Separation and Purification Technology 80 (2011) 246–261 Contents lists available at ScienceDirect Separation and Purification Technology journal homepage: www.elsevier.com/locate/seppur A systematic study on triazine retention by fouled with humic substances NF/ULPRO membranes Konstantinos V. Plakas, Anastasios J. Karabelas ⇑ Chemical Process Engineering Research Institute, Centre for Research and Technology – Hellas, P.O. Box 60361, 6th km Charilaou-Thermi Road, Thermi, Thessaloniki, GR 570-01, Greece a r t i c l e i n f o Article history: Received 15 February 2011 Received in revised form 3 May 2011 Accepted 4 May 2011 Available online 17 May 2011 Keywords: Water membrane nanofiltration Triazine herbicides retention Organic fouling Cake-layer resistance a b s t r a c t Naturally occurring organic compounds tend to form complexes with divalent cations and micropollutants, and to foul membrane surfaces; both phenomena have a significant effect on pollutant rejection. Previous study results show the significant influence of triazine herbicides complexation, with dissolved humic substances, on their rejection by NF/ULPRO membranes. The net effect of fouled membranes, on triazine retention, is systematically investigated herewith by comparing the performance of three types of clean and fouled membranes and relating them to changes observed in membrane surface characteristics. Two typical triazines (atrazine and prometryn) and three well characterized humic substances (HS) are employed. The results show that humic substances deposited on the membrane surfaces cause considerable changes of their characteristics, including the contact angle and salt retention, which affect water permeability and triazine retention. Specifically, a strong correlation is identified between the hydrophobicity/ hydrophilicity of dissolved HS and the resistance to flow of the fouling layer, which affects the retention of the smaller-size triazines. This trend is related to the condition of organic layers on the membrane. Generally, relatively loose fouling layers on the membranes are associated with reduced triazine retention. However, rather dense fouling layers formed by complexes of HS with calcium exhibit significant flux decline and an improved sieving effect on triazines. Moreover, tight and hydrophobic membranes display significant changes of triazine retention when fouled by HS of increased hydrophobicity. On the contrary, porous and hydrophilic membranes display significant changes of triazine retention only when fouled by HS of reduced hydrophobicity. The new results highlight the need for good knowledge of the properties characterizing the organic matter present in natural waters, before their treatment process is designed. Ó 2011 Elsevier B.V. All rights reserved. 1. Introduction The increasingly complex cocktail of anthropogenic organic chemicals detected at very small concentrations (i.e. at the level of ppb) in water bodies, and consequently in drinking water supplies, constitutes a major threat to humans, since many of these chemicals and their residues (e.g. hormones, polycyclic aromatic hydrocarbons and organochlorine pesticides) are toxic with known adverse health outcomes. The need for essentially complete removal of these compounds from water has increased the interest in membrane processes, such as nanofiltration (NF) and ultra-low pressure reverse osmosis (ULPRO), which are considered efficient and cost effective treatment methods for potable water production of high and stable quality. A significant number of studies have been carried out on the retention characteristics of pesticide active ingredients, pharmaceutically active compounds (PhACs), polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), endocrine disrupting compounds ⇑ Corresponding author. E-mail address: karabaj@cperi.certh.gr (A.J. Karabelas). 1383-5866/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.seppur.2011.05.003 (EDCs) and for many other toxic organic substances, which have improved our understanding of the prevailing mechanisms and factors affecting membrane retention efficiency [1–10]. The majority of the related experimental research so far is focused on the removal of organic residues of crop protection products (pesticides) and their metabolites due to their widespread application in agriculture and their significant toxicity which has led authorities to take measures to protect human health [11]. Laboratory research, pilot and industrial-scale activity show that the retention of pesticides residues varies from very good, by some membranes (mainly tight NF and RO membranes), to moderate or low removal by others (loose NF membranes). Recent reviews indicate the importance of solute–solute and solute–membrane interactions in membrane performance, which in turn is influenced by solute parameters (molecular size, charge and polarity), membrane properties (pore size, surface charge and hydrophobicity/hydrophilicity), feed water composition and process conditions [12,13]. The significant number of parameters affecting pesticide retention is characteristic of the complex interactions taking place, which can be further influenced by the changes occurring in membrane surface properties as a result of fouling. Indeed, the retention 247 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 of organic micropollutants (EDCs, PhACs, pesticides, etc.) is determined by electrostatic, steric and hydrophobic/hydrophilic solute– membrane interactions, which can be modified due to foulants depositing on the membrane surface. The effect of fouling on organic micropollutant retention has been the subject of extensive research in the past decade [14–23]. Investigations on the influence of colloidal and/or organic fouling on retention of various trace organics suggest that two cases may be distinguished, depending on the relative solute selectivities of the fouling layer and the membrane. First, if the membrane rejects solutes better than the deposited layer, hindered back diffusion of solutes (by the fouling layer) would cause solute accumulation near the membrane surface. This enhanced concentration polarization results in greater concentration gradient across the membrane and, hence, an increase of solute permeation. Second, if solutes are rejected better by the deposited layer than the membrane, the fouling layer controls solute retention which tends to improve. For example, the severe decline observed [14] in the retention of two steroid hormones due to accumulation of colloidal silica particles at the RO membrane surface was attributed to the increased hormone adsorption onto the membrane polymeric skin layer as a result of the cake enhanced concentration polarization, which in turn facilitated their diffusion across the skin layer to the permeate side. The effect of cake-enhanced concentration polarization on the retention of 21 PhACs was also recently invoked by Verliefde et al. [22] to explain the observed retention variations depending on type of fouling. Specifically, filtration of surface water, which was first pretreated with anionic exchange resins, resulted in the deposition of a mainly colloidal fouling layer which led to increased flux decline and to significant changes in pharmaceuticals retention. On the other hand, retention by membranes fouled by natural organic matter (surface water pretreated with ultrafiltration) or by untreated surface water exhibited smaller variation as a result of steric and electrostatic effects [22]. The study by Nghiem and Hawkes [19] highlighted the significance of NF membrane pore size on fouling and indirectly on the retention of low MW pharmaceuticals, which was mainly attributed to pore restriction and cake enhanced concentration polarization phenomena. In particular, the retention behavior of a very loose NF membrane was apparently dominated by pore restriction type of fouling, while there was evidence of the cake enhanced concentration polarization effect with the smaller pore size NF membranes [19]. Experiments with NF membranes [18], pre-fouled by activated sludge and landfill leachate, showed a marked reduction of retention of the larger contaminants, and an increased retention of smaller MW contaminants (36 neutral trace plastic additives, EDCs and other organics tested). Fouling experiments using a microfiltered secondary effluent [15] resulted in an increased adsorption capacity and reduced mass transport through partitioning and diffusion in NF/RO membranes for several organic compounds, representative of emerging organic contaminants. Interestingly, membrane fouling by a mixture of polysaccharides, silicate colloids and organic sulfonic acids (i.e. the major foulants in the microfiltered secondary effluent) had little effect on retention by thin film composite RO membranes [15]. In a recent work by Yangali-Quintanilla et al. [23], membrane fouling by a hydrophilic anionic polysaccharide (sodium alginate) led to different retention behavior, depending on the NF membrane used. Specifically, there was a significant reduction of retention for all 14 organic micropollutants used (including atrazine) when treated with a fouled NF200 membrane (a hydrophilic one); on the contrary, a fouled NF90 membrane (a hydrophobic one) exhibited negligible or increased retention improvement, depending on the hydrophobic/hydrophilic character of the organic micropollutants [23]. Similar observations were made by Bellona et al. [20]. From the above studies it is evident that membrane fouling may significantly affect the retention of low MW organic compounds depending mainly on foulants characteristics, membrane properties, and chemical composition of feed water. It is well known that humic substances (HS) make up more than 50% of the NOM in surface waters and are considered a major cause of NF fouling [24–27]. Moreover, low MW humic substances may still be present in UF pretreated feed water [28], which can cause organic fouling of the NF/ULPRO membranes used for the retention of micropollutants. In preliminary work [16], the effect of fouling (by humic and fulvic acids) on pesticides retention by a relatively porous NF membrane (NF270) was evident. The observed significant change of membrane performance due to organic fouling pointed to the need for a more detailed study of this topic. The same conclusion was reached in a recent study [29], where the capacity of triazines for complexation with dissolved humic substances could not fully account for the observed triazine retention during such solution filtration through NF/ULPRO membranes. To clarify the net effect of organic fouling on both triazine and salt retention, and to obtain an improved understanding of the role of humic substances on triazine rejection by NF/ULPRO membranes, this systematic study was undertaken, involving a multi-step filtration protocol, three well characterized humic substances (IHSS standards), two triazine herbicides (atrazine and prometryn) and three types of NF/ULPRO membranes. In the following, experimental conditions are described first; experimental results and discussion are presented next. 2. Experimental work 2.1. Materials and methods 2.1.1. Membranes and their characterization Three flat sheet type commercial membranes (Dow, Filmtec) denoted as NF90, NF270 and XLE were used in this study. NF90 and NF270 are nanofiltration membranes while XLE is described by the manufacturer as extra low energy membrane (ULPRO). Their characteristic properties are summarized in Table 1. For the experiments, flat sheet membrane specimens with an active membrane area of 14.6 cm2 were employed. Before use, all membrane specimens were rinsed with tap water for several minutes and stored at 4 °C in a 0.75% Na2S2O5 aqueous solution (to suppress development of micro-organisms), which was regularly replaced. Preliminary filtration tests with ultra-pure water and separate solutions of three different salts (2 mM NaCl or CaCl2 or Na2SO4, at 5 bar) showed that the rather porous and hydrophilic NF270 membrane exhibits the highest permeability among the three membranes tested, as well as fairly high retention of charged ions. On the other hand, the rather hydrophobic NF90 and XLE membranes exhibit Table 1 Characteristics of membranes used in this study; manufacturer Dow (Filmtec). MWCOa (Da) Membrane pore sizeb (nm) Contact anglec (°) Zeta potentiald (mV) NaCl retentione (%) CaCl2 retentione (%) Na2SO4 retentione (%) Specific fluxf (L m2 h1 bar1) a NF270 NF90 XLE 200 0.71 ± 0.14 28 ± 2 21.6 66.4 77.1 96.5 11.5 ± 0.3 200 0.55 ± 0.13 (0.82) 62 ± 2 24.9 99.5 98.8 99.9 5.8 ± 0.3 <100 (0.67) 65 ± 2 3.2 95.9 97 98.1 5.7 ± 0.4 Molecular weight cut off, as reported by the manufacturer. Values obtained from Ref. [30]. Values in the brackets obtained from Ref. [31]. c Sessile drop contact angle measurements (G10 Krüss). d Measured at pH 7 and 30 lS/cm KCl solution (PAAR EKA-Electro Kinetic Analyzer RV. 4.0). e Filtration tests with 2 mM salt concentration (5 bar). f Determined at 5 bar (filtration of Milli-Q water at 25 °C). b 248 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 similar permeability values and almost complete retention of charged ions. The relatively better desalting performance of NF90 membrane, in contrast to XLE, is attributed to its greater negative surface charge (measured at neutral pH), which does not greatly differ from the one characterizing the NF270 membrane (Table 1). Membrane specimens were dried after the experiments at room temperature for at least 24 h prior to surface analysis. Membrane hydrophobicity was characterized by sessile drop contact angle measurements with ultra-pure water, using contact angle measuring equipment (Krüss G10, Hamburg). To account for differences of surface morphology at least five readings were taken at different positions across the samples. A JEOL scanning electron microscope (JSM-6300) with an ISIS energy dispersive X-ray spectrometer (ISIS-EDS, Oxford Instruments) were used to examine the morphology and the elemental composition of the fouled membranes, respectively. 2.1.2. Triazines Two herbicides belonging in the triazine family, i.e. atrazine and prometryn, were selected for this study. Both herbicides were of analytical grade, purchased from Riedel de-Haën (Sigma–Aldrich): atrazine (purity 97.4%), prometryn (purity 99.2%). Information on the chemical structures and the physicochemical properties of the triazines are included in Table 2. Both triazines are considered as non-ionic, hydrophobic (log Kow >2) compounds that are moderately soluble in water (weak polar compounds). Between the two, prometryn is the largest molecule due to its branched structure, whereas its greater pKa value indicates a greater basicity in comparison to atrazine. Concentrations of the two triazines in feed, permeate and concentration samples were determined using an Agilent Technologies Model 7890A gas chromatograph system interfaced with an Agilent Technologies Model 5975C mass-selective detector (MSD). Pre-concentration of the water samples and transfer to the organic phase was based on off-line solid phase extraction (SPE) which took place prior to the chromatographic analysis. Simazine (Riedel de-Haën, purity 99.9%), a triazine compound similar to atrazine and prometryn, was used as internal standard in order to assess the recovery of the two triazines during the SPE-GC-MSD analyses. Recoveries from pure triazine solutions varied between 82% and 112%, while the limit of quantification (LOQ) was 1 lg/L for both triazine compounds, i.e. well above the lower concentrations expected in the permeate samples in the case of 10 lg/L feed concentrations and a SPE concentration factor equal to 25 (from 50 mL to a final volume of 2 mL). The chromatographic conditions and the SPE protocol followed are described in detail in previous publications [16,29]. 2.1.3. Humic substances Three types of water born humic substances (HS), purchased from the International Humic Substances Society (IHSS, University of Minnesota, SWC Dept.), were used in this investigation, which are denoted as Suwannee River humic acid (HA), fulvic acid (FA) and natural organic matter (NOM). They are well characterized reference materials of known origin, widely used in research. The IHSS humic and fulvic acids contain mainly hydrophobic organic acids, while the NOM sample contains not only hydrophobic and hydrophilic acids but also other soluble substances that are present in natural waters [35]. The characteristic chemical parameters of the three humic substances (HS) employed are summarized in Table 3, while the average molecular masses reported in the literature are presented in Table 4. The distribution of carbon among different functional groups of the three HS is shown in Table 5. Humic acids (HA) have the highest MW among the three HS, followed by NOM and FA. However, the carboxyl content of HS seems to be inversely related to the MW, since the relatively smaller FA exhibits higher carboxylic and total acidity, and therefore greater charge. UV absorbance at particular wavelengths shows that HA is more aromatic than NOM, and NOM slightly more aromatic than FA, which is in line with the 13C NMR estimates of carbon distribution in the three IHSS samples (Table 5). Measurement of the HS concentration (more specifically of the aromaticity) was performed in a UV–Vis spectrophotometer (UV1700, Shimadzu, Japan) at 254 nm in the case of Suwannee River HA and NOM, and at 275 nm in the case of FA. Quality control of the data was achieved by means of calibration and repeated analyses according to the procedure described in a previous publication [29]. 2.1.4. Feed solutions According to the filtration protocol, three different feed solutions were prepared; i.e. solutions of NaCl in ultra-pure water with concentration 2 mM, solutions of atrazine–prometryn in ultra-pure water with concentration 10 lg/L for each triazine, and foulant solutions of one humic substance in the presence or not of calcium Table 2 Properties of the herbicides used in this work [32]. Herbicide Atrazine Prometryn C8H14ClN5 Cl-Triazine 215.69 0.788 2.68 33 2.460 1.7 C10H19N5S S-Triazine 241.35  3.08 33  4.09 Chemical structure Molecular formula Chemical class Molecular weight (Da) Molecular sizea (nm) Log Kow Aqueous solubility (mg/L) Dipole momentb (debye) pKa (20 °C) a b Obtained from Ref. [33]. Obtained from Ref. [34]. 249 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Table 3 Acidicity and elemental composition of HS used in this study [35]. Humic substance Cat. no HA FA NOM a b Acidity (meq/g C) IHSS 2S101H IHSS 2S101F IHSS 1R101N Elemental composition Carboxylic Phenolic Total Ca Sa H2Ob 9.13 11.87 9.85 3.72 2.84 3.94 12.85 14.01 13.79 52.63 52.34 48.80 0.54 0.46 0.02 20.4 16.9 8.15 Elemental composition in %(w/w) of a dry, ash-free sample. %(w/w) of H2O in the air-equilibrated sample (a function of relative humidity). Table 4 Molecular mass data of IHSS organics employed in this study. Humic substance Technique [ref.]  na M  wb M  wc M n M Suwannee River HA FFF [36] Ultracentrifugation [37] Suwannee River FA FFF [36] Ultracentrifugation [37] HPSEC [38] 1580  1150  1760 4390 4260 ± 280 1910 1460 ± 80 2360 2.78  1.66  1.34 Suwannee River NOM a b c Number average molecular weight. Weight average molecular weight. Polydispersivity index; Suwannee River HA has a much larger polydispersivity compare to FA. Table 5 13 C NMR estimates of carbon distribution in IHSS samples used in this study [35]. Humic substance Carbonyl 220–190 ppm Carboxyl 190–165 ppm Aromatic 165–110 ppm Acetal 110–90 ppm Heteroaliphatic 90–60 ppm Aliphatic 60–0 ppm HA, 2S101H FA, 2S101F NOM, 1R101N 6 5 8 15 17 20 31 22 23 7 6 7 13 16 15 29 35 27 ions. In the latter case, model solutions of the three selected humic substances were prepared with ultra-pure water, of concentration 10 or 20 mg/L, at neutral pH. The HS were obtained in powder form and used without further purification as the bound iron and ash contents were very low. The calcium ion content in the corresponding foulant solutions was fixed at 40 mg/L (1 mM) by adding calcium chloride (J.T. Baker). Standard stock solutions of the two triazines (of concentration 100 mg/L) were prepared in high-performance liquid chromatography grade methanol and stored at 4 °C. The feed herbicide solutions were prepared from ultra-pure water by dilution of stock solution at a level of 10 lg/L for each triazine. Finally, all feed solutions were neutralized (pH 7.0) prior to their introduction in the stirred cells. The ultra-pure water used (resistivity >18 MXcm) for solution preparation was obtained from a Milli-Q purification system (Millipore, Milford, MA, USA). 2.2. Experimental set-up and filtration protocol Tests were conducted in the batch mode (dead-end experiments) using an experimental set-up, comprised of three high pressure filtration cells (volume of 300 mL each) made of stainless steel which does not allow adsorption of the two herbicides [7]. The three test cells were connected to a nitrogen cylinder to impose/control a constant filtration pressure and were operated simultaneously to assess the reproducibility and accuracy of the results. For mixing and minimization of concentration polarization phenomena, the test cells were equipped with Teflon-coated magnetic stirring elements, rotating at a rate of 250 rpm (measured with an Extech Instruments Model 461893 digital photo tachometer). Electronic balances connected to PCs were used to monitor permeate fluxes. The temperature of water in the test cells was kept constant at 25 ± 0.2 °C by a water cooling system. A new membrane specimen was used in each filtration test. A filtration protocol was designed to determine the impact of membrane fouling by humic substances on triazine and salt retention, as schematically shown in Fig. 1. First, virgin membrane specimens are placed in the stirred cells and rinsed with Milli-Q water for 1 h at a transmembrane pressure of 10 bar to remove the preservatives and to compact the membranes (step a). The initial pure water flux is measured by filtering Milli-Q water through the membrane at 5 bar for 30 min (step b). The next two filtration steps (c and d) involve retention experiments of NaCl and triazines by the clean membranes at 5 bar. The purpose of using separate feed solutions of NaCl and triazines is twofold; i.e. to avoid the effect of the increased ionic strength on triazine retention, as previously observed [7], and to permit a comparison between the retention values measured for salts and triazines with virgin and fouled membranes. Afterwards, the fouling procedure takes place, in which the foulant solution is filtered in three sequential cycles (e–g). Specifically, a 200 mL foulant solution is filtered at 5 bar until 100 mL of permeate are collected (50% recovery). After the first and second filtration, the permeate is returned to the stirred cell, so that fluid is filtered three times. Then, the final permeate and concentrate samples are collected to determine overall retention. This permeate recycle procedure was adopted from Schäfer [39] and was considered appropriate for fouling studies in the case of batch-type filtration experiments, to achieve stabilization of the organic deposits on the membrane surface. Afterwards, the retention efficiency of fouled membranes is tested by filtering triazine and NaCl solutions in two separate cycles (h and i), at 5 bar. Finally (step j), pure water flux was measured at 5 bar for 30 min to determine the final flux decline after the termination of all filtration cycles, for comparison with the initial pure water flux (step b). 250 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 2.3. Calculation of main parameters Experimental results are expressed in terms of the retention by the membrane of the organic and inorganic compounds, which was calculated according to the method described in a previous study [7]. In the present work, the retention of the two triazines is determined from the so-called ‘‘stable’’ permeate sample (j = 2) which corresponds to 50% water recovery. The amount of the two triazines and of the humic substances adsorbed on the membrane surface (MHS) is determined from mass balances as follows: j P Adsorbed organics ¼ 1  i¼1 V pi C pi þ C r V r Cf V f ð1Þ where C pi , Cr, Cf are the concentrations of specific organic species in the permeate sample i, retentate and feed, respectively, whereas V pi , V r , V f are the respective volumes of permeate sample i, retentate and feed. In the case of the humic substances, Cp, Cr, Vp and Vr correspond to the final permeate and retentate samples, collected after the termination of the fouling phase (third filtration cycle; step g, Fig. 1). The extent of membrane fouling is described by three parameters: percent flux reduction due to fouling (FR), mass of deposited humic substances on the membrane surface (MHS) and the average specific resistance (a) of the fouling layer [40]. FR is determined as a percentage reduction of water flux right before (JWb) and after (JWa) the fouling phase (i.e. between steps d and h in Fig. 1): 120 a 100 80 Flux (L/m2h) The evolution of flux as a function of filtration time, for all membranes and materials used, followed a similar pattern with the one presented in Fig. 1. More specifically, the differences observed between the initial pure water flux (straight line in Fig. 1) and the fluxes in each filtration step can be interpreted as follows: Step c: NaCl-solution filtration; some flux reduction due to the increased osmotic pressure of the rejected ions (concentration polarization) and/or due to ion adsorption to membranes. Step d: Herbicide-solution filtration; relatively stable water flux attributed to the negligible osmotic pressure of the herbicide solution. Step e: Flux decline mainly due to fouling in this first filtration with foulants. Steps f, g: The similar fluxes measured during the second and the third filtration cycles indicate that fouling has reached a pseudo steady-state condition. Step h: The flux reduction (compared to step d) during filtration of the herbicide solution is attributed to the humic layer existing on the membrane surface. Step i: Similar flux with that during herbicide filtration (step h). The limited flux decline is attributed to the small solution osmotic pressure. Step j: Flux decline after final pure water filtration. All filtration steps were performed with a feed solution of 200 mL; the tests performed with the triazine or the sodium chloride solutions were carried out until 100 mL of permeate were collected (50% recovery). In the case of triazines, a feed sample, two permeate samples (50 mL each) and the final concentrate were retained for chromatographic analysis. As in previous studies [7,29], the triazine feed solutions were initially stirred in the cells for 1 h without pressure. In this way, triazine adsorption on the membrane surface was considered to reach equilibrium, as shown by preliminary adsorption experiments (data not presented here). [After termination of the filtration experiments, the membrane specimens were dried for subsequent analysis, while the test cells were thoroughly washed with acetone and repeatedly rinsed with Milli-Q water.] b 60 d c 40 e f g 400 500 j i h 20 0 0 100 200 300 600 700 800 900 Filtration time (min) Fig. 1. Flux as a function of filtration time according to the filtration protocol implemented in the present study; experiment with NF270 membrane: (a) membrane conditioning at 10 bar, (b) initial pure water flux measurement at 5 bar, (c) filtration of 2 mM NaCl solution at 5 bar, (d) filtration of atrazine– prometryn solution (10 lg/L each) at 5 bar, (e) first filtration cycle of the foulant solution (10 mg/L HA + 1 mM Ca2+) at 5 bar, (f) second filtration cycle of the same foulant solution at 5 bar, (g) third filtration cycle of the same foulant solution at 5 bar, (h) filtration of atrazine–prometryn solution (10 lg/L each) at 5 bar, (i) filtration of 2 mM NaCl solution at 5 bar, (j) final pure water flux measurement at 5 bar. FR ¼ J Wb  J Wa  100 J Wb ð2Þ The average specific resistance, a, of the fouling layer was determined as follows. The fouling behavior during membrane filtration can be described by the common resistance-in-series model: J¼ 1 dV DP  Dpm ¼ A dt gðRm þ Rf Þ ð3Þ where J is the permeate flux, A the active membrane area, ðDP  Dpm Þ the effective transmembrane pressure, g the permeate viscosity, Rm the clean membrane resistance and Rf the total foulant resistance. In the case of calcium free solutions, the osmotic pressure difference, Dpm, is practically zero due to the negligible osmotic pressure caused by the selected humic substances. In the case of NF/ULPRO filtration of mixed HS/Ca solutions, the effective transmembrane pressure (DH-Dpm) is determined by estimating the osmotic pressure difference, including concentration polarization effects (see Appendix); the estimated osmotic pressure difference is also negligible (of order 105 bar). Therefore, Eq. (3) can be simplified for the present experiments as: J¼ DP gðRm þ Rf Þ ð4Þ Under constant pressure, the fouling layer resistance, Rf, is related to an average specific resistance a [40] through the expression Rf ¼ aC b;HS V A ð5Þ where Cb,HS is the bulk humic substance concentration. The value of specific resistance a is calculated by interpreting the fouling data in t/V vs V plots, where the slope of the linear region of the filtration curve, k, is characteristic of a foulant layer formed on the membrane surface, under constant pressure (DH) [40], i.e. k¼ gaC b;HS 2DPA2 ð6Þ It should be pointed out that the parameters FR, MHS and a are used in the present work to interpret triazine retention in relation to membrane fouling. 251 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 3. Results and discussion 40 Fig. 2 shows typical filtration data (plotted in terms of t/V vs V), obtained in repetitive filtration of a HA/Ca2+ solution through the NF90 membrane. Similar plots were obtained for all three membranes and dissolved humic substances, in the presence or not of calcium ions, from which the fouling layer specific resistances, a, were calculated. Specifically, resistance a was calculated in all cases from the slope of the linear region of the first filtration curve, which is typical of the cake formation process. The increased level of the quantity (t/V) in the 2nd and 3rd filtration cycle (Fig. 2) is attributed to a greater amount of deposited HA and perhaps to greater compactness of the fouling layer. Furthermore, as is wellknown (e.g. [41]), fouling layers in the presence of Ca2+ exhibit smaller permeability compared to those forming in the absence of calcium. This is clearly shown in data obtained with the NF90 membrane (Fig. 3), as well as with the other two membranes, for solutions with and without Ca2+. Data on percent flux reduction (FR), adsorption of humic substances (MHS) and of specific cake resistance (a) for the fouled membranes are summarized in Table 6. These data suggest that fouling is related to the type of foulant, the concentration of the humic substance, and the presence of calcium ions in the feed solution. In the absence of calcium ions, humic substances did not cause much fouling, as also observed in previous studies with UF and RO membranes; e.g. [40,41]. This is reflected in the slight flux decline and the rather small HS adsorption on the membrane surfaces; the latter varies between 0 and 0.171 g/m2, depending on the membrane and the humic substance type. Specifically, NF270 and NF90 membranes appear to absorb HA to a larger extent than NOM and FA, while XLE membranes seem to be more susceptible to fulvic acids fouling. These differences in organic adsorption are likely related to the characteristic acidity and aromaticity of the three Suwannee River substances (Tables 3 and 5), since the more aromatic and, therefore, hydrophobic humic acids are more susceptible to hydrophobic interactions with the membranes, especially with the hydrophobic ones (NF90). Moreover, the lower acidity of the HA and NOM compounds is likely leading to reduced electrostatic repulsion by the negatively charged membranes, thus promoting adsorption. On the other hand, very small interactions likely occur between FA and the two NF membranes due to the smaller aromaticity and the greater acidity of the fulvic acid mole- Permeate flux (L/m2h) 3.1. Changes in membrane permeability due to fouling 30 20 1mM Ca2+ 10 10mg/L HA 10mg/L HA+1mM Ca2+ 0 0 40 80 120 160 200 Filtration time (min) Fig. 3. Permeate flux temporal variation for filtration of various feed waters with NF90 membrane. cules. Specifically, the higher negative charge of NF90 and NF270 membranes renders them less susceptible to adsorption of negatively charged organics; this is not observed in the case of neutrally charged XLE membrane. The very significant effect of calcium ions on organic fouling (reflected in the data of Table 6) has been well documented in the literature [24,40–45], where the increased flux decline is attributed to two different mechanisms. The first is related to the charge neutralization capacity of the calcium ions, which bind specifically through complex formation with the acidic functional groups (predominantly carboxylic) of the humic materials. The concomitant charge reduction of the humic compounds is considered [24] to result in the formation of small, coiled humic macromolecules which are readily deposited on the membrane surface, forming quite compact organic layers. The second mechanism is related to the unique intermolecular bridging capability of calcium ions which can induce the arrangement of humic acid molecules into a ‘‘cross-linked’’ structure in the fouling layer, rendering it very compact [44]. Moreover, calcium ions may directly link the humic molecules with the membrane surface, leading to a strong and highly resistant (to mechanical and hydrodynamic forces) fouling layer. It is interesting to note that the specific cake resistances obtained in dead-end experiments [40] with UF membranes and humic acids (from Aldrich) are in substantial agreement with the results of the present NF experiments. 3.0 o Experimental data --- Linear curve fitting t/V (min/mL) 2.5 2.0 1.5 1.0 1st cycle 0.5 2nd cycle 3rd cycle 0.0 0 20 40 60 80 100 120 Cumulative volume, V (mL) Fig. 2. Filtration data of HA plus calcium solutions plotted as t/V versus V for three sequential cycles (NF90 membrane). Feed solution composition 10 mg/L HA, 1 mM Ca2+. Pressure 5 bar, pH 7, stirring rate 250 rpm. 252 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Table 6 Humic substance deposition (MHS), flux reduction (FR%) and specific cake resistance (a) of fouled NF270, NF90 and XLE membranes (pressure 5 bar, pH 7). Foulant solution (mg/LHS/mg/LCa2+) NF270 NF90 a a MHS (g/m2) FR (%) (1015 m/kg) XLE a MHS (g/m2) FR (%) (1015 m/kg) a MHS (g/m2) FR (%)a (1015 m/kg) a HA/Ca2+ 10/0 20/0 10/40 20/40 2.9 4.1 10.0 16.0 0.075 0.158 0.301 0.781 4.1 7.9 17.3 22.4 (1.1) (7.7) (14.5) (18.3) 4.3 16.0 39.0 32.0 0.027 0.082 0.158 0.144 3.6 10.7 30.1 36.6 (1.2) (9.2) (27.4) (19.7) 4.3 8.6 36.0 27.0 0.021 0.075 0.219 0.151 3.5 9.6 36.1 32.3 (11.1) (16.1) (25.8) (22.0) FA/Ca2+ 10/0 20/0 10/40 20/40 b 0.7 4.3 10.0 0.0 0.034 0.062 0.151 1.1 3.6 8.2 20.3 (1.9) (3.1) (8.9) (16.0) 8.6 14.0 16.0 17.0 0.068 0.082 0.096 0.096 9.7 7.3 17.0 18.8 (12.5) (9.3) (19.7) (14.5) 5.8 20.0 10.0 26.0 0.082 0.171 0.116 0.212 5.7 6.8 20.1 29.5 (9.6) (10.3) (17.8) (20.3) NOM/Ca2+ 10/0 20/0 10/40 20/40 b 1.4 4.3 7.2 0.0 0.010 0.048 0.068 1.4 6.8 9.2 16.6 (4.4) (5.5) (5.4) (15.9) 2.9 12.0 12.0 29.0 0.014 0.096 0.082 0.253 6.1 11.9 10.9 25.3 (8.7) (14.6) (17.0) (19.7) 13.0 12.0 15.0 13.2 0.068 0.055 0.197 0.103 11.7 14.0 24.4 19.6 (15.8) (17.5) (21.3) (16.6) a Values in the brackets describe the pure water flux reduction; comparison between the stabilized permeate fluxes measured at the beginning and the end of the experiment (steps a and j, Fig. 2). b No positive slope in the respective t/V vs V curves (linear curve fittings). It is occasionally reported in the literature (e.g. [19]) that organic fouling on NF membranes is well correlated with the membrane pore size, meaning that flux decline is more severe on loose NF membranes compared to the tighter ones. Such correlation was a not observed in this study (Table 6). The present findings rather underpin the marked effect of the type of the organic species present in the raw water on NF fouling, which appears to be directly related to the surface characteristics of the membranes. 