# 2I central values: no forced zero in the galpol field

Routine sprint, 13 November 2026, ~12:30–1:20 PM PST. Follows the open item "2I central values" in `open-questions.md` (after `octa-central-certificate-20261113.md` and `octahedral-exponent-20261113.md`). Scripts and outputs: `icosa1113/`.

Labels: [known] established, with source; [project] computed here; [speculative] unproved. **No RH claim.** This computes the conductor, root number and central values of two Artin L-functions. It finds, moves or excludes no zero off the line.

## 1. Field and coefficients

**[project]** K = PARI galpol real Galois field of degree 120, Gal(K/ℚ) = 2I = SL(2,5), ramified only at 2, 11, 23 (`polyhedral-fields-20261112.md`, `poly1112/measure_2I.py`). Frobenius quaternions were already known for unramified p < 20 000 except 17 primes where the automorphism denominators block the F_q test.

- `frob_blocked.py`: those 17 primes (3, 5, 151, …, 13649) via a p-maximal order (`nfinit([Q, list])`), `idealprimedec`, `nfmodprinit`, testing σ(θ) ≡ θ^q mod 𝔓 for every automorphism of the right order. Exactly one hit at each prime. Example: Frob₃ has order 10 (class with Re = 0.809), Frob₅ order 3. Below 20 000 the non-squarefree primes of Q mod q are exactly the blocked list plus 11, 23, so the earlier table was not affected by index divisors.
- `frob_extend.py`: same F_q method as `measure_2I.py` for 20 000 < p ≤ 400 000; 12 further index divisors (Q mod q not squarefree: 30893, 33599, 37337, 40819, 69827, 76079, 76949, 114073, 160183, 254627, 272407, 278591) handled by `frob_nonsqfree.py` (p-maximal method). Total: Frobenius class at all 33 860 unramified primes below 400 000.

a_p = 2 Re(Frob_p) for ρ; the Galois conjugate ρ′ (√5 → −√5) swaps the order-5 traces (√5−1)/2 ↔ (−√5−1)/2 and the order-10 traces (√5+1)/2 ↔ (1−√5)/2. Unramified Euler factor 1 − a_p x + x² (det ρ = 1).

Ramified Euler factors are 1. **[project, argument]** From v_p(disc K) = 120(1 − 1/e) for tame p: e = 6 at 11 (100 = 120·5/6) and e = 3 at 23 (80 = 120·2/3). An inertia group of even order contains −1 (the only involution of 2I), which acts as −1 in a faithful ρ; an order-3 inertia group acts with eigenvalues ω, ω̄. Inertia at 2 is a nontrivial 2-group, so it contains −1. So ρ has no inertia invariants at 2, 11, 23, and the tame conductor exponents at 11 and 23 are 2.

## 2. Conductor and root number

**[project, float numerics]** `lfun_2I.py`: PARI `lfuncreate([a_n, 0, Γ_ℝ(s)², 1, N, ε])` (K totally real, so ρ is even), scanning N = 2^a·11²·23² for a = 2…12 and ε = ±1, with `lfuncheckfeq`:

| | best (a, ε) | lfuncheckfeq | next best |
|---|---|---:|---:|
| ρ | (8, +1) | −55 (−56 at 38 digits) | −14 |
| ρ′ | (8, +1) | −51 (−55 at 38 digits) | −6 |

With ε = −1 at a = 8 the check gives 0 / +1 (no fit). So both conductors are **2⁸·11²·23² = 16 386 304** and both root numbers are **+1**. The bound 2^−55 reflects double-precision coefficients.

## 3. Central values

**[project, float numerics, not interval certificates]**

- L(ρ, ½) ≈ 5.85161879797644, L′(ρ, ½) ≈ −17.1674487527607
- L(ρ′, ½) ≈ 0.513258485079238, L′(ρ′, ½) ≈ −1.50579507034267
- Cross-check: L(ρ, 2) = 1.19529970692672 (functional-equation evaluation) against the Euler product over p ≤ 400 000, 1.1952997062 (agrees to 7·10⁻¹⁰, the size of the truncated tail); L(ρ′, 2) 0.869633217339 against 0.869633216608.

## 4. What this says

- **[project]** In this 2I field nothing forces a zero at ½: both faithful 2-dimensional L-functions have root number +1 and clearly nonzero central values. The forced central zeros used on the page for Q8, 2T and the 2O ρ come from the particular fields (or characters), not from the quaternion group. The same 2O field already showed both signs (ρ: −1, ρ⊗ε: +1).
- **[known]** For quaternion fields both root numbers occur (Fröhlich 1972, Invent. Math. 17, "Artin root numbers and normal integral bases for quaternion fields"); Armitage (1972, Invent. Math. 15, "Zeta functions with a zero at s = ½") gave a quaternion-field example with ζ_K(½) = 0.
- **[speculative]** If a 2I analogue of the 2T/2O aligned-product identity is written down, this field's central exponent would get no contribution from a forced zero of L(ρ), only from ζ(2s)-type factors and whatever other L-values vanish at ½. Not computed.

## 5. Open

- Interval (Arb) certificates of L(ρ, ½), L(ρ′, ½) ≠ 0 and of the root numbers.
- The 2I aligned-product identity (the 2T 3/8 / 2O 5/24 analogue) and its central exponent.
- A 2I field with root number −1 (e.g. among the LMFDB 24T201 records, whose Sign field came back as 0 = not stored in our API pull).
