Humic acid is one of the natural organic materials (NOMs) that has been considered as one of the main contaminants in surface water. The presence of HA in water causes to produce disinfection by-products such as trihalomethanes after chlorination, unpleasant color and taste [1-4]. Consequently, it is necessary to remove them from water using effective treatment processes. Various treatment technologies including chemical coagulation/flocculation [5, 6], electrocoagulation [7-9], ultrafiltration [10-13] and adsorption [14-17] have been extensively studied for humic acid removal. Adsorption is considered as very effective, economical, versatile and simple process. Despite the expansion and application of adsorption in HA removal, difficulties in diffusion to access inner pores and separation from water solutions are still critical problems for the wider application of adsorption [18, 19]. In recent years, special attention was given to use adsorptive membranes in water treatment as a promising alternative method to solve the problem of adsorption [20, 21]. Most of these studies have mainly focused on heavy metal removal from water [21-25]. Nevertheless, there are rare reports in published literature about the application of adsorptive membranes in humic acid removal. According to Thuyavan et al. embedding of 5 wt% zirconia in polyethersulfone membrane as adsorptive membrane had 97% HA rejection . Panda et al. studied HA removal by chitosan coated iron-oxide-polyacrylonitrile adsorptive membranes and evaluated their antifouling behavior. The obtained results revealed that electrostatic interaction between humic acid and positive charge of chitosan improves the HA adsorption .
Hence, current research has focused on developing an efﬁcient adsorptive membrane by embedding positively charged adsorptive particles for the maximum removal of HA.
In our previous work, arginine and lysine amino acids were used as modifiers for montmorillonite (MMT) on the basis of their two positive charge chains in natural pH conditions . The obtained results showed that modiﬁed MMT with arginine (MMT-A) can improve the adsorption capacity and regeneration capability of MMT for arsenic oxyanions removal. Consequently, the MMT-A was used for preparing PSf based adsorptive membranes in other studies [17, 29, 30]. The adsorption properties of modified MMT with positively charged amino acids and studies related to its applications in arsenic removal suggest that PSf/MMT-A adsorptive membranes can be useful for humic acid removal.
In order to improve the membrane performance for HA removal, the adsorption capacity of the membrane can be further improved through the use of the proper structure of polymer matrix. Studies related to the effect of membrane morphology on the behavior of adsorptive membranes have been very few. Liu et al. prepared polysulfone- graft -monoazabenzo-15-crown-5 ether adsorptive membrane for separation of lithium isotopes . They found that sponge-like structure leads to an increase in contact area between adsorption sites and lithium ions. Therefore, it is necessary to enhance our understanding of the effect of membrane morphology on the adsorption behavior of membranes to extend the use of these membranes for water treatment. In this study, polyethylene glycol (PEG) is employed as an additive to control the morphology of membranes. In the following, MMT-A was embedded in PSf membranes with finger-like and sponge-like pore structures. The prepared membranes were extensively evaluated in terms of HA adsorption capacity, mechanism, kinetics, pH effect and regeneration capability. Additionally, the description of the equilibrium and kinetic data using theoretical models was carried out.
Materials and methods
The polymer used as a matrix in this research was polysulfone provided by Solvay Advanced Polymer LLC. Natural montmorillonite (Cloisite Na+) with particle size of 9-13 µm was supplied by Southern Clay Products Inc., USA. Other chemicals including N-methyl-2-pyrrolidon (NMP) (ρ: 1030 kg/m3), PEG (Mw: 20000 g/mol), arginine, NaOH and HCl were purchased from Merck. All chemicals and reagents were analytical grades.
MMT was modified with arginine by using the method introduced in the literature . Adsorptive membranes (1.5 wt% of MMT-A) were prepared by using NIPS method . This amount of MMT-A was used to maximize its effect on the pure water flux and adsorption capacity of membranes, simultaneously. The composition of each casting solution and weight percentage of the casting solution is given in Table 1.
