Arxiv:1007.2990V2 [Gr-Qc]

Home , Trapped surface

arXiv:1007.2990v2 [gr-qc] 13 Aug 2010 ∗ † ‡ § slkl ocnan npicpe nifiienme of are number but infinite intersect that an horizons principle, dynamical in and contain, trapping horizon to trapping suit- likely a given, contains is any that that spacetime a means general, of this ably surface particular why the In trapping of reason definition hole. over main black a of preferred as the horizons choice been dynamical is the often and this have on horizons and on event i.e. coordinate, depends foliation, time spacetime surfaces the of trapped choice of the existence and location [1]. asymptotic isometries the exact on or rely behaviour to needing strong as without the such from region hole, directly gravity black momentum, the angular about and can mass data its horizons extract to dynamical used and be also Trapping thermodynamic their [1–3]. particular properties in horizons, sim- event properties to many ilar have has surfaces these It that shown surfaces. obtained been trapped 3-surfaces marginally viewed of certain time-evolutions be ignoring by can which, of conditions, both technical horizons work horizons Recent dynamical trapping and hole. as such black concepts a related of introduced relativ- has existence numerical the in indicate used to singular- frequently ity Penrose are surfaces. the and in trapped theorem element ity key marginally a are and surfaces These trapped of existence lcrncaddress: Electronic lcrncaddress: Electronic lcrncaddress: Electronic lcrncaddress: Electronic h lcn eedneo o-peial ymti quasi symmetric non-spherically of dependence slicing The oee,i a enkonfraln ieta the that time long a for known been has it However, the is spacetimes hole black of feature important An 2 3 etrfrCmuain&Tcnlg,26Jhso al Lou Hall, Johnston 216 Technology, & Computation for Center eateto hsc srnm,22NcosnHl,Loui Hall, Nicholson 202 Astronomy, & Physics of Department 1 a-lnkIsiu ¨rGaiainpyi,Albert-Ei für Gravitationsphysik, Max-Planck-Institut lc oe efidta h ieec a emn reso mag 04.70.Bw of 04.70.-s, orders 04.25.dg, many numbers: PACS be can holes. difference black the large sufficiently that for find area We horizo quasi-local hole. symmetric black spherically the between area raaedpneto h hieo oito.Hwvr h a the However, foliation. as of that choice find the We on 0 dependent horizon. event are ho and trapping area axisymmetric horizon found trapping numerically symmetric the corresponding for function area ev function the mass and Vaidya accreting sym the both of non-spherically include choices investigate on we horizons o spacetimes the different measures The as compare horizon such we as notions such paper, for quantities implications physical has This quasi-local spacetime. a in . 3%frasol vlighrznwt ˙ with horizon evolving slowly a for 035% ti elkonta us-oa lc oehrzn depend horizons hole black quasi-local that known well is It lxB Nielsen, B. Alex .INTRODUCTION I. [email protected] [email protected] [email protected] [email protected] 1, ∗ ihe Jasiulek, Michael Spacetimes m 1, 0 = † senIsiu,A ¨hebr ,D146Gl,Germany Golm, D-14476 Mühlenberg 1, Am nstein-Institut, ar Krishnan, Badri . 2 eas aclt nltclytedffrnein difference the analytically calculate also We 02. noe ihams n pn n ovstefil equa- field the solves of powers and in spin, particles to tions and point wish as mass holes a theory we black with the Post-Newtonian if endowed treats hand and space. other apart, the parameter on far large very a holes cover black we review). if the recent impossible start a computationally way for becomes the [7] this all e.g. However, phase (see black inspiral coalescence binary the the through a from Nu- evolve starting principle system hole in relativity. from can numerical relativity results merical and combine coa- to theory hole important post-Newtonian black becomes wave binary gravitational it modeling observed in lescence the example, For from parame- physical source signal. the the infer of to ters seeks one when astronomy angular etc. mass, momentum in- as how such linear to momentum, properties and important related horizons their distinct is affects these this it of spacetimes location meaningful the hole extract vestigate black to horizons from quasi-local data use. to use choose to we horizon Thus, these which of on evolution depend time spin will the hori- and quantities quasi-locally, mass different hole say black assign these the to of to wish we all if to Nevertheless, applicable zons. sufficiently are are [6], and energy laws general the balance and for momentum mechanics horizons, angular hole and quasi-local black of for laws results the key example The [5]. distinct 4.Frorproe,w oeta h w frameworks theory two full post-Newtonian the [8– differently; that a very e.g. note holes and see we black paper; large treat purposes, this is our of For topic scope 14]. the this beyond on frame- is two literature discussion the hole The combine black to binary works. essential the solve becomes fully it to coales- problem order the in near down Thus, breaks and cence. the apart when far best are works holes approximation black This holes. black the hsisei oetal eeatfrgaiainlwave gravitational for relevant potentially is issue This saaSaeUiest,BtnRue A783 USA 70803, LA Rouge, Baton University, State isiana in tt nvriy ao og,L 00,USA 70803, LA Rouge, Baton University, State siana n vn oio o lwyaccreting slowly a for horizon event and n asadaglrmmnu.I this In momentum. angular and mass f oclas fahlo hl,w compare we shell, hollow a of collapse to prtn lc oe.Frsm simple some For holes. black aporating ioswt h rao h spherically the of area the with rizons ercsiig fViy spacetimes. Vaidya of slicings metric ufc ftebakhl n loon also and hole black the of surface e aito sntlre forder of large, not is variation rea ntecoc fatm coordinate time a of choice the on 1, xetd ohtelcto and location the both expected, ‡ iuelre hntePlanck the than larger nitude n rkSchnetter Erik and v/c lclhrzn nVaidya in horizons -local with v en h ria eoiyof velocity orbital the being ,3, 2, § 2 treats the black holes as point particles while numerical not consider the case of timelike trapping horizons. Ben- relativity deals with the horizons non-perturbatively. It Dov [21] showed how to find marginally outer trapped turns out that using different flavors of Post-Newtonian surfaces arbitrarily close to the event horizon. It was approximants leads to biases in the values of physical also shown that there are flat regions inside the event parameters [15] (as infered from the gravitational wave- horizon (the region contained within the shell) where no forms). Numerical relativity on the other hand deals with trapped surface passes. Bengtsson and Senovilla [22] an- black hole horizons, and if we decide to use quasi-local alytically constructed trapped surfaces that pass through horizon measures for calculating physical quantities, the the flat region. This is not a contradiction; while there values of the black hole parameters will be affected by the are some flat regions where no trapped surface can be choice of foliation. This effect is absent when the black located, there are trapped surfaces in other parts of the holes are isolated; thus we expect it to be negligible in flat region. In spherically symmetric spacetimes spheri- the inspiral phase and to be larger in the dynamical co- cally symmetric marginally trapped surfaces are easy to alescence phase. In comparing the results from the two find and in the case of Vaidya are given by the condition frameworks, it is useful then to quantify this source of un- r =2m(v), where r is the areal coordinate (r = A/16π certainty and to show that errors in the physical param- with A being the area of the MOTS) of the surfacesp of eters are smaller than the biases between, say, different spherical isometry. It is also easy to show that such sur- post-Newtonian approximants. faces foliate a trapping horizon which we will refer to Another area where the location of horizons is impor- as the spherically symmetric trapping horizon. Due to tant is in black hole thermodynamics. If the area of results in [5], any other dynamical horizon will lie par- a black hole horizon is to play a role analogous to en- tially outside the spherically symmetric dynamical hori- tropy via the Bekenstein-Hawking relation it is impor- zon somewhere. For example, in [23] analytic solutions tant to know which area this relation should be applied were presented where closed trapped surfaces extend into to in which situations. For stationary black holes, cross- the region between the spherically symmetric trapping sections of the event horizon are marginally trapped sur- horizon and the event horizon. The surfaces we find here faces and thus the different notions of black hole hori- do the same. zons (i.e. event horizons and trapping horizons) coincide. In this article we will look at the variation of black However, in a dynamical spacetime the event horizon will hole parameters for horizons located on various different not coincide with the location of any of the trapping hori- slicings of the same spacetime. We will examine several zons. In general they will have different areas on a given different spherically symmetric mass functions; a linear spacelike hypersurface [16, 17]. Furthermore, although mass function designed to see the behaviour for slowing the location of the event horizon does not depend on the evolving black holes and a tanh log mass function, de- choice of spacetime foliation, when the event horizon is signed to see the behaviour in a short collapse of a shell to growing or shrinking, its area will depend on the choice a black hole. We compare marginally outer trapped sur- of foliation. faces found on non-spherically symmetric hypersurfaces Non-stationary spacetimes with dynamical black holes with those found on the spherically symmetric slicings. are complicated, and only a few exact dynamical black We are able to locate the spherically symmetric horizons hole solutions are known. One class of solutions is the analytically but the non-spherically symmetric horizons Vaidya solutions, which describe the evolution of a spher- are located numerically using the AHFinderDirect thorn ically symmetric, radially moving, pressureless null fluid [24, 25] of the Cactus framework [26, 27]. A number with a freely specifiable mass function m(v). This is a of useful analytic relations for the Vaidya spacetime are useful and non-trivial toy-model for a dynamical black given in the appendix. Background detail on the proper- hole. These spacetimes can be used to qualitatively ties of trapping horizons and their thermodynamics can model the physical accretion of matter by a black hole. be found in [1] and [3]. Some authors [18, 19] have even suggested using the Vaidya spacetimes to simulate the decrease in area of a black hole due to Hawking radiation, by allowing the II. THE VAIDYA METRIC infalling matter to be negative. The uniqueness of dynamical horizons in general space- To examine the slicing dependence of quasi-local hori- times was investigated in [5] where it was shown that zons we need to consider dynamical spacetimes. The the foliation of a given dynamical horizon by marginally Vaidya solutions have a number of nice properties that trapped surfaces is unique, and that it is not possible to make them popular for investigations of this type [20–22]. foliate a region with dynamical horizons since they inter- The Vaidya spacetimes [4] are a class of spherically sym- sect. In the Vaidya solution it has already been shown ex- metric, non-vacuum spacetimes with line element in ad- plicitly that marginally outer trapped surfaces (MOTS) vanced null Eddington-Finkelstein coordinates (v,r,θ,φ): can be found at various different locations depending on the slicing [20], although this reference did not investi- 2m(v) ds2 = 1 dv2 +2dvdr+r2dθ2 +r2 sin2 θdφ2. gate how the parameters of the black hole, such as area, − − r mass and angular momentum vary with slicing and did (1) 3

The function m(v) is a a freely specifiable mass func- black holes are typically found with luminosities between tion; it coincides with the Misner-Sharp mass [29] in 10 and 100% of the Eddington limit [30]. For a black hole this case. In advanced Eddington-Finkelstein coordi- accreting at a tenth of the Eddington limit the dimen- nates the metric is well defined across future horizons. sionless accretion rate is approximately [16] The Vaidya metric solves the Einstein equations with an M energy-momentum tensor of the form m˙ 10−22 . (7) ≃ M⊙ Tab = Enanb, (2) This is 1038 ergs per second for a solar mass black where na is an ingoing radial null direction and E is a hole. The∼ matter falling into the black hole is ten times function that depends on the normalization of na. For the energy being emitted as light. Since this accretion is example, for the case where na = ∂av then we have usually associated with a disk though it will not be spher- − ically symmetric. For a black hole accreting purely from a b m˙ Tabℓ ℓ = , (3) the Cosmic Microwave Background, which is assumed to 4πr2 be isotropic, we have from the Stefan-Boltzmann law, a dm(v) approximately a when n ℓ = 1 and definingm ˙ = dv . The Vaidya solution can be− interpreted as describing the radial collapse 4 2 −50 T M of pressureless null dust (i.e. infalling radiation). From m˙ 10 . (8) ≃ T3K M the appendix we can see that, in a spherically symmet- ⊙ ric foliation, trapping horizons will occur at r = 2m(v) For a black hole whose dynamics are dominated by evap- provided we keep m(v) > 0. oration through Hawking radiation we have Here we will study various mass functions for m(v). M 2 The simplest is a linear mass function of the form m˙ 10−81 ⊙ . (9) ≃− M m(v)= m0 +mv. ˙ (4) For numerical purposes we will investigate evolutions The linear mass function is suited to situations where the with mass rates much higher than astrophysical rates, black hole is accreting matter at a constant rate (for a typically m˙ 0.01. To the extent that 2√m˙ 1, finite time duration to ensure that the mass stays finite). the spherically| | ∼ symmetric trapping horizon will still≪ be While the assumption of spherical symmetry is some- a slowly evolving horizon in the sense of [32]. We will what artificial, real astrophysical black holes do have very also look at mass functions of the tanh log form, with small mass accretion rates whether from the accretion of m(v)=0 for v < 0 and for v> 0 surrounding stars and gas or purely from the accretion 2 of photons from the cosmic microwave background. In m0 v m0v m(v)= 1 + tanh log = 2 2 (10) the Vaidya solutions the Misner-Sharp mass on each in- 2 T v + T going constant v surface is a constant. This reflects the The first derivative of this tanh log mass function van- fact that the mass is flowing inwards at the speed of light ishes at v = 0. Therefore the metric and its first deriva- and the mass contained within a shell of radius r is con- tive are contiunous at v = 0. For T and m0 greater than stant as the radius decreases. Although the spacetime is zero, this mass function models the collapse of a hollow dynamical we can still define observers who remain at a spherical shell of matter m(v = 0) = 0 that asymptoti- fixed areal radius r and fixed θ and φ. In terms of the cally settles down to an isolated black hole of mass m0 for proper time τ of such observers we have v . These types of mass functions model situations → ∞ dv 1 where the black hole grows initially quite rapidly and = . (5) then asymptotically settles down to its final static state. dτ 1 2m(v)/r − The mass function reaches half its asymptotic value when p v = T . At large distances from the black hole r m(v) we can The maximum ofm ˙ for these tanh log functions occurs use this to relate directly the mass flux in terms≫ of the null at v = T/√3 and takes a value m /T . Therefore these coordinate v to the mass flux as seen by constant r, θ, φ 0 solutions will not be slowly evolving∼ in the sense of [32] observers who would be static observers in an exactly for m and T of similar sizes. static spacetime, 0 dm(v) dm(v) . (6) A. Location of spherically symmetric horizons dv ∼ dτ

