Designing desalination MXene membranes by machine learning and global optimization algorithm

The development of energy-efficient and low-cost desalination techniques is crucial. Among these techniques, reverse osmosis (RO) is considered one of the most promising solutions for addressing the global water crisis. It has been widely implemented for both large-scale and distributed water desalination. One such promising desalination membrane is MXene with nanopores, thanks to its unique properties. However, accurately predicting the performance of the desalination process or designing new materials is challenging due to the complexity of the process and the various tunable properties of the membranes and nanopores themselves. The combination of machine learning (ML) and global optimization algorithms offers a superior approach to material design from multiple perspectives. Our study demonstrates that Particle Swarm Optimization (PSO) exhibits faster convergence speed and stability compared to Genetic Algorithms. Ti3C2O2 with a specific charge on the nanopore mouth is an excellent candidate, as determined by the particle swarm optimization algorithm. By analyzing water density and ion density along the nanopore, we gain a deep understanding of how pore charge and functionalized groups affect salt rejection and water permeation. The integration of ML and global optimization algorithms can facilitate the design of materials with outstanding desalination performance.