Metric inequalities with scalar curvature (Q724265)

We establish several inequalities for manifolds with positive scalar curvature and, more generally, for the scalar curvature bounded from below. As far as geometry is concerned, these inequalities appear as generalisations of the classical bounds on the distances between conjugate points in surfaces with positive sectional curvatures. The technique of our proofs is based on the Schoen-Yau descent method via minimal hypersurfaces, while the overall logic of our arguments is inspired by and closely related to the torus splitting argument in Novikov's proof of the topological invariance of the rational Pontryagin classes. This large article (82 pages) is organized in 11 sections, the last of which, an appendix, has 12 subsections. In the first section the key inequalities, that are generalized in the next nine sections, are formulated and commented. These inequalities are: the torical \(\frac{2\pi}{n}\)-inequality, the normal radius inequality for \(T^{n-1}\subset S^n\), and the quadratic decay theorem. The torical \(\frac{2\pi}{n}\)-inequality is for torical bands which are smooth manifolds homeomorphic to tori times intervals. If \(V=\mathbb{T}^{n-1}\times [-1,1]\) is an n-dimensional torical band with the boundary \(\partial(V)=\partial_{-}(V)\cup \partial_{+}(V)=(\mathbb{T}^{n-1}\times {-1})\cup (\mathbb{T}^{n-1}\times {+1})\), and if \(g\) is a smooth Riemannian metric on \(V\), with scalar curvature \(Sc(g)\) satisfying \(Sc(g)\geq \sigma>0\), then the distance between the two boundary components of \(V\) satisfies the inequality \(\mathrm{dist}_g(\partial_{-}(V),\partial_{+1}(V))\leq 2\pi\sqrt{\frac{n-1}{\sigma n}}<\frac{2\pi}{\sqrt{n}}.\) For all \(n\)-dimensional torical bands this inequality can be written \[ \mathrm{dist}(\partial_{-},\partial_{+1})\leq \frac{2\pi}{n}. \] Then the normal focal radius of a smooth submanifold \(Y\) in a Riemannian manifold \(X\), denoted \(\mathrm{rad}^\odot(Y)=\mathrm{rad}^\odot(Y\subset X)\) is the maximal \(r\) such that the normal exponential map \(exp:T_{\perp}(Y)=T(X)|Y\ominus T(Y)\rightarrow X\) is locally one-to-one on the subset of vectors \(\nu\in T_{\perp}(Y)\), such that \(\|\nu\|<r\). Then the normal radius inequality for \(\mathbb{T}^{n-1}\subset S^n\) asserts that if a smooth hypersurface \(Y\) in the unit \(n\)-sphere is homeomorphic to the \((n-1)\)-torus, then \[ \mathrm{rad}^\odot(Y)\leq \frac{\pi}{n}. \] Finally, for \(X\) a complete Riemannian manifolds \(\underset{B(R)}{\min} Sc(X)\) denotes the minimum of scalar curvature (function) of \(X\) on the ball \(B(R)=B_{x_0}(R)\subset X\) for some centre point \(x_0\in X\). The quadratic decay theorem states that if \(X\) is homeomorphic to \(\mathbb{T}^{n-2}\times \mathbb{R}^2\), then there exists a constant \(R_0=R_0(X,x_0)\), such that \[ \underset{B(R)}{\min} Sc(X)\leq \frac{4\pi^2}{(R-R_0)^2},\forall R\geq R_0. \] In Section 2 these propositions are proved in the form of first generalizations as follows. A band is a manifold \(V\) with two disjoint non-empty subsets in the boundary \(\partial(V)=\partial_{-}(V)\cup \partial_{+}(V)\). A band \(V\) is called proper if \(\partial_{\pm}(V)\) are unions of connected components of \(\partial(V)\). Band maps \(V\rightarrow \underline{V} \) are those continuous ones which respect to these \(\pm\)-boundaries, \(\partial_{\pm}(V)\rightarrow \partial_{\pm}(\underline{V})\). If a band \(V\) is endowed with a Riemannian metric then the width of \(V\) is the distance between \(\partial_{-}(V)\) and \(\partial_{+}(V)\), that is the infimum of length of curves in \(V\) between \(\partial_{-}(V)\) and \(\partial_{+}(V)\). A compact proper orientable band \(V\) is called over-torical if it admits a band map to the toric band \(f:V\rightarrow \underline{V}=\mathbb{T}^{n-1}\times [-1,1], n=\dim V\), with non-zero degree. Then the \(\frac{2\pi}{n}\)-inequality for over-torical bands states: Over-torical bands with scalar curvature at least \(Sc(S^n)=n(n-1)\) satisfy the inequality \[ \mathrm{width}(V)\leq\frac{2\pi}{n}. \] Then the generalisation of the quadratic decay theorem states: If a complete orientable Riemannian \(n\)-manifold \(X\) admits a proper continuous map \(X\rightarrow \mathbb{T}^{n-2}\times \mathbb{R}^2\) of non-zero degree, then the minima of the scalar curvature of \(X\) over concentric \(R\)-balls in \(X\) satisfy \[ \underset{B(R)}{\min} Sc(X)\leq \frac{4\pi^2}{(R-R_0)^2}, \] for some \(R_0\) and all \(R\geq R_0\). In Sections 3--9, the author formulates and proves further generalisations and refinements of the above propositions. So, if for \(f\) a continuous map of non-zero degree from an oriented Riemannian \(n\)-manifold \(X\) to \(S^n\), \(\mathrm{Lip}(f)=\underset{x_1\neq x_2}{\sup}\frac{\mathrm{dist}_{S^n}(f(x_1),f(x_2))}{\mathrm{dist}_X(x_1,x_2)}\), then the spherical Lipschitz bound Theorem (Section 3) states: If the scalar curvature of a (possibly incomplete) Riemannian \(n\)-manifold is bounded from below by \(Sc(S^n)=n(n-1)\), then all continuous maps \(f\) from \(X\) to the sphere \(S^n\) (and also to the hemisphere \(S^n_{+}\)) of non-zero degrees satisfy \[ \mathrm{Lip}(f)>\frac{c}{\pi\sqrt{n}}, \] for a certain constant \(c>\frac{1}{3}\). Then the following topics are considered: \(\frac{4\pi}{n}\)-bound on width and related inequalities for iso-enlargeable bands (Section 4), \(\frac{4\pi}{n}\)-inequality for SYS (Shoen-Yau-Schick)-bands (Section 5), \(\frac{2\pi}{n}\)-inequality for SYSE (SYS-Enlargeable)-bands (Section 6), an inequality for the principal curvatures of all smooth immersions \(Y\rightarrow B(1)\subset \mathbb{R}^n\) (Section 7), Symmetrization of Riemannian manifolds with point-wise control of the scalar curvature (Sections 8 and 9). Throughout these sections the author indicates additional applications and articulates several conjectures. But he also refers to possible limitations of his arguments. Section 10 is entitled ``Comparison with results obtained with twisted Dirac operators'', and it is a suggestion for a new approach. The appendix is intended to help readers less aware of the topic to be able to track the article. The author explains in greater detail the following: (a) Smoothing hypersurfaces with no decrease of their mean curvatures, and (b) Smoothing Riemannian metrics with no decrease of theirs scalar curvatures. Next, he (c) elucidates some properties of minimal hypersurfaces in open manifolds needed for the width inequalities. Finally, the author (d) summarises the topological obstructions for \(Sc>0\) on closed manifolds which follow from his inequalities, and (e) highlights several conjectures mentioned in the main body of the article. Conjecture A. If a closed manifold \(X\) is SYS over a non-torsion cohomology class \(h\) in an aspherical space \(K\), then \(X\) admits no metrics with \(Sc>0\). Conjecture B. Products of manifolds \(X\) as the above A by \([\widetilde{\uparrow}0]\)-oversymplectic ones admit no metrics with \(Sc>0\). Conjecture C. If a closed manifold \(V_0\) of dimension \(n-1\geq 5\), admits no metric with \(Sc>0\) then Riemannian bands \(V\) diffeomorphic to \(V_0\times [-1,1]\) which have \(Sc(V)\geq \sigma >0\), satisfy the sharp width inequality, \(\mathrm{width}(V)\leq 2\pi\sqrt{\frac{n-1}{n}}\). Conjecture D. Let \(g_0\) stands for the standard Riemannian metric on the unit sphere \(S^n\) with sectional curvature 1. If a Riemannian metric \(g\) on \(S^n\) minus a point satisfies \(g\geq g_0\) and \(Sc(g)\geq Sc(g_0)=n(n-1)\), then \(g_0=g\). Conjecture D is concerned with closed \(n\)-manifolds \(X\), such that \(X\) minus a point admits no complete metric with \(Sc>0\). Conjecture E. Let \(V\) be a Riemannian manifold homeomorphic to \(\mathbb{T}^2\times [-1,1]\) with sectional curvature greater than or equal to 1 everywhere. Then \(\mathrm{width} (V)=\mathrm{dist}(\partial_{-}(V),\partial_{+}(V))\leq \frac{\pi}{2}\). Conjecture \(\mathrm{E}_{+\infty}\). Let \(Y\subset S^N, N=n,n+1,\dots,\infty\), be a submanifold homeomorphic to the product of \(n\) closed manifolds of dimensions at least 1, e.g., homeomorphic to the \(N\)-torus and let \(U\supset Y\) be a neighbourhood of \(Y\) in \(S^N\) which admits a retraction to \(Y\). Then \(\operatorname{dist}(Y,\partial U)\leq \operatorname{arcsin} \frac{1}{\sqrt{n}}\). Conjecture -E. Every compact smooth manifold \(Y^n\) of dimension \(n\) admits a smooth embedding to the sphere \(S^{2n}\) such that all principal curvatures of the image satisfy \(\mathrm{curv}(Y^n\subset S^{2n})\leq\mathrm{const}<\infty\), say for \(\mathrm{const}=1000\). Conjecture \(\mathrm{E}_{T^n}\). The minimal constant \(\beta\), such that the \(n\)-torus admits a smooth immersion to \(S^{n+1}\) with principal curvatures at most \(\beta\) is asymptotic, for \(n\rightarrow \infty\), to \(\mathrm{const}. n^\beta\), for some \(\beta>1\). Problem F. Let \(U\subset \mathbb{R}^N\) and \(V\subset \mathbb R^n\), \(n\leq N\), be open subsets, e.g., balls \(B^N(r)\) and \(B^n(R)\), \(R\geq r\). Evaluate the smallest \(\beta=\beta(U,V)>0\), such that \(V\) admits a smooth locally expanding immersion/embedding to \(U\) with the principal curvatures at most \(\beta\). Every conjecture is argued and discussed. In closing the article, the author concludes: The above A--F are only tips of the iceberg of what we don't know about the scalar curvature and nearabouts.