We can give a very rough indication of the order of mag- The trapping horizons are three-dimensional surfaces nitude form ˙ that might be expected for certain astro- H, foliated by closed spacelike two surfaces for which the physical cases. The Eddington rate is used to estimate future directed null normals ℓa and na satisfy the maximal rate at which infalling matter can be sup- ported by its own radiation pressure and large, luminous θ ℓ =0 , θ n < 0 , nθ ℓ < 0 . (11) ( ) ( ) L ( ) 4

a a Here θ(ℓ) and θ(n) are the expansions of ℓ and n respec- black hole at some point settles down to a Schwarzschild a tively, and \ is the Lie derivative along n . Dynami- black hole with no further matter accreting, then the cal horizonsL are also three-dimensional surfaces foliated, event horizon can be located by tracing back the null as above, by spheres with θ(ℓ) = 0, θ(n) < 0. How- rays from the future Schwarzschild radius. However, in ever, nθ(ℓ) < 0 is replaced by the requirement that H the situation where the black hole is accreting matter at be spacelike.L For the spherically symmetric horizons in a steady rate and is a suitably long way from changing Vaidya, these notions coincide [1]. to a different state one can find the approximate location For spherically symmetric slicings, the null normals as- of the event horizon by imposing the condition sociated with the trapping horizon will be radial null vec- d2r tors and the location of the trapping horizon is just given =0. (16) by dv2 This just reflects the fact that the event horizon is grow- r =2m(v) . (12) ing at a steady rate [16]. In this case, equation (15) has This is a spacelike surface form ˙ > 0, a null surface for the general solution m˙ = 0 and a timelike surface form ˙ < 0. For the linear m(v) mass function, in terms of the timelike coordinate t = r = 1 √1 16m ˙ . (17) v r, the horizons will be located at 4m ˙ − − − Form ˙ 1 this gives 2m0 + 2mt ˙ ≪ r = . (13) 1 2m ˙ r =2m(v) 1 + 4m ˙ + (m ˙ 2) (18) − O For the tanh log mass function the horizons are located and thus we expect that the event horizon will be outside at the solution of the cubic the spherically symmetric trapping horizon form> ˙ 0 but inside form ˙ < 0. For a solar-mass black hole accreting r3 + (2t 2m ) r2 + t2 + T 2 4m t r 2m t2 =0. 0 0 0 at a tenth of the Eddington rate the difference in areas − − − (14) between the event horizon and the spherically symmetric The equation for the horizon r = 2m(v) has a single trapping horizon will be around 1056 in units of Planck unique solution on each surface of constant v. But it area, while for a supermassive black hole of mass 108 solar can have multiple solutions, corresponding to multiple masses, accreting purely form the CMB, the difference in horizons, on surfaces of constant t. For m > 0 this 0 areas will be around 1060 in Planck units [16]. In terms cubic function always has at least one positive real root of the coordinate t = v r for the linear mass function and guarantees that there will always be at least one the event horizon has radial− coordinate horizon. In fact, it can be shown that there will only be a single horizon for t > 2m 4m2 T 2. For small 2m + 2mt ˙ 8mm˙ 0 − 0 − r + . (19) values of t there are multiple horizonsp and this will also ∼ 1 2m ˙ 1 2m ˙ be the case for other mass functions. − − This is just the location of the spherically symmetric We will use numerical methods to solve the trapping trapping horizon with a constant offset of 8mm˙ provided horizon equations for non-spherically symmetric slicings m˙ 1. In this approximation the generators of both the and postpone discussion of them until the next section. trapping≪ horizon and the event horizon have the same Due to results in [5] we expect that the non-spherically components but the norm of the generators is 4m ˙ for the symmetric trapping horizons will intersect the spherically trapping horizon and zero for the event horizon. symmetric ones. The event horizons are defined as the past causal boundary of future null infinity and are generated by null B. An axisymmetric spacetime slicing geodesics that fail to reach infinity. The event horizon is always a null surface since it is a causal boundary. In the Vaidya spacetimes it is generated by radial outgoing null We consider a simple axisymmetric slicing of the form vectors that satisfy t¯= v r αz, (20) − − dr 1 2m(v) = 1 . (15) where z = r cos θ and α is a parameter that determines dv 2 − r how far away from spherically symmetry the constant t¯ This first order ordinary differential equation generates surface is. We reserve the symbol t for spherically sym- the path of all outgoing radial null geodesics. In order to metric surfaces where α = 0, which just gives the usual give the location of the event horizon it requires a bound- Eddington-Finkelstein time coordinate (although not the ary condition that corresponds to the known location of Schwarzschild time coordinate). Hypersurfaces of constant t¯ are always timelike for α < 1 since the normal the event horizon at some particular point. In practice ¯a | | this is usually given by the position of the event horizon to a given hypersurface, t , has norm at some future point, either when the black hole evapo- a 2 2m(v) 2 2 t¯ t¯a = α 1 1+ α cos θ + α cos θ . (21) rates entirely or settles down to a stationary state. If the − − r 5

On the slice t¯ = 0 we have v > 0 everywhere for α < spherically symmetric surface with α = 0 but the same 1. In addition, since 1+ α cos θ + α2 cos2 θ is always| | constant value of t we see that the area will be greater if greater than 3/4, for real α this hypersurface will become m>˙ 0 and smaller ifm< ˙ 0. timelike near the horizon for α> √7/2 1.323. On each slice with a constant value of∼t¯, a two dimensional marginally outer trapped surface (MOTS) can be C. Horizon deformations searched for, satisfying just θ(ℓ) = 0. In some spacetimes there may be multiple MOTS on a given hypersurface. The choice of different foliations leads to different This will not occur in the Vaidya spacetimes with linear MOTS that can be deformed into each other. Using mass function since the horizon condition, r = 2m(v), these deformations it is possible to examine how certain is linear in r. The marginally outer trapped surfaces geometric properties of surfaces change with the defor- can typically be stacked to form three dimensional trap- mation [31]. For example, it is possible to see how the ping horizons, provided the other conditions θ(n) < 0 and a area element, ǫ, changes as MOTS are deformed into one n aθ n = 0 are satisfied too (or a dynamical horizon if ∇ ( ) another. the resulting surface is spacelike). We can search for MOTS on slices with different val- For the case α = 0 the slicings will be spherically sym- ues of t¯but because these marginally trapped surfaces do metric. The orbit of spherical rotations of each point will not lie on the same hypersurface it is difficult to compare lie entirely in the hypersurface. This will lead to MOTS them. If they do not intersect (and typically they don’t) that are spherically symmetric, where each surface is just it may or may not be possible to find a trapping horizon the orbit of spherical rotations. Commonly used slicings that connects both surfaces. We want to distinguish be- in numerical relativity such as constant mean curvature tween marginally trapped surfaces that are evolutions of surfaces will typically be spherically symmetric in the one another and marginally trapped surfaces that cannot Vaidya spacetime. In this case the horizons will be lo- be evolved into one another along a trapping horizon. cated uniquely at r = 2m(v) and every point on the For a given choice of t¯but two different values of α the surface will have the same value of the advanced time v. two slices thus defined will intersect when z = 0. The This will not be the case for surfaces with α = 0. In these 6 slice t = 0 with α = 0 does not lie entirely to the past cases the mass parameter m(v) will not necessarily take or future of the slice t¯ = 0 with α = 1. However, the the same value on different points of the surface. The horizons on these two slices will typically not intersect. MOTS will then extend into regions where the matter is On the two dimensional surface where these two slices more compact and regions where it is less compact. Sur- do intersect, z = 0, the horizon for α = 0 is found to lie faces with α> 0 will be the same as surfaces with α< 0, outside the horizon with α = 0. with the north pole (θ = 0) interchanged with the south We can define a vector field,6 X, on surfaces that can pole (θ = π) or, equivalently cos θ cos(π θ). be used to define variations of this surface, δX . Following ¯ → − Each different choice of t will lead to different MOTS [32] we can write this vector field on a marginally trapped than can be stacked to form different trapping horizons. surface in terms of the two null normals to the surface as Each of these trapping horizons will have an associated a a a natural foliation, the one that makes each two dimen- X = Bℓ Cn , (24) sional surface a MOTS. However we can also consider the − intersection of a given trapping horizon with a spacelike where B and C are functions on the surface. For deforma- hypersurface of a given α value different from the natural tions of the spherically symmetric surface where = 0 one. We can compute the area of the these closed two we have △ spheres, although for α = 0 they are not MOTS. For 6 the spherically symmetric trapping horizon (SSTH) in a ∂ ∂ X = B + C (25) Vaidya spacetime with linear mass function andm ˙ 0, ∂t ∂r the area will be ≪ for the Eddington-Finkelstein coordinates t and r. The t¯ 4 variation of intrinsic geometrical scalar properties can be A =4π (2m)2 1 + 4m ˙ + 2m ˙ + mα˙ 2 . (22) SSTH m 3 calculated using

