Pre-ozonation for high recovery of nanofiltration (NF) membrane system: Membrane fouling reduction and trace organic compound attenuation

Author’s Accepted Manuscript Pre-ozonation for high recovery of nanofiltration (NF) membrane system: Membrane fouling reduction and trace organic compound attenuation Minkyu Park, Tarun Anumol, Julien Simon, Flavia Zraick, Shane A. Snyder www.elsevier.com/locate/memsci

PII: DOI: Reference:

S0376-7388(16)30857-2 http://dx.doi.org/10.1016/j.memsci.2016.09.051 MEMSCI14772

To appear in: Journal of Membrane Science Received date: 30 June 2016 Revised date: 11 September 2016 Accepted date: 12 September 2016 Cite this article as: Minkyu Park, Tarun Anumol, Julien Simon, Flavia Zraick and Shane A. Snyder, Pre-ozonation for high recovery of nanofiltration (NF) membrane system: Membrane fouling reduction and trace organic compound a t t e n u a t i o n , Journal of Membrane Science, http://dx.doi.org/10.1016/j.memsci.2016.09.051 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Pre-ozonation for high recovery of nanofiltration (NF) membrane system: Membrane fouling reduction and trace organic compound attenuation Minkyu Parka, Tarun Anumola,b, Julien Simonc, Flavia Zraickc, Shane A. Snydera*

a

Department of Chemical & Environmental Engineering, University of Arizona, 1133 E James E Rogers Way, Harshbarger 108, Tucson, AZ 85721-0011, USA b c

Agilent Technologies Inc., 2850 Centerville Road, Wilmington, DE 19808, USA

CIRSEE, Suez Environnement, 38 rue du Président Wilson, 78230 Le Pecq, France

*

Corresponding author. [email protected]

Abstract This study demonstrated the use of nanofiltration (NF) membrane with pre-ozonation for surface water brine treatment. Pre-ozonation, used for fouling control, with relatively low ozone doses (0.1−0.4 mg O3/mg DOC ratios) significantly reduced organic fouling potential. A classical pore blocking model revealed that the dominant fouling mechanism was cake filtration. Based on a statistical correlation test, macromolecules (apparent molecular weight >10k Da) of typtophanlike aromatic protein and soluble microbial-like matter exhibited strong correlation to fouling potential. Small molecules (<1600 Da) of UV-absorbing matter were also correlated with fouling potential whereas bulk parameters such as SUVA displayed relatively lower correlation with fouling potential. Trace organic compound (TOrC) attenuation by ozone, membranes, and combination of ozone and membranes were also examined. Greater than 70% removal of the 17 TOrCs was achieved at 0.4 mg O3/mg DOC ratio. The NF membrane (NF90) alone also could remove more than 90% of TOrCs. When pre-ozonation was performed with the NF membrane,

all TOrCs tested were removed to less than the limits of detection. In conclusion, the application of pre-ozonation to NF membrane not only can alleviate organic fouling potential, but also can function as a dual barrier for ensuring removal of TOrCs for water treatment. Keywords: Nanofiltration; Brine; Ozone; Fouling; Micropollutant 1. Introduction With rapid increase in water scarcity, a significant amount of the population is struggling for access to reliable water sources [1]. For instance, over 4 billion people are reported to be threatened by inadequate access to freshwater [2]. To alleviate these water stresses, several measures such as water conservation, repair of existing infrastructure, and investment in new distribution systems [3] have been considered. However, these solutions can only improve utilization of existing water resources, but cannot overcome the existing hydrological cycle or provide new sources of water. Two viable options independent of hydrological cycle are desalination and water reuse. Membrane technology can be critical in implementation of both these schemes. Among membrane technologies, pressure-driven membrane processes such as microfiltration (MF), ultrafiltration (UF), nanofiltration (NF) and reverse osmosis (RO) are dominant in the water treatment market and proven to be the most effective against removing chemical and biological contaminants, thus providing high quality water suitable for potable use [4]. Benefits of membrane technologies include ease of operation, no phase change, and temperature-insensitive nature enabling operation at ambient temperature [5]. In addition, no byproduct formation is a benefit over oxidation treatment technologies such as ozone, UV, and chlorination [6-8]. However, inherent production of brine is an increasing problem [9, 10].

There are several options for brine management. Marine discharge, that is the discharge of brine into marine water bodies including bays, estuaries, and/or oceans, is one of the dominant practices [11]. However, this practice may contaminate receiving water bodies and is not generally viable for in-land communities. Further, U.S. states like Florida are considering banning offshore brine discharge which will require alternative strategies of brine disposal [12]. While brine disposal into sewers can also be an option, the addition of inorganic and organic contaminants into sewage lowers biochemical oxygen demand (BOD) removal and increase total dissolved solids, hence potentially aggravating activated sludge processes when this water gets to the water treatment plant [13]. Zero liquid discharge (ZLD) is a combination of processes inducing very high recovery rate (nearly 99 %) using techniques such as thermal distillation/evaporation and can be a viable option for brine treatment. However, currently available ZLD technologies raise operating costs significantly [14]. In this study, a NF membrane system was considered to achieve additional water recovery from brine. RO membrane has been used as a pre-concentration step in ZLD applications, which can enhance energy and cost efficiencies of subsequent energy-intensive desalting processes such as thermal distillation [15]. Even though RO membranes are adequate for such processes due to high rejection rate of organic and inorganic matter, NF membrane has benefits over RO membrane in that it requires lower specific energy consumption due to lower hydraulic membrane resistance while maintaining reasonable rejection of organic matter and salts. Although NF membrane is not suitable for desalination of highly saline water such as seawater, its application to less saline water such as surface water brine would expect substantial cost benefit. In order to apply NF membrane to brine treatment, fouling and/or scaling are main constraints to overcome [16]. While inorganic scaling is often a limiting factor to achieve high

recovery, organic fouling substantially increases operating cost and reduces membrane life span. Therefore, this study focuses on reduction of organic fouling potential. As a means of reducing organic fouling potential, pre-ozonation has been actively studied recently [17-21]. An ozone pilot plant testing for MF fouling control, for instance, is currently in practice at Edward C. Little Water Recycling Facility operated by the West Basin Municipal Water District in the State of California [22]. However, very few studies on the effects of pre-ozonation, particularly on use of NF membrane for brine fouling control have been conducted [17]. In addition to concerns of fouling, significantly greater amount of trace organic compounds (TOrCs) such as pharmaceuticals, personal care products, steroid hormones and industrial compounds is present in concentrate stream relative to the feed that may cause unacceptable concentrations in permeate. Hence, it is crucial to investigate the removal of TOrCs in highly concentrated stream (brine) and possible synergism of combination of pre-ozonation and membrane filtration for TOrCs removal in brine treatment. The primary aim of this study is to demonstrate application of pre-ozonation to fouling control for a NF-based pre-concentration process in a ZLD based on bench-scale testing. To this end, fouling propensity test for NF brines at two ozone doses was conducted using a flat sheet membrane system. In attempts to investigate possible alteration of fouling mechanism, classical pore blocking and cake filtration models were applied. For quantitative comparison, the change in modified fouling index (MFI)—an indicator of fouling potential—of brine before and after ozonation was calculated. Furthermore, the fouling potential (MFI) was statistically correlated with various organic quality parameters such as specific UV absorbance (SUVA), total fluorescence (TF) and chromophores with different apparent molecular weights. In addition, the

effects of pre-ozonation on the removal of TOrCs in permeate was also studied in order to scrutinize possible beneficial effects of pre-ozonation on the removal of TOrCs in a NF membrane system. 2. Methods This study was based on a scenario to treat brine from NF system with 85% water recovery. To this end, multiple processes were implemented in sequence. First, brine was generated as a feed water for fouling propensity test. To this end, a 4-inch spiral wound membrane system was utilized (Section 2.1). Thereafter, the generated brine was ozonated with two different ozone doses (Section 2.2), then fouling propensity of ozonated and non-oznated brines were tested using a bench-scale flat sheet membrane system (Section 2.3). The detailed experimental procedure is as follows. 2.1. Brine generation The brine was prepared for fouling propensity tests by using a spiral wound membrane system with two pressure vessels operated in parallel. Two Dow Filmtec NF270-4040 membrane elements were employed and a one-micron filter was installed prior to the membrane elements to remove particulate matter present in the feed water. Constant pressure and feed flow rate was employed during the brine generation (4.5 bar and 23 L/min, respectively). Desired recovery rate (85 %) was achieved by recirculating the concentrate stream back into the feed tank while the permeate stream was discarded. As feed water, Colorado river water (CRW) was used and 50 ng/L of 17 TOrCs were spiked into the feed water before brine generation [23]. 1.5 mg/L of antiscalant (Nalco Permatreat PC 191) was spiked into the feed water, which was calculated using PC Optimize (ver. 1.2.0) provided by Nalco, USA.

2.2. Ozonation setup and decay testing A ozone decay test of the produced brine was conducted using a bench-scale ozone generator (Modular 8HC ozone generator, Xylem Wedeco, Germany). First, ozone stock solution was prepared by diffusing ozone gas through DI water at 1°C. A concentrated stock of dissolved ozone at ~1.2 mM was then added into the brine to get the desired ozone dose for decay testing. Dissolved ozone concentration was measured using the indigotrisulfonate ozone residual method [24]. The residual ozone concentration in the brine was measured at 10, 20, 30, 60, 120 and 240 sec. Based on the result of ozone decay test, two ozone doses (1.8 mg/L and 7.5 mg/L corresponding to 0.1 mg O3/mg DOC and 0.4 mg O3/mg DOC ratios, respectively) were chosen. The selection of the ozone doses was described in detail in Section 3.1. The generated brine was ozonated at the two selected ozone doses using a pilot-scale ozone flow-through generator (Modular 8HC, Xylem Wedeco, Germany). 2.3. Fouling propensity test The fouling propensity test unit consists of a membrane cell, feed tank, temperature controller, pump, and data acquisition system (Fig. S1 in the supplementary material). 5 L of the brine as feed water was tested and the concentrate stream was recirculated into the feed tank. A constant temperature (25°C) was maintained by circulating cooling water through a stainless steel coil immersed in the feed tank during the fouling test. A flat sheet membrane cell (SEPA CF cell, Sterlitech Corp., Kent, WA, USA) was used with a 47-mil feed spacer and a 10-mil shim. Feed flow was controlled at a constant rate of 2 L/min using a variable frequency drive (VFD) and desired pressure was controlled using a pressure regulator. Dow Filmtec NF90 membrane was compacted using DI water at the same pressure as one applied for the test water (brine) for 24

hours, then was used for fouling propensity test. Flowrates of the concentrate stream and permeate stream were automatically recorded using a data acquisition system. Samples for analysis of TOrCs in permeate water were collected at 24 hours after initiation of fouling propensity test. 2.4. Analysis Dissolved organic carbon (DOC) was measured using Shimadzu TOC-L CSH Total Organic Carbon Analyzer (Shimadzu Corp., Japan). Before analysis, samples were filtered using 0.45 µm PVDF syringe filter (EMD Millipore), then acidified to pH 3 or lower using HCl (ACS grade, 37%, Sigma Aldrich). Metals were analyzed by an Agilent 7700x Series ICP-MS in helium collision gas mode. Samples for metal analysis were acidified with 2% (v/v) HNO3 using OPTIMA grade nitric acid (Fisher Scientific). 20 µL of an aqueous solution containing equal concentrations (5 mg/L each) of scandium, germanium, indium, and bismuth was then added to each tube as an internal standard and the contents vortexted vigorously. Reported measurements of metals in each sample represent the average of 3 measurements. Apparent molecular weight (AMW) of dissolved organic matters (DOMs) was measured using size exclusion chromatography (SEC). An Agilent 1290 high performance liquid chromatography (HPLC) (Palo Alto, CA) hyphenated with diode array detector (Agilent 1260 DAD) at wavelength of 254 nm and fluorescence detector (Agilent 1290 FLD) was applied. Five pairs of excitation (Ex) and emission (Em) wavelengths were employed for FLD: Ex/Em of 225nm/297nm, 225nm/343nm, 225nm/440nm, 280nm/345nm and 330nm/440nm, which represent tyrosine-like aromatic protein, tryptophan-like aromatic protein, fulvic acid-like matter, soluble microbial product (SMP)-like matter and humic acid-like matter, respectively. A silica-

