# Octonion mirror lines: local reciprocity, formal completions, and a dual obstruction

2026-11-11 10:52 AM PST. Grant's continuation t192u. Scripts/data under `mirror1111/`. No messages sent.

## Plain explanation for Grant

**[project, proved] A mirror does not trap zeros on itself.** G's numerator has a mirror at ½, yet four zeros sit on it and four sit off in reflected pairs. R has a different numerator mirror, about 0.7925. S₁ has none. None of those is the inherited zeta reference at 1/16.

We can write an exact equation with a meromorphic, prime-2 “gamma-like” ratio. But it is made directly from the already known correction polynomial, so it contributes no new restriction on zeros. A formal mirrored partner can also be written down. In the same additive-index normalization its first coefficients are irrational, whereas numbers of ideals must be whole numbers. It is not a counted dual order of the tested form.

**[project, proved] Two useful positive results:** a short quartic argument proves exactly why four G zeros land on ½, and linearized alternativity gives a conceptual proof of the period-four operator normalizer for **G and S₁**. That normalizer had previously only been checked by computer. S₁ has the same normalizer but no reciprocal numerator, so the normalizer alone cannot supply a functional equation.

## 1. Name the object and the mirror correctly

The parameter s is complex; x=2^(-s). A mirror **center c** means the analytic reflection **s↦2c−s**, not s↦c−s. With real coefficients, the zero-set reflection in the upper half-plane is s↦2c−conjugate(s), so horizontal arrows in the figure preserve height.

Three objects must be kept distinct:

- numerator Q_S(x);
- full local factor F₂=Q_S(x)/(1−x⁴) (E₈: 1/(1−x⁸));
- whole global correction D_S=(1+x⁴)Q_S (E₈: 1), so Z_S(s)=ζ(8s)D_S(2^(-s)).

A monomial reciprocal identity has the form Q(x)=K x^d Q(A/x), A>0. It gives c=−log(A)/(2log2). Rescaling by powers of 2 only changes K and the way A is written. The numerator degree and the endpoint coefficients fix K when Q(0)=1. Anti-palindromicity after positive rescaling is impossible for these positive-coefficient polynomials: comparing the constant and leading terms forces a positive multiplier, not a minus sign.

### Exact classification [project, proved]

| object | identity or result | center |
|---|---|---|
| Q_G, degree 8 | Q_G(x)=16x⁸ Q_G(1/(2x)) | **½** |
| Q_S₁, degree 5 | no positive-real reciprocal center | none |
| Q_R=1+3x² | Q_R(x)=3x² Q_R(1/(3x)) | **log3/log4≈0.792481** |
| B=1+x⁴ | B(x)=x⁴B(1/x) | **0** |
| Q_E8=1 | constant, hence no distinguished root mirror | no distinguished center |
| D_G,D_S₁,D_R | none has a single positive-real reciprocal center | none |
| F₂ for G,S₁,R | no monomial self-functional equation | none |
| F₂ for E₈ | F₂(s)=−2^(8s)F₂(−s) | **0**, a pole reflection |
| global E₈ completion ξ(8s) | ξ(8s)=ξ(1−8s) | **1/16** |

The R center is log3/log4, not a reflection about ½; writing A=2^(-b) gives b=log₂3, not an integer power. Calling R's numerator “without any functional equation” was too broad: it lacks the half-line reflection, but does have this shifted elementary reciprocal identity.

**Why S₁ has none.** For Q=Σa_k x^k, a₀=1, a_d>0, reciprocity forces K=a_d, A^d=a_d^(-2), and a_(d−k)=a_d a_k A^k. For Q_S₁ the k=1 coefficients force A=½, but A⁵ must equal 1/196, contradiction. The global D tests in the preceding note use the same exact criterion: G and S₁ force A=½ but fail the degree relation, while R forces A²=1/9 but would need A⁶=1/9. These are rational contradictions, not failed root searches.

