strongly converges to ϕϵ\phi^{\epsilon} in L2L^{2} on any compact set K[r++ϵ,)K\subset[r_{+}+\epsilon,\infty). Next, we prove that ϕϵ\phi^{\epsilon} also belongs to ϵ\mathcal{F}^{\epsilon}. By the strong L2L^{2} convergence on a compact set KK, we have

Kr2ΩM2(ϕϵ)2dr1.\int_{K}\frac{r^{2}}{\Omega_{M}^{2}}(\phi^{\epsilon})^{2}\mathrm{d}r\leq 1.

Passing to the limit, we have

r++ϵr2ΩM2(ϕϵ)2dr1.\int_{r_{+}+\epsilon}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}\left(\phi^{\epsilon}\right)^{2}\mathrm{d}r\leq 1.

Assume r++ϵr2ΩM2(ϕϵ)2dr<1\int_{r_{+}+\epsilon}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}(\phi^{\epsilon})^{2}\mathrm{d}r<1, then for any NN, there exist infinite many fnϵf_{n}^{\epsilon} such that

Nr2ΩM2(fnϵ)2dr>12(1r++ϵr2ΩM2(ϕϵ)2dr).\int_{N}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}(f_{n}^{\epsilon})^{2}\mathrm{d}r>\frac{1}{2}\left(1-\int_{r_{+}+\epsilon}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}(\phi^{\epsilon})^{2}\mathrm{d}r\right).

Then by (6.13)\eqref{core equality1}, we have

Cxv(VM,32q02A2ΩM2)r2(fnϵ)2drcN2N(fnϵ)2dr,C\geq\int_{x_{v}}^{\infty}\left(V_{M,\frac{3}{2}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}(f_{n}^{\epsilon})^{2}\mathrm{d}r\geq cN^{2}\int_{N}^{\infty}(f_{n}^{\epsilon})^{2}\mathrm{d}r,

which provides a contradiction if NN is large enough. Hence LM,32ϵ[f]L_{M,\frac{3}{2}}^{\epsilon}[f] can attain its minimum LM,32ϵ[ϕϵ]L_{M,\frac{3}{2}}^{\epsilon}[\phi^{\epsilon}] in ϵ\mathcal{F}^{\epsilon}. ∎

Remark 6.7.

The key point in obtaining the minimizer in the above proof is that the twisted potential VM,32V_{M,\frac{3}{2}} is strictly positive near the infinity. This property allows us to increase the integrability of the functions in the minimizing sequence and prove that ϕϵϵ\phi^{\epsilon}\in\mathcal{F}^{\epsilon}. For the q0=0q_{0}=0 case, we have a more refined description of VM,32V_{M,\frac{3}{2}}: VM,32V_{M,\frac{3}{2}} is negative on (r+,xv)(r_{+},x_{v}) and positive on (xv,)(x_{v},\infty). The presence of the charge q0q_{0} complicates the sign of the term VM,32q02A2ΩM2V_{M,\frac{3}{2}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}} on (r+,xv)(r_{+},x_{v}).

By the constrained variational principle, we can get the Euler–Lagrange equation for ϕϵ\phi^{\epsilon}

ddr(r2ΩM2dϕϵdr)+((Λ3)αq02A2ΩM2)r2ϕϵ=λMϵr2ΩM2ϕϵ,-\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}\frac{d\phi^{\epsilon}}{dr}\right)+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi^{\epsilon}=-\lambda_{M}^{\epsilon}\frac{r^{2}}{\Omega_{M}^{2}}\phi^{\epsilon}, (6.16)

where λMϵ\lambda_{M}^{\epsilon} is the minimum of LM,32ϵL_{M,\frac{3}{2}}^{\epsilon} in ϵ\mathcal{F}^{\epsilon}.

Remark 6.8.

One can still have Lemma 6.6 and estimate (6.15)\eqref{increase the integrability} without assuming the negative energy bound state. The proof follows line by line. However, the negative energy bound state condition allows us to derive the Euler-Lagrange equation with negative eigenvalue λM-\lambda_{M}.

Applying the asymptotic analysis to the solution of (6.16)\eqref{Euler for phi}, asymptotically we have ϕϵr32Δ\phi^{\epsilon}\approx r^{-\frac{3}{2}-\Delta} since ϕϵL2\|\phi^{\epsilon}\|_{L^{2}} is bounded. Moreover, by using the energy estimate, we can get the following uniform bound for ϕϵ\phi^{\epsilon} independent of ϵ\epsilon.

Proposition 6.9.

Let ϕϵ\phi^{\epsilon} be the solution of (6.16)\eqref{Euler for phi} with sub-extremal parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) obtained above. Then ϕϵ\phi^{\epsilon} is asymptotically Cr32ΔCr^{-\frac{3}{2}-\Delta} when rr approaches infinity. Moreover, we have

r++ϵr2+Δ(ϕϵ)2dr<C,\int_{r_{+}+\epsilon}^{\infty}r^{2+\Delta}(\phi^{\epsilon})^{2}\mathrm{d}r<C, (6.17)

where CC is a constant independent of ϵ\epsilon.

Proof.

By the asymptotic analysis of the equation(3.3)\eqref{stationary Klein-Gordon} at r=r=\infty and L2L^{2} boundedness of rϕϵr\phi^{\epsilon}, we know

ϕϵAr32Δ,r>N.\phi^{\epsilon}\approx Ar^{-\frac{3}{2}-\Delta},\quad r>N.

It remains to prove (6.17)\eqref{gain}. Let

ϕϵ=CDϵ(r)uD(r)+CNϵ(r)uN(r),\phi^{\epsilon}=C_{D}^{\epsilon}(r)u_{D}(r)+C_{N}^{\epsilon}(r)u_{N}(r), (6.18)

where uDu_{D} and uNu_{N} are the local basis of the equation (3.3)\eqref{stationary Klein-Gordon} with the same sub-extremal parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}) on (r+,)(r_{+},\infty), defined in Proposition 4.4. By the asymptotic behavior of ϕϵ\phi^{\epsilon} near r=r=\infty, we have

limrCNϵ(r)=0.\lim_{r\rightarrow\infty}C_{N}^{\epsilon}(r)=0.

Substituting (6.18)\eqref{varitional constant} into (6.16)\eqref{Euler for phi}, we have

(CDϵ)(r)uD(r)+(CNϵ)(r)uN(r)\displaystyle(C_{D}^{\epsilon})^{\prime}(r)u_{D}(r)+(C_{N}^{\epsilon})^{\prime}(r)u_{N}(r) =0,\displaystyle=0,
(CDϵ)(r)uD(r)+(CNϵ)(r)uN(r)\displaystyle(C_{D}^{\epsilon})^{\prime}(r)u_{D}^{\prime}(r)+(C_{N}^{\epsilon})^{\prime}(r)u_{N}^{\prime}(r) =λMϵϕϵ(ΩM2)2.\displaystyle=\frac{-\lambda_{M}^{\epsilon}\phi^{\epsilon}}{(\Omega_{M}^{2})^{2}}.

Then we have

CNϵ(r)=rλMϵϕϵuD(ΩM2)2(uNuDuNuD)dr¯.\displaystyle C_{N}^{\epsilon}(r)=\int_{\infty}^{r}\frac{\lambda_{M}^{\epsilon}\phi^{\epsilon}u_{D}}{(\Omega_{M}^{2})^{2}\left(u_{N}u_{D}^{\prime}-u_{N}^{\prime}u_{D}\right)}\mathrm{d}\bar{r}.

Using (5.3)\eqref{infty wronskian} and (5.4)\eqref{horizon wronskian}, we have

|uDuNuDuNuD1(ΩM2)2|Cr32Δ,r.\displaystyle\left|\frac{u_{D}}{u_{N}^{\prime}u_{D}-u_{N}u^{\prime}_{D}}\frac{1}{(\Omega_{M}^{2})^{2}}\right|\leq Cr^{-\frac{3}{2}-\Delta},\quad r\rightarrow\infty.

Hence for r>r++1r>r_{+}+1, we have

|CNϵ(r)|\displaystyle|C_{N}^{\epsilon}(r)| CλMϵr|uDuNuDuNuDϕϵ(ΩM2)2|dr¯\displaystyle\leq C\lambda_{M}^{\epsilon}\int_{r}^{\infty}\left|\frac{u_{D}}{u_{N}^{\prime}u_{D}-u_{N}u^{\prime}_{D}}\frac{\phi^{\epsilon}}{(\Omega_{M}^{2})^{2}}\right|\mathrm{d}\bar{r} (6.19)
CλMϵ(rr¯2(ϕϵ)2dr¯)12(r1r¯2(uDuNuDuNuD)2(1(ΩM2)2)2dr¯)12\displaystyle\leq C\lambda_{M}^{\epsilon}\left(\int_{r}^{\infty}\bar{r}^{2}(\phi^{\epsilon})^{2}\mathrm{d}\bar{r}\right)^{\frac{1}{2}}\left(\int_{r}^{\infty}\frac{1}{\bar{r}^{2}}\left(\frac{u_{D}}{u_{N}^{\prime}u_{D}-u_{N}u^{\prime}_{D}}\right)^{2}\left(\frac{1}{(\Omega_{M}^{2})^{2}}\right)^{2}\mathrm{d}\bar{r}\right)^{\frac{1}{2}}
CλMϵr2Δ,r>r++1.\displaystyle\leq C\lambda_{M}^{\epsilon}r^{-2-\Delta},\quad r>r_{+}+1.

where CC here is a constant independent of ϵ\epsilon. Similarly, for CDϵC_{D}^{\epsilon} and r>r++1r>r_{+}+1, we have

|CDϵ(r)CDϵ(r++1)|C(r++1r(1ruNuNuDuNuD1(ΩM2)2)2dr¯)12C.\left|C_{D}^{\epsilon}(r)-C_{D}^{\epsilon}(r_{+}+1)\right|\leq C\left(\int_{r_{+}+1}^{r}\left(\frac{1}{r}\frac{u_{N}}{u^{\prime}_{N}u_{D}-u_{N}u^{\prime}_{D}}\frac{1}{(\Omega_{M}^{2})^{2}}\right)^{2}\mathrm{d}\bar{r}\right)^{\frac{1}{2}}\leq C. (6.20)

Thus we only need to show that CDϵ(r++1)C_{D}^{\epsilon}(r_{+}+1) is uniformly bounded in ϵ\epsilon. Note that

CDϵ(r)uD(r)=ϕϵCNϵ(r)uN(r).C_{D}^{\epsilon}(r)u_{D}(r)=\phi^{\epsilon}-C_{N}^{\epsilon}(r)u_{N}(r).

