Solute and Water Transport in Forward Osmosis Using Polydopamine Modiﬁed Thin ﬁlm Composite Membranes

Desalination 343 (2014) 8–16

Contents lists available at ScienceDirect

Desalination

journal homepage: www.elsevier.com/locate/desal

Solute and water transport in forward osmosis using polydopamine modiﬁed thin ﬁlm composite membranes

Jason T. Arena a, Seetha S Manickam a,KevinK.Reimunda, Benny D. Freeman b, Jeffrey R. McCutcheon a,⁎ a University of Connecticut, Department of Chemical and Biomolecular Engineering, Center for Environmental Sciences and Engineering, Storrs, CT 06269, USA b The University of Texas at Austin, Department of Chemical Engineering, Austin, TX 78712, USA

HIGHLIGHTS

• Reverse osmosis membrane support layers were modiﬁed with polydopamine. • The modiﬁed membranes were tested under forward osmosis conditions. • An ammonia–carbon dioxide draw solution was used for the testing. • Sodium ion rejection was found to be low, while chloride ion rejection was high. • Differing ion rejection is an evidence of ion exchange between the feed and draw.

article info abstract

Article history: Forward osmosis is a rapidly emerging technology that has potential to enable low cost water treatment and Received 5 August 2013 desalination. Previous investigations have found that reverse osmosis (RO) membranes were unsuitable for Received in revised form 4 December 2013 forward osmosis in part due to their hydrophobic support layers, which inhibit wetting. Poor wetting hin- Accepted 2 January 2014 ders water and solute transport in the support layer, dramatically increasing the severity of internal concen- Available online 2 April 2014 tration polarization. In this study, RO membrane support layers were modified with polydopamine (PDA) to increase their hydrophilicity and promote wetting. The results indicate that the modified RO membranes Keywords: fl Forward osmosis exhibited a four to six fold increase in forward osmosis (FO) water ux under test conditions relative to Pressure retarded osmosis unmodified membranes. Additional tests were performed under model desalination conditions using an Thin film composite membrane ammonia–carbon dioxide draw solution with a sodium chloride feed. The sodium and chloride rejections Polydopamine were measured independently and in some instances substantial differences were observed. Additionally sodi- Membrane modification um and chloride rejections were lower than anticipated with a peak rejection of 90%. The substantial difference between sodium and chloride rejections was attributed to a cationic exchange effect between the draw and feed solutions. © 2014 Elsevier B.V. All rights reserved.

1. Introduction processes [12]. The CA membrane while offering acceptable permselectivity and desirable hydrophilicity has inherent chemical Forward osmosis (FO) is an emerging process being considered compatibility drawbacks, notably hydrolysis in alkaline conditions [13– for the desalination, puriﬁcation, and treatment of water [1–6].A 15]. Hydrolysis reduces salt rejection, which in FO translates to higher functional FO process requires an easily recoverable draw solution draw solute cross-over and a lower osmotic pressure difference across capable of generating high osmotic pressures as well as a highly pro- the membrane [14,15].TheNH3–CO2 draw solution will hydrolyze CA ductive and selective membrane [1,4,7]. Various draw solutes exist, as this draw solution can be expected to have pHs above 7.7 [16,17]. but only the ammonia–carbon dioxide (NH3–CO2)drawsolution This leads to the consideration of alternative membrane chemistries has been demonstrated as both an effective and recyclable solute for use with the NH3–CO2 draw solution. The commercial alternative that may enable osmotically driven desalination [1,4,7–10].Amongst to the CA membranes is the thin ﬁlm composite (TFC) membrane the most commonly studied membrane for forward osmosis is the platform. These membranes, typically used in reverse osmosis, com- asymmetric cellulose acetate (CA) manufactured by Hydration Tech- prise an ultra-thin aromatic polyamide layer supported by a polysulfone nology Innovations™ (HTI) [1,4,9–12,16]. This membrane's morpholo- (PSu) or polyethersulfone (PES) layer that has been cast onto a polyes- gy has been optimized for use in osmotically driven membrane ter (PET) nonwoven [18]. Each of these layers is capable of withstanding a broad range of pH and temperature conditions making them suitable – ⁎ Corresponding author. Tel.: +1 860 486 4601; fax: +1 860 486 2959. for use with the NH3 CO2 draw solution. Despite these desirable charac- E-mail address: [email protected] (J.R. McCutcheon). teristics for use FO processes early studies which attempted to use TFC

