Preparation of a Molecularly Imprinted Silica Nanoparticles Embedded Microfiltration Membrane for Selective Separation of Tetrabromobisphenol A from Water

The ubiquitous presence of tetrabromobisphenol A (TBBPA) in aquatic environments has caused severe environmental and public health concerns; it is therefore of great significance to develop effective techniques to remove this compound from contaminated waters. Herein, a TBBPA imprinted membrane was successfully fabricated via incorporating imprinted silica nanoparticles (SiO2 NPs). The TBBPA imprinted layer was synthesized on the 3-(methacryloyloxy) propyltrimethoxysilane (KH-570) modified SiO2 NPs via surface imprinting. Eluted TBBPA molecularly imprinted nanoparticles (E-TBBPA-MINs) were incorporated onto a polyvinylidene difluoride (PVDF) microfiltration membrane via vacuum-assisted filtration. The obtained E-TBBPA-MINs embedded membrane (E-TBBPA-MIM) showed appreciable permeation selectivity toward the structurally analogous to TBBPA (i.e., 6.74, 5.24 and 6.31 of the permselectivity factors for p-tert-butylphenol (BP), bisphenol A (BPA) and 4,4′-dihydroxybiphenyl (DDBP), respectively), far superior to the non-imprinted membrane (i.e., 1.47, 1.17 and 1.56 for BP, BPA and DDBP, respectively). The permselectivity mechanism of E-TBBPA-MIM could be attributed to the specific chemical adsorption and spatial complementation of TBBPA molecules by the imprinted cavities. The resulting E-TBBPA-MIM exhibited good stability after five adsorption/desorption cycles. The findings of this study validated the feasibility of developing nanoparticles embedded molecularly imprinted membrane for efficient separation and removal of TBBPA from water.


Introduction
Endocrine-disrupting compounds (EDCs), which are ubiquitously occurring in contaminated waters, have been drawing increasing attention for their potential risks to human health, such as disrupting hormonal balance and inducing reproductive abnormality [1][2][3]. Tetrabromobisphenol A (TBBPA) (i.e., a kind of EDC)-which has been frequently detected in natural waters, both from domestic and industrial discharges-exhibits greater toxic (DDBP, 97%) were purchased from Sigma Aldrich (St. Louis, MO, USA). The solvents used in this study, including acetonitrile, ethanol, toluene, methanol, ammonia, and acetic acid, were obtained from China Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). In particular, TEOS was used as the raw material for preparing SiO2 NPs, and the KH-570 as the silane coupling agent for SiO2 NPs modification. TBBPA, EGDMA, 4-VP, and AIBN were adopted as the template molecule, crosslinking agent, functional monomer, and initiator for fabricating the imprinted SiO2 NPs, respectively. BPA, BP, and DDBP were employed as the competitive compounds of TBBPA in the selective adsorption experiments.