If we can find a sequence ϵn\epsilon_{n} such that |CDϵn(r++1)||C_{D}^{\epsilon_{n}}(r_{+}+1)| goes to infinity, then by (6.20)\eqref{large Cd}, |CDϵn(r)||C_{D}^{\epsilon_{n}}(r)| goes to infinity uniformly in rr for r>r++1r>r_{+}+1. Hence we have

r++1(CDϵn)2(r)uD2(r)dr,\int_{r_{+}+1}^{\infty}(C_{D}^{\epsilon_{n}})^{2}(r)u^{2}_{D}(r)\mathrm{d}r\rightarrow\infty,

which is a contradiction since ϕϵ\phi^{\epsilon} and CNuNC_{N}u_{N} are uniformly L2L^{2} bounded in ϵ\epsilon. Therefore we obtain the uniform boundedness of CN(r)C_{N}(r). Hence we can prove (6.17)\eqref{gain}. ∎

Next, we want to get rid of ϵ\epsilon. We can prove the following proposition.

Proposition 6.10.

If LM[f]L_{M}[f] has a negative energy bound state, then there exists a non-zero solution of the equation

ddr(r2ΩM2dϕdr)+((Λ3)αq02A2ΩM2)r2ϕ=λMr2ΩM2ϕ,r+<r<.-\frac{d}{dr}(r^{2}\Omega_{M}^{2}\frac{d\phi}{dr})+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi=-\frac{\lambda_{M}r^{2}}{\Omega_{M}^{2}}\phi,\quad r_{+}<r<\infty. (6.21)

Moreover, ϕ\phi satisfies the Dirichlet boundary condition and can be extended continuously to the event horizon.

Proof.

Let Kn=[r++1n,n]K_{n}=[r_{+}+\frac{1}{n},n]. By (6.13)\eqref{core equality1}, we have

Kn(dϕϵdr)2drC(n).\int_{K_{n}}\left(\frac{d\phi^{\epsilon}}{dr}\right)^{2}\mathrm{d}r\leq C(n).

Thus we have the uniform boundedness for ϕϵ\phi^{\epsilon} on KnK_{n}

ϕϵH1(Kn)C(n).\displaystyle\|\phi^{\epsilon}\|_{H^{1}(K_{n})}\leq C(n). (6.22)

Hence ϕϵ\phi^{\epsilon} weakly converges to ϕ\phi in Hloc1H^{1}_{loc} and strongly converges to ϕ\phi in Lloc2L^{2}_{loc} up to a subsequence. By (6.17)\eqref{gain}, we have

Cr++ϵr2+Δ(ϕϵ)2drNΔNr2(ϕϵ)2dr.C\geq\int_{r_{+}+\epsilon}^{\infty}r^{2+\Delta}(\phi^{\epsilon})^{2}\mathrm{d}r\geq N^{\Delta}\int_{N}^{\infty}r^{2}(\phi^{\epsilon})^{2}\mathrm{d}r.

Then we have

Nr2(ϕϵ)2drCNΔ,\int_{N}^{\infty}r^{2}(\phi^{\epsilon})^{2}\mathrm{d}r\leq\frac{C}{N^{\Delta}}, (6.23)

where NN is a positive constant which will be chosen very large later.

When T>(Λ3)αr+βT>-\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha\frac{r_{+}}{\beta}, we have VM,β(r+)>0V_{M,\beta}(r_{+})>0. Now for our fixed parameters (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}), we can find β0\beta_{0} large enough, such that dΩM2dr(r+)>(Λ3)αr+β0\frac{d\Omega_{M}^{2}}{dr}(r_{+})>-\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha\frac{r_{+}}{\beta_{0}}. Then we have

VM,β0(r+)\displaystyle V_{M,\beta_{0}}(r_{+}) >0,\displaystyle>0,
limrVM,β0(r)\displaystyle\lim_{r\rightarrow\infty}V_{M,\beta_{0}}(r) =(Λ3)(β0(3β0)+α)<0.\displaystyle=\left(-\frac{\Lambda}{3}\right)\left(\beta_{0}(3-\beta_{0})+\alpha\right)<0.

Basic calculation shows that VM,β0V_{M,\beta_{0}} will have exactly one zero point xβ0x_{\beta_{0}}. Hence there exists x~β0>r+\widetilde{x}_{\beta_{0}}>r_{+} such that VM,β0q02A2ΩM2V_{M,\beta_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}} is positive on (r+,x~β0)(r_{+},\widetilde{x}_{\beta_{0}}).

Considering the energy functional LM,β0ϵ[f]L_{M,\beta_{0}}^{\epsilon}[f], we have

LM,β0ϵ[ϕϵ]+x~β0(r2)(VM,β0q02A2ΩM2)(ϕϵ)2dr\displaystyle L^{\epsilon}_{M,\beta_{0}}[\phi^{\epsilon}]+\int_{\widetilde{x}_{\beta_{0}}}^{\infty}(-r^{2})\left(V_{M,\beta_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)(\phi^{\epsilon})^{2}\mathrm{d}r (6.24)
=\displaystyle= r++ϵr2β0+2ΩM2(ddr(rβ0ϕϵ))2dr+r++ϵx~β0r2(VM,β0q02A2ΩM2)(ϕϵ)2dr.\displaystyle\int_{r_{+}+\epsilon}^{\infty}r^{-2\beta_{0}+2}\Omega_{M}^{2}\left(\frac{d}{dr}(r^{\beta_{0}}\phi^{\epsilon})\right)^{2}\mathrm{d}r+\int_{r_{+}+\epsilon}^{\widetilde{x}_{\beta_{0}}}r^{2}\left(V_{M,\beta_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)(\phi^{\epsilon})^{2}\mathrm{d}r.

Since

LM,β0[ϕϵ]=λMϵ,L_{M,\beta_{0}}[\phi^{\epsilon}]=-\lambda^{\epsilon}_{M},

where λMϵ\lambda_{M}^{\epsilon} is bounded above due to the fact that LM[f]L_{M}[f] has a negative energy bound state. Since the term (VM,α0q02A2ΩM2)\left(V_{M,\alpha_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right) on the left hand side of (6.24)\eqref{core2} is bounded, we have

x~α0r2(ϕϵ)2drC1.\int_{\widetilde{x}_{\alpha_{0}}}^{\infty}r^{2}(\phi^{\epsilon})^{2}\mathrm{d}r\geq C_{1}. (6.25)

Combining (6.23)\eqref{c1} and (6.25)\eqref{c2}, we have

xβ0Nr2(ϕϵ)2drC1CNΔ>C12\int_{x_{\beta_{0}}}^{N}r^{2}(\phi^{\epsilon})^{2}\mathrm{d}r\geq C_{1}-\frac{C}{N^{\Delta}}>\frac{C_{1}}{2} (6.26)

by choosing NN large enough.

Thus ϕ\phi is non-zero. By the local convergence result, ϕ\phi satisfies the equation

ddr(r2ΩM2dϕdr)+((Λ3)αq02A2ΩM2)r2ϕ=λMr2ΩM2ϕ.-\frac{d}{dr}(r^{2}\Omega_{M}^{2}\frac{d\phi}{dr})+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi=-\lambda_{M}\frac{r^{2}}{\Omega_{M}^{2}}\phi.

By the asymptotic analysis of the equation (6.16)\eqref{Euler for phi}, we have

ϕ(r)A(rr+)TλM+B,rr+,\displaystyle\phi(r)\approx A(r-r_{+})^{T\sqrt{\lambda_{M}}}+B,\quad r\rightarrow r_{+},
ϕCDr32Δ+CNr32+Δ,r.\displaystyle\phi\approx C_{D}r^{-\frac{3}{2}-\Delta}+C_{N}r^{-\frac{3}{2}+\Delta},\quad r\rightarrow\infty.

Since ϕϵ\phi^{\epsilon} is locally convergent to ϕ\phi on any compact set K[r+,)K\subset[r_{+},\infty), we have

Kr2ΩM2ϕ2<C,\displaystyle\int_{K}\frac{r^{2}}{\Omega_{M}^{2}}\phi^{2}<C,
Kr2ΩM2(dϕdr)2+((Λ3)αq02A2ΩM2)r2ϕ2dr<C,\displaystyle\int_{K}r^{2}\Omega_{M}^{2}(\frac{d\phi}{dr})^{2}+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi^{2}\mathrm{d}r<C,

where CC is a constant independent of KK. Thus we conclude that B=CN=0B=C_{N}=0 and ϕ\phi is the desired solution of (6.16)\eqref{Euler for phi}. ∎

The immediate consequence of Proposition 6.10 is that, eλMtϕe^{\sqrt{\lambda_{M}}t}\phi is a growing mode solution to the uncharged Klein–Gordon equation (1.1) with the mode ω=iλ\omega=-i\sqrt{\lambda}.

Remark 6.11.

By the asymptotic behavior of ϕ\phi, we have

r+r2ΩM2ϕ2dr<.\int_{r_{+}}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}\phi^{2}\mathrm{d}r<\infty.

If we normalize ϕ\phi such that

r+r2ΩM2ϕ2dr=1,\int_{r_{+}}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}\phi^{2}\mathrm{d}r=1,

then we have

r+r2ΩM2(dϕdr)2+((Λ3)αq02A2ΩM2)r2ϕ2dr=λM.\int_{r_{+}}^{\infty}r^{2}\Omega_{M}^{2}\left(\frac{d\phi}{dr}\right)^{2}+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi^{2}\mathrm{d}r=-\lambda_{M}. (6.27)
Remark 6.12.

Note that in the above argument, to show ϕ\phi is non-zero, we need the fact that (6.25)\eqref{c2} is bounded away from 0 to have the lower bound (6.26)\eqref{coervicity}. This step relies on the existence of a negative energy bound state.

Now we are ready to prove Theorem 4.6.

Proof.

By Lemma 6.2, Lemma 6.5, and Proposition 6.6, we only need to find an MM such that λM=0\lambda_{M}=0. First, we derive the monotonicity of λM\lambda_{M} and λMϵ\lambda_{M}^{\epsilon} in terms of MM. For any fCc(r+,)f\in C_{c}^{\infty}(r_{+},\infty) and M1>M2M_{1}>M_{2}, we have

LM1ϵ[f]=\displaystyle L_{M_{1}}^{\epsilon}[f]= r++ϵr2ΩM12(dfdr)2+((Λ3)αq02e12(1r+1r)2ΩM12)r2f2dr\displaystyle\int_{r_{+}+\epsilon}^{\infty}r^{2}\Omega_{M_{1}}^{2}\left(\frac{df}{dr}\right)^{2}+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}e_{1}^{2}\left(\frac{1}{r_{+}}-\frac{1}{r}\right)^{2}}{\Omega_{M_{1}}^{2}}\right)r^{2}f^{2}\mathrm{d}r
=\displaystyle= r++ϵr2(ΩM12ΩM22)(dfdr)2+(q02e22(1r+1r)2ΩM22q02e12(1r+1r)2ΩM12)r2f2dr\displaystyle\int_{r_{+}+\epsilon}^{\infty}r^{2}(\Omega_{M_{1}}^{2}-\Omega_{M_{2}}^{2})\left(\frac{df}{dr}\right)^{2}+\left(\frac{q_{0}^{2}e_{2}^{2}\left(\frac{1}{r_{+}}-\frac{1}{r}\right)^{2}}{\Omega_{M_{2}}^{2}}-\frac{q_{0}^{2}e_{1}^{2}\left(\frac{1}{r_{+}}-\frac{1}{r}\right)^{2}}{\Omega_{M_{1}}^{2}}\right)r^{2}f^{2}\mathrm{d}r
+\displaystyle+ LM2ϵ\displaystyle L_{M_{2}}^{\epsilon}

Since for fixed (r+,Λ)(r_{+},\Lambda) and rr ΩM2\Omega_{M}^{2} is a decreasing function of MM and e2ΩM2\frac{e^{2}}{\Omega_{M}^{2}} is an increasing funciton of MM, we have

LM1ϵLM2ϵ.L_{M_{1}}^{\epsilon}\leq L_{M_{2}}^{\epsilon}.

Passing to the limit, we have λM\lambda_{M} and λMϵ\lambda_{M}^{\epsilon} are increasing functions of MM. By the same computation as above, we can further show that λMϵ\lambda_{M}^{\epsilon} is a Lipschitz function of MM with Lipschitz constant uniform in ϵ\epsilon:

0λM1λM2C(M1M2),0\leq\lambda_{M_{1}}-\lambda_{M_{2}}\leq C(M_{1}-M_{2}),

where CC is a uniform constant independent of ϵ\epsilon and MM. Passing to the limit we get λM\lambda_{M} is also an increasing uniformly Lipschitz function of MM. Let Mc=infM𝒜sM_{c}=\inf_{M}\mathcal{A}_{s}, where s=Mbs=M_{b} if one considers the large charge case and s=M0s=M_{0} if one considers the general fixed charge case. Then we can continuously extend λM\lambda_{M} to M=McM=M_{c} by letting

λMc=limMMcλM0.\lambda_{M_{c}}=\lim_{M\rightarrow M_{c}}\lambda_{M}\geq 0.

If λMc>0\lambda_{M_{c}}>0, then by remark 6.12 and all the construction above, we can find ϕcϵϵ\phi_{c}^{\epsilon}\in\mathcal{F}^{\epsilon} and λMcϵ>0\lambda_{M_{c}}^{\epsilon}>0 such that ϕcϵ\phi^{\epsilon}_{c} is the solution of

ddr(r2ΩMc2dϕcϵdr)+((Λ3α)q02A2ΩMc2)r2ϕcϵ=λMcϵr2ΩMc2ϕcϵ.-\frac{d}{dr}(r^{2}\Omega_{M_{c}}^{2}\frac{d\phi_{c}^{\epsilon}}{dr})+\left(\bigl{(}-\frac{\Lambda}{3}\alpha\bigr{)}-\frac{q_{0}^{2}A^{2}}{\Omega_{M_{c}}^{2}}\right)r^{2}\phi_{c}^{\epsilon}=-\lambda_{M_{c}}^{\epsilon}\frac{r^{2}}{\Omega_{M_{c}}^{2}}\phi_{c}^{\epsilon}.

Then LMcϵ[ϕϵ]<0L_{M_{c}}^{\epsilon}[\phi^{\epsilon}]<0 implies we can find a negative energy bound state for LMcL_{M_{c}}, which means Mc𝒜sM_{c}\in\mathcal{A}_{s}. Since λM\lambda_{M} is a contninuous function of MM, for Mcϵ<M<McM_{c}-\epsilon<M<M_{c} where ϵ\epsilon is a small positive number, we have λM<0\lambda_{M}<0. Then we can apply the above argument in Proposition 6.6 again to show that M𝒜sM\in\mathcal{A}_{s}, which contradicts to the definition of McM_{c}. Therefore λMc=0\lambda_{M_{c}}=0.

Last, we need to construct the corresponding function ϕMc\phi_{M_{c}}. We consider the solution ϕM,λ\phi_{M,\lambda} of the equation

ddr(r2ΩM2dϕdr)+((Λ3)αq02A2ΩM2)r2ϕ+λr2ΩM2ϕ=0-\frac{d}{dr}(r^{2}\Omega_{M}^{2}\frac{d\phi}{dr})+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi+\lambda\frac{r^{2}}{\Omega_{M}^{2}}\phi=0

with ϕ(r)(rr+)Tλ\phi(r)\approx(r-r_{+})^{T\sqrt{\lambda}} and Dirichlet boundary condition. By the asymptotic analysis, we write ϕ\phi as

ϕ(r,M,λ)=A(M,λ)ϕD(r,M,λ)+B(M,λ)ϕN(r,M,λ),\phi(r,M,\lambda)=A(M,\lambda)\phi_{D}(r,M,\lambda)+B(M,\lambda)\phi_{N}(r,M,\lambda),

where

ϕD(r,M,λ)r32Δ,ϕN(r,M,λ)r32+Δ,r.\phi_{D}(r,M,\lambda)\approx r^{-\frac{3}{2}-\Delta},\quad\phi_{N}(r,M,\lambda)\approx r^{-\frac{3}{2}+\Delta},\quad r\rightarrow\infty.

Since B(M,λM)=0B(M,\lambda_{M})=0 for M𝒜sM\in\mathcal{A}_{s}, we have B(Mc,λMc)=0B(M_{c},\lambda_{M_{c}})=0 by continuity. Then

ϕMc:=ϕ(r,Mc,0)\phi_{M_{c}}:=\phi(r,M_{c},0)

is the non-zero solution of 3.3 with Dirichlet boundary condition and can be extended continuously to the event horizon. ∎

7 Proof of Theorem 4.6 for Neumann boundary conditions

In this section, we begin to prove Theorem 4.6 for Neumann boundary conditions. We will elaborate on the new ingredients in this different boundary condition case while omitting some proofs similar to those used in the Dirichlet boundary condition case.

7.1 Outline of the proof

In Section 4.2, we discussed the outline of the proof for Dirichlet boundary conditions. Based on the discussion there and the proofs we used in Section 6, we further discuss the new challenges we will face in the case of Neumann boundary conditions.

First, if we want to apply a similar variational method, the immediate difficulty we will face is that the energy functional LM[f]L_{M}[f] is not even finite for functions with Neumann boundary conditions since functions with Neumann boundary conditions decay slower near the infinity. This can be overcome by using the appropriately designed twisted derivative, which was first raised by Breitenlohner and Freedman [6] and later was used in [55, 19, 41] to study the Klein–Gordon equations on asymptotic AdS spacetimes under Neumann boundary conditions.

Second, to achieve Neumann boundary conditions for the minimizer, we can no longer take ff in Lemma 6.1 to be compactly supported. Then Lemma 6.1 will fail in general since the boundary term generated from using the integration by parts in the proof of Lemma 6.1 is not finite. Recall that in the proof of Proposition 6.10, we need to use a different and in principle equivalent twisted energy functional to show that the limit ϕ\phi is non-trivial. This step has not worked for the Neumann boundary conditions since the failure of Lemma 6.1 for generic twist functions. Hence we have to come up with a more robust method. We overcome this difficulty by constructing a nice twisted energy functional.

Third, as in the case of the Dirichlet boundary conditions, we can only expect local convergence ϕϵϕ\phi^{\epsilon}\rightarrow\phi while boundary conditions concern the behavior of ϕ\phi at r=r=\infty. For the Dirichlet boundary conditions, in the proof of Proposition 6.10, LM[ϕ]L_{M}[\phi] is finite if and only if solution ϕ\phi satisfies the Dirichlet boundary condition. We can obtain the finiteness of the energy functional on any compact set KK and then pass to the limit to get the finiteness of LM[ϕ]L_{M}[\phi]. This strategy does not work for the Neumann boundary conditions, since the suitable twisted energy functional LM,h[ϕ]L_{M,h}[\phi] is finite for functions with Neumann or Dirichlet boundary conditions and we lose information at infinity when applying the local convergence argument. The boundary condition of ϕ\phi is achieved by establishing the uniform bound of ϕϵ\phi^{\epsilon}, analogously to Proposition 6.9

7.2 More general twisted derivatives

In this section we introduce a new function space and the twisted derivatives we will use.

First, to illustrate the idea of dealing with non-integrability of the energy function generated by slow decay nature of function with Neumann boundary conditions, we consider the following naive twisted energy functional.

Let h=r32+Δh=r^{-\frac{3}{2}+\Delta}. We can rewrite the equation (3.3)\eqref{stationary Klein-Gordon} by using the twisted derivative hr(h1)h\partial_{r}(h^{-1}\cdot):

r32Δddr(r2Δ1ΩM2dr32Δϕdr)+(VM,32Δq02A2ΩM2)r2ϕ=0,\displaystyle-r^{\frac{3}{2}-\Delta}\frac{d}{dr}\left(r^{2\Delta-1}\Omega_{M}^{2}\frac{dr^{\frac{3}{2}-\Delta}\phi}{dr}\right)+\left(V_{M,\frac{3}{2}-\Delta}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi=0, (7.1)
VM,32Δ=(32Δ)(Δ12)r2+2(32Δ)2Mr3(32Δ)(52Δ)e2r4.\displaystyle V_{M,\frac{3}{2}-\Delta}=\frac{(\frac{3}{2}-\Delta)(\Delta-\frac{1}{2})}{r^{2}}+\frac{2(\frac{3}{2}-\Delta)^{2}M}{r^{3}}-\frac{(\frac{3}{2}-\Delta)(\frac{5}{2}-\Delta)e^{2}}{r^{4}}. (7.2)

The twisted energy functional for this equation is

LM,32Δ[f]=r+r2Δ1ΩM2(dr32Δfdr)2+(VM,32Δq02A2ΩM2)r2f2dr,L_{M,\frac{3}{2}-\Delta}[f]=\int_{r_{+}}^{\infty}r^{2\Delta-1}\Omega_{M}^{2}\left(\frac{dr^{\frac{3}{2}-\Delta}f}{dr}\right)^{2}+\left(V_{M,\frac{3}{2}-\Delta}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}dr, (7.3)

which is finite for functions with Neumann boundary conditions. The secret in the above choice of the twisted derivative is that the leading order term in the potential is canceled. However, in this situation, the leading term of VM,32ΔV_{M,\frac{3}{2}-\Delta} is comparable to the term q0A2ΩM2\frac{q_{0}A^{2}}{\Omega_{M}^{2}}, which means the positivity of the potential term near r=r=\infty depends on the values of (M,r+,Λ,α,q0)(M,r_{+},\Lambda,\alpha,q_{0}). In view of Remark 6.7, the method in the Dirichlet boundary condition case fails for Neumann boundary conditions. To make our method more robust and cover all ranges of possible α\alpha where the negative energy bound state and local well-posedness hold, we define the following twist function h0h_{0}, which can be viewed as a suitable modification of the twist function r32+Δr^{-\frac{3}{2}+\Delta} preserving its structure at infinity:

h0={r32+Δ(1r1Δ),rr++2,g,r++1<r<r++2,eKr,r+r<r++1,h_{0}=\begin{cases}r^{-\frac{3}{2}+\Delta}(1-r^{-1-\Delta}),\quad&r\geq r_{+}+2,\\ g,\quad&r_{+}+1<r<r_{+}+2,\\ e^{-Kr},\quad&r_{+}\leq r<r_{+}+1,\end{cases} (7.4)

where KK is a large constant which will be chosen later and gg is a smooth function that makes hh smooth on (r+,)(r_{+},\infty). When rr\rightarrow\infty, we have

limrr1+Δ(VM,h(r)q02A2ΩM2)\displaystyle\lim_{r\rightarrow\infty}r^{1+\Delta}\left(V_{M,h}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right) (7.5)
=\displaystyle= limrr1+Δ(1+Δr12+Δ(1r1Δ)ddr(r32ΩM2)+(32Δ)(1+Δ)r12+Δ(1r1Δ)r52ΩM2)\displaystyle\lim_{r\rightarrow\infty}r^{1+\Delta}\left(-\frac{1+\Delta}{r^{\frac{1}{2}+\Delta}(1-r^{-1-\Delta})}\frac{d}{dr}\left(r^{-\frac{3}{2}}\Omega_{M}^{2}\right)+\frac{(\frac{3}{2}-\Delta)(1+\Delta)}{r^{\frac{1}{2}+\Delta}(1-r^{-1-\Delta})}r^{-\frac{5}{2}}\Omega_{M}^{2}\right)
=\displaystyle= (α54)(Λ3)>0.\displaystyle\left(-\alpha-\frac{5}{4}\right)\left(-\frac{\Lambda}{3}\right)>0.

When rr+r\rightarrow r_{+}, we have

limrr+VM,h0(r)q02A2ΩM2=(Λ3)α+KT>0\lim_{r\rightarrow r_{+}}V_{M,h_{0}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}=\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha+KT>0 (7.6)

for KK large enough.

To define the new function space, let χ\chi be a smooth function which is 11 on [r+,R)[r_{+},R) and vanishes near infinity. Then we can define N\mathcal{F}_{N}:

N:={fL2(r+,),χfCc(r+,),r+r2ΩM2f2dr=1},\mathcal{F}_{N}:=\left\{f\in L^{2}(r_{+},\infty),\quad\chi f\in C_{c}^{\infty}(r_{+},\infty),\quad\int_{r_{+}}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}f^{2}\mathrm{d}r=1\right\},

fNf\in\mathcal{F}_{N} can be viewed as a function supported away from r=r+r=r_{+}. The difference between N\mathcal{F}_{N} and \mathcal{F} defined in Section 6 is that ff is no longer compactly supported. Consequently, Lemma 6.1 is not true generically since the non-vanishing boundary terms.

7.3 Negative energy bound state and the eigenvalue solution

Since to find the negative energy bound state, we only need to consider LM,h0[f]L_{M,h_{0}}[f] with ff compactly supported. Hence we can immediately get the existence of the negative energy bound state for LM,h0[f]L_{M,h_{0}}[f]. Specifically, we have the following two lemmas, the proofs of which follow line by line from the case of Dirichlet boundary conditions in Lemma 6.2 and Lemma 6.5.

Lemma 7.1.

For any fixed parameters (r+,Λ,α)(r_{+},\Lambda,\alpha) satisfying the bound 94<α<54-\frac{9}{4}<\alpha<-\frac{5}{4}, and for any Me=0<Mb<M0M_{e=0}<M_{b}<M_{0}, there exists a q1(Mb,r+,Λ,α)>0q_{1}(M_{b},r_{+},\Lambda,\alpha)>0 such that for any |q0|>q1|q_{0}|>q_{1}, 𝒜Mb\mathcal{A}_{M_{b}} is non-empty.

Lemma 7.2.

For every fixed parameters (r+,Λ,α,q0)(r_{+},\Lambda,\alpha,q_{0}) satisfying

94<α<{32+q022(Λ3),54},\displaystyle-\frac{9}{4}<\alpha<\left\{-\frac{3}{2}+\frac{q_{0}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}},-\frac{5}{4}\right\}, (7.7)
(Λ3)r+2>R0,\displaystyle\bigl{(}-\frac{\Lambda}{3}\bigr{)}r_{+}^{2}>R_{0}, (7.8)

where R0R_{0} is defined in (1.6), 𝒜M0\mathcal{A}_{M_{0}} is non-empty.

Next, we prove the existence of a minimizer for the energy functional LM,h0ϵ[f]L_{M,h_{0}}^{\epsilon}[f].

Proposition 7.3.

If LM,h0[f]L_{M,h_{0}}[f] has a negative energy bound state, then for any ϵ>0\epsilon>0 small enough, LM,h0ϵ[f]L_{M,h_{0}}^{\epsilon}[f] can attain its negative minimum LM,h0ϵ[ϕϵ]L_{M,h_{0}}^{\epsilon}[\phi^{\epsilon}]in the function class Nϵ\mathcal{F}_{N}^{\epsilon}. Moreover, ϕϵ\phi^{\epsilon} satisfies the Neumann boundary condition.

Proof.

The proof here is almost the same as the proof of Proposition 6.6. Consider the twisted energy functional

LM,h0ϵ[f]=r++ϵr2ΩM2h02(dh01fdr)2+(VM,h0(r)q02A2ΩM2)r2f2dr.L_{M,h_{0}}^{\epsilon}[f]=\int_{r_{+}+\epsilon}^{\infty}r^{2}\Omega_{M}^{2}h_{0}^{2}\left(\frac{dh_{0}^{-1}f}{dr}\right)^{2}+\left(V_{M,h_{0}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}dr.

By the limit (7.5)\eqref{positive near infinity}, we have VM,h0q02A2ΩM2V_{M,h_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}} is positive and asymptotically r1Δr^{-1-\Delta} when rr\rightarrow\infty. By the limit (7.6)\eqref{positive near horizon}, we have VM,h0q02A2ΩM2V_{M,h_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}} is positive near the event horizon. Let x1x_{1} and x2x_{2} be the smallest and largest root of VM,h0q02A2ΩM2V_{M,h_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}} respectively. We have

LM,h0ϵ[f]=\displaystyle L_{M,h_{0}}^{\epsilon}[f]= r++ϵr2ΩM2h02(dh01fdr)2𝑑r+(r++ϵx1+x2)(VM,h0q02A2ΩM2)r2f2dr\displaystyle\int_{r_{+}+\epsilon}^{\infty}r^{2}\Omega_{M}^{2}h_{0}^{2}\left(\frac{dh_{0}^{-1}f}{dr}\right)^{2}dr+\left(\int_{r_{+}+\epsilon}^{x_{1}}+\int_{x_{2}}^{\infty}\right)\left(V_{M,h_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}f^{2}dr (7.9)
x1x2(q02A2ΩM2VM,h0)r2f2𝑑r\displaystyle-\int_{x_{1}}^{x_{2}}\left(\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}-V_{M,h_{0}}\right)r^{2}f^{2}dr
\displaystyle\geq Cx1x2r1Δr2f2𝑑rCx1x2f2𝑑r>C.\displaystyle-C\int_{x_{1}}^{x_{2}}r^{-1-\Delta}r^{2}f^{2}dr\geq-C\int_{x_{1}}^{x_{2}}f^{2}dr>-C.

Hence LM,h0ϵL_{M,h_{0}}^{\epsilon} is lower bounded in Nϵ\mathcal{F}^{\epsilon}_{N}. Let ϕnϵ\phi_{n}^{\epsilon} be the minimizing sequence. Then we have

dϕnϵdrL2C(ϵ).\left\|\frac{d\phi_{n}^{\epsilon}}{dr}\right\|_{L^{2}}\leq C(\epsilon). (7.10)

Therefore ϕnϵ\phi_{n}^{\epsilon} is H1H^{1} bounded. Then we have ϕnϵ\phi_{n}^{\epsilon} is weakly convergent to ϕϵ\phi^{\epsilon} and strongly convergent to ϕϵ\phi^{\epsilon} on any compact set K(r+,)K\subset(r_{+},\infty). Then by (7.9)\eqref{coercivity}, we have in fact

r++ϵr1Δ(ϕnϵ)2𝑑r<C.\int_{r_{+}+\epsilon}^{\infty}r^{1-\Delta}(\phi_{n}^{\epsilon})^{2}dr<C.

Hence by the same argument as in Proposition 6.6, we have

r++ϵr2ΩM2(ϕϵ)2dr=1\int_{r_{+}+\epsilon}^{\infty}\frac{r^{2}}{\Omega_{M}^{2}}(\phi^{\epsilon})^{2}\mathrm{d}r=1

.

Since ϕϵ\phi^{\epsilon} is the minimizer of the energy functional LM,h0ϵ[f]L_{M,h_{0}}^{\epsilon}[f], by variational principle (see Chapter 88 in [20]), we have ϕϵ\phi^{\epsilon} solves the equation

h01ddr(r2ΩM2h02dh01ϕϵdr)+(VM,h0(r)q02A2ΩM2)r2ϕϵ=λMϵr2ΩM2ϕϵ-h_{0}^{-1}\frac{d}{dr}(r^{2}\Omega_{M}^{2}h_{0}^{2}\frac{dh^{-1}_{0}\phi^{\epsilon}}{dr})+\left(V_{M,h_{0}}(r)-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi^{\epsilon}=-\lambda_{M}^{\epsilon}\frac{r^{2}}{\Omega_{M}^{2}}\phi^{\epsilon} (7.11)

with Neumann boundary condition. Last, note we can also write the equation (7.11) in the untwisted form:

ddr(r2ΩM2dϕϵdr)+((Λ3)αq02A2ΩM2)r2ϕϵ=λMϵr2ΩM2ϕϵ.-\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}\frac{d\phi^{\epsilon}}{dr}\right)+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi^{\epsilon}=-\lambda_{M}^{\epsilon}\frac{r^{2}}{\Omega_{M}^{2}}\phi^{\epsilon}. (7.12)

Remark 7.4.

One should note that, the Dirichlet boundary condition in the proof of Proposition 6.6 is obtained by looking at the finite energy. However, since functions in our function class N\mathcal{F}_{N} no longer vanishes near infinity, the variational principle itself gives information about the boundary condition of ϕϵ\phi^{\epsilon}.

We can get rid of ϵ\epsilon in the following proposition:

Proposition 7.5.

If there exists a negative energy bound state of LM,h0[f]L_{M,h_{0}}[f], then we can find a non-zero solution ϕM\phi_{M} of the equation

ddr(r2ΩM2dϕMdr)+((Λ3)αq02A2ΩM2)r2ϕM=λMr2ΩM2ϕM.-\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}\frac{d\phi_{M}}{dr}\right)+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi_{M}=-\lambda_{M}\frac{r^{2}}{\Omega_{M}^{2}}\phi_{M}. (7.13)

Moreover, ϕM\phi_{M} satisfies the Neumann boundary condition and can be extended continuously to the event horizon r=r+r=r_{+}.

Proof.

By using the same argument as in the proof of Proposition 6.10, we can show that ϕϵ\phi^{\epsilon} is weakly convergent to ϕM\phi_{M} in Hloc1H^{1}_{loc} and strongly convergent to ϕM\phi_{M} in Lloc2L^{2}_{loc}. By (7.9)\eqref{coercivity}, we have

x1x2(ϕϵ)2𝑑r>C.\int_{x_{1}}^{x_{2}}(\phi^{\epsilon})^{2}dr>C.

Hence ϕM\phi_{M} is non-trival and solves the equation

h01ddr(r2ΩM2h02dh01ϕdr)+(VM,h0q02A2ΩM2)r2ϕ=λMr2ΩM2ϕ,-h_{0}^{-1}\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}h_{0}^{2}\frac{dh_{0}^{-1}\phi}{dr}\right)+\left(V_{M,h_{0}}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\phi=-\lambda_{M}\frac{r^{2}}{\Omega_{M}^{2}}\phi, (7.14)

which is equivalent to (7.13).

It remains to prove that ϕM\phi_{M} still satisfies the Neumann boundary condition. This part does not follow trivially since ϕM\phi_{M} is obtained by the local convergence and we may lose information near the infinity. For any compact set K(r+,)K\subset(r_{+},\infty), by using the same argument as in Proposition 6.9, we have

|Kr2Δ+1ddr(r32Δϕϵ)r1Δ2dr|<C\left|\int_{K}r^{2\Delta+1}\frac{d}{dr}\left(r^{\frac{3}{2}-\Delta}\phi^{\epsilon}\right)r^{\frac{-1-\Delta}{2}}\mathrm{d}r\right|<C (7.15)

is uniformly bounded in ϵ\epsilon and KK. Then by the weakly convergence results, we have

|Kr2Δ+1ddr(r32ΔϕM)r1Δ2dr|<C.\left|\int_{K}r^{2\Delta+1}\frac{d}{dr}\left(r^{\frac{3}{2}-\Delta}\phi_{M}\right)r^{\frac{-1-\Delta}{2}}\mathrm{d}r\right|<C. (7.16)

However, if ϕ\phi has the Dirichlet branch, then (7.16)\eqref{contraction} can not be uniformly bounded when KK is located near the infinity, which is a contradiction. ∎

As in the Dirichlet boundary condition case, one immediate corollary of Proposition 7.5 is that one can construct growing mode solutions to the uncharged Klein–Gordon equation (1.1) for all Mc<M<M0M_{c}<M<M_{0}.

Now by the exact same argument as in the proof of Theorem 4.6 for Dirichlet boundary conditions, we can finish the proof of Theorem 4.6 for Neumann boundary conditions.

8 Proof of growing mode solution

In this section, we finally close the proof of Theorem 4.1.

8.1 Growing mode solutions for the large charge case and the general fixed charge case under Dirichlet boundary condition

Theorem 8.1.

Let (Mc,r+,Λ,α,q0)(M_{c},r_{+},\Lambda,\alpha,q_{0}) be the parameters where the stationary solutions ϕ0\phi_{0} defined in Theorem 4.6 exist. Then there exists ϵ0>0\epsilon_{0}>0 such that there exist analytic functions Me=0M(ϵ)<M0M_{e=0}\leq M(\epsilon)<M_{0} and ωR(ϵ)\omega_{R}(\epsilon)\in\mathbb{R} for ϵ0<ϵϵ0-\epsilon_{0}<\epsilon\leq\epsilon_{0} with M(0)=McM(0)=M_{c} and ωR(0)=0\omega_{R}(0)=0 such that mode solutions ϕϵ=ei(ωR+iϵ)ψϵ\phi_{\epsilon}=e^{i(\omega_{R}+i\epsilon)}\psi_{\epsilon} to the equation (1.1) with parameters (M(ϵ),r+,Λ,α,q0)(M(\epsilon),r_{+},\Lambda,\alpha,q_{0}) exist under the Dirichlet boundary condition. Moreover, ϕ\phi can be continuously extended to the event horizon {r=r+}\{r=r_{+}\} and

dMdϵ(0)<0,\displaystyle\frac{dM}{d\epsilon}(0)<0, (8.1)
q0edωRdϵ(0)<0,q0e0.\displaystyle q_{0}e\frac{d\omega_{R}}{d\epsilon}(0)<0,\quad q_{0}e\neq 0. (8.2)

We can write the solution of (3.8)\eqref{Klein-Gordon equation for psi} in the form of

ψ=A(M,ω)uD(r,M,ω)+B(M,ω)uN(r,M,ω),\psi=A(M,\omega)u_{D}(r,M,\omega)+B(M,\omega)u_{N}(r,M,\omega), (8.3)

where {uD,uN}\{u_{D},u_{N}\} is the local basis of solutions of (3.8)\eqref{Klein-Gordon equation for psi} with uDu_{D} satisfying the Dirichlet boundary condition and uNu_{N} satisfying the Neumann boundary condition. Furthermore, if ω\omega is a real number while uDu_{D} and uNu_{N} are not real functions, then uD¯\overline{u_{D}} and uN¯\overline{u_{N}} are also two linearly independent solutions and satisfy the same boundary conditions. Hence we can always take uDu_{D} and uNu_{N} to be real if ω\omega is real.

Proof.

Since ϕ0\phi_{0} is a stationary solution, we have B(Mc,0)=0B(M_{c},0)=0. Let ω=ωR+iωI\omega=\omega_{R}+i\omega_{I}, A=AR+iAIA=A_{R}+iA_{I}, and B=BR+iBIB=B_{R}+iB_{I}, then by the implicit function theorem, to prove the existence of growing mode solutions under Dirichlet boundary conditions, we only need to show

det[BRωRBRMBIωRBIM]0.\det\begin{bmatrix}\frac{\partial B_{R}}{\partial\omega_{R}}&\frac{\partial B_{R}}{\partial M}\\ \frac{\partial B_{I}}{\partial{\omega_{R}}}&\frac{\partial B_{I}}{\partial M}\end{bmatrix}\neq 0. (8.4)

For ω=ωR\omega=\omega_{R} real, multiplying ψ¯0\overline{\psi}_{0} and taking the imaginary part, we have

(r2ΩM2dψdrψ¯)(r+)=limr(r2ΩM2dψdrψ¯)(r).\Im\left(r^{2}\Omega_{M}^{2}\frac{d\psi}{dr}\overline{\psi}\right)(r_{+})=\lim_{r\rightarrow\infty}\Im\left(r^{2}\Omega_{M}^{2}\frac{d\psi}{dr}\overline{\psi}\right)(r). (8.5)

By the near horizon and near infinity behaviors of ϕ0=ψ0\phi_{0}=\psi_{0} in (3.12)\eqref{near horizon for psi} and (8.3)\eqref{near infinity of psi}, we have

r+2ωR=limr(Λ3)((AB¯)uDuNr+(A¯B)uNuDr)=(Λ3)2Δ(AIBRARBI).\displaystyle r_{+}^{2}\omega_{R}=\lim_{r\rightarrow\infty}\bigl{(}-\frac{\Lambda}{3}\bigr{)}\left(\Im(A\overline{B})u_{D}\frac{\partial u_{N}}{\partial r}+\Im(\overline{A}B)u_{N}\frac{\partial u_{D}}{\partial r}\right)=\bigl{(}-\frac{\Lambda}{3}\bigr{)}2\Delta(A_{I}B_{R}-A_{R}B_{I}).

Taking the ωR\omega_{R} and MM derivatives respectively and evaluating at (ωR=0,M=Mc)(\omega_{R}=0,M=M_{c}), we have

AIBRωRARBIωR\displaystyle A_{I}\frac{\partial B_{R}}{\partial\omega_{R}}-A_{R}\frac{\partial B_{I}}{\partial\omega_{R}} =r+22(Λ3)Δ,\displaystyle=\frac{r_{+}^{2}}{2\bigl{(}-\frac{\Lambda}{3}\bigr{)}\Delta}, (8.6)
AIBRMARBIM\displaystyle A_{I}\frac{\partial B_{R}}{\partial M}-A_{R}\frac{\partial B_{I}}{\partial M} =0.\displaystyle=0.

Since (AR,AI)(Mc,0)0(A_{R},A_{I})(M_{c},0)\neq 0, (8.4)\eqref{det condition} holds if and only if BM0\frac{\partial B}{\partial M}\neq 0. Assume

BM=0.\frac{\partial B}{\partial M}=0. (8.7)

Then differentiating the equation (3.8)\eqref{Klein-Gordon equation for psi} at (M=Mc,ω=0)(M=M_{c},\omega=0), we have

ddr(r2ΩM2ddrψ0M)+ddr(r2ΩM2Mdψ0dr)=((Λ3)αq02A2ΩM2)r2ψ0MM(q02A2ΩM2)r2ψ0.\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}\frac{d}{dr}\frac{\partial\psi_{0}}{\partial M}\right)+\frac{d}{dr}\left(r^{2}\frac{\partial\Omega_{M}^{2}}{\partial M}\frac{d\psi_{0}}{dr}\right)=\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\frac{\partial\psi_{0}}{\partial M}-\frac{\partial}{\partial M}\left(\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\psi_{0}. (8.8)

We can multiply the above equation (8.8)\eqref{equation for psi M} by ψ¯0\overline{\psi}_{0} and use the integration by parts. Since by our assumption (8.7)\eqref{contradiction argument}, the boundary terms appear in the integration by parts vanish. Hence we have

r+(ddr(r2ΩM2dψ¯0dr)((Λ3)αq02A2ΩM2)r2ψ¯0)ψ0M𝑑r\displaystyle\int_{r_{+}}^{\infty}\left(\frac{d}{dr}\left(r^{2}\Omega_{M}^{2}\frac{d\overline{\psi}_{0}}{dr}\right)-\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}\alpha-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\overline{\psi}_{0}\right)\frac{\partial\psi_{0}}{\partial M}dr (8.9)
=\displaystyle= r+r2ΩM2M|dψ0dr|2M(q02A2ΩM2)r2|ψ|2dr.\displaystyle\int_{r_{+}}^{\infty}r^{2}\frac{\partial\Omega_{M}^{2}}{\partial M}\left|\frac{d\psi_{0}}{dr}\right|^{2}-\frac{\partial}{\partial M}\left(\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}|\psi|^{2}dr.

Since ψ¯0\overline{\psi}_{0} is also the solution of (3.8)\eqref{Klein-Gordon equation for psi}, the left hand side of the above equation is 0. Since Ω2M<0\frac{\partial\Omega^{2}}{\partial M}<0 on (r+,)(r_{+},\infty) for parameters (M,r+,Λ)(M,r_{+},\Lambda), we have

ψ0=dψ0dr=0,\psi_{0}=\frac{d\psi_{0}}{dr}=0,

which contradicts to the fact that ψ0\psi_{0} is non-trivial. Hence BM0\frac{\partial B}{\partial M}\neq 0 and by the implicit function theorem, for ϵ\epsilon small enough, there exist real analytic functions ωR(ϵ)\omega_{R}(\epsilon) and M(ϵ)M(\epsilon) on (ϵ,ϵ)(-\epsilon,\epsilon), such that for parameters (M(ϵ),r+,Λ,α,q0)\left(M(\epsilon),r_{+},\Lambda,\alpha,q_{0}\right), we have the growing mode solution ϕ(r,ϵ)=eiωR(ϵ)teϵtψ(r,ϵ)\phi(r,\epsilon)=e^{i\omega_{R}(\epsilon)t}e^{-\epsilon t}\psi(r,\epsilon) satisfying the Dirichlet boundary condition.

