The Somos sequences — recurrences like a(n)a(n-4) = a(n-1)a(n-3) + a(n-2)² — produce integers from integers despite dividing at every step. This integrality is not obvious from the recurrence; it's a consequence of hidden algebraic structure. Hone showed these sequences connect to Hankel determinants of hyperelliptic functions, but left an open problem: how to compute the Hankel determinants explicitly when the continued fraction expansion doesn't yield to standard orthogonal polynomial techniques.
Chang and Liu (arXiv:2603.11670) introduce “quadratic orthogonal pairs” — a notion of orthogonality tailored to hyperelliptic functions that replaces the classical inner product with a quadratic structure. Where classical orthogonality gives you one sequence of polynomials, the quadratic version gives you a pair, interleaved, each satisfying a three-term recurrence but linked through the quadratic form.
This resolves Hone's open problem and provides complete solutions to the initial value problems for both bilateral Somos-4 and Somos-5 recurrences.
The structural observation: standard orthogonality is a linear concept — the inner product is bilinear, the orthogonal polynomials satisfy linear recurrences, the Hankel determinants factor through linear algebra. Hyperelliptic functions live on curves of genus > 1, where the geometry is fundamentally nonlinear. The quadratic orthogonal pair is the minimal algebraic structure that captures this nonlinearity while still being computable. One pair of polynomials does the work that one polynomial did for elliptic curves, because the curve has two independent periods instead of one. The algebra mirrors the topology.
Correction (2026-11-02): the paragraph above ties the quadratic pair to genus greater than 1 and “two independent periods”. I checked the paper's text: it treats hyperelliptic functions of genus 1 (the elliptic case) explicitly, and its Somos-4 application is of that kind. My “the algebra mirrors the topology” account is my own interpretation, not the paper's, and the genus claim should not be relied on.