based column (Waters Protein-Pa 125; particle size 10 µm, diameter 7.8 mm, and length 300 mm) was used for separation. Eluent was prepared with 2.4 mM sodium phosphate, 1.6 mM disodium hydrogen phosphate, and 96.0 mM sodium chloride [25]. Polystyrene sulfonates with molecular weights (MWs) of 1100, 3610, 6520, 14900, and 32900 Da were used as MW standards (Polymer Standards Service, Mainz, Germany). UV and fluorescence spectra were measured simultaneously using a Horiba Aqualog fluorometer. UV spectra were scanned from 200 nm to 800 nm wavelength and the obtained UV spectra were used to correct inner filter effects and to obtain UV absorbance at 254 nm wavelength [26]. Regional integration method was conducted and the detailed methods can be found in the Chen et. al’s work [27]. In brief, five operationallydefined fluorescent regions were selected and the fluorescence intensity is integrated under the each region. Seventeen TOrCs were measured using an online solid-phase extraction module coupled to an ultra-HPLC tandem mass spectrometer (UHPLC-MS/MS) from Agilent Technologies (Palo Alto, CA). This setup employed a FlexCube module to perform automated online solid phase extraction using a polymeric solid phase extraction cartridge that is attached to an Agilent 1290 Infinity UHPLC, which in turn is coupled to an Agilent 6460 LC/MS Triple Quadrupole mass spectrometer. The method required only 1.7 mL of sample to achieve low ng/L detection limits for all seventeen TOrCs and performed simultaneous positive and negative electrospray ionization (ESI) allowing for just one sample injection to measure all analytes. Details of the analytical method including method performance and validation have been published previously [28]. Data was processed using Agilent Mass Hunter software (Ver. 06.01) and samples were quantified using the isotope dilution method.

3. Results and discussion 3.1. Ozone decay test to determine ozone dose An ozone residual in feed water can cause membrane damage since polymeric membranes including polyamide membranes are chemically intolerant to ozone [29]. Oxidants such as ozone and chlorine attack polymeric membranes and cause rapid membrane failure by increasing both water permeability and salt permeability [30]. In practice, residual ozone can be quenched by oxidant quenching agents such as sodium bisulfite, but is not attractive in terms of extra operating cost. Hence, the best practice would be to determine an ozone dose which can instantaneously be consumed in the water, but still provides sufficient treatment. To this end, ozone decay experiments can be performed to estimate residual ozone dose. In this study, the residual ozone in the brine was rapidly consumed (<60 sec) (Fig. 1). Instantaneous ozone demand (IOD) is defined as the concentration of ozone decomposed within the initial 20 sec [31] and IOD of the brine was determined to be 7.3 mg/L. Application of the lowest ozone dose to achieve the required treatment is always desirable as it minimizes operating costs and reduces potential for formation of any additional oxidation byproducts. Therefore, two ozone doses were selected based on IOD: a low value (1.8 mg/L corresponding to 0.1 mg O3/mg DOC ratio) and one at close to IOD (7.5 mg/L corresponding to 0.4 mg O3/mg DOC ratio). 3.2. Fouling reduction by pre-ozonation In order to investigate the impacts of ozonation on fouling propensity, 10 and 25 bar were applied on the membrane. It is important to clarify that the ‘brine’ hereinafter refers to concentrated CRW water used as feed to the flat sheet membrane. Fig. 2 illustrates clearly that increasing ozone dose reduces fouling propensity for the both applied pressures. As expected,

there was little effect due to ozone on inorganic water quality parameters on fouling including ionic strength, divalent cation concentrations as well as pH (Table 1) [32]. Hence, the reduced fouling potential is not expected to be caused by the change in inorganic chemistry. In other words, the change in fouling propensity due to ozonation would be closely related to the change of NOM characteristics in the brine. This will be further scrutinized in the later sections.

In NF membranes, cake filtration is known to be the major fouling mechanism [33]. To grasp the extent of fouling reduction by ozonation, it is important to quantify the change in fouling potential. The modified fouling index (MFI) can be utilized for the quantification of fouling potential when cake filtration plays a dominant role in fouling [34, 35]. MFI was derived by Schippers et al. based on cake filtration mechanism for colloidal fouling and its value is linearly proportional to the concentration of colloidal matter at a given pressure [34]. The MFI can be obtained from the slope of the linear region in the plot of t/V versus V as expressed in Eq. (1).

hI t h Rm = + V . V DPA 2DPA P 2

(1)

MFI

where η is the dynamic viscosity, Rm is the intrinsic membrane resistance, ΔP is the applied transmembrane pressure, I is a measure of the membrane fouling potential of the feed water, and A is the membrane area. As shown in Eq. (1), MFI is a function of the fouling potential and the transmembrane applied pressure, so MFI can be a good indicator to compare fouling potential among different feed waters. It should be noted that MFI was used as an indicator of organic fouling and not specifically limited to colloidal fouling.

MFI was calculated by plotting t/V versus the accumulated filtrate volume (Fig. 3(a) and (b)) and its values are shown in Fig. 3(c). Fouling potential of the brine for both 0.1 and 0.4 mg O3/mg DOC ratios was significantly reduced (Fig. 3) compared with the non-ozonated brine. For example, MFI values were reduced by 56.6% and 79.1% for 25 bar and 10 bar applied pressure respectively at 0.4 mg O3/mg DOC ratio. This implies that the pre-ozonation was as equivalently effective as reducing 56.6% and 79.1% of foulant, respectively. This is interesting since applied ozone doses were low, but effective to reduce fouling potential of the brine nonetheless. The MFI of the 10 bar applied pressure is greater than the 25 bar pressure due to lower convective flux of NOM towards the membrane surface because the extent of fouling is proportional to the permeate flux [36]. In addition, higher pressures can compress cake layer and render it less porous and denser, thereby increasing hydraulic resistance. The reduced fouling potential will be further correlated with organic properties such as SUVA and fluorescence properties in the later sections. 3.3. Fouling mechanism The reduction in organic fouling potential by ozonation was observed in the preceding section and it is invaluable to investigate the fouling mechanism change if there is. Organic fouling consists of three main sequential phases: pore blocking, cake filtration (or gel filtration), and cake filtration with compression [34]. In high pressure membrane processes such as NF and RO, cake filtration plays a dominant role in membrane fouling since pore blocking mechanisms generally occur if foulant size is similar or smaller than the pore size of the membrane. In order to confirm this, graphical analysis of classical pore blocking model was performed using the following equation:

