The existence of equivariant pure free resolutions (Q660438)

The paper under review was (when it first appeared, as an electronic preprint, in 2007 [\url{arXiv:0709.1529}]) the first step towards the proof of the conjecture of \textit{M. Boij} and \textit{J. Söderberg} [J. Lond. Math. Soc., II Ser. 78, No. 1, 85--106 (2008; Zbl 1189.13008)] concerning graded Betti numbers of Cohen-Macaulay modules over polynomial rings, a proof accomplished by \textit{D. Eisenbud} and \textit{F.-O. Schreyer} [J. Am. Math. Soc. 22, No. 3, 859--888 (2009; Zbl 1213.13032)]. The first part of this conjecture asserts that if \(A = K[x_1, \dots ,x_n]\) is a polynomial ring over a field \(K\) and if \(d_0 < d_1 < \dots < d_n\) is an increasing sequence of integers then there exists an exact complex of graded free \(A\)-modules of the following form: \[ 0 \longrightarrow A(-d_n)^{b_n} \longrightarrow \cdots \longrightarrow A(-d_0)^{b_0} \] for some integers \(b_0 > 0, \dots ,b_n > 0\) (actually, for any such complex, \(d_0, \dots ,d_n\) and \(b_0\) uniquely determine \(b_1, \dots ,b_n\)). In the paper under review, the authors construct, assuming \(\text{char}\, K = 0\), a \(\text{GL}(n)\)-equivariant complex with the above mentioned properties. In order to state their results, one has to recall some facts concerning the representation theory of the general linear group. A \textit{composition} is a sequence of non-negative integers \(\mu = (\mu_1, \dots ,\mu_n) \in {\mathbb N}^n\). If \(\mu_1 + \cdots + \mu_n = k\), one says that \(\mu\) is a composition of \(k\). A \textit{partition} is a composition \(\lambda = (\lambda_1 ,\dots ,\lambda_n)\) with \(\lambda_1 \geq \dots \geq \lambda_n\). Let \(E\) be an \(n\)-dimensional \(K\)-vector space. To each partition \(\lambda\) one can associate an irreducible representation \(S^\lambda E\) of the group \(\text{GL}(E) \simeq \text{GL}(n)\). \(S^\lambda E\) is a subrepresentation of \(S^{\lambda_1}E\otimes \cdots \otimes S^{\lambda_n}E\). Actually, there is a result asserting that if one denotes by \(c_{p,q}\) the composite map: \[ S^pE \otimes S^qE \longrightarrow S^pE \otimes E \otimes S^{q-1}E \longrightarrow S^{p+1}E \otimes S^{q-1}E \] then \(S^\lambda E\) is the intersection of the kernels of the maps \(\text{id}_{\, S^{\lambda_1}} \otimes \cdots \otimes c_{\lambda_i,\lambda_{i+1}} \otimes \cdots \otimes \text{id}_{\, S^{\lambda_n}}\). \textit{Pieri's Formula} asserts that \[ S^\lambda E \otimes S^eE \simeq \bigoplus_\mu S^{\lambda + \mu}E \] where the sum is taken over the compositions \(\mu\) of \(e\) such that \(\lambda_{i-1} \geq \lambda_i + \mu_i\), \(i = 2, \dots ,n\). Now, assuming \(d_0 \geq 0\), put \(e_0 = d_0\), \(e_i = d_i - d_{i-1}\), \(i = 1,\dots ,n\), and \(\lambda_i = e_0 + \sum_{j = i+1}^n (e_j - 1)\), \(i = 0,\dots ,n\). Consider the partitions \(\alpha(0) = (\lambda_1, \dots , \lambda_n)\) and \(\alpha(i) = (\lambda_1 + e_1 , \dots ,\lambda_i + e_i, \lambda_{i+1}, \dots ,\lambda_n)\), \(i = 1 ,\dots ,n\). Using Pieri's Formula one verifies easily that there exist unique non-zero \(\text{GL}(E)\)-equivariant maps \[ S^{\alpha(i)}E \longrightarrow S^{\alpha(i-1)}E \otimes S^{e_i}E,\;i\geq 1 \, , \] and that the composite maps \[ S^{\alpha(i+1)}E \rightarrow S^{\alpha(i)}E \otimes S^{e_{i+1}}E \rightarrow S^{\alpha(i-1)}E \otimes S^{e_i}E \otimes S^{e_{i+1}}E \rightarrow S^{\alpha(i-1)}E \otimes S^{e_i + e_{i+1}}E \] are zero. One constructs, in this way, a \(\text{GL}(E)\)-equivariant complex \(0 \rightarrow F_n \rightarrow \cdots \rightarrow F_0\) of graded free \(A\)-modules, with \(F_i = S^{\alpha(i)}E \otimes_KA(-d_i)\). Here, \(A\) is the symmetric algebra \(S(E) \simeq K[x_1, \dots, x_n]\). The main result of the paper under review asserts that this complex is exact. Despite the elementary character of the statement, the proof is quite involved. It uses geometric arguments based on a special case of Bott's theorem. The authors also provide a second construction of a pure complex with given degree shifts \(d_0 < \dots < d_n\), resolving a Cohen-Macaulay module supported on the degeneracy locus of a generic map of free modules \(G \rightarrow F\) of ranks \(n\) and \(m\), respectively, over the polynomial ring \(B = k[(x_{ij} \, | \, 1\leq i \leq m,1 \leq j \leq n)]\). This complex is \(\text{GL}(m) \times \text{GL}(n)\)-equivariant. The authors show, moreover, that both results hold in the more general context of complexes of graded free modules over rings of the form \(A\otimes_K\Lambda\), where \(A\) is a polynomial ring and \(\Lambda\) is an exterior algebra.