research.md

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

4298 bytes, as of the version from 2026-11-19 01:06 (current). Recorded here, not run.

# Checked research for T761 (build stage)

Instructions: archived ../20261118T213826Z-build-skills/manifest.json; every local/live archived resource was SHA256-verified, with equal bundle digest b5c3c7985b7c9e178b4899d083a9e3bc4284989dff76f05e5b0e52835d47eed2. The exact manifest is also skill-manifest.json.

Checked sources and normalization:

- Roberts–Rodriguez Villegas, arXiv:2109.00027v1, Section 5 “The completely separated case”, (5.2): the MUM condition gives denominator (T-1)^n and Hodge vector (1,...,1); c_3=4. Degree-three totient partitions yield precisely 2.2.2,2.3,2.4,2.6. Section 2 fixes the hypergeometric equation and Section 10 the prime exclusions. Downloaded RRV.html and extracted RRV.txt; URLs and hashes in source-manifest.json.
- DLMF 16.12.2 gives Clausen's square. Substituting a/2,(1-a)/2 gives 3F2(1/2,a,1-a;1,1;z). DLMF 15.8.18 gives the quadratic argument w=(1-sqrt(1-z))/2, first on Re(w)<1/2 and then by the stated continuation. For the upper z-bank, sqrt(1-z)=-i*sqrt(z-1), so Im(w)>0. DLMF 15.4.20 gives the finite gamma quotient at z=1. These passages were inspected through web browsing; direct .tex downloads returned 403 (not missing mathematical evidence).
- DLMF 15.9.24 uses elliptic modulus k. The table uses T25's parameter m=k^2, hence 2F1(1/2,1/2;1;w)=2K(w)/pi in parameter convention. DLMF 16.2 and 16.8.3 support the displayed defining series and differential equation.
- Naskrecki arXiv:1702.07738v2, abstract, Introduction and Section 7: a K3 realization for Phi_2*Phi_4/Phi_1^3. It is not cited as a four-template classification, nor is an integral-cycle comparison claimed. The numerical definition never uses its geometric t-coordinate.
- Independent BCC coefficient derivation: a 2n-step return in three independent sign coordinates has probability binomial(2n,n)^3/64^n. The Green-function discount 8/k gives z=64/k^2. At z=1 the coefficients are O(n^(-3/2)), so the endpoint sum converges. The independent gamma control agrees with the stored interval.

Live discovery: catalogue.json is the one authenticated catalogue for this planning pass; screen.json records already_here and already_asked, with active #346 and originating #261 checked. Metadata was refreshed for plausible matches before the claim. lookup files show the expected endpoint in T393 and no matching family for the nontrivial interior sample. T178 and T348 store Gauss ingredients, not their four motive-indexed squared periods. T679 and T738 have ranks four and five. T731 is a marked ratio, with underlying 2.4 solution at z=t^(-4). T749 is the BCC Green function; its endpoint and T393's BCC Watson integral legitimately share the 2.2.2,z=1 value in distinct contexts. No published table was edited.

Tags: tags.json checks exact spellings and variants. Reused number theory, special functions, special values, period and K3 surfaces. The absent hypergeometric motive tag was added within this draft for analytic/arithmetic invariants of specified hypergeometric motives. K3 surfaces is already registered despite zero published usage. No universal CM, Asai, modular-form or p-adic property is claimed.

Computation: generate.py uses exact QQ inputs and the off-cut quadratic Gauss pullback in Arb at 493 bits. build_checks.py invokes the existing dry_run.py/check.py and independently checks all 88 values by the half-parameter Clausen square (gamma at 1), with 476 minimum component accuracy bits. stored_check.py imports no generator and checks all stored intervals using mpmath at 160/210 digits, 40 direct defining-series checks, 22 elliptic checks, endpoint and deliberately wrong controls. verify.log reports 88/88 reproducibility. Results and versions are JSON artifacts, not an independent review verdict.

Pending: the arithmetic companion is only a proposal, so no nonexistent table link was inserted. The controller/family review should add the (template,z) directed companion link when its actual T-number exists. The single structural API audit warning about four template names is intentional object indexing of one normalized-period quantity, justified by the complete classified family; it is retained for the fresh critic. No unresolved numerical discrepancy is known. Independent scientific and argument-selection reviews remain mandatory.