arXiv:gr-qc/0604015 v1 4 Apr 2006 ste motn olo o etrwy oetatmore extract to It ways 7]. better 6, for 5, look ringdown 4, to 3, the important 2, to then [1, year is up past right the in inspiral possible the become the have from of starting orbits process, few Sim- merger last years. entire few the last of the ulations in black improved of greatly simulations have Numerical holes horizon. hole of dynamics black the the namely, ingredient regime, important this an understanding fea- discuss for we paper, non-perturbative this new, In tures. qualitatively regime contain This equations. may Einstein non-linearities full the beyond the of go confront dynamics and to to us and require grav- theory will the star, perturbation or a holes of black collapse two itational of coalescence the merger the the understanding of say, phase to, However, due arising waveforms gravitational holes. ade- processes black usually astrophysical are involving many particle, understanding point which a for as quate calculations hole black post-Newtonian the treat with along thereof. perturbations black This, and axisymmet- regarding stationary, solutions, Kerr-Newman the intuition ric studying our from of comes Most holes gen- of relativity. them regime eral about dynamical known non-perturbative, is fully little the relatively in holes, black of ing de/ ¶ § ‡ † ∗ URL: fl[email protected] address: Electronic [email protected] address: Electronic [email protected] address: Electronic URL: nsieo udmna dacsi u understand- our in advances fundamental of spite In http://www.cct.lsu.edu/about/focus/numerical/ http://numrel.aei.mpg.de/ 2 a-lnkIsiu ¨rGaiainpyi,Albert-Ei f¨ur Gravitationsphysik, Max-Planck-Institut h ue feeg rsigtehrzn hc ecie ho describes do which radiation. settles horizon, and/or hole the black crossing energy the of that fluxes verify the to used be can calcul which We p hole (iii) were which situations. momentum, dynamical to angular horizons and isolated mass hole tim black spacelike, the import either of be An tion can which evolution. horizon, the time of their signature study In and situations. surfaces time-dependent marginal to horizons hori isolated dynamical on the within work done is This simulations. merical ASnmes 42.m 47.w 53.f 97.60.Lf, 95.30.Sf, 04.70.Bw, 04.25.Dm, numbers: PACS w collision hole black non-axisymmetric mentum. bl a two and of star, collision neutron head-on axisymmetric the namely, cases, hspprpeet us-oa ehdo tdigtephy the studying of method quasi-local a presents paper This edsrb u ueia mlmnaino hs concepts these of implementation numerical our describe We .INTRODUCTION I. nrdcint yaia oiosi ueia relativit numerical in horizons dynamical to Introduction rkSchnetter, Erik URL: ; 1 oiin tt nvriy ao og,L 00,USA 70803, LA Rouge, Baton University, State Louisiana etrfrCmuainadTcnlg,32Jhso Hall, Johnston 302 Technology, and Computation for Center http://www.aei.mpg. ,2, 1, ∗ ar Krishnan, Badri Dtd pi ,2006) 4, April (Dated: senIsiu,A ¨hebr ,D146Gl,Germany Golm, D-14476 M¨uhlenberg 1, Am nstein-Institut, usiuefreethrzn,d aemn useful many have do not horizons, while event surfaces, for trapped substitute it marginally a above, that mentioned shown horizons theoreti- is quasi-local the in on Similarly, works time. cal real in on surface holes Cauchy black to locate a common to instead surfaces is trapped marginally it use simulations, sur- in- numerical Cauchy the In given a locate face. with to horizon event used the be of tersection can known no that is from condition there hori- quasi-local and, perspective, event experimentally relativity that observed numerical implies space- the be This the never of it. can history locate zons entire to order the in know time to have we cept, without quasi-locally holes horizons. black event study The to reference reviews. to for is 26] 25, aim [24, See basic to 23]. horizons related 22, trapping [21, on closely Hayward work by very earlier the turn, by stationary motivated in and Both isolated are, spacetime. models frameworks dynamical which otherwise these an 20], sig- framework in 19, holes a horizon black 18, are isolated 17, horizons the Dynamical [16, of extension holes. nificant black study to methods. non-gauge-dependent extend analysis emphasise to to want want We space- we and non-vacuum 13]). times, and 12, non-axisymmetric 11, to 10, work 9, this [8, axisym- in e.g. detail (see in metry past the in studied numerically been ingredient. necessary a dy- is understanding holes and black itself, namical in task non-trivial a This be etc. can conditions gauge using systems, coordinate performed different simulations different compare two to from and results simulations from information physical ic h vn oio sagoa,tlooia con- teleological global, a is horizon event the Since 15] [14, horizons dynamical of formalism the use We have horizons event and apparent of dynamics The t h oremlioemmnso h black the of moments multipole source the ate ,† 2, lk,o ul i)W eeaietecalcula- the generalize We (ii) null. or elike, o rmwr,wihetnsteearlier the extends which framework, zon c oe,teaiymti olpeo a of collapse axisymmetric the holes, ack lc oegosa tacee matter accretes it as grows hole black a w of kinds various locate We (i) particular: nt erslto.(v eas study also We (iv) solution. Kerr a to wn n lra Beyer Florian and t o-eoiiilobtlaglrmo- angular orbital initial non-zero ith n nrdeti h aclto fthe of calculation the is ingredient ant n pl hmt he pcfi test specific three to them apply and iso yaia lc oe nnu- in holes black dynamical of sics eiul endfraxisymmetric for defined reviously § ,‡ 2, E-0608 LSU-REL-033006 AEI-2006-018, y ¶ 2 properties and can be used fruitfully to study black hole II. BASIC NOTIONS AND DEFINITIONS physics. Thus, it is likely that ideas and results from dynamical/trapping horizons can be very useful for nu- A. Trapped surfaces and apparent horizons merical relativity. Information obtained from the quasi- local horizons complements the information obtained Let S be a closed, orientable spacelike 2-surface in from the event horizon. Once a simulation is complete a 4-dimensional spacetime ( , gab). The expansion of and ready for post-processing, event horizons are use- any such surface can be definedM invariantly without any ful for studying global properties and the causal struc- reference to a time slicing of the spacetime. Since S is ture of the spacetime, and also phenomena such as the smooth, spacelike, and 2-dimensional, the set of vec- topology change of the horizon during a black hole co- tors orthogonal to it at any point form a 2-dimensional alescence. Reliable and computationally efficient codes Minkowskian vector space. Thus, we can define two lin- are now available for locating event horizons (see e.g. early independent, future-directed, null vectors ℓa and [27]). Such information cannot be obtained at the quasi- na orthogonal to S such that local level, which is instead better for tracking the physi- cal parameters and geometry of a black hole in real time. g ℓanb = 1. (2.1) ab − In particular, we consider the following applications: Note that this convention is different from that used in (i) We study the behavior of various marginally trapped [15]. We shall assume that we know a priori what the surfaces under time evolution. This leads to greater outgoing and ingoing directions on are. By conven- insights about the trapped region of a spacetime. An tion, ℓa will denote an outgoing nullM normal and na an important ingredient here is the signature of the world ingoing one. The null normals are specified only up to a tube of marginally trapped surfaces. This world tube is boost transformation known to be null for isolated horizons, and more gener- a a a 1 a ally, it can be either spacelike or timelike; we show that ℓ f ℓ , n f − n (2.2) → → both types occur frequently in numerical simulations. (ii) We give meaningful definitions for the angular mo- where f is a, positive definite, smooth function on S. All mentum, mass, and higher multipole moments for the physical quantities must be invariant under this gauge dynamical black hole. The multipole moments capture transformation. gauge invariant geometrical information regarding the The Riemannian 2-metric q˜ab on S induced by the horizon geometry, and should be useful for understand- spacetime metric gab is ing fundamental issues such as the final state of black q˜ = g + ℓ n + n ℓ . (2.3) hole collapse. For example, we would expect that after ab ab a b a b a black hole has formed and settled down, its multipole The tensor q˜b can be viewed as a projection operator on moments should be identical to the source multipoles of a to S. The null expansions are a Kerr black hole. We show that it is, in principle, pos- sible to verify this conjecture and to calculate the rate at ab ab Θ ℓ = q˜ ℓ , Θ = q˜ n . (2.4) which a black hole approaches equilibrium. (iii) We also ( ) ∇a b (n) ∇a b describe and implement methods for calculating the en- These expansions tell us how the area element of S ergy flux falling into the horizon. This gives us detailed changes as it is deformed along ℓa and na respectively. information on how black holes grow as they swallow ℓa The shear of , σ(ℓ)ab, is the symmetric trace-free part matter and radiation. of the projection of ℓ : ∇a b

This paper is organized as follows. Section II sets c d 1 σ ℓ = q˜ q˜ ℓ Θ ℓ q˜ . (2.5) up notation, and summarizes the basic definitions and ( )ab a b∇(c d) − 2 ( ) ab properties of trapped surfaces and dynamical horizons. a Section III describes the various physical quantities that Similarly, the shear of n is we calculate using dynamical horizons, and also their 1 c d Θ numerical implementation. Section IV presents three σ(n)ab = q˜aq˜b (cnd) (n)q˜ab . (2.6) concrete, well known numerical examples where these ∇ − 2 concepts are applied and finally, section V discusses Note that these definitions only involve derivatives tan- some open issues and directions for further work. Un- gential to S. Thus ℓa and na can, if necessary, be ex- less mentioned otherwise, we use geometrical units tended arbitrarily away from S while computing these with G = c = 1, the spacetime signature is ( , +, +, +), − quantities. all manifolds and fields are assumed to be smooth, and The closed 2-surface S is said to be a trapped surface the Penrose abstract index notation is used throughout. if both expansions Θ(ℓ) and Θ(n) are strictly negative. The derivative operator compatible with the spacetime This is very different from a sphere in normal flat space metric g is and, following Wald [28], the Riemann ab ∇a which has positive outgoing expansion and negative in- tensor is defined via ( )ω = R dω . going expansion. This definition was first introduced ∇a∇b − ∇b∇a c abc d 3 by Penrose [29], who recognized its importance in the a smooth 3-surface foliated by MOTSs. formation of singularities. On a marginal surface, one The existence of MTTs: Numerically, it has been ob- of the two null expansions vanish. Of particular in- served that marginal surfaces (though not apparent terest are the marginally outer trapped surfaces (MOTSs), horizons – see below) usually behave smoothly under for which the outgoing null rays along ℓa have zero ex- time evolution and produce a smooth MTT. This obser- pansion. In addition, we shall mostly deal with future vation is placed on a more rigorous footing by the re- marginally outer trapped surfaces (FMOTSs), i.e., MOTSs cent result of Andersson et al. [33], which proves the lo- with Θ < 0. (n) cal existence of MTTs for a large class of MOTSs. Their There are three main reasons why closed trapped sur- results require the MOTS to be strictly-stably-outermost. faces are important for studying black holes. First, the An MOTS S on Σ is said to be strictly-stably-outermost existence of a trapped surface implies the existence of if there exists an infinitesimal first order outward defor- a singularity in the future [29, 30]. Secondly, they are mation which makes S strictly untrapped. Working with guaranteed to always lie within the event horizon. Fi- a radial coordinate r on Σ such that S is a level set of nally, in stationary spacetimes, the null generators of the r, and r increases in the outward direction, a sufficient event horizon have zero expansion. Thus for stationary (but not necessary) condition for S to be strictly-stably- spacetimes, the cross-section of the event horizon is a 1 ∂ Θ ℓ (r) > S MOTS. outermost is r ( ) 0 everywhere on . Here it is understood that we obtain Θ ℓ as a function of r by While trapped and marginally outer trapped surfaces ( ) calculating Θ ℓ for the constant-r surfaces in the vicin- are defined in the full four dimensional spacetime, in ( ) ity of S. In principle, for an unfortunate choice of r, it numerical relativity, one usually considers trapped sur- might happen that ∂ Θ ℓ < 0 even though there is a faces in conjunction with a foliation of (partial) Cauchy r ( ) different choice for which this condition is satisfied. In surfaces containing S; it is numerically much easier to any case, this is sufficient for verifying that S is strictly- look for closed surfaces on the Cauchy surface rather 2 than in the full spacetime manifold. For concreteness, stably-outermost. This condition, unlike the outermost we shall work in the ADM formalism where the rele- condition for an AH, is a quasi-local condition. We have vant portion of spacetime is foliated by spacelike sur- found in our simulations that most physically interest- faces, and Σ shall denote one of the leaves of this folia- ing MOTSs, such as ones which asymptote to the event tion. However, it will be obvious that the formalism is horizon, and also AHs, satisfy this condition quite gen- applicable no matter how Einstein’s equations are im- erally. However, as we shall see, there exist also MOTSs plemented. which are not strictly-stably-outermost. In practice, in- stead of checking ∂ Θ ℓ > 0 directly, we look for a sur- The trapped region Σ on Σ is defined to be the set of r ( ) points in Σ through whichT there passes a trapped sur- face with a small positive (or negative) non-vanishing face contained entirely in Σ. Note that there could be expansion, and check that it lies completely outside (or inside) the MOTS. points in Σ not contained in Σ, but through which there T passes a trapped surface not contained in Σ. Thus, Σ is It is shown in [33] that if a MOTS S is strictly-stably- T a subset of the intersection of Σ with the 4-dimensional outermost, then at least locally in time, S is a cross- trapped region in the full spacetime. A connected com- section of a smooth MTT. More explicitly, this result ponent of the boundary of Σ is called an apparent hori- shows that given a foliation of the spacetime by Cauchy T zon (AH). Under suitable regularity conditions, the AH surfaces Σt, if there is a MOTS S0 on Σ0 which is strictly- can be shown to be a MOTS [31, 32]. Thus, an appar- stably-outermost, then MOTSs S exist on Σ for ǫ < t t − ent horizon is the outermost MOTS on Σ. Due to this t < ǫ (for sufficiently small ǫ) such that the union St “outermost” property, an AH is not a quasi-local object is a smooth MTT. The MTT will exist for at least as longS on Σ. The behavior of AHs under time evolution can be as the MOTS remains strictly-stably-outermost. This is quite irregular. For example, they can “jump” discontin- a conceptually important result for numerical relativity uously. On the other hand, as we shall soon see, MOTSs because it shows that a large class of MOTSs behave reg- are more regular. ularly under time evolution. How is this to be reconciled with the known fact that AHs can “jump” during a time evolution? The reason is simply because of the outer- B. Dynamical horizons most property. It is possible that a new MOTS can ap- pear on the outside of a given MOTS. The “old” MOTS is 1. Definition and examples then no longer the globally outermost one even though

We can use marginal surfaces to extract physically in- teresting information about the black hole. The key idea 1 More precisely, ∂ Θ ℓ (r) 0 with ∂ Θ ℓ (r) > 0 somewhere on S. is to look not at a single MOTS by itself, but rather a r ( ) ≥ r ( ) 2 It is harder to show that a MOTS is not strictly-stably-outermost. world tube H of MOTSs constructed by stacking up the This can be done by calculating the signature of the horizon (see MOTSs obtained by time evolution. Such a world tube is below) or by calculating the principle eigenvalue of the stability op- called a Marginally Trapped Tube (MTT). An MTT is thus erator defined in [33]. 4 it is locally outermost, and it continues to evolve in a for this spacetime is shown in figure 1. The 4-metric is perfectly regular manner, but it is no longer an AH. There are, as yet, no similar existence proofs for 2M(v) g = 1 ∂ v∂ v + 2∂ v∂ r MOTSs which are not strictly-stably-outermost. How- ab −  − r  a b (a b) ever, as we shall see later, we find in all the examples + r2(∂ θ∂ θ + sin2 θ∂ φ∂ φ) , (2.7) we have looked at, that MOTSs evolve smoothly even a b a b in this case, forming a regular world tube. and the stress energy tensor is Isolated and dynamical horizons: An MTT is null in equi- librium situations when no matter or radiation is falling ˙ into it; the rest of the spacetime is still allowed to be M(v) Tab = ∂av∂bv (2.8) highly dynamical. This situation is formalized by the 4πr2 notion of an isolated horizon [16, 17, 18, 19, 20]. Us- ing isolated horizons, it has been possible to derive the where M˙ (v) dM/dv. The coordinates (v, r, θ, φ) are ≡ laws of black hole mechanics, use it as a basis for the analogous to the ingoing Eddington-Finkelstein coordi- quantum black hole entropy calculations and find unex- nates for Schwarzschild spacetime. The prescribed mass pected properties of hairy black holes in Einstein-Yang- function M(v) is a positive, non-decreasing function of Mills theory; see [24] and references therein. Most im- the retarded time coordinate v; the Schwarzschild space- portantly for our purposes, isolated horizons have also time is recovered when M(v) is a positive constant. Just proved to be useful in numerical relativity. For exam- as in Schwarzschild, the 2-surfaces r = 2M(v) (for con- ple, isolated horizons provide a coordinate invariant stant v) are FMOTSs. Unlike in the Schwarzschild space- method of calculating the angular momentum and mass time where the 3-surface r = 2M is null and coincides of a black hole [34]. They can be used to obtain bound- with the event horizon, in this case the r = 2M(v) sur- ary conditions for constructing quasi-equilibrium initial face is spacelike if M˙ > 0, and it lies strictly inside the data sets [35, 36]. They might have a role in waveform event horizon. While H is the only spherically symmet- extraction [20]. A pedagogical review of isolated hori- ric dynamical horizon in this spacetime, and there are zons from the numerical relativity perspective can be no spherically symmetric MOTSs outside H, it is shown found in [26]. in [44] that H is not the boundary of the trapped re- In this paper, we are more interested in the dynamical gion. There exist many more non-spherically-symmetric MOTSs and MTTS which come arbitrarily close to the regime when the MTT is not null. A spacelike MTT con- sisting of future-marginally trapped surfaces is called event horizon. Thus the event horizon is the most likely candidate for the boundary of the trapped region [45]. a Dynamical Horizon (DH). Thus, a dynamical horizon is a spacelike 3-surface equipped with a given foliation More generally, figure 2(a) depicts a dynamical hori- by FMOTSs. The properties of a dynamical horizon are zon H bounded by two MOTSs S1 and S2. S is a typi- a studied in detail in [14, 15, 37]. The case when the hori- cal member of the foliation. The vector τˆ is the future a zon is very close to being isolated but still evolving dy- directed unit timelike normal to H, rˆ is tangent to H namically has been studied in [38, 39] and its Hamilto- and is the unit outward pointing spacelike normal to the nian treatment is considered in [40]. Note that the lo- cross-sections. A fiducial set of null normals is cal existence of DHs follows from the local existence of < 1 MTTs because if Θ(n) 0 at any given time, it will con- ℓa = (τˆ a + rˆa) , (2.9) tinue to be strictly negative for at least a short duration. √2 We elaborate on the spacelike property below. 1 na = (τˆ a rˆa) . (2.10) A timelike MTT will be called a timelike membrane √2 − (TLM). A TLM cannot be considered to represent the < surface of a black hole since a time-like surface is not As before, Θ(ℓ) = 0 and Θ(n) 0. The area of a cross- a one-way membrane, and both ingoing and outgoing section S will be denoted by AS and its radius by RS := causal curves can pass through it. In some instances, we AS/4π. A radial coordinate on H will be denoted by shall use the term “horizon” loosely to refer to a generic rp; the cross sections of H are the constant r surfaces. The marginal surface or a MTT without any further quali- 3-metric and extrinsic curvature of H will be denoted fiers. The exact meaning should hopefully be clear from respectively by qab and Kab, and q˜ab is the 2-metric on S. the context. An explicit example of a dynamical horizon is pro- Figure 2(b) shows a Cauchy surface Σ intersecting a vided by the Vaidya spacetime which describes the dynamical horizon H. This intersection S will always gravitational collapse of null dust [41, 42] (see [43] for be assumed to be one of the given cross-sections of H. further examples). This example is not meant to be an The unit timelike normal to the horizon is Ta and the astrophysically realistic model of gravitational collapse, unit outward pointing spacelike normal to S within Σ but it nevertheless provides a good illustration of the is Ra. The three metric and extrinsic curvature of Σ are properties of a dynamical horizon. The Penrose diagram denoted by q¯ab and K¯ ab respectively. The fiducial set of 5

i+ S2

+ H I ℓa a E τˆ io a rˆ na H S v = 0 1 0 = r (a) Dynamical horizon v bounded by S1 and S2.

S2 i− Ta FIG. 1: The Vaidya spacetime. The dashed line indicates the S singularity. This diagram is valid for a strictly increasing mass Σ Ra function M(v) which vanishes for v 0, and asymptotes to a finite value for v ∞. Furthermore,≤ in drawing this diagram, → ˙ < M(v) is assumed to satisfy M(0) 1/16 so that the singular S1 point r = 0, v = 0 is locally naked, and for large v, M(v) is such that the singularity is not globally naked. The Penrose (b) A dynamical horizon intersecting a Cauchy diagram may change qualitatively for other mass functions; surface. see [42] for details. The dynamical horizon is the spacelike surface r = 2M(v) denoted by H, and the event horizon is FIG. 2: The top panel shows a dynamical horizon H. S1 and denoted by E. The shaded portion of the spacetime is flat. a S2 are the initial and final FMOTSs, ℓ is the outgoing null normal, na is the ingoing null normal, rˆa is the unit space- like normal to the cross-sections, and τˆa is the unit timelike Σ normal to H. The bottom panel shows a dynamical horizon null normals to S arising naturally from are and a Cauchy surface Σ intersecting in a 2-sphere S. Ta is the Σ a 1 unit timelike normal to and R is the unit space-like outward ℓ¯a = (Ta + Ra) , (2.11) pointing vector normal to S and tangent to Σ. √2 1 n¯ a = (Ta Ra) . (2.12) √ − 2 Second Law: The area of the cross-sections of a DH in- a A boost transformation of the form of equation (2.2) con- creases along rˆ . This is simply a consequence of < nects (ℓa, na) and (ℓ¯a, n¯ a): Θ(ℓ) = 0 and Θ(n) 0: ℓa ℓa a 1 a 1 = f ¯ , n = f − n¯ . (2.13) Darˆ = q˜ab (ℓ n ) a 2 ∇a b − b When the horizon settles down and becomes null, an 1 a a = (Θ ℓ Θ ) > 0. (2.14) infinite boost ( f ∞) is required to go from (ℓ¯ , n¯ ) to 2 ( ) − (n) (ℓa, na). → If we choose a time evolution vector field ta for which t rˆ > 0, then the area of the dynamical · 2. Summary of basic properties horizon will increase in time, and this result can be called the second law for dynamical horizons. An analogous calculation for TLMs shows that the Topology: The cross-sections of a DH can be either area decreases if Θ < 0, and increases if Θ > 0. spherical or toroidal [14, 15, 21, 33]. Toroidal topol- (n) (n) ogy is possible only in exceptional cases when Foliation: Any given spacelike surface cannot have ℓb σ(ℓ)ab, the scalar curvature R˜ of S, ℓΘ(ℓ), Rab , more than one distinct dynamical horizon struc- and ζa (defined in section III) all vanishL on S [15]. ture on it [37]. This means that a DH can have We shall therefore always take the cross-sections to one, and only one foliation by FMOTSs. This im- be spherical. There are no similar results for cross- plies that if a Cauchy surface Σ does not inter- sections of TLMs. However, we use an apparent sect a given DH in one of the preferred cross- horizon tracker which can only locate spherical sections, then the intersection cannot be a MOTS AHs [46] and therefore all observed MOTSs have at all. Thus, different choices of Cauchy surfaces spherical topology. in general lead to different dynamical horizons. 6

Uniqueness: As mentioned above, the dynamical hori- Σ zon depends on the choice of foliation t, i.e., Singularity choosing a different time slicing would lead to a different dynamical horizon (if it exists). There are however some constraints on the location of dy- a H a ℓa a ℓ 2 namical horizons and trapped surfaces as proved n n H1 by Ashtekar and Galloway [37]. For example, they Σ show that given a dynamical horizon H (along with a mild genericity assumption), there cannot be any trapped surfaces (and therefore no DHs) EH contained entirely in the past domain of depen- Surface of star dence of H. See also [44] for further discussion.

III. APPLICATIONS FIG. 3: Portion of Oppenheimer-Snyder collapse spacetime de- scribing spherically symmetric collapse of pressure-less dust. This section discusses some possible applications of Trapped surfaces are formed the point where the surface of the dynamical horizons. These ideas are illustrated using star intersects the event horizon. There are two sets of trapped concrete numerical examples later in Section IV. surfaces. The ones lying on the timelike surface H1 are the in- ner marginally trapped surfaces while the ones lying on the null surface H2, which is part of the event horizon and also a A. The signatureofa MTT an isolated horizon, are the outermost ones. Since n points away from the trapped region, deforming the inner MOTSs on a > H1 along n makes them untrapped. Therefore nΘ(ℓ) 0 for 1. Background L H1.

