![Arxiv:2007.10920V3 [Math.DG] 7 Feb 2021 Ouino Isenfil Qain,Teamms N H Ce the and Prom Mass Physical Undisputed ADM Their the Besides Equations, Ones](http://data.docslib.org/img/40a8ecc215e9624096f29edd6bb7afc2-1.webp)
arXiv:2007.10920v3 [math.DG] 7 Feb 2021 ouino isenfil qain,teAMms n h ce the and prom mass physical undisputed ADM their the Besides equations, ones. field relevant most Einstein of solution a naypoial a Riemannian flat asymptotically an n .d iahsbe atal upre yCP/rzlgr CNPq/Brazil by supported 00068.01.00/15. partially grant FUNCAP/CNPq/PRONEX been by has supported Lima de L. and mn h ag cl nainsta a eatce ymean by attached be can that invariants scale large the Among pedxE h nqeeso h rebudr M hemisphe CMC boundary free the of uniqueness The References 4.1 Proposition me of the proof E. of The Appendix terms in curvature Gauss-Kronecker D. The Appendix setup variational C. The Appendix B. Appendix pedxA h ag cl sprmti ect n h ma the and deficits isoperimetric scale T large of The proof the proof hemispheres: the isoperimetric A. hemispheres: relative Appendix CMC Large Theo boundary of free proof by the foliation 5. case: The boundaryless the in foliations 4. results The the of statements and 3. Preliminaries 2. Introduction 1. .Amrzhsbe atal upre yCP/rzlgran CNPq/Brazil by supported partially been has Almaraz S. AS ETRO ASADIOEIER NAYPOIAL FLA ASYMPTOTICALLY IN ISOPERIMETRY AND MASS OF CENTER MASS, etro h oito otecne fms ftemnfl a manifold the of mass cas of each center in the Also, to foliation Metzger. the J. of and center Eichmair M. by result enclosed brated large sufficiently all whic for hemispheres, surfaces curvature Schwarzschisoperimetric mean a constant under boundary again free by obtain, additionally we tr where holds ary, also deficit isoperimetric f relative a corresponding of Recov existence conditions. the curvature establish corresponding to the satisfying adapted be functi may implicit Huang the L.-H. how indicate by we positive is mass the that edn lsia eutdet .Hikn et ne Sc a under Next, Huisken. m G. to ADM due the result recover classical curvature a case, tending mean boundaryless total the the In involving deficits boundary). non-compact a with and o fcranioeiercqoinsi h etn fasy of setting the in quotients isoperimetric certain of ior A BSTRACT ervstteitrlybtentems,tecne fmas of center the mass, the between interplay the revisit We . . n 2.22 and 2.9 EGOAMRZADLV OE ELIMA DE LOPES LEVI AND ALMARAZ SERGIO 3 mnfl,vee ste(iesmerc nta aaset data initial (time-symmetric) the as viewed -manifold, .I 1. 3 -MANIFOLDS NTRODUCTION C ONTENTS 1 00721- n AEJBai rn 202.802/2019, grant FAPERJ/Brazil and 309007/2016-0 t oue hsetnigt hsstigacele- a setting this to extending thus volume, ei h rsneo o-opc bound- non-compact a of presence the in ue n 14521-.Bt uhr aebe partially been have authors Both 312485/2018-2. ant rn h asa h smttclmto the of limit asymptotic the as mass the ering poial flat mptotically nmto inee yR eadrefined and Ye R. by pioneered method on endb aitna methods. Hamiltonian by defined s wrshl smttc n assuming and asymptotics hwarzschild rae eew eaetegeometric the relate we here treated e l smttc,aflaina infinity at foliation a asymptotics, ild r hw ob h nqerelative unique the be to shown are h laino egbrodo infinity of neighborhood a of oliation efis hc htteisoperimetric the that check first we s nteaypoi ii,tu ex- thus limit, asymptotic the in ass nne h td fteeinvariants these of study the inence, erm22 22 2.28 heorem e .812 2.18 rem s h roso Theorems of proofs the ss: tro assadota the as out stand mass of nter n h ag cl behav- scale large the and s 3 fHmloinmtosto methods Hamiltonian of s fTerm22 18 2.27 Theorem of mnfls(ohwithout (both -manifolds ncraue31 curvature an e 35 res T 36 33 28 24 of 4 1 2 SERGIO ALMARAZ AND LEVI LOPES DE LIMA also reveals deep connections with several areas of Geometric Analysis, including the Yamabe problem [61, 41, 11], the inverse mean curvature flow [37], the construction of canonical foliations at infinity [38, 63, 44, 33, 35, 49] and the existence of isoperimetric surfaces for sufficiently large enclosed volumes [26]. We recall that in order to have the center of mass well defined, we must supplement the standard ADM decay assumptions with the so-called Regge-Teitelboim condi- tions [53, 9]. Even though some of the results discussed below do not require the fulfillment of these extra conditions, throughout this Introduction we assume that this is always the case in order to simplify the presentation. Motivated by questions related to the Yamabe problem on manifold with boundary [2], a ver- sion of the positive mass theorem has been established for asymptotically flat manifolds carrying a non-compact boundary [3, 5]. Moreover, a notion of center of mass for this class of manifolds has been introduced in [24], again under a suitable Regge-Teitelboim-type condition. The main purpose of this paper is to confirm that, similarly to what happens in the boundaryless case, these asymptotic invariants also play a central role in the investigation of large scale isoperimet- ric properties in the presence of a non-compact boundary. With this goal in mind, we start by slightly broadening our perspective and checking that already in the boundaryless case the ADM mass is recovered as the asymptotic limit of certain isoperimetric deficits involving the enclosed volume, the area and the total mean curvature in various combinations (Theorem 2.9). This provides interesting extensions of a well-known result due to Huisken [36], where the area/volume case is dealt with. Moreover, it clearly suggests the existence of stable foliations in a neighborhood of infinity satisfying the corresponding curvature conditions, which besides the mean curvature should additionally involve (a modified version of) the Gauss-Kronecker curvature (the product of the principal curvatures). Under a Schwarzschild- type asymptotics, and assuming as usual that the ADM mass is positive, we prove the existence of the foliations by adapting the well-known implicit function method pioneered by Ye [63] and refined by Huang [33] (Theorem 2.18). Their approach is made feasible here by means of a simple calculation expressing the Gauss-Kronecker curvature of large coordinate spheres in terms of the corresponding mean curvature up to a remainder decaying fast enough (Proposition 3.3). This is a quite direct consequence of the existence of an almost conformal vector field at infinity and allows us to rely on the computations for the mean curvature in [33, 34]. As a by-product of this procedure we are able to check in each case that the geometric center of the foliation and the center of mass of the manifold always remain at a finite distance from each other and even coincide for certain values of the parameters defining the asymptotics. Besides their intrinsic interest, the existence of the foliations suggest that their leaves may be realized as isoperimetric surfaces for the corresponding isoperimetric problems, which involve minimizing the total mean curvature for large prescribed values of the volume or area, at least if some convexity assumption on the competing surfaces is imposed. Notice that this is in alignment with the well-known fact that round spheres in R3 constitute global minimizers to these problems among convex surfaces [60, 31, 16]; see also Remark 2.20 for more on this point. Our next goal is to confirm that some of the classical results referred to above may be suitably extended to the class of asymptotically flat manifolds introduced in [3]. We first prove that in the presence of a non-compact boundary the relative isoperimetric deficit involving the area of coordinate hemispheres and the volume they enclose jointly with the boundary also recovers the mass in the asymptotic limit (Theorem 2.22). As before, this preliminary remark is accompanied by a result ensuring, again under a Schwarzschild asymptotics and assuming that the mass is pos- itive, that a neighborhood of infinity is foliated by stable free boundary hemispheres of constant mean curvature (CMC) and moreover that the geometric center of mass of this foliation coincides with the center of mass as defined by Hamiltonian methods (Theorem 2.27). We may view these MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 3 results as the large scale analogues of the main theorems in [28, 46], even though the technical de- tails are quite distinct in nature. A key step here is the establishment of a certain identity relating the center of mass to the integral over large coordinate hemispheres of the higher order terms in the expansion of the corresponding mean curvature against the asymptotic coordinates along the non-compact boundary (Proposition 4.1). In the boundaryless case, this kind of identity was first proved in [33, 34] by means of a quite delicate density argument. Subsequently, an elementary proof appeared in [27] and we succeed in checking that this latter reasoning adapts well in the presence of a non-compact boundary. As in these works, the identity is used here to remove the obstruction to the invertibility of the relevant linearized operator coming from the invariance of the mean curvature under translational isometries preserving the asymptotic boundary, besides playing a crucial role in checking that the geometric and Hamiltonian centers of mass coincide. Moreover, it provides an alternate expression for the center of mass as an asymptotic integral involving the mean curvature of large coordinate hemispheres (Corollary D.2). Finally, we com- plete our analysis of the large scale geometry of this class of manifolds by checking that these free boundary CMC hemispheres constitute the unique relative isoperimetric surfaces for sufficiently large values of the enclosed volume (Theorem 2.28). This extends to our setting a previous result by Eichmair-Metzger in the boundaryless case [26]; see also [47]. In order to keep to a minimum the technical features of the exposition, we work here under a Schwarzschild asymptotics whenever needed. In particular, we are usually quite generous when imposing orders of decay rates for the asymptotics of geometric quantities. Also, we have chosen to avoid the consideration of inner minimal horizons (black holes). Moreover, we restrict ourselves to the time-symmetric case, thus bypassing the complications coming from the extrinsic geometry of the given initial data set. As a consequence, we are unable to consider here the rather appealing question of determining how the asymptotic quantities vary as the initial data set evolves in time under the field equations. Nevertheless, we believe that the results established here under these mildly restrictive assumptions may be suitably extended to larger classes of asymptotically flat manifolds, in the line of [35, 44, 48, 49, 26, 19, 39, 13] for instance. Another topic we refrain from discussing here in detail is the uniqueness of the leaves as solutions of the corresponding variational problems. However, we remark that in the specific case of the free boundary CMC hemispheres in Theorem 2.27, the appropriate uniqueness is easily obtained by adapting an argument in [38, Section 4]; see Appendix E. We note that this uniqueness is crucial when identifying those hemispheres to the relative isoperimetric hemispheres in Theorem 2.28 and it eventually guarantees that they are unique in the class of relative isoperimetric surfaces enclosing large volumes. Finally, it would be interesting to investigate how our constructions fit into the quasi-local approach to conserved quantities summarized in [17]. We hope to address some of these questions elsewhere. This paper is organized as follows. In section 2, after a brief motivation intended to illustrate the local interplay between isoperimetric quotients and the scalar curvature, we provide precise statements of all the results mentioned above. The arguments leading to the existence of the stable foliations are presented in Sections 3 and 4. The proof of Theorem 2.28, which provides the relative isoperimetric regions in the presence of a boundary, is explained in Section 5. In order not to interrupt the exposition in the bulk of the paper, we defer to the appendices the proofs of a few technical results, including a discussion of the isoperimetric variational theories involved, specially in regard to the corresponding stability criteria. These appendices are also used to introduce much of the notation used in the paper. Finally, we also intersperse along the text a few interesting problems in this area of research.
Acknowledgments. We thank A. Freitas, E. Lima, J.F. Montenegro and S. Nardulli for conversa- tions at an early stage of this project. 4 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
2. PRELIMINARIES AND STATEMENTS OF THE RESULTS For the sake of motivation, we start by considering an arbitrary Riemmanian 3-manifold (M,g). We fix q M and introduce normal coordinates z = (z1,z2,z3) around q in the usual way. If ∈ M r = z is the geodesic distance to q, let Br (q) be the geodesic ball of radius r centered at q and M | | M M Sr (q)= ∂Br (q) the corresponding geodesic sphere. If N is the outward unit normal to Sr (q), let Wr = N be the shape operator of Sr(q), where is the Levi-Civita connection. For each ∇ ∇ k =1, 2 we may consider the curvature integral
M;k M (2.1) Qr (q)= σ2−k(Wr)dSr (q), M ZSr (q) where σi(Wr) is the elementary symmetric function of degree i in the eingenvalues of Wr (the principal curvatures κ1,κ2). Thus, σ0 = 1, σ1 = κ1 + κ2, the mean curvature, also denoted here by H, and σ2 = κ1κ2, the Gauss-Kronecker curvature, also denoted here by K. Finally, we set M;3 M Qr (q) = volg(Br (q)) by convention. For 1 k
M;2 d M;3 M;1 d M;2 Qr (q)= Q (q),Qr (q)= Q (q), dr ρ dr ρ we see that M;2 R Qr (q) g(q) r2 r4 R3;2 =1 + O( ), Qr (~0) − 18 and M;1 R Qr (q) g(q) r2 r4 R3;1 =1 + O( ). Qr (~0) − 9 From these, the expansions in (2.3) follow easily.
M;h,k h,k This result says that the local behavior of the relative isoperimetric quotients Ir (q)/I is completely determined by the sign of Rg(q). We note however that in general this isoperimetric comparison result only holds true at small scales since the argument depends on the correspond- ing asymptotic expansions as r 0. Our first remark here is that large scale versions of this → MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 5 principle hold true in the setting of asymptotically flat manifolds with non-negative scalar curva- ture and, as we shall see, this is closely related to the positive mass theorem in General Relativity. In the case (h, k)=(3, 2), this was first observed by Huisken [36], as we now pass to describe. Definition 2.2. A manifold (M,g) is asymptotically flat if there exists a compact subset U M ⊂ and a diffeomorhism M U = R3 B (~0) such that in the corresponding asymptotic coordinates \ ∼ \ 1 x = (x1, x2, x3) there holds 1 (2.5) e := g δ = O (r−τ ), τ> − 2 2 and −3−σ (2.6) Rg = O(r ), σ> 0, as r = x + . | |δ → ∞ Note that (2.6) implies R L1(M). Also, for a tensor f = f(x) in the asympotic region, we g ∈ say that f = O (r−τ ) if (∂ f)(x) = O(rτ−|α|) for any multi-index α with 0 α k. k | α | ≤ | |≤ Any manifold as in Definition 2.2 may be viewed as the time-symmetric initial data set of a so- lution to the Einstein field equations whose geometry at spatial infinity is essentially Minkowsk- ian. In particular, the appropriate Hamiltonian version of Noether’s theorem may be employed to attach to the Riemannian manifold (M,g) certain asymptotic invariants capturing common physical quantities associated to the isolated gravitational system modeled by the solution [7, 53, 9, 20, 32, 23]. The most prominent of these invariants is the so-called ADM mass, which is given by
1 U 1 x 2 (2.7) mADM = lim ( ,e) dSr . r→+∞ 16π S2 r Z r Here, U(f,e)= f(div e dtr e) i e + tr e df, δ − δ − ∇δ f δ f : R3 R is a smooth function, 1 is the function identically equal to 1, and S2 is the coordinate → r sphere of radius r in the asymptotic region centered at the origin. 2 Let us denote by A(r) (respectively V (r)) the area of Sr (respectively the volume of the compact 2 region enclosed by Sr ). We now recall a classical definition due to Huisken. Definition 2.3. [36] Under the conditions above, we set
;3 2 2 1 3 (2.8) J M , = V (r) A(r) 2 . r A(r) − 6π1/2 Remark 2.4. This may be expressed as
M;3,2 M;3,2 V (r) Ir (2.9) Jr =2 1 3,2 , A(r) − I ! b where A(r)3/2 IM;3,2 = r V (r) is the large scale analogue of the isoperimetricb quotient in (2.2) with (h, k) = (3, 2) and I3,2 = 1/2 R3 M;3,2 6π is the corresponding quotient evaluated on round spheres in ( ,δ). Thus, Jr should be thought of as the asymptotic analogue of the isoperimetric deficit in (2.8) with (h, k)=(3, 2). 6 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
M;3,2 As remarked above, the isoperimetric deficit Jr should somehow be controlled in case the standard dominant energy condition R 0 holds; here we view (M,g) as a time-symmetric ini- g ≥ tial data set propagating in time to a solution of Einstein field equations, so this energy condition is justified on physical grounds. This link between scalar curvature and isoperimetry holds true indeed and the key result goes as follows. Theorem 2.5. [36] One has M;3,2 (2.10) lim Jr = mADM . r→+∞ Notice that this justifies the appearance of the extra factor V (r)/A(r) in the right-hand side of (2.4) as the ADM mass has the dimension of length. Combining this with the positive mass theorem [62] we obtain the following remarkable result. Corollary 2.6. If (M,g) is asymptotically flat with scalar curvature R 0 then for r> 0 large enough, g ≥ IM;3,2 I3,2, r ≤ with the strict inequality holding unless (M,g) = (R3,δ) isometrically. b Proof. We know that m 0 by the positive mass theorem. If m =0 then (M,g) = (R3,δ) ADM ≥ ADM isometrically and the equality holds. Otherwise, mADM > 0 and the strict inequality holds. Remark 2.7. The discussion above leads naturally to a conjectured C0 version of the positive mass theorem. More precisely, the subisoperimetry condition in (2.4) with (h, k)= (3, 2) may be inter- preted as the statement that R (q) 0 for a metric g which is only assumed to be C0. This led g ≥ Huisken to conjecture that the validity of this subisoperimetry condition at any q M implies ∈ that the mass of an asymptotically flat manifold (M,g) is nonnegative, where g is of class C0, with the equality holding if and only if (M,g) = (R3,δ) isometrically. Of course, here the mass of (M,g) is defined by the limit in the left-hand side of (2.10), whenever it exists. To our knowledge, this conjecture, which constitutes a far-reaching generalization of the standard positive mass the- orem in [62], is still wide open. In any case, we observe that a similar question may be formulated in the presence of a non-compact boundary; see Remark 2.24 below. A proof of Theorem 2.5 appears in [29] and for our purposes a key observation is that a simple variation of their computations, which we reproduce in Appendix A below, proves that this re- markable connection between isoperimetry and mass also holds true for the other isoperimetric quotients involving the total mean curvature
2 M(r) := HdSr 2 ZSr 2 of the coordinate sphere Sr . They correspond to the cases (h, k)=(3, 1) and (h, k)=(2, 1) in (2.3) and are defined as follows. Definition 2.8. Under the conditions above, set M;3,1 M;3,1 4 1 3 4 V (r) Ir Jr = V (r) 7 2 M(r) = 1 3,1 , 3rM(r) − 3 2 π 3 rM(r) − I ! · · b and M;2,1 M;2,1 1 1 2 A(r) Ir Jr = A(r) M(r) = 1 2,1 , M(r) − 16π M(r) − I ! b where M(r)3 M(r)2 IM;3,1 = , IM;2,1 = , I3,1 =3 27 π2, I2,1 = 16π. r V (r) r A(r) · · b b MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 7
Theorem 2.9. If (M,g) is asymptotically flat then M;3,1 M;2,1 (2.11) lim J = mADM , lim J = mADM . r→+∞ r r→+∞ r Again, if combined with the positive mass theorem, this result has the following nice conse- quence. Corollary 2.10. If (M,g) is asymptotically flat with scalar curvature R 0 then for all r > 0 large g ≥ enough, we have the inequalities IM;3,1 I3,1, IM;2,1 I2,1, r ≤ r ≤ with the strict inequality holding in each case unless (M,g) = (R3,δ) isometrically. b b We present the proof of Theorem 2.9 in Appendix A. Remark 2.11. The subisoperimetry condition in (2.4) with either (h, k) = (3, 1) or (h, k) = (2, 1) may be interpreted as the statement that R (q) 0 for a metric g which is only assumed to be g ≥ C1. Thus, similarly to the discussion in Remark 2.7, we may ask whether the validity of any of these subisoperimetry conditions everywhere on an asymptotically flat manifold endowed with a C1 metric implies that the associated mass in nonnegative, with the rigidity statement holding as well. Here, the mass should be defined by the corresponding limit in the left-hand sides of (2.11), whenever it exists. It would be interesting to examine this question in light of recent developments on the positive mass theorem for metrics with low regularity; see [40] and the references therein. We now recall the connection between isoperimetry and another basic invariant of asymptoti- cally flat 3-manifolds, namely, the center of mass [53, 9]. In the following, given a tensor f = f(x) on the asymptotic region, we set f (1)(x) = (f(x)+ f( x))/2 and f (−1) = f f (1), so that f (1) and (−1) − (1) − −τ f are the even and odd parts of f, respectively. Also, we say that f = Ok(r ) (respectively, (−1) −τ ((−1)|α|) −τ−|α| ((−1)1+|α|) −τ−|α| f = Ok(r )) if (∂αf) (x)= O(r ) (respectively, (∂αf) (x)= O(r )) for any multi-index α with 0 α k. b ≤ | |≤ b Definition 2.12. We say that an asymptotically flat 3-manifold as in Definition 2.2 satisfies the Regge-Teitelboim (RT) conditions if there holds 1 (2.12) g(−1)(x)= O (r−1−τ ), τ> , 2 2 and b R(−1) −4−σ (2.13) g (x)= O(r ), σ> 0.
(−1) 1 (1) Note that (2.13) guarantees that each x Rg L (M) for i =1, 2, 3. Also, it is clear that x Rg i b ∈ i has the property that its integral over the region enclosed by two coordinate spheres vanishes. So, we may use the method in [45], with xi as a static potential, to ensure that to any manifold as in Definition 2.12 with m =0 we may attach a (Hamiltonian) center of mass defined by ADM 6 1 U x 2 (2.14) i = lim (xi,e) dSr , i =1, 2, 3. C r→+∞ 16πmADM S2 r Z r Example 2.13. Recall that the Schwarzschild metric on R3 c , c R3, is given by \{ } ∈ m 4 (2.15) g (x)= 1+ δ, m,c 2 x c | − |δ 8 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
R where m . We denote gm = gm,~0 for simplicity. Clearly, gm,c is asymptotically flat and since R ∈ gm,c = 0, it also satisfies the RT conditions. The associated asymptotic invariants may be easily determined if we expand (2.15) as r = x + : | |δ → ∞ 2m 2mc x 3m2 (2.16) g (x)= 1+ + · + + O(r−3) δ, m,c r r3 2r2 where the dot is the Euclidean inner product. A direct computation then gives mADM = m and = c. C This example indicates that the mass and the center of mass are somehow captured by, respec- tively, the first order term and the odd part of the second order term in the asymptotic expansion of the metric in r−1. This motivates the consideration of the following important class of exam- ples. Definition 2.14. We say that an asymptotically flat 3-manifold (M,g) (with τ =1) is asymptotically Schwarzschild (aS) if in the asymptotic region there holds 2m (2.17) g 1+ δ = O (r−2). − r 3 In general, an aS manifold may not satisfy the RT conditions as (2.13) might be violated. In this case, we would still have m = m but if m = 0 the limit in (2.14) might not exist. In ADM 6 fact, convergence takes place if and only if each xiRg is integrable [15]. Examples where the limit defining the center of mass diverges may be found in [12]. Definition 2.15. If an aS manifold satisfies the RT conditions we say that it is an aSRT manifold. Motivated by (2.16), we also consider a more flexible kind of asymptotics. Definition 2.16. We say that an asymptotically flat 3-manifold (M,g) (with τ = 1) is ǫ-asymptot- ically Schwarzschild (ǫ-aS) if in the asymptotic region there holds 2m c x 1 (2.18) g f γ1,γ2 δ = O (r−2−ǫ), f γ1,γ2 =1+ + γ · + γ − m,c 3 m,c r 1 r3 2 r2 with ǫ 0, m =0, γ ,γ R and c R3. ≥ 6 1 2 ∈ ∈ Remark 2.17. Clearly, an ǫ-aS manifold is aS. Also, the case ǫ = 0 does not impose any further conditions on an aS 3-manifold and γ1, γ2 and c are irrelevant in this case. If ǫ > 0 then (M,g) is aSRT with center of mass γ = 1 c. C 2m 2 The most relevant choice in this case is γ1 = 2m and γ2 = 3m /2, when satisfying (2.18) is equivalent to g g = O (r−2−ǫ). − m,c 3 With the appropriate definitions at hand, we now come back to the isoperimetry discussion motivated by Theorems 2.5 and 2.9. These results suggest that for aSRT manifolds with m > 0, large coordinate spheres might be perturbed to yield global solutions to the corresponding isoperimetric problems (here we should assume that the competing surfaces are convex in a suit- able sense so as to make sure that M(r) > 0). In the classical case (h, k) = (3, 2), this has been established in [26]. Very likely, a similar result should hold in the remaining cases treated in Theorem 2.9 and we hope to address this issue elsewhere. In any case, a first check toward this goal is to show that under these same asymptotic condi- tions a neighborhood of infinity can be foliated by surfaces satisfying the corresponding curvature MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 9 conditions and moreover that the geometric center of the foliations relate to the center of mass of (M,g) defined by Hamiltonian methods as in (2.14). In the case (h, k) = (3, 2), the existence of a canonical foliation by stable CMC spheres and the identification of the geometric center of this foliation with the Hamiltonian center of mass has been first established by Huisken and Yau [38] and then investigated further by many authors under varying asymptotic assumptions; see for instance [63, 52, 22, 44, 33, 35, 27, 19, 48, 49, 13]. Here we rely on the implicit function approach developed in [63, 33, 35, 34] to treat the cases (h, k)=(3, 1) and (h, k)=(2, 1) as follows. Theorem 2.18. Let (M,g) be an aSRT 3-manifold with positive ADM mass. Then there exists a neigh- borhood of infinity which is foliated by strictly stable spheres satisfying any of the curvature conditions 1 (2.19) K := K Ric (ν,ν) = const., − 2 g or e (2.20) K = γH. Here, K is the Gauss-Kronecker curvature of the leaf, ν is its outward unit normal and γ is a suitably chosen e Lagrangian multiplier (varying with the leaf). Also, the geometric center of any of these foliations CF remains at a finite distance from the Hamiltonian center of mass of (M,g). Finally, if we further assume C that the manifold is ǫ-aS with ǫ> 0 then = . CF C The curvature conditions (2.19) and (2.20) express the fact that the surfaces are critical configu- rations for the corresponding isoperimetric quotients, i.e., they extremize the total mean curvature under a volume or area constraint, respectively. The stability statement should be understood in this variational sense; see Appendix B for a discussion on this issue. The proof of Theorem 2.18 is presented in Section 3. Remark 2.19. It is not clear whether the rather unsatisfactory assumption ǫ > 0 in the last state- ment of Theorem 2.18 is an artifact of our method of proof or an intrinsic feature of the invariants involved. Remark 2.20. It is well-known that in Euclidean space (R3,δ) the isoperimetric quotients in (2.8) with (h, k) = (3, 1) or (h, k) = (2, 1) are minimized at round spheres if some convexity assump- tion is imposed on the competing surfaces. In the convex case (K 0) this follows from classical ≥ Brunn-Minkowski theory [60]. The result holds more generally if the competing surfaces are as- sumed to be mean convex (H 0) and star-shaped [31] and it is conjectured that star-shapedness ≥ may be dispensed with; see [16] for a survey of recent contributions in this direction. The problem of extending these isoperimetric inequalities (and their analogues involving quermassintegrals of higher order) to other geometries, again under suitable convexity requirements, is quite challeng- ing from a technical viewpoint. The class of aSRT 3-manifolds seems to be a natural choice for starting this research program, as it is reasonable to conjecture that the leaves of the foliations in Theorem 2.18 constitute global solutions to the corresponding isoperimetric problems for large values of the area or volume. See [25] for a possible approach to a version of this problem in space forms using methods from Integral Geometry. We now turn our attention to asymptotically flat 3-manifolds carrying a non-compact bound- R3 R3 aryas in [3]. We set + = x ; x3 0 , the Euclidean half-space with the standard flat metric + { ∈ ≥ } δ = δ R3 . | + Definition 2.21. ([3]) A 3-manifold (M,g) is asymptotically flat with a non-compact boundary Σ if there exists a compact subset V M and a diffeomorhism M V = R3 B (~0) such that in the ⊂ \ ∼ +\ 1 10 SERGIO ALMARAZ AND LEVI LOPES DE LIMA corresponding asymptotic coordinates x = (x1, x2, x3) there holds 1 (2.21) e+ := g δ = O (r−τ ), τ> , − + 3 2 and −3−σ −2−σ (2.22) Rg = O(r ), HΣ = O(r ), σ> 0, as r = x + . Here, H is the mean curvature of Σ. | |δ → ∞ Σ Note that (2.22) implies that R L1(M) and H L1(Σ). g ∈ Σ ∈ In this setting, the asymptotic invariant corresponding to (2.7) has been defined in [3] and is given by
m 1 U 1 + x 2 + x 1 (2.23) = lim ( ,e ) dSr,+ e , ϑ dSr , r→+∞ 16π 2 r − 1 r Sr,+ Sr ! Z Z 2 R3 where Sr,+ is the coordinate hemisphere of radius r, centered at the origin of , in the asymptotic region, S1 = ∂S2 Σ and ϑ is the outward unit co-normal vector field to S2 along S1 (with r r,+ ⊂ r,+ r respect to the flat metric δ+). In order to introduce the corresponding relative isoperimetric deficit 2 we denote by (r) (respectively (r)) the area of Sr,+ (respectively, the volume of the compact A 2 V region enclosed by Sr,+ and Σ). Similarly to (2.8) we may consider M;3,2 ;3 2 1 1 3 (r) r M , = (r) (r) 2 = V 1 I , Jr (r) V − 3 21/2π1/2 A (r) − 3,2 A · A I where 3 (r) 2 M;3,2 = A Ir (r) V and 3,2 =3 21/2π1/2 is the relative isoperimetric quotient of a hemisphere centered at a point in R2 I R3 · = ∂ +. Similarly to Theorem 2.5 we now have Theorem 2.22. Under the conditions above, (2.24) lim M;3,2 = m. r→+∞ Jr This result, whose proof is postponed to Appendix A, has the following notable consequence. Theorem 2.23. Let (M,g) be asymptotically flat with a non-compact boundary Σ as above. Assume further that R 0 everywhere and H 0 along the boundary. Then for all r> 0 large enough, g ≥ Σ ≥ M;3,2 3,2, Ir ≤ I R3 + with the strict inequality holding unless (M,g) = ( +,δ ) isometrically. Proof. Apply the positive mass theorem in [3]. Remark 2.24. Pick any smooth Riemannian 3-manifold (M,g) with boundary ∂M and fix q ∈ ∂M. After introducing Fermi coordinates around q, we may consider the coordinate hemisphere M r Sr,+(q) of small radius > 0 centered at q. By using the calculations in [28, 46] we may easily check that, as r 0, → M,∂M;3,2 r (q) 1 I = c+H (q)r + O(r2). − 3,2 ∂M Here, c+ > 0 is a universal constant andI r 3/2 M,∂M;3,2 ( ) r (q)= A , I (r) V MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 11
M,∂M where (r) and (r) are respectively the area of Sr (q) and the volume it encloses jointly with A V ,+ ∂M, and H∂M (q) is the mean curvature of ∂M at q. Thus, if the metric g is assumed to be merely C0, we may interpret the (boundary) subisoperimetry condition
M,∂M;3,2 3,2 (2.25) r (q) , I ≤ I for all r > 0 small enough, as the statement that H (q) 0 in this weak sense. Thus, in the ∂M ≥ spirit of Remark 2.7 above, we are led to conjecture that for an asymptotically flat 3-manifold with a non-compact boundary endowed with a C0 metric, the fulfillments of the subisoperimetry conditions (2.4) for q in the interior and (2.25) for q on the boundary imply that the mass of the R3 + manifold is nonnegative, with the equality taking place if only if it is isometric to ( +,δ ). Of course, here the mass should be defined by the limit in the left-hand side of (2.24), whenever it exists. Presumably, after a doubling across the boundary, this turns out to be equivalent to the original conjecture in Remark 2.7. In any case, this would lead to a far-reaching generalization of the main result in [3]. In order to have a well defined center of mass in the setting of Definition 2.21, we need the analogue of the RT conditions (2.12) and (2.13). Here, given a tensor f = f(x) in the asymptotic (1′) ′ ′ ′ (−1′) (1′) region, we set f (x) = (f(x , x3)+f( x , x3))/2, with x = (x1, x2), and f = f f . Since, −′ 2 − ′ 2 in this region, x3 is an even function of x which vanishes on Σ, namely, x3 = r x , we may ′ ′ − | | think of f (1 ) and f (−1 ) as the even and odd parts of f. In particular, we can mimic the discussion ′ p ′ preceding Definition 2.12 so as to make sense of any of the conditions f (1 ),f (−1 ) O (r−τ ). ∈ k Definition 2.25. We say that an asymptotically flat 3-manifold with a non-compact boundaryb as in Definition 2.21 satisfies the Regge-Teitelboim (RT) conditions if there holds
(−1′) −1−τ (2.26) g (x)= O2(r ), τ> 1/2, and b ′ R(−1′) −4−σ (−1 ) −3−σ (2.27) g (x)= O(r ), HΣ = O(r ), σ> 0.
′ ′ R(−1 ) 1 (−1 ) 1 Note that (2.27) implies that eachb xα g L (M)band each xαHΣ L (Σ), α = 1, 2. (1′) ∈ ∈ Also, xαRg has the property that its integral over the region enclosed by two large coordinate (1′) hemispheres vanishes, and similarly for the integral of xαHΣ over the region in the boundary enclosed by two coordinate circles. Thus, we may use the method in [45], as adapted in [5] and with x as a static potential, to ensure that to any manifold as in Definition 2.25 with m = 0 we α 6 may attach a (Hamiltonian) center of mass by
+ 1 U + x 2 + x 1 (2.28) α = lim (xα,e ) dSr,+ xαe , ϑ dSr , α =1, 2. C r→+∞ 16πm 2 r − 1 r Sr,+ Sr ! Z Z This invariant has been introduced in [24]. + As in the boundaryless case, we may also consider the half-Schwarzschild metric gm = gm R3 ~ . | +\{0} Definition 2.26. An asymptotically flat 3-manifold with a non-compact boudnary (M, g, Σ) is asymptotically half-Schwarzschild (ahS) if a neighborhood of infinity is diffeomorphic to the com- R3 plement of a hemisphere in + so that 2m (2.29) g = 1+ δ+ + p+, p+ = O (r−2). r 3 12 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
+ As usual, if we assume further that each xαRg and xαHΣ are integrable then the RT condi- tions are satisfied and the center of mass + is well defined for ahS manifolds. C Theorem 2.23 suggests that for an ahS manifold with m > 0, large coordinate hemispheres may be perturbed to yield global solutions of the corresponding relative isoperimetric problem, where each competing surface S satisfies ∂S Σ and int S Σ= , with the relevant constrained ⊂ ∩ ∅ volume being the one enclosed by S and Σ. As a first step towards this goal we establish here the following result:
Theorem 2.27. Assume that (M, g, Σ) is an ahS 3-manifold with a non-compact boundary Σ and satis- fying the RT+ conditions. If m = m/2 > 0 then there exists a neighborhood of infinity which is foliated by strictly stable free boundary CMC hemispheres. Moreover, the geometric center of this foliation coincides with the Hamiltonian center of mass + of (M, g, Σ). C The proof of this result is presented in Section 4. Again, the stability statement above should be interpreted in the sense of Appendix B. Our next result solves the relative isoperimetric problem referred to above by extending a celebrated result due to Eichmair-Metzger [26] to our setting. To state it, recall that the relative isoperimetric profile I : [0, + ) [0, + ) of (M, g, Σ) is given by g ∞ → ∞ ∗ o Ig(V ) = inf (∂ Ω,M ), Ω P where M o = M Σ is the interior of M and ∂∗Ω is the relative reduced boundary of a Borel set \ Ω M satisfying: ⊂ o i) vol(Ω) = V ; ii) the relative perimeter P (∂∗Ω,M ) is finite. We stress that in the computation of (∂∗Ω,M o) only that part of the boundary area inside M o P counts, hence the qualification “relative”. A solution to the relative isoperimetric problem is a ∗ o Borel subset Ω such that vol(Ω) = V and (∂ Ω,M )= Ig(V ) for some V > 0. We then say that Ω P o is a (relative) isoperimetric region and its boundary ∂Ω M is a (relative) isoperimetric surface. ∩ Theorem 2.28. Let (M, g, Σ) be as in Theorem 2.27. Then there exists V > 0 such that for any V V 0 ≥ 0 a bounded isoperimetric region with volume V exists. That region has a connected and smooth relative boundary which remains close to a centered coordinate hemisphere. Moreover, it sweeps out the whole manifold as V + . In particular, the corresponding isoperimetric surfaces coincide with the leaves of → ∞ the foliation in Theorem 2.27, thus being unique for each value of the enclosed volume.
This result, whose proof is presented in Section 5, provides a very precise description of the relative isoperimetric profile of (M, g, Σ) as above for all sufficiently large volume values.
3. THEFOLIATIONSINTHEBOUNDARYLESSCASE: THE PROOF OF THEOREM 2.18 The purpose of this section is to explain how the well-known implicit function method pre- sented in [63, 33] may be adapted to prove Theorem 2.18. We consider, in an aSRT manifold as in that theorem, the coordinate sphere S2(a) of radius ρ > 0 centered at some a R3 to be chosen ρ ∈ later. A justifiable assumption here is that a varies in a bounded region and we will take it for granted. In particular, we will sometimes ommit the dependence on a in the estimates below. The following two results play a central role in our argument.
2 Proposition 3.1. [33, Equation (5.1)] In an aS manifold, the mean curvature of Sρ (a) at a point x is 2 4m 9m2 6m(x a) a (3.1) H (x)= + + − · + G (x)+ O(ρ−4), ρ,a ρ − ρ2 ρ3 ρ4 ρ,a MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 13 where 1 p (x) G (x) = p (x)r r r +2 ij r r p (x)r ρ,a 2 ij,k i j k ρ i j − ij,i j p (x) 1 (3.2) ii + p (x)r , − ρ 2 ii,j j with p = g 1+ 2m δ and r = (x a)/ρ. − r − Proposition 3.2. [33, Lemma 5.1] The center of mass of an aSRT manifold satisfies C 2,δ −1 (3.3) (xi ai)Gρ,a(x)dSρ (a)= 8πm i + O(ρ ), i =1, 2, 3, 2 − − C ZSρ (a) 2,δ where dSρ (a) is the area element induced by the flat metric. Let (M,g) be an aSRT 3-manifold as in Theorem 2.18 and assume that it is also ǫ-aS with ǫ 0 ≥ and some fixed c R3 (recall that c is irrelevant for the case ǫ = 0). The key observation now is ∈ 2 that we can express the Gauss-Kronecker curvature Kρ,a of Sρ(a) in terms of the mean curvature Hρ,a up to terms decaying fast enough. This will allow us to make use of Propositions 3.1 and 3.2. Proposition 3.3. There holds 2 2m 9m 3γ2 3m 3γ1 (3.4) 2ρK = H 1 + − + x a x c + O(ρ−2−ǫ) + O(ρ−5). ρ,a ρ,a − ρ 2ρ2 ρ3 · − 2ρ3 · The proof of Proposition 3.3, which relies on the fact that the Schwarzschild space carries a radial conformal vector field, is deferred to Appendix C. Corollary 3.4. There holds 2 6m 26m2 3cγ 12m 3γ (3.5) 2ρK = + − 2 + x a 1 x c + G (x)+ O(ρ−3−ǫ). ρ,a ρ − ρ2 ρ3 ρ4 · − ρ4 · ρ,a Proof. Combinee (3.4), (3.1) and the fact that 2m (3.6) Ric (ν,ν)= + O(ρ−4−ǫ). g − ρ3
2,α 2 To proceed we consider for a function φ C (Sρ (a)) the corresponding normal graphical 2 ∈ surface over Sρ (a): (3.7) S2(a, φ)= x + ρ−θφ(x)ν(x); x S2(a) , θ (0, 1). ρ { ∈ ρ } ∈ −θ 2 By the Taylor’s formula, the modified Gauss-Kronecker curvature Kρ(a,ρ φ) of Sρ (a, φ) ex- pands as 1 e (3.8) K (a,ρ−θφ)= K (a, 0) + dK (a, 0)(ρ−θφ)+ (1 s)d2K (a,sρ−θφ)(ρ−θφ, ρ−θφ)ds. ρ ρ ρ − ρ Z0 e e e 2 e We now observe that Kρ(a, 0) = Kρ,a and dKρ(a, 0) = LSρ(a) is the Jacobi operator appearing in (B.14), whose asymptotic behavior we need to determine. With this goal in mind we introduce 2 local coordinates y1,ye2 on S (a)eand let δρebe the induced Euclidean metric, so that h = g 2 { } ρ |Sρ(a) is given by 2m h = 1+ δ + O(ρ−2). ρ ρ 14 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
Proposition 3.5. One has
1 2 −4 L 2 = ∆ρ + + O(ρ ), Sρ(a) −ρ ρ2 where ∆ρ is the Laplacian with respect to δρ. Proof. With the notation of Appendix B we have
1 √ AC B ΛS2(a) = ∂B det hh ΠA ∂C , ∂A = ∂/∂yA. ρ √det h Since 2m 2m −1 √det h = 1+ det δ + O(ρ−2), hAC = 1+ δAC + O(ρ−2), ρ ρ ρ ρ and p B −1 B −2 (3.9) ΠA = ρ δA + O(ρ ), we compute that
1 −1 BC −1 ΛS2(a) = ∂B ρ det δρ δρ ∂C + O(ρ ) ρ √det h 1 p = ∆ + O(ρ−4). ρ ρ On the other hand, from (3.1) and (3.5), 2 H K = + O(ρ−4). ρ,a ρ,a ρ3 Also, from (3.6), (3.9) and the fact that 1 (3.10) Riemν = Ric (ν,ν)h + O(ρ−4), g 2 g we see that 1 (3.11) ( Ric )(ν,ν) tr (ΠRiemν )= O(ρ−4). 2 ∇ν g − h g The result follows. We now have at our disposal the ingredients needed to prove the existence of a foliation satis- fying (2.19) in Theorem 2.18. Indeed, from (3.5) and (3.8) we see that finding φ so that 1 3m (3.12) K (a,ρ−θφ)= ρ ρ2 − ρ3 is equivalent to solving e 26m2 3γ 12m 3γ (3.13) 2ρ−θ∆ φ(x)= − 2 + x a 1 x c + G (x)+ E (x), (ρ) ρ3 ρ4 · − ρ4 · ρ,a ρ,φ −2 where ∆(ρ) = ∆ρ +2ρ and the remainder term Eρ,φ is controlled as (3.14) E (x)= O ρ−3−θ φ + ρ−3−2θ φ 2 + ρ−1−2θ φ ∂2φ + O(ρ−3−ǫ). ρ,φ | | | | | || | We next pull back this equation under the map F : S2(~0) S2(a), F (r) = a + ρr, so as to 1 → ρ ∗ 2 ~ obtain an equation for ψ = F φ on S1 (0):
(3.15) 2∆(1)ψ(r)= F (r,a,ψ) MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 15 where ∆(1) = ∆1 +2 and 26m2 3γ 12m 3γ F (r,a,ψ)= − 2 + r a 1 r c + ρ2+θF ∗G (r) ρ1−θ ρ1−θ · − ρ1−θ · ρ,a + ρ2+θF ∗E (r)+12mρ−2+θ a 2 3γ ρ−2+θa c. ρ,φ | | − 1 · We note that the primary obstruction to solving (3.15) is the fact that the operator ∆ : C2,α(S2(~0)) Cα(S2(~0)) (1) 1 → 1 has a nontrivial cokernel generated by the functions ri, i = 1, 2, 3. Thus, in order to remove this obstruction, we shall calculate r F r 2,δ ~ i ( ,a,ψ) dS1 (0). 2 ~ ZS1 (0) Note that by symmetry 2 r 26m 3γ2 −2+θ 2 −2+θ 2,δ ~ i 1−−θ + 12mρ a 3γ1ρ a c dS1 (0)=0. ~ ρ | | − · ZS1(0) Using (3.3) we easily see that
r 12m r 3γ1 r 2+θ ∗ r 2,δ ~ i 1−θ a 1−θ c + ρ F Gρ,a( ) dS1 (0) 2 ~ ρ · − ρ · ZS1 (0) 16πm γ = a 1 c Ci + O(ρθ−2) ρ1−θ i − 4m i − 2 16πm γ = a 1 c + O(ρθ−2), ρ1−θ i − 2m i where in the last step we used that = γ1 c, according to Remark 2.17. The remaining integral C 2m r 2+θ ∗ r 2,δ ~ iρ F Eρ,φ( )dS1 (0) ~ ZS1(0) may be estimated so as to yield
r F r 2,δ ~ 16πm γ1 1 (3.16) i ( ,a,ψ)dS1 (0) = 1−θ ai ci + 1−θ Ei, S2(~0) ρ − 2m ρ Z 1 where b −1 −θ −ǫ (3.17) E = O((ρ + ρ ) ψ 2 )+O(ρ ). i k kC 3 Thus, for each ρ large enough we may choose aρ R such that b ∈ γ 1 (3.18) (a ) = 1 c E , ρ i 2m i − 16πm i so as to have b r F r 2,δ ~ (3.19) i ( ,aρ, ψ) dS1 (0)=0, i =1, 2, 3, 2 ~ ZS1 (0) for any ψ with ψ 2 bounded. k kC With the obstruction so removed we may now use a standard fixed point argument to check that (3.13) has a unique solution φρ for all such ρ. More precisely, from (3.19) we see that F (r,aρ, ψ) lies in Ran ∆ if ψ 2,α 1. Therefore, we may uniquely solve (1) k kC ≤ (3.20) 2∆(1)ψ = F (r,aρ, ψ),
e 16 SERGIO ALMARAZ AND LEVI LOPES DE LIMA for ψ C2,α(S2(~0)) (ker ∆ )⊥ satisfying ∈ 1 ∩ (1) ′ max{−θ,θ−1,θ−1−ǫ} ′ max{−θ,θ−1} (3.21) ψ 2,α C F (r,a , ψ) 0,α C ρ = C ρ , e k kC ≤ k ρ kC ≤ and this is 1 if ρ is large enough. Thus, the map ψ ψ has a fixed point which yields a solution ≤ e 7→ ψρ of (3.15) and hence a solution φρ of (3.13). In particular, the graphical surface associated to 2 φρ, denoted simply Sρ (φρ), has constant modified Gauss-Kronekere curvature given by the right- hand side of (3.12). Moreover, if we choose θ = 1/2 in (3.21) then this analysis guarantees that φ C2,α(S2(a )) satisfies ρ ∈ ρ ρ (3.22) ρ|I| ∂ φ + ρ2+α[∂ φ ] C′′ρ1/2, | I ρ| I ρ α ≤ |XI|≤2 |XI|=2 2 −1/2 so the corresponding graphical surface Sρ (φρ), which actually involves the function ρ φρ, re- mains at a fixed distance of S2(a ) while becoming rounder as ρ + . The geometric center of ρ ρ → ∞ mass of the foliation is given by 2,δ 2 z dS (φρ) Sρ(φρ) ρ F = lim , ρ→+∞ 2,δ C R 2 dSρ (φρ) Sρ (φρ) where z is the position vector. We easily see fromR (3.17), (3.18) and (3.22) that either F remains Cγ1 at a finite distance from if ǫ = 0 or there holds F = if ǫ > 0 (recall that = 2m c). Also, C 2C C C each such surface may be viewed as a graph over Sρ (c). For simplicity of notation, we still denote 2 such a surface by Sρ (φρ). Finally, for further use we note that by (3.22) we may determine the 2 asymptotic expansions of the geometric invariants of Sρ (φρ). The result is K = ρ−2 4mρ−3 + O(ρ−4), H = 2ρ−1− 4mρ−2 + O(ρ−3), 2 −2 − −3 −4 (3.23) W = 2ρ 8mρ + O(ρ ), | | −−3 −4 Ric(ν,ν) = 2mρ + O(ρ ), −−2 −3 −4 KG = ρ 2mρ + O(ρ ). − Here, K is the Gaussian curvature. G It remains to check that for ρ large enough, the family of surfaces S2(φ ), ρ ρ , defines 0 ρ ρ ≥ 0 a foliation whose leaves are strictly stable in the appropriate sense. We first tackle the stability issue. 2 Theorem 3.6. If m> 0 then Sρ (φρ) is strictly stable for all ρ large enough. Proof. According to Proposition B.3, we must estimate from below the quadratic form 1 V (f)= Π f, f f 2 HK + ( Ric )(ν,ν)+tr (ΠRiemν ) dS, h ∇S ∇S i− 2 ∇ν g S g ZS where f (S) and here we set S = S2(φ ) for simplicity. From (3.23) we have ∈ G ρ ρ 2 12m HK = + + O(ρ−5). − −ρ3 ρ4 Also, from (3.6), (3.9) and (3.10), we may improve (3.11) to 1 m ( Ric )(ν,ν) tr (ΠRiemν )= + O(ρ−5). −2 ∇ν g − S g −ρ4 Thus, 2 11m (3.24) V (f) Π f, f dS + + + O(ρ−5) f 2dS. ≥ h ∇S ∇S i −ρ3 ρ4 ZS ZS MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 17
2 We now observe that the Newton tensor of S = Sρ (φρ) satisfies 1 2m (3.25) Π= I + O(ρ−3). ρ − ρ2 On the other hand, by the well-known Lichnerowicz eigenvalue bound,
f 2dS 2 inf K f 2dS, f (S). |∇S | ≥ G ∈ G ZS ZS where by (3.23), 1 2m inf K Cρ−4, C> 0. G ≥ ρ2 − ρ3 − We thus conclude that 3m V (f) Cρ−5 f 2dS, ≥ ρ4 − ZS and the result follows. We now check that the surfaces define a foliation. Since the argument, as explained for instance in [34], is well-known by now and it may be easily adapted to our setting, here we merely sketch the proof.
2 Proposition 3.7. Let ζ0 < ζ1 be the first two (unconstrained) eigenvalues of LSρ(φρ). Then 2 9m m ζ = + + O(ρ−5), ζ Cρ−5. 0 −ρ3 ρ4 1 ≥ ρ4 − 2,α 2 α 2 In particular, L 2 : C (S (φρ)) C (S (φρ)) is invertible for all ρ large enough. Sρ(φρ) ρ → ρ The proof of this statement is basically a refinement of the stability analysis above. In any case, for any such fixed ρ large enough, it implies the existence of a unique f C2,α(S2 (φ )) such 0 ρ0 ∈ ρ0 ρ0 2 that LS (φ )fρ0 =1. ρ0 ρ0 2 Proposition 3.8. The function fρ0 vanishes nowhere on Sρ0 (φρ0 ). Proof. This follows immediately from the estimate −1 sup fρ0 fρ0 Cρ0 fρ0 , 2 | − |≤ | | Sρ0 (φρ0 ) 2 where C > 0 is a constant depending only g and the overline stands for the average over Sρ0 (φρ0 ). The method of proof, which is explained in [34, Section 5.2], makes use of Nash-Moser iteration and equally applies here due to the fact that LS2 (φ ) is elliptic of divergence type; see Appendix ρ0 ρ0 B. 2 Now, we may organize the graphical surfaces Sρ0 (φρ) with ρ close to ρ0 in a smooth deforma- tion F : (K ǫ, K + ǫ) S2 (φ ) M, ǫ> 0, ρ0 − ρ0 × ρ0 ρ0 → 2 where Kρ is the modified Gauss-Kronecker curvature of S (φρ ) and F (K, ) has constant mod- 0 e e ρ0 0 · ified Gauss-Kronecker curvature equal to K. Clearly, e e 2 e F (K, x) = expS (φρ )(fK ν), e ρ0 0 2 for some function f e on Sρ (φρ0 ) with f e 0. Let K 0 e Kρ0 ≡ e e ∂fK ∂F e ee e e e f0 := = = = ,ν . ∂K |K Kρ0 ∂K |K Kρ0 e e e e 18 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
By (B.13), d 2 LS (φ )f0 = Kρ0 + k k=0 =1, ρ0 ρ0 dk | so that f0 = fρ0 by uniqueness. In particular,e fe0 never vanishes and we may use the inverse 2 function theorem to conclude that F is a diffeomorphism onto a small neighborhood of Sρ0 (φρ0 ) in M. Thise shows that the surfaces define a foliatione and completes the proof of the first part of Theorem 2.18. We now sketch the proof of the existence of a stable foliation satisfying the curvature condition in (2.20), whose leaves correspond to surfaces extremizing the total mean curvature under an area constraint by Appendix B. The Lagrange multiplier γ is determined by observing that (3.4) leads to
Kρ,a 1 m −3 = 2 + O(ρ ). Hρ,a 2ρ − 2ρ e With this choice of γ = O(ρ−1) > 0, it then follows from (3.25), (B.13) and (B.9) that the quadratic form associated to the linearization of (2.20) at S = Sρ(a) has 1 3m (Π γI) f, f dS = + O(ρ−3) f 2dS h − ∇S ∇S i 2ρ − 2ρ2 |∇S | ZS ZS as its principal part. Hence, the linearization is selfadjoint and elliptic. In fact, a computation shows that the rescaled linearization on S2(~0) is 2(1 γ)∆ ; compare with the left-hand side of 1 − (1) (3.15). Thus, we are in a position to run the implicit function method above in order to construct a graphical surface over Sρ(a) satisfying (2.20). As in the preceding case, this step only uses that m = 0. If m > 0 then a further analysis shows that these graphical surfaces comprise a foliation 6 of a neighborhood of infinity whose leaves are strictly stable in the appropriate sense. Moreover, the geometric center of this foliation remains at a finite distance from the Hamiltonian center of mass. In this way, the proof of Theorem 2.18 is completed.
4. THE FOLIATION BY FREE BOUNDARY CMC HEMISPHERES: THE PROOF OF THEOREM 2.27 In this section we present the proof of Theorem 2.27. In any ahS manifold as in that theorem we consider the coordinate hemisphere S2 (b) centered at some b R2 and use the notation of ρ,+ ∈ Appendix B. As in Proposition 3.1, we compute the mean curvature Hρ,+,b of this hemisphere to obtain 2 4m 9m2 6m(x b) b (4.1) H (x)= + + − · + G (x)+ O(ρ−4), ρ,+,b ρ − ρ2 ρ3 ρ4 ρ.b where G (x) is as in (3.2) with r = (x b)/ρ. We also need the analogue of Proposition 3.2, which ρ.b − goes as follows. Proposition 4.1. If (M, g, Σ) is an ahS manifold meeting the RT+ condition then its center of mass + C satisfies
2,δ+ m + −1 (4.2) (xα bα)Gρ.b(x)dSρ,+ (b)= 8π α + O(ρ ), α =1, 2, S2 (b) − − C Z ρ,+
2,δ+ where dSρ,+ (b) is the area element induced by the flat metric. The proof of this proposition, which adapts an argument first appearing in [27, Appendix F], is presented in Appendix D. MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 19
We now proceed to the proof of Theorem 2.27 via the standard implicit function method [63, 2,α 2 33]. We consider, for a function φ C (Sρ,+(b)) satisfying Neumann boundary condition along 1 2 ∈ 2 Sρ = ∂Sρ,+, the corresponding normal graphical surface over Sρ,+(b): (4.3) S2 (b, φ)= x + ρ−θφ(x)ν(x); x S2 (b) , θ (0, 1). ρ,+ { ∈ ρ,+ } ∈ −θ 2 By the Taylor’s formula, the mean curvature Hρ,+(b,ρ φ) of Sρ,+(b, φ) expands as 1 H (b,ρ−θφ)= H (b, 0)+ dH (b, 0)(ρ−θφ)+ (1 s)d2H (b,sρ−θφ)(ρ−θφ, ρ−θφ)ds. ρ,+ ρ,+ ρ,+ − ρ,+ Z0 We now observe that H (b, 0) = H and dH (a, 0) = L 2 is the Jacobi operator ap- ρ,+ ρ,+,b ρ,+ Sρ,+(b) pearing in (B.7). Thus, finding φ so that 2 4m (4.4) H (b,ρ−θφ)= ρ,+ ρ − ρ2 is equivalent to solving 9m2 6m(x b) b (4.5) ρ−θ∆ φ = + − · + G (x)+ E (x), (ρ) ρ3 ρ4 ρ,b ρ,φ+ −2 + where as usual ∆(ρ) = ∆ρ +2ρ (recall that ∆ρ is the Laplacian with respect to δρ , the induced 2 round metric on Sρ,+) and the remainder Eρ,φ+ has the the same bound as in (3.14). We next pull 2 2 r r back this equation under the map F : S1,+(0) Sρ,+(b), F ( )= b+ρ , so as to obtain an equation ∗ 2 → for ψ = F φ on S1,+(0): 9m2 6mr b (4.6) ∆ ψ = + · + ρ2+θF ∗G (r)+ ρ2+θF ∗E , (1) ρ1−θ ρ1−θ ρ,b ρ,φ+ where ∆(1) = ∆1 +2. We now recall that the operator ∆ : C2,α(S2 (~0)) Cα(S2 (~0)), where the star means (1) ⋆ 1,+ → 1,+ that we impose the Neumann boundary condition, has a nontrivial cokernel generated by the functions rα, α =1, 2. Clearly, this poses an obstruction to solving (4.6). However, using (4.2) we easily calculate that
2 r + m r 9m 6m b 2+θ ∗ r 2,δ ~ 8π + α 1−θ + 1−·θ + ρ F Gρ,b( ) dS1,+ (0) = 1−θ bα α S2 (~0) ρ ρ ρ −C Z 1,+ −1 +O(ρ ψ 2 ), k kC so we end up with
+ m r G r 2,δ ~ 8π + 1 α ( ,b,ψ)dS1,+ (0) = 1−θ (bα α )+ 1−θ Eα, S2 (~0) ρ −C ρ Z 1,+ where G (r,b,ψ) is the right-hand side of (4.6) and b −1 −θ (4.7) E = O((ρ + ρ ) ψ 2 ) α k kC Thus, for each ρ large enough we may choose bρ such that b 1 (4.8) (b ) = + E , ρ α Cα − 8πm α so as to have b r G r 2,δ+ ~ (4.9) α ( ,bρ, ψ) dS1,+ (0)=0, α =1, 2, S2 (~0) Z 1,+ for any ψ with ψ 2 bounded. This eliminates the obstruction mentioned earlier. k kC 20 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
As in the proof of Theorem 2.18 above, we may now use the standard fixed point argument to check that (4.5) has a unique solution φρ for all such ρ. In particular, the graphical surface corresponding to φρ as in (4.3), denoted Sρ,+(φρ), has constant mean curvature given by the right- hand side of (4.4). Also, the Neumann condition imposed on φρ implies that this graphical surface 2 is free boundary. Moreover, an estimate similar to (3.22) also holds true here, so Sρ,+(φρ) remains at a fixed distance of S2 (b ) while becoming rounder as ρ + . ρ,+ ρ → ∞ The geometric center of mass of this family of surfaces, is given by
2,δ+ 2 z dSρ,+ (φρ) + Sρ,+(φρ) = lim + , CH ρ→+∞ 2,δ R 2 dSρ,+ (φρ) Sρ,+(φρ) + R + where z is the position vector. Clearly, H = . Also, each such surface may be viewed as 2 + C C a graph over Sρ ( ). For simplicity we retain the notation and still denote such a surface by 2 C Sρ,+(φρ). It remains to check that this family of free boundary CMC hemispheres defines a fo- liation of a neighborhood of infinity with stable leaves. As usual we first consider the stability issue.
