Powder XRD patterns and sample fits for Automated Interpretation Framework (AIF) vs best Rwp

This archive contains powder X-ray diffraction (PXRD) refinement fits supporting the paper "Automating Chemical Reasoning in High-Throughput Phase Identification with a Probabilistic, LLM-Guided Framework" (O. Dartsi, L. Walters, A. E. Trewartha, S. B. Torrisi, A. Volk, L. Joyce, G. Ceder, A. Jain, "Automating Chemical Reasoning in High-Throughput Phase Identification with a Probabilistic, LLM-Guided Framework," Advanced Science(2026), in press. DOI: [pending].) The paper introduces the Automated Interpretation Framework (AIF), which ranks competing phase interpretations of PXRD patterns using a Bayesian combination of refinement metrics (Rwp, peak matching) and chemistry-aware priors (composition balance, LLM-derived plausibility). The dataset corresponds to the 80-sample Precursor Genome re-evaluation subset described in the paper. Samples were synthesized by the A-Lab autonomous laboratory at Lawrence Berkeley National Laboratory via pairwise solid-state reactions of common inorganic precursors, and refinements were generated with the DARA automated refinement engine. The archive is a single zip file containing one folder per sample (e.g., PG_25). Each folder contains four files comparing the two interpretations evaluated in the paper: files named {sample}aif* show the refinement fit for the AIF top-ranked interpretation, and files named {sample}rwp* show the fit for the lowest-Rwp interpretation. Each fit is provided both as an interactive HTML plot and as a static PNG render. Plots display the observed pattern, calculated fit, background, and difference curve, along with the phases in the interpretation (with reference database identifiers), the refined Rwp, and annotations of missing and extra peaks relative to the fit. The internal plot title encodes the precursor pair, synthesis temperature, and dwell time for each sample. The AIF software used to generate rankings and trustworthiness scores is available at https://github.com/hackingmaterials/AIF (release v1.0). This work was supported by the Toyota Research Institute through the Accelerated Materials Design and Discovery program.