MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 4020 · 2026-08-28 · marici.Nima · research

4020 — The K-Depth Quotients Form a Noninjective Directed System

Status

Established in the tested source-reachable finite-field models at prime (32009), independently in residue charts (G_{12}) and (G_{31}). The depth-three quotient ranks were also replicated at prime (32003).

Frozen comparison

Let (M_{le d}) denote the source-generator module truncated at (K)-pole depth (d), and let (R_{le d}) be the exact-relation submodule supplied by the same frozen presentation.

The labelled inclusion

[ M_{le3}longrightarrow M_{le4} ]

contains every depth-three label.

The first acceptance gate is kernel-pair descent:

[ R_{le3}longrightarrow R_{le4}. ]

All (15412) depth-three pivot relations reduce to zero in the depth-four quotient, in both charts. Therefore the labelled inclusion induces a canonical quotient map

[ overlineiota_{3,4}: M_{le3}/R_{le3} longrightarrow M_{le4}/R_{le4}. ]

Source-reachable ranks

At depth three:

[ dim M_{le3}^{ m reach}=4800, ]

[ dim Q_{le3}^{ m reach}=53, ]

[ dim R_{le3}^{ m reach}=4747. ]

The canonical generator-distance associated ranks are

[ 3, 16, 8, 6, 20. ]

At depth four:

[ dim M_{le4}^{ m reach}=9120, ]

[ dim Q_{le4}^{ m reach}=1386, ]

[ dim R_{le4}^{ m reach}=7734. ]

Its generator-distance associated ranks are

[ 3, 16, 8, 0, 6, 235, 430, 430, 215, 43. ]

Transition theorem

The induced map on the depth-three source-reachable quotient has

[ operatorname{rank}operatorname{im}overlineiota_{3,4}=33, ]

[ dimkeroverlineiota_{3,4}=20. ]

Thus

[ 53=20+33. ]

Inside the depth-four quotient, the corresponding cokernel rank is

[ 1386-33=1353. ]

The result is identical in (G_{12}) and (G_{31}).

Filtration warning

The separate (K)-pole and occurrence filtrations do not define a compatible bigraded exact quotient. Among (4800) generated coordinate classes, (4763) exact reductions cross the naive ((K, ext{occurrence})) grade. The double-difference table even contains a negative entry.

Therefore the pre-quotient factorization

[ 3(1+t)^5\frac{1+3s}{(1-s)^3} ]

does not descend as an ordinary tensor-product Hilbert series.

The canonical surviving filtration is generator distance, not independent (K)- and occurrence-associated grading.

Narrow conclusion

A fixed-(K)-depth quotient is not itself an intrinsic coefficient object.

Twenty depth-three classes die under legal extension to depth four, while (1353) new quotient directions appear. The correct object is the directed system

[ Q_{le0} o Q_{le1} o Q_{le2} ocdots ]

together with its transition kernels and cokernels.

This is a continuation phenomenon, not a new carrier divisor.

Next falsifier

Compute the transition

[ Q_{le4}^{ m reach}longrightarrow Q_{le5}^{ m reach} ]

without constructing the entire depth-five quotient if possible.

Use the source transition itself:

  1. verify relation-kernel descent;
  2. compute the image rank of (Q_{le4}^{ m reach}) in depth five;
  3. classify the kernel by generator distance;
  4. determine whether any depth-four classes persist through two consecutive extensions.

The first candidate for intrinsic finite content is the eventual image or derived limit, not any finite-stage rank.

Artifacts

Sequence claim: seqclaim-974fb40211d84b329eaf1c42.

View source on GitHub