Answers are grounded in this paper and cite their sources.
Let A(d,s) be the largest number of points on the unit sphere in R^d with pairwise inner products at most s. We prove that A(d,s) ≤ 2d(d+2)(1−s²)/(3−(d+2)s²) for d ≥ 4096 and 0 < s² ≤ 1/d + 1/(100d²). At s = d−1/2 the bound equals d(d+2) and is attained exactly by antipodal lifts of complete systems of real mutually unbiased bases. Together with the results of Cohn, de Laat and Leijenhorst for 16 ≤ d ≤ 1024, this shows that the Kerdock spherical codes and the binary Kerdock codes are optimal in every dimension 4^k, k ≥ 2. The proof also gives a stability estimate and an asymptotic version for |s⁻² − d| = o(d1/8). Files: the paper (PDF) and a Lean 4 supplement formalizing the main cardinality bound, the stability estimate, the equality conditions and the binary bound. See the README inside the archive.
No takes yet. Share an insight, caveat, or question.
YouJun Cheng (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: