Link Bundles of Compact Toric Varieties of Real Dimension 8

The proof can be found in [banagl2024link, Proposition 14]

The proof can be found in [banagl2024link, Proposition 19]

Using the stratification introduced in [banagl2024link], we obtain the following stratification on \mathcal{X}:

where

To show that the link bundle of \mathcal{X}, considered as a pseudomanifold, is trivial, we follow a similar line of reasoning as in [banagl2024link]. Accordingly, one can easily demonstrate that the link of a connected component of \mathcal{X}_{1} has the same CW structure as the link of an isolated singularity in a 6-dimensional toric variety. More precisely, this means that the underlying polytope of the link of a point x \in S^{1} \subset \mathcal{X}_{1} is 2-dimensional. To its interior, we attach a T^{3}, while to each of its 1-dimensional and 0-dimensional faces, we attach a T^{2} and an S^{1}, respectively. The collapsing data are derived from the cone corresponding to the isolated point in the real 8-dimensional toric variety for which \mathcal{X} serves as the link of this point. This allows us to easily determine whether a connected component of \mathcal{X}_{1} is, at least rationally, singular or regular. If f_{1} = 4 and f_{2} = 6, then the underlying polytope of the link is a tetrahedron. Hence, all points x \in \mathcal{X}_{1} are rationally regular, and the link \mathcal{X} satisfies Poincaré duality rationally. It follows that b_{1} = 0 and b_{2} = 0.

As mentioned in the proof of Corollary ?, \mathcal{X} is a pseudomanifold with the following stratification

Following the introduced construction of the links in [banagl2024link, Section 3], we can show that the link of the top stratum is a homological 3-sphere. Hence, the claim follows.

We mainly mimic the proof of [banagl2010intersection, Theorem 2.47]. Let (M, \partial M) be the \mathbb{Q}-manifold obtained by cutting out the n-c-dimensional singularities of X. Note that [banagl2024link, Theorem 47] implies that (M, \partial M) satisfies Lefschetz duality rationally, as mentioned. The topological space \partial M is a \mathbb{Q}-manifold by [banagl2024link, Proposition 50], and therefore, it rationally satisfies Poincaré duality using [banagl2024link, Corollary 49]. Hence, we can use [banagl2010intersection, Lemma 2.45] and [banagl2010intersection, Lemma 2.46] with rational coefficients. Consequently, we obtain a similar commutative diagram as in the proof of [banagl2010intersection, Theorem 2.47] with rational coefficients.

The link of a point x_{i} \in \mathcal{X}_{1}, \mathcal{L}_{i}, has the same structure as the link of an isolated singularity in a 6-dimensional toric variety. More specifically, this means the following. The underlying polytope of \mathcal{L}_{i}, \mathcal{M}_{i}, is a 2-dimensional polytope. We attach a torus T^3 to the interior of the polytope \mathcal{M}_{i}, and each 1-dimensional face and vertex of \mathcal{M}_{i} is associated with a torus T^2 and a circle S^1, respectively. Recall that we equipped each T^{2} with the following CW structure:

where \alpha is defined in Remark 16. From each T^{2}, we omit one 2-cell. Despite omitting the 2-cells from the tori, the 2-skeleton of the link has not changed. The rest of the proof proceeds along the same lines as in [banagl2024link, Proposition 67].

Let \pi: \mathcal{X} \longrightarrow \mathcal{M} be the natural projection. Note that \mathcal{X}_{1} = \bigsqcup_{i} S^{1}_{i}. Let \mathcal{L} be the link of S^{1} \subset \mathcal{X}_{1}, and let \mathcal{P} \subset \mathcal{M} be an embedding, as in the construction of links in [banagl2024link, Section 3]. In other words, we choose \mathcal{P} to be the intersection of a 2-dimensional plane with \mathcal{M}, ensuring that it intersects all the 1-dimensional faces of \mathcal{M} for which \pi(S^{1}) is a proper face. Then, we have \mathcal{L} \cong \sfrac{\pi^{-1}(\mathcal{P})}{S^{1}}. Hence, we obtain \mathcal{L} by attaching T^3, T^2, and S^1 to the 2-dimensional and 1-dimensional faces and vertices of \mathcal{P}, respectively. The attaching maps are the corresponding collapses derived from the cone associated with the isolated singularity in the 8-dimensional toric variety. Following [banagl2024link, Proposition 33], we can form the boundary operators and calculate the Betti numbers.

The space M is a \mathbb{Q}-manifold satisfying Lefschetz duality rationally. Thus, we have

From the CW structure of M it is clear that

because coning off the boundary, \partial M, only modifies the lower dimensional cells. Finally, we use \widetilde{H}_{j}(\sfrac{M}{ \partial M }) \cong \widetilde{H}_{j}(M, \partial M).

For the proof, we use the Mayer-Vietoris sequence. Let M be the \mathbb{Q}-manifold obtained by cutting out the singularities of \mathcal{X}. We set \mathcal{U} \simeq M. Let \mathcal{V} \cong \mathcal{C}(\sqcup_{i} S^{1} \times (\mathcal{L}_{i})_{<k} ), where i goes over the singular connected components of \mathcal{X}_{1}. Hence, we get \cap := \mathcal{U} \cap \mathcal{V} \simeq \sqcup_{i} S^{1} \times (\mathcal{L}_{i})_{<k}, and I^{\bar{n}}\mathcal{X} = \mathcal{U} \cup \mathcal{V}. Consequently, we have

Thus, we arrive at b_{1}^{I \mathcal{X}}=m-1. Since the singularities of \mathcal{X} are even-co-dimensional, we have I^{ \bar{n} }\mathcal{X}=I^{ \bar{m} }\mathcal{X}, and duality yields b_{6}^{I \mathcal{X}}= m-1. The more interesting part of the long exact sequence is the following:

Note that for a non-singular connected component of S^{1}_{i} \subset \mathcal{X}_{1}, we have f_{1}^{i}=3. Using the well-known Euler formula for 3-dimensional polytopes implies \sum_{i=1}^{m}(f_{1}^{i}-3)= \sum_{i=1}^{f_{2}-f_{1}+2}(f_{1}^{i}-3). Counting the 1-dimensional faces of the underlying polytope of the link of S^{1}_{i}, \mathcal{P}_{i}, for each connected component S^{1}_{i} equals counting the neighboring 1-dimensional faces of each vertex in the underlying polytope of \mathcal{X}, \mathcal{M}. Since each 1-dimensional face of the polytope \mathcal{M} has precisely two vertices in its topological boundary, we have \sum_{i=1}^{f_{2}-f_{1}+2}f_{1}^{i}=2 f_{2}. Thus, we get \sum_{i=1}^{f_{2}-f_{1}+2}(f_{1}^{i}-3)=2f_{2}-3(2-f_{1}+f_{2})=3 f_{1}-f_{2} -6. Since \mathcal{L} is rationally 3-segmented, we have

The definition yields H_{i}(M)\cong H_{i}(\mathcal{U}). Consider the above Mayer-Vietoris exact sequence and the exact sequence of the relative homology of the pair ( M ,\partial M) in the following diagram:

The commutativity of the diagram

ensures that the outer left and right squares in Diagram ? commute. From [banagl2010intersection, Lemma 2.46], it follows that there is a homomorphism H_{3}(I^{ \bar{n} }\mathcal{X}) \longrightarrow H_{3}(M, \partial M) that makes Diagram ? commutative. Using the 5-lemma yields b_{3}^{I\mathcal{X}}=b_{3}^{( M ,\partial M)}=b_{4}^{M}, whereas, in the last equality, we used Lefschetz duality. By using the duality of Betti numbers of intersection spaces and Lemma 29, we get b_{4}^{M}=b_{3}^{I\mathcal{X}}=b_{4}^{I\mathcal{X}}=3f_{1}-f_{2}-6-(b_{1}+b_{2}). Employing the exactness at H_{3}(\cap) \longrightarrow H_{3}(M) and considering that 0 \leq b_{1}, it immediately follows that b_{1}=0, where b_{1} is defined in Equation (?). Finally, we get b_{2}^{I\mathcal{X}}=f_{2}-2-b_{2}.

We consider the long exact sequence of the relative homology of (M, \partial M). Using rational Lefschetz duality and Equation (Proof), we obtain

By employing the exactness of the above sequence, we arrive at b_{2} = \operatorname{rk}(\operatorname{ker}(H_{3}(\partial M) \longrightarrow H_{3} (M))).

In the proof of Theorem 30, we showed that b_{1}=0. Consequently, the claim follows from Proposition 17 and Lemma 31.

First, we prove the statement for a point in the bottom stratum of X_{\Sigma}. We use the same notation as in the above discussion. Using the short exact sequence ? and [cox2024toric, Theorem 3.3.19], we conclude that X_{\mathring{\Sigma}_{\nu}} is a fiber bundle over X_{\Sigma^{\prime}_{\nu} } with fiber X_{\Sigma_{0}}, where \mathring{\Sigma}_{\nu} \subseteq N, \Sigma^{\prime}_{\nu} \subset N^{\prime}, and \Sigma_{0}\subseteq N_{0}. Thus, we arrive at the following fiber bundle:

where the maps \operatorname{id}_{\mathbb{R}} and \operatorname{pr} denote the identity map and the projection to a point, respectively. It follows that

is also a fiber bundle. This completes the proof for the case where x lies in the bottom stratum of X_{\Sigma}. For an arbitrary point x, let \Sigma_{x} \in \Sigma be the corresponding cone to the orbit in which x lies. Consider \Sigma_{x} as a cone in the sublattice N := \operatorname{span}(\Sigma_{x}) \cap \mathbb{Z}^{\dim(\Sigma) } \cong \mathbb{Z}^{\dim ( \Sigma_{x} ) }. Hence, we arrive at X_{\Sigma_{x} } \cong \mathring{\mathcal{C}} (\mathcal{L}_{x}), where \Sigma_{x} \subset N. Let \mathring{\Sigma}_{x} \subset N be the fan obtained from \Sigma_{x} by removing its interior. It follows that X_{\mathring{\Sigma}_{x}} \cong \mathcal{L}_{x} \times \mathbb{R}. The fan \mathring{\Sigma}_{x} splits by \Sigma_{0} = \{ 0\} and \Sigma^{\prime}_{x} := \pi^{\prime} (\mathring{\Sigma}_{x}). The map \pi^{\prime} : N \longrightarrow ( \sfrac{N_{\mathbb{R}}}{ (N_{0})_{\mathbb{R}} } ) \cap N is defined where N_0 \subset N is the lattice spanned by a 1-dimensional cone lying in the interior of \Sigma_{x}. The rest of the proof follows the same lines as in the case of the points in the bottom stratum.

We use the long exact sequence of homotopy groups induced by S^{1} \hookrightarrow \mathcal{L}_{x} \rightarrow X_{\Sigma^{\prime}}.

We begin by studying the aforementioned long exact sequence of homology groups. Since \dim(\mathcal{L}_{x})=7, we obtain

Hence, we arrive at the following exact sequences.

The exact sequences above yield the following equalities between Betti numbers.

where we used b_{6}^{\mathcal{L}_{x}}=b_{1}^{\mathcal{L}_{x}}=0 and b_{6}^{X_{\Sigma}}=b_{0}^{X_{\Sigma}}=1. Note that from the construction of the projection map \pi^{\prime}, defined in the proof of Theorem 36, the numbers of 1-dimensional and 2-dimensional faces of the underlying polytopes of \mathcal{L}_{x} and X_{\Sigma} are equal. Hence, the numbers f_{1}^{\mathcal{L}_{x}} and f_{2}^{\mathcal{L}_{x}}, defined in Theorem 32, equal the numbers f_{1}^{X_{\Sigma}} and f_{2}^{X_{\Sigma}}, which are, respectively, the numbers of 1-dimensional and 2-dimensional cones in the underlying fan of X_{\Sigma}. Therefore, by using the explicit form of the Betti numbers of X_{\Sigma} provided in [mcconnell1989rational] or [banagl2024link], the claim follows.

The homomorphisms \Psi^{\prime}:H_{p}(B) \longrightarrow H_{p-d-1}(B) associated to an S^{d}-bundle have been studied in [chern1950homology]. For the S^{1}-bundle \pi:\mathcal{L}_{x} \longrightarrow X_{\Sigma}, the claim follows from the exactness of sequence

More precisely, one obtains \operatorname{rk}\big(\operatorname{ker}(H_{4}(X_{\Sigma}) \longrightarrow H_{2}(X_{\Sigma}))\big)=b_{4}^{\mathcal{L}_{x}}-b_{3}^{X_{\Sigma}}. On the other hand, b_{4}^{\mathcal{L}_{x} }=3 f_{1}^{\mathcal{L}_{x} }-f_{2}^{\mathcal{L}_{x}}-6 and b_{3}^{X_{\Sigma} }=3 f_{1}^{X_{\Sigma} }-f_{2}^{X_{\Sigma}}-6-b. The numbers f_{1} and f_{2} denote the number of 2-dimensional and 1-dimensional faces of the underlying polytopes of the corresponding space. As noted in the proof of Theorem 38, f_{1}^{\mathcal{L}_{x} }=f_{1}^{X_{\Sigma} } and f_{2}^{\mathcal{L}_{x} }=f_{2}^{X_{\Sigma} }. The claim then follows.