d 2t æ dt ö =kç ÷ 2 dV è dV ø

n

(2)

where t is the filtration time, V is the total filtered volume, and k is the fouling coefficient [37, 38]. Classical pore blocking can be expressed by the exponent value: n=0 for cake filtration, n=1 for intermediate pore blocking, n=1.5 for standard pore blocking, and n=2 for complete pore blocking. n value can be obtained graphically by logarithmically plotting d2t/dV2 versus dt/dV, which can be calculated as follows:

dt 1 = dV JA

(3)

d 2t æ 1 öæ dJ ö = - ç 3 2 ÷ç ÷ 2 dV è J A øè dt ø

(4)

where J is the filtrate flux and A is the membrane area. The value of n can often be determined graphically [39]. Fig. 4 shows a plot of d2t/dV2 versus dt/dV for the applied pressure of 25 bars (a) and 10 bars (b). Note that spline curve fitting was conducted to help show the trend. It is noteworthy that higher value of dt/dV indicates longer filtration time (see Eq. (3). dt/dV is inversely proportional to the permeate flux which declines as increasing filtration time). As seen in the figure, the slopes for the both applied pressures are close to zero at early filtration time, then become negative as increasing dV/dt for the tested applied pressures and ozone doses. The zero slope (n=0) at shorter filtration time indicates that the organic fouling transpired under the cake filtration mechanism at the beginning of fouling. The following decreasing trend (i.e., negative slope) cannot be explained by any classical fouling model. In MF processes, the transition from zero to negative

slope is a direct result of the transition from pore blockage to cake filtration [40]. In NF membrane process; however, the decreasing trend may be attributed to additional flux decline by cake-enhanced osmotic pressure (CEOP) since the built-up cake layer provokes additional resistance against diffusive solute transport, which cannot be explained by the classical pore blocking model [41]. No noticeable change of graph trends for the both applied pressures were found after ozonation, which indicates that ozonation did not alter main fouling mechanism (i.e., cake filtration). In sum, cake filtration was dominant fouling mechanism, hence the alteration of cake layer characteristics would be the main cause for the reduced fouling potential. NOM characteristics change will be explained in the subsequent sections using spectroscopic analyses. 3.4. Change in NOM characteristics by ozone oxidation Specific UV absorbance (SUVA) is UV absorbance at 254 nm wavelength normalized by DOC concentration. As shown in Fig. 5(a), SUVA value was significantly reduced by ozonation [42], similar to other studies [7, 43]. Ozone is highly reactive with electrophilic aromatic moieties, thereby reducing attraction between NOM and membrane. Relatively fast reaction of aromatic moieties at low ozone dose can be beneficial in the application of pre-ozonation for fouling potential reduction since it can reduce operating costs. In addition, the selective removal of aromatic moieties of foulant enables pre-ozonation to be effective in reduction of fouling potential of feed water with high NOM. Hydrophobic interaction is known to be a major fouling mechanism in NF membrane filtration [44]. For instance, hydrophobic fraction of NOM can favorably adsorbs onto surfaces of hydrophobic membrane. Here, three-dimensional excitation-emission matrix (EEM) fluorescence spectra was utilized to investigate hydrophobicity change by ozone. Total fluorescence (TF) is

the summation of integrated fluorescence values over the operationally-defined excitationemission boundaries (Fig. 5b) and can be an indicator of hydrophobicity of NOM [27]. According to Chen et al. [27], hydrophobic fractions of NOM exhibited larger normalized TF (TF divided by DOC) than base and hydrophilic fractions. Similar to SUVA, normalized TF was substantially reduced by ozonation, which indicates the reduction in hydrophobicity. When investigating each fluorescent regions (Fig. 5(c)), it is worth pointing out that aromatic fluorescent fractions of NOM such as aromatic proteins including tryrosine-like (Region 1), tryptophan-like (Region 2), and SMP-like fluorescent organic matter (Region 4) were attenuated with greater efficacy compared to the other regions. Fulvic-like (Region 3) and humic-like (Region 5) fluorescent fractions of NOM were more recalcitrant to ozone oxidation compared to the other aromatic fractions. The greater removal of fluorophores at region 1, 2 and 4 were attributed to the greater reactivity of their large size fractions, which was observed from the SEC analysis. Fig. 6 depicts the AMW distribution of UV-absorbing and fluorescent NOM. NF membranes generally possess small pores (<1,000 Da) and the molecular weight cut-off (MWCO) of the Dow Filmtec NF90 membrane used in this study was reported between 200 and 300 Da [45, 46]. Hence, the main separation mechanism of NOM would be the size exclusion. No shift of retention time (equivalently AMW) at which local maxima of chromatograms were observed was found after ozonation while intensities of chromatographic peaks of UV-absorbing NOM at given AMWs substantially decreased. This does not indicate no AMW change due to ozonation since UV-absorbing capacity of newly formed molecules would be very little, which cannot be captured by DAD. Hence, this study considers only intensity change in spectroscopic parameters at given AMWs. However, some insights can be obtained from literature. Previous studies

reported moderate shift of molecular size of NOM by ozonation at an O3/DOC of 1 mg O3/mg DOC ratio [47, 48]. This study employed much lower O3/DOC (0.1 and 0.4 mg O3/mg DOC ratios), therefore little size shift by ozonation was expected. Similarly to SEC-UV, small fluorescent molecules exhibited similar AMW distribution between 1000 Da and 1700 Da regardless of fluorescent region (Fig. 6(b)). A major distinction from SECUV chromatogram, however, is that macromolecules (AMW>10k Da) were observed for Region 1, 2 and 4. These macromolecules are likely to be polysaccharides, proteins and/or amino sugars [49]. Almost complete reduction of fluorescence intensities of the macromolecules was observed at 0.4 mg O3/mg DOC ratio whereas molecules with AMW between 1000 Da and 1700 Da displayed lower extent of fluorescence decrease. This was attributed to the fact that reaction rate increase as increasing molecular size [50], which can be a reason for higher reduction of fluorescent for Region 1, 2 and 4 of EEM in the previous section. The effects of the reduction in the macromolecules on fouling potential will be further explained in the subsequent section. 3.5. Statistical correlation of spectroscopic organic properties with fouling potential In this study, MFI was calculated as an indicator of the fouling potential reduction of the brine by ozonation while spectral analyses described above were also conducted. Since there are many measured quantities that can explain the fouling potential reduction, statistical correlation can provide comprehensive comparison among the measured values. To this end, the chromatographic peaks were integrated under five AMW regions (named AMW 1 to AMW 5 from smaller to larger AMW ranges as shown in Fig. 7) and used as variables for the statistical analysis as well as bulk spectroscopic parameters such as SUVA and TF. Pearson-correlation test was performed for the applied pressure of 10 and 25 bar. Since the results for the both pressures

are almost identical, the correlation coefficients only for 10 bar were shown in Fig. 7. It should be noted that the linearity test (Pearson correlation test) can be meaningful based upon the premise that the MFI value is linearly proportional to the concentration of foulant conceptually. In general, positive correlation with MFI regardless of spectroscopic variables was found (correlation coefficient>0.79, Fig. 7) since all the spectroscopic variables as well as MFI monotonically reduced with increasing ozone dose. The macromolecules (AMW5) of tryptophan-like aromatic proteins (Region 2) and SMP-like matter (Region 4) displayed the highest correlation with MFI (correlation coefficient >0.97) (Fig. 7). These macromolecules are known to provoke the formation of highly-resistant cake layer [51] and hence influence MFI. Further, according to previous research [52], SMP exhibited fouling propensity in a given water compared to other model foulants tested including bovine serum albumin, colloidal NOM and Suwannee river humic acid. In addition to their high cake layer resistance, reduced deposition of macromolecules to membrane surface is also beneficial to mitigate the cake layer resistance in poly-dispersed feed water. In the feed water with the mixture of organic matter with different sizes (e.g., NOM), the void space of cake layer formed by macromolecules hinders back diffusion of small molecules within the cake layer and provokes greater flux decline compared to monodispersed foulant [53]. Hence, reduced cake layer of macromolecules by ozonation can mitigate the synergistic effects of the mixture. Along with the macromolecules, UVA254 of NOM fractions with AMW between 600 Da and 1600 Da (AMW 1 to 4) exhibited strong correlation with MFI too. In contrast, bulk parameters such as SUVA and TF displayed relatively lower correlation. Hence, specific size fractions of NOM such as fluorescent macromolecules (AMW>10k Da) and/or small UV-absorbing NOM (600 Da
however, further experiments using the combination of various water qualities and membranes would be valuable to bolster the finding. 3.6. Trace organic compounds (TOrCs) attenuation When greater recoveries from membrane processes are desired, the resulting brine will contain higher concentrations of organic contaminants. Therefore, it is important to evaluate the removal of TOrCs when treating brines via NF membranes. Fig. 8(a) illustrates the concentrations of TOrCs in the brine without ozonation and with the two ozone doses. Ozone oxidation and hence TOrC removal is due to molecular ozone and hydroxyl radical. [54]. More than 90 % removal of diltiazem (calcium channel blocker/antihypertensive) was observed at 0.1 mg O3/mg DOC ratio while the removal of the other compounds varied from 0 to 80 %. For instance, sulfamethoxazole, hydrocortisone, diphenhydramine, trimethoprim and naproxen (antibiotic, steroid, antihistamine, antibiotics and nonsteroidal anti-inflammatory drug (NSAID), respectively) had greater than 50 % removal at O3/DOC of 0.1 mg O3/mg DOC ratio. These compounds are known to have high reactivity with both ozone and hydroxyl radical [55]. The TOrCs with low removal rates such as ibuprofen, atrazine, and meprobamate (NSAID, herbicide and antianxiety drug, respectively) were attributed to low reactivity with ozone and hydroxyl radical [56]. At 0.4 mg O3/mg DOC ratio, all the tested TOrCs were attenuated at greater than 70%. Although some TOrCs such as clofibric acid and meprobamate possess low reactivity with molecular ozone, higher ozone doses resulted in greater hydroxyl radical formation, thereby attenuating recalcitrant TOrCs more efficiently. The NF membrane used in this study was very successful in treatment of TOrCs (>90% removal) in the given water quality (Fig. 8b). When combining membrane with pre-ozonation, all the

compounds were not detected above method detection limits (MDLs) whose values are shown in the supplementary material (Table S1). Due to the excellent rejection of NF90 membrane, the exact removal rates of TOrCs by the pre-ozonation followed by NF membrane could not be accurately quantified. Hence, possible synergism of the combination of NF membrane with preozonation is only theoretically proposed here. Roverall = RO3 + Rmembrane (1 - RO3 )

(5)

where Roverall , RO3, and Rmembrane indicate removal rate of TOrCs by the membrane with preozonation, ozone alone, and membrane alone, respectively. Its derivation procedure and a contour plot of Roverall with respect to RO3 and Rmembrane are shown in the supplementary material (Fig. S2). Based on Eq. (5), overall TOrCs attenuation is the sum of TOrCs attenuation by ozone ( RO3 ) and removal of remaining TOrCs by membrane after ozonation ( Rmembrane (1 - RO3 ) ), thereby provides synergism for TOrC attenuation. For instance, ozone has fairly good removal of diphenhydramine at 0.1 mg O3/mg DOC ratio (69.2%) whereas NF membrane removed greater than 92.3% of diphenhydramine (its concentration in the permeate was below MDL, therefore its removal is greater than 92.3%). Based on these values, theoretical overall removal of diphenhydramine of NF membrane with pre-ozonation at 0.1 mg O3/mg DOC ratio based on Eq. (5) is greater than 97.6%. Dual barrier approach with two different removal mechanisms is beneficial since one can compensate for the other while allowing redundancy in the treatment train in case of process failure in one of the systems. Therefore, it will be beneficial to use membrane with pre-ozonation for the management of TOrCs because not only can be higher removal of TOrCs in permeate achieved, but low residual of TOrCs in waste stream from membrane systems can also be obtained.

4. Conclusion Pre-ozonation was examined to reduce fouling potential of a highly-concentrated surface water NF brine (DOC ~18 mg/L). With relatively low ozone doses employed in this study (0.1 and 0.4 mg O3/mg DOC ratios), significant reduction in fouling potential quantified by MFI was achieved. The application of classical pore blocking model revealed that cake filtration was the dominant fouling mechanism and not altered after ozonation. The reduction of macromolecules (AMW>10k Da) of typtophan-like aromatic protein and SMP-like matter exhibited statistically strong correlation with MFI reduction by ozonation. Furthermore, small AMW fractions (AMW between 600 Da and 1600 Da) of UV-absorbing matter were better correlated with MFI whereas bulk parameters such as SUVA and TF displayed relatively lower correlation. It was found that the Dow NF90 membrane (NF90) used in this study was efficacious for TOrC attenuation (>90% removal) whereas the removal efficacy of TOrCs by ozone was dependent on reaction kinetics of each TOrC as well as ozone dose. With the combination of pre-ozonation with the NF membrane, all the TOrCs tested could not be detected above the MDLs, which indicates that the combined treatment process can function as a successful dual barrier for TOrC control. In conclusion, the reduction of organic fouling potential by pre-ozonation can be a viable option to achieve higher recovery of NF membrane systems and/or to treat organic-rich water such as brine. Furthermore, pre-ozonation enhanced overall attenuation of TOrCs in a NF membrane system as well as reduced organic fouling potential. Possible alteration of biofouling propensity by ozone in a long term operation is suggested as a future area of study.

Acknowledgement We thank CIRSEE—Suez Environnement for financial and technical support of this project. The authors would like to acknowledge Mojtaba Azadi Aghadam and Norma Nohemi VillagomezMarquez for assistance in brine generation. In addition, we also would like to thank Dr. Armando Durazo for assistance of analysis for metals. Minkyu Park was partially supported by a fellowship from Water Sustainability Program (WSP), University of Arizona. We also acknowledge Xylem Inc. for providing the ozone generators and Dow FilmTech for providing the NF membranes. Finally, the authors would like to thank Agilent Technologies for technical assistance with instruments employed for the measurement of TOrCs.

References [1] M.A. Shannon, P.W. Bohn, M. Elimelech, J.G. Georgiadis, B.J. Mariñas, A.M. Mayes, Science and technology for water purification in the coming decades, Nature, 452 (2008) 301310. [2] S.B. Grant, J.-D. Saphores, D.L. Feldman, A.J. Hamilton, T.D. Fletcher, P.L.M. Cook, M. Stewardson, B.F. Sanders, L.A. Levin, R.F. Ambrose, A. Deletic, R. Brown, S.C. Jiang, D. Rosso, W.J. Cooper, I. Marusic, Taking the “Waste” Out of “Wastewater” for Human Water Security and Ecosystem Sustainability, Science, 337 (2012) 681-686. [3] M. Elimelech, W.A. Phillip, The future of seawater desalination: energy, technology, and the environment, Science, 333 (2011) 712-717. [4] K. Kimura, G. Amy, J.E. Drewes, T. Heberer, T.-U. Kim, Y. Watanabe, Rejection of organic micropollutants (disinfection by-products, endocrine disrupting compounds, and pharmaceutically active compounds) by NF/RO membranes, Journal of Membrane Science, 227 (2003) 113-121. [5] M. Soltanieh, W.N. Gill, REVIEW OF REVERSE OSMOSIS MEMBRANES AND TRANSPORT MODELS, Chemical Engineering Communications, 12 (1981) 279-363. [6] S.-Y. Kwak, S.H. Kim, S.S. Kim, Hybrid Organic/Inorganic Reverse Osmosis (RO) Membrane for Bactericidal Anti-Fouling. 1. Preparation and Characterization of TiO2 Nanoparticle Self-Assembled Aromatic Polyamide Thin-Film-Composite (TFC) Membrane, Environmental Science & Technology, 35 (2001) 2388-2394. [7] E.C. Wert, F.L. Rosario-Ortiz, D.D. Drury, S.A. Snyder, Formation of oxidation byproducts from ozonation of wastewater, Water Research, 41 (2007) 1481-1490.

