Canonical graph decompositions and local separations: From infinite coverings to a finite combinatorial theory
Johannes Carmesin, Raphael W. Jacobs, Paul Knappe, Jan Kurkofka
TL;DR
This work develops a finite, combinatorial theory of r-local graph decompositions that faithfully encode local connectivity via r-local coverings. By introducing r-local separations, bottlenecks, and a displacement parameter Δ_r(G), the authors build canonical graph-decompositions H_r^{≤k}(G) that capture all r-local bottlenecks up to order k and are computable by a two-step algorithm. The theory harmonizes with existing local coverings, provides lifting/projection tools, and extends the Tree-of-Tangles paradigm to a local-bottlenecks setting, enabling a finite, algorithmic handle on graphs with complex local structure. In particular, the results unify local and global perspectives, recover trees-of-tangles in special cases, and offer a practical path to finite descriptions of otherwise infinite coverings, with potential applications to sparse networks and group-related graph structure.
Abstract
Every finite graph $G$ can be decomposed in a canonical way that displays its local connectivity-structure [DJKK22]. For this, they consider a suitable covering of $G$, an inherently infinite concept from Topology, and project its tangle-tree structure to $G$. We present a construction of these decompositions via a finite combinatorial theory of local separations, which we introduce here and which is compatible with the covering-perspective.
