Reviewed on 2026-11-21. This is an internal mathematical review, not external peer review or proof-assistant verification.
Verdict
The written argument proves the complete published Conjectures 2, 5, and 8 of López Pouso and Segura, with strict monotonicity. No mathematical gap was found. The three related assertions should be recorded as one named result package and one campaign count. The separate finite-degree Gauss–Legendre conjectures are not established by this proof.
Source comparison
The published article, DOI https://doi.org/10.1186/s13660-025-03283-w>, was inspected directly. Equation (16) and Conjecture 2 agree with the manuscript's indexing of the trapezoidal-error sequence. Equation (18) and Conjecture 5 agree with the midpoint-error sequence, including the limit minus pi squared over six. The nearby source prose contains a conflicting denominator; the numbered conjecture and numerical value identify the intended constant. Equation (19) and Conjecture 8 agree with the weighted first-moment expression. Appendix D establishes finiteness of the first limit, not its conjectured value or monotonicity.
Analytic checks
Nicholson's order-zero integral has factor eight over pi squared. Substituting twice the square root of the argument times the hyperbolic sine gives exactly the manuscript's factor four. Squaring and applying the beta integral gives the stated positive measure. Its mass is four over pi squared and its first moment is one over pi squared. Differentiation is locally dominated, and the second derivative of the squared auxiliary function is strictly positive because the measure is supported at positive values.
The phase starts at minus pi over two, reaches pi over two at the first positive zero, and increases strictly by the Wronskian. Thus the positive zeros occur at the claimed integer phase values. Both the inverse-phase differential identity and its third derivative have the stated factors. The endpoint value and first derivative tend to zero. Taking limits preserves convexity at zero, and strict convexity remains on every nonempty interval.
The first sequence is exactly half the composite trapezoidal error for the inverse-phase derivative. The difference of consecutive terms adds one positive local error. The second sequence's consecutive difference is exactly a negative midpoint-rule error. These facts prove signs and strict monotonicity for all indices.
The radial Green-function computation has the correct boundary sign: integration by parts contributes one from the singular endpoint. Subtracting the zero-parameter Green function makes the Fourier–Bessel coefficients summable, of order the inverse cube of the index. The bound on the order-zero Bessel function gives uniform convergence on the closed interval. Hilbert-space completeness therefore gives equality at the endpoint as well as in the interior. Differentiation with respect to the positive parameter is locally uniform, so the resolvent identity follows.
The leading reciprocal terms in the zero and amplitude asymptotics cancel. This is essential: it makes the renormalized weight errors absolutely summable. Dominated convergence handles their contribution to the scaled resolvent difference. The denominator perturbation is bounded by splitting the displayed sum at the large parameter and tends to zero after scaling.
The shifted Euler–Maclaurin correction has the required sign: the second Bernoulli polynomial at three quarters is minus one over forty-eight. The rational test function and its first four derivatives have the decay required by the remainder estimate. Subtraction gives the renormalized sum and both conjectured limits. Recomputing the arithmetic sums reproduces the constants.
Expanding the squared sum of weights proves the exact first-moment weighted-average identity. Every added value is larger than every preceding value, so the average increases strictly. Divergence of the total weight and convergence of the averaged sequence imply the claimed limit by a finite-prefix and tail argument.
The additional circle-constant monotonicity follows from the same positive measure. The manuscript correctly stops short of treating it as a proof of the uniform finite-degree circle conjecture.
Limits of this review
The proof uses classical Bessel identities, Fourier–Bessel completeness, and asymptotic expansions. Their exact source locations belong in the bibliography. This review checks the mathematical deductions; it does not certify worldwide priority. The separate source audit records the literature search and its limits.