As discussed above, MTTs have been shown to exist for a large and physically interesting class of MOTSs, by H1 and H2 in figure 3; H2 is an isolated horizon, i.e., and this is borne out in a large number of numeri- it is null, is part of the event horizon E, and has constant cal simulations where MOTSs are located and evolved area. H1 has decreasing area, is timelike, and goes into smoothly. How many of these MTTs are actually dy- the singularity. A Cauchy surface such as Σ would con- namical horizons? In other words, when is a MTT space- tain two FMOTSs. As expected, Θ ℓ < 0 for H , but Ln ( ) 2 like? The first result in this direction was obtained by nΘ(ℓ) > 0 for H1. There are no spherically symmetric Hayward [21] (see also [34]). Using the Raychaudhuri dynamicalL horizons in this dynamical black hole space- equation for ℓa, it can be shown that an MTT is space- time. like if α < 0, null if α = 0 and timelike if α > 0, where The issue of the signature has been considered in [33]. There it is shown that if a MOTS S is strictly stably out- ab ℓaℓb ab ℓaℓb ermost, and if the quantity σ(ℓ)abσ(ℓ) + Rab is non- σ(ℓ)abσ(ℓ) + Rab α . (3.1) zero somewhere on S (and assuming the null energy con- ≡ Θ ℓ Ln ( ) dition), then the MTT containing S is spacelike in a neighborhood of S. This result is stronger than Hay- In writing this expression, it is assumed that ℓa and a ward’s result (Eq. (3.1)) and it shows clearly that the n are extended off H geodetically, so that Θ ℓ is Ln ( ) spacelike case is physically the most interesting because meaningful. The term in the numerator is strictly pos- ab a b σ ℓ σ + R ℓ ℓ will not vanish in a non-stationary itive in the case of dynamical horizons if the matter ( )ab (ℓ) ab fields satisfy, say, the null energy condition. It vanishes situation. It also shows, somewhat surprisingly, that for isolated horizons. The denominator is negative for even if matter or radiation is falling into a black hole the Vaidya spacetime and also for the stationary Kerr- only in the form of say, a single narrow beam from a par- Newman family. This captures the notion that as we ticular direction, the entire MTT is spacelike. One might go inside the black hole, the outgoing null rays become naively have thought that the MTT would be spacelike more and more converging. Assuming that the numer- only on portions where the energy flux is non-zero, and ator of Eq. (3.1) is nowhere vanishing on H, the hypoth- null otherwise. This is not the case because of the elliptic nature of the equations governing the deformations of a esis that H is spacelike is equivalent to nΘ(ℓ) < 0. As shown by Ben-Dov [47], this lastL condition is not MOTS. satisfied for all MTTs; in Oppenheimer-Snyder collapse In figure 3, the inner MOTS on Σ is not strictly sta- [48], there exists a timelike world tube of FMOTSs with bly outermost; an outward deformation takes it into the trapped region. The results of [33] do not place any re- nΘ(ℓ) > 0. This is illustrated in figure 3 showing a por- tionL of the Oppenheimer-Snyder spacetime. There are strictions on the signature of H1. While the outer MOTS ab ℓaℓb two sets of FMOTSs in this spacetime which are denoted is strictly-stably-outermost, σ(ℓ)abσ(ℓ) + Rab vanishes 7

identically on it. Thus, the results of [33] are not appli-




















































































































cable, and neither H nor H are spacelike. This case is






































1 2













































































S a non-generic in the sense that the outermost FMOTS be- S






































out in r





































































































comes isolated as soon as it is formed, and it becomes a















































































































































































cross-section of an isolated horizon. This happens due

















































S






































(1)







































































































































to the discontinuity of the matter fields at the surface of


















































































































































the star. Choosing a smooth density profile in the initial



































































































S

















































(2) data will smoothen the transition between H1 and H2.










































































































































H will be initially spacelike and eventually settle down






































2













































































to an isolated horizon. Such examples in spherical sym-




















































































































metry are studied in [43].



























































































































































In all the examples we present later, just as in the






































Oppenheimer-Snyder case, it turns out that MOTSs






































form in pairs, i.e., just after a MOTS appears initially, it bifurcates into an “outer” and and “inner” MTT. The FIG. 4: Two MOTSs S(1) and S(2) surrounded by a common outer MTT is spacelike and is a DH. Furthermore, un- MOTS Sout. Spheres lying just inside these FMOTSs must have like the Oppenheimer-Snyder case which has a discon- negative outgoing expansion. Thus, there must be a inner tinuous matter distribution, the transition between the trapped horizon Sin inside Sout which encloses S(1) and S(2). outer and inner MTT is smooth, as far as we can tell nu- merically, in all the cases we have looked at. The inner MTT is, by continuity, initially spacelike. However, it by the sign of the determinant of qab which is gauge in- soon acquires a mixed signature and becomes more and dependent; note that the determinant is itself gauge de- more timelike, and ends up as a TLM. The MOTSs on a pendent. To calculate qab we find a frame e(i) (i = 1,2,3) the inner MTT are not strictly-stably-outermost and thus on H, i.e., three smooth vector fields on H which are this MTT is not required to be strictly spacelike accord- pointwise linearly independent. We then simply need ing to the results of [33]. We strongly suspect that such to compute the determinant of the matrix a bifurcation is a general phenomenon whenever a new MOTS is formed. a a q(i)(j) := gabe(i)e(j) . (3.2) There is one case where the existence of the inner MTT is easy to motivate. Figure 4 shows two MOTSs S(1),(2) We construct a frame on H as follows. Let (t, xi) surrounded by a common MOTS Sout; Θ(ℓ) vanishes on (i = 1, 2, 3) be the spacetime coordinates on used in M all these surfaces. Let us assume that S(1), S(2), and the numerical simulation. The MTT H is topologically Sout are all strictly-stably-outermost and that deforming I S2 (I some interval in R) so that we can assume co- S(1) and S(2) outward yields strictly untrapped surfaces ordinates× (r, θ, φ) on it. Here (θ, φ) are standard coor- 2 S(′1) and S(′2). Similarly, suppose that deforming Sout in- dinates on S and r is a radial coordinate. We can use wards gives a strictly trapped surface Sout′ . Then, since the time coordinate t as the radial coordinate r on H by Θ (ℓ) must change sign somewhere between Sout′ and S(′1) considering H to be embedded into the spacetime by means of the map M or S(′2), it is plausible that there is a MOTS Sin in the intermediate region inside S and outside S and S . out 1 2 F(r, θ, φ) = (t = r, xi = Fi(r, θ, φ)) . (3.3) This argument is supported by a recent result by Schoen [49] which shows the existence of a MOTS between a The maps Fi are known as soon as the MOTSs are found trapped (in our case S ) and an untrapped surface (in out′ by the AH tracker. As a frame on H we choose our case S(′1) S(′2)). It might be possible to extend this proof to rigorouslyS prove the existence of Sin in our case, 1 e(1) = ∂θ, e(2) = ∂φ, e(3) = ∂r. (3.4) and to check whether it is topologically a sphere. S(1), sin θ S(2), and Sout are cross sections of a dynamical horizon Hence, e connects a point on a MOTS at a certain in- while Sin is a cross-section of an MTT, not necessarily a (3) dynamical horizon. stant of time with a corresponding point on the MOTS at the next instant of time. Note that this choice of frame breaks down at the poles of the sphere. To apply for- mula (3.2), the frame (3.4) on H must be pushed forward 2. Numerical implementation to by means of the embedding F in the standard way: M From a numerical standpoint, it is more convenient to 1 2 3 deduce the signature of H by directly calculating the in- e(3) = (1, ∂rF , ∂r F , ∂r F ). (3.5) duced metric qab, rather than from Eq. (3.1) by calculat- a a ing nΘ(ℓ) which requires extensions of ℓ and n away This enables us to calculate q(i)(j) using the 3-metric on fromL the horizon. The signature of H is then determined the Cauchy surface, and the lapse and shift. 8

Having calculated the matrix q(i)(j) and assuming by a boost transformation, i.e., there must exist a posi- its determinant to be positive, we can easily calculate tive function f on S such that the unit vector rˆa. It is simply the outward pointing a a a 1 a unit spacelike vector which is a linear combination of ℓ = f ℓ¯ and n = f − n¯ . (3.8) (e(1), e(2), e(3)), and is orthogonal to e(1) and e(2). This construction of rˆa will also work in the timelike case, but After some simple algebra, the integrand of equation not in the null case where q(i)(j) becomes degenerate. (3.6) can be written as: ϕarˆbK = ϕaRbK¯ + ln f . (3.9) ab ab Lϕ B. Angular momentum and mass Therefore, the angular momentum is

Let ϕa be a rotational vector field on H tangent to (ϕ) 1 ¯ a b 2 2 JS = KabR ϕ d V + ϕ ln f d V . each cross-section.3 The angular momentum of a cross- − 8π IS IS L  section S associated with ϕa is given by The second integral vanishes precisely when ϕa is di- vergence free, i.e., when ϕa is a symmetry of the area (ϕ) 1 a b 2 JS = Kab ϕ rˆ d V . (3.6) element on S. In this case: − 8π IS 1 (ϕ) = ¯ a b 2 We refer to [15] for a justification for this formula. The JS KabR ϕ d V . (3.10) − 8π IS (ϕ) interpretation of J as angular momentum is most S a clear cut when ϕa is a rotational symmetry on H, i.e., In particular, this will be true when ϕ is a symmetry of the metric q˜ , but the divergence free condition is when ϕKab = 0 and ϕqab = 0. See [34] for a method ab of findingL Killing vectorsL suitable for numerical imple- much weaker than this. For example, following [24], mentation. Booth and Fairhurst have shown that this we can always construct a divergence free vector field formula also arises from a Hamiltonian calculation [40]. on a 2-sphere even in the absence of axisymmetry as (ϕ) follows. Let h be any smooth function on S, and g an- As we shall see below, JS is also gauge invariant when ab ϕa is only divergence free, and not necessarily a sym- other smooth function satisfying ǫ˜ ∂ah∂b g = 0, where (ϕ) ǫ˜ab is the volume form on S. It is easy to check explic- metry vector. However, JS is not meaningful for more a itly that the following vector field is automatically di- general ϕ . This is not a real restriction because, while vergence free: every 2-sphere metric does not have a symmetry vector, every 2-sphere metric always admits a divergence free ϕ˜a = gǫ˜ab∂ h . (3.11) vector. b If a cross-section S has radius RS and angular momen- The integral curves of ϕ˜ a are the level curves of h. Inpar- (ϕ) ticular, if h is chosen to be a geometric quantity such as, tum JS , we can meaningfully talk about the mass: say, the curvature R˜ , and g chosen such that ϕ˜ a has affine length 2π, then ϕ˜a will coincide with an axial Killing (ϕ) 1 4 (ϕ) M = R + 4(J )2 . (3.7) vector, if it exists. Therefore, ϕ˜a can be viewed as an er- S 2R q S S S satz axial symmetry vector field even in the absence of This mass has the same dependence on the area and axisymmetry. angular momentum as in the Kerr solution. There is a However, we haven’t as yet satisfactorily imple- meaningful balance law for the mass and furthermore, mented the above construction due to numerical diffi- culties arising from errors in taking derivatives of the it satisfies a physical process version of the first law a [14, 15]. scalar curvature. Furthermore, the ϕ coming from eq. (3.11) may not look like a rotational vector field; in par- Equation (3.6) uses the metric q and K and extrin- ab ab ticular it may vanish at more than just two points on the sic curvature of the dynamical horizon. It is more con- 4 venient to recast this in terms of the metric q¯ and ex- sphere even when S is close to axisymmetry. This is ab work in progress. The results presented below all use trinsic curvature K¯ ab of the partial Cauchy surface Σ (see figure 2(b)). It is convenient to work with the null nor- the method described in [34] of finding Killing vectors mals (ℓ¯a, n¯ a) defined in equation (2.11). It is clear that based on the Killing transport equations. This reduces (ℓ¯a, n¯ a) must be related to the old null normals (ℓa, na) the problem of finding Killing vectors on a sphere to the diagonalization of a 3 3 matrix, and integrating a 1-dimensional ordinary differential× equation. We have

