Environ. Sci. Technol. 2008, 42, 619–623
NOM Removal by Adsorption and Membrane Filtration Using Heated Aluminum Oxide Particles ZHENXIAO CAI, JAESHIN KIM, AND MARK M. BENJAMIN* Department of Civil and Environmental Engineering, Box 352700, University of Washington, Seattle Washington 98195-2700
Received August 27, 2007. Revised manuscript received October 31, 2007. Accepted November 2, 2007.
Heated aluminum oxide particles (HAOPs) are a newly synthesized adsorbent with attractive properties for use in hybrid adsorption/membrane filtration systems. This study compared removal of natural organic matter (NOM) from water by adsorption onto HAOPs with that by adsorption onto powdered activated carbon (PAC) or coagulation with alum or ferric chloride (FeCl3); explored the overlap between the NOM molecules that preferentially adsorb to HAOPs and those that are removed by the more conventional approaches; and evaluatedNOMremovalandfoulinginhybridadsorbent/membrane systems. For equivalent molar doses of the trivalent metals, HAOPs remove more NOM, and NOM with higher SUVA254, than alum or FeCl3. Most of the HAOPs-nonadsorbable fraction of the NOM can be adsorbed by PAC; in fact, that fraction appears to be preferentially adsorbed compared to the average NOM in untreated water. Predeposition of the adsorbents on a microfiltration membrane improves system performance. For the water tested, at a flux of 100 L/m2-hr, predeposition of 11 mg/L PAC and 5 mg/L HAOPs (as Al3+) allowed the system to operate 5 times as long before the transmembrane pressure increased by 1 psi and to remove 10–20 times as much NOM as when no adsorbents were added.
Introduction Natural organic matter (NOM) is a major concern in water treatment, both because it serves as a precursor for the formation of disinfection byproduct (DBPs) (1, 2) and because it fouls microfiltration (MF) and ultrafiltration (UF) membranes (3–5). It has long been recognized that NOM can be partially removed from solution by coagulation with alum or ferric chloride (FeCl3) (1, 6–9), with FeCl3 usually able to remove more NOM than alum can (10). Although the removal occurs by a combination of precipitation of metal-NOM solids and sorption of NOM onto the precipitated metal hydroxides, the latter mechanism dominates under normal water treatment conditions (7), and it is common to refer to all NOM removal by coagulants as adsorption; that convention is adopted here. The nonadsorbable NOM has been attributed to smaller, more hydrophilic, and/or neutral molecules, based in part on its relatively low specific UV absorbance at 254 nm (SUVA254, defined as the absorbance divided by the dissolved organic carbon concentration) (11–13). Powdered activated carbon (PAC), on the other hand, * Corresponding author phone: 206-543-7645; fax: 206-685-9185; e-mail:
[email protected]. 10.1021/es7021285 CCC: $40.75
Published on Web 12/15/2007
2008 American Chemical Society
has been reported to selectively remove medium- and/or low-molecular weight NOM (7, 14), including molecules with low SUVA254 (15). NOM removal from solution by coagulation or adsorption, either upstream of or directly in combination with a membrane process, often reduces membrane fouling (4, 6, 16–19). Possible mechanisms include removal of the foulants before they reach the membrane surface and an increase in the permeability of the cake/gel layer that forms at the membrane surface. However, in several studies, the addition of coagulants or adsorbents has exacerbated fouling, even though some NOM was removed (6, 16, 20–22). Kim et al. (23) found that heated aluminum oxide particles (HAOPs) are unusually effective at interfering with NOM fouling of a microfiltration membrane. This paper compares NOM removal by HAOPs with that by conventional coagulants and adsorbents and also describes the effect of such removal on fouling of a different membrane from the one used by Kim et al.
