Facile Modification of Reverse Osmosis Membranes by Surfactant

Jan 26, 2017 - The top polyamide layer of composite reverse osmosis (RO) membranes has a fascinatingly complex structure, yet nanoscale nonuniformitie...
16 downloads 28 Views 4MB Size
Article pubs.acs.org/est

Facile Modification of Reverse Osmosis Membranes by SurfactantAssisted Acrylate Grafting for Enhanced Selectivity Katie Baransi-Karkaby, Maria Bass, Stanislav Levchenko, Shahar Eitan, and Viatcheslav Freger* Technion - Israel Institute of Technology, Wolfson Department of Chemical Engineering, Technion City, 32000 Haifa, Israel S Supporting Information *

ABSTRACT: The top polyamide layer of composite reverse osmosis (RO) membranes has a fascinatingly complex structure, yet nanoscale nonuniformities inherently present in polyamide layer may reduce selectivity, e.g., for boron rejection. This study examines improving selectivity by in situ “caulking” such nonuniformities using concentration polarization-enhanced graft-polymerization with a surfactant added to the reactive solution. The surfactant appears to enhance both polarization (via monomer solubilization in surfactant micelles) and adherence of graft-polymer to the membrane surface, which facilitates grafting and reduces monomer consumption. The effect of surfactant was particularly notable for a hydrophobic monomer glycidyl methacrylate combined with a nonionic surfactant Triton X-100. With Triton added at an optimal level, close to critical micellization concentration (CMC), monomer gets solubilized and highly concentrated within micelles, which results in a significantly increased degree of grafting and uniformity of the coating compared to a procedure with no surfactant added. Notably, no improvement was obtained for an anionic surfactant SDS or the cationic surfactant DTAB, in which cases the high CMC of surfactant precludes high monomer concentration within micelles. The modification procedure was also up-scalable to membranes elements and resulted in elements with permeability comparable to commercial brackish water RO elements with superior boric acid rejection.

1. INTRODUCTION Nanofiltration (NF) and reverse osmosis (RO) membrane processes are widely used today for desalination and water treatment. The most common RO and NF membranes are thin-film composite (TFC) membranes with a polyamide top layer, which appear to have an exceptionally favorable combination of high permeability, good salt rejection, robustness and durability.1,2 However, they still have a few drawbacks, for example, poor selectivity toward small uncharged solutes, such as boric acid3−14 and many organic micropollutants.15−22 One cause of under-performance could be nanoscopic defects and nonuniformities in the top active layer, which could impair solute retention.23,24 Redox-initiated radical graft-polymerization onto polyamide RO or NF membranes surface is a well-established approach to modifying and coating surfaces for improved performance.25−30 An attractive and facile version of this approach is concentration polarization (CP)-enhanced grafting, whereby a dilute monomer solution with freshly added initiators is filtered through the membrane, enhancing surface grafting as a result of CP effect.2 This method allows obtaining similar grafting using monomer concentrations lower by 1−2 orders of magnitude. Feasibility of improving NF/RO selectivity using this approach was demonstrated using various monomers. Glycidyl methacrylate (GMA) monomer was found to be particularly efficient for improving selectivity, which might be due to improved © XXXX American Chemical Society

coating adherence via epoxide groups reacting with residual amine groups of polyamide,31−33 in addition to covalent bonding of acrylate groups and physical adhesion by hydrophobic interactions.2,3,19 Notably, the method is upscalable to commercial elements and thus potentially allows tuning the membrane performance in situ to the needs of a specific process.4 However, changes in surface properties after modification indicate that the coating layer formed on the membrane surface could not be as well-defined and uniform as desired. Surface characterization often shows nonuniform modification degrees even at different points of the same sample. For instance, surface N content measured by X-ray photoelectron spectroscopy (XPS) does not drop sharply after modification, suggesting that coatings might not cover the entire surface.2,31,34 To address this problem, the present study considers a modification to the CP-enhanced grafting that may facilitate technological implementation of this approach and robust formation of a uniform coating. This is achieved by adding to the monomer solution a surfactant that would solubilize and Received: Revised: Accepted: Published: A

October 18, 2016 January 23, 2017 January 26, 2017 January 26, 2017 DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology Table 1. Characteristics of Monomer, Cross Linker and Surfactants Used in This Study

membrane area 11.3 cm2 (diameter 38 mm), magnetically stirred and pressurized with nitrogen. The flux was determined by collecting and weighing permeate and solute passage by measuring concentration in the feed and permeate. NaCl concentration was measured using electric conductivity meter (WTW inoLab Cond 7110). Boric acid (B) concentration was measured using inductively coupled plasma (Thermo Scientific -ICP-OES Spectromerer iCap 6000 Series) at a wavelength 248 nm. The passage P and rejection R of the solutes were calculated as follows:

concentrate monomers, as surfactants do to hydrophobic molecules in general.35−37 A polymer can then be formed in the solution when free radicals enter the micelles interior with solubilized monomers, enhancing polymerization in a way similar to emulsion polymerization.38,39 For instance, Seko et al. reported such grafting of GMA onto polyethylene fibers in emulsion using sodium n-dodecyl sulfate (SDS) as a surfactant.40 Furthermore, being much larger than monomers, micelles diffuse 2−3 orders of magnitude more slowly and undergo a much larger CP hence concentrate strongly next to the membrane surface. That further facilitate and preferentially direct polymerization toward surface grafting, either via polymer growth from surface (“grafting from”) or attachment of oligomers or polymers already formed within micelles to the surface (“grafting to”). The added surfactant may also adsorb onto the membrane surface,41 and thus reduce the interfacial free energy35,42 and promote adhesion and, subsequently, chemical attachment of the polymer to the surface. The above features are beneficial for having a coating procedure that would be robust, yet easily implementable and tunable to specific technological needs via composition of reactive solution. In this study, we analyze benefits of this novel surfactantassisted approach using a nonionic surfactant Triton X-100 as additive for CP-enhanced grafting of GMA for improving boric acid rejection of low-pressure RO (LPRO) membranes. We elucidate enhancement mechanism by comparing different surfactants, and, ultimately, demonstrate up-scalability and selectivity enhancement by implementing the optimized procedure in commercial LPRO elements.

P=1−R=

CP CF

(1)

where CP and CF are the concentrations measured in permeate and feed solutions, respectively. Filtration experiments with spiral wound membrane elements ESPA1−252143 were carried out in a pilot system including a 100 L feed tank, a membrane housing, a high pressure pump and a pressure regulator valve. Water flux was monitored using flow meters installed in the concentrate and permeate lines, water temperature was controlled using a heat exchanger. The system was operated in a closed loop recirculating the permeate, concentrate and bypass flow back to the feed tank. Membrane permeability was tested at a constant concentrate flow rate, the pressure was monitored at 10 to 20 bar, two ranges of feed flow rates were applied, 330− 400 L/h and 600−680 L/h. Sodium chloride and boric acid rejection were tested in the standard conditions specified for the element by the manufacturer (10.3 bar, 1500 ppm of NaCl, 5 ppm boric acid, 10% recovery)43 at pH ≈ 7.3 or in similar conditions after modification. In addition, rejection at 15 and 20 bar were also tested using the feed flow rates in the range 350−680 L/h. Membrane Modification. Lab scale tests used flat sheets of ESPA-1 mounted in dead-end cells. The cells were then filled with a solution of GMA monomer with added surfactant at selected concentrations. Cross-linker (0.003 mM N,N′methylenebisacrilamide, MBAA, see Table 1) and initiators (4 mM potassium persulfate and 2 mM potassium metabisulfite, dissolved separately in water) were added shortly before starting the modification and the cell was immediately closed and pressurized with nitrogen to 20 bar.

2. MATERIALS AND METHODS Materials. All chemicals were purchased from SigmaAldrich and used without purification. The fully aromatic low pressure polyamide (PA) flat sheets of ESPA1 membrane (Hydranautics) were kindly supplied by the manufacturers; membrane elements were purchased from Hydranautics. Before testing flat sheet coupons were first washed in 50% water− ethanol solution in ultrasonic bath for 15 min to ensure complete pore wetting, and elements were washed with deionized water. Membrane Testing. Laboratory-scale modification and filtration tests before and after modification were performed in 165 mL cylindrical stainless steel dead-end cells of circular wet B

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 1. Effect of added surfactant and monomer concentration on degree of grafting. (a) Triton X-100 with 2 mM GMA, (b) SDS with 2 mM GMA, (c) DTAB with 2 mM GMA, (d) GMA added to 0.09 mM Triton. Reaction solution also contained 0.003 mM MBAA, 4 mM K2S2O8 and 2 mM K2S2O5. Modification conditions: filtration pressure 20 bar, filtration time 30 min. Filled and cross-hatched bars correspond to modifications with and without added Triton, respectively. The DG based on the 907 cm−1 band is shown only when it exceeded the lower limit 0.05.

Membrane Characterization. ATR-FTIR spectra of pristine and modified membranes were recorded (average of 64 scans at 4 cm−1 resolution) on a Nicolet 8700 FTIR spectrometer (Thermo-Electron Co) equipped with a MIRacle ATR diamond smart accessory (Pike). The degree of grafting (DG) was determined as

Modification was carried out for 30 min; at the end, without disassembling the cells, membranes were washed several times with deionized water (DW) and soaked in DW overnight before testing the NaCl passage. Thereafter the membranes were unmounted, dried and examined by attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy, contact angle, scanning electron microscopy (SEM) and XPS. Modification of 2521 ESPA-1 elements was carried out at a separate setup using 8 L of GMA solution with added Triton. The solution was filtered through the element at a feed pressure 20 bar for 5 min. The feed flow rate was kept at ∼130 L/h, the permeate recovery was 30−40%. After stabilization the crosslinker and initiators dissolved separately in DW at concentrations similar to dead end experiments were added in the feed tank to start the modification, which was carried out at temperature 24 °C for 30 min. Afterward, the monomer solution was discarded and the membrane was washed by circulating DW several times, then thoroughly washed with DW for 24 h before testing performance. Tangential flow velocity during modification of elements (about 1.8 cm/s) was much lower than in typical operations (about 20 cm/s), in order to enhance concentration polarization and surface grafting reaction. In addition, in contrast to the previous study where the recovery was 100%, i.e., the retentate line closed,4 in this study the retentate flow was allowed (about 2/3 of the feed flow). Recirculation of the retentate containing the monomer helped avoid a strong variation of monomer concentration and resulted in a more uniform modification along the module. After completing modification and all performance tests, the modules were washed first with dilute NaOH solution (pH = 10, 25 min, 4 bar) and then with DW several times and cut open for membrane surface characterization using ATR-FTIR, SEM, contact angle, and XPS.

