What was already known
Projective spins identify opposite directions and permit Z₂ vortices. Numerical and heuristic work left possible algebraic quadrupolar correlations disputed. The proof adapts the released fourth-order response and annular methods for O(n) spins to squared interactions and projective observables.
AI involvement
Masaki Oshikawa says he asked for the result without supplying the scientific or mathematical insight behind its proof. GPT-6 Astra generated the proof and formalization; Claude reviewed and suggested changes. Oshikawa chose the models and generalizations, supplied physical context and prepared the release.
Importance
Rules out a power-law tail of this order parameter at every finite temperature within the stated models.
Limitations
The bound is slower than exponential and does not rule out every phase transition or prove vortex unbinding. The selected checks cover finite-volume quadrupolar statements, not the paper’s heuristic renormalization-group discussion or every local-limit claim. Oshikawa says he is not qualified to verify the mathematical proof; the written exposition may differ from the formal proof.
Technical verification
The exact challenge defines normalized spherical Gibbs measures, squared interactions and the two centered observables. Palomar reports successful protected comparison and Lean, NanoDa and con-ron checks of both log-square bounds, without an assumed decay or annular-response certificate.
Formal checking
Protected comparison and independent kernel checks of ProjectiveDecay.rp2_log_square and rp3_log_square under standard axioms
- Checker
- Palomar: Lean, NanoDa and con-ron
- Scope
- In the two-dimensional nearest-neighbor RP² and RP³ projective spin models, the quadrupolar two-point correlation is bounded by Cβ exp[−cβ(log(2+r))²] for every finite coupling bound β. The positive constants are uniform over finite square-lattice subgraphs, signed couplings |bₑ|≤β and observation sites; r is their lattice ℓ∞ distance. The spherical observables are (nₓ·nᵧ)²−1/3 and (qₓ·qᵧ)²−1/4.
- Basis
- Reported by the linked checker
Checking a formal statement does not establish that it matches the original problem.
Human understanding
Few is the default when evidence for a higher level is insufficient. Oshikawa explicitly says he cannot verify the mathematical proof in a short period and credits the proof exposition to AI. His physical interpretation does not establish a human account of the central mathematical mechanism.
Correction and publication history
Added a reviewed, scoped central-AI contribution with checking and attribution limits.
Read this published revision