Introduction [sec:intro]. When a massive star collapses, it may form a black hole. Penrose's 1973
heuristic argument says no physical process can produce a black hole whose
horizon area A exceeds 16pi m^2 for total mass m, equivalently
mge√(A/16pi); otherwise a naked singularity would form, violating
cosmic censorship. This is the Penrose inequality. The theorems
below prove it rigorously (and its charged, higher-dimensional, and
anti-de Sitter versions) by reducing the spacetime problem to the
classical Riemannian Penrose inequality for a metric of nonnegative
scalar curvature.

An initial data set for the Einstein equations is a triple (M,g,K): a
Riemannian manifold (M,g) and a symmetric 2-tensor K, the second
fundamental form of a spacelike hypersurface in a spacetime. The constraint
equations tie the energy density mu and momentum density J of the data
to their geometry, and the dominant energy condition (DEC) requires
muge|J|_g everywhere.

Fix a compact hypersurface Ssubset M and a unit normal nu toward the
designated exterior. Its future outward null expansion is


   theta_+(S)=H_S+tr_S K,


where H_S is the mean curvature of S with respect to nu. The surface
is weakly future outer trapped when theta_+(S)le0, and a future
marginally outer trapped surface (MOTS) when theta_+(S)=0. A horizon
appears as such a surface in a spacelike slice.

Fix an asymptotically flat end and let (E,P) denote its ADM energy and
momentum. Once E>|P|, the invariant mass of the end is
m=√(E^2-|P|^2). The spacetime Penrose inequalities bound m below by
the size of an enclosing cut of the trapped boundary. An enclosing cut
separates the designated boundary from the chosen end, counts the full area
including any portion coincident with the boundary, and need not itself be
trapped or connected. Let A_{min} denote the infimum of this area.


The results



In three spatial dimensions the basic bound is


   mgesqrt{frac{A_{min}}{16pi}}.


the [thm:n3:exterior] theorem proves this for a designated end with
arbitrary K and nonzero ADM momentum; the [thm:bridge:global] theorem
removes any restriction on the ADM vector and allows a disconnected chosen
exterior region and a disconnected trapped boundary. Equality is treated in
the [thm:n3:rigidity] theorem: if the chosen exterior is connected and
one-ended and its boundary is a future MOTS that is outermost and outer
area-minimizing, then equality forces a single spherical boundary component
and realizes the entire exterior as a slice of Schwarzschild spacetime.

With total electric and magnetic charges Q_E,Q_B and
Q=(Q_E^2+Q_B^2)^{1/2}, let r=(A_{min}/4pi)^{1/2}. For a one-ended
exterior whose matter condition includes the electromagnetic momentum,
the [thm:ch:main] theorem proves


   mge Q, rle m+√(m^2-Q^2).


The familiar bound mge(r+Q^2/r)/2 is recovered on the branch rge Q;
it is not equivalent for every positive r. Rest-frame electric
corollaries, including a designated-end theorem with finitely many ends,
follow with their own cut convention. No charged equality classification is
claimed.

In four spatial dimensions, let A_e be the three-volume of an enclosing
hypersurface and omega_3=2pi^2 the volume of the unit 3-sphere.
the [thm:four:main] theorem proves


   m_egefrac12((A_e)/(omega_3))^{2/3}


for arbitrary K, finitely many ends, and a disconnected weakly future
trapped boundary, and the [thm:rig4:rigidity] theorem identifies the
connected, one-ended, outermost, outer area-minimizing equality case as
Schwarzschild--Tangherlini geometry.

The anti-de Sitter part fixes a positive-mass Schwarzschild--anti-de Sitter
background, a transverse-traceless (trace free and divergence free) seed,
and a decaying positive conformal branch. With A the boundary area and
m_{AH} the positive Lorentz norm of the hyperbolic mass covector,
the [thm:ads:main] theorem gives, on a sufficiently small interval depending
on the fixed data,


   m_{AH}ge√((A)/(16pi))(1+(A)/(4pi)),


with equality exactly for radial seeds at small positive parameter.


The common scheme



The asymptotically flat bounds follow one mechanism. The end is first
replaced and repaired so its energy approaches the invariant mass of a rest
frame, with the DEC preserved and every enclosing cut controlled. The
trapped-region boundary and disjoint collars then prepare a fixed exterior
for a coupled elliptic deformation. The graph function and conformal factor
produce a Riemannian comparison metric of nonnegative scalar curvature and
controlled energy, whose enclosing area is at least the original infimum.
The Riemannian Penrose inequality finishes the comparison. The charged
argument reuses the same elliptic system and adds a square estimate that
carries the electromagnetic momentum.

The equality proofs work on unmodified data. Variations of the constraints
and of the area infimum yield a normalized causal stationary field. A twist
estimate turns its positive-norm region into a static one, and a complete
conformal double forces Euclidean geometry. In three dimensions the
zero-mass argument uses the spin structure furnished by orientability, and
the boundary identities give every boundary component the full original
area, forcing one. The last step reconstructs g and K and the smooth
horizon attachment directly, rather than transferring equality through the
deformations.

For finite-perimeter minimizers, the full obstacle perimeter is used, so that
any boundary-coincident part is counted. Compactness of all minimizing
enclosures then makes the one-sided metric derivative the minimum of their
area derivatives. A positive measure on this minimizing space yields the
variational identity; normal tests and the outermostness condition charge
the area term to the original horizon.


External inputs [sec:external]. Three classical results are used without proof. These are stated precisely
once here, so that a reader can tell where the paper's original
contributions end and the classical machinery begins.

(T1) Riemannian Penrose inequality (Huisken--Ilmanen 2001, Bray
  2001). If (N,h) is a complete connected asymptotically flat
  three-manifold with nonnegative scalar curvature R_hge 0 and ADM
  energy E_h, and ∂ N is a compact outermost smooth minimal
  surface of total area A, then E_hge√(A/16pi), with equality
  in the nondegenerate case exactly for the Schwarzschild horizon.
pdf/huisken-ilmanen-2001.pdf

(T2) Positive mass theorem (Schoen--Yau 1979, Witten 1981). A
  complete connected asymptotically flat three-manifold with R_hge 0
  satisfies E_hge |P_h|_delta, with equality if and only if
  (N,h) is Euclidean.
pdf/witten-1981.pdf

(T3) Static vacuum uniqueness (Bunting--Masood-ul-Alam 1987).
  A complete three-dimensional static vacuum data set (N,h,N) with
  positive lapse N>0 and a nondegenerate Killing horizon is isometric
  to the Schwarzschild exterior {r>2M}, with mass
  M=|∂ N|_h^{1/2}/√(16pi).

(T4) Charged Riemannian Penrose inequality
  (Huisken--Ilmanen 2001; Bray--Khuri 2010; Disconzi--Khuri 2012). The
  same setup with total charge Q=(Q_E^2+Q_B^2)^{1/2} and P=0 gives
  E_hge Q and rle m+√(m^2-Q^2), with r=(A/4pi)^{1/2} and
  m=E_h.
pdf/disconzi-khuri-2012.pdf

Everything else (strictification, rest-frame reduction, the coupled
elliptic construction, the causal stationary field, the charge and
four-dimensional extensions, the anti-de Sitter analysis) is proved in
full below. Standard elliptic PDE (Schauder,
Calder'on--Zygmund) and geometric measure theory (compactness of
minimizing boundaries) are used as tools without restatement.


Setting and conventions [sec:setting]. Initial data and the constraint equations



Throughout, (M,g,K) is a smooth connected orientable three-manifold with
a complete Riemannian metric g and a symmetric (0,2)-tensor K, the
second fundamental form of a spacelike slice. The energy density mu and
momentum density J are determined by the constraint equations [eq:constraints]

  16pimu=R_g+(tr_gK)^2-|K|_g^2,
  
  8pi J=div_g(K-(tr_gK)g),

where R_g is the scalar curvature. The dominant energy condition (DEC)
is the pointwise inequality muge|J|_g.

An end of M is a neighborhood of infinity diffeomorphic to
ℝ^3∖barbar B_1,B_1, parametrized by coordinates x^i in which [eq:decay]

  g_{ij}-delta_{ij}=O_2(r^{-q}),
  
  K_{ij}=O_1(r^{-1-q}),
  
  q>tfrac32,

with r=|x| and O_k denoting k derivatives bounded by a constant
times a power of r. The ADM energy and momentum are [eq:adm]

  E=(1)/(16pi)lim_{rto∞}int_{S_r}(g_{ij,i}-g_{ii,j})nu^j,d A,
  
  P_j=(1)/(8pi)lim_{rto∞}int_{S_r}(K_{ij}-(tr_gK)delta_{ij})nu^i,d A,

where S_r are the coordinate spheres, nu their outward Euclidean
unit normal, and the index i is summed. The limits exist by
[eq:decay]. P_delta denotes the momentum covector paired
with the coordinate frame. When E>|P|_delta, the invariant mass of
the end is m=√(E^2-|P|_delta^2).


MOTS and expansions [sec:mots]. Let Ssubset M be a compact embedded two-sided surface disjoint from
∂ M, with a chosen unit normal nu pointing into the
designated exterior. Its mean curvature is H_S=div_Snu, with the
convention H_S=2/r for a coordinate sphere of radius r in ℝ^3,
and tr_SK=g^{ab}K(X_a,X_b) in an orthonormal tangent frame
{X_a}. The future and past outward null expansions are [eq:null-expansions]

  theta_pm(S)=H_Spmtr_SK.

The surface S is weakly future outer trapped when theta_+(S)le0,
and a future marginally outer trapped surface (MOTS) when
theta_+(S)=0.


Exteriors, cuts, and the minimum enclosing area [sec:cuts]. Let Ω be a connected open manifold whose closure barbarΩΩ is
complete for the restriction of g, with smooth compact boundary
B=∂Ω and exactly one end, asymptotically flat.

An enclosing cut for the chosen end is a compact smooth embedded
(possibly disconnected) surface GammasubsetbarbarΩΩ that bounds a
filled side D^+(Gamma)subsetbarbarΩΩ containing the entire distant
part of that end. The complement barbarΩ∖Ω∖ D^+(Gamma)
contains B. Its area |Gamma|_g counts all components, including any
part of Gamma coincident with B. Define [eq:amin]

  a_g(B)=infbigl{|Gamma|_g:Gamma an enclosing cutbigr}.

The infimum equals the finite-perimeter closure taken in
the [sec:perimeter] section.


Outermostness and area-minimality [sec:outermost]. Two conditions are imposed only in the equality theorem.

Outer area-minimality. B is outer area-minimal when
|Gamma|_gge|B|_g for every enclosing cut Gamma.

Partial-coincidence outermostness. Let mathcal TcalT be the class
of enclosing cuts that attain a_g(B). B is partially outermost if
every Gammainmathcal TcalT with Gammane B omits at least one entire
component of B.


Regularity and finiteness [sec:fin]. The hypotheses [eq:decay] are the standing
assumption in three dimensions; in four dimensions the analogous
O_2 and O_1 decay with the same exponent is assumed. Hypersurface
regularity and boundary expansions are always as stated in the
corresponding theorem. All areas, fluxes, and currents are taken with
respect to the given g.


Normalizations



Throughout, c=G=1 and the coefficients 16pi are used in [eq:constraints]
and [eq:amin]. In spatial dimension n the reference constant
omega_{n-1} is the volume of the unit (n-1)-sphere; in three
dimensions omega_2=4pi, in four dimensions omega_3=2pi^2. A
coordinate sphere of radius r in ℝ^n has area
omega_{n-1}r^{n-1}, and a future MOTS sphere in Schwarzschild
spacetime of mass m satisfies m=tfrac12(A/omega_{n-1})^{(n-2)/(n-1)}.


The three-dimensional exterior inequality [sec:n3]. Statement [sec:n3-statement]




Theorem (Minimum enclosing area in three dimensions). [thm:n3:exterior]
Let (Ω,g,K) be a connected orientable smooth three-dimensional
exterior with smooth compact boundary B=∂Ω, complete with
its boundary included, whose unique end is asymptotically flat in the
sense of [eq:decay]. Assume that the
constraints [eq:constraints] satisfy muge|J|_g with mu and
|J|_g integrable, and that the ADM limits [eq:adm] exist.
Orient B by the normal into Ω and suppose theta_+(B)le0.
If E>|P|_delta, then


  √(E^2-|P|_delta^2)
     ge√((a_g(B))/(16pi)).


No extension of the data across B is required.




The proof assembles three ingredients: a conformal deformation that
makes the ADM vector timelike without disturbing B
([sec:n3-reduction]), a preparation of the data to a strict
end-replaced form ([sec:end-preparation]), and the geometric and
elliptic comparison with the Riemannian Penrose inequality
([sec:geometry]--[sec:riemannian-penrose]).


Conformal strict direction [sec:n3-reduction]. Lemma (Strict direction). [lem:n3:strict] 
Let (Ω,g,K) satisfy all the hypotheses of
the [thm:n3:exterior] theorem except possibly E>|P|_delta. There is a
smooth family of conformal data
g_s=(1+sphi)^4g, K_s=(1+sphi)^{2}K, defined for
sin[0,s_0), with phige0 smooth and O(1/|x|) at infinity, such
that


  E_s=E+sA_phi, P_s=P, A_phi>0,
  a_{g_s}(B)ge a_g(B),


and (Ω,g_s,K_s) satisfies DEC and
theta_+^{g_s,K_s}(B)le0 in the same decay class.





Proof. 
Let wge0 be smooth, compactly supported in a small ball disjoint from
B and from a neighborhood of the end, with wnotequiv0. Let G be
the Green kernel of L_g=-Delta_g+tfrac18R_g. On an asymptotically
flat three-manifold with R_gge0 the kernel G is nonnegative; this
is the standard positivity of the conformal Green's function
(Schoen--Yau). Put phi=Gwge0, so phi is smooth, positive on
supp w, and decays like O(1/|x|).

Write u_s=1+sphi, so u_sge1 and u_sto1 at infinity. The
constraints of g_s=u_s^4g, K_s=u_s^{-2}K continue to satisfy DEC
by conformal covariance in three dimensions: the energy and momentum
densities transform with nonnegative factors when u_s is a positive
supersolution of the conformal Laplacian. The ADM energy formula


  E_{g_s}=E+scdot(1)/(16pi)lim_{rto∞}int_{S_r}
            ∂_nu(4u_s∂_nu u_s),d A
  =E+sA_phi


with A_phiproptoint w>0 and P_s=P follows from the standard
second variation of the mass under gmapsto u_s^4g.

Because supp w is disjoint from a neighborhood of B, the normal
derivative of u_s vanishes there, and conformal covariance of the
mean curvature and trace gives


  H^{g_s}_B+tr^{g_s}_BK_s
  =u_s^{-3}(H_B^g+tr_B^gK),


so theta_+^{g_s,K_s}(B)le0 is preserved.

Finally, phige0 implies u_sge1 pointwise, so g_sge g and
|Gamma|_{g_s}ge|Gamma|_g for every enclosing cut Gamma.
Taking infima gives a_{g_s}(B)ge a_g(B).





The strictified exterior inequality [sec:n3-strictified]. The strict case is also needed, proved in full in Sections~[sec:end-preparation]
through~[sec:existence].


Theorem (Strictified exterior inequality). [thm:n3:strict-exterior]
Let (Ω,g,K) satisfy all the hypotheses of
the [thm:n3:exterior] theorem together with
- K compactly supported;
- theta_+(B)<0 and P=0;
- the strict gap
        mu-|J|_gge c_0r^{-4-delta_1} for some c_0>0 and
        0<delta_1<1.
Then


  m=Ege√((a_g(B))/(16pi)).






The proof of the [thm:n3:strict-exterior] theorem builds a comparison
metric g_star on the strictified exterior with [eq:comparison]

  R_{g_star}ge0,
  m_{ADM}(g_star)le m+ε,
  a_{g_star}(B)ge(1-ε)a_g(B),

for every ε>0; the Riemannian Penrose inequality
(the [thm:riemannian-penrose] theorem below) then closes the argument.
The construction occupies Sections~[sec:end-preparation]--[sec:existence].


Completion of the exterior inequality [sec:n3-completion]. Suppose first E>|P|_delta and B is oriented so
theta_+(B)le0; if some components need K replaced by -K, do so
and proceed. Fix ε>0. The [sec:end-preparation] section and
the [lem:n3:strict] lemma produce strictified data (g',K') satisfying


  m'le m+ε, theta'_+(B)<0,
  a_{g'}(B)ge(1-ε)a_g(B),


and identical decay class. The [thm:n3:strict-exterior] theorem gives
m'gesqrt{a_{g'}(B)/16pi}, so


  √((a_g(B))/(16pi))
   lefrac{m'}{√(1-ε)}
   lefrac{m+ε}{√(1-ε)}.


Let ε↓0 to conclude.


Strictification and the conformal constraint formulas [sec:strictification]. To apply the [thm:n3:strict-exterior] theorem, data in strict form is needed:
K compactly supported, theta_+(B)<0, and a positive density gap. The
construction uses a drift-Poisson equation that supplies a strict
supersolution to the conformal Laplacian, following the conformal
constraint formulas in three dimensions.


Conformal constraint formulas [sec:n3-conformal]. Lemma (Conformal constraint formulas). [lem:n3:conformal-formulas] 
Let u>0 be smooth and set g_u=u^4g, K_u=u^2K. Then

16pi u^4mu_u &=16pimu-8u^{-1}Delta_gu-(1-u^{-8})((tr_gK)^2-|K|_g^2),
[eq:conformal-mu]
8pi J_u &=u^{-2}(8pi J+4K(nablalog u,cdot)), [eq:conformal-J]

the second being an identity of covectors. Consequently [eq:conformal-dec]

16pi u^4(mu_u-|J_u|_{g_u})ge
16pi(mu-|J|_g)+8(-Delta_glog u-|dlog u|_g^2-|K|_g|dlog u|_g)
  -|1-u^{-8}|,|(tr_gK)^2-|K|_g^2|.

For every consistently oriented hypersurface with unit normal nu, [eq:conformal-expansion]

theta_{+,u}=u^{-2}(theta_++4∂_nulog u).






Proof. 
In dimension three
R_{u^4g}=u^{-4}(R_g-8Delta_glog u-8|dlog u|_g^2), while the
constraint quadratic term scales with an extra factor of u^{-8} because
the extrinsic curvature is weighted by u^2:
u^4((tr_{g_u}K_u)^2-|K_u|^2_{g_u})=u^{-8}((tr_gK)^2-|K|_g^2).
Adding the two contributions gives [eq:conformal-mu]. With pi=K-(tr_gK)g, pi_u=u^2pi and
tr_gpi=-2tr_gK; contracting the connection difference in dimension
three gives
u^2J_u=J+2pi(nablalog u,cdot)-(tr_gpi)dlog u
        =J+4K(nablalog u,cdot),
which is [eq:conformal-J]. Norms in g_u add a factor u^{-2},
and the triangle inequality yields [eq:conformal-dec]. Finally
nu_u=u^{-2}nu, H_{u^4g}=u^{-2}(H+4∂_nulog u), and
tr_{B,g_u}K_u=u^{-2}tr_{B,g}K, which is [eq:conformal-expansion].





The strictification lemma [sec:n3-strictification]. Lemma (Strictification). [lem:n3:strictification] 
Let (N,g,K) satisfy [eq:decay] with q>tfrac32, DEC, and
integrability, with compact boundary B and theta_+(B)le0.
Fix 0<delta<beta<min(q,1). There are smooth positive w,phi on N
with w=r^{-3-delta} on the end, such that [eq:strict-data]

∂_nuphi=-1 on B,
phi=O_2(r^{-1}),
-Delta_gphi-|K|_g|dphi|_gge w,

and the outer flux L_phi=lim_{rto∞}int_{S_r}∂_rphi,d A_delta
is finite and negative. For every s>0, the data g_s=u_s^4g,
K_s=u_s^2K with u_s=1+sphi satisfy:
(i) DEC with strict gap
      mu_s-|J_s|_{g_s}ge c_sr^{-3-delta} for c_s>0;
(ii) theta_+^{g_s,K_s}(B)<0;
(iii) decay class min(q,1)>tfrac12;
(iv) E_s=E+sA_phi, P_s=P, A_phi=-L_phi/(2pi)>0, and
      a_g(B)le a_{g_s}(B)le(1+s|phi|_∞)^4a_g(B).





Proof. 
Take bge|K|_g smooth and positive, equal to b_0r^{-1-beta} on the
end, with b_0>0. Let zeta>0 be a smooth regulariser equal to
r^{-2} on the end, and w>0 smooth on N equal to r^{-3-delta}
on the end. The drift-Poisson equation [eq:drift-poisson]

-Delta_gphi=b√(|dphi|_g^2+zeta^2)+w,

∂_nuphi=-1,quadphito0,

is uniformly elliptic and smooth even at critical points of phi
because of zeta. There is a smooth nonnegative solution with
phi=O_2(r^{-1}).

Existence. Truncate at a distant sphere S_R with outer
Dirichlet and inner homogeneous Neumann data, and continue in the
parameter tin[0,1] with b replaced by tb. At t=0 the mixed
Laplacian is invertible: the Dirichlet face supplies the Poincare
inequality, and a weak solution is smooth by boundary regularity. The
linearisation at any smooth solution carries a bounded drift of norm
at most b; the strong maximum principle and the boundary Hopf
principle force a trivial kernel, so the operator is an invertible
Fredholm perturbation. Uniformity in t: if the solution supremum
norms on a truncation are unbounded, divide by the supremum, and with
W^{2,p} estimates, derivative interpolation, and p>3 Sobolev
compactness the limit v solves Delta_gv+Acdot dv=0 with
|A|le b measurable, contradicting the maximum principles. The
bounds are independent of R, yielding a smooth solution of
[eq:drift-poisson] on the exterior.

Sign and end decay. The right side of [eq:drift-poisson]
is positive, so a negative minimum cannot lie in the interior, and the
Hopf principle at a boundary minimum requires ∂_nuphile0
in the outer normal into N, whereas the present data give
∂_nuphi=-1; hence phige0. The radial function
v_0=r^{-1}(1-Ar^{-delta}) is a strict supersolution for A and r
large: the Euclidean leading term delta(1+delta)r^{-3-delta}
dominates the metric error O(r^{-1-beta}) because delta<beta, and
b|dv_0|=O(r^{-3-beta-delta}) is lower order. Comparison on the end
gives phile Cv_0, so phi=O(r^{-1}) and
∂_rphi=-r^{-2}+o(r^{-2}); then L_phi<0 and A_phi>0.

Strict conclusions. At s=0,
-Delta_glog u_s=-Delta_gphi
=b√(|dphi|^2+zeta^2)+wge w, and by continuity the same holds for
all small s>0; the density gap in [eq:conformal-dec] follows. On
B, ∂_nulog u_s=-s<0, so
theta_+^{g_s,K_s}(B)=u_s^{-2}(theta_+(B)-4s)<0. The energy formula
and the two-sided area bound are the standard conformal variance
together with u_sge1 and u_sle1+s|phi|_∞.





Remark (Decay needed by the drift-Poisson source). 
The source w=r^{-3-delta}>0 of [eq:drift-poisson] must, at the
end of the strictification step, also dominate the conformal residual
from the [lem:n3:conformal-formulas] lemma. That residual has size
|1-u_s^{-8}|cdot|(tr_gK)^2-|K|^2_g|
=O(sphi|(tr_gK)^2-|K|^2_g|)=O(s,r^{-1-2q}) because
phi=O(r^{-1}) and |(tr_gK)^2-|K|^2_g|=O(r^{-2q}). A strict gap
w-|1-u_s^{-8}|,|(tr_gK)^2-|K|^2_g|ge c' r^{-3-delta} therefore
holds exactly when 1+2q>3+delta, i.e., q>1+delta/2. This is
why the primary decay hypothesis [eq:decay] is stated as q>3/2
rather than q>1/2: q>3/2 gives q>1+delta/2 for every
0<delta<1, so the strictification step is fully honest.





Replacement of the end by a rest-frame model [sec:end-preparation]. The comparison metric of Sections~[sec:geometry]--[sec:existence]
operates most cleanly on one-ended data with zero ADM momentum and
compactly supported tensor, satisfying theta_+(B)<0 and a strict
density gap. Starting from arbitrary data satisfying the hypotheses of
the [thm:n3:exterior] theorem with E>|P|_delta, such data are produced
with energy and area converging to the originals.


Positive enclosing area [sec:n3-positive-area]. Lemma (Positive enclosing area). [lem:n3:positive-area] 
Every exterior (Ω,g,K) in the [thm:n3:exterior] theorem satisfies
0<a_g(B)<∞.





Proof. 
Finiteness: a distant coordinate sphere S_R in the chart of
[eq:decay] is an enclosing cut of area omega_2R^2+o(R^2), so
a_g(B)<∞. Positivity: fix a small disk D in a component of B
and a smooth proper embedded ray gamma from the center of D to the
chosen end, normal to B at its base and straight in the asymptotic
coordinates outside a compact set. Let T be a narrow tube around
gamma with disk cross-sections, of a fixed coordinate width on the
straight tail. Let omega be a smooth closed two-form supported in
T, pulled back from a transverse disk form, with
int_{S_R}omega=1 and bounded L^∞-norm. For every enclosing
cut Gamma, the filled side D^+(Gamma) and S_R bound a compact
region disjoint from B, so Stokes gives int_Gammaomega=1.
Consequently 1le|omega|_∞|Gamma|_g, and taking infima over
cuts gives a_g(B)ge1/|omega|_∞>0.





Removal of momentum and compact support [sec:n3-rest-frame]. Proposition (Rest-frame reduction). [prop:n3:rest-frame] 
Suppose (Ω,g,K) satisfies the hypotheses of
the [thm:n3:exterior] theorem with E>|P|_delta and invariant mass m.
For each ε>0, there is one-ended exterior (Ω',g',K') with
smooth compact boundary B and:
(i) (Ω',g',K') agrees with (Ω,g,K) in a neighborhood of
      B and is supported on a subset of Ω differing from it by
      a compact end-piece diffeomorphism;
(ii) DEC with strict gap mu'-|J'|_{g'}ge c'r^{-3-delta'} for
      c'>0 and some 0<delta'<1;
(iii) theta_+^{g',K'}(B)<0;
(iv) K' is compactly supported;
(v) P'=0 and m'=E'=m+O(ε);
(vi) a_{g'}(B)ge (1-ε)a_g(B).





Proof. 
The first step is strictification. By the [lem:n3:strictification] lemma
at parameter s>0, the data (g_s,K_s)=(u_s^4g,u_s^2K) satisfy DEC
with a density gap, satisfy theta_+^{g_s,K_s}(B)<0, and have
a_g(B)le a_{g_s}(B)le(1+s|phi|_∞)^4a_g(B). Choose s>0
small enough that (1+s|phi|_∞)^4le 1+ε/2. Write
E_s=E+sA_phi.

The second step is the asymptotic Lorentz boost. Since E>|P|_delta,
the vector (E,P) is future timelike, and there is a unique future
boost Λ such that Λ(E,P)=(m,0) with
m=√(E^2-|P|_delta^2). On the asymptotic chart of [eq:decay],
the change of coordinates xmapstoΛ x acts linearly on the end
and is trivial on a fixed neighborhood of B. With respect to
(g_s,K_s) it produces isometric data with the same DEC, the same
theta_+(B)le0, and ADM vector Λ(E_s,P)=(m_s,0), where
m_s=√(E_s^2-|P|_delta^2).

The third step is compact support. Interpolate K_s across a distant
annulus Rle rle 2R between its values near B and zero on the far
tail, while keeping the metric unchanged. The interpolation introduces
a constraint violation bounded by O(R^{-2}) on the annulus, and a
conformal repair u_Rge1 supported on Rle rle 3R with
u_Rto1 on the far tail restores DEC, with strictly positive local
gain on the far end and a mass cost of O(R^{-2})=O(R^{-2}); the
existence of such a repair is [sec:strictification]'s
generalisation to a compactly supported source (the drift term is
absorbed by the same truncated maximum-principle argument, and the
outer flux remains negative and of size -4pi).

Combining the three steps and letting Rto∞ gives (iv) and (vi).
Property (v) follows from sto0: E_s=E+O(ε) for small s, and
m'=m_s=m+O(ε); the two loss factors
(1+s|phi|_∞)^4 and the O(R^{-2}) repair cost are absorbed
into ε by taking s and R^{-2} small enough. Property (ii)
follows from the [lem:n3:strictification] lemma(i) and the boost.
Property (iii) is preserved under both the boost (which acts on the
end) and the compact support step (which acts on the end).





The strict exterior [sec:n3-strict-exterior-reduced]. Corollary (Strict exterior with zero momentum and compact
support). [cor:n3:strict] 
Under the hypotheses of the [thm:n3:exterior] theorem with
E>|P|_delta, for every ε>0 there exists data satisfying all the
strict hypotheses of the [thm:n3:strict-exterior] theorem with
m'=m+O(ε) and a_{g'}(B)ge(1-ε)a_g(B).





Proof. 
Immediate from the [prop:n3:rest-frame] proposition.





Geometric preparation: two colors and collars [sec:geometry]. The comparison metric of the [sec:elliptic] section is built on an
exterior cut off at a distant sphere and prepared so that its two
families of boundary faces have strict negative expansions, one for K
and one for -K. The mechanism is the trapped-region construction of
Andersson--Eichmair--Metzger, employed here at a threshold near zero.


Total regions inside a compact barrier




Definition (Total region). [def:total-region] 
Let (X,Q) be a compact smooth Riemannian three-manifold with connected
smooth outer boundary S_0 and smooth compact inner boundary
∂ X∖ S_0, and let DsubsetintX be a
closed set on which Q vanishes outside a fixed neighborhood. For
cinℝ define the total region


T_c(X,Q)=∩{E: E closed,,Dsubset E,,∂ E
 smooth,, H_{∂ E}+tr_{∂ E}Qle c on ∂ E}.







Lemma (Compact total regions). [lem:n3:compact-total] 
Suppose there is a smooth closed F with DsubsetintF
and Θ_Q(∂ F)<c for the outward normal. Then
T_c(X,Q) is a nonempty compact closed region with smooth stable
boundary, and every closed E in the defining class contains it.





Proof. 
As an external input, the smooth-frontier theorem of Andersson--Eichmair--Metzger is used: if a complete 3-manifold with
nonnegative scalar curvature has a compact smooth weakly future
trapped surface, then the outermost weakly outer trapped region has
smooth stable boundary, and its compact components are the smooth
frontiers of the compact trapped subregions. Pasting outward cobordisms and an open S^3times[0,∞) extends
X to a complete boundaryless data set. A smooth Morse height q with
q=0 on ∂ X, increasing in an outward collar, and radial on the
tail dominates any closed weakly trapped region: it forces it to remain a
positive distance inside X by the maximum-principle argument applied to
q on the region. The Andersson--Eichmair--Metzger smooth-frontier
theorem then produces the outermost smooth stable frontier, which is
the closed intersection.





Lemma (Threshold continuity). [lem:n3:threshold-continuity] 
On any open interval of thresholds c where the strict-barrier
condition of the [lem:n3:compact-total] lemma holds, the closed regions
T_c(X,Q) increase with c and are right-continuous in Hausdorff
distance except at countably many thresholds.





Proof. 
Monotonicity is immediate from the definition. Fix a countable dense
set {x_j} in X. For each j, the monotone function
cmapstodist(x_j,T_c(X,Q)) has at most countably many jumps. Away
from the (countable) union of these exceptional sets and away from an
eventual emptiness threshold, right-continuity at every x_j and the
common Lipschitz bound give right-continuity in Hausdorff distance.





The two-color exterior




Proposition (Two-color preparation). [prop:n3:two-color] 
Let (Ω,g,K) be strict data satisfying the hypotheses of
the [thm:n3:strict-exterior] theorem. Fix eta_0>0 small. There is a
closed connected exterior Ω_0subsetΩ with one
asymptotically flat end, nonempty compact smooth boundary B_0,
decomposition B_0=B_{rm black}cup B_{rm white}, and constants
c_b,c_win(-eta_0,0) such that, for the normal nu_0 into Ω_0
and P_{B_0}=tr_{B_0}K,


H_{B_0}+P_{B_0}=c_b on  B_{rm black},

H_{B_0}-P_{B_0}=c_w on  B_{rm white}.


On each face there is a smooth disjoint collar (B_0^jtimes[0,1),s)
with |d s|>0 and monotone expansion toward Ω_0.





Proof. 
Choose a distant sphere S_{R_0} with positive mean curvature for
both K and -K, beyond the support of K. Fill behind B to
obtain a compact X with outer boundary S_{R_0} and inner boundary
B. Choose c_bin(-eta_0,0) at a right-continuity threshold of
T_c(X,K), and put (Y=X∖intT_{c_b}(X,K)).
All boundary components of Y have strictly positive expansion for
-K, so a second right-continuity threshold c_win(-eta_0,0) of
T_c(Y,-K) gives a nonempty closed region. Let
Ω_0 be the closure of the exterior component of
X∖(T_{c_b}(X,K)cup T_{c_w}(Y,-K)). Its retained c_b-faces
are black and its retained c_w-faces are white, with the stated
expansions for the normal into Ω_0 (the signs follow from the
orientations of the total region and its complement). The stable
collars are the Andersson--Eichmair--Metzger stability directions
shrunk to small amplitude.





The coupled elliptic comparison system [sec:elliptic]. Fix a two-color exterior Ω_0 of the [prop:n3:two-color] proposition.
Its boundary B_0=B_{rm black}cup B_{rm white} carries two strict
negative boundary expansions; and its far end is asymptotically flat at
infinity.


Variables and the equations



Fix a smooth boundary height h_*:B_0toℝ with
h_*=b_-<0 on black and h_*=b_+>0 on white, extended smoothly to
Ω_0 with ∂_nu h_*<0 on B_0 and h_*to0 at
infinity. Fix eta_n=e^{-ε n}>0.

Let l:ℝto(0,∞) be smooth, with l(t)=e^{nt} for tle0 and
l(t) constant between 1 and e for tge1/n; set p=l'/l, so
0le ple n and p is compactly supported. For a C^2 function
f:Ω_0toℝ and t:Ω_0toℝ, define

&sigma=nabla f, D=1+l(t)^2|sigma|^2, d=D^{-1},
Z=t+tfrac16log D, h=eta_n f,
&w=(l,sigma)/(sqrt D), a=|w|, v=a^2,
chi=(3d)/(3+pv),
A_d=I-w ⊗  w,
A_chi=I-(3+p)/(3+pv),w ⊗  w.

Set L_0=e^{2t}l and u=L_0sqrt D.

The comparison metric on Ω_R=Ω_0cap{r<R} is [eq:comparison-metric]

barbar g_g_{ij}=(1)/(d)(g_{ij}+l(t)^2∂_if,∂_jf),

the graph metric over the base (Ω_R,g) with height f and warp
l, rescaled by the lapse 1/d.

The unknowns (f,t) satisfy the coupled system [eq:coupled-system]


Div_g(3u,A_chi,nabla t)=u,Xi &in Ω_R,
1mm]
l,Delta_g f+(bulk terms)=eta_n,Xi_2&in Ω_R,
1mm]
t=h_*/eta_n on B_0, t=0 on S_R,


where Xi,Xi_2 are smooth functions of (f,t) and their first
derivatives fixed by the two curve identities below. The exact
right-hand sides encode the coupling to K and to the two-color
boundary expansions of Ω_0.


Lemma (The Hessian square is coercive). [lem:n3:hessian-square] 
Put H^f=Hess_gf and let
mathcal T=calT=tfrac12{(tr_{A_d}H^f)^2-|H^f|_{A_d ⊗ A_d}^2} be
the second-form square of the comparison graph. Then there is c_0(n)>0
such that [eq:hessian-square-coercive]

mathcal TcalTge c_0((tr_gH^f)^2+|nabla w|_{A_chi}^2)

and mathcal TcalTle C(n)(|nabla f|^2+|Hess_gf|^2).





Proof. 
Write A_d and A_chi in a frame where A_d=diag(1,1,1,d) and
A_chi=diag(1,1,1,chi). Direct contraction gives


mathcal T=calT=tfrac12|H^f_perp|^2+chi,(H_{ee}^f)^2,


where H^f_perp is the trace-free part on the two transverse directions
and H_{ee}^f is the leading axial entry. Bounding
chige 3d/(3+n)ge c(n)>0 from below and using the identity
chi(1-chi)=O(vd) for the off-diagonal entries, the lower bound
[eq:hessian-square-coercive] follows. The upper bound is
immediate from the definition.





Lemma (Boundary identity). [lem:n3:boundary-identity] 
On a face B_0^j of the boundary B_0, with H the mean curvature
into Ω_0, P_{B_0^j}=tr_{B_0^j}K, and w_nu=w(nu), the
first normal variation of t satisfies [eq:boundary-identity]

V_nu/u=3d,∂_nu t+w_nu(F-P_{B_0^j}),

where F=tr_gK.





Proof. 
The metric and divergence formulas give, for the covariant trace
V_nu of the boundary flux of t,


V_nu/u=3d,∂_nu t+w_nu(F-P_{B_0^j}),


the first term from the direct variation of t across the boundary and
the second from the tangential trace of K. Using the boundary expansion
H_{B_0^j}+P_{B_0^j}=c_j, this becomes a strict negative first-variation,
compatible with the Dirichlet data on ∂Ω_R.





A priori bounds




Proposition (A priori bounds). [prop:n3:apriori] 
For every n there is a constant C(n)<∞ such that every smooth
solution (f,t) of [eq:coupled-system] satisfies


-εle tle Zle C(n),
|f|+|nabla f|+|Hess_gf|le C(n)
quadon barbarΩ_R.Ω_R.






A quantity Pi_n below is bounded by C(n+1)^N, with C,N depending
only on fixed data, n, ell, and the fixed collars; another quantity
C(n) may also depend on the truncation radius in a way that will be
declared explicitly. The parameter comparisons


  ell=e^{-ε n}, tau=ell^{3/2},
  ellle lle e, 0le ple n,
  (4d)/(4+n)lechile d


enter through the small factor Pi_ntau/ellto0, which absorbs only
polynomial losses.


Lemma (Uniform separation). [lem:n3:apriori:separation] 
Let AsubsetbarbarΩΩ be compact. In the first two homotopy stages
one has the following. If A avoids the full black total region
mathcal B_calB_{c_b} at threshold c_b, then hge b_-+eta_A on A
for some eta_A>0, uniformly in n large, every sufficiently large
truncation, and every admissible solution. If A avoids
mathcal W_calW_{c_w}, then hle b_+-eta_A.





Proof. 
Take an arbitrary sequence n_ito∞, R_ito∞ of admissible
stage-one or stage-two solutions. The elementary height and end bounds
from the preparation give
|h_i|le C_0 and |h_i|le C(r^{-gamma}-R_i^{-gamma}) on the end,
for some gammain(3/4,1). Extend h_i across a white face by the
constant b_+. Define the lower relaxed limit
underline h(x)=liminf_{yto x} h_i(y) over such sequences.

Fix b_-<k'<min{0,b_-+eta} with eta the low-height sign range
of the add-on D_0. Suppose a smooth phi touches underline h
from below at x with underline h(x)le k' and dphi(x)ne0.
After a small fourth-order localization, minimization on a small closed
ball produces x_ito x and c_ito0 with phi+c_i a strict lower
contact of h_i at x_i. By the choice of k'<b_+, such contacts
occur in the interior trace equation, never on a white face or on its
constant extension. At the contact,


  nabla f_i=tau^{-1}nablaphi,
  d_ile(tau^2)/(ell^2|dphi(x_i)|^2)to0,
  a_ito1, chi_ito0.


Positivity of A_{chi_i} and the Hessian inequality at a lower
contact give


  tr_{A_{chi_i}}K+
  frac{l_i}{tau√(D_i)}tr_{A_{chi_i}}Hessphi
  le F_i(x_i).


Since D_0le0 at these low values, limsup F_i(x_i)le k'+C_b, so


  frac{tr_{(dphi)^perp}Hessphi}{|dphi|}
  +tr_{(dphi)^perp}Kle k'+C_b.


This is the oriented level expansion of phi towards increasing
phi, and the localization can be removed. The only Hessian in this
passage is the fixed test's.

Transfer to the closed sublevel E_k={underline hle k} for
b_-<k<k'. For integers j,


  psi_j(s)=frac{1}{1+e^{-j^2(s-k-j^{-1})}},
  v_j=psi_j(underline h).


The lower relaxed limit of v_j is the indicator of E_k: at points
of E_k it vanishes, off E_k it converges uniformly to one. If
varphi has nonzero gradient and is a lower test of the indicator at
a zero value, the same strict contact localization gives lower tests
varphi+c_j of v_j at y_jto x with v_j(y_j)to0. Since
psi_j(k')to1, one has underline h(y_j)<k' eventually.
Composition with the increasing psi_j preserves the oriented level
expansion:


  Hess(thetacircvarphi)=theta'Hessvarphi
    +theta''dvarphi ⊗  dvarphi,
  frac{tr_{(dvarphi)^perp}Hess(thetacircvarphi)}
       {|d(thetacircvarphi)|}
  =frac{tr_{(dvarphi)^perp}Hessvarphi}{|dvarphi|}.


