research.md

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

3303 bytes, as of the version from 2026-11-15 23:03 (current). Recorded here, not run.

# Research for T718

Checked sources (copies and SHA-256 hashes in sources/manifest.json):
- Goldmakher, Theorem 1.1, PDF page 2, identifies the primitive quadratic
  character chi_D of conductor |D| and gives the fundamental-discriminant
  enumeration immediately afterward. This checks the whole D domain.
- Zagier, *Modular Forms of One Variable*, Utrecht 1991, Appendix A4,
  printed page 60 (PDF page 62), states the quadratic functional equation
  with exactly (q/pi)^(s/2) Gamma((s+a)/2)L(s,chi), in both parities.
- Lagarias, arXiv:math/0404394v4, equations (1.3)-(1.5), Lemma 2.2,
  Theorem 2.2: checked derivative indexing, factor n, generating functions,
  symmetric zero summation and the real/self-dual specialization. A finite
  computed prefix is not evidence for a universal positivity assertion.
- DLMF 25.15.3 gives the finite Hurwitz decomposition. Its common pole at 1
  is removed analytically because the nonprincipal character sum is zero.
- Kedlaya, sections 6.1 and 6.2, supplies the two theta-Mellin parity formulas.
  numerical-evidence.md derives the split integral and both explicit tails.
- FLINT's documented Taylor jets are factorial-normalized. The runtime probe
  checks actual python-flint 0.9.0/FLINT 3.6.0 operations. Production directly
  uses deflated Hurwitz-zeta series; the checker uses neither zeta nor L-jets.

Normalization settled: D is the fundamental discriminant, not the radicand;
D=1 is excluded. Order n starts at 1 and parameter order is (D,n). Signed
Li lambda_n is stored, not lambda_n/n. No s(s-1) factor is used. An alternative
completion multiplied by (q/pi)^(a/2) leaves Li coefficients unchanged but
would scale the proposed central-series companion; do not transfer that
constant silently. No unresolved source convention remains in this build.

Live discovery: catalogue.json is the one lightweight planning-pass cache;
discovery.json and preclaim.json preserve screen.already_here, already_asked,
text searches, metadata refreshes, tags, and three API lookups. No duplicate
was found. The selected proposal is #319, contributing only partially to #274;
those exact links are in every revision's edit message, outside table prose.

Related original tables, inspected as context rather than numerical oracles:
T31 zeta Li coefficients; T4 Dirichlet zeros; T142 Gauss sums and root number;
T145 uncompleted positive-integer values; T128 L(1) residues; T230 quadratic
Euler-Kronecker constants. Search/catalogue supplies the actual links. The
central-series table in #319 is a planned companion, not built here.

Numerical facts actually checked: all 520 independent theta comparisons,
all 26 exact Conrey matches and primitive-character checks, all stored interval
containments, known-answer controls and failure controls. Production minimum
accuracy 479 bits; checker minimum 433 bits. File hashes are in
attachment-readback.json and evidence-manifest.json. sources/lagarias-passages.txt
and sources/zagier-passages.txt preserve source locations without relying on
another worker's transcript. These builder checks are not the required fresh
scientific or argument-selection verdicts.

Instruction archive: ../20261115T193637Z-build-skills/manifest.json;
matching full bundle d6661c2f21613524889fa3a57d3c95f8330344ca62e5a856211c08a282b7dbab.