**Why full F₂ is different.** Its genuine poles lie on Re s=0 (the Q and denominator are coprime). A self-equation with a nonvanishing monomial multiplier must reflect that pole line to itself, forcing center0. But Q_G,Q_S₁,Q_R have no center0 reciprocity. Thus their numerators' mirrors are not simple monomial self-equations for the whole rational local factor. For example G's actual rational identity is

    F_G,2(s) = 2^(4−8s) * (1−2^(4s−4))/(1−2^(−4s)) * F_G,2(1−s),

an equation with an extra rational multiplier. For E₈ its local pole reflection at0 also does not conflict with its global1/16 mirror: the prime-2 denominator cancels when the whole global zeta is assembled.

## 2. Why exactly four G roots lie on the half-line

**[project, proved]** Let

    Q_G(x)=1+x+7x²+7x³+14x⁴+14x⁵+28x⁶+8x⁷+16x⁸.
    u = x + 1/(2x),
    P(u)=8u⁴+4u³−2u²+u−3.

A direct polynomial expansion gives

    Q_G(x) = 2x⁴ P(x+1/(2x)).

The transformation makes the reciprocal symmetry automatic. On the critical circle |x|=1/√2, writing x=e^(iθ)/√2 gives u=√2 cosθ, real in [−√2,√2]. Conversely every real u strictly inside that interval gives two distinct roots

    x=(u ± i√(2−u²))/2,

both of modulus1/√2, hence Re s=½.

It remains to count the real roots of P. A Sturm chain, with each member positively rescaled for readability, is

    P,
    32u³+12u²−4u+1,
    44u²−28u+97,
    102u+133,
    −1.

Signs at −∞ are (+,−,+,−,−), giving 3 variations; at +∞ they are (+,+,+,+,−), giving 1 variation. Therefore P has **exactly two real roots**. Moreover

    P(−5/4)=139/32>0, P(−1)=−2<0,
    P(0)=−3<0, P(1)=8>0.

So those roots lie in (−5/4,−1) and (0,1), both strictly inside (−√2,√2). They give exactly four circle roots. The other two P roots are nonreal, so their four x-preimages cannot lie on the circle, because circle points always give real u. The discriminant −4585472 is nonzero, consistent with all roots being simple. This is a complete rational-algebra/Sturm proof; it does not infer circle membership from rounded moduli. Approximate real u-roots −1.10572463387 and 0.69914840048 are only diagram coordinates.

### Why the quaternion Chebyshev theorem does not transfer

The earlier Solomon/Chebyshev argument proves that **all** transformed roots lie in the real interval for its specific weighted-square family. Reciprocity alone proves no such confinement. Even the positive palindrome

    1+6x²+4x⁴ = (2x²)² [1+6(1/(2x))²+4(1/(2x))⁴]

has no critical-circle roots: y=√2x gives y⁴+3y²+1, whose y² roots are (−3±√5)/2, both off the unit circle in modulus. G itself already supplies the relevant arithmetic counterexample: four roots on its mirror, four off. No RH implication follows.

## 3. A meromorphic gamma-like equation exists, formally

Let r(s)=1/8−s, A=2^(-1/8), and H_S(s)=ξ(8s)D_S(2^(-s)). Since ξ(8s)=ξ(8r(s)), we have **exactly**

    H_S(s) = gamma_S(s) H_S(r(s)),
    gamma_S(s) = D_S(2^(-s))/D_S(2^(-r(s))),
    gamma_S(s) gamma_S(r(s)) = 1.

Label: **[project, proved; formal meromorphic identity]**. It is meaningful as an identity of meromorphic functions, including by continuation at zeros/poles of the quotient. It is not an equation with a nonvanishing conductor multiplier, and it gives no new zero constraint: the ratio has exactly the zeros and poles needed to compensate for the absent mirror partners. Calling this a newly discovered arithmetic gamma factor would be misleading.

Two modifications are immediately possible but do not preserve the original arithmetic zero set:

- divide by D to recover ξ(8s), removing order-specific zeros;
- multiply by the reflected D to obtain ξ(8s)D(2^(-s))D(2^(-r(s))), symmetric by construction but adding missing mirror zeros.

This does not exclude another genuinely motivated completion in a broader specified class. It settles the above formal constructions and the usual-completion obstruction, not a universal nonexistence theorem.

## 4. Formal dual pairing and why it is not an ideal-count dual

For D(x)=Σa_k x^k of degree d, a₀=1, a_d>0, define its uniquely normalized reciprocal partner at the inherited center by

    D^vee(x) = x^d D(A/x)/(a_d A^d),  A=2^(-1/8).

It has constant term1. With H^vee(s)=ξ(8s)D^vee(2^(-s)), a direct calculation proves

    H_S(s) = a_d 2^(−ds) H^vee(r(s)).

This is a genuine **formal paired identity**. But the forced coefficients are

    b_k = (a_(d−k)/a_d) 2^(k/8).

For the three corrections:

| S | first problematic coefficient | minimal polynomial |
|---|---|---|
| G | b₁=2^(1/8)/2, coefficient at index2 | 128t⁸−1 |
| S₁ | b₁=2^(1/8), coefficient at index2 | t⁸−2 |
| R | b₂=2^(1/4)/3, coefficient at index4 | 81t⁴−2 |

They are irrational. Multiplying by ζ(8s) cannot repair those coefficients: its first nonconstant Dirichlet term is at index2⁸=256. An ordinary additive-index ideal zeta must have nonnegative **integer** coefficients. Therefore this normalized reciprocal partner is **not** an ideal-count zeta. In particular it is not any of G,S₁,R,E₈, nor any integral finite-correction dual in this same prime-2 normalization with the usual nonvanishing monomial pairing multiplier. This is the precise scope of the obstruction; changing the counting object, exponent convention, allowed gamma factors, or inherited family is outside the claim.

### G's trace dual and conductor candidates

The trace pairing tr(x·conjugate(y)) has Gram matrix2I₈ on G. Thus **G^#=(1/2)G**. It is a fractional lattice, not an order: (1/2·1)²=1/4·1 is not in (1/2)G. Its right multiplier order is again G (homothety cancels in the defining stability condition). As a G-stable fractional lattice it also has the same normalized sublattice count as G, by scaling. This does not produce a new counted order or module with a different global factor.

Fix a maximal Coxeter order O containing G. The largest O-ideal contained in G is **2O**, because 2O⊆G, O⊄G, and the known O-ideals are exactly mO. Its multiplier order is O, giving the already known E₈ factor **if it is counted as an ideal lattice for its own multiplier order**. Counting G-stable sublattices of 2O is a different module-zeta problem, not the same as counting O-ideals. No such module count is silently identified with the E₈ count here; the irrational-coefficient obstruction still rules out any candidate order or module whose normalized index zeta is asserted to be the forced finite-polynomial reciprocal partner above. This order conductor is not the multiplication-algebra conductor **8M₈(Z)** from the earlier computation. We do not silently assume an associative invertible-different formalism for octonions.

## 5. An actual Atkin–Lehner-type operator: conceptual period-four proof

**[project, proved]** This strengthens the previous numerical normalizer check for G and S₁.

Let T_i be right multiplication by an imaginary basis unit e_i on the 8-dimensional octonion vector space, and let A_S be the associative algebra of linear operators generated by all right multiplications from S. Linearized right alternativity implies

    T_i²=−I,
    T_i T_j + T_j T_i = 0  (i≠j).

These are identities of linear operators, not assertions that octonion products associate. Put g=I+T_i, so g^(-1)=(I−T_i)/2. Then

    g T_i g^(-1) = T_i,
    g T_j g^(-1) = T_i T_j  (j≠i).

The inverse conjugation replaces T_iT_j by −T_iT_j. Thus g normalizes A_G, without an HNF computation.

For S₁=G+Zω, ω=(1+e₁+⋯+e₇)/2, the additional generator is T_ω=(I+ΣT_j)/2. The same anticommutation calculation gives

    g T_ω g^(-1) = I + T_i T_ω,
    g^(-1) T_ω g = T_i − T_i T_ω.

Both lie in A_S1. Hence g also normalizes A_S1. Moreover

    det(g)=16,  g²=2T_i.

For an A_S-stable lattice L, gL is stable, gL⊆L, [L:gL]=2⁴, and g²L=2L because T_i is an invertible operator in A_S. This proves the chain mechanism **L ⊃ gL ⊃ 2L** and explains the period-four denominator for both orders. Exact matrices from the preceding Fano certificate corroborate every identity for all seven choices of i; they are not needed for the conceptual derivation.

This operator is a useful Atkin–Lehner-type analogy, **not an inner octonion-order automorphism or a new dual order**. S₁ has the same normalizer but no reciprocal Q, so normalization/periodicity alone does not force a numerator palindrome or a global FE. A conceptual lattice-duality explanation for the specific Q_G palindrome remains separate and unresolved; its explicit polynomial identity and the circle count above are proved.

## 6. Bushnell–Reiner: what was checked and what does not transfer

Primary source re-read this turn as rendered pages156–158: Bushnell–Reiner, *Zeta Functions of Arithmetic Orders and Solomon's Conjecture*, Math. Z. **173** (1980),135–162; [EuDML record](https://eudml.org/doc/172975). Local PDF `night-sources-20261110/bushnell-reiner-1980.pdf`; page images `mirror1111/br-23.png` through `br-25.png`.

Theorem2 and Remark1 use an associative arithmetic order, its contragredient lattice, a trace/Fourier duality, and local zeta integrals on the multiplicative group of a simple algebra. Remark3 writes a monomial reflection for the quotient by a maximal-order zeta. The Appendix explicitly assumes a simple Qₚ-algebra. The earlier note `night-br-atkin-lehner-20261111.md` explains a plausible extension when a **quaternion** trace-dual lattice is a normalizing translate; it remains an argument reading, not a source theorem or a line-by-line checked extension.

An octonion algebra is nonassociative, and its invertible elements are not an associative multiplicative group. Merely observing G^#=(1/2)G or the normalizer g does not establish the hypotheses or import that proof. Our formal gamma identity and operator normalizer are proved directly; no Bushnell–Reiner theorem for octonion ideal counts is asserted. R's shifted Q mirror also shows why “no half-line FE” must not be restated as “no identity whatsoever”.

## Files / status

- `mirror1111/compute.py`, `results.json`: exact reciprocal identities, rational Sturm proof, formal-dual coefficients/minimal polynomials, and normalizer checks. Run with `/workspace/tools/pari-venv/bin/python` (sympy/numpy/mpmath).
- `mirror1111/figure.py`: desktop/mobile diagrams; uses previous numerical root positions and previously certified inherited ordinates. Arrows show zero-set mirrors s↦2c−conjugate(s); red crosses are explicitly missing reflected partners, not extra zeros.
- `figures/octonion-mirror-lines{,-mobile}.svg` → site `img/hyper-octonion-mirror-lines{,-mobile}.svg`.
- PNG: `/workspace/math/octonion-mirror-lines-20261111.png`, mobile variant also saved.
- `mirror1111/test_acceptance.py` checks proof outputs and built section. No new numerical ideal enumeration, no new RH verification, no outside referee.
- Closed now: conceptual normalizer/period-four mechanism for G and S₁; exact four-on-circle reduction for Q_G; reciprocal/formal-dual classification in the stated class. Open: a genuinely arithmetic alternative completion outside that class, and a conceptual duality explanation of the G numerator palindrome.
- Publication receipt goes in `mirror1111/live-receipt.json` and `/workspace/math/DEPLOY-LOG.md` after live verification.
