Viability of nanofiltration and ultra-low pressure reverse osmosis membranes for multi-beneficial use of methane produced water

Separation and Purification Technology 52 (2006) 67–76

Viability of nanofiltration and ultra-low pressure reverse osmosis membranes for multi-beneficial use of methane produced water Pei Xu ∗ , J¨org E. Drewes Environmental Science & Engineering Division, Colorado School of Mines, Golden, CO 80401-1887, USA Received 25 October 2005; received in revised form 22 March 2006; accepted 23 March 2006

Abstract Produced water management has become one of the key factors to sustainable development of natural gas/oil resources. The substantial quantities of saline water present intractable environmental threats and also increase oil/gas production costs through produced water disposal such as deep well reinjection. Developing high-efficient and flexible treatment systems that can be operated at low costs is of high interests for producers and state regulators. Beneficial use of produced water could represent a new water resource especially for areas with inadequate existing supplies. Furthermore, some produced waters generated are also characterized by elevated concentrations of recoverable constituents, for example iodide. Recovering iodide from brine could offer additional benefits besides providing methane gas, reusing produced water, or reducing brine disposal volume. The advent of ultra-low pressure reverse osmosis (ULPRO) membranes and nanofiltration (NF) membranes with high desalting degree might offer a viable option for produced water treatment because these membranes can be as effective as reverse osmosis (RO) in removing certain solutes from water while requiring considerably less feed pressure resulting in lower operating costs. The objectives of this research were to investigate the viability of ULPRO and NF membranes as potential techniques to treat produced water by meeting water quality standards and concentrating iodide in the brine. The produced water extracted from sandstone aquifer in Eastern Montana was characterized as brackish groundwater of sodium chloride type with total dissolved solids (TDS) concentration of 5300 mg/L, absence of hydrocarbons, and average iodide concentrations of 55 mg/L. The produced water exhibited a very high potential to membrane fouling indicated by silt density index (SDI) measurements due to the presence of small particles and inorganic constituents. The studied candidate membranes included one RO membrane (TFC-HR, Koch Membrane Systems), three ULPRO membranes XLE (Dow/Filmtec), TFC-ULP (Koch) and TMG-10 (Toray America), and three NF membranes NF-90 (Dow/Filmtec), TFC-S (Koch), and ESNA (Hydranautics). Bench-scale cross-flow flat sheet test units were employed to assess the candidate membranes using the produced water with focus on fouling potential, iodide recovery, and general salt rejection. The degree of flux decline was found to be dependent upon the combination of permeate drag force and physico-chemical properties of the membranes. The membranes with higher permeability generally displayed faster initial flux decline. In addition, hydrophobic and rough membranes exhibited a higher flux decline and lower chemical cleaning efficiency than smooth and/or hydrophilic membranes. Flux decline experiments, in situ microscopic techniques, analysis of elemental composition and functional groups revealed that the pretreatment including microfiltration, pH adjustment and addition of antiscalants could alleviate membrane fouling significantly. Chemical cleaning using caustic and anionic surfactant solutions restored membrane permeability more efficiently than hydraulic cleaning or using acids and metal chelating agents. This study showed that TFC-ULP, TMG-10, and NF-90 membranes exhibited competitive efficiency regarding salt rejection, iodide recovery and adjusted specific flux as compared to a conventional RO membrane. © 2006 Elsevier B.V. All rights reserved. Keywords: Produced water; Membrane technology; Desalination; Membrane fouling; Water reuse

1. Introduction

∗

Corresponding author. Tel.: +1 303 273 3932; fax: +1 303 273 3413. E-mail addresses: [email protected] (P. Xu), [email protected] (J.E. Drewes). 1383-5866/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.seppur.2006.03.019

Exploration of natural gas fields is growing very rapidly to meet increased energy demand in the world. Large volumes of produced water are generated in gas production by dewatering the trapped gas bubbles from water. Produced water management has become one of the key factors in the feasibility of

68

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

gas field development. In the United States, more than 60% by volume of produced water is presently reinjected into specially designed wells which are deemed to be geologically isolated from potential underground sources of drinking water [1]. The reinjection costs vary at US$ 0.50–1.75 per barrel in wells with installations costs from US$ 400,000 to 3,000,000 per well [2]. Although this is a widely adopted technique for disposing produced water, many industries are seeking alternative methods for cost-effective produced water management due to limited reinjection capacities and increased regulatory restrictions. Since regions where methane exploration is underway often are lacking sufficient supplies for drinking water and irrigation needs, beneficial use of produced water has become an attractive solution to produced water management by providing additional and reliable water supplies and reducing the cost for disposal of produced water. Produced waters vary widely in composition since they originate from different geological formations, gas hydrocarbon compositions, and differ by well development and maintenance. This affects significantly the technical and economic feasibility of employing treatment technologies for utilizing produced water for beneficial use and meeting regulatory criteria for the targeted end use. The common constituents usually found in produced water include organic compounds such as oil, grease, benzene, and phenol; and inorganic compounds such as sodium, potassium, iron, calcium, magnesium, chloride, sulfate, carbonate, bicarbonate, silicate, and borate. Based on the produced water quality and composition, desalination is often imperative to meet potable and non-potable water reuse standards. High-pressure membranes such as reverse osmosis (RO) have been used to desalinate seawater and brine water for more than 30 years and could offer a possible solution for a beneficial treatment of produced water. Previous pilot tests have shown that good quality water could be produced by removing a large fraction of organic and inorganic constituents from produced water [3,4]. Membrane fouling, however, deteriorates the membrane performance very quickly and results in increased operation costs due to increasing operating pressure [1,5]. Depending upon the produced water composition, free and dissolved oil can adhere onto membrane surface resulting in a loss of permeability. The soluble hydrocarbons including methane, volatile acids and BTEX (benzene, toluene, ethylbenzene, and xylene) can support growth of a biofilm on membrane surfaces. Particles and soluble salts can precipitate onto membrane surfaces causing membrane scaling. Rigorous and complex pre-treatment is often required for the success of RO membrane application. The advent of ultra-low pressure RO (ULPRO) membranes and nanofiltration (NF) membranes with high desalting degree offers a viable option for produced water treatment because they can be as effective as RO in removing certain solutes from water while requiring considerably less feed pressure. The produced water studied was generated during methane exploration from sandstone aquifers. The tested water was characterized as brackish groundwater of sodium chloride type with TDS concentration of 5300 mg/L, with no detectable hydrocarbons but elevated concentrations of iodide. Iodine is an essential and rare element with increasing demand in many industrial

applications. The major end-uses of iodine include animal feed supplements, catalysts, inks, and colorants, pharmaceuticals, photographic chemicals and films, sanitary and industrial disinfectants, and stabilizers [6]. Key elements of this investigation were to select appropriate NF and ULPRO membranes to treat produced water with respect to iodide recovery and water qualities suitable for non-potable and potable reuse. The study also identified pretreatment, fouling, and cleaning issues of membranes employed for produced water treatment through flux decline experiments and in situ membrane characterization techniques. 2. Materials and methods 2.1. Water quality The water sample collected at a natural gas production facility in Eastern Montana was frozen after sample collection and no preservation of the sample occurred during shipping. A comprehensive water analysis was conducted employing Standard Methods [7]. Iodide was analyzed using an EA 940 Expanded Ionanalyzer (Orion Research Inc., Boston, MA) with ThermoOrion iodide and Ag/AgCl electrodes. Inorganic cations were measured with an Optima 3000 inductive coupled plasma (ICP) Spectrometer (Perkin Elmer, Norwalk, CT). Samples were prepared by 0.45 ␮m filtration and then acidified to pH 2.0 using phosphoric acid prior to analysis. Inorganic anions were measured with a Dionex DS600 Ion Chromatograph (IC) (Dionex, Sunnyvale, CA) using an AS14A column and a sodium hydroxide eluent. Prior to IC analysis, samples were prepared by 0.45 ␮m filtration. The produced water from a sandstone aquifer was characterized as a brackish groundwater of sodium chloride type with a pH of 8.45 ± 0.26. Total dissolved solids (TDS) concentration was quantified as 5243 ± 561 mg/L with a specific conductance of 9647 ± 652 ␮S/cm. Beside sodium and chloride, major constituents with concentrations of less than 100 mg/L were calcium, magnesium, bromide and iodide. The well water was relatively rich in iodide with a concentration of 55.6 ± 10.8 mg/L. The water was classified as hard (total hardness of 123 ± 25 mg/L as CaCO3 ) with an alkalinity of 195 ± 30 mg/L as CaCO3 . Minor constituents with concentrations of less than 10 mg/L were aluminum, boron, barium, potassium, silicon, and strontium. Due to the elevated pH of the produced water, iron was present in the form of small particulate precipitates while the dissolved iron concentration was below the detection limit of 2 ␮g/L using ICP. No other inorganic constituents were detected in the well water above the detection limit of the analytical methods. The dissolved organic carbon (DOC) concentration in the well water was 1.75 ± 0.20 mg/L and organics were characterized by moderate to high aromaticity (specific UV absorbance equals 4.0 ± 0.45 L/(mg m)). Hydrocarbons and BTEX compounds, however, were not detected in the well water sample provided. Membrane fouling potential was assessed by the silt density index (SDI) calculated from the rate of plugging of a 0.45 ␮m

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

membrane filter (47 mm in diameter, Nuclepore polyester, Whatman, Clifton, NJ). The SDI was measured at 30 psi using the standard test method D4189-95 (2002) according to Eq. (1). Deionized water (type II) was used as non-logging reference water. Since the time to collect 500 mL of the tested water represented more than 110% of the non-plugging time, a sample size of 100 mL was used for the SDI measurement.      %P30 100 ti SDI = × = 1− (1) T tf T where %P30 is the percent at 30 psi feed pressure, T the total elapsed flow time (min), ti the time to collect initial 100 mL of sample (s), tf the time to collect final 100 mL of sample (s) after the time T. For the SDI measurement, total elapsed flow time, T, is often set for 15 min. However due to the high fouling potential of the tested water, the %P30 exceeded 75% in this study, which is the recommended value of the method. Thus 5 min was used as the total elapsed time T. Water temperature before and after the test was 19.4 ◦ C. Even for a total elapsed time of 5 min, the tf required to collect a final volume of 100 mL water was too long to calculate an appropriate SDI value. Therefore, SDI could not be used to analyze the particulate matter in the untreated well water. Microfiltration was selected as a pretreatment to avoid membrane fouling. Subsequently, the well water sample was processed through a filtration device using a 5 ␮m cartridge and a 0.45 ␮m filter bag (Cole-Parmer, Vernon Hills, IL). The microfiltered water sample had a SDI of 19.2 ± 0.1, representing a very high value for membrane treatment. A SDI range of approximately 3–5 has been found to result in successful operation of spiral-wound RO membranes. The elevated value pointed to the fact that some precipitates would quickly build-up at the membrane surface potentially resulting in a substantial declining flux. 2.2. Selected membranes and characterization Membranes selected for this study are characterized as thinfilm composite polyamide membranes and are all commercially available in the United States. The candidate membranes included one RO membrane (TFC-HR, Koch Membrane Systems), three ULPRO membranes XLE (Dow/Filmtec), TFCULP (Koch) and TMG-10 (Toray America), and three NF membranes NF-90 (Dow/Filmtec), TFC-S (Koch), and ESNA (Hydranautics). The RO and ULPRO membranes TFC-HR,

69

TMG-10, TFC-ULP, and XLE had MWCO below 100 Da and high salt rejection above 99% reported by the manufacturers. The NF membranes NF-90, TFC-S, and ESNA had a MWCO of 200 Da with varying desalting degree from 90%, 85% to 70%, respectively. All the studied membranes were prepared by interfacial polymerization of mphenylenediamine. Due to different reactants employed during membrane treatment and modification, the membranes exhibited a range of different physico-chemical properties as described in Table 1. Membrane hydrophobicity was characterized by sessile drop contact angle measurement using a NRL Goniometer-Model 100-00 (Ram´e-hart, Inc., Surface Science Instrument, Landing, NJ). Membrane surface structure and morphology was characterized by environmental scanning electron microscopy (ESEM) Quanta 600 (FEI Company, Hillsboro, OR) and a digital instrument atomic force microscopy (AFM) mounted in MultiMode Scanning Probe Microscopy (Digital Instruments, Santa Barbara, CA) in tapping mode. Elemental compositions of virgin and fouled membrane specimens were quantified by the energy dispersive spectroscopy (EDS) equipped in the ESEM. Prior to microscopic analysis, the membrane specimens were coated with a thin layer of gold in a Hummer VI sputtering system (Technic Inc., Providence, RI) for ESEM imaging; or coated with a carbon layer in a Denton DV-502 vacuum evaporator (Cherry Hill, NJ) for EDS analysis. Functional group characteristics of virgin and fouled membrane specimens were measured using a Nicolet Nexus 870 Fourier transform infrared (FTIR) spectrometer (Nicolet, Madison, WI) with a ZnSe flat plate crystal. Using a liquid nitrogen cooled mercury cadmium telluride (MCT) detector, the spectra were recorded by the attenuated total reflection (ATR) method with 500 scans and a wave number resolution of 2.0 cm−1 . Virgin membrane specimens were thoroughly rinsed in deionized water and stored at 4 ◦ C. Prior to membrane characterization, virgin, and fouled membrane specimens were rinsed with deionized water and dried at room temperature for 24 h. To reduce the interference of water in membrane samples, specimens were dried in a desiccator for 3 days prior to FTIR measurement. The ESEM micrographs and AFM images showed the virgin membranes all displayed a ridge-valley structure with different surface roughness (from 29 nm for ESNA to 73 nm for XLE and TFC-S). All membranes were negatively charged at pH 6 [8–11]. Based on contact angle measurements, the active layer of TFC-HR and TFC-ULP was characterized as hydrophilic and all other membranes were hydrophobic. In this study, the pure

Table 1 Characterization of membrane properties Membrane type

TFC-HR

XLE

TMG-10

TFC-ULP

TFC-S

NF-90

ESNA

Salt rejection (manufacturer, %) pH range (long term) Pure water permeability [L/(m2 day kPa)] (25 ◦ C) Contact angle (◦ ) Mean roughness (nm)

99.4 4–11 0.84 35 64

99 4–11 2.16 66 73

99.4 2–11 2.20 55 44

99 4–11 1.95 38 42

85 4–11 2.64 57 73

90 4–11 2.49 63 64

70 3–10 1.05 57 29

70

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

water permeability (at 25 ◦ C) of the seven selected membranes varied from 0.84 to 2.64 L/(m2 day kPa). 2.3. Membrane filtration tests Two standard laboratory cross-flow membrane filtration units (Sepa CF II, GE Osmonics, Minnetonka, MN) were employed in rejection tests and membrane fouling experiments. The cells have dimensions of 14.6 cm × 9.5 cm × 0.86 mm for channel length, width, and height, respectively. These channel dimensions provide an effective membrane area of 139 cm2 per unit and a cross-sectional flow area of 0.82 cm2 . Given the channel height of 34 mil (0.86 mm) and controlled flow rate, the test cell can simulate hydrodynamic conditions of a spiral wound element that often has a spacer thickness of 31 mil. The test cells of the units were rated for operating pressures up to 100 psi (689.5 kPa). The experiments were carried out at a pressure of 80 psi (551.6 kPa) and a feed flow rate of 500 mL/min, resulting in a feed cross-flow velocity of 10 cm/s. The feed water temperature was kept at 11 ± 1 ◦ C by a stainless steel water cooling system, simulating the in situ well water temperature. Feedwater to the units was the microfiltered produced water processed through a 5 ␮m cartridge and a 0.45 ␮m filter bag (Cole-Parmer, Vernon Hills, IL). To avoid inorganic scaling, the pH of the microfiltered produced water was adjusted to 6.0 using HCl solution. Approximately 50 L of the microfiltered water was stored in a drum and pumped into the parallel membrane test cell set-up by splitting feedwater into two parallel streams. The fouling tests were operated in a recycling mode where all concentrates and permeates were recirculated into the feed drum. The membrane specimens were preserved for membrane characterization and future reference. Prior to fouling and rejection experiments, virgin membrane specimens were placed in the units and rinsed with deionized water at 80 psi for 2 h to compact the membrane and eliminate impurities attached to the membrane surface. Pure water permeability was recorded for each membrane specimen to ensure that the membranes used for the experiments were comparable. The effect of adding antiscalants on flux decline was assessed using the antiscalant Hypersperse MDC700 from GE Betz (Trevose, PA) at concentration of 3 mg/L. In addition to hydraulic cleaning, the efficiency of chemical cleaning was examined with hydrochloric acid, citric acid, NaOH as an alkaline solution, sodium ethylenediaminetetraacetate (EDTA) as a metal chelating agent and sodium dodecyl sulfate (SDS) as an anionic surfactant. These chemical agents are common ingredients in commercial chemical cleaning solutions for organic and inorganic fouled membranes. Hydraulic and chemical cleaning was conducted for 10 min at ambient temperature. After chemical cleaning, the membrane was flushed with type II water for 10 min to remove any residual agents and to measure pure water permeability. All chemicals used were of reagent grade from Mallinckrodt (St. Louis, MO), and Fisher Scientific Inc. (Fairlawn, NJ). Deionized water (DI) (type II) was obtained from a laboratory water purification system (U.S. Filters, Warrendale, PA).

3. Results and discussion 3.1. Potential of membrane fouling during produced water treatment The SDI measurement indicated a serious fouling potential of the produced water without appropriate pretreatment. The permeate flux decline of NF-90 and TMG-10 with and without addition of antiscalants is compared in Fig. 1. The addition of antiscalants solution resulted in significantly lower flux declines during produced water filtration for the tested membranes. As the flux decline stabilized after 48 h with the addition of antiscalants (Fig. 2), the filtration time used to assess membrane performance was shortened to 75 h in the study. The normalized permeate flux for all the studied membranes is shown in Fig. 3. The TFC-HR, TFC-ULP, TMG-10, and ESNA membranes exhibited a similar degree of fouling resulting in a permeate flux decline of 23% over 75 h. The NF-90 and TFC-S membranes showed a lower fouling resistance to produced water with a permeate flux decline of 34% and 37%, respectively, over the course of the experiment. The XLE membrane was observed the least resistant to fouling with a permeate flux decline of 52%. All the tested membranes exhibited an initial high permeate flux decline in 20 h and reached stable flux conditions throughout the remainder of the experiments. Several factors contributing to membrane fouling have been identified in previous studies, including surface charge [12,13],

Fig. 1. Reduction in permeate flux over time (a) NF-90 membrane and (b) TMG-10 membrane using 0.45 ␮m microfiltered produced water without and with addition of 3 mg/L antiscalants. Applied pressure 80 psi, pH 6.0, temperature 11 ± 1 ◦ C, initial specific permeate flux 0.38 L/(m2 day kPa) (NF-90) and 0.24 L/(m2 day kPa) (TMG-10).

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

71

Fig. 4. Normalized permeate flux (J/Jo ) vs. initial specific flux and pure water permeability. Fig. 2. Reduction in permeate flux over time using 0.45 ␮m microfiltered produced water with addition of 3 mg/L antiscalants. Applied pressure 80 psi, pH 6.0, temperature 11 ± 1 ◦ C, initial specific permeate flux 0.30 L/(m2 day kPa) (TFC-S) and 0.36 L/(m2 day kPa) (XLE).

hydrophobicity [14], roughness [15], molecular weight cutoff [16], initial permeate rate, and drag force [12,17], feedwater quality, such as natural organic matter [18–20], monovalent and divalent cations [12] and microorganisms [21]. To better understand the factors affecting membrane fouling during produced water treatment, correlations between normalized permeate flux (J/J0 ) and specific flux, hydrophobicity (contact angle), and membrane morphology (surface roughness) were plotted. Membranes with higher permeability generally displayed more severe flux decline during filtration (Fig. 4). The regression coefficients (R2 ) of these relationships were low, implying that flux decline could not be predicted accurately from the permeability alone. The scattered data points indicated that other factors likely also contribute to flux decline besides permeability. A correlation with physico-chemical properties of the membrane suggested that the flux decline was also dependent upon the hydrophophicity and roughness of the virgin membranes (Fig. 5). Hydrophilic and/or smooth membrane surfaces, such as TFC-HR, TFCULP, ESNA, and TMG-10, exhibited a less potential to interact with the inorganic and organics components in produced water, thus having a lower flux decline during filtration. Rough and

Fig. 3. Reduction in permeate flux over time using 0.45 ␮m microfiltered produced water water with addition of 3 mg/L antiscalants. Applied pressure 80 psi, pH 6.0, temperature 11 ± 1 ◦ C, initial specific permeate flux (L/(m2 day kPa)) 0.12 (TFC-HR), 0.25 (TFC-ULP), 0.24 (TMG-10), 0.38 (NF-90), 0.30 (TFC-S), 0.36 (XLE) and 0.28 (ESNA).

hydrophobic membranes, such as XLE, TFC-S, and NF-90, however, showed a higher degree of flux decline over the course of the filtration experiment. Overall, membrane fouling during produced water treatment was determined by the combined effect of membrane permeability, hydrophobicity, and roughness. 3.2. Characterization of membrane fouling The ESEM micrographs of the virgin NF-90 membrane specimen and NF-90 membranes fouled in the presence and absence of antiscalants are compared in Fig. 6. After 75 h of filtration without addition of antiscalants, the NF-90 virgin membrane surface was completely covered by a gel-like fouling layer (Fig. 6(b)). With the addition of antiscalants, membrane fouling was significantly alleviated within the same filtration time. The ESEM graph supports that the ridge-valley surface of the virgin membrane was only partially covered by the fouling layer (Fig. 6(c)). The ESEM micrographs of other membranes also indicated that the addition of antiscalants prevented precipitation of foulants on the membrane surface and no significant foulants were observed on the membranes contacted with the produced water within the first 75 h (Fig. 7).

Fig. 5. Correlation of normalized permeate flux (J/Jo ) as a function of contact angle and surface roughness of virgin membranes.

72

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

Fig. 6. ESEM micrographs of: (a) NF-90 virgin membrane, (b) NF-90 membranes fouled without antiscalants, and (c) NF-90 membranes fouled with antiscalants.

The comparison of elemental composition of virgin and fouled NF-90 membrane specimens by EDS showed the presence of trace amount of iron in the membrane fouled without addition of antiscalants. With the addition of antiscalants, EDS analysis of the fouled membranes did not detect any inorganic precipitates on the fouled membranes. The functional groups of the virgin and fouled membranes were characterized by ATR-FTIR. All the virgin and fouled membranes exhibited similar distinct and sharp absorption peaks including the NF-90 membrane fouled without addition of antiscalants (only the spectra of NF-90 and XLE membranes are shown in Fig. 8). Infrared light likely penetrated through the thin active layer (about 250 nm) resulting in detection of the polysulfone microporous support layer. Therefore, all the polyamide membranes exhibited almost the same ATR-FTIR spectra with indicative peaks at 1650 cm−1 (amide groups), 1592 cm−1 and 1110 cm−1 (aromatic double bonded carbon), 1016 cm−1 (ester groups), 1492 cm−1 (methyl groups), and at 1151 cm−1 and 694 cm−1 (sulfone groups). The ATR-FTIR did not identify any functional groups typical of foulants on the fouled membranes. The findings imply that organic fouling may not be the major

mechanisms for the flux decline during produced water treatment. 3.3. Assessment of cleaning procedures Different cleaning procedures were assessed to recover membrane permeate flux. In order to shorten the time for experiments, the microfiltered well water without addition of antiscalant was used where the tested membranes displayed a significant flux decline. Periodic cleaning of the membranes by hydraulic and a variety of chemical cleaning strategies resulted in restoring a portion of the membrane permeability (the variation of flux over time for TMG-10 membrane is shown in Fig. 9 as an example). For all the membranes, cleaning by deionized water, hydrochloric acid, citric acid, and EDTA seemed less efficient in removing foulants than using NaOH and SDS solutions. The efficiency of the latter two agents and hydraulic cleaning was tested to clean the membranes fouled by produced water with addition of antiscalants (the reduction of flux over time for NF-90 and TMG-10 membranes with the addition of antiscalants is shown

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

73

Fig. 7. ESEM micrographs of virgin and fouled membranes with addition of antiscalants (a) XLE virgin (b) XLE fouled (c) TMG-10 virgin (d) TMG-10 fouled (e) TFC-ULP virgin, and (f) TFC-ULP fouled.

Fig. 8. ATR-FTIR spectra of virgin and fouled membranes (a) NF-90 and (b) XLE.

74

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76 Table 2 Rejection performance of the selected membranes using 0.45 ␮m microfiltered produced water at pH 6.0 Membrane type

Rejection (%) Conductivity

Iodide

TOC

UVA-254

RO

TFC-HR

91.4 ± 0.0

92.4 ± 0.0

80.4 ± 3.4

87.2 ± 0.3

ULPRO

TMG-10 TFC-ULP XLE

78.0 ± 1.5 75.4 ± 1.9 73.1 ± 1.7

87.6 ± 0.6 82.6 ± 1.5 80.1 ± 0.8

79.1 ± 4.7 84.8 ± 3.5 84.3 ± 5.6

87.2 ± 3.8 82.1 ± 0.6 80.8 ± 1.3

NF

NF-90 TFC-S ESNA

72.7 ± 5.4 62.7 ± 2.3 52.5 ± 2.6

78.3 ± 1.3 69.5 ± 1.5 55.6 ± 1.4

87.6 ± 0.6 75.2 ± 1.9 57.8 ± 1.2

63.8 ± 2.1 73.3 ± 1.9 63.7 ± 1.1

Fig. 9. Permeate flux of TMG-10 membrane at different stages of fouling, hydraulic cleaning, and chemical cleaning procedures. Feedwater is 0.45 ␮m microfiltered produced water, applied pressure 80 psi, pH 6.0, temperature 11 ± 1 ◦ C, initial specific permeate flux 0.24 L/(m2 day kPa). (a) Flux after hydraulic cleaning, (b) flux after cleaning using 0.01 M NaOH solution, pH 12.0, (c) flux after cleaning using 0.01 M HCl solution, pH 2.1, (d) flux after cleaning using 0.01 M citric acid, pH 2.6, (e) flux after cleaning using 0.01 M NaOH solution, pH 12.0, (f) flux after cleaning using 0.01 M EDTA solution, pH 4.6, and (g) flux after cleaning using 0.01 M SDS solution, pH 8.0.

tant cleanings could only restore 10% of the declined flux. The flux decline caused by the irreversible fouling on NF-90 membrane remained at 23% over the course of experiments.

in Fig. 10 as example). Hydraulic cleaning using deionized water did not restore the membrane permeability for both NF-90 and TMG-10 membranes (Fig. 10). The chemical cleaning using NaOH and SDS solutions showed a high efficiency to restore the declined flux for the TMG-10 membrane, and the foulants appeared to be easily removed as compared to without addition of antiscalants. The NF-90 membrane, however, showed more resistance to chemical cleaning. Both caustic and surfac-

The average rejection of iodide and bulk parameters by the tested membranes during operation is summarized in Table 2. Rejection was stable during membrane filtration and not affected by the degree of fouling and type of membrane cleaning over the course of the experiments. The RO membrane, TFC-HR, exhibited the highest rejection in terms of specific conductance and iodide above 91% and 92%, respectively. All tested membranes showed a low organic concentration in the permeate water samples.

3.4. Rejection performance

3.5. Membrane selection

Fig. 10. Permeate flux of: (a) TMG-10 and (b) NF-90 membranes at different stages of fouling, hydraulic cleaning (with deionized water), and chemical cleaning procedures (with 0.01 M NaOH solution, pH 12.0 and 0.01 M SDS solution, pH 8.0). Feedwater is 0.45 ␮m microfiltered produced water, applied pressure 80 psi, pH 6.0, temperature 11 ± 1 ◦ C, 3 mg/L antiscalants, initial specific permeate flux 0.24 L/(m2 day kPa) (TMG-10) and 0.38 L/(m2 day kPa) (NF-90).

The selection criteria for evaluating candidate membranes in this study include assessing operational performance as well as rejection of salts for beneficial water reuse and recovery of iodide. Operational performance was evaluated by considering both the specific flux and the flux decline measured during the filtration of produced water. The specific flux of non-fouled membranes related to operating pressure using the produced water, and was considered the starting point for comparing membrane operational performance. The membrane specific flux was measured at the beginning of the filtration tests using the produced water. The flux decline was measured during the flat sheet fouling experiments, and was used to correct the virgin membrane flux for fouling. These two terms were incorporated into a single term, called adjusted specific flux, in order to allow a comparison among the target membranes. The adjusted specific flux is defined as the difference between the specific flux and the flux decline due to fouling (Eq. (2)). Although the fouling experiments did not simulate actual filtration time or pressure conditions (75 h and 80 psi) typical for full-scale systems, the membranes were tested under the same conditions, and it is assumed that the measured flux decline is relative, allowing the use of this data for comparison of the target membranes. Adjusted specific flux = specific flux ×normalized permeate flux J/Jo

(2)

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

The adjusted specific fluxes [L/(m2 day kPa)] of the candidate membranes rank as follows: NF-90 (0.252) > ESNA (0.214) > TFC-ULP (0.191) > TFC-S (0.189) > TMG-10 (0.187) > XLE (0.168) > TFC-HR (0.090). The NF membranes NF-90 and ESNA displayed a high adjusted specific flux due to their large specific flux and low fouling potential. Although the XLE membrane showed a high initial specific flux, the adjusted flux was low as a result of membrane fouling. The RO membrane TFC-HR exhibited the lowest adjusted specific flux as the initial specific flux was the smallest. The TFC-ULP, TFC-S and TMG-10 membranes exhibited similar adjusted specific fluxes. The candidate membranes are also ranked according to the rejection of salt and recovery of iodide as: The iodide and salt rejections by the ULPRO and NF membranes rank as TFCHR > TMG-10 > TFC-ULP > XLE > NF-90 > TFC-S > ESNA. In summary, based on the membrane performance regarding adjusted specific flux and rejection of salt and iodide, the TMG-10, TFC-ULP, and NF-90 membranes were selected for laboratory-scale testing using 4040 spiral wound elements in a two-stage testing unit employing a 1:1 array.

75

however, was also determined by physico-chemical properties of the virgin membranes. Hydrophobic and rough membranes such as XLE, NF-90 and TFC-S exhibited higher flux decline than smooth and/or hydrophilic membranes. Meanwhile, chemical cleaning could not totally restore the reduced flux of the hydrophobic and rough membranes due to irreversible fouling. Based on the membrane performance with regard to adjusted specific flux and rejection of salt and iodide, the ULPRO and NF membranes having high salt rejection and permeability, provide a viable technique for produced water treatment and beneficial reuse. Acknowledgments The authors thank the U.S. Bureau of Reclamation (BOR) for its financial, technical, and administrative assistance in funding and managing the project through which this information was derived. The comments and views detailed herein may not necessarily reflect the views of the BOR, its officers, directors, affiliates or agents. The authors also thank Paul Mendell with Mendell Energy, Inc. for his financial and technical support.

4. Conclusions References This study employed bench-scale filtration tests and in situ characterization techniques to examine the viability of ULPRO and NF membranes in multi-beneficial use of produced water with focus on salt rejection, iodide recovery and operating performance. The tested produced water from sandstone aquifers was characterized as a brackish groundwater of sodium chloride type with a total dissolved solids (TDS) concentration of 5300 mg/L and a specific conductance of 9650 ␮S/cm. The average iodide concentration was 55 mg/L. The critical water constituents in the produced water were identified as small particulate matters, and other inorganic compounds that could cause fouling/scaling including iron, barium, calcium, magnesium, and silica. The produced water exhibited a very high potential to membrane fouling indicated by silt density index (SDI) measurement. Seven candidate membranes were tested including one RO membrane TFC-HR (Koch) and three ULPRO membranes XLE (Dow/Filmtec), TMG-10 (Toray) and TFC-ULP (Koch), and three NF membranes NF-90 (Dow/Filmtec), TFC-S (Koch) and ESNA (Hydranautics). The TFC-HR membrane, a traditional RO membrane widely used for wastewater treatment and desalination, was used as a benchmark in the study to evaluate the treatment efficiency of ULPRO and NF membranes. The observed flux decline and characterization by ESEM/EDS and ATR-FTIR revealed that the pretreatments including microfiltration, pH adjustment and addition of antiscalant solution was effective to alleviate membrane fouling/scaling. The cleaning using caustic or surfactant could restore the permeability of the fouled membrane specimens to different extent depending upon membrane properties. Membrane fouling was found to be dependent upon the initial permeability. The degree of flux decline, in general, increased with increasing permeability. Membrane fouling,

[1] T. Hayes, D. Arthur, Overview of emerging produced water treatment technologies, in: Proceedings of the 11th Annual International Petroleum Environmental Conference, Albuquerque, NM, October 12–15, 2004. [2] Argonne National Laboratory Report, Produced Water Treatment and Management. http://www.anl.gov/LabDB2/current/Ext/H557-text.001.html. [3] R. Funston, R. Ganesh, L.Y.C. Leong, Evaluation of technical and economic feasibility of treating oilfield produced water to create a “New” water resource, in: Proceedings of 2002 GWPC Produced Water Conference, Colorado Spring, CO, October 16–17, 2002. [4] C. Bellona, J.E. Drewes, Reuse of produced water using nanofiltration and ultra-low pressure reverse osmosis to meet future water demands, in: Proceedings of 2002 GWPC Produced Water Conference, Colorado Spring, CO, October 16–17, 2002. [5] C. Visvanathan, P. Svenstrup, P. Ariyamethee, Volume reduction of produced water generated from natural gas production process using membrane technology, Water Sci. Technol. 41 (10–11) (2000) 117– 123. [6] Roskill Report, The Economics of Iodine, eighth ed., 2002. [7] American Public Health Association (APHA), American Water Works Association (AWWA), and Water Environment Federation (WEF), Standards Methods for the Examination of Water and Wastewater, 20th ed., 1998. [8] P. Xu, J.E. Drewes, T.-U. Kim, C. Bellona, G. Amy, Effect of Membrane Fouling on Transport of Organic Contaminants in NF/RO Membrane Applications. J. Membr. Sci. (2006), in press. [9] J. Cho, J. Son, H. Choi, I.S. Kim, G. Amy, Effects of molecular weight cutoff, f/k ratio (a hydrodynamic condition), and hydrophobic interactions on natural organic matter rejection and fouling in membranes, J. Water Supply: Res. Technol.–AQUA 51.2 (2002). [10] M. Nystrom, A. Pihlajamai, R. Liikanen, M. Manttari, Influence of process conditions and membrane/particle interaction in NF of wastewaters, Desalination 156 (2003) 379–387. [11] W. Peng, I.C. Escobar, Evaluation of factors influencing membrane performance, Environ. Prog. 24 (4) (2005) 392–399. [12] S.K. Hong, M. Elimelech, Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes, J. Membr. Sci. 132 (1997) 159–181.

76

P. Xu, J.E. Drewes / Separation and Purification Technology 52 (2006) 67–76

[13] R. Gerard, H. Hachisuka, M. Hirose, New membrane developments expending the horizon for the application of reverse osmosis technology, Desalination 119 (1998) 47–55. [14] M. M¨antt¨ari, L. Puro, J. Nuortila-Jokinen, M. Nystr¨om, Fouling effect of polysaccharides and humic acid in nanofiltration, J. Membr. Sci. 165 (2000) 1–17. [15] E.M. Vrijenhoek, S.K. Hong, M. Elimelech, Influence of membrane surface properties on initial rate of colloidal fouling of reverse osmosis and nanofiltration membranes, J. Membr. Sci. 188 (2001) 115–128. [16] C. Jarusuthhirak, G. Amy, Membrane filtration of wastewater effluents for reuse: effluent organic matter rejection and fouling, Water Sci. Technol. 43 (2001) 225–232. [17] X. Zhu, M. Elimelech, Colloidal fouling of reverse osmosis membranes: measurement and fouling mechanisms, Environ. Sci. Technol. 31 (2) (1997) 3654–3662.

[18] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic matter: comparison of flux decline, NOM rejection, and foulants during filtration with three UF membranes, Desalination 127 (2000) 283– 298. [19] N. Her, G. Amy, C. Jarusutthirak, Seasonal variations of nanofiltration (NF) foulants: identification and control, Desalination 132 (2000) 143– 160. [20] A. Seidel, M. Elimelech, Coupling between chemical and physical interactions in natural organic matter (NOM) fouling of nanofiltration membranes: implications for fouling control, J. Membr. Sci. 203 (2002) 245– 255. [21] J.S. Vrouwenvelder, D. Van der Kooij, Diagnosis, prediction and prevention of biofouling of NF and RO membranes, Desalination 139 (2001) 65– 71.