research.md

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

4231 bytes, as of the version from 2026-11-18 20:02 (current). Recorded here, not run.

# T757 research and handoff leads

Checked: Suzuki, Aspects of the screw function corresponding to the Riemann
zeta-function, arXiv:2206.03682v4: Theorem 1.1(3), (1.13), (1.15), Theorem
8.1 and (8.1),(8.3). This is the sequence of raw moments of Psi=-g with
weight exp(-t/2)/4; n>=0, no n! division. Absolute convergence is unconditional.
The determinant inequalities and positive-measure interpretation need their
stated all-order hypotheses. No RH assumption is used in generation.

Checked independently from the original source: DLMF 25.5.13–14 supplies the
theta-integral completed-xi representation. FLINT's native Taylor-series docs
and actual Sage operations confirm the zeta deflation convention and ball
zeta/gamma/log series. Runtime probes are retained. The local environment has
Sage 10.9 and FLINT 3.4.0; python-flint is absent but not needed.

Live discovery: one authenticated lightweight catalogue pass (753 entries),
already_here/already_asked, synonym searches and fresh metadata found no table
of these moments. The only matching current proposal is #344. T756 is the
function-value companion, already offered by its controller; no edits were
made to it. T31 holds Li's lambda_n, and T36 holds sigma_n and sigma_n/n in
their checked definitions. Neither is this transformed family. api/lookup
found the expected mu_0 coincidence in T31/T36 and no hits for mu_1,mu_2,
mu_10,mu_50. The selected origins are proposal #344, contributes to #274
only; the latter does not close that wanted issue. Its source discussion was
read, including the proof-facing experiment and false-positive-control request.

Tags checked in the live registry: L-function, number theory, moments,
integral, Taylor series, screw functions. Singular/case screw variants were
absent; the exact plural tag exists from the function companion. All six
suggestions apply and are retained. None implies unconditional positivity;
no eigenvalue, polynomial or determinant tag is asserted for a moment table.

Selection: every n=0,...,50, both parities, one genuine discrete index.
argument-selection.md explains the practical cutoff and absence of couplings.
No new input families, table splitting, quantiles or statistical critical
values were introduced.

Computation: generate.py uses the pole-removed completed product and the
sigma-coordinate finite transform. independent_check.py uses bounded theta
quadrature, a fresh Li recurrence and Suzuki (8.3). All 51 independently
computed balls lie strictly inside the stored 100-digit decimal intervals.
Full package verification also matches 51/51, and the displayed example ran.
These are builder checks, not a fresh independent critic's verdict.

Proof-facing experiment: changing the moment Gram matrix to the exact Li
polynomial basis only rescales its determinant and preserves inertia. A signed
control has positive diagonals yet determinant -1/8, so diagonal positivity
fails. Native complex-ball matrices validate all development orders 0..10 and
withheld orders 11..20 at high precision. Generic real-ball attempts lacked
useful precision; their sources/results are preserved as failures. No new
positive decomposition or theorem was found. See numerical-evidence.md and
basis-results.json; this experiment creates no additional table.

Larger artifacts in this directory: suzuki-v4.html, dlmf25-5.html,
flint-series.html, wanted274.json, catalogue.json, discovery.json, lookup.json,
production.json, theta-checkpoint.json, check-results.json, basis-results.json,
stored-document.json, rendered HTML/screenshots, versions.json. Their exact
SHA-256 hashes are indexed by evidence-sha256.json. The coordinated skill
manifest is ../20261118T181011Z-build-skills/manifest.json and attached as
skill-manifest.json. Handoff preserves these paths rather than copying a
sibling table's conclusions as independent evidence.

Remaining controller work: fresh independent scientific critique and
argument-selection review, then any warranted bounded repair and family
consistency review. Future determinant companions have no assigned T-number
yet; add their directed links only once real drafts exist, under their own
authorization. T757 currently links T756/T31/T36; no unrelated tables changed.