A similar calculation can be performed for the intersec- δX φ = X φ, (26) tions of the event horizon with the constant t¯ surfaces. L Taking the position of the event horizon to be (18) and and for general tensors, by projecting the Lie derivative again the approximationm ˙ 0 we ﬁnd onto the surface ≪ ¯ c d 2 t 4 2 δX wab = qaqb X wab. (27) AEH =4π (2m) 1 + 12m ˙ + 2m ˙ + mα˙ . (23) L m 3 The variation of the area element ǫ for example satisﬁes If we compare the area of the event horizon on a constant t¯ hypersurface with the area of the event horizon on a δX ǫ = Cθ n ǫ. (28) − ( ) 6

Variations of extrinsic properties are a little harder to toolkit [26, 27] which is widely used in numerical rela- calculate. We are particularly interested in deformations tivity, for example to perform binary black hole, neutron of the form star, or stellar core collapse simulations. For each leaf we write the 3+1-decomposed Vaidya metric onto the Carte- δX θ(ℓ) =0, (29) sian grid of Cactus. In each ’time step’ we write the next leaf of the slicing onto the Cartesian grid. This mimics since this will generate a class of marginally outer a numerical time evolution of an initial Cauchy slice of trapped surfaces. The calculations in [32] give a dynamical BH spacetime. To locate the trapping hori- 2 a zon at each ’instant of time’ we use the apparent horizon δX θ ℓ = d C +2˜ω daC Cδ n θ ℓ + Bδ ℓ θ ℓ , (30) ( ) − − ( ) ( ) ( ) ( ) finder AHFinderDirect [24, 25]. This way we obtain the where d is the covariant derivative compatible with the location and area of the trapping horizon as a function intrinsic metric of the two-surface. Notice that for radial of time. In the following we compare these functions for ℓa we have (3) and thus the different trapping horizons we detect on the axisymmetric slicings from above. 2m ˙ δ ℓ θ ℓ = , (31) ( ) ( ) − r2 A. Location of tilted versus untilted horizons since the shear is zero. Furthermore, axisymmetric variations from the spherically symmetric surface will satisfy The MOTS lying on the hypersurfaces t¯ = v r − − d2C dC αz = 0 with different values of α do not intersect. An + cot θ + C 2mB ˙ =0. (32) dθ2 dθ − examination of the marginally outer trapped surface with α = 0.83 shows that the points at the north pole θ = 0, In the case where both B and C are constant we get equator θ = π/2 and south pole θ = π have the four- the evolution along the spherically symmetric trapping dimensional spacetime coordinates given in Table (I) horizon. In [32] it was assumed that the deformations were l-oriented such that B > 0 everywhere. But another solution of this equation is TABLE I: Coordinate location of various points on surface with t¯ = 0, α = 0.83 and m(v) = 1.0 + 0.01v k cos θ B = , (33) 2m ˙ surface point r v t zmax (north pole) 2.0454 3.7498 1.7044 C = k cos θ, (34) xmax (equator) 2.0386 2.0386 0.0000 − zmin (south pole) 2.0247 0.3374 -1.6872 where k is a constant of integration. Since the variation in this case is dependent on θ it will not generate spherically symmetric marginal surfaces, and since B = C = 0 for Trapping horizons can be formed by stacking surfaces θ = π/2 this will not generate evolution along a trapping found on different hypersurfaces with different values of horizon. The area change in this case is easily computed t¯ but the same value of α. These trapping horizons, each as with their own value of α will intersect one another. This 2k can also been seen in Table (I). The points on the equator δX ǫ = Cθ n ǫ = cos θ sin θdθdφ =0. (35) Z − Z ( ) r Z lie inside the horizon located on the spherically symmetric slicing t = v r = 0, where every point on the horizon In this case, at least, the area of the spherically sym- has r = 2.0408− and t = 0. The point at the north pole metric surface is extremal. For a general axisymmetric lies inside the spherically symmetric horizon on the slice variation expanded as a Fourier series, the area of the t =1.7044, which lies at r =2.0756, and the point at the spherically symmetric surface is extremal provided the south pole lies outside the spherically symmetric horizon coefficients a(n) of the cos(nθ) terms in C are zero for on the slice t = 1.6872, which lies at r =2.0064. even n. In terms of the− deformations discussed above (25) we see that to deform the spherically symmetric surface at t = 0 we need B > 0 and C > 0 at the north pole and III. NUMERICAL RESULTS FOR B < 0 and C < 0 at the south pole, where in both cases AXISYMMETRIC SLICINGS the size of C is much smaller than B. Figure (1) shows the locations (r coordinate) of the We now compare the different trapping horizons one north pole, equator and south poles for MOTS against α obtains for the axisymmetric slicings given by (20) where for a mass function of the form m(v)=1.0+0.01v. For we consider the spherically symmetric mass functions (4), values of α less than zero the location of the MOTS is (10). For the numerical calculation in the following we the same with the north pole θ = 0 and south pole θ = π use, as in previous work [20], the Cactus computational interchanged. Near the spherically symmetric MOTS at PSfrag replacements m˙ =0.02 7 α A / max(A) 1.00000 2.045 north pole m˙ A(α =0.0) − A(α =0.83) 0.99995 2.040 equator ) A m˙ =0.02 Time ( 0.99990 A 2.035 0.99985 / max Radius south pole PSfrag replacements A 2.030 0.99980

0.99975 2.025 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 α α

1.00 FIG. 1: The location (r-coordinate) of the north pole, equator and south pole versus diﬀerent values of α for a linear mass 0.98 m v . . v α

function ( ) = 1 0 + 0 01 . The positions for negative are ) α 0.96 the same as for positive with the north pole and south pole A ( interchanged. Near α = 0 there is no change in the position 0.94 of the equator and the change in the position of the north 0.92 pole and south pole are equal but opposite in sign. / max

PSfrag replacements A 0.90 α = 0 the change in the r coordinate of the equatorm˙ =0.02 0.88 is zero, whereas the change in the r coordinates of the north and south poles is equal but opposite in sign. This 0.0 0.2 0.4 0.6 0.8 1.0 supports the conjecture that for these cases C is of the α form (34). FIG. 2: Slicing dependence of the area at t¯for the linear mass function (top) with m = 1.0 andm ˙ = 0.02 and for the tanh log mass function (bottom) with m = 1.0 and T = 1.0 B. Area dependence on α

The results for the area dependence of α are shown in Fig.(2) for the linear mass case and Fig.(2) for the much larger than for the apparent horizon. tanh log mass function. The α = 0 surfaces have the largest area. As expected the area of the tilted horizons is unchanged under a change in sign of α. The difference between the area on α = 0 slices and α = 0 slices depends 6 onm ˙ as shown in Fig.(3) for the linear mass function. It C. Dependence of rate of change of area on α is zero form ˙ = 0, where the spacetime just reduces to the Schwarzschild solution and the horizon is an isolated horizon. The difference between the area of tilted and For the linear mass function the α = 0 surfaces seem untilted horizons increases with increasing magnitude of to have the same rate of change of area6 as the α = 0 m˙ . The difference is slightly larger for positive values of surfaces, at least to within numerical accuracy. For the m˙ than for negative values. This difference in area does tanh log mass function, as depicted in Fig.(4), the area of not seem to follow a simple power law and it is difficult to tilted horizons grows at a greater rate than the untilted extrapolate what the difference in areas will be for values areas and both converge to the asymptotic isolated hori- ofm ˙ different form the values examined. However, the zon area, which will also be the asymptotic area of the difference in areas for values ofm ˙ much lower than the event horizon. In this sense the spherically symmetric lowest positive value ofm ˙ considered here (m ˙ = 0.02) trapping horizon plays a special role as it is the one with are likely to be much less than the 0.035% difference the smallest rate of increase of area. This would seem ∼ found here for that case. to contradict the conjecture made in [33] that the pre- Finally, Fig. 5 compares the area of the event horizon ferred trapping horizon should be the one where the area (or more precisely, the intersection of the event horizon increases the most and approaches the event horizon the with the spatial slices corresponding to different values fastest. However, it is not known whether this behaviour of α) with the area of the MOTS as functions of α. It is is repeated in more general situations with other mass interesting that the variation in the event horizon area is functions or other slicing conditions. PSfrag replacements 8 m˙ =0.02 α A / max(A) 64. 54.64 0.0 m˙ A(α =0.0) − A(α =0.83) 63.5 54.62 83)

. - 0.5 Time A 63. 54.59 =0 - 1.0 α A (α), m˙ =0.02

α EH ( Radius 62.5 54.56

A - A α m . PSfrag replacements 1.5 north pole AH( ), ˙ =0 02

− equator 62. 54.53 -2.0 0) m . south pole Area of event horizon .

˙ =0 02 Area of apparent horizon 61.5 54.51 -2.5

α =0

A A α 61. 54.48 / max( ) ( -3.0 0. 0.2 0.4 0.6 0.8 1. A α -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 m ˙ FIG. 5: Area of the event horizon AEH (dashed line) and the MOTS AAH for different values of α for the linear mass FIG. 3: Area difference for two slices in different Vaidya space- function withm ˙ = 0.02. The y-axis on the left refers to the times with linear mass function and m = 1.0, The difference area of the event horizon and the axis on the right to the between the areas increases for largem ˙ . The untilted areas MOTS; the scale of AAH has been expanded to better show are larger than the tilted areas form ˙ both positive and nega- its variation. The variation of AEH is much larger than the tive. However, the difference is much larger form> ˙ 0. Only variation of AAH and the variation of AAH is small relative to form ˙ = 0 do the areas coincide. the difference between AEH and AAH.

vary by much, although the larger the rate at which the 49 black hole accretes matter the larger the diﬀerence in PSfrag replacements 48 the areas. For them ˙ = 0.02 Vaidya solution with linear mass function the area can vary by approximately 0.03% m˙ =0.02 47 as α is varied from 0 to 1. Beyond α = 1 the Cauchy

