UiO-66 Benchmark Dataset

Description This record provides the complete benchmark dataset generated for the thesis “Towards a Benchmark for Machine Learning in MOF Synthesis: Intelligent Exploration of Parameter Space”. It contains raw and processed experimental data, associated analysis code, and X-ray diffraction (XRD) diffractograms for preliminary experiments, the main dataset, and repeat syntheses. The dataset focuses on the zirconium-based metal–organic framework UiO-66, synthesised in a high-throughput microwave setup guided by the Edison 4.0 platform. The aim was to create a machine-learning benchmark dataset for machine-learning applications, enabling systematic exploration of synthesis–structure relationships, phase identity, phase purity, and crystallinity. Contents 1. Dataset/ Contains CSV files with synthesis parameters, outcomes, and evaluation results. Dataset.csv — 290 unique UiO-66 syntheses, exported directly from Edison 4.0. Columns: synthesis parameters (metal, linker, modulator concentration; reaction time; temperature), outcomes (yield = Amount, Crystallinity, Phase Identity), subscores, go/no-go flags, and aggregated total score. Dataset_combined.csv — the same 290 syntheses with updated column naming: Phase Identity → Phase Purity, isgo_Phase Identity → isgo_Phase Purity, new column isgo_Phase Identity added, plus a Sample ID field. repeats.csv — 30 repeat syntheses for reproducibility assessment, reported with original Phase Identity assignments. repeats_reviewed.csv — the same 30 repeats with re-evaluated assignments (old Phase Identity preserved as Phase Purity, re-evaluated assignment stored under Phase Identity). 2. Python Code/ Jupyter notebooks implementing the data processing, evaluation, and visualisation workflows described in the thesis. Basic Evaluation.ipynb — Pearson correlation analysis and Random Forest classification metrics. Regression_Model.ipynb — Random Forest regression model implementation and evaluation. Feature variation Phase Purity.ipynb — Feature importance and variation analysis. Feature Removal Phase Purity.ipynb — Feature removal and re-evaluation analysis. Scatterplot Phase Purity.ipynb — Scatterplots relating synthesis parameters to outcomes (phase purity). Scatterplot Phase Identity.ipynb — Scatterplots and model evaluation based on phase identity. 2D Slices.ipynb — 2D parameter space slices illustrating model trends. The notebooks include code for cleaning and structuring raw synthesis data, assigning phase identity and crystallinity, and generating figures used in the thesis. 3. XRD/ X-ray diffraction data underlying phase assignment, crystallinity evaluation, and reproducibility analysis. Provided in multiple formats: Original.Diffrac.Suite.Raw/ — vendor raw files (Bruker DIFFRAC.SUITE format). Original.xy/ — exported ASCII diffraction patterns (2θ vs intensity). Background.xy/ — background curves used for subtraction. Background Subtracted.xy/ — processed XRDs used in analysis. Gold Peaks.xy/ — reference peak files for UiO-66 identification. Subfolders: Preliminary Experiments/ — diffractograms from early trials establishing UiO-66 synthesis. Main Dataset/ — diffractograms for all 290 benchmark syntheses. Repetitions/ — diffractograms for the 30 repeat syntheses used in reproducibility and noise assessment. Usage The dataset can be used to: Train and benchmark machine learning models for MOF synthesis prediction. Study the influence of synthesis parameters on crystallinity, phase identity, and phase purity. Reproduce or extend the analyses presented in the thesis using the provided Jupyter notebooks. Explore raw and processed XRD diffractograms for UiO-66 and evaluate background subtraction effects. Citation When using this dataset, please cite:Holz, N. (2025). UiO-66 Benchmark Study: Benchmark Dataset (v1.0). Zenodo. https://doi.org/10.5281/zenodo.17211171