Materials Project-Derived Dataset of 19,154 Inorganic Compounds for Thermochemical Energy Storage Materials Screening and Machine Learning
Description
This dataset presents a comprehensive collection of 19,154 inorganic material records containing 66 physicochemical, thermodynamic, crystallographic, electronic, and compositional descriptors relevant to the computational investigation of thermochemical energy storage (TCES) materials. The dataset was compiled from the Materials Project database and organized to facilitate systematic materials screening, comparative analysis, and machine learning-assisted identification of potentially suitable energy storage materials. The dataset encompasses eight material families: metal oxides (15,615 records), anhydrous salts (992), carbonates (883), salt hydrates (606), metal hydrides (450), hydroxides (431), ammoniates and amides (102), and gas-reference compounds (75). These categories represent a broad range of inorganic chemical systems relevant to the investigation of reversible thermochemical reactions and associated material properties. Each record contains material identifiers, chemical formulas, elemental compositions, crystal structures, space groups, formation energies, formation enthalpies, energies above the convex hull, stability indicators, densities, molar masses, molar volumes, and electronic properties. Additional descriptors include magnetic characteristics, available mechanical properties, elemental-presence indicators, and information distinguishing experimentally reported structures from theoretical material entries. The dataset is designed to support data-driven materials discovery, exploratory statistical analysis, structure–property relationship investigation, material classification, and predictive machine learning applications. It enables researchers to examine trends in thermodynamic stability and other relevant properties across multiple inorganic material families. Importantly, the dataset contains computational materials descriptors rather than experimentally validated thermochemical storage performance measurements. Material stability indicators should not be interpreted as confirmation of reversible reaction performance, storage capacity, reaction kinetics, or long-term cycling stability. Additional reaction-level thermodynamic analysis and experimental validation are necessary to establish practical TCES suitability. The dataset is provided in comma-separated values (CSV) format to facilitate accessibility, computational reproducibility, and integration with Python, MATLAB, and other data analysis and machine learning environments. It may serve as a reference resource for researchers investigating thermochemical energy storage, computational materials science, and artificial intelligence-assisted energy materials discovery.
Files
Steps to reproduce
The dataset was generated using a Python-based workflow with computational materials data obtained from the Materials Project database (https://next-gen.materialsproject.org/) through its official Application Programming Interface (API): https://next-gen.materialsproject.org/api. Data were retrieved using the `mp-api` Python package with an authenticated API key, based on Materials Project database version 2026.04.13. Material records were extracted through family-specific queries to the Materials Project summary endpoint. Eight inorganic material families relevant to thermochemical energy storage (TCES) were considered: metal oxides, hydroxides, carbonates, salt hydrates, anhydrous salts, metal hydrides, ammoniates/amides, and gas-reference compounds. Retrieved materials were classified using predefined elemental-composition criteria, including required elements, permitted cations, additional elements, and composition limits. Python-based filtering was applied to ensure compliance with the selection rules. Material properties, including chemical formulas, crystal structures, formation energies, formation enthalpies, energy above the convex hull, densities, molar masses, electronic properties, and stability indicators, were extracted and organized into a structured dataset. Data preprocessing included property standardization, chemical classification, formatting corrections, and preservation of Materials Project identifiers. The final dataset contains 19,154 material records and 66 descriptors, including 19,079 solid-state entries and 75 gas-reference entries. The processed data were exported in CSV format for computational analysis and machine learning applications. To reproduce the dataset, researchers should obtain a Materials Project API key, install the required Python packages (`mp-api`, `pymatgen`, `pandas`, and `numpy`), and execute the supplied extraction workflow using identical selection criteria. Differences may arise when newer Materials Project database releases are used.