back to table · edit · history · where entries came from · files · download
3742 bytes, as of the version from 2026-11-20 05:53 (current). Recorded here, not run.
# T772 research (builder-checked leads; independent critique pending)
Quantity and domain: smallest positive real period w1 of dx/(2y+a1*x+a3) for
rank-zero rational elliptic curves, using integral global minimal models.
Parameters N,c4,c6 identify individual curves. The source roster is frozen at
ecdata 25cec5ecfec8b9f016eb1631ac633194c2bed39f; coverage N<=100 is finite only.
Source bytes and hashes are in source-manifest.json; roster.py reconstructs all
284 curves, 76 classes, preserving actual Cremona/LMFDB numbering. 11a1=11.a2.
Inspected primary passages: Cremona data page Tables 1/2 and Generators;
Sage period_lattice.py lines 265-298 and 713-744 (_compute_periods_real);
PARI elliptic documentation members roots/omega and ellrank unconditional
bounds r1<=rank<=r2. Raw copies/hashes: sources/ and source-downloads.json.
The Sage documentation HTML fetch via Python was refused, but its primary
page was inspected through the web tool and actual source downloaded.
LMFDB 11.a2 JSON is a short-precision external control, not a certificate.
Live discovery: catalogue.json (768 tables including own visible drafts),
discovery.json and candidate-metadata.json, extra-metadata.json. T72-T74 are
positive-rank primitive periods; T294 is over real quadratic fields; T765/T766
have weights >=4; T211 is a lattice-area height; T771 is class L(E,1).
No matching rank-zero family was found. Lookup of 1.2692093042795534 found no
hit; 37a1's period found T72, and a CM-period probe found a distinct beta
context, a legitimate repeated value. Registered tags elliptic curves,
number theory, period all apply. No aliases or new tags are needed.
Draft T772 created under zeta3 with the supplied exact #353, contributes-to
#266 message. It is private and will remain unoffered. Source sharing from the
family dossier is research reuse, not an independent numerical conclusion.
The local/live archived skill hashes match; skill-manifest.json is copied
from ../20261120T053812Z-build-skills/manifest.json, SHA256
7629363ce07d2e1340528294fe639e6fc7a5b18da808ec52558687fc1db91534.
All 284 production values passed independent PARI omega comparisons at 832/1024
bits and certified direct defining-integral checks (full coverage, both signs:
147 positive and 137 negative discriminants). Minimum retained accuracies are
594 production bits and 471 quadrature bits. Rank bounds are [0,0] throughout.
The CM control is 32a2, Gamma(1/4)^2/(2*sqrt(2*pi)); 11a1 and 37a1 external
controls and a deliberately doubled period passed/rejected as intended.
No mathematical disagreement remains in builder evidence.
The unscaled integral failed to return a finite enclosure at 77b3. Rescaling
u=D^(1/4)*v and evaluating the denominator as
(2*t-1)^2+(alpha+2)*t^2*(1-t)^2 resolved the peak without relaxing quadrature
error goals. The checker retained 279 completed records under their original
checker/input hashes; the final five complete the roster. See accuracy-assessment.md
for the overstrict internal 600-bit assertion at 99d3 (594 accurate bits still
far exceeds the requested 100 decimal digits). Original attempts are preserved.
Final generator source changes only the docstring's measured precision number;
documentation-change.json verifies unchanged computational AST.
Primary data are archived in full, with exact SHA256; gzip source attachments
preserve those bytes while fitting the per-file limit. Reproduction requires
decompression, described in README.md. Both displayed Programs ran successfully.
The additive upload reported 284 added, zero updated/removed. verify() matched
284/284; independent stored-interval readback and rendered-page/audit evidence
are recorded separately. This is builder checking, not a scientific verdict.