This paper presents a data-driven artificial intelligence (AI) framework for designing efficient graphene nanopores for water desalination. The framework combines deep reinforcement learning (DRL) and convolutional neural networks (CNN) to rapidly create and screen thousands of nanopores, identifying those with high water flux and ion rejection rates. Molecular dynamics (MD) simulations validate the AI-designed pores' superior performance compared to conventional circular nanopores, attributing this to their irregular shape with rough edges.

Yuyang Wang, Zhonglin Cao, Amir Barati Farimani