Enhanced gypsum scaling by organic fouling layer on nanofiltration membrane: Characteristics and mechanisms

Water Research 91 (2016) 203e213

Contents lists available at ScienceDirect

Water Research journal homepage: www.elsevier.com/locate/watres

Enhanced gypsum scaling by organic fouling layer on nanofiltration membrane: Characteristics and mechanisms Jiaxuan Wang a, b, Lei Wang a, *, Rui Miao a, Yongtao Lv a, Xudong Wang a, Xiaorong Meng c, Ruosong Yang a, Xiaoting Zhang a a b c

School of Environmental & Municipal Engineering, Xi'an University of Architecture and Technology, Yan Ta Road, No. 13, Xi'an 710055, China Leibniz Institute of Surface Modification, Permoserstraße 15, Leipzig D-04318, Germany School of Science, Xi'an University of Architecture and Technology, Yan Ta Road, No. 13, Xi'an 710055, China

a r t i c l e i n f o

a b s t r a c t

Article history: Received 3 November 2015 Received in revised form 6 January 2016 Accepted 10 January 2016 Available online 12 January 2016

To investigate how the characteristics of pregenerated organic fouling layers on nanofiltration (NF) membranes influence the subsequent gypsum scaling behavior, filtration experiments with gypsum were carried out with organic-fouled poly(piperazineamide) NF membranes. Organic fouling layer on membrane was induced by bovine serum albumin (BSA), humic acid (HA), and sodium alginate (SA), respectively. The morphology and components of the scalants, the role of Ca2þ adsorption on the organic fouling layer during gypsum crystallization, and the interaction forces of gypsum on the membrane surface were investigated. The results indicated that SA- and HA-fouled membranes had higher surface crystallization tendency along with more severe flux decline during gypsum scaling than BSA-fouled and virgin membranes because HA and SA macromolecules acted as nuclei for crystallization. Based on the analyses of Ca2þ adsorption onto organic adlayers and adhesion forces, it was found that the flux decline rate and extent in the gypsum scaling experiment was positively related to the Ca2þ-binding capacity of the organic matter. Although the dominant gypsum scaling mechanism was affected by coupling physicochemical effects, the controlling factors varied among foulants. Nevertheless, the carboxyl density of organic matter played an important role in determining surface crystallization on organic-fouled membrane. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Nanofiltration membrane Organic fouling Gypsum scaling Quartz crystal microbalance with dissipation (QCM-D) Interaction force

1. Introduction With the increasing requirement for high quality water, membrane processes have attracted worldwide attention because of their potential to provide efficient and enhanced water purification as well as environmental and commercial benefit. Nanofiltration (NF) membranes are widely used in systems for seawater desalination, drinking water purification, wastewater treatment and reclamation, because they are capable of removing a broad range of organic, inorganic, and microbial contaminants in a single treatment step with their nanoscale pore size and/or surface charge (Hong and Elimelech, 1997; Van Der Bruggen et al., 2003). Removal of hardness at low operating pressure, high productivity, and low operational cost also make NF membranes one of the more favorable alternatives in the drinking water purification industry.

* Corresponding author. E-mail address: [email protected] (L. Wang). http://dx.doi.org/10.1016/j.watres.2016.01.019 0043-1354/© 2016 Elsevier Ltd. All rights reserved.

Although its application in the water industry is noteworthy (Mohammad et al., 2015; Van Der Bruggen et al., 2003), just like other membrane processes, NF also suffers from the enduring problem of membrane fouling caused by organic solutes, inorganic solutes, colloids, or biological solids (Mohammad et al., 2015). Membrane fouling has been a major obstacle for most applications of NF processes in the water industry, especially in cases where high concentrations of natural organic matters (NOMs) and inorganic constituents occur (Le Gouellec and Elimelech, 2002; Thorsen, 2004). Because of the retention effect of NF membranes, the ionic concentrations near or on the NF membrane surface will increase, and might exceed the solubility limit of sparingly soluble salts. Consequently, inorganic scalants could form. This situation usually happens in membrane desalination systems with high product water recovery. The most common constituents of scale are calcium carbonate, gypsum (CaSO4$2H2O), barium/strontium sulfate and €fer et al., 2005). silica, although other potential scalants exist (Scha

204

J. Wang et al. / Water Research 91 (2016) 203e213

Among the range of sparingly soluble salts that lead to scaling in membrane processes for seawater/brackish water desalination, gypsum is one of the most ubiquitous scaling sources because of its relatively high concentration in natural waters (Dydo et al., 2003; Le Gouellec and Elimelech, 2002). However, control of gypsum is a major challenge in the development of membrane processes for desalination, since it cannot be effectively prevented by lowering feedwater pH or chemical cleaning. As a result of scaling, membrane flux declines severely, permeate quality reduces, and the life of the membrane system is significantly shortened (Lee et al., 1999; Lee and Lee, 2000). Consequently, obtaining a full understanding of gypsum scaling mechanisms as well as how the related factors affect gypsum scaling is very important to mitigate gypsum scaling. Various parameters affecting the crystallization process have already been identified, such as temperature (Hoang et al., 2007), pH (Her et al., 2000), operating pressure (Lee and Lee, 2000), flow velocity (Lee and Lee, 2000), salt concentration (Le Gouellec and Elimelech, 2002), types of antiscalant (Rahman, 2013), coexisting sparingly soluble salts (Sheikholeslami, 2003), and other metal ions (Hamdona and Al Hadad, 2007). In addition, coexisting NOM has also been considered to interfere with the formation of gypsum precipitates (Barcelona and Atwood, 1978) and the performance of NOM on various forms of scaling on reverse osmosis (RO)/NF membranes has also been investigated (Wiesner, 2007). A previous study demonstrated that NF membrane flux decline because of gypsum scaling was much slower in the presence of 3 mg L1 humic acid (HA) than in the absence of HA (Le Gouellec and Elimelech, 2002). A similar phenomenon was found in gypsum scale formation on RO/NF membranes in the presence of HA, where HA acted like an antiscalant and substantially decreased the rate of formation of gypsum scaling (Lee et al., 2009). Organic fouling occurs widely in NF membrane processes (Hong and Elimelech, 1997; Seidel and Elimelech, 2002) because organic macromolecules are ubiquitous in all natural water sources (e.g., seawater, surface water, and groundwater). Although membrane cleaning has the ability to remove the adhesive foulants and restore the flux to various degrees (Mi and Elimelech, 2008, 2010bbib_Mi_and_Elimelech_2010b), the accumulation of organic foulants on the membrane surface over time is inevitable. It is likely that membrane organic fouling usually precedes inorganic scaling in typical NF systems (Huiting et al., 2001; Yang et al., 2008). Hence, a pregenerated organic fouling layer would probably change the physicochemical properties of the membrane surface. Consequently, the subsequent inorganic scaling behaviors will likely be affected by the specifics of the pregenerated organic fouling layer. For instance, in a forward osmosis (FO) membrane process, when gypsum coexisted with alginate, the permeate flux decreased faster than for alginate/gypsum alone; moreover, the most severe flux decline was observed when alginate was precoated on the FO membrane surface (Liu and Mi, 2012). Therefore, it is significant to investigate how the surface characteristics of NF membranes are altered by organic fouling layers and further influence gypsum scaling for the development of useful protocols for fouling mitigation. Nevertheless, to our knowledge, there has been no reported research on this aspect to date. This study aimed to investigate the influence of pregenerated organic fouling layers on the subsequent gypsum scaling of NF membranes, and determine the characteristics of gypsum scaling behavior of diverse organic conditions as well as virgin membrane to elucidate the underlying mechanisms. Three typical organic foulantsdbovine serum albumin (BSA), HA, and sodium alginate (SA)dwere used to represent widespread organic foulants (proteins, humics, and polysaccharides, respectively). To distinguish the influence of the organic fouling layer in a pressure-driven filtration

process, bench-scale experiments on gypsum scaling and subsequent cleaning of different conditioned membranes were conducted and compared. Scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) analyses of the scalants were performed to determine whether the characteristics of the membrane surface impacted the gypsum scaling morphology and chemical composition. Quartz crystal microbalance with dissipation (QCM-D) was utilized to investigate the influence of the chemical characteristics of each organic layer on the adsorption of calcium ions and quantitatively evaluate the role of Ca2þ adsorption on the structural characteristics of the adhesion layer. Atomic force microscopy (AFM) force measurements were conducted to determine the interaction characteristics of different organic fouling layers on gypsum scaling and to elucidate the scaling mechanisms at the nanoscale. The ultimate goal was to provide useful information about targeted removal of organic matter for feedwater pretreatment in desalination processes. 2. Materials and methods 2.1. Poly(piperazineamide) NF membrane The membranes used in this study were poly(piperazineamide) NF membranes with polysulfone support. The detailed process of manufacture of the fabricated NF membranes is given in Section S1 of the Supporting Information. The average pure water flux of the NF membrane tested under 0.6 MPa was 48 L m2 h1. The divalent salt rejection was 94%, determined with a 2 g L1 MgSO4 feed solution at an applied pressure of 0.6 MPa and a cross-flow velocity of 8 cm s1. Other properties and AFM images of the prepared poly(piperazineamide) NF membrane are shown in Table S1 and Fig. S1 in Section S2. 2.2. Organic foulants and inorganic salts Model organic foulants (BSA, HA, and SA) were purchased from SigmaeAldrich Co. LLC (St. Louis, MO) and received in powder form. The stock solution of each organic foulant (2 g L1) was prepared as detailed in Section S3. Inorganic salts including sodium chloride (NaCl), sodium sulfate (Na2SO4), and calcium chloride (CaCl2) were provided by Tianli Chemical Reagent Co. Ltd. (Tianjin, China). All inorganic reagents were analytical grade and used without further purification. Deionized (DI) water (conductivity < 2 ms cm1, pH ¼ 7e7.5) was used throughout this study unless otherwise stated. 2.3. Protocols for the NF membrane fouling, scaling and cleaning experiments The fouling, scaling, and cleaning experiments were conducted with a bench-scale cross-flow system that comprised a rectangular plate-and-frame membrane cell with a dimensioned rectangular channel (12 cm long, 4 cm wide, and 0.3 cm deep) inside. A schematic diagram of the setup is depicted in Fig. S2. The protocols for the entire experiment are illustrated in Fig. 1. Each experiment included five different and consecutive stages, namely compaction, conditioning, organic fouling, gypsum scaling, and water cleaning, as follows: (i) the clean membrane was compacted and equilibrated with DI water at 0.7 MPa for at least 5 h; (ii) the membrane was conditioned under baseline conditions of around 0.6 MPa (the pressure was slightly adjusted as needed) for at least 3 h until a satisfactory steady initial baseline flux was obtained; (iii) the organic fouling experiment was continuously performed with freshly prepared organic foulant solution for 24 h to introduce an organic fouling layer on the membrane surface; (iv)

J. Wang et al. / Water Research 91 (2016) 203e213

205

generating an organic adlayer on the modified quartz crystal sensor. The behavior of Ca2þ adsorbed onto organic adlayers as well as the structure of adhesion layers was characterized using a QCMD instrument (E1, Q-Sense, Sweden). A schematic diagram of the synthesis of the poly(piperazineamide)-coated quartz crystal sensor is shown in Fig. 2. The following sequence of steps was employed: (i) goldcoated crystal sensor (QSX 301 Au, Q-Sense) surface cleaning; (ii) deposition of a 2-aminoethanethiol adhesion layer; (iii) absorption of TMC onto the aminated sensor surface; (iv) polymerization of poly(piperazineamide) film through the reaction of PIP with TMC; and (v) construction of a multilayer poly(piperazineamide) surface. The prepared crystal sensor was analyzed to determine its chemical composition, hydrophilicity, and surface topography. The specific preparation procedures and detailed characterizations are described in Section S5. Four kinds of test solutions were included, namely ultrapure water, background electrolyte solution, organic matter solution, and CaCl2 solution, as detailed in Section S5. During each QCM-D experiment, the test temperature was kept at 23  C and the cross-flow rate was fixed at 100 mL min1. First, a freshly modified crystal sensor was housed in the flow chamber of the QCM-D system. Subsequently, the baselines of frequency shift (Df, Hz) and energy dissipation change (DD) were established in ultrapure water (~10 min), then in background electrolyte solution (~15 min) to stabilize the sensor. After a stable baseline was established in the background electrolyte solution, an organic matter solution (i.e., BSA, HA, or SA) was introduced to allow adsorption of organic molecules to occur. Adsorption equilibrium was assumed when Df and DD stabilized. After the adsorption of organic macromolecules reached equilibrium, the feed was then switched back to the background electrolyte solution to remove any unstable adsorbed organic matter. Then, ultrapure water was delivered into the flow cell to acquire another baseline stage for the evaluation of subsequent Ca2þ adsorption on each organic adlayer. After this, CaCl2 solution was pumped through the system until a new equilibrium was obtained. Finally, the system was rinsed with ultrapure water to eliminate the uncombined Ca2þ. The variations in Df and DD were measured for five overtones (n ¼ 3, 5, 7, 9) and the 5th overtone is presented for further evaluation.

Fig. 1. Schematic illustration of the experimental protocols.

the gypsum scaling experiment was immediately carried out for about 24 h by switching the organic foulant solution to newly prepared gypsum scaling solution; (v) membrane cleaning was conducted by simple surface flushing with DI water for 20 min with the cross-flow rate increased to 15 cm s1 and in the absence of permeate flux. The baseline experiment was then performed again to determine the membrane flux recovery. The compositions of the feed solutions are summarized in Table 1. Note that the solubility product of Ca2þ and SO2 4 concentrations in the feed solution of the scaling experiment was made slightly higher (with a saturation index [SI] of 1.3) than that of gypsum so that scaling could take place at a reasonable speed (Mi and Elimelech, 2010a). A high organic foulant concentration (100 mg L1) was chosen to accelerate the organic fouling process. Other details about the preparation of feed solutions and experimental conditions are given in Section S4. The initial permeate flux of each organic fouling/gypsum scaling stage was used for flux normalization since no organic/inorganic fouling occurred. To confirm the reproducibility of the results, all experimental processes were conducted in duplicate. Membrane specimens for EDS analysis were prepared in a separate batch of experiments (without the water cleaning stage).

2.5. Analytical technologies Surface morphology and roughness of membranes and quartz crystal sensors were obtained by AFM (MultiMode 8.0, Bruker, Germany) equipped with a NanoScope V controller testing in contact mode. SEM (JSM-6510LV, JOEL Co., Japan) was applied to obtain the surface images of membranes prepared by static adsorption experiments which were conducted by vertically soaking organic-fouled and virgin membranes in supersaturated CaSO4 solution (SI ¼ 3.1; 38 mM NaCl, 40 mM Na2SO4, and 70 mM CaCl2) at 25  C for 24 h. EDS (Quanta 200, FEI, USA) was used to analyze the elemental composition of the scaling layers. The chemical compositions of modified quartz crystal sensors and poly(piperazineamide) NF membrane were analyzed by an X-ray photoelectron spectrometer system (K-Alpha, Thermo Fisher

2.4. QCM-D measurements Quartz crystal sensors can be modified to yield special surface chemistry (Contreras et al., 2011; Steiner et al., 2011). Here, a poly(piperazineamide)-coated quartz crystal sensor was prepared using a polymerization reaction with piperazine (PIP) and trimesoyl chloride (TMC), which produced a surface chemistry similar to that of the poly(piperazineamide) NF membrane. Thus, the organic fouling layer on the NF membrane surface could be imitated by

Table 1 Chemical composition of the feed solutions. Experiment

NaCl (mM)

Organic matters (mg L1)

Na2SO4 (mM)

CaCl2 (mM)

pH

Total ionic strength (mM)

Baseline BSA fouling HA fouling SA fouling Gypsum scaling

150 150 150 150 19

0 100 100 100 0

0 0 0 0 20

1 1 1 1 35

7.5 7.5 7.5 7.5 7.5

150 150 150 150 150

206

J. Wang et al. / Water Research 91 (2016) 203e213

Fig. 2. The schematic diagram of the synthesis of the poly(piperazineamide)-coated quartz crystal sensor.

2.6. AFM force measurements AFM in conjunction with a gypsum particle probe was utilized to measure the gypsumevirgin/organic-fouled membrane and gypsumegypsum interaction forces. A functionalized probe was prepared by gluing a single gypsum crystal particle (6 mm in diameter) onto a commercial tipless SiN AFM cantilever (NP-10, Bruker, Germany), following the same procedures as described in detail by Mi and Elimelech (2010a). An SEM image of the gypsum probe is shown in Fig. S5. Because it is difficult to use an organic-coated membrane sample prepared with a cross-flow system in the AFM liquid cell without disturbing the thick fouling adlayer, organicfouled membranes were prepared by soaking clean membranes vertically in the relevant organic foulant solutions. The composition of each organic solution was identical to that used in the organic fouling experiment and the soak time was 24 h so that a layer of organic molecules could adsorb onto the clean membrane surface. The adhesion force measurements were performed in a fluid cell filled with the test solution using a closed inlet/outlet loop under contact mode. The procedures and test solutions for force measurements were similar to those used previously (Mi and Elimelech, 2010a). It is worth noting that the force measurements were performed with either saturated or supersaturated (SI ¼ 3.1) gypsum solutions to prevent dissolution of the gypsum particle probe as well as to obtain the interfacial forces during the different stages of gypsum scaling. The higher SI in the force measurements (SI ¼ 3.1) than in the bench-scale experiments (SI ¼ 1.3) was intended to accelerate the nucleation progress (Mi and Elimelech, 2010a). The interfacial forces were first measured with saturated gypsum solution, where no precipitation or dissolution took place, to estimate gypsumemembrane interactions. After this, a freshly prepared supersaturated gypsum solution was injected into the fluid cell and equilibrated for 50 min so that gypsum had already started precipitating in the solution and/or on the membrane surface to

simulate the scaling layer (Mi and Elimelech, 2010a). Force measurements were performed at five different locations on each sample and 20 force curves were taken at each location. 3. Results and discussion 3.1. Bench-scale experiments 3.1.1. Membrane flux behavior under organic fouling conditions The flux of each organic condition became steady after a high concentration organic foulant (BSA, HA, or SA) was introduced for a certain period (Fig. 3), indicating that a layer of stable adhesive organic macromolecules had been generated on the membrane surface. Accelerated flux loss in HA and SA conditions compared with the BSA condition probably resulted from the formation of a dense, thick HA/SA fouling layer that caused an enormous increase in cake resistance and poor permeate capacity. These behaviors are also consistent with organic fouling characteristics observed in NF (Hong and Elimelech, 1997; Li and Elimelech, 2004; Seidel and Elimelech, 2002), RO (Lee et al., 2010; Lee and Elimelech, 2006), and FO systems (Lee et al., 2010; Mi and Elimelech, 2008). 3.1.2. Membrane flux behavior under gypsum scaling conditions To demonstrate better the effect of different membrane surface conditions on gypsum scaling, normalized water flux is presented

BSA HA SA

1.0

0.8

Normalized Flux

Scientific Inc., USA) equipped with an Al Ka X-ray source and a monochromator. Contact angle measurements were carried out by a contact angle analyzer (SL200A, USA KINO Industry CO. Ltd.) using the sessile drop method with ultrapure water at 25  C. Streaming potential measurement was performed with a DelsaNano C instrument (Beckman Coulter, Inc., USA) and conducted with the solution used in the bench-scale baseline experiment. The solutions used for preparing organic-fouled membranes were identical to those used in the corresponding organic fouling experiments. The zeta potentials of virgin/organic-fouled NF membrane surfaces were calculated from the measured streaming potentials using the HelmholtzeSmoluchowski equation (Abramson, 1933). All measurements were performed at least three times.

0.6

0.4 0.2

0.0

0

200

400

600

800

1000 1200 1400 1600

T (min) Fig. 3. Normalized flux decline curves obtained during the NF organic fouling runs with BSA, HA, and SA. All fouling runs were performed with identical solutions (i.e., 100 mg L1 organic foulant concentration, 1 mM CaCl2, 150 mM NaCl, and pH 7.5 ± 0.1) in recycling mode. Experimental conditions: cross-flow velocity of 8 cm s1, initial flux (J0) of 35 L m2 h1, and temperature of 25 ± 1  C.

J. Wang et al. / Water Research 91 (2016) 203e213

0.6

0.4 BSA HA SA Virgin membrane

0.2

0.0

0

200

400

600

800

1000

1200

1400

T (min) Fig. 4. Normalized flux decline curves obtained during gypsum scaling on the organicfouled and virgin membranes. The scaling solution contains 19 mM NaCl, 20 mM Na2SO4, and 35 mM CaCl2, with a gypsum saturation index (SI) of 1.3. Initial flux (J1) of 21 L m2 h1, other experimental conditions of the scaling experiments were identical to those described in Fig. 3.

as a function of time in Fig. 4. To identify any possible effects of the organic fouling adlayer on subsequent inorganic scaling, a gypsum scaling experiment with virgin membrane was also performed for comparison. The results presented in Fig. 4 are quite notable, demonstrating significant diverse impacts of different membrane surface conditions on flux diminution in the gypsum scaling process. The degree and rate of flux decline increased in the order of virgin membrane < BSA < HA < SA. There were some other notable characteristics of gypsum scaling behavior under different conditions. Initially (i.e., at the preliminary stage), the flux decline was caused by concentration polarization that resulted in additional resistance to water transporting through the membrane. Compared with other conditions, the substantial flux reductions under HA and SA conditions at the preliminary stage were likely because of enhanced concentration polarization upon HA- and SA-fouled membrane surfaces. Moreover, in all conditions, the continued and evident flux decrease started after a certain period of plateau (i.e., the flux remained almost constant during this stage). This phenomenon is consistent with a previous study where CaSO4 deposition on the NF membrane surface started after an induction period characterized as a “delayed” period before flux decline occurred (Lin et al., 2005). This delay period could be attributed to the time needed to generate sufficient nuclei from the supersaturated solution (Le Gouellec and Elimelech, 2002; Subir Bhattacharjee, 2002). Therefore, HA and SA conditions had a shorter induction period than the other conditions tested, probably because of faster gypsum nuclei generation. The induction period under BSA condition was shorter than that of virgin membrane, perhaps because of more rapid nucleation, which was caused by cake-enhanced concentration polarization (CECP) (Hoek and Elimelech, 2003; Lee et al., 2005). Once sufficient CaSO4 nuclei were formed, significant crystallization of the solids could occur. The crystallization and subsequent deposition of the solids on the membrane surface led to cake formation (Lin et al., 2005). The greatest flux reduction in all conditions occurred during this stage of gypsum scaling, possibly as a result of the growth of cake thickness and compactness. Specifically, flux declines under HA and SA conditions were more serious and rapid than the other conditions, probably because of thicker and denser gypsum scaling

3.1.3. Cleaning reversibility The normalized fluxes of all conditioned NF membranes after scaling and cleaning experiments are shown in Fig. 5. These fluxes are normalized by the clean membrane flux obtained from the baseline experiment. Flux was almost fully reversible (94%) by simple physical cleaning with DI water in the virgin membrane

0.5

0.4

1.0

Normalized flux after gypsum scaling Normalized flux recovery by cleaning

0.8 0.3

0.6

0.2

0.4

0.1

0.0

0.2

BSA

HA

SA

0.0 Virgin membrane

Normalized flux recovery by cleaning

Normalized Flux

0.8

layers. Furthermore, during the last period of the experiment, the fluxes under both HA and SA conditions had reached a steady state, where the solid deposition was balanced by the shearing force caused by the cross-flow feed solution. In conclusion, the flux decline during gypsum scaling on virgin/organic-fouled membrane in this study underwent four stages: concentration polarization, nucleation, crystallization, and steady state (only SA and HA treatments reached this last stage). The gypsum scaling behavior on organic-fouled membrane could be controlled by the following impacts. It is well known that Ca2þ is able to bind to the carboxylic groups of organic molecules (Grant et al., 1973; Mo et al., 2008; Tipping, 2002). According to previous studies, the carboxylic acidities for BSA, HA, and SA were about 1, 3.4, and 3.5 meq g1, respectively (Ang and Elimelech, 2007; Hong and Elimelech, 1997), indicating that all organic foulants can bind Ca2þ with the ability in the order of SA > HA > BSA (Mi and Elimelech, 2008). Therefore, under gypsum scaling experimental conditions, Ca2þecarboxyl specific interaction could result in increased organic layer cross-linking and Ca2þ concentration on the membrane surface. Moreover, the accumulation of a highly organized organic fouling layer on the membrane surface hindered the transport of salts through the membrane and reduced back diffusion of salts, which led to increased CECP, further resulting in an elevated salt concentration at the membrane surface. In addition, the adhered organic foulants increased the negative charge of the membrane surface (see Table S3, Supporting Information), thus enhancing the Donnan effect between divalent ions and the organic-fouled membrane. Consequently, the concentrations of both Ca2þ and SO2 4 increased close to the organicfouled membrane surface, which led to supersaturation of gypsum near/at the membrane surface. Therefore, the formation of gypsum prenucleation clusters was initiated, and subsequently amorphous gypsum nanoparticles and polycrystals were generated on the organic-fouled membrane surface. However, the overall process performance probably results from synthetic influences; thus, further investigations are needed to reveal the underlying mechanisms.

Normalized flux after gypsum scaling

1.0

207

Gypsum scaling condition Fig. 5. Cleaning efficiencies of DI water on the organic-fouled and virgin membranes after gypsum scaling experiments. Cleaning conditions: crossflow rate of 15 cm s1; time of 20 min; temperature of 25 ± 1  C; and no applied hydraulic pressure.

