Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Claim. The infimum of , over all degrees and all monic whose roots lie in the closed unit disc, is . Tang's note proves the upper bound on the model polynomials : the open set has congruent components, each with boundary length , and this tends to as . For the lower bound the note splits on . If , the note applies Pommerenke's Theorem 3 of 1961 to the component of in the open set and obtains a diameter of at least , the largest modulus of a root (equality at , where and is the unit disc). Pommerenke's statement bounds the component of in the closed set , which can be larger than when components of the open set touch at a critical point, so this step rests on Pommerenke's argument carried over to the open component rather than on his statement as printed; the formalization linked below proves that open-component version (diameter above ) itself. The diameter of a bounded open set equals that of its boundary, and a rectifiable closed curve has length at least twice its diameter, so has length at least . If , every root lies on the unit circle, is a boundary point of some component, that component contains a root and so reaches the unit circle, and its boundary has diameter at least and length at least . The note also raises the fixed-degree question, whether minimizes among monic polynomials of degree with roots in the disc, proves it for and , and leaves open; that question is not part of the problem.
Sources used. Pommerenke's Theorem 3 (Michigan Math. J. 8 (1961), p. 99) is the lower bound's one outside input; it is compiled on the result page theorem_3 of the card pommerenke_1961_metric_properties_complex_polynomials. Tang's note is held only at its public address, a PDF with its TeX source in the author's repository, uploaded on 2026-01-05 and announced in the site's thread the same day; there is no journal version.
Acceptance. Reviewed: Terence Tao wrote in the thread on 2026-01-14 that Tao had read the note and thought the argument works, resting on Pommerenke's Theorem 3, which Tao saw no reason to doubt; and the site's curator, T. F. Bloom, labels the problem solved and credits the resolution to Tang. No refereed publication exists. Tang disclosed in the thread on 2026-01-21 that ChatGPT-5.2 Pro was used in Step 1 for numerical exploration, which suggested as the likely extremizer, and that a purported proof it produced was discarded as incorrect. Nothing here is independently reviewed by this project.
Lean. Lorenzo Luccioli's development Erdos1044, published as a gist on
2026-04-28, declares itself a formalization of Tang's note and proves
Erdos1044.erdos_problem_1044 : lambdaInf = 2, importing only Mathlib, as
its header states. Luccioli wrote in the site's thread on 2026-04-28 that they
had formalized Tang's solution using Aristotle; the gist's header does not
name the system. A second posting of the same development, in Boris
Alexeev's repository of Lean proofs, names Tang as informal author and
Aristotle and Luccioli as formal authors; both postings are linked above.
The development proves the open-component form of Pommerenke's Theorem 3
(the component of has diameter above when ) within the
file, which its docstring presents as a corrected statement of the cited
theorem. It is the Lean qualifier of the site's label. The
formal-conjectures statement file for the problem, at its revision of
2026-09-18, credits Tang, states the problem with every theorem's proof
left as sorry, and carries a formal_proof attribute pointing to the
copy in Alexeev's repository. This corpus has not built or audited the
development, so no formalized evidence is listed.
Depends on. Nothing on the wiki; the one outside input is the 1961 theorem named under Sources used.