Ferreira–Simon's Wright log-concavity conjecture

What Lean checks

The actual infinite Wright series has a continuous beta-zero counterexample family, with admissible alpha approaching one. Strict compilation, the complete 25-declaration owned audit including eight private declarations, and independent stock-kernel replay passed. The positive-beta refinements in the written proof and the combined repository gate remain separate and pending.

Disproved. The parameter threshold in Ferreira and Simon's §6.4 conjecture does not imply log concavity of the Wright density. The proof gives continuous families of counterexamples, including parameters with β>0\beta>0.

Source: R. A. C. Ferreira and T. Simon, On the log-concavity of the Wright function, Constructive Approximation 60 (2024), 309–338; DOI, primary author preprint. Tracking and review: issue 100.

Lean verification

Conjectures.FerreiraSimonWright proves the original conjecture false. It defines the Wright function by the source's infinite reciprocal-Gamma series and uses the source normalization for the density. For every sufficiently small positive t, the actual density with alpha = 1 - t and beta = 0 violates log concavity at the three points 1/4, 1/2, and 3/4. The source sets alpha_star(0) = 1, so these parameters satisfy the conjectured threshold.

The proof establishes the infinite-series limit wright (-(1-t)) 0 (-x) / t → x/(1-x)^2 for every 0 ≤ x < 1. Its values at the three selected points are 4/9, 2, and 12; hence the square of the middle value is strictly smaller than the product of the endpoint values. Gamma reflection and a uniform coefficient bound justify the limit by dominated convergence. No series truncation or numerical counterexample is a premise.

Strict compilation with warnings treated as errors, the complete owned-module axiom audit, and independent replay with the pinned stock Lean kernel passed. The audit includes all 25 declarations, eight of them private, and permits only propext, Classical.choice, and Quot.sound. Validation receipt. The combined repository gate remains pending.

The Lean result is a complete disproof of the universal source conjecture. The additional positive-beta families, general slope formulas, and derivative limits in the written proof below are not claimed to be formalized.

Counterexample intervals

For every sufficiently small t>0t>0, each of the following parameter choices satisfies α<α*(β)\alpha\lt \alpha_*(\beta), while its normalized density is not log concave:

Parameters Why the threshold holds Limiting curvature at x=1/2x=1/2
α=1−t\alpha=1-t, β=0\beta=0 The source sets α*(0)=1\alpha_*(0)=1. −16-16
α=1−3t\alpha=1-3t, β=t\beta=t The curvature at zero, divided by t2t^2, tends to 2>02>0. The source proves this sign characterizes the threshold. −448-448

The last column is the limit of (ft′2−ftft″)/t2(f_t'^2-f_tf_t'')/t^2 for the unnormalized Wright density. Multiplication by the positive density-normalization constant preserves the sign. Negative curvature in this convention means that logft\log f_t has a positive second derivative, contradicting log concavity.

Proof mechanism

For fixed a>0a>0 and b≥0b\geq0, the proof establishes

ϕ(−(1−at),bt,−x)t→b+(a−b)x(1−x)2,|x|<1,\frac{\phi(-(1-at),bt,-x)}{t} \longrightarrow \frac{b+(a-b)x}{(1-x)^2}, \qquad |x|\lt 1,

locally uniformly together with every fixed derivative order. Gamma reflection and log convexity give the uniform coefficient bound an+ban+b for n≥1n\geq1. This justifies passing to the curvature limit.

The general limiting curvature at x=1/2x=1/2 is −16(a2+6ab+b2)-16(a^2+6ab+b^2). Its value at zero is a2−2ab−b2a^2-2ab-b^2. Every fixed slope satisfying a2>2ab+b2a^2>2ab+b^2 therefore yields counterexamples for all sufficiently small t>0t>0.

The analytic proof establishes whole intervals. The computation below only checks its exact algebra; it does not supply a numerical substitute for the proof. The separate equivalence between density log concavity and reciprocal Mittag–Leffler convexity is not classified here.

Files and review

Run python3 results/ferreira-simon-wright/verify.py from the repository root. The exact check passes. Separate internal reviews by the root and algebra agents found no mathematical gap; these are not outside peer review.

No earlier resolution of the exact conjecture was located in the sources inspected as of 20 November 2026. The novelty record identifies the inaccessible published full text and the abstract-only screen of a recent related manuscript.