Machine learning and DSPM-DE assisted insights into resolving the flux–rejection trade-off in ZIF-67/ZIF-8 pH-responsive nanofiltration membranes

Membrane processes during pollutant separation typically face an inherent flux–rejection trade-off, as both parameters are often governed by a common set of structural and chemical factors. In this study, a pH-responsive membrane containing ZIF-8 in the top layer and ZIF-67 in the sublayer is introduced to overcome this limitation. As the pH increased from 3 to 5, flux and rejection improved by 36% and 45%, respectively. This membrane maintained its performance under real wastewater conditions (94% doxorubicin removal), with negligible ZIFs release and no evidence of structural degradation under long-term pH cycling tests. To elucidate the mechanism underlying the mitigation of trade-off, an innovative approach based on the integration of the DSPM-DE and ML was employed. First, membrane parameters that are not directly measurable experimentally were estimated using DSPM-DE. These parameters …