argument-selection.md

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

2680 bytes, as of the version from 2026-11-16 11:55 (current). Recorded here, not run.

# Argument selection: even-row Bessel period determinants

The sole parameter n (type Z) is the total Bessel factor count, equivalently the symmetric-power index in Fresán–Sabbah–Yu, arXiv:2006.02702v2, Notation 8.2, (8.3), Theorem 8.4 and Table 1. This is a discrete_index, with infinite natural domain n>=5. No continuous coupling, Bessel order, mass or scale is independently sampled. All functions have order zero and common unit scale, on the positive real integration path.

Set d(n)=floor(n/4)-1 if 4 divides n, and floor(n/4) otherwise. The named quantity is the raw signed determinant of M(2i,n-2i;2j-1), i,j=1,...,d(n), in increasing order, where M(a,b;r)=integral_0^infinity I0(t)^a K0(t)^b t^r dt. The source prescribes the even rows and, at multiples of four, deletion of the final row and column. These internal indices are not free parameters. Both parity tables use odd powers of t. No pi factors, factorials, column signs, absolute values or regularizations are absorbed. n=4 would be a zero-dimensional reduced object and is outside this table.

Store the complete admissible prefix n=5,...,16: dimensions 1,1,1,1,2,2,2,2,3,3,3,3. The cutoff 16 is a noncanonical cost and coverage choice. It covers every residue class modulo four, all initial scalar cases, genuine minors of orders two and three, and reduced orders one, two and three at n=8,12,16. Keeping the full prefix avoids selecting only closed-form or famous cases. Beyond 16, the next four-factor block introduces larger unreduced even minors; numerical integration cost and determinant cancellation motivate stopping at this initial block boundary. This does not assert that 16 is an intrinsic endpoint or that all periods are known critical L-values. Set complete: no.

Convergence is checked before computation: every stored row has b>a, while all columns have r>=1. Near zero the integrand is O(t^r(1+|log t|)^b), and at infinity it is O(t^(r-n/2) exp(-(b-a)t)). The unreduced n=4m boundary row would instead include a logarithmically divergent entry; it is never evaluated.

The generator enumeration, Programs and completeness note must all use n=5,...,16, the same reduction, and the single named E_n. T521 has individual moments, T647 consecutive-row full determinants, T648 vacuum constructions, and T724 odd-row minors. None is this family. Source locations and numerical evidence are preserved separately; sibling values are not independent scientific evidence. The controller must obtain a fresh, fingerprint-bound independent argument review before offering.

Instruction evidence: skill-manifest.json, bundle SHA-256 d6661c2f21613524889fa3a57d3c95f8330344ca62e5a856211c08a282b7dbab.