Horizons in a binary black hole merger I: Geometry and area increase

Daniel Pook-Kolb,1, 2 Ofek Birnholtz,3 Jos´eLuis Jaramillo,4 Badri Krishnan,1, 2 and Erik Schnetter5, 6, 7 1Max-Planck-Institut f¨urGravitationsphysik (Albert Einstein Institute), Callinstr. 38, 30167 Hannover, Germany 2Leibniz Universit¨atHannover, 30167 Hannover, Germany 3Department of Physics, Bar-Ilan University, Ramat-Gan 5290002, Israel 4Institut de Math´ematiquesde Bourgogne (IMB), UMR 5584, CNRS, Universit´ede Bourgogne Franche-Comt´e,F-21000 Dijon, France 5Perimeter Institute for Theoretical Physics, Waterloo, ON N2L 2Y5, Canada 6Physics & Astronomy Department, University of Waterloo, Waterloo, ON N2L 3G1, Canada 7Center for Computation & Technology, Louisiana State University, Baton Rouge, LA 70803, USA Recent advances in numerical relativity have revealed how marginally trapped surfaces behave when black holes merge. It is now known that interesting topological features emerge during the merger, and marginally trapped surfaces can have self-intersections. This paper presents the most detailed study yet of the physical and geometric aspects of this scenario. For the case of a head-on collision of non-spinning black holes, we study in detail the world tube formed by the evolution of marginally trapped surfaces. In the first of this two-part study, we focus on geometrical properties of the dynamical horizons, i.e. the world tube traced out by the time evolution of marginally outer trapped surfaces. We show that even the simple case of a head-on collision of non-spinning black holes contains a rich variety of geometric and topological properties and is generally more complex than considered previously in the literature. The dynamical horizons are shown to have mixed signature and are not future marginally trapped everywhere. We analyze the area increase of the marginal surfaces along a sequence which connects the two initially disjoint horizons with the final common horizon. While the area does increase overall along this sequence, it is not monotonic. We find short durations of anomalous area change which, given the connection of area with entropy, might have interesting physical consequences. We investigate the possible reasons for this effect and show that it is consistent with existing proofs of the area increase law.

I. INTRODUCTION Numerous merger events where two black holes merge to form a larger remnant black hole have now been ob- served by gravitational wave detectors [3–8]. The number One of the remarkable predictions of general relativity of detections will increase by orders of magnitude in the is the existence of black holes, purely geometric objects in next years as the detectors become more sensitive and a curved spacetime which behave like compact physical new generations of detectors are built. It is common to objects in numerous physical situations, and can power understand such mergers using event horizons. A well the most energetic phenomena in our universe. The prop- known example is [9], the “pair of pants picture”, which erties of spacetime near black holes are unlike anything clearly shows the merger of two disjoint surfaces to yield observed in flat space, even for very large black holes a final horizon. How should one think of the merger pro- where the curvature near the horizon is not necessarily cess in terms of marginally trapped surfaces, and does large. One of these unusual phenomena is the existence this yield a picture analogous to the “pair of pants”? of closed trapped surfaces. These are 2-dimensional closed The scenario of how two MOTSs merge has been re- surfaces which have the unusual property that even out- cently established for the first time numerically, and is going light rays emanating from the surface are conver- summarized in Fig. 1. This figure is obtained from a nu- gent. Such surfaces cannot exist completely contained merical solution of the full vacuum Einstein equations for arXiv:2006.03939v1 [gr-qc] 6 Jun 2020 in flat spacetime regions. In fact, within classical gen- the head-on collision of two unequal mass black holes. We eral relativity, they can only exist in geodesically incom- start with simplest puncture initial data where the black plete spacetimes, usually taken to indicate the presence holes have no spin or initial linear momentum, namely of a singularity [1, 2]. Marginally outer trapped surfaces Brill-Lindquist initial data [10]. The initial data is pre- (MOTS) are limiting cases of trapped surfaces where the scribed on Euclidean space with two points (the “punc- outgoing light rays have vanishing convergence. The out- tures”) removed. The data is time symmetric, i.e. the ermost MOTS on a given constant time hypersurface, extrinsic curvature vanishes. The 3-metric hab is confor- 4 also known as an apparent horizon, can be shown lo- mally flat: hab = Φ δab. The conformal factor at a point cally to have the property of a one-way membrane, i.e. r is any material particle having fallen into it cannot cross it m1 m2 again. The outermost MOTS can also be shown to have Φ(r) = 1 + + , (1) 2r1 2r2 non-decreasing area and furthermore these are found to satisfy the laws of black hole mechanics. Black hole space- where r1 and r2 are the distances from r to the two punc- times however contain a much wider variety of MOTSs tures, d the distance between the punctures, and m1, m2 which have interesting physical and geometric properties. are the bare masses associated with the punctures. The 2

ADM mass is seen to be MADM = m1 + m2. There turns These results were first reported in [16], and the numer- out be a rich variety of MOTSs even in this simple initial ical method is detailed in [11, 16]. In particular, the exis- data; see [11] for a detailed study. We choose a partic- tence of self-intersecting MOTSs has been proven in de- ular configuration m1 = 0.5 and m2 = 0.8, and the ini- tail with high accuracy (cf. also the recent [17]). A num- tial coordinate separation between the two punctures is ber of questions still remain to be answered about this d0/MADM = 1. For these parameters, there are initially scenario. These include understanding geometric prop- only two disjoint MOTSs (representing the two black erties of the world tubes such as the status of the area holes) surrounding the two punctures. Throughout, we increase law, physical properties such as multipole mo- state times in units of := M /1.3. ments, fluxes of energy across the horizons, and the sta- M ADM As we evolve this initial data using the Einstein equa- bility properties. The goal of the present series of papers tions, the result is shown in Fig. 1. This is the analog is to discuss these physical properties in detail for this of the “pair of pants picture”. In this figure time goes same initial configuration. In this paper, the first of two vertically upwards and horizontal sections of the tubes parts, we shall discuss the geometric properties of the yield sections of the MOTSs at a given time (the MOTSs world tubes shown in Fig. 1. This includes the signa- are axially symmetric, and thus the full MOTS can be ture of the world tubes, the expansion of the ingoing null obtained by revolving these sections around their respec- rays, and most importantly, the area increase law. The tive axes). The tubes in red and purple are the world area of the final apparent horizon outer at late times is certainly larger than the sum of theS areas of and tubes of the two individual MOTSs. These get closer S1 S2 to each other and eventually touch at a time labeled at early times. Moreover, we can trace a sequence of sur- faces which takes us from the initially disjoint surfaces Ttouch, and go through each other after Ttouch. At a time and to the single final horizon . However, the Tbifurcate which is somewhat earlier than Ttouch, when the S1 S2 Souter two horizons get sufficiently close to each other, a com- area does not increase monotonically along this sequence; mon MOTS is formed outside the individual ones. This there are in fact short durations of area decrease along common MOTS immediately bifurcates into an inner and this sequence. This fact needs an explanation, and might outer branch shown respectively as a green mesh and in have important physical implications. In short, the geo- blue. The outer branch (in blue) becomes more symmet- metrical quantities studied in this first paper show how ric and reaches an equilibrium state corresponding to a the area increases. The second paper (henceforth paper MOTS of a Schwarzschild black hole horizon. The inner II) addresses the question of why the area increases, i.e. branch on the other hand becomes increasingly distorted. the energy fluxes across the horizon. Eventually it merges with the two individual MOTSs pre- The plan for the rest of the paper is the following. cisely at the time Ttouch. Finally, the inner MOTS devel- Sec. II sets up notation and briefly summarizes the basic ops self-intersections immediately after Ttouch. For refer- notions and results that we shall use later. The behavior ence, we find Ttouch 5.53781 ( 4.25985 MADM) and of MOTSs under time evolution, even for the simple case T 1.37460 ≈ ( 1.05738M ≈M ). bifurcate ≈ M ≈ ADM of a head-on collision, exhibits a rich variety of geometric It is useful to rephrase the above in terms of sections and physical properties, and we will need the full machin- of the world tubes of Fig. 1. Before time Tbifurcate the ery of quasi-local horizons to describe these features. For intersection of the world tubes with a Cauchy surface this reason we summarize different notions of quasi-local will consist of two disjoint spherical surfaces which we horizons, and we shall find it appropriate to modify ex- shall denote 1 and 2. The 3-dimensional world tubes isting terminology in some cases. Sec. III discusses a very generated byS them willS be denoted and respec- basic geometric and physical aspect, namely the area of H1 H2 tively. Between Tbifurcate and Ttouch, 1 and 2 continue the MOTSs, and the status of the area increase law. This to exist separately, but are now surroundedS byS a pair of will involve issues such as the signature of the world tube marginally trapped surfaces which enclose both and and whether the MOTSs are future or past trapped. The S1 2. The inner of these is denoted inner and the outer one classic area increase law for event horizons might lead us (theS apparent horizon) by S. The world tubes gen- to believe that area should always increase to the future. Souter erated by them are inner and outer. At Ttouch, 1 and The situation will be somewhat more complicated for us. touch and at laterH times theyH go through eachS other While the area of the MOTS does increase overall, we S2 while remaining spherical. Also at Ttouch, inner coincides find that there are small durations where this does not with , i.e. it forms a cusp. After TS , the cusps hold. Sec. IV studies in detail the expansion of the ingo- S1 ∪ S2 touch of inner develop into knots, i.e. self-intersections, which ing null normal. Future trapped surfaces have negative becomeS larger with time. The eventual fate of , and ingoing expansion everywhere, indicating the presence of S1 S2 inner is not yet known [12, 13]; these get closer to the a singularity to the future. We shall see that the dynami- puncturesS whence they become difficult to track numeri- cal horizons are not everywhere future trapped, and have cally (though constraints on their possible dynamics fol- portions with positive ingoing expansion. Sec. V studies low from general results precluding the change of topol- the signature of the dynamical horizons and it shows that ogy of Cauchy hypersurfaces during evolution [14, 15]). the horizons have both timelike and spacelike portions. outer continues moving outwards becoming ever more Sec. VI revisits the area increase, and considers the cor- symmetricS as it reaches its final equilibrium fate. respondence of geometric fields on the horizon with prop- 3

FIG. 1: The behavior of MOTSs in a binary black hole merger obtained from a numerical simulation of the full vacuum Einstein equations. The left panel shows an overview up to T = 8 > Ttouch and the right panel a close-up of the end section of the two individual horizons (purple, red) and the innerM common horizon (green) with the self-intersections visible. The blue contour describes the outer common horizon. See text for details. erties of a 2-dimensional fluid, first suggested within the consider higher genus surfaces, these are generically ex- membrane paradigm. Sec. VII concludes by discussing pected to be unstable [26] and we restrict ourselves to open questions and possible directions for future work. spherical topology in this paper. is naturally endowed Appendix A presents a detailed comparison of our re- with two null-normal fields denotedS `a and na. These sults with the proof of area monotonicity on a dynamical vector fields are required to be future-directed, null and horizon. Appendix B extends the membrane paradigm orthogonal to . We are free to rescale them by positive- analogy to spinning black holes, and finally Appendix C definite functions.S This rescaling freedom can be reduced speculates on a possible geometric interpretation of self- by fixing the inner-product ` n = 1 which ties the intersecting MOTSs. rescalings of ` and n: · −

1 ` f` , n f − n , f > 0 . (2) → → II. BASIC NOTIONS Since is spacelike, the spacetime metric gab, when re- strictedS to the tangent space of , yields a Riemannian A. The optical scalars and marginally trapped metric S surfaces qab = gab + `anb + na`b . (3)

We collect here the basic notions and definitions re- We denote the volume 2-form on by  , and the deriva- lated to quasi-local black hole horizons we shall need e S ab tive operator on compatible with qab is denoted . later. While we will try to be as self contained as possible, Integrals over willS be written with dA as the measure.D our goal in this section is not to provide a comprehen- S e Of special importance for us will bee the so called opti- sive overview of the subject, but rather to summarize the cal scalars, i.e. the expansion, shear, and twist of `a and connections between the different notions with suitable a n . The derivative a`b projected on can be separated references to the literature. Reviews with diverse view- into a symmetric and∇ anti-symmetricS part, and the sym- points can be found in e.g. [18–25]. metric part can in turn be decomposed into a trace and Spacetime is a 4-dimensional manifold M with a trace-free part: Lorentzian metric g with signature ( , +, +, +). The ab − derivative operator compatible with gab will be denoted c d 1 (`) (`) qaqb c`d = Θ(`)qab + σ + ω . (4) a, and the Riemann tensor Rabcd is defined according ∇ 2 ab ab ∇ d to ( a b b a)Xc = Rabc Xd for an arbitrary 1- ∇ ∇ − ∇ ∇ Here Θ is the expansion of `a, the symmetric tracefree form Xc. Let M be a smooth, orientable, closed (`) e e e S ⊂ (`) 2-dimensional spacelike surface. While it is possible to tensor σab is the shear, and the anti-symmetric tensor 4

(`) a a ωab is the twist of ` . The expansion, shear, twist of n require the Cauchy data, i.e. the 3-metric and extrinsic are defined analogously: curvature. For a closed 2-surface in Σ, let Ra be the spacelike outward pointing normalS to , and T a the unit- S c d 1 (n) (n) timelike normal to Σ. A convenient choice for the null q q cnd = Θ(n)qab + σ + ω . (5) a b ∇ 2 ab ab normals is

The moste importante objects fore us will be the two expan- a 1 a a a 1 a a (`) ` = (T + R ) , n = (T R ) . (8) sions, and σab . In general, whenever we refer to shear or √2 √2 − (`) σab, we shall mean σab unless indicated otherwise. The twist will vanish because, by construction, ` and n are The expansion condition Θ(`) = 0 can be written in terms of the extrinsic curvature K of Σ M : orthogonal to a smooth surface . ab ⊂ It will often be useful to completeS (`, n) to a null-tetrad D Ra + K RaRb K = 0 (9) (`, n, m, m¯ ) where ma is a complex null vector orthogonal a ab − a to both ` and n. Thus m is tangent to and satisfies where D is the derivative operator on Σ. Taking to m m¯ = 1 (m ¯ is the complex conjugate ofSm). Given the a S · be the level set of a suitable function h, this equation null-tetrad, the information contained in the symmetric- can, in turn, be expressed as a second order non-linear tracefree tensor σ on can be reduced to a single com- ab S differential equation for h. Our choice for h is based on plex field: choosing a reference surface sufficiently close to . Details on how the reference surface is chosen, the associatedS co- σ := σ mamb = mamb ` . (6) ab ∇a b ordinate system, and the numerical method for solving The choice of m can be changed by a phase: m eiψm. the above equation can be found in [11, 27] and our im- Under this change, the shear transforms as σ → e2iψσ plementation is available at [28]. This is an extension of which means that σ has spin weight +2. This implies→ that the widely used method developed in [29–34]. Our nu- Einstein Toolkit σ can be expanded into angular modes on using spin- merical calculations use the [35, 36]. We TwoPunctures weighted spherical harmonics of spin-weightS +2. This will use [37, 38] to set up initial conditions and McLachlan play a very important role in paper II. an axisymmetric version of [39] to solve the Einstein equations, which uses Kranc [40, 41] to generate The projections of a`b and anb given in Eqs. (4) and (5) are the two∇ extrinsic curvatures∇ of embed- efficient C++ code. For the results in the current series of papers, we performed simulations with three spatial ded in spacetime M . The other important quantityS is the connection on the normal bundle of . Since has resolutions 1/∆x = 960, 480, 60, running, respectively, S S until T = 7 , 20 , 50 . Further details of the co-dimension 2, the connection on the normal bundle is max M M M given by a single 1-form ω defined as simulation specific to our problem are detailed in [27]. a We now collect basic definitions and results pertaining c b to quasi-local horizons that we shall use while presenting ωa = nbq c` . (7) − a∇ our results. The goal here is not a detailed review of the This 1-form determines the angular momentum associ- subject, but mainly to orient the reader and to set up e ated with the horizon; we will always deal with non- notation and terminology. spinning black holes and will have ωa = 0 in this paper. is said to be a future-marginally-outer-trapped sur- S B. The MOTS stability operator face if Θ(`) = 0 and Θ(n) < 0. Note that the expan- sions Θ(`) and Θ(n) are also rescaled under the trans- formation of Eq. (2), but these conditions remain un- Starting with a MOTS on a Cauchy surface, it is nat- changed since `a and na remain future directed. If instead ural to ask how it behaves under time evolution. It is not Θ(n) > 0 (and still requiring Θ(`) = 0), then is said to a priori obvious that a MOTS should evolve smoothly. be past-marginally-outer-trapped. We shall oftenS just re- It is now known that the behavior of a MOTS under fer to marginally trapped surfaces with the understanding time evolution is controlled by a second order non-self- that we are referring to future-marginally-outer-trapped adjoint elliptic operator LΣ known as the stability oper- surfaces. Surfaces satisfying only Θ(`) = 0 (with no con- ator [11, 42–44]. LΣ is constructed from variations of dition on Θ ) are the marginally outer trapped surfaces, [26]. Given a surface , let be a family of surfaces pa-S (n) S Sλ or MOTS in short – these are the basic objects that we rameterized by a real variable λ; the surfaces λ depend shall study in this paper. As we shall see, Θ < 0 will smoothly on λ. We take to coincide withS , and λ (n) Sλ=0 S not always be satisfied; it is therefore necessary to keep can take values in an interval ( , ). The variation λ track of this condition in the various definitions and re- is assumed to be smooth which− implies that if we pickS sults. a point p on , the variation produces a smooth curve A MOTS is a geometric concept in the full 4- passing throughS p, and the tangent vector to these curves dimensional spacetime independent of any spatial slices. at λ = 0 defines a vector field X on . With this struc- In numerical simulations however, they are connected to ture, one can naturally define the variationS of geometric Cauchy surfaces because in order to locate them, we only quantities on [45, 46]. Of particular importance is the S 5 variation of Θ denoted δ Θ . On each construct C. Marginally trapped tubes and dynamical (`) X (`) Sλ the null normals `(λ) and n(λ) as for itself. This defines horizons (λ) S the expansion Θ(`) for all λ, and allows us to differentiate it: With the time evolution understood, we consider the three-dimensional world tube traced out by a MOTS. (λ) This world tube is known as a marginally outer trapped dΘ(`) δX Θ := . (10) tube. More formally (following [43]): (`)  dλ  λ=0 Definition 1 . A   (Marginally Outer Trapped Tube) This variation should not be confused with usual deriva- smooth 3-dimensional surface in a spacetime is said to be a marginally outer trappedH tube (MOTT) if tives of Θ(`). In particular, while δcX Θ(`) = cδX Θ(`) for constants c, it turns out that δ Θ = ψδ Θ if ψ is a ψX (`) 6 X (`) it has topology S2 R, i.e. it admits a foliation by non-constant function. If X is tangent to and Θ(`) = 0 • 2-spheres, × on , then it is obvious that δ Θ = 0.S Thus, we only S X (`) need to consider variation fields X normal to . One each leaf of the foliation is a MOTS. could consider X to be proportional to ` or n, butS in the • context of a Cauchy evolution, it is natural to take X Note that is allowed to have arbitrary signature and a a a H along the normal R , X = ψR . Thus, we define the no restrictions are placed on the ingoing expansion Θ(n) stability operator associated with Σ as for any of the MOTSs which constitute . As we have ⊃ S seen, the results involving the stabilityH operator men- LΣ [ψ] := √2 δψRΘ(`) , (11) tioned above do not assume any condition on Θ(n), and hold for any MOTT. The classic examples of MOTTs in where a global constant positive factor can be arbitrarily spherical symmetry are the well known Vaidya [50] and chosen, and √2 is chosen to simplify later expressions. Oppenheimer-Snyder [51] solutions. Further examples in An explicit calculation shows that LΣ is a second order spherical symmetry can be found in e.g. [52–54]. These elliptic operator but it is not necessarily self-adjoint. In examples already show the wide variety of cases that can vacuum spacetimes, the expression for LΣ is the follow- appear even in spherical symmetry. See [55] for a con- ing: struction of the spacetime locally near such a horizon. See e.g. [12, 34, 56, 57] for previous examples of numerical a LΣ [ψ] = ∆ ψ + 2ω aψ studies concerning dynamical horizons in various physi- − S D 1 cal situations. The present paper shall provide the most + + ωa ω ωa σ σab ψ . (12) 2R Da − a − ab detailed numerical study yet of these horizons in a black   hole merger. Here ∆ is the Laplacian on and a is the derivative Imposing additionally Θ(n) < 0 leads to the definition operatorS on . In the presentS case,D we shall deal with of a marginally trapped tube (following [58]): the head-on collisionS of non-spinning black holes so that Definition 2 (Marginally Trapped Tube). A MOTT is ωa = 0, whence LΣ will be self-adjoint and will have real eigenvalues. said to be a marginally trapped tube (MTT) if it satisfies In the dynamical evolution setting, the importance of in addition Θ(n) < 0 everywhere. LΣ lies in the following result [11, 42–44]: As we shall discuss below in Sec. III, the condition Θ(n) < 0 is employed in different proofs of the area in- A MOTS evolves smoothly as long as LΣ is invert- • ible, i.e. as long as none of its eigenvalues vanish. crease law. The proof of the area increase law [59–61] holds for an MTT of arbitrary signature, though with In simple cases when the smallest eigenvalue Λ0 is strictly additional technical assumptions we shall discuss later positive, then LΣ is obviously invertible and the MOTS (MTTs are referred to as holographic screens in this evolves smoothly. In a binary black hole merger, this is work). what happens for the outermost MOTS and the two in- Additional restrictions can be placed on a MOTT de- dividual MOTSs. However, as shown in [11, 16, 27], the pending on the physical situation one is interested in. inner common MOTS is more complicated. It is born at When the marginally trapped tube is in equilibrium, i.e. T with Λ = 0 which immediately becomes nega- there is no energy flux across , we need the notion of a bifurcate 0 H tive. None of the other eigenvalues cross 0 and the MOTS non-expanding horizon [62]: continues to evolve smoothly. It is clear from this that the Definition 3 . A smooth 3- complete spectrum of L , and not just its principal eigen- (Non-expanding horizon) Σ dimensional surface in a spacetime is said to be a non- value, is potentially of interest. This is especially true for ∆ expanding-horizon if spinning black holes when the eigenvalues can be com- plex, thus leading to the full MOTS-spectral problem for- ∆ has topology S2 R mulated in [47] and initiated in [48, 49]. We shall explore • × the spectrum of L in paper II. ∆ is null Σ • 6

Any null normal to ∆, denoted `a, has vanishing this paper. A general MOTT will be called a dynamical • expansion (Θ(`) = 0). horizon and qualifiers will be added as appropriate. We conclude this section by a short discussion of the The Einstein field equations hold at ∆ and, if Tab area increase and fluxes across dynamical horizons. Con- • is the stress energy tensor, a b is future directed Tb ` sider a portion ∆ between two MOTSs with initial area and causal when `a is future− directed. H Ai and final area Af . As shown in [45, 76, 82], the area From the properties of a null surface, it can be shown change Af Ai can be written as an integral over ∆ , with the integrand− being local fields on ∆ . The inte-H that every complete cross-section of ∆ is a MOTS. Thus, H a non-expanding horizon is, in essence, a MOTT with grand can be viewed as a flux, whence the area is seen null signature. The last condition is an energy condition to change due to the flux of radiation across the horizon. and is implied by, for example, the dominant energy con- We shall discuss the fluxes in great detail in paper II, but dition. It can also be shown that each cross-section of ∆ here we just mention two points: i) the dominant contri- (`) has the same area; the black hole here is in equilibrium bution is due to the shear σab , which was recently seen in an otherwise dynamical spacetime. Not all geometric to be closely correlated with the outgoing flux measured fields on a non-expanding horizon are time independent. at infinity [83]. Thus, the fluxes provide a critical link Further physical restrictions requiring the derivative op- between horizon dynamics and observations of gravita- erator to be time independent lead to the notion of an tional waves. ii) the fluxes are manifestly positive definite isolated horizon [63–65]. It is interesting to note that a for spacelike dynamical horizons, but not so for timelike version of the stability operator also appears in going cases [77]. from non-expanding to isolated horizons [65], and again, Besides these flux laws, there is an alternate formula- the invertibility of the stability operator turns out to be tion of the area change. The starting point is the mem- the relevant condition. brane paradigm for black hole event horizons [84–87]. Local constructions of spacetime neighborhoods near By applying the Einstein equations to an event horizon, non-expanding horizons is given in [55, 63, 66–71]. All Damour showed a close analogy between evolution equa- stationary black holes and Killing horizons, including tions on the horizon and the Navier-Stokes equation for of course the Schwarzschild and Kerr black holes, are a 2-dimensional fluid [84, 85]. In this way, it is possible non-expanding horizons. A detailed study of the Kerr- to relate fields on the event horizon and physical proper- Newman black hole viewed as a non-expanding horizon ties of a corresponding 2-dimensional fluid such as energy can be found in [72]. Apart from these, there are also density, pressure, bulk and shear viscosity. An interesting the so-called distorted black holes representing station- feature of this correspondence is that one obtains a nega- ary black holes in the presence of external fields [73] tive bulk viscosity for the fluid, suggesting an instability. (see also [74] for exact solutions representing charged dis- As shown in [88–90], this correspondence also holds for torted black holes). Distorted black holes can potentially dynamical horizons and one can similarly obtain coun- have positive Θ(n) [75]. terparts to the various physical quantities listed above. Moving now to the general dynamical case, we will In particular, the bulk viscosity now turns out to be pos- work with a general MOTT. As we shall see, a MOTT itive as expected. We shall explore certain aspects of this can be spacelike or timelike, or even have sections of analogy later in Sec. VI. mixed signature. In addition it can have positive or neg- ative Θ(n). In principle we could add qualifiers in front of MOTT and refer to, for example, spacelike or timelike III. THE AREA INCREASE LAW MOTTs. However, to minimize the number of acronyms and to perhaps make it easier to remember: The laws of black hole thermodynamics can be sat- Definition 4 (Dynamical Horizons). We shall refer to a isfactorily formulated using quasi-local horizons [62, 64, generic MOTT as a dynamical horizon. Additional qual- 76, 77, 91–93]. For example, formulations of the first law ifiers will be added as appropriate. Thus we can have of black hole mechanics based on event horizons [94, 95] spacelike or timelike dynamical horizons depending on use a mixture of quantities defined at the horizon (such the signature, and future or past depending on whether as the area) and infinity (such as the ADM mass, and Θ < 0 or > 0 respectively. also surface gravity which uses the timelike Killing vec- (n) tor normalized at infinity). The quasi-local formulation The reader might be aware that previously, dynam- of the first law satisfactorily addresses this problem, and ical horizons referred to spacelike MTTs [76, 77] (this coherently uses quantities defined only at the horizon. is closely related to but not the same as a future outer Here we shall not review all aspects of black hole ther- trapping horizon [78–81]). However, already in [77] (Ap- modynamics, and instead focus on one aspect, namely pendix B), timelike cases were considered and referred the area increase law. to as timelike dynamical horizons. Dynamical horizons The areas of the various horizons for our particular were always meant to refer to a general MOTT and the configuration are straightforward to calculate and have spacelike case was initially thought to be the most rele- been presented previously [16, 27]. Here we present the vant case. We shall therefore use different terminology in same data first in terms of the radii of the black holes; 7

see Fig. 2. For a spherical surface with area A, one can Then, with Θ(`b) = 0, we easily get define a radius R according to A/4π, known as the Θ a (nb) area-radius. This is straightforward to define for inner Dar = > 0 , (15) and , and their radii willp be denoted R S and − √2 Souter inner Router respectively. For the two individual horizons 1 where D is the derivative operator on . This shows S a b H and 2, we can similarly define an effective area radius that the area of the cross-sections of increases along S H R1+2 as (A1 + A2)/4π. Fig. 2 plots these radii as func- ra. Similarly, for a timelike dynamical horizon, the roles tions of time. To connect the two initial horizons to the of ra and τ a are interchanged and it is now τ a which is final one,p one can follow the curves along the segment I , a tangentb to . We get for the divergence of τ : then follow segment II backwards in time, and then seg- H b b Θ b ment III which takes us to the final remnant black hole. a (nb) Daτ = + < 0 , b (16) While the overall area change is of course positive, the √2 area does not change monotonically along I + II + III . a whence the area decreasesb along τ . The reader should There is a small duration of anomalous area increase on bear in mind that the null normals na and na are related segment II just prior to T . The second panel shows a touch by a scaling as in Eq. (2). Thusb the expansions of n a close-up near T and the anomalous area increase a touch and n are also related by a scaling,b but the sign of the of inner. For reference, the local minimum of the area expansion remains unchanged. S b occurs at Tmin = 5.50594 . These simple calculations illustrate the importance of M Fig. 3 shows the irreducible masses Mirr = A/16π of the sign of Θ(n) and the signature. The proof of the area the various horizons. For the two individual horizons increase law by Bousso & Engelhardt [59, 60] which we p S1 and 2, we show them separately and also the sum will discuss below, does not make any assumption on the S signature but does assume Θ(n) < 0. With this assump- tion, and additional technical requirements which will be (1) (2) Mirr + Mirr = A1/16π + A2/16π . (13) important, it can be shown that the area must be mono- tonic on a future dynamical horizon. Now, viewing p p Hinner This measure takes into account the interaction energy and outer as a single dynamical horizon (i.e. consider the H III II II between the two black holes, which in fact is quite signif- union of the segments , , and 0 shown in Fig. 2), except for the anomalous icant given that the separation between the black holes the area is in fact monotonic area increase starting from is small. Thus unlike Fig. 2, in Fig. 3, the curve for Tmin. We need to explain (1) (2) which of the conditions assumed in the proof are violated. Mirr + Mirr lies above the curves for the outer hori- outer Some obvious questions for us are thus: Do the various zons. For the apparent horizon, the value of Mirr at dynamical horizons behave as one might have naively ex- late times is a good approximation to the Bondi mass, pected? For example, is the outer common horizon every- i.e. the mass left in the spacetime after all the gravita- where spacelike and does it have Θ(n) < 0? Similarly, is tional radiation has left the system. Since this is a very the inner horizon always timelike? What happens during short simulation, the amount of radiation is small and the anomalous area increase shown in the second panel of the difference between this estimate of the Bondi mass Fig. 2? We shall now proceed to address these questions. and the ADM mass is smaller than 0.1%. In the next sections we shall examine the signature In the next sections we will go deeper into the var- and the behavior of Θ(n) for all the horizons. Keep in ious ingredients which determine the area change. Let mind the different role played by these two aspects: Θ(n) us therefore conclude this section by outlining why the is part of the extrinsic curvature of a MOTS, i.e. it is expansion Θ(n), and the signature of the dynamical hori- determined by how is embedded in a spacetime mani- zons, are the main ingredients we should be looking fold. The signature ofS the dynamical horizons , on the at. There are some simplified cases when the area in- other hand, necessarily involves MOTSs at differentH time crease law can be easily proved, namely for purely future- steps; we need to obtain at least a small portion of in spacelike and purely future-timelike dynamical horizons, order to evaluate its signature. Thus, we first investigateH i.e. we assume everywhere Θ(n) < 0 and fix the signature. Θ(n) followed by the signature. Let us start with the spacelike case, and let be the dy- namical horizon. The null normal choices ofH Eq. (8) are tied to a Cauchy surface intersecting the dynamical hori- IV. THE EXPANSION OF THE INGOING NULL zon. We need instead null-normals entirely determined RAYS by . Since is spacelike, it has a unit-timelike normal τ a.H The foliationH on determines a unit-spacelike vector As we have seen, the expansion of the inward point- a H field r , which we take to be outward pointing. Then, ing null rays, Θ(n), is of great importance for the area analogous to Eq. (8), a suitable choice of null normals is b increase law. The average Θ(n) over a closed 2-surface S b with area A and area 2-form  is S a a a a a τ + r a τ r 1 ` = , n = − . (14) Θ(n) = Θ(n) . (17) √2 √2 A e S IS b b b b b b e 8

2.3227 2.6 III Router

2.5

II R1+2 at Ttouch M M 2.3226 Rinner touch R/ R/ 2.4 T ∆R

bifurcate I 0 T I min. of Rinner

2.3 R1+2 II 0 min touch T T 2.3225 0 2 4 6 8 10 5.450 5.475 5.500 5.525 5.550 T/ T/ M M FIG. 2: The first panel shows the radii of the various horizons as functions of time. The orange dotted line is an “effective” area-radius for the two individual horizons, the blue curve is the radius of the apparent horizon, while the solid green curve is the radius of the inner common MOTS. It is possible to connect the initial radii with the final one by following the segment I + II + III (the segment II is followed backwards in time). The second panel shows a close-up near Ttouch where an anomalous increase in the area of the inner common horizon is observed.

1.6 M (1) + M (2) Souter irr irr 0.0 inner S1 touch

T S2 1.4 outer ) S n

Mirr ( Θ 0.5 − M 1.2 / inner irr Mirr M 1.0 1.0 (2) − Mirr 0 5 10 15 20 T/ 0.8 M (1) bifurcate touch M

T T irr FIG. 4: The average ingoing expansion 0.6 0 2 4 6 8 10 12 14 16 18 20 T/ M as one might have expected. Upon formation, there is a FIG. 3: Irreducible masses of the various horizons as small region with Θ(n) > 0 around the “waist” defined functions of time. as follows. Each MOTS is axisymmetric, and thus has an axial symmetry vector ϕa. This vanishes at two points which defines the two poles. We can calculate the proper length of each orbit of ϕa. The proper length vanishes at The average of Θ(n) is shown in Fig. 4 for all four horizons as functions of time. The initial data is time symmetric, the poles and, for a regular round sphere, it is maximum at the equator. However, for some of the MOTSs in our which implies that Θ(n) = 0 for 1,2 initially. The average becomes negative and remains negativeS at all times. The problem, we find that the proper circumference has a lo- cal minimum around the equator. This is most obvious common horizons are born with negative Θ(n) and they remain negative at all times. for inner near Ttouch where it looks like a figure eight (seeH second panel of Fig. 6), but it is also true for the The behavior of Θ(n) beyond the averages is more in- teresting. The individual horizons however remain bor- apparent horizon just when it is born at Tbifurcate. ing: Θ 0 for the individual horizons at all times. For , this portion soon disappears and we have (n) ≤ Houter Since the initial data is time symmetric, the individual the conventional Θ(n) < 0 at all points on the appar- black holes initially have Θ(n) = 0 but it is strictly nega- ent horizons after this. See Fig. 5. A similar feature was tive thereafter. The individual horizons are therefore con- also seen in [57]. Evidently, the small portion with the ventional future-dynamical horizons. The outer dynami- “wrong” sign of Θ(n) does not affect the area increase cal horizon generated by the apparent horizons are some- law for ; the relevant portion of the horizon is too Houter what more interesting: they do not always have Θ(n) < 0 small to have an overall effect and the apparent horizon 9 area is monotonically increasing. intuitively, the term along T a dominates and is positive a As might be expected, the inner horizon inner is yet and thus V is timelike. Similarly, when b and c have op- S a more interesting. As shown in Fig. 6, inner never truly posite signs, the term along R dominates. When b > c, becomes future marginally trapped, i.e.S it always has i.e. when b > 0 and c < 0, V a points outwards (i.e. along a a portion (around its “waist”) with positive Θ(n). This R by definition), and in the opposite case when b < c shrinks with time and eventually vanishes momentarily then it points inwards. We note that in our simulations, at Ttouch, but reappears immediately afterwards. Thus we by construction the MOTS is found on Cauchy surfaces see that inner is never truly future marginally trapped. referring to a given time, and we essentially construct H a However, just after Tbifurcate, inner is decreasing in area V by connecting a MOTS at a given time with another (as it should) despite this effect.S MOTS at a later time. This means that when it is time- We see then that dynamical horizons generally do not like, V a can never be past directed and thus case IV have Θ(n) < 0 everywhere – this is the takeaway message cannot occur in our simulations. from this section. This condition has been widely used in The null case corresponds to either of b or c vanishing, previous literature on quasi-local horizons to prove the and it is usually assumed that this does not occur on open area increase law. In particular, it is used in the proof sets (this is borne out in our numerical results). The null of the area increase law for future dynamical horizons of portions arise when transitions between any two of the arbitrary signature [59, 60]. Thus, these proofs are not four cases listed above.H Furthermore, only one of the signs directly applicable for inner and there is, strictly speak- can change in a transition; the vanishing of both b and ing, no contradiction.H However, there is a more subtle c means that V a vanishes which cannot happen as long reason why the proof of [59, 60] does not apply to inner, as the foliation of the MOTT is regular. If  is the area and we shall return to this point shortly. H 2-form on , and when is a MOTS so that Θ = 0, S S (`) then e

V. THE SIGNATURE OF THE DYNAMICAL V  = cΘ(n) . (20) HORIZONS L Locally, the area increase is determined by the product e e Given a MOTS and a dynamical horizon obtained cΘ(n). Thus, c plays a double role: the product bc deter- H mines the signature while the product cΘ(n) determines by evolving it, the signature of can be computed in a a straightforward way. For any pointH p on the world tube, the change in area. V is strictly null only on a set of we can construct three linearly independent vectors tan- measure zero. As discussed earlier, the null case is never- theless very important for conceptual reasons. All of the gent to , say eA (A = 1, 2, 3) and find their inner- productsHq := e e . The eigenvalues of q then well known stationary Kerr and Schwarzschild horizons AB A · B AB are null. Moreover, has b large and positive, and c yield the signature; if it has a negative eigenvalue it is Houter timelike, and it is spacelike if all eigenvalues are positive. small but negative in the limit of late times as it reaches If the matrix is degenerate then is a null surface. equilibrium, so that V V = 2bc ' 0. The same holds H for ad at early times.· − Alternatively, we can consider properties of a “time H1 H2 evolution” vector on the dynamical horizon. Consider a With this understanding, we can now present our re- sults regarding the signature. , and always dynamical horizon of arbitrary signature and arbitrary H1 H2 Houter a H turn out to be spacelike; this is consistent with them Θ(n). Let V be a vector field on such that it is orthog- onal to the leaves of the MOTSs whichH constitute , and being stable in the sense of the principal eigenvalue of it maps one foliation to the next. Thus, if on a dynamicalH the stability operator being positive [43] (this will be dis- cussed further in paper II). Only shows interesting horizon the MOTSs are labeled by a parameter λ, then Hinner a behavior in this regard. The signature of this world tube we can choose λ such that V ∂aλ = 1. Each MOTS is taken to lie on a given Cauchy surface and thus equippedS is shown in Figs. 8. As discussed earlier, this horizon de- with null normals (`a, na) according to Eq. (8). Since V a velops cusps and self intersections. From Fig. 8 we see clearly that is mostly timelike, but there are in- is orthogonal to , there must exist functions b and c on Hinner such that S teresting and non-negligible portions which are spacelike. S When it is initially formed at Tbifurcate, it is completely a a a V = b` + cn . (18) spacelike; it must of course agree there with outer which is always spacelike. However, it remains fullyH spacelike for Since ` n = 1, we have V V = 2bc. Thus, the signa- · − · − only a few time-steps after which most portions become ture of is controlled by the sign of bc; is spacelike if timelike; the region around the “waist” remains spacelike b and c Hhave different signs, and timelikeH if they have the for the longest. After this, inner remains entirely time- same signs. Readers more familiar with T a and Ra might H a like until just before Ttouch when the portion around the find the following expression for V more illuminating: larger black hole develops spacelike portions. The second √2V a = (b + c)T a + (b c)Ra . (19) panel of Fig. 8 shows a close-up of a portion around the − self-intersecting knot. We see there is only one change We identify the 4 cases shown in Fig. 7 depending on of signature as we traverse each knot. Fig. 9 shows the the signs of b and c. If b and c are both positive, then distance of the waist to the point where this change hap- 10

FIG. 5: The sign of Θ(n) for outer. Upon formation, there is a small portion around the waist that has Θ(n) > 0. This portion soon disappearsH after which time the outer dynamical horizon is a conventional spacelike future-trapped dynamical horizon. The second panel shows a zoom of just after it is formed. Houter pens. The region around the knot is becoming increas- vious paragraph. However, this proof requires the exis- ingly spacelike. tence of a MOTS which has c < 0 everywhere, i.e. it We can dissect this behavior further in terms of the requires that the spacelike portion contains at least one complete MOTS. We see that around T , has functions b and c introduced earlier. The apparent hori- touch Hinner zon at late times is easiest to understand: V a is space- complete MOTSs in the timelike portion, but none in the spacelike portion. Moreover, after T , violates like and outward pointing, but strongly tilted towards touch Hinner `a. Thus, c is small and negative, and b is positive which another requirement assumed in [59, 60], namely that means that, at late times, the dynamical horizon each MOTS should have disjoint “inside” and “outside” Houter generated by outer must be of type I . The same holds regions. Thus, again, there is no contradiction with the for and Sat early times. Closer to T when proof. Further details can be found in Appendix A. The S1 S2 bifurcate outer is growing rapidly, we must have b > 0 and c < 0, appendix also shows that a straightforward extension of Sbut c will not be small. The inner horizon is radically the proof to our case does not work. different. As it is born, it moves inwards rapidly and it is spacelike: it has b < 0 and c > 0 and is of type III . The spacelike portions at early times shown in Fig. 8 (at VI. THE ANOMALOUS AREA INCREASE AND the bottom of the world tube) are of type III . This world THE MEMBRANE PARADIGM tube soon becomes timelike of type II wherein b > 0 and c > 0. This continues till we approach Ttouch. Shortly be- In this final section, we indulge in some speculations on fore Ttouch, a part of inner again becomes spacelike: this the anomalous area increase. We have mentioned briefly is the portion which envelopsS the larger MOTS . How- S2 earlier that because of the relation between horizon area ever, in this spacelike portion it turns out that we have and entropy, the anomalous area increase of inner might b > 0 and c < 0, i.e. it is of type I . The inner horizon thus be physically significant. One approach whereH this might shows the following transitions: III II I (partially). → → play a role is in the fluid-gravity correspondence. For It is in fact this spacelike portion of type I which is re- black holes in d + 1-dimensional anti-deSitter space, it sponsible for the anomalous area increase shown in the is suggested in [96] that the area increase law for the right panel of Fig. 2. event horizon has a dual description in terms of an “en- To explain this, we need to go back to the Bousso- tropy current” defined for a relativistic fluid living on Engelhardt proof of the area increase law [59, 60]. A key the d-dimensional boundary. The calculation presented intermediate result in this work is Theorem IV.2 of [59] in [96] is perturbative, where many of the complications which shows that c cannot change sign. This would seem of full non-linear general relativity, such as those we have to rule out the transition II I described in the pre- studied, do not arise. In order to extend this correspon- → 11

FIG. 6: The sign of Θ for . As for the apparent horizon, has a portion around its waist with Θ > 0, (n) Sinner Sinner (n) and this portion becomes smaller over time but does not disappear. The second panel shows details near Ttouch. We see that the portion with positive Θ(n) momentarily vanishes at Ttouch, but reappears again immediately afterwards.

nature of the event horizon, it would be unusual if its properties could be mapped to a local hydrodynamics description (except in situations where it can be treated perturbatively). At present, a viable proposal for the dual entropy cur- rent for dynamical horizons is lacking. We suggest that binary mergers might provide an interesting test case. If each of the horizons 1 and 2 at early times have a dual hydrodynamics descriptionH H and so does the final horizon outer at late times, then the overall increase in area mightH be viewed as the increase in entropy due to the interaction and mixing between the two fluids. As we have detailed in this paper, the inner horizon inner provides the link between the initial and the final states.H Thus, if such a dual description is generally viable then inner, and in particular the quantity cΘ(n) appearing in FIG. 7: The four different types of time evolution vector Eq.H (20), is likely to play an important role. The product fields on a generic dynamical horizon. The right and left cΘ yields the time derivative of the area according to I III (n) quadrants, and respectively, refer to spacelike Eq. (20) averaged over the horizons: dynamical horizons. The top and the bottom quadrants, II IV and respectively, are timelike. Horizons in the ˙ V  = A = cΘ(n) dA . (21) right quadrant are moving outwards while those in the L S left quadrant move inwards. We do not have any IS IS IV The integrals of cΘe(n) as functions of time for 1, 2 horizons in the bottom quadrant ( ) since in our H H simulations V a can never be past directed. and outer are shown in Fig. 10 and these are consistent withH Fig. 2. Similarly, Fig. 11 for the inner horizon is consistent with the right panel of Fig. 2; the minimum of the area in the right panel of Fig. 2 is consistent with dence to non-perturbative situations, it has been argued the zero of cΘ(n) in Fig. 11. While these results are that the event horizon might not be the appropriate con- guaranteed mathematically, it is still a useful numerical cept, and one should consider dynamical horizons instead check since Eq.R (21) is an independent calculation of A˙ . [97, 98]. Indeed, because of the teleological and non-local There exists in fact a different description of a blackS 12

FIG. 8: The signature of the inner-horizon inner. Blue shows the spacelike portions and red is timelike. The null portions where the horizon transitions betweenH spacelike and timelike are shown as dotted lines. Upon formation (at Tbifurcate) inner is entirely spacelike. This phase however lasts for a very short time (and is not easy to make out in the figure).H It develops timelike portions and soon becomes fully timelike. The portions around the “waist” persist in remaining spacelike for the longest. At a later time, a little bit before Ttouch, the part of inner surrounding the larger black hole reverts to being spacelike. The right panel shows details near the self-intersection.H The thin dashed curve is the location of the self-intersection.

tioned there, the analogy between fields on black hole 1 horizons and a 2-dimensional fluid also works for . One of these quantities is the energy density which,H it turns out, is proportional to the expansion of V a. Since

