Active learning for the development of new metallic nanoparticles: applications in electrocatalysis

• Grant number: 24/16774-0
• Support Opportunities: Scholarships in Brazil - Doctorate (Direct)
• Start date: March 01, 2025
• End date: February 28, 2030
• Field of knowledge: Physical Sciences and Mathematics - Physics - Condensed Matter Physics
• Principal Investigator: James Moraes de Almeida
• Grantee: Duanny Silva Onório
• Host Institution: Centro de Ciências Naturais e Humanas (CCNH). Universidade Federal do ABC (UFABC). Ministério da Educação (Brasil). Santo André , SP, Brazil
• Associated research grant: 23/09820-2 - Materials by design: from quantum materials to energy applications, AP.TEM

Abstract

The current energy scenario in the world is in a process of energy transition, originated by the finiteness of fossil fuels and their environmental impacts, so renewable energies have been gaining ground, but they face problems of seasonality and intermittency. Therefore, it is essential to generate clean and efficient forms of energy storage. This project aims to find new catalysts for the oxygen reduction (ORR) and hydrogen evolution (HER) processes, fundamental processes for energy generation in fuel cells. Currently, good catalysts are compositions of noble metals such as platinum or iridium, which makes scalability unfeasible due to high costs. Therefore, studies have been carried out with more abundant compounds, but laboratory research progress is slow. Thus, the proposed solution for this work is the use of active machine learning to track a large number of compositions, so that the model is iteratively fed with Density Functional Theory simulations, strategically suggested for acquisition functions, such as the Bayesian function, thus improving the results with each iteration, until the model is accurate enough to find promising discoveries. This would speed up the process of locating good candidates, in addition to finding suggestions that human intuition could discard. The active learning process has already shown very relevant results in different areas, but it has not yet been used in the screening of discoveries for ORR and HER, as proposed by the project. In the methodological part, we will mainly use the Gibbs free energy (¿G) of adsorption as a guide to evaluate the events, as it is already quite successful in the literature. Therefore, the project seeks ways to improve the catalytic process, which may result in an advance in the generation of clean energy through the discovery of new efficient studies, using computational simulation methods and active learning models to minimize or eliminate the use of rare and high-cost elements, promoting viable and scalable solutions for fuel cells and catalytic reactions.