Materials for energy, the Politecnico discovers new compounds with Ai
The result is a drastic reduction in analysis timeand a broadeningof knowledge
A group of researchers at the Politecnico di Torino has developed an innovative protocol based on artificial intelligence capable of identifying, among hundreds of thousands of theoretically stable but still unexplored materials, those most promising for energy applications.
The protocol uses a two-step approach: the first consists of a system of 'artificial experts' who - voting by majority vote - identify the compounds most likely to possess useful properties for energy applications; subsequently, other suitably trained models accurately estimate their key parameters.
The protocol was presented in a study, published in the journal Energy and AI, which introduces Energy-GNoME, the first 'evolutionary' database that integrates machine learning algorithms with data from the GNoME (Graph Networks for Materials Exploration) project, implemented by Google DeepMind.
The GNoME project made an unprecedented archive available to the scientific community: thousands of never-before-studied materials, identified by means of generative artificial intelligence techniques. However, these materials had not yet been characterised, i.e. their possible technological applications were unknown. Energy-GNoME was created to fill this gap, selecting and classifying the most suitable structures for energy storage, conversion and transport.
The protocol is the result of the work of Paolo De Angelis, Giulio Barletta, Giovanni Trezza, Pietro Asinari and Eliodoro Chiavazzo, of the SMaLL (Smart Materials and Living Lab) laboratory at the Politecnico's Energy Department. The result is a drastic reduction in analysis time and an expansion of the knowledge base useful for the energy sector. "With Energy-GNoME,' explains Paolo De Angelis, first author of the study, 'we wanted to show how artificial intelligence can be not only an analysis tool, but a true accelerator of scientific discovery. The aim is to move from a random generation of materials to research oriented towards engineering functionality, because a crystal is just a chemical compound: it is its function that transforms it into a useful material'.
