Dirac and Plateau billiards in domains with corners (Q2248114)

The article intends to develop a theory of singular spaces in order to extend the well-known theorems on manifolds of positive scalar curvature. The author exposes many examples and conjectures, some of them included in this review. A byproduct would be to reveal the relations between Spin Geometry and Minimal Hypersurface Theory, not yet clear, since both are fundamental tools used to prove the main theorems belonging to the topic. The question which compact manifolds admit a metric with positive scalar curvature goes back to \textit{A. Lichnerowicz} [C. R. Acad. Sci., Paris 257, 7--9 (1963; Zbl 0136.18401)] who used spin geometry techniques. Later \textit{N. J. Hitchin} [Adv. Math. 14, 1--55 (1974; Zbl 0284.58016)] applied the Atiyah-Singer index theorem to prove that whenever the topological invariant \(\hat{A}\)-genus is non-zero there is no Riemannian metric \(g\) on \(X\) such that \(\mathrm{scal}_{g}\geq 0\). In the 70s, \textit{R. Schoen} and \textit{S.-T. Yau} [Manuscr. Math. 28, 159--183 (1979; Zbl 0423.53032)], using minimal hypersurfaces, proved similar non-existence result by assuming that \(X\) has enough codimension on homology classes intersecting non-trivially. Indeed, they settled Geroch's conjecture, in particular they proved that ``the torus \(T^{n}\) does not admit a metric of positive scalar curvature''. The present article focusses on two main issues: {\parindent=6mm\begin{itemize}\item[(1)] \(C^{0}\)-stability of a set of metrics whose scalar curvatures are bounded below by a continuous function. \item[(2)] the extension of the smooth techniques over singular spaces. \end{itemize}} Main Theorem. Let \(X\) be a smooth manifold and \(\kappa:X\rightarrow \mathbb{R}\) a \(C^{0}\)-function. Associated to \(\kappa\), consider the set \(\mathcal{U}_{\kappa}\) of \(C^{2}\)-smooth Riemannian metrics \(g\) on \(X\) such that the scalar curvature \(\mathrm{scal}_{g}\) of \(g\) satisfies \(\mathrm{scal}_{g}(x)\geq \kappa(x)\), for all \(x\in X\). So, \(\mathcal{U}_{\kappa}\) is closed under \(C^{0}\)-limits of Riemannian metrics on \(X\). In this way, if \(\{g_{n}\}_{n\in\mathbb{N}}\subset \mathcal{U}_{\kappa}\) and \(\lim g_{n}=g\) in the \(C^{0}\)-topology, then \(g\in\mathcal{U}_{\kappa}\). This results is an important stability theorem as the author believes that the geometry of manifolds of all dimensions with \(\mathrm{scal}_{g}\geq 0\) (and, possibly, with \(\mathrm{scal}\geq \kappa\)) is governed by integrals of the scalar curvature function \(\mathrm{scal}_{g}:X\rightarrow\mathbb{R}\) over surfaces in \(X\). Curiously, the main Theorem contrasts with Lohkamp's theorem which claims that every Riemannian metric \(g\) on a smooth manifold \(X\) can be \(C^{0}\) approximated by a sequence \(\{g_{n}\}\) of metrics with \(\mathrm{scal}_{g_{n}}<0\), and even \(\mathrm{Ricci}_{g_{n}}<0\). The following questions is set along the lines. Question: Is there a single theory of manifolds \(X\) with \(\mathrm{scal}(X)\geq \kappa\) or could there be several theories associated with different model spaces (homogeneous, symmetric) that serves as extreme for geometric and topological inequalities in spaces with \(\mathrm{scal}\geq \kappa\)? The author claims that all geometric and topological inequalities, whenever these are sharp, are accompanied by rigidity problems. Besides, one may ask for a description of \(X\) where such inequalities become equalities. In particular, the inequalities concerning (smoothed as well as cornered) manifolds \(X\) with \(\mathrm{scal}(X)\geq \kappa\). Though, it is not at all clear in what sense such \(X\) is stable as shown by the flat tori \(\mathbb{T}^{n}\) example, whose stability guaranteed by the Geroch's conjecture can be perturbed by blowing up huge bubbles with positive scalar curvatures confined into \(k\geq 3\)-codimension sets in \(X\). The article interprets the situations as follows: It may exists a particular weak metric \(d_{w}\) in the space of \(n\)-manifolds such that, for example, by giving a sequence of tori \(\{T_{g_{n}}\}\) with \(\mathrm{scal}_{g_{n}}\geq -\epsilon\) and \(d_{w}(T_{g_{n}},\mathbb{T}^{n})\to 0\), when properly normalized, there exists a subsequence converging to a flat torus \(\mathbb{T}^{n}\). The article calls attention upon the spin geometry and minimal hypersurfaces techniques used to obtain non-existence theorems. By quoting the author: ``The Dirac operator \(D\) and minimal hypersurfaces seem to belong to different worlds but they unexpectedly meet in spaces with lower bounds on their scalar curvatures. Is there a deeper link between Dirac and Minimal?''. The Dirac operator \(D\) relates to the geometry of \((X,g)\) by the Lichnerowicz formula \(D^{2}=\nabla^{2}+\mathrm{scal}_{g}/4\), while the application of minimal hypersurfaces depends on the second variation formula in the Schoen-Yau form making it similar to the Lichnerowicz formula. Quoting the author: ``Are there further formulas of this kind mediating between these two?'' The singular spaces considered are the closed subsets \(P\) (in a smooth \(n\)-manifold \(X\)) named cornered or curve-faced polyhedral \(n\)-domains of depth \(d=0,1,\dots,n\) whose boundary \(\partial P\) is decomposed into the union of a countable locally finite family of disconnected \((n-1)\)-faces \(Q^{n-1}_{i}\) with a distinguished set of adjacent pairs of faces \((Q_{i},Q_{j})\), such that {\parindent=6mm\begin{itemize}\item[(i)] \(Q_{i}\) is contained in a smooth hypersurface \(Y^{n-1}_{i}\subset X\), where \(Y_{i}\) is transversal to \(Y_{j}\) for all adjacent pairs of \((n-1)\)-faces \((Q_{i},Q_{j})\);\item[(ii)] for each \(i\), \(\partial Q_{i}=\bigcup_{j} Q_{i}\cap Q_{j}\) has a polyhedral \((n-1)\)-domain structures of depth \((d-1)\). \end{itemize}} The author defines a preconvex Riemannian manifold \(P\) with corners as \textit{mean curvature convex}, and writes \(\text{mn.curv}(\partial(P))\geq 0\), if all \((n-1)\)-faces \(Q_{i}\) have positive mean curvatures, i.e., if the variations of the \((n-1)\)-volumes of the faces \(Q_{i}\) are non-positive under infinitesimal inward deformations of \(Q_{i}\). Thus the following question is stated: Question: When does a Riemannian manifold \(X\), e.g., \(X=\mathbb{R}^{n}\), contain a mean curvature convex polyhedron \(P=P_{\leq \alpha_{ij}}\) of a given combinatorial type, where all its dihedral angles are bounded by \(\angle_{ij}(P)\leq \alpha_{ij}\) for given constants \(\alpha_{ij}\)? The author stresses that the answer is unknown even for \(X=\mathbb{R}^{3}\), the available answer being only for ``normal'' mean curvature convex domains \(P\subset\mathbb{R}^{3}\) that are combinatorially equivalent to prisms. By ``normal'' he means that the dihedral angles at the top and the bottom of these prisms are equal to \(\pi/2\). He claims that in this case \(\sum_{ij}(\pi-\alpha_{ij})\leq 2\pi\). Equality is achieved only for convex prisms \(P\subset\mathbb{R}^{3}\) with flat faces. In the article, the following Gauss-Bonnet prism inequality is quoted: Let \(P\) be a mean curvature convex 3-dimensional Riemannian manifold with corners that is combinatorially equivalent to a prism and has non-negative scalar curvatures, so \[ \sum_{ij=1}{k}(\pi-\alpha_{ij})\leq 2\pi. \] The mean curvature convexity concept seems to play a similar role as the scalar curvature does, e.g., it is the case for surgery. Let \(P\) be a strictly mean curvature convex polyhedral domain in a Riemannian \(n\)-manifold \(X\) with all dihedral angles of \(P\) bounded as \(\angle_{ij}(P)\leq \alpha_{ij}\). Let \(P'\) be obtained by surgery along some union \(B^{m-1}\subset P\) of \((m-1)\)-faces that themselves have no corner. By the same argument as in [\textit{M. Gromov} and \textit{H. B. Lawson jun.}, Ann. Math. (2) 111, 423--434 (1980; Zbl 0463.53025)], if \(n-m\geq 2\), then \(P'\) can be immersed in \(X\) with strictly positive mean curvature of all its faces and with all dihedral angles satisfying \(\angle_{ij}(P)\leq \alpha_{ij}\). So, the existence of \(P=P_{<\alpha_{ij}}\) is invariant under codimension-\(k\) surgery, \(k\geq 2\). The existence of \(P_{<\alpha_{ij}}\) is conjectured to be stable under smooth perturbations of the metric \(g\) in \(X\supset P\) that are \(\epsilon\)-small in \(C^{0}\)-topology. Definition: {\parindent=6mm\begin{itemize}\item[(i)] A mean curvature convex polyhedron \(P\subset X\), e.g., a convex polyhedron in \(\mathbb{R}^{n}\), is called dihedrally extremal if no \(C^{0}\) deformation on \(P\) can diminish its dihedral angles while keeping the faces mean curvature convex\item[(ii)] A polyhedral domain \(P\subset X\) is called a (poly) Plateau-hedron, or PP-hedron, if all its \((n-1)\)-faces have zero mean curvatures and the dihedral angle functions \(\angle_{ij}=\angle(Q_{i}\cap Q_{j})\) are constant on the edges that are \((n-2)\)-faces \(Q_{ij}=Q_{i}\cap Q_{j}\). \end{itemize}} Thus, dihedrally extremal mean curvature convex polyhedra \(P\) in Riemannian manifolds are PP-hedra. The fundamental questions concerning these objects is written as follows: Question: convexity \(\Leftrightarrow\) dihedrally extremal? {\parindent=6mm\begin{itemize}\item[(i)] Which convex polyhedra \(P\subset\mathbb{R}^{n}\) are dihedrally extremal? \item[(ii)] Are there non-convex dihedrally extremal \(P\subset\mathbb{R}^{n}\)? \end{itemize}} It is shown that convex Euclidean prisms are extremal for the prism inequality, moreover they are dihedrally rigid. Conjecture: The dihedral angles can not be diminished by deforming without developing negative mean curvature in some of the faces. Question: Can one prove the prism inequalities by a purely Dirac-type argument with no localization to minimal hypersurfaces? The article contains more ideas, examples, questions and conjectures about {\parindent=6mm \begin{itemize}\item[--] gluing around edges and dihedral extremity theorem,\item[--] billiards, pure edges and ramified coverings\item[--] \(\mu\)-area and \(\mu\)-bubbles \item[--] plateau traps\item[--] Reifenberg flatness\item[--] plateau \(m\)-web \end{itemize}}