The weight of the complement of
ProvedBunkbedAux.xc_weightIn the involution argument of Gladkov--Pak--Zimin, the configurations meeting cancel between the two connection events, and what remains is the complementary set : the configurations in which every hyperedge avoids . This theorem computes the per-hyperedge weight of that complement, which is the normalizing constant of the conditional measure on .
Write for the determinant-like quantity of Eq. [1], and let be a probability distribution on the five Wierman--Ziff states, so . Then the total weight of the legal refined states of a single hyperedge lying outside is
and the two variants alone contribute .
Where the two terms come from. Of the legal refined states, lie in , leaving . Two of these are the variants of the split pairs, each of weight , giving ; these are exactly the configurations in which the hyperedge is fully present on one level and fully absent on the other. The other nine are the pairs having the state in at least one coordinate: five of the form , five of the form , with the pair counted in both. Their total weight is
using that sums to . This is the only place the normalisation hypothesis is needed.
Mathematical role. The ratio is the conditional probability, given , that a single hyperedge is in a state; raised to the sixth power it is the probability of the event that all hyperedges are, and conditioned on the Wierman--Ziff model reduces to the alternative bunkbed hypergraph model, where Hollom's computation applies. The hypothesis Eq. [1], , enters precisely here: it forces to be small compared with , hence to have probability close to one, and the sign of the difference is then inherited from Hollom's versus .
import Definitions.Def_BunkbedWZRefined open Bunkbed.Hyper Finset
namespace BunkbedAux
/-- **The weight of the complement of `Λ²₊`.** Summed over the legal refined states of a single
hyperedge that do *not* lie in `Λ²₊`, the weight is the normalizing constant
`Z = 2W + 2 P_a|bc - P_a|bc ^ 2` of Gladkov-Pak-Zimin, where
`W = P_abc * P_a|b|c - P_ab|c * P_ac|b`; and the two `-` variants alone contribute `2W`. -/
theorem xc_weight (P : WZ → ℚ)
(hsum : P .abc + P .a_b_c + P .a_bc + P .ab_c + P .ac_b = 1) :
(∑ q ∈ Finset.univ.filter
(fun q : WZExt => WZExt.legal q = true ∧ WZExt.inLamSq q = false), WZExt.wt P q)
= 2 * (P .abc * P .a_b_c - P .ab_c * P .ac_b) + 2 * P .a_bc - P .a_bc * P .a_bc
∧ (WZExt.wt P (Sum.inr 1) + WZExt.wt P (Sum.inr 3))
= 2 * (P .abc * P .a_b_c - P .ab_c * P .ac_b) := by
sorry
end BunkbedAux