The graph of monomial ideals (Q2570675): Difference between revisions

The paper continues the authors' investigations [see \textit{K. Altmann} and \textit{J. A. Christophersen}, Manuscr. Math. 115, No. 3, 361--378 (2004; Zbl 1071.13008) and \textit{M. Haiman} and \textit{B. Sturmfels}, J. Algebr. Geom. 13, No. 4, 725--769 (2004; Zbl 1072.14007)] on the graph \({\mathcal G}\) of monomial ideals in the polynomial ring \(R=k[x_1,\dots,x_n]\), \(k\) a field. \({\mathcal G}\) is the infinite graph with the monomial ideals in \(R\) as vertex set. Two monomial ideals \(M_1\) and \(M_2\) are connected by an edge if there exists an ideal \(I\) in \(R\) such that the set of all initial monomial ideals of \(I\), with respect to all term orders, is precisely \(\{M_1,M_2\}\). \(I\) is called edge providing in this case. It is well known that \(I,M_1,M_2\) have many invariants in common. Each invariant yields a stratification of \({\mathcal G}\). A first proposition concerns the subgraph \({\mathcal G^r}\) obtained by restriction to artinian ideals of colength \(r\): Each such stratum is a connected component of \({\mathcal G}\). The main result (theorem 8) characterizes edge providing ideals as ``very homogeneous'': There exists upto multiples a single \(c\in\mathbb Z^n\) such that \(I\) is \(A\)-graded for \(A=\mathbb Z^n/c\mathbb Z\). This allows to define the Schubert scheme \(\Omega_c(M_1,M_2)=\Omega(M_1,M_2)\) of all \(A\)-homogeneous edge providing ideals connecting \(M_1\) and \(M_2\). A thorough analysis of the settings yields an algorithm that computes, for given \(M_1\) and \(M_2\), the direction \(c\) and \(\Omega(M_1,M_2)\) as affine scheme. More generally, the authors consider \(A\)-homogeneous ideals for arbitrary gradings \(\deg\:\mathbb Z^n\to A\) (including the standard one). Define \(h_I\:A\to \mathbb N\) as the Hilbert function of \(I\), i.e., \(h_I(a),\;a\in A\), is the \(k\)-dimension of the \(a\)-homogeneous part of \(I\). In the above situation, for \(I\in \Omega(M_1,M_2)\) the Hilbert functions of \(I\), \(M_1\) and \(M_2\) coincide. For positive gradings, i.e., \(\mathbb N^n\cap \text{ker}(\deg)=(0)\), this is also the general situation: \(\Omega_c(M_1,M_2)\not=\emptyset\) implies \(\deg(c)=0\) and hence the Schubert schemes describe an essential part of the multigraded Hilbert scheme \(\text{Hilb}_h\) of all \(A\)-homogeneous ideals with given Hilbert function \(h\). This part is sufficient to detect connectedness: Over \(k=\mathbb R\) or \(k=\mathbb C\), \(\text{Hilb}_h\) is connected if and only if the induced subgraph \({\mathcal G}(\text{Hilb}_h)\) is connected. Section 4 discusses properties of the Schubert schemes for square-free monomial ideals. The results are more technical and continue the investigations started by \textit{K. Altmann} and \textit{J. A. Christophersen} [loc. cit.]. In particular, it turns out that neighboring square-free ideals are connected by a generalization of the bistellar flip construction [see, e.g., \textit{O. Viro}, Proc. Workshop Differential Geometry Topology, Alghero 1992, World Scientific. 244--264 (1993; Zbl 0884.57015) or \textit{D. Maclagan} and \textit{R. R. Thomas}, Discrete Comput. Geom. 27, No. 2, 249--272 (2002; Zbl 1073.14503)]. The paper ends with a list of open problems about the graph \({\mathcal G}\) and the Schubert schemes.