Construction of the two-dimensional O(3) nonlinear sigma model as an interacting continuum quantum field theory with a mass gap
The two-dimensional nonlinear sigma model is the lattice measure with . Perturbative renormalization (Polyakov 1975; Brezin-Zinn-Justin 1976) predicts asymptotic freedom: as and the lattice spacing shrinks, a finite physical mass survives, and integrability (Zamolodchikov-Zamolodchikov; Hasenfratz-Maggiore-Niedermayer) describes the expected massive continuum theory. Rigorous constructions existed only for large (Kupiainen, Kopper), for a hierarchical model (Gawedzki-Kupiainen), or in perturbation theory (Mitter-Ramadas), with Balaban's program partial. Does the nearest-neighbour lattice model have a continuum limit that satisfies the Osterwalder-Schrader axioms, is interacting (non-Gaussian), and has a positive mass gap?
- Result
- Proved(see note)
- Status
- Candidate (review pending)
- AI contribution
- AI-discovered
- Method
- Argument
- Field
- Constructive quantum field theory; asymptotically free sigma models
- Posed by
- Long-standing goal of constructive field theory; the manuscript frames it via Polyakov (1975), Brezin-Zinn-Justin (1976) and work of Kupiainen, Gawedzki-Kupiainen and Balaban, naming no single poser
- Year posed
- —
- Years open
- —
- Solved
- 2026-10-04
- Model
- Unreleased internal OpenAI model
- Vendor
- OpenAI
- Collaborators
- —
- Verification
- Unreviewed
- Publication
- Announced
- Collection
- OpenAI math release (October 2026), version adc7f12
- Significance
- 62 / 100
- Disclosed cost
- —
- Wikipedia
- No dedicated article
What was actually shown
Theorem 1.2: along bare couplings with spacing , suitably normalized smeared lattice spins converge in law and in all moments; the limiting Schwinger distributions are Euclidean and invariant, reflection positive and clustering with factorial bounds, reconstruct a local relativistic theory in dimensions with a unique vacuum and a Hamiltonian gap on the whole vacuum complement, and have a nonzero connected four-point function, so the theory is not Gaussian. Theorem 1.1: normalized by susceptibility and second-moment length, the limit exists along every real without subsequences. The companion proves an isolated one-particle mass atom in the spin two-point spectral measure. Not shown: the value of the mass, the S-matrix of the constructed theory, or a continuum construction for other .
What the AI did
The release README says the results were produced by an unreleased internal OpenAI model with a fixed procedure, on average about three hours of ChatGPT Pro thinking compute per result, and that some outputs build on earlier model results. This result is not among the README's exceptions (the Hodge conjecture for CM abelian varieties and the Re(s) > 11/12 zero-free region). The manuscripts are authored 'OpenAI' and name no human author. The principal manuscript (October 4, 2026) adapts the exact blocking and angular-integration estimates of the family's O(4) manuscript ('Sharp mass bounds', September 23), proving the O(3)-specific changes itself; the same-day companion 'An Isolated Particle Pole' builds on this construction.
Verification
No independent mathematician has checked this yet. Checked here: Theorems 1.1 (canonical continuum limit) and 1.2 (construction at a fixed physical scale) and Corollary 1.3 were read against the constructive problem as the introduction states it. The proof (exact block renormalization, cutoff comparison at fixed physical scale, source tracking and OS reconstruction) was not refereed. No Lean formalization: the family's Lean scope note says only the O(n) exponential-decay theorem is formalized. The construction relies on intermediate estimates imported from the unreviewed O(4) companion of the same release, with stated hypotheses. Scope the paper itself states: one lattice action, periodic thermodynamic state, no theta term and no external field; no comparison with other lattice actions; no use of the factorized S-matrix.