VibeMathedMath problems solved with AI

Mathieu's spectral isometry conjecture

Let AA and BB be unital complex C∗C^*-algebras. A linear map T:A→BT:A\to B is a spectral isometry if r(Ta)=r(a)r(Ta)=r(a) for every a∈Aa\in A, where rr is the spectral radius. Mathieu conjectured that every unital surjective spectral isometry between unital C∗C^*-algebras is a Jordan isomorphism, that is, a linear bijection with T(a2)=T(a)2T(a^2)=T(a)^2 for all aa. It was known for commutative algebras, for properly infinite von Neumann algebras, and for real rank zero algebras without tracial states. Is it true for all unital C∗C^*-algebras?

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-discovered
Method
Argument
Field
Operator algebras; preserver problems in Banach algebras
Posed by
Martin Mathieu (Mathieu and Schick, First results on spectrally bounded operators, Studia Math. 152 (2002), Conjecture 4.6; restated as Problem 5 in Mathieu's 2009 problem collection)
Year posed
2002
Years open
24y
Solved
2026-10-29
Model
GPT-6 Astra
Vendor
OpenAI
Collaborators
Matthew J. Colbrook, George Stepaniants
Verification
Lean-checked, statement unaudited
Publication
Preprint
Significance
26 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

(Theorem 1.1) Let AA and BB be unital complex C∗C^*-algebras, and let T:A→BT:A\to B be a surjective complex-linear map such thatT(1A)=1B,rB(T(a))=rA(a)(a∈A). T(1_A)=1_B, \qquad r_B(T(a))=r_A(a) \qquad (a\in A). Then TT is a Jordan isomorphism. In particular,T(a2)=T(a)2(a∈A). T(a^2)=T(a)^2 \qquad (a\in A).

What the AI did

From the paper: the proof was found by GPT-6 Astra after a sequence of simple prompts. The authors then discussed it further with ChatGPT 6 to understand the mechanisms and wrote the paper themselves, as a human digestion of the AI proof; ChatGPT 6 was used only for minor language editing of the penultimate draft.

Verification

No independent mathematician has checked this yet. Lean-checked on this site's ladder, not Lean-verified. Checked here on 15 November 2026: arXiv 2610.40470 (TeX source), including Theorem 1.1, the AI statement and the citation of Conjecture 4.6 of Mathieu and Schick (2002); the attribution was confirmed in Mathieu and Sourour (arXiv 1110.6856), who write that Mathieu conjectured it there. Palomar entry PALOMAR-2026-10-28-000031 v1 read: source MColbrook/Proof-of-the-spectral-isometry-conjecture-for-C-star-algebras at commit e08ba67, Lean v4.35.0-rc3, Comparator theorems main_result, entire_group and normalization, permitted axioms propext, Classical.choice and Quot.sound, kernel replays by Palomar (nanoda, con-ron). Challenge.lean (62 lines, Mathlib imports only) was read: arbitrary CStarAlgebra instances, a complex-linear map, surjective, T 1 = 1, spectralRadius preserved for every element, concluding bijective and square-preserving. No boundedness or adjoint hypothesis is added, so it is the conjecture as posed. A grep of all 360 Lean files (about 66,600 lines, including vendored OAI library files) found no sorry outside the challenge stub and no axiom, admit, native_decide, implemented_by, extern or unsafe. The repository's own CI (Lean verification) passed on commit 223af31, which differs from e08ba67 only in the README. Not rebuilt here. No independent formal statement (Formal Conjectures or a tracker) was found.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion