Prime-power structure of the stable regime for modular Schur numbers

Sm(k,ℓ) = m/gcd(m,ℓ−1) − 1 for all k ≥ n−1, with a Lean 4 formalization behind an independent comparator gate.

  • A closed form that holds for every modulus. With n = m/gcd(m,ℓ−1), we prove Sm(k,ℓ) = n − 1 for all m,ℓ ≥ 2 and k ≥ n−1. This settles the m ≥ 8 case left open by D’orville–Sim–Wong–Ho (2025) and subsumes the earlier per-modulus results. The proof is checked by the Lean 4 kernel against mathlib alone: no sorry, no project axioms, no native_decide.
  • The color threshold is exact at prime moduli, and composite moduli behave differently. For a prime p with ℓ ≥ p−1 and ℓ ≢ 1 (mod p), the threshold is k0(p,ℓ) = p − 1. In ℤ/12 the pair {1, 5} is a safe two‑element class, which refutes the natural guess that only singletons are safe.
  • The composite threshold reduces to one finite question, which stays open. In the stable regime the threshold depends only on ℓ mod m and splits as K + τ. We prove a coordinate‑union edge rule and a separated‑fiber criterion for the residual cover τ, but certified fibers (n = 880, and n ∈ {330, 660}) break the matching witness a uniform closed form would need, so τ at composite m remains conjectural.

Prove2.me missions

Missions I captain on Prove2.me, where every Lean 4 statement and proof is checked by the platform. Each link goes to the mission page, its milestones, and its verified theorems.

Tooling

  • turbotools — Claude Code plugins
  • super-cow-powers — a Claude Code skill for Debian packaging
  • nthdegree — persistent cross-session memory and hybrid document search for AI coding agents coming soon
  • piqd — SAT job daemon with a content-addressed CNF store, CaDiCaL and Kissat backends, and DRAT/LRAT proof support coming soon