Global smoothing of singular Fano and Calabi-Yau varieties
Anda Tenie
TL;DR
The paper develops global smoothing criteria for Fano and Calabi–Yau varieties with isolated rational lci singularities by linking local deformation directions to global deformations through the framework of 1-rational and 1-Du Bois singularities. It establishes that Fano and CY varieties can be deformed to stricter singularity types (1-rational or 1-Du Bois) or become smoothable under natural global conditions, including a Hodge–Du Bois numerical criterion in the presence of 1-liminal singularities. The results generalize classical threefold smoothing theorems (Namikawa–Steenbrink, Friedman, Gross, Friedman–Laza) to higher dimensions and to lci singularities beyond hypersurfaces, with unobstructedness statements $\mathbb{T}^i_Y=0$ for $i\ge 2$ in the Fano case and analogous CY unobstructedness. The work provides a unified deformation-theoretic approach to smoothing that leverages the Du Bois framework, Hodge theory, and local-to-global exact sequences, yielding new smoothing criteria and broadening the scope of known smoothability results.
Abstract
We study the problem of smoothing Fano and Calabi-Yau varieties with isolated rational lci singularities. For Fano varieties we show that any such $Y$ admits a deformation to a Fano variety with only $1$-rational singularities, and if all the singularities of $Y$ are not $1$-rational, then $Y$ is smoothable. For Calabi-Yau varieties, we show first that any such $Y$ deforms to a Calabi-Yau variety with only $1$-Du Bois singularities. Moreover, if all the singularities of $Y$ are not $1$-Du Bois then $Y$ is smoothable. When allowing $1$-liminal singularities, we give a global criterion in terms of the Hodge-Du Bois numbers of $Y$ which ensures $Y$ is smoothable. These theorems recover and generalize results for threefolds of Friedman, Namikawa, Namikawa-Steenbrink, Gross, and Friedman-Laza. In higher dimensions, our results provide alternative smoothing conditions and also extend the work of Friedman-Laza from the case of hypersurface singularities to lci singularities.
