# Gravesian octonions Z^8: the 2-local factor Q(x)/(1−x⁴) is now proved (Block 4, Nov 11, 2026, from 1:36 AM PST)

Status change: the formula P(x) = Q(x)/(1−x⁴) for the 2-part of the left- (and right-) stable sublattice count of G = Z^8 moves from **[project conjecture]** (beyond index 2^18) to **[project, proved — computer-assisted, double-computed]**. The period-4 mechanism is also explained (Step 5): left multiplication by 1+e_k.

## Setup
- G = Z^8 with the Fano labelling of the project scripts (lines {i, i+1, i+3} mod 7). A = the Z-algebra of 8×8 integer matrices generated by the left multiplications L_0..L_7 (likewise R_0..R_7 for the right). "Left-stable" means stable under A.
- a_k = number of A-stable L ⊆ G with [G:L] = 2^k; x = 2^{-s}; P(x) = Σ a_k x^k.

## Step 1. Every finite A-quotient of 2-power order has only the trivial simple factor F_2  [project, proved]
- [project, computed exactly] Mod 2 the signs vanish, and the index table e_i e_j = ±e_{t(i,j)} is commutative, associative and has e_i² = 1 mod 2; the map (Z/2)³ → {e_0..e_7} generated by e_1, e_2, e_3 is a bijective homomorphism (`octonion_gravesian_conductor.py`, output `octonion-gravesian-conductor.json`). So G/2G ≅ F_2[(Z/2)³] ≅ F_2[x,y,z]/(x²,y²,z²), a local ring, derived directly. This replaces reliance on the Albuquerque–Majid citation for this point (that paper was still not read).
- On G/2G each L_i acts as a permutation matrix of the regular representation of this local commutative algebra, so G/2G has only F_2 (all L_i acting as 1) as composition factor.
- Every A-stable L is a lattice in the same Q_2-representation, so by Brauer–Nesbitt (semisimplifications of reductions of lattices in one representation agree) L/2L has the same composition factors. Any finite A-quotient L/L' of 2-power order is a subquotient of L/2^mL, so all its composition factors are F_2. Hence every stable L ⊆ G is reached from G by a chain of stable index-2 steps, and the breadth-first searches are complete. (This tightens the earlier "G/2G is local" sentence, which spoke about G only.)

## Step 2. A contains 8·M_8(Z)  [project, computed exactly, double-computed]
- Method 1 (`octonion_gravesian_conductor.py`): close the Z-span of the L_i under left multiplication by generators until the PARI HNF stabilises. Rank 64, [M_8(Z):A] = 2^96, and the least c with 2^c E_ij ∈ A for all i,j is c = 3. Right side identical.
- Method 2 (`block4/conductor_check2.py`): span of all 128 distinct words of length ≤ 5 (numpy products, one HNF). Same index 2^96; 8E_ij ∈ span for all i,j; 4E_ij not all. Right side identical.
- Only "8E_ij ∈ A" is needed below, and method 2 shows it with a subset of A, so it does not depend on the closure being complete.

## Step 3. Homothety reduction  [project, proved]
- If L is stable, so is 2L, with index ×2^8. Call L ⊆ G *primitive* if L ⊄ 2G. Every stable L ⊆ G is uniquely 2^m·L₀ with L₀ primitive and stable. So P(x) = Π(x)/(1−x⁸), where Π(x) = Σ p_k x^k counts primitive stable lattices, and p_k = a_k − a_{k−8}.
- Bound: take v ∈ L \ 2G. Then v is primitive in G ⊗ Z_2, so M_8(Z_2)v = G ⊗ Z_2 and L ⊇ A v ⊇ 8M_8(Z)v, giving L ⊇ 8G (2-adically, which is all that matters for 2-power index). Hence [G:L] ≤ 2^24: **p_k = 0 for k > 24.**

## Step 4. The counts to index 2^26  [project, computed, three runs]
- `octonion_gravesian_levels.py` left 26 and right 26, and `octonion_gravesian_levels_pari.py` 26 (independent PARI HNF + direct stability test) all give
  a_k = 1, 1, 7, 7, 15, 15, 35, 15, 31, 15, 35, 15, 31, 15, 35, 15, 31, 15, 35, 15, 31, 15, 35, 15, 31, 15, 35 (k = 0..26).
  Records: `block4/levels-{left,right,pari}-26.json` and `-run.txt`. (The main-folder JSONs were restored to their k ≤ 18 / k ≤ 12 versions so older references stay valid.)
- Primitive counts: p_k = 1, 1, 7, 7, 15, 15, 35, 15, 30, 14, 28, 8, 16 for k = 0..12, and p_k = 0 for 13 ≤ k ≤ 26 (`block4/proof-check-run.txt`).
- Π(x) = Q(x)(1 + x⁴) exactly, with Q = 1 + x + 7x² + 7x³ + 14x⁴ + 14x⁵ + 28x⁶ + 8x⁷ + 16x⁸.

## Conclusion
**[project, proved — computer-assisted]** P(x) = Q(x)(1+x⁴)/(1−x⁸) = Q(x)/(1−x⁴) for all k, left and right. The functional equation Q(x) = (2x²)⁴Q(1/(2x)) and the zero locations (4 on Re s = ½, 4 at Re s ≈ 1.183 / −0.183) from Block 3 therefore hold for the true factor, not just a fitted one.

Side fact [project, computed]: there are Σ p_k = 192 homothety classes of left-stable Z_2-lattices for the Gravesian order.

## Step 5. Why period 4: a norm-2 left multiplication  [project, proved from exact computations, double-computed]
- Let g = L_{1+e_k} (any k = 1..7). Then det g = N(1+e_k)⁴ = 16.
- [computed exactly, `block4/period4_explain.py`] g normalises A: g A g⁻¹ = A (equal 64-dim HNFs), for every k.
- [known identity, also checked] Octonions are alternative, so L_u² = L_{u²}; with u = 1+e_k, u² = 2e_k, so g² = 2·L_{e_k}, and L_{e_k} is a unit of A (its inverse is −L_{e_k}).
- Consequences [project, proved]: for stable L, gL is stable (A gL = g A L = gL), gL ⊆ L because g ∈ A, [L : gL] = 2⁴, and g²L = 2L_{e_k}L = 2L. So the stable lattices fall into chains …, L, gL, 2L, 2gL, … with index step 2⁴, i.e. g is a "square root of 2" acting on them, the octonion analogue of an Atkin–Lehner element. Counting the first member inside G of each chain gives P(x) = Q(x)/(1−x⁴) with Q(x) = Σ_k #{stable L ⊆ G of index 2^k with L ⊄ gG} x^k.
- [project, computed, a third independent count of Q] `block4/q_direct.py` counts exactly that set level by level: 1, 1, 7, 7, 14, 14, 28, 8, 16, then 0 for k = 9..12, which is Q. So Q(1) = 96 chains, half of the 192 homothety classes.
- [computed, `block4/period4_shift.py`] Each of the 56 left multiplications L_u by a norm-2 element u = e_i ± e_j (i < j, e_0 = 1) preserves left stability on all 223 stable lattices of index ≤ 2^12. None of the 56 right multiplications R_u does (`block4/period4-shift.json`).

## Still open
- [open] A conceptual proof (not computational) that g normalises A, and a lattice-duality proof of Q(x) = (2x²)⁴Q(1/(2x)).
- The Albuquerque–Majid paper itself remains unread; the mod-2 structure no longer depends on it.
- Item 4 of the Block 4 plan was done for one intermediate order: see `octonion-intermediate-order.md`.

No RH content: this is one finite local factor of a non-maximal octonion order.
