back to table · edit · history · where entries came from · files · download
3958 bytes, as of the version from 2026-11-20 03:35 (current). Recorded here, not run.
# T770 build research
Selected quantity: B_p(P_i,P_j)=(H(P_i+P_j)-H(P_i)-H(P_j))/2 on the frozen
Cremona models and points, in Sage's MTT normalization. Iwasawa log(p)=0;
H(P)=-2/m^2 log_p(sigma_p(t(mP))/d), t=-x/y, positive denominator d.
No division by log_p(1+p). No L-value, period or Sha input is used.
Checked source locations:
- Sage 10.9 padics.py, padic_height and padic_height_pairing_matrix: the
explicit -2 sign, /2 polarization, multiplier, adjusted precision, and
platform-independent fixed-basis examples on 37a1,389a1,5077a1.
- Harvey arXiv:0708.3404v1 eq.(3), sections 2.2,4,5: the +2 printed sign is
opposite to this convention; multiplier, sigma and precision algorithm.
- PARI manual ellpadicheight, ellpadics2 and official ellpadic.c coverage
source: de Rham components [f,g], g=-log_E(P)^2, ordinary scalar f-cg.
The source constructs rational formal logarithm/Bernardi sigma by
integrations. For c=ellpadics2, log(sigma_p/sigma_B)=-c*z^2/2, so f-cg
gives -2 log(sigma_p/d)/m^2. This derivation fixes the sign independently
of numerical fitting. The same exact model/differential enters both routes.
- Cremona documentation TABLE ONE/TWO: ordinary projective coordinates,
source free-point order, imported generator claim and 990h3 optimality
exception. Pinned allgens source matches the proposal hash. input_check.py
reconstructs the complete finite selection, checks every point equation,
companion model/rank and exact finite-field prime eligibility.
Live discovery: 766 lightweight records including two unrelated offered
zeta3 drafts; no ordinary point-pairing family. already_here and
already_asked refreshed (the latter's title query was empty, not treated as
proof). T412 is rank-one regulator determinants in the same MTT convention;
T64/65/66 and T293 are real regulators, T211 and T670 are Faltings heights.
The 37a1 sample modulo 5^10 matches T412; the 389a1 cross-pairing sample has
no match. Stored exact tag spellings p-adic, elliptic curves, number theory
are confirmed. The absent narrow p-adic height/heights tags do not prevent
using these sound existing tags. No L-function tag is justified by this
point-height calculation. metadata JSON and search/lookup logs are retained.
Frozen input: points.txt SHA-256
049273399057ba95739695c3a848bc2bc1791bee35a4e5d90c8048e11471fc21.
Full source SHA-256 values and URLs are in sources.json and input-evidence.json.
The exact source rank is an imported selection attribute. Independent PARI
descent bounds, checked saturation coverage (none newly claimed), and subgroup
versus full-lattice status are separate ledger fields. Pairings of these
explicit points do not require an independent full-lattice certificate.
The known control bases are checked exactly: 389a1 uses columns (-1,1),(0,1)
relative to the frozen points. 5077a1 uses (0,-1,-1),(0,1,0),(-1,-1,-1),
with determinant -1. The full printed source matrices passed at p=5. p=13
checks bounded integer combinations, retaining P/-P ambiguities. This is a
known-identity exercise, not a uniqueness theorem or a new point discovery.
Computation: compute.py logs resumable OBJECT records. Each nonzero pairing
retains k_p relative digits, k=(144,119,97,90). Sage and PARI are each run at
absolute exponents k+24 and k+44; complete units, precision losses and times
are preserved. Zero balls are never promoted to exact zeros. Actual successful
coverage is reported by the COMPLETE record and numerical-evidence.json, not
assumed from the planned count. Earlier failed import/probe and zero-identity
precision-handling attempts are retained. Independent critique remains pending.
Canonical archive: ../20261120T030205Z-build-skills/manifest.json; local/live
resources verified, bundle b5c3c7985b7c9e178b4899d083a9e3bc4284989dff76f05e5b0e52835d47eed2.
No mathematical convention disagreement remains open. Implementation evidence
and source hashes are not an independent scientific verdict.