verified certificate script asserted: CERTIFIED chi(CS^2) >= 3, by an argument declared exact throughout (no floating-point margin)
1 run; 0.005 CPU-hours; 0.005 h wall-clock; on Apple silicon laptop, one core; about 15 s for the certificate; claude-fable-5-1, wrote the exact certificate.
Settles χ(|CS〉⊗2) = 3 with the verified three-term witness, against the product bound 4. Exact throughout: the 36,720 four-qubit stabilizer states are enumerated from the verifier's parametrization as vectors over Z[i] (reduced-row-echelon flats, strictly upper-triangular Q, l in Z4^k; count checked against 2^4 (2+1)(4+1)(8+1)(16+1), pairwise distinct up to phase, and matched one to one against the numerical dictionary as a consistency check nothing rests on). With P = 4|CS〉⊗2 the component q(s) = <P,P> s - <P,s> P of each state off the target stays in Z[i], and P lies in span(s, t) exactly when q(s) and q(t) are parallel, which is decided by dividing each q(s) by its first nonzero entry in Q(i) and comparing: no two coincide and none vanishes. The design note's numerical search (2026-09-20) found the same with closest approach to parallel 0.763 against the threshold 1 - 1e-6; this certificate removes the margin. Controls: |++++〉 is reported as rank 1, |CS〉 ⊗ |++〉 is spanned by pairs, and on two qubits the pair {|++〉, CZ|++〉} is found for |CS〉. The script also prints χ(CS^3) ≥ 3 by projection monotonicity, (I ⊗ 〈00|)|CS〉⊗3 = |CS〉⊗2 / 2.
Vincent Russo · stabrank verify_challenge · exact: no pair of the 36,720 four-qubit stabilizer states spans |CS〉⊗2, decided in Q(i) by canonicalizing each state's component off the target
{
"schema_version": "0.1",
"orbit": "CS",
"m": 2,
"direction": "lower",
"rank": 3,
"certificate": {
"script": "verify_challenge/cert_cs_m2_rank2.py",
"expect": "CERTIFIED chi(CS^2) >= 3",
"exact": true
},
"provenance": {
"author": "Vincent Russo",
"reference": "stabrank verify_challenge",
"method": "exact: no pair of the 36,720 four-qubit stabilizer states spans |CS>^2, decided in Q(i) by canonicalizing each state's component off the target",
"date": "2026-10-21",
"github": [
"vprusso"
],
"compute": {
"cpu_hours": 0.005,
"wall_clock_hours": 0.005,
"runs": 1,
"hardware": "Apple silicon laptop, one core; about 15 s for the certificate",
"llm": [
{
"model": "claude-fable-5-1",
"role": "wrote the exact certificate"
}
]
}
},
"notes": "Settles chi(|CS>^2) = 3 with the verified three-term witness, against the product bound 4. Exact throughout: the 36,720 four-qubit stabilizer states are enumerated from the verifier's parametrization as vectors over Z[i] (reduced-row-echelon flats, strictly upper-triangular Q, l in Z_4^k; count checked against 2^4 (2+1)(4+1)(8+1)(16+1), pairwise distinct up to phase, and matched one to one against the numerical dictionary as a consistency check nothing rests on). With P = 4|CS>^2 the component q(s) = <P,P> s - <P,s> P of each state off the target stays in Z[i], and P lies in span(s, t) exactly when q(s) and q(t) are parallel, which is decided by dividing each q(s) by its first nonzero entry in Q(i) and comparing: no two coincide and none vanishes. The design note's numerical search (2026-09-20) found the same with closest approach to parallel 0.763 against the threshold 1 - 1e-6; this certificate removes the margin. Controls: |++++> is reported as rank 1, |CS> (x) |++> is spanned by pairs, and on two qubits the pair {|++>, CZ|++>} is found for |CS>. The script also prints chi(CS^3) >= 3 by projection monotonicity, (I (x) <00|)|CS>^3 = |CS>^2 / 2."
}