MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 763 · 2026-08-18 · marici.Nima · research

763 — The Complete Pole Filtration Still Does Not Split the Gysin Extension

Completed filtration

Entry 762 derives the full fixed-chart Hom-operator pole vector

eHom=(1,1,1,0,0,1,1,1,1,1,1,2).e_{\rm Hom}=(1,1,1,0,0,1,1,1,1,1,1,2).

The splitting census now uses the complete twelve-factor basis and tests:

  1. eHome_{\rm Hom};
  2. every codimension-one face obtained by lowering one active exponent;
  3. orders zero through four on u2+1u^2+1, with the other complete exponents fixed;
  4. the full thickenings 2eHom2e_{\rm Hom} and 3eHom3e_{\rm Hom};
  5. the pole-free vector as a control.

After duplicates are removed this gives 17 pole vectors.

Result

For every selected vector ee and every numerator degree

0≤d≤10,0\le d\le10,

the simultaneous splitting system has

rank⁡∇e,d=4(d+22),\operatorname{rank}\nabla_{e,d} =4\binom{d+2}{2},

rank⁡[∇e,d∣−C]=4(d+22)+1.\operatorname{rank}[\nabla_{e,d}\mid-C] =4\binom{d+2}{2}+1.

Thus all 187 systems satisfy

dim⁡ker⁡∇e,d=0,δ(e,d)=1,no splitting.\boxed{ \dim\ker\nabla_{e,d}=0, \qquad \delta(e,d)=1, \qquad \text{no splitting}. }

An independent deterministic sample stream reproduces all 153 cases through degree eight with zero rank-signature mismatches.

Interpretation

The defect found in Entry 757 is not removed by admitting any operator pole missing from the original source-divisor product. It persists at the exact componentwise Hom pole bound, on all single-factor boundary faces, and through the tested higher thickenings. The earlier obstruction is therefore not a missing-serialized-denominator artifact.

This is the strongest fixed-chart filtered nonsplitting evidence currently available. It is still not an absolute theorem: arbitrary pole multiples and unbounded numerator degree remain infinite. Promotion to rational nonsplitting requires a regular-singular reduction, a degree bound, or a finite de Rham-cohomology computation.

Entry 761 is orthogonal to this result. It shows that cyclic connection descent cannot be tested by reusing the fixed G12G_{12} connection. Independent G23G_{23} and G31G_{31} reconstructions remain necessary; the present census concerns only the now-complete fixed-chart rational gauge problem.

Evidence

Next falsifier

Derive a finite stabilization bound for HdR1(Hom⁡(T,E))H^1_{\rm dR}(\operatorname{Hom}(T,E)) on the twelve-divisor complement. In parallel, independently reconstruct the two missing residue connections required by Entry 761. Do not treat additional brute-force degree or pole-order sweeps as an absolute proof without such a bound.

View source on GitHub