208

J. Wang et al. / Water Research 91 (2016) 203e213

condition. In contrast, cleaning the gypsum scaling of BSA-, HA-, and SA-fouled membranes would recover the fluxes to 78%, 71%, and 64% of the original clean membrane flux, respectively. Different cleaning effects of organic-conditioned membranes implied that the fouling layer on each membrane surface was incompletely wiped off by physical cleaning. The higher flux recovery of virgin membrane compared with organic-conditioned membranes was likely because of weaker gypsumemembrane interactions than the interactions between organics and membrane, organics and organics, or between organic-associated gypsum scalants. In addition, the latter interactions were also possibly the reason for flux recovery differences between diverse organic-fouled membranes. These results indicate that not only was gypsum scaling enhanced but also the cleaning effect was weakened in organicfouled membrane conditions. This suggests that gypsum scaling of differently conditioned membranes might require diverse cleaning intensity and/or cleaning reagents. Therefore, further optimization of the cleaning process is necessary to reduce the duration, the amount of water spent, and/or the different cleaning reagents required for membrane cleaning. 3.1.4. Influence of organic macromolecules on gypsum scaling composition and morphology (SEM-EDS) EDS was conducted to determine the elemental composition of the obtained scaling layer. Carbon and sulfur were selected as typical elements of organic (BSA, HA, and SA) and inorganic (gypsum) substances in the gypsum scaling layer. It can be seen from Fig. 6, the gypsum scaling layer on either virgin or SA-fouled membrane contained very high sulfur and relatively low carbon concentrations. These results were attributed to the pure gypsum scaling layer on the virgin membrane and indicated that gypsum crystals dominated in the deposited layer on SA-fouled membrane. In contrast, the scaling layer on the BSA-fouled membrane had very low sulfur and the highest carbon content, demonstrating that organic matter was dominant in the deposited layer. The scaling layer on the HA-fouled membrane contained relatively low sulfur and high carbon concentrations. Accordingly, it is speculated that the surface topology and/or chemical characteristics of the organic layer would significantly impact the subsequent gypsum scaling behavior. To verify further that the gypsum scaling behaviors on diverse membrane surface conditions were dominated by different scaling mechanisms, static adsorption experiments (not pressure-driven)

20 Carbon

Concentration (%)

24

18

Sulfur

20

16

16

14

12

12

8 10

4

Concentration (%)

28

8

0 BSA

HA

SA

Virgin membrane

Gypsum scaling condition Fig. 6. EDS data of the gypsum scaling layers on the organic-fouled and virgin membranes, respectively. Carbon and sulfur are selected as representative organic and inorganic contents in the gypsum scaling layer.

were carried out to investigate surface crystallization on different conditioned membranes. SEM images demonstrate that only a few scattered plate-like gypsum crystals were found on the virgin membrane surface (Fig. 7(d)) and there were some rod-like gypsum crystals on the BSA-fouled membrane surface (Fig. 7(a)). In contrast, numerous needle-shaped gypsum crystals were formed on the HA-fouled membrane surface (Fig. 7(b)) and a large proportion of the SA-fouled membrane surface was covered by serried gypsum crystal clusters (Fig. 7(c)). The results demonstrate that SAand HA-fouled membranes had higher surface crystallization tendency than the BSA-fouled and virgin membranes, indicating that HA and SA macromolecules acted as nuclei for crystallization and thus markedly accelerating gypsum surface crystallization. A similar result was observed by Liu and Mi (2012), who found that alginate molecules might serve as nuclei for the formation of gypsum crystals during combined fouling by alginate and gypsum in the FO process. Based on EDS and SEM analyses, it is interesting to note that the surface crystallization tendency was closely correlated to the carboxylic content of the membrane surface, which means that the interaction between organic molecules/virgin membrane and Ca2þ could play an important role in crystal formation. This hypothesis needs to be examined on a much smaller scale with the aid of QCMD and AFM, as discussed below. 3.2. QCM-D measurements of calcium ions adsorbing onto organic adlayers As noted, the adsorption of different organic foulants onto the membrane surface is important as it is related to the initial fouling rate on a clean membrane, as well as the change in membrane surface properties. Moreover, the generated organic fouling adlayer impacts the Ca2þ concentration nearby and further inorganic scaling behavior. In this study, QCM-D was used to examine the hypothesis that Ca2þecarboxyl interaction played a considerable role in initiating gypsum scaling and quantitatively evaluate the role of Ca2þ adsorption on the structural characteristics of the adhesion layer. 3.2.1. Ca2þ adsorption capacity of different organic fouling layers The normalized DD associated with Df for the 5th overtone as a function of time during the overall QCM-D experiments is presented in Fig. 8(aec). The vertical lines indicate the injection times of ultrapure water (t0, t4, and t6), baseline solution (t1 and t3), organic foulant solution (t2), and CaCl2 solution (t5), respectively. The adsorption of both organic and inorganic substances at the crystal sensor surface consisted of reversible and irreversible adsorption as indicated by the changes in Df over time (Fig. 8(aec)). After removing the unstable adsorbed Ca2þ with ultrapure water and comparison with the stable stage before CaCl2 solution was introduced into the system, frequency decrease and dissipation increase were observed for the HA and SA conditions, indicating that calcium ions were bound onto the organic-coated sensor surface. Under BSA condition, neither frequency nor dissipation changed distinctly before or after Ca2þ was introduced, suggesting that little calcium ions were adsorbed by BSA molecules. 3.2.2. Effect of Ca2þ adsorption on organic adhesion layer structure Changes in the frequency versus dissipation shifts at different adsorption equilibrium stages were analyzed to investigate the impacts of Ca2þ on the organic adsorbed layer structure. This DD/Df analysis is commonly used to analyze structural information of the adsorbed layer (Contreras et al., 2011; Feiler et al., 2007). For the BSA condition, the DD/Df values were nearly unaltered and were the lowest among the three conditions both before and after Ca2þ

J. Wang et al. / Water Research 91 (2016) 203e213

209

Fig. 7. SEM images of the gypsum generation on (a) BSA-fouled, (b) HA-fouled, (c) SA-fouled, and (d) virgin membranes after soaking corresponding membrane specimens in supersaturated gypsum solutions, respectively. The membrane coupons were vertically soaked for 24 h in 38 mM NaCl, 40 mM Na2SO4, and 70 mM CaCl2, with a gypsum saturation index of 3.1. All images were taken at the same magnification.

Fig. 8. The normalized DD associated with Df for the 5th overtone as a function of time during the overall experiment at conditions (a) BSA, (b) HA, (c) SA. The vertical lines indicate the injection times of ultrapure water (t0 and t6), background electrolyte solution (t1, t3, and t5), organic foulant solution (t2), and CaCl2 solution (t4), respectively. (d) Calculated DD/ Df at adsorption equilibrium stages before and after Ca2þ introduced. Experimental conditions: background electrolyte solution (105 mM NaCl), organic foulant solution (1 g L1 organic matter, 105 mM NaCl), and CaCl2 solution (35 mM), pH of 7.5 ± 0.1, test temperature of 23  C, flow rate of 100 mL min1. (d) DD/Df values at adsorption equilibrium stage for different organic conditions before and after calcium ions adsorption. Note that horizontal and vertical scales of the images (aec) are not identical.

was introduced (Fig. 8(d)), indicating that the BSA adlayer was rigid and Ca2þ did not affect the structure of the adsorbed layer at

equilibrium. Both HA and SA adlayers exhibited much higher DD/Df values than the BSA layer, demonstrating that SA and HA adsorbed

210

J. Wang et al. / Water Research 91 (2016) 203e213

layers were looser and more elastic than the BSA layer. However, once Ca2þ was introduced, different responses were observed for the HA and SA conditions. The introduction of Ca2þ enlarged the DD/Df for the HA condition, indicating that the HA adlayer became much softer after the adsorption of Ca2þ. The increased softness was most likely caused by higher water content in the HA layer (Liu and Mi, 2014). In contrast, the DD/Df value for the SA condition decreased after the exposure to Ca2þ, implying that the interaction between calcium and SA molecules made the SA film more rigid than the initial adlayer. The phenomenon probably resulted from Ca2þealginate binding and calcium-induced alginate intermolecular bridging by forming a zigzag structure that eventually made the SA adlayer denser and more rigid (Lee and Elimelech, 2007). Similar results were found by Liu and Mi (2014) who investigated the adsorption behaviors of Ca2þ on organic conditioning layers by using silica wafers. 3.3. AFM adhesion force measurements AFM force measurement is widely used to evaluate the role of membraneefoulant and foulantefoulant interactions in membrane fouling processes (Guillen-Burrieza et al., 2013; Lee and Elimelech, 2006; Li and Elimelech, 2004; Mi and Elimelech, 2010a; Wang et al., 2013). To understand better the mechanisms underlying gypsum scaling on specific membrane surfaces, AFM in conjunction with the gypsum particle probe was utilized to characterize interfacial phenomena at the nanoscale during gypsum scale formation. During the initial stage of gypsum scaling, the interaction forces (i.e., gypsumemembrane interactions) between gypsum and the different conditioned surfaces (i.e., BSA-, HA-, SA-fouled, and virgin membrane) controlled the gypsum nuclei formation. After the membrane surface was covered by the inorganic foulants, gypsumemembrane interactions would be replaced by gypsumegypsum interactions, which governed the subsequent gypsum scaling rate. 3.3.1. Gypsumemembrane interactions Representative adhesion force curves of gypsumemembrane interactions are displayed in Fig. 9(a); the frequency distributions of the corresponding forces are shown in Fig. 9(b) to acquire the maximum distribution range of the obtained adhesion force values. The average gypsumemembrane adhesion forces of different conditions increased in the order of BSA (0.191 mN m1) < virgin membrane (0.245 mN m1) < HA (0.292 mN m1) < SA (0.349 mN m1). Compared with the BSA condition, the SA- and HA-adhesive membrane surfaces that had more plentiful carboxylic groups and higher zeta potential (Table S3) showed stronger interactions with the gypsum probe. Although HA and SA have similar carboxylic acidity (Mi and Elimelech, 2008) and the HA-fouled membrane surface had the highest zeta potential (i.e., the highest electrostatic attraction between gypsum), the adhesive forces of the HA condition were lower than those of the SA condition. This implied that the dominant effect was not electrostatic interaction but likely Ca2þecarboxyl chemical interplay, which was affected by HA and SA molecular properties along with the structural characteristics of the resulting fouling layers (Mi and Elimelech, 2008). It was clear that the trend in adhesion forces between gypsum and organic-fouled membranes corresponded to the calcium-binding ability of the organic layers (see the QCM-D measurements in Section 3.2). For poly(piperazineamide) NF membrane, the predominant functional groups were deprotonated carboxylic groups at pH 7.5 ± 0.1 that could bind Ca2þ. Moreover, the largest extension distance and stepwise disruption feature observed under the virgin membrane condition might result from the nonuniform membrane

surface morphology that affected pulling away the probe from the membrane surface. The results further indicate that the gypsum scaling potential on organic-fouled membrane surfaces is related to the calciumbinding capacity of the coated organic matters, which impacts Ca2þ concentration near the membrane surface, then gypsum nucleation, and the subsequent gypsum scaling progress.

3.3.2. Gypsumegypsum interactions Representative normalized adhesion force curves and frequency distributions of the corresponding forces of the gypsumegypsum type are presented in Fig. 9(c, d). The average gypsumegypsum adhesion forces of BSA-, HA-, SA-fouled, and virgin membranes were 0.356, 0.501, 0.603, and 0.411 mN m1, respectively (Fig. 9(b, d)). The extension distances of SA and HA conditions were larger than those of BSA-fouled and virgin membranes. The greater interaction forces and extension distances of the SA and HA conditions could be attributed to the more abundant heterogeneous surface crystallization, which is in agreement with the results of the static adsorption experiments. If the two series of interfacial force measurements and the results of the bench-scale gypsum scaling experiments are combined, it is interesting to find that the adhesion forces were remarkably correlated with flux declines for diverse organic-conditioned membranes during different crystallization stages. In other words, for the same gypsum scaling stage, the stronger the adhesion force, the more gypsum crystallized and the greater the decline in flux. In particular, both the interaction forces and zeta potential (see Table S3) of the virgin membrane condition were larger than those of the BSA condition; however, the holistic gypsum scaling rate relationship was the reverse. The opposite results were likely due to CECP caused by the BSA fouling layer in the pressure-driven filtration process. When comparing Fig. 9(a) and (c), it is clear that the sequence of gypsumegypsum interactions for different membrane surface conditions was in agreement with the gypsumemembrane result. Furthermore, the adhesion forces and extension distances of all conditions increased compared with those of the gypsumemembrane situation, which suggests that gypsum crystallization increased the number of interaction sites between the gypsum probe and the membrane surface, but did not change the magnitude of the adhesion forces. Consequently, eliminating the gypsumegypsum adhesion force is important for controlling gypsum scaling of membranes. Nevertheless, the role of the interaction between gypsum and membrane in gypsum scaling behavior is important given that after a certain amount of gypsum accumulation on the membrane surface, gypsumemembrane interactions will be displaced by gypsumegypsum interactions, which control the subsequent gypsum scaling behavior. This conclusion is analogous to membrane organic fouling characteristics found in previous studies (Li and Elimelech, 2004; Wang et al., 2013). The above outcomes demonstrate that the interaction force measured by the gypsum particle probe is a good indicator for predicting gypsum scaling behavior and revealing the dominant gypsum scaling mechanism of different conditioned NF membranes. A combination of the results of interaction force measurements and bench-scale experiments suggests that a pretreatment process that removes primarily SA-like foulants, then HA-like foulants, rather than other substances, could be significant for reducing the amount of such foulants in the organic fouling layer, and further mitigating gypsum scaling of poly(piperazineamide) NF membrane.

J. Wang et al. / Water Research 91 (2016) 203e213

211

Fig. 9. Representative normalized adhesive force curves of (a) gypsumemembrane type, and (b) the frequency distribution of the corresponding forces. The adhesion force measurements were conducted in a saturated gypsum solution, without gypsum precipitation. Representative normalized adhesive force curves of (c) gypsumegypsum type, and (d) the frequency distribution of the corresponding forces. The adhesion forces were measured with a supersaturated gypsum solution (SI ¼ 3.1), with gypsum precipitation taken place.

3.4. Mechanisms underlying gypsum scaling on organic-fouled membranes Gypsum scaling in membrane processes can be governed by one of two different mechanisms: (1) heterogeneous or surface crystallization, during which crystals grow directly on the membrane surface; and (2) homogeneous or bulk crystallization, where crystals are formed in the bulk solution and then deposit on the membrane surface (Lee et al., 1999; Lee and Lee, 2000). A previous study found that variation in membrane materials/surface properties might change the dominant gypsum scaling mechanism from homogeneous to heterogeneous crystallization (Mi and Elimelech, 2010a). Similarly, the organic fouling layer could affect gypsum scaling by modifying the membrane surface physicochemical properties. In the present study, flux decline was aggravated in all organic-fouled membranes compared with the virgin membrane. Organic macromolecules deposited on membranes can accelerate gypsum scaling by three possible mechanisms, namely size

exclusion, Donnan exclusion, and calcium bridging/complexation. The organic fouling layer performs as an “active membrane,” which acts as a barrier to salt transport as well as water molecules through the membrane, resulting in the CECP effect. Moreover, the negative charge of the membrane surface could be increased by the adhered organic foulants, thus enhancing the Donnan effect between divalent ions and the organic-fouled membrane. Organic macromolecules might behave as the nuclei for crystallization, thereby initiating heterogeneous crystallization around organic molecular nuclei, leading to subsequent crystal growth on the membrane surface. It is interesting to note that the induction time length and crystallization rate of gypsum scaling were coincident with the tendency of the Ca2þ-binding capacity of organic matter, which was confirmed by the QCM-D experiments. In addition, the strong adhesion forces and expanded extension distance caused by heterogeneous crystallizationdand likewise the fact that much faster water flux decline took place in surfaces rich in carboxylic functional groupsdpoint to a dominant scaling mechanism described

212

J. Wang et al. / Water Research 91 (2016) 203e213

as follows. The dominant gypsum scaling mechanisms of different organic-conditioned membranes mainly depend on organic foulant properties especially carboxyl density. That is, heterogeneous crystallization becomes the dominant mechanism of gypsum scaling on organic-fouled membranes with high carboxyl density (i.e., HA and SA). However, for membranes fouled with lower carboxyl density organic matters (i.e., BSA), the governing gypsum scaling mechanism likely depends on the synthesis of various factors, including carboxyl density, organic molecular structure, fouling layer structure, surface charge and roughness, and others. 4. Conclusion This study investigated the influence of organic fouling layer on the subsequent gypsum scaling behavior of NF membrane. The deposited organic fouling layer could change the behavior of gypsum scaling, as observed in bench-scale experiments as well as demonstrated by SEM-EDS measurements. Organic molecules probably acted as the nuclei for gypsum crystal growth, and resulted in a combined network of gypsum crystaleorganics. Moreover, a positive correlation between gypsum scaling and QCM-D measurements as well as intermolecular adhesion forces was observed, indicating that the carboxyl density of the membrane surface played an important role in determining the rate and extent of initiation of gypsum crystal growth and subsequent gypsum scaling behavior. Nevertheless, gypsum scaling on diverse organic-fouled NF membranes could be the result of a combination of interactions. The dominant mechanism of gypsum scaling on organic-fouled membrane controlled by various factors can vary for different foulants. Ca2þ-binding capacity of organics, organic fouling layer structure, and surface charge are the major factors governing the development of a gypsum scaling layer on the organic-fouled membrane surface, of which Ca2þ-binding capacity of organics plays a crucial role. Based on our experimental results, it is suggested that SA-like and HA-like foulants, rather than BSA-like substances, should be the target organic compounds for removal from feedwater. Furthermore, QCM-D measurements and the AFM adhesion force test proved to be valuable techniques that could provide useful information to optimize process technologies including membrane modification, feedwater pretreatment, optimization of module arrangement and process conditions, and periodic membrane cleaning, which could mitigate or control gypsum scaling in the NF process and extend the lifetime of the NF membrane. Acknowledgments Financial support for this study was provided by the National Natural Science Foundation of China (Grant No. 51178378, No. 51278408), the Shaanxi Province Science and Technology Innovation Projects (Grant No. 2012KTCL03-06, No. 2013KTCL03-16), the Innovative Research Team of Xi'an University of Architecture and Technology. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.watres.2016.01.019. References Abramson, H.A., 1933. Electrokinetic phenomena and their application to biology and medicine. J. Phys. Chem. 38 (8), 1128e1129. Ang, W.S., Elimelech, M., 2007. Protein (BSA) fouling of reverse osmosis membranes: implications for wastewater reclamation. J. Membr. Sci. 296 (1e2), 83e92.

Barcelona, M.J., Atwood, D.K., 1978. Gypsum-organic interactions in natural seawater: effect of organics on precipitation kinetics and crystal morphology. Mar. Chem. 6 (2), 99e115. Contreras, A.E., Steiner, Z., Miao, J., Kasher, R., Li, Q., 2011. Studying the role of common membrane surface functionalities on adsorption and cleaning of organic foulants using QCM-D. Environ. Sci. Technol. 45 (15), 6309e6315. Dydo, P., Turek, M., Ciba, J., 2003. Scaling analysis of nanofiltration systems fed with saturated calcium sulfate solutions in the presence of carbonate ions. Desalination 159 (3), 245e251. Feiler, A.A., Sahlholm, A., Sandberg, T., Caldwell, K.D., 2007. Adsorption and viscoelastic properties of fractionated mucin (BSM) and bovine serum albumin (BSA) studied with quartz crystal microbalance (QCM-D). J. Colloid Interface Sci. 315 (2), 475e481. Grant, G.T., Morris, E.R., Rees, D.A., Smith, P.J.C., Thom, D., 1973. Biological interactions between polysaccharides and divalent cations: the egg-box model. FEBS Lett. 32 (1), 195e198. Guillen-Burrieza, E., Thomas, R., Mansoor, B., Johnson, D., Hilal, N., Arafat, H., 2013. Effect of dry-out on the fouling of PVDF and PTFE membranes under conditions simulating intermittent seawater membrane distillation (SWMD). J. Membr. Sci. 438, 126e139. Hamdona, S.K., Al Hadad, U.A., 2007. Crystallization of calcium sulfate dihydrate in the presence of some metal ions. J. Cryst. Growth 299 (1), 146e151. Her, N., Amy, G., Jarusutthirak, C., 2000. Seasonal variations of nanofiltration (NF) foulants: identification and control. Desalination 132 (1e3), 143e160. Hoek, E.M.V., Elimelech, M., 2003. Cake-enhanced concentration polarization: a new fouling mechanism for salt-rejecting membranes. Environ. Sci. Technol. 37 (24), 5581e5588. Hong, S., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Membr. Sci. 132 (2), 159e181. Hoang, T.A., Ang, H.M., Rohl, A.L., 2007. Effects of temperature on the scaling of calcium sulphate in pipes. Powder Technol. 179 (1e2), 31e37. Huiting, H., Kappelhof, J.W.N.M., Bosklopper, T.G.J., 2001. Operation of NF/RO plants: from reactive to proactive. Desalination 139 (1e3), 183e189. Le Gouellec, Y.A., Elimelech, M., 2002. Calcium sulfate (gypsum) scaling in nanofiltration of agricultural drainage water. J. Membr. Sci. 205 (1e2), 279e291. Lee, S., Boo, C., Elimelech, M., Hong, S., 2010. Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO). J. Membr. Sci. 365 (1e2), 34e39. Lee, S., Cho, J., Elimelech, M., 2005. Combined influence of natural organic matter (NOM) and colloidal particles on nanofiltration membrane fouling. J. Membr. Sci. 262 (1e2), 27e41. Lee, S., Choi, J.-S., Lee, C.-H., 2009. Behaviors of dissolved organic matter in membrane desalination. Desalination 238 (1e3), 109e116. Lee, S., Elimelech, M., 2006. Relating organic fouling of reverse osmosis membranes to intermolecular adhesion forces. Environ. Sci. Technol. 40 (3), 980e987. Lee, S., Elimelech, M., 2007. Salt cleaning of organic-fouled reverse osmosis membranes. Water Res. 41 (5), 1134e1142. Lee, S., Kim, J., Lee, C.-H., 1999. Analysis of CaSO4 scale formation mechanism in various nanofiltration modules. J. Membr. Sci. 163 (1), 63e74. Lee, S., Lee, C.-H., 2000. Effect of operating conditions on CaSO4 scale formation mechanism in nanofiltration for water softening. Water Res. 34 (15), 3854e3866. Li, Q., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environ. Sci. Technol. 38 (17), 4683e4693. Lin, C.-J., Shirazi, S., Rao, P., 2005. Mechanistic model for CaSO4 fouling on nanofiltration membrane. J. Environ. Eng. 131 (10), 1387e1392. Liu, Y., Mi, B., 2012. Combined fouling of forward osmosis membranes: synergistic foulant interaction and direct observation of fouling layer formation. J. Membr. Sci. 407e408, 136e144. Liu, Y., Mi, B., 2014. Effects of organic macromolecular conditioning on gypsum scaling of forward osmosis membranes. J. Membr. Sci. 450, 153e161. Mi, B., Elimelech, M., 2008. Chemical and physical aspects of organic fouling of forward osmosis membranes. J. Membr. Sci. 320 (1e2), 292e302. Mi, B., Elimelech, M., 2010a. Gypsum scaling and cleaning in forward osmosis: measurements and mechanisms. Environ. Sci. Technol. 44 (6), 2022e2028. Mi, B., Elimelech, M., 2010b. Organic fouling of forward osmosis membranes: fouling reversibility and cleaning without chemical reagents. J. Membr. Sci. 348 (1e2), 337e345. Mo, H., Tay, K.G., Ng, H.Y., 2008. Fouling of reverse osmosis membrane by protein (BSA): effects of pH, calcium, magnesium, ionic strength and temperature. J. Membr. Sci. 315 (1e2), 28e35. Mohammad, A.W., Teow, Y.H., Ang, W.L., Chung, Y.T., Oatley-Radcliffe, D.L., Hilal, N., 2015. Nanofiltration membranes review: recent advances and future prospects. Desalination 356, 226e254. Rahman, F., 2013. Calcium sulfate precipitation studies with scale inhibitors for reverse osmosis desalination. Desalination 319, 79e84. €m, M., Sch€ afer, A., Andritsos, N., Karabelas, A.J., Hoek, E.M.V., Schneider, R., Nystro 2005. Chapter 8: fouling in nanofiltration. In: Schaefer, A., Fane, A.G., Waite, T.D. (Eds.), Nanofiltration: Principles and Applications. Elsevier, Oxford, UK, pp. 169e239. Seidel, A., Elimelech, M., 2002. Coupling between chemical and physical interactions in natural organic matter (NOM) fouling of nanofiltration membranes: implications for fouling control. J. Membr. Sci. 203 (1e2), 245e255.

J. Wang et al. / Water Research 91 (2016) 203e213 Sheikholeslami, R., 2003. Mixed saltsdscaling limits and propensity. Desalination 154 (2), 117e127. Steiner, Z., Miao, J., Kasher, R., 2011. Development of an oligoamide coating as a surface mimetic for aromatic polyamide films used in reverse osmosis membranes. Chem. Commun. 47 (8), 2384e2386. Subir Bhattacharjee, G.M.J., 2002. A model of membrane fouling by salt precipitation from multicomponent ionic mixtures in crossflow nanofiltration. Environ. Eng. Sci. 19 (6), 399e412. Thorsen, T., 2004. Concentration polarisation by natural organic matter (NOM) in NF and UF. J. Membr. Sci. 233 (1e2), 79e91. Tipping, E., 2002. Cation Binding by Humic Substances. Cambridge University Press, Cambridge.

213

Van Der Bruggen, B., Vandecasteele, C., Van Gestel, T., Doyen, W., Leysen, R., 2003. A review of pressure-driven membrane processes in wastewater treatment and drinking water production. Environ. Prog. 22 (1), 46e56. Wang, L., Miao, R., Wang, X., Lv, Y., Meng, X., Yang, Y., Huang, D., Feng, L., Liu, Z., Ju, K., 2013. Fouling behavior of typical organic foulants in polyvinylidene fluoride ultrafiltration membranes: characterization from microforces. Environ. Sci. Technol. 47 (8), 3708e3714. Wiesner, I.K.a.M.R., 2007. Morphological variations of precipitated salts on NF and RO membranes. Environ. Eng. Sci. 24 (5), 602e614. Yang, H.L., Huang, C., Pan, J.R., 2008. Characteristics of RO foulants in a brackish water desalination plant. Desalination 220 (1e3), 353e358.