A surface becomes timelike for certain values of θ. This is α 46 A / max(A) the smallest value ofm ˙ that we investigated but it is still 45 much larger than the mass accretion rate expected for m˙ astrophysical black holes. A(α =0.0) − A(α =0.83) 44 From a purely practical point of view we’ve found ev- 43 idence that the parameters of the black hole, derived 2 4 6 8 10 from its trapping horizon, do not change significantly Time when looking at reasonably simple slicings. It is still un- known whether the properties would change drastically FIG. 4: Area versus time t¯ for α = 0 and α = 0.83 for the for certain unusual slicings, but these are unlikely to oc- tanh log mass function with m = 1.0 and T = 1.0. Both areas cur in normal numerical simulations. It is also reassuring converge rapidly to the asymptotic isolated horizon area 16π. that these differences are much smaller than the param- The tilted area grows faster since it starts from a lower point. eter biases of up to 10% found between different post- Newtonian models [15]. This variation of the parameters with foliation will also occur if one uses the event horizon. IV. CONCLUSION AND DISCUSSION From a conceptual point of view, we have demon- strated explicitly the known result that the location of We have investigated the slicing dependence of the ap- the black hole surface and some of its properties such parent horizons in the Vaidya spacetime. For a given slice as the area depend on the choice of the spacetime slic- we have looked at three types of horizons; the MOTS that ing. There are a variety of different possible responses lie on the slice, the intersection of the slice with the event to the issue of non-uniqueness of the trapping horizons horizon and the intersection of the slice with the spher- in a given spacetime. The oldest approach is to focus ically symmetric trapping horizon. We have examined purely on the event horizon as the unique indicator of rather simple axisymmetric slicings in this simple space- the black hole and its properties. The location of the time. These simple slicings do not exhaust all possible event horizon is independent of the choice of foliation. axisymmetric slicings and there are many other possible This approach however causes trouble in certain quan- slicings that are not axisymmetric. tum inspired spacetimes [3, 18] where no event horizon Explicitly the location of the horizons is different for exists. In numerical settings, the acausal definition of the different choices of horizon and different choices of slicing. event horizon means that it is only known after a simula- However, for slowly evolving horizons, the areas do not tion is finished, preventing it from being used to analyse 9 the simulation’s state as it proceeds. takes the form Another approach is to accept all horizons on an equal 2 2 2 2 footing as purely a property of the geometry. In this pic- ds = dv + 2drdv + r dΩ , (36) −△ ture it is not clear how to associate unique properties to 2m(v) ′ 2m(v) ˙ 2m ˙ the black hole such as a horizon area or horizon angular where = 1 r , = r2 , = r andm ˙ = △ − △ △ − momentum. Because the horizons intersect one cannot ∂vm(v). rely on specifying, for example, the outermost trapping a horizon. However, it may be possible to formulate a gen- Radial null vectors, with canonical normalization ℓ na = eralized second law for each possible foliation of space- 1 and na = ∂vr: − − time and in this context use the horizon defined by the a chosen foliation. ℓ = 1, △, 0, 0 , (37) A third approach is to look for the boundary of the re- 2 gion that admits trapped surfaces. This surface should be ℓa = △, 1, 0, 0 , (38) spherically symmetric in a spherically symmetric space- − 2 time. If this surface lies strictly outside the spherically a n = (0, 1, 0, 0) , (39) symmetric trapping horizon then, by the results of [5], it − na = ( 1, 0, 0, 0) . (40) cannot itself admit the structure of a dynamical horizon. − This surface will have a location that is independent of any given foliation, although it may not have a simple Expansions of the null normals: thermodynamic interpretation. 2 A fourth approach is to look for properties that select θ(ℓ) = △ , θ(n) = . (41) r − r out a certain preferred trapping horizon in the spacetime [34], such as the spherically symmetric horizon in The location of the spherically symmetric MOTS is ob- a spherically symmetric spacetime. This would require tained by setting θ(ℓ) = 0 leading to r = 2m(v). The some additional condition to be imposed that selects out variations of the expansions along na: a unique trapping horizon from the many that occur in ′ dynamical black hole spacetimes. a n aθ ℓ = △ △ , (42) Despite their quasi-local nature, closed marginally ∇ ( ) r2 − r 2 ′ trapped surfaces do have some non-local behaviour. a ˙ ℓ aθ ℓ = △ △ + △△ , (43) Their dependence on the slicing is but one manifesta- ∇ ( ) r − 2r2 2r tion of this behaviour. The fact that marginally outer a 2 n aθ n = , (44) trapped surfaces can be found all the way to the event ∇ ( ) −r2 horizon [21] is another manifestation. Although the event a horizon is a fully non-local teleologically defined surface, ℓ aθ n = △ . (45) ∇ ( ) r2 it still acts as the boundary of outer trapped surfaces in a Vaidya spacetime that satisfies the null energy condition Other optical scalars: and asymptotes to a static Schwarzschild solution in the far future. σ(ℓ) = σ(n) = ω(ℓ) = ω(n) =0 . (46) This behaviour is related to both the choice of the surface null normals and the requirement that the Surface gravities: marginally trapped surface be closed. Closed trapped ′ a b surface cannot be found entirely in a flat spacetime. How- κ(ℓ) = n ℓ bℓa = △ , (47) ever, parts of a closed marginally trapped surface can be − ∇ 2 a b found passing through a region of flat space where both κ(n) = ℓ n bna =0 . (48) − ∇ null expansions are negative [22]. It may be that in the quantum context only an effec- Components of the energy-momentum tensor: tive horizon is meaningful or that the existence of many m˙ ab a b intersecting horizons contribute to the full path integral. T = 2 n n , (49) For purely classical situations one may be able to refine 4πr a b m˙ the definition of the surface of a black hole. Tabℓ ℓ = , (50) 4πr2 a b a b Tabn ℓ = Tabn n =0 . (51) V. APPENDIX - SOME USEFUL RELATIONS Normal to the trapping horizon = 0: IN THE VAIDYA SPACETIME △ a N = (1, 2m, ˙ 0, 0) , (52) These relations are provided here as a repository − for reference in the main text. In the advanced null Eddington-Finkelstein coordinates (v,r,θ,φ) the metric Na = ( 2m, ˙ 1, 0, 0) . (53) − 10

