Bunkbed percolation factorises over edge-disjoint parts
ProvedBunkbedAux.bbProb_biUnion_disjointIn bunkbed percolation the two levels are percolated independently on the same edge set, so is a double sum over pairs of configurations. If that edge set is a disjoint union , then each of the two configurations further factorises across the parts.
This theorem records the resulting fourfold-indexed form: for a pairwise disjoint family and any weight function ,
both sums running over tuples with .
No hypothesis on is required: this is an algebraic rearrangement, obtained by applying the single-level factorisation once in each summation variable. The underlying facts are that configurations on a disjoint union correspond to tuples of configurations on the parts, and that the configuration weight factorises as a product of per-part weights.
Mathematical role. This is what makes a substituted graph tractable. If a graph is built by replacing each hyperedge of a hypergraph with a gadget copy, the copies are edge-disjoint, so this theorem presents bunkbed percolation on the whole graph as an independent choice of a configuration in each copy on each level. Combined with the fact that connectivity depends only on the partition each copy induces on its attachment vertices, it identifies the process with a Wierman--Ziff model on the hypergraph: one independent pair of states per hyperedge, one for each level.
import Definitions.Def_BunkbedPercolation open Bunkbed Finset SimpleGraph
namespace BunkbedAux
/-- **Bunkbed percolation factorises over edge-disjoint parts.** Both levels are percolated
independently, so for a pairwise disjoint family of edge sets the double sum defining `bbProb`
becomes a double sum over tuples of per-part configurations. -/
theorem bbProb_biUnion_disjoint {V : Type*} [Fintype V] [DecidableEq V]
{ι : Type*} [Fintype ι] [DecidableEq ι]
(E : ι → Finset (Sym2 V)) (hdisj : ∀ i j, i ≠ j → Disjoint (E i) (E j))
(w : Sym2 V → ℚ) (T : Finset V) (x y : V × Fin 2) :
bbProb (Finset.univ.biUnion E) w T x y
= ∑ S₀ ∈ Fintype.piFinset (fun i => (E i).powerset),
∑ S₁ ∈ Fintype.piFinset (fun i => (E i).powerset),
(if (bbGraph T (Finset.univ.biUnion (fun i => S₀ i))
(Finset.univ.biUnion (fun i => S₁ i))).Reachable x y then (1 : ℚ) else 0)
* (∏ i, weight (E i) w (S₀ i)) * (∏ i, weight (E i) w (S₁ i)) := by
sorry
end BunkbedAux