Environ. Sci. Technol. 2007, 41, 4767-4773
Membrane Independent Limiting Flux for RO and NF Membranes Fouled by Humic Acid CHUYANG Y. TANG* AND JAMES O. LECKIE Environmental Engineering and Science, Department of Civil and Environmental Engineering, Stanford University, Stanford, California 94305-4020
The flux decline of reverse osmosis and nanofiltration membranes was investigated under constant pressure conditions during humic acid fouling tests. For a given membrane type under a given feedwater composition, increasing pressure resulted in increased flux reduction and foulant accumulation. A limiting flux seems to exist beyond which the membrane flux cannot be sustained. Membranes with initial fluxes greater than the limiting flux experienced severe fouling and their pseudo stable fluxes approached the limiting flux. Flux reduction was much milder when the initial flux was lower than the limiting flux. Furthermore, the limiting flux seems to be independent of membrane properties, probably due to the dominance of foulantdeposited-foulant interaction upon complete foulant coverage over membrane surfaces. On the other hand, strong dependence of the limiting flux on the feedwater composition was observed. The limiting flux was reduced at higher proton, calcium, and/or background electrolytes concentrations, likely due to reduced electrostatic repulsion under these conditions.
1. Introduction Reverse osmosis (RO) and nanofiltration (NF) membranes have been increasingly used in water treatment and wastewater reclamation in recent years (1). The effectiveness of these membrane separation processes, however, is greatly affected by foulingsthe reduction in permeate flux due to accumulation of colloidal matter, organic molecules, sparingly soluble inorganic compounds, and microorganisms on membrane surfaces or in membrane pores (1, 2). Fouling is a complex process influenced by numerous factors, including feedwater characteristics (such as types of foulants, pH, and ionic compositions), membrane properties (such as surface roughness, charge properties, and hydrophobicity), and operational conditions (such as flux, crossflow velocity, and temperature) (1, 2). The effect of permeate flux on fouling is particularly interesting. It has been well documented that high flux tends to promote fouling by proteins (3), natural organic matter (4-6), polysaccharides (7), surfactants (8, 9), microorganisms (10, 11), and inorganic colloids (12, 13). Increasing flux not only increases the effective concentrations of foulant and background electrolytes near a membrane surface (concentration polarization) (2), but also increases the rate of foulant deposition due to increased hydrodynamic drag force on * Corresponding author phone: 65-6790-6871; fax: 65-6791-0676; e-mail:
[email protected]. 10.1021/es063105w CCC: $37.00 Published on Web 05/22/2007
2007 American Chemical Society
foulants toward the membrane surface (12, 13). If the initial flux is below a critical value, the flux decline due to fouling is insignificant during constant pressure fouling tests (5, 1115). The critical flux concept suggests that below the critical flux foulant accumulation and, therefore fouling, is insignificant and that membrane use can continue without flux decline. Cohen and Probstein (12) observed a critical flux during colloidal iron oxide fouling, which they attributed to the balance of negative colloidal interaction forces (such as the electrical double layer forces) and positive hydrodynamic drag force. These authors hypothesized that foulant deposition occurs only if the hydrodynamic drag force exceeds the interaction forces. Similarly, Palecek and Zydney (3) suggested that the flux decline of microfiltration membranes during protein filtration is principally governed by both foulantdeposited-foulant interaction (electrostatic repulsion and van der Waals force) and the hydrodynamic drag force. In practice, the critical flux during a constant pressure fouling test is often operationally defined as the initial flux that corresponds to some threshold flux reduction (typically 2-10%) over a given period. Thus, depending on the threshold value chosen and the duration of the fouling test, very different critical values may be reported for a given membrane system, which makes it extremely difficult to compare values reported in the literature. While many existing bench-scale fouling tests have been performed for relatively short durations (24 h or less), it may be desirable to foul a membrane until its flux becomes pseudo stable (quasi steady), i.e., further change in the membrane flux is considerably slow. As a natural extension of the critical flux concept, it might be hypothesized that (1) there is a maximum or limiting value for pseudo stable flux, and that (2) any membrane flux exceeding this limiting flux will decline to the limiting flux. The hypotheses are consistent with experimental observations that, at high initial fluxes, membranes with different initial fluxes have the tendency to decline to an identical pseudo stable flux (4, 5, 8, 16). Assuming the drag force and interaction forces are the governing factors controlling the deposition of foulants, the flux of a membrane will continue to decrease until the hydrodynamic drag force is lower or sufficiently close to the foulant-membrane and/or foulantdeposited-foulant interaction forces (3, 12). Thus, the limiting flux is probably determined by the interaction forces experienced by the foulant, and is independent of a membrane’s initial flux. Although the idea of a limiting flux is conceptually simple and might be a potentially powerful tool for understanding fouling behavior, there is a lack of systematic experimental investigation on the existence of such limiting flux, its relationship to critical flux, and the dependence of the limiting flux on membrane properties and feedwater composition. The purpose of the current investigation was (1) to demonstrate the existence of a membrane independent limiting flux for a given feedwater composition, and (2) to study the effect of feedwater compositions (pH, ionic strength, [Ca2+], and foulant concentration) on limiting flux. Purified Aldrich humic acid (PAHA) was used as a model macromolecular foulant, and the flux behavior of both RO and NF membranes were determined by bench-scale fouling tests under constant applied pressure.
2. Materials and Methods 2.1. Chemicals and Materials. Unless otherwise specified, all reagents and chemicals were analytical grade. MilliQ water was supplied from a Millipore water system (Billerica, MA) with a resistivity of 18.2 Mohm-cm. Sodium chloride, calcium VOL. 41, NO. 13, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4767
chloride, sodium hydroxide, and hydrochloric acid were purchased from Fisher Scientific (Santa Clara, CA). Aldrich humic acid (Sigma-Aldrich, H16752, technical grade, St. Louis, MO) was used as a model foulant. It was pretreated extensively to reduce fulvic acid, heavy metals, and ash content based on a slightly modified method from the International Humic Substances Society (IHSS) method (5). Briefly, the humic acid sample was washed with an HCl solution (pH ∼1) and centrifuged. The residue was dissolved in NaOH (pH ∼13), then centrifuged, and the resulting supernatant was filtered twice through a 0.2 µm polyethersulfone filter. The filtered suspension was centrifuged at pH 1, and the residue dialyzed and freeze-dried. Aldrich humic acid (AHA) is a peat-derived humic material, and is larger than typical aquatic humic acids. Purified AHA (PAHA) has a weight-averaged molecular weight ∼4000-20 000 (17, 18) with a polydispersivity of ∼2.5 (18). Its mean particle size is ∼4 nm (19). In the presence of calcium, PAHA molecules tend to aggregate and form slightly larger particles (19). PAHA has an acidity of ∼5-6 meq/g (4, 5) and an elemental composition of 55.5% C, 38.9% O, 4.6% H, and 0.6% N (17). It has been extensively used as a model foulant by many membrane researchers (4, 20) due to its easy availability and well-characterized properties. Four RO membranes (LFC1, LFC3, BW30, and ESPA3) and seven NF membranes (NF270, HL, DK, DL, NE70, NE90, and NF90) were used in the current study. BW30, NF270, and NF90 were provided by Dow FilmTec (Minneapolis, MN), LFC1, LFC3, and ESPA3 were provided by Hydranautics (Oceanside, CA), HL, DK, and DL were obtained from GE Osmonics (Minnetonka, MN), and NE70 and NE90 were gifts from Saehan Industries Inc (South Korea). All the membranes are polyamide (PA) thin film composite (TFC) membranes. On the basis of the PA synthesis chemistry, the membranes can be classified as fully aromatic PA membranes (LFC1, LFC3, BW30, ESPA3, NE90, NF90, formed by trimesoyl chloride and m-phenylene-diamine) and semi-aromatic poly(piperazinamide) based membranes (NF270, HL, DK, DL, NE70) (21, 22). The surfaces of the semi-aromatic membranes are much smoother (root-mean-square roughness Rrms ∼ 10 nm) than those for fully aromatic ones (Rrms on the order of 100 nm) (9, 22). The fully aromatic membranes LFC1, LFC3, and BW30 are commercially coated with a neutral polyalcohol layer rich in -COH groups (21). These membranes tend to be more hydrophilic (contact angle ∼20°) and less charged (zeta potential ∼ -10 mV at pH 7) than the uncoated fully aromatic PA membranes (contact angle >40° and zeta potential -10 to -30 mV) (21, 22). All membranes, except NE70 and NE90, were supplied and stored as dry coupons in the dark. NE70 and NE90 were supplied as dry coupons, and were stored in MilliQ water bath immediately upon receiving following the manufacturer’s advice. 2.2. Membrane Fouling Tests. The setup and experimental procedure used for membrane fouling tests, described elsewhere (5, 6, 8), is briefly summarized here. Flat membrane coupons (14.6 × 9.5 cm) were extensively rinsed and soaked in MilliQ water for 24 h before being loaded into CEPA CFII crossflow testing cells (GE Osmonics, Minnetonka, MN). The coupons were precompacted with MilliQ water for 48 h under pressure, and then equilibrated with background electrolytes (NaCl and/or CaCl2) at the desired pH for 24 h before PAHA was added to make a 50 L feed solution. Constant pressure (target pressure (6.9 kPa) was applied throughout the different phases of a given test run (precompaction, equilibration with electrolytes, and fouling). Unless otherwise specified, the following conditions were applied: feed tank temperature at 25 ( 1 °C, crossflow velocity at 20 cm/s, 5.0 mg/L PAHA with a total ionic strength of 10 mM (adjusted by addition of NaCl and/or CaCl2). Both permeate and retentate were recycled back to the feed tank during the tests. 4768
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 13, 2007
Fouling tests were terminated when membrane fluxes became pseudo-stable such that the change in flux over a 24 h period was less than 0.015 m/day (0.3-1.5% of the initial flux for typical runs). The typical duration of fouling was 4-10 days. Fouled membrane samples were subsequently characterized for the amount of PAHA accumulation on membrane surfaces by extracting deposited PAHA with a 0.1 N sodium hydroxide solution overnight. Detailed description of the extraction procedures can be found in Tang et al. (6), and similar procedures have been used by Hong and Elimelech (4). The PAHA concentration in the extract was measured by a UV/vis spectrophotometer (Uvikon XL, Bio-Tek Instruments) at a wavelength of 254 nm (UV254). The calibration curves, run before each batch of samples, had R2 values better than 0.99 over a concentration range of 1-10 mg PAHA/mL.
3. Results and Discussions 3.1. Existence of Limiting Flux. Figure 1a shows the flux performance of a nanofiltration membrane NF90 over 10 day fouling tests. The applied pressure ranged from 3452069 kPa (50-300 psi), and the feedwater had 5.0 mg/L PAHA in 10 mM NaCl background electrolyte at pH 7. Clearly, the initial flux of NF90 was greater at higher applied pressure. However, membrane samples with elevated initial fluxes (>2 m/day) were significantly fouled over the 10 day period, and their fluxes declined continuously. The rate of flux decline was greater for samples with greater initial fluxes. In addition, the rate of flux decline under a given pressure was greatly reduced at longer filtration time when the flux became much lower than the corresponding initial flux. Near the end of the 10 day period, further changes in permeate fluxes for all the membrane samples became reasonably slow (rate of flux decline 2 m/day) ended up with almost identical pseudo stable fluxes. On the other hand, samples with low initial fluxes (
