Four-dimensional disk embedding and Wall's conjecture
From OpenAI Aftermath
The disk embedding conjecture in dimension four asks whether immersed disks in a four-manifold with algebraically dual spheres and vanishing intersection invariants can always be replaced by disjoint locally flat embedded disks with the same boundary. Freedman proved it for good fundamental groups, which makes topological surgery and the five-dimensional s-cobordism theorem work for them; whether the free group is good has been a central open question. Wall's question asks whether every Poincaré duality group is the fundamental group of a closed aspherical manifold. Three manuscripts in OpenAI's public mathematics collection (github.com/openai/math), all dated 24 September 2026, claim that the unrestricted disk embedding conjecture fails, that the free group on two generators is not good, and that a finitely presented four-dimensional Poincaré duality group has no closed aspherical manifold model. The family has no Lean formalization in the repository's catalogue. This page summarizes the published claims and lists publicly available follow-up work. Statements are the respective authors' claims; the publication of a paper is not presented here as independent verification.
Background
- As recounted in the introductions of the manuscripts:
- In dimension four the Whitney disks needed to cancel intersections may themselves meet the surfaces. Casson's handles and Freedman's theorem that every Casson handle is topologically an open two-handle gave the simply connected theory; Freedman and Quinn developed it with capped gropes, and the disk embedding theorem holds for good groups.
- Freedman and Teichner extended the good class to groups of subexponential growth, and Krushkal and Quinn gave another proof; the class is closed under subgroups, quotients, extensions and directed colimits. Powell, Ray and Teichner supplied a missing construction of geometric dual spheres in the proof and record that the nonabelian free-group case was open in 2025.
- Freedman asked whether the Borromean rings are A–B slice; Krushkal and Freedman–Krushkal obtained positive results for weakened versions of this embedding problem.
- Wall asked whether every integral Poincaré duality group is the fundamental group of a closed aspherical manifold. Davis used reflection groups to turn finiteness examples of Bestvina and Brady into integral Poincaré duality groups in every dimension at least four that are not finitely presented, and so cannot be such fundamental groups.
The original results
- Claim (boundary-only obstruction). The main theorem of A boundary-only obstruction to four-dimensional disk embedding states that there is a compact oriented smooth four-manifold containing finitely many disk maps with framed algebraic dual spheres satisfying the intersection conditions of the disk embedding theorem, whose boundary circles bound no disjoint locally flat disks, even when no homotopy classes or output framings are prescribed. Its corollary states, through a theorem of Powell, Ray and Teichner, that the free group on two generators is not good in the sense of Freedman and Quinn, and neither is any group containing it.
- Claim (marked tensor obstruction). A marked tensor obstruction to four-dimensional disk embedding states a counterexample to the unrestricted disk embedding conjecture with immersed disks and framed algebraic dual spheres satisfying the usual equivariant intersection and reduced self-intersection conditions, but no pairwise disjoint locally flat replacements that preserve the boundary maps and induced normal framings.
- Claim (Wall's question). A PD4 group without an aspherical manifold model states that there is a finitely presented integral Poincaré duality group of dimension four with a finite classifying space that is not the fundamental group of any closed aspherical topological four-manifold. It builds on the marked tensor obstruction and on Davis's reflection method.
- Method, as described: the construction starts with a smooth four-manifold of the homotopy type of a finite graph, in which capped gropes encode triple commutators of end generators. Hypothetical disjoint locally flat disks would allow the manifold to be cut along three-dimensional walls; boundary-polarized Chern–Simons potentials on the walls then give exact identities for the equations of a representation chart of the cut manifold, and after reduction to positive characteristic a finite algebraic tensor obstruction rules them out. The third manuscript applies Davis's reflection construction to a chamber whose marked boundary has no filling of the required graph type.
- Later use in the collection. The manuscript of The Borel conjecture in dimension four imports results of the marked tensor obstruction and of the Poincaré duality group manuscript.
- Sources (openai/math repository, commit
7ef5c0c): A boundary-only obstruction to four-dimensional disk embedding; A marked tensor obstruction to four-dimensional disk embedding; A PD4 group without an aspherical manifold model; all dated 24 September 2026.
Public follow-up work
| Date | Work | Relation | Public claim | Public status (checked 30 October 2026) |
|---|---|---|---|---|
| 2026-10-07 | Disk embedding and manifold realization: boundary data and tensor obstructions, an article of the Latent Seminar blog at lowrank.github.io | Reading (exposition) | Explains disks, duals and boundary data, the three conclusions, the passage from hypothetical disks to smooth cuts, potentials and actual equations, the marked finite tensor obstruction, why five active factors suffice and the passage from a forbidden marked filling to a closed realization problem, with proof outlines on the homotopy that repairs the kernel condition, the three-parameter calculation and the final surgery calculation. | Blog article; it states that its review covers statements, introductions and selected proof mechanisms and is not an independent verification of the complete proofs. |
| 2026-10-16 | J. Jardine, Lattice24, referee reviews rank_217 (boundary-only obstruction), rank_281 (marked tensor obstruction) and rank_439 (Poincaré duality group) in JJardine919/openai-math-audit (archived at Zenodo 10.5281/zenodo.23422995) | Review (AI-assisted audit) | Reports no gap in any of the three manuscripts. The boundary-only review has no notes. The marked tensor review notes that the algebraicity step depends on a comparison of shifted symplectic forms in an appendix that it only skimmed. The Poincaré duality group review has four minor notes: the transfer of the companion obstruction to an arbitrary marked filling is argued only in outline, and the review calls the implication that free groups are not good an extraordinary consequence; a separation step for unboundedly many collections is asserted rather than checked; and two steps are left implicit or slightly misstated but reconstructed by the reviewer. | Part of an audit that its README describes as complete for the collection's manuscripts without a Lean formalization. The reviews are published under CC BY 4.0. |
| 2026-10-18 | AI Proved What?, episode 305 Free groups not good?, remade version of 23 October 2026 | Reading (exposition) | Video explainer of the three claims and their history. The episode states that the papers do not say which of topological surgery or the five-dimensional s-cobordism theorem fails, do not settle Freedman's named test cases such as the Whitehead double of the Borromean rings, and say nothing about homology manifolds or higher dimensions; if correct, the result goes the way Freedman guessed in 1983. | Web page and video; the site states that its scripts and animations are produced with AI assistance and that the narration is an AI-generated voice. Status record last checked 27 October 2026: posted as preprints: OpenAI, three papers, 133 pages, dated September 24, 2026; under expert review: not yet; peer reviewed: not yet; checked in Lean: no formalization for this item. The page states that none of the papers has been revised and that it found no public specialist comment and no reported error. |
Notes
- The background summarizes the manuscripts' own account of earlier work.
- The list covers follow-ups found in public sources up to the date above; it is not a priority or completeness claim.
