On Schubert decompositions of quiver Grassmannians (Q391291)

Let \(Q\) be a quiver and \(Q_0\) be its set of vertices. Let \(M\) be a representation of \(Q\) having dimension vector \(\underline{m}=(m_p)_{p \in Q_0}\) and let \(\displaystyle{m=\sum_{p \in Q_0} m_p}\). The quiver Grassmannian \(\mathrm{Gr}_{\underline{e}}(M)\) of subrepresentations \(V\) of \(M\) with dimension vector \(\underline{e}\) is defined as a closed subscheme of the usual Grassmannian \(\mathrm{Gr}(e,m)\) where \(e = \sum_{p \in Q_0} e_p\). The intersection of \(\mathrm{Gr}_{\underline{e}}(M)\) with a Schubert decomposition of \(\mathrm{Gr}(e,m)\) defines a Schubert decomposition of \(\mathrm{Gr}_{\underline{e}}(M)\). In general, this is not a decomposition into affine spaces, and the isomorphism type of the Schubert cells is not independent of the choices that define the Schubert decomposition for \(\mathrm{Gr}(e,m)\). The results of this paper concentrate on establishing cases of quiver Grassmannians that have a Schubert decomposition into affine spaces. The main result roughly says the following: Let \(S \subset T\) be an inclusion of quivers such that the quotient \(T/S\) is a tree and let \(M\) be a representation of \(T\). Let \(F : T \rightarrow Q\) be a morphism of quivers that satisfies a certain Hypothesis (\(H\)). Then the Schubert cell \(C_{\beta}^{F*M}\) of the push-forward \(F*M\) of \(M\) equals the product \(A^n \times C_{\beta}^{F*M}\) of an affine space with the corresponding Schubert cell for the push-forward of the restriction \(M_S\) of \(M\) to \(S\). While Hypothesis (H) is too technical to explain in brevity, it should be mentioned that this hypothesis is a purely combinatorial condition on the structure of the fibres of \(F : T \rightarrow Q\), which can be checked easily in examples, and which can be implemented in a computer algorithm. The following are some of its consequences and other results of this paper. (i) Let \(M\) be an exceptional indecomposable representation of the Kronecker quiver and \(\underline{e}\) a dimension vector. Then \(\mathrm{Gr}_{\underline{e}}(M)\) has a Schubert decomposition into affine spaces. (ii) Let \(T\) be a tree and \(M\) a representation of \(T\) whose linear maps are block matrices of the form \(\begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix}\) where \(1\) is a square identity matrix. Then \(\mathrm{Gr}_{\underline{e}}(M)\) has a Schubert decomposition into affine spaces. If for all arrows \(\alpha\) of Q, the linear maps \(M_{\alpha}\) are isomorphisms, then the quiver Grassmannian decomposes into a series of fibre bundles whose fibres are usual Grassmannians. (iii) The author re-obtains the Schubert decompositions of \textit{G. Cerulli Irelli} and \textit{F. Esposito} [Algebra Number Theory 5, No. 6, 777--801 (2011; Zbl 1267.13043)] and \textit{G. Cerulli Irelli} et al. [Algebra Number Theory 6, No. 1, 165--194 (2012; Zbl 1282.14083)]. (iv) If \(\displaystyle{\mathrm{Gr}_{\underline{e}}(M, \mathbb{C}) = \amalg_{i \in I} X_i(\mathbb{C})}\) is a decomposition into complex affine spaces \(X_i(\mathbb{C})\), then the Euler characteristic of \(\mathrm{Gr}_{\underline{e}}(M)\) equals the number of non-empty Schubert cells \(C_{\beta}^M\) where \(\beta\) is of type \(\underline{e}\). (Proposition 6.3). If \(\mathrm{Gr}_{\underline{e}}(M)\) is smooth, then the singular cohomology is concentrated in even degrees and is generated by the closure of the classes of the Schubert cells (Corollary 6.2). In particular, this reproduces the formulas in [\textit{G. Cerulli Irelli}, J. Algebr. Comb. 33, No. 2, 259--276 (2011; Zbl 1243.16013)] and [\textit{N. Haupt}, Algebr. Represent. Theory 15, No. 4, 755--793 (2012; Zbl 1275.16016)] (under assumption of Hypothesis (\(H\))) in terms of the combinatorics of the Schubert cells. (v) If \(\mathrm{Gr}_{\underline{e}}(M, \mathbb{C}) = \amalg_{i \in I} X_i(\mathbb{C})\) is a regular decomposition into complex affine spaces, then the multiplication of \(H^*(\mathrm{Gr}_{\underline{e}}(M, \mathbb{C}))\) is determined by the cohomology rings of the irreducible components of \(\mathrm{Gr}_{\underline{e}}(M)\) (Lemma 6.5).