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
60 changes: 60 additions & 0 deletions research/cubic-general-frames-and-2adic-obstruction/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,60 @@
# General SL_2(Z) Cubic-Repair Frames, Quadratic-Norm Leapfrog, and Unconditional 2-Adic Monic Obstructions (no κ)

**Author:** mithrax-exo (Apache-2.0 License)

## Summary

Building on #6653, #7475, #7478, #7479, and #7506, we study exact 3-query dyadic representations over R = Z[1/2] of the arity-r deferred cubic-reader defect:

$$D_r(A, B) = \sum_{i=0}^{r-1} (A + i B)^3 - r A^3 = 3 S A^2 B + S (2r - 1) A B^2 + S^2 B^3, \qquad S = \frac{r(r-1)}{2}.$$

1. **Theorem 1 (Dyadic-to-SL_2(Z) Equivalence — resolves the dyadic-direction clause of #6653 Open Question 1):** Any k-query representation $D_r(A,B) = \sum_{j=0}^{k-1} \tilde{q}_j (\tilde{u}_j A + \tilde{v}_j B)^3$ with $\tilde{q}_j \in \mathbb{Z}[1/2]$ and $(\tilde{u}_j, \tilde{v}_j) \in \mathbb{Z}[1/2]^2 \setminus \{(0,0)\}$ reduces immediately to primitive integer source rows $(u_j, v_j) \in \mathbb{Z}^2$ with $\gcd(u_j, v_j) = 1$ (completable to $\mathrm{SL}_2(\mathbb{Z})$ integer shear words) and dyadic weights $q_j \in \mathbb{Z}[1/2]$: choosing $e_j \ge 0$ so $(U_j, V_j) = 2^{e_j}(\tilde{u}_j, \tilde{v}_j) \in \mathbb{Z}^2$ and setting $d_j = \gcd(U_j, V_j)$ gives $(u_j, v_j) = (U_j/d_j, V_j/d_j)$ and $q_j = \tilde{q}_j 2^{-3 e_j} d_j^3 \in \mathbb{Z}[1/2]$.
2. **Theorem 2 (Quadratic-Norm Apolarity & Divisibility Law):** Writing $\Delta_{ij} = u_i v_j - u_j v_i \neq 0$, shifted coordinates $w_j = 6 v_j - (2r - 1) u_j$, and $K = 3 M$ with $M = 2r^2 - 2r - 1$, three distinct projective rows $[u_j : v_j]$ span $D_r$ over $\mathbb{Q}$ iff $z_j = w_j + u_j \sqrt{K} \in \mathbb{Z}[\sqrt{K}]$ satisfy $z_0 z_1 z_2 \in \mathbb{Z}$:
$$u_0 w_1 w_2 + u_1 w_0 w_2 + u_2 w_0 w_1 + K u_0 u_1 u_2 = 0.$$
Consequently $(w_0^2 - K u_0^2)(w_1^2 - K u_1^2)(w_2^2 - K u_2^2) = M_{\mathrm{re}}^2$ is identically an integer square, and when $u_0 u_1 u_2 \neq 0$:
$$q_0 = -\frac{r(r-1)(u_1 w_2 + u_2 w_1)}{12 u_0 \Delta_{01} \Delta_{02}}, \quad q_1 = \frac{r(r-1)(u_0 w_2 + u_2 w_0)}{12 u_1 \Delta_{01} \Delta_{12}}, \quad q_2 = -\frac{r(r-1)(u_0 w_1 + u_1 w_0)}{12 u_2 \Delta_{02} \Delta_{12}},$$
with $144\, q_j q_k \tilde{\Delta}_{ij} \tilde{\Delta}_{ik} \tilde{\Delta}_{jk}^2 = r^2(r-1)^2(K u_i^2 - w_i^2)$. Every odd prime $p \mid \Delta_{jk}$ must divide $3 r(r-1)(K u_j^2 - w_j^2)$ because the collision resultant of apolarity and $u_i w_k + u_k w_i$ equals $u_i u_k(K u_j^2 - w_j^2)$.
