Which state are you converging? A spin audit of sample-based quantum diagonalization benchmarks on iron–sulfur clusters — supplementary code and data

This is the data-and-code archive for the preprint "Which state are you converging? A spin audit of sample-based quantum diagonalization benchmarks on iron–sulfur clusters."

Sample-based quantum diagonalization (SQD, also called QSCI) selects electronic configurations from quantum-circuit samples and diagonalizes the Hamiltonian in that subspace. Its iron–sulfur demonstrations, on the [2Fe-2S] and [4Fe-4S] clusters, are among the headline evidence for utility-scale quantum chemistry. This work asks a question the original benchmarks did not: when the projected energy converges, has the calculation actually converged to the target spin state? Using a rotation-invariant spin audit, exact determinant-space ⟨S²⟩, invariant S²-Gram spectra across near-degenerate manifolds, and higher spin moments, it finds that no audited execution returns the named singlet at a competitive energy. At the largest published dimension (5.625×10⁷ determinants), the numerically stable lowest root lies about 170 mHa below the published energy but at ⟨S²⟩ = 1.37, nowhere near a singlet, and the released hardware samples show no energy advantage over their own uniform-random control. The recommendation: measured spin moments should become a required benchmark output.

The archive lets a reader reproduce and check the analysis. Start with START_HERE.md inside the zip. The package holds the paper (Markdown, LaTeX, and PDF), every analysis and figure script, the per-run result archives, a reproduction map tying each headline claim to the artifact and command that regenerate it, and the audit trail.

Four large files ship next to the zip as standalone companions rather than inside it: the flagship-dimension amplitude archive and three certification vector blocks. SHA-256 hashes for the zip and all four companions are recorded in SHA256SUMS inside the package, so every file can be integrity-checked before use.