Local cohomology of bigraded Rees algebras and normal Hilbert coefficients (Q392515)

Let \((R,\mathfrak m,k)\) be an analytically unramified local ring of dimension \(d\) and let \(I\) be an \(\mathfrak m\)-primary ideal of \(R\). There exists a polynomial \(\overline{P}_I(x)\in \mathbb{Q}[x]\) of degree \(d\) such that \(\lambda(R/\overline{I^n})=\overline{P}_I(n)\) for all \(n\gg 0,\) where \(\lambda(M)\) denotes the length of an \(R\)-module \(M\). One can write \[ \overline{P}_I(n+1)=\overline{e}_0(I){n+d\choose d}-\overline{e}_1(I) {n+d-1\choose d-1}+\dots +(-1)^d\overline{e}_d(I) \] for some integers \(\overline{e}_i(I).\) Let \(J\) be another \(\mathfrak m\)-primary ideal of \(R\) and let \(\overline{\mathcal{R'}}(I,J)\) denote the extented Rees ring \(\bigoplus_{r,s\in \mathbb{Z}}\overline{I^rJ^s}t_1^rt_2^s\). Rees has proved that there exists a polynomial \(\overline{P}_{I,J}(x,y)\in \mathbb{Q}[x,y]\) of total degree \(d\) such that \(\lambda(R/\overline {I^rJ^s})=\overline{P}_{I,J}(r,s)\) for all \(r,s\gg 0.\) Again, one may write \[ \overline{P}_I(x,y)=\sum_{i+j\leq d} (-1)^{d-(i+j)}\overline{e}_{(i,j)}(I,J){x+i-1\choose i}{y+j-1\choose j} \] for some integers \(\overline{e}_{(i,j)}(I,J)\). From now on assume furthermore that \(R\) is Cohen-Macaulay, \(d=2\) and \(k\) is infinite. The main achievement of this paper is a generalization of one of the important theorems of Rees. It asserts that \(\overline{e}_2(IJ)= \overline{e}_2(I)+\overline{e}_2(J)\) if and only if \(\overline{I^{r+1}J^{s+1}}=a\overline{I^rJ^{s+1}}+ b\overline{I^{r+1}J^s}\) for all \(r,s\geq 0\), where \(a\in I\) and \(b\in J\) is a good joint reduction of \(I\) and \(J\). Using this, Rees deduced that the product of any two integrally closed ideals is integrally closed in 2-dimensional pseudo-rational local rings. He showed that regular local rings are pseudo-rational, and so his later result generalizes Zariski's product theorem. Let \(a\in I\) and \(b\in J\) be a good joint reduction of \(I\) and \(J\) and \(r_0,s_0\geq 0\) be two integers. As a generalization of Rees's Theorem, the authors prove that the following statements are equivalent: {\parindent=6mm \begin{itemize}\item[1)] \(\overline{e}_2(I)+\overline{e}_2(J)-\overline{e}_2(IJ)= \lambda(R/\overline{I^{r_0}J^{s_0}})-g_{r_0}(I,J)-h_{s_0}(I,J)-r_0s_0\overline{e}_{(1,1)}(I,J),\) \item[2)] \([H^2_{(at_1,bt_2)}(\overline{\mathcal{R'}}(I,J))]_{(r_0,s_0)}=0,\) \item[3)] \(\overline{I^{r+1}J^{s+1}}=a\overline{I^rJ^{s+1}}+b\overline{I^{r+1}J^s}\) for all \(r\geq r_0, s\geq s_0,\) \end{itemize}} where \(g_{r_0}(I,J), h_{s_0}(I,J)\) satisfy \(\lambda(\overline{J^s}/\overline{I^{r_0}J^s})=\overline{e}_ {(1,1)}(I,J)r_0s+g_{r_0}(I,J)\) for \(s\gg 0\) and \(\lambda(\overline{I^r}/\overline{I^rJ^{s_0}})=\overline{e}_{(1,1)}(I,J)rs_0+h_{s_0}(I,J)\) for \(r\gg 0\).