The five partition probabilities of a triple exhaust the measure
ProvedBunkbedAux.partition_sum_eq_totalFix a finite edge set , a weight function , and three vertices . In the percolation, the connected components induce a partition of , and there are exactly five possibilities: all three together, one of the three pairs together with the third vertex apart, or all three separate. Writing for the probabilities of these five events,
The content is that the five events are mutually exclusive and exhaustive. This holds because reachability in any fixed open subgraph is an equivalence relation: the triple of truth values of can therefore only be , , , or , and these five patterns correspond bijectively to the five quantities. The three patterns ruled out by transitivity — those with exactly two of the relations true — are precisely what makes the count five rather than eight.
Combined with the fact that the total mass is , this gives the relation used to eliminate from any expression in the five, which is how the source converts its determinant-shaped quantity into connection probabilities.
Formalization Note The right-hand side is left as the total mass rather than , so that this lemma is independent of the normalisation lemma and the two can be combined as needed. The five definitions in BunkbedPercolation are phrased asymmetrically — for instance negates while also negates — and they agree with the symmetric descriptions only via transitivity, which the proof must supply.
import Definitions.Def_BunkbedPercolation open Bunkbed Finset
namespace BunkbedAux
/-- The five partition probabilities of a triple exhaust the measure. -/
theorem partition_sum_eq_total {V : Type*} [Fintype V] [DecidableEq V] (E : Finset (Sym2 V))
(w : Sym2 V → ℚ) (a b c : V) :
Pabc E w a b c + Pab_c E w a b c + Pac_b E w a b c + Pa_bc E w a b c + Pa_b_c E w a b c
= ∑ S ∈ E.powerset, weight E w S := by sorry
end BunkbedAux