Theorem 4.2. If m = m/2 > 0 then Sρ,+(φρ) is strictly stable for all ρ large enough.
For the proof we first note that the geometric invariants of Sρ,+(φρ) expand as W 2 = 2ρ−2 8mρ−3 + O(ρ−4), | | − (4.10) Ric(ν,ν) = 2mρ−3 + O(ρ−4), K = −ρ−2 2mρ−3 + O(ρ−4). G − We also need the asymptotic expansion of the second fundamental of Σ, the non-compact bound- ary of M. Lemma 4.3. The second fundamental form of Σ satisfies B (4.11) = O(ρ−3), α,β =1, 2. Bαβ Proof. Recall that 2m g = 1+ δ+ + p+, p+ = O(ρ−2). r Outside a compact subset of M, Σ is defined by x3 = 0 and its tangent space is generated by ∂ , ∂ . If η = ηi∂ is the inward unit normal along Σ then { 1 2} i = η, ∂ =Γi η, ∂ =Γ3 g ηi. Bαβ h ∇∂α βi αβh ii αβ 3i Since 1 Γ3 = g33 (g + g g )+ O(ρ−4) αβ 2 α3,β β3,α − αβ,3 1 2m −1 = 1+ g + O(ρ−3), −2 r αβ,3 and
−2 ∂r −3 gαβ,3 = 2mr δαβ + O(ρ ) − ∂x3 = 2mr−3x δ + O(ρ−3) − 3 αβ = O(ρ−3), the result follows. MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 21
By Proposition B.1, the proof of Theorem 4.2 involves estimating from below the quadratic form Q(f)= f 2 W 2 + Ric(ν,ν) f 2 dS κf 2d∂S, f (S), |∇S | − | | − ∈F ZS Z∂S 2 where here we set S = Sρ,+(φρ) for simplicity. We may assume that S f dS = 1, which implies f = O(ρ−1). Hence, using (4.10), R 2 10m (4.12) Q(f)= ( f 2dS κf 2d∂S + + O(ρ−4). |∇S | − − ρ2 ρ3 ZS Z∂S Thus, we are left with the task of estimating from below the quadratic form
Q(f)= ( f 2dS κf 2d∂S, f (S), |∇S | − ∈F ZS Z∂S which is equivalent to estimatingb from below the first eigenvalue λ of the eigenvalue problem
∆Sf = λf in S (4.13) −∂f b = κf on ∂S ∂µ where f (S). Notice that a comparison with the first eigenvalue 2/ρ2 of the Neumann (κ =0) ∈F eigenvalue problem on (S (a),δ+) ֒ (R3 ,δ+) already shows that λ > 0 and provides the ρ,+ ρ → + preliminary but useful estimate λ = O(ρ−2). If f is an eigenfunction of (4.13) with eigenvalue λ = λ, we see that b b (4.14) f 2dS = κf 2d∂S + λ. |∇S | b ZS Z∂S α β −3 2 −4 −4 From (4.11), κ = αβν ν = O(ρ ), so that κf d∂S = O(bρ ) and ∂f/∂µ = O(ρ ), where B ∂S (4.13) was used in the latter step. Thus, f 2dS = O(ρ−2), so that f = O(ρ−2) and, S R|∇S | ∇S moreover, from (4.13) we get ∆ f = O(ρ−3). S R We now apply a well-known integral identity due to Reilly [54]. In our setting (dim S = 2) it simplifies to ∂f (4.15) (∆ f)2 2 f 2 dS = 2 ∆ fd∂S S − |∇S | ∂µ ∂S ZS Z∂S ∂f 2 + H + f 2 d∂S ∂S ∂µ |∇∂S | Z∂S ! + K f 2dS, G|∇S | ZS where H∂S is the mean (in fact, geodesic) curvature of ∂S in S. Since S is free boundary, H∂S = (T,T )= O(ρ−3), where T is a unit tangent vector along ∂S and we used (4.11). Thus, the second B integral in the right-hand side equals
H f 2d∂S = O(ρ−6). ∂S |∇S | Z∂S On the other hand, ∂2f ∆ f = ∆ f = O(ρ−3), ∂S S − ∂µ2 so that the first integral in the right-hand side is also O(ρ−6). Finally, using (4.10) and (4.14),
2 1 2m −4 −4 KG Sf dS 2 3 Cρ λ + O(ρ ) , C> 0, S |∇ | ≥ ρ − ρ − Z b 22 SERGIO ALMARAZ AND LEVI LOPES DE LIMA and combining this with the fact that 2 f 2 (∆ f)2/2 and λ = O(ρ−2) we get from (4.15) that |∇S | ≥ S 2 4m λ λ + C′ρ−6, Cb′ > 0. − ρ2 ρ3 ≥− Since we already know that λb−1b= O(ρ2) is positive, this gives 2 4m b λ C′′ρ−4, C′′ > 0. ≥ ρ2 − ρ3 − Combining this with (4.12) we finallyb have 6m Q(f) C′′ρ−4, ≥ ρ3 − which completes the proof of Theorem 4.2. From this point on, the proof that the family of free boundary CMC hemispheres comprises a foliation follows from a simple variation of the standard argument. Indeed, if χ0 < χ1 are the L first two (unconstrained) eigenvalues of Sρ,+(φρ), a spin-off of the analysis above leads to 2 10m 6m (4.16) χ = + + O(ρ−4), χ Cρ−4. 0 −ρ2 ρ3 1 ≥ ρ3 − In other words, L : C2,α(S (φ )) Cα(S (φ )) is injective, where the bullet indicates Sρ,+(φρ) • ρ,+ ρ → ρ,+ ρ that the boundary condition in (4.13) is imposed. Since it is known that this Jacobi operator is Fredholm of index zero [42, Section 2], we see that it is surjective as well. In particular, there exists f C2,α(S (φ )) such that L f = 1. On the other hand, just like in the dis- ρ ∈ • ρ,+ ρ Sρ,+(φR) ρ cussion after the proof of Proposition 3.8 above, we may realize fρ as the variational function 2 associated to a deformation of Sρ,+(φρ) by the graphical free boundary CMC hemispheres, now parameterized by their mean curvature H (H ǫ,H + ǫ), ǫ > 0. Since the Nash-Moser ∈ ρ − ρ scheme may be implemented to make sure that fρ never vanishes, the standard argument using the inverse function theorem shows that this deformation actually provides a diffeomorphism of (H ǫ,H + ǫ) S2 (φ ) onto a small neighborhood of S2 (φ ) in M. This proves the existence ρ − ρ × ρ,+ ρ ρ,+ ρ of the foliation and completes the proof of Theorem 2.27.
5. LARGE RELATIVE ISOPERIMETRIC HEMISPHERES: THE PROOF OF THEOREM 2.28 Here we follow [26] closely and present the proof of Theorem 2.28. We start by briefly review- ing the argument leading to their main result, which is based on three ingredients: An effective area comparison result for large volume, off-center regions in Schwarzschild • space, which refines Bray’s characterization of isoperimetric regions as being those en- closed by centered spheres (together with the minimal horizon) [10, 21]. This is then transplanted to an effective estimate for large volume, off-center regions in asymptoti- cally Schwarzschild manifolds; see Proposition 3.3 and Theorem 3.4 in [26]. A precise understanding of the behavior of minimizing sequence of regions attaining the • corresponding isoperimetric profile, to the effect that they split as the disjoint union of a (possibly empty) isoperimetric region (for the volume it encloses) that remains at a finite distance of a given point and a coordinate ball of radius r 0 which slides away toward ≥ the asymptotic region. Moreover, if none of these regions degenerate (in particular, r> 0) then the boundary of the isoperimetric region left behind has constant mean curvature 2/r; see [26, Proposition 4.2], which relies on [56, Theorem 2.1]. Existence of a foliation by CMC spheres filling out the asymptotic region as in [38, 63, 33]. • MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 23
The first item above is used to make sure that the boundary of a sufficiently large isoperimet- ric region remains close to a centered coordinate sphere bounding the same volume in the sense that the scale invariant C2-norm of the function describing such large isoperimetric surface as a normal graph over the centered sphere tends to zero as the enclosed volume goes to infinity. Otherwise, by suitably scaling down the region one is able to check that it is off-center, hence not isoperimetric by the effective area estimate, a contradiction; see [26, Theorem 5.1]. With this information at hand, one sees from the second item above that for a large enclosed volume the worst case scenario takes place whenever the runaway ball does not degenerate, for in this case the isoperimetric region attaining this volume splits as the disjoint union of two large balls each roughly with the same radius. Since this configuration is far from being isoperimetric, we get a contradiction. Thus, the runaway ball actually disappears and the isoperimetric region starts fill- ing out the whole manifold as the enclosed volume diverges. Moreover, as its boundary remains close to a centered coordinate sphere, it has to coincide with a leaf of the foliation appearing in the third item. It turns out that all of these ingredients also work fine in our setting. Of course, the existence of the relevant foliation, in our case by free boundary CMC hemispheres, is the content of Theorem 2.27 above. We now discuss the validity of the remaining ones. First, it is clear that in half-Schwarzschild space, the region bounded by the minimal horizon r = m/2 and a coordinate sphere of radius r >m/2 is the only one attaining the relative isoperi- metric profile for the corresponding volume. Otherwise, after reflecting upon the totally geodesic boundary x3 = 0 we obtain a region in “boundaryless” Schwarzschild space which is isoperi- metric but differs from any of the symmetric regions realizing the corresponding isoperimetric profile as in Bray’s result. Essentially the same argument yields an effective area comparison for off-center regions in ahS manifolds. Definition 5.1. Let (M, g, Σ) be as in Theorem 2.28. Given (τ, η) (1, + ) (0, 1), a bounded ∈ ∞ × Borel set Ω M of finite relative perimeter is said to be (τ, η)-off-center if: ⊂ 2 (1) there exists a large coordinate hemisphere Sr,+, r > 1, whose enclosed region, say Mr, has the same volume as Ω; (2) 2(∂∗Ω M ) η (S2 ). Hg \ τr ≥ H2 r,+ Here, 2 is Hausdorff measure with respect to g. Hg The next proposition provides the analogue of the first item above to our setting. Proposition 5.2. Let (M, g, Σ) be as in Theorem 2.28. For every (τ, η) (1, + ) (0, 1) there ex- ∈ ∞ × ists V > 0 and Θ > 0 such that the following holds. Let Ω M be a bounded Borel set with 0 ⊂ finite relative perimeter whose volume is at least V0 and which is (τ, η)-off-center and further satisfies 2(∂∗Ω)1/2vol (Ω)−1/3 Θ and 2(M ∂∗Ω) Θσ2 for all σ 1. Then, Hg g ≤ Hg σ ∩ ≤ ≥ (5.1) 2(S2 )+ cr 2(∂∗Ω), c = c(m,τ,η) > 0. Hg r,+ ≤ Hg Proof. First note that an effective bound similar to (5.1) holds in case Ω is a subset of the half- Schwarzschild space. Indeed, upon reflecting this Ω across the totally geodesic boundary x3 =0 we obtain a region in the exact “boundaryless” Schwarzschild space to which [26, Proposition 3.3] applies. By halving the so obtained estimate, the sought for bound follows. As already emphasized in [26], this bound is robust enough to provide, via a suitable scaling argument, the effective area estimate (5.1) for a (τ, η)-off-center region in a ahS manifold as in the theorem. The argument is virtually identical to the one appearing in the proof of [26, Theorem 3.4], so it is omitted here. By a scaling argument as in the proof of [26, Theorem 5.1], we check that isoperimetric sur- faces remain close to a centered hemisphere as the volume diverges. We now take a divergent 24 SERGIO ALMARAZ AND LEVI LOPES DE LIMA sequence of volumes V + . Arguing as in [26, Proposition 4.2], we see that there exits a fixed i → ∞ isoperimetric region Ω and a coordinate half-ball of radius r 0 which is disjoint from Ω and i i ≥ i contributes to the attained isoperimetric profile in the expected manner: 2πr3 vol (Ω )+ i = V , 2(∂∗Ω )+2πr2 = I (V ). 3 i 3 i Hg i i g i
This is our analogue of the volume splitting in the second item above. Moreover, if ri > 0 as i + then the mean curvature of ∂∗Ω is 2/r , so the relative isoperimetric region associated to → ∞ i i Vi encompasses two disjoints half-balls each roughly with the same radius ri. This contradiction shows that ri = 0 for all i large enough. Thus, to each volume greater than some V0 the corre- sponding isoperimetric region stays at a finite distance from a given point on the manifold. Since we already known that this region centers around a large coordinate hemisphere, it certainly sweeps out the whole manifold as the volume diverges and its boundary necessarily coincides with a free boundary CMC hemisphere described in Theorem 2.27; see Appendix E in regard to this last point. This completes our sketch of the proof of Theorem 2.28. Remark 5.3. As already pointed out in [26], the existence of relative isoperimetric regions for sufficiently large enclosed volumes via the argument above only requires that g is C0-asymptotic to half-Schwarzschild. The higher order asymptotics, the Regge-Teitelboim condition included, are only needed to make sure that a foliation exists as in Theorem 2.27, so its leaves may be identified to the isoperimetric hemispheres. Remark 5.4. The argument above assumes the well-known fact that relative isoperimetric surfaces are sufficiently regular (indeed smooth) up to the boundary and hence are stable free boundary CMC surfaces as explained in Appendix B. This most desirable property is explicitly stated in [56, Proposition 2.4] and we refer to the discussion there for the pertaining sources. Remark 5.5. Very likely an analogue of Theorem 2.27 holds true for the class of asymptotically hyperbolic 3-manifolds with a non-compact boundary introduced in [4]. This would extend a series of results in the boundaryless case literature starting with [55, 50, 51, 43]; see also [49] and the references therein for a recent account of the status of this line of research. In the same vein, it might also be possible to characterize the corresponding large relative isoperimetric regions in the line of the main result in [18], so as to extend Theorem 2.28 accordingly.
APPENDIX A. THELARGESCALEISOPERIMETRICDEFICITSANDTHEMASS: THE PROOFS OF THEOREMS 2.9 AND 2.22 The arguments to prove Theorems 2.9 and 2.22 are simple variations on the computation ap- pearing in [29, Section 2], where a proof of Theorem 2.5 appears. This justifies the inclusion of a somewhat detailed account of their calculation in what follows. If (M,g) is asymptotically flat as in Definition 2.2, we first observe that, since ∂r/∂xi = xi/r, we have x ∂ (A.1) r = gij i . ∇ r ∂xj and hence x x x x (A.2) r 2 = gij i j =1 e i j + O(r−2τ ). |∇ | r2 − ij r2 2 Also, if ν is the outward unit normal vector field to the coordinate 2-sphere Sr then x (A.3) ν = + O(r−τ ). r MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 25
2,δ 2 2,δ Let dSr = r dS1 be the area element of the Euclidean sphere of radius r. It follows that the 2 area element of the corresponding coordinate sphere Sr expands as 1 (A.4) dS2 = 1+ hij e + O(r−2τ ) dS2,δ, r 2 ij r where xixj (A.5) h = g ν ν = δ + O(r−τ ) ij ij − i j ij − r2 2 is the induced metric (extended to vanish in the radial direction). Thus, the area of Sr is
2 1 ij 2,δ 2−2τ (A.6) A(r)=4πr + h eij dSr + O(r ). 2 2 ZSr From this we obtain
d 1 ij xk 2,δ 1 ij 2,δ 1−2τ A(r) = 8πr + h eij,kdSr + h eij dSr + O(r ), dr 2 2 r r 2 ZSr ZSr where the comma means partial differentiation. Using (A.5) we get
d 1 xk 2,δ 1 xixj xk 2,δ A(r) = 8πr + eii,k dSr eij,k 3 dSr dr 2 2 r − 2 2 r ZSr ZSr 1 ij 2,δ 1−2τ + h eij dSr + O(r ). r 2 ZSr We now work out the third term in the right-hand side. We first note that
∂ xj δij xixj (A.7) = 3 . ∂xi r r − r We then compute:
∂ xj xixk 2,δ xixj xk 2,δ eij 2 dSr = eij,k 3 dSr S2 ∂xk r r S2 r Z r Z r δjk xj xk xixk 2,δ + eij 3 2 dSr 0 r − r r ZSr xixj xk 2,δ = eij,k 3 dSr , 2 r ZSr so we have
xixj xk 2,δ ∂ xj xixk 2,δ eij,k 3 dSr = eij 2 dSr S2 r S2 ∂xk r r Z r Z r ∂ xj 2,δ = eij dSr S2 ∂xi r Z r (I)
| {z xixk } ∂ xj 2,δ δik 2 eij dSr . − S2 − r ∂xk r Z r (II) Using (A.5) we have | {z }
xj 2,δ 1 ij 2,δ 1−2τ (I)= eij,i dSr + h eij dSr + O(r ). 2 r r 2 ZSr ZSr 26 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
Also, integration by parts together with (A.7) gives
∂ xixk xj 2,δ xixj 2,δ (A.8) (II)= 2 eij dSr = 2 eij 3 dSr , − S2 ∂xk r r − S2 r Z r Z r so that
xixj xk 2,δ xixj 2,δ xj 2,δ eij,k 3 dSr = 2 eij 3 dSr + eij,i dSr 2 r − 2 r 2 r ZSr ZSr ZSr 1 ij 2,δ 1−2τ + h eij dSr + O(r ). r 2 ZSr Thus,
d 1 xj 2,δ A(r) = 8πr + (eii,j eij,i) dSr dr 2 2 − r ZSr xixj 2,δ 1 ij 2,δ 1−2τ + eij 3 dSr + h eij dSr + O(r ) 2 r 2r 2 ZSr ZSr xixj 2,δ 1 ij 2,δ (A.9) = 8πr 8πm + eij 3 dSr + h eij dSr + o(1). − 2 r 2r 2 ZSr ZSr Combining this with (A.6), we get
d A(r) xixj 2,δ (A.10) A(r)= +4πr 8πm + eij 3 dSr + o(1). dr r − 2 r ZSr 2 We now look at the volume V (r) enclosed by Sr . By the co-area formula, (A.2) and (A.4),
1 d 1 −1 2 V (r) = r dSr r dr r 2 |∇ | ZSr A(r) 1 xj xj 2,δ (A.11) = + eij 3 dSr + o(1). r 2 2 r ZSr We may now eliminate the integral term in (A.10) and (A.11). The result is d d (rA(r)) =4πr2 8πmr +2 V (r)+ o(r). dr − dr Integrating we obtain a formula relating the volume and area, namely, 1 2π (A.12) V (r)= rA(r) r3 +2πmr2 + o(r2), 2 − 3 which gives 1 4πr2 r 2r A(r) 2 J M;3,2 = r + m + o(1). r A(r) − 3 − 3 4πr2 On the other hand, from (A.6),
A(r) −2τ 1 ij 2,δ −τ 2 =1+ + O(r ), := 2 h eij dSr = O(r ) 4πr I I 8πr 2 ZSr so that r 2r 1 J M;3,2 = r + 1 + O(r−2τ ) m 1+ + O(r−2τ ) + o(1) r − I − 3 − 3 2I = m + o(1), which gives the proof of Theorem 2.5. MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 27
So far we have been following [29] closely. We now explain how a little variation yields the proof of Theorem 2.9. We will make use of the well-known expansion 2 (A.13) H = + O(r−τ−1). r Together with (A.4) this gives M(r) (A.14) =1+ + O(r−2τ ). 8πr I Also, by the first variation formula for the area,
d ∂ 2 A(r) = ,Hν dSr dr 2 ∂r ZSr (A.2)+(A.13) −1 2 = r HdSr 2 |∇ | ZSr xixj 2,δ −2τ+1 = M(r)+ eij 3 dSr + O(r ), 2 r ZSr and combining this with (A.10) we get 1 1 (A.15) r2M(r)= rA(r)+2πr3 4πmr2 + o(r2). 2 2 − We now use (A.12) to eliminate the area term. Solving for the volume we get 1 8π (A.16) V (r)= r2M(r) r3 +6πmr2 + o(r2), 2 − 3 so that, using (A.14), 2 8πr 4 2 M(r) 2 J M;3,1 = r + m r r + o(1) r 3 M(r) − 9 − 9 8πr 2 4 2 = r + 1 + O(r−2τ ) m r r 1+2 + O(r−2τ ) + o(1) 3 − I − 9 − 9 I = m + o (1), which finishes the proof of the first equality in (2.11). As for the second one, note that by (A.15) and (A.14), 4πr 1 J M;2,1 = r + (2m r) M(r)+ o(1) r M(r) − − 16π 1 r = r + 1 + O(r−2τ ) (2m r) 1+ + O(r−2τ ) + o(1) 2 − I − − 2 I = m + o(1) , which completes the proof of Theorem 2.9. We now present the proof of Theorem 2.22. We first observe that instead of (A.6) we now have
+ 2 1 ij + 2,δ 2−2τ (A.17) (r)=2πr + h eij dSr,+ + O(r ). A 2 S2 Z r,+ Also, the integration by parts leading to (A.8) now produces an extra term, so that (II) gets replaced by
+ + + xixj 2,δ + xj k 1,δ+ −2τ+1 (A.18) (II )= 2 eij 3 dSr,+ ekj ϑ dSr + O(r ). − S2 r − S1 r Z r,+ Z r 28 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
Thus, instead of (A.10) we now have
d (r) m + xixj 2,δ+ (A.19) (r)= A +2πr 8π + eij 3 dSr,+ + o(1). dr A r − S2 r Z r,+ Hence, proceeding exactly as before we now get 1 π (A.20) (r)= r (r) r3 +2πmr2 + o(r2), V 2 A − 3 which gives
1 r 2πr2 r r (r) 2 M;3,2 = + m A + o(1) Jr 2 (r) − 6 − 3 2πr2 A r r r 1 = + 1 I + O(r−2τ ) m 1+ I + O(r−2τ ) + o(1) 2 − − 6 − 3 2 m = + o(1), b b where 1 ij + 2 −τ I = 2 h eij dSr,+ = O(r ). 4πr 2 ZSr,+ This completes the proof of Theoremb 2.22.
APPENDIX B. THE VARIATIONAL SETUP Here we address the variational issues needed in the bulk of the paper. Our aim is twofold. First, we review the well-known variational theory of free boundary constant mean curvature surfaces [58, 57] in a way that is convenient for our purposes. Next, we discuss the much less known variational theory of closed surfaces which are critical for the total mean curvature func- tional under a volume preserving constraint and develop the corresponding stability theory. We remark that the variational theory associated to curvature integrals involving elementary sym- metric functions of the principal curvatures (quermassintegrals) of hypersurfaces in space forms is a well established subject; see [8] and the references therein. We start by considering a one-parameter family of compact, embedded surfaces t ( ε,ε) 3 ∈ − 7→ St in an arbitrary Riemannian manifold (M ,g) evolving as ∂x (B.1) t = Y , ∂t t where xt is the smooth map defining the embedding and Yt is a vector field along St, a (not necessarily normal) section of TM restricted to St. As usual, if νt is the unit normal vector field along S , let W = ν be the shape operator of S , so the corresponding principal curvatures t ∇ t t (the eigenvalues of W ) are κ1 and κ2. Thus, the mean curvature is H = κ1 + κ2 and the Gauss- Kronecker curvature is K = κ1κ2. For later reference, we recall that 1 K = K Ric (ν,ν) − 2 g is the modified Gauss-Kronecker curvature.e A well-known computation gives d (B.2) A(t) = div Y dS, dt |t=0 S ZS MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 29 where A(t) is the area of St and we agree to drop the subscript t upon evaluation at t = 0. Next we decompose Yt into its normal and tangential components: (B.3) Y = f ν + Y ⊤, f = Y ,ν . t t t t t h t ti Thus, if we assume further that St carries a boundary ∂St, d A(t) = fHdS + div Y ⊤dS dt |t=0 S ZS ZS (B.4) = fHdS + Y,µ d∂S, h i ZS Z∂S where µ is the outward unit normal vector field along ∂S and we used that H = divSν. Let us assume now that M also carries a boundary, say Σ, with the variation being admissible in the sense that ∂S Σ. It follows that S = S is critical for the area under such variations t ⊂ 0 satisfying the volume preserving condition
(B.5) fdS =0 ZS if and only if the mean curvature is constant and S meets Σ orthogonally along ∂S. We then say that S is a free boundary constant mean curvature (CMC) surface. We now recall the corresponding notion of stability. Assuming that S = S0 is a free boundary CMC as above, a well-known computation [57] gives the second variational formula for the area: d2A ∂f (B.6) = fL fdS + f κf d∂S, dt2 |t=0 S ∂µ − ZS Z∂S where (B.7) L = ∆ W 2 + Ric (ν,ν) , S − S − | | g κ = ν, ν and = η is the shape operator of the embedding Σ ֒ M. Here, η is the outward h W i W ∇ → unit normal vector to M along Σ. Recall that S = S0 is strictly stable (as a free boundary CMC surface) if the right-hand side of (B.6) is positive for any f =0 satisfying (B.5). Accordingly, we define 6 (S)= f H1(S); fdS =0 . F ∈ ZS Proposition B.1. A free boundary CMC surface S as above is strictly stable if and only if the first eigen- value λLS of the eigenvalue problem
LSf = λf in S, ∂f = κf on ∂S, ∂µ is positive, where f (S). Equivalently, for any 0 = f (S), ∈F 6 ∈F f 2 W 2 + Ric(ν,ν) f 2 dS κf 2d∂S > 0. |∇S | − | | − ZS Z∂S We now turn to the variational theory of the total mean curvature functional S HdS. Here we assume that S is closed (∂S = ) and the variation is normal (Y = fν). A simple computation t ∅ R shows that the shape operator evolves as ∂W (B.8) = 2 f (W 2 + Riemν )f, ∂t −∇S − g where 2 , the Hessian of f,isviewedasa (1, 1)-tensor, Riemν ( ) = Riem ( ,ν)ν and W 2 = W W . ∇S g · g · ◦ 30 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
Proposition B.2. In a Riemannian 3-manifold (M,g) as above, a closed surface extremizes the total mean curvature under volume (respectively, area) preserving variations if and only if K = const (respectively, K = γH, where γ is a constant). e Proof. From ∂dS /∂t = fHdS , the fact that the mean curvature evolves as e t t ∂H (B.9) = L f, ∂t S and the algebraic identity W 2 = H2 2K, we immediately see that | | − ∂ HdS =2 KfdS , ∂t t|t=0 t ZSt ZS which proves the first statement. As for the second one, juste combine the computation above with (B.4) and take into account that ∂S = . ∅ In order to discuss the stability of this variational problem, we now compute the variation of K. First, from ∂ν/∂t = f, −∇S ∂ (B.10)e Ricg(ν,ν)= f( ν Ricg)(ν,ν) 2Ricg( Sf,ν). ∂t ∇ − ∇ As for the variation of K, we first recall the well-known formula ∂ ∂ K = tr Π W , ∂t S ∂t where Π= HI W is the Newton tensor [59]. Using (B.8) we then get − ∂ (B.11) K = tr (Π 2 f) ftr (Π W 2) ftr (Π Riemν ). ∂t − S ∇S − S − S g To proceed we choose an orthonormal frame e , A =1, 2, tangent to S with ( ) e =0 at the A ∇S eA B given point. We compute tr (Π 2 f) = ΠAB ( ) f,e S ∇S h ∇S eA ∇S Bi = ΠAB e f,e ΠAB f, ( ) e Ah∇S Bi− h∇S ∇S eA Bi = (ΠAB f B) ΠAB f B, ∇S ;A − ;A∇S where the semicolon denotes covariant derivation. By Codazzi equations, recalling that h = g , |S ΠAB = (HhAB) W AB ;A ;A − ;A = (e H)hAB W AA R(e ,e )ν,e A − ;B − h A B Ai = e H e H Ric (ν,e ) B − B − g B = Ric (ν,e ), − g B so that tr (Π 2 f) = div (Π f) + Ric (ν, f). S ∇S S ∇S g ∇S 2 Thus, from (B.11) and the algebraic identity trS(ΠW )= HK, which holds in dimension 3, ∂ K = Λ f Ric (ν, f) fHK ftr (ΠRiemν ), ∂t − S − g ∇S − − S g where (B.12) Λ f = div (Π f). S S ∇S MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 31
Together with (B.10) this finally gives ∂ (B.13) K = L f, ∂t S where e 1 (B.14) L = Λ HK ( Ric )(ν,ν) tr (ΠRiemν ) S − S− − 2 ∇ν g − S g is the corresponding Jacobi operator. We note that
(B.15) fΛ fdS˜ = Π f, f˜ dS, − S h ∇S ∇S i ZS ZS for any functions f and f˜. In particular, LS is always self-adjoint. Moreover, it is easy to check that this operator is elliptic whenever Π is positive definite. We now consider a surface S M satisfying K = const. and with the property that Π is ⊂ positive definite everywhere. We then say that S is strictly stable if d2 e HdS > 0, dt2 |t=0 ZS for any normal variation as in (B.1) with f =0. As before let us set 6 (S)= f H1(S); fdS =0 . G ∈ ZS Proposition B.3. S is strictly stable if and only if 1 Π f, f f 2 HK + ( Ric )(ν,ν)+tr (ΠRiemν ) dS > 0, h ∇S ∇S i− 2 ∇ν g S g ZS for any 0 = f (S). Equivalently, the first eigenvalue λ of the eigenvalue problem 6 ∈ G LS L f = λf, f (S), S ∈ G is positive.
APPENDIX C. THE GAUSS-KRONECKER CURVATURE IN TERMS OF THE MEAN CURVATURE In this section we prove Proposition 3.3. Thus we aim to prove the identity (3.4) which ex- presses the Gauss-Kronecker curvature in terms of the mean curvature up to terms decaying fast enough at infinity. Our starting point is the fact that the radial vector field ∂ X = (xi ai) − ∂xi is conformal with respect to the Euclidean metric, i.e., X δ = 2δ, where is the Lie derivative. L γ1,γ2L From this we see that X is also conformal with respect to the metric fm,c δ where 2m c x 1 f γ1,γ2 =1+ + γ · + γ m,c r 1 r3 2 r2 for some γ ,γ R and c R3. Indeed, there holds (f γ1,γ2 δ)=2ξf γ1,γ2 δ, with 1 2 ∈ ∈ LX m,c m,c 1 1 (C.1) ξ(x) = f γ1,γ2 + ∂ f γ1,γ2 (x a ) f γ1,γ2 m,c 2 k m,c k − k m,c m 2m2 γ m γ = 1 + − 2 + x a 1 x c + O(r−3) − r r2 r3 · − r3 · m 2m2 γ 2m γ = 1 + − 2 + x a 1 x c + O(ρ−3), − ρ ρ2 ρ3 · − ρ3 · 32 SERGIO ALMARAZ AND LEVI LOPES DE LIMA where in the last step we used that (x a) a (C.2) rk = ρk + k − · + O(ρ−2+k), k R. ρ2−k ∈
γ1,γ2 −2−ǫ Let us consider an aS metric of the form g = fm,c δ + p, where p = O(r ), which satisfies (2.18) with ǫ 0. ≥ Proposition C.1. The vector field X is almost conformal with respect to g in the sense that (C.3) g =2ξg + B, where B = O(ρ−2−ǫ). LX Proof. A direct computation shows that (C.3) holds with B = p 2ξp. Note however that LX − ( p) = Xi p + p Xi + p Xi, LX jk ∇i jk ik∇j ij ∇k and the result follows given that X = O(r).
We now take e ,e to be a local orthonormal frame on S2(a) and ν its outward unit normal { 1 2} ρ vector. Recall that W = ν is the shape operator of S2(a) and Π = HI W denotes its Newton ∇ ρ − tensor. If X⊤ = X X,ν ν is the tangential component of X, then − h i ⊤ g(Πe ,e )= L g(Πe ,e ) 2 X,ν W (Πe ,e ), LX A A X A A − h i A A and we obtain from (C.3) that X⊤,e + X⊤, Πe =2ξ Πe ,e 2 X,ν W Πe ,e + B(Πe ,e ). h∇ΠeA Ai h∇eA Ai h A Ai− h ih A Ai A A Since X⊤, Πe = e , X⊤ , h∇eA Ai h A ∇ΠeA i which is easily verified if we take the frame to be principal with respect to the shape operator W , this simplifies to 1 X⊤, Πe = ξ Πe ,e X,ν W Πe ,e + B(Πe ,e ). h∇eA Ai h A Ai − h ih A Ai 2 A A Thus, summing over A and using that H2 W 2 =2K , we obtain a,ρ − | | a,ρ 1 (C.4) X⊤, Πe = ξH 2 X,ν K + B(Πe ,e ). h∇eA Ai a,ρ − h i a,ρ 2 A A XA XA In order to make use of this identity, which first appeared in [1], we need to determine the asymp- totics of X⊤. Proposition C.2. One has X⊤ = O(ρ−1−ǫ). Proof. Recalling that r = (x a)/ρ, so that X = ρr ∂/∂x , one computes − i i −1/2 ∂ γ1,γ2 r −2−ǫ (C.5) ν = fm,c i + O(ρ ), ∂xi so that 1/2 (C.6) X,ν = f γ1,γ2 ρ + O(ρ−1−ǫ) h i m,c Thus, ∂ −1−ǫ X,ν ν = ρ ri + O(ρ ), h i ∂xi and the result follows. MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 33
We now observe that by (3.1) we may rewrite (3.25) as 1 Π= H I + O(ρ−3), 2 ρ,a so that 1 1 −5 B(ΠeA,eA)= Hρ,atr 2 B + O(ρ ), 2 4 Sρ(a) XA where we used that B = O(ρ−2). Also, the left-hand side of (C.4) may be treated similarly. Indeed, by Proposition C.2,
⊤ 1 ⊤ −5 e X , ΠeA = Hρ,adiv 2 X + O(ρ ). h∇ A i 2 Sρ (a) XA Putting all the pieces of our computation together and using (C.6) we get
ξ 1 1 1 ⊤ −6 2Ka,ρ = + tr 2 B div 2 X Ha,ρ + O(ρ ) X,ν X,ν 4 Sρ(a) − 2 Sρ(a) h i h i ξ = + O(ρ−3−ǫ) H + O(ρ−6). X,ν a,ρ h i The proof of Proposition 3.3 is completed if we note that by (C.1) and (C.6),
ρ ξ −1/2 = ξ f γ1,γ2 + O(ρ−2−ǫ) X,ν m,c h i m 2m2 γ 2m γ = 1 + − 2 + x a 1 x c + O(ρ−3) − ρ ρ2 ρ3 · − ρ3 · m 3m2 γ m γ 1 + − 2 + x a 1 x c + O(ρ−2−ǫ) · − ρ 2ρ2 ρ3 · − 2ρ3 · 2m 9m2 3γ 3m 3γ = 1 + − 2 + x a 1 x c + O(ρ−2−ǫ). − ρ 2ρ2 ρ3 · − 2ρ3 ·
APPENDIX D. THE PROOF OF PROPOSITION 4.1 Here we indicate how the argument in [27, Appendix F] may be used to prove Proposition 4.1. In fact, this method allows us to approach the problem in the category of manifolds considered in Definition 2.25. Proposition D.1. If (M,g) is an asymptotically flat 3-manifold with a non-compact boundary satisfying the RT condition then
+ 2 2,δ m + −τ (D.1) (xα bα) Hρ,+,b dSρ,+ (b)=8π bα Cα + O(ρ ), α =1, 2. S2 (b) − − ρ − Z ρ,+ Corollary D.2. There holds
1 + + = lim x H dS2,δ (~0). α m α ρ,+,~0 ρ,+ C − ρ→+∞ 8π S2 (~0) Z ρ,+ 2 The key ingredient in the proof is an integral identity derived from the fact that Sρ,+(b) is a free boundary CMC surface with mean curvature 2/ρ with respect to the metric δ+. 34 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
Proposition D.3. There holds
+ r r + 1 + r r r 2,δ 1 + r + i j 2,δ (xα bα) eij,k i j kdSρ,+ (b) = (xα bα) eij,k j 2eij dSρ,+ (b) 2 S2 (b) − S2 (b) − 2 − ρ Z ρ,+ Z ρ,+ 1 +r + r 2,δ+ + eii α + eiα i dSρ,+ (b) 2 S2 (b) Z ρ,+ 1 + r 1,δ+ (xα bα) ei3 idSρ (b), −2 1 − ZSρ(b) where S1(b)= ∂S2 (b) and recall that r = (x b)/ρ. ρ ρ,+ − Proof. Apply the identity that follows from equating the right-hand sides of (B.2) and (B.4) with µ = ∂/∂x to the vector field Y = (x b )e+r ∂/∂x by taking into account that − 3 (α) α − α ij i j + e+ + r eii ij r r + r + r r r divS2 (b) Yα = eiα i + (xα bα) 2 i j + eij,j i eij,k i j k . ρ,+ − ρ − ρ − !
We now recall the expansion 2 1 r r 1 e+ H = e+ r r r +2e+ i j e+ r + e+ r ii + E, ρ,+,b − ρ 2 ij,k i j k ij ρ − ij,i j 2 ii,j j − ρ
′ where the remainder satisfies E = O(ρ−1−2τ ) and E(−1 ) = O(ρ−2−2τ ); see [34, Lemma 2.1]. This reduces to (4.1) if we take e+ =2mr−1δ+ +O(r−2), which provides the link between Propositions D.1 and 4.1. It follows that
2 2,δ+ 1 + + r 2,δ+ (xα bα) Hρ,+,b dSρ,+ (b) = (xα bα) eij,i eii,j j dSρ,+ (b) S2 (b) − − ρ −2 S2 (b) − − Z ρ,+ Z ρ,+ 1 + r +r 2,δ+ + eiα i eii α dSρ,+ (b) 2 S2 (b) − Z ρ,+ 1 + r 1,δ+ −τ (xα bα) ei3 idSρ (b)+ O(ρ ), −2 1 − ZSρ(b) where Proposition D.3 has been used to make sure that only those terms which are linear in r survive in the right-hand side. We now observe that under the decay assumptions (including Regge-Teitelboim) the integrals
+ + bj 2,δ+ + + bj 2,δ+ xα eij,i eii,j dSρ,+ (b), eij,i eii,j dSρ,+ (b), S2 (b) − ρ S2 (b) − ρ Z ρ,+ Z ρ,+ and + + bi + bα 2,δ eiα eii dSρ,+ (b) S2 (b) ρ − ρ Z ρ,+ are O(ρ−τ ), the same happening to the boundary integrals
bα + 1,δ+ bαbi + 1,δ+ xαei3dSρ (b), ei3dSρ (b). ρ 1 ρ 1 ZSρ (b) ZSρ (b) MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 35
Thus, we end up with
2 2,δ+ 1 + + xj 2,δ+ (xα bα) Hρ,+,b dSρ,+ (b) = xα eij,i eii,j dSρ,+ (b) S2 (b) − − ρ −2 S2 (b) − ρ Z ρ,+ Z ρ,+ 1 + xi + xα 2,δ+ + eiα eii dSρ,+ (b) 2 S2 (b) ρ − ρ Z ρ,+ 1 + xi 1,δ+ xαei3 dSρ (b) −2 1 ρ ZSρ(b) + 1 + + xj 2,δ + bα eij,i eii,j dSρ,+ (b) 2 S2 (b) − ρ Z ρ,+ 1 + xi 1,δ+ −τ + bα ei3 dSρ (b)+ O(ρ ). 2 1 ρ ZSρ(b) Comparing the right-hand side of the above with the definitions of m and +, the proof of Propo- C sition D.1, and hence of Proposition 4.1, follows. Remark D.4. The upshot of Corollary D.2 is another expression for the center of mass +, besides C (2.28), derived from Hamiltonian methods, and the isoperimetric one appearing in Theorems 2.27 and 2.28. Another rendition of this invariant comes from [24, Theorem 2.4], this time in terms of certain asymptotic flux integrals involving the Einstein tensor of the metric in the interior and the Newton tensor along the boundary; see also [14]. It is remarkable indeed that this kind of invariant admits so many distinct manifestations.
