Identifies and resolves the divergence problem in spectral residuals, indicating new foundational insights in random matrix theory.
For fifty years, random matrix theorists have decomposed spacing distribution residuals into polynomial modes weighted by the Wigner surmise — and for fifty years, that decomposition has been silently divergent above β = 1. Somewhat unnoticed, because everyone truncated the series and reported power fractions as if they were meaningful. They weren't: at β = 4 (GSE), the polynomial coefficients grow by a factor of 400 over ten modes. The "kurtosis dominance" reported at large β was never a physical transition — it was the highest retained mode absorbing the divergence. This paper identifies the problem and fixes it. The natural decomposition variable is the Hellinger deviation h = √(P_bulk/P_W) − 1, which lies in L²(P_W) at all β by a trivial but apparently unnoticed bound. In the convergent Hellinger basis, the variance mode dominates at every repulsion strength — from 80% at GOE to 51% at GSE — meaning the Wigner surmise residual is always primarily a width change, never a shape distortion. The "dominance reversal" was an artifact of expanding a function that doesn't belong to the space. Three additional results: An exact, closed-form identity connecting the variance mode coefficient to the variance excess ε(β) via gamma-function ratios (Proposition 2), explaining the empirically observed c₂/ε ≈ 0.73 as a near-cancellation. A mode hierarchy inversion in Berry–Robnik mixed spectra at chaotic fraction ρ ≈ 0.6, with a measurement horizon at ρ ≈ 0.97 beyond which the geoid becomes featureless. A negative result on β_c = π: five natural geometric constructions fail to detect it as a curvature feature; the critical value is algebraic (a property of the variance excess formula), not geometric (a property of any statistical manifold). The paper completes the two-point layer of the information-geometric programme for spectral statistics, complementing the one-point Brody manifold geometry developed in the companion papers. Paper Role DOI The Instability Compression Principle ICP empirical foundation 10.5281/zenodo.18099118 The Compressibility of Chaos (Ordo ab Chao) ICP theoretical derivation: α₀ = 5π² 10.5281/zenodo.18834609 Variance Excess ε(β) formula, β_c = π — the quantity this paper decomposes 10.5281/zenodo.18650473 Information Geometry of the Brody Distribution Fisher metric, one-point layer — the layer this paper complements 10.5281/zenodo.18879754 The α-Connection Structure of the Brody Manifold Amari–Chentsov tensor, γ_E cancellation → ζ-stratification origin 10.5281/zenodo.19151206 Dual Symmetries of the Brody Statistical Manifold Z₂×Z₂ symmetry, β*=√2 — the one-point special value vs this paper's β_c=π 10.5281/zenodo.19239285 The Duality Web of the Brody Statistical Manifold Conjugation-singularity theorem, interval arithmetic methods 10.5281/zenodo.19389065 Spiked Random Matrix Signatures of the L-H Transition Experimental Krylov bridge endpoint: σ²(bₙ) diagnostic connected here via §6.3 10.5281/zenodo.19423076 The Spectral Geoid Two-point layer: Hellinger convergence theorem, mode hierarchy, ζ-stratification this paper
No takes yet. Share an insight, caveat, or question.
Jon Wiberg (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: