Generalized virtual polytopes and quasitoric manifolds (Q2099705)

The paper under review is devoted to generalized virtual polytopes based on the study of topological properties of affine subspace arrangements in real spaces. Let \(Q\) be a polynomial of degree at most \(k\) (homogeneous polynomial of degree \(k\)) on \(\mathbb{R}^{n}\), and denote by \(\omega = dx_{1} \wedge \dots \wedge dx_{n}\) the standard volume form on \(\mathbb{R}^{n}\). Let \(C_{s}\) be the cone of strictly convex bodies \(\triangle \subset \mathbb{R}^{n}\) with a smooth boundary. Then the function \[ F(\triangle) = \int_{\triangle}Q_{\omega} \] on the cone \(C_{s}\) is a polynomial of degree at most \(k+n\) (homogeneous polynomial of degree \(k+n\)). Now, to extend the domain of the integral functional to the entire vector space generated by the cone \(C_{s}\), the authors introduce the notion of a virtual convex body as a formal difference of two convex bodies with the identification \(\triangle_{1} - \triangle_{2} = \triangle_{3} - \triangle_{4} \iff \triangle_{1} + \triangle_{4} = \triangle_{2}+\triangle_{3}\). The main result provided by the authors, in the setting presented above, can be formulated as follows. Theorem A. Let \(M\) be the space of virtual convex bodies representable as a difference of convex bodies from the cone \(C_{s}\). Then the functional \(F\) on \(C_{s}\) can be extended as an integral of the form \(Q\omega\) over the chain of virtual convex bodies. Moreover, such an extension will be a polynomial on \(M\). Next, the authors study homological properties of unions \(X\) of (finite) arrangements of affine subspaces \(\{L_{i}\}\) in the real space \(L:=\mathbb{R}^{n}\) by means of the nerves \(K_{X}\) and their closed coverings by the sets \(L_{i}\). Consider a set \(\{L_{i}\}\) of affine subspaces in \(L\) indexed by elements \(i \in I\) and let \(X = \bigcup_{i \in I}L_{i}\) be their union. Notice that \(X\), as a topological space, has a natural covering by the affine subspaces \(L_{i}\). The nerve \(K_{X}\) of the natural covering of \(X\) is the simplicial complex with the vertex set indexed by \(I\) (one vertex for each index \(i \in I\)). A set of vertices \(v_{i_{1}}, \dots, v_{i_{k}}\) defines a simplex in \(K_{X}\) if and only if the intersection \(L_{i_{1}} \cap \dots \cap L_{i_{k}}\) is not empty. Consider another collection of affine subspaces \(\{M_{i}\}\) in a vector space \(M\) indexed by the same set \(I\), with a nerve \(K_{Y}\) corresponding to the natural covering of their union \(Y\). We say that the nerve \(K_{X}\) of the collection \(\{L_{i}\}\) dominates the nerve \(K_{Y}\) of the collection \(\{M_{i}\}\) if \[ \bigcap_{j \in J} L_{j} \neq \emptyset \quad \Rightarrow \quad \bigcap_{j \in J} M_{j} \neq \emptyset \,\, \text{ for all }J \subset I. \] We write in such a case that \(K_{X}\geq K_{Y}\). We say that a continuous map \(f : X \rightarrow Y\) is compatible with \(K_{X}\) and \(K_{Y}\) if \[ x \in L_{i_{1}} \cap \dots \cap L_{i_{k}} \Rightarrow f(x) \in M_{i_{1}} \cap \dots \cap M_{i_{k}}. \] In order to study homological properties of unions of affine subspace arrangements, the authors show the following result. Theorem B. \begin{itemize} \item[i)] If a map \(f : X \rightarrow Y\) compatible with \(K_{X}\) and \(K_{Y}\) exists, then the condition \(K_{X}\geq K_{Y}\) holds. \item[ii)] If a map \(f : X \rightarrow Y\) compatible with \(K_{X}\) and \(K_{Y}\) exists, then it is unique up to homotopy. \item[iii)] If \(K_{X}\) is isomorphic to \(K_{Y}\), then the map \(f : K_{X} \rightarrow K_{Y}\) compatible with \(K_{X}\) and \(K_{Y}\) provides a homotopy equivalence between \(X\) and \(Y\). \end{itemize} Recall that a good triangulation of \(X = \bigcup_{i\in I} L_{i}\) is a triangulation such that the following condition holds: the vertex set of a simplex \(S\) in a good triangulation is totally ordered, i.e., there is an order on the vertex set \(\{v_{i_{1}},\dots, v_{i_{s}}\}\) of \(S\) such that \(I(v_{i_{1}}) \subset\dots \subset I(v_{i_{s}})\), where for a point \(x \in X\) by \(I(x)\) we mean the subset of indices such that \(x \in L_{i}\) if and only if \(i \in I(x)\). Theorem C. For any finite union \(X = \bigcup_{i\in I} L_{i}\) of affine subspaces \(L_{i}\) in a linear space \(L\), one can construct a good triangulation of \(X\). Now we pass to topological properties. Suppose that we have an arrangement of affine hyperplanes \(\{H_{i}\}\) in \(L:=\mathbb{R}^{n}\). We call it non-degenerate if there is no proper linear subspace \(V \subset \mathbb{R}^{n}\) which is parallel to all the hyperplanes \(H_{i}\). Then the union \(X\) of such an arrangement has the homotopy type of a wedge of \((n-1)\)-dimensional spheres, in which the number of spheres is equal to the number of bounded regions in \(L\setminus X\). Therefore each cycle \(\Gamma \in H_{n-1}(X,\mathbb{Z})\) can be represented as a linear combination \(\Gamma = \sum_{j} \lambda_{j} \partial \triangle_{j}\), where each coefficient \(\lambda_{j}\) equals the winding number of the cycle \(\Gamma\) around a point \(a_{J} \in \triangle_{j} \setminus \partial \triangle_{j}\). Here by \(\triangle_{j}\) we denote the closure of a bounded open polyhedron representing a bounded component of \(L \setminus X\). We say that two hyperplane arrangements \(\mathcal{H}_{1}\) and \(\mathcal{H}_{2}\) are combinatorially equivalent if the corresponding nerves \(K_{\mathcal{H}_{1}}\) and \(K_{\mathcal{H}_{2}}\) are isomorphic. Theorem D. Suppose that \(\mathcal{H} = \{H_{1}, \dots, H_{s}\}\) and \(\mathcal{H}' = \{H_{1}^{'}, \dots, H_{s}^{'}\}\) are two combinatorially equivalent hyperplane arrangements, and let \(X = \bigcup_{i} H_{i}\) and \(Y = \bigcup_{i}H_{i}^{'}\) be the corresponding unions of hyperplanes. Then there exists a canonical homotopy equivalence \(f : X \rightarrow Y\). In order to study the homotopy type of a union of affine subspaces in \(\mathbb{R}^{n}\), one considers finite unions \(U \subset \mathbb{R}^{n}\) of open convex bodies \(U = \bigcup_{i} U_{i}\). Recall that a tail cone \(\mathrm{tail}(U)\) of a convex body \(U\) is the set of points \(v \in \mathbb{R}^{n}\) such that the inclusion \(a+tv \in U\) holds for any \(a \in U\) and \(t\geq 0\). If the set \(\mathrm{tail}(U_{i})\) is a linear space \(L_{i}\), then along with \(U_{i}\) one can consider a shifted space \(a_{i} + L_{i} \subset U_{i}\), where \(a_{i}\) is an arbitrary point in \(U_{i}\). Theorem E. Let \(U \subset \mathbb{R}^{n}\) be a finite union of open convex bodies. Then the set \(\mathbb{R}^{n} \setminus U\) is homotopy equivalent to the set \(\mathbb{R}^{n} \setminus \bigcup\{a_{i} + L_{i}\}\), where the union is taken over all indices \(i\) such that \(\mathrm{tail}(U_{i})\) is a vector space. Assume that all the linear spaces \(V_{i} = \mathrm{tail}(U_{i})\) above are equal to the same linear space \(V\) and denote by \(T\) a subspace transversal to \(V\), i.e., a linear subspace of \(\mathbb{R}^{n}\) such that \(\mathbb{R}^{n} = T \oplus V\). Then the set \(\mathbb{R}^{N} \setminus U\) is homotopy equivalent to \(T \setminus \{b_{i}\}\), where \(b_{i}:= T \cap \{a_{i} + V_{i}\}\). The statement describes completely the homotopy type of the set \(\mathbb{R}^{n} \setminus \bigcup H_{i}\), where \(\{H_{i}\}\) is any collection of affine hyperplanes in \(\mathbb{R}^{n}\), and this comes from the fact that the complement \(\mathbb{R}^{n}\setminus \bigcup H_{i}\) is a union of open convex sets. The second part of the paper is devoted to volumes of generalized virtual polytopes, and this is done to study the cohomology rings of generalized quasitoric manifolds. Suppose that \(\triangle\) is a triangulation of an \((n-1)\)-dimensional sphere on the vertex set \(V(\triangle) = \{v_{1}, \dots, v_{m}\}\). In what follows, we will identify a simplex of \(\triangle\) with the set of its vertices viewed as a subset in \(\{1, \dots, m\}\). A map \(\lambda : V(\triangle) \rightarrow (\mathbb{R}^{n})^{*}\) is called a characteristic map if for any vertices \(v_{i_{1}}, \dots, v_{i_{r}}\) belonging to the same simplex of \(\triangle\) the images \(\lambda(v_{i_{1}}), \dots, \lambda(v_{i_{r}})\) are linearly independent over \(\mathbb{R}\). Such a map defines an \(m\)-dimensional family of hyperplane arrangements \(\mathcal{AP}\) in the following way. For any \(h = (h_{1}, \dots, h_{m}) \in \mathbb{R}^{m}\), the arrangement \(\mathcal{AP}\) is given by \[ \mathcal{AP} = \{H_{1}, \dots, H_{m}\} \quad \text{ with } \quad H_{i} = \{\ell_{i}(x) = h_{i}\}, \] where \(\ell_{i}\) denotes the linear function \(\lambda(v_{i})\) for each \(i \in \{1, \dots, m\}\). Given a subset \(I \subset \{1, \dots,m\}\), we also define \(H_I = \bigcap_{j \in I} H_{j}\). If \(I \in \triangle\), then \(\Gamma_{I}\) denotes the face dual to \(I\) in the polyhedral complex \(\triangle^{\perp}\) dual to the simplicial complex \(\triangle\). A generalized virtual polytope is defined as a map \[ f \, : \triangle^{\perp} \rightarrow \bigcup_{\mathcal{AP}(h)}H_{i} \] subordinate to the characteristic map \(\lambda\), that is for any \(I \subset \{1, \dots, m\}\) one has \[ f(\Gamma_{I}) \subset H_{I}. \] Let \(U\) be a bounded region of \(\mathbb{R}^{n} \setminus \bigcup_{\mathcal{AP}(h)}H_{i}\) and \(W(U,f)\) be a winding number of \(f\). For a given polynomial \(Q\) on \(\mathbb{R}^{n}\), consider \[ I_{Q}(f) := \sum W(U,f) \int_{U}Q\omega . \] Theorem F. Let \(I \subset \{i_{1}, \dots, i_{r}\} \subset \{1, \dots, m\}\) be such that \(I \not\in \triangle\) and \(k_{1}, \dots, k_{r}\) be positive integers. Then we have \[ \partial_{i_{1}}^{k_{1}} \cdots \partial_{i_{r}}^{k_{r}}(I_{Q})(f) = 0. \] However, if \(r=n=\dim \triangle +1\) and \(I\) is a simplex in \(\triangle\) dual to a vertex \(A \in \triangle^{\perp}\), then we have \[ \partial_{I}(I_{Q})(f) = \mathrm{sgn}(I)Q(A)|\det(e_{i_{1}}, \dots, e_{i_{n}})|. \] In the last part of the paper, the authors describe the cohomology rings of quasitoric manifolds. Assume that \(K = K_{\Sigma}\) is a star-shaped sphere which is an intersection of a complete simplicial fan \(\Sigma\) in \(\mathbb{R}^{n} \simeq N \otimes_{\mathbb{Z}}\mathbb{R}\) with the unit sphere \(S^{n-1} \subset \mathbb{R}^{n}\). Let \(\mathcal{Z}_{K}\) be the moment-angle complex and denote by \(\Lambda : \Sigma(1) \rightarrow N\) a characteristic map. Then the \((m-n)\)-dimensional subtorus \(H_{\Lambda} := \ker \exp \lambda \subset (S^{1})^{m}\) acts freely on \(\mathcal{Z}_{K}\), and the smooth manifold \(X_{\Sigma,\Lambda} :=\mathcal{Z}_{K} / H_{\Lambda}\) is called a generalized quasitoric manifold. In that setting, the authors show the following results. Theorem G. Let \(X_{\Sigma, \Lambda}\) be a generalized quasitoric manifold. Then \(X_{\Sigma, \Lambda}\) has a cellular decomposition with only even-dimensional cells. The cells in this decomposition are in a bijection with the maximal cones \(\tau\) in \(\Sigma\), and the Euler characteristic of the manifold \(X_{\Sigma,\Lambda}\) is equal to the number of maximal cones in \(\Sigma\). As it is pointed out by the authors, in order to describe the cohomology ring of a generalized quasitoric manifold, it is enough to compute the self-intersection polynomial \[ h_{1}[D_{1}] + \dots + h_{m}[D_{m}] \mapsto \langle(h_{1}[D_{1}] + \dots + h_{m}[D_{m}])^{m}, [X_{\Sigma,\Lambda}]\rangle \] on the space of all linear combinations of classes of codimension \(2\) characteristic submanifolds. Theorem H. Let \(X_{\Sigma,\Lambda}\) be a generalized quasitoric manifold with codimension \(2\) characteristic submanifolds \(D_{1}, \dots, D_{m}\). Then the following identity holds: \[ \langle(h_{1}[D_{1}] + \dots + h_{m}[D_{m}])^{m}, [X_{\Sigma,\Lambda}]\rangle = n! \cdot \mathrm{Vol}(f_{h}), \] where \(f_{h} \in \mathcal{P}_{\Sigma,\Lambda}\) is a generalized virtual polytope associated with the simplicial complex \(K_{\Sigma}\), the characteristic map \(\Lambda\), and the set of parameters \(h = (h_{1}, \dots, h_{m})\). Theorem I. Let \(X_{\Sigma, \Lambda}\) be a generalized quasitoric manifold and let \(\mathcal{P}_{\Sigma,\Lambda}\) be the space of generalized virtual polytopes associated with it. Then the cohomology ring \(H^{\ast}(X_{\Sigma,\Lambda})\) can be computed as \[ H^{\ast}(X_{\Sigma, \Lambda}) = \mathrm{Diff}(\mathcal{P}_{\Sigma, \Lambda})/ \mathrm{Ann}(\mathrm{Vol}), \] where \(\mathrm{Diff}(\mathcal{P}_{\Sigma,\Lambda})\) is the ring of differential operators with constant coefficients on \(\mathcal{P}_{\Sigma,\Lambda}\) and \(\mathrm{Ann}(\mathrm{Vol})\) is the annihilator ideal of the volume polynomial.