MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 350 · 2026-08-17 · marici.Nima · research

20260817-350 — Saturated Soft-Support Extension Test at (X_2=0)

Question

Does the final four-master algebraic--elliptic extension acquire a class supported at the soft intersection

[ E_T=X_2=0 ]

that is invisible on the generic total-energy divisor?

Frozen input

Work in the homogeneous chart

[ X_1=1,\qquad u=E_T,\qquad y=X_2=\frac{u+v}{2}-1. ]

The source-defined algebraic kernel generator degenerates at (u=0,v=2) as

[ v_{ m alg}=y^2\bigl(0,1-y^2,2,-2\bigr). ]

The supported test therefore uses the saturated generator

[ widetilde v_{ m alg}=\frac{v_{ m alg}}{X_2^2} ]

before specializing to (X_2=0). No carrier cell, support summand, or fitted normalization is added.

Finite falsifier

  1. Reconstruct the exact rational (u)-connection of the final four-master block.
  2. Extract its logarithmic residue along (u=0).
  3. Conjugate that residue into the regular saturated frame [ (e_6,widetilde v_{ m alg},e_7,widetilde e_8). ]
  4. Expand at the soft normal (v-2=0).
  5. Read the kernel-to-quotient principal part. A nonzero principal part would be a soft-supported extension class on the existing soft carrier.

Result

At both independent finite fields,

[ p_1=2305843009213693951, \qquad p_2=2305843009213693921, ]

with disjoint deterministic reconstruction streams, every rational fit closes and the Laurent orders agree.

The only soft pole in the saturated total-energy residue lies in the elliptic quotient block:

[ operatorname{PP}{X_2=0} R{E_T}

\begin{pmatrix} 0&0&0&0
0&0&0&0
0&0&0&-\frac12
0&0&0&0 end{pmatrix}. ]

The algebraic-to-elliptic extension principal part is

[ \boxed{ operatorname{PP}{X_2=0} \bigl(R{E_T}^{ m ext}\bigr)=0. } ]

A regular finite off-diagonal coefficient survives:

[ operatorname{FP}{X_2=0} \bigl(R{E_T}^{ m ext}\bigr)

\begin{pmatrix} 0&-\frac14
0&0 end{pmatrix}. ]

Thus the test distinguishes vanishing of a supported logarithmic class from vanishing of all finite mixing: only the former is established.

Narrow conclusion

For the final four-master block, after the source-forced (X_2^2)-saturation of (v_{ m alg}), no kernel-to-elliptic extension class is supported at the generic (X_2=0) soft intersection of the total-energy divisor.

The remaining pole is the expected quotient degeneration, and the regular (-1/4) mixing is coefficient-level connection data. This calculation requires no new carrier datum.

Classification:

[ \boxed{ ext{existing soft carrier} + ext{elliptic quotient degeneration} + ext{regular coefficient mixing}; quad ext{no new supported extension detected}. } ]

Scope

This is a modular, generic fiberwise de Rham calculation in the final four-master sector. It does not establish:

Evidence

Next falsifier

Repeat the saturation analysis at the other site-soft branch (X_1=0), then test the complete nine-master algebraic kernel. Any nonzero supported kernel-to-elliptic principal part must be classified as soft-support coefficient data unless it requires an incidence divisor absent from the frozen energy carrier.

View source on GitHub