Machine learning teaches membranes to sort by chemical affinity
Ultrafiltration membranes used in pharmaceutical manufacturing and other industrial processes have long relied on separating molecules by size. Now, Cornell researchers have created porous materials that filter molecules by their chemical makeup. Two molecules of identical size and weight but different chemistry, such as antibodies with distinct molecular structure, are difficult to separate using current ultrafiltration (UF) membrane technology. But in a study published in Nature Communications, researchers find that blending chemically distinct block copolymer micelles—tiny self-assembling polymer spheres—could be applied to making membranes capable of filtering molecules by chemical affinity. "This is the first real pathway to creating UF membranes with chemically diverse pore surfaces," said Ulrich Wiesner, the Spencer T. Olin Professor of Materials Science and Engineering, and the study's senior author. "In principle, post-fabrication processes may achieve this, but the cost would be prohibitive for industry to adopt it. This new approach could truly revolutionize ultrafiltration."
Read more.








