Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
159 changes: 159 additions & 0 deletions research/quinary-resonant-disjointness-repair/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,159 @@
# Repeated-root disjointness repair in characteristic five

## Statement and source boundary

Let F be any field of characteristic five. For h=4k with k≥1, let A_h be
indexed by the nonempty proper subsets of [h], with entry1 exactly when
the two subsets are disjoint. This is the same literal source matrix as
Ali Romman's PR4962 and Nick Petalas's PR5173, now in the repeated-root
coefficient characteristic. Put B_h=I−A_h.

Define nonnegative integers c_(h,j),1≤j≤5, by c_(0,1)=1 and all other
initial entries zero. The update replaces each block label j as follows:

1 → 2; 2 → 1+3; 3 → 2+4; 4 → 3+5; 5 → 5+5. (1)

Multiplicity is additive, and all c are ordinary integers, not residues.
Let b_h=sum_j c_(h,j).

**Theorem.** For every h=4k≥4,

rank_F(B_h)=2^h−1−b_h. (2)

Exactly this many additive scalar rank-one terms are necessary and
sufficient to repair A_h to the identity. For example h=4 gives c=(2,0,3,0,1)
and rank9; h=8 gives c=(13,0,21,0,36) and rank185.

The theorem covers the entire resonant family h≡0 mod4. Other dimensions,
nonadditive repairs, physical quinary costs and selected κ are not claimed.
The separate even-characteristic-three problem in Discussion5194 is not
part of this work. Its semisimple proof cannot replace the Jordan analysis here.

## 1. The repeated-root tensor

Include the empty and full subsets. The full disjointness matrix is
T_h=K^⊗h, K=[[1,1],[1,0]]. Over F5,

K=3I+N, N=[[3,1],[1,2]], N²=0, N≠0.

Consequently U=K/3 is similar to the unipotent block J2=I+N2, where N2
has a single1 above its diagonal. Similarity can be exhibited over F5:
with columns (3,1)^T,(1,0)^T, the matrix S has determinant−1 and
S^-1 N S=N2; rescale the second column by3 to obtain S' with
S'^-1 (N/3) S'=N2. Thus U^⊗h is similar to J2^⊗h over every F.

Since U^5=I, at h divisible by4 we have 3^h=1 and

T_h=U^⊗h, T_h^5=I, M_h:=I−T_h satisfies M_h^5=0. (3)

Unlike the characteristic-three cases, the full operator is nonsemisimple.

## 2. Complete finite Jordan rule and induction

For 1≤j≤5 write Jj=I+Nj, Nj the j-by-j upper shift. The nilpotent part
of J2⊗Jj is the explicit 2j-by-2j matrix

L_j = [[Nj, I+Nj],[0,Nj]].

For each r≥1 its power is

L_j^r = [[Nj^r, r Nj^(r−1)(I+Nj)],[0,Nj^r]], (4)

by block multiplication; coefficients are reduced modulo5. In particular
L_j^5=0. Direct reduction of these five explicitly specified matrices gives
ranks of powers r=1,...,5:

| j | ranks | resulting Jordan sizes |
|---|---|---|
|1|1,0,0,0,0|2|
|2|2,1,0,0,0|1,3|
|3|4,2,1,0,0|2,4|
|4|6,4,2,1,0|3,5|
|5|8,6,4,2,0|5,5|

This is a fixed finite algebra certificate: entries are given by (4), and
exact elimination over F5 checks all25 ranks independently of h. A separate
kernel enumeration on the smallest cases and rank-nullity differences
corroborate it. Rank of a matrix over the prime subfield is unchanged under
field extension, so the table holds over every F. The ranks of all powers
of a nilpotent matrix determine its Jordan sizes: the number of blocks of
size at least r is rank(L^(r−1))−rank(L^r). This proves (1), not just a
numerical pattern at a few large dimensions.

Start with the one-dimensional identity J1 at h=0. Tensoring once more
by J2 applies this complete table separately to every Jordan block.
Induction proves that c_(h,j) counts the actual blocks of U^⊗h for every h.
The ordinary-integer dimension invariant sum_j j c_(h,j)=2^h follows from
(1). At h divisible by4, (3) therefore implies

nullity(M_h)=b_h, rank(M_h)=2^h−b_h. (5)

## 3. The endpoint witness

Let e0 be the empty-subset coordinate and P=M_h^4. As M_h^5=0, P M_h=0.
In characteristic five, (I−T)^4=I+T+T²+T³+T⁴. The needed local powers are

K^0=I, K^1=[[1,1],[1,0]], K²=[[2,1],[1,1]],
K³=[[3,2],[2,1]], K⁴=[[0,3],[3,2]].

At an interior subset of size r,1≤r≤h−1, the coordinate of P e0 is

1+2^(h−r)+3^(h−r)2^r = 1+3^r+4^r,

because h is divisible by4. For r≡0,1,2,3 mod4 these are respectively
3,3,1,2 in F5, all nonzero. In particular P e0≠0, and e0 is not in the
image of M_h. This explicit witness is the endpoint fact absent from a
mere eigenvalue or rank count.

## 4. Exact deletion of empty and full subsets

In the order empty, proper nonempty, full, direct disjointness gives

M_h(x0,z,xF)=(-1^T z−xF, −x0*1+B_h z, −x0+xF). (6)

If B_h z=0, then M_h(0,z,0)=−(1^T z)e0. Apply P. The nonzero witness
of §3 forces 1^T z=0. As B_h is symmetric, its image is the orthogonal
complement of its kernel for the nondegenerate standard bilinear pairing.
Thus the all-ones vector belongs to its image. Choose w with B_h w=1.
By (6), M_h(1,w,1)=−(1^T w+1)e0. Applying P again gives 1^T w=−1,
so (1,w,1) lies in ker M_h.

Every kernel vector of M_h has xF=x0 by its last coordinate. Those with
x0=0 are exactly (0,z,0) with z in ker B_h, using the established sum-zero
condition. The vector (1,w,1) supplies precisely one additional independent
kernel direction. Therefore

nullity(M_h)=nullity(B_h)+1,
rank(M_h)=rank(B_h)+1.

Together with (5) this proves (2) over every characteristic-five field.

## 5. Additive correction cost and limits

If sum_(j=1)^ell u_j v_j^T=I−A_h, rank subadditivity gives
ell≥rank(B_h). Conversely a rank factorization over F writes B_h as the
sum of exactly rank(B_h) nonzero scalar rank-one matrices. This proves the
stated exact algebraic correction count. It supplies no permission to reuse
the binary physical ledger or to treat coefficient preparation/restoration
as free in a physical construction.

The universal result rests on the five finite Jordan rules followed by
induction and the all-h endpoint witness. Literal h4/h8 controls corroborate
source indexing but do not establish its quantifier. No Lean, large census,
physical supplier, full multiplication theorem or selected-frontier change
is part of this package.

## Open next questions and credit

Analyze the nonresonant characteristic-five dimensions with their own
endpoint restriction; this theorem does not claim their rank. Any physical
quinary repair requires an explicit caller and complete preparation,
routing, precision, dirty-return and tape invoice before expansion.

Ali Romman PR4962 supplies the binary tensor/deletion predecessor; Nick
Petalas PR5173 supplies the distinct ternary inverse; mikey1201 PR172 and
icekylinx/#144 retain the source endpoint/compiler lineage. Jordan tensor
classification and rank factorization retain their classical attribution.
Prepared by JoyboyBrian with substantial OpenAI Codex assistance; Apache-2.0.
This is a written proof with a fixed exact algebra certificate, not human or
formal proof-kernel approval.
49 changes: 49 additions & 0 deletions research/quinary-resonant-disjointness-repair/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
# Repeated-root disjointness repair in characteristic five

For every h=4k>=4 over every characteristic-five field, the actual proper-subset
disjointness repair has rank **2^h−1−b_h**, where b_h counts the Jordan blocks
in the explicit five-state recurrence in [PROOF.md](PROOF.md). This is also
its exact additive scalar rank-one correction count. First values are9 at
h4 and185 at h8.

The new mechanism is nonsemisimple: the source tensor factor has a repeated
root in characteristic five. Five fixed Jordan tensor identities yield the
all-h block multiplicities; a nonzero fourth-power endpoint witness proves
that deleting empty/full subsets lowers rank by exactly one. The written
induction and endpoint argument supply the quantifier. Finite h4/h8 ranks
are calibration, not extrapolation.

This follows the source matrix and attribution of Ali Romman's PR4962 and
Nick Petalas's distinct odd-characteristic-three PR5173. Their exact proof
bytes are live inputs pinned in SOURCES.json. The separately claimed even
characteristic-three scope in Discussion5194 is excluded. No binary invoice
is reused, and no physical quinary supplier, nonadditive repair, selected κ
change, human review or proof-kernel acceptance is claimed.

## Reproduction

```sh
python3.12 -B research/quinary-resonant-disjointness-repair/verify.py
python3.12 -B research/quinary-resonant-disjointness-repair/independent.py
python3.12 scripts/research_package.py --package research/quinary-resonant-disjointness-repair
```

Standard library, below one second for each direct check. The mandatory typed
RESULT records all five fixed tensor rules, source ranks, actual endpoint
witnesses and domain controls. The independent checker uses literal Kronecker
products, column-space reduction and a complete625-vector tiny kernel census.
No large predecessor replay is required.

## Open next questions

Nonresonant dimensions require their own endpoint analysis. A physical quinary
application requires an actual caller and paid preparation, movement, precision,
restoration and retained backing; none is supplied. TASK.json records these
separate obligations.

## Credit

Ali Romman PR4962; Nick Petalas PR5173; mikey1201 PR172; icekylinx/#144 and the
retained endpoint/compiler lineage. Classical Jordan theory and rank
factorization retain their attribution. Prepared by JoyboyBrian with substantial
OpenAI Codex assistance; Apache-2.0.
114 changes: 114 additions & 0 deletions research/quinary-resonant-disjointness-repair/RESULT.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,114 @@
{
"scope": "all resonant h proof in PROOF.md; bounded exact controls here",
"characteristic": 5,
"resonant_modulus": 4,
"rules": [
{
"j": 1,
"blocks": [
2
],
"ranks": [
1,
0,
0,
0,
0
]
},
{
"j": 2,
"blocks": [
1,
3
],
"ranks": [
2,
1,
0,
0,
0
]
},
{
"j": 3,
"blocks": [
2,
4
],
"ranks": [
4,
2,
1,
0,
0
]
},
{
"j": 4,
"blocks": [
3,
5
],
"ranks": [
6,
4,
2,
1,
0
]
},
{
"j": 5,
"blocks": [
5,
5
],
"ranks": [
8,
6,
4,
2,
0
]
}
],
"fixtures": [
{
"h": 4,
"rank": 9,
"multiplicities": [
2,
0,
3,
0,
1
],
"witness_sha256": "059cca9e05ed7a044d840be7086413d9b14cf264748917c607245b9e00a9e7a3"
},
{
"h": 8,
"rank": 185,
"multiplicities": [
13,
0,
21,
0,
36
],
"witness_sha256": "c475ee092b0f28f84a44dcd34624dedd7a1244e318b6b4e66ec1c56d4a58331e"
}
],
"invalid_domains_rejected": [
"0",
"2",
"3",
"True"
],
"kernel_census": 5,
"source_sha256": {
"research/boolean-disjointness-repair/PROOF.txt": "f831bc6fe5af6ad87dd028070c6db60d55f8bd1ffad052ed27092b671ab2a427",
"research/ternary-disjointness-repair/PROOF.md": "43d61f2c7ea623e6a39783b486c12c8edea62197a2ae2512a4e4f9596f8920e0"
},
"no_physical_supplier": true
}
30 changes: 30 additions & 0 deletions research/quinary-resonant-disjointness-repair/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
{
"schema_version": 1,
"base_commit": "ec2780e8aeb1c88acaf5e9e470ef8c68e7fbbb6c",
"files": [
{
"path": "certificates/selected-result.json",
"usage": "provenance",
"git_blob": "c35482cd9035843b15a634cc8276a7e1e40e8395",
"sha256": "6fe0a099b89aa2aed796a13c3a21dafc63f3388ebf397464e44e0451893b9375"
},
{
"path": "docs/research/agent-task-catalog.json",
"usage": "provenance",
"git_blob": "f126dd2e2dad2b669e75d1d0d0bd05c600725602",
"sha256": "f7a48bfa2659b88378282a8eee7d9fe52cdf505594124c4d1686b58510fffd72"
},
{
"path": "research/boolean-disjointness-repair/PROOF.txt",
"usage": "input",
"git_blob": "96b0e770da33660eff812df8b74302b15c557e59",
"sha256": "f831bc6fe5af6ad87dd028070c6db60d55f8bd1ffad052ed27092b671ab2a427"
},
{
"path": "research/ternary-disjointness-repair/PROOF.md",
"usage": "input",
"git_blob": "680e0adb269a3cd3420c2fe2312bea99c4b76ded",
"sha256": "43d61f2c7ea623e6a39783b486c12c8edea62197a2ae2512a4e4f9596f8920e0"
}
]
}
Loading