0011-9164/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.desal.2014.01.009 J.T. Arena et al. / Desalination 343 (2014) 8–16 9 membranes in FO found the performance of TFC RO membranes to be in a series of three deionized water baths for 45 min each. Following the inferior to that of HTI's CA FO membrane [1,2].Inlaterwork,thelack IPA wetting and DI water rinsing, the membranes were stored in DI of TFC support layer wetting was demonstrated as a hindrance to os- water at 4 °C before being modified with PDA. The dopamine polymer- motic flux due to a reduced effective porosity and enhanced internal ization took place within a custom built coating container where the concentration polarization (ICP) [19,22]. To address this problem the membrane separates two reservoirs. This container ensures that nearly use of TFC membranes with an intrinsically hydrophilic support all of the PDA polymerizes within the PSu layer and not the selective would be desirable. This would require a retuning of the delicate in- layer (which would negatively impact permeability [26–28]). Both terfacial polymerization process, which can be impacted by the sup- sides of the membrane were placed in contact with a pH 8.8 Tris buffer port layer properties [21,22]. Furthermore, hydrophilic supports solution. Dopamine-HCl was added to the solution in contact with may plasticize in the presence of water and cause damage to the membranes' PSu support layers to bring the support layer coating solu- fragile selective layer. Ideally, one could start with a TFC membrane tion to a concentration of 2 g·L−1 dopamine. Polymerization occurs at made from a non-swelling hydrophobic support that also exhibits room temperature with non-agitated solutions exposed to the air. The good permselectivity; then modify that membrane's support layer PDA polymerization can be observed upon the addition of dopamine to increase its hydrophilicity. Recently commercial TFC FO mem- where the formation of PDA is indicated by the change in color of the branes have just begun to enter the market with limited availability, polymerizing dopamine solution from clear to orange and finally to with only HTI providing theirs for sale at the time of writing brown. [6,11,23,24]. A recently developed technique to impart a hydrophilic character 2.3. Mercury intrusion porosimetry onto microfiltration, ultrafiltration, and reverse osmosis membrane selective layers for enhanced fouling resistance to oil/water emulsions A mercury intrusion porosimeter (MIP) (AutoPoreIV, Micrometrics) and protein mixtures was reported by McCloskey and co-workers was used to characterize the membranes for pore diameter and total using polydopamine (PDA) [25–28]. PDA is a polymer with a chemistry pore volume. The Washburn equation was used to calculate the pore di- similar to the adhesive secretions of mussels [29–31]. It is formed from ameters from the intrusion pressure. the spontaneous polymerization of dopamine in an alkaline aqueous solution. A subsequent study by Arena examined the first use of PDA mod- −4γ cosðÞθ d ¼ : ð1Þ ified membranes for osmotically driven membrane process. This was P done through the application of PDA to TFC membrane support layer(s). Significant improvements in the water flux of PDA modified TFC RO In Eq. (1), P is the intrusion pressure (MPa), d is the pore diameter membranes were observed in the pressure retarded osmosis (PRO) ori- (μm), γ is the surface tension of mercury (485 dyn·cm−1)andθ is the entation [32]. Others, such as Han, adopted this technique prior to syn- contact angle of mercury (a value of 130° was assumed) with the sam- thesis of the membrane [33]. ple. The sample was tested in the pressure range of 1–720 bar. It is to be With the improved performance of these membranes in the PRO noted that Eq. (1) assumes that measured pore diameters are cylindri- mode, we hypothesized that a similar improvement would be possible cal. While this assumption is idealized for the membrane supports test- in the FO mode as well. The excellent selectivity of these membranes ed in this study, the resulting values for d calculated in Eq. (1) represent as well as tolerance to the often used ammonia–carbon dioxide draw the equivalent cylindrical pore diameters of the support. It is also to be solution make such a platform appealing. Membranes were tested for noted that the intrusion technique can detect both through and blind desalination performance using this draw solution in hopes of demon- pores but not closed pores [38]. strating the promise of these modified membranes; however, it was found that rejection, especially for cations, was far lower than anticipat- 2.4. Fourier transform infrared spectroscopy ed. This was attributed to an ion exchange phenomenon occurring through the polyamide selective layer. The modified TFC RO membranes were tested in Fourier transform infrared (FTIR) spectroscopy to examine the surface functional 2. Materials and methods groups of the membranes' selective layers. Membranes were tested, after drying, in a Thermo Scientific (Waltham, MA) Nicolet iS10 FTIR 2.1. Selected membranes and chemicals spectrophotometer with Smart iTR attachment was used to perform these measurements on a dried membrane. Measurements were taken The membranes selected for this investigation are the Dow Water & on the selective layer using 64 scans with a resolution of 4 cm−1. Process Solutions™ BW30 and SW30-XLE. Both membranes' support layers are made of PSu supported by a PET nonwoven [34]. These mem- 2.5. Osmotic flux testing of modified membranes branes were chosen for their high permselectivity, use in earlier studies, and reported properties [46]. Sodium chloride, tris–hydrochloride, sodi- 2.5.1. Sodium chloride as the draw solute um hydroxide, ammonium bicarbonate, and ammonium hydroxide Both neat and modified TFC RO membranes were tested under os- were purchased from Fisher Scientific (Pittsburgh, PA). Dopamine- motic flux conditions with the membrane oriented in the FO mode hydrochloride was purchased from Sigma-Aldrich (St. Louis, MO). (with the support layer facing the draw solution) [35]. Both membranes Isopropanol, sodium tetraphenyl boron, potassium chromate, calcium were tested in each of the four following varieties neat (used as- nitrate, and silver nitrate were purchased from Acros Organics (Geel, received), no PET (where the PET backing has been removed from the Belgium). Water used in this study was ultrapure Milli-Q water produce membrane), one hour PDA modified, and forty-two hour PDA modified by a Millipore Integral 10 water system (Millipore Corporation, Billerica, membranes (both modified using the method reported above). Mem- MA). branes not modified with PDA were tested following storage in DI water. Prior to testing no wetting technique was implemented. The 2.2. PDA modification of TFC membranes membrane area exposed to the feed and draw solutions were ap- proximately 19 cm2 (3 in.2). Sodium chloride (NaCl) was used as The PDA modification followed the procedure set forth in previous the draw solute at concentrations of 0.05, 0.1, 0.5, 1.0, and 1.5 M. work [32]. Since the PDA formation only occurs in the aqueous phase, The osmotic flux testing procedure has been described previously it was necessary to prewet the support in isopropanol (IPA) prior to [11,22,32,37,39]. Temperature was maintained at 23 ± 1 °C. Flux PDA modification. The support was soaked IPA for 1 hr and then washed was measured gravimetrically using a balance (Denver Instruments 10 J.T. Arena et al. / Desalination 343 (2014) 8–16

mass transfer limitations can be represented by the following equation [37]. 8 9 > J S J > <>π exp − w −π exp − w => d;b D f ;b k J ¼ A ð2Þ w > B J J S > : 1 þ exp w − exp − w ; Jw k D

