Elucidating the Rejection Mechanisms of PPCPs by Nanofiltration and Reverse Osmosis Membranes

In this study, the rejection mechanisms of six commonly detected pharmaceutical and personal care products (PPCPs) were systematically studied with three commercial thin film composite polyamide reverse osmosis (RO, XLE) and nanofiltration (NF, NF90, and NF270) membranes at pH 3-10. The amount of PPCP adsorption on membrane surfaces was also extracted and calculated so as to determine the contribution of adsorption mechanism on PPCP rejection using adsorption kinetics and adsorption isotherms. At low pHs, PPCP rejection was the highest for XLE followed by NF90 and NF270. As pH increased to 10, PPCP rejection increased significantly for NF90 and NF270, attributed to the enhanced electrostatic repulsion between the negatively charged membrane surface and ionized PPCPs, while being slightly increased for XLE due to the dominant mechanism of steric hindrance. The simplified charge concentration polarization model predicted well for most cases, which demonstrated the contribution of steric hindrance and electrostatic repulsion mechanisms in PPCP rejection by NF and RO membranes. However, two groups of bias were observed during model prediction. One is the underestimated group of small ionized PPCPs at high pH values (8 and 10) by NF270 because of the electrostatic repulsion between the small ionized PPCPs and NF270 overwhelmed the increase in permeate flux by membrane swelling. The other group is the overestimated rejection of triclosan (TRI), a highly hydrophobic compound adsorbing onto membrane surfaces leading to its diffusion and penetration through membranes, which can be confirmed by the extraction of TRI mainly from the top polyamide plus polysulfone membrane layers and also the bottom polyester layer after filtration experiments even at high pH values. The competitive adsorption of TRI and ibuprofen was observed in static adsorption kinetic experiments, and the adsorption can be explained well by the first-order reaction model. The adsorption isotherm data fitted best with the Freundlich model in all cases with the n value indicating chemical adsorption, which would be a hydrophobic interaction between PPCPs and the membrane surface in this study.