Cartan-Hadamard isoperimetric conjecture

Cartan-Hadamard isoperimetric conjecture

OpenAI's claim

Model isoperimetry and equality rigidity: The manuscript claims sharp comparison with the constant-curvature model ball for every finite-volume finite-perimeter set in a simply connected complete manifold whose sectional curvatures do not exceed $\kappa\le0$; in the Euclidean comparison, bounded equality sets are precisely regions isometric to round Euclidean balls, up to null sets.

Sharp fillings in CAT$0$: The paper claims the Euclidean-optimal integral filling inequality for compactly supported integer cycles of dimension at least two in proper CAT$0$ spaces; arbitrary integer multiplicities and ambient dimensions are permitted.

Mathematical statement of the claimed result

Mathematical claim 1.

Let $n\ge2$, $\kappa\le0$, and assume $(M^n,g)$ is a smooth Riemannian manifold, simply connected and complete, satisfying $\sec_g\le\kappa$. For a measurable set of finite volume and finite ambient perimeter, the claim is \[P_g(E)\ge\mathcal I_{\sqrt{-\kappa}}(\operatorname{vol}_g E).\] Here $s_b(r)=r$ for $b=0$ and $s_b(r)=\sinh(br)/b$ otherwise, $A_b(r)=n\omega_ns_b(r)^{n-1}$, $V_b(r)=n\omega_n\int_0^rs_b(t)^{n-1}\,dt$, and $\mathcal I_b=A_b\circ V_b^{-1}$; $\omega_n$ is unit-ball volume. When $\kappa=0$, equality $P_g(E)=n\omega_n^{1/n}\operatorname{vol}(E)^{(n-1)/n}$ for bounded positive-volume $E$ holds exactly when, up to a volume-zero set, $E$ is an open region isometric to a round Euclidean ball.

Mathematical claim 2.

For every proper CAT$(0)$ metric space $X$ and integer $n\ge2$, each compactly supported integral $n$-current $T$ with $\partial T=0$ is claimed to bound a compactly supported integral $(n+1)$-current $S$ with \[\partial S=T,\qquad\mathbf M(S)\le\frac{\mathbf M(T)^{(n+1)/n}}{(n+1)^{(n+1)/n}\omega_{n+1}^{1/n}}.\] No ambient-dimension or multiplicity restriction is imposed, and the constant is claimed optimal. Its manifold corollary gives $\operatorname{area}(\partial\Omega)\ge d\omega_d^{1/d}\operatorname{vol}(\Omega)^{(d-1)/d}$ for smooth bounded domains in a smooth Cartan--Hadamard $d$-manifold, $d\ge3$. For relatively compact measurable $E$ of finite perimeter the assertion instead uses the ambient variation $P_g(E)=|D\mathbf1_E|_g(M)$: \[P_g(E)\ge d\omega_d^{1/d}\operatorname{vol}_g(E)^{(d-1)/d}.\] For smooth $\Omega$, $P_g(\Omega)=\operatorname{area}_g(\partial\Omega)$.

Outline of the argument

Model isoperimetry and equality rigidity: A sharp critical Sobolev inequality is established on immersed hypersurfaces using radial sphere maps and positive weights compatible with Jacobi comparison. This controls the boundary geometry needed for the isoperimetric estimate. Approximation extends the bound to finite-perimeter sets, and the equality analysis forces the geometric comparison to be exact, identifying a Euclidean ball region.

Sharp fillings in CAT$0$: Induction reduces a failing inequality to a compact CAT$0$ space and an extremizing cycle of small normalized mass. Inward geodesic variations estimate its curvature-like behavior in current charts. The resulting sharp mass comparison contradicts extremality, and Euclidean examples show that the coefficient cannot be improved.

Formal verification

Family 337 describes Lean coverage for the sharp integral-current CAT$0$ filling estimate and its optimal coefficient; the curved-model perimeter theorem is separate.

Model isoperimetry and equality rigidity: No manuscript-specific Comparator is listed in the pinned inventory.

The documented selection does not cover the entire manuscript assertion relevant here.

Sharp fillings in CAT$0$: The manuscript links FillingCoefficient.lean, SharpCAT0Filling.lean.

The documented selection includes this manuscript’s main assertion; that assertion retains the restrictions stated above.

This site has not run the Lean development. These links describe OpenAI’s selected formal statements and should be read with their stated scope.

Status caveat

OpenAI reports that an internal model produced the manuscripts. Their arguments have not been independently verified as proofs by this site. Any portion outside the documented formal scope may contain mistakes, and the catalogue label reports the announced mathematical claim.

Sources

• OpenAI, Generalized Cartan–Hadamard isoperimetry and Euclidean equality rigidity, 2026-09-23; citation key OAI:Generalized-Cartan-Hadamard-isoperimetry-and-Euclidean-equality-rigidity-September-23-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main; Statement source: thm:equality.
• OpenAI, Sharp integral fillings in CAT$0$ spaces, 2026-09-23; citation key OAI:Sharp-integral-fillings-in-CAT(0)-spaces-September-23-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: cor:cartan-hadamard; Statement source: thm:main.
• Family 337: Lean scope.
• Comparator FillingCoefficient.json.
• Comparator FillingCoefficient.lean.
• Comparator SharpCAT0Filling.json.
• Comparator SharpCAT0Filling.lean.
• OpenAI catalogue entry.

Cartan-Hadamard isoperimetric conjecture

Definitions and mathematical statement

A Cartan--Hadamard manifold is a complete simply connected Riemannian manifold with sectional curvature at most zero. For a bounded measurable region $\Omega$ of finite perimeter and positive volume, write $V=\operatorname{vol}(\Omega)$ and $P(\Omega)$ for the total variation of its indicator function, agreeing with boundary area for smooth regions. If $\omega_n$ is the Euclidean unit-ball volume, the conjecture is \[ P(\Omega)\ge n\omega_n^{1/n}V^{(n-1)/n}. \] Equality is to occur only when the region, up to null sets, is isometric to a Euclidean ball in the source's rigidity sense. This does not force the entire ambient manifold to be Euclidean. The assertion is not limited to geodesic balls or smooth convex regions.

Formulation and concept references: [S1].

Short English statement

Does nonpositive sectional curvature preserve the Euclidean lower bound for perimeter at fixed volume, with equality only for a Euclidean ball?

Research attempt

Scope and current-source qualification.

This attempt was prepared by GPT-6 Astra Ultra on 27 September 2026. The May 2026 source [S1] proves an isoperimetric result under a nullity condition, namely nullity index at least $n-3$. Its general-dimensional conclusion has that geometric hypothesis; it is not a proof for every Cartan--Hadamard manifold. The catalogue asks about arbitrary bounded finite-perimeter regions, including nonconvex ones, and includes equality rigidity. Those requirements are retained.

The present work proves the sharp inequality and its rigidity for geodesic balls in every Cartan--Hadamard manifold. It then tests why the radial comparison used in that proof does not cover arbitrary regions. The result is a restricted theorem, not the general conjecture. Assume $n\ge2$ below. In dimension one, a complete simply connected manifold is a line and a nonempty bounded finite-perimeter region has at least two boundary points, with equality for an interval up to null sets.

Radial comparison with its sign checked.

Fix $o\in M$. The Cartan--Hadamard theorem gives global geodesic polar coordinates away from $o$, without cut or conjugate points. Write \[ d\operatorname{vol}=j(r,\theta)\,dr\,d\theta, \qquad A(r)=\int_{\mathbb S^{n-1}}j(r,\theta)\,d\theta, \qquad V(r)=\int_0^r A(s)\,ds. \] Here $d\theta$ is the standard unit-sphere measure, of total mass $n\omega_n$. We claim \[ \partial_r\log j(r,\theta)\ge\frac{n-1}{r}, \qquad j(r,\theta)\ge r^{n-1}. \tag{1} \] The geometric input can be seen directly from the index form. For a transverse endpoint vector $v$ along a unit radial geodesic $\gamma:[0,r]\to M$, let $J$ be its Jacobi field with $J(0)=0$ and $J(r)=v$. The second fundamental form of the distance sphere is \[ \operatorname{Hess}r(v,v)= \int_0^r\left(|D_sJ|^2- \langle R(J,\dot\gamma)\dot\gamma,J\rangle\right)\,ds. \] Sectional curvature at most zero makes the curvature contribution nonnegative. Parallel-transporting to one tangent space and applying Cauchy--Schwarz gives $\int_0^r|D_sJ|^2\,ds\ge|v|^2/r$. Taking the transverse trace proves the first inequality in $1$. Since $j(r,\theta)/r^{n-1}\to1$ as $r\downarrow0$, integration gives the second and also monotonicity of that ratio. The sign is expansion relative to Euclidean space, not contraction.

A sharp inequality for every geodesic ball.

For $0<s\le r$, monotonicity of $j(s,\theta)/s^{n-1}$ gives \[ A(s)\le A(r)(s/r)^{n-1},\qquad V(r)\le rA(r)/n. \tag{2} \] Also $A(r)\ge n\omega_n r^{n-1}$ by $1$. Combining the inequalities in the correct direction, \[ \begin{split} A(r)^n &=A(r)^{n-1}A(r)\\ &\ge (nV(r)/r)^{n-1}\,n\omega_n r^{n-1} =n^n\omega_n V(r)^{n-1}. \end{split} \] Therefore \[ A(r)\ge n\omega_n^{1/n}V(r)^{(n-1)/n}. \tag{3} \] This proves the selected constant for all radii and centers within the class of geodesic balls. Merely knowing both $A(r)$ and $V(r)$ are at least their Euclidean counterparts would not suffice; the upper relation between $V(r)$ and $A(r)$ in $2$ is essential.

Rigidity for the restricted theorem.

Suppose equality holds in $3$ for a positive radius. Equality must hold in both factors used above, in particular $A(r)=n\omega_n r^{n-1}$. Since $j(r,\theta)\ge r^{n-1}$ continuously in $\theta$, equality of the integrals forces equality for every direction. Monotonicity in $1$ then gives $j(s,\theta)=s^{n-1}$ for $0<s\le r$.

The shape operator $S_s$ of every radial sphere satisfies $S_s\ge s^{-1}I$ by the index-form proof and has trace $(n-1)/s$. Thus $S_s=s^{-1}I$. If $g_s$ denotes the angular metric in polar coordinates, the relation $\partial_sg_s=2S_sg_s$ implies $\partial_s(s^{-2}g_s)=0$. Its limit at zero is the round unit-sphere metric, so \[ g=ds^2+s^2g_{\mathbb S^{n-1}} \] throughout the ball. Hence the ball is isometric to a Euclidean ball. This argument imposes no curvature condition outside it. It therefore has exactly the local form of rigidity requested, within the class we have actually proved.

A model-space check.

In constant curvature $-\kappa^2$ with $\kappa>0$, put $f(r)=\sinh(\kappa r)/\kappa$ and $c=n\omega_n$. Then $A(r)=cf(r)^{n-1}$ and $V(r)=c\int_0^rf(s)^{n-1}\,ds$. Since $f'(r)=\cosh(\kappa r)\ge1$, \[ \int_0^rf(s)^{n-1}\,ds\le f(r)^n/n. \] Substitution gives $3$ immediately, with strict inequality for $r>0$. This independent check detects both the perimeter exponent and the normalization of $\omega_n$.

There is also a local curvature check. The usual normal-coordinate volume-density expansion gives, at a center $o$, \[ \begin{split} V(r)&=\omega_nr^n\left(1- \frac{\operatorname{Scal}(o)}{6(n+2)}r^2+O(r^3)\right),\\ A(r)&=n\omega_nr^{n-1}\left(1- \frac{\operatorname{Scal}(o)}{6n}r^2+O(r^3)\right). \end{split} \] Consequently the ratio of the two sides of $3$ is \[ 1-\frac{\operatorname{Scal}(o)}{2n(n+2)}r^2+O(r^3). \tag{4} \] Nonpositive sectional curvature implies nonpositive scalar curvature, so the leading correction has the expected sign. This Taylor calculation is a consistency check, not a substitute for the all-radius proof or a proof about arbitrary shapes.

Why polar comparison does not settle arbitrary regions.

Even for a smooth star-shaped region described by $0<r<R(\theta)$, the volume is $\int\int_0^{R(\theta)}j(s,\theta)\,ds\,d\theta$, whereas its boundary area involves angular derivatives of $R$ and the full angular metric. Pointwise lower bounds for the radial density increase both integrals and do not order their isoperimetric ratio.

The failed shortcut is already visible in Euclidean space. Ignoring the angular-gradient term gives only \[ P\ge\int R^{n-1}\,d\theta,\qquad V=\frac1n\int R^n\,d\theta. \] On the sphere equipped with normalized probability measure, the $L^{n-1}$ norm is at most the $L^n$ norm. Thus these two formulas alone give the wrong direction for the desired perimeter lower bound. Taking $R$ large on a small angular patch and small elsewhere makes the ratio of those radial moments particularly small. Smoothing such a profile produces a large angular-gradient contribution to the true area; omitting it is the error, not a counterexample to isoperimetry.

A region may also fail to be star-shaped about every center, or have several components and holes. Choosing a single center then does not provide a one-valued radial boundary function at all. The radial-ball argument has not addressed those cases.

The finite-perimeter extension requires a real premise.

If the sharp inequality were established for every bounded smooth region in a fixed manifold, it would extend to bounded finite-perimeter regions by strict smooth approximation: choose smooth regions $\Omega_j$ with characteristic functions converging in $L^1$, volumes converging to $V$, and perimeters converging to $P(\Omega)$. Passing to the limit in the smooth inequality then gives the same bound. Local mollification in coordinate charts and coarea provide this standard BV approximation in a relatively compact neighborhood.

Lower semicontinuity alone, $P(\Omega)\le\liminf_jP(\Omega_j)$, would not transfer a lower bound from the approximants: it has the wrong direction. Perimeter convergence is the crucial extra feature. Moreover, approximating arbitrary regions by smooth ones does not approximate them by geodesic balls, so it cannot extend our restricted theorem $3$ to the whole target. Equality rigidity for arbitrary BV regions would require a further argument even after the inequality.

Verification and final status.

The index form, the two inequality directions in $2$, and the equality case have been checked separately. The local script checks the scalar coefficient in $4$ algebraically and tests the hyperbolic formulas numerically only as a normalization check. The infinite-dimensional geometric proof is not certified by those samples.

UNRESOLVED. The sharp inequality and Euclidean-ball rigidity are proved here for geodesic balls. The first missing step is a perimeter-versus-volume estimate for arbitrary smooth regions without the radial-ball assumption. Concrete next targets are an appropriate rearrangement preserving the necessary area control, a global boundary curvature argument beyond the nullity class of [S1], and equality control under strict BV approximation. None has been supplied for general Cartan--Hadamard manifolds.

Sources

• S1 Mohammad Ghomi, Isoperimetric and total curvature inequalities in Cartan-Hadamard manifolds with nullity, arXiv:2605.24638v1, 2026. https://arxiv.org/html/2605.24638v1. Formulation source.