research.md

back to table · edit · history · where entries came from · files · download

3033 bytes, as of the version from 2026-11-18 14:14. Recorded here, not run.

# Laurent coefficients of the secondary zeta function at 1

Checked: Arias de Reyna 2006.04869v1 introduction (1) defines the upper-half
zero variable alpha=(rho-1/2)/i without RH. Its stated pole list puts the next
pole after s=1 at s=-1. Kawalec 2403.15741v2 section 4, (64)--(65), (69),
(71), defines the regular part and C_n as derivatives, hence c_n=C_n/n!.
Its Table 9 supplies independent published agreement controls, not certificates
or a reason to impose RH. The ordinate harmonic constant is not tabulated.
Voros 1403.4558v2 (13)--(19), (25) fixes the confluent and central-xi conventions;
Voros Xi=2*xi, whose factor cancels in logarithmic derivatives.

Live discovery saved in catalogue.json and discovery.json: no matching Laurent
family or draft; api/lookup of C0,C1,C2 returns no hits. already_asked returns
proposal #342. T752 is the separate special-value companion. T733's current
definition gives H_heat(t)=Theta(4*pi*t)/2. T36 is the expansion of xi'/xi at
1, not this spectral Laurent family. Current full documents and candidate
metadata are saved. Only #275 is an originating wanted issue; #342 is the
proposal and #329 is contextual previous heat work. The initial edit cites
the exact #342 and #275 links, without importing request history into prose.

Live registry: Taylor series, L-function and spectral zeta functions exist.
Laurent series and checked case variants do not; this justified new tag names
expansions with principal parts. Taylor series separately describes the regular
part. L-function identifies the Riemann-zeta divisor context, without asserting
an Euler product for the secondary function. No material tag departure.

Generator route: complex FLINT A-P+E-S balls, then Cauchy extraction on |z|=1/2.
An interval covering of |z|=7/8 bounds the analytic regular part; the geometric
alias sum supplies an explicit truncation error. The source code and upstream
hashes are copied from the shared family computation, not sibling numerical
values. FLINT documentation explicitly states unconditional rigorous tails.
Boundary-radius 1 was rejected because its interval arcs crossed the removable
s=0 singularity of an internal decomposition, producing NaN; 7/8 avoids it.
This is an evaluator-domain issue, not disagreement about the mathematics.

Independent build check planned: Mellin transform of the central-xi logarithmic
derivative, after subtracting its large-t logarithm, with moments differentiated
analytically. This replaces the dossier's suggested heat-Mellin route: it uses
different arithmetic and representation from A-P+E-S and covers all 21 orders.
Quadrature/depth/precision agreement will be labelled heuristic checking;
certification rests on the generator's complete ball and alias bound.

Source snapshots: arias.html, kawalec.html, voros.html, flint-doc.html.
Larger artifacts and actual SHA-256 hashes will be indexed in
evidence-manifest.json. See argument-selection.md and skill-manifest.json.
Build checks are not the required fresh independent scientific verdict.