Answers are grounded in this paper and cite their sources.
Unrefereed research candidate. In one sentence: for every value of the repulsion parameter β, the points of the Sine-β process, counted outward from a fixed origin, have log-concave and positively dependent joint laws, which proves a 2024 conjecture of Johnston and Prochno; as a consequence the typical spacing between neighbouring points obeys explicit Gamma-type bounds whose shape β+1 cannot be improved. The object. Eigenvalues of a large random matrix, looked at on the scale of one spacing, form the same random pattern for a very wide range of models: the Sine-β point process. The parameter β measures how strongly neighbouring points repel; β = 2 is the sine process of the Gaussian unitary ensemble. Exact formulas are rare, so shape properties that hold for every β are useful. The question and the obstacle. Johnston and Prochno (arXiv:2410.10754v2, Conjecture 2.2) conjectured that, for the sine process, every finite selection of points numbered from a fixed origin has a log-concave joint density. Edge results of Baslingker, Krishnapur and Madiman work because the k-th point from an edge is a fixed coordinate. In the middle of the spectrum the number of points to the left of the origin is random, and mixing log-concave laws can destroy log-concavity. The idea and the results. Put the particles on a circle (the circular β-ensemble) and cut the circle at the origin: the first point to the right of the origin is then literally the first coordinate. The cut density is log-concave and multivariate totally positive of order two (MTP2), and both properties survive the limit (Valkó–Virág). Proved for every β > 0: (1) the conjecture, together with MTP2; (2) a stronger statement that keeps the repulsion: the density divided by the β-th power of the Vandermonde product of the selected points is still log-concave and MTP2, also for the process seen from a typical point; (3) the mean-one typical spacing G is dominated in convex order by Gamma(β+1, β+1), so Var(G) ≤ 1/(β+1), with matching moment and tail bounds; the shape β+1 is optimal, small gaps have probability of exact order tβ+1, and the variance bound becomes exact as β → 0; (4) the same Gamma comparison for any eigenvalue difference, divided by its mean, in a finite β-ensemble with convex potential, at every matrix size, which improves moment bounds and far-tail bounds of Johnston (2026) for GUE gaps; (5) for the Hua–Pickrell process (a point charge δ at the origin) all labelled laws are MTP2 for Re δ > −1/2 and log-concave exactly when Re δ ≥ 0 (as laws; with densities at β = 2); (6) at the hard edge, the positive labelled Bessel-type points are log-concave for a ≥ 1/β − 1, a range not covered by earlier work. Limits. At β = 2 the variance bound is 1/3 against the true value 0.180; the large-gap bounds decay only exponentially, while the truth decays like a Gaussian (the bound on P(G ≥ 3) is 0.067 against the true 4.9×10−5). Nothing here concerns zeros of the Riemann zeta function. The exact hard-edge threshold, densities of Hua–Pickrell laws for β ≠ 2, Gamma shapes for sums of spacings and the sign of dependence between spacings are open. The proofs rely on strong known limit theorems (Valkó–Virág; Li–Valkó) and on classical closure properties of log-concavity and total positivity. Contents. The paper (LaTeX source and PDF), two Python scripts with their recorded outputs and the figure: verify.py runs 9,694 finite falsification checks of the formulas used in the proofs, including negative controls below each threshold; sine2_calibration.py computes the exact β = 2 values by Bornemann's Fredholm-determinant method. Also a dated prior-art search record, the bibliography from registry records, a machine-readable index, a research-gate record, a replay receipt and SHA-256 sums. Assurance. Written analytic proofs; the computations are finite checks, not part of any proof. Internal AI-assisted review only (five reviewer roles, an adversarial review by a second AI model, a re-review and two proof audits): not peer review, independent reproduction or formal verification. Creator: Anonymous. Text, data and figures CC0-1.0; code MIT.
No takes yet. Share an insight, caveat, or question.
A 2026 study studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: