Metal-Core-Specific Screening with Machine Learning: Accelerating the Discovery of Metal Oxide Clusters for Enhanced...

Obtaining effective extreme ultraviolet lithography (EUVL) materials for pragmatic applications remains challenging. The experimental design and conventional theoretical prediction are time-consuming and costly and hardly affordable to accelerate the discovery of commercial EUVL materials. In this work, we employed the machine learning (ML) technique to predict the ionization potential of promising EUVL materials, which is closely related to the photoresists' solubility switch. The developed ML model presents a strong generalization ability and can predict new EUVL materials containing different metals (i.e., Mg, Cu, Ca, Cd, Ni). Furthermore, feature analysis indicates that the number of hydrogen bond donors in a compound plays a vital role in determining the ionization potential of EUVL materials. The work provides not only an effective ML model to predict EUVL materials but also crucial insights into the correlation between the structure and properties. Finally, the developed ML model has been integrated into an online platform (https://zinc-oxo-cluster-predictor.streamlit.app/), allowing users to quickly evaluate their designed materials and develop a comprehensive scheme for discovering promising EUVL compounds based on our platform.