Skip to content
Open
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
41 changes: 41 additions & 0 deletions research/two-prime-alphabet-descent/CREDITS.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
# Sources and credit

Ali Romman (@aliromman) prepared the simultaneous-column descent proof,
sharp three-column family and exact controls with substantial OpenAI
Codex assistance. New material is Apache-2.0. No independent human or
formal proof-kernel acceptance is claimed.

Ali Romman's [PR #7769](https://github.com/CrocSwap/integer-mult-bounds/pull/7769)
supplied the [complete fifteenth-root capacity proof and attaining matrix](https://github.com/aliromman/integer-mult-bounds/blob/6dbe6e9503f26189833d5ef213aee708679835f2/research/fifteenth-root-column-capacity/PROOF.md),
[full finite certificate](https://github.com/aliromman/integer-mult-bounds/blob/6dbe6e9503f26189833d5ef213aee708679835f2/research/fifteenth-root-column-capacity/RESULT.json)
and [checker](https://github.com/aliromman/integer-mult-bounds/blob/6dbe6e9503f26189833d5ef213aee708679835f2/research/fifteenth-root-column-capacity/verify.py).
All 28 source package files remain byte-exact. The full verifier runs
unchanged. The source exact capacity is an input to the all-power
corollary; it is not claimed again as a new finite result.

The retained source includes Ali Romman's
[PR #7762](https://github.com/CrocSwap/integer-mult-bounds/pull/7762)
[eight-root census and projected-star proof](https://github.com/aliromman/integer-mult-bounds/blob/d33fe8562c86888307cbba4a2aa92490212c8d78/research/mixed-cycle-column-obstruction/PROOF.md),
and the motivation in
[PR #7768](https://github.com/CrocSwap/integer-mult-bounds/pull/7768)'s
[general defect bound](https://github.com/aliromman/integer-mult-bounds/blob/fefebaf8285f3c46f75a815657acbf48780763fb/research/mixed-cycle-column-defect/PROOF.md).
The new descent proof does not use the latter spectral bound.

Nick Petalas (@petalas) supplied the two-prime projection and original
completion question in
[PR #7744](https://github.com/CrocSwap/integer-mult-bounds/pull/7744).
Its [proof and question](https://github.com/petalas/integer-mult-bounds/blob/7dc85513e8ee9b771a5856f027cf199936f85af0/research/two-prime-column-completion/README.md)
and checker remain in the source closure and replay unchanged. All
predecessor notices, earlier completion credit and David Ganz's atlas
credit remain intact.

T. Y. Lam and K. H. Leung supplied the classical two-prime minimal-cycle
classification in
[On Vanishing Sums of Roots of Unity, Corollary 3.4](https://arxiv.org/pdf/math/9511209#page=8).
It is an explicit mathematical input to the new all-power proof.
Cyclotomic polynomial arithmetic, Fourier triples and phase-gauge
invariants are also classical methods, not new general theory.

SOURCE-PINS.json authenticates the retained closure; SOURCES.json records
canonical-main provenance. Source commits are content pins, not ancestry
claims. No global literature priority is claimed.
184 changes: 184 additions & 0 deletions research/two-prime-alphabet-descent/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,184 @@
# Higher prime powers do not enlarge this column alphabet

Ali Romman with substantial OpenAI Codex assistance. Apache-2.0.

**Theorem.** Let q>=5 be prime and let a,b>=1. Let H have q+3 rows,
at least four pairwise orthogonal columns, and entries in the group
of roots `mu_m`, where `m=3^a q^b`. Row and column phases from mu_m
change H into a matrix with entries in `mu_(3q)`.

Equivalently, every two-by-two multiplicative cross ratio of H lies
in mu_(3q). The explicit double dephasing is

```text
H'_rj = H_rj conjugate(H_r0) conjugate(H_0j) H_00.
```

The theorem says that every entry of H' lies in the smaller alphabet.
It preserves all column inner products and their norms.

**Corollary.** The maximum number of pairwise orthogonal eight-entry
columns over `mu_(3^a 5^b)` is exactly four for every a,b>=1. This follows
from the complete fifteenth-root capacity proof in
[PR #7769](https://github.com/CrocSwap/integer-mult-bounds/pull/7769).
Its [attaining matrix and exhaustive certificate](https://github.com/aliromman/integer-mult-bounds/blob/6dbe6e9503f26189833d5ef213aee708679835f2/research/fifteenth-root-column-capacity/PROOF.md)
remain unchanged and replay here. Four columns attain the bound in
every larger alphabet because mu_15 is a subgroup.

This is an exact alphabet theorem, not a physical supplier. No new
kappa or change to the selected frontier follows.

## 1. Pairwise zero sums have only two cycles

The classical two-prime result of
[T. Y. Lam and K. H. Leung, Corollary 3.4](https://arxiv.org/pdf/math/9511209#page=8)
states that minimal positive zero sums over mu_m are rotated prime
cycles. Decompose any positive zero sum into such minimal sums.
For a zero sum of length q+3, the only nonnegative solution of
`3u+qv=q+3` is u=v=1. Thus each pair of orthogonal columns gives
exactly one rotated 3-cycle and one rotated q-cycle, as a multiset.

Two consequences are needed. First, each of those cycles lies inside
one coset of mu_(3q) in mu_m. Its coset has multiplicity three or q;
if the two cosets coincide, all q+3 terms lie in that one coset.
Second, a single root has multiplicity at most two in the zero sum.
Each individual prime cycle has no repeated root, so their union
cannot contain a root three times.

This use of the classical theorem is explicit. The new argument below
concerns simultaneous columns; the cycle classification is not claimed
as new theory.

## 2. Every nonconstant quotient column uses the same three rows

Multiply rows to make the first column constant one. Work in the
additive quotient group

```text
K = mu_m / mu_(3q).
```

Its order is `3^(a-1) q^(b-1)`, which is odd. By Section 1, each column's
quotient entries are either constant or have two values with counts
three and q. A column phase makes the value on q rows zero. A
nonconstant column then has the form `t 1_S`, where t is nonzero in K
and S is a three-row subset. A constant column can be made identically
zero in the quotient.

Suppose two nonconstant columns have different subsets S,T. Put
`k=|S intersection T|`, so k is 0,1 or 2. Their quotient difference has
these values and multiplicities before coincident values are combined:

| Row region | Quotient value | Multiplicity |
| --- | --- | --- |
| Outside S and T | 0 | q-3+k |
| S only | t | 3-k |
| T only | -u | 3-k |
| Intersection | t-u | k |

Here t,u are nonzero. If k>0, the only possible equalities between the
four values are t=u or t=-u. These cannot both hold in an odd-order
group. At least three distinct values therefore remain, contrary to
Section 1.

If k=0, there are three occupied values unless t=-u. In that exceptional
case there are two, with multiplicities q-3 and six. They cannot be
three and q for a prime q>=5: that equality would require q=6. Nor
is the quotient difference constant. This is again impossible.

Thus all nonconstant quotient columns have one common three-row subset
S. Every quotient column is `t_j 1_S`, allowing t_j=0 for the constant
columns. If no nonconstant quotient column exists, column phases
already place every entry in mu_(3q), and the theorem follows.

## 3. An essential higher phase permits at most three columns

Assume that a nonconstant quotient column exists. The first column has
t_j=0, so there are at least two distinct quotient values among columns.
Partition the columns by t_j.

For columns in different parts, their quotient difference has value
`t_j-t_k` on S and zero on the other q rows. The two cycles in their
inner product must therefore be supported separately: the 3-cycle is
on S and the q-cycle is on its complement. In particular, their
restrictions to S are orthogonal in C^3.

Now take two columns x,y in the same part, and choose a column z from
a different part. The three coordinate ratios x/z form a rotated
3-cycle, and so do y/z. After their scalar factors are removed, each
is a permutation of `(1,omega,omega^2)`. Up to scalar, such permutations
lie on only two lines:

```text
span(1,omega,omega^2), span(1,omega^2,omega).
```

These lines are orthogonal. Multiplication by the unit-modulus entries
of z preserves this conclusion. Therefore x and y restricted to S are
either proportional or orthogonal.

They cannot be proportional. If they were, the three terms of their
inner product on S would be exactly the same root. The complete
q+3-term zero sum would then contain that root at least three times,
contrary to Section 1. Hence x and y are also orthogonal on S.

All column restrictions to S are consequently pairwise orthogonal
nonzero vectors in C^3. There can be at most three columns. This
contradicts the hypothesis of at least four and proves the theorem.

## 4. The four-column threshold is sharp

Suppose m>3q and write eta for a primitive m-th root, omega for a
primitive cube root and zeta for a primitive q-th root. The following
(q+3)-by-3 matrix has columns indexed by j=0,1,2:

```text
first three rows: H_rj = eta^j omega^(rj), r=0,1,2,
last q rows: H_rj = zeta^(rj), r=0,...,q-1.
```

Any two columns are orthogonal on each block separately. Their
nonzero index differences are one or two, so both geometric sums
vanish. All column norms squared are q+3.

Use the first row of each block and columns zero and one. Their
multiplicative cross ratio is eta, which is not in mu_(3q).
Cross ratios are unchanged by row and column phases. Thus this
three-column matrix cannot descend to the smaller alphabet, even
if arbitrary unit-modulus row and column phases are permitted.
This gives a uniform counterexample to a three-column version of
the theorem. The requirement of four columns is substantive.

## Evidence and limits

The complete source capacity verifier runs unchanged, including its
full 95-vertex extension graph, triangle exclusion, attaining matrix
and nested predecessor checks. All 28 retained files are authenticated.
The new checker uses only exact integers and cyclotomic polynomial
reduction. It checks:

- 10,272 canonical intersection/group cases and 173,600 full subset-pair
cases for the quotient collision lemma. A composite-q=6 pattern
exhibits the excluded multiplicities three and six.
- All prime-cycle pairs in alphabets of orders 15,45,75,225 and
21,63,147,441: 14,400 pairs. Their quotient counts and maximum root
multiplicity agree with the proof. This is the complete cycle-pair
family in those alphabets, not an exhaustive search of arbitrary
root multisets without the classical theorem.
- Every normalized three-root zero sum in these eight alphabets, by
exact basis lookup. Each alphabet has exactly the two displayed
normalized Fourier triples. All line-pair inner products agree.
- The explicit three-column matrices and their nontrivial higher-phase
cross ratios, plus four-column attainment over each 3/5 alphabet.
Full row/column phase changes dephase exactly to the source matrix.
In total, 200 exact Gram entries are checked.

The all-prime/all-power statement is a written proof using the cited
classical theorem. The finite checks do not enumerate all matrices or
replace that theorem. The source capacity result remains a finite
certificate rather than a formal proof-kernel artifact. Other row counts,
a third prime divisor, approximate orthogonality, full equivalence
classification, native preparation and complete caller prices remain
outside the result. No selected cold replay, Lean, Docker, independent
human acceptance or formal-kernel acceptance is claimed. All five
framework assumptions and the selected frontier remain unchanged.
34 changes: 34 additions & 0 deletions research/two-prime-alphabet-descent/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
# Descent from higher prime-power phase alphabets

For every prime q>=5 and positive a,b, a q+3-row matrix with at least
four pairwise orthogonal columns over mu_(3^a q^b) can be changed by
row and column phases into a matrix over mu_(3q).

Consequently the maximum number of orthogonal eight-entry columns is
**exactly four for every alphabet mu_(3^a 5^b)**. Three-column examples
with essential higher phases show that the column threshold is sharp.
No new kappa or physical supplier follows.

Read [PROOF.md](PROOF.md), [RESULT.json](RESULT.json), [verify.py](verify.py),
[SOURCE-PINS.json](SOURCE-PINS.json) and [CREDITS.md](CREDITS.md).

Use Python 3.12 and the standard library:

```sh
python3.12 -B scripts/research_package.py --package research/two-prime-alphabet-descent
```

Both modes print `PASS_TWO_PRIME_ALPHABET_DESCENT`. The unchanged
fifteenth-root capacity verifier and its nested source checks run.
New exact checks cover quotient patterns, eight cyclotomic alphabets,
14,400 prime-cycle pairs, the sharp three-column examples and lifted
four-column attainment. No network is needed.

## Open next questions

- Determine which changed row counts permit an essential higher phase
with four or more columns. First identify the permitted cycle types.
- Test an alphabet with a third prime divisor under a specified matrix
contract. The two-cycle proof does not cover that setting.

Ali Romman with substantial OpenAI Codex assistance. Apache-2.0.
Loading
Loading