To prove (8.1) and (8.2), differentiating the equation (3.8) with respect to ϵ\epsilon at ϵ=0\epsilon=0, we have

ddr(r2ϵ(ΩM2dψ0dr))+((Λ3)q02A2ΩM2)r2ψ0ϵ=(q02ϵ(A2ΩM2)2q0AΩM2(ωRϵ+i))r2ψ0.-\frac{d}{dr}\left(r^{2}\frac{\partial}{\partial\epsilon}\bigl{(}\Omega_{M}^{2}\frac{d\psi_{0}}{dr}\bigr{)}\right)+\left(\bigl{(}-\frac{\Lambda}{3}\bigr{)}-\frac{q_{0}^{2}A^{2}}{\Omega_{M}^{2}}\right)r^{2}\frac{\partial\psi_{0}}{\partial\epsilon}=\left(q_{0}^{2}\frac{\partial}{\partial\epsilon}\bigl{(}\frac{A^{2}}{\Omega_{M}^{2}}\bigr{)}-\frac{2q_{0}A}{\Omega_{M}^{2}}\bigl{(}\frac{\partial\omega_{R}}{\partial\epsilon}+i\bigr{)}\right)r^{2}\psi_{0}. (8.10)

By the boundary conditions (3.13) and (3.12) for ψ0\psi_{0} at the event horizon, we have

ϵ(ΩM2dψ0dr)ψ0¯(r+)=1+idωRdϵ,\displaystyle\frac{\partial}{\partial\epsilon}\bigl{(}\Omega_{M}^{2}\frac{d\psi_{0}}{dr}\bigr{)}\overline{\psi_{0}}(r_{+})=-1+i\frac{d\omega_{R}}{d\epsilon}, (8.11)
limrr+ΩM2dψ¯0drψ0ϵ(r,0)=0\displaystyle\lim_{r\rightarrow r_{+}}\Omega_{M}^{2}\frac{d\overline{\psi}_{0}}{dr}\frac{\partial\psi_{0}}{\partial\epsilon}(r,0)=0 (8.12)

Since ψ(r,ϵ)\psi(r,\epsilon) satisfies the Dirichlet boundary condition, the boundary terms at the infinity vanish. For the equation (8.10), multiplying ψ¯0\overline{\psi}_{0}, taking the real and imaginary part respectively, and integrating by parts, we have

r+2dωRdϵ=r+2q0AΩM2r2|ψ0(r,0)|2dr,\displaystyle r_{+}^{2}\frac{d\omega_{R}}{d\epsilon}=\int_{r_{+}}^{\infty}\frac{-2q_{0}A}{\Omega_{M}^{2}}r^{2}|\psi_{0}(r,0)|^{2}\mathrm{d}r, (8.13)
r+2=r+r2ΩM2ϵ|dψ0dr|2+(q02ϵ(A2ΩM2)+2q0AωRϵΩM2)r2|ψ0|2dr.\displaystyle r_{+}^{2}=\int_{r_{+}}^{\infty}r^{2}\frac{\partial\Omega_{M}^{2}}{\partial\epsilon}\left|\frac{d\psi_{0}}{dr}\right|^{2}+\left(-q_{0}^{2}\frac{\partial}{\partial\epsilon}\bigl{(}\frac{A^{2}}{\Omega_{M}^{2}}\bigr{)}+\frac{2q_{0}A\frac{\partial\omega_{R}}{\partial\epsilon}}{\Omega_{M}^{2}}\right)r^{2}|\psi_{0}|^{2}\mathrm{d}r. (8.14)

Hence we have

q0edωRdϵ<0,\displaystyle q_{0}e\frac{d\omega_{R}}{d\epsilon}<0,
dMdϵ<0.\displaystyle\frac{dM}{d\epsilon}<0.

8.2 Growing mode solutions for the weakly charged case under Dirichlet boundary condition

The existence of growing mode solutions to the uncharged Klein–Gordon equation under Dirichlet boundary conditions, as stated in the weakly charged case of Theorem 4.1, follows from Proposition 6.10, Proposition 3.4, and the equation (3.8). It remains to show the existence of growing mode solutions near the extremality when q00q_{0}\neq 0 is small. Let McM_{c} and λM\lambda_{M} for Mc<M<M0M_{c}<M<M_{0} as in Theorem 4.6, we have the following result:

Theorem 8.2.

Let ϕ=eλMtψ\phi=e^{\sqrt{\lambda_{M}}t}\psi be the growing mode solution to the uncharged Klein–Gordon equation (1.1) with parameters (M,r+,Λ,α,q0=0)(M,r_{+},\Lambda,\alpha,q_{0}=0) under Dirichlet boundary. Then there exist a small positive number ϵ0>0\epsilon_{0}>0 and real analytic functions ωR(ϵ)\omega_{R}(\epsilon) and ωI(ϵ)<0\omega_{I}(\epsilon)<0 on (ϵ0,ϵ0)(-\epsilon_{0},\epsilon_{0}) with ωI(0)=λM\omega_{I}(0)=-\sqrt{\lambda_{M}} such that a growing mode solution of the form ei(ωR(ϵ)+iωI(ϵ))ψϵe^{i(\omega_{R}(\epsilon)+i\omega_{I}(\epsilon))}\psi_{\epsilon} to (1.1) with parameters (M,r+,Λ,α,ϵ)(M,r_{+},\Lambda,\alpha,\epsilon) under the Dirichlet boundary condition exists.

Proof.

We can write the solution of (3.8)\eqref{Klein-Gordon equation for psi} in the form of

ψ=A(q0,ωR,ωI)uD(r,q0,ωR,ωI)+B(q0,ωR,ωI)uN(r,q0,ωR,ωI),\psi=A(q_{0},\omega_{R},\omega_{I})u_{D}(r,q_{0},\omega_{R},\omega_{I})+B(q_{0},\omega_{R},\omega_{I})u_{N}(r,q_{0},\omega_{R},\omega_{I}), (8.15)

as in Section 8.1. Assume that for parameters (M,r+,Λ,α,q0=0)(M,r_{+},\Lambda,\alpha,q_{0}=0), the uncharged Klein–Gordon equation (1.1) admits a growing mode solution eiωtψe^{i\omega t}\psi with mode ω=iλM\omega=-i\sqrt{\lambda_{M}} under the Dirichlet boundary condition. The boundary conditions for ψ\psi at the event horizon are given by ψ(r+)=0\psi(r_{+})=0 and ΩM2dψdr=iωψ\Omega_{M}^{2}\frac{d\psi}{dr}=i\omega\psi. We can deduce that

ddωI(ΩM2dψdr)ψ¯=ddωR(ΩM2dψdr)ψ¯=0.\frac{d}{d\omega_{I}}(\Omega_{M}^{2}\frac{d\psi}{dr})\bar{\psi}=\frac{d}{d\omega_{R}}(\Omega_{M}^{2}\frac{d\psi}{dr})\bar{\psi}=0. (8.16)

Differentiating the equation (3.8) with ωR\omega_{R} and ωI\omega_{I} respectively at q0=0q_{0}=0 and ωI=λM\omega_{I}=-\sqrt{\lambda_{M}}, taking the imaginary part, and integrating over (r+,)(r_{+},\infty), analogously to the computation in (8.9), we have

ARBIωRAIBRωR=λM(Λ3)Δ,\displaystyle A_{R}\frac{\partial B_{I}}{\partial\omega_{R}}-A_{I}\frac{\partial B_{R}}{\partial\omega_{R}}=\frac{\sqrt{\lambda_{M}}}{\bigl{(}-\frac{\Lambda}{3}\bigr{)}\Delta}, (8.17)
ARBIωIAIBRωI=0.\displaystyle A_{R}\frac{\partial B_{I}}{\partial\omega_{I}}-A_{I}\frac{\partial B_{R}}{\partial\omega_{I}}=0. (8.18)

Analogously to the proof of Theorem 8.1, assuming BωI=0\frac{\partial B}{\partial\omega_{I}}=0, differentiating equation (3.8) with respect to ωI\omega_{I}, multiplying ψ¯\overline{\psi}, and integrating by parts over (r+,)(r_{+},\infty), we can get a contradiction. Hence BωI0\frac{\partial B}{\partial\omega_{I}}\neq 0. Then by the implicit function theorem, we can finish the proof. ∎

8.3 Growing mode solution for Neumann boundary condition

The proof of Theorem 4.1 under Neumann boundary conditions follows almost line by line as in the previous section by considering the twisted equation (3.9) for ψ\psi.

