WorksPhysics and Astronomy
Rotational Splitting of Reflectionless Matching and Virtual Absorption in a Semiopen BTZ Cavity
Abstract
We study complex frequency reflection zeros of a scalar field in a finite rotating BTZ region. The region is bounded by a partially reflecting near horizon wall and connected to an exterior propagation channel. We use exact radial solutions to determine the conditions for zero reflection and test their finite time behavior through wave packet evolution. Rotation splits the reflection zeros in the sectors. For a representative pair of branches, there is a range of reflection coefficients for a passive wall in which the co-rotating incident wave must grow in time, while the counter-rotating one must decay. In the near horizon limit, the magnitude of the wall reflection coefficient mainly controls the growth or decay rate, while its phase mainly controls the oscillation frequency. Exterior matching introduces corrections to these relations. We examine four truncated wave packets, including one tuned case and three reference cases. Tuning reduces the reflected energy fraction from – to , with little change in wall absorption or the peak energy stored in the cavity. We also find that two co-rotating branches merge at a second-order exceptional point (EP2). A near horizon expansion determines its location and the small wall reflectivity required for its formation. Under perturbations, the reflection zeros exhibit square-root splitting, while the poles of the local-source Green's function do not merge and its response remains linear. These results distinguish scattering zeros from the intrinsic resonances of the system.
Introduction
Quasinormal modes (QNMs) are the poles of the black hole Green’s function: the real and imaginary parts of their complex frequencies give, respectively, the oscillation frequency and the damping rate of the ringdown [1, 2, 3], and hence form the basis for spectroscopic tests of merger remnants [4, 5]; meanwhile, the residues of the QNMs determine the strength with which a given source excites each mode [6]. In asymptotically anti–de Sitter (AdS) spacetimes, the same set of QNM poles also governs the retarded boundary correlators [7, 8, 9]. Unlike the eigenvalue problem above, reflection zeros answer a different kind of scattering question: which incident waves undergo no reflection at all during scattering? If a partially reflecting surface is introduced near the location where a horizon would otherwise form [10, 11], this question becomes connected to a class of delayed-signal models; in such models, propagation, external scattering, and surface reflection together produce the delayed signal [12, 13].
For a specified incident direction, the condition of “vanishing reflection” defines the totally transmissive modes in the study of Schwarzschild and Kerr perturbations [14, 15, 16]. Reflectionless scattering theory formulates this condition in terms of complex frequency reflection zeros [17]; in a dissipative single channel system, zeros lying on the real frequency axis correspond to coherent perfect absorption [18]. When the zeros move off the real axis, their positions further dictate the growth or decay of the incident wave in time: an exponentially growing incident wave can exhibit coherent virtual absorption [19, 20]. Recent studies have, on the one hand, linked reflection zeros to echo phenomena [21, 22], and on the other hand examined black hole scattering as well as tailored incidence in the semi-open Schwarzschild–de Sitter geometry [23, 24].
Rotation breaks the equivalence between the two sectors of opposite azimuthal angular momentum; the same effect appears in the echoes of rotating bodies and in the totally transmissive modes of acoustic black holes [13, 25]. The rotating BTZ geometry admits exact solutions of the radial equation for a scalar field, so that its absorption problem can be treated analytically [26, 27, 28, 29]. In this setting, the radius-dependent combination allows us to distinguish the distinct roles of the wall, the propagation, and the external matching: the rotational correction cannot simply be reduced to a constant frequency shift. For this purpose, we retain a finite BTZ interval and connect it to an external channel with definite incident and reflected amplitudes, without imposing a boundary condition at AdS infinity [30, 31].1 The second question arises when two reflection zeros coalesce at an exceptional point (EP): the two then share a single eigenfunction, and their frequency splitting takes a square-root form [32, 33]. Previous studies have examined EPs in black hole QNM spectra [34, 35, 36, 37], including the response of the rotating BTZ [38] and EPs in absorption zeros [39]. Together with studies on the sensitivity of totally transmissive modes [40, 41], these results prompt us to investigate where the smooth tuning of reflection zeros breaks down.
After defining the scattering problem in Sec. II, we obtain three sets of results. First, rotation separates the critical wall coefficients corresponding to the paired reflection zeros; between the two thresholds of a representative paired branch, one matched incident wave grows with time while the other decays (Sec. III). Second, for a truncated wave packet, tuning to the reflection zero reduces the reflected fraction from – to , while the wall absorption and the peak energy in the region remain almost unchanged (Sec. IV). Third, the two corotating reflection-zero branches merge at an EP, whose location and the requisite small wall reflection coefficient can both be obtained from a near-horizon inverse mapping; we also compare their square-root splitting with the response induced by a localized source (§ 1.4).
Scattering in a rotating BTZ cavity
We study scattering through a finite interval of the rotating BTZ geometry (the BTZ interval), bounded by a partially reflecting inner wall and by an exterior matching surface that connects to an idealized propagation channel (Figure 1). At a reflection zero (a frequency at which the reflection vanishes), a purely incident exterior wave, after propagating through the cavity, satisfies the wall condition.
Figure 1. The BTZ interval extends from the partially reflecting wall at to the exterior matching surface at , with exterior amplitudes and . The region used for energy integration, with , includes part of the exterior channel.
The rotating BTZ metric, with AdS radius and horizon radii , is
For a scalar field satisfying , the ansatz , with and , gives
Here is the tortoise coordinate and ; as the outer horizon is approached, and at the AdS boundary of the untruncated geometry. We introduce the dimensionless variables , , and ; the horizon angular velocity is .
The BTZ interval is restricted to . At the near-horizon wall, we impose
In the local wave decomposition associated with this Robin condition, is the wall amplitude reflection coefficient and is the flux reflectivity. We use a real, frequency-independent for wave packets and allow complex in reflection-phase studies. A frequency-dependent membrane response provides a possible extension [42].
For , freezing and at their matching-surface values defines the exterior channel,
We take on the real-frequency propagation branch and analytically continue this branch to complex frequencies. Reflectionless incidence, with , requires
Let and denote the solutions obtained by continuing unit-amplitude, purely incident and purely outgoing exterior waves inward, respectively. Their scaled wall residuals give
Along regular analytic branches away from , with corresponds to reflectionless incidence, whereas with corresponds to purely outgoing resonances (). These conditions are analogous, respectively, to the TTM and QNM boundary prescriptions in black-hole scattering [14, 15, 43, 44]. Together, the frequency-dependent equation and the boundary conditions define a non-self-adjoint nonlinear spectral problem. The potential, exact radial solutions and scattering amplitudes are given in Supplement A. This scattering problem differs from the BTZ brick-wall models, which impose an inner Dirichlet wall and normalizability at AdS infinity [45, 46].
A complex reflection zero fixes both the oscillation frequency and the incident envelope through . Upper-half-plane zeros select exponentially growing incident waves, termed virtual absorption modes. For real and , with , and , we have , and flux conservation then gives
| Physical quantity | Reference value |
| Geometry | , <br> |
| Scalar field | |
| Azimuthal sectors | |
| Inner boundary | |
| Matching surface | |
| Wall reflection coefficient | (spectral catalog)<br> (wave packets) |
| Observation point |
Table 1. Reference parameters. Changes are stated where used.
A real-frequency reflection zero therefore yields perfect monochromatic absorption at the wall. Finite-duration wave packets are studied in Sec. IV. Reference parameters are listed in Table I.
Unless stated otherwise, spin scans keep and fixed, so both the circumferential matching radius and the cavity length in tortoise coordinates vary with spin. Frequency shifts therefore include changes in both radial propagation and coupling to the exterior channel. Numerical methods and coverage checks are summarized in Supplement C.1.
Rotation and wall control of reflection zeros
Rotation splits the reflection zeros for through the radial term , which changes sign with and produces different propagation amplitudes and phases. At , the two sectors are degenerate because the radial equation and boundary conditions depend only on . Continuation from this common limit identifies the paired branches. We label them by in order of increasing nonrotating frequency and use as a representative pair; the correspondence to the numerical catalog is given in the Supplemental Material. The radial dependence of means that rotation changes propagation throughout the interval, not just the frequency measured at the wall.
A real-frequency reflection zero corresponds to constant-envelope incidence and, for a passive wall, critical coupling [17, 18]. Figure 2 shows distinct critical wall reflection magnitudes: and at . The symmetry implies , where , for regular, locally unique critical crossings. Consequently,
Figure 2. Critical coupling and imaginary frequencies for . Shading marks the interval between and . The black dot and open circles identify the wave-packet parameters at ; the horizontal line in (b) marks .
For , and , as determined by radial matching (Supplement A.2). The resulting interval
selects an exponentially growing corotating wave and a decaying counterrotating wave under the convention. The reference choice lies in this interval and is used for the wave-packet comparisons.
The imaginary parts of the zeros in Figure 3 vary differently with spin on different radial branches. At , four lower corotating branches cross the real axis twice: their reflectionless incident waves change from decaying to growing and then back to decaying as increases. The branch crosses once, while several higher branches do not cross. The real parts also evolve differently between sectors: corotating frequencies can turn, whereas the counterrotating frequencies decrease over the displayed range. The incident frequency and growth rate must therefore be chosen for a particular branch, even when the same wall is used.
Figure 3. Spin evolution of twelve tracked reflection zeros per panel. Arrows indicate increasing ; small circles and triangles mark the endpoints , , and larger markers identify reference pairs at . The corotating branches form the EP pair, while is used for the wave packets. Labels are fixed along each trajectory.
All six pairs in Figure 4 satisfy for , with pair dependent splitting and turning points. For the representative pair, this ordering also survives attachment through a smooth transition layer (Supplemental Table V), displacements of and (Supplemental Table IV), and scans at fixed circumferential matching radius instead of fixed (Supplemental Fig. 12). The critical values shift in each comparison. Their ordering is a result for these branches and geometries, rather than a consequence of rotation-reversal symmetry.
Figure 4. Critical wall reflection magnitudes for six branch pairs. Solid and dashed curves denote and ; colors distinguish branches, with coral marking . Each pair is degenerate at .
The dependence on follows from the near-horizon radial solutions. Let denote the ratio of outgoing to ingoing horizon-basis coefficients of the solution that is purely incident at [Eq. (A13)]. To leading order in , the wall condition becomes
Its modulus gives , while its phase gives , with fixed along a local phase branch. At a real-frequency zero, the modulus equation reduces to the critical-coupling condition . It is the exterior coefficient, together with the phase condition, that determines which wall reflection magnitude gives zero reflection.
For a regular simple zero with , implicit differentiation gives
Here . The logarithm is taken on a local analytic branch; the exact first line also retains the finite-wall corrections. When , this yields . Thus primarily changes the growth or decay rate, and changes the oscillation frequency, on the same scale . Figure 5 shows these slopes and the sector-dependent deviations due to . The delay scale is even in ; the directional dependence remains in both and . The propagation phase is related to those entering echo transfer functions [12, 13], although is not a measured echo spacing.
Figure 5. Wall control at for , , . Frequency changes are scaled by . Dashed lines show the leading near-horizon slopes and .
Wave-packet response and energy balance
A truncated packet has a finite bandwidth on the real axis, so its reflection depends on across that band and not on the complex zero alone [19, 23, 24]. Figure 6 shows the reflectance together with the incident spectra of the four packets defined below. We retain propagating components with and (Supplement C.3).
Figure 6. Real-frequency reflectance and incident spectra at . The upper panels show for , ; the middle panels show the energy-normalized incident spectra of four packets evolved at . The lower maps show wall dependence, with the two settings marked. Gray regions are nonpropagating; white circles mark refined real-frequency zeros. The map floor is .
At the observation point, incident and outgoing amplitudes are separated by Eq. (A4), including during temporal overlap. For , the incident energy and predicted full-record reflected energy are
Here is the initial total energy on . All retained frequencies satisfy , giving a positive spectral weight. The plotted density obeys . The measured full-record reflected energy follows by replacing with in Eq. (12), where . The propagation phase has unit modulus on the retained band.
We compare four packets with the same initial envelope and truncation at : an packet tuned to the growing zero ; an packet with a detuned oscillation frequency and the same growth rate; an packet at ; and an packet with the oscillation frequency of its own zero but the same positive growth rate. The last packet is not tuned to that zero, which requires a decaying wave. We use each packet’s measured incident spectrum and its group-velocity-dependent incidence interval (Supplement C.3). Table II and Table III report the reflected energy during incidence and over the full record, including the response to truncation, respectively.
| Packet | Reflected energy during incidence | |
| , zero tuned | 0.999971 | |
| , detuned | 0.999978 | |
| , same complex frequency | 0.999994 | |
| , branch aligned same growth rate | 0.999989 |
Table II. Incident and reflected energy during incidence, normalized by , at . The interval runs from the packet front reaching the wall until before the truncated packet center returns; its endpoints depend on group velocity.
| Packet | Full-record reflection | Spectral prediction |
| , zero tuned | ||
| , detuned | ||
| , same complex frequency | ||
| , branch aligned same growth rate |
Table III. Full-record reflected energy and spectral predictions at , normalized by . The record includes the response after truncation.
The zero-tuned packet reflects least by either measure (Table II and Table III). Over the full record, it reflects of its initial energy, compared with for the three controls, a reduction by factors of . Equation (13), evaluated with each measured incident spectrum, reproduces all four full-record values to within . For the counterrotating controls, moving the oscillation frequency onto the branch lowers the reflected energy during incidence from to of the initial energy, but raises the full-record fraction from to . This packet retains a growing envelope, whereas the zero requires decay; the comparison tests a change of oscillation frequency at fixed growth rate.
The energy between the wall and the observation point is , where and . The integration interval includes the exterior segment . With positive flux toward increasing , conservation gives
Figure 7?? plots and . The four histories are nearly indistinguishable: the regional energy peaks near , and more than of has crossed the wall by the end of each record. For all four packets, the weakly reflecting wall absorbs nearly all the energy, so tuning to the zero changes the reflected remainder of order while barely changing wall loss or peak regional energy.
Figure 7. Energy evolution for four packets at , normalized by their respective . In (a), open endpoints show spectral predictions for the full-record reflected energy. Panel (b) shows regional energy changes and cumulative wall loss ; the region includes part of the exterior channel. Shading marks the zero-tuned packet’s incidence interval.
The wall is passive in the sense , with [Eq. (C5)]. This differs from the flux because ; the cumulative loss is nevertheless nonnegative for all four packets. Spatial refinement changes the full-record reflected energies by at most , with their ordering unchanged (Supplement C.3).
Exceptional point of the reflection zeros
Coalescence and local structure
Two corotating reflection zeros, , coalesce at , , and . The negative imaginary part requires an exponentially decaying incident wave; the growing wave packets of Sec. IV use the distinct pair. As decreases at , the higher-frequency zeros in Figure 8 descend with shifts in smaller than 0.008. The low-frequency pair approaches, coalesces, and separates primarily in : one decay rate decreases while the other increases. Beyond the coalescence, the plotted continuation denotes the less-damped member and the more-damped member. This convention follows the two local square-root signs; the straight real- path does not uniquely identify an individual branch through the EP.
Figure 8. Reflection-zero trajectories at and , with decreasing from to along the negative real axis. Panel ?? shows six branches; panel ?? shows the imaginary parts of the pair. The black diamond marks the EP. Solid and dashed segments lie at and , respectively.
The coalescence occurs where the ordinary frequency-control relation of Sec. III ceases to be invertible. Since with , a double zero at fixed is a stationary point of . In the near-horizon approximation, at this point: the exterior term is now as large as the near-horizon term in Eq. (11). Near a simple zero of , write , where . The stationary point of then gives
At , and , giving . The negative exponentially suppresses the required . Along the cavity-length family below, the leading prediction for the complex differs from the numerical value by 10–11%; including the curvature of reduces this to 1.2–1.4% (Supplement B.1). In this leading approximation, is also the zero for . The finite-position Robin condition adds an correction, so the exact zero at is instead defined by .
At the numerical EP,
The purely incident condition at fixes the radial solution up to normalization, so the double root has a single eigenfunction. The frequency-dependent boundary-value problem has a Jordan chain of length two, including the derivatives of both boundary conditions [47]; Supplement B gives the proof. Expanding at fixed yields
A circuit around the EP in the real plane exchanges the two zeros; two circuits restore them (Supplemental Fig. 14). For a simple zero, the sensitivity in Eq. (11) consequently diverges as , where at and . The pink spin sweep in Supplemental Fig. 13 passes through the same EP.
Continuation in cavity length
The double-root conditions are four real equations for the complex frequency and the controls. At fixed wall position, the real spin and real negative provide the two controls needed for an isolated EP. Varying the wall position traces the line in Figure 9, with the circumferential matching radius held fixed. For –26.54, increases from 0.533 to 0.639 and increases by about 36%. If the spin is instead fixed at and is complex, and are independent controls; over the same range of wall positions the required magnitude decreases from to .
Figure 9. (a) Spin determined by the real wall condition. (b) Reflection magnitude under two constraints. Dependence of the reflection-zero EP on cavity length, with and the remaining field parameters fixed as in Table I. Blue circles give the negative real path, with both the spin and the reflection magnitude retuned to maintain the double root. The cavity length is . The black point marks the reference wall position . Magenta squares correspond to with complex , included as a mathematical comparison. Markers are EP solutions and lines connect neighboring wall positions.
The difference follows from the phase of the same near-horizon map. Keeping negative and real forces the spin to change with the wall position through Eq. (B10). Along this path, increases enough to outweigh the length and damping terms in Eq. (B11). The independently adjustable phase of complex removes this constraint. The comparison covers the stated finite range of wall positions. The EP also persists at shifted parameters when a smooth transition layer connects the BTZ region to the exterior channel (Supplemental Table VI).
Response to a localized source
For a reduced unit source at and an observation point in the exterior channel, the retarded Green function is
where satisfies Eq. (3) and . Its denominator is the outgoing-wave function in [12]. At fixed geometry and frequency, both and are affine in , so the source response is analytic in wherever . With this normalization, and the sampled minimum of in Figure 10 is 0.118. The reflection zeros merge in the upper row, while the lower row shows the Green function in the same lower-half-plane window.
Figure 10. Reflection zeros and the analytically continued retarded Green function near the EP, with , , and , at the fixed reference EP spin and geometry. The upper row shows , with a display floor of for the reflection amplitude. Circles mark simple zeros, and diamonds mark double zeros. The lower row shows using a common color scale and the same column parameters as the upper row. All panels cover the same window in the lower half of the complex-frequency plane.
For and , the relative band norm of Eq. (B12), integrated over , scales linearly with for . Over that interval the half splitting of the zeros scales as . Thus the square-root displacement of the zeros does not describe the perturbation of the field generated by a fixed source. At larger changes in , Figure 11 compares both the amplitude and phase of this field. For and , the responses at and differ from the response by about 32% and 1.6%, respectively, in the same band norm. The oscillations and complex plane loops are more pronounced at ; the two smaller magnitudes give nearly coincident curves. The source normalization and independent propagation check are given in Supplement B.2. For the paired branches considered here, the corotating and counterrotating modes possess different critical wall reflection coefficients, so that there exists a parameter range in which the corotating incident wave required for zero reflection grows with time while the counterrotating incident wave decays with time. A near horizon analysis further reveals how the zero frequencies depend on the logarithmic amplitude and the phase of the wall reflection coefficient. In the time domain, we find that tuning a truncated incident wave packet to a reflection zero further suppresses the residual reflection, but leaves the wall absorption and the energy accumulation in the cavity almost unchanged. This indicates that tuning to a reflection zero mainly controls the reflected fraction in the external scattering channel, rather than the overall energy dissipation of the system.
Figure 11. Response to a localized source at , , and . The upper panels show amplitude spectra at real frequencies, while the lower panels show trajectories of the complex Green function as increases from 0.55 to 2.5. Open circles and crosses mark the two ends of the frequency interval; squares and diamonds identify the same two sampled frequencies in all panels, and open arrowheads indicate the direction of increasing . All panels use the same three wall settings and normalization to a reduced unit source, but the axis ranges differ between the two sectors. In both sectors, denotes the wall coefficient corresponding to the EP.
Conclusions
This paper studies the scattering of a scalar field by a partially reflecting surface located near the horizon in the rotating BTZ spacetime, and determines the conditions for the existence of reflection zeros, the effects of rotation, and their exceptional point (EP) structure. We find that, even when the near horizon wall has nonzero reflectivity, the scattering system as a whole can achieve zero reflection for incident waves of specific complex frequencies. This shows that reflection zeros are not determined solely by the local reflective properties of the wall; rather, they arise from the combined action of the near horizon boundary condition, wave propagation, and external scattering.
Rotation significantly modifies the distribution of reflection zeros and their time-domain behaviour. For the paired branches considered here, the corotating and counterrotating modes possess different critical wall reflection coefficients, so that there exists a parameter range in which the corotating incident wave required for zero reflection grows with time while the counterrotating incident wave decays with time. A near horizon analysis further reveals how the zero frequencies depend on the logarithmic amplitude and the phase of the wall reflection coefficient.
Another main result of this paper is the finding that two corotating reflection-zero branches can merge at an EP and exhibit a characteristic square-root frequency splitting. Using a near horizon inverse mapping, we determine the location of the EP and the corresponding wall reflection condition. Notably, the merging of reflection zeros does not imply that the poles of the Green’s function for a localized source also merge: the latter are still determined by an independent outgoing boundary condition. Therefore, the spectral sensitivity embodied by an EP of reflection zeros cannot be identified directly with an enhancement of the signal excited by a physical source.
These results reveal a class of nontrivial scattering structures arising from the combined action of near horizon reflection and rotation, and they also illustrate the importance of distinguishing scattering zeros from the intrinsic resonances of the system. This paper adopts a finite BTZ interval and prescribed external matching conditions, so the critical reflection coefficients and the specific location of the EP are not universal. It would be worthwhile in the future to extend this analysis to black hole geometries with more realistic external scattering potentials, and to introduce a frequency dependent wall response that respects causality, in order to study how reflection zeros and their EPs affect the signals produced by physical sources, in particular the observable late-time echo structure [12, 13, 21, 42].
Acknowledgments
This work was supported by the National Natural Science Foundation of China (Grants No. 12547143, 12305076, 12405055 and 12305063), the Key Project of the Department of Education of Hunan Province (No. 25A0084), the Hunan Provincial Furong Plan Youth Talent Support Program (Grant No. 2026ZRB-QN100).
References
- [1]K. D. Kokkotas and B. G. Schmidt, Living Rev. Rel. 2, 2 (1999), arXiv:gr-qc/9909058.DOI
- [2]E. Berti, V. Cardoso, and A. O. Starinets, Class. Quant. Grav. 26, 163001 (2009), arXiv:0905.2975 [gr-qc].DOI
- [3]R. A. Konoplya and A. Zhidenko, Rev. Mod. Phys. 83, 793 (2011), arXiv:1102.4014 [gr-qc].DOI
- [4]O. Dreyer, B. Kelly, K. Krishnan, L. S. Finn, D. Garrison, and R. Lopez-Aleman, Class. Quant. Grav. 21, 787 (2004), arXiv:gr-qc/0309007.DOI
- [5]E. Berti, V. Cardoso, and C. M. Will, Phys. Rev. D 73, 064030 (2006), arXiv:gr-qc/0512160.DOI
- [6]E. Berti and V. Cardoso, Phys. Rev. D 74, 104020 (2006), arXiv:gr-qc/0605118.DOI
- [7]G. T. Horbatsch and V. E. Hubeny, Phys. Rev. D 62, 024027 (2000), arXiv:hep-th/9909056.
