Answers are grounded in this paper and cite their sources.
An elementary twist-ratio bound for the SU(2) Wilson lattice, formalised in Lean 4 (note v1.1, 12 November 2026). Version 1.1: literature corrections only; no statement, proof, constant or Lean file is changed. What the note is. A technical note on a sign-averaging inequality: multiplying by −1 all links of a column and averaging over independent column signs. The centre-multiplication tool is classical (Tomboulis–Yaffe 1985, Appendix I.C, p. 332; Kovács–Tomboulis 2002, eq. (8)); the twist ratio is the classical quantity of ’t Hooft, exp(−Fel) = ½(1 − Z1/Z0), and (Z0 − Z1)/(Z0 + Z1) = f/(1 − f) with f = exp(−Fel). No novelty and no priority is claimed (the paper of Borgs, 1985, could not be read). Results. Twenty-three Lean 4 declarations: the abstract bound |⟨W⟩| ≤ tanh(κ)|A| and its sharpness; the elementary area law tanh(β|F|)area on (rectangular plate) × (finite fibre); parity and image lemmas on the discrete torus; the twist-ratio bound |Z0 − Z1| ≤ tanh(κ)N(Z0 + Z1); and a bridge theorem stated on the lattice, the Wilson energy and the SU(2) of the public Lean development of OpenAI (compiled inside that project): for every period l ≥ 2 and β ≥ 0, |Z0 − Z1| ≤ tanh(β l2)l2 (Z0 + Z1), where Z1 has the sign of the plaquettes of one base point of the (0,1)-plane reversed (all positions along axes 2 and 3). Status. The 23 declarations are checked by Lean 4.32.0 and 4.32.2 and by the independent kernel nanoda 0.4.32 (40099 declarations, no errors), standard axioms only. On 12 November 2026 an AI model that did not write them (Claude Sonnet 5.5) rebuilt the cumulative Lean file from the git archive of the package under Lean 4.32.0 on the same rented machine and environment (code 0, standard axioms only) and nanoda re-checked the closure (40099 declarations); Lean 4.32.2 and a rebuild of the environment were not redone. A transfer-matrix consequence for the odd/even centre-parity sectors of a tube (λ1/λ0 ≤ tanh(β L2)l) is proved on paper only and checked by AI models only. Written by AI models and read by other AI models; no human has checked the proofs; not refereed. Version 1.1 (literature corrections). The note now places the statements next to the classical ones: composed with the classical flux inequality of Tomboulis–Yaffe, Kovács–Tomboulis (eq. (11)) and Kanazawa (Theorems 1 and 2), the twist-ratio bound gives |⟨W[C]⟩| ≤ 2 tanh(β l2)|AC| on the periodic lattice of period l for the loops to which that inequality applies (a consequence of the cited statements as read; not formalised; the divisibility condition of Kanazawa’s proof is assumed). The misplaced location of the centre flip in version 1.0 (“Section 3” of Tomboulis–Yaffe) and the sentence that presented a paper of Borgs as the most likely neighbour of the area law are corrected. Version 1.0 remains available as its own record. Limitations. The bound is elementary and valid at every coupling but tends to a vacuous statement when the transverse section grows at fixed β; it says nothing about the mass gap in dimension four, the weak-coupling regime, the continuum limit or the neutral sector. Prepared with the assistance of Claude (Anthropic), GPT-6.1 Sol and GPT-6 Astra (OpenAI), Qwen3.8 and DeepSeek V4 Pro; the roles, including those of the literature search of 12 November 2026, are listed in the note.
No takes yet. Share an insight, caveat, or question.
Kevin Remondière (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: