The projective dimension of three cubics is at most 5 (Q1634838)

Let \(R = k[x_1, \dots, x_N]\) be a polynomial ring over a field \(k\), and \(f_1, \ldots , f_n\) a minimal system of homogeneous generators of a homogeneous ideal \(I\). Possibly coming from the computational side, Stillman [\textit{I. Peeva} and \textit{M. Stillman}, J. Commut. Algebra 1, No. 1, 159--195 (2009; Zbl 1187.13010)] has asked whether one can find an upper bound for the projective dimension \(\mathrm{pd}(R/I)\) solely based on the number of minimal generators of \(I\) and their degrees. More precisely, is there a universal bound depending solely on \(n\) and the degrees of the given generators? The author's main result gives an affirmative answer in the case where \(n=3\) and \(\deg(f_i)=3\) throughout: \(\mathrm{pd}(R/I)\leq 5\). The general question has been answered by Ananyan and Hochster (also Erman, Sam and Snowden, and Draisma, Laso and Leykin), but the bounds are awfully large and often not entirely explicit. Sharp bounds were known for quadrics: if ht\((I)=2\) then \(\mathrm{pd}(R/I)\leq 2n-2\) (sharp) and for arbitrary height, \(\mathrm{pd}(R/I)\leq 4\) (sharp). The present proof is a case-by-case chase based on the multiplicity bound \(e(R/I)\leq 7\), noted by the authors and previously (partially) by \textit{B. Engheta} [J. Symb. Comput. 45, No. 1, 60--73 (2010; Zbl 1179.13009)]. The general strategy of proving the main result goes as follows: \(\bullet\) In several special situations, the bound on the \(3\) cubics stems from previous results by Engheta and previous papers by the authors (some in additional collaboration). An alternative has been to use linkage, or variables counting, or suitable short exact sequences. These methods and results are collected in Section 2 of the paper, a true facilitation for the interested reader. \(\bullet\) Next, they consider the possible primary decompositions of \(I^{\mathrm{un}}\) when \(e(R/I)\leq 4\), or of a suitable ideal \(L\) linked to \(I^{\mathrm{un}}\) if \(e(R/I)>4\). In each case they carefully go through the structure of \(I^{\mathrm{un}}\) and/or of a the obtained link \(L\) of \(I^{\mathrm{un}}\). From this follows fundamental structural information about a system of generators of \(I\). \(\bullet\) Finally, equipped with the information collected so far, they draw that either the \(I\) falls in one of the situations pointed out above, or, alternatively, the information on the structure of \(I\) is so explicit that by using short exact sequences or additional linkage process they obtain the desired bound. Throughout the paper is very clearly written and the strategy is nicely diversified. The stretched case-by-case analysis is probably inherent to getting tight bonds on the projective dimension. Of course, one would dream about some further results that not only bring out explicit sharp bounds but give some perception of why such bounds take place.