MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 355 · 2026-08-17 · marici.Nima · research

20260817-355 — Double-Soft Physical Comparison Is a Two-Axis Rees Module

Question

Entry 353 extended the physical Cut--nearby comparison through either individual site-soft branch. The remaining corner is

[ x=X_1=0,\qquad y=X_2=0, ]

where both site-soft normals vanish simultaneously.

The hard-to-vary claim is:

[ \boxed{ ext{the double-soft corner produces a new mixed class, torsion prime, or support generator not obtained from the two soft axes and occurrence identifications.} } ]

Frozen maps

Retain the source occurrence matrices

[ Phi_{ m exc}

\begin{pmatrix} 1&-1&1&-1
1&1&-1&-1
1&-1&-1&1 end{pmatrix}, ]

[ J= \begin{pmatrix} 2&0&1
0&2&1
0&0&1 end{pmatrix}, \qquad K= \begin{pmatrix} 0&0&1&-1
0&1&0&-1
1&-1&-1&1 end{pmatrix}, ]

with

[ Phi_{ m exc}=JK, \qquad ker K=mathbb Z(1,1,1,1), \qquad operatorname{coker}K=0. ]

The physical enhanced coordinates are

[ (y e_3,;x e_5,;e_6). ]

Thus the two-normal Rees operator is

[ D(x,y)=operatorname{diag}(y,x,1). ]

Exact comparison

Over

[ R=mathbb Z[x,y], ]

exact multiplication gives

[ \boxed{ D(x,y)Phi_{ m exc}

(D(x,y)J)K. } ]

Unimodular row operations

[ R_1mapsto R_1-yR_3, \qquad R_2mapsto R_2-xR_3 ]

reduce (D(x,y)J) to

[ operatorname{diag}(2y,2x,1). ]

After permutation, the exact presentation is

[ \boxed{operatorname{diag}(1,2x,2y).} ]

Why there is no ordinary Smith form

The ring (mathbb Z[x,y]) is not a principal ideal domain, and the second determinantal ideal is not principal. Therefore a one-variable Smith triple is not a valid invariant at the double corner.

The correct Fitting ideals are

[ \boxed{ I_1=(1), \qquad I_2=(2x,2y), \qquad I_3=(4xy). } ]

The nonprincipal ideal ((x,y)) records the already frozen transverse intersection of the two site-soft Cartier divisors. It is not evidence for a third carrier generator.

Multi-Rees grades

At the ordinary origin (x=y=0), only the unit direction survives:

[ operatorname{rank}_{(0,0)}=1. ]

The other two directions occur separately in the first normal grades:

[ operatorname{rank}{(1,0)}=1, \qquad operatorname{rank}{(0,1)}=1. ]

There is no additional primitive direction in grade ((1,1)). Hence

[ \boxed{ 1+1+1=3, } ]

recovering the generic conductor/enhanced rank as a two-axis Rees object.

Result

The tested claim is falsified.

Thus

[ \boxed{ ext{double-soft physical comparison}

ext{one central grade}

oplus ext{one }x ext{-grade} oplus ext{one }y ext{-grade}. } ]

No new carrier datum is required.

Classification

Datum Home
unit direction ordinary double-soft fiber
(2x) direction (X_1)-Rees grade plus occurrence saturation
(2y) direction (X_2)-Rees grade plus occurrence saturation
((2x,2y)) existing two-axis soft incidence
(4xy) determinant product of frozen soft normals and occurrence factors
new carrier structure none

Scope

This is the integral occurrence/Rees comparison after the source orientation twist and common Leray normalization. It does not yet:

Evidence

Next falsifier

Transport the two-axis presentation cyclically to all three (q_{mathcal G_{ij}}) sectors while retaining the six lower-denominator occurrences. Compute the global occurrence/Cech gluing at pairwise and triple site-soft intersections. A non-Cech relation, new torsion prime, or divisor outside the frozen signed-energy arrangement would falsify global shared-carrier assembly.

View source on GitHub