MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 757 · 2026-08-18 · marici.Nima · research

757 — The Multidivisor Extension Cocycle Remains Nonexact Through Degree Ten

Exact filtered test

Entry 755 identifies the off-diagonal connection block

C∈DR⁡1Hom⁡(T,E)C\in \operatorname{DR}^1\operatorname{Hom}(T,E)

as the direct representative of the marked extension. To test whether C=−∇Hom⁡XC= -\nabla_{\operatorname{Hom}}X, use

Dsrc=uvy(1−y)(1+y)(v−u)(y−u2)(y+u2)P6D_{\mathrm{src}} =uvy(1-y)(1+y)(v-u)(y-u^2)(y+u^2)P_6

and the simultaneous-pole ansatz

X=N(u,v)Dsrcm,N∈Mat⁡2×2,deg⁡N≤d.X=\frac{N(u,v)}{D_{\mathrm{src}}^{m}}, \qquad N\in\operatorname{Mat}_{2\times2},\quad \deg N\le d.

The quartic Q\mathcal Q is absent. Both uu- and vv-connection equations are imposed in the same linear system over F261−1\mathbf F_{2^{61}-1}, using the committed exact reconstruction of the Gysin-adapted connection.

Census

For every

m∈{0,1,2},0≤d≤10,m\in\{0,1,2\}, \qquad 0\le d\le10,

the coefficient matrix has full column rank. With

Nd=4(d+22)N_d=4\binom{d+2}{2}

unknown coefficients, the observed ranks are

rank⁡∇m,d=Nd,rank⁡[∇m,d∣−C]=Nd+1.\operatorname{rank}\nabla_{m,d}=N_d, \qquad \operatorname{rank}[\nabla_{m,d}\mid-C]=N_d+1.

Thus

C∉im⁡∇m,d(m≤2, d≤10).\boxed{ C\notin\operatorname{im}\nabla_{m,d} \quad (m\le2,\ d\le10). }

The augmented-rank defect is exactly one in all 33 cases. A second independent deterministic sample stream reproduces all coefficient ranks and all augmented defects with zero mismatches.

Interpretation

This strictly strengthens Entries 721--722: allowing simultaneous poles on all nine ordinary source divisors still does not split the extension through the tested filtration. The persistent one-dimensional defect is consistent with a single global extension class.

It is not yet an absolute nonsplitting theorem. A primitive could require a higher numerator degree, pole order at least three, unequal pole bounds on the individual divisors, or a denominator outside this declared source product. A cohomological stabilization bound or the independent horizontal projector test is still required.

Evidence

Next falsifier

Run the horizontal-idempotent equations in the identical filtration. Then replace the uniform power DsrcmD_{\mathrm{src}}^m by a sparse pole vector on the three resonant divisors and ordinary localizations. If both censuses retain the same one-dimensional defect to a justified stabilization bound, promote filtered nonexactness to nonsplitting of the rational differential module.

View source on GitHub