Materials and Methods Reagents and Water Source. All chemicals were reagentgrade. Deionized (DI) water (Millipore Milli-Q) was employed to prepare stock solutions. PAC was purchased (Integra Chemical Company) and used as received. HAOPs were prepared by neutralizing 1.5 M Al2(SO4)3 · 18H2O with NaOH to pH 7.0 and heating the resulting suspension in a closed container at 110 °C for 24 h (23). After cooling to room temperature, the suspension was rinsed with DI water and stored for later use. Raw water was collected on several occasions in spring 2007 from the northwestern shore of Lake Washington (LW). The sample was passed through a 0.45 µm filter within 24 h and was stored at 4 °C; it was brought to room temperature (around 20 °C) immediately prior to use in the tests. The water was at pH 7.6 and contained 2.2∼2.4 mg/L DOC, and its UV absorbance at 254 nm (UV254) was 0.062∼0.067 cm-1; the corresponding SUVA254 was 2.58∼3.03 L/mg-m. Adsorption Tests. Adsorption of LW NOM was studied by mixing various doses of HAOPs, alum, ferric chloride, or PAC with 100 mL of the prefiltered lake water. After the coagulants or adsorbents were added, the pH was adjusted to 7.0 ( 0.1 with 0.1 M NaOH or HCl, and the flasks were placed on a rotary shaker operating at 100 rpm for two hours. The solids were separated from solution by centrifugation followed by filtration through a 0.45 µm syringe filter. Prior studies indicated that, for all four solids, an approximate equilibrium condition was reached within 90 min. Sequential adsorption experiments were also conducted, in which the residual NOM in the filtrate from one experiment was used as the initial solution in a subsequent adsorption experiment with the same or a different adsorbent. Membrane Filtration Tests. Membrane filtration tests used dead-end filtration through 47 mm disk membranes made of polyethersulfone (MP005, Microdyn-Nadir). The membranes had a nominal pore size of 0.05 µm and an effective surface area of 9.62 cm2. Transmembrane pressure (TMP) was measured online using a transducer (PX302–100GV, Omega Engineering, Inc.). Lake water was delivered to the membrane by a peristaltic pump at a constant flux of 100 L/m2-hr. Permeate samples were collected for analysis of DOC and UV254. Experiments were terminated when the TMP reached ∼10 psi. The progress of the experiments is reported here in terms of the specific volume filtered, Vsp, defined as the cumulative volume filtered per unit area of membrane. VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
619
FIGURE 1. NOM removal from LW water by sorption onto various solids, pH 7.0 ( 0.1. Curves are empirical fits to the data and are solely to assist in visualization. In hybrid adsorption/membrane experiments, an adsorbent layer was predeposited on the membrane by passing a small volume of a concentrated slurry of HAOPs, Al(OH)3 (prepared by neutralizing alum with NaOH to pH 7.0), or PAC through the filter. The adsorbent dose was 0.052 mmol/ cm2 as Al3+for either HAOPs or Al(OH)3, and 3.1 mg/cm2 for PAC. These applied doses correspond to 0.25 mmol/L Al3+ (6.75 mg/L) and 15 mg/L PAC when Vsp is 2000 L/m2; the nominal dose decreases as Vsp increases. In some experiments, two adsorbents were predeposited on the membrane, either sequentially or simultaneously (by mixing the adsorbents before they were deposited), and in one experiment, the adsorbents were deposited on membranes in separate cartridges through which the feed was passed in sequence. Analyses. UV254 was measured using a dual-beam Lambda¨ berlingen 18 spectrophotometer (Perkin-Elmer Gmbh., U Germany) with a 1 cm cell, and DOC was determined on filtered samples with a TOC analyzer (Shimadzu, TOC-VCSH). The HAOPs particle size distribution was determined using a Saturn DigiSizer (model 5200, Micromeritics) on a sample of the suspension that had been sonicated overnight. The BET surface area of freeze-dried HAOPs was analyzed with a Micromeritics FLOWSORB II 2300 system. The aluminum content of HAOPs was determined by dissolving freeze-dried solids in 3 M NaOH, adding concentrated nitric acid to lower the pH to
