Answers are grounded in this paper and cite their sources.
Computational study reveals a lossless conic-relaxation prefilter reduces certification work 546-fold in an elliptic curve family, ruling out ranks of 21 or greater across tested slices.
Elkies and Klagsbrun exhibited a family of elliptic K3 surfaces over Q(t) with torsion Z/2Z and Mordell–Weil rank 9, and describe the tension between an aggressive heuristic cutoff in a record search and the risk of eliminating promising fibers. We remove that tension for this family with a conic relaxation of the 2-isogeny descent bound, evaluated by an F_2-rank of a matrix of Hilbert symbols over the family's freely given factorization. It computes an upper bound rank(E) ≤ S in about 0.6 ms per fiber, provably never discards a fiber, and composes with an exact 2-isogeny Selmer stage S′. At target threshold 21 the cascade reduces certification work by a factor of about 546. With it we exhaustively decided the S ≥ 21 slices of three branches of the family (about 1.67 million fibers certified unconditionally with PARI/GP ellrank); the outcome there is exclusion of rank ≥ 21 in each box, with the certified slices attaining at most exact rank 16/15/15. Byproducts include a fiber of unconditionally exact rank 18 (listed in Dujella's Z/2Z tables) and nine fibers of exact rank 16, now curves #481–#490 of the ICARM Elliptic Curve Rank Leaderboard. All claims carry their verification method; all computations are reproducible from the accompanying dataset. Provenance: the author is a software developer, not a professional mathematician; substantial AI assistance (Anthropic Claude) was used throughout, as stated in the note. Every numerical claim names the check that guards it. Version 2 (erratum and addition). v1 defined exact rank r as PARI/GP ellrank returning coinciding bounds r1 = r2 = r and presented this as unconditional. ellrank may return r1 = r2 while finding fewer than r independent points, filling the gap by a parity argument that assumes finiteness of the Tate–Shafarevich group. No exclusion and no rank ≥ 16 claim is affected; the rank-15/14 tallies are split into unconditional (r independent points, height-matrix rank r) and parity-only cases, recounted from the v1 dataset (script audit-paritaet.py). v2 also adds a section on steering the search by measured branch screening, which yielded four further fibers of exact rank 18 on branch u = 29/13 on 2 September 2026.
No takes yet. Share an insight, caveat, or question.
Matthias Breddin (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: