VibeMathedMath problems solved with AI

The supercritical Ten Martini Problem: Cantor spectrum for every analytic one-frequency Schrödinger operator with positive Lyapunov exponent

For a real analytic, nonconstant vv on T=R/Z\mathbb{T}=\mathbb{R}/\mathbb{Z}, an irrational frequency α\alpha and a phase θ\theta, consider the Schrödinger operator on ℓ2(Z)\ell^2(\mathbb{Z}) (Hv,α,θu)n=un+1+un−1+v(θ+nα)un.(H_{v,\alpha,\theta}u)_n=u_{n+1}+u_{n-1}+v(\theta+n\alpha)u_n. Its spectrum Σv,α\Sigma_{v,\alpha} does not depend on θ\theta. The Ten Martini Problem, posed for the almost Mathieu operator v(x)=2λcos⁡2πxv(x)=2\lambda\cos 2\pi x and solved by Avila and Jitomirskaya, asks that the spectrum be a Cantor set for every irrational α\alpha. For general analytic vv this can fail where the Lyapunov exponent L(E)L(E) vanishes (examples announced by Argentieri and Avila). The supercritical Ten Martini Problem is the conjecture that it never fails where L(E)>0L(E)>0: for every nonconstant analytic vv and every irrational α\alpha, the part of the spectrum in the regime of positive Lyapunov exponent is a Cantor set.

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-assisted
Method
Argument
Field
Spectral theory; quasi-periodic Schrödinger operators
Posed by
Jiangong You, Sci. Sin. Math. 54 (2024), no. 6, 863-870, Conjecture 1; see also Marx and Jitomirskaya (2017)
Year posed
2024
Years open
2y
Solved
2026-11-09
Model
Unspecified LLM
Vendor
—
Collaborators
Lingrui Ge, Svetlana Jitomirskaya, Jiangong You
Verification
Unreviewed
Publication
Preprint
Significance
42 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1 of arXiv:2611.13068 (Ge, Jitomirskaya, You): for every nonconstant v∈Cω(T,R)v\in C^\omega(\mathbb{T},\mathbb{R}) and every irrational α\alpha, the set Σv,α∩O‾\Sigma_{v,\alpha}\cap\overline{\mathcal O} is a Cantor set, where O={E:L(E)>0}\mathcal O=\{E: L(E)>0\}. In particular the whole spectrum is Cantor whenever L(E)>0L(E)>0 throughout the spectrum, which by Sorets and Spencer holds for λv\lambda v with ∣λ∣|\lambda| large. Nothing is claimed for subcritical energies, where intervals can occur. The paper notes the result cannot in general be strengthened to the dry version (all gaps open), citing a counterexample in preparation. The proof builds on two preprints by the same group, arXiv:2407.08866v3 (the intrinsic symplectic center) and arXiv:2609.15812 (The Robust Ten Martini Problem), both cited by the paper as arXiv preprints.

What the AI did

From the paper's section "Acknowledgement and AI-use": the authors conceived the entire proof strategy in 2025 without AI. Their first use of AI came after the strategy, the intrinsic symplectic center and the extension of Kotani theory were in place: "We then used an LLM to help carry out the renormalization and local parameter KAM steps. In particular, it helped us derive the renormalization estimate in Lemma 5.1 and discover the monotonicity property in the Lemma 3.2. Its assistance was particularly valuable in finding how to use this property to simplify our original local KAM argument with parameters." An LLM was also used to speed up preparation of the manuscript, stress-test the arguments, find overlooked cases and improve the exposition. The model is not named.

Verification

No independent mathematician has checked this yet. Read by this site on 19 November 2026 in the TeX source of arXiv:2611.13068v1 (60 pages, posted 9 November 2026): the introduction, Theorem 1.1, the outline of the proof, the history section and the AI-use section. The theorem is the conjecture as the paper states it, restricted to the closure of the positive-Lyapunov set and with no arithmetic condition on the frequency. You's 2024 article, which poses the conjecture and is in Chinese, could not be opened here, so the wording of the problem is taken from this paper, of which You is a coauthor. The proof was not refereed and there is no formal proof. It depends on two earlier preprints by the same authors. An arXiv search for the problem found no competing or earlier claim; expert reaction outside arXiv was not searched for.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion