betaFeedback

WorksPhysics and Astronomy

Separation of gravitational perturbations of Myers-Perry black holes

Abstract

We study linearized gravitational perturbations of five-dimensional asymptotically flat Myers–Perry black holes with two unequal spins. We construct separated solutions for the five coupled gauge-invariant polarizations for arbitrary azimuthal mode numbers. The key observation is that the Killing–Yano symmetry induces a first-order angular operator that, together with a commuting second-order operator, determines the angular profiles of all five gravitational polarizations. The Carter interchange relates the scalar angular and radial equations, and the five tidal components of each separated solution share a single radial amplitude. We prove completeness of the angular harmonics at real frequencies and show that the resulting physical perturbations admit local metric reconstruction. In the singly spinning limit, we recover Santos's angular operators and radial equation for perturbations with mψ=0m_ψ=0 and even parity under ψ→−ψψ\to-ψ. These results provide a framework for studying quasinormal modes and gravitational stability of five-dimensional rotating black holes with two independent spins.

Introduction

Black hole perturbation theory provides a fundamental framework for understanding gravitational dynamics, stability, and radiation. The characteristic oscillations of perturbed black holes, known as quasinormal modes, underpin black hole spectroscopy and offer a powerful probe of strong-field gravity [1, 2]. A cornerstone of this framework is the separability of gravitational perturbation equations. For Schwarzschild black holes, gravitational perturbations decompose into odd- and even-parity sectors, each governed by a radial wave equation [3, 4]. For rotating Kerr black holes, the extreme Weyl scalars satisfy decoupled equations that separate into radial and angular parts [5, 6]. These landmark developments established the foundations of modern black hole perturbation theory and provided a systematic approach to understanding the intricate dynamics of gravitational perturbations.

Extending the separation of gravitational perturbations to higher dimensions remains challenging. Five-dimensional Myers-Perry black holes [7] have two independent rotations and five gravitational-wave polarizations. Although master equations exist for static black holes with maximally symmetric horizons [8], rotation complicates the problem. Hidden symmetries allow geodesic motion and scalar waves to separate [9, 10, 11], but the optical properties of the principal null congruences obstruct direct decoupling of the extreme Weyl components [12]. Nevertheless, coupling need not prevent separability, as demonstrated for Kerr-Newman gravitioelectromagnetic perturbations [13], followed by quasinormal-mode calculations [14]. For Myers-Perry black holes, progress includes Killing-Yano symmetry in selected equal-spin gravitational modes [15], separated gravitational solutions for special spin configurations, including cases with all spins nonzero [16], and, more recently, separation of the transverse-sphere-invariant sector of singly spinning black holes [17].

A useful geometric clue comes from Kaluza-Klein reduction. The principal tensor of the Myers-Perry geometry selects a Killing direction ξK=ab ∂t+b ∂ϕ+a ∂ψ\xi_{K}=ab\,\partial_{t}+b\,\partial_{\phi}+a\,\partial_{\psi}, along which the five-dimensional metric reduces locally to a four-dimensional metric conformal to a Carter-type metric of Petrov type D, together with a gauge field and a dilaton. This geometric structure motivates our use of Carter coordinates adapted to KCK_{C} and suggests that separation may persist despite the coupling among gravitational polarizations.

In this work, we construct separated radiative solutions of the linearized vacuum equations for nonextremal, asymptotically flat five-dimensional Myers-Perry black holes with two unequal, nonzero spins. A first-order angular operator induced by the Killing-Yano symmetry, together with a commuting second-order operator, determines the relative profiles of all five tidal polarizations and yields a complete angular basis at real frequencies. The Carter interchange yields the radial equations, while the linearized Einstein equations determine which angular and radial branches can be combined to form physical gravitational perturbations. On each regular matched branch, all five polarizations share a single radial amplitude, allowing local metric reconstruction up to gauge.

Model and setup

We consider the five-dimensional, asymptotically flat Myers-Perry black hole [7]. In Carter coordinates (T,r,y,P,w)(T,r,y,P,w) with signature (−+++++)(-+++++), its line element reads [18]

ds2=−XΣA2+Σ(dr2X+dy2Y)+YΣB2+r2y2C2(1)ds^2=-\frac{X}{\Sigma}\mathcal{A}^2+\Sigma\left(\frac{dr^2}{X}+\frac{dy^2}{Y}\right)+\frac{Y}{\Sigma}\mathcal{B}^2+r^2y^2\mathcal{C}^2 \tag*{(1)}

where A=dT+y2dP\mathcal{A}=dT+y^2dP, B=dT−r2dP\mathcal{B}=dT-r^2dP, and C=dw−abΣ(Ar2+By2)\mathcal{C}=dw-\frac{ab}{\Sigma}\left(\frac{\mathcal{A}}{r^2}+\frac{\mathcal{B}}{y^2}\right). Here Σ=r2+y2\Sigma=r^2+y^2, X=Δ/r2X=\Delta/r^2, Y=(a2−y2)(y2−b2)/y2Y=(a^2-y^2)(y^2-b^2)/y^2, and Δ=(r2+a2)(r2+b2)−μr2\Delta=(r^2+a^2)(r^2+b^2)-\mu r^2, with μ\mu the mass parameter and a,ba,b the two rotation parameters. We consider nonextremal black holes with μ>(∣a∣+∣b∣)2\mu>(|a|+|b|)^2. The exterior region is r>r+r>r_+, where r+r_+ is the largest positive root of Δ\Delta.

For two nonzero spins of unequal magnitude, we take a>b>0a>b>0 without loss of generality. The latitude coordinate yy satisfies b≤y≤ab\le y\le a. The Carter coordinates are related to the Boyer–Lindquist coordinates (t,r,θ,ϕ,ψ)(t,r,\theta,\phi,\psi) by y2=a2cos⁡2θ+b2sin⁡2θy^2=a^2\cos^2\theta+b^2\sin^2\theta, t=T+(a2+b2)P+ab wt=T+(a^2+b^2)P+ab\,w, ϕ=aP+bw\phi=aP+bw, and ψ=bP+aw\psi=bP+aw [19, 20]. Here rr is unchanged, 0≤θ≤π/20\le\theta\le\pi/2, and both ϕ\phi and ψ\psi have period 2π2\pi. For unequal spin magnitudes, cases with ab<0ab<0 are included by reversing ϕ\phi or ψ\psi. For equal spin magnitudes, a2=b2>0a^2=b^2>0, yy becomes constant and the Carter coordinates degenerate, while the enhanced U(2)U(2) symmetry organizes the perturbations into Wigner harmonics [21].

In what follows, we focus on gravitational perturbations of Myers–Perry black holes with unequal spin magnitudes. For a metric perturbation habh_{ab}, we seek Fourier solutions of the linearized vacuum equations,

δRab[h]=0,hab∝e−iωT+ikPP+ikww.(2)\delta R_{ab}[h]=0,\qquad h_{ab}\propto e^{-i\omega T+ik_P P+ik_w w}. \tag*{(2)}

Although TT differs from Boyer–Lindquist time tt, ∂T=∂t\partial_T=\partial_t, so ω\omega is the same in both coordinate systems. In Boyer–Lindquist coordinates, the Fourier factor is e−iωt+imϕϕ+imψψe^{-i\omega t+im_\phi\phi+im_\psi\psi}, where the 2π2\pi periods of ϕ\phi and ψ\psi require mϕ,mψ∈Zm_\phi,m_\psi\in\mathbb{Z}. Using the transformation between Boyer–Lindquist and Carter coordinates and comparing with (2), we obtain kP=amϕ+bmψ−(a2+b2)ωk_P=am_\phi+bm_\psi-(a^2+b^2)\omega and kw=bmϕ+amψ−abωk_w=bm_\phi+am_\psi-ab\omega. Thus, at fixed ω\omega, the Fourier coefficients kPk_P and kwk_w are determined by the azimuthal numbers mϕm_\phi and mψm_\psi.

Curvature variables and polarization.— The type-D structure of the Myers–Perry background allows us to construct gauge-invariant curvature variables for its perturbations [12] [22]. Following the higher-dimensional Newman–Penrose approach [23], we choose a null frame (L,N,Ei)(L,N,E_i), with i=+,−,3i=+,-,3, where LL and NN are aligned with the two Weyl-aligned null directions. In Carter coordinates (T,r,y,P,w)(T,r,y,P,w), the frame vectors are La=X−1(r2,−X,0,1,ab/r2)L^a=X^{-1}(r^2,-X,0,1,ab/r^2), Na=(2Σ)−1(r2,X,0,1,ab/r2)N^a=(2\Sigma)^{-1}(r^2,X,0,1,ab/r^2), E±a=(p±2Y)−1(0,0,∓iY,1,−ab/y2)E_\pm^a=(p_\pm\sqrt{2Y})^{-1}(0,0,\mp iY,1,-ab/y^2), and E3a=(ry)−1(0,0,0,0,1)E_3^a=(ry)^{-1}(0,0,0,0,1), where p±=r±iyp_\pm=r\pm iy. They satisfy LaNa=−1L^aN_a=-1, E+aE−a=1E_+^aE_{-a}=1, and E3aE3a=1E_3^aE_{3a}=1, with EiE_i forming a complex basis for the three-dimensional space orthogonal to LL and NN.

In this frame, the background Weyl projections CabcdLaEibLcEjdC_{abcd}L^aE_i^bL^cE_j^d vanish, where CabcdC_{abcd} is the Weyl tensor. We introduce their linear perturbations Ωij[h]=δ(CabcdLaEibLcEjd)\Omega_{ij}[h]=\delta(C_{abcd}L^aE_i^bL^cE_j^d), where δ\delta denotes the first-order variation induced by habh_{ab}, including the variation of the null frame. The resulting components are gauge invariant and form a symmetric trace-free tensor on the three-dimensional transverse space [12] [23]. In particular, the trace condition gives Ω−−=−Ω33/2\Omega_{--}=-\Omega_{33}/2, leaving five independent components (Ω++,Ω−−,Ω+3,Ω−3,Ω33)(\Omega_{++},\Omega_{--},\Omega_{+3},\Omega_{-3},\Omega_{33}) that describe the five gravitational-wave polarizations.

Killing–Yano symmetry and angular harmonics.—The Killing–Yano three-form ff, dual to the principal two-form, defines a first-order symmetry operator acting on metric perturbations as (Qh)ab=facd∇chbd+fbcd∇chad(Qh)_{ab}=f_a{}^{cd}\nabla_ch_{bd}+f_b{}^{cd}\nabla_ch_{ad} [11] [24]. On this Ricci-flat background, the operator QQ constructed from ff maps each solution hh of the linearized Einstein equation to another solution QhQh [24]. Evaluating the gauge-invariant curvature components Ωij[Qh]\Omega_{ij}[Qh] and using δRab[h]=0\delta R_{ab}[h]=0, we find

Ω[Qh]=FΩ[h],(3)\Omega[Qh]=F\Omega[h], \tag*{(3)}

where FF is a first-order angular differential operator.

