摘要：Creation of freeways in a membrane accelerates water transport for efficient desalination;however,the synthesis of target membrane materials is very *** the present study,the molecular-design strategy is developed for the synthesis of porous aromatic framework(PAF)membranes with rigid and open networks to regulate water ***-based monomers in triangle shape are designed for the construction of PAF-98 and its sister frameworks,and these PAF membranes are formed through trifluoromethanesulfonic acid *** rigidity and geometry of the monomers endow PAFs with unique structures of connected pores with edging size(4.5–6.7Å)between water(2.8Å)and hydrated sodium molecules(7.2Å),enabling effective differentiation of water and *** fabricated membranes with high flexibility and continuity are ideal for water desalination,and reverse osmosis experiments demonstrate that the three PAF membranes are selective for permeation of water molecules over hydrated sodium *** water flux of 547.1 L m^(-2) h^(-1) and high rejection degree of 90%are obtained on a representative membrane of PAF-98-E;and the desalination performance is superior to currently available membrane ***,PAF-98-E membrane is capable of rejecting various cations with high efficiency in real seawater,holding great promise in practical water desalination application.

摘要：发现和优化纳米复合材料可以通过融合从头算方法和机器学习(ML)来实现,目前的瓶颈在于缺乏原子尺度上描述单体间相互作用的ML模型.我们开发了名为双区域网络(DRN)的深度学习算法,通过训练学习原子之间的多尺度交互作用,预测了由337种不同分子组成的分子对在随机构型中的电子耦合强度,平均绝对误差低至2.8 meV.基于两种截止半径的层次化材料表象被证明对DRN模型的高迁移性和鲁棒性至关重要,这种材料表征方式不仅捕获了共轭片段的局部特征,而且在模型训练之前编码了重要的分子片段间相互作用.本研究建立的ML模型为描述单体间相互作用提供了普适的建模框架,为复杂纳米复合材料的逆向设计打下了基础.