VibeMathedMath problems solved with AI

Morel's purity conjecture for A1\mathbb A^1-homotopy sheaves of pure S1S^1-spectra

Let SS be a separated noetherian scheme of finite Krull dimension. Call an S1S^1-spectrum over SS pure if it lies in the smallest triangulated subcategory of the stable A1\mathbb A^1-homotopy category containing the suspension spectra of smooth projective SS-schemes and closed under arbitrary coproducts and retracts; call a Nisnevich sheaf on SmS\mathrm{Sm}_S pure if Zariski and Nisnevich cohomology agree and cohomology is insensitive, in a range of degrees, to removing closed subsets of high codimension. Morel conjectured (Conjecture 6.4.12) that for every pure S1S^1-spectrum EE and every nn, the sheaf π‾nA1(E)\underline{\pi}^{\mathbb A^1}_n(E) is pure. Is this true?

Result
Disproved(see note)
Status
Candidate (review pending)
AI contribution
AI-assisted
Method
Construction
Field
Algebraic K-theory, motivic homotopy theory
Posed by
Fabien Morel, The stable A1-connectivity theorems, K-Theory 35 (2005), Conjecture 6.4.12
Year posed
2005
Years open
21y
Solved
2026-11-02
Model
GPT-6 Astra; Claude Opus 5.5
Vendor
OpenAI; Anthropic
Collaborators
Niels Feld
Verification
Unreviewed
Publication
Preprint
Significance
15 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Disproof claimed. Theorem A: over a regular base SS, KS∧MK_S\wedge M is pure for every spectrum MM. Theorem B: injectivity of restriction to dense opens for its homotopy sheaves is equivalent to Gersten injectivity at the henselized local rings of smooth SS-schemes. Corollary D: with A=Z3[[x,y]]/(3+x2−y3)A=\mathbb Z_3[[x,y]]/(3+x^2-y^3) and the author's earlier class in K2(A;Z/3)K_2(A;\mathbb Z/3) that dies on the fraction field (arXiv 2608.05005), π‾2A1(KS/3)(S)→π‾2A1(KS/3)(U)\underline{\pi}^{\mathbb A^1}_2(K_S/3)(S)\to\underline{\pi}^{\mathbb A^1}_2(K_S/3)(U) is not injective, so the conjecture fails over S=Spec AS=\mathrm{Spec}\,A, and also over Spec Z[1/2][x,y]/(3+x2−y3)\mathrm{Spec}\,\mathbb Z[1/2][x,y]/(3+x^2-y^3). Integral and rational failures are left conditional on unrefereed Gersten counterexamples.

What the AI did

An earlier version of the manuscript was drafted with GPT-6 Astra. Claude Opus 5.5 checked the proof, searched the literature, and proposed some arguments.

Verification

No independent mathematician has checked this yet. Read by this site on 15 November 2026: the abstract, introduction, Morel's definitions and Conjecture 2.6 as quoted from [Mor05, Conjecture 6.4.12], Section 6 (the finite-coefficient counterexample), Remark 8.1 and the AI disclosure of hexagon 2611.00017v1. The paper states the conjecture as Morel did, without a regularity hypothesis, and disproves it over a regular two-dimensional local base. The arithmetic input, a nonzero class in K2(A;Z/3)K_2(A;\mathbb Z/3) vanishing on the fraction field, is the author's earlier single-author preprint arXiv 2608.05005 (v2, September 2026), whose abstract was read but not its proof. The deduction follows Morel's own remark that purity implies Gersten injectivity. Hexagon is a moderated, not peer-reviewed, repository; it labels this text primarily AI-generated with human understanding of all parts. No formalization.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion