# Q8 zero counts to height 100, with Turing averaging, per Artin factor

Dated 10 November 2026, afternoon, America/Los_Angeles. Notes and figures only; nothing was deployed.

This is a **count certificate** for the primitive factors of the Dedekind zeta of the Q8 field `2.2304.8t5.b.a`, through height 100. It is not a statement about the Riemann hypothesis, and it is distinct from the existing height-50 argument-principle counts in `q8-extended.py` (Arb/flint contour, Ma bound).

The Artin decomposition is

ζ_K(s) = ζ(s) L(s, χ_8) L(s, χ_12) L(s, χ_24) L(s, ρ)^2,

with ρ the faithful 2-dimensional irreducible representation of Q8. Primitive owners: zeta, the three Dirichlet L-functions, and the Artin L-function once.

## Counts at height 100

| Factor | Degree | Conductor | Gamma shifts | N_+(100) | Weyl main term | Method | Isolated? |
|---|---|---|---|---|---|---|---|
| zeta | 1 | 1 | [0] | 29 | 29.00 | existing Turing certificate at height 20000 | yes, already |
| L(χ_8) | 1 | 8 | [0] (even) | 61 | 61.22 | Rumely/Trudgian, t0=100>50, h=6 | yes, [60.532, 61.439] |
| L(χ_12) | 1 | 12 | [0] (even) | 67 | 67.68 | Rumely/Trudgian, t0=100>50, h=6 | yes, [66.570, 67.449] |
| L(χ_24) | 1 | 24 | [0] (even) | 79 | 78.71 | Rumely/Trudgian, t0=100>50, h=6 | yes, [78.587, 79.415] |
| Artin L(ρ) | 2 | 2304 | [1, 1] | 179 | 179.48 | Booker Theorem 4.5, t0=100, h=8 | yes, N=179.5 in [178.773, 180.216] |

The Artin factor also has an exact central zero from the root number W=−1, excluded from the positive list. Booker's N(t) = Φ(t)+S(t) equals N_+(t)+1/2 on this L-function, because the completed Lambda is purely imaginary on the critical line and arg Lambda starts at π/2 just above s=1/2. Every window width h in {8,10,12,16,20,24,30} isolates the same half-integer 179.5. Including the centre, there are 180 zeros with 0 ≤ Im ρ ≤ 100.

## Sources and conventions  [known]

- **Zeta.** Trudgian's published zeta integral bound requires t > 168π ≈ 528, so it is not applied at T=100. The count 29 is read from the already-certified ordinary zeros of the order-height Turing integral at height 20000, source_id `eb130d8f89bc716ed830276079a08260450e0e9997208f7173cb7d6dc70fba5d`, 22,491 zeros, verification PASS. Last zero at or below 100: 98.83119421819369; next: 101.31785100573138. PARI's `lfunzeros` list through 100 agrees with the certified midpoints to 1.4×10^{-14}.
- **Dirichlet.** Rumely's bound, as recorded in Trudgian, *Improvements to Turing's method*, Math. Comp. (2011), arXiv:0903.1885, Theorem 3.1: for t2 > t1 > 50, |∫_{t1}^{t2} S(t,χ) dt| ≤ 1.8397 + 0.1242 log(q t2 / 2π). This applies at T=100.
- **Artin / Hecke.** Booker, *Artin's conjecture, Turing's method and the Riemann hypothesis*, arXiv:math/0507502, Theorem 4.5, with c_0 ≈ 5.65055, θ=0 (unitary), X=40. Condition (4.19) holds at these heights (t=70 already satisfies it with X=40 and shifts μ_j=1). The L-function is monomial (induced from a Hecke character of the C4 fixed field), hence entire without any appeal to Artin's conjecture.
- **Parity.** All three quadratic subfields of this CM octic are real — Q(√2), Q(√3), Q(√6), fundamental discriminants 8, 12, 24 — so χ_8, χ_12, χ_24 are even. Confirmed by PARI 2.17.2: `lfunparams(lfuncreate(D)) = [D, 1, [0]]` for D=8,12,24, whereas negative D give gamma [1]. An early draft of the script marked 12 and 24 odd; that raised Φ by 1/4 and is corrected.

Φ(t) is Booker's archimedean/conductor phase, using `scipy.special.loggamma` on the critical line (no branch cut for Re > 0). Empirical integrals of S over every window used are well inside the published bounds (typical |∫S| is a few tenths, against B ≈ 2.5 for Dirichlet and B ≈ 5.5 for Artin).

## What is certified, and what is not  [project]

The ordinates themselves are PARI `lfunzeros` candidates, already computed to T=200 in `zero-textures-q8-zeros.json` (decimal precision 40). The Turing integral promotes the *count* at T=100 to an isolated integer (or half-integer), provided the zeros used in the averaging window are genuine. That proviso was checked by evaluating PARI's Hardy Z-function at the quarter, mid, and three-quarter point of every gap in the windows [80,120] (Dirichlet) and [70,130] (Artin):

- every consecutive midpoint changes sign;
- the three interior samples in each gap share a sign (no extra zero in a gap);
- minimum |Z| at midpoints: 0.58, 1.01, 2.06 (Dirichlet), 0.20 (Artin).

This does not replace an interval isolation of each individual ordinate, and it does not extend the existing height-50 argument-principle certificate. It also does not treat the other two Q8 bases (`2.2304.8t5.a.a` and the conductor-42025 field); those remain at the earlier heights.

A pad of 0.02 was used on every candidate when forming upper and lower endpoints. The narrowest gap in (70,130) is 0.078 (Artin), so the pad does not mix adjacent zeros.

## Figures

Desktop and 360px mobile, SVG and PNG, under `figures/`.

- `q8-turing-counts`: isolated N_+(100) against the Weyl main term, one bar per primitive factor.
- `q8-turing-bounds`: the isolating interval at the chosen window width, and (right panel / lower mobile panel) the Artin bounds versus h, squeezing to 179.5.

## Scripts and data

- `q8_turing.py` — Φ, Rumely and Booker integrals, isolation.
- `q8_turing_figures.py` — the figures.
- `q8-turing-results.json`, `zero-textures/q8-turing-results.json` — numeric record.
- `q8-turing-hardy.json` — Hardy sign checks.

No writes to `order-height/`, no edits to `zeta-zeros-build`, no deploy, no messages.

## What is still open

- Turing isolation at T=100 for the other Q8 bases, and a literal interval (Arb) isolation of each Artin ordinate rather than a count.
- Extending the existing argument-principle code `q8-extended.py` to T=100, which is a different method (Ma bound grows like exp(T arcsin(2/3)) and needs more Gauss nodes); not done here.
- Items 3–6 of the 10 November afternoon queue (coefficient audit, negative-D twists, longer number-variance window, octonion analogues).