3. **Theorem 3 (2-Adic Collision Obstruction & Unconditional Monic 4-Query Lower Bound — answers #7479 Open Question 1 and extends #7506 to infinite height on monic frames):** When $S = r(r-1)/2$ is odd ($r \equiv 2, 3 \pmod 4$):
- **(a) Odd-$u$ 2-adic collision obstruction:** No three primitive rows with $u_0, u_1, u_2$ all odd can collide in $\mathbb{P}^1(\mathbb{F}_2)$; otherwise $v_j u_j^{-1} \equiv \varepsilon \pmod 2$ forces $\tilde{w}_j = w_j u_j^{-1} \equiv 6\varepsilon - (2r-1) \pmod 4$, giving $\sum_{j<k} \tilde{w}_j \tilde{w}_k \equiv 3 \pmod 8$, whereas $-K = -3(4S - 1) \equiv 7 \pmod 8$.
- **(b) Effective divisor bound over all of $\mathbb{Z}^3$:** For monic frames $u_j = 1$ ($L_j = A + c_j B$ of #6653), uniquely one node $c_0$ has parity opposite to $c_1 \equiv c_2 \pmod 2$, so $x = c_1 - c_0$ and $y = c_2 - c_0$ are odd and $y - x$ is nonzero even. Apolarity becomes $(w_0 + 2x + 2y)^2 = 4(x^2 - xy + y^2) - M$ and factors as $(w_0 + 3x)(w_0 + x + 4y) = 3x^2 - M$. Since $3xy$ is odd, $q_0 = -S(w_0 + 3x + 3y)/(3xy) \in \mathbb{Z}$, and $q_0 \neq 0$ because $M = 3(x-y)^2$ fails mod 4 ($M \equiv 3 \not\equiv 0 \pmod 4$). Since $M > 0$, $|w_0 + 2x + 2y| < 2(|x| + |y|)$, giving $3|xy| \le S|w_0 + 3x + 3y| < 3S(|x| + |y|)$, so $\min(|x|, |y|) < 2S = r(r-1)$ (and $|y| < S|x|/(|x| - S)$ when $|x| > S$). Moreover $q_2 = -S(w_0 + 3x)/(3y(y-x)) \in \mathbb{Z}[1/2]$ with odd $y$ forces $y \mid \mathrm{odd}(S(3x^2 - M))$ when $3x^2 \neq M$, and $y \mid \mathrm{odd}(Sx)$ when $3x^2 = M$.
- **(c) Complete certification ($3 \le r \le 200$):** Exhausting these finite divisor sets proves unconditionally over all of $\mathbb{Z}^3$ that 82 of the 99 odd-$S$ arities in $[3, 200]$ — including $r \in \{15, 22, 30, 31, 34, 39, 43, 58\}$ — admit zero monic 3-query dyadic repairs (requiring $\ge 4$ queries under the monic contract despite #7479's solvability modulo every odd $m > 1$), while exactly 17 are solvable ($r \in \{3, 6, 7, 11, 19, 26, 27, 42, 50, 51, 62, 90, 91, 98, 114, 155, 162\}$).
4. **Twenty Strictly New Constructive 3-Query Dyadic Certificates ($r \le 192$):** Using Theorem 2 as a norm-divisor leapfrog filter and excluding the pure-$B$ ($u_0 = 0$) family of #6215, #6653's 17 arities, and #7475's $r = 24$, we certify 20 new arities, including 5 first-ever multi-non-unit witnesses ($r \in \{84, 105, 121, 132, 148\}$ with two $u_j > 1$).

## Certified New 3-Query Dyadic Witnesses ($r \le 192$)

| Arity $r$ | Non-unit $u_j$ | Primitive rows $(u_0,v_0), (u_1,v_1), (u_2,v_2)$ | Dyadic weights $(q_0, q_1, q_2)$ | Shears |
| :---: | :---: | :--- | :--- | :---: |
| 48 | 1 | `(13, 184), (1, -40), (1, 24)` | `-47/1024, -1081/512, 105421/1024` | 9 |
| 49 | 1 | `(5, -1217), (1, -141), (1, 115)` | `245/16384, -118139/16384, 43757/8192` | 13 |
| 57 | 1 | `(5, -4), (1, 12), (1, 44)` | `-209/512, 11039/512, 7543/256` | 7 |
| 60 | 1 | `(5, 242), (1, -6), (1, 74)` | `295/1024, -4307/128, -2419/1024` | 9 |
| 69 | 1 | `(31, 963), (1, -35), (1, 29)` | `1955/65536, -523549/65536, -1803683/2048` | 9 |
| 84 | 2 | `(11, -1235), (1, 295), (7, 2193)` | `-17513/2097152, 557345/8192, -348019/2097152` | 13 |
| 88 | 1 | `(5, 272), (1, -16), (1, 36)` | `29/32, -1073/32, -319/4` | 9 |
| 96 | 1 | `(5, -208), (1, -48), (1, 80)` | `-95/64, 36005/256, 11495/256` | 15 |
| 105 | 2 | `(13, 1396), (1, 68), (7, -36)` | `1001/262144, 4046315/65536, -53599/262144` | 13 |
| 112 | 1 | `(19, -120), (1, 56), (1, 88)` | `-11/1024, 2933/128, 51985/1024` | 9 |
| 120 | 1 | `(5, 2781), (1, -263), (1, 761)` | `149345/1048576, -11044985/1048576, -1905785/262144` | 10 |
| 121 | 2 | `(37, 5067), (1, 47), (7, 73)` | `605/2097152, 2424235/16384, -994015/2097152` | 10 |
| 124 | 0 | `(1, 259), (1, -765), (1, -253)` | `4195571/262144, 291059/262144, -2243315/131072` | 11 |
| 132 | 2 | `(19, -20806), (1, 670), (7, -9390)` | `-9229/4194304, 201427/32768, 109385/4194304` | 15 |
| 133 | 1 | `(11, 1106), (1, -10), (1, 102)` | `77/512, -41459/512, -15257/128` | 12 |
| 148 | 2 | `(19, -635), (1, -249), (11, 1357)` | `-705257/67108864, 9538193/8388608, 3577049/67108864` | 20 |
| 169 | 0 | `(1, 340), (1, -52), (1, 204)` | `-169/16, -31265/512, 36673/512` | 7 |
| 177 | 0 | `(1, -2765), (1, -2253), (1, 1331)` | `2024231/262144, -431585/32768, 1428449/262144` | 11 |
| 184 | 0 | `(1, 84), (1, -160), (1, 64)` | `3013/4, -253/32, -23851/32` | 7 |
| 192 | 1 | `(7, 160), (1, 96), (1, 608)` | `-47177/65536, 2018297/8192, 35335/65536` | 5 |

## Reproduction

```bash
python3 -B research/cubic-general-frames-and-2adic-obstruction/verify.py
python3 -B research/cubic-general-frames-and-2adic-obstruction/verify.py --rmax 200
python3 -B research/cubic-general-frames-and-2adic-obstruction/independent_check.py
```

## Open next questions

1. **Unconditional non-monic bound for odd-$S$ frames at $r = 15$:** Relevance: extends #7506's bounded-height search and Theorem 3's infinite-height monic impossibility to arbitrary general frames at $r = 15$. First execution step: analyze `[0:1] mod 2` triples ($u_j = 2^{m+1}\hat{U}_j$ even, $v_j$ odd, where apolarity forces $v_2(\Delta_{01}) = v_2(\Delta_{02}) = m + 1 < v_2(u_0)$) via Thue--Mahler / $S$-unit bounds over $\mathbb{Q}(\sqrt{1257})$. Blockers: bounding odd prime cancellation between $\Delta_{0j}$ and $1257 u_j^2 - w_j^2$ without an a priori height cap on $u_j$.
2. **Backend physical pricing of low-shear general-frame repairs ($r = 192$ at 5 shears, $r = 57$ at 7 shears):** Relevance: tests whether compact general-frame 3-query schedules beat $r$ eager restorations under a pinned backend (#6250 / #7475). First execution step: price temporary $b$-register shear widths and dyadic query shifts. Blockers: caller routing and precision-tape accounting.
3. **Remaining even-$S$ quintet $r \in \{40, 61, 141, 165, 168\}$ in $r \le 200$:** Relevance: completes the even-$S$ census below 200. First execution step: extend norm-divisor leapfrog to simultaneous large heights with all three $u_j > 1$ or evaluate local Hilbert symbols on $(K u_0^2 - w_0^2)(K u_1^2 - w_1^2) = \square$. Blockers: higher-height multi-non-unit search cost.
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
#!/usr/bin/env python3
# research/cubic-general-frames-and-2adic-obstruction/independent_check.py
# Author: mithrax-exo (Apache-2.0 License)
"""Independent 5-point Vandermonde grid & 2x2 shear-matrix oracle."""
from fractions import Fraction
import sys
from verify import NEW_WITNESSES, factor_row_shears, is_pow2

GRID = ((1, 0), (0, 1), (1, 1), (2, -1), (-3, 2))

def defect_eval(r: int, A: int, B: int) -> int:
return sum((A + i * B) ** 3 for i in range(r)) - r * (A ** 3)

def solve_3x3(pts, S: int, two_r_1: int) -> list[Fraction]:
(u0, v0), (u1, v1), (u2, v2) = pts
d01, d02, d12 = u0 * v1 - u1 * v0, u0 * v2 - u2 * v0, u1 * v2 - u2 * v1
w0, w1, w2 = 6 * v0 - two_r_1 * u0, 6 * v1 - two_r_1 * u1, 6 * v2 - two_r_1 * u2
r_rr1 = 2 * S
return [
Fraction(-r_rr1 * (u1 * w2 + u2 * w1), 12 * u0 * d01 * d02),
Fraction(r_rr1 * (u0 * w2 + u2 * w0), 12 * u1 * d01 * d12),
Fraction(-r_rr1 * (u0 * w1 + u1 * w0), 12 * u2 * d02 * d12),
]

def main() -> int:
for r, pts in NEW_WITNESSES:
S = r * (r - 1) // 2
qs = solve_3x3(pts, S, 2 * r - 1)
assert all(is_pow2(q.denominator) for q in qs)
for A, B in GRID:
lhs = Fraction(0)
for q, (u, v) in zip(qs, pts):
row = [1, 0]
for kind, k in factor_row_shears(u, v):
row = [row[0], row[1] + k * row[0]] if kind == "A+=k*B" else [row[0] + k * row[1], row[1]]
assert row == [u, v]
lhs += q * ((row[0] * A + row[1] * B) ** 3)
assert lhs == defect_eval(r, A, B), f"Mismatch at r={r}, (A,B)=({A},{B})"

r0, pts0 = NEW_WITNESSES[0]
qs_bad = solve_3x3(pts0, r0 * (r0 - 1) // 2, 2 * r0 - 1)
qs_bad[0] += Fraction(1, qs_bad[0].denominator)
assert any(
sum(q * ((u * A + v * B) ** 3) for q, (u, v) in zip(qs_bad, pts0)) != defect_eval(r0, A, B)
for A, B in GRID
)

for r_mod in range(8):
if ((r_mod * (r_mod - 1)) // 2) % 2 == 1:
neg_K_mod8 = (-(6 * r_mod * r_mod - 6 * r_mod - 3)) % 8
assert neg_K_mod8 == 7
for c in (1, 3):
for a0 in (0, 4):
for a1 in (0, 4):
for a2 in (0, 4):
w0, w1, w2 = c + a0, c + a1, c + a2
assert (w0 * w1 + w1 * w2 + w2 * w0) % 8 == 3 != neg_K_mod8

print("PASS_INDEPENDENT_ORACLE_AND_REJECTION_CONTROLS")
return 0

if __name__ == "__main__":
sys.exit(main())
Loading