3 This means that ϕa is tangent to S, has closed integral curves, and is normalized so that its integral curves have an affine length of 2π, and it vanishes at exactly two points on S. 4 We thank Ivan Booth for this comment. 9 found this method to be quite reliable for the cases when The moments of these quantities will give the desired c b the horizon is sufficiently close to axisymmetry, even multipole moments. We could also use q˜aKcbrˆ instead in cases when the coordinate system is not adapted to c ¯ b of q˜aKcbR above; the two expressions are related by a the axial symmetry. Thus, it works well for the head- boost transformation. Just as for angular momentum, on collision and axisymmetric neutron star collapse, but the final expressions for the multipole moments given only at very early and late times for a non-axisymmetric below will be boost invariant if the ϕa used in their def- black hole collision. This caveat only affects the exam- inition is divergence free. To define the moments, we ple of section IV B. It is important to keep in mind that need a preferred coordinate system on S so that we can this Killing transport method is not reliable for check- define the preferred spherical harmonics. ing whether the horizon is close to axisymmetry; this re- The construction of the preferred coordinate system quires an independent calculation of q˜ to verify that Lϕ ab (θ, φ) on S is the same as given in [50]: φ [0,2π) is it is sufficiently small. Finally, we emphasize that this the affine parameter along ϕa and ζ := cos θ∈ [ 1,1] is method is also not guaranteed to produce a divergence defined by the condition ∈ − free rotational vector field; this must also be checked in- dependently. ˜ 1 a Daζ = 2 ǫ˜ba ϕ . (3.16) RS C. Multipole moments The freedom to add a constant to ζ is removed by requir- 2 ing its integral over S to vanish: S ζ d V = 0. When ap- The notion of multipole moments play a very impor- plied to a Kerr black hole, theseH invariant coordinates tant role in Newtonian gravity and classical electrody- turn out to be the same as the usual Boyer-Lindquist namics. Let us focus on classical electrodynamics in (θ, φ) coordinates. Minkowski space with axisymmetric charge and current The mass and angular multipole moments are then re- distributions ρ and ja respectively, given on a sphere S of spectively: radius RS. Let (θ, φ) be coordinates on S; ρ and ja, being n RS MS 2 axisymmetric, are functions only of θ. The electric mul- Mn = ˜ Pn(ζ) d V , (3.17) 8π IS R tipoles En and magnetic multipoles Bn are respectively  n+1 defined as RS ab c 2 Jn = ǫ˜ (∂bPn(ζ))KacR d V n 2 − 8π IS n o En = RS ρPn(cos θ)d V , (3.12) I Rn 1 S− ¯ a b 2 ~ = Pn′ (ζ)Kab ϕ R d V (3.18) n+1 ~ ˜ 2 8π IS Bn = RS j ∂Pn(cos θ) ndˆ V , (3.13) − IS  ×  · where Pn′ (ζ)= dPn(ζ)/dζ. We have used equation (3.16) th ˜ where Pn is the n Legendre polynomial, ∂ denotes the to obtain the final expression for Jn above. This form standard derivative operator on a sphere, and nˆ is the clarifies the relation of Jn to the angular momentum and a unit outward normal to the sphere. For black holes, the also demonstrates the gauge invariance of Jn when ϕ is analogs of the electric and magnetic multipole moments divergence free. Using the Gauss-Bonnet theorem, it is are respectively the mass and angular momentum mul- trivial to check that M0 = MS and J1 = JS. J0 vanishes tipole moments. Motivated by this analogy, there ex- because we do not consider any topological defects. Fur- ist meaningful definitions of the source multipole mo- thermore, these expressions are well suited for numeri- ments for an isolated horizon [50]. Roughly speaking, cal computation because they involve only quantities on these definitions correspond to taking the moments of the Cauchy surface and an integral over the MOTS. the free data on an axisymmetric isolated horizon, and knowledge of these moments is sufficient to construct the entire horizon geometry. D. The energy and angular momentum fluxes For dynamical horizons, we can generalize the con- struction of [50] to construct a set of multipole moments Hawking’s area theorem shows that if matter satis- which capture the geometry of a dynamical horizon at fies the null energy condition, then the area of the event any instant of time, and which are furthermore equal to horizon can never decrease. This is one of the central the isolated horizon multipole moments when the black results of black hole physics, and it leads to the classical hole is isolated. The analog of charge density is (propor- picture of the black hole growing inexorably as it swal- tional to) the scalar curvature on S: lows matter and radiation. Therefore, one might expect 1 there to be a balance law relating the increase in area to ρ = M ˜ , (3.14) S 8π SR fluxes of matter and radiation crossing the event hori- zon. However, the teleological nature of event horizons and the angular momentum current is is again a problem; there cannot exist any such local bal- 1 ance law for the area of the event horizon. A clear ex- j = q˜c K¯ Rb . (3.15) a − 8π a cb ample is seen in the Vaidya spacetime where the event 10 horizon is formed in flat space and its area increases in small quantity on the horizon with a very large one. This anticipation of matter falling into the black hole at a later is consistent with the results of [38] where it is found 2 time; see figure 1. that σ ℓ¯ is the most important when the horizon is | ( )| For DHs, it is possible to obtain an exact balance close to equilibrium. law for the area increase [14, 15]; i.e., given two cross- Let t be the time coordinate used to label the Cauchy sections S1 and S2 with radii R1 and R2 respectively, and surfaces. Using this coordinate, we can identify the di- with S2 lying to the outside of S1, the increase in the ra- vergence of various terms appearing in (g). We start F dius is given by the sum of the energy flux due to matter by rewriting (g) as: ( (m)) and gravitational radiation ( (g)), both of which F F F are manifestly positive.: (g) 1 2 2 dR 2 = σ(ℓ) + ζ d Vdt . (3.25) F 8π ZH n| | | | o dt R R 2 − 1 = (m) + (g) , (3.19) The integrand on the right hand side can be expanded 2 F F as where 2 2 σ(ℓ) + ζ R˙ = | | | |  (m) aℓb 2 ˙ 4 2 ˙ 2 2 ˙ 2 = √2Tabτˆ dRd V , (3.20) R f κ¯ + R f ( σ(ℓ¯) ω¯ κ¯)+ R ω¯ . (3.26) F ZH | | | | − · | | Let us look at the various terms in this expression. First, (g) 1 2 2 2 = σ(ℓ) + ζ dRd V . (3.21) ω¯ a F 8π Z | | | | can be shown to be equal to the angular momentum H n o current; for an axial symmetry vector ϕa, the angular a 2 ab 2 a a momentum is simply the integral of ϕ ω¯ a over the cross Here σ(ℓ) := σ(ℓ)abσ ℓ , ζ := ζaζ where ζ is a vector | | ( ) | | section of the MTT. Thus, ω¯ need not vanish even when on S defined as a the MTT becomes an isolated horizon. The ω¯ 2 term in | | a ab c the flux can, in some sense, be viewed as the flux of ro- ζ := √2q˜ rˆ cℓ , (3.22) ∇ b tational energy entering the horizon. Now consider κ¯a. ℓ¯b ℓ¯a ℓ¯a and d2V is the natural geometric volume element on H. For an isolated horizon, b ∝ because in this case ℓ¯a ∇ a The extra factors of 2 and √2 in the above equations as is guaranteed to be geodetic. This implies κ¯ = 0. compared to the corresponding equations in [15], arise On the dynamical horizon side, we can choose suitable ℓ¯a a because of our normalization convention ℓ n = 1; [15] extensions of (and n¯ ) away from the MTT so that · − a uses ℓ n = 2. κ¯ = 0. The shear σ(ℓ¯) on the other hand contains most See [15]· for− additional reasons why (g) has the right of the non-trivial information about the radiation falling properties to be viewed as the flux of gravitationalF radi- into the black hole. It vanishes on an isolated horizon ation. Equation (3.19) is an exact statement about black as it should, and it is independent of any extensions of a a holes in full non-linear general relativity, and it is the ℓ¯ , n¯ away from the MTT. Therefore, in the examples of analog of the Bondi mass balance law at null infinity. section IV, we shall usually plot σ(ℓ¯) to show the energy From a numerical point of view, (g) is inconve- flux falling into the horizon. nient to calculate, especially when the horizonF is settling The angular momentum also obeys a balance law sim- down and is close to being null. First of all, we have di- ilar to equation (3.19): a a rect access only to the fiducial null normals (ℓ¯ , n¯ ) de- ( ) (g) J J = m + (3.27) fined in eq. (2.11) and not to (ℓa, na) themselves. The 2 − 1 Jϕ Jϕ two sets of null normals are related to each other by a where a a 1 a boost transformation ℓ = f ℓ¯ , n = f − n¯ . Under this (m) = T τˆ a ϕbd3V transformation, σℓ = f σℓ¯. Similarly, it is easy to show ϕ ab , (3.28) J − Z∆H that (g) 1 ab 3 ϕ = P ϕqabd V (3.29) ζa = f 2κ¯a ω¯ a , (3.23) J − 16π Z∆H L − where Pab := Kab Kqab. Unlike the energy flux (g), where − F the angular momentum flux (g) is not positive definite. a abℓc ℓ a ab c ℓ (g) Ja κ¯ = q˜ ¯ c ¯b and ω¯ = q˜ n¯ c ¯b . (3.24) Also, vanishes when ϕ is an axial Killing vector ∇ ∇ on H.J Thus, angular momentum is conserved in the ax- Here κ¯a and ω¯ a are tangent to the cross-sections of the isymmetric vacuum case, as it should be. DH. When the DH approaches equilibrium, f ∞. → However, the value of (g) itself remains finite. All fields with a bar remainF finite even when the horizon IV. EXAMPLE NUMERICAL SIMULATIONS becomes null even though f diverges While this is not a problem analytically, this does cause numerical errors In this section, we apply the ideas discussed in the in the transition to equilibrium when we multiply a very previous sections to three concrete numerical simula- 11 tions: i) A head-on collision of two black holes starting assuming that the data is time symmetric, i.e., K¯ ab = 0. with Brill-Lindquist initial data; ii) A non-axisymmetric The number of punctures is equal to the number of black black hole collision using puncture initial data with non- holes. The only equation to be solved is the flat space vanishing linear momentum and iii) Axisymmetric col- Laplace equation for the conformal factor: lapse of a neutron star. Each of these three cases is quite well known in the numerical relativity literature, and all ∆ψ = 0. (4.8) have been well studied. This section aims to further ex- Let x (i = 1, 2, 3) be Cartesian coordinates on Σ such plore these examples using the tools described in Sec- (i) that in these coordinates hab = diag(1,1,1). We consider tion III. ~ the case of two punctures located at the points x(1) and ~ x(2). The solution satisfying the fall-off conditions at in- A. Head-on collision with Brill-Lindquist data finity is 2 α(i) 1. Brill-Lindquist data ψ = 1 + ∑ , (4.9) 2 ~x ~x i=1 | − (i)| Let us first briefly review the Brill-Lindquist initial th where α(i) characterizes the mass of the i black hole data and the conformal method of solving the con- and ~x is its location. We shall denote the distance be- straints. For vacuum general relativity, the initial data (i) tween the two punctures as d = x x . Note that d on a manifold Σ embedded in a four dimensional space- (1) (2) is the distance as measured with| respect− to| the fictitious time consists of the induced Riemannian metric q¯ , and ab flat background metric; the physical distance between the extrinsic curvature K¯ ; as before, barred quantities ab the punctures is actually infinite. It was shown in [53] refer to the physical fields on a Cauchy surface Σ. If D¯ a that each of the punctures is actually an asymptotically is the derivative operator on Σ compatible with q¯ , then ab flat region. As shown in [53], the total ADM mass of the the constraint equations are: common asymptotic region is a D¯ K¯ ab D¯ aK¯ = 0, (4.1) − mADM = 2α(1) + 2α(2) , (4.10) 2 ab R¯ + K¯ K¯ abK¯ = 0. (4.2) − and the ADM masses of the two punctures are The data is assumed to be asymptotically flat so that in 2α( )α( ) the exterior of a compact ball in Σ, i.e., in the asymptotic ADM 1 2 m(1) = 2α(1) + (4.11) region, we have the fall-off conditions d α α ADM 2 (1) (2) m(2) = 2α(2) + . (4.12) 2m 2 d q¯ = 1 + δ + (r˜− ) , (4.3) ab  r˜  ab O (4.13) 2 K¯ = (r˜− ) , (4.4) ab O These are exact results, irrespective of the distance d be- tween the punctures. In the next two sub-sections, we where r˜ is a the radial coordinate in a flat coordinate sys- look at two different regimes (i) the far limit when d is tem in the asymptotic region. In the conformal method large and (ii) the merger of the two holes starting from of solving the initial value constraints [51, 52], we start relatively small values of d. by defining the conformal metric and extrinsic curva- ture

(c) 4 (c) 10 2. Thefarlimit hab = ψ− q¯ab , Kab = ψ K¯ ab . (4.5)

Quantities with the superscript (c)( ) are meant to be Before presenting the results from the numerical evo- conformally rescaled quantities. We· · · shall restrict our- lution of this data, it is instructive to look at a special selves to conformally flat initial data in this article so case which is amenable to analytic treatment, namely, (c) in the far limit where the separation between the holes that hab = δab. In this section we also take the data ¯ is very large: d α , α . In this case, there are to be maximal, i.e., K = 0. In terms of the confor- ≫ (1) (2) mally rescaled quantities, the constraint equations then two MOTSs surrounding each of the punctures with- become out any common MOTS surrounding them. The angu- lar momenta of the two black holes are trivially zero a(c) ∂ Kab = 0, (4.6) because the extrinsic curvature vanishes. What about the mass? Should mADM and mADM be identified with the a 1 (c) (c) ab 7 (1) (2) ∂ ∂ ψ = K K ψ− . (4.7) a − 8 ab masses of the black holes? There are three difficulties with this. First, these ADM masses also include contri- The Brill-Lindquist data consists of taking the manifold butions from radiation present in the respective asymp- Σ to be R3 with n points removed (the punctures) and totic regions. Secondly, if this identification is correct, 12 mADM (i = 1, 2) is supposed to be the mass of the black (i) Horizon shapes at t=1 hole for all values of d, even when the two black holes are very close to each other. Shouldn’t the mass of the 1 individual horizons black holes in this regime also include, say, contribu- inner horizon tions from the tidal distortions produced by the other outer horizon hole? Finally, the strategy of using the asymptotic re- 0.5

gions to define black hole masses is not applicable gen- x erally, say in the case when there are matter fields and 0 the topology of Σ is just R3, or in Misner data [54] where the two black holes do not have their own individual -0.5 asymptotic regions. From the isolated/dynamical horizon perspective, -1.5 -1 -0.5 0 0.5 1 1.5 since the black holes have zero angular momentum, from equation (3.10), the irreducible mass is the correct z measure of mass in this case: m(i) = a(i)/16π where FIG. 5: Coordinate shapes of the horizons at t = 1 in the xz q a(i) is the area of the MOTS around each of the punc- plane. A common horizon has formed, and the inner and outer tures. Let us then calculate the mass of the black holes as common horizons have already separated. Compare figure 4. a power series in 1/d. To simplify calculations, put the origin of coordinates at the location of the first puncture and the other puncture on the z-axis at (0,0, d). Intro- zons are isolated even for relatively small values of d duce the usual spherical coordinates (r, θ, φ) so that the once the common MOTS has formed. conformal factor becomes explicitly

1 2 2 α(1) α(2) 2d cos θ d − 3. Numerical results for the merger phase φ(r, θ)= 1 + + 1 + . r r  − r r2  (4.14) We performed a numerical evolution starting with We see that due to axisymmetry, there is no dependence Brill-Lindquist initial data. Working in units where on φ. Let the surface of the FMOTS around the origin be the total ADM mass is unity, the punctures were lo- given by the equation r = h(θ). In the limit when d cated at z = 0.5, and the individual black holes had ∞ → ± , the initial data reduces to Schwarzschild in isotropic equal masses. Thus 2α(1) = 2α(2) = 0.5. The domain coordinates so that the horizon is located at r = α(1). had an explicit octant symmetry and extended up to Higher order effects can also be explicitly calculated. x, y, z = 96. Near the outer boundary the spatial res- 3 It turns out [55] that up to (d− ), the location of the olution was h = 1.6, and near the punctures we used MOTS is given by O mesh refinement to increase the resolution successively up to h = 0.0125, so that the individual horizon diam- α α α α (1) (2) (1) (2) eters contained initially 32 grid points. We used fourth r = α(1) + (α(2) α(1) cos θ) − d d − order accurate spatial differencing operators, and a third α α (1) (2) 2 order Runge–Kutta time integrator. α(2) 3α(1)α(2) cos θ − 3  − We excised [56] coordinate spheres with a radius of 5 2 4 re = 0.0625 about the punctures from the domain, cor- + α P (cos θ) + (d− ) (4.15) 7 (1) 2  O responding to a diameter of 10 grid points. We used the AEI BSSN formulation [56, 57] for time evolution, where P2 is the second Legendre polynomial. Using this using the boundary conditions also described in [56]. result, the horizon mass m(i) = a(i)/16π can be calcu- These boundary conditions are known to be incompati- q lated and, somewhat surprisingly, the mass is the same ble with the Einstein equations. We used a 1 + log slic- as the ADM mass even up to third order: ing condition [58] starting from α = 1, and a zero shift. This makes both the individual and the outer common horizon grow in coordinate space. We used the Cactus 2α(1)α(2) 4 m = 2α + + (d− ) . (4.16) (1) (1) d O framework [59, 60], the Carpet mesh refinement driver [61, 62], and the CactusEinstein infrastructure. We lo- This relation was verified numerically for a sequence of cated the apparent horizon surfaces with J. Thornburg’s BL data with different values of d. However, we did not AHFinderDirect [46]. have sufficient resolution to estimate the leading order In this setup, the apparent horizon has two discon- ADM deviation between m(1) and m(1) . Similarly, the shear nected components in the initial data, and a common of the horizon vanishes up to third order indicating that MOTS forms shortly after t = 0.5. As discussed in sec- the individual horizons are isolated to an excellent ap- tion III A 1 and figure 4, the common MOTS appears proximation. As we shall see below, the individual hori- as a pair: an outer horizon which is strictly-stably- 13

Irreducible mass Horizon metric determinant at t=0.6 1 7 0.99 inner horizon individual horizon 0.98 outer horizon 6 inner horizon 0.97 5 outer horizon 0.96 4

M 0.95 3 0.94 det q 2 0.93 0.92 1 0.91 0 0.9 -1 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 π π 0 /2 t θ FIG. 7: Irreducible mass vs. time for the individual and the (a) t = 0.6 common MTTs. The outer common MTT grows and accretes mass, while the inner MTT shrinks and loses mass.

Horizon metric determinant at t=1 0.6 the outer MTT tends to become null (as expected), while individual horizon 0.4 the inner MTT becomes completely timelike, and then inner horizon becomes so distorted at about t = 1.2 that it cannot be 0.2 outer horizon reliably tracked any more. This coordinate distortion is 0 already evident in figure 5, and the horizon discretisa- -0.2 tion used in the apparent horizon finder is inaccurate det q -0.4 near the neck of the inner horizon [46]. Figure 7 shows -0.6 the time evolution of the masses M = AS/16π of the -0.8 individual and the common horizonsp (in this case, the ∞ -1 angular momentum vanishes identically). If M is the 0 π/2 π asymptotic value of the mass of the outermost horizon ∞ θ at late times, then MADM M is, in principle, a reliable way of estimating the amount− of energy radiated away (b) t = 1 to infinity in the form of gravitational waves. This dif- ference could be used as a consistency check on other es- timates using the extracted waveforms at large distances FIG. 6: Determinant of the horizon world tube’s three-metric from the black holes. However, our emphasis in this pa- vs. latitude θ at t = 0.6 and t = 1. The individual MTTs are per is on the dynamics of the merger and not on long q˜ = null, i.e., det 0 (up to numerical errors). The common duration stable evolutions. Our simulations do not last outer MTT is spacelike (i.e., det q˜ > 0) and it tends to null ∞ at late times. The inner common MTT is partially timelike at long enough to estimate M reliably. t = 0.6; later it becomes completely timelike. Another feature of the horizons, shown in figure 7, is that while the common outer MTT increases in area as expected, the area of the common inner MTT decreases monotonically. This is explained as follows. Initially, outermost, and an inner one which becomes strictly un- when the common MOTS is just formed, by continu- trapped on being deformed inwards. Figure 5 shows the ity with the outer MTT, the inner MTT is spacelike for a shapes of the individual and the inner and outer com- very short duration (much before t = 0.6) and it is thus mon MOTSs at time t = 1, where the inner and outer a DH for this duration. However, this DH is being tra- common MTTs have already noticeably separated. The versed in the inwards direction (i.e., along rˆa) so that its individual horizons are null up to numerical errors (con- area appears to decrease. Shortly after its− formation, the sistent with the result on the smallness of σ(ℓ) in the far inner MTT becomes partly timelike and later fully time- < limit), and their masses are essentially constant up to like. Recall that for a TLM, the area decreases if Θ(n) 0. numerical error. The common horizons form at the same Thus, both the spacelike and timelike portions of the in- time as a single surface and then split into two MTTs. ner MTT contribute to its monotonic area decrease. This As expected, the outer MTT is purely spacelike while behavior of the outer MTT is roughly similar to what the inner MTT, being spacelike initially, becomes partly was found in [43] for spherically symmetric horizons; timelike quickly. Figure 6 shows the horizon world tube however due to spherical symmetry, the horizons in [43] metric signature at t = 0.6 and t = 1. At later times, did not have any cross sections of mixed signature. 14

Mass quadrupole moment Energy flux at t=0.6 1 0.007 inner horizon outer horizon 0.5 outer horizon 0.006 0 0.005 -0.5 2

0.004 M -1 0.003 E / dA dt

2 -1.5 d 0.002 -2 0.001 -2.5 0 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 π π 0 /2 t θ Mass multipole moment l=4 4 3 inner horizon Shear at t=0.6 2 outer horizon 0.25 1 outer horizon 0 0.2 4 -1 M -2 0.15 -3 2 |

σ -4 | 0.1 -5 -6 0.05 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 t 0 0 π/2 π Mass multipole moment l=6 θ 20 inner horizon 15 outer horizon Total energy flux 10

1 6

outer horizon M 5

0.1 0

-5 dE/dt 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 0.01 t

FIG. 9: Some mass multipole moments vs. time for the inner 0.001 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 and outer MTTs for the head-on collision. The multipole mo- ments for the outer horizon all approach their Schwarzschild t values (i.e., 0) but the inner horizon does not seem to do so.

2 FIG. 8: Energy flux and the shear σ ℓ¯ through the outer com- | ( )| mon horizon vs. latitude θ at t = 0.6, and the total energy flux but it becomes exponentially “more and more null” vs. time. The shear vanishes at the poles and the black hole at late times in the sense that det q becomes smaller settles down exponentially. and smaller, and the flux also becomes correspondingly smaller.

Let us now consider the higher mass multipoles Mn Figure 8 demonstrates how the common outer ap- (all the Jns vanish identically). Here, since all quanti- parent horizon grows. The energy flux vanishes at the ties are symmetric with respect to a reflection about the poles, and the shear (but not the total flux) is maximum equatorial plane, Mn = 0 for odd n. Figure 9 plots the at the equator. The horizon is spacelike all the time, mass quadrupole moment M2 and also M4 and M6 of 15 the outer and inner common MTTs as a function of time. kinds of initial data incorporating this are the so called We expect that the black hole should eventually settle “puncture” data introduced by Brandt and Br ¨ugmann down to a Schwarzschild solution by radiating away all [63], which is a generalization of the Brill-Lindquist con- of its higher multipole moments. Clearly, for the outer struction. MTT, M2, M4 and M6 all become smaller with time, ap- The data is still taken to be conformally flat, but now proaching zero. However, the run did not last long no longer assumed to be time symmetric, i.e., K¯ ab does enough for us to obtain the asymptotic fall-off rate. It not necessarily vanish. We therefore need to solve the is interesting to note that, as far as we can tell, the mul- momentum constraint equation (4.6). For a single black tipole moments for the inner MTT do not vanish asymp- hole, such a solution has been found explicitly by York totically. This tells us that the spacetime near the inner [52]: MTT is not close to Schwarzschild even at late times. ( ) 3 At even later times, all the inner horizons presumably (c)K i = 2P n (g n n )Pcn ab 2r2 (a b) − ab − a b c cease to exist (see next paragraph) and the spacetime ap-  proaches Schwarzschild everywhere. 6 c d + n(aǫb)cdS n . (4.17) We conclude this section with some remarks on the r3 eventual fate of the inner MTT. First of all, as expected, Here, Pa and Sa characterize respectively the linear and the outer MTT eventually settles down and approaches angular momenta of the black hole. Since the momen- future timelike infinity. The inner MTT shrinks and ap- tum constraint equation (4.6) is linear, the solution for proaches the two individual horizons which are essen- multiple black holes is found by a linear superposition: tially stationary. It is interesting to speculate on how, if at all, the inner MTT will merge with the two individual n (c) (c) (i) MTTs. Does the inner MTT “pinch off” into two indi- Kab = ∑ Kab . (4.18) vidual horizons? If the inner MTT is indeed the one pre- i=1 dicted by [49], then it has a priori curvature bounds. If The puncture data consists of substituting this extrinsic these curvature bounds are maintained in the limit, then curvature into the momentum constraint equation (4.7) the inner horizon cannot pinch off. It is more likely that and solving the resulting elliptic equation for the con- the two individual MTTs merge first with each other and formal factor. then later, perhaps also with the inner MTT. It would be interesting to investigate this question further. If the inner MTT does indeed merge smoothly with the 2. Numerical results two individual MTTs, then the set of all MTTs in this case would form one single smooth 3-manifold. Fur- We performed a numerical evolution of puncture ini- thermore, the area of the cross-section of this manifold tial data corresponding to the innermost stable circular would be monotonic in the outward direction – travers- orbit as predicted in [64], which applies the effective po- ing this manifold in the outward direction means going tential techniques of [65]. This model was also studied forward in time on the individual and outer MTTs, and as “QC-0” with the Lazarus perturbative matching tech- backward in time on the inner MTT. nique [66, 67] and later in [2, 3, 5, 6, 68]. In our setup, We are not able to settle these issues numerically in a the punctures were located at x = 1.168642873, and conclusive manner because the inner MTT becomes so their mass parameters were m = 0.453,± and their mo- distorted at late times that the AH tracker is no longer menta were p = 0.3331917498. The domain had an able to track it. This is because the AH tracker can only y explicit rotating quadrant± symmetry and extended up locate star-shaped surfaces and, as is clear from figure to x, y, z = 10. Near the outer boundary the spatial res- 5, the inner MTT will not necessarily be star-shaped at olution was h = 0.4, and near the punctures we used later times. Furthermore, our gauge choice in which we mesh refinement to increase the resolution successively allow the outer MTT to grow in coordinate space, makes up to h = 0.025, so that the individual horizon diam- the inner MTT shrink and therefore harder to resolve at eters contained initially 16 grid points. We used fourth later times. order accurate spatial differencing operators, and a third order Runge–Kutta time integrator. We excised [56] coordinate spheres with a radius of B. Non-axisymmetric black hole collision re = 0.075 about the punctures from the domain, corre- sponding to a diameter of 6 grid points. We used again 1. The initial data the AEI BSSN formulation [56, 57] for time evolution, a 1 + log slicing condition [58] starting from a lapse that is The head-on collision described above does not incor- one at infinity and zero at the punctures, and a Γ driver porate any effects of angular momentum. In this sec- shift condition, starting from a rigid co-rotation with an tion, we remove the restriction of axisymmetry by tak- angular velocity of ω = 0.06. We also used a drift cor- ing initial configurations in which the black holes are recting shift term similar to [69, 70] to keep the individ- orbiting around each other. Perhaps one of the simplest ual horizons centered about their initial locations. 16

Horizon shapes at t=18 Irreducible mass 2 0.9 individual horizon inner horizon 1.5 inner horizon 0.88 outer horizon 1 outer horizon 0.86 0.5 0.84 y

0 M 0.82 -0.5 -1 0.8 -1.5 0.78 -2 0.76 -3 -2 -1 0 1 2 3 17.5 18 18.5 19 x t

FIG. 10: Coordinate shapes of the MOTSs at t = 18 for the FIG. 12: A plot of the irreducible mass Mirr = √A/16π as non-axisymmetric black hole collision. Note that the individ- a function of time for the outer an inner MTTs in the non- ual horizons are locked in place through the co-rotating coor- axisymmetric black hole collision. As expected, the outer MTT dinate system and through an adaptive shift condition. has increasing area while the inner MTT shrinks.

Horizon metric determinant at t=18 zon has formed. As before, there is a pair of common 0.25 MOTSs, an outer and an inner one, surrounding the in- individual horizon 0.2 inner horizon dividual horizons. The inner and outer common MOTSs outer horizon have already separated significantly. The outer MOTS 0.15 is strictly-stably-outermost, and the MTT formed by the 0.1 outer horizons is spacelike, as expected. See figure 11.

det q 0.05 As before, the inner MTT is timelike and not strictly- stably-outermost. The behavior of the MTTs is quali- 0 tatively similar to what was observed in the head-on -0.05 case. Thus, the inner MTT is initially spacelike for a very -0.1 short duration after which it becomes partly timelike 0 π/2 π and, eventually, fully timelike. The MTTS for the two θ individual horizons are null (up to numerical errors). Figure 12 shows the irreducible mass of the outer and FIG. 11: Determinant of the MTT three-metric at t = 18. As inner MTTs as a function of time. Again, the behavior in the head-on case, the outer MTT is purely spacelike while is qualitatively the same as we saw in the head-on col- the inner MTT is partly spacelike and partly timelike. At later lision. The outer MTT increases in area while the inner times, it becomes purely timelike. The individual MTTs are one has decreasing area. null at this time. Figure 13 shows the flux of GW energy falling into the 2 outer horizon at t = 18.4 and also the shear σ ℓ¯ at the | ( )| As previously, we used the Cactus framework [59, 60], same time, for the outer and individual horizons. The 2- 2 d contour plots of the shear σ ℓ¯ and the total flux on the Carpet mesh refinement driver [61, 62], and the | ( )| CactusEinstein infrastructure. We solved the initial the horizon shows in detail how gravitational radiation data equation with M. Ansorg’s TwoPuncture solver is falling into the horizon. Unlike in the head-on case [71], and we located the apparent horizon surfaces with (fig. 8), the shear and the flux are now no longer axisym- J. Thornburg’s AHFinderDirect [46]. metric. Therefore, the flux is no longer constant along This setup contains two initially separated horizons the φ direction but its maxima still lie on the equator. that rotate around each other for a fraction of an orbit The shear on the other hand, now has its maximum on before a common horizon forms [66, 68]. Its ADM mass the poles and its minima lie on the equator. It would be 2 is M = 1.00788, the initial proper horizon separation interesting to further investigate the behavior of σ ℓ¯ ADM | ( )| is L 4.99 MADM, and the horizons have initially the an- and the energy flux as a function of time and for differ- ≈ 2 gular momentum J 0.78 MADM and angular velocity ent physical situations to gain a better understanding of Ω 0.17/M . The≈ common apparent horizon forms how a black hole grows. ≈ ADM at about t = 17.5, which we verified through pretracking Let us now turn to the rotational vector ϕa on the [72]. outer horizon and the quantities such as angular mo- Figure 10 shows the shape of the various MOTSs at mentum, mass, and multipole moments associated with a time t = 18, a short while after the common hori- it. The simulation presented here was run only up to 17

t 19.4, and the final black hole has not settled down Energy flux d2E / dA dt at t=18.4 0.01 sufficiently,≈ and has not attained axisymmetry at this 0 0.0075 0.005 point. Figure 14 shows the Lie derivative of the 2-metric a 0.0025 ϕq˜ab on the horizon at t = 18, where ϕ is the Killing vectorL candidate found by the algorithm presented in [34]. It is clear that q˜ is very far from 0 at this time. π ϕ ab /2 This means that theL angular momentum, mass, and mul- tipole moments associated with this ϕa are not mean- ingful at this point. This is to be expected, since the fi- nal black hole should attain axisymmetry only on a time π scale set by the quasi-normal mode ringdown, which 0 π/2 π 3π/2 2π has a period of 15.9MADM in this case. It is interesting φ to see that our Killing vector field candidate is indeed Killing on the equator. This is by construction, since we choose the Killing vector field candidate by an integral a 2 along the equator; see [34]. However, the vector field ϕ Shear |σ| at t=18.4 0.1 is far from Killing away from the equator. 0 0.075 0.05 A word of caution is due here regarding the Killing 0.025 vector finding algorithm of [34]. First of all, the algo- rithm only produces a candidate for a Killing vector, and an independent check is required to see whether q˜ π ϕ ab /2 is sufficiently small or not. Furthermore, as mentionedL previously, this method reduces the problem of finding a Killing vector on a sphere to diagonalizing a 3 3 matrix followed by integrating a 1-dimensional ODE.× In partic- π ular, the method requires that one of the eigenvalues of π π π π 0 /2 3 /2 2 this matrix is sufficiently close to unity. While this is fine φ when the horizon is exactly axisymmetric, the subtlety 2 arises when the horizon is only approximately axisym- FIG. 13: Energy flux through the horizons and shear σ ¯ on | (ℓ)| metric. It is not clear how close the eigenvalue must the horizon at t = 18.4 in (θ, φ)-coordinates. be to unity for the horizon to be regarded as approxi- mately axisymmetric. Work is in progress to understand this better and to also investigate an alternate method of finding an appropriate ϕa as discussed in section III B, Lie derivative of the two-metric at t=18 which is guaranteed to produce a divergence free vec- 20 tor. |Lθ qab| |qab| 15 C. Axisymmetric gravitational collapse

10 1. The initial configuration

5 Up to now, all of our examples have involved only vacuum spacetimes. In this section, we present an exam- 0 ple of the gravitational collapse of a neutron star to form 0 π/2 π a black hole in an axisymmetric spacetime. These sim- θ ulations were performed using the Whisky code which deals with the matter terms of the Einstein equations in FIG. 14: Lie derivative of the two-metric q˜ at t = 18 on the framework of the Cactus toolkit. Thus, the Whisky Lϕ ab the φ = 0 line. The two-metric q˜ab is also shown for compar- code solves the conservation equations for the stress en- a ison. The quantity shown in the plots are actually the norms ergy tensor Tab and for the matter current density J : 2 2 ∑ab( ϕq˜ab) and ∑ab(q˜ab) in the coordinate system (θ, φ) a a q L Tab = 0, a J = 0. (4.19) on the horizon. Thep vector field ϕ is Killing on the equator (see ∇ ∇ main text), but not everywhere. This shows that the horizon is For details about the Whisky code and the implementa- not (yet) axisymmetric. We expect it to become axisymmetric tion of the above equations, we refer the reader to [73] at later times. Note that we have only shown the plots along and references therein. Here we shall restrict ourselves the φ = 0 curve and we do not have axisymmetry here. to describing the initial stellar configuration which is one of the configurations studied in [73]. 18

The neutron star is modeled as a uniformly rotating Average coordinate radius ball of perfect fluid. The equation of state is taken to be a K = 100, Γ = 2 polytrope so that the pressure p and rest- 3 Γ outer horizon mass density ρ are related according to p = Kρ . The 2.5 inner horizon equilibrium configuration is determined by the mass 2 MNS, central density ρc, and the angular momentum NS

JNS; when necessary, the subscript is used in order to r 1.5 avoid any confusion with previously defined symbols. The model we take is the one denoted as “D4” in [73] 1 15 3 which has MNS = 1.86M , ρc = 1.934 10 g cm− , and 0.5 2 ⊙ × JNS = 0.543MNS. This leads to a ratio of polar to equa- torial coordinate radii of 0.65, a circumferential equa- 0 torial radius of 14.22 km, and a rotational frequency 130 140 150 160 170 180 190 200 of 1295.34Hz. This equilibrium configuration turns out t to be dynamically unstable. In practice, the instability Areal radius is induced by uniformly reducing the pressure slightly 3.6 throughout the star. outer horizon 3.4 inner horizon 3.2 2. Numerical results 3 R 2.8 We simulated the above system on a grid with an ex- plicit rotating octant symmetry. The outer boundary 2.6 was at x, y, z = 150, and the grid spacing near the outer 2.4 boundary was h = 3. We used mesh refinement to in- 2.2 crease our spatial resolution in the center of the domain 130 140 150 160 170 180 190 200 to h = 0.375 at the initial time, and progressively intro- t duced more mesh refinement levels to increase the cen- = tral resolution up to h 0.046875 as the neutron star FIG. 15: The average coordinate radius and the area radius as collapsed, based on the maximum density in the star a function of time for the outer and inner MTTs for the neutron [74, 75]. We also apply third order Kreiss–Oliger dissipa- star collapse. The inner horizon is not to be trusted after t tion [76] to the spacetime (but not the hydrodynamics) 140 due to lack of resolution, since its coordinate radius ha≈s variables. become very small by that time. We find an apparent horizon starting at about t = 130; this time is mainly dependent on the details of how the collapse is induced and has no intrinsic meaning. olution at later times. The horizon is born with an irreducible mass of about Figure 16 shows the determinant of the metric on the Mirr = 1.51 and an angular momentum of J = 0.89 MTTs. The outer MTT is initially spacelike, which is con- (a = 0.38), giving it a total mass of MH = 1.54. Some time after t = 185, a singularity forms in the spacetime, sistent with its growing, and exponentially approaches and the simulation aborts because we do not use exci- null at late times. After about t = 160, the simula- sion inside the apparent horizon. As before, a pair of tion cannot distinguish the horizon world tube signa- MOTSs is formed, an outer and an inner one. The outer ture from null any more. As an example we also show the determinant as a function of the latitude θ at t 138, MTT is spacelike, has increasing area, and tends to null ≈ at late times. In this case, the inner MTT remains space- and the average value of the determinant over the hori- like. However, its area decreases because we are travers- zon as a function of time. The inner MTT is also space- ing it in the inward direction; in other words, the time like and becomes more and more null at least as long as evolution vector ta is such that at the inner MTT, t rˆ < 0 we are able to track it reliably. so that that the area decreases along ta. Our gauge· con- Figure 17 shows the outer horizon has grown at t = ditions are such that the outer horizon grows in coordi- 155 to an irreducible mass of Mirr = 1.80 and an angu- nate space while the inner horizon shrinks. After about lar momentum of J = 1.93 (a = 0.55), giving it a total t = 140, the inner horizon is so small that we do not mass of MH = 1.87. For comparison, the correspond- have enough resolution to track it beyond that time. See ing ADM quantities are MADM = 1.86 and JADM = 1.88 figure 15. The areal radius of the outer MTT increases (a = 0.54). Because the spacetime is axially symmet- but not as rapidly as the coordinate radius; it levels off ric, gravitational waves cannot carry away angular mo- at later times. The area radius of the inner horizon de- mentum. That means that the spin a = J/M2 is ap- creases initially and shows an increase at later times, but proximately correct at late times. Unlike in the non- this is probably just a numerical artefact due to poor res- axisymmetric black hole collision discussed earlier, the 19

Horizon metric determinant Total mass 10 1.9 outer horizon outer horizon 1 inner horizon 1.8 inner horizon 1.7 ADM mass 0.1 1.6

0.01 M 1.5 0.001 1.4 1.3 average of det q 0.0001 1.2 1e-05 1.1 130 140 150 160 170 180 190 200 130 140 150 160 170 180 190 200 t t

Horizon metric determinant at t=138.24 Angular momentum 0.024 2 0.022 outer horizon 0.02 inner horizon 1.5 0.018 0.016 1

