Data Sources

The Cambridge Structural Database (CSD), maintained by the Cambridge Crystallographic Data Centre (CCDC), is the world's largest curated collection of experimentally determined crystal structures, containing more than 1.3 million accurate 3D structures derived from X ray, neutron, and electron diffraction analyses. It is widely used across research and industry including pharmaceuticals, agrochemicals, and fine chemicals, to support structure prediction, materials design, and data driven discovery. Within PSDI, access to the CSD is available to authenticated users from the UK academic community. Users can search the CSD through the Find a Substance option in the PSDI Cross Data Search service or via the dedicated webCSD interface. CSD software tools are also available through PSDI's Remote Desktop Access. Use the Visit or resource landing page link to search the CSD via PSDI Cross Data Search, and see Linked Resources for links to webCSD and CSD software. Find out more using the links in the Further Information section.

The Chemical Availability Search (ChASe) data source enables users to search and compare commercially available chemicals using the PSDI Cross Data Search service. Enter product or vendor keywords, CAS number, InChI or SMILES to search pricing and supplier information for over 250,000 unique chemicals from UK suppliers. Search results include product name, supplier name, quantity, purity, prices, CAS number, and catalogue number with a link to the supplier's website. Check the supplier's website for the most up to date product information.

Propersea contains calculated predictions for a wide range of molecular and physicochemical properties for small molecules, such as melting and boiling points, density, solubility, polarizability, and more. It employs various algorithms, including RDKit, semi-empirical quantum methods, Bayesian regression trees, and transformer neural networks. Propersea also contains predicted IUPAC names generated using a machine learning model. Propersea can be searched using the PSDI Cross Data Search service using InChIs, SMILES or by drawing a molecule. Results include predicted values, confidence intervals and reliability scores for the prediction.

AI ready datasets give physical sciences researchers data they can trust and reuse. These AI-Ready datasets are hosted as a PSDI Community Data Collection and are designed to support AI and machine learning workflows, ranging from general purpose collections that researchers can shape for their own models, to task-specific datasets that already include annotations and predefined training, validation, and test splits. Every record includes one or more datafiles accompanied by a Croissant format metadata description (https://mlcommons.org/working-groups/data/croissant/), ensuring that structure, provenance, and context are captured in a machine readable way.

AFMDB is a centralised collection of Atomic Force Microscopy (AFM) data covering a range of physical and biological materials, hosted within the PSDI Community Data Collections. Each entry includes the original AFM files along with a processed image generated during analysis. Initial examples demonstrate the output from our open-source tool TopoStats which has been used to process images and standardised the data output. By making these datasets openly available, AFMDB supports good data sharing practices and provides high quality, standardised examples that can be used to train and evaluate deep learning methods for enhanced AFM analysis. Researchers are encouraged to contribute AFM data from their publications and to test TopoStats, helping refine the tool through community feedback.

BenchmarkSet1500 is an open-access multireference excited-state database established to provide the first dedicated high-accuracy benchmark set for organic semiconductor research. The repository comprises 1,500 small organic molecules with consistently computed vertical excited-state properties obtained using state-averaged complete active space self-consistent field (SA-CASSCF) and strongly contracted N-electron valence state second-order perturbation theory (SC-NEVPT2), alongside the full reproducible workflow code used to generate the dataset. The dataset focuses on systems where single-reference approaches (e.g. TD-DFT) are known to fail, including molecules exhibiting strong static correlation and inverted singlet-triplet gaps. BenchmarkSet1500 is designed to support rigorous method benchmarking, systematic assessment of theory-level performance, development of predictive models, and screening for technologically relevant organic semiconductors.

Topology, trajectory, and AiiDA data provenance files from molecular dynamics simulations of biomolecules are stored in a repository specifically designed for their preservation and accessibility.

The Catalysis Data Infrastructure (CDI) Data Objects resource provides access to the UK Catalysis Hub (UKCH) published data objects that support research outputs. Through the CDI Portal, PSDI Cross Data Search and PSDI Index search users can search, explore, and reuse these data resources. In the Catalysis Data Infrastructure (CDI) knowledge graph, data objects are linked with researchers, institutions, publications, and research themes, fostering data discovery, highlighting collaborations, and supporting analysis of data-driven research. CDI Data Objects can be extended to include domain-specific data resources such as chemical entities, materials, or analytical methods.

The Catalysis Data Infrastructure (CDI) Publications resource provides access to the UK Catalysis Hub (UKCH) Publications, including peer-reviewed articles, books, and theses. CDI Publications offer a dedicated entry point into the CDI knowledge graph through the CDI Portal. PSDI Cross Data Search and PSDI Index search can be used to explore and analyse research publications. Within CDI, publications are linked to researchers (authors), institutions, research themes, and associated data objects. The linking enhances publication discovery, displays research relationships, and supports insight into research activity. CDI Publications can be expanded to include domain-specific indexes about chemicals, materials, and analysis methods.

The Collaborative Computational Project for NMR Crystallography (CCP-NC) database of calculated solid-state NMR data from DFT codes in MAGRES format.

This dataset is derived from the National Institute of Standards and Technology (NIST) publication, "Critical Micelle Concentrations of Aqueous Surfactant Systems" (http://doi.org/10.6028/NBS.NSRDS.36). The data was extracted using the AI-driven Data Revival service (http://www.data-revival.com/), then manually enhanced by adding SMILES strings, and finally reprocessed by Data Revival to be fully converted into a machine-readable format for improved accessibility and interoperability.

The Crystallography Open Database (COD) is an open-access collection of crystal structures for organic, inorganic, metal-organic compounds and minerals, excluding biopolymers. It contains over half a million entries from a wide range of published and unpublished sources, including large databases, small collections, and individual contributors. All data are made available under a permissive open-access licence. This dataset was originally developed and is maintained by the Crystallography Open Database project at https://www.crystallography.net/cod/. PSDI provides a mirror of both the COD website and its corresponding OPTIMADE endpoint to offer a UK-based access point, improving availability, resilience and performance.

Data-to-Knowledge Collections provide curated datasets specifically designed for use in machine learning or generated through machine learning processes. An example is the Machine Learning Interatomic Potentials (MLIPs) data collection, which includes MLIP XYZ files used for training, the trained MLIP models, and, where available, additional metadata such as AIIDA provenance information. By making these datasets accessible, researchers without the resources to generate such data themselves can leverage them for machine learning. Additionally, the MLIP models can be directly applied in modeling tasks, enabling broader exploration and advancements in research.

The Reproducible XAFS Analyses Zenodo Community provides the published results of X-ray absorption fine structure (XAFS) analyses reproduced using X-ray absorption spectroscopy (XAS) Galaxy tools developed in collaboration with PSDI, and use cases and data from the UK Catalysis Hub. The results shared in the Reproducible XAFS Analyses Zenodo Community are in the form of FAIR Data Objects in RO-Crate format and provide examples relevant to the catalysis domain.

The Flow Battery Experimental Metadata Repository provides a structured place for contributing and viewing metadata about experimental practices, equipment choices, and procedural decisions made during multi-institution flow battery reproducibility studies. These studies bring together several of the world's leading research groups, national laboratories, and industrial partners to understand how experimental practices influence the repeatability, replicability, and reproducibility of flow battery electrochemical testing. Structured metadata collection allows researchers to compare approaches, identify factors affecting performance variation, and support the development of more consistent testing methodology. In the longer term, the study aims to provide benchmark datasets for use by the wider community and help new entrants to the field by lowering the barrier to consistent and reliable experimental practices. At present, the catalogue is populated with metadata gathered through the multi-institution "Phase 2" studies conducted across 2025/26, with contributions from over approximately 40 participants worldwide. The resource is openly accessible for browsing, allowing the broader community to explore experimental set-ups and methodological variability. In future phases, the catalogue will be open to submissions from the wider community and anticipate running further benchmarking studies to expand the collection. The repository is supported and hosted by PSDI partners on behalf of the flow battery research community.

This collection of datasets focuses on solubility-related information, including boiling points, Henry's law constants, LogS, melting points, and mole fractions. The Physical Chemistry Properties Data Collection provides carefully curated and standardized data on key chemical properties, empowering researchers to effectively model, analyze, and drive innovation in the field of physical chemistry. This dataset collection is available for download from a Github repository as zipped csv files.

These datasets focus on solubility-related information, including boiling points, Henry's law constants, LogS, melting points, and mole fractions. The Physical Chemistry Properties Data Collection provides carefully curated and standardized data on key chemical properties, empowering researchers to effectively model, analyze, and drive innovation in the field of physical chemistry.

Project M (https://www.diamond.ac.uk/ProjectM/) was a citizen science initiative run with UK secondary schools that generated an open collection of more than 650 calcium carbonate diffraction datasets, crystallised under a wide range of additives and conditions, with all associated fits and parameters included. These records are hosted as a PSDI Community Data Collection and contain the raw X ray powder diffraction data from Diamond Light Source, accompanied by the essential experimental metadata.

SimpNMR_DB is a shared collection for ab initio calculations of hyperfine interactions that are used as inputs for SimpNMR. SimpNMR is an open-source python package for analysing paramagnetic NMR data that links Density Functional Theory (DFT) calculations with NMR spectra and helps to assign the peaks and extract the magnetic susceptibility tensor form paramagnetic shifts. This ab initio data is useful for the community involved in computing magnetic properties of molecules due to the difficulty in accurate prediction. It provides access to computational results that are often time consuming to generate, giving researchers a consistent reference point for comparison, method testing, and reproducibility. As the database grows, it will also support benchmarking studies and the development of machine learning models. Therefore we encourage researchers to share their computational files in SimpNMR_DB to benefit the wider community. Note that this repository is currently under rapid and active development.

TEM-ParticlesDB is an open repository of Transmission Electron Microscopy (TEM) nanoparticle images and associated segmentation data that have been automatically segmented using machine learning to identify particle size, shape and morphology across a range of substrates. Each entry includes the original experimental TEM images, the synthetic training data used to develop the segmentation models, and the resulting segmented particle outlines. The synthetic data is generated using the TEMPOS (Transmission Electron Microscopy Particle Outline Segmentation) software and used to train segmentation models via transfer learning, before being applied to the experimental images.

2DMatpedia, an open computational database of two-dimensional materials from top-down and bottom-up approaches

Description of provider (Alexandria): A collection of databases from the group of Prof Miguel A. L. Marques at Ruhr University Bochum. Description of dataset (alexandria): A dataset of 2.5m+ stable and metastable materials calculated with the PBE functional

Description of provider (Alexandria): A collection of databases from the group of Prof Miguel A. L. Marques at Ruhr University Bochum. Description of dataset (alexandria): A new dataset of 415k stable and metastable materials calculated with the PBEsol and SCAN functionals

The Chemotion Repository is an open repository of well-structured and machine-readable chemical data created by the Karlsruhe Institute of Technology. The focus of the repository is synthetic and analytical chemistry data. It includes data pertaining to reactions, molecules and physical samples. Data in the repository is uploaded by the community, and is subject to peer review.

Material-Property-Descriptor Database (MPDD) of atomic structures. Optimized for the high-throughput deployment of material featurizers and ML models. Maintained by Phases Research Lab (phaseslab.org) at The Pennsylvania State University