back to table · edit · history · where entries came from · files · download
3299 bytes, as of the version from 2026-11-16 12:25 (current). Recorded here, not run.
# T724 research (build, not independent critique)
## Checked sources and conventions
- Fresán–Sabbah–Yu, arXiv:2006.02702v2, Notation 8.2, (8.3),
Theorem 8.4/Table 1: odd I0 exponents, odd integration powers,
dimension floor((n+1)/4); delete the final row/column when 4 divides n.
Source HTML archived as fsy.html. It fixes an actual raw determinant;
abstract critical periods have rational equivalence ambiguity.
- Zhou, arXiv:2012.03523v3, (1.21), Proposition 4.5: reflection between
odd/even minors for odd factor count. Equations (4.37)–(4.38):
O5 is Gamma(1/15)Gamma(2/15)Gamma(4/15)Gamma(8/15)/(240 sqrt(5)).
Archived zhou.html. No general L-value identity is asserted here.
- Bailey–Borwein–Broadhurst–Glasser, arXiv:0801.0891v2, (7), (23):
the two nonzero moment controls and O3=pi/(3 sqrt(3)). Archived bbbg.html.
- DLMF 10.32.1 and 10.32.9 inspected with web tools: positive integral
representations yield I0(t)<=exp(t), K0(t)<=sqrt(pi/(2t))exp(-t),
and K0(t)<=2-log(t) for 0<t<=1. The generator derives both tail bounds.
- DLMF 13.6.10 inspected: K0(t)=sqrt(pi)exp(-t)U(1/2,1,2t).
Actual Sage 10.9 native bessel_K loses precision around t=150..200
at 525 bits. hyper-probe.py checked the U formula against 2200-bit
native Bessel evaluations at 1,20,50,100,150,200,300; all overlap.
The integration callback rejects balls leaving Re(t)>0.
The proposal reports a published typesetting discrepancy. This builder
uses the explicit v2 formula and Zhou's independently consistent dimensions;
it did not inspect the published PDF and makes no new claim about that PDF.
## Live discovery
catalogue.json, screen.json, preclaim.json and lookup-extra.json preserve the
live screening. T521 contains individual moments, T647 full consecutive-row
determinants, T648 vacuum constructions. Their full definitions and current
metadata were inspected. An initial unvalidated 1.048... probe was not used in the decision.
Sample digit lookups at O3, O5, O6 find these scalar
contexts, not the odd-row family. No duplicate family or title collision was
found. Four older drafts were left untouched. The proposed Feynman integrals
tag was absent; this table uses the four verified existing tags instead.
T720 defines the scalar coincidences O6=B5(1), O8=B7(1). No sibling values are
numerical inputs. The U formula and endpoint bounds are source checked;
T720-generate.py is reused only as a computation-component lead.
## Selection and evidence
argument-selection.md justifies all 13 n values and the practical cutoff 16.
The core and all six task-relevant references were read from the coordinated
local archive only; complete archive hashes matched the live version.
See skill-manifest.json, bundle SHA-256
`d6661c2f21613524889fa3a57d3c95f8330344ca62e5a856211c08a282b7dbab`.
All 13 independent mpmath computations at 130 working decimal digits passed,
as did the two scalar closed forms, three reflection identities and negative
controls. Actual results are in check-results.json. Stored readback, audit and
presentation findings are recorded in the build handoff. No independent scientific verdict is issued
by this build. A future even-row table should link T724; no new sibling was
created here. The wanted request #263 remains only partially addressed.