Constructing Moisture-Induced Degradation Pathways in Metal-Oxide Resists from Two-Phase Active Learning of Deep...

Tin-based metal-oxide resist (Sn-MOR) is a promising candidate material for next-generation photolithography, yet its shelf and in-process stability are significantly undermined by reactions with ambient moisture. While an in-depth computational study is required to understand the reaction mechanism and provide design rules balancing moisture stability and photospeed, the long time scale of degradation reactions hampers ab initio calculations. In this work, we introduce a two-phase active learning (2P-AL) framework, which couples deep potential molecular dynamics with well-tempered metadynamics (WT-MetaD) to capture the rare-event reaction dynamics of solvated Sn-MOR molecules in aqueous solution. The first phase focuses on exploration to broaden structural diversity under a fixed simulation budget, and the subsequent phase focuses on convergence to systematically improve force field fidelity and ultimately achieve ab initio-level accuracy. WT-MetaD simulations with this model successfully constructed the free energy landscape of moisture-induced degradation of Sn-MOR molecules and uncovered a plausible pathway by which ambient moisture can promote degradation in Sn-MORs, offering molecular-level insight into their stability challenges. In parallel, the proposed 2P-AL framework offers an adaptable and efficient approach to investigate reaction dynamics in solution-phase reactive systems, yielding direct molecular insight into moisture-induced degradation in disordered Sn-MOR materials.