Radial tangent to the trapping horizon For this null vector to be the generator of an isolated a b ab a horizon it is required to satisfy T v v = 0 giving A =0 V = (1, 2m, ˙ 0, 0) , (54) and therefore from above C = 0. Thus any null vector generating an isolated horizon must be proportional to θ(n) but from the appendix we see θ(n) = 2/r. This Va = (2m, ˙ 1, 0, 0) . (55) only vanishes at inﬁnity. There may be a− degenerate planar isolated horizon “at” inﬁnity. Variation of the area along the trapping horizon

a 4Am˙ V aA = . (56) ∇ r Connection on the normal cotangent bundle

b c ωã = qãnc bℓ = (0, 0, 0, 0) . (57) − ∇ Scalar curvature of the horizon 2 R˜ = . (58) VI. ACKNOWLEDGMENTS r2 The null energy condition is satisfied if E > 0. This is easy to see in a fiducial spherically symmetric null tetrad a a a a a Alex Nielsen gratefully acknowledges financial support ℓ ,n ,m , m¯ since for a general null vector v with from the Alexander von Humboldt Foundation and hos- a a a a ¯ a pitality at the Albert Einstein Institute; Erik Schnet- v = Aℓ + Bn + Cm + Cm¯ , (59) ter acknowledges support from the NSF awards 0701566, a b 2 a b m˙ 2 a 2 0721915, and 0904015. The numerical solutions were then Tabv v = A Tabℓ ℓ = 2 A (ℓ na) . The null 4πr found using the AHFinderDirect thorn [24, 25] of the energy condition is closely related to whether there is Cactus framework [26, 27] as part of the Einstein Toolkit positive or negative energy flowing into the black hole. [28]. We also thank Alberto Sesana and Jose Luis Jaramillo for helpful discussions. Ifm ˙ = 0 there are no isolated horizons in the Vaidya spacetime6 since the norm of the generalized null vector (59) is During the work for this paper we were saddened to hear of the passing of Professor P. C. Vaidya, upon whose a v va = 2AB +2CC¯ =0. (60) solution this work is based. −

[1] A. Ashtekar and B. Krishnan, Living Rev. Rel. 7 (2004) [16] A. B. Nielsen, arXiv:1006.2448 [gr-qc]. 10 [arXiv:gr-qc/0407042]. [17] I. Booth and J. Martin, arXiv:1007.1642 [gr-qc]. [2] I. Booth, Can. J. Phys. 83 (2005) 1073 [arXiv:gr- [18] S. A. Hayward, Phys. Rev. Lett. 96 (2006) 031103 qc/0508107]. [arXiv:gr-qc/0506126]. [3] A. B. Nielsen, Gen. Rel. Grav. 41 (2009) 1539 [19] W. A. Hiscock, Phys. Rev. D 23 (1981) 2813. [arXiv:0809.3850 [hep-th]]. [20] E. Schnetter and B. Krishnan, Phys. Rev. D 73 (2006) [4] P. C. Vaidya, Phys. Rev. 83, 10 (1951). 021502 [arXiv:gr-qc/0511017]. [5] A. Ashtekar and G. J. Galloway, Adv. Theor. Math. [21] I. Ben-Dov, Phys. Rev. D 75 (2007) 064007 [arXiv:gr- Phys. 9 (2005) 1 [arXiv:gr-qc/0503109]. qc/0611057]. [6] A. Ashtekar and B. Krishnan, Phys. Rev. D 68, 104030 [22] I. Bengtsson and J. M. M. Senovilla, Phys. Rev. D 79 (2003) [arXiv:gr-qc/0308033]. (2009) 024027 [arXiv:0809.2213 [gr-qc]]. [7] I. Hinder, Class. Quant. Grav. 27, 114004 (2010) [23] J. E. Aman, I. Bengtsson and J. M. M. Senovilla, J. Phys. [arXiv:1001.5161 [gr-qc]]. Conf. Ser. 229 (2010) 012004 [arXiv:0912.3691 [gr-qc]]. [8] A. Buonanno, G.B. Cook, and F. Pretorius, Phys. Rev. [24] J. Thornburg, Phys. Rev. D 54 (1996) 4899 [arXiv:gr- D 75 (2007) 124018 qc/9508014]. [9] J.G. Baker et al., Phys. Rev. Lett. 99 (2007) 181101 [25] J. Thornburg, Class. Quant. Grav. 21 (2004) 743 [10] M. Campanelli et al., Phys. Rev. D 79 (2009) 084010 [arXiv:gr-qc/0306056]. [11] M. Boyle et al., Phys. Rev. D 76 (2007) 124038 [26] T. Goodale, G. Allen, G. Lanfermann, J. Mass´o, [12] M. Hannam et al., Phys. Rev. D 77 (2008) 044020 T. Radke, E. Seidel and J. Shalf, “The Cactus Frame- [13] I. Hinder, F. Herrmann, P. Laguna and D. Shoemaker, work and Toolkit: Design and Applications,” Vector and Phys. Rev. D 82 (2010) 024033 [arXiv:0806.1037 [gr-qc]]. Parallel Processing – VECPAR’2002, 5th International [14] L. Santamaria et al., arXiv:1005.3306 [gr-qc]. Conference, Lecture Notes in Computer Science, (2003), [15] A. Buonanno et al., Phys. rev. D 80 (2009), 084043 Springer, Berlin, url = http://edoc.mpg.de/3341, 11

[27] Cactus Computational Toolkit, http://www. 95 (2005) 111102 [arXiv:gr-qc/0506013]. cactuscode.org [32] I. Booth and S. Fairhurst, Phys. Rev. D 75 (2007) 084019 [28] Einstein Toolkit, http://einsteintoolkit.org [arXiv:gr-qc/0610032]. [29] C.W. Misner and D. H. Sharp, Phys. Rev. 136, B571 [33] E. Gourgoulhon and J. L. Jaramillo, Phys. Rev. D 74 (1964). (2006) 087502 [arXiv:gr-qc/0607050]. [30] J. A. Kollmeier et al., Astrophys. J. 648 (2006) 128 [34] S. A. Hayward, arXiv:0906.2528 [gr-qc]. [arXiv:astro-ph/0508657]. [31] L. Andersson, M. Mars and W. Simon, Phys. Rev. Lett.