Abstract
The discovery of energy materials increasingly couples high-throughput computation with machine learning, but the data and models behind a given study are rarely left in a state that lets others, or the original authors months later, re-run or extend the work. IEMAP, Italian Energy Materials Acceleration Platform, was built within the Italian Mission Innovation programme to support the management, analysis and reuse of heterogeneous experimental and computational data on energy materials, with electrochemical storage as its primary use case. Data are organized as projects, each described by structured metadata kept in MongoDB and linked to raw files held in a Ceph object store by content hash. The platform offers interactive access through a web interface and programmatic access through an open-source Python client (iemap-mi) that lets a researcher script ingestion, query and inference together. A trained graph neural network (geoCGNN) is served as an endpoint returning the formation energy and redox potential of a candidate crystal, so the model runs against the same store that holds its inputs. We report a component-level FAIR (Findable, Accessible, Interoperable, Reusable) assessment of the platform, publish the domain ontology that underpins its interoperability, and outline a roadmap towards native semantic querying and knowledge-graph integration. As a demonstration, we reproduce a study of Ni/Ti-doped P2-NaMnO2 sodium-ion cathodes as a single, openly accessible IEMAP workflow that screens a large compositional space, identifies promising chemistries, and stores all intermediate and final results for reuse. The contribution is the infrastructure that turns a one-off computation-and-ML study into a reusable resource for energy-materials research.
https://www.mdpi.com/2313-0105/12/9/355