The angular structure (3) is key to separating the perturbation equations. Suppressing the Fourier factor, we seek separated waves of the form

Ωij=R(r)Aij(y).(4)\Omega_{ij}=R(r)A_{ij}(y). \tag*{(4)}

The five angular functions Aij(y)A_{ij}(y), assembled into A(y)=(A++,A−−,A+3,A−3,A33)TA(y)=(A_{++},A_{--},A_{+3},A_{-3},A_{33})^T, describe the angular dependence of the gravitational-wave polarizations, while R(r)R(r) describes their common radial dependence. Since FF involves only angular derivatives, we have FΩ=R(r)FA(y)F\Omega=R(r)FA(y). To further distinguish the angular modes, we construct a second-order angular operator JJ satisfying [F,J]=0[F,J]=0. We then seek joint angular eigenfunctions satisfying

FA=νA,JA=λA.(5)FA=\nu A,\qquad JA=\lambda A. \tag*{(5)}

The angular operators take the form F=α1∂y+α0F=\alpha_1\partial_y+\alpha_0 and J=γ2∂y2+β1∂y+β0J=\gamma_2\partial_y^2+\beta_1\partial_y+\beta_0, where II is the 5×55\times5 identity matrix. The eigenvalues λ\lambda and ν\nu are not independent but satisfy H(λ,ν)=0H(\lambda,\nu)=0. The explicit operator coefficients and the polynomial HH are given in the Supplemental Material (SM). The joint eigenvalue equations for FF and JJ, supplemented by smoothness at both axes, determine the angular profiles of all five components Aij(y)A_{ij}(y).

We may also formulate the angular equations in terms of a single scalar s=2A++/Ys=2A_{++}/Y, from which all five angular components can be reconstructed. This scalar formulation also provides a route to the radial equation for R(r)R(r) through the Carter interchange. For ω≠0\omega\ne0 and fixed λ\lambda, every root ν\nu of H(λ,ν)=0H(\lambda,\nu)=0 gives a two-dimensional space of local angular solutions. At ω=0\omega=0, regularity at both axes ensures that all roots in ν\nu are simple. Wherever (s,s′)(s,s') provides a nonsingular parametrization of this space, all five angular components can be reconstructed as A=u0s′+v0sA=u_0s'+v_0s, with u0u_0 and $v_0 determined by the joint angular equations. The coupled angular system then reduces to a single second-order scalar equation,

d2sdy2=c0(y)s+c1(y)dsdy,(6)\frac{\mathrm{d}^{2}s}{\mathrm{d}y^{2}}=c_{0}(y)s+c_{1}(y)\frac{\mathrm{d}s}{\mathrm{d}y}, \tag*{(6)}

where c0c_{0} and c1c_{1} follow algebraically from FF and JJ. The normalization 2/Y2/Y accounts for the frame transformation under the Carter interchange, allowing the angular equation to be continued into the radial equation. The explicit construction is given in the Supplemental Material (SM).

Radial separation.—The radial equation can be obtained from the angular equation through the Carter interchange

r~=−iy,y~=ir,μ~=−μ,\tilde{r}=-iy,\qquad\tilde{y}=ir,\qquad\tilde{\mu}=-\mu,
a~2+b~2=a2+b2−μ,a~b~=ab.(7)\tilde{a}^{2}+\tilde{b}^{2}=a^{2}+b^{2}-\mu,\qquad\tilde{a}\tilde{b}=ab. \tag*{(7)}

This transformation exchanges the metric functions X(r)X(r) and Y(y)Y(y). Keeping the Fourier labels (ω,kP,kw)(\omega,k_{P},k_{w}) fixed, the eigenvalues of FF and JJ transform as ν~=ν\tilde{\nu}=\nu and λ~=λ+μω2\tilde{\lambda}=\lambda+\mu\omega^{2}, respectively. Applying this continuation to (6) yields

d2Rdr2=−c~0(ir)R+ic~1(ir)dRdr,(8)\frac{\mathrm{d}^{2}R}{\mathrm{d}r^{2}}=-\tilde{c}_{0}(ir)R+i\tilde{c}_{1}(ir)\frac{\mathrm{d}R}{\mathrm{d}r}, \tag*{(8)}

where c~0\tilde{c}_{0} and c~1\tilde{c}_{1} are obtained from the angular coefficients with the continued parameters.

The radial equation also follows directly by substituting Ωij=R(r)Aij(y)\Omega_{ij}=R(r)A_{ij}(y) into the vacuum curvature constraints and eliminating angular derivatives using (5). On a regular matched branch its coefficients agree with (8) wherever the Wronskian is nonzero.

At fixed ν≠0\nu\ne0, the spectral relation H(λ,ν)=0H(\lambda,\nu)=0 is quadratic in λ\lambda. For generic parameters, its two distinct roots label two angular branches and, through the Carter interchange, two radial branches. Of the four candidate pairings, only the two satisfying λ~=λ+μω2\tilde{\lambda}=\lambda+\mu\omega^{2} obey the linearized vacuum Einstein equations on a regular reconstruction patch (see the SM).

Angular completeness.—The angular eigenvalue equations must be supplemented by regularity conditions at the two rotation axes. At y=ay=a the ϕ\phi circle collapses, whereas at y=by=b the ψ\psi circle collapses. Since the complex transverse frame EiE_i is not regular at these axes, smoothness is imposed on the curvature tensor represented by Ωij\Omega_{ij}, expressed in a smooth local Cartesian frame, rather than on its individual components in the Carter frame. For fixed (ω,mϕ,mψ)(\omega,m_{\phi},m_{\psi}), these conditions select the allowed joint eigenvalues (λ,ν)(\lambda,\nu).

For real ω\omega, the angular operators FF and JJ are symmetric with respect to the positive inner product

⟨A,B⟩ang=∫bay dy A†WB,(9)\langle A,B\rangle_{\mathrm{ang}}=\int_{b}^{a}y\,\mathrm{d}y\,A^{\dagger}WB, \tag*{(9)}

where AA and BB are five-component angular functions with the same Fourier labels, and W=diag⁡(1,1,2,2,32)W=\operatorname{diag}(1,1,2,2,\frac{3}{2}) is determined by the norm of the transverse tidal tensor Ωij\Omega_{ij}. As shown in the SM, the smooth-axis realization of JJ is self-adjoint and elliptic on the compact angular three-sphere. Since [F,J]=0[F,J]=0, the symmetric operator FF can be diagonalized within each finite-dimensional eigenspace of JJ. The resulting regular joint angular harmonics form a complete orthonormal basis.

Consequently, for real ω\omega, any smooth tidal field with fixed Fourier labels admits the expansion

Ω(r,y)=∑jRj(r)A(j)(y),(10)\Omega(r,y)=\sum_{j}R_{j}(r)A^{(j)}(y), \tag*{(10)}

where {A(j)}\{A^{(j)}\} is a complete orthonormal basis of joint angular harmonics, including possible degeneracies, and Rj(r)=⟨A(j),Ω(r,⋅)⟩angR_{j}(r)=\langle A^{(j)},\Omega(r,\cdot)\rangle_{\mathrm{ang}}. If Ω\Omega arises from a vacuum metric perturbation, its projection onto each joint (λ,ν)(\lambda,\nu) eigenspace preserves the vacuum curvature compatibility conditions on the regular reconstruction patches specified in the SM. For a degenerate eigenvalue pair, the projection includes all harmonics sharing that pair.

Radial boundary conditions.—The future event horizon is generated by the Killing vector χ=(r+2∂t+∂ϕ+ab r+−2∂ψ)/μ\chi=(r_{+}^{2}\partial_{t}+\partial_{\phi}+ab\,r_{+}^{-2}\partial_{\psi})/\mu, with unit coefficient of the asymptotically normalized time translation ∂t\partial_{t}. Its action on a Fourier mode defines the frequency relative to the rotating horizon [10],

ωH=ω−amϕr+2+a2−bmψr+2+b2.(11)\omega_{H}=\omega-\frac{am_{\phi}}{r_{+}^{2}+a^{2}}-\frac{bm_{\psi}}{r_{+}^{2}+b^{2}}. \tag*{(11)}

For a nonextremal horizon, the surface gravity is κH=Δ′(r+)/(2μr+2)\kappa_{H}=\Delta'(r_{+})/(2\mu r_{+}^{2}). For perturbations that are purely ingoing at the horizon and purely outgoing at spatial infinity, the radial amplitude has the asymptotic behavior

R(r)∼{(r−r+)−iωH2κH,r→r+(ingoing),r−3/2eiωr,r→∞(outgoing).(12)R(r)\sim \begin{cases} (r-r_{+})^{-\frac{i\omega_{H}}{2\kappa_{H}}}, & r\to r_{+}\quad\text{(ingoing)},\\ r^{-3/2}e^{i\omega r}, & r\to\infty\quad\text{(outgoing)}. \end{cases} \tag*{(12)}

Together with the regular angular eigenvalue conditions, these boundary conditions determine the complex quasinormal-mode frequencies of the perturbations.

Recovering the metric perturbation.—Although the angular and radial equations separate the five tidal components, a separated curvature field must also arise from a metric perturbation satisfying the full linearized Einstein equations. This imposes additional compatibility conditions on Ωij\Omega_{ij}. As shown in the SM, a rank analysis yields ten independent fourth-order differential conditions and establishes that they are locally sufficient for reconstructing a vacuum metric perturbation up to gauge. We choose the gauge hTT=hTP=hTw=hPP=hPw=0h_{TT}=h_{TP}=h_{Tw}=h_{PP}=h_{Pw}=0, in which the tidal components and their derivatives through fourth order determine the remaining ten metric components wherever the reconstruction matrices are nonsingular.

For a separated mode Ωij=R(r)Aij(y)\Omega_{ij}=R(r)A_{ij}(y), the higher derivatives can be eliminated using the scalar angular and radial equations (6) and (8). The gauge-fixed Fourier amplitude then takes the form

Habgf=Mab(r,y)(Rs,Rs′,Rs′,Rs′)T.(13)H_{ab}^{\mathrm{gf}}=M_{ab}(r,y)(R_s,R_s',R_s',R_s')^{T}. \tag*{(13)}

Here MabM_{ab} is determined by the curvature projections and the linearized Einstein equations, as detailed in the SM. The compatibility conditions select the physical angular–radial pairings discussed above. On each regular matched branch, the two independent angular and radial solutions yield four local separated solutions. The reconstructed metric reproduces the prescribed curvature Ωij=RAij\Omega_{ij}=RA_{ij} and satisfies the linearized vacuum Einstein equations δRab[hgf]=0\delta R_{ab}[h^{\mathrm{gf}}]=0, confirming that these solutions describe genuine gravitational perturbations. For a fixed angular harmonic, the radial initial data (R,R′)(R,R') at an ordinary point of Eq. (8) determine the metric perturbation locally up to gauge. Stationary parameter perturbations with vanishing principal tidal curvature require separate treatment.

Single-spin limit.—The singly spinning limit provides both a nontrivial consistency check with Santos [17] and an illustration of the broader perturbation sectors covered by our construction. To make direct contact with Santos’s results, we set b=mψ=0b=m_{\psi}=0 and restrict to perturbations even under ψ↦−ψ\psi\mapsto-\psi. Then kw=0k_w=0, Y=a2−y2Y=a^2-y^2, and the three even components are (A−−,A++,A33)(A_{--},A_{++},A_{33}). Since JJ preserves parity whereas FF reverses it, the angular operators acting within this sector are JeJ_e and (F2)e(F^2)_e. Introducing x=y/a=cos⁡θx=y/a=\cos\theta and D=diag⁡(1/2,1/2,1)D=\operatorname{diag}(1/2,1/2,1), we obtain the exact operator identities

D−1JeD=KC−(mϕ2+3)I,D^{-1}J_eD=K_C-(m_\phi^2+3)I,
D−1(F2)eD=3K3−KC+[(aω−mϕ)2+6]I.(14)D^{-1}(F^2)_eD=3K_3-K_C+\left[(a\omega-m_\phi)^2+6\right]I. \tag*{(14)}

Here KCK_C and K3K_3 are the angular operators of [17]. The corresponding separation constants are related by

λS=λ+mϕ2+3,\lambda_S=\lambda+m_\phi^2+3,
3νS=ν2+λS−(aω−mϕ)2−6.(15)3\nu_S=\nu^2+\lambda_S-(a\omega-m_\phi)^2-6. \tag*{(15)}

With this identification and the Fourier-label reversal (ωS,mS)=(−ω,−mϕ)(\omega_S,m_S)=(-\omega,-m_\phi) required by the different principal-null conventions, our angular spectral relation and second-order radial equation agree exactly with [17].

Apart from the reflection-even sector (A−−,A++,A33)(A_{--},A_{++},A_{33}), our construction also includes the reflection-odd sector (A+3,A−3)(A_{+3},A_{-3}) at mψ=0m_\psi=0, as well as perturbations with mψ≠0m_\psi\ne0 in the singly spinning limit. For the general five-dimensional Myers–Perry black hole with two nonzero, unequal spins, the joint angular harmonics involve all five tidal polarizations and form a complete basis at real frequency for arbitrary azimuthal mode numbers. On each regular matched radial branch, the five components share a single radial amplitude governed by a second-order scalar equation.

Discussion.—We have established the separation of gravitational perturbations of five-dimensional Myers–Perry black holes with two nonzero spins of unequal magnitude (∣a∣≠∣b∣|a|\ne|b|). A key observation is the remarkably simple relation Q[Qh]=FΩ[h]\mathcal{Q}[\mathit{Qh}]=F\Omega[h] in Eq. (3), which shows that the Killing–Yano symmetry acts on the five gauge-invariant tidal components through a purely angular differential operator FF. Together with a second, commuting operator JJ, this structure allows the coupled gravitational perturbation equations to be separated. The joint angular eigenvalue problem determines the relative profiles of the five tidal polarizations, while their radial dynamics reduce to a single second-order scalar equation. Thus, the coupling between gravitational polarizations does not obstruct separability, revealing a hidden structure in the linearized Einstein equations and providing a practical framework for black-hole spectroscopy. We have also discussed angular completeness at real frequencies, radial boundary conditions, and local metric reconstruction. As a nontrivial consistency check, the singly spinning limit reproduces the five-dimensional separation equations of Santos [17] in the corresponding symmetry sector.

The separated equations provide a framework for investigating gravitational quasinormal modes and stability across the two-spin parameter space of five-dimensional Myers–Perry black holes, as well as for exploring non-Hermitian exceptional points, excitation factors, and resonant response [25]. Several extensions are worth exploring. One is to include a cosmological constant and examine whether the separation mechanism persists for five-dimensional rotating (anti-)de Sitter black holes [18]. Another is to extend the construction to charged Chong–Cvetič–Lü–Pope (CCLP) black holes, whose generalized Killing–Yano symmetry with torsion suggests a possible route to separating coupled gravitoelectromagnetic perturbations [26, 27]. Extension to higher-dimensional Myers–Perry black holes is more challenging. In five dimensions, our operator QQ uses the Hodge dual of the principal two-form. In DD dimensions this dual is a (D−2)(D-2)-form [11]. Although Killing–Yano three-forms can be constructed from higher powers of the principal tensor in higher odd dimensions [28], whether they induce suitable angular symmetry operators for gravitational perturbations remains unclear. The presence of multiple latitude variables and D(D−3)/2D(D-3)/2 gravitational polarizations presents further challenges.

Acknowledgments.—We thank Li Li, Peng-Cheng Li, and Tieguang Zi for helpful discussions. J.-K. Z. is particularly grateful to Li Li for his encouragement during this project. This work was supported in part by the National Natural Science Foundation of China (Grant Nos. 12505087, 12447134 and 12365011). H.-D. L. was also supported by Postdoctoral Innovation Project of Shandong Province (SDCX-ZG-202503036).

References

Supplemental Material for Separation of gravitational perturbations of Myers-Perry black holes

Jun-Kun Zhao; Hong-Da Lyu

SM.pdf

Supplemental Material for Separation of gravitational perturbations of Myers-Perry black holes

Carter coordinates and the tidal field

Coordinate identifications. For the angular regularity analysis below, we use the inverse of the coordinate transformation given in the Letter,

(T,P,w)=(t,0,0)+1a2−b2(−a3ϕ+b3ψ,aϕ−bψ,aψ−bϕ).(16)(T,P,w)=(t,0,0)+\frac{1}{a^2-b^2}\left(-a^3\phi+b^3\psi,a\phi-b\psi,a\psi-b\phi\right). \tag*{(16)}

At fixed tt, the 2π2\pi periodicities of ϕ\phi and ψ\psi determine the corresponding identifications of (T,P,w)(T,P,w). The ϕ\phi and ψ\psi circles collapse at y=ay=a and y=by=b, respectively.

Curvature along a principal null ray. For the tidal-curvature calculation, we use the metric duals of the principal frame vectors defined in the Letter,

L♭=−A−ΣX dr,N♭=−X2Σ(A−ΣX dr),E±♭=Y/2r±iy(−B∓iΣY dy),E3♭=ry C.(17)L^\flat=-A-\frac{\Sigma}{X}\,dr,\qquad N^\flat=-\frac{X}{2\Sigma}\left(A-\frac{\Sigma}{X}\,dr\right),\qquad E^\flat_{\pm}=\frac{\sqrt{Y/2}}{r\pm iy}\left(-B\mp i\frac{\Sigma}{Y}\,dy\right),\qquad E^\flat_3=ry\,\mathcal{C}. \tag*{(17)}

Set p±=r±iyp_{\pm}=r\pm iy and p3=rp_3=r. The principal frame satisfies [L,Ei]=Ei/pi[L,E_i]=E_i/p_i, while the only nonvanishing transverse component of dL♭dL^\flat is dL♭(E+,E−)=2iy/ΣdL^\flat(E_+,E_-)=2iy/\Sigma. Since the transverse inner products are constant, the Koszul formula gives

2g(∇LEi,Ej)=dL♭(Ei,Ej)+(1pi−1pj)g(Ei,Ej)=0.(18)2g(\nabla_L E_i,E_j)=dL^\flat(E_i,E_j)+\left(\frac{1}{p_i}-\frac{1}{p_j}\right)g(E_i,E_j)=0. \tag*{(18)}

Thus (∇LEi)⊥=0(\nabla_L E_i)_\perp=0, where ⊥\perp denotes projection onto the transverse screen.

The tidal perturbation defined in the Letter includes the variation of the null frame. Since the background Weyl components of boost weights +2+2 and +1+1 vanish, only the component of δL\delta L along NN contributes. Preserving the null condition gives the convenient choice δL=12hLLN\delta L=\frac{1}{2}h_{LL}N, where hLL=habLaLbh_{LL}=h_{ab}L^aL^b. For a vacuum perturbation, the tidal curvature becomes

Ωij[h]=δR(L,Ei,L,Ej)+12hLL[R(N,Ei,L,Ej)+R(L,Ei,N,Ej)].(19)\Omega_{ij}[h]=\delta R(L,E_i,L,E_j)+\frac{1}{2}h_{LL}\left[R(N,E_i,L,E_j)+R(L,E_i,N,E_j)\right]. \tag*{(19)}

Here δR\delta R is evaluated in the fixed background frame, with the curvature convention δRijkl=(hil,jk+hjk,il−hjl,ik−hik,jl)/2+⋯\delta R_{ijkl}=(h_{il,jk}+h_{jk,il}-h_{jl,ik}-h_{ik,jl})/2+\cdots.

Projecting the metric symmetry onto the tidal field

To fix the normalization of the Killing–Yano symmetry operator QQ introduced in the Letter, we choose the orientation ϵTryPw=ryΣ\epsilon_{TryPw}=ry\Sigma, so that ϵTryPw=−1/(ryΣ)\epsilon^{TryPw}=-1/(ry\Sigma). The corresponding Killing–Yano three-form [1] is

f=⋆(−r dr∧A+y dy∧B).(20)f=\star\left(-r\,dr\wedge A+y\,dy\wedge B\right). \tag*{(20)}

The kernel of the principal two-form is generated by ∂w\partial_w. Its metric dual is r2y2Cr^2y^2\mathcal{C}, so the local Einstein-frame quotient is ry(g5−r2y2C2)ry(g_5-r^2y^2\mathcal{C}^2) up to a constant [2]. The metric in parentheses has the four-dimensional Carter form of Petrov type D [3, 4].

Consider either of the simultaneous sign reversals (a,ϕ,mϕ)→(−a,−ϕ,−mϕ)(a,\phi,m_{\phi}) \to(-a,-\phi,-m_{\phi}) or (b,ψ,mψ)→(−b,−ψ,−mψ)(b,\psi,m_{\psi}) \to(-b,-\psi,-m_{\psi}). The coordinate and Fourier relations in the Letter then imply that (T,P,r,y,ω,kP)(T,P,r,y,\omega,k_{P}) remain unchanged, whereas (w,kw)→(−w,−kw)(w,k_{w}) \to(-w,-k_{w}). This transformation reverses E3E_{3} and the orientation entering the Hodge dual, and hence changes the sign of the Killing–Yano three-form. Consequently, the angular operators transform as

F→−DrefFDref,J→DrefJDref.(21)F \to-D_{\mathrm{ref}}FD_{\mathrm{ref}}, \qquad J \to D_{\mathrm{ref}}JD_{\mathrm{ref}}. \tag*{(21)}

where Dref=diag⁡(1,1,−1,−1,1)D_{\mathrm{ref}}=\operatorname{diag}(1,1,-1,-1,1). The corresponding eigenvalues therefore transform as (λ,ν)→(λ,−ν)(\lambda,\nu) \to(\lambda,-\nu).

For the explicit coefficients, we introduce the angular Fourier combination

W0=−ωy2−kP+abkwy2=Y(ω−amϕa2−y2+bmψy2−b2).(22)W_{0}=-\omega y^{2}-k_{P}+\frac{abk_{w}}{y^{2}}=Y\left(\omega-\frac{am_{\phi}}{a^{2}-y^{2}}+\frac{bm_{\psi}}{y^{2}-b^{2}}\right). \tag*{(22)}

In the normalized frame,

F=α1∂y+α0,J=YI∂y2+β1∂y+β0.(23)F=\alpha_{1}\partial_{y}+\alpha_{0}, \qquad J=YI\partial_{y}^{2}+\beta_{1}\partial_{y}+\beta_{0}. \tag*{(23)}

Let h=diag⁡(2,−2,1,−1,0)h=\operatorname{diag}(2,-2,1,-1,0) in the component order (++,−−,+3,−3,33)(++,--,+3,-3,33), and write i=ddi=dd, j=d3j=d3, jˉ=(−d)3\bar{j}=(-d)3, k=33k=33 for d=±1d=\pm1. Primes denote yy derivatives. The curvature projection gives

(α0)nn=(6−2hnn2)aby2+hnnkwy,(\alpha_{0})_{nn}=(6-2h_{nn}^{2})\frac{ab}{y^{2}}+h_{nn}\frac{k_{w}}{y},
((α1)ij,(α1)ji,(α1)jk,(α1)kj)=Y2(2,−1,3/2,−2).(24)((\alpha_{1})_{ij},(\alpha_{1})_{ji},(\alpha_{1})_{jk},(\alpha_{1})_{kj})=\frac{\sqrt{Y}}{2}(2,-1,3/2,-2). \tag*{(24)}

The off-diagonal coefficients are

mnmn(α0)mn/(α1)mn(\alpha_{0})_{mn}/(\alpha_{1})_{mn}y2(β0)mn/(α1)mny^{2}(\beta_{0})_{mn}/(\alpha_{1})_{mn}
ijij2/y+(dW0−Y′/2)/Y2/y+(dW_{0}-Y'/2)/Y4ab/y+ab(Y′−2dW0)/Y−2dkw4ab/y+ab(Y'-2dW_{0})/Y-2dk_{w}
jiji−1/y+(Y′−dW0)/Y-1/y+(Y'-dW_{0})/Y−2ab/y+2ab(dW0−Y′)/Y+2dkw-2ab/y+2ab(dW_{0}-Y')/Y+2dk_{w}
jkjk1/y+dW0/Y1/y+dW_{0}/Y2ab/y−2dabW0/Y−2dkw2ab/y-2dabW_{0}/Y-2dk_{w}
kjkj(Y′/2−dW0)/Y(Y'/2-dW_{0})/Yab(2dW0−Y′)/Y+2dkwab(2dW_{0}-Y')/Y+2dk_{w}
β1=(Y′+Y/y)I−2aby2α1,\beta_{1}=(Y'+Y/y)I-\frac{2ab}{y^{2}}\alpha_{1},
(β0)nn=−(W0−hnnY′/2)2Y−(a2+b2)ω2−2kPω−kw2+3(a2+b2)y2−3a2b2y4+hnn2Y2y2+2hnn(abkwy3+2ωy).(25)(\beta_{0})_{nn}=-\frac{(W_{0}-h_{nn}Y'/2)^{2}}{Y}-(a^{2}+b^{2})\omega^{2}-2k_{P}\omega-\frac{k_{w}^{2}+3(a^{2}+b^{2})}{y^{2}}-\frac{3a^{2}b^{2}}{y^{4}}+\frac{h_{nn}^{2}Y}{2y^{2}}+2h_{nn}\left(\frac{abk_{w}}{y^{3}}+2\omega y\right). \tag*{(25)}

In addition to the four slots in Eq. (S10), the remaining nonzero off-diagonal entries of β0\beta_{0} are

(β0)ik=−3Y2y2,(β0)ki=−Yy2,(β0)jjˉ=3Y2y2.(\beta_{0})_{ik}=-\frac{3Y}{2y^{2}}, \qquad(\beta_{0})_{ki}=-\frac{Y}{y^{2}}, \qquad(\beta_{0})_{j\bar{j}}=\frac{3Y}{2y^{2}}.

All unlisted entries vanish. These coefficients satisfy [F,J]=0[F,J]=0 and H(J,F)=0H(J,F)=0, with

H(λ,ν)=ν5+(5ℓ−3)ν3−42ωkwν2+(4ℓ2−12ℓ+72kPω)ν−24ωkwℓ+144abω2,H(\lambda,\nu)=\nu^{5}+(5\ell-3)\nu^{3}-42\omega k_{w}\nu^{2}+(4\ell^{2}-12\ell+72k_{P}\omega)\nu-24\omega k_{w}\ell+144ab\omega^{2},
ℓ=λ+(a2+b2)ω2+2kPω.(26)\ell=\lambda+(a^{2}+b^{2})\omega^{2}+2k_{P}\omega. \tag*{(26)}

Consequently, every joint eigenmode satisfies H(λ,ν)=0H(\lambda,\nu)=0.

To verify the angular action of the Killing–Yano symmetry, we evaluate FΩ[h]−Ω[Qh]F\Omega[h]-\Omega[Qh] without imposing the linearized vacuum equations. For this calculation we use the Riemann projection in Eq. (19) off shell. It agrees with the Weyl projection when δRab=0\delta R_{ab}=0 [5]. Let Tab=δRab[h]T_{ab}=\delta R_{ab}[h], pd=r+idyp_{d}=r+idy for d=±1d=\pm1, and Wr=−ωr2+kP+abkw/r2W_{r}=-\omega r^{2}+k_{P}+abk_{w}/r^{2}. Writing Rij=(FΩ[h]−Ω[Qh])ijR_{ij}=(F\Omega[h]-\Omega[Qh])_{ij}, we obtain

Rdd=0,R_{dd}=0,
Rd3=Y/2y[(12−y2pd2)TLL+y(r−idy)2pd(∂y+dW0Y)TLL]+y(∂r+2pd−iWrX)TLEd,R_{d3}=\frac{\sqrt{Y}/2}{y}\left[\left(\frac{1}{2}-\frac{y^{2}}{p_{d}^{2}}\right)T_{LL}+\frac{y(r-idy)}{2p_{d}}\left(\partial_{y}+\frac{dW_{0}}{Y}\right)T_{LL}\right]+y\left(\partial_{r}+\frac{2}{p_{d}}-\frac{iW_{r}}{X}\right)T_{LE_{d}},
R33=[2ab(1y2−1r2)+2ikwr]TLL+2y(∂r+2r−iWrX)TLE3.(27)R_{33}=\left[2ab\left(\frac{1}{y^{2}}-\frac{1}{r^{2}}\right)+\frac{2ik_{w}}{r}\right]T_{LL}+2y\left(\partial_{r}+\frac{2}{r}-\frac{iW_{r}}{X}\right)T_{LE_{3}}. \tag*{(27)}

Here the derivatives act on the complete contractions of TabT_{ab} with the background frame. For a vacuum perturbation, Tab=0T_{ab}=0, and hence Rij=0R_{ij}=0, establishing Ω[Qh]=FΩ[h]\Omega[Qh]=F\Omega[h].

Obtaining the angular and radial equations

At an ordinary latitude JA=λAJA = \lambda A gives

A′′=U1A+U2A′,U1=(λI−β0)/Y,U2=−β1/Y.(28)A^{\prime\prime}=U_1A+U_2A^\prime,\qquad U_1=(\lambda I-\beta_0)/Y,\qquad U_2=-\beta_1/Y. \tag*{(28)}

The evolution of (A,A′)T(A,A^\prime)^T and the action of FF on these data are represented by

Dy=(0IU1U2),Fdata=(α0α1α0′+α1U1α0+α1′+α1U2).(29)D_y=\begin{pmatrix}0&I\\U_1&U_2\end{pmatrix},\qquad F_{\mathrm{data}}=\begin{pmatrix}\alpha_0&\alpha_1\\\alpha_0^\prime+\alpha_1U_1&\alpha_0+\alpha_1^\prime+\alpha_1U_2\end{pmatrix}. \tag*{(29)}

Since [F,J]=0[F,J]=0, these matrices satisfy

Fdata′=[Dy,Fdata],(30)F_{\mathrm{data}}^\prime=[D_y,F_{\mathrm{data}}], \tag*{(30)}

so the characteristic polynomial of FdataF_{\mathrm{data}} is independent of yy. Direct evaluation at an ordinary latitude gives

det⁡(νI10−Fdata)=H(λ,ν)2,(31)\det(\nu I_{10}-F_{\mathrm{data}})=H(\lambda,\nu)^2, \tag*{(31)}

where I10I_{10} is the ten-dimensional identity matrix. Moreover, the operator identity H(J,F)=0H(J,F)=0 implies H(λ,Fdata)=0H(\lambda,F_{\mathrm{data}})=0. Thus, each simple root ν\nu of H(λ,ν)H(\lambda,\nu) corresponds to two independent local angular solutions. For ω≠0\omega\ne0 this count holds at every root, including a multiple one. If ν\nu has multiplicity mm in HH, the generalized eigenspace has dimension 2m2m by Eq. (31), while the annihilating polynomial bounds every Jordan block by mm. Hence the kernel has dimension at least two. Before imposing the azimuthal identifications, vary abab at fixed a2+b2a^2+b^2, yy, ω\omega, kPk_P, kwk_w, λ\lambda, ν\nu. At an ordinary latitude the matrix varies analytically and ∂(ab)H=144ω2≠0\partial_{(ab)}H=144\omega^2\ne0. The determinant therefore vanishes to order two in abab. For a kernel of dimension dd, a constant basis beginning with its dd kernel vectors makes dd columns vanish at the root, so the determinant vanishes to order at least dd. Hence d=2d=2.

For s=2A++/Ys=2A_{++}/Y put p=(2/Y)(1,0,0,0,0)p=(2/Y)(1,0,0,0,0). When ker⁡(Fdata−νI10)\ker(F_{\mathrm{data}}-\nu I_{10}) has dimension two, let the columns of EE be any basis for this kernel and define

U0=(p0p′p),L0=E(U0E)−1,det⁡(U0E)≠0.(32)U_0=\begin{pmatrix}p&0\\p^\prime&p\end{pmatrix},\qquad L_0=E(U_0E)^{-1},\qquad\det(U_0E)\ne0. \tag*{(32)}

Then (A,A′)T=L0(s,s′)T(A,A^\prime)^T=L_0(s,s^\prime)^T, so that A=u0s+v0s′A=u_0s+v_0s^\prime and s′′=c0s+c1s′s^{\prime\prime}=c_0s+c_1s^\prime with

(u0,v0)=(I,0)L0,(c0,c1)=(p′′+pU1,  2p′+pU2)L0.(33)(u_0,v_0)=(I,0)L_0,\qquad(c_0,c_1)=(p^{\prime\prime}+pU_1,\;2p^\prime+pU_2)L_0. \tag*{(33)}

The tensor equations continue AA through ordinary latitudes at which det⁡(U0E)=0\det(U_0E)=0.

Under the Carter interchange specified in the Letter, with the cyclic coordinates (T,P,w)(T,P,w) and Fourier labels (ω,kP,kw)(\omega,k_P,k_w) held fixed, the background quantities transform as

X~(r~)=Y(y),Y~(y~)=X(r),Σ~=−Σ,A~=B,B~=A,C~=C.(34)\widetilde{X}(\widetilde{r})=Y(y),\qquad\widetilde{Y}(\widetilde{y})=X(r),\qquad\widetilde{\Sigma}=-\Sigma,\qquad\widetilde{A}=B,\qquad\widetilde{B}=A,\qquad\widetilde{C}=C. \tag*{(34)}

These relations preserve the metric and the principal two-form. The oriented volume is also preserved, since the Jacobian of the (r,y)(r,y) transformation and the ratio r~y~Σ~/(ryΣ)\widetilde{r}\widetilde{y}\widetilde{\Sigma}/(ry\Sigma) are both −1-1. In particular, the principal null covector transforms as

L~♭=−B−iΣ~Y dy=p+Y/2E~+♭,p~+=−p+.(35)\widetilde{L}^{\flat}=-\mathcal{B}-\mathrm{i}\frac{\widetilde{\Sigma}}{Y}\,dy=\frac{p_+}{\sqrt{Y/2}}\widetilde{E}_+^{\flat},\qquad\widetilde{p}_+=-p_+. \tag*{(35)}

The transverse frame transforms accordingly as

E~+b=−X/2p+Lb.(36)\widetilde{E}_+^{b}=-\frac{\sqrt{X/2}}{p_+}L^b. \tag*{(36)}

Together with Eq. (35), this shows that the Riemann projection acquires a factor X/YX/Y under the interchange. The null-frame correction in Eq. (19) transforms with the same factor, giving

Ω~++=XYΩ++,2Ω~++X=2Ω++Y.(37)\widetilde{\Omega}_{++}=\frac{X}{Y}\Omega_{++},\qquad\frac{2\widetilde{\Omega}_{++}}{X}=\frac{2\Omega_{++}}{Y}. \tag*{(37)}

Thus the normalized component 2Ω++/Y2\Omega_{++}/Y is preserved by the interchange, allowing its angular equation to be continued directly into the radial equation.

The Carter interchange preserves the first-order eigenvalue, ν~=ν\widetilde{\nu}=\nu. Since the Fourier labels remain fixed, preserving ℓ\ell in Eq. (26) requires λ~=λ+μω2\widetilde{\lambda}=\lambda+\mu\omega^{2}. Let s~\widetilde{s} denote the angular scalar obtained with the continued parameters and eigenvalue λ~\widetilde{\lambda}. Setting R(r)=s~(ir)R(r)=\widetilde{s}(ir), the chain rule gives R′=is~′R'=i\widetilde{s}' and R′′=−s~′′R''=-\widetilde{s}'', where the tilded functions are evaluated at y~=ir\widetilde{y}=ir. Equation (33) then becomes

R′′(r)=−c~0(ir)R(r)+ic~1(ir)R′(r).(38)R''(r)=-\widetilde{c}_{0}(ir)R(r)+i\widetilde{c}_{1}(ir)R'(r). \tag*{(38)}

For generic ν≠0\nu\ne0, the spectral relation (S12) is quadratic in λ\lambda and admits two distinct branches. Although the construction above determines the corresponding angular and radial equations, it does not establish that their separated solutions arise from a vacuum metric perturbation. To identify the physical radial–angular pairings, we first reconstruct the metric from the tidal field and then impose the linearized Einstein equations.

Recovering the metric from the tidal field

We derive the curvature compatibility conditions and the local metric reconstruction stated in the Letter by comparing the finite-order derivatives of the metric and tidal curvature. At fixed Fourier labels, let hNh_N collect the 15 metric components and their rr- and yy-derivatives through total order NN. Each component contributes tN=(N+1)(N+2)/2t_N=(N+1)(N+2)/2 entries. Let ENE_N map hNh_N to the linearized Ricci tensor and its derivatives through order N−2N-2, and let GNG_N map displacement derivatives through order N+1N+1 to pure-gauge metric perturbations. Since ENGN=0E_NG_N=0 on the vacuum background, the dimension of the independent vacuum metric data modulo gauge is

dN=dim⁡(ker⁡EN/im⁡GN).(39)d_N=\dim\left(\ker E_N/\operatorname{im}G_N\right). \tag*{(39)}

We next count the independent tidal-curvature data produced by these vacuum metric jets. To express the curvature map with rational coefficients, we introduce

Tp=diag⁡(2p+2/Y,2p−2/Y,yp+p3/Y/2,yp−p3/Y/2,y2p32),TNhN+2=jet⁡n(TpΩ[h]),(40)T_p=\operatorname{diag}\left(2p_{+}^{2}/Y,2p_{-}^{2}/Y,yp_{+}p_{3}/\sqrt{Y/2},yp_{-}p_{3}/\sqrt{Y/2},y^{2}p_{3}^{2}\right),\qquad T_Nh_{N+2}=\operatorname{jet}^{n}\left(T_p\Omega[h]\right), \tag*{(40)}

where jet⁡n\operatorname{jet}^{n} collects all rr- and yy-derivatives through total order nn. The factors in TpT_p remove the denominators and square roots associated with the complex transverse frame. At an ordinary point, TpT_p is invertible, and the product rule relates the derivatives of TpΩT_p\Omega to those of Ω\Omega by a block triangular matrix with TpT_p on each diagonal block. Thus the weighting does not change the number of independent curvature data. We denote this number by

sN=rank⁡(TN−2∣ker⁡EN).(41)s_N=\operatorname{rank}\left(\left.T_{N-2}\right|_{\ker E_N}\right). \tag*{(41)}

The dimensions of the vacuum metric and tidal-curvature jets are summarized in Table S1. Since im⁡GN⊂ker⁡EN\operatorname{im}G_N\subset\ker E_N, Eq. (39) gives dN=15tN−rank⁡EN−rank⁡GNd_N=15t_N-\operatorname{rank}E_N-\operatorname{rank}G_N. For N=4,5N=4,5, we find sN<dNs_N<d_N, so the tidal-curvature data do not determine all vacuum metric data modulo gauge. At N=6N=6, the dimensions first coincide, s6=d6=65s_6=d_6=65, and the equality persists at higher orders shown in the table. Thus curvature derivatives through fourth order are sufficient to distinguish the vacuum metric jets modulo gauge at a generic ordinary point.

NNrank⁡EN\operatorname{rank}E_Nrank⁡GN\operatorname{rank}G_NdNd_NsNs_N
4751054530
51201405550
61751806565
72402257575
83152758585

Table S1. Ranks of the vacuum constraints, gauge transformations, and tidal-curvature map at a generic ordinary point. The equality sN=dNs_N=d_N first occurs at N=6N=6.

The contracted Bianchi identities impose five relations among the derivatives of the linearized Ricci tensor, giving the upper bounds

rank⁡EN≤15tN−2−5tN−3,sN≤min⁡(5tN−2,dN).(42)\operatorname{rank}E_N\le15t_{N-2}-5t_{N-3},\qquad s_N\le\min(5t_{N-2},d_N). \tag*{(42)}

The gauge map also satisfies rank⁡GN≤5tN+1\operatorname{rank} G_N \le5t_{N+1}. To establish the generic ranks, we evaluate the matrices at (a,b,μ,ω,mϕ,mψ)=(2/3,1/2,2,3/5,2,0)(a,b,\mu,\omega,m_{\phi},m_{\psi})=(2/3,1/2,2,3/5,2,0) and (r,y)=(3,3/5)(r,y)=(3,3/5). Exact calculations modulo 65537, with i→256i \to256, give nonzero minors of the maximal allowed sizes. Since these minors are rational functions of the parameters, their nonvanishing at this reference point establishes the corresponding generic lower bounds. These bounds saturate the upper bounds and give the ranks listed in Table S1. The rational weighting in Eq. (40) allows the calculation to be performed without introducing an extension of the finite field.

At N=6N=6, we combine the linearized Ricci equations and the curvature derivatives through fourth order into

A6=(E6T4),rank⁡A6=rank⁡E6+s6=240.(43)A_6=\begin{pmatrix} E_6 \\ T_4 \end{pmatrix}, \qquad\operatorname{rank} A_6=\operatorname{rank} E_6+s_6=240. \tag*{(43)}

This matrix has 300 rows and 420 columns. We select 240 independent rows, comprising 175 Ricci rows and 65 curvature rows, together with 240 columns such that the square submatrix M=A6[I,Lc]M=A_6[I,L_c] is invertible. The curvature rows are selected in order of increasing derivative order. For a prescribed tidal field Ω\Omega, let bI[Ω]b_I[\Omega] contain zeros in the selected Ricci rows and the corresponding derivatives of TpΩT_p\Omega in the selected curvature rows. The remaining ten curvature rows must then satisfy

[jet⁡4(TpΩ)]u=A6[u,Lc]M−1bI[Ω],(44)\left[\operatorname{jet}^{4}(T_p\Omega)\right]_u=A_6[u,L_c]M^{-1}b_I[\Omega], \tag*{(44)}

where uu labels the ten unselected curvature rows. These are the ten independent fourth-order compatibility conditions required for the tidal field to arise from a vacuum metric perturbation.

Since rank⁡A6=240\operatorname{rank} A_6=240, its kernel has dimension 420−240=180420-240=180, equal to rank⁡G6\operatorname{rank} G_6. Gauge invariance implies im⁡G6⊆ker⁡A6\operatorname{im}G_6\subseteq\ker A_6, and hence ker⁡A6=im⁡G6\ker A_6=\operatorname{im}G_6. Thus the Ricci and curvature equations determine the metric jet uniquely up to a gauge transformation.

To fix this freedom, we impose the five conditions U(hgf)=0U(h^{\mathrm{gf}})=0, where

U(h)=(hTT,hTP,hTw,hPP,hPw)T.(45)U(h)=(h_{TT},h_{TP},h_{Tw},h_{PP},h_{Pw})^{T}. \tag*{(45)}

For a displacement ξ\xi with the prescribed Fourier dependence, the transformation of these components is U(Lξg)=DξU(\mathcal{L}_{\xi}g)=D\xi, with

DAB,c=∂cgAB+ikAgcB+ikBgAc,(46)D_{AB,c}=\partial_cg_{AB}+ik_Ag_{cB}+ik_Bg_{Ac}, \tag*{(46)}

where (AB)=(TT,TP,Tw,PP,Pw)(AB)=(TT,TP,Tw,PP,Pw) and (kT,kP,kw)=(−ω,kP,kw)(k_T,k_P,k_w)=(-\omega,k_P,k_w). When DD is invertible, the choice ξ=D−1U(h)\xi=D^{-1}U(h) brings the metric into the desired gauge.

Let VhV_h denote the resulting linear map from a metric jet to the gauge-fixed metric components, h−LD−1U(h)gh-\mathcal{L}_{D^{-1}U(h)}g. Since VhV_h annihilates ker⁡A6\ker A_6, the local reconstruction is

hgf=Vh[:,Lc]M−1bI[Ω].(47)h^{\mathrm{gf}}=V_h[:,L_c]M^{-1}b_I[\Omega]. \tag*{(47)}

Here Vh[:,Lc]V_h[:,L_c] denotes the columns of VhV_h selected by LcL_c.

The reconstruction in Eq. (47) is valid where both DD and MM are invertible. In particular, when ω=mϕ=mψ=0\omega=m_{\phi}=m_{\psi}=0, all cyclic Fourier labels vanish. Since the background metric is independent of (T,P,w)(T,P,w), the corresponding three columns of DD then vanish, giving rank⁡D≤2\operatorname{rank}D\le2. The chosen gauge therefore does not provide a general reconstruction chart in this sector. An example is a stationary variation of the mass parameter μ\mu, which changes the ADM mass while leaving the principal tidal curvature zero. The same gauge can also fail for stationary nonaxisymmetric modes. When ω=kP=0\omega=k_P=0, the TT,TP,PPTT,TP,PP rows of Eq. (46) can have nonzero entries only in the r,yr,y columns, so DD is singular. For example, a=2ba=2b and (mϕ,mψ)=(1,−2)(m_{\phi},m_{\psi})=(1,-2) give kw=−3bk_w=-3b. When differentiating Eq. (47), derivatives must also act on the position-dependent matrices D−1D^{-1} and M−1M^{-1}.

We now show that the compatibility conditions (S30) are sufficient for local metric reconstruction. Differentiating these ten fourth-order equations once and twice gives additional conditions on the curvature derivatives through orders five and six. Exact minors of the highest-derivative matrices have ranks 10, 20, and 25 at orders four, five, and six, respectively. The space of curvature jets through sixth order satisfying these conditions therefore has dimension at most

5t6−(10+20+25)=140−55=85.(48)5t_6-(10+20+25)=140-55=85. \tag*{(48)}

On the other hand, Table S1 gives s8=85s_8=85 independent curvature jets generated by vacuum metric perturbations. Since these jets satisfy Eq. (44) and its derivatives, they saturate the bound. Consequently, every compatible curvature six-jet can be matched, at an ordinary point, by a vacuum metric eight-jet.

Let Ω\Omega be a C6C^{6} tidal field satisfying Eq. (44) on a reconstruction patch. The dimension count above implies that, at each point, there exists a metric eight-jet h8h_{8} satisfying E8h8=0E_{8}h_{8}=0 and T6h8=jet⁡6(TpΩ)\mathcal{T}_{6}h_{8}=\operatorname{jet}^{6}(T_{p}\Omega). The reconstruction map obeys the off-shell identity

Vhh6=Vh[:,Lc]M−1A6[I,:]h6,(49)V_{h}h_{6}=V_{h}[:,L_{c}]M^{-1}A_{6}[I,:]h_{6}, \tag*{(49)}

which follows from ker⁡A6=im⁡G6\ker A_{6}=\operatorname{im}G_{6} and the fact that VhV_{h} annihilates pure-gauge jets. Applying Eq. (49) and its first two derivatives to h8h_{8} shows that the two-jet of the reconstructed metric in Eq. (47) coincides with the gauge-fixed two-jet of h8h_{8}. Since the linearized Ricci tensor and the tidal curvature depend on at most two metric derivatives, we obtain

δRab[hgf]=0,Ω[hgf]=Ω.(50)\delta R_{ab}[h^{\mathrm{gf}}]=0,\qquad\Omega[h^{\mathrm{gf}}]=\Omega. \tag*{(50)}

Thus the ten compatibility conditions are locally sufficient for reconstructing a vacuum metric perturbation where the reconstruction matrices remain invertible.

For a separated tidal field Ωij=R(r)Aij(y)\Omega_{ij}=R(r)A_{ij}(y), the derivatives entering bI[Ω]b_{I}[\Omega] can be reduced using the scalar equations. Equation (33) expresses the angular components and their derivatives in terms of ss and s′s', while Eq. (38) reduces all higher radial derivatives to RR and R′R'. Substituting these relations into Eq. (47) gives the gauge-fixed Fourier amplitude as a linear combination of RsRs, Rs′Rs', R′sR's, and R′s′R's', yielding the reconstruction matrix Mab(r,y)M_{ab}(r,y) quoted in the Letter. The five components fixed by Eq. (45) vanish identically, and the remaining ten are determined without introducing additional independent functions. For any separated field satisfying the compatibility conditions, Eq. (50) ensures that the reconstructed metric satisfies the vacuum equations and reproduces the prescribed tidal curvature.

Physical pairing. The angular and radial scalar equations admit more separated products than are compatible with the linearized Einstein equations. The compatibility conditions derived above select the physical combinations. For generic ν≠0\nu\ne0, Eq. (26) has two distinct roots λ1\lambda_{1} and λ2\lambda_{2}. Let Aj(q)A_{j}^{(q)}, q=1,2q=1,2, denote independent angular solutions associated with λj\lambda_{j}, and let Ri(p)R_{i}^{(p)}, p=1,2p=1,2, denote independent radial solutions with λ~i=λi+μω2\widetilde{\lambda}_{i}=\lambda_{i}+\mu\omega^{2}. The four possible radial–angular pairings span

Eij=span⁡{Ri(p)Aj(q)∣p,q=1,2},i,j=1,2.(51)\mathcal{E}_{ij}=\operatorname{span}\left\{R_{i}^{(p)}A_{j}^{(q)}\mid p,q=1,2\right\},\qquad i,j=1,2. \tag*{(51)}

Each space has four independent local solutions. We call E11\mathcal{E}_{11} and E22\mathcal{E}_{22} the matched pairings, while E12\mathcal{E}_{12} and E21\mathcal{E}_{21} are the crossed pairings.

To determine which products admit vacuum metric reconstruction, we substitute them into the ten compatibility conditions DΩ=0D\Omega=0 defined by Eq. (44). All higher derivatives are eliminated using Eqs. (33) and (S24). For the reference black hole used in Table S1, we choose

ν=83,(λ1,λ2)=(−3943900,−563180).(52)\nu=\frac{8}{3},\qquad(\lambda_{1},\lambda_{2})=\left(-\frac{3943}{900},-\frac{563}{180}\right). \tag*{(52)}

At (r,y)=(3,3/5)(r,y)=(3,3/5), exact evaluation gives

(rank⁡D∣E11,rank⁡D∣E12,rank⁡D∣E21,rank⁡D∣E22)=(0,4,4,0),(53)\left(\operatorname{rank}\mathcal{D}|_{\mathcal{E}_{11}},\operatorname{rank}\mathcal{D}|_{\mathcal{E}_{12}},\operatorname{rank}\mathcal{D}|_{\mathcal{E}_{21}},\operatorname{rank}\mathcal{D}|_{\mathcal{E}_{22}}\right)=(0,4,4,0), \tag*{(53)}
rank⁡D∣⨁i,jEij=8.\operatorname{rank}\mathcal{D}\big|_{\bigoplus_{i,j}\mathcal{E}_{ij}}=8.

Thus the compatibility conditions eliminate all crossed products at the reference point. For the matched pairings, the four independent products in each space also satisfy the full reconstruction equations: substitution into the fifteen Ricci and five curvature residuals yields identically vanishing rational expressions in rr and yy. These identities establish the physical character of the eight matched solutions throughout the regular reference patch, rather than only at the point where the ranks were evaluated.

We next extend this result beyond the reference parameters. On a noncharacteristic surface r=r∗r=r_{*}, the linearized Einstein constraints, together with the gauge conditions (S31), determine the constrained metric data wherever the gauge matrix (S32) is invertible. The remaining initial data consist of five metric components and five conjugate momenta, specified as analytic functions of yy. Imposing the eigenvalue condition associated with the commuting symmetry QQ gives a system of linear ordinary differential equations for these functions. Local analytic Einstein evolution then allows the independent reference solutions to be continued order by order in the black-hole and Fourier parameters.

It remains to determine whether these continuations can develop crossed components. The rank-eight minor in Eq. (53) remains nonzero near the reference parameters. Consequently, any crossed amplitudes compatible with the constraints are determined by the matched amplitudes through rational linear relations. Since these relations vanish at the reference parameters, a nonzero crossed contribution would first appear at some positive order in the parameter expansion. Its leading coefficient would define a nonzero rational intertwiner between two distinct second-order scalar equations.

This possibility can be excluded using their pp-curvatures [6]. Reducing the reference equations modulo 1717, with i↦4i \mapsto4, gives the characteristic polynomials

z2−7,z2+7(angular),z2−7,z2+3(radial).(54)\begin{aligned} z^2 - 7,\quad&z^2 + 7 \quad\text{(angular)},\\ z^2 - 7,\quad&z^2 + 3 \quad\text{(radial)}. \tag*{(54)} \end{aligned}

The two polynomials in each pair are coprime. A rational intertwiner must also intertwine the corresponding pp-curvature matrices; their coprime characteristic polynomials therefore force the intertwiner to vanish. Hence no crossed contribution can arise in the formal continuation.

The eight continued vacuum solutions consequently remain in the matched spaces. Their compatibility residuals vanish to every order in the parameter expansion and therefore vanish identically as rational functions on the regular simple-root branches. Together with the nonvanishing rank-eight minor, this establishes

ker⁡D∣⨁i,jEij=E11⊕E22.(55)\left.\ker\mathcal{D}\right|_{\bigoplus_{i,j}\mathcal{E}_{ij}}=\mathcal{E}_{11}\oplus\mathcal{E}_{22}. \tag*{(55)}

By Eq. (50), these and only these local separated products admit vacuum metric reconstruction within the stated class.

For ν=0\nu= 0 and kwω≠0k_w\omega\ne0, Eq. (26) fixes ℓ=6abω/kw\ell= 6ab\omega/k_w and leaves a single angular branch. At the reference parameters, the four products associated with λ=3/20\lambda= 3/20 satisfy the compatibility conditions identically in rr and yy. Their analytic continuation gives the corresponding four physical solutions on regular reconstruction patches.

Completeness of the angular harmonics at real frequency

Preservation of the vacuum constraints. Completeness of the angular eigenfunctions does not by itself ensure that their spectral projections represent vacuum gravitational perturbations. We must also show that the projections preserve the compatibility conditions derived in Eq. (44). Let D\mathcal{D} denote these ten fourth-order differential operators acting on the tidal curvature Ω\Omega. Since JJ is second order, DJ\mathcal{D}J involves curvature derivatives through order six. We establish its vanishing on vacuum perturbations by constructing separated physical solutions that span the corresponding vacuum curvature jets.

At fixed λ\lambda, consider the angular and radial initial-data spaces together with the polynomial

H(λ,x)−H(λ,x′)x−x′,x=Fdata⊗I,x′=I⊗F~data,(56)\frac{H(\lambda,x)-H(\lambda,x')}{x-x'},\qquad x=F_{\mathrm{data}}\otimes I,\quad x'=I\otimes\widetilde{F}_{\mathrm{data}}, \tag*{(56)}

where the radial factor is evaluated at λ~=λ+μω2\widetilde{\lambda}=\lambda+\mu\omega^2. The quotient is a polynomial in the commuting matrices xx and x′x', so it does not require inverting x−x′x-x'. On distinct roots of H(λ,ν)H(\lambda,\nu) it vanishes, whereas on a matched root it equals ∂νH(λ,ν)\partial_\nu H(\lambda,\nu). Thus it selects the matched angular–radial subspaces whenever the roots are simple.

To test whether these local separated solutions span the vacuum curvature jets, we use Eqs. (33) and (38) to eliminate their higher derivatives through total order six. At the reference parameters, successively including the matched solutions at λ=0,1,2,3,4\lambda=0,1,2,3,4 and evaluating their jets of diag⁡(2/Y,2/Y,y/Y/2,y/Y/2,y2)Ω\operatorname{diag}(2/Y,2/Y,y/\sqrt{Y/2},y/\sqrt{Y/2},y^2)\Omega gives ranks 20, 40, 60, 80, 85 modulo 65537. An 85 × 85 minor has nonzero determinant 22410 in this field. These values of λ\lambda are used only for the local calculation; they need not satisfy the global angular regularity conditions.

Since s8=85s_8=85 by Table S1, the separated solutions span the full space of vacuum curvature six-jets wherever this minor remains nonzero. Each separated solution satisfies DΩ=0\mathcal{D}\Omega=0 and JΩ=λΩJ\Omega=\lambda\Omega, and therefore DJΩ=0\mathcal{D}J\Omega=0. Because DJ\mathcal{D}J depends only on curvature derivatives through order six, its vanishing on this spanning set implies that

DJΩ=0(57)\mathcal{D}J\Omega=0 \tag*{(57)}

for every curvature field arising from a local vacuum metric perturbation on this patch.

