verified certificate script asserted: CERTIFIED chi(cat_5) >= 3, by an argument declared exact throughout (no floating-point margin)
3 runs; 0.01 CPU-hours; 0.01 h wall-clock; on Apple silicon laptop, one core; claude-opus-5-5, designed, wrote and ran the certificate and its cross-checks.
Machine check of Lemma 3 in the appendix of Qassim, Pashayan, and Gosset (arXiv:2106.07740), χ(cat_5) > 2, which they prove by a computer comparison of Pauli spectra; the rank is theirs, the check is new. With bounds/cat-m5-upper-3.json the cell is exact, χ(cat_5) = 3, as in their Table 1. Method: |cat5〉 lies in the stabilizer code Z^{(x)5} = +1, and projecting a decomposition onto that code keeps each term a stabilizer state or zero; the CNOT chain onto the last qubit identifies the code with four qubits, so χ(cat_5) is the rank of the four-qubit state c5(y) = cat_5(y, parity(y)), whose entries are fourth roots of unity. All 36,720 four-qubit stabilizer states are generated exactly as Gaussian-integer vectors (count asserted). With q(s) = s - s0 t, exact over Z[i], the target lies in the span of two states iff their quotients are parallel, decided by exact reduced Gaussian-rational keys; no floating point enters the decision. Controls in the script: compressed cat_4 (χ = 2) shows its 15 rank-2 classes, a stabilizer target is flagged as rank 1, and the closed form of |catm〉 matches stabrank_verify.cat_vector exactly for m = 4, 5, 6. About 8 s on one core.
Michel de Messieres · stabrank verify_challenge · exact rank-2 exclusion over the 36,720 four-qubit stabilizer states after compressing |cat5〉 into the even-weight code Z^5 = +1
{
"schema_version": "0.1",
"orbit": "cat",
"m": 5,
"direction": "lower",
"rank": 3,
"certificate": {
"script": "verify_challenge/cert_cat_m5_rank2.py",
"exact": true,
"expect": "CERTIFIED chi(cat_5) >= 3"
},
"provenance": {
"author": "Michel de Messieres",
"reference": "stabrank verify_challenge",
"date": "2026-10-16",
"github": [
"MicheldeMessieres"
],
"method": "exact rank-2 exclusion over the 36,720 four-qubit stabilizer states after compressing |cat_5> into the even-weight code Z^5 = +1",
"compute": {
"cpu_hours": 0.01,
"wall_clock_hours": 0.01,
"runs": 3,
"hardware": "Apple silicon laptop, one core",
"llm": [
{
"model": "claude-opus-5-5",
"role": "designed, wrote and ran the certificate and its cross-checks"
}
]
}
},
"notes": "Machine check of Lemma 3 in the appendix of Qassim, Pashayan, and Gosset (arXiv:2106.07740), chi(cat_5) > 2, which they prove by a computer comparison of Pauli spectra; the rank is theirs, the check is new. With bounds/cat-m5-upper-3.json the cell is exact, chi(cat_5) = 3, as in their Table 1. Method: |cat_5> lies in the stabilizer code Z^{(x)5} = +1, and projecting a decomposition onto that code keeps each term a stabilizer state or zero; the CNOT chain onto the last qubit identifies the code with four qubits, so chi(cat_5) is the rank of the four-qubit state c_5(y) = cat_5(y, parity(y)), whose entries are fourth roots of unity. All 36,720 four-qubit stabilizer states are generated exactly as Gaussian-integer vectors (count asserted). With q(s) = s - s_0 t, exact over Z[i], the target lies in the span of two states iff their quotients are parallel, decided by exact reduced Gaussian-rational keys; no floating point enters the decision. Controls in the script: compressed cat_4 (chi = 2) shows its 15 rank-2 classes, a stabilizer target is flagged as rank 1, and the closed form of |cat_m> matches stabrank_verify.cat_vector exactly for m = 4, 5, 6. About 8 s on one core."
}