2 touch T 0 Θ(V ) = cΘ(n), we see that the energy density is propor- M tional to cΘ . The interpretation of cΘ as an energy l/ 5.53 5.54 5.55 (n) (n)

∆ density means that has large negative energy when 1 Sinner it is formed, and its energy becomes positive after Tmin. To explain this more fully, we revisit the discussion of the quasi-local membrane paradigm in [88–90], adapting 0 it to MOTTs of arbitrary signature. Given a hypersurface 5.5 6.0 6.5 7.0 7.5 8.0 , we introduce the orthogonal vector W a (compare with T/ Ha M V in Eq. (18)) W a = b`a cna , (22) FIG. 9: Proper length ∆l of the curve segment − connecting the waist of inner to the point where the signature changes fromS timelike to spacelike. This satisfying V W = 0, W W = 2bc = V V . The evolution of its expansion· Θ =· bΘ cΘ − along· is written corresponds to the proper distance measured along the (W ) (`) − (n) H MOTS when going from the waist to the dotted line in as (cf. e.g. [45, 99]) Fig. 8. We cannot numerically resolve whether this 1 Θ = κ(V )Θ Θ Θ σ(V )σab signature change happens precisely at the cusp (i.e. LV (W ) (V ) − 2 (V ) (W ) − ab (W ) ∆l = 0) when T = T . touch G V aW b + ( b)Θ ( c)Θ − ab LV (`) − LV (n) + a (b c c b 2bc ω ) , (23) D Da − Da − a where κ(V ) = nbV a ` , σ(V ) = bσ(`) + cσ(n), σ(W ) = hole horizon in terms of fluids, namely that arising in − ∇a b ab ab ab ab the membrane paradigm mentioned in Sec. II C. As men- bσ(`) cσ(n) and G is the Einstein tensor. First we ab − ab ab 13

102 101 outer 1 S S2 0 1 10 10− S dA dA ) )

n 2 n 3 ( 10− ( 10− Θ Θ c c

R 4 R 5 10− 10− touch touch T T 6 7 10− 10− 0 5 10 15 20 0 2 4 6 8 T/ T/ M M

FIG. 10: Plot of the integral of cΘ(n) as a function of time for outer (left panel) and 1, 2 (right panel), on a logarithmic scale. As explained in the text, this quantity is essentiallyS the rate of changeS ofS the area. As expected, and consistent with Fig. 2, this shows that the area of outer increases rapidly on formation (just after Tbifurcate) and then settles down. The dip near 13 will be explainedS in paper II. The individual horizons and show the ∼ M S1 S2 opposite behavior, i.e. increasing rapidly only as Ttouch is approached.

100 0.2 Souter 50 Sinner Souter dA dA 0.1 inner

) ) S n n

( 0 ( Θ Θ c c

R R 0.0 50 − min touch T T 100 − 1.37 1.38 1.39 1.40 1.41 1.42 5.50 5.52 5.54 T/ T/ M M FIG. 11: Plot of the integral of cΘ for the inner horizon . The left panel shows the values just after T (n) Sinner bifurcate when inner is formed and rapidly decreases in area, while the right panel shows values near Ttouch. The results are consistentS with the area results. In particular, as we see from the second panel of Fig. 2, the area of has a Sinner minimum at Tmin. In the second panel of this plot, we get a zero at Tmin, consistent with it being the rate of change of the area. note that, if is a smooth event horizon (so in par- in terms of a total surface energy ticular a nullH hypersurface), by making b = 1, c = 0 E ˙ so that V a = W a = `a, we immediately recover the 1 1 A = ε  = Θ(V ) = cΘ(n) = S . (25) null Raychaudhuri equation. This equation was inter- E 8π 8π 8π IS IS IS preted by Price and Thorne in [86, 87] as an energy The rate of change of balance law by introducing the surface energy density e E e e ¨ ε = Θ(W )/8π = Θ(`)/8π. For a dynamical horizon, ˙ A − − = S = V (ε) = V ε + Θ(V )ε  , (26) namely with Θ(`) = 0 and thus Θ(W ) = Θ(V ), we can E 8π L L − IS IS write
is controlled by Eq. (24).e This can be cast ase an energy balance law 2 (V ) 1 2 (V ) ab V Θ(V ) + Θ(V ) = κ Θ(V ) + Θ(V ) + σab σ(W ) (V ) L − 2 κ Θ(V ) ε + Θ ε = Θ + Θ +G V aW b + ( ln c)Θ LV (V ) − 8π (V ) (V ) 16π ab LV (V )     a ab + (c ab b ac + 2bc ωa) . (24) σ D D − D (V ) (W ) a +σab + Π aQ , (27) 8π ! − D Identifying again ε = Θ(W )/8π as a formal surface energy density, that in− the MOTT case translates into where κ(V )/8π is a surface tension (2-dimensional pres- ˙ ε = Θ(V )/8π = cΘ(n)/8π, we can interpret A in Eq. (21) sure term), Θ(V ) is the fluid expansion (so that a bulk 14

(V ) viscosity coefficient ζ = 1/16 can be identified), σab and in part by the Government of Canada through the De- (W ) partment of Innovation, Science and Economic Develop- σab /8π are, respectively, the shear strain and stress ten- sors (in general not proportional, so MOTTs do not corre- ment Canada and by the Province of Ontario through spond to Newtonian fluids and therefore a shear viscosity the Ministry of Colleges and Universities. We also thank a b 1 the French EIPHI Graduate School (ANR-17-EURE- µ cannot be defined), Π := TabV W + 8π ( V ln c)Θ(V ) is an external energy production rate (enforcedL by the Ein- 0002) and the Spanish FIS2017-86497-C2-1 project (with stein equations) and Qa := (b ac c ab 2bc ωa)/8π is FEDER contribution) for support. heat flux. D − D − This fluid description arising in the membrane paradigm is however only a formal analogy. There is, Appendix A: Comparison with the unlike in the fluid-gravity correspondence, no deeper in- Bousso-Engelhardt area increase law terpretation in terms of any dual boundary description or otherwise. Nevertheless, it is still interesting that the In this Appendix, we show that the anomalous area analogy goes through for the inner horizon as well. In increase does not violate any of the existing proofs of Appendix B, we show that the analogy also extends to the area increase laws. The most general proof to-date is spinning black holes, i.e. to include the rotation 1-form due to Bousso & Engelhardt [59, 60], and an extension ωa, which serves to define a momentum density on . thereof due to Sanches & Weinberg [61]. These proofs for- We shall make further use of this analogy in paper II.H In mally apply to future dynamical horizons, or holographic particular following [100, 101] in this viscous fluid pic- screens in the terminology of [59, 60]. We note that this ture, the evolution equation for Θ(V ) will be employed proof is in fact an application of the maximum principle to introduce decay and oscillation timescales leading to a for elliptic operators. The application of the maximum slowness parameter [102] for the approach of a dynamical principle to null-surfaces was studied previously by Gal- horizon to equilibrium. loway [103]. These methods have previously been applied by Ashtekar & Galloway to spacelike dynamical horizons to show e.g. the uniqueness of the foliation by MOTSs VII. CONCLUSIONS [58]. We shall discuss the proof in more detail below, but roughly speaking, these results would naively indi- In this paper we have studied geometrical properties cate that the inner horizon should have decreasing area of the world tube of marginally outer trapped surfaces in and would seem to rule out the anomalous area increase. a binary black hole merger. This includes the status of While these results assume Θ(n) < 0, this is only a suf- the area increase law, and the different ingredients which ficient but not necessary condition. The regions on the go into the rate of change of the area, i.e. the expan- horizon with Θ(n) > 0 are small near Ttouch and do not, sion of the ingoing null normal Θ(n) and the signature by themselves, explain the area increase. In other words, of the world tube. We have seen that the horizons are of we can split the area (and similarly also its rate of change) mixed signature with various transitions between space- as the sum of two terms depending on the sign of Θ(n): like and timelike portions, especially for the inner hori- zon. Cross-sections of the inner horizon can be of mixed A = dA + dA . (A1) signature. Similarly, the condition Θ < 0 is not sat- S (n) Θ(n)<0 Θ(n)>0 isfied everywhere for the inner horizon. The anomalous Z Z area increase apparently contradicts existing proofs of the The area over the Θ(n) < 0 portion turns out to be much area increase law, in particular the Bousso-Engelhardt re- larger than the second term, and it has a correspondingly sult. We have argued that technical assumptions required larger effect on A˙ . We saw for example that the area of S for this proof do not hold in our case. We have briefly outer increases as expected despite it not having Θ(n) < discussed the anomalous area increase in terms of the S0 everywhere. membrane-paradigm analogy using the energy density of The difficulty, and apparent contradiction, lies else- a fictitious 2-dimensional fluid. The deeper physical sig- where. It is the spacelike portion on the inner horizon nificance of the anomalous area increase, if any, is still near Ttouch that leads to the anomalous area increase. not understood. A key ingredient of the Bousso-Engelhardt results is an The second paper will continue this study and consider intermediate step showing that the function c is not al- physical quantities such as energy fluxes, multipole mo- lowed to change sign. Thus, if we actually had Θ(n) < 0, ments and the stability operator on all of these horizons. then cΘ(n) cannot change sign, and the area increase law follows directly from Eq. (20) (it must increase with time for regions I and IV , and decrease with time for regions ACKNOWLEDGMENTS II and III ). In the transition from timelike to spacelike of the inner horizon near Ttouch, which we have seen is We are indebted to Abhay Ashtekar, Ivan Booth and a transition II I of Fig. 7, this is precisely what goes Ricardo Uribe-Vargas for valuable discussions and sug- wrong: c goes from→ positive to negative. Moreover, this gestions. Research at Perimeter Institute is supported intermediate result only uses the condition Θ(`) = 0 and 15

FIG. 13: A pictorial sketch of the Bousso-Engelhardt FIG. 12: As a MOTS evolves in a spatially outward proof. The portion of with the offending timelike H direction from to , the null surface is nowhere region of type II with c > 0 is +. The rest of the 1 2 2 H S to theS future of . N horizon is of type I . The parameter λ increases N1 upwards. λ0 is the MOTS at λ0, and λ1 is the earliest + S + + that appears. is the portion of λ1+ within . The dashedH curveS depicts the intersectionS of withH the does not rely upon Θ(n) < 0. So how can this intermedi- H ate result be consistent with the results of this paper? To null surface − (generated by the past-inward directed null curves alongN `a starting from +). The point q is understand this we need to delve into some of the techni- − S cal conditions required in the Bousso-Engelhardt proof. where λ has a minimum over the dashed curve. At q, the null generator of the − coincides with the We shall not spell out the details of the proof, but we N outgoing null normal of the MOTS ? which is, by instead offer a pictorial description which will make the Sλ result plausible. definition, supposed to have vanishing expansion. The simplest case which we sketch here is the proof that the transition I II is ruled out. We thus start out with a MOTS evolving→ spatially outward, which later Since  is chosen sufficiently small, − lies to the past of ? N so that λ > λ . Let q ? be the point where the partially switches over to a future timelike direction. For Sλ0 0 ∈ Sλ a MOTS , we construct the outgoing null surface minimum is achieved. This construction is shown pictori- generatedS by the null-rays starting from `a. Denote theN ally in Figs. 13 and 14. Fig. 13 depicts the world tube, its part of to the future of by +, and the portion to timelike portion, and the intersection with −. The same N S N N the past by −. If we move 1 spatially outwards along situation is shown in Fig. 14 in terms of how these sur- a N S V to a new MOTS 2, the null surface 2 is easily seen faces would appear on Cauchy surfaces at various times. to be nowhere to theS future of ; seeN Fig. 12. If V a N1 partially changes to a future timelike direction, this will Then, it can be shown that − is tangent to λ? ; they N S eventually cease to hold. share the same null normal at q so that by a suitable a a rescaling we can set k = ` at q. However, since ? is Consider now the dynamical horizon, i.e. the world Sλ tube spanned by as it evolves along V a. Let the part “inside” −, it has larger curvature, and thus we must S N of with c > 0 be denoted as +, the part with c < 0 have Θ(`)(q) Θ(k)(q) where Θ(k)(q) is the expansion of H H 0 a ≥ as −, and the part with c = 0 as . Assume that k at q. From the Raychaudhuri equation, and assuming H H I H the null energy condition and the positivity of σ(k) 2, it has initially a leaf completely of type , i.e. we assume | | c < 0 on a complete leaf. Let λ be the affine parameter follows that Θ(k)(q) > 0. This implies then that Θ(`)(q) a a ≥ along V , i.e. V ∂aλ = 1, and let λ = λ0 at the initial Θ(k)(q) > 0 which contradicts Θ(`)(q) = 0 (the defining MOTS (which has c < 0). condition for a MOTS). Thus, + cannot exist and the Sλ0 H Let us take + to lie in the λ > λ region; this will transition I II is ruled out. H 0 → be shown to lead to a contradiction. Let λ1 > λ0 be the A similar argument, works for the transition from re- smallest value of λ on +, i.e. when the timelike portion gion II I , except we follow the null surface to the H future, and→ it is required to have a complete MOTS with first appears. Clearly, λ1 lies nowhere to the future of N c < 0. The other forbidden transitions where c changes λ0 . We can move further infinitesimally to λ1 +  still withoutN entering the future of . Consider then the sign are IV III and III IV . In both of these cases, λ0 → → MOTS and its subset +N = +. Then a direct attempt at applying the above argument does Sλ1+ S Sλ1+ ∩ H trace back the null rays to the past along −, and let not work. Instead, one needs to reverse the direction of a N k denote the null generator of −. Let us follow − V so that these are reduced to the previous two cases. back in time and look at its intersectionN with CauchyN The details of this proof are given in [59, 60]; see surfaces corresponding to various values of λ < λ1. The also [58, 103]. Here we list the technical conditions and intersection − is generically a curve on . Let the Lemmas which must hold for the above argument to go minimum valueN of∩ Hλ on this intersection be denotedH λ?. through: 16

area increase which occurs before Ttouch. The culprit is then condition (2): We see that near Ttouch, we have c < 0 only over a part of the MOTSs before Ttouch. Thus, the proof of [59, 60] does not rule out the anomalous area increase scenario presented above. Repeating the proof by dropping the requirement of a complete MOTS with c < 0, and replacing it with requiring that there should be a complete MOTS in region II (as is the case here) is seen to not work either. To show this, let us look explicitly at the case of inter- est to us, namely a transition II I (future-timelike to outward-spacelike). The transition→ is partial and we do not have any section which is entirely spacelike. We do however have a complete MOTS in the timelike portion. The picture is very similar to Fig. 13, with the timelike and spacelike portions interchanged. Thus, + becomes + H − and can be labeled −. Similarly, Fig. 14 can H S S be reused but with the curves for − and the various FIG. 14: Same as Fig. 13, but now showing how the N interchanged. Since − is spacelike, it contracts faster various surfaces would appear on Cauchy surfaces. We S S ? than −, and thus it goes inside −. At λ , we again start with λ1 +  where we start tracing back −. The N N N have − tangent to λ? , but critically, λ? now lies “out- color scheme is consistent with Fig. 13: the portion in N S S side” −. Thus, while we still have Θ (q) > 0, but red is the timelike portion +, the blue curve the N (k) S now Θ(`)(q) Θ(k)(q), and no contradiction arises with MOTS, while the dashed curve shows the intersection of ≤ Θ(`)(q) = 0. − with the Cauchy surfaces. The timelike portion will Ncontract slower than the null surface, and will thus stay ? outside the null portion. Eventually, at λ , the Appendix B: Damour-Navier-Stokes equation in intersection is the point q. ? is tangent to − at q, Sλ N MOTTs and the rest of it lies “inside” −. The maximum principle is applied in a neighborhoodN of q. For completeness, in the context of the quasi-local membrane paradigm discussed in section VI, we present here the equation for the evolution of the rotation form 1. Each MOTS λ which foliate must have an “in- S H ωa, interpreted as a (Damour)-Navier-Stokes equation for side” and an “outside”, i.e. if they lie on a Cauchy the momentum density of the two-dimensional fluid. surface Σ, they must split it into two disjoint por- As we did for Θ(V ), we evaluate now the evolution of tions. ω along a hypersurface a H 2. The existence of a MOTS which has c < 0 every- 1 S0 ω + Θ ω = κ(V ) + Θ σ(W )b (B1) where – required for the point q which minimizes λ LV a (V ) a Da 2Da (V ) − Db a to exist. +q bG W c Θ b + Θ c . a bc − (`)Da (n)Da 3. Genericity condition on the zeros of b and c, ∂ + = Making first b = 1 and c = 0, i.e. V a = W a = `a, we 0 H = ∂ − – this excludes non-generic zeros of c or recover the evolution equation of the rotation form on a H H b (e.g. the zeros must not coincide with extrema, null hypersurface [21], in particular the one satisfied on and the functions cannot vanish in an open set). a general (smooth) event horizon 2 1 4. On any , σ must be positive definite (in the ω + Θ ω = κ(`) + Θ σ(`)b presenceS of| matter,| we would include the null en- LV a (`) a Da 2Da (`) − Db a b c ergy condition) - this ensures the positivity of +qa GbcW . (B2) Θ(k)(q). Following Damour [84, 85], this equation can be inter- 5. The proof of Θ (q) Θ (q) is given in [104]; preted as a Navier-Stokes equation by defining a mo- (`) ≥ (k) see [103] for a more general proof with, e.g. weaker mentum surface density πa = ωa/8π, leading to a New- − smoothness assumptions, explicitly using the max- tonian viscous fluid picture with negative bulk viscos- imum principle. ity ζ = 1/16π and shear viscosity µ = 1/16π. In our − MOTT case, making Θ(`) = 0 in Eq. (B1), we get All of these conditions do hold on , and Souter S1 S2 but not on where the first two are violated. (1) (V ) 1 (W )b inner V ωa + Θ(V )ωa = aκ + aΘ(V ) bσ a is violated byS the self intersecting MOTSs but this only L D 2D − D happens after T and is not relevant for the anomalous +q bG W c + Θ ln c . (B3) touch a bc (V )Da 17

Defining, as in the event horizon case, a momentum sur- described transition corresponds indeed to the generic face density πa = ωa/8π we get [88] curve metamorphoses (“perestroikas”) happening when − projecting a three-dimensional curve into a plane with (V ) κ Θ(V ) the projection direction changing with a parameter [106]. V πa + Θ πa = a + a L (V ) −D 8π D 16π Indeed, as illustrated in [17], such self-intersections seem     (W )b to have a “genericity” flavor in trapped regions. σa + b + fa , (B4) Such a projection view suggests a picture with the D 8π ! “lifted” curve as the fundamental object and self- intersections as an artifact of the projections. The natural that corresponds to a viscous fluid with the interpreta- question is: what is the appropriate higher-dimensional tion of the terms given after Eq. (27), namely a New- space in which MOTSs “truly” live? A suggestive ten- tonian fluid in the trace part of the viscous stress ten- tative answer is given in terms of contact structures sor (since it is proportional to the trace of the strain [107, 108]. tensor, i.e. the expansion Θ ) with positive bulk vis- (V ) Specifically, and dwelling now beyond axisymmetry, let cosity ζ = 1/16π, non-Newtonian in the shear stress (W ) us consider a n-dimensional spacetime M and its cotan- tensor part (since σ /8π is not in general propor- a ab gent bundle T ∗M with natural Liouville form Λ = padx . (V ) tional to the strain shear tensor σab ). The term fa := After removing the vanishing one-forms, T ∗M 0M∗ , q bT W c 1 Θ ln c, where Einstein equations are we can take quotient by one-form rescalings with\{ non-} − a bc − 8π (V )Da imposed, corresponds to an external force surface den- vanishing real numbers R∗. The resulting space PT ∗M = 1 sity . See [105] for a critical account of this viscous fluid (T ∗M 0M∗ )/R∗ is the “projectified cotangent bundle”. interpretation. Crucially,\{ at} each point, the kernel of Λ is invariant un- Equations (B1) and (B4) are valid for general signature der such R∗-rescalings and projects onto hyperplanes in a a MOTTs, where V can be proportional n . In contrast the tangent space of PT ∗M : such field of (non-integrable) with the Θ(V ) evolution and the energy balance equation hyperplanes defines a “contact structure” on PT ∗M . In a a discussed in section VI, where the “heat flux” term Q a local chart (x , pa), with say pn = 0, we can con- i 6 must be generalized, Eqs. (B1) and (B4) coincide exactly sider the “affine” chart (x , pi, z), i 1, . . . , n 1 with n i ∈ { − } with those in [88–90], that were originally restricted to z = x , pn = 1. Then Λ0 = pidx + dz is the contact form a deviations around the outgoing null vector ` . of the (2n 1)-contact manifold PT ∗M (cf. [108] for a discussion in− a general relativistic setting; see also [109]). The relevance of such contact structures is that they Appendix C: Contact structures: a tentative bridge rule the properties of light propagation in the geomet- from immersed MOTS to wavefronts and caustics ric optics approximation, in particular the geometry of light wavefronts and the formation of caustics [110]. In As observed in Fig. 8 and fully discussed in Section this sense, and given the constitutive relation between VI of [27], the inner common horizon transitions (marginally) trapped surfaces and light convergence, such Sinner from i) a smooth “embedded curve” before time Ttouch, a higher-dimensional geometry seems a promising setting into ii) a singular curve (vanishing of its differential) with in which lifts of spacetime MOTSs could exist as embed- cusps at Ttouch, and then to iii) an “immersed curve” with ded, properly not self-intersecting, surfaces. Elucidating self-intersections (knot) after Ttouch. In a complementary their ultimate relation to wavefronts and caustics could view to this “parametrised curve” perspective, we could shed light onto the here discussed patterns of MOTS dy- approach such a sequence in terms of projections: the namics and the understanding of the trapped region.

[1] Roger Penrose, “Gravitational collapse and space-time 1: A Gravitational-Wave Transient Catalog of Compact singularities,” Phys. Rev. Lett. 14, 57–59 (1965). Binary Mergers Observed by LIGO and Virgo during [2] S. W. Hawking and R. Penrose, “The Singularities of the First and Second Observing Runs,” Phys. Rev. X 9, gravitational collapse and cosmology,” Proc. Roy. Soc. 031040 (2019), arXiv:1811.12907 [astro-ph.HE]. Lond. A314, 529–548 (1970). [4] B. P. Abbott et al. (LIGO Scientific, Virgo), “Binary [3] B.P. Abbott et al. (LIGO Scientific, Virgo), “GWTC- Black Hole Mergers in the first Advanced LIGO Ob- serving Run,” Phys. Rev. X6, 041015 (2016), [erratum: Phys. Rev.X8,no.3,039903(2018)], arXiv:1606.04856 [gr- qc]. 1 To complete this picture, it is interesting to note that the prin- [5] Alexander H. Nitz, Collin Capano, Alex B. Nielsen, cipal eigenvalue Λo of the MOTS stability operator, that will Steven Reyes, Rebecca White, Duncan A. Brown, and be systematically studied in Paper II, also admits a membrane Badri Krishnan, “1-OGC: The first open gravitational- paradigm interpretation. Namely, it can be seen as the pressure wave catalog of binary mergers from analysis of public difference at the interphase between two fluids, in the spirit of a Advanced LIGO data,” Astrophys. J. 872, 195 (2019), Young-Laplace law [47]. 18

arXiv:1811.01921 [gr-qc]. arXiv:0901.4365 [gr-qc]. [6] Alexander H. Nitz, Thomas Dent, Gareth S. Davies, [23] Jose Luis Jaramillo, “An introduction to local Black Sumit Kumar, Collin D. Capano, Ian Harry, Simone Hole horizons in the 3+1 approach to General Rel- Mozzon, Laura Nuttall, Andrew Lundgren, and Mr- ativity,” Int. J. Mod. Phys. D20, 2169 (2011), ton Tpai, “2-OGC: Open Gravitational-wave Catalog arXiv:1108.2408 [gr-qc]. of binary mergers from analysis of public Advanced [24] S.A. Hayward, Black Holes: New Horizons, New hori- LIGO and Virgo data,” Astrophys. J. 891, 123 (2019), zons (World Scientific, 2013). arXiv:1910.05331 [astro-ph.HE]. [25] Valerio Faraoni and Angus Prain, “Understanding dy- [7] Tejaswi Venumadhav, Barak Zackay, Javier Roulet, namical black hole apparent horizons,” Lecture Notes Liang Dai, and Matias Zaldarriaga, “New Binary Black in Physics 907, 1–199 (2015), arXiv:1511.07775 [gr-qc]. Hole Mergers in the Second Observing Run of Advanced [26] R.P.A.C. Newman, “Topology and stability of marginal LIGO and Advanced Virgo,” (2019), arXiv:1904.07214 2-surfaces,” Class. and Quant. Grav. 4, 277–290 (1987). [astro-ph.HE]. [27] Daniel Pook-Kolb, Ofek Birnholtz, Badri Krishnan, [8] Barak Zackay, Tejaswi Venumadhav, Liang Dai, Javier and Erik Schnetter, “Self-intersecting marginally outer Roulet, and Matias Zaldarriaga, “Highly spinning and trapped surfaces,” Phys. Rev. D 100, 084044 (2019). aligned binary black hole merger in the Advanced LIGO [28] Daniel Pook-Kolb, Ofek Birnholtz, Badri Krishnan, and first observing run,” Phys. Rev. D 100, 023007 (2019), Erik Schnetter, “MOTS Finder version 1.1,” (2019), arXiv:1902.10331 [astro-ph.HE]. 10.5281/zenodo.2591105. [9] R. A. Matzner, H. E. Seidel, Stuart L. Shapiro, [29] Jonathan Thornburg, “A Fast Apparent-Horizon Finder L. Smarr, W. M. Suen, Saul A. Teukolsky, and J. Wini- for 3-Dimensional Cartesian Grids in Numerical Rel- cour, “Geometry of a black hole collision,” Science 270, ativity,” Class. Quant. Grav. 21, 743–766 (2004), 941–947 (1995). arXiv:gr-qc/0306056. [10] Dieter R. Brill and Richard W. Lindquist, “Interaction [30] Jonathan Thornburg, “Event and Apparent Horizon energy in geometrostatics,” Phys. Rev. 131, 471–476 Finders for 3 + 1 Numerical Relativity,” Living Rev. (1963). Rel. 10, 3 (2007), arXiv:gr-qc/0512169. [11] Daniel Pook-Kolb, Ofek Birnholtz, Badri Krishnan, and [31] Jonathan Thornburg, “Finding apparent horizons in nu- Erik Schnetter, “Existence and stability of marginally merical relativity,” Phys. Rev. D 54, 4899–4918 (1996), trapped surfaces in black-hole spacetimes,” Phys. Rev. arXiv:gr-qc/9508014. D 99, 064005 (2019). [32] Deirdre M. Shoemaker, Mijan F. Huq, and Richard A. [12] P. M¨osta, L. Andersson, J. Metzger, B. Szil´agyi, Matzner, “Generic Tracking of Multiple Apparent Hori- and J. Winicour, “The Merger of Small and Large zons with Level Flow,” Phys. Rev. D62, 124005 (2000), Black Holes,” Class. Quant. Grav. 32, 235003 (2015), arXiv:gr-qc/0004062. arXiv:1501.05358 [gr-qc]. [33] Lap-Ming Lin and Jerome Novak, “A new spectral [13] Christopher Evans, Deborah Ferguson, Bhavesh apparent horizon finder for 3D numerical relativity,” Khamesra, Pablo Laguna, and Deirdre Shoemaker, Class. Quant. Grav. 24, 2665–2676 (2007), arXiv:gr- “Inside the Final Black Hole: Puncture and Trapped qc/0702038. Surface Dynamics,” (2020), arXiv:2004.11979 [gr-qc]. [34] Jose Luis Jaramillo, Marcus Ansorg, and Nicolas Vas- [14] D. Gannon, “Singularities in nonsimply connected set, “Application of initial data sequences to the study space-times,” J. Math. Phys. 16, 2364 (1975). of black hole dynamical trapping horizons,” Physics [15] D. Gannon, “On the topology of spacelike hypersur- and mathematical of gravitation. Proceedings, Span- faces, singularities and black holes,” Gen. Rel. Grav. ish Relativity Meeting, Salamanca, Spain, September 7, 219 (1976). 15-19, 2008, AIP Conf. Proc. 1122, 308–311 (2009), [16] Daniel Pook-Kolb, Ofek Birnholtz, Badri Krishnan, and arXiv:1103.6180 [gr-qc]. Erik Schnetter, “Interior of a binary black hole merger,” [35] Frank L¨offler,Joshua Faber, Eloisa Bentivegna, Tanja Phys. Rev. Lett. 123, 171102 (2019). Bode, Peter Diener, Roland Haas, Ian Hinder, Bruno C. [17] Ivan Booth, Robie Hennigar, and Saikat Mondal, Mundim, Christian D. Ott, Erik Schnetter, Gabrielle “Marginally outer trapped (open) surfaces and extreme Allen, Manuela Campanelli, and Pablo Laguna, “The mass ratio mergers,” (2020), arXiv:2005.05350 [gr-qc]. Einstein Toolkit: A Community Computational Infras- [18] Sean A. Hayward, “Black holes: New horizons,” in Re- tructure for Relativistic Astrophysics,” Class. Quantum cent developments in theoretical and experimental gen- Grav. 29, 115001 (2012), arXiv:1111.3344 [gr-qc]. eral relativity, gravitation and relativistic field theories. [36] EinsteinToolkit, “Einstein Toolkit: Open software for Proceedings, 9th Marcel Grossmann Meeting, MG’9, relativistic astrophysics,” http://einsteintoolkit. Rome, Italy, July 2-8, 2000. Pts. A-C (2000) pp. 568– org/. 580, arXiv:gr-qc/0008071 [gr-qc]. [37] Marcus Ansorg, Bernd Br¨ugmann,and Wolfgang Tichy, [19] Ivan Booth, “Black hole boundaries,” Can. J. Phys. 83, “A single-domain spectral method for black hole punc- 1073–1099 (2005), arXiv:gr-qc/0508107. ture data,” Phys. Rev. D 70, 064011 (2004), arXiv:gr- [20] Abhay Ashtekar and Badri Krishnan, “Isolated and dy- qc/0404056. namical horizons and their applications,” Living Rev. [38] Marcus Ansorg, “Double-domain spectral method for Rel. 7, 10 (2004), arXiv:gr-qc/0407042. black hole excision data,” Phys. Rev. D72, 024018 [21] Eric Gourgoulhon and Jose Luis Jaramillo, “A 3+1 per- (2005), arXiv:gr-qc/0505059 [gr-qc]. spective on null hypersurfaces and isolated horizons,” [39] J. David Brown, Peter Diener, Olivier Sarbach, Erik Phys. Rept. 423, 159–294 (2006), arXiv:gr-qc/0503113. Schnetter, and Manuel Tiglio, “Turduckening black [22] Matt Visser, “Black holes in general relativity,” holes: an analytical and computational study,” Phys. PoS BHSGRANDSTRINGS2008, 001 (2008), Rev. D 79, 044023 (2009), arXiv:0809.3533 [gr-qc]. 19

[40] Sascha Husa, Ian Hinder, and Christiane Lechner, proach to equilibrium,” Phys. Rev. D97, 084028 (2018), “Kranc: a Mathematica application to generate nu- arXiv:1801.07048 [gr-qc]. merical codes for tensorial evolution equations,” Com- [58] Abhay Ashtekar and Gregory J. Galloway, “Some put. Phys. Commun. 174, 983–1004 (2006), arXiv:gr- uniqueness results for dynamical horizons,” Adv. Theor. qc/0404023. Math. Phys. 9, 1–30 (2005), arXiv:gr-qc/0503109. [41] Kranc, “Kranc: Kranc assembles numerical code,” . [59] Raphael Bousso and Netta Engelhardt, “Proof of a [42] Lars Andersson, Marc Mars, and Walter Simon, “Lo- New Area Law in General Relativity,” Phys. Rev. D92, cal existence of dynamical and trapping horizons,” 044031 (2015), arXiv:1504.07660 [gr-qc]. Phys.Rev.Lett. 95, 111102 (2005), arXiv:gr-qc/0506013 [60] Raphael Bousso and Netta Engelhardt, “New Area Law [gr-qc]. in General Relativity,” Phys. Rev. Lett. 115, 081301 [43] Lars Andersson, Marc Mars, and Walter Si- (2015), arXiv:1504.07627 [hep-th]. mon, “Stability of marginally outer trapped surfaces [61] Fabio Sanches and Sean J. Weinberg, “Refinement of and existence of marginally outer trapped tubes,” the Bousso-Engelhardt Area Law,” Phys. Rev. D 94, Adv.Theor.Math.Phys. 12 (2008), arXiv:0704.2889 [gr- 021502 (2016), arXiv:1604.04919 [hep-th]. qc]. [62] Abhay Ashtekar, Stephen Fairhurst, and Badri Krish- [44] Lars Andersson, Marc Mars, Jan Metzger, and nan, “Isolated horizons: Hamiltonian evolution and the Walter Simon, “The Time evolution of marginally first law,” Phys. Rev. D62, 104025 (2000), arXiv:gr- trapped surfaces,” Class.Quant.Grav. 26, 085018 qc/0005083. (2009), arXiv:0811.4721 [gr-qc]. [63] Badri Krishnan, “The spacetime in the neighborhood [45] Ivan Booth and Stephen Fairhurst, “Isolated, slowly of a general isolated black hole,” Class.Quant.Grav. 29, evolving, and dynamical trapping horizons: geometry 205006 (2012), arXiv:1204.4345 [gr-qc]. and mechanics from surface deformations,” Phys. Rev. [64] Abhay Ashtekar, Christopher Beetle, and Jerzy D75, 084019 (2007), arXiv:gr-qc/0610032. Lewandowski, “Mechanics of Rotating Isolated Hori- [46] Changjun Gao, Xuelei Chen, Valerio Faraoni, and You- zons,” Phys. Rev. D64, 044016 (2001), arXiv:gr- Gen Shen, “Does the mass of a black hole decrease due qc/0103026. to the accretion of phantom energy,” Phys. Rev. D78, [65] Abhay Ashtekar, Christopher Beetle, and Jerzy 024008 (2008), arXiv:0802.1298 [gr-qc]. Lewandowski, “Geometry of Generic Isolated Hori- [47] Jos´eLuis Jaramillo, “A Young-Laplace law for black zons,” Class. Quant. Grav. 19, 1195–1225 (2002), hole horizons,” Phys. Rev. D89, 021502 (2014), arXiv:gr-qc/0111067. arXiv:1309.6593 [gr-qc]. [66] Jerzy Lewandowski, “Spacetimes Admitting Isolated [48] Jos´eLuis Jaramillo, “Black hole horizons and quan- Horizons,” Class. Quant. Grav. 17, L53–L59 (2000), tum charged particles,” Class. Quant. Grav. 32, 132001 arXiv:gr-qc/9907058. (2015), arXiv:1410.0509 [gr-qc]. [67] Sharmila Gunasekaran and Ivan Booth, “Horizons as [49] Jos Luis Jaramillo, “A perspective on Black Hole Hori- boundary conditions in spherical symmetry,” Phys. Rev. zons from the Quantum Charged Particle,” Proceedings, D 100, 064019 (2019), arXiv:1905.02748 [gr-qc]. Spanish Relativity Meeting: Almost 100 years after Ein- [68] Norman G¨urlebeck, “No-hair theorem for Black Holes stein Revolution (ERE 2014): Valencia, Spain, Septem- in Astrophysical Environments,” Phys. Rev. Lett. 114, ber 1-5, 2014, J. Phys. Conf. Ser. 600, 012037 (2015), 151102 (2015), arXiv:1503.03240 [gr-qc]. arXiv:1608.05963 [gr-qc]. [69] Aleˇ s Flandera, “Geometry of isolated horizons,” [50] P.C. Vaidya, “The External Field of a Radiating Star (2016), arXiv:1611.02215 [gr-qc]. in Relativity,” Gen. Rel. Grav. 31, 119–120 (1999). [70] Jerzy Lewandowski and Tomasz Pawlowski, “Neighbor- [51] J.R. Oppenheimer and H. Snyder, “On Continued grav- hoods of isolated horizons and their stationarity,” Class. itational contraction,” Phys. Rev. 56, 455–459 (1939). Quant. Grav. 31, 175012 (2014), arXiv:1404.7836 [gr- [52] Ivan Booth, Lionel Brits, Jose A. Gonzalez, and Chris qc]. Van Den Broeck, “Marginally trapped tubes and dy- [71] Jerzy Lewandowski and Carmen Li, “Spacetime near namical horizons,” Class. Quant. Grav. 23, 413–440 Kerr isolated horizon,” (2018), arXiv:1809.04715 [gr- (2006), arXiv:gr-qc/0506119 [gr-qc]. qc]. [53] Alexis Helou, Ilia Musco, and John C. Miller, “Causal [72] Martin Scholtz, Ales Flandera, and Norman Guer- Nature and Dynamics of Trapping Horizons in Black lebeck, “Kerr-Newman black hole in the formalism of Hole Collapse,” Class. Quant. Grav. 34, 135012 (2017), isolated horizons,” Phys. Rev. D96, 064024 (2017), arXiv:1601.05109 [gr-qc]. arXiv:1708.06383 [gr-qc]. [54] Ayan Chatterjee, Amit Ghosh, and Suresh Jaryal, [73] Robert P. Geroch and J. B. Hartle, “Distorted black “Marginally Trapped Surfaces in Spherical Gravita- holes,” J. Math. Phys. 23, 680 (1982). tional Collapse,” (2020), arXiv:2004.11266 [gr-qc]. [74] Stephen Fairhurst and Badri Krishnan, “Distorted black [55] Ivan Booth, “Spacetime near isolated and dynamical holes with charge,” Int. J. Mod. Phys. D10, 691–710 trapping horizons,” Phys. Rev. D 87, 024008 (2013), (2001), arXiv:gr-qc/0010088. arXiv:1207.6955 [gr-qc]. [75] Terry Pilkington, Alexandre Melanson, Joseph Fitzger- [56] Erik Schnetter, Badri Krishnan, and Florian Beyer, ald, and Ivan Booth, “Trapped and marginally trapped “Introduction to dynamical horizons in numerical rel- surfaces in Weyl-distorted Schwarzschild solutions,” ativity,” Phys. Rev. D74, 024028 (2006), arXiv:gr- Class. Quant. Grav. 28, 125018 (2011), arXiv:1102.0999 qc/0604015. [gr-qc]. [57] Anshu Gupta, Badri Krishnan, Alex Nielsen, and [76] Abhay Ashtekar and Badri Krishnan, “Dynamical hori- Erik Schnetter, “Dynamics of marginally trapped sur- zons: Energy, angular momentum, fluxes and balance faces in a binary black hole merger: Growth and ap- laws,” Phys. Rev. Lett. 89, 261101 (2002), arXiv:gr- 20

qc/0207080. raz Minwalla, and Mukund Rangamani, “Nonlinear [77] Abhay Ashtekar and Badri Krishnan, “Dynamical hori- Fluid Dynamics from Gravity,” JHEP 02, 045 (2008), zons and their properties,” Phys. Rev. D68, 104030 arXiv:0712.2456 [hep-th]. (2003), arXiv:gr-qc/0308033. [97] Ivan Booth, Michal P. Heller, Grzegorz Plewa, and [78] S.A. Hayward, “General laws of black hole dynamics,” Michal Spalinski, “On the apparent horizon in fluid- Phys.Rev. D49, 6467–6474 (1994). gravity duality,” Phys. Rev. D 83, 106005 (2011), [79] Sean A. Hayward, “Spin coefficient form of the new laws arXiv:1102.2885 [hep-th]. of black hole dynamics,” Class. Quant. Grav. 11, 3025– [98] Pau Figueras, Veronika E. Hubeny, Mukund Ranga- 3036 (1994), arXiv:gr-qc/9406033. mani, and Simon F. Ross, “Dynamical black holes [80] Sean A. Hayward, “Energy and entropy conservation for and expanding plasmas,” JHEP 04, 137 (2009), dynamical black holes,” Phys. Rev. D70, 104027 (2004), arXiv:0902.4696 [hep-th]. arXiv:gr-qc/0408008. [99] Li-Ming Cao, “Deformation of Codimension-2 Surface [81] Sean A. Hayward, “Angular momentum conservation and Horizon Thermodynamics,” JHEP 03, 112 (2011), for dynamical black holes,” Phys. Rev. D74, 104013 arXiv:1009.4540 [gr-qc]. (2006), arXiv:gr-qc/0609008. [100] Jose Luis Jaramillo, Rodrigo P. Macedo, Philipp M¨osta, [82] Ivan Booth and Stephen Fairhurst, “The first law for and Luciano Rezzolla, “Black-hole horizons as probes slowly evolving horizons,” Phys. Rev. Lett. 92, 011102 of black-hole dynamics II: geometrical insights,” Phys. (2004), arXiv:gr-qc/0307087. Rev. D85, 084031 (2012), arXiv:1108.0061 [gr-qc]. [83] Vaishak Prasad, Anshu Gupta, Sukanta Bose, Badri Kr- [101] J. L. Jaramillo, R. P. Macedo, P. M¨osta, and ishnan, and Erik Schnetter, “News from horizons in L. Rezzolla, “Towards a cross-correlation approach to binary black hole mergers,” (2020), arXiv:2003.06215 strong-field dynamics in Black Hole spacetimes,” Pro- [gr-qc]. ceedings, Spanish Relativity Meeting : Towards new [84] T Damour, Quelques propri´et´es m´ecaniques, paradigms. (ERE 2011): Madrid, Spain, August 29- ´electromagn´etiques, thermodynamiques et quantiques September 2, 2011, AIP Conf. Proc. 1458, 158–173 des trous noirs, Ph.D. thesis, University of Paris (1979). (2011), arXiv:1205.3902 [gr-qc]. [85] Thibaut Damour, Surface effects in black hole physics [102] Richard H. Price, Gaurav Khanna, and Scott A. (na, 1982). Hughes, “Systematics of black hole binary inspiral [86] R. H. Price and K. S. Thorne, “Membrane Viewpoint on kicks and the slowness approximation,” Phys.Rev. D83, Black Holes: Properties and Evolution of the Stretched 124002 (2011), arXiv:1104.0387 [gr-qc]. Horizon,” Phys. Rev. D33, 915–941 (1986). [103] Gregory J. Galloway, “Maximum principles for null hy- [87] K. S. Thorne, R. H. Price, and D. A. MacDonald, Black persurfaces and null splitting theorems,” Annales Henri holes: The membrane paradigm (Yale University Press, Poincare 1, 543–567 (2000), arXiv:math/9909158. 1986). [104] Aron C. Wall, “The Generalized Second Law implies a [88] Eric Gourgoulhon, “A generalized Damour-Navier- Quantum Singularity Theorem,” Class. Quant. Grav. Stokes equation applied to trapping horizons,” Phys. 30, 165003 (2013), [Erratum: Class.Quant.Grav. 30, Rev. D72, 104007 (2005), arXiv:gr-qc/0508003. 199501 (2013)], arXiv:1010.5513 [gr-qc]. [89] Eric Gourgoulhon and Jose Luis Jaramillo, “Area evo- [105] T. Padmanabhan, “Entropy density of spacetime and lution, bulk viscosity and entropy principles for dynami- the navier-stokes fluid dynamics of null surfaces,” Phys. cal horizons,” Phys. Rev. D74, 087502 (2006), arXiv:gr- Rev. D 83, 044048 (2011). qc/0607050. [106] Ricardo Uribe-Vargas, “On singularities, pere- [90] Eric Gourgoulhon and Jose Luis Jaramillo, “New the- stroikas and differential geometry of space curves,” oretical approaches to black holes,” New Astron. Rev. L’Enseignement Math´ematique. IIe S´erie 50 (2004), 51, 791–798 (2008), arXiv:0803.2944 [astro-ph]. 10.5169/seals-2641. [91] Abhay Ashtekar, Christopher Beetle, and Stephen [107] V.I. Arnold´ and International Mathematical Union, Fairhurst, “Isolated horizons: A generalization of black Contact Geometry and Wave Propagation: Lec- hole mechanics,” Class. Quant. Grav. 16, L1–L7 (1999), tures Given at the University of Oxford Under arXiv:gr-qc/9812065. the Sponsorship of the International Mathematical [92] Abhay Ashtekar, Christopher Beetle, and Stephen Union, Monographie de lEnseignement´ math´ematique Fairhurst, “Mechanics of Isolated Horizons,” Class. (L’Enseignement math´ematique,Universit´ede Gen`eve, Quant. Grav. 17, 253–298 (2000), arXiv:gr-qc/9907068. 1989). [93] Abhay Ashtekar and Alejandro Corichi, “Laws govern- [108] H. Friedrich and J.M. Stewart, “Characteristic initial ing isolated horizons: Inclusion of dilaton couplings,” data and wave front singularities in general relativity,” Class. Quant. Grav. 17, 1317–1332 (2000), arXiv:gr- Proc. Roy. Soc. Lond. A A385, 345–371 (1983). qc/9910068. [109] J¨urgenEhlers and Ezra Newman, “The theory of caus- [94] Jacob D. Bekenstein, “Black holes and entropy,” Phys. tics and wave front singularities with physical appli- Rev. D7, 2333–2346 (1973). cations,” Journal of Mathematical Physics, v.41, 3344- [95] James M. Bardeen, B. Carter, and S. W. Hawking, 3378 (2000) 41 (2000), 10.1063/1.533316. “The Four laws of black hole mechanics,” Commun. [110] V. Arnold, Singularities of Caustics and Wave Fronts, Math. Phys. 31, 161–170 (1973). Mathematics and its Applications (Springer Nether- [96] Sayantani Bhattacharyya, Veronika E Hubeny, Shi- lands, 2013).