Reconstruction Conjecture — Minimal separators are cliques #
The capstone of the chordal structure layer: in a chordal graph, an
inclusion-minimal separator is a clique. The proof contradicts chordality by
exhibiting an induced cycle of length ≥ 4: take two non-adjacent vertices
x, y of the separator S; by minimality each has a neighbour in two distinct
components A, B of G − S; the two shortest x–y paths confined to
A ∪ {x,y} and B ∪ {x,y} are chordless arcs sharing only x, y, with no
cross-edges, so they glue (inducedCycleEmbedding_of_paths) into an induced
cycle — impossible in a chordal graph.
Main results #
SimpleGraph.exists_chordless_arc— a chordlessx–yarc through a prescribed component (a geodesic in the induced subgraph on that component plusx, y).
A chordless arc through a component. Given non-adjacent x ≠ y, a
component of G − S represented by u₀ ∉ S, and neighbours of x and of y
in that component, there is a chordless x–y path P of length ≥ 2 whose
interior lies entirely in that component. It is a shortest path in the induced
subgraph on the component together with x, y; being a geodesic makes it
chordless (geodesic_not_adj_of_lt), and the induced subgraph confines its
interior to the component.
Minimal separators are cliques (Dirac's lemma — the chordal-structure
crux). In a chordal graph, an inclusion-minimal separator induces a clique:
two non-adjacent separator vertices would, by minimality, have neighbours in two
distinct components of G − S, and the two shortest paths confined to those
components glue into an induced cycle of length ≥ 4, impossible in a chordal
graph.