Independent review of López Pouso–Segura's Bessel-weight conjectures

What Lean checks

Lean proves all three signs, strict monotonicity and exact limits conditionally on the stated finite mixture, phase ODE, endpoint, zero/weight asymptotics and two resolvent expansions. It also proves the reference kernel integral and the dominated-convergence and comparison steps linking those inputs to the limits. The Bessel identification, special-function inputs, reference Euler–Maclaurin remainder and full repository gate remain open.

Reviewed by the algebra agent on 21 November 2026. Draft reviewed: agent-analysis/lopez-pouso-segura-bessel/PROOF.md.

Verdict: no substantive mathematical gap found. This is an independent proof review, not an independent historical-priority certification.

The original Springer article was read at equations (16), (18), and (19), and Conjectures 2, 5, and 8. Its definitions and index ranges agree with the draft. The separate uniform Gauss–Legendre assertions are correctly left outside the claimed result.

Checks

  • Nicholson's integral in DLMF 10.9.30 gives the coefficient 4/pi^2 after the stated change of variables. Squaring and using the beta integral gives the positive measure coefficient 16/pi^5. Its mass is 4/pi^2 and its first moment is 1/pi^2.
  • The branch of the phase starts at -pi/2; its first positive Bessel zero corresponds to phase coordinate one. The Wronskian gives lambda'(k)=4a_k. Differentiating the inverse-phase ODE gives precisely lambda'''=(pi^6/2) H (H^2)'' > 0.
  • The trapezoidal expression for 2 C_i and midpoint expression for E_i-E_(i-1) have the asserted signs and constants. The initial interval is valid because lambda'(0+)=0.
  • In the Green identity, the lower endpoint contributes -1 to r(phi F_s' - F_s phi'), so the integrated boundary difference is +1. This yields the coefficient 1/(lambda_k+s^2) with the correct sign.
  • The difference series has coefficients O(k^-3), hence uniform convergence on [0,1] from |J_0|<=1. The limit at zero is log(s/2)+gamma+K_0(s)/I_0(s). Differentiation yields the factor 4s in the resolvent identity.
  • McMahon's expansion DLMF 10.21.19 and the Nicholson-integral amplitude expansion cancel the reciprocal term in a_k, leaving an absolutely summable O(k^-3) correction. The denominator replacement contributes O((1+log s)/s^2) after the prescribed multiplication by s^4.
  • In the shifted Euler–Maclaurin calculation, B_2(3/4)=-1/48, so the correction to the scaled sum is +1/(96A). Consequently the second resolvent coefficient is -1/16+pi^2/192 and the renormalized sum is 1/16-pi^2/192. The integrable fourth derivative gives the required O(A^-3) sum remainder.
  • The arithmetic limits are C_i -> pi^2/12 and E_i -> -pi^2/6. The identity expressing K_i as the a_j-weighted average of C_(j-1) is exact and proves positivity, strict increase, and the limit.
  • The additional circle-constant consequence agrees with the measure identity and tends monotonically to 1/4; it does not establish the separate finite-degree circle conjecture.

Citation refinement

DLMF 10.23.18–20 supplies a direct Fourier–Bessel expansion theorem: weighted integrability and local bounded variation imply equality at each interior point. The smooth difference F_0-F_s satisfies these hypotheses. This provides a precise source for the interior equality before the uniformly convergent series is extended to zero. It can replace or supplement the draft's general reference to Hilbert-space completeness.