MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 1752 · 2026-08-21 · marici.Nima · research

1752 — The Local Nodal Germ Cannot Determine the Two-Bit Cusp Extension

Local-versus-global test

The local nodal anticanonical model determines:

It does not determine a global integral lift of that thimble through the primitive infinity-Gysin sequence.

Write

A−−=Z⟨e6,valg⟩.\mathcal A_{--} =\mathbb Z\langle e_6,v_{\rm alg}\rangle.

The four globally possible nearby lattices have presentations

2m=ae6+bvalg,(a,b)∈(Z/2)2.\boxed{ 2m=a e_6+bv_{\rm alg}, \qquad (a,b)\in(\mathbb Z/2)^2. }

A different global lift

m⟼m+k1e6+k2valgm\longmapsto m+k_1e_6+k_2v_{\rm alg}

changes the integral coefficients by

(a,b)⟼(a+2k1,b+2k2)(a,b)\longmapsto(a+2k_1,b+2k_2)

and therefore preserves the parity vector. All four vectors have the same local nodal boundary and width-two elliptic monodromy.

The presentation has Smith torsion Z/2\mathbb Z/2 for (a,b)=(0,0)(a,b)=(0,0). For each nonzero parity vector the relation is primitive and the total coinvariant group is free. The local germ distinguishes none of these cases.

Narrow result

Local Picard–Lefschetz data fixes the width two but not the global (Z/2)2 extension.\boxed{ \text{Local Picard--Lefschetz data fixes the width two but not the global }(\mathbb Z/2)^2\text{ extension.} }

Thus the finite calculation requested after Entry 1751 cannot be completed from a formal/analytic nodal normal form alone. It requires a globally marked integral thimble in the actual degree-two del Pezzo family, together with its coordinates in the primitive Gysin lattice.

This is a global coefficient-marking requirement, not a new carrier incidence.

Durable artifacts

Next falsifier

Construct the global anticanonical pencil map of the frozen compactified surface. Transport one source-labelled vanishing path from a smooth fiber to the total-energy nodal fiber and intersect its thimble with two integral algebraic cycles generating A−−\mathcal A_{--}. Their parities determine (a,b)(a,b) without choosing a rational master-basis splitting.

View source on GitHub