Homotopy $n$-types of cubical sets and graphs
Chris Kapulkin, Udit Mavinkurve
TL;DR
This work develops a robust framework for homotopy $n$-types in discrete settings by transporting two parallel model-structure constructions from simplicial sets to cubical sets and then applying them to graphs via the graph nerve. It shows that a right-transferred model structure along $\mathrm{cosk}_{n+1}$ yields a cubical $n$-type model whose fibrant objects align with vanishing higher homotopy, and establishes a Quillen equivalence with the canonical Cisinski $n$-type model. The paper then builds a fibration category on graphs with weak equivalences given by discrete $n$-equivalences and proves that for $n=1$ this recovers the classical groupoid model, via the fundamental groupoid functor. Overall, it offers a bridge between classical homotopy theory and discrete graph-homotopy, enabling transfer of techniques like Mayer–Vietoris and Blakers–Massey to combinatorial contexts.
Abstract
We give a new construction of the model structure on the category of simplicial sets for homotopy $n$-types, originally due to Elvira-Donazar and Hernandez-Paricio, using a right transfer along the coskeleton functor. We observe that an analogous model structure can be constructed on the category of cubical sets, and use it to equip the category of (simple) graphs with a fibration category structure whose weak equivalences are discrete $n$-equivalences.
