Numerical Gorenstein elliptic singularities (Q1769030)

Let \((X,p)\) be a normal complex surface singularity and \(\pi:(M,A)\rightarrow (X,p),\;\pi^{-1}(p)=A\) be a minimal resolution. It is assumed that \((X,p)\) is numerical Gorenstein, i.e., the canonical divisor in \(M\) is numerically equivalent to a cycle supported in \(A\). For any cycle \(B\) in \(A\) let \(Z_B\) be the minimal positive cycle supported in Supp\((B)\) such that \(-Z_B\) is nef on \(B\), i.e., \(-Z_B\cdot C\geq 0\) for all irreducible components \(C\) of \(B\). Let \(C_t:=\sum_{i=0}^t Z_{B_{i}}\) and \((X_j,p_j)\) be the singularity obtained by contracting the support of \(Z_{B_j}\), where \(B_0,B_1,\dots, B_m\) is the Yau sequence as in [\textit{A. Némethi}, Invent. Math. 137, No.~1, 145--167 (1999; Zbl 0934.32018)]. Let also \( {\mathcal A}_f=\{j\mid 0\leq j\leq m-1, H^0({\mathcal O}_M(-C_{j+1}))\subsetneqq H^0({\mathcal O}_M(-C_j))\}\cup\{m\}\), \({\mathcal A}_g=\{j\mid 0\leq j\leq m, X_j \text{ is Gorenstein }\} \), \(\alpha=\min {\mathcal A}_g\) and \(\beta=\min{\mathcal A}_f\). The main theorem of this article is the following: Let \(d=-C_\beta^2\). Then one of the following cases hold: 1) \(-Z_{B_\beta}^2\geq 2\), or \(-Z_{B_\beta}^2=1\) and \(\alpha<\beta\). In this case \({\mathcal O}_M(-C_\beta)\) is free and \(\text{ mult}(X,p)=d\geq 2\). If \((X,p)\) is Gorenstein then \(\text{ embdim}(X,p)=\max\{d,3\}\). 2) \(-Z_{B_\beta}^2=1\), \(\alpha=\beta\) and \(C_\beta\) is the maximal ideal cycle; \(\text{ mult}(X,p)=\text{embdim}(X,p)-1=d+1\). 3) \(-Z_{B_\beta}^2=1\), \(\alpha=\beta>0\) and \(C_\beta\) is not the maximal ideal cycle; \(d\geq 2\) and \(\text{mult}(X,p)=\text{embdim}(X,p)-1=d+2\). The author also proves the following statement: For \(j\geq \alpha\), \(X_j\) is a \(\mathbb Q\)-Gorenstein singularity of index \(\gamma/(j-\alpha,\gamma)\). Thus \(\gamma | (m-\alpha)\) and \({\mathcal A}_g=\{\alpha+i\gamma\mid 0\leq i\leq (m-\alpha)/\gamma\}\). Since \(p_g=\#{\mathcal A}_g\), it follows that \(p_g=(m-\alpha)/\gamma+1\). We also have \(0\leq\beta-\alpha<\gamma\). If \((X,p)\) is Gorenstein, then \(\beta=\gamma-1\). The above theorems generalize the results of the above cited article to Gorenstein singularities.