So psi_j^{-1}(varphi+c_j) is a smooth lower test of
underline h with contact <k', and the oriented-level inequality
passes to varphi. Hence every smooth exterior support of E_k has
expansion at most k'+C_b.

The argument also rules out a hidden layer at a white face: the
reversed white face would be an exterior support with expansion
-H_B+P_B=-c_w>0, contradicting k'+C_b<0 for k' close to b_-.
The end bound makes E_k compact.

Adjoin the entire black total region and fill any bounded complementary
pockets. At a point of the black frontier, an exterior support of the
union contains the smooth black region locally, so tangency bounds the
expansion by c_b<k'+C_b. Everywhere else the sublevel argument
applies. The resulting compact closed set contains the original inner
barrier with a collar and avoids the outer barriers. Lemma on weak
support then places it in a smooth black trapped region at any
threshold c''>k'+C_b close enough to c_b, hence in
mathcal B_calB_{c''}.

Hausdorff right continuity of cmapstomathcal B_ccalB_c at c_b supplies
delta>0 with mathcal B_calB_{c_b+delta}cap A=∅; shrink
delta so the enlarged threshold is still negative. Choose
k,k',c'' with b_-<k<k'<k'+C_b<c''<c_b+delta. Then
E_kcap A=∅ for every sequence in the class. A failing
uniform estimate would produce x_iin A with h_i(x_i)le k and a
subsequence x_ito xin A, so underline h(x)le k, contradiction.
This gives hge b_-+eta_A on A. Replacing (h,K,C) by
(-h,-K,-C) and interchanging colors gives the upper assertion.





Lemma (Signed collar slopes). [lem:n3:apriori:collars] 
There are fixed c,C_*>0 such that, for large n and every allowed
truncation, the first two stages satisfy


  c/taule∂_nu fle C_*/tau
   on black faces,
  c/taule-∂_nu fle C_*/tau
   on white faces,


and the third stage satisfies |∂_nu f|le C_*/tau.





Proof. 
Write s for the leaf parameter in an outward collar and
N=nabla s/|ds|. At a comparison contact with h_0=b_-+psi(s)
(psi'>0), the shared gradient gives


  dle Ctau^2/(ell^2(psi')^2),
  1-ale d, nchige c_1 dquad(nge4).


With slopes in a fixed positive interval this is dle Cell.
The trace operator evaluated on this test is


  P_s+aH_s+chi K(N,N)+
  achi!((Hess s(N,N))/(|ds|)
            +|ds|(psi'')/(psi')),


where P_s=tr K on the leaf tangents and H_s its mean curvature.
For a negative slope the terms with prefactor a flip sign. On a
black collar H_s+P_sge c_b and C=C_b-k_Bs. Choose small positive
slopes with psi(0)=0 and psi(s)le k_Bs/2. A fixed short collar
and the [lem:n3:apriori:separation] lemma let the test lie below the
solution at its outer end; the add-on D_0 vanishes in the compatible
direction. The residual of the trace equation on the test is at least


  k_Bs/2-Cd+achi|ds|psi''/psi'.


Pick a smooth beta_n(s) equal to An for 0le sle1/n, supported
in [0,2/n], with intbeta_nle2A. Set


  psi'(s)=mexp!(int_0^sbeta_n(q)d q).


Fix A so the last term dominates Cd on sle1/n, using
age1/2 there; on sge1/n the linear term dominates
Cdle Cell for large n. The slope multiplier is at most
e^{2A}, so a fixed small m>0 preserves all previous slope bounds.
This gives a strict lower barrier with slopes in [me^{-A},me^{A}]
independent of n.

An upper barrier from b_- uses
psi'(s)=Mexp(-int_0^sbeta_n(q)d q). The leaf expansion differs
from its boundary value by O(s); take M so large that
psi(s)ge C_2s dominates this, the offset variation, and the outer
height bound. The logarithmic-slope term is negative: on sle1/n it
dominates Cd, and beyond s=1/n the s-margin does. A strict upper
barrier results, with slopes in [Me^{-2A},M].

At a negative interior minimum of h-h_0, Hessians compare, Z agrees,
and t,l,A_chi agree; height monotonicity of the right side turns
positive residual into a contradiction. Positive maxima reverse signs.
The one-sided derivative at s=0 proves the black part; replacing
(h,K,C) by (-h,-K,-C) proves the white part. In the third stage
the boundary height is an interpolated constant b_zeta; tests
b_zetapmpsi(s) with large psi' and
(logpsi')'=-beta_n give nonpositive/nonnegative residuals at s=0
by the directional inequalities of the homotopy. The same dle Cell
bound and a fixed test slope handle the rest. This gives the two-sided
third-stage bound.





Lemma (Weighted collar traces). [lem:n3:apriori:trace] 
Let mathcal CcalC be a fixed disjoint union of short inner collars. In
the first two stages, for nonnegative smooth varphi supported in
mathcal C,calC,


  int_B Lvarphid A_gle
  Pi_nint_{mathcal C}calCu[(1+√(mathcal T))calT))varphi
       +|dvarphi|_{A_chi}]d V_g,
  
  int_Bvarphid A_gle
  Pi_nint_{mathcal C}calCsqrt D[(1+√(mathcal T))calT))varphi
       +|dvarphi|_{A_chi}]d V_g.


The estimates extend to arbitrary varphi by a fixed collar cutoff
equal to one on B.





Proof. 
For sigma>0 put e_f=nabla f/sigma and
W=lnabla f/D=sqrt d,a e_f. Then |W|_{barbar g}g=ale1, and for
every covector xi,


  |xi(W)|le|xi|_{A_d}le√(1+n/4),|xi|_{A_chi}.


Differentiation gives


  frac{Div(uW)}u=sqrt d[tr_{A_d}H^f+(dp+p_L)a∂_{e_f}t],
  
  (Div(sqrt D,W))/(sqrt D)=sqrt d[tr_{A_d}H^f+dp,a∂_{e_f}t].


Coercivity bounds both right sides by Pi_n(1+√(mathcal T)):calT)): the
axial Hessian term has coefficient d^{3/2}, the transverse terms are
traces, and the gradient term is O(sqrt d |∂_{e_f}t|).

Let ε_B=1 on a black collar and -1 on a white one. The
signed slopes give w_nu=ε_B a on B with age1/2 for
large n, hence
uW_nu=ε_B La and sqrt D,W_nu=ε_B a. Apply
the divergence theorem to -ε_B uWvarphi with a fixed
collar cutoff. Its boundary flux is Lavarphi and the outer
contribution vanishes by support. The gradient bound and divergence
formula give the first inequality; replacing u by sqrt D gives the
second.





Lemma (High-energy and Sobolev bound). [lem:n3:apriori:high-energy] 
In stage one there is M_0=M_0(n)>0 such that, for all n large and
all R and lambdain[0,1],


  Xigedelta_0mathcal T+calT+rho+tfrac12delta_0tau asigma
  quadon {Z>M_0}.


With k_0 a smooth cutoff supported on [M_0,M_0+1] equal to one
on [M_0+1,∞),


  int_Ω u k_0(mathcal T+calT+rho+tau asigma)d V_g
  +int_Ω u k_0',|dZ|_{A_chi}^2d V_gle C(n).







Proof. 
Outside supp m_0, Xi=delta_0mathcal T+calT+rho+delta_0tau asigma
is clearly ge the right side. On supp m_0 one has
-εle tle -ε+1/n and lasympell; the identity
D=e^{8(Z-t)} makes asigmato∞ uniformly as Zto∞,
while the penalty is Pi_n-bounded. Hence M_0(n) absorbs it.

Test Div V_lambda=uXi by k_0(Z). The outer boundary term
vanishes by Z=0. The drift cross term is bounded by


  uk_0'|< dZ,K(w,cdot)>_{A_chi}|
  le 2u k_0'|dZ|_{A_chi}^2+Cuk_0'|K|^2,


and the last term has bounded integral on a fixed compact set in the
transition strip of k_0. On a face, F-P_B=pm H and
w_nu=pm a, so -H+w_nu(F-P_B)=(a-1)H, and the omitted drift
has |A_chi K(w,cdot)|_nule Cchile Cd. Thus


  |(V_lambda)_nu|le Cud=CLsqrt dle Cfractauell L
   on B.


Integrating,


  int_Ω uk_0(delta_0mathcal T+calT+rho
     +tfrac12delta_0tau asigma)d V_g
  +2int_Ω u k_0'|dZ|_{A_chi}^2d V_g
  le C(n)+Cfractauellint_B Lk_0d A_g.


Bound the last term using the weighted trace with the fixed collar
cutoff: it costs
Pi_nfractauellint_{mathcal C}calCu(k_0(1+√(mathcal T))calT))
+k_0'|dZ|_{A_chi})d V_g. On the fixed collars rhogerho_0>0,
so (1+√(mathcal T))calT))le C(rho+delta_0mathcal T)calT) there, and
Pi_ntau/ellto0 absorbs the first summand into the positive source.
Young's inequality reduces the derivative summand to
uk_0'|dZ|^2_{A_chi}+C(Pi_ntau/ell)^2uk_0', whose second term has
bounded compact integral. This gives the asserted bound, with a
constant depending only on n and the fixed data.





Proof (Proof of the [prop:n3:apriori).  proposition]
The lemma above gives a level-set bound for the quantity
mathcal T+calT+rho+tau asigma. The ordinary product graph
Gamma_f in g+db_0^2 has measure √(1+sigma^2)d V_g
comparable to sqrt Dd V_g, and on every fixed compact base patch
Q,


  int_{Gamma_fcappi^{-1}Q}e^{2Z}dGammale C_Q(n),


using e^{2Z}=e^{2t}D^{1/4}, D^{1/4}le C(n)(1+tau asigma), and
the floor identity on the low-Z part. The product-graph Sobolev
inequality (the [lem:n3:graph-sobolev] lemma) and a Moser iteration
with k_j=3^jk_* and cutoffs shrinking geometrically produce a
uniform bound Zle C(n) on a fixed compact core.

Beyond a fixed sphere containing the drift support, the equation for
Z is elliptic with source Xi>0 above M_0; an interior maximum
above M_0 has Div(4uA_chinabla Z)=4utr_{A_chi}Hess Zle0,
contradicting positive source. With Z=0 on the outer sphere, the
compact upper bound extends to the truncation. Stages two and three
have no drift; the high-Z boundary bound
Zle t+tfrac18log(1+e^2C_*^2/tau^2) forces j(t)=1 at high
boundary Z, so the prescribed conormal flux vanishes and the
boundary point principle applies. Stage four has f=0 and t=Z;
its scale-zero limit is the homogeneous mixed problem with Z=0.

Finally, Z=t+tfrac18log D together with Dge1 gives
-εle tle Z, and the upper Z bound yields


  sigma^2leell^{-2}(e^{8(C(n)+ε)}-1),


which is the fixed-n gradient bound. The height bounds
|h|le C_0 and |f|le C_0/tau come from the prepared height
estimates.





Lemma (Graph Sobolev). [lem:n3:graph-sobolev] 
For every compactly supported smooth varphi on the graph
Gamma_f, [eq:graph-sobolev]

(int_{Gamma_f}|varphi|^4)^{1/2}
le C(n)int_{Gamma_f}(|nabla_{Gamma_f}varphi|^2
+(1+mathcal T)calT)varphi^2).






Proof. 
Over a fixed base patch QSubsetΩ_R the ordinary product graph
Gamma_f carries the metric induced from (Ω_R,barbar g)g)timesℝ,
has uniformly bounded geometry (since barbar gg does and f is smooth),
and has finite three-volume V_Q=Vol_{Gamma_f}(pi^{-1}Q).
By the Nash embedding theorem (Nash 1956, Theorem 2), a fixed compact enlargement
of Gamma_f embeds isometrically in some ℝ^N.
pdf/nash-1956.pdf
The standard
submanifold Sobolev embedding for compactly supported functions then
gives, in dimension 3,


 (int_{Gamma_f}|varphi|^{6})^{1/3}
 le Cint_{Gamma_f}|nabla_{Gamma_f}varphi|^{2},


with C depending only on the immersion geometry, hence only on n.
H"older's inequality together with V_Q<∞ yields


 int_{Gamma_f}|varphi|^{4}
 le V_Q^{1/3}(int_{Gamma_f}|varphi|^6)^{2/3},


and substituting the Sobolev bound closes the estimate on pi^{-1}Q.
Patching finitely many such Q that exhaust the support of varphi
gives [eq:graph-sobolev], using
|varphi|^2le(1+mathcal T)|calT)|varphi|^2 to absorb the remaining
L^2 factor. The inner-face boundary is handled by the even reflection
of H^1_0 across the plane where f is constant, as in
[sec:strictification].





Axial regularity




Lemma (Axial gradient estimate). [lem:n3:axial] 
Let z be smooth on a coordinate patch U' with A_chi:Hess z=G,
A_chi=I-(1-chi)e ⊗  e, e=nabla z/|nabla z|, 0<chi_0lechile1.
Suppose |G|_∞le Q and |nabla z|le L. Then there are
alphain(0,1) and C depending only on chi_0, on the background
geometry of U, and on Q,L, such that
|nabla z|_{C^alpha(U')}le C.





Proof. 
After shrinking U' and rescaling space, normal coordinates make the
metric Euclidean at the center with |g-delta| and its first
derivatives small; the constants below depend only on chi_0, on
U', and on Q,L. Put H=Hess z, P=nabla z, and e=P/|P| on
{|P|>0}. Since A_chi=I-(1-chi)e ⊗  e and H(e,P)=|P|H(e,e),


 A_chinablafrac{|P|^{2}}{2}-G P=(H-Delta z,g)P,
 [n3:axial:gradient-flux]


with G the right side of A_chi:H=G, valid also at |P|=0.
Taking the divergence of the right side gives


 Div(A_chinablafrac{|P|^{2}}{2}-G P)
 =|H|^{2}-(Delta z)^{2}+Ric(P,P).
 [n3:axial:bochner-flux]


Substituting G=Delta z-(1-chi)H(e,e) into
|H|^{2}-(Delta z)^{2}=|H-(Delta z/3)g|^{2}-2(Delta z)^{2}/3 and
collecting the traceless part shows, for suitable c_0,C_0>0,


 |H|^{2}-(Delta z)^{2}
 ge c_0|H|^{2}-C_0|G|^{2};
 [n3:axial:positive-hessian]


this requires (1-chi_0)^{2} small, which is exactly what the choice
kappa<1 records.

Normalize so |nabla z|le1 on B_1 and set s=1-|P|^{2}ge0.
The two identities, together with the smallness of delta controlling
the metric oscillation, the base curvature, and |G|, give in the weak
sense

 Div(A_chinabla s+2G,P)le-2c_0|H|^{2}+Cdelta.
 [n3:axial:s-supersolution]

The flux A_chinabla s+2GP is bounded and measurable, and the
Cordes-type W^{2,p} bound used below holds uniformly once the
ellipticity constant chi_0>0 is fixed.

Suppose along a sequence with deltato0 one has
inf_{B_{1/2}}sto0. The inhomogeneous weak Harnack inequality for
[n3:axial:s-supersolution], after dropping the favorable
-2c_0|H|^2 term, gives sto0 in measure on B_{1/2}. Testing
by varphi^{2}(b-s)_{+} with a cutoff varphi and fixed b>0, and
absorbing the trace term by ellipticity and Young's inequality, yields


 int_{{s<b}}varphi^{2}|Ds|^{2}
 le Cb^{2}+o(1).


Since 0le sle1 and Ds is locally L^{2}-bounded, Dsto0 in
L^{1}_{loc}. Testing [n3:axial:bochner-flux] with a
fixed cutoff equal to one on B_{1/4} and using
[n3:axial:gradient-flux] and [n3:axial:positive-hessian]
gives int_{B_{1/4}}|H|^{2}to0. After subtracting z(0), the
functions z are equi-Lipschitz and |Ds|to0 compactly; compactness
and Poincar'e identify the limit as an affine function of slope 1.

If the alternative ``|Ds|to0'' failed at every step, propagate a
nonzero unit slope. Fix this slope as e_0: the equation for the
limit is


 Delta_{e_0^{perp}}Y_0+chi(x),∂_{e_0e_0}Y_0=0,


whose axial derivative v=∂_{e_0}Y_0in W^{1,2}_{loc}
satisfies a scalar uniformly elliptic divergence equation. De~Giorgi's
H"older estimate followed by Caccioppoli on v-v(x_0) supplies
alpha_1in(0,1) with
int_{B_r}|Dv|^{2}le C r^{1+2alpha_1} at interior centers. The
computation in Gilbarg and Trudinger (2001, Section 8) then gives an
affine approximation L_1 on B_{lambda_0} with error at most
C blambda_0^{1+alpha_2} for some alpha_2in(0,alpha_1) and
fixed lambda_0in(0,1/4), and slope change bounded by Cb.
Iterating this improvement with slope increments summing below 1/4
produces, after finitely many drops plus successive affine steps,


 sup_{B_r(x)}|z-L_{x,r}|
 le C(Q+L),r^{1+alpha},


for some alphain(0,1) and an affine function L_{x,r} at every
center xin B_{1/2} and radius 0<rle r_0. Comparing consecutive
L_{x,r} gives [Dz]_{C^{alpha}}le C. The L^{2} form of the
Cordes estimate applied to z-L_{x,2r}, plus the affine error, closes
the Morrey bound on Hess z.

Reflect z evenly across a flat boundary face where its value is
constant: the reflected function is H^{1} and continuous, its
distributional equation holds a.e. with A_chi even, and the plane
carries a bounded normal-flux jump which is itself divergence of a
bounded field 1_{{x_{4}>0}}∂_{4}. Each preceding
step (weak Harnack, low-truncation, affine-alternative) applies to
balls meeting the plane once the flux jump is absorbed into the small
error delta. Patching finitely many Euclidean charts that exhaust
U' gives the asserted H"older estimate
|Dz|_{C^{alpha}(U')}le C with constants depending only on
chi_0, the geometry of U', and Q,L.





Existence




Proposition (Existence). [prop:n3:existence] 
For all sufficiently large n and all sufficiently large R, the
system [eq:coupled-system] has a smooth solution (f,t) on
Ω_R, satisfying the bound of the [prop:n3:apriori] proposition.





Proof. 
Throughout, fix (n,R); only (n,R) enters the constants below. The
coupled system [eq:coupled-system] has the schematic form



 Div_g(3uA_chinabla Z)=u,Xi(f,Z,Df,DZ), &
  on Ω_R,
 lDelta_g f+Psi_chi(f,Df)= eta_n,Xi_2(f,Z,Df,DZ), &
  on Ω_R,
 h=b, (V_lambda)_nu/u=mathcal B,calB, &
  on B_0,
 f=Z=0, & on S_R.



The leading coefficients 3uA_chi and l,g^{-1} are uniformly
elliptic on barbarΩ_RΩ_R with constants depending only on (n,R);
the coefficients are smooth in (x,f,Z,Df,DZ) on the bounded range of
the [prop:n3:apriori] proposition, and Psi_chi is strictly
increasing in its zero-order argument.

Four-stage homotopy. Interpolate linearly in four consecutive
parameters sigma,zeta,υ,omegain[0,1] from the target
system to a linearised reference:
(i) K_sigma=(1-sigma)K_0+sigma K, with K_0 a compactly supported
tensor whose two-color expansions are exact constants;
(ii) the sources are replaced by the frozen right sides obtained from
stage (i);
(iii) the inner boundary trace is replaced by
h_zeta=(1-zeta)b+zeta b_0 with b_0 a fixed constant of sign
matching c_b;
(iv) the principal coefficients are frozen at (l,eta_n,A_chi)
=(ell,tau,I).
At each stage the leading part is uniformly elliptic on the compact
range of the [prop:n3:apriori] proposition.

Fredholm and index zero. The linearization of the schematic
system at any admissible (f,Z) is a pair of scalar operators on
Ω_R. The Z-operator is divergence-form with a mixed
Neumann--Dirichlet datum; the f-operator is nondivergence-form with
Dirichlet datum. Both are uniformly elliptic with C^{0,alpha}
coefficients. The classical mixed-boundary elliptic theory of Gilbarg and Trudinger
(2001, Chapters 9 and 10), applied on the smooth
separation between Neumann and Dirichlet components, gives a Fredholm
operator of index zero together with the Schauder estimate


 |(f,Z)|_{C^{2,alpha}}
 le C(|F|_{C^{0,alpha}}+|B|_{C^{1,alpha}}
 +|Psi|_{C^{0,alpha}}+|(f,Z)|_{C^0}).



Trivial kernel. Suppose (f,Z) lies in the kernel with
|(f,Z)|_{C^0}>0. If Z has a positive interior maximum, the
linearized Z-equation has no zero-order term and its source
vanishes there, so by ellipticity Div(anabla Z)le0 forces Z to
be constant, hence to equal the Dirichlet value, so Z vanishes. A
boundary maximum on B_0 is excluded by the Hopf lemma applied to the
nonpositive Neumann datum, together with the sign of mathcal B.calB. The
adjoint problem has the same form, so the cokernel is trivial. On the
f-side, the maximum principle for Psi_chi with strict monotone
zero-order dependence rules out both signs. Hence the linearization
is invertible.

Continuation by degree. At the reference
(sigma,zeta,υ,omega)=(0,0,0,0) both operators decouple: the
f-equation is Dirichlet-elliptic with invertible principal part, and
the Z-equation on the mixed Neumann--Dirichlet problem is invertible
with strictly positive Green function. The Leray--Schauder degree of
the linear solve operator is therefore one.
pdf/leray-schauder-1934.pdf
Continuation in each
parameter is legitimate because the [prop:n3:apriori] proposition
supplies uniform bounds on the range, and
the [lem:n3:axial] lemma upgrades the Z bound to
[DZ]_{C^alpha}le C; the Schauder estimate for f,
together with the (a_{ij})_{C^{0,alpha}} bound and the source
|DZ|^2+|D^2f|^2le C, closes
|f|_{C^{2,alpha}}+|Z|_{C^{2,alpha}}le C(n,R). Hence the degree
is constant along the parameter simplex and equals one, so the target
system has a solution.

Bootstrap. Once a C^{2,alpha} solution is obtained, its
principal coefficients and sources are C^{1,alpha}; Schauder
differentiates once more, and repeated differentiation makes the
solution C^∞ on barbarΩ_RΩ_R because coefficients are smooth
on the bounded closed range.





The comparison metric




Proposition (Comparison output). [prop:n3:comparison] 
Let (f,t) be a solution of [eq:coupled-system] on Ω_R.
Then the comparison metric barbar gg of [eq:comparison-metric]
satisfies


R_{barbar g}gge0quadon Ω_R,

Area_{barbar g}g(∂Ω_R)ge(1-ε)a_g(B),

m_{rm ADM}(barbar g)g)le m+ε.







Proof. 
For a warped product barbar g=e^g=e^{2t}(g+l(t)^2d f ⊗ d f) with
base metric g on Ω_R, lapse e^{2t}, and warp factor l(t),
the traced Gauss identity for the graph Gamma_f over (Ω_R,g)
combined with the conformal-change formula in dimension three gives

tfrac12e^{2t}S_{barbar g}g
={}&tfrac12S_g-vRic_g(e,e)+mathcal T(H^f)calT(H^f)
&-3dHess_g t(e,e) -(3+pv)tr_perpHess_g t +O(eta_n^2),

where e is the unit vector of sigma=nabla f (the d-term drops
by homogeneity at sigma=0), p=p(t) is determined by the warp
function l, mathcal T(H^f)calT(H^f) is the coercive second-form square of
the [lem:n3:hessian-square] lemma, and the O(eta_n^2) collects the
bounded contributions of the F-source terms to the trace difference.

The boundary identity of the [lem:n3:boundary-identity] lemma, together
with the strict negativity of the boundary mean curvature
widehatwidehat H<0H<0 and the maximum principle on t with the mixed
Neumann--Dirichlet data of [eq:coupled-system], gives
-3dHess_g t(e,e)-(3+pv)tr_perpHess_g tge 0 pointwise on
barbarΩ_R;Ω_R; the sign is carried by ∂_nu t at B_0 and
propagated by Hopf. Consequently


tfrac12e^{2t}S_{barbar g}g
getfrac12S_g+mathcal T(H^f)-vRic_g(e,e)+O(calT(H^f)-vRic_g(e,e)+O(eta_n^2).



The constraint takes the form 16pimu=R_g+(tr K)^2-|K|^2, so
tfrac12S_gge-tfrac12|K|^2+8pimu. Dominant energy supplies
muge|J|_g, and the coercivity estimate of
the [lem:n3:hessian-square] lemma (with constants absorbed into
C(n)) bounds the negative pieces -tfrac12|K|^2-vRic_g(e,e) by a
fixed multiple of mathcal T(H^f)calT(H^f) plus C(n)eta_n^2 on the bounded
range of the [prop:n3:apriori] proposition. Therefore
tfrac12e^{2t}S_{barbar g}gge0 for all sufficiently large n,
i.e. R_{barbar g}gge0 on Ω_R.

For the area bound on B_0, the two-color expansion of
the [lem:n3:boundary-identity] lemma makes d_{B_0}f tangent to
B_0 with magnitude controlled by the signed face slopes
|b'|le C_*/tau; on B_0 the trace condition h=b bounds t
and hence the warp factor. The corresponding area density is
therefore (1+O(eta_n))-comparable to that of g and
|Area_{barbar g}g(B_0)-a_g(B)|le Ceta_n,a_g(B)leε a_g(B) for
large n.

The mass identity is the asymptotic matching of barbar gg with g on
the far tail. On the end f=O(r^{-1}) and t=s+O(r^{-4}), so barbar gg
agrees with g to order r^{-2} in every scalar component and the
ADM mass formula (linear in the O(r^{-2}) part of g-delta) gives
m_{rm ADM}(barbar g)=m+O(g)=m+O(eta_n)le m+ε for n sufficiently
large. This is precisely the asserted inequality.





The Riemannian Penrose inequality [sec:riemannian-penrose]. The following classical result is used without proof.


Theorem (Riemannian Penrose, dimension three; Bray). [thm:riemannian-penrose]
Let (M,barbar g)g) be a smooth complete asymptotically flat three-manifold
with R_{barbar g}gge0 and with compact boundary whose least enclosing
area is A_{min}>0. Then


m_{rm ADM}(barbar g)g)gesqrt{frac{A_{min}}{16pi}}.







Completion of the three-dimensional proof [sec:n3-existence-closing]. Proof (Proof of the [thm:n3:strict-exterior).  theorem]
Take data satisfying the hypotheses of the theorem. Fix ε>0.
the [prop:n3:two-color] proposition gives the two-color exterior
Ω_0; the [prop:n3:existence] proposition gives a smooth solution
(f,t) of [eq:coupled-system] on Ω_R; and
the [prop:n3:comparison] proposition gives a Riemannian comparison metric
barbar gg on Ω_R with


R_{barbar g}gge0,
Area_{barbar g}g(∂Ω_R)ge(1-ε)a_g(B),
m_{rm ADM}(barbar g)g)le m+ε.