Synthesis of E-TBBPA-MINs and E-TBBPA-MIM
SiO2 NPs were fabricated with the Stӧber method [21]. The experimental setup for the synthesis and modification of SiO2 NPs is shown in Figure S1. The schematic diagram of the preparation route of E-TBBPA-MINs and E-TBBPA-MIM is exhibited in Figure 1. To prepare SiO2 NPs, 10 mL TEOS was added into a mixture of ammonia aqueous solution (28 wt%, 20 mL), deionized water (50 mL), and ethanol (110 mL). The mixture was stirred for 2 h. SiO2 NPs were then obtained after centrifugation, ethanol-washing (three times), and 12 h of drying at 60 °C. KH-570 was grafted on the surface of SiO2 NPs according to the reported procedures [10,[22][23][24]. In brief, the SiO2 NPs (0.15 g) were dispersed in toluene (30 mL), followed by the slow addition of 4 mL KH-570 and stirring. Subsequently, the mixture was flushed with nitrogen for 10 min to remove oxygen and heated at 50 °C for 12 h. The obtained products (termed K-SiO2 NPs) were collected following centrifugation, ethanol-washing (three times), and 12 h of drying at 60 °C. The fabrication procedure of E-TBBPA-MINs was conducted as follows: First, TBBPA (0.1 mmol) and 4-VP (0.1-0.8 mmol) were mixed in toluene (30 mL) and mechanically stirred under 30 °C for 1 h. Then, 0.15 g of K-SiO2 NPs were dispersed in the mixture, followed by adding EGDMA (0.8-2.4 mmol) and AIBN (20 mg). Next, the mixture reacted at 60 °C for 24 h under N2 protection to achieve the TBBPA molecularly imprinted nanoparticles (termed TBBPA-MINs). Subsequently, unreacted monomers absorbed on the imprinted nanoparticles were removed using ethanol-washing thrice. Finally, E-TBBPA-MINs were obtained after removing the template molecule (i.e., TBBPA) from the TBBPA-MINs with an eluent and dried at 60 °C. Note that the eluent was a mixture of acetic acid and methanol (vol./vol. = 1:9), which was commonly used for rinsing the template The fabrication procedure of E-TBBPA-MINs was conducted as follows: First, TBBPA (0.1 mmol) and 4-VP (0.1-0.8 mmol) were mixed in toluene (30 mL) and mechanically stirred under 30 • C for 1 h. Then, 0.15 g of K-SiO 2 NPs were dispersed in the mixture, followed by adding EGDMA (0.8-2.4 mmol) and AIBN (20 mg). Next, the mixture reacted at 60 • C for 24 h under N 2 protection to achieve the TBBPA molecularly imprinted nanoparticles (termed TBBPA-MINs). Subsequently, unreacted monomers absorbed on the imprinted nanoparticles were removed using ethanol-washing thrice. Finally, E-TBBPA-MINs were obtained after removing the template molecule (i.e., TBBPA) from the TBBPA-MINs with an eluent and dried at 60 • C. Note that the eluent was a mixture of acetic acid and methanol (vol./vol. = 1:9), which was commonly used for rinsing the template molecules from MIPs [25,26]. The elution process was completed until no TBBPA was detected in the eluent.
Prior to the preparation of E-TBBPA-MIM, the optimal dosage of 4-VP and EGDMA was assured by comparing the measured Q e values (i.e., equilibrium adsorption capacity in mg/g. The calculation procedure can be found in Section 2.4.1) after suspending 10 mg E-TBBPA-MINs in 50 mg/L TBBPA solution for 2 h at 25 • C. Thereafter, the E-TBBPA-MIM was fabricated using the E-TBBPA-MINs with the optimal TBBPA/4-VP/EGDMA molar ratio. In brief, 5 mg of E-TBBPA-MINs was dispersed in ethanol (50 mL), and then vacuumfiltrated onto a PVDF microfiltration membrane (pore size = 0.45 µm), followed by drying at 40 • C overnight to obtain the E-TBBPA-MIM. Note that the non-imprinted nanoparticles and membrane (without adding the TBBPA, termed NINs and NIM, respectively) were manufactured according to the same procedure.

Characterization of E-TBBPA-MINs and E-TBBPA-MIM
Morphologies and/or elemental compositions of E-TBBPA-MINs and E-TBBPA-MIM were determined by scanning electron microscopy equipped with an energy-dispersive spectrometer (SEM-EDS, Tescan Mira4, Brno, Czech Republic), transmission electron microscopy (TEM, JEOL F200, Tokyo, Japan), and atomic force microscopy (AFM, Bruker Dimension Icon, Bremen, Germany). The surface functional groups and chemical states of key elements of SiO 2 NPs, E-TBBPA-MINs, and NINs were measured with a Fourier transform infrared spectrometer (FTIR, Thermo Scientific Nicolet 670, Waltham, MA, USA) and X-ray photoelectron spectroscopy (XPS, Thermo Scientific Nexsa, Waltham, MA, USA). The PerkinElmer system was utilized to perform thermo-gravimetric analysis (TGA) of SiO 2 NPs and E-TBBPA-MINs. A Brunauer-Emmett-Teller (BET, Micromeritics ASAP 2460, Norcross, GA, USA) analysis was conducted to compare the characteristics of N 2 adsorption/desorption isotherm, specific surface area, average pore diameter, and pore volume of E-TBBPA-MINs, and NINs. Membrane-wetting behaviors of E-TBBPA-MIM and pristine PVDF membrane were evaluated using a contact angle meter (SDC 100S, Beijing, China). Measurements of the porosity and water permeability of these two membranes were conducted according to the procedures described previously [27,28].

Adsorption Isotherms of E-TBBPA-MIM
E-TBBPA-MIM and the control membrane (NIM) were placed in a series of 30 mL TBBPA solutions with the initial concentration of TBBPA ranging from 10 to 100 mg/L (i.e., 10, 20, 40, 60, 80, and 100 mg/L). The series of mixtures were shaken for 2 h at 25 • C, and then the equilibrium concentration of TBBPA was measured by high-performance liquid chromatography (HPLC). All adsorption experiments were conducted at 25 • C, which was among the commonly used temperature range (20-35 • C) to investigate the adsorption behaviors of MIPs for organic contaminants [29][30][31]. The binding amount at equilibrium Q e was calculated by the following equation: where C 0 (mg/L) and C e (mg/L) represent the initial and equilibrium concentration of TBBPA, respectively. W (mg) and V (mL) signify the mass of membranes and the volume of the TBBPA solution, respectively.

Kinetic Absorption of E-TBBPA-MIM
E-TBBPA-MIM and NIM were placed in the TBBPA solution (60 mg/L) and adsorbed within a series of different times (i.e., 10, 20, 40, 60, 80, 100, 120, 150 and 180 min). After adsorption, the supernatant was analyzed by HPLC, and the absorption capacity Q t (mg/g) was calculated at each contact time (t) as: where C t (mg/L) represents the concentration of TBBPA at time t.

Selective Rebinding Experiments
The selectivity of E-TBBPA-MIM was investigated in the presence of bisphenols compounds (i.e., TBBPA, BPA, BP, and DDBP) with the same initial concentration (60 mg/L). In detail, E-TBBPA-MIM and NIM were placed in the solution of TBBPA, BPA, BP, and DDBP, respectively, followed by incubation at 25 • C. The concentration of TBBPA and its structural analogues in supernatant was examined with HPLC. The adsorption capacity (Q, mg/g), static allocation coefficient (K, L/g), and imprinting factor α were computed according to the following equations: where C 0 (mg/L) and C (mg/L) represent the concentration of TBBPA before and after adsorption, respectively.

Stability Tests
To investigate the regeneration and stability of E-TBBPA-MIM, the membranes were placed in TBBPA solution (60 mg/L) for adsorption (2 h, 25 • C). After each adsorption experiment, the E-TBBPA-MIM was subject to the elution process, as described in Section 2.2. The E-TBBPA-MIM was used for the subsequent five cycles (adsorption/desorption) to evaluate the regeneration capability. The turbidity of the eluents was measured by a portable turbidimeter (Hach 2100P, Loveland, CO, USA) to examine whether SiO 2 MINs were released from the E-TBBPA-MIM.

Selective Permeation Experiments
A cross-flow filtration mode was conducted to evaluate the permeation performance of E-TBBPA-MIM and NIM ( Figure S2). The effective filtration area of membranes was 19.6 cm 2 , and the effluent flow rate was controlled at 24 mL/min by a peristaltic pump (LeadFluid YZ 15, Baoding, China). The synthetic feeding solution contained TBBPA, BPA, BP, and DDBP (60 mg/L, 120 mL). At the predetermined time intervals, i.e., 5, 10, 20, 30, 60, 90, and 120 min, the effluent was collected to measure the concentration of TBBPA and its analogues by HPLC. The permeation flux J (mg/cm 2 ·min), permeation coefficient P (L/cm·min), and permselectivity factors β were computed as follows: where ∆C R /∆t (mg/L·min) indicates the variation of TBBPA and its analogues concentration in effluent vs. time. V R (L), A(cm 2 ), and d (cm) are the solution volume of influent, active membrane area, and thickness of the membrane, respectively. (C F − C R ) (mg/L) is the concentration difference of TBBPA and its analogues between influent and effluent.

Optimization of E-TBBPA-MINs Composition
Appropriate dosages of functional monomers and cross-linker are crucial to the imprinting performance of MINs since they regulate the quantity and structure of recognition sites [32,33]. As shown in Figure S3A, the Q e of E-TBBPA-MINs increased with the augment of TBBPA/4-VP molar ratio in the range of 1:1 to 1:4, and further increasing this ratio resulted in the diminished Q e . This can be attributed to the fact that though a larger dosage of functional monomer can give rise to more formation of functional monomer-template complexes (i.e., the precursor of imprinted cavities), the self-association effect of excess functional monomers dooms to happen, thus decreasing the interaction between them and templates [34,35]. An adequate dosage of cross-linker, which determines the rigidity of the imprinted polymer skeleton, is a major prerequisite for producing imprinted cavities with high stability and binding capacity for templates [33]. Figure S3B shows that the optimum molar ratio of 4-VP/EGDMA was 1:5, and a higher molar ratio led to a decreased Q e . A plausible explanation is that TBBPA became difficult to elute from the imprinted polymer with an excessively high cross-linking degree, incurred by the overdose of EGDMA. Hence, the following experiments were carried out using the E-TBBPA-MINs/MIM prepared at the optimal TBBPA/4-VP/EGDMA molar ratio of 1:4:20.

Characterization of E-TBBPA-MINs and E-TBBPA-MIMs
TEM images, as shown in Figure 2A,B, present the morphologies of SiO 2 NPs and E-TBBPA-MINs. It can be observed that SiO 2 NPs have a regular diameter of 300 nm with good dispersibility. After TBBPA was loaded and eluted, a dense shell of 60 nm thickness representing the layer of MIPs was well coated on the surface of the SiO 2 NPs. FTIR analysis of SiO 2 NPs, K-SiO 2 NPs, E-TBBPA-MINs, and NINs is illustrated in Figure 2C. The peaks around 1100 cm −1 , representing the Si-O-Si antisymmetric stretching vibration [25], occurred for SiO 2 NPs, K-SiO 2 NPs, TBBPA-MINs, and E-TBBPA-MINs, suggesting that the imprinting process did not alter the characteristics of SiO 2 NPs. In the spectrum of K-SiO 2 NPs, the characteristic peaks of C=C and C=O were detected at 1634 cm −1 and 1707 cm −1 [36], respectively, indicating the successful grafting of KH-570 on the surface of SiO 2 NPs. The presence of EGDMA can be confirmed by the appearance of a stretching vibration adsorption peak at 1438 cm −1 [33]. The EDS analysis results of the elemental composition of all NPs are shown in Table 1. With the introduction of TBBPA, the amount of nitrogen and bromine increased significantly to 0.6% and 0.2%, respectively, while after the elution of TBBPA, the bromine content decreased to 0%. In line with this, as can be observed from the comparison of XPS C1s spectra of TBBPA-MINs and E-TBBPA-MINs, only the C-Br peak at the binding energy of 287.2 eV disappeared after the elution ( Figure S4A,B). These results corroborated that the elution process was capable of completely removing TBBPA without destroying the polymeric network of MIPs. The influence of the imprinting process on the thermal stability of SiO 2 NPs was investigated by TGA from 20 to 800 • C ( Figure 2D). It can be observed that E-TBBPA-MINs exhibited three distinct weight loss stages. Figure 2D shows that~74% of weight loss occurred in the second stage between 280 and 450 • C, likely associated with the decomposition of MIPs on the surface of SiO 2 NPs [37,38]. In the first and third stages, the weight loss process of E-TBBPA-MINs might be ascribed to the disappearance of residual water and carbonization of polymer chains on the surface of SiO 2 NPs, respectively [39]. Figure 2E shows the N 2 adsorption-desorption isotherms of E-TBBPA-MINs and NINs. It can be inferred that both NPs have similar mesoporous structures, given their physisorption isotherms with the Type IV and H3 hysteresis loop features. Furthermore, it can be seen from Table S1 that the specific surface area, average pore size, and pore volume of E-TBBPA- MINs and NINs were quite similar, implying that compared to NINs, the pore structure might not be an important factor influencing the adsorption ability of E-TBBPA-MINs. The influence of the imprinting process on the thermal stability of SiO2 NPs was investigated by TGA from 20 to 800 °C ( Figure 2D). It can be observed that E-TBBPA-MINs exhibited three distinct weight loss stages. Figure 2D shows that ~74% of weight loss occurred in the second stage between 280 and 450 °C, likely associated with the decomposition of MIPs on the surface of SiO2 NPs [37,38]. In the first and third stages, the weight loss process of E-TBBPA-MINs might be ascribed to the disappearance of residual water and carbonization of polymer chains on the surface of SiO2 NPs, respectively [39]. Figure  2E shows the N2 adsorption-desorption isotherms of E-TBBPA-MINs and NINs. It can be inferred that both NPs have similar mesoporous structures, given their physisorption isotherms with the Type IV and H3 hysteresis loop features. Furthermore, it can be seen from Table S1 that the specific surface area, average pore size, and pore volume of E-TBBPA-MINs and NINs were quite similar, implying that compared to NINs, the pore structure might not be an important factor influencing the adsorption ability of E-TBBPA-MINs. Figure 3A-C presents the surface morphologies of E-TBBPA-MINs embedded membrane and the pristine PVDF membrane (i.e., E-TBBPA-MIM). It can be seen that for the E-TBBPA-MIM, a large quantity of spherical NPs was closely attached onto the pore walls of the membrane matrix, suggesting the successful loading of E-TBBPA-MINs. Furthermore, after the immobilization of E-TBBPA-MINs, compared to the pristine membrane, the average surface roughness (Ra) of E-TBBPA-MIM significantly increased by ~61%, as shown in Figure 3D,E. Table 2 shows the membrane intrinsic properties of E-TBBPA-MIM and the pristine PVDF membrane. The introduction of E-TBBPA-MINs had a remarkably positive impact on the hydrophilicity of E-TBBPA-MIM, given the smaller contact angle of E-TBBPA-MIM than pristine PVDF membrane (14.3 ± 1.6° vs. 97.5 ± 1.7°). This can be ascribed to the coverage of the hydrophilic SiO2 NPS on the membrane surface. However, significant changes in the membrane permeability and porosity were also observed, suggesting that the coverage of NPs on the membrane surface might block up the pores. The equilibrium adsorption capacity (Qe) of E-TBBPA-MIM reached 19.2 ± 0.3 mg/g, significantly greater than that (0.5 ± 0.2 mg/g) of the pristine PVDF membrane.   Figure 3D,E. Table 2 shows the membrane intrinsic properties of E-TBBPA-MIM and the pristine PVDF membrane. The introduction of E-TBBPA-MINs had a remarkably positive impact on the hydrophilicity of E-TBBPA-MIM, given the smaller contact angle of E-TBBPA-MIM than pristine PVDF membrane (14.3 ± 1.6 • vs. 97.5 ± 1.7 • ). This can be ascribed to the coverage of the hydrophilic SiO 2 NP S on the membrane surface. However, significant changes in the membrane permeability and porosity were also observed, suggesting that the coverage of NPs on the membrane surface might block up the pores. The equilibrium adsorption capacity (Q e ) of E-TBBPA-MIM reached 19.2 ± 0.3 mg/g, significantly greater than that (0.5 ± 0.2 mg/g) of the pristine PVDF membrane.

Adsorption Performance of E-TBBPA-MIM
The adsorption capacities of E-TBBPA-MIM and NINs embedded PVDF membrane (NIM) toward TBBPA (10-100 mg/L) were investigated by HPLC. It can be observed from Figure 4A that in the case of all tested TBBPA concentrations, the Qe values of E-TBBPA-MIM were much higher than that of NIM, which could be due to the existence of highaffinity recognition sites on E-TBBPA-MINs for TBBPA molecules [40]. In contrast, the randomly arranged functional groups on the NINs could not specifically bind the template molecules [41], thus the NIM did not possess the imprinting effect. Moreover, the Qe of E-TBBPA-MIM for TBBPA increased with increasing TBBPA concentration from 10 to 80 mg/L, attributed to the enhanced diffusion of TBBPA toward the recognition sites.

Adsorption Performance of E-TBBPA-MIM
The adsorption capacities of E-TBBPA-MIM and NINs embedded PVDF membrane (NIM) toward TBBPA (10-100 mg/L) were investigated by HPLC. It can be observed from Figure 4A that in the case of all tested TBBPA concentrations, the Q e values of E-TBBPA-MIM were much higher than that of NIM, which could be due to the existence of high-affinity recognition sites on E-TBBPA-MINs for TBBPA molecules [40]. In contrast, the randomly arranged functional groups on the NINs could not specifically bind the template molecules [41], thus the NIM did not possess the imprinting effect. Moreover, the Q e of E-TBBPA-MIM for TBBPA increased with increasing TBBPA concentration from 10 to 80 mg/L, attributed to the enhanced diffusion of TBBPA toward the recognition sites.
To further evaluate the adsorption mechanisms, the Langmuir (9) and Freundlich (10) model equations [42,43] were used to assess the equilibrium adsorption data of E-TBBPA-MIM and NIM: Q e = K L Q m C e 1 + K L C e (9) where Q e (mg/g) and Q m (mg/g) indicate the equilibrium and maximum rebinding capacity of TBBPA, respectively. C e (mg/L) is the equilibrium concentration of TBBPA. K L (L/mg), K F [(mg/g)·(L/mg) 1/n ] are the Langmuir and Freundlich binding coefficients, respectively. n is the Freundlich binding constant. As shown in Figure 4A and Table S2, it can be seen that the linear regression values of E-TBBPA-MIM and NIM fitted better with Langmuir model (R 2 = 0.9926 and 0.9882 for E-TBBPA-MIM and NIM, respectively) than those of Freundlich model (R 2 = 0.9594 and 0.9478 for E-TBBPA-MIM and NIM, respectively). This can be ascribed to the homogeneous distribution of E-TBBPA-MINs on the E-TBBPA-MIM, which was consistent with the observation from SEM image in Figure 3B, because Freundlich isotherm is more suitable to study the adsorption behaviors occurring in heterogeneous system [44]. Moreover, the better accuracy of Langmuir model validated that the adsorption of TBBPA on the E-TBBPA-MIM was a monolayer adsorption process. To further evaluate the adsorption mechanisms, the Langmuir (9) and Freundlich (10) model equations [42,43] were used to assess the equilibrium adsorption data of E-TBBPA-MIM and NIM: where Qe (mg/g) and Qm (mg/g) indicate the equilibrium and maximum rebinding capacity of TBBPA, respectively. Ce (mg/L) is the equilibrium concentration of TBBPA. KL (L/mg), KF [(mg/g)·(L/mg) 1/n ] are the Langmuir and Freundlich binding coefficients, respectively. n is the Freundlich binding constant. As shown in Figure 4A and Table S2 Figure 3B, because Freundlich isotherm is more suitable to study the adsorption behaviors occurring in heterogeneous system [44]. Moreover, the better accuracy of Langmuir model validated that the adsorption of TBBPA on the E-TBBPA-MIM was a monolayer adsorption process. The adsorption kinetics analysis of TBBPA on the E-TBBPA-MIM and NIM were conducted to elucidate the rate-controlling and rebinding mechanism of TBBPA molecules. Thus, the plots of Figure 4B were fitted by the pseudo-first-order (Equation (11)) and pseudo-second-order models (Equation (12)): The adsorption kinetics analysis of TBBPA on the E-TBBPA-MIM and NIM were conducted to elucidate the rate-controlling and rebinding mechanism of TBBPA molecules. Thus, the plots of Figure 4B were fitted by the pseudo-first-order (Equation (11)) and pseudo-second-order models (Equation (12)): where Q e and Q t are the amounts of TBBPA adhered in the case of rebinding equilibrium and different time t (mg/g), respectively. K 1 and K 2 represent the equilibrium rate constants of the pseudo-first-order and pseudo-second-order kinetic models (g/mg min), respectively. It can be observed from Figure 4B and Table S3 that the rebinding capacity of E-TBBPA-MIM increased with the increase in time at the beginning. Subsequently, the curve arrived at a plateau over 60 min, mainly due to the fact that the recognition sites of E-TBBPA-MIM were almost saturated after a quick adsorption process (within one hour). In contrast, the NIM exhibited a similar curve but was inferior regarding rebinding capacity. The fitting obtained was better with the pseudo-second-order model (R 2 = 0.9938) rather than pseudo-first-order model (R 2 = 0.9673), implying that the adsorption behavior of TBBPA on E-TBBPA-MIM was controlled by chemical adsorption [45].
To examine the selectivity of E-TBBPA-MIM, the adsorption capabilities of E-TBBPA-MIM and NIM for TBBPA and competitive molecules (BP, BPA and DDBP) were evaluated. It can be seen from Figure 4C that E-TBBPA-MIM showed higher adsorption capacity of TBBPA than those of BP, BPA and DDBP. The imprinting factors (α) of BP, BPA and DDBP were 4.52, 2.88 and 6.07, respectively, suggesting that E-TBBPA-MIM had no selectivity toward non-template molecules. In contrast, there was no significant difference shown between the adsorption capabilities of E-TBBPA-MIM and NIM for competitive molecules. Figure 4D clearly indicates that the rebinding capacity of E-TBBPA-MIM still reached around 90% that of the initial after 5 adsorption/desorption cycles, demonstrating its remarkable adsorption stability. Likewise, previous studies reported that the rebinding capacities of the developed MIMs merely decreased by 7-12% after 4-6 times reuse [37,41]. In addition, during the stability tests, no turbidity could be detected in the eluents, indicating the tight coverage of SiO 2 MINs on the E-TBBPA-MIM.

Permselectivity Performance and Mechanism of E-TBBPA-MIM
Permselectivity behaviors toward the template molecule, i.e., TPPBA, were elucidated to better understand the specific adsorption separation properties of E-TBBPA-MIM. Time-dependent permselectivity behaviors of E-TBBPA-MIM and NIM for TBBPA were investigated using BP, BPA, and DDBP as the competitive molecules ( Figure 5A,B). As shown in Figure 5A,B, at any time interval, E-TBBPA-MIM and NIM resulted in the similar effluent concentrations of BP, BPA, and DDBP, indicating that the imprinted cavities on the E-TBBPA-MIM showed no specific recognition capacity toward the structural analogues of TBBPA, since either the presence or absence of imprinted cavities did not influence the permeation behaviors of these compounds. This is unsurprising because molecules, which cannot be specifically recognized by the MIMs, are known to penetrate the membranes by diffusion/convection [46]. It can be also observed that the E-TBBPA-MIM exhibited significantly greater separation performance for TBBPA than other molecules, but there was no obvious difference between the separation performance of NIM for all the types of molecules. This may be due to the fact that there existed no specific sites on the NIM available for TBBPA adsorption, thus TBBPA penetrated the membranes with other molecules in the same manner [37]. As depicted in Table S4, the permselectivity factors (β) of E-TBBPA-MIM toward BP, BPA, and DDBP were 6.74, 5.24, and 6.31, much greater than those of NIM (i.e., 1.47, 1.17 and 1.56 for BP, BPA, and DDBP, respectively), indicating the successful construction of specific recognition sites, which contributed to the excellent separation performance of E-TBBPA-MIM for TBBPA.
The selective separation behaviors of MIMs are often associated with two opposite mechanisms, i.e., facilitated and retarded permeation mechanisms [46,47]. Given the strong retention capacity of E-TBBPA-MIM for TBBPA, as corroborated by the results of above permselectivity experiments, the retarded permeation mechanism played the dominant role in the separation process of E-TBBPA-MIM for TBBPA. The proposed permselectivity mechanism of E-TBBPA-MIM is shown in Figure 5C. There are two reasons accounting for the selective separation of E-TBBPA-MIM for TBBPA. On one hand, the hydrogen bond interaction between the N atoms in the benzene ring of MIPs and hydroxyl groups of TBBPA resulted in the chemical adsorption of TBBPA by the imprinted cavities. For another, the chemical structure of TBBPA was perfectly matched to the shape, size, and spatial arrangement of imprinted cavities. The rebinding capacity of E-TBBPA-MIM could be recovered after the TBBPA removal by the eluent. The BP, BPA, and DDBP, which are nonmatched to the spatial structure of imprinted cavities, could not enter into the cavities to produce a hydrogen bond interaction with MIPs. Therefore, the E-TBBPA-MIM showed poor adsorption capacity for these structural analogues of TBBPA.
Membranes 2023, 13, x FOR PEER REVIEW 12 of 16 groups of TBBPA resulted in the chemical adsorption of TBBPA by the imprinted cavities. For another, the chemical structure of TBBPA was perfectly matched to the shape, size, and spatial arrangement of imprinted cavities. The rebinding capacity of E-TBBPA-MIM could be recovered after the TBBPA removal by the eluent. The BP, BPA, and DDBP, which are nonmatched to the spatial structure of imprinted cavities, could not enter into the cavities to produce a hydrogen bond interaction with MIPs. Therefore, the E-TBBPA-MIM showed poor adsorption capacity for these structural analogues of TBBPA.

Conclusions
In this study, a novel TBBPA imprinted SiO2 NPs embedded PVDF microfiltration membrane (E-TBBPA-MIM) with enhanced permselectivity and rebinding capacity were successfully developed. The adsorption capacity of E-TBBPA-MIM for TBBPA was markedly greater than the pristine PVDF membrane and NIM. Adsorption data of TBBPA on the E-TBBPA-MIM fitted well with the Langmuir and pseudo-second-order models, suggesting that the adsorption process was a monolayer chemical adsorption process. Especially, the permselectivity factors of E-TBBPA-MIM for structural analogues (i.e., BP, BPA and DDBP) of TBBPA reached 5.24-6.71, significantly greater than those (1.17-1.56) of NIM. The rebinding capacity of E-TBBPA-MIM could still reach around 90% that of the initial after 5 adsorption/desorption cycles. The excellent adsorption performance, permselectivity and stability indicated the developed E-TBBPA-MIM as a promising option for efficient separation and removal of TBBPA from water.

Supplementary Materials:
The following supporting information can be downloaded at: www.mdpi.com/xxx/s1, Figure S1: Picture of experimental setup for synthesis and modification of SiO2 NPs; Figure S2: Picture of experimental setup for cross-flow filtration tests; Figure S3: Effects of (A) TBBPA/4-VP and (B) 4-VP/EGDMA molar ratios on Qe of MINs; Figure S4: XPS C1s spectra of

Conclusions
In this study, a novel TBBPA imprinted SiO 2 NPs embedded PVDF microfiltration membrane (E-TBBPA-MIM) with enhanced permselectivity and rebinding capacity were successfully developed. The adsorption capacity of E-TBBPA-MIM for TBBPA was markedly greater than the pristine PVDF membrane and NIM. Adsorption data of TBBPA on the E-TBBPA-MIM fitted well with the Langmuir and pseudo-second-order models, suggesting that the adsorption process was a monolayer chemical adsorption process. Especially, the permselectivity factors of E-TBBPA-MIM for structural analogues (i.e., BP, BPA and DDBP) of TBBPA reached 5.24-6.71, significantly greater than those (1.17-1.56) of NIM. The rebinding capacity of E-TBBPA-MIM could still reach around 90% that of the initial after 5 adsorption/desorption cycles. The excellent adsorption performance, permselectivity and stability indicated the developed E-TBBPA-MIM as a promising option for efficient separation and removal of TBBPA from water.

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/membranes13060571/s1, Figure S1: Picture of experimental setup for synthesis and modification of SiO 2 NPs; Figure S2: Picture of experimental setup for cross-flow filtration tests; Figure S3: Effects of (A) TBBPA/4-VP and (B) 4-VP/EGDMA molar ratios on Q e of MINs; Figure S4: XPS C1s spectra of (A) TBBPA-MINs and (B) E-TBBPA-MINs; Table S1: Specific surface area, average pore size and total pore volume of E-TBBPA-MINs and NINs; Table S2: Isotherm adsorption constants of E-TBBPA-MIM and NIM for TBBPA; Table S3: Kinetics constants for adsorption of TBBPA onto E-TBBPA-MIM and NIM; Table S4: Selective permeation parameters of E-TBBPA-MIM and NIM for TBBPA and its structural analogues.  Institutional Review Board Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.