0.014 J det q 0.012 0.5 0.01 outer horizon 0.008 0 inner horizon 0.006 ADM J 0.004 -0.5 0 π/2 π 130 140 150 160 170 180 190 200 θ t

FIG. 16: Average of the determinant of the horizon world FIG. 17: The total mass MH,, and angular momentum J as a tube’s three metric vs. time, and vs. latitude θ at t = 138.24 function of time for the outer and inner MTTs for the neutron for the inner and outer horizons for the neutron star collapse. star collapse.

present case is explicitly axisymmetric and there are no black hole physics. Marginally trapped surfaces behave problems with locating the rotational symmetry vector. more regularly that one might have expected previously, Figure 18 shows the mass quadrupole moment M2 and they are useful for extracting interesting physical and the angular momentum octopole moment of the information about the horizon. We have shown how outer and inner MTTs as a function of time. Given that the mass, angular momentum, multipole moments, and we know the asymptotic values of the area and angu- the flux of energy due to in-falling gravitational radia- lar momentum of these MTTs (the ADM values), we can tion of matter can be calculated in a coordinate indepen- also calculate the expected values of M2 and J3 at late dent way (given a particular time slicing of our space- times. The plots clearly show that the values of M2 and time). We have implemented these ideas numerically J3 approach the Kerr values at later times (though this and shown three concrete examples. In these examples, matching is not exact, presumably due to numerical er- we see how the black hole is formed, how it grows, and rors). Also note that M2 is noisy. We have observed such how it settles down to an isolated Kerr black hole. We noise only in simulations that include matter, and we have also seen that the dynamical horizon formalism is find that this noise is much improved by using artificial valuable for exploring the geometry of the trapped re- dissipation on the spacetime variables (which we do). gion. It allows us to classify various types of trapped The angular momentum multipoles seem unaffected. surfaces which might appear during the course of a gravitational collapse or a black hole coalescence. Fi- nally, these ideas can also be viewed as a set of diag- V. DISCUSSION nostic tools which allow us to keep track of what is go- ing on during the course of a numerical simulation, and In this article, we have applied the dynamical hori- whether numerical results make sense and satisfy some zon formalism to numerical simulations of black hole basic, but non-trivial properties in the strong field re- spacetimes. The main theme in this formalism is to gion. take trapped surfaces seriously as a way of describing Some suggestions for future work: 20

simulations. Thus, we have not been conclusively Mass quadrupole moment able to prove that the black hole settles down to 0 Kerr (though there are strong indications that this -0.2 outer horizon M for Kerr does happen). We have not been able to extract the -0.4 2 rate at which equilibrium is reached, thereby ex- -0.6 -0.8 tending Price’s law (see [77] and e.g. [78]) to more 2 -1 general situations, but this is, in principle possible M -1.2 and requires more stable and accurate simulations. -1.4 Similarly, we have not been able to accurately cal- -1.6 culate the asymptotic value of the black hole mass ∞ ∞ -1.8 M . The difference MADM M is, in principle, a -2 reliable estimate of the amount− of energy radiated 130 140 150 160 170 180 190 200 to infinity. While the ADM mass is hard to calcu- t late reliably during the simulation because of the finite grid and low resolution in the asymptotic re- Angular momentum octupole moment gion, it can usually be calculated accurately from 0 the initial data itself. Calculating M∞ and under- outer horizon standing this estimate of the radiated energy re- -2 J3 for Kerr -4 quires more accurate and stable runs, applied to diverse and realistic initial data. The results of [38] -6

3 could also be used to study the approach to equi- J -8 librium. -10 iii. It would be useful numerically to have a gauge con- -12 dition which ensures that the horizon stays at the -14 same coordinate location at all times. While such 130 140 150 160 170 180 190 200 conditions are not difficult to find in the isolated t case, dynamical situations are harder. Given the location of an outer MOTS at a particular instant of time, the results and methods of [33] can be used FIG. 18: Horizon mass quadrupole moment M2 and angular momentum octopole moment J3 vs. time for the neutron star to predict the location of the MOTS at the next in- collapse. For comparison, the values for a Kerr black hole with stant by solving an elliptic equation on the MOTS. the same ADM mass and angular momentum as the initial This could be used to construct appropriate gauge data are also shown. conditions and evolution schemes which take the horizon geometry into account [45, 79]. i. As mentioned in the text, the calculation of the ax- iv. The structure of the horizon can be used to con- ial vector ϕa for non-axisymmetric cases is not yet struct a wave extraction method. These methods satisfactory. We have used the method suggested typically involve starting with a preferred cross- in [34] which works well enough at early and late section of the horizon and flowing it outwards, ei- times, when the horizon is approximately axisym- ther in the past along a null direction as in [20], or metric. However, in general, the result is not guar- outward along the Cauchy surface as in [80]. The anteed to be divergence free and thus the angu- radiation is then encoded essentially in the shear lar momentum not guaranteed to be gauge invari- of the outgoing null normal. a ant. Furthermore, the resulting ϕ in these situ- v. What happens to the inner horizon of figures 4, 5, and ations may be numerically ill-behaved and may 10? As described in section IV A 3, the eventual sometimes have discontinuities. This is particu- fate of these inner MTTs and the two individual larly problematic for calculating the angular mo- horizons is not yet known, and would be inter- mentum fluxes in the non-axisymmetric case. The esting to investigate further. This requires simu- generalization described in section III B resolves lations with higher resolution near the inner hori- many of these issues, but requires greater numer- zons, different gauge conditions, and perhaps also ical resolution at the horizon because it requires AH trackers capable of handling non-star-shaped taking derivatives of the scalar curvature. We have surfaces, and perhaps also higher genus surfaces. not yet implemented this satisfactorily, and this is work in progress. vi. Can the methods of [33] be extended for MOTSs which are not strictly-stably-outermost? In this re- ii. The accuracy of the numerical examples that we have gard, it would be interesting to study the stabil- shown decreases with time, and this is a common ity operator LΣ introduced in [33]. For a strictly- feature of most present day black hole numerical stably-outermost MOTS, the principle eigenvalue 21

of LΣ turns out to be strictly positive and this is Acknowledgments an important ingredient in the existence results. A numerical computation of the eigenvalues of this operator, especially during the transition be- We are grateful to Lars Andersson and Abhay tween inner and outer MTTs and for the inner non- Ashtekar for many valuable suggestions and fruit- spacelike MTTs might lead to further insights. ful discussions. We also thank Ivan Booth, Sergio Dain, Steve Fairhurst, Greg Galloway, Ian Hawke, Sean vii. Do all the trapped surfaces (for a given time-slicing Hayward, Jan Metzger, Denis Pollney, Reinhard Prix, of the spacetime) form a single smooth MTT? This Jonathan Thornburg, and Robert Wald for useful discus- is suggested in [33]. In our examples, this is clearly sions. true of the two branches of the outer horizon. The area decreases for the inner horizon and increases As always, our numerical calculations would have for the outer one. Thus, if we traverse the MTT been impossible without the large number of people outwards, i.e., backwards in time along the inner who made their work available to the public: we used horizon and forward on the outer, then the area the Cactus framework [59, 60] and the CactusEinstein is monotonic. Thus, the irreducible mass of ev- infrastructure [84] with a number of locally developed ery cross-section of this MTT is lesser than the thorns, such as the initial data solver TwoPunctures ADM mass, i.e., each of these trapped surfaces by M. Ansorg, the mesh refinement criteria set up via satisfies the Penrose inequality [81, 82, 83]. Simi- WhiskyCarpetRegrid by C. D. Ott and I. Hawke, and larly, if the scenario described at the end of IV A 3. the horizon finder AHFinderDirect by J. Thornburg. We above is correct, and the inner horizon splits into also used the general relativistic hydrodynamics code two and merges smoothly with the two individ- Whisky [85] developed by the authors of [73], and the ual horizons, then the area is monotonic on this initial data generator RNSID by N. Stergioulas, which whole MTT representing the merger of two black were both developed during the EU training network holes. The Penrose inequality is then valid for “Sources of Gravitational Waves”. The code uses rou- all the MOTSs constituting this MTT. Can this ar- tines of the LAPACK [86, 87] and BLAS [88] libraries from gument be made more precise, and what are the the Netlib Repository [89], the Numerical Recipes [90], limits of its validity? Note that a counterexam- and the UMFPACK [91] library. The numerical simulations ple to the Penrose inequality for apparent horizons were performed on the Peyote Beowulf Cluster at the was found in [47] which relied on having an inner AEI. ES was partly funded by the DFG’s special research horizon with increasing area. This requires a past centre SFB TR/7 “Gravitational Wave Astronomy”. This > marginally trapped surface (Θ(n) 0, Θ(ℓ) = 0) work was supported by the Albert–Einstein–Institut and a discontinuous matter distribution. and the Center for Computation & Technology at LSU.

[1] F. Pretorius, Evolution of binary black hole spacetimes, Phys. [9] P. Anninos, D. Bernstein, S. R. Brandt, D. Hobill, E. Sei- Rev. Lett. 95, 121101 (2005), gr-qc/0507014. del, and L. Smarr, Dynamics of black hole apparent horizons, [2] M. Campanelli, C. O. Lousto, P. Marronetti, and Y. Zlo- Phys. Rev. D 50, 3801 (1994), URL http://link.aps.org/ chower, Accurate evolutions of orbiting black-hole binaries abstract/PRD/v50/p3801. without excision (2005), unpublished, gr-qc/0511048, gr- [10] P. Anninos, D. Bernstein, S. R. Brandt, J. Libson, J. Mass´o, qc/0511048. E. Seidel, L. Smarr, W.-M. Suen, and P. Walker, Dynam- [3] J. G. Baker, J. Centrella, D.-I. Choi, M. Koppitz, and J. van ics of apparent and event horizons, Phys. Rev. Lett. 74, Meter, Gravitational wave extraction from an inspiraling con- 630 (1995), gr-qc/9403011, URL http://link.aps.org/ figuration of merging black holes (2005), gr-qc/0511103. abstract/PRL/v74/p630. [4] P. Diener, F. Herrmann, D. Pollney, E. Schnetter, E. Seidel, [11] S. R. Brandt and E. Seidel, Evolution of distorted rotating R. Takahashi, J. Thornburg, and J. Ventrella, Accurate evo- black holes II: Dynamics and analysis, Phys. Rev. D 52, 870 lution of orbiting binary black holes (2006), gr-qc/0512108. (1995), URL http://link.aps.org/abstract/PRD/v52/ [5] F. Herrmann, D. Shoemaker, and P. Laguna, Unequal-mass p870. binary black hole inspirals (2006), gr-qc/0601026. [12] J. Libson, J. Mass´o, E. Seidel, W.-M. Suen, and P. Walker, [6] M. Campanelli, C. O. Lousto, and Y. Zlochower, The last Event horizons in numerical relativity: Methods and tests, orbit of binary black holes, Phys. Rev. D 73, 061501 (2006), Phys. Rev. D 53, 4335 (1996), URL http://link.aps.org/ gr-qc/0601091. abstract/PRD/v53/p4335. [7] J. G. Baker, J. Centrella, D.-I. Choi, M. Koppitz, and J. van [13] J. Mass´o, E. Seidel, W.-M. Suen, and P. Walker, Event hori- Meter, Binary black hole merger dynamics and waveforms zons in numerical relativity II: Analyzing the horizon, Phys. (2006), unpublished, gr-qc/0602026, gr-qc/0602026. Rev. D 59, 064015 (1999), gr-qc/9804059, URL http:// [8] S. Hughes, I. Keeton, Charles R., P. Walker, K. Walsh, S. L. link.aps.org/abstract/PRD/v59/e064015. Shapiro, and S. A. Teukolsky, Finding black holes in numer- [14] A. Ashtekar and B. Krishnan, Dynamical Horizons: Energy, ical spacetimes, Phys. Rev. D 49, 4004 (1994), URL http:// angular momentum, fluxes, and balance laws, Phys. Rev. Lett. link.aps.org/abstract/PRD/v49/p4004. 89, 261101 (2002), gr-qc/0207080. 22

