VibeMathedMath problems solved with AI

Orthogonal bases of symmetric unitary matrices in odd dimensions (mixed-unitary rank of the symmetric Werner-Holevo channel)

Let SnS_n be the space of complex symmetric n×nn\times n matrices, of dimension (n+12)\binom{n+1}{2}, with the Hilbert-Schmidt inner product ⟨A,B⟩=Tr(AB∗)\langle A,B\rangle=\mathrm{Tr}(AB^*). Does SnS_n always have an orthogonal basis consisting of unitary matrices? Girard, Leung, Levick, Li, Paulsen, Poon and Watrous showed this is equivalent to the symmetric Werner-Holevo channel on MnM_n having mixed-unitary rank equal to its Choi rank (n+12)\binom{n+1}{2}, proved it for even nn and n=3n=3, and conjectured it for every nn. The open cases are odd nn.

Result
Proved(see note)
Status
Partial result
AI contribution
AI-assisted
Method
Computation
Field
Quantum channels; matrix analysis
Posed by
Luftbahnfahrer, MathOverflow question 316985 (2018); conjectured for all n by Girard, Leung, Levick, Li, Paulsen, Poon and Watrous (arXiv 2003.14405)
Year posed
2018
Years open
8y
Solved
2026-11-15
Model
OpenAI Codex; Anthropic Claude (versions not recorded)
Vendor
OpenAI; Anthropic
Collaborators
Qiyuan Ma
Verification
Site-confirmed
Publication
Preprint
Significance
8 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

For every odd nn with 3≤n≤673\le n\le 67, SnS_n has an orthogonal basis of symmetric unitary matrices; with the known even case the conjecture of Girard et al. holds for all n≤68n\le 68. The basis consists of chirped circulants DcCDcD_cCD_c and their diagonal shifts; orthogonality reduces to a square system of (n2−1)/4(n^2-1)/4 real equations in as many phases. For n=5n=5 the basis is explicit, with entries in a number field of degree 8; for the other nn existence is a computer-assisted proof (Krawczyk operator in interval arithmetic). Previously known odd cases below 70: n=3,7,23,31,47n=3,7,23,31,47. This is a finite range, not a proof for all odd nn: the general case stays open, and the paper states it as a conjecture about its own system.

What the AI did

From the paper's acknowledgements and the Hexagon provenance statement: AI assistance (Codex from OpenAI and Claude from Anthropic, versions not recorded) was used in the literature search, in designing and running the computational experiments, in writing the verification and certification code, and in drafting the manuscript. Hexagon labels the paper 'Primarily AI-generated text', human understanding 'some parts'; the author reports that some parts, but not all, have been checked and understood, without saying which. AI is not an author.

Verification

Re-checked by this site on 19 November 2026 with code written here, not the author's. The author's archive (Zenodo 24246318) supplies, for each odd nn from 3 to 67, a vector of phases defining (n+12)\binom{n+1}{2} 'chirped circulant' matrices. (1) An independent Krawczyk test in Arb ball arithmetic (python-flint 0.9.0, 128 bits, radius 10−910^{-9}) certified a zero of the square system FnF_n near every one of the 33 archived centres, with the same deviations as the paper's log (for example 1.888×10−121.888\times10^{-12} at n=67n=67). (2) Independently of the paper's reduction theorem, all the matrices were built in full in floating point and tested directly: symmetric and unitary to 10−1410^{-14}, and the whole Gram matrix equal to nn times the identity to 5×10−135\times10^{-13}, up to the 2278 matrices of n=67n=67. The reduction (Theorem 5), a one-page computation, was read and agrees with (2). The MathOverflow question (no answers) and the abstracts of Girard et al. and Levick-Rahaman were read. Not run here: the author's own scripts, including the exact number-field verification for n=5n=5. No independent expert has commented. The general odd case is open, so the status is Partial.

Sources

Submitted by RustyDingo867 on

Changelog2 changes

Discussion