▎ 摘　　要

Freshwater shortage has become a looming global crisis with consistent population growth, industrialization, and climate variability, and desalination is regarded as a promising solution. Two-dimensional (2D) crystalline materials are valued for their excellent molecular permeability and selectivity. However, most of the reported desalination membranes suffer from high contamination tendency and a limitation of permeability-selectivity trade-off. In this work, we screened a 2D graphene-based nanomaterial (PO-C32) consisting of dodecagonal, octagonal, hexagonal, and pentagonal carbon rings by density functional theory (DFT) and machine learning (ML). The results show that the periodic pores of 0.55 nm ensure the material achieves nearly 100 % desalination efficiency. The most optimal Na adsorption is located at the hollow site of the dodecagonal ring, and the strong adsorption ability provides the membrane with satisfactory salt ions rejection. The suitable pore density and homogeneity enhance anti-pressure stability, mechanical stability, and service life. In addition, inherent interlinked electron structure produces a self-cleaning effect that contributes to recycling. These features endow PO-C32 with great potential as desalination membrane, providing a promising candidate idea for efficient and low-energy desalination technology.