VibeMathedMath problems solved with AI

Eremenko's conjecture: a bound linear in nn for the number of zeros of a polynomial P(z,zˉ)P(z,\bar z) of bidegree (n,m)(n,m)

A polyanalytic polynomial is a polynomial P(z,zˉ)P(z,\bar z) in a complex variable and its conjugate. Let PP have degree nn in zz and degree m<nm<n in zˉ\bar z, and suppose the equation P(z,zˉ)=0P(z,\bar z)=0 has finitely many solutions. Bézout's theorem bounds their number by a quadratic expression in nn (Khavinson, Lundberg and Perry: n2+m2n^2+m^2 under a coprimality assumption). For m=1m=1 the sharp bound is 5n−55n-5 (Khavinson and Neumann), linear in nn. Eremenko conjectured that this is general: for each m≥1m\ge1 there are constants AmA_m, BmB_m such that the number of solutions is at most Amn+BmA_m n+B_m. Wilmshurst's more precise conjecture for harmonic polynomials, 3n−2+m(m−1)3n-2+m(m-1), had been disproved in 2015.

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Argument
Field
Complex analysis; valence of harmonic and polyanalytic polynomials
Posed by
A. Eremenko, The number of solutions of some non-holomorphic equations, unpublished note, 7 September 2026
Year posed
2026
Years open
0y
Solved
2026-11-09
Model
GPT-6.1
Vendor
OpenAI
Collaborators
Erik Lundberg, Zvi Rosen, Teng Zhang
Verification
Unreviewed
Publication
Preprint
Significance
18 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1 of arXiv:2611.13278 (Lundberg, Rosen, Zhang): if n>m≥1n>m\ge1, f∈C[z,w]f\in\mathbb{C}[z,w] has deg⁡zf=n\deg_z f=n, deg⁡wf=m\deg_w f=m and f(z,zˉ)f(z,\bar z) has finitely many zeros, then their number is at most (4m+1)n−5m(4m+1)n-5m. So Eremenko's conjecture holds with Am=4m+1A_m=4m+1, Bm=−5mB_m=-5m; for m=1m=1 this recovers 5n−55n-5. Proposition 4.1: for each fixed mm and all large nn there are examples with at least (4m+1)(n−m)(4m+1)(n-m) zeros, so the leading coefficient 4m+14m+1 is optimal. The exact maximum for m≥2m\ge2 stays open, and the examples are not harmonic, so the conjectured bound 2m(n−1)+n2m(n-1)+n for harmonic polynomials (Lee, Lerario, Lundberg) is neither proved nor refuted. The proof builds a smooth surface of degree n−mn-m in CP2\mathbb{CP}^2 whose genus is controlled by the number of zeros and applies the Kronheimer-Mrowka theorem (the Thom conjecture).

What the AI did

From the abstract: "The proof uses an approach suggested by GPT-6.1 based on a topological result of Kronheimer and Mrowka (Thom's Conjecture)." From the paper's AI tools disclosure: "The idea to use a topological approach based on the theorem of Kronheimer and Mrowka was suggested in an interaction with GPT-6.1. The model was also used in subsequent discussions to explore and refine the argument and to help create Figure 1. The final proofs were written and checked by the authors." The work was supported by a grant of access to OpenAI models through the ChatGPT for Academic Researchers program.

Verification

Read by this site on 19 November 2026 in the TeX source of arXiv:2611.13278v1 (13 pages): the introduction, Theorem 1.1, the generic-case construction through the immersion lemma, the statement of the sharpness proposition and the AI disclosure. Eremenko's note of 7 September 2026 was opened on his Purdue page: it states the conjecture of a bound linear in nn for fixed mm, and the theorem answers it, with the finiteness assumption made explicit. Recomputed here: the genus formula g=(n−1)(m−1)+(n2+m2−N)/2g=(n-1)(m-1)+(n^2+m^2-N)/2 combined with the Kronheimer-Mrowka bound for degree n−mn-m gives exactly N≤(4m+1)n−5mN\le(4m+1)n-5m, and m=1m=1 gives the known sharp 5n−55n-5. The construction of the surface was read only as far as the immersion lemma and the perturbation argument was not read; neither was refereed. No formal proof. An arXiv search found no competing or earlier claim; expert reaction outside arXiv was not searched for.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion