# Refinement 12 Nov 2026 (~9:30–10:30 AM PST): is the Q8 Artin L(ρ, s) a Maass newform? (Andersen–Thorner applies)

Closes Selberg–Levinson open question 1(a) (lit/selberg-levinson/open-questions.md) and the "[not yet cited]" caveat in
#selberg-levinson-transfer on /critical-line-hypercomplex. Labels: [known] published, with source; [project] checked here.
**Nothing here is RH or GRH progress.** It identifies which published theorem covers an ordinary classical L-function.

## Question
The page uses three Q8 Artin L-functions L(ρ, s), ρ the 2-dim faithful representation: 2.2304.8t5.b.a, 2.2304.8t5.a.a,
2.42025.8t5.a.a. Andersen–Thorner prove κ_f > 0.0696 (more than 6.96% of nontrivial zeros on Re s = ½) for every holomorphic or
Hecke–Maass newform f on Γ₀(N_f) (trivial nebentypus), any N_f ≥ 1. Is each L(ρ, s) = L(s, f) for such an f?

## Answer: yes — cited chain + computed hypotheses
1. [known] **Theorem used.** N. Andersen, J. Thorner, "Zeros of GL₂ L-functions on the critical line", Forum Math. 33 (2021),
   no. 2, 477–491, doi 10.1515/forum-2020-0104, arXiv 2004.03581, Theorem 1.1: "Let f be a holomorphic or Hecke–Maaß newform on
   Γ₀(N_f) for any N_f ≥ 1. Then κ_f > 0.0696. Under the generalized Ramanujan conjecture, κ_f > 0.0896." (Read on arXiv 12 Nov
   2026; §2 sets f with trivial nebentypus, Laplace eigenvalue λ = ¼ + r², r ∈ ℝ or ir ∈ [0, ½); λ = ¼ is r = 0.) Journal data:
   Andersen's CV and the MaRDI/zbMATH record (zbMATH 7509461).
2. [known] **Dihedral → Maass.** For a quadratic field K and a finite-order Hecke character χ of K with χ ≠ χ^c, there is a
   unique normalized cuspidal Hecke newform g_χ with L(s, g_χ) = L(s, χ, K), attached to ρ_χ = Ind χ; g_χ is holomorphic of weight
   1 if ρ_χ is odd, and **a Maass form with Laplace eigenvalue ¼ if ρ_χ is even** (ρ_χ(c) = ±Id), with Fourier coefficients
   χ(𝔞)·√y K₀(2πN𝔞 y)·2cos(2πN𝔞 x) if ρ(c) = Id and 2sin(…) if ρ(c) = −Id.
   Source: W.-C. W. Li, Z. Rudnick, "Pair arithmetical equivalence for quadratic fields", arXiv 2007.13147, §1.3 (read 12 Nov
   2026). Classical origin: Maass (1949) for real quadratic fields; automorphy of all non-icosahedral 2-dim ρ: Langlands (1980),
   Tunnell (1981), as summarised in A. R. Booker, Ann. of Math. 158 (2003) 1089–1098, introduction (read 12 Nov 2026), which also
   records that even ρ has γ(s) = π^{−s}Γ((s+a)/2)², a ∈ {0,1}.
3. [known, standard] Level = Artin conductor, nebentypus = det ρ (Dirichlet character). The conductor of Ind χ is |d_K|·N(𝔣_χ).
   Not re-read in a primary source today; both are checked numerically below.

## Hypotheses checked [project]
Script maass1112/maass_check.py (PARI via cypari2, /workspace/tools/pari-venv), results maass1112/maass-check-results.json,
log run.log. Method A: PARI `lfunartin` (Artin side, Galois closure + character table). Method B: PARI Hecke L-function
`lfuncreate([bnr, ψ])` for an order-4 ray class character ψ of each quadratic subfield F, with bnr from `rnfconductor(F, K/F)`
(the dihedral/induced side). For n ≤ 20000:

| field | N | Vga | W | subfields F (disc) | ψ ramified at ∞ | d_F·N(𝔣_ψ) | max |a_n(A) − a_n(B)| |
|---|---|---|---|---|---|---|---|
| 2.2304.8t5.b.a | 2304 | [1,1] | −1 | 12, 24, 8 (all real) | both real places | 2304 ×3 | 0 (exact) ×3 |
| 2.2304.8t5.a.a | 2304 | [0,0] | +1 | 12, 24, 8 (all real) | neither | 2304 ×3 | 0 ×3 |
| 2.42025.8t5.a.a | 42025 | [0,0] | −1 | 205, 5, 41 (all real) | neither | 42025 ×3 | 0 ×3 |

Also, for every field: Galois group [8,4] = Q8; `lfuncheckfeq` ≈ 10^−27 … 10^−32 (functional equation holds to working
precision); multiplicativity a_mn = a_m a_n (gcd 1, m, n ≤ 140) exact; |a_p| ≤ 2 for all unramified p ≤ 20000 (Ramanujan bound,
automatic for Artin); and a_p² − a_{p²} = 1 for all unramified p ≤ 141, i.e. det ρ(Frob_p) = 1: **trivial nebentypus**.

Reading: ρ is induced from a finite-order character of a *real* quadratic field (three ways), ρ is even (Vga [0,0] means
ρ(c) = +Id, [1,1] means ρ(c) = −Id; the latter is the "sin" / odd Maass form), det ρ = 1, conductor = d_F·N(𝔣). By (2)–(3)
L(ρ, s) = L(s, f) with f a Hecke–Maass newform on Γ₀(N), trivial character, λ = ¼ (N = 2304 or 42025). By (1):

**κ > 0.0696 for each of the three L(ρ, s)** [known: Andersen–Thorner Thm 1.1, applied; hypotheses checked here]. Under the
generalized Ramanujan conjecture the theorem gives 0.0896; whether the Ramanujan bound for these dihedral f themselves (true:
|a_p| ≤ 2) suffices is NOT claimed — their θ also enters through other forms in the spectral decomposition (open question 1(b)).

## What this is not
- Not a statement about the quaternion "half" product F (the aligned Model A product); F has no functional equation and no
  Z-function (primer §10.2). The theorem is about the classical L(ρ, s) only.
- Not new: positive proportion for these L(ρ) follows from published work; we only matched hypotheses.
- Small-height data (all 43 / 42 zeros with 0 ≤ t ≤ 30 on the line, q8_hardy_z.json) is consistent but is not a proportion.

## Status labels for the page
[known] Andersen–Thorner 2021, Li–Rudnick §1.3, Booker 2003 intro; [project] the PARI identification table above.
Page: /critical-line-hypercomplex#q8-maass-newform; sl-tag-1 text and /critical-line-proportion §10.1 updated to point there.