MMT-A was dissolved in NMP and sonicated for 30 min using prob sonicator (Sonopuls HD 3200, Bandelin) at room temperature. After the addition of PSf for M3 membrane and PSf-PEG for M4 membrane, the mixture was heated at 60°C for 8h. Then the resulted homogenous solution was sonicated for 15 min and then enough time was given for bubbles to be completely released. The resulted homogenous solution was cast at 150µm thickness. Immediately after casting, the ﬁlm was immersed in a water bath to initiate phase inversion and then dried.
Neat membranes were prepared from a casting solution of PSf (15 wt%) with the same procedure as mentioned above without the addition of particles.
The morphology of the membranes was characterized by FE-SEM (MIRA3 FEG-SEM, Tescan). The hydrophilicity of membranes was evaluated by measuring the contact angle between the membrane surface and water droplet using a contact angle goniometer (PGX, Thwing-Albert Instrument Co.). To measure the membrane porosity, samples were cut into specific sizes before being weighed in a digital balance. The porosity of membranes was calculated using the following equation :
where ε (٪) is membrane porosity, ww is the mass of wet membrane in g, wd is the mass of dry membrane in g, ρw, and ρp are the density of isobutanol and polymer in (g/cm3), respectively. In the determination of the porosity, the average of 3 samples of each membrane was reported.
Pure water flux of membranes was determined using a dead-end filtration system having 5cm2 of membrane area. To minimize compaction effects, the pre-wetted membranes were compacted for 30 min at 2 bar. Then the pressure was reduced to 1.5 bar and after reaching steady state, the water flux was calculated.
The surface porosity of membranes was determined using Digimizer Image Analysis software. To determine the mean pore size of the membranes, the filtration velocity method was used following produce already established [29,30].
The static adsorption behavior of the prepared membranes was investigated with the batch experiments for humic acid solutions in the appropriate concentrations ranging from 2 to 20 ppm. For batch adsorption, adsorptive membranes (total weight 0.1g) were immersed in the HA solution (100 mL) in tap water and stirred at room temperature, while the pH of the solution was adjusted to 7.0. HA was adsorbed on the membranes thereby reducing HA concentration in the bulk until equilibrium was reached.
To determine the adsorption kinetics, the concentrations of residual HA in the prepared solutions were analyzed by UV spectrophotometer (Bio Quest CE2501) at different time intervals. The equilibrium adsorption capacity and removal efficiency of membranes were calculated as follows:
Where Co (mg/L), Ce (mg/L) and Ct (mg/L) are concentrations at the initial, equilibrium and time t in the solution, respectively, V is the total volume (L) of the HA solution and Mm is the mass (g) of dry membrane used in the adsorption study.
To study the eﬀect of pH on the adsorption behavior, the membranes were added into HA solutions, which were prepared at different pH (3-9). The samples were equilibrated at 25 °C in a shaking incubator and after 48 h, the concentration of HA was measured.
Filtration experiments were carried out in a dead-end filtration set up with an effective surface area of 5cm2 for membrane. The concentration of HA in the feed solution was set at 1ppm and pH of the solution was adjusted to 7.0. While 5 mL of effluent was collected at different time intervals to measure the respective HA concentrations. After saturation of membranes by HA, 50 mL NaOH solution with pH=9 filtered through the membrane. After regeneration, the second cycle of experiments was initiated and this process was repeated for five cycles.
RESULTS AND DISCUSSION
Characterization of membranes
A detailed characterization of the structural changes of MMT which take place after modification with arginine involving XRD, FTIR, zeta potential methods were reported earlier . In this paper only the examination of the adsorptive membrane structure and its effect on adsorption properties were discussed.
The morphology imaging towards the surface and cross-sections of membranes can be found in Fig.1. The FE-SEM images towards the surface and cross-section of M1 and M2 membranes can be found in Fig. 1a, 1b, respectively. It can be seen clearly that the M1 membrane has a sponge structure. In contrast to this, the M2 membrane has macrovoids which are contributed with a small tear like voids throughout. The different structure is due to PEG added to the polymer solution of the membrane. It is obvious that membrane structure in NIPS method depends on the thermodynamic equilibrium properties and transfer rates of solvent and non-solvent. The formation of macrovoids in M2 membrane is explained in terms of increase in mass ratio of non-solvent inflow and solvent outflow due to hydrophilic properties of PEG .
By adding MMT-A in PSf solution with PEG, macrovoids quickly transitioned into narrower voids that spanned the entire cross-section of the M4 membrane (Fig. 1d). The addition of MMT-A causes to increase the solution viscosity and can help to extend the macrovoids. In contrast, no differences in cross-sectional morphologies were observed upon MMT-A addition for the M3 membrane (Fig. 1c).
Obviously, a significant increment in surface pore sizes can be observed visually by comparing the FE-SEM top surface images of M1 (Fig. 1e); and M2 (Fig. 1f). The main reason is due to the increase in the ratio of non-solvent inﬂow to solvent outﬂow by adding PEG. The top surface images of adsorptive membranes, Fig. 1g, 1h show that the number of pores increases by incorporating MMT-A for M3 and M4.
The results of the membrane characterization in Table 2 display the porosity, mean pore size, surface porosity and the pure water flux of the prepared membranes.
As illustrated in Table 2, the porosity of the PSf membrane increases from 70 to 78% after the PEG addition and further increases to 74 and 84 for M3 and M4 membranes after embedding MMT-A, respectively. As can be seen from Table 2, embedding the MMT-A in PSf-PEG causes a decrease in the surface pore size and increase in the surface porosity significantly. The decrease of surface pore size is may be due to the increased viscosity of the solutions by embedding MMT-A . However, the surface pore size and porosity values were not affected by adding PEG and MMT-A in PSf membrane.
Similar to the membrane porosity, the PWF of membranes increases by embedding MMT-A. However, a significant increase in water flux was found for the M4 membrane in comparison to the M3 membrane. These results were consistent with the observation from the FE-SEM images.
Fig. 2 depicts the water contact angle of prepared membranes. It could be found that the presence of PEG resulted in improved membrane hydrophilicity (lower contact angles). Furthermore, all the adsorptive membranes have less contact angle than the neat membranes because of hydrophilic polar amine functional groups of MMT-A.
Static humic acid adsorption behavior
The experimental isotherm curves of prepared membranes at different initial concentrations of HA are shown in Fig. 3. From the obtained results, it is evident that HA adsorption capacity is negligible for M1 and M2 membranes and increases as the MMT-A is embedded in the polymer solution. It is obvious that adsorption capacity increases for M3 and M4 with the increase of HA concentration. Furthermore, qe for M3 is higher than M4, which implies that cellular structure provides effective surface area available for adsorption.
In order to understand the adsorption in more detail, Langmuir and Freundlich isotherms as two common models were used. Langmuir model assumes monolayer adsorption of adsorbate at specific homogeneous sites without any interactions among the adsorbate molecules. In contrast, Freundlich model assumes multilayer adsorption that can occur on heterogeneous surfaces. Langmuir and Freundlich models can be expressed as Eqs. (4) and (5), respectively:
Where KL and KF are the Langmuir and the Freundlich adsorption constants, respectively. Moreover, qmax is the maximal adsorption capacity and n is the heterogeneity factor.
The fitting of the Langmuir and Freundlich models to the adsorption data of HA on the M3 and M4 membranes was carried out and the results were shown in Fig. 4. The corresponding fitting parameters obtained from both Langmuir and Freundlich models were illustrated in Table 3.
According to the correlation coefficients (R2), the Freundlich model describes the adsorption process better than the Langmuir model. Due to heterogeneity and complexity of the HA composition, various adsorption mechanisms can occur on adsorption sites. Ligand exchange and electrostatic attraction are proved to be main mechanism of HA adsorption on the edges of montmorillonite . Moreover, electrostatic interaction between positive charge of arginine and negative charge of HA causes HA adsorption.
In order to better understand the rate of adsorption, adsorption kinetics was investigated and the effect of time on removal efficiency of M3 and M4 membranes was shown in Fig. 5.
The experimental outcomes indicate that the HA adsorption capacity increases with time. The rate of the adsorption for M3 and M4 membranes increased quickly in the first 4h and 1h, respectively and then the less marked effect was observed on increasing the contact time for both membranes.
The rapid adsorption of HA on the M4 respect to M3 may be due to the presence of large pores at the M4 membrane which facilitates the solution penetration and adsorption of HA.
In this study, the kinetics of the adsorption processes were analyzed using pseudo-first-order and pseudo-second-order equations; their mathematical equations are as below:
Where qe and qt are the amounts of HA adsorbed at equilibrium and time t (h), respectively; k1(h-1) and k2 (g.mg-1.h-1) are the equilibrium constants of the pseudo-first-order and pseudo-second-order models, respectively.
The progress of the HA adsorption was examined by ﬁtting the experimental data using pseudo-first-order and pseudo-second-order models. As observed from Fig. 6, it is evident that R2 for the pseudo-second-order is higher than the pseudo-first-order. Furthermore, this indicates that the adsorption is controlled by a chemisorption process. The corresponding fitting parameters for the adsorption kinetic models from the analysis were listed in Table 4. By comparison, the values of qe in pseudo-second order model were more in accordance with the experimental data.
The dependence of HA adsorption upon the solution pH for M3 and M4 as shown in Fig. 7. According to Fig.8, it becomes clear that HA adsorption decreases with increasing solution pH for both membranes, which can be interpreted by the surface charges of adsorptive particles embedded in membranes. The isoelectric point of MMT-A was 8 according to previous work . When the pH of the solution is below 8, the surface of MMT-A has positive charge which resulted in the electrostatic attraction with negatively charged carboxyl and phenol groups of HA. It is also evident from Fig.7 that adsorption capacity of M3 is higher than M4 at all pH ranges.
Dynamic humic acid adsorption behavior
In the study of the dynamic adsorption of HA, M3 and M4 membranes were tested. Breakthrough point as the position where the concentration of pollutant rises to 10% of the feed concentration was determined for membranes. As shown in Fig. 8 (a), the M3 and M4 can effectively treat more than 100 mL and 50 mL of HA-contaminated water to ensure the HA concentration of effluents less than the MCL (1ppm), respectively. According to the results, the treatment capacity of the M3 and M4 membranes can be calculated as 200 L/m and 100 L/m, respectively. Consequently, M3 with high treatment capacity was selected for adsorption-desorption study. The reusability of M3 membrane for filtration of 100 mL HA solution (10 ppm) was investigated by running five adsorption-desorption cycles. As is apparent in Fig. 8 (b) there is low reduction in the removal efficiency of M3 even after five cycles. These results implied that M3 can be simply regenerated by treatment.
Modified montmorillonite with amino acid (MMT-A) embedded in PSf with different morphologies. The results showed that MMT-A greatly improved the hydrophilicity, permeability, and porosity of the membrane with a finger-like pore structure. No signiﬁcant improvement was observed in the cellular structure membrane. Cellular membranes maintained higher HA adsorption capacity during the batch adsorption, indicating that this structure enhanced available adsorption sites. The faster HA adsorption was obtained for finger-like structure due to the presence of large pores. The cellular membrane demonstrated the highest adsorption capacity and its suitability in HA removal investigated through dead-end filtration. The obtained results suggest that this PSf adsorptive membrane with cellular structure can be effectively used to HA removal from water for multiple adsorption-desorption cycles.
CONFLICTS OF INTEREST
There are no conflicts to declare.