back to table · edit · history · where entries came from · files · download
3324 bytes, as of the version from 2026-11-17 00:36 (current). Recorded here, not run.
# Argument selection for Primitive Frobenius polynomials of Fermat varieties
The ordered parameters are n (even dimension at least 2), d (degree at least 3), and p (a rational prime not dividing d). These are discrete indices of the fixed global model X_1^d+...+X_(n+1)^d=X_0^d. The polynomial variable T is not an input. There are no continuous or probability parameters.
The default enumeration contains precisely 48 triples in increasing (n,d,p) order: all n=2, 3<=d<=6, primes p<=31 not dividing d (10,10,10,9 rows by degree), and n=4,d=3, all primes 5<=p<=31 (9 rows). No entry is pruned for a zero trace, repetition, splitting behavior or simple factorization. Completeness is no for the natural infinite family.
Dimension n: discrete_index. Surfaces give the first positive even dimension and connect the rational cubic, quartic K3 and general-type surfaces to their complex Pham periods and linear-cycle invariants, already in T722/T723. Cubic fourfolds broaden dimension without an expensive high-degree polynomial. Dimensions above four remain uncomputed. This is a practical bound, not an intrinsic endpoint.
Degree d: discrete_index. All consecutive degrees 3 through 6 are retained for surfaces, covering odd/even and prime/composite degrees. Primitive polynomial degrees are 6,21,52,105. The modest degree-105 excess over the skill's approximate degree-100 starting point includes the first even composite general-type surface. The fourfold slice stops at the cubic (degree 22), because the next degrees produce degrees 183,820,2605. These unequal ranges address meaningful breadth and polynomial readability; the cutoff is noncanonical, not famous-only pruning.
Prime p: discrete_index. A prime is an actual good-reduction place of this rational model, not a relabeled real input. Every good prime <=31 is retained for surfaces, including nonsplit regimes. Cubic fourfolds use p>3 to respect the characteristic hypothesis of Brünjes's Section 4 setup before Theorem 4.6; nonsplit primes are treated over their quadratic splitting field and descended by character orbits. The good reduction at p=2 is deliberately uncomputed in that slice; p=3 is bad. The upper bound 31 is a noncanonical initial reference extent. The largest orbit field is 23^4=279841. All p=23,29,31 entries are retained without fitting a recognition hypothesis.
Sources inspected: Griffon, arXiv:1612.08721v2, Definition 2.2, Theorem 2.3 and Remark 2.4 (surface Jacobi factors, arbitrary good primes and coefficient twist); Brünjes, arXiv:math/0301186, Section 4 setup, Definitions 4.3/4.4, Theorem 4.6 and Example 4.7 (character lines, split-field Frobenius and cubic-fourfold control). Kloosterman, arXiv:math/0703120v2, Lemma 6.1 and Definition 6.8/Corollary 6.9 give supporting character-action context, with a different complement/inverse convention. Live definitions of T722/T723 establish the shared model, not independent arithmetic evidence.
The equation's negative X_0 coefficient is retained in every field. Primitive removes only the ambient hyperplane-power class, not all algebraic classes. No transcendental projector or cycle-lifting assertion is made. The archived skill manifest is ../20261117T002128Z-build-skills/manifest.json (bundle SHA-256 984965681e2f8925c511a121c118efcde59581c7b0d906cac8b641bf6925e55a).