- [8]P. K. Kovtun and A. O. Starinets, Phys. Rev. D 72, 086009 (2005), arXiv:hep-th/0506184.DOI
- [9]D. T. Son and A. O. Starinets, JHEP 2002 (09), 042, arXiv:hep-th/0205051.
- [10]V. Cardoso, E. Franzin, and P. Pani, Phys. Rev. Lett. 116, 171101 (2016), arXiv:1602.07309 [gr-qc].DOI
- [11]C. P. Burgess, R. Plestid, and M. Rummel, JHEP 2018 (09), 113, arXiv:1808.00847 [gr-qc].DOI
- [12]Z. Mark, A. Zimmerman, S. M. Du, and Y. Chen, Phys. Rev. D 96, 084002 (2017), arXiv:1706.06155 [gr-qc].DOI
- [13]E. Maggio, A. Testa, S. Bhagwat, and P. Pani, Phys. Rev. D 100, 064056 (2019), arXiv:1907.03091 [gr-qc].DOI
- [14]A. Maassen van den Brink, Phys. Rev. D 62, 064009 (2000), arXiv:gr-qc/0001032.DOI
- [15]G. B. Cook, L. S. Annichiarico, and D. J. Vickers, Phys. Rev. D 99, 024008 (2019), arXiv:1808.00987 [gr-qc].DOI
- [16]G. B. Cook and S. Lu, Phys. Rev. D 107, 044043 (2023), arXiv:2211.14955 [gr-qc].DOI
- [17]W. R. Sweeney, C. W. Hsu, and A. D. Stone, Phys. Rev. A 102, 063511 (2020), arXiv:1909.04017 [physics.optics].DOI
- [18]Y. D. Chong, L. Ge, H. Cao, and A. D. Stone, Phys. Rev. Lett. 105, 053901 (2010), arXiv:1003.4968 [physics.optics].DOI
- [19]D. G. Baranov, A. Krasnok, and A. Alù, Optica 4, 1457 (2017), arXiv:1710.09345 [physics.optics].arxiv.org/abs/1710.09345
- [20]Y. Ra’di, A. Krasnok, and A. Alù, ACS Photonics 7, 1468 (2020), arXiv:2002.00487 [physics.optics].
- [21]R. F. Rosato, S. Biswas, S. Chakraborty, and P. Pani, Phys. Rev. D 111, 084051 (2025), arXiv:2501.16433 [gr-qc].DOI
- [22]W.-L. Qian, Q. Pan, R. G. Daghigh, B. Wang, and R.-H. Yue, Phys. Rev. D 113, 124021 (2026), arXiv:2511.00565 [gr-qc].arxiv.org/abs/2511.00565
- [23]F. Tuncer, V. Cardoso, R. P. Macedo, and T. F. M. Spieksma, Phys. Rev. D 113, 104028 (2026), arXiv:2509.19451 [gr-qc].arxiv.org/abs/2509.19451
- [24]L.-B. Wu, Y.-S. Zhou, Z. Yu, M.-F. Ji, and L.-M. Cao, Phys. Rev. D 10.1103/lcwf-4bkn (2026), accepted for publication, arXiv:2603.22897 [gr-qc].arxiv.org/abs/2603.22897
- [25]Z. Yu and L.-B. Wu, Total transmission modes in draining bathtub model with vorticity (2026), arXiv:2605.03730 [gr-qc].arxiv.org/abs/2605.03730
- [26]M. Bañados, C. Teitelboim, and J. Zanelli, Phys. Rev. Lett. 69, 1849 (1992), arXiv:hep-th/9204099.DOI
- [27]V. Cardoso and J. P. S. Lemos, Phys. Rev. D 63, 124015 (2001), arXiv:gr-qc/0101052.DOI
- [28]D. Birmingham, I. Sachs, and S. N. Solodukhin, Phys. Rev. Lett. 88, 151301 (2002), arXiv:hep-th/0112055.DOI
- [29]D. Birmingham, I. Sachs, and S. Sen, Phys. Lett. B 413, 281 (1997), arXiv:hep-th/9707188.