The spanning calculation establishes this identity where the chosen 85×8585 \times85 minor is nonzero. To extend it beyond that particular minor, write Ω=Wh\Omega= Wh, where WW is the metric-to-curvature map. Since DJWD_JW involves metric derivatives through order eight, the preceding identity is equivalent to

rank⁡(E8DJW)=rank⁡E8=315.(58)\operatorname{rank}\left(\frac{E_8}{D_JW}\right)=\operatorname{rank} E_8=315. \tag*{(58)}

After the invertible curvature weighting used above, the matrix entries are rational functions. The vanishing of the augmented 316×316316 \times316 minors therefore extends identically from the generic patch. Equation (57) continues to hold wherever the Ricci-jet matrix has rank 315 and the compatibility operators are defined, even if the particular separated-solution minor vanishes. The diagonal weight is invertible for b<y<ab<y<a.

Regularity at the rotation axes. For real ω\omega and a>b>0a>b>0, we first determine the inner product of the angular operators. On a section of fixed Boyer–Lindquist time tt, Eq. (16) gives ∣dϕ∧dψ∣=(a2−b2)∣dP∧dw∣\lvert d\phi\wedge d\psi\rvert=(a^2-b^2)\lvert dP\wedge dw\rvert. With the measure y dyy\,dy of the Letter, the component weight is

W=diag⁡(1,1,2,2,3/2).(59)W=\operatorname{diag}(1,1,2,2,3/2). \tag*{(59)}

For real Fourier parameters, the coefficients satisfy

Wα1=−α1TW,y(Wα0−α0TW)=(yWα1)′,W\alpha_1=-\alpha_1^{T}W,\qquad y(W\alpha_0-\alpha_0^{T}W)=(yW\alpha_1)',
KT=−K,y(Wβ0−β0TW)=K′,K=yWβ1−(yY)′W.(60)K^{T}=-K,\qquad y(W\beta_0-\beta_0^{T}W)=K',\qquad K=yW\beta_1-(yY)'W. \tag*{(60)}

Integration by parts gives the formal symmetry of FF and JJ in this inner product.

To determine the regularity conditions at the rotation axes, we introduce polar coordinates in the collapsing two-plane,

(ρ2,χ)={(a2−y2a2−b2,ϕ),y→a,(y2−b2a2−b2,ψ),y→b.(61)(\rho^2,\chi)= \begin{cases} \left(\dfrac{a^2-y^2}{a^2-b^2},\phi\right), & y\to a,\\ \left(\dfrac{y^2-b^2}{a^2-b^2},\psi\right), & y\to b. \end{cases} \tag*{(61)}

In terms of the Cartesian coordinate z=ρeiχz=\rho e^{i\chi}, the combinations e−idχEde^{-id\chi}E_d (d=±1d=\pm1) and E3E_3 extend smoothly across either axis. Indeed, the leading part of EdE_d in the collapsing plane is proportional to ∂ρ−idρ−1∂χ\partial_\rho-id\rho^{-1}\partial_\chi, whose rotating phase is removed by e−idχe^{-id\chi}.

For the five components ordered as (++,−−,+3,−3,33)(++ ,-- ,+3 ,-3 ,33), the corresponding angular weights are h=(2,−2,1,−1,0)h=(2,-2,1,-1,0). A Fourier mode with azimuthal number mm along the collapsing circle is therefore regular precisely when

Aj=ρ∣m−hj∣bj(ρ2),(62)A_j=\rho^{\lvert m-h_j\rvert}b_j(\rho^2), \tag*{(62)}

where bjb_j is smooth. To check the operator domain in this frame, act with FF or JJ on a Cartesian monomial zuzˉvz^u\bar z^v in component jj. Its rotating-frame amplitude is ρp\rho^p with p=u+vp=u+v and m=hj+u−vm=h_j+u-v. For axis y0=a,by_0=a,b,

∂yρpρp=pyy2−y02,∂y2ρpρp=py2−y02+p(p−2)y2(y2−y02)2.\frac{\partial_y\rho^p}{\rho^p}=\frac{py}{y^2-y_0^2},\qquad \frac{\partial_y^2\rho^p}{\rho^p}=\frac{p}{y^2-y_0^2}+\frac{p(p-2)y^2}{(y^2-y_0^2)^2}.

Substitution in Eqs. (24)–(25) gives each output component in the form (S48) at both axes. Taking p≤1p\leq1 for FF and p≤2p\leq2 for JJ, with arbitrary momentum on the surviving circle, successively tests all Cartesian coefficients of these first- and second-order operators.

We now establish the completeness of the angular eigenfunctions. The terms of β0\beta_0 quadratic in the azimuthal numbers are

−(a2−b2)(mϕ2a2−y2+mψ2y2−b2)I,(63)-(a^2-b^2)\left(\frac{m_\phi^2}{a^2-y^2}+\frac{m_\psi^2}{y^2-b^2}\right)I, \tag*{(63)}

while β1\beta_1 is independent of these numbers. Together with Y∂y2Y\partial_y^2, these terms give −J-J the positive scalar principal symbol of the Laplacian on the round S3S^3. Thus JJ is elliptic. Its formal symmetry and the smooth-axis conditions established above define a self-adjoint realization on the compact angular manifold, with a discrete real spectrum and finite-dimensional eigenspaces.

Since [F,J]=0[F,J]=0 and both operators preserve the smooth-axis domain, FF acts within each eigenspace of JJ. Its restriction is Hermitian and can therefore be diagonalized there. The resulting simultaneous eigenfunctions form a complete orthogonal basis in each integer Fourier sector for real ω\omega.

At ω=0\omega=0 the polynomial (S12) factors as ν(ν2+4λ)(ν2+λ−3)\nu(\nu^2+4\lambda)(\nu^2+\lambda-3). To locate its roots on the regular spectrum, put K=α1/Y\mathcal{K}=\alpha_1/\sqrt{Y}, U=[h,K]U=[h,\mathcal{K}], with the diagonal matrix hh of Eq. (24). Write cos⁡η=acos⁡θ/y\cos\eta=a\cos\theta/y, sin⁡η=bsin⁡θ/y\sin\eta=b\sin\theta/y and define

Dθ=∂θ+aby2K,Cϕ=cos⁡θ(cos⁡η h+sin⁡η U),Cψ=sin⁡θ(sin⁡η h−cos⁡η U).D_\theta=\partial_\theta+\frac{ab}{y^2}\mathcal{K},\qquad C_\phi=\cos\theta(\cos\eta\,h+\sin\eta\,U),\qquad C_\psi=\sin\theta(\sin\eta\,h-\cos\eta\,U).

In the component inner product K\mathcal{K} is skew-adjoint and Cϕ,CψC_\phi,C_\psi are self-adjoint. Substituting these expressions in the coefficients of J(0)J(0) and integrating by parts gives

(−λ−3)∥A∥2=∥DθA∥2+∥(mϕI−Cϕ)Asin⁡θ∥2+∥(mψI−Cψ)Acos⁡θ∥2.(-\lambda-3)\lVert A\rVert^2=\lVert D_\theta A\rVert^2+\left\lVert\frac{(m_\phi I-C_\phi)A}{\sin\theta}\right\rVert^2+\left\lVert\frac{(m_\psi I-C_\psi)A}{\cos\theta}\right\rVert^2.

The norms use the measure sin⁡θcos⁡θ dθ\sin\theta\cos\theta\,\mathrm{d}\theta and the component weights in Eq. (59). Axis regularity makes the endpoint term sin⁡θcos⁡θ A†WDθA\sin\theta\cos\theta\,A^\dagger W D_\theta A vanish. Thus λ≤−3\lambda\leq-3, excluding all possible coincidences λ=−1,0,3\lambda=-1,0,3. Each stationary eigenvalue pair admitted by the two-axis problem therefore has two independent local angular solutions.

Spectral projections of physical perturbations. Consider a tidal field u=Ωu=\Omega that is jointly analytic in the radial coordinate and in regular angular frames on I×S3I\times S^3, where II is an ordinary radial interval. Let Πλ\Pi_\lambda denote the spectral projector of JJ with eigenvalue λ\lambda. Since JJ is elliptic, analytic elliptic estimates imply exponential decay of the angular spectral coefficients [7].

More explicitly, choose c>sup⁡spec⁡Jc>\sup\operatorname{spec}J and define tλ=c−λt_\lambda=\sqrt{c-\lambda}. On any compact subinterval of II, Cauchy estimates give

∥(c−J)k∂rju∥≤C0(C1k)2k,0≤j≤4.(64)\left\lVert(c-J)^k\partial_r^j u\right\rVert\leq C_0(C_1k)^{2k},\qquad0\leq j\leq4. \tag*{(64)}

Projecting onto a JJ eigenspace and optimizing over kk then yields

∥Πλ∂rju∥≤Ce−ϵtλ,0≤j≤4.(65)\left\lVert\Pi_\lambda\partial_r^j u\right\rVert\leq Ce^{-\epsilon t_\lambda},\qquad0\leq j\leq4. \tag*{(65)}

with constants uniform on the chosen radial subinterval. Elliptic regularity and the polynomial growth of the angular spectrum ensure absolute convergence after applying any fixed number of angular derivatives. In particular, the spectral expansion may be differentiated and substituted into the physical equations term by term.

We now show that each angular spectral projection preserves the vacuum constraints. Let DD denote any row of the constraint operator D\mathcal{D}, and define dλ=(DΠλΩ)(r,y)d_\lambda=(D\Pi_\lambda\Omega)(r,y) at a fixed point. Since JJ preserves the vacuum constraints, we have

0=D(c−J)kΩ=∑λtλ2kdλ,k=0,1,2,…(66)0=D(c-J)^k\Omega=\sum_\lambda t_\lambda^{2k}d_\lambda,\qquad k=0,1,2,\ldots \tag*{(66)}

The exponential bound in Eq. (65) ensures that ∑λdλcos⁡(tλz)\sum_\lambda d_\lambda\cos(t_\lambda z) is holomorphic in a strip containing the real axis. Equation (S52) implies that all its derivatives at z=0z=0 vanish, so the function is identically zero.

Because the frequencies tλt_\lambda are positive and distinct for different eigenvalues, averaging against e−itλ0ze^{-it_{\lambda_0}z} along the real axis isolates dλ0d_{\lambda_0}. Hence

DΠλΩ=0(67)D\Pi_\lambda\Omega=0 \tag*{(67)}

for every λ\lambda. Each JJ projection therefore remains physical on the local reconstruction patch. Within a fixed JJ eigenspace, the projectors onto the eigenvalues of FF are polynomials in FF. Since FF is induced by a symmetry of the vacuum equations, these further projections also preserve the physical constraints. This selects the entire joint eigenspace when several harmonics share (λ,ν)(\lambda,\nu). On the matched patches of the preceding section, every angular solution on that branch carries the same radial equation. There the radial evolution is scalar on the angular multiplicity space, so any constant angular basis within it gives the same radial equation for each coefficient.

The analyticity required above follows for the tidal curvature of a smooth vacuum Fourier-mode metric perturbation in the open exterior. Away from the rotation axes, the principal symbol of the reduced wave operator in the (r,y)(r,y) directions is positive definite, since X>0X>0 and Y>0Y>0 for r>r+r>r_+ and b<y<ab<y<a.

