Source data and code for: Silica drawdown and the ferrous-silicate switch in marine iron deposits

Source data and code for the PNAS research report "Silica drawdown and the ferrous-silicate switch in marine iron deposits" (N. Galili, Department of Earth and Planetary Sciences, Weizmann Institute of Science). Contents (two zip files that unpack into one Source_Data_and_Code/ tree): Source_Data_and_Code.zip: the frozen mineral compilation (the Dataset S1 workbook), the memberships and values behind every figure and table, the numerical summaries, the attributed published-source measurements used in the chemical and isotope comparisons, PHREEQC inputs and outputs for the finite-parcel equilibrium models (Thermoddem v1.10 database), the Kremikovtsi composition calculations, claim-source provenance, and the Python scripts that reproduce the reported counts, bootstrap intervals, chemical response calculations and all 22 figures. RELEASE_MANIFEST.json lists the SHA-256 hash of every file. Source_Data_and_Code_predictions.zip: the two large held-out prediction tables of the predictive sensitivity analysis (source_data/analysis/predictive_stress/statistics_predictions.csv and refit_inputs/empirical_predictions.csv), separated only because of upload size. Reproduction: Python 3.10 or later with NumPy; the commands are listed in source_data/README.txt. Figure regeneration additionally needs pandas, matplotlib and PyMuPDF. The chemical arithmetic verifier and the saved-score audit use the Python standard library only.