C053 - For both ions in the {down,e} subspace, U_i(epsilon) creates an error with probability p...
Verdict: verified
Location: Supp. Mat. S2
Type / expected artifact: math / math
Claim: For both ions in the {down,e} subspace, U_i(epsilon) creates an error with probability p ~= 2 (epsilon/2)^2 for p<<1, so the process is parametrized by epsilon = sqrt(2 p).
Models: extraction claude-opus-4-8; verification gpt-5; verification_chain claude-opus-4-8 -> gpt-5; verdict_chain verified -> verified.
Source location(s): source/supp_content.tex:82 (Supp. Mat. S2).
Conclusion
P(exactly one ion flipped) for U_x(eps)|dd> = 2 sin^2(eps/2) cos^2(eps/2) = eps^2/2 - eps^4/6 + O(eps^6) (sympy); leading coeff 1/2 => p ~= 2(eps/2)^2 = eps^2/2, so eps=sqrt(2p). Numeric eps=0.05 -> 1.2490e-3 (paper hint 1.25e-3); p_one/(eps^2/2) -> 1 as eps->0; round-trip eps=sqrt(2p) recovers p. Clean asymptotic math fact -> verified.
Verification details
Executable rerun: run.py exited 0 in 0.756s; log verification/C053/attempts/R002/run.log.
Output excerpt:
P_one (exact) = 2 sin^2(e/2) cos^2(e/2)
series to O(eps^4): epsilon**2*(3 - epsilon**2)/6
leading coeff P_one/eps^2 -> 1/2 (= 1/2 => p ~= eps^2/2)
eps p_one(numeric) eps^2/2 ratio
0.0500 1.24895868e-03 1.25000000e-03 0.999167
0.0200 1.99973335e-04 2.00000000e-04 0.999867
0.0100 4.99983334e-05 5.00000000e-05 0.999967
0.0050 1.24998958e-05 1.25000000e-05 0.999992
ground-truth check: eps=0.05 -> p_one=1.248959e-03 (paper hint 1.25e-3)
eps=sqrt(2p) with p=0.001: eps=0.044721 -> p_one=9.993335e-04 (target 1.000000e-03)