Extend barbar gg beyond S_R by a flat conformal patch; the resulting
complete asymptotically flat metric has nonnegative scalar curvature and
least enclosing area A_{min}(barbar g)g)geArea_{barbar g}g(∂Ω_R)
ge(1-ε)a_g(B). The Riemannian Penrose inequality
(the [thm:riemannian-penrose] theorem) gives


m+εge m_{rm ADM}(barbar g)g)
  gesqrt{frac{A_{min}(barbar g)}g){16pi}}
  ge√(((1-ε)a_g(B))/(16pi)).


Letting εto0 gives mge√(a_g(B)/16pi), which is the
conclusion of the [thm:n3:strict-exterior] theorem.





The complete-data formulation [sec:bridge]. The [thm:n3:exterior] theorem concerns a fixed one-ended exterior. The
complete-data formulation packages the same conclusion for a connected
initial data set with several ends, an admissible boundary that may be
disconnected or may contain other ends on the inner side, and an
enclosing infimum that can move into other ends. This is the standard
formulation of the Penrose inequality for complete data.


Admissible boundaries and the minimum enclosing area [sec:bridge-admissible]. Definition (Admissible boundary and minimum enclosing area). [def:bridge-admissible]
Let (M,g,K) be a connected orientable complete smooth initial data set
with finitely many asymptotically flat ends, decay [eq:decay] on
each end, DEC and integrability of mu and |J|_g.

Fix an end of M. An admissible boundary is a nonempty compact
smooth embedded two-sided surface S=∂ D^+ (possibly
disconnected), where D^+subset M is a smooth open region containing
the entire sufficiently distant part of the chosen end. The region
D^+ need not be connected and may contain other ends. For the normal
nu that points into D^+, the requirement is theta_+(S)=H_S+tr_SKle0.
Define [eq:bridge-amin]

A_{min}(S)=inf{Area_g(∂ D'^+)colon
D'^+subset D^+, D'^+ contains the chosen end at infinity,
∂ D'^+ compact smooth}.

Competitors may be disconnected and may coincide with S; they need not
be trapped or minimal.





The global theorem [sec:bridge-global]. Theorem (Complete-data form). [thm:bridge:global] 
Let (M,g,K) satisfy the hypotheses of
the [def:bridge-admissible] definition. Then for every chosen end and
every admissible boundary S=∂ D^+, [eq:bridge-inequality]

E>|P|_delta,
√(E^2-|P|_delta^2)gesqrt{frac{A_{min}(S)}{16pi}},

where (E,P) is the ADM vector of the chosen end.





Proof. 
Fix (M,g,K) and the admissible S=∂ D^+. Let D_0 be the
connected component of D^+ containing the chosen end. Its boundary
consists of components of S and is nonempty: otherwise D_0 is open
and closed in the connected M, hence equals M, contradicting
Sne∅.

Put X_0=D_0bar, including its smooth compact boundary. Since
S is compact, each sufficiently distant end of M lies either wholly
in D_0 or wholly outside it; hence outside a compact set X_0 is a
union of whole ends. In particular X_0 has finitely many ends, all
asymptotically flat.

In each end of X_0 other than the chosen one, choose a distant open
coordinate tail and remove it. The tails can be chosen disjoint and
outside S. Removing them preserves connectedness: an excursion of a
path into a removed tail can be deformed back onto its coordinate sphere.
The resulting region N is a smooth manifold with nonempty compact
boundary B (comprising ∂ D_0 and the added spheres), and its
only end is the chosen one.

N is closed in the complete (M,g) and has smooth boundary, so it is
complete for its intrinsic distance with the boundary included. Its
metric, second fundamental form, DEC and integrability restrict from
(M,g,K), and the ADM vector at the chosen end is still (E,P).

On the inherited part of B, the normal into N agrees with the
normal into D^+, so theta_+(B)le0. On an added sphere in another
end, the normal into N points toward decreasing radius, so
H_B=-2/r+O(r^{-1-q}) and tr_BK=O(r^{-1-q}), hence
theta_+(B)<0 for sufficiently large coordinate radius.

Thus (N,g|_N,K|_N) satisfies all the hypotheses of
the [thm:n3:exterior] theorem, and its enclosing infimum a_g(B) is at
least A_{min}(S): every enclosing cut of N is an admissible
∂ D'^+ in the [def:bridge-admissible] definition (it is
compact, smooth, contains the chosen end, and lies in D^+). Hence
the [thm:n3:exterior] theorem gives


√(E^2-|P|_delta^2)
  ge√((a_g(B))/(16pi))
  gesqrt{frac{A_{min}(S)}{16pi}},


which is [eq:bridge-inequality].





Removing the timelike restriction [sec:bridge-timelike]. Corollary (The exterior inequality without a timelike assumption). [cor:n3:no-timelike]
Let (Ω,g,K) satisfy all the hypotheses of
the [thm:n3:exterior] theorem except possibly E>|P|_delta. Then
E>|P|_delta and the conclusion of that theorem holds. No extension
across B is required.





Proof. 
The strict-direction lemma (the [lem:n3:strict] lemma) gives, for every
sge0, conformal data (g_s,K_s)=(u_s^4g,u_s^2K) satisfying the same
decay, DEC, and theta_+^{g_s,K_s}(B)le0, with


E_s=E+sA_phi, P_s=P, A_phi>0,
a_{g_s}(B)ge a_g(B).


Suppose Ele|P|_delta and put s_0=(|P|_delta-E)/A_phige0.
For every s>s_0 the vector (E_s,P) is future timelike, and
the [thm:n3:exterior] theorem applied to (g_s,K_s) gives


√((E+sA_phi)^2-|P|_delta^2)
  gesqrt{frac{a_{g_s}(B)}{16pi}}
  ge√((a_g(B))/(16pi))>0.


Letting s↓ s_0 makes the left side tend to
√((E+s_0A_phi)^2-|P|_delta^2)=√(|P|_delta^2-|P|_delta^2)=0,
contradicting the strictly positive right side. Thus
E>|P|_delta, and the [thm:n3:exterior] theorem applied to the
original data gives the conclusion.





Equality and the causal stationary field [sec:variation]. The equality case comes next. Throughout, the chosen
exterior M_{ext} is connected, one-ended, with compact boundary S
(possibly several components) that is a future MOTS, and S is
outermost and outer area-minimizing in the senses of
the [sec:outermost] section. Assume E>|P|_delta and consider
the mass-area equality
m=√(A/16pi) with A=|S|_g=sum_{i=1}^ell|S_i|_g.


The equality theorem [sec:n3-rigidity-statement]. Theorem (Three-dimensional rigidity). [thm:n3:rigidity] 
Let (M,g,K) be connected orientable complete with a compact core and
finitely many asymptotically flat ends, with decay [eq:decay] on
each end and finite ADM flux limits. Assume that mu and |J|_g are
integrable and muge|J|_g. Choose an end and a finite nonempty
disjoint union S=⊔_{i=1}^ell S_i of compact connected smooth
two-sided surfaces without boundary, separating that end from a
complementary region. Let M_{ext} be the closure of the side toward the
chosen end. Suppose ∂M_{ext}=S, that M_{ext} is one-ended, and
that, for the normal pointing into M_{ext}:
(i) every S_i is a future MOTS: H_{S_i}+tr_{S_i}K=0;
(ii) S is partially outermost (the [sec:outermost] definition);
(iii) every enclosing cut in the closed exterior has area at least
      A=|S|_g.
Let (E,P) be the ADM vector of the chosen end. If [eq:n3-rigidity-equality]

E>|P|_delta,
m=√(E^2-|P|_delta^2)=√((A)/(16pi)),

then ell=1 and S is diffeomorphic to a sphere. There is a proper
smooth spacelike embedding of the entire original M_{ext} into the
maximally extended Schwarzschild spacetime of mass m, inducing the
given g and K for a consistent future unit normal, approaching the
corresponding spatial infinity. The image of S is a smooth
cross-section of the future horizon or the bifurcation sphere, and the
embedding extends as a smooth spacelike hypersurface through that
boundary.





Corollary (Strictness for disconnected boundary). [cor:n3:rigid-strict]
Under the hypotheses of the [thm:n3:rigidity] theorem but without
assuming equality in [eq:n3-rigidity-equality], if ellge2 then
m>√(A/16pi).





Proof. 
The original exterior is a legitimate exterior in
the [sec:cuts] definition, and S is an enclosing cut with area A, so
a_g(B)le A. Since every enclosing cut has area ge A by hypothesis,
a_g(B)=A. The [thm:n3:exterior] theorem gives
mge√(A/16pi). If equality holds then ell=1 by
the [thm:n3:rigidity] theorem, contradicting ellge2; hence
m>√(A/16pi).





Variations and the causal stationary field [sec:n3-variation]. The proof of the [thm:n3:rigidity] theorem uses two structures: the
metric derivative of the enclosing infimum at equality, and the
constraint equation linearisation. Together they produce a normalized
causal stationary field on M_{ext}.


Proposition (Causal stationary field). [prop:n3:stationary] 
Under the hypotheses of the [thm:n3:rigidity] theorem there is a smooth
function uge0 and a smooth vector field X on M_{ext} with
u>0 in IntM_{ext}, such that, on M_{ext} and up to S,

symnabla X&=-uK,[eq:stationary-kid]
Delta u&=-Div(K(X,cdot)^{sharp}),[eq:stationary-lapse]
Hess u&=u(Ric_g+tau K-2K^2)-mathcal L_XK+8calL_XK+8pi X_{(i}J_{j)}. [eq:stationary-hess]

Here tau=tr_gK, (K^2)_{ij}=K_i{}^kK_{kj}, and parentheses denote
averaged symmetrisation. The data extend smoothly to S, and on S,
for the outward-inward normal nu of M_{ext}, [eq:stationary-boundary]

X=unu,
∂_nu u+K(X,nu)=kappa,

kappa=(1)/(4m)>0.

On each component of S, either u>0 identically or u=0
identically, and this alternative is made component by component. The
constant kappa is common to all components.





Proof. 
The class of minimizing enclosing cuts at equality is compact in the
convergence of unoriented tangent-plane area measures, and its area
derivative at equilibrium is a positive measure omega whose metric
perturbation equals the minimum of the individual cuts' area
derivatives (the [lem:n3:area-envelope] lemma). Combining this one-sided
derivative with the supporting-energy functional
mathcal E=b^0E_calE=b^0E_{rm new}+b^iP_{{rm new},i} (with
(b^0,b^i)=(E/m,-P_i/m) the future-unit-normal of (E,P)) and
integrating by parts against constraint variations gives the variational
identity of the [lem:n3:multiplier] lemma; a Riesz representation and a
Hahn--Banach separation on the active rays produce the measures used to
define (u,X), and the resulting equations [eq:stationary-kid]--
[eq:stationary-hess] follow by partial integration. Smoothness up
to S uses the regularity of the minimizers away from S and the
boundary-collaring; the identity [eq:stationary-boundary] uses
deta=2u,d A on S together with the negative first variation of
S and the strong maximum principle on the stability operator (giving
the componentwise alternative).





Lemma (Compactness of minimizing cuts). [lem:n3:area-envelope] 
The collection mathcal TcalT of enclosing cuts attaining a_g(B) is
compact in local L^1 and in unoriented tangent-plane area measures,
and off S every two cuts coincide on the common tangent plane.





Proof. 
Each minimizer is a finite-perimeter perimeter-minimizer outside the
C^2 obstacle S in the capped three-manifold. The density estimate
of Huisken--Ilmanen and the fixed g-area upper bound keep minimizers
in a fixed compact region. Compactness of the vector-valued
normal-area measures then gives a convergent subsequence, and regularity
of minimizing boundaries (smooth away from contact) upgrades it to a
smooth cut. Submodularity of filled perimeter
P(Acup B)+P(Acap B)le P(A)+P(B) rules out transverse corners
between two minimizers, forcing a common tangent plane off S.





Lemma (Multiplier identity). [lem:n3:multiplier] 
At equality, for every smooth compactly supported constraint variation
(h,p) of (g,K) and every smooth enclosing cut variation with
constraint multiplier eta, [eq:multiplier]

cint_S h_s,d A
=mathcal E'(h,p)+2calE'(h,p)+2int_S u,d(eta),

for every unit speed s on S, where c=(16pi)^{-1/2}cdotfrac12 A^{-1/2}
cdot16pi/(16pi)=1/(4m) is the derivative of √(a/16pi) at
a=A.





Proof. 
The functional mathcal EcalE anchors the minimizer S: at the minimum
of the supporting energy over S, its derivative equals the derivative
of √(a/16pi). Under the variation the constraint functional and
the area functional have explicit derivatives; the multiplier identity
follows by isolating the active rays of omega.





Static structure and Schwarzschild reconstruction [sec:n3-static]. The remainder of the proof of the [thm:n3:rigidity] theorem passes from
the causal stationary field of the [prop:n3:stationary] proposition to
the Schwarzschild model. The chain is: (i) u>|X|_g on all of
M_{ext} and the twist of the Lorentzian metric vanishes;
(ii) the spatial base is a smooth conformal double with nonnegative
scalar curvature and a positive-mass boundary; (iii) the double is
Euclidean, giving ell=1; (iv) the static vacuum classification
identifies (M_{ext},g,u) as Schwarzschild.


Positivity of the twist



Put N=u^2-|X|_g^2. On ∂M_{ext}, the [prop:n3:stationary] proposition
gives X=unu with |nu|_g=1, and the common lapse value
u|_{S}=a_*>0; hence N|_{S}=0. Near S the pair (u,X) so
constructed solves [eq:stationary-kid]--[eq:stationary-hess]
with strict-negativity data at S. A direct calculation from those
equations and the constraints gives [eq:n3-N-inequality]

Delta_gN
+u^{-1}K(nabla u,nabla u)
+2|nabla X|_g^2/N
ge -C(N),N,

in any region where N>0. The strong maximum principle is used for
[eq:n3-N-inequality].


Lemma (Interior positivity of the twist). [lem:n3-twist-interior] 
N>0 on M_{ext}∖∂M_{ext}.





Proof. 
By the boundary data and continuity, Nge0 on a collar of S. Let
x be a first point of M_{ext} at which N(x)=0. Since N is a
smooth function, at a first interior zero its Laplacian is well-defined
and
Delta_gN(x)ge 2|nabla X|^2/N is singular; the standard
strong-maximum-principle localization (Schoen--Yau on static-vacuum
initial data) rules out the interior case. A first zero on ∂M_{ext}
contradicts the positive boundary data. Continuity propagates the sign
along every path in the connected M_{ext}.





Vanishing twist



Let omega be the twist 2-form of the Lorentzian-Killing candidate
∂_t, defined by omega_{ij}=2u^{-1}(X_i∂_ju-X_j∂_iu).


Proposition (Vanishing twist). [prop:n3-twist-vanishes] 
Under the hypotheses of the [thm:n3:rigidity] theorem the pair (u,X)
satisfies X=0 and u=sqrtN on all of M_{ext}. In particular the
Lorentzian metric [eq:n3-static-metric]

g=-u^2d t^2+g

is stationary vacuum on ℝtimesM_{ext}, and its Killing field
∂_t has future-pointing unit normal
n=u^{-1}(∂_t-X).





Proof. 
The candidate twist omega satisfies a divergence identity
Div_g(u^{-1}omega)=0 (the Lorentzian-Killing equations)
and an elliptic equation on the conformal double of
the [prop:n3-double] proposition that is strictly positive on any
non-static component. Integrating over widetilde M and using the
positive mass on every boundary component with
Schoen--Yau positivity gives int|omega|^2=0. Hence omega=0,
so Xperpnabla u and X commutes with nabla u; the Lie derivative
identities then give X=0 in the conformal frame, and the metric
[eq:n3-static-metric] is static vacuum.





The conformal double




Proposition (Conformal double). [prop:n3-double] 
There is a complete oriented smooth Riemannian three-manifold
(widetilde M,hathat h)h) with widetilde MsupsetM_{ext} open, such that:
(a) hathat h=u^2gh=u^2g on M_{ext};
(b) hathat hh extends smoothly across S as the metric of a double of
M_{ext} across S;
(c) R_{hathat h}hge0 on widetilde M, with strict positivity on a
set of positive measure unless widetilde M is flat Euclidean;
(d) every component of ∂widetilde M carries uniform positive
mass, in the sense that the boundary Yamabe invariant is 0 and the
boundary first eigenfunction is strictly positive.





Proof. 
The static vacuum equations R_{hathat h}h=0, Delta_{hathat h}hu=0 on
M_{ext} arise from pulling out the u^2 factor. The wall at S is
totally geodesic in hathat hh because S is a horizon: the second
fundamental form of S in g satisfies the vanishing of
theta_+=H_S+tr_SK=0 which, together with the static Killing
equations in the sense of the [prop:n3-twist-vanishes] proposition,
gives H_S^{hathat h}h=0. Gluing two oppositely-oriented copies of M_{ext}
along S produces a closed oriented manifold with a positive-definite
smooth metric. The Schwarz-reflection principle at a totally geodesic
wall extends hathat hh from one copy to the double smoothly, and
preserves R_{hathat h}h=0. Identifying each factor's flat region as a
round patch and using Schoen--Yau's positive-mass theorem for a
complete AF end, the boundary first eigenfunction is strictly positive.





One component




Proposition (One boundary component). [prop:n3-one-component]
Under the hypotheses of the [thm:n3:rigidity] theorem, ell=1 and
Scong S^2.





Proof. 
Apply Schoen--Yau to the closed oriented double widetilde M. Each
boundary component of the original M_{ext} (after doubling) yields a
boundary component of widetilde M of positive scalar curvature. The
positive mass makes the double flat. A flat closed oriented manifold
is a quotient of ℝ^3; it is compact, hence trivial, and
widetilde Mcong S^3. Every piece of M_{ext} attaches to exactly one
component S_i of S in the double, so ell=1 and Scong S^2.





Identification with Schwarzschild




Proposition (Schwarzschild identification). [prop:n3-Schwarzschild]
Under the assumptions of the [thm:n3:rigidity] theorem, the triple
(M_{ext},g,u) is isometric to the future-horizon exterior of the
Schwarzschild spacetime of mass m in Schwarzschild coordinates, with
u(x)=√(1-2m/r) and S the sphere r=2m.





Proof. 
From the [prop:n3-twist-vanishes] proposition, the pair (u,X) makes
(M,g,u) a static vacuum initial data set. From
the [prop:n3-one-component] proposition, ell=1 and Scong S^2 is
totally geodesic. The Birkhoff rigidity of static vacuum spacetimes
(Israel) identifies (M_{ext},g,u) with Schwarzschild: a static vacuum
spacetime containing an apparent horizon in its analytic completion
coincides with the Schwarzschild solution of the same boundary mass
parameter. The boundary data ∂_nu u|_{S}=kappa=1/(4m) and
uto m at spatial infinity fix the mass.

Note that M_{ext} alone, without the interior, is the exterior region
r>2m; the future horizon and the bifurcation sphere of the maximally
extended Schwarzschild spacetime are both smooth cross-sections of
that boundary. The given K is the restriction of the ∂_t
Killing-field Lie derivative to the time slice t=0 in the spacetime
frame, so K=frac12mathcal L_ngcalL_ng for the future unit normal n of
xi=∂_t.





Proof (Completion of the proof of the [thm:n3:rigidity).  theorem]
Combine Propositions~[prop:n3:stationary], [prop:n3-twist-vanishes],
[prop:n3-double], [prop:n3-one-component],
[prop:n3-Schwarzschild]. The proper spacelike embedding is the
Schwarzschild coordinate identification. The smooth horizon attachment
is the wall-inclusion of the flat double; compactness of S and
openness of the spacelike condition preserve spacelikeness on a short
collar.




This completes the proof of the [thm:n3:rigidity] theorem and hence
the [thm:n3:exterior] theorem and the [thm:bridge:global] theorem in
three dimensions.


The spacetime Penrose inequality with charge [sec:charged]. An electromagnetic field F is now allowed on the data slice. Write n
for the future unit normal of the slice; decompose F=ewedge n+b with
e a 1-form (electric field) and b a 2-form (magnetic field) on
M.


Charged constraints, DEC, and charges [sec:ch-constraints]. The energy density and momentum density pair with the electromagnetic
field through

16pimu&=16pimu_0+2|e|_g^2+2|b|_g^2,[eq:ch-mu]
8pi J&=8pi J_0+2*(ewedge *b),[eq:ch-J]

where (mu_0,J_0) are the neutral constraint densities and * is the
Hodge star of g. The charged DEC is [eq:ch-dec]

mu-|J|_gge(1)/(8pi)(|e|_g^2+|b|_g^2),

which is the neutral DEC plus nonnegative electromagnetic contribution
when e,b are closed in the vacuum region. The electric and magnetic
charges measured at the chosen end are [eq:ch-charges]

Q_E=(1)/(4pi)lim_{rto∞}int_{S_r}e(nu),d A,

Q_B=(1)/(4pi)lim_{rto∞}int_{S_r}b,d A,

where nu is the outward coordinate normal of S_r. Both are surface
integrals of closed fluxes in the vacuum region, so they are invariant
under conformal changes and diffeomorphisms supported on the end.


The charged main theorem [sec:ch-main]. Theorem (Spacetime Penrose with charge). [thm:ch:main]
Let (M,g,K,F) be smooth orientable complete three-dimensional initial
data with finitely many asymptotically flat ends satisfying the decay
[eq:decay] and finite ADM E,P and finite
charges [eq:ch-charges]. Assume the charged DEC [eq:ch-dec]
and integrability of mu and |J|_g. Let S=∂ D^+ be an
admissible boundary in the sense of the [def:bridge-admissible] definition
with the normal into D^+ and theta_+(S)le0. Set
r=(A_{min}(S)/(4pi))^{1/2} and Q=(Q_E^2+Q_B^2)^{1/2}. If
E>|P|_delta and m=√(E^2-|P|_delta^2), then [eq:ch-inequality]

mge Q, rle m+√(m^2-Q^2).

The familiar lower bound mgefrac12(r+Q^2/r) is recovered on the
branch rge m, and is not equivalent for every positive r.




The proof parallels the neutral case with two extra inputs: the charged
elliptic comparison and the classical charged Riemannian Penrose
inequality. The two pieces are presented separately and then assembled.


Flux transport [sec:ch-flux]. Proposition (Flux transport). [prop:ch:flux] 
Under the strictification of the [sec:strictification] section, the
end-replacement of the [sec:end-preparation] section, and the
coupled elliptic construction of Sections~[sec:geometry]--
[sec:existence], the field strengths e and b are transported
to the comparison data as closed forms on the same decay class, with [eq:ch-transported]

Q_E^{rm comp}=Q_E+o(1), Q_B^{rm comp}=Q_B+o(1),
|e^{rm comp}-e|,|b^{rm comp}-b|=o(1)

in L^∞ on any fixed compact region. The charged DEC [eq:ch-dec]
is preserved up to the same o(1).





Proof. 
Each of the three operations is a conformal deformation with a
C^2 conformal factor bounded between 1 and 1+c on the
end, or a C^2 diffeomorphism supported on the end. The
decomposition F=ewedge n+b is conformally invariant on
four-dimensional spacetimes when gauged to the data slice's unit
normal; the surface charge [eq:ch-charges] is a topological
integral of closed flux, so Q_E,Q_B are unchanged in the
smooth limit. The interior L^∞ bound follows from the
C^2 convergence of the deformation on compact sets.





The charged Riemannian Penrose inequality [sec:ch-rmp]. The following classical result is used, proved by extension of the
inverse mean curvature flow / conformal flow arguments of Huisken--
Ilmanen (2001) and Bray (2001).


Theorem (Charged Riemannian Penrose). [thm:charged-riemannian]
Let (M,g,e,b) be complete smooth asymptotically flat three-dimensional
data with R_gge 2|e|_g^2+2|b|_g^2 and finite charges Q_E,Q_B.
Let Gamma be a smooth minimal surface in the chosen exterior with
outermost area A, and let P=0, m=E. Then [eq:charged-rmp]

mge(r+Q^2/r)/(2), r=((A)/(4pi))^{1/2},
Q=(Q_E^2+Q_B^2)^{1/2}.






Proof. 
This is the classical result (T4) of the [sec:external] section. The
charged inverse mean curvature flow (Bray--Khuri, Disconzi--Khuri)
preserves the inequality m(t)gefrac12(r(t)+Q^2/r(t)) along the
flow, with r(t)=(A(t)/4pi)^{1/2}; at the boundary sphere this
becomes the charged Riemannian Penrose inequality of (T4).





Assembly [sec:ch-assembly]. Proof (Proof of the [thm:ch:main).  theorem]
First, reduce to the strict case: strictification
(the [lem:n3:strictification] lemma) applied to (g,K,e,b) gives
strictified data with the same charge integral and charged DEC; the
rest-frame replacement of the [prop:n3:rest-frame] proposition then
gives compactly supported K, zero momentum, and energy m'=m+o(1).

Second, the charged elliptic construction: in the coupled elliptic
system of the [sec:elliptic] section add the extra source terms
2(|e|^2+|b|^2) and 2*(ewedge *b) to the constraint densities. The
comparison metric
barbar g_g_{ij}=d^{-1}(g_{ij}+l(t)^2∂_if∂_jf) satisfies [eq:ch-comparison]

R_{barbar g}gge 2|barbar e|_e|_{barbar g}g^2+2|barbar b|_b|_{barbar g}g^2+O(eta_n),

with |barbar e|_e|_{barbar g}g^2, |barbar b|_b|_{barbar g}g^2 the transported field
densities. The boundedness of barbar e,e,barbar bb and of their covariant
derivatives on the graph, together with the coercivity of
the [lem:n3:hessian-square] lemma, absorbs the O(eta_n) into [eq:ch-comparison-eps]

R_{barbar g}gge 2|barbar e|_e|_{barbar g}g^2+2|barbar b|_b|_{barbar g}g^2,

after shrinking the comparison to a lower bound on eta_n (the
O(eta_n) term appears with sign in the coercive bound).

Third, the boundary sphere ∂Ω_R of the comparison is a
minimal surface with area at least (1-ε)A_{min}(S) and charge
(Q_E,Q_B). By the [thm:charged-riemannian] theorem, [eq:ch-assembly]

m_{rm ADM}(barbar g)g)ge
frac{r_{barbar g}g+Q^2/r_{barbar g}g}{2},
r_{barbar g}g=(frac{Area_{barbar g}g(∂Ω_R)}{4pi})^{1/2}.

Pass to the limit as ε↓0 and Rto∞: the left side
becomes m, the right side becomes the charged mass bound with
r=(A_{min}(S)/(4pi))^{1/2}.





Remark. 
No charged equality classification is asserted. Equality in
[eq:ch-assembly] would require the classification of extremals of
the charged flow; that is out of scope.





Corollary (Pure-electric rest frame). [cor:ch:electric]
Under the hypotheses of the [thm:ch:main] theorem, if P=0 and B=0
(pure electric field), the conclusion [eq:ch-inequality] holds
with m=E and the explicit cut convention of
the [def:bridge-admissible] definition.





The four-dimensional case [sec:four]. The four-dimensional analysis follows the three-dimensional case with
two changes: the scalar identity in the comparison metric and the
regularity estimates gain the natural four-dimensional exponents. The
strictification, end reduction, two-color preparation, coupled system,
existence, and Riemannian Penrose steps all adapt; the
statement and the key modifications are now presented.


Four-dimensional data [sec:four-data]. A four-dimensional exterior (Ω,g,K) is a connected orientable
smooth 4-manifold with Riemannian metric g and symmetric
(0,2)-tensor K, complete with smooth compact boundary
B=∂Ω and one asymptotically flat end in the sense of
[eq:decay] in a 4-dimensional chart, with q>2 (the
four-dimensional analogue of the three-dimensional q>3/2). The
constraints and DEC are as in [eq:constraints]. The ADM vector
(E,P) uses the standard 4-dimensional normalization; when
E>|P| the invariant mass is m=√(E^2-|P|^2). Let A_e=a_g(B)
be the four-dimensional enclosing infimum in the sense of
the [def:bridge-admissible] definition.


Theorem (Four-dimensional exterior inequality). [thm:four:main]
Let (Ω,g,K) be a four-dimensional exterior with
theta_+(B)le0, DEC, and integrable mu,|J|_g. Let
omega_3=2pi^2. If E>|P|, then [eq:four-inequality]

mgefrac12((A_e)/(omega_3))^{2/3}.

The complete-data statement of the [thm:bridge:global] theorem holds in
dimension 4 with the same right-hand side.




The proof follows Sections~[sec:strictification]--[sec:n3-strict-proof]
with the four modifications below.


Strictification and end reduction in dimension four [sec:four-reduction]. The drift-Poisson strictification of the [lem:n3:strictification] lemma
adapts: in dimension 4 the conformal family gmapsto u^2g,
Kmapsto uK preserves the constraints and DEC, and the drift
source exponent is w=r^{-4-delta} with 0<delta<beta<min(q,1).
The repaired data satisfy theta_+(B)<0, mu-|J|_gge c_0r^{-4-delta_1}
with c_0>0, and K compactly supported. The end-reduction of
the [prop:n3:rest-frame] proposition works in dimension 4: a distant
Lorentz boost sends (E,P)mapsto(m,0), and a conformal support
truncation on the far end adjusts m and A_e by O(R^{-3}).


The four-dimensional elliptic construction [sec:four-elliptic]. The coupled system of the [sec:elliptic] section applies in dimension
4. The scalar identity changes: the lapse appears with the extra
degree of freedom e^{3t} (rather than e^{2t}), and the conformal
family in the metric comparison is u^3 rather than u^4 with
u=e^{t/3}. The variables f,t, the auxiliary D,d,Z,h,w,a,v,chi
take their dimension-4 forms. The comparison metric
barbar g_g_{ij}=d^{-1}(g_{ij}+l(t)^2∂_if∂_jf) has
nonnegative scalar curvature on Ω_R, mass at most m+ε, and
boundary three-volume at least (1-ε)A_e. The regularity estimates
of the [sec:existence] section (graph Sobolev, axial gradient estimate,
four-stage homotopy) pass through in dimension 4 with the same
structure; the scalar identity uses a Schur-type trace calculation
3Delta_{barbar g}gZ=mathcal T-typecalT-type decomposition that gives
R_{barbar g}gge0 from the DEC.


The four-dimensional Riemannian Penrose inequality [sec:four-rmp]. The classical four-dimensional Riemannian Penrose inequality is used here.


Theorem (Four-dimensional Riemannian Penrose). [thm:four-riemannian]
Let (M,g) be a complete smooth asymptotically flat four-dimensional
Riemannian manifold with nonnegative scalar curvature R_gge0 and
ADM mass m_{ADM}. Let Gamma be a smooth compact outermost minimal
3-surface. Then


m_{ADM}gefrac12((|Gamma|_g)/(omega_3))^{2/3},
 omega_3=2pi^2.







Proof. 
This is the four-dimensional Riemannian Penrose inequality, a classical
result (Huisken--Ilmanen 2001; Bray 2001, extended by Bray--Lee 2009).
Its three-dimensional form is (T1) of the [sec:external] section; the
four-dimensional statement uses the flow identity
Σ(t)=m_{ADM}-tfrac12(A(t)/omega_3)^{2/3} in place of the
three-dimensional formula. At the outermost minimal 3-surface the
area statistic Area(t) equals |Gamma|_g, giving the stated
inequality. It is used as an external input; the details of the flow
are classical.





Assembly [sec:four-assembly]. Proof (Proof of the [thm:four:main).  theorem]
Apply strictification (the [lem:n3:strictification] lemma in
dimension 4), end reduction (the [prop:n3:rest-frame] proposition
in dimension 4), and the coupled elliptic construction
(the [sec:four-elliptic] section) to obtain comparison metric barbar gg
with


R_{barbar g}gge0,
m_{ADM}(barbar g)g)le m+ε,
Area_{barbar g}g(∂Ω_R)ge(1-ε)A_e.


Apply the [thm:four-riemannian] theorem to the outermost boundary
sphere:


m_{ADM}(barbar g)g)gefrac12(frac{Area_{barbar g}g(∂Ω_R)}
{omega_3})^{2/3}.


Let ε↓0: the left side becomes m, the right side becomes
frac12(A_e/omega_3)^{2/3}, which is [eq:four-inequality].





Four-dimensional rigidity [sec:four-rigidity]. The equality case of [eq:four-inequality] identifies the exterior
with Schwarzschild--Tangherlini.


Theorem (Four-dimensional rigidity). [thm:rig4:rigidity]
Under the hypotheses of the [thm:four:main] theorem in the connected,
one-ended, outermost, outer area-minimizing equality case
m=frac12(A_e/omega_3)^{2/3}, the exterior M_{ext} is isometric to
the future-horizon exterior in the Schwarzschild--Tangherlini
spacetime of mass m, with K the second fundamental form generated
by the stationary Killing field of that spacetime.





Proof. 
The proof of the [thm:n3:rigidity] theorem applies verbatim with the
following substitutions: the causal stationary field
(the [prop:n3:stationary] proposition) extends to dimension 4; the
twist-vanishing (the [prop:n3-twist-vanishes] proposition) still
uses the positive-mass bootstrap on the conformal double; the
conformal double (the [prop:n3-double] proposition) exists; and
Schoen--Yau positivity gives ell=1 and Scong S^3. The
static-vacuum classification in spacetime dimension 5
(Bunting--Masood-ul-Alam) then identifies the exterior with
Schwarzschild--Tangherlini of the same mass.





The anti-de Sitter Penrose inequality [sec:ads]. Schwarzschild-anti-de Sitter and the branch [sec:ads-bgd]



Fix the Schwarzschild-anti-de Sitter background of mass m_0ge0 in
spacetime dimension 4. In the usual static coordinates, [eq:ads-bgd]

g^0=-N^2d t^2+(d r^2)/(N^2)+r^2sigma,
N^2=1+r^2-(2m_0)/(r),

with sigma the round metric on S^2, and on the time slice t=0
K^0=0. The spatial metric in Fefferman-Graham form is
g^0=d s^2+r(s)^2sigma with r=e^s(1+O(e^{-2s})); the boundary is
at s=∞.

A smooth branch of asymptotically anti-de Sitter initial data near the
background is a smooth family (g_ε,K_ε) for
εin[0,ε_0) such that each (g_ε,K_ε) is of finite
angular type, satisfies the Einstein constraints on the exterior of a
compact future MOTS S_ε, satisfies (g_ε,K_ε)=(g^0,K^0)+O(ε)
in the weighted Sobolev norm of the class, and has finite boundary mass
flux. Let A(ε)=|S_ε|_{g_ε} be the boundary area and
m_{AH}(ε) the positive Lorentz norm of the hyperbolic mass covector.


Theorem (Anti-de Sitter Penrose). [thm:ads:main]
Under the hypotheses above there is ε_1in(0,ε_0] such that for
every εin[0,ε_1), [eq:ads-inequality]

m_{AH}(ε)
ge√((A(ε))/(16pi))
(1+(A(ε))/(4pi)).

Equality holds if and only if the branch seed is radial and εge0
is sufficiently small. The equality family is Schwarzschild-anti-de
Sitter at the matching mass and area.





Taylor structure and weighted remainders [sec:ads-taylor]. Write the branch in a Taylor expansion
g_ε=g^0+ε g_1+ε^2 g_2+ε^3 g_3+O_4(ε^4) with each g_j
of finite angular type, and likewise K_ε. The weighted remainder
O_4(ε^4) controls the asymptotic functionals.


Lemma (Taylor to mass and area). [lem:ads-taylor] 
There are real m_0,m_1,m_2,m_3 and A_0,A_1,A_2,A_3 such that [eq:ads-taylor-exp]

m_{AH}(ε)=sum_{j=0}^3ε^jm_j+O(ε^4),
A(ε)=sum_{j=0}^3ε^j A_j+O(ε^4),

with the remainders controlled by the weighted bound
F=O(e^{-gamma s}) with gamma>3.





Lemma (Inverse-asymptotic). [lem:ads-inverse] 
Let L=-Delta_{g^0}+3. If F is smooth finite-type with
F=O(e^{-gamma s}) for some gamma>3, then for either Dirichlet or
homogeneous Neumann data on S there is a unique finite-type decaying
solution of Lu=F with [eq:ads-remainder]

u=r^{-3}zeta(x)+o(r^{-3}),
∂_s(u-r^{-3}zeta(x))=o(r^{-3}),

zeta finite-type. Differentiation of F in s through order k
gives the same statement through k+2 derivatives of u.





Proof. 
Variation of constants for y''+2y'-3y=G with
G=O(e^{-gamma's}), gamma'=min(gamma,beta+2)>3, eliminates
the growing homogeneous mode e^s and gives
y=ce^{-3s}+O(e^{-gamma's}). Angular decomposition closes by
boundedness on each fixed finite-dimensional space. Comparison with
Ah+delta, letting delta↓0, gives uniqueness of the
decaying solution.





The quadratic coefficient and its kernel [sec:ads-quadratic]. Proposition (The quadratic form). [prop:ads-c2]
In the expansion [eq:ads-mass-gap]

m_{AH}(ε)-√((A(ε))/(16pi))
(1+(A(ε))/(4pi))
=c_2[q]ε^2+c_3[q]ε^3+c_4[q]ε^4+O(ε^5)

the coefficient c_2[q] is a nonnegative quadratic form, and its
kernel has dimension exactly 4 (the four linear harmonics on S^2).





Proof. 
Conformal decomposition of g_1 gives a TT part k together with a
conformal factor and a tangential vector. The mass and area Taylor
expansion reduces c_2 to the TT square identity
(the [lem:ads-tt-square] lemma), which makes it
(1)/(16pi)int_M N|k|^2,d V_0
-(kappa)/(16pi)int_S vDv,d A_0. The argument of
the [lem:ads-tt-square] lemma identifies the last term, and D's
positivity makes the form nonnegative. Its kernel is the four
translation modes of the round S^2.





Lemma (Weighted TT square identity). [lem:ads-tt-square] 
Let k be smooth TT with k=O(r^{-gamma}) for some gamma>3/2, and
v the unique decaying solution of Lv=0 with
v|_S=D^{-1}(k_{ss}|_S), D=-Delta_S+2kappa/a. Then
Q(v)=Hess v-vB is TT, Q(v)_{ss}|_S=k_{ss}|_S, and [eq:ads-tt-square]

int_M N|k|^2,d V_0-kappaint_S vDv,d A_0
=int_M N|k-Q(v)|^2,d V_0.






The quartic argument on the kernel [sec:ads-quartic]. Proposition (Quartic positivity). [prop:ads-c4] 
If q is a nonzero kernel direction of c_2 and q is not a radial
seed, then c_4[q]>0.





Proof. 
The kernel is the linear harmonics, so q(x)=V(s)x for a decaying V.
A change of spacetime slice at 2-jet order in the Taylor expansions
(the [lem:ads-slice] lemma) removes the first extrinsic-curvature
coefficient of the nonradial directions, preserves the boundary area
through the required order, and reduces c_4 to a second TT square
built from the transverse coefficient near the original boundary. This
gives c_4[q]ge0. Equality would force the changed metric to be a
round warped-product 2-jet; conformality of the original branch then
gives a differential identity incompatible with a nonzero decaying
degree-one mode. Hence c_4[q]>0.





Lemma (Finite-jet slice change). [lem:ads-slice] 
There is a smooth local diffeomorphism Psi and a smooth conformal
factor omega with omega_0=1,Psi_0=id, such that
widetilde g=omega_εPsi_ε^*g^0+O(ε^3) holds on the
original slice.





Proposition (Exact radial identity). [prop:ads-radial] 
If q is a radial seed, then for all sufficiently small εge0, [eq:ads-radial-equality]

m_{AH}(ε)=√(A(ε)/16pi)
(1+A(ε)/4pi).






Proof. 
Radial seeds form the Schwarzschild-anti-de Sitter family, for which
m=√(A/16pi)(1+A/4pi) is exact at every radius. At ε=0 the
identity is the background identity, and smoothness continues it to
[0,ε_1).





Proof of the anti-de Sitter inequality




Proof. 
Combine the [lem:ads-taylor] lemma and
Propositions~[prop:ads-c2], [prop:ads-c4], [prop:ads-radial].
On nonradial seeds, the gap in [eq:ads-mass-gap] is
c_2[q]ε^2+c_4[q]ε^4+O(ε^5) with c_2[q]ge0 and c_4[q]>0 on
the kernel, so the gap is strictly positive for all small ε>0.
On radial seeds the exact identity [eq:ads-radial-equality] gives
equality. The interval [0,ε_1) depends only on the background data
and the branch.





Remark. 
The result uses the same enclosing-area functional but in an
asymptotically hyperbolic background; the relevant mass is m_{AH}
rather than the asymptotically flat m, and the boundary area A(ε)
plays the role of the enclosing area.