The paper studies a particular class of geometric objects called Calabi-Yau threefolds, which are six-dimensional spaces (in real dimensions) that play a central role in string theory and algebraic geometry. These spaces are constructed as hypersurfaces inside higher-dimensional toric varieties, which are geometric spaces built from combinatorial data encoded in polytopes, essentially higher-dimensional generalizations of polygons and polyhedra. The specific polytopes in question are "reflexive four-polytopes," and a major classification of them by Kreuzer and Skarke identified over 470 million examples, giving rise to a vast landscape of Calabi-Yau threefolds. A key question is whether a given Calabi-Yau threefold can be "smoothed," meaning whether a singular space can be continuously deformed into a nearby non-singular one.
The authors develop a concrete method for smoothing certain singular Calabi-Yau threefolds by carefully deforming the surrounding toric space. Their approach works face by face: for each two-dimensional face of the polytope whose associated geometric cone has singularities, they write down explicit polynomial equations (trinomials in a standard coordinate system) that locally resolve the problem. Two conditions must be met for a given face: the singularity of its cone must be smoothable in a known sense, and the corresponding dual edge of the polytope must contain a lattice point in its interior. When these conditions hold everywhere on the polytope, the entire Calabi-Yau hypersurface can be smoothed. The authors also construct a specific seven-vertex example that demonstrates the second condition is not just a technical convenience but is genuinely necessary.
Applying their method to the full Kreuzer-Skarke database, the authors find that their face-by-face criterion explains 3,774 of the 30,241 known smoothable cases in the Batyrev-Kreuzer census. The remaining roughly 26,000 smoothings require more global reasoning, specifically interactions between exceptional curve classes coming from different faces of the polytope, which their local method does not capture. This result both validates the approach and clearly maps out its limitations, pointing toward the additional geometric structure needed to understand the broader landscape of smoothable Calabi-Yau threefolds.