OPERA-GC: software and data for globular-cluster proper-motion analysis

Abstract

OPERA-GC version 1.3.0 contains Python source code, configuration, bundled inputs, numerical results and method documentation for heterogeneous HST and Gaia globular-cluster proper-motion analysis. Product records retain catalogue selection, radial support, uncertainty and reference-frame definitions. The implementation applies catalogue masks, rotates proper motions and measurement errors, fits separate component likelihoods, restricts aggregate comparisons to measured radial overlap, joins HST photometry by source ID, matches catalogues with mutual nearest neighbours and evaluates an HST selection grid. Method documentation defines the likelihood, optimization bounds, inferred annuli, output flags and statistical approximations; source maps link operations to implemented functions. The twelve-cluster transfer includes the NGC 6397 benchmark. Its published HACKS profile has seven bin counts, 42, 96, 96, 96, 96, 96 and 95, summing to 617 profile tracers. All 324 declared sensitivity variants satisfy the operational profile-discrepancy rule; 97 also have 494–740 stars inside the inferred annuli. There are 282 variants with usable fits for both same-star components. Variant 244 has 417 selected HST stars, 393 inside those annuli and 24 also matched to Gaia. A dictionary defines all 70 columns in the six principal Stage 2E CSV tables. Run python scripts/reproduce.py --all from the software directory after installing requirements.txt. The recorded Linux/Python 3.12 execution checks source integrity, runs 64 stage tests and recomputes the 324-variant Stage 2E envelope using bundled inputs. Of the 64 tests, 63 are retained original tests and one verifies independence from document templates. Stage 1B/1C tests assert archived output contracts; their simulations are not rerun. Stage 2F writes numerical summaries, CSV tables, diagnostic figures and machine-specific benchmark records. Its document-writing code and template dependency have been removed. Stages 0–2E scientific algorithms and thresholds are unchanged. Start with README_OPERA_GC.md, followed by OPERA_GC_IMPLEMENTED_METHOD.md, OPERA_GC_VARIABLES_AND_COUNTS.md and OPERA_GC_REPRODUCIBILITY.md. The archive includes source, result tables, figure inputs, SHA256 manifests, test reports, licences and file-level provenance. The variants are correlated sensitivity cases. The profile discrepancy is not a calibrated goodness-of-fit probability; same-star ratio errors omit covariance between the fitted dispersions. Profile/count agreement does not recover the original 617 source IDs. Original code uses MIT; original derived tables, figures and software documentation use CC BY 4.0. Third-party products retain their own terms and attribution.