- [30]A. Ishibashi and R. M. Wald, Class. Quant. Grav. 21, 2981 (2004), arXiv:hep-th/0402184.DOI
- [31]C. Dappiaggi, H. R. C. Ferreira, and C. A. R. Herdeiro, Phys. Lett. B 778, 146 (2018), arXiv:1710.08039 [gr-qc].DOI
- [32]T. Kato, Perturbation Theory for Linear Operators (Springer, Berlin, 1995).
- [33]W. D. Heiss, J. Phys. A 45, 444016 (2012), arXiv:1210.7536.DOI
- [34]H. Motohashi, Phys. Rev. Lett. 134, 141401 (2025), arXiv:2407.15191 [gr-qc].DOI
- [35]R. Panosso Macedo, T. Katagiri, K.-i. Kubota, and H. Motohashi, Phys. Rev. D 113, L121504 (2026), arXiv:2512.02110 [gr-qc].DOI
- [36]J. P. Cavalcante, M. Richartz, and B. C. da Cunha, Phys. Rev. Lett. 133, 261401 (2024), arXiv:2407.20850 [gr-qc].DOI
- [37]N. Nakamoto and N. Oshita, Phys. Rev. D 113, 104032 (2026), arXiv:2601.00704 [gr-qc].DOI
- [38]S. Long, Y. Shi, Z. Xia, H. Gong, Z. Cao, Q. Tan, and Q. Pan, Exceptional points and near-extremal boundary response in rotating BTZ black holes (2026), arXiv:2609.30612 [gr-qc].arxiv.org/abs/2609.30612
- [39]W. R. Sweeney, C. W. Hsu, S. Rotter, and A. D. Stone, Phys. Rev. Lett. 122, 093901 (2019), arXiv:1807.08805 [physics.optics].DOI
- [40]Y.-S. Zhou, M.-F. Ji, L.-B. Wu, and L.-M. Cao, Phys. Rev. D 113, 124042 (2026), arXiv:2512.16372 [gr-qc].DOI
- [41]X.-N. Dong, L.-B. Wu, Y.-S. Zhou, and Z.-K. Guo, The (in)stability on total transmission modes with small bumps (2026), arXiv:2609.07167 [gr-qc].arxiv.org/abs/2609.07167
- [42]S. Chakraborty, E. Maggio, A. Mazumdar, and P. Pani, Phys. Rev. D 106, 024041 (2022), arXiv:2202.09111 [gr-qc].DOI
- [43]N. Andersson, Class. Quant. Grav. 11, L39 (1994).
- [44]G. B. Cook and M. Zalutskiy, Phys. Rev. D 90, 124021 (2014), arXiv:1410.7698 [gr-qc].DOI
- [45]S. Das, C. Dundu, JHEP 2024 (02), 049, arXiv:2310.06438 [hep-th].arxiv.org/abs/2310.06438
- [46]A. I. Belokon, M. Čubrović, V. Đukić, V. V. Pushkarev, and L. Umrikhim, Another Brick in the Wall: Spectral Form Factor of Normal Modes in Horizonless Geometries (2026), arXiv:2610.37038 [hep-th].arxiv.org/abs/2610.37038
- [47]S. Güttel and F. Tisseur, Acta Numerica 26, 1 (2017).










