argument-selection.md

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

3545 bytes, as of the version from 2026-11-16 11:10 (current). Recorded here, not run.

# T722 argument selection

Quantity: the signed complex scalar integral of the weighted Euler form eta_a
over the single marked Pham cycle product(1-T_j) Delta on
X_1^d+...+X_(n+1)^d=X_0^d. This is one named quantity, not a period-matrix
entry selector or a list of Hodge invariants. Ordered coordinates, positive
roots on Delta and dt_1 wedge ... wedge dt_n fix the normalization.

All three parameters have category `discrete_index`:

* `n` (Z) is the even complex dimension, naturally n>=2. Select n=2 and 4:
  surfaces provide geometric controls and fourfolds broaden the middle-period
  context beyond curves. This practical cutoff leaves n>=6 uncomputed.
* `d` (Z) is the defining degree, naturally d>=3. For each selected dimension
  include every degree 3 through 6. This consecutive small-degree range covers
  prime and composite degrees, the quartic K3 surface, the cubic fourfold and
  the sextic Calabi-Yau fourfold. The bound is a noncanonical reference budget
  choice, not a theorem or an application claim for each entry. It keeps every
  monomial-character orbit in each selected model inspectable. Breadth across
  fourfolds takes priority over deeper surface-only coverage.
* `a` (Symbolic) is an integer character tuple, encoded with periods, of length
  n+1, with 1<=a_j<=d-1, nondecreasing, and sum(a_j) not divisible by d.
  Each model includes every such tuple. These are all permutation orbits of
  the affine character labels; the distinguished homogeneous coordinate is
  not permuted. Every filtration grade present in these models is retained.
  This is a genuine discrete label and has no continuous coupling beside it.

Checked support: Movasati's author manuscript dated 27 April 2021, Chapter 15,
equations (15.2), (15.4)-(15.8), Propositions 15.2 and 15.3, gives the affine
models, integral cycles, weighted Euler forms and weight/Hodge filtration.
Duque Franco and Villaflor Loyola, arXiv:2112.14818v2, Section 2,
Theorem 2.1, gives the projective character decomposition and Hodge grades.
These sources support the family and constraints; the finite dimension/degree
bounds are the explicit coverage choice above, not attributed to a theorem.

The counts independently enumerated by ordered tuples and by multisets agree:

| n | d | representatives | ordered classes |
|---|---|---|---|
|2|3|2|6|
|2|4|8|21|
|2|5|16|52|
|2|6|29|105|
|4|3|4|22|
|4|4|17|183|
|4|5|44|820|
|4|6|106|2605|

There are 226 representatives and 3814 ordered classes. The count of ordered
classes per model is ((d-1)^(n+2)+(d-1))/d. A root-of-unity filter proves this:
among m=n+1 nonzero residues the zero-sum count is
((d-1)^m+(d-1)(-1)^m)/d; subtract from (d-1)^m, with m odd.

No curve rows, integral-A boundary classes, phase-translate duplicate rows,
famous-only character pruning, statistical quantiles, arbitrary meshes,
reference_sample groups or hidden selectors occur. The natural family remains
infinite and `complete: no`. Generator enumeration, Programs and completeness
metadata specify the same eight models. The stored entries block is about
59 KB as JSON, below the 160 KB working target. Every row records 100 decimal
digits per nonzero component. Precision does not determine the range.

Canonical instruction archive: ../20261116T021518Z-build-skills/manifest.json;
core SHA-256 c440c9e106fe2752785dc7189021172e6cf0eb91f89673462ba43e49d57b971f;
matching complete bundle d6661c2f21613524889fa3a57d3c95f8330344ca62e5a856211c08a282b7dbab.
This is the builder's rationale, not an independent argument-review verdict.