Some linear syzygy conjectures (Q1179964)

Let \({\mathcal M}\) be a graded module, generated by elements of degree 0, over the polynomial ring \(F[W]=F[W_ 1,\dots,W_ w]\) where \(F\) is an arbitrary field and \(W\) denotes the space of 1-forms. The linear syzygy conjectures concern the linear part of a graded minimal free resolution of \({\mathcal M}\), or in more abstract terms, the dimensions of the \(F\)- vector spaces \(\text{Tor}_ i^{F[W]}(F,{\mathcal M})_ i\) (the last index \(i\) denotes the \(i\)-th graded component). It is clear that the linear part of the resolution only depends on \(M={\mathcal M}_ 0\) and the multiplication map \(W\otimes M\to M_ 1={\mathcal M}_ 1\). Let \(m=\dim M\). The first section of the paper contains an algebraic proof of Green's vanishing theorem, which says that \(\text{Tor}_ i^{F[W]}(F,{\mathcal M})_ i=0\) for \(i\geq m\), provided \({\mathcal M}\) is a torsionfree module over \(F[W]/{\mathfrak p}\) for some absolutely irreducibe prime ideal \({\mathfrak p}\) not containing linear forms. --- The conjectures on linear syzygies are formulated in section 2. They extend and give quantitative sharpenings of Green's result. To state their main conjecture, the authors identify \(W\otimes M\) with \(\text{Hom}_ F(M^*,W)\) and consider the subvarieties \(R_ j\) of \(R=\text{Ker}(W\otimes M\to M_ 1)\) consisting of the elements of rank at most \(j\). They conjecture that if \(\text{Tor}_ k^{F[W]}(F,{\mathcal M})_ k\neq 0\), then (i) \(m\leq k\Rightarrow\dim R_ 1\geq k\), and (ii) \(m\geq k\Rightarrow\dim R_{m- k+1}\geq k\). As \(R_ 1=\{w\otimes m:wm=0\}\), this conjecture considerably strengthens Green's vanishing theorem. --- To understand the further (and stronger) conjectures one must consider \(\text{Tor}_ k^{F[W]}(F,{\mathcal M})_ k\) as being computed by the Koszul complex of \(W_ 1,\dots,W_ m\) with values in \({\mathcal M}\). The linear \(k\)-th syzygies then correspond to those elements of \(\bigwedge^ kW\otimes M\) which go to zero under the map \(\bigwedge^ kW\otimes M\to\bigwedge^{k-1}W\otimes M_ 1\) induced by the Koszul complex. An element of \(\bigwedge^ kW\otimes M\) may be interpreted as a linear map \(M^*\to\bigwedge^ kW\). Together with the operation of \(\bigwedge^{k-1}W^*\) on \(\bigwedge^ kW\) it induces a map \(\bigwedge^{k-1}W^*\to\text{Hom}_ F(M^*,W)\). Whereas the conjecture above makes an assertion about the image of \(\bigwedge^{k-1}W^*\) in \(\text{Hom}_ F(M^*,W)\) under a map coming from a linear \(k\)-th syzygy, the stronger conjectures require dimension inequalities for the images of certain subvarieties \(\Gamma_ s\) of the set of pure \((k-1)\)-vectors contained in \(\bigwedge^{k-1}W^*\). --- Belief in the truth of the conjectures is supported by the fact that the authors can indeed prove the desired dimension inequalities for the \(\Gamma_ s\) themselves. This proof is carried out in section 4, after some useful general results on linear syzygies have been derived in section 3. It uses Schubert calculus techniques. The remaining sections of the paper are devoted to extreme cases of the conjectures (and to a relatively easy `monomial' case). It is shown that the conjectures hold in the cases \(k=w\) and \(k=w-1\) where \(w=\dim W\), and also in the case \(k=2\). In particular the conjectures hold true whenever \(w\leq 4\).