Fix the elliptic curve E : y2 + y = x3 − 2x + 1 of conductor 163 (Cremona label 163a1), with j-invariant −218 · 33 · 53 · 233 · 293 / 1632 and Mordell–Weil rank one, generator P = (0, 0). Write f = ∑ an qn for its associated weight-2 newform.
For each prime p ≠ 163, the Hecke operator Tp acts on f by the scalar ap ∈ Z. We study the six smallest such primes: p ∈ {3, 7, 11, 19, 43, 67}. These are the six gates of the automaton.
Gate classification
For an integer n ≠ 0, write v3(n) for its 3-adic valuation and u3(n) := (n / 3v3(n)) mod 3 ∈ F3× for its leading 3-adic unit. The gate type is determined by ap as follows:
If ap = 0, then Tp is an annihilator.
If ap ≠ 0 and v3(ap) = 0, then Tp is a permutation gate with multiplier u3(ap) ∈ F3×. Multiplier 1 gives an identity; multiplier 2 gives a swap.
If ap ≠ 0 and v3(ap) ≥ 1, then Tp is a shift gate of depth v3(ap) and unit multiplier u3(ap).
Applied to conductor 163:
p
ap
v3(ap)
u3(ap)
Gate type
3
0
—
—
Annihilator
7
2
0
2
Swap (involution)
11
−6
1
1
Shift (+1, ×1)
19
−6
1
1
Shift (+1, ×1)
43
7
0
1
Identity
67
−2
0
1
Identity
Proof: direct computation. 7 ≡ 1 (mod 3), so u3(7) = 1 and T43 is identity. −2 ≡ 1 (mod 3), so u3(−2) = 1 and T67 is identity. 2 ¬≡ 1 (mod 3), so T7 is swap. −6 = −2 · 3 has v3 = 1 and unit −2 ≡ 1 (mod 3), so T11 and T19 are shifts.
with six input symbols {Tp : p ∈ {3, 7, 11, 19, 43, 67}} and transition function δ : Q × {Tp} → Q given by:
δ(VAC, Tp) = VAC (VAC is absorbing)
δ((r, k), T3) = VAC (annihilator)
δ((r, k), Tp) = (u3(ap) · r mod 3, k + v3(ap)) for p ≠ 3
Since u3(ap) ∈ F3× = {1, 2} for all p ≠ 3 and r ∈ F3×, the residue component never collapses to 0 under the non-annihilator gates. The only path into VAC from a unit state is through T3.
Three theorems
Let Σ = {T3, T7, T11, T19, T43, T67} and Σ* = Σ \ {T3}. For a word w = g1 … gn ∈ Σn and a state q ∈ Q, write δ(q, w) for the iterated transition.
Theorem 1 (Monotonicity)
For every word w ∈ (Σ*)* and every unit state (r, k), the result δ((r, k), w) is again a unit state (r', k') with
k' = k + ∑i v3(ap(gi)) ≥ k.
The shell coordinate is a non-decreasing additive counter, strictly increasing iff w contains at least one shift gate T11 or T19.
Proof. Every g ∈ Σ* has either v3(ap) = 0 (swap, identity) or v3(ap) = 1 (shifts T11, T19). The transition (r, k) ↦ (u3(ap) · r mod 3, k + v3(ap)) never produces VAC since u3(ap) ∈ {1, 2} and r ∈ {1, 2}. Iterating, the shell coordinate accumulates ∑ v3(ap(gi)) ≥ 0, with strict inequality iff some gi ∈ {T11, T19}.
Theorem 2 (Boundary Z/2Z)
Let π : Q \ {VAC} → F3× be projection to the residue coordinate. The induced action of Σ* on F3× factors through the group F3× ≅ Z/2Z, with T7 ↦ (nontrivial involution) and T11, T19, T43, T67 ↦ (identity). In particular SWAP2 = ID.
Proof. The residue acts by multiplication by u3(ap) ∈ F3×. From the gate table: u3(a7) = 2 while u3(ap) = 1 for p ∈ {11, 19, 43, 67}. Since F3× = {1, 2} has order 2, we have 22 = 4 ≡ 1 (mod 3).
Theorem 3 (Confinement)
Let S ⊆ F3× × Z≥0 be the reachable set from (1, 0) under (Σ*)*. Then S = F3× × Z≥0, but the subset {(r, 0) : r ∈ F3×} (the matter or boundary sector) is reachable only by words containing no shift gate. Once T11 or T19 has been applied, the trajectory enters the vacuum sector F3× × Z≥1 and never returns. Equivalently: the 3-adic ideal m = 3Z3 is forward-invariant under Σ*, and the matter complement Z3× is forward-absorbing into m.
Proof. Reachability: T7 realises both residues from r = 1, and T11k reaches every shell k ≥ 0. Forward invariance of m: by Theorem 1, the shell coordinate never decreases, so once k ≥ 1 it remains ≥ 1. Absorption: both shift gates have v3 = 1 > 0, so a single application sends k = 0 to k = 1.
Corollary. The annihilator T3 is the unique gate that exits both sectors into VAC. Removing T3 from the alphabet yields a deterministic automaton on F3× × Z≥0 in which the matter sector is one-way and the vacuum sector is closed.
The five numbers
(i) Sum of squared eigenvalues. ∑p ap2 = 0 + 4 + 36 + 36 + 49 + 4 = 129 = 3 · 43. The sum factors through the Heegner set.
(ii) Involution. SWAP2 = ID on residues (Theorem 2).
(iv) Kernel sum. The kernel of the residue action Σ* → F3× is the submonoid generated by {T11, T19, T43, T67}, with ∑p ∈ ker ap2 = 36 + 36 + 49 + 4 = 125 = 53.
(v) Shift symmetry. The two shift primes contribute equally: a112 = a192 = 36 = 4 · 32.
Items (iv) and (v) are stated as observations on the gate table, not as structural theorems. Whether 125 = 53 is meaningful (e.g. as a degree of an associated congruence) or coincidental is open.
What is not proved here
This paper makes only the claims stated above. In particular:
We do not claim the formal-group orbit of the Mordell–Weil generator P ∈ E(Q) is monotone under multiplication-by-n. Numerical data show the shell pattern v3(z(nP)) is not monotone in n beyond the first few rungs. The monotonicity proved here is a property of gate sequences acting on the state space, not of the integer orbit on the curve.
We do not prove the automaton is universal among curves with similar Hecke data. The empirical universality classes observed across ten conductors are deferred.
The connection to the singularity category Dsg(R) for R = Z3[η]/(η2 − 3η) is sketched only as motivation; the Z/2Z of Theorem 2 is established here purely as a property of residue arithmetic.
Reproducibility
The gate classification and the arithmetic of ∑ ap2 = 129 are reproduced by src/heegner/heegner_gate_automaton.py, which uses Sage to compute ap(E) from the curve equation directly and applies the gate classification rule. The full transition table of A163 truncated to shell depth 4 is emitted as heegner_gate_automaton.json.