Spacetime Penrose inequalities: enclosing area, charge, rotation, and rigidity

Source and verification

This article explains family 260 of the October 6, 2026 OpenAI collection. Research conclusions are attributed to its manuscripts. The review covers statements, introductions, and selected proof mechanisms; it is not an independent verification of the complete proofs.

🏷️ Initial data, trapped boundaries, and mass

A spacetime is a manifold with a Lorentzian metric: one direction measures time, while spacelike tangent planes have positive definite metrics. A spacelike slice carries a Riemannian metric $g$ and a symmetric tensor $K$, its second fundamental form. The tensor $K$ records how the slice bends in spacetime. Time symmetry means $K=0$; maximality means only ${\mathrm{tr}}_{g}K=0$. Neither follows from the other geometric hypotheses below.

For zero cosmological constant, use geometric constraint densities

$$\begin{array}{rl}2\mu & =R_g+({\mathrm{tr}}_{g}K{)}^2-|K{|}_g^2,\\ j& ={\mathrm{div}}_g(K-({\mathrm{tr}}_{g}K)g).\end{array}$$

The dominant energy condition is $\mu \ge |j{|}_g$. In a physical normalization these two densities are divided by $8\pi$; the inequality is unchanged. Scalar curvature $R_g$ measures a contraction of the curvature of the slice. In particular, time-symmetric data satisfying this condition have $R_g\ge 0$.

Let $\Omega$ be a smooth connected orientable exterior, complete with its compact nonempty boundary $S$ included, with one asymptotically flat end. This means that outside a compact region there are Euclidean coordinates in which $g$ approaches the Euclidean metric $\delta$ and $K$ decays. The derivative notation $O_j(r^{-q})$ requires the derivative of order $i\le j$ to decay like $r^{-q-i}$.

The ADM quantities $(E,P)$ are the energy and linear momentum computed by surface fluxes at spatial infinity. With the usual Schwarzschild normalization, their invariant mass is

$$m=\sqrt{E^2-|P{|}^2},E>|P|.$$

The inequality $E>|P|$ makes this a positive real mass. Changing to a rest frame sets $P=0$ and makes energy equal to mass; controlling energy in an arbitrary frame would give a weaker comparison.

Orient $S$ toward the exterior and infinity. Its future outward null expansion is

$${\theta }_{+}=H_S+{\mathrm{tr}}_{S}K.$$

Here $H_S$ is the sum of principal curvatures, positive for an outward Euclidean sphere. A weakly future outer trapped boundary has ${\theta }_{+}\le 0$; a marginally outer trapped surface, abbreviated MOTS, has ${\theta }_{+}=0$. These conditions describe the initial rate of area change along outgoing light rays. They impose no condition on the other null expansion.

The Penrose problem asks whether mass must be large enough to support the trapped region. This family contains thirteen manuscripts: neutral comparison and rigidity, charged versions, rotating versions and counterexamples to a weaker formulation, anti-de Sitter versions, and a transfer to hyperboloidal ends. Their different area and asymptotic conventions are essential.

🏷️ The area being compared

A full enclosing cut is the entire compact intrinsic boundary $\Gamma =\partial D$ of a connected smooth codimension-zero submanifold $D\subset \Omega$. The region $D$ is closed in $\Omega$, contains the entire sufficiently distant end, and has interior in $\mathrm{int}\Omega$. Its boundary is smooth, embedded, and two-sided. Every component counts, including portions coinciding with $S$.

Define

$$A_{\min }(g)=\underset{\Gamma }{\inf }|\Gamma {|}_g.$$

The competitor $D=\Omega$ contributes $|S{|}_g$, not zero: intrinsic manifold boundary differs from relative topological boundary. Cuts need not be connected, minimal, or trapped, and the infimum need not be attained. In general $A_{\min }\le |S{|}_g$.

If $S$ is outer area-minimizing, every full cut has area at least $|S{|}_g$, so equality holds between these two areas. Outermostness is a separate condition: in the connected horizon classes below, it excludes a full weakly future trapped cut lying entirely in the open exterior. Minimizing area and having a specified null expansion are different requirements.

For spatial dimension $n$, put

$$\begin{array}{c}{\omega }_{n-1}=|{\mathbb{S}}^{n-1}|,\\ r_A=(A_{\min }/{\omega }_{n-1}{)}^{1/(n-1)}.\end{array}$$

This area radius is chosen so that a round sphere of radius $r_A$ has the same area as the enclosing infimum. The neutral comparison is

$$m\ge \frac{1}{2}{r}_A^{n-2}.$$

In dimension three this is $m\ge \sqrt{A_{\min }/(16\pi )}$. The Schwarzschild horizon has area $16\pi m^2$, explaining the normalization. Replacing $A_{\min }$ by an arbitrary trapped boundary’s larger area strengthens the statement and is not justified.

🏷️ Neutral comparison and its analytic construction

The September numerical manuscript claims the neutral bound for every $n\ge 3$ under strong differentiated decay

$$\begin{array}{c}g-\delta =O_6(r^{-q}),\\ K=O_5(r^{-1-q}),\\ \frac{n-2}{2}<q<n-2,\end{array}$$

with the smooth one-ended geometry above, integrable constraint densities, finite ADM limits, the dominant energy condition, and weak future trapping on every boundary component. Its initial theorem assumes $A_{\min }>0$ and $E>|P|$; a subsequent corollary removes the timelike assumption while retaining positive enclosing area. In dimensions three and four the companion results permit $O_2$ metric decay and $O_1$ tensor decay with $q>(n-2)/2$, and establish positivity and timelikeness as part of the conclusion. No maximality, symmetry, or outermostness is required for these numerical statements. (OpenAI, 2026)

The October synthesis also treats designated ends in complete data with finitely many asymptotically flat ends. Its three- and four-dimensional admissible exterior classes differ in which other ends must be excluded. These extensions use their own full-cut definitions; their area infima cannot simply be identified with the one-ended infimum defined here. The synthesis collects the weak-decay neutral results, charged numerical results, neutral rigidity, and the local conformal anti-de Sitter argument. It does not assert charged rigidity. (OpenAI, 2026a)

The main deformation has two unknowns: a graph height $f$ and a conformal factor. A representative boundary-graph construction uses

$$\hat{g}=e^{4t/(n-2)}(g+\ell (t{)}^{2}df\otimes df).$$

The nonnegative graph term increases lengths, but a small conformal factor could destroy the area. A lower bound $t\ge -\epsilon$ prevents that. Indeed, if $h\ge cg$ as quadratic forms, every eigenvalue of $h$ restricted to an $(n-1)$-plane is at least $c$ times the corresponding reference eigenvalue. Taking the square root of the determinant gives

$$\begin{array}{rl}|\Gamma {|}_h& \ge c^{(n-1)/2}|\Gamma {|}_g,\\ A_{\min }(h)& \ge c^{(n-1)/2}{A}_{\min }(g).\end{array}$$

This controls all cuts at once, without choosing a minimizing surface.

The graph scalar-curvature identity expresses $R_{\hat{g}}$ through the original constraints, favorable squares, and a divergence term. The coupled scalar equation controls that divergence. Its integral also controls the energy change at infinity, so the curvature and mass estimates are linked. On an exterior with an actual inner boundary, the inner flux must also have the correct sign or a sufficiently small negative error.

The boundary-deformation companion explains the required order: prepare strict energy and trapping inequalities; construct geometric height and slope barriers; obtain the conformal lower bound; then derive uniform ellipticity and solve by continuation. Classical regularity constants may depend on a fixed large deformation parameter. Only the estimates used in the final flux limit must have the stronger quantitative dependence. A formal scalar identity alone does not establish existence or these bounds. The separate four-dimensional maximal construction retains a noncompact tensor tail and has its own hypotheses. (OpenAI, 2026a)

After the deformation produces nonnegative scalar curvature and a strict inner barrier, a minimizing enclosure supplies the Riemannian comparison boundary. The conformal-flow companion treats every dimension, including frontiers with a singular set of dimension at most $n-8$. Its hypotheses require a smooth metric extending through the frontier, outer minimization, local perimeter minimization on both sides, integrable nonnegative scalar curvature with the stated pointwise decay, and an asymptotically flat end. The conclusion counts the full perimeter and requires neither spin nor connectedness of the frontier. (OpenAI, 2026b)

In its harmonic-flat reduction, the conformal flow preserves frontier area while its mass satisfies

$$m^{\prime }(s)=-2(m(s)-c(s))\le 0.$$

Here $c(s)$ is the normalized capacity, a Dirichlet-energy quantity for the harmonic function equal to one on the frontier and zero at infinity. A mass–capacity comparison gives the sign. Estimates of the limiting exterior radius then give the sharp area bound. The detailed treatment of singular frontiers is a substantive source input, not a consequence of this one-line monotonicity formula.

🏷️ Charged bounds and the outer branch

In three dimensions let $\mathcal{E},\mathcal{B}$ be electric and magnetic fields with zero divergence. Their charges are their fluxes at infinity divided by $4\pi$. Write $Q=(Q_E^2+Q_B^2{)}^{1/2}$. In the geometric density normalization, the charged matter condition is

$$\begin{array}{c}{\mu }_m=\mu -|\mathcal{E}{|}^2-|\mathcal{B}{|}^2,\\ j_m=j-2(\mathcal{E}\times \mathcal{B}{)}^{♭},\\ {\mu }_m\ge |j_m|.\end{array}$$

Subtracting electromagnetic energy but forgetting its momentum would change the hypothesis. The field flux forms ${\iota }_{\mathcal{E}}d{V}_g$ and ${\iota }_{\mathcal{B}}d{V}_g$ are closed; retaining these forms during metric changes preserves the fluxes without requiring global electromagnetic potentials.

The charged three-dimensional paper assumes $O_2(r^{-q})$ metric decay and $O_1(r^{-1-q})$ tensor decay with $q>1/2$, fields $O_1(r^{-2})$, integrable total constraint densities, and finite ADM and charge limits. It allows arbitrary ADM momentum and either the future trapping sign on all components or a fixed independent future/past sign on each component. It claims positivity of enclosing area, a timelike ADM vector, and

$$m\ge Q,r_A\le m+\sqrt{m^2-Q^2}.$$

For $r_A>Q$, this is equivalent to the familiar polynomial bound

$$m\ge \frac{1}{2}\left(r_A+\frac{Q^2}{r_A}\right).$$

For $0<r_A\le Q$, the upper-area bound follows already from $m\ge Q$; the polynomial lower bound is not equivalent there. At $r_A=Q$ the two numerical lower bounds coincide. (OpenAI, 2026e)

To see the branch issue directly, put $r_{\pm }=m\pm \sqrt{m^2-Q^2}$. They satisfy $r_{-}{r}_{+}=Q^2$ and $r_{-}\le Q\le r_{+}$. When $r_A>Q$, requiring $r_A\le r_{+}$ places $r_A$ between the two roots of $r^2-2mr+Q^2$, so the polynomial is nonpositive. When $r_A<Q$, the upper-area condition does not exclude $r_A<r_{-}$, where that polynomial is positive.

For $n\ge 4$, the charged reduction uses a normalized purely electric field. Set $k=n-1$ and

$$\begin{array}{c}Q={\omega }_k^{-1}{\int }_{\infty }g(\mathcal{E},\nu )d{A}_g,\\ \begin{array}{rl}2{\mu }_m& =R_g+(\mathrm{tr}K{)}^2-|K{|}^2\\ & -k(k-1)|\mathcal{E}{|}^2.\end{array}\end{array}$$

The electric contribution is dimension dependent. With $j_m=\mathrm{div}(K-(\mathrm{tr}K)g)$, the assumptions include ${\mu }_m\ge |j_m|$, $\mathrm{div}\mathcal{E}=0$, field decay $O_1(r^{-k})$, the weak metric/tensor decay with $q>(n-2)/2$, integrability, finite fluxes, and the same componentwise trapping convention. Defining $X_A=r_A^{n-2}$, the conclusion is

$$\begin{array}{c}m\ge |Q|,\\ X_A\le m+\sqrt{m^2-Q^2}.\end{array}$$

The equivalent polynomial bound uses $(X_A+Q^2/X_A)/2$ only on the outer branch $X_A>|Q|$, with equality of formulations at the threshold. This paper explicitly takes the strong-decay neutral numerical theorem as its input. It does not assume a higher-dimensional charged theorem. Its second scalar equation converts electric flux into an energy decrease while retaining a lower bound for every enclosing area; the neutral comparison then completes the reduction. (OpenAI, 2026a)

🏷️ Rotation and electromagnetic tails

For an axisymmetric exterior let $\eta$ generate rotations with period $2\pi$. Angular momentum is a surface flux involving $K(\nu ,\eta )$. In electrovacuum, electromagnetic momentum can move between surfaces, so the gravitational flux alone need not be conserved. After a duality rotation and a suitable magnetic potential $\alpha$, the conserved quantity is

$$\begin{array}{rl}{J}_{\mathrm{tot}}(\Sigma )& =\frac{1}{8\pi }{\int }_{\Sigma }(K-\tau g)(\nu ,\eta )dA\\ & +\frac{1}{4\pi }{\int }_{\Sigma }\alpha (\eta )g({\mathcal{E}}^{\prime },\nu )dA,\end{array}$$

where $\tau =\mathrm{tr}K$ and ${\mathcal{E}}^{\prime }$ is the rotated electric field.

The Kerr–Newman manuscript concerns smooth source-free electrovacuum data on a three-dimensional ball exterior with one end, rotational symmetry, a connected spherical outermost and outer area-minimizing future MOTS, and positive mass in the rest frame $P=0$. It specifies

$$\begin{array}{c}g-\delta =O_4(r^{-1}),\\ K=O_3(r^{-3}),\\ {\mathcal{E}}^i=Q_E{x}^i/r^3+O_3(r^{-3}),\\ {\mathcal{B}}^i=Q_B{x}^i/r^3+O_3(r^{-3}).\end{array}$$

Integrable constraint densities and finite fluxes are also required. These Coulomb leading terms make the electromagnetic correction vanish at infinity; thus the conserved total flux agrees there with gravitational ADM angular momentum. The physical area branch

$$A\ge 4\pi \sqrt{Q^4+4J^2}$$

is an explicit hypothesis, with $A=|S|=A_{\min }$ and $J=J_{\mathrm{tot}}$. The claimed inequality is

$$m^2\ge \frac{A}{16\pi }+\frac{Q^2}{2}+\frac{\pi (Q^4+4J^2)}{A}.$$

Maximality is not assumed. Equality on the strict area branch identifies the original data as an admissible slice of a subextremal dyonic Kerr–Newman exterior, smoothly meeting its future horizon or bifurcation sphere. The numerical endpoint is included, but extremal rigidity is not claimed. (OpenAI, 2026e)

The electromagnetic-tail companion changes two features: it assumes only $O_3(r^{-2})$ fields, without Coulomb leading terms, and uses the bare gravitational ADM flux as $J$. It constructs maximal electrovacuum examples with both charges zero but both fields nonzero. For amplitude $s\in [0,1]$ and large transition radius $L$,

$$\begin{array}{c}J=-4s^2/15,\\ m=2+O(s^2/L),\\ A=64\pi +O(s^2/L^2).\end{array}$$

The electromagnetic correction at infinity is $+4s^2/15$. Thus an angular momentum of fixed size survives in the gravitational flux even as the extra field energy becomes small. At $s=1$, the defect in the displayed Kerr–Newman inequality tends to $-1/225$: the terms $m^2$ and $A/(16\pi )$ both tend to $4$, while $4\pi J^2/A$ tends to $1/225$.

The examples retain a connected outermost outer area-minimizing future MOTS and the strict physical area branch. A continuity argument also gives equality examples that are not Kerr–Newman slices. This does not contradict the preceding theorem, whose asymptotics and conserved-flux definition are different. (OpenAI, 2026d)

🏷️ Equality must recover the original slice

A comparison metric may have scalar curvature and area properties that the original metric lacks. Equality after taking a sequence of numerical limits does not identify that comparison metric with the original pair $(g,K)$.

The neutral rigidity companion instead varies the original mass–area deficit. The enclosing infimum can change minimizer, so its one-sided metric derivative is the minimum of area derivatives over all minimizing enclosures:

$$a^{\prime }(g;h)=\underset{C}{\min }\frac{1}{2}{\int }_{{\partial }^{\ast }C}{\mathrm{tr}}_{T{\partial }^{\ast }C}hd{A}_g.$$

Here $a(g)=A_{\min }(g)$, $h$ is a metric variation, and ${\partial }^{\ast }C$ is the reduced boundary of a finite-perimeter enclosure. Separating the constraint and area variations produces a lapse, a shift, and initially a measure supported on minimizing sheets. Outermostness is used to eliminate unwanted interior area sources. The remaining adjoint equations produce a stationary field; further estimates make the relevant quotient static. Classification of that quotient is followed by recovery of the original metric, second fundamental form, and smooth horizon attachment. (OpenAI, 2026e)

The neutral all-dimensional strong-decay theorem assumes a connected outermost outer area-minimizing future MOTS and a timelike ADM vector. Its model is Schwarzschild–Tangherlini. The weak-decay four-dimensional class also has a connected horizon. The weak-decay three-dimensional finite-component theorem additionally places the exterior in complete boundaryless data with finitely many ends and uses an outermostness condition that permits competitors retaining some original components. Equality then forces one component. Consequently disconnected horizons in that class have strict inequality, without a uniform positive deficit being asserted.

The charged three-dimensional and higher-dimensional papers classify nonextremal equality in their respective connected outermost outer area-minimizing horizon classes. The models are Reissner–Nordström and Reissner–Nordström–Tangherlini, with the actual original invariant mass and charges. A smooth horizon cross-section or the bifurcation sphere is allowed. In higher dimensions the induced magnetic two-form must vanish: a generic boosted charged slice need not satisfy that purely electric requirement. No classification behind the boundary, at extremality, or for disconnected charged equality boundaries is obtained. (OpenAI, 2026i) (OpenAI, 2026a)

🏷️ Anti-de Sitter ends and local conformal branches

With cosmological constant $\Lambda =-3$, the three-dimensional reference metric is hyperbolic,

$$b=\frac{d{r}^2}{1+r^2}+r^2{\sigma }_{{\mathbb{S}}^2}.$$

The global anti-de Sitter papers take $(g,K)$ asymptotic to $(b,0)$. Their constraints, now in physical normalization, include

$$16\pi \mu =R_g+6+(\mathrm{tr}K{)}^2-|K{|}^2.$$

They require one spherical end, $g-b\in C_{\tau }^{2,\alpha }$ and $K\in C_{\tau }^{1,\alpha }$ for $3/2<\tau <3$, and the corresponding weighted integrability of $\mu$, $|j|$, and $|K{|}^2$. Here the weight controls tensor norms and covariant derivatives in $b$, not Euclidean coordinate components.

Four metric fluxes against the hyperbolic static potentials $\sqrt{1+r^2}$ and $r{\omega }_i$ give a mass covector $(p_0,p_1,p_2,p_3)$. Future timelikeness is an assumption, and

$$m_{\mathrm{AH}}=\sqrt{p_0^2-p_1^2-p_2^2-p_3^2}.$$

The spatial components are metric mass charges, not ADM momentum. Their Lorentz covariance concerns spatial changes of hyperbolic chart; it does not assert invariance under arbitrary changes of spacetime slice.

For the dominant energy condition and the same future trapping sign on every boundary component, the claimed bound is

$$\begin{array}{c}{m}_{\mathrm{AH}}\ge \frac{r_A+r_A^3}{2},\\ r_A=\sqrt{A_{\min }/(4\pi )}.\end{array}$$

The maximal companion assumes $\mathrm{tr}K=0$ and makes no equality classification. The nonmaximal companion removes that assumption and adds original-data rigidity in the connected outermost outer area-minimizing MOTS class. Both numerical statements permit disconnected boundary and arbitrary compact topology. (OpenAI, 2026h) (OpenAI, 2026h)

The model explains the cubic term: the Schwarzschild–AdS horizon solves $1+r^2-2M/r=0$, giving $M=(r+r^3)/2$. The nonmaximal equality theorem includes the match between $M$ and the computed metric mass. An arbitrary slice of the model does not automatically have that mass match.

The local conformal companion gives a different theorem. Fix a positive-mass Schwarzschild–AdS metric $g_0$, a decaying transverse-traceless seed $q$ satisfying ${\mathrm{div}}_{g_0}q=0$ and ${\mathrm{tr}}_{g_0}q=0$, and a specified decaying positive solution branch of the conformal constraint problem. Its data have the form

$$g_{\epsilon }={\varphi }_{\epsilon }^4{g}_0,K_{\epsilon }=\epsilon {\varphi }_{\epsilon }^{-2}q.$$

The theorem uses the actual MOTS boundary area, without requiring outermostness or outer area minimization. It assumes the stated finite future-timelike mass fluxes and proves the inequality for sufficiently small nonnegative $\epsilon$, with the interval depending on the fixed seed and branch. It does not construct a branch for every seed. Radial seeds give equality; nonradial seeds give strict inequality for sufficiently small positive parameters. (OpenAI, 2026b)

The deficit has zero constant and linear terms. Its quadratic coefficient is nonnegative but has a four-dimensional kernel, so second order alone does not settle the theorem. A fourth-order calculation proves positivity on the nonradial quadratic-null directions. Radial directions instead satisfy an exact conserved-mass identity. Weighted remainder estimates connect these finite expansions to the actual mass and area. This explains both the fixed-branch quantifier and the need to treat the quadratic kernel.

🏷️ Hyperboloidal replacement and the Bondi bound

The CKS companion has zero cosmological constant and a hyperboloidal end asymptotic to $(b,b)$, rather than the anti-de Sitter end $(b,0)$. Its specified Cha–Khuri–Sakovich expansions involve leading angular coefficients in both $g$ and $K$. Their prescribed combination defines an initial-data Bondi charge $(E_B,P_B)$; the theorem assumes $E_B>|P_B|$ and writes $m_B=\sqrt{E_B^2-|P_B{|}^2}$. Identifying this functional with a Bondi–Sachs charge at null infinity requires an appropriate compatible spacetime development, which is not assumed here. (OpenAI, 2026c)

The geometric replacement theorem constructs asymptotically flat data $(G_R,k_R)$ that agree with the original data on every prescribed compact interior for large $R$, preserve completeness and the dominant energy condition, and satisfy

$$\begin{array}{c}{G}_R\ge (1-{\epsilon }_R)g,{\epsilon }_R\to 0,\\ P_R=0,E_R\to m_B.\end{array}$$

The metric comparison implies $A_{\min }(G_R)\ge (1-{\epsilon }_R)A_{\min }(g)$ in dimension three. If the original boundary is weakly future trapped and its enclosing area is positive, the neutral asymptotically flat inequality gives

$$E_R\ge \sqrt{\frac{(1-{\epsilon }_R)A_{\min }(g)}{16\pi }}.$$

Taking $R\to \infty$ yields $m_B\ge \sqrt{A_{\min }(g)/(16\pi )}$, including disconnected boundaries. Canonical horizon-regular Schwarzschild hyperboloidal examples attain equality at every positive mass. The replacement theorem itself needs no boundary expansion condition; that condition enters when applying the neutral comparison.

🏷️ Key steps and proof outlines

The routes share control of all enclosing cuts, but their analytic mechanisms differ. Equality uses an additional variation of the original data; it does not follow merely by declaring every approximate comparison sharp.

flowchart TD
    A["Original initial data and full enclosing cuts"] --> B["Graph deformation and flux estimates"]
    B --> C["Neutral curvature and area comparison"]
    B --> D["Charge converted to mass decrease"]
    B --> E["AdS stress or bad-volume area transfer"]
    C --> F["Numerical mass-area inequality"]
    D --> F
    E --> F
    A --> G["Vary the original deficit at equality"]
    G --> H["Constraint adjoint and stationary classification"]

The charged graph identity preserves flux before estimating energy

The three-dimensional charged manuscript’s filled construction uses $\bar{g}=g+l(t{)}^{2}df\otimes df$ and $\hat{g}=e^{2t}\bar{g}$. Its scalar unknown is $Z=t+\frac{1}{4}\log D$, where $D=1+l(t{)}^2|df{|}_g^2$. This is a legitimate global change of variables: if $p=l^{\prime }/l\ge 0$ and $v=l^2|df{|}^2/D$, differentiation in $t$ gives $1+pv/2\ge 1$. Hence every pair $(Z,df)$ determines one smooth $t$, including at zero slope.

Why this combination? Consider the auxiliary Lorentzian metric $-l^2(dz-df{)}^2+\bar{g}$. In the original $z$ coordinates it has spatial metric $g$, while in the coordinate $z-f$ it is static over $\bar{g}$. Computing its scalar curvature in these two descriptions identifies a divergence and the difference between the original second fundamental form $K$ and the auxiliary stationary one. The conformal scalar-curvature formula then introduces derivatives of $t$. The differential identity

$$2dZ=(2+pv)dt+H^f(w,\cdot ),H^f=l{D}^{-1/2}{\nabla }^{2}f,w=l{D}^{-1/2}\nabla f,$$

collects these derivatives into the flux used by the coupled equation. Its remaining quadratic terms are treated by square completion. This is an exact geometric identity; existence of a solution, coercivity and the lower bound on $t$ are separate analytic steps. (OpenAI, 2026m)

The October synthesis spells out how electromagnetic momentum enters the comparison. Rotate the electric and magnetic fields by a constant orthogonal matrix so their charges become $(Q,0)$; call the resulting fields $X_e,X_o$. Their electromagnetic energy-plus-momentum expression is unchanged. Completing a square gives

$$\begin{array}{rl}{I}_g(w)& =|X_o+w\times X_e{|}^2+|X_e{|}^2-|w\times X_e{|}^2\\ & \ge |\alpha {|}_{\bar{g}}^2,\alpha ={\iota }_{X_e}d{V}_g.\end{array}$$

Indeed the inverse graph metric has eigenvalues $(d,1,1)$ in a frame parallel to $w$, with $d=1-|w{|}^2$. For $X_e=(x_1,x_2,x_3)$, the two-form norm is $x_1^2+d(x_2^2+x_3^2)$, exactly the last expression. Retain the rotated flux pair $(\alpha ,0)$ and rotate back. These closed forms have the original two charges, even though their pointwise fields have changed. This explains how the construction can preserve both total charges without claiming preservation of the original electromagnetic fields. (OpenAI, 2026h)

Charge is converted into a controlled mass decrease

The higher-dimensional reduction adds a second scalar equation after the graph construction. For $k=n-1$, let $X_A=(A_{\min }/{\omega }_k{)}^{(k-1)/k}$. Choose $0<s<1$. Its conformal floor is

$$y_0=\frac{\log (1-s^2)}{k-1}.$$

The metric change $\gamma =e^{2y}{g}_b$ therefore retains a factor $e^{k{y}_0}$ in every full-cut area, hence a factor $1-s^2$ in its area power $X_A$. Meanwhile a flux profile with boundary value approaching $ks$ decreases the ADM energy by a quantity approaching $s|Q|$. The profile’s differential inequality is chosen so that the mixed gradient-electric term can be absorbed by a nonnegative quadratic form; retained gradient coercivity pays for the transition band. Applying the neutral inequality to the resulting smooth trapped exterior gives

$$m\ge s|Q|+\frac{1-s^2}{2}{X}_A.$$

For $X_A>|Q|$, maximize at $s=|Q|/X_A$ (or take $s\downarrow 0$ when $Q=0$) to obtain $(X_A+Q^2/X_A)/2$. For $X_A\le |Q|$, the supremum is reached as $s\uparrow 1$ and gives only $m\ge |Q|$. This optimization is the precise reason the outer branch cannot be replaced by the polynomial bound for arbitrarily small enclosing area. (OpenAI, 2026a)

Rotation turns enclosing area into a meridional crossing constraint

In the Kerr–Newman comparison, write the axisymmetric metric as $g_r=h_r+X_r^{2}d{\varphi }^2$. A meridional arc has swept area $2\pi$ times its length in ${\mathfrak{h}}_r=X_r^2{h}_r$. Removing loops from a crossing arc and rounding its axis endpoints changes that weighted length arbitrarily little, because $X_r$ vanishes linearly on the axis. Thus the enclosing-area lower bound becomes a lower bound for every crossing between the two original axis sides.

The scalar-curvature inequality then becomes a two-dimensional curvature inequality. With $u=\log X_r$ and the electromagnetic/twist potential map $\Phi$,

$$X_r^2{K}_{{\mathfrak{h}}_r}=\frac{1}{2}{R}_{g_r}+|du{|}_{h_r}^2,K_{{\mathfrak{h}}_r}\ge |d\Phi {|}_{{\mathfrak{h}}_r,G}^2.$$

Here $G$ is the target metric specified by those potentials. A minimizing enclosure can be chosen rotation invariant: perimeter submodularity shows unions of minimizers remain minimizing, so the largest-volume minimizer is unique and inherits the symmetry. Stability and positive scalar curvature make its components spheres. Only the sphere separating the original axis sides has the prescribed charge and angular-momentum jumps; additional spheres have zero jumps. These facts connect the geometric cut constraint to the model comparison while retaining the conserved potential normalization. (OpenAI, 2026i)

The tail counterexamples identify exactly why the normalization matters. For the axial Killing field $\eta$ and magnetic potential $hd\varphi$, Cartan’s identity gives ${\iota }_{\eta }d(hd\varphi )=-dh$. Combining it with the electrovacuum momentum constraint yields

$${\mathrm{div}}_g(K(\cdot ,\eta {)}^{{♯}_g}+2h\mathcal{E})=0.$$

The examples have gravitational angular flux $-4s^{2}c(r{)}^2/15$ and electromagnetic correction $+4s^{2}c(r{)}^2/15$, where $c(r)$ is the radial cutoff. Their conserved sum is zero at every radius. The nonzero limiting gravitational term alone therefore cannot be substituted for the conserved quantity when the electromagnetic tail is allowed. The asymptotic mass and area calculation above measures the resulting defect explicitly. (OpenAI, 2026f)

Neutral deformation: a floor is useful only if its flux cost vanishes

The neutral numerical manuscript couples the graph equation to a divergence equation. Its lower-floor penalty has a favorable sign in the scalar-curvature identity, but the same term can increase ADM energy when integrated. These are two separate estimates. If $\ell$ is the small lapse scale near the floor and $\eta$ is the graph regularization parameter, the source bounds the adverse integrated cost by

$${\mathcal{P}}_N\left(\ell +\frac{{\ell }^2}{\eta }\right),$$

where ${\mathcal{P}}_N$ is polynomial in the deformation parameter $N$. The height-derivative estimate instead costs ${\mathcal{P}}_N\eta /\ell$. Choosing

$$\eta ={\ell }^{3/2}$$

makes both competing ratios equal to ${\ell }^{1/2}$. At fixed floor tolerance, $\ell$ decreases faster than these polynomial losses grow. Thus the floor protects every enclosing area while its mass penalty disappears.

The radial weight exponent is equally deliberate:

$$\frac{n}{4}<b_1<\min (n-2,n/2).$$

The first inequality makes the square of the weight $r^{-2b_1}$ integrable; the upper bounds support the height barrier and the pointwise floor contradiction. This is how pointwise curvature, global area, and the flux limit are made compatible. Existence of solutions to the coupled system remains a separate elliptic input. (OpenAI, 2026l)

Boundary graphs: obtain a trace bound before invoking regularity

The boundary-graph companion cannot simply discard the inner boundary term. Let $\bar{g}=g+l^{2}d{f}^2$, $D=1+l^2|df{|}_g^2$, and

$$W=l\bar{\nabla }f=\frac{l\nabla f}{D}.$$

Its $\bar{g}$-norm is at most one. In the source’s weighted equations, $uW=Lw$; $u,L$ are positive scalar weights and $w$ is the normalized graph-gradient vector. Direct differentiation bounds

$$|\mathrm{div}(uW)|\le {\mathcal{P}}_{N}u(1+\sqrt{\mathcal{T}}),$$

where $\mathcal{T}$ is the coercive quadratic energy. The corresponding covector contraction is controlled by the elliptic energy norm, again with polynomial constants.

The already established signed boundary slopes give $w_{\nu }={ϵ}_{B}a$, with $a\ge 1/2$ and ${ϵ}_B=\pm 1$ fixed on each collar. For a smooth function $\phi \ge 0$ supported in the collar, apply the divergence theorem to $-{ϵ}_{B}uW\phi$, remembering that the outward domain normal is $-\nu$. Its boundary contribution is $La\phi \ge L\phi /2$. This produces the weighted collar trace inequality; replacing $u$ by $\sqrt{D}$ produces its unweighted counterpart. No inverse power of the tiny floor scale $\ell$ enters. The proof therefore supplies the boundary control needed for energy iteration before later regularity estimates are invoked. (OpenAI, 2026d)

Singular frontiers: preserve perimeter mass in blow-ups

In ambient dimension $n\ge 8$, the conformal-flow companion treats singular minimizing frontiers through a quantitative estimate for weighted perimeter minimizers. In lower dimensions, the regularity input gives an empty singular set. Absorb a positive perimeter density $\hat{\rho }$ into

$$g_{\ast }={\hat{\rho }}^{2/(n-1)}g.$$

The variational problem becomes ordinary $g_{\ast }$-perimeter minus a smooth volume forcing. Under a blow-up of radius $r_j$, the metric tends to Euclidean and the forcing is multiplied by $r_j$, so it disappears.

Convergence of characteristic functions alone could hide canceling sheets. To obtain convergence of perimeter as well, glue the limiting set inside a ball to the approximating set outside. Coarea chooses the cutting radius where their trace discrepancy tends to zero. Minimality bounds the approximating perimeter by the limit perimeter plus this vanishing seam cost and the vanishing forcing term. Lower semicontinuity supplies the reverse inequality. The resulting strict convergence controls tangent-plane measures and prevents lost varifold mass.

If singular blow-ups could approach cones with $n-7$ translation directions, this compactness would identify the limit with a minimizing boundary. Splitting off those directions leaves a minimizing cone in ambient dimension at most seven, hence a halfspace by the imported regularity theorem. Flat-boundary regularity would then contradict singularity of the approximants. The exclusion holds uniformly at every sufficiently small scale. The cited quantitative-stratification theorem converts it into

$$\mathrm{Vol}(B_t(\mathcal{S}))\le C{t}^8$$

for compact singular subsets. Both minimization and the all-scale exclusion are necessary; bounded mean curvature alone would not give codimension eight. (OpenAI, 2026e)

Original-data rigidity: differentiate an infimum without selecting one surface

Let $\mathcal{T}(g)$ be the compact family of minimizing enclosures. The rigidity proof needs continuity of their tangent-plane measures, not just convergence of areas. Uniform confinement, finite-perimeter compactness, and convergence of total perimeter provide precisely that continuity. Consequently

$$a_E^{\prime }(h)=\frac{1}{2}{\int }_{{\partial }^{\ast }E}{\mathrm{tr}}_{T{\partial }^{\ast }E}hd{A}_g$$

depends continuously on $E\in \mathcal{T}(g)$.

For a smooth metric path $g_t=g+th+O(t^2)$, the area density has a Taylor expansion uniform over the compact Grassmann bundle:

$$P_{g_t}(F)=P_g(F)+t{a}_F^{\prime }(h)+O(t^2)P_g(F).$$

Testing each fixed minimizer bounds the upper right derivative by ${\min }_E{a}_E^{\prime }(h)$. Conversely choose any $g_t$-minimizer $E_t$. Since $P_g(E_t)\ge a(g)$,

$$\frac{a(g_t)-a(g)}{t}\ge a_{E_t}^{\prime }(h)-O(t).$$

Every sequence $t\downarrow 0$ has a subsequence converging, with plane measures, to a $g$-minimizer. This gives the matching lower bound. Thus the right derivative is the minimum of all active area derivatives even when minimizers switch. This envelope calculation is the correct input to separation and the later area-supported multiplier; assuming a differentiable choice of one minimizing surface would omit possible active constraints. (OpenAI, 2026g, p.c ompactness and derivative of least area)

Nonmaximal AdS data: coarea saturates a bounded flux

In the nonmaximal AdS construction, a nonlinear stress potential $f_D$ equals a large constant $D$ on an inner boundary $T$ and tends to zero at infinity. Write $A_{\ast }$ for the infimum of the areas of its full enclosing cuts; attainment is unnecessary. The constitutive function $t(\sigma )$ is chosen so that

$$X_D=t(|d{f}_D|)\frac{\nabla f_D}{|d{f}_D|}$$

is divergence free and has norm at most one; its value at critical points is defined continuously. The end expansion $f_D=z{C}_D(\omega )+O(z^2)$ gives flux

$${\mathcal{F}}_D={\int }_{{\mathbb{S}}^2}{C}_D^{2}d\omega \le A_{\ast }.$$

Integrating $\mathrm{div}(f_D{X}_D)$ also gives

$$D{\mathcal{F}}_D=\int t(|d{f}_D|)|d{f}_D|dV.$$

For $0<\lambda <1$, the superlevel region $f_D>\lambda D$ has volume bounded independently of $D$, while $\sigma (1-t(\sigma ))$ is bounded. Coarea and full-cut minimality therefore yield

$$D{\mathcal{F}}_D\ge (1-\lambda )D{A}_{\ast }-C_{\lambda }.$$

Let $D\to \infty$ and then $\lambda \downarrow 0$: the flux saturates the full enclosing area. Filling bounded complementary components ensures that almost every level really supplies a full cut. Hölder then turns this quadratic flux into the cubic moment

$$\frac{1}{4\pi }{\int }_{{\mathbb{S}}^2}{C}_D^{3}d\omega \ge {\left(\frac{{\mathcal{F}}_D}{4\pi }\right)}^{3/2},$$

which enters the stress/bending mass comparison. This mechanism explains how the AdS cubic area term arises from an area-saturating flux. (OpenAI, 2026j)

Maximal AdS data: a small bad volume protects the area

The maximal companion permits a small region where the conformal factor loses its pointwise lower bound. Its area-transfer theorem shows why that region cannot carry an entire low-area enclosure. Write the comparison metric schematically as $\hat{h}\approx x\bar{g}$, with $\bar{g}\ge g_{\mathrm{base}}$ and $d{V}_{\bar{g}}=Wd{V}_{\mathrm{base}}$. The bad set is $\mathcal{B}=\{x<1-\delta \}$. A projection weight $\rho$ satisfies ${\int }_N\rho d{A}_{\mathrm{base}}\ge (1-q)A_{\ast }$ for every relevant enclosure.

If the initial minimal boundary had area below $(1-ϵ)A_{\ast }$, weak inverse-mean-curvature flow would pass through a fixed positive time interval with areas between $(1-ϵ)A_{\ast }$ and $(1-ϵ/2)A_{\ast }$. On every such level, the good region contributes too little projected area; therefore

$${\int }_{N_t\cap \mathcal{B}}\rho x^{-1}d{A}_{\hat{h}}\ge c_A>0.$$

The scalar-curvature and topology estimates bound $J(t)={\int }_{N_t}{H}^{2}dA$. Hölder gives

$${\int }_{N_t\cap \mathcal{B}}\frac{{\rho }^{3/2}{x}^{-3/2}}{H}dA\ge \frac{c_A^{3/2}}{\sqrt{J(t)}}.$$

Integrating in flow time and applying coarea forces a positive lower bound on ${\int }_{\mathcal{B}}{\rho }^{3/2}Wd{V}_{\mathrm{base}}$, contradicting the constructed small bad volume. The factor $x^{-3/2}$ cancels the three-dimensional conformal volume factor. This explains both the weight exponent and the uniformity as the initial area varies. (OpenAI, 2026k)

Local AdS perturbations: the boundary correction is part of the square

For the local conformal theorem, put $B={\mathrm{Ric}}_{g_0}+3g_0$, $L={\Delta }_{g_0}-3$, and $Q(v)=\mathrm{Hess}v-vB$. Let $N$ be the background static lapse, $a$ the horizon radius, and $\kappa ={\partial }_{s}N{|}_S>0$. The positive boundary operator is $D_S=-{\Delta }_S+2\kappa /a$. For a smooth transverse-traceless tensor $k$ of finite angular type, with decay $k=O(r^{-\gamma })$ for $\gamma >3/2$, take the decaying solution of

$$Lv=0,v{|}_S=D_S^{-1}(k_{ss}{|}_S).$$

Then $Q(v)$ is transverse-traceless and has the same normal-normal boundary value as $k$. The one-form $\omega =Ndv-vdN$ satisfies

$$\mathrm{sym}\nabla \omega =NQ(v).$$

Integration by parts against divergence-free $k$, with the correct inner normal, gives

$$\int N|k{|}^2-\kappa {\int }_{S}v{D}_{S}v=\int N|k-Q(v){|}^2.$$

Thus the boundary term is absorbed exactly, without assuming a sign for $k_{ss}$.

On a quadratic-null direction, the source changes the slice through fourth-order jets and transports the boundary along a null direction. The remaining minimal-graph correction contributes an order-four boundary area term of precisely the form subtracted above. The quartic coefficient becomes a sum of the corrected tensor square, a trace-free leaf-curvature square, and a nonnegative spherical lapse energy. Vanishing forces the first two angular geometric jets to be round; the separate degree-one obstruction then excludes nonradial kernel directions. These are finite Taylor constructions, with the source’s remainder estimates needed to conclude positivity on the actual prescribed branch. (OpenAI, 2026b)

CKS replacement: angular heat flow cancels the leading constraint error

The CKS replacement preserves the angular mass aspect while changing the end. After balancing, its initial function $f_{\ast }$ has mean $m_B$ and no degree-one spherical harmonic. Evolve it by ${\partial }_{t}f={\Delta }_{{\mathbb{S}}^2}f$. Writing ${\lambda }_{\ell }=\ell (\ell +1)$, define

$$\psi (t)=\sum _{\ell \ge 2}\frac{e^{-{\lambda }_{\ell }t}}{1-{\lambda }_{\ell }/2}{f}_{\ast ,\ell },\mathcal{T}=({\nabla }^2\psi {)}^{\mathrm{TF}}.$$

The missing degree-one mode is essential: it would make the denominator zero. The spherical identity $\mathrm{div}({\nabla }^2\psi {)}^{\mathrm{TF}}=d(\psi +\Delta \psi /2)$ gives $\mathrm{div}\mathcal{T}=df$. All nonconstant modes decay at least as $e^{-6t}$, while the mean remains $m_B$.

Let $v/r$ be the tangential tensor coefficient being lowered from one to zero. Away from the activation cutoff, choose

$$t^{\prime }(r)=\frac{1}{2r(1+v^2)},\tau =\frac{r(v^{\prime }-v/r)}{1+v^2}\mathcal{T}.$$

The heat equation cancels the leading angular Laplacian in the scalar constraint; the divergence identity cancels the leading tangential momentum error. A positive integrable radial mass padding absorbs the controlled remainders. The final cutoff of $v$ is delayed to $r\asymp R^2$, where $v$ and its scaled derivatives are already $O(R^{-1})$. No division by a vanishing $v$ is needed. This preserves the mass mean while producing a smooth asymptotically flat, zero-momentum tail. (OpenAI, 2026c)

🏷️ Connections and scope

Spectral scalar curvature and width discusses a different geometric comparison driven by scalar curvature; the present family protects an enclosing area while controlling mass at infinity. The collection guide places these geometric PDE arguments alongside the differential geometry articles. The preceding discussion audits hypotheses and selected mechanisms, with elementary derivations of area transfer, the charged branch distinction, the tail defect, and the Bondi limit. It does not independently verify the global elliptic constructions, singular-frontier flow, stationary classification, or complete research proofs.

🏷️ Manuscripts and verification scope

• The spacetime Penrose inequality with charge and original-data rigidity — pinned manuscript and companion source.
• A Charged Reduction of the Spacetime Penrose Inequality in Spatial Dimensions at Least Four — pinned manuscript and companion source.
• Spacetime Penrose inequalities: enclosing area, charge, and rigidity — pinned manuscript and companion source.
• The Kerr–Newman Penrose Inequality for Axisymmetric Electrovacuum Exteriors — pinned manuscript and companion source.
• Electromagnetic tails and the Kerr–Newman Penrose inequality — pinned manuscript and companion source.
• The nonmaximal anti-de Sitter Penrose Inequality and original-data rigidity — pinned manuscript and companion source.
• The Penrose inequality for maximal asymptotically hyperbolic initial data — pinned manuscript and companion source.
• A local Penrose inequality for conformal perturbations of Schwarzschild–anti-de Sitter data — pinned manuscript and companion source.
• The spacetime Penrose inequality and enclosing area — pinned manuscript and companion source.
• Conformal flow and the Riemannian Penrose inequality with minimizing frontiers — pinned manuscript and companion source.
• Equality and rigidity in the spacetime Penrose inequality — pinned manuscript and companion source.
• Boundary graph deformations for the spacetime Penrose inequality — pinned manuscript and companion source.
• Area-controlled end replacement and the Bondi–Penrose inequality in the CKS class — pinned manuscript and companion source.

The family scope page describes selected formalization coverage. Only the CKS end-replacement manuscript is listed. The selected coverage is area-controlled replacement and canonical Schwarzschild equality examples. The general Bondi–Penrose inequality for possibly disconnected weakly trapped boundaries remains unformalized. A retained conditional result for connected marginal boundaries assumes the asymptotically flat inequality and is not selected as a substitute for the main inequality. No Lean build, axiom audit, or manuscript-to-formal-statement equivalence check was performed for these notes.

OpenAI collection provides prerequisites and the reading sequence; coverage register records completed and pending families.

🐻  OpenAI 2026a. A Charged Reduction of the Spacetime Penrose Inequality in Spatial Dimensions at Least Four.

🐻  OpenAI 2026b. A local Penrose inequality for conformal perturbations of Schwarzschild–anti-de Sitter data.

🐻  OpenAI 2026c. Area-controlled end replacement and the Bondi–Penrose inequality in the CKS class.

🐻  OpenAI 2026d. Boundary graph deformations for the spacetime Penrose inequality.

🐻  OpenAI 2026e. Conformal flow and the Riemannian Penrose inequality with minimizing frontiers.

🐻  OpenAI 2026f. Electromagnetic tails and the Kerr–Newman Penrose inequality.

🐻  OpenAI 2026g. Equality and rigidity in the spacetime Penrose inequality.

🐻  OpenAI 2026h. Spacetime Penrose inequalities: enclosing area, charge, and rigidity.

🐻  OpenAI 2026i. The Kerr–Newman Penrose Inequality for Axisymmetric Electrovacuum Exteriors.

🐻  OpenAI 2026j. The nonmaximal anti-de Sitter Penrose Inequality and original-data rigidity.

🐻  OpenAI 2026k. The Penrose inequality for maximal asymptotically hyperbolic initial data.

🐻  OpenAI 2026l. The spacetime Penrose inequality and enclosing area.

🐻  OpenAI 2026m. The spacetime Penrose inequality with charge and original-data rigidity.