Alphabeta Math
TheoremStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck passverified 2026-10-22 (gpt-6.1-sol)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Periodic-torus RSW lower bounds at the FK self-dual point

Statement

For each q≥1 and aspect ratio α>1, there is a constant c(α,q)>0 such that the following holds. Let n,m be positive integers with m>⌈αn⌉, let Tm=(Z/mZ)2 with its nearest-neighbor horizontal and vertical edges and periodic boundary conditions, and give its edges the FK law with no boundary wiring, parameter q, and psd(q)=q/(1+q). The probability of a horizontal open crossing of the vertex rectangle {0,…,⌈αn⌉}×{0,…,n} is at least c(α,q), uniformly in n and m.

Facts & Assumptions

Given: Real q≥1, α>1, positive integers n,m with m>⌈αn⌉, and the finite periodic square-torus FK law at p=q/(1+q).

[F1]

Finite FK laws on labelled multigraphs use the stated component weights (Finite FK random-cluster weights for real q and boundary partitions).

[F2]

Finite FK has positive association, exact spatial Markov conditioning, and conditional edge-open probability at least ρ=p/[p+q(1−p)] (Finite FK association, spatial Markov property, and boundary monotonicity).

[F3]

The dual-complement law on every labelled square torus has likelihood bounded by q−2 and q2 relative to the ordinary dual FK law, including thin quotients (Cellular component counts and self-dual FK likelihoods on a finite square torus).

[F4]

In the rotated lattice P={(u,v)∈Z2:u≡v(mod2)} with diagonal edges, use diagonal torus periods (m,m) and (−m,m). For N∈4Z>0 the closed square has each crossing probability at least a=ρ/(1+q2) when N<m, and the closed width-3N/2, height-N rectangle has horizontal crossing probability at least b=a3/[2(1+q2)] when 3N/2<m (Finite mixed-boundary reflection and gluing on the rotated FK torus). These are actual-edge crossing events with the literal closed-side conventions.

[F5]

A simple polygonal Jordan curve separates the plane into two components (Polygonal Jordan curve theorem: a polygon has exactly two complementary regions and is the frontier of each). Consequently opposite-direction paths in any of the finite planar rotated rectangles intersect: a loop-erased horizontal path, completed through exterior side collars, separates the bottom from the top. All primal diagonal edges meet only at common vertices.

Proof

1.1F1F3F4givenconstructalgebra

The invertible linear map u=x+y, v=y−x sends the ordinary square lattice to P and its nearest-neighbor edges to the diagonal edges. Its inverse is x=(u−v)/2, y=(u+v)/2. Ordinary periods (m,0),(0,m) become (m,−m),(m,m), generating exactly the diagonal-period lattice in [F4]. It is therefore a labelled graph isomorphism, preserving the weights in [F1]. A rotated patch with both coordinate spans less than m embeds: every nonzero period has at least one coordinate of absolute value at least m. We may compute probabilities of such patches in these coordinates. The cellular torus comparison [F3] is the likelihood input used in [F4], not an unproved planar substitution for torus topology.

1.2F2F4F5constructalgebra

A closed rotated vertical rectangle of width S∈4Z>0 and height 2S has crossing probability at least d=b2a if 2S<m. Indeed take the vertical crossings of [0,S]×[0,3S/2] and [0,S]×[S/2,2S], and a horizontal crossing of their overlap square [0,S]×[S/2,3S/2]. The vertical bounds are [F4] after coordinate interchange; the shift S/2 is even and preserves P. The square bound is also [F4]. Each horizontal path intersects both vertical paths by [F5], so their union gives the required vertical crossing. FKG [F2] gives b2a. Translation by any integer vector with even coordinate sum preserves the lattice, the periodic FK law, and this estimate. Restricting a taller crossing at its first hit of any intermediate integer height gives a crossing of the truncated rectangle.

2.1F4step 1.1step 1.2givenconstructalgebra

Suppose n≥32. Set k=⌊n/32⌋, L=16k, h=4k=L/4, W=⌈αn⌉ and J=⌈W/h⌉. For 0≤j≤J put aj=jh, bj=−jh, and take the closed squares Qj=[aj,aj+L]×[bj,bj+L]. For 0≤j<J take the closed connectors Cj=[aj+h,aj+L]×[bj−h,bj+L]. Each connector has width 12k and height 20k. Since aj+bj=0, every vertex of every square and connector has original coordinate 0≤y=(u+v)/2≤L≤n. The corner translations (jh,−jh) preserve P. Every square embeds and has horizontal crossing probability at least a by [F4], since L<m. A vertical aspect-two rectangle of the same connector width S=12k, starting at its bottom, has height 24k<n<m. Step 1.2 bounds its vertical crossing probability below by d. Truncation at height 20k implies a vertical crossing of Cj, so that event too has probability at least d. Its witness may use the taller box to establish the numerical bound, but the event being intersected below is supported on Cj itself.

3.1F5step 1.1step 2.1constructalgebra

Require horizontal crossings of all Qj and vertical crossings of all Cj. In Qj∩Cj the horizontal path has a subpath crossing its width-12k span, while the vertical path has a subpath crossing its full height-L span. They intersect by [F5]. The same applies to Qj+1∩Cj, whose height-L span is shifted down by h. Thus the union is a connected actual-open-edge set. A left endpoint of the Q0 crossing has u=0 and hence x≤0. A right endpoint of the QJ crossing has u=Jh+L and v≤−Jh+L, hence x≥Jh≥W. Their joined path stays in 0≤y≤n by step 2.1. Truncate this path between its last vertex on x=0 before its first hit of x=W. It gives the required ordinary-grid crossing of {0,…,W}×{0,…,n}. The patches need not jointly embed: lift the periodic configuration to the plane and choose the guaranteed crossings in these fixed lifts. Their adjacent overlaps are literal planar overlaps. The final target itself embeds, since W<m and n<m, so its projected crossing is unambiguous.

4.1F2step 2.1step 3.1algebra∎

By finite FKG [F2], the joint event in step 3.1 has probability at least aJ+1dJ≥min⁡(a,d)2J+1. Because k≥n/64 for n≥32, h≥n/16 and h≥4. Hence J≤W/h+1≤16α+5/4 and 2J+1≤32α+7/2. Its probability is at least min⁡(a,d)⌈32α+5⌉>0, independently of n,m. When 1≤n<32, use the fixed ordinary horizontal row of W edges. All its labels are distinct since W<m. Repeated conditioning with [F2] gives probability at least ρW≥ρ⌈32α⌉. The minimum of these two positive constants is a permissible c(α,q). This proves the stronger ceiling-width statement for all stated sizes.

Remarks

The auxiliary rotated boxes have closed sides and scales divisible by four; these conventions are used only in the proof. The target is the stated ordinary square-grid rectangle for every positive integer size. The local mixed-reflection supplier uses first intersection along the vertical path; the earlier first-along-reflected-path prescription is not used. Its finite counterexample and correction history are retained in the owner research directory.

Depends on

Dependency tree · 2 levels

Sources