Arxiv:0808.1745V1 [Gr-Qc] 12 Aug 2008 Atri H Nevnn Pc.I Pctm Is Spacetime If Space

Home , Spacetime symmetries

arXiv:0808.1745v1 [gr-qc] 12 Aug 2008 atri h nevnn pc.I pctm is spacetime If space. of intervening ordinary stress-energy the the in the precisely matter from is computed 2-spheres integrals momentum different these angular of between pair difference a the for axi-symmetric, vacuum, is not spacetime but When infinity. at momentum measured can angular spacetime well-defined in physically 2-sphere the arbitrary reproduce an over 2] [1, integral field vector symmetry Killing rotational a example, a by For generated admits spacetime vacuum relativity. a general when and in energy momentum as such angular quantities conserved important cally iuu.Te eedntol nte2-sphere the on only not am- depend rather They become formulae biguous. such however, symmetric, ehiusietf rfre 2-spheres these describe Essentially, preferred may identify regions. that techniques other spacetime in non- a dynamics isolated, in interesting appro- an hole an modeling on black 5] or dynamical [4, review), horizon recent isolated a flat priate for asymptotically mo- [3] in angular (see infinity spacetimes at instance, well-defined For is mentum conditions. favorable ically integrand. the define to hc n nerts u loo h etrfield vector the on also but integrates, one which eutn us-oa omlehv h eea Brown– general form The the [6] have reduce York momentum. formulae thereby angular and quasi-local the integrate, resulting defining to in ambiguity which the over etc.) zon, oetm u ahrta hyhave they that rather but momentum, have surfaces ymtyo n idadteeoe emnl,n pre- no seemingly, field therefore, vector and ferred kind no any have of generally symmetry processes astrophysical symmetry of simulations that spacetime.) least, of bulk at the (Now, into it. extend not select need to used be can ihnΣ h ai rbe ean:tevco field vector the remains: problem basic ϕ The Σ. within where savco edon field vector a is xrni uvtr fasailsieΣcnann it, containing Σ slice spatial a of curvature extrinsic a pctm ymtisaeesnilfrdfiigphysi- defining for essential are symmetries Spacetime hs icliscnb atal vie ne phys- under avoided partially be can difficulties These h oioso lc oe eutn rmnumerical from resulting holes black of horizons The sabtayunless arbitrary is S sa(ehp rfre)2-sphere, preferred) (perhaps a is ceeta ol eo meit rcia s nnumerical in use practical based immediate fields of Killing be overlo could approximate been that have calculate scheme to to seems sugge how which definition Matzner, describes is by a proposal it relates older However, paper an This relativity. general geometrically. in defined holes black symmetric J no prxmt iln etrfilsaeepce ohl define help to expected are fields vector Killing Approximate [ ϕ := ] esnbedfiiinfrteangular the for definition reasonable ϕ S eateto hsc,FoiaAlni nvriy oaR Boca University, Atlantic Florida Physics, of Department a n d and , prxmt iln ilsa nEgnau Problem Eigenvalue an as Fields Killing Approximate 8 h rbe sntta such that not is problem The . S πG 1 a an has I S S b ϕ steae lmn of element area the is intrinsic a ϕ K a ab hnaKomar-type a then , d S S nntl many infinitely b ymtythat symmetry , ifiiy hori- (infinity, K ab Dtd uy1,2007) 13, July (Dated: not ϕ hitpe Beetle Christopher S a sthe is used over axi- (1) ϕ S a . hr soefreeyvco field vector every for one is There etrfil,adteovostigt oi ose a seek to preferred is a do pick to to thing technique obvious the a and is field, needed vector is What zon. ht nsm es,i scoea osbet Killing a the to to intuitively possible leads an This as of present. close idea is as none if is even sense, field, some in that, in[0 fa prxmt iln edo compact a on field Killing approximate an manifold of [10] Whiting. tion and Cook of relationship that intimate and scheme the this elucidates between numer- and in holes of relativity, black adaptation ical for momentum novel computing angular of a preferred eigen- problem a specific suggests an the solving also to approach ap- on It Matzner’s older based an Matzner problem. revives to value paper due This [10] by [9]. proach scheme Whiting minimization and a recently Cook most Killing and include conformal [8], [7], These vectors transport Killing on 2-spheres. based on schemes of fields definitions elegant Killing proposed approximate recently have groups eral yaKligvco field vector Killing a by metric throughout ξ emtywt otnossmer disa least equation at the admits to solution symmetry non-trivial continuous one a with geometry eemndby determined eaesilfe ose ouin o hseuto.But equation. this any. for find solutions not will seek we to generically free is still geometry are the we when even principle, In nagnrcgoer.Teoeao ∆ operator The geometry. generic a on erctno ed nReana aiod 1] re- A ∆ [12]. denoted manifolds operator, Riemannian sym- lated on of fields decomposition tensor transverse metric the in naturally arises a e sbgnwt re eiwo aze’ defini- Matzner’s of review brief a with begin us Let sev- above, discussed issues general the by Motivated osdrteegnau problem eigenvalue the Consider and g tdrcnl yCo n htn to Whiting and Cook by recently sted o biu o uhfilssol be should fields such how obvious not ∇ ab nteepooasuiga efficient an using proposals these on M relativity. a otnossmer of symmetry continuous A . kdi h eetltrtr.I also It literature. recent the in oked approximate hsclymaigu pn o non- for spins meaningful physically M steuiu oso-recneto on connection torsion-free unique the is fdimension of ∆ where , g K ab tn lrd,33431 Florida, aton, L ξ aigadvrec,w e htany that see we divergence, a Taking . a − ξ := g 2 ab L ∇ − iln field. Killing ξ 2 = ξ b 2 eoe h i eiaiealong derivative Lie the denotes a n L ∇ ∇ rssi h aewyfrom way same the in arises , satisfying qipdwt Riemannian a with equipped ( ∇ b b ∇ ξ ( a a ( ) ξ b b 0 = ξ ϕ ) a a (2) 0 = ) agn otehori- the to tangent . = K κξ g not ab pern here appearing a sgenerated is symmetric, (3) (4) ϕ M a 2 the conformal Killing equation, and plays a similar role us to solve Eq. (4) for ξa. However, it does make an ap- in the transverse-traceless decomposition of such fields. proach to Matzner’s eigenvalue problem via a variational Its application to the initial-data problem in general rel- principle like Eq. (6) rather complicated. There is no al- ativity is very well known indeed [13]. gebraic equation to determine the Lagrange multiplier. Eq. (4) of course admits solutions ξa only for a cer- Indeed, κ is determined in this problem precisely by the tain spectrum of eigenvalues κ, and zero may or may not condition that it be an eigenvalue of ∆K , and there is no be among these. Matzner [10] establishes the following algebraic equation giving these. Minimizing QK [ξ; κ) in four results: the spectrum of eigenvalues κ of Eq. (4) Eq. (6) by solving the associated Euler–Lagrange equa- on a compact manifold is (a) discrete, (b) non-negative, tions is neither easier nor harder than solving the eigen- (c) corresponds to a complete set of real vector eigenvalue problem in Eq. (4). fields ξa, and (d) contains κ = 0 if and only if the cor- Cook and Whiting’s recent definition [9] of an approx- responding eigenfield ξa is a genuine Killing field. That imate Killing field uses a variational principle based on is, the zero eigenspace of Eq. (4) is precisely the finite- a quadratic form closely related to that of Eq. (6). How- dimensional vector space of Killing fields of the metric gab ever, it differs in a two important details. First, it focuses on M. Therefore, Eq. (3) admits no solution if the met- on the case where M ≃ S is topologically a 2-sphere, and a ric gab on M has no continuous symmetries, as claimed restricts ξ to be area-preserving: above. However, we assert that the best approximation L c to a Killing field on a manifold with no actual symmetry ξ ǫab = (∇c ξ ) ǫab =0. (8) is the unique vector eigenfield ξa of Eq. 4 with the min- The motivation for this restriction arises from the tech- imum eigenvalue κ > 0. This is Matzner’s definition of nical details of an eventual application to calculating the an approximate Killing field, and it has several desirable angular momentum of a non-symmetric black hole [4]. features. It exists generically, reduces to the correct an- Second, it is based on a non-standard inner product swer when symmetries do exist, and, like a true Killing

field on a symmetric manifold, is naturally defined only a up to an overall (i.e., constant over M) scaling. hζ, ξiR := hζ,Rξi = ζa ξ R ǫ. (9) IS Like any eigenvalue problem, Eq. (4) admits a variational formulation. Recall the natural L2 inner product These choices change the form, but not the basic content, of the resulting equations. They still describe a sort of a self-adjoint eigenvalue problem. hζ, ξi := ζa ξ ǫ (5) IM To restrict to area-preserving vector fields, it is easiest simply to recall that any divergenceless vector field ˚ξa on the space of (complex) vector fields over M. Here, ǫ on a 2-sphere topology is described by a unique scalar denotes the canonical n-form volume element induced on potential Θ such that M by the metric gab. We minimize the quadratic form a a ab 1 1 ˚ξ = (∗dΘ) := −ǫ ∇b Θ and Θ ǫ =0. (10) QK [ξ; κ) := 2 hξ, ∆K ξi− 2 κ hξ, ξi− 1 , (6) IS where κ is constant over M and here plays the role of Now consider the restricted eigenvalue problem a Lagrange multiplier. Minimizing this functional pro- ˚ ˚a ˚ ˚ ˚a ˚a duces the Euler–Lagrange equations ∆K ξ := P ∆K P ξ = ˚κ ξ , (11) a a where P˚ denotes the orthogonal projection onto the sub- ∆K ξ = κξ and hξ, ξi =1, (7) space of area-preserving vector fields within the Hilbert a a the solutions of which are clearly the vector eigenfields of space of Eq. (5). A given ˚ξ = (∗dΘ) solves this equa- Eq. (4), normalized to unity in Hilbert space. tion if and only if, for all other ˚ζa = (∗dΦ)a, we have Many variational problems are solved by initially solving the first, differential equation in Eq. (7) for ξa as h∗dΦ, ∆K ∗dΘi = ˚κ h∗dΦ, ∗dΘi. (12) a function of an arbitrary Lagrange multiplier κ, and Integrating both sides by parts, and using positivity of then using that result in the second, algebraic equation the standard L2 inner product of scalar functions over S, to impose the constraint and determine κ. This does not we find that Eq. (11) is equivalent to happen for Eq. (7) because the differential equation is a 2 a linear in ξ , and therefore leaves the overall scaling of ∆˚∆K Θ := ∆ Θ+ ∇ (R ∇a Θ) =˚κ ∆Θ, (13) ξa undetermined. The second equation serves only to fix a that scaling, and cannot also determine κ. The Lagrange where ∆ := −∇ ∇a denotes the standard scalar Lapla- multiplier therefore must be fixed when we solve the first cian. We have shown that the scalar functions Θ solv- equation; for general κ, no solution exists. This is hardly ing Eq. (13) generate, via Eq. (10), solutions ˚ξa of the surprising since of course only true eigenvalues κ allow restricted eigenvalue problem of Eq. (11). Because the 3 projection P˚ does not typically commute with ∆K , these which the total scalar curvature remains everywhere pos- ˚ξa do not generally solve Eq. (4), and the restricted eigen- itive. In this region, there is no obvious problem with the values ˚κ are generally distinct from the eigenvalues κ in Cook–Whiting scheme, but nothing particular to recom- the full Hilbert space. In fact, we should generally ex- mend it either. The question could presumably be settled pect that ˚κmin >κmin. However, the area-preserving vec- [11] by comparing qualitative features of the approximate tor eigenfield corresponding to this minimum restricted Killing fields computed from Eqs. (11) and (15). eigenvalue can also be considered a best approximation Matzner’s eigenvalue definition of an approximate to a Killing field, albeit within a restricted class. Killing field is unambiguous, universally applicable, and To recover the Cook–Whiting approximate Killing reproduces the usual Killing fields on a symmetric man- field, we must repeat the previous calculation in the inner ifold. But it is not necessarily efficient in practice. In- product of Eq. (9). The operator ∆K is then no longer deed, it would prohibitively expensive to solve any of the −1 self-adjoint, but R times ∆K is. Accordingly, we seek eigenvalue problems in Eqs. (4), (11) or (15) on the ap- ˚a a vector fields ξR = (∗dΘ) satisfying parent horizon at every moment of time of a black hole in a numerical simulation. Fortunately, however, there , R−1 κ , h∗dΦ ∆K ∗dΘiR = ˚R h∗dΦ ∗dΘiR (14) is a simple approximation to speed the process up on a for all ˚ζa = (∗dΦ)a. Integrating by parts, and once again generic geometry. This approximation is based on the invoking positivity of the standard inner product of scalar Rayleigh–Ritz method [15], and works so long as we only lowest functions, we find that Eq. (14) is equivalent to want to find the eigenvalue and the corresponding vector eigenfield. a ∆˚∆K Θ= −˚κR ∇ (R ∇a Θ). (15) Consider the Rayleigh–Ritz functional