[8] A.W. Zularisam, A.F. Ismail, R. Salim, Behaviours of natural organic matter in membrane filtration for surface water treatment — a review, Desalination, 194 (2006) 211-231. [9] S.G.J. Heijman, H. Guo, S. Li, J.C. van Dijk, L.P. Wessels, Zero liquid discharge: Heading for 99% recovery in nanofiltration and reverse osmosis, Desalination, 236 (2009) 357-362. [10] L.F. Greenlee, D.F. Lawler, B.D. Freeman, B. Marrot, P. Moulin, Reverse osmosis desalination: Water sources, technology, and today's challenges, Water Research, 43 (2009) 2317-2348. [11] P. Chelme-Ayala, D.W. Smith, M.G. El-Din, Membrane concentrate management options: a comprehensive critical reviewA paper submitted to the Journal of Environmental Engineering and Science, Canadian Journal of Civil Engineering, 36 (2009) 1107-1119. [12] Florida-Senate, Leah Schad Memorial Ocean Outfall Program, in: 403.08601, 2009. [13] G.W. Lawton, C.V. Eggert, Effect of High Sodium Chloride Concentration on Trickling Filter Slimes, Sewage and Industrial Wastes, 29 (1957) 1228-1236. [14] R. Bond, B. Batchelor, T. Davis, B. Klayman, Zero liquid discharge desalination of brackish water with an innovative form of electrodialysis: Electrodialysis metathesis, Florida Water Resources Journal, 63 (2011) 36-44. [15] A. Robertson, L. Nghiem, Treatment of high TDS liquid waste: Is zero liquid discharge feasible?, (2011). [16] T. Tong, M. Elimelech, The Global Rise of Zero Liquid Discharge for Wastewater Management: Drivers, Technologies, and Future Directions, Environmental Science & Technology, 50 (2016) 6846-6855. [17] S. Van Geluwe, C. Vinckier, L. Braeken, B. Van der Bruggen, Ozone oxidation of nanofiltration concentrates alleviates membrane fouling in drinking water industry, Journal of Membrane Science, 378 (2011) 128-137. [18] A. Bes-Pía, A. Iborra-Clar, J.A. Mendoza-Roca, M.I. Iborra-Clar, M.I. Alcaina-Miranda, Nanofiltration of biologically treated textile effluents using ozone as a pre-treatment, Desalination, 167 (2004) 387-392. [19] S.-H. You, D.-H. Tseng, W.-C. Hsu, Effect and mechanism of ultrafiltration membrane fouling removal by ozonation, Desalination, 202 (2007) 224-230. [20] B.D. Stanford, J.F. Leising, R.G. Bond, S.A. Snyder, Chapter 11 Inland Desalination: Current Practices, Environmental Implications, and Case Studies in Las Vegas, NV, in: C.E. Isabel, I.S. Andrea (Eds.) Sustainability Science and Engineering, Elsevier, 2010, pp. 327-350. [21] B.D. Stanford, A.N. Pisarenko, R.D. Holbrook, S.A. Snyder, Preozonation Effects on the Reduction of Reverse Osmosis Membrane Fouling in Water Reuse, Ozone: Science & Engineering, 33 (2011) 379-388. [22] G.D. Wetterau, R.B. Chalmers, P. Liu, W. Pearce, Advancing indirect potable reuse in California, Water Practice and Technology, 8 (2013). [23] K.J. Ormerod, C.A. Scott, Drinking Wastewater: Public Trust in Potable Reuse, Science, Technology & Human Values, 38 (2013) 351-373. [24] K.L. Rakness, E.C. Wert, M. Elovitz, S. Mahoney, Operator-Friendly Technique and Quality Control Considerations for Indigo Colorimetric Measurement of Ozone Residual, Ozone: Science & Engineering, 32 (2010) 33-42. [25] B. Kwon, J. Cho, N. Park, J. Pellegrino, Organic nanocolloid fouling in UF membranes, Journal of Membrane Science, 279 (2006) 209-219. [26] J.R. Lakowicz, Principles of Fluorescence Spectroscopy, 3rd ed., Springer, 2006.

[27] W. Chen, P. Westerhoff, J.A. Leenheer, K. Booksh, Fluorescence Excitation−Emission Matrix Regional Integration to Quantify Spectra for Dissolved Organic Matter, Environmental Science & Technology, 37 (2003) 5701-5710. [28] T. Anumol, S.A. Snyder, Rapid analysis of trace organic compounds in water by automated online solid-phase extraction coupled to liquid chromatography–tandem mass spectrometry, Talanta, 132 (2015) 77-86. [29] J. Glater, M.R. Zachariah, S.B. McCray, J.W. McCutchan, Reverse osmosis membrane sensitivity to ozone and halogen disinfectants, Desalination, 48 (1983) 1-16. [30] J. Glater, S.-k. Hong, M. Elimelech, The search for a chlorine-resistant reverse osmosis membrane, Desalination, 95 (1994) 325-345. [31] M.-O. Buffle, J. Schumacher, E. Salhi, M. Jekel, U. Von Gunten, Measurement of the initial phase of ozone decomposition in water and wastewater by means of a continuous quench-flow system: application to disinfection and pharmaceutical oxidation, Water research, 40 (2006) 1884-1894. [32] A. Braghetta, F. DiGiano, W. Ball, Nanofiltration of Natural Organic Matter: pH and Ionic Strength Effects, Journal of Environmental Engineering, 123 (1997) 628-641. [33] K. Listiarini, D.D. Sun, J.O. Leckie, Organic fouling of nanofiltration membranes: Evaluating the effects of humic acid, calcium, alum coagulant and their combinations on the specific cake resistance, Journal of Membrane Science, 332 (2009) 56-62. [34] J. Schippers, J. Verdouw, The modified fouling index, a method of determining the fouling characteristics of water, Desalination, 32 (1980) 137-148. [35] S.F.E. Boerlage, M.D. Kennedy, M.R. Dickson, D.E.Y. El-Hodali, J.C. Schippers, The modified fouling index using ultrafiltration membranes (MFI-UF): characterisation, filtration mechanisms and proposed reference membrane, Journal of Membrane Science, 197 (2002) 1-21. [36] L. Song, K.L. Chen, S.L. Ong, W.J. Ng, A new normalization method for determination of colloidal fouling potential in membrane processes, Journal of colloid and interface science, 271 (2004) 426-433. [37] J. Hermia, Constant pressure blocking filtration laws-Applications to power-law nonnewtonian fluids, Trans. IChemE., 60 (1982) 183-187. [38] S.-J. Lee, M. Dilaver, P.-K. Park, J.-H. Kim, Comparative analysis of fouling characteristics of ceramic and polymeric microfiltration membranes using filtration models, Journal of Membrane Science, 432 (2013) 97-105. [39] L. Palacio, C.-C. Ho, A.L. Zydney, Application of a pore-blockage—Cake-filtration model to protein fouling during microfiltration, Biotechnology and Bioengineering, 79 (2002) 260-270. [40] W. Yuan, A. Kocic, A.L. Zydney, Analysis of humic acid fouling during microfiltration using a pore blockage–cake filtration model, Journal of Membrane Science, 198 (2002) 51-62. [41] E.M.V. Hoek, M. Elimelech, Cake-Enhanced Concentration Polarization:  A New Fouling Mechanism for Salt-Rejecting Membranes, Environmental Science & Technology, 37 (2003) 5581-5588. [42] J. Świetlik, A. Dąbrowska, U. Raczyk-Stanisławiak, J. Nawrocki, Reactivity of natural organic matter fractions with chlorine dioxide and ozone, Water Research, 38 (2004) 547-558. [43] E.C. Wert, F.L. Rosario-Ortiz, S.A. Snyder, Effect of ozone exposure on the oxidation of trace organic contaminants in wastewater, Water Research, 43 (2009) 1005-1014. [44] S. Hong, M. Elimelech, Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes, Journal of Membrane Science, 132 (1997) 159-181.

[45] Á. de la Rubia, M. Rodríguez, V.M. León, D. Prats, Removal of natural organic matter and THM formation potential by ultra- and nanofiltration of surface water, Water Research, 42 (2008) 714-722. [46] H.M. Krieg, S.J. Modise, K. Keizer, H.W.J.P. Neomagus, Salt rejection in nanofiltration for single and binary salt mixtures in view of sulphate removal, Desalination, 171 (2005) 205-215. [47] E.C. Wert, S. Gonzales, M.M. Dong, F.L. Rosario-Ortiz, Evaluation of enhanced coagulation pretreatment to improve ozone oxidation efficiency in wastewater, Water Research, 45 (2011) 5191-5199. [48] P. Bose, B.K. Bezbarua, D.A. Reckhow, Effect Of Ozonation On Some Physical and ChemicalProperties Of Aquatic Natural Organic Matter, Ozone: Science & Engineering, 16 (1994) 89-112. [49] C. Jarusutthirak, G. Amy, Understanding soluble microbial products (SMP) as a component of effluent organic matter (EfOM), Water Research, 41 (2007) 2787-2793. [50] P. Westerhoff, G. Aiken, G. Amy, J. Debroux, Relationships between the structure of natural organic matter and its reactivity towards molecular ozone and hydroxyl radicals, Water Research, 33 (1999) 2265-2276. [51] C. Jarusutthirak, G. Amy, Role of Soluble Microbial Products (SMP) in Membrane Fouling and Flux Decline, Environmental Science & Technology, 40 (2006) 969-974. [52] N. Park, B. Kwon, I.S. Kim, J. Cho, Biofouling potential of various NF membranes with respect to bacteria and their soluble microbial products (SMP): Characterizations, flux decline, and transport parameters, Journal of Membrane Science, 258 (2005) 43-54. [53] A.E. Contreras, A. Kim, Q. Li, Combined fouling of nanofiltration membranes: Mechanisms and effect of organic matter, Journal of Membrane Science, 327 (2009) 87-95. [54] U. von Gunten, Ozonation of drinking water: Part I. Oxidation kinetics and product formation, Water Research, 37 (2003) 1443-1467. [55] D. Gerrity, S. Gamage, D. Jones, G.V. Korshin, Y. Lee, A. Pisarenko, R.A. Trenholm, U. von Gunten, E.C. Wert, S.A. Snyder, Development of surrogate correlation models to predict trace organic contaminant oxidation and microbial inactivation during ozonation, Water Research, 46 (2012) 6257-6272. [56] Y. Lee, D. Gerrity, M. Lee, A.E. Bogeat, E. Salhi, S. Gamage, R.A. Trenholm, E.C. Wert, S.A. Snyder, U. von Gunten, Prediction of Micropollutant Elimination during Ozonation of Municipal Wastewater Effluents: Use of Kinetic and Water Specific Information, Environmental Science & Technology, 47 (2013) 5872-5881.

Table 1 Bulk water quality parameters and metal concentrations Parameter

Colorado river water (CRW)

Brine

Brine with 0.1

Brine with 0.4

mg O3/mg C

mg O3/mg C

DOC (mg/L)

3.5

18.3

16.7

17.2

pH

8.1

8.3

8.2

8.2

Alkalinity (mg/L CaCO3)

130

310

302

308

Conductivity (μS/cm)

946

2360

2310

2340

Na (mg/L)

65.7

212

212

212

Mg (mg/L)

19.1

101

112

107

K (mg/L)

4.9

11.4

11.3

11.3

Ca (mg/L)

77.5

291

285

278

Ba (mg/L)

0.08

0.54

0.50

0.53

Highlights ·

Brine treatment using nanofiltration (NF) membrane with pre-ozonation

·

Fouling reduction by ozone oxidation

·

Reduced aromaticity of NOM after ozonation was major fouling reduction mechanism

·

Cake filtration was dominant fouling mechanism regardless of ozonation

·

Synergism of NF membrane with pre-ozonation for trace organic compounds attenuation

Figure captions Fig. 1. Ozone decay curve of NF brine (8 mg/L ozone applied resulting in instantaneous ozone demand (IOD) 7.3 mg/L). Fig. 2. Flux decline of brine with and without ozonation. Two ozone doses (0.1 mg O3/mg C and 0.4 mg O3/mg C, respectively) were employed for the applied pressure of (a) 25 bars and (b) 10 bars. Fig. 3. Filtration curves for the brine with and without ozonation for the applied pressure of (a) 25 bars and (b) 10 bars. (c) Modified fouling index (MFI) values were obtained by calculating slopes of t/V versus V for each fouling propensity test. Fig. 4. Flux decline analysis for the brine with and without ozonation for the applied pressure of (a) 25 bars and (b) 10 bars. Spline curve fitting was applied to help show trend. Fig. 5. (a) Specific UV absorbance (SUVA) in the left axis and normalized total fluorescence (TF) in the right axis. (b) Excitation-emission matrix (EEM) of the brine. Operationally-defined

excitation and emission boundaries into five regions. Region 1 to 5 refers to tyrosine-like aromatic protein, tryptophan-like aromatic protein, fulvic acid-like matter, soluble microbial product (SMP)-like matter and humic acid-like matter, respectively. (c) Relative fluorescence integrated in each operationally-defined excitation/emission wavelength region. Fig. 6. Apparent molecular weight (AMW) distribution measured by size-exclusion chromatography (SEC) with (a) a UV-diode array detector (254 nm) and (b) fluorescence detector with the five pairs of excitation/emission wavelengths. Fig. 7. Correlation coefficients of MFI with spectroscopic variables for the applied pressure of 10 bar. Fig. 8. (a) Removal of trace organic contaminants (TOrCs) by ozonation. Note that the concentration of compounds marked with asterisk for the brine with 0.4 mg O3/mg C was below method detection limit (MDL). All the compounds in the brine with 0.1 mg O3/mg C were detected above MDL. (b) Removal of TOrCs by NF membrane filtration without ozonation. The concentration of the compounds with asterisk exhibited below MDL. MDL values are available in the supplementary information (Table S1).

Park et. al

Fig. 1

Park et. al

Fig. 2.

Park et. al

Fig. 3

Park et. al

Fig. 4

Park et. al

Fig. 5

Fig. 6 Park et. al

Fig. 7 Park et. al

Fig. 8. Park et. al

33