APPENDIX E. THEUNIQUENESSOFTHEFREEBOUNDARY CMC HEMISPHERES The very last piece of the argument leading to Theorem 2.28 uses the appropriate uniqueness of the free boundary CMC hemispheres in Theorem 2.27. Here we justify this step by following the reasoning in [38, Section 4]. We know from the analysis in Section 4 that for each ρ large enough the corresponding hemisphere is a strictly stable free boundary CMC surface graphically described by a function φ on S2 ( +) satisfying the bound ρ ρ,+ C −1/2 (ρ) ρ φ 2,α C, k ρkC ≤ where C > 0 is an absolute constant and the weighted H¨older norm is defined as in the left- hand side of (3.22). The uniqueness claim is that, for ρ large enough depending only on C, any other free boundary CMC hemisphere with the same mean curvature and which is graphed by a function satisfying this H¨older bound should coincide with (the graph of) φ0 := φρ. Indeed, assume there exists another such hemisphere, say associated to a function φ1. As in [38, Propo- sition 2.1], the asymptotic roundness of the graphs means that we may interpolate between the corresponding embeddings by setting F (x)= F (x)+ tu(x)ν(x), t [0, 1], t 0 ∈ for some function u(x) = ~a,ν(x) + q(x), where ~a R2 is a vector and q = O(ρ−1). A crucial h i ∈ remark at this point is that all of these surfaces are free boundary (with a possibly non-constant mean curvature HFt for 0
Y C implies Y =0. We next check that Y (equivalently, ~a ) may be chosen small enough so | |≤ 2 | | | | as to fulfill this vanishing criterion if ρ is large. We first note that, again because HF0 = HF1 ,
(E.1) dHF0 Y sup (dHFt dHF0 )Y . k k≤ t k − k As in [6, Proposition 16] we compute that L ⊤ dHFt Y = Ft u + Y HFt , where Y ⊤ is the tangential component of Y . Starting with (C.5) we obtain Y ⊤ = O(ρ−3) and | | hence Y ⊤H = O(ρ−4). Combining this with (4.10) we see that the right-hand side of (E.1) is | Ft | O(ρ−4). On the other hand, since ~a,ν is an approximate eigenfunction of L under Neumann h i F0 boundary condition with eigenvalue close to 6mρ−3, the left-hand side of (E.1) is C ~a ρ−3. ≥ 3| | Thus, ~a C ρ−1 and the uniqueness claim follows provided we take ρ C−1C . | |≤ 4 ≥ 2 4
REFERENCES [1] Luis J Al´ıas, Aldir Brasil, and Antonio G Colares. Integral formulae for spacelike hypersurfaces in conformally sta- tionary spacetimes and applications. Proceedings of the Edinburgh Mathematical Society, 46(2):465–488, 2003. [2] S´ergio Almaraz. Convergence of scalar-flat metrics on manifolds with boundary under a Yamabe-type flow. Journal of Differential Equations, 259(7):2626–2694, 2015. [3] S´ergio Almaraz, Ezequiel Barbosa, and Levi L de Lima. A positive mass theorem for asymptotically flat manifolds with a non-compact boundary. Communications in Analysis and Geometry, 24(4):673–715, 2016. [4] S´ergio Almaraz and Levi L de Lima. The mass of an asymptotically hyperbolic manifold with a non-compact bound- ary. Annales Henri Poincar´e, 21(11):3727–3756, 2020. [5] S´ergio Almaraz, Levi L de Lima, and Luciano Mari. Spacetime positive mass theorems for initial data sets with noncompact boundary. International Mathematics Research Notices, 2020. [6] Lucas C Ambrozio. Rigidity of area-minimizing free boundary surfaces in mean convex three-manifolds. The Journal of Geometric Analysis, 25(2):1001–1017, 2015. [7] Richard Arnowitt, Stanley Deser, and Charles W Misner. The dynamics of general relativity. Gravitation: an introduc- tion to current research, 1962. [8] Jo˜ao Lucas M Barbosa and Antˆonio G Colares. Stability of hypersurfaces with constant r-mean curvature. Annals of Global Analysis and Geometry, 15(3):277–297, 1997. [9] Robert Beig and Niall O´ Murchadha. The Poincar´egroup as the symmetry group of canonical general relativity. Annals of Physics, 174(2):463–498, 1987. [10] Hubert L Bray. The Penrose inequality in general relativity and volume comparison theorems involving scalar curvature. PhD thesis, Stanford University, 1997. [11] Simon Brendle and Fernando Marques. Recent progress on the Yamabe problem. Adv. Lect. Math., 20:29–47, 2011. [12] Carla Cederbaum and Christopher Nerz. Explicit riemannian manifolds with unexpectedly behaving center of mass. Annales Henri Poincar´e, 16(7):1609–1631, 2015. [13] Carla Cederbaum and Anna Sakovich. On center of mass and foliations by constant spacetime mean curvature sur- faces for isolated systems in general relativity. arXiv preprint arXiv:1901.00028, 2018. [14] Xiaoxiang Chai. Two quasi-local masses evaluated on surfaces with boundary. arXiv preprint arXiv:1811.06168, 2018. [15] Pak-Yeung Chan and Luen-Fai Tam. A note on center of mass. Comm. Anal. Geo., 24(3):471–486, 2016. [16] Sun-Yung Alice Chang and Yi Wang. On Aleksandrov-Fenchel inequalities for k-convex domains. Milan Journal of Mathematics, 79(1):13–38, 2011. [17] Po-Ning Chen, Mu-Tao Wang, and Shing-Tung Yau. Quasilocal angular momentum and center of mass in general relativity. Advances in Theoretical Mathematical Physics, 20:671–682, 2016. [18] Otis Chodosh. Large isoperimetric regions in asymptotically hyperbolic manifolds. Communications in Mathematical Physics, 343(2):393–443, 2016. [19] Otis Chodosh, Michael Eichmair, Yuguang Shi, and Haobin Yu. Isoperimetry, scalar curvature, and mass in asymp- totically flat riemannian 3-manifolds. arXiv:1606.04626, 2016. [20] Demetrios Christodoulou. Mathematical problems of general relativity I, volume 1. European Mathematical Society, 2008. [21] Justin Corvino, Aydin Gerek, Michael Greenberg, and Brian Krummel. On isoperimetric surfaces in general relativity. Pacific Journal of Mathematics, 231(1):63–84, 2007. [22] Justin Corvino and Haotian Wu. On the center of mass of isolated systems. Classical and Quantum Gravity, 25(8):085008, 2008. MASS, CENTER OF MASS AND ISOPERIMETRY IN ASYMPTOTICALLY FLAT 3-MANIFOLDS 37
[23] Levi L de Lima. Conserved quantities in General Relativity: the case of initial data sets with a non-compact boundary. In preparation, 2021. [24] Levi L de Lima, Frederico Gir˜ao, and Amilcar Montalb´an. The mass in terms of Einstein and Newton. Classical and Quantum Gravity, 36(7):075017, 2019. [25] Levi L de Lima and Isaac C L´azaro. A Cauchy-Crofton formula and monotonicity inequalities for the Barbosa-Colares functionals. The Asian Journal of Mathematics, 7(1):81–89, 2003. [26] Michael Eichmair and Jan Metzger. Large isoperimetric surfaces in initial data sets. Journal of Differential Geometry, 94(1):159–186, 2013. [27] Michael Eichmair and Jan Metzger. Unique isoperimetric foliations of asymptotically flat manifolds in all dimensions. Inventiones mathematicae, 194(3):591–630, 2013. [28] Mouhamed M Fall. Area-minimizing regions with small volume in riemannian manifolds with boundary. Pacific Journal of Mathematics, 244(2):235–260, 2009. [29] Xu-Qian Fan, Yuguang Shi, and Luen-Fai Tam. Large-sphere and small-sphere limits of the brown-york mass. Com- munications in Analysis and Geometry, 17(2):37–72, 2009. [30] Alfred Gray. Tubes, volume 221 of Progress in Mathematics. Birkh¨auser, 2012. [31] Pengfei Guan and Junfang Li. The quermassintegral inequalities for k-convex starshaped domains. Advances in Math- ematics, 221(5):1725–1732, 2009. [32] Daniel Harlow and Jie-qiang Wu. Covariant phase space with boundaries. Journal of High Energy Physics, 2020(10):1– 52, 2020. [33] Lan-Hsuan Huang. On the center of mass of isolated systems with general asymptotics. Classical and Quantum Grav- ity, 26(1):015012, 2008. [34] Lan-Hsuan Huang. Center of mass and constant mean curvature foliations for isolated systems. MSRI Lecture Notes, 2009. [35] Lan-Hsuan Huang. Foliations by stable spheres with constant mean curvature for isolated systems with general asymptotics. Communications in Mathematical Physics, 300(2):331–373, 2010. [36] Gerhard Huisken. An isoperimetric concept for mass and quasilocal mass. Oberwolfach Rep, 3(1):87–88, 2006. [37] Gerhard Huisken and Tom Ilmanen. The inverse mean curvature flow and the riemannian penrose inequality. Journal of Differential Geometry, 59(3):353–437, 2001. [38] Gerhard Huisken and Shing-Tung Yau. Definition of center of mass for isolated physical systems and unique folia- tions by stable spheres with constant mean curvature. Inventiones mathematicae, 124(1-3):281–311, 1996. 0 [39] Jeffrey L Jauregui and Dan A Lee. Lower semicontinuity of mass under C convergence and Huisken’s isoperimetric mass. Journal f¨ur die reine und angewandte Mathematik, 2019(756):227–257, 2019. [40] Dan A Lee and Philippe G LeFloch. The positive mass theorem for manifolds with distributional curvature. Commu- nications in Mathematical Physics, 339(1):99–120, 2015. [41] John M Lee and T Parker. The Yamabe problem. Bulletin of AMS, 17(1):37–91, 1987. [42] Davi M´aximo, Ivaldo Nunes, and Graham Smith. Free boundary minimal annuli in convex three-manifolds. Journal of Differential Geometry, 106(1):139–186, 2017. [43] Rafe Mazzeo and Frank Pacard. Constant curvature foliations in asymptotically hyperbolic spaces. Revista Matematica Iberoamericana, 27(1):303–333, 2011. [44] Jan Metzger. Foliations of asymptotically flat 3-manifolds by 2-surfaces of prescribed mean curvature. Journal of Differential Geometry, 77(2):201–236, 2007. [45] Benoıt Michel. Geometric invariance of mass-like asymptotic invariants. Journal of Mathematical Physics, 52(5):052504, 2011. [46] Jos´eFabio Montenegro. Foliation by free boundary constant mean curvature leaves. arXiv:1904.11867, 2019. [47] Abraham E Munoz Flores and Stefano Nardulli. The isoperimetric problem of a complete riemannian manifold with a finite number of asymptotically schwarzschild ends. Communications in Analysis and Geometry, 28(7):1577–1601, 2020. [48] Christopher Nerz. Foliations by stable spheres with constant mean curvature for isolated systems without asymptotic symmetry. Calculus of Variations and Partial Differential Equations, 54(2):1911–1946, 2015. [49] Christopher Nerz. Foliations by spheres with constant expansion for isolated systems without asymptotic symmetry. Journal of Differential Geometry, 109(2):257–289, 2018. [50] Andr´eNeves and Gang Tian. Existence and uniqueness of constant mean curvature foliation of asymptotically hy- perbolic 3-manifolds. Geometric and Functional Analysis, 19(3):910, 2009. [51] Andr´eNeves and Gang Tian. Existence and uniqueness of constant mean curvature foliation of asymptotically hy- perbolic 3-manifolds II. Journal f¨ur die reine und angewandte Mathematik, 2010(641):69–93, 2010. [52] Jie Qing and Gang Tian. On the uniqueness of the foliation of spheres of constant mean curvature in asymptotically flat 3-manifolds. Journal of the American Mathematical Society, 20(4):1091–1110, 2007. [53] Tullio Regge and Claudio Teitelboim. Role of surface integrals in the Hamiltonian formulation of general relativity. Annals of physics, 88(1):286–318, 1974. 38 SERGIO ALMARAZ AND LEVI LOPES DE LIMA
[54] Robert C Reilly. Applications of the Hessian operator in a riemannian manifold. Indiana University Mathematics Jour- nal, 26(3):459–472, 1977. [55] Ralf Rigger. The foliation of asymptotically hyperbolic manifolds by surfaces of constant mean curvature (including the evolution equations and estimates). Manuscripta Mathematica, 113(4):403–421, 2004. [56] Manuel Ritor´eand C´esar Rosales. Existence and characterization of regions minimizing perimeter under a volume constraint inside Euclidean cones. Transactions of the American Mathematical Society, 356(11):4601–4622, 2004. [57] Antonio Ros and Rabah Souam. On stability of capillary surfaces in a ball. Pacific Journal of Mathematics, 178(2):345– 361, 1997. [58] Antonio Ros and Enaldo Vergasta. Stability for hypersurfaces of constant mean curvature with free boundary. Ge- ometriae Dedicata, 56(1):19–33, 1995. [59] Harold Rosenberg. Hypersurfaces of constant curvature in space forms. Bulletin des Sciences Math´ematiques, 117(2):211–239, 1993. [60] Rolf Schneider. Convex bodies: the Brunn–Minkowski theory. Number 151 in Encyclopedia of Mathematics and its Ap- plications. Cambridge University Press, 2014. [61] Richard Schoen. Conformal deformation of a riemannian metric to constant scalar curvature. Journal of Differential Geometry, 20(2):479–495, 1984. [62] Richard Schoen and Shing-Tung Yau. On the proof of the positive mass conjecture in general relativity. Communica- tions in Mathematical Physics, 65(1):45–76, 1979. [63] Rugang Ye. Foliation by constant mean curvature spheres on asymptotically flat manifolds. Geometric Analysis and the Calculus of Variations, 1996.
UNIVERSIDADE FEDERAL FLUMINENSE (UFF), INSTITUTO DE MATEMATICA´ , CAMPUS DO GRAGOATA´ , RUA PROF. MARCOS WALDEMAR DE FREITAS, S/N, 24210-201, NITEROI´ ,RJ,BRAZIL. Email address: [email protected]
UNIVERSIDADE FEDERAL DO CEARA´ (UFC), DEPARTAMENTO DE MATEMATICA´ , CAMPUS DO PICI,AV. HUMBERTO MONTE, S/N,BLOCO 914, 60455-760, FORTALEZA,CE,BRAZIL. Email address: [email protected]