References

  • [1] Lars Andersson and Pieter Blue. Hidden symmetries and decay for the wave equation on the Kerr spacetime. Annals of Mathematics, pages 787–853, 2015.
  • [2] Alain Bachelot. Superradiance and scattering of the charged klein–gordon field by a step-like electrostatic potential. Journal de mathématiques pures et appliquées, 83(10):1179–1239, 2004.
  • [3] Carolina L Benone, Luís CB Crispino, Carlos Herdeiro, and Eugen Radu. Kerr-Newman scalar clouds. Physical Review D, 90(10):104024, 2014.
  • [4] Nicolas Besset. Decay of the local energy for the charged klein–gordon equation in the exterior de sitter–reissner–nordström spacetime. In Annales Henri Poincaré, volume 21, pages 2433–2484. Springer, 2020.
  • [5] Nicolas Besset and Dietrich Häfner. Existence of exponentially growing finite energy solutions for the charged klein–gordon equation on the de sitter–kerr–newman metric. Journal of Hyperbolic Differential Equations, 18(02):293–310, 2021.
  • [6] Peter Breitenlohner and Daniel Z Freedman. Stability in gauged extended supergravity. Annals of physics, 144(2):249–281, 1982.
  • [7] Yves Brihaye, Carlos Herdeiro, and Eugen Radu. Myers–Perry black holes with scalar hair and a mass gap. Physics Letters B, 739:1–7, 2014.
  • [8] Richard Brito, Vitor Cardoso, Paolo Pani, et al. Superradiance, volume 10. Springer, 2020.
  • [9] Otis Chodosh and Yakov Shlapentokh-Rothman. Time-periodic Einstein–Klein–Gordon bifurcations of Kerr. Communications in Mathematical Physics, 356:1155–1250, 2017.
  • [10] Pedro VP Cunha, Carlos AR Herdeiro, Eugen Radu, and Helgi F Rúnarsson. Shadows of Kerr black holes with scalar hair. Physical review letters, 115(21):211102, 2015.
  • [11] Mihalis Dafermos and Gustav Holzegel. Dynamic instability of solitons in 4+ 1 dimensional gravity with negative cosmological constant. In Seminar at DAMTP. University of Cambridge Cambridge, 2006.
  • [12] Mihalis Dafermos and Gustav Holzegel. On the nonlinear stability of higher dimensional triaxial Bianchi-{\{IX}\} black holes. Advances in Theoretical and Mathematical Physics, 10(4):503–523, 2006.
  • [13] Mihalis Dafermos and Igor Rodnianski. Decay for solutions of the wave equation on Kerr exterior spacetimes i-ii: The cases— a—¡¡ m or axisymmetry. arXiv preprint arXiv:1010.5132, 2010.
  • [14] Mihalis Dafermos, Igor Rodnianski, and Yakov Shlapentokh-Rothman. Decay for solutions of the wave equation on Kerr exterior spacetimes III: The full subextremal case— a—¡ m. Annals of mathematics, pages 787–913, 2016.
  • [15] Th Damour, N Deruelle, and R Ruffini. On quantum resonances in stationary geometries. Lettere al Nuovo Cimento (1971-1985), 15:257–262, 1976.
  • [16] G Denardo and R Ruffini. On the energetics of Reissner–Nordström geometries. Physics Letters B, 45(3):259–262, 1973.
  • [17] Steven Detweiler. Klein-Gordon equation and rotating black holes. Physical Review D, 22(10):2323, 1980.
  • [18] Laurent Di Menza and Jean-Philippe Nicolas. Superradiance on the Reissner–Nordstrøm metric. Classical and Quantum Gravity, 32(14):145013, 2015.
  • [19] Dominic Dold. Unstable mode solutions to the Klein–Gordon equation in Kerr-anti-de Sitter spacetimes. Communications in Mathematical Physics, 350:639–697, 2017.
  • [20] Lawrence C Evans. Partial differential equations, volume 19. American Mathematical Society, 2022.
  • [21] Oran Gannot. Elliptic boundary value problems for Bessel operators, with applications to anti-de Sitter spacetimes. Comptes Rendus Mathematique, 356(10):988–1029, 2018.
  • [22] Elena Giorgi. The linear stability of Reissner–Nordström spacetime: the full subextremal range— q—¡ m. Communications in Mathematical Physics, 380(3):1313–1360, 2020.
  • [23] Olivier Graf and Gustav Holzegel. Mode stability results for the Teukolsky equations on Kerr–anti-de Sitter spacetimes. Classical and Quantum Gravity, 40(4):045003, 2023.
  • [24] Olivier Graf and Gustav Holzegel. Linear stability of schwarzschild-anti-de sitter spacetimes i: The system of gravitational perturbations. arXiv preprint arXiv:2408.02251, 2024.
  • [25] Olivier Graf and Gustav Holzegel. Linear stability of schwarzschild-anti-de sitter spacetimes ii: Logarithmic decay of solutions to the teukolsky system. arXiv preprint arXiv:2408.02252, 2024.
  • [26] Steven S Gubser. Phase transitions near black hole horizons. Classical and Quantum Gravity, 22(23):5121, 2005.
  • [27] Steven S Gubser. Breaking an Abelian gauge symmetry near a black hole horizon. Physical Review D, 78(6):065034, 2008.
  • [28] Sean A Hartnoll, Christopher P Herzog, and Gary T Horowitz. Building an AdS/CFT superconductor. arXiv preprint arXiv:0803.3295, 2008.
  • [29] Sean A Hartnoll, Christopher P Herzog, and Gary T Horowitz. Holographic superconductors. Journal of High Energy Physics, 2008(12):015, 2008.
  • [30] Carlos Herdeiro and Eugen Radu. Ergosurfaces for Kerr black holes with scalar hair. Physical Review D, 89(12):124018, 2014.
  • [31] Carlos Herdeiro, Eugen Radu, and Helgi Rúnarsson. Non-linear Q-clouds around Kerr black holes. Physics Letters B, 739:302–307, 2014.
  • [32] Carlos AR Herdeiro and Eugen Radu. Kerr black holes with scalar hair. Physical review letters, 112(22):221101, 2014.
  • [33] Markus Heusler. Stationary black holes: Uniqueness and beyond. Living Reviews in Relativity, 1:1–57, 1998.
  • [34] Gustav Holzegel. On the massive wave equation on slowly rotating Kerr-AdS spacetimes. Communications in Mathematical Physics, 294(1):169–197, 2010.
  • [35] Gustav Holzegel. Well-posedness for the massive wave equation on asymptotically anti-de Sitter spacetimes. Journal of Hyperbolic Differential Equations, 9(02):239–261, 2012.
  • [36] Gustav Holzegel, Jonathan Luk, Jacques Smulevici, and Claude Warnick. Asymptotic properties of linear field equations in anti-de Sitter space. Communications in Mathematical Physics, 374(2):1125–1178, 2020.
  • [37] Gustav Holzegel and Jacques Smulevici. Self-gravitating Klein–Gordon fields in asymptotically anti-de-Sitter spacetimes. In Annales Henri Poincare, volume 13, pages 991–1038. Springer, 2012.
  • [38] Gustav Holzegel and Jacques Smulevici. Decay properties of Klein-gordon fields on Kerr-AdS spacetimes. Communications on Pure and Applied Mathematics, 66(11):1751–1802, 2013.
  • [39] Gustav Holzegel and Jacques Smulevici. Stability of Schwarzschild-ads for the spherically symmetric Einstein-Klein-Gordon system. Communications in Mathematical Physics, 317:205–251, 2013.
  • [40] Gustav Holzegel and Jacques Smulevici. Quasimodes and a lower bound on the uniform energy decay rate for Kerr–AdS spacetimes. Analysis & PDE, 7(5):1057–1090, 2014.
  • [41] Gustav H Holzegel and Claude M Warnick. Boundedness and growth for the massive wave equation on asymptotically anti-de Sitter black holes. Journal of Functional Analysis, 266(4):2436–2485, 2014.
  • [42] Gary T Horowitz. Introduction to holographic superconductors. From Gravity to Thermal Gauge Theories: The AdS/CFT Correspondence: The AdS/CFT Correspondence, pages 313–347, 2011.
  • [43] Warren Li and Maxime Van de Moortel. Kasner inversions and fluctuating collapse inside hairy black holes with charged matter. arXiv preprint arXiv:2302.00046, 2023.
  • [44] Juan Maldacena. The large-N limit of superconformal field theories and supergravity. International journal of theoretical physics, 38(4):1113–1133, 1999.
  • [45] Juan Maldacena. Eternal black holes in anti-de Sitter. Journal of High Energy Physics, 2003(04):021, 2003.
  • [46] Georgios Moschidis. The Einstein–null dust system in spherical symmetry with an inner mirror: structure of the maximal development and Cauchy stability. arXiv preprint arXiv:1704.08685, 2017.
  • [47] Georgios Moschidis. The characteristic initial-boundary value problem for the Einstein–massless Vlasov system in spherical symmetry. arXiv preprint arXiv:1812.04274, 2018.
  • [48] Georgios Moschidis. A proof of the instability of AdS for the Einstein-null dust system with an inner mirror. Analysis & PDE, 13(6):1671–1754, 2020.
  • [49] Georgios Moschidis. A proof of the instability of AdS for the Einstein-massless Vlasov system. Inventiones mathematicae, 231(2):467–672, 2023.
  • [50] William H Press and Saul A Teukolsky. Floating orbits, superradiant scattering and the black-hole bomb. Nature, 238(5361):211–212, 1972.
  • [51] Yakov Shlapentokh-Rothman. Exponentially growing finite energy solutions for the Klein–Gordon equation on sub-extremal Kerr spacetimes. Communications in Mathematical Physics, 329:859–891, 2014.
  • [52] Yakov Shlapentokh-Rothman. Quantitative mode stability for the wave equation on the Kerr spacetime. In Annales Henri Poincaré, volume 16, pages 289–345. Springer, 2015.
  • [53] Saul A Teukolsky. Rotating black holes: Separable wave equations for gravitational and electromagnetic perturbations. Physical Review Letters, 29(16):1114, 1972.
  • [54] Maxime Van de Moortel. Violent nonlinear collapse in the interior of charged hairy black holes. Archive for Rational Mechanics and Analysis, 248(5):89, 2024.
  • [55] Claude M Warnick. The massive wave equation in asymptotically AdS spacetimes. Communications in mathematical physics, 321(1):85–111, 2013.
  • [56] Bernard F Whiting. Mode stability of the Kerr black hole. Journal of Mathematical Physics, 30(6):1301–1305, 1989.
  • [57] Edward Witten. Anti de Sitter space and holography. arXiv preprint hep-th/9802150, 1998.
  • [58] Weihao Zheng. Asymptotically anti-de sitter spherically symmetric hairy black holes. In preparation, 2024.
  • [59] Theodoros JM Zouros and Douglas M Eardley. Instabilities of massive scalar perturbations of a rotating black hole. Annals of physics, 118(1):139–155, 1979.