[15] A. Ashtekar and B. Krishnan, Dynamical horizons and their qc/0407063. properties, Phys. Rev. D 68, 104030 (2003), gr-qc/0308033. [37] A. Ashtekar and G. Galloway, Some uniqueness results for [16] A. Ashtekar, C. Beetle, and S. Fairhurst, Isolated horizons: dynamical horizons, Advances in Theoretical and Mathe- A generalization of black hole mechanics, Class. Quantum matical Physics to appear (2005), gr-qc/0503109. Grav. 16, L1 (1999), gr-qc/9812065. [38] I. Booth and S. Fairhurst, The first law for slowly evolving [17] A. Ashtekar, C. Beetle, and S. Fairhurst, Mechanics of iso- horizons, Phys. Rev. Lett 92, 011102 (2004), gr-qc/0307087. lated horizons, Class. Quantum Grav. 17, 253 (2000), gr- [39] W. Kavanagh and I. Booth, Spacetimes containing slowly qc/9907068. evolving horizons (2006), gr-qc/0603074. [18] A. Ashtekar, C. Beetle, and J. Lewandowski, Mechanics of [40] I. Booth and S. Fairhurst, Horizon energy and angular mo- rotating isolated horizons, Phys. Rev. D 64, 044016 (2001), mentum from a hamiltonian perspective, Class. Quant. Grav. gr-qc/0103026. 22, 4515 (2005), gr-qc/0505049. [19] A. Ashtekar, S. Fairhurst, and B. Krishnan, Isolated hori- [41] P. C. Vaidya, The gravitational field of a radiating star, Proc. zons: Hamiltonian evolution and the first law, Phys. Rev. D Ind. Acad. Sci. A 33, 264 (1951). 62, 104025 (2000), gr-qc/0005083. [42] Y. Kuroda, Naked singularities in the Vaidya spacetime, Prog. [20] A. Ashtekar, C. Beetle, O. Dreyer, S. Fairhurst, B. Krish- Theor. Phys. 72, 63 (1984). nan, J. Lewandowski, and J. Wisniewski, Generic isolated [43] I. Booth, L. Brits, J. A. Gonzalez, and C. V. D. Broeck, horizons and their applications, Phys. Rev. Lett. 85, 3564 Marginally trapped tubes and dynamical horizons, Class. (2000), gr-qc/0006006. Quant. Grav. 23, 413 (2006), gr-qc/0506119. [21] S. A. Hayward, General laws of black hole dynamics, Phys. [44] E. Schnetter and B. Krishnan, Non-symmetric trapped sur- Rev. D 49, 6467 (1994), gr-qc/9306006, URL http://link. faces in the Schwarzschild and Vaidya spacetimes, Phys. Rev. aps.org/abstract/PRD/v49/p6467. D 73, 021502(R) (2006), gr-qc/0511017. [22] S. Hayward, Spin-coefficient form of the new laws of black [45] D. M. Eardley, Black hole boundary conditions and coordinate hole dynamics, Class. Quantum Grav. 11, 3025 (1994), gr- conditions, Phys. Rev. D 57, 2299 (1998). qc/9406033. [46] J. Thornburg, A fast apparent-horizon finder for 3- [23] S. Hayward, Energy and entropy conservation for dynamical dimensional Cartesian grids in numerical relativity, Class. black holes, Phys. Rev. D 70, 104027 (2004), gr-qc/0408008. Quantum Grav. 21, 743 (2004), gr-qc/0306056, URL [24] A. Ashtekar and B. Krishnan, Isolated and dynamical hori- http://stacks.iop.org/0264-9381/21/743. zons and their applications, Living Rev. Rel. 7, 10 (2004), gr- [47] I. Ben-Dov, The penrose inequality and apparent horizons, qc/0407042. Phys. Rev. D 70, 124031 (2004), gr-qc/0408066. [25] I. Booth, Black hole boundaries, Can. J. Phys. 83, 1073 (2005), [48] J. R. Oppenheimer and H. Snyder, On continued gravita- gr-qc/0508107. tional contraction, Phys. Rev. D 56, 455 (1939). [26] E. Gourgoulhon and J. L. Jaramillo, A 3 + 1 perspective [49] R. Schoen (2005), presentation at the Miami waves confer- on null hypersurfaces and isolated horizons, Physics Reports ence. 423, 159 (2006), gr-qc/0503113. [50] A. Ashtekar, J. Engle, T. Pawlowski, and C. Van Den [27] P. Diener, A new general purpose event horizon finder for Broeck, Multipole moments of isolated horizons, Class. Quan- 3D numerical spacetimes, Class. Quantum Grav. 20, 4901 tum Grav. 21, 2549 (2004), gr-qc/0401114. (2003), gr-qc/0305039, URL http://stacks.iop.org/ [51] A. Lichnerowicz, L’int´egration des ´equations de la gravitation 0264-9381/20/4901. relativiste et la probl`eme des n corps, J. Math. Pures et Appl. [28] R. M. Wald, General relativity (The University of Chicago 23, 37 (1944). Press, Chicago, 1984), ISBN 0-226-87032-4 (hardcover), 0- [52] J. W. York, Gravitational degrees of freedom and the initial- 226-87033-2 (paperback). value problem, Phys. Rev. Lett. 26, 1656 (1971). [29] R. Penrose, Gravitational collapse and space-time singulari- [53] D. S. Brilland R. W. Lindquist, Interaction energy in geomet- ties, Phys. Rev. Lett. 14, 57 (1965). rostatics, Phys. Rev. 131, 471 (1963). [30] R. Penrose and S. W. Hawking, The singularities of gravita- [54] C. W. Misner, The method of images in geometrostatics, Ann. tional collapse and cosmology, Proc. Roy. Soc. Lond. A 314, Phys. 24, 102 (1963). 529 (1970). [55] B. Krishnan, Isolated horizons in numerical relativity, [31] S. W. Hawking and G. F. R. Ellis, The large scale structure of Ph.D. thesis, Pennsylvania State University (2002), URL spacetime (Cambridge University Press, Cambridge, Eng- http://etda.libraries.psu.edu/theses/approved/ land, 1973), ISBN 0-521-09906-4. WorldWideIndex/ETD-177/index.html. [32] M. Kriele and S. A. Hayward, Outer trapped surfaces and [56] M. Alcubierre and B. Br ¨ugmann, Simple excision of a black their apparent horizon, J. Math. Phys. 38, 1593 (1997). hole in 3+1 numerical relativity, Phys. Rev. D 63, 104006 [33] L. Andersson, M. Mars, and W. Simon, Local existence of (2001), gr-qc/0008067. dynamical and trapping horizons, Phys. Rev. Lett. 95, 111102 [57] M. Alcubierre, B. Br ¨ugmann, D. Pollney, E. Seidel, and (2005), gr-qc/0506013. R. Takahashi, Black hole excision for dynamic black holes, [34] O. Dreyer, B. Krishnan, D. Shoemaker, and E. Schnet- Phys. Rev. D 64, 061501(R) (2001), gr-qc/0104020. ter, Introduction to Isolated Horizons in Numerical Relativ- [58] M. Alcubierre, B. Br ¨ugmann, P. Diener, M. Koppitz, ity, Phys. Rev. D 67, 024018 (2003), gr-qc/0206008, URL D. Pollney, E. Seidel, and R. Takahashi, Gauge conditions http://link.aps.org/abstract/PRD/v67/e024018. for long-term numerical black hole evolutions without excision, [35] S. Dain, J. L. Jaramillo, and B. Krishnan, On the existence Phys. Rev. D 67, 084023 (2003), gr-qc/0206072. of initial data containing isolated black holes, Phys. Rev. D 71, [59] T. Goodale, G. Allen, G. Lanfermann, J. Mass´o, T. Radke, 064003 (2005), gr-qc/0412061. E. Seidel, and J. Shalf, The Cactus framework and toolkit: [36] J. L. Jaramillo, E. Gourgoulhon, and G. A. Mena Marugan, Design and applications, in Vector and Parallel Processing – Inner boundary conditions for black hole initial data derived VECPAR’2002, 5th International Conference, Lecture Notes from isolated horizons, Phys. Rev. D 70, 124036 (2004), gr- in Computer Science (Springer, Berlin, 2003), URL http:// 23

www.cactuscode.org/Publications/. gr-qc/0503016. [60] Cactus Computational Toolkit home page, URL http:// [75] C. D. Ott, H. Dimmelmeier, I. Hawke, E. Schnetter, www.cactuscode.org/. B. Zink, E. M¨uller, and E. Seidel, Fully consistent 3D general [61] E. Schnetter, S. H. Hawley, and I. Hawke, Evolutions in relativistic rotating stellar core collapse with mesh refinement: 3D numerical relativity using fixed mesh refinement, Class. Comparison to 2D CFC-based approach (2006), in prepara- Quantum Grav. 21, 1465 (2004), gr-qc/0310042. tion. [62] Mesh Refinement with Carpet, URL http://www. [76] H.-O. Kreiss and J. Oliger, Methods for the approximate so- carpetcode.org/. lution of time dependent problems, Global atmospheric re- [63] S. R. Brandt and B. Br ¨ugmann, A simple construction of ini- search programme publications series 10 (1973). tial data for multiple black holes, Phys. Rev. Lett. 78, 3606 [77] R. Price, Nonspherical perturbations of relativistic gravita- (1997), gr-qc/9703066. tional collapse. I. scalar and gravitational perturtbations, Phys. [64] T. W. Baumgarte, Innermost stable circular orbit of binary Rev. D 5, 2419 (1972). black holes, Phys. Rev. D 62, 024018 (2000), gr-qc/0004050. [78] M. Dafermos and I. Rodnianski, A proof of price’s law for the [65] G. B. Cook, Three-dimensional initial data for the collision of collapse of a self-gravitating scalar field, Invent. Math. 162, two black holes II: Quasi-circular orbits for equal-mass black 381 (2005), gr-qc/0309115. holes, Phys. Rev. D 50, 5025 (1994). [79] P. Anninos, G. Daues, J. Mass´o, E. Seidel, and W.-M. Suen, [66] J. Baker, M. Campanelli, C. O. Lousto, and R. Takahashi, Horizon boundary conditions for black hole spacetimes, Phys. Modeling gravitational radiation from coalescing binary black Rev. D 51, 5562 (1995). holes, Phys. Rev. D 65, 124012 (2002), astro-ph/0202469. [80] S. Hayward, Gravitational radiation from dynamical black [67] J. Baker, M. Campanelli, C. O. Lousto, and R. Takahashi, holes, Class. Quantum Grav. 23, L15 (2006), gr- The final plunge of spinning binary black holes (2003), astro- qc/0505080. ph/0305287. [81] R. Penrose, Naked singularities, Ann. N.Y. Acad. Sci. 224, [68] M. Alcubierre, B. Br ¨ugmann, P. Diener, F. S. Guzm´an, 125 (1973). I. Hawke, S. Hawley, F. Herrmann, M. Koppitz, D. Poll- [82] G. Huisken and T. Ilmanen, The inverse mean curvature flow ney, E. Seidel, et al., Dynamical evolution of quasi-circular and the riemannian penrose inequality, J. Diff. Geom. 59, 353 binary black hole data, Phys. Rev. D 72, 044004 (2005), gr- (2001). qc/0411149, URL http://link.aps.org/abstract/PRD/ [83] H. L. Bray, Proof of the riemannian penrose inequality using v72/e044004. the positive mass theorem, J. Diff. Geom. 59, 177 (2001). [69] B. Br ¨ugmann, W. Tichy, and N. Jansen, Numerical simula- [84] CactusEinstein Toolkit home page, URL http://www. tion of orbiting black holes, Phys. Rev. Lett. 92, 211101 (2004), cactuscode.org/Community/numericalRelativity/. gr-qc/0312112. [85] Whisky, EU Network GR Hydrodynamics Code, URL [70] M. Alcubierre, P. Diener, F. S. Guzm´an, S. Hawley, http://www.whiskycode.org/. M. Koppitz, D. Pollney, and E. Seidel, Shift Conditions for [86] E. Anderson, Z. Bai, C. Bischof, S. Blackford, J. Demmel, Orbiting Binaries in Numerical Relativity (2006), in prepara- J. Dongarra, J. Du Croz, A. Greenbaum, S. Hammarling, tion. A. McKenney, et al., LAPACK Users’ Guide (Society for [71] M. Ansorg, B. Br ¨ugmann, and W. Tichy, A single-domain Industrial and Applied Mathematics, Philadelphia, PA, spectral method for black hole puncture data, Phys. Rev. D 70, 1999), 3rd ed., ISBN 0-89871-447-8 (paperback). 064011 (2004), gr-qc/0404056. [87] LAPACK: Linear Algebra Package, URL http://www. [72] E. Schnetter, F. Herrmann, and D. Pollney, Horizon pre- netlib.org/lapack/. tracking, Phys. Rev. D 71, 044033 (2005), gr-qc/0410081. [88] BLAS: Basic Linear Algebra Subroutines, URL http:// [73] L. Baiotti, I. Hawke, P. J. Montero, F. L¨offler, L. Rez- www.netlib.org/blas/. zolla, N. Stergioulas, J. A. Font, and E. Seidel, Three- [89] Netlib Repository, URL http://www.netlib.org/. dimensional relativistic simulations of rotating neutron star [90] W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vet- collapse to a kerr black hole, Phys. Rev. D 71, 024035 (2005), terling, Numerical Recipes (Cambridge University Press, gr-qc/0403029. Cambridge, England, 1986), URL http://www.nr.com/. [74] L. Baiotti, I. Hawke, L. Rezzolla, and E. Schnetter, [91] UMFPACK, URL http://www.cise.ufl.edu/research/ Gravitational-wave emission from rotating gravitational col- sparse/umfpack/. lapse in three dimensions, Phys. Rev. Lett. 94, 131101 (2005),