Integral scalar curvature and simplicial volume

Source and verification

This article explains family 335 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.

🏷️ Preliminaries: scalar curvature and simplicial volume

Scalar curvature ${\mathrm{Scal}}_g$ is the trace of the Ricci tensor. With the convention that round spheres have positive curvature, it equals twice the sum of sectional curvatures over coordinate two-planes in an orthonormal basis. In dimension two it is twice Gaussian curvature. In higher dimensions it retains much less directional information than a Ricci-curvature bound.

For a closed connected oriented $n$-manifold, real simplicial volume measures the least total absolute coefficient needed to represent its fundamental class:

$$\begin{array}{c}\Vert M\Vert =\underset{[c]=[M{]}_{\mathbb{R}}}{\inf }\Vert c{\Vert }_1,\\ c=\sum _j{a}_j{\sigma }_j,\Vert c{\Vert }_1=\sum _j|a_j|.\end{array}$$

Here the ${\sigma }_j$ are singular simplices and the sum is a finite real cycle: its oriented boundary cancels. The fundamental class records the orientation of the whole manifold. Real coefficients matter; this is not a minimum number of simplices in a triangulation.

Define the negative part of scalar curvature by

$${\mathrm{Scal}}_g^{-}=\max (0,-{\mathrm{Scal}}_g).$$

The family asks whether a topologically complicated fundamental cycle forces a definite amount of negative scalar curvature, even when that curvature is concentrated in a small region.

Integral manuscript result

For every $n\ge 3$, there is $a_n>0$ depending only on $n$ such that every closed connected oriented smooth $n$-manifold and every smooth metric satisfy

$${\int }_M({\mathrm{Scal}}_g^{-}{)}^{n/2}d{V}_g\ge a_n\Vert M\Vert .$$

No spin or curvature-concentration restriction is imposed. (OpenAI, 2026)

The prerequisites are Riemannian geometry, integration, Hölder’s inequality, and the basic meaning of cycles and degree. More elaborate topology is used below through stated detection and realization interfaces.

🏷️ Scaling and the companion’s qualitative theorem

Replacing $g$ by ${\lambda }^{2}g$ multiplies scalar curvature by ${\lambda }^{-2}$ and volume by ${\lambda }^n$. Thus

$$\begin{array}{rl}& {\int }_M({\mathrm{Scal}}_{{\lambda }^{2}g}^{-}{)}^{n/2}d{V}_{{\lambda }^{2}g}\\ & ={\int }_M({\mathrm{Scal}}_g^{-}{)}^{n/2}d{V}_g.\end{array}$$

The exponent $n/2$ is exactly the scale-invariant one. A pointwise bound ${\mathrm{Scal}}_g\ge -{\kappa }^2$ gives

$$\Vert M\Vert \le a_n^{-1}{\kappa }^n{\mathrm{Vol}}_g(M).$$

The integral paper uses one major theorem from its companion: nonnegative scalar curvature on a closed oriented manifold of positive dimension forces zero real simplicial volume. The companion proves a stronger obstruction under strictly positive scalar curvature. (OpenAI, 2026b)

To state that obstruction, let $c_M:M\to B{\pi }_1(M)$ classify the universal cover. The target is a space with the same fundamental group and contractible universal cover. The manifold is rationally essential when its rational fundamental class has nonzero image under $c_M$.

The companion claims that a closed connected oriented manifold of dimension at least two with strictly positive scalar curvature is rationally inessential. The statement has no spin, fundamental-group, or upper-dimension restriction. Its proof constructs geometric degree data from the detected class; no development of group theory is needed for the discussion here.

Strict positivity cannot be replaced by nonnegativity in this rational-class conclusion. A positive-dimensional torus is rationally essential and has a flat metric, while its real simplicial volume is zero. Nonnegative scalar curvature does suffice for the weaker simplicial-volume vanishing statement, which treats a Ricci-flat alternative separately.

🏷️ A conformal reduction with an explicit dependency

Assume $\Vert M\Vert >0$. The companion’s vanishing theorem rules out every nonnegative-scalar-curvature metric on $M$. In particular, the constant scalar curvature of a Yamabe minimizer in any conformal class must be negative.

For the class of $g$, the Yamabe quotient is

$$\begin{array}{c}{Q}_g(u)=\frac{A_g(u)}{{\left({\int }_M{u}^{p}d{V}_g\right)}^{2/p}},\\ \begin{array}{rl}{A}_g(u)& =c_n{\int }_M|du{|}^{2}d{V}_g\\ & +{\int }_M{\mathrm{Scal}}_g{u}^{2}d{V}_g,\end{array}\end{array}$$

where $u>0$, $c_n=4(n-1)/(n-2)$, and $p=2n/(n-2)$. The Yamabe constant is its infimum. Existence of a smooth positive minimizer is an input from the Yamabe theorem.

Normalize the minimizing metric $h$ to ${\mathrm{Scal}}_h=-1$. Then

$$-Y(M,[g])={\mathrm{Vol}}_h(M{)}^{2/n}.$$

Let $I_g={\int }_M({\mathrm{Scal}}_g^{-}{)}^{n/2}d{V}_g$. Hölder gives

$$\begin{array}{rl}& {\int }_M{\mathrm{Scal}}_g{u}^{2}d{V}_g\\ & \ge -I_g^{2/n}{\left({\int }_M{u}^{p}d{V}_g\right)}^{2/p}.\end{array}$$

The gradient term is nonnegative, so $Y(M,[g])\ge -I_g^{2/n}$ and therefore

$${\mathrm{Vol}}_h(M)\le I_g.$$

It remains to prove $\Vert M\Vert \le C_n{\mathrm{Vol}}_h(M)$ for scalar curvature exactly $-1$. This reduction explains why qualitative vanishing alone is insufficient: the remaining argument must supply a constant uniform over every topology and every normalized metric.

🏷️ Detecting the fundamental cycle

For the remaining volume estimate, write $g$ for the normalized metric of scalar curvature $-1$.

Duality for simplicial volume supplies a bounded cocycle of norm at most one whose pairing $I$ with the fundamental class satisfies

$$I\ge \frac{1}{2}\Vert M\Vert .$$

One may think of the cocycle as a uniformly bounded rule assigning compatible numbers to simplices. Its compatibility makes its total evaluation depend only on the represented homology class.

The construction realizes this rule as a closed differential form on a simplex of probability vectors. A probability-valued map from the universal cover, compatible with deck transformations, pulls that form back to a detector with integral $I$ on $M$.

This converts a topological quantity into an integral to be bounded geometrically. The detector remains fixed while the metric and probability map are changed together.

🏷️ Weighted curvature and metric enlargement

For a metric $h$ and a weight $\varphi$, use

$$\mathcal{S}(h,\varphi )={\mathrm{Scal}}_h-2{\Delta }_h\varphi -|d\varphi {|}_h^2.$$

Graph deformations enlarge the metric while changing the weight so that the weighted volume is controlled. Prescribed weighted mean curvature contributes positive squares and favorable derivative terms to the scalar identity.

Applied successively to probability coordinates, the construction produces a measured family of auxiliary geometries, a retained density $\rho \le d{V}_g$, and a large energy coefficient $E$. Put $W=\log (d{V}_h/\rho )$. The central estimate has the form

$$\begin{array}{rl}\mathcal{S}(h,\varphi )\ge & -C_n+2W-n\log E\\ & +E|dF{|}_{h,\mathrm{HS}}^2.\end{array}$$

The logarithmic volume term will cancel volume expansion. The energy term pays for derivatives of the comparison map. The absolute smallness of $|dF|$ also controls overlaps among its labels.

The parameter $E$ cannot be chosen without regard to the later construction. The transfer theorem first specifies how large it must be in terms of the comparison complexity. The deformation theorem then supplies that value. This order prevents an implicit circular parameter choice.

🏷️ Degree estimates and the final cancellation

A second theorem bounds degrees of maps from controlled measured bands to cubes. It uses minimizing boundaries for a perimeter-minus-forcing functional, often called a $\mu$-bubble functional. Ordered boundaries reduce the dimension one coordinate at a time.

Their normal speed, or lapse, appears in both the scalar identity and the coarea formula. Keeping it in both places provides a cancellation. In high dimensions the minimizing boundaries may be singular, so the construction moves the residual map away from the relevant target near singular regions and pays for the change with weighted distance potentials. Smooth descent cannot simply be assumed.

The cocycle is then transferred to a balanced chain of dual cubes. Unbiased integer rounding preserves the balance identities. A reflected realization and an oppositely oriented mirror close unmatched faces while retaining a comparison side with the original oriented pushforward.

Here is the decisive dimensional bookkeeping. Let the dual cube have dimension $m$, put $q=n+m$, and choose comparison lengths

$$L_T=b_{\ast }\sqrt{E},L_N=b_{\ast }\sqrt{qE}.$$

Tangent cube directions contribute $L_T^m$ to volume. The degree bound divides by $L_T^m{L}_N^n$, while the scalar estimate supplies $E^{n/2}{e}^{-W}$. Thus

$$\begin{array}{rl}& \frac{L_T^m}{L_T^m{L}_N^n}{E}^{n/2}{e}^{-W}d{V}_h\\ & =b_{\ast }^{-n}{q}^{-n/2}\rho .\end{array}$$

The remaining detector and derivative estimates cost at most $C_n{q}^{n/2}$ on the straight comparison part, canceling the displayed power of $q$. For grid spacing $r$, the cell coefficient, dual-cube volume, and center density similarly satisfy

$$r^n{r}^m{r}^{-q}=1.$$

Seam contributions are controlled separately by small coordinate volume. The resulting bound is

$$I\le C_n\int \rho \le C_n{\mathrm{Vol}}_g(M).$$

Together with the detector lower bound and conformal reduction, this gives the attributed integral inequality. The measured deformation, singular degree theorem, and dual-chain realization are substantive source inputs; the final cancellation explains how they fit together.

🏷️ The companion’s positive-curvature obstruction

The qualitative companion begins with a nonzero rational classifying-map pairing. Rational essentiality does not automatically supply arbitrarily contracting nonzero-degree maps to spheres. The paper instead adds a parameter manifold and realizes the pairing as local degree.

At a large comparison scale $R$, it obtains finitely many coordinates with small derivatives in the original manifold directions:

$$\underset{i}{\max }|d{G}_i|\le J/R,\log q\le JR.$$

Parameter-direction derivatives are controlled later by enlarging the parameter metric. A graph change and a reciprocal warped circle shorten one coordinate. Across a complete pass, decreases of the inverse metric telescope on an $n$-dimensional cotangent space. Their trace is bounded by $n$, independently of the number $q$ of coordinates.

Consequently, the scalar loss depends logarithmically on $q$, rather than linearly. Successive passes divide the derivative bound by four and the allowed coordinate error by two. Both the total error and scalar loss are summable; the boundary gap preserving local degree survives.

The resulting nonzero-degree sphere maps become arbitrarily contracting while a positive scalar lower bound remains on suitable stabilized products. A local torical obstruction gives a contradiction with a threshold independent of the number of added circles. The order is important: first choose $R$, fixing the comparison dimension, then take a sufficiently large finite number of shortening passes.

The terminal obstruction is imported from the cited scalar-curvature literature. The companion’s contribution is the conversion of the detected class into data to which it applies, with the required scalar and differential estimates.

🏷️ Consequences and interpretation

An aspherical manifold has contractible universal cover, so its classifying map preserves its nonzero fundamental class. The companion therefore rules out positive scalar curvature on every closed aspherical manifold; the nonorientable statement follows through the orientation cover. It also claims that a nonnegative-scalar-curvature metric on a closed aspherical manifold is flat, using an additional Ricci-flat reduction and splitting argument.

For the integral theorem, positive simplicial volume forces a definite amount of negative scalar curvature in the scale-invariant integral. The theorem does not require negativity to be evenly distributed and does not provide an optimized numerical value of $a_n$.

🏷️ Key steps and proof outlines

The entropy increment has a telescoping scalar budget

The quantitative theorem needs more than the qualitative obstruction to nonnegative scalar curvature. For the normalized metric with scalar curvature $-1$, a sweep increases a coefficient $\epsilon$ to $(1+l)\epsilon$ while changing the metric and weight. Let $W_p$ be the logarithmic volume increase during this sweep. The source proves an increment inequality whose essential terms are

$$\begin{array}{rl}{\mathcal{S}}_{+}-(1+l)\epsilon e_{+}\ge & {\mathcal{S}}_0-\epsilon e_0+2W_p-n\log (1+l)\\ & -C_n{l}^2+P_p,\end{array}$$

where $e_0,e_{+}$ are the probability-map energies and $P_p\ge 0$ is a retained surplus. The volume increments add, while the logarithms record the multiplicative growth of the energy coefficient. Summing yields the combination $2W-n\log E$ in the final contract. The small quadratic errors are scheduled to have controlled total size. (OpenAI, 2026a)

The matrix identity behind the budget is particularly informative. If $h_{+}-h_0$ is the sum of the graph-coordinate square tensors, their energy measured in the final metric is

$$m_f=n-\mathrm{tr}(h_{+}^{-1}{h}_0)\le n(1-e^{-2W_p/n}).$$

The last inequality is arithmetic–geometric mean for the positive eigenvalues of $h_{+}^{-1}{h}_0$, whose product is $e^{-2W_p}$. It ties the energy gained by enlargement to its precise volume cost.

The bootstrap starts the iteration without spending volume

Initially the useful energy coefficient may be too small for the entropy increment. The preliminary sweeps are followed by the conformal reset $k=e^{-2W_p/(n-1)}{h}_{+}$. Its volume density satisfies

$$d{V}_k=e^{-W_p/(n-1)}d{V}_{h_0}\le d{V}_{h_0}.$$

The scalar conformal identity supplies a favorable gradient square. Repeated resets raise the energy coefficient into a fixed dimensional interval while keeping the negative scalar constant bounded. This produces the retained density $\rho \le d{V}_g$ that ultimately pays for the topological pairing. Metric enlargement alone would lose precisely this control.

Rounding must preserve a whole system of boundary equations

The bounded cocycle is transferred to coefficients on dual cubes. Rounding each coefficient independently would destroy the cancellation equations at common faces and hence change the represented pairing. Instead, the source encodes all local types in one finite vector and rounds inside the kernel of its incidence equations. The rounded vector has rational coordinates, exactly satisfies all balance equations, and has the original vector as its conditional expectation.

Finiteness of this vector follows from the label graph and a coloring that separates labels at distance at most two. A sufficiently large color coordinate identifies an anchor label and the appropriate flat sheet. Only the finite neighborhood of that anchor can affect the local coefficient. This also prevents the flat-sheet realization from branching where charts overlap.

The conditional transfer theorem first supplies a complexity threshold $E\ge {\Psi }_n(q)$. Only then is the entropy theorem invoked with a prescribed function dominating that threshold. The output estimate bounds the detected pairing by a dimensional constant times $\int \rho$, and hence by the original normalized volume. Together with the conformal reduction, this closes the scale-invariant integral inequality. The companion’s rational-inessentiality theorem supplies the qualitative exclusion needed at the start; it does not replace the quantitative entropy and transfer argument.

See Also

Spectral curvature and codimension-two width uses related weighted curvature, metric enlargement, and hypersurface methods to control whole fibers of maps. Its conclusion concerns geometric width rather than the simplicial norm of a fundamental cycle.

🏷️ Manuscripts and verification scope

• An integral scalar curvature bound for real simplicial volume — pinned manuscript and companion source.
• Positive scalar curvature forces rational inessentiality — pinned manuscript and companion source.

No family scope page was present at lean/docs/335.md in the pinned snapshot. This does not establish that all related formalization code is absent. 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. An integral scalar curvature bound for real simplicial volume.

🐻  OpenAI 2026b. Positive scalar curvature forces rational inessentiality.