On canonical singularities as total spaces of deformations (Q915801)

This paper gives criteria to recognize canonical singularities from their hyperplane sections. The first result gives that a Gorenstein singularity is terminal, if for some \(f:\;(X,p)\to ({\mathbb{C}},0)\) the singularity \(f^{-1}(0)\) is rational and \(f^{-1}(t)\) is terminal for \(t\neq 0\). This is a converse to Reid's theorem that the general hyperplane section through a Gorenstein canonical point has a rational or elliptic singularity [see \textit{M. Reid} in Journées de géométrie algébrique, Angers/France 1979, 273-310 (1980; Zbl 0451.14014)]. In the elliptic case X is canonical if \(X_ 0\) is isolated log-canonical and \(X_ t\) is canonical. The same method proves that the existence of an anticanonical divisor with rational singularities on a normal singularity X of \(index\quad m\) implies that X is canonical; moreover, if \(X\setminus \{p\}\) is terminal, then X is terminal. This answers a question of \textit{M. Reid} [Algebraic geometry, Proc. Summer Res. Inst., Brunswick/Maine 1985, part 1, Proc. Symp. Pure Math. 46, 345-414 (1987; Zbl 0634.14003); 6.5.2]. The paper concludes with some remarks on the relation between deformations of a singularity and its canonical cover.