# A second non-maximal octonion order: no functional equation, zeros off ½  (Block 4, Nov 11, 2026, 1:42–1:53 AM PST)

Question (Block 4 plan, item 4): is the Gravesian pattern (4 local zeros on Re s = ½, a mirror pair off it, exact s ↦ 1−s symmetry) typical of non-maximal octonion orders?

> **Update, 11 Nov 2026, 7:50 AM PST:** R contains (1+e1+…+e7)/2, so it sits above the index-2 order S1 (G ⊂ S1 ⊂ R). Every order between G and E8, S1's factor, and a Gorenstein-type test (R passes it) are in `octonion-intermediate-orders-20261111.md`.

## The order  [project, computed exactly]
R = the ring generated by G = Z^8 and h = (1+e1+e2+e4)/2 ({1,2,4} is a Fano line in the project labelling). Closure adds (e3+e5+e6+e7)/2; [R:G] = 4; R is closed under multiplication (checked on all basis products) and contains 1. Script `block4/intermediate_order.py`.

## Counts of left-stable sublattices of 2-power index  [project, double-computed]
- Complete method (`block4/intermediate_order_full.py`): children of each stable L are all stable N with 2L ⊆ N ⊊ L, from the full submodule lattice of L/2L. Complete by Nakayama. Counts for k = 0..12: **1, 0, 3, 0, 1, 0, 3, 0, 1, 0, 3, 0, 1**.
- Independent brute force over every HNF sublattice (`block4/intermediate_bruteforce.py`, `block4/intermediate_conductor.py`): index 2: 0 stable of 255; index 4: 3 of 43,435; index 8: 0 of 6,347,715. Agrees.
- Note: the index-2 breadth-first search used for G is **not** complete here. The left action on R/2R has no common eigenvector over F_2, so stable steps have index 4, consistent with a 2-dimensional simple factor (my reading; a residue field F_4 was not checked directly).

## The local factor  [project, proved — computer-assisted]
The left multiplication algebra of R has index 2^40 in M_8(Z) and contains 2·M_8(Z) (c = 1, shown with a subset of words, so the bound is safe). As for G, primitive stable lattices then have index ≤ 2^8, and the counts to 2^12 give
  Σ_L [R:L]^{-s} (2-part) = (1 + 3x²)/(1 − x⁴),  x = 2^{-s}.
- Zeros: 1 + 3·4^{-s} = 0, so **Re s = log 3 / log 4 ≈ 0.79248**, Im s = (2m+1)·π/(2 log 2) ≈ ±2.26618, … No zeros on Re s = ½.
- No functional equation of the Eichler/Gravesian shape: 1+3y (y = x²) is not self-reciprocal under y ↦ 1/(2y) or y ↦ 1/(4y) up to a constant, and there is no mirror zero at Re s ≈ 0.2075.

## Reading  [project; analogy labelled]
- [project] Among the octonion orders checked so far: the Coxeter (maximal) order has no local factor at all (ζ(8s)); G has a self-dual degree-8 factor with half its zeros on ½; R, between them, has a non-self-dual factor with all zeros off ½. So "local zeros on the critical line" is **not** a general feature of octonion orders; in the quaternion Eichler family it came with Gorenstein/Bass structure and a local functional equation. [project conjecture, untested] R fails a Gorenstein-type condition; this was not checked.
- [analogy] In Grant's zero-movement language, moving from G to R (adding one Hurwitz-type half element) removes the self-dual structure and both kinds of local zeros, leaving a single off-line family. That is a comparison of two separate orders, not a continuous movement.
- No RH content: these are finite local factors.

Figure: `figures/octonion-orders-local-zeros{,-mobile}.{svg,png}` (zeros of both factors in one period).