At the axes, ellipticity can be established in regular Cartesian coordinates. The Gram determinants of the stationary Killing field and the surviving axial Killing field are

det⁡g∣(∂T,∂ψ)=−Δr2+a2,y=a,\det g|_{(\partial_T,\partial_\psi)}=-\frac{\Delta}{r^2+a^2},\qquad y=a,
det⁡g∣(∂T,∂ϕ)=−Δr2+b2,y=b,(68)\det g|_{(\partial_T,\partial_\phi)}=-\frac{\Delta}{r^2+b^2},\qquad y=b, \tag*{(68)}

where Δ=r2X>0\Delta=r^2X>0 outside the horizon. The Killing directions therefore span a Lorentzian two-plane, leaving a positive definite reduced principal symbol near either axis.

A local de Donder gauge can then be imposed by solving the corresponding elliptic equation for the gauge displacement. In this gauge, the linearized vacuum equations have analytic coefficients and a scalar elliptic principal part. Analytic elliptic regularity [8] implies that the gauge-fixed metric perturbation is locally analytic. The gauge-invariant tidal curvature Ω\Omega, expressed in regular local frames, consequently inherits joint analyticity throughout the open exterior.

This argument applies to perturbations represented by a smooth full metric. Smooth extension of a metric reconstructed from an individual angular eigenfunction requires a separate regularity analysis.

Radial boundary conditions. With dr∗/dr=(r2+a2)(r2+b2)/Δ\mathrm{d}r_*/\mathrm{d}r=(r^2+a^2)(r^2+b^2)/\Delta, regular advanced coordinates satisfy

dv=dt+dr∗,dϕin=dϕ+a(r2+b2)Δdr,dψin=dψ+b(r2+a2)Δdr.\mathrm{d}v=\mathrm{d}t+\mathrm{d}r_*,\qquad\mathrm{d}\phi_{\mathrm{in}}=\mathrm{d}\phi+\frac{a(r^2+b^2)}{\Delta}\mathrm{d}r,\qquad\mathrm{d}\psi_{\mathrm{in}}=\mathrm{d}\psi+\frac{b(r^2+a^2)}{\Delta}\mathrm{d}r.

for the nonextremal Myers–Perry horizon [9]. In these coordinates L=−∂rL=-\partial_r and the transverse frame is regular at an ordinary latitude. The Boyer–Lindquist Fourier factor contributes e+iωHr∗e^{+i\omega_H r_*} times a regular nonvanishing factor to the ingoing-coordinate amplitude. Since r∗=(2κH)−1log⁡(r−r+)+O(1)r_*=(2\kappa_H)^{-1}\log(r-r_+)+O(1), a smooth ingoing solution has R=(r−r+)−iωH/(2κH)R=(r-r_+)^{-i\omega_H/(2\kappa_H)} times a regular factor. If both local solutions are regular in the advanced frame at a resonant frequency, the ingoing branch is specified by continuation from nonresonant frequencies wherever that continuation exists.

At infinity the leading outgoing radiative curvature in a normalized frame falls as r−(5−2)/2r^{-(5-2)/2} [10]. Here LL approaches ∂t−∂r\partial_t-\partial_r, so its contractions retain the leading outgoing radiation for ω≠0\omega\ne0. The time convention of the Letter then gives Rout∝r−3/2eiωrR_{\mathrm{out}}\propto r^{-3/2}e^{i\omega r} to leading order for a nonzero radiative amplitude. For a complex frequency, the outgoing condition sets the incoming coefficient to zero by continuation of the outgoing solution [11]. At zero frequency one imposes stationary boundary conditions separately.

Single-spin limit

Set b=mψ=0b=m_\psi=0 and restrict to the reflection-even components (A−−,A++,A33)(A_{--},A_{++},A_{33}). The operator JJ preserves this sector, and FF interchanges the even and odd sectors. With x=y/ax=y/a and D=diag⁡(1/2,1/2,1)D=\operatorname{diag}(1/2,1/2,1) as in the Letter, the coefficient of each power of ∂x\partial_x in D−1JeDD^{-1}J_eD and D−1(F2)eDD^{-1}(F^2)_eD follows from Eq. (23) by the product rule. The three coefficients give the two angular operator identities in the Letter upon substitution of Eqs. (19) and (20) of Ref. [12] at n=1n=1.

With λS=λ+mϕ2+3\lambda_S=\lambda+m_\phi^2+3 and 3νS=ν2+ℓ−33\nu_S=\nu^2+\ell-3, the angular polynomial in Eq. (21) of Ref. [12] is exactly −νH(λ,ν)/9-\nu H(\lambda,\nu)/9. Thus ν=0\nu=0 also belongs to the even spectral curve. The identification uses (F2)e(F^2)_e and requires no division by ν\nu or by a difference of roots.

For the radial comparison, (t,ϕ)↦(−t,−ϕ)(t,\phi)\mapsto(-t,-\phi) exchanges the principal null directions, so (ωS,mS)=(−ω,−mϕ)(\omega_S,m_S)=(-\omega,-m_\phi). In the notation of the Letter X=r2+a2−μX=r^2+a^2-\mu is the function denoted by Δ\Delta in Ref. [12]. Its radial amplitudes can be chosen as

f=R=2A~++(ir)X,g=−A~33(ir)2r2.(69)f=R=\frac{2\widetilde{A}_{++}(ir)}{X},\qquad g=-\frac{\widetilde{A}_{33}(ir)}{2r^2}. \tag*{(69)}

With these substitutions the ++++ row of J~A~=λ~A~\widetilde{J}\widetilde{A}=\widetilde{\lambda}\widetilde{A} is precisely Eq. (7) of Ref. [12], including its source 6g6g. The two first-order equations are

(νS+χ)(f′g′)=−(PfPgQfRK(νS+χ)−χ′−Pf)(fg).(70)(\nu_S+\chi)\begin{pmatrix}f'\\g'\end{pmatrix} =-\begin{pmatrix}P_f&P_g\\Q_f&R_K(\nu_S+\chi)-\chi'-P_f\end{pmatrix} \begin{pmatrix}f\\g\end{pmatrix}. \tag*{(70)}

where primes now denote rr derivatives and all five coefficients are those of Eqs. (42)–(44) of Ref. [12].

To verify Eq. (70), bring both sides to the left, multiply by XJXJ of Eq. (39) of that reference, and express the result as two rows acting on (A~e,∂yeA~e)(\widetilde{A}_{e},\partial_{y_e}\widetilde{A}_{e}). Use ∂r=i∂ye\partial_r=i\partial_{y_e} on the continued angular functions. Each row is quadratic in ν2\nu^2. Substituting the even block of F~data2\widetilde{F}^{2}_{\mathrm{data}} from Eq. (29) for ν2\nu^2, with matrix powers acting on the right of their coefficient rows, gives the zero 2×62\times6 matrix identically in r,a,μ,ω,mϕ,λr,a,\mu,\omega,m_{\phi},\lambda. On joint data the even block has eigenvalue ν2\nu^2, proving both equations also at coincident roots. Eliminating gg from Eq. (7) of that reference with the first row of Eq. (70) therefore gives Eq. (38) for RR.

For completeness, the even characteristic polynomial is [zH(λ,z)]2[\sqrt{z}H(\lambda,\sqrt{z})]^2, by the off-diagonal parity blocks of FdataF_{\mathrm{data}} and Eq. (31). The polynomial zH(λ,z)\sqrt{z}H(\lambda,\sqrt{z}) annihilates the squared even block, by FdataH(λ,Fdata)=0F_{\mathrm{data}}H(\lambda,F_{\mathrm{data}})=0. Thus every root has at least two independent even data, by the same Jordan-block count as in Eq. (31). If RR vanishes on an ordinary interval, the ++++ and 3333 rows of the continued JJ equation successively force A~33=A~−−=0\widetilde{A}_{33}=\widetilde{A}_{--}=0. The map to RR is therefore injective and spans both solutions of Eq. (38) wherever its leading coefficient is nonzero.

For μ>a2>0\mu>a^2>0 the only additional scalar denominator in the open exterior is Santos’s Π\Pi. For ωS≠0\omega_S\ne0 its leading term at infinity is −ωS2r2/2-\omega_S^2r^2/2. For ωS=0\omega_S=0, mS≠0m_S\ne0, vanishing of its r−1r^{-1} and r−2r^{-2} coefficients would require both νS=−1\nu_S=-1 and a2−μ=a2mS2a^2-\mu=a^2m_S^2, which is impossible. Thus the comparison holds on every ordinary interval for all parameters of the Santos problem, with the tensor equation providing continuation through isolated zeros of Π\Pi. Ref. [12] excludes the stationary axisymmetric mode ωS=mS=0\omega_S=m_S=0.

† [email protected]

References

References

  1. [1]V. P. Frolov and D. Kubiznak, Hidden Symmetries of Higher Dimensional Rotating Black Holes, Phys. Rev. Lett. 98, 011101 (2007), arXiv:gr-qc/0605058.DOI
  2. [2]D. Rasheed, The Rotating dyonic black holes of Kaluza-Klein theory, Nucl. Phys. B 454, 379 (1995), arXiv:hep-th/9505038.DOI
  3. [3]B. Carter, Hamilton-Jacobi and Schrodinger separable solutions of Einstein’s equations, Commun. Math. Phys. 10, 280 (1968).DOI
  4. [4]P. Krtouš, V. P. Frolov, and D. Kubiznak, Hidden Symmetries of Higher Dimensional Black Holes and Uniqueness of the Kerr-NUT-(A)dS spacetime, Phys. Rev. D 78, 064022 (2008), arXiv:0804.4705 [hep-th].DOI
  5. [5]M. Durkee and H. S. Reall, Perturbations of higher-dimensional spacetimes, Class. Quant. Grav. 28, 035011 (2011), arXiv:1009.0015 [gr-qc].DOI
  6. [6]N. M. Katz, Nilpotent connections and the monodromy theorem: applications of a result of Turrittin, Publications Mathématiques de l’IHÉS 39, 175 (1970).DOI
  7. [7]R. T. Seeley, Eigenfunction expansions of analytic functions, Proceedings of the American Mathematical Society 21, 734 (1969).DOI
  8. [8]C. B. Morrey and L. Nirenberg, On the analyticity of the solutions of linear elliptic systems of partial differential equations, Communications on Pure and Applied Mathematics 10, 271 (1957).DOI
  9. [9]V. P. Frolov and D. Stojkovic, Quantum radiation from a five-dimensional rotating black hole, Phys. Rev. D 67, 084004 (2003), arXiv:gr-qc/0211055.
  10. [10]M. Godazgar and H. S. Reall, Peeling of the Weyl tensor and gravitational radiation in higher dimensions, Phys. Rev. D 85, 084021 (2012), arXiv:1201.4373 [gr-qc].DOI
  11. [11]E. Berti, V. Cardoso, and A. O. Starinets, Quasinormal modes of black holes and black branes, Class. Quant. Grav. 26, 163001 (2009), arXiv:0905.2975 [gr-qc].DOI
  12. [12]J. E. Santos, Separation of gravitational perturbations of singly spinning Myers-Perry black holes (2026), arXiv:2610.18956 [hep-th].arxiv.org/abs/2610.18956

Paper details

vlnc:W_wxjpejvr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents