纳滤（NF）和反渗透（RO）技术能够有效去除水中的药物等微量有机物，在饮用水深度处理领域具有广阔的应用前景。NF/RO膜对药物的截留主要依靠尺寸筛分效应。但由溶质与膜之间多种相互作用导致的吸附现象会增加药物在膜表面的分配，从而对药物本身的截留具有不利影响。本文研究了四种商品化NF/RO膜对多种药物的截留特性，定量化评估了吸附对截留的影响程度，深入解析了各种溶质-膜相互作用及吸附机理，并在具有“选择性分离药物和无机盐”性能的新型NF膜制备方面进行了探索。首先考察了两种致密膜（ESPA1和NF90）和两种疏松膜（NF270和HL）对药物的去除效果，将中性pH和膜等电点条件下的实测截留率与仅基于尺寸筛分效应的模型预测截留率进行比较。结果表明，在中性pH条件下，吸附导致致密膜和疏松膜对药物的截留率偏低程度分别为<6%和7–36%。与非静电作用（疏水作用、氢键等）相比，静电吸引对正电性药物截留率的不利影响更大。为了解析溶质-膜相互作用，分别采用动态过滤实验和静态吸附实验测定了药物在疏松膜和致密膜聚酰胺（PA）活性层上的吸附量。在静态吸附实验中，膜的聚砜（PSf）支撑层对吸附容量和达到吸附平衡所需时间均有显著影响，验证了测定吸附量时将膜PA层与PSf层分离的必要性。不同药物的吸附量差异主要由静电吸引/排斥和疏水作用导致，其中静电吸引对正电性药物吸附量的贡献比例整体高于65%。通过分子对接这种计算化学方法模拟膜PA层与药物的结合模式并计算结合能，进一步识别了吸附过程中涉及的多种特异性相互作用（氢键、π-π堆积、π-阳离子相互作用和离子桥结合）。使用结合能和药物分子描述符（溶解度和摩尔折射率）分别表征溶质与膜亲和力和溶剂效应，能够较好地解释药物的实测吸附量。最后，为了减轻吸附对药物截留的不利影响并降低膜对无机盐（NaCl）的截留率，探索制备了膜孔尺寸合适、表面负电荷密度较低且亲水性较高的NF膜。通过适当提高界面聚合反应中哌嗪单体的浓度，以及在初生膜表面接枝亲水性单体二乙醇胺或一乙醇胺，所制得的膜对药物整体截留率较高（~91%）同时对NaCl的截留率较低（<33%），在实际应用中具有一定优势。本文对药物与膜之间相互作用的系统性研究，能够为水处理过程中膜的合理选用及制备提供参考，对有机物传质模型的修正也具有重要意义。

Nanofiltration （NF） and reverse osmosis （RO） are considered as promising technologies for the removal of trace organic compounds including pharmaceutically active compounds （PhACs） during advanced treatment of drinking water. Rejection of PhACs by NF/RO primarily depends on the size exclusion mechanism, while the adsorption of PhACs on membranes resulting from solute–membrane interactions was found to adversely affect the steady-state rejection by promoting the solute partitioning. In this work, based on the rejection performance of several PhACs by four commercial NF/RO membranes, the impact of adsorption on rejection was quantitatively evaluated, and various solute–membrane interactions and adsorption mechanisms were comprehensively elucidated. The preparation of novel NF membranes was also explored to target selective separation of PhACs and inorganic salts.The removal efficiency of PhACs by two tight membranes （ESPA1 and NF90） and two loose membranes （NF270 and HL） was firstly investigated. The measured rejection at neutral pH and that at the membrane isoelectric point were compared with the predicted rejection based solely on size exclusion. Results showed that at neutral pH, the relative decrease of rejection due to adsorption was ＜ 6% for the tight membranes and 7–36% for the loose membranes. Compared to non-electrostatic interactions （hydrophobic interaction, hydrogen bonding, etc.）, electrostatic attraction had a larger adverse effect on the rejection of the positively charged PhACs.To better understand solute–membrane interactions, the adsorbed amounts of PhACs on the polyamide （PA） active layers of the loose and tight membranes were respectively determined through dynamic filtration and static adsorption experiments. During the static adsorption process, both the adsorption capacity for PhACs and the time needed to reach the adsorption equilibrium were obviously influenced by the presence of the polysulfone （PSf） support layer of membranes. This demonstrated the necessity of isolating the PA layer from the PSf layer when determining the adsorbed amounts on the PA layer. Experimental results showed that the difference in adsorbed amounts among various PhACs was closely associated with electrostatic attraction/repulsion and hydrophobic interaction, and the adsorption of the positively charged PhACs was mainly contributed by electrostatic attraction （generally ＞ 65%）. Molecular docking, a computational chemistry approach, was further applied to the simulation of the binding mode between each PhAC and the constructed PA model as well as the calculation of the corresponding binding energy. Based on the binding modes, a variety of specific interactions （including hydrogen bonding, π-π stacking, π-cation interaction and ionic bridge binding） involved in adsorption were identified. The experimental adsorbed amounts of PhACs could be well interpreted by utilizing binding energies and molecular descriptors （solubility and molar refraction） which respectively indicated the solute-membrane affinity and the solvent effect. In order to alleviate the negative influence of adsorption on rejection and reduce the rejection of inorganic salts （e.g. NaCl）, efforts were made to prepare the NF membranes with a proper pore size, a low surface negative charge density and high hydrophilicity. By appropriately increasing the concentration of piperazine monomers for interfacial polymerization and then grafting hydrophilic monomers （diethanolamine or monoethanolamine） on the nascent membrane surface, a high rejection of PhACs （~91%） and a low NaCl rejection （＜ 33%） were simultaneously achieved, which endowed the membranes with certain advantages in practical applications.The systematic investigations on the interactions between PhACs and NF/RO membranes in this study can provide references to the selection and fabrication of membranes during water treatment, and also contribute to the improvement of the transport model of PhACs through membranes.