Vibrational Sample-Based Quantum Diagonalization with Unitary-Cluster-Jastrow Ansätze on IBM QPUs

Complete per-parameter-setting measurement ledgers supporting "Vibrational Sample-Based Quantum Diagonalization with Unitary-Cluster-Jastrow Ansätze on IBM QPUs." This dataset contains four self-contained, interactive HTML tables that can be opened directly in any browser with no server or installation required. The files report every real-hardware result underlying the paper's figures: 240 Ground-State Cells: 120 on ibm_kingston and 120 on ibm_phoenix across 3 molecules (H2O, CH2O, CH2ClF) × 4 modal truncations × 2 one-mode bases × 5 ansätze. 24 Excited-State Comparison Cells: Side-by-side results on both backends. 12 Uniform-Random-Sampler Control Cells. Each row provides: Circuit CZ gate count and depth One-hot retention percentage Before- and after-recovery energy Number of recovery iterations Real IBM Quantum Job ID Transpiled circuit physical-qubit layout and coupling edges (where captured) Note on Data Completeness: Two fields present in the schema—per-cell classical/transpile/queue timing breakdown and current device-calibration snapshots for a subset of cells—were only captured for part of the measurement campaign and are left blank (not zero) where unavailable. The paper's text provides verified aggregate and ballpark timing figures for the campaign as a whole. Job IDs allow independent verification against IBM Quantum's public job-retrieval API for as long as the jobs remain retrievable. File Descriptions recovery_table_artifact_kingston.html Contents: Ground-state results on real ibm_kingston (IBM Heron r2) hardware. Coverage: 120 rows—one per (molecule, modal count, basis, ansatz) combination across H2O, CH2O, and CH2ClF (2, 4, 6, 8 modals per mode; harmonic-oscillator and VSCF-modal bases; UVCCSD, VLUCJ, VIm-uCJ, Vg-uCJ, CHC). Details: Each row reports the real job ID, shot count, transpiled CZ count and depth, one-hot retention, before- and after-recovery energy, number of self-consistent recovery iterations, and actual physical qubits and coupling edges used (for 116 of 120 rows). The remaining 4 rows (CH2ClF, 6 modals/mode, harmonic-oscillator basis, non-UVCCSD ansätze) have no saved transpiled-circuit file for this specific layer setting and are left blank rather than estimated. Live device-calibration and per-job QPU billing data are included where available. recovery_table_artifact_phoenix.html Contents: Ground-state comparison run on real ibm_phoenix (IBM Nighthawk r2) hardware. Coverage: 120 rows—identical parameters and coverage as the Kingston ledger. Details: Physical-qubit layout and coupling edges are complete for all 120 rows. excited_comparison_artifact.html Contents: Excited-state fundamental-frequency results comparing both backends side by side. Coverage: 24 rows—one per (molecule, modal count, basis) combination. Details: Nests the ibm_kingston and ibm_phoenix UVCCSD results alongside the classical VCC(M) target used for comparison. Reports the recovered zero-point energy, fundamental transition energy, one-hot retention for both ground- and excited-state circuits, and real job IDs for both backends. recovery_table_artifact_uniform.html Contents: Uniform-random-sampler control. Coverage: 12 rows (3 molecules × 4 modal counts in harmonic-oscillator basis). Details: Compares the self-consistent recovery loop's result when fed a structureless uniform random sample (no circuit, no backend, zero QPU cost) against the best real-hardware result at the matching parameter setting. Fully populated with no missing fields.