By @josebrox

613b16a / docs/results_20261119_v81.md

Raw
# Paired contacts, amplified rationality descent, and the remaining realization problems

The ordinary rational abc quartic now uses **56 total variables**, improving
the published 58-variable construction and passing the target of 57.
Its 46 quadratic residuals have a sum of squares with degree four, content
one, 809 coordinate monomials, and exact coefficient height

```math
H(F_B)=B^2+377888002299\qquad(B\in\mathbb Z_{>0}).
```

Every sufficiently large fixed odd B defines the integers over the
rationals, under ordinary rational abc and the explicitly retained
elliptic, formal-comparison, trace-index and prime-pattern premises.
The pole realization uses Burungale's published Theorem B in its
stated square classes. The sufficient B is ineffective.

The [complete proof](../papers/quartic_existence/paired_contact56_20261119_v81.tex),
[coefficient table](../computation/variable_reductions_v81/abc/quartic56/polynomial.json),
[quadratic equations](../computation/variable_reductions_v81/abc/quartic56/quadrics.json),
[independent reconstruction](../computation/variable_reductions_v81/abc/check56.py),
and [mathematical review](../computation/variable_reductions_v81/abc/proof_review56.md)
give the full specification and scope. The shared detector and Laurent
packing originate in **notes of Gabe Conant with Astra 6.2**. Their
contributions, the v79 derivative transfer, the v80 normalizer, and the
new v81 rechart are distinguished in the proof.

## 1. A nonlinear contact change saves two coordinates

Put

```math
M=B+2(X^8+T_{\rm sc}^2),\quad a=M^4+16,\quad
T=a^2-z^2,\quad S=a^2+z^2,\quad P=T^2.
```

The sixteen linear factors from the eight shifts
8, 9, 12, 15, 17, 20, 23, 24 become four factors
P−2r²S+r⁴, for r=1,4,7,8. Pair radii (1,8) and (4,7), obtaining
A1 and A2. If H_i=2T(A_i)_P+(A_i)_S, their equal sums of squared
radii give the exact identity

```math
H_2-H_1=720(-4T+8S-130).
```

Thus only one derivative coordinate is stored. Storing F/a, with
F=A1 A2, also removes the extra product coordinate from the derivative
quotient because a is always positive. These are equivalences on every
rational fibre: no zero intermediate coordinate is removed. The contact
block uses nine coordinates instead of eleven; the one-helper
normalizer remains unchanged.

The height proof works for this fixed positive shift set with the same
strict margin 92221/99700. A new uniform two-adic lemma proves that
n² Tr(x([n]R0)) is integral at two for every integer witness. It makes
a a two-adic square for every odd B and supplies the exact actual-input
class needed by the native pole theorem. Negative odd valuations at
the remaining primes are covered by the v80 normalizer.

## 2. What the Pasten–Vidaux paper contributes

We checked the primary version of
[Hector Pasten and Xavier Vidaux, *On Hilbert's tenth problem for number fields*, arXiv:2611.24748v1](https://arxiv.org/html/2611.24748v1).
Its key architecture uses an elliptic curve E/Q whose positive rank
is unchanged over K. Elliptic formal-parameter ratios then lie in Q
and approximate integral targets. A local pole predicate propagates
one odd contact to all primes above the underlying rational prime.
The difference t−σ(t), for an irrational input, gives an upper height
bound that contradicts the abc contact lower bound outside a finite
set. We found no gap in that architecture under its stated assumptions.

The [amplified descent manuscript](../papers/number_fields/amplified_rationality_descent_20261119_v81.tex)
proves the following extensions, with the original descent method
attributed to Pasten and Vidaux:

- Elliptic ratios approximate **every rational target**, simultaneously
  at the real place and finitely many finite places, with finite
  avoidance. This removes the final quotient of two independently
  defined sets.
- The first pole input treats **both signs** of an odd valuation. An
  odd power in the second input amplifies the extracted congruence
  without a new witness coordinate.
- For a reduced rational divisor of degree r supported at odd-order
  zeros and poles, a lower estimate `N^(1)(D,t) >= theta*h(t)-C`
  suffices whenever `2*theta-r > 2/k`.
- The cubic rational map

```math
f(T)=\frac{T^3-2}{T^3-3},\qquad k=2,
```

  has six simple zero/pole contacts and no rational zero, pole, or
  fibre above −1. Only its one fixed divisor estimate with theta>7/2
  is needed. Ordinary abc over K supplies it. Alternatively a polynomial
  of degree five suffices with amplification k=3 in the positive-contact
  architecture.
- Only **prepared intrinsic pole inputs** are needed. A sixteen-element
  multiplier menu handles real and two-adic square classes while
  retaining the universal negative-odd-valuation obstruction. Ordered
  approximation prepares the actual inputs, including every new prime
  of the first chosen ratio.
- A direct construction defines Q inside K in **16 total K-coordinates**
  using integer coefficients. All point charts, divisions, finite
  correction and equation aggregation are included. A separate
  explicit 52-branch menu also makes the full source pole predicate
  quantitative.

The rank hypothesis is `0 < rank E(Q) = rank E(K)` for an E/Q.
[Zywina's Theorem 1.2](https://arxiv.org/html/2505.16960v1) supplies
such a curve for each quadratic K/Q. Its theorem is not being read as
the same assertion for every extension of higher degree. Finite
exception removal is ineffective. The definition of Q inside K does
not by itself lower the separate counts for Z over Q.

## 3. Finite algebra reduces the field count to ten

The [compressed construction](../papers/number_fields/ten_coordinate_rationality_20261119_v81.tex)
then improves the transparent sixteen-coordinate definition to **ten**.
It retains the input, four elliptic abscissas, four pole parameters and
one encoder. The four point ordinates and two pole roots are stored in
a monic finite algebra of rank 64, with all source divisions explicitly
cleared and guarded.

For completeness the elliptic approximation is carried out in mE(Q),
so its four preimages are rational. All characters of the initial
point algebra merely change ordinate signs. Finite avoidance keeps
both detector inputs nonzero at every such character. The actual
prepared pole witnesses are Zariski dense by the published class-seven
twist theorem, so they can avoid every other-character radicand zero.
This gives the full unit guards without imposing an additional equality.

For six generators r1,...,r6, the finite menu

```math
\alpha_j=\sum_{i=1}^{6}j^{i-1}r_i,\qquad 1\le j\le10081,
```

contains a primitive element at every guarded specialization: each of
the 2016 character pairs forbids at most five values of j. Let Delta_j
be its power-basis determinant and let G_j include Delta_j and every
source guard. The encoder is

```math
\det\bigl(G_j(ZI-M_{\alpha_j})-I\bigr)=0.
```

At a missing guard it equals one. At any other root it recovers a
field-valued character through the eigenvalue Z−1/G_j. This remains
valid when the primitive value is zero. A finite product over the
separator and pole-multiplier menus supplies one integer polynomial of
content one. The complete determinant specification has 2,580,736
factors; it has not been expanded. The paper gives finite degree,
support and coefficient bounds in the chosen circuit data.

Thus, under the stated hypotheses, an N-variable definition of Z over
Q gives one in **10N variables over K**. The ring of integers uses
**10N[K:Q]**. Our N=11 construction gives **110** for the rational
integers in K and **220** for the ring of integers of a quadratic K.
All original Z-over-Q arithmetic assumptions remain in this composition.
These are not quartic-degree or effective-coefficient claims.

The [independent algebra checker](../computation/number_fields/pasten_vidaux_v81/compression_check.py)
checks the scalar clearing, a complete rank-64 split multiplication
algebra, its exact primitive determinant, a zero primitive value,
failed guards, and a nonsplit triangular example recovering the
original roots. Its finite fixtures are separated from the universal
[proof review](../computation/number_fields/pasten_vidaux_v81/compression_proof_review.md).

## 4. What remains open for 25/11 and 165/180

The current general counts **25, 165 and 180**, and the unrestricted-degree
ordinary-abc count **11**, remain unchanged. The new approximation
lemma permits finite local conditions and avoidance of algebraic
equalities; it does not impose a new exact rational-square equality
at the original elliptic detector tuple.

For the eleven-variable route, the concrete target remains to arrange
`1+d` as a rational square at the **actual** detector input, with the
same point and contact tuple, all guards, and the original extraction
inequality. The known identity `P_d(0,s)=4d^4` when `s²=1+d` is useful
only after this realization is proved. Rational points on freely
chosen auxiliary fibres, real recurrence, or finitely many local
squares do not prove it.

For the general quartics, eliminating the shared R² register still
requires an exact complete quadratic compilation with a scalar detector.
The [general-quartic checkpoint](../computation/variable_reductions_v81/general/README.md)
records the new two-term field scalar and the unresolved scaling
compatibility. Its local implications are proved separately from
global completeness and a coordinate saving. No 164/179 construction
or global minimality claim is made.

The next ordinary-abc quartic target is another nonlinear contact
change or sharing with the retained point block. Simply dropping a
register from the new rechart is not justified by a dimension count.
Every proposed saving must retain the entire rational fibre, the
height estimate, and complete integer witnesses.

## 5. Verification and publication scope

Run

```sh
python -B scripts/verify_v81.py --report /tmp/v81-verification.json
```

The gate reconstructs the new coefficients independently, regenerates
the producer output in a temporary tree, checks the field identities
and coordinate census, and checks immutable source and proof-review
hashes. Every earlier regression gate remains in the full workflow.
The manuscripts are compiled and all rendered pages inspected.

These checks establish the stated finite identities and exact data.
They do not prove abc, replace the cited arithmetic theorems, or
constitute an independent audit of every inherited foundation.