A smart framework to design membranes for organic micropollutants removal

Developing polymeric membranes that effectively remove organic micropollutants (OMPs) is important for water management. However, the structural diversity and physiochemical variability of OMPs make it challenging to develop such membranes. Here we present a data-mechanism-integrated approach to assist membrane design. This approach integrates molecular fingerprint and physical models within the machine learning framework to quantify how functional groups in OMPs affect removal by polymeric membranes and to elucidate the removal mechanisms. We uncovered an anomalous multigroup coupling effect in membrane-based OMP removal and showed that the efficiency of removal depends on the influence of the functional group coupling in the molecular structure. This finding challenges the conventional approach in membrane screening and design that focuses on the properties of isolated functional groups. By combining this knowledge with assessments of OMP types and membrane properties, we reveal a comprehensive interaction framework for tailoring OMP-removal membranes. Overall, the data-mechanism co-driven paradigm has the potential to facilitate the development of advanced water-treatment membranes, eventually contributing to sustainable water management and the preservation of a safe water environment.