This note separates three layers that should not be conflated:
B0 <-> R <-> B3,3-state Hecke machine on [w, T_3 w, T_11 w],u^m e0, u^m e3, u^m x, u^m y.For the pair factor
R = Z_3[eta]/(eta^2 - 3 eta),
the three CM indecomposables are B0, R, B3. Using the explicit branch maps
i0(x)=(3x,0), q0(a,b)=a,
i3(x)=(0,3x), q3(a,b)=b,
eta(a,b)=(0,3b),
all quiver relations were verified exhaustively modulo 81.
Verified relations:
q0 i0 = 3 id_B0, q3 i3 = 3 id_B3,
i0 q0 = 3 - eta, i3 q3 = eta,
q0 i3 = q3 i0 = 0,
eta^2 = 3 eta.
So every round-trip through the singular mediator carries the arithmetic cost 3 already in the CM quiver, before any stable quotient is taken.
The absence of direct cross-branch arrows is exact over Z_3:
Hom_R(B0,B3) = Hom_R(B3,B0) = 0,
because an R-linear map is multiplication by lambda with 3 lambda = 0 in Z_3, hence lambda = 0. Finite truncations mod 3^k do show torsion ghost maps, but these disappear in the inverse limit:
| modulus | solutions to 3 lambda = 0 mod 3^k |
|---|---|
3 |
[0, 1, 2] |
9 |
[0, 3, 6] |
27 |
[0, 9, 18] |
81 |
[0, 27, 54] |
3-state flag is intrinsicIn the cyclic basis
[w, T_3 w, T_11 w],
every Heegner operator preserves the flag
0 < <T_11 w> < <T_3 w, T_11 w> < <w, T_3 w, T_11 w>.
| operator | preserves <T11w> |
preserves <T3w,T11w> |
top | middle | bottom |
|---|---|---|---|---|---|
T_3 |
True |
True |
0 |
1 |
0 |
T_7 |
True |
True |
2 |
2 |
2 |
T_11 |
True |
True |
0 |
0 |
0 |
T_19 |
True |
True |
0 |
2 |
0 |
T_43 |
True |
True |
1 |
2 |
1 |
T_67 |
True |
True |
1 |
2 |
1 |
T_163 |
True |
True |
2 |
1 |
2 |
The scalar pattern is exactly pair -> v7 -> pair from bottom to top, so the residue machine is a genuine 3-step filtered Hecke object, not just a decorative basis choice.
C6Reducing modulo 3 gives
A = F_3 x F_3[eps]/(eps^2).
Normalizing projective classes so the first coordinate is 1, the six classes are
[[1, 1, 0], [1, 1, 1], [1, 1, 2], [1, 2, 0], [1, 2, 1], [1, 2, 2]]
and the class of T_43 generates all of them.
| Hecke operator | residue coords [e7,epair,n] |
normalized projective class | order |
|---|---|---|---|
T7 |
[2, 2, 2] |
[1, 1, 1] |
3 |
T43 |
[2, 1, 1] |
[1, 2, 2] |
6 |
T67 |
[2, 1, 1] |
[1, 2, 2] |
6 |
T163 |
[1, 2, 0] |
[1, 2, 0] |
2 |
The powers of the generator class [T_43] are
[[1, 1, 0], [1, 2, 2], [1, 1, 2], [1, 2, 0], [1, 1, 1], [1, 2, 1]]
so in particular
[T_43]^3 = [T_163], [T_43]^4 = [T_7],
which is a concrete cyclic order-6 shadow of the singular node.
In the stable two-object quotient, use generators
e0, e3, x, y, u
with
yx = u e0, xy = u e3,
e0 x = x = x e3, e3 y = y = y e0,
e0 e3 = e3 e0 = 0.
The pairwise multiplication table is already a finite-state reducer:
| left | right | result |
|---|---|---|
e0 |
e0 |
{'u_power_added': 0, 'symbol': 'e0'} |
e0 |
e3 |
0 |
e0 |
x |
{'u_power_added': 0, 'symbol': 'x'} |
e0 |
y |
0 |
e3 |
e0 |
0 |
e3 |
e3 |
{'u_power_added': 0, 'symbol': 'e3'} |
e3 |
x |
0 |
e3 |
y |
{'u_power_added': 0, 'symbol': 'y'} |
x |
e0 |
0 |
x |
e3 |
{'u_power_added': 0, 'symbol': 'x'} |
x |
x |
0 |
x |
y |
{'u_power_added': 1, 'symbol': 'e3'} |
y |
e0 |
{'u_power_added': 0, 'symbol': 'y'} |
y |
e3 |
0 |
y |
x |
{'u_power_added': 1, 'symbol': 'e0'} |
y |
y |
0 |
Every nonzero word in e0,e3,x,y of length at most 8 reduced to one of the four normal forms
{ u^m e0, u^m e3, u^m x, u^m y }.
Sample reductions:
| word | reduction |
|---|---|
e0 |
u^0 e0 |
e3 |
u^0 e3 |
x |
u^0 x |
y |
u^0 y |
e0 e0 |
u^0 e0 |
e0 x |
u^0 x |
e3 e3 |
u^0 e3 |
e3 y |
u^0 y |
x e3 |
u^0 x |
x y |
u^1 e3 |
y e0 |
u^0 y |
y x |
u^1 e0 |
So the stable category is the reduced interaction algebra, while the CM skeleton is the unreduced mediator graph. The arithmetic node contains both levels at once; the stable quotient is not the whole story.