Although the notation here differs slightly, this is pre- (a b) 2 ∇(a ξb) · ∇ ξ · ǫ cisely the Euler–Lagrange equation that Cook and Whit- hξ, ∆K ξi I F [ξ] := = M (17) ing find [9] by minimizing a quadratic form similar to hξ, ξi ξ ξa ǫ ˚a a Eq. (6). Once again, the solutions (ξR,˚κR) of this eigen- IM value problem generally differ from the solutions (ξa,κ) on the full Hilbert space of Eq. (5), with the zero vec- ˚a of Eq. (4) and from the solutions (ξ ,˚κ) of Eq. (11). tor removed. The local extrema of Eq. (17) occur when Let us now make two technical comments. First, any a ξ is a vector eigenfield of ∆K , and the value of F [ξ] constant function Θ = c will give zero on both sides of at each such extremum is the corresponding eigenvalue Eqs. (11) and (15) for all values of˚κ or ˚κR, respectively. κ. Note that the numerator here, which arises via in- These are spurious solutions. They arise only because tegration by parts of the second-order operator ∆K in we have used potentials to describe the subspace of area- Eq. (4), is precisely one half the square integral of Lξ gab preserving vector fields. These solutions are ruled out by from Eq. (2). Thus, among all vector fields with fixed the the second condition in Eq. (10), which makes the L2-norm on S, diffeomorphisms along the approximate ˚a correspondence between ξ and Θ an isomorphism. Killing field modify the metric least in a quantifiable, L2 Second, the Cook–Whiting inner product in Eq (9) sense. looks a little odd, but it is not immediately clear whether It is still not practicable to find the genuine abso- there is anything technically wrong with it. There cer- lute minimum of Eq. (17) on the computer, which of tainly can be problems. Recall that the scalar curvature course would yield Matzner’s approximate Killing field. in two dimensions varies as But one can approximate that minimum by minimizing 2 b a 1 2 a F [ξ] within an appropriate space of trial vector fields. δ R = ∇ ∇ a δgb − Rδga (16) [ ] 2 This idea is familiar from elementary quantum mechan- under a perturbation δgab of the metric. If this pertur- ics, where just such a variational principle is routinely bation varies sufficiently rapidly over S, then the first used to approximate the ground-state wave-function of a term here can easily dominate the second, as well as the complicated system. Unless the subspace of trial fields unperturbed, background value 2R. The result is that a one chooses happens to be orthogonal, or nearly so, to a generic spherical geometry, even if perturbatively close the true minimum ξtrue of F [ξ] in all of Hilbert space, the a to a round sphere in the sense that δgab has small ampli- dominant component of the minimizing trial field ξtrial a tude, can have regions of negative scalar curvature. (This should lie along ξtrue in Hilbert space. Most randomly- a is intuitively obvious if we imagine “pinching” the surface chosen trial spaces will not be orthogonal to ξtrue. This of a round sphere to create a small, saddle-shaped region idea allows us to approximate Matzner’s approximate of negative curvature.) On such geometries, the “inner Killing field. product” of Eq. (9) is not positive-definite, and does not There is a natural candidate for the trial space of vec- define a Hilbert space. However, in the space of all spher- tor fields in which to minimize Eq. (17) in the specific ical geometries, there should be some finite region of “suf- case M ≃ S of a 2-sphere horizon of a quiescent black ficiently smooth” perturbations of the round sphere for hole in numerical relativity. One striking feature of many 4 recent numerical simulations (e.g, [14]) is that the hori- Acknowledgements zons at late times often look fairly regular in the fiducial spacetime coordinates used to do the evolution. There- The author would like to thank Ivan Booth, Manuela a fore, it is natural to try a space of trial fields ξ based Campanelli, Greg Cook, Stephen Fairhurst, Greg Gal- simply on those coordinates. A specific proposal follows. loway, Carlos Lousto, Charles Torre, Bernard Whiting Use the fiducial spacetime coordinates in which the nu- and Yosef Zlochower for stimulating discussions related merical evolution occurs to induce spherical coordinates to this question. This work has been supported by NSF (θ, φ) on the black-hole horizon in some more-or-less nat- grants PHY 0400588 and PHY 0555644, and by NASA ural, but fundamentally ad hoc, way. Then, take the grant ATP03-0001-0027. space of scalar trial potentials

