argument-selection.md

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

3709 bytes, as of the version from 2026-11-20 07:17 (current). Recorded here, not run.

# Argument selection for T773

Quantity: the single real bilinear pairing B_E(P_i,P_j), with B_E(P,P)=H_E(P)
and H_E(P)=lim 4^(-n) log(max(|numerator(x(2^n P))|,|denominator(x(2^n P))|)).
This is the larger Cremona/BSD normalization, twice Silverman's half-height.
E is the exact reduced global minimal Weierstrass model over Q. The stable
parameter order is N,c4,c6,i,j. All rational inputs are exact.

The natural family allows any such curve and explicitly fixed rational point
catalogue. The finite reference selection is source curve number 1 of each
ecdata isogeny class in commit 25cec5ecfec8b9f016eb1631ac633194c2bed39f,
allgens/allgens.00000-09999: source rank 1 with N<=175, source rank 2 with
N<=2000, and source rank>=3 with N<=10000. These disjoint budgets cover 60,
97 and 1 curves, respectively. All r(r+1)/2 upper-triangular components are
included, giving 357 entries. No curves in a selected slice are cherry-picked.
The first r listed projective points are the free points in original order;
torsion follows them. Ordinary [X:Y:Z] means (X/Z,Y/Z).

N identifies conductor, a discrete arithmetic measure of complexity. Its bounds
are explicitly noncanonical practical reference budgets, wider for rank two
and higher to include multi-point interactions rather than only rank-one
heights already related to regulators. The purpose is exploratory recognition
and comparison of specified low-conductor point lattices across arithmetic
ranks. This is a reference_sample with policy: purpose as stated; bounds as
stated; coverage rationale is broad low-conductor isogeny-class coverage and
all matrix components, not a requested count; noncanonical_acknowledged=true,
notation_independent=true, sources_checked=true. No arbitrary continuous grid
or independent coupling is present.

c4 and c6 jointly identify the individual curve and are determined by its exact
model. They are arithmetic_structure parameters conditional on the chosen
source curve, not two additional independently swept ranges. All selected
models in a conductor slice are retained under the source number-1 rule; this
is a practical common reference, not a claim that this curve always has a
universal distinguished role. Exact coefficient lists and both Cremona and
LMFDB labels are in points.json and points.txt.

i and j are discrete_index parameters selecting the actual frozen points,
1<=i<=j<=r. Their finite bounds exhaust the upper triangle; symmetry recovers
the other half without a quantity selector. Source rank r is explanatory
metadata, not a second independent free parameter. Diagonal point heights
and signed off-diagonal pairings are entries of one bilinear object. All 158
source-derived models and ordered lists match live T770 metadata; no sibling
numerical output was used. Saturation and full-lattice index are not assumed.

Sources inspected: Cremona's ecdata documentation, TABLE TWO: GENERATORS
(original free-point order, torsion last and ordinary projective examples);
Sage ell_point.height definition and implementation (normalization and cache),
non_archimedean_local_height and archimedean_local_height; PARI ellheight and
ellheightmatrix documentation (polarization); exact frozen source and label
crosswalk. Source file hashes are in source-manifest.json, and the matching
canonical skill bundle is recorded in skill-manifest.json.

Coverage is incomplete in the natural family: other isogenous curves, larger
conductors, higher ranks outside this file, other point choices and non-basis
catalogues are absent. The generator, Programs, completeness clause and point
attachments describe the same selection. The independent argument-selection
verdict is reserved for a fresh critique session.