Betti numbers of determinantal ideals (Q2470364): Difference between revisions

Let \(R = k[x_1,\dots,x_n]\) be the polynomial ring over a field \(k\), with \(\deg x_i = 1\) for all \(i\). Let \(I\) be a homogeneous ideal of arbitrary codimension. Consider the minimal free resolution of \(R/I\), writing the \(i\)th free module as \(\bigoplus_{j \in \mathbb Z} R(-j)^{\beta_{i,j}}\), where \(\beta_{i,j} = \dim \text{Tor}_i^R(R/I,k)_j\). We define \(m_i = \min \{ j \in \mathbb Z \mid\beta_{i,j} \neq 0 \}\) and \(M_i = \max \{ j \in \mathbb Z \mid\beta_{i,j} \neq 0 \}\); these are the smallest and the largest twists (in absolute value), respectively, that appear in the \(i\)th free module in the minimal free resolution. Let \(c\) be the codimension of \(R/I\). The Multiplicity Conjecture of Huneke, Herzog and Srinivasan stated that if \(R/I\) is Cohen-Macaulay then the multiplicity \(e(R/I)\) should be bounded by \[ \frac{1}{c!} \prod_{i=1}^c m_i \leq e(R/I) \leq \frac{1}{c!} \prod_{i=1}^c M_i. \] Furthermore, the upper bound should hold even if \(R/I\) is not Cohen-Macaulay. This conjecture has recently been solved in a series of papers by Boij and Söderberg, Eisenbud, Fløystad and Weyman, Eisenbud and Schreyer, and again Boij and Söderberg. The present paper, however, deals instead with a related question due to Römer for Cohen-Macaulay algebras. Denoting by \(\beta_i = \sum_{j \in \mathbb Z} \beta_{i,j}\), the goal is to bound any \(\beta_i\) from above by a function of \(i\) involving the various \(M_j\), and from below by a function of \(i\) involving the various \(m_j\). Römer suggested specific functions, and showed that they hold in the strictly quasi-pure case, as well as the codimension 2 and Gorenstein codimension 3 cases. In the current paper this work is extended by proving that Römer's bounds also hold for certain kinds of determinantal ideals, as well as for arithmetically Cohen-Macaulay divisors on a variety of minimal degree. The author also discusses when these bounds are sharp, which primarily comes when the resolution is pure. She combines all these results into the formal conjecture that for \(R/I\) Cohen-Macaulay of codimension \(c\), we have \[ \prod_{1 \leq j < i} \frac{m_j}{M_i - m_j} \prod_{i < j \leq c} \frac{m_j}{M_j - m_i} \leq \beta_i \leq \frac{1}{(i-1)!(c-i)!} \prod_{ j\neq i} M_j. \] She points out that while this work is inspired by the Multiplicity Conjecture, it is not at all clear if either conjecture implies the other.