Arxiv:2006.14628V2 [Gr-Qc] 9 Sep 2020

How does a dark compact object ringdown?

Elisa Maggio,1, ∗ Luca Buoninfante,2, † Anupam Mazumdar,3, ‡ and Paolo Pani1, § 1Dipartimento di Fisica, “Sapienza” Universitàdi Roma & Sezione INFN Roma1, Roma 00185, Italy 2Department of Physics, Tokyo Institute of Technology, Tokyo 152-8551, Japan 3Van Swinderen Institute, University of Groningen, 9747 AG, Groningen, The Netherlands A generic feature of nearly out-of-equilibrium dissipative systems is that they resonate through a set of quasinormal modes. Black holes – the absorbing objects par excellence – are no exception. When formed in a merger, black holes vibrate in a process called “ringdown”, which leaves the gravitational-wave footprint of the event horizon. In some models of quantum gravity which at- tempt to solve the information-loss paradox and the singularities of General Relativity, black holes are replaced by regular, horizonless objects with a tiny effective reflectivity. Motivated by these scenarios, here we develop a generic framework to the study of the ringdown of a compact object with various shades of darkness. By extending the black-hole membrane paradigm, we map the interior of any compact object in terms of the bulk and shear viscosities of a fictitious fluid located at the surface, with the black-hole limit being a single point in a three-dimensional parameter space. We unveil some remarkable features of the ringdown and some universal properties of the light ring in this framework. We also identify the region of the parameter space which can be probed by current and future gravitational-wave detectors. A general feature is the appearance of mode doublets which are degenerate only in the black-hole limit. We argue that the merger event GW150914 already imposes a strong lower bound on the compactness of the merger remnant of approximately 99% of the black-hole compactness. This places model-independent constraints on black-hole alternatives such as diffuse “fuzzballs” and nonlocal stars.

I. INTRODUCTION provides the key to reading BH physics [17]: the strong redshift near these objects, the existence of a photon- At the beginning of the last century, the quantum de- sphere [18, 19], the number of independent “charges”, scription of atomic interactions had been dramatically their multipole moments, their vanishing tidal deforma- revolutionized by precision atomic spectroscopy. The ob- bility, and last but not least their QNM spectrum. servation of a small effect such as the Lamb shift in the Nonetheless, the presence of an event horizon poses energy levels of the hydrogen atom [1] had a paramount some theoretical and fundamental problems [17], the impact in shaping the development of quantum electro- most notable ones being the presence of a curvature dynamics. Following the historical detection of the grav- singularity inside BHs (where Einstein’s theory breaks itational waves (GWs) from binary mergers [2,3], black- down) and the Hawking information loss paradox (ac- hole (BH) spectroscopy [4–10] – i.e., the measurement of cording to which unitarity of quantum mechanics is lost the GW spectrum of a BH through post-merger “ring- if a spacetime has an event horizon [20]). Different at- down” observations [2, 11] – may play a similar major tempts to solve these long-standing problems have been role in probing the gravitational interaction and funda- proposed. Many of them share the common feature that mental physics [12–15]. classical BHs are replaced by horizonless, singularity-free solutions. This is the case of the “fuzzball” proposal [21– Indeed, BHs can be considered as the “hydrogen atom” 23], according to which a large number of regular, hori- of gravity, and their GW spectrum is their characteristic zonless microstate geometries with the same asymptotic footprint. At variance with the hydrogen atom, though, charges of a BH emerge as solutions to supergravity and BHs are dissipative systems and their spectrum is de- string theories [23–27]. Another example are nonlocal arXiv:2006.14628v2 [gr-qc] 9 Sep 2020 scribed by quasinormal modes (QNMs), each one defined stars [28] that emerge in theories with infinite-derivatives by an oscillation frequency ω and a damping time τ re- in which the nonlocality of the gravitational interaction lated to the dissipation due to the GW emission at infin- can smear out the curvature singularity and avoid the ity and the absorption at the horizon. presence of a horizon [29–34]. The defining feature of a BH is its event horizon, i.e., a In this context, the devising of observational tests to hypersurface delimiting the spacetime region from which distinguish a classical BH from another dark compact no signal can escape. Accordingly, BHs are perfect ab- 1 object is of utmost importance, and current GW obser- sorbers at the classical level . The presence of a horizon vations (such as GW190814 [35]) do not exclude the existence of exotic compact objects other than BHs and neutron stars. Arguably, GW spectroscopy is the most

∗ [email protected] † [email protected] ‡ [email protected] 1 For astrophysical BHs, the semi-classical phenomenon of Hawk- § [email protected] ing radiation [16] is negligible. 2

nonlocal eﬀects [28]. These solutions can be less compact than BHs; therefore, they could emit a ringdown signal which diﬀers from the BH ringdown at early stages. In this paper, we propose a generic and (as much as possible) model-independent framework to study the ringdown of a dark compact object in the context of Gen- eral Relativity or some of its high-energy extensions. We unveil some universal features of the ringdown and show that current GW observations can already place very stringent constraints on the compactness of a dark object.

FIG. 1. Schematic representation of the BH membrane paradigm [36, 37]. A static observer outside the horizon can II. MEMBRANE PARADIGM FOR replace the interior of a perturbed BH (left) with a perturbed PERTURBED COMPACT OBJECTS fictitious membrane (right). The membrane is made of a viscous fluid whose properties (density, pressure, viscosity) are A. Setup fixed to reproduce the BH phenomenology, in particular the QNM spectrum. In this paper we extend this concept to a generic dark compact object, parametrizing its properties in Our framework extends the seminal work by Damour, terms of the membrane’s physical quantities. MacDonald, Price, Thorne, Znajek, and others on the BH membrane paradigm [36, 37, 71]. In its standard formulation (see Fig.1), a static observer can replace direct way to test the nature of a remnant formed in the the BH interior with a fictitious membrane located at highly-dynamical aftermath of a gravitational merger. the horizon. The features of the internal spacetime are Even small deviations in the QNMs predicted for a BH projected onto the membrane, whose physical properties in General Relativity might provide indication for new are fixed by the Israel-Darmois junction conditions [72– physics, similarly to the Lamb shift in the spectrum of 74] (we assume G = c = 1 units throughout), the hydrogen atom. [[hab]] = 0 , [[Kab − habK]] = −8πTab , (1) A dark compact object can be identified by two parameters, quantifying its shades of “darkness” and its where hab is the induced metric on the membrane, Kab ab “compactness” [17]. The darkness of the object depends is the extrinsic curvature, K = Kabh , Tab is the stress- on its reflectivity, whereas the compactness is related energy tensor of the matter distribution located on the to the gravitational redshift at its surface. It has been membrane, and [[...]] denotes the jump of a quantity shown that the prompt ringdown signal which follows across the membrane. The junction conditions impose the merger is universal if the remnant is sufficiently com- that the fictitious membrane is described by a viscous pact [38–40]. Putative near-horizon structures, which fluid whose thermodynamical properties (density, pres- can be modelled in terms of a reflectivity coefficient at the sure, viscosity parameters) are uniquely determined if the surface of the object [41, 42], show up only at late times membrane is demanded to act as a BH in terms of ob- in the form of repeated and modulated “GW echoes” [38– servable effects [37, 75]. We work in the framework of 40, 43]. The time delay between echoes depends on the General Relativity2 and, for simplicity, we assume spher- compactness of the object, whereas the relative ampli- ical symmetry. tude between subsequent echoes is related to the reflectivity. Considerable attention has been devoted to the study of this signal in a variety of models of near-horizon quan- B. Background geometry tum structures [44–49], in the context of exotic compact objects in General Relativity [50, 51], and for modified- Let us assume that the fictitious membrane is lo- gravity scenarios [28, 52–54]. Phenomenological wave- cated at the surface of the compact object, r = R, in form models have also been analyzed in detail [41, 55– 62] and implemented in actual searches in LIGO/Virgo data [43, 44, 60, 63–68] (see Refs. [17, 69] for some re- 2 Although our framework can be applied to any theory of gravity, views). in practice, we assume that General Relativity works sufficiently The above studies focused on the case of objects almost well near the radius of the compact object. This assumption as compact as a BH, but with some nonvanishing reflec- is well justified also for solutions to modified gravity in which tivity. However, in what can be arguably considered as putative extra degrees of freedom are heavy. In this case, all corrections to the metric and field equations are suppressed by the best-motivated attempts to solve the information-loss powers of `P /R 1, where R is the object radius, and `P is paradox, the reflectivity is expected to be very small, for the Planck length or the scale of new physics. For compact as- example owing to the enormous number of degrees of free- trophysical objects, such corrections are negligible. No specific dom that can be excited within a fuzzball [70] or due to theory is assumed to describe the object interior. 3 some coordinate system. Owing to Birkhoff’s theorem, density of the fluid the spacetime geometry for r > R is described by the p Schwarzschild metric, 2f(R) + Rf 0(R) f(R) p(R) = , ρ(R) = − , (5) 16πRpf(R) 4πR 1 ds2 = −f(r)dt2 + dr2 + r2 dθ2 + sin2 θdϕ2 , (2) f(r) which parametrically fix a barotropic equation of state, p = p(ρ). In the BH limit, the density vanishes and the where f(r) = 1−2M/r and M is the total mass of the ob- pressure diverges as the redshift factor −1/2. The speed ject. It is convenient to parametrize the circumferential p of sound, cs = ∂p/∂ρ, diverges both in the BH limit radius of the membrane as ( → 0) and in the light-ring limit ( → 1/2). However, it is important to stress that these pathologies of the fluid R = 2M(1 + ) , (3) are not an issue since the membrane is fictitious. which defines the compactness of the object, M/R. When → 0 the membrane location coincides with C. Linear perturbations & boundary conditions the Schwarzschild’s horizon, R = 2M, whereas when = 1/2 it coincides with the light ring (or photon- sphere), R = 3M, where circular, unstable photon orbits Gravitational perturbations in the exterior reside. As we shall show, in the latter limit some re- Schwarzschild geometry are governed by two simple markable and universal properties of the QNM spectrum equations that can be both written in a Schrödinger-like emerge. The case of 1 corresponds to ultracom- form [5, 78–80], pact configurations characterized, for instance, by micro- d2ψ(x) scopic corrections at the horizon scale, whereas the case 2 2 + ω − V (r) ψ(x) = 0 , (6) of ∼ O(0.1 − 1) corresponds to less compact objects dx whose compactness is comparable to or larger than that where the effective potential V (r) reads of a neutron star. In the context of fuzzballs, the former and latter cases describe tight and diffuse fuzzball mod- l(l + 1) 6M V =f − , (7) els [70], respectively. The model of nonlocal stars intro- axial r2 r3 duced in [28] is characterized by a maximal compactness q2(q + 1)r3 + 3q2Mr2 + 9M 2(qr + M) of the order of the Buchdahl limit [76], i.e., ' 0.125, Vpolar =2f 3 2 ,(8) and therefore belongs to the latter category. r (qr + 3M) Following the original membrane paradigm, we assume for axial (i.e., magnetic or odd-parity) and polar (i.e., that the fictitious membrane is such that the extrin- − electric or even-parity) perturbations, respectively. Here sic curvature of the internal spacetime vanishes, Kab = x is the “tortoise” coordinate such that dx/dr = 1/f(r), 0 [37]. The details of the computation are given in Ap- 1 q = 2 (l − 1)(l + 2), and l is the angular number of the pendixA, here we only summarize the main results. perturbation. The junction conditions are compatible with a mem- By imposing boundary conditions at infinity and at brane described by the stress-energy tensor of a dissi- the inner boundary, Eq. (6) defines an eigenvalue problem pative fluid [37, 77]: whose complex3 eigenvalues are the QNMs of the system, ω = ωR +iωI . The effective potential vanishes at infinity, Tab = ρuaub + (p − ζΘ) γab − 2ησab , (4) so the general solution to Eq. (6) in the asymptotic region is a superposition of ingoing and outgoing waves. The where ρ, p, and u are the density, the pressure, and the a QNMs are defined by imposing purely outgoing waves at 3-velocity of the fluid, respectively, whereas Θ is the ex- infinity pansion, σab is the shear tensor, and γab = hab + uaub is the projector tensor. The parameters ζ and η are ψ(x) ∼ eiωx , x → ∞ . (9) the bulk and the shear viscosities that govern the response of the fluid to external perturbations. For a given In the BH case, regularity at the horizon (located at model of the interior of the compact object, η and ζ x → −∞ in tortoise coordinates) implies purely ingoing are uniquely determined. In general, they depend on waves at the inner boundary, the density (and the temperature) of the fluid and can −iωx be frequency-dependent, complex numbers. Since energy ψBH(x) ∼ e , x → −∞ , (10) dissipation is absent when <[η] = 0 = <[ζ], for simplicity in the following we shall consider η and ζ to be real con- for both axial and polar perturbations. stants, although our framework can be straightforwardly generalized. When the object is unperturbed, the viscosity is irrele- 3 vant and the spherical membrane is described by a perfect In our conventions a stable mode corresponds to ωI < 0, whereas fluid. The junction conditions then give the pressure and an unstable mode corresponds to ωI > 0. 4

In the general case, the condition on the inner bound- Unlike the axial sector, the polar boundary condition ary depends on the properties of the object, in particular depends explicitly on the bulk viscosity ζ. However, on its compactness and reflectivity. In order to derive when R → 2M the dependence on ζ disappears, so this these boundary conditions generically (i.e., without as- parameter is not fixed by the linear perturbation anal- suming any specific solution for the interior of the ob- ysis in the BH limit (see AppendixA for details). On ject), we rely on the membrane paradigm and impose the other hand, according to the membrane paradigm, a junction conditions on the perturbations at the mem- generic BH can be mapped in terms of a fluid also en- brane. As detailed in AppendixA, the generic boundary dowed with a negative bulk viscosity, ζ = −1/(16π) ≡ condition for the axial case turns out to be ζBH. This peculiar property is associated with the teleo- logical nature of the event horizon [37]. ψ0(x) iω R2 = − − Vaxial(R) , x → xR (11) One might wonder about the physical meaning of the ψ(x) 16πη 2(R − 3M) shear viscosity as given in Eq. (14). For this purpose, it is 2 where the prime denotes the derivative with respect to instructive to compute the reflectivity coefficient |R| of the object. The latter is defined through the scattering of the tortoise coordinate, and xR ≡ x(R). The polar case is more involved but a similar computation yields a wave coming from infinity with (say) unitary amplitude and being partly reflected back, i.e. ψ0(x) = −16πiηω + G(R, ω, η, ζ) , x → x (12) iωx −iωx ψ(x) R ψ(x) ∼ Re + e , x → ∞ , (15) where G is a complicated function given in AppendixA. after being subjected to the boundary conditions (11) The boundary conditions given by Eq. (11) and or (12) at r = R. Eq. (12) are one of our main results: they are valid for It is straightforward to compute R in the large- any compact object described by a Schwarzschild exterior frequency4 limit (Mω 1), where the effective potential and have far-reaching consequences, which we discuss in in Eq. (6) can be neglected. In this case, we obtain a the following. simple analytical result valid for both axial and polar The linear response of the object in the time domain perturbations: 2 2 is governed by the partial differential equation [∂t − ∂x + ˆ ˆ 1 − η/η 2 V ]ψ(t, x) = 0, where ψ(t, x) is the inverse Fourier trans- |R|2 = BH . (16) form of ψ(ω, x). In the axial sector, the time-domain 1 + η/ηBH response can be computed straightforwardly, since the inverse Fourier transform of the boundary condition (11) This shows that the object is a perfect absorber (|R|2 = reads 0) of high-frequency waves if and only if η = ηBH, 2 2 whereas it becomes a perfect reflector (|R| = 1) of high- ˆ 1 ˆ R Vaxial(R) ˆ ∂xψ(t, xR) = ∂tψ(t, xR) − ψ(t, xR) . frequency waves when either η = 0 or η → ∞. In the case 16πη 2(R − 3M) of an ultracompact object with 1 and for any fre- (13) quency, the former and latter cases correspond to Dirich- The polar sector is more involved since the dependence let (ψ = 0) and Neumann (ψ0 = 0) boundary conditions, on ω in Eq. (12) is complicated and the inverse Fourier respectively, for the axial sector. As clear from Eq. (12), transform cannot be obtained in closed form. the opposite is true for the polar sector: η = 0 (η → ∞) corresponds to the Neumann (Dirichlet) boundary condition. These perfectly reflecting boundary conditions D. Effective viscosity, reflectivity, and universality were studied in Refs. [42, 81], of which our results are a generalization. In the BH limit the boundary conditions in Eqs. (11) Although η is formally a free parameter, we expect and (12) must reduce to the BH one. A straightforward the most interesting range to be η ∈ [0, ηBH]. Indeed, inspection of the above formulas shows that, when R → from Eq. (16) negative values of η would correspond to 2M, the BH boundary condition in Eq. (10) is recovered |R|2 > 1, which leads to superradiant instabilities [82] only if even in the static case. Similarly, for η > ηBH the reflec- 1 tivity coefficient is a monotonically growing function of η = ≡ η . (14) the shear viscosity, which is also unphysical. 16π BH The reflectivity coefficient for generic frequencies can In this case Eq. (13) reduces to ∂xψ(t, x) = ∂tψ(t, x), be computed numerically with standard methods (see i.e., the standard boundary condition for an ingoing e.g., Ref. [82]). The result is shown in Fig.2 for axial (left wave ∼ e−iω(t+x) near the horizon. Thus, a perturbed panel) and polar (right panel) perturbations and some BH of mass M behaves as a (2 + 1)-dimensional viscous fluid located at r = R ∼ 2M with shear viscosity η = 1/(16π). This result agrees with the stan- 4 dard BH membrane paradigm and extends the results of For an object with M = 10M , radio waves with frequency Refs. [69, 77] to the polar sector. 3 kHz are already in the large-frequency regime. 5

Axial,η=η BH Polar,η=η BH,ζ=ζ BH 1.0 1.0

0.8 0.8

2 0.6 2 0.6

| ℛ BH | ℛ BH 0.4 ϵ=0.1 0.4 ϵ=0.1 ϵ=0.2 ϵ=0.2 ϵ=0.3 ϵ=0.3 0.2 ϵ=0.4 0.2 ϵ=0.4 ϵ=0.499 ϵ=0.499 0.0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 Mω Mω

FIG. 2. Reflectivity coefficient of a dark compact object for axial (left panel) and polar (right panel) perturbations in terms of the shear (η = ηBH) and the bulk (ζ = ζBH) viscosities of a fictitious fluid located at the surface of the object, R = 2M(1 + ), and for several values of . At high frequency, Mω 1, the object is perfectly absorbing for any value of [see Eq. (16)]. Interestingly, for → 1/2 the object becomes a perfect reflector of axial GWs for all frequencies and regardless of the values of η (not shown in the plot). In both panels the vertical dashed line corresponds to the fundamental QNM frequency of a BH Schwarzschild BH, MωR ≈ 0.3737. representative choices of the parameters. Besides recov- These effects have dramatic consequences for the ring- ering all the expected limits with respect to the BH case down and the QNM spectrum, as we are going to discuss and in the large-frequency regime, this plot also shows in the next section. some other interesting effects. Noticeably, as → 1/2 (i.e., as the surface of the object approaches the light ring), the reflectivity coefficient of axial GWs tends to unity for any frequency. This can be III. RINGDOWN & QNM SPECTRUM understood by looking at the boundary condition (11), whose last term diverges as R → 3M, thus imposing Equation (6) with boundary conditions (9) and ei- ψ = 0 for any frequency and for any η ∈ C. There- ther (11) (for the axial case) or (12) (for the polar case) fore, regardless of its internal structure, an object with can be solved numerically as an eigenvalue problem to 5 R = 3M is a perfect reflector of axial GWs . The polar derive the QNM spectrum of a given model, identified sector does not enjoy the same universality. Moreover, in by the parameters (, η, ζ). For each value of l = 2, 3, ..., some regions of the parameter space the polar reflectiv- there exists a countably infinite set of QNMs which can ity shows some distinctive peak at specific frequencies, as be ordered (n = 0, 1, 2, ...) in terms of their imaginary shown in Fig.2 for = 0.1 (for the choice η = η and BH part, =(ω) ≡ ωI : the fundamental mode (n = 0) has the ζ = 0 not shown in the plot the peak can be as high as longest decay time, τ = −1/ω , whereas its overtones 2 I |R| = 1). (n = 1, 2, ...) have a shorter lifetime. It is also interesting to note that the reflectivity of the In the axial case the QNM spectrum depends on the spacetime at intermediate frequencies (Mω = O(0.1−1), two-dimensional parameter space (, η) – i.e., on the com- those relevant in the ringdown [11]) can be larger than pactness and viscosity of the membrane – whereas in the the BH reflectivity even if the object is a perfect absorber polar sector there is an additional dependence on the bulk at higher frequencies. For example, at the fundamental viscosity ζ. As previously discussed, in the → 0 limit BH QNM frequency, Mω ≈ 0.3737 (shown in Fig.2 as a the polar boundary conditions (and hence the QNMs) vertical dashed line), the reflectivity is |R|2 ≈ 1 for axial 2 are independent of ζ. perturbations, = 0.4 and η = ηBH, although |R| → 0 as Mω 1. We compute the QNM spectrum using two numerical methods [85]: a shooting method based on the direct integration of Eq. (6), and another one based on continued fractions [10, 86] in a variant adapted from the case of 5 The only exception is when η → − 3iω (R − 3M) as R → 3M, in 16πq compact stars [87–89]. The continued-fraction method is which case the divergence in Eq. (11) cancels out. However, for any purely imaginary η the axial reflectivity is unity at any fre- more robust as it is well suited also for overtones with quency, so the object remains a perfect reflector. Although very MωI 1, for which the direct integration fails. When fine-tuned, this particular case corresponds to thin-shell gravas- both applicable, the two methods are in excellent agree- tars [83, 84]. ment. 6

A. Prompt ringdown versus GW echoes 0.4

Before proceeding with the QNM analysis, let us look 0.2 at the linear response in the time domain by studying a wave packet scattered off the object. Since the external 0.0 spacetime for r > R is described by the Schwarzschild -0.2 metric, the initial, prompt ringdown is associated with BH ψ [ r = 10M,t ] -8 the scattering of the wave packet off the effective po- ϵ=10 , η=0.1ηBH -0.4 -8 tential V (r) [see Eq. (6)] independently of the boundary ϵ=10 , η=0.5ηBH ϵ=10-4, η=0.5η conditions. Indeed, if is sufficiently small the follow- -0.6 BH ing causality argument shows that the boundary conditions cannot affect the prompt ringdown [38]. The 0 50 100 150 200 decay time scale of the prompt ringdown is associated t/M with the instability time scale of circular photon orbits at the light ring [17, 18], or equivalently to the decay 0.6 time of the fundamental QNM of a Schwarzschild BH, 0.4 τ = −1/ωI ≈ 10M. Thus, the boundary conditions at r = R (< 3M) do not have time to modify the prompt 0.2 ringdown if the round-trip time of the radiation from the 0.0 photon sphere to the boundary and back [17, 39, 40], -0.2

ψ [ r = 10M,t ] BH τecho = 2M [1 − 2 − 2 log(2)] , (17) ϵ=0.1, η=ηBH -0.4 ϵ=0.4, η=ηBH is much longer than ≈ 10M. This imposes ϵ=0.49, η=ηBH O(0.01) [17]. Objects with such a large compactness -0.6 were dubbed ClePhOs, since they have a clean photon- 0 10 20 30 40 50 60 70 sphere [17, 40]. On the other hand, if & O(0.01) the object’s interior is expected to affect the prompt ring- t/M down. These expectations are confirmed by the ringdown FIG. 3. Ringdown of a dark compact object with ra- waveforms shown in Fig.3, which are obtained by solv- dius R = 2M(1 + ) and effective shear viscosity η. We consider axial perturbations and an initial Gaussian profile, ing the linearized problem for axial perturbations with 2 ψ(x, 0) = 0, ∂tψ(x, 0) = exp[−(x − 7) ]. Top panel: when boundary condition (13) using a fourth-order Runge- 0.01 GW echoes appear for several values of η related Kutta finite-difference scheme. The top panel shows the to the reflectivity of the object. Bottom panel: a selection → 0 limit. Confirming previous results [38–40], in this of ringdown waveforms for > 0.01. In this case echoes are case the prompt ringdown is universal and the details absent but the prompt ringdown is governed by the modified of the object’s interior appear only in late-time echoes QNMs of the object. of the initial ringdown. Their period is roughly given by Eq. (17), and their frequency content and phase are modulated by the boundary conditions and by the tunneling higher peaks in the prompt ringdown. Crucially, as dis- through the potential barrier [61, 62]. Finally, their am- cussed in the following section, at late time the prompt plitude depends on the shear viscosity η. In particular, ringdown is dominated by the fundamental QNM of the η ≈ 0 corresponds to |R|2 ≈ 1 for which subsequent object, which is not the mode of the universal prompt echoes amplitudes are only mildly damped [61], whereas ringdown in the BH case. absorption is maximized as η → ηBH. In the latter case Finally, one might wonder why there are no echoes for the linear response is identical to that of a BH, since the & O(0.01), even though the reflectivity of the object boundary conditions are the same when η → ηBH and at intermediate frequencies is high, as shown in Fig.2. → 0. The reason is that only waves with frequency V (R) < 2 The bottom panel of Fig.3 focuses on the case & ω < Vmax can be trapped between the object surface and O(0.01), where the prompt ringdown is modified and no the potential barrier. Therefore, when the compactness subsequent echoes appear. The changes to the prompt decreases, the resulting cavity is small. Furthermore, the ringdown can be understood by considering that the part transmission coefficient of the potential barrier is large 2 of the wave packet that initially tunnels through the bar- when ω . Vmax, which implies that these frequencies rier has enough time to be reflected at r = R and to cannot be trapped efficiently. This also explains why tunnel again to infinity. This process results in a su- when R → 3M (corresponding to perfect reflection of perposition of the two pulses (the one directly reflected axial GWs) there are no echoes, since the membrane is by the potential barrier and the one reflected by the ob- very close to the maximum of the potential. In practice, ject), which interfere in a complicated pattern. When for & O(0.01) one sees only the interference between the the two pulses sum in phase, the interference produces prompt ringdown and the first echo, while subsequent 7

2.5 1.1 Axial Polar ζ=ζBH 1.0 Polar ζ=0 2.0 Polar ζ=10 BH

BH Polar ζ=-10 R 0.9 I / ω / ω Axial I R

ω 1.5 ω 0.8 Polar ζ=ζBH Polar ζ=0 Polar ζ=10 0.7 ζ=- Polar 10 1.0 0.6 10-4 0.001 0.010 0.100 1 10-4 0.001 0.010 0.100 1 ϵ ϵ

FIG. 4. Real (left panel) and imaginary (right panel) part of the fundamental (l = 2) QNM of a dark compact object with effective shear viscosity η = ηBH, with respect to their corresponding value in the BH case. The QNMs are functions of the parameter which is related to the compactness of the object. For → 0 we recover the fundamental BH QNM and axial and polar modes are isospectral, whereas for larger values of (less compact object) the QNMs differ from the BH case and isospectrality is broken. For large ζ or for → 0, the polar QNMs are independent of ζ. As a reference, the shaded regions correspond to the error bars (10% and 15%, respectively) on the frequency and the damping time for the merger event GW150914 [2, 11]. For & 0.01, these QNMs appear in the modified prompt ringdown, see e.g. Fig.3. reflections are strongly suppressed or absent, as in the Polar modes are qualitatively similar to the axial ones bottom panel of Fig.3. and show a mild dependence on the bulk viscosity ζ. The QNM spectrum is independent of ζ in the large-ζ limit, as shown in Fig.4 by the ζ = ±10 curves. This is expected B. QNM dependence on the compactness and since, when ζ → ∞, the polar boundary condition (12) is viscosity independent of ζ, and the second term of the right-hand 3M(R−2M) side reduces to G = R2(3M+qR) . Let us first analyze the QNM spectrum of a dark com- Finally, Fig.5 shows the complex QNM plane of a dark pact object with η = ηBH. Figure4 shows the real and compact object for axial perturbations. For a given value imaginary part of the axial and polar QNMs as a func- of , each curve is parametrized by the shear viscosity6 η. tion of , with respect to the BH case. As expected, As the location of the surface approaches the light ring for → 0 we recover the fundamental l = 2 QNM of a ( → 1/2), the axial QNMs become independent of η. Schwarzschild BH, MωBH ≈ 0.3737 − i0.089. For larger Indeed, they tend to a universal mode which, for l = 2, values of , the compactness of the object decreases and reads the QNMs depart from the BH case. From the previous discussion, for 0.01 the (small) deviations to the Mωaxial ≈ 0.3601 − i0.2149 , → 1/2 , (18) QNMs do not affect the prompt ringdown. Furthermore, for η = ηBH echoes are suppressed, so an object with for any value of η. As previously discussed, in this limit 0.01 and η = ηBH is indistinguishable from a BH in the object becomes a perfect reflector of axial GWs, re- terms of its ringdown. gardless of the value of η. Therefore, a compact object For & 0.01, the prompt ringdown is affected by with R = 3M has a universal axial QNM spectrum, re- the modified QNMs. Indeed, by fitting the time-domain gardless of its properties (but see footnote5 for a loop- waveform at late times with a damped sinusoid, Ψ(t) ∼ hole). Unfortunately, this remarkable universality does −t/τ A cos(ωRt + φ)e , we can verify that in this range the not apply to the polar sector, since there does not exist prompt ringdown is governed by the modified fundamen- a value of R at which the function G defined in Eq. (12) tal QNM of the system. This noteworthy feature carries diverges for any value of η, ζ, and ω. a characteristic imprint of the structure of the remnant. The QNMs of the Schwarzschild BH enjoy a remarkable “isospectrality” [5, 90]: the entire spectrum is the same for the axial and the polar gravitational sectors. 6 We focus on η > 0 because, consistently with the previous anal- This isospectrality is broken for finite values of (and ysis of the reflectivity coefficient, when η < 0 we found unsta- for η 6= ηBH, as discussed below). For each point of the ble modes (i.e., modes with ωI > 0). This is expected since a parameter space, the l = 2 fundamental QNMs form a negative shear viscosity corresponds to energy injected into the doublet. system, i.e., to a super-emitter with |R|2 > 1 [82]. 8

Axial Axial

0.100 0.100 0.050 I 0.050 ϵ=10-4 I ϵ=10-2 ϵ=0.125 - M ω - M ω ϵ=0.3 0.010 ϵ=0.4 0.005 0.010 ϵ=0.45 BH n=0 ϵ=0.49 BH n=1 ϵ=0.499 0.005 0.36 0.38 0.40 0.42 0.44 10-12 10-9 10-6 10-3 1

MωR 1-η/ηBH

FIG. 5. The complex QNM plane for axial modes. Each FIG. 6. Imaginary part of the axial QNMs of a dark compact curve is parametrized in terms of the effective shear viscosity object with radius R = 2M(1+) and = 10−10 as a function η ∈ [10−4, 100]. As → 1/2, the curves collapse to a sin- of the effective shear viscosity, tracking the fundamental (n = gle universal point (black empty circle), corresponding to the 0, blue curve) and first overtone (n = 1, red curve) QNM of a universal QNM in Eq. (18). As a reference, the fundamental Schwarzschild BH (leftmost part of the plot) down to the limit QNM of a Schwarzschild BH is marked by a black dot. of perfectly reflecting object (η = 0, rightmost part). There exists a crossing point at which the imaginary part of the BH overtone becomes smaller than that of the BH fundamental C. Transition from fundamental modes to mode. The polar case is qualitatively similar. overtones & instability of the QNM spectrum

Indeed, neither of the modes shown in Fig.6 for η → 0 As previously discussed, the computation of the QNMs correspond to the fundamental mode of the perfectly re- in the → 0 limit is less relevant from a phenomeno- flecting case. The latter is a long-lived mode with a logical point of view, since these modes do not govern tiny imaginary part [81]. Interestingly enough, track- the prompt ringdown but only the late-time echo struc- ing the fundamental mode from the perfectly reflecting ture [17]. Nonetheless, from a theoretical standpoint it case back to η is extremely challenging. We suspect is enlightening to study their dependence on η, since this BH that this is connected to the fact that – when tracked to parameter interpolates between the BH limit (η = η ) BH η → η – this mode corresponds to a very high-order and the perfectly reflecting case (η → 0, ∞). BH overtone in the BH limit, which are extremely difficult to For concreteness, we consider = 10−10 and track the compute numerically, due to their large imaginary part. QNMs from the BH case (η = η ) down to η = 0. In BH Furthermore, an interesting property of the BH QNM Fig.6, we show the imaginary part of the axial QNMs ob- spectrum is its instability against small deformations of tained with this procedure starting from the fundamen- the eigenvalue problem, such as bumps in the effective tal (n = 0, blue curve) mode and from its first overtone potential [98–100] or slightly different boundary condi- (n = 1, red curve). As expected, at the starting point tions [38] (see Ref. [101] for recent work on this topic). In we recover the modes of a Schwarzschild BH. However, our numerical analysis we have seen hints of this instabil- immediately after η decreases (going on the right in the ity (which is more severe for high-order overtones [101]) plot) there is a crossing point between the two curves: due to finite- effects and to the slightly different bound- already when η < 0.99997η the first BH overtone be- BH ary conditions when approaching the BH limit. We de- comes longer lived (i.e., smaller ω ) than the fundamental I fer a detailed analysis of this interesting problem to the BH mode, showing that the former becomes more rele- future, stressing that our framework is well-suited to ex- vant for the late-time dynamics of the system7. Although plore this problem in full generality. tracking higher overtones is challenging, we have indica- tions that this behavior is rather general: higher overtones increasingly become more relevant in the η → 0 limit. D. Detectability

Figure7 shows the relative percentage diﬀerence between the BH QNM and the QNMs of a dark compact ob- 7 Recently, it has been realized that the overtones might play an ject in the (, η) parameter space. Based on the previous important role in the ringdown of loud mergers [91–97]. Study- discussion, we focus on the region which directly aﬀects ing the implication of our results for overtones is an interesting the prompt ringdown, i.e., & O(0.01) and η ∈ [0, ηBH]. extension of this work. As a reference, in Fig.7 we show the contour lines 9

(ω -ω BH)/ω BH% BH BH I I I (ωR -ωR )/ωR % 150 20

10 100

0 50

-10

-20

-50 -30

(ω -ω BH)/ω BH% BH BH I I I (ωR -ωR )/ωR % 75

20 50

10 25

0 0

-25

-10 -50

-20 -75

FIG. 7. Relative percentage difference of the real (left panel) and imaginary (right panel) part of the gravitational l = 2 QNMs of a dark compact object with respect to the BH QNM. The difference is parametrized by the shear viscosity η ∈ [0, ηBH] of a fictitious fluid located at R = 2M(1 + ) and by the compactness of the object through the parameter ∈ [0.01, 1]. As a reference, the contour lines correspond to the accuracy in the measurement of the fundamental QNM of GW150914 and the dashed areas are the regions of the parameter space potentially excluded by individual measurements of ωR (left) or ωI (right). The dark shaded region in all panels is the area not excluded by a simultaneous measurement of the frequency and the damping time with the same accuracy as GW150914. corresponding to 10% and 15% of relative difference for the QNM spectrum is compatible with a BH in General ωR and ωI , respectively, which roughly correspond to the Relativity. However, independent estimates of the mass accuracy within which the fundamental QNM of the rem- and spin can be extracted from the inspiral-merger signal, nant in GW150914 has been measured. Although in the assuming a binary BH in General Relativity [2, 11]. Any latter case the remnant is spinning [11], we can assume deviation from the expected fundamental QNM would ≈ (10 − 15)% as an approximate estimate for the ring- imply a failure of this test, suggesting either a departure down resolution of LIGO/Virgo for “golden” events such from General Relativity or that the remnant is not a Kerr as GW150914. For the same type of events, an overall BH, or both. improvement of the signal-to-noise ratio by an order of magnitude will provide a resolution at the percent level. Interestingly, already with the current LIGO/Virgo accuracy one can potentially place very strong constraints Measuring a single mode in the ringdown only allows to on the parameter space of our model, excluding the re- extract the mass and the spin of the object and one would gions of the parameter space hatched in Fig.7. By com- need to measure at least two QNMs to verify whether bining the information from the real and the imaginary 10 part of the QNMs, Fig.7 shows that only a small region Log10[ρres] of the parameter space with . 0.1 and η ∼ ηBH (dark 5 shaded area) is compatible with the current constraints for both axial and polar perturbations. For polar perturbations, also a region with & 0.1 and 0 < η < ηBH is compatible with the current constraints. Our results 4 suggest that an improvement by a factor of a few in the detector sensitivity relative to LIGO first observational run would firmly rule out any modification of the prompt 3 ringdown in our model. Furthermore, as previously discussed, a generic feature of departure from the BH picture is the presence of a mode doublet. The frequencies of axial and polar modes 2 are similar in most of the parameter space, with some regions featuring relative differences as large as ≈ 40%; however, the relative difference of the damping times can 1 be much bigger (as large as ≈ 1500% in some regions). A necessary condition for resolving these differences is based on the Rayleigh resolvability criterion [8, 93]

max[σf , σf ] < |f1 − f2| , (19) 1 2 FIG. 8. Minimum signal-to-noise ratio required for the re- max[σQ1 , σQ2 ] < |Q1 − Q2| , (20) solvability of the axial-polar QNM doublet according to the (i) Rayleigh criterion. where fi = ωR /(2π) and Qi = πfiτi are the frequency and quality factor of the i-th mode, and σX is the un- certainty in recovering a quantity X. In Fig.8 we show of a BH, larger values of η would result in echoes that the minimum signal-to-noise ratio, ρ , required to re- res are too damped to be detectable with current facilities. solve the axial and polar QNMs according to the above criterion. We computed the uncertainties on the parameters with a simple Fisher analysis [8, 93] adapted from the case of BH overtones [95, 97], and assuming that IV. CONCLUDING REMARKS the amplitude ratio between the axial and polar modes 3 is 1/10. In the η ≈ ηBH region ρres > 10 , but it can If an object is almost as compact as a BH and absorbs be much smaller for η < ηBH. Although, a comparison radiation almost as much as BH, then its QNM spectrum between Fig.8 and Fig.7 shows that detecting devia- is almost indistinguishable from that of a BH. Besides tions from the BH QNMs should require a much smaller confirming this reasonable expectation, our analysis pro- signal-to-noise ratio than resolving the axial-polar dou- vides a general framework to quantify the next natural blet, the existence of this doublet is a generic smoking question: when does the ringdown of a dark compact ob- gun of objects other than BHs, so it might be useful to ject become distinguishable from that of a BH? place model-independent constraints. If an object with radius R = 2M(1 + ) is sufficiently Finally, let us look at the opposite regime, where compact ( 0.01), the prompt ringdown is universal 0.01. In this case the prompt ringdown is not mod- and the properties of the object’s interior can appear ified and the constraints on the model rely on the de- only after the scrambling time τecho ∼ M| log | [38–40], tectability of the echoes. This problem has been recently provided that the viscosity is sufficiently small. Thus, analyzed in detail [41, 55–57, 59–62, 67, 68]. In particu- a negative echo search places an upper bound on the lar, Refs. [61, 62] analyzed the detectability as a function viscosity of the model, with current (future) detectors of the intrinsic reflectivity of the object, which coincides being able to probe values of η as large as 6% (60%) of with Eq. (16) at high frequencies. By using the latter the BH viscosity. equation, we can map previous constraints on the object’s For less compact objects ( & 0.01), there are no reflectivity to the viscosity parameter. In particular, us- repeated echoes but the prompt ringdown is affected ing the results of Refs. [61, 62], our analysis suggests that by the properties of the object’s interior. In particu- echoes are detectable with current GW interferometers at lar, the late-time ringdown is governed by the modified 2σ confidence level if η < 0.06ηBH. Future detectors such QNMs of the system. When the object is a perfect ab- as the Einstein Telescope [102] and LISA [103] will allow sorber at high frequency (η = ηBH) the fundamental to probe η < 0.4ηBH and η < 0.6ηBH, respectively, at the QNM deviates as much as O(10%) and O(100%) (for same confidence level, while future LISA+ concepts [104] the real and imaginary part, respectively) from the BH can reach almost η ≈ ηBH at 3σ level. In other words, case when = O(0.1). Although a more refined analysis current GW interferometers could detect echoes only if is needed (in particular including the spin of the object), the effective viscosity of the object is at most 6% of that this suggests that inspiral-merger-ringdown consistency 11 tests [2,3] can already put a strong constraint on of the and KAKENHI Grant-in-Aid for Scientific Research order of . 0.01, i.e., the compactness of the remnant No. JP19F19164. P.P. and E.M. acknowledge financial cannot be smaller than 99% that of a BH (see Fig.7). support provided under the European Union’s H2020 Future ringdown detections have the capability to rule ERC, Starting Grant agreement no. DarkGRA–757480, out any deviation of the prompt ringdown in our model. under the MIUR PRIN and FARE programmes (GW- A generic feature that emerges from our framework is NEXT, CUP: B84I20000100001), and support from the the fact that objects with & O(0.1) have a significant Amaldi Research Center funded by the MIUR program reflectivity at the typical ringdown frequency (even when “Dipartimento di Eccellenza” (CUP: B81I18001170001). η = ηBH) compared to the BH case. This enhances the A.M. is supported by Netherlands Organization for Sci- effects of the object’s interior in the ringdown. This phe- entific Research (NWO) Grant number 680-91-119. nomenon is most dramatic in the axial case, when the QNM spectrum becomes independent of η as → 1/2 and the reflectivity is unity. An object with radius equal Appendix A: Membrane paradigm to the light ring becomes a perfect reflector of axial GWs and its axial QNM spectrum is universal. In this appendix we provide details on the computation Another generic feature is the breaking of isospectral- leading to the boundary conditions (11) and (12). Let ity. Axial and polar modes are generically different ex- M be the whole spacetime manifold. We can define a cept for the measure-one subspace ( → 0, η = ηBH) 3-dimensional spherical membrane (or shell) generated of the whole three-dimensional (, η, ζ) parameter space. by time-like geodesics of radius R = 2M(1 + ), whose Assuming that axial and polar modes are excited with interior M− and exterior M+ regions are described by − − + + comparable amplitude during the merger, this implies two different metrics gµν (x ) and gµν (x ), respectively, that a dark object other than a BH should vibrate with a with x± being the respective coordinates. The exterior mode doublet. Resolving these doublet when η ≈ ηBH re- region is described by the Schwarzschild metric (2). The quires a signal-to-noise ratio in the ringdown signal above two sets of coordinates are 103, whereas it takes much smaller sensitivity in other regions of the parameter space. x−µ = (t−, r−, θ−, ϕ−) , x+µ = (t+, r+, θ+, ϕ+) , Our framework can be applied to specific models of (A1) µ extreme compact objects [17], such as wormholes [105], while the membrane’s coordinates are xm = (t, R, θ, ϕ), gravastars [83, 84, 106, 107], superspinars [108], nonlocal so that the intrinsic 3-dimensional coordinates on the stars [28], and horizonless microstate geometries [23–27]. shell are

In specific models the viscosity parameters might be com- a plex and frequency dependent. This extension should be y = (t, θ, ϕ) . (A2) straightforward and is left for the future. The normal vector nµ to the time-like membrane is space- Likewise, we focused on nonrotating objects for sim- µ like, i.e., nµn = 1. A basis of three independent gener- plicity, but our approach is conceptually valid also in the ators for the shell can be chosen as case of rotation. In that case, we envisage two technical ∂xµ obstacles: (i) the junction conditions require the analysis eµ = , (A3) of metric perturbations, which should therefore be re- a ∂ya constructed from the Teukolsky function [109, 110]; (ii) the absence of Birkhoff’s theorem in axisymmetry implies through which we can define the induced metric that the exterior of the object is not necessarily described h = eµeν g . (A4) by the Kerr metric. While deviations might be small in ab a b µν the → 0 limit [111], when = O(0.1 − 1) a case-by- on the 3-dimensional surface. The extrinsic curvature on case analysis might be required. In the spinning case, it the membrane is defined as would also be interesting to study the ergoregion insta- µ ν bility [112–114] and how the latter can be quenched in Kab = ea eb ∇µnν , (A5) particular models [42, 77, 81]. ab and its trace is K = h Kab. To embed the 3-dimensional membrane in the manifold M we need to impose a junc- ACKNOWLEDGMENTS tion condition that glues the interior region with the exterior [73]:

+ + − − The authors are grateful to Emanuele Berti for use- (K hab − Kab) − K hab − Kab = 8πTab , (A6) ful correspondence on the continued fraction method + − for stars, and to Samir Mathur, Gabriel Andres Pio- where Kab and Kab are the extrinsic curvatures defined vano, Guilherme Raposo, and Lucas Tonetto for use- on the top and on the bottom of the shell, respectively; ful discussions. L.B. thanks Sapienza University of Tab is the stress-energy tensor of some matter distribution Rome for the hospitality in the first stages of this located on the membrane creating a discontinuity in the project. L.B. acknowledges financial support from JSPS extrinsic curvature. 12

0 The crucial and fundamental assumption of the mem- R (2f(R) + Rf (R)) 2 θθ : = 8πR p0 , brane paradigm is that such a stress-energy tensor is fic- 2pf(R) titious and its role is to carry all the information about 2 0 Rsin θ (2f(R) + rf (R)) 2 2 the interior. In the standard paradigm, one assumes that ϕϕ : p = 8πR sin θp0 . (A12) the inside region is such that 2 f(R) − The (tt) component gives the unperturbed density of the Kab = 0 , (A7) membrane [69] so that the junction condition (A6) reduces to [37] pf(R) Khab − Kab = 8πTab , (A8) ρ (R) = − ; (A13) 0 4πR + where for simplicity we have redefined Kab ≡ Kab, and henceforth we denote the coordinates outside simply as while the angular components give the unperturbed pres- xµ = (t, r, θ, ϕ). sure of the membrane The junction conditions are compatible with the stress- 2f(R) + Rf 0(R) energy tensor of a dissipative fluid, Eq. (4), where ρ and p0(R) = . (A14) 16πRpf(R) p are the density and pressure, respectively; ua is the 3-velocity of the fluid defined in terms of its 4-velocity µ a The speed of sound is Uµ as ua = ea Uµ. Moreover, Θ = u;a is the expansion, σ = 1 (u γc + u γc − Θγ ) is the shear ten- s s ab 2 a;c b b;c a ab 2 ∂p0 1 + 2 + 4 sor, where γab = hab + uaub is the projector tensor, and cs ≡ = , (A15) ub;a is the 3-dimensional covariant derivative compatible ∂ρ0 8(1/2 − ) with the induced metric hab [115, 116]. which diverges as → 0, 1/2 and is complex for > 1/2. Note, however, that the properties of the fluid do not 1. Unperturbed background need to be physical, since the membrane is fictitious.

In the static background [Eq. (2)] the viscosity parameters η and ζ do not play any role. Indeed, it is easy 2. Gravitational perturbations to show that the expansion and shear are zero, so the membrane stress-energy tensor reduces to that a perfect We work in the Regge-Wheeler gauge [78] and study ﬂuid. In this case, the 3-metric reads separately the axial and polar sectors of gravitational perturbations, since they decouple in presence of a static hϕϕ 2 htt = −f(R) , hθθ = 2 = R . (A9) background. sin θ The perturbed metric can be cast in the following gen- The normal vector nµ can be uniquely determined by eral form µ µ imposing the four conditions n nµ = 1 and ea nµ = 0; its 0 components read gµν = gµν (r) + δgµν (r, θ, t) , (A16)

1 where, without loss of generality, the perturbation δg nt = 0 , nr = , nθ = 0, nϕ = 0. (A10) µν pf(r) does not depend on the azimuthal angle ϕ, owing to the 0 spherical symmetry of the background gµν . The extrinsic curvature is also diagonal, Because of the metric perturbations, the dissipative 1p 0 Kϕϕ p components of the stress-energy tensor are switched on, Ktt = − f(R)f (R) ,Kθθ = = R f(R) , 2 sin2 θ and both density and pressure are perturbed as follows where a prime denotes partial derivative with respect to ρ = ρ0 + δρ(t, θ) , (A17) the argument. The only nonvanishing component of the fluid 4- p = p0 + δp(t, θ) . (A18) µ t r θ ϕ t p velocity U = (U ,U ,U ,U ) is U = 1/ f(R). This Also the location of the membrane is affected by the per- yields the the 3-velocity turbation, and we can parametrize the deviation as ! µ a ab 1 ua = e Uµ , u = h ub = , 0, 0 . (A11) rm(t, θ) = R + δR(t, θ) . (A19) a pf(R) We can also easily find the expressions of the perturbed Therefore, the nonvanishing components of the junction µ tangential vectors ea introduced in Eq. (A3). Since the condition (A8) at the background level are 4-dimensional coordinates of the membrane are

2 3/2 tt : − f (R) = 8πf(R)ρ0 , µ R xm = (t, R + δR(t, θ), θ, ϕ) , (A20) 13 we have The perturbed components of the stress-energy tensor are µ et = (1, ∂tδR(t, θ), 0, 0) , −iωt µ δTtϕ = −εe ρ0h0 sin θ∂θPl(cos θ) eθ = (0, ∂θδR(t, θ), 1, 0) , µ 2p 2 ϕ eϕ = (0, 0, 0, 1) . (A21) +R f sin θ(p0 + ρ0)δu (t, θ) , δT = −η R2 sin2 θ∂ δuϕ(t, θ) . (A26) Note that, δρ(t, θ), δp(t, θ) and δR(t, θ) are scalar θϕ θ quantities under rotation, so they are only aﬀected by The (θϕ) component of the junction condition then re- polar perturbations and can be decomposed as duces to

−iωt 0 0 δρ(t, θ) = ερ1Pl(cos θ)e , 1p 8πηR [h0f − f (iωh1 + h0)] fh1 = √ . (A27) −iωt 2 0 δp(t, θ) = εp1Pl(cos θ)e , f (2f − Rf ) −iωt δR(t, θ) = εδR0Pl(cos θ)e . (A22) In vacuum the Regge-Wheeler functions are related to each other [78], h (r) = − f(r) d [f(r)h (r)]. One can where ρ , p , δR only depend on the unperturbed ra- 0 iω dr 1 1 1 0 use this relation to eliminate h (R) and h0 (R) in terms dius R, P (cos θ) are the Legendre polynomials8 and the 0 0 l of h (R), h0 (R) and h00(R). Furthermore, one can replace parameter ε keeps track of the order in perturbation, so 1 1 1 h and its derivatives by introducing the Regge-Wheeler that all contributions of order O(ε2) are negligible. 1 function, In the following we treat the axial and polar sectors separately. f(r) ψ(r) = h (r) (A28) r 1

a. Axial sector which satisfies Eq. (6) with effective potential Vaxial. We finally obtain The nonvanishing components of the axial metric per- ∂ψ RVaxial(R) turbations in the Regge-Wheeler gauge [78] are ωψ(R) = i16πη + ψ(R) . ∂x 2f(R) − Rf 0(R) −iωt R δgtϕ = εe h0(r) sin θ∂θPl(cos θ) , (A29) −iωt which coincides with Eq. (11) in the main text. δgrϕ = εe h1(r) sin θ∂θPl(cos θ) . (A23) It follows that the only nonvanishing component of the 3-metric perturbation δhab is b. Polar sector

−iωt δhtϕ = εe h0(R) sin θ∂θPl(cos θ) . (A24) The nonvanishing components of the polar metric perturbation are In the case of axial perturbations the the four compo- −iωt nents of the normal vector up to O(ε) are simply given δgtt = εe Pl(cos θ)f(r)H(r) , by the unperturbed one in Eq. (A10). As a consequence, −iωt H(r) the nonvanishing components of the extrinsic curvature δgrr = εe Pl(cos θ) , f(r) perturbation are −iωt δgtr = εe Pl(cos θ)H1(r) , 1 −iωt p 0 δg δKtϕ = e ε f (iωh1 + h0) sin θ∂θPl(cos θ) , ϕϕ −iωt 2 2 δgθθ = 2 = εe Pl(cos θ)r K(r) , (A30) 1 sin θ δK = − e−iωtεpfh cos θ∂ + sin θ∂2 P (cos θ) . θϕ 2 1 θ θ l and the location of the membrane is perturbed as in Eq. (A22). t r As for the fluid velocity, the components U , U and By projecting on the 3-dimensional membrane, the θ ϕ U are not affected by axial perturbations, while δu 6= nonvanishing components of the 3-metric perturbation 0 and its expression can be found by solving the (tϕ) δhab read component of the junction condition, δh = ε (fH − f 0δR ) P (cos θ)e−iωt , −iωt 0 0 tt 0 l ϕ εe ∂θPl(cos θ)[h0f − f (iωh1 + h0)] δu = √ . (A25) hϕϕ 2 −iωt Rsinθ f (2f − Rf 0) δhθθ = = ε R K + 2RδR0 Pl(cos θ)e . (A31) sin2θ Moreover, the perturbed components of the normal vector for polar perturbations up to first order in ε are 8 Note that we can set m = 0 in the spherical harmonics −iωt Y m(θ, ϕ) ∝ eimϕP m(cos θ) without any loss of generality due ε iωe Pl(cos θ)δR0 l l δnt = , to the spherical symmetry of the background. pf(r) 14

−iωt εe Pl(cos θ)H(r) Moreover, δU ϕ = δuϕ = 0 and δU r = δnr = p , t −iωt θ θ 2 f(r) −εU iωe Pl(cos θ), while δU = δu is nonvanishing −iωt and can be determined by solving the (tθ) component of εe ∂θPl(cos θ)δR0 δnθ = − . (A32) the junction condition (A8), as we will see below. pf(r) As for the perturbed ﬂuid velocity, in the polar sector δU t can be uniquely determined from the condition on µ the norm UµU = −1,

ε (fH − δR f 0) δU t = δut = 0 P (cos θ)e−iωt . (A33) 2f 3/2 l

We can now compute the polar perturbation in the extrinsic curvature whose nonvanishing components up to ﬁrst order in ε are

−iωt h 2 02 00 0 2 0i εe Pl(cos θ) δR0 4ω − f + f (−2δR0f + 3Hf + 4iωH1) + 2f H δK = √ , tt 4 f εe−iωt f 2δR − RH + R2K0 + 2RK + δR Rf 0 − 2∂2 P (cos θ) δK = 0 √ 0 θ l , θθ 2 f ε sin2 θe−iωt f 2δR − RH + R2K0 + 2RK + δR (Rf 0 − 2 cot θ∂ ) P (cos θ) δK = 0 √ 0 θ l , ϕϕ 2 f εe−iωt∂ P (cos θ)(fH − 2iδR ω) δK = − θ l √ 1 0 . (A34) tθ 2 f The nonvanishing components of the perturbation to the stress-energy tensor are

−iωt 0 δTtt = εe Pl(cos θ)[ρ0f δR0 + f (ρ1 − ρ0H)] , R n δT = √ −pfR (ζ + η)∂ δuθ + (ζ − η) cot θδuθ θθ f θ −iωt hp io +εe Pl(cos θ) f (p0RK + 2p0δR0 + p1R) + iζω(RK + 2δR0) , R sin2 θ n δT = √ −pfR (ζ − η)∂ δuθ + (ζ + η) cot θδuθ ϕϕ f θ −iωt hp io +εe Pl(cos θ) f (p0RK + 2p0δR0 + p1R) + iζω(RK + 2δR0) ,

2p θ δTtθ = −R f(ρ0 + p0)δu . (A35)

(q + 1)M From the components (tt), (θθ), and (ϕϕ) of the junction + iωr2 + H (r) , (A38) conditions we obtain analytical (albeit cumbersome) ex- iωr 1 pressions for ρ1 and p1, and the deviation in the membrane location where q = (l − 1)(l + 2)/2. The last relation allows us to eliminate, say, H(r). 16πηRfH1 δR0 = , (A36) Moreover, we can rewrite H1(r) and K(r) in terms of 2f − Rf 0 − 32πηiωR the Zerilli function Z(x), which satisﬁes Eq. (6) with V = whereas from the (tθ) component we derive Vpolar [79, 80]. Indeed, √ εe−iωt∂ P (cos θ)R fH dZ(x) δuθ = θ l 1 . (A37) H1(r) = ωh(r)Z(x) + ωk(r) , (A39) 2f − Rf 0 − 32πηiωR dx dZ(x) K(r) = g(r)Z(x) + , (A40) This still leaves the metric perturbations, H(r), H1(r) dx and K(r), unknown. In vacuum, these three functions are related by the following algebraic equation [79, 80]: where [79, 80] 1 ω2r4 r − 3M 3qMr − qr2 + 3M 2 H(r) = qr − + M K(r) h(r) = i , qr + 3M r − 2M r − 2M (r − 2M)(qr + 3M) 15

r2 k(r) = −i , where the sound speed squared is given by Eq. (A15). r − 2M By substituting the above algebraic equations, after a q(q + 1)r2 + 3qMr + 6M 2 tedious but conceptually simple calculation we obtain the g(r) = 2 . (A41) r (qr + 3M) following boundary condition The last condition which closes the system of equations and uniquely determines the boundary conditions for the ∂ Z(x) polar metric perturbations at r = R is found from the x = −16πηiω + G(R, ω, η, ζ) , (A43) barotropic equation of state p = p(ρ), at ﬁrst order in Z(x) perturbation, which gives

2 δp = csδρ , (A42) where G = A/B and

n 2 A = (y − 2) 9 −3 − 3w2y3 + w2y4 − 48iπwy2ζ + y(2 + 96iπwζ) + q3y2 −3i + iy + 16πwy2η +3q −9 + 3y2(1 + 64iπwζ) − 32iπwy3(3ζ + 2η) + wy4(−3w + 16iπη) + w2y5 1 + 256π2η2 +q2y −18 + 6y2(1 + 16iπw(ζ − 2η)) + 256π2w2y5η2 + 32πwy4η(i + 24πwη) − iy3(−i + 48πw(ζ + η)) , (A44) B = y2(3 + qy) 3 −3 − 3w2y3 + w2y4 − 48iπwy2ζ + y(2 + 96iπwζ) +q −9 + 9y + w2y5 + y2(−3 + 192iπwζ) + y3(1 − 32iπw(3ζ − η)) − wy4(3w + 16iπη) +q2y 3 − 3y − 16iπwy3(ζ + η) + y2(1 + 32iπw(ζ + η)) , (A45)

and we have deﬁned the dimensionless quantities y = density ρ and the expansion Θ vanish for a perturbed R/M, w = Mω. Note that in the BH limit G(2M) = Schwarzschild spacetime. In the BH limit the eﬀective 0, and the BH boundary condition is recovered for η = stress-energy tensor (4) reduces to 1/(16π), consistently with the axial case. BH Tab = pγab − 2ηBHσab ,R → 2M, (A46) Finally, from our computations it follows that both the which is independent of ζ, as discussed in the main text.

[1] W. E. Lamb and R. C. Retherford, “Fine structure of gravitational wave observations,” Class. Quant. Grav. the hydrogen atom by a microwave method,” Phys. 21 (2004) 787–804, arXiv:gr-qc/0309007 [gr-qc]. Rev. 72 (Aug, 1947) 241–243. https: [8] E. Berti, V. Cardoso, and C. M. Will, “On //link.aps.org/doi/10.1103/PhysRev.72.241. gravitational-wave spectroscopy of massive black holes [2] LIGO Scientific, Virgo Collaboration, B. Abbott with the space interferometer LISA,” Phys. Rev. D73 et al., “Tests of general relativity with GW150914,” (2006) 064030, arXiv:gr-qc/0512160 [gr-qc]. Phys. Rev. Lett. 116 no. 22, (2016) 221101, [9] K. D. Kokkotas and B. G. Schmidt, “Quasinormal arXiv:1602.03841 [gr-qc]. [Erratum: Phys.Rev.Lett. modes of stars and black holes,” Living Rev. Rel. 2 121, 129902 (2018)]. (1999) 2, arXiv:gr-qc/9909058 [gr-qc]. [3] LIGO Scientific, Virgo Collaboration, B. Abbott [10] E. Berti, V. Cardoso, and A. O. Starinets, et al., “Tests of General Relativity with the Binary “Quasinormal modes of black holes and black branes,” Black Hole Signals from the LIGO-Virgo Catalog Class. Quantum Grav. 26 (2009) 163001, GWTC-1,” Phys. Rev. D 100 no. 10, (2019) 104036, arXiv:0905.2975 [gr-qc]. arXiv:1903.04467 [gr-qc]. [11] The LIGO/Virgo Scientific Collaboration [4] C. V. Vishveshwara, “Stability of the schwarzschild Collaboration, B. P. Abbott et al., “Observation of metric,” Phys. Rev. D1 (1970) 2870–2879. Gravitational Waves from a Binary Black Hole [5] S. Chandrasekhar, The mathematical theory of black Merger,” Phys. Rev. Lett. 116 no. 6, (2016) 061102, holes. 1985. arXiv:1602.03837 [gr-qc]. [6] S. Detweiler, “Black holes and gravitational waves. III [12] E. Berti et al., “Testing General Relativity with - The resonant frequencies of rotating holes,” Present and Future Astrophysical Observations,” Astrophys. J. 239 (July, 1980) 292–295. Class. Quant. Grav. 32 (2015) 243001, [7] O. Dreyer, B. J. Kelly, B. Krishnan, L. S. Finn, arXiv:1501.07274 [gr-qc]. D. Garrison, and R. Lopez-Aleman, “Black hole [13] L. Barack et al., “Black holes, gravitational waves and spectroscopy: Testing general relativity through fundamental physics: a roadmap,” arXiv:1806.05195 16

[gr-qc]. arXiv:1804.08195 [gr-qc]. [14] E. Berti, K. Yagi, H. Yang, and N. Yunes, “Extreme [32] L. Buoninfante, A. S. Cornell, G. Harmsen, A. S. Gravity Tests with Gravitational Waves from Compact Koshelev, G. Lambiase, J. Marto, and A. Mazumdar, Binary Coalescences: (II) Ringdown,” Gen. Rel. Grav. “Towards nonsingular rotating compact object in 50 no. 5, (2018) 49, arXiv:1801.03587 [gr-qc]. ghost-free infinite derivative gravity,” Phys. Rev. D 98 [15] B. S. Sathyaprakash et al., “Extreme Gravity and no. 8, (2018) 084041, arXiv:1807.08896 [gr-qc]. Fundamental Physics,” arXiv:1903.09221 [33] T. Biswas, E. Gerwick, T. Koivisto, and [astro-ph.HE]. A. Mazumdar, “Towards singularity and ghost free [16] S. Hawking, “Particle Creation by Black Holes,” theories of gravity,” Phys. Rev. Lett. 108 (2012) Commun. Math. Phys. 43 (1975) 199–220. [Erratum: 031101, arXiv:1110.5249 [gr-qc]. Commun.Math.Phys. 46, 206 (1976)]. [34] V. P. Frolov, “Mass-gap for black hole formation in [17] V. Cardoso and P. Pani, “Testing the nature of dark higher derivative and ghost free gravity,” Phys. Rev. compact objects: a status report,” Living Rev. Rel. 22 Lett. 115 no. 5, (2015) 051102, arXiv:1505.00492 no. 1, (2019) 4, arXiv:1904.05363 [gr-qc]. [hep-th]. [18] V. Cardoso, L. C. B. Crispino, C. F. B. Macedo, [35] LIGO Scientific, Virgo Collaboration, R. Abbott H. Okawa, and P. Pani, “Light rings as observational et al., “GW190814: Gravitational Waves from the evidence for event horizons: long-lived modes, Coalescence of a 23 Solar Mass Black Hole with a 2.6 ergoregions and nonlinear instabilities of ultracompact Solar Mass Compact Object,” Astrophys. J. 896 no. 2, objects,” Phys. Rev. D90 no. 4, (2014) 044069, (2020) L44, arXiv:2006.12611 [astro-ph.HE]. arXiv:1406.5510 [gr-qc]. [36] T. Damour, “Surface Effects in Black-Hole Physics,” in [19] P. V. Cunha and C. A. Herdeiro, “Stationary black Marcel Grossmann Meeting: General Relativity, p. 587. holes and light rings,” Phys. Rev. Lett. 124 no. 18, Jan., 1982. (2020) 181101, arXiv:2003.06445 [gr-qc]. [37] K. S. Thorne, R. Price, and D. Macdonald, eds., [20] S. Hawking, “Breakdown of Predictability in BLACK HOLES: THE MEMBRANE PARADIGM. Gravitational Collapse,” Phys. Rev. D 14 (1976) 1986. 2460–2473. [38] V. Cardoso, E. Franzin, and P. Pani, “Is the [21] O. Lunin and S. D. Mathur, “AdS / CFT duality and gravitational-wave ringdown a probe of the event the black hole information paradox,” Nucl. Phys. B horizon?,” Phys. Rev. Lett. 116 no. 17, (2016) 171101, 623 (2002) 342–394, arXiv:hep-th/0109154. arXiv:1602.07309 [gr-qc]. [22] O. Lunin and S. D. Mathur, “Statistical interpretation [39] V. Cardoso, S. Hopper, C. F. B. Macedo, of Bekenstein entropy for systems with a stretched C. Palenzuela, and P. Pani, “Gravitational-wave horizon,” Phys. Rev. Lett. 88 (2002) 211303, signatures of exotic compact objects and of quantum arXiv:hep-th/0202072 [hep-th]. corrections at the horizon scale,” Phys. Rev. D94 [23] S. D. Mathur, “The Fuzzball proposal for black holes: no. 8, (2016) 084031, arXiv:1608.08637 [gr-qc]. An Elementary review,” Fortsch. Phys. 53 (2005) [40] V. Cardoso and P. Pani, “Tests for the existence of 793–827, arXiv:hep-th/0502050 [hep-th]. black holes through gravitational wave echoes,” Nat. [24] R. C. Myers, “Pure states don’t wear black,” Gen. Rel. Astron. 1 no. 9, (2017) 586–591, arXiv:1709.01525 Grav. 29 (1997) 1217–1222, arXiv:gr-qc/9705065 [gr-qc]. [gr-qc]. [41] Z. Mark, A. Zimmerman, S. M. Du, and Y. Chen, “A [25] I. Bena and N. P. Warner, “Black holes, black rings recipe for echoes from exotic compact objects,” Phys. and their microstates,” Lect. Notes Phys. 755 (2008) Rev. D96 no. 8, (2017) 084002, arXiv:1706.06155 1–92, arXiv:hep-th/0701216 [hep-th]. [gr-qc]. [26] V. Balasubramanian, J. de Boer, S. El-Showk, and [42] E. Maggio, P. Pani, and V. Ferrari, “Exotic Compact I. Messamah, “Black Holes as Effective Geometries,” Objects and How to Quench their Ergoregion Class. Quant. Grav. 25 (2008) 214004, Instability,” Phys. Rev. D96 no. 10, (2017) 104047, arXiv:0811.0263 [hep-th]. arXiv:1703.03696 [gr-qc]. [27] I. Bena and N. P. Warner, “Resolving the Structure of [43] J. Abedi, H. Dykaar, and N. Afshordi, “Echoes from Black Holes: Philosophizing with a Hammer,” the Abyss: Tentative evidence for Planck-scale arXiv:1311.4538 [hep-th]. structure at black hole horizons,” Phys. Rev. D96 [28] L. Buoninfante and A. Mazumdar, “Nonlocal star as a no. 8, (2017) 082004, arXiv:1612.00266 [gr-qc]. blackhole mimicker,” Phys. Rev. D 100 no. 2, (2019) [44] R. S. Conklin, B. Holdom, and J. Ren, “Gravitational 024031, arXiv:1903.01542 [gr-qc]. wave echoes through new windows,” Phys. Rev. D98 [29] P. Nicolini, A. Smailagic, and E. Spallucci, no. 4, (2018) 044021, arXiv:1712.06517 [gr-qc]. “Noncommutative geometry inspired Schwarzschild [45] N. Oshita and N. Afshordi, “Probing microstructure of black hole,” Phys. Lett. B 632 (2006) 547–551, black hole spacetimes with gravitational wave echoes,” arXiv:gr-qc/0510112. Phys. Rev. D99 no. 4, (2019) 044002, [30] A. S. Koshelev and A. Mazumdar, “Do massive arXiv:1807.10287 [gr-qc]. compact objects without event horizon exist in infinite [46] Q. Wang, N. Oshita, and N. Afshordi, “Echoes from derivative gravity?,” Phys. Rev. D96 no. 8, (2017) Quantum Black Holes,” Phys. Rev. D 101 no. 2, 084069, arXiv:1707.00273 [gr-qc]. (2020) 024031, arXiv:1905.00446 [gr-qc]. [31] L. Buoninfante, A. S. Koshelev, G. Lambiase, [47] V. Cardoso, V. F. Foit, and M. Kleban, “Gravitational J. Marto, and A. Mazumdar, “Conformally-flat, wave echoes from black hole area quantization,” JCAP non-singular static metric in infinite derivative 08 (2019) 006, arXiv:1902.10164 [hep-th]. gravity,” JCAP 1806 no. 06, (2018) 014, [48] A. Coates, S. H. Völkel, and K. D. Kokkotas, “Spectral 17

Lines of Quantized, Spinning Black Holes and their M. Cabero, C. Capano, T. Dent, B. Krishnan, Astrophysical Relevance,” Phys. Rev. Lett. 123 no. 17, G. Meadors, and A. H. Nitz, “Low significance of (2019) 171104, arXiv:1909.01254 [gr-qc]. evidence for black hole echoes in gravitational wave [49] L. Buoninfante, “Echoes from corpuscular black data,” Phys. Rev. D97 no. 12, (2018) 124037, holes,” arXiv:2005.08426 [gr-qc]. arXiv:1712.09966 [gr-qc]. [50] V. Ferrari and K. D. Kokkotas, “Scattering of particles [66] J. Abedi, H. Dykaar, and N. Afshordi, “Comment on: by neutron stars: Time evolutions for axial ”Low significance of evidence for black hole echoes in perturbations,” Phys. Rev. D62 (2000) 107504, gravitational wave data”,” arXiv:1803.08565 arXiv:gr-qc/0008057 [gr-qc]. [gr-qc]. [51] P. Pani and V. Ferrari, “On gravitational-wave echoes [67] R. S. Conklin and B. Holdom, “Gravitational Wave from neutron-star binary coalescences,” Class. Quant. ”Echo” Spectra,” arXiv:1905.09370 [gr-qc]. Grav. 35 no. 15, (2018) 15LT01, arXiv:1804.01444 [68] K. W. Tsang, A. Ghosh, A. Samajdar, [gr-qc]. K. Chatziioannou, S. Mastrogiovanni, M. Agathos, and [52] B. Holdom and J. Ren, “Not quite a black hole,” Phys. C. Van Den Broeck, “A morphology-independent Rev. D95 no. 8, (2017) 084034, arXiv:1612.04889 search for gravitational wave echoes in data from the [gr-qc]. first and second observing runs of Advanced LIGO and [53] L. Buoninfante, A. Mazumdar, and J. Peng, Advanced Virgo,” Phys. Rev. D 101 no. 6, (2020) “Nonlocality amplifies echoes,” Phys. Rev. D 100 064012, arXiv:1906.11168 [gr-qc]. no. 10, (2019) 104059, arXiv:1906.03624 [gr-qc]. [69] J. Abedi, N. Afshordi, N. Oshita, and Q. Wang, [54] A. Delhom, C. F. Macedo, G. J. Olmo, and L. C. “Quantum Black Holes in the Sky,” Universe 6 no. 3, Crispino, “Absorption by black hole remnants in (2020) 43, arXiv:2001.09553 [gr-qc]. metric-affine gravity,” Phys. Rev. D 100 no. 2, (2019) [70] B. Guo, S. Hampton, and S. D. Mathur, “Can we 024016, arXiv:1906.06411 [gr-qc]. observe fuzzballs or firewalls?,” JHEP 07 (2018) 162, [55] H. Nakano, N. Sago, H. Tagoshi, and T. Tanaka, arXiv:1711.01617 [hep-th]. “Black hole ringdown echoes and howls,” PTEP 2017 [71] R. Price and K. Thorne, “Membrane Viewpoint on no. 7, (2017) 071E01, arXiv:1704.07175 [gr-qc]. Black Holes: Properties and Evolution of the Stretched [56] P. Bueno, P. A. Cano, F. Goelen, T. Hertog, and Horizon,” Phys. Rev. D 33 (1986) 915–941. B. Vercnocke, “Echoes of Kerr-like wormholes,” Phys. [72] G. Darmois, Les équationsde la gravitation Rev. D97 no. 2, (2018) 024040, arXiv:1711.00391 einsteinienne. Gauthier-Villars, 1927. [gr-qc]. http://eudml.org/doc/192556. [57] Y.-T. Wang, Z.-P. Li, J. Zhang, S.-Y. Zhou, and Y.-S. [73] W. Israel, “Singular hypersurfaces and thin shells in Piao, “Are gravitational wave ringdown echoes always general relativity,” Nuovo Cim. B44S10 (1966) 1. equal-interval?,” Eur. Phys. J. C78 no. 6, (2018) 482, [Nuovo Cim.B44,1(1966)]. arXiv:1802.02003 [gr-qc]. [74] M. Visser, Lorentzian wormholes: From Einstein to [58] M. R. Correia and V. Cardoso, “Characterization of Hawking. AIP, Woodbury, USA, 1996. echoes: A Dyson-series representation of individual [75] T. Jacobson, A. Mohd, and S. Sarkar, “Membrane pulses,” Phys. Rev. D97 no. 8, (2018) 084030, paradigm for Einstein-Gauss-Bonnet gravity,” Phys. arXiv:1802.07735 [gr-qc]. Rev. D 95 no. 6, (2017) 064036, arXiv:1107.1260 [59] Q. Wang and N. Afshordi, “Black hole echology: The [gr-qc]. observer’s manual,” Phys. Rev. D97 no. 12, (2018) [76] H. A. Buchdahl, “General Relativistic Fluid Spheres,” 124044, arXiv:1803.02845 [gr-qc]. Phys. Rev. 116 (1959) 1027. [60] N. Uchikata, H. Nakano, T. Narikawa, N. Sago, [77] N. Oshita, Q. Wang, and N. Afshordi, “On Reflectivity H. Tagoshi, and T. Tanaka, “Searching for black hole of Quantum Black Hole Horizons,” JCAP 04 (2020) echoes from the LIGO-Virgo Catalog GWTC-1,” Phys. 016, arXiv:1905.00464 [hep-th]. Rev. D 100 no. 6, (2019) 062006, arXiv:1906.00838 [78] T. Regge and J. A. Wheeler, “Stability of a [gr-qc]. Schwarzschild singularity,” Phys. Rev. 108 (1957) [61] A. Testa and P. Pani, “Analytical template for 1063–1069. gravitational-wave echoes: signal characterization and [79] F. J. Zerilli, “Effective potential for even parity prospects of detection with current and future Regge-Wheeler gravitational perturbation equations,” interferometers,” Phys. Rev. D98 no. 4, (2018) 044018, Phys. Rev. Lett. 24 (1970) 737–738. arXiv:1806.04253 [gr-qc]. [80] F. Zerilli, “Gravitational field of a particle falling in a [62] E. Maggio, A. Testa, S. Bhagwat, and P. Pani, schwarzschild geometry analyzed in tensor harmonics,” “Analytical model for gravitational-wave echoes from Phys. Rev. D2 (1970) 2141–2160. spinning remnants,” Phys. Rev. D 100 no. 6, (2019) [81] E. Maggio, V. Cardoso, S. R. Dolan, and P. Pani, 064056, arXiv:1907.03091 [gr-qc]. “Ergoregion instability of exotic compact objects: [63] G. Ashton, O. Birnholtz, M. Cabero, C. Capano, electromagnetic and gravitational perturbations and T. Dent, B. Krishnan, G. D. Meadors, A. B. Nielsen, the role of absorption,” Phys. Rev. D99 no. 6, (2019) A. Nitz, and J. Westerweck, “Comments on: ”Echoes 064007, arXiv:1807.08840 [gr-qc]. from the abyss: Evidence for Planck-scale structure at [82] R. Brito, V. Cardoso, and P. Pani, “Superradiance,” black hole horizons”,” arXiv:1612.05625 [gr-qc]. Lect. Notes Phys. 906 (2015) pp.1–237, [64] J. Abedi, H. Dykaar, and N. Afshordi, “Echoes from arXiv:1501.06570 [gr-qc]. the Abyss: The Holiday Edition!,” arXiv:1701.03485 [83] M. Visser and D. L. Wiltshire, “Stable gravastars: An [gr-qc]. Alternative to black holes?,” Class. Quant. Grav. 21 [65] J. Westerweck, A. Nielsen, O. Fischer-Birnholtz, (2004) 1135–1152, arXiv:gr-qc/0310107 [gr-qc]. 18

[84] P. Pani, E. Berti, V. Cardoso, Y. Chen, and R. Norte, arXiv:gr-qc/9903032. “Gravitational-wave signatures of the absence of an [100] E. Barausse, V. Cardoso, and P. Pani, “Can event horizon. II. Extreme mass ratio inspirals in the environmental effects spoil precision gravitational-wave spacetime of a thin-shell gravastar,” Phys. Rev. D81 astrophysics?,” Phys. Rev. D89 no. 10, (2014) 104059, (2010) 084011, arXiv:1001.3031 [gr-qc]. arXiv:1404.7149 [gr-qc]. [85] P. Pani, “Advanced Methods in Black-Hole [101] J. L. Jaramillo, R. Panosso Macedo, and L. Al Sheikh, Perturbation Theory,” Int. J. Mod. Phys. A28 (2013) “Pseudospectrum and black hole quasi-normal mode 1340018, arXiv:1305.6759 [gr-qc]. (in)stability,” arXiv:2004.06434 [gr-qc]. [86] E. Leaver, “An Analytic representation for the quasi [102] S. Hild et al., “Sensitivity Studies for Third-Generation normal modes of Kerr black holes,” Proc. Roy. Soc. Gravitational Wave Observatories,” Class. Quant. Lond. A A402 (1985) 285–298. Grav. 28 (2011) 094013, arXiv:1012.0908 [gr-qc]. [87] M. Leins, H. Nollert, and M. Soffel, “Nonradial [103] H. Audley, S. Babak, J. Baker, E. Barausse, oscillations of neutron stars: A New branch of strongly P. Bender, E. Berti, P. Binetruy, M. Born, damped normal modes,” Phys. Rev. D 48 (1993) D. Bortoluzzi, J. Camp, C. Caprini, V. Cardoso, 3467–3472. M. Colpi, J. Conklin, N. Cornish, C. Cutler, et al., [88] O. Benhar, E. Berti, and V. Ferrari, “The Imprint of “Laser Interferometer Space Antenna,” ArXiv e-prints the equation of state on the axial w modes of (Feb., 2017) , arXiv:1702.00786 [astro-ph.IM]. oscillating neutron stars,” ICTP Lect. Notes Ser. 3 [104] V. Baibhav et al., “Probing the Nature of Black Holes: (2001) 35–46, arXiv:gr-qc/9901037. Deep in the mHz Gravitational-Wave Sky,” [89] P. Pani, E. Berti, V. Cardoso, Y. Chen, and R. Norte, arXiv:1908.11390 [astro-ph.HE]. “Gravitational wave signatures of the absence of an [105] T. Damour and S. N. Solodukhin, “Wormholes as event horizon. I. Nonradial oscillations of a thin-shell black hole foils,” Phys. Rev. D76 (2007) 024016, gravastar,” Phys. Rev. D80 (2009) 124047, arXiv:0704.2667 [gr-qc]. arXiv:0909.0287 [gr-qc]. [106] P. O. Mazur and E. Mottola, “Gravitational vacuum [90] S. Chandrasekhar and S. L. Detweiler, “The condensate stars,” Proc. Nat. Acad. Sci. 101 (2004) quasi-normal modes of the Schwarzschild black hole,” 9545–9550, arXiv:gr-qc/0407075 [gr-qc]. Proc. Roy. Soc. Lond. A A344 (1975) 441–452. [107] C. B. M. H. Chirenti and L. Rezzolla, “How to tell a [91] M. Giesler, M. Isi, M. A. Scheel, and S. Teukolsky, gravastar from a black hole,” Class. Quant. Grav. 24 “Black Hole Ringdown: The Importance of (2007) 4191–4206, arXiv:0706.1513 [gr-qc]. Overtones,” Phys. Rev. X 9 no. 4, (2019) 041060, [108] E. G. Gimon and P. Horava, “Astrophysical violations arXiv:1903.08284 [gr-qc]. of the Kerr bound as a possible signature of string [92] M. Isi, M. Giesler, W. M. Farr, M. A. Scheel, and S. A. theory,” Phys. Lett. B672 (2009) 299–302, Teukolsky, “Testing the no-hair theorem with arXiv:0706.2873 [hep-th]. GW150914,” Phys. Rev. Lett. 123 no. 11, (2019) [109] C. O. Lousto and B. F. Whiting, “Reconstruction of 111102, arXiv:1905.00869 [gr-qc]. black hole metric perturbations from Weyl curvature,” [93] E. Berti, J. Cardoso, V. Cardoso, and M. Cavaglia, Phys. Rev. D 66 (2002) 024026, arXiv:gr-qc/0203061. “Matched-filtering and parameter estimation of [110] B. Whiting and L. Price, “Metric reconstruction from ringdown waveforms,” Phys. Rev. D76 (2007) 104044, Weyl scalars,” Class. Quant. Grav. 22 (2005) arXiv:0707.1202 [gr-qc]. S589–S604. [94] V. Baibhav, E. Berti, V. Cardoso, and G. Khanna, [111] G. Raposo, P. Pani, and R. Emparan, “Exotic “Black Hole Spectroscopy: Systematic Errors and compact objects with soft hair,” Phys. Rev. D99 Ringdown Energy Estimates,” Phys. Rev. D97 no. 4, no. 10, (2019) 104050, arXiv:1812.07615 [gr-qc]. (2018) 044048, arXiv:1710.02156 [gr-qc]. [112] V. Cardoso, P. Pani, M. Cadoni, and M. Cavaglia, [95] S. Bhagwat, X. J. Forteza, P. Pani, and V. Ferrari, “Ergoregion instability of ultracompact astrophysical “Ringdown overtones, black hole spectroscopy, and objects,” Phys. Rev. D77 (2008) 124044, no-hair theorem tests,” Phys. Rev. D 101 no. 4, (2020) arXiv:0709.0532 [gr-qc]. 044033, arXiv:1910.08708 [gr-qc]. [113] V. Cardoso, P. Pani, M. Cadoni, and M. Cavaglia, [96] I. Ota and C. Chirenti, “Overtones or higher “Instability of hyper-compact Kerr-like objects,” harmonics? Prospects for testing the no-hair theorem Class. Quant. Grav. 25 (2008) 195010, with gravitational wave detections,” Phys. Rev. D 101 arXiv:0808.1615 [gr-qc]. no. 10, (2020) 104005, arXiv:1911.00440 [gr-qc]. [114] P. Pani, E. Barausse, E. Berti, and V. Cardoso, [97] X. J. Forteza, S. Bhagwat, P. Pani, and V. Ferrari, “Gravitational instabilities of superspinars,” Phys. “On the spectroscopy of binary black hole ringdown Rev. D82 (2010) 044009, arXiv:1006.1863 [gr-qc]. using overtones and angular modes,” [115] Y. Kojima, “Equations governing the nonradial arXiv:2005.03260 [gr-qc]. oscillations of a slowly rotating relativistic star,” Phys. [98] H.-P. Nollert, “About the significance of quasinormal Rev. D 46 (1992) 4289–4303. modes of black holes,” Phys. Rev. D 53 (1996) [116] N. Uchikata, S. Yoshida, and P. Pani, “Tidal 4397–4402, arXiv:gr-qc/9602032. deformability and I-Love-Q relations for gravastars [99] P. Leung, Y. Liu, W. Suen, C. Tam, and K. Young, with polytropic thin shells,” Phys. Rev. D94 no. 6, “Perturbative approach to the quasinormal modes of (2016) 064015, arXiv:1607.03593 [gr-qc]. dirty black holes,” Phys. Rev. D 59 (1999) 044034,