−1 In Eq. (2), Jw is the water ﬂux (m·s ), A is the water permeance −1 −1 (m·s ·bar ), πd,b is the osmotic pressure of the draw solution (bar), S is the structural parameter (m), D is the solute diffusivity in water (m2·s−1), t is the thickness of the membrane support layer(s)

(m), πf,b is the osmotic pressure of the feed solution (bar), and B is the solute permeability (m·s−1). Osmotic pressures can be calculated using the van't Hoff equation [38]. The water permeance (A)andsolute permeability (B) are commonly determined using reverse osmosis [32]. The structural parameter can be determined from a numerical solution to Eq. (2) using experimental water flux data. The structural parameter Fig. 1. Porosity data from MIP of BW30 (solid) and SW30-XLE (cross-hatched) membranes is often defined as a function of support layer thickness (t), porosity (ε), (neat, no PET and PDA modified). and tortuosity (τ)(S = tτ / ε), and is representative of the effective diffusion distance through the support; however, rather than measuring PI-4002, Denver Instruments Bohemia, NY) connected to a computer each of these values individually (which can be difficult to do accurate- measuring the mass of the draw solution tank once per minute. The ly), fitting S to the model above using experimental data provides an “ef- osmotic pressures produced by these draw solutions (as presented fective structural parameter” since the approach accounts for poor in the figures) were calculated using the van't Hoff equation. Tests wetting in the support since these regions not available for solute trans- were run in triplicate using fresh membrane samples. Reverse solute port [19,21,32]. flux was monitored by measuring of the feed solution conductivity. 2.5.3. Ammonia–carbon dioxide draw solution PDA modified TFC membranes were tested for NaCl rejection in

2.5.2. Determination of the effective structural parameter forward osmosis using an NH3–CO2 based draw solution. These The structural parameter is a measure of the effective diffusive dis- tests were performed in a laboratory scale osmosis test systems tance of a solute through a porous media [5,11,35–37]. Solutes and using a 2.0 M draw solution with an ammonia to carbon dioxide water are assumed to be only capable of diffusing only through a wetted ratio of 1.2:1 on a molar basis and a feed solution of 0.25 M sodium pore, thus a lack of wetting can have a large impact on the effective chloride. These solutions were run counter-current with a cross structural parameter of a porous material [19,21,32]. The importance ﬂow velocity of 0.25 m·s−1 at 23 ± 1 °C, matching the testing conditions of the structural parameter is shown in the governing equation for using the NaCl draw solution. The membrane support layer was in con- water ﬂux in the FO orientation, which including feed solution external tact with the NH3–CO2 draw solution (FO mode). Experiments were

Fig. 2. Pore diameter histograms from MIP for of BW30 and SW30-XLE membranes (with PET removed and PDA modiﬁed). J.T. Arena et al. / Desalination 343 (2014) 8–16 11

Fig. 3. FTIR spectra of the selective layer for PDA modified commercial TFC membranes at wave numbers from 1800 to 600 cm−1. The peaks are consistent with those typically found for the fully aromatic TFC [45]. also run for a short time with the draw solution against a DI water feed to Waltham, MA) equipped with a sodium cathode lamp (Perkin-Elmer measure the pure water flux for the NH3–CO2 draw solution. Intensitron Part# 303-6065, Perkin-Elmer, Waltham, MA). Solutions Loss of draw solutes via permeation through the membrane neg- were analyzed using an air-acetylene flame with the detector set atively impacts the overall cost and efficiency of FO processes be- at 589 nm. Standard solutions were made with sodium chloride in cause draw solutes that are lost must be replaced after draw diluted ammonium bicarbonate solution at concentrations ranging solution recovery [6,40,41]. Unrecovered draw solutes may also con- from 2 to 12 ppm. The instrument was blanked against an ammonium taminate the brine complicating its disposal. The flux of ammonia bicarbonate draw solution with the same dilution factor as the sodium species (both as ammonia and ammonium) from the draw to the feed chloride-containing draw solution. Ammonium bicarbonate draw sam- solution was measured gravimetrically using sodium tetraphenyl ples were diluted to give an absorbance in the range of the standard boron as a precipitating agent [42,43]. A small sample of feed solution solutions. was removed from the feed tank and analyzed. When added to a solu- Chloride flux was determined from a mass balance based on the tion containing ammonia species, ammonium tetraphenyl borate is final concentration of chloride in the draw solution, which was de- formed and precipitates out of solution. This precipitate was captured termined using the Mohr titration [42]. In the Mohr titration chloride using fine porosity filter paper, washed with 1 °C DI water, dried, and is titrated with silver nitrate in the presence of a potassium chromate massed on an analytical balance (Denver Instruments PI-114, Denver indicator. At the end point of the titration excess silver ions form Instruments Bohemia, NY). Following filtration of the ammonium silver chromate producing a reddish brown color within the solution. tetraphenyl borate mixture a small amount of sodium tetraphenyl Due to the presence of bicarbonate in the draw solution being ana- boron was added to the filtered solution to ensure that all of the ammo- lyzed the solution was boiled to dryness prior to the titration to vol- nia species in solution were precipitated. atilize all of the ammonia and carbon dioxide within the solution. Sodium flux was determined from a mass balance based on the Following drying, the residual solutes were rehydrated and the final concentration of sodium in the draw solution, analyzed via atomic resulting solution was titrated. A complete validation of this tech- absorption spectroscopy in a Perkin-Elmer 3100 AA (Perkin-Elmer, nique is presented in the Supplementary material.

3. Results and discussion

3.1. Porosimetry characterization

Mercury intrusion porosimetry (MIP) was performed on both modified and unmodified membranes to examine the effect of the PDA modification on membrane support layer pore diameters and porosity. As shown in Fig. 1,theporosityforbothmembranetypes (i.e., BW30 and SW30-XLE) decreased as a result of polydopamine deposition, and the samples exposed to the dopamine coating solution for a longer time (i.e. 42 hr) had a lower porosity than those treated for only 1 hr. This decrease in porosity directly competes with the increased wettability as measured by contact angle goniometry as reported by Arena [32]. Fig. 2 presents the effective pore size distribution for the membranes considered in this study. There were minimal changes in the pore diameter distribution for membranes with higher coating times. These membranes exhibited a slight shift toward smaller pores, but given the thinness of PDA layers [26,44] the pore diameter distributions do not Fig. 4. FTIR spectra of the selective layer for both PDA modified and unmodified commer- − change dramatically. Care should be taken when scrutinizing MIP data cial TFC membranes at wave numbers from 3700 to 2700 cm 1. The peak from 3000 to 2800 is likely attributed to solid state hydrogen bonded hydroxyl stretch in the polyamide too closely as the high pressures employed by cause irreversible sample layer's carboxylic acid moieties [49]. compression and skew results; however, for comparative purposes the 12 J.T. Arena et al. / Desalination 343 (2014) 8–16

Fig. 5. Osmotic flux performance of BW30 and SW30-XLE membranes (neat, PET removed and PDA modified) at 23 ± 1 °C, 0.25 m/s feed and draw cross-flow velocity, and no transmembrane hydrostatic pressure. unmodified and modified membranes would deform similarly and so these charged groups may play a significant role in other transport pro- this technique is reasonable for comparing porosity changes. cesses during FO.

3.2. FTIR spectra 3.3. Osmotic flux performance The FTIR spectra for these membranes, shown in Figs. 3 and 4,are 3.3.1. Water flux for a sodium chloride draw solution characteristic for those membranes based upon a fully aromatic Fig. 5 shows that osmotic water flux was increased significantly polyamide [45]. The strong similarities in the FTIR spectra for the following modification of the BW30 and SW30-XLE membranes BW30 and SW30-XLE membranes aretobeexpectedgiventheir with PDA. PDA modification caused water flux to increase by up to common lineage stemming from the FT30 membrane originally de- a factor of 4 for the BW30 and up to a factor of 6 for the SW30-XLE veloped by Cadotte [46,47]. Also, based upon the FTIR spectra PDA membrane. This observation is similar to those reported previously, cannot be detected. This is unsurprising many of the functional where the PDA modified BW30 and SW30-XLE membranes exhibited groups characteristic of PDA are already present in an aromatic poly- an 8 and 12 fold increase in flux, respectively [32].The42hourPDA amide, which based upon the structure proposed by Dreyer consists modified membranes showed slightly decreased (but not statistical- of an indole or indoline like structure (containing a N–H), carbonyl ly significant) water flux when compared to the 1 hour PDA modified and hydroxyl functional groups [48]. Overall the application of PDA membrane. This can be explained to be a result of decrease porosity tothemembranesupportlayersdoesnotappeartosignificantly within the membrane support layers as shown in Fig. 1.Theincrease alter the surface functional groups of the membrane's selective layer. water flux for the PDA modified membranes may be attributed to the An interesting peak of the spectra (found in Fig. 4) is the 3000– increased wettability of membrane support layer increasing the rate 2800 cm−1 peak. This peak can be only attributed to a hydrogen of draw solute transport through the membrane support layer. This bonded hydroxyl stretch of a solid state carboxylic acid [49]. This will increase the concentration, and osmotic pressure, of the draw peak implies incomplete cross-linking between the trimesoyl chloride solution at the membrane interface. and m-phenylene diamine monomers of the polyamide, producing a functional group that can be expected to de-protonate at elevated pHs. This deprotonation of the polyamide selective layer would give 3.3.2. Reverse solute flux for a sodium chloride draw solution rise to negative surface charges of the membranes as detailed in the The salt flux increased (Fig. 6) after PDA modification for both the literature [50,51]. Additionally, deprotonation of carboxylic acid groups BW30 and SW30-XLE membranes as a result of the improved wetta- of a polyamide can also be attributed to improved rejections of these bility of the membranes' support layer. As support layer wetting membranes at slightly basic pHs [51,52]. As will be discussed below, improves, solutes can more easily diffuse through a membrane's

Fig. 6. Salt (sodium chloride) flux of BW30 and SW30-XLE membranes (neat, PET removed and PDA modified) at 23 ± 1 °C, 0.25 m/s feed and draw cross-flow velocity, and no transmembrane hydrostatic pressure. J.T. Arena et al. / Desalination 343 (2014) 8–16 13

PDA coating despite the fact that the porosity of the support layer is decreased as a result of PDA coating.

3.4. Desalination performance of PDA modiﬁed TFC membranes

3.4.1. Water ﬂux in forward osmosis desalination

By comparing the pure water ﬂuxes for the NH3–CO2 draw solution in Fig. 8 to water ﬂuxes for a NaCl draw solution presented in Fig. 5 it

becomes apparent that the NH3–CO2 draw solution produces similar water ﬂuxes the 1 M sodium chloride draw solution under these test conditions. Upon addition of sodium chloride water, ﬂuxes decreased by more than 50%. This is likely due to external concentration polarization effects, increasing the osmotic pressure of the feed solution at the membrane selective layer interface.

3.4.2. Solute flux in forward osmosis desalination Reverse solute flux was measured for the ammonia species per- meating through the membrane in both the molecular and ionic Fig. 7. Structural parameters of BW30 (solid bars) and SW30-XLE (cross-hatched bars) forms (as ammonia and ammonium respectively) from the draw so- fi membranes (neat, no PET and PDA modi ed), calculated from RO data presented in lution into the feed solution. The ammonia species crossover was Arena et al. [32]. measured between 0.75 and 0.9 mol·m−2·h−1. Ammonia being polar molecule like water and of similar size to water with a more mobile hydration shell than ammonium [54] prevents the membrane from easily discriminating between water and ammonia molecules [15]. support layer. This increases the concentration of those solutes at the Sodium and chloride ion fluxes are given in Fig. 9. As would be ex- selective layer interface and results in increased solute flux. pected, the SW30-XLE exhibited significantly lower forward sodium flux (cross-hatched bars) than the BW30 due to its higher selectivity. 3.3.3. Membrane structural parameters On the other hand, chloride flux is dramatically lower for both mem- Effective structural parameters for the membranes considered in branes. This was an unanticipated finding since, in early studies on FO this study were calculated using Eq. (2). Water permeance and sodium desalination using HTI's CA membrane and this draw solute found chloride permeability values reported in Arena [32] were used for this high NaCl rejections [1,20]. analysis. As shown in Fig. 7, removal of the PET backing layer resulted The unequal sodium and chloride ion fluxes must mean that a cation in a 70% reduction in the effective structural parameters for both the from the draw solution is moving to the feed solution from the draw BW30 and SW30-XLE. Following removal of the PET layers these mem- solution, since electroneutrality must be maintained. The only cation branes still exhibit structural parameter orders of magnitude higher available in the draw solute is ammonium. It is interesting to note than their structure would suggest is possible based upon their thick- that in all instances the ammonia flux was greater than or equivalent ness and porosity [53]. to the sodium flux. This supports evidence of ion exchange since it This finding suggests that the poor wetting of the PSu layer is the would close the mass balance for both ammonia and sodium moving primary cause of the high effective structural parameters for both the between the two solutions. BW30 and SW30-XLE; however, poor wetting of the PSu layer seems The ion flux data is converted to rejection values in Fig. 10 (done to be more severe for the SW30-XLE as shown by this membrane's by multiplying forward solute flux by water flux to determine the higher effective structural parameters. Modification of these concentration of water passing through the membrane then dividing membrane's PSu layer with PDA resulted in a near order of magni- this by the concentration of the feed water). The SW30-XLE had tude decrease in the effective structural parameter for both mem- better sodium and chloride rejection under these process conditions branes. This result is particularly interesting given that membrane with around 65% rejection of sodium and 85–90% rejection of chlo- porosity is reduced by the PDA coating process, as shown in Fig. 1. ride for both the 1 hour and 42 hour PDA modification. The BW30 ex- That is, the resistance to mass transfer has been reduced due to hibited a large disparity in sodium and chloride rejections. The

Fig. 8. Osmotic ﬂux data for pure water (solid bars) and 0.25 M sodium chloride (cross-hatched bars) feed solutions against a 2 M NH3/CO2 solution at 23 ± 1 °C, 0.25 m/s draw and feed cross-ﬂow velocity, and no transmembrane hydrostatic pressure. 14 J.T. Arena et al. / Desalination 343 (2014) 8–16

Fig. 9. Solute fluxes for osmotically driven sodium chloride rejection. The lined bar represents ammonia species reverse solute fluxes, the solid bar represents sodium ion forward solute fluxes, and the cross-hatched bard represents chloride ion forward solute fluxes at 23 ± 1 °C, 0.25 m/s draw and feed cross-flow velocity, and no transmembrane hydrostatic pressure. rejections of the sodium ion were 15–25% while the chloride ion functional groups of a membrane's selective layer allow for preferen- rejections were 80–85%. The cation exchange occurring between tial transport of cations. the 0.25 M NaCl feed and the 2.0 M NH3–CO2 draw solutions present Similar ion exchange behavior to those illustrated in Fig. 9 (this phenomena never directly reported. These data could also explain being unequal anion to cation transport for electrolytes) was report- low sodium chloride rejections [55] or uneven anion and cation ed in a recent publication by Coday observed unequal feed solute ion rejections of various electrolytes [24] reported by others using TFC transport using non-volatile solutes with commercial TFC FO mem- membranes. branes [24]. As these solutes do not exist in equilibrium between a charged and uncharged species this would imply that the membrane chemistry is the dominating factor in ion transport behavior. This is 3.4.3. Ion exchange mechanisms further reinforced by observations also by Coday where HTI's CA FO There are two possible mechanisms for the ion exchange behav- membrane was also tested displaying lower cation fluxes than TFC ior exhibited between the NaCl feed and NH3–CO2 draw solutions. FO membranes [24]. Additionally the high rejections of sodium chlo- The first is reliant upon the equilibria amongst ammonia species ride in studies using HTI's CA membrane with the NH3–CO2 draw so- within the draw solution [17]. Three nitrogen containing species lution further demonstrate the importance of membrane chemistry are present within the draw solution: ammonia, ammonium, and [1,20]. carbamate. These species are in equilibrium, but ammonia is un- A classical cation exchange resin should be a cross-linked water charged, has chemical interactions similar to water, and a less rigid insoluble structure with acidic functional groups (i.e. sulfonic, hydration shell (in relation cation and anion species) [54].Assuch carboxylic, and phenolic). These acid functional groups when ammonia can easily diffuse through the membrane selective layer deprotonated would have a negative charge allowing for ionic inter- without affecting electroneutrality between the two solutions. actions with cations, specifically cations residing within the polymer Ammonia present within feed solution can now speciate into ammo- structure and exchanging cations in solution. Cation exchangers nium, causing an imbalance of charge. This charge imbalance drives a with carboxylic acid functionality are pH sensitive only functioning sodium ion (the only cation available on the feed side) to diffuse into as such at pHs above 7 [42]. The FTIR spectra for these membranes the draw solution thus producing the unequal feed solution ion indicate that carboxylic functional groups are part of the polyamide fluxes. The second mechanism for ion exchange is the selective selective layers of these membranes (see Fig. 4) [49].Sothemost layer functioning as a cation exchanger where negatively charged likely reason for the ion exchange behavior is the deprotonation of

Fig. 10. Observed rejection for a 2 M NH3/CO2 draw solution versus 0.25 M sodium chloride feed. The solid bars represent sodium rejection and the cross-hatched bars represent chloride rejection at 23 ± 1 °C, 0.25 m/s draw and feed cross-flow velocity, and no transmembrane hydrostatic pressure. J.T. Arena et al. / Desalination 343 (2014) 8–16 15 carboxylic acid functional groups of polyamide making available [14] J.C. Watters, E. Klein, M. Fleischman, J.S. Roberts, B. Hall, Rejection spectra of reverse osmosis membranes degraded by hydrolysis or chlorine attack, Desalination 60 cation exchange site within the polyamide layer [42,51,52].Thisal- (1986) 93–110. lows for the movement of cations between the feed and draw solu- [15] R.W. Baker, Membrane Technology and Applications, Second edition John Wiley & tions; therefore, in order to mitigate this behavior in polyamide Sons Ltd, West Sussex, England, 2004. [16] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, Desalination by ammonia–carbon di- based TFC membranes the pH would need to be below 7 (not possi- oxide forward osmosis: influence of draw and feed solution concentrations on pro- ble with all draw solutions). Alternatively other selectively layer cess performance, J. Membr. Sci. 278 (2006) 114–123. chemistries can be synthesized or revisited. [17] F. Mani, M. Peruzzini, P. Stoppioni, CO2 absorption by aqueous NH3 solutions: speci- ation of ammonium carbamate, bicarbonate and carbonate by a 13C NMR study, Green Chem. 8 (2006) 995–1000. fi 4. Conclusions [18] R.J. Petersen, Composite reverse osmosis and nano ltration membranes, J. Membr. Sci. 83 (1993) 81–150. [19] J.R. McCutcheon, M. Elimelech, Influence of membrane support layer hydrophobic- This study explored the impact on FO properties resulting from ity on water flux in osmotically driven membrane processes, J. Membr. Sci. 318 the application of a thin film of PDA on the PSu support structure of (2008) 458–466. [20] J.R. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal con- a commercial TFC RO membrane. A four and six fold enhancement centration polarization on flux behavior in forward osmosis, J. Membr. Sci. 284 in the FO mode osmotic flux of the BW30 and SW30-XLE membranes, (2006) 237–247. respectively, was observed after modification with PDA. Overall, [21] A.K. Ghosh, E.M. Hoek, Impacts of support membrane structure and chemistry on – fl polyamide polysulfone interfacial composite membranes, J. Membr. Sci. 336 these membranes were shown to have modest ux under desalina- (2009) 140–148. tion conditions with a 2.0 M NH3–CO2 draw solution and a 0.25 M so- [22] N. Bui, M.L. Lind, E.M. Hoek, J.R. McCutcheon, Electrospun nanofiber supported thin dium chloride feed; however, low sodium rejections were observed film composite membranes for engineered osmosis, J. Membr. Sci. 385–386 (2011) – due to cation exchange between the draw and feed solutions. 10 19. [23] Hydration Technology Innovations™, HTI's New Thin Film Forward Osmosis Mem- Evidence for this ion exchange is provided by the unequal fluxes brane in Production, http://www.htiwater.com/news/press-room/content/2012/ between sodium and chloride. A tuning of process conditions or press-HTI-HTIThinFilmMembrane042512.pdf. membrane chemistry may enable higher rejections for both ions [24] B.D. Coday, D.M. Heil, P. Xu, T.Y. Cath, Effects of transmembrane hydraulic pressure on performance of forward osmosis membranes, Environ. Sci. Technol. 47 (2013) within the feed solution. 2386–2393. [25] B.D. McCloskey, H.B. Park, H. Ju, B.W. Rowe, D.J. Miller, B.D. Freeman, A bioinspired fouling-resistant surface modification for water purification membranes, J. Membr. Acknowledgments Sci. 413–414 (2012) 82–90. [26] B.D. McCloskey, H.B. Park, H. Ju, B.W. Rowe, D.J. Miller, B.J. Chun, K. Kin, B.D. Freeman, Influence of polydopamine deposition conditions on pure water flux and foulant ad- The authors acknowledge funding from the NSF CBET Chemical hesion resistance of reverse osmosis, ultrafiltration, and microfiltration membranes, and Biological Separations Program #1160098 and #1160069, and Polymer 51 (2010) 3472–3485. USEPA (Project No. R834872). The authors also acknowledge the [27] D.J.Miller,P.A.Araújo,P.B.Correia,M.M.Ramsey,J.C.Kruithof,M.C.M.van Loosdrecht, B.D. Freeman, D.R. Paul, M. Whiteley, J.S. Vrouwenvelder, NWRI-AMTA Fellowship for Membrane Technology and National Sci- Short-term adhesion and long-term biofouling testing of polydopamine and ence Foundation GK-12 Program, which provided support for Jason poly(ethylene glycol) surface modifications of membranes and feed spacers T. Arena. The authors also wish to thank Dow Water & Process Solutions for biofouling control, Water Res. 46 (2012) 3737–3753. for providing membranes for this study. Additionally the authors would [28] S. Kasemset, A. Lee, D.J. Miller, B.D. Freeman, M.M. Sharma, Effect of polydopamine deposition conditions on fouling resistance, physical properties, and permeation like to thank Dr. James V. Arena at Central Connecticut State University properties of reverse osmosis membranes in oil/water separation, J. Membr. Sci. and Dr. Abhay Vaze at the University of Connecticut for providing 425–426 (2013) 208–216. technical assistance and feedback for some of the analytical tech- [29] H. Lee, N.F. Scherer, P.B. Messersmith, Single-molecule mechanics of mussel adhesion, Proc. Natl. Acad. Sci. U. S. A. 103 (2006) 12999–13003. niques used. [30] H. Lee, Y. Lee, A.R. Statz, J. Rho, T.G. Park, P.B. Messersmith, Substrate-independent layer-by-layer assembly by using mussel-adhesive-inspired polymers, Adv. Mater. 20 (2008) 1619–1623. References [31] H. Lee, S.M. Dellatore, W.M. Miller,P.B.Messersmith,Mussel-inspired surface chemistry for multifunctional coatings, Science 318 (2007) 426–430. [1] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, A novel ammonia–carbon dioxide for- [32] J.T. Arena, B.D. McCloskey, B.D. Freeman, J.R. McCutcheon, Surface modification of ward (direct) osmosis desalination process, Desalination 174 (2005) 1–11. thin film composite membrane support layers with polydopamine: enabling use [2] T.Y. Cath, S. Gormly, E.G. Beaudry, M.T. Flynn, V.D. Adams, A.E. Childress, Membrane of reverse osmosis membranes in pressure retarded osmosis, J. Membr. Sci. 375 contactor processes for wastewater reclamation in space: part I. Direct osmotic con- (2011) 55–62. centration as pretreatment for reverse osmosis, J. Membr. Sci. 257 (2005) 85–98. [33] G. Han, S. Zhang, X. Li, N. Widjojo, T. Chung, Thin film composite forward osmo- [3] R.E. Kravath, J.A. Davis, Desalination of sea water by direct osmosis, Desalination 16 sis membranes based on polydopamine modified polysulfone substrates with (1975) 151–155. enhancements in both water flux and salt rejection, Chem. Eng. Sci. 80 (2012) [4] A. Achilli, T.Y. Cath, A.E. Childress, Selection of inorganic-based draw solutions for 219–231. forward osmosis applications, J. Membr. Sci. 364 (2010) 233–241. [34] S.S. Manickam, J.R. McCutcheon, Characterization of polymeric nonwovens using [5] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications, and porosimetry, porometry and X-ray computed tomography, J. Membr. Sci. 407–408 recent developments, J. Membr. Sci. 281 (2006) 70–87. (2012) 108–115. [6] R.L. McGinnis, N.T. Hancock, M.S. Nowosielski-Slepowron, G.D. McGurgan, Pilot [35] J.R. McCutcheon, M. Elimelech, Modeling water flux in forward osmosis: implica- – demonstration of the NH3/CO2 forward osmosis desalination process on high salin- tions for improved membrane design, AIChE J 53 (2007) 1736 1744. ity brines, Desalination 312 (2013) 67–74. [36] N.Y. Yip, M. Elimelech, Performance limiting effects in power generation from [7] R.L. McGinnis, M. Elimelech, Energy requirements of ammonia–carbon dioxide salinity gradients by pressure retarded osmosis, Environ. Sci. Technol. 45 forward osmosis desalination, Desalination 207 (2007) 370–382. (2011) 10273–10282. [8] S.K. Yen, F. Mehnas Haja N, M. Su, K.Y. Wang, T. Chung, Study of draw solutes using [37] A. Tiraferri, N.Y. Yip, A.P. Straub, S. Romero-Vargas Castrillon, M. Elimelech, A meth- 2-methylimidazole-based compounds in forward osmosis, J. Membr. Sci. 364 (2010) od for the simultaneous determination of transport and structural parameters of 242–252. forward osmosis membranes, 444 (2013) 523–538. [9] Q. Ge, J. Su, T. Chung, G. Amy, Hydrophilic superparamagnetic nanoparticles: synthe- [38] R.A. Robinson, R.H. Stokes, Electrolyte Solutions, Second Revised edition Dover Pub- sis, characterization, and performance in forward osmosis processes, Ind. Eng. lications, Inc., Mineola, New York, 2002. Chem. Res. 50 (2010) 382–388. [39] W.A. Phillip, J.S. Yong, M. Elimelech, Reverse draw solute permeation in forward [10] M.M. Ling, K.Y. Wang, T. Chung, Highly water-soluble magnetic nanoparticles as osmosis: modeling and experiments, Environ. Sci. Technol. 44 (2010) 5170–5176. novel draw solutes in forward osmosis for water reuse, Ind. Eng. Chem. Res. 49 [40] N.T. Hancock, T.Y. Cath, Solute coupled diffusion in osmotically driven membrane (2010) 5869–5876. processes, Environ. Sci. Technol. 43 (2009) 6769–6775. [11] T.Y. Cath, M. Elimelech, J.R. McCutcheon, R.L. McGinnis, A. Achilli, D. Anastasio, A.R. [41] N.T. Hancock, W.A. Phillip, M. Elimelech, T.Y. Cath, Bidirectional permeation of elec- Brady, A.E. Childress, I.V. Farr, N.T. Hancock, J. Lampi, L.D. Nghiem, M. Xie, N.Y. Yip, trolytes in osmotically driven membrane processes, Environ. Sci. Technol. 45 (2011) Standard methodology for evaluating membrane performance in osmotically driven 10642–10651. membrane processes, Desalination 312 (2013) 31–38. [42] G.H.Jeffery,J.Bassett,J.Mendham,R.C.Denney, Vogel's Textbook of Quantitative Chem- [12] J. Herron, Asymmetric Forward Osmosis Membranes, 2008. ical Analysis, Fifth edition Longman Scientific & Technical, Essex, England, 1989. [13] K.D. Vos, F.O. Burris, R.L. Riley, Kinetic study of the hydrolysis of cellulose acetate in [43] D.C. Harris, Quantitative Chemical Analysis, Sixth edition W. H. Freeman and Com- the pH range of 2-10, J. Appl. Polym. Sci. 10 (1966) 825–832. pany, New York, NY, 2003. 16 J.T. Arena et al. / Desalination 343 (2014) 8–16

[44] F. Pan, H. Jia, S. Qiao, Z. Jiang, J. Wang, B. Wang, Y. Zhong, Bioinspired fabrication of [50] C.Y. Tang, Y. Kwon, J.O. Leckie, Characterization of humic acid fouled reverse osmosis high performance composite membranes with ultrathin defect-free skin layer, J. and nanofiltration membranes by transmission electron microscopy and streaming Membr. Sci. 341 (2009) 279–285. potential measurements, Environ. Sci. Technol. 41 (2006) 942–949. [45] C.Y. Tang, Y. Kwon, J.O. Leckie, Effect of membrane chemistry and coating layer [51] E.M. Van Wagner, B.D. Freeman, M.M. Sharma, M.A. Hickner, S.J. Altman, Polyamide on physiochemical properties of thin film composite polyamide RO and NF Desalination Membrane Characterization and Surface Modification to Enhance Foul- membranes: I. FTIR and XPS characterization of polyamide and coating layer ing Resistance, Sandia National Laboratories, Albuquerque, New Mexico, 2010. chemistry, Desalination 242 (2009) 149–167. [52] E.M. Van Wagner, A.C. Sagle, M.M. Sharma, B.D. Freeman, Effect of crossflow testing [46] Dow Water & Process Solutions, FILMTEC™ Reverse Osmosis Membranes Technical conditions, including feed ph and continuous feed filtration, on commercial reverse Manual, Form No. 609-00071-0309. osmosis membrane performance, 345 (2009) 97–109. [47] J. Cadotte, R. Petersen, R. Larson, E. Erickson, A new thin-film composite seawater re- [53] S. Soundara Manickam, J. Gelb, J. McCutcheon, Characterization of thin film compos- verse osmosis membrane, Desalination 32 (1980) 25–31. ite membranes using porosimetry and X-ray microscopy, Microsc. Microanal. 19 [48] D.R. Dreyer, D.J. Miller, B.D. Freeman, D.R. Paul, C.W. Bielawski, Elucidating the struc- (2013) 634–635. ture of poly(dopamine), 28 (2012) 6428–6435. [54] H. Hesske, K. Gloe, Hydration behavior of alkyl amines and their corresponding proton- [49] J.B. Lambert, H.F. Shurvell, D.A. Lightner, R.G. Cooks, Organic Structural Spectrosco- ated forms. 1. Ammonia and methylamine, J. Phys. Chem. A 111 (2007) 9848–9853. py, Prentice Hall, Inc., Upper Saddle River, New Jersey, 1998. [55] S.C. Low, Preliminary studies of seawater desalination using forward osmosis, Desalin. Water Treat. 7 (2009) 41–46.