On the defining equations of hypersurface purely elliptic singularities (Q558626)

A purely elliptic singularity is a normal isolated singularity \((X,x)\) of a complex analytic space of dimension at least two, such that the \(l^2\)-plurigenus \(\delta _m (X,x)=1\) for all positive integer \(m\). These singularities, as well as some generalizations thereof (such as \({\mathbb Q}\)-Gorenstein singularities) have been studied by several mathematicians (S. Ishii, M. Tomari, K. Watanabe, among others). Purely elliptic singularities have many interesting properties and are related, perhaps surprisingly, to other mathematical objects. For instance, if \(\pi: {\widetilde X} \to X\) is a \textit{good} resolution of \(X\) (meaning that the exceptional divisor \(E\) has simple normal crossings), then it is possible to define a divisor \(E_J \subseteq E\), called the essential divisor, which plays an important role in the theory. Under suitable assumptions (e.g., the existence of a a certain partial desingularization, called the canonical modification of \((X,x)\)), it is proved that the irreducible components of \(E_J\) are birationally equivalent to Calabi-Yau varieties (i.e., projective varieties \(V\) with \(H_i(V, {\mathcal O}_V)=0\) for \(0 <i < n\), \(V\) might be singular, but at worst it has Gorenstein canonical singularities). This was originally proved in a rather abstract fashion. In the present paper the author finds a more explicit proof of this fact, in the case of a purely elliptic singularity of an \(r\)-dimensional hypersurface \(X\), defined by a non-degenerate polynomial \(f=f(x_0, \ldots, x_r)\). This means: if we consider the Newton polyhedron \( \Gamma _{+} (f) \subset {\mathbb R}^{r+1}\) of \(f\), then the partials of \(f\) do not simultaneously vanish on any compact face of the polyhedron. It was known that then there is a single compact face \(\gamma _{\mathbf 1}\) of \(\Gamma _{+} (f)\) containing the point \((1, \ldots, 1)\). Let \(d = \dim \, (\gamma _{\mathbf 1}) \). Kanesaka shows that each irreducible component of \(E_J\) (satisfying a certain condition) is birationally equivalent to the product of a Calabi-Yau variety \(Y\) of dimension \(d-1\) times \({\mathbb P}^{r-d}\). Moreover, the birational class of \(Y\) depends on the leading part of the polynomial \(f\) only, and not on the choice of the component. The leading part of \(f\) is the sum of the terms whose coefficients are in \(\gamma _{\mathbf 1}\). The techniques of the proof are primarily from toric and birational geometry (in the sense of Mori's program). An introductory section is devoted to background material, it contains many useful bibliographical references.