# Real-axis zero map of maximal quaternion-order zetas (10 Nov 2026, 9:15 AM PT refinement run)

Object: ideal-index zeta of a maximal order in a definite rational quaternion algebra of discriminant D,
Z_D(s) = zeta(2s) zeta(2s-1) prod_{p|D} (1 - p^(1-2s))   (Voight, Thm 26.4.9 / §26.8). r = number of ramified primes (odd).

## Claim [known consequence, derived here from standard facts about zeta]
On the real axis, Z_D has
- simple zeros at every negative half-integer s = -1/2, -1, -3/2, -2, ... (integers from zeta(2s) trivial zeros, half-integers from zeta(2s-1) trivial zeros);
- order r-1 at s = 1/2 (r local zeros against the simple pole of zeta(2s); nonzero value -log(p)/2 when r = 1);
- a simple pole at s = 1 with residue zeta(2)/2 * prod(1 - 1/p);
- nothing else.

Proof sketch. Real zeros of zeta are exactly the negative even integers (zeta(x) != 0 on (-2,0), (0,1), (1,inf); zeta(0) = -1/2), zeta has its only pole at 1, and the local factor 1 - p^(1-2s) vanishes on the real axis only at s = 1/2. At s = -m the factor zeta(2s-1) = zeta(-2m-1) != 0; at s = 1/2 - m (m >= 1), zeta(2s) = zeta(1-2m) != 0. Trivial zeros of zeta are simple, so every listed zero is simple. In (0,1/2) and (1/2,1) both zeta factors are nonzero. Hence the list is complete.

Coordinate comparison (open-questions item 6). In the reduced-norm coordinate w = 2s the same divisor is: simple zeros at every negative integer w = -1, -2, -3, ..., order r-1 at w = 1, pole at w = 2. The mirror centre moves from s = 1/2 to w = 1; the zeros only get renamed, as with the E8 rescaling.

## Check [project, 60-digit numerical, NOT interval]
order-real-axis.py: sign-change scan of s in [-8.2, 0.98], step 1/400, 60 digits, for D = 2, 3, 30. Found exactly 16 roots per D, matching -1/2..-8 to better than 1e-40, each with nonzero derivative. Leading coefficient at 1/2: -0.34657359028 (D=2), -0.549306144334 (D=3), h^2 coefficient -2.45117397417 (D=30), reproducing the earlier values. Residue estimates at s=1 match zeta(2)/2 prod(1-1/p) to ~15 digits. A sign scan cannot see even-order zeros; the analytic argument excludes them.

Caveats: real axis only; complex zero families are on the page already. Nothing here bears on RH, and none of it is an Artin forced zero. Figure: figures/order-real-axis.svg.