80 Contact Angle (o) 70 60 50 40 30 20 10 0 NF270 Virgin 20mg/L FA b NF90 10mg/L HA 10mg/L NOM 20mg/L HA 20mg/L NOM XLE 10mg/L FA 80 Contact Angle (o) 70 60 50 40 30 20 10 0 NF270 Virgin 10mg/L FA + 1mM Ca2+ 20mg/L NOM + 1mM Ca2+ NF90 10mg/L HA + 1mM Ca2+ 20mg/L FA + 1mM Ca2+ X LE 20mg/L HA + 1mM Ca2+ 10mg/L NOM + 1mM Ca2+ Fig. 4. Membrane surface contact angle values; (a) after the filtration of humic substances alone, and (b) after the filtration of humic substances in the presence of calcium ions. 253 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Contact Angle (o) a 80 70 60 50 40 30 NF270 20 NF90 10 XLE 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 MHA (g/m2) b 80 Contact Angle (o) 70 60 50 40 30 NF270 20 NF90 10 XLE 0 0 0.05 0.1 0.15 0.2 0.25 MFA (g/m2) c 80 Contact Angle (o) 70 60 50 bars correspond to the minimum and maximum values measured for each membrane specimen (from at least five readings). These data suggest that the effect of fouling on membrane hydrophobicity is not the same for the three membranes. In the absence of calcium, humic substance filtration through the denser NF90 and XLE membranes resulted in small changes of the membrane surface contact angle. However, the NF270 membrane became less hydrophilic due to fouling. On the other hand, the increased adsorption of HS-calcium complexes on the membrane surfaces enhanced significantly the hydrophobicity of the loose NF270 membrane, but altered slightly the hydrophobic character of the NF90 and XLE membranes. An exception is observed in the case of the FA-calcium complexes which rendered the XLE membrane considerably more hydrophilic, and the NF270 membrane slightly less hydrophilic than the virgin one. The results shown in Fig. 4 are indicative of the adsorption of humic substances onto the membranes and consequently of the effect of foulant physicochemical properties on membrane surface characteristics. This is evident in Fig. 5, in which the membrane contact angle appears to be linearly related with adsorbed foulant mass; this effect is either positive or negative, possibly depending on the number of the hydrophilic and/or hydrophobic functional groups characterizing the deposited molecules. For instance, the hydrophobic humic acids are related with increased values of membrane contact angles, while fulvic acids tend to reduce hydrophobicity as a result of their limited number of hydrophobic groups. A similar trend is also observed in the work performed by Xu et al. [46], where the slight decrease observed in contact angle values of the NF90 and XLE membranes could be attributed to the hydrophilic character of the foulants used in their study. Although it is difficult to clearly isolate the effect of organic fouling on membrane hydrophobicity/hydrophilicity, the results of the present study support the commonly held opinion that the greater the extent of fouling, the greater the contribution of foulant properties on membrane surface characteristics. 40 30 NF270 20 NF90 10 XLE 0 0 0.05 0.1 0.15 0.2 0.25 0.3 MNOM (g/m2) Fig. 5. Membrane contact angle vs deposited mass surface density for (a) humic acids (MHA), (b) fulvic acids (MFA), and (c) NOM (MNOM). 3.2. Characterization of fouled membranes 3.2.1. Contact angle measurements The mean contact angle values of the virgin and the fouled with humic substances membranes are depicted in Fig. 4, in which the 3.2.2. Energy dispersive X-ray spectroscopy measurements (X-ray EDS) The elemental composition of the fouled membranes is given in Table 7. The major inorganic constituents of the foulants quantified by energy dispersive spectroscopy (EDS) are Cu, Fe, Si, and Ca, while the large values of %S are most likely due to the active layer support (made of polysulfone). In support of this interpretation is the low content of sulfur characterizing the three IHSS substances (Table 3), as well as the small thickness of the membrane active layers (<100 nm) which may allow probing deep into the membrane. Therefore, one may suggest that the sulfur content (%S) of HS-fouled membranes (determined through X-ray EDS and X-ray photoelectron spectroscopy (XPS) measurements performed in previous studies [22]), could serve as a qualitative indicator of the deposition of organic substances (especially of the hydrophobic ones) on the membrane surface. For instance, the high %S in membranes fouled by fulvic acids and calcium ions may be indicative of Table 7 Elemental composition (%) of fouled by humic substances (10 mg/L) and calcium ions (1 mM) membranes, as determined by X-ray EDS. % Relative composition Al Si S Ca Fe Ni Cu Cr NF270 NF90 XLE HA/Ca2+ FA/Ca2+ NOM/Ca2+ HA/Ca2+ FA/Ca2+ NOM/Ca2+ HA/Ca2+ FA/Ca2+ NOM/Ca2+ 0.96 2.06 42.85 15.79 8.35 4.46 25.54   0.70 83.33 3.71 1.81  6.83 4.54 1.46 1.96 57.94 6.71 15.04 0.71 17.04  0.41 12.46 35.02 5.18 2.23 1.07 43.62  0.88 0.72 86.93 0.66 4.85  8.75  0.41 7.35 78.75 5.07 2.07  7.71 0.33 5.84 15.52 5.41 8.08 21.24 0.77 35.48 7.66 0.65 0.44 89.18 0.81 0.49 0.89 7.95  3.73 29.87 57.12 1.76 0.79  8.83  254 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 reduced fouling, which is in accord with the smaller flux decline (Table 6) and limited changes on the membrane surface morphology (as evidenced by SEM images not shown here). 3.3. Effect of fouling on salt retention Difference in NaCl rejection with virgin membrane (%) The influence of organic fouling on salt retention (expressed as conductivity ratio), for all humic substances and membranes tested, is shown in Fig. 6. The retention values are plotted as the difference in NaCl retention observed between the virgin and the fouled membranes (steps c and i in Fig. 1, respectively); positive values correspond to increased retention by the fouled membranes, compared to the virgin ones. The values in Fig. 6 correspond to the final measurements made after the recovery of 50% water as permeate (mean values). The error bars designate the minimum and maximum values measured for each membrane specimen; i.e. values obtained from triplicate experiments with different membrane specimen of the same type. The differences in salt retention vary between the three membranes, being more pronounced for the relatively loose NF270 membrane. In general, significant improvement in membrane desalting efficiency was observed in the case of membranes fouled by calcium complexes of all foulants. Salt retention was greatly improved in the case of membranes fouled by HA-Ca2+, being greater for higher HA concentrations (20 mg/L). In the absence of calcium ions, different behavior was observed in the case of the relatively loose NF270 membrane; i.e. fouling by all HS used led to significantly reduced salt retention compared to the virgin membrane. Although it is difficult to isolate the contribution of organic fouling and membrane zeta potential on salt retention, the above results suggest that foulant deposits may significantly change the membrane desalting capacity as a result of modified charge, steric and hydrophobic effects. Interestingly, the differences observed in the retention values appear to be related with the acidity properties of the adsorbed HS and specifically with their carboxylic acidity [23]. As shown in Fig. 7a, an increased carboxylic content of the adsorbed HS (calculations based on carbon composition, humic concentration and carboxylic acidity of HS used) is related to increased salt retention. This trend is possibly related to both steric and charge effects, due to pore restriction and increased negative charge on the fouled membrane surfaces, respectively. Specifically, the adsorbed carboxylic acidity inside the membrane pores can lead to increased intra membrane repulsion (within the membrane pores) which may cause a reduction of the effective pore size, leading to reduced permeability of the membranes (Fig 7b), and consequently to a better retention of hydrated ionic species. Fig 7b indicates that this mechanism may be involved in the case of rather porous NF270 membrane, for which the adsorbed carboxylic acidity appears to be positively correlated with the pure water flux reduction, something that is not clearly observed for the tight NF90 and XLE membranes. Moreover, as shown in Fig. 7a, the filtration of low acidity HS through NF270 tends to increase the ionic permeability, and consequently reduce rejection. This is in accord with the expected reduced steric effects of the loose NF270 membranes after the preliminary treatment with HS of low carboxylic acidity. On the other hand, increased charge effects may be attributed to the adsorbed humic acidity on the membrane surfaces, which possibly account for the increased NaCl rejection by the tight NF90 and XLE membranes; this is especially true for the neutrally charged XLE membrane. 3.4. Effect of membrane fouling on triazine retention Figs. 8–10 show the mean values of triazine retention by fouled with humic substances, in the absence or presence of calcium ions, for NF270, NF90 and XLE membranes, respectively. The error bars correspond to the minimum and maximum values obtained from triplicate experiments with different membrane specimen of the same type. In the same figures, the adsorption capacity of the two triazines on the virgin and fouled membranes is also included (as a percentage of the initial feed concentration). Triazine retention between fouled and virgin membranes varied from 34.8% to +17.1% in the case of NF270 membranes, from 28.0% to +19.5% in the case of NF90 membranes, and from 17.7% to +14.3% for the XLE membranes, depending on the type of the humic substances and the presence or not of calcium ions in the foulant solution. The corresponding range of the differences observed in triazine retention (51.9%, 47.5% and 32% for the NF270, NF90 and XLE membranes, respectively) indicates that the 12 NF270 HA NF90 XLE 10 NOM 8 6 FA 4 2 0 -2 -4 -6 10mg/L HA 20mg/L HA 10mg/L HA 20mg/L HA 10mg/L FA 20mg/L FA 10mg/L FA 20mg/L FA + 1mM + 1mM + 1mM +1mM Ca 2+ Ca2+ Ca 2+ Ca 2+ 10mg/L NOM 20mg/L NOM 10mg/L NOM + 1mM Ca2+ 20mg/L NOM + 1mM Ca 2+ Fig. 6. Difference in salt retention between fouled and virgin NF270, NF90 and XLE membranes. Negative and positive values correspond to a decreased and increased retention, respectively. Rejection values for clean NF270, NF90 and XLE membranes: 66.4% ± 1.3%, 99.5% ± 1.8% and 95.9% ± 0.5%, respectively. 255 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Difference in NaCl rejection with virgin membrane (%) a Pure water flux reduction (%) b 8 6 4 2 0 -2 NF270 -4 NF90 XLE -6 0 1 2 3 Carboxylic acidity of adsorbed HS 4 5 6 (meqx10-3) 30 25 20 15 10 NF270 5 NF90 XLE 0 0 1 2 3 4 5 6 Carboxylic acidity of adsorbed HS (meqx10-3) Fig. 7. (a) Difference in salt retention between fouled and virgin membranes versus the estimated? total carboxylic acidity characterizing the adsorbed humic substances (rejection values for clean NF270, NF90 and XLE membranes: 66.4% ± 1.3%, 99.5% ± 1.8% and 95.9% ± 0.5%, respectively). (b) Pure water flux reduction versus the adsorbed humic carboxylic acidity. influence of organic fouling on triazine retention is more significant for the more porous NF270 membrane. This is consistent with results from previous studies [15,19] which suggest that the magnitude of the influence of membrane fouling on retention, either positive or negative, decreases as the membrane pore size decreases. It is interesting to note that the influence of organic fouling on triazine retention by the NF270 and NF90 membranes differs from that observed in a previous study [19], where these membranes exhibited small changes in the retention of three pharmaceutically active compounds when fouled by a cocktail consisting of soil-born humic acids (Sigma–Aldrich) and background electrolytes. This trend may be attributed to the different solute and foulant characteristics, since the smaller solubility of the larger MW Sigma–Aldrich HA used may have resulted in greater fouling and as a consequence in improved rejection for all solutes tested in that study [19]. Previous work on triazine retention by virgin membranes showed that there is a pronounced effect of molecular size and hydrophobicity (log Kow) of the solute and of the physical properties of membrane surface (porosity, hydrophobicity) [7]. Therefore, differences in triazine retention between fouled and virgin membranes could be explained by the changes taking place in mem- brane–solute hydrophobic interactions and membrane–solute size exclusion as a result of the modified membrane surface characteristics. Indeed, findings of the present study indicate that the transport of the relatively hydrophobic triazine compounds is dependent on solute solubility and diffusion through the fouled membranes. This is highlighted by the adsorption data summarized in Figs. 8–10, which strongly suggest that the changes in triazine retention are related to the observed differences in triazine adsorption between virgin and fouled membranes. Specifically, from Figs. 8–10, the following observations can be made:  An increased triazine adsorption on fouled membranes is normally related to lower retention; the opposite is observed for reduced adsorption on the fouled membranes.  The formation of loose organic layer (Table 6) of intermediate hydrophobicity (Fig. 5) on the membrane surface, results in an increased diffusion of the rather neutral and hydrophobic triazines, thereby facilitating the triazine transport to the permeate side; this is the case of HS fouling in the absence of calcium ions.  The formation of a rather dense fouling layer (Table 6) formed by HS-calcium complexes may hinder solute diffusion, thus improving triazine retention. An exception is observed in the 256 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Rej. Atrazine Rej. Prometryn Ads. Atrazine Ads. Prometryn 30 100 90 Triazine rejection (%) 70 20 60 15 50 40 10 30 20 Triazine adsorption (%) 25 80 5 10 0 0 Virgin Fouled-10mg/L HA Fouled-20mg/L HA Fouled-10mg/L HA+1mM Ca2+ Fouled-20mg/L HA+1mM Ca2+ 60 100 90 Triazine rejection (%) 70 40 60 30 50 40 20 30 20 Triazine adsorption (%) 50 80 10 10 0 0 Fouled-10mg/L FA Fouled-20mg/L FA Fouled-10mg/L FA+1mM Ca2+ Fouled-20mg/L FA+1mM Ca2+ 100 18 90 16 Triazine rejection (%) 80 14 70 12 60 10 50 8 40 6 30 Triazine adsorption (%) Virgin 4 20 2 10 0 0 Virgin Fouled-10mg/L NOM Fouled-20mg/L NOM Fouled-10mg/L NOM+1mM Ca2+ Fouled-20mg/L NOM+1mM Ca2+ Fig. 8. Atrazine and prometryn rejection (columns) and adsorption (points) by virgin and fouled (with HA, FA and NOM) NF270 membranes; fouling layers formed in the presence or absence of Ca2+. 257 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Rej. Atrazine Rej. Prometryn Ads. Atrazine Ads. Prometryn 35 100 90 30 25 70 60 20 50 15 40 30 10 Triazine adsorption (%) Triazine rejection (%) 80 20 5 10 0 0 Virgin Fouled-10mg/L HA Fouled-20mg/L HA Fouled-10mg/L HA+1mM Ca2+ Fouled-20mg/L HA+1mM Ca2+ 30 100 90 Triazine rejection (%) 70 20 60 15 50 40 10 30 20 Triazine adsorption (%) 25 80 5 10 0 0 Virgin Fouled-10mg/L FA Fouled-20mg/L FA Fouled-10mg/L FA+1mM Ca2+ Fouled-20mg/L FA+1mM Ca2+ 100 35 90 30 25 70 60 20 50 15 40 30 10 Triazine adsorption (%) Triazine rejection (%) 80 20 5 10 0 0 Virgin Fouled-10mg/L NOM Fouled-20mg/L NOM Fouled-10mg/L NOM+1mM Ca2+ Fouled-20mg/L NOM+1mM Ca2+ Fig. 9. Atrazine and prometryn rejection (columns) and adsorption (points) by virgin and fouled (with HA, FA and NOM) NF90 membranes; fouling layers formed in the presence or absence of Ca2+. 258 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 Rej. Atrazine Rej. Prometryn Ads. Atrazine Ads. Prometryn 100 40 90 35 30 70 25 60 50 20 40 15 30 10 Triazine adsorption (%) Triazine rejection (%) 80 20 5 10 0 0 Virgin Fouled-10mg/L HA Fouled-20mg/L HA Fouled-10mg/L HA+1mM Ca2+ Fouled-20mg/L HA+1mM Ca2+ 25 100 Triazine rejection (%) 80 20 70 60 15 50 40 10 30 20 Triazine adsorption (%) 90 5 10 0 0 Fouled-10mg/L FA Fouled-20mg/L FA Fouled-10mg/L FA+1mM Ca2+ Fouled-20mg/L FA+1mM Ca2+ 100 45 90 40 Triazine rejection (%) 80 35 70 30 60 25 50 20 40 15 30 10 20 Triazine adsorption (%) Virgin 5 10 0 0 Virgin Fouled-10mg/L NOM Fouled-20mg/L NOM Fouled-10mg/L NOM+1mM Ca2+ Fouled-20mg/L NOM+1mM Ca2+ Fig. 10. Atrazine and prometryn rejection (columns) and adsorption (points) by virgin and fouled with HA, FA and NOM, XLE membranes; fouling layers formed in the presence or absence of Ca2+. 259 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 case of NF270 membranes fouled by NOM–Ca2+; i.e., atrazine and prometryn retention decreased by 35% and 13%, respectively, irrespective of NOM concentration in the foulant solution.  Tight and hydrophobic membranes (NF90 and XLE) display significant changes of triazine retention when fouled by humic substances of increased hydrophobicity (HA, NOM). On the other hand, membrane fouling by rather hydrophilic humic substances (FA) does not greatly affect triazine retention.  On the contrary, porous and hydrophilic membranes (NF270) display significant changes of triazine retention when fouled by humic substances of decreased hydrophobicity (FA).  Size exclusion apparently plays a significant role in the transport of hydrophobic non-ionic triazines across fouled membranes; the retention of prometryn was always higher than It should be noted that it is impossible to distinguish between adsorption of the triazines on the fouling layers and on the membrane itself. Considering that a preliminary membrane ‘‘saturation’’ with the two triazines took place prior to membrane fouling, it can be assumed that the variations in triazine adsorption, and, therefore, in triazine retention are due to the modified membrane characteristics after their treatment with HS solutions, which can either increase or reduce solute diffusion. For instance, the formation of a loose organic layer of intermediate hydrophobicity enhances the hydrophobic interactions on the membrane surfaces which favor the accumulation of the two triazines near the membranes, thus 20 Difference in prometryn rejection with virgin NF270 membranes (%) 20 virgin NF270 membranes (%) Difference in atrazine rejection with a that of atrazine (especially for the tighter NF90 and XLE membranes), likely due to the larger molecular weight and size of prometryn. 10 0 -10 -20 HA FA -30 NOM 0 5 10 15 Specific cake resistance, α (x10 -5 HA FA -10 NOM 0 5 10 20 15 Specific cake resistance, α (x10 15 m/kg) 30 10 0 -10 -20 HA FA -30 NOM 0 10 20 40 30 15 Specific cake resistance, α (x10 20 10 0 -10 HA -20 FA NOM -30 50 0 10 20 30 40 15 m/kg) Specific cake resistance, α (x10 50 m/kg) 20 Difference in prometryn rejection with virgin XLE membranes (%) 20 virgin XLE membranes (%) 0 20 -40 Difference in atrazine rejection with 5 m/kg) Difference in prometryn rejection with virgin NF90 membranes (%) virgin NF90 membranes (%) Difference in atrazine rejection with 15 20 c 10 -15 -40 b 15 10 0 -10 HA FA -20 NOM -30 10 0 HA FA NOM -10 0 5 10 15 20 25 30 15 Specific cake resistance, α (x10 m/kg) 35 40 0 5 10 15 20 25 30 15 Specific cake resistance, α (x10 35 40 m/kg) Fig. 11. Difference in atrazine (left) and prometryn (right) retention between fouled and virgin membranes, as a function of the specific cake-layer resistances of fouled (a) NF270, (b) NF90 and (c) XLE membranes. 260 K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 hindering their diffusion back to the bulk solution. This effect is described in the literature [14,19,21,22] as cake-enhanced concentration polarization leading to reduced retention. However, cake-enhanced concentration polarization phenomena cannot be invoked to explain the reduced rejection of triazines through essentially non-fouled membranes (with negligible HS adsorption and flux decline). Indeed, it is interesting to note that the filtration of FA and NOM through the relatively hydrophilic and porous NF270 membrane resulted in an increased membrane hydrophobicity and a reduced triazine rejection, although the amount of the two humic substances adsorbed on the membrane surfaces was insignificant. On the other hand, a rather dense fouling layer serves as additional barrier, enhancing the sieving effect (increased retention), as previously observed [16]. This phenomenon appears to be independent of the hydrophobicity of the fouled membranes. Specifically, the greater hydrophobicity measured in the case of the dense HS/Ca layer was not accompanied by a greater adsorption and diffusion of the triazine compounds, as in the case of the loose HS layers. Probably, the hydrophobic interactions between the hydrophobic humic-calcium complexes and the triazine compounds are more important than the respective interactions with the hydrophobic groups of the membrane surfaces. The experimental results indicate that an intermediate case (between the above two cases) is dependent on the characteristics of the organics accumulated on the membrane surfaces. As can be seen in Fig. 11, there seems to be a rather positive correlation between the specific cake layer resistance (a) of the HS deposits and the differences observed in triazine retention between fouled and virgin NF membranes (NF270, NF90). In the case of the non-porous and relatively uncharged ULPRO membranes (XLE, Fig. 11c) such correlation was only observed in the case of the smaller atrazine molecules. On the other hand, humic substances characterized by a greater hydrophilicity and acidity (like the fulvic acids) result in fouling resistances which exhibit a positive but rather smaller correlation with triazine retention, especially in the case of hydrophilic and smooth membrane surfaces (NF270). Such behavior is not observed in the case of tight and hydrophobic NF/ULPRO membranes (NF90 and XLE membranes), which are characterized by a similar retention performance regardless of the FA deposition on their surface. The above results seem to explain the differences in solute retention observed in the literature [14–23], since solute transport through fouled membranes can be strongly affected by the physicochemical properties of both the foulants (hydrophobicity and charge) and the solutes (size and hydrophobicity), as well as by the characteristics of the membranes (MWCO, hydrophobicity, surface charge). Moreover, the results of the present study are in accord with the differences observed in triazine retention when humic substances are filtered together with the two triazines [29]. Indeed, the differences observed [29] in solute rejection could not be fully attributed to the interactions taking place in the bulk between triazines and humic substances (i.e. formation of pseudo-complexes), thereby, indicating the important role of membrane fouling on process effectiveness. For instance, the somewhat reduced retention of the two triazines observed in the case of XLE membrane, when filtered together with humic acids [29], is probably the result of the increased triazine adsorption on the membrane, which, according to present study results, is facilitated by the fouling layer formed on the membrane surface. 4. Conclusions Organic fouling, depending on the nature and the relative concentration of humic substances as well as on the presence of cal- cium, results in considerable changes of the membrane surface characteristics, including the contact angle (an indicator of hydrophobicity) and salt retention (an indicator of surface charge and Donnan effects), which in turn can significantly affect the retention of triazine herbicides. In general, the permeation or rejection of tested, relatively hydrophobic and non-ionic, triazine compounds appears to be influenced by their solubility and diffusivity through the fouled membranes; this influence depends on the extent of membrane fouling. Specifically, limited deposition of organic foulants on the membrane surfaces favors the diffusion of the two triazines across the membranes, and therefore, their permeation (reduced retention). The magnitude of this permeation is related to the physicochemical properties of the foulants (hydrophobicity and charge) and the characteristics of the membranes used (MWCO, hydrophobicity and surface charge), being greater for smaller organic compounds (e.g. atrazine) and rather loose NF membranes (e.g. NF270). On the other hand, the formation of a dense fouling layer, comprised of calcium complexes with humic substances, results in an increased retention for both triazine compounds. In particular, dense layers can serve as additional barriers which enhance the sieving effect. This phenomenon appears to be more pronounced in the case of tight and negatively charged NF membranes (e.g. NF90). The new results support the view that there is a need for good understanding of the relevant characteristics of the organic matter present in natural waters. This would aid the prediction of the organic fouling tendency of the selected membrane, which in turn may significantly affect the retention of pesticides, and in general of low MW organic compounds, from potable water by NF/ULPRO membranes. Appendix A A.1. Calculation of the osmotic pressure difference (Dpm) in the case of 1 mM Ca2+ feed solution Taking into account the concentration polarization effects, the osmotic pressure difference can be estimated as follows: Dpm ¼ 3C b;Ca RTRo exp   J k ðA:1Þ where Cb,Ca is the bulk calcium concentration, R is the universal gas constant, T is the absolute temperature, Ro is the observed calcium ion rejection, while constant 3 accounts for a 1:2 electrolyte solution (CaCl2) at low to moderate salt concentrations where the van’t Hoff law is valid. The broadly accepted concentration polarization factor [exp(J/k)] is determined by estimating the bulk mass transfer coefficient k, for stirred-cells, from the correlation presented by Smith et al. [48]: Sh ¼ k r sc ¼ 0:27Re0:567 Sc0:33 Dw ðA:2Þ where rsc is the stirred cell radius, Dw the bulk diffusivity of the solute (CaCl2), Re the Reynolds number ð¼ xr2sc =v Þ, m the solution kinematic viscosity (equal to that of water at 25 °C,=0.8926  106 m2/ s), x the stirring speed in rad/s (=2pN/60, where N = 250 rpm) and Sc the Schmidt number (=m/Dw). Using a constant bulk diffusion coefficient of calcium chloride in water, Dw = 1.45  109 m2/s [47], the estimated mass transfer coefficient is k  3.38  105 m/s. The magnitude of the concentration polarization factor [exp(J/k)] varies between 1.2 and 1.4 for fluxes between 22 L/m2 h (NF90, XLE membranes) and 42 L/m2 h (NF270 membranes). This correction is applied in Eq. (A.1), from which the estimated osmotic pressure difference is found to be negligible (in the order of 105 bar). K.V. Plakas, A.J. Karabelas / Separation and Purification Technology 80 (2011) 246–261 References [1] K.O. Agenson, J.-I. Oh, T. Urase, Retention of a wide variety of organic pollutants by different nanofiltration/reverse osmosis membranes: controlling parameters of process, J. Membr. Sci. 225 (2003) 91–103. [2] K. Kimura, S. Toshima, G. Amy, Y. Watanabe, Retention of neutral endocrine disrupting compounds (EDCs) and pharmaceutical active compounds (PhACs) by RO membranes, J. Membr. Sci. 245 (2004) 71–78. [3] L.D. Nghiem, A. Schäfer, M. Elimelech, Removal of natural hormones by nanofiltration membranes: measurement, modeling, and mechanisms, Environ. Sci. Technol. 38 (2004) 1888–1896. [4] L.D. Nghiem, A.I. Schäfer, M. Elimelech, Pharmaceutical retention mechanisms by nanofiltration membranes, Environ. Sci. Technol. 39 (2005) 7698–7705. [5] P. Xu, J.E. Drewes, C. Bellona, G. Amy, T.U. Kim, M. Adam, T. Heberer, Retention of emerging organic micropollutants in nanofiltration–reverse osmosis membrane applications, Wat. Env. Res. 77 (1) (2005) 40–48. [6] Y. Yoon, P. Westerhoff, S.A. Snyder, E.C. Wert, Nanofiltration and ultrafiltration of endocrine disrupting compounds, pharmaceuticals and personal care products, J. Membr. Sci. 270 (2006) 88–100. [7] K.V. Plakas, A.J. Karabelas, Membrane retention of herbicides from single and multi-solute media: the effect of ionic environment, J. Membr. Sci. 320 (2008) 325–334. [8] Y. Zhang, C. Causserand, P. Aimar, J.P. Cravedi, Removal of bisphenol A by a nanofiltration membrane in view of drinking water production, Wat. Res. 40 (2006) 3793–3799. [9] D. Libotean, J. Giralt, R. Rallo, Y. Cohen, F. Giralt, H.F. Ridgway, G. Rodriguez, D. Phipps, Organic compounds passage through RO membranes, J. Membr. Sci. 313 (2008) 23–43. [10] I. Koyuncu, O.A. Arikan, M.R. Wiesner, C. Rice, Removal of hormones and antibiotics by nanofiltration membranes, J. Membr. Sci. 309 (2008) 94–101. [11] A.J. Karabelas, K.V. Plakas, E.S. Solomou, V. Drossou, D.A. Sarigiannis, Impact of European legislation on marketed pesticides – a view from the standpoint of health impact assessment studies, Environ. Intern. 35 (2009) 1096–1107. [12] C. Bellona, J.E. Drewes, P. Xu, G. Amy, Factors affecting the retention of organic solutes during NF/RO treatment – a literature review, Wat. Res. 38 (2004) 2795–2809. [13] A.J. Karabelas, K.V. Plakas, Membrane treatment of potable water for pesticides removal, in: S. Soloneski, M. Larramendy (Eds.), Herbicides, Theory and Applications, InTech Open Access Publisher, Croatia, 2011, pp. 369–408. [14] H.Y. Ng, M. Elimelech, Influence of colloidal fouling on retention of trace organic contaminants by reverse osmosis, J. Membr. Sci. 244 (2004) 215–226. [15] P. Xu, J.E. Drewes, T.-U. Kim, C. Bellona, G. Amy, Effect of membrane fouling on transport of organic contaminants in NF/RO membrane applications, J. Membr. Sci. 279 (2006) 165–175. [16] K.V. Plakas, A.J. Karabelas, T. Wintgens, T. Melin, A study of selected herbicides retention by nanofiltration membranes – the role of organic fouling, J. Membr. Sci. 284 (2006) 291–300. [17] E. Steinle-Darling, M. Zedda, M.H. Plumlee, H.F. Ridgway, M. Reinhard, Evaluating the impacts of membrane type, coating, fouling, chemical properties and water chemistry on reverse osmosis retention of seven nitrosoalklyamines, including NDMA, Wat. Res. 41 (2007) 3959–3967. [18] K.O. Agenson, T. Urase, Change in membrane performance due to organic fouling in nanofiltration (NF)/reverse osmosis (RO) applications, Sep. Pur. Technol. 55 (2007) 147–156. [19] L.D. Nghiem, S. Hawkes, Effects of membrane fouling on the nanofiltration of pharmaceutically active compounds (PhACs): mechanisms and role of membrane pore size, Sep. Pur. Technol. 57 (2007) 176–184. [20] C. Bellona, M. Marts, J.E. Drewes, The effect of organic membrane fouling on the properties and retention characteristics of nanofiltration membranes, Sep. Pur. Technol. 74 (2010) 44–54. [21] L.D. Nghiem, P.J. Coleman, NF/RO filtration of the hydrophobic ionogenic compound triclosan: transport mechanisms and the influence of membrane fouling, Sep. Pur. Technol. 62 (2008) 711–718. [22] A.R.D. Verliefde, E.R. Cornelissen, S.G.J. Heijman, I. Petrinic, T. Luxbacher, G.L. Amy, B. Van der Bruggen, J.C. van Dijk, Influence of membrane fouling by (pretreated) surface water on retention of pharmaceutically active compounds (PhACs) by nanofiltration membranes, J. Membr. Sci. 330 (2009) 90–103. [23] V. Yangali-Quintanilla, A. Sadmani, M. McConville, M. Kennedy, G. Amy, Retention of pharmaceutically active compounds and endocrine disrupting compounds by clean and fouled nanofiltration membranes, Wat. Res. 43 (2009) 2349–2362. 261 [24] S. Hong, M. Elimelech, Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes, J. Membr. Sci. 132 (1997) 159– 181. [25] J. Cho, G.L. Amy, J. Pellegrino, Membrane filtration of natural organic matter: initial comparison of retention and flux decline characteristics with ultrafiltration and nanofiltration membranes, Wat. Res. 33 (1999) 2517–2526. [26] M. Mänttäri, L. Puro, J. Nuortila-Jokinen, M. Nyström, Fouling effects of polysaccharides and humic acid in nanofiltration, J. Membr. Sci. 165 (2000) 1– 17. [27] J.E. Kilduff, S. Mattaraj, G. Belfort, Flux decline during nanofiltration of naturally-occurring dissolved organic matter: effects of osmotic pressure, membrane permeability, and cake formation, J. Membr. Sci. 239 (2004) 39–53. [28] H. Winters, Twenty years experience in seawater reverse osmosis and how chemicals in pretreatment affect fouling of membranes, Desalination 110 (1997) 93–96. [29] K.V. Plakas, A.J. Karabelas, Triazine retention by nanofiltration in the presence of organic matter: the role of humic substance characteristics, J. Membr. Sci. 336 (2009) 86–100. [30] N. Hilal, H. AI-Zoubi, N.A. Darwish, A.W. Mohammad, Characterisation of nanofiltration membranes using atomic force microscopy, Desalination 177 (2005) 187–199. [31] K. Košutić, D. Dolar, D. Ašperger, B. Kunst, Removal of antibiotics from a model wastewater by RO/NF membranes, Sep. Purif. Technol. 53 (2007) 244–249. [32] Weed Science Society of America (WSSA), Herbicide Handbook, seventh ed., 1994. [33] B. Van der Bruggen, J. Schaep, W. Maes, D. Wilms, C. Vandecasteele, Nanofiltration as a treatment method for the removal of pesticides from ground waters, Desalination 117 (1998) 139–147. [34] K.N. Reddy, M.A. Locke, Molecular properties as descriptors of octanol–water partition coefficients of herbicides, Water, Air, Soil Pollut. 86 (1996) 389–405. [35] International Humic Substances Society, http://ihss.gatech.edu/ihss2/ index.html. [36] R. Beckett, Z. Jue, J.C. Giddings, Determination of molecular weight distributions of fulvic and humic acids using flow field-flow fractionation, Env. Sci. Technol. 21 (1987) 289–295. [37] P.M. Reid, A.E. Wilkinson, E. Tipping, M.N. Jones, Determination of molecular weights of humic substances by analytical (UV scanning) ultracentrifugation, Geochim. Cosmochim. Acta 54 (1990) 131–138. [38] S. Lee, B. Kwon, M. Sun, J. Cho, Characterizations of NOM included in NF and UF membrane permeates, Desalination 173 (2005) 131–142. [39] A.I. Schäfer, Natural Organic Removal Using Membranes, PhD thesis, The University of New South Wales, Chemical Engineering and Industrial Chemistry, Australia, 1999. [40] D.C. Sioutopoulos, S.G. Yiantsios, A.J. Karabelas, Relation between fouling characteristics of RO and UF membranes in experiments with colloidal organic and inorganic species, J. Membr. Sci. 350 (2010) 62–82. [41] K. Katsoufidou, S.G. Yiantsios, A.J. Karabelas, A study of ultrafiltration membrane fouling by humic acids and flux recovery by backwashing: experiments and modeling, J. Membr. Sci. 266 (2005) 40–50. [42] S.-H. Yoon, C.-H. Lee, K.-J. Kim, A.G. Fane, Effect of calcium ion on the fouling of nanofilter by humic acid in drinking water production, Wat. Res. 32 (1998) 2180–2186. [43] A.I. Schäfer, A.G. Fane, T.D. Waite, Nanofiltration of natural organic matter: removal, fouling and the influence of multivalent ions, Desalination 118 (1998) 109–122. [44] Q. Li, M. Elimelech, Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms, Env. Sci. Technol. 38 (17) (2004) 4683–4693. [45] S. Lee, M. Elimelech, Relating organic fouling of reverse osmosis membranes to intermolecular adhesion forces, Env. Sci. Technol. 40 (3) (2006) 980–987. [46] P. Xu, J.E. Drewes, Viability of nanofiltration and ultra-low pressure reverse osmosis membranes for multi-beneficial use of methane produced water, Sep. Pur. Technol. 52 (2006) 67–76. [47] J.M.M. Peeters, J.P. Boom, M.H.V. Mulder, H. Strathmann, Retention measurements of nanofiltration membranes with electrolyte solutions, J. Membr. Sci. 145 (1998) 199–209. [48] K.A. Smith, C.K. Colton, E.W. Merrill, L.B. Evans, Convective transport in a batch dialyzer: determination of true membrane permeability from a single measurement, Chem. Eng. Prog. Symposium Ser. No 84, 64 (1968) 45–58.