DG =

A mon A mem

(2)

Amon is the area of the characteristic band of acrylic monomers and polymers, either the 1724−1730 cm−1 band of the carbonyl group or 907 cm−1 band of epoxy group of GMA. Amem is the intensity of the 1586 cm−1 bands of polysulfone (part of the original membrane).44 (Representative ATR-FTIR spectra are presented in the Supporting Information.) Surface topography was examined by HR-SEM using a Zeiss Ultra-Plus SEM microscope. Prior to imaging the samples were coated with carbon. Contact angle was measured by sessile drop method using DSA 100 instrument (Kruss). XPS measurements were conducted using ESCALAB 250 (Thermo Fisher Scientific Inc., Waltham, UK) with Al X-ray source and monochromator. Dynamic light scattering (DLS) measurements were performed using a Zetasizer Nano ZSP (Malvern Instruments).

3. RESULTS AND DISCUSSION Type of Surfactant Matters. In order to understand the effect of the surfactant type on grafting, different concentrations of Triton X-100, a nonionic surfactant, sodium dodecyl sulfate (SDS), an anionic surfactant, and dodecyltrimethylammonium bromide (DTAB) a cationic surfactant were added to 2 mM GMA solution for modifications. To verify consistency, two DG estimates based on two different IR bands of GMA, 1725 and 907 cm−1 were used. The carbonyl band at 1725 cm−1 was C

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 2. SEM images of (a) pristine reverse osmosis ESPA1 membrane, (b) ESPA1 membrane modified with 2 mM GMA, (c) ESPA1 membrane modified with 2 mM GMA with the addition of 0.09 mM Triton, and (d) membrane from ESPA1−2521 module modified with 2 mM GMA with the addition of 0.045 mM Triton.

half CMC (∼0.12 mM). Given there are ∼120 Triton molecules per micelle,52 each micelle will contain ∼2000 GMA molecules, assuming 100% solubilization of 2 mM GMA. On the other hand, when surfactant is 5 times CMC (∼1 mM of surfactant assembled in micelles), this number will decrease to ∼250 GMA molecules and GMA will be more diluted with Triton within micelles. (Note Triton molecule is about 4.5 times larger than GMA.) Parallels may be seen between this behavior and occurrence of an optimal surfactant concentration for preparing latexes with narrowest particle size distribution by emulsion polymerization.39 The above mechanism should in principle be similar for any surfactant that forms micelles and can solubilize GMA. This assumption was tested by replacing Triton with SDS. Due to repulsion of the negatively charge sulfate head-groups, SDS has a much higher CMC than Triton, 8−8.3 mM for pure SDS.53,54 Therefore, the range of SDS concentrations was extended to above CMC to ensure formation of micelles and GMA solubilization. The resulting degrees of grafting as a function of [SDS] are shown in Figure 1b. Clearly, there was no increase in degree of grafting within the employed range of concentration. In fact, DG was lower than without added surfactant, regardless of whether [SDS] was below, close or above CMC. One possible reason for the negative impact of SDS on degree of grafting could be that, relative to Triton, the larger surfactant concentrations resulted in a much less GMA per micelle and much smaller GMA:surfactant ratio. An estimate analogous to that for Triton shows that for a SDS concentration 50% above its CMC, a micelle containing ∼80 SDS molecules42 will contain just ∼40 GMA molecules, that is, ∼ 50 times less than for Triton and the monomer will be strongly diluted within micelles. Furthermore, the average diameter of Triton micelle obtained in DLS is about 7.5−10 nm (see Supporting Information), while for SDS it is about 1.8 nm.55 Such a

relatively strong and reliably measured for all samples, however, the 907 cm−1 band of the epoxy group was relatively weak and overlapped with another weak band of the pristine membranes and was only measurable for the highest DGs. When that was the case (roughly, for DG (907) > 0.05), the latter band was used to verify that both bands yielded consistent trends, unaffected by noise and other possible artifacts (see Figure 1). Figure 1a shows that dependence of DG on Triton concentrations was nonmonotonic and highest DGs were achieved between 0.09 and 0.14 mM Triton. Apparently, the increase in DG when Triton was below this optimal range manifests onset of micellization and solubilization of GMA. Notably, this optimal range was slightly lower than the critical micelle concentration (CMC) of Triton (∼0.24 mM45−47). One possibility is that presence of GMA in solution might affect micellization of Triton, as solubilizable hydrophobic solutes are known to reduce CMC of surfactants.42,48−51 We however could not detect a significant shift in CMC of Triton at 2 mM GMA using DLS and only a slight change was found for such system after GMA was let polymerize for 30 min following addition of initiators (see Supporting Information for more details). A more likely reason is then the effect of CP, whereby Triton concentration near surface is increased thus it may surpass its CMC and effectively solubilize GMA next to the membrane surface, while in solution it was still well below CMC. However, when the surfactant concentration increased above the optimal range, DG dropped. Apparently, in this case the concentration of micelles next to the surface increased while GMA concentration stayed constant, thereby the number of monomers solubilized within each micelle decreased. This best explains occurrence of an optimal Triton concentration, close to CMC, at which the number of monomers solubilized in a micelle is at maximum. For instance, a simple estimate may consider a surfactant concentration just 50% above CMC, i.e., the concentration of surfactant assembled in micelles of about D

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Table 2. Contact angle (deg.) and Elemental Ratios Obtained by XPS for ESPA1 Membranes (From Flat Sheets and 2521 Elements) and Modified with GMA Membranes, Without and with Triton Addition (Triton Concentration Was 0.09 mM in All Samples except for ESPA1-2521 Module Modified With 2 mM GMA, Where Triton Concentration Was 0.045 mM) contact angle (deg.)

a

ESPA1 sample

modification solution

flat sheet flat sheet flat sheet flat sheet element element element

not modified 0.5 mM GMA 1.5 mM GMA 2.5 mM GMA not modified 1.5 mM GMA 2 mM GMA

no triton 28 25 33 35 26

± ± ± ± ±

5 1 1 1 1

C:O

triton added

no triton

N:O triton added

3.9 24 ± 1 38 ± 9 50 ± 2 45 ± 1 57 ± 2

2.8a

no triton

triton added

0.6 2.9a 2.4a 2.1 2.3

0.16a

0.09a 0.05a 0.14 0.2

Samples washed with DW water at 60 °C.

Surface Characterization. Further insight into the effect of added Triton to GMA grafting could be gained from examining the morphology and surface characteristics of modified membranes. Figure 2 compares SEM images of membranes modified without and with added surfactant. The typical ridgeand-valley structure of the pristine membrane is clearly seen in Figure 2a. Modification with GMA without added Triton resulted in inhomogeneous grafting, with polymer filling the valleys on the membrane surface, producing DG = 0.58 based on 1725 cm−1 band of GMA (Figure 2b). Added surfactant resulted in a more uniform coating with a higher degree of grafting (DG = 2.04), which completely covered and hid the ridge-and-valley structure of the original ESPA1 membrane (Figure 2c). In addition to ATR-FTIR and SEM, more surface characteristics were gained through contact angle (CA) and XPS. Table 2 displays the changes in CA and surface elemental composition as a function of GMA concentration for membranes in dead-end cells. Note that XPS has a much smaller penetration depth than ATR-FTIR, therefore it may help reveal whether the pristine surface was fully covered by examining C:O and N:O ratios. It was expected that the contact angle would increase with increasing DG, since the poly-GMA coating is more hydrophobic than the original membrane surface. However, it can be seen that, without added Triton, the change was small and insignificant, indicating that the modification was apparently not uniform and a significant fraction of original surface might remain exposed, especially, for lower GMA concentrations. However, with Triton added at near-optimal concentration a higher DG and a large and consistent increase in contact angle were obtained (cf. DG values in Figure 1d). Apparently, Triton helped form a more uniform coating, which is also seen in the SEM images in Figure 2c. The more uniform and complete surface coverage with added Triton brings the value of the contact angle closer to that of the coating. Elemental ratios of C:O and N:O by XPS for pristine membranes, ∼4 and 0.6, are usually slightly lower than ideal values for aromatic polyamide (6 and 1, respectively) due to incomplete cross-linking,2 yet they are significantly higher than the theoretical values for the poly-GMA, 2.3 and 0, respectively. The XPS data for selected flat sheet samples in Table 2, all after 2 h wash at 60 °C to simulate a long exposure, indeed show much lower ratios after modification, especially, when Triton was added. While for modifications using 1.5 mM GMA the final C:O ratio was close to the poly-GMA, both with and without Triton, the more indicative N:O ratio dropped more with added Triton. When GMA concentration was increased to

difference may yield a larger polymerization degree within a micelle and a higher CP for Triton. However, even for the smallest SDS micelles the diffusivity (∼10−10 m2/s as predicted by Einstein-Stokes equation) hence the mass transfer coefficient will be so low that CP is likely to lead to micelle deposition on the surface in the form of a dense layer. In such case micelles composition (estimated above) will determine the monomer content in this layer and, ultimately, DG, which is obviously superior in the case of Triton. The negative charge and repulsion between membrane and SDS micelles, as opposite to neutral Triton, could presumably play some role in its low grafting efficiency as well. Although the experiments with SDS were conducted at pH ∼ 3.1−3.4, close to the isoelectric point of ESPA1 membrane (pI ∼ 3.6, according to streaming potential measurements), repulsion between micelles could still prevent their concentration next to the surface. To rule out such interference, SDS was replaced with a cationic surfactant DTAB, having a CMC ∼ 15.4 mM,56 close to that of SDS, and somewhat larger micelle diameter (3.5−6.5 nm).46,57 Experiments were performed above pI at pH 8.1−8.7, when ESPA-1 is negatively charged. Despite the fact that a strong attraction between DTAB micelles and membrane was expected in this case, results in Figure 1c show a picture similar to SDS. From well below to well above CMC of DTAB, DG monotonically decreased with increasing DTAB concentrations, similar to SDS. This indicates the main factor responsible for enhancing grafting is the monomer concentration within micelles, maximized by a low CMC of surfactant, rather than surfactant’s charge or micelle size. Addition of Triton X-100 to GMA monomer solution also seems to improve coating homogeneity and/or reduce the monomer consumption. Solid bars in Figure 1d display the degree of grafting (DG) obtained for different concentrations of GMA after adding 0.09 mM Triton, which was optimal for 2 mM GMA. For comparison, cross-hatched bars show the degrees of grafting for the same GMA concentrations without added Triton. It can be seen that at the higher range of GMA concentrations there was an increase in DG with Triton added. Clearly, surfactant reduces the monomer concentration required to achieve a certain DG for 1.5, 2, and 2.5 mM GMA. However, for lower GMA, 0.5 and 1 mM, the effect of Triton was less significant. Lack of enhancement could be related to the polymerization kinetics being much slower, since DG was much lower for these GMA concentrations regardless of surfactant (note log scale of DG). However, this could also reflect the effect of GMA on CMC of Triton (see above) thus 0.09 mM Triton could become suboptimal and could not enhance grafting for lower GMA concentrations. E

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology 2.5 mM with added Triton, the elemental ratios became still lower and closer to theoretical values, indicating a nearly complete surface coverage with poly-GMA. Some residual N likely reflects the relatively small thickness of the coating. A quantitative method developed recently58 estimates that the average coating thickness was in the range 25 to 400 nm for the flat sheet samples tested by XPS in Table 2. Such thickness was well above penetration depth of XPS,59 yet irregular morphology in Figure 2 suggests that at some locations it could well be lower. Upscaling to Commercial Elements. The final step of this study was to demonstrate up-scalability of the optimal procedure, including added Triton, as devised from dead-end experiments. Since commercial LPRO elements operate in the cross-flow rather than dead-end mode, appropriate adjustments were necessary (see Materials and Methods). Thus, opposite to the previous work performed in dead-end mode (100% recovery),4 we allowed for some retentate recirculation at a 35% recovery rate to prevent retentate getting excessively concentrated toward the exit. The downside is a smaller polarization hence smaller grafting, however, more uniform conditions produced a more uniform modification. Figure 3 shows the results of autopsy of ESPA1 elements after modification at different locations relative to the entry and

Figure 4. Boric acid passage and permeability of commercial BWRO elements as reported in the literature (squares), pristine ESPA1 flat sheet coupons and elements tested in our lab (triangles), and ESPA1 elements modified using 1.5 and 2 mM GMA with Triton addition (circles). The two “high rejection” commercial elements are also shown (diamonds).

focus on B comes from the motivation to improve this parameter for BW desalination without impairing salt rejection and permeability, which is highly challenging and for which benefits of GMA coating and good coating stability were demonstrated previously.3 Since B passage is generally correlated with water permeability, the presentation in Figure 4 highlights the trade-off between the two. Note that small coupons in laboratory cells usually perform more poorly than elements (cf. pristine coupons in Figure 4), therefore for consistency modified membranes are represented only by elements. It is seen that compared to nonmodified elements, the modified modules show a consistently reduced boric acid passage, even though the permeability drops as well. Nevertheless, while compared with most BWRO elements of similar permeability, the modified ESPA1 elements have a superior B rejection. Modifications also resulted in a significant reduction of NaCl passage, to 0.25−0.5%, compared to the typical 0.5− 1% values of BWRO or 0.5−0.7% values of pristine ESPA1 modules (see Supporting Information), consistent with previous studies.2−4,64 As previously, we attribute the improved rejection to better utilization of the polyamide properties by sealing less selective areas (“defects”) in the top layer3,19 and, possibly, actual defects introduced during element manufacturing, e.g., during glueing. Curiously, the defect-blocking effect of modification may be compared to a similar effect of fouling,64 which is obviously too unpredictable and variable to be an alternative to permanent modification. Notably, “high-rejection” commercial elements in Figure 4, HRLE 440i and BW30-440i that might be modified during manufacturing,65,66 show a similar performance. However, the present in situ modification may still have advantages in cases where the performance of already installed and underperforming elements may need to be improved, tuned, or restored. Based on the defect-blocking mechanism, the method is inherently more suitable for open-type membranes, such as LPRO or BWRO. Surface characteristics obtained after autopsy of modified membrane after about 1 week of testing confirm the desired change. The surface morphology (Figure 2d) was similar to membranes modified in dead-end cells (Figure 2b and c), but

Figure 3. Degree of grafting based on 1725 cm−1 for ESPA1−2521 elements modified with GMA at different relative locations along (distance form feed inlet) and across (different distances from the permeate-collecting tube) the element. Modification conditions: first module modified with 2 mM GMA with 0.045 mM Triton, and second module modified 1.5 mM GMA with 0.09 mM GMA; 0.003 mM MBAA-cross-linker, 20 bar, 30 min.

permeate edge. Degrees of grafting were determined based on the 1725 cm−1 band only, as the 907 cm−1 band was under the detection limit. DGs within elements were lower, that is, coating was thinner, than in dead-end cell experiments. Surprisingly, DG was higher for 1.5 mM GMA, however the standard deviation was also high. It seems that in this case the 1725 cm−1 band may not represent only the carbonyl band of GMA and it can be partly contributed by other part of the membrane, for example, the polyamide layer. The latter is known to contain some carboxylic groups showing the same band60 and some groups could also be formed by hydrolysis of a fraction of amide bonds following washing the module with NaOH solution at pH ≈10 in the end of the experiment. However, in either case DG appears to be relatively uniform, both along and across the modules. The performance of modified modules in terms of B passage and permeability and comparison with other data for commercial brackish water (BW) RO elements compiled from literature reports12,61−63 is shown in Figure 4. The F

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology the coating filling the ridge-and-valley structure seems to be thinner than for flat sheets. The thinner coating, as compared to dead-end results, presumably resulting from a different flow regime and smaller CP, is also consistent with the smaller DG (cf. Figure 1) and larger N:O ratio (Table 2). Yet the larger contact angles and low C:O ratios in Table 2 suggest a good coverage of the underlying polyamide surface. In summary, we have seen that surfactant addition may assist in modifying polyamide membranes by CP-enhanced radical graft-polymerization, improve coating uniformity and reduce monomers consumption. The beneficial effect of the surfactant could be attributed to solubilization and concentration of GMA within surfactant micelles and facile adherence of neutral micelle to the membrane surface with subsequent monomer or polymer release. This mechanism is consistent with existence of an optimum Triton concentration range, which was absent in the case of the surfactants with a high CMC (SDS and DTAB). This points to the low CMC of the surfactant, rather than its charge or micelle size, as the dominant factor in grafting enhancement. We also show the process can be straightforwardly up-scaled to commercial RO elements. Surface examination shows that a stable thin and uniform coating could be obtained of membranes within commercial elements, resulting in improved selectivity-permeability trade-off compared to most nonmodified commercial elements. The proposed method may be a promising way for in situ upgrading, tuning or restoring membrane performance in industrial installations.



(2) Bernstein, R.; Belfer, S.; Freger, V. Surface modification of dense membranes using radical graft polymerization enhanced by monomer filtration. Langmuir 2010, 26 (14), 12358−12365. (3) Bernstein, R.; Belfer, S.; Freger, V. Toward improved boron removal in RO by membrane modification: Feasibility and challenges. Environ. Sci. Technol. 2011, 45 (8), 3613−3620. (4) Bernstein, R.; Belfer, S.; Freger, V. Improving performance of spiral wound RO elements by in situ concentration polarizationenhanced radical graft polymerization. J. Membr. Sci. 2012, 405−406, 79−84. (5) Borokhov Akerman, E.; V. Simhon, M.; Gitis, V. Advanced treatment options to remove boron from seawater. Desalin. Water Treat. 2012, 46 (1−3), 285−294. (6) Wolska, J.; Bryjak, M. Methods for boron removal from aqueous solutions  A review. Desalination 2013, 310, 18−24. (7) Fujioka, T.; Oshima, N.; Suzuki, R.; Price, W. E.; Nghiem, L. D. Probing the internal structure of reverse osmosis membranes by positron annihilation spectroscopy: Gaining more insight into the transport of water and small solutes. J. Membr. Sci. 2015, 486, 106− 118. (8) Arias, M. F. C.; Bru, L. V.; Rico, D. P.; Galvan, P. V. Approximate cost of the elimination of boron in desalinated water by reverse osmosis and ion exchange resins. Desalination 2011, 273 (2−3), 421− 427. (9) Tu, K. L.; Nghiem, L. D.; Chivas, A. R. Boron removal by reverse osmosis membranes in seawater desalination applications. Sep. Purif. Technol. 2010, 75 (2), 87−101. (10) Greenlee, L. F.; Lawler, D. F.; Freeman, B. D.; Marrot, B.; Moulin, P. Reverse osmosis desalination: Water sources, technology, and today’s challenges. Water Res. 2009, 43 (9), 2317−2348. (11) Oo, M. H.; Song, L. Effect of pH and ionic strength on boron removal by RO membranes. Desalination 2009, 246 (1−3), 605−612. (12) Prats, D.; Chillon-Arias, M. F.; Rodriguez-Pastor, M. Analysis of the influence of pH and pressure on the elimination of boron in reverse osmosis. Desalination 2000, 128 (3), 269−273. (13) Sagiv, A.; Semiat, R. Analysis of parameters affecting boron permeation through reverse osmosis membranes. J. Membr. Sci. 2004, 243 (1−2), 79−87. (14) Kabay, N.; Bryjak, M.; Hilal, N.; Yoshizuka, K.; Nishihama, S. Boron Separation Processes 2015, 219. (15) Bellona, C.; Drewes, J. E.; Xu, P.; Amy, G. Factors affecting the rejection of organic solutes during NF/RO treatment - A literature review. Water Res. 2004, 38 (12), 2795−2809. (16) Kimura, K.; Toshima, S.; Amy, G.; Watanabe, Y. Rejection of neutral endocrine disrupting compounds (EDCs) and pharmaceutical active compounds (PhACs) by RO membranes. J. Membr. Sci. 2004, 245 (1−2), 71−78. (17) Schäfer, A. I.; Nghiem, L. D.; Waite, T. D. Removal of the natural hormone estrone from aqueous solutions using nanofiltration and reverse osmosis. Environ. Sci. Technol. 2003, 37 (1), 182−188. (18) Nghiem, L. D.; Schäfer, A. I. Critical risk points of nanofiltration and reverse osmosis processes in water recycling applications. Desalination 2006, 187 (February 2005), 303−312. (19) Ben-David, A.; Bernstein, R.; Oren, Y.; Belfer, S.; Dosoretz, C.; Freger, V. Facile surface modification of nanofiltration membranes to target the removal of endocrine-disrupting compounds. J. Membr. Sci. 2010, 357 (1−2), 152−159. (20) Plakas, K. V.; Karabelas, A. J. Removal of pesticides from water by NF and RO membranes  A review. Desalination 2012, 287, 255− 265. (21) Radjenović, J.; Petrović, M.; Ventura, F.; Barceló, D. Rejection of pharmaceuticals in nanofiltration and reverse osmosis membrane drinking water treatment. Water Res. 2008, 42 (14), 3601−3610. (22) Yüksel, S.; Kabay, N.; Yüksel, M. Removal of bisphenol A (BPA) from water by various nanofiltration (NF) and reverse osmosis (RO) membranes. J. Hazard. Mater. 2013, 263 (Pt 2), 307−310. (23) Singh, P. S.; Joshi, S. V.; Trivedi, J. J.; Devmurari, C. V.; Rao, A. P.; Ghosh, P. K. Probing the structural variations of thin film composite RO membranes obtained by coating polyamide over

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.6b05260. DLS results for Triton micellization and micelle size, representative ATR-FTIR spectra of pristine and modified flat sheets membranes, and comparison of NaCl passage of ESPA1 elements before and after modification (PDF)



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Viatcheslav Freger: 0000-0001-8067-052X Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by Israel Water Authority grant No 4500687020; KBK acknowledges the scholarship by the Israel Ministry of Science Technology and Space for the Israeli Arab Ph.D. students and the Rieger Foundation-Jewish National Fund for the Fellowship in Environmental Studies. Finally, we acknowledge Miguel Tirrado’s inspiring mistake that motivated this study and Roy Bernstein’s help with XPS.



REFERENCES

(1) Freger, V. Swelling and morphology of the skin layer of polyamide composite membranes: An atomic force microscopy study. Environ. Sci. Technol. 2004, 38 (11), 3168−3175. G

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology polysulfone membranes of different pore dimensions. J. Membr. Sci. 2006, 278 (1−2), 19−25. (24) Kolev, V.; Freger, V. Hydration, porosity and water dynamics in the polyamide layer of reverse osmosis membranes: A molecular dynamics study. Polymer 2014, 55 (6), 1420−1426. (25) Belfer, S.; Purinson, Y.; Kedem, O. Surface modification of commercial polyamide reverse osmosis membranes by radical grafting. An ATR-FTIR study 1998, 574−582. (26) Belfer, S.; Purinson, Y.; Fainshtein, R.; Radchenko, Y.; Kedem, O. J. Membr. Sci. 1998, 139, 175−181. (27) Belfer, S.; Gilron, J.; Purinson, Y.; Fainshtain, R.; Daltrophe, N.; Priel, M.; Tenzer, B.; Toma, a. Effect of surface modification in preventing fouling of commercial SWRO membranes at the Eilat seawater desalination pilot plant. Desalination 2001, 139 (1−3), 169− 176. (28) Belfer, S.; Fainshtain, R.; Purinson, Y.; Gilron, J.; Nyström, M.; Mänttäri, M. Modification of NF membrane properties by in situ redox initiated graft polymerization with hydrophilic monomers. J. Membr. Sci. 2004, 239 (1), 55−64. (29) Freger, V.; Gilron, J.; Belfer, S. TFC polyamide membranes modified by grafting of hydrophilic polymers. J. Membr. Sci. 2002, 209, 283−292. (30) Kang, G.; Cao, Y. Development of antifouling reverse osmosis membranes for water treatment: A review. Water Res. 2012, 46 (3), 584−600. (31) Van Wagner, E. M.; Sagle, A. C.; Sharma, M. M.; La, Y.-H.; Freeman, B. D. Surface modification of commercial polyamide desalination membranes using poly(ethylene glycol) diglycidyl ether to enhance membrane fouling resistance. J. Membr. Sci. 2011, 367 (1− 2), 273−287. (32) Groups, E.; Copolymers, M. Reactions of Epoxide Groups of Glycidyl. 1974, 166 (47), 155−166. (33) Nabe, A.; Staude, E.; Belfort, G. Surface modification of polysulfone ultrafiltration membranes and fouling by BSA solutions. J. Membr. Sci. 1997, 133 (1), 57−72. (34) Sagle, A. C.; Van Wagner, E. M.; Ju, H.; McCloskey, B. D.; Freeman, B. D.; Sharma, M. M. PEG-coated reverse osmosis membranes: Desalination properties and fouling resistance. J. Membr. Sci. 2009, 340 (1−2), 92−108. (35) Rangel-Yagui, C. D. O.; Pessoa, A.; Tavares, L. C. Micellar solubilization of drugs. J. Pharm. Pharm. Sci. 2005, 8 (2), 147−163. (36) Kumar, Y.; Popat, K. M.; Brahmbhatt, H.; Ganguly, B.; Bhattacharya, a. Pentachlorophenol removal from water using surfactant-enhanced filtration through low-pressure thin film composite membranes. J. Hazard. Mater. 2008, 154 (1−3), 426−431. (37) Tehrani-Bagha, A. R.; Holmberg, K. Solubilization of hydrophobic dyes in surfactant solutions. Materials 2013, 6 (2), 580−608. (38) Jenkins, D. W.; Hudson, S. M. Review of vinyl graft copolymerization featuring recent advances toward controlled radical-based reactions and illustrated with chitin/chitosan trunk polymers. Chem. Rev. 2001, 101 (11), 3245−3273. (39) Chern, C. S. Emulsion polymerization mechanisms and kinetics. Prog. Polym. Sci. 2006, 31 (5), 443−486. (40) Seko, N.; Ninh, N. T. Y.; Tamada, M. Emulsion grafting of glycidyl methacrylate onto polyethylene fiber. Radiat. Phys. Chem. 2010, 79 (1), 22−26. (41) Sritapunya, T.; Kitiyanan, B.; Scamehorn, J. F.; Grady, B. P.; Chavadej, S. Wetting of polymer surfaces by aqueous surfactant solutions. Colloids Surf., A 2012, 409, 30−41. (42) Rosen, M. J.; Kunjappu, J. T. Surfactants and Interfacial Phenomena, 4th ed, 2012. (43) Http://www.lenntech.com/Data-sheets/Hydranautics-ESPA2521-L.pdf. Application data. 2005, 92058. (44) Bernstein, R.; Belfer, S.; Freger, V. Bacterial attachment to RO membranes surface-modified by concentration-polarization-enhanced graft polymerization. Environ. Sci. Technol. 2011, 45 (14), 5973−5980. (45) Tung, C.-C.; Yang, Y.-M.; Chang, C.-H.; Maa, J.-R. Removal of copper ions and dissolved phenol from water using micellar-enhanced

ultrafiltration with mixed surfactants. Waste Manage. 2002, 22 (7), 695−701. (46) Surfactant micelle characterization using dynamic light scattering. Zetasizer Nano application note (2006), 1−5. (47) Lin, S.-Y.; Mckeigue, K.; Maldarelli, C. Diffusion-Controlled Surfactant Adsorption Studied by Pendant Drop Digitization. AIChE J. 1990, 36 (12), 1785−1795. (48) Mukerjee, P. Solubilization in Aqueous Micellar Systems. In Solution Chemistry of Surfactants: Vol. 1; Mittal, K. L., Ed.; Springer New York: Boston, MA, 1979; pp 153−174. (49) Zana, R. Aqueous surfactant-alcohol systems: A review. Adv. Colloid Interface Sci. 1995, 57 (C), 1−64. (50) Rao, I. V.; Ruckenstein, E. Micellization behavior in the presence of alcohols. J. Colloid Interface Sci. 1986, 113 (2), 375−387. (51) Gu, T.; Galera-Gomez, P. A. The effect of different alcohols and other polar organic additives on the cloud point of Triton X-100 in water. Colloids Surf., A 1999, 147 (3), 365−370. (52) Biaselle, C. J.; Millar, D. B. Studies on triton X-100 detergent micelles. Biophys. Chem. 1975, 3 (4), 355−361. (53) Rahman, a.; Brown, C. W. Effect of pH on the Critical Micelle Concentration of Sodium Dodecyl Sulphate. J. Appl. Polym. Sci. 1983, 28 (1983), 1331−1334. (54) Bales, B. L.; Messina, L.; Vidal, A.; Peric, M.; Nascimento, O. R. Precision Relative Aggregation Number Determinations of SDS Micelles Using a Spin Probe. A Model of Micelle Surface Hydration. J. Phys. Chem. B 1998, 102 (50), 10347−10358. (55) Almgren, M.; Swarup, S. Size of sodium dodecyl sulfate micelles in the presence of additives i. alcohols and other polar compounds. J. Colloid Interface Sci. 1983, 91 (1), 256−266. (56) Mata, J.; Varade, D.; Bahadur, P. Aggregation behavior of quaternary salt based cationic surfactants. Thermochim. Acta 2005, 428 (1−2), 147−155. (57) Pisárčik, M.; Devínsky, F.; Švajdlenka, E. Spherical dodecyltrimethylammonium bromide micelles in the limit region of transition to rod-like micelles. A light scattering study. Colloids Surf., A 1996, 119 (2−3), 115−122. (58) Bass, M.; Freger, V. Facile evaluation of coating thickness on membranes using ATR-FTIR. J. Membr. Sci. 2015, 492, 348−354. (59) Bernstein, R.; Kaufman, Y.; Freger, V. Membrane characterization. Encycl. Membr. Sci. Technol. 2013, 41. (60) Gershevitz, O.; Sukenik, C. N. In Situ FTIR-ATR Analysis and Titration of Carboxylic Acid-Terminated SAMs. J. Am. Chem. Soc. 2004, 126 (2), 482−483. (61) Xu, P.; Drewes, J. E.; Heil, D. Beneficial use of co-produced water through membrane treatment: technical-economic assessment. Desalination 2008, 225 (1−3), 139−155. (62) Redondo, J.; Busch, M.; De Witte, J. P. Boron removal from seawater using FILMTEC high rejection SWRO membranes. Desalination 2003, 156 (1−3), 229−238. (63) Murray-Gulde, C.; Heatley, J. E.; Karanfil, T.; Rodgers, J. H.; Myers, J. E. Performance of a hybrid reverse osmosis-constructed wetland treatment system for brackish oil field produced water. Water Res. 2003, 37 (3), 705−713. (64) Van Wagner, E. M.; Sagle, A. C.; Sharma, M. M.; Freeman, B. D. Effect of crossflow testing conditions, including feed pH and continuous feed filtration, on commercial reverse osmosis membrane performance. J. Membr. Sci. 2009, 345 (1−2), 97−109. (65) Mickols, W. E. Composite membrane with polyalkylene oxide modified polyamide surface. United States Pat. No.6,280,853 B1 2001, 1 (12). (66) Niu, Q. J.; Mickols, W. E.; Zhang, C. Modified polyamide membrane. United States Pat. No.7,905,361 B2 2011, 2 (12).

H

DOI: 10.1021/acs.est.6b05260 Environ. Sci. Technol. XXXX, XXX, XXX−XXX