Answers are grounded in this paper and cite their sources.
Fixed-integrand quadrature error bounds do not describe what happens when a parameter-dependent endpoint singularity approaches the integration interval as the Gauss–Legendre order increases. We study the ordinary, unmodified rule for nearly singular kernels on [0,2], under the coupled scaling δ ≍ 𝑁−4 , where 𝑁 = n + 1/2. Under explicit joint C 1 regularity and positivity conditions, the signed quadrature error has a locally uniform limit determined by endpoint data and a convergent Bessel-zero series. Analytic tail bounds and outward-rounded interval enclosures establish precisely one simple limiting zero in each of two disjoint parameter intervals. With additional C 1 regularity of the coefficients frozen at the singular parameter, we derive a next-order 𝑁 −2 correction comprising a canonical contribution and two endpoint-slope terms. A lacunary Legendre-polynomial construction shows that joint C 1 regularity alone does not guarantee this correction. The next-order formula determines the displacement of any selected sequence of exact finite-order error-cancellation roots near either limiting zero, without assuming uniqueness at finite quadrature order. For a conservative bipolar electrostatic oscillator, the endpoint-slope terms vanish, leading to an explicit 𝑁 −6 correction to its cancellation amplitude. Independent numerical checks of several kernel families and the oscillator complement the analytic results. The conclusions characterize signed-error transitions and isolated cancellation scales for the classical quadrature rule rather than proposing an alternative integration scheme.
No takes yet. Share an insight, caveat, or question.
Shu et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: