Resolution of ideals associated to subspace arrangements (Q2166201): Difference between revisions

In the paper under review, the authors study resolutions of ideals that are associated with arrangements of subspaces. Recall that a subspace arrangement \(\mathcal{V}\) is a finite collection \(V_{1},\dots, V_{n}\) of vector subspaces of a given vector space \(V\) over an infinite field \(\mathbb{F}\). For each \(i \in \{1, \dots, n\}\) let \(\{f_{ij} \, : \, j \in \{1, \dots, d_{i}\}\}\) be an ordered \(\mathbb{F}\)-basis of \(V_{i}\). The arrangement of vectors \[ \{ f_{ij} \, : \, i \in \{1, .\dots, n\}, j \in \{1, \dots, d_{i}\}\} \] is a collection of bases of \(\mathcal{V}\). We made an assumption that for a given \(\mathcal{V} = \{V_{1}, \dots, V_{n}\}\) the collection of bases \(\{f_{ij}\}\) is generic in the sense that for all \(a=(a_{1}, \dots, a_{n})\in \mathbb{N}^{n}\) with \(a_{i}\leq d_{i}\) the dimension of \(W_{a}\) is the largest possible, where \[ W_{a} = \langle f_{ij} \, : \, i \in \{1, \dots, n\} \text{ and } j \in \{1, \dots, a_{i}\}\rangle, \] which clearly depends on the subspace arrangement but also on the collection of bases chosen. Notice that a collection of bases satisfying the above assumption always exists. Here the authors consider the product \(J\) of the ideals \(I_{i}\) generated by \(V_{i}\) in the polynomial ring \(S\), i.e., \(S\) is the symmetric algebra of \(V\). It turned out, as an example, that the Hilbert function of \(J\) is combinatorial invariant which means that it just depends on the rank function \[ \mathrm{rk}_{\mathcal{V}} : 2^{[n]} \rightarrow \mathbb{N}, \quad A \subset \{1, \dots, n\}, \quad \mathrm{rk}_{\mathcal{V}}(A) = \dim_{\mathbb{F}} \sum_{i \in A} V_{i}. \] Attached to the rank function we have a discrete polymatroid \[ P(\mathcal{V}) = \bigg\{ x \in \mathbb{N}^{n} \, : \, \sum_{i \in A} x_{i} \leq \mathrm{rk}_{\mathcal{V}}(A) \, : \text{ for all } A \subset \{1, \dots, n\}\bigg\}, \] and its subpolymatroid \[ P(\mathcal{V})^{*} = \bigg\{x \in \mathbb{N}^{n} \, : \, \sum_{i \in A} x_{i} \leq \mathrm{rk}_{\mathcal{V}}(A)-1 \, : \text{ for all } \emptyset \neq A \subset \{1, \dots, n\}\bigg\} \] whose ran function \(\mathrm{rk}_{\mathcal{V}}^{*}\) is obtained by the so-called Dilworth truncation, i.e., \[ \mathrm{rk}_{\mathcal{V}}^{*}(A) = \min\bigg\{ \sum_{i=1}^{p}\mathrm{rk}_{\mathcal{V}}(A_{i}) - p \, : \, A_{1}, \dots, A_{p} \text{ is a partition of } A \bigg\}. \] The authors show, among others, that the ideal \(J\) has linear quotients (which is much stronger than the property of having a linear resolution) and they provide an explicit formula for the Betti numbers and the projective dimension of \(J\). As it was showed, the Betti numbers \(\beta_{i}(J)\) of \(J\) are given by \[ \sum_{i\geq 0}\beta_{i}(J)z^{i} = \sum_{i \geq 0}\gamma_{i}(\mathcal{V})(1+z)^{i}, \] where \(\gamma_{i}(\mathcal{V}) = \#\{ x \in P(\mathcal{V})^{*} \, : |x|=i\}\), and the projective dimension of \(J\) is given by \[ \mathrm{proj.dim}\, J = \mathrm{rk}_{\mathcal{V}}^{*}(\{1, \dots, n\}) = \min\bigg\{\sum_{i=1}^{p} \mathrm{ rk}_{\mathcal{V}}(A_{i}) - p \, : \, A_{1}, \dots, A_{p} \text{ is a partition of } \{1, \dots, n\} \bigg\}. \] Furthermore, the authors show that the associated primes of \(J\) are exactly the ideals of the form \(\sum_{i \in A}I_{i}\), where \(A \subset \{1, \dots, n\}\) such that \(\mathrm{rk}_{\mathcal{V}}(A) < \mathrm{rk}_{\mathcal{V}}(B)\) for all \(A \subsetneq B\) and \(\mathrm{rk}_{\mathcal{V}}^{*}(A) = \mathrm{rk}_{\mathcal{V}}(A)-1\). Finally, the authors show that the minimal free resolution of \(J\) can be realized as a subcomplex of the tensor product of the truncated Koszul complexes associated with generic generators of \(V_{i}\)'s.