lmax l lm Θ(θ, φ) := Θ Yˆlm(θ, φ), (18) Xl=1 mX=−l [1] A. Komar. Covariant Conservation Laws in General Rel- ativity. Phys. Rev. 113 (1959) 934-936. where Yˆlm(θ, φ) are the ordinary scalar spherical har- [2] R.M. Wald. General Relativity. University of Chicago monic functions on a round sphere, Θlm are arbitrary Press, Chicago, 1984. constants, and lmax is a chosen cut-off. Each of these [3] L.B. Szabados. Quasi-Local Energy-Momentum and An- potentials generates an area-preserving vector field via gular Momentum in General Relativity: A Review Arti- Eq. (10), and this will be our trial space [16] within the cle. Living Rev. Relativity 7 (2004) 4. Cited 8 July 2007. full Hilbert space of Eq. (5). Therefore, minimize [4] A. Ashtekar, C. Beetle and J. Lewandowski. Mechanics of rotating isolated horizons. Phys. Rev. D 64 (2001) h∗dΘ, ∆ ∗dΘi hΘ, ∆˚∆ Θi 044016. F˚[Θ] := K = K (19) h∗dΘ, ∗dΘi hΘ, ∆Θi [5] A. Ashtekar and B. Krishnan. Isolated and Dynamical Horizons and Their Applications. Living Rev. Relativity ac bd ab cd 7 (2004), 10. Cited 8 July 2007. 2 g g − g g ∇a ∇b Θ · ∇c ∇d Θ ǫ [6] J.D. Brown and J.W. York, Jr. Quasilocal energy and IS = conserved charges derived from the gravitational action. a − Θ · ∇ ∇a Θ · ǫ Phys. Rev. D 47 (1993) 1407-1419. IS [7] Olaf Dreyer, B. Krishnan, D. Shoemaker and E. Schnet- within the trial space of potentials given by Eq. (18). ter. Introduction to isolated horizons in numerical relativity. Phys. Rev. D 67 (2003) 024018. Generally, we should expect that the minimizing po- [8] M. Caudill, G.B. Cook, J.D. Grigsby and H.P. Pfeiffer. a tential will generate a vector field ξtrial fairly close to Circular orbits and spin in black-hole initial data. Phys. the minimum-eigenvalue area-preserving vector eigen- Rev. D 74 (2006) 064011. field ˚ξa of Eq. (11). This, in turn, should approxi- [9] G.B. Cook and B.F. Whiting. Approximate Killing Vec- true 2 mate Matzner’s approximate Killing field from Eq. (4). tors on S . E-Print arXiv: 0706.0199v1 [gr-qc]. 2007. To check the approximation, one could imagine increas- [10] R.A. Matzner. Almost Symmetric Spaces and Gravita- J. Math. Phys. 9 ing l until ξa doesn’t vary much with the cut-off. tional Radiation. (1968) 1657-1668. max trial [11] C. Beetle, M. Campanelli, C.O. Lousto and Y. Zlochower. Equivalently, one could use a fairly large cut-off—perhaps In preparation. lmax = 5 would be enough— from the start, and check [12] J.W. York, Jr. Covariant decompositions of symmetric that the amplitudes Θlm are small for large l. If one tensors in the theory of gravitation. Ann. Inst. Henri prefers to approximate the Cook–Whiting approximate Poincaré 21 (1974) 319-332. Killing field, one need only insert a factor of the scalar [13] N. O´ Murchadha and J.W. York, Jr. Initial-value problem curvature R between the gradients in the denominator of of general relativity. I. General forumlation and physical Eq. (19). interpretation. Phys. Rev. D 10 (1974) 428-436. [14] M. Campanelli, C.O. Lousto, Y. Zlochower, B. Krishnan There is one significant issue that has not been ad- and D. Merritt. Spin flips and precession in black-hole- dressed in this discussion. Even once an approximate binary mergers. Phys. Rev. D 75 (2007) 064030. a Killing field ξ has been found from the eigenvalue ap- [15] J. Mathews and R.L. Walker. Mathematical Methods proach, it is still determined only up to overall normal- of Physics. Addison-Wesley, Redwood City, California, ization on S. For a proper rotational Killing field on 1970. a symmetric apparent horizon, the correct normalization [16] The space of potentials in Eq. (18) is usually not orthog- would demand that the affine length of each Killing orbit onal, in the sense of Eq. (10), to the space of constant functions on S. However, the key point is that standard should be 2π. It is not immediately clear what conven- properties of the ordinary spherical harmonics show that tion might be used in the general case, without symme- this space of trial potentials contains no actual constant try, to fix a normalization that goes over to this correct functions. This is why we have taken lmin = 1 in Eq. (18). one in the limit of a symmetric manifold. This issue will Therefore, Eq. (10) maps our space of trial potentials be discussed more thoroughly, in the context of practical faithfully to a space of trial vector fields with the same applications, in a forthcoming paper [11]. dimension, lmax (lmax + 2).