Proceedings of MATSUS Spring 2025 Conference (MATSUSSpring25)
DOI: https://doi.org/10.29363/nanoge.matsusspring.2025.261
Publication date: 16th December 2024
The organic electrode materials (OEM) are emerging as a promising alternative to develop greener and sustainable battery technologies. However, significant improvements are still required in their cycling stability, rate capability and energy density. This can only be achieved through a fundamental understanding of the electrochemistry at molecular level establishing the structure-properties relationships. To contribute to this end, we are developing methodologies based on evolutionary algorithms (EA) and artificial neural networks (ANN) at interplay with density functional theory (DFT) based calculations. The EA has initially been employed to predict the structure and electrochemistry of a set of dicarboxylates. In a second stage, we have developed a wider database of organic materials for energy applications that contains information about molecular geometries and high-level features extracted from DFT calculations. Based on this database, we have developed a machine learning approach to predict the redox potentials by giving only chemical species and molecular structures, completely by-passing the computer-demanding DFT calculations. A number of learning algorithms based on ANN have been investigated along with different molecular representations. Here we have also included a layer to predict redox-stable compounds. This machinery has been employed to screen a large organic materials library leading to the discovery of novel cathode materials [1,2], which is actually one of the bottlenecks on the development of organic batteries. The computational materials design platform developed here has the potential of significantly contribute to accelerate the discovery of organic electrode materials with superior properties.